IOP PUBLISHING Eur. J. Phys.

30 (2009) 965972

EUROPEAN JOURNAL OF PHYSICS

doi:10.1088/0143-0807/30/5/005

Subtleties in energy calculations in the image method

Instituto de Fsica, UFRJ, CP 68528, Rio de Janeiro 21945-970, Brazil Escola de Cencias e Tecnologia, Universidade Federal do Rio Grande de Norte, Natal, Brazil

E-mail: marciotaddei@if.ufrj.br, tarciromendes@ect.ufrn.br and farina@if.ufrj.br

Received 26 March 2009, in nal form 10 June 2009 Published 8 July 2009 Online at stacks.iop.org/EJP/30/965
Abstract

In this pedagogical work, we point out a subtle mistake that can be made by undergraduate or graduate students in the computation of the electrostatic energy of a system containing charges and perfect conductors if they naively use the image method. Specically, we show that naive expressions for the electrostatic energy for these systems obtained directly from the image method are wrong by a factor of 1/2. We start our discussion with well-known examples, namely point chargeperfectly conducting wall and point charge perfectly conducting sphere, and then proceed to the demonstration of general results, valid for conductors of arbitrary shapes.

1. Introduction

The typical problem in electrostatics consists of determining in all relevant space the electrostatic eld generated by some set of charges from the known charge distribution itself as well as from the appropriate boundary conditions pertaining to the situation. Solving this problem amounts to nding a static potential (x) obeying Poissons equation, namely 2 (x) = (x)
0

(1)

subject to suitable boundary conditions. Although one can approach this task in many different wayswhose convenience depends on the particular problem being dealt withPoissons equation (given denite boundary conditions) has a unique solution for each charge distribution (x). This allows one to look for solutions in any desired fashion: if one nds a potential that obeys both the boundary conditions and Poissons equation for the correct (x), it must be the correct potential for the given conguration. A especially suitable method for simple situations with point charges and dipoles in the presence of conductors is called the image method. It consists of nding a different conguration in which the conductors are replaced by some charge distribution so that the
0143-0807/09/050965+08$30.00 c 2009 IOP Publishing Ltd Printed in the UK 965

966

M M Taddei et al

Figure 1. Real and image charges for the charge-wall case.

potential of the entire set-up in the region of physical interest (i.e. outside the conductors) ts the appropriate boundary conditions and also obeys Poissons equation. These ctitious charges put in place of the conductors are called image charges. Since any potential created by charges obeys Poissons equation (equation (1)), once the boundary conditions are satised, the eld created by all charges (real and image) is the eld obtained in the actual conguration outside the conductors. It is important to state that in the ctitious conguration the charges outside the conductors must be in the same place as in the real distribution, or else one would nd a solution to Poissons equation with a different (x): all image charges must be placed in the space originally occupied by the conductors. The method can then provide the force acting on a charge in the presence of conductors. However, if one tries to naively use the image method to compute the electrostatic energy of that conguration, an incorrect result will be found, namely one arrives at twice the correct energy, as Grifths [1] and Franklin [2] have shown for particular cases. In simple situations such as the point chargeplane wall set-up, it is rather easy to realize that the real conguration has half the energy of the ctitious one, but in more involved and less symmetrical geometries this is no longer obvious. In the case of a charge in front of a sphere, one could lucidly expect to nd a prefactor depending on the sphere radius which would only tend to 1/2 if the radius tended to innity (reobtaining the chargewall result). Our goal is to show that, whatever the shape and number of perfectly conducting bodies near a point charge, the electrostatic energy of the system is half the Coulombic energy of the interaction between the real charge and each image charge of the problem. We shall then generalize this result for more than one source charge. Our paper is organized as follows. We begin in section 2 with the familiar problem of a charge and a conducting wall, and also comment on the case of a wedge. We then proceed, in section 3, to a less symmetrical geometry, that of a conducting sphere. Section 4 is dedicated to the theorem demonstration in the general case, and we leave the last section for conclusion and remarks.
2. Usual case: point charge and conducting wall

In order to state the problem and emphasize the important point in the clearest way, we start our discussion by considering in this section the simplest problem that can be solved by the image method, namely nding the electrostatic eld of a point charge in the presence of a perfectly conducting wall. This problem can be found in many standard textbooks [14], so we go directly to the point here. For convenience, let us choose our Cartesian axes OX YZ such that the region z < 0 is lled with a perfectly conducting material and a point charge q is at position (0, 0, z), as shown in gure 1. According to the image method, the force exerted on the real charge q by the charge distribution induced on the surface of the conductor is given by Fq = 1 q2 z. 4 0 (2z)2 (2)

Subtleties in energy calculations in the image method

967

One could naively think that the electrostatic energy of the system shown in gure 1 would also be given simply by the Coulombic energy between point charge q and its image, q, namely U = (q 2 /4 0 )(1/2z). However, this is not true, as we can easily verify if we take the gradient of U: U = z q2 1 z 4 0 (2z) = 2 q2 1 4 0 (2z)2 z = Fq . (3)

The correct expression for U has an additional factor of 1/2 and can be readily obtained if we start with the very denition of U as the total external work to bring all the real charges (including the surface distribution) from innity to the nal static conguration Cf . The easiest way to compute it is to picture the set-up of all charges of the system along their paths from innity to Cf , always consistent with the presence of the conductor. This guarantees that the surface charges move across regions of constant potential, without any work required to bring them. Under these conditions (in the electrostatic context, radiation effects are negligible), we have U = Wext =
Cf

Fq (r ) dr .

(4)

Substituting equation (2) into the above equation, we get U = Wext = q2 4 0

dz 1 q2 1 . = (2z )2 2 4 0 (2z)

(5)

In other words, the electrostatic energy of the system formed by point charge q and the conducting region is half the electrostatic energy of a point charge q located at (0, 0, z) and a point charge q located at (0, 0, z). This result is in agreement with (2) as can be readily seen. This kind of discussion can be found in many textbooks, such as Grifthss [1], among others. In this simple case, the factor 1/2 could also be anticipated by symmetry arguments, as follows. First, recall that (1/2) 0 E2 is the energy density of the electrostatic eld. With this in mind, we easily see that the energy of a system formed by the charges q at (0, 0, z) and q at (0, 0, z) (with no conductor at all) is equally divided between the regions z > 0 and z < 0. In this calculation, we must, of course, exclude the self-energies of each charge. Symmetry also allows us to deal with the case of a charge near an innite wedge whose aperture angle equals /n for any positive integer n (n = 1 corresponding to the plane wall). In these cases, the entire space can be divided into 2n sectors with that same angle, one corresponding to the outside of the conductor and the remaining 2n 1 to the space lled by the conductor. An image charge will be in each sector, except for outside the conductor, where the real charge q1 lies. We shall label the sectors and the pertaining charges with integers; i = 1 refers to the real charge and i = (2, . . . , 2n) to the image ones. Symmetry allows us to say that the conguration energy would be 1/2n the Coulombic energy of all 2n charges, i.e. U= 1 2n
2n 2n

qi Vj (ri ) =
i=1 j =i+1

1 2n

qi qj , 4 0 rij i=1 j =i+1

2n

2n

(6)

where Vj (ri ) is the potential created by charge qj at the position ri of charge qi and rij is the distance between charges qi and qj . This seems to indicate that the prefactor depends on n, i.e. on the wedge angle at hand, but we must pay closer attention to the expression we are comparing the energy with. The interaction energy of the chargewedge system is 1/2n the energy of the total system composed of 2n charges; we wish to compare it with the Coulombic interaction between the real charge and each of the images (imageimage

968

M M Taddei et al

Figure 2. Real and image charges for the chargesphere case. The image charges are q = qR/a

and q .

interactions not being included). This distinction can be very subtle; in the case n = 2 we wish to compare the actual energy with the energy of the pairs (q1 , q2 ), (q1 , q3 ), (q1 , q4 ), while the double summation on equation (6) includes these plus the pairs (q2 , q3 ), (q2 , q4 ), (q3 , q4 ). More generally, the double summation on equation (6) comprises n(2n 1) pairs. If one only counts the interaction between the real charge and each image, one nds (2n 1) pairs. Moreover, using the fact that odd-numbered charges have the value q and even-numbered ones the value q, together with the symmetry of the conguration, one can see that qi qj q1 qj =n . rij r1j i=1 j =i+1 j =2
2n 2n 2n

(7)

An interested reader may verify equation (7) for any particular value of n. We thus conclude that U= 1 1 n q1 Vj (r1 ) = 2n 2 j =2
2n 2n

q1 Vj (r1 ),
j =2

(8)

and once more the energy of the conguration is half of the interaction energy between the real charge and each image. A natural question then arises: what happens in less symmetric situations or even in situations where there is no symmetry at all? From now on, this answer is our main concern. However, we shall do that in two steps. First, we shall consider in the following section another example and work out the result explicitly. Then, we shall attack a completely general situation of one (or more) charged particle(s) in the vicinity of N grounded or neutral perfect conductors of arbitrary shapes.
3. Point charge and a perfectly conducting sphere

Let us consider as our next example a point charge q near a perfectly conducting grounded sphere of radius R. Suppose the distance from charge q to the centre of the sphere is a, a > R. For simplicity, we choose the axis OX with its origin at the centre of the sphere so that the position of charge q is given by (a, 0, 0), as shown in gure 2. It is well known that the surface charge distribution on the sphere is such that the force on q is the same as if there were no

Subtleties in energy calculations in the image method

969

sphere at all and a charge q = (R/a)q were located at (R 2 /a, 0, 0) (see [1]). Therefore, the force exerted by the surface distribution of the sphere on the point charge q is given by Fq = ax q(qR/a) q 2 R x = . 2 /a)2 2 R 2 )2 4 0 (a R 4 0 (a (9)

According to the previously presented discussion, we do not expect the electrostatic energy of the chargesphere system shown in gure 2 to be given by the Coulombic interaction energy between real and image charges. However, we have no reason, a priori, to say that the correct answer is obtained simply by including an additional factor of 1/2 as occurred in the cases discussed in section 2. It would be natural, though, to expect a factor depending on R and a that, for R with R a kept constant, reduces to the previous factor 1/2, since that limit reproduces the chargewall case. Let us then perform the explicit calculation using, as before, the very denition of the electrostatic energy of a conguration in light of the comments made before equation (4). Doing that along the axis OX and using equation (9), we may write U = Wext = x q 2R a dx 2 R 2 )2 4 0 (x 1 q(Rq/a) = , 2 4 0 (a R 2 /a)

(10)

which is nothing but half the Coulombic energy between charge q and its image q = Rq/a. At rst sight, it seems amazing that the same factor 1/2 appears. This suggests that this will happen for conductors of general shapes. In fact, this is precisely what happens, as we shall demonstrate in the following section.
4. General case
4.1. One source charge

We shall now consider one point charge q in the vicinity of a set of N perfect conductors of arbitrary shapes, which can be either neutral or grounded. Let x0 be its position in space with respect to some reference frame. Our purpose here is to obtain an expression for the electrostatic energy of this conguration in terms of the Coulombic interaction energy between q and each image charge necessary to solve the problem, which would be the energy necessary to bring in the charge q from innity with every image held xed at its nal position. For convenience, we shall anticipate the nal result in the form of a theorem, namely The electrostatic energy of a point charge q near N perfect conductors of arbitrary shapes, each conductor being either neutral or grounded, is half the Coulombic energy between the charge q and all the image charges. We now present a simple demonstration of this theorem. It is convenient to start with the following expression for the electrostatic energy U of a general charge distribution: U= 1 2 (x) (x) dV ,
R

(11)

where (x) is the charge volumar density at position x, (x) is the electrostatic potential at position x due to all charge distributions and R is a region of space containing all charges. Of course, whenever point charges are present in the distribution, we must subtract from the above expression the corresponding innite self-energies. It has already been shown in the

970

M M Taddei et al

literature (for instance, in [1]) that this expression is equivalent to computing the external work to bring the charges from innity. Since our distribution consists of a point charge q located at position x0 and surface charge distributions on the conductors, equation (11) can be written as U= 1 1 q (x0 ) + 2 2
N

k (x) (x) dAk ,

(12)

where (x0 ) is the potential at position x0 created by all charges of the system except point charge q and k describes the charge distribution on the surface Sk of the kth conductor. Writing (x0 ) instead of (x0 ) in the rst term of the rhs of the previous equation is equivalent to subtracting the (innite) self-energy of the point charge q. Recalling that each surface Sk is an equipotential surface, whose potential we denote by k , we obtain U= 1 1 q (x0 ) + 2 2
N k k=1 Sk

k (x) dAk .

(13)

Since each conductor is either grounded, from which k = 0, or neutral, from which the surface integral is zero, the last term on the rhs of equation (13) always vanishes, and U = 1 q (x0 ). 2 (14)

Now, all we need to do is to invoke the image method to nish our demonstration. Image charges are, by denition, imaginary charges situated in the nonphysical regions (inside the conductors) that create at any point of the physical region (outside the conductors) the same eld as created by all surface distributions of all conductors. Hence, (x0 ) is precisely the electrostatic potential at position x0 due to all image charges, so that we can write symbolically U = 1 q (x0 ) = 1 U (q; {images}), 2 2 (15)

where U (q; {images}) means all Coulombic interactions between point charge q and each image. This completes the demonstration for one point charge q.
4.2. Many source charges

The result can be further generalized to accommodate the presence of more source charges. We shall now consider a set of M point charges q1 , q2 , . . . , qM in the vicinity of a set of N perfect conductors of arbitrary shapes, which can be either neutral or grounded. Let xi be the position of charge qi in space with respect to some reference frame. The expression for the electrostatic energy of this conguration shall also include terms due to interaction of the real charges. Anticipating the nal result once more, The electrostatic energy of a set of M point charges q1 , q2 , . . . , qM near N perfect conductors of arbitrary shapes, each conductor being either neutral or grounded, is the Coulombic interaction energy between the real point charges plus half the sum, from i = 1 to i = M, of the Coulombic energies between charge qi and all the image charges. Let us demonstrate this second theorem. We start once again with equation (11): U= 1 2 (x) (x) dV .
R

(16)

Subtleties in energy calculations in the image method

971

We can write the expression separating (x) into the contributions due to each point charge qi located at position xi and due to the surface charge distributions on the conductors: U= 1 2
M N

qi i (xi ) +
i=1 k=1

1 2

k (x) (x) dAk .

(17)

We subtract the self-energy of point charge qi by substituting the potential (xi ) by i (xi ), the potential at position xi created by all charges of the system except point charge qi itself. The second term on the rhs of this equation vanishes analogous to how the second term on the rhs of equation (12) does. We obtain U= 1 2
M

qi i (xi ).
i=1

(18)

The potential i can be split into the potential due to the other point charges qj and the potential due to all surface charges together surf : U= 1 2
M M

i=1 j =1 j =i

qi qj 1 + 4 0 |xi xj | 2

qi
i=1

surf (xi ).

(19)

The rst term can be readily recognized as the Coulombic energy between the source charges, as can be seen in the literature [1] (we remember that each pair is counted twice in that double summation). Invoking the image method as before, we can state that surf (xi ) is the electrostatic potential at position xi due to all image charges, allowing us to write, symbolically, U = U ({all source charges}) + 1 2
M

U (qi ; {images}),
i=1

(20)

thus completing our most general demonstration. Equation (20) can be interpreted in terms of pairwise Coulombic energies as follows: the energy of each pair composed of two real charges enters the expression with a prefactor of 1; pairs that comprise a real charge and an image one take on a factor of 1/2; pairs of image charges are not present in equation (20).
5. Conclusions and nal remarks

We have calculated the electrostatic energy of systems composed of a point charge and conductors of various geometries. We started with the simple case of a plane wall and obtained as a result one-half of the Coulombic energy between real and image charges, which could be easily understood considering the elds energy density and the symmetry of the problem. We then indicated, solely on symmetry arguments, that the energy of a system composed of a charge and a wedge of aperture angle /n (n a positive integer) should also be one-half of the Coulombic energy between the charge and every image. We then moved on to a less symmetrical geometry, a spherical one, and there, too, we found the same factor 1/2 when comparing the energy of the actual system to the pairwise Coulombic energy between the real charge and each image. We then proved the theorem that this same factor 1/2 arises in every problem of a point charge in the presence of conductors that is solvable by the image method, whatever its geometry may be. The argument was completed by generalizing the theorem for the case of more than one source charge.

972

M M Taddei et al

Acknowledgments

The authors wish to thank P A Maia Neto, A Ten rio, M V Cougo-Pinto, I Waga and o V Miranda for enlightening discussions, as well as CNPq (Brazils National Research Council) and Faperj (Research Support Foundation of the State of Rio de Janeiro) for partial nancial support.
References
[1] [2] [3] [4] Grifths D J 1999 Introduction to Electrodynamics 3rd edn (Englewood Cliffs, NJ: Prentice-Hall) pp 1234, 126 Franklin J 2005 Classical Electromagnetism (San Francisco, CA: Addison-Wesley) pp 6970, 86 Jackson J D 1999 Classical Electrodynamics 3rd edn (New York: Wiley) Stratton J 1941 Electrodynamic Theory (New York: McGraw-Hill)