Chapter 9: Green’s functions for time-independent problems

Introductory examples

One-dimensional heat equation

Consider the one-dimensional heat equation

∂u ∂2u
=k 2 (1)
∂t ∂x
with boundary conditions
u(0, t) = 0 (2)
u(L, t) = 0 (3)
and initial condition
u(x, 0) = f (x) (4)
We already know that the solution of this problem is given by
∞
X nπx −k(nπ/L)2 t
u(x, t) = an sin e (5)
n=1
L

where an are the Fourier sine series coefficients of f (x).

∞
X nπx
f (x) = an sin
n=1
L
Z L
2 nπx
an = f (x) sin dx (6)
L 0 L
After replacing (6) in (5) we may express the solution as
∞
" Z #
X 2 L nπx0 nπx −k(nπ/L)2 t
u(x, t) = f (x0 ) sin dx0 sin e
n=1
L 0 L L
R P
which may be written ( after interchanging and )
∞
Z L !
X 2 nπx0 nπx −k(nπ/L)2 t
u(x, t) = f (x0 ) sin sin e dx0 (7)
0 n=1
L L L
We define the quantity
∞
X 2 nπx0 nπx −k(nπ/L)2 t
sin sin e
n=1
L L L
as the influence function for the initial condition. For every point x0 this quantity shows the influence of
the initial temperature at x0 on the temperature at position x and time t.
Further insight may be obtained by considering the heat equation with sources

∂u ∂2u
= k 2 + Q(x, t) (8)
∂t ∂x
with boundary conditions (2-3) and the initial condition (4).

1
Q: Show that the solution of (8), (2-3), (4) may be expressed as
Z L Z L Z t
u(x, t) = f (x0 )G(x, t; x0 , 0)dx0 + Q(x0 , t0 )G(x, t; x0 , t0 )dt0 dx0 (9)
0 0 0

where G(x, t; x0 , t0 ) is given by

∞
X 2 nπx0 nπx −k(nπ/L)2 (t−t0 )
G(x, t; x0 , t0 ) = sin sin e (10)
n=1
L L L

The function G(x, t; x0 , t0 ) defined by (10) is called the Green’s function for the heat equation problem
(8), (2-3), (4).
At t0 = 0, G(x, t; x0 , t0 ) expresses the influence of the initial temperature at x0 on the temperature at
position x and time t. In addition, G(x, t; x0 , t0 ) shows the influence of the source/sink term Q(x0 , t0 ) at
position x0 and time t0 on the temperature at position x and time t.

Notice that the Green’s function depends only on the elapsed time t − t0 since

G(x, t; x0 , t0 ) = G(x, t − t0 ; x0 , 0)

Green’s functions for boundary value problems for ODE’s

In this section we investigate the Green’s function for a Sturm-Liouville nonhomogeneous ODE

L(u) = f (x)

subject to two homogeneous boundary conditions.

The simplest example is the steady-state heat equation

d2 x
= f (x)
dx2
with homogeneous boundary conditions

u(0) = 0, u(L) = 0

The method of variation of parameters

Consider the linear nonhomogeneous problem

L(u) = f (x) (11)

where u = u(x), a < x < b satisfies homogeneous boundary conditions and L is the Sturm-Liouville operator
 
d d
L≡ p +q
dx dx

If u1 and u2 are two linearly independent solutions of the homogeneous problem L(u) = 0, the general
solution of the homogeneous problem is
u = c1 u1 + c2 u2
where c1 and c2 are arbitrary constants. To solve the nonhomogeneous problem, we use the method of
variation of parameters and search for a particular solution of (11) of the form

u(x) = v1 (x)u1 (x) + v2 (x)u2 (x) (12)

2
where v1 (x) and v2 (x) are functions to be determined such that (12) satisfies (11). Since we have only one
equation to be satisfied and two unknown functions, we impose an additional constraint
dv1 dv2
u1 + u2 = 0 (13)
dx dx
Using (13), we obtain from (12)
du du1 du2
= v1 + v2
dx dx dx
and the equation (11) is satisfied if
dv1 du1 dv2 du2
p + p = f (x) (14)
dx dx dx dx
From (13) and (14) we obtain
dv1 −f u2
= (15)
dx c
dv2 f u1
= (16)
dx c
where  
du2 du1
c = p u1 − u2 (17)
dx dx
Remark: The quantity  
du2 du1  u u2 
W = u1 − u2 =  du11 du2 
dx dx dx dx
is called the Wronskian of u1 and u2 and satisfies the differential equation
dW 1 dp
=− W
dx p dx
Q: Using the Wronskian, show that expression (17) is constant.
Then, integrating (15) and (16) we obtain

1 x
Z
v1 (x) = − f (x0 )u2 (x0 )dx0 + c1 (18)
c a
1 x
Z
v2 (x) = f (x0 )u1 (x0 )dx0 + c2 (19)
c a
such that the general solution of the nonhomogeneous problem (11) is

u(x) = v1 (x)u1 (x) + v2 (x)u2 (x) (20)

u1 (x) x u2 (x) x
Z Z
= c1 u1 (x) + c2 u2 (x) − f (x0 )u2 (x0 )dx0 + f (x0 )u1 (x0 )dx0 (21)
c a c a

The constants c1 and c2 are determined by the boundary conditions.

A simple example. Consider the following problem

d2 u
= f (x)
dx2
with homogeneous boundary conditions

u(0) = 0, u(L) = 0

Two linearly independent solutions of the homogeneous differential equation are

u1 (x) = x (22)

3
 u2 (x) = L − x (23)
Then, W = u1 du
dx
2
− u2 du
dx
1
= −L such that from (18) and (19) we obtain

1 x
Z
v1 (x) = f (x0 )(L − x0 )dx0 + c1 (24)
L 0
1 x
Z
v2 (x) = − f (x0 )x0 dx0 + c2 (25)
L 0
Using the boundary conditions,

u(0) = 0 ⇒ c2 = 0

Z L
1
u(L) = 0 ⇒ c1 = − f (x0 )(L − x0 )dx0
L 0

and replacing in (24), (25) it follows that

Z x Z L Z L
1 1 1
v1 (x) = f (x0 )(L − x0 )dx0 − f (x0 )(L − x0 )dx0 = − f (x0 )(L − x0 )dx0 (26)
L 0 L 0 L x
Z x
1
v2 (x) = − f (x0 )x0 dx0 (27)
L 0
and the solution of the nonhomogeneous problem is

u(x) = v1 (x)u1 (x) + v2 (x)u2 (x) (28)

x L L−x x
Z Z
= − f (x0 )(L − x0 )dx0 − f (x0 )x0 dx0 (29)
L x L 0

This solution may be written as

Z L
u(x) = f (x0 )G(x, x0 )dx0 (30)
0

where the Green’s function G(x, x0 ) is given by

( −x(L−x0 )
L , x < x0
G(x, x0 ) = (31)
−x0 (L−x)
L , x > x0
The method of eigenfunction expansion for Green’s functions

Consider a general Sturm-Liouville nonhomogeneous ODE

L(u) = f (x), a < x < b

subject to two homogeneous boundary conditions. We know that the eigenfunctions Φn (x) of the related
eigenvalue problem
L(Φ) = −λσΦ
(for an arbitrary weight function σ) subject to the same homogeneous boundary conditions form a ”complete
set”, such that u(x) may be expressed as a generalized Fourier series of eigenfunctions
∞
X
u(x) = an Φn (x) (32)
n=1

Term-by-term differentiation of (32) implies (together with the linearity of L)

4
 ∞ ∞ ∞
!
X X X
L(u) = L an Φn (x) = an L(Φn ) = − an λn σΦn
n=1 n=1 n=1

In the last relation above we used L(Φn ) = −λn σΦn . Therefore,

∞
X
L(u) = f (x) ⇒ − an λn σΦn = f (x)
n=1

and using the orthogonality of the eigenfunctions (with weight σ) we obtain the coefficients
Rb
f (x)Φn dx a
an =
Rb (33)
−λn a φ2n σdx
R P
Replacing (33) in (32) we obtain ( after interchanging and )
∞ Rb b ∞
f (x)Φn (x)dx
Z
X
a
X Φn (x)Φn (x0 )
u(x) = Rb Φn (x) = f (x0 ) Rb dx0 (34)
n=1 −λn a φ2n σdx a 2
n=1 −λn a φn σdx

For this problem,

Z b
u(x) = f (x0 )G(x, x0 )dx0
a

where the Green’s function G(x, x0 ) is

∞
X Φn (x)Φn (x0 )
G(x, x0 ) = Rb (35)
2
n=1 −λn a φn σdx

Remark: Formula (35) implies that the Green’s function is symmetric:

G(x, x0 ) = G(x0 , x)

The Dirac Delta Function and its relationship to Green’s function

In the previous section we proved that the solution of the nonhomogeneous problem

L(u) = f (x)
subject to homogeneous boundary conditions is
Z b
u(x) = f (x0 )G(x, x0 )dx0
a

In this section we want to give an interpretation of the Green’s function. Let xs , a < xs < b represent an
arbitrary fixed point. We consider a perturbation of the source term

f (x) + δ (x)

where δ (x) is a continuous function with the following properties:

Rb
1. a
δ (x)dx = 1
2. δ (x) = 0 if x ≥ xs +  or x ≤ xs − 

5
A possible choice for δ (x) is the hat function


 0, a < x ≤ xs − 
 1
2 (x − xs + ), xs −  < x < xs
δ (x) = 1 (36)
− 2 (x − xs − ), xs < x < xs + 
 


0, xs +  ≤ x < b

Denote u (x) the solution of the perturbed problem

L(u ) = f (x) + δ (x)

subject to the same homogeneous boundary conditions as u. The variation ∆u = u − u in the solution u
due to the perturbation of the source term satisfy the following problem:

L(∆u) = δ (x)

with homogeneous boundary conditions, such that we have

Z b Z xs + Z xs +
∆u(x) = δ (x0 )G(x, x0 )dx0 = δ (x0 )G(x, x0 )dx0 = G(x, x ) δ (x0 )dx0
a xs − xs −

where x is a point in the interval (xs − , xs + ). The existence of such point is a consequence of the mean
value theorem, since G(x, x0 ) is continuous. Notice that
Z xs + Z b
δ (x0 )dx0 = δ (x0 )dx0 = 1
xs − a

If we let  → 0, we obtain
∆u(x) = G(x, xs )
The limit
lim δ = δ(x − xs ) (37)
→0

represents an infinitely concentrated pulse at xs , that is zero everywhere, except at x = xs


0, x 6= xs
δ(x − xs ) = (38)
∞, x = xs

We define the Dirac delta function δ(x − xs ) as an operator with the property (38) such that for every
continuous function f (x) we have Z ∞
f (x) = f (xs )δ(x − xs )dxs
−∞

The Dirac delta function has unit area

Z ∞
δ(x − xs )dxs = 1
−∞

and is even in the argument x − xs

δ(x − xs ) = δ(xs − x)
Then G(x, xs ) satisfies
L[G(x, xs )] = δ(x − xs )
and the homogeneous boundary conditions at x = a and x = b, such that we obtain the following interpreta-
tion: the Green’s function G(x, xs ) is the response at x due to a concentrated source at xs . The
symmetry of the Green’s function
G(x, xs ) = G(xs , x)

6
implies that the response at x due to a concentrated source at xs is the same as the response at
xs due to a concentrated source at x. This property is known as Maxwell’s reciprocity.

Jump conditions

Next we show how the Green’s function may be obtained by directly solving
L[G(x, x0 )] = δ(x − x0 ) (39)
subject to homogeneous boundary conditions.

Example consider the steady-state problem

u00 (x) = f (x)
u(0) = 0, u(L) = 0
The Green’s function to this problem is then a solution to
d2 G(x, x0 )
= δ(x − x0 ) (40)
dx2
G(0, x0 ) = 0, G(L, x0 ) = 0 (41)
2
d G(x,x0 )
It follows then that dx2 = 0, x 6= x0 , therefore G(x, x0 ) must be a linear function on each interval
x < x0 and x > x0 (
a + bx, x < x0
G(x, x0 ) = (42)
c + dx, x > x0
The constants a, b, c, d need to be determined. From the boundary conditions we have
G(0, x0 ) = 0 ⇒ a = 0
G(L, x0 ) = 0 → c = −dL
such that we obtain (
bx, x < x0
G(x, x0 ) = (43)
d(x − L), x > x0
To find b and d, first we require that G(x, x0 ) must be continuous at x0
G(x− +
0 , x0 ) = G(x0 , x0 ) ⇒ bx0 = d(x0 − L) (44)
Next, for any  > 0, integrating (40) in (x0 − , x0 + )
d d
G(x0 + , x0 ) − G(x0 − , x0 ) = 1
dx dx
then passing to the limit as  → 0, we get the jump condition
d d
G(x+
0 , x0 ) − G(x−
0 , x0 ) = 1 ⇒ d − b = 1 (45)
dx dx
Solving (44-45) we get
x0 x0 − L
d= , b=
L L
such that the Green’s function (43) is
x
(
−L (L − x0 ), x < x0
G(x, x0 ) = (46)
− xL0 (L − x), x > x0
which is the same expression we obtained using the variation of parameters.

7
Nonhomogeneous boundary conditions

The Green’s function may be used to solve problems with nonhomogeneous boundary conditions e.g.,

u00 = f (x) (47)

u(0) = α, u(L) = β (48)

The solution u(x) may be written as the sum

u = u1 + u2

where u1 (x) solves the nonhomogenous ODE with homogeneous boundary conditions

u001 = f (x), u1 (0) = u1 (L) = 0

and u2 (x) solves the homogenous ODE with nonhomogeneous boundary conditions

u002 = 0, u2 (0) = α, u2 (L) = β

Then,
Z L
u1 (x) = f (x0 )G(x, x0 ) dx0
0
β−α
u2 (x) = α + x
L
such that
L
β−α
Z
u(x) = u1 (x) + u2 (x) = f (x0 )G(x, x0 ) dx0 + α + x
0 L
is the solution to the nonhomogeneous problem (47-48).