Chapter 2

Bessel Functions
2.1 Bessel, Neumann, and Hankel Functions: J
n
(x), N
n
(x), H
(1)
n
(x), H
(2)
n
(x)
Bessel functions are solutions of the following dierential equation:
x
2
y

+ xy

+ (x
2

2
)y = 0 (2.1)
which is called the Bessels dierential equation. This is a second order dierential equation and has two
linearly independent solutions. Any two of the following functions are linearly independent solutions of (2.1)
J

(x) N

(x) H
(1)

(x) H
(2)

(x) (2.2)
Thus, the general solution for (2.1) can be written as a linear combination of any two of the above functions.
Usually the general solution is written as one of these forms:
AJ

(x) + BN

(x) AH
(1)

(x) + BH
(2)

(x) (2.3)
When is not an integer ( = n) J

and J

are also linearly independent solutions of (2.1), however, we

usually never use J

. The Neumann function is related to J

and J

:
N

(x) =
cos J

(x) J

(x)
sin
(2.4)
N
n
(x) = lim
n
cos J

(x) J

(x)
sin
(2.5)
and Hankel functions of the rst and second kind are related to Bessel and Neumann functions:
H
(1)

(x) = J

(x) + jN

(x) (2.6)
H
(2)

(x) = J

(x) jN

(x) (2.7)
With a variable transformation x = equation (2.1) can be transformed into:

2
y

+ y

+ (
2

2
)y = 0 (2.8)
whose independent solutions are J

() and N

(). When = n is an integer J

n
and J
n
are not
independent anymore and we have:
J
n
(x) = (1)
n
J
n
(x) N
n
(x) = (1)
n
N
n
(x)
J
n
(x) = (1)
n
J
n
(x) N
n
(x) = (1)
n
[2 j J
n
(x) + N
n
(x)]
(2.9)
Plots of the rst three Bessel functions are shown in Fig. 2.1 and the rst three Neumann functions are
shown in Fig. 2.2
8
Amir Borji Special Functions and Laplace Equation 9
0.6
0.2
0.4
0.6
0.8
1
0 2 4 6 8 10 12 14
0.2
0.4
0
J
1
(x)
J
2
(x)
J
0
(x)
Figure 2.1: Bessel functions of the rst kind
2.1.1 Small and Large Argument Approximations
Small Argument Limit |x| 1
J
n
(x)
1
n!
_
x
2
_
n
(2.10)
N
n
(x)
(n 1)!

_
2
x
_
n
n = 0 (2.11)
Y
0
(x)
2

ln
x
2
= 1.78107241799 . . . Eulers constant (2.12)
Large Argument Limit |x| 1
J
n
(x)
_
2
x
cos
_
x

4

n
2
_
(2.13)
N
n
(x)
_
2
x
sin
_
x

4

n
2
_
(2.14)
2.1.2 Orthogonality Relationships and Fourier-Bessel Expansions
Bessel equation (2.8) can be written in the following form:
(y

+
_

2

n
2

_
y = 0 (2.15)
If you compare (2.15) with (1.5), you can see that it is a Sturm-Liouville equation with p() = , q() =
n
2

,
w() = , and =
2
. With appropriate boundary conditions on a nite interval such as [a, b] we can
have a Sturm-Liouville eigenvalue problem (regular if a > 0 and irregular if a = 0). Therefore, all the
properties of Sturm-Liouville eigenfunctions and eigenvalues will be applicable to this equation. Here we
only consider two irregular cases [?].
First consider (2.15) on the interval 0 b with boundary condition y(b) = 0. This is an irregular
problem and we have to impose an extra condition at 0 in order to have orthogonal eigenfunctions.
This extra condition is that the solution and its derivative must be bounded as 0. Since N
n
()
as 0, therefore, we can not have N
n
() in our solution. Consequently, eigenfunctions must be in the
form of J
n
() and since y(b) = 0 we obtain the eigenvalues:
J
n
(b) = 0 =
m
=

nm
b
=
m
= (
m
)
2
=
_

nm
b
_
2
(2.16)
Amir Borji Special Functions and Laplace Equation 10
3
1
0.5
0
0.5
1
0 2 4 6 8 10 12 14
2
2.5
1.5
N
2
(x)
N
1
(x)
N
0
(x)
Figure 2.2: Bessel functions of the second kind
in which
nm
is the m
th
root of the Bessel function J
n
(x) = 0, i.e. J
n
(
nm
) = 0. These eigenvalues are
all real and have all the properties that we explained for Sturm-Liouville problem. We have the following
orthogonality property over the interval [0, b] with respect to weight function w() = :
_
b
0
J
n
_

nm
b

_
J
n
_

nk
b

_
d =
_
_
_
0, m = k
b
2
2
[J
n+1
(
nm
)]
2
, m = k
(2.17)
Any piecewise continuous function f() for which f(b) = 0 and is dened on the interval [0, b] can be
expanded in a series of above eigenfunctions:
f() =

m=1
A
m
J
n
_

nm
b

_
(2.18)
in which the coecients A
m
are obtained by using the orthogonality property (2.17)
A
m
=
2
b
2
[J
n+1
(
nm
)]
2
_
b
0
f()J
n
_

nm
b

_
d (2.19)
Expression (2.18) is called the Fourier-Bessel series expansion of f().
As another case, consider (2.15) on the interval 0 b with boundary condition y

(b) = 0. This
is an irregular problem and we have to impose an extra condition at 0 in order to have orthogonal
eigenfunctions. This extra condition is that the solution and its derivative must be bounded as 0. Since
N
n
() as 0, therefore, we can not have N
n
() in our solution. Consequently, eigenfunctions
must be in the form of J
n
() and since y

(b) = 0 we obtain the eigenvalues:

J

n
(b) =
dJ
n
()
d

=b
= 0 =
m
=

nm
b
=
m
= (
m
)
2
=
_

nm
b
_
2
(2.20)
in which

nm
is the m
th
root of the derivative of Bessel function J

n
(x) = 0, i.e. J

n
(

nm
) = 0. We have the
following orthogonality property over the interval [0, b] with respect to weight function w() = :
_
b
0
J
n
_

nm
b

_
J
n
_

nk
b

_
d =
_
_
_
0, m = k
b
2
2
_
1
n
2

2
nm
_
[J
n
(

nm
)]
2
, m = k
(2.21)
Amir Borji Special Functions and Laplace Equation 11
Any piecewise continuous function f() for which f

(b) = 0 and is dened on the interval [0, b] can be

expanded in a series of above eigenfunctions:
f() =

m=1
A
m
J
n
_

nm
b

_
(2.22)
in which the coecients A
m
are obtained by using the orthogonality property (2.21)
A
m
=
2
b
2
_
1
n
2

2
nm
_
[J
n
(

nm
)]
2
_
b
0
f()J
n
_

nm
b

_
d (2.23)
If the interval is [a, b] and a > 0, then the general form of eigenfunctions would be A
m
J
n
(
m
)+B
m
N
n
(
m
).
The boundary conditions at = a and = b will determine the eigenvalues
m
and we have similar
orthogonality property between the eigenfunctions as well. However, we will not study this case further.
2.1.3 Recursion Relationships and Wronskians
Consider Z
n
(x) to be either of J
n
(x) or N
n
(x) or any linear combination of these two functions. Then, the
following recursive formulas are applicable (n can be any number):
Z
n1
(x) + Z
n+1
(x) =
2n
x
Z
n
(x) (2.24)
Z
n1
(x) Z
n+1
(x) = 2Z

n
(x) (2.25)
Z

n
(x) +
n
x
Z
n
(x) = Z
n1
(x) (2.26)
Z

n
(x)
n
x
Z
n
(x) = Z
n+1
(x) (2.27)
(x
n
Z
n
(x))

= x
n
Z
n1
(x) (2.28)
_
x
n
Z
n
(x)
_

= x
n
Z
n+1
(x) (2.29)
in particular J

0
(x) = J
1
(x). Equation (2.28) and (2.29) are very useful when integrating over Bessel
functions.
J
n
(x)Y
n+1
(x) J
n+1
(x)N
n
(x) = J

n
(x)N
n
(x) J
n
(x)Y

n
(x) =
2
x
(2.30)
2.1.4 Series and Integral Relationships
exp
_
x
2
_
t
1
t
__
=
+

n=
J
n
(x) t
n
(2.31)
e
ixcos
=
+

n=
i
n
J
n
(x)e
in
(2.32)
_
Z
n
(x)Z
n
(x)xdx = x
Z
n
(x)Z
n1
(x) Z
n1
(x)Z
n
(x)

2
(2.33)
_
Z
2
n
(x)xdx =
x
2
2
_
Z
2
n
(x) Z
n1
(x)Z
n+1
(x)

(2.34)
_
x
n+1
Z
n
(x) dx = x
n+1
Z
n+1
(x) (2.35)
J
n
(x) =
1
2
_

e
ixsin in
d =
1

_

0
cos(n xsin ) d (2.36)
Z
n
(x) can be any of J
n
(x) or N
n
(x) or AJ
n
(x) + BN
n
(x)