J. Phys. Commun. 5 (2021) 115010 https://doi.org/10.

1088/2399-6528/ac3461

PAPER

Theory of perturbation of electric potential by a 3D object made of an

OPEN ACCESS
anisotropic dielectric material
RECEIVED
24 August 2021
Akhlesh Lakhtakia1 , Hamad M Alkhoori2 and Nikolaos L Tsitsas3
REVISED 1
7 October 2021 Department of Engineering Science and Mechanics, The Pennsylvania State University, University Park, Pennsylvania 16802, United
States of America
ACCEPTED FOR PUBLICATION 2
Department of Electrical Engineering, United Arab Emirates University, P O Box 15551, Al Ain, United Arab Emirates
28 October 2021 3
School of Informatics, Aristotle University of Thessaloniki, Thessaloniki 54124, Greece
PUBLISHED
11 November 2021 E-mail: hamad.alkhoori@uaeu.ac.ae

Keywords: anisotropic dielectric, electrostatics, extended boundary condition method

Original content from this
work may be used under
the terms of the Creative
Commons Attribution 4.0
licence.
Abstract
Any further distribution of The extended boundary condition method (EBCM) was formulated for the perturbation of a source
this work must maintain
attribution to the
electric potential by a 3D object composed of a homogeneous anisotropic dielectric medium whose
author(s) and the title of relative permittivity dyadic is positive definite. The formulation required the application of Green’s
the work, journal citation
and DOI. second identity to the exterior region to deduce the electrostatic counterpart of the Ewald–Oseen
extinction theorem. The electric potential inside the object was represented using a basis obtained by
implementing an affine bijective transformation of space to the Gauss equation for the electric field.
The EBCM yields a transition matrix that depends on the geometry and the composition of the 3D
object, but not on the source potential.

1. Introduction

The disturbance of a field in a homogeneous medium by an object embedded therein is a problem of intense
interest in diverse branches of physics such as acoustics [1–3], electromagnetics [2–5], solid mechanics [2, 3, 6],
and fluid mechanics [7, 8]. So long as linear constitutive relations hold in the ambient medium as well as in the
disturbing object, analogous mathematical approaches can be used in all of these branches, as is exemplified by
the Ewald–Oseen extinction theorem, which emerged in electromagnetics [9, 10] but has been applied in
acoustics [11], elastodynamics [12], and fluid mechanics [13].
Dating back more than a hundred years, the Ewald–Oseen extinction theorem states that when a 3D object is
illuminated by an incident time-harmonic electromagnetic field, the electric and magnetic surface current
densities induced on the exterior side of its surface produce an electromagnetic field that cancels the incident
field throughout the interior of the object. This theorem is a cornerstone of the extended boundary condition
method (EBCM) [14] since its inception in 1965 [15]. Since a bilinear expansion of the free-space dyadic Green
function is known [15, 16], the EBCM can be used to investigate scattering by an object composed of any linear
homogeneous medium for which a basis exists to represent the electromagnetic field therein. This requirement
can be satisfied by isotropic dielectric-magnetic mediums, isotropic chiral mediums, and orthorhombic
dielectric-magnetic mediums with gyrotropic magnetoelectric properties [17]. In addition, bases have been
numerically synthesized for certain gyrotropic dielectric-magnetic mediums [18–20]. The EBCM literature
continues to expand as time marches on [14, 21–23].
Historically, the situation has been markedly different in electrostatics. Green’s second identity yields an
expression [24, 25] that was used by Farafonov [26] in 2014 to formulate the EBCM for the perturbation (i.e.,
‘scattering’) of a source electric potential (the electrostatic counterpart of an incident time-harmonic
electromagnetic field) by a 3D object composed of a homogeneous isotropic medium. In the Farafonov
formulation, Green’s second identity is applied separately to the exterior and interior regions. The potentials in
the two regions are certainly different but, as the Green function

© 2021 The Author(s). Published by IOP Publishing Ltd

J. Phys. Commun. 5 (2021) 115010 A Lakhtakia et al

Figure 1. Schematic of the boundary-value problem. The region Vout is vacuous and the region Vin is filled with an anisotropic
dielectric material of relative permittivity dyadic e rel , given by equations (3) and (4).

1
G (r , r ¢) = (1)
4p∣r - r ¢∣

for the Poisson equation is applicable to both regions, there is no need to set up the electrostatic counterpart of
the Ewald–Oseen extinction theorem. Eigenfunctions of the Laplace equation are used in infinite-series
representations of the source potential everywhere, the perturbation potential in the exterior region, as well as
the potential induced inside the isotropic dielectric object, since the right side of equation (1) has a bilinear
expansion [27] in terms of those eigenfunctions. All three series are suitably terminated and a transition matrix is
obtained to characterize the perturbation of the source potential by the object. Although the Farafonov
formulation does not require the object to be axisymmetric, it has been used only for such objects [28–30].
Our objective in this paper is to generalize the EBCM for electrostatic problems in which the perturbing 3D
object is composed of a homogeneous anisotropic dielectric medium whose permittivity dyadic is positive
definite. We use Green’s second identity only in the exterior region and deduce the electrostatic counterpart of
the Ewald–Oseen extinction theorem therefrom. As the Laplace and Poisson equations apply in the exterior
region, our representations of the source and perturbation potentials are the same as in the Farafonov
formulation. But, the Laplace equation is inapplicable in any region occupied by a homogeneous anisotropic
dielectric medium. Therefore, an affine transformation of space in the Gauss equation for the electric field is
implemented in order to generate a basis for representing the electric potential induced inside the object [31].
This basis is premised on the eigenfunctions of the Laplace equation in the spherical coordinate system.
There is no requirement of axisymmetry in our formulation, as is explicitly demonstrated with numerical results
for perturbation by ellipsoids. Hence, this work can be useful for examining the effects of inclusions and cracks in the
electrical characteristics (such as the capacitance) of semiconductors (e.g., silicon and germanium), ceramics (e.g.,
porcelain and titanium carbide), refractory materials (e.g., zirconium nitride and magnesium carbide), glasses, and
gemstones. As such, it can be applied for quality control of a large variety of materials and devices.
The plan of this paper is as follows. Section 2 describes the boundary-value problem, section 3 contains the
derivation of integral equations underlying the EBCM for electrostatics, section 4 describes the formulation of
the transition matrix that relates the expansion coefficients of the perturbation potential to those of the source
potential, and section 5 presents and discusses illustrative numerical results. The paper concludes in section 6
with remarks pertaining to future work.

2. boundary-value problem

Let the region Vin be the interior of the surface S, whereas the region Vout be bounded by the surfaces S and S∞,
as shown in figure 1. The coordinate system has its origin lying inside Vin, the sphere with the origin as its center

2
J. Phys. Commun. 5 (2021) 115010 A Lakhtakia et al

and inscribed in Vin has radius rin, the sphere circumscribing Vin has radius rout, and S∞ is a sphere of radius r∞.
The surface S is described by a continuous and once-differentiable function.
The electric potential Φ(r) satisfies the Poisson equation
2F (r) = - e-
0 r (r) ,
1
r Î Vout , (2)
in the region Vout, with ε0 as the permittivity of free space (i.e., vacuum) and the charge density ρ(r) being non-
zero only for r ä Vso ⊂ Vout. Every point in Vso is at least a distance rso away from the origin, as shown in figure 1.
The charge-free region Vin is filled with an anisotropic dielectric medium with permittivity dyadic
e = e0 e rel = e0 er A · A, (3)
wherein the scalar εr > 0 and the diagonal dyadic
A = a-x 1 xx
ˆ ˆ + a-y 1 yy
ˆ ˆ + zz
ˆˆ (4)
contains the anisotropy parameters αx > 0 and αy > 0. The constitutive principal axes of this medium are thus
parallel to x̂ , ŷ , and ẑ .
Any real symmetric dyadic can be written in the form of equations (3) and (4) by virtue of the principal axis
theorem [32, 33]. Anisotropic materials exist in nature [34], and homogenizable composite materials of this kind
can be engineered by properly dispersing dielectric fibers in some host isotropic dielectric material [35].

3. Integral equations

Application of Green’s second identity to the exterior region Vout yields [24, 36]
∭Vout F (r ¢) d (r - r ¢) d 3r ¢
= e- 1
0 ∭ r (r ¢) G (r , r ¢) d 3r ¢
Vso

+∬ [F+(r ¢) nˆ (r ¢) · ¢G (r , r ¢)
S È S¥

- G (r , r ¢) nˆ (r ¢) · ¢F+(r ¢)] d 2r ¢, (5)

where F+(r ¢) is the value of F (r ¢) on the exterior side of S, d (r - r ¢) is the Dirac delta, and the unit normal
vector nˆ (r) at r ä S ∪ S∞ points into Vout.
For finite r and with S∞ taken to be a spherical surface of radius r∞, the potential F (r ¢) at r ¢ Î S¥ drops off
as 1 r ¢ as r∞ → ∞ . Accordingly, the integral on S∞ in equation (5) vanishes. Furthermore, let us define the
source potential

Fso (r) = e-
0
1
∭ r (r ¢) G (r , r ¢) d 3r ¢, (6)
Vso

because it exists everywhere when Vin is vacuous (just like Vout actually is). Equation (5) then delivers the twin
equations
0 = Fso (r) + ∬ [F+(r ¢) nˆ (r ¢) · ¢G (r , r ¢)
S
- G (r , r ¢) nˆ (r ¢) · ¢F+(r ¢)] d 2r ¢, r Î Vin , (7a)
and
F (r) = Fso (r) + ∬ [F+(r ¢) nˆ (r ¢) · ¢G (r , r ¢)
S
- G (r , r ¢) nˆ (r ¢) · ¢F+(r ¢)] d 2r ¢, r Î Vout. (7b)
Equation (7a) is the electrostatic equivalent of the Ewald–Oseen extinction theorem [9, 10, 14], and is the
cornerstone of our EBCM formulation. It is convenient to rewrite it as
Fso (r) = - ∬ [F+(r ¢) nˆ (r ¢) · ¢G (r , r ¢)
S
- G (r , r ¢) nˆ (r ¢) · ¢F+(r ¢)] d 2r ¢, r Î Vin. (8a)
According to equation (7b) the electric potential in Vout has two components, one of which is the source
potential Φso(r) and the other is the perturbation potential Φpert(r) due to the region Vin being non-vacuous [24];
indeed, Φpert(r) ≡ 0 if Vin is vacuous (just like Vout). Hence, equation (7b) yields
Fpert (r) = ∬ [F+(r ¢) nˆ (r ¢) · ¢G (r , r ¢)
S
- G (r , r ¢) nˆ (r ¢) · ¢F+(r ¢)] d 2r ¢, r Î Vout. (8b)

3
J. Phys. Commun. 5 (2021) 115010 A Lakhtakia et al

With the usual requirement of the electric field having a finite magnitude everywhere on S, the potential
must be continuous across that interface; hence,
F+(r) = F-(r) , r Î S, (9a)
where F-(r) is the value of F (r) on the interior side of S. Likewise, with the assumption of S being charge-free,
the normal component of the electric displacement must be continuous across that surface; hence,
nˆ (r) · F+(r) = nˆ (r) · e rel · F-(r) , r Î S. (9b)
Substitution of equations (9a) and (9b) in equations (8a) and (8b) delivers
Fso (r) ⎫
=  ∬ [F-(r ¢) nˆ (r ¢) · ¢G (r , r ¢)
Fpert (r)⎬
⎭
S

- G (r , r ¢) nˆ (r ¢) · e rel · ¢F-(r ¢)] d 2r ¢, ⎧ r Î Vin , (10)

⎩ r Î Vout.
⎨
Since Φ−(r) at r ä S is nothing but the internal potential Φint(r) evaluated on the interior side of S, we finally get
Fso (r) ⎫
=  ∬ [Fint (r ¢) nˆ (r ¢) · ¢G (r , r ¢)
Fpert (r)⎬
⎭
S

- G (r , r ¢) nˆ (r ¢) · e rel · ¢Fint (r ¢)] d 2r ¢, ⎧ r Î Vin , (11)

⎩ r Î Vout.
⎨
Knowing Φso, we first have to find Φint everywhere in Vin after choosing r ä Vin in equations (11). Thereafter,
knowing Φint, we can find Φpert everywhere in Vout after choosing r ä Vout in equations (11). Thus, the
electrostatic counterpart of the Ewald–Oseen formulation is a crucial ingredient in our EBCM formulation,
which follows the style of the Waterman formulation for electromagnetics [15, 16].

4. Extended boundary condition method

4.1. Source and perturbation potentials

The solutions of the Laplace equation in the spherical coordinate system being known [27], the source potential
can be represented as
¥ n
Fso (r) = å å å [Emn  smn r n Ysmn (q, f)] , r < rso. (12)
s Î {e , o}n = 0 m = 0

Here, the expansion coefficients  smn are supposed to be known,

2n + 1 (n - m)!
Emn = (2 - dm0 ) (13)
4p (n + m)!
is a normalization factor with dmm¢ as the Kronecker delta, and the spherical harmonics
Yemn (q , f) = Pnm (cos q ) cos (mf)⎫
(14)
Yomn (q , f) = Pnm (cos q ) sin (mf) ⎬
⎭
involve the associated Legendre function Pnm (cos q ) [27]. The definitions of the spherical harmonics mandate
that  o0n = 0 " n Î [0, ¥). Likewise, the perturbation potential can be represented as
¥ n
Fpert (r) = å å å [Emn smnr -(n+ 1) Ysmn (q, f)] , r  rout, (15)
s Î {e , o}n = 0 m = 0

with unknown expansion coefficients smn , where o0n = 0 " n Î [0, ¥).

4.2. Algebraic equations

The Green function G (r, r ¢) has the bilinear expansion [27]

G (r , r ¢) = ås Î {e, o}å ¥ n
n= 0 åm = 0
⎡ Emn Ysmn (q , f) Ysmn (q ¢, f ¢)
⎣ 2n + 1
-(n + 1)
⎧ r (r ¢) ⎧ r < r ¢,
n
⎫ ⎤, (16)
- + ) (r )n ⎬ ⎥
¢ ⎩ r > r ¢.
⎨ r (n 1 ⎨
⎩ ⎭⎦
Now, we use the orthogonality relationships of the spherical harmonics on spherical surfaces. First, let the
surface r = ain < rin be chosen for applying equation (11)top. The source potential on the left side of this equation
can be replaced by the series on the right side of equation (12). The bilinear expansion of G (r, r ¢) for r < r ¢ has

4
J. Phys. Commun. 5 (2021) 115010 A Lakhtakia et al

to be used on the right side of equation (11)top, since r = ain and r ¢ Î S. After multiplying both sides of the
resulting equation by Ys¢m¢n¢ (q, f ) and integrating over the surface r = ain, we obtain
1
 smn = - ∬ {Fint (r ¢) nˆ (r ¢) · ¢ [(r ¢)-(n+ 1) Ysmn (q ¢, f ¢)]
2n + 1 S
- (r ¢)-(n + 1) Ysmn (q ¢, f ¢) nˆ (r ¢) · e rel · ¢Fint (r ¢)} d 2r ¢. (17a)
Second, we choose the surface r = aout > rout for applying equation (11)bot. The perturbation potential on
the left side of this equation can be replaced by the series on the right side of equation (15). On the right side of
equation (11)bot the bilinear expansion of G (r, r ¢) for r > r ¢ must be used, since r = aout and r ¢ Î S. Then, after
multiplying both sides of the resulting equation by Ys¢m¢n¢ (q, f ) and integrating over the surface r = aout, we get
1
smn = ∬ {Fint (r ¢) nˆ (r ¢) · ¢ [(r ¢)n Ysmn (q ¢, f ¢)]
2n + 1 S
- (r ¢)n Ysmn (q ¢, f ¢) nˆ (r ¢) · e rel · ¢Fint (r ¢)} d 2r ¢. (17b)

The integral equations (17a) and (17b) incorporate the boundary conditions (9a) and (9b) prevailing on the
surface S of the perturbing object Vin, but the orthogonalities of the spherical harmonics were not applied on
that surface. This is the reason for ‘extended boundary condition’ in the name of EBCM [14, 15].

4.3. Internal potential

The internal potential Φint(r) satisfies the equation
 · { e · [Fint (r)]} = 0, r Î Vin , (18)
which emerges from the Gauss equation for the electric field. Equation (18) is not the Laplace equation, although
it does simplify to the Laplace equation when αx = αy = 1 (i.e., for an isotropic medium [26, 28–30]).
After applying a bijective affine transformation of space that maps a sphere into an ellipsoid, the solution of
equation (18) can be written as the series [31]
¥ n
Fint (r) = å å å [ smnZsmn (r)] , r Î Vin , (19)
s Î {e , o}n = 0 m = 0

where

r ay ⎫
Z emn (r) = ∣ A-1 · r∣n Pnm ⎡ -1 cos q⎤ cos ⎡m tan-1 ⎛ tan f⎞ ⎤⎪ ⎜ ⎟
⎢ ∣ A · r∣
⎣
⎥
⎦
⎢
⎣ a
⎝ x ⎥
⎠ ⎦⎪
. (20)
- m⎡ r ⎤ ⎡ - ⎛ ay ⎞ ⎤
⎬
Z omn (r) = ∣ A · r∣ Pn
1 n cos q sin m tan 1 tan f ⎪ ⎜ ⎟
⎢ ∣ A-1 · r∣ ⎥ ⎢ ⎝ ax ⎠⎥
⎣ ⎦ ⎣ ⎦⎪⎭
Since the basis formed by the functions Zsmn(r) is constructed by a spatial transformation of the eigenfunctions
of the Laplace equation in the spherical coordinate system, care must be taken to ensure that the angle
r
qq = cos-1 ⎛⎜ -1 cos q ⎞⎟ (21a)
⎝ ∣ A · r∣ ⎠
lies in the same quadrant as θ, and the angle
ay
fq = tan-1 ⎛ tan f⎞ ⎜ ⎟ (21b)
⎝ ax ⎠
lies in the same quadrant as f.

4.4. Transition matrix

Substitution of equation (19) in equation (17a) delivers the algebraic equation
¥ n¢
 smn = - å å å ( (ss1¢)mm¢nn¢  s¢m¢n¢) , (22)
s ¢Î {e, o}n ¢= 0 m ¢= 0

where the surface integral

1
 (ss1¢)mm¢nn¢ = ∬ {nˆ (r ¢) · ¢ [(r ¢)-(n+ 1) Ysmn (q ¢, f ¢)] Z s¢m¢n¢ (r ¢)
2n + 1 S
- (r ¢)-(n + 1) Ysmn (q ¢, f ¢) nˆ (r ¢) · e rel · ¢Z s¢m¢n¢ (r ¢)} d 2r ¢. (23)

5
J. Phys. Commun. 5 (2021) 115010 A Lakhtakia et al

Likewise, substitution of equation (19) in equation (17b) leads to

¥ n¢
smn = å å å ( (ss3¢)mm¢nn¢  s¢m¢n¢) , (24)
s ¢Î {e, o}n ¢= 0 m ¢= 0

where the surface integral

1
 (ss3¢)mm¢nn¢ = ∬ {nˆ (r ¢) · ¢ [(r ¢)n Ysmn (q ¢, f ¢)] Z s¢m¢n¢ (r ¢)
2n + 1 S
- (r ¢)n Ysmn (q ¢, f ¢) nˆ (r ¢) · e rel · ¢Z s¢m¢n¢ (r ¢)} d 2r ¢. (25)
Clearly, the dependence of Φpert(r) on the shape and size of Vin is coded in the surface integrals  (ss1¢)mm¢nn¢
and  (ss3¢)mm¢nn¢ .
After the indexes n and n¢ in the foregoing equations are restricted to the range [0, N], equation (22) leads to
the matrix equation (in abbreviated notation)
 = -  (1) · , (26)
whereas equation (24) yields the matrix equation
 =  (3) · . (27)
Then,
 =  · , (28)
where the transition matrix
 = -  (3) · ( (1))-1 (29)
relates the expansion coefficients of the perturbation potential to those of the source potential. Very importantly,
this matrix does not depend on the source potential.

4.5. Asymptotic expression for perturbation potential

The perturbation potential defined in equation (15) can be written as
1 (1) 1 (2)
Fpert (r) = f pert + 2 f pert (q , f )
r r
N
1 n
+ lim ås Î {e, o} å n + 1 å [Emn smn Ysmn (q , f)] , r  rout, (30a)
N ¥ n= 2 r m=0

where
1
(1)
f pert = e00 (30b)
4p
and
3
(2)
f pert (q , f ) =
[e01 cos q + 2(e11 cos f + o11 sin f) sin q ]. (30c )
4p
Thus, far away from the object, the perturbation potential can be asymptotically stated as
1 (1) 1 (2) 1
Fpert (r) = f + 2 f pert (q , f ) +  ⎛ 3 ⎞ , r  ¥ , (31)
r pert r ⎝r ⎠
and it is determined only by the four perturbational-potential coefficients e00 , e01, e11, and o11.
We have explicitly retained the two lowest-order terms in equation (31), because f pert (1)
is not always nonzero.
For instance, e00 = 0 when the source potential varies linearly in a fixed direction and object is axisymmetric as
well as composed of an isotropic medium [26, 28, 30]. Likewise, e00 = 0 when the object is a sphere and the
(1)
source is either a point charge or a point dipole, so that f pert = 0 and Φpert(r) ∝ r−2 as r → ∞ [31]. More
(1) (2) (1)
generally, the distance after which f pert dominates f pert can depend on the direction, because f pert is
independent of θ and f whereas f pert
(2)
is not. Indeed, the first term on the right side of equation (30c) does not
exist for θ = π/2, whereas the second and third terms on the right side of the same equation are absent
for q Î {0, p}.

5. Numerical results and discussion

By virtue of equations (23), (25), (28), and (29), the perturbation potential must depend on (i) the geometry of
the perturbing object described via S; (ii) the composition of the object, quantitated by e rel ; and (iii) the spatial

6
J. Phys. Commun. 5 (2021) 115010 A Lakhtakia et al

profile of the source potential Φso. We present numerical results in this section to illustrate the effects of S, e rel ,
and Φso on the perturbation potential Φpert.

5.1. Preliminaries
5.1.1. Geometry
We chose S to be an ellipsoidal surface defined by the position vector
r (q , f) = a S · U · S-1 · [(xˆ cos f + yˆ sin f) sin q + zˆ cos q ] ,
q Î [0, p ] , f Î [0, 2p ). (32)
The rotation dyadic
S = R z (gs) · R y (bs) · R z (as) (33a)
is a product of three rotation dyadics, with
R z (z ) = (xx
ˆ ˆ + yy
ˆ ˆ ) cos z - (xy
ˆ ˆ - yx
ˆ ˆ) sin z + zz
ˆ ˆ , z Î {as , gs}, (33b)
and
R y (bs) = (xx
ˆ ˆ + zz
ˆ ˆ) cos bs - (ˆzxˆ - xz
ˆ ˆ) sin bs + yy
ˆ ˆ. (33c )

Sequentially, R z (as ) represents a rotation by αs ä [0, π] about the z axis, R y (bs ) represents a rotation by βs ä [0, π]
about the new y axis, and R z (gs ) represents a rotation by γs ä [0, π] around the new z axis. The shape dyadic is given
by
U = m xx
ˆ ˆ + n yy
ˆ ˆ + zz.
ˆˆ (34)
Thus, the shape principal axes of the ellipsoidal object are parallel to the unit vectors S · xˆ , S · yˆ , and S · zˆ .
The semi-axes of this object are given by aμ parallel to S · xˆ , aν parallel to S · yˆ , and a parallel to S · zˆ , where
μ > 0 and ν > 0 are the two aspect ratios of the ellipsoid; thus, rout = max {a , am, an}. Finally, with the
geometric mean aave of the three semi-major axes fixed, the length
a = a ave (mn )-1 3 (35)
becomes a function of the aspect ratios μ and ν. Note that a = aave when μ = ν = 1 (i.e., the ellipsoid reduces to a
sphere).

5.1.2. Composition
With the arithmetic mean εave of the three eigenvalues of e rel fixed, the relative permittivity scalar
3 eave
er = -2 (36)
a x + a-y 2 + 1
is a function of the anisotropy parameters αx and αy. Note that εr = εave when αx = αy = 1 (i.e., the material
becomes isotropic).

5.1.3. Source
Sources of two types were considered for illustrative numerical results: (i) a point-charge source and (ii) a point-
dipole source. The coefficient [24]
Q 1
 smn = ro-(n + 1) Ysmn (qo, fo) (37)
e0 2n + 1
for a point charge Q located at ro ≡ (ro, θo, fo) with ro … rout, θo ä [0, π], and fo ä [0, 2π). For a point dipole
p = p p̂ located at the same point, the coefficient [37]
p 1
 smn = pˆ · o [ro-(n + 1) Ysmn (qo, fo)] , (38)
e0 2n + 1
where p > 0; the unit vector
pˆ = (xˆ cos fp + yˆ sin fp) sin qp + zˆ cos qp (39)
contains the angles θp ä [0, π] and fp ä [0, 2π) that define the orientation of the dipole, and ∇o(...) denotes the
gradient with respect to ro.

5.2. Convergence
A MathematicaTM program was written to calculate the transition matrix  of an anisotropic ellipsoid for a
chosen N. The column vector  containing the perturbation-potential coefficients was then calculated using
equation (28). Also, we determined the spatial profile of Φpert(r, θ, f).

7
J. Phys. Commun. 5 (2021) 115010 A Lakhtakia et al

Figure 2. Φpert(r, θ, 0) versus θ for a prolate spheroid when aave = 3.82 cm, μ = ν = 2/3, εave = 3, and αx = αy = 1. The source is a
point charge with Q = (10-9 36p ) C located at ro = 2a, θo = 0. (a) r = 1.1a, (b) r = 2a, (c) r = 4a, and (d) r = 10a.

Figure 3. Fpert (˜r rout, q, f ) versus θ and f when aave = 5 cm, μ = 0.8, ν = 1.2, εave = 3, αx = 0.5, αy = 1.5, and αs = βs = γs = 0.
The source is a point charge Q = 10−10 C located at ro = 2rout, θo = π/4, and fo = π/6. (a) r˜ = 1.1, (b) r˜ = 2, (c) r˜ = 4 , and (d)
r˜ = 10 .

A convergence test was carried out with respect to N, by calculating the integral
2p p
I (r ) = òf=0 òq=0 F2pert (r, q, f) sin q dq df, (40)

8
J. Phys. Commun. 5 (2021) 115010 A Lakhtakia et al

Figure 4. Same as figure 3, except that αs = 2π/3, βs = 3π/4 and γs = 5π/9.

at diverse values of r ä [1, 10] rout as N was incremented by unity. The iterative process of increasing N was
terminated when I(r) converged within a preset tolerance of 1% for a specific value of r. The adequate value of N
was higher for lower r, with N = 7 sufficient for r … 1.1rout.

5.3. Code validation

Validation of the EBCM code was done in two steps. First, the results for the perturbation of the potential of a
point charge by a prolate spheroid (i.e., μ = ν < 1) composed of an isotropic dielectric medium were compared
with the corresponding series solution obtained using the eigenfunctions of the Laplace equation in the prolate-
spheroidal coordinate system [38].
Figure 2 shows plots of the perturbation potential Φpert(r, θ, 0) evaluated at r˜ = r rout Î {1.1, 2, 4, 10}.
These plots were obtained using the series solution [38] and our EBCM when the source is a point charge
Q = (10-9 36p ) C located at ro = 2rout and θo = 0 (with fo being irrelevant as sin qo = 0), whereas the
perturbing object is characterized by aave = 3.82 cm, μ = ν = 2/3, εave = 3, and αx = αy = 1. Good agreement
can be seen in figure 2(a) between the two solutions when r˜ = 1.1, except in the neighborhood of θ = 0. The
difference arises because the spherical harmonics used for representing Φpert and Φint do not follow the shape of
the prolate spheroid well away from the equator, an issue associated with EBCM even for time-harmonic
problems [39, 40]. However, the difference diminishes as r̃ increases. Indeed, the difference is vanishingly small
for r˜ = 10 in figure 2(d).
Second, the same comparison was done for a sphere (i.e., μ = ν = 1) made of an anisotropic dielectric
medium, the series solution for a sphere being available using the eigenfunctions of the Laplace equation in the
spherical coordinate system [31]. The difference in values of Φpert(r) calculated using the two methods for any
r … a was always less than 0.001%.

5.4. Numerical results for anisotropic dielectric ellipsoids

Now we present the perturbation potential’s variations with respect to r̃ and S when the source is either a point
charge or a point dipole. For definiteness, we fixed aave = 5 cm, μ = 0.8, ν = 1.2, εave = 3, αx = 0.5 and αy = 1.5.
Also, the point source was located at ro = 2rout, θo = π/4 and fo = π/6.

9
J. Phys. Commun. 5 (2021) 115010 A Lakhtakia et al

Figure 5. Same as figure 3, except for a point-dipole source with p = 10−10 C m. Also, θp = π/4 and fp = π/3.

5.4.1. Point-charge source

In order to focus on the effect of r̃ exclusively, we set S = I (i.e., the shape principal axes coincide with the
constitutive principal axes). Figures 3(a)–(d) present the angular profiles of Fpert (˜r rout, q, f ) for
r˜ Î {1.1, 2, 4, 10} for a point-charge source with Q = 10−10 C, when αs = βs = γs = 0. The perturbation
potential Fpert (˜r rout, q, f ) decreases with increase of r̃ , as becomes evident by comparing figures 3(a)–(d). The
angular profiles vary significantly for r˜  2, but they change very little as r˜ - 2 increases. We have observed
through diverse computations (results not shown) that the foregoing observation is valid even when ro exceeds
2rout.
Next, we considered the effect of S ¹ I (i.e., the shape principal axes are rotated with respect to the
constitutive principal axes). We repeated the calculations of the previous case when αs = 2π/3, βs = 3π/4, and
γs = 5π/9; the results are depicted in figure 4. By comparing figures 3 with 4, it can be seen that the fact S ¹ I
alters the rate of increase/decrease of Φpert(r, θ, f) in any specific direction. Furthermore, the locations of the
extremums of Φpert(r, θ, f) in the θf-plane are affected by the difference S - I when r˜ < 2; see figures 3(a) and
4(a), for instance. The impact of S - I diminishes as r˜ - 2 increases, as can be seen on comparing figures 3(c)
and 4(c) for instance.

5.4.2. Point-dipole source

Figures 5(a)–(d) present the angular profiles of Fpert (˜r rout, q, f ) for r˜ Î {1.1, 2, 4, 10} for a point-dipole
source with p = 10−10 C m, θp = π/4 and fp = π/3 when S = I . Just as with the point-charge source in
figures 3(a–d), Fpert (˜r rout, q, f ) decreases with increase of r;
˜ however, the rate of decrease for the point-dipole
source is smaller compared to that for the point-charge source. The effect of the type of source is visually evident
on comparing figures 3 and 5 for the locations of the extremums of Φpert(r, θ, f) in the θf-plane.
Figures 6(a)–(d) present the same angular profiles as figures 5(a)–(d), except that αs = 2π/3, βs = 3π/4 and
γs = 5π/9 (i.e., S ¹ I ). After comparing figures 5 and 6, we concluded that (similarly to figure 4) S ¹ I alters
the increase/decrease rate of Φpert(r, θ, f). As in section 5.4.1, the locations of the extremums of Φpert(r, θ, f) in
the θf-plane are mainly affected by the difference S - I when r˜ < 2.

10
J. Phys. Commun. 5 (2021) 115010 A Lakhtakia et al

Figure 6. Same as figure 5, except when αs = 2π/3, βs = 3π/4 and γs = 5π/9.

6. Concluding remarks

We have formulated the EBCM for the perturbation of a source electric potential by a 3D object composed of a
homogeneous anisotropic dielectric medium whose relative permittivity dyadic is positive definite. The
electrostatic counterpart of the Ewald–Oseen extinction theorem was derived for that purpose, and the electric
potential inside the object was represented using a basis obtained by implementing an affine bijective
transformation of space to the Gauss equation for the electric field. This formulation is different from the
Farafonov formulation [26, 28–30], which is applicable only when the object is composed of a homogeneous
isotropic medium.
Even though the object is nonspherical, all potentials were represented in terms of the eigenfunctions of the
Laplace equation in the spherical coordinate system. Numerical results have shown that our EBCM formulation
fails to yield convergent results when the shape of the perturbing object deviates too much from spherical (i.e.,
when either |μ − 1| is substantial and/or |ν − 1| is substantial). The same problem bedevils the analogous
formulation of the EBCM even for time-harmonic problems [39, 40].
As the Laplace equation is either separable or R-separable in several coordinate systems [41], we plan to use
bases in future work that will conform better to the shape of the object. A bijective affine transformation of space
[31] will be applied to map each basis for the Laplace equation into a basis for equation (18). Furthermore, the
partitioning of Vin into several subregions is also possible, with different bases used in different subregions
[39, 40, 42–45].

Acknowledgments

AL is grateful to the Charles Godfrey Binder Endowment for ongoing support of his research activities. H M
Alkhoori acknowledges the United Arab Emirates University for funding this work.

11
J. Phys. Commun. 5 (2021) 115010 A Lakhtakia et al

Data availability statement

The data that support the findings of this study are available upon reasonable request from the authors.

ORCID iDs

Akhlesh Lakhtakia https://orcid.org/0000-0002-2179-2313

Hamad M Alkhoori https://orcid.org/0000-0002-8793-1769
Nikolaos L Tsitsas https://orcid.org/0000-0003-1409-2631

References
[1] Pierce A D 1989 Acoustics: An Introduction to Its Physical Principles and Applications (Woodbury, NY, USA: Acoustical Society of
America)
[2] de Hoop A T 1995 Handbook of Radiation and Scattering of Waves (San Diego, CA, USA: Academic)
[3] Varadan V V, Lakhtakia A and Varadan V K (ed) 1991 Field Representations and Introduction to Scattering (Amsterdam, The
Netherlands: Elsevier)
[4] Felsen L B and Marcuvitz N 1973 Radiation and Scattering of Waves (Upper Saddle River, NJ, USA: Prentice-Hall)
[5] Chew W C 1990 Waves and Fields in Inhomogeneous Media (New York, NY, USA: Van Nostrand Reinhold)
[6] Varadan V K and Varadan V V (ed) 1982 Elastic Wave Scattering (Ann Arbor, MI, USA: Ann Arbor Science)
[7] Happel J and Brenner H 1973 Low Reynolds Number Hydrodynamics (The Netherlands: Noordhoff, Groningen)
[8] Nečasová Š and Kračmar S 2016 Navier-Stokes Flow Around a Rotating Obstacle (Paris, France: Atlantis Press)
[9] Ewald P P 1916 Zur Begründung der Kristalloptik Ann. Phys. 49 117–43
[10] Oseen C W 1915 Über die Wechselwirkung zwischen zwei elektrischen Dipolen und über die Drehung der Polarisationsebene in
Kristallen und Flüssigkeiten Ann. Phys. 48 1–56
[11] Waterman P C 1969 New formulation of acoustic scattering J. Acoust. Soc. Am. 45 1417–29
[12] Varatharajulu V and Pao Y-H 1976 Scattering matrix for elastic waves. I. Theory J. Acoust. Soc. Am. 60 556–66
[13] Wymer S A, Lakhtakia A and Engel R S 1996 The Huygens principle for flow around an arbitrary body in a viscous incompressible fluid
Fluid Dynam. Res. 17 213–23
[14] Lakhtakia A 2018 The Ewald-Oseen extinction theorem and the extended boundary condition method The World of Applied
Electromagnetics ed A Lakhtakia and C M Furse (Cham, Switzerland: Springer) 481–513
[15] Waterman P C 1965 Matrix formulation of electromagnetic scattering Proc. IEEE 53 805–12
[16] Waterman P C 1969 Scattering by dielectric obstacles Alta Freq. (Speciale) 38 348–52
[17] Faryad M and Lakhtakia A 2018 Infinite-Space Dyadic Green Functions in Electromagnetism (Bristol, United Kingdom: Institute of
Physics) Chap. 4
[18] Lin Z and Chui S T 2004 Electromagnetic scattering by optically anisotropic magnetic particle Phys. Rev. E 69 056614
[19] Schmidt V and Wriedt T 2012 The T-matrix for particle with arbitrary permittivity tensor and parallelization of the computational
code J. Quant. Spectrosc. Radiat. Transf. 113 1712–8
[20] Zouros G P, Kolezas G D, Stefanou N and Wriedt T 2021 EBCM for electromagnetic modeling of gyrotropic BoRs IEEE Trans.
Antennas Propagat. 69 6134–9
[21] Doicu A, Wriedt T and Khebbache N 2019 An overview of the methods for deriving recurrence relations for T -matrix calculation
J. Quant. Spectrosc. Radiat. Transf. 224 1289–302
[22] Alkhoori H M, Lakhtakia A, Breakall J K and Bohren C F 2019 Plane-wave scattering by an ellipsoid composed of an orthorhombic
dielectric-magnetic material with arbitrarily oriented constitutive principal axes J. Opt. Soc. Am. B 36 F60–71
[23] Kahnert M and Rother T 2020 Convergence of the iterative T-matrix method Opt. Express 28 28269–82
[24] Jackson J D 1999 Classical Electrodynamics III edn (Hoboken, NJ, USA: Wiley) Sec. 1.8
[25] Colton D and Kress R 2013 Integral Equation Methods in Scattering Theory (Philadelphia, PA, USA: SIAM) Theorem 3.1
[26] Farafonov V G 2014 The Rayleigh hypothesis and the region of applicability of the extended boundary condition method in
electrostatic problems for nonspherical particles Opt. Spectrosc. 117 923–35
[27] Morse P M and Feshbach H 1953 Methods of Theoretical Physics vol 2 (New York, NY, USA: McGraw-Hill) pp 1264–74
[28] Farafonov V G and Ustimov V I 2015 Analysis of the extended boundary condition method: an electrostatic problem for Chebyshev
particles Opt. Spectrosc. 118 445–59
[29] Farafonov V G, Il’in V, Ustimov V I and Prokopjeva M 2016 On the analysis of Waterman’s approach in the electrostatic case J. Quant.
Spectrosc. Radiat. Transf. 178 176–91
[30] Majić M R A, Gray F, Auguié B and Le Ru E C 2017 Electrostatic limit of the T-matrix for electromagnetic scattering: Exact results for
spheroidal particles J. Quant. Spectrosc. Radiat. Transf. 200 50–8
[31] Lakhtakia A, Tsitsas N L and Alkhoori H M 2021 Theory of perturbation of electrostatic field by an anisotropic dielectric sphere Quart.
J. Mech. Appl. Math. hbab013
[32] Charnow A and Charnow E 1986 Fields for which the principal axis theorem is valid Math. Mag. 59 222–5
[33] Strang G 2016 Introduction to Linear Algebra V edn (Wellesley, MA, USA: Wellesley-Cambridge) p 339
[34] Auld B A 1990 Acoustic Fields and Waves in Solids, Vol I II edn (Malabar, FL, USA: Krieger) p 387
[35] Mackay T G and Lakhtakia A 2020 Modern Analytical Electromagnetic Homogenization with Mathematica® II edn (Bristol, United
Kingdom: Institute of Physics) Sec. 6.5
[36] Smythe W R 1950 Static and Dynamic Electricity II edn (New York, NY, USA: McGraw-Hill) Sec. 3.09
[37] Tsitsas N L and Martin P A 2012 Finding a source inside a sphere Inverse Prob. 28 015003
[38] Redžić D V 1994 An electrostatic problem: a point charge outside a prolate dielectric spheroid Am. J. Phys. 62 1118–21
[39] Iskander M F, Lakhtakia A and Durney C H 1983 A new procedure for improving the solution stability and extending the frequency
range of the EBCM IEEE Trans. Antennas Propagat. 31 317–24
[40] Lakhtakia A, Varadan V K and Varadan V V 1984 Iterative extended boundary condition method for scattering by objects of high aspect
ratios J. Acoust. Soc. Am. 76 906–12

12
J. Phys. Commun. 5 (2021) 115010 A Lakhtakia et al

[41] Moon P and Spencer D E 1952 Separability conditions for the Laplace and Helmholtz equations J. Franklin Inst. 253 585–600
[42] Lakhtakia A, Iskander M F and Durney C H 1983 An iterative extended boundary condition method for solving the absorption
characteristics of lossy dielectric objects of large aspect ratios IEEE Trans. Microwave Theory Tech. 31 640–7
[43] Lakhtakia A, Varadan V K and Varadan V V 1984 Iterative extended boundary condition method for scattering by objects of high aspect
ratios J. Acoust. Soc. Am. 76 906–12
[44] Ballisti R and Hafner C 1983 The multiple multipole method (MMP) in electro- and magnetostatic problems IEEE Trans. Magnetics 19
2367–70
[45] Hafner C 1990 The Generalized Multipole Technique for Computational Electromagnetics (Boston, MA, USA: Artech House)

13