Bessel’s Function

In standard form, Bessel’s differential equation is

d2 y dy
x2 2
+ x + (x2 − n2 )y = 0 (1)
dx dx
where n ≥ 0 is a real number.
Another useful form of Bessel’s differential equation is
d2 y dy
x2 + x + (m2 x2 − n2 )y = 0 (2)
dx2 dx
When developing the properties of Bessel’s function, the form (1) will be used. Bessel’s
differential equation has a regular singular point at x = 0.
∞
ar xr+c be the trial solution of (1). Then substituting the value of y, y 0 and
P
Let y(x) =
r=0
y 00 in (1), we get
∞ ∞
(r + c)(r + c − 1)ar xr+c + (r + c)ar xr+c +
P P
r=0 r=0
∞ ∞
ar xr+c+2 − n2 ar xr+c = 0
P P
r=0 r=0

Shifting the summation index in the third summation and collecting terms under a single
summation gives
∞
(c2 − n2 )a0 xc + [(c + 1)2 − n2 ]a1 xc+1 + [(r + c + n)(r + c − n)ar + ar−2 ]xr+c = 0
P
r=2

Equating the coefficients of powers of xc , xc+1 and xr+c to zero shows the following:
Coefficient of xc : (c2 − n2 )a0 = 0 with a0 6= 0 gives c = ±n.
Coefficient of xc+1 : [(c + 1)2 − n2 ]a1 = 0.
Coefficient of xr+c :
[(r + c)2 − n2 ]ar + ar−2 = 0 (3)

Since with c = ±n, (c + 1)2 − n2 6= 0, we must have a1 = 0. It follows from (3) that
a1 = a3 = a5 = · · · = 0, i.e., ar = 0 for all odd integer r. As only even indices r are
involved, in the recurrence relation we set r = 2m with m = 0, 1, 2, . . . .
Substituting c = n in (3), we get
1
a2m = − a2m−2 , m = 1, 2, 3, . . . (4)
4m(m + n)
1
Since a0 is arbitrary, set a0 = 2n Γ(n+1) , where Γ(n+1) is the gammma function. We define
R ∞ −t n−1
Γ(n) = 0 e t dt, where n > 0
We have Γ(n + 1) = nΓ(n) = n! for n = 0, 1, 2, . . . . Therefore,
a0 1 a0 1
a2 = − 22 (1+n) = − 22+n 1!Γ(2+n) , a4 = − 22 2(2+n) = 24+n 2!Γ(3+n) ,

1
2

(−1)m a0
and in general, a2m = for m = 1, 2, . . ..
22m+n m!Γ(m+1+n)
∞
ar xr+c , we obtain a particular solution of
P
By inserting these coefficients in y(x) =
r=0
(1) which is denoted by Jn (x) and is defined as
∞
X (−1)m x
Jn (x) = ( )2m+n , n = 0, 1, 2, . . . , (5)
m=0
m!(m + n)! 2

which is known as the Bessel function of order n.

Bessel function
∞
of the first kind of order zero:
P (−1)m x 2m x2 x4 x6
J0 (x) = ( ) = 1 − 22 (1!)
(m!)2 2 2 + 24 (2!)2 − 26 (3!)2
+ ···
m=0

Bessel function
∞
of the first kind of order one:
(−1)m 3 x5 x7
( x )2m+1 = x2 − 23x1!2! +
P
J1 (x) = m!(m+1)! 2 25 2!3!
− 27 3!4!
+ ···
m=0
p 2
p 2
Problem: Prove that (i) J 1 (x) = ( πx )sinx ; (ii) J− 1 (x) = ( πx )cosx.
2 2
Proof: We have
xn x2 x4
Jn (x) = [1 − + − ··· (6)
2n Γ(n + 1) 2 × 2(n + 1) 2 × 4 × 22 (n + 1)(n + 2)
1
(i) Substituting n = 2
in (6), we get
1
2 4
J 1 (x) = 1 1 [1 − 2×2(x1 +1) + 2×4×22 ( x1 +1)( 1 +2) − · · ·]
x2
2
√ 2 Γ( 2 +1)
2 2 2 2

(x) 2 4 p 3 5
=√ 3
[1 − x3! + x5! − · · ·] = ( x2 ) 1 Γ(1 1 ) x1 [x − x3! + x5! − · · ·]
p(2)Γ(
2
2
)
1
p 2 2

= ( πx )sinx with Γ( 2 ) = (π).

(ii) Substituting n = − 21 in (6), we get
−1 2 4
J− 1 (x) = − 1 x 21 [1 − 2×2(− x
1
+1)
+ 2×4×22 (− x1 +1)(− 1 +2) − · · ·]
2 2 2 Γ(− 2 +1) 2 2 2
√
(2) 2 4 p 2
=√ 1
x x
[1 − 2! + 4! − · · ·] = ( πx )cosx.
(x)Γ( 2 )

Recurrence relations
Form I: xJn0 (x) = nJn (x) − xJn+1 (x).
Proof: We have∞
(−1)m
( x )2m+n
P
Jn (x) = m!Γ(m+n+1) 2
m=0
Differentiating with respect to x, we get
∞
(−1)m (2m+n) x 2m+n−1 1
Jn0 (x) =
P
m!Γ(m+n+1) 2
( ) ·2
m=0
∞
(−1)m
xJn0 (x) = n ( x )2m+n
P
m!Γ(m+n+1) 2
m=0
∞
(−1)m (2m)
( x )2m+n−1
P
+x 2(m!)Γ(m+n+1) 2
m=1
 3

∞
(−1)m
( x )2m+n−1
P
= nJn (x) + x (m−1)!Γ(m+n+1) 2
m=1

Putting m − 1 = p, we get
∞
(−1)p+1
xJn0 (x) = nJn (x) + x ( x )2p+n+1
P
p!Γ(p+n+2) 2
p=0
∞
(−1)p
( x )2p+(n+1)
P
= nJn (x) − x p!Γ(p+(n+1)+1) 2
p=0
= nJn (x) − xJn+1 (x).
Form II: xJn0 (x) = −nJn (x) + xJn−1 (x).
Proof: We have
∞
(−1)m
( x )2m+n
P
Jn (x) = m!Γ(m+n+1) 2
m=0
Differentiating with respect to x, we get
∞
(−1)m (2m+n) x 2m+n−1 1
Jn0 (x) =
P
m!Γ(m+n+1) 2
( ) ·2
m=0
∞
(−1)m (2m+n) x 2m+n
xJn0 (x) =
P
m!Γ(m+n+1) 2
( )
m=0
∞
P (−1)m {2(m+n)−n} x 2m+n
= m!Γ(m+n+1)
(2)
m=0
∞ ∞
(−1)m (2m+2n) x 2m+n (−1)m
( x )2m+n
P P
= m!Γ(m+n+1) 2
( ) −n m!Γ(m+n+1) 2
m=0 m=0
∞
P (−1)m ·2 x 2m+n
= m!Γ(m+n) 2
( ) − nJn (x)
m=0
∞
(−1)m
( x )2m+n−1
P
= m!Γ{m+(n−1)+1} 2
− nJn (x)
m=0
= xJn−1 (x) − nJn (x).
Form III: 2Jn0 (x) = Jn−1 (x) − Jn+1 (x).
Proof: We have
xJn0 (x) = nJn (x) − xJn+1 (x)
and
xJn0 (x) = −nJn (x) + xJn−1 (x)
Adding the above two equations, we get
2xJn0 (x) = −xJn+1 (x) + xJn−1 (x)
Thus the result follows.
Form IV: 2nJn (x) = x[Jn−1 (x) + Jn+1 (x)].
Proof: We have
xJn0 (x) = nJn (x) − xJn+1 (x) (1)
and
xJn0 (x) = −nJn (x) + xJn−1 (x) (2)
Subtracting (2) from (1), we get
0 = 2nJn (x) − xJn+1 (x) − xJn−1 (x).
Thus the result follows.
4

Form V: dxd
[x−n Jn (x)] = −x−n Jn+1 (x).
Proof: We have
xJn0 (x) = nJn (x) − xJn+1 (x) (1)
Multiplying (1) by x−n−1 , we get
x−n Jn0 (x) = nx−n−1 Jn (x) − x−n Jn+1 (x)
x−n Jn0 (x) − nx−n−1 Jn (x) = −x−n Jn+1 (x)
Thus d
dx
[x−n Jn (x)] = −x−n Jn+1 (x).
d
Form VI: dx [xn Jn (x)] = xn Jn−1 (x).
Proof: We have
xJn0 (x) = −nJn (x) + xJn−1 (x) (2)
Multiplying (2) by xn−1 , we get
xn Jn0 (x) + nxn−1 Jn (x) = xn Jn−1 (x)
d
Thus dx
[xn Jn (x)] = xn Jn−1 (x).
Orthogonality property of Bessel’s function
(A) Show that
Z 1
xJn (αx)Jn (βx)dx = 0 (7)
0

where α and β are the roots of Jn (x) = 0.

R1
(B) Show that 0 x[Jn (αx)]2 dx = 21 [Jn+1 (α)]2 .
Proof: (A) We have
d2 y dy
x2 2
+ x + (α2 x2 − n2 )y = 0 (8)
dx dx
and

d2 z dz
x2 2
+ x + (β 2 x2 − n2 )z = 0 (9)
dx dx

Solutions of (8) and (9) are Jn (αx), Jn (βx), respectively.

−y
Multiplying (8) by xz , (9) by x
and then adding the equations, we get
2
d y 2 dy
d z dz 2 2
x(z dx2 − y dx2 ) + (z dx − y dx ) + (α − β )xyz = 0

d dy dz
dx
[x(z dx − y dx )] + (α2 − β 2 )xyz = 0 (i)
Integrating (i) with respect to x and taking limit from 0 to 1, we get
dy dz 1
R1
x(z dx − y dx )|0 + (α2 − β 2 ) 0 xyzdx = 0
R1 dy dz
⇒ (β 2 − α2 ) 0 xyzdx = [z dx − y dx ]x=1 (ii)
 5

dy
We have y = Jn (αx) and z = Jn (βx). So that dx = αJn0 (αx) and dx
dz
= βJn0 (βx)
Therefore,R (ii) becomes
1
(β 2 − α2 ) 0 xJn (αx)Jn (βx)dx = [αJn0 (αx)Jn (βx) − βJn0 (βx)Jn (αx)]x=1
= αJn0 (α)Jn (β) − βJn0 (β)Jn (α) (iii)
Since α and β are the roots of Jn (x) = 0, so Jn (α) = Jn (β) = 0. Hence (iii) becomes
R1
(β 2 − α2 ) 0 xJn (αx)Jn (βx)dx = 0
Thus the result follows.
(B) We have R
1
(β 2 − α2 ) 0 xJn (αx)Jn (βx)dx = αJn0 (α)Jn (β) − βJn0 (β)Jn (α) (iii)
Putting α = β. We have Jn (α) = 0. Let β be a neighboring value of α, i.e., β → α. Then
R1 0
limβ→α 0 xJn (αx)Jn (βx)dx = limβ→α 0+αJβn2(α)J −α2
n (β)

which is an indeterminate form of type 00 .

By RL’Hospital’s rule, we have 0 0
1
0
xJn2 (αx)dx = limβ→α αJn (α)J2β
n (β)
= 12 [Jn0 (α)]2 .
Again from recurrence relation, we know
Jn0 (x) = nx Jn (x) − Jn+1 (x)
Therefore, Jn0 (α) = αn Jn (α) − Jn+1 (α)
ButRJn (α) = 0. So that Jn0 (α) = −Jn+1 (α). Thus
1
0
x[Jn (αx)]2 dx = 12 [Jn+1 (α)]2 . 2

p 2 p 2
We know J 1 (x) = ( πx )sinx ; and J− 1 (x) = ( πx )cosx.
2 2
Also from the recurrence formula IV for Jn (x), we have 2nJn (x) = x[Jn−1 (x)+Jn+1 (x)].
Prove the following
p 2identities:
(i) J 3 (x) = ( πx )[ sinx − cosx].
2 p 2x cosx
(ii) J− 3 (x) = − ( πx )[ x + sinx].
2 p 2 3 sinx
(iii) J 5 (x) = ( πx )[ ( − cosx) − sinx].
2 p 2 x 3 xcosx
(iv) J− 5 (x) = ( πx )[ x ( x + sinx) − cosx].
2