Lecture 4: Solving differential equations via Greens

functions

Learning outcomes: By the end of this lecture, the students should be able to

solve ordinary differential equations via Greens functions

find the Greens function of partial differential equations by means of Fourier trans-
forms

Using the formalism of distributions, we can solve inhomogeneous linear differential equa-
tions by means of Greens functions.

Theorem (Greens function). Suppose we want to find the solution u the inhomogeneous
ordinary or partial differential equation

Lu = f

with a linear differential operator L and an arbitrary source f for given boundary conditions.
The Greens function G is the solution to the associated differential equation with point-like
inhomogeneity,
LG = ,

and identical boundary conditions. It is alternatively known as the fundamental solution. Us-
ing the Greens function, the solution to the original differential equation can be constructed
as
u = Gf.

Proof: Recalling the properties (f ) = f and f = f , we have

Lu = L(Gf ) = (LG)f = f = f.

Remark: The solution u hence obtained is unique in the space of distributions. To see this,
assume that two functions u1 , u2 are both solutions to Lu = f . Their difference is then

u1 u2 = (u1 u2 ) = (u1 u2 )LG = (Lu1 )G (Lu2 )G = f G f G = 0.

1
I. ORDINARY LINEAR DIFFERENTIAL EQUATIONS

We consider an ordinary linear, inhomogeneous differential equation

L(t)u(t) = f (t)

with boundary conditions

u(0) = u (0) = . . . = u(n2) (0) = 0, u(n1) (0) = 1

where
n
X dk
L(t) = ak (t) with an (t) 1
k=0
dtk
is a (normalised) linear differential operator and f (t) is a driving or source term for u(t). For
these particular boundary conditions, the Greens function can be constructed in a simple
way by using the step distribution.

Theorem (Greens function for ordinary differential equation). The Greens function for an
ordinary linear differential equation

L(t)u(t) = f (t)

with boundary conditions

u(0) = u (0) = . . . = u(n2) (0) = 0, u(n1) (0) = 1

is given by
G(t) = (t)Z(t)

where Z(t) is the solution to the associated homogeneous equation

L(t)Z(t) = 0

with the same boundary conditions

Z(0) = Z (0) = . . . = Z (n2) (0) = 0, Z (n1) (0) = 1.

Proof: Repeated use of the identity

[(t)Z(t)] = (t)Z(t) + (t)Z (t) = (t)Z(0) + (t)Z (t)

2
together with the boundary conditions shows that
n n
X dk X dk
L(t)G(t) = ak (t) k [(t)Z(t)] = (t) ak (t) k Z(t) + an (t)(t)Z (n1) (t)
k=0
dt k=0
dt
= (t)L(t)Z(t) + (t) = (t)

because an (t) = 1 and L(t)Z(t) = 0 by construction.

Example: For the first-order differential equation
d
u(t) + au(t) = f (t) with u(0) = 1,
dt
we note that the homogeneous equation
d
Z(t) + aZ(t) = 0 with Z(0) = 1
dt
is uniquely solved by
Z(t) = eat .

Using the resulting Greens function

G(t) = (t)Z(t) = (t)eat ,

one can construct the solution to the inhomogeneous differential equation as

Z Z t
a(tt )
u(t) = G(t)f (t) =
dt (t t )e
f (t ) = dt ea(tt ) f (t ).

Partial differential equations cannot be solved in this simple way. Instead, we have
to employ Fourier transforms to convert the differential equation into a simple algebraic
equation. This shall be demonstrated in the following two sections.

II. HEAT CONDUCTION EQUATION

The heat conduction equation

U(r, t) U(r, t) = (r, t)
t
describes the evolution of a temperature profile U(r, t) via heat conduction in a material
with thermal diffusivity under the influence of sources (r, t). To solve it for arbitrary
sources, we seek the Greens function as the solution to

G(r, t) G(r, t) = (r)(t).
t

3
Performing Fourier transforms with respect to r by using

F (r) = 1(k), F U(r, t) = (ik)2 F U(r, t) = k 2 U(k, t)

the heat conduction equation takes the form

G(k, t) + k 2 G(k, t) = (t).
t
As shown above, it is solved by

2
G(k, t) = (t)ek t .

Performing the inverse Fourier transform, we thus find the Greens function for the heat
conduction equation,
1 1
Z Z
3 2
G(r, t) = 3
d ke ikr
G(k, t) = (t) 3
d3 k eikrk t
(2) (2)
2
 
(t) r
= exp
(2 t ) 3 4t
which is alternatively known as the heat kernel. Here, we have used
2 
x2i
Z Z    
iki xi ki2 t ixi
dki e = dki exp t ki + exp
2t 4t
2
r  
x
= exp i .
t 4t

III. SCALAR HELMHOLTZ EQUATION

The scalar potential for the electromagnetic field in Lorentz gauge is the solution to the
scalar Helmholtz equation

1 2 (r, t)
(r, t) 2 2
(r, t) = .
c t 0
After carrying out a Fourier transform in the time domain, it can be cast in the form
(r, )
(r, ) + k02 (r, ) =
0
with k0 = /c.
The Greens function for the scalar Helmholtz equation is the solution to

(r, ) + k02 (r, ) = (r).

4
Fourier transformation leads to

k 2 (k, ) + k02 (k, ) = 1,

so that
1
(k, ) = .
k02 k2
The Greens function is hence given by
Z 2 Z Z ikr cos
1 1 2 e
Z
3 ikr
G(r, ) = d k e G(k, ) = d d sin dk k
(2)3 (2)3 0 0 0 k02 k 2
Z 1 Z
1 2 eikrz
= dz dk k
(2)2 1 0 (k0 + i0 + k)(k0 + i0 k)
eikr
Z
1
= dk k
(2)2 ir (k + k0 + i0)(k k0 i0)
ik0 r
e
= .
4r
Here, we have regularized the k-integral by assuming that k0 has a small imaginary part
k0 7 k0 + i0 and have carried it out by closing the integration contour in the upper half
plane.

FURTHER READING

Generalized Functions in Mathematical Physics, V. S. Vladimirov (Mir Publishers,

Moscow, 1979).
Methods of the Theory of Generalized Functions, V. S. Vladimirov (Taylor & Francis,
London, 2002).
Generalized Functions: Theory and Technique, R. P. Kanwal, 2nd ed. (Birkhauser,
Boston, 1998).