1 Bessel Equation and Bessel Functions

In physics, the Bessel equation appears while solving problems with Laplace and Helmhotz
equations in cylindrical and spherical polar coordinates. These problems appear in quan-
tum mechanics, electrodynamics, classical mechanics, heat conduction, vibrations, etc.

1.1 The Bessel Equation

The Bessel equation is
x2 y ′′ + xy ′ + (x2 − ν 2 )y = 0
This is a linear homogeneous equation with non-essential singularity at x = 0. The
Frobenius method provides two solutions (the indicial equation gives s = ±ν),
 
ν Γ(1 + ν)  x 2 Γ(1 + ν)  x 4 Γ(1 + ν)  x 6
y(x) = a0 x 1 − + − + ···
Γ(2 + ν) 2 2! Γ(3 + ν) 2 3! Γ(4 + ν) 2
and an expression by replacing ν by −ν in the above expression. The series is convergent.
If we choose,
1
a0 = ν
2 Γ(1 + ν)
the solution is called the Bessel functions of First Kind or just Bessel function,
∞
X (−1)m  x 2m±ν
J±ν (x) = Bessel function
m=0
m! Γ(m ± ν + 1) 2

In physics, we encounter Bessel equation only for integral or half-integral values of ν.

So we will confine ourselves to the cases when ν is either an integer or a half-integer.

Bessel Fuctions of Second Kind (Neumann Function) are defined as,

cos(νπ)Jν (x) − J−ν (x)
Nν (x) = Neumann function
sin(νπ)

For ν = n (integer) this function is indeterminate, and can be determined by L’Hospital’s

rule. For non-integral values of ν, we have no problem.

Bessel Fuctions of Third Kind (Hankel Functions) are defined as,

Hν(+) (x) = Jν (x) + iNν (x) Hankel function

Hν(−) (x) = Jν (x) − iNν (x) Hankel function

Modified Bessel Fuctions (Hyperbolic Bessel Functions) are defined as,

Iν (x) = i−ν Jν (ix)

π ν+1 (+) π I−ν (x) − Iν (x)
Kν (x) = i Hν (ix) =
2 2 sin(νπ)
All of these Bessel functions appear in physics. The good news is, the properties of all
these functins (specially, the recursion relations) are either identical or can be directly
obtained from those of Jν (x). So we confine ourselves to the study of Jν (x) only.
Important note: In physics problems, normally we come across an equation which is
slightly different (note a2 inside the parentheses),

x2 y ′′ + xy ′ + (a2 x2 − ν 2 )y = 0
(+) (−)
The solutions to this equation are Jν (ax), Nν (ax), Hν (ax) and Hν (ax).

The Bessel equation is not self-adjoint, but can be transformed into a self-adjoint form,

ν2 ν2
   
′′ ′ 2 d h ′i 2
xy + y + a x − y = 0 =⇒ xy + a x− y=0
x dx x

ν2
with p(x) = x, q(x) = − , λ = a2 and w(x) = x. We could have chosen, q(x) = a2 x and
x
λ = −ν 2 instead.

1.2 Limiting Cases

The expressions for Jν (x) for very small x and for very large x (asymptotic limit) can be
derived from the series form. These are given below.
 ν
1  x ν Γ(ν) 2
Small x: Jν (x) ∼ Nν (x) ∼ −
Γ(1 + ν) 2 π x
r    
2 1 π
Large x: Jν (x) ∼ cos x − ν +
πx 2 2
r    
2 1 π
Nν (x) ∼ sin x − ν +
πx 2 2
We can see that,
Jν (0) = δν0 , and lim Jν (x) = 0
x→∞

And there are several zeros for each Bessel functions in between. We will denote the m-th
zero of Jν (x) as ανm . Normally, the first zero (at x = 0) is counted as the 0-th zero. See
Table 11.1 in Arfken for the list of zeros of Jn (x).

1.3 Recurrence Relations

Some well known recursion relations are given below. These are satisfied by the all three
(+) (−)
types of the Bessel functions, Jν , Nν , Hν and Hν . For the modified Bessel functions,
Iν and Kν , these need to be modified slightly.

Proof: Multiply the expression of

d  ν ν Jν (x) by xν and differentiate to get the

RR-1 dx x J ν (x) = x Jν−1 (x)
result.
 Proof: Multiply the expression of
d  −ν
Jν (x) by x−ν and differentiate to get
 −ν
RR-2 dx x J ν (x) = −x Jν+1 (x)
the result.

ν
RR-3 Jν−1 (x) = x Jν (x) + Jν′ (x) Proof: Expand RR-1 to get the result.

ν
RR-4 Jν+1 (x) = x Jν (x) − Jν′ (x) Proof: Expand RR-2 to get the result.

2ν Proof: Add RR-3 and RR-4 to get the

RR-5 Jν+1 (x) + Jν−1 (x) = x Jν (x)
result.

Proof: Subtract RR-4 from RR-3 to

RR-6 Jν−1 (x) − Jν+1 (x) = 2Jν′ (x)
get the result.

Two Important Conclusions:

1. At every extremum x0 (maxima or minima) of Jν (x0 ), Jν−1 (x0 ) = Jν+1 (x0 ).

2. At every positive zero ανn > 0 of Jν (x), Jν−1 (ανn ) = −Jν+1 (ανn ).

In physics, we encounter Bessel functions of either integral order
(ν = n) [in prob-
lems with cylindrical symmetery] or half-integral order ν = n + 12 [in problems with
spherical symmetry]. So now, we will divide our discussions into two distinct cases: one
for ν = n and one for ν = n + 12 . In the later case, the functions are called Spherical
Bessel functions. We will discuss them completely separately. So, in the following, first
we discuss the Bessel functions for ν = n.
 Bessel Functions of Integral Order
The Bessel equation is
x2 y ′′ + xy ′ + (x2 − n2 )y = 0 =⇒ y = Jn (x) or Nn (x)
x2 y ′′ + xy ′ + (a2 x2 − n2 )y = 0 =⇒ y = Jn (ax) or Nn (ax)
The Bessel functions are
∞
X (−1)m  x 2m+n
Jn (x) = n = 0, ±1, ±2, · · · Bessel function
m=0
m! (m + n)! 2

1.4 Recurrence Relations

We have already obtained these for general ν. Just put ν = n.

d  n  n
RR-1 dx x J n (x) = x Jn−1 (x)

d  −n  −n
RR-2 dx x J n (x) = −x Jn+1 (x)

n
RR-3 Jn−1 (x) = x Jn (x) + Jn′ (x)

n
RR-4 Jn+1 (x) = x Jn (x) − Jn′ (x)

2n
RR-5 Jn+1 (x) + Jn−1 (x) = x Jn (x)

RR-6 Jn−1 (x) − Jn+1 (x) = 2Jn′ (x)

1.5 Rodrigue’s formula (Well, not quite!)

For integral values of n, we can write
 n
n 1 d
Jn (x) = x − J0 (x)
x dx
This can be proved by writing RR-2 as
h i  1 d h i
−n
x Jn (x) = − x−(n−1) Jn−1 (x)
x dx
Note that the terms inside the square brackets on two sides have exactly the same form,
and the terms on the right-hand-side are obtained by replacing n by (n − 1). Repeating
this scheme n times gets you the above relation. This is not exactly a Rodrigue’s formula
because there is no simple expression for J0 (x).
1.6 Generating function of Jn (x)
The generating function for the Bessel functions of integral order is
   ∞
x 1 X
Φ(x, t) = exp t− = tn Jn (x), |t| ≤ 1
2 t n=−∞

This can be used to find the values of Jn (x) at certain special values of x. It can also be
used in solving many problems much more easily than can be solved using other methods.

1.7 Special Values and Parity

The generating function is a very useful tool to determine the parity as well as the values
of Jn (x) for special values of x. It can be shown that, for even (odd) n, Jn (x) have even
(odd) parity. The values of Jn (x) at the special values of x are given below.
Jn (−x) = (−1)n Jn (x) Jn′ (−x) = (−1)n+1 Jn′ (x) Jn′′ (−x) = (−1)n Jn′′ (x)
1 1 1 1 1
Jn (0) = δn0 , Jn′ (0) = − δn,1 + δn,−1 Jn′′ (0) = − δn,0 + δn,2 + δn,−2
2 2 2 4 4
lim Jn (x) = 0, lim Jn′ (x) = 0, lim Jn′′ (x) = 0
x→∞ x→∞ x→∞

1.8 Some Interesting Formulae

We can derive some very useful and interesting relations by using the generating function.

Put t = 1, in the generating function to gets,

X∞ ∞ h
X i
n
1= Jn (x) = J0 (x) + 1 + (−1) Jn (x)
n=−∞ n=1

which immediately reduces to

∞
X
J0 (x) + 2 J2n (x) = 1
n=1
If we differentiate the generating function and then put t = 1, we can get the relation
(left for the students to prove)
∞
X
′ ′
J0 (x) + 2 J2n+1 (x) = 0
n=0
iθ
Put t = e and simplify to get
∞ ∞
eix sin θ = J0 (x) + 2i sin[(2n − 1) θ] J2n−1 (x) + 2 cos (2nθ) J2n (x)
X X

n=1 n=1
This immediately gives us
∞
X
cos (x sin θ) = J0 (x) + 2 cos (2nθ) J2n (x)
n=1
and ∞
X
sin (x sin θ) = 2 sin[(2n − 1) θ] J2n−1 (x)
n=1
1.9 Orthogonality of Bessel Functions
Write the Bessel functions for two different orders, n and m,
i  n2

d h ′
x Jn (x) + x − Jn (x) = 0
dx x
i  m2

d h ′
x Jm (x) + x − Jm (x) = 0
dx x
Multiply the first equation by Jm (x) and the second equation by Jn (x), then subtract one
from the other to get
d h i
Jn (x) Jm (x)
Jn (x) x Jm (x) − Jn (x) x Jm (x) = m2 − n2
′ ′
dx
dx x
Integrating from 0 to ∞, the left-hand-side vanisjes. So we get

∞ Jn (x) Jm (x)
Z
2 2
m −n dx = 0
0 x
This implies the orthogonality condition,
Z ∞
Jn (x) Jm (x)
dx = An δmn , Obviously An > 0
0 x
Evaluation of An is delegated to the Exercise.

However, in physics, we rarely come across an integral of this type. What we encounter
is a completely different type of integral involving two Bessel functions of same order
but different arguements. For those integrals, we write the Bessel equation as,
n2
   
d d 2
x Jn (ax) + a x − Jn (ax) = 0
dx dx x
n2
   
d d 2
x Jn (bx) + b x − Jn (bx) = 0
dx dx x
Multiply the first equation by Jn (bx) and the second one by Jn (ax) and subtract to get
 
d d d
Jn (ax) = a2 − b2 Jn (ax) Jn (bx) x


x Jn (ax) Jn (bx) − x Jn (bx)
dx dx dx

Integrating from 0 to 1, we get (Be careful!! dJndx(bx)  = b Jn′ (b), and so on)

x=1
h i Z 1
2 2
Jn′ Jn′


b Jn (a) (b) − a (a) Jn (b) = a − b Jn (ax) Jn (bx) x dx
0

If we let a = αnj and b = αnk , then the left-hand-side vanishes, and we get
Z 1
Jn (αnj x) Jn (αnk x) x dx = Bjn δjk , Obviously Bjn > 0
0

To get the value Bjn (normalisation), we proceed as follows. In the previous equation, we
put a = αnj and keep b as it is to get

1
Z
′ 2 2
−αnj Jn (αnj ) Jn (b) = αnj − b Jn (αnj x) Jn (bx) x dx
0
 1
−αnj Jn′ (αnj ) Jn (b)
Z
=⇒ Jn (αnj x) Jn (bx) x dx = 2


0 αnj − b2
When b → αnj , the left-hand-side is Bjn , but the righ-hand-side is indeterminate. We use
the L’Hospitals’s rule to determine the right-hand-side when b → αnj . We get
1
1h ′
Z i2
Jn (αnj x) Jn (αnj x) x dx = Jn (αnj )
0 2
Use RR-3 and RR-4 to finally write the orthonormality condition as
Z 1
1h ′ i2
Jn (αnj x) Jn (αnk x) x dx = J (αnj ) δjk
0 2 n
1h i2 1h i2
= Jn+1 (αnj ) δjk = Jn−1 (αnj ) δjk
2 2
ρ
Sometimes, the upper limit is other than 1. If we let x = , then we get
a
Z a 
αnj ρ  α ρ
nk a2 h i2 a2 h i2
Jn Jn ρ dρ = Jn+1 (αnj ) δjk = Jn−1 (αnj ) δjk
0 a a 2 2

1.10 Differential Equation with Bessel Function as the Solution:

It can be easily shown that (This appears as an exercise Bessel 39), the solution of the
differential equation

a2 − ν 2 c 2
 
′′ 1 − 2a ′ c−1 2


y + y + bc x + y=0
x x2

has the solution of the form,

y(x) = xa Jν (b xc )
Below, some examples are given.

y ′′ + ω 2 y = 0
y = x1/2 J1/2 (ωx)
=⇒
 
′′ 1/2 2 3/2
y + xy = 0 =⇒ y = x J1/2 x
3
y′ y
y ′′ + + = 0 =⇒ y = J0 2x1/2


x x
 Exercises

Bessel 1 It is claimed that, the following function is the generating function for the Bessel
functions of integral order.
   ∞
x 1 X
Φ(x, t) = exp t− = tn Jn (x)
2 t n=−∞

Prove this statement. Hint: First show that,

2
∂ 2 Φ(x, t)

2 ∂Φ(x, t) 2 ∂
x +x + x Φ(x, t) − t Φ(x, t) = 0
∂x2 ∂x ∂t

Bessel 2 Show that,

∞
X
2
[J0 (x)] + 2 [Jn (x)]2 = 1
n=1

Hint: Use the product of the generating functions, Φ(x, t) · Φ(x, −t).

Then, show that,

1
|J0 (x)| ≤ 1 and |Jn (x)| ≤ √ for n ≥ 1
2

Bessel 3 Put t = ieiθ in the generating function to get

∞
ix cos θ
X
e = J0 (x) + 2 im Jm (x) cos(mθ)
m=1

Then use this relation to show that,

∞
X
cos (x cos θ) = J0 (x) + 2 (−1)m cos (2mθ) J2m (x)
m=1

and ∞
X
sin (x cos θ) = 2 (−1)m+1 cos[(2m − 1) θ] J2m−1 (x)
m=1

Bessel 4 Show that, the Bessel functions of integral order can be written as the following
integrals.
Z 2π
1
(a) Jn (x) = ei(x sin θ−nθ) dθ
2π 0
i−n 2π i(x cos θ−nθ)
Z
(b) Jn (x) = e dθ
2π 0
Bessel 5 Show that, the Bessel function of integral order can be written as the following
integral
1 π
Z
Jn (x) = cos (nθ − x sin θ) dθ
π 0
Hint: Use the expressions for cos(x sin θ) and sin(x sin θ) obtained from the gener-
ating function. Or alternatively, use the results of problem Bessel 4 and compare
the real and imaginaty parts on both sides.

Use part (b) of problem Bessel 4 to show that, Jn (x) can also be written as,
(−1)n 2π
Z
(a) J2n (x) = cos [x cos θ − 2nθ] dθ
2π 0

2π
(−1)n
Z
J2n+1 (x) = sin [x cos θ − (2n + 1)θ] dθ
2π 0

Then show that

1 ∞
2 cos(xt) 2 sin(xt)
Z Z
(b) J0 (x) = √ dt = √ dt
π 0 1 − t2 π 1 t2 − 1

Bessel 6 Show that

π/2
sin x
Z
(a) J0 (x cos θ) cos θ dθ =
0 x
π/2
1 − cos x
Z
(b) J1 (x cos θ) dθ =
0 x
Hint: Use the series forms of J0 and J1 and then integrate.

Bessel 7 Show that,

∞
1
Z
x−n Jn+1 (x) dx = n≥0
0 2n n!
Bessel 8 Show that,
∞
1
Z
J0 (ax) e−px dx = p
0 p 2 + a2
Hint: Write the Bessel equation for n = 0 as
x2 y ′′ (x) + xy ′ (x) + a2 x2 y(x) = 0 with y(x) = J0 (ax)
So, we have y(0) = J0 (0) = 1 and y ′ (0) = −J1 (0) = 0. Use the method of Laplace
transform for ODEs to show that the Laplace transform of y(x) = J0 (ax) is
Z ∞
1
Y (p) = J0 (ax) e−px dx = p
0 p + a2
2
 Bessel 9 (a) Show that,
Z ∞ Z ∞ Z ∞ Z ∞
J0 (x) dx = J2 (x) dx = J4 (x) dx = · · · = J2n (x) dx
0 0 0 0

and
Z ∞ Z ∞ Z ∞ Z ∞
J1 (x) dx = J3 (x) dx = J5 (x) dx = · · · = J2n+1 (x) dx
0 0 0 0

(b) Show that, Z ∞

J1 (x) dx = 1
0
and use problem Bessel 8 to show that,
Z ∞
J0 (x) dx = 1
0

Note: When parts (a) and (b) are used together, they prove that
Z ∞
Jn (x) dx = 1 n = 0, 1, 2, 3, · · ·
0

Bessel 10 Prove the identity

′ 2
Jν (x) J−ν (x) − Jν′ (x) J−ν (x) = − sin(νπ)
πx
Then show that,
2
Jν (x) Nν′ (x) − Jν′ (x) Nν (x) =
πx
Bessel 11 Show that,
2
Jν (x) Nν+1 (x) − Jν+1 (x) Nν (x) = −
πx
Bessel 12 Show that,
∞
Jn (x) Jm (x) 1
Z
dx = δnm , n, m > 0
0 x 2n

Bessel 13 We denote βnj as the j-th zero of the first derivative of Jn (x), i.e. Jn′ (βnj ) = 0.
Then show that,
Z a 
a2 n2 h
 
ρ  ρ i2
Jn βnj Jn βnk ρ dρ = 1− 2 Jn (βnj ) δjk
0 a a 2 βnj

Bessel 14 Show that,

Z α0n  
Z α0n
x 1
1− J0 (x) x dx = J0 (x) dx
α0n α0n
0 0
Bessel 15 In nuclear scattering theory, we come across a quantity called the scattering ampli-
tude which is a function of the scattering angle θ and is given by
ik 2π
Z Z R
f (θ) = − dφ eikr sin θ sin φ r dr
2π 0 0

where R is the nuclear radius and k is the momentum of the incident particles. The
differential scattering crossection is given by
dσ
= |f (θ)|2
dΩ
Show that,  2
dσ R J1 (kR sin θ)
=
dΩ sin θ
Bessel 16 The transmission coefficient for a circular aperture for normal incidence of light rays
is given by Z 2ka Z 2ka
J2 (x) 1
T =2 dx − J2 (x) dx
0 x 2ka 0
Show that, T can be written as
∞ 2ka
1 X 1
Z
(a) T =1− J2n+1 (2ka) (b) T =1− J0 (x) dx
ka n=0 2ka 0

Bessel 17 In the theory of Fraunhoffer diffraction through a circular aperture, we encounter

an integral of the type’
Z a Z 2π
Φ= r dr eibr cos θ dθ
0 0

2πa
Show that, Φ= J1 (ba).
b

Bessel 18 A function f (x) is expressed as Bessel series:

∞
X
f (x) = ak Jn (αnk x) , 0≤x≤1
k=1

(a) Show that,

1
2
Z
ak = f (x) Jn (αnk x) x dx
[Jn+1 (αnk )]2 0

(b) Show that (This is called the Parseval relation),

Z 1 ∞
2 1X 2
[f (x)] x dx = ak [Jn+1 (αnk )]2
0 2 k=1

(c) Then show that,

∞
X 1 1
2
=
α
k=1 nk
4(n + 1)
Hint: Expnand xn in a Bessel series and use Parseval relation.
2 Spherical Bessel Functions
These are related to the Bessel functions of half-integral order and appear in physics
problems as the solutions of the Bessel’s Equation in quantum mechanics, electrodynam-
ics, classical mechanics, etc.

2.1 The Spherical Bessel Equation

The spherical Bessel equation is
h i
x2 y ′′ + 2xy ′ + x2 − ℓ(ℓ + 1) y = 0 ℓ = 0, 1, 2, · · ·

This is a self-adjoint linear homogeneous equation with non-essential singularity at x = 0,

because it can be written as
  h
d 2 dy
i
x + x2 − ℓ(ℓ + 1) y = 0
dx dx
Comparing with the general form of the self-adjoint equation, we see that (well, at this
point) p(x) = x2 , q(x) = x2 , w(x) = 1, and λ = ℓ(ℓ + 1).

However, the most important point to note is that the solution of the above equation can
be written as
1
yℓ (x) = 1/2 Jℓ+ 1 (x)
x 2

This is left as an exercise for the students (Hint: Use the equation in Section 4.10). So
there is no need to derive the solution using the Frobenius method nor there is any need
to derive the recurrence relations. Those can be obtained from the original relations by
replacing ν = ℓ + 12 and taking care of the extra term 1/x1/2 . If we multiply yℓ (x) by
p
2/π (this is the convention), we get the spherical Bessel functions:
r
π
jℓ (x) = J 1 (x) Spherical Bessel function
2x ℓ+ 2
r
π
nℓ (x) = N 1 (x) Spherical Neumann function
2x ℓ+ 2
r
(±) π (±)
hℓ (x) = jℓ (x) ± i nℓ (x) = Hℓ+ 1 (x) Spherical Hankel functions
2x 2

Important note: In physics, normally we come across an equation which is slightly

different (note k 2 inside the parentheses),
  h
d 2 dR(r) i
r + k 2 r2 − ℓ(ℓ + 1) R(r) = 0
dr dr
(±)
The solution to this equation is one of jℓ (kr), nℓ (kr) and hℓ (kr).

Note that, the Spherical Bessel equation in the above written form is self-adjoint, with
with p(r) = r2 , q(r) = −ℓ(ℓ+1), λ = k 2 and w(r) = r2 . We could have chosen, q(r) = k 2 r2
and λ = −ℓ(ℓ + 1) instead. The specific choice depends upon the specific problems at
hand. We will come to this point when we actually face a problem in physics.
2.2 Special Values
The values of the spherical Bessel functions can be obtained from those of the Bessel
functions. The question of parity of the spherical Bessel functions is a case of formal
interest only for physicists, because we do not need it.

jℓ (−x) = (−1)ℓ jℓ (x) jℓ′ (−x) = (−1)ℓ+1 jℓ′ (x)

1 (±)
jℓ (0) = δℓ0 , jℓ′ (0) = δℓ,1 nℓ (0) = −∞ hℓ (0) = ∓i∞
3
(±)
lim jℓ (x) = 0, lim nℓ (x) = 0 lim hℓ (x) = 0
x→∞ x→∞ x→∞

xℓ −(2ℓ − 1)!! (±)

Small x: jℓ (x) ∼ nℓ (x) ∼ ℓ+1
hℓ (x) ≈ ±inℓ (x)
(2ℓ + 1)!! x
 
1 ℓπ
Large x: jℓ (x) ∼ sin x −
x 2
(∓i)ℓ+1 ±ix
 
−1 ℓπ (±)
nℓ (x) ∼ cos x − hℓ (x) = e
x 2 x
We will denote the k-th zero of jℓ (x) as γℓk . Normally, the first zero (at x = 0) is counted
as the 0-th zero.

2.3 Recurrence Relations

p
These can be obtained by writing jℓ (x) = ( 2/πx) Jℓ+ 1 (x) in the original six recurrence
2
relations.

d  ℓ+1  ℓ+1 d  −ℓ  −ℓ
RR-1 dx x j ℓ (x) = x jℓ−1 (x) RR-2 dx x jℓ (x) = −x jℓ+1 (x)

ℓ+1 ℓ
RR-3 jℓ−1 (x) = x jℓ (x) + jℓ′ (x) RR-4 jℓ+1 (x) = x jℓ (x) − jℓ′ (x)

2ℓ+1 1
RR-5 jℓ+1 (x) + jℓ−1 (x) = RR-6 jℓ−1 (x) − jℓ+1 (x) = x jℓ (x) + 2jℓ′ (x)
x jℓ (x)

We note that, at every positive zero γℓk > 0 of jℓ (x), jℓ+1 (γℓk ) = −jℓ−1 (γℓk ).

2.4 The Rayleigh Formulae

The spherical Bessel functions can be written as (these are called Rayleigh formulae),
 ℓ    ℓ 
ℓ 1 d sin x ℓ 1 d cos x 
jℓ (x) = x − nℓ (x) = −x −
x dx x x dx x
The proof is very easy. Use RR-2 to write
h i  1 d h i
−ℓ −(ℓ−1)
x jℓ (x) = − x jℓ−1 (x)
x dx
Note that the square brackets on the left-hand side and the right-hand-side contain iden-
tical expressions if ℓ is replaced by (ℓ − 1). So, going down the ladder ℓ times, we get
h i  1 d ℓ
−ℓ
x jℓ (x) = − j0 (x)
x dx
Then solve the spherical Bessel equation for ℓ = 0 to get j0 (x).

So, the spherical Bessel functions can be expressed entirely in terms of sin x, cos x and
powers of (1/x).

2.5 Orthogonality of Spherical Bessel Functions

Write the spherical Bessel equation for two different indices, ℓ and m,
d h 2 ′ i  2 
x jℓ (x) + x − ℓ (ℓ + 1) jℓ (x) = 0
dx
d h 2 ′ i 
x jm (x) + x2 − m (m + 1) jm (x) = 0

dx
Multiply the first equation by jm (x) and the second equation by jℓ (x), then subtract one
from the other to get
d h i h i
jm (x) x2 jℓ′ (x) − jℓ (x) x2 jm
′
(x) = ℓ(ℓ + 1) − m(m + 1) jℓ (x) jm (x)
dx
Integrate the above expression from 0 to ∞, and get the orthogonality condition
Z ∞
jℓ (x) jm (x) dx = Aℓ δℓm
0

Determination of Aℓ of left to the students as an exercise.

Sometimes, we face integration of product of two spherical Bessel functions of same

index ℓ, but with different arguements. For those integrals, we write the spherical Bessel
equation as,   h
d 2 d
i
x jℓ (ax) + a2 x2 − ℓ(ℓ + 1) jℓ (ax) = 0
dx dx
  h
d 2 d
i
x jℓ (bx) + b2 x2 − ℓ(ℓ + 1) jℓ (bx) = 0
dx dx
Multiply the first equation by jℓ (bx) and the second one by jℓ (ax) and subtract to get
 
d 2 d 2 d
jℓ (ax) = a2 − b2 jℓ (ax) jℓ (bx) x2


x jℓ (ax) jℓ (bx) − x jℓ (bx)
dx dx dx

Integrating from 0 to 1, we get (Be careful!! djℓdx
(bx) 
 = b jℓ′ (b), and so on)
x=1
h i Z 1
2 2
jℓ′ jℓ′ jℓ (ax) jℓ (bx) x2 dx


b jℓ (a) (b) − a (a) jℓ (b) = a − b
0

If we let a = γℓi and b = γℓk , then the left-hand-side vanishes, and we get
Z 1
jℓ (γℓi x) jℓ (γℓk x) x2 dx = Bkℓ δik , Obviously Bkℓ > 0
0
To get the value Bkℓ (normalisation), we proceed as follows. In the previous equation, we
put a = γℓi and keep b as it is to get

1
Z
2 2
′
−γℓi jℓ (γℓi ) jℓ (b) = γℓi − b jℓ (γℓi x) jℓ (bx) x2 dx
0

1
−γℓi jℓ′ (γℓi ) jℓ (b)
Z
=⇒ jℓ (γℓi x) jℓ (bx) x2 dx =
0 (γℓi2 − b2 )
When b → γℓi , the left-hand-side is Bjn , but the righ-hand-side is indeterminate. We use
the L’Hospitals’s rule to determine the right-hand-side when b → γℓi . We get
Z 1
1h ′ i2
jℓ (γℓi x) jℓ (γℓi x) x2 dx = jℓ (γℓi )
0 2
Use RR-3 and RR-4 to finally write the orthonormality condition as
Z 1
1h ′ i2
jℓ (γℓi x) jℓ (γℓk x) x2 dx = jℓ (γℓi ) δik
0 2
1h i2 1h i2
= jℓ+1 (γℓi ) δik = jℓ−1 (γℓi ) δik
2 2
r
Sometimes, the upper limit is other than 1. If we let x = , then we get
a
Z a 
γℓi r   γℓk r  2 a3 h i2 a3 h i2
jℓ jℓ r dr = jℓ+1 (γℓi ) δik = jℓ−1 (γℓi ) δik
0 a a 2 2

2.6 The Rayleigh Equation

The spherical Bessel functions and the Legendre polynomials have a special relationship.
This is expressed as the Raleigh equation,
∞
X
ixω
e = iℓ (2ℓ + 1) jℓ (x) Pℓ (ω), − ∞ < x < ∞, −1≤ω ≤1
ℓ=0

This is very useful in many physics problems, most importantly in scattering theory. We
will prove this in the following.
Proof: Any function in [−1, 1] can be expanded in a Legendre series. So we let
∞ 1
2ℓ + 1
Z
eixω = eixω Pℓ (ω) dω
X
aℓ (x) Pℓ (ω) =⇒ aℓ (x) =
ℓ=0
2 −1

Writing the series expansion of the exponential, we get

∞
2ℓ + 1 X im xm 1 m
Z
aℓ (x) = ω Pℓ (ω) dω
2 m=0 m! −1

The integral is non-zero only when m ≥ ℓ. So the sum actually starts at m = ℓ:

∞ 1 ∞ 1
2ℓ + 1 X im xm 2ℓ + 1 X in+ℓ xn+ℓ
Z Z
m [n=(m−ℓ)]
aℓ (x) = ω Pℓ (ω) dω ======⇒ ω n+ℓ Pℓ (ω) dω
2 m=ℓ m! −1 2 n=0 (n + ℓ)! −1
The integral is non-zero, only when n + 2ℓ = even =⇒ n is even. So we let n = 2m. Then
∞
2ℓ + 1 X i2m+ℓ x2m+ℓ 1 2m+ℓ
Z
aℓ (x) = ω Pℓ (ω) dω
2 m=0 (2m + ℓ)! −1

Use Legendre 18 to get the integral, and simplify

∞
2ℓ + 1 ℓ X (−1)m x2m+ℓ 2 (2m + ℓ)!
aℓ (x) = i
2 m=0
(2m + ℓ)! (2m + 2ℓ + 1)!! (2m)!!

∞
ℓ (−1)m x2m+ℓ
X
=⇒ aℓ (x) = i (2ℓ + 1) =⇒ aℓ (x) = iℓ (2ℓ + 1) jℓ (x)
m=0
(2m + 2ℓ + 1)!! (2m)!!
This proves the Raleigh equation.

When we put x = kr and ω = cos θ, the left-hand side represents a plane wave with θ
being the angle between the wave-vector ~k and the position vector ~r . This relation is
then called the Rayleigh equation and is very useful in scattering theory.

The Rayleigh equation implies a direct relationship between Pℓ (ω) and jℓ (x), which are
Z 1 Z ∞
1 ixω 1
jℓ (x) = ℓ e Pℓ (ω) dω Pℓ (ω) = ℓ eixω jℓ (x) dx
2i −1 πi −∞

So, jℓ (x) and Pℓ (ω) are Fourier transforms of each other (apart from the constants).

If we put ω = 1 in the Rayleigh equation we get,

∞
X
ix
e = iℓ (2ℓ + 1) jℓ (x), −∞<x<∞
ℓ=0

The real and imaginary parts give us

∞
X ∞
X
cos x = (−1)ℓ (4ℓ + 1) j2ℓ (x) sin x = (−1)ℓ (4ℓ + 3) j2ℓ+1 (x)
ℓ=0 ℓ=0

′
Multiply eixω and e−ix ω and integrate over ω from −1 to 1 to get
∞
X
(2ℓ + 1) jℓ (x) jℓ (x′ ) = π δ(x − x′ ) [Closure Relation for jℓ (x)]
ℓ=0

Try finding a similar Closure Relation for Pℓ (ω),

∞
X
(2ℓ + 1) Pℓ (ω) Pℓ (ω ′ ) = 2 δ(ω − ω ′ )
ℓ=0
 Exercises
Sph-Bessel 1 The spherical Bessel functions are related to the corresponding Bessel functions as
follows: r r
π π
jℓ (x) = Jℓ+ 1 (x) and nℓ (x) = N 1 (x)
2x 2 2x ℓ+ 2
Using the series form of Jν (x) and Nν (x), show that the series forms of jℓ (x) and
nℓ (x) are
∞ ∞
X (−1)m x2m+ℓ X (−1)m+ℓ+1 x2m−ℓ−1
jℓ (x) = nℓ (x) =
m=0
(2m)!! (2m + 2ℓ + 1)!! m=0
(2m)!! (2m − 2ℓ − 1)!!

Sph-Bessel 2 Use the series form of jℓ (x) to show that,

dℓ
 
ℓ!
jℓ (x) =
dxℓ x=0 (2ℓ + 1)!!

Sph-Bessel 3 Prove the identity

1
jℓ (x) n′ℓ (x) − jℓ′ (x) nℓ (x) =
x2
Sph-Bessel 4 Show that,
1
jℓ (x) nℓ+1 (x) − jℓ+1 (x) nℓ (x) = −
x2
Sph-Bessel 5 Show that,
∞
π
Z
jℓ (x) jm (x) dx = δℓm
0 2(2ℓ + 1)
Hint: Use the Raleigh equation.

Sph-Bessel 6 Show that,

∞  2
jℓ (x) π
Z
dx = ℓ>0
0 x (2ℓ − 1)(2ℓ + 1)(2ℓ + 3)

Hint: Square RR-5 and integrate from 0 to ∞ using Sph-Bessel 5.

Sph-Bessel 7 Show that,

∞
[jℓ (x)]2 1
Z
(a) dx = , ℓ>0
0 x 2ℓ(ℓ + 1)

∞
jℓ (x) jℓ+1 (x) π
Z
(b) dx =
0 x 2(2ℓ + 1)(2ℓ + 3)

∞
jℓ (x) jℓ+2 (x) 1
Z
(c) dx =
0 x 6(ℓ + 1)(ℓ + 2)
 Sph-Bessel 8 Show that,
∞
1
Z
jℓ (x) jℓ′ (x) dx = − δℓ0
0 2
Sph-Bessel 9 Show that,
Z ∞ Z ∞
(ℓ + 1) jℓ+1 (x) dx = ℓ jℓ−1 (x) dx, ℓ>0
0 0

Hint: Integrate RR-3 and RR-4 and manipulate or first show that,

(2ℓ + 1)jℓ′ (x) = ℓ jℓ−1 (x) − (ℓ + 1) jℓ+1 (x)

Sph-Bessel 10 Show that,

∞ ∞
(2ℓ − 1)!!  π  (2ℓ)!!
Z Z
j2ℓ (x) dx = and j2ℓ+1 (x) dx =
0 (2ℓ)!! 2 0 (2ℓ + 1)!!

Hint: Use problem Sph-Bessel 9 to show that

Z ∞
(2ℓ − 1)!! ∞
Z
j2ℓ (x) dx = j0 (x) dx
0 (2ℓ)!! 0

and ∞ ∞
(2ℓ)!!
Z Z
j2ℓ+1 (x) dx = j1 (x) dx
0 (2ℓ + 1)!! 0
∞ Z ∞
π
Z
Then show that, j0 (x) dx = and j1 (x) dx = 1.
0 2 0
(Use the complex contour integration method for the first integration, see Arfken,
page 468, and use one of the recurrence relations for the second integration).

Sph-Bessel 11 Show that,

∞
1
Z
x−ℓ jℓ+1 (x) dx =
0 (2ℓ + 1)!!
Sph-Bessel 12 Show that,
∞
X h i2
(2ℓ + 1) jℓ (x) =1
ℓ=0

−1 1
This relation then implies that, √ ≤ jℓ (x) ≤ √ .
2ℓ + 1 2ℓ + 1