The Method of Green’s Functions: Examples

General Strategy The general form of a linear ODE with homogeneous boundary conditions is

dn u dn−1 u du
an (x) + a n−1 (x) + . . . + a1 (x) + a0 (x)u ≡ Lu = f (x) (1)
dxn dxn−1 dx
Bu = 0 , (2)

where Bu = 0 specifies any linear boundary conditions at the endpoints of the domain x ∈ [a, b]. Here we
assume that all the coefficients, aj (x), are continuous functions.
If the operator L is non-singular, which is equivalent to L not having any zero eigenvalues (i.e. there
are no non-trivial solutions to the associated homogeneous problem with f (x) = 0), then the solution to (1)
may be written as
Z b
u(x) = G(x, y)f (y) dy . (3)
a
The kernel G(x, y) involved in this integral form of the solution is known as the Green’s Function, which is
defined to satisfy the ODE (in the sense of distributions)

Lx G(x, y) = δ(x − y) . (4)

The subscript x on the operator L serves to remind you that differentiation is performed with respect to the
variable x.
The Green’s function G(x, y) is chosen to satisfy the homogeneous boundary conditions (2). It must
further satisfy certain continuity and jump conditions owing to the singular delta distribution appearing in
its defining equation (4). If one integrates (4) through the singularity at x = y, one finds that the (n − 1) th
derivative of G(x, y) has a jump discontinuity:
x=y+
dn−1 G  1
lim n−1
= (5)
→0 dx  an (y)
x=y−
n−1
In other words, the distribution ddxn−1G is like a Heaviside function for which the jump is 1/an (y), rather
than unity. Continuing this line of argument, all lower order derivatives are continuous at x = y:
x=y+
dk G 
lim = 0, (k = 0, 1, . . . , n − 2). (6)
→0 dxk x=y−
The jump condition (5) and the continuity conditions (6), coupled with the boundary conditions (2), defines
the Green’s function uniquely if the operator L is non-singular.

Example 1 Consider the following first order ODE over the half-infinite domain x ∈ [0, ∞):

du
+ a0 (x)u = f (x)
dx
u(0) = 0.

The Green’s function G(x, y) is defined by the solution of

1
 dG
+ a0 (x)G = δ(x − y)
dx
G(0, y) = 0 .

Integration of this equation gives

 Rx
− a (s)ds
 c1 e 0 0
 x>y
G(x, y) = Rx
a (s)ds
 −
c2 e 0 0 x<y ,


where c1 (y) and c2 (y) are integration constants to be determined. The boundary condition G(0, y) = 0
demands c2 = 0. The jump condition is thus

lim G|x=y+ = 1 ,
→0

which gives
Ry
a0 (s)ds
c1 (y) = e 0 .

Note that we do not impose any further constraints on G(x, y), such as the continuity conditions (6), because
the defining ODE is only 1st order. The Green’s function is then
 Rx
 e− y a0 (s)ds

x>y
G(x, y) =

0 x < y.


The final solution for u(x) is

Z ∞ Z x Rx
− a0 (s)ds
u(x) = G(x, y)f (y) dy = e y f (y) dy .
0 0

You may verify this solution with a couple of simple examples. If a0 = 1 and f = 1, then you may verify the
solution u(x) = 1 − e−x . In the case a0 = x and f = x2 , one finds the solution
r  
π − 1 x2 x
u(x) = x − e 2 erfi √ ,
2 2
where the imaginary error function is defined by
z
2
Z
2
erfi(z) = √ et dt .
π 0

Example 2 An example of the Green’s function method for a second order equation is

d2 u du
−η = f (x)
dx2 dx
u(−∞) = 0
u(0) = 0,

where η > 0 is a constant and the spatial domain is x ∈ (−∞, 0]. The Green’s function is given by solving
the system

2
 d2 G dG
2
−η = δ(x − y) (7)
dx dx
G(−∞, 0) = 0 (8)
G(0, y) = 0 (9)
x=y+
dG 
lim = 1 (10)
→0 dx 
x=y−
x=y+
lim G|x=y− = 0. (11)
→0

A couple of simple integrations yields


 α + βeηx x>y
G(x, y) = ,
γ + ξeηx x<y


where the constants {α, β, γ, ξ} are determined by conditions (8)-(11). Application of the boundary condition
(8) gives γ = 0. Boundary condition (9) demands β = −α. Up to this point the Green’s function is then

 α (1 − eηx ) x>y
G(x, y) = .
ξeηx x<y


The jump condition (10) and continuity condition (11) are, respectively,

−αηeηy − ξηeηy = 1
α (1 − eηy ) = ξeηy ,

the solution of which is

1
α = −
η
1

ξ = 1 − e−ηy .
η
Finally, we may express the Green’s function by
1
(eηx − 1)

 η x>y
G(x, y) = ,
 1
η (1 − e−ηy ) eηx x<y

and the solution of the ODE is

Z 0
eηx 0
Z x
1
Z
G(x, y)f (y) dy = (eηx − 1)


u(x) = f (y) dy + 1 − e−ηy f (y) dy .
−∞ η −∞ η x