Chapter 8

Green’s Functions

“I do not want to claim that every topic should be probed in detail. Following one’s
own talent and inclination, one should select at least one topic, and study it in
depth. In the others, one should follow the example of the bee which sucks a drop of
nectar from each flower ...” - Jacopo Francesco Riccati (1676-1754)
“The young theoretical physicists of a generation or two earlier subscribed to
the belief that: If you haven’t done something important by age 30, you never will.
Obviously, they were unfamiliar with the history of George Green, the miller of
Nottingham.” Julian Schwinger (1918-1994)

8.1 Introduction

In this chapter we will investigate the solution of nonhomogeneous

differential equations using Green’s functions. Our goal is to solve the non-
homogeneous differential equation

L[y] = f ,

where L is a differential operator. The solution is formally given by

y = L −1 [ f ] .

The inverse of a differential operator is an integral operator, which we seek

to write in the form Z
y = G ( x, ξ ) f (ξ ) dξ. (8.1)

The function G ( x, ξ ) is referred to as the kernel of the integral operator and

is called the Green’s function.
The history of the Green’s function dates back to 1828, when George
Green published work in which he sought solutions of Poisson’s equation
∇2 u = f for the electric potential u defined inside a bounded volume with
specified boundary conditions on the surface of the volume. He introduced
a function now identified as what Riemann later coined the “Green’s func-
tion”. A more in depth history can be found in Section 8.8.
We will restrict our discussion to Green’s functions for ordinary differen-
tial equations. Extensions to partial differential equations are typically one
264 differential equations

of the subjects of a PDE course. We will begin our investigations by examin-

ing solutions of nonhomogeneous second order linear differential equations
using the Method of Variation of Parameters, which is typically seen in a
first course on differential equations. We will identify the Green’s function
for both initial value and boundary value problems. We will then focus on
boundary value Green’s functions and their properties. Determination of
Green’s functions is also possible using Sturm-Liouville theory. This leads
to series representation of Green’s functions, which we will study in the last
section of this chapter.

8.2 What Are Green’s Functions?

Adapted from .
Russell L. Herman. Laplace Trans-
forms and Green’s Functions. In A Green’s function is a mathematical tool used in many branches
V. Martinez-Luaces, editor, Engineering of physics, engineering, and applied mathematics. It is a function that de-
Problems and the Laplace Transform, pages
1–112. Nova Science, 2025 scribes the response of a linear system to a localized input or “forcing”
function. In other words, it is a solution to a differential equation with a
1
Paul Adrien Maurice Dirac (1902-1984) specific forcing term that is often represented by a Dirac delta function.1
introduced the modern concept of the We use the Green’s function to determine the response of the system to an
delta function in 1927 [Dirac, 1927] and
put it in his 1932 quantum mechanics input by expressing the input as a sum of impulses and convolving each
text [Dirac, 1981]. It was a more rig- impulse function with the Green’s function. This technique is known as the
orous approach than earlier attempts to
describe impulse functions, leading to Green’s function method. In Section 8.8 we review some of the history of
the modern theory of distributions. [See Green’s functions.
[Lützen, 2012]]. The Dirac delta function The Green’s function itself is the response to a simple impulse. We can
is defined by two properties, δ( x ) = 0,
x ̸= 0 and demonstrate this idea using a simple example. Consider what happens
Z ∞ when we let f (t) = δ(t − t0 ) in the initial value problem,
δ( x ) dx = 1.
−∞
y′ (t) + cy(t) = δ(t − t0 ), y(0) = 0.
We will often use the sifting property
(8.2) in what follows. More details can
be found in Section 8.5.1. Here δ(t − t0 ) represents a unit impulse at time t = t0 . Using the solution
in Equation (8.1), we have

Z t  G (t, t0 ) 0 < t0 < t,

y(t) = G (t, τ )δ(τ − t0 ) dτ =
0
0, otherwise.



Here we made use of the sifting property of the Dirac delta function [Her-
man, 2016],
Z b
f (t)δ(t − t0 ) dt = f (t0 ), a < t0 < b. (8.2)
a
For a boundary value problem, we consider

y′′ ( x ) − a2 y( x ) = δ( x − x ′ ), y(0) = y( L) = 0.

Using the general solution in Equation (8.1), we have

Z L
y( x ) = G ( x, ξ )δ(ξ − x ′ ) dξ = G ( x, x ′ ), 0 < x ′ < L.
0

We further note that the boundary conditions require that G (0, x ′ ) = 0 =

G ( L, x ′ ).
 green’s functions 265

Therefore, the solutions to these problems are essentially Green’s func-

tions. In essence, the Green’s function of a system is a way to “encode”
the behavior of the system in response to an input in a single function.2 2
A number of people tried to define the
It has many important applications, such as in solving partial differential notion of an impulse before Dirac. As
noted by Katz and Tall [2013], Cauchy
equations, studying the behavior of electromagnetic fields, and modeling [1816] wrote in 1815 (published 1827) the
the behavior of mechanical systems. formula
Z a+ϵ
1 a dµ π
F (µ) = F ( a ).
2 a−ϵ a2 − ( µ − a )2 2
8.3 The Method of Variation of Parameters
However, Fourier [1808] wrote in his
memoir a similar expression,
Z ∞ Z ∞
1
We are interested in solving nonhomogeneous second order linear f (x) = da f ( a) dp cos( px − pa).
2π −∞ −∞
differential equations of the form
More can be found in Katz and Tall
′′ ′ [2013] and in Van der Pol and Bremmer
a2 ( x ) y ( x ) + a1 ( x ) y ( x ) + a0 ( x ) y ( x ) = f ( x ). (8.3) [1950], Chapter V.
In some sense, Green [1852, 1871a]
The general solution of this nonhomogeneous second order linear differen- and Heaviside were also interested in
tial equation is found as a sum of the general solution of the homogeneous specifying unit sources for their prob-
lems. Heaviside was interested in be-
equation, ing able to find the “disturbance in
a2 ( x )y′′ ( x ) + a1 ( x )y′ ( x ) + a0 ( x )y( x ) = 0, (8.4) any electrical system resulting from the
sudden application at time t = 0 of
and a particular solution of the nonhomogeneous equation. Recall from a given force” [Van der Pol, 1929].
Rayleigh [1911] referenced how Kelvin
Chapter 1 that there are several approaches to finding particular solutions of [1856] derived special solutions from
nonhomogeneous equations. Any guess would be sufficient. An intelligent point sources or a point source rate.
guess, based upon the Method of Undetermined Coefficients, was reviewed
previously in Chapter 1. However, a more methodical method, which is first
seen in a first course in differential equations, is the Method of Variation of
Parameters. Also, we explored the matrix version of this method in Section
2.9. We will review this method in this section and extend it to the solution
of boundary value problems.
While it is sufficient to derive the method for the general differential
equation above, we will instead consider solving equations that are in Sturm-
Liouville, or self-adjoint, form. Therefore, we will apply the Method of
Variation of Parameters to the equation

dy( x )
 
d
p( x ) + q ( x ) y ( x ) = f ( x ). (8.5)
dx dx

Note that f ( x ) in this equation is not the same function as in the general
equation posed at the beginning of this section.
We begin by assuming that we have determined two linearly independent
solutions of the homogeneous equation. The general solution is then given
by
y ( x ) = c1 y1 ( x ) + c2 y2 ( x ). (8.6)
In order to determine a particular solution of the nonhomogeneous equa-
tion, we vary the parameters c1 and c2 in the solution of the homogeneous
problem by making them functions of the independent variable. Thus, we
seek a particular solution of the nonhomogeneous equation in the form

y p ( x ) = c1 ( x ) y1 ( x ) + c2 ( x ) y2 ( x ). (8.7)
266 differential equations

In order for this to be a solution, we need to show that it satisfies the

differential equation. We first compute the derivatives of y p ( x ). The first
derivative is

y′p ( x ) = c1 ( x )y1′ ( x ) + c2 ( x )y2′ ( x ) + c1′ ( x )y1 ( x ) + c2′ ( x )y2 ( x ).

Without loss of generality, we will set the sum of the last two terms to zero.
(One can show that the same results would be obtained if we did not. See
Problem 2.) Then, we have

c1′ ( x )y1 ( x ) + c2′ ( x )y2 ( x ) = 0. (8.8)

Now, we take the second derivative of the remaining terms to obtain

y′′p ( x ) = c1 ( x )y1′′ ( x ) + c2 ( x )y2′′ ( x ) + c1′ ( x )y1′ ( x ) + c2′ ( x )y2′ ( x ).

Expanding the derivative term in Equation (8.5),

p( x )y′′p ( x ) + p′ ( x )y′p ( x ) + q( x )y p ( x ) = f ( x ),

and inserting the expressions for y p , y′p ( x ), and y′′p ( x ), we have

f ( x ) = p( x ) c1 ( x )y1′′ ( x ) + c2 ( x )y2′′ ( x ) + c1′ ( x )y1′ ( x ) + c2′ ( x )y2′ ( x )

 

+ p′ ( x ) c1 ( x )y1′ ( x ) + c2 ( x )y2′ ( x ) + q( x ) [c1 ( x )y1 ( x ) + c2 ( x )y2 ( x )] .

 

Rearranging terms, we find

f ( x ) = c1 ( x ) p( x )y1′′ ( x ) + p′ ( x )y1′ ( x ) + q( x )y1 ( x )

 

+c2 ( x ) p( x )y2′′ ( x ) + p′ ( x )y2′ ( x ) + q( x )y2 ( x )

 

+ p( x ) c1′ ( x )y1′ ( x ) + c2′ ( x )y2′ ( x ) .

 
(8.9)

Since y1 ( x ) and y2 ( x ) are both solutions of the homogeneous equation. The

first two bracketed expressions vanish. Dividing by p( x ), we have that
f (x)
c1′ ( x )y1′ ( x ) + c2′ ( x )y2′ ( x ) = . (8.10)
p( x )
Our goal is to determine c1 ( x ) and c2 ( x ). In this analysis, we have found
that the derivatives of these functions satisfy a linear system of equations
(in the ci ’s):

Linear System for Variation of Parameters

c1′ ( x )y1 ( x ) + c2′ ( x )y2 ( x ) = 0.

f (x)
c1′ ( x )y1′ ( x ) + c2′ ( x )y2′ ( x ) = . (8.11)
p( x )

This system is easily solved to give

f ( x ) y2 ( x )
c1′ ( x ) = −
p( x ) y1 ( x )y2′ ( x ) − y1′ ( x )y2 ( x )
 

f ( x ) y1 ( x )
c2′ ( x ) = . (8.12)
p( x ) y1 ( x )y2′ ( x ) − y1′ ( x )y2 ( x )

 green’s functions 267

We note that the denominator in these expressions involves the Wron-

skian of the solutions to the homogeneous problem. Recall that

y1 ( x ) y2 ( x )
W (y1 , y2 )( x ) = .
y1′ ( x ) y2′ ( x )

Furthermore, we can show that the denominator, p( x )W ( x ), is constant.

Differentiating this expression and using the homogeneous form of the dif-
ferential equation proves this assertion.
d d 
p( x ) y1 ( x )y2′ ( x ) − y1′ ( x )y2 ( x )


( p( x )W ( x )) =
dx dx
d
= y1 ( x ) ( p( x )y2′ ( x ))) + p( x )y2′ ( x )y1′ ( x )
dx
d
−y2 ( x ) ( p( x )y1′ ( x ))) − p( x )y1′ ( x )y2′ ( x )
dx
= −y1 ( x )q( x )y2 ( x ) + y2 ( x )q( x )y1 ( x ) = 0. (8.13)

Therefore,
p( x )W ( x ) = constant.
So, after an integration, we find the parameters as
Z x
f ( ξ ) y2 ( ξ )
c1 ( x ) = − dξ
x0 p ( ξ )W ( ξ )
Z x
f ( ξ ) y1 ( ξ )
c2 ( x ) = dξ, (8.14)
x1 p ( ξ )W ( ξ )
where x0 and x1 are arbitrary constants to be determined later.
Therefore, the particular solution of (8.5) can be written as
Z x Z x
f ( ξ ) y1 ( ξ ) f ( ξ ) y2 ( ξ )
y p ( x ) = y2 ( x ) dξ − y1 ( x ) dξ. (8.15)
x1 p ( ξ )W ( ξ ) x0 p ( ξ )W ( ξ )
As a further note, we usually do not rewrite the initial value problems in
self-adjoint form. Recall that for an equation of the form

a2 ( x )y′′ ( x ) + a1 ( x )y′ ( x ) + a0 ( x )y( x ) = g( x ). (8.16)

we obtained the self-adjoint form by multiplying the equation by

a1 ( x )
1 1
R
a2 ( x )
dx
e = p ( x ).
a2 ( x ) a2 ( x )
This gives the standard form

( p( x )y′ ( x ))′ + q( x )y( x ) = f ( x ),

where
1
f (x) = p ( x ) g ( x ).
a2 ( x )
With this in mind, Equation (8.15) becomes
Z x Z x
g ( ξ ) y1 ( ξ ) g ( ξ ) y2 ( ξ )
y p ( x ) = y2 ( x ) dξ − y1 ( x ) dξ. (8.17)
x1 a 2 ( ξ )W ( ξ ) x0 a ( ξ )W ( ξ )
268 differential equations

Example 8.1. Consider the nonhomogeneous differential equation

y′′ − y′ − 6y = 20e−2x .

We seek a particular solution to this equation. First, we note two

linearly independent solutions of this equation are

y1 ( x ) = e3x , y2 ( x ) = e−2x .

So, the particular solution takes the form

y p ( x ) = c1 ( x )e3x + c2 ( x )e−2x .

We just need to determine the ci ’s.

Since this problem is not in self-adjoint form, we will use
f (x) g( x )
= = 20e−2x
p( x ) a2 ( x )
as seen above. Then the linear system we have to solve is

c1′ ( x )e3x + c2′ ( x )e−2x = 0.

3c1′ ( x )e3x − 2c2′ ( x )e−2x = 20e−2x . (8.18)

Multiplying the first equation by 2 and adding the equations yields

5c1′ ( x )e3x = 20e−2x ,

or
c1′ ( x ) = 4e−5x .
Inserting this back into the first equation in the system, we have

4e−2x + c2′ ( x )e−2x = 0,

leading to
c2′ ( x ) = −4.
These equations are easily integrated to give
4
c1 ( x ) = − e−5x , c2 ( x ) = −4x.
5
Therefore, the particular solution has been found as

y p (x) = c1 ( x )e3x + c2 ( x )e−2x

4
= − e−5x e3x − 4xe−2x
5
4
= − e−2x − 4xe−2x . (8.19)
5
Noting that the first term can be absorbed into the solution of the
homogeneous problem. So, the particular solution can simply be writ-
ten as
y p ( x ) = −4xe−2x .
This is the answer you would have found had you used the Modified
Method of Undetermined Coefficients from Section 1.1.4.
 green’s functions 269

Example 8.2. Revisiting the last example, y′′ − y′ − 6y = 20e−2x .

The formal solution in Equation (8.15) was not used in the last ex-
ample. Instead, we proceeded from the Linear System for Variation of
Parameters earlier in this section. This is the more natural approach
towards finding the particular solution of the nonhomogeneous equa-
tion. Since we will be using Equation (8.15) to obtain solutions to
initial value and boundary value problems, it might be useful to use it
to solve this problem.
From the last example we have
y1 ( x ) = e3x , y2 ( x ) = e−2x .
We need to compute the Wronskian:

e3x e−2x
W ( x ) = W (y1 , y2 )( x ) = = −5e x .
3e3x −2e−2x
Also, we need p( x ), which is given by
 Z 
p( x ) = exp − dx = e− x .

So, we see that p( x )W ( x ) = −5. It is indeed constant, just as we had

proven earlier.
Finally, we need f ( x ). Here is where one needs to be careful as
the original problem was not in self-adjoint form. We have from the
original equation that g( x ) = 20e−2x and a2 ( x ) = 1. So,
p( x )
f (x) = g( x ) = 20e−3x .
a2 ( x )
Now we are ready to construct the solution.
Z x Z x
f ( ξ ) y1 ( ξ ) f ( ξ ) y2 ( ξ )
y p (x) = y2 ( x ) dξ − y1 ( x ) dξ
x1 p ( ξ )W ( ξ ) x0 p ( ξ )W ( ξ )
20e−3ξ e3ξ 20e−3ξ e−2ξ
Z x Z x
= e−2x dξ − e3x dξ
x1 −5 x0 −5
Z x Z x
= −4e−2x dξ + 4e3x e−5x dξ
x1 x0
4 x x
= −4ξe−2x − e3x e−5ξ
x1 5 x0
4 4
= −4xe−2x − e−2x + 4x1 e−2x + e−5x0 e3x . (8.20)
5 5
Note that the first two terms we had found in the last example.
The remaining two terms are simply linear combinations of y1 and y2 .
Thus, we really have the solution to the homogeneous problem con-
tained within the solution when we use the arbitrary constant limits in
the integrals. In the next section we will make use of these constants
when solving initial value and boundary value problems.
In the next section we will determine the unknown constants subject to ei-
ther initial conditions or boundary conditions. This will allow us to combine
the two integrals and then determine the appropriate Green’s functions.
270 differential equations

8.4 Initial and Boundary Value Green’s Functions

We begin with the particular solution (8.15) of the nonhomo-

geneous differential equation (8.5). This can be combined with the
general solution of the homogeneous problem to give the general solution
of the nonhomogeneous differential equation:

Z x Z x
f ( ξ ) y1 ( ξ ) f ( ξ ) y2 ( ξ )
y ( x ) = c1 y1 ( x ) + c2 y2 ( x ) + y2 ( x ) dξ − y1 ( x )
dξ.
x1 p ( ξ )W ( ξ )
p ( ξ )W ( ξ ) x0
(8.21)
As seen in the last section, an appropriate choice of x0 and x1 could be
found so that we need not explicitly write out the solution to the homoge-
neous problem, c1 y1 ( x ) + c2 y2 ( x ). However, setting up the solution in this
form will allow us to use x0 and x1 to determine particular solutions which
satisfies certain homogeneous conditions.
We will now consider initial value and boundary value problems. Each
type of problem will lead to a solution of the form
Z b
y ( x ) = c1 y1 ( x ) + c2 y2 ( x ) + G ( x, ξ ) f (ξ ) dξ, (8.22)
a
where the function G ( x, ξ ) will be identified as the Green’s function and
the integration limits will be found on the integral. Having identified the
Green’s function, we will look at other methods in the last section for deter-
mining the Green’s function.

8.4.1 Initial Value Green’s Function

We begin by considering the solution of the initial value problem

dy( x )
 
d
p( x ) + q ( x ) y ( x ) = f ( x ).
dx dx
y (0) = y0 , y ′ (0) = v0 . (8.23)

Of course, we could have studied the original form of the differential equa-
tion without writing it in self-adjoint form. However, this form is useful
when studying boundary value problems. We will return to this point later.
We first note that we can solve this initial value problem by solving two
separate initial value problems. We assume that the solution of the homo-
geneous problem satisfies the original initial conditions:
dyh ( x )
 
d
p( x ) + q( x )yh ( x ) = 0.
dx dx
y h (0) = y0 , y′h (0) = v0 . (8.24)

We then assume that the particular solution satisfies the problem

dy p ( x )
 
d
p( x ) + q ( x ) y p ( x ) = f ( x ).
dx dx
y p (0) = 0, y′p (0) = 0. (8.25)
 green’s functions 271

Since the differential equation is linear, then we know that y( x ) = yh ( x ) +

y p ( x ) is a solution of the nonhomogeneous equation. However, this solution
satisfies the initial conditions:

y (0) = y h (0) + y p (0) = y0 + 0 = y0 ,

y′ (0) = y′h (0) + y′p (0) = v0 + 0 = v0 .

Therefore, we need only focus on solving for the particular solution that
satisfies homogeneous initial conditions.
Recall Equation (8.15) from the last section,
Z x Z x
f ( ξ ) y1 ( ξ ) f ( ξ ) y2 ( ξ )
y p ( x ) = y2 ( x ) dξ − y1 ( x ) dξ. (8.26)
x1 p ( ξ )W ( ξ ) x0 p ( ξ )W ( ξ )

We now seek values for x0 and x1 which satisfies the homogeneous initial
conditions, y p (0) = 0 and y′p (0) = 0.
First, we consider y p (0) = 0. We have
Z 0 Z 0
f ( ξ ) y1 ( ξ ) f ( ξ ) y2 ( ξ )
y p (0) = y2 (0) dξ − y1 (0) dξ. (8.27)
x1 p ( ξ )W ( ξ ) x0 p ( ξ )W ( ξ )

Here, y1 ( x ) and y2 ( x ) are taken to be any solutions of the homogeneous

differential equation. Let’s assume that y1 (0) = 0 and y2 ̸= (0) = 0. Then
we have Z 0
f ( ξ ) y1 ( ξ )
y p (0) = y2 (0) dξ. (8.28)
x 1 p ( ξ )W ( ξ )

We can force y p (0) = 0 if we set x1 = 0.

Now, we consider y′p (0) = 0. First we differentiate the solution and find
that
Z x Z x
f ( ξ ) y1 ( ξ ) f ( ξ ) y2 ( ξ )
y′p ( x ) = y2′ ( x ) dξ − y1′ ( x ) dξ, (8.29)
0 p ( ξ )W ( ξ ) x0 p ( ξ )W ( ξ )

since the contributions from differentiating the integrals will cancel. Evalu-
ating this result at x = 0, we have
Z 0
f ( ξ ) y2 ( ξ )
y′p (0) = −y1′ (0) dξ. (8.30)
x0 p ( ξ )W ( ξ )

Assuming that y1′ (0) ̸= 0, we can set x0 = 0.

Thus, we have found that
x f (ξ )y (ξ )
Z x f (ξ )y (ξ )
Z
1 2
y p (x) = y2 ( x ) dξ − y1 ( x ) dξ.
0 p ( ξ )W ( ξ ) 0 p ( ξ )W ( ξ )
Z x
y1 ( ξ ) y2 ( x ) − y1 ( x ) y2 ( ξ )
 
= f (ξ ) dξ. (8.31)
0 p(ξ )Wξ )

This result is in the correct form and we can identify the temporal, or
initial value, Green’s function. So, the particular solution is given as
Z x
y p (x) = G ( x, ξ ) f (ξ ) dξ, (8.32)
0
272 differential equations

where the initial value Green’s function is defined as

y1 ( ξ ) y2 ( x ) − y1 ( x ) y2 ( ξ )
G ( x, ξ ) = .
p(ξ )Wξ )
We summarize

Solution of Initial Value Problem (8.23)

The solution of the initial value problem (8.23) takes the form
Z x
y( x ) = yh ( x ) + G ( x, ξ ) f (ξ ) dξ, (8.33)
0

where
y1 ( ξ ) y2 ( x ) − y1 ( x ) y2 ( ξ )
G ( x, ξ ) =
p(ξ )Wξ )
and the solution of the homogeneous problem satisfies the initial condi-
tions,
yh (0) = y0 , y′h (0) = v0 .
Here y1 ( x ) and y2 ( x ) are solutions of the homogeneous equation satisfy-
ing
y1 (0) = 0, y2 (0) ̸= 0, y1′ (0) ̸= 0, y2′ (0) = 0.

Example 8.3. Solve the forced oscillator problem

x ′′ + x = 2 cos t, x (0) = 4, x ′ (0) = 0.

This is a simple problem which could be solved using the Modi-

fied Method of Undetermined Coefficients in Section 1.1.4. Here we
will determine the initial value Green’s function. We first solve the
homogeneous problem with nonhomogeneous initial conditions:

xh′′ + xh = 0, xh (0) = 4, xh′ (0) = 0.

The solution is easily seen to be xh (t) = 4 cos t.

Next, we construct the Green’s function. We need two linearly inde-
pendent solutions, y1 ( x ), y2 ( x ), to the homogeneous differential equa-
tion satisfying y1 (0) = 0 and y2′ (0) = 0. So, we pick y1 (t) = sin t and
y2 (t) = cos t. The Wronskian is found as

W (t) = y1 (t)y2′ (t) − y1′ (t)y2 (t) = − sin2 t − cos2 t = −1.

Since p(t) = 1 in this problem, we have

y1 ( τ ) y2 ( t ) − y1 ( t ) y2 ( τ )
G (t, τ ) =
p(τ )Wτ )
= sin t cos τ − sin τ cos t
= sin(t − τ ). (8.34)

Note that the Green’s function depends on t − τ. While this is useful

in some contexts, we will use the expanded form.
 green’s functions 273

We can now determine the particular solution of the nonhomoge-

neous differential equation. We have
Z t
x p (t) = G (t, τ ) f (τ ) dτ
0
Z t
= (sin t cos τ − sin τ cos t) (2 cos τ ) dτ
0
Z t Z t
2
= 2 sin t cos τdτ − 2 cos t sin τ cos τdτ
0 0
 t  t
τ 1 1 2
= 2 sin t + sin 2τ − 2 cos t sin τ
2 2 0 2 0
= t sin t. (8.35)

Therefore, the particular solution is x (t) = 4 cos t + t sin t. This is what

you would obtain using the Modified Method of Undetermined Coef-
ficients.
Remember, you should verify your solutions. First, x (0) = 4 and
′
x (0) = −4 sin(0) + sin(0) = 0. Furthermore,

x ′′ = −2 cos t − t sin t = 2 cos t − x (t).

So, x ′′ + x = 2 cos t and x (0) = 4, x ′ (0) = 0.

As noted in an earlier section, we usually are not given the differential
equation in self-adjoint form. Generally, it takes the form

a2 ( x )y′′ ( x ) + a1 ( x )y′ ( x ) + a0 ( x )y( x ) = g( x ). (8.36)

Using Equation (6.9), the driving term becomes

g( x )
f ( x ) = p( x ) .
a2 ( x )

Inserting this into the Green’s function form of the particular solution, we
obtain the following:

Solution Using the Initial Value Green’s Function

The solution of the initial value problem,

a2 ( x )y′′ ( x ) + a1 ( x )y′ ( x ) + a0 ( x )y( x ) = g( x ), yh (0) = y0 , y′h (0) = v0

takes the form Z x

y( x ) = yh ( x ) + G ( x, ξ ) g(ξ ) dξ, (8.37)
0
where the Green’s function is the piecewise defined function

y1 ( ξ ) y2 ( x ) − y1 ( x ) y2 ( ξ )
G ( x, ξ ) = (8.38)
a 2 ( ξ )W ( ξ )

and y1 ( x ) and y2 ( x ) are solutions of the homogeneous equation satisfying

y1 (0) = 0, y2 (0) ̸= 0, y1′ (0) ̸= 0, y2′ (0) = 0.

274 differential equations

8.4.2 Boundary Value Green’s Function

We now turn to boundary value problems. We will focus on the

problem

dy( x )
 
d
p( x ) + q ( x ) y ( x ) = f ( x ), a < x < b,
dx dx
y( a) = 0, y(b) = 0. (8.39)

However, the general theory works for other forms of homogeneous bound-
ary conditions.
Once again, we seek x0 and x1 in the form
Z x Z x
f ( ξ ) y1 ( ξ ) f ( ξ ) y2 ( ξ )
y ( x ) = y2 ( x ) dξ − y1 ( x ) dξ
x1 p ( ξ )W ( ξ ) x0 p ( ξ )W ( ξ )

so that the solution to the boundary value problem can be written as a

single integral involving a Green’s function. Here we absorb yh ( x ) into the
integrals with an appropriate choice of lower limits on the integrals.
We first pick solutions of the homogeneous differential equation such
that y1 ( a) = 0, y2 (b) = 0 and y1 (b) ̸= 0, y2 ( a) ̸= 0. So, we have
Z a Z a
f ( ξ ) y1 ( ξ ) f ( ξ ) y2 ( ξ )
y( a) = y2 ( a ) dξ − y1 ( a) dξ
x1 p ( ξ )W ( ξ ) x0 p ( ξ )W ( ξ )
Z a
f ( ξ ) y1 ( ξ )
= y2 ( a ) dξ. (8.40)
x1 p ( ξ )W ( ξ )

This expression is zero if x1 = a.

At x = b we find that
Z b Z b
f ( ξ ) y1 ( ξ ) f ( ξ ) y2 ( ξ )
y(b) = y2 ( b ) dξ − y1 (b) dξ
x1 p ( ξ )W ( ξ ) x0 p ( ξ )W ( ξ )
Z b
f ( ξ ) y2 ( ξ )
= − y1 ( b ) dξ. (8.41)
x0 p ( ξ )W ( ξ )

This vanishes for x0 = b.

So, we have found that
Z x Z x
f ( ξ ) y1 ( ξ ) f ( ξ ) y2 ( ξ )
y ( x ) = y2 ( x ) dξ − y1 ( x ) dξ. (8.42)
a p ( ξ )W ( ξ ) b p ( ξ )W ( ξ )

We are seeking a Green’s function so that the solution can be written as one
integral. We can move the functions of x under the integral. Also, since
a < x < b, we can flip the limits in the second integral. This gives
Z x Z b
f ( ξ ) y1 ( ξ ) y2 ( x ) f ( ξ ) y1 ( x ) y2 ( ξ )
y( x ) = dξ + dξ. (8.43)
a p ( ξ )W ( ξ ) x p ( ξ )W ( ξ )

This result can be written in a compact form:

 green’s functions 275

Boundary Value Green’s Function

The solution of the boundary value problem takes the form

Z b
y( x ) = G ( x, ξ ) f (ξ ) dξ, (8.44)
a

where the Green’s function is the piecewise defined function


 y1 ( ξ ) y2 ( x ) , a ≤ ξ ≤ x

pW

G ( x, ξ ) = . (8.45)
y 1 ( x ) y2 ( ξ )
x≤ξ≤b


pW


The Green’s function satisfies several properties, which we will explore

further in the next section. For example, the Green’s function satisfies the
boundary conditions at x = a and x = b. Thus,

y1 ( a ) y2 ( ξ )
G ( a, ξ ) = = 0,
pW

y1 ( ξ ) y2 ( b )
G (b, ξ ) = = 0.
pW
Also, the Green’s function is symmetric in its arguments. Interchanging the
arguments gives

 y1 ( x ) y2 ( ξ ) , a ≤ x ≤ ξ

pW

G (ξ, x ) = . (8.46)
y 1 ( ξ ) y2 ( x )
ξ≤x≤b


pW


But a careful look at the original form shows that This property is often referred to as reci-
procity. Another was to look at this
is to say that the effect at x from a
G ( x, ξ ) = G (ξ, x ).
source at ξ is the same as the effect at
ξ from a source at x. This is a conse-
We will make use of these properties in the next section to quickly deter- quence of Green’s Reciprocation Theo-
mine the Green’s functions for other boundary value problems. rem, as found in Problem 1.11 in .
John David Jackson. Classical Electrody-
Example 8.4. Solve the boundary value problem y′′ = x2 , y(0) = namics. John Wiley & Sons, 1962
0 = y(1) using the boundary value Green’s function.
We first solve the homogeneous equation, y′′ = 0. After two integra-
tions, we have y( x ) = Ax + B, for A and B constants to be determined.
We need one solution satisfying y1 (0) = 0 Thus, 0 = y1 (0) = B. So,
we can pick y1 ( x ) = x, since A is arbitrary.
The other solution has to satisfy y2 (1) = 0. So, 0 = y2 (1) = A + B.
This can be solved for B = − A. Again, A is arbitrary and we will
choose A = −1. Thus, y2 ( x ) = 1 − x.
For this problem p( x ) = 1. Thus, for y1 ( x ) = x and y2 ( x ) = 1 − x,

p( x )W ( x ) = y1 ( x )y2′ ( x ) − y1′ ( x )y2 ( x ) = x (−1) − 1(1 − x ) = −1.

Note that p( x )W ( x ) is a constant, as it should be.

276 differential equations

Now we construct the Green’s function. We have


 − ξ (1 − x ), 0 ≤ ξ ≤ x

G ( x, ξ ) = . (8.47)
 − x (1 − ξ ), x ≤ ξ ≤ 1


Notice the symmetry between the two branches of the Green’s func-
tion. Also, the Green’s function satisfies homogeneous boundary con-
ditions: G (0, ξ ) = 0, from the lower branch, and G (1, ξ ) = 0, from the
upper branch.
Finally, we insert the Green’s function into the integral form of the
solution:
Z 1
y( x ) = G ( x, ξ ) f (ξ ) dξ
0
Z 1
= G ( x, ξ )ξ 2 dξ
0
Z x Z 1
= − ξ (1 − x )ξ 2 dξ − x (1 − ξ )ξ 2 dξ
0 x
Z x Z 1
= −(1 − x ) ξ 3 dξ − x (ξ 2 − ξ 3 ) dξ
0 x
x 1
ξ4
 3
ξ4

ξ
= −(1 − x ) −x −
4 0 3 4 x
1 1 1
= − (1 − x ) x4 − x (4 − 3) + x (4x3 − 3x4 )
4 12 12
1 4
= ( x − x ). (8.48)
12

8.5 Properties of Green’s Functions

We have noted some properties of Green’s functions in the last

section. In this section we will elaborate on some of these properties as a tool
for quickly constructing Green’s functions for boundary value problems.
Here is a list of the properties based upon our previous solution. We now
show how a knowledge of these properties allows one to quickly construct
a Green’s function.

Properties of the Green’s Function

3
Note that this is a partial differential 1. Differential Equation: 3
∂G ( x, ξ )
 
equation since G ( x, ξ ) is a function of ∂
two variables. We can treat ξ as a pa- p( x ) + q( x ) G ( x, ξ ) = 0, x ̸= ξ
rameter and solve the eqwuation as if it
∂x ∂x
were an ordinary differential equation. For x < ξ we are on the second branch and G ( x, ξ ) is proportional to
y1 ( x ). Thus, since y1 ( x ) is a solution of the homogeneous equation, then
so is G ( x, ξ ). For x > ξ we are on the first branch and G ( x, ξ ) is propor-
tional to y2 ( x ). So, once again G ( x, ξ ) is a solution of the homogeneous
problem.
2. Boundary Conditions:
 green’s functions 277

For x = a we are on the second branch and G ( x, ξ ) is proportional

to y1 ( x ). Thus, whatever condition y1 ( x ) satisfies, G ( x, ξ ) will satisfy. A
similar statement can be made for x = b.
3. Symmetry or Reciprocity: G ( x, ξ ) = G (ξ, x )
We had shown this in the last section.
4. Continuity of G at x = ξ: G (ξ + , ξ ) = G (ξ − , ξ )
Here we have defined

G (ξ + , x ) = lim G ( x, ξ ), x > ξ,
x ↓ξ

G (ξ − , x ) = lim G ( x, ξ ), x < ξ.
x ↑ξ

Setting x = ξ in both branches, we have

y1 ( ξ ) y2 ( ξ ) y ( ξ ) y2 ( ξ )
= 1 .
pW pW

∂G
5. Jump Discontinuity of at x = ξ:
∂x
∂G (ξ + , ξ ) ∂G (ξ − , ξ ) 1
− =
∂x ∂x p(ξ )

This case is not as obvious. We first compute the derivatives by noting

which branch is involved and then evaluate the derivatives and subtract
them. Thus, we have

∂G (ξ + , ξ ) ∂G (ξ − , ξ ) 1 1 ′
− = − y1 (ξ )y2′ (ξ ) + y ( ξ ) y2 ( ξ )
∂x ∂x pW pW 1
y1′ (ξ )y2 (ξ ) − y1 (ξ )y2′ (ξ )
= −
p(ξ )(y1 (ξ )y2′ (ξ ) − y1′ (ξ )y2 (ξ ))
1
= . (8.49)
p(ξ )

Example 8.5. Construct the Green’s function for the problem

y′′ + ω 2 y = f ( x ), 0 < x < 1,

y (0) = 0 = y (1),
with ω ̸= 0.

I. Find solutions to the homogeneous equation.

A general solution to the homogeneous equation is given as

yh ( x ) = c1 sin ωx + c2 cos ωx.

Thus, for x ̸= ξ,

G ( x, ξ ) = c1 (ξ ) sin ωx + c2 (ξ ) cos ωx.

278 differential equations

II. Boundary Conditions.

First, we have G (0, ξ ) = 0 for 0 ≤ x ≤ ξ. So,

G (0, ξ ) = c2 (ξ ) cos ωx = 0.

So,
G ( x, ξ ) = c1 (ξ ) sin ωx, 0 ≤ x ≤ ξ.

Second, we have G (1, ξ ) = 0 for ξ ≤ x ≤ 1. So,

G (1, ξ ) = c1 (ξ ) sin ω + c2 (ξ ) cos ω. = 0

A solution can be chosen with

c2 (ξ ) = −c1 (ξ ) tan ω.

This gives
G ( x, ξ ) = c1 (ξ ) sin ωx − c1 (ξ ) tan ω cos ωx.
This can be simplified by factoring out the c1 (ξ ) and placing the remain-
ing terms over a common denominator. The result is

c1 ( ξ )
G ( x, ξ ) = [sin ωx cos ω − sin ω cos ωx ]
cos ω
c (ξ )
= − 1 sin ω (1 − x ). (8.50)
cos ω
Since the coefficient is arbitrary at this point, as can write the result as

G ( x, ξ ) = d1 (ξ ) sin ω (1 − x ), ξ ≤ x ≤ 1.

We note that we could have started with y2 ( x ) = sin ω (1 − x ) as one

of the linearly independent solutions of the homogeneous problem in
anticipation that y2 ( x ) satisfies the second boundary condition.

III. Symmetry or Reciprocity

We now impose that G ( x, ξ ) = G (ξ, x ). To this point we have that

c1 (ξ ) sin ωx, 0≤x≤ξ


G ( x, ξ ) = .
 d1 (ξ ) sin ω (1 − x ), ξ ≤ x ≤ 1


We can make the branches symmetric by picking the right forms for c1 (ξ )
and d1 (ξ ). We choose c1 (ξ ) = C sin ω (1 − ξ ) and d1 (ξ ) = C sin ωξ. Then,

 C sin ω (1 − ξ ) sin ωx, 0 ≤ x ≤ ξ

G ( x, ξ ) = .
 C sin ω (1 − x ) sin ωξ, ξ ≤ x ≤ 1


Now the Green’s function is symmetric and we still have to determine

the constant C. We note that we could have gotten to this point using the
1
Method of Variation of Parameters result where C = .
pW
 green’s functions 279

IV. Continuity of G ( x, ξ )
We note that we already have continuity by virtue of the symmetry im-
posed in the last step.
∂
V. Jump Discontinuity in G ( x, ξ ).
∂x
We still need to determine C. We can do this using the jump discontinuity
of the derivative:
∂G (ξ + , ξ ) ∂G (ξ − , ξ ) 1
− = .
∂x ∂x p(ξ )
For our problem p( x ) = 1. So, inserting the Green’s function, we have
∂G (ξ + , ξ ) ∂G (ξ − , ξ )
1 = −
∂x ∂x
∂ ∂
= [C sin ω (1 − x ) sin ωξ ] x=ξ − [C sin ω (1 − ξ ) sin ωx ] x=ξ
∂x ∂x
= −ωC cos ω (1 − ξ ) sin ωξ − ωC sin ω (1 − ξ ) cos ωξ
= −ωC sin ω (ξ + 1 − ξ )
= −ωC sin ω. (8.51)

Therefore,
1
C=− .
ω sin ω
Finally, we have the Green’s function:

 − sin ω (1 − ξ ) sin ωx ,

 0≤x≤ξ
G ( x, ξ ) = ω sin ω . (8.52)
 − sin ω (1 − x ) sin ωξ ,

ξ≤x≤1
ω sin ω
It is instructive to compare this result to the Variation of Parameters re-
sult. We have the functions y1 ( x ) = sin ωx and y2 ( x ) = sin ω (1 − x ) as the
solutions of the homogeneous equation satisfying y1 (0) = 0 and y2 (1) = 0.
We need to compute pW:

p ( x )W ( x ) = y1 ( x )y2′ ( x ) − y1′ ( x )y2 ( x )

= −ω sin ωx cos ω (1 − x ) − ω cos ωx sin ω (1 − x )
= −ω sin ω (8.53)

Inserting this result into the Variation of Parameters result for the Green’s
function leads to the same Green’s function as above.

8.5.1 The Dirac Delta Function

We will develop a more general theory of Green’s functions for

ordinary differential equations which encompasses some of the listed prop-
erties. The Green’s function satisfies a homogeneous differential equation
for x ̸= ξ,
∂G ( x, ξ )
 
∂
p( x ) + q( x ) G ( x, ξ ) = 0, x ̸= ξ. (8.54)
∂x ∂x
280 differential equations

When x = ξ, we saw that the derivative has a jump in its value. This is
similar to the step, or Heaviside, function,

 1, x > 0

H (x) = .
 0, x < 0


In the case of the step function, the derivative is zero everywhere except at
the jump. At the jump, there is an infinite slope, though technically, we have
learned that there is no derivative at this point. We will try to remedy this
by introducing the Dirac delta function,

d
δ( x ) = H ( x ).
dx
We will then show that the Green’s function satisfies the differential equa-
tion
∂G ( x, ξ )
 
∂
p( x ) + q( x ) G ( x, ξ ) = δ( x − ξ ). (8.55)
∂x ∂x
The Dirac delta function, δ( x ), is one example of what is known as a gen-
eralized function, or a distribution. Dirac [1927] had introduced this function
in the 1920’s in his study of quantum mechanics as a useful tool. It was
later studied in a general theory of distributions and found to be more than
a simple tool used by physicists. The Dirac delta function, as any distribu-
tion, only makes sense under an integral.
Before defining the Dirac delta function and introducing some of its prop-
erties, we will look at some representations that lead to the definition. We
will consider the limits of two sequences of functions.
First we define the sequence of functions
4

 0, | x | > 1

n
f n (x) = .
n
 , |x| < 1

2 n

This is a sequence of functions as shown in Figure 8.1. As n → ∞, we find

the limit is zero for x ̸= 0 and is infinite for x = 0. However, the area
under each member of the sequences is one since each box has height n2 and
2 width n2 . Thus, the limiting function is zero at most points but has area one.
(At this point the reader who is new to this should be doing some head
scratching!)
The limit is not really a function. It is a generalized function. It is called
1
the Dirac delta function, which is defined by

1. δ( x ) = 0 for x ̸= 0.
R∞
2. −∞ δ( x ) dx = 1.

- 21 - 41 - 18 1 1 1 x Another example is the sequence defined by

8 4 2

Figure 8.1: A plot of the functions f n ( x ) 2 sin nx

for n = 2, 4, 8. Dn ( x ) = . (8.56)
x
 green’s functions 281

We can graph this function. We first rewrite this function as y

8
sin nx
Dn ( x ) = 2n .
nx
Now it is easy to see that as x → 0, Dn ( x ) → 2n. For large x, The function
tends to zero. A plot of this function is in Figure 8.2. For large n the peak
grows and the values of Dn ( x ) for x ̸= 0 tend to zero as show in Figure 8.3.
We note that in the limit n → ∞, Dn ( x ) = 0 for x ̸= 0 and it is infinite at x
x = 0. However, using complex analysis one can show that the area is −5 5
Z ∞ −2
Dn ( x ) dx = 2π.
−∞
Figure 8.2: A plot of the function Dn ( x )
for n = 4.
Thus, the area is constant for each n.
y
There are two main properties that define a Dirac delta function. First
80
one has that the area under the delta function is one,
Z ∞
δ( x ) dx = 1.
−∞

Integration over more general intervals gives

Z b
δ( x ) dx = 1, 0 ∈ [ a, b]
a

and Z b
δ( x ) dx = 0, 0∈
/ [ a, b]. x
a −3 3
Another common property is what is sometimes called the sifting property.
Namely, integrating the product of a function and the delta function “sifts” −20
out a specific value of the function. It is given by Figure 8.3: A plot of the function DΩ ( x )
Z ∞ for Ω = 40.
δ( x − a) f ( x ) dx = f ( a).
−∞

This can be seen by noting that the delta function is zero everywhere except
at x = a. Therefore, the integrand is zero everywhere and the only contribu-
tion from f ( x ) will be from x = a. So, we can replace f ( x ) with f ( a) under
the integral. Since f ( a) is a constant, we have that
Z ∞ Z ∞ Z ∞
δ( x − a) f ( x ) dx = δ( x − a) f ( a) dx = f ( a) δ( x − a) dx = f ( a).
−∞ −∞ −∞

Another property results from using a scaled argument, ax. In this case
we show that
δ( ax ) = | a|−1 δ( x ). (8.57)
As usual, this only has meaning under an integral sign. So, we place δ( ax )
inside an integral and make a substitution y = ax:
Z ∞ Z L
δ( ax ) dx = lim δ( ax ) dx
−∞ L→∞ − L
Z aL
1
= lim δ(y) dy. (8.58)
L→∞ a − aL
282 differential equations

If a > 0 then Z ∞ Z ∞
1
δ( ax ) dx = δ(y) dy.
−∞ a −∞

However, if a < 0 then

Z ∞ Z −∞ Z ∞
1 1
δ( ax ) dx = δ(y) dy = − δ(y) dy.
−∞ a ∞ a −∞

The overall difference in a multiplicative minus sign can be absorbed into

one expression by changing the factor 1/a to 1/| a|. Thus,
Z ∞ Z ∞
1
δ( ax ) dx = δ(y) dy. (8.59)
−∞ | a| −∞
R∞
Example 8.6. Evaluate −∞ (5x + 1)δ(4( x − 2)) dx. This is a straight
forward integration:
Z ∞ Z ∞
1 11
(5x + 1)δ(4( x − 2)) dx = (5x + 1)δ( x − 2) dx = .
−∞ 4 −∞ 4

A more general scaling of the argument takes the form δ( f ( x )). The in-
tegral of δ( f ( x )) can be evaluated depending upon the number of zeros of
f ( x ). If there is only one zero, f ( x1 ) = 0, then one has that
Z ∞ Z ∞
1
δ( f ( x )) dx = δ( x − x1 ) dx.
−∞ −∞ | f ′ ( x1 )|

This can be proven using the substitution y = f ( x ) and is left as an exercise

for the reader. This result is often written as
1
δ( f ( x )) = δ ( x − x1 ).
| f ′ ( x1 )|
R∞
Example 8.7. Evaluate −∞ δ(3x − 2) x2 dx.
This is not a simple δ( x − a). So, we need to find the zeros of
f ( x ) = 3x − 2. There is only one, x = 23 . Also, | f ′ ( x )| = 3. Therefore,
we have
Z ∞ Z ∞  2
1 2 1 2 4
δ(3x − 2) x2 dx = δ( x − ) x2 dx = = .
−∞ −∞ 3 3 3 3 27

Note that this integral can be evaluated the long way by using the
substitution y = 3x − 2. Then, dy = 3dx and x = (y + 2)/3. This gives
Z ∞ Z ∞ 2
y+2
  
21 1 4 4
δ(3x − 2) x dx = δ(y) dy = = .
−∞ 3 −∞ 3 3 9 27

More generally, one can show that when f ( x j ) = 0 and f ′ ( x j ) ̸= 0 for x j ,

j = 1, 2, . . . , n, (i.e.; when one has n simple zeros), then

n
1
δ( f ( x )) = ∑ | f ′ (x j )| δ(x − x j ).
j =1
 green’s functions 283

R 2π
Example 8.8. Evaluate 0 cos x δ( x2 − π 2 ) dx.
In this case the argument of the delta function has two simple roots.
Namely, f ( x ) = x2 − π 2 = 0 when x = ±π. Furthermore, f ′ ( x ) = 2x.
Therefore, | f ′ (±π )| = 2π. This gives
1
δ( x2 − π 2 ) = [δ( x − π ) + δ( x + π )].
2π
Inserting this expression into the integral and noting that x = −π is
not in the integration interval, we have
Z 2π 2π
1
Z
cos x δ( x2 − π 2 ) dx = cos x [δ( x − π ) + δ( x + π )] dx
0 2π 0
1 1
= cos π = − . (8.60) H (x)
2π 2π
1
Finally, we previously noted there is a relationship between the Heavi-
side, or step, function and the Dirac delta function. x
0
Example 8.9. Show H ′ ( x ) = δ( x ), where the Heaviside function (or,
Figure 8.4: The Heaviside step function,
step function) is defined as H ( x ).

 0, x < 0

H (x) =
 1, x > 0


and is shown in Figure 8.4.

Looking at the plot, it is easy to see that H ′ ( x ) = 0 for x ̸= 0 and
is undefined for x = 0. In order to check that this gives the delta
function, we need to compute the area integral. Therefore, we have
Z ∞ ∞
H ′ ( x ) dx = H ( x ) = 1 − 0 = 1.
−∞ −∞

Thus, H ′ ( x ) satisfies the two properties of the Dirac delta function.

This argument might not be satisfying to some. Here we seem to
have 
 0, x ̸= 0

′
H (x) =
 ∞, x = 0


This is not a normal function. It is what is called a distribution, or gen-

eralized function. It is zero everywhere except it has an infinite spike at
x = 0.
Distributions only make sense when integrated against a test function.
So, we need a class of test functions, ϕ( x ). These are infinitely differentiable
functions, C ∞ , which vanish at infinity, lim x → ±∞ = 0.
Now, we consider H ′ ( x ) under the an integral with a test function,.
Z ∞ Z ∞
∞
H ′ ( x )ϕ( x ) dx = H ( x )ϕ( x ) −∞
− H ( x )ϕ′ ( x ) dx
−∞ −∞
Z ∞
= − ϕ′ ( x ) dx = ϕ(0). (8.61)
0

Thus, we see that H ′ ( x ) sifts out the value of the test function at x = 0. This
is precisely the behavior we have seen with the Dirac delta function.
284 differential equations

8.5.2 Green’s Function Differential Equation

4
At the end of this section we show As noted, the Green’s function satisfies the differential equation4
that the Green’s function is the response
∂G ( x, ξ )
 
(solution) to the impulse (source term) ∂
as governed by the differential equation p( x ) + q( x ) G ( x, ξ ) = δ( x − ξ ) (8.62)
∂x ∂x
(8.62.
and satisfies homogeneous conditions. We have used the Green’s function
to solve the nonhomogeneous equation
dy( x )
 
d
p( x ) + q ( x ) y ( x ) = f ( x ). (8.63)
dx dx
These equations can be written in the more compact forms

L[y] = f ( x )
L[ G ] = δ( x − ξ ). (8.64)

Multiplying the first equation by G ( x, ξ ), the second equation by y( x ), and

then subtracting, we have

G L[y] − yL[ G ] = f ( x ) G ( x, ξ ) − δ( x − ξ )y( x ).

Now, integrate both sides from x = a to x = b. The left side becomes

Z b Z b
[ f ( x ) G ( x, ξ ) − δ( x − ξ )y( x )] dx = f ( x ) G ( x, ξ ) dx − y(ξ )
a a

and, using Green’s Identity, the right side is

Z b    x=b
′ ∂G
( G L[y] − yL[ G ]) dx = p( x ) G ( x, ξ )y ( x ) − y( x ) ( x, ξ ) .
a ∂x x=a

Combining these results and rearranging, we obtain

Z b    x=b
∂G ′
y(ξ ) = f ( x ) G ( x, ξ ) dx − p( x ) y( x ) ( x, ξ ) − G ( x, ξ )y ( x ) .
a ∂x x=a
(8.65)
Next, one uses the boundary conditions in the problem in order to deter-
mine which conditions the Green’s function needs to satisfy. For example, if
we have the boundary condition y( a) = 0 and y(b) = 0, then the boundary
terms yield
Z b   
∂G ′
y(ξ ) = f ( x ) G ( x, ξ ) dx − p(b) y(b) (b, ξ ) − G (b, ξ )y (b)
a ∂x
  
∂G ′
+ p( a) y( a) ( a, ξ ) − G ( a, ξ )y ( a)
∂x
Z b
= f ( x ) G ( x, ξ ) dx + p(b) G (b, ξ )y′ (b) − p( a) G ( a, ξ )y′ ( a). (8.66)
a

The right hand side will only vanish if G ( x, ξ ) also satisfies these homoge-
neous boundary conditions. This then leaves us with the solution
Z b
y(ξ ) = f ( x ) G ( x, ξ ) dx.
a
 green’s functions 285

We should rewrite this as a function of x. So, we replace ξ with x and x

with ξ. This gives
Z b
y( x ) = f (ξ ) G (ξ, x ) dξ. (8.67)
a
However, this is not yet in the desirable form. The arguments of the Green’s
function are reversed. But, G ( x, ξ ) is symmetric in its arguments. So, we
can simply switch the arguments getting the desired result.5 5
If f ( x ) = δ( x − x0 ) in Equation (8.63),
We can now see that the theory works for other boundary conditions. If then the solution is f ( x ) = G ( x, x0 ).
However, from Equation (8.67) we have
we had y′ ( a) = 0, then the y( a) ∂G
∂x ( a, ξ ) term in the boundary terms could be Z b
made to vanish if we set ∂x ( a, ξ ) = 0. So, this confirms that other boundary
∂G G ( x, x0 ) = δ(ξ − x0 ) G (ξ, x ) dξ
a
value problems can be posed besides the one elaborated upon in the chapter = G ( x0 , x ).
so far. Therefore, we have established reci-
We can even adapt this theory to nonhomogeneous boundary conditions. procity. Note that this works this way
because L is self-adjoint. If it were not,
We first rewrite Equation (8.65) as
then we would need the adjoint Green’s
Z b   ξ =b function.
∂G
y( x ) = G ( x, ξ ) f (ξ ) dξ − p(ξ ) y(ξ ) ( x, ξ ) − G ( x, ξ )y′ (ξ ) .
a ∂ξ ξ =a
(8.68)
′
Let’s consider the boundary conditions y( a) = α and y (b) = β. We also
assume that G ( x, ξ ) satisfies homogeneous boundary conditions,
∂G
G ( a, ξ ) = 0, (b, ξ ) = 0.
∂ξ
in both x and ξ since the Green’s function is symmetric in its variables.
Then, we need only focus on the boundary terms to examine the effect on
the solution. We have
  ξ =b
∂G ′
p(ξ ) y(ξ ) ( x, ξ ) − G ( x, ξ )y (ξ )
∂ξ ξ =a
  
∂G
= p(b) y(b) ( x, b) − G ( x, b)y′ (b)
∂ξ
  
∂G
− p( a) y( a) ( x, a) − G ( x, a)y′ ( a)
∂ξ
∂G
= − βp(b) G ( x, b) − αp( a) ( x, a). (8.69)
∂ξ
Therefore, we have the solution
Z b
∂G
y( x ) = G ( x, ξ ) f (ξ ) dξ + βp(b) G ( x, b) + αp( a) ( x, a). (8.70)
a ∂ξ
This solution satisfies the nonhomogeneous boundary conditions. Let’s see
how it works.
Example 8.10. Modify Example 8.4 to solve the boundary value prob-
lem y′′ = x2 , y(0) = 1, y(1) = 2 using the boundary value Green’s
function that we found:

 − ξ (1 − x ), 0 ≤ ξ ≤ x

G ( x, ξ ) = . (8.71)
 − x (1 − ξ ), x ≤ ξ ≤ 1

286 differential equations

We insert the Green’s function into the solution and use the given
conditions to obtain
Z 1   ξ =1
2 ∂G ′
y( x ) = G ( x, ξ )ξ dξ − y(ξ ) ( x, ξ ) − G ( x, ξ )y (ξ )
0 ∂ξ ξ =0
Z x Z 1
= ( x − 1)ξ 3 dξ + x (ξ − 1)ξ 2 dξ
0 x
∂G ∂G
+ y (0) ( x, 0) − y(1) ( x, 1)
∂ξ ∂ξ
1 4
= ( x − x ) + (1 − x ) + 2x
12
x4 11
= + x + 1. (8.72)
12 12
Of course, this problem can be solved more directly by direct inte-
gration. The general solution is
x4
+ c1 x + c2 .
y( x ) =
12
Inserting this solution into each boundary condition yields the same
result.
We have seen how the introduction of the Dirac delta function in the
differential equation satisfied by the Green’s function, Equation (8.62), can
lead to the solution of boundary value problems. The Dirac delta function
also aids in our interpretation of the Green’s function. We note that the
Green’s function is a solution of an equation in which the nonhomogeneous
function is δ( x − ξ ). Note that if we multiply the delta function by f (ξ ) and
integrate we obtain
Z ∞
δ( x − ξ ) f (ξ ) dξ = f ( x ).
−∞
We can view the delta function as a unit impulse at x = ξ which can be
used to build f ( x ) as a sum of impulses of different strengths, f (ξ ). Thus,
the Green’s function is the response to the impulse as governed by the dif-
ferential equation and given boundary conditions.
In particular, the delta function forced equation can be used to derive the
jump condition. We begin with the equation in the form
∂G ( x, ξ )
 
∂
p( x ) + q( x ) G ( x, ξ ) = δ( x − ξ ). (8.73)
∂x ∂x
Now, integrate both sides from ξ − ϵ to ξ + ϵ and take the limit as ϵ → 0.
Then,
Z ξ +ϵ  Z ξ +ϵ
∂G ( x, ξ )
  
∂
lim p( x ) + q( x ) G ( x, ξ ) dx = lim δ( x − ξ ) dx
ϵ →0 ξ − ϵ ∂x ∂x ϵ →0 ξ − ϵ
= 1. (8.74)

Since the q( x ) term is continuous, the limit of that term vanishes. Using the
Fundamental Theorem of Calculus, we then have
∂G ( x, ξ ) ξ +ϵ
 
lim p( x ) = 1. (8.75)
ϵ →0 ∂x ξ −ϵ
 green’s functions 287

This is the jump condition that we have been using!

8.6 The Adjoint Green’s Function

In the previous discussion, we have relied on the differential opera-

tor to be self-adjoint and constructed Green’s functions based on properties
of Sturm-Liouville operators. However, we can start with a general differ-
ential operator and not put it into Sturm-Liouville form. So, if G ( x, ξ ) is
the Green’s function satisfying LG ( x, ξ ) = δ( x − ξ ), then there is an adjoint
Green’s function, G A ( x, ξ ), satisfying L† G A ( x, ξ ) = δ( x − ξ ). We will see
that the adjoint Green’s function will come in handy to prove that we can
use G ( x, ξ ) in ways similar to the earlier examples.
We will show how this works by solving a specific problem of the form
π
y′′ + 2y′ + 2y = f ( x ), y(0) = 0, y = 0. (8.76)
2

This problem takes the form Ly = f , where L = D2 + 2D + 2 is not a self-

adjoint operator. The adjoint operator is given by L† = D2 − 2D + 2. We can
also put this problem in Sturm-Liouville form,

dy( x )
 
d
Ly( x ) = e2x + 2e2x y( x ) = e2x f ( x ). (8.77)
dx dx

We seek the Green’s functions associated with each operator satisfying

homogeneous boundary conditions and the equations

LG ( x, ξ ) = δ( x − ξ )
† A
L G ( x, ξ ) = δ( x − ξ )
LG( x, ξ ) = δ( x − ξ ). (8.78)

We then use G ( x, ξ ) and G( x, ξ ) to construct solutions to the boundary value

problem.
We first seek the Green’s function, G ( x, ξ ), satisfying

∂2 G ( x, ξ ) ∂G ( x, ξ )
2
+2 + 2G ( x, ξ ) = δ( x − ξ ), (8.79)
∂x ∂x


and the boundary conditions G (0, ξ ) = 0, G π2 , ξ = 0. We will see that this
Green’s function is not symmetric.
We will then find the adjoint Green’s function, G A ( x, ξ ), satisfying

∂2 G A ( x, ξ ) ∂G A ( x, ξ )
− 2 + 2G A ( x, ξ ) = δ( x − ξ ), (8.80)
∂x2 ∂x

and the boundary conditions G A (0, ξ ) = 0, G A π2 , ξ = 0. We will show

that G (ξ, x ) = G A ( x, ξ ) and use both functions to find the solution to the
boundary value problem. We will then show that this solution is the same
as using the Sturm-Liouville operator.
288 differential equations

Example 8.11. Find the Green’s function satisfying Equation (8.79).

Defining g( x ) = G ( x, ξ ), then for x ̸= ξ,
π
g′′ + 2g′ + 2g = 0, g(0) = 0, g = 0.
2
The characteristic equation is r2 + 2r + 2 = 0. So, r = −1 ± i. This
gives the general solution as

g( x ) = e− x ( a cos x + b sin x ).

For 0 ≤ x ≤ ξ, we find the solution g1 ( x ) satisfying the boundary

condition g1 (0) = 0.

g1 (0) = e0 ( a cos 0 + b sin 0) = a = 0.

So, g1 ( x ) = be− x sin x.

Similarly, we find the solution g2 ( x ), ξ ≤ x ≤ π2 , satisfying the
boundary condition g2 π2 = 0. In this case we find g2 ( x ) = ae− x cos x.

Now we construct the Green’s function. So far, we have the piece-

wise defined function

 be− x sin x, 0 ≤ x ≤ ξ,

G ( x, ξ ) = (8.81)
 ae− x cos x, ξ ≤ x ≤ π .

2

The first condition is that G ( x, ξ ) be continuous at x = ξ. This gives

be−ξ sin ξ = ae−ξ cos ξ.

This can be satisfied by defining

a = c sin ξ, b = c cos ξ.

So, we have

 ce− x sin x cos ξ,

0 ≤ x ≤ ξ,
G ( x, ξ ) = (8.82)
 ce− x sin ξ cos x, ξ≤x≤ π
2.


∂G ( x,ξ )
The next condition is that ∂x is discontinuous at x = ξ. We show
this by integrating Equation (8.79) over the interval x ∈ [ξ − ϵ, ξ +
ϵ]. Using the definition of the Dirac delta function and continuity of
G ( x, ξ ), we let ϵ approach zero to obtain
Z ξ +ϵ  2 Z ξ +ϵ
∂ G ( x, ξ ) ∂G ( x, ξ )

+2 + 2G ( x, ξ ) dx = δ( x − ξ ) dx
ξ −ϵ ∂x2 ∂x ξ −ϵ
ξ +ϵ Z ∞
∂G ( x, ξ )

lim + 2G ( x, ξ ) = δ( x − ξ ) dx
ϵ →0 ∂x ξ −ϵ −∞
ξ +
∂G ( x, ξ )

= 1 (8.83)
∂x ξ−
 green’s functions 289

This gives a jump condition for the discontinuity of the derivative at

x = ξ where ξ + is the value above x = ξ and ξ − is the value below
x = ξ.
We can apply this to G ( x, ξ ). Namely, we have
 +
∂G ( x, ξ ) ξ

1 =
∂x ξ−
h i
= ce−ξ − sin ξ cos ξ − sin2 ξ − (− sin ξ cos ξ + cos2 ξ )
= −ce−ξ .
So, c = −eξ and the Green’s function is

 −eξ − x sin x cos ξ,

0 ≤ x ≤ ξ,
G ( x, ξ ) = (8.84)
 −eξ − x sin ξ cos x, ξ≤x≤ π
2.


We see that G ( x, ξ ) is not symmetric, G ( x, ξ ) ̸= G (ξ, x ). Thus, it seems

that G ( x, ξ ) does not satisfy a reciprocity condition.
Example 8.12. Find the adjoint Green’s function satisfying Equation
(8.80).
The derivation parallels that for G ( x, ξ ). However, in this case we
start with the general solution g A ( x ) = G A ( x, ξ ), for
′′ ′
π
g A − 2g A + 2g A = 0, x ̸= ξ, g A (0) = 0, g A =0
2
as
g A ( x ) = e x ( a cos x + b sin x ).
Following the same steps as before, we find the adjoint Green’s
function,

 −e x−ξ sin x cos ξ, 0 ≤ x ≤ ξ,

G A ( x, ξ ) = (8.85)
 −e x−ξ sin ξ cos x, ξ ≤ x ≤ π .

2

While G A ( x, ξ ) ̸= G A (ξ, x ), we do have that G ( x, ξ ) = G A (ξ, x )

gives the form of a reciprocity condition.
Example 8.13. Derive a general relation between G ( x, ξ ) and G A ( x, ξ ).
From the last example we have found G ( x, ξ ) = G A (ξ, x ) for a spe-
cific problem. In general these Green’s functions satisfy the equations
for x ∈ [ a, b]:

LG ( x, ξ ) = δ( x − ξ )
′
A A
L G ( x, ξ ) = δ ( x − ξ ′ ). (8.86)

Multiply the first equation by G A ( x, ξ ′ ) and the second equation by

G ( x, ξ ). Subtract and integrate
Z bh i
G A ( x, ξ ′ ) LG ( x, ξ ) − G ( x, ξ ) L A G A ( x, ξ ′ ) dx
a
Z bh i
= G A ( x, ξ ′ )δ( x − ξ ) − G ( x, ξ )δ( x − ξ ′ ) dx. (8.87)
a
290 differential equations

Assuming appropriate boundary conditions, we have

Z b Z b
G A ( x, ξ ′ ) LG ( x, ξ ) dx = G ( x, ξ ) L A G A ( x, ξ ′ ) dx.
a a

So, after applying the Dirac delta function integrations, we have

G A (ξ, ξ ′ ) = G (ξ ′ , ξ ).

Now we can return to the original problem but adding nonhomogeneous

boundary conditions.
Example 8.14. Use the adjoint Green’s function to solve
π
y′′ + 2y′ + 2y = f ( x ), y(0) = A, y = B. (8.88)
2
Defining L = D2 + 2D + 2, we have

Ly( x ) = f (x)
A A
L G ( x, ξ ) = δ ( x − ξ ). (8.89)

As with the previous example, we multiply the first equation by G A ( x, ξ )

and the second equation by y( x ). Subtracting and integrating we have
Z π/2 h
G A ( x, ξ ) y′′ ( x ) + 2y′ ( x ) + 2y( x )


0
∂2 G A ( x, ξ ) ∂G A ( x, ξ )
 
−y − 2 + 2G A ( x, ξ ) dx
∂x2 ∂x
Z π/2 h i
= G A ( x, ξ ) f ( x ) − y( x )δ( x − ξ ) dx (8.90)
0

Cleaning this up, we find that

∂G A ( x, ξ )
Z π/2  
∂
G A ( x, ξ )y′ ( x ) − y( x ) + 2y( x ) G A ( x, ξ ) dx
0 ∂x ∂x
Z π/2
= G A ( x, ξ ) f ( x ) dx − y(ξ ), (8.91)
0

or, after applying boundary conditions,

Z π/2
y(ξ ) = G A ( x, ξ ) f ( x ) dx
0
π/2
∂G A ( x, ξ )

− G A ( x, ξ )y′ ( x ) − y( x ) + 2y( x ) G A ( x, ξ )
∂x 0
∂G A ( π2 , ξ ) ∂G A (0, ξ )
Z π/2
= G A ( x, ξ ) f ( x ) dx + B− A. (8.92)
0 ∂x ∂x
Since G A ( x, ξ ) = G (ξ, x ), we can exchange variables to obtain the so-
lution in terms of the Green’s function, G ( x, ξ ):

∂G A ( π2 , x ) ∂G A (0, x )
Z π/2
y( x ) = G A (ξ, x ) f (ξ ) dξ + B− A
0 ∂ξ ∂ξ
∂G ( x, π2 ) ∂G ( x, 0)
Z π/2
= G ( x, ξ ) f (ξ ) dξ + B− A. (8.93)
0 ∂ξ ∂ξ
 green’s functions 291

We now apply this general solution to a specific problem.

Example 8.15. Use Equation (8.93) with the Green’s function in Equa-
tion (8.84) to solve
π
y′′ + 2y′ + 2y = 5 sin x, y(0) = 2e, y = 0. (8.94)
2

∂G ( x, π2 ) ∂G ( x, 0)
Z π/2
y( x ) = G ( x, ξ ) f (ξ ) dξ + B− A
0 ∂ξ ∂ξ
Z x
∂G ( x, 0)
Z π/2
= G ( x, ξ ) f (ξ ) dξ + G ( x, ξ ) f (ξ ) dξ − 2e
0 x ∂ξ
Z xh i Z π/2 h i
= −eξ − x sin ξ cos x 5 sin(ξ ) dξ + −eξ − x sin x cos ξ 5 sin(ξ ) dξ
0 x
∂ h ξ −x i
−2e −e sin ξ cos x
∂ξ ξ =0
Z x Z π/2
= −5e− x cos x eξ sin2 ξ dξ − 5e− x sin x eξ sin ξ cos ξ dξ + 2e1− x cos x
0 x
 
−x 2 1 x 2 2 x 2 x
= −5e − + e sin x − e sin x cos x + e cos x
5 5 5 5
  
1 1 1 π
+ − e x sin 2x + ( e x cos 2x + e 2 sin x + 2e1− x cos x
10 5 5
= 2e− x cos x − 2 cos x + sin x − e 2 − x sin x + 2e1− x cos x
π

 
= 2 e− x (1 + e) − 1 cos x + 1 − e 2 − x sin x.
  π
(8.95)

We can check this solution by using the Method of Undetermined

Coefficients to obtain the solution. We know the solution to the homo-
geneous equation is

yh ( x ) = e− x (c1 sin x + c2 cos x ).

We seek a particular solution,

y p ( x ) = c3 sin x + c4 cos x.

Inserting into the differential equation, we have

[c3 − 2c4 ] sin x + [c4 + 2c3 ] cos x = 5 sin x.

This is true when c3 = 1 and c4 = −2. So, the general solution to the
nonhomogeneous equation is

y( x ) = e− x (c1 sin x + c2 cos x ) + sin x − 2 cos x.

For the solution of the boundary value problem, we need to satisfy

the boundary conditions.

y (0) = c2 − 2 = 2e,
π
= e− 2 c1 + 1 = 0.
π
y (8.96)
2
292 differential equations

π
So, c2 = 2(1 + e) and c1 = −e 2 and the solution is
= e− x (−e 2 sin x + 2(1 + e) cos x ) + sin x − 2 cos x
π
y( x )
 
= 2 e− x (1 + e) − 1 cos x + 1 − e 2 − x sin x.
  π
(8.97)
So, the solutions agree.
We have seen how we can solve for and use the Green’s function and
adjoint Green’s function in an example inolving a non-Hermitian operator.
However, we also know that we can cast the problem in Sturm-Liouville
form. So, how do these methods differ if we used the Sturm-Liouville oper-
ator and its Green’s function?
Example 8.16. Consider the boundary value problem
π
y′′ + 2y′ + 2y = f ( x ), y(0) = A, y = B. (8.98)
2
Put this in Sturm-Liouville form, find its Green’s function, and write
the solution in terms of the Green’s function.
The Sturm-Liouville form of the differential equation is
dy( x )
 
d
e2x + 2e2x y( x ) = e2x f ( x ). (8.99)
dx dx
The associated Green’s function would then satisfy
2x ∂ G( x, ξ )
 
∂
e + 2e2x G( x, ξ ) = δ( x − ξ ). (8.100)
∂x ∂x
Following Example 8.11, we find the Green’s function satisfying the


homogeneous boundary conditions, G (0, ξ ) = 0, G π2 , ξ = 0, takes
the form

 ce− x sin x cos ξ, 0 ≤ x ≤ ξ,

G( x, ξ ) = (8.101)
 ce− x sin ξ cos x, ξ ≤ x ≤ π .

2

So far, this function is continuous at x = ξ.

We need to derive the jump condition for the discontinuity of the
derivative of the Green’s function. As before, we integrate the Green’s
function equation over the interval x ∈ [ξ − ϵ, ξ + ϵ] to obtain
Z ξ +ϵ  Z ξ +ϵ
2x ∂ G( x, ξ )
  
∂ 2x
e + 2e G( x, ξ ) dx = δ( x − ξ ) dx
ξ −ϵ ∂x ∂x ξ −ϵ

∂G( x, ξ ) ξ +ϵ
  Z ∞
lim e2x = δ( x − ξ ) dx
ϵ →0 ∂x ξ −ϵ −∞
ξ +
2x ∂ G( x, ξ )

e = 1 (8.102)
∂x ξ−

We can apply this to G( x, ξ ). Namely, we have

 +
∂G( x, ξ ) ξ

1 = e2x
∂x ξ−
h i
= ce2ξ e−ξ − sin ξ cos ξ − sin2 ξ − (− sin ξ cos ξ + cos2 ξ )
= −ceξ .
 green’s functions 293

So, c = −e−ξ and the Green’s function is


 −e− x−ξ sin x cos ξ, 0 ≤ x ≤ ξ,

G( x, ξ ) = (8.103)
 −e− x−ξ sin ξ cos x, ξ ≤ x ≤ π .

2

We see that G( x, ξ ) is symmetric, G( x, ξ ) = G(ξ, x ).

We can now find the solution to Equation (8.98) using this Green’s
function. Defining
 
d 2x d
L= e + 2e2x ,
dx dx

y( x ) and G satisfy the differential equations

Ly( x ) = e2x f ( x )
LG( x, ξ ) = δ( x − ξ ). (8.104)

As with the previous example, we multiply the first equation by G( x, ξ )

and the second equation by y( x ). Subtracting and integrating we have
Z π Z π h i
2 2
[G( x, ξ )Ly( x ) − y( x )LG( x, ξ )] dx = e2x f ( x )G( x, ξ ) − y( x )δ( x − ξ ) dx
0 0
 π Z π
dy( x ) ∂G( x, ξ )
 
2 2 2x
e2x G( x, ξ ) − y( x ) = e f ( x )G( x, ξ ) dx − y(ξ )
dx ∂x 0 0
Z π
2
y(ξ ) = e2x f ( x )G( x, ξ ) dx
0
∂G( π2 , ξ ) ∂G(0, ξ )
+eπ B −A . (8.105)
∂x ∂x
Computing the derivative of the Green’s function,

∂G( x, ξ )  e− x−ξ (sin x cos ξ − cos x cos ξ ), 0 ≤ x ≤ ξ,

= (8.106)
∂x  e− x−ξ (sin ξ cos x + sin ξ sin x ), ξ ≤ x ≤ π .

2

and evaluating it at the boundary points, we have

∂G(0, ξ ) ∂G( π2 , ξ )
= −e−ξ cos ξ, = e− 2 −ξ sin ξ.
π

∂x ∂x
The solution can be written as
Z π
2
e2x f ( x )G( x, ξ ) dx + Be 2 −ξ sin ξ + Ae−ξ cos ξ
π
y(ξ ) =
0

or Z π
2
e2ξ f (ξ )G( x, ξ ) dx + Be 2 − x sin x + Ae− x cos x.
π
y( x ) =
0
Let’s compare this with the solution in Equation (8.93). We need

∂G ( x, ξ )  −eξ − x (sin x cos ξ − sin x sin ξ ), 0 ≤ x ≤ ξ,

= (8.107)
∂ξ  −eξ − x (sin ξ cos x + cos ξ cos x ), ξ ≤ x ≤ π .

2
294 differential equations

Then,
∂G ( x, 0) ∂G ( x, π2 )
= −e− x cos x, = e 2 − x sin x.
π

∂ξ ∂x
Inserting these values and noting that G ( x, ξ ) = e2ξ G( x, ξ )
∂G ( x, π2 ) ∂G ( x, 0)
Z π/2
y( x ) = G ( x, ξ ) f (ξ ) dξ + B− A
0 ∂ξ ∂ξ
Z π/2
e2ξ G( x, ξ ) f (ξ ) dξ + Be 2 − x sin x + Ae− x cos x. (8.108)
π
=
0

Thus, the solutions using the Sturm-Liouville form of the problem and the
original form are the same.

8.7 Series Representations of Green’s Functions

There are times that it might not be so simple to find the Green’s
function in the simple closed form that we have seen so far. However,
there is a method for determining the Green’s functions of Sturm-Liouville
boundary value problems in the form of an eigenfunction expansion. We
will finish our discussion of Green’s functions for ordinary differential equa-
tions by showing how one obtains such series representations. (Note that
we are really just repeating the steps towards developing eigenfunction ex-
pansion which we had seen in Chapter 6.3.)
We will make use of the complete set of eigenfunctions of the differential
operator, L, satisfying the homogeneous boundary conditions:

L[ϕn ] = −λn σϕn , n = 1, 2, . . .

We want to find the particular solution y satisfying L[y] = f and homo-

geneous boundary conditions. We assume that
∞
y( x ) = ∑ an ϕn (x).
n =1

Inserting this into the differential equation, we obtain

∞ ∞
L[y] = ∑ an L[ϕn ] = − ∑ λn an σϕn = f.
n =1 n =1

This has resulted in the generalized Fourier expansion

∞
f (x) = ∑ cn σϕn (x)
n =1

with coefficients
cn = −λn an .
We have seen how to compute these coefficients earlier in the text. We
multiply both sides by ϕk ( x ) and integrate. Using the orthogonality of the
eigenfunctions,
Z b
ϕn ( x )ϕk ( x )σ ( x ) dx = Nk δnk ,
a
 green’s functions 295

one obtains the expansion coefficients (if λk ̸= 0)

( f , ϕk )
ak = − ,
Nk λk
Rb
where ( f , ϕk ) ≡ a f ( x )ϕk ( x ) dx.
As before, we can rearrange the solution to obtain the Green’s function.
Namely, we have
∞ Z b ∞
( f , ϕn ) ϕn ( x )ϕn (ξ )
y( x ) = ∑ − Nn λn
ϕn ( x ) = ∑
a n =1 − Nn λn
f (ξ ) dξ
n =1
| {z }
G ( x,ξ )

Therefore, we have found the Green’s function as an expansion in the eigen-

functions:
∞
ϕn ( x )ϕn (ξ )
G ( x, ξ ) = ∑ . (8.109)
n =1
−λn Nn
Example 8.17. Eigenfunction Expansion Example
We will conclude this discussion with an example. Consider the
boundary value problem

y′′ + 4y = x2 , x ∈ (0, 1), y(0) = y(1) = 0.

The Green’s function for this problem can be constructed fairly quickly
for this problem once the eigenvalue problem is solved. We will solve
this problem three different ways in order to summarize the methods
we have used in the text.
The eigenvalue problem is

ϕ′′ ( x ) + 4ϕ( x ) = −λϕ( x ),

where ϕ(0) = 0 and ϕ(1) = 0. The general solution is obtained by

rewriting the equation as

ϕ′′ ( x ) + k2 ϕ( x ) = 0,

where
k2 = 4 + λ.
Solutions satisfying the boundary condition at x = 0 are of the form

ϕ( x ) = A sin kx.

Forcing ϕ(1) = 0 gives

0 = A sin k ⇒ k = nπ, k = 1, 2, 3 . . . .

So, the eigenvalues are

λn = n2 π 2 − 4, n = 1, 2, . . .

and the eigenfunctions are

ϕn = sin nπx, n = 1, 2, . . . .
296 differential equations

We need the normalization constant, Nn . We have that

Z 1
1
Nn = ∥ϕn ∥2 = sin2 nπx = .
0 2

We can now construct the Green’s function for this problem using
Equation (8.109).
∞
sin nπx sin nπξ
G ( x, ξ ) = 2 ∑ (4 − n2 π 2 )
. (8.110)
n =1

We can use this Green’s function to determine the solution of the

boundary value problem. Thus, we have
Z 1
y( x ) = G ( x, ξ ) f (ξ ) dξ
0
∞
Z 1
!
sin nπx sin nπξ
= 2∑ ξ 2 dξ
0 n =1 (4 − n2 π 2 )
∞ Z 1
sin nπx
= 2 ∑ (4 − n2 π 2 ) 0
ξ 2 sin nπξ dξ
n =1
∞
(2 − n2 π 2 )(−1)n − 2
 
sin nπx
= 2 ∑ 2 2 n3 π 3
. (8.111)
n =1 (4 − n π )

We can compare this solution to the one one would obtain if we

did not employ Green’s functions directly. The eigenfunction expan-
sion method for solving boundary value problems, which we saw in
Section 6.3 proceeds as follows. We assume that the solution is in the
form
∞
y( x ) = ∑ cn ϕn (x).
n =1

Inserting this into the differential equation L[y] = x2 gives

∞
" #
x2 = L ∑ cn sin nπx
n =1
∞
d2
 
= ∑ cn dx2
(sin nπx ) + 4 sin nπx
n =1
∞
= ∑ cn [4 − n2 π2 ] sin nπx (8.112)
n =1

We need the Fourier sine series expansion of x2 on [0, 1] in order to

determine the cn ’s. The Fourier sine series coefficients are given by
Z 1
2
bn = x2 sin nπx
1 0
(2 − n2 π 2 )(−1)n − 2
 
= 2 , n = 1, 2, . . . . (8.113)
n3 π 3
Therefore,
∞
(2 − n2 π 2 )(−1)n − 2
 
x2 = 2 ∑ n3 π 3
sin nπx.
n =1
 green’s functions 297

Inserting this series in Equation (8.112), we find

∞  ∞
(2 − n2 π 2 )(−1)n − 2

2∑ sin nπx = ∑ cn [4 − n2 π 2 ] sin nπx.
n =1 n3 π 3 n =1

Due to the linear independence of the eigenfunctions, we can solve for

the unknown coefficients to obtain
(2 − n2 π 2 )(−1)n − 2
cn = 2 .
(4 − n2 π 2 ) n3 π 3
Therefore, the solution using the eigenfunction expansion method is
∞
y( x ) = ∑ cn ϕn (x)
n =1
∞
(2 − n2 π 2 )(−1)n − 2
 
sin nπx
= 2 ∑ 2 2 n3 π 3
. (8.114)
n =1 (4 − n π )

We note that this is the same solution as in Equation 8.111 which we

had obtained using the Green’s function as an eigenfunction expan-
sion.
One remaining question is the following: Is there a closed form for
the Green’s function and the solution to this problem? The answer
is yes! We note that the differential operator is a special case of the
Example 8.5 done is section 8.4.2. Namely, we pick ω = 2. The Green’s
function was already found in that section. For this special case, we
have

 − sin 2(1 − ξ ) sin 2x , 0 ≤ x ≤ ξ



G ( x, ξ ) = 2 sin 2 . (8.115)
 − sin 2(1 − x ) sin 2ξ , ξ ≤ x ≤ 1

2 sin 2
What about the solution to the boundary value problem? This so-
lution is given by

y( x )
Z 1
= G ( x, ξ ) f (ξ ) dξ
0
Z x Z 1
sin 2(1 − x ) sin 2ξ 2 sin 2(ξ − 1) sin 2x 2
= − ξ dξ + ξ dξ
0 2 sin 2 x 2 sin 2
1 h 2 i
= − − x sin 2 − sin 2 cos2 x + sin 2 + cos 2 sin x cos x + sin x cos x .
4 sin 2
1 h 2 i
= − − x sin 2 + (1 − cos2 x ) sin 2 + sin x cos x (1 + cos 2) .
4 sin 2
1 h 2 i
= − − x sin 2 + 2 sin2 x sin 1 cos 1 + 2 sin x cos x cos2 1) .
4 sin 2
1 h i
= − − x2 sin 2 + 2 sin x cos 1(sin x sin 1 + cos x cos 1) .
8 sin 1 cos 1
x2 sin x cos(1 − x )
= − . (8.116)
4 4 sin 1
In Figure 8.5 we show a plot of this solution along with the first five
terms of the series solution. The series solution converges quickly.
298 differential equations

Figure 8.5: Plots of the exact solution to x

Example 8.17 with the first five terms of 0.0 0.5 1.0
the series solution.
0.0

−0.02

−0.04

−0.06

As one last check, we solve the boundary value problem directly using
the Method of Undetermined Coefficients. Again, the problem is

y′′ + 4y = x2 , x ∈ (0, 1), y(0) = y(1) = 0.

The problem has the general solution

y( x ) = c1 cos 2x + c2 sin 2x + y p ( x ),

where y p is a particular solution of the nonhomogeneous differential equa-

tion. Using the Method of Undetermined Coefficients, we assume a solution
of the form
y p ( x ) = Ax2 + Bx + C.

Inserting this in the nonhomogeneous equation, we have

2A + 4( Ax2 + Bx + C ) = x2 ,

Thus, B = 0, 4A = 1 and 2A + 4C = 0. The solution of this system is

1 1
A= , B = 0, C=− .
4 8
So, the general solution of the nonhomogeneous differential equation is

x2 1
y( x ) = c1 cos 2x + c2 sin 2x + − .
4 8
We now determine the arbitrary constants using the boundary condi-
tions. We have

0 = y (0)
1
= c1 −
8
0 = y (1)
1
= c1 cos 2 + c2 sin 2 + (8.117)
8
1
Thus, c1 = 8 and
1
+ 18 cos 2
c2 = − 8 .
sin 2
 green’s functions 299

Inserting these constants in the solution we find the same solution as before.
" #
1 1
1 8 + 8 cos 2 x2 1
y( x ) = cos 2x − sin 2x + −
8 sin 2 4 8
cos 2x sin 2 − sin 2x cos 2 − sin 2x x2 1
= + −
8 sin 2 4 8
sin2 x sin 1 + sin x cos x cos 1 x2
= − +
4 sin 1 4
x 2 sin x cos(1 − x )
= − . (8.118)
4 4 sin 1

8.8 Appendix: The Rise of Green’s Functions

Green’s functions were first introduced by the British mathemati-

cian George Green (1793-1841) in 1828 [Green, 1889]. George Green sought
solutions of Poisson’s equation Adapted from .
Russell L. Herman. Laplace Trans-
∇2 u (r) = f (r) forms and Green’s Functions. In
V. Martinez-Luaces, editor, Engineering
Problems and the Laplace Transform, pages
for the electric potential u defined inside a bounded volume with specified 1–112. Nova Science, 2025
boundary conditions on the surface of the volume. He introduced a func-
tion now identified as what Riemann later coined the “Green’s function”6 6
Grattan-Guinness [1995] suggests that
[Cannell, 2001, Cannell and Lord, 1993, Duffy, 2015]. However, the concept both Riemann and Carl Neumann (1832-
1925) contributed to the name.
of Green’s functions was not widely understood or used until much later,
and it wasn’t until the 20th century that they became a central tool in many
areas of physics and applied mathematics.7 7
Green also introduced potential theory
George Green was self-taught as he had about one year of formal school- and Green’s Theorem in his studies of
electricity and magnetism.
ing between 8 and 9. [See [Cannell, 2001, Cannell and Lord, 1993].] He
lived most of his life in Sneinton, Nottinghamshire. His father, George, was
a baker who built and owned a brick windmill to grind grain. In 1828 the
younger George Green published his famous essay, An Essay on the Applica-
tion of Mathematical Analysis to the Theories of Electricity and Magnetism. He
published it privately at his expense and sold 51 copies to mostly friends.
[For more on the history, see [Cannell, 2001, Cannell and Lord, 1993, Challis
and Sheard, 2003, Tazzioli, 2001].]
Green’s father died in 1829 and left an inheritance to his son and daugh-
ter. Green later met a wealthy landowner and mathematician, Edward
Bromhead (1789 - 1855), who had bought a copy of the essay and encour-
aged Green to further his studies. But, Green did not contact Bromhead
for two years when in 1833 Green enrolled in Gonville and Caius College,
Cambridge. He wrote several papers but in his final years at Cambridge,
Green became ill and in 1840 he returned to Sneinton, only to die a year
later [Cannell, 1999, 2001, Cannell and Lord, 1993, Grattan-Guinness, 1995].
Green’s work was not well known during his lifetime. He had done little
to promote his essay. But he had shared copies with Carl Jacobi (1804-1851)
and William Hopkins (1793-1866) [Cannell and Lord, 1993, Craik, 2008,
300 differential equations

Grattan-Guinness, 1995]. In 1845 a young William Thomson (1824-1907),

later to be known as Lord Kelvin and who was interested in electricity and
magnetism, saw a footnote in Robert Murphy’s (1806-1843) 1832 paper, On
the inverse method of definite integrals, with physical applications, referring to
Green’s essay. He searched for the essay and on the evening before going
to Paris on a fellowship he had obtained copies of the essay from William
8
William Hopkins (1793-1866) was fa- Hopkins 8 . While in Paris he met with a number of mathematicians sharing
mous for coaching students at Cam-
what he had found, including Joseph Liouville (1809-1882), Michel Chasles
bridge University to become senior
wranglers by scoring the best on the (1793-1880). Jacques Charles François Sturm (1803-1855), Augustin-Louis
Mathematical Tripos. Thomson and Cauchy (1789-1857) and Jean-Baptiste Biot (1775-1862).
Stokes were two of many to later be no-
table mathematicians. Thomson saw the importance of Green’s work and saw to it that it was
published and popularised. The essay was published in three parts in
Crelle’s Journal 1850, 1852, and 1854 [Cannell, 1999]. In 1871 Norman M.
Ferrers (1829 - 1903) assembled The Mathematical Papers of the Late George
9
Green’s essay can now be found at Green9 [Green, 1871a].
arXiv:0807.0088. There were several contributions that Green made in these papers. In
his 1928 essay Green introduced potential functions, although there were
others working in the area [Grattan-Guinness, 1995].It is believed that Green
coined the word “potential.” He also introduced what we now call Green’s
Theorem and Green’s identities. He made use of these in the introduction
of what we now call Green’s functions.
Green later wrote a paper on electrostatics and then turned to attrac-
tions of ellipsoids, elasticity, and water waves. In one of his papers, On
the Motion of Waves in a Variable Canal ..., Green gave an early version of
10
The WKB Approximation, named the Wentzel–Kramers–Brillouin (WKB) approximation10 [Merzbacher, 1998,
after Wentzel–Kramers–Brillouin, is a Berera and Del Debbio, 2021]. He wrote on the motion of small water waves
method to approximate solutions to
linear differential equations with spa- in a thin, shallow rectangular canal. According to Cannell and Lord [1993],
tially varying coefficients. According to he was first to note that “particles in a deep sea wave execute circular mo-
Schlissel [1977], this method was dis-
cussed by several researchers such as tion” and was an early commentator on John Scott Russell (1808 - 1882)
George Green (1793-1841), Joseph Liou- work on solitary waves [Green, 1871b, Herman, 1992]. Also, in 1839 Green
ville (1809-1882), Lord Rayleigh (1842- contributed to the study of vibrations of an elastic solid and the Cauchy-
1909), Richard Gans (1880-1954), Harold
Jeffreys (1891-1989), only to have the Green tensor is named after him.
names of Gregor Wentzel (1898-1978), William Thomson and George Stokes (1819-1903), both prominent math-
Henrich A. Kramers (1884-1952), and
Leon Brillouin (1889-1969) stick as the ematicians and physicists of the 19th century, did not explicitly use the term
“WKB” approximation (or, method). “Green’s function” in their work, as the concept was not yet widely under-
stood or named as such. However, they both made significant contributions
to the development of the underlying mathematical concepts that would
later become known as Green’s functions.
Thomson, for example, developed what is now known as the Kelvin
method, which is a technique for solving Laplace’s equation by decompos-
ing the solution into a sum of simpler functions. This method is closely
related to the use of Green’s functions to solve partial differential equations.
Thomson also introduced the method of images and the Kelvin transforma-
tion, or inversion in a sphere.
Stokes, on the other hand, made important contributions to the study of
fluid dynamics and the behavior of viscous fluids. He introduced what is
 green’s functions 301

now known as the Stokes equation, which describes the motion of a viscous
fluid in response to an external force. The solution to this equation involves
the use of integral operators that are closely related to Green’s functions. He
is best know to calculus students for an integral theorem, Stokes’ Theorem.
11 11
Histories of this theorem can be found
So while neither Thomson nor Stokes explicitly used the term “Green’s in [Crowe, 1984, Katz, 1979]. Maxwell sat
for the Smith Prize in 1854 and took first
function” in their work, they both made important contributions to the un- place with Edward Routh and Maxwell
derlying mathematical concepts and techniques that would later become [1873], stated “This theorem was given
by Professor Stokes. Smith’s Prize Ex-
central to the use of Green’s functions in physics and applied mathematics. amination, 1854, question 8. It is proved
Also, Kirchhoff [1857, 1883] mentions Green’s functions, has a delta function in Thomson and Tait’s Natural Philoso-
and references work by J. Fröhlich [1878] and Woldemar Voigt (1850–1918)12 phy, § 190 (f)” [Kelvin and Tait, 1867].
Copies of Smith Prize questions are in
[Voigt, 1878]. Gustav Kirchhoff (1824-1887) introduced what is now known [Stokes, 1905]. [Tait, 1870] later relates
as the Kirchhoff integral theorem, which is essentially a form of Green’s Stokes’ Theorem to Green’s Theorem us-
ing quaternions and the ∇ operator.
function. Similarly, Bernhard Riemann (1826-1866) introduced the concept 12
Kirchhoff, Voight and Minnigerode
of a “potentialschrift,” which is essentially a Green’s function for the Laplace were students of Franz Neumann (1798-
equation. In [Riemann, 1857] he used Abelian integrals to determine a 1895), father of Carl Neumann.
Green’s function. Bottazzini and Tazzioli [1995] describe Riemann’s work
in natural philosophy (physics) and how his “complex function theory is
deeply connected with potential theory in two dimensions.”
These researchers used Green’s functions to study a range of physical
phenomena, such as the propagation of waves [Riemann, 1860], the behav-
ior of electric and magnetic fields, and the diffusion of heat. One of the
earliest uses of Green’s functions for the heat equation was by Bernhard
Minnigerode (1837–1896)) [Minnigerode, 1862, 1886].
So while the term "Green’s function" may not have been in use in the
1800s, the concept itself was certainly being explored and applied by a
number of researchers in the field of mathematics and physics. More re-
cently, Green’s functions found there way into quantum field theory. In
1928, P. A. M. Dirac (1902-1984) used Green’s functions to develop his the-
ory of quantum mechanics [Dirac, 1928]. Richard Feynman (1918-1988) used
Green’s functions to develop his path integral formulation of quantum me-
chanics [Feynman, 1950, Feynman et al., 2010]. Julian Schwinger (1918-1994)
used Green’s functions to develop his theory of quantum electrodynamics
[Schwinger, 1948, 2001] and Freeman Dyson (1923-2020) [Dyson, 1949, 2001]
added to the understanding of these works. The Schwinger–Dyson equa-
tion is used to find the Green functions of quantum field theories. Green’s
functions are also commonly taught in electromagnetism [Franklin, 2017].

Problems

1. Use the Method of Variation of Parameters to determine the general

solution for the following problems.

a. y′′ + y = tan x.
b. y′′ − 4y′ + 4y = 6xe2x .
302 differential equations

2. Instead of assuming that c1′ y1 + c2′ y2 = 0 in the derivation of the solu-

tion using Variation of Parameters, assume that c1′ y1 + c2′ y2 = h( x ) for an
arbitrary function h( x ) and show that one gets the same particular solution.

3. Find the solution of each initial value problem using the appropriate
initial value Green’s function.

a. y′′ − 3y′ + 2y = 20e−2x , y(0) = 0, y′ (0) = 6.

b. y′′ + y = 2 sin 3x, y(0) = 5, y′ (0) = 0.
c. y′′ + y = 1 + 2 cos x, y(0) = 2, y′ (0) = 0.
d. x2 y′′ − 2xy′ + 2y = 3x2 − x, y(1) = π, y′ (1) = 0.

4. Consider the problem y′′ = sin x, y′ (0) = 0, y(π ) = 0.

a. Solve by direct integration.

b. Determine the Green’s function.
c. Solve the boundary value problem using the Green’s function.
d. Change the boundary conditions to y′ (0) = 5, y(π ) = −3.
i. Solve by direct integration.
ii. Solve using the Green’s function.

5. Consider the problem:

∂2 G ∂G
= δ ( x − x0 ), (0, x0 ) = 0, G (π, x0 ) = 0.
∂x2 ∂x
a. Solve by direct integration.
b. Compare this result to the Green’s function in part b of the last
problem.
c. Verify that G is symmetric in its arguments.

6. In this problem you will show that the sequence of functions

 
n 1
f n (x) =
π 1 + n2 x 2

approaches δ( x ) as n → ∞. Use the following to support your argument:

a. Show that limn→∞ f n ( x ) = 0 for x ̸= 0.

b. Show that the area under each function is one.

7. Verify that the sequence of functions { f n ( x )}∞

n=1 , defined by f n ( x ) =
n −n| x |
2e ,
approaches a delta function.

8. Evaluate the following integrals:

Rπ

a. 0 sin xδ x − π2 dx.
R∞
b. −∞ δ x− 5 2x
3x2 − 7x + 2 dx.



3 e
c. 0 x2 δ x + π2 dx.
Rπ

 green’s functions 303

R∞
d. 0 e−2x δ( x2 − 5x + 6) dx. [See Problem 10.]
R∞ 2
e. −∞ ( x − 2x + 3)δ( x2 − 9) dx. [See Problem 10.]

9. Find a Fourier series representation of the Dirac delta function, δ( x ), on

[− L, L].

10. For the case that a function has multiple simple roots, f ( xi ) = 0,
f ′ ( xi ) ̸= 0, i = 1, 2, . . . , it can be shown that
n
δ ( x − xi )
δ( f ( x )) = ∑ | f ′ ( xi )|
.
i =1
R∞
Use this result to evaluate −∞ δ( x2 − 5x + 6)(3x2 − 7x + 2) dx.

11. Consider the boundary value problem: y′′ − y = x, x ∈ (0, 1), with
boundary conditions y(0) = y(1) = 0.

a. Find a closed form solution without using Green’s functions.

b. Determine the closed form Green’s function using the properties
of Green’s functions. Use this Green’s function to obtain a solution
of the boundary value problem.
c. Determine a series representation of the Green’s function. Use
this Green’s function to obtain a solution of the boundary value
problem.
d. Confirm that all of the solutions obtained give the same results.