23 The Green Function

Contents
23.1 The Green function

23.2 The Hankel Transform

Keywords: Green function, Hankel transform

Ref: I. N. Sneddon: Elements of Partial Differential Equations; R. Courant

and D. Hilbert: Methods of Mathematical Physics; J.D. Jackson: Classical

Electrodynamics.

23.1 The Green function

The next problem we take up is the wave equation with a known source term

f (r, t),

2 1 ∂ 2 ψ(r, t)
∇ ψ(r, t) − 2 = −4πf (r, t). (23.1)
c ∂t2

One would now like to solve for ψ. We first define the transforms,
Z ∞
1
ψ(r, t) = √ φ(r, ω)e−iωt dω,
2π Z−∞
∞
1
f (r, t) = √ g(r, ω)e−iωt dω. (23.2)
2π −∞
2 23 The Green Function

Substituting the transforms (23.2) in (23.1), we have a Helmholtz equation

with a source term.

ω2
∇2 φ(r, ω) + φ(r, ω) = −4πg(r, ω). (23.3)
c2
To solve the equation (23.3) one first finds the Green function. The Green

function, G(r − r′ , ω) for the Helmholtz operator is defined as,

ω2
∇ G(r − r , ω) + 2 G(r − r′ , ω) = −δ 3 (r − r′ ),
2 ′
(23.4)
c
by replacing the source term of the equation (23.3) by a three dimensional

delta function. Next we define a three dimensional Fourier transform for the

Green function as the following,

∞
1
Z
′
′
G(r − r , ω) = G̃(k, ω)eik·(r−r ) d3 k. (23.5)
(2π)3/2 −∞

Substituting (23.5), in (23.4), we have

2 ω2 1
(−k + 2 )G̃(k, ω) = − (23.6)
c (2π)3/2
1
R ∞ ik·(r−r′ ) 3
since, δ 3 (r − r′ ) = (2π) 3 −∞ e d k. The Green function is now
Z ∞
′ 1 1 ik·(r−r′ ) 3
G(r − r , ω) = e d k. (23.7)
(2π)3 −∞ k 2 − (ω/c)2
Now we have to perform a 3-dimensional k integral. We choose the z-axis of

the k-space along the the vector, r − r′ . So,

Z ∞ Z 1 Z 2π
1 1 ik|r−r′ | cos θ
G(r − r′ , ω) = k 2
dk d cos θ dφ e
(2π)3 0 −1 0 k 2 − (ω/c)2
(23.8)

NPTEL Course: Wave Propagation in Continuous Media

23.1 The Green function 3

and the angular integrals can be quickly performed to give the result,
Z ∞
1 1
G(r − r′ , ω) = 2
kdk 2 2
(eikr − e−ikr )
(2π) ir 0 k − (ω/c)
Z ∞
1 1
= 2 kdk 2 2
(eikr − e−ikr ), (23.9)
8π ir −∞ k − (ω/c)

where |r − r′ | = r. The integral, (36.18), is evaluated with a contour inte-

gration in complex k-plane with Feynman technique. In this method first

the poles of the integrand are slightly displaced from real line with a small

imaginary term and after the evaluation of the integral this terms is set to

zero. We elaborate the method in the following. We define,

∞
1 1
Z
± ′
G (r − r , ω) = 2 kdk 2 2
(eikr − e−ikr ), (23.10)
8π ir −∞ k − (ω/c) ∓ iǫ

where ǫ is a very small quantity. Let us now evaluate G− (r − r′ , ω). The

ǫc
poles of the integrand in this case are at k = ±( ωc − i 2ω ). Now the integral

(23.10) can be written in two parts as the following.

Z ∞
− ′ 1 1 ikr
G (r − r , ω) = 2 kdk 2 ω ǫc 2 e
8π ir −∞ k − ( c − i 2ω )
Z ∞
1 1 −ikr
− 2 kdk 2 ω ǫc 2 e . (23.11)
8π ir −∞ k − ( c − i 2ω )

In the above integral, (23.11), the for the first term a semicircular contour

is chosen in the upper half plane with a large radius so that the integrand

vanishes on the semicircular contour and for the second integral contour is

chosen in the lower half plane. Both the contours are shown in fig. 23.1.

NPTEL Course: Wave Propagation in Continuous Media

4 23 The Green Function

k
k

C1

✘ ✘
ω/c ω/c
ω/c ω/c
✘ ✘

C2

Fig. 23.1: Contours and poles for the Green function integral

1 1
Z
G− (r − r′ , ω) = 2 kdk 2 ω ǫc 2 e
ikr
8π ir C1 k − ( c − i 2ω )
1 1
Z
−ikr
− 2 kdk 2 ω ǫc 2 e . (23.12)
8π ir C2 k − ( c − i 2ω )
The first contour captures the pole, k = − ωc + i 2ω
ǫc
, whereas the second one
ω ǫc
includes the pole, k = c − i 2ω . Now using Cauchy residue theorem it is

evaluate these integrals,

− ′ 1  − ωc −i ω r  1  ωc −i ω r 
G (r − r , ω) = 2πi 2 e c − (−2πi) 2 2ω e
c
8π ir − 2ω c
8π ir c
1 ω ′
= ′
e−i c |r−r | . (23.13)
4π|r − r |
Similarly one obtains

1 i ωc |r−r′ |
G+ (r − r′ , ω) = e . (23.14)
4π|r − r′ |
G+ is known as the retarded Green function and G− is known as the advanced

Green function.

Problem 1: Evaluate G+ explicitly to obtain the result given in (23.14).

NPTEL Course: Wave Propagation in Continuous Media

23.2 The Hankel Transform 5

Now the solution φ(r, ω) of the equation (23.3) is written as

1
Z
i ωc |r−r′ |
φ(r, ω) = dr′ 4πg(r′ , ω) e . (23.15)
4π|r − r′ |

and hence the required solution ψ(r, t) of the original equation is given by
Z ∞Z
1 ′ ′ 1 iω( |r−r
′|
−t)
ψ(r, t) = √ dr g(r , ω) ′
e c dω (23.16)
2π −∞ |r − r |

Now with the inverse Fourier transform the integral over ω can be performed

to obtain,
|r − r′ | 1
Z
′ ′
ψ(r, t) = dr f (r , t − ) . (23.17)
c |r − r′ |
Notice that in the solution (23.17) in the argument of the source function
′
f (r′ , t) the time t is replaced by t − |r−r |
c , which is called the ‘retarded time’.

We can interpret this as the following. The signal emitted at r′ reaches the
|r−r′ |
observation point, r, after a finite time, c , where c is the speed of the

signal propagation.

23.2 The Hankel Transform

In our next example we introduce the Hankel transform. Suppose we have to

find the solution of the equation,

∂ 2 φ 1 ∂φ ∂ 2 φ
+ + =0 (23.18)
∂r2 r ∂r ∂z 2

in the region r ≥ 0, z ≥ 0 with the boundary condition φ(∞, z) = φ(r, ∞) = 0

and φ(r, 0) = χ(r).

NPTEL Course: Wave Propagation in Continuous Media

6 23 The Green Function

We introduce the the Bessel function of zero order, J0 (kr), which satisfies the

Bessel differential equation,

∂ 2 J0 (kr) 1 ∂J0 (kr)

2
+ + k 2 J0 (kr) = 0. (23.19)
∂r r ∂r

Multiply the equation (23.18) with rJ0 (kr) and integrate with respect to r

from zero to infinity.

Z ∞ Z ∞ Z ∞
∂ 2φ 1 ∂φ ∂ 2φ
rJ0 (kr) 2 dr + rJ0 (kr) dr + rJ0 (kr) 2 dr = 0 (23.20)
0 ∂r 0 r ∂r 0 ∂z

Now we integrate by parts,

∞ Z ∞
∂φ  ∂(rJ0 (kr)) ∂φ
rJ0 (kr)  − dr
∂r 0 0 ∂r ∂r
Z ∞ Z ∞
∂φ ∂2
+ J0 (kr) dr + 2 rJ0 (kr)φdr = 0. (23.21)
0 ∂r ∂z 0

The first term vanishes by putting the limits and using the boundary condi-

tions. We are left with,

Z ∞
∂J0 (kr) ∂φ ∂ 2 ψ(k, z)
− r dr + = 0. (23.22)
0 ∂r ∂r ∂z 2

Where we have defined the Hankel transform, ψ(k, z), as,

Z ∞
ψ(k, z) = rJ0 (kr)φ(r, z)dr. (23.23)
0

Equation (23.22) is again integrated by parts to give (using the boundary

conditions),
∞
∂ 2 J0 (kr) 1 ∂J0 (kr)
d2 ψ(k, z)
Z
r + φdr + . (23.24)
0 ∂r2 r ∂r dz 2

NPTEL Course: Wave Propagation in Continuous Media

23.2 The Hankel Transform 7

This can be further be simplified using (23.19)

d2 ψ(k, z)
2
− k 2 ψ(k, z) = 0. (23.25)
dz

The solution of (23.25) is simple, and using the boundary condition, we ob-

tain,

ψ(k, z) = ψ(k, 0)e−kz , (23.26)

where,
Z ∞
ψ(k, 0) = rJ0 (kr)χ(r)dr (23.27)
0

The solution is of φ(r, z) is now obtained using the inverse Hankel transform,
Z ∞ Z ∞
φ(r, z) = kJ0 (kr)ψ(k, z)dk = kJ0 (kr)ψ(k, 0)e−kz dk. (23.28)
0 0

Problem 2: Show that the inverse Hankel transform is indeed,

Z ∞
φ(r, z) = kJ0 (kr)ψ(k, z)dk.
0

[ You may use, the following orthogonality relation of Bessel functions.]

∞
1
Z
rJ0 (kr)J0 (k ′ r)dr = δ(k − k ′ ).
0 k

NPTEL Course: Wave Propagation in Continuous Media