A3.

13
Section A3.4: Power Series Solution

Many times a closed form solution is difficult (or impossible) to find for an
ordinary differential equation. A power series solution can be found under certain
conditions. Sometimes the power series solution is useful enough to warrant its own
name such as the Sine, Legendre or Bessel function. The series solution must converge
for the function to be defined. The series must uniformly converge for the solution to be
integrable or differentiable.

Topic A3.4.1: Solution Near an Ordinary Point

The solution near ordinary points is particularly easy to find. Recall the
definitions of ordinary points.
1. The point x
o
is an ordinary point of a function f(x) if that function can be Taylor
expanded at the point x
o
and in a small neighborhood about x
o
.
2. The point x
o
is an ordinary point of the equation
) x ( r y ) x ( q ' y ) x ( p ' ' y = + +
if p(x) and q(x) have Taylor expansions at x
o
.
3. The Taylor expansion is defined as
( ) ( )
n
o
x x
n
n
n
o
x
o
x x
dx
f d
! n
1
x x
dx
df
) x ( f ) x ( f
o
o

|
|
.
|

\
|
= + + =
=

Second order, ordinary differential equations have series solutions at ordinary
points. Singular points are values of x which are not ordinary points of a function.

Example A3.4.1: What are the singular points of
( ) 0 y 3 ' xy 6 ' ' y 4 x x
2 2
= + +
The point x=0 is a double root of ( ) 4 x x
2 2
+ and i 2 x = are two single roots. Note that
( ) ( ) 4 x x
6
4 x x
x 6
) x ( p
2 2 2
+
=
+
=
Therefore x=0 is a singular point of p(x). Also i 2 x = are singular points but if the
problem is restricted to real values of x, these two point are not considered.
---------

The point x
o
is an ordinary point of
0 y ) x ( q ' y ) x ( p ' ' y = + +
if x
o
is an ordinary point of p(x) and q(x). Near ordinary points, an ordinary differential
equation 0 y ) x ( q ' y ) x ( p ' ' y = + + has two power series solutions of the form

=
=
0 n
n
n
x a y
The radius of convergence is from the origin to the magnitude of the first singular point
(even if it is complex). Notice that r(x)=0 in the ODE but can be included so long as it
can be Taylor expanded. The reader can refer to Rainville Bedients book Elementary
Differential Equations as an excellent reference.

A3.14

Example A3.4.2: Solve
( ) 0 y 4 ' xy 6 ' ' y x 1
2
= or
( ) ( )
0 y
x 1
4
' y
x 1
x 6
' ' y
2 2
=

using

=
=
0 n
n
n
x a y (A3.4.1)
The region of convergence is seen to be ( ) 1 , 1 x by looking at the singular points of
the coefficient of ' y . The solution to
( ) 0 y 4 ' xy 6 ' ' y x 1
2
= (A3.4.2)
is found by substituting the power series Equation A3.4.1 into Equation A3.4.2 to get
( ) ( ) 0 x a 4 x a n x 6 x a 1 n n x 1
0 n
n
n
1 n
n
0 n
2 n
n
0 n
2
=


=

=

Multiply the coefficients
( ) ( ) | | 0 x a 4 x na 6 x a 1 n n x a 1 n n
0 n
n
n
n
n
n
n
2 n
n
0 n
= + +


=

=

which gives
( ) ( ) | | 0 x a 4 n 6 1 n n x a 1 n n
n
n
0 n
2 n
n
0 n
= + +


=

=
(A3.4.3)
Notice that, for the first summation in Equation A3.4.3, n=0 and n=1 do not contribute to
the sum, so redefine the sum starting at m=n-2. The first summation now becomes
( )( )

=
+
+ +
0 m
m
2 m
x a 1 m 2 m
or, noting m is a dummy variable and can be written in terms of n.
( )( )

=
+
+ +
0 m
n
2 n
x a 1 n 2 n
Equation A3.4.3 becomes
( )( ) ( )( ) | | 0 x a 4 n 1 n a 1 n 2 n
n
0 m
n 2 n
= + + + +

=
+
(A3.4.4)
It is a simple matter to check that the functions
{ } , x , x , 1
2

are independent. Therefore, the only way for Equation A3.4.4 to be zero is if all the
coefficients are zero (by definition of independent functions).
( )( )
( )( )
( )
( )
n n 2 n
a
2 n
4 n
a
2 n 1 n
4 n 1 n
a
+
+
=
+ +
+ +
=
+
0 n
Some people prefer to work with n 2 n +
2 n n
a
n
2 n
a

+
= 2 n (A3.4.5)
The allowed values of n show that a
o
and a
1
are undetermined. These two coefficients
lead to two separate series solutions. Equation A3.4.5 is a recursion relation for the
coefficients. A formula can be found for a
n
in terms of a
o
or a
1
by forming a table for the
A3.15
coefficients. Set k 2 n = for the terms associated with a
o
and set 1 k 2 n + = for terms
related to a
1
.

k Even k Odd

0
0 0
a a = 0
1 1
a a =
1
0 2
a
1
2
a =
1
1 3
a
3
5
a =
2
2 4
a
2
3
a =
2
3 5
a
5
7
a =
3
4 6
a
3
4
a =
3
5 7
a
7
9
a =

k
2 k 2 k 2
a
k
1 k
a

+
=
k
1 k 2 1 k 2
a
1 k 2
3 k 2
a
+
+
+
=

For the even column, multiply all of the entries to find (except for k=0)
( )( ) ( )( )
( )( ) ( )( )
0 2 2 k 2 2 2 k 2 k 2
a a a
1 2 1 k k
2 3 k 1 k
a a a

+
=
and rewrite it as
( )
( )
0 0 k 2
a 1 k a
! k
! 1 k
a + =
+
=
Similarly, the entries in the odd column can be multiplied to find
1 1 k 2
a
3
3 k 2
a
+
=
+

The solution to the ordinary differential equation (A3.4.2) consists of two terms, both of
which are derived from the assumed series solution (A3.4.1).


=
+
+

=
+ = =
0 k
1 k 2
1 k 2
0 k
k 2
k 2
0 n
n
n
x a x a x a y
Substitute to get
( )
( )


=
+

=
+
+ + =
0 k
1 k 2
1
0 k
k 2
0
x
3
3 k 2
a x 1 k a y
There are two independent solutions with the coefficients a
0
and a
1
taking the place of the
arbitrary constants.


Topic 3.4.2: Power Series and the Indicial Equation

Some ordinary differential equations
0 y ) x ( q ' y ) x ( p ' ' y = + +
can be solved by a power series technique even though there are singular points for p(x)
and q(x). The singularities of the functions p(x) and q(x) can be no worse than regular
singular points x
o
which appear in the functions p(x) and q(x) as
A3.16
o
x x
) x ( d
~ ) x ( p

and
( )
2
o
x x
) x ( c
~ ) x ( q

where d(x) and c(x) are analytic at x
o
.


Example A3.4.3: Determine the regular and singular points for
( ) ( ) ( ) 0 y 1 x 2 x ' y x ' ' y 2 x 1 x x
3 2 2
= + + +
This equation is equivalent to
( ) ( )
( )
( ) ( )
0 y
2 x 1 x x
1 x 2 x
' y
2 x 1 x x
x
' ' y
2
3
2
2
=
+
+

+
+
x=-2 is a regular singular point
x=1 is NOT a regular singular point
x=0 is a regular singular points
all other points are regular points.
------------

If x=x
o
is a regular singular point of
0 y ) x ( q ' y ) x ( p ' ' y = + +
then the substitution
o
x x x changes the equation to one with x=0 as the regular
singular point. Therefore consider the following discussion with x=0 as the regular
singular point.
If x=0 is a regular singular point of
0 y ) x ( q ' y ) x ( p ' ' y = + + (A3.4.6)
then this equation always a solution that depends on two constants c
1
and c
2
. These two
constants are solutions to the so called indicial equation which is found by substituting x
c

into the ODE. The solutions to the ODE are obtained from the power series

=
+
=
0 n
c n
n
x a y
The
c
x term allows the series to start with terms such as
+ +
x
1
x
1
2

Equation A3.4.6 has two independent solution since it is a second order ODE. The two
solutions are obtained through the two constants c
1
and c
2
. Generally the two constants
are ordered such that
2 1
c Re c Re .
The solution to the ODE
0 y ) x ( q ' y ) x ( p ' ' y = + +
with x=0 as a regular singular point is


=
+

=
+
+ =
0 n
c n
n
0 n
c n
n
2 1
x b B x a A y for
2 1
c c integer
or
| |


=
+

=
+
+ + =
0 n
c n
n
0 n
c n
n
2 1
x b B x a ) x ( Ln B A y for =
2 1
c c integer
A3.17
An example is best to illustrate the procedure procedure.

Example A3.4.4: Suppose the two roots to the indicial equation c
1
and c
2
do not differ by
an integer
2 1
c c integer. Solve
( ) 0 y 2 ' y x 1 ' ' xy 2 = + + (A3.4.7)
using

=
+
=
0 n
c n
n
x a y (A3.4.8)
First find the roots to the indicial equation by substituting x
c
into the ODE (A3.4.7).
( ) ( ) 0 x 2 cx x 1 x 1 c xc 2
c 1 c 2 c
= +


Collect terms and use only the lowest order one for the indicial equation
( ) 0 c 1 c c 2 = +
The two roots of the indicial equation is
2
1
c
1
= and 0 c
2
=
Next substitute the power series (A3.4.8) into the ODE (A3.4.7). Note that n=0 just
gives the roots to the indicial equation again. For 1 n
( )( )
1 n n
a
1 c 2 n 2 c n
3 c n
a

+ +
+
=
Obviously two separate sequence of coefficients emerge when c
1
and c
2
are substituted
for c so long as
2 1
c c . Consider two cases for the two roots.

Case 1: c=c
1
=1/2 with a
0
and a
1
arbitrary constants
( )
1 n n
a
n 2 1 n 2
5 n 2
a

+

=
Making a table similar to the previous example and multiplying entries provides
( )
( )( )( ) 1 n 2 1 n 2 3 n 2 ! n 2
a 3 1
a
n
0
n
n
+

=
The solution is


=
+

=
+
+ = =
1 n
c n
n
c
0
0 n
c n
n 1
x a x a x a y
The term
c
0
x a was separated from the summation since the recursion relation is only for
1 n . The first solution is therefore
( )
( )( )( )
(

+ = + = =


=
+
=
+

=
+
1 n
n
2 / 1 n
0
n
2 / 1
0
1 n
c n
n
c
0
0 n
c n
n 1
1 n 2 1 n 2 3 n 2 ! n 2
x a 3 1
x a x a x a x a y
A3.18
Case 2: c=c
2
=0
The recursion relation is
( )( ) ( )
1 n 1 n n
a
1 n 2 n
3 n
a
1 c 2 n 2 c n
3 c n
a

=
+ +
+
=
which gives the following table
0 1
a
1 1
2
a

=
1 2
a
3 2
1
a

=

( )
1 n n
a
1 n 2 n
3 n
a

=

Note that a
3
=0 terminates the sequence of coefficients. Therefore
0 1
a 2 a =
0 2
a
3
1
a = 3 n 0 a
n
=
The solution is
2
0 0 0
1 n
c n
n
c
0
0 n
c n
n 2
x a
3
1
x a 2 a x a x a x a y + + = + = =


=
+

=
+

or
|
.
|

\
|
+ + =
2
0 2
x
3
1
x 2 1 ' a y
where the prime is inserted on a
o
to distinguish it from the a
o
in the y
1
solution. The total
solution is y=y
1
+y
2
.

Topic 3.4.3: Indicial Equations with Equal Roots

As noted in the previous topic, if there are two distinct roots to the indicial
equation then there are obviously two series solutions to the equation since the
coefficients depend on the roots
( ) c a a
n n
=
For equal roots of the indicial equation, say c=a for example, substitute
) c , x ( y x a y
0 n
c n
n
=

=
+

into the ODE L[y]=0 and consider c to be a variable. Obtain
( ) a , x y y
1 1
= (A3.4.9)
as usual. The second solution is obtained by using
a c
1
2
c
y
y
=

= (A3.4.10)

Example A3.4.5: Solve
( ) 0 y x 2 1 ' xy 3 ' ' y x Ly
2
= + + = (A3.4.11)
A3.19
Start by substituting
c
x y = to find the indicial equation
( ) 0 1 c 3 1 c c = + +
The two roots are identical 1 c = . Next, substitute the assumed series solution

=
+
=
0 n
c n
n
x a y
into Equation A3.4.11.
( )( ) ( ) | |

=
+ + + + +
+ + + + + =
0 n
1 c n
n
c n
n
c n
n
c n
n
x a 2 x a x c n a 3 x 1 c n c n a 0 (A3.4.12)
n=0 gives the indicial equation. This last equation can be rewritten by setting n 1 n +
in the
1 c n
x
+ +
term to get
( ) ( ) | |


=
+

=
+
+ + + + =
1 n
c n
1 n
0 n
c n 2
n
x a 2 x 1 c n 2 c n a 0
The recursion relation is seen to be
( ) ( ) 1 c n 2 c n
a 2
a
2
1 n
n
+ + + +
=

(A3.4.13)
The first solution is found by setting 1 c = into Equations A3.4.13 and
+ + + = + = = = =


=

=
+
x a a
x
a
x a
x
a
x a ) 1 c , x ( y ) c , x ( y
2 1
0
1 n
1 n
n
0
0 n
c n
n 1 1 1

Notice the term 1/x matches the 1/x term in p(x). The second solution is found by
differentiating the first solution with respect to c as required by Equation A3.4.10.
1 c
1 n
c n
n
0
1 c
1
2
x a
x
a
c c
y
) x ( y
=

=
+
= )
`

=


Often a trick is required to evaluate y
2
. If u=fg then
u
g
' g
f
' f
dx
du
(

+ =
Use
) x ( Ln c c
e x u = = so that
) x ( Ln x ) x ( Ln e
c
u
c ) x ( Ln c
= =