You are on page 1of 15

62 IEEE Antennas and Propagation Magazine, Vol. 55, No.

2, April 2013
General Analytical Solution for the Dispersion
Equation of an Ionized Gas
in a Constant Background Magnetic Field
John W. Arthur
School of Engineering
The University of Edinburgh
Edinburgh, Scotland
E-mail: john.arthur@tiscali.co.uk
Abstract
Understanding how radio waves propagate in the Earths upper atmosphere requires a model for the conductivity of an
ionized gas. A basic linear model requires the solution of three simultaneous vector equations for
i
v ,
e
v , and
n
v ,
respectively the ion, electron, and neutral gas velocities. Of these,
i
v and
e
v alone determine the current density. When
a uniform background magnetic eld is present, some of the coefcients in the equation are cross products, rather than
just simple scalars. This article demonstrates that in contrast to the more obvious route of nding an approximate
solution or one computed via 9 9 matrices, an algebraic approach leads to a systematic analytical solution of this
reasonably complex problem. For example, the dispersion equations for plane electromagnetic waves may be found
exactly, and using the effective resistivity requires less effort than the conductivity. In addition, an interesting method of
nding
i e
v v develops out of an equation of the form
i i e e
a b a b + = + v v v v .
Keywords: Algebra; conductivity; dispersive media; electromagnetic analysis; electromagnetic engineering education;
electromagnetic propagation; electromagnetic propagation in plasma media; ionospheric electromagnetic propagation;
matrix inversion
1. Introduction
I
t has been well known for many years that the electrical
conductivity of ionized layers within the upper atmosphere
has a signifcant impact on the propagation of electromagnetic
waves, particularly in the case of radio communications. The
problem is complex. There are both positive and negative
charge carriers within an atmosphere of mainly neutral gas. The
positive and negative carriers have very different masses, so
that their reactions to time-varying electromagnetic felds are
also very different. Under the infuence of the Earths magnetic
feld, their motion may be gyratory rather than lin ear. Each
constituent gas exerts a frictional drag on the drift of the two
others. Finally, the atmosphere as a whole may well be in
motion. Understanding how such an ionized gas behaves in the
presence of an electromagnetic feld is therefore key to
understanding how radio waves are propagated or refected.
The simplest model to embody all the essential features is the
set of magnetoionic kinetic equations [1, p. 415; 2; 3, p. 14] for
three constituent gas types, namely positive ions, i, elec trons, e,
and neutral atoms, n, respectively with drift velocities
i
v

,
e
v ,
and
n
v . In SI units, these equations are
( ) ( )
i i t i e e ei i e i i i i n
N m N m v N m v + + v v v v v
( )
i i i
N e = + + E v B F ,
( ) ( )
e e t e e e ei e i e e e e n
N m N m v N m v + + v v v v v
( )
e e e
N e = + + E v B F , (1)
( ) ( )
n n t n i i i n i e e e n e n
N m N m v N m v + + = v v v v v F .

It is assumed here that the forces acting on each constituent gas
are due to the Lorentz force arising from a time-varying electric
feld, E, and a magnetic feld, B , which includes a constant
ISSN 1045-9243/2012/$26 2013 IEEE
background magnetic feld,
0
B ; external forces,
l
F ; and
collisions between the different types of particles. The effects
of collisions are introduced by way of the collision frequencies,
ei


,
e
, and
i


, which respectively refer to colli sions between
electrons and ions, electrons and neutrals, and ions and neutrals,
together with the relative drift velocities of the different
particles. These collision frequencies [1, pp. 89-91; 4, pp. 101-
104] are taken so as to provide the correct rates of momentum
transfer in the above equations, with
l
N being the number
density and
l
m being the mass of the lth type of particle. It is
assumed that:
The ionized gas is uniform and can be ionized to
any degree
There is no net charge, so that
i e
N N =
The ions are singly charged and have the same
aver age mass as neutrals, i.e.,
i n
m m =
The response to the electromagnetic feld is linear
and, in particular,
l
v B may be taken as
0 l
v B
The time derivatives may be treated as partial rather
than total
All external forces are absent or negligible.
Finally, we assume harmonic time variation with circular fre-
quency , leaving us with the three simultaneous equations
written symbolically as

i ei ei i
ei e ei e
i e e i
j
j
j



+ + + (
(
+ +
(
( + +


0
i
e
e
n
e
m
( (
( (
=
( (
( (

v E
v E
v
, (2)
where
e
e m B = . implies taking the cross product with
the corresponding vector in the column matrix to the right.
e i
m m = and ( )
i i n n e n
N m N m N N = , which ranges
from 0 (no ionization) to (complete ionization). We have
taken 1 as being effectively one, since
5
2 10

,
whereas the relative inaccuracies in the gas parameters (such as
the number density and collision frequencies) are likely to be
orders of magnitude greater. Equation (2) then represents a
straightforward linear model that would be hard to simplify in
the absence of further information, such as the likely relative
magnitudes of the number densities or collision frequencies.
However, our chief purpose is not to adapt or analyze the model,
but to take it as it stands and study the method of solu tion.
Although we have referred to the example of the Earths
atmosphere (more specifcally, the ionosphere lying between
approximately 80 km to 300 km in altitude), with the appro-
priate choice of parameters it applies in principle to any sort of
uniform ionized gas in which the external forces are purely
electromagnetic and the response to them is linear.
While the cross products with and the vector entries in
the column matrices make Equation (2) slightly nonstan dard,
the meaning is clear in an operational sense. If all the coeffcients
in our 3 3 matrix were simple scalars, we could readily invert
it to fnd the gas velocities in terms of the elec tric feld, E.
However, the operator itself may be repre sented by a 3 3
matrix (we will use bold italics to differen tiate square
matrices and rank-2 tensors from vectors, which are represented
using bold without italics) such that for any vector u ,
u u =

| |
0

0
0
z y
x
z x y
y x
z
u
u
u



( (
=
( (

( (
( (


x y z
. (3)
This causes the symbolic square and column matrices in
Equation (2) to infate to their proper nine-dimensional form,
with everything expressed purely in terms of components. Here,
we show just the frst six rows and columns as an illus tration:
AP_Mag_Apr_2013_Final.indd 62 5/18/2013 8:57:02 PM
IEEE Antennas and Propagation Magazine, Vol. 55, No. 2, April 2013 63
General Analytical Solution for the Dispersion
Equation of an Ionized Gas
in a Constant Background Magnetic Field
John W. Arthur
School of Engineering
The University of Edinburgh
Edinburgh, Scotland
E-mail: john.arthur@tiscali.co.uk
Abstract
Understanding how radio waves propagate in the Earths upper atmosphere requires a model for the conductivity of an
ionized gas. A basic linear model requires the solution of three simultaneous vector equations for
i
v ,
e
v , and
n
v ,
respectively the ion, electron, and neutral gas velocities. Of these,
i
v and
e
v alone determine the current density. When
a uniform background magnetic eld is present, some of the coefcients in the equation are cross products, rather than
just simple scalars. This article demonstrates that in contrast to the more obvious route of nding an approximate
solution or one computed via 9 9 matrices, an algebraic approach leads to a systematic analytical solution of this
reasonably complex problem. For example, the dispersion equations for plane electromagnetic waves may be found
exactly, and using the effective resistivity requires less effort than the conductivity. In addition, an interesting method of
nding
i e
v v develops out of an equation of the form
i i e e
a b a b + = + v v v v .
Keywords: Algebra; conductivity; dispersive media; electromagnetic analysis; electromagnetic engineering education;
electromagnetic propagation; electromagnetic propagation in plasma media; ionospheric electromagnetic propagation;
matrix inversion
1. Introduction
I
t has been well known for many years that the electrical
conductivity of ionized layers within the upper atmosphere
has a signifcant impact on the propagation of electromagnetic
waves, particularly in the case of radio communications. The
problem is complex. There are both positive and negative
charge carriers within an atmosphere of mainly neutral gas. The
positive and negative carriers have very different masses, so
that their reactions to time-varying electromagnetic felds are
also very different. Under the infuence of the Earths magnetic
feld, their motion may be gyratory rather than lin ear. Each
constituent gas exerts a frictional drag on the drift of the two
others. Finally, the atmosphere as a whole may well be in
motion. Understanding how such an ionized gas behaves in the
presence of an electromagnetic feld is therefore key to
understanding how radio waves are propagated or refected.
The simplest model to embody all the essential features is the
set of magnetoionic kinetic equations [1, p. 415; 2; 3, p. 14] for
three constituent gas types, namely positive ions, i, elec trons, e,
and neutral atoms, n, respectively with drift velocities
i
v

,
e
v ,
and
n
v . In SI units, these equations are
( ) ( )
i i t i e e ei i e i i i i n
N m N m v N m v + + v v v v v
( )
i i i
N e = + + E v B F ,
( ) ( )
e e t e e e ei e i e e e e n
N m N m v N m v + + v v v v v
( )
e e e
N e = + + E v B F , (1)
( ) ( )
n n t n i i i n i e e e n e n
N m N m v N m v + + = v v v v v F .

It is assumed here that the forces acting on each constituent gas
are due to the Lorentz force arising from a time-varying electric
feld, E, and a magnetic feld, B , which includes a constant
background magnetic feld,
0
B ; external forces,
l
F ; and
collisions between the different types of particles. The effects
of collisions are introduced by way of the collision frequencies,
ei


,
e
, and
i


, which respectively refer to colli sions between
electrons and ions, electrons and neutrals, and ions and neutrals,
together with the relative drift velocities of the different
particles. These collision frequencies [1, pp. 89-91; 4, pp. 101-
104] are taken so as to provide the correct rates of momentum
transfer in the above equations, with
l
N being the number
density and
l
m being the mass of the lth type of particle. It is
assumed that:
The ionized gas is uniform and can be ionized to
any degree
There is no net charge, so that
i e
N N =
The ions are singly charged and have the same
aver age mass as neutrals, i.e.,
i n
m m =
The response to the electromagnetic feld is linear
and, in particular,
l
v B may be taken as
0 l
v B
The time derivatives may be treated as partial rather
than total
All external forces are absent or negligible.
Finally, we assume harmonic time variation with circular fre-
quency , leaving us with the three simultaneous equations
written symbolically as

i ei ei i
ei e ei e
i e e i
j
j
j



+ + + (
(
+ +
(
( + +


0
i
e
e
n
e
m
( (
( (
=
( (
( (

v E
v E
v
, (2)
where
e
e m B = . implies taking the cross product with
the corresponding vector in the column matrix to the right.
e i
m m = and ( )
i i n n e n
N m N m N N = , which ranges
from 0 (no ionization) to (complete ionization). We have
taken 1 as being effectively one, since
5
2 10

,
whereas the relative inaccuracies in the gas parameters (such as
the number density and collision frequencies) are likely to be
orders of magnitude greater. Equation (2) then represents a
straightforward linear model that would be hard to simplify in
the absence of further information, such as the likely relative
magnitudes of the number densities or collision frequencies.
However, our chief purpose is not to adapt or analyze the model,
but to take it as it stands and study the method of solu tion.
Although we have referred to the example of the Earths
atmosphere (more specifcally, the ionosphere lying between
approximately 80 km to 300 km in altitude), with the appro-
priate choice of parameters it applies in principle to any sort of
uniform ionized gas in which the external forces are purely
electromagnetic and the response to them is linear.
While the cross products with and the vector entries in
the column matrices make Equation (2) slightly nonstan dard,
the meaning is clear in an operational sense. If all the coeffcients
in our 3 3 matrix were simple scalars, we could readily invert
it to fnd the gas velocities in terms of the elec tric feld, E.
However, the operator itself may be repre sented by a 3 3
matrix (we will use bold italics to differen tiate square
matrices and rank-2 tensors from vectors, which are represented
using bold without italics) such that for any vector u ,
u u =

| |
0

0
0
z y
x
z x y
y x
z
u
u
u



( (
=
( (

( (
( (


x y z
. (3)
This causes the symbolic square and column matrices in
Equation (2) to infate to their proper nine-dimensional form,
with everything expressed purely in terms of components. Here,
we show just the frst six rows and columns as an illus tration:
AP_Mag_Apr_2013_Final.indd 63 5/18/2013 8:57:02 PM
64 IEEE Antennas and Propagation Magazine, Vol. 55, No. 2, April 2013

0 0
0 0
0 0
0 0
0 0
0 0
i ei z y ei
ix
z i ei x ei iy
y x i ei ei
iz
ei e ei z y ex
ey
ei z e ei x
ei y x e ei ez
j
j
j
j
j
j







+ + (
(
(
(
+ +
(
(
(
( + +
(
(
( + +
(
(
(
+ +
(

+ +
(


x
y
z
x
e
y
z
E
E
E
e
E
m
E
E

(
(
(
(
(
=
(
(

( (
( (

( (
( (

(4)
At this point, most of us would reach for the handiest
computer application, introduce suitable expressions for the gas
parameters, and invert the complex 9 9

matrix to obtain the
gas velocities and thence the current density as functions of the
electric feld. Of course, this must be done for each fre quency of
interest. Once we have taken the step from Equa tion (2) to
Equation (4), any hope of an analytic solution is effectively lost.
However, combining an analytic solution with Maxwells
equations would provide a model for electromag netic waves in
the ionized gas at any frequency, and it would consequently be
free from any problem associated with the numerical inversion
of matrices. While this route was actually investigated some
time ago [5-9; 10, Ch. 7; 11-13], some form of restriction,
simplifying assumption, o r approximation was generally
involved. For example, Abbas assumed the neutral gas was
stationary, while Piddington assumed the conductivity tensor
was that of a fully ionized gas, and introduced colli sions only
between ions and neutrals.
Debate often emerges as a result of the differences
between solutions based on different physical assumptions and
approximations. Unless an exact solution can be found, it is
often diffcult to tell how good or bad any given approximate
solution might be. If approximations are available to help sim-
plify the problem, it is in the nature of matrix inversion that
small errors on the input side may well be magnifed on the
output side. Contrary to expectation, an exact, general analyti-
cal solution for the linear three-gas model expressed by Equa-
tion (2) can be found [14, 15]. The purpose of this paper is
not simply to demonstrate this complete solution. Rather, it is
intended as a tutorial paper, with the aim of reminding us to
be alert to novel or unusual techniques, no matter how simple,
and of the potential benefts of an analytic solution in com-
parison with the results of expedient approximation methods or
computer models.
2. Solution
While it is by no means obvious, it is possible to solve
Equations (1) by introducing variables that include the current
density, J , a plasma velocity,
p
v ,

and a total gas velocity,
t
v ,
and then manipulating the three equations until both
p
v and
t
v are eliminated [11]. This leads to the crucial result that
relates J directly to E. However, there is a subtler but more
direct route, starting from Equation (2), which requires us only
to consider it in a slightly different way. Let us look again at the
matrix . By defnition, = u u

for any vector u . We
can evaluate
2
as a matrix, and indeed it is given by

2
2 2 2
2
1 0 0
0 1 0
0 0 1
x x y x z
x y y y z
x z y z z



(
(
(
(
(
=
(
(
(
(


,
(5)
where
( )
1/ 2
2 2 2
0 x y z e
eB m = + + = is the electron
cyclo tron frequency. However, going one step further, the
curious thing is that we fnd
3 2
= . This is illustrated
by taking B along z , say, which gives

0 1 0
1 0 0
0 0 0

(
(
=
(
(

,

2 2
1 0 0
0 1 0
0 0 0

(
(
=
(
(

, (6)

3 3 2
0 1 0
1 0 0
0 0 0

(
(
= =
(
(

.
The impact of this is twofold: not only can we treat

as an
algebraic operator, any polynomial in must reduce to a
simple quadratic (the Cayley-Hamilton theorem [16, p. 299]).
We now replace with , but instead of thinking of it
primarily as a 3 3 matrix, we think of it algebraically, i.e.,
without reference to any vector basis such as | |
x y z . Not
only does any polynomial in reduce to a quadratic, it also
turns out that any non-null expression of the form
2
1 p q + +
has an inverse given by

( )
1
2 2
1 1 p q P Q

+ + = + + , (7a)
where


( )
2
2 2 2
1
p
P
p q

=
+
(7b)
and

( )
( )
2 2
2
2 2 2
1
1
p q q
Q
p q



=
+
. (7c)
Therefore, in this particular context, by treating symboli-
cally as though it were an ordinary number, there is no need to
fall into the trap of turning Equation (2) into a complex 9 9

matrix. We may then simply rewrite Equation (2) in a subtly
different form, with no cross products to distract us:

i ei ei i i
ei e ei e e
i e e i n
j
j
j



+ + + ( (
( (
+ +
( (
( ( + +

v
v
v


0
e
e
m
(
(
=
(
(

E
E . (8)
We may now do useful things, such as evaluating , the sym-
bolic determinant of the 3 3

matrix in Equation (8), and then
fnding its inverse. Proceeding by the usual rules, we obtain
( ) ( ) 1 1
e i ei i e
j ( = + + + + (

( ) ( ) ( )
2 3
1 1 1
i ei e
j + + + + + (


( ) ( ) { } ( )
2
1 1 1 1 1
i e
j
(
+ ( (



( )
2
i e
j ( + +

. (9)
Note that the scalar contribution vanishes at 0 = . This result
is exact, yet relatively simple. Since is quadratic in , from
Equation (7) it must possess a formal inverse
1
, which is a
prerequisite for fnding the inverse of our algebraic matrix in
Equation (8) according to the usual prescription

1
1
A B C EK FH CH KB BF EC
D E F GF KD AK CG CD AF
G H K DH EG BG AH AE DB

( (
( (
=
( (
( (


(10)
It is to be emphasized here that since the terms A and E both
include , the entries in our matrix are not necessarily scalar.
Nevertheless, we may go on to write

1
1
0
i
e
e
n
A B C
e
D E F
m
G H K

( ( (
( ( (
=
( ( (
( ( (

v
v E
v

( )
( )
( )
1
e
EK FH KB CH
e
GF KD CG AK
m
DH EG AH BG

+ (
(
= +
(
(
+

E . (11)
Note that E is treated here not as a 3 1 column vector, but as
a single entity, just as though it were a scalar. Since we are
mainly interested in the current density, J , we fnd
( )
e i e
N e = J v v
( ) {
2
1 e
e
N e
EK FH KB CH
m

= + (

( ) }
GF KD CG AK + (

E
(12)
( ) ( ) ( ) { }
2
1
1
e
i e i e
e
N e
j j
m

( ( = + + + + +

( )
}
2
i e
+ E
( )
( )
2
1 2 2
0
1
e
e
e
N j O
m

(
= + + +
(

E ,
where ( )( )
0
1
i e
= + + . Provided is taken from
Equation (9), this result is an exact solution, to frst order in
, for the given model equations. As no approximations are
involved in the method of solution, its accuracy essentially
depends only on our knowledge of the gas parameters. How-
ever, since
5
2 10

, there is little point in retaining in
expressions such as 1 + and 1 + , which respectively
occur in Equation (12) and in the defnition of
0
. Remarka bly,
both
ei
and the operator are completely absent from the
result, so that they are both represented in the solution only
through their role in the determinant .
Rather than having to write an equation of the form
( )
p
= + J E v B , where is the true conductivity tensor
for the ionized gas, we may now defne as being an effec tive
conductivity tensor in which the effects of the background
magnetic feld have been included [4, pp. 118-127]. This returns
us to the more familiar form = J E , in which the off-
diagonal components of correspond to Hall conductivi ties.
Therefore, to order zero in ,

AP_Mag_Apr_2013_Final.indd 64 5/18/2013 8:57:03 PM
IEEE Antennas and Propagation Magazine, Vol. 55, No. 2, April 2013 65

0 0
0 0
0 0
0 0
0 0
0 0
i ei z y ei
ix
z i ei x ei iy
y x i ei ei
iz
ei e ei z y ex
ey
ei z e ei x
ei y x e ei ez
j
j
j
j
j
j







+ + (
(
(
(
+ +
(
(
(
( + +
(
(
( + +
(
(
(
+ +
(

+ +
(


x
y
z
x
e
y
z
E
E
E
e
E
m
E
E

(
(
(
(
(
=
(
(

( (
( (

( (
( (

(4)
At this point, most of us would reach for the handiest
computer application, introduce suitable expressions for the gas
parameters, and invert the complex 9 9

matrix to obtain the
gas velocities and thence the current density as functions of the
electric feld. Of course, this must be done for each fre quency of
interest. Once we have taken the step from Equa tion (2) to
Equation (4), any hope of an analytic solution is effectively lost.
However, combining an analytic solution with Maxwells
equations would provide a model for electromag netic waves in
the ionized gas at any frequency, and it would consequently be
free from any problem associated with the numerical inversion
of matrices. While this route was actually investigated some
time ago [5-9; 10, Ch. 7; 11-13], some form of restriction,
simplifying assumption, o r approximation was generally
involved. For example, Abbas assumed the neutral gas was
stationary, while Piddington assumed the conductivity tensor
was that of a fully ionized gas, and introduced colli sions only
between ions and neutrals.
Debate often emerges as a result of the differences
between solutions based on different physical assumptions and
approximations. Unless an exact solution can be found, it is
often diffcult to tell how good or bad any given approximate
solution might be. If approximations are available to help sim-
plify the problem, it is in the nature of matrix inversion that
small errors on the input side may well be magnifed on the
output side. Contrary to expectation, an exact, general analyti-
cal solution for the linear three-gas model expressed by Equa-
tion (2) can be found [14, 15]. The purpose of this paper is
not simply to demonstrate this complete solution. Rather, it is
intended as a tutorial paper, with the aim of reminding us to
be alert to novel or unusual techniques, no matter how simple,
and of the potential benefts of an analytic solution in com-
parison with the results of expedient approximation methods or
computer models.
2. Solution
While it is by no means obvious, it is possible to solve
Equations (1) by introducing variables that include the current
density, J , a plasma velocity,
p
v ,

and a total gas velocity,
t
v ,
and then manipulating the three equations until both
p
v and
t
v are eliminated [11]. This leads to the crucial result that
relates J directly to E. However, there is a subtler but more
direct route, starting from Equation (2), which requires us only
to consider it in a slightly different way. Let us look again at the
matrix . By defnition, = u u

for any vector u . We
can evaluate
2
as a matrix, and indeed it is given by

2
2 2 2
2
1 0 0
0 1 0
0 0 1
x x y x z
x y y y z
x z y z z



(
(
(
(
(
=
(
(
(
(


,
(5)
where
( )
1/ 2
2 2 2
0 x y z e
eB m = + + = is the electron
cyclo tron frequency. However, going one step further, the
curious thing is that we fnd
3 2
= . This is illustrated
by taking B along z , say, which gives

0 1 0
1 0 0
0 0 0

(
(
=
(
(

,

2 2
1 0 0
0 1 0
0 0 0

(
(
=
(
(

, (6)

3 3 2
0 1 0
1 0 0
0 0 0

(
(
= =
(
(

.
The impact of this is twofold: not only can we treat

as an
algebraic operator, any polynomial in must reduce to a
simple quadratic (the Cayley-Hamilton theorem [16, p. 299]).
We now replace with , but instead of thinking of it
primarily as a 3 3 matrix, we think of it algebraically, i.e.,
without reference to any vector basis such as | |
x y z . Not
only does any polynomial in reduce to a quadratic, it also
turns out that any non-null expression of the form
2
1 p q + +
has an inverse given by

( )
1
2 2
1 1 p q P Q

+ + = + + , (7a)
where


( )
2
2 2 2
1
p
P
p q

=
+
(7b)
and

( )
( )
2 2
2
2 2 2
1
1
p q q
Q
p q



=
+
. (7c)
Therefore, in this particular context, by treating symboli-
cally as though it were an ordinary number, there is no need to
fall into the trap of turning Equation (2) into a complex 9 9

matrix. We may then simply rewrite Equation (2) in a subtly
different form, with no cross products to distract us:

i ei ei i i
ei e ei e e
i e e i n
j
j
j



+ + + ( (
( (
+ +
( (
( ( + +

v
v
v


0
e
e
m
(
(
=
(
(

E
E . (8)
We may now do useful things, such as evaluating , the sym-
bolic determinant of the 3 3

matrix in Equation (8), and then
fnding its inverse. Proceeding by the usual rules, we obtain
( ) ( ) 1 1
e i ei i e
j ( = + + + + (

( ) ( ) ( )
2 3
1 1 1
i ei e
j + + + + + (


( ) ( ) { } ( )
2
1 1 1 1 1
i e
j
(
+ ( (



( )
2
i e
j ( + +

. (9)
Note that the scalar contribution vanishes at 0 = . This result
is exact, yet relatively simple. Since is quadratic in , from
Equation (7) it must possess a formal inverse
1
, which is a
prerequisite for fnding the inverse of our algebraic matrix in
Equation (8) according to the usual prescription

1
1
A B C EK FH CH KB BF EC
D E F GF KD AK CG CD AF
G H K DH EG BG AH AE DB

( (
( (
=
( (
( (


(10)
It is to be emphasized here that since the terms A and E both
include , the entries in our matrix are not necessarily scalar.
Nevertheless, we may go on to write

1
1
0
i
e
e
n
A B C
e
D E F
m
G H K

( ( (
( ( (
=
( ( (
( ( (

v
v E
v

( )
( )
( )
1
e
EK FH KB CH
e
GF KD CG AK
m
DH EG AH BG

+ (
(
= +
(
(
+

E . (11)
Note that E is treated here not as a 3 1 column vector, but as
a single entity, just as though it were a scalar. Since we are
mainly interested in the current density, J , we fnd
( )
e i e
N e = J v v
( ) {
2
1 e
e
N e
EK FH KB CH
m

= + (

( ) }
GF KD CG AK + (

E
(12)
( ) ( ) ( ) { }
2
1
1
e
i e i e
e
N e
j j
m

( ( = + + + + +

( )
}
2
i e
+ E
( )
( )
2
1 2 2
0
1
e
e
e
N j O
m

(
= + + +
(

E ,
where ( )( )
0
1
i e
= + + . Provided is taken from
Equation (9), this result is an exact solution, to frst order in
, for the given model equations. As no approximations are
involved in the method of solution, its accuracy essentially
depends only on our knowledge of the gas parameters. How-
ever, since
5
2 10

, there is little point in retaining in
expressions such as 1 + and 1 + , which respectively
occur in Equation (12) and in the defnition of
0
. Remarka bly,
both
ei
and the operator are completely absent from the
result, so that they are both represented in the solution only
through their role in the determinant .
Rather than having to write an equation of the form
( )
p
= + J E v B , where is the true conductivity tensor
for the ionized gas, we may now defne as being an effec tive
conductivity tensor in which the effects of the background
magnetic feld have been included [4, pp. 118-127]. This returns
us to the more familiar form = J E , in which the off-
diagonal components of correspond to Hall conductivi ties.
Therefore, to order zero in ,

AP_Mag_Apr_2013_Final.indd 65 5/18/2013 8:57:03 PM
66 IEEE Antennas and Propagation Magazine, Vol. 55, No. 2, April 2013
( )
2
1
0
e
e
N e
j j v
m

= + . (13)
Here,
1
is to be evaluated by applying Equation (7) to Equa-
tion (9), which is likely to be more tedious than diffcult. This
result provides a generalization of the simple dc form as given,
for example, by Jackson [17, 10.10, p. 345]. However, a
moments thought prompts us that the resistivity tensor R will
be more easily obtained because there is then no need to invert
:


1
= R
(14)

( )
2
0
e
i
m
j j v
N e

=
+

.
By this means, inversion of is now only a symbolic manipu-
lation for the purposes of reaching Equation (12). In order to
apply R to J , we simply replace the symbolic form
throughout by using either its 3 3 matrix representa tion or
. However, it remains to be seen whether it will prove as
easy to use R to fnd the dispersion equation for the propagation
of plane waves.
3. The Dispersion Equation
While Equation (9) for is exact within the scope of the
model, a simpler approximation may be used when the ratio
i e
h = turns out to be a small number. This is the case in the
Earths ionosphere (see Table 1), where it is of the order of
4 2 0.03 [1, p. 335]. In addition, since
4
10

< , we may
take any expression of the form 1 b + to order zero in
unless 1 b . That is to say, 1 1 b + except in the case of
1 + when 1 , i.e., the gas is close to complete ioniza-
tion. On this basis, we fnd
( ) ( ) ( )
2 3
1 1
i e ei ei e
j h j = + + + + (


( ) ( ) ( )
2 2
1
i i
j j O h
(
+ + +


(15)
Equation (15) retains the essential features of Equation (9),
the only difference being the accuracy with which the terms
involving the collision frequencies are specifed. However,
this is hardly signifcant given that these are unlikely to be
determined to an accuracy that is even comparable to h.
On combining Equations (14) and (15), we fnd that
= J E R

takes the form

( )
( ) ( )
{
2
0
1
1
e
i e ei
e
m
j
j j
N e


+ +
+

( ) ( )
2 3 2
1 1
ei e i
h j j
(
+ + + (

( )
}
2
i
j + = J E , (16)
where, to a consistent degree of accuracy, we may now write
( )
0
1
i
+ .
Only now do we introduce an axis system. If we consider
plane waves propagating along an arbitrary direction z , it will
be useful to take
0
B to be in the xz plane, and so we represent
it as ( )
0
B + x z , where sin = and cos = , with
being the angle between the background magnetic feld and the
propagation direction. At this point, we rule out any high-power
electromagnetic wave with a magnetic feld so strong that it
might infuence the motion of the charges. That is to say, we
may continue to consider as being entirely due to static feld
0
B , giving
Table 1. Typical ionospheric data at times of moderate
ionization. The altitudes of 80 km, 125 km, and 300 km are
respectively representative of the D, E, and F regions, and
0
B was taken as 45 T (0.45 G) throughout. The collision
frequencies were calculated using the formulae quoted in
[10, p. 90].
Region D E F Units
Altitude 80 125 300 km
T 200 600 1000 K
n
N
20
2.3 10
17
2.3 10
15
3.7 10
3
m

e
N
9
1. 10
11
1. 10
12
1. 10
3
m

5
2.1 10

5
2.1 10

5
3.4 10

6
2. 10
3
3. 10
1
6. 10
1
s

ei

1
2. 10
2
3. 10
3
1. 10
1
s

5
1. 10
2
1. 10
0
2. 10
1
s

6
1.8 10
7
1.8 10
7
5.6 10
1
s

6
7.9 10
6
7.9 10
6
7.9 10
1
s


0
e
eB
m
= ,

0 0
0
0 0

(
(
=
(
(

, (17)

2
2 2
2 0
2
2
0
0 1 0
0
e
e B
m


(

(
=
(
(


.
We may therefore render Equation (16) in conventional matrix
form as

2
0 2 1 2
1 0 2 1
2
2
2 1 0 2
e
e
r r r r
m
r r r r
N e
r r r r



(
+
(
+ =
(
(
+

J E, (18)
where the components
0
r ,
l
r ,

and
2
r are simple scalars with
dimensions of
1
s

, and are given by



( ) ( ) ( )
( )
2
0
0
1
i e ei ei e i
j
r
j


+ + + + +
=
+
,

1
r = , (19)

( )
( )
2
2
0
i
j
r
j j


+
=
+
.
For a weakly ionized gas, we may take 0 = , whereas for a
highly ionized gas, we take the limit . In the latter case,
e
and
i
vanish, so that
0 ei
r j + and
2
2
( ) r j
. However, the singularity in r
2
at 0 = is always present, and
needs some explanation. Since
2
r const j for
i
,
the normal dc resistivity can become very large in directions
perpendicular to
0
B . This corresponds to the fact that under the
infuence of the mag netic feld, the current fow ends up being
perpendicular to E, rather than along it. In a fnite volume of
gas (or any other conductor or semiconductor), this gives rise to
the build-up charge that is responsible for the Hall effect. The
presence of the Hall feld stops the current fowing perpendicular
to the applied feld, and thereby restores normal fow parallel to
the applied feld, so that the dc resistivity reassumes its basic
form, as though no background magnetic feld were present.
Since we are dealing with all the electromagnetic sources
explicitly through J , the wave equation for E comes directly
from Maxwells equations in free space:

( )
2
2
0
1
t t
c
= + E J E . (20)
We make no assumption as in some of the works cited about
either the displacement current or J being negligi ble.
Taking plane electromagnetic waves varying as
( )
e
j t k r

where k = k z , we employ
t
j and j k in the
usual way. We then proceed to the eigen value equation by
multiplying through by R so as to be able to substitute E for
J R , as follows:
( )
2
2 0
j
c

+ = k k E E J

( )
2 2 2
0
c j = E E R K (21)

( )
2 2 2
0
0 c j
(
+ =
(

E R K ,

where

0 0
0 0
0 0 0
jk
jk
(
(
=
(
(

K

and

2
2 2
0 0
0 0
0 0 0
k
k
(
(
( =
(
(

K .
Note that because K is of the same general form as , it may
be treated in the same symbolic fashion, with the result that
j k E may be replaced with

E K . However, a key point is
that K and will not commute unless
0
B is paral lel to k .
Since R is a quadratic in , we cannot therefore expect to be
able to commute
2
K with R . Bearing this in mind,
Equations (18) and (21) may still be manipulated in matrix
form to yield the dispersion equation that relates k to :
AP_Mag_Apr_2013_Final.indd 66 5/18/2013 8:57:05 PM
IEEE Antennas and Propagation Magazine, Vol. 55, No. 2, April 2013 67
( )
2
1
0
e
e
N e
j j v
m

= + . (13)
Here,
1
is to be evaluated by applying Equation (7) to Equa-
tion (9), which is likely to be more tedious than diffcult. This
result provides a generalization of the simple dc form as given,
for example, by Jackson [17, 10.10, p. 345]. However, a
moments thought prompts us that the resistivity tensor R will
be more easily obtained because there is then no need to invert
:


1
= R
(14)

( )
2
0
e
i
m
j j v
N e

=
+

.
By this means, inversion of is now only a symbolic manipu-
lation for the purposes of reaching Equation (12). In order to
apply R to J , we simply replace the symbolic form
throughout by using either its 3 3 matrix representa tion or
. However, it remains to be seen whether it will prove as
easy to use R to fnd the dispersion equation for the propagation
of plane waves.
3. The Dispersion Equation
While Equation (9) for is exact within the scope of the
model, a simpler approximation may be used when the ratio
i e
h = turns out to be a small number. This is the case in the
Earths ionosphere (see Table 1), where it is of the order of
4 2 0.03 [1, p. 335]. In addition, since
4
10

< , we may
take any expression of the form 1 b + to order zero in
unless 1 b . That is to say, 1 1 b + except in the case of
1 + when 1 , i.e., the gas is close to complete ioniza-
tion. On this basis, we fnd
( ) ( ) ( )
2 3
1 1
i e ei ei e
j h j = + + + + (


( ) ( ) ( )
2 2
1
i i
j j O h
(
+ + +


(15)
Equation (15) retains the essential features of Equation (9),
the only difference being the accuracy with which the terms
involving the collision frequencies are specifed. However,
this is hardly signifcant given that these are unlikely to be
determined to an accuracy that is even comparable to h.
On combining Equations (14) and (15), we fnd that
= J E R

takes the form

( )
( ) ( )
{
2
0
1
1
e
i e ei
e
m
j
j j
N e


+ +
+

( ) ( )
2 3 2
1 1
ei e i
h j j
(
+ + + (

( )
}
2
i
j + = J E , (16)
where, to a consistent degree of accuracy, we may now write
( )
0
1
i
+ .
Only now do we introduce an axis system. If we consider
plane waves propagating along an arbitrary direction z , it will
be useful to take
0
B to be in the xz plane, and so we represent
it as ( )
0
B + x z , where sin = and cos = , with
being the angle between the background magnetic feld and the
propagation direction. At this point, we rule out any high-power
electromagnetic wave with a magnetic feld so strong that it
might infuence the motion of the charges. That is to say, we
may continue to consider as being entirely due to static feld
0
B , giving
Table 1. Typical ionospheric data at times of moderate
ionization. The altitudes of 80 km, 125 km, and 300 km are
respectively representative of the D, E, and F regions, and
0
B was taken as 45 T (0.45 G) throughout. The collision
frequencies were calculated using the formulae quoted in
[10, p. 90].
Region D E F Units
Altitude 80 125 300 km
T 200 600 1000 K
n
N
20
2.3 10
17
2.3 10
15
3.7 10
3
m

e
N
9
1. 10
11
1. 10
12
1. 10
3
m

5
2.1 10

5
2.1 10

5
3.4 10

6
2. 10
3
3. 10
1
6. 10
1
s

ei

1
2. 10
2
3. 10
3
1. 10
1
s

5
1. 10
2
1. 10
0
2. 10
1
s

6
1.8 10
7
1.8 10
7
5.6 10
1
s

6
7.9 10
6
7.9 10
6
7.9 10
1
s


0
e
eB
m
= ,

0 0
0
0 0

(
(
=
(
(

, (17)

2
2 2
2 0
2
2
0
0 1 0
0
e
e B
m


(

(
=
(
(


.
We may therefore render Equation (16) in conventional matrix
form as

2
0 2 1 2
1 0 2 1
2
2
2 1 0 2
e
e
r r r r
m
r r r r
N e
r r r r



(
+
(
+ =
(
(
+

J E, (18)
where the components
0
r ,
l
r ,

and
2
r are simple scalars with
dimensions of
1
s

, and are given by



( ) ( ) ( )
( )
2
0
0
1
i e ei ei e i
j
r
j


+ + + + +
=
+
,

1
r = , (19)

( )
( )
2
2
0
i
j
r
j j


+
=
+
.
For a weakly ionized gas, we may take 0 = , whereas for a
highly ionized gas, we take the limit . In the latter case,
e
and
i
vanish, so that
0 ei
r j + and
2
2
( ) r j
. However, the singularity in r
2
at 0 = is always present, and
needs some explanation. Since
2
r const j for
i
,
the normal dc resistivity can become very large in directions
perpendicular to
0
B . This corresponds to the fact that under the
infuence of the mag netic feld, the current fow ends up being
perpendicular to E, rather than along it. In a fnite volume of
gas (or any other conductor or semiconductor), this gives rise to
the build-up charge that is responsible for the Hall effect. The
presence of the Hall feld stops the current fowing perpendicular
to the applied feld, and thereby restores normal fow parallel to
the applied feld, so that the dc resistivity reassumes its basic
form, as though no background magnetic feld were present.
Since we are dealing with all the electromagnetic sources
explicitly through J , the wave equation for E comes directly
from Maxwells equations in free space:

( )
2
2
0
1
t t
c
= + E J E . (20)
We make no assumption as in some of the works cited about
either the displacement current or J being negligi ble.
Taking plane electromagnetic waves varying as
( )
e
j t k r

where k = k z , we employ
t
j and j k in the
usual way. We then proceed to the eigen value equation by
multiplying through by R so as to be able to substitute E for
J R , as follows:
( )
2
2 0
j
c

+ = k k E E J

( )
2 2 2
0
c j = E E R K (21)

( )
2 2 2
0
0 c j
(
+ =
(

E R K ,

where

0 0
0 0
0 0 0
jk
jk
(
(
=
(
(

K

and

2
2 2
0 0
0 0
0 0 0
k
k
(
(
( =
(
(

K .
Note that because K is of the same general form as , it may
be treated in the same symbolic fashion, with the result that
j k E may be replaced with

E K . However, a key point is
that K and will not commute unless
0
B is paral lel to k .
Since R is a quadratic in , we cannot therefore expect to be
able to commute
2
K with R . Bearing this in mind,
Equations (18) and (21) may still be manipulated in matrix
form to yield the dispersion equation that relates k to :
AP_Mag_Apr_2013_Final.indd 67 5/18/2013 8:57:05 PM
68 IEEE Antennas and Propagation Magazine, Vol. 55, No. 2, April 2013
2 2 2 2
0 2 1 2
2 2 2 2
1 0 2 1
2 2
2 1 0 2
0 0
0 0 0
0 0
1
p
r r r r c k
r r r r c k j
r r r r



( (
+
( (

( ( + + =
`
( (

+ ( (

)
E
(22)

( )( ) ( )
( )
( )
( )
( ) ( ) ( )
2 2 2 2 2 2 2 2 2
0 2 1 2
2 2 2 2 2 2 2 2
1 0 2 1
2 2 2 2 2 2 2 2 2
2 1 0 2
0
p
p
p
r r c k j r c k r
r c k r r c k j r
r c k r c k r r j



+ +
+ + =
+ +
.
The dispersion equation is therefore found from 0 = , where
is now the determinant shown in Equation (22) involving:
the wave frequency, ; the electron plasma frequency,
p
,
given by ( )
2 2
0 p e e
N e m = ; the angle, , between the propa-
gation direction and the background magnetic feld; and, fnally,
the three resistivity functions,
l
r , given in Equa tion (19). These,
in turn, depend on: the lth power of the elec tron cyclotron
frequency, ; the excitation frequency, ; the three collision
frequencies,
e
,
i
, and
ei


; and the ioniza tion parameter,
. The overall result is a quadratic in
2
k that

may be expressed
as
4 2
2 0 0
0 k k = + + = , or, more conveniently, in terms
of
2 2 2
1 k c , as

( ) ( )
2
2 2 2 2 2 2
2 1 0
1 1 0 k c k c + + = ,
where
2
=
( )
( )
( )
( )
( )
2
2 2 2 2 2 2
0 2 0 2 1 0 0 2 1 p
r r r r r j r r r r
(
+ + + + + +
(

1
=
( )
( )
( ) { }
2 2 2 2 2
0 2 0 2 0 2 1
2 1 2
p
j r r j r r r r r
( (
+ + + + + +
( (


( )
2 2 2
0 0 2 p p
j r r
(
= + +
(

. (23)
The roots of Equation (23) provide a straightforward, exact
formulation the generalized Appleton-Hartree equation [18, pp.
183-193]. However, while we could in principle use the exact
forms of
0
r ,
1
r , and
2
r from Equation (19), it would generally
be more appropriate to use them to generate simpler approximate
forms that can be used with some confdence in certain
scenarios, e.g., for the cases of either low or high degrees of
ionization, or to cover particular frequency ranges of interest. In
particular, for the case of the low degrees of ionization that are
typical throughout most of the ionosphere, the resistivity
components given in Equations (19) reduce to

2
0 e ei
i
r
j



+
+
,

1
r = , (24)

2
2
i
r
j

+
.
4. Typical Results
The main characteristics of the ionosphere are to be found
throughout references [2, 4, 10]. The ionospheres onset is
approximately 80 km in altitude, where
e
N is relatively low,
26
n p
m m , and
5
2 10

. Near the ionospheres upper
limit, say around 300 km,
e
N is near its maximum, and
16
i p
m m because nearly all the ions are O
+
. Conse quently,
increases slightly to
5
3 10

, which is still very small, but


the degree of ionization itself remains quite low, with
3
10

<

. The atmospheric density,


l l
l
m N

, and to an even greater


extent,
e
N , vary according to such factors as the time of year,
time of day, solar activity, as well as latitude, longitude, and
altitude [19, 20]. The collision frequencies as derived from
these parameters [21; 10, p. 90] also depend on the atmospheric
composition and temperature, both of which also vary with
altitude and latitude. Finally, the Earths mag netic feld depends
on latitude, longitude, calendar year, and altitude. However, the
numbers given in Table 1 could be con sidered to be a typical
data set at times of moderate ionization.
For propagation either parallel to the magnetic feld
( ) 0 = or perpendicular to it ( ) 90 = , Equation (22) takes
the more tractable forms

( )
( ) ( )
( )
( )
( )
( )
( )
( )
( )
( )
2 2 2 2 2 2 2
0 2 1
2 2 2 2 2 2 2
1 0 2
2 2
0
2 2 2 2
0
2 2 2 2 2
0 2 1
2 2 2 2 2
1 0 2
0
0 0 0
0 0
0 0
90 0 0
0
p
p
p
p
p
p
r r c k j r c k
r c k r r c k j
r j
r c k j
r r c k j r
r c k r r j






(
+ +
(
(
= + + =
(
(
+ (

(
+
(
(
= + + =
(
(
( + +

E
E
, (25)
which go along with the dispersion relations shown in Table 2,
taken from the zero-valued determinants, together with the
eigenvectors found from applying these dispersion relations to
the specifc 1 1 or 2 2 submatrix from that they derive. The
parameter ( )
1
Re c c ck

= gives the relative phase velocity
of electromagnetic waves in the gas (the reciprocal of its
refractive index), whereas ( ) ( ) Re Im ck ck gives the 1 e
attenuation distance relative to the actual wavelength, .
Cases (i) and (ii) in Table 2 are the alternative roots offered by
the manifest factorization of the secular equation. Case (i)
corresponds to the simple root, which does not depend on the
background magnetic feld because E, and hence the motion of
the charges, is along
0
B . When 0 = , both E and B

lie
along k, implying that this has to be a longitudinal, and there-
fore non-propagating, mode. This is confrmed by the fact that
the dispersion relation, into which we have introduced the
approximation for
0
r given in Equation (24), is independent of
k. The physical interpretation is plasma oscillation, which
becomes clearer if we ignore
i
, for then we readily obtain
( )
1/ 2
2 2
p e e
j = + . Here, the term ( )
1
2
e e ei
= +
depresses the resonant frequency below the bare plasma fre-
quency
p
, and introduces damping through the imaginary
term. With 90 = , both E and B are orthogonal to k, giv ing
a linearly polarized propagating wave with a relatively simple
dispersion equation, referred to as the ordinary wave, or
O-wave.
In contrast to case (i), case (ii) in Table 2 does depend on
the background magnetic feld, as demonstrated by the pres ence
of
1
r and
2
r in the equations. Since E is now perpendicu lar to
0
B , the motion of the charges is governed by the full Lorentz
force, rather than by the electric feld alone. For propagation
perpendicular to the feld, 90 = , there is only one root. With
the driving electric feld along y , a Hall effect produces a
component along z that causes the resultant electric feld to
deviate from the familiar condition of being perpendicular to
the propagation direction. This is described as being the
extraordinary wave, or X-wave. However, for propagation
Table 2. Equation (26): The dispersion relations that go along with Equation (25).
( )
( )
( )
( )
( ) ( )
( ) ( )
( )
2 2
3 2
0 2 1 2 2
0
0
2 2
0
Case (i) Case (ii)
1
0
0 exp
exp
exp

1
90 exp
p
e ei
i e ei p p i
p
x
x
ck
j
j r r jr
j
j j t k
E j t k k
E j j t k B
B
ck
j r
E j t k
k
E

| |
+
= +
|
+
\ .
(
+ + + + =

= =
=
= +
=
| |
= +
|
\ .
= =
=
E x y z
E z z
B y x z z
B z
E x z
B

( )
( )
( ) ( )
( ) ( )
( )
( ) ( )
2 2
0 2
2
2
2
0 2 1 0 2
2
0
0
2
1
2
2 2
0 2 1 0 2
1
exp
exp
exp
p
p
y z
y
p
z y
p
j r r
ck
j r r r r r
E E j t k
k
E j t k B
j t k B
r
E E
r r j r r r


+ +
| |
= +
|
\ .
(
+ + +
(

= +
= +
+

=
(
+ + + +
(

E y z z
B z x x
y z x
AP_Mag_Apr_2013_Final.indd 68 5/18/2013 8:57:05 PM
IEEE Antennas and Propagation Magazine, Vol. 55, No. 2, April 2013 69
2 2 2 2
0 2 1 2
2 2 2 2
1 0 2 1
2 2
2 1 0 2
0 0
0 0 0
0 0
1
p
r r r r c k
r r r r c k j
r r r r



( (
+
( (

( ( + + =
`
( (

+ ( (

)
E
(22)

( )( ) ( )
( )
( )
( )
( ) ( ) ( )
2 2 2 2 2 2 2 2 2
0 2 1 2
2 2 2 2 2 2 2 2
1 0 2 1
2 2 2 2 2 2 2 2 2
2 1 0 2
0
p
p
p
r r c k j r c k r
r c k r r c k j r
r c k r c k r r j



+ +
+ + =
+ +
.
The dispersion equation is therefore found from 0 = , where
is now the determinant shown in Equation (22) involving:
the wave frequency, ; the electron plasma frequency,
p
,
given by ( )
2 2
0 p e e
N e m = ; the angle, , between the propa-
gation direction and the background magnetic feld; and, fnally,
the three resistivity functions,
l
r , given in Equa tion (19). These,
in turn, depend on: the lth power of the elec tron cyclotron
frequency, ; the excitation frequency, ; the three collision
frequencies,
e
,
i
, and
ei


; and the ioniza tion parameter,
. The overall result is a quadratic in
2
k that

may be expressed
as
4 2
2 0 0
0 k k = + + = , or, more conveniently, in terms
of
2 2 2
1 k c , as

( ) ( )
2
2 2 2 2 2 2
2 1 0
1 1 0 k c k c + + = ,
where
2
=
( )
( )
( )
( )
( )
2
2 2 2 2 2 2
0 2 0 2 1 0 0 2 1 p
r r r r r j r r r r
(
+ + + + + +
(

1
=
( )
( )
( ) { }
2 2 2 2 2
0 2 0 2 0 2 1
2 1 2
p
j r r j r r r r r
( (
+ + + + + +
( (


( )
2 2 2
0 0 2 p p
j r r
(
= + +
(

. (23)
The roots of Equation (23) provide a straightforward, exact
formulation the generalized Appleton-Hartree equation [18, pp.
183-193]. However, while we could in principle use the exact
forms of
0
r ,
1
r , and
2
r from Equation (19), it would generally
be more appropriate to use them to generate simpler approximate
forms that can be used with some confdence in certain
scenarios, e.g., for the cases of either low or high degrees of
ionization, or to cover particular frequency ranges of interest. In
particular, for the case of the low degrees of ionization that are
typical throughout most of the ionosphere, the resistivity
components given in Equations (19) reduce to

2
0 e ei
i
r
j



+
+
,

1
r = , (24)

2
2
i
r
j

+
.
4. Typical Results
The main characteristics of the ionosphere are to be found
throughout references [2, 4, 10]. The ionospheres onset is
approximately 80 km in altitude, where
e
N is relatively low,
26
n p
m m , and
5
2 10

. Near the ionospheres upper
limit, say around 300 km,
e
N is near its maximum, and
16
i p
m m because nearly all the ions are O
+
. Conse quently,
increases slightly to
5
3 10

, which is still very small, but


the degree of ionization itself remains quite low, with
3
10

<

. The atmospheric density,


l l
l
m N

, and to an even greater


extent,
e
N , vary according to such factors as the time of year,
time of day, solar activity, as well as latitude, longitude, and
altitude [19, 20]. The collision frequencies as derived from
these parameters [21; 10, p. 90] also depend on the atmospheric
composition and temperature, both of which also vary with
altitude and latitude. Finally, the Earths mag netic feld depends
on latitude, longitude, calendar year, and altitude. However, the
numbers given in Table 1 could be con sidered to be a typical
data set at times of moderate ionization.
For propagation either parallel to the magnetic feld
( ) 0 = or perpendicular to it ( ) 90 = , Equation (22) takes
the more tractable forms

( )
( ) ( )
( )
( )
( )
( )
( )
( )
( )
( )
2 2 2 2 2 2 2
0 2 1
2 2 2 2 2 2 2
1 0 2
2 2
0
2 2 2 2
0
2 2 2 2 2
0 2 1
2 2 2 2 2
1 0 2
0
0 0 0
0 0
0 0
90 0 0
0
p
p
p
p
p
p
r r c k j r c k
r c k r r c k j
r j
r c k j
r r c k j r
r c k r r j






(
+ +
(
(
= + + =
(
(
+ (

(
+
(
(
= + + =
(
(
( + +

E
E
, (25)
which go along with the dispersion relations shown in Table 2,
taken from the zero-valued determinants, together with the
eigenvectors found from applying these dispersion relations to
the specifc 1 1 or 2 2 submatrix from that they derive. The
parameter ( )
1
Re c c ck

= gives the relative phase velocity
of electromagnetic waves in the gas (the reciprocal of its
refractive index), whereas ( ) ( ) Re Im ck ck gives the 1 e
attenuation distance relative to the actual wavelength, .
Cases (i) and (ii) in Table 2 are the alternative roots offered by
the manifest factorization of the secular equation. Case (i)
corresponds to the simple root, which does not depend on the
background magnetic feld because E, and hence the motion of
the charges, is along
0
B . When 0 = , both E and B

lie
along k, implying that this has to be a longitudinal, and there-
fore non-propagating, mode. This is confrmed by the fact that
the dispersion relation, into which we have introduced the
approximation for
0
r given in Equation (24), is independent of
k. The physical interpretation is plasma oscillation, which
becomes clearer if we ignore
i
, for then we readily obtain
( )
1/ 2
2 2
p e e
j = + . Here, the term ( )
1
2
e e ei
= +
depresses the resonant frequency below the bare plasma fre-
quency
p
, and introduces damping through the imaginary
term. With 90 = , both E and B are orthogonal to k, giv ing
a linearly polarized propagating wave with a relatively simple
dispersion equation, referred to as the ordinary wave, or
O-wave.
In contrast to case (i), case (ii) in Table 2 does depend on
the background magnetic feld, as demonstrated by the pres ence
of
1
r and
2
r in the equations. Since E is now perpendicu lar to
0
B , the motion of the charges is governed by the full Lorentz
force, rather than by the electric feld alone. For propagation
perpendicular to the feld, 90 = , there is only one root. With
the driving electric feld along y , a Hall effect produces a
component along z that causes the resultant electric feld to
deviate from the familiar condition of being perpendicular to
the propagation direction. This is described as being the
extraordinary wave, or X-wave. However, for propagation
Table 2. Equation (26): The dispersion relations that go along with Equation (25).
( )
( )
( )
( )
( ) ( )
( ) ( )
( )
2 2
3 2
0 2 1 2 2
0
0
2 2
0
Case (i) Case (ii)
1
0
0 exp
exp
exp

1
90 exp
p
e ei
i e ei p p i
p
x
x
ck
j
j r r jr
j
j j t k
E j t k k
E j j t k B
B
ck
j r
E j t k
k
E

| |
+
= +
|
+
\ .
(
+ + + + =

= =
=
= +
=
| |
= +
|
\ .
= =
=
E x y z
E z z
B y x z z
B z
E x z
B

( )
( )
( ) ( )
( ) ( )
( )
( ) ( )
2 2
0 2
2
2
2
0 2 1 0 2
2
0
0
2
1
2
2 2
0 2 1 0 2
1
exp
exp
exp
p
p
y z
y
p
z y
p
j r r
ck
j r r r r r
E E j t k
k
E j t k B
j t k B
r
E E
r r j r r r


+ +
| |
= +
|
\ .
(
+ + +
(

= +
= +
+

=
(
+ + + +
(

E y z z
B z x x
y z x
AP_Mag_Apr_2013_Final.indd 69 5/18/2013 8:57:06 PM
70 IEEE Antennas and Propagation Magazine, Vol. 55, No. 2, April 2013
Figure 1. The relative speed of light (phase velocity) [] and relative 1 e distance [ ] plotted as a function of fre quency for
the representative data of Table 1 for the D region of the ionosphere.
Figure 2. The relative speed of light (phase velocity) [] and relative 1 e distance [ ] plotted as a function of fre quency for
the representative data of Table 1 for the E region of the ionosphere.
AP_Mag_Apr_2013_Final.indd 70 5/18/2013 8:57:06 PM
IEEE Antennas and Propagation Magazine, Vol. 55, No. 2, April 2013 71
Figure 1. The relative speed of light (phase velocity) [] and relative 1 e distance [ ] plotted as a function of fre quency for
the representative data of Table 1 for the D region of the ionosphere.
Figure 2. The relative speed of light (phase velocity) [] and relative 1 e distance [ ] plotted as a function of fre quency for
the representative data of Table 1 for the E region of the ionosphere.
AP_Mag_Apr_2013_Final.indd 71 5/18/2013 8:57:06 PM
72 IEEE Antennas and Propagation Magazine, Vol. 55, No. 2, April 2013
parallel to
0
B there are two roots, each of which corresponds to
a different sense of circular polarization, known as the R-wave
and the L-wave. Because each root ck depends, through
1
r ,
on the sense of polarization, this causes a splitting of the
dispersion curves that is determined by the magnitude of
0
B .
Gyromagnetic behavior of the elec trons and ions gives rise to
this type of phenomenon, which is one that may be more
familiar in the context of non-conduct ing magnetic media, such
as ferrites. When the background magnetic feld is removed,
both case (ii) solutions degenerate to an ordinary wave.
When collisions are ignored,
0
r j and
( )
2
2 r j , upon which the solutions given in Equa-
tion (26) (Table 2) reduce to the much simpler forms obtained
from the Appleton-Hartree equation in the absence of damp ing.
However, Equations (23) and (26) are only a little more
complicated. Within the scope of the given three-component
gas model free of external forces, approximations enter into our
solution only when estimating the gas parameters required to
fnd the resistivity components
l
r . As an illustration, we have
used the representative data of Table 1 to generate the
l
r from
Equation (24), and thence, via Equation (23), the disper sion
curves that are displayed in Figures 1-3. However, this is as
good as we can hope to achieve, and is in contrast to the situation
in which approximations are implicit in the construc tion of the
solution itself. For example, solutions based on the Appleton-
Hartree formulae with damping do not generally include the
effects of ion-neutral collisions or ion-cyclotron resonance,
which is equivalent to making
0
,
i
, and van ish from
Equations (19) and (24), whereas here these features are
retained.
Figures 1-3 are included as an illustration of solutions
calculated from Equation (23), which allows to be freely
chosen. The horizontal axis represents the natural frequency, f,
in Hz. Low attenuation occurs when d , the relative 1 e
attenuation distance, is much greater than one, and vice versa.
Here, d is the absolute 1 e attenuation distance, and is the
actual wavelength, 2 Re k , rather than the free-space wave-
length.
When 90 = , modes 1 and 2 are commonly referred to
as the ordinary and extraordinary waves. On the other hand,
when 0 = , they correspond to the choices of either left or
right circular polarization. These circularly polarized modes
behave very differently, as in one case the gyration of the
electrons is in the same sense as the polarization, whereas in the
other case it is contrary. This gives rise to nonreciprocal
propagation and the rapidly changing behavior of mode 2 in the
region of
1
2
p
f

= .
Intermediate values of give rise to a linear combina tion
of these principal modes, with the results changing smoothly
from the one limit to the other. However, in the E and F regions,
as shown in the comparison between subfgures (a), (c) and (e)
or (b), (d) and (f), they change more rapidly more as
approaches 90.
Dispersion curves such as those in Figures 1-3 reveal
interesting features about radio propagation in the upper
atmosphere. For example, we may readily fnd those modes and
frequency ranges that will support communications. For this,
we require 1 d , i.e., the relative 1 e distance must be
large, while at the same time the relative phase velocity c c
should be free of dispersion, i.e., fat, over the bandwidth of
interest. However, these parameters may vary considerably
with altitude. The situation when clear propagation exists in the
D region, but one of the upper layers is highly refective or
refractive, is of particular interest. For example, Figure 1e
shows that propagation is viable in the D region by means of the
O-wave at a frequency of 2 MHz ( 100 15 d = km) whereas
in Figure 2e we see that the E region is cut off, and highly
refective, below ~2.8 MHz. As is well known to ama teur radio
enthusiasts, these circumstances allow a radio wave to be
bounced between terrestrial sites over distances that exceed
normal line-of-sight propagation. Quite a separate feature is
that the

R-wave ( 0 = , mode 2, in Figures 1b, 2b, and 3b)
allows radio propagation over a range of audio fre quencies.
However, from the positive slope of the curve it is clear that this
mode is dispersive, with the higher frequencies traveling faster
than the lower frequencies. In order to main tain 0 = , these
modes must follow the Earths magnetic-feld lines, so that a
highly energetic electromagnetic distur bance originating from,
say, a lightning strike may take the path of a feld line and
propagate over a considerable distance through the ionosphere.
With velocities very much lower than c, the propagation delay
of these waves may then be of the order of seconds. When
received, they give rise to an audible down-chirp, or whistler,
because of the signifcant degree of dispersion involved.
Although they are of little use for com munication purposes,
they provide a means of detecting elec trical storms at great
distances.
5. Other Useful Results and Conclusions
It is of interest to explore further uses for the treatment of as
an algebraic operator. For example, it can be applied to fnding
the gas velocities. Although it seems that calculating the
velocity of the neutral gas from Equation (11) would require
evaluation of
1
, we can actually get a more useful result by
substituting
( )
( )
1
1 2
0 e e
m N e j j

( = +

from
Equation (13) to give
Figure 3. The relative speed of light (phase velocity) [] and relative 1 e distance [ ] plotted as a function of fre quency for
the representative data of Table 1 for the F region of the ionosphere.
AP_Mag_Apr_2013_Final.indd 72 5/18/2013 8:57:07 PM
IEEE Antennas and Propagation Magazine, Vol. 55, No. 2, April 2013 73
parallel to
0
B there are two roots, each of which corresponds to
a different sense of circular polarization, known as the R-wave
and the L-wave. Because each root ck depends, through
1
r ,
on the sense of polarization, this causes a splitting of the
dispersion curves that is determined by the magnitude of
0
B .
Gyromagnetic behavior of the elec trons and ions gives rise to
this type of phenomenon, which is one that may be more
familiar in the context of non-conduct ing magnetic media, such
as ferrites. When the background magnetic feld is removed,
both case (ii) solutions degenerate to an ordinary wave.
When collisions are ignored,
0
r j and
( )
2
2 r j , upon which the solutions given in Equa-
tion (26) (Table 2) reduce to the much simpler forms obtained
from the Appleton-Hartree equation in the absence of damp ing.
However, Equations (23) and (26) are only a little more
complicated. Within the scope of the given three-component
gas model free of external forces, approximations enter into our
solution only when estimating the gas parameters required to
fnd the resistivity components
l
r . As an illustration, we have
used the representative data of Table 1 to generate the
l
r from
Equation (24), and thence, via Equation (23), the disper sion
curves that are displayed in Figures 1-3. However, this is as
good as we can hope to achieve, and is in contrast to the situation
in which approximations are implicit in the construc tion of the
solution itself. For example, solutions based on the Appleton-
Hartree formulae with damping do not generally include the
effects of ion-neutral collisions or ion-cyclotron resonance,
which is equivalent to making
0
,
i
, and van ish from
Equations (19) and (24), whereas here these features are
retained.
Figures 1-3 are included as an illustration of solutions
calculated from Equation (23), which allows to be freely
chosen. The horizontal axis represents the natural frequency, f,
in Hz. Low attenuation occurs when d , the relative 1 e
attenuation distance, is much greater than one, and vice versa.
Here, d is the absolute 1 e attenuation distance, and is the
actual wavelength, 2 Re k , rather than the free-space wave-
length.
When 90 = , modes 1 and 2 are commonly referred to
as the ordinary and extraordinary waves. On the other hand,
when 0 = , they correspond to the choices of either left or
right circular polarization. These circularly polarized modes
behave very differently, as in one case the gyration of the
electrons is in the same sense as the polarization, whereas in the
other case it is contrary. This gives rise to nonreciprocal
propagation and the rapidly changing behavior of mode 2 in the
region of
1
2
p
f

= .
Intermediate values of give rise to a linear combina tion
of these principal modes, with the results changing smoothly
from the one limit to the other. However, in the E and F regions,
as shown in the comparison between subfgures (a), (c) and (e)
or (b), (d) and (f), they change more rapidly more as
approaches 90.
Dispersion curves such as those in Figures 1-3 reveal
interesting features about radio propagation in the upper
atmosphere. For example, we may readily fnd those modes and
frequency ranges that will support communications. For this,
we require 1 d , i.e., the relative 1 e distance must be
large, while at the same time the relative phase velocity c c
should be free of dispersion, i.e., fat, over the bandwidth of
interest. However, these parameters may vary considerably
with altitude. The situation when clear propagation exists in the
D region, but one of the upper layers is highly refective or
refractive, is of particular interest. For example, Figure 1e
shows that propagation is viable in the D region by means of the
O-wave at a frequency of 2 MHz ( 100 15 d = km) whereas
in Figure 2e we see that the E region is cut off, and highly
refective, below ~2.8 MHz. As is well known to ama teur radio
enthusiasts, these circumstances allow a radio wave to be
bounced between terrestrial sites over distances that exceed
normal line-of-sight propagation. Quite a separate feature is
that the

R-wave ( 0 = , mode 2, in Figures 1b, 2b, and 3b)
allows radio propagation over a range of audio fre quencies.
However, from the positive slope of the curve it is clear that this
mode is dispersive, with the higher frequencies traveling faster
than the lower frequencies. In order to main tain 0 = , these
modes must follow the Earths magnetic-feld lines, so that a
highly energetic electromagnetic distur bance originating from,
say, a lightning strike may take the path of a feld line and
propagate over a considerable distance through the ionosphere.
With velocities very much lower than c, the propagation delay
of these waves may then be of the order of seconds. When
received, they give rise to an audible down-chirp, or whistler,
because of the signifcant degree of dispersion involved.
Although they are of little use for com munication purposes,
they provide a means of detecting elec trical storms at great
distances.
5. Other Useful Results and Conclusions
It is of interest to explore further uses for the treatment of as
an algebraic operator. For example, it can be applied to fnding
the gas velocities. Although it seems that calculating the
velocity of the neutral gas from Equation (11) would require
evaluation of
1
, we can actually get a more useful result by
substituting
( )
( )
1
1 2
0 e e
m N e j j

( = +

from
Equation (13) to give
Figure 3. The relative speed of light (phase velocity) [] and relative 1 e distance [ ] plotted as a function of fre quency for
the representative data of Table 1 for the F region of the ionosphere.
AP_Mag_Apr_2013_Final.indd 73 5/18/2013 8:57:07 PM
74 IEEE Antennas and Propagation Magazine, Vol. 55, No. 2, April 2013
( )
( )
2
0
e
n
e
e
m e
DH EG AH BG
m j j
N e


(
= + ( (

+
(

v E


( )( )
( ) ( ) ( )
0
1
e i i e
e
j
j j N e


( = + +

+
E
(27)

( )
( ) ( )
( )
0
e i i e
e
j
j j N e



( = + +

+
J


( )
( )
( )
0
e
e i
j h
j j


+

+
v v

.
Note that since is a polynomial in , it must commute with
, allowing us to reorder the equation so that is brought
together with E. This result allows
n
v to be com pared directly
with
e i
v v .
In contrast, making a comparison of
i
v with respect to
e
v

does require the use of Equation (7) to invert a function of .
This is of interest because it leads to the conditions upon which
the ion motion may be neglected, a problem considered by
Abbas [12]. From Equation (11),
( ) ( )
1
i e
e m EK FH KB CH

= + (

v ,
( ) ( )
1
e e
e m GF KD CG AK

= + (

v ,
( ) ( )
i e
GF KD CG AK EK FH KB CH + = + ( (

v v
( ) ( )
{ }
2
i e i i e i
j j ( ( + + + + +

v
( ) ( )
{ }
2
i e e i e e
j j ( ( = + + + +

v
(28)
which is of the simple form
( ) ( )
i e
a b a b + = + v v ,
( )
i e
b j ( = + +

,
(28)

i
a b j = + ,
( )
e
a b j = + .
Although
1


is once again superfuous, we must invert
a b +

to be able to obtain
i
v in terms of
e
v . Equation (7)
gives
( )
( )
2 2 2 2 2
1
2 2 2
a b ab b
a b
a a b

+ +
+ =
+

. (29)
Applying this to Equation (28) allows us to fnd
i e
v v as the
mandatory quadratic in :

( )
( )
2 2 2 2 2
2 2 2
i
e
a b ab b
a b
a a b

+ +
= +
+
v
v


( )
( )
( ) ( )
2 2 2 2 2
2 2 2
1
a a b ab a a b a a
a a b

(
= + +
(

+

( )
( )
2
2 2 2
b a a
a b
a a
a b

| |
= +
|
\ .
+
. (30)
Once again, the result is exact in principle, but while it would
be interesting, it is not within our scope to explore it further.
Rather, our purpose has been to demonstrate the usefulness of
the algebraic approach in obtaining a simple, methodical tech-
nique that leads to the analytical solution of certain types of
electromagnetic problems that would be otherwise diffcult.
While it is comparatively easy to solve an equation such as
i i e e
a b a b + = + v v v v by standard means when is
along a chosen basis vector, it is considerably more trouble-
some when this is not the case. Equation (30) involves no such
problem, as and
2
are easily written down in any basis
from Equations (3) and (5).
Despite the undeniable benefts of computer modeling
packages, we hope to have demonstrated in this article that it is
still worthwhile to attempt a systemic analysis of a problem
before resorting to other methods that are simply expedient. We
have focused on just one such example where the search
happens to be fruitful, but the principle applies to any problem
that falls into the middle range of diffculty. The simplifying
factor may be some special feature that allows a particular form
of approach, for example, it sometimes happens that a problem
that seems diffcult in the time domain may simplify in the
frequency domain, or vice versa. If a search for a suit able
trapdoor is not fruitful, we are then justifed in resorting to
other means; but when a key simplifying fact does emerge, as
shown here, the benefts can be signifcant. The humble algebra
of the operator turned out to be the simplifying factor that
allowed us to avoid a cumbersome 9 9 matrix representation
that would generally be considered to be ana lytically intractable.
Ad hoc methods of solution, by their very nature, lack systematic
methodology and so may prove to be cumbersome, which
encourages simplifying approximations to be brought in early.
By avoiding such pitfalls, we were able to proceed directly to a
readily solvable 3 3 matrix problem in which the
approximations apply only to the available data, not to the
method of solution.
6. Acknowledgement
This article is dedicated to Professor C. O. Hines, who
gave the author his grounding in electromagnetic theory and set
the original project at the University of Toronto.
7. References
1. S. Chapman and R. G. Cowling, Mathematical Theory
of Non-uniform Gases, Cambridge UK, Cambridge Uni-
versity Press, 1952.
2. J. A. Ratcliffe, The Magneto-Ionic Theory, Cambridge
UK, Cambridge University Press, 1959.
3. L. S. Alperovich and E. N. Fedorov, Hydromagnetic Waves
in the Magnetosphere and the Ionosphere, Dordrecht, Nether-
lands, Springer, 2007.
4. J. A. Ratcliffe, An Introduction to the Ionosphere and
Magnetosphere, Cambridge UK, Cambridge University
Press, 1972.
5. J. H. Piddington, Solar Atmospheric Heating by Hydro-
magnetic Waves, Monthly Notices of the Royal Astronomi cal
Society, 116, 1956, pp. 314-323.
6. J. H. Piddington, Geomagnetic Storms, Auroras and
Associated Effects, Space Science Reviews, 3, 5-6, 1964,
pp. 724-780.
7. C. O. Hines and H. Bondi, Generalized Magneto-Hydro-
dynamic Formulae, Proceedings of the Cambridge Phi-
losophical Society, 49, 1953, pp. 299-.
8. C. O. Hines, Notes on Waves in Plasmas, in E. C.
Jordan (ed.), Symposium on Electromagnetic Theory and
Antennas, Copenhagen, June 25-30 1962, Oxford, Per-
gammon Press, 1963.
9. J. A. Fejer, Hydromagnetic Wave Propagation in
the Ionosphere, Journal of Atmospheric and Terrestrial
Physics, 18, 1960, pp. 135-146.
10. C. O. Hines, I. Paghis, et al. (eds.), Physics of the
Earths Upper Atmosphere, Englewood Cliffs, NJ, Pren tice-
Hall, 1965.
11. S-I. Akasofu, Dispersion Relation of Magneto-Hydro-
dynamic Waves in the Ionosphere and its Application to
Shock Wave, Science reports of Tohoku University, Series 5,
Geophysics, 8, 1, 1956, pp. 24-40.
12. M. Abbas, Conditions for Neglecting Ion Motion for Radio
Waves in a Plasma, IEEE Transactions on Antennas and
Propagation, 18, 3, May 1970, pp. 436-438.
13. B. S. Tanenbaum and D. Mintzer, Wave Propaga tion
in a Partly Ionized Gas, Physics of Fluids, 5, 1962, pp.
1226-37.
14. J. W. Arthur, Final Year Physics 430 Project Report,
Supervisor C. O. Hines, University of Toronto, 1971.
15. C. O. Hines, The Relation Between Magnetohydrody-
namic Waves and Magneto-Ionic Theory in C. O. Hines
(ed.), The Upper Atmosphere in Motion: a Selection of Papers
with Annotation (American Geophysical Union Monograph),
Worcester, MA, Heffernan Press, 1974, Paper 36, pp. 862-872
and Note 36.1, p. 875.
16. G. Birkhoff and S. Mac Lane, A Survey of Modern Alge bra,
Third Edition, New York NY, Macmillan, 1965.
17. J. D. Jackson, Classical Electrodynamics, New York NY,
Wiley, 1967.
18. K. C. Yeh and C-H. Liu, Theory of Ionospheric Waves, New
York, Academic Press, 1972.
19. D. Bilitza, International Reference Ionosphere IRI-2007,
online at http://omniweb.gsfc.nasa.gov/vitmo/iri_vitmo.html.
20. A. E. Hedin, MSIS-E-90 Atmosphere Model, online at http://
omniweb.gsfc.nasa.gov/vitmo/msis_vitmo.html.
21. S. Chapman, The Electrical Conductivity of the Atmos-
phere: A Review, Nuovo Cimento, 4, 10, Suppl. 4, 1956, pp.
1385-1412.
AP_Mag_Apr_2013_Final.indd 74 5/18/2013 8:57:08 PM
IEEE Antennas and Propagation Magazine, Vol. 55, No. 2, April 2013 75
( )
( )
2
0
e
n
e
e
m e
DH EG AH BG
m j j
N e


(
= + ( (

+
(

v E


( )( )
( ) ( ) ( )
0
1
e i i e
e
j
j j N e


( = + +

+
E
(27)

( )
( ) ( )
( )
0
e i i e
e
j
j j N e



( = + +

+
J


( )
( )
( )
0
e
e i
j h
j j


+

+
v v

.
Note that since is a polynomial in , it must commute with
, allowing us to reorder the equation so that is brought
together with E. This result allows
n
v to be com pared directly
with
e i
v v .
In contrast, making a comparison of
i
v with respect to
e
v

does require the use of Equation (7) to invert a function of .
This is of interest because it leads to the conditions upon which
the ion motion may be neglected, a problem considered by
Abbas [12]. From Equation (11),
( ) ( )
1
i e
e m EK FH KB CH

= + (

v ,
( ) ( )
1
e e
e m GF KD CG AK

= + (

v ,
( ) ( )
i e
GF KD CG AK EK FH KB CH + = + ( (

v v
( ) ( )
{ }
2
i e i i e i
j j ( ( + + + + +

v
( ) ( )
{ }
2
i e e i e e
j j ( ( = + + + +

v
(28)
which is of the simple form
( ) ( )
i e
a b a b + = + v v ,
( )
i e
b j ( = + +

,
(28)

i
a b j = + ,
( )
e
a b j = + .
Although
1


is once again superfuous, we must invert
a b +

to be able to obtain
i
v in terms of
e
v . Equation (7)
gives
( )
( )
2 2 2 2 2
1
2 2 2
a b ab b
a b
a a b

+ +
+ =
+

. (29)
Applying this to Equation (28) allows us to fnd
i e
v v as the
mandatory quadratic in :

( )
( )
2 2 2 2 2
2 2 2
i
e
a b ab b
a b
a a b

+ +
= +
+
v
v


( )
( )
( ) ( )
2 2 2 2 2
2 2 2
1
a a b ab a a b a a
a a b

(
= + +
(

+

( )
( )
2
2 2 2
b a a
a b
a a
a b

| |
= +
|
\ .
+
. (30)
Once again, the result is exact in principle, but while it would
be interesting, it is not within our scope to explore it further.
Rather, our purpose has been to demonstrate the usefulness of
the algebraic approach in obtaining a simple, methodical tech-
nique that leads to the analytical solution of certain types of
electromagnetic problems that would be otherwise diffcult.
While it is comparatively easy to solve an equation such as
i i e e
a b a b + = + v v v v by standard means when is
along a chosen basis vector, it is considerably more trouble-
some when this is not the case. Equation (30) involves no such
problem, as and
2
are easily written down in any basis
from Equations (3) and (5).
Despite the undeniable benefts of computer modeling
packages, we hope to have demonstrated in this article that it is
still worthwhile to attempt a systemic analysis of a problem
before resorting to other methods that are simply expedient. We
have focused on just one such example where the search
happens to be fruitful, but the principle applies to any problem
that falls into the middle range of diffculty. The simplifying
factor may be some special feature that allows a particular form
of approach, for example, it sometimes happens that a problem
that seems diffcult in the time domain may simplify in the
frequency domain, or vice versa. If a search for a suit able
trapdoor is not fruitful, we are then justifed in resorting to
other means; but when a key simplifying fact does emerge, as
shown here, the benefts can be signifcant. The humble algebra
of the operator turned out to be the simplifying factor that
allowed us to avoid a cumbersome 9 9 matrix representation
that would generally be considered to be ana lytically intractable.
Ad hoc methods of solution, by their very nature, lack systematic
methodology and so may prove to be cumbersome, which
encourages simplifying approximations to be brought in early.
By avoiding such pitfalls, we were able to proceed directly to a
readily solvable 3 3 matrix problem in which the
approximations apply only to the available data, not to the
method of solution.
6. Acknowledgement
This article is dedicated to Professor C. O. Hines, who
gave the author his grounding in electromagnetic theory and set
the original project at the University of Toronto.
7. References
1. S. Chapman and R. G. Cowling, Mathematical Theory
of Non-uniform Gases, Cambridge UK, Cambridge Uni-
versity Press, 1952.
2. J. A. Ratcliffe, The Magneto-Ionic Theory, Cambridge
UK, Cambridge University Press, 1959.
3. L. S. Alperovich and E. N. Fedorov, Hydromagnetic Waves
in the Magnetosphere and the Ionosphere, Dordrecht, Nether-
lands, Springer, 2007.
4. J. A. Ratcliffe, An Introduction to the Ionosphere and
Magnetosphere, Cambridge UK, Cambridge University
Press, 1972.
5. J. H. Piddington, Solar Atmospheric Heating by Hydro-
magnetic Waves, Monthly Notices of the Royal Astronomi cal
Society, 116, 1956, pp. 314-323.
6. J. H. Piddington, Geomagnetic Storms, Auroras and
Associated Effects, Space Science Reviews, 3, 5-6, 1964,
pp. 724-780.
7. C. O. Hines and H. Bondi, Generalized Magneto-Hydro-
dynamic Formulae, Proceedings of the Cambridge Phi-
losophical Society, 49, 1953, pp. 299-.
8. C. O. Hines, Notes on Waves in Plasmas, in E. C.
Jordan (ed.), Symposium on Electromagnetic Theory and
Antennas, Copenhagen, June 25-30 1962, Oxford, Per-
gammon Press, 1963.
9. J. A. Fejer, Hydromagnetic Wave Propagation in
the Ionosphere, Journal of Atmospheric and Terrestrial
Physics, 18, 1960, pp. 135-146.
10. C. O. Hines, I. Paghis, et al. (eds.), Physics of the
Earths Upper Atmosphere, Englewood Cliffs, NJ, Pren tice-
Hall, 1965.
11. S-I. Akasofu, Dispersion Relation of Magneto-Hydro-
dynamic Waves in the Ionosphere and its Application to
Shock Wave, Science reports of Tohoku University, Series 5,
Geophysics, 8, 1, 1956, pp. 24-40.
12. M. Abbas, Conditions for Neglecting Ion Motion for Radio
Waves in a Plasma, IEEE Transactions on Antennas and
Propagation, 18, 3, May 1970, pp. 436-438.
13. B. S. Tanenbaum and D. Mintzer, Wave Propaga tion
in a Partly Ionized Gas, Physics of Fluids, 5, 1962, pp.
1226-37.
14. J. W. Arthur, Final Year Physics 430 Project Report,
Supervisor C. O. Hines, University of Toronto, 1971.
15. C. O. Hines, The Relation Between Magnetohydrody-
namic Waves and Magneto-Ionic Theory in C. O. Hines
(ed.), The Upper Atmosphere in Motion: a Selection of Papers
with Annotation (American Geophysical Union Monograph),
Worcester, MA, Heffernan Press, 1974, Paper 36, pp. 862-872
and Note 36.1, p. 875.
16. G. Birkhoff and S. Mac Lane, A Survey of Modern Alge bra,
Third Edition, New York NY, Macmillan, 1965.
17. J. D. Jackson, Classical Electrodynamics, New York NY,
Wiley, 1967.
18. K. C. Yeh and C-H. Liu, Theory of Ionospheric Waves, New
York, Academic Press, 1972.
19. D. Bilitza, International Reference Ionosphere IRI-2007,
online at http://omniweb.gsfc.nasa.gov/vitmo/iri_vitmo.html.
20. A. E. Hedin, MSIS-E-90 Atmosphere Model, online at http://
omniweb.gsfc.nasa.gov/vitmo/msis_vitmo.html.
21. S. Chapman, The Electrical Conductivity of the Atmos-
phere: A Review, Nuovo Cimento, 4, 10, Suppl. 4, 1956, pp.
1385-1412.
AP_Mag_Apr_2013_Final.indd 75 5/18/2013 8:57:08 PM
76 IEEE Antennas and Propagation Magazine, Vol. 55, No. 2, April 2013
Introducing the Feature Article Author
John Arthur trained as a physicist, gaining a BSc from
the University of Toronto and a PhD from the University of
Edinburgh. After initially working on light scattering from
lattice vibrations in crystals, he took up a career as an engi-
neer. He frst researched on the applications of CCD FIR fl-
ters, and then moved to an industrial post with Racal MESL
Ltd. to develop applications for SAW devices in r adar, com-
munications, and EW. While there, he also worked in the felds
of electronics, signal processing, and microwaves, becoming
frst a board member with the company and then Technical
Director. During this time, he led the teams that gained the
company a Queens Award for Technology and a Design
Council Millennium Product Award. He was a council member
of the Scottish Optoelectronic Association, and also served
with local universities as an external examiner and a member of
industrial advisory boards.
From 2005 to 2012, he was an Honorary Fellow in the
School of Engineering at the University of Edinburgh. In 2012,
he was appointed as a trustee of the James Clerk Maxwell
Foundation. He has recently authored an AP-S spon sored
entitled Understanding Geometric Algebra for Electro magnetic
Theory (Wiley/IEEE). During his career, he has instigated
several patents and published a number of papers in physics
and engineering. He was the recipient of the 1997 IEE Oliver
Lodge premium, and the 2010 IEEE Donald Fink award for
best tutorial paper. A Senior Member of the IEEE, he is also a
Fellow of the Royal Society of Edinburgh, the Royal Academy
of Engineering, the Institute of Engineering and Technology,
and of the Institute of Physics (UK).
An Engineered Conductor for
Gain and Effciency Improvement of
Miniaturized Microstrip Antennas
Saeed I. Latif, Lotfollah Shafai, and Cyrus Shafai
Department of Electrical and Computer Engineering
University of Manitoba
75 Chancellors Circle, Winnipeg, Manitoba, Canada R3T 5V6
Tel: +1 (204) 474-9615; Fax: +1 (204) 269-0381
E-mail: latif@cc.umanitoba.ca; Lot.Shafai@ad.umanitoba.ca; Cyrus.Shafai@ad.umanitoba.ca
Abstract
This paper reviews the concept of an engineered conductor, introduced by the authors in order to reduce ohmic losses
of miniaturized microstrip antennas [1-4]. Because of the miniaturization, the ohmic losses of microstrip antennas
increase, which essentially signicantly reduces their gain and efcien cy. By the use of the engineered conductor
concept, these two important parameters of such antennas can be improved without altering the antennas geometry.
The concept is based on using multiple laminated thin conductors, rather than one thick conducting layer, to form
the microstrip antenna. The technique is applied to several miniaturized microstrip antennas in order to reduce their
ohmic losses. The concept is explained using a microstrip line to demonstrate that the conductor loss can be reduced
by increasing the number of layers in the lamination, while keeping the total thickness constant. Using conventional
metallized substrates, the lamination reduces the thickness of each conductor layer to approximately or less than the
skin depth in that conductor. Studies of two miniaturized antennas namely, the square-ring antenna and the modied
open-ring antenna have respectively provided about 4.6 dB and 1.5 dB improvements in the gain, and from 30%
to 40.7% improvement in the efciency. Experimental investigations are also presented that conrmed the simulated
results.
Keywords: Gain and efciency enhancement; miniaturized microstrip antennas; multiple laminated patch conductors;
ohmic loss reduction
1. Introduction
A
ntenna miniaturization has become an essential part
in todays wireless equipment, because the handheld
devices are experiencing a huge decrease in physical dimen-
sions. Miniaturized antennas are associated with several
problems, such as narrow impedance bandwidth, poor radia tion
resistance, low gain and effciency, and high cross polari zation
[5-8]. Among these, the reduction in gain and effciency are the
signifcant problems; they occur due to increased ohmic losses
caused by miniaturization.
We presented a new technique of an engineered conduc-
tor [1-4], which utilizes the concept of laminated conductors,
similar to that widely used in coaxial cables and transmission
lines. There, thin laminated conductors have been used instead
of a single conductor, which essentially increases the penetra-
tion depth of the electromagnetic wave [9, 10]. The attenua-
tion is therefore less [11], which indicates a reduction in ohmic
losses. In the case of miniaturized microstrip antennas, the
patch conductor is engineered with a few laminating dielectric
layers, and signifcant improvements in gain and effciency are
observed.
This paper discusses the concept of laminating conduc tors,
and reviews the work conducted on this technique. In Section 2,
the concept is applied to two different miniaturized microstrip
antennas in order to enhance their gain and eff ciency. The frst
example is a signifcantly reduced-sized modifed square-ring
antenna with a gap loading, on a dielec tric substrate. Despite
AP_Mag_Apr_2013_Final.indd 76 5/18/2013 8:57:08 PM

You might also like