S.
Ghorai 1
Lecture XV
Bessel’s equation, Bessel’s function
1 Gamma function
Gamma function is defined by
Z ∞
Γ(p) = e−t tp−1 dt, p > 0. (1)
0
The integral in (1) is convergent that can be proved easily. Some special properties of
the gamma function are the following:
i. It is readily seen that Γ(p + 1) = pΓ(p), since
Z T
Γ(p + 1) = lim e−t tp dt
T →∞ 0
" Z T #
T
= lim −e−t tp +p −t p−1
e t dt
T →∞ 0 0
Z ∞
= p e−t tp−1 dt = pΓ(p).
0
ii. Γ(1) = 1 (trivial proof)
iii. If p = m, a positive integer, then Γ(m + 1) = m! (use i. repeatedly)
√
iv. Γ(1/2) = π. This can be proved as follows:
Z ∞ Z ∞
−t −1/2 2
I = Γ(1/2) = e t dt = 2 e−u du.
0 0
Hence Z ∞ Z ∞ Z ∞Z ∞
2 2 2 +y 2 )
I2 = 4 e−u du e−v dv = 4 e−(x dx dy.
0 0 0 0
Using polar coordinates ρ, θ, the above becomes
Z ∞ Z π/2
2 √
I2 = 4 e−ρ ρ dρ dθ ⇒ I 2 = π ⇒ I = π
0 0
v. Using relation in i., we can extend the definition of Γ(p) for p < 0. Suppose N is
a positive integer and −N < p < −N + 1. Now using relation of i., we find
Γ(p + 1) Γ(p + 2) Γ(p + N )
Γ(p) = = = ··· = .
p p(p + 1) p(p + 1) · · · (p + N − 1)
Since p + N > 0, the above relation is well defined.
vi. Γ(p) is not defined when p is zero or a negative integer. For small positive ,
Γ(1 ± ) 1
Γ(±) = ≈ → ±∞ as → 0.
± ±
Since Γ(0) is undefined, Γ(p) is also undefined when p is a negative integer.
S. Ghorai 2
2 Bessel’s equation
Bessel’s equation of order ν (ν ≥ 0) is given by
x2 y 00 + xy 0 + (x2 − ν 2 )y = 0. (2)
Obviously, x = 0 is regular singular point. Since p(0) = 1, q(0) = −ν 2 , the indicial
equation is given by
r2 − ν 2 = 0.
Hence, r1 = ν, r2 = −ν and r1 − r2 = 2ν. A Frobenius series solution exists for the
larger root r = r1 = ν. To find this series, we substitute
∞
y = xr an x n ,
X
x>0
n=0
into (2) and (after some manipulation) find
∞ ∞
n
an−2 xn = 0
X X
ρ(n + r)an x +
n=0 n=2
where ρ(r) = r2 − ν 2 . This equation is rearranged as
∞
ρ(n + r)an + an−2 xn = 0.
X
ρ(r)a0 + ρ(r + 1)a1 x +
n=2
Hence, we find (since a0 6= 0)
ρ(r) = 0, ρ(r + 1)a1 = 0, ρ(r + n)an = −an−2 , n ≥ 2.
From the first relation, we get r1 = ν, r2 = −ν. Now with the larger root r = r1 we
find
an−2
a1 = 0, an = − , n ≥ 2.
n(n + 2ν)
Iterating we find (by induction),
1
a2n+1 = 0, a2n = (−1)n a0 , n ≥ 1.
22n n!(ν + 1)(ν + 2) · · · (ν + n)
Hence ∞
(−1)n x2n
!
ν
X
y1 (x) = a0 x 1+ 2n
. (3)
n=1 2 n!(ν + 1)(ν + 2) · · · (ν + n)
Here it is usual to choose (instead of a0 = 1 as was done in lecture 14 )
1
a0 = .
2ν Γ(ν + 1)
Then the Frobenius series solution (3) is called the Bessel function of order ν of the
first kind and is denoted by Jν (x):
∞
(−1)n
2n+ν
X x
Jν (x) = . (4)
n=0 n!Γ(n + ν + 1) 2
To find the second independent solution, we consider the following three cases:
S. Ghorai 3
A. r1 − r2 = 2ν is not a nonnegative integer: We know that a second Frobenius
series solution for r2 = −ν exist. We do similar calculation as in the case of r1
and it turns out that the resulting series is given by (4) with ν replaced by −ν.
Hence, the second solution is given by
∞
(−1)n
2n−ν
X x
J−ν (x) = . (5)
n=0 n!Γ(n − ν + 1) 2
B. r1 = r2 : Obviously this corresponds to ν = 0 and a second Frobenius series
solution does not exist.
C. r1 − r2 = 2ν is a positive integer: Now there are two cases. We discuss them
separately.
C.i ν is not a positive integer: Clearly ν = (2k+1)/2, where k ∈ {0, 1, 2, · · ·}.
Now we have found earlier that (since a0 6= 0)
ρ(r) = 0, ρ(r + 1)a1 = 0, ρ(r + n)an = −an−2 , n ≥ 2.
With r = r2 = −ν, we get
ρ(r) = 0; 1 · 1 − (2k + 1) a1 = 0; n · n − (2k + 1) an = −an−2 , n ≥ 2.
It is clear that the even terms a2n can be determined uniquely. For odd
terms, a1 = a3 = · · · = a2k−1 = 0 but for a2k+1 we must have
n · 0 · a2k+1 = −a2k−1 ⇒ 0 · a2k+1 = 0.
This can be satisfied by taking any value of a2k+1 and for simplicity, we can
take a2k+1 = 0. Rest of the odd terms thus also vanish. Hence, the second
solution in this case is also given by (5), i.e.
∞
(−1)n
2n−ν
X x
J−ν (x) = . (6)
n=0 n!Γ(n − ν + 1) 2
[Link] ν is a positive integer: Clearly ν = k, where k ∈ {1, 2, 3, · · ·}. Now we
find (since a0 6= 0)
ρ(r) = 0, ρ(r + 1)a1 = 0, ρ(r + n)an = −an−2 , n ≥ 2.
With r = r2 = −ν, we get
ρ(r) = 0; 1 · 1 − 2k a1 = 0; n · n − 2k an = −an−2 , n ≥ 2.
It is clear that all the odd terms a2n+1 vanish. For even terms, a2 , a4 , · · · , a2k−2
each is nonzero. For a2k we must have
n · 0 · a2k = −a2k−2 ⇒ 0 · a2k 6= 0,
which is a contradiction. Thus, a second Frobenius series solution does not
exist in this case.
S. Ghorai 4
Summary of solutions for Bessel’s equation: The Bessel’s equation of order ν
(ν ≥ 0)
x2 y 00 + xy 0 + (x2 − ν ) y = 0,
has two independent Frobenius series solutions Jν and J−ν when ν is not an (nonneg-
ative) integer:
∞ ∞
(−1)n (−1)n
2n+ν 2n−ν
X x X x
Jν (x) = , J−ν (x) = .
n=0 n!Γ(n + ν + 1) 2 n=0 n!Γ(n − ν + 1) 2
Thus the general solution, when ν is not an (nonnegative) integer, is
y(x) = C1 Jν (x) + C2 J−ν (x).
When ν is a (nonnegative) integer, a second solution, which is independent of Jν , can
be found. This solution is called Bessel function of second kind and is denoted by Yν .
Hence, the general solution, when ν is an (nonnegative) integer, is
y(x) = C1 Jν (x) + C2 Yν (x).
3 Linear dependence of Jm and J−m , m is a +ve integer
When ν = m is a positive integer, then
∞ ∞
(−1)n (−1)n
2n+m 2n+m
X x X x
Jm (x) = = ,
n=0 n!Γ(n + m + 1) 2 n=0 n!(n + m)! 2
since Γ(n + m + 1) = (n + m)!.
Since Γ(±0) = ±∞, we define 1/Γ(k) to be zero when k is nonpositive integer. Now
∞
(−1)n
2n−m
X x
J−m (x) = .
n=0 n!Γ(n − m + 1) 2
Now each term in the sum corresponding to n = 0 to n = m − 1 is zero since 1/Γ(k)
is zero when k is nonpositive integer. Hence, we write the sum starting from n = m:
∞
(−1)n
2n−m
X x
J−m (x) = .
n=m n!Γ(n − m + 1) 2
Substituting n − m = k, we find
∞
(−1)k+m
2(m+k)−m
X x
J−m (x) =
k=0 (m + k)!Γ(k + 1) 2
∞
(−1)k
2k+m
x
= (−1)m
X
k=0 k!(m + k)! 2
= (−1)m Jm (x).
Hence Jm and J−m becomes linearly dependent when m is a positive integer.
S. Ghorai 5
4 Properties of Bessel function
Few important relationships are very useful in application. These are described here.
A. From the expression for Jν given in (4), we find
∞
(−1)n
2n+2ν
ν
X x
x Jν (x) =
n=0 n!Γ(n + ν + 1) 2
Taking derivative with respect to x we find
∞ ∞
(−1)n (n + ν) x (−1)n
2n+2ν−1 2n+2ν−1
0 x
xν Jν (x)
X X
= = ,
n=0 n!Γ(n + ν + 1) 2 n=0 n!Γ(n + ν) 2
where we have used Γ(n + ν + 1) = (n + ν)Γ(n + ν). We can write the above
relation as
∞
(−1)n
2n+ν−1
ν
0
ν
X x
x Jν (x) =x .
n=0 n!Γ n + (ν − 1) + 1 2
Hence, 0
xν Jν (x) = xν Jν−1 (x). (7)
B. From (4), we find
∞
−ν (−1)n
x2n .
X
x Jν (x) = 2n+ν
n=0 2 n!Γ(n + ν + 1)
Taking derivative with respect to x we find
∞
0 (−1)n
x−ν Jν (x) x2n−1 .
X
= 2n+ν−1 (n − 1)!Γ(n + ν + 1)
n=1 2
Note that the sum runs from n = 1 (in contrast to that in A). Let k = n − 1,
then we obtain
∞
(−1)k+1
2k+ν+1
−ν
0
−ν
X x
x Jν (x) = x
k=0 k!Γ k + (ν + 1) + 1 2
∞
(−1)k
2k+ν+1
−ν
X x
= −x .
k=0 k!Γ k + (ν + 1) + 1 2
Hence, 0
x−ν Jν (x) = −x−ν Jν+1 (x). (8)
Note: In the first relation A, while taking derivative, we keep the sum running
from n = 0. This is true only when ν > 0. In the second relation B, we only
need ν ≥ 0. Taking ν = 0 in B, we find J00 = −J1 . If we put ν = 0 in A, then we
find J00 = J−1 . But J−1 = −J1 and hence we find the same relation as that in B.
Hence, the first relation is also valid for ν ≥ 0.
S. Ghorai 6
C. From A and B, we get
ν
Jν0 + Jν = Jν−1
x
0 ν
Jν − Jν = −Jν+1
x
Adding and subtracting we find
Jν−1 − Jν+1 = 2Jν0 (9)
and
2ν
Jν−1 + Jν+1 = Jν . (10)
x
