2401 Dierential Equations: Bessel Functions Page 1

Bessels Equation
There are many special functions which arise as solutions to dierential equations (Hermite, Legendre, Chebyshev, etc.).
Here we will look at how one important class of functions, Bessel Functions, arise through a series solution to a dierential
equation. Bessel Functions of the First Kind are particularly important in the study of partial dierential equations, and
arise in the study of vibrating circular drumheads, heat equations, and many other areas where cylindrical symmetry is
present.
Bessels equation is a dierential equation of the form
x
2
y

+ xy

+ (x
2
v
2
)y = 0,
where v is a real number.
The point x = 0 is a singular point, since p(x) =
1
x
and q(x) = 1
v
2
x
2
are not analytic at x = 0. Recall that analytic
means there exists a Taylor series of the function about the point with nonzero radius of convergence.
Further, since xp(x) = 1 and x
2
q(x) = x
2
v
2
are analytic at x = 0, the point x = 0 is a regular singular point.
Therefore, we can try to determine at least one series solution of Bessels equation which has the form y =

n=0
a
n
x
r+n
.
Bessel Functions of the First Kind
Dierentiating y =

n=0
a
n
x
r+n
and substituting into Bessels equation, we nd
y =

n=0
a
n
x
r+n
, y

n=0
(r + n)a
n
x
r+n1
, y

n=0
(r + n)(r + n 1)a
n
x
r+n2
,
x
2
_

n=0
(r + n)(r + n 1)a
n
x
r+n2
_
+ x
_

n=0
(r + n)a
n
x
r+n1
_
+ (x
2
v
2
)
_

n=0
a
n
x
r+n
_
= 0

n=0
(r + n)(r + n 1)a
n
x
r+n
+

n=0
(r + n)a
n
x
r+n
+

n=0
a
n
x
r+n+2

n=0
v
2
a
n
x
r+n
= 0

n=0
(r + n)(r + n 1)a
n
x
r+n
+

n=0
(r + n)a
n
x
r+n
+

n=2
a
n2
x
r+n

n=0
v
2
a
n
x
r+n
= 0
_

n=0
(r + n)(r + n 1)a
n
x
n
+

n=0
(r + n)a
n
x
n
+

n=2
a
n2
x
n

n=0
v
2
a
n
x
n
_
x
r
= 0

n=0
(r + n)(r + n 1)a
n
x
n
+

n=0
(r + n)a
n
x
n
+

n=2
a
n2
x
n

n=0
v
2
a
n
x
n
= 0
_
r(r 1)a
0
+ ra
0
v
2
a
0

x
0
+
_
(r + 1)ra
1
+ (r + 1)a
1
v
2
a
1

x
1
+

n=2
(r + n)(r + n 1)a
n
x
n
+

n=2
(r + n)a
n
x
n
+

n=2
a
n2
x
n

n=2
v
2
a
n
x
n
= 0
2401 Dierential Equations: Bessel Functions Page 2
_
r(r 1)a
0
+ ra
0
v
2
a
0

x
0
+
_
(r + 1)ra
1
+ (r + 1)a
1
v
2
a
1

x
1
+

n=2
_
(r + n)
2
a
n
+ a
n2
v
2
a
n

x
n
= 0
We want this to be true for all values of x. Therefore, the coecients of powers of x must be zero. This leads to the
equations:
r(r 1)a
0
+ ra
0
v
2
a
0
= 0 (1)
(r + 1)ra
1
+ (r + 1)a
1
v
2
a
1
= 0 (2)
(r + n)
2
a
n
+ a
n2
v
2
a
n
= 0, n = 2, 3, 4, . . . (3)
These equations are the recursion and indicial equation. We can choose any one of them as the indicial equation; lets
choose the indicial equation to be Eq. (1),
r(r 1)a
0
+ ra
0
v
2
a
0
= 0.
If we assume a
0
= 0, the solutions to the indicial equation are r = v. We can get one solution to Bessels equation if we
choose to work with the larger root from the indicial equation, so lets assume r = v > 0 and proceed.
The recursions equations (Eq. (2) and (3)) become
(2v + 1)a
1
= 0 (4)
a
n
=
a
n2
n(n + 2v)
, n = 2, 3, 4, . . . (5)
Equation (4) tells us that a
1
= 0, unless v = 1/2, in which case the equation is satised regardless of the value of a
1
.
Since we have assumed that v > 0, this situation does not arise, and we may safely assume a
1
= 0.
We can now use Eq (5) to generate the rst few a
n
, and then try to determine the pattern.
a
0
= unspecied, does not equal zero
a
1
= 0
a
2
=
a
0
2(2 + 2v)
a
3
= 0
a
4
=
a
2
4(4 + 2v)
=
a
0
2 4(2 + 2v)(4 + 2v)
=
a
0
2
22
2!(1 + v)(2 + v)
a
5
= 0
a
6
=
a
4
6(6 + 2v)
=
a
0
2 4 6(2 + 2v)(4 + 2v)(6 + 2v)
=
a
0
2
23
3!(1 + v)(2 + v)(3 + v)
This is enough for us to recognize the pattern. First, lets work with (1 + v)(2 + v) (m + v). If v is an integer, we
observe that
1
(1 + v)(2 + v) (m + v)
=
1 2 3 (v 1)(v)
1 2 3 (v 1)(v)(1 + v)(2 + v) (m + v)
=
v!
(v + m)!
.
2401 Dierential Equations: Bessel Functions Page 3
If v is not an integer, we can replace the factorial with a gamma function using the relation w! = (1 + w). Therefore,
1
(1 + v)(2 + v) (m + v)
=
(1 + v)
(1 + v + m)
.
If you check this on Mathematica, you will nd that this product can also be represented in terms of the Pochhammer
symbol.
The general term in our expansion is therefore
a
2m
=
(1)
m
a
0
(1 + v)
2
2m
m! (1 + v + m)
, m = 0, 1, 2, . . . .
A solution to Bessels equation is therefore
y =

n=0
a
n
x
r+n
=

m=0
a
2m
x
v+2m
=

m=0
(1)
m
a
0
(1 + v)
2
2m
m! (1 + v + m)
x
v+2m
= a
0
x
v
(1 + v)

m=0
(1)
m
x
2m
2
2m
m! (1 + v + m)
Bessel Functions of The First Kind of order v are obtained by choosing a
0
= 1/(2
v
(1 + v)), and they are commonly
denoted J
v
(x),
J
v
(x) =
_
x
2
_
v

m=0
(1)
m
x
2m
2
2m
m! (1 + v + m)
Although we assumed in our analysis that v > 0, we can see thatJ
v
(x) is dened for v R. In fact, it can be extended to
v C. What is important in the solution to partial dierential equations is that J
v
(x) is nite at x = 0 for integer values
of v.
Here is a plot of J
v
(x), for v = 0, 1, 2, 3, 4, 5. Note that J
0
(0) = 1.
2401 Dierential Equations: Bessel Functions Page 4
Bessel Functions of the Second Kind
A second solution to Bessels equation can be found using reduction of order, since we now know a rst solution. Here we
use the formula we derived for reduction of order earlier, where we identify p(x) = 1/x from Bessels equation.
y2(x) = y
1
(x)
_
exp
_

_
p(x) dx
_
y
2
1
(x)
dx
= J
v
(x)
_
exp
_

_
(1/x) dx
_
J
2
v
(x)
dx
= J
v
(x)
_
1
xJ
2
v
(x)
dx
=

2
Y
v
(x)
where the last integral is found from a book of integrals [1] or a more in-depth study of Bessel Functions than we have
time for.
Notice that this procedure found a solution which is a constant multiple of Y
v
(x), so we can drop the constant and take
as the second solution y
2
(x) = Y
v
(x).
Here is a plot of Y
v
(x), for v = 0, 1, 2, 3, 4, 5.
Notice that lim
x0
Y
v
(x) = for integer values of v. This is a logarithmic singularity that we sometimes see with
second solutions to dierential equations about a regular singular point. This excludes, for physical reasons, Y
v
(x) from
solutions in many applications.
The derivation of the second solution using other techniques can be a dicult task, but realize that although our solution
looks simple, we had an integral that we worked out using a table of integrals, so to understand this integral we would
likely have to put in an equal amount of work.
References
[1] Gradshteyn, I. S.; Ryzhik, I. M. Table of Integrals, Series, and Products, 6th ed., Academic Press, New York, 2000.
Page 664 #6.539.1