Journal of Electrostatics 66 (2008) 561–563

Contents lists available at ScienceDirect

Journal of Electrostatics
journal homepage: www.elsevier.com/locate/elstat

Proof of Thomson’s theorem of electrostatics

Ezzat G. Bakhoum
Department of Electrical and Computer Engineering, University of West Florida, United States

a r t i c l e i n f o a b s t r a c t

Article history: A 100 years old formula that was given by Thomson [Footnote on p. 154 of J.J. Thomson, Maxwell’s
Received 7 August 2006 Treatise on Electricity and Magnetism, Dover, New York, NY, 1954.] recently found numerous applications
Accepted 18 June 2008 in computational electrostatics and electromagnetics. Thomson himself never gave a proof for the for-
Available online 15 July 2008
mula; but a proof based on Differential Geometry was suggested by Jackson [Classical Electrodynamics,
Wiley, New York, NY, 1962, p. 51] and later published by Pappas [Differential geometric solution of a
Keywords: problem in electrostatics, SIAM Rev. 28 (1986) 225]. Unfortunately, Differential Geometry, being a spe-
Thomson’s theorem of electrostatics
cialized branch of mathematics, is normally inaccessible to the majority of scientists and engineers. This
Potential gradient
Curvature
paper provides for the first time a proof that does not depend on Differential Geometry.
Equipotential surface Ó 2008 Elsevier B.V. All rights reserved.
Numerical solutions of field problems

1. Introduction authors, however, treated the problem using an inappropriate

In 1891, Thomson, the discoverer of the electron, stated without
a proof a formula [1] that relates the vertical potential gradient (or
electric field intensity) at any point on the surface of a charged
conductor to the mean curvature of that surface. The formula
remained unrecognized and unused for nearly 100 years, until the
emerging new techniques of the rapid numerical solution of elec-
trostatic and electromagnetic field problems finally put Thomson’s
formula in the spotlight [4–8]. The formula was also found to be
extremely valuable in a similar application, namely, numerical
geodesy (the study of the geoid, or the equipotential surface of the
earth) [9].
While Thomson didn’t originally provide a proof for his
theorem, there seem to be hardly any proof for it in the scientific
literature of the past 100 years either. The few attempts to prove the
theorem that have in the past appeared in the literature show that
the theorem is in fact quite challenging to prove (except for trivial
geometrical configurations). The first attempt to prove the theorem
was apparently the one suggested (but not published in a detailed
manner) by Jackson in his popular book Classical Electrodynamics
[2]. The solution is based on Gauss’s theorem of Differential
Geometry (note that this is not the same as Gauss’s theorem of
electrostatics!). Unfortunately, Differential Geometry, being a spe-
cialized branch of mathematics, is normally inaccessible to many
scientists and engineers (and students!). A second attempt to prove
Thomson’s theorem (aimed at providing a simpler proof for the
theorem) was published by Estevez and Bhuiyan [10]. Those
E-mail address: ebakhoum@uwf.edu
mathematical approach (power series) and essentially failed to give
a valid proof for the formula. Finally, a proof based on Jackson’s
differential-geometric idea was published by Pappas [3]. From 1986
to the present the scientific literature was, once again, silent about
the origin of Thomson’s theorem.
The purpose of this paper is to demonstrate, for the first time
since 1891, that Thomson’s theorem can be in fact proved in a
straightforward manner directly from the fundamental laws of
electrostatics, without invoking Differential Geometry or any
advanced mathematical concepts. Since such a proof does not exist
in the literature, it will be of important archival value for the
researchers who are currently using the theorem in various
applications.
2. Thomson’s formula and its proof
In a small footnote written in 1891 [1], Thomson gave the fol-
lowing formula:
 
vjEj 1 1
¼ jEj þ : (1)
vz R1 R2
This formula relates the normal derivative of the magnitude of the
!
electric field intensity E to the so called ‘‘principal radii of curva-
!
ture’’, R1 and R2, of an equipotential surface to which E is per-
pendicular. Here, z is the normal direction to the surface at the
point under consideration (see Fig. 1).
We shall now proceed to prove Eq. (1) from the very basic
principles of electrostatics. We start by noting that the potential

0304-3886/$ – see front matter Ó 2008 Elsevier B.V. All rights reserved.
doi:10.1016/j.elstat.2008.06.002
562 E.G. Bakhoum / Journal of Electrostatics 66 (2008) 561–563

Z vU vU vU
VU ¼ b y; b
x; b z ¼ Ex b y ; Ez b
x ; Ey b z; (7)
vx vy vz

E where b y; b
x; b z are unit vectors along the coordinate system axes.
Hence Laplace’s equation can be written as

vEx vEy vEz

V2 U ¼ 0 ¼ þ þ (8)
vx vy vz
X !
(notice that this is the same as div E ¼ 0). Substituting in the last
equation from the three partial derivatives in Eq. (5) yields
Y  
vdx vdy vdz vjEj
0 ¼ jEj þ þ þ : (9)
Fig. 1. An orthogonal coordinate system at a point on an equipotential surface. Z is vx vy vz vz
normal to the surface and X, Y are oriented along the ‘‘principal directions’’ (the di-
! Now we note that
rections of maximum and minimum curvature) on the surface.1 The electric field E is
directed along the Z axis. qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
jEj ¼ Ex2 þ Ey2 þ Ez2 ; (10)

U(x, y, z) is a constant on the equipotential surface. Furthermore, the from which we have
!
field intensity E is given by the gradient  
vjEj 1 vEx vEy vEz
! ¼ 2Ex þ 2Ey þ 2Ez
E ¼ VU; (2) vz 2jEj vz vz vz
vEx vEy vEz vEz
according to the theory of electrostatics [11]. In Fig. 1, we shall as- ¼ dx þ dy þ dz ¼ (11)
! vz vz vz vz
sume that the direction cosines of the vector E are dx, dy and dz
(shorthand notation for cos qx, cos qy and cos qz, or the cosines (note that dx ¼ dy ¼ 0 and dz ¼ 1). From Eq. (11) and the last of Eq. (5)
of the angles formed between the vector and the principal axes). we must conclude that
!
Since E is directed along the Z axis, then dz ¼ cos 0 ¼ 1 and
dx ¼ dy ¼ cos 90 ¼ 0. We now write an expression for the compo- vdz
! jEj ¼ 0; (12)
nents of E in terms of its direction cosines: vz
and hence vdz/vz ¼ 0 (this result, as a matter of fact, should have
Ei ¼ jEjdi ; (3) been an obvious result). Eq. (9) now reduces to
where i ¼ x, y, z, and where jEj is the magnitude of the vector. Even  
vjEj vdx vdy
though we realize that the components Ex ¼ Ey ¼ 0, we are ¼ jEj þ : (13)
interested in the partial derivatives of those components, which do vz vx vy
not vanish. We now take the partial derivative of each term in Eq. The rate of change of a direction cosine, such as dx or dy, is related to
(3) with respect to the coordinate i, obtaining the concept of ‘‘curvature’’ of a curve by a well known relationship.
vEi vd vjEj This relationship can be derived in a simple manner without in-
¼ jEj i þ di : (4) voking Differential Geometry, as shown in Fig. 2.
vi vi vi
In Fig. 2, let the arbitrary curve shown be a curve embedded in
Now we write this equation explicitly for each of the coordinates, x, the surface under consideration and directed along one of the
y and z, noting that dz ¼ 1 and dx ¼ dy ¼ 0. We have principal directions. At the specific point where the electric field
!
vector E is considered, the X axis is tangent to the curve and R is the
vEx vdx
¼ jEj ; radius of a circle that is also tangent to the curve. In theory, if we
vx vx consider R to be the ‘‘radius of curvature’’ of the curve at the point
under consideration, then the circle is also known as the ‘‘oscu-
vEy vdy !
¼ jEj ; lating’’ circle. For an infinitesimal deviation of the vector E from
vy vy the vertical direction, the cosine of the angle qx can be determined
from the small triangle shown in the figure. We have
vEz vdz vjEj
¼ jEj þ : (5)
vz vz vz
E
The potential U satisfies Laplace’s equation [11], that is

v2 U v2 U v2 U
V2 U ¼ þ 2 þ 2 ¼ 0: (6)
vx2 vy vz θx
x
But from Eq. (2) we note that X
R

1
Note that no prior knowledge of Differential Geometry is required to un-
derstand the derivations that will follow. Now, it may seem odd to the reader who
is not familiar with Differential Geometry that the maximum and the minimum
curvatures at any point on a surface will occur at orthogonal directions. However,
Differential Geometry says that this is necessarily so. See Ref. [12] for proof.
Fig. 2. An arbitrary curve and its tangent direction, X, at a point under consideration.
For a small deviation from the vertical, the cosine of the angle qx, formed between X
!
and the electric field vector E , is approximately equal to x/R, where R is the radius of
the osculating circle that defines the ‘‘radius of curvature’’ of the curve.
 E.G. Bakhoum / Journal of Electrostatics 66 (2008) 561–563 563

 
vjEj 1 1
¼ jEj þ ; (16)
vz R1 R2
where R1 and R2 are the so called ‘‘principal radii’’ of curvature of
the surface at the point under consideration. This concludes the
proof of Thomson’s formula. The reader who is interested in
learning more about the concept of curvature and about Differen-
tial Geometry in general can refer to Ref. [12]. It is to be noted that
when the formula is applied to the case of geodesy (the study of the
earth’s equipotential surface), we merely replace the electric field
! !
intensity E by the earth’s acceleration of gravity g . The problem is
otherwise mathematically identical.
The proof given in this paper does not exist in the scientific
literature, and it is hoped that it will be a valuable reference for the
researchers who are currently using Thomson’s theorem as well as
to the future generations of students.

3. Applications of Thomson’s theorem

Thomson’s theorem is currently being used extensively in the

tracing and the visualization of curvilinear squares field maps; since
the theorem is essentially a ‘‘visual Laplace solver’’. Fig. 3(a) shows
a plot of the potential function V ¼ X2  Y2, which satisfies Laplace’s
equation. Fig. 3(b) is a plot of the equipotentials that exist between
two electrodes, where one of the electrodes is a sphere and the other
is a plane (the electric field lines have been suppressed for clarity).
A detailed algorithm for using Thomson’s theorem in the solu-
tion of electrostatic field problems can be found in Ref. [6]. It is to be
noted, however, that other tools do exist for the graphical mapping
of field problems and that the real importance of Thomson’s the-
orem lies in its applications in the rapid solution of those problems
(see Ref. [6] for a discussion of those applications).

References

[1] Footnote on p. 154 of J.J. Thomson, Maxwell’s Treatise on Electricity and

Magnetism, Dover, New York, NY, 1954.
[2] J.D. Jackson, Classical Electrodynamics, Wiley, New York, NY, 1962, p. 51.
[3] R.C. Pappas, Differential geometric solution of a problem in electrostatics,
SIAM Rev. 28 (1986) 225.
Fig. 3. (a) Plot of the potential function V ¼ X2  Y2. (b) Solution of Laplace’s equation
[4] R.J. Haywood, M. Renksizbulut, G.D. Raithby, Transient deformation of freely
between a spherical and a planar electrodes. suspended liquid droplets in electrostatic fields, J. Am. Inst. Chem. Eng. 37 (9)
(1991) 1305.
[5] J.F. Holmes et al., Laser Modulation of LMI Sources, US Patent No. 5015862, 1991.
[6] E. Bakhoum, J.A. Board Jr., The geometric solution of Laplace’s equation,
x J. Comput. Phys. 123 (1996) 274.
dx ¼ cos qx z : (14) [7] J.A. Carretero-Benignos, Numerical Simulation of a Single Emitter Colloid
R
Thruster in Pure Droplet Cone-Jet Mode, Ph.D. thesis, Massachusetts Institute
of Technology, February 2005.
[8] J. Zhanga, K. Adamiak, G.S. Castle, Numerical modeling of negative corona
Hence,
discharge in oxygen under different pressures, J. Electrostatics 65 (3) (2007)
174.
vdx vcos qx 1 [9] W.H. Smith, Seafloor tectonic fabric from satellite altimetry, Annu. Rev. Earth
¼ z : (15) Planet. Sci. 26 (1998) 697.
vx vx R
[10] G.A. Estevez, L.B. Bhuiyan, Power series expansion solution to a classical
The rate of change of the direction cosine with respect to X is problem in electrostatics, Am. J. Phys. 53 (1985) 133.
therefore equal to the reciprocal of the radius of curvature along [11] W.H. Hayt Jr., J.A. Buck, Engineering Electromagnetics, McGraw-Hill, New York,
NY, 2006.
that principal direction. Since we have two principal directions on [12] Manfredo P. DoCarmo, Differential Geometry of Curves and Surfaces, Prentice
the surface, X and Y, then Eq. (13) can be written as follows: Hall, Englewood Cliffs, NJ, 1976.