C

op
yr
ig
ht
ed
M
at

Chapter 1
er
ia
l–

Historical Development
Ta
y lo
r&

One of the fundamental problems of field theory1 is the construction

Fr

of solutions to linear differential equations when there is a specified source

an

and the differential equation must satisfy certain boundary conditions. The
purpose of this book is to show how Green’s functions provide a powerful
ci
s

method for obtaining these solutions. In this chapter, we present a historical

overview of their evolution.

1.1 MR. GREEN’S ESSAY

In 1828 George Green (1793–1841) published an Essay on the Application

of Mathematical Analysis to the Theory of Electricity and Magnetism. In this
seminal work of mathematical physics, Green sought to determine the electric
potential within a vacuum bounded by conductors with specified potentials.
In today’s notation we would say that he examined the solutions of ∇2 u =
−f within a volume V that satisfy certain boundary conditions along the
boundary S.

1 Any theory in which the basic quantities are fields, such as electromagnetic theory.

© 2015 by Taylor & Francis Group, LLC

2 Green’s Functions with Applications

To solve this problem, Green first considered a problem where the source
is a point charge. In modern notation, he sought to solve the partial differen-
tial equation:
∇2 g(r|r0 ) = −4πδ(r − r0 ), (1.1.1)
where δ(r − r0 ) is the Dirac delta function. We now know that the solution
to Equation 1.1.1 is g = 1/R, where R2 = (x − ξ)2 + (y − η)2 + (z − ζ)2 .
Recognizing the singular nature of g, he proceeded as follows:
First Green proved the theorem that bears his name:
""" ""
# 2 2
$
ϕ∇ χ − χ∇ ϕ dV = ⊂⊃ (ϕ∇χ − χ∇ϕ) · n dS, (1.1.2)
V S

where the outwardly pointing normal is denoted by n and χ and ϕ are scalar
functions that possess bounded derivatives. Then, by introducing a small ball
C
op

about the singularity at r0 (because Equation 1.1.2 cannot apply there) and
yr

then excluding it from the volume V , he obtained

ig

""" "" """ ""

ht

g∇2 u dV + ⊂⊃ g∇u · n dS = u∇2 g dV + ⊂⊃ u∇g · n dS − 4πu(r0 )

ed

V S V S
(1.1.3)
M

because the surface integral over the small ball is 4πu(r0 ) as the radius of
at
er

the ball tends to zero. Next, Green required that g satisfies the homogeneous
ia

boundary condition g = 0 along the surface S. Since ∇2 u = −f and ∇2 g = 0

l–

within V (recall that the point r0 is excluded from V ), he found that

Ta

""
1
u(r) = ⊂⊃ u ∇g · n dS, (1.1.4)
y lo

4π S
r&

when f = 0 (Laplace’s equation) for any point r within S. Here u denotes the
Fr

value of u on S. This solved the boundary-value problem once g was found.

an

Green knew that g had to exist; it physically described the electrical potential
ci

from a point charge located at r0 .

s

Green’s essay remained relatively unknown until it was published2 at

the urging of Kelvin between 1850 and 1854. Later Poincaré3 summarized
our knowledge of Green’s functions near the turn of the twentieth century.
The subsequent evolution of Green’s functions can be divided into two parts:
before and after the publication in 1946 of Methods of Theoretical Physics by
P. M. Morse and H. Feshbach.4 In this paper-back version of classnotes that

2 Green, G., 1850, 1852, 1854: An essay on the application of mathematical analysis to

the theories of electricity and magnetism. J. Reine Angew. Math., 39, 73–89; 44, 356–374;
47, 161–221.
3Poincaré, H., 1894: Sur les équations de la physique mathématique. Rend. Circ. Mat.
Palermo, 8, 57–156.
4 Morse, P. M., and H. Feshbach, 1946: Methods of Theoretical Physics. MIT Technol-

ogy Press, 497 pp.

© 2015 by Taylor & Francis Group, LLC

Historical Development 3

they developed since the late 1930s to teach mathematical methods to physics
graduate students, they laid out the four properties that a Green’s function
must possess. Using the Sturm-Liouville problem given by
% &
d dy
f (x) + p(x)y = −q(x), (1.1.5)
dx dx

these four properties are:

• The Green’s function satisfies the homogeneous differential equation when

x &= ξ, the source point.
• The Green’s function satisfies homogeneous boundary conditions.
• The Green’s function is symmetric in the variables x,ξ .
• The Green’s function g(x|ξ) satisfies the condition
C
op

' '
dg '' dg '' 1
yr

' − ' =− . (1.1.6)

ig

dx x=ξ+ dx x=ξ− f (ξ)

ht
ed

Prior to the publication of Morse and Feshbach’s notes, authors used var-
M

ious tricks to find Green’s functions that satisfied these four properties. Morse
at

and Feshbach’s great contribution was to show that the “Green’s function is
er

the point source solution [to a boundary-value problem] satisfying appropriate

ia

boundary conditions.” Thus the Green’s function could be found by simply

l–

solving (in the case of Sturm-Liouville problem)

Ta

% &
y

d dg
lo

f (x) + p(x)g = −δ(x − ξ) (1.1.7)

dx dx
r&
Fr

with homogeneous boundary conditions, where δ(x − ξ) was the recently in-
an

troduced delta function by Dirac. The advantage of this formulation was

ci

that the powerful techniques of eigenvalue expansions and transform methods

s

could be used in a straightforward manner to find Green’s functions. They

will be the primary methods used in this book.
By the 1960’s many textbooks began to champion the use of Green’s
functions. For example, in Mackie’s 1965 book5 he sought “to give a general
account of how certain mathematical techniques, notably those of Green’s
functions and of integral transforms, can be used to solve important and com-
monly occurring boundary value problems in ordinary and partial differential
equations.” In the following sections we turn to the development of Green’s
functions as they evolved within each general class of differential equations.

5 Mackie, A. G., 1965: Boundary Value Problems. Oliver & Boyd, 252 pp.

© 2015 by Taylor & Francis Group, LLC

4 Green’s Functions with Applications

1.2 POTENTIAL EQUATION

Shortly after the publication of Green’s monograph on the European con-

tinent, the German mathematician and pedagogue Carl Gottfried Neumann
(1832–1925) developed the concept of Green’s function as it applies to the
two-dimensional (in contrast to three-dimensional) potential equation.6 He
defined the two-dimensional Green’s function, showed that it possesses the
property of reciprocity, and found that it behaves as ln(r) as r → ∞. Using
elliptic coordinates he rederived Poisson’s integral formula and developed an
eigenfunction expansion for the two-dimensional Green’s function. In 1875
Paul Meutzner (1849–1914) extended Neumann’s work.7 In particular, he ob-
tained the Green’s function for the region within an ellipse (Ellipsenfläche)
and a circle (Ringfläche). Finally, in his book on the logarithmic potential,
A. Harnack8 (1851–1888) gave the Green’s function for a circle and rectangle.
C
op

All of these authors used a technique that would become one of the fun-
damental techniques in constructing a Green’s function, namely eigenfunction
yr
ig

expansions. The investigator would first find an eigenfunction expansion that

ht

satisfied both the homogeneous differential equation and boundary conditions.

ed

The geometry of the problem would determine the coordinate system that was
used. Then the Fourier coefficients would be chosen so that the Green’s func-
M
at

tion exhibited the proper behavior (such as 1/r) near the source point. Later
er

on,9 John Dougall (1867–1960) derived three-dimensional Green’s functions

ia

in cylindrical and spherical coordinates.

l–

In 1879 Alfred George Greenhill10 (1847–1927) applied the method of

Ta

images to construct the Green’s function for a rectangular parallelepiped.

Because his results are expressed as an infinite summation of theta functions,
y lo

it was not very useful and has essentially been forgotten.

r&

Hector Munro Macdonald11 (1865–1935) took a slightly different ap-

Fr

proach in the 1890s. As before, he began with the eigenfunction expansion

an

∂2Φ
ci

6 Neumann, C., 1861: Ueber die Integration der partiellen Differentialgleichung: +

∂x2
s

2
∂ Φ
∂y 2
= 0. J. Reine Angew. Math., 59, 335–366.
7 Meutzner, P., 1875: Untersuchungen im Gebiete des logarithmischen Potentiales.

Math. Ann., 8, 319–338. For an alternative derivation, see Sections 15 and 17 in Neu-
mann, C., 1906: Über das logarithmische Potential. Ber. Verh. K. Sachs. Ges. Wiss.
Leipzig, Math.-Phys. Klasse, 58, 482–559.
8 See Chapter 2 in Harnack, A., 1887: Die Grundlagen der Theorie des logarithmischen

Potentiales und der eindeutigen Potentialfunktion in der Ebene. Leipzig, B. G. Teubner,

170 pp.
9 Dougall, J., 1900: The determination of Green’s function by means of cylindrical or

spherical harmonics. Proc. Edinburgh Math. Soc., Ser. 1 , 18, 33–83.

10
Greenhill, A. G., 1879: On Green’s function for a rectangular parallelepiped. Proc.
Cambridge Philos. Soc., 3, 289–293.
11 Macdonald, H. M., 1895: The electrical distribution on a conductor bounded by two

spherical surfaces cutting at any angle. Proc. London Math. Soc., Ser. 1 , 26, 156–172;

© 2015 by Taylor & Francis Group, LLC

Historical Development 5

C
op
yr
ig
ht
ed
M

Figure 1.2.1: Carl Gottfried Neumann (1832–1925) was a leading German mathematician
at

and teacher. Today he is best known for his work on the Dirichlet principle and inte-
er

gral equations, and his co-founding with Alfred Clebsch of Mathematische Annalen. Left
ia

c
photograph !Universitätarchiv c
Leipzig; right photograph !Photo Deutsches Museum.
l–
Ta

that satisfied the boundary conditions. But now, the Fourier coefficients were
chosen so that the expansion satisfied the general Poisson equation. Then
y lo

he considered the special case of a point source. We illustrate his method in

r&

Example 6.8.2. Because you must solve the general Poisson equation first, his
technique never became popular.
Fr

In the late 1890’s Arnold Sommerfeld12 (1868–1951) developed a tech-

an

nique using integration on the complex plane to extend the method of images
ci
s

to several other useful geometries in three dimensions. Ernst William Hob-

son (1856–1933) then used this method13 to find the Green’s function for a
circular disk. Later, Ludwig Waldmann (1913–1980), a young assistant to
Sommerfeld, applied this technique in electrostatic calculations of an electron
lens.14 Unfortunately this technique is very complicated and we will present

Macdonald, H. M., 1900: Demonstration of Green’s formula for electric density near the
vertex of a right cone. Trans. Cambridge Philos. Soc., 18, 292–297.
12 Sommerfeld, A., 1897: Über verzweigte Potentiale im Raum. Proc. London Math.

Soc., Ser. 1, 28, 395–429.

13 Hobson, E. W., 1900: On Green’s function for a circular disc, with applications to

electrostatic problems. Trans. Cambridge Philos. Soc., 18, 277–291.

14 Waldmann, L., 1937: Zwei Anwendungen der Sommerfeld’schen Methode der ver-

zweigten Potentiale. Phys. Z., 38, 654–663.

© 2015 by Taylor & Francis Group, LLC

6 Green’s Functions with Applications

an improved version in Example 6.2.5.

At the beginning of twentieth century the method of bilinear expansions
was developed:

(∞
ψn (x, y, z)ψn (ξ, η,ζ )
g(x, y, z|ξ, η,ζ ) = , (1.2.1)
n=1
λn

where λn and ψn (x, y, z) are the nth eigenvalue and eigenfunction, respec-
tively. Adolf Kneser15 (1862–1930) showed that the Green’s function was the
symmetric kernel of the integral equation
"""
ψn (ξ, η,ζ ) = λn g(x, y, z|ξ, η,ζ )ψn (x, y, z) dx dy dz. (1.2.2)
C
op

Assuming that the Green’s function can be expressed as an eigenfunction

yr

expansion, Equation 1.2.1 follows. As examples, Kneser found the bilinear

ig

expansion for rectangular and circular areas and for the surface of a sphere.
ht

In summary then, by 1950 there were essentially three methods16 for

ed

finding Green functions. The first method simply used a Green’s function
M

developed for Helmholtz’s equation ∇2 u + k02 u = 0 and took the limit as k0 →

at

0. The second method wrote the Green’s function as a sum of eigenfunctions

er

that satisfied the boundary conditions. The coefficients were then chosen so
ia

that the correct singular behavior occurred at the source point. Finally, the
l–

third method wrote the Green’s function as the sum of the free-space solution
Ta

plus a harmonic function.17 The harmonic solution was chosen so that the
y

Green’s function satisfied the boundary conditions.

lo

Later on, Kelvin’s classic inversion18 that maps the interior of a circle
r&

or sphere to the exterior and vice versa was developed to find the Green’s
Fr

function for Poisson’s equation. We will illustrate this technique following

an

Equation 6.3.37 and in Section 6.8.

ci

Finally Green’s functions have been used to solve mixed boundary-value

s

problems involving the two-dimensional Poisson’s equation. These problems

occur when the boundary condition changes along a given boundary from a

15 Kneser, A., 1911: Integralgleichungen und ihre Anwendungen in der mathematischen

Physik . Braunschweig, 293 pp.

16 See, for example, Bouwkamp, C. J., and N. G. de Bruijn, 1947: The electrostatic field

of a point charge inside a cylinder, in connection with wave guide theory. J. Appl. Phys.,
18, 562–577. This paper is of particular note because of its use of the modern definition of
the delta function. See their Equation 41.
17 Weber, E., 1939: The electrostatic potential produced by a point charge on the axis

of a cylinder. J. Appl. Phys., 10, 663–666.

18 Thomson, W., 1845: Extrait d’une lettre de M. William Thomson à M. Liouville. J.

Math. Pures Appl., 10, 364–367; Thomson, W., 1847: Extraits de deux lettres addressées
à M. Liouville. J. Math. Pures Appl., 12, 256–264.

© 2015 by Taylor & Francis Group, LLC

Historical Development 7

Dirichlet condition to a Neumann condition and vice versa. We will explore

this topic in Section 6.10.

1.3 HEAT EQUATION

The development of using Green’s function to solve the heat equation

consists of two parts. In the nineteenth century a synthetical method was
developed which replaces the actual distribution by sets of sources and sinks
distributed over the boundaries and throughout the region under investiga-
tion. The twentieth century has been dominated by the use of transform
methods.
Our tale begins with William Thomson (Lord Kelvin)19 (1824–1907) and
his solution of the one-dimensional heat equation
C

∂u ∂2u
op

= , 0 < x < ∞, 0 < t, (1.3.1)

∂t ∂x2
yr
ig

with the boundary conditions

ht

!
ed

V, 0 < t < T,
u(0, t) = lim u(x, t) → 0, 0 < t, (1.3.2)
0, T < t,
M

x→∞
at

and the initial conditions

er
ia

u(x, 0) = 0, 0 < x < ∞. (1.3.3)

l–
Ta

Using Fourier integrals, Kelvin obtained the solution

y

" T % &
lo

Vx dτ x2
r&

u(x, t) = √ exp − . (1.3.4)

2 π 0 (t − τ )3/2 4(t − τ )
Fr

He also showed that the solution could also be obtained “synthetically” for
an

small T if he introduced a source at x = 0 and a sink at x = α and took the

ci

limit as α → 0. Furthermore, reporting on some correspondence with George

s

G. Stokes (1819–1903), he gave the solution in a form that we now call the
superposition integral:
" t % &
x x2 f (τ )
u(x, t) = √ exp − dτ, (1.3.5)
2 π 0 4(t − τ ) t − τ

where u(0, t) = f (t).

In 1887 Ernst William Hobson20 (1856–1933) generalized Kelvin’s syn-
thetical method to two and three dimensions. In general, his effort must

19Thomson, W., 1854/55: On the theory of the electric telegraph. Proc. R. Soc.
London, 7, 382–399.
20 Hobson, E. W., 1887: Synthetical solutions in the conduction of heat. Proc. London

Math. Soc., Ser. 1 , 19, 279–299.

© 2015 by Taylor & Francis Group, LLC

8 Green’s Functions with Applications

C
op
yr
ig
ht
ed
M
at
er
ia
l–

Figure 1.3.1: William Thomson, Baron Kelvin of Largs, (1824–1907) was one of the lead-
Ta

ing mathematical physicists of the nineteenth century. During his work on thermodynamics
he realized that there was a lower limit to temperature. He became well known to the public
y lo

for his prediction on the speed of a signal in a transatlantic submarine cable that was being
r&

laid in the 1850s. (Portrait courtesy of the Royal Society.)

Fr

be considered a failure since it resulted in expressions that were difficult to

an

interpret. For example, in another paper,21 Hobson redid Kelvin’s problem

ci

of one-dimensional heat conduction when the boundary condition at x = 0

s

changed to ux (0, t) = h[u(0, t) − f (t)], where h denotes the external conduc-

tivity. He found that
" ∞" ∞ ! % & % &)
h (x + ζ − ξ)2 (x + ζ + ξ)2
u(x, t) = √ e−hζ exp − − exp −
2 π 0 0 4t 4t
× [u(ξ, 0) − ux (ξ, 0)] dξd ζ. (1.3.6)
Hobson’s solution fails if the initial temperature distribution is discontinuous.
In 1892, George H. Bryan22 (1864–1928) improved the synthetical meth-
od. He cleverly wrote the solution as the sum of two parts: a source term

21
Hobson, E. W., 1888: On a radiation problem. Math. Proc. Cambridge Philos. Soc.,
6, 184–187.
22 Bryan, G. H., 1892: Note on a problem in the linear conduction of heat. Math. Proc.

Cambridge Philos. Soc., 7, 246–248.

© 2015 by Taylor & Francis Group, LLC

Historical Development 9

(located at x = ξ) plus a homogeneous solution so that the total solution

satisfies the boundary condition. Using this technique to redo Hobson’s 1888
calculation, he found that
" ∞! % & % &)
1 (x − ξ)2 (x + ξ)2
u(x, t) = √ exp − + exp − f (ξ) dξ
2 πt 0 4t 4t
" ∞" ∞ % &
h (x + ζ + ξ)2
−√ e−hζ exp − f (ξ) dξd ζ. (1.3.7)
πt 0 0 4t

Later on, Sommerfeld23 generalized Kelvin’s results by considering sub-

stances that have different thermal properties. He considered two cases: (a)
two semi-infinite domains with the interface at x = 0 and (b) two semi-infinite
slabs separated a finite layer lying between a < x < b. In 1939 P. V. Solovi-
C

eff 24 used mirror images to construct Green’s functions for a n+1 dimensional
op

heat equation when one or more boundaries are moving.

yr

By the turn of the twentieth century, John Dougall25 (1867–1960) intro-

ig

duced the concept of contour integration to find the mathematical description

ht

of how a sphere cools in a well-stirred liquid. Although Dougall’s analysis

ed

made no use of Green’s functions, it provided the necessary insight that al-
M

lowed H. S. Carslaw to synthesize all of the ideas on how to apply Green’s

at

function to heat conduction problems. Carslaw’s 1902 paper26 began by deriv-

er

ing how any conduction problem without sources can be expressed in terms
ia
l–

of its Green’s function, the initial condition and the solution’s value along
the boundary surrounding the domain of interest. The question then turned
Ta

on the question of finding the Green’s function for various geometries and
y

boundary conditions. In the case of unbounded domains, he used the Green’s

lo
r&

functions given by the synthetical method. The remaining Green’s functions

were obtained using Dougall’s method of contour integration. For example, to
Fr

find the Green’s function for linear heat flow over the interval (0, a), when the
an

Green’s function vanishes at both ends, Carslaw first introduced the Green’s
ci

function
s

! % & % &)
1 (x − ξ)2 (x + ξ)2
g1 (x, t|ξ, 0) = √ exp − − exp − . (1.3.8)
2 πκt 4κt 4κt

23 Sommerfeld, A., 1894: Zur analytischen Theorie der Wärmeleitung. Math. Ann., 45,

263–277.
24 Solovieff, P. V., 1939: Die Greensche Funktion der Wärmeleitungsgleichung. Dokl.

Acad. Sci. USSR, 23, 132–134; Solovieff, P. V., 1939: Fonctions de Green des équations
paraboliques. Dokl. Acad. Sci. USSR, 24, 107–109.
25 Dougall, J., 1901: Note on the application of complex integration to the equation of

conduction of heat, with special reference to Dr. Peddie’s problem. Proc. Edinburgh Math.
Soc., Ser. 1 , 19, 50–56.
26 Carslaw, H. S., 1902: The use of Green’s functions in the mathematical theory of the

conduction of heat. Proc. Edinburgh Math. Soc., Ser. 1 , 21, 40–64.

© 2015 by Taylor & Francis Group, LLC

10 Green’s Functions with Applications

x
C

Figure 1.3.2: The contour P and closed contour used in Equation 1.3.9 and Equation
op

1.3.10, respectively.
yr
ig

The first term is the free-space Green’s function given by Kelvin while the
ht

second term assures that g1 (0, t|ξ, 0) = 0. Carslaw then showed that this
ed

Green’s function could be expressed by the contour integral

M

"
at

1 2
g1 (x, t|ξ, 0) = e−κz t sin(x< z)e−izx> dz, (1.3.9)
er

πi P
ia
l–

where P is the contour shown in Figure 1.3.2. On the right side, the contour
Ta

must lie between 0 < arg(z) < π /4 as |z| → ∞ while on the left side the
contour lies between 3π/4 < arg(z) < π as |z| → ∞.
y lo

Next, Carslaw introduced the new Green’s function

r&

"
1 t sin(x< z) sin(x> z) iaz
Fr

2
g2 (x, t|ξ, 0) = − e−κz e dz. (1.3.10)
an

πi P sin(az)
ci
s

Why did Carslaw create this new Green’s function? Because a linear combi-
nation of g1 (x, t|ξ, 0) and g2 (x, t|ξ, 0) yields the Green’s function
"
1 2
t sin(x< z) sin[(a − x> )z]
g(x, t|ξ, 0) = e−κz dz (1.3.11)
πi P sin(az)

which satisfies the boundary conditions g(0, t|ξ, 0) = g(a, t|ξ, 0) = 0. Further-
more, Carslaw was also able to show that this Green’s function satisfied the
initial condition and had the correct behavior at the singularity x = ξ. Using
Cauchy’s residue theorem and the closed contour showed in Figure 1.3.2, this
Green’s function could be rewritten in the convenient form of
, - % &
2 ( * nπx +
∞
nπξ κn2 π 2 (t − τ )
g(x, t|ξ,τ ) = sin sin exp − . (1.3.12)
a n=1 a a a2

© 2015 by Taylor & Francis Group, LLC

Historical Development 11

The difficulty of this method is quite apparent: It is not easy to choose

a complex representation for the Green’s function that satisfies the boundary
conditions, initial condition and the singular nature at the point of excitation.
This difficulty became academic with the advent of Laplace transforms.
In the mid-1920s, Gustav Doetsch (1892–1977) wrote a series of papers on
heat conduction.27 In particular, he considered the heat conduction problem

∂u ∂2u
= , 0 < x < c, 0 < t, (1.3.13)
∂t ∂x2
with the boundary conditions

u(0, t) = A(t), u(c, t) = B(t), 0 < t, (1.3.14)

C

and the initial condition

op

u(x, 0) = Φ(x), 0 < x < c. (1.3.15)

yr
ig
ht

He showed that the solution u(x, t) could be expressed by the one-dimensional

ed

version of Equation 5.0.11 (see Problem 1 in Chapter 5) and the Green’s

function
M
at

∞
2 ( −n2 π2 t/c2
er

+
g(x, t|ξ, 0 ) = e sin(nπξ/c) sin(nπx/c). (1.3.16)
ia

c n=1
l–

Furthermore, he showed that the Green’s function is symmetric. Finally, the

Ta

limit of c → ∞ yielded the Green’s functions on a semi-infinite domain found

y lo

by Kelvin and Carslaw. The revolutionary aspect of Doetsch’s approach was

r&

his use of Laplace transforms.

In 1932, Sydney Goldstein28 (1903–1989) showed how Laplace transforms
Fr

could be used to solve many heat conduction problems whose derivation up

an

to that time had been very cumbersome. Indeed he pointed out that many
ci

of the Green’s functions found by Carslaw’s contour integral method was in

s

reality the “operational method.”

27 Bernstein, F., and G. Doetsch, 1925: Probleme aus der Theorie der Wärmeleitung.

I. Mitteilung. Eine neue Methode zur Integration partieller Differentialgleichungen. Der

lineare Wärmeleiter mit verschwindender Anfangstemperatur. Math. Z., 22, 285–292;
Doetsch, G., 1925: Probleme aus der Theorie der Wärmeleitung. II. Mitteilung. Der lin-
eare Wärmeleiter mit verschwindender Anfangstemperatur. Die allgemeinste Lösung und
die Frage der Eindeutigkeit. Math. Z., 22, 293–306; Doetsch, G., 1925: Probleme aus der
Theorie der Wärmeleitung. III. Mitteilung. Der lineare Wärmeleiter mit beliebiger An-
fangstemperatur. Die zeitliche Fortsetzung des Wärmezustandes. Math. Z., 25, 608–626;
Bernstein, F., and G. Doetsch, 1927: Probleme aus der Theorie der Wärmeleitung. IV.
Mitteilung. Die räumliche Fortsetzung des Temperaturablaufs (Bolometerproblem). Math.
Z., 26, 89–98.
28 Goldstein, S., 1932: Some two-dimensional diffusion problems with circular symmetry.

Proc. London Math. Soc., Ser. 2 , 34, 51–88.

© 2015 by Taylor & Francis Group, LLC

12 Green’s Functions with Applications

Consider, for example, the one-dimensional heat conduction problem

∂u ∂2u
= a2 2 , 0 < x < ∞, 0 < t. (1.3.17)
∂t ∂x
Taking the Laplace transform of Equation 1.3.17, we have that

d2 U
= q 2 U, 0 < x < ∞, (1.3.18)
dx2
where s = a2 q 2 . The solution to Equation 1.3.18 is

U (x, s) = Ae−qx , (1.3.19)

where we have discarded the exponentially growing solution as x → ∞. Now,

C

" x
op

# $
2 U (x, s) dx = 2A 1 − e−qx /q. (1.3.20)
yr

0
ig
ht

To find the Green’s function, we must choose A so that it represents an in-

ed

stantaneous plane source of heat from x = −∞ to x = ∞ in the limit of t → 0.

M

In that case, the left side of Equation 1.3.20 equals one as q → ∞, or 2A = q.

at

Therefore,
er

G(x, s|0, 0) = 12 qe−qx . (1.3.21)

ia
l–

Taking the inverse of G(x, s|0, 0), we have that

, -
Ta

1 x2
g(x, t|0, 0) = √ exp − 2 , (1.3.22)
y

2 πa2 t 4a t
lo
r&

the same result that Kelvin found. Goldstein used similar methods to find
Fr

the Green’s function for an axisymmetric problem in the plane and outside of
an

a cylinder.
ci

During the 1930s and 1940s several authors found the Green’s functions
s

for cylindrical and spherical geometries by using Byran’s method of writing

the Green’s function as a sum of a free-space Green’s function plus a homo-
geneous solution which satisfies the boundary conditions. They obtained the
homogeneous solution using Laplace transforms. For example,29 Lowan re-
did Bryan’s original problem of finding the Green’s function in a semi-infinite
planar solid that is radiating at the x = 0 face. Later, Lowan applied this
technique to heat conduction in cylindrical30 and spherical31 coordinates.

29 Lowan, A. N., 1937: On the operational determination of Green’s functions in the

theory of heat conduction. Philos. Mag., Ser. 7 , 24, 62–70.

30 Lowan, A. N., 1938: On the operational determination of two dimensional Green’s
function in the theory of heat conduction. Bull. Amer. Math. Soc., 44, 125–133.
31 Lowan, A. N., 1939: On Green’s functions in the theory of heat conduction in spherical

coordinates. Bull. Amer. Math. Soc., 45, 310–315 and 951–952.

© 2015 by Taylor & Francis Group, LLC

Historical Development 13

C
op
yr
ig
ht

Figure 1.3.3: John Conrad Jaeger’s (1907–1979) (right portrait) association with Ho-
ed

ratio Scott Carslaw (1870–1954) began during Jaeger’s undergraduate education at the
University of Sydney. After Jaeger’s undergraduate education and subsequent studies at
M

the University of Cambridge, the contact with Carslaw was renewed by irregular trips to
at

Carslaw’s retirement farm when Jaeger returned to Tasmania in 1936. These meetings lead
er

to a collaboration on the application of Laplace transforms to find the Green’s function

ia

for the heat equation. (Carslaw’s portrait courtesy of the University of Sydney Archives,
l–

Image G3/224/0695; Jaeger’s portrait courtesy of the Royal Society.)

Ta

During this same period Carslaw and Jaeger also found Green’s functions
y lo

using Laplace transforms. In their earliest paper32 they found the Green’s
r&

function for the region outside of a cylinder using both the Laplace transforms
and the contour method. In a subsequent paper Carslaw33 found the Green’s
Fr
an

function for heat conduction in two semi-infinite solids of different materials

with a common boundary at x = 0 as well as the case when the two semi-
ci
s

infinite solids are separated by a third solid of thickness 2a. In the case of
three dimension problems, they34 wrote the Green’s function as a sum of the
free-space Green’s function plus a homogeneous solution of the heat equation.
They then used Laplace transforms to find the homogeneous solution. Finally
Carslaw and Jaeger35 applied Bryan’s technique to cylindrical problems where

32 Carslaw, H. S., and J. C. Jaeger, 1939: On Green’s functions in the theory of heat

conduction. Bull. Amer. Math. Soc., 45, 407–413.

33 Carslaw, H. S., 1940: A simple application of the Laplace transformation. Philos.

Mag., Ser. 7 , 30, 414–417.

34 Carslaw, H. S., and J. C. Jaeger, 1940: The determination of Green’s function for the

equation of conduction of heat in cylindrical coordinates by the Laplace transformation. J.

London Math. Soc., 15, 273–281.
35 Carslaw, H. S., and J. C. Jaeger, 1941: The determination of Green’s function for

© 2015 by Taylor & Francis Group, LLC

14 Green’s Functions with Applications

they used line sources for the free-space Green’s functions.

1.4 HELMHOLTZ’S EQUATION

A partial differential equation which is quite similar to Laplace’s equa-

tion is Helmholtz’s equation. It arises in the study of forced (steady-state)
vibrations governed by the wave equation; the most famous application is the
diffraction of acoustic and visible light waves.
The history of Green’s function involving Helmholtz’s equation begins
with the theoretical work of Hermann von Helmholtz (1821–1894) during
his study of acoustics.36 He showed that the free-space Green’s function
is g(x, y, z|ξ, η,ζ ) = cos(k0 r)/r, where r2 = (x − ξ)2 + (y − η)2 + (z − ζ)2 .
Helmholtz used this Green’s function to express the solution of ∇2 u + k02 u = 0
C
op

in a region R with the boundary S which gives the solution and its derivative
yr

on the boundary.
ig

F. Pockels (1865–1913) epic book37 on Helmholtz’s equation summarized

ht

our knowledge of Green’s function at the end of the nineteenth century. In

ed

Part IV (Bestimmung der Functionen u aus gegebenen Randwerthen und ver-

M

wandten Bedingungen) he reviewed the Dirichlet principle and showed how

at

Green’s functions could be used to solve this problem. Next Pockels discussed
er

the expansion of Green’s function in term of eigenfunctions. For example, he

ia

gave the Green’s function within a circle in terms of Fourier series.

l–

For us, Section 4 (Lösung der Randwerthaufgaben für die Functionen u

Ta

mit Hülfe verallgemeinerter Green’scher Functionen) of Part IV is of partic-

y

ular interest. In this section Pockels gave the free-space Green’s function in
lo

three-dimensional space and on xy-plane, discussed reciprocity and presented

r&

the boundary-value solution in terms of the Green’s function and the type
Fr

of boundary condition. In summary, the nineteenth century saw the full de-
an

velopment of the concept of the Green’s functions but presented few actual
ci

functions. This began to change in the twentieth century.

s

In the early twentieth century several major studies appeared on the

Green’s functions for Helmholtz’s equation. The first paper38 was by A. Som-
merfeld (1868–1951). It consisted of two parts: Green’s function for a bounded
and unbounded region. For a finite domain he showed that the Green’s func-

line sources for the equation of conduction of heat in cylindrical coordinates by the Laplace
transformation. Philos. Mag., Ser. 7 , 31, 204–208.
36 Helmholtz, H., 1860: Theorie der Luftschwingungen in Röhren mit offenen Enden. J.

Reine Angew. Math., 57, 1–72.

37 Pockels, F., 1891: Über die partielle Differentialgleichung ∆u + k 2 u = 0 und deren

Auftreten in der mathematischer Physik . Leipzig, Teubner, 339 pp.

38 Sommerfeld, A., 1912: Die Greensche Funktion der Schwingungsgleichung. Jahresber.

Deutsch. Math.-Verein., 21, 309–353.

© 2015 by Taylor & Francis Group, LLC

Historical Development 15

tion can be expressed by the eigenfunction expansion:

( um (O)um (P )
g= , (1.4.1)
m
k02 − km
2

where um (O) and um (P ) are the mth eigenfunction to the eigenvalue problem
∇2 um + km 2
um = 0 with um = 0 along the boundary and point O denotes
a general point within the domain while the point P is the location of the
singularity. Sommerfeld proved that this expansion satisfies the partial dif-
ferential equation, g = 0 at the boundary, is indefinite when point P and
point O are collocated, and satisfies reciprocity. As examples, he gave free-
space Green’s functions in two and three dimensions as well as for a circular
membrane with a fixed boundary and a three-dimensional parallelepiped with
Neumann boundary conditions. In two subsequent papers H. S. Carslaw39
C

extended Sommerfeld’s results for a wide variety of three dimensional spaces

op

involving cylindrical and spherical coordinates.

yr

Turning to unlimited domains, Sommerfeld gave the free-space Green’s

ig

function in two and three dimensions. More importantly, he derived his fa-
ht

mous “radiation condition” that required outwardly propagating waves from

ed

physical considerations.40
M

By 1950 Green’s functions for Helmholtz’s equation were used to find the
at

wave motions due to flow over a mountain41 and in acoustics.42

er
ia

1.5 WAVE EQUATION

l–
Ta

Soon after the publication of Green’s essay, Green’s functions were used
to solve the wave equation. In 1860 Bernhard Riemann43 (1826–1866) applied
y lo

the method of Green’s functions to integrate the hyperbolic equation that de-
r&

scribes the propagation of sound waves.44

Fr
an

39 Carslaw, H. S., 1912: Integral equations and the determination of Green’s functions
ci

in the theory of potential. Proc. Edinburgh Math. Soc., Ser. 1 , 31, 71–89; Carslaw, H.
s

S., 1914: The Green’s function for the equation ∇2 u + k 2 u = 0. Proc. London Math. Soc.,
Ser. 2 , 15, 236–257.
40 See Schot, S. M., 1992: Eighty years of Sommerfeld’s radiation condition. Hist. Math.,

19, 385–401.
41 Lyra, G., 1943: Theorie der stationären Leewellenströmung in freier Atmosphäre.

Zeit. Angew. Math. Mech., 23, 1–28.

42 Foldy, L. L., and H. Primakoff, 1945: A general theory of passive linear electroacoustic

transducers and the electroacoustic reciprocity theorem. I. J. Acoust. Soc. Am., 17, 109–
120.
43 Riemann, B., 1860: Ueber die Fortpflanzung ebener Luftwellen von endlicher Schwing-
ungsweite. Abh. d. Kön. Ges. der Wiss. zu Göttingen, 8, 43–65. An English translation
appears in Johnson, J. N., and R. Chéret, 1998: Classic Papers in Shock Compression
Science. Springer-Verlag, 524 pp.
44 Mackie, A. G., 1964/65: Green’s function and Riemann’s method. Proc. Edinburgh

Math. Soc., Ser. 2 , 14, 293–302.

© 2015 by Taylor & Francis Group, LLC

16 Green’s Functions with Applications

C
op
yr
ig
ht
ed
M
at
er
ia

Figure 1.5.1: Originally drawn to mathematics, Arnold Johannes Wilhelm Sommerfeld

l–

(1868–1951) migrated into physics due to Klein’s interest in applying the theory of complex
variables and other pure mathematics to a range of physical topics from astronomy to dy-
Ta

namics. Later on, Sommerfeld contributed to quantum mechanics and statistical mechanics.
y

(AIP Emilio Segrè Visual Archives, Margrethe Bohr Collection.)

lo
r&

For a linear hyperbolic equation of second order in two independent vari-

Fr

ables
∂2u ∂2u ∂u ∂u
an

− 2 + 2a − 2b + cu = 0, (1.5.1)
∂x2 ∂y ∂x ∂y
ci
s

where x and y are chosen so that the two families of characteristic are x ± y =
constant and a, b and c are functions only of x and y, the solution45 at the
point P with coordinates (ξ,η ) is
1
u(ξ,η ) = 2 (uG)|A + 12 (uG)|B (1.5.2)
"
+ 12 (Guy − uGy + 2b uG) dx + (Gux − uvx + 2a uG) dy.
AB

Here G denotes the Riemann-Green function and is given by the adjoint equa-
tion
∂2G ∂2G ∂(aG) ∂(bG)
− −2 +2 + cG = 0 (1.5.3)
∂x2 ∂y 2 ∂x ∂y

45 For the derivation, see Section 73 in Webster, A. G., 1966: Partial Differential Equa-

tions of Mathematical Physics. Dover, 446 pp.

© 2015 by Taylor & Francis Group, LLC

Historical Development 17

such that
∂G ∂G
+ = (a + b)G on y − x = η − ξ, (1.5.4)
∂x ∂y
∂G ∂G
− = (a − b)G on y + x = η + ξ, (1.5.5)
∂x ∂y
and G(ξ,η ) = 1. The values of u and its first derivative are specified along the
arc AB which is chosen so that no characteristic cuts it more than at one point;
the arcs P A and P B are characteristics. Although there are several methods
for finding Riemmann-Green functions,46 actually finding one is very difficult.
The greatest success with this technique involved the equation of telegraphy.47
The next important development of Green’s functions for the wave equa-
tion lies with Gustav Robert Kirchhoff 48 (1824–1887), who used it during his
study of the three-dimensional wave equation. Starting with Green’s second
C

formula, he was able to show that the three-dimensional Green’s function is

op
yr

δ(t − τ − R/c)
ig

g(x, y, z, t|ξ, η, ζ,τ ) = , (1.5.6)

4πR
ht

.
ed

where R = (x − ξ)2 + (y − η)2 + (z − ζ)2 (modern terminology). Although

he did not call his solution a Green’s function,49 he clearly grasped the concept
M
at

that this solution involved a function that we now call the Dirac delta function
er

(see pg. 667 of his Annalen d. Physik ’s paper). He used this solution to de-
ia

rive his famous Kirchhoff ’s theorem, which is the mathematical expression for
l–

Huygen’s principle: energy always propagates out to infinity.

Ta

The early twentieth century saw the development of Laplace transforms

y

to solve the wave equation in addition to the previous known method of Fourier
lo
r&

46 See Copson, E. T., 1958: On the Riemann-Green function. Arch. Rat. Mech. Anal.,
Fr

1, 324–348.
an

47 Picard, É., 1894: Sur l’équation aux dérivées partielles qui se recontre dans la théorie
ci

de la propagation de l’électricité. Acad. Sci., Compt. Rend., 118, 16–17; Bois-Reymond,

s

P. du, 1889: Über lineare partielle Differentialgleichungen zweiter Ordnung. J. Reine An-
gew. Math., 104, 241–301; Voigt, W., 1899: Ueber die Aenderung der Schwingungsform des
Lichtes beim Fortschreiten in einem dispergirenden oder absorbirenden Mittel. Ann. Phys.,
304, 598–603; Gray, M. C., 1923: The equation of telegraphy. Proc. Edinburgh Math. Soc.,
Ser. 2 , 42, 14–28; Rademacker, H., and R. Iglisch, 1961: Randwertprobleme der partiellen
Differentialgleichungen zweiter Ordnung, 779–828 in Frank, Ph., and R. von Mises, 1961:
Die Differential- und Integralgleichungen der Mechanik und Physik. I. Mathematischer
Teil . Dover, 916 pp.; Section 74 in Webster, A. G., 1966: Partial Differential Equation
of Mathematical Physics. Dover, 446 pp.; Wahlberg, C., 1977: Riemann’s function for a
Klein-Gordon equation with a non-constant coefficient. J. Phys., Ser. A, 10, 867–878;
Asfar, O. R., 1990: Riemann-Green function solution of transient electromagnetic plane
waves in lossy media. IEEE Trans. Electromagn. Compat., EMC-32, 228–231.
48 Kirchhoff, G., 1882: Zur Theorie der Lichtstrahlen. Sitzber. K. Preuss. Akad. Wiss.
Berlin, 641–669; reprinted a year later in Ann. Phys. Chem., Neue Folge, 18, 663–695.
49 This appears to have been done by Gutzmer, A., 1895: Über den analytischen Aus-

druck des Huygens’schen Princips. J. Reine Angew. Math., 114, 333–337.

© 2015 by Taylor & Francis Group, LLC

18 Green’s Functions with Applications

C
op
yr
ig
ht
ed
M
at

Figure 1.5.2: Gustav Robert Kirchhoff’s (1824–1887) most celebrated contributions to

er

physics are the joint founding with Robert Bunsen of the science of spectroscopy, and the
ia

discovery of the fundamental law of electromagnetic radiation. Kirchhoff’s work on light

l–

coincides with his final years as a professor of theoretical physics at Berlin. (Portrait taken
from frontispiece of Kirchhoff, G., 1882: Gesammelte Abhandlungen. J. A. Barth, 641 pp.)
Ta
y lo

transforms.50 In 1914 T. J. I’A. Bromwich (1875–1929) showed how Laplace

r&

transforms51 can be used to solve the wave equation by eliminating the tempo-
Fr

ral dependence, leaving a boundary-value problem. Interestly he then solved

an

this boundary-value problem using Green’s functions. Then, unknowingly

ci

he found as an example the Green’s function for the one-dimensional wave

s

equation with fixed ends (see his Example 5 on page 438). A. N. Lowan
(1898–1962) applied Bromwich’s idea to finding the wave motions within a
wedge52 of infinite radius and an infinite solid53 which is exterior to a cylin-
der or sphere.

50 Poincaré, H., 1893: Sur la propagation de l’électricité. Acad. Sci., Compt. Rend.,

117, 1027–1032; Webster, A. G., 1966: Partial Differential Equation of Mathematical

Physics. Dover, Section 46.
51 Bromwich, T. J. I’A., 1914: Normal coordinates in dynamical systems. Proc. London

Math. Soc., Ser. 2 , 15, 401–448.

52 Lowan, A. N., 1941: On the problem of wave-motion for the wedge of an angle. Philos.
Mag., Ser.7 , 31, 373–381.
53 Lowan, A. N., 1939: On wave motion in an infinite solid bounded internally by a

cylinder or a sphere. Bull. Amer. Math. Soc., 45, 316–325.

© 2015 by Taylor & Francis Group, LLC

Historical Development 19

One difficulty of finding the Green’s function for the wave equation lies
in its definition. Around 1950 A. G. Walters wrote a series of papers54 (that
have been essentially forgotten) on finding the Green’s function of transient
partial differential equations. In the case of the wave equation, his Green’s
vibrational function g(P, P1 , t − τ ) satisfied the wave equation

∂2g
= c2 D(g), (1.5.7)
∂t2

where D(·) is a differential operator in one or more dimensions, and satisfies

the integral conditions
"
lim g(P, P1 , t − τ ) dV = 0, (1.5.8)
τ →t
C

V
op

and "
yr

lim gt (P, P1 , t − τ ) dV = 1, (1.5.9)

ig

τ →t
ht

V
ed

where V is any region enclosed by the boundaries and contains the point P1 ,
which is defined below. He then showed how the solution to the wave equa-
M
at

tion can be expressed in terms of volume integrals involving the initial condi-
er

tions, boundary conditions, and any source terms. To compute g(P, P1 , t − τ ),

ia

Walters proved that the Green’s function is the inverse Laplace transform of
l–

Φ(P, P1 , s2 /c2 ), where Φ satisfies the boundary-value problem

Ta

s2
y

D(Φ) = Φ. (1.5.10)
lo

c2
r&

Note that Φ satisfies Equation 1.5.10 except for a singular point located at
Fr

P1 .
an

Our modern definition of the Green’s function for the wave equation as
ci

the response of this equation to impulse forcing appears to originate with H.

s

J. Bhabha55 (1909–1966) in his study of the meson field of neutrons. As we

will show in Example 4.1.1, he used transform methods to find the Green’s
function for the Klein-Gordon equation.
Van der Pol and Bremmer were the first to introduce the general com-
munity to the concept that the Green’s function56 of the wave equation is a

54 Walters, A. G., 1949: The solution of some transient differential equations by means

of Green’s functions. Proc. Cambridge Philos. Soc., 45, 69–80; Walters, A. G., 1951: On
the propagation of disturbances from moving sources. Proc. Cambridge Philos. Soc., 47,
109–126.
55 Bhabha, H. J., 1939: Classical theory of mesons. Proc. R. Soc. London, Ser. A, 172,

384–409.
56 Van der Pol, B., and H. Bremmer, 1964: Operational Calculus Based on the Two-Sided

Laplace Transform. Cambridge, 415 pp.

© 2015 by Taylor & Francis Group, LLC

20 Green’s Functions with Applications

particular solution to the wave equation when it is forced by a point source

both in space and time. They then derived the Green’s function for the n-
dimensional wave equation as well as the three-dimensional wave equation
with dispersion. Shortly after Van der Pol’s book, P. M. Morse and H. Fesh-
bach further developed the theory of Green’s functions as it applies to the wave
equation. From Van der Pol’s definition, they obtained the reciprocity rela-
tion and the free-space Green’s function in one, two, and three dimensions.57
Significantly Morse and Feshbach did not use transform methods but derived
their results from heuristic arguments and repeated integration.
The use of transform methods to find Green’s function for the wave equa-
tion rapidly occurred after the publication of Van der Pol’s book. For example,
in Friedlander’s examination58 of pulses by a circular cylinder, he found the
approximate Green’s function for the two-dimensional wave equation exte-
rior to a cylinder of radius 1 where gr (1, θ , t|ρ,θ $ , τ ) = 0. He used many of
C
op

the techniques in this book. In particular, Laplace transforms were used to

yr

eliminate time and Fourier series were employed to give the θ dependence.
ig

An important class of wave propagation involves the diffraction of a direct

ht

wave by an infinitesimally thin barrier along the x-axis. In 1935 L. Cagniard

ed

used Laplace transforms to find the diffraction of a step function direction

M

by a half-plane.59 The Green’s function follows by simply taking the time

at

derivative of Cagnaird’s solution.60

er

The Green’s function for the corresponding two-dimensional problem is

ia

more difficult. R. D. Turner found the earliest representation using Laplace

l–

transforms.61 G. Schouten62 has given a closed form solution for the Green’s
Ta

function. See Section 4.8.

y lo
r&

1.6 ORDINARY DIFFERENTIAL EQUATIONS

Fr

The application of Green’s functions to ordinary differential equations

an

began in 1894. Noting the use of Green’s functions in solving the two and three
ci
s

57 Morse, P. M., and H. Feshbach, 1953: Methods of Theoretical Physics. Part I: Chap-
ters 1 to 8. McGraw-Hill, 997 pp.
58 Friedlander, F. G., 1954: Diffraction of pulses by a circular cylinder. Commun. Pure

Appl. Math., 7, 705–732.

59 Cagniard, L., 1935: Diffraction d’une onde progressive par un écran en forme de demi-

plan. J. Phys. Radium, Ser. 7 , 6, 310–318; Cagniard, L., 1935: Diffraction d’une onde
harmonique par un écran en forme de demi-plan. J. Phys. Radium, Ser. 7 , 6, 369–372.
60 Schouten, G., 1999: Two-dimensional effects in the edge sound of vortices and dipoles.

J. Acoust. Soc. Am., 106, 3167–3177.

61Turner, R. D., 1956: The diffraction of a cylindrical pulse by a half-plane. Q. Appl.
Math., 14, 63–73.
62 Schouten, op. cit., p. 3170.

© 2015 by Taylor & Francis Group, LLC

Historical Development 21

dimensional Poisson equation, H. Burkhardt63 (1861–1914) asked whether

they could be used to solve

d2 y
= f (x), a < x < b. (1.6.1)
dx2
He showed that the solution to this problem can be written
" x " b
(b − x)(ξ − a) (b − ξ)(x − a)
y(x) = − f (ξ) dξ − f (ξ) dξ, (1.6.2)
a b−a x b−a

which he wrote " b

y(x) = − g(x|ξ)f (ξ) dξ, (1.6.3)
a
C

where
op

(b − x> )(x< − a)
g(x|ξ) = . (1.6.4)
yr

b−a
ig
ht

Burkhardt’s Green’s function g(x|ξ) enjoyed the classic properties:

ed

• The Green’s function satisfies the differential equation g $$ = 0.

M

• The Green’s function is finite and continuous on the interval (a, b) except
at

for x = ξ.
er

• The first derivative of g(x|ξ) is continuous except at x = ξ where it

ia

possesses two distinct values that differ by 1.

l–

• g(a|ξ) = g(b|ξ) = 0.
Ta

• The Green’s function has the symmetry property: g(x|ξ) = g(ξ|x).

y lo

In 1903 M. Bôcher64 (1867–1918) gave the properties of the Green’s function

r&

for homogeneous, linear, n-th order ordinary differential equation

Fr

dn y dn−1 y dy
an

n
+ a1 n−1 + · · · + an−1 + an y = 0, a ≤ x ≤ b, (1.6.5)
dx dx dx
ci
s

where a1 , a2 , . . . , an are real functions of the real variable x and the coefficients
of the n linearly independent boundary conditions are constants. William M.
Whyburn (1901–1972) would list65 the properties of the Green’s function for
the general self-adjoint linear, second-order homogeneous differential equation
% &
d dy
p(x,λ ) − q(x,λ )y = 0, a < x < b, (1.6.6)
dx dx

63 Burkhardt, M. H., 1894: Sur les fonctions de Green relatives a un domaine d’une

dimension. Bull. Soc. Math., 22, 71–75.

64 Bôcher, M., 1901: Green’s function in space of one dimension. Bull. Amer. Math.
Soc., Ser. 2, 7, 297–299.
65 Whyburn, W. M., 1924: An extension of the definition of the Green’s function in one

dimension. Ann. Math., Ser. 2 , 26, 125–130.

© 2015 by Taylor & Francis Group, LLC

22 Green’s Functions with Applications

C
op
yr
ig
ht
ed
M
at
er
ia
l–

Figure 1.6.1: Educated at Harvard and Göttingen, Maxime Bôcher (1867–1918) spent his
entire career teaching at Harvard. He published approximately 100 papers on differential
Ta

equations, series and algebra. Photograph courtesy of the American Mathematical Society
y

(www.ams.org).
lo
r&

with self-adjoint boundary conditions. The jump in the first derivative at the
point of excitation x = ξ now becomes 1/p(x,λ ).
Fr
an

Burkhardt also found the Green’s function for Equation 1.6.1 when the
boundary condition reads
ci
s

y(a) + χy $ (a) = 0, y(b) + χy $ (b) = 0. (1.6.7)

In that case,
(b + χ − x> )(x< − a + χ)
g(x|ξ) = . (1.6.8)
b − a + 2χ
He noted special difficulties when χ = ∞ and χ = (a − b)/2. In the first case,
the boundary condition becomes g $ (a|ξ) = g $ (b|ξ) = 0. For the solution to
/b
exist, a f (x) dx = 0. In that case,
" x
y(x) = f (ξ)(x − ξ) dξ + λ, (1.6.9)
a
where λ is arbitrary. In the second case, the solution exists if
" b, -
a+b
x− f (x) dx = 0. (1.6.10)
a 2

© 2015 by Taylor & Francis Group, LLC

Historical Development 23

If this condition holds, then

" x , -
a+b
y(x) = f (ξ)(x − ξ) dξ + λ x − . (1.6.11)
a 2

All of these results were merely stated, not proved. In 1911 M. Bôcher66
(1867–1918) provided the mathematical justification for these results.
Ince,67 in his great treatise on ordinary differential equations, introduced
these results to the general community. In his Section 11.1 he proved that
the Green’s function for Equation 1.6.5 exists and is unique. Furthermore he
showed that the Green function is symmetrical g(x|ξ) = g(ξ|x) if the ordinary
differential equation is self-adjoint. In Section 11.11 Ince further proved that
"
C

b
op

y(x) = g(x|ξ)f (ξ) dξ (1.6.12)

a
yr
ig

is the solution to the non-homogeneous equation

ht
ed

dn y dn−1 y dy
+ a + · · · + an−1 + an y = f (x), a ≤ x ≤ b. (1.6.13)
M

n 1 n−1
dx dx dx
at
er

As we will see in Section 3.4 an important concept in the theoretical

ia

development of Green’s functions as they apply to ordinary differential equa-

l–

tions involves the adjoint . In 1908 G. D. Birkhoff 68 formulated the Green’s

Ta

function in terms of the adjoint differential equations and the eigenvalues and
y

eigenfunctions of the differential equations. He expressed the Green’s function

lo

as a ratio of determinants in terms of linearly independent solutions to the dif-

r&

ferential equation and the boundary conditions. Then, in 1909 E. Bounitzky69

Fr

extended Birkhoff results (although he did not reference his paper) to systems
an

of first-order differential equations.

ci

Of all of the possible ordinary differential equations that possess a Green’s

s

function, several are of fundamental importance because these arise in impor-

tant physical problems. One of these is

d2 u
+ [0(x) + λ2 ]u = 0, 0 < x < 1, (1.6.14)
dx2

66 Bôcher, M., 1911/12: Boundary problems and Green’s functions for linear differential

and difference equations. Ann. Math., Ser. 2 , 13, 71–88.

67 Ince, E. L., 1956: Ordinary Differential Equations. Dover, 558 pp.
68 Birkhoff, G. D., 1908: Boundary value and expansion problems of ordinary linear

differential equations. Trans. Amer. Math. Soc., 9, 373–395.

69 Bounitzky, E., 1909: Sur la fonction de Green des équations différentielles linéaires

ordinaires. J. Math. Pures Appl., Ser. 6 , 5, 65–126.

© 2015 by Taylor & Francis Group, LLC

24 Green’s Functions with Applications

with u(0) = u(1) = 0. In 1911 Hilb70 (1882–1929) wrote down the Green’s
functions in terms of linearly homogeneous solutions to Equation 1.6.14. Fur-
thermore, he found the bilinear representation of the Green’s function,
( ϕj (ξ)ϕj (x)
g(x|ξ) = , (1.6.15)
j
λ2j − λ2

where λj and ϕj (x) are the eigenvalues and eigenfunctions, respectively, of

the boundary-value problem.
Another differential equation that arises from physics is the simple har-
monic oscillator. In this case, the Green’s function is governed by

g $$ + k 2 g = δ(x − ξ). (1.6.16)

C

In 1910 A. Sommerfeld (1868–1951) examined the solution71 to Equation

op

1.6.16 over a finite interval 0 < x < L when the boundary conditions are
yr

g(0|ξ) = g(L|ξ) = 0. He expressed the Green’s function by the expansion

ig
ht

(∞
um (x)um (ξ)
ed

g(x|ξ) = , (1.6.17)
k 2 − km
2
M

m=1
at

where km and um (x) are the eigenvalues and eigenfunctions of the Sturm-
er

Liouville problem: u$$m + k 2 um = 0 with um (0) = um (L) = 0.

ia

The bilinear expansion found by Sommerfeld for the harmonic oscillator

l–

is a simple example of the general Sturm-Liouville problem:

Ta

% &
d du
y lo

− p(x) + q(x)u = λu, 0 < x < 1, (1.6.18)

dx dx
r&

with u(0) = u(1) = 0. In 1911, Adolf Kneser72 (1862–1930) showed how

Fr

the symmetric Green’s function for this problem could be formulated as the
an

homogeneous integral equation

ci
s

" 1
u(ξ) = λ g(x|ξ)u(x) dx. (1.6.19)
0

He showed that this integral equation lead directly to the general bilinear
formula
(∞
ψm (x)ψm (ξ)
g(x|ξ) = , (1.6.20)
m=1
λm

70 Hilb, E., 1911: Über Reihenentwicklungen nach den Eigenfunktionen linearer Diffe-
rentialgleichungen 2ter Ordnung. Math. Ann., 71, 76–87.
71 Sommerfeld, A., 1910: Die Greensche Funktion der Schwingungsgleichungen für ein

beliebiges Gebiet. Phys. Z., 11, 1057–1066.

72 Kneser, op. cit.

© 2015 by Taylor & Francis Group, LLC

Historical Development 25

where λm and ψm (x) are the mth eigenvalue and eigenfunction, respectively,
of Equation 1.6.18.
In the following year Sommerfeld considered the case when the interval
is infinite −∞ < x < ∞. He showed73 that the Green’s function in this case
is
1 ±ik(x−ξ)
g(x|ξ) = e , (1.6.21)
2ik
where the positive sign holds if x > ξ and the negative sign holds when
x < ξ. To find this solution, Sommerfeld74 introduced his famous “radiation
condition” that requires that energy always propagates out to infinity.
Another important ordinary differential equation whose Green’s function
was studied in the early twentieth century is the one that governs the deflection
of a beam75 with clamped ends:
C

g iv = −δ(x − ξ), 0 < x < L, (1.6.22)

op
yr

with the boundary conditions

ig
ht

g(0|ξ) = g(L|ξ) = g $$ (0|ξ) = g $$ (L|ξ) = 0. (1.6.23)

ed

In Chapter 3 we will show that

M
at

x< (x> − L) # 2 $
er

g(x|ξ) = − x< + x2> − 2Lx> . (1.6.24)

6L
ia
l–

A quick check shows that Equation 1.6.24 satisfies the ordinary differential
equation and boundary conditions. Furthermore, g $ (ξ|ξ) and g $$ (ξ|ξ) is con-
Ta

tinuous while
y

'ξ +
lo

d3 g ''
r&

= −1. (1.6.25)
dx3 ' − ξ
Fr

76
In 1914 A. Kneser published a paper on the integral equation that
an

expresses the vibrations of a string where the right boundary is attached to a

ci

mass and a spring. His analysis involved solving the Green’s function problem
s

(in modern notation) of

g $$ + λ2 g = −δ(x − ξ), 0 < x, ξ < L, (1.6.26)

73 Sommerfeld, A., 1912: Die Greensche Funktion der Schwingungsgleichung. Jahrber.

Deutsch. Math.-Verein., 21, 309–353.

74 Schot, S. M., 1992: Eighty years of Sommerfeld’s radiation condition. Hist. Math.,

19, 385–401.
75 See Section 1.22 and 1.23 in Bateman, H., 1959: Partial Differential Equations of
Mathematical Physics. Cambridge, 522 pp. See also Von Mises, R., Ph. Frank, H. Weber,
and B. Riemann, 1925: Die Differential- und Integralgleichungen der Mechanik und Physik.
Vol I. Braunschweig, F. Vieweg, 687 pp.
76 Kneser, A., 1914: Belastete Integralgleichungen. Rend. Circ. Matem. Palermo, 37,

169–197.

© 2015 by Taylor & Francis Group, LLC

26 Green’s Functions with Applications

with
g(0|ξ) = g $ (L|ξ) + (p − qλ2 )g(L|ξ) = 0, (1.6.27)
where λ &= λn , the eigenvalue of the system, and q > 0. This problem is
of particular interest because λ appears in the differential equation and the
boundary condition. He found that the corresponding Green’s function is

{(qλ2 − p) sin[λ(L − x> )] − λ cos[λ(L − x> )]} sin(λx< )

g(x|ξ) = . (1.6.28)
λ(qλ2 − p) sin(λL) − λ2 cos(λL)

He also found the bilinear expansion for this Green’s function.

In the papers cited above, the differential equations were incompatible
- there are no nonvanishing solutions which satisfy both the homogeneous
differential equation and the boundary conditions. However, as early as 1904,
C

Hilbert77 gave simple examples of compatible differential equations for linear

op

second-order, boundary-value problems of the form

yr
ig

L[u] = [p(t)u$ ]$ − q(t)u = 0. (1.6.29)

ht
ed

He then constructed a generalized Green’s function (Greensche Funktionen im

M

erweiterten Sinne) to treat such cases. In 1909, Westfall78 proved Hilbert’s

at

results. Almost two decades later, Elliott79 generalized Hilbert’s results by

er
ia

finding the generalized Green’s function for a nth order differential system.
l–

His results are not limited to self-adjoint systems.

Since these early studies, the study of generalized Green’s functions has
Ta

languished. Loud80 used regular differential operator theory to explain gen-

y lo

eralized Green’s functions and developed techniques to find them. In 1977

r&

Locker81 explored the properties of generalized Green’s functions.

Fr
an
ci
s

77 See pp. 44−45 in Hilbert, D., 1912: Grundzüge einer allgemeiner Theorie der linearen

Integralgleichungen. Leipzig, B. G. Teubner, 312 pp.

78 Westfall, W. D. A., 1909: Existence of the generalized Green’s function. Ann. Math.,

Ser. 2 , 10, 177–180.

79 Elliott, W. W., 1928: Generalized Green’s function for compatible differential systems.

Amer. J. Math., 50, 243–258; Elliott, W. W., 1929: Green’s function for differential systems
containing a parameter. Amer. J. Math., 51, 397–416.
80 Loud, W. S., 1970: Some examples of generalized Green’s functions and generalized

Green’s matrices. SIAM Rev., 12, 194–210.

81 Locker, J., 1977: The generalized Green’s function for the nth order linear differential

operator. Trans. Amer. Math. Soc., 228, 243–268.

© 2015 by Taylor & Francis Group, LLC