Scattering of a Multi-layer Sphere

G. Kevin Zhu

Created on: September 9, 2009

Modified on: December 18, 2018

1 Problem Statement and Solution Overview

The problem of a plane wave scattered by an L-layer medium is a boundary value problem in solving partial
differential equations [1, Ch. 3.7].
Fig. 1 shows a plane wave polarized in the +x direction and propagating in the +z direction. The plane
wave is scattered by an L-layer sphere. The centre of the sphere coincides with the origin of the spherical
coordinate system.
Solving such problems includes the following steps:
1. Selecting a coordinate system so that the boundaries between layers conform to some of the principal
axes of the coordinate system.
2. Selecting some potentials, which convert the vector Maxwell equations to the scalar Helmholtz equation.
3. Solving the Helmholtz equation using the method of separation of variables, which decomposes it into
three linear ordinary differential equations.
4. Applying the homogeneous boundary condition to solve for the eigen functions of the ordinary differ-
ential equations.
5. Expressing the incident plane wave in terms of potentials, which are in series forms of the products of
the eigen functions.
6. Expressing the scattered potentials and the potentials inside each layer in the series forms with some
unknown coefficients.
7. Applying the conditions at the boundaries between layers, which leads to a system of linear equations.
8. Solving this linear system gives the coefficients of the potentials.
9. Expressing the fields in terms of the potentials.

R(1)
r
R(L)
θ
y
φ

Figure 1: An x-polarized plane wave, propagating in the z direction, impinges upon an L-layer sphere.
Starting from the inner core, each layer is characterized by its radius, permittivity, and permeability.

1
2 Potentials and Construction of Solutions
The purpose of using potentials instead of the electric and magnetic field vectors is that the Maxwell equations
expressed in terms of potentials become scalar equations. This reduces the number of unknowns to be
calculated.1 Here, we follow the approach in [4, 129] to our problem, in which the magnetic and electric
vector potentials are introduced to the scattering problem of a dielectric sphere. The magnetic vector
~ and the electric vector potential F~ are defined so that
potential A

∇×A ~ = µH ~ ,
∇ × F~ = − E
~ ,

where  and µ denote the permittivity and permeability. We assume ejωt to be the time-harmonic factor.2
The Maxwell equations can be equivalently expressed in terms of the vector potentials as

∇×∇×A ~ − k2 A ~ = jωµ∇Φe , (1a)

∇ × ∇ × F~ − k 2 F~ = jωµ∇Φm , (1b)
√
where k is the wave number and k = ω µ, and ω is the angular frequency. Φm and Φe are some arbitrary
scalar field quantities that will be discussed shortly. The electric and magnetic field can be expressed in
terms of the potentials as

~ = − 1 ∇ × F~ + 1 ∇ × ∇ × A
E ~ , (2a)
 jωµ
~ = 1∇×A
H ~ + 1 ∇ × ∇ × F~ . (2b)
µ jωµ
A convenient choice of the vector potentials in the scattering problems in a spherical coordinate system is
~ = Ar~ar and F~ = Fr~ar . Substituting this choice to (1a) leads to
to let A

1 ∂ ∂Ar 1 ∂ 2 Ar 2 ∂Φe
− sin θ − 2 − k A r = −jωµ , (3a)
r2 sin θ ∂θ ∂θ r2 sin θ ∂φ2 ∂r
∂ 2 Ar ∂Φe
= −jωµ , (3b)
∂r∂θ ∂θ
2 e
∂ Ar ∂Φ
= −jωµ . (3c)
∂r∂φ ∂φ
Because the choice of Φe is arbitrary, we set
∂Ar
= −jωµΦe , (4)
∂r
1 The choice of the potentials is not unique as their divergence can be freely defined. The Debye potential is used for scattering

problems. The Hertzian potential is often for radiation problems [2, 3].
2 Some texts [5, 6] use e−jωt as the time-harmonic factor due to the preference to represent a wave propagating in the z

direction as ejkz . To cross check solutions based on these two conventions, it should be noted that the field is a physical
quantity. The same wave propagating in the +z direction can be expressed as <{Aejkz e−jωt } or <{Āe−jkz ejωt }, where (·) ¯
denotes the conjugate. Thus, the phasors expressed in these two different conventions are conjugate pairs.

2
 Ar
to satisfy both (3b) and (3c), and (3a) becomes the scalar Helmholtz equation of r , which is discussed in
the next section. Eq. (2) becomes
 2 
1 ∂ 2
Er = + k Ar , (5a)
jωµ ∂r2
1 1 ∂2 1 1 ∂
Eθ = Ar − Fr , (5b)
jωµ r ∂r∂θ  r sin θ ∂φ
1 1 ∂2 11 ∂
Eφ = Ar + Fr , (5c)
jωµ r sin θ ∂r∂φ  r ∂θ
 2 
1 ∂
Hr = + k 2 Fr , (5d)
jωµ ∂r2
1 1 ∂ 1 1 ∂2
Hθ = Ar + Fr , (5e)
µ r sin θ ∂φ jωµ r ∂r∂θ
11 ∂ 1 1 ∂2
Hφ = − Ar + Fr . (5f)
µ r ∂θ jωµ r sin θ ∂r∂φ

3 Helmholtz’s Equation in a Spherical Coordinate System

The Helmholtz equation of a scalar quantity Φ, in the spherical coordinate system, is

∂2Φ
   
1 ∂ 2 ∂Φ 1 ∂ ∂Φ 1
r + sin θ + + k2 Φ = 0 . (6)
r2 ∂r ∂r r2 sin θ ∂θ ∂θ r2 sin2 θ ∂φ2
It is a linear partial differential equation and is solved by the method of separation of variables. Let
Φ(r, θ, φ) = A(r)B(θ)C(φ). Substituting it to (6), we obtain three linear ordinary differential equations,
which are
 
d dA
r2 + [(kr)2 − n(n + 1)]A = 0 , (7a)
dr dr
m2
 
1 d dB
sin θ + [n(n + 1) − ]B = 0 , (7b)
sin θ dθ dθ sin2 θ
d2 C
+ m2 C = 0 , (7c)
dφ2
where n and m are non-negative integers and m ≤ n. The operators associated with these differential
equations are of the Sturm-Liouville type [7, ch. 5]; they are self-adjoint operators and their eigen functions
are orthogonal to each other. The eigen functions form a complete set; any piecewise smooth function can
be represented by a linear combination of these eigen functions.3
If we let x = kr, eq. (7a) becomes the spherical Bessel equation. Its eigen functions are known as the
family of the spherical Bessel functions, denoted as bn (x). Within this family, the ordinary spherical Bessel
function is denoted as jn (x), representing the standing wave. The spherical Hankel function of the first and
(1) (2)
second kind are denoted as hn (x) and hn (x) representing the inward and outward propagating waves. If
we let x = cos θ, eq. (7b) is known as the associated Legendre equation and one of its eigen functions is the
associated Legendre polynomial denoted as Pnm (x). When m = 0, we have the ordinary Legendre polynomial
denoted as Pn (x). Eq. (7c) is known as the Harmonic equation and its eigen functions are e±jmφ . Any
solution to (6) can be expressed as the infinite sum of the products of these eigen functions, expressed as
XX
cnm bn (x)|x=kr Pnm (x)|x=cos θ e±jmφ ,
n m

where cnm denotes some coefficient.

3 This explains why so many functions are supported by the wave equation to propagate freely in space.

3
4 Wave Transformation
A plane electromagnetic wave polarized in the x direction propagating in the z direction is expressed in terms
of the eigen functions in the Cartesian coordinate system.4 The electric part of the plane wave is expressed
as
~ inc = e−jkz~ax ,
E (8)

where the magnitude is 1 V/m. The magnetic part of the plane wave is expressed as
inc
~ inc = Ex e−jkz~ay .
H (9)
η
Expressing electric part of the plane wave in the form of the product of the eigen functions gives
∞
X
~ inc =
E j −n (2n + 1)jn (x)|x=kr Pn (x)|x=cos θ ~ax (10a)
n=0
=Er~ar + Eθ~aθ + Eφ~aφ (10b)
∞
X
= j −n (2n + 1)jn (x)|x=kr Pn (x)|x=cos θ (sin θ cos φ~ar + cos θ cos φ~aθ − sin φ~aφ ) . (10c)
n=0

The conversion from (8) to (10a) is known as the wave transformation and (10c) is the expression in the
spherical coordinate system.
Since the component in the direction of ~ar in (10c) solely depends on Ar shown in (5a), substituting the
component in the direction of ~ar in (10c) to (5a) gives
∞
cos φ X −n 2n + 1 ˆ
Ainc
r = j Jn (x)|x=kr Pn1 (x)|x=cos θ , (11)
ω n=1 n(n + 1)

(1) (1) (2) (2)

where we define B̂n (x) = xbn (x).5 Thus, Jˆn (x) = xjn (x), Ĥn (x) = xhn (x), and Ĥn (x) = xhn (x).
Similarly, expressing the magnetic part of the plane wave in the form of the product of the eigen functions
gives
∞
sin φ X −n 2n + 1 ˆ
Frinc = j Jn (x)|x=kr Pn1 (x)|x=cos θ . (12)
ωη n=1 n(n + 1)

The fields due to Ainc

r are known as the TM fields. The fields due to Frinc are known as the TE fields.

5 Expressions of the Potentials

We use the superscript (·)(l) to identify the quantities in the lth layer of the multi-layer sphere. The super-
script (·)(0) denotes the background. The scattered potentials, Ascattr and Frscatt , propagate outward and
4 It is easy to show that e−jkz is an eigen function of the Helmholtz equation expressed in the Cartesian coordinate system,
2
which is ∂∂zΦ 2
2 + k Φ = 0.
5 These functions are known as the Riccati-Bessel functions in [8, pp. 283]. However, a different definition of Riccati-Besssel

functions exists in the literature [9, ]. They are defined as

Sn (x) =xjn (x) ,
Cn (x) = − xyn (x) ,
(1)
ξ(x) =xhn (x) ,
(2)
ζ(x) =xhn (x) .
There is a sign difference in the relation to yn (x).

4
take the following form
∞
cos φ X (0) (2)
Ascatt
r = a Ĥ (x)|x=k(0) r Pn1 (x)|x=cos θ , (13a)
ω n=1 n n
∞
sin φ X (0) (2)
Frscatt = b Ĥ (x)|x=k(0) r Pn1 (x)|x=cos θ . (13b)
ωη (0) n=1 n n

(l) (l)
The potentials inside the lth layer, Ar and Fr , contain both the outward and inward propagating wave
and take the following form
∞
cos φ X h (l) (2) i
A(l)
r = an Ĥn (x)|x=k(l) r + c(l) (1) 1
n Ĥn (x)|x=k(l) r Pn (x)|x=cos θ , (14a)
ω n=1
∞
sin φ X h (l) (2) i
Fr(l) = b n Ĥ n (x)|x=k
(l) (1) 1
(l) r + dn Ĥn (x)|x=k (l) r Pn (x)|x=cos θ . (14b)
ωη (l) n=1

(L) (L)
The potentials inside the Lth layer, i.e. the inner core, Ar and Fr , contain the standing wave and take
the following form
∞
cos φ X (L) ˆ
A(L)
r = a Jn (x)|x=k(L) r Pn1 (x)|x=cos θ , (15a)
ω n=1 n
∞
sin φ X (L) ˆ
Fr(L) = b Jn (x)|x=k(L) r Pn1 (x)|x=cos θ . (15b)
ωη (L) n=1 n

Substituting the potentials to (5) gives the expressions of the electric fields and magnetic field. The scattered
electric fields are given as [10, pp. 654]

∞
X h i
Erscat = − j cos φ a(0)
n Ĥn(2) 00 (x)|x=k(0) r + Ĥn(2) (x)|x=k(0) r Pn1 (x)|x=cos φ , (16a)
n=1
∞ 
Pn1 (x)|x=cos θ

cos φ X (0) (2) 0 10
Eθscat = (0) (0) (2)
jan Ĥn (x)|x=k(0) r sin θPn (x)|x=cos θ − bn Ĥn (x)|x=k(0) r , (16b)
k r n=1 sin θ
∞ 
Pn1 (x)|x=cos θ

sin φ X (0) (2) 0 10
Eφscat = (0) jan Ĥn (x)|x=k(0) r (0) (2)
− bn Ĥn (x)|x=k(0) r sin θPn (x)|x=cos θ , (16c)
k r n=1 sin θ
∞
−j sin φ X (0) h (2) 00 i
Hrscat = b Ĥ (x)| (0)
x=k r + Ĥ (2)
(x)| 1
x=k r Pn (x)|x=cos φ ,
(0) (16d)
η (0) n=1 n n n

∞ 
Pn1 (x)|x=cos θ

sin φ X (2) 0 10
Hθscat = (0) (0) jb(0)
n Ĥn (x)| x=k(0) r sin θP n (x)|x=cos θ − a(0) (2)
n Ĥ n (x)|x=k(0) r , (16e)
η k r n=1 sin θ
∞ 
Pn1 (x)|x=cos θ

− cos φ X (0) (2) 0 10
Hφscat = (0) (0) jbn Ĥn (x)|x=k(0) r (0) (2)
− an Ĥn (x)|x=k(0) r sin θPn (x)|x=cos θ . (16f)
η k r n=1 sin θ

Substituting the potentials to (5) gives the expressions of the electric fields and magnetic field in the lth

5
layer.

∞
X h i
Er(l) = − j cos φ a(l)
n Ĥn(2) 00 (x)|x=k(l) r + Ĥn(2) (x)|x=k(l) r Pn1 (x)|x=cos φ
n=1
h i
+ c(l)
n Ĥn(1) 00 (x)|x=k(l) r + Ĥn(1) (x)|x=k(l) r Pn1 (x)|x=cos φ
∞ 
(l) cos φ X (2) 0 10 Pn1 (x)|x=cos θ
Eθ = (l) ja(l)n Ĥn (x)|x=k(l) r sin θPn (x)|x=cos θ − bn Ĥn (x)|x=k(l) r
(l) (2)
k r n=1 sin θ
Pn1 (x)|x=cos θ

(1) 0 10
+jc(l)n Ĥ n (x)| x=k(l) r sin θP n (x)| x=cos θ − d(l) (1)
n Ĥ n (x)| x=k(l) r
sin θ
∞  1
(l) sin φ X (2) 0 Pn (x)|x=cos θ 10
Eφ = (l) ja(l)
n Ĥn (x)|x=k(l) r − b(l) (2)
n Ĥn (x)|x=k(l) r sin θPn (x)|x=cos θ
k r n=1 sin θ
Pn1 (x)|x=cos θ

(l) (1) 0 (l) (1) 10
+jcn Ĥn (x)|x=k(l) r − dn Ĥn (x)|x=k(l) r sin θPn (x)|x=cos θ
sin θ
∞
−j sin φ X (l) h (2) 00 i
Hr(l) = b Ĥ (x)| x=k r(l) + Ĥ (2)
(x)| 1
x=k r Pn (x)|x=cos φ
(l)
η (l) n=1 n n n

h i
+ d(l)
n Ĥn(1) 00 (x)|x=k(l) r + Ĥn(1) (x)|x=k(l) r Pn1 (x)|x=cos φ
∞ 
(l) sin φ X (2) 0 10 Pn1 (x)|x=cos θ
Hθ = (l) (l) jb(l)
n Ĥ n (x)| x=k r(l) sin θP n (x)| x=cos θ − a (l) (2)
n Ĥn (x)| (l)
x=k r
η k r n=1 sin θ
1

(1) 0 10 Pn (x)|x=cos θ
+jd(l)n Ĥn (x)|x=k(l) r sin θPn (x)|x=cos θ − cn Ĥn (x)|x=k(l) r
(l) (1)
sin θ
∞  1
(l) − cos φ X
(2) 0 Pn (x)|x=cos θ 10
Hφ = (l) (l) jb(l)
n Ĥn (x)|x=k(l) r − a(l) (2)
n Ĥn (x)|x=k(l) r sin θPn (x)|x=cos θ
η k r n=1 sin θ
Pn1 (x)|x=cos θ

(1) 0 10
+jd(l)n Ĥ n (x)| x=k r(l) − c(l) (1)
n Ĥn (x)| x=k r(l) sin θPn (x)|x=cos θ
sin θ

Substituting the potentials to (5) gives the expressions of the electric fields and magnetic field in the Lth
layer.

∞
X h i
Er(L) = − j cos φ a(L)
n Ĥn(2) 00 (x)|x=k(L) r + Ĥn(2) (x)|x=k(L) r Pn1 (x)|x=cos θ ,
n=1
∞ 
Pn1 (x)|x=cos θ

(L) cos φ X
(L) (2) 00 10 (L) (2)
Eθ = jan Ĥn (x)|x=k(L) r sin θPn (x)|x=cos θ − bn Ĥn (x)|x=k(L) r ,
k (L) r n=1
sin θ
∞ 
Pn1 (x)|x=cos θ

(L) sin φ X
(L) (2) 00 (L) (2) 10
Eφ = jan Ĥn (x)|x=k(L) r − bn Ĥn (x)|x=k(L) r sin θPn (x)|x=cos θ ,
k (L) r n=1
sin θ
∞
−j sin φ X (L) h (2) 00 i
Hr(L) = b Ĥ (x)| x=k (L) r + Ĥ (2)
(x)| x=k (L)
1
r Pn (x)|x=cos θ ,
η (L) n=1 n n n

∞ 
Pn1 (x)|x=cos θ

(L) sin φ X (2) 00 10
Hθ = (L) (L) jb(L)
n Ĥn (x)| x=k (L) r sin θP n (x)|x=cos θ − a(L) (2)
n Ĥn (x)| x=k (L) r ,
η k r n=1 sin θ
∞ 
Pn1 (x)|x=cos θ

(L) − cos φ X (L) (2) 00 (L) (2) 10
Hφ = (L) (L) jbn Ĥn (x)|x=k(L) r − an Ĥn (x)|x=k(L) r sin θPn (x)|x=cos θ .
η k r n=1 sin θ

6
6 Boundary Conditions
The boundary conditions at the interface between the lth layer and the (l + 1)th layer are the continuities of
the tangential components of the electric fields and magnetic fields. They are stated as

(l) (l+1)
Eθ (r(l+1) ) =Eθ (r(l+1) ) , (19a)
(l) (l+1) (l+1)
Hθ (r(l+1) ) =Hθ (r ) . (19b)
(l) (l+1) (l+1)
Eφ (r(l+1) ) =Eφ (r ) , (19c)
(l) (l+1) (l+1)
Hφ (r(l+1) ) =Hφ (r ) . (19d)

The electric and magnetic fields are dervied from the potentials from (5). Applying the boundary conditions
in (19) to the electric and magnetic fields in the background and the first layer gives
" # " #
2n+1 ˆ 0
(2)
k (1) Ĥn 0 (x)|x=k(0) r(1) k (1) j −n n(n+1) Jn (x)|x=k(0) r(1)
a(0)
n (2) + 2n+1 ˆ
µ(1) Ĥn (x)|x=k(0) r(1) µ(1) j −n n(n+1) Jn (x)|x=k(0) r(1)
" #" #
(2) (1) (1)
k (0) Ĥn 0 (x)|x=k(1) r(1) k (0) Ĥn 0 (x)|x=k(1) r(1) an
= (2) (1) (1) , (20)
µ(0) Ĥn (x)|x=k(1) r(1) µ(0) Ĥn (x)|x=k(1) r(1) cn

" # " #
2n+1 ˆ
(2)
k (1) Ĥn (x)|x=k(0) r(1) k (1) j −n n(n+1) Jn (x)|x=k(0) r(1)
b(0)
n (2) + (1) −n 2n+1 ˆ 0
µ(1) Ĥn 0 (x)|x=k(0) r(1) µ j n(n+1) Jn (x)|x=k(0) r(1)
" #" #
(2) (1) (1)
k (0) Ĥn (x)|x=k(1) r(1) k (0) Ĥn (x)|x=k(1) r(1) bn
= (0) (2) (1) (1) . (21)
µ Ĥn 0 (x)|x=k(1) r(1) µ(0) Ĥn 0 (x)|x=k(1) r(1) dn

Applying the boundary conditions in (19) to the electric and magnetic fields in the lth layer and the (l + 1)th
layer gives
" #" #
(2) (1) (l)
k (l+1) Ĥn 0 (x)|x=k(l) r(l+1) k (l+1) Ĥn 0 (x)|x=k(l) r(l+1) an
(2) (l+1) (1) (l)
µ(l+1) Ĥn (x)|x=kl r(l+1) µ Ĥn (x)|x=kl r(l+1) cn
" #" #
(2) (1) (l+1)
k (l) Ĥn 0 (x)|x=k(l+1) r(l+1) k (l) Ĥn 0 (x)|x=k(l+1) r(l+1) an
= (2) (1) (l+1) , (22)
µ(l) Ĥn (x)|x=k(l+1) r(l+1) µ(l) Ĥn (x)|x=k(l+1) r(l+1) cn

" #" #
(2) (1) (l)
k (l+1) Ĥn (x)|x=k(l) r(l+1) k (l+1) Ĥn (x)|x=k(l) r(l+1) bn
(2) (1) (l)
µ(l+1) Ĥn 0 (x)|x=k(l) r(l+1) µ(l+1) Ĥn 0 (x)|x=k(l) r(l+1) dn
" #" #
(2) (1) (l+1)
k (l) Ĥn (x)|x=k(l+1) r(l+1) k (l) Ĥn (x)|x=k(l+1) r(l+1) bn
= (l) (2) 0 (1) (l+1) . (23)
µ Ĥn (x)|x=k(l+1) r(l+1) µ(l) Ĥn 0 (x)|x=k(l+1) r(l+1) dn

Applying the boundary condition in (19) to the electric and magnetic fields in the (L − 1)th layer and the
Lth layer gives
" #" #
(2) (1) (L−1)
k (L) Ĥn 0 (x)|x=k(L−1) r(L) k (L) Ĥn 0 (x)|x=k(L−1) r(L) an Jˆn (x)|x=k(L) r(L)
 (L−1) 0 
(L) k
(2) (1) (L−1) = an ˆ , (24)
µ(L) Ĥn (x)|x=k(L−1) r(L) µ(L) Ĥn (x)|x=k(L−1) r(L) cn µ(L−1) J(x)|x=k(L) r (L)

" #" #
(2) (1) (L−1)
k (L) Ĥn (x)|x=k(L−1) r(L) k (L) Ĥn (x)|x=k(L−1) r(L) Jˆn (x)|x=k(L) r(L)
 (L−1) 
bn (L) k
(L−1) = bn . (25)
(2)
µ(L) Ĥn 0 (x)|x=k(L−1) r(L)
(1)
µ(L) Ĥn 0 (x)|x=k(L−1) r(L) dn µ(L−1) Jˆn 0 (x)|x=k(L) r(L)

7
7 Implementation Issues
The infinite sum in (16) is approximated by a finite sum. The number of terms, Nmax , is empirically
determined by Nmax = max(Nstop , |m(l) x(l) |, |m(l) x(l−1) |) + 15, l = 1, 2, ...L, where m(l) is the relative
refractive index m(l) = n(l) /n(l−1) and x(l) is related to the size of each layer as x(l) = k (l−1) r(l) [11]. The
value of Nstop is the integer closest to

(L) (L) 1
x + 4(x ) 3 + 1,
 0.02 ≤ x(L) < 8 ,
1
Nstop = x(L) + 4.05(x(L) ) 3 + 2, 8 ≤ x(L) < 4200 ,

 (L) 1
x + 4(x(L) ) 3 + 2, 4200 ≤ x(L) < 20000 .

The derivatives of B̂n (x) in (16) are computed from their recurrent relations
 
0 1 1
B̂n (x) = B̂n−1 (x) + B̂n (x) − B̂n+1 (x) ,
2 x
 
00 1 0 1 0 1 0
B̂n (x) = B̂n−1 (x) + B̂n (x) − 2 B̂n (x) − B̂n+1 (x) ,
2 x x

which are easily derived from the recurrent relations of spherical Bessel functions. The associated Legendre
polynomial is defined as6
dm
Pnm (x) = (−1)m (1 − x2 )m/2 m Pn (x) .
dx
In this study, we are interested in the scattered field at θ = π. Directly substituting θ = π into (16b)
Pn1 (x) 10
and (16c) incurs numerical difficulty when evaluating sin θ |x=cos θ and sin θPn (x)|x=cos θ . This numerical
difficulty is resolved by the recurrent relations of the associated Legendre polynomial. It is shown that [10,
pp. 656]

Pn1 (x) n(n + 1)

|x=cos θ =(−1)n , θ = π, (26a)
sin θ 2
n(n + 1)
sin θPn1 0 (x)|x=cos θ =(−1)n , θ = π. (26b)
2
To prove (26a), we use one of the recurrent relation of the associated Legendre polynomial, which is

1 2n + 1 1 n+1 1
Pn+1 (x) = xPn (x) − Pn−1 ,
n n
where x = cos θ. and thus
1 1
Pn+1 (x) 2n + 1 Pn1 (x) n + 1 Pn−1 (x)
= x − . (27)
sin θ n sin θ n sin θ
Pn1 (x)
Let us define cn (x) = sin θ , then (27) can be rewritten as

2n + 1 n+1
cn+1 (x) = xcn (x) − cn−1 (x) . (28)
n n
Since

P11 (x)|x=cos θ = − sin θ ,

P21 (x)|x=cos θ = − 3 cos θ sin θ ,
3
P31 (x)|x=cos θ = − (5 cos2 θ − 1) sin θ ,
2
6 This definition of the associated Legendre polynomial contains the term (−1)m , known as the Condon-Shortly phase [12,

pp. 772]. The definition including this term is adopted in [10, pp. 951], [5, pp. 108], and [4, pp. 468]. Care must be exercised
when comparing results in the books that use different definitions.

8
 Normalized radar cross section
2
10

1
10

0
10

−1
10

−2
10
0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2
Sphere radius relative to the free−space wavelength

Figure 2: Normalized mono-static radar cross section of a dielectric sphere with a relative permittivity of
2.25 and a radius of one background wavelength [13].

then c1 = −1, c2 = −3 cos θ and c3 = − 32 (5 cos2 θ − 1). Together with (28), we can calculate all terms of
cn for n > 2. When θ = π and n = 2, the first three terms, c1 = −1, c2 = 3, and c3 = −6, satisfy (26a).
Assume that (26a) is true when n = k, then from (28) we have

2k + 1 k+1
ck+1 = − ck − ck−1
k k
2k + 1 k(k + 1) k + 1 k(k − 1)
=− (−1)k − (−1)k−1
k 2 k 2
(k + 1)(k + 2)
=(−1)k+1
2
which is (26a) at n = k + 1. 
The proof of (26b) relies on the recurrent relation
1
Pn−1 (x) P 1 (x)
sin θPn1 0 (x) = (n + 1) − n cos θ n
sin θ sin θ
which can be rewritten as
sin θPn1 0 (x)| = (n + 1)cn−1 − nxcn . (29)
We can use cn and (29) to calculate all terms of sin θPn1 0 (x) for n > 1 and sin θP11 0 (x) = cos θ. When θ = π,
(29) becomes

n(n − 1) n(n + 1)
sin θPn1 0 (cos θ) =(n + 1) (−1)n−1 + n (−1)n
2 2
n(n + 1)
=(−1)n
2

which is (26b).

8 Numerical Validation
Fig. 2 shows the normalized mono-static radar cross sections of a dielectric sphere with a relative permittivity
of 2.25 and a radius of one background wavelength. The figure is consistent with the result reported in [13].
Fig. 3 demonstrates the scenario that an x-polarized plane wave propagating in the z direction is scattered
by a three-layer dielectric sphere with the relative permittivity of 2, 8 and 4 and the radii of 1.5, 1 and 0.6
normalized to the background wavelength. The total field of |Hr | and |Hθ | on the y-z plane passing through

9
 0.022
0
0.02
z (background wavelength) 0.018
0.5 0.016
0.014
0.012
1
0.01
0.008
0.006
1.5
0.004
0.002
2
0 0.5 1 1.5 2
y (background wavelength)
(a)

0 0.025
z (background wavelength)

0.02
0.5

0.015

0.01

1.5
0.005

2
0 0.5 1 1.5 2
y (background wavelength)
(b)

Figure 3: Total field of (a) |Hr | and (b) |Hθ | as the x-polarized plane wave interacts with a three-layer
dielectric sphere with the relative permittivity of 2, 8 and 4 and the radii of 1.5, 1 and 0.6 of a background
wavelength.

10
the origin are shown in Fig. 3. The continuity of these tangential components demonstrates the correctness
in the solution to the linear systems and the implementation of (16).

References
[1] W. C. Chew, Waves and Fields in Inhomogeneous Media. Piscataway, NJ: IEEE Press, 1995.
[2] A. Ishimaru, Electromagnetic Wave Propagation, Radiation, and Scattering. Upper Saddle River, NJ:
Prentice Hall, Inc., 1991.

[3] W. C. Chew and W. H. Weedon, “A 3-D perfectly matched medium from modified Maxwell’s equations
with stretched coordinates,” Microwave and optical technology letters, vol. 7, no. 13, pp. 90–92, 1995.
[4] R. F. Harrington, Time-Harmonic Electromagnetic Fields. Piscataway, NJ: IEEE Press, 2001.
[5] J. D. Jackson, Classical Electrodynamics, 3rd ed. New York, NY: John Wiley & Sons, Inc., 1998.

[6] J. A. Stratton, Electromagnetic Theory. New York, NY: McGraw-Hill Companies Inc., 1941.
[7] R. Haberman, Applied Partial Differential Equations: with Fourier Series and Boundary Value Prob-
lems, 4th ed. Upper Saddle River, NJ: Prentice Hall, Inc., 2003.
[8] S. Zhang and J. Jin, Computation of Special Functions. New York, NY: John Wiley & Sons, Inc., 1996.

[9] C. F. Bohren and D. R. Huffman, Absorption and Scattering of Light by Small Particles. New York,
NY: John Wiley & Sons, Inc., 1983.
[10] C. A. Balanis, Advanced Engineering Electromagnetics. New York, NY: John Wiley & Sons, Inc., 1989.
[11] W. J. Wiscombe, “Improved Mie scattering algorithms,” Applied Optics, vol. 19, no. 9, pp. 1505–1509,
1980.

[12] G. B. Arfken and H. J. Weber, Mathematical Methods for Physicists, 6th ed. Amsterdam, Holland:
Academic Press, 2001.
[13] M. Djordjević and B. M. Notarǒs, “Double higher order method of moments for surface integral equation
model of metallic and dielectric antennas and scatterers,” IEEE Trans. Antennas Propag., vol. 52, no. 8,
pp. 2118–2129, 2004.

11