BESSEL FUNCTIONS/pp391 MODERN-QM NOTES BY @EUPHORICRHINO

1. Generating Function
Let
1
•  ‹˜
x
g(x, t) = exp t− (1)
2 t
then we define, for each order n, the Bessel function of the first kind Jn (x) in terms of the expansion
∞
X
g(x, t) = Jn (x)t n (2)
n=−∞

We can match Jn (x) to each of the t n power terms in the expansion of (1)

g(x, t) = e x t/2 · e−x/2t

  x p    x q 
∞ tp ∞ (−1)q t −q
2 2
X X
= · 
p=0
p! q=0
q!
 x  p+q
∞ (−1)q t p−q
2
X
= (3)
p,q=0
p!q!

Now the double sum of p and q runs P∞ all the grid points of the p-q plane’s first quadrant. If we define n = p −q,
P∞through
we can convert the double sum p=0 q=0 into a double sum whose outer sum runs through all the diagonal rays
representing different n (from −∞ to ∞), and whose inner sum runs through all different grid points along the
diagonal ray for a given n. But depending on whether n ≥ 0 or n < 0, the inner sum’s index will be chosen differently
(see figure below).

q P−1 P∞
n=−∞ p=0
2

1
−

0
=

=
n

5
1
=
n

4
2
=
n

3
P∞ P∞
2 n=0 q=0

0
0 1 2 3 4 5 p

The sum in (3) is equivalent to

 x 2p−n  x 2q+n
−1 ∞ (−1) p−n tn ∞ X
∞ (−1)q tn
2 2
X X X
g(x, t) = + (4)
n=−∞ p=0
p!(p − n)! n=0 q=0
q!(q + n)!

Compare (4) with (2), we can easily see that for n ≥ 0,

 x 2q+n
∞ (−1) q
2
X
Jn (x) = (5)
q=0
q!(q + n)!
J−n (x) = (−1)n Jn (x) (6)
2. Non-integer Order
Now we consider to generalize Jn (x) to non-integer orders, denoted Jν (x), where ν is real but not necessarily an
integer. Our starting point is (5) where (x/2)2q+ν is already well defined, so we only need to generalize (q + ν)!. The
natural generalization of factorial is the Γ -function (keep in mind that Γ (n + 1) = n! for positive integer n). Therefore
we define, for arbitrary real order ν, the Bessel function of the first kind Jν (x) as
 x 2q+ν
∞ (−1)q
2
X
Jν (x) = (7)
q=0
q!Γ (q + ν + 1)

Since Γ function is singular at non-positive integers, this means when ν is a negative integer P −n, for (7) to be
Pwell
∞ ∞
defined, we have to restrict q to avoid those singularities. I.e., when ν = −n, the sum has to be q=n instead of q=0 .
The new definition (7) is compatible with (6), since when ν = −n, by (7) we have (note we have shifted q to start
from n instead of 0 given the comments above)
 x 2q−n
∞ (−1) q
2
X
J−n (x) = (relabel q′ = q − n)
q=n
q!Γ (q − n + 1)
 x 2q′ +n
∞ (−1)q +n
′

2
X
= (note (q′ + n)! = Γ (q′ + n + 1), Γ (q′ + 1) = q′ !)
q =0
′ (q ′ + n)!Γ (q ′ + 1)

= (−1)n Jn (x) (8)

3. Recurrence Relations
There are many recurrence relations for Jν (x), here we prove one (which can be used to give the exact form of
spherical Bessel functions later).

2ν
Jν+1 (x) = Jν (x) − Jν−1 (x) (9)
x

Proof. By definition (7), we have

 x 2q+ν+1
x ∞ (−1)q
 x X
Jν+1 (x) = 2
2 2 q=0
q!Γ (q + ν + 2)
 x 2(q+1)+ν
∞ (−1) q
2
X
= (relabel q′ = q + 1)
q=0
q!Γ (q + ν + 2)
 x 2q′ +ν
′
∞ −(−1)q q ′
2
X
= ′ !Γ (q ′ + ν + 1)
(add q′ = 0 term )
q =1
′ q
 x 2q′ +ν
′
∞ −(−1)q q ′
2
X
= ′ !Γ (q ′ + ν + 1)
(10)
q =0
′ q
 x 2q+ν−1
x  x X∞ (−1)q
Jν−1 (x) = 2
2 2 q=0 q!Γ (q + ν)
 x 2q+ν
∞ (−1)q
2
X
= (11)
q=0
q!Γ (q + ν)
 Adding (10) and (11) gives
∞ •  x 2q+ν
1
˜
x X q
[Jν+1 (x) + Jν−1 (x)] = − (−1)q (note Γ (x + 1) = xΓ (x))
2 q=0
q!Γ (q + ν) q!Γ (q + ν + 1) 2
∞
q+ν  x 2q+ν
• ˜
X q
= − (−1)q
q=0
q!Γ (q + ν + 1) q!Γ (q + ν + 1) 2
 x 2q+ν
∞ (−1)q
2
X
=ν·
q=0
q!Γ (q + ν + 1)
= νJν (x) (12)

4. Bessel Equation
We claim Jν (x) satisfies the Bessel equation

x 2 F ′′ (x) + x F ′ (x) + (x 2 − ν2 )F (x) = 0 (13)

Proof. By taking the derivative of Jν (x), we have

2q + ν  x 2q+ν−1
 ‹
∞ (−1)
q
X 2 2
x Jν′ (x) = x ·
q=0
q!Γ (q + ν + 1)
 x 2q+ν
∞ (−1)q (2q + ν)
2
X
= (14)
q=0
q!Γ (q + ν + 1)

q (2q + ν)(2q + ν − 1)
• ˜  2q+ν−2
x
∞ (−1)
X 2 · 2 2
x 2 Jν′′ (x) = x 2 ·
q=0
q!Γ (q + ν + 1)
 x 2q+ν
∞ (−1)q (2q + ν)(2q + ν − 1)
2
X
= (15)
q=0
q!Γ (q + ν + 1)

Adding (14) and (15) gives

 x 2q+ν
∞ (−1)q (2q + ν)2
2
X
x Jν′ (x) + x 2 Jν′′ (x) = (16)
q=0
q!Γ (q + ν + 1)

On the other hand

 x 2q+ν
∞ (−1) q
2
X
x 2 Jν (x) = x 2
q=0
q!Γ (q + ν + 1)
 x 2(q+1)+ν
∞ (−1)q 4
2
X
= (relabel q′ = q + 1)
q=0
q!Γ (q + ν + 1)
 x 2q′ +ν
′
∞ −(−1)q 4q ′ (q ′ + ν)
2
X
= (17)
q′ =0
q′ !Γ (q′ + ν + 1)

Finally (13) is proved by adding (16) and (17) since (2q + ν)2 − 4q(q + ν) = ν2 .

5. Orthonormality
The notion of orthonormality between Bessel functions is different than usual. Let Jν (x) be a Bessel function of real
order ν, and let ανl and ανk be its l-th and k-th root. Then the orthonomality condition claims
Z L 
x  x 1
Jν ανl Jν ανk x d x = L 2 [Jν′ (ανk )]2 δlk (18)
0
L L 2
Note that it applies to Bessel functions of the same order ν, whose arguments are scaled by different roots.

Proof. Since the order ν is fixed, it will be omitted from the subscripts in this proof. The Bessel equation (13) can be
rewritten as
ν2
 
′′ ′
x J (x) + J (x) + x − J(x) = 0 =⇒
x
ν2
• ˜  
d d J(x)
x + x− J(x) = 0 (19)
dx dx x
Now for a given real number β, define x = β y, then (19) can be rewritten as
ν2
˜  
1 d d J(β y) 1
•
βy + βy− J(β y) = 0 =⇒
β dy dy β βy
ν2
• ˜  
d d J(β y)
y 2
+ β y− J(β y) = 0 (20)
dy dy y
Denote J(β y) = Jβ ( y), and similarly define Jγ ( y) for another factor γ, together we have
dJβ ( y) ν2
   
d
y + β y−
2
Jβ ( y) = 0 (21)
dy dy y
dJγ ( y) ν2
   
d
y 2
+ γ y− Jγ ( y) = 0 (22)
dy dy y

Multiply (21) by Jγ and multiply (22) by Jβ and subtract

   
d d Jβ d dJγ

Jγ y − Jβ y = γ2 − β 2 y Jβ Jγ (23)
dy dy dy dy

Integrate (23) from 0 to L, we have

Z L

L
    Z
d d Jβ d dJγ
Jγ y − Jβ y 2
dy = γ −β 2
Jβ Jγ y d y (24)
0
dy dy dy dy 0
whose LHS gives
L L L L
– Z ™ – Z ™
d Jβ dJβ dJγ dJγ dJγ dJβ
LHS = Jγ y − y d y − Jβ y − y dy
dy 0 0
dy dy dy 0 0
dy dy
d Jβ dJγ
= Jγ (L)L − Jβ (L)L (25)
dy L dy L

Then (24) becomes

Z L
d Jβ dJγ

Jγ (L)L − Jβ (L)L = γ −β
2 2
Jβ Jγ y d y (26)
dy L dy L 0

When we take β = αl /L and γ = αk /L, where l ̸= k and αl , αk are roots of J(x), LHS of (24) vanishes, which implies
the integral on the RHS must vanish, proving orthogonality.
Now choose γ = α/L and β = αk /L, where α is not necessarily a root, (26) becomes
α2 − α2k L  y   y 
 ‹ Z
L d J(x)
J α L = J α J αk yd y =⇒
L d x/β x=αk L2 0
L L
 
0
Z L  z }| {
y  y L2  J(α) − J(αk )  αk L 2
J α J αk y d y = J(α)αk J ′ (αk ) =
 α−α
′
 α + α J (αk )
 (27)
0
L L α2−α 2
k k k

| {z }
(∆J/∆α)|αk

(18) is proved by taking the limit α → αk .

6. Bessel Function of the Second Kind
The Bessel equation (13) is a second order ODE, so it should have two linearly independent solutions. Since (13) is
dependent on ν2 , it’s clear that both Jν (x) and J−ν are solutions. It is also easy to see that for non-integer ν, the two
solutions Jν and J−ν are linearly independent. This is because by definition
 x 2q+ν
∞ (−1)q
2
X
Jν (x) =
q=0
q!Γ (q + ν + 1)
 x 2q−ν
∞ (−1)q
2
X
J−ν (x) =
q=0
q!Γ (q − ν + 1)

they cannot be related by a multiplicative factor unless they have the same set of x power terms, i.e., there must be
q, q′ to satisfy
2q + ν = 2q′ − ν =⇒ ν = q′ − q
in other words, ν must be integer for J±ν to be linearly dependent. In fact, due to (6), this condition is also sufficient.
Thus for ν a non-integer, we have two linearly independent solutions Jν (x) and J−ν (x). But the Bessel function of the
second kind is defined as
Jν (x) cos νπ − J−ν (x)
Yν (x) = (ν non-integer) (28)
sin νπ

Note Yν (x), being an linear combination of J±ν (x), is indeed a solution of (13).
Such a definition will yield convenient relations when the order is half integer ν = n + 1/2:

J−(n+1/2) (x) = (−1)n+1 Yn+1/2 (x) (29)

Y−(n+1/2) (x) = (−1) Jn+1/2 (x)
n
(30)

which follow trivially from (28).

In these notes, we will omit the discussion of Yν where ν is integer. Subsequent discussion will focus on spherical
Bessel functions which are defined using half-integer order Bessel functions of both kinds.

7. Spherical Bessel Functions

The radial part of the wave function of a free particle is the solution to the equation (see Sakurai eq (3.281)):

d 2 R 2 dR l(l + 1)
• ˜
+ + 1 − R=0 (31)
dρ 2 ρ dρ ρ2

If we write R(ρ) = ρ −1/2 u(ρ), we will have

1
R′ = − ρ −3/2 u + ρ −1/2 u′ =⇒
2
2 ′
R = −ρ −5/2 u + 2ρ −3/2 u′ (32)
ρ
3 1
R′′ = ρ −5/2 u − 2 · ρ −3/2 u′ + ρ −1/2 u′′ (33)
4 2
l(l + 1)
• ˜
1− R = ρ −1/2 u − l(l + 1)ρ −5/2 u (34)
ρ2

Add (32)-(34):

1
§ • ˜ ª
ρ −1/2 u′′ + ρ −3/2 u′ + ρ −1/2 − l(l + 1) + ρ −5/2 u = 0 (35)
4

Multiply ρ 5/2 to both sides

1 2
  ‹ 
ρ 2 u′′ + ρu′ + ρ 2 − l + u=0 (36)
2
 which is exactly (13) with ν = l + 1/2. This means
1
R(ρ) = p J±(l+1/2) (ρ) (37)
ρ
will be the solutions of (31).
We call
π
s
jl (x) = Jl+1/2 (x) (38)
2x
π π
s s
yl (x) = Yl+1/2 (x) = (−1) l+1
J−(l+1/2) (x) (39)
2x 2x
the Spherical Bessel Functions of order l. They are the two linearly independent solutions to equation (31).

8. Closed Form of Spherical Bessel Functions

Unlike ordinary Bessel functions, spherical Bessel functions can be written in closed forms. First, let’s calculate j0 (x).
 x 2q+1/2
∞ (−1)q
π
s
2
X
j0 (x) =
2x q=0 q!Γ (q + 3/2)
 x 2q
∞ (−1)q
π
s X
= 2
4 q=0 q!Γ (q + 3/2)
 2q 
x
∞ (−1) q
π 4q
s X
=
4 q=0 (2q + 2)! p
q! q+1 π
4 (q + 1)!
∞
1 X (−1)q x 2q 4(q + 1)
=
2 q=0 (2q + 2)!
∞
1 X (−1)q x 2q+1 sin x
= = (40)
x q=0 (2q + 1)! x

1

p
where in the third line, we have used Γ n + 2 = π(2n)!/(4n n!). Similarly, for y0 (x):
 x 2q−1/2
∞ (−1)q
π
s
2
X
y0 (x) = −
2x q=0 q!Γ (q + 1/2)
 x 2q
p X ∞ (−1)q
π 2
=−
x q=0 q!Γ (q + 1/2)
 2q 
x
p X ∞ (−1) q
π 4q
=−
x q=0 (2q)! p
q! q π
4 q!
∞
1 X (−1)q x 2q cos x
=− =− (41)
x q=0 (2q)! x

p
Here we see that the π/2 factor in the definition of spherical Bessel functions is to give j0 and y0 a "normalized"
form.
We can use (9) to continue to obtain j1 (x), y1 (x):
˜ s v
π π 1 π t 2x 1
s s
sin x cos x
• • ˜
j1 (x) = J3/2 (x) = J1/2 (x) − J−1/2 (x) = j0 (x) + y0 (x) = 2 − (42)
2x 2x x 2x π x x x
˜ s v
π π π
s s
1 2x 1 cos x sin x
• t • ˜
y1 (x) = J−3/2 (x) = − J−1/2 (x) − J1/2 (x) = y0 (x) − j0 (x) = − 2 − (43)
2x 2x x 2x π x x x
 The general recurrence relation for jl , yl can be obtained from (9), (38) and (39):

2l + 1 2l + 1
Jl+3/2 = Jl+1/2 − Jl−1/2 =⇒ jl+1 = jl − jl−1 (44)
x x
2l + 1 2l + 1
J−(l+3/2) = − J−(l+1/2) − J−(l−1/2) =⇒ yl+1 = yl − yl−1 (45)
x x

We can then calculate the closed form formula for any spherical Bessel functions jl , yl using (40)-(45).

9. Rayleigh’s Formulae
There is neat differential relation of spherical Bessel functions, called Rayleigh’s Formulae

1 d l sin x 1 d l
 ‹  ‹
jl (x) = (−x)l = (−x)l j0 (x) (46)
x dx x x dx
1 d l cos x 1 d l
 ‹  ‹
yl (x) = (−x)l+1 = (−x)l y0 (x) (47)
x dx x x dx

Proof. Since they are of the same form, we are going to use f l to denote either jl or yl . Now define

f l (x)
g l (x) = (48)
(−x)l

We are going to use induction for the proof, for which the l = 0, 1 cases are trivially true (for both j and y). Now the
goal is to inductively prove

f l+1 (x) 1 d l
 
1 d 1
‹
g l+1 (x) = = f0 (x) = g l′ (x) (49)
(−x) l+1 x dx x dx x

To prove (49), we first establish the differential equation g l (x) must satisfy. Take the first and second derivative of
(48), we have

f l′ = −l(−x)l−1 g l + (−x)l g l′ (50)

′′
f l = l(l − 1)(−x) l−2
g l − 2l(−x)l−1 g l′ + (−x) l
g l′′ (51)

Then (31) becomes

0 = x 2 f l′′ + 2x f l′ + [x 2 − l(l + 1)] f l

= x 2 l(l − 1)(−x)l−2 g l − 2l(−x)l−1 g l′ + (−x)l g l′′ + 2x −l(−x)l−1 g l + (−x)l g l′ + [x 2 − l(l + 1)](−x)l g l
   

= (−x)l+2 g l′′ + (−x)l+1 (−2l − 2)g l′ + (−x)l [l(l − 1) + 2l + x 2 − l(l + 1)]g l

= (−x)l+2 g l′′ − 2(l + 1)(−x)l+1 g l′ + (−x)l+2 g l (52)

After canceling the common factor (−x)l+1 , we get the differential equation satisfied by all g l s:

x g l′′ + 2(l + 1)g l′ + x g l = 0 (53)

To prove (49), by (44) or (45), it’s equivalent to prove

2l + 1 1
(−x)l g l − (−x)l−1 g l−1 = (−x)l+1 g l′ (by canceling (−x)l−1 ) ⇐⇒
x x
1 ′
−(2l + 1)g l − g l−1 = x g l′ (by induction assumption g l = g ) ⇐⇒
x l−1
2l + 1 ′ 1 ′′ 1 ′
 ‹
− g l−1 − g l−1 = x g l−1 − 2 g l−1 ⇐⇒
x x x
′′ ′
x g l−1 + 2l g l−1 + x g l−1 = 0 (54)

which is exactly what we have shown in (53) with l → l − 1.