CHAP6

Download as pdf or txt
Download as pdf or txt
You are on page 1of 42

CHAPTER 6

SERIES SOLUTIONS OF DIFFERENTIAL EQUATIONS

In the previous chapters we saw that a homogeneous linear differential equation whose
coefficients are constant can be solved by algebraic methods, and the solutions are
elementary functions known from calculus. However, if those coefficients are not
constant but depend on x , the situation is more complicated and the solutions may be
non-elementary functions. Bessel’s equation, Legendre’s equation and the
hypergeometric equation for example are of this type. Since these and other equations
and their solutions play an important role in science and engineering mathematics, we
shall now consider a method for solving such equations. The solutions will appear in the
form of power series. For this reason, the method is called the power series method.

6.1 Power Series Solutions About An Ordinary point

A. Basic Concepts and Results


Consider the second-order homogeneous linear differential equation
a 0 ( x ) y ′′ + a1 ( x ) y ′ + a 2 (x ) y = 0 , (6.1)

and suppose that this equation has no solution that is expressible as a linear combination
of known elementary functions. Let us assume that it does have a solution that can be
expressed in the form of an infinite series. Specifically, we assume that it has a solution
expressible in the form

c0 + c1 ( x − x 0 ) + c 2 ( x − x0 ) + ... = ∑ c n ( x − x0 ) ,
2 n
(6.2)
n =0

where c0 , c1 , c 2 , ... are constants called the coefficients of the series, x 0 is a constant

called the centre of the series, and x is a variable. An expression of the form (6.2) is
called a power series in ( x − x0 ) . We thus assumed that the equation (6.1) has a so-called
power series solution of the form (6.2). Assuming that this assumption is valid, we can
proceed to determine the coefficients c0 , c1 , c 2 , ... in (6.2). in such a manner that the

195
expression (6.2) does satisfy the equation (6.1). To see the validity of the assumption of
the existence of the solution of the equation (6.1), we shall first introduce certain
definitions. We write the differential equation (6.1) in the equivalent normalized form
y ′′ + P1 ( x ) y ′ + P2 ( x ) y = 0 , (6.3)
where
a1 ( x ) a2 (x )
P1 ( x ) = and P2 ( x ) =
a0 (x ) a0 (x )

Definition 6.1
A function f is said to be analytic at x 0 if its Taylor series about x 0 ,

f (n ) ( x0 )
∑ ( x − x 0 )n
n =0 n!
exists and converges to f ( x ) for all x in some open interval including x 0 .

We note that all polynomial functions are analytic everywhere; so also are the functions
with values e x , sin x , and cos x . A rational function is analytic except at those values of
x at which its denominator is zero.

Definition 6.2
The point x 0 is called an ordinary point of the differential equation (6.1) if both of the

functions P1 and P2 in the equivalent normalized equation (6.3) are analytic at x 0 . If

either (or both) of these functions is not analytic at x 0 , then x 0 is called a singular point

of the differential equation (6.1).

Example 6.1
Consider the differential equation
( )
y ′′ + xy ′ + x 2 + 2 y = 0 . (6.4)

196
Here P1 ( x ) = x and P2 ( x ) = x 2 + 2 . Both of the functions P1 and P2 are polynomial
functions and so they are analytic everywhere. Thus all points are ordinary points of this
differential equation.

Example 6.2
Consider the differential equation

(x − 1) y ′′ + xy ′ + 1 y = 0 . (6.5)
x
We express (6.5) in the normalized form, obtaining
x 1
y ′′ + y′ + y = 0.
x −1 x( x − 1)
Here

P1 ( x ) = and P2 ( x ) =
x 1
.
x −1 x( x − 1)

The function P1 is analytic, except at x = 1 , and P2 is analytic, except at x = 0 and


x = 1 . Thus x = 0 and x = 1 are singular points of the differential equation under
consideration. All other points are ordinary points. We note that x = 0 is a singular point,
even though P1 is analytic at x = 0 since both P1 and P2 must be analytic at x 0 in order

for x 0 to be an ordinary point.

Theorem 6.1
Hypothesis: The point x 0 is an ordinary point of the differential equation (6.1).

Conclusion: The differential equation (6.1) has two nontrivial linearly independent
power series solutions of the form

∑ c (x − x )
n
n 0 ,
n =0

and these power series converge in some interval x − x0 < R (where

R > 0 ) about x 0

197
x − x0 < R is called the convergence interval. If this interval is finite, it has the midpoint

x 0 . The number R is called the radius of convergence of the power series (6.2). It can

be obtained from either of the formulas


1 1
R= or R= (6.6)
Lim m a m a
m →∞ Lim m +1
m→∞ a
m

provided these limits exist and are not zero. If they are infinite, then the power series
(6.2) converges only at the centre x 0 .

This theorem gives us sufficient condition for the existence of power series solutions of
the differential equation (6.1). Thus if x 0 is an ordinary point of (6.1), we may obtain the
general solution of (6.1) as a linear combination of these two linearly independent power
series

Example 6.3
Find the radius of convergence of the series

(− 1)m x3 x6 x9

m =0 8m
x 3m = 1 − + −
8 64 512
+ −.....

Solution:
By the ratio test, the limiting ratio, r , is given by

am +1
r = lim
m →∞ am

x3( m +1) . ( −1) ( −1)


m +1 m
x 3m
= lim
m →∞ 8m +1 8m

x 3( m +1) 8m
= lim m +1 . 3m
m →∞ 8 x

198
x 3 m . x 3 8m
= lim .
m →∞ 8m.8 x3m

x3
= lim
m →∞ 8

x3
=
8
3
x
=
8

This converges if x < 8 . This implies that x < 2 and thus R = 2 .


3

Operations on Power Series


We shall now consider the operations on power series that are used in connection with
the power series method.

(a) Termwise differentiation


A power series may be differentiated term by term. More precisely: if

y (x ) = ∑ a m (x − x0 )
m

m =0

Converges for x − x0 < R , where R > 0 , then the series obtained by differentiating term

by term also converges for those x and represents the derivative y ′ of y for those x ,
that is,

y ′( x ) = ∑ ma m ( x − x0 )
m −1

m =1

(b) Termwise addition


Two power series may be added term by term. More precisely: if the series
∞ ∞

∑ a m (x − x0 )
m =0
m
and ∑ b (x − x )
m =0
m 0
m
(6.7)

Have positive radii of convergence and their sums are f ( x ) and g ( x ) , then the series

199

∑ (a + bm )( x − x0 )
m
m
m =0

Converges and represents f (x ) + g (x ) for each x which lies in the interior of the
convergence interval of each of the given series.

(c) Termwise multiplication


Two power series may be multiplied term by term. More precisely: Suppose that the
series (6.7) have positive radii of convergence and let f ( x ) and g ( x ) be their sums. Then
the series obtained by multiplying each term of the first series by each term of the second
series and collecting like powers of ( x − x0 ) , that is,

∑ (a b
m =0
0 m + a1bm −1 + ... + a m b0 )( x − x0 )
m

= a0 b0 + (a0 b1 + a1b0 )(x − x0 ) + (a0 b2 + a1b1 + a 2 b0 )( x − x0 ) + ...


2

Converges and represents f ( x )g ( x ) for each x in the interior of the convergence interval
of each of the given series.

B. The Method of Solution


Now that we are assured that under appropriate hypotheses, equation (6.1) actually does
have power series solutions of the form (6.2), we now proceed to determine the
coefficients c0 , c1 , c 2 , ... in the expression

∑ c (x − x )
n
n 0
n =0

so that this expression actually does satisfy equation (6.1).

Assume that x 0 is an ordinary point of the differential equation (6.1), so that solutions in

powers of ( x − x0 ) actually do exist. We denote such solutions by



y = c0 + c1 ( x − x0 ) + c 2 ( x − x 0 ) + ... = ∑ c n (x − x0 ) .
2 n
(6.8)
n=0

200
Since the series in (6.8) converges on an interval x − x0 < R about x 0 , it may be

differentiated term by term on this interval twice in succession to obtain



y ′ = c1 + 2c 2 ( x − x0 ) + 3c3 ( x − x 0 ) + ... = ∑ nc n ( x − x0 )
2 n −1
(6.9)
n =1

and

y ′′ = 2c 2 + 6c3 ( x − x0 ) + 12c 4 ( x − x0 ) + ... = ∑ n(n − 1)c n ( x − x0 )
2 n−2
(6.10)
n=2

respectively.

We then substitute the series in the right members of (6.8), (6.9) and (6.10) in the
differential equation (6.1). We then simplify the resulting expression so that it takes the
form
K 0 + K1 ( x − x0 ) + K 2 (x − x0 ) + ... = 0
2
(6.11)

where the coefficients K i (i = 0,1,2,...) are functions of certain coefficients cn of the

solution (6.8). In order that (6.11) be valid for all x in the interval of convergence
x − x0 < R , we must set

K 0 = K 1 = K 2 = ... = 0 .

This leads to a set of conditions that must be satisfied by the various coefficients cn in

the series (6.8) in order that (6.8) be a solution of the differential equation (6.1). We shall
illustrate this procedure with some examples.

Example 6.4
Find the power series solution of the differential equation
( )
y ′′ + xy ′ + x 2 + 2 y = 0 (6.12)
In powers of x (that is, about x0 = 0 )

Solution

201
Observe that x0 = 0 is an ordinary point of the differential equation (6.12) and so two

linearly independent solutions of the desired type actually exist. We therefore assume a
solution of the form (6.8) with x0 = 0 . Thus

y = ∑ cn x n . (6.13)
n=0

Differentiate (6.13) term by term to obtain



y ′ = ∑ nc n x n −1 (6.14)
n =1

and

y ′′ = ∑ n(n − 1)c n x n − 2 . (6.15)
n=2

Substituting the series (6.13), (6.14) and (6.15) in the differential equation (6.12) we
obtain
∞ ∞ ∞ ∞

∑ n(n − 1)cn x n−2 + x∑ nc n x n−1 + x 2 ∑ cn x n + 2∑ cn x n = 0 ,


n=2 n =1 x =0 n =0

or
∞ ∞ ∞ ∞
. ∑ n(n − 1)c n x n − 2 + ∑ nc n x n + ∑ c n x n + 2 + 2∑ c n x n = 0 (6.16)
n=2 n =1 x =0 n=0

To put the left member of the equation (6.16) in the form (6.11), we shall rewrite the first
and third summations in (6.16) so that x in each of the se summations will have the
exponent n . In the first term we replace n by n + 2 and in the third term we replace n
by n − 2 . Thus the series ((6.16) becomes
∞ ∞ ∞ ∞

∑ (n + 2)(n + 1)cn+ 2 x n + ∑ ncn x n + ∑ cn−2 x n + 2∑ cn x n


n=0 n =1 x=2 n=0
= 0. (6.17)

Although x has the same exponent n in each summation in (6.17), the ranges of the
various summations are not the same. The common range is from 2 to ∞ . We now write
out individually the terms in each summation that do not belong to this common range,
and continue to employ the summation notation to denote the remainder of the terms.

202
Equation (6.17) thus becomes
∞ ∞
2c 2 + 6c3 x + ∑ (n + 2 )(n + 1)c n + 2 x n + c1 x + ∑ nc n x n
n=2 n=2

∞ ∞
+ ∑ c n- 2 x n + 2c0 + 2c1 x + 2∑ c n x n = 0 .
n=2 n=2

We now combine like powers of x to obtain


(2c0 + 2c2 ) + (3c1 + 6c3 )x

+ ∑ [(n + 2 )(n + 1)c n + 2 + (n + 2 )c n + c n − 2 ]x n = 0 (6.18)
n=2

Equation (6.18) is in the desired form (6.11). For (6.18) to be valid for all x in the
interval of convergence x − x0 < R , its coefficient of each power of x must be equal to

zero. This leads to the conditions

2c0 + 2c 2 = 0 (6.19)

3c1 + 6c3 = 0 (6.20)

(n + 2)(n + 1)cn+ 2 + (n + 2)cn + c n−2 = 0 , n≥2 (6.21)

From (6.19) and (6.20), we have


c 2 = −c 0 , (6.22)

and
c3 = − 12 c1 . (6.23)

The condition (6.21) is called a recurrence formula. It enables us to express each


coefficient c n + 2 for n ≥ 2 in terms of the previous coefficients cn and c n − 2 , thus
obtaining
(n + 2)c n + cn −2 n ≥ 2,
cn+ 2 = , (6.24)
(n + 1)(n + 2)

203
For n = 2 , we have
4c 2 + c0
c4 = − .
12
Using (6.22), this reduces to
c 4 = 14 c 0 (6.25)

For n = 3 , (6.24) yields


5c3 + c1
c5 = − .
20

Using (6.23), this reduces to


c5 = 3
40 c1 . (6.26)

In the same way we can express each even term coefficient in terms of c0 and each odd

coefficient in terms of c1 .

Substituting the values of c2 , c3 , c4 and c5 in the assumed solution (6.13) we obtain

y = c 0 + c1 − c 0 x 2 − 12 c1 x 3 + 14 c 0 x 4 + 3
40 c1 x 5 + ...

Collecting terms in c0 and c1 together, finally we have

( ) (
y = c 0 1 − x 2 + 14 x 4 + ... + c1 x − 12 x 3 + 3
40 )
x 5 + ... , (6.27)

which gives the solution of the differential equation (6.12)

The two series in parentheses in (6.27) are power series expansions of two linearly
independent solution of (6.12), and c0 and c1 are arbitrary constants which can be
determined by the some initial or boundary condition as demonstrated by the following
example.
g

Example 6.5

204
Find a power series solution of the initial-value problem
(x 2
)
− 1 y ′′ + 3xy ′ + xy = 0 , (6.28)

y(0) = 4 , (6.29)

y ′(0) = 6 . (6.30)

Solution:
We first observe that all points except x = ±1 are ordinary points for the differential
equation (6.28). Thus we can assume solutions of the form (6.8) for any x0 ≠ ±1 .

However, since the initial values are prescribed at x = 0 , we shall choose x0 = 0 and

seek solutions in power of x . Thus we assume



y = ∑ cn x n . (6.31)
n=0

Differentiate (6.31) term by term to obtain



y ′ = ∑ nc n x n −1 (6.32)
n =1

and

y ′′ = ∑ n(n − 1)c n x n − 2 . (6.33)
n=2

Substituting the series (6.31), (6.32) and (6.33) in the differential equation (6.28) we
obtain
∞ ∞ ∞ ∞

∑ n(n − 1)cn x n − ∑ n(n − 1)cn x n−2 + 3∑ nc n x n + ∑ x n+1 = 0


n=2 n=2 n =1 n =0
(6.34)

We now rewrite the second and fourth summations in (6.34) so that x in each of these
summations has the exponent n so that
∞ ∞ ∞ ∞

∑ n(n − 1)c
n=2
n x n − ∑ (n + 2 )(n + 1)c n + 2 x n + 3∑ nc n x n + ∑ c n −1 x n = 0
n =0 n =1 n =1
(6.35)

205
The common range is from 2 to ∞ . We now write out individually the terms in each
summation that do not belong to this common range, and continue to employ the
summation notation to denote the remainder of the terms. Thus
∞ ∞

∑ n(n − 1)cn x n − 2c2 − 6c3 x − ∑ (n + 2)(n + 1)cn+ 2 x n


n=2 n=2

∞ ∞
+ 3c1 x + 3∑ nc n x n + c0 x + ∑ c n −1 x n = 0 (6.36)
n=2 n=2

Combining like powers of x we get


− 2c 2 + (c 0 + 3c1 − 6c3 )x

+ ∑ [− (n + 2 )(n + 1)c n − 2 + n(n + 2 )c n + c n −1 ]x n = 0 (6.37)
n=2

For (6.37) to be valid for all x in the interval of convergence x − x0 < R , its coefficient

of each power of x must be equal to zero. This leads to the conditions


− 2c 2 = 0 , (6.38)
c0 + 3c1 − 6c3 = 0 , (6.39)

− (n + 2 )(n + 1)c n + 2 + n(n + 2)c n + c n −1 = 0 , n≥2 (6.40)

Form (6.38), we have c2 = 0 ; and from (6.39), c3 = 16 c 0 + 12 c1 . The recurrence formula


(6.40) gives
n(n + 2 )c n + c n −1
cn+2 = , n ≥ 2.
(n + 1)(n + 2)

Using this, we find successively


8c 2 + c1 1
c4 = = c1 ,
12 12
15c3 + c 2 1 3
c5 = = c0 + c1 .
20 8 8

206
Substituting the values of c2 , c3 , c4 c5 … in the assumed solution (6.31) we obtain

⎛c c ⎞ c ⎛c 3c ⎞
y = c 0 + c1 x + ⎜ 0 + 1 ⎟ x 2 + 1 x 4 + ⎜ 0 + 1 ⎟ x 5 + ...
⎝6 2⎠ 12 ⎝8 8 ⎠
or
( ) (
y = c 0 1 + 16 x 3 + 18 x 5 + ... + c1 x + 12 x 3 + 121 x 4 + 83 x 5 + ... . ) (6.41)

The solution (6.41) is the general solution of the differential equation (6.28) in powers of
x.

We now apply the initial conditions (6.29) and (6.30). Applying (6.29) to (6.41) we find
that
c0 = 4 .

Differentiating (6.41, we find that


y ′ = c0 ( 1
2 ) ( )
x 2 + 85 x 4 + ... + c1 1 + 32 x 2 + 13 x 3 + 158 x 4 + ... . (6.42)

Applying (6.30) to (6.42) we find that


c1 = 6
so that the solution of the initial-value problem in powers of x is now given by
( ) (
y = 4 1 + 16 x 3 + 18 x 5 + ... + 6 x + 12 x 3 + 121 x 4 + 83 x 5 + ... )
or
y = 4 + 6 x + 113 x 3 + 12 x 4 + 114 x 5 + ... .

Remarks:
Suppose the initial values of y and y ′ in conditions (6.29) and (6.30) of example 6.5 are
prescribed at x = 2 , instead of x = 0 . Then we have the initial-value problem

207
(x )d
2
y dy
2
−1 2
+ 3x + xy = 0
dx dx
(6.43)
y (2) = 4, y ′(2) = 6

We then seek solutions of the for



y = ∑ cn (x − 2) .
n
(6.44)
n=0

The simplest procedure for obtaining a solution of the form (6.44) is first to make the
substitution t = x − 2 . This replaces the initial-value problem (6.43) by equivalent
problem

(t )d
2
+ (3t + 6 ) + (t + 2 ) y = 0
y dy
2
+ 4t + 3 2
dt dt

y (0) = 4, y ′(0) = 6
In which t is the independent variable and the initial values are prescribed at t = 0 . One
then seeks a solution of the for

y = ∑ cn t n . (6.45)
n =0

Replacing t by x − 2 in resulting solution, one obtains the desired solution of the original
problem.

Exercise 6.1
Find power series solutions in powers of x for each of the following differential
equations:
1. y ′′ + xy ′ + y = 0 . 2. y ′′ − y ′ + 2 xy = 0 .

3. y ′′ + xy ′ + (2 x 2 + 1)y = 0 . 4. y ′′ + xy ′ + (3x + 2) y = 0 .

5. y ′′ − (x 3 + 2)y ′ − 6 x 2 y = 0 . 6. (x 2
+ 1)y ′′ + xy ′ + xy = 0 .

7. (x 3
− 1)y ′′ + x 2 y ′ + xy = 0 .

208
Find the power series solution of each of the following initial-value problems:
8. y ′′ − xy ′ − y = 0 , y(0) = 1 , y ′(0) = 0 .

9. y ′′ + x 2 y ′ + x 2 y = 0 , y(0) = 2 , y(0) = 4 .

10. (2 x 2
− 3)y ′′ − 2 xy ′ + y = 0 y(0) = −1 , y(0) = 5

Find the power series solution in powers of (x − 1) for each of the following differential
equations:
11. x 2 y ′′ + xy ′ + y = 0 . 12. x 2 y ′′ + 3xy ′ − y = 0

13. Find the power series solution in powers of (x − 1) for the following initial-value
problem:
xy ′′ + y ′ + 2 y = 0 , y(1) = 2 , y ′(1) = 4

14. The differential equation


(1 − x )y ′′ − 2 xy ′ + n(n + 1)y = 0 ,
2

where n is a constant, is called Legendre’s differential equation.


(a) Show that x = 0 is an ordinary point of this differential equation, and find two
linearly independent power series solutions in powers of x .
(b) Show that if n is a nonnegative integer, then one of the two solutions found in
part (a) is a polynomial of degree n .

6.2 Solutions About Singular Points: The Method of Frobenius

A. Regular Singular Points


Consider again the homogeneous linear differential equation
a 0 ( x ) y ′′ + a1 ( x ) y ′ + a 2 (x ) y = 0 , (6.46)

and assume that x 0 is a singular point of (6.46). Then Theorem 6.1 does not apply at the

point x 0 , and therefore we are not assured of a power series solution

209

y = ∑ c n (x − x0 )
n
(6.47)
n=0

in powers of ( x − x0 ) . We must then seek a different type of solution for such a case. It
happens that under certain conditions we are justified in assuming a solution of the form
r ∞
y = x − x0 ∑ c (x − x )
n =0
n 0
n
, (6.48)

where r is a certain (real or complex) constant. In order to be able to state conditions


under which a solution of this form is assured, we first classify singular points.

We again write the differential equation (6.46) in the equivalent normalized form
y ′′ + P1 ( x ) y ′ + P2 ( x ) y = 0 , (6.49)
where
a1 ( x ) a2 (x )
P1 ( x ) = and P2 ( x ) =
a0 (x ) a0 (x )

Definition 6.3
Consider the differential equation (6.46), and assume that at least one of the functions P1

and P2 in the equivalent normalized equation (6.49) is not analytic at x 0 , so that x 0 is a


singular point of (6.46). If the functions defined by the products
(x − x0 )P1 (x ) and (x − x0 )2 P2 (x ) (6.50)

Are both analytic at x 0 , then x 0 is called a regular singular point of the differential
equation (6.46). If either (or both) of the functions defined by the products (6.50) is not
analytic at x 0 , then x 0 is called an irregular singular point of (6.46).

Example 6.6
Consider the differential equation
2 x 2 y ′′ − xy ′ + ( x − 5) y = 0 . (6.51)
Writing this in the normalized form (6.49), we have
1 x−5
y ′′ − y′ + y =0.
2x 2x 2

210
x−5
Here P1 ( x ) = − and P2 ( x ) =
1
. It is apparent that both P1 and P2 fail to be
2x 2x 2
analytic at x = 0 and therefore we conclude that x = 0 is a singular point of (6.51). We
now consider the functions defined by the products
x−5
xP1 ( x ) = − x 2 P2 (x ) =
1
and
2 2
of the form (6.50). Both of these product functions are analytic at x = 0 , and so x = 0 is
a regular point of the differential equation (6.51).

Example 6.7
Consider the differential equation
x 2 ( x − 2 ) y ′′ + 2( x − 2 ) y ′ + ( x + 1) y = 0 .
2
(6.52)
In the normalized form (6.49), we have
2 x +1
y ′′ + y′ + 2 y = 0.
x (x − 2)
2
x (x − 2)
2

Here
x +1
P1 ( x ) = P2 ( x ) =
2
and .
x (x − 2)
2
x (x − 2)
2 2

Clearly the singular points of the differential equation (6.52) are x = 0 and x = 2 .

Consider x = 0 and form the functions defined by the products


x +1
xP1 ( x ) =
2
and x 2 P2 (x ) = .
x( x − 2 ) ( x − 2 )2
The product function x 2 P2 ( x ) is analytic at x = 0 , but that defined by xP1 ( x ) is not. Thus
x = 0 is an irregular singular point of (6.52).

Now consider x = 2 . From (6.50) we have


x +1
(x − 2)P1 (x ) = 2
and (x − 2)2 P2 (x ) = .
x2 x2
Both of the functions thus defined are analytic at x = 2 , and hence x = 2 is a regular
singular point of (6.52).

211
Now that we can distinguish between regular and irregular singular points, we shall state
a basic theorem concerning solutions of the form (6.48) about regular singular points..

Theorem 6.2
Hypothesis: the point x 0 is a regular singular point of the differential equation (6.46).

Conclusion: The differential equation (6.46) has at least one nontrivial solution of the
form

x − x0 ∑ c (x − x )
r n
n 0 . (6.53)
n=0

where r is a definite (real or complex) constant which may be determined,


and this solution is valid in some deleted interval 0 < x − x0 < R (where

R > 0 ) about x 0

Example 6.8
In example (6.6) we saw that x = 0 is a regular singular point of the differential equation
2 x 2 y ′′ − xy ′ + ( x − 5) y = 0 . (6.54)

By Theorem (6.2) we conclude that this equation has at least one nontrivial solution of
the form

∑c
r
x n xn ,
n =0

valid in some deleted interval 0 < x < R about x = 0 .

Example 6.9
In example (6.7) we saw that x = 2 is a regular singular point of the differential equation
x 2 ( x − 2 ) y ′′ + 2( x − 2 ) y ′ + ( x + 1) y = 0 .
2
(6.55)

212
Thus we know that this equation has at least one nontrivial solution of the form

x−2 ∑ c (x − 2)
r n
n ,
n =0

in some deleted interval 0 < x − 2 < R about x = 2 .

We also observed that x = 0 is a singular point of equation (6.55). However, this singular
point is irregular and so Theorem 6.2 does not apply to it. We are therefore not assured
the differential equation (6.55) has a solution of the form

∑c
r
x n xn
n =0

in any deleted interval about x = 0

B. The Method of Frobenius


Now we are assured of at least one solution of the form (6.53) about a regular singular
point x 0 of the differential equation (6.46), we now proceed to determine the coefficients

cn and the number r in this solution. The procedure is similar to that introduced in
section 6.1 and is commonly called the method of Frobenius.

Outline of the Method of Frobenius


1. Let x 0 be a regular singular point of the differential equation (6.46), assume a
solution of the form

y = (x − x0 ) ∑ c (x − x )
r n
n 0
n =0

valid in some interval 0 < x − x0 < R where c0 ≠ 0 . We write this solution in the
form

y = ( x − x0 )∑ c n ( x − x 0 )
n+r
, (6.56)
n =0

2. Assuming term-by-term differentiation of (6.56) is valid, we obtain

213

y ′ = ∑ (n + r )c n ( x − x0 )
n + r −1
(6.57)
n=0

and

y ′′ = ∑ (n + r )(n + r − 1)c n ( x − x0 )
n+ r −2
. (6.58)
n =0

Substitute (6.56), (6.57) and (6.58) into the differential equation (6.46).

3. Simplify the resulting expression so that it takes the form


K 0 (x − x0 ) + K1 ( x − x0 ) + K 2 ( x − x0 )
r +k r + k +1 r +k +2
+ ... = 0 , (6.59)

where k is a certain integer and the coefficients K i (i = 0,1,2,...) are functions of

r and certain of the coefficients cn of the solution (6.56).

4. For (6.59) to be valid for all x in the deleted interval 0 < x − x0 < R , we must set

K 0 = K 1 = K 2 = ... = 0 .

5. Upon equating to zero the coefficient K 0 of the lowest power r + k of ( x − x0 ) ,


we obtain a quadratic equation in r , called the indicial equation of the
differential equation (6.46). We denote the roots of the indicial equation by r1 and
r2 , where Re(r1 ) ≥ Re(r2 )

6. Now equate the remaining coefficients K1 , K 2 ... in (6.59) to zero. We are now led
to conditions involving the constant r , which must be satisfied by various
coefficients cn in (6.56)

7. We now substitute the root r1 for r into the conditions obtained in step 6, and
then choose the cn to satisfy these conditions. The resulting series (6.56) with

r = r1 is the desired solution.

There are three possible forms for the second linearly independent solution corresponding
to the following cases:

214
Case 1: r1 and r2 differ but not by an integer.
A basis is of the form
( )
y1 ( x ) = x r1 c 0 + c1 x + c 2 x 2 + ... ,

and
(
y 2 (x ) = x r2 c 0* + c1* x + c 2* x 2 + ...)

Case 2: r1 = r2 .
A basis is
(
y1 ( x ) = x r1 c 0 + c1 x + c 2 x 2 + ... )
and
( )
y 2 ( x ) = y1 ( x ) ln x + x r2 c 0* + c1* x + c 2* x 2 + ... ,

Case 3: r1 and r2 differ by a nonzero integer.


A basis is
(
y1 ( x ) = x r1 c 0 + c1 x + c 2 x 2 + ... )
and
( )
y 2 ( x ) = ky1 ( x ) ln x + x r2 c 0* + c1* x + c 2* x 2 + ...

where the roots are so denoted that r1 − r2 > 0 and may turn out to be zero.

Note that in Case 2 we must have logarithm whereas in Case 3 we may or may not.

We now consider the three cases separately.

Case 1: r1 and r2 differ but not by an integer.


This is the easiest case, since equation (6.49) has two solutions, for x > 0 , of the forms
( )
y1 ( x ) = x r1 c 0 + c1 x + c 2 x 2 + ... , c0 = 1 ,

215
(
y 2 (x ) = x r2 c 0* + c1* x + c 2* x 2 + ... . ) c 0* = 1 .

Thus y1 and y 2 are linearly independent since y1 cannot be constant. The coefficients
y2

c1 , c2 , ... are obtained by following the above steps. We demonstrate the procedure with
the following two examples.

Example 6.10
Solve the Euler equation
1 1
y ′′ + y′ + 2 y = 0 , x > 0. (6.60)
4x 8x

Solution
We substitute

( )
∞ ∞
y = x r c0 + c1 x + c 2 x 2 + ... = x r ∑ c n x n = ∑ c n x r + n
n =0 n =0

into equation (6.60) to obtain



1 ∞ ∞

∑ (r + n )(r + n − 1)cn x r + n−2 +


n =0

4 x n =0
(r + n )cn x r + n −1 + 1 2
8x
∑c
n =0
n x r +n

∞ ∞ ∞
= ∑ (r + n )(r + n − 1)c n x r + n − 2 + ∑ 14 (r + n )c n x r + n − 2 + ∑ 18 c n x r + n − 2
n =0 n =0 n =0


= ∑ c n [(r + n )(r + n − 1) + 1
4
(r + n ) + 18 ]x r + n−2 =0. (6.61)
n =0

The indicial equation is obtained by setting the expression in brackets, for n = 0 , equal to
zero. Thus
r (r − 1) + 14 r + 18 = r 2 − 34 r + 18 = (r − 12 )(r − 14 ) = 0 ,

with roots r = 1
4 and r = 1
2 that do not differ by an integer. Assume a solution of the form

(
y = x 4 c0 + c1 x + c2 x 2 + ... .
1
)
Then with r = 14 , equation (6.61) becomes

216

= ∑ c n [( 14 + n )( 14 + n − 1) + ( 14 + n ) + 18 ]x 4+n−2
= 0,
1
1
4
n =0

or, after a bit of algebra,



⎡ 2 n ⎤ n−7 4
∑c
n =0
n ⎢⎣n − 4 ⎥⎦ x .

Equating all terms of this series to zero, we get


n(n − 14 )c n = 0 ,

which holds only if c n = 0 , for n > 0 . Thus

y1 ( x ) = c0 x 4 = x 4 .
1 1

To find the second solution, we set r = 12 in equation (6.61), obtaining



= ∑ c n [( 12 + n )( 12 + n − 1) + ( 12 + n ) + 18 ]x 2+n−2
= 0.
1
1
2
n =0

from which we get n(n + 14 )c n* = 0 . Thus c n* = 0 for n > 0 . So

y 2 ( x ) = c0* x .
Hence the general solution to equation (6.60) is
y = Ax 4 + Bx 2 .
1 1

The roots of the indicial equation may also be complex, as the following example
illustrates.

Example 6.11
Find the general solution of the equation
1 1
y ′′ + y′ + 2 y = 0 , for x>0 (6.62)
x x

Solution

Substitute y = ∑ c n x r + n and its derivatives into (6.62) into to obtain
n =0

217

∑ c [(r + n )(r + n − 1) + (r + n ) + 1]x


n =0
n
r + n−2
= 0. (6.63)

The indicial equation obtained by setting n = 0 in (6.63) is


r (r − 1) + r + 1 = r 2 + 1 = 0 ,

with the roots r1 = i and r2 = −i , which do not differ by an integer. Setting r = i in


(6.63), we have

∑ c [(i + n )(i + n − 1) + (i + n ) + 1]x


n =0
n
i +n−2
= 0.

Equating all coefficients of this series to zero, we have, after combining terms, and using
the fact that (i + n ) = n 2 + 2in − 1 ,
2

[ ] ( )
0 = cn (i + n ) + 1 = c n n 2 + 2in = cn n(n + 2i ) ,
2

which holds only if c n = 0 for n > 0 . Note that

e ln u = u
and that
e iu = cos u + i sin u ,
for any u > 0 . This suggests that

x i = e ln x = e i ln x = cos ln x + i sin ln x .
i

Thus, setting c0 = 1 (any constant gives us a solution),

y1 ( x ) = x i = e i (ln x ) = [cos(ln x ) + i sin (ln x )] .

Similarly, substituting r = −i into (6.63) yields the series


∑ c [(− i + n )(− i + n − 1) + (− i + n ) + 1]x


n =0
*
n
−i + n − 2
=0

whose coefficients satisfy the condition


[
c n* n 2 − 2in = 0 . ]
Thus c n* = 0 for n > 0 . Hence, setting c n* = 1 ,

y1 ( x ) = x − i = [cos(ln x ) − i sin (ln x )]

218
Finally since linear combinations of solutions are solutions, the real and imaginary parts
of y1 and y 2 ,

y1* ( x ) = ( y1 + y 2 ) = cos(ln x ) ,
1
2

y 2 (x ) = ( y1 − y 2 ) = sin (ln x )
1
2i
are solutions of (6.62). y1* and y 2* are linearly independent since ( y 2* y1* ) = tan (ln x ) ,
which is nonconstant. Hence the general solution of (6.62) is
y = A cos(ln x ) + B sin(ln x ) , x>0
g

Case 2: r1 = r2
Example 6.12
Solve the differential equation
x(x − 1) y ′′ + (3x − 1) y ′ + y = 0 . (6.64)

Solution

Substitute y = ∑ c n x r + n and its derivatives into (6.64) into to obtain
n =0

∞ ∞

∑ (n + r )(n + r − 1)c
n =0
n x n + r − ∑ (n + r )(n + r − 1)c n x n + r −1
n =0

∞ ∞ ∞
+ 3∑ (n + r )c n x n + r − ∑ (n + r )c m x n + r −1 + ∑ c n x n + r = 0 . (6.65)
n =0 n =0 n =0

By equating the sum of the smallest power of x , which is x r −1 , we obtain


[− r (r − 1) _ r ]c0 = 0 or r2 = 0 .

Hence this indicial equation has the double root r = 0 .

First Solution

219
We insert this value into (6.65) and equate the sum of the coefficients of the power x s to
zero., getting
s(s − 1)a s − (s + 1)sa s +1 + 3sa s − (s + 1)a s +1 + a s = 0
0r
a s +1 = a s .

Hence a 0 = a1 = a 2 = ... , and by choosing a 0 = 1 we obtain the solution



y1 ( x ) = ∑ x n =
1
.
n =0 1− x

Second solution
We now substitute y 2 = uy1 and its derivatives into (6.64), to get
x( x − 1)(u ′′y1 + 2u ′y1′ + uy1′′) + (3x − 1)(u ′y1 + uy1′ + uy1′ ) + uy1 = 0 .
Since y1 is a solution of (6.64), this reduces to
x( x − 1)(u ′′y1 + 2u ′y1′ ) + (3x − 1)u ′y1 = 0 .
Inserting the expressions for y1 and y 2 , we first have

⎛ 1 ⎞
x( x − 1)⎜⎜ u ′′ ⎟ + (3x − 1)u ′
1 1
+ 2u ′′ 2 ⎟
= 0.
⎝ 1− x (1 − x ) ⎠ 1− x

or
xu ′′ + u ′ = 0 .
By integrating twice, we get
1 1
ln u ′ = − ln x = ln , u′ = , u = ln x .
x x
Hence a second independent solution of (6.64) is
ln x
y 2 = uy1 =
1− x
g

Case 3: Second solution with logarithmic term

220
Example 6.13
Solve
(x 2
− 1)x 2 y ′′ − (x 2 + 1)xy ′ + (x 2 + 1)y = 0 .

Solution

Substitute y = ∑ c n x r + n and its derivatives into the differential equation to get
n =0

(x ) ( )
∞ ∞
2
− 1 ∑ (n + r )(n + r − 1)c n x n + r − x 2 + 1 ∑ c n x n + r = 0 .
n =0 n =0

or
∞ ∞

∑ (n + r − 1) cn x n+ r + 2 − ∑ (n + r + 1)(n + r − 1)cn x n+ r = 0.
2
(6.66)
n =0 n =0

By equating the coefficients of x r to zero we get the indicial equation


(r + 1)(r − 1) = 0 .
The roots r1 = 1 and r2 = −1 differ by an integer.

First solution
By equating the coefficients of x r +1 in (6.66) to zero we have
− (r + 2)ra1 = 0 .
For r1 = 1 and r2 = −1 , it follows that a1 = 0 . By equating the sum of the coefficients of

x s + r + 2 in (6.66) to zero we find that

(s + r − 1)2 a s = (s + r + 3)(s + r + 1)a s+ 2 , (s = 0,1,...) . (6.67)

Inserting r = r1 = 1 and solving for a s + 2 , we see that

s2
as+2 = a , (s = 0,1,...) .
(s + 4)(s + 2) s
Since a1 = 0 , it follows that a3 = 0 , a5 = 0 , etc. For s = 0 we obtain a 2 = 0 and from

this by taking s = 2,4,... , we have a 4 0 , a6 = 0 , etc. Hence to the larger root r1 = 1 there
corresponds the solution
y1 = a 0 x .

221
Second solution
If we insert the smaller root r = r2 = −1 into (6.67), we get

(s − 2)2 a s = (s + 2)sa s + 2 , (s = 0,1,...) .


With s = 0 this becomes 4a0 = 0 , which implies that a0 = 0 . This contradicts the initial

assumption that a0 ≠ 0 . Therefore the equation does not have a second independent

solution y 2 ( x ) of the form y = ∑ c n x r + n corresponding to the smaller root r2
n =0

To obtain a second solution we from


( )
y 2 ( x ) = ky1 ( x ) ln x + x r2 c 0* + c1* x + c 2* x 2 + ... (6.68)

with r2 = −1 . It is simpler to use the method of reduction of order directly; that is, since
x is a solution then we may set
y 2 ( x ) = xu( x ) .
Substituting this into the present differential equation and simplifying, we find that
(x 3
− x )u ′′ + (x 2 − 3)u ′ = 0
or
u ′′ 3 − x 2 3 1 1
= 3 =− + + .
u x −x x x +1 x −1
Integrating both sides we have
x2 −1
ln u ′ = −3 ln x + ln( x + 1) + ln(x − 1) = ln .
x3
Taking exponentials and integrating again, we obtain
1
u = ln x + .
2x 2
Therefore the second independent solution is

y 2 ( x ) = xu ( x ) = x ln x +
1
.
2x
g

222
Exercise 6.2
Locate and classify the singular points of each of the differential equations:
1. (x 2
− 3x )y ′′ + ( x + 2) y ′ + y = 0 . 2. (x 3
+ x 2 )y ′′ + (x 2 − 2 x )y ′ + 4 y = 0

3. (x 4
− 2 x 3 + x 2 )y ′′ + 2( x − 1) y ′ + x 2 y = 0

4. (x 5
+ x 4 − 6 x 3 )y ′′ + x 2 y ′ + ( x − 2) y = 0
Use the method of Frobenius to find solutions of each of the following differential
equations
5. xy ′′ + 2 y ′ + 4 xy = 0 . 6. xy ′′ + (1 − 2 x ) y ′ + ( x − 1) y = 0

7. (x − 1)xy ′′ + (4 x − 2) y ′ + 2 y = 0 8. x 2 y ′′ − 5 xy ′ + 9 y = 0

9. x(1 − x ) y ′′ + 12 ( x + 1) y ′ − 12 y = 0 10. (x + 2)2 y ′′ + (x + 2) y ′ − y = 0


11. (x − 1)2 y ′′ + (x − 1) y ′ − 4 y = 0
12. x(1 − x ) y ′′ + ( 32 − 72 x )y ′ − 32 y = 0

13. x(1 + x ) y ′′ − (1 + 2 x )xy ′ + (1 + 2 x ) y = 0

14. 2 x(1 − 2 x ) y ′′ + (12 x 2 − 4 x + 1)y ′ − 2(4 x 2 + 1)y = 0

6.3 Bessel Functions


The differential equation
x 2 y ′′ + xy ′ + (x 2 − p 2 )y = 0 (6.69)

which known as the Bessel equation of order p (≥ 0) , is one of the most important
differential equations in applied mathematics. The equation was first investigated in 1703
by Jacob Bernoulli in connection with the oscillatory behaviour of a hanging chain, and
later by the German mathematician Friedrich Wilhelm Bessel (1784-1846) in his studies
of planetary motion. Since then, the Bessel functions have been used in the studies of
elasticity, fluid motion, potential theory, diffusion, and the propagation of waves.

From the previous section, a solution of the form



y(x ) = ∑ cn x r + n , x ≠ 0, c0 = 1 , (6.70)
n =0

223
exists for the Bessel equation of order p . Upon dividing equation (6.69), we get

1 ⎛ p2 ⎞
y ′′ + y ′ + ⎜⎜1 − 2 ⎟⎟ y = 0 .
x ⎝ x ⎠

Substitution of equation (6.70) yields (after setting k = n + 2 in ∑c
n =0
n x r +n )

∞ ∞

∑ c (r + n )(r + n − 1)x
n =0
n
r + n−2
+ ∑ c n (r + n )x r + n − 2
n =0

∞ ∞
+ ∑ − p 2 cn x r +n−2 + ∑ cn−2 x r +n−2 = 0
n =0 n=2

or
( ) [
c0 r 2 − p 2 x r − 2 + c1 (r + 1) − p 2 x r −1
2
]
{ [ ] }

+ ∑ c n (n + r ) − p 2 + c n − 2 x r + n − 2 = 0
2
(6.71)
n=2

The indicial equation is r 2 − p 2 = 0 , with roots r1 = p (≥ 0) and r2 = − p . Setting r = p


in (6.71) yields

(1 + 2 p )c1 x p −1 + ∑ [n(n + 2 p )c n + c n−2 ]x n + p −2 =0
n=2

Indicating that c1 = 0 and that


cn−2
cn = − , for n ≥ 2 . (6.72)
n(n + 2 p )
Hence all the coefficients with odd-numbered subscripts c 2 j +1 are zero, since by equation

(6.72) they can all be expressed as a multiple of c1 . Letting n = 2 j + 2 , we see that the
coefficients with even-numbered subscripts satisfy the equation
c2 j
c 2( j +1) = − , for j ≥ 0 ,
2 2 ( j + 1)( p + j + 1)
which yields
c0
c2 = − ,
2 ( p + 1)
2

c2 c0
c4 = − = 2 ,
2 .2( p + 2) 2 .2!( p + 1)( p + 2 )
2

224
c4 c0
c6 = − =− 4 ,…
2 .3( p + 3)
2
2 .3!( p + 1)( p + 2)
Hence series (6.70) becomes

p⎡ c c0 ⎤
y1 ( x ) = x ⎢c0 − 2 0 x2 + 4 x 4 − ...⎥
⎣ 2 ( p + 1) 2 .2!( p + 1)( p + 2) ⎦

x 2n
= c0 x ∑ (− 1) 2 n
p n
. (6.73)
n =0 2 .n!( p + 1)( p + 2)...( p + n )

To write equation (6.73) in a more compact form, we define the gamma function for all
values p > −1 :

Γ( p + 1) = ∫ e −t t p dt .
0

Integrating Γ( p + 1) by parts, we get


∞ ∞
Γ( p + 1) = ∫ e −t t p dt = − e −t t p [ ]

0 + p ∫ e −t t p −1 dt .
0 0

= 0 + pΓ( p )
= pΓ( p )
Since

Γ(1) = ∫ e −t dt = − e −t [ ] ∞
0 =1,
0

it follows that Γ(2) = 1.Γ(1) = 1! , Γ(3) = 2Γ(2) = 2! , …, and in general, Γ(n + 1) = n!.

It is customary in equation (6.73) to let c 0 = 2 p Γ( p + 1) [ ]


−1
Then equation (6.73) becomes

J p (x ) =
x
p ∞

(− 1) ( x 2)
2n

∑ x ≠ 0.
n
, (6.74)
2 n=0 n!Γ( p + n + 1)
which is known as the Bessel function of the first kind of order p . Thus J p (x ) is the

first solution of equation (6.69). It can be shown that the series J p ( x ) converges for all

real x .

225
To find the second solution, we must consider the difference r1 − r2 = 2 p . By case 1 of
1
the previous section, if p is not a multiple of 2 , we can again apply the method of

Frobenius with r = − p to find the second solution.. We set r = − p in equation (6.71)


and assume that p is not an integer. We then obtain

(1 − 2 p )c1 x − p −1 + ∑ [n(n − 2 p )cn + c n −2 ]x n − p −2 = 0, (6.75)
n=2

Indicating that c1 = 0 if p ≠ 1
2 and that

cn−2
cn = − . (6.76)
n(n − 2 p )
1
Note that when p is not a multiple of 2 , all the coefficients cn with odd-numbered

subscripts are zero. If p = (2m + 1) 2 , then c 2 m +1 is arbitrary and the recurrence (6.76)
yields the coefficients
c 2 m +1 c 2 m +1
c2 m+3 = − , c 2 m +5 = − , ….
2 ( p + 1)
2
2 .2!( p + 1)( p + 2)
4

Hence the odd-numbered coefficients generate the series

−p ⎡ 2 m +1 c 2 m +1 2 m+3 c 2 m +1 x 2 m +5 ⎤
x ⎢c 2 m +1 x − 2 x + 4 − ...⎥
⎣ 2 ( p + 1) 2 .2!( p + 1)( p + 2 ) ⎦

p⎡ x2 x4 ⎤
= c 2 m +1 x ⎢1 − 2 + 4 − ...⎥ ,
⎣ 2 ( p + 1) 2 .2!( p + 1)( p + 2) ⎦
Which is a multiple of J p (x ) . Thus we can ignore the odd-numbered coefficients and

concentrate on using equation (6.75) to calculate the coefficients with even-numbered


subscripts. Then
c0 c2 c0
c2 = − , c4 = − = 4 ,…,
2 (1 − p )
2
2 .2(2 − p ) 2 .2!(1 − p )(2 − p )
2

And the second solution is the convergence series

J − p (x ) =
−p ∞
( x 2 )2 n .
∑ (− 1)
x n
(6.77)
2 n =0 n!Γ(n − p + 1)
Therefore, if p is not an integer, then
y ( x ) = AJ p ( x ) + BJ − p ( x )

226
Is the general solution of the Bessel equation for all valued x ≠ 0 .

If p is an integer, then the term (n − 2 p ) in the recurrence (6.77) is zero for the integer
n = 2 p . Hence c 2 p − 2 is zero, and iterating equation (6.76) repeatedly, we see that

c 2 p − 2 = c 2 p − 4 = ... = c 2 = c 0 = 0 . But this contradicts the assumed form (6.70). The second

linearly independent solution of equation (6.69) must be calculated using the method of
case 3 with the logarithmic term. After a long but straightforward calculation, we obtain

1 ⎡ p −1 ( p − k − 1)! ⎛ x ⎞
2k − p
hp ⎛ x ⎞
y 2 ( x ) = J p ( x ) ln x − ⎢∑ ⎜ ⎟ + ⎜ ⎟
2 ⎢⎣ k =0 k! ⎝2⎠ p! ⎝ 2 ⎠

∞(− 1)k [hk + h p + k ]⎛ x ⎞ 2k + p ⎤


+∑ ⎜ ⎟ ⎥ (6.78)
k =1 k!( p + k )! ⎝ 2 ⎠ ⎥⎦
where
1 1 1
hp = 1 + + + ... + (6.79)
2 3 p
and p is a positive integer.

It is customary to replace equation (6.78) by the linear combination of solutions

Y p (x ) =
2
π
[y (x ) + (γ − ln 2)J (x )] ,
2 p p = 0,1,2,... ,

where
γ = lim (h p − ln p ) = 0.5772156649...
p →∞

is the Euler constant. This particular solution is obviously independent of J p ( x ) and is

called the Bessel function of the second kind of order p , or Neumann’s function of
order p . It is defined by the formula

⎞ 1 ⎡ ( p − k − 1)!⎛ x ⎞
p −1 2k − p
⎛ x hp ⎛ x ⎞ p
Y p ( x ) = J p ( x )⎜ ln + γ ⎟ − ⎢∑
2
⎜ ⎟ + ⎜ ⎟
π ⎝ 2 ⎠ π ⎣ k =0 k! ⎝2⎠ p! ⎝ 2 ⎠


+ ∑ (− 1)
[h
k ]
+ h p+k ⎛ x ⎞ 2k + p ⎤
⎜ ⎟
k

k =1 k!( p + k )! ⎝ 2 ⎠ ⎥⎦

227
for all integers p = 0,1,2,... .

Therefore the general solution of the Bessel equation of order p is


y ( x ) = AJ p ( x ) + BY p ( x ) , x ≠ 0.

Graphs of the functions J 0 , J 1 , J 2 are shown below:

Note that J 0 has the highest y - intercept, followed by J1 and then J 2 .

Graphs of the functions Y0 , Y1 , Y2 are shown below:

Y0 has the least x - intercept, followed by Y1 and then Y2 .

Properties of Bessel Functions


Now that we have the expressions for J p ( x ) and Y p (x ) , we can drive a number of

important expressions involving Bessel functions and their derivatives. We assume


x > 0 . From equation (6.74), we have

228
n 2 (x 2) (x 2) 2(n + p )
2 n+ 2 p p −1 2 n + 2 p −1
d ∞ p ∞

∑ (− 1) = ∑ (− 1) n 2

dx n =0 n!Γ( p + n + 1) n =0 n!Γ( p + n + 1)

(− 1) ( x 2)
2 n + p −1
= x p J p −1 ( x ) ,
=x ∑
p n

n =0 n!Γ( p + n )
since Γ( p + n + 1) = ( p + n )Γ( p + n ) .

Thus
d p
dx
[ ]
x J p ( x ) = x p J p −1 ( x ) (6.80)

d −p
dx
[ ]
x J p ( x ) = − x − p J p +1 ( x ) (6.81)

Expanding the left-hand sides of equations (6.80) and (6.81), we have


x p J ′p + px p −1 J p = x p J p −1

and
x − p J ′p − px − p −1 J p = − x − p J p +1

which may be simplified to yield the identities


xJ ′p = xJ p −1 − pJ p , (6.82)

xJ ′p = pJ p − xJ p +1 . (6.83)

Subtracting equation (6.83) from equation (6.82), we obtain the recursion relations
xJ p +1 − 2 pJ p + xJ p −1 = 0 . (6.84)

Adding the two together yields


2 J ′p = J p −1 − J p −1 . (6.85)

Equations (6.80) through (6.85) are important in solving problems involving Bessel
functions since they allow us to express Bessel functions of higher order in terms of
lower-order functions.

Example 6.14
Express J 3 (x ) in terms of J 0 ( x ) and J 1 ( x ) .

229
Solution:
Letting p = 2 in equation (6.84) we get
xJ 3 = 4 J 2 − xJ 1 .

If we apply (6.84) with p = 1 to J 2 , we obtain


xJ 2 = 2 J 1 − xJ 0 .
Thus
⎛ 8 ⎞
J 3 (x ) = J 2 − J 1 = 2 (2 J 1 − xJ 0 ) − J 1 = ⎜ 2 − 1⎟ J 1 (x ) − J 0 ( x )
4 4 4
x x ⎝x ⎠ x
g

Example 6.15
Evaluate the integral

∫ x J (x )dx .
4
1 (6.86)

Solution:
Integrating (6.86) by parts, we have, by equation (6.80),

∫ x [x J (x )]dx = x (x )
J 2 − ∫ x 2 J 2 .2 xdx = x 4 J 2 ( x ) − 2 ∫ x 3 J 2 dx .
2 2 2 2
1

Applying equation (6.80) to the last integral, we obtain

∫ x J (x )dx = x J 2 (x ) − 2 x 3 J 3 (x ) + C .
4 4
1

Example 6.16
Solve the equation
y ′′ + k 2 xy = 0 (6.87)

Solution
The following substitution reduces equation (6.87) to a Bessel equation: Let u = y x
3
and z = 2kx 2
3 . Then

230
du du dx y ′ y
= = −
dz dz dx kx 2kx 2
and
d ⎛ du ⎞
⎜ ⎟
d u dx ⎝ dz ⎠
2
y ′′ 3 y′ y
2
= = 3
− 5
+ 7
.
dz dz dx 2
k x 2 2
2k x 2 2
k x 2

Hence
d 2u du 4 3 2 1 y
z2 2
+z = x y ′′ + ,
dz dz 9 9 x
and using equation (6.87) for y ′′ , we obtain

d 2u du 4 ⎛ y ⎞ 1 y ⎛ 1⎞
z 2
+z = − k 2 x 3 ⎜⎜ ⎟⎟ + = −⎜ z 2 − ⎟u ,
⎝ 9⎠
2
dz dz 9 ⎝ x⎠ 9 x
or
d 2u du ⎛ 2 1 ⎞
z 2
+z + ⎜ z − ⎟u = 0 , (6.88)
dz 2
dz ⎝ 9⎠
which is the Bessel equation of order one-third. Since equation (6.88) has the general
solution
u ( z ) = AJ 1 ( z ) + BJ −1 ( z ) ,
3 3

then equation (6.87) has the solution


⎡ ⎛2 3 ⎞ ⎛ 2 3 ⎞⎤
y ( x ) = x ⎢ AJ 1 ⎜ kx 2 ⎟ + BJ −1 ⎜ kx 2 ⎟⎥ .
⎣ 3⎝3 ⎠ 3⎝3 ⎠⎦
g

Exercise 6.3
1. Express J 5 (x ) in terms of J 0 ( x ) and J 1 ( x ) .

2. Show that 4 J ′p′ ( x ) = J p + 2 ( x ) − 2 J p ( x ) + J p − 2 ( x ) .

3. [ ′
]
Show that x p Y p ( x ) = x p Y p −1 ( x ) .

4. [ ′
]
Prove that x − p Y p ( x ) = − x − p Y p +1 (x ) .

231
5. Prove the following identities:
(a) ∫ J (x )dx = − J (x ) + C
1 0

∫ x J (x )dx = 2 xJ (x ) − x J 0 (x ) + C
2 2
(b) 1 1

(c) ∫ xJ (x )dx = xJ (x ) + C
0 1

∫ x J (x )dx = (x − 4 x )J 1 ( x ) + 2 x 2 J 0 ( x ) + C
3 3
(d) 0

6. Show that ∫ J ( x )dx = 2


0 x J1 ( x )+ C
7. Show that ∫ J (x )dx = 2[J (x ) + J (x ) + J (x ) + ...] + C
0 1 3 5

In problems 8-11 reduce each given equation to a Bessel equation and solve it
8. 4 x 2 y ′′ + 4 xy ′ + (x 2 − p 2 )y = 0
9. xy ′′ − y ′ + xy = 0

10. xy ′′ + (1 + 2k ) y ′ + xy = 0

11. y ′′ + k 2 x 4 y = 0

6.4 Legendre Polynomials


Another very important differential equation that arises in many applications is the
Legendre differential equation
(1 − x )y ′′ − 2 xy ′ + p( p + 1)y = 0 ,
2
(6.89)
where p is a fixed parameter. Any solution of equation (6.89) is called a Legendre
function. This has a regular singular point at 1, and hence a series solution for (6.89)

about x = 1 may be obtained by the method of Frobenius. Substituting y = ∑ c n x n and
n=0

its derivatives into equation (6.89), we obtain

(1 − x )∑ c n(n − 1)x
∞ ∞ ∞
2
n
n−2
− 2 x ∑ c n nx n −1 + p ( p + 1)∑ c n x n = 0 ,
n=2 n =1 n=0

or

∑ {(n + 2)(n + 1)c n+ 2 − c n [n(n + 1) − p( p + 1)]}x n = 0 (6.90)


n=0

Setting the coefficients of the sum (6.90) to zero, we obtain the recurrence relation

232
(n + 2)(n + 1)c n + 2 = c n (n 2 + n − p 2 − p ) = c n (n − p )(n + p + 1) .

Thus we have

cn+ 2 = −
( p − n )( p + n + 1) c . (6.91)
(n + 2)(n + 1) n
Therefore
p ( p + 1) ( p − 1)( p + 2 ) c
c2 = − c0 , c3 = 1
2! 3!

c4 = −
( p − 2)( p + 3) = ( p − 2) p( p + 1)( p + 3) c , …
(4)(3) 4!
0

Inserting these values for the coefficients into the power series expansion for y( x ) yields
y ( x ) = c0 y1 ( x ) + c1 y 2 ( x ) , (6.92)
where
x2 x4
y1 ( x ) = 1 − p( p + 1) + ( p − 2) p( p + 1)( p + 3) − ... , (6.93)
2! 4!
x3 x5
y 2 ( x ) = x − ( p − 1)( p + 2) + ( p − 3)( p − 1)( p + 2)( p + 4) − ... (6.94)
3! 5!
Dividing equation (6.94) by equation (6.93), we have
y 2 (x )
= x+
( p 2 + p + 1) x 3 + ... ,
y1 ( x ) 3

which is nonconstant, implying that y1 and y 2 are linearly independent. Thus equation

(6.92) is the general solution of the Legrendre equation (6.89) for x < 1 .

In many applications, the parameter p in the Legendre equation is nonnegative integer.


When this occurs, the right-hand side of equation (6.91) is zero for n = p , implying that
c p + 2 = c p + 4 = c p + 6 = ... = 0 . Thus one of the equations (6.93) or (6.94) reduces to a

polynomial of degree p (for even p , it is y1 ; for odd p , it is y 2 ). These polynomials,


multiplied by an appropriate constant, are called the Legendre polynomials. It
customary to set

cp =
(2 p )! , p = 0, 1, 2, ... , (6.95)
2 p ( p!)
2

233
so, by equation (6.91)
p( p − 1) (2 p − 2)!
c p −2 = − cp = − p
2(2 p − 1) 2 ( p − 1)! ( p − 2)!

c p −4 = −
( p − 2)( p − 3) c =
(2 p − 4)! , …,
4(2 p − 3) 2!( p − 2 )!( p − 4 )!
p−2 p
2
and in general

c p−2k =
(− 1)k (2 p − 2k )! .
2 p k!( p − k )!( p − 2k )!
Then the Legendre polynomials of degree p are given by

Pp ( x ) = ∑
M
(− 1)k (2 p − 2k )! x p −2 k , p = 0, 1, 2,... , (6.96)
k = 0 2 k!( p − k )!( p − 2 k )!
p

where M is the largest integer not greater than p 2 . In particular, we have

P0 ( x ) = 1 , P1 ( x ) = x , P2 ( x ) = 1
2 (3x 2
− 1) , P3 ( x ) = 1
2 (5 x 3
− 3x ) ,

P4 ( x ) = 1
6 (35 x 4
− 30 x 2 + 3 , … )
As a consequence, we have Pp (1) = 1 and Pp (− 1) = (− 1) for all integers p ≥ 0 .
p

We observe that, from equation (6.96), we can write

Pp ( x ) = ∑
M
(− 1)k (x ) ,
d p 2 p−2k
k = 0 2 k!( p − k )! dx
p p

since
d p −1 2 p − 2 k −1
dx p
(
d p 2 p−2k
x ) = (2 p − 2 k )
dx p −1
(x ),
= (2 p − 2k )...( p − 2k + 1)x p − 2 k =
(2 p − 2k )! x p −2 k .
( p − 2k )!
Hence

∑ (− 1) k!( p − k )! (x )
dp M
Pp ( x ) =
1 p!
k 2 p−k
.
2 p p! dx p k =0

234
We may extend the range of this sum by letting k range from zero to p . This extension
does not affect the result, since the added terms are a polynomial of degree < p , so the

p th derivative is zero. Thus

∑ k!( p − k )! (x ) (− 1)
1dp p
p! 2 p −k
Pp =
k
,
2 p p! dx p k =0

and by the binomial formula we have

Pp ( x ) = p
1dp 2
2 p! dx p
(x − 1) ,
p
p = 0, 1, 2, ... (6.97)

Equation (6.97), called the Rodrigues formula, provides an easy way of computing
successive Legendre polynomials.

Legendre polynomials also satisfy the recurrence formula


( p + 1)Pp +1 + pPp −1 = (2 p + 1)xPp , p = 0, 1, 2, ... . (6.98)

Example 6.17
Show that P2 ( x ) = 1
2 (3x 2
−1)

Solution
By the Rodrigues formula, we have

P2 ( x ) =
1 d2 4
2
2 2! dx 2
(x − 2 x 2 + 1) = (12 x 2 − 4) = (3x 2 − 1)
1
8
1
2
g

Example 6.18
Starting with P0 = 1 and P1 = x , calculate the polynomials P2 , P3 and P4 .

Solution
By equation (6.98), we have

235
(2 p + 1)xPp − pPp −1
Pp +1 = ,
p +1
so
3xP1 − P0 3x 2 − 1
P2 = = ,
2 2
5 xP2 − 2 P1 15 x 3 − 5 x − 4 x 5 x 3 − 3x
P3 = = = ,
3 6 2
7 xP3 − 3P2 35 x 4 − 21x 2 − 9 x 2 + 3 35 x 4 − 30 x 2 + 3
P4 = = = .
4 8 8
g

Exercise 6.4
1. Calculate P4 , by means of the Rodrigues formula.
2. Calculate P5 , P6 , P7 and P8 by means of equation (6.98).

3. Prove that P2 p +1 (0 ) = 0 , for all integers p ≥ 0 .

4. Prove that P2 p (0) =


(− 1) p (2 p )! , for p ≥ 0.
2 2 p ( p!)
2

236

You might also like