Professional Documents
Culture Documents
Unit 4 Bessel Functions
Unit 4 Bessel Functions
4.1 INTRODUCTION
In Unit 3, we studied Legendre, Hermite and Laguerre polynomials and discussed a
number of their special properties. In this unit we shall study about Bessel functions.
Bessel functions were first discovered in 1732 by D. Bernoulli (1700-1782), who
provided a series solution (representing a Bessel function) for the oscillatory
displacement of a heavy hanging chain. In 1764, Euler developed a series similar to
that of Bernoulli, in connection with his work on the vibration of a circular drumhead.
J.Fourier (1768-1836) also used Bessel functions in his classical treatise on heat in
1822, but it was Bessel the German astronomer who derived the differential equation
bearing his name and studied the general properties of its solutions (now called Bessel
functions) in his 1824 memoir. These functions occur so frequently in practice that
they are undoubtedly the most important functions beyond the elementary ones.
We shall start the discussion in Sec.4.2 with the Bessel’s differential equation and
obtain its power series solution in terms of Bessel functions of first and second kind.
We shall also discuss various properties of Bessel functions here. In Vibrational
phenomena, one of the fundamental problems is to describe the wave motion of a
natural or mechanical system: such as a vibrating membrane, subject to certain initial
and boundary conditions. In Sec.4.4 we shall discuss this problem of vibrating
membrane and illustrate the role played by Bessel functions in getting its solution. We
shall also discuss here the problem concerning the buckling of vertical columns under
a compressive load.
Objectives
After studying this unit you should be able to
• identify Bessel’s differential equation;
• obtain the series solutions of Bessel’s differential equation;
• derive Rodrigue’s formula, generating function, recurrence relations and
orthogonal property of Bessel functions;
• use recurrence relations and other properties of Bessel functions occurring in
equations governing physical phenomenon, for finding their solutions.
∑
m =0
(m + k ) (m + k − 1) c m x m + k + ∑
m=0
(m + k ) c m x m + k + ∑
m =0
c m x m+k +2 − υ 2 ∑c
m =0
mx
m+k
=0
∞ ∞
or, ∑ [(m + k) (m + k − 1) + (m + k) − υ ]c
m=0
2
mx
m+ k
+ ∑c
m =0
mx
m+k +2
=0
∞ ∞
or, ∑ [(m + k + υ) (m + k − υ)]c m x m+ k + ∑ c m x m +k +2 = 0
m =0 m=0
∞ ∞
or, ∑ [(m + k)
m=0
2
]
− υ 2 c m x m+ k + ∑c
m=0
mx
m+k +2
=0 (3)
Equating to zero the coefficient of smallest power of x (i.e. x k ) to zero, we get the
indicial equation as
c 0 ( k 2 − υ 2 ) = 0 ⇒ k = υ, − υ (Q c 0 ≠ 0) (4)
Equating the coefficient of x k +1 to zero in Eqn.(3), we obtain
[(k + 1) 2 − υ 2 ) c1 = 0 , or c1 = 0, for k = ± υ
To obtain the recurrence relation, we equate to zero the coefficients of x k + m in
Eqn.(3). This yields
( m + k + υ) ( m + k − υ) c m = − c m − 2
−1
i.e, cm = c m −2 (5)
(m + k + υ) (m + k − υ)
For k = υ , relation (5) simplifies to
−1
cm = c m−2 (6)
( m − 2 υ) m
Since c1 = 0 , putting m = 3, 5, 7, K in relation (6), we get
c3 = c5 = c7 = K = 0 (7)
and for m = 2, 4, 6, K , we obtain
−c −c c0
c2 = 2 0 , c4 = 3 2 = 4
2 (1 + υ) 2 (2 + υ ) 2 (2 !) (1 + υ) (2 + υ)
− c4 − c0
c6 = 2 = 6 L
2 (3) (3 + υ) 2 (3! ) (1 + υ) (2 + υ) (3 + υ) 3
In general, we can write
(−1) m c 0
c 2m = , m = 1, 2, K (8)
2 2 m (m !) (1 + υ) (2 + υ) L (m + υ)
122
Bessel Functions
Putting these values of c' s in Eqn.(2), we obtain one of the linearly independent
solution of the Bessel’s equation as
x2 x4
y1 ( x ) = c 0 x υ 1 − 2 + 4 −L
2 (1 + υ) 2 (2! ) (1 + υ) (2 + υ)
(−1) m x 2 m
L+ m + L (9)
2 (m !) (1 + υ) (2 + υ) K (m + υ)
∫
Γ ( υ) = e − t t υ−1 dt
0
(11)
This y( x ) defines the Bessel function of the first kind of order υ and is denoted by
J υ ( x ) . Therefore we have one solution of Eqn.(1) as
∞ 2m+ υ
(−1) m x
J υ (x) = ∑
m =0
m!Γ (m + υ + 1) 2
υ ∞ 2m
x (−1) m x
=
2
∑
m =0
m!Γ ( m + υ + 1) 2
(14)
To obtain a second linearly independent solution, we turn to the other indicial root,
k = − υ . There is no need to repeat all the steps; all we need to do is to change
υ to − υ everywhere on the right side of Eqn.(14). Denoting the result as J − υ ( x ) , we
have the second solution of Eqn.(1), the Bessel function of the first kind of order
− υ as
∞−υ 2m
x (−1) m x
J − υ (x ) = ∑
2 m =0 m!Γ (m − υ + 1) 2
(15)
Using ratio test it can be shown that both the series (14) and (15) converge for all x .
The leading terms of the series in (14) and (15) are constant times x υ and x − υ ,
respectively so neither of the solutions J υ ( x ) and J − υ ( x ) is a scalar multiple of the
other. Thus they are linearly independent, and we conclude that, when υ is not an
integer
y(x) = A J υ (x) + B J − υ (x) (16)
is the general solution of Bessel’s Eqn.(1).
Writing Eqns.(14) and (15) in the form
1 1
J υ (x) = x υ υ
− υ+ 2
x 2 + L (17)
Γ ( υ + 1) 2 Γ ( υ + 2) 2
123
Ordinary Differential 1 1
Equations J − υ (x ) = x −υ −υ
− −υ+2
x 2 + L (18)
Γ (1 − υ) 2 Γ ( 2 − υ) 2
you may observe that the power series within the square brackets tend to
1 Γ( υ + 1) 2 υ and 1 Γ (1 − υ) 2 − υ , respectively, as x → 0 . We see that
[ ] [ ]
J υ ( x ) ~ 1 Γ ( υ + 1) 2 υ x υ and J − υ ( x ) ~ 1 Γ (1 − υ) 2 − υ x − υ as x → 0 . In more concise
υ −υ
form we can write J υ ( x ) = O( x ) and J − υ ( x ) = O( x ) as x → 0 . Thus, J υ ( x ) ′s
tend to zero and the J − υ ( x ) ′s tend to infinity as x → 0 . In Fig.1, we have plotted
1.5
J (x) J 1 / 2 ( x ) and J −1 / 2 ( x ) , for υ = 1/2 .
1 −1 / 2
J1 / 2 (x ) 1 3 5
0 .5 The cases where υ is half-integral i.e. υ = ± , ± , ± , L , are of special interest
2 2 2
0
because these Bessel functions are actually elementary functions. In particular, the
− 0.5 1
case υ = leads to the interesting results where
0 5 10 15 2
x
2 2
Fig.1 J1 / 2 (x) = sin x and J −1 / 2 ( x ) = cos x
πx πx
***
Example 2: Transform the equation
x 2 y ′′ + xy ′ + 4( x 4 − υ 2 ) y = 0
using the substitution x 2 = z and hence find the general solution of the equation.
dy dy dz dy dy
y′ = = . = 2x =2 z
dx dz dx dz dz
2 2
d y dy d y dy d2y
and y ′′ = 2 = 2 + 4x 2 2 = 2 + 4z 2
dx dz dz dz dz
On substituting y ′ and y ′′ from above, the given equation reduces to
d2y dy dy
z 4z 2 + 2 + 2z + 4 (z 2 − υ 2 ) y = 0
dz dz dz
2
d y dy
or, z2 2 + z + (z 2 − υ 2 ) y = 0
dz dz
which is a Bessel’s differential equation of order υ . Hence the general solution is
given by
y( z ) = A J υ ( z ) + B J − υ ( z )
or, y( x ) = A J υ ( x 2 ) + B J − υ ( x 2 )
***
124
Bessel Functions
You may now try the following exercises.
1
E1) Using Eqns.(14) and (15) and the relations Γ = π , show that
2
2 2
J1 / 2 (x) = sin x , J −1 / 2 ( x ) = cos x .
πx πx
E2) Show that the given functions are solutions of the corresponding equations
2kx 3 / 2
a) y ′′ + k 2 xy = 0 , y = x J 1 / 3
3
k x2
b) y ′′ + k 2 x 2 y = 0 , y = x J 1 / 4
2
kx
3
c) y ′′ + k 2 x 4 y = 0 , y = x J 1 / 6
3
E3) Using the substitutions given alongside, transform the following differential
equations and hence, find their general solution.
a) 4 x 2 y ′′ + 4 x y ′ + ( x − υ 2 ) y = 0; ( x = z)
b) xy ′′ − y ′ + xy = 0; ( y = xu )
c) xy ′′ + (1 + 2k ) y ′ + xy = 0; (y = u / x k )
where we have used the fact that Γ(m − n + 1) is undefined when its argument is zero
or a negative integer, namely, for m = 0, 1, K , n − 1 . To start the series at zero once
again, we make the change of index m − n = k in Eqn.(21), which yields
2k + n 2k+n
∞
(−1) n + k x
∞
(−1) n + k x
J −n (x ) = ∑
k =0
(n + k )!Γ(k + 1) 2
= ∑
k =0
k!(n + k ) ! 2
(22)
If (−1) n is factored out the series that remains on the right hand side of En.(22) is the
same as that given in Eqn.(20), so that
J − n ( x ) = (−1) n J n ( x ) (23)
This shows that J n ( x ) and J − n ( x ) are linearly dependent. Thus, whereas J υ ( x ) and
J − υ ( x ) are linearly independent and give the general solution (16) of Eqn.(1) when υ
is not an integer, we have only one linearly independent solutions so far for the case
125
Ordinary Differential where υ = integer, namely, y 1 ( x ) = J n ( x ) given by Eqn.(20). The second linearly
Equations
independent solution is yet to be determined which we shall take up later in Sec.4.2.2.
The Bessel functions that arise most frequently in practice are J 0 (x) and J 1 (x) ,
whose series representations are
∞ 2m
(−1) m x
J 0 (x) = ∑
m =0
( m !) 2 2
x2 x4 x6
= 1− + − +L (24)
22 2 4 (2 !) 2 2 6 (3 ! ) 2
∞ 2 m +1
(−1) m x
and J1 (x) = ∑
m =0
m !(m + 1)! 2
x x3 x5 x7
= − 3 + 5 − 7 +L (25)
2 2 1! 2! 2 2 !3! 2 3 ! 4!
At x = 0 , we see that J 0 (0) = 1 while J 1 (0) = 0 and from Eqn.(20) it follows that
J n (0) = 0, n = 2, 3, 4, K . For x near zero, the m = 0 term of Eqn.(20) yields the
asymptotic formulas
J 0 ( x ) ~ 1 as x → 0 + (26)
( x / 2) n
J n (x) ~ as x → 0 + , n = 1, 2, 3, K (27)
n!
The graphs of J 0 ( x ), J 1 ( x ) are shown in Fig.2. Observe that these functions exhibit
an oscillatory behaviour somewhat like that of the sinusoidal functions, except that the
amplitude (maximum departure from the x-axis) of the Bessel function diminishes as
x increases and the (infinitely many) zeros of these functions are not evenly spaced.
The location of these zeros is of great theoretical and practical importance, but we
shall not be going into the details of the theory here. However, the first five zeros of
J 0 ( x ) and J 1 ( x ) are given below:
J n (x)
J0
1
J1
5
0
1 2 9 x
−1
Fig.2
You may now try the following exercises.
E4) Show that J n ( x ) is an even function for n even and an odd function for n odd
where n is an integer.
∫x
υ
J υ−1 (x) dx = x υ J υ (x) + c (29)
2 k + υ +1
∞
(−1) k +1 x
= x −υ ∑
k =0
k ! Γ ( k + υ + 2) 2
−υ
= − x J υ +1 ( x )
d −υ
Therefore, [ x J υ (x) ] = − x − υ J υ + 1 (x) (30)
dx
Integrating Eqn.(30), we obtain
∫x
−υ
J υ+1 (x) dx = −x − υ J υ (x) + c (31)
If we carry out the differentiation on the left-hand sides of Eqns.(28) and (30) and
divide the results by the factors x υ −1 and x − υ−1 , respectively, we deduce that
x J ′υ (x) + υ J υ (x) = x J υ −1 (x) (32)
and x J ′υ (x) − υ J υ (x) = −x J υ+1 (x) (33)
When υ is an integer n , we write
x J ′n (x) = x J n −1 (x) − n J n (x) (34)
and x J ′n (x) = n J n (x) − x J n +1 (x) (35)
Finally, the sum and difference of relations (32) and (33) yields, respectively
127
Ordinary Differential 2 J ′υ (x) = J υ −1 (x) − J υ +1 (x) (36)
Equations
and 2 υJ υ (x) = x J υ−1 (x) + x J υ+1 (x) (37)
Let us look at the following examples.
8 6 24
Example 3: Show that J 4 ( x ) = 2 − 1 J 1 − 2 − 1 J 0 .
x x x
Solution: Using recurrence relation (37), we obtain
2n
J n +1 ( x ) = J n ( x ) − J n −1 ( x )
x
6
∴ J 4 (x ) = J 3 (x) − J 2 (x)
x
6 8 4 2
= 2 − 1 J 1 − J 0 − J 1 + J 0
x x x x
8 6 24
= 2 − 1 J 1 − 2 − 1 J 0
xx x
***
Example 4: Show that
a) [
[ x J n ( x ) J n +1 ( x )]′ = x J 2n ( x ) − J 2n +1 ( x ) ]
1
b) J 1′′ ( x ) = J 2 ( x ) − J 1 ( x )
x
2 1 3
c) J −5 / 2 ( x ) = 2
(3 − x 2 ) cos x + sin x
πx x x
Solution:
a) [ x J ( x ) J ( x )]′ = [ {x − n J ( x ) } {x n +1 J ( x ) }]′
n n +1 n n +1
−n n +1
= x J n (x ) [x J n ( x ) ] + x n +1 J n +1 [− x − n J n +1 ( x ) ]
[using relations (28) and (30)]
=x J 2n ( x )
−x (x )J 2n +1
b) Using relation (36) with υ = 1 , we get
2J 1′ ( x ) = J 0 ( x ) − J 2 ( x )
Differentiating it we obtain
2 J 1′′ ( x ) = J ′0 ( x ) − J ′2 ( x )
Using relation (32) with υ = 2 , and the fact that J ′0 ( x ) = −J 1 ( x ) , we get
2
2 J 1′′ ( x ) = −J 1 ( x ) − J 1 ( x ) − J 2 ( x )
x
1
or, J 1′′ ( x ) = J 2 ( x ) − J 1 ( x ) .
x
c) Using relation (37) for υ = − 3/2 and υ = − 1/2 , we obtain
− 3 J −3 / 2 ( x ) = x [ J −5 / 2 ( x ) + J −1 / 2 ( x ) ]
and − J −1 / 2 ( x ) = x [ J −3 / 2 ( x ) + J 1 / 2 ( x ) ]
Eliminating J −3 / 2 ( x ) from the above two equations we obtain
x 2 J −5 / 2 ( x ) = 3 J −1 / 2 ( x ) − x 2 J −1 / 2 ( x ) + 3x J 1 / 2 ( x )
2 2
Using J 1 / 2 ( x ) = sin x and J 1 / 2 ( x ) = cos x , in the above
πx πx
equation we have,
128
Bessel Functions
2 1 3
J −5 / 2 ( x ) = 2
(3 − x 2 ) cos x + sin x
πx x x
***
Example 5: Prove the following
1
a) ∫
J 0 ( x ) J 1 ( x ) dx = − J 02 ( x ) + c
2
2
x
b) ∫
x J 02 ( x ) dx =
2
[J 02 ( x ) + J 12 ( x ) ] + c
2
c) ∫
J 3 ( x ) dx = − J 2 ( x ) − J 1 ( x ) + c
x
Solution: a) We know that
[J 02 ( x ) ]′ = 2 J 0 ( x ) J ′0 ( x ) = −2J 0 ( x ) J 1 ( x ) [Q J′0 ( x ) = − J 1 ( x ) ]
−1 2
∴ ∫J 0 ( x ) J 1 ( x ) dx =
2
J 0 (x ) + c
x2
b) ∫ x J 02 ( x ) dx =
2 ∫
J 02 ( x ) + x 2 J 0 ( x ) J 1 ( x ) dx [Q J′0 ( x ) = − J1 ( x ) ]
x2
=
2 ∫
J 02 ( x ) + ( x J 0 ( x )) ( x J 1 ( x ) ) dx
x2
=
2 ∫
J 02 ( x ) + ( x J 1 ( x ) ) 2 − ( x J 1 ( x )) ( x J 1 ( x ) ) ′ dx [ using relation (29)]
x2 (x J1 (x) ) 2
= J 02 ( x ) + ( x J 1 ( x )) 2 − +c
2 2
x2
= [ J 02 ( x ) + J 12 ( x ) ] + c
2
c) ∫J 3 ( x ) dx ∫
= x 2 ( x −2 J 3 ( x ) ) dx
∫
= x 2 (− x −2 J 2 ( x ) ) + x −2 J 2 ( x ) 2x dx [ using relation (31)]
−1
= − J 2 ( x ) + 2( − x J 1 ( x ) ) + c [ using relation (31)]
2
= −J 2 ( x ) − J1 (x ) + c
x
***
∫x
3
Example 6: Evaluate J 3 ( x ) dx .
∫x ∫
J 3 ( x ) dx = x 5 [ x −2 J 3 ( x ) ] dx
3
Solution:
∫
= x 5 [− x −2 J 2 ( x )] − [ − x −2 J 2 ( x )] 5x 4 dx [using relation (31)]
∫
= − x 3 J 2 ( x ) + 5 x 2 J 2 ( x ) dx
∫x ∫
2 3 −1
Now J 2 ( x ) dx = x [ x J 2 ( x ) ] dx
∫
= x 3 [ − x −1 J 1 ( x ) ] − [− x −1 J 1 ( x ) ] 3x 2 dx
∫
2
= − x J 1 ( x ) + 3 x J 1 ( x ) dx
∫ x J (x ) dx = −∫ x J ′ (x) dx = −[x J
1 0 0 (x) ∫
− J 0 ( x ) dx ]
= − x J0 (x ) + ∫ J 0 ( x ) dx
∫x ∫
3
Then J 3 ( x ) dx = − x 3 J 2 ( x ) − 5x 2 J 1 ( x ) − 15x J 0 ( x ) + 15 J 0 ( x ) dx
129
Ordinary Differential
Equations
The integral ∫J 0 ( x ) dx cannot be obtained in closed form.
***
You may now try the following exercises.
b) ∫J 2 ( x ) dx ∫
= −2J 1 ( x ) + J 0 ( x ) dx + c
4 8
c) ∫J 5 ( x ) dx = −J 4 ( x ) −
3
J 3 (x) − 2 J 2 (x) + c
x
∫x
3
d) J 0 ( x ) dx = x ( x 2 − 4) J 1 ( x ) + 2 x 2 J 0 ( x ) + c
∫x
5
e) J 0 ( x ) dx = x 5 J 1 ( x ) − 4 x 4 J 2 ( x ) + 8x 3 J 3 ( x ) + c
a
a
f) ∫xJ
0
0 ( r , x ) dx =
r
J 1 (ar), a , r are constant .
∫ (x − x
3
g) )J 0 ( x ) dx = 4 J 1 (1) − 2 J 0 (1)
0
We shall now show that the Bessel function of the first kind, J n ( x ) can also be
obtained using a generating function.
Generating Function
We shall show that
1 1 ∞
x t−
e 2 t
= ∑J
n = −∞
n (x) t
n
, t≠0. (38)
The function on the left hand side is called the generating function of the Bessel
function J n ( x ).
We have
x 1
t− ∞ ( xt / 2) r ∞ (− x / 2 t ) m
e 2 t
= e xt/2 e − x/2t =
r =0
∑
r ! m =0 m!
∑
130
∞ ∞ Bessel Functions
(−1) m ( x / 2) r + m t r − m
=
r = 0 m =0
∑∑ r! m!
(39)
Our goal is to obtain a single series in powers of t . Thus we make the change of
index n = r − m , and because both r and m have an infinite range of values, it
follows that n varies from − ∞ to ∞ . Then the sum on the right hand side of Eqn.(39)
becomes
∞ ∞
(−1) m ( x / 2) n + 2 m t n ∞
∞ (−1) m ( x / 2) n + 2 m n
∑∑
n = −∞ m = 0 ( n + m) ! m!
=
n = −∞ m = 0
∑ ∑
m!(n + m)!
t
∞
= ∑
n = −∞
J n (x) t n .
iθ
and set t = e . Then
x i θ −i θ ) ∞ ∞
(e − e
e2 = e i x sin θ = ∑
n = −∞
J n ( x )e i n θ = ∑
n = −∞
J n ( x ) [cos nθ + i sin nθ]
0,m ≠ n
π
where we have used the result cos m θ cos nθ dθ = π
0
∫
2 , m = n
.
Similarly, multiplying both sides of Eqn.(42) by sin nθ , integrating over the interval
π 0,m ≠ n
0
∫
[ 0, π ] and using sin mθ sin nθ dθ = π
2 , m = n
, we obtain
π
1 J ( x ), n is odd
π ∫
sin ( x sin θ) sin nθ dθ = n
0 , n is even
(44)
0
Adding Eqns.(43) and (44), we obtain
131
Ordinary Differential π
1
Equations
π0 ∫
cos nθ cos ( x sin θ) + sin nθ sin ( x sin θ) dθ = J n ( x )
π
1
or,
π0 ∫
cos(n θ − x sin θ ) dθ = J n (x) (45)
***
Example 8: Using the generating function for J n ( x ) , prove that
2n
J n −1 ( x ) + J n +1 ( x ) = J n ( x ) , for integer values of n .
x
Solution: We know that
x 1 ∞
t−
e 2 t
= ∑
n = −∞
J n (x) t n .
x 1 ∞ ∞
or, 1 + 2
2 t n = −∞
∑
J n (x) t n =
n = −∞
∑
n J n ( x ) t n −1
∞ ∞
x 1
i.e., ∑
n = −∞ 2 t
n
1 + 2 J n ( x ) t =
n = −∞
∑
n J n ( x ) t n −1
∞ ∞ ∞
x x
or, ∑
n = −∞ 2
J n (x) t n +
n = −∞ 2
∑
J n + 2 (x ) t n =
n = −∞
∑
(n + 1) J n +1 ( x ) t n
∞ ∞
x x n
i.e. ∑
n = −∞
2
J n ( x ) +
2
J n + 2 ( x )
t =
n = −∞
∑
(n + 1) J n +1 ( x ) t n
***
π
1
Example 10: Show that J 0 ( x ) =
π0∫cos ( x cos θ) dθ
Solution: Using the result obtained in Example 9 and integrating Eqn.(50) over the
interval [ 0, π ] , we obtain
π π
∫
0
∫
cos ( x cos θ) dθ = J 0 ( x ) dθ = π J 0 ( x ) .
0
since all other terms vanish.
***
It is often necessary in mathematical physics to expand a given function in terms of
Bessel functions, where the particular type of expansion depends on the problem at
hand. The simplest and most useful expansions of this kind are series of the form
∞
f (x) = ∑a
n =1
n J υ (λ n x ) = a 1 J υ (λ 1 x ) + a 2 J υ ( λ 2 x ) + L (52)
where f ( x ) is defined on the interval 0 ≤ x ≤ 1 , and the λ n are the positive zeros of
some fixed Bessel function J υ ( x ) with υ ≥ 0 . We have chosen the interval 0 ≤ x ≤ 1
only for the sake of simplicity. By a simple change of variable a function defined on
an interval of the form 0 ≤ x ≤ R , can be considered. Such expansions play a major
role in physical problems. We shall illustrate the use of expansion (52) in solving the
two-dimensional wave equation for a vibrating circular membrane in Sec.4.3.
Determination of the coefficients in Eqn.(52) depend on the orthogonality property of
the Bessel functions J υ (λ n x ) which we shall discuss now.
Orthogonality of Bessel Functions
In λ n and λ m are the positive zeros of Bessel functions J υ ( x ) with υ ≥ 0 then we shall
show that
133
Ordinary Differential 1 0 , if m ≠ n
∫
Equations
x J υ (λ m x ) J υ (λ n x ) dx = 1 2 (53)
J υ +1 (λ n ), if m = n
0 2
Formula (53) shows that the functions J υ (λ n x ) are orthogonal with respect to the
weight functions x on the interval 0 ≤ x ≤ 1 .
∫x J
0
υ (λ m x ) J υ (λ n x ) = 0 (58)
which is the first part of (53). Now evaluation of integral (58) has to be done when
m=n.
Multiplying Eqn.(54) by 2 x 2 u ′ , it becomes
2 x 2 u ′ u ′′ + 2 x u ′ 2 + 2a 2 x 2 u u ′ − 2υ 2 u u ′ = 0
d d 2 2 2 d 2 2
or, (x 2 u ′ 2 ) + ( a x u ) − 2a 2 x u 2 − (υ u ) = 0
dx dx dx
∫
2 a 2 x u 2 dx = [ x 2 u ′ 2 + (a 2 x 2 − υ 2 ) u 2 ]10
0
(59)
134
1 Bessel Functions
1 1
∫x
0
J 2υ (λ n x ) dx = J ′υ2 (λ n ) = J 2υ+1 (λ n )
2 2
(60)
υ
where we have used the relation J ′υ ( x ) −
J υ ( x ) = J υ +1 ( x ) in Eqn.(60).
x
Thus formula (53) is completely established.
As an illustration of this formula let us see how it is used in determining the
coefficients of the series (52).
If an expansion of the form (52) is assumed to be possible, then multiplying Eqn.(52)
throughout by x J υ (λ m x ) , integrating term by term from 0 to 1, and using formula
(53), we obtain
1
a
0
∫
x f ( x ) J υ (λ m x ) dx = m J 2υ +1 (λ m )
2
and on replacing m by n we obtain the following formula for a n :
1
2
an = 2
J υ +1 (λ n ) ∫ x f (x) J
0
υ (λ n x ) dx (61)
The series (52), with its coefficients calculated by (61), is called the Bessel series or
sometimes the Fourier-Bessel series of the function f ( x ) .
As an illustration consider the following example.
Example 11: Compute the Bessel series of the function f ( x ) = 1 , for the interval
0 ≤ x ≤ 1 , in terms of the functions J υ (λ n x ) , where λ n are the positive zeros of
J 0 (x) .
Solution: In this case Eqn.(61), yields
1
2
an = 2
J 1 (λ n ) ∫
0
x J 0 (λ n x ) dx
135
Ordinary Differential 1
Equations 1 , 0 ≤ x ≤ 2
1 1
f (x) = , x=
2 2
0 , 1 < x ≤ 1
2
show that
∞
J 1 ( λ n / 2)
f (x) = ∑ 2
n =1 λ n J 1 (λ n )
J 0 (λ n x )
E13) If f ( x ) = x υ for the interval 0 ≤ x < 1 , show that its Bessel series in the
functions J υ (λ n x ) , where the λ n are the positive zeros of J υ ( x ) , is
∞
2
xυ = ∑
n =1 λ n J υ +1 (λ n )
J υ (λ n x ) .
We have shown in Sec.4.2 that when υ is not an integer then J υ ( x ) and J − υ ( x ) are
linearly independent solutions of Eqn.(1) and the complete solution is given by
y( x ) = A J υ ( x ) + B J − υ ( x ) .
However, when υ = n is an integer, J n ( x ) and J − n ( x ) are linearly dependent since
J − n ( x ) = (−1) n J n ( x ) . We therefore have only one independent solution. In order to
find the general solution, we now obtain the second linearly independent solution of
the Bessel’s equation.
4.2.2 Bessel Function of the Second Kind
For obtaining the second solution of the Bessel’s equation for the case when υ = n is
an integer, we once again make use of the power series method discussed in Unit 2.
Let us illustrate the method for the case n = 0 , for Bessel’s equations of order zero
x y ′′ + y ′ + xy = 0 (62)
∞
If we substitute the series solution y( x ) = ∑c
m=0
mx
m+k
in Eqn.(62), we have the case of
repeated indicial roots k = 0 . If you recall Case 3, discussed in Sec.2.4.2, Unit 2, then
you know that two linearly independent solutions are given by [ y ( x ) ] k = 0 and
[∂y / ∂k ] k =0 . Accordingly, if y1 ( x ) = [ y ( x ) ] k =0 , then second solution is of the form
y 2 ( x ) = y1 ( x ) ln x + [ c1 x + c 2 x 2 + L] , x > 0
= J 0 ( x ) ln x + [c1 x + c 2 x 2 + L ]
since y1 ( x ) = J 0 ( x ) .
Further, we have
1
y ′2 ( x ) = J ′0 ln x + J 0 + [ c 1 + 2c 2 x + 3c 3 x 2 + L ]
x
2 1
y ′2′ ( x ) = J ′0′ ln x + J ′0 − 2 J 0 + [ 2c 2 + 6c 3 x + L ]
x x
Substituting the above values of y 2 ( x ), y ′2 ( x ) and y ′2′ ( x ) in Eqn.(62), we obtain
[ x J ′0′ + J ′0 + x J 0 ] ln x + 2J ′0 + [ c1 + 4c 2 x + (c1 + 9c 3 ) x 2 + L
+ [c m −1 + (m + 1) 2 c m +1 ] x m + L] = 0 (63)
in Eqn.(63), we obtain
x3 x5
− x + 2
− 4
+ L + [c 1 + 4c 2 x + (c 1 + 9c 3 ) x 2 + (c 2 + 16c 4 ) x 4 + L] = 0
(2! )2 (3! ) (2! )2
Equating the coefficients of even power of x to zero, we obtain
c1 = 0, c 1 + 9c 3 = 0 ⇒ c 3 = 0, K , c 2 m −1 + (2m + 1) 2 c 2 m +1 = 0, m = 1, 2, K
Therefore, c1 = 0 = c 3 = c 5 = L
For m = 2 , we have
−1
+ c 4 + 36 c 6 = 0
(3! ) (2! ) 2 4
1 1 1 1 1 1 1
or, c6 = + 2 2 1 + = 2 2 2 1 + 2 + 3 , K
62 (3!) (2!) 2
4
2 .4 2 2 .4 .6
and is called the Neumann function of order zero. However, any other linear
combination of y 1 ( x ) and y 2 ( x ) can also be taken as the second linearly independent
solution. In practice, it is convenient and standard to use
2
y *2 ( x ) = [ y 2 ( x ) + ( γ − ln 2) J 0 ( x ) ] (64)
π
where γ is the Euler constant ( γ = 0.5772157) , as the second linearly independent
solution. This solution is called the Bessel function of the second kind of order
zero. It is denoted by Y0 ( x ) .
Hence,
2 x x
2
1 x4 1 1 x6
Y0 ( x ) = J 0 ( x ) ln + γ + 2 − 1 + 4 2
+ 1 + + 6 2
− L (65)
π 2 2 2 2 (2!) 2 3 2 (3!)
2 x
It can be seen that Y0 ( x ) ~ J 0 ( x ) ln + γ as x → 0 . However, since J 0 ( x ) ~ 1
π 2
and | ln x | >> γ − ln 2 for small x we deduce that Y0 ( x ) has the logarithmic
2
behaviour Y0 ( x ) ~ ln x as x → 0 (see Fig.3).
π
137
Ordinary Differential
Equations y
1 y 0 (x )
2 4 6 7
0 x
1 3 5
−1
Fig.3: Bessel function of the second kind Y0 ( x )
For n = 1,2, K the solution can be obtained in a similar manner. It can be seen that the
indicial roots k = ± n differ by an integer and the second linearly independent solution
always contains a logarithmic term (see Case 2, Sec.2.4.2, Unit 2).
This solution can be written as
n −1 2 j− n
1 (n − j − 1) ! x
Yn ( x ) = J n ( x ) ln x −
2
∑
j= 0 j!
2
∞ j 2 j+ n
1 (−1) [ φ( j) + φ ( j + n ) ] x
−
2
∑
j= 0 j! ( j + n ) !
2
(66)
1 1 1
where, φ(0) = 0 and φ ( j) = 1 + + + L + , for j ≥ 1 .
2 3 j
Thus, we have two different general solutions for Eqn.(1), depending on whether υ
is an integer or not.
y( x ) = A J υ ( x ) + B J − υ ( x ), if υ is not an integer
and y( x ) = A J n ( x ) + B Yn ( x ) , if υ = n , an integer.
Without proof we state the result and define the Bessel function of the second kind of
order υ as
(cos υ π) J υ ( x ) − J − υ ( x )
Yυ ( x ) = (67)
sin υ π
and Yn ( x ) = lim Yυ ( x )
υ→n
that is, the limit of Yυ ( x ) as υ → n (n = 0,1, 2, K) give the same result as (66).
Further, you may observe that for υ > 0 , J υ ( x ) ~ 0 for x → 0 and thus
− J −υ (x)
Yυ ( x ) ~
sin υ π
− ( x / 2) − υ
~ , υ>0 , x→0
Γ(1 − υ) sin υπ
But since Γ( x )Γ(1 − x ) = π / sin πx , we finally obtain the asymptotic formula
υ
Γ ( υ) 2
Yυ ( x ) ~ − , υ > 0, x → 0 (68)
π x
Also since J υ ( x ) and Yυ ( x ) are linearly independent, the general solution of the
Bessel’s differential equation for all values of υ , is
y(x) = A J υ (x) + B Yυ (x) (69)
where Yυ ( x ) is defined by Eqn.(67).
138
Bessel Functions
Let us look at the following examples.
Example 12: Find the solution of the differential equation
x 2 y ′′ + xy ′ + ( x 2 − 16) y = 0
Solution: The given equation is the Bessel’s equation with υ = 4 . Its general
solution is thus written as
y( x ) = A J 4 ( x ) + B Y4 ( x ) .
***
Example 13: Using the substitution z = 2 x , obtain the solution of the differential
equation xy ′′ + y ′ + y = 0 .
***
You may now try the following exercise.
E14) Using the substitutions given in E10) reduce the following equations to the
Bessel’s equations and hence find their solution
1
a) x 2 y ′′ + xy ′ + ( x − n 2 ) y = 0
4
b) xy ′′ + y = 0
c) xy ′′ + y ′ + y / 4 = 0 .
n =1 n=2 n =3
The constants a n and b n are selected in such a way that the initial conditions (73) are
satisfied. Therefore, we find
∞
u ( r , 0) = f ( r ) = ∑
n =1
a n J 0 (k n r ) , 0 < r < 1 (84)
and
∞
∂u
∂t
( r , 0) = g ( r ) = ∑ n =1
k n c b n J 0 (k n r ), 0 < r < 1 (85)
Eqns.(84) and (85) are the Bessel series for f (r ) and g (r ) , respectively, where
1
2
an =
[J 1 (k n )] 2 ∫ r f (r ) J
0
0 (k n r ) dr, n = 1, 2, K (86)
and
1
2
k nc bn =
[J 1 (k n ) ] 2 ∫ r g( r ) J
0
0 (k n r ) dr, n = 1, 2, K (87)
141
Ordinary Differential give the coefficients of the series (83) and the required solution is completely
Equations determined.
Before we consider another application of Bessel functions you may try the following
exercises.
E15) Show that if u = u (r, t ) , then in polar coordinate system Eqn.(70) reduces to
Eqn.(71).
E16) If the initial conditions (73) are given by
u (r, 0) = A J 0 (k 3 r ) (A constant )
∂u
( r , 0) = 0
∂r
where J 0 (k 3 ) = 0 , show that the subsequent displacements of the membrane
are described by
u (r, t ) = A J 0 (k 3 r ) cos k 3 ct .
E17) If the initial conditions (73) are given by f (r ) = 0 and g (r ) = 1 , show that
the subsequent displacement of the membrane are described by
2 ∞ sin k n ct
u (r, t ) = ∑
c n =1 k 2n J 1 (k n )
J 0 (k n r )
where, J 0 (k n ) = 0 (n = 1, 2, K) .
x=b
x=0 y
1
= lim x C1 k x − C 2
x →0
πk x
1
= −C 2 (95)
πk
which can vanish only if C 2 = 0 . The second boundary condition then requires that
y ( b ) = C1 b J 1 ( 2 k b ) = 0 (96)
143
Ordinary Differential There are infinitely many solutions of J 1 (2k b ) = 0 , the first of which yields the
Equations
Euler critical load P1 . That is, if we let k 1 denote the smallest value of k for which
Eqn.(96) is satisfied, the corresponding Euler critical load is
P1 = Eαk 12 (97)
leading to the fundamental deflection mode
y = C1 x J 1 ( 2 k 1 x ) (98)
where C1 remains undetermined.
E18) Show that the solution of the boundary value problem (89) for constant E, I
EI
and P gives the critical load P1 = π 2 2 and corresponding deflection
b
πx
y = C sin , where C is an arbitrary constant.
b
E19) Verify that on substituting y = x z and t = x , Eqn.(92) reduces to Eqn.(93).
We now end this unit by giving a summary of what we have covered in it.
4.4 SUMMARY
In this unit we have learnt the following points:
1. The differential equation x 2 y ′′ + xy ′ + ( x 2 − υ 2 ) y = 0 , where υ is a non-
negative real number is known as Bessel’s equation of order υ and its
solutions are known as Bessel functions.
2. When υ is not an integer, one of the linearly independent series solution of
Bessel’s equation is the Bessel function of the first kind of order υ and is
denoted by J υ ( x ) where
∞ 2m+ υ
(−1) m x
J υ (x) = ∑ m =0
m!Γ( m + υ + 1) 2
3. The second linearly independent solution of Bessel’s equation is J − υ ( x ) and
thus the general solution of Bessel’s equation when υ is not an integer is
y( x ) = A J υ ( x ) + B J − υ ( x ) , where A and B are arbitrary constants.
∞ 2m+ n
(−1) m x
4. When υ= n is an integer then J n ( x ) = ∑
m = 0 m! ( m + n ) ! 2
is one linearly
144
Bessel Functions
d
[ x υJ υ(x) ] = x υJ υ−1 (x)
dx
d
[ x − υJ υ(x) ] = −x − υJ υ+1 (x)
dx
x J ′υ(x) + υJ υ(x) = x J υ−1 (x)
x J ′υ(x) − υJ υ(x) = −x J υ+1 (x)
2 J ′υ(x) = J υ−1 (x) − J υ+1 (x)
2υJ υ(x) = xJ υ−1 (x) + xJ υ+1 (x)
7. The function on the left side of equation
x 1 ∞
t−
e2 t
= ∑
n = −∞
J n (x ) t n , t ≠ 0
4.5 SOLUTIONS/ANSWERS
∞
(−1) r ( x / 2)1 / 2 + 2 r ( x / 2)1 / 2 ( x / 2) 5 / 2 ( x / 2) 9 / 2
E1) J1 / 2 (x) = ∑
r =0 r ! Γ ( r + 3 / 2)
= − +
Γ(3 / 2) 1!Γ(5 / 2) 2 !Γ(7 / 2)
−L
( x / 2) 1 / 2 ( x / 2) 5 / 2 ( x / 2) 7 / 2
= − + −L
(1 / 2) π 1!(3 / 2) (1 / 2) π 2 !(5 / 2) (3 / 2) (1 / 2) π
( x / 2) 1 / 2 x 2 x 4 ( x / 2)1 / 2 sin x 2
= 1 − + − L = = sin x
(1 / 2) π 3! 5 ! (1 / 2) π x πx
∞
(−1) r ( x / 2)1 / 2+ 2 r ( x / 2) −1 / 2 ( x / 2) 3 / 2 ( x / 2) 7 / 2
J −1 / 2 ( x ) = ∑
r =0 r ! Γ ( r + 1 / 2)
=
Γ(1 / 2)
−
1!Γ(3 / 2)
+
2 !Γ(5 / 2)
−L
( x / 2) −1 / 2 x 2 x 4 2
= 1 − + − L = cos x
π 2! 4! πx
E2) a) y ′′ + k 2 xy = 0
2k 3 / 2
Put y = u x and z = x
3
y′ =
d
dx
(
u x = x u′ + ) 1
u
2 x
1 1 1
y ′′ = x u ′′ + u′ + u′ − u
2 x 2 x 4( x ) 3 / 2
1 1
Then , y ′′ + k 2 xy = 0 ⇒ x u ′′ + u′ − 3/ 2
u + k 2 (x) 3 / 2 u = 0
x 4( x )
du du dz 2k 3 1 / 2 du du
u′ = = . = . .x . = kx 1 / 2
dx dz dx 3 2 dz dz
2 2
d 1 / 2 du k du 1 / 2 d u dz k du 2 d u
u ′′ = kx = + kx . = + k x
dx dz 2 x dz dz 2 dx 2 x dz dz 2
The given equation reduces to
d2u k du 1 du 2 3 / 2 1
x k 2 x 2 + + k x + k x − u=0
dz 2 x dz x dz 4x 3 / 2
145
Ordinary Differential
d 2 u 1 du 1
2
Equations or, + + 1 − u = 0
dz 2 z dz 3z
du 2 1
2
d2u
or, z2 2 + z + z − u = 0 , which is Bessel’s equation of
dz dz 3
order 1 / 3 and its solution is
u = J 1 / 3 (z)
2k 3 / 2
⇒ y = x u = x J 1 / 3 (z) = x J 1 / 3 x .
3
2
kx
b) Put y = u x and z = , then the given equation transforms to
2
du 2 1
2
d2u 1
z2 2 + z + z − u = 0 , which is Bessel’s equation of order .
dz dz 4 4
kx 2
Hence, y = x u = x J 1 / 4 .
2
3
kx
c) Put y = u x and z =
3
Then given equation transforms to
du 2 1
2 2
2 d u
z +z + z − u = 0 ,
dz 2 dz 6
which is Bessel’s equation of order 1 / 6 and its solution is
kx 3
y = x u = x J 1 / 6 (z) = x J 1 / 6 .
3
dy 1 dy d 2 y 1 d 2 y 1 dy
E3) a) z= x ⇒ = , 2
= 2
− 3/ 2
dx 2 x dz dx 4 x dz 4x dz
Substituting these values in the given equation, we obtain
d2y dy
x 2 + x + (x − υ 2 ) y = 0
dz dz
∴y = A J υ ( x ) + B J −υ ( x )
c) y = u / x k , y ′ = x − k u ′ − kx − k −1 u
y ′′ = x − k u ′′ − 2kx − k −1 u ′ + k (k + 1) x − k − 2 u
The given equation reduces to
x 2 u ′′ + xu ′ + ( x 2 − k 2 ) u = 0
Its general solution is
u = A J k (x) + B J −k (x)
or, y = x − k [A J k ( x ) + B J − k ( x )]
2 8 4
E6) J 2 (x) = J1 (x) − J 0 (x ) , J 3 ( x ) = 2 − 1 J 1 ( x ) − J 0 ( x )
x x x
∫x ∫x
5
e) J 0 dx = x 5 J 1 − 4 4
J 1 dx
∫
= x 5 J 1 − 4 x 4 J 2 + 8 x 3 J 2 dx = x 5 J 1 − 4 x 4 J 2 + 8x 3 J 2 + C
f) Let r x = t and use Eqn.(29).
g) Using Eqn.(29), we obtain
1
∫ (x − x
3
) J 0 dx = 2J 2 (1) = 2 [2J 1 (1) − J 0 (1) ] [using Eqn.(37)]
0
h) Integrate by parts, use Eqn.(32) and again integrate by parts.
147
Ordinary Differential For n = 0 , we have
Equations ∞
J 0 ( x + y) = ∑
k = −∞
J − k ( x ) J k ( y)
∞ ∞
= J 0 ( x )J 0 ( y) + ∑
k =1
J − k ( x ) J k ( y) + ∑
k =1
J k ( x ) J − k ( y)
∞
= J 0 ( x ) J 0 ( y) + ∑
k =1
(−1) k [ J k ( x ) J k ( y) + J k ( x ) J k ( y) ]
∞
= J 0 ( x )J 0 ( y) + ∑
k =1
(−1) k 2J k ( x ) J k ( y)
or, J 0 ( x + y) = J 0 ( x ) J 0 ( y) − 2J 1 ( x ) J 1 ( y) + 2J 2 ( x ) J 2 ( y) − L
Replacing y by − x and using the fact that J n ( x ) is even or odd according as
n is even or odd, the required identity is obtained.
1
2
E12) We have, a n = 2
J 1 (λ n ) ∫
0
x f ( x ) J 0 (λ n x ) dx
1 1/ 2 1/ 2
x 1
Now ∫
0
x f ( x )J 0 (λ n x ) dx = ∫
0
x J 0 (λ n x ) dx =
λn
J 1 (λ n x )
0
=
2λ n
J 1 ( λ n / 2)
1 1
∴a n = 2 J 1 ( λ n / 2)
J 1 (λ n ) λ n
∞
J 1 ( λ n / 2)
and it follows that, f ( x ) = ∑
n =1 λ n J 12 (λ n )
J 0 (λ n x )
1
2
∫x
υ +1
E13) We have, a n = 2 J υ (λ n x ) dx
J υ +1 (λ n ) 0
1
1 υ+1 1
∫x
υ+1 1
Now J υ (λ n x ) dx = x J υ+1 (λ n x ) 0
= J υ+1 (λ n )
0
λn λn
2 1
∴ an = J υ +1 (λ n )
J 2υ+1 (λ n ) λ n
∞
2
and x υ = ∑
n =1 λ n J υ+1 (λ n )
J υ (λ n x )
1
E14) a) Comparing with E10), we get a = 0, c = , b = 1 , therefore
2
y = u , t = x , z = t and y = A J n ( x ) + B J − n ( x ) or,
A J n ( x ) + B Yn ( x ) , depending on n being a non-integer or an integer.
b) Comparing with E10), we get a = 1 / 2, c = 1 / 2, b = 2, n = 1 or
y = u x , t = x , z = 2t , y = x [A J 1 (2 x ) + B Y1 (2 x ) ] .
c) Comparing with E10), we get y = u, t = x , z = t ,
y = A J 0 ( x ) + B Y0 ( x ) .
⇒ b n = 0, n = 1, 2, K
and a n ' s are given by
1
2
an = 2
J 1 (k n ) ∫0
r AJ 0 (k 3 r )J 0 ( k n r )dr, 0 < r < 1
1
Now since, ∫
0
rJ 0 (k 3 r )J 0 (k n r ) dr = 0 , if n ≠ 3
1 2
= J 1 (k 3 ) , for n = 3
2
∴ a n = 0 for n ≠ 3
2A
1 2
= J 1 (k 3 ) = A for n = 3
J 12 (k 3 )
2
∴ solution (83) reduces to
u (r, t ) = AJ 0 (k 3 r ) cos k 3 ct .
⇒ a n = 0 , n = 1, 2, K
∞
and 1 = ∑
n =1
k n cb n J 0 (k n r ) , 0 < r < 1
1
2
∴k n cb n = 2
J 1 (k n ) ∫
0
r J o (k n r ) dr
1
1 1
∫
1
Now r J 0 (k n r ) dr = r J 1 (k n r ) 0
= J 1 (k n )
0
kn kn
2 1
∴b n = 2
c k n J 1 (k n )
149
Ordinary Differential P
Equations E18) Eqn.(89) gives y ′′ + y = 0 , its solution is given by
EI
P P
y( x ) = B cos x + C sin x , B and C arbitrary constants
EI EI
P
y(0) = 0 ⇒ B = 0 , therefore y = C sin x
EI
P P EIπ 2
y(b) = C sin b = 0 , if C ≠ 0 then b = π or , P = 2
EI EI b
πx
∴ y( x ) = C sin
b
150