Sunteți pe pagina 1din 7

1 io

IEEE TRANSACTIONS ON EDUCATION , VOL. 3 6 , NO. 4 . NOVEMBER 1993

Solution of the Maxwell Field Equations


in Vacuum for Arbitrary Charge and
Current Distributions Using the
Methods of Matrix Algebra
Richard P. Bocker and B. Roy Frieden

Abstruct- Presented in this paper is a new matrix representation of classical electromagnetic theory. The basis
4 )f
this representation is a space-time, eight-by-eight difrerential matrix operator. This matrix operator is initially
lormulated from the differential form of the Maxwell field
equations for vacuum. The resulting matrix formulation of
Maxwells equations allows simple and direct derivation of:
i he electromagnetic wave and charge continuity equations;
I he Lorentz conditions and definition of the electromagnetic
jotentials; the Lorentz and Coulomb gauges; the electromagnetic
2otential wave equations; and Poyntings conservation of
1:nergy theorem. A four-dimensional Fourier transform of
:he matrix equations casts them into an eight-dimensional
transfer theorem. The transfer function has an inverse, and
this allows the equations to be inverted. This expresses the
fields directly in terms of the charge and current source
distributions.

--

__

-ay

dX

dr
dz
d
- a_
dz
dr
d
0

d
dz

--

ax
O

-dT

d
dT

--

dz

where the temporal component of a space-time point is den0 ed


by

I. INTRODUCTION

RESENT day mathematical descriptions and computations in the field of classical electrodynamics rely heavily
on the use of vector calculus. Matrix calculus, on the other
hand, finds widespread use in other areas of physics, engineering, and applied mathematics. Structural analysis, vibrational
analysis, electrical circuit analysis, and the theory of elasticity
are just a few of the disciplines where matrix calculus [l]
has been successfully employed routinely. This paper presents
a new formulation of electromagnetic theory from a matrix
calculus point of view. As will be shown, the formulation
arises from use of a skew-Hermitian space-time eight-by-eight
differential matrix operator

11. DERIVATION
OF THE [ M ]MATRIXREPRESENTATION

The fundamental equations of classical electromagnetic phenomena, namely the Maxwell field equations in vector f o m ,
serve as our starting point in this investigation. In this section
the space-time operator [MI, which is of paramount importance to the remainder of the paper, is constructed from the
Maxwell field equations. The four Maxwell field equations,
in SI (Systeme International) units [2], are given by the
following:
Ampere-Maxwell Law

d
V x H ( r ,t ) = - D ( r , t ) + J e ( r t, )

at

( 3a)

Gauss Law for Electricity

V . D ( r .t ) = p e ( r , t )

0162-8828/93$03.00 0 1993 IEEE

- ~-

(2)

The imaginary unit


is denoted by z and c represents the
speed of light in vacuum.

Manuscript received April 1991; revised April 1992.


R. P. Bocker is with the Naval Command, Control and Ocean Surveillance
Center, San Diego, CA 92152.
B. R. Frieden is with the Optical Sciences Center, University of Arizona,
Tucson, AZ 85721.
IEEE Log Number 9211757.

-.

= ict.

( 3b)

IEEE TRANSACTIONS ON EDUCATION,

VOL. 3 6 , NO. 4 ,

NOVEMBER 1993

35 1

F iraday's Law

d
V x E ( r ,t ) = - - B ( r , t ) - P
dt

( r ,t )

C auss' Law for Magnetism

V . B ( r ,t ) = p m ( r . t ) .

(34

The order in which the four equations appear above is impor ant in the construction of the space-time operator [MI.
Thi scalar and vector physical quantities appearing in the
Ma well field equations are: E ( r ,t ) the electric field, D ( r ,t )
the electric displacement, B ( r .t ) the magnetic induction,
H(~I-,
t ) the magnetic field, J e ( r ,t ) the electric current density,
s' : r ,t ) the magnetic current density, p e ( r , t ) the electric
cha-ge density, p m ( r , t ) the magnetic charge density, and ( r .t )
a sf ace-time point. Both magnetic charge and current densities
[2] have been included in Maxwell's equations for purposes
of :ompleteness. They, of course, may be set equal to zero
sin( e magnetic charge has not been discovered in nature.
'I'he electric displacement and electric field, as well as the
ma:;netic induction and magnetic field, are related [2] through
the expressions
1

D ( r ,t ) = c,E(r, t )

+ P ( r ,t ) and
+ p o M ( r .t )

B ( r ,t ) = p , H ( r , t )

(4)

Other authors have formed electromagnetic matrix operators


similar to (6), but either under less general conditions, or
with different goals in mind [3], [4]. Mayes [3] assumes the
special case of time-harmonic fields due to radiating antennas,
so that time is suppressed in the formulation. The effect is
suppression of Maxwell's equations (3b) and (3d), resulting in
a six-by-six matrix operator formulation instead of eight-by-six
matrix operator formulation (6). Eisele and Mason [4] assume
zero source strengths, and form complex field quantities,
E, /E,, E ? .H , iHY.H,. with the goal of showing that
Maxwell's equations reduce to a Dirac-like equation for a
four-component electromagnetic spinor field with zero mass.
Also, no other authors, as far as we know, take the step
described next of broadening the long (eight-by-six) matrix
in (6) into an associated square (eight-by-eight) matrix, with
attendant theoretical benefits.
We are now in a position to construct the desired eight-byeight space-time operator [MI. The key to this construction
is to first replace the eight-by-six matrix operator appearing
in (6) by an eight-by-eight matrix operator containing two
additional columns. That is,

wh :re P ( r ,t ) is the macroscopic polarization, M ( r ,t ) is


the macroscopic magnetization, E , is the permittivity of free
spaze, and po is the permeability of free space. In this paper,
die ectric and magnetic materials are not considered. Hence,
we let

P ( r ,t ) = 0 and M ( r ,t ) = 0.

(5)

'45th the use of (4) and (5), the four Maxwell's equations
car be rewritten as eight scalar equations. The eight scalar
eqi ations can then be combined into a single matrix equation
involving an eight-by-six differential matrix

lo
0
0

(7)

dx
0

d
-

_-

dY
0

ax
0
0

0
d

ay

%I
6,

--

az

Columns 4 and 8 are the newly inserted columns. Their


elements mzJare as yet undefined. To null-out the effects of
the two new columns, additional zeros must be inserted in
the electric and magnetic field vector at positions 4 and 8,

352

IEEE TRANSACTIONS ON EDUCATION , VOL. 3 6 , NO 4 , NOVEMBER

as shown above. This eight-by-eight matrix operator representation of Maxwell's equations is now entirely equivalent to
the original vector representation, regardless of choice of the
added elements h i j . This is easy to verify. We next choose
useful, specific values for the added k i j . The motivation for
choice of the added elements h i j is as follows. The offdiagonal known elements in the eight-by-eight matrix have
an antisymmetric arrangement. For example, element h 2 5 is
d / d z and m 5 2 is -d/dz. This suggests that we choose the
added elements h i j such that antisymmetry is preserved. This
requires that we choose the off diagonal elements hi? as
follows:

m i 4 =m5g

= --,
dX

h 2 4

=m6g =

and

The left-hand eight-by-eight matrix is the matrix (1) we


previously called [MI. Derivation of this matrix form (12) of
the Maxwell field equations is the chief result of this sectim.
This matrix will be used throughout the remainder of the paper.
As will be seen, it permits simple matrix operations [upon (12)]
to derive the major phenomena of electromagnetic theory. This
is the subject of Section 111. It also permits direct inversion for
the fields E and H in terms of the sources, without the need
for calculating intermediary potentials. This is the subject of
Sections IV and V.
Equation (12) can be rewritten in a more compact form

in terms of four-by-four submatrices


Next, the diagonal elements h 4 4 and h g g have to be specified. If these elements are chosen according to the equation

then the conditions for the Lorentz gauge are satisfied. This
will be seen in the next section. If, however, these elements
are chosen according to the equation

and
0
[M21

then the conditions of the Coulomb gauge are satisfied [ 5 ] .


Substituting the Lorentz gauge choice (lo), and the antisym0
metric choices (8), (9) in the general form (7) leads to the
following eight-by-eight covariant matrix representation [SI four-by-one vectoi fields
of Maxwell's equations:

_ _8
d7
0
0
d

d
dz

__

--

d7

o - 0

a7

ax

dz

_d _ _d

d
_

d7
0

_ _a
dz
0

ay

ax

--

87

--

d
d7
0

a
-

a
-

dx

ay

dX

ay

__

- _ _ _ - _ -

%
_ _Y
_
dz
?Jl
0
-

ax
d

--

?Jld

-dX

a
-_
az

82
0

0
and

0
0

o - - dXa
0

and four-by-one vector sources

--

a y
dz

dz

dr

I:

J,"

_ _ _ _
dr

993

Is11

a - -a
_

=a[
icp"

and

353

IEEE TRANSACTIONS ON EDUCATION, VOL. 3 6 , NO. 4 , NOVEMBER 1993

B. Electromagnetic Potentials and Lorentz Conditions

L icpm 1
in summary, the eight-by-eight matrix representation (13)
is equivalent to the Maxwell field equations. The matrix
rei resentation (13) is in covariant form. This development
wi 11 continue in Section IV, where four-dimensional Fourier
traisforms will be taken in order to achieve inversion of the
m: trix field equations.

We found before that the space-time operator [MI defines


the eight scalar Maxwell field equations. We now observe that
the complex conjugate of [MI also provides the definition
of the electromagnetic fields in terms of the familiar vector
and scalar potentials. In particular, for the Lorentz gauge, the
matrix equation

where
111. MATRIXFORMULATION
OF OTHER
ELECTROMAGNETIC
PHENOMENA

The matrix representation (13) of the Maxwell field equatiois will now be shown to imply, in simple fashion, some
milestone effects of classical electromagnetic theory. Simple
mzitrix operations upon the space-time operator [MI constricted in the previous section will be shown to derive (a)
thc electromagnetic wave and charge continuity equations, (b)
thc Lorentz conditions and definition of the electromagnetic
PO entials, (c) the electromagnetic potential wave equations,
an 1 (d) Poynting's conservation of energy theorem.

and

AT

(51

-a
b21=

a-

c
A. Electromagnetic Field Wave Equations and
CI urge Continuity Equations

Multiply both sides of matrix equation (13) by the conjugate


of the space-time operator [MI. This gives

is equivalent to eight scalar equations. Six of these are equivalent to the two vector equations which define the relationship
between the electromagnetic potentials and the electromagnetic fields

E=

Vp" - -Ae

at

x A"'

to

and
The matrix [D]is defined in terms of the D'Alembertian
o p :rator0

r0

01

The remaining two scalar equations correspond to the Lorentz


conditions

Lo

OJ

V . A " + P E -4"=0

"at

and

[(.I

is the 4-by-4 null matrix. The matrix representation (17)


is equivalent to eight scalar equations. It is easily shown that
th :se are equivalent to the vector electromagnetic field wave
ec uations

a2

V 2 E - --E
c2 at2
1

a2

C 2 H - --H
c2 at2

= -Vp"
E,

= -Vp"
p,

+ p , -ataJ e + V x J" (19a)


a - V x Je(19b)
+ E,-J"
at

arid the electric and magnetic charge continuity equations


and

V . J " + - p "a= o .

at

V.A"+p

E
O

-4"=O.

"at

(24'

Results (23) and (24) are easily verified by explicit multipli.


cation of the right-hand side of matrix equation (21).
A comparison of matrix equations (13) and (21) discloses
the curious fact that the same operator [MI that operates upon
the fields fi,i = 1 , 2 to give the sources s, also operates upon
the potentials p , to give the fields. There is an interesting
symmetry between the two effects.

C. Electromagnetic Potential Wave Equations


It is well known that the electromagnetic vector and scalar
potentials satisfy inhomogeneous wave equations. Once again,
we show this by simple matrix multiplication. First substitutc:

3: 4

IEEE TRANSACTIONS ON EDUCATION , VOL. 3 6 , NO. 4 , NOVEMBER 19'13

tlie matrix expression (21) into the matrix representation (13)


of Maxwell's equations. This gives

where S is the Poynting vector representing energy flow.


Again, this is easily verified by explicit multiplication out of
the matrix products in (30).

Mi
M2

Mi M2
M I ] M2 M I ] '

M2

[;:I

(25)
IV. ELECTROMAGNETIC
FIELD
FOURIERTRANSFORM
REPRESENTATION

I- owever, for the Lorentz gauge-choice of [ M ]

b y direct evaluation of the matrix product. Then, by (25) and


( l6), the following matrix representation of the electromagretic potential wave equations is obtained

By explicit evaluation of the left-hand matrix product, this


single matrix equation is equivalent to eight scalar equations.
Cix of these scalar equations are equivalent to the following
fair of wave equations for the electric and magnetic vector
I otentials

1 d2
V 2 A "- --.-A"
c2 at2

= - p o J"

and

A method for inverting the Maxwell field equations is given


in this section. The Maxwell field equations have been written
in a compact matrix form involving the square eight-by-eight
space-time differential operator [MI. With the use of fourdimensional Fourier transform theory, the matrix form of
Maxwell's equations can be easily cast from the differentid
form into an algebraic form. The algebraic form is an eightdimensional transfer theorem in K-space, linearly connectirig
the Fourier transform of the fields to those of the sources. The
connection is through an eight-dimensional transfer function,
whose inverse is known. This, in turn, permits inversion of
the field equations. The fields can then be calculated in temis
of the charge and current source distributions.
A three (lower-) dimensional Fourier transform approach to
direct inversion is described by Mayes [6]. This was for the
special case of time-harmonic fields due to antennas. Also,
inversion was made by recourse to Green's functions (i.e., use
of vector calculus) rather than by direct use of matrix transfer
theorem (38) below.
We first define the following four-dimensional Fourier transform pairs for the fields

-'he remaining two represent the wave equations for the


L lectric and magnetic scalar potentials

and

[:I =&/.I
J1:

>. Energy Conservation; Poynting 's Theorem

The law of conservation of energy, often called Poynting's


heorem [3], is an important milestone of electromagnetic the)ry. Again simply matrix manipulation of (13) will accomplish
he derivation. Multiply both sides of (13) by the Hermitian
:onjugate of the electromagnetic field vector.

[ ;:] [E: 21 I:;[

[ ; I " [::I.

=
(30)
rhis is a single scalar equation representing Poynting's the,rem
d
V . S -U
E . J" H . J"" = 0
(31)

+ dt +

If

we define quantity
U G

+ip,H. H

(35b)

R = ( Z I Y l Z , T ) = (7-,T)l
dR = dz dy dz dt
and
7

= ict

(:16)

K = (kz, k y , k z , k )= (k,h-),
dK = d k , dky d k z dw
and

(32)

where U is the total energy density of the electromagnetic


fields, and define S by

S=ExH,

[ i!] dK

where by definition

to obey

:,E.E

e+i(K'R)

(33)

k, = iw/c.

(37)

Next, multiplying both sides of (34b) by the space-tine


operator [MI and using (35b) with (13) leads to the following Fourier transform representation of the Maxwell fi1:ld

355

IEEl: TRANSACTIONS ON EDUCATION , VOL. 36 , NO. 4 , NOVEMBER 1993

Poles at K = 0 are avoided by recourse to Cauchy principal


value integration [8] in analytic problems; or by simply
avoiding the points K = 0 during numerical integration of
(43) in nonanalytic problems.

eq iations

wtlere

v.

ELECTROMAGNETIC
FIELDPOINT
SPREAD
FUNCTIONREPRESENTATION

--k,
0
0

L
0
kz
-k ,
- 0

0
-IC,
0

IC,
-IC,
0

IC,
0

0
0

-IC,

-IC,
-IC,
-k,

0
IC,

-IC,

IC,

-IC,

k,
-IC,
0
0

0
0
0
0

-IC,
0
0
k,

-IC,
0
k,
0
0

-IC,
0
k,

k,
-kz
0
0
0
0
-k,
k,

0
0
0
0
-k,
4,

Inverse-solution (43) is essentially the Fourier transform of


the product of two functions. By the convolution theorem [9]
this equals the convolution of the Fourier transforms of the
two functions. The convolution is then in direct-space (and
not K-space), and hence is in terms of source functions s1
and s 2 directly. This can be of computational advantage. We
develop such a direct-space solution next.
We note the following Fourier transform pair relations

-IC,
-k,

= [K].

(39)

Ecluation (38) represents an electromagnetic transfer theorem,


thl. direct counterpart to well-known transfer theorems in
optics [7], circuit theory, etc. The LHS matrix [K] is an
eiliht-by-eight transfer function. It has an inverse, as discussed
neut.
Transfer function matrix [K] has elements (39) that are not
derivative operators. Its elements (39) are purely algebraic, i.e.,
pure numbers in specific cases. Therefore, it can potentially
have an inverse in contrast to matrix [MI in direct-space
re-)resentation (13), which consists of differential operators and
hence cannot have an inverse in the ordinary sense. The very
re ison for taking a Fourier approach was, in fact, to achieve
a epresentation with a realizable inverse. The inverse of [K]
is found to exist, and is given by [SI

(444

for a Greens function [8] or point spread function given by

G ( R )=

(45)

Irl

The greek letter 6 denotes Dirac delta function. Next we form


the following partial derivatives of the Greens function from
(444,

\Y

= x , y, z , or T .

(46a:)

iere

K2 = k:

+ IC: + k:

-w2/cz,

(41)

Equation (46a) shows that


transform mate the quantity

(a/au)G ( R ) has

as its Fourier

TI& is easily verified by explicitly evaluating the matrix


pr 3duct of the RHS of (40) with [K]. The result is a diagonal
id1:ntity matrix, as required.
With [KIP known, (38) may be inverted for the fields in
Fc lurier-space
Note that quantity k , in identity (46b) represents the general
element of matrix [K] given by (39). With repeated use of
identity (46b) for each element of [K] we form the following
Fourier transform pair:

b! identity (40). Finally, substituting (42) into (34b) gives the


f i c Ids per se,

(43)

3 i6

IEEE TRANSACTIONS ON EDUCATION , VOL. 3 6 , NO. 4 , NOVEMBER 1993

\!here

It is noted that the extension of the matrix [ M ]into a square


form, described above, is essential to the above electromagnetic derivations. Because of the power of matrix operatiors,
we were able to derive the key effects of electromagnetic
theory without the need for the usual plethora of vector
calculus identities that have become the standard in these
derivations: Stokes theorem, the Divergence theorem, Greens
theorem, the formula for the curl of a curl, etc. Instead,
the simple matrix operations of multiplication and inversion
were used. We find these to be easier to understand and
remember than the preceding vector calculus identifies. Hence,
we believe that electromagnetic theory is easier to learn (a id
is perhaps best taught) from the matrix viewpoint.

QT

0
0
QZ

0
QZ

-QY

0
QT

0
0

QT

QY

QZ

-&Z

0
Qx

QY
-QZ

0
0

.vith elements

d
dU

Qu(R) - G ( R ) .

REFERENCES
(49)

[ l ] L. A. Pipes, Matrix Methods for Engineering. Englewood Cliffs, NI:


Prentice-Hall, 1963.
[2] L. M. Magid, Electromagnetic Fields, Energy, and Waves. New Ycrk:
kcording to plan, we now apply the convolution theorem to
Wiley, 1972.
nversion formula (43). It becomes the convolution of [ s ~ s z ] ~ 131 P. E. Mayes, Second Forum, Electromagnetics and Antennas, a Seiies
of Lectures from the Univ. of Illinois, College of Engin., Urbana, IL,
with the Fourier transform of - [ K ] /n( K 2 ) .By identity (47a)
1967.
he latter is [NI. Hence, the inversion formula becomes
[4] J. A. Eisele and R. M. Mason, Applied Matrix and Tensor Analysis.
New York: Wiley-Intersci., 1970.
[5] R. P. Bocker and B. R. Frieden, A new matrix formulation of classical
electrodynamics: Part I-Vacuum,
NOSC TD 2259, Naval Oc8:an
Systems Center, San Diego, CA, 1992.
[6] P. E. Mayes, IEEE Antennas and Propagat. Soc. Int. Symp. Dig., IEEE
Catalog 85CH2128-7, Library of Congress 84-62888, June 1985, pp.
51-54.
[7] J. W. Goodman, Introduction to Fourier Optics. New York: McGr.iwWe have achieved our goal of solving for the electromagnetic
Hill, 1968.
fields directly in terms of the sources.
[ E ] J. D. Jackson, Classical Electrodynamics. New York: Wiley, 1961.
[9] R. Bracewell, The Fourier Transform and Its Applications. New York:
McGraw-Hill Book, 1965.
VI. SUMMARY AND CONCLUSION

The Maxwell field equations have a natural matrix form (6)


where the matrix is not square. With the intuitive goal of making the matrix square, and continuing the skew-Hermitian trend
of the existing elements of the matrix, two additional columns
of elements are added with compensating zeros in the field
vector (7). Some of the elements remain unfixed. Interestingly,
the particular choices (10) or (11) for the elements lead to
either the Lorentz gauge or the Coulomb gauge-representations
of the fields, respectively.
The tactic of squaring matrix [MI has apparently not been
done before, but has many benefits, as summarized next. Use
of the square matrix operator [MI allows Maxwells equations
to be placed in a compact, covariant form (13). Once in this
form, use of the simple matrix operation of multiplication upon
(13) derives, in turn, the landmark effects of electromagnetic
theory: i) the electromagetic wave and charge continuity
equations (17); ii) the Lorentz gauge/Coulomb gauge definitions (21) of the electromagnetic potentials; iii) the wave
equations (27) for the potentials; and iv) Poyntings theorem
(30) on energy conservation. Alternatively, taking the fourdimensional Fourier transforms of (13) leads to v) a Fourier
representation (38) of Maxwells equations; vi) their inversion
(43) for the fields directly in terms of the sources (but in
Fourier-space); vii) corresponding inversion formula (50) in
direct-space.

Richard P. Bocker received the B.Sc. degree fiom


the University of Utah, in 1965, the M.Sc. dei:ree
from the University of California, Los Angeles, in
1968, and the Ph.D. degree from the Universit:, of
Arizona in 1975.
He has been a research physicist with the Department of Navy for 25 years. His research ir terests are in digital image processing, optical sil:nal
processing and computing, and matrix methods in
electromagnetic wave propagation.

B. Roy Frieden received the Ph.D. degree frorr the


Institute of Optics, University of Rochester, NJ, in
1966.
Since then, he has been a professor at the Optical
Sciences Center, University of Arizona, Tucson.
He has developed and taught courses on image
restoration, information theory, optical propertics of
the atmosphere and oceans, statistical optics, and
estimation methods in optics. His book, Pr&bility, Statistical Optics and Data Testing, 2nd ed.
(Springer Verlag, NY, 1991) is widely used as an
introducion to these subjec&. His research interests are in image fornution
(invented 3-D optical transfer theory), image reconstruction (invented niaximum entropy image restoration), and information physics.
Dr. Frieden is a fellow of the Optical Society of America, SPIE, and the
AAAS.

S-ar putea să vă placă și