You are on page 1of 19

Chapter 3

Radiation Integrals and Auxiliary Potential Functions

M. Bataineh
Communications Engineeering Department
Yarmouk Uinversity
3.1 Introduction
v v
The aim of radiation problems is to find electromagnetic fields ( E and H ) due to
v
certain sources ( and / or J ). This is the analysis part of the problem. If we were asked
to tailor certain sources to radiate desired radiation fields, then synthesis or design these
sources is what actually required. Instead of finding the radiation fields directly from the
integral form of Maxwells equations, an alternative easier intermediate stage, in which
nonphysical function is derived. This function is referred to as auxiliary or vector
v v
potential function. These are mathematical tools by which the evaluation of E and H
are made easier. This two-stage procedure is illustrated in Fig. 3.1.

v v v
Sources , J Radiation fields E , H
Direct Integration

Integration Differentiation

Vector Potential
Functions

v v
Fig. 3.1 An overview of evaluating E and H .

In order to appreciate the difficulty involved by following the direct integration path in
Fig. 3.1, it is enough to draw attention that one has to integrate the following de
coupled equations:
v
v v J
E o o E = j 1
2
(1)
o
and


v v v
H 2 o o H = J (2)

v r v
To relate E and H to the source, here is J . Methods of integrating such equations are
convoluted and out of the scope of our introductory treatise, the references may be
consulted to have more details.

In both paths, however, the integration is always done over the source bounds (primed
variables are used to refer to sources coordinates). Differentiation (spatial) is done using
the observation, or field or distant, point coordinates (unprimed variables are used).

v
3.2. The Vector Potential A
In this section the vector potential function required in the intermediate stage of the
alternative path is defined. From Maxwells equations there are two sources to a given
v
electromagnetic field. These are J , the electric current density (A/m2), and , the
v v
volume charge density, (C/m3). These are related to H and E by:

v v v
H = J + j o E (3)
v
and E = /o (4)

A potential function could be defined by using the fact that the divergence of the curl of
any vector is identically zero. This idea is used in the divergenceless Maxwells
equation:
v
B = 0 (5)

Which implies that no isolated magnetic charges could be found. Therefore, magnetic
fields are solenoidal and can be expressed as the curt of another vector as:

v v
B = A (6)
v
Where A is an arbitrary function yet not totally determined1.

A vector field is determined if both its divergence and its curt are specified everywhere.


Now substitute (6) in the Faraday-Lenzs equation:
v v
E = j B (7)

to get: v v
E = ( j A)
v v
or [ E + j A] = 0 (8)

If one browse any vector identities table one would notice that the curl of the gradient of
any scalar function is zero, i.e., we can write:

v v
E + j A = (9)
v
Where is an arbitrary scalar function. Remember from electrostatics that E = - hence
a minus sign in the right hand side of Eq. (9) has been set and we will refer to as the
v
scalar electric potential in contrast to a vector magnetic potential attributed to A .

From (9) we can write:


v v
E = j A (10)
v
The relation between A and is made clearer if we substitute (10) in one of Maxwells
equations containing one of the sources. Take for example:

v v
H = J + j o E (11)

From Eqs. (6) and (10):

v v v
A = o J + j o o ( j A )
v v
= o J + 2 o o A j o o (12)

The left-hand side of (12) is:


v v
A = ( ) 2 A (13)



2
2
Use this in Eq. (12) and noting that o o = = k2, k is the wave number, to
c
obtain:
v v v v
2 A + k 2 A = o J + ( A + j o o ) (14)

This complicated equation could be made easier if we choose


v
A = j o o (15)

Which is what usually referred to as Lorentz gauge. Eq. (14) now reads:
v v v
2 A + k 2 A = o J (16)

Let us rewrite this equation in the following form:


v
v v J
A + k A = 1
2 2
(17)
o
v
This equation is referred to as the VECTOR WAVE EQUATION. Upon obtaining A
from (17), i.e., solving a second order partial differential equation, and using the Lorentz
v v
gauge given in Eq. (15) we can find E and H everywhere surrounding the source via:

v v A
E = j A j (18)
o o
v
v A
H= (19)

v
Actually E could also be found from:
v
v H
E= (20)
j
v
This is so since J = 0 at points for away from the source.


v
Exercise: Substitute (10) in E = o to get the wave equation for .
Solution:

v
E = (21)
o

From Eq. (10) we can write:

v
( j A ) = (22)
o
v
j A = (23)
o

Using Lorentz condition (gauge) to write:


j ( j o o ) =
o

2 k 2 =


or 2 + k 2 = (24)
o

Note, in particular, the similarity between (17) and (24), both of them governed a
potential wave function, with a right hand side representing a reduced source
v
J
distribution either 1 or , respectively. Solving Eq. (24), the simpler, should guide
o o

to deduce the solution of Eq. (17). But before going further, attention is drawn to the
vector identity used in Eq. (13). This identity is permissible only in rectangular
coordinates, because the unit vectors could be factored out of the Laplacian operator
because they are not functions of angular coordinates ( and ).

3.3. Solution of the Wave Equation

A solution of the inhomogeneous second order wave partial differential equation given in
v v
Eq. (17) is required before being able to find the field quantities E and H , and to form
v v
E H , i.e., to find most antennas parameters studied earlier. A first order


approximation solution is easily deduced from the static case (k = 0) of Eq. (24), i.e., the
Poissons equation:

2 = o (25)

Whose solution is known and welldocumented and is given by.

1 1
= 4
v'
dv'
o r
(26)

Where r is the distance between the volume charge density , assumed located at the
origin, and an observation point located at any point in space. Each static component of
Eq. (17) has a solution identical to that given in (26) with one component of the reduced
v
J
current source 1 , replaces the reduced charge source . The most notable
o o
v
feature of these solutions is that A , as well as , varies as 1/r. This helps in developing a
solution for Eq. (17), i.e., the full wave equation.

If it were assumed that an infinitesimal current source is located at the origin of the
rectangular coordinate system and is z directed, then Eq. (17) would reduce to:

J
2 Az + k 2 Az = 1z (27)
o

At points far away from the origin, i.e., points removed from the source Jz = 0, Eq. (27)
becomes:
2 Az + k 2 Az = 0 (28)

Because of symmetry, Az will not be a function of the angular spherical coordinates


and/or , it will vary only with radial distance r. The dependence is known to be 1/r in
the static case.

Expanding the Laplacian in spherical coordinates to write Eq. (28) in form:


1 d 2 d Az
+ k Az = 0
2
2 r (29)
r dr dr

In Eq. (29) partial derivatives have been replaced by ordinary derivatives since Az is only
a function of r. Eq. (29) is a second order homogenous, linear, and with variable
coefficients differential equation. The fast variation static radial dependence is removed
by assuming a solution of the form:

(r )
Az ( r ) = (30)
r

Where (r) is an arbitrary function of r to be determined later. Differentiating twice and


substituting in (29) to obtain the following equation governing the variation of :

d2
2
+ k2 = 0 (31)
dr

Note that the result of assuming (30) as a general form of solution results in a second
order, linear, homogenous, and with constant coefficients differential equation whose
solution may be easily verified to be:

(r) = C1 e-jkr + C2 ejkr (32)

here, C1 and C2 are arbitrary constants to be determined by the conditions imposed by the
source. The general solution for Az after writing the time dependence explicity takes the
form:

1 j(t - kr) 1 j(t + kr)


Az (r,t) = C1 e + C2 e (33)
r r

These two terms forming a linearly independent set, thus gives the general solution for
(28). These two terms are constants only at spheres surrounding the source, hence they
are referred to as spherical waves. The first term in (33) represents a spherical wave
propagating away (in the radial direction) from the source, and the second term describes
an inwardly traveling wave. From the physical specifications of the problem C2 is not


part of the solution since there is nothing, as described in the problem statement, to cause
reflection of an outgoing wave. Therefore, Eq. (33) reduces to:

1 j(t - kr)
Az (r,t) = C1 e (34a)
r
r
j (t )
1 e c
= C1 (34b)
r

The spherical wave written in (34a) has a phase retardation kr or the time is delayed by
r/c, as (34b) suggests. The only undetermined constant of integration, C1, remained in
Eq. (34) is now evaluated.

3.3.1. A Particular Solution of the Wave Equation

The constant, C1, appeared in (34) is determined by the conditions imposed by the source
v
J
. Before undertaking the details of this step let us have another look at Az. The only
o1

difference between Az in static case and dynamic case is the factor e-jkr, as indicating by
Eq. (34). Therefore, Az in static case should amended by this factor to have a particular
solution for Az in time varying conditions. This is deduced by the similarities, already
noted, between the governing equations for and Az, i.e., Eqs. (24) and (27).

For static case:


1 Jz 1
Az = 4
v'
1 dv'
r
o
(35)

for a filamentary current source of length dl and cross sectional area ds, dv in Eq. (35)
could be replaced by:

dv' = ds' dl'

then:
J z d s' dl ' Idl
Az = 4
v'
1
o r
=
4 o1 r
(36)


For time varying conditions it is sufficient to multiply Eq. (36) by e-jkr and it is
understood that the time dependence is ejt. Therefore, the particular solution of Az
appears as:

Idl 1 j (t kr )
Az = e (37)
4 1
o
r

Comparing Eqs. (37) and (34) we deduce that:

Idl
C1 = (38)
4 o1 r

The same result would be obtained by taking the volume integration of Eq. (27) over a
very small spherical volume of radius ro. Evaluating this on a term-by-term basis. For
the first term we have:

lim 2 Az dv' = lim Az dv' = lim Az ro2 sin d d a r (39)


ro 0 ro 0 ro 0
v' v' s'
Divergence Theorem

Using Eq. (34) to write:

e j k r
Az = - (1 + jkr) C1 a r (40)
r2 r = ro

Substituting in Eq. (39) and taking the limit to get, for the first term:

- 4 C1 (41)

1
The second term vanishes since Az varies as and dv' = ro2 sin d d dr. Therefore,
r
if ro is chosen vanishingly small the volume integral of Az would be zero. The volume
integral of the right hand side, for an infinitesimal current source, is:

Jz Idl

v'
d s ' dl ' = 1
1
o o
(42)

Collecting the terms to get:


Idl
C1 = (43)
4 o1

as deduced earilier in Eq. (38).

The conclusion of the foregoing discussion is that for an infinitesimal current source,
corresponding to an isolated charge in electrostatics, the z component of the magnetic
vector potential, suppressing the time dependence, is given by:

( Idl )
Az = e jkr (44)
4 o1 r

Eq. (44) is written in a form to remind us with the static electric potential due to an
isolated charge Q which is known to be

(Q)
= (45)
4 o r

Note that the factor ejkr was needed in Eq. (44) to include time varying effects, as
discussed previously, however.

The extension from point sources (either charge or infinitesimal current distribution)
v
located at r ' = 0, i.e., at the origin of the coordinate system, to the case of arbitrary
distribution of charges or current over a volume is straightforward. For the case of
current distribution over an arbitrary volume v shown in Fig. 3.2, the contribution of
v
current of each dv located at r ' is evaluated. The current element and the distance in Eq.
v v v
(44) are replaced by J (r ' ) dv and r - r ' , respectively, i.e.,

v v
Idl J (r ' ) dv
v v
r r - r'

v
r
v v v
R = r - r'
dv
v
v r' y

x
Fig. 3.2 An arbitrary distribution of current

v v
Then the amount of A due to J dv is only:
v
v J dv' e jkR
dA = (46)
4 o1 R
v v v
Where R = R = r - r ' . The distance R is taken between the source point and the
observation, or field, point. Summing the potential created by each infinitesimal amount
of current (Eq. (46)) due to each infinitesimal volume element dv to have the total
magnetic vector potential:

v v
v v J (r ' ) e jkR
A (r ) = dv' (47)
v' 4 o1 R

Of course, the integral in Eq. (47) is possible because Maxwells equations are linear.
Therefore, we can simply sum the potentials contributed by each incremental amount of
current acting alone to get the total potential. For this reason the integral in Eq. (47) is
also called superposition integral.

v
After the integration in (47) is carried out we get A in terms of the unprimed variables,
v
i.e., observation point coordinate system. Remember that A is also a function of time.
Restoring the time dependence in Eq. (47) to read as:


v v
v v J (r ' ) e j (t kR )
A (r ; t ) = dv' (48)
v' 4 o
1
R

The exponential factor, in particular, in Eq. (48) could be written as

e j(t R/c) (49)

Which indicates that the vector magnetic potential at a distance R from the source at time
v
t depends on the value of J at an earlier time (t R/c). It takes time R/c for the effect of
v v
J to be felt at distance R. For this reason A is called a retarded vector potential. It is
now made clearer why C2 in Eq. (33) should be zero since that term can not be a
physically useful solution because it would lead to the impossible situation that the effect
v
of J would be felt at a distant point before it occurs at the source (non causal).

Another interesting point, which is drawn from the factor given in Eq. (49), is that it takes
time for electromagnetic waves to travel and the effect of source disturbance travel with a
finite velocity (though it is 3 108 m/s!!) to the observation point, this is in contrast with
the instant response always assumed in dealing with circuit problems.

Let us stress the point that R in Eq. (48) could not be taken outside the integral sign since
it is a function of the observation coordinate point as well as the source coordinate point.
To explore this assume that the source point is located at (x, y, z) and the field at (x, y, z)
is desired then:

v v v
R= R = r - r ' = [(x - x)2 + (y - y )2 + (z - z )2)]

R = [(x 2 + y 2 + z 2) - (x2 + y2 + z2) 2 (x x + y y + z z )] (50)

From this we note that:


v
r = r = x 2 + y 2 + z 2

Is the usual spherical radial distance from the origin to the field, or observation, point,

and


v
r = r ' = x 2 + y 2 + z 2
The same like r but to the source point. It is not so strong assumption to assume that

r << r

Replacing the observation point rectangular coordinate variables x, y, z by their spherical


equivalents, i.e.,

x = r sin cos
y = r sin sin
z = r cos

to cast Eq. (50) in the following form:

R = [r 2 2 r (x cos sin + y sin sin + z cos )]


2
= r [1 - (x cos sin + y sin sin + z cos )] (51)
r

Upon using the binomial expansion and retaining only r-1 terms, Eq. (51) could be written
as:

1
R r [1 - (x cos sin + y sin sin + z cos )]
r
= r - (x cos sin + y sin sin + z cos ) (52)

Eq. (52) shows explecitly the dependence of R on both primed (source) and unprimed
(observation) coordinate variables. It could be written in an elegant, easy to remember
form if it is noted that the parenthesized scalar term in Eq. (52) could be written as an
inner product of two vectors. The first is easily written as:

v
r ' = x a x + y a y + z a z (53)

Which is the radius vector from the origin to the source point. The second vector should
take the form:

sin cos a x + sin sin a y + cos a z (54)


and it is not going to take us long to realize that this vector is not but a r. Therefore, Eq.
(52) now takes the compact form:

v
R = r - r a r (55)
v
i.e., the difference between R and r is the projection of a r onto the radius vector r ' .
Returning now to Eq. (47) and take all the unprimed variables outside the integral sign
v
to have the following most general form for A :

v e jkr v v v

jka r r '
A = o J ( r ' ) e dv' (56)
4 r v'

Note, in particular, that R has been substituted by r in the amplitude term since this has a
negligible effect on the amplitude of each elementary contribution, especially when
r>>r, as the usual case indeed. In the phase factor we must be more accurate, however,
since the path length differences may be a sizable fraction of a wavelength which is very
much affect the phase factor.

v
The form of A as given in Eq. (56) can be thought of coming about from product of two
factors. The first is an outgoing (scalar) spherical wave factor:

e jkr
(57)
4 o 1
r

and all the directional properties are contained in the other factor:
v v v v
g ( , ) =
jka r r '
J ( r ' ) e dv' (58)
v'

v
hence g is referred to as the directional weighting function. Because the integration will
be carried out over x, y, z the result will be a function of the angular variables and .
v
Therefore, A could be written as:


v e j (t kr ) v
A (r , , , t ) = g ( , ) (59)
4 o1 r
v
Eq. (59) indicates indeed an important property of the dependence of A on coordinate
variables. The dependence on r is separated from the dependence on and . This is a
direct consequence of writing Eq. (56) as a multiplication of Eqs. (57) and (58). The
v
form of Eq. (59) will enable us to deduce useful formulae to the field quantities from A ,
as will be shown in the following section.

3.4 Radiated Fields

The form of the magnetic vector potential as given in Eq. (59) is now utilized in order to
v v
develop easier relations between E and H than that which were given in Eqs. (18) and
(19). The formulae which are about to be developed are especially useful to be applied at
the far field, or distant, or radiation, or Fraunhofer, region. At that region r >> r and
kr>>1, therefore, in the following analysis terms varying as 1/r2 or higher will be
negligible in comparison with those varying as 1/r.

Starting from Eq. (18) and perform all the needed operations on a step-by-step basis. Let
us reproduce Eq. (18) here:

v
v v A
E = j A j (60)
o o
Where
v e jkr v
A = g ( , ) (61)
4 r 1
o

As given previously in Eq. (59)

The first term in Eq. (60) will be:

o e jkr v
j g ( , ) (62)
4 r


v
The second term requires us first to find the divergence of A then the gradient of the
resulting scalar. Undertaking this step1 using spherical coordinate system and keeping
terms O (1/r), where the symbol O stands for of the order of, results in:

j k 2 e jkr v
g a r (63)
4 o r

Therefore, Eq. (60) now takes the form:

v v k2 v e jkr
E = j g ( , ) 2 ( g a r ) a r (64)
o o 4 o r
1

k2
Note that = 1 , therefore Eq. (64) could be easily written as,
2

v e jkr v
E = j g T ( , ) (65)
4 r 1
o

The subscript T appeared in Eq. (65) denotes the transverse component of the
directional weighting function. At points far away from the source the electric field
intensity is given by:

v v
E = j AT = j ( A a + A a ) (66)
v
Where Eq. (66) was written by remembering that A is a product of a spherical wave
factor and the directional weighting factor, as given previously in Eq. (59).

The magnetic field intensity could be found from Eq. (19) or from:
v
v E
H = j (67)
o
v
Substituting E from (65) to get1

See Appendix A


v 1 e jkr v
H = ( g T ( , )) (68)
4 r
v
a r E
= (69)

o
Where = is the free space wave impedance. It can be concluded from a study of
o
v v v
Eqs. (66) and (69) that the radiated E and H fields are entirely transverse, also E
v v v
differs from A T by a multiplicative constant, and H is perpendicular to E , i.e.,
v v
E and H form a Transverse ElectroMagnetic (TEM) wave in the far-field.
3.5 Closure
This chapter introduces all the equations needed to find all antennas parameter discussed
so far. The aim was to find the fields for a given distribution of actual sources. This is
v
done via an intermediate stage of magnetic vector potential A which is related to the
source and made to obey all Maxwells equations and whose variations are governed by
the wave equation. A solution of this equation is sought taking into consideration the
v
limiting case of time invariant (i.e., static) case. Finding A is so crucial to find
v v
E and H . The procedure of finding the fields is summarized below:

v
1. Find A in rectangular coordinates using Eq. (56) for a given current distribution.
v
2. Express A in spherical coordinates.
3. (a) If far-field are desired use of Eqs. (66) and (69).
(b) If the fields everywhere are required use of Eqs. (18) and (19).
v v
4. Form E H to find any antenna parameter required.

In timevarying case, it is always the current distribution, not the charge distribution, is
needed. This is so because they are related through the continuity equation:

v
J = j (70)
Therefore, solving for the fields using one of the sources is sufficient.


Appendix A : Formulae Verification

A.1 Verification of Eq. (63)


v
A jk 2 e jkr
j = g r ( , ) a r
o o 4 o r

Perform this step by step:


v 1 (r Ar )
2
1 (sin A ) 1 A
A = 2 + + (A.1)
r r r sin r sin

The second and third terms are negligible compared with the first since both of them are
O(1/r2). The first term will be:
v r e jkr v
r 2 Ar = r 2 A a r = g a r
4 o1

r e jkr
= g r ( , )
4 o1

1
(r 2 Ar ) = (r e jkr ) g r ( , )
r r 4 o1

g r ( , )
= [ 1 jkr ] e jkr
4 1
o

1 e jkr 1
(r Ar ) = j k
2
g r ( , ) + O ( 2 )
r r
2
4 o r1
r

Now take the gradient of this scalar requires:

v e jkr jk
A = j k g r ( , ) = a r g r ( , )
4 o r1
4 o1

jk (e jkr / r ) 1
= a r g r ( , ) + O( 2 )
4 o1 r r

jk e jkr 1
= a r g r ( , ) j k + O( 2 )
4 o1 r r

1
Ignoring all the terms of O ( ) results in Eq. (63):
r2


j v k j k jkr jk 2 e jkr
A = g r ( , ) e a r = g r ( , ) a r
oo 4 o r 4 o r

A.2 Verification of Eq. (69)


v
v 1 e jkr v a r E
H = ( g T ( , )) =
4 r

We could use the vector identity:


v v
( f G ) = f G + f ( G )

v e jkr v
Where f and G correspond to and g T ( , ) , respectively.
r
e jkr e jkr 1
= a r j k + O( 2 )
r r r

e jkr v e jkr v
then g T ( , ) = j k a r g T ( , )
r r
e jkr
( gv T ( , ) ) = O ( 12 )
r r
1 e jkr v e jkr v
then, g T ( , ) = j k a r g T ( , )
4 r 4 r
v jk e jkr v
or H = a r g T ( , )
o 4 o r
1

jk v
= a r AT
o
v
jk E v
= a r E
o ( j )
k v
= a r E
o
v
H a r
o o v
= a r E
o
v v
a r E a r E
= =
o / o

You might also like