Proceedings of the Physical Society.

Section A

Related content
- A Scalar Representation of
A Scalar Representation of Electromagnetic Fields Electromagnetic Fields: II
E Wolf

To cite this article: H S Green and E Wolf 1953 Proc. Phys. Soc. A 66 1129 - Diffraction Theory
C J Bouwkamp

- The role of the Riemann–Silberstein vector

in classical and quantum theories of
electromagnetism
View the article online for updates and enhancements. Iwo Bialynicki-Birula and Zofia Bialynicka-
Birula

Recent citations
- Reactive energy in nondiffracting localized
waves
Peeter Saari and Ioannis Besieris

- X-ray Fokker–Planck equation for paraxial

imaging
David M. Paganin and Kaye S. Morgan

- Kurt E. Oughstun

This content was downloaded from IP address 132.203.227.62 on 09/02/2021 at 18:32

 1129

A Scalar Representation of Electromagnetic Fields

BY H. S. GREEN? AND E. WOLF

Department of Mathematical Physics, University of Edinburgh

Comnzriiricnted by M m Born; M S . received 16th Julte 1953, and in amendcd form

31st August 1953

Abstract. I t is shown that in a region which is free of currents and charges, any
electromagnetic field may be rigorously derived from a single, generally complex,
scalar wave function V ( X ,t ) . I n terms of this function the momentum density
g(x, t ) and the energy density w ( x , t ) of the field may be defined i n such a way
that they are represented by expressions analogous to the formulae for the
probability current and the probability density in quantum mechanics ; in a
homogeneous isotropic medium
1 .
g(x, t ) = - -[V*VV+V V Y * ] ,
8WOC

CL0 1.
The densities defined in this way differ from those given by the usual expressions,
but the choice is justified since the differences disappear on integration over
any arbitrary macroscopic domain. (The corresponding Lagrangian densities
differ by a four divergence.) When V is of the form Vo(x)e-iwt,g is found to form
a solenoidal field which is orthogonal to the co-phasal surfaces arg V , =constant.

$1. INTRODUCTION

I
N a region of space which is free of charges and currents an electromagnetic
field is fully specified by the magnetic vector potential A. From it the
electric and the magnetic field vectors may be derived by means of well-
known relations.
In a wide class of problems, particularly in those encountered in optics, the
actual behaviour of the field vectors is of little interest. What one primarily
wishes to know is the average energy or the average flux, and one is led to wonder
whether for such purposes the derivation from a vector potential is really the
most suitable one. Except in the so-called ‘rigorous’ diffraction theory, one
does in fact often employ in optical considerations of energy a single, generally
complex, scalar wave function (usually called the disturbance, or the complex
amplitude), whose squared modulus is taken as the measure of the light intensity.
This simple procedure has been employed in optics since about the time of
Fresnel (the disturbance then being considered to be the displacement of a
Particle in an ‘ elastic ether ’) and has been the subject of much criticism, in spite
of the fact that under fairly general conditions it gives results which are found
t o be in excellent agreement with experiment. T h e velidity of the scalar
1. O n leave ofabsence from the University of Adelaide, Adelaide, South Australia.
PXOC. PHYS. SOC. LXVI, 12-A 75
1130 H . S. Green and E. Wolf

theory has been justified, at least as an approximation or as a time average, for

a number of special cases only: ( a ) for certain two-dimensional problems and
for monochromatic fields with rotational symmetry (cf. Braunbek 1951), (b) in
applications to diffraction problems encountered in optical instruments of usual
design and working with non-polarized and almost monochromatic light (Picht
1931, Luneberg 1947-8, Theimer, Wassermann and Wolf 1952).
Now it is well known that the energy density and the momentum density
(and consequently the light intensity) are not uniquely defined by the electric
and the magnetic field vectors. One may always add to the Lagrangian density
a four divergence which gives no contribution to the field equations, though
it may alter the local (unobservable) values of the densities of the energy and
the momentum. Such alteration must of course lead to no observable changes
in the total amount of .energy and momentum contained in any extended
(macroscopic) domain. T h e question is therefore open as to whether one
could not define the energy and momentum densities in such a way that they
would always be expressible in a simple manner in terms of a single complex
scalar wave function, the field vectors remaining unchanged. In the present
paper it is shown that this in fact is possible.
We find that in regions where no currents or charges are present the
magnetic potential may be rigorously derived from a single, generally complex,
scalar wave function V(x, t ) , which we call the complex potential. I n terms of
this function the momentum density g ( x , t ) and the energy density eo(x,t) in
a homogeneous isotropic medium may be defined by means of formulae analogous
to the expressions for the probability current and the probability density in
quantum mechanics :
1 .
g(x, t ) = - -[V*VV+ VVV"],
8WOC
I
I
and k . .. . . .(l.l)
1
W ( X , t ) = &[2vv*+-vv.vv*
PO
these quantities satisfy a conservation law of the usual form :
I; J
.(1.2)
The corresponding Lagrangian density L is
1
~VV*--vv.vv* . ... ..(1.3)
PO
here eo, po are the dielectric constant and the magnetic permeability and c is the
vacuum velocity of light.
An interesting consequence of the present analysis is the result that in the
case when the time enters V only through the factor e-iwt, i.e. when V is of the
form
' real) ......( 1.4)
V ( x , t ) = u ( x ) e x p ( i [ k c Y ( x ) - u t ] ) (U and d
(w =frequency, k = wlc), g and w become independent of time and (1,l) reduces
to
g(x)= &
."
v2VcY. ... . . . ( 1.5)
.The conservation law now becomes
v .g(x)= 0 . . .. . . . ( 1 4
 A Scalar Representation of Electromagnetic Fields 1131

Equation (1.5) shows that the energy flow is orthogonal to the surfaces of constant
phase of the complex potential and (1.6) expresses the fact that the vector field g
is solenoidal. One is thus led rigorously to the concept of ‘ clectromagnetic rays’,
and one obtains for a class of electromagnetic fields a simple model which may
be regarded as a natural generalization of geometrical optics.

3 2. DEFINITION OF THE COMPLEX POTENTIAL

We consider an electromagnetic field in a homogeneous, isotropic medium.
In regions free of currents and charges the field quantities may be derived from
a single vector potential a(x,t ) which satisfies the homogeneous weve equation
and also the divergence condition
V.El=O. ...... (2.1)
In general, currents and charges will, of course, be present in some parts
of the x space. It will be convenient to imagine these enclosed in boxes and
to consider in place of the function A defined by
A(x, t ) =a(x, t ) outside and on the boxes
=O inside the boxes.
A is formally defined over the whole x space and may be represented by a
Fourier integral
. + .
A(x, t ) = [a(k, t ) cos (k x) b(k, t ) sin (k x)] dk, . . . . . .(2.2)
where, since A is real, the integration is carried over half of k space (e.g. K, 30).
On account of (2.1) the (real) vectors a(k, t ) and b(k, t ) are orthogond to k,
a . k = b. k =O. ...... (2.3)
With each k we associate two unit vectors l,(k) and l,(k) such that l,,1, and k
form a right-handed orthogonal triad. This may be done for example by
choosing a constant vector n and taking as 1, a unit vector perpendicular to the
(k, n) plane and as 1, a unit vector perpendicular to k and 1, :
nhk
11( k) = -nhk 1
Inhk(’
Inhkl’
khl ksn-(n.k)k ,
..... .(2.4)
l,(k) = 2=
Ikhll) IK2n-(n.k)kJ )
Equations (2.3) show that a and b lie in the plane of 1, and 1, and may therefore
be resolved along these vectors :
a, =a. l,,
b, = b . I,,
Next we form the complex combinations
1
a,=a. I,,
b, = b. 1,.
...... (2.5)

1
a(k, t ) =a,(k, t )+ia& t ) ,
......(2.6)
1
S( k, t ) = h(k, t ) + %(k, t ) ,
and regard cc and ,8 as Fourier coefficients of a function V :
. .
V ( x ,t ) = [a(k, t ) cos (k x) + /3( k, t ) sin (k x)] dk. ...... (2.7)
We call V ( x ,t ) the complex potential of the field.
We have now replaced the magnetic vector potentia1 by a complex scalar
function. Conversely it is easily seen that from the complex potential V , the
75-2
 1132 H . S. Green and E. Wolf

magnetic vector potential A and consequently the field vectors may be uniquely
derived. For in the first place, from the knowledge of V the quantities cc and p
may be obtained by applying the Fourier inversion formula. Then U,, u2,b, and
b, may be determined from (2.6). From them the Fourier components a and b
of the complex potential are obtained :
a =all, + a212, b = b,l, + b,12. ......(2.8)
. .
Finally, if one forms the combinations a COS (k X) + b sin (k X) and integrates
over all k, one obtains the vector potential A. Hence the scalar V ( x , t )
completely specifies the field.-*
I t is easily seen that in a homogeneous isotropic medium of dielectric
constant eo and magnetic permeability po the complex potential satisfies the
wave equation
v2v- €323p=o*
C2
......(2.9)
For, since the vector potential A satisfies the wave equation

(2.10)
one must have, for each component,
A,=a(k,t)cos(k.x)+b(k, t)sin(k.x) ......(2.11)
of A,
(2.12)

It then follows from (2.12), (2.5) and (2.6) that the corresponding terms V , of
V , i.e. . .
V k =cc(k, t ) cos (k x) + P(k, t ) sin (k x), ...... (2.13)
must satisfy the scalar wave equation
...... (2.14)
Consequently the complex potential satisfies the homogeneous wave equation (2.9).

8 3. THEMOMENTUM DENSITY AND THE ENERGY DENSITY

3.1. We now derive expressions for the momentum density and the energy
density of the field in terms of the complex potential. We restrict our discussion
t o a homogeneous isotropic medium.
I n terms of the vector potential, the electric and the magnetic field vectors
are given by 1.
E = -.-A, ~,H=VAA, ......(3.1)
C

and the total momentum G of the field is

1 1 ..
G(t)=G j ( E n H ) d x = - 4 X C J
A A ( V A A ) ~ X ., . . . . .(3.2)

the integration being taken throughout the x space.

* Note a d h d in p o u f . Since this paper was written our attention has been drawn to a
paper by E. T. Whittaker (Proc. Land. Muih. Soc., 1904, 1, 367), where it is shown that
in Z'UCZ(O at points not occupied by electrons a field produced by any number of electrons
moving in any arbitrary manner may be derived from two real scalar wave functions.
T h e connection between Whittaker's analysis and that of the present section is being
investigated and it is hoped to publish the results in a later communication.
 A Scalar Representation of Electromagnetic Fields 1133
T o express G in terms of V we first substitute into (3.2) from (2.2). We have
j”
1
A = [acos (k. x ) + b sin (k. x ) ]dk,

.
k A [ - a sin (k x ) + b cos (k .x ) ]dk. 1
J
......(3.3)

Hence G = - ___ j”dxjj[a(k‘, t ) os(k’.x)+b(k’,t)sin(k’.x)]

+POC
h[k A (- a(k, t ) sin(k x) .
+ b(k, t )cos (k. x ) ] dkdk’. ...... (3.4)
This becomes, with the help of the Fourier integral theorem,

G=--
72

2POC
1 a(k, t ) A [k A b(k, t ) ] - b(k, t ) A [k ~ a ( kt ), ]dk, ...... (3.5)
or, using ( 2 4 ,
[(a,bl+a,b2)-(b,u,+b,u2)]kdk

5j[(.Xj+xj*)-(k*p+$*)]kdk
= 2POC

=- - j (1( i . * v V + ~ v v * ) d x . ..... .(3.6)

8vLL,c
Hence the density g of the momentum may be defined as

g(x, t ) = -
8VOC
-[ v * v V + VVV*]. ......
The electric and the magnetic energy densities may also easily be defined
in terms of the complex potential. T h e total electric energy W , is

J E2dx
W e ( t )= 5-
877

= 2
87c2
J’ dx 11 .
(h(k, t ) cos (k x ) + b(k, t ) sin (k. x))

.
. {u(k‘, t ) cos (k’ .X) + h(k‘, t ) sin (k’ x ) )dk dk’

2 ~ J’2(ala+ az2+ b12+ bZ2)d k

X2E0
=

=
7 A 0
-
2c2
.)_(a*,*
+ ,&a*)dk
= *f(VV*)dx.
87c2
..... . ( 3 . 8 )
The total magnetic energy W,,,is
W,ll(t)= 21 H2dx,

and one finds, by a similar,calculation as above,

.. 1 . . (3.9)
 1134 H . S. Green and E. Wolf

Equations (3.8) and (3.9) show that one may define the electric and magnetic
densities w, and w,,,by the expressions
E ' . 1
W,(X,t ) = d2 VV*, w,,,(x,t ) = -( V V .VV*). ......(3.10)
87rc 8 W O
I t is easily verified, with the help of the wave equation (2.9), that the total
energy density w = w, + w,,, and the energy flux S = cg satisfy the conservation law
dW
-+v.s=o.
dt
.... . .(3.11)
3.2. V is in general a complex function. Let v denote its amplitude and
4 its phase : V(x,t) = ~ ( xt),exp {i+(x, t ) } . ......(3.12)
If one substitutes from (3.12) into the wave equation (2.9) and separates real
and imaginary parts, one obtains the following two equations :
V2li'-li'(V4)2- 3E..(dj-vd;2)=o,
C2
......(3.13)
. . .
2(Vv). (V+) + vV24 - wo
C2
(2U+ +v+) = 0. ...... (3.14)
I n terms of li' and 4 the momentum density g and the energy densities w,,and
1
g(x, t ) = - -(UVV + V"V+),

w,(x, t ) = &2
4 T O C

(62 + &2),
1 ......(3.15)
1
W,,l(X, t ) = -((vv)2 + v2(v+)z}. ~

8 W O J
Of particular interest is the case when the time enters only through a factor?
e- iwt . Then v is independent of time and is of the form +
4(~, t ) = k@(x) - ut ......(3.16)
( k = w / c ) . ' The eqns. (3.13) and (3.14) reduce to
1
- k2v
(VCP')~ -V2v = n2, ...... (3.17)
v v . v30 + +vv2CV = 0, ...... (3.18)
where n2 = eopo. ......(3.19)
The expressions (3.7) and (3.10) for the momentum densities and the energy
densities become c k2
g(x)= L2
4xn
v2 VCV , .... ..(3.20)
....(3.21)

t
8nn
(VCq2 + 1
(V log v)2

Note added in proof: A general monochromatic wave must be represented by a

1. ......(3.22)

Z' function of the form

V ( x ,t ) = Vl(x)e-iwt+ Ve(x)eimt,
where VI and V, are complex functions of positrons. It appears that the case here
considered (Vz=O) represents a monochromatic wave of arbitrary shape but circularly
polarized, at least in the sense of a space average over a macroscopic domain.
 A Scalar Representation of Electromagnetic Fields 1135
Equation (3.20) shows that the energy flow is orthogonal to the surfaces
@(x) =constant. ......(3.23)
Moreover, (3.18) may be written in the form
v. (V2VH) = 0 ; ...... (3.24)
this relation (which also follows from the conservation law (3.11)) shows that the
vector field g is solenoidal.
The two results which we have just established show a close analogy with
the model presented by geometrical optics. The surfaces 30 =constant may be
regarded as generalized wave fronts and the energy may be considered as being
propagated along the curves (rays) orthogonal to these surfaces.
To determine the wave fronts =constant one must in general solve two
equations (3.17) and (3.18). But we note that the second term in (3.17) contains
as multiplicative factor the second power of the small quantity 1/k. Except in
special regions this term will be very small in comparison with n2 = eopoand
may therefore be neglected. This implies that the geometrical wave fronts
given by the solution of the eikonal equation
(V30)z = n2 ......(3.25)
are in general a very good approximation to the waves associated with the complex
potential.? Better approximations may be obtained by the application of the
W.K.B. method, or with the help of Huygens' principle (or Kirchhoff's integral
formula).
One can also easily write down an expression for the variation of the amplitude
along each ' electromagnetic ray' (orthogonal trajectory to the 30 surfaces). For
if 8/87 denotes differentiation along a particular ray, a/& = V& . V, and (3.18)
gives an/& + +(V2N)u= 0 ; hence
a(.) =~ exp
( 7 ~ ) ......(3.26)
$4. A COMPARISON WITH THE CLASSICAL THEORY
The energy and the momentum of the electromagnetic field may be derived
by a variational principle from a Lagrangian density L' which is Lorentz
invariant :
......(4.1)
The variational principle leads to the vector wave equation (2.10) for the vector
potential A.
I n the present theory we have, in place of (2.10), the scalar wave equation (2.9)
for the complex potential V. This equation may also be derived from the
variational principle by replacing L' by the Lagrangian density$
L= 1 -(" vv*- -vv.vv*}.
1 ....., (4.2)
.
877 c2 Po
7 In this connection mention must be made of recent researches of Luneberg (cf. 1949,
particularly pp. 55- 56), which have shown that for small wavelengths geometrical optics,
in its extended form, including also description of vectorial properties of the field, gives
a very good first approximation to a steady state field.
$ It is of interest to note the analogy with the Lagrangian density for a scalar meson
with zero rest mass.
 1136 H. S. Green and E. Wolf

Since L' and L lead to the same field equations, their difference must be
the form
L - L' = P-k d i v a . ......(4.3)
I n consequence, the momentum density g and the energy density w of the present
theory will differ from the usual expressions g' and w' associated with (4.1). In
fact one has relations of the type
1

g-g'=gradP, w-w'= - d i v a , .. . . . a (4.4)

where P = P and Q = Q if P and Q do not involve the field variables. The effect
of these differences will be negligible on integration over an arbitrary domain
which is large compared with the wavelength.
That the usual definition of energy flow, based on the Poynting vector, is not
the only possible one is of course well known.* Our present definition has
several advantages. We saw, for example, that in a field represented by a complex
potential of the form V(x, t ) = V,,(x)e-iot . . . . . .(4.5)
g, w, and w,,,are independent of time and g is orthogonal to the surfaces of
constant phase of the complex potential. But even in more complicated fields
the densities as defined by the present theory seem to lead to simpler results,
T o illustrate this point we consider an example (essentially due to Braunbek
1951) :
Suppose that the vector potential of a field in vacuo (c0 =po = 1) is
+ .
A(x, t ) = C{n, cos [k(n, .x - c t ) ] n3 cos [k(n, x - c t ) ] , . . . . . . (4.6)
where C is a constant and n,, n,, n3 is a right-handed orthogonal triad of unit
vectors. A straightforward calculation gives for the time average
c(EAH)
(S')= ( F )
of the Poynting- vector:
ck2C
(S')=
-[(n,+n,)-n,cos[K(n,-n,)
857
.XI. . . . . . .(4.7)
This expression does not satisfy the relation (S' ) . curl (S') = 0 which would
obtain if (S') possessed orthogonal trajectories (cf. Weatherburn 1930, p. 217).
On the other hand one easily finds that the average energy flux S = cg as defined
by (3.7) possess orthogonal trajectories. For one has in this case (taking
n=n,+n,)
V ( X t, ) =C
(:2
From (4.8) and (3.7) one obtains, by a straightforward calculation,
I
-(1 - i ) cos [K(n,. x-cct] +icos[k(n, .x-ct)] . . . . . . . (4.8)

ck2C
c 1
( S ) = ( c g ) = -[n,+n,] 1- -cos(k(n,-n,).x)]
877 d2
, . . . . ..(4.9)
and it follows from (4.9) that (S) satisfied the required condition for the existence
of orthogonal trajectories : (s). curl (s) =0.
It is seen that (S') and (S) differ by a space-periodic function, to which clearly
no physical meaning can be attached. For div {(S') - (S)} = 0, and consequently
the (time-averaged) energy which crosses any closed surface in the field, will be
the same whether S or S' is taken to define the energy flow.
* cf. Stratton 1941, p. 134; also Hines 1952.
 A Scalar Representation of Electromagnetic Fields 1137

Finally we give a summary of our main formula and the corresponding

formulae of the classical theory :
Usual definition Present definition
Basic field quantity A V
(E=- 1 ~. , p o H = v t , A
c

Lagrangian density

Momentum density

Electric energy density

I

Magnetic energy density -vv.vv*

8TO

ACKNOWLEDGMENT
In conclusion we wish to thank Professor Max Born for his interest in this
work and for stimulating discussions.

REFERENCES
BRAUXBEG, W., 1951, Z . Naturforsch., Cia, 12.
r.
HINES,C. O., 1952, Canad. Phys., 30, 123.
LUKEBERG, R. K., 1947-8, Propagation of Electromagnetic Waves (mimeographed notes
of lectures delivered at New York University] ; 1949, Asymptotic Development of
Steady State Electromagnetic Fields, New York Univwsity, Mathematics Research
Gvoup, Report No. EM-14.
PICHT,J., 1931, Optische Abbildung (Braunschweig : Vieweg).
STRATTON, J. A., 1941, Electromagnetic Theory (New York : McGraw-Hill).
THEINER, O., WASSERMANN, G. D., and WOLF, E., 1952, Proc. Roy. Soc. A, 212, 426.
WEATHERBURN, C. E., 1930, Dazwential Geometry of Three Dimensions, Vol. 11 (Cam-
bridge : University Press).