MA-203: Note on Green’s Function

April 19, 2020

Introduction:
Consider a homogeneous differential equations

d2 y
=0 (1)
dx2
which can be solved very easily and we get y(x) = ax + b.
Similarly consider
d2 y
+ λ2 y = 0 (2)
dx2
This differential equation can be solved to get the solution y as
y(x) = Acos(λx) + B sin(λx)
i.e. there are simply ways to solve the homogeneous differential equations.
However, if we consider the inhomogeneous differential equations
d2 y d2 y
2
= ln(x), + λ2 y = tan(x)
dx dx2
then the problem become difficult to solve. In such cases the Green’s function
may be used to obtain the solution of differential equations. Green’s functions
are named after the mathematician, George Green, who first developed the
concept in the 1830s.
For the solution of boundary value problems associated with either ordinary
or partial differential equations, one requires a brief knowledge about Green’s
function.
Here, we will study strategies for solving the boundary value problems of in-
homogeneous linear differential equation Ly = f (x). The tool we will use is
the Green function, which is an integral kernel representing the inverse operator
L−1 .
Consider the differential equation, Ly(x) = f (x), where L is an ordinary
linear differential operator, f (x) is a known function while y(x) is an unknown
function. To solve above equation, one method is to find the inverse operator
to L in the form of an integral operator with a kernel G(x,t ) such that,
Z
−1
y(x) = L f (x) = G(x, t)f (t)dt,

1
 The kernel of this integral operator is called Green’s function for the differ-
ential operator. Thus the solution to the non-homogeneous differential equation
Ly = f (x), can be written down, once the Green’s function for the problem
is known. For this reason, the Green’s function is also sometimes called the
fundamental solution associated to the operator L. It can be used to solve both
partial and exact differential equations.
Constructing Green’s function : Fundamental concept
We will illustrate the construction of Green’s function by considering
the Sturm Liouville differential equation and the corresponding boundary
value problem (BVP) with homogeneous boundary conditions (SL problem) as
follows:
Consider a inhomogeneous differential equation
Ly(x) + f (x) = 0, a≤x≤b (3)
 
d dy
where a,b are real numbers s.t. a < b, L = [ dx p(x) dx + q(x)] is the Sturm
Liouville operator and p(x), p0 (x), q(x), are continuous real valued functions in
[a, b]. p(x) is not equal to zero in [a,b].
Above Eq. (3) is a non-homogeneous linear differential equation.
The homogeneous part of equation (3) can be written as
Ly(x) = 0 (4)
We add the following general type homogeneous boundary conditions with Eq.
(3) and (4):
α1 y(a) + α2 y 0 (a) = 0 (5)
β1 y(b) + β2 y 0 (b) = 0 (6)
where α1 , α2 (not both zero) and β1 , β2 (not both zero) are the real constants.
The solution of equation (3) is formally given by y(x) = L−1 [−f (x)]. The
inverse of the differential operator L is an integral operator, which we seek to
Rb
write in the form y(x) = a G(x, t)f (t)dt. The function G(x, t) is referred to as
the kernel of the integral operator and is called the Green’s function.

Definition: If the system of (4-6) has only trivial solution y(x) = 0, then there
exists a Green’s function G(x, t), defined on [a, b] × [a, b], and there exists a
unique solution of differential equation (3) along with the boundary conditions
(5-6), where G(x, t) satisfy the following properties for a given t :
(I) Differential equation: G(x,t) is taken as solution of Ly(x)=0, for
x 6= t as

2
 (
G1 (x, t), a≤x<t
G(x, t) = (7)
G2 (x, t), t<x≤b
such that
LG1 (x, t) = 0, a≤x<t
,
LG2 (x, t) = 0, t<x≤b
This implies that for x < t , G(x, t) is proportional to y1 (x). Thus, since
y1 (x) is a solution of the homogeneous equation, then so is G(x, t). For x > t
, G(x, t) is proportional to y2 (x). So, once again G(x, t) is a solution of the
homogeneous problem.
(II) Boundary conditions:
G1 (x, t) satisfies the boundary condition (5). G2 (x, t) satisfies
the boundary condition (6).
(III) Continuity of G(x,t) at x = t :

G(x, t) is continuous at x = t , i.e.

G(x, t)|x=t− = G(x, t)|x=t+
⇒ G1 (t, t) = G2 (t, t)

(IV) Jump discontinuity of ∂G(x,t) ∂x at x = t : There is a finite jump

discontinuity of ∂G(x,t)
∂x at x = t as
 
∂G2 (x, t)  ∂G1 (x, t)  1
⇒ − =−
∂x 
x=t ∂x 
x=t p(t)

Construction of G(x,t):
Let y1 (x) and y2 (x) be two linearly independent solutions of Ly(x) = 0 on
(a,b) and let y1 (x) satisfies the boundary condition (5) and y2 (x) satisfies the
boundary condition (6).
Hence for some constants c1 and c2 , we define G1 (x, t) = c1 y1 (x),and G2 (x, t) =
c2 y2 (x),
Let

c1 y1 (x), if x < t
G(x, t) = (8)
c2 y2 (x), if x > t
Choose c1 and c2 such that

c2 y2 (t) − c1 y1 (t) = 0 (9)

1
c2 y20 (t) − c1 y10 (t) = − (10)
p(t)

3
 With ths choice of c1 and c2 , the function G(x,t) defined by the relation (8)
has all properties of the Green’s function. Since y1 (x) and y2 (x) satisfy Ly(x)
= 0, it follows that
d
y1 (py20 )0 − y2 (py10 )0 = [p(y1 y20 − y2 y10 )] = 0
dx
Hence p(y1 y20 − y2 y10 )] = A for all x in [a,b],
where A is a non-zero constant, because p(x ) is not equal to zero and y1 (x)
and y2 (x) are linearly independent solutions of Ly(x) = 0.
In particular, by solving equations (9) and (10) it is seen that

[y1 (t)y20 (t) − y2 (t)y10 (t)] = A/p(t), (11)

where A = p(t)[y1 (t)y20 (t)−y2 (t)y10 (t)] = p(t)W (y1 , y2 )(t). Here W (y1 , y2 )(t)
is the value of Wronskian of functions y1 (x) and y2 (x) at x = t.
From equations (10) and (11), we find that

c1 = −y2 (t)/A, c2 = −y1 (t)/A (12)

Hence, in view of (8) and (12), the Green’s function G(x,t) for the present
BVP is obtained as
( y1 (x)y2 (t)
− p(t)W (y1 ,y2 )(t) , if x < t
G(x, t) = y2 (x)y1 (t) (13)
− p(t)W (y1 ,y2 )(t) , if x > t
From equation (13), we find that Green’s function G(x,t) is sym-
metric, i.e., G(x,t)=G(t,x) for the present problem.

Theorem: Let G(x, t) be given by the relation (13). Then y(x) is a solution of
equation (3), and boundary conditions (5-6) if and only if
Z b
y(x) = G(x, t)f (t)dt (4)
a
Proof: Let the relation
"Z (14) hold. Then equation (13) yields #
x Z b
y(x) = − y2 (x)y1 (t)f (t)dt + y2 (t)y1 (x)f (t)dt /A (14)
a x

"Z (14) w.r.t x yields Z

differentiating equation #
x b
y 0 (x) = − y20 (x)y1 (t)f (t)dt + y10 (x)y2 (t)f (t)dt /A
a x

again differentiating w.r.t x, we get

y 00 (x) = − [y20 (x)y1 (x) − y10 (x)y2 (x)] f (x)/A

"Z #
x Z b
− y200 (x)y1 (t)f (t)dt + y100 (x)y2 (t)f (t)dt /A
a x

4
Substituting y, y 0 and y 00 in Left hand side of equation (3), we have

L(y(x)) + f (x) = (p(x)y 0 (x))0 + q(x)y(x) + f (x)

= p(x)y 00 (x) + p0 (x)y 0 (x) + q(x)y(x) + f (x) = 0 (15)

Here, we have used the relation − y1 (t)y20 (t) y10 (t)y2 (t)
= A/p(t), and the fact
that L(y1 (x)) = 0, L(y2 (x)) = 0.
Further, from the relation
( (15), it is seen Rthat
b
Ay(a) = −y1 (a) a y2 (t)f (t)dt
b (16)
Ay 0 (a) = −y10 (a) a y2 (t)f (t)dt
R

Since y1 (x) satisfies the boundary condition given in equation (5), it follows
from (17) that y(x) also satisfies the boundary condition (5). Similarly y(x)
satisfies the boundary condition (6). This proves that y(x) satisfies equation
(3) and boundary conditions (5-6).
Conversely, let y(x) satisfy equations (3),(5-6). Then from equation (3) it is
clear that Z b Z b
− G(x, t)L(y(t))dt = G(x, t)f (t)dt (17)
a a
The left side of equation (18) is
Z x Z b
− G1 (x, t)L(y(t))dt − G2 (x, t)L(y(t))dt (18)
a x
Now, a well known result is used that if u and v are two functions which admits
the continuous derivatives in [t1 , t2 ], then
Z t2 Z t2
u(t)L(v(t))dt = v(t)L(u(t))dt + [p(t)(u(t)v 0 (t) − u0 (t)v(t))]tt21 (19)
t1 t1

Applying the identity (20) in expression (19) and using the properties of G1 (x, t)
and G2 (x, t) the left side of equation (18) becomes
   
∂G1 (x, t)  ∂G2 (x, t) 
−p(x) [G1 (x, x)y 0 (x) − y(x)] − [G2 (x, x)y 0 (x) − y(x)] (20)
∂x 
t=x ∂x 
t=x
The first and third term in equation (20) cancel each other because of continuity
property of G(x, t) at t = x. The property (IV) in the definition of Green’s
function now shows that the value of the expression (21) is y(x). But expression
Rb
(21) is left side of equation (18), which means y(x) = a G(x, t)f (t)dt. This
completes the proof of the theorem.
Hence, we obtain the following:
If y = 0 is the only trivial solution of (4), (5-6), then there exists a unique
solution of differential equation (3), and bounadry conditions given by (5-6) as
Z b
y(x) = G(x, t)f (t)dt
a
Z x Z b
= G2 (x, t)f (t)dt + G1 (x, t)f (t)dt
a x

5
 Z x Z b
y2 (x)y1 (t)f (t) y1 (x)y2 (t)f (t)
= −{ dt + dt}
a p(t)W (y1 , y2 )(t) x p(t)W (y1 , y2 )(t)
Hence, solution of Ly(x)= f(x) subjected to boundary conditions (5-6) is
given by
Z x Z b
y2 (x)y1 (t)f (t) y1 (x)y2 (t)f (t)
y(x) = { dt + dt}
a p(t)W (y1 , y2 )(t) x p(t)W (y1 , y2 )(t)
However, if there is a non-trivial solution of Ly(x) = 0, and conditions (5-6)
then there may not have any solution or may have many solution of the BVP
given by (3), (5-6). Concept of Modified Green’s function technique is used in
such cases.
—————–
Now we will illustrate the construction of Green’s function and hence
solution of BVP with the following Examples:
Problem 1: Find the Green’s function, if it exists for BVP
y 00 (x) = 0, y(0) = y(l) = 0

Solution: The given differential equation is

y 00 (x) = 0
⇒ y(x) = Ax + B and using the boundary conditions y(0) = y(l) = 0, we have
A = B = 0. Therefore y = 0 is solution of given system. Thus Green’s functions
exists.
Ax + B, 0 ≤ x < t
Let G(x, t) = {
Cx + D, t < x ≤ l
To get G(x,t), we will now use the properties of Green’s function (I)-(IV).
Property (II) implies G1 (0, t) = 0 and G2 (l, t) = 0 =⇒ B = 0 and Cl + D = 0
....(i)

Again property (III) gives G1 (t, t) = G2 (t, t) =⇒ At = Ct + D

=⇒ D = (A − C)t .................(ii)
 
Aplying propert (IV), we find ∂G∂x 2 (x,t)  ∂G1 (x,t) 
 − ∂x  1
= − p(t) , since p(x) = 1
x=t x=t
=⇒ C − A = −1 ...........(iii)

Now from (ii) & (iii), we get D =
t, and using this value in (i) we get C = − tl .
again from (ii), we get A = 1 − tl .

Hence, by using the values of A, B, C and D, we obtain the Green’s function as

x 1 − tl
, 0 ≤ x < t


G(x, t) = { .
t 1 − xl , t < x ≤ l
Problem 2: Solve y 00 + µ2 y = 0, subjected to y(0) = y(1) = 0
Solution: y(x) = Acos(µx) + Bsin(µx)

6
y(0) = 0 ⇒ A = 0
y(1) = 0 ⇒ Bsin µ = 0 ⇒ B = 0.
Therefore, y = 0. (trivial solution).

Now, consider the Green’s function (using property (I)) as

Acos(µx) + Bsin(µx), 0 ≤ x < t
G(x, t) = { (i)
Ccos(µx) + Dsin(µx), t < x ≤ 1

We will apply the properties of the Green’s function to get G(x,t).

Property (II) gives G1 (0, t) = 0 and G2 (1, t) = 0 ⇒ A = 0 and Ccos(µ) +
Dsin(µ) = 0
A=0
⇒ (ii)
C = − Dsin(µ)
cos(µ)

Using property (III), we get G1 (t, t) = G2 (t, t)

⇒ Bsin(µt) = Ccos(µt)+Dsin(µt) (iii)

Now using equation (ii) in above equation, we have

Bsin(µt) = − Dsin(µ)
cos(µ) cos(µt) + Dsin(µt) = D
sin(t−1)µ
cos µ
sin(t−1)µ
⇒ B = D cos µ sin µt (iv)
 
Further property (IV) gives ∂G∂x
2 (x,t) 
 − ∂G∂x
1 (x,t) 
 1
= − p(t) , here p(x) = 1
x=t x=t
⇒ −Cµsin(µt) + Dµcos(µt) − Bµcos(µt) = −1
Hence, using equation (ii-iv) in above equation we get
⇒ Dsin(µ) sin(t−1)µ 1
cos(µ) sin(µt) + Dcos(µt) − D cos µ sin µt cos(µt) = − µ
cos µ
⇒ D = − µ sin µ sin(µt) (v)

From equation (iv) we obtain B = − sin(t−1)µ

µ sin µ
and from equation (v), we get C = sin(µt)/µ
Hence, using the values of A, B, C and Din equation (i) we have the Greens
function as
− sin(t−1)µ
µ sin µ sin(µx), 0 ≤ x < t
G(x, t) = { sin(x−1)µ
µ sin µ sin(µt), t < x ≤ 1

Problem 3: Solve y 00 + y = x,with boundary condition y(0) = y(π/2) = 0

Solution: Consider the homogeneous differential equation y 00 + y = 0 ⇒ y =
Acos x + Bsin x
Now, y(0) = 0 ⇒ A = 0 and y(π/2) = 0 ⇒ B = 0.
i.e. trivial solution is the only solution of Ly = 0 and the given boundary
conditions.
Assume the Green’s function as

7
 Acos(x) + Bsin(x), 0 ≤ x < t
G(x, t) = { (i)
Ccos(x) + Dsin(x), t < x ≤ π/2
From the properties of Green’s function, we have
Property (II): G1 (0, t) = 0 and G2 (π/2, t) = 0
⇒ A = 0 and D = 0 (ii)

Proerty (III): G1 (t, t) = G2 (t, t)⇒ Bsin t = Ccos t

cos t
⇒ B = C sin t (iii)
 
Proerty (IV): ∂G∂x 2 (x,t) 
 − ∂G∂x
1 (x,t) 
 1
= − p(t) , here p(x) = 1
x=t x=t
⇒ −Csin t − Bcos t = −1
using equation (iii) in above equation, we have
cos t
Csin t + C sin t cos t = 1
⇒ C = sin t (iv)

By using equation (iv) in equation (iii), we get

B = cos t
Therefore, using the values of A, B, C and D in equation (i), we obtain the
Green’s function in the form
cos t sin x, 0 ≤ x < t
G(x, t) = {
cos x sin t, t < x ≤ π/2
Hence, the solution of the differential equation can be written as
R π/2
y(x) = 0 G(x, t)f (t)dt, here f (t) = −t.
Therefore, we get the solution of the present BVP as
Z x Z π/2
y(x) = − t cos x sin t dt − t cos t sin x dt
0 x
x π/2
= −cos x (t cos t + sin t)|0 − sin x (t sin t + sin t)|0
π
= x − sin x.
2
Problem 4: Construct the Green’s function for the system
d2 y
dx2 −y = 0 (i)

y(0) = y 0 (0)
(ii)
y(l) + λy 0 (l) = 0
2
d y x −x
Solution: Given dx 2 − y = 0 ⇒ y(x) = Ae + Be
0 x −x
also, y (x) = Ae + Be
Now, using the boundary condition
y(0) = y 0 (0) ⇒ A + B = A − B ⇒ B = 0
y(l) + λy 0 (l) = 0 ⇒ Ael + λAel = 0 ⇒ A = 0
Therefore, y = 0 is the solution.

8
 Consider the Green’s function as,
Aex + Be−x , 0 ≤ x < t
G(x, t) = { (iii)
Cex + De−x , t < x ≤ l
Now, we use the properties of Green’s function
Property (II): G1 (0, t) = G01 (0, t) ⇒ A + B = A − B ⇒ B = 0
and G2 (l, t) + λG02 (l, t) = 0 ⇒ Cel + De−l + λ(Cel − De−l ) = 0
−2l
⇒ C = −D 1−λ 1+λ e (iv)

Property (III):G1 (t, t) = G2 (t, t) ⇒ Aet = Cet + De−t

using equation (iv) we find
−2l
A =−D 1−λ
1+λ e +De−2t (v)
 
Property (IV): ∂G∂x2 (x,t) 
 − ∂G∂x 1 (x,t) 
 1
= − p(t) , here p(t) = 1
x=t x=t
⇒Cet − De−t − Aet = −1
−2l
⇒−D 1−λ
1+λ e − De−2t + D 1−λ
1+λ e
−2l
− De−2t = −1
⇒D = et /2 (vi)

Using equation (vi) in equation (iv), we get

(1−λ) −2l t
C = − 2(1+λ) e e
Again using equation (vi) in equation (v), we have
(1−λ) t −2l −t
A = − 2(1+λ) ee + e2

Hence, by using the values of A, B, C and D in equation (iii) we obtain the

Green’s function as,
(1−λ) x+t−2l ex−t
− 2(1+λ) e + 2 , 0≤x<t
G(x, t) = { (1−λ) x+t−2l e−x+t
− 2(1+λ) e + 2 , t<x≤l

Task:
1. Find the Green’s function for the problem and hence solve it:
y 00 = x2 , y(0) = 0, y(1) = 0
2. Construct the Green’s function for the system
xy 00 + y 0 = 0

where y(0) is bounded and y( 1) = 0

3. Construct the Green’s function for the system
d2 y
dx2 −y =0

y(0) = 0
y(1) = 0

9
4. Find the Green’s function for the problem and hence solve it:
y 00 + π 2 y = cosπx, y(0) = y(1), y 0 (0) = y 0 (1)
———————————-

Special Note:
There is a more general theory of Green’s functions for ordinary and partial
differential equations. Much of this theory relies on understanding that the
Green’s function really is the system response function to a point source. This
begins with the concept that the boundary value Green’s function satisfies a
homogeneous differential equation for x 6= t,
∂ ∂
[p(x) G(x, t)] + q(x)G(x, t) = 0, x 6= t
∂x ∂x
For x = t, we have seen that the derivative of G(x,t) has a jump in its value.
This is similar to the step, or Heaviside, function, H(x).
The Dirac delta function and Heaviside unit function satisfy the relation
d
δ(x) = dx H(x).
Hence, it can be shown that Green’s function satisfies
∂ ∂
LG(x, t) = [p(x) G(x, t)] + q(x)G(x, t) = δ(x − t)
∂x ∂x
Further, the solution of the non-homogeneous differential equation
Ly(x) = f (x)
subjected to homogeneous boundary conditions can be found out in terms of
Green’s function G(x, t) (that satisfies LG(x, t) = δ(x−t) ) and boundary terms.
This concept can also be extended to solve non-homogeneous boundary value
problems.
Hence, Green’s function is used as an integral kernel that can solve differential
equations from a large number of families. In the modern study of linear differ-
ential equations, Green’s functions are studied largely from the point of view of
fundamental solutions instead. It is an important mathematical tool that has
application in many areas of theoretical physics including mechanics, electro-
magnetism, solid-state physics, thermal physics, and the theory of elementary
particles.
Reference Books: (1) Higher Engineering Mathematics by Dr. BS
Grewal, (2) Boundary Value problem by Rai Singhania

(3) Interested one may aso use the available note:

http://people.uncw.edu/hermanr/pde1/PDEbook/Green.pdf
(part of Chapter 7 (Green’s Functions and Nonhomogeneous Problems) in
the book of “Introduction to Partial Differential equations by Russell L. Herman)
*************

10