On the Electrostatic Forces acting on Point Charges

in the Presence of a Dielectric Slab in Between

Long T. Nguyen,1 Tuan K. Do,2, ∗ and Trung V. Phan3, †
1
High School for Gifted Students, Hanoi University of Science, Vietnam National University,
182 Luong The Vinh Str., Thanh Xuan, Hanoi 100000, Vietnam.
2
Department of Mathematics, Princeton University, Princeton, NJ 08544, USA
3
Department of Molecular, Cellular, and Developmental Biology,
Yale University, New Haven, CT 06520, USA
(Dated: August 2, 2022)
Abstract
arXiv:2105.04722v3 [physics.class-ph] 31 Jul 2022

How will the electrostatic interaction between two point charges change if we introduce a dielectric slab in the middle of
them? While the physical setting of this electromagnetic shielding problem is relatively simple, it is easy to be wronged and
the solution is surprisingly complicated. Here we will show the correct answer by using the method of images, in which an
infinite number of image charges is needed to represent the interaction. We also obtain analytical and algebraic results in some
special cases.

I. THE COMMON MISTAKE USING NAIVE where 0 is the dielectric constant of vacuum, q1 and q2
SPATIAL EXPANSION are two point charges of interests and d is the distance be-
tween them. These electrostatic forces acting on the two
We are interested in determining the electrostatic charges are opposite in direction but equal in magnitude,
forces that act on the two point charges q1 and q2 placed and
√ this value is the same as if the charges are distance
in vacuum due to presence of an infinitely large dielectric d apart in vacuum. Due to that, when dealing with
slab of thickness h and relative dielectric constant  in- the setting in Fig. 1, many students in correctly assume
serted in between. The distance between the charges and that the present of the dielectric slab is no different than
the slab are d1 and d2 , and we consider the simple case having the √
region of vacuum space it is filling expands by
where the line connecting the charges is perpendicular to the factor , thus arrive at:
the slab as shown in Fig. 1. While this is a simple setting
that represents the phenomena of electromagnetic shield- 1 q1 q2
f= √ . (2)
ing, which has a wide variety of engineering applications 4π0 (d1 + h + d2 )2
in industries1 , it is curious that the answer for this ques-
tion cannot be found in textbooks. This paper is set out Here, we will show that not only this value is wrong, but
to fill in this gap. also these forces are non-reciprocal: the magnitudes of
the forces acting on charge q1 and q2 are in general not
the same, |f1 | 6= |f2 |. The image charges configuration
and the calculation is quite non-trivial, and to the best of
our knowledge it has not been carried out and analyzed
in much details.

II. THE CORRECT SOLUTION USING AN INFI-

NITE NUMBER OF IMAGE CHARGES

The key insight dealing with this problem is to realize

that the dielectric slab doesn’t effectively expand vac-
uum space as viewed by electrostatic interaction, but the
freely-moving charges on the slab redistributed to screen
FIG. 1: An infinitely large dielectric slab is inserted in
out electrical field so that the boundary conditions for
between two point charges.
the displacement field are satisfied.
The standard dielectric image method is for the set-
In many undergraduate introductory and advance high ting where the space is filled with two mediums of dif-
school physics courses2–4 , the students are taught that ferent relative dielectric constant 1 and 2 , separated by
the electrostatic force f between charges inside a medium an infinite-planar interface (see Fig. 2a). Place a point
of relative dielectric constant  is given by the Coulomb’s charge q inside region 1 , the final distribution of free-
equation: moving charges in space will give electric field in region 1
1 q1 q2 equivalent to that of two charge: the original charge q and
f= , (1) an image charge q 0 which q 0 /q = (1 − 2 )/(1 + 2 ) at the
4π0  d2

1
mirror-reflected position of charge q through the interface Thus the electrostatic force f1 acting on charge q1 can be
(see Fig. 2b), while in in region 2 it is equivalent to that calculated as an infinite summation series:
of another image charge q 00 which q 00 /q = 21 /(1 + 2 ) ∞ (1,k)
at the same position with q (see Fig. 2c).5,6 q1 q10 X q1
f1 = 2
+
2
We consider the electrical field created by a system 4π0 (2d1 )
k=0 2(d1 + h) + 2kh
placed in vacuum consists of a point charge q1 at distance ∞
! (7)
(1,k)
d1 from an infinitely large dielectric slab of thickness h
X q2
+
2 .
and relative dielectric constant . The slab divides space k=0 (d1 + h + d2 ) + 2kh
into three regions: region 1 where the charge q1 is in,
region 2 where the slab is filling and region 3 has the Define the ratios α1 = d1 /h, α2 = d2 /h and define the
rest of the space. Using the method of images to deal summation Sa (z):
consecutively with the two infinite-planar interfaces, we ∞
X z 2k
obtain an infinite series of image charges for any region. Sa (z) = , (8)
The electrical field seen inside the region 1 is equivalent to (a + 2k)2
k=0
the electrical field produced by an infinite series of point we can rewrite the force equation as:
charges: q1 , an image charge q10 placed outside region 1 at "
(1,k)  
distance 2d1 and image charges q1 where k = 0, 1, 2, ... q1 −1 1
f1 = − q 1
placed outside region 1 at distance 2(d1 + h) + 2kh (see 4π0 h 2  + 1 (2α1 )2
Fig. 3b):  2    
2 −1 −1
 2  2k+1 + q1  S2(α1 +1) (9)
−1 (1,k) 2 −1 +1 +1 +1
q10 = − q1 , q1 = q1 .
+1 +1 +1  2  #
2 −1
(3) + q2  Sα1 +1+α2 .
The electrical field seen inside the region 0 is equivalent to +1 +1
the electrical field produced by an infinite series of point
(0,k) Similarly, we get f2 acting on charge q2 , which magnitude
charges: image charges q1 where k = 0, 2, 4, ... placed can be shown to be different from that of f1 in general.
(0,k)
inside region 1 at distance kh and image charges q1 This means the interaction between these two charges
where k = 1, 3, 5, ... placed outside region 1 at distance mediated by the dielectric slab is non-reciprocal.
2(d1 + h) + (k − 1)h (see Fig. 3c): The result here can also by found by solving the Pois-
  k son equation. For the sake of demonstration, we will
(0,k) 2 −1 sketch a derivation in the Appendix A for the simple
q1 = q1 . (4)
+1 +1 case where the setting is symmetric.
The electrical field seen inside the region 2 is equivalent
to the electrical field produced by an infinite series of B. Some special cases
(0,k)
point charges: image charges q1 where k = 0, 1, 2, ...
placed inside region 1 at distance 2kh (see Fig. 3d): There are special cases where Eq. (9) can be repre-
 2  2k sented by special functions and even have a nice short
(2,k) 2 −1
q1 = q1 . (5) algebraic form. P∞ k 2
+1 +1 The dilogarithm function7 Li2 (z) = k=1 z /k can
be used to rewrite the infinite sum in Eq. (8) with finite
terms. For examples, when a ∈ 2Z+ :
A. The general case P(a−2)/2 z2k
Li2 (z 2 ) − k=1 k2
Sa (z) = , (10)
4z a
Now we have enough ingredient to solve the origi-
nal problem by considering the superposition of all the and when a ∈ 2Z+ + 1:
charges and image charges as we introduced charge q2 at P(a−3)/2 z 2k+1
Li2 (z) − Li2 (z 2 )/4 − k=0 (2k+1)2
region 2 distance d2 away from the dielectric slab (see Sa (z) = . (11)
Fig. 4a). For example, the electrical as seen in re- za
gion 1 is equivalent to the electrical field produced by When d1 = h = d2 , we get the ratios α1 = α2 = 1 and
(1,k)
not only charge q1 and image charges q10 , q1 where the electrostatic force in Eq. (9) becomes:
(1,k)
k = 0, 1, 2, ... but also q2 placed inside region 2 at dis-
!
tance (d1 + h + d2 ) + 2kh away from charge q1 (see Fig. q12 ζ (1 − ζ 2 ) Li2 (ζ 2 ) − ζ 2
f1 = − +
4b): 4π0 h2 4 4ζ 3

! (12)

2
2 
−1
2k q1 q2 (1 − ζ 2 ) 4Li2 (ζ) − Li2 (ζ 2 ) − 4ζ
(1,k) + ,
q2 = q2 . (6) 4π0 h2 4ζ 3
+1 +1

2
FIG. 2: (a) The space is divided into two regions having different dielectric constants by an infinite-planar interface.
A point charge q is introduced inside on of these regions. (b) The electrical field seen inside the same region as the
charge q is equivalent to the electrical field produced by two point charges: q and image charge q 0 . (c) The electrical
field seen inside the different region as the charge is equivalent to the electrical field produced by one point charge:
image charge q 00

FIG. 3: (a) The space is divided into three regions by the infinitely large dielectric slab. A point charge q1 is
introduced outside of the slab. (b) The electrical field seen inside the same region as the charge q is equivalent to the
(1,0) (1,1) (1,2)
electrical field produced by an infinite series of point charges: q1 and image charges q10 , q1 , q1 , q1 , ... (c) The
electrical field seen inside the slab is equivalent to the electrical field produced by an infinite series of point charges:
(0,0) (0,1) (0,2)
image charges q1 , q1 , q1 , ... (d) The electrical field seen inside the vacuum region without charge q1 is
(2,0) (2,1) (2,2)
equivalent to the electrical field produced by an infinite series of point charges: image charges q1 , q1 , q1 , ...

where ζ = ( − 1)/( + 1). Consider further simplification For the relative dielectric constant  = 3, we get ζ = 1/2.
with q1 = q2 = q, the interaction is now reciprocal f1 = Since Li2 (1/2) = π 2 /12 − ln2 2/2,8 we can arrive at the
f2 = f and we also arrive at:

q2
 
1 1 Li2 (ζ) Li2 (ζ)
f= 1− − 2− + . (13)
4π0 h2 4ζ ζ ζ ζ3

3
 FIG. 4: (a) Introduce charge q2 on the opposite side of charge q1 with respect to the slab. (b) Superposing the
image charges from charge q1 and q2 , as seen from region 1.

following algebraic form: ansatz  = 3+δ in which δ  3, then in the leading order
of δ Eq. (13) can be written as:
q2 π 2 − 7 − 6 ln2 2
f= ×
4π0 h2 2 q2 π 2 − 7 − 6 ln2 2
2
(14) f≈ ×
q 4π0 h2 2
× −6.6 × 10−3 .


≈ 2 (18)
4π0 h !
−11π 2 + 51 + 66 ln2 2 + 36 ln 2
+ δ ,
The “−” sign indicates that the electrostatic forces are 24
pulling the charges toward the dielectric slab. Compare
with Eq. (2) which comes from the naive assumption of thus f = 0 can be satisfied around:
expanding space:
12π 2 − 84 − 72 ln2 2
q 2 δ= ≈ −0.1744, (19)
f= √ −11π 2 + 51 + 66 ln2 2 + 36 ln 2
4π0 (2 + 3)2 h2
(15) which results in  ≈ 2.8256. A numerical investigation of
q2
× +7.2 × 10−2 ,


≈ Eq. (13), which gives Fig. 5, is indeed in good agreement
4π0 h2
as f = 0 when  ≈ 2.8308.
this wrong answer has incorrect sign and the interaction
strength is off by an order of magnitude.
There are other interesting limits that can be read-off
easily from Eq. (13). For the relative dielectric constant
satisfies | − 1|  1, we have ζ ≈ ( − 1)/2, Li2 (ζ) ≈
( − 1)/2 and get the approximation:

q2
 
1 3( − 1)
f≈ − >0, (16)
4π0 h2 9 32

which is consistent with the slab vanishing as  → 1. For

the relative dielectric constant satisfies   1, we have
ζ ≈ 1, Li2 (ζ) ≈ π 2 /6 and get the approximation:

q2 q2
 
1
f≈ − = − <0, (17)
4π0 h2 4 4π0 (2d1 )2 FIG. 5: The electrostatic force f as a function of
dielectric constant , using Eq. (13).
which is consistent with the slab being a conductor (the
interaction is mostly between q1 and its image charges
q10 ≈ −q1 while the electrical field from q2 is perfectly
shielded out).
There exists a value  in which the electrostatic forces III. CONCLUSION
acting on the charges vanish, which corresponds to the
perfect shielding scenario. From (15), (16) and (17), that We have calculated the interaction between charges af-
value of  can be guessed to be close to  = 3. Try the ter separating them by a dielectric shielding, using the

4
method of images (which should be familiar to most un- The jumping conditions of ∂z Ṽk (z) at the charges’ posi-
dergraduate students majored in physics and electrical tions give:
engineering) and representing the answer by an infinite
q  
summation series (which can be numerically evaluated). − = −Ake−k(d+h/2)
We hope that this paper can fill in the gap often left in 0
  (A4)
introductory physics course or at the very least eliminate − −Bke−k(d+h/2) + Cke+k(d+h/2) .
the incorrect understanding on a simple setting of great
relevant to many industrial applications. The continuity conditions of ∂z Ṽk (z) at the interfaces
give:
 
IV. ACKNOWLEDGEMENT 0 = −Bke−kh/2 + Cke+kh/2
  (A5)
− D −ke−kh/2 + ke+kh/2 .
We thank Tung X. Tran, Duy V. Nguyen, Nam H.
Nguyen and the xPhO journal club for their support to Four unknowns A, B, C, D can be solved with four equa-
share this finding to a wider audience. tions (A3), (A4), (A5):

2qekh cosh(kd) cosh( kh kh

2 ) +  sinh(kd) sinh( 2 )
A= ,
Appendix A: Solving the Poisson Equations
0 k ((1 − ) + (1 + )ekh )
qe−k(d−h/2) (1 + ) + (1 − )ekh


B= ,
Consider the simple case where the setting is symmet- 20 k ((1 − ) + (1 + )ekh )
ric, q1 = q2 = q and d1 = d2 = d. Choose a Cartesian qe−k(d+h/2)
Oxyz coordinate system, in which the origin is inside the C= ,
20 k
dielectric slab at the middle of two charges, the xy-plane
qe−k(d−h/2)
is parallel to the slab thus the z-axis passes through both D= .
charges. Decompose the potential into Fourier-modes in 0 k ((1 − ) + (1 + )ekh )
the xy-plane: (A6)
Z Now let’s look at the charge’s position z = d + h/2 and
dkx dky ikx x+iky y
V (x, y, z) = e Ṽk (z) , (A1) the self-contribution:
(2π)2
h (s) (s)
z >d+ : Ṽk (z) = A+ e−kz ,
2
q
where k = kx2 + ky2 is a sufficient index due to rotational (A7)
h (s) (s)
symmetry in the xy-plane. Excluding the two interfaces d+ > z : Ṽk (z) = A− e+kz .
2
and the positions of two charges we get the Laplace’s
equation ∇2 V = 0, hence the solution will be of the form (s)
The continuity condition of Ṽk (z):
(due to the setting is symmetric and far-away potential
should go to 0): (s) (s)
A+ e−k(d+h/2) = A− e+k(d+h/2) . (A8)

h (s)
z >d+ : Ṽk (z) = Ae−kz , The jumping conditions of Ṽk (z):
2
h h q 
(s)
 
(s)

d+ >z > : Ṽk (z) = Be−kz + Ce+kz , − = −A+ ke−k(d+h/2) − A− ke+k(d+h/2) . (A9)
2 2 0
h
> |z| : Ṽk (z) = D e−kz + e+kz , From (A8) and (A9), we get:


(A2)
2
h h qe±k(d+h/2)
d + > −z > : Ṽk (z) = Ce−kz + Be+kz , (s)
A± = (A10)
2 2 20 k
h
−z > d + : Ṽk (z) = Ae+kz + De+kz . (r) (s)
2 The regularized Ṽk (z) = Ṽk (z) − Ṽk (z) is continuous
and smooth at that charge’s position, and can be used to
The continuity conditions of Ṽk (z) at the charges’ posi- determine the gradient right there:
tions and the interfaces give:  
(r) (s)
∂z Ṽk (z) h
= A − A + ∂z e−kz h
z=d+ 2 z=d+ 2
Ae−k(d+h/2) = Be−k(d+h/2) + Ce−k(d+h)/2 , (A11)
(A3) qe−2kd 1 −  tanh( kh
2 )
=− .
 
Be−kh/2 + Ce+kh/2 = D e+kh/2 + e+kh/2 . 20 1 + kh
 tanh( 2 )

5
 Thus the force acting on the charge can be calculated
with:

f = −q∂z V (r) (0, 0, z)

z=d+ h 2
Z ∞
2πkdk (r)
= −q ∂z Ṽk (z) (A12)
0 (2π)2 z=d+ h
2
∞ kh
q2 1 −  tanh( 2 )
Z
= dkke−2kd kh
.
4π0 0 1 +  tanh( 2 )

Define χ = kh/2 then:

Z ∞
q2 1 −  tanh χ
f= 2
× 4 dχχe−4χd/h . (A13)
4π0 h 0 1 +  tanh χ

When d = h, we get back the result Eq. (13).

∗ 5
kdo@princeton.edu P. Hammond, Proceedings of the IEE-Part C: Monographs
†
trung.phan@yale.edu 107, 306 (1960).
1 6
S. Celozzi, R. Araneo, and G. Lovat, Electromagnetic shield- J. D. Jackson, Classical electrodynamics (1999).
7
ing (John Wiley & Sons, 2008). P. E. Cartier, B. Julia, P. Moussa, and P. Vanhove, Frontiers
2
E. M. Purcell and D. J. Morin, Electricity and magnetism in number theory, physics, and geometry II: on conformal
(Cambridge University Press, 2013). field theories, discrete groups and renormalization, Vol. 2
3
L. D. Landau, J. Bell, M. Kearsley, L. Pitaevskii, E. Lif- (Springer Science & Business Media, 2007).
8
shitz, and J. Sykes, Electrodynamics of continuous media, J. Loxton, Acta Arithmetica 43, 155 (1984).
Vol. 8 (elsevier, 2013).
4
J. A. Stratton, Electromagnetic theory, Vol. 33 (John Wiley
& Sons, 2007).