BESSEL’S DIFFERENTIAL EQUATION AND BESSEL FUNCTIONS:

In the process of following various applied problems, several important ODEs have been found to
precipitate special functions named after their discoverers. One of such equations is called
Bessel’s Equation named after a German astronomer Friedrich Wilhelm Bessel while formulating
a problem in planetary motion. It takes on the general form;

( )
x 2 y ' '+ xy'+ x 2 − n 2 y = 0

Where n is a real number and the system is called Bessel’s equation of the order n .

Another equation worth of mention here is Legendre’s equation after the French mathematician
and astronomer Adrian-Marie Legendre. This takes on the following form;

(1 − x )y' '−2 xy'+n(n + 1)y = 0

2

Our discussion on these equations will focus on;

i. Obtaining the associated series solution

ii. Examining the Fundamental properties.

The D.E x 2 y ' '+ xy'+(x 2 − n 2 )y = 0 where n may take on any real value but is taken to be an
integer is known as Bessel’s Equation of the order n . While seeking a power series solution to
Bessel equation, we observe that the origin is a regular singularity and the only singularity of the
ODE, implying that there exists atleast one Frobenius type series solution convergent on the
deleted neighborhood of the origin. 0  x   To Bessel‘s DE of the form;


y ( x ) =  a k x k + c , a0  0
k =o

To determine the parameters, we substitute the postulated solution and its derivatives into
Bessel’s equation. Having made the necessary provisions for combining the series terms under
one summation identically equal to zero, we obtain the following key results.

( )
x 2 y ' '+ xy'+ x 2 − n 2 y = 0

1
Substituting the associated derivatives we obtain;

   

 (k + c )(k + c − 1)ak x k +c +  (k + c )ak x k +c +  ak x k +c+2 − n 2  ak x k +c = 0

k =0 k =0 k =0 k =0

We equate the sum of coefficients x s + c to zero. Note that the power x s + c corresponds to k = s in
the first, second and fourth series, and to k = s − 2 in the third series. Hence for s = 0 and s = 1 , the
third series does not contribute since k  0 for s = 2,3,... all 4 series contribute, so that we get a
general formula for all these. Equating terms of similar powers in all series and equating to zero
gives;

1. c(c − 1)a 0 + ca 0 − n 2 a 0 = 0(s = 0); k = 0;

2. (c + 1)ca1 + (c + 1)a1 − n 2 a1 = 0; k = 1;

3. (k + c )(k + c − 1)a k + (k + c )a k + a k − 2 − n 2 a k = 0, k  2 i.e s = (2,3,...)

1) Since a 0  0 ;then

(c 2
)
− n 2 a0 = 0
c = n

(c + 1) 2

− n2 = 0
2) c = n
 (c + 1) − n 2  0; a1 = 0
2

− ak − 2
3) ak = , k2
(k + c )2 − n 2
Note;

+ x + (x 2 − n 2 )y = 0
2d2y dy
x 2
dx dx

x  + (x 2 − n 2 )y = 0
d  dy 
2
x2 
dx  dx 
Is Bessel differential of order n . These functions are of frequent use in solutions of many types of
potential problems involving circular cylindrical boundaries as well as in the applications, in such

2
fields as elasticity, fluid flow, electrical field theory, frequency and phase modulations and
aerodynamic flutter analysis.
We always seek to formally obtain solutions of Bessel D.E. we initially suppose that the constant
n is real since the quantity n 2 appears in the ODE, we may also consider it to be non-negative
without loss of generality.
We note that the Bessel differential equation is a typical example of linear variable-coefficient,
second-order differential equations given by the ODE above. It has only one singular
point x = 0 and this singular point is clearly a regular singular point since with p(x ) = x 2 , q(x ) = x

q(x ) x 2 r (x ) x 2 (x 2 − n 2 )
and r (x ) = x − n , we have lim x
2
= 2 = 1 and lim x 2
2
= lim = −n 2 both finite.
x →0 p( x ) x x →0 p(x ) x → 0 x 2

Consequently by the theorem we identified, there exists at least one Frobenius-type solution of the
form.

y =  ak x k + c
k =0

Which is convergent for all values of x . Substituting this series solution, along with its first and
second derivatives into the Bessel ODE, we obtain the expression.

x 2  (k + c )(k + c − 1)ak x k + c − 2 + x  (k + c )ak x k + c −1 + (x 2 − n 2 ) ak x k + c = 0

  

k =0 k =0 k =0

We rewrite

 (k + c )(k + c − 1) + (k + c ) − n a x
 
2
k
k +c
+  ak x k + c + 2 = 0
k =0 k =0

By separating the first two terms from the first summation and then adjusting the subscripts and
exponents in the remaining terms by the addition of 2 to the summation variable, we may absorb
the 2 summations into a single one and thus write.

(c   
− n 2 )a0 x c + (c + 1) − n 2 a1 x c +1 +  (c + k + 2 ) − n 2 ak + 2 + ak x k + c + 2 = 0  

2 2 2

k =0

This equation gives us the results earlier obtain

(c 2
)
− n 2 a0 = 0 a0  0 For non-degenerate Frobenius-types series solution

(c + 1) − n a = 0
2 2
1

(k + c + 2) − n a 2 2
k +2 + ak = 0; k = 0,1,2,...
3
From our Obtained results, the solution to Bessel’s equation depends on the nature of n from c = n
or c = − n ;

Case 1:
c = n, n0
− ak −2
ak = k 2
(k + c )2 − n 2
− a k −2
= k 2
k (k + 2n )

a0 = Arbitrary, a1 = 0

a3 = 0, a5 = 0, a2 m+1 = 0 All odd Coefficients

− a0
a2 =
2 1!(n + 1)
2

a0
a4 =
2  2!(n + 1)(n + 2)
4

− a0
a6 =
2  3!(n + 1)(n + 2)(n + 3)
6

Generalizing

a2 m =
(− 1)m a0
2 2 m  m!(n + m )(n + m − 1)...(n + 1)

a2 m =
(− 1)m a0
m
2 2 m  m! (n + s )
s =0

For the postulated solution


y =  ak x k + n
k =0

y = a0 x n 

(− 1)m x 2m n0
m
m =0
2 m! (n + s )
2m

s =1

 

y = x n  ak x k = x n a0 + a1 x + a2 x 2 + a3 x 3 + ...
k =0
4
Observe that the denominator in the k th term i.e. k = 2m contains that factor m!, also present in the
m
same is  (n + s ) which could become (n + m)!if multiplied by n! The k th term of the series
s =1

 x
involves the term   while the multiplying factor in front involves only x n
2
2m


(− 1)m  x 
y ( x ) = a0 x n  2 n0
m
n =0
m! (n + s )
s =1

x
A particular solution applying the above observed factorials and introducing in the factor
2
1
instead of x can be obtained by so choosing a0 = n
, yielding
2 n!
n+2m


(− 1)  x  m

y(x ) =  2 n0

m =0 m!(n + m )!
To allow for the possibility that n is any non-negative number, integral or non-integral, we are led
to invoke the generalization of the factorial function from the integral definition of the Gamma
function for n ,

(n ) =  u n −1e −u du
0

(n + 1) = n!
(n + m + 1) = (n + m)!
Then
n+2m


(− 1)m  x 
y(x ) =  2
n0
m = 0 m! (n + m + 1)

The above expression is the solution of Bessel’s equation conventionally denoted as

n+2m


(− 1)  x  m

J n (x ) =  2 n0
m = 0 m!(n + m + 1)

5
The solution is known as Bessel’s Function of the 1st kind of order n and is usually denoted
by J n ( x ) .

We shall investigate the solution of Bessel’s differential equation for the second indicial root a bit
later for the second root c = − n , n  0 .

Example 1:

Show that J 1 (x ) =
2
sin x and thereby invoking the variation of parameters, verify that the
2
x

(
solution to x 2 y ' '+ xy'+ x 2 − 1 4 y = 0 is ) A cos x + B sin x
x
, 0 x

Solution
n+2m


(− 1)m  x 
J n (x ) =  2 n0
m = 0 m!(n + m + 1)

1
2m+

J 1 (x ) = 

(− 1) m
x 2

1
m =0 2m+  1
2
2 2
m! m + 1 + 
 2
Invoking Legendre’s formula

 1  (2k )! 
 k +  =
 2 2 2 k k!

k = m + 1;

 1  (2m + 2 )! 
 m + 1 +  = 2 m + 2
 2 2 (m + 1)!
From the numerator (2m + 2)!= (2m + 1)!2(m + 1)

= (2m + 1)!2(m + 1)

 1  (2m + 1)!2(m + 1) 
 m + 1 +  =
 2 2 2 m + 2 (m + 1)!

6
 2(2m + 1)! 
=
2 2 m+ 2 m!
1
2m+

J 1 (x ) = 

(− 1) x 2
m

2.(2m + 1)! 
1
m =0 2 m + 2
2
2  m! 2m+2
2 .m!
1
2m+

J 1 (x ) = 

(− 1) m
x 2

m =0 2m+
1
m
(2m + 1)! 
 !
2 2
2 2 m +1
2 !
.m
1
2m+

=

(− 1)m x 2

m = 0 (2m + 1)! 
1
2
2

J 1 (x ) =
2 
(− 1)m x 2m+1
x

m =0 (2m + 1)!
2

2
= sin x
x
As required

(
x 2 y ' '+ xy'+ x 2 − 1 y = 0
4
)
, and J 1 (x ) =
1 2
This is Bessel’s equation with n = sin x
2 2
x
Dropping the constants, the 1st linearly independent solution is
sin x
y1 =
x
y2
The second linearly independent solution  c is found by the method of variation of
y1

parameters.
q(x )
−  p ( x ) dx
e
y2 = y1  dx
 y1 2
7
q( x ) = x , p(x ) = x 2
x
−  2 dx
sin x e x
x   sin x 
y2 = 2
dx
 x 
 
1
−  x dx
sin x e
x  sin x
y2 = 2
dx

x
1
−  x dx 1
e =
x
sin x
y2 = − cot x
x
cos x
y2 = −
x
A cos x + B sin x
y= , 0 x
x

Example 2:
u
By making the substitution y = . Find the solution to the ODE xy' '+5 y '+ xy = 0 that is bounded
x2
at the origin, Express your answer in the form y =  ak x k
k

Solution
u
Differentiate y =
x2
u ' 2u
y' = −
x 2 x3
u ' ' 2u ' 2u ' 6u
y' ' = 2 − 3 − 3 + 4
x x x x
 u ' ' 4u ' 6u   u ' 2u  xu
x  2 − 3 + 4  + 5 2 − 3  + 2 = 0
x x x  x x  x
u' ' u'  1 4 
+ +  − u = 0
x x 2  x x3 
8
Multiplying through by x 3
( )
x 2u ' '+ xu'+ x 2 − 4 u = 0
Which is Bessel’s form of ODE of order n = 2
The solution to Bessel’s equation bounded at the origin is of the form of the Bessel function of the
first kind of order n=2 obtained by substituting in the Generic Bessel function;

J n (x ) =  2 k + n
(− 1) x 2 k + n
k

k =0 2 k!(n + k + 1)

(− 1) x 2 k + 2
u ( x ) = J 2 ( x ) =  2 (k +1)
k

k =0 2 k!(k + 2 )!

u
But y =
x2

y=

(− 1)k x 2 k x 
k =0 2
2 ( k +1)
k!(k + 2 )!

Example 3:
By making the substitution y = ux3 solve the ODE xy' '−5 y '+ xy = 0 . Express your answer in the

form y =  ak x k
k

Solution
y = ux 3
y ' = 3ux 2 + u ' x 3
y ' ' = 3u ' x 2 + 6 xu + u ' ' x 3 + 3u ' x 2
y ' ' = u ' ' x 3 + 6 x 2 u ' + 6 xu

Substituting into the ODE

( ) ( ) ( )
x u ' ' x 3 + 6 x 2 u ' + 6 xu − 5 3ux 2 + u ' x 3 + x ux 3 = 0

u ' ' x 4 + 6 x 3u '+6 x 2u − 5u ' x 3 − 15ux2 + ux4 = 0

u ' ' x 4 + x 3u '−9u ' x 2 + ux4 = 0

( )
u ' ' x 4 + x 3u '+ x 2 x 2 − 9 u = 0 divide through by x2 to obtain Bessel ODE

The above ODE is a Bessel ODE of the order n = 3

9
The solution to Bessel’s equation bounded at the origin is

J n (x ) = 

(− 1)k x 2k +n
k =0 2 2 k + n k!(k + 1 + n)

u (x ) = J 3 ( x ) = 

(− 1)k x 2k +3
k =0 2 2 k +3 k!(k + 4)

u (x ) = 

(− 1)k x 2 k +3
k =0 2
2 k +3
k!(k + 3)!

y = ux3 = 

(− 1)k x 2 k +6 x 
k =0 2
2 k +3
k!(k + 3)!

y =  k +3
(− 1) x 2 (k +3 )
k
x 
k =0 2 k!(k + 3)!

Case 2: c = − n , n  0

We will note that in our treatment of this case, we could legitimately seek for a solution to
Bessel’s equation by considering the key results of substituting the postulated solution and its
derivatives into Bessel’s equation. We note however that such an investment is not necessary. All
we have to do is to replace n in the definition of Bessel’s function of order n of the 1st kind by - n .

(
J −n (x ) =  2 k −n
− 1) x 2 k −n
k
, n >0 (this is meaningful for all n and not just +ve integers)
k =0 2 k!(k − n + 1)
This implies that J −n (x ) is meaningful for all n >0 and represents a solution of Bessel’s equation.

We continue to note that for any positive integer n  (k + 1) , we have

1
= 0;
(− n + k + 1)
for k  (n − 1).
The solution may be re-written alternatively as

J −n (x ) = 

(− 1)k  x  2k −n
 
k =0 k!(k − n + 1)  2 

10
Question:

Verify that J (x ) = 2
cos x .
x
1
−
2

THE GENERAL SOLUTION OF BESSEL’S EQUATION

2k +n


(− 1)  x 
k

Given that J n (x ) =  2 n0

k = 0 k!(k + n + 1)

2k −n


(− 1)  x 
k

And J − n ( x ) =  2 n  0 if n is positive but not an integer, J −n (x ) is not bounded at

k = 0 k! (k + n + 1)

the origin while J n ( x ) may be bounded at the origin.

Illustration:

Expanding J n ( x ) and J −n (x )
n n+2 n+4 n+6
x x x x
J n (x ) =
1 1 1 1
− + − + ...
 
0!(n + 1)  2  1!(n + 2)  2  2!(n + 3)  2  3!(n + 4)  2 
And
−n − n+2 − n+4 − n+6
x x x x
J −n (x ) =
1 1 1 1
− + − + ...
0!(− n + 1)  2  1!(− n + 2)  2  2!(− n + 3)  2  3!(− n + 4)  2 

We note that ifn is a positive non-integer,

J n ( x ) Is bounded/finite (in fact zero) for x = 0 , while J −n (x ) is not bounded. This implies that for

non-integer values of n  0 , the two solutions are linearly independent. Consequently the general

11
solution of the Bessel differential equation in this case is given by the linearly independent linear
combination of J n ( x ) and J −n (x ) that is

y = An J n (x ) + Bn J −n (x ) ; n  0 And n  0,1,2,3,...

Note the exclusion of the case n  0 . This is because at n = 0 , both J n ( x ) and J −n (x ) degenerate

into J 0 ( x ) .

For positive integer values of n , we find that J n ( x ) and J −n (x ) are linearly dependent and given by

the general relationship

J −n (x ) = (− 1) J n (x )
n

For n = 0 ; J −n (x ) = J 0 (x ) and J n (x ) = J 0 (x )

In case n is an integer, the two solutions J n ( x ) and J −n (x ) are linearly dependent. In order to obtain

a solution that is linearly independent of J n ( x ) , we can use several techniques. If we apply the

method of variation of parameters, setting y1 = J n (x ) , we obtain the general solution.

y = An J n (x ) + Bn J n (x )
dx
xJ n (x )
2

The second solution which is linearly independent is given by

y2 = J n (x )
dx
xJ n (x )
2

This result is valid for any positive real n (regardless of whether n is an integer or not). The
disadvantage with this form of the second linearly independent solution is that we do not have an
explicit expression for it, which could be analyzed, graphed etc.

12
 DEFINITION OF THE EXPLICIT FORM OF THE GENERAL SOLUTION:

To offset the disadvantage discovered in the preceding text, a linear combination of J n ( x )

and J −n (x ) , defined by the expression.

J n (x ) cosn − J −n (x )
Yn (x ) =
sin n

Is often preferred, when n is a positive real non-integer.

If n is not an integer, it follows from the fact that J n ( x ) and J −n (x ) are both solutions of Bessel’s

equation and that

J n (x ) cosn − J −n (x )
Yn (x ) = Is a solution
sin n

From J 1 (x ) = sin x J 1 (x ) = cos x then Y1 (x ) = −

2 2 2
cos x
2 x −
2 x 2 x

If n is an integer, Yn (x ) assumes an indeterminate form

0
. We are led to consider as a possible
0
solution

J p (x ) cos p − J − p (x )
Yn (x ) = lim
p →n sin p

This limit can be determined by L’Hopital’s rule in calculus and this yields a solution Yn (x ) is aka

Bessel’s function of the 2nd kind of the order n or Neumann’s function of the order n . The general
solution to Bessel’s function is given as

Y = AJ n (x ) + BYn (x )

Irrespective of whether n is an integer or not where A and B are arbitrary constants.

13
 PROPERTIES OF BESSEL’S FUNCTIONS:

Bessel’s function of the 1st kind;

2k +n


(− 1)  x 
k

J n (x ) =  2 n0
k = 0 k!(k + n + 1)

2k 4


(− 1)k  x  2
 x
 
n = 0 J 0 (x ) =  2 = 1 − +  2 + − + ...
x 2
k =0 (k!)2 4 (2!)

BESSEL FUNCTIONS OF THE FIRST KIND OF ORDER ZERO AND ONE

a) Bessel function of the 1st kind of order zero J 0 ( x )

4
 x
 

J 0 (x ) = 
(− 1)  x 
k 2k
x 2
= 1 − +  2 +
2
2  
k = 0 (k!)  2  4 (2!)

Putting in some values

J 0 (0) = 1.00000, J 0 (1) = 0.76519, J 0 (2) = 0.22389, J 0 (3) = −0.26005, J 0 (4) = −0.39714.

The graph of J 0 ( x ) is oscillatory in nature, reminding us of graphs of damped vibrations. Its roots

(zeros) i.e. J 0 ( x ) = 0 are points of intersection with x -axis and there are infinitely many roots

and all these are real and positive.

b) Bessel function of the 1st kind of order one J1 (x )

J1 ( x ) = 
(− 1)k  x 2k +1
 
k =0 k!(k + 1)!  2 

Putting in some values

J1 (0) = 0.00000, J1 (1) = 0.44005, J1 (2) = 0.57672, J1 (3) = −0.33905, J1 (4) = −0.06604.
The graph is oscillatory in character resembling damped sinusoids.

Note: Graphs of J 0 ( x ) and J1 (x ) (Check addendum for graphs)

14
Summary of Observations from the graphs:

1) The graph is oscillatory in nature resembling the graph of damped oscillations.

2) The graph also reveals that there are roots of the equation J 0 ( x ) = 0 also known as the

zeros of J 0 ( x )

3) Investigation reveals that there are infinitely many roots which are all real.
4) The same applies to J1 (x ) .

5) Between two zeros of J 0 ( x ) , lies a zero of J1 (x ) .

6) The graphs of J n ( x ) for other values of n are similar to those of the plots above.

It is possible to show that the zeros of J n ( x ) are real and that between 2 successive positive zeros

of J n ( x ) , there is only one zero of J n +1 (x ) .

Table of positive zeros for some Bessel function orders

zeros 1 2 3 4 5
J 0 (x ) 2.4048 5.5201 8.6537 11.7915 14.9309

J1 ( x ) 3.8317 7.0156 10.1735 13.3237 16.4598

J 2 (x ) 5.1356 8.4172 11.6198 14.7960 17.9598

J 3 (x ) 6.3802 9.7610 13.0152 17.2235 19.4094

Note that if we take the difference between successive zeros, of J o , we obtain 3.1153, 3.1336,

and 3.1394.
Suggesting the conjecture that this difference approaches  = 3.14
A similar observation on the difference of successive zeros of J1 (x ) and J 2 (x ) also leads to such a
conjecture. This conjecture can actually be proved and since successive zeros of
sin x and cos x differ by  , the approximate description of the Bessel function of the 1st kind of
order n as a damped sine wave is further warranted. It can hence be shown that for large values
of x that

15
  n  
J n (x ) =
2
cos x  x − −
x  2 4 

 
J 1 (x ) =
2 2
cos x −  = sin x
2 x  4 x
Bessel’s function has been referred to as “a damped sine wave” since successive zeros of the
sine and cosine waves differ by  = 3.141592653...
The wave characteristics of Bessel functions seem very much like the shapes of water waves
which could be generalized, for example, by dropping a stone into the middle of a large puddle
of water. The water waves generated would then resemble the surface of revolution generated
by J 0 , J1 (x ),... about the J n ( x ) -axis ( y -axis). It does indeed turn out that in the theory of

hydrodynamics, circular vibrating membranes, etc, the shapes of the waves ensuing therein are
answered by Bessel functions.

BESSEL FUNCTIONS OF THE SECOND KIND;

(Order zero, one and two, Y0 (x ), Y1 (x ) and Y2 (x ) )

J n ( x ) cos n − J − n ( x )
Yn ( x ) =
sin n
In the same way, Bessel functions of the second kind Y0 ( x),Y1 (x ),...Y2 (x ) generate the graphs

below.

We also note that the difference between the successive roots of Yn (x ) n = 0,1,2,3,4,... tend to

some constant. In fact the data Y0 ( x) = 0 and Y1 (x ) = 0 clearly suggest that these differences

approach the constant  , as x recedes to infinity. No wonder then the Bessel functions Yn (x ) are

also often said to behave like “damped sine waves”.

16
IMPORNTANT IDENTITIES OF BESSEL’S FUNCTIONS OF THE 1ST KIND:

Identity 1:
d
dx
 
x n J n (x ) = x n J n −1 (x )

when  = 1

d n
dx
 
x J n ( x ) = x n J n −1 ( x )

 x J (x)dx = x J (x) + c
n n
n −1 n

Example 1:

 xJ (x)dx = xJ (x) + c
0 1

Proof;
d n
dx
 
x J n (x ) = x n J n −1 (x )

J n (x ) = 

(− 1)
k
x 
2k +n

 
k =0 k!(n + k + 1)  2 

=

(− 1)k  2 k + n  x  2k + n
 
k = 0 k!(n + k + 1)  2 

On multiplying both sides by x n , we obtain

(− 1) x n x 
2k +n
 k
x n
J n (x ) =   
k = 0 k!(n + k + 1)  2 

(− 1)k x 2(k + n )  x 
2k +n

=  
k = 0 k!(n + k + 1)  2 

Differentiating both sides, we get

17
 d   (− 1)  2 k + n x 2(k + n ) 
 
k
x J n (x ) =
d n
 
dx dx  k =0 k!2 2 k + n (n + k + 1)

=

(− 1)  2 k + n 2(n + k )x 2(k + n )−1
k

k =0 k!2 2 k + n (n + k + 1)

Using (k + 1) = k(k )

We get
  (− 1)k  2 k + n 2(n + k )x 2(k + n )−1 
d n

x J n (x ) =  
 k =0 k!2 (k + n )(n + k ) 
2k +n
dx

= x n 

(− 1)k  2 k + n−1 x 2(k + n )−1
k =0 k!2 2 k + n −1 (n + k )

= x n 

(− 1)k x  2k +n−1
 
k =0 k!(n + k )  2 

= x  n

(− 1)
k
x 
2 k + n −1

 
k =0 k!((n − 1) + k + 1)  2 

= x n J n−1 (x )

For  = 1 , we have
d n
dx
 
x J n ( x ) = x n J n −1 ( x )

Identity :2

 
x J n (x ) = −x − n J n +1 (x )
d −n
dx

When  = 1

 
x J n ( x ) = − x − n J n +1 ( x )
d −n
dx

x
−n
J n+1 (x )dx = − x − n J n (x ) + c

18
Example: 2

 J (x)dx = − J (x) + c
1 0

Identity: 3

J n +1 ( x ) = J n ( x ) − J n −1 ( x )
2n
x

J n −1 ( x ) + J n +1 ( x ) = J n (x )
2n
x

J n−1 (x ) − J n+1 (x ) = 2 J n (x )
'

(The recurrence Relationship for Bessel’s Function);

Bessel’s function for higher order is represented in terms of Bessel’s function for lower order.

Evaluate;

1) J 3 (x ) Using the above relationship.

2

J
1+
1 ( x ) = 2n J 1 ( x ) − J − 1 ( x )
2 x 2 2

2  sin x 
=  − cos x 
x  x 
2) J
−
3 (x ) ;
2

Using

J n −1 (x ) = J n (x ) + J 1 (x )
2n
x 2

1
n=− ;
2

J 1 (x ) + J 1 (x )
1
=−
x −2 2

19
 2  cos x 
=−  + sin x 
x  x 

EVALUATION OF INTEGRALS INVOLVING BESSEL:

Consider a general integral

I =  x m J n (x )dx n And m are non-negative integers,  is a constant.

Transforming the Integral

I =  x m−n−1 x n−1 J n (x )dx

Integrating by parts u = x m − n =1
du = (m − n − 1)x m − n − 2 dx
dv = x n +1 J n (x )dx

x n +1 J n +1 (x )
1
v=

J n −1 (x )dx = x n J n (x ) + c
1
x
n

J n (x )dx = x m J n +1 (x ) − (m − n − 1) x m−1 J n+1 (x )dx

1 1
x
m

 

The integral can also be recast in the form

x
m+ n −1
x −(n−1) J n (x )dx And applying integration by parts

u = x m + n −1 dv = (m + n − 1)x m+ n −2 dx

dx = x −(n−1) J n (x)dx
From the properties of Bessel Functions of the 1st kind

J n +1 (x )dx = − x − n J n (x )

1
x
−n

x −(n =1) J n −1 (x )

1
v=−


20
   x m J n (x )dx = − x m J n −1 (x ) + (m + n − 1) x m −1 J n −1 (x )dx
1 1
 

Analogous to the solution we have with trigonometric functions above constitute reduction
formulae which provide effective means for computing or evaluating integrals of the general
form
I =  x m J n (x )dx

In their applications, it is instructive to observe two facts

1) If the integers m and n are either or both even or odd, the
factors (m − n −1) and (m + n −1) can never vanish i.e. probably the last value they reduce to is 1
and this will always leave a residue integral that has to be evaluated by other techniques such as
Numerical methods.
2) If one of the integers m or n is odd and the other even, it is possible that the
factors (m − n −1) or (m + n −1) vanish at some stage and at that point, the integral will have been
fully evaluated.

21
APPENDIX : IMPORNTANT IDENTITIES FOR BESSEL FUNCTIONS:

d n
dx
 
x J n (x ) = x n J n −1 (x )

1. when  = 1
d n
dx
 
x J n ( x ) = x n J n −1 ( x )

 
x J n (x ) = −x − n J n +1 (x )
d −n
dx
2. when  = 1

 
x J n ( x ) = − x − n J n +1 ( x )
d −n
dx

J n (x ) = J n −1 (x ) − J n (x )

d n
dx x
3. when  = 1

J n ( x ) = J n −1 ( x ) − J n ( x )
d n
dx x
Recurrence

J n (x ) = J n (x ) − J n +1 (x )

d n
dx x
4. when  = 1

J n ( x ) = J n ( x ) − J n +1 ( x )
d n
dx x


J n (x ) = J n −1 (x ) − J n +1 (x )
d
dx 2
5. when  = 1

J n ( x ) = J n −1 ( x ) − J n +1 ( x )
d 1
dx 2

22
J n +1 ( x ) = J n ( x ) − J n −1 ( x ) ( Recurrence)
2n
x

J n −1 ( x ) + J n +1 ( x ) = J n (x )
2n
x

J n−1 (x ) − J n+1 (x ) = 2 J n (x )
'

23
EXAMPLES
Evaluate the following integrals;

 x J (x)dx  xJ (x)dx
3
a) 1 e) 1

 x J (x )dx x J 1 (x )dx
3 −1
b) 0 f)

 x J (x )dx  J (x)dx
6
c) 1 g) 2

d)  J (x)dx
3 h)  J 0 (x )cos xdx = xJ0 (x )cos x + xJ1 (x )sin x + c

i) Solving definite integrals in Erwin Kreyszig

ii) Advanced Engineering Mathematics by Wiley and Barett.

Solutions

a)

 x J (x)dx
3
1

Using I =  x m J n (x )dx

m = 3, n = 1,  = 1
m − n −1 = 3 −1 −1 = 1
m + n −1 = 3 + 1 −1 = 3

Then using the first relationship

 x J (x)dx = x J (x) −  x J (x)dx

3 3 2
1 2 2

At the next relationship, we have

x
2
J 2 (x )dx = − x 2 J 1 (x ) + 3 xJ 1 (x )dx

At the next relationship, we have

 xJ (x)dx = − xJ (x) +  J (x)dx

1 0 0

On adding up

 x J (x)dx = x J (x) + x J (x) + 3xJ (x) − 3 J (x)dx

3 3 2
1 2 1 0 0

24
b)

 x J (x)dx , m = 3, n = 0,  = 1, m − odd, n − even, we can’t have a residual integral,

3
0

Using relationship one,

 x J (x)dx = x J (x) − 2 x J (x)dx ,

3 3 2
0 1 1

Using the same identity the second time

 x J (x)dx = x J (x) + c
2 2
1 2

Using the property J n +1 ( x ) = J n ( x ) − J n −1 ( x )

2n
x

J 2 (x ) = J1 ( x ) − J 0 ( x )
2
x
2x2
x 2 J 2 (x ) = J1 ( x ) − J 0 ( x )
x
x 2 J 2 (x) = 2 xJ1 (x) − J 0 (x)


  x 3 J 0 (x )dx = x 3 J 1 (x ) − 2 2 xJ1 (x ) − x 2 J 0 (x ) 
= x3 J1 (x) − 4 xJ1 (x) + 2 x 2 J 0 (x)

 x J (x)dx
3
0  
= xJ1 (x) x 2 − 4 + 2 x 2 J 0 (x) + A

c)

 J (x)dx
3

The Integral can be rewritten as

 J (x)dx =  x x J 3 (x )dx
2 −2
3

Invoking the theorem of integration by parts

 udv = uv −  vdu
du
u = x2 , = 2 x; du = 2 xdx
dx
dv = x − 2 J 3 ( x )dx
25
Using the property

 
x J n (x ) = −x − n J n +1 (x )
d −n
dx
 − x
−n
J n +1 (x ) = x − n J n (x )

v = − x −2 J 2 (x )

 J (x)dx =  x x 
J 3 (x ) dx = − J 2 (x ) +  x −2 J 2 (x ) 2 xdx
2 −2
Hence 3

= − J 2 (x ) + 2 x −1 J 2 (x )dx

Using the reduction formula

−1
J n (x )dx = x m J n −1 (x ) + (m + n − 1) x m−1 J n−1 (x )dx
1
x
m

 
m = −1, n = 2,  = 1

x
−1
J 2 (x )dx = − x −1 J1 (x ) + 0 x −2 J1 (x )dx

= − x −1 J1 (x ) + c

 J (x )dx = − J (x ) + x J (x ) + A
2
3 2 1

Rewriting J 2 (x ) using the recurrence relationship

J n +1 ( x ) = J n ( x ) − J n −1 (x )
2n
x
For n = 1

J 2 (x ) = J1 ( x ) − J 0 ( x )
2
x

 J 3 (x )dx = − x J1 (x ) + J 0 (x ) + x J1 (x ) + A
2 2

 J (x )dx = J (x ) + A
3 0

26
 d) J 0 cos xdx

Using integration by parts

u = J 0 cos x

= J 0 ( x )(− sin x ) − J 1 ( x ) cos x

du
dx
= −J 0 ( x )sin x + J 1 ( x ) cos x 
dv
= 1, v = x
dx
Invoking integration by parts

 udv = uv −  vdu
 1 J 0 cos xdx = xJ0 cos x +  xJ 0 (x )sin x + xJ1 (x )cos xdx

 1 J 0 cos xdx = xJ0 cos x +  xJ 0 (x )sin x +  xJ1 (x )cos xdx

Integrating the 1st integral using integration by parts

= xJ 0 ( x )
dv
u = sin x
dx
= cos x, v = xJ1 ( x )
du
dx

 udv = uv −  vdu
 xJ (x )sin xdx = xJ (x )sin x −  xJ (x )cos xdx
0 1 1

 J (x )cos xdx = xJ (x )cos x + xJ (x )sin x −  xJ (x )cos xdx +  xJ (x )cos xdx

0 0 1 1 1

 J (x )cos xdx = xJ (x )cos x + xJ (x )sin x

0 0 1

27
Evaluate the integral
J 2 (3 x )
 x2
dx

Multiplying the denominator and numerator by 9x 2

9 x 2 J 2 (3x ) 2 J (3 x )
 9 x 2 x 2 dx = 9  (3x ) 2x 4 dx
1

Using integration by parts

u = (3x ) J 2 (3x ),
dx
dv =
2

x4
−1
u ' = (3x ) J1 (3x )  3dx v = 3
2

3x
1 − 3 dx 
=  J 2 (3x ) + 3 3xJ1 (3x )  2 
9 x x 

 (3x )J (3x ) x
dx
1 2

u = 3 xJ1 (3 x )
dx
dv =
x2
−1
u ' = 3  3 xJ 0 (3 x ) v=
x

= −3J1 (3x ) +  3  3J 0 (3x )dx

 x (3x )dx = 9  x J (3x ) + 3(− 3J (3x ) + 9 J (3x )dx)

J 2 1−3 
2 2 1 0

−1
= J 2 (3 x ) − J1 (3 x ) + 3 J 0 (3 x )dx
3x

28
Evaluate:

 x J (x )dx
3
1

m = 3, n = 1,  = 1
−1
remember  x m J n (x )dx = x m J n −1 (x ) + (m + n − 1) x m−1 J n−1 (x )dx
1
 
m = 3, n = 1
= − x 3 J 0 ( x ) + 3 x 3 J 0 ( x )dx
using the reduction formula

= − x 3 J 0 ( x ) + 3 x 3 J 1 ( x ) − 2  x 2 J 1 ( x )dx 
= − x 3 J 0 ( x ) + 3 x 3 J 1 ( x ) − 6  x 2 J 1 ( x )dx
using the identity

 x J (x )dx = x J (x ) + c
n n
n −1 n

 x J (x )dx = x J (x ) + c
2 2
1 2

= − x 3 J 0 ( x ) + 3 x 2 J1 ( x ) − 6 x 2 J 2 ( x ) + D

J n ( x ) − J n −1 ( x )
2n
J n +1 =
x
J1 ( x ) − J n −1 ( x )
2
J2 =
x
J 2 (x ) = J (x ) − J 0 (x )
2
x
= x J 0 ( x ) + 3 x 2 J1 ( x ) − 12xJ1 ( x ) + 6 x 2 J 0 ( x )
3

( ) (
= 6x 2 − x3 J 0 (x) + 3x 2 −12x J1 (x) + D )

29
Evaluate:

 x J (x )dx
6
1

m = 6, n = 1,  = 1, using identified relationship

 x J (x )dx = − x J (x ) + 6 x J (x )dx
6 6 5
1 0 0

using

J n (x )dx = x m J (n +1) (x ) − (m − n − 1) x m−1 J n+1 (x )dx

1 1
x
m

 
 x J (x )dx = x J (x ) − 4 x J (x )dx
5 5 4
0 1 1

and  x J ( x )dx = x J ( x ) − 2  x J ( x )dx

4
1
4
2
3
2

utilising the relationship

 x J (x )dx = x J (x ) + c , n = 3
n n
n −1 n

 x J (x )dx = x J (x ) + c
3 3
2 3

  x J ( x )dx = − x J ( x ) + 6x J ( x ) − 4(x

6
1
6
0
5
1
4
J 2 (x ) − 2 x 3 J 3 (x ) + c )
= − x 6 J 0 ( x ) + 6 x 5 J 1 ( x ) − 24 x 4 J 2 ( x ) + 48x 3 J 3 ( x ) + D

30
Show that  J (x ) cos xdx = xJ (x ) cos x + xJ (x )sin x + c
0 0 1

 J (x ) cos xdx
0

let u = J 0 ( x ) cos x
from
d −n
dx
( )
x J n ( x ) = x − n J n +1 ( x )

= J 0 ( x ) (cos x ) + cos x ( J 0 ( x ))
du d d
dx dx dx
= J 0 ( x )(− sin x ) − cos xJ1 ( x )
= −J 0 ( x ) sin x + J 1 ( x ) cos x 
dv = 1
v =  1dx = x

 udv = uv −  vdu
 J (x ) cos xdx = xJ (x ) cos x +  (sin xJ (x ) + cos xJ (x ))xdx
0 0 0 1

= xJ ( x ) cos x +  xJ ( x ) sin xdx +  xJ ( x ) cos xdx

0 0 1

=  xJ ( x ) sin xdx
0

reducing

 xJ (x )sin xdx
0

u = sin x
du = cos x
dv = xJ 0 ( x )
v =  xJ 0 ( x )dx

from
d n
dx
( )
x J n ( x ) = x n J n −1 ( x )

 x J (x )dx = x J (x ) + c
n n
n −1 n 1

 xJ (x )dx = xJ (x )
0 1

utilising

 udv = uv −  vdu
 x sin x =J (x )dx = x sin xJ (x ) cos xdx
0 1

substituting in the original expression

 J (x ) cos xdx = xJ (x ) cos x + xJ (x )sin x −  xJ (x ) cos xdx +  xJ (x ) cos xdx

0 0 1 1 1

 J (x ) cos xdx = xJ (x ) cos x + xJ (x )sin x + c

0 0 1

31
f)

x
−1
J1 ( x )dx =  x −2 x1 J1 ( x )dx
dv
= x −2
dx
v = − x −1
u = x1 J 1 ( x )

= xJ 0 ( x )
du
dx
utilising  udv = uv −  vdu

= − x −1  x1 J1 ( x ) +  x −1 xJ 0 ( x )

= − J1 (x ) +  J 0 (x )dx

g)

 J 2 ( x )dx
using relationship
−1
J n (x ) = x m J n −1 (x ) + (m + n − 1) x m−1 J n−1 (x )dx
1
x
m

 
= − J 1 ( x ) +  x −1 J 1 ( x )dx

= − J 1 ( x ) − J 1 ( x ) +  J 0 ( x )dx

= −2 J 1 (x ) +  J 0 ( x )dx

32
Show that;


2

 J (x cos )cosd = x
sin x
0
0

 
2 2
 (− 1)k (x cos )2 k 
0 0
J ( x cos ) cosd = 0  22k (k!)2  cosd



=  2k
(− 1)k x 2 k 2 (cos )2 k +1 d
2 
k = 0 2 (k!) 0

Using Wallis’s Formula;

2  4  6 (2k ) 2 2 k (k!)
2 2

 cos d =
2 k +1
=
0
1  3  5 (2k + 1) (2k + 1)!

=

(− 1)k x 2 k  2 2 k (k!)2 x x
k = 0 2 (k!)
2k 2
(2k + 1)! x
1  (− 1) x 2 k +1
k
= 
x k =0 (2k + 1)!
sin x
=
x

Walli’s second Formula; (Check out this in Kreyszig)


1 3  5...(2k − 1) 
0
2
cos2 kd =
(2k )
.
2

=
(2k )!  
2 (k!) 2
2k 2

33
THE ALTERNATIVE FORMS OF BESSEL’S DIFFERENTIAL EQUATION:

( )
x 2 y ' '+ xy'+ x 2 − n 2 y = 0
If x = z
dx = dz
dy dy dz
= .
dx dz dx
dz 1
=
dx 
dy 1 dy
=
dx  dz
d 2 y d  dy  1 d2y
 2 = = y ' ' =
dx dx  dx  2 dz 2

(z )2 1 d 2 y  z  dy

+   + (z ) − n 2 y = 0
 dz    dz
2 2
2

y  z  dy
 
2
(z )2 d +   + (z ) − n 2 y = 0
2

dz    dz
2

 y (z ) = c1 J n (z ) + c2Yn (z )

Question;

State the general solution of 4 x 2 y ' '+4 xy'+(2 x 2 − 1)y = 0 .

Solution;
This can be rewritten as
 1  2  1  2 
x y ' '+ xy'+ 
2
x −    y = 0
 2   2  
1
This is an alternative form of Bessel’s function of the order n = , and since n is not an integer
2

34
  x   x 
y = AJ 1   + BJ 1  
2 2 2 2
−

J 1 (z ) = sin z , J 1 ( z ) =
2 2
cos z
2 z −
2 z
 x   x 
a0 cos  + b0 sin 
y=  2  2
x

35
 1 3
Solve the ODE y' '+9 xy = 0 : y = x 2u
, z = 2x 2

Changing the independent variables

1 1
dz = x 2 x 2 dx
3
1
dz = 3 x 2 dx
1
dz 1 z 3
= 3 x 2 = 3 
dx 2

dy dy dz
y' = = 
dx dz dx
dy
dx
1 dy
=3 z 3
2 dz
( )
d 2 y d dy 
y' ' = =
dx 2 dx  dx
d 2 y d  z 13 dy 
= 3 ( )
but dx =
dz
dx 2 
dx  2 
dz  3z 3
2
1
( )
dy
dx 2
=3 z 3
2
1
( )
d  z 3 dy 
3
1

dz  2 dz 
 ( )
d2y
dx 2
=9 z
2
( ) 2
3 d 2 y  3 − 2 3 dy  13
+
dz 2  2 2 3
z
dz 
z

y' ' = 9 z( 2) 2
3 d 2 y  3 −13 dy 
+
dz 2  2 2 3
z
dz 

substituting and multiplying through w ith ( z ) (2) 3

1 2
3

( 2)
9z
d2y 3
2

dz 2
+
3
2 z
3
−1 dy
3
dz
+9 z
2
( ) 2
3
y=0
2
d 2 y dy
3z 2 + + 3 zy = 0
dx dz

36
Changing the dependent variables to the independent variables

y = x 2 u , z = 2( x )
1 3
2

y= z ( 2) 1
3
u

y = x 2 u and z = 2(x )
1 3
2

y= z ( 2) 1
3
u

dy d  dy  d  z  3  z
( )
1
1 du 1 1 −2
=   =   u  = 3
+  1 z 3u
dz dz  dz  dz  2   2 dz 2
  2 3

d 2 y d  dy  d  z  3 du 1 1 − 2 3 
1

=   =   +  z u
dz 2 dz  dz  dz  2  dz 3 2 13 

1
d2y  z  3
d 2u 1 1 2 3 du 1 1 − 2 3
=  + z +  z u
dz 2  2  dz 2 3 2 13 dz 3 2 13

Collecting like terms

1
z 3
d 2 u 2 1 − 2 3 du 2 1 − s 3
  +  z −  z u
2 dz 2 3 2 13 dz 9 2 13
1 2
3 3
Substituting in the ODE and multiplying through with 2 and z
 d 2u du 2 −1  du 1
3z  z 2 + 2 − z u + z + u + 3z 2 u = 0
3
 dz dz 9  dz 3
counting it as a 2 nd ordered D.E with dependent variable u
d 2u du  2 1 
3z2
+ 3z + 3 z − u = 0
dz 2
dz  3
dividing through by 3, we have Bessel's equqation of the order
n= 1 given as
3
d2y dy  2 1 
z2 2 + z + z − u = 0
dz dz  9
we know that it will have a solution stated as

37
u( z ) = AJ 1 ( z ) + BJ 1 ( z )
3 3
1
z 3
but y =   u
2
1
 z 3 
y( z ) =    AJ 1 ( z ) + BJ −1 ( z )
2  3 3

3
but z = zx 2

1
 3  3 
y( x ) = x 2  AJ 1  2 x 2  + BJ −1  2 x 2   0 x
 3   3   
1 1
by making the substitution y = x u , z = kx find the general soluion to he ODE
2 2

k2
y ' '+ y=0
4x

Question:

1 1
k2
Using the substitution y = x u and z = kx Find the solution to y ' '+
2 2
y=0
4x

Solution:

1
z = kx 2

1
1 −2
dz = kx dx
2

dz k 2 zdz
= 1 dx =
dx k2
2x 2

2
z
x= 
k

dz k k2
= =
dx  z  2z
2 
k

38
 dy dy dz
y' = = 
dx dz dx

dy  k 2 
=  
dz  2 z 

d 2 y d  dy 
y' ' = 2 =  
dx dx  dx 

d  k 2 dy 
=  
dx  2 z dz 

k 2 d  k 2 dy 
y' ' =  
2 zdz  2 z dz 

Substituting in the original equation

k2 k 2  − k 2 dy k 2 d 2 y  k 4
y ' '+ y=  + + y=0
4x 2 z  2 z 2 dz 2 z dz 2  4 z 2

k4  − 1 dy d 2 y 
=  + 2 + y = 0
4z 2  z dz dz 

4z 2
Multiplying through by
k4

d 2 y 1 dy
− +y=0
dz 2 z dz
d 2 y dy
z 2 − + zy = 0
dz dz

1
But y = x 2 u

39
 2
z
x= 
k
z
y= u
k

dy u z
= + u'
dz k k

d 2 y 1 dy z u'
2
= + u ' '+
dz k dz k k

z 2u '   u z  z 2u
z  u ' '+ − + u' + =0
k k   k k  k

Multiplying through by coefficient k

z 2u ' '+(2 z − u )u '−u + z 2u = 0

( )
z 2u ' '+ zu'+ z 2 − 1 u = 0

This is Bessel’s function of order n = 1

Solution is;

u(z ) = AJ1 (z ) + BY1 (z )

z
But y = u
k

y(z ) = AJ1 + BY1 ( z )

1 1
Y ( x) AJ1 (kx 2 ) + BY1 (kx 2 )

Home Work:


(−1) k x 2 k + 6
Using the substitution y = ux solve xy − 5 y + xy = 0 Ans: y = 
3
2 k +3
x 
k =0 2 k!(k + 3)!

40
LAPLACE TRANSFORM OF BESSEL’S FUNCTIONS:

The most natural way to determine the Laplace transform of J 0 (at ) is probably to transform the

infinite series of J 0 (at ) term by term using the formulae

J 0 (at )
Remember
  (− 1)n  at 
2n

J 0 (at ) =     
 n =0 (n!) 2 
2

=
(− 1)n a 2 n  t 2 n 


n = 0 2 (n!)
2n 2


(− 1) a 2 n (2n )!
n
=  2 n 2  2 n +1
n = 0 2 (n!) s
carrying out algebraic manipulati ons
1  (− 1) (2n )! a 2 
n n

=  2 n 2  2 
s n =0 2 (n!)  s 

n
2 n
n!= 2  4  6  8...(2n − 2 )(2n ) =  2k
k =0

(2n )!=  2k   (2k − 1)

n n

 k =1   k =1 

(− 1 ) (2k − 1) a
n
n
n
1 2

=  k =1
n
 2 
s 2 n! s 
n

 2k − 1  a 
n 2

 −  
1 n =0 k =1  2  s 
2

=
s n!
on expansion, the result will be

41
 n
 2k − 1  a 
2
−   2

1  n  2  s  1  1 a2 
1  3  2  1  3 s  2 3
2 2  a  − 2 2 2  a  + ...
 s

= 1−
2 s 2 +
2!  s 2   s2  
s n =o k =1 n!

3!   
The expression above is of the Binomial Expansion form given by

(1 + x )r = 1 + rx + r (r − 1) x 2 + r (r − 1)(r − 2) x 3 + ...
2! 3!
−1
1   a 2 
2

1 +  
s   s 2 
hence
(
J 0 (at ) = s 2 + a 2 )−1
2

numerous other transforms follow readily for instance

since
d
J 0 (at ) = −aJ , (at ), it follows from the theory
dt
that  f ' (t ) = s f (t )− f (0 )
−1 d
that J1 (at ) =  J 0 (at )
a dt
−1
= sJ 0 (at )− J 0 (0)
a
1 s2 
=  − 1
(
a  s 2 + a 2 12 
 ) 
−1  s − s2 + a2   s + s2 + a2 
=   
a  s 2 + a 2   s + s 2 + a 2 
on rationalising.

J1 (at ) =
a

s2 + a2 s + s2 + a2 

42
H0ME WORK:

1) Find the Laplace transform of the following

J 2 (at ) hint J n +1 ( x ) = J − n ( x ) − 2 J n (x ) , x = a
d
dx

2) Given that J 0 (at ) =

1
then
(s + a ) 2 2
1
2

 f (t ) = f (s ) and t f (t ) =
n (− 1) d f (s ) , n n

ds n
show that tJ 0 (at ) =
s
+ a2 2 (s 2
) 3

3) show by mathematical induction that

 
a)  t n J n −1 (at ) =
(2n )!a n−1 s
(
2 n n! s 2 + a 2 )
n+ 1
2

 
b)  t n J n (at ) =
(2n )!a n
(
2n n! s 2 + a 2 )
n+ 1
2

43
 LEGENDRE’S EQUATIONS:

ADRIEN MARIE LEGENDRE (1752-1833), French mathematician, who became a professor

in Paris in 1975 and made important contributions to special functions , elliptic integrals, number
theory and the calculations of variations.
d2y
The D.E (1 − x ) − 2 x + n(n + 1) y = 0 where n is a real number.
2 dy
2
dx dx
Is known as Legendre’s differential equation of order n . It has two singular points only namely
x = 1 and x = −1 . n is a non-integer.
We observe that the origin, like all other points x  1 is an ordinary point of the equation and
there exists at least one maclaurin’s type series solution of the form;

y =  ak x k x  1
k =0

Upon substitution of this solution, along with its 1st and 2nd derivatives in the equation, we get
 

 k (k − 1)a xk
k −2
−  k (k − 1) + 2k − n(n + 1)ak x k = 0
k =0 k =0

Note that the first 2 terms of the first summation for k = 0 and k = 1 are both zeros and we may
write
 

 k (k − 1)ak x k −2 −  k (k − 1) + 2k − n(n + 1)ak x k = 0

k =2 k =0

On collecting coefficients of like powers of x , we may now amalgamate the two summations into
the single summation expression


 (k + 1)(k + 2)a

k =0
k +2 + (n − k )(n + k + 1)ak x k = 0

This is now an identity, which must hold true for any value of x . In particular, it must hold true
for no-zero values of x . For this to happen, the coefficient of x k must be identically zero for all
permissible values of k ; that is
(k + 1)(k + 2)ak +2 + (n − k )(n + k + 1)ak  = o ; k = 0,1,2...

 ak +2 = −
(n − k )(n + k + 1) a ; k = 0,1,2,3,...
(k + 1)(k + 2) k
44
 
From y =  ak x k = a0 + a1 x1 + a2 x 2 + a3 x 3 + a4 x 4 + a5 x 5 + ...
k =0

We attempt to compute the coefficients

a0 and a1 are arbitrary constants

− n(n + 1)a0 − (n − 1)(n − 2)a1

a2 = , a3 =
2! 3!
n(n + 1)(n − 2)(n + 3)a0 (n − 3)(n − 1)(n − 2)(n + 4)a1
a4 = , a5 =
4! 5!
Substituting back into the postulated solution, it yields
y = a0 y1 + a1 y2 = a0 ye + a1 y0 e -Even, 0 -odd.

n(n + 1)x 2 (n − 2)(n + 1)(n + 3)x 4

y1 = ye = 1 − + + ...
2! 4!

y2 = y0 = x −
(n − 1)(n − 2)x 3 + (n − 3)(n − 1)(n − 2)(n + 4)x 5
3! 4!
y1 And y 2 are linearly independent solutions
We may write the series solution in the form
  
y =  ak x k =  a2 k x 2 k +  a2 k +1 x 2 k +1
k =0 k =0 k =0

even  odd 

Where the general two forms a2 k and a2 k +1 represent the coefficient ak + 2

y = un (x )even + un (x )odd

u n ( x )even =  a2 k x 2 k Contains even powers of x only
k =0


u n ( x )odd =  a 2 k +1 x 2 k +1 Contains odd powers of x only
k =0

un (x )even And un (x )odd Are linearly independent since

u n ( x )even
 const
u n ( x )odd

At the same time, we should note that un (x )even and un (x )odd separately are also solutions in their

own rights, since each satisfies the D.E

45
In Engineering applications, we are usually most interested in those infinite series solutions that
converge to some finite value or sum.
If n is not an integer, both y1 and y 2 converge on the interval x  1 and diverge at the

interval x = 1 In this case, y1 and y 2 are known as Legendre functions of the second kind. If n is

a non-negative integer, one of the series ( y1 or y 2 ), terminates i.e. it reduces to a polynomial of

order n and is thereby convergent for all x while the other series converges for x  1 and diverges

at x = 1 .
More precisely if n is even, y1 will reduce to a polynomial of order n in either case, the series
that reduces to a polynomial is known as a Legendre polynomial or Zonal Harmonic of order n
The usual standard form of the Legendre polynomial is obtained by
multiplying y1 and y 2 whichever is by a constant so chosen that the resulting polynomial has a
value of 1 when x = 1 . Conventionally, Legendre’s polynomial of order n is represented as
pn ( x ) by requiring p n (1) = 1

In so doing, the coefficient of the highest power of x in pn (x ) is given by

an =
(2n )!
2 n (n!)
2

Other coefficients are obtained from

− (k + 1)(k + 2)
ak = a , k  n−2
(n − k )(n + k + 1) k + 2
pn (x) = an x n + an−2 x n−2 + an−4 x n−4 + ...

The general coefficient term in pn (x ) is given by; (whether its odd or even)

ak = n
(− 1) (2n − 2k )!
k
k = 0,1,...n
2 k!(n − k )!(n − 2k )!

pn ( x ) = 
N
(− 1)k (2n − 2k )! x n−2k
k = 0 2 k!(n − k )!(n − 2k )!
n

46
 n
 , n = even
Where N =  2
n −1
 , n = odd
 2
By inspection, when
n = 0 , p0 (x ) = 1
n = 1, p1 (x ) = x

n = 2, p2 (x ) =
1 2
2
(
3x − 1 )
1
p 2 (x ) = 
(− 1)k (4 − 2k )! x 2−2k
k =0 2 k!(2 − k )!(2 − 2k )!
2

4! 2!
= x2 − 2
2 (2!)
2 2
2

= (3 x 2 − 1)
1
2

n = 3, p3 ( x ) =
1
2
(
5 x 3 − 3x )
n = 4, p4 ( x ) =
1
8
(
35x 4 − 30x 2 + 3 )
n = 5, p5 (x ) =
1
8
(
63x 5 − 70x 3 + 15x )
Since the argument of Legendre’s polynomials Can be substituted with x = cos , we may write
the polynomials alternatively as
p0 (cos ) = 1

p1 (cos ) = cos

p2 (cos ) =
1
2
(
3 cos2  − 1 )
(
p3 (cos ) = 5 cos3  − 3 cos ;
1
2
)
Special values of Legendre polynomials, which are in fact the analogues of the values of the
circular (trigonometric) function cos evaluated at  =   are the following;
pn (1) = pn (cos ) = 1;

47
pn (− 1) = pn (cos ) = (− 1)
n

Legendre’s polynomials are the only solutions of Legendre’s equation which are bounded on the
interval x  1 with the knowledge that pn (x ) is a solution to Legendre’s equation.

A general solution is available by the method of variation of parameters.

Note:
The other solution of Legendre’s differential equation for a non-negative n is then
either un (x )odd if n is even or un (x )even if n is odd, in either case, the usual notation assigned to it

for any given choice of the arbitrary constants a0 or a1 is Qn (x ) . So that the general solution may

be written in the form

y = An pn (x ) + BnQn (x )

The function Qn (x ) is also sometimes referred to as Legendre function of the second kind.

When variation of parameters is used to obtain the other solution, we obtain;

y = An pn (x ) + Bn pn (x )
dx
( )
x − 1  pn (x )
2 2

This form is of course not explicit and may often not be preferred.

Exercise:
Prove using the same approach as for p2 (x ) , for p3 ( x ) , p4 (x ) , p5 ( x ) , p6 (x ) , p7 ( x ) , p8 ( x ) , …etc

Graphs of Legendre’s Polynomials (See Addendum)

48
Illustration:

Find the general solution of the O.D.E (1 − x 2 )y ' '−2 xy'+6 y = 0 for 0  x  1 . Satisfying the

condition y 1 ( x) = 10
2

The O.D.E in question is a Legendre’s equation of the n = 2 of the

form (1 − x 2 )y ' '−2 xy'+ n(n + 1) y = 0
The solution is a Legendre’s equation of order n = 2

p2 ( x ) =
1
2
(
3x 2 − 1 )
1 1
p2   = − But our initial conditions are such that;
2 8

1
y   = 0 i.e.
2
y = Ap2 (x )
y 1 = 10
2

1
10 = Ax − ; A = −80
8
y = 40 − 120x 2 , x  

SOME IMPORNTANT RESULTS INVOLVING LEGENDRE’S POLYNOMIALS:

1. Recurrence formula;
2n + 1
pn +1 ( x ) = xpn ( x ) − pn −1 ( x )
n
n +1 n +1
2. Rodrigue’s Formula;

pn ( x ) =
1 dn 2
n
2 n! dx n
(x − 1)
n

Use Rodrigue’s formula to find Legendre’s Polynomials of the order n = 3

49
p3 (x ) =
1 d3 2
3
2 3! dx 3
(x − 1)
3

d 2
dx
(3
)2
x −1 = 3 x2 −1  2x ( )
(
= 6x x2 −1 )2

d2
dx 2
 2

6 x(x 2 − 1) = 6 x  2(x 2 − 1)  2 x + (x 2 − 1)  6
1 2

(
) ( )
= 24x 2 x 2 − 1 + 6 x 2 − 1
2

= 6(x − 1)4 x + (x − 1)

2 2 2

6(x − 1)4 x + (x − 1)

3
d
= 3
2 2 2

dx
= 12(x − 1) 2 x + 96x − 48x
2 3

= 24x 3 − 24x + 96x 3 − 48x

f ' ' ' ( x ) = 120x 3 − 72x = 24 5 x 3 − 3 x ( )
p3 ( x ) =
1
48
(
 24 5 x 3 − 3x )

p3 (x ) =
1 3
2
(
5 x − 3x )

ORTHOGONALITY OF LEGENDRE’S POLYNOMIALS:


0 ,m  n
( ) ( )
1
−1 p m x p n x dx = 
 2 ,m = n
 2n + 1

  p (x ) dx = 2n + 1
2
1 2
n Where m and n are non-negative integers
−1

Example:

Evaluate  x 2 pn +1 ( x ) pn −1 ( x )dx , n = 1,2,3,... hence or otherwise, evaluate  x p (x ) p (x )dx

1 1
2
3 1
−1 0

1. Using the recurrence relationship

50
 n +1
xpn ( x ) = pn +1 (x ) + pn −1 (x ) ,
n
2n + 1 2n + 1
Express the integral in question as a sum of terms of the form

An  pm ( x ) pn ( x )dx And Bn   pn (x ) dx
1 1 2

−1 −1

Where An and Bn are constants dependent on 1

2. Invoke the orthogonality property of Legendre’s Polynomials

 0, m  n
−1 pm (x ) pn (x )dx =  2 , m = n
1

 2n + 1

−1  pn (x ) dx =
2
1 2
2n + 1
3. Make the appropriate deductions
(a) Setting n → n + 1the recurrence relationship becomes
n+2 n +1
xpn +1 (x ) = pn + 2 ( x ) + pn ( x )
2n + 3 2n + 3
Setting n → n −1
n −1
xpn −1 ( x ) = pn ( x ) + pn − 2 ( x )
n
2n − 1 2n − 1
 n + 2  n  1 2
( ) ( )   x pn + 2 ( x ) pn ( x )dx
1
−1 =  
2
x p + x p − x dx
 2n + 3  2n − 1  −1
n 1 n 1

 n + 2  n  1 2
+    x pn + 2 ( x ) pn − 2 ( x )dx
 2n + 3  2n − 1  −1
 n + 1  n  1
+   −1  pn ( x ) dx
2

 2 n + 3  2 n − 1 
 n + 1  n  1
+    pn ( x ) pn − 2 ( x )dx But pn −2 (x ) = 0
 2n + 3  2n − 1  −1
 n + 1  n  2n 
  x 2 pn +1 ( x ) pn −1 ( x )dx = 
1
  
−1
 2n + 3  2n − 1  2n + 1 

x 2 pn +1 ( x ) pn −1 (x )dx = 2 x 2 p3 ( x ) p1 ( x )dx
1 1
Hence 
−1 0

51
By comparing n = 2 , our expression evaluated becomes
 2 + 1  2  2 
=   
 4 + 3  2 x 2 − 1  2 x 2 + 1 
3 2 2
=   
7 3 5
=2
35

Method II:
Using Rodrigue’s formula

p1 ( x ) =
1 d 2
2 dx
(
x −1 )
p1 ( x ) = (2 x )
1
2
p1 ( x ) = x

1  d3 2 3
p3 ( x ) = 3  3 (x − 1) 
2 3!  dx 
p3 ( x ) = (5x − 3x )
1 3
2
Therefore
1
x
0
2
p3 ( x ) p1 ( x )dx =  x 2
1

0
1
2
( )
5 x 3 − 3x  xdx

=  x3
1

0
1 3
2
(
5 x − 3x dx )
=
1 1 6
2 0
(
5 x − 3 x 4 dx)
1  5 x 7 3x 5  0
=  − 
2 7 5 1

1  5 3  
=  − −0
2  7 5  

1 4 2
=  =
2 35 35

52
Evaluate the Integral

xpn ( x ) pn −1 ( x )dx
1
−1

Using the recurrence formula;

n +1
xpn ( x ) = pn +1 (x ) + pn −1 (x )
n
2n + 1 2n + 1
n +1
xpn (x ) pn −1 ( x )dx =  pn +1 ( x ) pn −1 (x )dx +  pn −1 ( x ) pn −1 (x )dx
1 1 1 n
Thus −1 −1 2n + 1 −1 2n + 1

−1  pn (x ) dx =
2
But pn +1 (x ) = 0 And
1 2
2n + 1
Setting n → n −1

−1  pn−1 (x ) dx =
2
1 2
2(n − 1) + 1

n
=
2n − 1

n
Multiplying by
2n + 1
n 2
= 
2n + 1 2n − 1
2n
=
4n 2 − 1

xp5 ( x ) p4 ( x )dx
1
Evaluate −1

xpn ( x ) pn −1 ( x )dx =
1 2n
−1 4n 2 − 1
n = 5;

xp5 ( x ) p4 ( x )dx =
1 2 x5 10 10
−1
= =
4 x5 − 1 100 − 1 99
2

53
 x p (x ) p (x )dx = 2 x 99 = 99
1 1 10 5
5 4
0

Find the solution of (1 − x 2 )y ' '−2 xy'+12 y = 0 , which is bound for −1  x  1 and satisfies the

1
condition y   = 47
4
Comparing the O.D.E with Legendre’s equation
Implying the O.D.E is Legendre’s equation n = 3 . The solution of Legendre’s equation bound for
−1  x  1 is pn (x ) . Substituting n = 3 in Rodrigue’s formula gives;

1 d3 2
p3 (x ) = 3
2 3! dx 3
(x − 1) Hence p3 (x ) = (5 x 3 − 3x )
3 1
2

1 1 1  1 
3

; Since p4 (x )  47 ,
43
p3   =  5  − 3   = −
 4  2   4   4   128

1 1 43
y  = ap3   Implying that 47 = −
4 4 128
a  −140 And hence the solution of the O.D.E in question that is bound −1  x  1 and satisfies

the condition y   = 47 is y(x ) = 70(3x − 5 x 3 )

1
4

Evaluate  x 2 p2 n −1 ( x ) p2 n −1 (x )dx, n = 1,2,3, and Hence Evaluate  x p (x ) p (x )dx

1 1
2
3 1
−1 0

From the recurrence relation

(n + 1) pn+1 (x ) = (2n + 1)xpn+1 (x ) + npn−1 (x )
We can write
n +1
xpn ( x ) = pn +1 (x ) + pn −1 (x )
n
2n + 1 2n + 1
Setting n → 2n + 1 yields
2n + 2 2n + 1
xp2 n +1 ( x ) = p2 n + 2 ( x ) + p2 n ( x )
4n + 3 4n + 3
Setting n → 2n −1yields

54
 2n − 1
xp2 n −1 ( x ) = p2 n ( x ) + p2 n − 2 ( x )
2n
4n − 1 4n − 1
Multiplying xp2 n +1 (x ) with xp2 n −1 (x ) , and integrating the resultant product with respect

to x over −1  x  1 yields
1  2n + 2  2n 
( ) ( ) −1  4n + 3  4n − 1  p2n+ 2 (x ) p2n (x )dx
1
−1 =
2
x p 2 n +1 x p 2 n −1 x dx

1  2 n + 2  2 n − 1 
+ 
−1 4n + 3
  p2 n + 2 ( x ) p2 n − 2 ( x )dx
  4n − 1 
1  2 n + 1  2 n 
+ 
−1 4 n + 3
  p2 n ( x ) dx
2

  4n − 1 
1  2 n + 1  2 n − 1 
+ 
−1 4 n + 3
  p2 n ( x ) p2 n − 2 ( x )dx
  4n − 1 
The 1st, 2nd and 4th integrals on the on the right hand side of the above expression are zero by
virtue of the orthogonality property of the Legendre’s polynomials.

pm (x ) pn (x )dx = 0; m  n
1
−1

x 2 p2 n +1 ( x ) p2 n −1 ( x )dx
1
We can thus write −1

 2n + 1  2n  1
=   −1  p2 n ( x ) dx
2

 4 n + 3  4 n − 1 
It can be deduced that

−1  p2n (x ) dx = By setting n → 2n in   pn ( x ) dx =

2
1 2 2 1 2
4n + 1 −1 2n + 1
 2n + 1  2n  2 
x 2 p2 n +1 ( x ) p2 n −1 ( x )dx = 
1
−1
  
 4n + 3  4n − 1  4n + 1 
4n(2n + 1)
=
(
(4n + 3) 16n 2 − 1 )
Since  p2 n (x ) is an even function
2

2n(2n + 1)
p2 n +1 (x ) p2 n −1 (x )dx =
1
x
2
0 (4n + 3)(16n 2 − 1)
Hence substituting n = 1 into the above expression, we have

55
 2(2 + 1)
 x p (x ) p (x )dx = (4 + 3)(16 − 1)
1
2
3 1
0

= 0.057(3 SF calc.)
2
=
35

QUESTIONS:
a) Show that if x is replaced by x where  is a constant, Bessel’s
equation x 2 y ' '+ xy'+(x 2 − n 2 )y = 0 , becomes x 2 y ' '+ xy'+(2 x 2 − n 2 )y = 0 And hence or otherwise ,

solve the O.D.E 4 x 2 y ' '+4 xy'+(22 x 2 − 1)y = 0

b) Evaluate the following integrals

 x J (x)dx
3
i. 2

ii. x
−4
J 5 (x )dx
1
iii. x 2
J 1 (x )dx
2

c) Find a solution of (1 − x 2 )y ' '+2 xy'+6 y = 0 which is bound for −1  x  1 And satisfies the

1
condition y   = 10
2
d) Use Rodrigue’s formula for LPs to find the Legendre polynomials of order n = 5
e) Evaluate the integral

p2 n +1 ( x ) p2 n −1 ( x )dx n = 1,2,3,...
1
x
2
−1

1
f) Find the Laplace Transform of the function f (t ) = t 2
g) The figure below shows a spring-mass system. The displacement of the masses from their
equilibrium positions are x1 and x2 . A periodic force sin t is applied to the 4 kg at time t = 0 ; when

56
the system is at rest in equilibrium. The governing equations of the ensuing motion
d 2 x1 d 2 x2
are 4 x 2 + 4 x1 − 2 x 2 = sin t and − 2 x1 + 2 + 3x2 = 0 with initial conditions
dt 2 dt 2
• •
x1 (0 ) = x1 (0 ) = x2 (0 ) = x2 (0 ) = 0 Find x1 (t ) and x2 (t )

57