6.

Bessel Functions
6.1 Bessel Functions of the First Kind,
) (x J
v
Bessel functions appear in a wide variety of physical problems. For example,
separation of the Helmholtz or wave equation in circular cylindrical coordinates leads to
Bessels equation.
Generating function, integer order, J
n
(x)
Although Bessel functions are of interest primarily as solutions of differential
equations, it is instructive and convenient to develop them from a completely different
approach, that of the generating function. Let us introduce a function of two variables,
. ) , (
) 1 )( 2 ( t t x
e t x g

(6.1)
Expanding it in a Laurent series, we obtain
. ) (
) 1 )( 2 (

n
n
n
t t x
t x J e (6.2)
The coefficient of t
n
, J
n
, is defined to be Bessel function of the first kind of integer
order n. Expanding the exponentials, we have
.
! 2
) 1 (
! 2
0 0
2 2

,
_

,
_

s
s
s
s
r
r
r
t x xt
s
t x
r
t x
e e (6.3)
Setting n =r- s, yields
.
!
)
2
( ) 1 (
)! ( 2
0

+
+

,
_

s s n
s
s s
s n
s n
s
t x
s n
t x
(6.4)
Since

[ ] , 0
)! (
1

< s n
s n
(note:
0
)! (
1

m
for a positive integer m)
we have
{ } { } { }.
0 0 0

n s s n s s n
The coefficient of t
n
is then
1

.
)! 1 ( 2 ! 2
)
2
(
)! ( !
) 1 (
) (
2
2
0
2
+
+

+

+
+

n
x
n
x x
s n s
x J
n
n
n
n
s
s n
s
n (6.5)
This series form exhibits the behavior of the Bessel function J
n
for small x. The results
for J
0
, J
1
, and J
2
are shown in Fig.6.1. The Bessel functions oscillate but are not periodic.
Figure 6.1 Bessel function,
) (
0
x J
, ) (
1
x J , and ) (
2
x J .
Eq.(6.5) actually holds for n < 0 , also giving

0
2
)
2
(
)! ( !
) 1 (
) (
s
n s
s
n
x
n s s
x J
. (6.6)
Since the terms for n<s (corresponding to the negative integer (s-n) ) vanish, the series
can be considered to start with s=n. Replacing s by s + n, we obtain
). ( ) 1 ( )
2
(
)! ( !
) 1 (
) (
0
2
x J
x
n s s
x J
n
n
s
n s
n s
n

+

+
+

(6.7)
These series expressions may be used with n replaced by v to define J
v
and J
-v
for non-
integer v.
Recurrence relations
Differentiating Eq.(6.1) partially with respect to t, we find that
, ) ( )
1
1 (
2
1
) , (
1 ) 1 )( 2 (
2

n
n
n
t t x
t x nJ e
t
x t x g
t
(6.9)
and substituting Eq.(6.2) for the exponential and equating the coefficients of t
n-1
, we
obtain
). (
2
) ( ) (
1 1
x J
x
n
x J x J
n n n
+
+
(6.10)
2
This is a three-term recurrence relation.
On the other hand, differentiating Eq.(6.1) partially with respect to x, we have
. ) ( )
1
(
2
1
) , (
) 1 )( 2 (

n
n
n
t t x
t x J e
t
t t x g
x
(6.11)
Again, substituting in Eq.(6.2) and equating the coefficients of t
n
, we obtain the result
). ( 2 ) ( ) (
1 1
x J x J x J
n n n

+
(6.12)
As a special case,
). ( ) (
1 0
x J x J
(6.13)
Adding Eqs.(6.10) and (6.12) and dividing by 2, we have
). ( ) ( ) (
1
x J x J
x
n
x J
n n n
+

(6.14)
Multiplying by x
n
and rearranging terms produces
[ ] ). ( ) (
1
x J x x J x
dx
d
n
n
n
n

(6.15)
Subtracting Eq.(6.12) from (6.10) and dividing by 2 yields
). ( ) ( ) (
1
x J x J
x
n
x J
n n n

+
(6.16)
Multiplying by x
-n
and rearranging terms, we obtain
[ ] ). ( ) (
1
x J x x J x
dx
d
n
n
n
n
+

(6.17)
Bessels differential equation
Suppose we consider a set of functions
) (x Z
v
which satisfies the basic recurrence
relations (Eq. (6.10) and (6.12)), but with v not necessarily an integer and
) (x Z
v
not
necessarily given by the series (Eq. (6.5)). Equation (6.14) may be rewritten (nv) as
). ( ) ( ) (
1
x vZ x xZ x Z x
v v v



(6.18)
3
On differentiating with respect to x, we have
. 0 ) 1 ( ) (
1 1
+ +
v v v v
Z Z x Z v x Z x
(6.19)
Multiplying by x and subtracting Eq (6.18) multiplied by v given us
. 0 ) 1 (
1
2
1
2 2
+ +
v v v v v
Z x xZ v Z v Z x Z x (6.20)
Now we rewrite Eq. (6.16) and replace n by v-1.
. ) 1 (
1 1 v v v
xZ Z v Z x

(6.21)
Using this to eliminate
1 v
Z
and
1

v
Z
from Eq. (6.20), we finally get
. 0 ) (
2 2 2
+ +
v v v
Z v x Z x Z x (6.22)
This is just Bessels equation. Hence any function,
) (x Z
v
, that satisfy the recurrence
relations (Eq. (6.10) and (6.12), (6.14) and (6.16), or (6.15) and (6.17)) satisfy Bessels
function; that is,the unknown
v
Z
are Bessels function. In particular, we have shown
that the functions J
n
(x), defined by our generating funcition, satisfy Bessels function. If
the argument is k rather than x, Eq. (6.22) becomes
. 0 ) ( ) ( ) ( ) (
2 2 2
2
2
2
+ +

k Z v k k Z
d
d
k Z
d
d
v v v
(6.22)
Integral representation

A particular useful and powerful way of treating Bessel functions employs integral
representations. If we return to the generating function (Eq. (6.2)), and substitute t=e
i
,
), 3 sin ) ( sin ) ( ( 2
) 4 cos ) ( 2 cos ) ( ( 2 ) (
3 1
4 2 0
sin

+ + +
+ + +

x J x J i
x J x J x J e
ix
(6.23)
in which we have used the relations
, sin ) ( 2
) )( ( ) ( ) (
1
1 1 1

x iJ
e e x J e x J e x J
i i i i

(6.24)
, cos ) ( 2 ) ( ) (
2
2
2
2
2


x J e x J e x J
i i
+

and so on.

In summation notation
4

[ ], ) 1 2 cos( ) ( 2 ) sin sin(
, ) 2 cos( ) ( 2 ) ( ) sin cos(
1
1 2
1
2 0

+
n
n
n
n
n x J x
n x J x J x


(6.25)
equating real and imaginary parts, respectively. It might be noted that angle (in radius)
has no dimensions. Likewise sin has no dimensions and function cos(xsin) is perfectly
proper from a dimensional point of view.
By employing the orthogonality properties of cousine and sine,

nm
d m n

2
cos cos
0

(6.26a)

nm
d m n

2
sin sin
0

(6.26b)
in which n and m are positive integers (zero is excluded), we obtain

'

odd n
even n x J
d n x
n
0
) (
cos ) sin cos(
1
0

(6.27)

'

odd n x J
even n
d n x
n
) (
0
sin ) sin sin(
1
0

(6.28)
If these two equations are added together

[ ]
, ) sin cos(
1
sin ) sin sin( cos ) sin cos(
1
) (
0
0

d x n
d n x n x x J
n

, 3 , 2 , 1 , 0 n
(6.29)
As a special case,
. ) sin cos(
1
) (
0
0

d x x J (6.30)
Nothing that
) sin cos( x
repeats itself in all four quadrants (
2 1
, ,
+
4 4
, ), we may write Eq. (6.30) as
. ) sin cos(
2
1
) (
2
0
0

d x x J (6.30a)
On the other hand,
) sin sin( x
reverses its sign in the third and fourth quadrants so that
. 0 ) sin sin(
2
1
2
0

d x (6.30b)
Adding Eq. (6.30a) and i times Eq. (6.30b), we obtain the complex exponential
representation
.
2
1
2
1
) (
2
0
cos
2
0
sin
0

d e d e x J
ix ix
(6.30c)
5
This integral representation (Eq. (6.30c)) may be obtained somewhat more directly by
employing contour integration.
Example 6.11 Fraunhofer Diffraction, Circular Aperture
In the theory of diffraction through a circular aperture we encounter the integral
. ~
0
2
0
cos

a
ibr
rdr d e

(6.31)

for , the amplitude of the diffracted wave. Here is an azimuth angle in the plane of
the circular aperture of radius a, and

, is the angle defined by a point on a screen

below the circular aperture relative to the normal through the center point. The parameter
b is given by

sin
2
b (6.32)
with the wavelength of the incident wave. The other symbols are defined by Fig. 6.2
From Eq. (6.30c), we get
. ) ( 2 ~
0
0

a
rdr br J (6.33)
Figure 6.2 Fraunhofer diffraction circular aperture
Equation (6.15) enables us to integrate Eq. (6.33) immediately to obtain
). sin
2
(
sin
~ ) (
2
~
1 1 2

a
J
a
ab J
b
ab
(6.34)
6
The intensity of the light in the diffraction pattern is proportional to
2

and

( ) [ ]
2
1 2
sin
sin 2
~

'

a J
. (6.35)
6.2 Orthogonality
If Bessels equation, Eq.(6.22a), is divide by

, we see that it becomes self-adjoint(or

Hermitian). Therefore, by the theory presented in Chapter 4, the solutions are expected to
be orthogonal, provided that appropriate boundary conditions are satisfied. For v> 0, J
v
(0)=0. Thus, for a finite interval [0, a], when
vm

is the mth zero of J

v
(i.e.
0 ) (
vm v
J
), we are able to have
if m n, . 0 ) ( ) (
0

a
vn v vm v
d
a
J
a
J

(6.49)
This gives us orthogonality over the interval [0, a].
Normalization
The normalization result may be written as
[ ] . ) (
2
) (
2
1
2
0
2
vm v
a
vm v
J
a
d
a
J

1
]
1

(6.50)
Bessel series
If we assume that the set of Bessel functions
) ( a J
vm v

(v fixed, m=1,2, ) is
complete, then any well-behaved function
) ( f
may be expanded in a Bessel series

1
) ( ) (
m
vm v vm
a
J c f


,
, 0 a
. 1 > v (6.51)
The coefficients c
vm
are determined by using Eq.(6.50),
[ ]
. ) ( ) (
) (
2
0
2
1
2
+

a
vm v
vm v
vm
d
a
J f
J a
c

(6.52)
Continuum form
7
If a , then the series forms may be expected to go over into integrals. The discrete
roots become a continuous variable

. A key relation is the Bessel function closure

equation
), (
1
) ( ) (
0

d J J
v v
.
2
1
> v (6.59)
Figure 6.3 Neumann functions , ) (
0
x N , ) (
1
x N , and ) (
2
x N .
6.3 Neumann function, Bessel function of the second kind,
) (x N
v

From the theory of the differential equations it is known that Bessels equation has
two independent solutions, Indeed, for nonintegral order v we have already found two
solutions and labeled them
) (x J
v
and
) (x J
v
, using the infinite series (Eq. (6.5)).
The trouble is that when v is integral Eq.(6.8) holds and we have but one independent
solution. A second solution may be developed by the method of Section 3.6. This yields
a perfectly good solution of Bessels equation but is not the usual standard form.
Definition and series form
As an alternate approach, we have the particular linear combination of
) (x J
v
and
) (x J
v
.
sin
) ( ) ( cos
) (

v
x J x J v
x N
v v
v

(6.60)
This is Neumann function (Fig. 6.3). For nonintegral v ,
) (x N
v
clearly satisfies
Bessels function, for it is a linear combination of known solutions,
) (x J
v
and
) (x J
v
.
To verify that
) (x N
v
, our Neumann function or Bessel function of the second kind,
actually does satisfy Bessels equation for integral n , we may process as follows.
8
LHospitals rule applied to Eq. (6.60) yields

[ ]
[ ]
.
) (
) 1 (
) ( 1
cos
cos ) ( sin
sin ) (
) ( ) ( cos ) (
) (
n v
v n v
n v
v v n
n v
v v
n
v
x J
v
x J
n
v J v J n x J n
v dv d
x J x J v dv d
x N

1
]
1

(6.65)
Differentiating Besself equation for ) (x J
t
with repect to v , we have
. 2 ) ( ) ( ) (
2 2
2
2
2
v
v v v
vJ
v
J
v x
v
J
dx
d
x
v
J
dx
d
x
t
t t t

(6.66)
Multiplying the equation for
) (x J
v
by (-1)
v
, substracting from the equation for
) (x J
v
(as suggested by Eq. (6.65)), and taking the limit n v , we obtain
[ ]. ) 1 (
2
) (
2 2
2
2
2
n
n
n n n n
J J
n
N v x N
dx
d
x N
dx
d
x

+ +

(6.67)

For n v , an integer, the right-hand side vanishes by Eq. (6.8) and
) (x N
n
is seen to
be a solution of Bessels equation. The most general solution for any v can be written as

). ( ) ( ) ( x BN x AJ x y
v v
+
(6.68)
Example Coaxial Wave Guides
We are interested in an electromagnetic wave confined between the concentric ,
conducting cylindrical surfaces
a
and
b
. Most of the mathematics is worked
out in Section 3.3. From EM knowledge,
. 0
2 2 2
+
z z
E c E (
Z
E : electrical field along z axis)
Let
ikz im
z
e e P E

) ( , we have

[ ] . 0 ) ( ) (
2 2 2 2
+ P m k c
d
dP
d
d

This is the Bessel equation. If

finite P ) 0 (
, the solution is
) (
m
J
with
2 2 2 2
k c . But , for the coaxial wave guide one generalization is needed. The
origin
0
is now excluded (
b a 0
). Hence the Neumann function
) (
m
N
may
not be excluded. ) , , , ( t z E
z
becomes
9
[ ] . ) ( ) (
,
) (

t
+
n m
t kz i im
mn mn m mn z
e e N c J b E


(6.79)
With the condition
, 0
z
H (6.80)
we have the basic equatios for a TM (transverse magnetic ) wave.
The (tangential) electric field must vanish at the conducting surfaces (Direchlet
boundary condition) or

0 ) ( ) ( + a N c a J b
mn mn m mn

(6.81)

0 ) ( ) ( + b N c b J b
mn mn m mn

(6.82)
these transcendental equations may be solved for
) (
mn

and the ratio
mn mn
b c
. From
the relation
2
2
2
2

c
k (6.83)
and since
2
k must be positive for a real wave, the minimum frequency that will be
propagated (in this TM mode) is
, c
(6.84)
with

fixed by the boundary conditions, Eqs. (6.81) and (6.82). This is the cutoff
frequency of the wave guide.
6.4 Hankel function

Many authors perfer to introduce the Hankel functions by means of integral
representations and then use them to define the Neumann function,
) (z N
m
. An outline
of this approach is given at the end of this section.
Defintions
As we have already obtained the Neumann function by more elementary (and less
powerful) techniques, we may use it to define the Hankel functions, ) (
) 1 (
x H
v
and
) (
) 2 (
x H
v
:
) ( ) ( ) (
) 1 (
x iN x J x H
v v v
+ (6.85)
) ( ) ( ) (
) 2 (
x iN x J x H
v v v
(6.86)
This is exactly analogous to taking
. sin cos

i e
i
t
t
(6.87)
For real arguments
) 1 (
v
H and
) 2 (
v
H are complex conjugates. The extent of the analogy
will be seen better when the asymptotic forms are considered . Indeed, it is their
10
asymptotic behavior that makes the Hankel functuions useful!
6.5 Modified Bessel function ,
) (x I
v
and
) (x K
v
The Helmholtz equation,
0
2 2
+ k ,
separated in circular cylindrical coordinates, leads to Eq. (6.22a), the Bessel equation.
Equation (6.22a) is satisfied by the Bessel and Neumann functions
) ( k J
v
and
) ( k N
v
and any linear combination such as the Hankel functions ) (
) 1 (
k H
v
and
) (
) 2 (
k H
v
. Now the Helmholtz equation describes the space part of wave phenomena. If
instesd we have a diffusion problem, then the Helmholtz equation is replaced by
0
2 2
k . (6.88)
The analog to Eq. (6.22a) is
. 0 ) ( ) ( ) ( ) (
2 2 2
2
2
2
+ +

k Y v k k Y
d
d
k Y
d
d
v v v
(6.89)
The Helmholtz equation may be transformed into the diffusion equation by the
transformation ik k . Similarly, ik k changes Eq. (6.22a) into Eq. (6.89) and shows
that

) ( ) ( ik Z k Y
v v

.
The solution of Eq. (6.89) are Bessel function of imaginary argument. To obtain a
solution that is regular at the origin, we take
v
Z
as the regular Bessel function
v
J
. It is
customary (and convenient) to choose the normalization so that
). ( ) ( ) ( ix J i x I k Y
v
v
v v

(6.90)
(Here the variable
k
is being replaced by x for simplicity.) Often this is written as
). ( ) (
2 2 i
v
iv
v
xe J e x I

(6.91)
Series form
In the terms of infinite series this is equivalent to removing the
s
) 1 ( sign in Eq.
(6.5) and writing
, )
2
(
)! ( !
1
) (
, )
2
(
)! ( !
1
) (
0
2
0
2

s
v s
v
s
v s
v
x
v s s
x I
x
v s s
x I
(6.92)
The extra
v
i

normalization cancels the

v
i from each term and leaves
) (x I
v
real. For
11
integral v this yields

). ( ) ( x I x I
n n

(6.93)
Recurrence relations
The recurrence relations satisfied by
) (x I
v
may be developed from the series
expansions, but it is easier to work from the existing recurrence relations for
) (x J
v
. Let
us replace x by ix and rewrite Eq. (6.90) as

). ( ) ( ix I i x J
v
v
v
(6.94)
Then Eq. (6.10) becomes
). (
2
) ( ) (
1
1
1
1
ix I i
x
v
ix I i ix I i
v
v
v
v
v
v
+
+
+

Repalcing x by ix, we have a recurrence relation for

) (x I
v
,
). (
2
) ( ) (
1 1
x I
x
v
x I x I
v v v

+
(6.95)
Equation (6.12) transforms to
). ( 2 ) ( ) (
1 1
x I x I x I
v v v
+
+
(6.96)
From Eq. (6.93) it is seen that we have but one independent solution when v is an
integer, exactly as in the Bessel function
v
J
The choice of a second, independent
solution of Eq. (6.108) is essentially a matter od convenience.
We choose to define a second solution in terms of the Hankel function
) 1 (
v
H by
[ ]. ) ( ) (
2
) (
2
) (
1 ) 1 ( 1
ix iN ix J i ix H i x K
v v
v
v
v
v
+
+ +

(6.97)
The factor
1 + v
i makes
) (x K
v
rela when x is real. Using Eqs. (6.60) and (6.90), we may
transform Eq. (6.97) to
,
sin
) ( ) (
2
) (

v
x I x I
x K
v v
v

(6.98)
analogous to Eq. (6.60) for
) (x N
v
. The choice od Eq. (6.97) as a definition is
somewhat unfortunate in that the function
) (x K
v
does not satisfy the same recurrence
relations as
) (x I
v
. To avoid this annoyance other authors have included an additional
factor of cosine n . This permits
) (x K
v
satisfy the same recurrence relations as
) (x I
v
, but it has the disadvantage of making
0 ) ( x K
v
for , , ,
2
5
2
3
2
1
v .
12
To put the modified Bessel functions
) (x I
v
and
) (x K
v
in proper perspective, we
introduce them here because:
1. These functions are solutions of the frequently encountered modified Bessel
equation.
2. They are needed for specific physical problems such as diffusion problems.
Figure 6.4 Modified Bessel functions
6.6 Asmptotic behaviors
Frequently in physical problems there is a need to know how a given Bessel or
modified Bessel functions for large values of argument, that is, the asymptotic behavior.
Using the method of stepest descent studied in Chapter 2, we are able to derive the
asymptotic behaviors of Hankel functions (see page 450 in the text book for details) and
related functions:
1.

'

1
]
1

+
2
)
2
1
( exp
2
) (
) 1 (

v z i
z
z H
v

. 2 arg < < z
(6.99)
2. The second kind Hankel function is just the complex conjugate of the first (for real argument),

'

1
]
1

+
2
)
2
1
( exp
2
) (
) 2 (

v z i
z
z H
v

. arg 2 < < z
(6.100)
3. Since
) (z J
v
is the real part of ) (
) 1 (
z H
v
,

1
]
1

+
2
)
2
1
( cos
2
) (

v z
z
z J
v

. arg < < z
(6.101)
13
4. The Neumann function is the imaginary part of ) (
) 1 (
z H
v
, or

1
]
1

+
2
)
2
1
( sin
2
) (

v z
z
z N
v
(6.102)
5. Finally, the regular hyperbolic or modified Bessel function
) (z I
v
is given by
) ( ) ( iz J i z I
v
v
v

(6.103)
or

z
e
z I
z
v
2
) ( .
2
arg
2

< < z (6.104)
14