621.3.013:621.319.

7 The Institution of Electrical Engineers

Monograph No. 379
May 1960

ELECTRIC AND MAGNETIC IMAGES

By P. HAMMOND, M.A., Member.
{The paper was first received Ylth November, 1959, and in revised form Ylth February, 1960. // was published as an INSTITUTION MONOGRAPH
in May, 1960.)
SUMMARY it is usual to consider the effect of the earth on the field pattern
The method of images as applied to electrostatic, magnetostatic and of an aerial to be the same as that of the image of the aerial.
electromagnetic fields is investigated. By considering the uniqueness Because of the wide field of application and great practical
of the field it is shown within what limits the method can safely be interest of the method of images it would seem reasonable to
used, and rules are given for its use. The application of the method conclude that the method is well understood and can be safely
is illustrated by a discussion of the electric field near a cylindrical applied to new problems. Various factors have, however, con-
cathode and the magnetic fields near the end-windings of electrical
machines. vinced the author that this conclusion is mistaken; among these
might be listed the denial by a writer recently that there is any
image force on electrons around a cylindrical cathode.1 Even
LIST OF PRINCIPAL SYMBOLS more seriously, there was in the author's mind the doubt whether
A, C = Vector potential functions. the method of images is applicable to magnetic boundaries of
B= Magnetic flux density. finite permeability, since the original proof of the image theory
c= Velocity of light. apparently depends on the provision of boundaries at constant
D= Electric flux density. potential. This doubt was reinforced by the observation that
E= Electric field strength. writers on images in electromagnetic theory confined their
H— Magnetic field strength. attention to problems of infinitely conducting boundaries.
/ = Current. The paper is the outcome of a close examination of the method
i= Surface current. of images and has been written with two aims in view: the first
/ = Current density. is to examine the magneticfieldof current distributions near iron
k = cole boundaries with special reference to thefieldsof the end-windings
n= Unit vector normal to a surface. of rotating machines; the second is the more general aim of
P, Q = Vector functions. arriving at a clearer understanding of the method of images in
q, q' = Electric charges. order to determine the conditions under which this method can
R= Radius. be safely applied.
r— Polar co-ordinate.
S= Surface area. (2) HISTORICAL BACKGROUND
v= Volume. The method of images is due to Lord Kelvin,2 who in 1848
x, y, z = Cartesian co-ordinates, published a paper in which he showed that thefieldof an electric
a, j8 = Angles. charge in front of a conducting sheet could be correctly repre-
e= Base of natural logarithms. sented as the field of the charge alone plus the field of its mirror
e0 = Primary electric constant. image in the conducting sheet. Lord Kelvin used the term
e= Permittivity. 'image' in order to point out the close similarity of electrical
with optical images. In optics an image is defined as a point
Ho = Primary magnetic constant. or system of points which, if it existed, would emit the system
fj, = Permeability. of rays which are actually due to reflected light from a mirror
p = Volume density of electric charge. or lens. Two kinds of images are distinguished in optics:
a = Surface density of electric charge or magnetic pole virtual images are located on the far side of the mirror or lens
strength. and real images are located on the near side. Lord Kelvin's
<f>,X,tp = Scalar potential functions. electrical images were all virtual ones.
co = Angular frequency. Lord Kelvin confined his attention to electrostatic problems.
V = Hamilton's vector operator. He was followed by Maxwell,3 who greatly enlarged the
treatment. Maxwell was mainly concerned with problems con-
(1) INTRODUCTION cerning spheres, and he lists all the combinations of spheres
The method of images is of great use to electrical engineers and planes which in the presence of electric charges give a finite
in the determination of electric and magnetic fields in cases number of images. Maxwell, like Lord Kelvin, was chiefly
where there are reflecting boundaries. A typical example is the interested in the image treatment of electrostatic problems.
calculation of the forces on electrons emitted from a cathode: There are, however, in his treatise passing references to the
these forces can be easily calculated by a consideration of the method of images applied to the electric potential of steady
images of the electrons in the cathode surface. In magneto- current conduction, to magnetostatics4 and to the magnetic
images of slowly varying current sheets.5 Searle6 successfully
statics, too, it is customary to calculate the attraction between applied the method of images to magnetic problems of finite
a current and a parallel iron boundary by considering the image permeability, and Hague,7 in an excellent presentation of the
of the current in the iron. Again, in high-frequency problems subject, pays great attention to Searle's method. Both Searle
Correspondence on Monographs is invited for consideration with a view to and Hague, however, present the method of images as a ready-
publication. made solution to certain problems. It is clear that the method
Mr. Hammond is in the Department of Engineering, University of Cambridge.
[3061
 HAMMOND: ELECTRIC AND MAGNETIC IMAGES 307

is correct, but the reader is left wondering how the solution was necessary. All that is necessary is that X shall be constant on
obtained. Other writers follow Maxwell's treatment and restrict S, since we have then
their attention to electrostatic problems involving conducting
boundaries. Among these, Jeans8 bases his proof, like Maxwell,
on the provision of equipotential boundaries,* but then para-
doxically shows that the method is also applicable to problems The criterion of uniqueness is therefore that either X shall be
involving dielectric boundaries, in which there is an electric field constant or dX/d« shall be zero everywhere on S. This means
along the boundary. that either the tangential or the normal components of the elec-
tric field must be specified everywhere on S. A particular case
(3) GENERAL CONSIDERATIONS of great importance arises when S is an equipotential surface,
The basic requirement of the method of images is that the and it is this case which has occupied the attention of writers
effect of the boundaries shall be correctly represented by the on image methods. Almost invariably their discussion starts
images. In electrostatic problems, for example, there will with the consideration of a surface S at earth potential. In
generally be induced charges on the boundaries of the region such a case
considered. The field of the image charges within this region
must be identical with that of the induced charges. We are
thus led to examine how the field within a bounded region of
space depends on the charges at the boundary. When we have and if <f> is zero (or some other known constant value), it is
discovered how the field within a volume depends on the field clear that the field is unique. Such a method starts with the
at the boundaries, we can then postulate an image distribution consideration of charges near perfect conductors, or of currents
of sources which will give the same boundary field and therefore near surfaces of infinite permeability. Of course, such discus-
the same field within the volume considered. This image sion is perfectly valid, but it obscures the issue. The condition
distribution will then correctly simulate the effect of the that S shall be a conducting surface is sufficient to determine
boundaries. the uniqueness of the field within v, but this condition is by no
(3.1) Electrostatic Fields means a necessary one. All that is necessary is that the value
of the tangential or normal electric field shall be specified at
An electrostatic field can be described by a scalar potential cf>.
every point on the surface. Thus there can be fields tangential
In seeking a relationship between surface and volume effects to the surface S. It is therefore not at all surprising that image
we can make use of Gauss's identity methods can be applied to dielectric boundaries. From a
consideration of eqn. (4) we arrive at the following formulation
(1) of the image problem in electrostatics:
Consider a surface S which totally encloses a volume v. Deter-
where P is an arbitrary vector, n is a unit vector drawn from mine the tangential electricfield(or the normalfield)on S due to
within v normal to S and the mathematical surface S completely the induced charges on the boundaries. Find a distribution of
charges outside 5 to give on S the same tangential field (or the
encloses the volume v. If we choose P = «/»V</>, where ift and <f> same normal field). Then this distribution of charges will give
are two scalar functions, then within v the samefieldas is given by the induced charges on the
boundaries. Thus it is the image distribution required.
. . (2) It should be noted that the image distribution is to be placed
outside S. In the language of optics, we require a virtual image.
and this is known as Green's theorem.9 Real images, i.e. charges within v, are inadmissible, because they
do violence to the physical conditions of the problem. The only
sources of an electrostatic field are electric charges and it is
incorrect to insert charges unless these are made inaccessible.
The image charges must always lie outside the region under
consideration, because we require the field of these charges but
We seek for a unique relationship between the surface field and not the charges themselves. Furthermore, although the field of
the volume field. Consider then the possibility of two scalar the image distribution is unique, no such uniqueness is required
potentials <j>\ and cf>2 to satisfy eqn. (3). Then V 2 ^ = — />/eof0, the image distribution itself. It is not uniqueness but con-
where p is the charge density within v and V2<£2 = — p/eo- If venience that is the criterion of the image distribution.
X = <f>x - cf>2, then V2X = 0.
Applying eqn. (3) to the function X gives (3.1.1) An Example: Line Charge Parallel to the Face of a Semi-Infinite
Dielectric Slab.
Fig. 1 shows an infinite line charge q at a distance d from the
(4) face of a semi-infinite dielectric slab of permittivity e. Let the
density of induced charge at P be a; then E normal to the slab
If the left-hand side of this equation is zero, the integrand of at P due to the charge q is given by
the right-hand side must be zero. In such a case VX = 0 and
both X and (j>{ — <f>2 are constants. Thus the field derivable
from 0! is identical with that derivable from <f>2. The field is x2 )
therefore unique. and a is derived from
Reference to eqn. (4) shows that the field will be unique if
?)Xfbn is zero everywhere on the surface. If X is zero on the
surface uniqueness is again achieved, but this condition is not
• He writes in article 209: The principle underlying this method (of images) is that
of finding a system of electric charges such that a certain surface, ultimately to be
made into a conductor, is caused to coincide with the equipotential V «• 0.
308 HAMMOND: ELECTRIC AND MAGNETIC IMAGES
it is convenient to describe the field by a vector potential A,
where B, the magneticfluxdensity, is related to A by B = V x A.
Stratton11 gives a vector form of Green's theorem, which he
obtains as follows:
In Gauss's identity [eqn. (1)] let the vector be P X V x Q.
Then
\\(Px V X Q). ndS = JJJ V . (P x V x Q)dv (6)

Therefore

Fig. 1.—Line charge parallel to the face of a semi-infinite VxP).(V X 0

dielectric slab.
s
- P . ( V x V x Q)dv] (7)
Then the normal field of the distribution a is given by
It will be seen that eqn. (7) is closely analogous to eqn. (2):
qd it is, in fact, Green's theorem in vector form. We proceed as
2e0 x2) in Section 3.1. Let P = Q «= A. Then
Comparing this with the normal field due to q we note that it is
(V x A)2 - A . (V x V x A)]dv
a field which would be produced by a charge of line density fJS {A x V x A). ndS =
—<?[(e — l)/(e 4- 1)] placed at distance d from O within the . . . . (8)
dielectric. This is the required image charge. Let A be the vector potential. Then V x A = B and
(3.1.2) The Physical Meaning of the Boundary Conditions. V x V x ^ = V x B = fiQJ, where J is the electric current
density. To find the criterion for uniqueness we examine the
It is of great interest to examine the physical meaning of the possibility of two solutions, A\ and A , to satisfy eqn. (8).
surface conditions, which are sufficient to guarantee the unique- Then V x V x A = fi J and V x V x 2^ = /x /. Let C =
x 0 2 0<
ness of an electrostatic field. On first looking at the problem A — A and apply eqn. (8) to this difference field. Then
x 2
it would seem essential to specify all three components of electric
field on S.* Green's theorem, however, teaches that it is suffi- V x V x C = 0 and
cient to specify either the normal or the tangential field
components. Thus in the example in Section 3.1.1 it was suffi- (C X V x C) . ndS = jjj (V x C)2dv (9)
cient to consider the normal component of field only.
The physical basis of this is as follows. An arbitrary value If the left-hand side of this equation is zero, the integrand of
of normal field can be introduced at any part of S by placing the right-hand side must be zero. In such a case V x C = 0,
close to S a surface distribution of electric charge. Green's i.e. V x Ax = V x A2 and Bx — B2. The magneticfieldis then
theorem states that it is possible to produce any arbitrary field unique. The left-hand side of eqn. (9) can be written
within a closed volume solely by a layer of surface charge. A
special case is that of zero electric field. It is clear that zero
electric field can always be produced by surface charge on IT (C x V x C).ndS= f I* (V x O x n . CdS
conductors of any arbitrary shape. Green's theorem shows
that not only zero field but any other field can be produced in
this manner. = \\(n x C).(V x QdS
Just as a step in the normal field is due to a charge layer, so
a step in the tangential field can be produced by a layer of Now if n x C is constant on S we have
electric doublets.10 This is of great interest in magnetostatic
problems and is mentioned in Section 3.2.2.
The fact that we can specify either tangential or normal field | J (n x O . (V x QdS = (n x C) J"J (V x C) x ndS
seems to imply a possibility of choice of surface distributions.
This, however, is only a mathematical possibility, since layers
of electric doublets do not occur in nature, f If they did exist, = n x (V x V x C)dv = 0
they would correspond to a surface distribution of steady
magnetic current. Thus the criterion for uniqueness is that either the tangential
components of C shall be constant or the tangential components
(3.2) Magnetostatic Fields of V x C shall be zero everywhere on S. This means that
Magnetostatic fields can be regarded as being due either to either normal B or tangential B must be specified everywhere
a distribution of steady electric currents or to a distribution of on S. The formulation of the image problem in the magneto-
magnetic poles (or dipoles). We shall consider each point of static case is therefore as follows:
view in turn. Consider a surface S which totally encloses a volume v. Deter-
If the sources of the magnetostatic field are electric currents, mine the normal magnetic field (or the tangential magnetic field)
* It should be noted that the surface S considered here is not the actual boundary on S due to the induced currents on the boundaries. Find a
but an adjacent surface just to the right of the boundary. The reason for this is that distribution of current outside S to give on S the same normal or
~b4>l~bn is discontinuous at the boundary and Green's theorem [eqn. (3)] applies only tangential B. Then this distribution of currents will give within v
if i<t>lt>n is continuous throughout v. the same field as is given by the induced currents on the boundaries.
t It should be pointed out that we are here dealing with movable surface distribu-
tions like electric charges and currents which can be induced on a boundary. It is, Thus it is the image distribution required.
of course, true that fixed layers of electric doublets occur very commonly, if not
universally, on the surfaces of solids and are necessary for an explanation of the As in the electrostatic case the images must be virtual, i.e.
existence of the 'potential barrier' which is observed as an electron passes into or
out of a solid. they must lie outside v.
 HAMMOND: ELECTRIC AND MAGNETIC IMAGES 309
The alternative view of the magnetostatic field is to regard its not physically. A better physical description is given by the
sources as magnetic dipoles. All currents must be replaced by surface polarity a, which gives the average magnetic effect of
equivalent magnetic shells. The magnetic field can then be domains which have been rotated by the current /.
derived from a scalar potential and the treatment in Section 3.1 If, on the other hand, all sources of the magnetic field are to
applies throughout. The formulation of the image problem is be poles, / must be replaced by a magnetic shell, and this is a
therefore identical whether one proceeds from consideration of mathematical device without physical content.
currents or dipoles. We are thus led to the conclusion that there are two types of
source—electric currents and magnetic poles. These poles may,
(3.2.1) An Example: Line Current Parallel to the Face of a Semi-Infinite in turn, be due to the orbital and spin motions of electrons. In
Magnetic Slab. this case we should have to talk about two types of current—
Fig. 2 shows an infinite line current / at a distance d from the normal electric current and atomic electric current. But because
face of a semi-infinite magnetic slab of permeability JU,. of the difficulties of a fully relativistic treatment, engineers who
use magnetic materials find the notion of magnetic pole-strength
very convenient.
tORff
Eqns. (3) and (8) now assume a new significance. Tangential B
is a double layer of pole strength and normal B is a single layer.
Thus tangential B represents a current loop and normal B a
magnetized surface. Similarly A is a single layer of pole
strength and B an electric current layer.12 In other words, the
field can be uniquely specified either by electric currents or by
magnets or by a combination of both. This is, of course,
reasonable. We can use air-cored coils, permanent magnets or
iron-cored coils to produce a magnetostatic field.
In the method of images we can therefore use either current
Fig. 2.- -Line current parallel to the face of a semi-infinite images or magnetic images or a combination of both types.
magnetic slab. This greatly enlarges the scope of the method.

First, let all sources of the magnetic field be electric currents. (3.3) Electromagnetic Fields
Let / be the induced current density at P, and let Bt be the Consideration of the uniqueness of electromagnetic fields is
tangential magnetic field due to / at P. Then complicated by the fact that the sources of the electric field are
not only electric charges but also changing magnetic fields. In
. . . (10) other words, the electric field has both divergence and curl, and
^ -*t= -H-5- for its definition requires both a scalar and a vector potential.
At first sight the magnetic field offers a simpler approach,
Therefore because it has no divergence and can still be derived solely from
<n +1> a vector potential. But this potential is now a 'delayed' potential
The tangential field due to / just to the right of the boundary is and the sources of the magnetic field include displacement
current as well as conduction current. Some writers, including
Stratton,13 discuss the uniqueness of the electromagnetic field
by using Poynting's theorem, but this discussion seems to the
author to be inconclusive.15 We shall therefore approach the
Thus the image current is I[(fi — 1 )/(/*, + 1)] at a distance d question in a different manner: we shall consider only steady-
to the left of O. state variation in time, and since any such variation can be
Alternatively, let all sources of the magnetic field be mag- expressed by a Fourier series, we need only consider a single
netic poles. Let a be the induced pole strength at P, and Bn frequency.
be the normal magnetic field due to the magnetic shell bounded In the method of images we are concerned with the field of
at/. the induced currents and charges on the boundaries. Discussion
can therefore be limited to the case where there are no currents
Then (11) or charges within v. Then it is shown by Stratton14 that the
electric field within v at any point x, y, z is given by

/+ = - -1 f [
Thus the image magnetic shell can be represented by a current . . . . (12)
I[Qi - i)/Qi + 1)] as before.
where t = eJkrlr and r is the distance measured from an element
(3.2.2) The Physical Meaning of the Magnetostatic Boundary Conditions. of surface dS to the point of observation x, y, z. The frequency
There are some unsatisfactory features in the discussion of is given by co, and k = co/c where c is the velocity of light.
the example in Section 3.2.1 which throw a great deal of light Now n x B, the tangential components of B, are defined by a
on the physical nature of the magnetostatic field. If we surface electric current. Similarly n x E is a surface 'magnetic
consider that all sources of this field are electric currents, we current' and n . E a surface charge. It therefore appears that
must postulate a surface current i in a magnetic material. we need to specify tangential B and normal and tangential E
Physically, however, no such current will flow. This current / everywhere on S. But tangential B and normal E are related
would have to be greatest at O, but if the magnetic material is by Maxwell's equation V x H = dD/df. In the same manner
iron, having a domain structure, there will be no effect at all the surface current and charge distributions are related by the
at O. Thus i describes the magnetic effect mathematically but equation of continuity of charge. Thus a knowledge of
310 HAMMOND: ELECTRIC AND MAGNETIC IMAGES
tangential E and H on S is sufficient to make the field unique
within v.
The formulation of the image problem is therefore as follows:
Consider a surface S which totally encloses a volume v. Deter-
mine on S the tangential components of the electric and magnetic
field due to the induced currents and charges on the boundaries.1
Find a distribution of current and charge outside S to give on S
the same tangential components of electric and magnetic field. Then
this distribution will give within v the same field as is given by the
induced currents and charges. Thus it is the image distribution
required.
Unfortunately it is in general impossible to follow this pro- Fig. 3.—Aerial near to the face of an infinite plane
cedure. For a simple image we should need both the tangential conducting sheet.
and normal electric field of the image to have a constant ratio
to the tangential and normal electric field of the source. Now Moreover, the tangential magnetic field must obey the
the induced currents and charges will modify the normal electric relationships
field in some definite ratio; but the tangential field will not vary
in the same manner, because it depends, not only on the currents
and charges, but also on the conductivity of the boundaries. (14)
If the boundary could adjust the tangential electric field inde-
pendently of the normal electric field, an image might be
found; but a step in tangential E implies a surface 'magnetic
current' and this does not occur in nature. To adjust the two The normal electricfieldcan be derived from Maxwell's equations
types of electric field we need two types of surface current.
Since only one type occurs in nature, the double adjustment is
impossible. Iz
• • (15)
There is, however, one important case in which the method of *Hx2
images can be applied. If the boundaries have infinite con- ~dt
ductivity, the tangential electric field of the boundary currents
and charges will be equal and opposite to the tangential electric Therefore EyX = Ey2. In terms of charge density a, Ey2 = a/2e 0
field of the source. Furthermore the normal electric fields and and therefore Eyl = a/2e0. Hence the normal electric force
the tangential magnetic fields will be equal and of the same sign.* within the conducting sheet is given by EyX + Ey2 — a/e 0 = 0,
Then an image can be found. It should be noted that the and this is the required value. Thus the specification of
boundaries must be plane in order that there shall be no path tangential H has included the specification of normal E.
difference between the source and the image. In electromag- The image aerial is therefore required to give at the surface S
netic problems the method of images can therefore be used only (just to the right of the conducting sheet) —ExU +EyU —Ezl.
when the boundaries are fiat sheets of very large conductivity.16*17 This is the field of an equal and opposite aerial at a distance d
It should be noted that attention has again been confined to to the left of O.
virtual images in spite of the fact that we are dealing with
electromagnetism and that light is an electromagnetic pheno- (4) APPLICATIONS OF THE METHOD OF IMAGES
menon. The real images of optics are points of intense electric (4.1) Image Forces on a Cylindrical Cathode
field and are caused by changing magnetic fields. But in the
method of electromagnetic images we require an image source Refer to Fig. 4 and consider a line charge of strength +q
of the same characteristics as the actual source. Thus the image parallel to a conducting cylinder. Let the charge be placed at B.
source must have currents and charges and we cannot insert
such a source into the volume under consideration without doing
violence to the physical system. It must be concluded that the
real images of optics are approximations only and do not
accurately reproduce the characteristics of the true source of
light.
(3.3.1) An Example: Aerial near the Face of an Infinite Plane Conducting
Sheed of Zero Resistivity.
In Fig. 3 let the surface current density be / and the surface
charge density be a; let the suffix 1 refer to the field of the
source and the suffix 2 to the field of / and a. Then the tangen- Fig. 4.—Line charge parallel to an infinitely long conducting cylinder.
tial electric field must obey the relationships
If there is an image charge it must, by symmetry, lie on the
p — diameter through B. Assume that its position is at A and its
-*-'.* 1 — (13) strength is —q'. In order that this image may correctly specify
and
the field outside the cylinder, it is required that its normal field
* It might be asked how it comes about that an image which ensures that tangential
E is zero over the boundary also ensures that tangential H is correctly adjusted, i.e. just outside the cylinder shall be equal to the normal field of the
that tangential H is doubled just inside the boundary. The physical reason for this
double adjustment lies in the fact that electromagnetic energy cannot be transferred
charge +q. Consider a typical point P on the cylinder as shown
across a surface of infinite conductivity. Consider this with reference to the Poynting in Fig. 4.
vector E x H. Since there is no energy flow across the boundary, it is necessary
that the surface integral of the Poynting vector of the image must be equal and opposite If we choose Z.OPA = Z.PBA = a, the angles will be as
to that of the source. But tangential E is already equal and opposite, which demands shown in the Figure and A and B will be inverse points with
that tangential H must be equal and of the same sign. Thus the tangential H is
doubled. respect to the circle.
 HAMMOND: ELECTRIC AND MAGNETIC IMAGES 311
The normal force of —q' will be ^— —q. Because of this charge there will be a field outside the
27T6 0 PA cylinder even when the external charge +q is arranged in the
form of a tube around the cylinder. Thus there will be a force
and of +q cos on a distribution of electrons around an earthed cathode, and
PB the paradox is resolved.
If we choose q — q' the difference of the normal forces will be (4.2) Images of the Current Distributions found in Electrical
Machines
q j~cos a cos )8~J q 1 q 1
In electrical machines such as motors and generators it is of
2TT€01 PA PB 27re 0 O P 27T60 i?
great importance to determine the magnetic field of those parts
where i? is the radius of the cylinder. of the winding which are known as the end-connections. Unlike
Thus the charge —q at A is not sufficient by itself to give the the slot portion of the windings, the end-connections do not con-
correct normal field at P. If, however, an additional charge + q tribute to the useful torque or electromotive force. Their
is placed on the axis of the cylinder, the normal field will be importance lies in the fact that they cause eddy-current losses,
correct. Thus the required image distribution consists of a which it is the designer's object to reduce. Moreover, these
charge — q at A and a charge +q at O.* If the single line end-windings are particularly vulnerable to forces which may
charge +q outside the cylinder is replaced by a tube of charge, occur on short-circuit and they have to be carefully braced.
it is clear that the image system will be a tube of opposite sign The end-windings are not embedded in iron, but they are
passing through the inverse points inside the cylinder plus a close to the iron masses of the machine. The method of images
line charge of the same sign at the centre of the cylinder. This offers a useful approach to the problem because it should be
system is shown in Fig. 5. There will then be no electric field possible to estimate the contribution to the magnetic field made
by the iron surfaces in terms of an image of the end-winding.
Because the end-windings are of complicated shape, it is helpful
to examine briefly the effect of small current loops and current
elements before considering a complete winding.
(4.2.1) Images of Current Loops and Current Elements.
\+q Consider first a loop of constant current / opposite an iron
boundary of permeability [x. The images are shown in Fig. 6.
It therefore appears that currents parallel to the iron are reflected
with equal sign and perpendicular currents with opposite sign.

Fig. 5.—Image charges on a long cylindrical cathode surrounded by a

tubular charge distribution.
o
PERMEABILITY p
§ / / / / / / / / / / / / / / / / / /

between the cylinder and the tube outside it. The two images
entirely cancel and it could justly be said that there is no image
at all. There is certainly no image force. If then the outside Fig. 6.—Images of current loops in a plane magnetic surface.
tube represents a uniform cloud of electrons around a cathode,
there will be no image forces on these electrons.
This conclusion appears very paradoxical to engineers familiar
with the behaviour of valves. The paradox is, however,
explained if we consider the potential of the cathode. In our PERMEABILITY y. / / / / / / / / / / /
discussion the cathode has so far been represented by an isolated
uncharged metal cylinder. If now this cylinder is kept at earth
potential by an external connection, the potential of the external
charge plus the image charges must be zero at the surface of the
cylinder. Fig. 7.—Supposed images of current elements in a plane
magnetic surface.
Consider then the sum of the potentials of a charge +q at
B and its image charge — q at A. At any point P on the surface It would, in fact, be tempting to formulate a rule for current
of the cylinder the potential is elements as shown in Fig. 7.
<7 , PB OB OB A rule like this is given by Schelkunoff18 for images in con-
=
— l08
AP = ducting sheets. Consideration will show, however, that Fig. 7
is not justified. Consider the vertical current element: if we
If B is very near the surface of the cylinder OB ~ OC and <f> use Biot and Savart's expression for the magnetic field of a
tends to zero. It follows that, if the potential of the cylinder is current element,
to be kept constant, it will be necessary to remove the second
image charge -\-q at O. The cylinder then has a net charge of
* It should be noted that the matching of the normal field automatically ensures
that the charge is conserved, i.e. that there is no net charge within the cylinder. If we note that the vertical current element has no component of
we had made use of the tangential force, the additional charge on the axis of the
cylinder would have had to be inserted by invoking the conservation of charge magnetic force perpendicular to the surface of the iron. Hence
explicitly. In dealing with a field of divergence sources it is thus always safer to no polarity is induced and the field outside the iron is not
match the normal field components. Similarly, in dealing with a field of curl sources
it is safer to match the tangential field components. affected by the presence of the iron. Thus it could justly be
312 HAMMOND: ELECTRIC AND MAGNETIC IMAGES
said that the vertical current element has no image; but this is loops embedded in the iron. It is seen that the iron has the
not conclusive, because the Biot-Savart law is not unique19 effect of adding a current loop of strength I[(fx — l)/(/x + 1)]
unless it is applied to a Heaviside rational current element.20 in the same place as the current loop /. The effect of current
Moreover, the Heaviside element requires to be immersed in and iron combined is therefore the same as if the iron were
an infinite conducting medium, and the iron surface would removed and the current strength were increased to 2;U.//(/A + 1).
interfere with this medium. The plain fact is that a current For large values of p this means doubling the current.
element is not a physical entity and no amount of theory can
make it so. It is impossible to apply the method of images to
isolated current elements. -2q
Although steady current elements are a physical impossibility,
-2q +2q
it is possible to have electric doublets, i.e. alternating current + 2q
elements with charges at their ends. The images of these in iron . . • / » > ,

are not easy to determine, because currents will flow in the iron 'INFINITE PERMEABILITY/
'////////////'//////A
by virtue of the alternating fields. Let us approach this problem
by a consideration of images in a sheet of infinite conductivity. Fig. 11.—Images of electric doublets embedded in an infinitely
Fig. 8 shows the images of electric doublets and of small loops permeable material.
of alternating current. Comparison of Fig. 6 with Fig. 8 reveals
the fact that the image of an alternating current loop in a con-
Fig. 11 shows image effects outside a region of infinite

r - q +q
o
permeability. It will be seen that the strength of the electric
doublets has been increased twofold.

(4.2.2) The Method of Images applied to the End-Windings of Electrical

o
• -q Machines.
INFINITE ELECTRIC CONDUCTIVITY
The discussion in the previous Section shows that a distinction
tq -q must be made between direct and alternating currents. With
alternating currents the conductors can be broken up into electric
doublets terminated by finite charges. Direct-current elements
Fig. 8. Images of electric doublets and current loops in a plane
perfect electric conductor. are a physical impossibility, but they can be regarded as the
limiting case of an electric doublet as the frequency tends to zero.
A further distinction must be made between images in magnetic
surfaces and conducting surfaces. The two types of image are
of opposite sign, as shown in Figs. 8 and 9. In treating the

i: -q
field of a machine end-winding it has therefore to be decided
whether the eddy-currents in the core end-plates make the surface
of these plates into a conducting sheet, or whether they are

r
INFINITE MAGNETIC CONDUCTIVITY
sufficiently inhibited to allow the end-plate to act as a magnetic
sheet.
-q
Coils of any arbitrary shape can be treated as a succession of
electric doublets, and the method is therefore completely general.
Fig. 9.—Images of electric doublets in a plane perfect Fig. 12 shows the image of a typical coil-end in a perfectly con-
magnetic conductor.
ducting sheet; Fig. 13 shows the image effect outside the iron
of a coil emerging from the iron; and Fig. 14 shows the image
ducting surface is opposite in sign to the image of a steady effect inside the iron for the same coil.
current loop in a magnetic surface. Fig. 9 shows the images of In considering the field outside the iron it is likely that Fig. 13
electric doublets in a perfect magnetic conductor (if such a
substance existed). If will be noted that the images are of
opposite sign to the images in a perfect electrical conductor.
The images of the current loops, current elements and electric
doublets described in Figs. 6, 7 and 9 arise when the current is
outside the magnetic material. In electrical machines, however, ^INFINITELY CONDUCTING
parts of the conductors are embedded in iron. We must there-
fore consider the image distribution of currents inside a magnetic
material. This image distribution is easily derived, by the Fig. 12.—Images of a coil-end near an infinitely conducting sheet.
method of this paper, for the fields both within and outside the
iron.
Fig. 10 shows the image effect outside the iron of current

//KtHMtABILITY u,////.
o.7 + 7 A
,7+7 >-

Fig. 13.—Image effect in air of a coil-end emerging from a block of

Fig. 10.—Images of current loops embedded in a magnetic material. magnetic material.
 HAMMOND: ELECTRIC AND MAGNETIC IMAGES 313
(6) REFERENCES
(1) MOULLIN, E. B.: 'On the Amplification Factor of the
Triode', Proceedings I.E.E., Monograph No. 211 R,
November, 1956 (104 C, p. 222).
(2) LORD KELVIN: 'Reprint of Papers on Electrostatics and
Magnetism' (Macmillan, 1872), pp. 52 and 144.
(3) MAXWELL, J. C : 'Electricity and Magnetism' (Clarendon
Press, 1892), Chapter XI.
(4) Ibid., Article 318.
(5) Ibid., Article 662.
(6) SEARLE, G. F. C.: 'On the Magnetic Field due to a Current
Fig. 14.—Image effect in magnetic material of a coil-end emerging in a Wire Placed Parallel to the Axis of a Cylinder of
from a block of magnetic material. Iron', The Electrician, 28th January, 1898, pp. 453
and 510.
is nearer the truth than Fig. 12. Nevertheless, the eddy-currents (7) HAGUE, B.: 'Electromagnetic Problems in Electrical Engi-
will reduce the image effect shown in Fig. 13. Recourse must neering' (Oxford University Press, 1929), Chapter IV.
be had to measurement rather than calculation, but the method (8) JEANS, SIR JAMES: 'Electricity and Magnetism' (Cambridge
of images has pointed the way to fruitful experimental University Press, 1925), Articles 208-225.
investigation. (9) GREEN, G.: 'An Essay on the Application of Mathematical
(5) CONCLUSIONS Analysis to the Theories of Electricity and Magnetism'
The method of images has a very wide field of application. (Printed privately at Nottingham, 1828), p. 11.
In electrostatics and magnetostatics the method is not confined (10) ABRAHAM, M., and BECKER, R.: 'Electricity and Magnetism'
to problems involving equipotential boundaries. If the problem (Blackie, 1950), Chapter II, Section 8.
is expressed in terms of a scalar potential, an image source is (11) STRATTON, J. A.: 'Electromagnetic Theory' (McGraw-Hill,
adequately specified if the normal component of the field is 1941), p. 250.
specified everywhere on a closed surface. If the problem is (12) Ibid., pp. 247 and 243.
expressed in terms of a vector potential, it is necessary and (13) Ibid., pp. 486-488.
sufficient to specify the tangential components of the field every- (14) Ibid., pp. 464- 468.
where on a closed surface. (15) HAMMOND, P.: 'Electromagnetic Energy Transfer', Proceed-
In electromagnetic problems the tangential components of ings I.E.E., Monograph No. 286, February, 1958 (105 C,
both the electric and magnetic field have to be specified on a p. 352).
closed surface. In general this is impossible, and the image (16) MOULLIN, E. B.: 'Radio Aerials' (Clarendon Press, 1949),
method must be confined to problems involving infinitely con- p. 121.
ducting boundaries. (17) AHARONI, J.: 'Antennae' (Clarendon Press, 1946), pp. 212
All electric and magnetic images are virtual, and the analogy and 216.
with optical images must not be pressed too far. (18) SCHELKUNOFF, S. A.: 'Electromagnetic Waves' (Van
The application of image methods to problems involving low- Nostrand, 1943), p. 171.
frequency currents near iron boundaries is difficult. For certain (19) WHITTAKER, SIR EDMUND: 'A History of the Theories of
frequencies there are likely to be very small image effects. Aether and Electricity' (Nelson, 1951), Vol. 1. p. 82.
Experimental work is required to investigate this tentative (20) HEAVISIDE, O.: 'Electrical Papers' (Macmillan, 1892),
conclusion. Vol. 2, Article 48.