1

I. INTRODUCTION
Some problems regarding the elliptical cylinders can be solved by using an analytical
approach like that applied to circular cylinders: one separates the variables and the exact
solution is given by expansions involving angular and radial Mathieu functions. These
functions have been introduced by Emile Mathieu in 1868 by investigating the vibrating
modes in an elliptic membrane [1]. Details (tables or relations) concerning the Mathieu
functions can be found for example in [212]. For circular cylinders the solutions involve
readily available trigonometric and Bessel functions, while for elliptical cylinders there are
still controversial and incomplete algorithms for computing the Mathieu functions. One
reason for the lack of algorithms for Mathieu functions was probably the complicated and
various notation existent in the literature. A main purpose for us was to simplify as much
as possible the notation. With a simplied and self-contained notation, the use of Mathieu
functions should be as simple as the use of Bessel functions. We largely followed the notations
used by Stratton [6] and Stamnes [11, 12], but we introduced further simplications. All
formulas required to get the Mathieu functions are given explicitly. Tables of numerical
values are provided. Several examples are given. Mathieu functions applied to plane wave
scattering by elliptical cylinders can be found in [13, 14].
II. FUNDAMENTALS
A. Elliptical cylinder coordinates
Let consider an ellipse in the plane (x, y) dened by equation (x/x
0
)
2
+ (y/y
0
)
2
= 1
with x
0
> y
0
. The semifocal distance f is given by f
2
= x
2
0
y
2
0
and the eccentricity is
e = f/x
0
< 1. The elliptic cylindrical coordinates (u, v, z) are dened by relations
x = f cosh u cosv, y = f sinh u sin v, z = z (1)
with 0 u < and 0 v 2. In terms of (, , z), with = cosh u and = cos v, the
elliptic cylindrical coordinates are dened by relations
x = f , y = f

(
2
1)(1
2
), z = z. (2)
The contours of constant u are confocal ellipses (of semiaxes x
0
= f, y
0
= f

2
1)
and those of constant v are confocal hyperbolas (see Fig. 1). The z axis coincides with the
2
cylinder axis. The scale factors h
j
, with j = , , z, are dened like as for any coordinate
transformation [6],
h

= f

2
1
, h

= f

1
2
, h
z
= 1. (3)

u=const

v=const
2f < >
^
> x
y
F
1
F
2
FIG. 1: Elliptic cylindrical coordinates. F
1
and F
2
are the foci of the ellipse. f is the semifocal
length.
B. Wave equation in elliptic cylindrical coordinates
The scalar wave equation (
2
+ k
2
)U(r) = 0, where r is the position vector, k is the
wave number, k = 2

/, is the permittivity, and is the wavelength in vacuum, when

expressed in elliptic cylindrical coordinates becomes

2
f
2
(cosh 2u cos 2v)

2
u
2
+

2
v
2

+

2
z
2
+ k
2

U(u, v, z) = 0. (4)
3
Using a solution of the form U = Z(z)S(v)R(u) gives

d
2
dz
2
+ k
2
z

Z(z) = 0, (5)

d
2
dv
2
+ (a 2q cos 2v)

S(v) = 0, (6)

d
2
du
2
(a 2q cosh 2u)

R(u) = 0, (7)
where k
z
is the wave vector component on z direction, q = k
2

f
2
/4, with k
2

= k
2
k
2
z
, and a
is separation constant. Equation (5) has solution Z(z) = exp (ik
z
z). Equations (6) and (7)
are known as the angular and radial Mathieu equations, respectively.
III. ANGULAR MATHIEU FUNCTIONS
In this version, only the periodic solutions of period or 2 are considered. For a given
order n, there are four categories of periodic solutions satisfying (6):
1 even-even: S
ee
(v, q, n) =

j=0
A
(2j)
ee
(q, n) cos(2jv),
2 even-odd: S
eo
(v, q, n) =

j=0
A
(2j+1)
eo
(q, n) cos[(2j + 1)v],
3 odd-even: S
oe
(v, q, n) =

j=1
A
(2j)
oe
(q, n) sin(2jv),
4 odd-odd: S
oo
(v, q, n) =

j=0
A
(2j+1)
oo
(q, n) sin[(2j + 1)v].
(8)
A
pm
with p, m = e, o are expansion coecients. In the following, the angular Mathieu
functions are denoted S
pm
(v, q, n), with p, m = e, o. Instead of two angular Mathieu func-
tions, even S
ep
and odd S
op
, with p = e, o [12], a single angular Mathieu function S
pm
, with
p, m = e, o, is considered refering to all the four categories. For a given value of q there
exist four innite sequences of characteristic values (eigenvalues) a, for either value of a
corresponding an innite sequence (eigenvector) of expansion coecients.
4
A. Characteristic values and coecients
By subsituting (8) in (6), the following recurrence relations among the expansion coe-
cients result
1 even-even:
aA
(0)
ee
qA
(2)
ee
= 0,
(a 4)A
(2)
ee
q[2A
(0)
ee
+ A
(4)
ee
] = 0,
[a (2j)
2
]A
(2j)
ee
q[A
(2j2)
ee
+ A
(2j+2)
ee
] = 0, j = 2, 3, 4 (9)
2 even-odd:
(a 1)A
(1)
eo
q[A
(1)
eo
+ A
(3)
eo
] = 0,
[a (2j + 1)
2
]A
(2j+1)
eo
q[A
(2j1)
eo
+ A
(2j+3)
eo
] = 0, j = 1, 2, 3 (10)
3 odd-even:
(a 4)A
(2)
oe
qA
(4)
oe
= 0,
[a (2j)
2
]A
(2j)
oe
q[A
(2j2)
oe
+ A
(2j+2)
oe
] = 0, j = 2, 3, 4 (11)
4 odd-odd:
(a 1)A
(1)
oo
+ q[A
(1)
oo
A
(3)
oo
] = 0,
[a (2j + 1)
2
]A
(2j+1)
oo
q[A
(2j1)
oo
+ A
(2j+3)
oo
] = 0, j = 1, 2, 3 . (12)
The recurrence relations can be written in matrix form [11],
1 even-even:

a q 0 0 0 0
2q 2
2
a q 0 0 0
0 q 4
2
a q 0 0
0 0 q 6
2
a q 0
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

A
(0)
ee
A
(2)
ee
A
(4)
ee
A
(6)
ee
.
.
.

= 0, (13)
2 even-odd:

1 + q a q 0 0 0 0
q 3
2
a q 0 0 0
0 q 5
2
a q 0 0
0 0 q 7
2
a q 0
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

A
(1)
eo
A
(3)
eo
A
(5)
eo
A
(7)
eo
.
.
.

= 0, (14)
5
3 odd-even:

2
2
a q 0 0 0 0
q 4
2
a q 0 0 0
0 q 6
2
a q 0 0
0 0 q 8
2
a q 0
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

A
(2)
oe
A
(4)
oe
A
(6)
oe
A
(8)
oe
.
.
.

= 0, (15)
4 odd-odd:

1 q a q 0 0 0 0
q 3
2
a q 0 0 0
0 q 5
2
a q 0 0
0 0 q 7
2
a q 0
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

A
(1)
oo
A
(3)
oo
A
(5)
oo
A
(7)
oo
.
.
.

= 0. (16)
The matrices are real, tridiagonal, and symmetric for all categories, with the exception
of the 1 even-even category where the matrix is slightly non-symmetric. The eigenvalue
problem is accurately solved in Matlab. In other computer programs it could be necessary to
transform the slightly non-symmetric matrix in a symmetric one [11]. Both the eigenvalues
a and the corresponding eigenvectors (A
pm
, with p, m = e, o) are determined for either
category at any order n. The order n takes dierent values for each category of Mathieu
functions. For the purpose of avoiding any confusion, a distinction must be done between
the n
th
order (in the succession of all orders) and the true value of that order. Thus, let
denote n the order in the succession of all orders, and t the true value of order n. The values
of n and t for the four categories of Mathieu functions are
1 even-even: n = 0, 1, 2 t = 0, 2, 4 ,
2 even-odd: n = 0, 1, 2 t = 1, 3, 5 ,
3 odd-even: n = 1, 2, 3 t = 2, 4, 6 ,
4 odd-odd: n = 0, 1, 2 t = 1, 3, 5 .
Note that, if the notation is self-contained by all routines of Mathieu functions, there is no
need to determine the specic values of n and t for either category of Mathieu functions
since it is done automatically.
6
B. Normalization and orthogonality
Following [6, 11], the angular Mathieu functions are normalized by requiring that
S
ep
(0, q, n) = 1,

dS
op
(v, q, n)
dv

v=0
= 1, p = e, o. (17)
These requirements imply that,
1 even-even:

j=0
A
(2j)
ee
(q, n) = 1,
2 even-odd:

j=0
A
(2j+1)
eo
(q, n) = 1,
3 odd-even:

j=1
2jA
(2j)
oe
(q, n) = 1,
4 odd-odd:

j=0
(2j + 1)A
(2j+1)
oo
(q, n) = 1.
(18)
The orthogonality relation for the angular Mathieu functions is

2
0
S
pm
(v, q, n)S
pm
(v, q, n) dv = N
pm

mm
, p, m, m

= e, o, (19)
where N
pm
is normalization factor,
mm
equals 1 if m = m

and equals 0 otherwise. Then,

the following relations for the normalization factor result,
1 even-even: N
ee
(q, n) = 2[A
(0)
ee
(q, n)]
2
+

j=1
[A
(2j)
ee
(q, n)]
2
,
2 even-odd: N
eo
(q, n) =

j=0
[A
(2j+1)
eo
(q, n)]
2
,
3 odd-even: N
oe
(q, n) =

j=1
[A
(2j)
oe
(q, n)]
2
,
4 odd-odd: N
oo
(q, n) =

j=0
[A
(2j+1)
oo
(q, n)]
2
.
(20)
Since dierent normalization schemes have been adopted in the literature, much attention
should be paid when numerical results provided by dierent authors are compared ones
against the others.
7
C. Correlation factors
Let consider two regions of dierent permittivities, and

. The parameter q being

dierent in the two regions, q = q

, the characteristic values and expansion coecients are

also dierent. Let S
pm
and S

pm
be the respective angular Mathieu functions. The correlation
factors C
pm
(q, q

, n), with p, m = e, o, between the angular Mathieu functions S

pm
and S

pm
are dened by relation
C
pm
(q, q

, n) =
mm

2
0
S
pm
(v, q, n)S

pm
(v, q

, n) dv, p, m, m

= e, o. (21)
Using (8) gives
1 even-even: C
ee
(q, q

, n) = 2A
(0)
ee
(q, n)A
(0)
ee
(q

, n)
+

j=1
A
(2j)
ee
(q, n)A
(2j)
ee
(q

, n),
2 even-odd: C
eo
(q, q

, n) =

j=0
A
(2j+1)
eo
(q, n)A
(2j+1)
eo
(q

, n), (22)
3 odd-even: C
oe
(q, q

, n) =

j=1
A
(2j)
oe
(q, n)A
(2j)
oe
(q

, n),
4 odd-odd: C
oo
(q, q

, n) =

j=0
A
(2j+1)
oo
(q, n)A
(2j+1)
oo
(q

, n).
D. Derivatives of angular Mathieu functions
The derivatives of the angular Mathieu functions follow readily from (8),
1 even-even:
dS
ee
(v, q, n)
dv
=

j=1
2jA
(2j)
ee
(q, n) sin(2jv),
2 even-odd:
dS
eo
(v, q, n)
dv
=

j=0
(2j + 1)A
(2j+1)
eo
(q, n) sin[(2j + 1)v], (23)
3 odd-even:
dS
oe
(v, q, n)
dv
=

j=1
2jA
(2j)
oe
(q, n) cos(2jv),
4 odd-odd:
dS
oo
(v, q, n)
dv
=

j=0
(2j + 1)A
(2j+1)
oo
(q, n) cos[(2j + 1)v].
8
IV. RADIAL MATHIEU FUNCTIONS
Solutions of (7) can be obtained from (8) by replacing v by iu. Instead of sin v and cos v,
the terms of the series now involve sinh u and cosh u. The convergence is low unless |u|
is small. Better convergence of series results by expressing the solutions of (7) in terms of
Bessel functions associated with the same expansion coecients that are determined once for
both the angular and radial Mathieu functions. Either pair of angular and radial Mathieu
functions are proportional to one another [6],
S
ep
(iu, q, n) =

2g
ep
(q, n)J
ep
(u, q, n), p = e, o, (24)
where J
ep
are even radial Mathieu functions of the rst kind and g
ep
are joining factors.
When u = 0,
S
ep
(0, q, n) = 1, J
ep
(0, q, n) =
1

2g
ep
(q, n)
, p = e, o. (25)
Thus, one obtains,
1 even-even: g
ee
(q, n) =
(1)
r
A
(0)
ee
(q, n)
S
ee
(/2, q, n), r = t/2,
2 even-odd: g
eo
(q, n) =
(1)
r

qA
(1)
eo
(q, n)

dS
eo
(v, q, n)
dv

v=/2
, r = (t 1)/2. (26)
Similarly [6],
iS
op
(iu, q, n) =

2g
op
(q, n)J
op
(u, q, n), p = e, o. (27)
When u = 0,
J
op
(0, q, n) = 0,

dJ
op
(u, q, n)
du

u=0
=
1

2g
op
(q, n)
, p = e, o. (28)
Thus, one obtains,
3 odd-even: g
oe
(q, n) =
(1)
r
qA
(2)
oe
(q, n)

dS
oe
(v, q, n)
dv

v=/2
, r = t/2,
4 odd-odd: g
oo
(q, n) =
(1)
r

qA
(1)
oo
(q, n)
S
oo
(/2, q, n), r = (t 1)/2. (29)
Remember that t is the true value of order n.
9
A. Radial Mathieu functions of the rst kind
Since rapidly converging series are those expressed in terms of products of Bessel functions
[10, 11], in the following relations refer only to them. Similarly to the angular Mathieu
functions, one may distinct four categories of radial Mathieu functions of the rst kind
which are denoted J
pm
(u, q, n), with p, m = e, o,
1 even-even: J
ee
(u, q, n) =

2
(1)
r
A
(0)
ee
(q, n)

j=0
(1)
j
A
(2j)
ee
(q, n)J
j
(v
1
)J
j
(v
2
),
r = t/2,
2 even-odd: J
eo
(u, q, n) =

2
(1)
r
A
(1)
eo
(q, n)

j=0
(1)
j
A
(2j+1)
eo
(q, n)[J
j
(v
1
)J
j+1
(v
2
)
+ J
j
(v
2
)J
j+1
(v
1
)], r = (t 1)/2, (30)
3 odd-even: J
oe
(u, q, n) =

2
(1)
r
A
(2)
oe
(q, n)

j=1
(1)
j
A
(2j)
oe
(q, n)[J
j1
(v
1
)J
j+1
(v
2
)
J
j1
(v
2
)J
j+1
(v
1
)], r = t/2,
4 odd-odd: J
oo
(u, q, n) =

2
(1)
r
A
(1)
oo
(q, n)

j=0
(1)
j
A
(2j+1)
oo
(q, n)[J
j
(v
1
)J
j+1
(v
2
)
J
j
(v
2
)J
j+1
(v
1
)], r = (t 1)/2,
where v
1
=

q exp (u) and v
2
=

q exp (u). The derivatives of the radial Mathieu func-
tions of the rst kind are
1 even-even: r = t/2,
dJ
ee
(u, q, n)
du
=

2
(1)
r
A
(0)
ee
(q, n)

j=0
(1)
j
A
(2j)
ee
(q, n)[v
1
J
j+1
(v
1
)J
j
(v
2
)
v
2
J
j
(v
1
)J
j+1
(v
2
)],
2 even-odd: r = (t 1)/2,
dJ
eo
(u, q, n)
du
=

2
(1)
r
A
(1)
eo
(q, n)

j=0
(1)
j
A
(2j+1)
eo
(q, n)

(v
2
v
1
)[J
j
(v
1
)J
j
(v
2
)
J
j+1
(v
1
)J
j+1
(v
2
)] + (2j + 1)[J
j+1
(v
1
)J
j
(v
2
) J
j
(v
1
)J
j+1
(v
2
)]

,
10
3 odd-even: r = t/2,
dJ
oe
(u, q, n)
du
=

2
(1)
r
A
(2)
oe
(q, n)

j=0
(1)
j+1
A
(2j+2)
oe
(q, n)(4j + 4)

J
j
(v
1
)J
j
(v
2
) (31)
+ cosh 2uJ
j+1
(v
1
)J
j+1
(v
2
) (j + 1)[
1
v
1
J
j+1
(v
1
)J
j
(v
2
) +
1
v
2
J
j
(v
1
)J
j+1
(v
2
)]

,
4 odd-odd: r = (t 1)/2,
dJ
oo
(u, q, n)
du
=

2
(1)
r
A
(1)
oo
(q, n)

j=0
(1)
j
A
(2j+1)
oo
(q, n)

(v
1
+ v
2
)[J
j
(v
1
)J
j
(v
2
)
+ J
j+1
(v
1
)J
j+1
(v
2
)] (2j + 1)[J
j+1
(v
1
)J
j
(v
2
) + J
j
(v
1
)J
j+1
(v
2
)]

.
B. Radial Mathieu functions of the second kind
A second independent solution of (7) is obtained by replacing the Bessel functions of the
rst kind J
n
(v
2
) in (30) by the Bessel functions of the second kind Y
n
(v
2
), [10, 11]. This
solution is denoted Y
pm
(u, q, n), with p, m = e, o.
1 even-even: Y
ee
(u, q, n) =

2
(1)
r
A
(0)
ee
(q, n)

j=0
(1)
j
A
(2j)
ee
(q, n)J
j
(v
1
)Y
j
(v
2
),
r = t/2,
2 even-odd: Y
eo
(u, q, n) =

2
(1)
r
A
(1)
eo
(q, n)

j=0
(1)
j
A
(2j+1)
eo
(q, n)[J
j
(v
1
)Y
j+1
(v
2
)
+ Y
j
(v
2
)J
j+1
(v
1
)], r = (t 1)/2, (32)
3 odd-even: Y
oe
(u, q, n) =

2
(1)
r
A
(2)
oe
(q, n)

j=1
(1)
j
A
(2j)
oe
(q, n)[J
j1
(v
1
)Y
j+1
(v
2
)
Y
j1
(v
2
)J
j+1
(v
1
)], r = t/2,
4 odd-odd: Y
oo
(u, q, n) =

2
(1)
r
A
(1)
oo
(q, n)

j=0
(1)
j
A
(2j+1)
oo
(q, n)[J
j
(v
1
)Y
j+1
(v
2
)
Y
j
(v
2
)J
j+1
(v
1
)], r = (t 1)/2,
The derivatives of the radial Mathieu functions of the second kind are
11
1 even-even: r = t/2,
dY
ee
(u, q, n)
du
=

2
(1)
r
A
(0)
ee
(q, n)

j=0
(1)
j
A
(2j)
ee
(q, n)[v
1
J
j+1
(v
1
)Y
j
(v
2
)
v
2
J
j
(v
1
)Y
j+1
(v
2
)],
2 even-odd: r = (t 1)/2,
dY
eo
(u, q, n)
du
=

2
(1)
r
A
(1)
eo
(q, n)

j=0
(1)
j
A
(2j+1)
eo
(q, n)

(v
2
v
1
)[J
j
(v
1
)Y
j
(v
2
)
J
j+1
(v
1
)Y
j+1
(v
2
)] + (2j + 1)[J
j+1
(v
1
)Y
j
(v
2
) J
j
(v
1
)Y
j+1
(v
2
)]

,
3 odd-even: r = t/2,
dY
oe
(u, q, n)
du
=

2
(1)
r
A
(2)
oe
(q, n)

j=0
(1)
j+1
A
(2j+2)
oe
(q, n)(4j + 4)

J
j
(v
1
)Y
j
(v
2
) (33)
+ cosh 2uJ
j+1
(v
1
)Y
j+1
(v
2
) (j + 1)[
1
v
1
J
j+1
(v
1
)Y
j
(v
2
) +
1
v
2
J
j
(v
1
)Y
j+1
(v
2
)]

,
4 odd-odd: r = (t 1)/2,
dY
oo
(u, q, n)
du
=

2
(1)
r
A
(1)
oo
(q, n)

j=0
(1)
j
A
(2j+1)
oo
(q, n)

(v
1
+ v
2
)[J
j
(v
1
)Y
j
(v
2
)
+ J
j+1
(v
1
)Y
j+1
(v
2
)] (2j + 1)[J
j+1
(v
1
)Y
j
(v
2
) + J
j
(v
1
)Y
j+1
(v
2
)]

.
C. Radial Mathieu functions of the third and the fourth kinds
Radial Mathieu functions of the third kind, analogous to the Hankel functions of the rst
kind are dened as follows [6, 11]
H
pm1
(u, q, n) = J
pm
(u, q, n) + iY
pm
(u, q, n), p, m = e, o. (34)
Similarly, radial Mathieu functions of the fourth kind, analogous to the Hankel functions of
the second kind are dened as follows [6, 11]
H
pm2
(u, q, n) = J
pm
(u, q, n) iY
pm
(u, q, n), p, m = e, o. (35)
12
V. IMPLEMENTATION OF MATHIEU FUNCTIONS IN MATLAB
Following the notation of the four categories of angular Mathieu functions, the
implementation in Matlab or in any other computer program is readily done by
introducing a function code KF. The rst step in any algorithm of Mathieu function
computation is to nd the characteristic values (eigenvalues) and the expansion coecients
(eigenvectors). This is done by routine eig Spm which has q as input parameter (see
Table I). Besides q, the function code KF should be specied. Thus, if KF = 1, the
routine eig Spm solves the eigenvalue problem for category 1 even-even of Mathieu
functions, if KF = 2 for category 2 even-odd, and so on. The number of expansion
coecients is the same, it is set equal to 25, for all categories of Mathieu functions.
Concerning the outputs of routine eig Spm, va is a line vector representing the
characteristic values a for all the 25 orders; mc is 25 25 matrix, where the columns
represent the eigenvectors (that is, the expansion coecients) for all orders; vt is a column
vector specifying the true value t for all orders. Note that the eigenvectors in mc were
processed to obey equation (18). For the purpose to save the time of computation, all the
other routines have mc as input (see Table I), the routine eig Spm being called once, at
the beginning of the computation, for any values of coordinates u and v that intervene in
that computation. Since in many cases the convergence is assured by the rst several
orders, all the other routines have nmax 25 as input. It means that those routines take
into account only the rst nmax orders, but for either order the length of the
corresponding eigenvector is the same, equal to 25. The routine extract one value can be
used to get a single value, and the routine extract one column to get a single
eigenvector, corresponding to the order t. The derivatives of S
pm
, with p, m = e, o, are
computed by routine dSpm. For both Spm and dSpm, v is expressed in radians,
with values in interval (0, 2). The normalization, correlation, and joining factors are
computed by routines Npm, Cpm, and gpm, respectively. The four kinds of radial
Mathieu functions, J
pm
, Y
pm
, H
pm1
, and H
pm2
, with p, m = e, o, are computed by routines
Jpm,Ypm,Hpm1, and Hpm2, respectively, and their derivatives with respect to u
by routines dJpm,dYpm,dHpm1, and dHpm2, respectively.
Numerical values of the separation constant a, of the angular Mathieu functions S
pm
and
their derivatives S

pm
, with p, m = e, o, where the prime denotes dierentiation with respect
13
to v, are given in Tables IIIV. They can be compared with data in [2]. With the purpose
to facilitate the comparison, since in [2] the normalization N
pm
= is applied, the data of
S
pm
and S

pm
in Tables IIIV are multiplied by

/N
pm
.
Concerning the radial Mathieu functions, numerical values of S
ep
(iu, q, n) and
iS
op
(iu, q, n) are given for u = 0.5 in Tables V and VI. They are multiplied by

/N
pm
and compared with data in [9]. Note that S
ep
is correlated to the radial Mathieu function
of the rst kind J
ep
by Eq. (24), whereas S
op
is correlated to J
op
by Eq. (27). We found
that, for parameters in [9], the values of S
ep
(iu, q, n) and iS
op
(iu, q, n) calculated with
Eqs. (24) and (27) dier from those obtained with Eq. (8) by less than 7.5 10
12
.
Four les are attached as examples of Mathieu functions computation: example1 Spm.m,
example2 Spm.m, example1 Jpm.m, and example2 Jpm.m. Results are shown in Figs. 25.
[1] E. Mathieu Le mouvement vibratoire dune membrane de forme elliptique, Jour. de Math.
Pures at Appliquees (Jour. de Liouville) 13, 137203 (1868).
[2] M. Abramowitz and I. Stegun Handbook of Mathematical Functions (New York, Dover,
1964).
[3] I. S. Gradshteyn and I. M. Ryzhik Tables of Integrals, Series, and Products (Academic
Press, San Diego, 1994).
[4] E. L. Ince Ordinary Dierential Equations (Dover, New York, 1967).
[5] N. W. McLachlan Theory and Application of Mathieu Functions (Oxford Press, London,
1951).
[6] J. A. Stratton Electromagnetic Theory (Mc-Graw Hill New York, 1941).
[7] J. A. Stratton and P. M. Morse Elliptic Cylinder and Spheroidal Wave Functions Including
Tables of Separation Constants and Coecients (John Wiley & Sons, New York, 1941).
[8] E. T. Whittaker and G. N. Watson A Course of Modern Analysis (Cambridge University
Press, Cambridge, 1950).
[9] E. T. Kirkpatrick, Tables of values of the modied Mathieu functions, Mathematics of
Computation 14 118129 (1960).
[10] J. C. Gutierrez-Vega, Formal analysis of the propagation of invariant optical elds in elliptic
coordinates, Ph. D. Thesis, INAOE, Mexico, 2000.
14
[11] J. J. Stamnes and B. Spjelkavik New method for computing eigenfunctions (Mathieu
functions) for scattering by elliptical cylinders, Pure Appl. Opt. 4 25162 (1995).
[12] J. J. Stamnes Exact two-dimensional scattering by perfectly reecting elliptical cylinders,
strips and slits, Pure Appl. Opt. 4 84155 (1995).
[13] E. Cojocaru, Mathieu functions approach to bidimensional scattering by dielectric elliptical
cylinders, e-print arXiv:0808.2123v1.
[14] E. Cojocaru, Elliptical cylindrical invisibility cloak, a semianalytical approach using
Mathieu functions, e-print arXiv:0808.1498v1.
15
0 0.1 0.2 0.3 0.4 0.5
1
0.5
0
0.5
1
v/pi
S
p
m
q=1; n=3:5; KF=1
0 0.1 0.2 0.3 0.4 0.5
1
0.5
0
0.5
1
v/pi
S
p
m
q=1; n=3:5; KF=2
0 0.1 0.2 0.3 0.4 0.5
1
0.5
0
0.5
1
v/pi
S
p
m
q=1; n=3:5; KF=3
0 0.1 0.2 0.3 0.4 0.5
1
0.5
0
0.5
1
v/pi
S
p
m
q=1; n=3:5; KF=4
FIG. 2: Examples of angular Mathieu functions at q = 1 and orders n = 3 (black), n = 4 (red),
and n = 5 (blue). The angular coordinate v is varied from 0 to /2.
16
0 0.5 1 1.5 2
0
10
20
30
v/pi
S
p
m
n=1; q=1:5; KF=1
0 0.5 1 1.5 2
5
0
5
v/pi
S
p
m
n=1; q=1:5; KF=2
0 0.5 1 1.5 2
2
1
0
1
2
v/pi
S
p
m
n=1; q=1:5; KF=3
0 0.5 1 1.5 2
10
5
0
5
10
v/pi
S
p
m
n=1; q=1:5; KF=4
FIG. 3: Examples of angular Mathieu functions at n = 1 and dierent values of q: q = 1 (black),
q = 2 (red), q = 3 (blue), q = 4 (cyan), and q = 5 (magenta). The angular coordinate v is varied
from 0 to /2.
17
0 0.5 1 1.5 2 2.5
0.4
0.2
0
0.2
0.4
0.6
u
J
p
m
q=1; n=3,6,9; KF=1
0 0.5 1 1.5 2 2.5
0.4
0.2
0
0.2
0.4
0.6
u
J
p
m
q=1; n=3,6,9; KF=2
0 0.5 1 1.5 2 2.5
0.4
0.2
0
0.2
0.4
0.6
u
J
p
m
q=1; n=3,6,9; KF=3
0 0.5 1 1.5 2 2.5
0.4
0.2
0
0.2
0.4
0.6
u
J
p
m
q=1; n=3,6,9; KF=4
FIG. 4: Examples of radial Mathieu functions of the rst kind at q = 1 and orders n = 3 (black),
n = 6 (red), and n = 9 (blue).
18
0 0.5 1 1.5 2 2.5
0.5
0
0.5
1
u
J
p
m
n=1; q=1:5; KF=1
0 0.5 1 1.5 2 2.5
0.5
0
0.5
1
u
J
p
m
n=1; q=1:5; KF=2
0 0.5 1 1.5 2 2.5
0.5
0
0.5
1
u
J
p
m
n=1; q=1:5; KF=3
0 0.5 1 1.5 2 2.5
0.5
0
0.5
1
u
J
p
m
n=1; q=1:5; KF=4
FIG. 5: Examples of radial Mathieu functions of the rst kind at n = 1 and dierent values of q:
q = 1 (black), q = 2 (red), q = 3 (blue), q = 4 (cyan), and q = 5 (magenta).
19
TABLE I: Routines comprised in the toolbox
Name of routine Routine call What the routine computes
eig Spm [va, mc, vt]=eig Spm(KF, q) Vector of characteristic values va, matrix of
coecients mc, and vector of orders vt, at given
function code KF and elliptical parameter q 0.
Spm y=Spm(KF, v, mc, nmax) Angular Mathieu functions S
pm
, [Eq. (8)].
dSpm y=dSpm(KF, v, mc, nmax) Derivatives with respect to v of S
pm
, [Eq. (23)].
Npm y=Npm(KF, mc, nmax) Normalizing factors of angular Mathieu functions
S
pm
, [Eqs. (19) and (20)].
Cpm y=Cpm(KF, mc, mc

, nmax) Correlation factors of S

pm
and S

pm
, having matrices
of coecients mc and mc

, [Eqs. (21) and (22)].

Jpm y=Jpm(KF, u, q, mc, nmax) Radial Mathieu functions of the rst kind J
pm
,
[Eq. (30)].
dJpm y=dJpm(KF, u, q, mc, nmax) Derivatives with respect to u of J
pm
, [Eq. (31)].
gpm y=gpm(KF, q, mc, nmax) Joining factors for pairs of angular, S
pm
and
radial, J
pm
Mathieu functions, [Eqs. (24)(29)].
Ypm y=Ypm(KF, u, q, mc, nmax) Radial Mathieu functions of the second kind
Y
pm
, [Eq. (32)].
dYpm y=dYpm(KF, u, q, mc, nmax) Derivatives with respect to u of Y
pm
, [Eq. (33)].
Hpm1 y=Hpm1(KF, u, q, mc, nmax) Radial Mathieu functions of the third kind H
pm1
,
[Eq. (34)].
dHpm1 y=dHpm1(KF, u, q, mc, namax) Derivatives with respect to u of H
pm1
.
Hpm2 y=Hpm2(KF, u, q, mc, nmax) Radial Mathieu functions of the fourth kind H
pm2
,
[Eq. (35)].
dHpm2 y=dHpm2(KF, u, q, mc, namax) Derivatives with respect to u of H
pm2
.
extract one column y=extract one column(KF, t, mc) Extracts one column from mc at given t.
extract one value y=extract one value(KF, t, vec) Extracts one value from vec at given t.
20
TABLE II: Values of S
ee
multiplied by
ee
=

/N
ee
to be compared with data in [2]
t q a
ee
S
ee
(0, q, n)
ee
S
ee
(/2, q, n)
0 0 0 0.7071067811865 0.7071067811865
5 -5.8000460208515 0.0448001816519 1.3348486746980
10 -13.9369799566589 0.0076265175709 1.4686604707129
15 -22.5130377608640 0.0019325083152 1.5501081466866
20 -31.3133900703364 0.0006037438292 1.6098908573959
25 -40.2567795465667 0.0002158630184 1.6575102983235
2 0 4.0000000000000 1.0000000000000 -1.0000000000000
5 7.4491097395292 0.7352943084007 -0.7244881519677
10 7.7173698497796 0.2458883492913 -0.9267592641263
15 5.0779831975435 0.0787928278464 -1.0199662260303
20 1.1542828852468 0.0286489431471 -1.0752932287797
25 -3.5221647271583 0.0115128663309 -1.1162789532953
10 0 100.0000000000000 1.0000000000000 -1.0000000000000
5 100.1263692161636 1.0259950270894 -0.9753474872360
10 100.5067700246816 1.0538159921009 -0.9516453181790
15 101.1452034473016 1.0841063118392 -0.9285480638845
20 102.0489160244372 1.1177886312594 -0.9057107845941
25 103.2302048044949 1.1562399186322 -0.8826919105637
21
TABLE III: Values of S
eo
and S

eo
multiplied by
eo
=

/N
eo
to be compared with data in [2]
t q a
eo
S
eo
(0, q, n)
eo
S

eo
(/2, q, n)
1 0 1.0000000000000 1.0000000000000 -1.0000000000000
5 1.8581875415478 0.2565428793224 -3.4690420034057
10 -2.3991424000363 0.0535987477472 -4.8504383044964
15 -8.1011051316418 0.0150400664538 -5.7642064390510
20 -14.4913014251748 0.0050518137647 -6.4905657825800
25 -21.3148996906657 0.0019110515067 -7.1067412352901
5 0 25.0000000000000 1.0000000000000 -5.0000000000000
5 25.5499717499816 1.1248072506385 -5.3924861549882
10 27.7037687339393 1.2580199413083 -5.3212765411609
15 31.9578212521729 1.1934322304131 -5.1191498884064
20 36.6449897341328 0.9365755314226 -5.7786752500644
25 40.0501909858077 0.6106943100507 -7.0598842916553
15 0 225.0000000000000 1.0000000000000 15.0000000000000
5 225.0558124767096 1.0112937325296 15.1636574720602
10 225.2233569749644 1.0228782824382 15.3198803056623
15 225.5029562446541 1.0347936522369 15.4687435032830
20 225.8951534162079 1.0470843441629 15.6102785232380
25 226.4007200447481 1.0598004418139 15.7444725050679
22
TABLE IV: Values of S
op
and S

op
multiplied by
op
=

/N
op
, where p = e, o, to be compared
with data in [2]
t q a
oe
S

oe
(0, q, n)
oe
S

oe
(/2, q, n)
2 0 4.0000000000000 2.0000000000000 -2.0000000000000
5 2.0994604454867 0.7331661960372 -3.6405178524082
10 -2.3821582359570 0.2488228403985 -4.8634220691653
15 -8.0993467988959 0.0918197143696 -5.7655737717278
20 -14.4910632559807 0.0370277776852 -6.4907522240373
25 -21.3148606222498 0.0160562170491 -7.1067719073739
10 0 100.0000000000000 10.0000000000000 -10.0000000000000
5 100.1263692156019 9.7341731518695 -10.2396462566908
10 100.5067694628784 9.4404054347686 -10.4539475316485
15 101.1451722929092 9.1157513395126 -10.6428998776563
20 102.0483928609361 8.7555450801360 -10.8057241781325
25 103.2256800423735 8.3526783655914 -10.9413538308191
t q a
oo
S

oo
(0, q, n)
oo
S
oo
(/2, q, n)
1 0 1.0000000000000 1.0000000000000 1.0000000000000
5 -5.7900805986378 0.1746754006198 1.3374338870223
10 -13.9365524792501 0.0440225659111 1.4687556641029
15 -22.5130034974235 0.0139251347875 1.5501150743576
20 -31.3133861669129 0.0050778849001 1.6098915926038
25 -40.2567789846842 0.0020443593656 1.6575103983745
5 0 25.0000000000000 5.0000000000000 1.0000000000000
5 25.5108160463032 4.3395700104946 0.9060779302024
10 26.7664263604801 3.4072267604013 0.8460384335355
15 27.9678805967175 2.4116664728002 0.8379493400125
20 28.4682213251027 1.5688968684857 0.8635431218534
25 28.0627658994543 0.9640716219024 0.8992683245108
15 0 225.0000000000000 15.0000000000000 -1.0000000000000
5 225.0558124767096 14.8287889732852 -0.9889607027406
10 225.2233569749643 14.6498600449581 -0.9781423471832
15 225.5029562446537 14.4630006940372 -0.9675137031855
20 225.8951534161767 14.2679460909928 -0.9570452540613
25 226.4007200438825 14.0643732956172 -0.9467086958781
23
TABLE V: Values of S
ep
(iu, q, n) for u = 0.5 multiplied by

/N
ep
, where p = e, o, compared
with data in [9]
t q Values at p = e Data in [9] t q Values at p = o Data in [9]
0 5 -0.019325304910071 -0.01932 1 5 0.021440743185527 0.02144
10 -0.007055239716193 -0.00705 10 -0.038634237458525 -0.03863
20 -0.000169411415735 -0.00016 20 -0.003373888309642 -0.00337
2 5 0.446937465741068 0.44693 3 5 1.205528267066838 1.2055
10 -0.063855921612085 -0.06385 10 0.235940782144547 0.23594
20 -0.024916657795101 -0.02491 20 -0.097385461808731 -0.09738
4 5 2.234088244534832 2.2341 5 5 3.864089377116713 3.8641
10 1.039103163573830 1.0391 10 2.285610444240526 2.2856
20 -0.143991090269732 -0.14399 20 0.274270780278172 0.27427
TABLE VI: Values of iS
op
(iu, q, n) for u = 0.5 multiplied by

/N
op
, with p = e, o, compared
with data in [9]
t q Values at p = e Data in [9] t q Values at p = o Data in [9]
2 5 0.238342768735937 0.23834 1 5 0.036613617783886 0.03661
10 0.028675814044625 0.02867 10 0.000750806874015 0.00075
20 -0.003176296415956 -0.00317 20 -0.000538258353937 -0.00053
4 5 1.883560277440876 1.8836 3 5 0.806555153528872 0.80655
10 0.769679129538722 0.76968 10 0.204495885546638 0.20449
20 0.040515136278697 0.04051 20 -0.005279473480675 -0.00527
6 5 6.6066602369876 6.6067 5 5 3.667530204538722 3.6675
10 4.1161420952367 4.1161 10 1.972361938552091 1.9724
20 1.1805904286267 1.1806 20 0.320398855944192 0.32040