Integral Transform Techniques For Green's Function: Kazumi Watanabe

Lecture Notes in Applied and Computational Mechanics 76
Kazumi Watanabe
Integral
Transform
Techniques for
Green’s Function
Second Edition
Lecture Notes in Applied and Computational
Mechanics
Volume 76
Series editors
Friedrich Pfeiffer, Technische Universität München, Garching, Germany
e-mail: pfeiffer@amm.mw.tum.de
Peter Wriggers, Universität Hannover, Hannover, Germany

e-mail: wriggers@ikm.uni-hannover.de
About this Series
This series aims to report new developments in applied and computational

mechanics—quickly, informally and at a high level. This includes the fields of fluid,
solid and structural mechanics, dynamics and control, and related disciplines. The
applied methods can be of analytical, numerical and computational nature.
More information about this series at http://www.springer.com/series/4623

Kazumi Watanabe
Integral Transform
Techniques for Green’s
Function
Second Edition
123
Kazumi Watanabe
Yamagata University
Yonezawa
Japan
ISSN 1613-7736 ISSN 1860-0816 (electronic)

Lecture Notes in Applied and Computational Mechanics
ISBN 978-3-319-17454-9 ISBN 978-3-319-17455-6 (eBook)
DOI 10.1007/978-3-319-17455-6
Library of Congress Control Number: 2013940095
Springer Cham Heidelberg New York Dordrecht London

© Springer International Publishing Switzerland 2014, 2015
This work is subject to copyright. All rights are reserved by the Publisher, whether the whole or part
of the material is concerned, specifically the rights of translation, reprinting, reuse of illustrations,
recitation, broadcasting, reproduction on microfilms or in any other physical way, and transmission
or information storage and retrieval, electronic adaptation, computer software, or by similar or
dissimilar methodology now known or hereafter developed.
The use of general descriptive names, registered names, trademarks, service marks, etc. in this
publication does not imply, even in the absence of a specific statement, that such names are exempt
from the relevant protective laws and regulations and therefore free for general use.
The publisher, the authors and the editors are safe to assume that the advice and information in this
book are believed to be true and accurate at the date of publication. Neither the publisher nor the
authors or the editors give a warranty, express or implied, with respect to the material contained
herein or for any errors or omissions that may have been made.
Printed on acid-free paper
Springer International Publishing AG Switzerland is part of Springer Science+Business Media

(www.springer.com)
Dedicated to my teachers,
Dr. Akira Atsumi (Late Professor,
Tohoku University)
and
Dr. Kyujiro Kino (Late Professor,
Osaka Institute of Technology)
Preface to the Second Edition
After publishing the original version, the author noticed that more detailed math-
ematical techniques should be included so that the young reader could learn the
traditional analytical techniques without any mathematical skip. That is, the dis-
cussion on square root functions. In many dynamic/wave problems, we frequently
encounter the square root function which is the typical multi-valued function and
have to introduce branch cuts in the complex plane for the inversion integral.
A simple elementary technique for the introduction of the branch cut and the
discussion on the argument of the square root function along the cut are included in
Chap. 1 as Sect. 1.3. This branch cut is employed throughout the book and applied
to the inversion integrals in Sect. 2.5 and other sections. Due to the introduction
of the unified branch cut, Sect. 2.5 for the time-harmonic Green’s function is wholly
rewritten.
In the revising process the author also noticed five exact closed-form solutions:
three are Green’s functions for torsion problems, the fourth one is for the reflection
problem and the last is for the scattering problem. Green’s functions for the torsion
problem are inserted in Chaps. 3 and 7. SH-wave reflection at a moving boundary in
Sect. 7.4 is slightly rewritten in order to include the closed-form solution. Section 7.5
is newly inserted and shows the exact closed-form solution for a wave scattering
problem in an inhomogeneous elastic solid. Further, employing the branch cut
described in Chap. 1, an excellent application technique of the complex integral is
explained in the last Sect. 7.6. It is the transformation of a semi-infinite integral to a
finite one that is suitable for numerical computations. Needless to say, many errors
and mistakes in the original version have also been corrected.
The author hopes the young reader can learn one of the traditional analytical
techniques, especially the application of the complex integral for the integral
transform. Thus, the present revised version is more instructive than the original
one, and every question and inquiry via email “kazy@yz.yamagata-u.ac.jp” is
welcome.
Hikoshima Island, Japan, January 2015 Kazumi Watanabe
vii
Preface
When I was a senior student, I found a book on the desk of my advisor professor
and asked him how to get it. His answer was negative, saying its content was too
hard, even for a senior student. Some weeks later, I found it again in a book store,
the biggest one in Osaka. This was my first encounter with “Fourier Transforms”
written by the late Prof. I. N. Sneddon. Since then, I have learned the power of
integral transform, i.e. the principle of superposition.
All phenomena, regardless of their fields of event, can be described by differ-
ential equations. The solution of the differential equation contains the crucial
information to understand the essential feature of the phenomena. Unfortunately,
we cannot solve every differential equation, and almost all phenomena are governed
by nonlinear differential equations, of which most are not tractable. The differential
equations that can be solved analytically are limited to a very small number. But
their solutions give us the essence of the event. The typical partial differential
equations that can be solved exactly are the Laplace, the diffusion and the wave
equations. These three partial differential equations, which are linearized for sim-
plicity, govern many basic phenomena in physical, chemical and social events. In
addition to single differential equations, some coupled linear partial differential
equations, which govern somewhat complicated phenomena, are also solvable and
their solutions give much information about, for example, the deformation of solid
media, propagation of seismic and acoustic waves, and fluid flows.
In a case where phenomena are described by linear differential equations, the
solutions can be expressed by superposition of basic/fundamental solutions. The
integral transform technique is a typical superposition technique. The integral
transform technique does not require any previous knowledge for solving differ-
ential equations. It simply transforms partial or ordinary differential equations to
reduced ordinary differential equations or to simple algebraic equations. However, a
substantial difficulty is present regarding the inversion process. Many inversion
integrals are tabulated in various formula books, but typically, this is not enough. If
a suitable integration formula cannot be found, the complex integral must be
considered and Cauchy’s integral theorem is applied to the inversion integral. Thus,
ix
x Preface
integral transform techniques are intrinsically connected with the theory of complex
integrals.
The present book intends to show how to apply integral transforms to partial
differential equations and how to invert the transformed solution into the actual
space-time domain. Not only the use of integration formula tabulated in books, but
also the application of Cauchy’s integral theorem for the inversion integrals are
described concisely and in detail. A particular solution for a differential equation
with a nonhomogeneous term of a point source is called the “Green’s function.”
The Green’s functions for coupled differential equations are called “Green’s dyadic.”
The Green’s function and Green’s dyadic are the basic and fundamental solution
of the differential equation and give the principal features of the event. Furthermore,
these Green’s functions and dyadics have many applications for numerical com-
putation techniques such as the Boundary Element Method. However, the Green’s
function and Green’s dyadic have been scattered in many branches of applied
mechanics and thus, their solution methods are not unified. This book intends to
present and illustrate a unified solution method, namely the method of integral
transform for the Green’s function and Green’s dyadic. Thus, the fundamental
Green’s function for the Laplace and wave equations and the Green’s dyadic for
elasticity equations are gathered in this single book so that the reader can have access
to a proper Green’s function and understand the mathematical process for its
derivation.
Chapter 1 describes roughly the definition of the integral transforms and the
distributions to be used throughout the book. Chapter 2 shows how to apply an
integral transform for solving a single partial differential equation such as the
Laplace equation and the wave equation. The basic technique of the integral
transform method is demonstrated. Especially, in the case of the time-harmonic
response for the wave equation, the integration path for the inversion integral is
discussed in detail. At the end of the chapter, the obtained Green’s functions are
listed in a table so that the reader can easily find the difference of the functional
form among the Green’s functions. An evaluation technique for a singular inversion
integral which arises in a 2D static problem of Laplace equation is also developed.
The Green’s dyadic for 2D and 3D elastodynamic problems are discussed in
Chap. 3. Three basic responses, impulsive, time-harmonic and static responses, are
obtained by the integral transform method. The time-harmonic response is derived
by the convolution integral of the impulsive response without solving the differ-
ential equations for the time-harmonic source.
Chapter 4 presents the governing equations for acoustic waves in a viscous fluid.
Introducing a small parameter, the nonlinear field equations are linearized and
reduced to a single partial differential equation for velocity potential or pressure
deviation. The Green’s function which gives the acoustic field in a uniform flow is
derived by the method of integral transform. A conversion technique for the
inversion integral is demonstrated. That is, to transform an inversion integral along
the complex line to that along the real axis in the complex plane. It enabled us to
apply the tabulated integration formula.
Preface xi
Chapter 5 presents Green’s functions for beams and plates. The dynamic
response produced by a point load on the surface of a beam and a plate is discussed.
The impulsive and time-harmonic responses are derived by the integral transform
method. In addition to the tabulated integration formulas, the inversion integrals are
evaluated by application of complex integral theory.
Chapter 6 presents a powerful inversion technique for transient problems of
elastodynamics, namely the Cagniard-de Hoop method. Transient response of an
elastic half space to a point impulsive load is discussed by the integral transform
method. Applying Cauchy’s complex integral theorem, the Fourier inversion
integral is converted to an integral of the Laplace transform and then its Laplace
inversion is carried out by inspection without using any integration formula. The
Green’s function for an SH-wave and Green’s dyadics for P, SV and SH-waves are
obtained.
The last Chap. 7 presents three special Green’s functions/dyadics. The 2D static
Green’s dyadic for an orthotropic elastic solid and that for an inhomogeneous solid
are derived. In the last section, a moving boundary problems is discussed. Two
different Laplace transforms are applied for a single problem, and a conversion
formula between two Laplace transforms is developed with use of Cauchy’s the-
orem. This conversion enables us to apply the integral transform technique to a
moving boundary problem.
The integral transform technique has been used for many years. The inversion
process inevitably requires a working knowledge of the theory of complex func-
tions. The author finds the challenge of a complex integral amusing, especially the
challenge of choosing the right contour for the inversion integral. He hopes that
young researchers will join the fun and carry on with the inversion techniques. In
this respect it must be mentioned that he feels a lack of mathematical skill in the
recent research activities, since some researchers tend to use numerical techniques
without considering the possibility of an analytical solution. The more mathemat-
ical techniques expand the horizon of the differential equations wider and one can
extract more firm knowledge from the nature which is described by the differential
equations. The author hopes that the present book gives one more technique to the
younger researchers.
Finally, the author wishes to express his sincere thanks to Dr. Mikael A.
Langthjem, Associate Professor of Yamagata University, for his advice and nice
comments.
Yonezawa, Japan, January 2013 Kazumi Watanabe

Contents
1 Definition of Integral Transforms and Distributions . . . . . . . . . . . 1

1.1 Integral Transforms . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.2 Distributions and Their Integration Formulas . . . . . . . . . . . . . 6
1.3 Branch Cut and Argument of Square Root Functions . . . . . . . 11
pffiffiffiffiffiffiffiffiffiffiffiffi
1.3.1 Square Root Function 1: gðzÞ ¼ pzffiffiffiffiffiffiffiffiffiffiffiffiffi z0 .ffi . . . . . . . . . 11
1.3.2 Square Root Function 2: gðzÞ ¼ z2 z20 . . . . . . . . . 14
1.4 Comments on Inversion Techniques and Integration
Formulas . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ....... 28
References. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ....... 32
2 Green’s Functions for Laplace and Wave Equations . . . . . . . . . . 33

2.1 1D Impulsive Source. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33
2.2 1D Time-Harmonic Source. . . . . . . . . . . . . . . . . . . . . . . . . . 38
2.3 2D Static Source. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44
2.6 3D Static Source. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 68
Appendix . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 76
References. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 76
3 Green’s Dyadic for an Isotropic Elastic Solid . . . . . . . . . . . . . . . . 77

3.1 2D Impulsive Source. . . . . . . . . . ..... . . . . . . . . . . . . . . . 79
3.2 2D Time-Harmonic Source. . . . . . ..... . . . . . . . . . . . . . . . 87
3.3 2D Static Source. . . . . . . . . . . . . ..... . . . . . . . . . . . . . . . 89
3.4 3D Impulsive Source. . . . . . . . . . ..... . . . . . . . . . . . . . . . 96
3.5 3D Time-Harmonic Source. . . . . . ..... . . . . . . . . . . . . . . . 107
3.6 3D Static Source. . . . . . . . . . . . . ..... . . . . . . . . . . . . . . . 108
xiii
xiv Contents
3.7 Torsional Source. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 109

3.7.1 Ring Source . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 110
3.7.2 Point Torque Source . . . . . . . . . . . . . . . . . . . . . . . . 113
Appendix . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 116
References. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 119
4 Acoustic Wave in a Uniform Flow . . . . . . . . . . . . . . . . . . . . . . . . 121

4.1 Compressive Viscous Fluid . . . . . . . . . . . . . . . . . . . . . . . . . 121
4.2 Linearization . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 123
4.3 Viscous Acoustic Fluid . . . . . . . . . . . . . . . . . . . . . . . . . . . . 126
4.4 Wave Radiation in a Uniform Flow. . . . . . . . . . . . . . . . . . . . 129
4.5 Time-Harmonic Wave in a Uniform Flow . . . . . . . . . . . . . . . 135
References. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 137
5 Green’s Functions for Beam and Plate. . . . . . . . . . . . . . . . . . . . . 139

5.1 An Impulsive Load on a Beam . . . . . . . . . . . . . . . . . . . . . . . 139
5.2 A Moving Time-Harmonic Load on a Beam . . . . . . . . . . . . . 142
5.3 An Impulsive Load on a Plate . . . . . . . . . . . . . . . . . . . . . . . 145
5.4 A Time-Harmonic Load on a Plate . . . . . . . . . . . . . . . . . . . . 148
Appendix . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 152
References. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 152
6 Cagniard-de Hoop Technique . . . . . . . . . . . . . . . . . . . . . . . . . . . 153

6.1 2D Anti-plane Deformation . . . . . . . . . . . . . . . . . . . . . . . . . 154
6.2 2D In-plane Deformation . . . . . . . . . . . . . . . . . . . . . . . . . . . 162
6.3 3D Dynamic Lamb’s Problem . . . . . . . . . . . . . . . . . . . . . . . 178
References. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 204
7 Miscellaneous Green’s Functions . . . . . . . . . . . . . . . . . . . . . . . .. 205

7.1 2D Static Green’s Dyadic for an Orthotropic Elastic Solid . . .. 205
7.2 2D Static Green’s Dyadic for an Inhomogeneous
Elastic Solid. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 213
7.2.1 2D Kelvin’s Solution for Homogeneous Media. . . . .. 221
7.3 Green’s Function for Torsional Waves in a Monoclinic
Material. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 222
7.4 Reflection of a Transient SH-Wave at a Moving Boundary . .. 227
7.5 Wave Scattering by a Rigid Inclusion in an Inhomogeneous
Elastic Solid. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 242
7.6 An Excellent Application of Cauchy Complex Integral . . . . .. 253
References. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 260
Index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 261
Chapter 1
Definition of Integral Transforms
and Distributions
This first chapter describes a brief definition of integral transforms, such as Laplace
and Fourier transforms, a rough definition of delta and step functions which are
frequently used as the source function, and a concise introduction of the branch cut
for a multi-valued square root function. The multiple integral transforms and their
notations are also explained. The introduction of the branch cut and the discussion
on the argument of the root function along the branch cut will be helpful for
applying the complex integral to the inverse transform. The last short comment lists
some important formula books which are crucial for the inverse transform, i.e. the
evaluation of the inversion integral.
1.1 Integral Transforms
For a well-defined function f ðxÞ; x 2 ða; bÞ, when the integral with the kernel
function Kðn; xÞ,
Zb
FðnÞ ¼ Kðn; xÞf ðxÞdx ð1:1:1Þ
a
has its inverse integral with another kernel function K ðn; xÞ,
Z
f ðxÞ ¼ K ðn; xÞFðnÞdn ð1:1:2Þ
L
we call this integration pair an “integral transform.” The function f ðxÞ is an original
function and the function FðnÞ is the “image or transformed function” in the
transformed domain. If the reciprocal f ðxÞ , FðnÞ holds, we call FðnÞ the “integral
© Springer International Publishing Switzerland 2015 1

K. Watanabe, Integral Transform Techniques for Green’s Function,
Lecture Notes in Applied and Computational Mechanics 76,
DOI 10.1007/978-3-319-17455-6_1
2 1 Definition of Integral Transforms and Distributions
transform” of f ðxÞ and the two-variable-functions Kðn; xÞ and K ðn; xÞ the kernels
of the integral transform.
We have already learned many integral transforms which are classified and
named depending on the kernel function and the integration range. A well-known
integral transform is the Laplace transform. The (one-sided) Laplace transform is, in
the present book, defined for the time-variable function f ðtÞ; t 2 ½0; 1Þ as
Z1

f ðsÞ ¼ f ðtÞ expðstÞdt ð1:1:3Þ
0
where “s” is the transform parameter and the transform kernel is expðstÞ. The
inverse transform is also defined by the integral along the complex line,
Z
cþi1
1
f ðtÞ ¼ f ðsÞ expðstÞds ð1:1:4Þ
2pi
ci1
where the integration path from c i1 to c þ i1 is called the “Bromwich line.”

The real constant “c” must be larger than the real part of any singular point of the
transformed function f ðsÞ. Thus this line is placed at far right from all singular
points in the complex s-plane.
In these definitions, the variable s is called the transform parameter and the two
kernels for the transform and the inverse transform are exponential functions:
1
Kðs; tÞ ¼ expðstÞ; K ðs; tÞ ¼ expðstÞ ð1:1:5Þ
2pi
Further, the integration ranges are also different from each other. The transform
integral is carried out along the semi-infinite real line ½0; 1Þ for the time and the
inverse transform is carried out along an infinite line ðc i1; c þ i1Þ in the
complex s-plane.
Since any notation for the transform parameter is available, one should be aware
of the notation of the parameter since some authors use “p” instead of “s.” When the
Laplace inversion is carried out by using some inversion formulas in a reference
book, not performing the inversion integral in the complex plane, the symbolic form
of the Laplace inversion
f ðtÞ ¼ L1 ½f ðsÞ ð1:1:6Þ
is used for the sake of simplicity. The present book also employs frequently this
simple expression for the Laplace inversion.
So far, many integral transform pairs have been found and defined. We choose
and use one suitable integral transform depending on the geometry (integration
range) and the simplicity of its application. The followings are typical integral
transforms which are much used in applications.
1.1 Integral Transforms 3
(1) Finite Fourier transform (Fourier series): f ðxÞ; x 2 ½p; p

When an original function is defined within a finite region, the Fourier finite
transform, i.e. Fourier series is used and its transform pair can be defined as follows:
(1.1) Complex Fourier series
Zp
1 X 1
fn ¼ f ðxÞ expðinxÞdx; f ðxÞ ¼ fn expðinxÞ ð1:1:7Þ
2p n¼1
p
where the index “n” is an integer.

(1.2) Fourier cosine series
Zp Zp
1 1X 1
a0 ¼ f ðxÞdx; an ¼ f ðxÞ cosðnxÞdx; f ðxÞ ¼ an cosðnxÞ ð1:1:8Þ
2 p n¼0
p p
(1.3) Fourier sine series
Zp
1X 1
bn ¼ f ðxÞ sinðnxÞdx; f ðxÞ ¼ bn sinðnxÞ ð1:1:9Þ
p n¼1
p
(2) Complex Fourier transform: f ðxÞ; x 2 ð1; 1Þ

When the function is defined in an infinite region, the complex Fourier transform
pair is defined by
Z1 Z1
f ðnÞ ¼ 1 f ðnÞ expðinxÞdn
f ðxÞ expðinxÞdx; f ðxÞ ¼ ð1:1:10Þ
2p
1 1
(2.1) Fourier transform with a non-uniform parameter.

The Fourier transform with a non-uniform parameter is defined as
Zb
f ðnÞ ¼ f ðxÞ expfþixhðnÞgdx ð1:1:11aÞ
a
Zþ1
1 f ðnÞ expfixhðnÞgh0 ðnÞdn
f ðxÞ ¼ ð1:1:11bÞ
2p
1
where hðnÞ is a monotonically increasing/decreasing function, and the lower and

upper integration limits, a and b, are arbitrary constants.
Let us verify the pair of Eq. (1.1.11), briefly. Since hðnÞ is the monotonically
increasing or decreasing function, we can define its inverse function uniquely as
g ¼ hðnÞ , n ¼ h1 ðgÞ ð1:1:12Þ
Using the above definition, Eq. (1.1.11a) is rewritten in the form of the regular
Fourier transform,
Zb
f ðh1 ðgÞÞ ¼ f ðxÞ expðþixgÞdx ð1:1:13Þ
a
Applying the inversion formula of Fourier transform in Eq. (1.1.10) to the above
equation, the Fourier inversion of Eq. (1.1.13) is given by
Zþ1
1 f ðh1 ðgÞÞ expðixgÞdg;
f ðxÞ ¼ a\x\b ð1:1:14Þ
2p
1
As we have defined the inverse function in Eq. (1.1.12), the integration variable g is
returned to the original variable n,
n ¼ h1 ðgÞ ) g ¼ hðnÞ; dg ¼ h0 ðnÞdn ð1:1:15Þ
Then, Eq. (1.1.14) is rewritten as
Zþ1
1 f ðnÞ expfixhðnÞgh0 ðnÞdn
f ðxÞ ¼ ð1:1:16Þ
2p
1
This is the second of the transform pair Eq. (1.1.11b). Therefore, the pair of the
Fourier transform with the non-uniform parameter hðnÞ is verified.
(3) Fourier cosine/sine transform: f ðxÞ; x 2 ½0; 1Þ
When an original function is even or odd in an infinite region, or the function is
defined in a semi-infinite region, we employ Fourier cosine or sine transform. They
are defined as follows:
1.1 Integral Transforms 5
(3.1) Fourier cosine transform
Z1 Z1
f ðCÞ ðnÞ ¼ 2 f ðCÞ ðnÞ cosðnxÞdn
f ðxÞ cosðnxÞdx; f ðxÞ ¼ ð1:1:17Þ
p
0 0
(3.2) Fourier sine transform
Z1 Z1
f ðSÞ ðnÞ ¼ 2 f ðSÞ ðnÞ sinðnxÞdn
f ðxÞ sinðnxÞdx; f ðxÞ ¼ ð1:1:18Þ
p
0 0
(4) Hankel transform: f ðxÞ; x 2 ½0; 1Þ

Another semi-infinite integral transform is Hankel transform (Sneddon 1951,
p. 48) defined by
Z1 Z1
~fn ðnÞ ¼ xf ðxÞJn ðnxÞdx; f ðxÞ ¼ n ~fn ðnÞJn ðnxÞdn ð1:1:19Þ
0 0
where Jn ðzÞ is the n-th order Bessel function of the first kind.
In order to guarantee the application of the integral transform, each integral must
converge. For example, the Fourier transform requires a convergence condition at
infinity,
f ðxÞ ) Oðjxjm Þ; m [ 1 ð1:1:20Þ

x!1
for the original function. However, if we employ the Hankel transform, it is

allowable for the original function to be finite since the Bessel function decays with
the order of the inverse square root at the infinity.
In some applications, where we encounter multi-variable functions, or where a
multiple integral transform is employed, a different mark for each transform is
defined such as f ; ~f ; f , and the multiple transform is denoted by piling up the
transform marks, like ~f . In the present book, we employ the Laplace transform

with respect to the time variable t as defined in Eqs. (1.1.3) and (1.1.4). For the
space variables ðx; y; zÞ, we apply three Fourier transforms with the transform
parameters ðn; g; fÞ respectively. They are defined as
Z1 Z1
f ðxÞ expðþinxÞdx; f ðxÞ ¼ ð1:1:21aÞ
2p
1 1
Z1 Z1
~f ðgÞ ¼ 1 ~f ðgÞ expðigyÞdg
f ðyÞ expðþigyÞdy; f ðyÞ ¼ ð1:1:21bÞ
2p
1 1
Z1 Z1
^f ðfÞ ¼ 1 ^f ðfÞ expðifzÞdf
f ðzÞ expðþifzÞdz; f ðzÞ ¼ ð1:1:21cÞ
2p
1 1
Please remember that the pair of the space variable and transform parameter is fixed
throughout the present book, such as the pair, ðx; nÞ; ðy; gÞ and ðz; fÞ.
1.2 Distributions and Their Integration Formulas
For modeling engineering phenomena, many mathematical functions are used.

Elementary and some special functions are used for continuous phenomena. But for
discontinuous phenomena, distributions such as delta and step functions are fre-
quently used. This subsection explains briefly the definition of three distributions:
Dirac’s delta function, Heaviside’s unit step function and Heisenberg’s delta
function.
(1) Heaviside’s unit step function: Hðx aÞ
Heaviside’s unit step function is defined by the graphical form in Fig. 1.1. This
function takes the value 0 when its argument is negative and +1 when the argument
is positive,

þ1; x [ a
Hðx aÞ ¼ ð1:2:1Þ
0; x\a
Due to this definition, the zero argument at x ¼ a is a discontinuous point. So, it takes
two limiting values from the positive and negative sides of the discontinuous point,
Hð0þ Þ ¼ þ1; Hð0 Þ ¼ 0 ð1:2:2Þ
Then, we have to understand that the step function is not defined at x ¼ a.
Fig. 1.1 Heaviside’s unit

step function
H ( x − a)
+1
x =a
1.2 Distributions and Their Integration Formulas 7
(a) (b) δ ( x − a)
↑
ε 1/ ε ε →0
0 0 0 0
x=a x=a
Fig. 1.2 Schematic definition of Dirac’s delta function a rectangular pulse b Dirac’s delta
function
(2) Dirac’s delta function: dðx aÞ

Dirac’s delta function is defined as the limit e ! 0 of a rectangular pulse with
width e and height 1=e as shown in Fig. 1.2.
The center of the rectangular distribution is fixed at x ¼ a in the limiting process
and we understand the point as an application point of the delta function. It is
denoted by dðx aÞ. In the limit, the width of the delta function vanishes and the
functional value becomes infinite. But, its internal invisible area is one due to the
definition ð1=eÞ e ¼ 1. This nature makes the evaluation of the integral very
simple, and we have the formula where the integrand includes the delta function as
Zb
f ðcÞ ; a \ c \ b
f ðxÞdðx cÞdx ¼ ð1:2:3Þ
0; c \ a or b \ c
a
Using this simple integration formula, we can obtain an integral representation

for the delta function. Applying Fourier transform defined by Eq. (1.1.10) to the
delta function, f ðxÞ ¼ dðx cÞ, we use the integration formula (1.2.3). The trans-
form integral is evaluated as
Z1
f ðnÞ ¼ dðx cÞ expðþinxÞdx ¼ expðþincÞ ð1:2:4Þ
1
The inverse Fourier transform is also applied to the above f ðnÞ and its integration
range is reduced to the semi-infinite,
Z1 Z1
1 1
f ðxÞ ¼ dðx cÞ ¼ expfinðx cÞgdn ¼ cosfnðx cÞgdn ð1:2:5Þ
2p p
1 0
Then, we have the integral representation for Dirac’s delta function, i.e.
Z1
1
dðx cÞ ¼ cosfnðx cÞgdn ð1:2:6Þ
p
0
Finally, we would like to add one useful formula between delta and step func-
tions. That is
d
dðx aÞ ¼ Hðx aÞ ð1:2:7Þ
dx
This relation will be understood from the graphical discussion on delta and step
functions.
(3) Heisenberg’s delta function: d ðxÞ
In many applications, we have to express a semi-infinite distribution of a
physical quantity, such as the uniform load over the surface. In order to treat the
semi-infinite distribution, another delta function is defined. That is Heisenberg’s
delta function which is the Fourier transform of the semi-infinite distribution.
Let us consider Fourier transform of Heaviside’s unit step function,

1; 0\x\ þ 1
HðxÞ ¼ ð1:2:8Þ
0; 1 \ x \ 0
Its formal Fourier transform is
Z1 Z1

hðnÞ ¼ HðxÞ expðþinxÞdx ¼ expðþinxÞdx ð1:2:9Þ
1 0
This integral cannot be evaluated in the regular sense of calculus. Instead, we shall
look for the transformed image function hðnÞ so that its Fourier inversion integral
results in the step function.
Remember the delta function whose Fourier transform is a constant, and assume
that the image function is the sum of delta function and a newly introduced
unknown function h1 ðnÞ,
hðnÞ ¼ pdðnÞ þ h1 ðnÞ ð1:2:10Þ
Let us apply the Fourier inversion integral to the above function and look for the
suitable form of the unknown function h1 ðnÞ. The inversion integral is processed as
1.2 Distributions and Their Integration Formulas 9
Z1 Z1
1 1
HðxÞ ¼
hðnÞ expðinxÞdn ¼ pdðnÞ þ
h1 ðnÞ expðinxÞdn
2p 2p
1 1
Z1
1 1 must be 1; x[0
¼ þ h1 ðnÞ expðinxÞdn ) ð1:2:11Þ
2 2p 0; x\0
1
From the last line in the above equation, we learn that the Fourier transform
(inversion) integral of the unknown function h1 ðnÞ must be þ1=2 in x [ 0 and
1=2 in x \ 0, i.e.
Z1 (
1 þ1=2 ; x[0
h1 ðnÞ expðinxÞdn ¼ ð1:2:12Þ
2p 1=2; x\0
1
We remember the integration formula that gives 1=2 in each semi-infinite

x-region (Erdélyi 1954, vol I, pp. 64, 3), that is
Z1 Z1 (
1 sinðxnÞ 1 i þ1=2 ; x[0
dn ¼ expðixnÞdn ¼ ð1:2:13Þ
p n 2p n 1=2 ; x\0
0 1
Comparing Eq. (1.2.12) with Eq. (1.2.13), we learn that the whole integrand in
Eq. (1.2.12) must be the function
sinðxnÞ
ð1:2:14Þ
n
and thus we can guess that the suitable functional form for h1 ðnÞ is
h1 ðnÞ ¼ i ð1:2:15Þ

n
We shall now examine whether this form is really suitable or not. Substitute the
above Eq. (1.2.15) into the last line in Eq. (1.2.11) and split the integral into real
and imaginary parts, we get
Z1 Z1 Z1
1 1 i 1 1 1 i 1
þ expðinxÞdx ¼ þ sinðnxÞdx þ cosðnxÞdx
2 2p n 2 2p n 2p n
1 1 1
ð1:2:16Þ
The third term on the right hand side vanishes due to the anti-symmetric nature of
the integrand. The second term is evaluated with the aid of formula (1.2.13) as
Z1 Z1
1 1 sinðnxÞ 1 1 sinðnxÞ 1 1 þ1 ; x[0
þ dn ¼ þ dn ¼ þ
2 2p n 2 p n 2 2 1 ; x\0
1 0
1; x[0
¼ ð1:2:17Þ
0; x\0
Then, we learn that the supposed function h1 ðnÞ in Eq. (1.2.15) is correct and that
the Fourier transform of the step function is given by
hðnÞ ¼ pdðnÞ þ i ð1:2:18Þ

n
This function is called “Heisenberg’s delta function” and its transform pair is
Z1 Z1
i 1
dþ ðnÞ ¼ pdðnÞ þ ¼ expðþinxÞdx; HðxÞ ¼ dþ ðnÞ expðinxÞdn
n 2p
0 1
ð1:2:19Þ
Heisenberg’s delta function dþ ðnÞ is the Fourier transform of the step function
HðxÞ. If we have the another step function HðxÞ, its transform pair is also given by
Z0 Z1
i 1
d ðnÞ ¼ pdðnÞ ¼ expðþinxÞdx; HðxÞ ¼ d ðnÞ expðinxÞdn
n 2p
1 1
ð1:2:20Þ
Then, Heisenberg’s delta function has two definitions as
Z1
i 1
d ðnÞ ¼ pdðnÞ ; HðxÞ ¼ d ðnÞ expðinxÞdn ð1:2:21Þ
n 2p
1
As a byproduct, we can obtain a new formula. Two operations, addition and

subtraction of Eqs. (1.2.19) and (1.2.20) give us the formulas as
Z1 Z1
1
cosðnxÞdx ¼ pdðnÞ; sinðnxÞdx ¼ ð1:2:22Þ
n
0 0
The formula on the left is the same as the integral in Eq. (1.2.6), but with c ¼ 0.
1.3 Branch Cut and Argument of Square Root Functions 11
1.3 Branch Cut and Argument of Square Root Functions
When we apply the multiple integral transform to the dynamic problem, some root
(radical) functions of the complex variable appear. The square root function is one of
multi-valued functions and we have to introduce branch cuts in order to make the root
function single-valued. Many textbooks for the complex analysis describe the branch
cut/line for very simple multi-valued functions. However, in the practical applica-
tions, we have to consider the branch cut for a little bit complicated function. In this
subsection, we show how to introduce the branch cut for the fundamental square root
function following the method of Ewing et al. (1957, Sect. 2.5, pp. 44–49). This
method is very simple and is easily understood even for college students. After
introducing the branch cut, we also show the argument of the root function along the
branch cut and that our introduction of the branch cut satisfies the regular conver-
gence condition in the complex plane. This condition is crucial for the inversion
integral.
We shall discuss the branch cut and the argument of two square root (radical)
pffiffiffiffiffiffiffiffiffiffiffiffi pffiffiffiffiffiffiffiffiffiffiffiffiffiffi
functions: gðzÞ ¼ z z0 ; z2 z20 where z0 is a complex constant. The branch
cut is introduced under the conditional: the real part of the root function is positive
in the complex z-plane, i.e. RefgðzÞg 0. Discussions on these two root functions
are carried out, separately.
1.3.1 Square Root Function 1: gðzÞ ¼ z z0
Let us consider the introduction of the branch cut/line for a simple square root
function which is defined by
gðzÞ ¼ z z0 ð1:3:1Þ
where zð¼ x þ iyÞ is the complex variable and z0 ð¼ a þ ibÞ is an arbitrary complex
constant. We impose the conditional
Reð z z0 Þ 0 ð1:3:2Þ
on the root function.

It is easily found that the branch point is z ¼ z0 which is derived from gðzÞ ¼ 0.
In order to discuss the branch cut, we introduce the capital letters, X and Y, which
denote the real and imaginary parts of the root function, respectively. The root
function is rewritten in the detailed form,
pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
X þ iY ¼ x þ iy ða þ ibÞ ð1:3:3Þ
Squaring both sides, and equating the real and imaginary parts, we have the
simultaneous equations for two unknowns, X and Y, as
(
X2 Y 2 ¼ x a
ð1:3:4Þ
XY ¼ ðy bÞ=2
Since the conditional is equivalent to

X ¼ Reð z z0 Þ 0 ð1:3:5Þ
the border for this conditional is X ¼ 0 and then, from the second equation in
(1.3.4), we have
yb¼0 ð1:3:6Þ
On the other hand, substituting X ¼ 0 into the first equation in (1.3.4), we have
Y 2 ¼ x a ð1:3:7Þ
Since Y 2 \ 0; x must be smaller than a,
x\a ð1:3:8Þ
The two conditions, Eqs. (1.3.6) and (1.3.8), give one semi-infinite line
ðx \ a; y ¼ bÞ in the complex z-plane. This semi-infinite horizontal line withthe
edge ðx ¼ a; y ¼ bÞ is called the branch cut or line. The conditional Re z z0 ¼
0 is satisfied on this branch cut which is shown by thethick line PQ in Fig. 1.3.
Now, we shall examine whether the conditional Re z z0 0 is satisfied or
not in the whole complex z-plane. We introduce the polar coordinates ðr; hÞ for the
complex variable as
z z0 ¼ r expðþihÞ ð1:3:9Þ
and measure its argument h from the horizontal line PC as shown in Fig. 1.3. Then
the root function is expressed in the polar form as
pffiffiffiffiffiffiffiffiffiffiffiffi pffiffi
z z0 ¼ r expðþih=2Þ ð1:3:10Þ
If the argument h is defined within p h þ p, the argument of the root function

is in the range, p=2 h=2 þ p=2, and thus the real part of the root function is
positive in the whole complex z-plane. Based on this argument definition, we can
specify the argument of the root function on the line along the branch cut. The
argument of the root function on the upper horizontal
pffiffiffiffiffiffiffiffiffiffiffiffiline AB in Fig. 1.3 is
measured from the horizontal line PC and is arg z z0 ¼ þp=2. On the lower
y = Im( z )
0 < arg z − z0 < π / 2

arg ( )
z − z0 = +π / 2
z − z0 = re + iθ
θ
B y=β θ = +π A
θ =0
C
Q A' P
B' θ = −π
x = Re( z )
x =α
arg ( )
z − z0 = −π / 2
0 > arg z − z0 > −π / 2
Fig. 1.3 Definition of the arguments of the variable z z0 and its square root function z z0
line A0 B0 , the argument is arg z z0 ¼ p=2. Thus, the argument of the root
function is in the range
p=2 argð z z0 Þ þ p=2 ð1:3:11Þ
Therefore, the real part of the root function is positive in the whole complex
z-plane.
In the above discussion, the argument of the variable z z0 is restricted in the
range p h þ p. However, if we take the positive argument only, i.e. 0 h, we
have to introduce two Riemann sheets. The two sheets are connected along the
branch cut as shown in Fig. 1.4 and the argument of the variable z z0 is measured
from the horizontal line as that in Fig. 1.3, but, when the argument exceeds þp
(cross the cut), we measure the argument in the lower sheet. After rounding one
anti-clockwise rotation in the lower sheet, the argument is then measured in the
upper sheet. The variation of the argument is shown by the solid and dotted thin
curves in Fig. 1.4, schematically. Due to this definition, the argument of the variable
is argðz z0 Þ ¼ þp on the upper line AB, and argðz z0 Þ ¼ þ3p on the lower line
A0 B0 . Thus, the argument of the square root function is positive in the upper
Riemann sheet, but, unfortunately, it is not continuous across the horizontal line PC
in Fig. 1.3, i.e. h ¼ 0 and h ¼ þ4p in Fig. 1.4. Due to this discontinuous nature of
the argument, the former definition is preferable for the argument of the root
function as defined in Fig. 1.3.
upper Riemann sheet
branch point
θ = +π
branch cut/line θ = +2π
θ = 0 or 4π
θ = +3π z = z0 θ = 0
θ = +4π
lower Riemann sheet
Fig. 1.4 Two Riemann sheets and argument variation of the variable z z0 (upper sheet: solid
curve, lower sheet: dotted curve)
qffiffiffiffiffiffiffiffiffiffiffiffiffiffi
1.3.2 Square Root Function 2: gðzÞ ¼ z2 z20
This square rootpfunction appears frequently in many dynamic problems. The

ffiffiffiffiffiffiffiffiffiffiffiffiffiffi
conditional, Re z z0 0 is also employed in order to guarantee the con-
2 2
vergence at infinity. We shall discuss the introduction of the branch cut under this
conditional.
(1) Branch cut
The branch cut and the argument of a little bit complicated square root function,
pffiffiffiffiffiffiffiffiffiffiffiffiffiffi
gðzÞ ¼ z2 z20 , is discussed. The conditional imposed on the function is similar
to that in the previous subsection, i.e. Re z2 z20 0. It should be understood
that the variable is z ¼ x þ iy and the complex constant is z0 ¼ a þ ib where a and
b are positive. It is easily found that two branch points of this root function are at
z ¼ ða þ ibÞ. In order to discuss the branch cut, we introduce the capital X and
Y for the real and imaginary parts of the function and rewrite the root function in the
explicit form,
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
X þ iY ¼ ðx þ iyÞ2 ða þ ibÞ2 ð1:3:12Þ
Squaring the both sides, and equating the real and imaginary parts, we have the
simultaneous equations for the unknowns, X and Y,

X 2 Y 2 ¼ x2 y2 ða2 b2 Þ
ð1:3:13Þ
XY ¼ xy ab
Since the conditional is X ¼ Re z2 z20 0, its border is X ¼ 0. Substituting
X ¼ 0 into the second of Eq. (1.3.13), we have the simple equation for the
hyperbola in the complex z-plane,
ab
y¼ ð1:3:14Þ
x
This is the border curve for the conditional, but its range must be specified. So, we
substitute X ¼ 0 into the first of Eq. (1.3.13),
Y 2 ¼ x2 y2 ða2 b2 Þ ð1:3:15Þ
As the left side of the above equation is negative, the right side also must be
negative,
x2 y2 ða2 b2 Þ 0 ð1:3:16Þ
Substituting the Eq. (1.3.14) into the above equation, we have the quadratic
inequality for x2 ,
x4 ða2 b2 Þx2 a2 b2 ¼ ðx2 a2 Þðx2 þ b2 Þ 0 ð1:3:17Þ
Solving for x2 , we have
b2 x2 þ a2 ð1:3:18Þ
Then, the admissible range for the variable x in the complex z-plane is
0 jxj a ð1:3:19Þ
The combination of two Eqs. (1.3.14) and

p (1.3.19) determines the branch cut which
satisfies the conditional Re z z0 ¼ 0. This shows two semi-infinite hyper-
2 2
bolic curves and their edges are corresponding to the branch points as shown in
Fig. 1.5.
root function y = Im( z )

z 2 − z 02
z0 = α + i β
branch point: z = z0
αβ
branch cut/line y=
x
β
−α
α
x = Re( z )
−β
αβ
y=
x
branch point: z = − z0 branch cut/line
Fig. 1.5 Branch cut/line for the square root function z2 z20 with z0 ¼ a þ ib
When the real or imaginary part of the constant z0 vanishes, the branch cut
becomes a simple line. In the case of the pure real constant, z0 ¼ a, the branch cut
for the root function z2 a2 is the line along the real and imaginary axes as
shown in Fig. 1.6a, where the cut along the real axis is slightly shifted from the axis
for the visual explanation. On the other hand, in the case of the pure imaginary
pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
constant z0 ¼ þib, the branch cut for the root function z2 þ b2 is the semi-infinite
straight line along the imaginary axis as shown in Fig. 1.6b.
It is easily anticipated that the branch cut/curve depends on the sign of the
imaginary part of the complex constant z0 . The former discussion is carried out for
the positive imaginary part of the constant. So, we shall reconsider the branch cut in
the case of the negative imaginary part of the constant, i.e. z0 ¼ a ib. Employing
the same conditional and procedure as those in the former, the branch points for the
root function z2 z20 are z ¼ ða ibÞ and the branch cuts are given by
ab
y¼ ; 0 jxj a ð1:3:20Þ
x
This is the semi-hyperbolic curve shown in Fig. 1.7. Two cuts are introduced in the
second and the fourth quadrants in the complex z-plane and the edge of the cut is
corresponding to each branch point, z ¼ ða ibÞ. When the imaginary part of the
constant vanishes, z0 ¼ a; b ¼ 0, the branch cut for the root function z2 a2
(a) y = Im( z )
root function: z2 − α 2
branch cut
x = −α x = +α
x = Re( z )
branch cut
(b) y = Im( z )
root function: z2 + β 2 branch cut
y = +β
x = Re( z )
y = −β
branch cut
pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
Fig. 1.6 a Branch cut for the root function z2 a2 . b Branch cut for the root function z2 þ b2
becomes the line on the real and imaginary axes as shown in Fig. 1.8a. On the other
hand, when the real part vanishes, z0 ¼ ib; a ¼ 0, the branch cut for the root
function z2 þ b2 becomes the two semi-infinite lines on the imaginary axis as
shown in Fig. 1.8b.
Comparing Fig. 1.8a with Fig. 1.6a, we learn that the branch cut on the real axis
is different in spite of the same form of the root function z2 a2 . This difference
y = Im( z ) root function

branch cut z 2 − z02
z0 = α − i β
branch point: z = − z0 αβ
y=−
x
β
α
−α
−β x = Re( z )
αβ
y=−
x branch point: z = z0
branch cut/line
Fig. 1.7 Branch cut/line for the square root function z2 z20 with z0 ¼ a ib
comes from the approaching direction of the zero imaginary part of the constant z0 .
When the imaginary part approaches to zero from above, the branch cut for the root
function z2 a2 is given by the lines in Fig. 1.6a. But, when the imaginary part
approaches to zero from below, the branch cut is given by the lines in Fig. 1.8a. In
the practical applications, the real constant a expresses the frequency x of the time-
harmonic vibration and the root function is given as z2 x2 . Therefore, we are
puzzled which is the suitable branch cut. This puzzle is solved by the sign of the
exponential time factor expðixtÞ and physical conditions. However the answer
and verification are given in the later chapters, a brief conclusion is stated here. The
branch cut in Fig. 1.6a is employed when the time factor is negative, i.e. expðixtÞ,
and the cut in Fig. 1.8a is employed when the time factor is positive, expðþixtÞ.
When the real part of the complex constant vanishes, a ! 0, the branch cut lays
on the imaginary axis as shown in Figs. 1.6b and 1.8b. These two figures are same.
Thus, the branch cut for the square root function z2 þ b2 is unchanged regardless
of the sign of the imaginary part b.
(2) Argument
We have just determined
p the branch cut for the square root function under the
conditional Re z z0 ¼ 0. In this subsection, we examine whether the real
2 2
part of the root function is positive or not in the whole complex plane. In order to
(a) y = Im( z )
branch cut
root function: z2 − α 2
x = +α
x = −α x = Re( z )
branch cut
(b) y = Im( z )
root function: z2 + β 2 branch cut
y = +β
x = Re( z )
y = −β
branch cut
pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
Fig. 1.8 a Branch cut for the root function z2 a2 . b Branch cut for the root function z2 þ b2
discuss the argument of the root function, we factorize the function as the product
of two simple root functions as
pffiffiffiffiffiffiffiffiffiffiffiffipffiffiffiffiffiffiffiffiffiffiffiffi
z2 z20 ¼ z z0 z þ z0 ð1:3:21Þ
and introduce the polar coordinate ðrj ; hj Þ; j ¼ 1; 2 for each variable as
z z0 ¼ r1 expðþih1 Þ; z þ z0 ¼ r2 expðþih2 Þ ð1:3:22Þ
where rj is the radial arm/distance from each branch point, and the arguments hj is
measured from the horizontal line parallel to the positive real axis. Then, the root
function is rewritten in the polar form,
pffiffiffiffiffiffiffiffi
z2 z20 ¼ r1 r2 expfþiðh1 þ h2 Þ=2g ð1:3:23Þ
and the discussion for the argument of the square root function is carried out on its
polar angle ðh1 þ h2 Þ=2. We assume that the real and imaginary parts of the con-
stant a and b are positive. Therefore, the branch points and cuts are in the first and
third quadrants in the complex z-plane as shown in Fig. 1.5. The argument of the
root function is discussed on three paths: (1) path along the branch cut, (2) path on
the non-branch cut hyperbola, and (3) the line which connects two branch points.
We shall discuss their arguments, separately.
(2:1) Right side of the upper branch cut
Firstly, we consider the argument on the right side of the upper branch cut in
Fig. 1.9a, whose branch cuts are the same as those in Fig. 1.5. The mark “P1 ”
denotes an arbitrary point on the upper branch cut and its coordinate is ðxP ; yP Þ in
the complex plane. The arguments h1 of the radial arm r1 ¼ O1 P1 is measured from
the horizontal line O1 A1 . Since the branch point is at ða; bÞ, the complementary
angle /1 of the radial arm r1 is introduced as
yP b
tan /1 ¼ ð1:3:24Þ
a xP
and thus the argument h1 is expressed by
h1 ¼ p /1 ð1:3:25Þ
On the other hand, the argument h2 of the radial arm r2 ¼ O2 P1 is measured from
the horizontal line O2 A2 . Since the lower branch point is at ða; bÞ, the argument
of the radial arm, i.e. h2 , is given by
b þ yP
tan h2 ¼ ð1:3:26Þ
a þ xP
(a) y = Im( z )
P1 ( xP , y P )
θ1
r1
θ1
φ1 (α , β )
B1 A1
r2 O1
θ2
x = Re( z )
O2
B2 A2
(−α , − β )
(b) y = Im( z )
(α , β )
B1 A1
O1 θ1
r1 Q1
r2 ( xQ , yQ ) x = Re( z )
O2 θ2
B2 A2
(−α , − β )
(c) y = Im( z )
(α , β )
B1 ψ A1
r1 O1
R θ1 = −(π −ψ )
r2
x = Re( z )
O2 θ2 = ψ
B2 r1
(−α , − β )
Fig. 1.9 Arguments, h1 and h2 , a along the upper branch cut, b on the upper non-branch cut
hyperbola, c on the line connecting two branch points, d along the lower branch cut and e on
the lower non-branch cut hyperbola
(d) y = Im( z )
(α , β )
B1 A1
ψ1 O1
r1 x = Re( z )
θ1
O2
B2 A2
(−α , − β ) φ2
θ2
r2
θ2
( − xP , − y P )
P2
(e) y = Im( z )
(α , β )
B1 A1
ψ1
O1
r1 θ1
Q2
θ2 x = Re( z )
(− xQ , − yQ ) r2 O2
ψ2
B2 A2
(−α , − β )
Fig. 1.9 (continued)
Two Eqs. (1.3.24) and (1.3.26) are looks different, but, the point P1 is on the
hyperbolic curve at the limit. The hyperbolic curve gives the relation between xP
and yP ,
ab
yP ¼ ð1:3:27Þ
xP
Substituting the above relation into the two Eqs. (1.3.24) and (1.3.26),
b
tan h2 ¼ tan /1 ¼ ð1:3:28Þ
xP
we find that h2 ¼ /1 . The argument of the root function at the right side of the
upper branch cut is thus given by
ðh1 þ h2 Þ=2 ¼ ðp /1 þ /1 Þ=2 ¼ þp=2 ð1:3:29Þ

This shows that the value of the square root function z2 z20 at the right side of
the upper branch cut is positive imaginary.
(2:2) Left side of the upper branch cut
We now consider the argument of the root function at the left side of the upper
branch cut at P1 . The argument h1 must be measured in the clockwise direction
from the horizontal line O1 A1 , since the arm r1 cannot cross its own branch cut.
However, the argument h2 can cross the cut, since the cut is not for the lower branch
point. Thus, from the geometry in Fig. 1.9a, we have
h1 ¼ ðp þ /1 Þ; h2 ¼ / 1 ð1:3:30Þ
and the argument of the root function is
ðh1 þ h2 Þ=2 ¼ p=2 ð1:3:31Þ
Then, the value of the root function at the left side of the upper branch cut is
negative imaginary.
(2:3) On the upper non-branch cut hyperbola
One arbitrary point Q1 is taken on the non-branch cut hyperbola as shown in
Fig. 1.9b and we assume that its coordinate is ðxQ ; yQ Þ in the complex plane. The
arm r1 from the upper branch point is denoted by the line O1 Q1 and the arm r2 from
the lower branch point is O2 Q1 . They are shown in Fig. 1.9b. The argument of the
arm r1 is measured from the horizontal line O1 A1 in the clockwise direction and
that of the arm r2 is measured from the horizontal line O2 A2 in the anti-clockwise
direction. A simple analytical geometry shows that the arguments h1 and h2 are
b yQ b þ yQ
tan h1 ¼ ; tan h2 ¼ ð1:3:32Þ
xQ a a þ xQ
Substituting the equation for the hyperbola yQ ¼ ab=xQ into the above equations,
we have
b b
tan h1 ¼ ; tan h2 ¼ ð1:3:33Þ
xQ xQ
Thus, h2 ¼ h1 , and the argument of the root function on the non-branch cut
hyperbola is
ðh1 þ h2 Þ=2 ¼ 0 ð1:3:34Þ

This shows the value of the root function z2 z20 is positive real.
(2:4) On the line connecting two branch points
We take an arbitrary point R on the line O1 O2 , which connects two branch
points, as shown in Fig. 1.9c. Since the argument of each arm from the branch point
is unchanged on this line, it is no need to specify the location of the point R.
Introducing the complementary angle w for the acute angle \B1 O1 O2 , the argument
h2 for the lower arm r2 is expressed by
h2 ¼ w ð1:3:35Þ
and that for the upper arm r1 is
h1 ¼ ðp wÞ ð1:3:36Þ
Then, the argument of the root function is given by
ðh1 þ h2 Þ=2 ¼ ðp=2 wÞ ð1:3:37Þ
This shows that the argument of the root function is negative on the connecting line,
since 0 \ w \ p=2. Thus, the real part of the square root function is positive.
(2:5) Right side of the lower branch cut
We take an arbitrary point P2 on the lower branch cut as shown in Fig. 1.9d, and
consider the argument of the root function on the right side of the cut. Assuming the
location of the point P2 as ðxP ; yP Þ, the complementary angle w1 ð¼\B1 O1 P2 Þ is
given by
b þ yP
tan w1 ¼ ð1:3:38Þ
a þ xP
Then, the argument of the radial arm r1 is measured in the clockwise direction,
h1 ¼ ðp w1 Þ ð1:3:39Þ
The argument h2 of the arm r2 ð¼ O2 P2 Þ is also measured in the clockwise direction.

It is given by
yP b
tan h2 ¼ ð1:3:40Þ
a xP
Substituting the equation for the hyperbola, yP ¼ ab=xP , into Eqs. (1.3.38) and
(1.3.40), we have
b b
tan w1 ¼ ; tan h2 ¼ ð1:3:41Þ
xP xP
Then,
h2 ¼ w1 ð1:3:42Þ
Two Eqs. (1.3.39) and (1.3.42) give the argument of the root function as
ðh1 þ h2 Þ=2 ¼ p=2 ð1:3:43Þ
This shows that the value of the root function on the right side of the lower branch
cut is negative imaginary.
(2:6) Left side of the lower branch cut
When we consider the argument on the left side of the lower branch cut, an
attention must be paid to the argument h2 since the arm r2 from the lower branch
point cannot cross its own branch cut. As shown in Fig. 1.9d, the argument h2 is
measured from the horizontal line O2 A2 in the counterclockwise direction. Intro-
ducing the complementary angle /2 as shown in Fig. 1.9d, the argument h2 is given
by
h2 ¼ 2p /2 ð1:3:44Þ
On the other hand, the arm r1 can cross the lower branch cut and its argument is
h1 ¼ ðp w1 Þ ð1:3:45Þ
The two complementary angles are given by
b þ yP yP b
tan w1 ¼ ; tan /2 ¼ ð1:3:46Þ
a þ xP a xP
Substituting the equation for the hyperbola, yP ¼ ab=xP , into the above, we find
that w1 ¼ /2 . Thus, the argument of the root function is
ðh1 þ h2 Þ=2 ¼ þp=2 ð1:3:47Þ
This shows that the value of the root function on the left side of the lower branch
cut is positive imaginary.
(2:7) On the lower non-branch cut hyperbola

We consider the argument on the lower non-branch cut hyperbola and take an
arbitrary point Q2 on it, as shown in Fig. 1.9e. Assuming the point Q2 is at
ðxQ ; yQ Þ, we introduce two complementary angles w1 ð¼\B1 O1 Q2 Þ and w2
ð¼\B2 O2 Q2 Þ. They are
b þ yQ b yQ
tan w1 ¼ ; tan w2 ¼ ð1:3:48Þ
a þ xQ xQ a
Since the point Q2 is on the hyperbola, we substitute yQ ¼ ab=xQ and then find that
w1 ¼ w2 . The two arguments are
h1 ¼ ðp w1 Þ; h2 ¼ p w2 ð1:3:49Þ
Thus, the argument of the root function is
ðh1 þ h2 Þ=2 ¼ 0 ð1:3:50Þ
This shows the value of the root function on the non-branch cut hyperbola is
positive real.
square root function y = Im( z )

g ( z ) = z − (α + i β )
2 2
0 < arg( g ) < +π / 2
arg( g ) = −π / 2 arg( g ) = +π / 2
O1
arg( g ) = 0
−π / 2 < arg( g ) < 0
arg( g ) = 0 O2 x = Re( z )
arg( g ) = +π / 2
arg( g ) = −π / 2
0 < arg( g ) < +π / 2
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
Fig. 1.10 Argument of the square root function z2 ða þ ibÞ2 (The imaginary part of the
constant is positive)

g ( z) = z − α
2 2
arg( g ) = +π / 2
arg( g ) = −π / 2
0 < arg( g ) < +π / 2
−π / 2 < arg( g ) < 0

+α
arg( g ) = 0
arg( g ) = 0
x = Re( z )
−α
0 < arg( g ) < +π / 2 −π / 2 < arg( g ) < 0
arg( g ) = −π / 2
arg( g ) = +π / 2
Fig. 1.11 Argument of the square root function z2 a2 (The positive imaginary part of the
complex constant vanishes)
Summarizing the above discussions in (2.1)–(2.7), the argument of the square

root function in the whole complex plane is shown schematically in Fig. 1.10.
However the argument in the non-specific point/region was not discussed, we can
easily guess the argument from our result. It is shown in the square box in the
figure. When the imaginary part of the complex constant vanishes, b ! 0, Fig. 1.10
is reduced to Fig. 1.11. On the other hand, when the real part of the complex
constant vanishes, a ! 0, Fig. 1.10 is reduced to Fig. 1.12. In any case, the
argument of the square root function in the whole complex z-plane is in the range
p=2 ðh1 þ h2 Þ=2 þ p=2 ð1:3:51Þ

This shows that the conditional Re z2 z20 0 is satisfied in the whole com-
plex plane and thus we have introduced the branch cuts and defined the argument of
the square root function, correctly.
If we employ the complex constant with the negative imaginary part,
z0 ¼ a ib, the branch cuts are in the second and the fourth quadrants in the
complex z-plane as shown in Fig. 1.7. The discussion in the above subsections can
be applied to the argument of the root function. Although the detailed discussions
for the argument are not repeated here, the final results are schematically shown in
Figs. 1.13, 1.14 and 1.15.

g ( z) = z + β
2 2
arg( g ) = +π / 2
arg( g ) = −π / 2
0 < arg( g ) < +π / 2
−π / 2 < arg( g ) < 0 +iβ
arg( g ) = 0 arg( g ) = 0
arg( g ) = 0
x = Re( z )
−iβ
0 < arg( g ) < +π / 2 −π / 2 < arg( g ) < 0
arg( g ) = −π / 2
arg( g ) = +π / 2
Fig. 1.12 Argument of the square root function z2 þ b2 in the complex plane (The real part of
the complex constant vanishes)
One comment should be repeated here. Comparing Fig. 1.11 with Fig. 1.14 in the
case of vanishing imaginary part of the complex constant, the square root function
z2 a2 on the real axis x ¼ ReðzÞ is the positive or negative purely imaginary as
pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi þipffiffiffiffiffiffiffiffiffiffiffiffiffiffi
a 2 x2
ffi
x a ¼
2 2 p ffiffiffiffiffiffiffiffiffiffiffiffiffiffi
ffi ; 0 jxj \ a ð1:3:52Þ
i a2 x2
The sign depends on the vanishing direction of the imaginary part of the complex
constant z0 . Thus, as was mentioned before, we have to pay special attention to the
square root function with the real constant z2 a2 .
1.4 Comments on Inversion Techniques and Integration

Formulas
The integral transform technique is a powerful tool for solving linear differential
equations. However, its success is up to the evaluation of the inversion integral. So
far, many integration formulas have been found and published. The most com-
prehensive formula books are Erdélyi (1954), Gradshteyn and Ryzhik (1980),
1.4 Comments on Inversion Techniques and Integration Formulas 29
y = Im( z ) square root function

arg( g ) = −π / 2
g ( z ) = z 2 − (α − i β ) 2
−π / 2 < arg( g ) < 0 arg( g ) = +π / 2
O1
arg( g ) = 0
x = Re( z )
0 < arg( g ) < +π / 2
arg( g ) = 0
O2
−π / 2 < arg( g ) < 0
arg( g ) = +π / 2
arg( g ) = −π / 2
Fig. 1.13 Argument of the square root function z2 ða ibÞ2 (The imaginary part of the
complex constant is negative)
Magnus et al. (1966), Watson (1966), Moriguchi et al. (1972) and Titchmarsh
(1948).
The books from Erdélyi (1954) to Watson (1966) are well-known and it is not
difficult to obtain access to them. The book Watson (1966) deals solely with Bessel
functions and not with integral transforms, but it gives many integration formulas
for the Hankel transform. The book Moriguchi et al. (1972), written in Japanese, is
very compact and is separated into three small handy books. In spite of its com-
pactness, the principal formulas which are included in Erdélyi (1954), Gradshteyn
and Ryzhik (1980), Magnus et al. (1966) and Watson (1966) are cited. The author
believes that the three handy books are most convenient as a “first aid.” The last
book Titchmarsh (1948) describes the mathematics of the theory of Fourier trans-
form. When someone needs more detailed mathematics for the integral transform,
this book will give proper answers.
If a desired formula cannot be found in these books, the complex integral is
employed to evaluate the inversion integral. The complex integral based on Cau-
chy’s integral theorem is the most useful evaluation technique and the discussion on

g ( z) = z − α
2 2
arg( g ) = +π / 2
arg( g ) = −π / 2
0 < arg( g ) < +π / 2
−π / 2 < arg( g ) < 0

+α
arg( g ) = 0 x = Re( z )
arg( g ) = 0
−α
−π / 2 < arg( g ) < 0

0 < arg( g ) < +π / 2
arg( g ) = −π / 2
arg( g ) = +π / 2
Fig. 1.14 Argument of the square root function z2 a2 in the complex plane (The negative
imaginary part of the complex constant vanishes)
the branch cut in the previous subsections will be helpful. If the complex integral
does not give any compact result, the inversion integral is left in its definition form
or is converted to the numerically tractable form by the complex integral.
Exercises
(1:1) Apply the finite Fourier transform, complex Fourier series defined by
Eq. (1.1.6), to Dirac’s delta function which is defined in a finite region
ðp; þpÞ as
dðx aÞ ; p \ a \ þ p ðaÞ
and then show the series form of the delta function,

" #
1 X1
dðx aÞ ¼ 1þ2 cosfnðx aÞg ðbÞ
2p n¼1
1.4 Comments on Inversion Techniques and Integration Formulas 31

g ( z) = z + β2 2
arg( g ) = +π / 2
arg( g ) = −π / 2
0 < arg( g ) < +π / 2
−π / 2 < arg( g ) < 0 +β
arg( g ) = 0
arg( g ) = 0
arg( g ) = 0
x = Re( z )
−β
0 < arg( g ) < +π / 2 −π / 2 < arg( g ) < 0
arg( g ) = −π / 2
arg( g ) = +π / 2
Fig. 1.15 Argument of the square root function z2 þ b2 in the complex plane (The real part of
the complex constant vanishes)
Fig. 1.16 Two branch cuts

pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi Im( z )
for the root function z2 þ a2 r1
+ia θ1
Re( z )
r2
θ2
−ia
(1:2) If we expand the x-range to ð1; þ1Þ, the Fourier series in the above
equation (b) gives an infinite sequence of delta function, i.e.
" #
1 X1 X
þ1
1þ2 cosfnðx aÞg ¼ dðx a 2mpÞ ðcÞ
2p n¼1 m¼1
Explain why the infinite sequence?

(1:3) For the root function z2 þ a2 , where a is a positive real constant, if we
introduce two branch cuts as shown in Fig. 1.16, how is the argument of the
root function along the branch cut?
References
Erdélyi A (ed) (1954) Tables of integral transforms, vol I and II. McGraw-Hill, New-York
Ewing WM, Jardetzky WS, Press F (1957) Elastic waves in layered media. McGraw-Hill,
New York
Gradshteyn IS, Ryzhik IM (Jefferey A, ed) (1980) Table of integrals, series, and products, 5th edn.
Academic Press, San Diego
Magnus W, Oberhettinger F, Soni RP (1966) Formulas and theorems for the special functions of
mathematical physics. Springer, New York
Moriguchi S, Udagawa K, Ichimatsu S (1972) Mathematical formulas, vols I–III (in Japanese).
Iwanami, Tokyo
Sneddon IN (1951) Fourier transforms. McGraw-Hill, New York
Titchmarsh EC (1948) Introduction to the theory of Fourier integrals, 2nd edn. Oxford, London
Watson GN (1966) A treatise on the theory of Bessel functions. Cambridge University Press,
Cambridge
Chapter 2
Green’s Functions for Laplace
and Wave Equations
This chapter shows the solution method for Green’s functions of 1, 2 and 3D
Laplace and wave equations. Lengthy and detailed explanations are given in order
to instruct the basic technique of the integral transform. Especially, Fourier
inversion integral for the time-harmonic Green’s function is discussed in detail, and
three evaluation techniques are introduced in Sect. 2.5.
2.1 1D Impulsive Source
We start from the simplest wave equation that has two variables: a single space
variable x and the time t,
@2/ 1 @2/
¼ PdðxÞdðtÞ ð2:1:1Þ
@x2 c2 @t2
The nonhomogeneous term represents a wave source with magnitude P. The two
Dirac’s delta functions, dðxÞ and dðtÞ, show the location and the impulsive nature of
the source. The Green’s function is a particular solution of the differential equation
corresponding to the impulsive source. The Green’s function is sought under the
quiescent condition at an initial time,

@/
/jt¼0 ¼ ¼ 0 ð2:1:2Þ
@t t¼0
and the convergence condition at infinity,

@/
/jx!1 ¼ ¼ 0 ð2:1:3Þ
@x x!1
To obtain the particular solution of the wave equation (2.1.1), we apply the
integral transforms and reduce the differential equation to an algebraic equation in
the transformed domain. Since the unknown function /ðx; tÞ has two variables, we

DOI 10.1007/978-3-319-17455-6_2
34 2 Green’s Functions for Laplace and Wave Equations
apply the double transform: Laplace transform with respect to the time variable t:
0 t\ þ 1,
Zþ1

f ðsÞ ¼ L½f ðtÞ ¼ f ðtÞ expðstÞdt ð2:1:4Þ
0
and Fourier transform with respect to the space variable x: 1 \ x \ þ 1,
Zþ1 Zþ1
f ðxÞ expðþinxÞdx; f ðxÞ ¼ ð2:1:5Þ
2p
1 1
Firstly, we multiply the kernel of the Laplace transform expðstÞ to both sides of
the differential equation (2.1.1),
Z1 2
@ / 1 @2/
¼ PdðxÞdðtÞ expðstÞdt ð2:1:6Þ
@x2 c2 @t2
0
and perform the Laplace transform integral term by term. The order of integration
and differentiation are interchanged for the first term in the left hand side of the
equation. The first term in the right hand side is left in its order and then integrated
by parts. The last nonhomogeneous term, which has the delta function, is evaluated
by using the formula (1.2.3). Then, Eq. (2.1.6) is rewritten as
Z1 Z1 Z1
d2 1 @2/
/ expðstÞdt ¼ 2 expðstÞdt PdðxÞ dðtÞ expðstÞdt ð2:1:7Þ
dx2 c @t2
0 0 0
We define the Laplace transform of the unknown function as
Z1

/ ¼ / expðstÞdt ð2:1:8Þ
0
The time-derivative term is integrated by parts,
Z1 t!1 Z1
@2/ @/
expðstÞdt ¼ expðstÞ þ ½ s/ expðstÞ t!1
þ s 2
/ expðstÞdt ¼ s2 /
@t2 @t t¼0
t¼0
0 0
ð2:1:9Þ
2.1 1D Impulsive Source 35
where the quiescent condition at the initial time is incorporated. To the last non-
homogeneous term, the simple integration formula for the delta function,
Z1
dðtÞ expðstÞdt ¼ 1 ð2:1:10Þ
0
is applied. Then, we have the ordinary differential equation for the function / in
the transformed domain,
d 2 /
¼ ðs=cÞ2 / PdðxÞ ð2:1:11Þ
dx2
It is possible to obtain the exact solution for the above ordinary differential equation
by the elementary method. However, in order to demonstrate the integral transform
technique, we further apply Fourier transform to the ordinary differential equa-
tion (2.1.11). The Fourier transform defined by Eq. (2.1.5) is applied to
Eq. (2.1.11),
Zþ1 2
d / 2
¼ ðs=cÞ / PdðxÞ expðþinxÞdx ð2:1:12Þ
dx2
1
The Fourier transform integral is applied to each term as
Zþ1 Zþ1 Zþ1

d 2 /
expðþinxÞdx ¼ ðs=cÞ2 / expðþinxÞdx P dðxÞ expðþinxÞdx
dx2
1 1 1
ð2:1:13Þ
The convergence condition is also transformed, as
Z1
@/ d/
/jx!1 ¼ ¼ 0 expðstÞdt ) / jx!1 ¼ ¼ 0
@x x!1 dx x!1
0
ð2:1:14Þ
Defining the Fourier transform of the Laplace transformed unknown function,
Zþ1
¼
/ / expðþinxÞdx ð2:1:15Þ
1
the transform of the space-derivative in Eq. (2.1.13) is carried out with the aid of the
convergence condition,
Zþ1 x!þ1 Zþ1

d 2 / d/
expðþinxÞdx ¼ expðþinxÞ in½ / expðþinxÞ x!þ1
x!1 n2
/ expðþinxÞdx
dx2 dx x!1
1 1
~
¼ n2 /
ð2:1:16Þ
The integration formula for the delta function
Zþ1
dðxÞ expðþinxÞdx ¼ 1 ð2:1:17Þ
1
is also used for evaluating the last term. The Fourier transform of the ordinary
differential equation (2.1.12) becomes a simple algebraic equation for the double
transformed unknown function / ,
¼ s2
n2 / / P ð2:1:18Þ
c2
,
Then we have the exact expression for the double transformed function /
¼ P
/ ð2:1:19Þ
n þ ðs=cÞ2
2
The unknown function has just been determined explicitly in the transformed
domain. We shall carry out two inverse transforms successively. As the first
inversion, we apply the Fourier inversion integral which is defined by the second of
Eq. (2.1.5). The formal Fourier inversion is given by the integral
Zþ1
1 P
/ ¼ expðinxÞdn ð2:1:20Þ
2p n þ ðs=cÞ2
2
1
Due to the symmetric nature of the integrand, the integral is reduced to the simpler
semi-infinite integral,
Z1
P 1
/ ¼ cosðnxÞdn ð2:1:21Þ
p n þ ðs=cÞ2
2
0
This is a simple integral and the integration formula (Erdélyi 1954, vol. I, pp. 8, 11),
Zþ1
1 p
cosðxyÞdx ¼ expðajyjÞ ð2:1:22Þ
x 2 þ a2 2a
0
can be applied. Then, Eq. (2.1.21) gives
P
/ ¼ expðsjxj=cÞ ð2:1:23Þ
2ðs=cÞ
The next step is to carry out the Laplace inversion. The symbolical form of the
Laplace inversion is given by

cP 1 1
/ ¼ L1 ½/ ¼ L expðsjxj=cÞ ð2:1:24Þ
2 s
Fortunately, we have the Laplace inversion formula (Erdélyi 1954, vol. I, pp. 241, 1),

1 0; t\a
L1 expðasÞ ¼ ¼ Hðt aÞ ð2:1:25Þ
s 1; t[a
where Hð:Þ is Heaviside’s unit step function. Applying this formula to Eq. (2.1.24),
the solution for the non-homogeneous 1D wave equation is obtained as
cP cP
/¼ Hðt jxj=cÞ ¼ Hðct jxjÞ ð2:1:26Þ
2 2
This solution shows an expanding (or out-going) 1D wave with the uniform
amplitude cP/2 and with the velocity c as shown in Fig. 2.1. Consequently, we get
the Green’s function for the 1D wave equation,
@2/ 1 @2/ cP
¼ PdðxÞdðtÞ ) /ðx; tÞ ¼ Hðct jxjÞ ð2:1:27Þ
@x2 c2 @t2 2
Fig. 2.1 1D expanding wave

from a source point
cP / 2
ct
2.2 1D Time-Harmonic Source
When the source is time-harmonic, the nonhomogeneous term in Eq. (2.1.1) is

replaced with a harmonic function, but the source location is unchanged. Thus, the
wave equation with a time-harmonic source is given by
@2/ 1 @2/
¼ QdðxÞ expðþixtÞ ð2:2:1Þ
@x2 c2 @t2
where Q is the source magnitude and x the frequency of the time-harmonic

vibration. We assume that its solution satisfies the convergence condition at infinity,
i.e.

@/
/jx!1 ¼ ¼ 0 ð2:2:2Þ
@x x!1
As the first step of the solution method, we assume that the solution is also time-
harmonic,
/ðx; tÞ ¼ /# ðxÞ expðþixtÞ ð2:2:3Þ
where /# is called the “amplitude function.” Due to this assumption, the conver-
gence condition (2.2.2) is rewritten for the amplitude function,

# d/#
/ x!1 ¼ ¼0 ð2:2:4Þ
dx x!1
Substituting the time-harmonic assumption of Eq. (2.2.3) into the wave equa-
tion (2.2.1), we have the ordinary differential equation for the amplitude function,
d 2 /#
þ ðx=cÞ2 /# ¼ QdðxÞ ð2:2:5Þ
dx2
The exact solution of this ordinary differential equation can also be obtained by the
elementary method. However, in order to demonstrate the integral transform
technique, we apply the Fourier transform, which is defined by Eq. (2.1.5), to the
ordinary differential equation (2.2.5),
Zþ1 2 #
d / 2 #
þ ðx=cÞ / ¼ QdðxÞ expðþinxÞdx ð2:2:6Þ
dx2
1
2.2 1D Time-Harmonic Source 39
Defining the Fourier transform of the amplitude function as

Zþ1
# ¼
/ /# expðþinxÞdx ð2:2:7Þ
1
the space derivative term in Eq. (2.2.6) is integrated by parts as

Zþ1 # x!þ1
d 2 /# d/ x!þ1
expðþinxÞdx ¼ expðþinxÞ in /# expðþinxÞ x!1
dx2 dx x!1
1
ð2:2:8Þ
Zþ1
n2 /# expðþinxÞdx
1
We apply the convergence condition of Eq. (2.2.4) and the definition of the Fourier
transform (2.2.7) to the above equation. The Fourier transform of the double
derivative is then reduced to
Zþ1
d 2 /# #
expðþinxÞdx ¼ n2 / ð2:2:9Þ
dx2
1
The nonhomogeneous term is evaluated as

Zþ1
Q dðxÞ expðþinxÞdx ¼ Q ð2:2:10Þ
1
Then, Eq. (2.2.6) gives a simple algebraic equation for the Fourier transformed
amplitude function,
# þ ðx=cÞ2 /
n2 / # ¼ Q ð2:2:11Þ
The Fourier transformed amplitude is determined completely,
# ¼ Q
/ ð2:2:12Þ
n ðx=cÞ2
2
Our next task is to invert the transformed amplitude. Applying the formal Fourier
inversion integral to Eq. (2.2.12), we get
Zþ1
# 1 P
/ ¼ expðinxÞdn ð2:2:13Þ
2p n ðx=cÞ2
2
1
Im(ξ )
+ω / c
? Re(ξ )
?
−ω / c
Fig. 2.2 Possible deformations of the integration path around the pole
Inspecting the integrand, we see that it has two simple poles at n ¼ ðx=cÞ, that is
to say, the poles are located on the integration path (real axis in the complex
n-plane). Since the integration cannot be performed through these singular points,
we have to distort the integration path around the poles. There are two ways of
deforming the path. One is through an upper semi-circle, another is through a lower
semi-circle as shown in Fig. 2.2. We have to determine which semi-circle is suit-
able. Discussing the nature of the initial wave equation (2.2.1), we learn that the
wave will expand to the outer region from the source point, i.e. wave radiation from
the source. Therefore, we have to choose the path so that the inversion integral
results in a radiating (out-going) wave from the source.
Still, it is somewhat complicated to explain the path selection. To aid the
understanding, two integrals with complex frequency are considered. After dis-
cussing the wave nature derived from each integral, we will determine and
understand the path distortion.
Let us introduce and add a small imaginary number e to the frequency in
Eq. (2.2.13) so that the poles are shifted from the real axis and are not on the
integration path. The complex frequency is considered in two ways, positive and
negative imaginary parts, x ! x ie. Employing the complex frequency -, we
consider the complex integral,
Z
1 P
U¼ expðinxÞdn ð2:2:14Þ
2p n ð-=cÞ2
2
C
where the integrand is the same as that in the Fourier inversion integral (2.2.13), but
the frequency is complex, - ¼ x ie. The integration loop C for the two cases of
complex frequency, with positive and negative imaginary parts, is discussed
separately.
(1) Small positive imaginary part: - ¼ x þ ie
When the frequency has a small positive imaginary part, the poles are shifted
from the real axis. The integration path C is chosen so that the integrand vanishes
on the large semi-circle with infinite radius. We employ the lower closed loop CðÞ
in the case of positive x and the upper loop CðþÞ in that of negative x as shown in
Fig. 2.3a.
Fig. 2.3 Integration loop for (a) Im(ξ )

the complex integral U in the
(+)
case of a the positive C
imaginary part of the
frequency, b the negative
imaginary part of the
frequency
x<0
(ω + iε ) / c
Re(ξ )
−(ω + iε ) / c
x>0
C (−)
Re(ξ )
(b)
(+)
C
x<0
−(ω − iε ) / c
Im(ξ )
(ω − iε ) / c
x>0
C (−)
Applying Cauchy’s integral theorem (Jordan’s lemma) to the complex integral U

with the loop C ðÞ in Fig. 2.3a, the integral along the real axis is evaluated as the
residue at the lower pole n ¼ -=c ¼ ðx þ ieÞ=c,
Zþ1 " #
1 P 2pi n þ -=c
expðinxÞdn ¼ P 2 expðinxÞ ð2:2:15Þ
2p n2 ð-=cÞ2 2p n ð-=cÞ2 n¼-=c
1
Rewriting the above equation, and taking the limit e ! 0, we have in the case of
positive x,
Zþ1
1 P iP
2
expðinxÞdn ¼ expðþixx=cÞ; x[0 ð2:2:16Þ
2p n ðx=cÞ
2 2ðx=cÞ
1
On the other hand, when x \ 0, we employ the upper loop C ðþÞ for the complex
integral U and have
Zþ1 " #
1 P 2pi n -=c
1
Then, we take the limit e ! 0,
Zþ1
1 P iP
2
expðinxÞdn ¼ expðixx=cÞ; x\0 ð2:2:18Þ
2p n ðx=cÞ
2 2ðx=cÞ
1
Unifying the two Eqs. (2.2.16) and (2.2.18), we have for the Fourier inversion
integral, where the positive imaginary part of the complex frequency tends to zero,
i.e.
Zþ1
1 P iP
2
expðinxÞdn ¼ expðþixjxj=cÞ; x þ ieje!0 ð2:2:19Þ
2p n ðx=cÞ
2 2ðx=cÞ
1
(2) Small negative imaginary part: - ¼ x ie

When the imaginary part of the complex frequency is negative, the poles are also
shifted from the real axis as shown in Fig. 2.3b. In order to guarantee the con-
vergence on the large semi-circle, the lower loop C ðÞ is employed in the case of
positive x, and the upper CðþÞ in that of negative x.
When x [ 0, we employ the loop C ðÞ and apply Jordan’s lemma to the complex
integral U in Eq. (2.2.14). The integral along the real axis is converted to the residue
at the lower pole, n ¼ -ð¼ x ieÞ=c,
Zþ1 " #
1 P 2pi n -=c
1
Rewriting the above and taking the limit e ! 0, we have in the case of positive x,
Zþ1
1 P iP
2
expðinxÞdn ¼ expðixx=cÞ; x [ 0 ð2:2:21Þ
2p n ðx=cÞ
2 2ðx=cÞ
1
Similarly, when we employ the upper loop C ðþÞ in the case of x\0,
Zþ1 " #
1 P 2pi n þ -=c
1
Taking the limit e ! 0, the Fourier inversion integral is evaluated as
Zþ1
1 P iP
2
expðinxÞdn ¼ expðþixx=cÞ; x\0 ð2:2:23Þ
2p n ðx=cÞ
2 2ðx=cÞ
1
Unifying two Eqs. (2.2.21) and (2.2.23), we have the unified expression when the
negative imaginary part of the frequency approached to zero,
Zþ1
1 P iP
2
expðinxÞdn ¼ expðixjxj=cÞ; x ieje!0 ð2:2:24Þ
2p n ðx=cÞ
2 2ðx=cÞ
1
(3) Selection of the imaginary part

Two expressions are obtained for the single Fourier inversion integral. They are
Eqs. (2.2.19) and (2.2.24) and are summarized as
8
< 2ðx=cÞ
iP
expðixjxj=cÞ; x ieje!0
/# ¼ ð2:2:25Þ
:þ iP
x þ ieje!0
2ðx=cÞ expðþixjxj=cÞ;
The wave nature of the two expressions is discussed by multiplying the time factor,
8
< 2ðx=cÞ
iP
expfþixðt jxj=cÞg; x ieje!0
/¼ ð2:2:26Þ
: þ iP expfþixðt þ jxj=cÞg; x þ iej
2ðx=cÞ e!0
Inspecting the argument of the exponential function in the above equation, the
upper solution shows an out-going (radiation) wave from the source, since the
argument of the exponential function is t jxj=c. On the other hand, the lower
includes t þ jxj=c and shows an incoming wave, coming from infinity. Since the
suitable solution must be the out-going wave, we employ the upper formula in
Eqs. (2.2.25) and (2.2.26). Thus, the complex frequency with the negative imagi-
nary part is the right assumption. Then, the suitable integration contour for the
Fourier inversion integral is that of Fig. 2.3b and the final result for the time-
harmonic response is given by
@2/ 1 @2/ iQ
¼ QdðxÞ expðþixtÞ ) /¼ expfþixðt jxj=cÞg
@x2 c2 @t2 2ðx=cÞ
ð2:2:27Þ
We have just learned that the right selection for the complex frequency is
- ¼ x ie, i.e. negative imaginary part, and the suitable integration loop is C ðÞ in
Fig. 2.3b. If we do not introduce the small imaginary part and keep the integration
path on the real axis, the poles is on the real axis and the integration path around the
pole must be deformed by a small semi-circle shown in Fig. 2.4a. This deformation
is valid only in the case of the positive time factor, expðþixtÞ. If we assume the
negative time factor expðixtÞ, the integration path on the real axis is deformed as
that shown in Fig. 2.4b. Thus, the selection of the deformed path around the pole
depends on the sign of the imaginary part of the complex frequency. Therefore, we
could answer to the initial question about the deformation of the integration path for
the Fourier inversion integral.
2.3 2D Static Source
Let us consider Green’s function for a typical partial differential equation, the so-
called Laplace equation. The Laplace equation with a source S is the nonhomo-
geneous differential equation,
@2/ @2/
þ ¼ SdðxÞdðyÞ ð2:3:1Þ
@x2 @y2
2.3 2D Static Source 45
(a)
Positive time factor Im(ξ )
exp(+iω t )
+ω / c
Re(ξ )
−ω / c
(b)
Negative time factor
Im(ξ )
exp(−iω t )
+ω / c
Re(ξ )
−ω / c
Fig. 2.4 Path deformation for the inversion integral. a Positive time factor, b Negative time factor
The product of two delta functions in the nonhomogeneous term shows the location
of the source, i.e. the source S is placed at the coordinate origin (0, 0) in (x, y)-plane.
The convergence condition at infinity,

@/
ffi ¼ @/
/jpffiffiffiffiffiffiffiffi ¼
x2 þy2 !1 @x pffiffiffiffiffiffiffiffi
ffi ffi ¼0
@y pffiffiffiffiffiffiffiffi
ð2:3:2Þ
x2 þy2 !1 x2 þy2 !1
is also imposed.
Now, we apply the integral transforms to Eq. (2.3.1). Since the unknown
function / has two space variables, we apply the double Fourier transform defined
by
Zþ1 Zþ1
1
/ðnÞ ¼ /ðxÞ expðþinxÞdx; /ðxÞ ¼ /ðnÞ expðinxÞdn ð2:3:3Þ
2p
1 1
Zþ1 Zþ1
~ 1 ~
/ðgÞ ¼ /ðyÞ expðþigyÞdy; /ðyÞ ¼ /ðgÞ expðigyÞdg ð2:3:4Þ
2p
1 1
to the differential equation (2.3.1) successively, as

0 1
Zþ1 Zþ1
@ @2/ @2/
þ ¼ SdðxÞdðyÞ expðþinxÞdxA expðþigyÞdy ð2:3:5Þ
@x2 @y2
1 1
With the aid of the convergence condition (2.3.2), each term is transformed as
follows:
0 1
Zþ1 Zþ1
@ @ 2
/ ~
expðþinxÞdxA expðþigyÞdy ¼ n2 /
@x2
1 1
0 þ1 1
Zþ1 Z
@ @ 2
/ ~
expðþinxÞdxA expðþigyÞdy ¼ g2 / ð2:3:6Þ
@y2
1 1
0 þ1 1
Zþ1 Z
@ SdðxÞdðyÞ expðþinxÞdxA expðþigyÞdy ¼ S
1 1
~ ¼ S for the double

Then, we have the simple algebraic equation ðn2 þ g2 Þ/
transformed function and its solution is given by
~ ¼
/
S
ð2:3:7Þ
n2 þ g2
The reader will find that the partial differential equation (2.3.1) is transformed to
the simple algebraic equation. There is thus no need of solving the differential
equation directly. The subsequent inversion process is however crucial for the
solution in the actual space. The formal Fourier inversion integral with respect to
the parameter g,
Zþ1 Z1
¼ 1
/
S
expðigyÞdg ¼
S 1
cosðgyÞdg ð2:3:8Þ
2p n þg
2 2 p n þ g2
2
1 0
is evaluated with the aid of the formula (2.1.22) and yields
¼ S expðjnjjyjÞ
/ ð2:3:9Þ
2jnj
The next inversion is to evaluate the Fourier inversion integral with respect to the
parameter n,
Zþ1
1 S
/¼ expðjnjjyjÞ expðinxÞdn ð2:3:10Þ
2p 2jnj
1
Inspecting the integrand, the singular point at n ¼ 0 lies on the real axis, i.e. on the
integration path. It is impossible to evaluate the integral in the regular sense. So, we
have to deform the integration path around the pole as that in the previous section.
But it was somewhat complicated to determine the path deformation. In order to
avoid this troublesome work, we employ a simpler way for evaluating the integral.
Since the trouble stems from the singular point at n ¼ 0, in order to avoid the
trouble, we differentiate equation (2.3.10) with respect to the space variables, x and
y, respectively,
Zþ1
@/ S in
¼ expðjnjjyjÞ expðinxÞdn ð2:3:11Þ
@x 4p jnj
1
Zþ1
@/ S
¼ sgnðyÞ expðjnjjyjÞ expðinxÞdn ð2:3:12Þ
@y 4p
1
where sgn(.) is the sign function defined by

þ1; y [ 0
sgnðyÞ ¼ ð2:3:13Þ
1; y\0
Using the symmetry of the integrand in Eqs. (2.3.11) and (2.3.12), the integrals are
reduced to the real valued semi-infinite integrals,
Z1
@/ S
¼ expðnjyjÞ sinðnxÞdn ð2:3:14Þ
@x 2p
0
Z1
@/ S
¼ sgnðyÞ expðnjyjÞ cosðnxÞdn ð2:3:15Þ
@y 2p
0
The two integrals in the above equations are well-known from Calculus and we
have the formulas,
Z1 Z1
x y
expðnyÞ sinðnxÞdn ¼ 2 ; expðnyÞ cosðnxÞdn ¼ ð2:3:16Þ
x þ y2 x2 þ y2
0 0
Then, Eqs. (2.3.14) and (2.3.15) are evaluated as
@/ S x @/ S y
¼ ; ¼ ð2:3:17Þ
@x 2p x2 þ y2 @y 2p x2 þ y2
The derivative of / is completely determined. We integrate the above two equations

with respect to x and y, respectively. This integration leads to two expressions for
the single function / as
@/ S
) /¼ log x2 þ y2 þ c0 ðyÞ
@x 4p
ð2:3:18Þ
@/ S
) / ¼ log x2 þ y2 þ c00 ðxÞ
@y 4p
Since the above two equations must be equal, two constant terms should be
identical,
c0 ðyÞ ¼ c00 ðxÞ ð2:3:19Þ
This condition is satisfied only when the two terms are pure constant and do not
include any space variable:
c0 ðyÞ ¼ c00 ðxÞ ¼ constant ð2:3:20Þ
Then, we have Green’s function for the Laplace equation,
@2/ @2/ S
þ ¼ SdðxÞdðyÞ ) /¼ log x2 þ y2 þ arbitrary constant
@x2 @y2 4p
ð2:3:21Þ
This Green’s function does not satisfy the convergence condition at infinity, since it
has the constant. This is because we could not carry out the Fourier inversion
integral of Eq. (2.3.10) directly.
The 2D wave equation with an impulsive source is given by
@2/ @2/ 1 @2/

þ ¼ PdðxÞdðyÞdðtÞ ð2:4:1Þ
@x2 @y2 c2 @t2
The nonhomogeneous term states that the source with magnitude P is placed at the
coordinate origin and is impulsive in time. The quiescent condition at an initial
time,

@/
/jt¼0 ¼ ¼0 ð2:4:2Þ
@t t¼0

@/ @/
/jpffiffiffiffiffiffiffiffiffi ¼ ffi ¼ @y pffiffiffiffiffiffiffiffiffi ¼ 0 ð2:4:3Þ
x2 þy2 !1 x2 þy2 !1
are also employed. As the wave equation (2.4.1) has two space variables x and y,
and one time variable t, the triple integral transform is applied to the differential
equation (2.4.1): the Laplace transform with respect to the time variable,
Z1

/ ðsÞ ¼ /ðtÞ expðstÞdt ð2:4:4Þ
0
and the double Fourier transform with respect to the space variables,
Zþ1 Zþ1
1
2p
1 1
Zþ1 Zþ1
~ 1 ~
2p
1 1
Applying this triple integral transform with the quiescent and convergence condi-
tions, the original differential equation (2.4.1) is transformed to the simple algebraic
equation for the unknown function / ~ ,
~ g2 /
n2 /
~ ¼ ðs=cÞ2 /
~ P

ð2:4:7Þ
~ in the transformed domain is given by

Then, the exact expression for /
~ ¼
/
P
ð2:4:8Þ
n þ
2
g2 þ ðs=cÞ2
For the inversion, three inversion integrals must be carried out successively. The
first one is the Fourier inversion integral with respect to the parameter g. This is
reduced to the semi-infinite integral as
Zþ1 Z1
¼ 1
/
P
expðigyÞdg ¼
P cosðgyÞ
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi2 dg
2
2p n þ g þ ðs=cÞ
2 2 p
1 0 g2 þ n2 þ ðs=cÞ2
ð2:4:9Þ
The integral on the far right side is easily evaluated by applying the formula
(2.1.22). Then, the first Fourier inversion integral in Eq. (2.4.9) yields
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
P
¼ qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
/ ffi exp jyj n2 þ ðs=cÞ2 ð2:4:10Þ
2 n2 þ ðs=cÞ2
Secondly, we apply the inversion integral with respect to the parameter n,

Zþ1 qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
1 P
/ ¼ qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi exp jyj n2 þ ðs=cÞ2 expðinxÞdn ð2:4:11Þ
2p
1 2 n2 þ ðs=cÞ2
The above integral is also reduced to the semi-infinite integral, as

Z1 qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
P
1
/ ¼ qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi exp jyj n2 þ ðs=cÞ2 cosðnxÞdn ð2:4:12Þ
2p
0 n2 þ ðs=cÞ2
and we apply the integration formula (Erdélyi 1954, vol. I, pp. 17, 27)
Z1 pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
1
pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi exp c x2 þ a2 cosðbxÞdx ¼ K0 a b2 þ c2 ð2:4:13Þ
x 2 þ a2
0
where K0 ð:Þ is the zeroth order modified Bessel function of the second kind. Then,
Eq. (2.4.12) takes the compact form
P s pffiffiffiffiffiffiffiffiffiffiffiffiffiffi
/ ¼ K0 x2 þ y2 ð2:4:14Þ
2p c
The last inversion is the Laplace inversion. The Laplace inversion is symboli-
cally expressed as
P 1 h s pffiffiffiffiffiffiffiffiffiffiffiffiffiffii
/¼ L K0 x2 þ y2 ð2:4:15Þ
2p c
We have the suitable inversion formula (Erdélyi 1954, vol. I, pp. 277, 8),

Hðt aÞ 0; t\a
L1 ½K0 ðasÞ ¼ pffiffiffiffiffiffiffiffiffiffiffiffiffiffi ¼ pffiffiffiffiffiffiffiffi
1 ffi
; t[a ð2:4:16Þ
t 2 a2 t a
2 2
Applying this formula to Eq. (2.4.15), the simple expression for / is obtained as
(
cP H ðct r Þ cP 0; ct\r
/¼ qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi ¼ pffiffiffiffiffiffiffiffiffiffiffiffi
1
; ct [r ð2:4:17Þ
2p 2p
ðctÞ2 r 2
2
ðctÞ r
2
where Hð:Þ is Heaviside’s unit step function and the radial distance r from the
source is defined by
r¼ x2 þ y2 ð2:4:18Þ
Consequently, we have the exact expression for Green’s function of the 2D wave
equation as
@2/ @2/ 1 @2/ cP Hðct rÞ

þ ¼ PdðxÞdðyÞdðtÞ ) /¼ qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi ð2:4:19Þ
@x2 @y2 c2 @t2 2p
ðctÞ2 r 2
When the source is vibrating harmonically, the nonhomogeneous 2D wave equation

is given by
@2/ @2/ 1 @2/

þ ¼ QdðxÞdðyÞ expðþixtÞ ð2:5:1Þ
@x2 @y2 c2 @t2
where Q and x are the magnitude and the frequency of the source, respectively. We
assume that its Green’s function satisfies the convergence condition at infinity,

p ffiffiffiffiffiffiffiffi
ffi @/ @/
/j x2 þy2 !1 ¼ ffi ¼ @y pffiffiffiffiffiffiffiffiffi ¼ 0 ð2:5:2Þ
@x pffiffiffiffiffiffiffiffi
x2 þy2 !1 x2 þy2 !1
As the standard technique for the time-harmonic response, the Green’s function
is assumed as the product
/ðx; y; tÞ ¼ /# ðx; yÞ expðþixtÞ ð2:5:3Þ
where /# ðx; yÞ is an amplitude function to be determined. Substituting this

assumption into the nonhomogeneous wave equation (2.5.1), we have the reduced
wave equation (so-called Helmholtz equation) for the amplitude function /# ðx; yÞ,
@ 2 /# @ 2 /#
þ þ ðx=cÞ2 /# ¼ QdðxÞdðyÞ ð2:5:4Þ
@x2 @y2
The convergence condition (2.5.2) is also rewritten for the amplitude function as

@/# @/#
/# pffiffiffiffiffiffiffiffi
ffi ¼
ffi ¼ @y pffiffiffiffiffiffiffiffiffi ¼ 0 ð2:5:5Þ
x2 þy2 !1 x2 þy2 !1
The double Fourier transform with respect to two space variables as defined by
Eqs. (2.4.5) and (2.4.6) is applied to the nonhomogeneous Helmholtz
equation (2.5.4),
2 3
Zþ1 Zþ1 2 # 2 #

4 @ / @ /
þ ¼ ðx=cÞ / QdðxÞdðyÞ expðþinxÞdx5 expðþigyÞdy
2 #
@x2 @y2
1 1
ð2:5:6Þ
Defining the double transform of the amplitude function as

2 3
Zþ1 Zþ1
~
#
/ ¼ 4 /# ðx; yÞ expðþinxÞdx5 expðþigyÞdy ð2:5:7Þ
1 1
Equation (2.5.6) is transformed into the algebraic equation for the unknown
~ # ,
amplitude function /
~ g2 /
n2 /
# ~ ¼ ðx=cÞ2 /
# ~ Q #
ð2:5:8Þ
Then, the explicit expression for the amplitude function in the transformed domain
is given by
~ # ¼
/
Q
ð2:5:9Þ
n þ
2
g2 ðx=cÞ2
As the first inversion, we apply the Fourier inversion integral with respect to the
parameter g. Its formal inversion integral is simplified as
Zþ1 Z1
# ¼ 1
/
Q
expðigyÞdg ¼
Q cosðgyÞ
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi2 dg
2
2p n þ g ðx=cÞ
2 2 p
1 0 g2 þ n2 ðx=cÞ2
ð2:5:10Þ
The far right integral is evaluated with the aid of the formula (2.1.22) and yields
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
# Q
/ ¼ qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi exp jyj n2 ðx=cÞ2 ð2:5:11Þ
2 n2 ðx=cÞ2
where the real part of the square root function n2 ðx=cÞ2 must be positive in
order to guarantee the convergence at infinity jyj ! 1, i.e.
Re n2 ðx=cÞ2 0 ð2:5:12Þ
The second Fourier inversion integral with respect to the parameter n is given by
Zþ1 qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
# Q 1
/ ¼ qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi exp jyj n2 ðx=cÞ2 expðinxÞdn ð2:5:13Þ
4p
1 n2 ðx=cÞ2
The integrand in the above inversion integral has two branch points at n ¼ x=c
which are on the integration path, and the square root function (radical)
n2 ðx=cÞ2 changes its sign around the branch point. Thus, we have to discuss
the path deformation around the branch point as was done in the case of the 1D
time-harmonic problem in Sect. 2.2. However, the singular point for the 1D
problem was the simple pole, but that for the present 2D problem is the branch point
for the square root function. Therefore, we have to discuss the introduction of the
branch cut in the complex plane. As the detailed discussion for the introduction of
branch cuts has been carried out in Sect. 1.3.2 in Chap. 1, we do not repeat, but, we
cite here its result. If the reader needs more about the introduction of the branch cut,
please return to Sect. 1.3 in Chap. 1.
In order to evaluate the inversion integral of Eq. (2.5.13), we define the integral I as
1 1
I¼ qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi exp inx jyj n2 ðx=cÞ2 dn ð2:5:14Þ
2p
1 n2 ðx=cÞ2
and consider the complex integral having the same integrand as that in the above
equation,
Z qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
1 1
U¼ qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi exp ifx jyj f2 ðx=cÞ2 df ð2:5:15Þ
2p
C f2 ðx=cÞ2
In the subsequent analysis, it should be understood that the real and imaginary parts
of the complex variable f are n and g, i.e. f ¼ n þ ig. The square root function
f2 ðx=cÞ2 has two branch points at f ¼ x=c, and the corresponding two
branch cuts must be introduced in the complex f-plane, in order to make the radical
be single-valued and to guarantee the convergence condition,
Re f2 ðx=cÞ2 0 ð2:5:16Þ
The discussion on the branch cut for the square root function in Sect. 1.3.2
shows two ways to consider the complex frequency: one is the complex frequency
with a small positive imaginary part and the other with a small negative imaginary
part. As the shape of the branch cut depends on the sign of the imaginary part, we
evaluate the integral I for these two cases of the complex frequency, separately, and
then determine the suitable imaginary part.
(1) Complex frequency with the negative imaginary part
If we add a small negative imaginary part to the frequency, the branch points are
shifted from the real axis and then two branch cuts are introduced along the
hyperbola in the second and the fourth quadrants in the complex f-plane as shown
in Figs. 1.7 and 1.8 in Chap. 1. When the imaginary part of the frequency tends to
zero as the limit, the branch cut lies on the real and imaginary axes as shown by the
thick lines in Fig. 1.8a. Therefore, we take the closed loop C for our complex
integral of Eq. (2.5.15) as shown in Fig. 2.5. We have two closed loops for C: one is
the upper closed loop C ðþÞ and the other is the lower loop C ðÞ . The selection of the
two loops depends on the convergence of the integrands at the infinity jfj ! 1.
The lower loop is employed when the parameter (space variable) x is positive, and
the upper loop is employed when x\0. The following evaluation is carried out for
the case of the positive space variable, x [ 0.
When x [ 0, the loop for the complex integral is C ðÞ as shown in Fig. 2.5

!
where the integration path is the straight line AOB on the real axis. In order to
exclude the lower branch cut, the path along the branch cut is introduced. The path
\
is composed of two lines, CDE and FGH along the branch cut, and a small circle EF
around the branch point. It should be understood that the radius of the small circle

!
vanishes as the limit. Two edges of the straight line AOB are connected to the edges
η = Im(ζ )
Loop C ( + )
x<0
+ω / c
A O B
D E
−ω / c ξ = Re(ζ )
G F
Loop C ( − )
C H x>0
Fig. 2.5 Closed loop C for the complex integral U in Eq. (2.5.15) (the negative imaginary part of
the complex frequency vanished)
\ \
of the path CDE and FGH through two quarter circles, AC and BH, respectively. It
is also understood that the radius of the quarter circle tends to infinity as the limit.
Then, the closed loop C ðÞ is the sum of the integrals along these paths. In the
clockwise direction, the closed loop is

! \ ! \ ! \
C ðÞ : AOB þ BH þHGF þ FE þEDC þ CA ð2:5:17Þ
The complex integral of Eq. (2.5.15) is decomposed into the integrals along the
path segments as
0 1
Z Z Z Z Z Z qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
1 B C 1
U¼ B þ þ þ þ þ C qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
ffi exp ifx jyj f2 ðx=cÞ2 df
2p @ A

! \ ! \ ! \ f2 ðx=cÞ2
AOB BH HGF FE EDC CA
ð2:5:18Þ
\
The integral along the small circle FE around the branch point vanishes as its radius
\ \
tends to zero. The two integrals along the large arcs BH and CA also vanish as the
radius tends to infinity due to the convergence condition of Eq. (2.5.16). Thus, since
no singular point is included in the closed loop and the total value of the complex

!
integral vanishes, i.e. U ¼ 0, the integral along the real axis AOB is converted to
the integrals along the branch cut,
1 1
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi exp ifx jyj f2 ðx=cÞ2 df
2p
! f2 ðx=cÞ2
AOB
0 1
Z Z qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
1 B
B
C 1
C qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
¼ þ ffi exp ifx jyj f2 ðx=cÞ2 df
2p @ A 2
! ! f ðx=cÞ
2
CDE FGH
ð2:5:19Þ
! ! ! !
where the directions of paths CDE and FGH are inverse of the paths EDC and HGF,
respectively. Equation (2.5.19) states that it is enough to consider the integrals
!
along the branch cut to evaluate the integral on the real axis. The line paths CDE
!
and FGH are further decomposed into the line segments along the real and imag-
inary axes. They are
! ! ! ! ! !
CDE ¼ CD þ DE ; FGH ¼ FG þ GH ð2:5:20Þ
The argument and value of the square root function have already been determined
in Sect. 1.3.2(2) in Chap. 1 and are tabulated in Table 2.1 with the integration
variable along each line segment. Applying these results, the integral along each
path is given by
1 1
2p

! f2 ðx=cÞ2
AOB
ð2:5:21Þ
Z1 qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
1 1
¼ qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi exp inx jyj n2 ðx=cÞ2 dn
2p
1 n2 ðx=cÞ2
1 1
2p
! f2 ðx=cÞ2
CD
ð2:5:22Þ
1 1
¼ qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi exp gx ijyj g2 þ ðx=cÞ2 dg
2p
0 g2 þ ðx=cÞ2
Table 2.1 Value and qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi

Path Variable f and its range
argument of the square root f2 ðx=cÞ2
function on the integration ! qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
f¼n
path in Fig. 2.5 AOB n2 ðx=cÞ2
1\n\ þ 1
! qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
f ¼ ig
CD þi g2 þ ðx=cÞ2
0\g\1
f ¼ þn
DE þi ðx=cÞ2 n2
0\n\x=c
f ¼ þn
FG i ðx=cÞ2 n2
0\n\x=c
n ¼ ig
GH i g2 þ ðx=cÞ2
0\g\1
!
n and g are positive real, except BOA
1 1
2p
! f2 ðx=cÞ2
DE
ð2:5:23Þ
Zx=c qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
i 1
¼ qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi exp inx ijyj ðx=cÞ2 n2 dn
2p
0 ðx=cÞ2 n2
1 1
2p
! f2 ðx=cÞ2
FG
ð2:5:24Þ
i 1
¼ qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi exp inx þ ijyj ðx=cÞ2 n2 dn
2p
0 ðx=cÞ2 n2
1 1
2p
! f2 ðx=cÞ2
GH
ð2:5:25Þ
1 1
¼ qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi exp gx þ ijyj g2 þ ðx=cÞ2 dg
2p
0 g2 þ ðx=cÞ2
Substituting the above equations into Eq. (2.5.19) with the path decomposition of
Eq. (2.5.20), we have for the integral I,
1 1
I¼ qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi exp inx jyj n2 ðx=cÞ2 dn
2p
1 n2 ðx=cÞ2
1 1
¼ qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi expðgxÞ cos y g2 þ ðx=cÞ2 dg
p
0 g2 þ ðx=cÞ2
ð2:5:26Þ
1 1
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi sinðnxÞ cos y ðx=cÞ2 n2 dn
p
0 ðx=cÞ2 n2
i 1
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi cosðnxÞ cos y ðx=cÞ2 n2 dn
p
0 ðx=cÞ2 n2
The Fourier inversion integral has just been converted to three real-valued integrals.
Fortunately, we can evaluate these integrals with the aid of integration formulas.
Two integration formulas (Gradshteyn and Ryzhik 1980, pp. 755, 6.677, No. 4)*
Z1 pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi Za pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
1 1
pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi cos b a2 þ x2 expðcxÞdx pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi cos b a2 x2 sinðcxÞdx
a þx
2 2 a x
2 2
ð2:5:27Þ
0 0
p pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
¼ Y0 a b2 þ c2
2
and (Erdélyi 1954, vol. I, pp. 28, 42)
Za pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
1 p pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi cos b a2 x2 cosðcxÞdx ¼ J0 a b2 þ c2 ð2:5:28Þ
a2 x 2 2
0
are applied to the right hand side of Eq. (2.5.26) and the Hankel function (Watson
1966, p. 73) is introduced. Then, Eq. (2.5.26) is simplified as
1 1
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi exp inx jyj n2 ðx=cÞ2 dn
2p
1 n2 ðx=cÞ2
1 x pffiffiffiffiffiffiffiffiffiffiffiffiffiffi i x pffiffiffiffiffiffiffiffiffiffiffiffiffiffi ð2:5:29Þ
¼ Y0 x2 þ y2 J0 x2 þ y2
2 c 2 c
i ð2Þ x pffiffiffiffiffiffiffiffiffiffiffiffiffiffi
¼ H0 x2 þ y2
2 c
where J0 ð:Þ and Y0 ð:Þ are Bessel functions of the first and second kind, respectively,
ð2Þ
and Hankel function of the second kind is defined by H0 ð:Þ ¼ J0 ð:Þ iY0 ð:Þ.
Finally, we have the simple expression for the Fourier inversion integral I. That is
1 1
2p
1 n2 ðx=cÞ2
¼ H0 x2 þ y2 ð2:5:30Þ
2 c
It is easily understood that the above equation is also valid for the negative space
variable, x\0, since Eq. (2.5.30) is the even function of the space variable x. So, if
we perform the complex integral with the upper closed loop C ðþÞ in Fig. 2.5, we can
obtain the same expression for the negative space variable. However, the result in
Eq. (2.5.30) is valid only when the negative imaginary part of the complex fre-
quency tends to zero.
(2) Complex frequency with the positive imaginary part
If we assume that the imaginary part of the complex frequency approaches to
zero from the positive, the branch cuts corresponding to two branch points are
introduced in the first and the third quadrants in the complex plane as was discussed
in Sect. 1.3.2 in Chap. 1 and are shown in Fig. 1.11 as its zero limit. Thus, the
η = Im(ζ )
Loop C ( + )
x<0
−ω / c
A O B
F´
G´ ξ = Re(ζ )
E´ D´ +ω / c
Loop C ( − )
C´ H´ x>0
Fig. 2.6 Closed loop C for the complex integral U in Eq. (2.5.15) (the positive imaginary part of
the complex frequency vanished)
Table 2.2 Value and qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi

Path Variable f and its range
argument of the root function f2 ðx=cÞ2
along the integration path in ! qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
f¼n
Fig. 2.6 AOB n2 ðx=cÞ2
1\n\ þ 1
f ¼ ig
C0 D0 þi g2 þ ðx=cÞ2
0\g\1
f ¼ n
D0 E0 þi ðx=cÞ2 n2
0\n\x=c
f ¼ n
F0 G0 i ðx=cÞ2 n2
0\n\x=c
n ¼ ig
G0 H0 i g2 þ ðx=cÞ2
0\g\1
!
n and g are positive real, except BOA
closed loop for the complex integral U is taken as C ðÞ in the complex f-plane as
shown in Fig. 2.6 and the value and argument of the square root function are listed
in Table 2.2. We can apply the same evaluation procedure as that in the previous
subsection. The analysis is carried out for the case of the positive space variable,
x [ 0. The integral on the real axis, i.e. the integral I, is given by
1 1
I¼ qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi exp ifx jyj f2 ðx=cÞ2 df
2p
! f2 ðx=cÞ2
AOB
ð2:5:31Þ
1 1
¼ q ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi exp inx jyj n2 ðx=cÞ2 dn
2p 2
1 n ðx=cÞ
2
and the integrals along the branch cut are

1 1
2p
! f2 ðx=cÞ2
0 0
CD
ð2:5:32Þ
1 1
¼ qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi exp gx ijyj g2 þ ðx=cÞ2 dg
2p
0 g2 þ ðx=cÞ2
1 1
2p
! f2 ðx=cÞ2
0 0
DE
ð2:5:33Þ
i 1
¼ q ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi exp þinx ijyj ðx=cÞ2 n2 dn
2p 2
0 ðx=cÞ n 2
1 1
2p
! f2 ðx=cÞ2
F0 G0
ð2:5:34Þ
i 1
¼ qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi exp þinx þ ijyj ðx=cÞ2 n2 dn
2p
0 ðx=cÞ2 n2
1 1
2p
! f2 ðx=cÞ2
0 0
GH
ð2:5:35Þ
1 1
¼ qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi exp gx þ ijyj g2 þ ðx=cÞ2 dg
2p
0 g2 þ ðx=cÞ2
Since no singular point is included in the closed loop, U ¼ 0. Thus, the integral I on
the real axis is given by
1 1
I¼ qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi exp inx jyj n2 ðx=cÞ2 dn
2p
1 n2 ðx=cÞ2
1 1
¼ qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi expðgxÞ cos y g2 þ ðx=cÞ2 dg
p
0 g2 þ ðx=cÞ2
ð2:5:36Þ
1 1
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi sinðnxÞ cos y ðx=cÞ2 n2 dn
p
0 ðx=cÞ2 n2
i 1
þ qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi cosðnxÞ cos y ðx=cÞ2 n2 dn
p
0 ðx=cÞ2 n2
The above equation can be simplified by applying the formulas (2.5.27) and
(2.5.28). We have
1 1
2p
1 f2 ðx=cÞ2
¼ þ H0 x2 þ y2 ð2:5:37Þ
2 c
ð1Þ
where Hankel function of the first kind is defined by H0 ð:Þ ¼ J0 ð:Þ þ iY0 ð:Þ.
*Derivation of formula (2.5.27).
We have the integration formula (Gradshteyn and Ryzhik 1980, pp. 755, 6.677),
8 qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
>
> 1
Z1
pffiffiffiffiffiffiffiffiffiffiffiffiffiffi > pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi sin z a2 b2 ; b\a
<
a2 b2 qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
Y0 a x2 þ z2 cosðbxÞdx ¼
>
> 1
0 : pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
> exp z b2 a2 ; b [ a
b a
2 2
ð2:5:38Þ
If we consider the above as the Fourier cosine transform with respect to the
transform parameter b, its inverse cosine transform is given by
Za qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
p pffiffiffiffiffiffiffiffiffiffiffiffiffiffi 1
Y0 a x2 þ z2 ¼ pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi sin z a2 b2 cosðbxÞdb
2 a 2 b2
0
ð2:5:39Þ
Z1 qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
1
pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi exp z b2 a2 cosðbxÞdb
b2 a2
a
Making the change of variable, u ¼ b2 a2 , for the second integral in the right
hand side, the formula (2.5.27) is obtained as
Za qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
p pffiffiffiffiffiffiffiffiffiffiffiffiffiffi 1
Y0 a x2 þ z2 ¼ pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi sin z a2 b2 cosðbxÞdb
2 a 2 b2
0
ð2:5:40Þ
Z1 pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
1
pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi expðzuÞ cos x u2 þ a2 du
u2 þ a2
0
(3) Selection of the branch cut and the integration loop

The Fourier inversion integral I has just been evaluated in two ways of the
branch cut and we get two expressions for the single integral I as
1 1
2p
1 n2 ðx=cÞ2
8 ð2:5:41Þ
> i ð2Þ
< H0 ðxr=cÞ; x x ieje!0
¼ 2
>
: þ i H ð1Þ ðxr=cÞ; x x þ ieje!0
2 0
where the radial distance r from the source point is defined by

r¼ x2 þ y2 ð2:5:42Þ
In order to determine the correct evaluation, we multiply the time factor expðþixtÞ
and replace the Hankel functions with their asymptotic forms (Watson 1966, p. 198)
as
rffiffiffiffiffi
ð1Þ 2
H0 ðzÞ expfiðz p=4Þg
rpz
ffiffiffiffiffi ; z!1 ð2:5:43Þ
ð2Þ 2
H0 ðzÞ expfþiðz p=4Þg
pz
The asymptotic form of the integral I yields to
expðþixtÞ 1
2p
1 n2 ðx=cÞ2
8 rffiffiffiffiffiffiffiffiffi
>
> ð2:5:44Þ
> i
<
2c
expðþpi=4Þ expfþixðt r=cÞg; x x ieje!0
2 pxr

rffiffiffiffiffiffiffiffiffi
r!1>> i 2c
>
:þ expðpi=4Þ expfþixðt þ r=cÞg; x x þ ieje!0
2 pxr
It is clear that the asymptotic form in the upper line shows the out-going wave from
the source point r = 0, and that in the lower line does the in-coming wave from the
infinity. Thus, the right evaluation for the integral I is
1 1 i
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi exp inx jyj n2 ðx=cÞ2 dn ¼ H0ð2Þ ðxr=cÞ
2p 2 2
1 n ðx=cÞ
2
ð2:5:45Þ
and the correct introduction of the branch cut and the integration loop are shown in
Fig. 2.5 for our positive time factor expðþixtÞ. However, if we employ the negative
time factor expðixtÞ, the branch cut and the loop in Fig. 2.6 are the right choices,
needless to say.
Anyway, we could evaluate the infinite integral of the Fourier inversion and
obtained the amplitude function in the closed form as
iQ ð2Þ pffiffiffiffiffiffiffiffiffiffiffiffiffiffi
/# ¼ H ðxr=cÞ; r¼ x2 þ y2 ð2:5:46Þ
4 0
Thus, the 2D Green’s function for the time-harmonic source is given by

@2/ @2/ 1 @2/ iQ ð2Þ

þ 2 ¼ 2 2 QdðxÞdðyÞ expðþixtÞ ) / ¼ H0 ðxr=cÞ expðþixtÞ
@x 2 @x c @t 4
ð2:5:47Þ
When the time factor is negative, we have
@2/ @2/ 1 @2/ iQ ð1Þ

þ 2 ¼ 2 2 QdðxÞdðyÞ expðixtÞ ) / ¼ þ H0 ðxr=cÞ expðixtÞ
@x 2 @x c @t 4
ð2:5:48Þ
(4) Convolution integral

The solution of Eq. (2.5.47) can be obtained directly by applying the convolu-
tion integral of the impulsive solution which is given by Eq. (2.4.19). Let us
consider its convolution. The convolution integral for the Laplace transform is
defined by
Zt
/ðx; y; tÞ ¼ /ðimplseÞ ðx; y; t0 Þ expfþixðt t0 Þgdt0 ð2:5:49Þ
0
where /ðimplseÞ is the solution for the impulsive source. We substitute the impulsive
Green’s function of Eq. (2.4.19) with the source magnitude Q into Eq. (2.5.49),
Zt
cQ Hðct0 rÞ
/ðx; y; tÞ ¼ qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi expfþixðt t0 Þgdt0 ð2:5:50Þ
2p
0 ðct0 Þ2 r 2
and examine the supporting region for the step function. We have
Zt
cQ 1
/ðx; y; tÞ ¼ expðþixtÞHðt r=cÞ qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi expðixt0 Þdt0 ð2:5:51Þ
2p
r=c ðct0 Þ2 r 2
As we are considering the steady-state response, the time in the upper integration
limit and that in the step function can be set to be infinite as
2 3
Zt
cQ 6 1 7
/ðx; y; tÞ ¼ expðþixtÞ lim 4Hðt r=cÞ qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi expðixt0 Þdt05 ð2:5:52Þ
2p t!1 2
r=c ðct0 Þ r 2
Taking the limit for the time, we have
Z1
Q 1
/ðx; y; tÞ ¼ expðþixtÞ pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi expðixu=cÞdu ð2:5:53Þ
2p u r2
2
r
This is just the integral representation for the Hankel function of the second kind
(Watson 1966, p. 170),
Z1
ð2Þ 2i 1
H0 ðxÞ ¼ pffiffiffiffiffiffiffiffiffiffiffiffiffi expðixuÞdu ð2:5:54Þ
p u 1
2
1
Thus, we have
iQ ð2Þ
/ðx; y; tÞ ¼ expðþixtÞH0 ðrx=cÞ ð2:5:55Þ
4
(5) Direct evaluation

In the former subsection (1) and (2), the Fourier inversion is carried out by
applying the Cauchy complex integral. It was lengthy troublesome work. Without
applying the Cauchy theorem, we can evaluate the integral I,
1 1
I¼ qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi exp inx jyj n2 ðx=cÞ2 dn ð2:5:56Þ
2p
1 n2 ðx=cÞ2
Firstly, we reduce the infinite integral to the semi-infinite integral,
1 1
I¼ qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi cosðnxÞ exp jyj n2 ðx=cÞ2 dn ð2:5:57Þ
p
0 n2 ðx=cÞ2
In the complex n-plane, the integration path for this integral is just on the real axis
and two branch cuts for the radical n2 ðx=cÞ2 are introduced on the real and
imaginary axes as was discussed in Sect. 1.3.2 in Chap. 1. The slightly shifted
branch cuts are shown in Fig. 2.7 where the integration path for the semi-infinite
integral is OP. The integration path in the region 0\n\x=c is slightly up from the
Im(ξ )
branch cut
+ω / c
Q P
O
Re(ξ )
−ω / c
branch cut
Fig. 2.7 Branch cuts and integration path for the inversion integral of Eq. (2.5.57)
lower branch cut. When the cut approaches to the real axis as was discussed in
Sect. 1.3.2 (2), the argument of the square root function in the integrand yields to
þp=2, i.e.
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
n ðx=cÞ ¼ þi ðx=cÞ2 n2 ;
2 2
jnj\x=c ð2:5:58Þ
In the other positive region n [ x=c on the real axis, the root function is positive
real. Then, the semi-infinite integral is decomposed into two integrals in the regions,
0\n\x=c and n [ x=c, as
1 1
I¼ qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi cosðnxÞ exp ijyj ðx=cÞ2 n2 dn
p
0 ðþiÞ ðx=cÞ2 n2
ð2:5:59Þ
1 1
þ qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi cosðnxÞ exp jyj n2 ðx=cÞ2 dn
p
x=c n2 ðx=cÞ2
The above integral I is further decomposed into the real and imaginary parts,
1 1
I¼ qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi cosðnxÞ sin jyj ðx=cÞ2 n2 dn
p
0 ðx=cÞ2 n2
1 1
þ qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi cosðnxÞ exp jyj n2 ðx=cÞ2 dn ð2:5:60Þ
p
x=c n2 ðx=cÞ2
i 1
p
0 ðx=cÞ2 n2
We make the change of variable for the first and second integrals in the above
equation. The changes are defined by
gdg
g¼ ðx=cÞ2 n2 ; n¼ ðx=cÞ2 g2 ; dn ¼ qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi ð2:5:61Þ
ðx=cÞ2 g2
for the first integral, and

gdg
g ¼ n2 ðx=cÞ2 ; n ¼ g2 þ ðx=cÞ2 ; dn ¼ qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi ð2:5:62Þ
g2 þ ðx=cÞ2
for the second integral. Equation (2.5.60) yields to
1 1
I¼ qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi cos x ðx=cÞ2 g2 sinðjyjgÞdg
p
0 ðx=cÞ2 g2
1 1
þ qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi cos x g2 þ ðx=cÞ2 expðjyjgÞdg ð2:5:63Þ
p
0 g2 þ ðx=cÞ2
i 1
p
0 ðx=cÞ2 n2
Then, we apply the formula of Eq. (2.5.27) to the first two integrals and the formula
of Eq. (2.5.28) to the third integral. The integral I is exactly evaluated as
1 x pffiffiffiffiffiffiffiffiffiffiffiffiffiffi i x pffiffiffiffiffiffiffiffiffiffiffiffiffiffi i ð2Þ x pffiffiffiffiffiffiffiffiffiffiffiffiffiffi

I ¼ Y0 x2 þ y2 J0 x2 þ y2 ¼ H0 x2 þ y2 ð2:5:64Þ
2 c 2 c 2 c
This is just the same as Eqs. (2.5.30) and (2.5.45) for the integral I.
The static 3D Green’s function for the Laplace equation is a particular solution of
the nonhomogeneous differential equation,
@2/ @2/ @2/

þ þ 2 ¼ SdðxÞdðyÞdðzÞ ð2:6:1Þ
@x2 @y2 @z
where S is the magnitude of the source which is placed at the coordinate origin
(0, 0, 0). The convergence condition at infinity,

p ffiffiffiffiffiffiffiffiffiffiffiffiffiffi @/ @/ @/
/j x2 þy2 þz2 !1 ¼ ¼ ¼ ¼0
@x pxffiffiffiffiffiffiffiffiffiffiffiffiffiffi
2 þy2 þz2 !1 @y pffiffiffiffiffiffiffiffiffiffiffiffiffiffi
x2 þy2 þz2 !1 @z pffiffiffiffiffiffiffiffiffiffiffiffiffiffi
x2 þy2 þz2 !1
ð2:6:2Þ
is applied to the Green’s function /.

The Green’s function is obtained by the method of integral transform. For three
space variables, the triple Fourier transform defined by
Zþ1 Zþ1
1
2p
1 1
Zþ1 Zþ1
~ 1 ~
2p
1 1
Zþ1 Zþ1
^ 1 ^ expðifzÞdf
/ðfÞ ¼ /ðzÞ expðþifzÞdz; /ðzÞ ¼ /ðfÞ ð2:6:5Þ
2p
1 1
is applied to Eq. (2.6.1),

2 3
Zþ1 * Zþ1 Zþ1 +
4 @2/ @2/ @2/ 5 expðþigyÞdy expðþifzÞdz
þ þ ¼ SdðxÞdðyÞdðzÞ expðþinxÞdx
@x2 @y2 @z2
1 1 1
ð2:6:6Þ
Applying the convergence condition (2.6.2), the above Eq. (2.6.6) is transformed to
^~
the algebraic equation for the triple transformed function /,
^~
ðn2 þ g2 þ f2 Þ/ ¼ S ð2:6:7Þ
The transformed function is determined explicitly,
^~ S
/ ¼ ð2:6:8Þ
n2 þ g2 þ f 2
The first Fourier inversion integral with respect to the parameter f is
Zþ1 Z1
~
¼ 1 S
expðifzÞdf ¼
S 1
/ pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi2 cosðfzÞdf ð2:6:9Þ
2p n þg þf
2 2 2 p f 2 þ n2 þ g2
1 0
The above integral is easily evaluated with the aid of the formula (2.1.22). It yields
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
~
S
/ ¼ pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi exp jzj n2 þ g2 ð2:6:10Þ
2 n þg
2 2
The second inversion integral with respect to the parameter g is reduced to the semi-
infinite integral as
Zþ1 qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
¼ 1
/
S
pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi exp jzj n2 þ g2 expðigyÞdg
2p 2 n2 þ g2
1
ð2:6:11Þ
S 1
¼ pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi exp jzj n2 þ g2 cosðgyÞdg
2p n2 þ g2
0
The latter semi-infinite integral can be evaluated by using the formula (2.4.13). It
follows that
¼ S K0 jnj y2 þ z2
/ ð2:6:12Þ
2p
The last inversion integral with respect to the parameter n is also reduced to the
Zþ1 pffiffiffiffiffiffiffiffiffiffiffiffiffiffi
S
/¼ 2 K0 jnj y2 þ z2 expðinxÞdn
4p
1
ð2:6:13Þ
Z1 pffiffiffiffiffiffiffiffiffiffiffiffiffiffi
S
¼ 2 K0 n y2 þ z2 cosðnxÞdn
2p
0
Fortunately, we have the suitable integration formula, (Erdélyi 1954, vol. I, pp. 49, 40)
Z1
p
K0 ðanÞ cosðbnÞdn ¼ pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi ð2:6:14Þ
2 a þ b2
2
0
Applying this formula to Eq. (2.6.13), the last inversion integral is exactly evaluated
as
S
/¼ pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi ð2:6:15Þ
4p x2 þ y2 þ z 2
Consequently, we have the static 3D Green’s function for the Laplace equation,
@2/ @2/ @2/ S

þ þ 2 ¼ SdðxÞdðyÞdðzÞ ) /¼ ð2:6:16Þ
@x2 @y2 @z 4pR
where R is the 3D radial distance from the source,

pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
R¼ x2 þ y2 þ z 2 ð2:6:17Þ
Green’s function for the 3D wave equation is discussed. The wave equation with an
impulsive point source located at the coordinate origin is given by
@2/ @2/ @2/ 1 @2/

þ þ 2 ¼ 2 2 PdðxÞdðyÞdðzÞdðtÞ ð2:7:1Þ
@x2 @y2 @z c @t
where P is the magnitude of the source. The quiescent condition at an initial time,

@/
/jt¼0 ¼ ¼0 ð2:7:2Þ
@t t¼0

@/ @/ @/
/jpxffiffiffiffiffiffiffiffiffiffiffiffiffiffi
2 þy2 þz2 !1 ¼ ¼ ¼ ¼0 ð2:7:3Þ
2 þy2 þz2 !1 @y pxffiffiffiffiffiffiffiffiffiffiffiffiffiffi
2 þy2 þz2 !1 @z pffiffiffiffiffiffiffiffiffiffiffiffiffiffi
x2 þy2 þz2 !1
are also imposed.

Our dynamic Green’s function has four variables: three space and one time
variables. Laplace transform with respect to the time,
Zþ1

/ ðsÞ ¼ /ðtÞ expðstÞdt ð2:7:4Þ
0
and the triple Fourier transform with respect to three space variables,
Zþ1 Zþ1
1
2p
1 1
Zþ1 Zþ1
~ 1 ~
2p
1 1
Zþ1 Zþ1
^ 1 ^ expðifzÞdf
/ðfÞ ¼ /ðzÞ expðþifzÞdz; /ðzÞ ¼ /ðfÞ ð2:7:7Þ
2p
1 1
are applied to the nonhomogeneous wave equation (2.7.1). With the aid of the
quiescent and convergence conditions, the wave equation is transformed to the
^~
simple algebraic equation for the multi-transformed unknown function / ,
^
~
¼ P
fn2 þ g2 þ f2 þ ðs=cÞ2 g/ ð2:7:8Þ
The inversion starts from the transformed function,
^~ P
/ ¼ ð2:7:9Þ
n þ
2
g2 þ f2 þ ðs=cÞ2
As the first inversion, the Fourier inversion integral with respect to the parameter f
Zþ1
~ ¼ 1
/
P
expðifzÞdz
2p n þ
2
g2 þ f2 þ ðs=cÞ2
1
Z1 ð2:7:10Þ
P cosðfzÞ
¼ qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi2 dz
p
0 f2 þ n2 þ g2 þ ðs=cÞ2
is carried out by applying the formula (2.1.22). It follows that

qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
~
/
P
¼ qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
ffi exp jzj n2 þ g2 þ ðs=cÞ2 ð2:7:11Þ
2 n2 þ g2 þ ðs=cÞ2
The second inversion is the integral with respect to the parameter g,
Zþ1 qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
¼ 1
/
P
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi exp jzj n2 þ g2 þ ðs=cÞ2 expðigyÞdg
2p 2
1 2 n þ g þ ðs=cÞ
2 2
Z1 qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
P 1
¼ qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi exp jzj n2 þ g2 þ ðs=cÞ2 cosðgyÞdg
2p
0 n2 þ g2 þ ðs=cÞ2
ð2:7:12Þ
The latter semi-infinite integral is also evaluated by applying the formula (2.4.13). It
yields
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffipffiffiffiffiffiffiffiffiffiffiffiffiffiffi
¼ P K0
/ n2 þ ðs=cÞ2 y2 þ z2 ð2:7:13Þ
2p
The third inversion integral with respect to the parameter n is
Zþ1 qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffipffiffiffiffiffiffiffiffiffiffiffiffiffiffi
P 1
/ ¼ K0 n2 þ ðs=cÞ2 y2 þ z2 expðinxÞdn
2p 2p
1
ð2:7:14Þ
Z1 qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffipffiffiffiffiffiffiffiffiffiffiffiffiffiffi
P
¼ 2 K0 n2 þ ðs=cÞ2 y2 þ z2 cosðnxÞdn
2p
0
The integration formula (Erdélyi 1954, vol. I, pp. 56, 43)
Z1 qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
p
K0 a n2 þ b2 cosðcnÞdn ¼ pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi exp b a2 þ c2 ð2:7:15Þ
2 a2 þ c 2
0
is very helpful for our task. Then, applying the above formula to the last integral in
Eq. (2.7.14), we have
P s pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
/ ¼ pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi exp x2 þ y2 þ z2 ð2:7:16Þ
4p x2 þ y2 þ z2 c
The last is the Laplace inversion. Its symbolical form is
P h s pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffii
/¼ pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi L1 exp x2 þ y2 þ z 2 ð2:7:17Þ
4p x2 þ y2 þ z2 c
The transform parameter “s” is included only in the argument of the exponential
function. We remember the simple Laplace inversion formula for the delta function,
L1 ½expðasÞ ¼ dðt aÞ ð2:7:18Þ
Thus, applying this inversion formula, Eq. (2.7.17) is fully inverted as
P
/¼ dðt R=cÞ ð2:7:19Þ
4pR
where the radial distance R from the source is defined by Eq. (2.6.17). Finally, we
have the 3D Green’s function for the wave equation with the impulsive point
source,
@2/ @2/ @2/ 1 @2/ P

þ þ 2 ¼ 2 2 PdðxÞdðyÞdðzÞdðtÞ ) / ¼ dðt R=cÞ
@x2 @y2 @z c @t 4pR
ð2:7:20Þ
This is the very simple expression in spite of the 3D nature!
This section derives the 3D Green’s function for a time-harmonic source. It is the
convolution integral of the impulsive Green’s function obtained in the previous
section. The wave equation with the time-harmonic source is given by
@2/ @2/ @2/ 1 @2/

þ þ 2 ¼ 2 2 QdðxÞdðyÞdðzÞ expðþixtÞ ð2:8:1Þ
@x2 @y2 @z c @t
where Q is the magnitude of the source and x the frequency of the time-harmonic
vibration. The convergence condition at infinity,

@/ @/ @/
/jpxffiffiffiffiffiffiffiffiffiffiffiffiffiffi
2 þy2 þz2 !1 ¼ ¼ ¼ ¼0 ð2:8:2Þ
2 þy2 þz2 !1 @y pxffiffiffiffiffiffiffiffiffiffiffiffiffiffi
2 þy2 þz2 !1 @z pffiffiffiffiffiffiffiffiffiffiffiffiffiffi
x2 þy2 þz2 !1
is also imposed.
The standard multiple integral transform technique is available for getting the
time-harmonic Green’s function. However, as shown in the case of 2D Green’s
function in Sect. 2.5 (4), we take a very simple way, i.e. the convolution integral of
the impulsive Green’s function.
In the previous section we get the Green’s function for the impulsive source.
Replacing the source magnitude with unit 1, the Green function is given by
@2/ @2/ @2/ 1 @2/ 1

þ þ ¼ dðxÞdðyÞdðzÞdðtÞ ) / ¼ dðt R=cÞ ð2:8:3Þ
@x2 @y2 @z2 c2 @t2 4pR
Employing the time-harmonic source with the frequency x and the magnitude Q
Q expfþixðt t0 Þg ð2:8:4Þ
the convolution integral for the Laplace transform is given by
Zt
Q
/¼ dðt0 R=cÞ expfþixðt t0 Þgdt0 ð2:8:5Þ
4pR
0
It is very easy to evaluate the above integral, since the integrand includes Dirac’s
delta function and we can apply the simple integration formula (1.2.3) in Sect. 1.2,
Zb
f ðcÞ; a\c\b
0; c\a or b\c
a
Then, we can evaluate the integral in Eq. (2.8.5) and have for /
Q
/¼ H ðt R=cÞ expfþixðt R=cÞg ð2:8:7Þ
4pR
The step function ahead of the equation means that Eq. (2.8.7) is the transient
response to the time-harmonic source and the disturbance starts from the wave
arrival t = R/c. When sufficient long time has passed and the response becomes
steady, the step function is meaningless. Then, we have the steady-state time-
harmonic response as
Q
/¼ expfþixðt R=cÞg ð2:8:8Þ
4pR
Therefore, the 3D Green’s function for the wave equation with the time-harmonic
source is given by
Table 2.3 Green’s function for Laplace and Wave equations
Differential equation Source Green’s function
2
@2 /
1D PdðxÞdðtÞ /ðx; tÞ ¼ cP
2 Hðct jxjÞ
@x2 ¼ c12 @@t/2 Source
QdðxÞ expðixtÞ iQ
/ðx; tÞ ¼ 2ðx=cÞ expfixðt jxj=cÞg
2 2 pffiffiffiffiffiffiffiffiffiffiffiffiffiffi
2D @ / SdðxÞdðyÞ S
@x2 þ @@y/2 ¼ Source /ðx; yÞ ¼ 4p logðrÞ þ arbitrary constant; r ¼ x2 þ y2
@2 / PdðxÞdðyÞdðtÞ cP
@x2 þ @@x/2 ¼ c12 @@t/2 Source /ðx; y; tÞ ¼ pffiffiffiffiffiffiffiffiffiffiffiffi H ðct r Þ; r ¼ x2 þ y2
2p ðctÞ2 r2
ð2Þ
QdðxÞdðyÞ expðþixtÞ /ðx; y; tÞ ¼ iQ x2 þ y2
4 H0 ðxr=cÞ expðþixtÞ; r ¼
ð1Þ
p ffiffiffiffiffiffiffiffiffiffiffiffiffiffi
QdðxÞdðyÞ expðixtÞ /ðx; y; tÞ ¼ þ iQ x2 þ y2
4 H0 ðxr=cÞ expðixtÞ; r ¼
2 2 2 pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
3D @ / SdðxÞdðyÞdðzÞ S
@x2 þ @@y/2 þ @@z/2 ¼ Source /ðx; y; zÞ ¼ 4pR ; R ¼ x2 þ y2 þ z2
@2 / @2 / @2 / @2 / PdðxÞdðyÞdðzÞdðtÞ P
@x2 þ @y2 þ @z2 ¼ c12 @t2 Source /ðx; y; z; tÞ ¼ 4pR dðt R=cÞ; R ¼ x2 þ y2 þ z2
Q
QdðxÞdðyÞdðzÞ expðixtÞ /ðx; y; z; tÞ ¼ 4pR expfixðt R=cÞg; R ¼ x2 þ y2 þ z2
75
@2/ @2/ @2/ 1 @2/

þ þ 2 ¼ 2 2 QdðxÞdðyÞdðzÞ expðþixtÞ
@x2 @y2 @z c @t ð2:8:9Þ
Q
)/¼ expfþixðt R=cÞg
4pR
Exercise
(2:1) Using the impulsive response (2.1.27) for 1D wave equation, derive the
time-harmonic Green’s function through the convolution integral and
compare it with the time-harmonic Green’s function (2.2.27).
Appendix
See Table 2.3.
References
Erdélyi A (ed) (1954) Tables of integral transforms, vols I, II, McGraw-Hill, New York
Gradshteyn IS, Ryzhik IM (1980) In: Jefferey A (ed) Table of integrals, series, and products, 5th
edn. Academic Press, San Diego
Watson GN (1966) A treatise on the theory of Bessel functions, Cambridge University Press,
Cambridge
Chapter 3
Green’s Dyadic for an Isotropic Elastic
Solid
The present chapter shows how to derive an exact closed form solution, the so-
called Green’s dyadic, for elasticity equations. Introducing the Cartesian coordinate
system (xi) ≡ (x, y, z), we employ the notation u ðui Þ for the displacement,
e ðeij Þ for the strain, r ðrij Þ for the stress, B ðBi Þ for the body force and ρ for
the density. The governing equations for the deformation of an isotropic elastic
solid are constituted by the following equations of motion,
@rxx @ryx @rzx @ 2 ux

þ þ þ qBx ¼ q 2
@x @y @z @t
@rxy @ryy @rzy @ 2 uy
þ þ þ qBy ¼ q 2 ð3:1Þ
@x @y @z @t
@rxz @ryz @rzz @ 2 uz
þ þ þ qBz ¼ q 2
@x @y @z @t
and by the stress-strain relation, the so-called Hooke’s law,

2 3 2 32 3
rxx k þ 2l k k 0 0 0 exx
6 ryy 7 6 k þ 2l k 0 0 0 7 6 eyy 7
6 7 6 76 7
6 rzz 7 6 k þ 2l 0 7 6 7
6 7 6 0 0 76 ezz 7
6 rxz 7 ¼ 6 0 7 6 7 ð3:2Þ
6 7 6 2l 0 76 exz 7
4 ryz 5 4 sym: 2l 0 54 eyz 5
rxy 2l exy
Here (λ, μ) are Lame’s constants which are expressed by Young’s modulus E and
Poisson ratio ν as
mE E
k¼ ; l¼ ð3:3Þ
ð1 þ mÞð1 2mÞ 2ð1 þ mÞ

DOI 10.1007/978-3-319-17455-6_3
78 3 Green’s Dyadic for an Isotropic Elastic Solid
The strains in Eq. (3.2) are defined by the displacement gradient,
@ux @uy @uz

exx ¼ ; eyy ¼ ; ezz ¼ ;
@x @y @z
ð3:4Þ
1 @ux @uy 1 @uy @uz 1 @ux @uz
exy ¼ þ ; eyz ¼ þ ; ezx ¼ þ
2 @y @x 2 @z @y 2 @z @x
It should be noticed that the symmetry relations for the stress and strain components
are satisfied:
rij ¼ rji ; eij ¼ eji ; i; j ¼ x; y; z ð3:5Þ
The elasticity equations constitute a set of coupled partial differential equations

with 15 unknowns. In order to reduce the differential equations to more compact
forms, the strain in Hooke’s law is replaced with the displacement gradient and then
the stress is substituted into the equations of motion. We then get a set of dis-
placement equations, the so-called Navier equations, with only three unknown
functions (displacement components),

@ @ux @uy @uz @ 2 ux @ 2 ux @ 2 ux 1 @ 2 ux 1
ðc2 1Þ þ þ þ 2 þ 2 þ 2 ¼ 2 2 2 Bx
@x @x @y @z @x @y @z cs @t cs

@ @ux @uy @uz @ uy @ uy @ uy
2 2 2
1 @ uy 1
2
ðc2 1Þ þ þ þ 2 þ 2 þ 2 ¼ 2 2 2 By ð3:6Þ
@y @x @y @z @x @y @z cs @t cs

@ @ux @uy @uz @ uz @ uz @ uz
2 2 2
1 @ uz 1
2
ðc2 1Þ þ þ þ 2 þ 2 þ 2 ¼ 2 2 2 Bz
@z @x @y @z @x @y @z cs @t cs
where cs and cd are the velocities of shear and dilatational waves, respectively, and
are defined by
rffiffiffi sffiffiffiffiffiffiffiffiffiffiffiffiffiffi
l k þ 2l
cs ¼ ; cd ¼ ð3:7Þ
q q
The velocity ratio γ is defined and expressed by Poisson ratio ν,

sffiffiffiffiffiffiffiffiffiffiffiffiffiffi rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
cd k þ 2l 2ð1 mÞ
c¼ ¼ ¼ ð3:8Þ
cs l 1 2m
Then, in this chapter we shall show a solution method for the coupled dis-
placement Eqs. (3.6) with nonhomogeneous body force terms Bi. The solution
method that we employ is the integral transforms: Fourier transforms with respect to
the space variables and Laplace transform with respect the time variable.
3 Green’s Dyadic for an Isotropic Elastic Solid 79
A particular solution corresponding to a point body force is called a Green’s

function. However, each displacement component produced by one source com-
ponent is called “Green’s dyadic.” Thus, the present chapter shows the solution
method for the Green’s dyadic.
We assume here the deformation of plane-strain. Take the 2D coordinate system (x,
y) in an infinite elastic solid and neglect the anti-plane deformation produced by the
anti-plane displacement uz. The in-plane displacement components (ux, uy) are
functions of two space variables (x, y) and the time t. As a wave source, we assume
an impulsive point body force with magnitude Pi placed at the coordinate origin.
Under these assumptions, the displacement equations (3.6) is reduced to the simpler
form

@ @ux @uy @ 2 ux @ 2 ux 1 @ 2 ux Px
ðc2 1Þ þ þ 2 þ 2 ¼ 2 2 2 dðxÞdðyÞdðtÞ
@x @x @y @x @y cs @t cs
ð3:1:1Þ
@ @ux @uy @ uy @ uy
2 2
1 @ uy Py
2
ðc2 1Þ þ þ 2 þ 2 ¼ 2 2 2 dðxÞdðyÞdðtÞ
@y @x @y @x @y cs @t cs
Now, we consider these coupled differential equations. The solutions are two
displacement components (ux, uy) corresponding to the nonhomogeneous body
force Pi. Our solution strategy is very simple, namely, to transform the differential
equations into the simultaneous algebraic equations in the transformed domain. We
employ Laplace transform with respect to the time, defined by
Z1
uj ¼ uj expðstÞdt ð3:1:2Þ
0
and the double Fourier transform with respect to the two space variables, defined by
Zþ1 Zþ1
1

uj ¼ uj expðþinxÞdx; uj ¼ uj expðinxÞdn
2p
1 1
ð3:1:3Þ
Zþ1 Zþ1
1
~
uj ¼ uj expðþigyÞdy; uj ¼ ~uj expðigyÞdg
2p
1 1
where the subscript j stands for x and y. The quiescent condition at an initial time,

@uj
uj t¼0 ¼ ¼0 ð3:1:4Þ
@t t¼0
and the convergence condition at infinity

@uj @uj
uj pffiffiffiffiffiffiffiffi
ffi ¼ ¼ ffi ¼0 ð3:1:5Þ
ffi
x2 þy2 !1 @y pffiffiffiffiffiffiffiffi
x2 þy2 !1
are assumed.
Applying the triple integral transform to the displacement equations (3.1.1), we
have the algebraic equations for the transformed displacement components ~uj ,

inðc2 1Þ in~ux ig~uy n2 þ g2 ~ux ¼ ðs=cs Þ2 ~ux Px =c2s
ð3:1:6Þ
igðc2 1Þ in~ux ig~uy n2 þ g2 ~uy ¼ ðs=cs Þ2 ~uy Py =c2s
Solving for the displacement, we have the exact expressions in the transformed
domain (ξ, η, s),

Px 1 2 1 1 1
~ux ¼ n Px þ ngPy ð3:1:7aÞ
c2s a2s s2 a2d a2s

Py 1 1 1 1
~uy ¼ ngP x þ g 2
Py ð3:1:7bÞ
c2s a2s s2 a2d a2s
where two radicals are defined by

qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
ad ¼ n2 þ g2 þ ðs=cd Þ2 ; as ¼ n2 þ g2 þ ðs=cs Þ2 ð3:1:8Þ
Our main task is to invert the transformed displacement to the original space
(x, y, t). Examining the Eqs. (3.1.7), we learn that the inversion of the four
transformed functions is enough for evaluating the displacement. These four
fundamental functions are
~I ¼ 1 ð3:1:9Þ
0
c2s a2s
2
~I ¼ n 1

1
ð3:1:10Þ
xx
s2 a2d a2s

~I ¼ ng 1 1 ð3:1:11Þ
xy
s2 a2d a2s
2

~I ¼ g 1 1
2 ð3:1:12Þ
yy 2
s2 ad as
If we could have the four inversions, the displacement is expressed as
ux ¼ Px ðI0 þ Ixx Þ þ Py Ixy ; uy ¼ Px Ixy þ Py ðI0 þ Iyy Þ ð3:1:13Þ
Then, we shall consider the inversion of the four fundamentals, successively.

(1) Inversion of ~ I 0
Rewriting ~I 0 as
~I ¼ 1 1
; ð3:1:14Þ
0
c2s n2 þ g2 þ ðs=cs Þ2
we can find that this equation is the same as Eq. (2.4.8) for the transformed Green’s
~
with the replacement P → 1/c2s and c → cs. Then, we apply the same
function /
mathematics as that in Sect. 2.4 and have the inversion,
1 Hðcc t rÞ
I0 ðx; y; tÞ ¼ qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi : ð3:1:15Þ
2pcs
ðcs tÞ2 r 2
where H(.) is Heaviside’s unit step function.

(2) Inversion of ~
I xx
As we have no previous result for the second inversion ~I xx , the inversion inte-

grals are carried out successively. The Fourier inversion integral with respect to the
parameter η is given by
Z1
Ixx
n2 1 1
¼ expðigyÞdg
2ps2 a2d a2s
1
ð3:1:16Þ
2 Z1 ( )
n 1 1
¼ 2 cosðgyÞdg
ps n2 þ g2 þ ðs=cd Þ2 n2 þ g2 þ ðs=cs Þ2
0
The semi-infinite integral is evaluated with the use of the formula (2.1.22) and is
arranged so that each term has the same radical,
8 9
>
< qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi 2 >
=
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
Ixx
1 ðs=cd Þ
¼ 2 n þ ðs=cd Þ qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi exp jyj n2 þ ðs=cd Þ2
2 2
2s >
: >
n2 þ ðs=cd Þ2 ;
8 9
>
< qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi 2 >
=
1 ðs=cs Þ
þ 2 n þ ðs=cs Þ qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi exp jyj n2 þ ðs=cs Þ2
2 2
2s >
: >
n2 þ ðs=cs Þ2 ;
ð3:1:17Þ
The next Fourier inversion integral with respect to the parameter ξ is reduced to the
semi-infinite integral as
8 9
Z1 >
< qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi 2 >
=
1 ðs=cd Þ
Ixx ¼ n 2
þ ðs=c d Þ 2
ffi exp jyj n2 þ ðs=cd Þ2 cosðnxÞdn
2ps2 > : 2> ;
0 n þ ðs=cd Þ
2
8 9
Z1 >
=
1 ðs=cs Þ
þ n þ ðs=cs Þ qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi exp jyj n2 þ ðs=cs Þ2 cosðnxÞdn
2 2
2ps2 > : >
0 n2 þ ðs=cs Þ2 ;
ð3:1:18Þ
We have already had the integration formula (2.4.13). Here, it is recited again,
1
x 2 þ a2
0
where K0(.) is the zeroth order modified Bessel function of the second kind. This
formula is applicable only for the second term in the bracket; another formula is
necessary for the first term. Differentiating the formula (3.1.19) twice with respect
to the parameter “c,” we have the new formula for our use,
x2 þ a2 exp c x2 þ a2 cosðbxÞdx
0 ð3:1:20Þ
a c pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
2 2
a b2 c2 pffiffiffiffiffiffiffiffiffiffiffiffiffiffi
ffi
¼ 2 K0 a b2 þ c2 pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi 2 K1 a b2 þ c2
b þc 2
b2 þ c 2 b þ c 2
Applying these two formulas to the integrals in Eq. (3.1.18), we have
1 x2 1 x2 y2
Ixx ¼þ K 0 ð rs=c d Þ þ K1 ðrs=cd Þ
2pc2d r 2 2psrcd r 2
ð3:1:21Þ
1 x2 1 x2 y2
K 0 ð rs=c s Þ K1 ðrs=cs Þ
2pc2s r 2 2psrcs r 2
where the radial distance from the source is defined by

r¼ x2 þ y2 ð3:1:22Þ
The last inversion is the Laplace inversion. Its inversion is expressed symboli-
cally as

1 x2 1 1 x2 y2 1 1
Ixx ¼ þ L ½K0 ðrs=cd Þ þ L K1 ðrs=cd Þ
2pc2d r 2 2prcd r 2 s
ð3:1:23Þ
1 x2 1 1 x2 y2 1 1
L ½ K0 ð rs=c s Þ L K 1 ð rs=c s Þ
2pc2s r 2 2prcs r 2 s
We have already used the inversion formula (2.4.16),
Hðt aÞ
L1 ½K 0 ðasÞ ¼ pffiffiffiffiffiffiffiffiffiffiffiffiffiffi ð3:1:24Þ
t 2 a2
One more inversion formula that includes the modified Bessel function of the
second kind is that (Erdélyi 1954, vol. I, pp. 277, 11),
1 1 t 2 a2
L K 1 ðasÞ ¼ Hðt aÞ ð3:1:25Þ
s a
Applying these two inversion formulas to Eq. (3.1.23), we have the final form for
Ixx,
8 9
>
< 2 qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi=
>
Hðt r=cd Þ x 2
1 x y
2
Ixx ðx; y; tÞ ¼ þ q ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi þ ðc d tÞ 2
r 2
2pcd >r 2
: r4 >
;
ðcd tÞ2 r 2
8 9
> qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi>
Hðt r=cs Þ <x2 1 x2 y2 =
q ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi þ ðcs tÞ2 r 2 ð3:1:26Þ
2pcs >
:r
2 r 4 >
;
ðcs tÞ2 r 2
(3) Inversion of ~
I yy
The same inversion procedure as that for ~I xx is applied to the inversion of ~I yy , and

its final form is

8 9
>
< 2 qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi=
>
Hðt r=cd Þ y 2
1 y x
2
Iyy ðx; y; tÞ ¼ þ q ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi þ ðc d tÞ 2
r 2
2pcd >
:r
2 r4 >
;
ðcd tÞ2 r 2
8 9
> qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi>
Hðt r=cs Þ <y2 1 y2 x2 =
q ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi þ ðc s tÞ 2
r 2 ð3:1:27Þ
2pcs >
:r
2 r4 >
;
ðcs tÞ2 r 2
The difference between Ixx in Eq. (3.1.26) and Iyy in Eq. (3.1.27) should be
noticed. The difference is in the space variables only. If we replace x with y in Ixx of
Eq. (3.1.26), we could have Iyy of Eq. (3.1.27). This is easily anticipated from the
comparison of the transformed forms in Eqs. (3.1.10) and (3.1.12). In these
equations, each transform parameter corresponds to the original space variable. If
we exchange ξ with η in Eq. (3.1.10), it yields Eq. (3.1.12). Similarly, in the original
space we can exchange the space variables.
(4) Inversion of ~
I xy
The inversion technique for ~I xy is essentially the same as that for the former two

cases, but the integration formulas are slightly different. The Fourier inversion
integral with respect to the parameter η is reduced to the semi-infinite integral
Z1 Z1
Ixy
1 1 1 in g g
¼ ng 2 2 expðigyÞdg ¼ 2 sinðgyÞdg
2ps2 ad as ps a2d a2s
1 0
ð3:1:28Þ
The integration formula (Erdélyi 1954, vol. I, pp. 65, 15)
Z1
x p
sinðbxÞdx ¼ sgnðbÞ expðajbjÞ; ð3:1:29Þ
x2 þa 2 2
0
is applied to Eq. (3.1.28). It then follows that

Ixy
in
¼ 2 sgnðyÞ exp jyj n2 þ ðs=cd Þ2 exp jyj n2 þ ðs=cs Þ2 ð3:1:30Þ
2s
where sgnð:Þ is the sign function defined by

þ1 ; x [ 0
sgnðxÞ ¼ ð3:1:31Þ
1 ; x \ 0
The next Fourier inversion integral with respect to the parameter ξ

Zþ1
sgnðyÞ 1
Ixy ¼ ðþinÞ exp jyj n2 þ ðs=cd Þ2 exp jyj n2 þ ðs=cs Þ2 einx dn
2s2 2p
1
ð3:1:32Þ
is also reduced to the semi-infinite integral
Z1
sgnðyÞ
Ixy ¼ n exp jyj n2 þ ðs=cd Þ2 exp jyj n2 þ ðs=cs Þ2 sinðnxÞdn
2ps2
0
ð3:1:33Þ
Applying the integration formula (Erdélyi 1954, vol. I, pp. 75, 35)
a2 bc
x exp b x2 þ a2 sinðcxÞdx ¼ 2 K 2 a b2 þ c 2 ð3:1:34Þ
b þ c2
0
to the integral in Eq. (3.1.33), we have for I*xy

xy 1 1
Ixy ¼ K2 ð rs=c d Þ K 2 ð rs=c s Þ : ð3:1:35Þ
2pr 2 c2d c2s
Lastly, the Laplace inversion formula (Erdélyi 1954, vol. I, pp. 277, 12)

L1 ½K2 ðbsÞ ¼ Hðt bÞ pffiffiffiffiffiffiffiffiffiffiffiffiffiffi þ 2 t2 b2 ð3:1:36Þ
t 2 b2 b
is applied to Eq. (3.1.35). The final form of Ixy is given by

8 9
>
< 2 qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi>
=
xy ðr=cd Þ
Ixy ¼ þ q ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
ffi þ 2 t 2 ðr=c Þ2 Hðt r=c Þ
d d
2pr 4 >
: t2 ðr=cd Þ2 >
;
8 9 ð3:1:37Þ
>
< 2 qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi>
=
xy ðr=cs Þ
q ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi þ 2 t 2 ðr=c Þ2 Hðt r=c Þ:
s s
2pr 4 >
: t2 ðr=cs Þ2 >
;
(5) Green’s dyadic
We have obtained the exact expressions for the three fundamentals, Ixx, Ixy and
Iyy, and the unified expression for these is given by
( pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi)
Hðt r=cd Þ xi xj 1 2xi xj ðcd t2 Þ r 2
Iij ðx; y; tÞ ¼ þ pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi þ dij
2pcd r 2 ðcd t2 Þ r 2 r2 r2
8 qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi9
> >
Hðt r=cs Þ <xi xj 1 2xi xj ðcs tÞ2 r 2 =
q ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
ffi þ d ij
2pcs >
:r
2 r2 r2 >
;
ðcs tÞ2 r 2
ð3:1:38Þ
where the subscripts i and j stand for x and y, and it should be understood that
xx x; xy y. Further, δij is Kronecker’s delta defined by

1; i¼j
dij ¼ ð3:1:39Þ
0; i 6¼ j
Thus, we have just obtained the exact expressions for the four fundamental func-
tions. Substituting Eqs. (3.1.15) and (3.1.38) into Eq. (3.1.13), the displacement in
the actual space (x, y, t) is expressed as
2 8 qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi9
>
< >
X 6Hðt r=cd Þ xi xj 1 2xi xj ðcd tÞ2 r 2 =
ui ðx; y; tÞ ¼ Pj 4 qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi þ dij
2pcd >
:r
2 r2 r2 >
;
j¼x;y ðcd tÞ2 r 2
> >
Hðt r=cs Þ <xi xj 1 2xi xj ðcs tÞ2 r 2 =7
d ij q ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
ffi þ d ij 5
2pcs >
: r
2 r2 r2 >
;
ðcs tÞ2 r 2
ð3:1:40Þ
In this equation, the operation of Heaviside’s unit step function determines the
disturbed region. H(t − r/cd) shows a circular cylindrical region disturbed by the
dilatational wave which expands with velocity cd, while H(t − r/cs) does that by the
shear wave with cs. These two waves are the basic disturbances in the 2D dynamic
deformation and are shown in Fig. 3.1.
Now, we rewrite the above Eq. (3.1.40) as
X
ui ðx; y; tÞ ¼ Pj Gij ðx; y; tÞ ð3:1:41Þ
j¼x;y
Fig. 3.1 Radiation of P and y

SV waves from a point source
cd t
cs t
x
The function Gij(x, y, t) is called “Green’s dyadic” and it expresses the displacement
component in the i-axis direction due to the unit body force in the j-axis direction.
The explicit form of the dyadic Gij is given by
>
< >
Hðt r=cd Þ xi xj 1 2xi xj ðcd tÞ2 r 2 =
Gij ðx; y; tÞ ¼ q ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi þ dij
2pcd > r2
: r2 r2 >
;
ðcd tÞ2 r 2
>
< >
Hðt r=cs Þ xi xj 1 2xi xj ðcs tÞ2 r2 =
d ij q ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
ffi þ d ij
2pcs >
: r
2 r2 r2 >
;
ðcs tÞ2 r 2
ð3:1:42Þ
When the source is harmonically vibrating with frequency ω, the non-homogeneous

body force term in the displacement equation (3.1.1) is replaced with the time-
harmonic source, i.e.

@ @ux @uy @ 2 ux @ 2 ux 1 @ 2 ux Qx
ðc 1Þ
2
þ þ 2 þ 2 ¼ 2 2 2 dðxÞdðyÞ expðixtÞ
@x @x @y @x @y cs @t cs
ð3:2:1aÞ

@ @ux @uy @ 2 uy @ 2 uy 1 @ 2 uy Qy
ðc2 1Þ þ þ 2 þ 2 ¼ 2 2 2 dðxÞdðyÞ expðixtÞ
@y @x @y @x @y cs @t cs
ð3:2:1bÞ
where Qi is the magnitude of the source component.

It is possible to apply the integral transform method for obtaining the time-
harmonic Green’s dyadic. However, we do not employ the method here since the
cumbersome complex integration must be discussed as that for the time-harmonic
Green’s function in Sect. 2.5. We employ a simpler way, i.e. the convolution
integral. The time-harmonic response can be obtained by the convolution integral of
the impulsive response as
Zt
ðimpulseÞ
ui ðx; y; tÞ ¼ lim ui ðx; y; t0 Þ expfþixðt t0 Þgdt0 ð3:2:2Þ
t!1
0
ðimpulseÞ
where ui ðx; y; t0 Þ is the impulsive response given by Eq. (3.1.41) with
(3.1.42), where the source magnitude Pi for the impulsive solution is replaced with
Qi. In this convolution integral, the time-harmonic function, exp(+iωt), is excluded
from the limit, but we keep the upper limit of the integral be infinite, since the
steady-state response takes place long time after the initial disturbance. Then, the
convolution integral for the time-harmonic response takes the form
tZ
!1
ðimpulseÞ
ui ðx; y; tÞ ¼ expðþixtÞ ui ðx; y; t0 Þ expðixt0 Þdt0 ð3:2:3Þ
0
Substituting Eqs. (3.1.41) with (3.1.42) into Eq. (3.2.3), we have

2 8 9
xi xj R
1
>
> pexpðixt
0
Þ
ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi dt0 >
>
6 >
X 6 1 < r=cd ðcd t Þ r
r 2
0 2 2 >
=
ui ðx; y; tÞ ¼ expðþixtÞ Qj 6
62pcd > 2x x R1 q ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
> >
j¼x;y 4 > þ ri2 j dij r12
: ðcd tÞ2 r 2 expðixt0 Þdt0 >
>
;
r=cd
8 R1 expðixt0 Þ 0 93
>
>
xi xj
dij pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
ffi dt >
>
>
< r2 >
=7
1 r=cs
ðcs t0 Þ2 r 2 7
q ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
ffi 7
2pcs > R
1 > 7
>
> 2x x
ðcs tÞ r 2 expðixt0 Þdt0 >
2 > 5
: þ r2 dij r2
1
;
i j
r=cs
ð3:2:4Þ
Two integrals in the above equation are the integral representations of Hankel
function of the second kind (Watson 1966, p. 169). They are
Z1
expðixzÞ ip
pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi dx ¼ H0ð2Þ ðazÞ
x a
2 2 2
a
ð3:2:5Þ
ipa ð2Þ
x2 a2 expðixzÞdx ¼ þ H ðazÞ
2z 1
a
Applying these formulas to the integrals in Eq. (3.2.4), the time-harmonic response
is expressed as
X
ui ðx; y; tÞ ¼ Qj gij ðx; yÞ expðþixtÞ ð3:2:6Þ
j¼x;y
where gij(x, y) is the Green’s dyadic for the time-harmonic response and is given by

i 1 xi xj ð2Þ 2xi xj cd ð2Þ
gij ðx; yÞ ¼ H ðrx=c d Þ þ dij H ðrx=cd Þ
4 c2d r 2 0 r 2 rx 1

ð3:2:7Þ
1 xi xj ð2Þ 2xi xj cs ð2Þ
2 dij H0 ðrx=cs Þ þ dij 2 H ðrx=cs Þ
cs r2 r rx 1
Then, Eq. (3.2.6) with Eq. (3.2.7) gives the particular solution of the displacement
equation (3.2.1).
If the frequency is negative, ω → −ω( =ωe−πi), and the time factor is exp(−iωt),
we apply the formula (Watson 1966, p. 75)
Hnð2Þ ðxepi Þ ¼ ð1Þnþ1 Hnð1Þ ðxÞ ð3:2:8Þ
to the Hankel function in Eq. (3.2.7), and then the displacement is expressed as
X
ui ðx; y; tÞ ¼ Qj gij ðx; yÞ expðixtÞ ð3:2:9Þ
j¼x;y
and its dyadic gij (x, y) is given by

gij ðx; yÞ ¼ þ H ðrx=c d Þ þ d ij H ðrx=c d Þ
4 c2d r 2 0 r 2 rx 1

2 d ij H 0 ðrx=c s Þ þ d ij H ðrx=c s Þ
cs r2 r 2 rx 1
ð3:2:10Þ
n (.) is the n-th order Hankel function of the first kind and defined by
where H(1)
Hnð1Þ ðxÞ ¼ Jn ðxÞ þ iYn ðxÞ ð3:2:11Þ
If we consider to derive the static solution from the time-harmonic solution, we

have to take the limit ω → 0 in the dyadic given by Eq. (3.2.7) or (3.2.10).
However, the limit does not exist since the static displacement in the 2D plane
deformation does not vanish at infinity. We have to start from the original differ-
ential equation and solve it. The static source is a non-time-dependent point body
force and the displacement equation which has no inertia term is given by

@ @ux @uy @ 2 ux @ 2 ux Sx
ðc 1Þ
2
þ þ 2 þ 2 ¼ 2 dðxÞdðyÞ
@x @x @y @x @y cs
ð3:3:1Þ
@ @ux @uy @ uy @ uy
2 2
Sy
ðc2 1Þ þ þ 2 þ 2 ¼ 2 dðxÞdðyÞ
@y @x @y @x @y cs
where the static body force is placed at the coordinate origin and its magnitude is Si.
For these coupled differential equations, the convergence condition at infinity

@uj @uj
ffi
uj x2 þy2 !1 ¼ ffi ¼ @y pffiffiffiffiffiffiffiffiffi ¼ 0
@x pxffiffiffiffiffiffiffiffi
ð3:3:2Þ
2 þy2 !1 x2 þy2 !1
is assumed in order to apply the Fourier transform (but, the final solution does not
satisfy the first condition in the above equation). We should understand that the
convergence condition guarantees the application of the Fourier transform.
In order to obtain the static Green’s dyadic, we apply the double Fourier
transform with respect to two space variables. Applying the double Fourier trans-
form defined by Eq. (3.1.3) to the displacement equations (3.3.1), we have the
algebraic equations for the transformed displacement components ~uj ,

inðc2 1Þ in~ux ig~uy n2 þ g2 ~ux ¼ Sx =c2s
ð3:3:3Þ
igðc2 1Þ in~ux ig~uy n2 þ g2 ~uy ¼ Sy =c2s
and their solutions are

( )
Sx 1 c2 1 n2 Sy c 2 1 ng
~

ux ¼ 2
cs n þg
2 2 c ðn þ g Þ
2 2 2 2 cs c ðn þ g2 Þ2
2 2 2
( ) ð3:3:4Þ
Sx c 2 1 ng Sy 1 c2 1 g2
~

uy ¼ 2 þ
cs c2 ðn2 þ g2 Þ2 c2s n2 þ g2 c2 ðn2 þ g2 Þ2
Examining the above equations, we learn that four inversions are needed for the full
Fourier inversion since the displacement is given by

Sx ~ c2 1 ~ Sy c2 1 ~
~ux ¼ I 0 I xx 2 I xy
2
cs c 2 cs c2
ð3:3:5Þ
S c2 1 ~ Sy ~ c2 1 ~
~uy ¼ x I xy þ I 0 I yy
c2s c2 c2s c2
where the four fundamentals to be inverted are
~
I 0 ¼ 1 ~I ¼ n2 ~I ¼ g2 ~I ¼ ng
; xx ; yy ; xy
n þ g2
2
ðn2 þ g2 Þ2 ðn þ2
g2 Þ 2 ðn þ g2 Þ2
2
ð3:3:6Þ
The inversion for each fundamental is carried out in the following subsections.
(1) Inversion of ~
I 0
We have already inverted this function in Sect. 2.3. The result is given by
Eq. (2.3.21), i.e.
~
¼
/
S
) /¼
S
log ðx2 þ y2 Þ þ const: ð3:3:7Þ
n2 þ g2 4p
Applying this result, we have the inversion for ~I 0 as
1
I0 ¼ log ðx2 þ y2 Þ þ const: ð3:3:8Þ
4p
(2) Inversion of ~
I xx
The Fourier inversion integral with respect to the parameter η is reduced to the
Zþ1 Z1
Ixx ¼ 1 n2 n2 1
expðigyÞdg ¼ cosðgyÞdg ð3:3:9Þ
2p ðn2 þ g2 Þ2 p ðn þ g2 Þ2
2
1 0
The integration formula (Gradshteyn and Ryzhik 1980, pp. 449, 3.729 1)
Z1
1 p
cosðaxÞdx ¼ ð1 þ abÞ expðabÞ; a [ 0; b[0 ð3:3:10Þ
ðx2 þ b2 Þ 2 4b3
0
is applied to the far right integral in Eq. (3.3.9). We have for Ixx
Z1
Ixx ¼ n
2
1 1 jyj
cosðgyÞdg ¼ þ expðjnjjyjÞ ð3:3:11Þ
p ðn2 þ g2 Þ2 4jnj 4
0
The last Fourier inversion with respect to the parameter ξ is given by the sum of two
integrals,
Zþ1
1 1 jyj
Ixx ¼ þ expðjnjjyjÞ expðinxÞdn
2p 4jnj 4
1
ð3:3:12Þ
Z1 Z1
1 1 jyj
¼ expðnjyjÞ cosðnxÞdn þ expðnjyjÞ cosðnxÞdn
4p n 4p
0 0
Inspecting the two integrals in the last equation, the second integral can be eval-
uated by applying the formula (2.3.16) in Chap. 2. The first integral, however, has
no formula since its integrand has the first order singularity at ξ = 0 and it is
impossible to evaluate the integral in this form. We extract the first integral
Z1
ð1Þ 1 1
Ixx ¼ expðnjyjÞ cosðnxÞdn ð3:3:13Þ
4p n
0
and consider its derivative with respect to each space variable as
ð1Þ Z1
@Ixx 1
¼ expðnjyjÞ sinðnxÞdn
@x 4p
0
ð3:3:14Þ
ð1Þ Z1
@Ixx 1
¼ expðnjyjÞ cosðnxÞdn
@jyj 4p
0
These two integrals are easily evaluated by the formulas (2.3.16) to yield
ð1Þ ð1Þ
@Ixx 1 x @Ixx 1 jyj
¼ ; ¼ ð3:3:15Þ
@x 4p x2 þ y2 @jyj 4p x2 þ y2
Then, we return the two derivatives to the original one by the integration with
respect to each space variable. Two expressions for the single I(1)
xx are obtained as
Z
ð1Þ 1 x 1
Ixx ¼ dx ¼ logðx2 þ y2 Þ þ C1 ðyÞ
4p x2 þ y2 8p
Z
ð1Þ 1 jyj 1
Ixx ¼ djyj ¼ logðx2 þ y2 Þ þ C2 ðxÞ ð3:3:16Þ
4p x þ y
2 2 8p
These two equations must be equal. This condition is satisfied by setting

C1(y) = C2(x) and thus the two functions must be a pure constant,
C1 ðyÞ ¼ C2 ðxÞ ¼ const: ð3:3:17Þ
Then, we have for I(1)

xx ,
ð1Þ 1
Ixx ¼ logðx2 þ y2 Þ þ const: ð3:3:18Þ
8p
The second integral in Eq. (3.3.12) is easily evaluated by the formula (2.3.16)
and it yields
Z1
ð2Þ jyj 1 y2
Ixx ¼ expðnjyjÞ cosðnxÞdn ¼ ð3:3:19Þ
4p 4p x2 þ y2
0
Finally, substituting Eqs. (3.3.18) and (3.3.19) into (3.3.12), we have the inversion
1 1 y2
Ixx ¼ logðx2 þ y2 Þ þ þ const: ð3:3:20Þ
8p 4p x2 þ y2
Further, since the constant in the above equation is arbitrary, we can rewrite Ixx as
1 1 x2
Ixx ¼ logðx2 þ y2 Þ þ const: ð3:3:21Þ
8p 4p x2 þ y2
(3) Inversion of ~
I yy
Rewriting the transformed equation in Eq. (3.3.5), we learn that this inversion
can be decomposed into the sum of the former two functions, as
~I ¼ 1 n2
yy ¼ ~I 0 ~I xx ð3:3:22Þ
n2 þ g2 ðn2 þ g2 Þ2
Then, we can apply the former results, Eqs. (3.3.8) and (3.3.21), to the above
equation,
Iyy ¼ I0 Ixx
1 1 1 y2
¼ logðx2 þ y2 Þ þ const: þ logðx2 þ y2 Þ þ const:
4p 8p 4p x2 þ y2
ð3:3:23Þ
The final form of Iyy is given by
1 1 y2
Iyy ¼ logðx2 þ y2 Þ þ const: ð3:3:24Þ
8p 4p x2 þ y2
This can be derived from Ixx in Eq. (3.3.21) by the exchange of variables. Changing
x with y in Eq. (3.3.21), we have the inversion for Iyy.
(4) Inversion of ~
I xy
The Fourier inversion integral with respect to the parameter η is reduced to the
Zþ1 Z1
Ixy ¼ 1 ng in g
2
expðigyÞdg ¼ sinðgyÞdg ð3:3:25Þ
2p ðn þ g Þ
2 2 p ðn þ g2 Þ2
2
1 0
The integration formula (Gradshteyn and Ryzhik 1980, pp. 449, 3.729 2),
Z1
x pa
sinðaxÞdx ¼ expðabÞ ð3:3:26Þ
ðx2 þ b2 Þ 2 4b
0
is applied to the last integral in Eq. (3.3.25). We have
Ixy ¼ iy sgnðnÞ expðjnjjyjÞ ð3:3:27Þ

4
The next Fourier inversion integral with respect to the parameter ξ is reduced to the
simple integral,
Zþ1
iy 1
Ixy ¼ sgnðnÞ expðjnjjyjÞ expðinxÞdn
4 2p
1
ð3:3:28Þ
Z1
y
¼ expðnjyjÞ sinðnxÞdn
4p
0
Applying the formula (2.3.16), we have for Ixy,
1 xy
Ixy ¼ ð3:3:29Þ
4p x2 þ y2
(5) 2D Kelvin’s solution
The four inversions in Eq. (3.3.5) have thus been completed. Since the dis-
placement in the original space is given by

Sx c2 1 Sy c 2 1
ux ¼ 2 I 0 I xx Ixy
cs c2 c2s c2
ð3:3:30Þ
Sx c 2 1 Sy c2 1
uy ¼ 2 I xy þ I 0 Iyy
cs c2 c2s c2
we substitute I0 and Iij given by Eqs. (3.3.8), (3.3.21), (3.3.24) and (3.3.29) into the
above Eq. (3.3.30). The final form for the displacement is given by

x2
Sx Sy xy
ux ¼ 2
ðc þ 1Þ logðrÞ þ ðc 1Þ
2 2
þ 2
ðc2 1Þ 2
4pcd r 4pcd r

y2 ð3:3:31Þ
Sx xy S y
uy ¼ ðc2 1Þ 2 þ ðc2 þ 1Þ logðrÞ þ ðc2 1Þ
4pc2d r 4pc2d r
where cd = γcs defined by Eq. (3.8) is used. The constant term is omitted since it
gives a simple rigid motion with no strain. The reader should notice that the
constant term in the displacement breaks the applicability of the Fourier transform;
thus the last inversion integral with respect to the parameter ξ has the singular point.
The expression for the displacement is rewritten in terms of the dyadic,
ðstaticÞ ðstaticÞ
ux ¼ Sx gxx ðx; yÞ þ Sy gxy ðx; yÞ
ðstaticÞ ðstaticÞ
ð3:3:32Þ
uy ¼ Sx gxy ðx; yÞ þ Sy gyy ðx; yÞ
where the dyadic for the static source is given by

x2
1
gðstaticÞ
xx ðx; yÞ ¼ ðc 2
þ 1Þ logðrÞ þ ðc 2
1Þ
4pc2d r
1 xy
gðstaticÞ
xy ðx; yÞ ¼ ðc2 1Þ 2 ð3:3:33Þ
4pc2d r

y2
1
gðstaticÞ
yy ðx; yÞ ¼ ðc 2
1Þ
4pc2d r
This static Green’s dyadic for the plane deformation is called the “two dimensional
Kelvin’s solution.”
An impulsive point source placed at the coordinate origin is expressed by the body
force
0 1 0 1
Bx Px
@ By A ¼ @ Py AdðxÞdðyÞdðzÞdðtÞ; ð3:4:1Þ
Bz Pz
where Pi is the magnitude in the i-direction. The displacement equation with this
impulsive source is given by

@ @ux @uy @uz @ 2 ux @ 2 ux @ 2 ux 1 @ 2 ux Px
ðc2 1Þ þ þ þ 2 þ 2 þ 2 ¼ 2 2 2 dðxÞdðyÞdðzÞdðtÞ
@x @x @y @z @x @y @z cs @t cs

@ @ux @u y @uz @ 2
uy @ 2
uy @ 2
uy 1 @ 2
uy P y
@y @x @y @z @x @y @z cs @t cs

@ @ux @uy @uz @ 2 uz @ 2 uz @ 2 uz 1 @ 2 uz Pz
@z @x @y @z @x @y @z cs @t cs
ð3:4:2Þ
where cs is the shear wave velocity and the velocity ratio γ is defined by Eq. (3.8).
On these displacement equations, we impose the quiescent condition at an initial
time,
@ui
ui jt¼0 ¼ j ¼0 ð3:4:3Þ
@t t¼0
@ui pffiffiffiffiffiffiffiffiffiffiffiffiffiffi @ui pffiffiffiffiffiffiffiffiffiffiffiffiffiffi @ui pffiffiffiffiffiffiffiffiffiffiffiffiffiffi

ui jpxffiffiffiffiffiffiffiffiffiffiffiffiffiffi
2 þy2 þz2 !1 ¼ j ¼ j ¼ j ¼0
@x x2 þy2 þz2 !1 @y x2 þy2 þz2 !1 @z x2 þy2 þz2 !1
ð3:4:4Þ
Laplace transform with respect to the time variable,
Z1
uj ¼ uj expðstÞdt ð3:4:5Þ
0
and the triple Fourier transform with respect to three space variables,
Zþ1 Zþ1
1

uj ¼ uj expðþinxÞdx; uj ¼ uj expðinxÞdn
2p
1 1
Zþ1 Zþ1
1
~
uj ¼ uj expðþigyÞdy; uj ¼ ~uj expðigyÞdg ð3:4:6Þ
2p
1 1
Zþ1 Zþ1
1
^
uj ¼ uj expðþifzÞdz; uj ¼ ûj expðifzÞdf
2p
1 1
are applied to the displacement equations (3.4.2). The simple algebraic equations
for the transformed displacement are
2
2^ ^ ^ 2 ^ s ^ Px
ðc 1Þ n
2 ~ ~ ~
ux nguy nfuz ðn þ g þ f Þux ¼
2 2 ~ ~ux
cs c2s
2
s ^ Py
ux g2 ^~uy gf^~uz ðn2 þ g2 þ f2 Þ^~uy ¼
ðc2 1Þ ng^

~
~uy ð3:4:7Þ
cs c2s

s 2 ^ Pz
ðc2 1Þ nf^
ux gf^~uy f2 ^~uz ðn2 þ g2 þ f2 Þ^~uz ¼

~
~uz
cs c2s
and the displacement components in the transformed domain are given by

!
^ Px 1 1 1
~

ux ¼ n2 Px þ ngPy þ nfPz 2
2
c2s bs s b2d b2s
!
^ Py 1 1 1
~

uy ¼ ngPx þ g2 Py þ gfPz 2 ð3:4:8Þ
2
2
cs bs s b2d b2s
!
^ Pz 1 1 1
~

uz ¼ nfPx þ gfPy þ f2 Pz 2
2
c2s bs s b2d b2d
where the radicals are defined by

qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
bd ¼ n2 þ g2 þ f2 þ ðs=cd Þ2 ; bs ¼ n2 þ g2 þ f2 þ ðs=cs Þ2 ð3:4:9Þ
In order to explore the convenient way for the inversion of the transformed
displacement, we inspect the expression of Eq. (3.4.8). It is found that seven
inversion formulas are necessary.
They are
^~ 1
I 0 ðn; g; 1; sÞ ¼ ; ð3:4:10Þ
c2s b2s
! !
2
^
I 11 ðn; g; f; sÞ ¼ n
~ 1 1 ^~ g2 1 1
; I 12 ðn; g; f; sÞ ¼ 2 ;
s2 b2d b2s s b2d b2s
! ð3:4:11Þ
2
^
~
I 13 ðn; g; f; sÞ ¼ f 1 1

s2 b2d b2s
! !
^
~
I 21 ðn; g; f; sÞ ¼ ng 1 1 ^~ gf 1 1
; I 22 ðn; g; f; sÞ ¼ 2 ;
s2 b2d b2s s b2d b2s
! ð3:4:12Þ
^
~I ðn; g; f; sÞ ¼
nf 1 1
23
s2 b2d b2s
^
The first inversion for ~I 0 is the same as that of the 3D wave equation in Sect. 2.7.
Applying that result to our inversion, we have
1
I0 ðx; y; z; tÞ ¼ dðt R=cs Þ ð3:4:13Þ
4pc2s R
where the 3D radial distance is defined by

R¼ x2 þ y2 þ z 2 ð3:4:14Þ
^
Subsequently, we consider the inversion of ~I 1j in Eq. (3.4.11). The three Fourier
transforms are the same in the form of definition integral and thus the exchange of
two space variables is equivalent to the exchange of two integration parameters.
That is
^
I 12 ðn; g; f; sÞ ¼ ^~I 11 ðg; n; f; sÞ
~
) I12 ðx; y; z; tÞ ¼ I11 ðy; x; z; tÞ
ð3:4:15Þ
^
I 13 ðn; g; f; sÞ ¼ ^~I 11 ðf; g; n; sÞ
~
) I13 ðx; y; z; tÞ ¼ I11 ðz; y; x; tÞ
^
Due to this exchangeability, if we could obtain the inversion for ~I 11 as F(x, y, z, t), the
^
inversion of ~
I is given by the exchange of the space variables as F(y, x, z, t) and the
12
^
inversion of ~
I 13 is also given by the variable exchange as F(z, y, x, t). Thus, we can
draw a schematic inversion diagram as
!
2
^
~
I 11 ðn; g; f; sÞ ¼ n 1 1
2
2 ) I11 ðx; y; z; tÞ ¼ Fðx; y; z; tÞ
s2 bd bs
# exchange n with g # exchange x with y ð3:4:16Þ
!
^
~
I 12 ðn; g; f; sÞ ¼ g 2
1 1
) I12 ðx; y; z; tÞ ¼ Fðy; x; z; tÞ
s2 b2d b2s
^
Similarly, the third inversion ~I 13 is also obtained by the exchange of the space
variables as
!
2
^
~
I 11 ðn; g; f; sÞ ¼ n 1 1
2
2 ) I11 ðx; y; z; tÞ ¼ Fðx; y; z; tÞ
s2 bd bs
# exchange n with f # exchange x with z ð3:4:17Þ
!
2
^
~
I 13 ðn; g; f; sÞ ¼ f 1 1
2
2 ) I13 ðx; y; z; tÞ ¼ Fðz; y; x; tÞ
s2 bd bs
Then, the necessary inversion formula is reduced to only one, that is

!
2
^~ n 1 1
I 11 ðn; g; f; sÞ ¼ 2 ð3:4:18Þ
s b2d b2s
As for the second group of the inversion in Eq. (3.4.12), we learn that the
necessary inversion formula is
!
^~ ng 1 1
I 21 ðn; g; f; sÞ ¼ 2 ð3:4:19Þ
s b2d b2s
If we could get the inversion as G(x, y, z, t), the other two are given by the exchange
of the space variables. They are
!
^
I 21 ðn; g; f; sÞ ¼ ng 1 1
~ ) I21 ðx; y; z; tÞ ¼ Gðx; y; z; tÞ
s2 b2d b2s
# exchange n with f # exchange x with z ð3:4:20Þ
!
^
I 22 ðn; g; f; sÞ ¼ gf 1 1
~ ) I22 ðx; y; z; tÞ ¼ Gðz; y; x; tÞ
s2 b2d b2s
and
!
^
I 21 ðn; g; f; sÞ ¼ ng 1 1
~ ) I21 ðx; y; z; tÞ ¼ Gðx; y; z; tÞ
s2 b2d b2s
# exchange g with f # exchange y with z ð3:4:21Þ
!
^
I 23 ðn; g; f; sÞ ¼ nf 1 1
~ ) I23 ðx; y; z; tÞ ¼ Gðx; z; y; tÞ
s2 b2d b2s
^ ^
Consequently, it is sufficient to develop the inversion formulas only for ~I 11 and ~I 21 .
^
(1) Inversion of ~I 11
The formal Fourier inversion integral with respect to the parameter f is given by
Zþ1 !
I 11 ðn; g; z; sÞ ¼ 1
~ n2 1 1
expðifzÞdf
2p s2 b2d b2s
1
! ð3:4:22Þ
2 Z1
n 1 1
¼ 2 cosðfzÞdf
ps f2 þ a2d f2 þ a2s
0
where
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
aj ¼ n2 þ g2 þ ðs=cj Þ2 ; j ¼ d; s ð3:4:23Þ
Applying the simple integration formula (2.1.22) to the integral in Eq. (3.4.22), we
have
2

I 11 ðn; g; z; sÞ ¼ n
~ 1
expða d jzjÞ
1
expða s jzjÞ ð3:4:24Þ
2s2 ad as
The second inversion integral with respect to the parameter η is given by
Zþ1

I11
1 n2 1 1
ðn; y; z; sÞ ¼ 2
expðad jzjÞ expðas jzjÞ expðigyÞdg
2p 2s ad as
1
2
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi3
Z1 exp jzj n2 þ g2 þ ðs=cd Þ2 exp jzj n2 þ g2 þ ðs=cs Þ2
n 2 6 7
¼ 6 qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi 7 cosðgyÞdg
2ps2 4 2 2
5
0 n 2
þ g 2 þ ðs=c Þ
d n 2
þ g 2 þ ðs=c Þ
s
ð3:4:25Þ
The integration formula (2.4.13) in Chap. 2,
1
x 2 þ a2
0
is applied to the integral in Eq. (3.4.25). We have for I11

qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
I11
n2
ðn; y; z; sÞ ¼ K0 r n þ ðs=cd Þ K0 r n2 þ ðs=cs Þ2
2 2
ð3:4:27Þ
2ps2
where K0(.) is the zeroth order modified Bessel function of the second kind and
r ¼ y2 þ z 2 ð3:4:28Þ
The last Fourier inversion integral with respect to the parameter ξ is given by
Zþ1
1 n2
I11 ðx; y; z; sÞ ¼ K 0
r n2
þ ðs=c d Þ 2
K 0
r n2 þ ðs=cs Þ2 expðinxÞdn
2p 2ps2
1
Z1
1
¼ 2 2 n2 K0 r n2 þ ðs=cd Þ2 K0 r n2 þ ðs=cs Þ2 cosðnxÞdn
2p s
0
ð3:4:29Þ
The last integral, which includes the modified Bessel function, has not yet been
tabulated in reference books. However, we have one formula which resembles our
integral, i.e.x
Z1 pffiffiffiffiffiffiffiffiffiffiffiffiffiffi pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
p exp c a2 þ b2
I¼ K0 a x þ c cosðbxÞdx ¼
2 2 pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi ð3:4:30Þ
2 a2 þ b2
0
*Notice: the above equation is just the Fourier cosine inversion of Eq. (3.4.26)!
If we differentiate this integration formula with respect to the parameter “b”
twice, it yields to our necessary formula,
Z1 ( pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi )
@2I pffiffiffiffiffiffiffiffiffiffiffiffiffiffi p @2 exp c a2 þ b2
2¼ x K0 a x2 þ c2 cosðbxÞdx ¼
2
@b 2 @b2 a2 þ b2
0
ð3:4:31Þ
That is,
Z1 pffiffiffiffiffiffiffiffiffiffiffiffiffiffi
x2 K0 a x2 þ c2 cosðbxÞdx
0

p 1 1 3b2 b2 c 2
¼ pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi þ c 1 2 pffiffiffiffiffiffiffiffiffiffiffiffiffiffi
ffi exp c a2 þ b2
2 a2 þ b2 a þ b2
2 a þ b2 a2 þ b2
ð3:4:32Þ
Applying this new formula (3.4.32) to the integral in Eq. (3.4.29), we have

1 3x2 1 3x2 1 x2
I11 ðx; y; z; sÞ ¼ 1 2 þ 1 expfsðR=cd Þg
4pR2 R Rs2 R2 cd s c2d R

3x2 1 3x2 1 x2
1 2 þ 1 expfsðR=c s Þg
R Rs2 R2 cs s c2s R
ð3:4:33Þ
where the 3D radial distance R is defined by Eq. (3.4.14).

The last Laplace inversion is carried out by applying the inversion formulas,

1 1; t [ a
L1 expðasÞ ¼ Hðt aÞ ¼ ð3:4:35Þ
s 0; t \ a

1 t a; t[a
L1 expðasÞ ¼ Hðt aÞðt aÞ ¼ ð3:4:36Þ
s2 0; t\a
where δ(.) and H(.) are Dirac’s delta and Heaviside’s unit step functions defined in
Sect. 1.2. Finally, the application of Laplace inversion formulas leads to the explicit
expression for the inversion:

1 x2 x2
I11 ðx; y; z; tÞ ¼ 2 2 dðt R=cd Þ þ 2 2 dðt R=cs Þ
4pR cd R cs R
2
ð3:4:37Þ
3x t
þ 1 2 Hðt R=c d ÞHðR=c s tÞ
R R2
^ ^
The other inversions for ~I 12 and ~I 13 can be obtained by the simple exchange of the
space variables as shown in Eqs. (3.4.16) and (3.4.17). They are

1 y2 y2
4pR cd R cs R
2
ð3:4:38Þ
3y t
þ 1 2 Hðt R=c d ÞHðR=c s tÞ
R R2

1 z2 z2
4pR cd R cs R
2
ð3:4:39Þ
3z t
þ 1 2 Hðt R=cd ÞHðR=cs tÞ
R R2
^
(2) Inversion of ~I 21
The Fourier inversion integral with respect to the parameter ζ
Zþ1 !
I 21 ðn; g; z; sÞ ¼ 1
~ ng 1 1
expðifzÞdf
2p s2 b2d b2s
1
! ð3:4:40Þ
Z1
ng 1 1
¼ 2 2 cosðfzÞdf
ps f þ ad f þ a2s
2 2
0
is evaluated by applying the formula (2.1.22) in Chap. 2. It is

I 21 ðn; g; z; sÞ ¼ ng 1 expðad jzjÞ 1 expðas jzjÞ
~ ð3:4:41Þ
2s2 ad as
The second inversion integral with respect to the parameter η is given by
Zþ1

I21
1 ng 1 1
ðn; y; z; sÞ ¼ 2
expðad jzjÞ expðas jzjÞ expðigyÞdg
2p 2s ad as
1
Z1 ( qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
)
in g g
¼ pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi exp jzj g2 þ c2d pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi exp jzj g2 þ c2s sinðgyÞdg
2ps2 g2 þ c2d g2 þ c2s
0
ð3:4:42Þ
where
cj ¼ n2 þ ðs=cj Þ2 ; j ¼ d; s ð3:4:43Þ
We apply the integration formula (Erdélyi 1954, vol. I, pp. 75, 36),
x ab
pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi exp c x2 þ a2 sinðbxÞdx ¼ pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi K1 a b2 þ c2
x 2 þ a2 b2 þ c 2
0
ð3:4:44Þ
to the integral in Eq. (3.4.42). It follows that

qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
I21
y
ðn; y; z; sÞ ¼ in n2 þ ðs=cd Þ2 K1 r n2 þ ðs=cd Þ2
2pr s2
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi ð3:4:45Þ
þ in n2 þ ðs=cs Þ2 K1 r n2 þ ðs=cs Þ2
where r has already been defined by Eq. (3.4.28). The last Fourier inversion integral
with respect to the parameter ξ is given by
Zþ1
y 1
I21 ðn; y; z; sÞ ¼ in n2 þ ðs=cd Þ2 K1 r n2 þ ðs=cd Þ2
2pr s2 2p
1
þ in n2 þ ðs=cs Þ2 K1 r n2 þ ðs=cs Þ2 expðinxÞdn
Z1
y
¼ 2 2 n n2 þ ðs=cd Þ2 K1 r n2 þ ðs=cd Þ2
2p r s
0
n n2 þ ðs=cs Þ2 K1 r n2 þ ðs=cs Þ2 sinðnxÞdn
ð3:4:46Þ
We have the formula (Erdélyi 1954, vol. I, pp. 113, 45)
x x2 þ b2 K1 a x2 þ b2 sinðcxÞdx
0 ð3:4:47Þ
2

p ab c 3 3
¼ 1 þ pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi þ 2 2 exp b a2 þ c2
2 ða2 þ c2 Þ3=2 b a2 þ c2 b ða þ c2 Þ
The integrals in Eq. (3.4.46) are evaluated and we have

xy 1 3cd 3c2d
I21 ðx; y; z; sÞ ¼ 1þ þ expfsðR=cd Þg
4pR3 c2d sR s2 R2

ð3:4:48Þ
1 3cs 3c2s
2 1þ þ 2 2 expfsðR=cs Þg
cs sR s R
The Laplace inversion is carried out by applying the formulas (3.4.34)–(3.4.36).

Then, the final form of I21(x, y, z, t) is given by

xy 1 1 3t
I21 ðx; y; z; tÞ ¼ dðt R=cd Þ 2 dðt R=cs Þ þ 2 Hðt R=cd ÞHðR=cs tÞ
4pR3 c2d cs R
ð3:4:49Þ
We have just obtained the exact expressions for the two principal parts: I11 and
I21. Other two inversions can be obtained by the exchange of the space variables as
shown in Eqs. (3.4.20) and (3.4.21). They are

yz 1 1 3t
I22 ðx; y; z; tÞ ¼ dðt R=cd Þ 2 dðt R=cs Þ þ 2 Hðt R=cd ÞHðR=cs tÞ
4pR3 c2d cs R
ð3:4:50Þ

xz 1 1 3t
I23 ðx; y; z; tÞ ¼ dðt R=cd Þ dðt R=cs Þ þ Hðt R=cd ÞHðR=cs tÞ
4pR3 c2d c2s R2
ð3:4:51Þ

We have just obtained the exact inversion of the fundamentals. Inspecting
Eq. (3.4.8), we learn that the displacement components can be expressed in terms of
Iij(x, y, z, t) as
ux ¼ Px fI0 ðx; y; z; tÞ I11 ðx; y; z; tÞg þ Py fI21 ðx; y; z; tÞg þ Pz fI23 ðx; y; z; tÞg
uy ¼ Px fI21 ðx; y; z; tÞg þ Py fI0 ðx; y; z; tÞ I12 ðx; y; z; tÞg þ Pz fI22 ðx; y; z; tÞg
uz ¼ Px fI23 ðx; y; z; tÞg þ Py fI22 ðx; y; z; tÞg þ Pz fI0 ðx; y; z; tÞ I13 ðx; y; z; tÞg
ð3:4:52Þ
Substituting Eqs. (3.4.13), (3.4.37–3.4.39) and (3.4.49–3.4.51) into Eq. (3.4.52), we

have the following explicit expressions for the displacement components:
8 2 9
> x 1 x2 >
>
> 2 dðt R=c Þ þ 1 dðt R=c s >
Þ >
Px < cd R2 =
d
c2s R2
ux ¼
4pR >
> t 3x2 >
>
>
: 1 2 Hðt R=cd ÞHðR=cs tÞ > ;
R 2 R

Py xy xy 3xyt
þ 2
dðt R=c d Þ dðt R=c s Þ þ Hðt R=c d ÞHðR=c s tÞ
4pR cd R2 c2s R2 R4

Pz xz xz 3xzt
þ dðt R=c d Þ dðt R=c s Þ þ Hðt R=cd ÞHðR=c s tÞ
4pR c2d R2 c2s R2 R4
ð3:4:53aÞ

Px xy xy 3xyt
uy ¼ dðt R=c d Þ dðt R=c s Þ þ Hðt R=cd ÞHðR=c s tÞ
8 2 2
9
>
>
y
dðt R=cd Þ þ 2 1 2 dðt R=cs Þ >
1 y
>
>
< 2 R2 >
=
Py cd c s R
þ
4pR >
> t 3y2 >
>
>
: 1 2 Hðt R=cd ÞHðR=cs tÞ > ;
R2 R

Pz yz yz 3yzt
þ dðt R=c d Þ dðt R=c s Þ þ Hðt R=c d ÞHðR=c s tÞ
ð3:4:53bÞ

Px xz xz 3xzt
uz ¼ dðt R=cd Þ dðt R=c s Þ þ Hðt R=cd ÞHðR=c s tÞ

Py yz yz 3yzt
þ dðt R=c d Þ dðt R=c s Þ þ Hðt R=c d ÞHðR=cs tÞ
8 2 2
9
>
>
z 1 z >
> 2 dðt R=c Þ þ 1 dðt R=c s >
Þ >
Pz < cd R2 =
d
c2s R2
þ
4pR >
> >
> t 1 3z Hðt R=c ÞHðR=c tÞ >
2
: >
;
2 2 d s
R R
ð3:4:53cÞ
In terms of the Green’s dyadic, we can rewrite the displacement components as

X
ui ¼ Pj Gij ðx; y; z; tÞ; i ¼ x; y; z ð3:4:54Þ
j¼x;y;z
where the dyadic Gij ðx; y; z; tÞ is given by

8 xx 1 xi xj 9
>
>
i j
dðt R=c Þ þ d dðt R=c Þ
s >
>
1 < c2d R2 =
d ij
c2s R2
Gij ðx; y; z; tÞ ¼ ð3:4:55Þ
>
4pR > t 3xi xj >
>
: d ij Hðt R=c d ÞHðR=c s tÞ ;
R2 R2
It is clear that there are two spherical waves. Their fronts are given by
cdt = R and cst = R for dilatational and shear waves, respectively. We also find that
the displacement has the singularity of the delta function (i.e. the first order sin-
gularity) at their wave fronts.
When the source is a time-harmonic vibration, the body force with magnitude Qi is
given by
0 1 0 1
Bx Qx
@ By A ¼ @ Qy AdðxÞdðyÞdðzÞ expðþixtÞ; ð3:5:1Þ
Bz Qz
and the displacement equation is

@ @ux @uy @uz @ 2 ux @ 2 ux @ 2 ux
ðc 1Þ
2
þ þ þ 2 þ 2 þ 2
@x @x @y @z @x @y @z
1 @ 2 ux Qx
¼ 2 2 2 dðxÞdðyÞdðzÞ expðixtÞ
cs @t cs

@ @ux @uy @uz @ 2 uy @ 2 uy @ 2 uy
ðc2 1Þ þ þ þ 2 þ 2 þ 2
@y @x @y @z @x @y @z
ð3:5:2Þ
1 @ uy Qy
2
cs @t cs

@ @ux @uy @uz @ 2 uz @ 2 uz @ 2 uz
ðc2 1Þ þ þ þ 2 þ 2 þ 2
@z @x @y @z @x @y @z
1 @ uz Qz
2
cs @t cs
In this section we do not employ the method of the integral transform for solving
the displacement equations. Instead, we take a short-cut. That is the convolution
integral with the impulsive Green’s dyadicobtained in the previous section. When
we express the displacement in terms of the time-harmonic Green’s dyadic,
X
ui ¼ Qj gij ðx; y; z; tÞ; i ¼ x; y; z ð3:5:3Þ
j¼x;y;z
the dyadic for the time-harmonic response can be derived by the convolution
integral of the impulsive dyadic given by Eq. (3.4.55). The convolution integral is
evaluated as follows
Zt
gij ðx; y; z; tÞ ¼ lim Gij ðx; y; z; t0 Þ expfþixðt t0 Þgdt0
t!1
0
tZ
!1
ð3:5:4Þ
¼ expðþixtÞ Gij ðx; y; z; t0 Þ expðixt0 Þdt0
0
Substituting Eq. (3.4.55) into the last integral in the above equation, we can easily
perform the integration and obtain the time-harmonic Green’s dyadic gij as
gij ðx; y; z; tÞ
( )
1 xi xj 1 xR 3xi xj
¼ þ 1þi dij 2 expfixðt R=cd Þg
4pR c2d R2 ðxRÞ2 cd R
( )
1 1 xi xj 1 xR 3xi xj
þ dij 2 1þi dij 2 expfixðt R=cs Þg
4pR c2s R ðxRÞ2 cs R
ð3:5:5Þ
When a static point force is placed at the coordinate origin, the body force with
magnitude Si is given by
0 1 0 1
Bx Sx
@ By A ¼ @ Sy AdðxÞdðyÞdðzÞ ð3:6:1Þ
Bz Sz
The static deformation can be considered as the time-harmonic response with zero-
frequency. Taking the limit ω → 0 in Eq. (3.5.4), the displacement can be
expressed in terms of the static dyadic as
X ðKÞ
ui ¼ Sj Gij ðx; y; zÞ; i ¼ x; y; z ð3:6:2Þ
j¼x;y;z
ðKÞ
where the source magnitude Qi is replaced with Si and the static dyadic Gij ðx; y; zÞ
is derived from the limit and is given by
ðKÞ 1 n 2 xi xj o
Gij ðx; y; zÞ ¼ lim gij ðx; y; z; tÞ ¼ 2
ðc þ 1Þdij þ ðc2 1Þ 2 ð3:6:3Þ
x!0 8pcd R R
Since the velocity ratio can be replaced with the simple function of Poisson ratio ν
as
2ð1 mÞ
c2 ¼ ð3:6:4Þ
1 2m
the dyadic can also be rewritten as
1 n xi xj o
ðKÞ
Gij ðx; y; zÞ ¼ ð3 4mÞd ij þ ð3:6:5Þ
8pð1 2mÞc2d R R2
This dyadic is called “Kelvin’s solution” for the static deformation.
3.7 Torsional Source
This section shows two impulsive Green’s functions for the axisymmetric torsion
problem. The first one is corresponding to a circular ring force, and the second to a
point torque. We take the cylindrical coordinate system (r, θ, z) in an infinite elastic
solid and assume that the symmetry axis of the deformation/torsion is the z-axis.
The ring force lies on the z = 0 plane and its symmetry axis is also the z-axis. On the
other hand, the point torque is place on the coordinate origin. Both produce the
axisymmetric torsional deformation.
Due to the axisymmetric nature of the deformation, the non-vanishing dis-
placement component is the circumferential displacement uθ only, and the gov-
erning equation for the axisymmetric torsional deformation is given by
@ 2 uh 1 @uh uh @ 2 uh 1 @ 2 uh 1
þ 2 þ 2 2 2 ¼ 2 Bh ðr; z; tÞ ð3:7:1Þ
@r 2 r @r r @z cs @t cs
where cs is the shear (torsional) wave velocity and Bθ is the circumferential body
force which represents the ring or point source. When the source is an impulsive
ring force with radius a, the body force is defined by
dðr aÞ
Bh ðr; z; tÞ ¼ Q dðzÞdðtÞ ð3:7:2Þ
r
where δ(·) is Dirac’s delta function and Q is the magnitude of the source. On the
other hand, when the source is an impulsive point torque, the body force is defined
by the limit of the ring force as

T dðr aÞ
Bh ðr; z; tÞ ¼ lim dðzÞdðtÞ ð3:7:3Þ
2pq a!0 ar
where T is the magnitude of the torque. As we have no suitable mathematical

formula for the limit in the square bracket, the limiting form for the point torque
source is retained. In this section, we shall discuss Green’s functions for these two
sources, separately.
3.7.1 Ring Source
In order to solve the differential equation (3.7.1) with the source (3.7.2), we apply
the triple integral transform which is defined in Chap. 1: Hankel transform with
respect to the radial variable r, Fourier transform with respect to the axial variable
z and Laplace transform with respect to the time t. The triple transform is expressed
as
8 2 9
>
> @ u 1 @u u @ 2
u 1 @ 2
uh>
2þ 2 2 2 >
h h h h
Z1 Z1 > 2 þ
Z1 < >
@r r @r r @z cs @t =
expðstÞdt expðþifzÞdz rJ ðnrÞdr
>
> 1 dðr aÞ > 1
0 1 0 >: ¼ 2Q dðzÞdðtÞ >
>
;
cs r
ð3:7:4Þ
The convergence condition at infinity

ffi @uh @uh
uh j r2 þz2 !1 ¼ ffi ¼ @z pffiffiffiffiffiffiffiffi ffi ¼0 ð3:7:5Þ
@r pffiffiffiffiffiffiffiffi
r2 þz2 !1 r 2 þz2 !1
3.7 Torsional Source 111
and the quiescent condition at the initial time

@uh
uh jt¼0 ¼ ¼0 ð3:7:6Þ
@t t¼0
are employed, and the transformed displacement is defined as
Z1 Z1 Z1
^
uh ¼
~ expðstÞdt expðþifzÞdz ruh J1 ðnrÞdr ð3:7:7Þ
0 1 0
The application of the triple integral transform yields to the simple algebraic
equation for the transformed displacement,
n o Q
n2 þ f2 þ ðs=cs Þ2 ^~uh ¼ 2 J1 ðnaÞ ð3:7:8Þ
cs
Then, we consider the inversion of the displacement,
^~u ¼ Q J1 ðnaÞ
h ð3:7:9Þ
c2s n2 þ f2 þ ðs=cs Þ2
As the first step, the Fourier inversion integral defined by Eq. (1.1.21c) in Chap. 1 is
applied and it is reduced to the semi-infinite integral as
Z1
Q 1
~uh ¼ 2 J1 ðnaÞ cosðfzÞdf ð3:7:10Þ
pcs n þ f þ ðs=cs Þ2
2 2
0
The simple integration formula (2.1.22) in Chap. 2 is applied to the above integral.
It follows that

Q J1 ðnaÞ
uh ¼
~ q ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi exp j z j n2 þ ðs=cs Þ2 ð3:7:11Þ
2c2s 2
n þ ðs=cs Þ
2
The second step is to apply the Hankel inversion with n = 1 defined by Eq. (1.1.19)
in Chap. 1. The Hankel inversion integral is given by
Z1
Q n
uh ¼ 2 qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi exp jzj n2 þ ðs=cs Þ2 J1 ðnaÞJ1 ðnrÞdn ð3:7:12Þ
2cs
0 n2 þ ðs=cs Þ2
The product of two Bessel functions is replaced with the integral form of the single
Bessel function (Watson 1966, p. 361),
Zp pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
1
J1 ðnaÞJ1 ðnrÞ ¼ J0 n r 2 þ a2 2ar cos u cos udu ð3:7:13Þ
p
0
and the order of integration is exchanged. Equation (3.7.12) yields
Zp Z1
Q n
uh ¼ cos udu qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi exp jzj n2 þ ðs=cs Þ2 J0 ðnZÞdn
2pc2s
0 0 n2 þ ðs=cs Þ2
ð3:7:14Þ
where
pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
Z¼ r 2 þ a2 2ar cos u ð3:7:15Þ
Fortunately, we have the nice integration formula (Erdélyi 1954, vol. II, pp. 9, 24)
for the inner integral, i.e.
Z1
n
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi exp jzj n2 þ ðs=cs Þ2 J0 ðnZÞdn
0 n2 þ ðs=cs Þ2 ð3:7:16Þ
1 n pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffio
¼ pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi exp ðs=cs Þ Z 2 þ z2
Z 2 þ z2
Applying this integration formula to Eq. (3.7.14), and recalling the definition of Z,
we have the single finite integral for the Laplace transformed displacement,
n pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffio
Zp exp ðs=cs Þ r 2 þ a2 2ar cos u þ z2
Q
uh ¼ pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi cos udu ð3:7:17Þ
2pc2s r 2 þ a2 2ar cos u þ z2
0
Before going to the Laplace inversion, the integral is simplified by the change of
variable, φ → u,
1 pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
ffi
u¼ r 2 þ a2 2ar cos u þ z2 ð3:7:18Þ
cs
Equation (3.7.17) is simplified as
Z2 =cs
R (sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffisffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi)
Q R22 ðcs uÞ2 ðcs uÞ2 R21
uh ¼ 2
expðsuÞdu ð3:7:19Þ
4parcs ðcs uÞ R21 R22 ðcs uÞ2
R1 =cs
where
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
R1 ¼ ðr aÞ2 þ z2 ; R2 ¼ ðr þ aÞ2 þ z2 ð3:7:20Þ
Examining the integrand, we learn that the Laplace transform parameter s is

included only in the argument of the exponential function. So, the simple Laplace
inversion formula for the delta function
is applied to Eq. (3.7.19). It yields
Z2 =cs
R (sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi)
Q R22 ðcs uÞ2 ðcs uÞ2 R21
uh ¼ dðt uÞdu ð3:7:22Þ
4parcs ðcs uÞ2 R21 R22 ðcs uÞ2
R1 =cs
In order to evaluate the integral, we examine the supporting region for the delta
function and applied the simple formula (1.2.3) in Chap. 1,
Zb
f ðcÞ; a\c\b
0; c \ a or b \ c
a
We have the simple form for the torsional displacement, i.e. Green’s function for
the torsional deformation,
(sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi)
Q R22 ðcs tÞ2 ðcs tÞ2 R21
uh ¼ HðR2 cs tÞHðcs t R1 Þ 2
ð3:7:24Þ
4pcs ar ðcs tÞ R21 R22 ðcs tÞ2
where H(.) is Heaviside’s unit step function defined in Sect. 1.2.
3.7.2 Point Torque Source
This subsection considers the Green’s function corresponding to the point torque
source defined by Eq. (3.7.3). As was employed in the previous subsection, we also
apply the triple integral transform to the governing Eq. (3.7.1) with the point torque
source,
8 2 9
Z1 Z1 Z1 >< @r2h þ 1r @rh r2h þ @z2h c12s @t2h >
@ u @u u @2 u @2u
=

expðstÞdt expðþifzÞdz rJ1 ðnrÞdr
>
: ¼ 2pl
T
lim dðraÞ dðzÞdðtÞ > ;
0 1 0 ar a!0
ð3:7:25Þ
The limit in the source term is taken after the Hankel transform. Its procedure is
Z1 Z1
dðr aÞ dðr aÞ J1 ðnaÞ n
lim rJ1 ðnrÞdr ¼ lim rJ1 ðnrÞdr ¼ lim ¼
a!0 ar a!0 ar a!0 a 2
0 0
ð3:7:26Þ
The transformed governing equation yields to the simple algebraic equation,

n o T
n2 þ f2 þ ðs=cs Þ2 ^~uh ¼ n ð3:7:27Þ
4pl
Thus, the triple transformed displacement is given by
^~u ¼ T n
h ð3:7:28Þ
4pl n þ f þ ðs=cs Þ2
2 2
Now, we consider the inversion. As was done for the inversion integral of
Eq. (3.7.10), the Fourier inversion integral with respect to the parameter ζ is easily
evaluated as

T n
uh ¼
~ qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi exp jzj n2 þ ðs=cs Þ2 ð3:7:29Þ
8pl
n2 þ ðs=cs Þ2
The second inversion, i.e. Hankel inversion integral, is separated into two terms as
Z1
T n2
uh ¼ qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi exp jzj n2 þ ðs=cs Þ2 J1 ðnrÞdn
8pl
0 n2 þ ðs=cs Þ2
8 9
Z1 >
=
T ðs=cs Þ
¼ n þ ðs=cs Þ qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi exp jzj n2 þ ðs=cs Þ2 J1 ðnrÞdn
2 2
8pl > : >
0 n2 þ ðs=cs Þ2 ;
ð3:7:30Þ
The second term in the last line can be evaluated by applying the formula (Erdélyi
1954, vol. II, pp. 19, 10)
Z1
1
pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi exp a n2 þ b2 J1 ðnyÞdn
n 2 þ b2
1 n pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffio
0
¼ expðabÞ exp b a2 þ y2 ð3:7:31Þ

by
As to the first term in the Hankel inversion integral (3.7.30), we derive a suitable
integration formula from the formula (3.7.31). Differentiating Eq. (3.7.31) twice
with respect to the parameter α, we have
n2 þ b2 exp a n2 þ b2 J1 ðnyÞdn
0
!
b 1 y a2 b pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
¼ expðabÞ þ pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi 2 p ffiffiffiffiffiffiffiffiffiffiffiffiffiffi
ffi exp b a2 þ y2
y a 2 þ y 2 a þ y 2 y a2 þ y 2
ð3:7:32Þ
Applying these two formulas to Eq. (3.7.30), we have the Laplace transformed
displacement as

T r s 1
uh ¼ þ expðsR=cs Þ ð3:7:33Þ
8pl R2 cs R
where R is the distance from the source point (coordinate origin),

R¼ r 2 þ z2 ð3:7:34Þ
The last Laplace inversion is very simple. Fortunately, we have two simple
inversion formulas,
L1 fexpðasÞg ¼ dðt aÞ; L1 fs expðasÞg ¼ d0 ðt aÞ ð3:7:35Þ
where d0 (.) is the derivative of the delta function as
ddðtÞ
d0 ðtÞ ¼ ð3:7:36Þ
dt
Applying these two inversion formulas to Eq. (3.7.33), we have the final form of the
torsional displacement as

T r 1 0 1
uh ¼ d ðt R=c s Þ þ dðt R=c s Þ ð3:7:37Þ
8pl R2 cs R
The above equation shows only one spherical wave and its front is R = cst. It should
be noticed that disturbance is only on the wave front and no disturbance its inside,
R < cst, due to the nature of the delta function.
Exercises
(3:1) From the unified expression for the Green’s dyadic (3.1.42), derive the
explicit expressions for Gxx, Gxy, Gyx,Gyy and show that Gxy = Gyx.
(3:2) Show that the explicit expression of the Green’s dyadic (3.4.55) is the same
as the corresponding one in the displacement equation (3.4.53).
(3:3) When the body force Bθ in the governing equation (3.7.1) is a suddenly
applied point torque,

T0 dðr aÞ
Bh ðr; z; tÞ ¼ lim dðzÞHðtÞ ð3:7:39Þ
2pq a!0 ar
where H(t) is Heaviside’s unit step function and T0 is the magnitude of the torque,
show that the corresponding Green’s function is given by

T0 r 1 1
uh ¼ dðt R=c s Þ þ Hðt R=c s Þ ð3:7:40Þ
8pl R2 cs R
Appendix
See Table 3.1.

Table 3.1 Green’s dyadic for elastodynamic equations
Displacement equations Source Bi Green’s dyadic
X
2D 2D inplane deformation (plane strain) Pi dðxÞdðyÞdðtÞ ui ðx; y; tÞ ¼ Pj Gij ðx; y; tÞ; i; j ¼ x; y
Appendix
j¼x;y
@ @ux @uy @ 2 ux @ 2 ux 1 @ 2 ux 1
ðc2 1Þ þ þ 2 þ 2 ¼ 2 2 2 Bx ðx; y; tÞ 8 qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi9
@x @x @y @x @y cs @t cs > >
Hðt r=cd Þ <xi xj 1 2xi xj ðcd tÞ2 r 2 =
@ @ux @uy @ 2 uy @ 2 uy 1 @ 2 uy 1 Gij ðx; y; tÞ ¼ qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi þ dij
ðc2 1Þ þ þ 2 þ 2 ¼ 2 2 2 By ðx; y; tÞ 2pcd > r2
: r2 r2 >
;
@y @x @y @x @y cs @t cs ðcd tÞ2 r 2
> >
Hðt r=cs Þ <xi xj 1 2xi xj ðcs tÞ2 r 2 =
2
dij qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi þ 2
dij 2
2pcs >
: r 2 r r >
;
ðcs tÞ r 2
X
Qi dðxÞdðyÞ expðþixtÞ ui ðx; y; tÞ ¼ Qj gij ðx; yÞ expðþixtÞ; i; j ¼ x; y
j¼x;y

gij ðx; yÞ ¼ H ðrx=cd Þ þ dij 2 H ðrx=cd Þ
4 c2d r 2 0 r rx 1

2 dij H0 ðrx=cs Þ þ dij 2 H ðrx=cs Þ
cs r2 r rx 1
X ðstaticÞ
Si dðxÞdðyÞ ui ¼ Sj gij ðx; yÞ; i; j ¼ x; y
j¼x;y

x2
ðstaticÞ 1
gxx ðx; yÞ ¼ ðc2 þ 1Þ logðrÞ þ ðc2 1Þ
4pc2d r
ðstaticÞ 1 xy
gxy ðx; yÞ ¼ ðc2 1Þ 2
4pc2d r

y2
ðstaticÞ 1 2
gyy ðx; yÞ ¼ ðc þ 1Þ logðrÞ þ ðc2 1Þ
4pc2d r
X
3D @ @ux @uy @uz @ 2 ux @ 2 ux @ 2 ux 1 @ 2 ux 1 Pi dðxÞdðyÞdðzÞdðtÞ ui ¼ Pj Gij ðx; y; z; tÞ; i; j ¼ x; y; z
ðc2 1Þ þ þ þ 2 þ 2 þ 2 ¼ 2 2 2 Bx ðx; y; z; tÞ
@x @x @y @z @x @y @z cs @t cs j¼x;y;z
8 9
@ @ux @uy @uz @ 2 uy @ 2 uy @ 2 uy 1 @ 2 uy 1 > xi xj 1 xi xj
ðc2 1Þ þ þ >
> >
>
dðt R=cd Þ þ 2 dij 2 dðt R=cs Þ >
þ 2 þ 2 þ 2 ¼ 2 2 2 By ðx; y; z; tÞ < c 2 R2 c s R =
@y @x @y @z @x @y @z cs @t cs 1 d
Gij ðx; y; z; tÞ ¼
@ @ux @uy @uz @ 2 uz @ 2 uz @ 2 uz 1 @ 2 uz 1 4pR >
> t 3xi xj >
>
ðc2 1Þ þ þ þ 2 þ 2 þ 2 ¼ 2 2 2 Bz ðx; y; z; tÞ >
: 2 dij 2 Hðt R=cd ÞHðR=cs tÞ > ;
@z @x @y @z @x @y @z cs @t cs R R
X
Qi dðxÞdðyÞdðzÞ ui ¼ Qj gij ðx; y; z; tÞ; i; j ¼ x; y; z
j¼x;y;z
expðþixtÞ ( )
1 xi xj 1 xR 3xi xj
gij ðx; y; z; tÞ ¼ 2 2
þ 2
1þi dij 2 expfixðt R=cd Þg
4pR cd R ðxRÞ cd R
( )
1 1 xi xj 1 xR 3xi xj
þ 2
dij 2 2
1þi dij 2 expfixðt R=cs Þg
4pR cs R ðxRÞ cs R
(continued)
117
Table 3.1 (continued)
118
Displacement equations Source Bi Green’s dyadic

X ðKÞ
Si dðxÞdðyÞdðzÞ ui ¼ Sj Gij ðx; y; zÞ; i; j ¼ x; y; z
j¼x;y;z
1 n xi xj o
ðKÞ
Gij ðx; y; zÞ ¼ ð3 4mÞdij þ 2
8pð1 2mÞc2d R R
(sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi)
Torsional ring source Q R22 ðcs tÞ2 ðcs tÞ2 R21
@ 2 uh 1 @uh uh @ 2 uh 1 @ 2 uh 1 dðraÞ uh ¼ H ðR2 cs tÞH ðcs t R1 Þ
@r 2 þ r @r r 2 þ @z2 c2 @t2 ¼ c2 Q r dðzÞdðtÞ 2 2
s s 4pcs ar ðcs tÞ R1 R22 ðcs tÞ2
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
R1 ¼ ðr aÞ2 þ z2 ; R2 ¼ ðr þ aÞ2 þ z2

Point torque source T r 1 0 1
uh ¼ 2
d ðt R=cs Þ þ dðt R=cs Þ
@ 2 uh h uh @ 2 uh 1 @ 2 uh T dðraÞ 8pl R cs R
@r 2 þ 1r @u
@r r 2 þ @z2 c2 @t2 ¼ 2pl lim ar dðzÞdðtÞ
s a!0 pffiffiffiffiffiffiffiffiffiffiffiffiffiffi
ddðtÞ 2 2
d0ðtÞ ¼ ; R¼ r þz
dt

Suddenly applied point torque T0 r 1 1
uh ¼ dðt R=cs Þ þ Hðt R=cs Þ
@ 2 uh h uh @ 2 uh 1 @ 2 uh T0 dðraÞ 8pl R2 cs R
@r 2 þ 1r @u
@r r 2 þ @z2 c2 @t2 ¼ 2pl lim ar dðzÞHðtÞ pffiffiffiffiffiffiffiffiffiffiffiffiffiffi
s a!0
R ¼ r2 þ z2
3 Green’s Dyadic for an Isotropic Elastic Solid
References 119
References
Erdélyi A (ed) (1954) Tables of integral transforms, vol I and II. MacGraw-Hill, New-York
Gradshteyn IS, Ryzhik IM (Jefferey A (ed)) (1980) Table of integrals, series, and products, 5th
edn. Academic Press, San Diego
Watson GN (1966) A treatise on the theory of bessel functions. Cambridge University Press,
Cambridge (1966)
Chapter 4
Acoustic Wave in a Uniform Flow
Traditionally, when we use the word “wave,” we think about acoustic waves or
water waves. The acoustic wave is much more familiar to our daily life and we
experience many wave phenomena such as reflection, refraction, diffraction, the
Doppler effects, etc. The governing equations for the acoustic wave are rigorously
derived from the fluid equations and Green’s function for the acoustic wave in a
flowing fluid is discussed by applying the method of integral transform.
4.1 Compressive Viscous Fluid
Motions and disturbances in fluids such as water, oil and gas are governed by four
groups of equations:
(1) Equations of motion,

@rxx @ryx @rzx @vx @vx @vx @vx
þ þ þ qBx ¼ q þ vx þ vy þ vz
@x @y @z @t @x @y @z

@rxy @ryy @rzy @vy @vy @vy @vy
þ þ þ qBy ¼ q þ vx þ vy þ vz ð4:1:1Þ
@x @y @z @t @x @y @z

@rxz @ryz @rzz @vz @vz @vz @vz
þ þ þ qBz ¼ q þ vx þ vy þ vz
@x @y @z @t @x @y @z
(2) Constitutive equations for the linear Newtonian fluid,

2l @vx @vy @vz @vx
rxx ¼ p þ þ þ 2l
3 @x @y @z @x

2l @vx @vy @vz @vy
ryy ¼ p þ þ þ 2l
3 @x @y @z @y
ð4:1:2Þ
2l @vx @vy @vz @vz
rzz ¼ p þ þ þ 2l
3 @x @y @z @z

@vx @vy @vy @vz @vx @vz
rxy ¼ l þ ; ryz ¼ l þ ; rzx ¼ l þ
@y @x @z @y @z @x

DOI 10.1007/978-3-319-17455-6_4
122 4 Acoustic Wave in a Uniform Flow
(3) Continuity equation
@q @ðqvx Þ @ðqvy Þ @ðqvz Þ

þ þ þ ¼0 ð4:1:3Þ
@t @x @y @z
(4) Equation of state for the acoustic medium (adiabatic change)

j
p q
¼ ð4:1:4Þ
p0 q0
In above equations, the viscosity l is a known/given constant that specifies the

nature of the fluid. The body force Bi is assumed to be a source and its functional
form must be specified. The stress rij , the particle velocity vi, the hydro-static
pressure p, and the density q are unknown functions to be determined. The sub-
script “0” stands for the quantities at a reference state. The constant parameter j is
the ratio of two specific heats. Thus, the governing equation for the linear fluid is a
set of strongly coupled partial differential equations with 11 unknowns.
Substituting the constitutive Eq. (4.1.2) into the equations of motion (4.1.1), we
obtain the well-known Navier-Stokes equations,
2
@p l @ @vx @vy @vz @ vx @ 2 vx @ 2 vx
þ þ þ þl þ 2 þ 2 þ qBx
@x 3 @x @x @y @z @x2 @y @z
ð4:1:5aÞ
@vx @vx @vx @vx
¼ q vx þ vy þ vz þ
@x @y @z @t
2
@p l @ @vx @vy @vz @ vy @ 2 vy @ 2 vy
þ þ þ þl þ 2 þ 2 þ qBy
@y 3 @y @x @y @z @x2 @y @z
ð4:1:5bÞ
@vy @vy @vy @vy
@x @y @z @t
2
@p l @ @vx @vy @vz @ vz @ 2 vz @ 2 vz
þ þ þ þl þ 2 þ 2 þ qBz
@z 3 @z @x @y @z @x2 @y @z
ð4:1:5cÞ
@vz @vz @vz @vz
@x @y @z @t
When the fluid is incompressible and no density change takes place, the con-
tinuity equation is simplified and the equation of state is unnecessary. Then, the
incompressible fluid is governed by a simplified form of the Navier-Stokes equa-
tions. However, when we discuss acoustic waves, the two reminder Eqs. (4.1.3) and
(4.1.4) must be fully incorporated since acoustic waves express the propagation of
density changes.
4.2 Linearization 123
4.2 Linearization
The nonlinearity stems from the acceleration terms in Navier-Stokes equations

(4.1.5), the product ðqvi Þ in the continuity equation (4.1.3) and the power of density
in the equation of state (4.1.4). Thus, the governing equations for acoustic waves
are fully nonlinear, coupled partial differential equations. In order to reduce the
differential equations to tractable ones, we introduce the assumption that the
acoustic wave is an infinitesimal small disturbance superimposed on a reference
nonlinear fluid motion. This assumption states that the infinitesimal small distur-
bance is a deviation from the nonlinear reference state. We introduce a small
parameter e, and represent the disturbance with ðvi ; p; q Þ. The five unknown
functions, the velocity component vi, the static hydro-pressure p, and the density q,
are approximated as the power of the small parameter e,
vi ¼ Vi þ evi þ Oðe2 Þ; p ¼ p0 þ ep þ Oðe2 Þ; q ¼ q0 þ e

q þ Oðe2 Þ ð4:2:1Þ
We also assume that the wave source described by the body force is a small
quantity of order e1 ,
i þ Oðe2 Þ
Bi ¼ eB ð4:2:2Þ
Substituting Eqs. (4.2.1–4.2.2) into Eq. (4.1.5a), and neglecting terms of higher
order than Oðe1 Þ, the equation is reduced to two parts, the zeroth and first order
terms. The zeroth order term is
2
@p0 l @ @Vx @Vy @Vz @ Vx @ 2 Vx @ 2 Vx
þ þ þ þl þ þ 2
@x 3 @x @x @y @z @x2 @y2 @z
ð4:2:3Þ
@Vx @Vx @Vx @Vx
¼ þ q0 þ Vx þ Vy þ Vz
@t @x @y @z
and the first order term is

2
@
p l @ @vx @vy @vz @ vx @ 2vx @ 2vx x
e þ þ þ þl þ þ þ qB
@x 3 @x @x @y @z @x2 @y2 @z2

@Vx @Vx @Vx
¼ þq Vx þ Vy þ Vz
@x @y @z

@vx @vx @Vx @vx @Vx @vx @Vx
þ q0 þ q0 Vx þ vx þ q 0 Vy þ vy þ q 0 Vz þ vz
@t @x @x @y @y @z @z
ð4:2:4Þ
Similarly, we substitute Eqs. (4.2.1–4.2.2) into the reminder of Eq. (4.1.5). From
Eq. (4.1.5b), we have the zeroth order term
2
@p0 l @ @Vx @Vy @Vz @ Vy @ 2 Vy @ 2 Vy
þ þ þ þl þ 2 þ 2
@y 3 @y @x @y @z @x2 @y @z
ð4:2:5Þ
@Vy @Vy @Vy @Vy
@t @x @y @z
and the first order term

2
@p l @ @vx @vy @vz @ vy @ 2vy @ 2vy y
@y 3 @y @x @y @z @x2 @y2 @z2

@Vy @Vy @Vy
@x @y @z

@vy @vy @Vy @vy @Vy @vy @Vy
þ q0 þ q0 Vx þ vx þ q0 Vy þ vy þ q0 Vz þ vz
@t @x @x @y @y @z @z
ð4:2:6Þ
From Eq. (4.1.5c), we also have for the zeroth order term
2
@p0 l @ @Vx @Vy @Vz @ Vz @ 2 Vz @ 2 Vz
þ þ þ þl þ 2 þ 2
@z 3 @z @x @y @z @x2 @y @z
ð4:2:7Þ
@Vz @Vz @Vz @Vz
@t @x @y @z
and for the first order term

2
@p l @ @vx @vy @vz @ vz @ 2vz @ 2vz z
@z 3 @z @x @y @z @x2 @y2 @z2

@Vz @Vz @Vz
@x @y @z

@vz @vz @Vz @vz @Vz @vz @Vz
@t @x @x @y @y @z @z
ð4:2:8Þ
Substituting Eq. (4.2.1) into Eq. (4.1.3), the continuity equation is decomposed
into the two parts,
@ðq0 þ eqÞ @ðq0 þ eqÞðVx þ evx Þ @ðq0 þ e

qÞðVy þ evy Þ @ðq0 þ e qÞðVz þ evz Þ
þ þ þ
@t @x @y @z
@q0 @ðq0 Vx Þ @ðq0 Vy Þ @ðq0 Vz Þ
¼ þ þ þ
@t @x @y @z

q @ðq0vx Þ @ðq0vy Þ @ðq0vz Þ @ðVx q
@ Þ @ðVy q Þ @ðVz q
Þ
þe þ þ þ þ þ þ
@t @x @y @z @x @y @z
þ Oðe2 Þ ¼ 0
ð4:2:9Þ
4.2 Linearization 125
The equation of state for the adiabatic change in Eq. (4.1.4) is approximated as

p0 þ ep q j
q0 þ e p j
q
p j
q
¼ ) 1þe ¼ 1 þ e þ Oðe2 Þ ) ¼ ð4:2:10Þ
p0 q0 p0 q0 p0 q0
Thus, we have the linear relation between the pressure and density deviations.
Consequently, we have the equations for the reference state as the Oðe0 Þ part,
2
@p0 l @ @Vx @Vy @Vz @ Vx @ 2 Vx @ 2 Vx
þ þ þ þl þ þ
@x 3 @x @x @y @z @x2 @y2 @z2
ð4:2:11aÞ
@Vx @Vx @Vx @Vx
@t @x @y @z
2
@p0 l @ @Vx @Vy @Vz @ Vy @ 2 Vy @ 2 Vy
þ þ þ þl þ þ
@y 3 @y @x @y @z @x2 @y2 @z2
ð4:2:11bÞ
@Vy @Vy @Vy @Vy
@t @x @y @z
2
@p0 l @ @Vx @Vy @Vz @ Vz @ 2 Vz @ 2 Vz
þ þ þ þl þ þ
@z 3 @z @x @y @z @x2 @y2 @z2
ð4:2:11cÞ
@Vz @Vz @Vz @Vz
@t @x @y @z
@q0 @ðq0 Vx Þ @ðq0 Vy Þ @ðq0 Vz Þ

þ þ þ ¼0 ð4:2:12Þ
@t @x @y @z
The disturbance of Oðe1 Þ is governed by the equations

2
@p l @ @vx @vy @vz @ vx @ 2vx @ 2vx x
þ þ þ þl þ þ þ q0 B
@x 3 @x @x @y @z @x2 @y2 @z2

@Vx @Vx @Vx
@x @y @z

@vx @vx @Vx @vx @Vx @vx @Vx
@t @x @x @y @y @z @z
ð4:2:13aÞ
2
@
p l @ @vx @vy @vz @ vy @ 2vy @ 2vy y
@y 3 @y @x @y @z @x2 @y2 @z2

@Vy @Vy @Vy
@x @y @z

@vy @vy @Vy @vy @Vy @vy @Vy
@t @x @x @y @y @z @z
ð4:2:13bÞ
2
@
p l @ @vx @vy @vz @ vz @ 2vz @ 2vz z
þ þ þ þl þ 2 þ 2 þ q0 B
@z 3 @z @x @y @z @x2 @y @z

@Vz @Vz @Vz
@x @y @z

@vz @vz @Vz @vz @Vz @vz @Vz
@t @x @x @y @y @z @z
ð4:2:13cÞ
Þ @ðVy q
q @ðq0vx Þ @ðq0vy Þ @ðq0vz Þ @ðVx q
@ Þ @ðVz q
Þ
þ þ þ þ þ þ ¼ 0 ð4:2:14Þ
@t @x @y @z @x @y @z
q0
¼
q p ð4:2:15Þ
jp0
We should understand that the zeroth order equations, which govern the reference
fluid flow, are already satisfied by the reference quantities ðVi ; p0 ; q0 Þ. On the other
hand, the first order equations, which include the change/deviation from the ref-
erence state, are the coupled differential equations for the 5 unknowns ðvi ; p; q Þ.
4.3 Viscous Acoustic Fluid
When the reference state is a uniform flow with viscosity l, we can assume
Vi ; p0 ; q0 ¼ const:, and then the first order Eq. (4.2.13) are reduced to the simpler
forms
2
@
p l @ @vx @vy @vz @ vx @ 2vx @ 2vx x
@x 3 @x @x @y @z @x2 @y2 @z2
ð4:3:1aÞ
@vx @vx @vx @vx
¼ þ q0 þ q0 Vx þ Vy þ Vz
@t @x @y @z
2
@
p l @ @vx @vy @vz @ vy @ 2vy @ 2vy y
@y 3 @y @x @y @z @x2 @y2 @z2
ð4:3:1bÞ
@vy @vy @vy @vy
@t @x @y @z
2
@
p l @ @vx @vy @vz @ vz @ 2vz @ 2vz z
@z 3 @z @x @y @z @x2 @y2 @z2
ð4:3:1cÞ
@vz @vz @vz @vz
@t @x @y @z
4.3 Viscous Acoustic Fluid 127
With use of the linear relation between the pressure and density deviations, the
continuity equation (4.2.14) can be rewritten in terms of the pressure and velocity
gradients as

@
p @
p @p @p @vx @vy @vz
þ Vx þ Vy þ Vz þ jp0 þ þ ¼0 ð4:3:2Þ
@t @x @y @z @x @y @z
The above four Eqs. (4.3.1) and (4.3.2) are the governing equations for linear-
ized acoustic waves in the uniformly flowing viscous fluid. The coupled differential
equations, which have four unknowns ðvi ; pÞ, can be reduced to a single differential
equation by the introduction of a velocity potential. We assume that the velocity
components ðvi Þ can be derived from a single velocity potential /ðx; y; z; tÞ as
@/ @/ @/
vx ¼ ; vy ¼ ; vz ¼ ð4:3:3Þ
@x @y @z
As to the body force, we also assume that there exists a body force potential B(x, y,
z, t) and each body force component is derived as,
x ¼ @B ;
B y ¼ @B ;
B z ¼ @B
B ð4:3:4Þ
@x @y @z
Substituting Eqs. (4.3.3) and (4.3.4) into (4.3.1a), we obtain

@
p @ @/ @/ @/ @/ 4l @ @ 2 / @ 2 / @ 2 / @B
þ q0 þ Vx þ Vy þ Vz þ þ 2 q0 ¼0
@x @x @t @x @y @z 3 @x @x2 @y2 @z @x
ð4:3:5Þ
Since all terms in the above equation are derivatives with respect to the space
variable x, we integrate it with respect to the variable x,

@/ @/ @/ @/ 4l @ 2 / @ 2 / @ 2 /

p ¼ q0 þ Vx þ Vy þ Vz þ þ þ þ q0 B
@t @x @y @z 3 @x2 @y2 @z2
ð4:3:6Þ
where the constant term is neglected since it is one of the reference quantities.
Substitution into Eqs. (4.3.1b) and (4.3.1c) leads to the same equation and thus the
equations of motion in all three directions are reduced to the single Eq. (4.3.6).
The velocity potential defined by Eq. (4.3.3) is also substituted into the conti-
nuity equation (4.3.2),
2
@
p @
p @p @p @ / @2/ @2/
þ Vx þ Vy þ Vz þ jp0 þ þ 2 ¼0 ð4:3:7Þ
@t @x @y @z @x2 @y2 @z
Then, we have the coupled differential Eqs. (4.3.6) and (4.3.7) in terms of only two
unknowns, the velocity potential and the pressure deviation. Further, if we sub-
stitute Eq. (4.3.6) into Eq. (4.3.7), we have just the single differential equation for
the velocity potential /,

4l @ @ @ @
r /þ 2
þ Vx þ Vy þ Vz r2 /
3jp0 @t @x @y @z

q @ @ @ @ 2 q @B @B @B @B
¼ 0 þ Vx þ Vy þ Vz / 0 þ Vx þ Vy þ Vz
jp0 @t @x @y @z jp0 @t @x @y @z
ð4:3:8Þ
where r2 is the Laplacian operator defined by
@2 @2 @2
r2 ¼ þ þ ð4:3:9Þ
@x2 @y2 @z2
We can also derive the differential equation for the pressure deviation. We apply
the Laplacian to both sides of Eq. (4.3.6),

@ @ @ @ 4l
r
p ¼ q0
2
þ Vx þ Vy þ Vz r2 / þ r2 r2 / þ q0 r2 B
@t @x @y @z 3
ð4:3:10Þ
and derive the Laplacian of the velocity potential from Eq. (4.3.7) as

1 @p @p @p @p
r2 / ¼ þ Vx þ Vy þ Vz ð4:3:11Þ
jp0 @t @x @y @z
The Laplacian of the velocity potential in Eq. (4.3.10) is replaced with the above
pressure gradient. We therefore have the single differential equation for the pressure
deviation as

4l @ @ @ @
r
pþ2
þ Vx þ Vy þ Vz r2 p
3jp0 @t @x @y @z
2 ð4:3:12Þ
q @ @ @ @ 2 @ B @2B @2B
¼ 0 þ Vx þ Vy þ Vz p þ q0 þ þ
jp0 @t @x @y @z @x2 @y2 @z2
Two differential Eqs. (4.3.8) and (4.3.12) for the potential and the pressure are
mathematically the same, but the non-homogeneous terms of the body force are
4.3 Viscous Acoustic Fluid 129
little bit different. We easily see that both equations are wave equations and their
wave velocity is
pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
c¼ ðjp0 Þ=q0 ð4:3:13Þ
If we introduce a reference length l and define Reynolds number,
Re ¼ ðq0 clÞ=l ð4:3:14Þ
and Mach numbers,
Mj ¼ Vj =c; j ¼ x; y; z ð4:3:15Þ
the governing equation for the viscous acoustic media is simplified to

4l 1 @ @ @ @
r2 / þ þ Mx þ My þ Mz r2 /
3Re c @t @x @y @z

1@ @ @ @ 2 1 1 @B @B @B @B
¼ þ Mx þ My þ Mz / þ Mx þ My þ Mz
c @t @x @y @z c c @t @x @y @z
ð4:3:16Þ
for the velocity potential, and to

4l 1 @ @ @ @
r2
pþ þ Mx þ My þ Mz r2 p
3Re c @t @x @y @z
2 ð4:3:17Þ
1@ @ @ @ 2 @ B @2B @2B
¼ þ Mx þ My þ Mz p þ q0 þ þ
c @t @x @y @z @x2 @y2 @z2
for the pressure deviation. These two differential equations have only two param-
eters, Reynolds and Mach numbers. Thus, the nature of acoustic waves in the
flowing viscous fluid is characterized by these two numbers.
4.4 Wave Radiation in a Uniform Flow
This section discusses a 2D wave propagation problem in a uniform flow. Take the
(x, y) plane and assume that all quantities are independent of z, and assume that the
non-viscous acoustic fluid (l ¼ 0) is flowing along the x-axis with the uniform
velocity Vx (see Fig. 4.1). Further, we also assume the Laplacian of the body force
potential as the wave source Q,
⎧⎪δ (t )
Qδ ( x )δ ( y ) ⎨ iω t
⎪⎩ e
Vx
x
Fig. 4.1 A point source in a uniformly flowing fluid
2
@ B @2B
q0 þ ¼ QdðxÞdðyÞdðtÞ ð4:4:1Þ
@x2 @y2
Then, we substitute Re ¼ 1; My ¼ Mz ¼ 0 into Eq. (4.3.17). The governing

equation for the pressure deviation is then reduced to the simple form
2
1@ @
r2 p ¼ þ Mx p þ QdðxÞdðyÞdðtÞ ð4:4:2Þ
c @t @x
where c is the acoustic wave velocity defined by Eq. (4.3.13), and Mach number Mx
is defined by Eq. (4.3.15). Furthermore, the pressure in this equation is the deviation
from the reference state and the correct notation is p (over bar). However, we do not
use the over bar for the pressure deviation since we apply the double Fourier
transform and one of the transforms is classified by the over bar. So, in order to
avoid the confusion, the over bar is dropped from the pressure deviation.
Since our acoustic field is of infinite extent, it is enough to obtain a particular
solution of Eq. (4.4.2). Laplace transform with respect to the time t and double
Fourier transform with respect to two space variables x and y are applied. The triple
integral transform is defined as
Z1

Laplace transform: f ðsÞ ¼ f ðtÞ expðstÞdt ð4:4:3Þ
0
Z1 Z1
1
Fourier transform: f ðnÞ ¼ f ðxÞ expðþinxÞdx; f ðxÞ ¼ f ðnÞ expðinxÞdn
2p
0 0
ð4:4:4Þ
Z1 Z1
1
Fourier transform: ~f ðgÞ ¼ f ðyÞ expðþigyÞdy; f ðyÞ ¼ ~f ðgÞ expðigyÞdg
2p
0 0
ð4:4:5Þ
4.4 Wave Radiation in a Uniform Flow 131
Applying the triple transform to Eq. (4.4.2), the simple algebraic equation for the
transformed pressure is obtained as
ðn2 þ g2 Þ~p ¼ ðs=c inMx Þ2 ~p þ Q ð4:4:6Þ
The pressure in the transformed domain is thus given by
Q Q
~p ¼ ¼ ð4:4:7Þ
n2 þ g2 þ ðs=c inMx Þ2 g2 þ q2
where q is written as
q2 ¼ n2 þ ðs=c inMx Þ2
( 2 2 )
iMx s=c
¼ 1 Mx 2
n ðs=cÞ þ
1 Mx2 1 Mx2
n o
¼ 1 Mx2 ðn iaMx Þ2 þa2 ð4:4:8Þ
The Laplace transform parameter s is included in the parameter a which is defined

by
s=c
a¼ ð4:4:9Þ
1 Mx2
As to the first inversion, we apply the Fourier inversion with respect to the
parameter g, and reduce it to the semi-infinite integral,
Zþ1 Z1
1 Q Q 1

p ¼ expðigyÞdg ¼ cosðgyÞdg ð4:4:10Þ
2p g þq
2 2 p g2 þ q2
1 0
The last integral is easily evaluated by the formula (2.1.22) in Chap. 2 and yields
Q
p ¼ expðqjyjÞ ð4:4:11Þ
2q
The next step is to apply the Fourier inversion integral with respect to the parameter n,
Zþ1
Q 1 1
p ¼ expðqjyjÞ expðinxÞdn ð4:4:12Þ
2 2p q
1
Recalling q in Eq. (4.4.8), the integral in the above equation is rewritten in the
explicit form,

pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiqffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
ffi
Zþ1 exp jyj 1 Mx ðn iaMx Þ þa inx
2 2 2
Q 1
p ¼ pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi dn
2 1 Mx 2p 2
ðn iaMx Þ2 þa2
1
ð4:4:13Þ
Introducing the variable transform, n ! 1, as
1 ¼ n iaMx ð4:4:14Þ
we obtain the simpler integrand, but the integration path is shifted to the complex
plane, not on the real axis, i.e.
pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffipffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
Z
þ1iaMx
exp jyj 1 Mx2 12 þ a2 i1x
Q expðþaMx xÞ 1
p ¼ pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi d1
2 1 Mx2 2p 1 2 þ a2
1iaMx
ð4:4:15Þ
The integration path for this integral is slightly shifted from the real axis in the
complex 1-plane and is shown by the line CD in Fig. 4.2. In order to transform the
integral to that along the real axis, we consider the complex integral U with the
!
closed loop LðCDBACÞ in the figure,
Z exp jyj 1 M 2 12 þ a2 i1x

Q expðþaMx xÞ 1 x
U¼ pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi d1 ð4:4:16Þ
2 1 Mx 2 2p 1 þ a2
2
L
The integrand has two branch points, at 1 ¼ ia. Two branch cuts are introduced
along the imaginary axis, as shown in the figure. Fortunately, if we assume subsonic
Fig. 4.2 Transform of Im(ς )

integration path
+iα
Branch cut
A B Re(ς )
C
−iα M x D
−iα
motion of the flow, Mx ð¼Vx =cÞ\1, these branch cuts do not cross the integration
line CD and no singular point are then included in the closed loop L. In addition, the
two integrals
along the line AC and BD, with an infinite real part, vanish since
Re 1 þ a [ 0. Then, applying Cauchy’s integral theorem to the complex
2 2
integral U in Eq. (4.4.16), the integral along the complex path CD is converted to
that along the real axis AB in the 1-plane. That is
Zþ1 exp jyj 1 M 2 12 þ a2 i1x

p ¼ pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi d1 ð4:4:17Þ
2 1 Mx2 2p 1 2 þ a2
1
The above integral can be further reduced to the semi-infinite integral,

Z1 exp jyj 1 M 2 12 þ a2
p ¼ pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi cosð1xÞd1 ð4:4:18Þ
2 1 Mx2 p 1 þ a2
2
0
Applying the integration identity, which is the integral representation of the mod-
ified Bessel function (Erdélyi 1954, vol. I, pp. 17, 27; Watson 1966, p. 172),
Z1 exp y 12 þ a2
K 0 a x2 þ y2 ¼ pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi cosð1xÞd1
1 2 þ a2
0 pffiffiffiffiffiffiffiffiffiffiffiffiffiffi
Z1 exp au x2 þ y2
¼ pffiffiffiffiffiffiffiffiffiffiffiffiffi du ð4:4:19Þ
u2 1
1
to Eq. (4.4.18), we have the other integral form for the pressure,
Z1 qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
Q expðþaMx xÞ 1
p ¼ pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi pffiffiffiffiffiffiffiffiffiffiffiffiffi exp au x2 þ 1 Mx2 y2 du ð4:4:20Þ
2p 1 Mx2 u2 1
1
Further, we recall the definition of a which includes the Laplace transform

parameter s. The above Eq. (4.4.20) can be rewritten as
n qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
o
Z1 exp s=c 2 u x2 þ 1 M 2 y2 Mx x
Q 1 1Mx x
p ¼ pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi pffiffiffiffiffiffiffiffiffiffiffiffiffi du ð4:4:21Þ
2p 1 Mx2 u 1
2
1
Introducing the variable transform, u ! t, as defined by


ux2 þ 1 Mx2 y2 Mx x 1 Mx2 ct þ Mx x
t¼ ; u ¼ qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
;
c 1 Mx2 x2 þ 1 Mx2 y2
ð4:4:22Þ
c 1 Mx2
du ¼ qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
dt
x2 þ 1 Mx2 y2
Equation (4.4.21) is converted to the form of the Laplace transform integral,
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi Z1
cQ expðstÞdt
p ¼ 1 Mx2 ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
q 2
2p pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
ffi 1 Mx2 ct þ Mx x x2 þ 1 Mx2 y2
x2 þð1Mx Þy2 Mx x
2
cð1Mx2 Þ
ð4:4:23Þ
The above equation is the Laplace transformed pressure, but it is just in the form
of the Laplace transform integral. Thus, the original pressure is its integrand with a
shifted starting time since the lower limit is not zero. Inspecting Eq. (4.4.23), we
can find the original pressure as
cQ 1 Mx2
p ¼ qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
2 ð4:4:24Þ
2p
1 Mx2 ct þ Mx x x2 þ 1 Mx2 y2
where its valid time range is


x2 þ 1 Mx2 y2 Mx x
\t\1 ð4:4:25Þ
c 1 Mx2
In order to obtain a more compact expression, the argument in the radical and the
time range are rewritten as

x2 þ 1 Mx2 y2 \ct 1 Mx2 þ Mx x ) ðx Vx tÞ2 þy2 \ðctÞ2 ð4:4:26Þ
2 n o
1 Mx2 ct þ Mx x x2 þ 1 Mx2 y2 ¼ 1 Mx2 ðctÞ2 ðx Vx tÞ2 y2
ð4:4:27Þ
Fig. 4.3 Disturbed circular y

region in a uniform flow
Vx t
Flow Vx ct
x
source
Finally, the pressure deviation in the flowing fluid is expressed in the simple form,
Q 1
p¼ rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
ffi H ct ðx Vx tÞ2 þy2 ð4:4:28Þ
2p 2 n o
t ðx Vx tÞ2 þy2 =c2
We have just arrived at the final form of the solution that gives the pressure
fluctuations in the flowing fluid with the stationary impulsive point source. The
disturbed region is easily derived from the argument of the step function: it is a
circle but its center is moving with the flow velocity,
ðx Vx tÞ2 þy2 ¼ ðctÞ2 ð4:4:29Þ
Figure 4.3 shows the typical wave front and the circular disturbed region in the
flowing fluid.
4.5 Time-Harmonic Wave in a Uniform Flow
The steady-state acoustic response produced by a stationary time-harmonic source

is governed by

1@ @ 2
r p
2
þ Mx p ¼ QdðxÞdðyÞ expðþixtÞ ð4:5:1Þ
c @t @x
A solution to this equation can be obtained by the convolution integral of the

impulsive response as
Zt
p ¼ lim pðimpulseÞ ðx; y; t0 Þ expfixðt t0 Þgdt0 ð4:5:2Þ
t!1
0
Substituting the impulsive response of Eq. (4.4.24), not the final one, we have
cQ 1 Mx2 expðixt0 Þdt0
p ¼ expðixtÞ qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
ffi
2p pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
ffi 2 ct 0 þ M x 2 x2 þ 1 M 2 y2
1 M x x x
2 2 2 x þð1Mx Þy Mx x
cð1Mx2 Þ
ð4:5:3Þ
where the source magnitude Q is not the same as that of the impulsive source. For
this integral, we make the change of variable, t0 ! u, as
u Mx x du
u ¼ 1 Mx2 ct0 þ Mx x; t0 ¼ ; dt0 ¼ ð4:5:4Þ
1 Mx2 c 1 Mx2 c
Equation (4.5.3) is now rewritten as

Z1 exp 1M ixu
Q 1 Mx x ð x2 Þc
p ¼ pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi exp ix t þ qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
du
2p 1 Mx 2 1 Mx c
2
pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi u2 x2 þ ð1 Mx2 Þy2
x2 þð1Mx2 Þy2
ð4:5:5Þ
Since the integral in the above equation is just the definition integral of the Hankel
function of the second kind (which is Eq. (3.2.5) in Sect. 3.2), we replace the
integral with the Hankel function. The final form for the time-harmonic response is
then given by
pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi!
iQ ð2Þ x x2 þ ð1 Mx2 Þy2 Mx x
p ¼ pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi H0 exp ix t þ ð4:5:6Þ
4 1 Mx 2 c 1 Mx2 1 Mx2 c
4.5 Time-Harmonic Wave in a Uniform Flow 137
Exercises
(4:1) When a wave source is moving with uniform velocity V along the x-axis,
how do you change the nonhomogeneous wave source term in Eq. (4.4.1)?
(4:2) Using the asymptotic formula for the Hankel function (Watson 1966, p. 198),
rffiffiffiffiffi
2
Hmð2Þ ðzÞ expfiðz mp=2 p=4Þg ðaÞ
pz
show that the time-harmonic response (4.5.6) is the out-going wave.
References
Erdélyi A (ed) (1954) Tables of integral transforms, vols I and II. McGraw-Hill, New York
Cambridge
Chapter 5
Green’s Functions for Beam and Plate
This chapter presents dynamic Green’s functions for the elastic beam and plate. The
Green’s function is the deflection response produced by a point load. The deflection
equation with the nonhomogeneous term of the applied load is discussed by
applying the method of integral transform.
5.1 An Impulsive Load on a Beam
We shall obtain Green’s function for an elastic beam. The deflection of the beam,
w(x, t), is governed by the well-known partial differential equation, so called beam/
deflection equation,
@4w @2w
EI þ qA 2 ¼ pðx; tÞ ð5:1:1Þ
@x 4 @t
where EI is bending rigidity, ρA mass per unit length and p(x, t) the load on the
beam. An impulsive point loadwith magnitude P is assumed by
pðx; tÞ ¼ PdðxÞdðtÞ ð5:1:2Þ
where dð:Þ is Dirac’s delta function. The nonhomogeneous deflection equation

(5.1.1) is rewritten as
@4w @2w
a4 þ 2 ¼ QdðxÞdðtÞ ð5:1:3Þ
@x4 @t
where
EI P
a4 ¼ ; Q¼ ð5:1:4Þ
qA qA

DOI 10.1007/978-3-319-17455-6_5
140 5 Green’s Functions for Beam and Plate
For solving the deflection equation, we employ the quiescent condition at an initial
time,

@w
wjt¼0 ¼ ¼0 ð5:1:5Þ
@x t¼0

@w @ 2 w @ 3 w
wjjxj!1 ¼ ¼ 2 ¼ ¼0 ð5:1:6Þ
@x jxj!1 @x jxj!1 @x3 jxj!1
The higher derivatives in the above equation mean the vanishing of the moment and
shear force at infinity.
In order to solve the nonhomogeneous deflection equation, we employ the
method of integral transform. Since the deflection depends on two variables, we
apply the following double transform: Laplace transform with respect to the time
variable,
Z1

0
and Fourier transform with respect to the space variable,
Zþ1 Zþ1
2p
1 1
Applying the double transform to the deflection equation (5.1.3), we have the
algebraic equation for the transformed deflection and then the deflection in the
transformed domain is given by
Q
¼
w ð5:1:9Þ
ðanÞ4 þ s2
The Laplace inversion is applied firstly. The Laplace inversion formula (Erdélyi
1954, vol. I, pp. 150, 1),

1 1 1
L ¼ sinðatÞ ð5:1:10Þ
s 2 þ a2 a
is applied to Eq. (5.1.9) and thus we have for the Fourier transformed deflection,
5.1 An Impulsive Load on a Beam 141
Q
¼
w 2
sinða2 tn2 Þ ð5:1:11Þ
ðanÞ
The Fourier inversion integral of the above equation is reduced to the semi-infinite
integral, as
Zþ1 Z1
1 Q Q 1
w¼ sinða tn Þ expðinxÞdn ¼ 2
2 2
sinða2 tn2 Þ cosðxnÞdn
2p ðanÞ2 pa n2
1 0
ð5:1:12Þ
Fortunately, we have a convenient formula for the inversion integral. That is the
formula (Erdélyi 1954, vol. I, pp. 23, 3)
Z1 2
1 p y y pffiffiffiffiffiffi y p
sinðax Þ cosðxyÞdx ¼ y S pffiffiffiffiffiffiffiffi C pffiffiffiffiffiffiffiffi
2
þ pa sin þ
x2 2 2pa 2pa 4a 4
0
ð5:1:13Þ
where C(x) and S(x) are Fresnel integrals/functions (Erdélyi 1954, vol. II, p. 431)
defined by
Zx
CðxÞ 1 1 sinðuÞ
¼ pffiffiffiffiffiffi pffiffiffi du ð5:1:14Þ
SðxÞ 2p u cosðuÞ
0
Applying this formula to the last integral in Eq. (5.1.12) and rewriting the
expression, we obtain the Green’s function for the dynamic deflection of the beam
as
rffiffiffi np o
Q t pz
wðx; tÞ ¼ pffiffiffi fSðzÞ CðzÞg þ sin ð2z2 þ 1Þ ð5:1:15Þ
a p 2 4
where the dimensionless variable z is defined by

x
z ¼ pffiffiffiffiffiffiffi ð5:1:16Þ
a 2pt
Note that the Green’s function given by Eq. (5.1.15) is the solution for the
impulsive response; however, it does not show any wave nature since the deflection
wave is dispersive and its wave velocity depends on the frequency. The impulsive
source includes an infinite frequency, i.e. the initial disturbance spreads all over the
beam at once without showing any wave nature.
5.2 A Moving Time-Harmonic Load on a Beam
When a time-harmonic load with frequency ω is moving with the uniform velocity
V, the deflection equation for the beam is given by,
@4w @2w
EI þ qA 2 ¼ Pdðx VtÞ expðþixtÞ ð5:2:1Þ
@x 4 @t
where the load location is expressed by the delta function and is at x ¼ Vt. The
convergence condition given by Eq. (5.1.6) is also assumed.
Since the time variable t is included not only in the argument of the exponential
function, but also in that of delta function, it is not good to assume a simple time-
harmonic vibration such as wðx; tÞ ¼ w# ðxÞ expðþixtÞ. We directly apply the
Fourier transform defined by Eq. (5.1.8) to the deflection equation (5.2.1),

d2 w
2
þ ðanÞ4 w
¼ Q expfþiðVn þ xÞtg ð5:2:2Þ
dt
where α and Q are defined by Eq. (5.1.4). The particular solution corresponding to
the nonhomogeneous loading term is easily obtained as
Q
¼
w expfþiðnV þ xÞtg ð5:2:3Þ
ðanÞ4 ðnV þ xÞ2
Factorizing the denominator, we get
ðanÞ4 ðnV þ xÞ2 ¼ a4 ðn nþ þ

1 Þðn n1 Þðn n2 Þðn n2 Þ ð5:2:4Þ
where the eigenvalues are given by

8 sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi 9 8 sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi 9
< 2 = < 2 =
1 V V 1 V V
n
1 ¼ xþ þ ; n
2 ¼ i x ð5:2:5Þ
a: 2a 2a; a: 2a 2a;
the Fourier inversion integral with respect to the parameter n is expressed as

Zþ1
Q expfinðx VtÞg
w¼ expðþixtÞ dn ð5:2:6Þ
2pa4 ðn nþ n
1 Þðn
þ
1 Þðn n2 Þðn n2 Þ
1
In order to evaluate the integral in Eq. (5.2.6), we apply the theory of complex
integrals (the residue theorem). Following the discussion in Sect. 2.2, the complex
frequency with a small negative imaginary part is assumed. Due to this assumption,
all singular points do not lie on the real axis. Two singular points, n þ
1 and n2 , are in
þ
the upper complex n-plane and the other two, n1 and n2 , are in the lower plane.
Then, the complex integral
5.2 A Moving Time-Harmonic Load on a Beam 143
Z
1 expfinðx VtÞg
U¼ dn ð5:2:7Þ
2p ðn nþ n
1 Þðn
þ
C
is to be discussed. Two closed loops C are shown in Fig. 5.1. We employ the circuit
C þ in the case of x Vt \ 0 and C in that of x Vt [ 0 in order to guarantee the
convergence on the large semi-circle. Jordan’s lemma is applied to the complex
integral Φ and after somewhat lengthy calculation we arrive at the expressions,
Zþ1
1 expfinðx VtÞg
dn
2p ðn nþ
1 Þðn n þ
1
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
2
exp ði=aÞ x þ ðV=2aÞ þ ðV=2aÞ ðx VtÞ
ia3
¼ 2
4 qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi2ffi qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi2ffi
x þ ðV=2aÞ x þ ðV=2aÞ þ ðV=2aÞ
2
exp ð1=aÞ x ðV=2aÞ þ iðV=2aÞ ðx VtÞ
a3
2 ; x Vt [ 0
4 qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi2ffi qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi2
x ðV=2aÞ x ðV=2aÞ iðV=2aÞ
2
exp ði=aÞ x þ ðV=2aÞ ðV=2aÞ ðx VtÞ
ia3
¼ 2
4 qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi2ffi qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi2ffi
x þ ðV=2aÞ x þ ðV=2aÞ ðV=2aÞ
2
exp ð1=aÞ x ðV=2aÞ þ iðV=2aÞ ðx VtÞ
a3
2 ; x Vt\0
4 qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi2ffi qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi2
x ðV=2aÞ x ðV=2aÞ þ iðV=2aÞ
ð5:2:8Þ
Then, the deflection produced by the moving time-harmonic load is given by

n
pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi o
iQ exp ði=aÞ x þ X2 þ X ðx VtÞ þ ixt
wðx; tÞ ¼ pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi 2
4a
x þ X2 x X2 þ X
n
Q exp ð1=aÞ x X þ iX ðx VtÞ þ ixt
2
pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi 2 ; x Vt [ 0
4a
x X2 x X2 iX
ð5:2:9aÞ
Im(ξ ) Im(ξ )
(c / 2α ) + ω + (c / 2α ) 2 −(c / 2α ) + i ω − (c / 2α ) 2
ξ1+ = ξ 2+ =
α α
Re(ξ )
C+
−
C
Re(ξ )
(c / 2α ) − ω + (c / 2α ) 2
−(c / 2α ) − i ω − (c / 2α ) 2 ξ1− =
ξ 2− = α
α
Fig. 5.1 Closed loops C for the complex integral Φ
n
iQ exp ði=aÞ x þ X 2
X ðx VtÞ þ ixt
wðx; tÞ ¼ pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
4a
x þ X2 x X2 X
n
Q exp ð1=aÞ x X 2
þ iX ðx VtÞ þ ixt
þ pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi 2 ; x Vt\0
4a
x X2 x X2 þ iX
ð5:2:9bÞ
where
V
X¼ ð5:2:10Þ
2a
Inspecting the above equations, the radical x X2 in the denominator gives a
critical velocity
pffiffiffiffi
Vcr ¼ 2a x ð5:2:11Þ
at which the deflection divergence takes place.

When the load is stationary at the coordinate origin, the simple time-harmonic
response is obtained by taking the limit X ! 0 ðV ! 0Þ in Eq. (5.2.10). It yields
pffiffiffiffi pffiffiffiffi
Q x x
wðx; tÞ ¼ pffiffiffiffi3 i exp þi xt jxj þ exp jxj þ ixt
4a x a a
ð5:2:12Þ
5.2 A Moving Time-Harmonic Load on a Beam 145
This is the particular solution of the deflection equation with the time-harmonic
stationary load,
@4w @2w
a4 þ 2 ¼ QdðxÞ expðþixtÞ ð5:2:13Þ
@x4 @t
5.3 An Impulsive Load on a Plate
Let us consider an impulsive response of an infinite elastic plate. Taking the

(x, y) coordinate on the neutral plane of the plate, the deflection equation for the
plate with an impulsive point loadis given by
4
@ w @4w @4w @2w
D þ 2 þ þ qh ¼ PdðxÞdðyÞdðtÞ ð5:3:1Þ
@x4 @x2 @y2 @y4 @t2
where D is bending rigidity, ρh density per unit area and P magnitude of the load.
The initial condition,

@w
wjt¼0 ¼ ¼0 ð5:3:2Þ
@t t¼0
and the convergence condition,

ffi @w @ 2 w @ 3 w
wj x2 þy2 !1 ¼ ffi ¼ @x2 pffiffiffiffiffiffiffiffiffi ¼ @x3 pffiffiffiffiffiffiffiffiffi ¼ 0
@x pxffiffiffiffiffiffiffiffi
2 þy2 !1 x2 þy2 !1 x2 þy2 !1
ð5:3:3Þ
@w 2
@ w 3
@ w 2
@ w
@y pffiffiffiffiffiffiffiffi ffi ¼ @y2 pffiffiffiffiffiffiffiffiffi ¼ @y3 pffiffiffiffiffiffiffiffiffi ¼ @x@ypffiffiffiffiffiffiffiffiffi ¼ 0
x2 þy2 !1 x2 þy2 !1 x2 þy2 !1 x2 þy2 !1
are applied to the deflection equation.

In order to obtain the particular solution corresponding to the non-homogeneous
loading term, we apply the triple integral transform to Eq. (5.3.1). The Laplace
transform with respect to the time t,
Z1

0
and the double Fourier transform with respect to the space variables (x, y),
Zþ1 Zþ1
2p
1 1
Zþ1 Zþ1
f ðyÞ expðþigyÞdy; f ðyÞ ¼ ð5:3:6Þ
2p
1 1
are applied to Eq. (5.3.1). The exact solution in the transformed domain is easily
obtained as
Q
~ ¼
w ð5:3:7Þ
s2 þ fb2 ðn2 þ g2 Þg2
where
D P
b4 ¼ ; Q¼ ð5:3:8Þ
qh qh
Now, let us consider the inversion. The Laplace inversion of Eq. (5.3.7) is easily
performed with use of the formula (5.1.10). It follows that
Q
~ ¼
w sinfb2 ðn2 þ g2 Þtg; t[0 ð5:3:9Þ
b ðn þ g2 Þ
2 2
The double Fourier inversion is expressed by the double integral as
Zþ1 Zþ1
1 Q
w¼ sinfb2 ðn2 þ g2 Þtg expfiðnx þ gyÞgdndg
ð2pÞ2 b2 ðn2 þ g2 Þ
1 1
ð5:3:10Þ
If we introduce the polar coordinates (r, θ) and the polar integration variables
(u, φ) as
x ¼ r cos h; y ¼ r sin h ð5:3:11Þ
n ¼ u sin u; g ¼ u cos u; dndg ¼ ududu ð5:3:12Þ
the double integral can be rewritten as
Z1 Z2p
Q 1
w¼ 2 2
sinðb2 tu2 Þ expfiur sinðu þ hÞgdudu ð5:3:13Þ
ð2pÞ b u
0 0
5.3 An Impulsive Load on a Plate 147
The exponential function in the integrand is the generating function of Bessel

function (Watson 1966, p. 22),
X
þ1
expfiur sinðu þ hÞg ¼ Jn ðurÞ expfinðu þ hÞg ð5:3:14Þ
1
where Jn ð:Þ is the n-th order Bessel function of the first kind. We substitute the
above into Eq. (5.3.13) and exchange the order of the summation and integration as
Z1 Z2p X
þ1
Q 1
w¼ sinðb2 tu2 Þ Jn ðurÞ expfinðu þ hÞgdudu
ð2pÞ2 b2 u 1
0 0
ð5:3:15Þ
þ1 Z Z
1 2p
Q X 1
¼ sinðb2 2
tu ÞJ n ðurÞdu expfinðu þ hÞgdu
ð2pÞ2 b2 1 u
0 0
The inner integral with respect to the angle variable φ is easily evaluated as
Z2p
2p; n ¼ 0
expfinðu þ hÞgdu ¼ ð5:3:16Þ
0; n 6¼ 0
0
Then, we have the single integral for the deflection
Z1
Q 1
w¼ sinðb2 tu2 ÞJ0 ðurÞdu ð5:3:17Þ
2pb2 u
0
Lastly, the integration formula (Erdélyi 1954, vol. II, pp. 11, 39),
Z1 2
1 1 b
sinðax2 ÞJ0 ðbxÞdx ¼ si ð5:3:18Þ
x 2 4a
0
where si(x) is one of sine integrals (Erdélyi 1954, vol. II, p. 430) defined by
Z1
sinðuÞ
si(xÞ ¼ du ð5:3:19Þ
u
x
is applied to Eq. (5.3.17). The Green’s function for the dynamic plate deflection is
thus given by
Q r
w¼ 2
si ; t [ 0; r¼ x2 þ y2 ð5:3:20Þ
4pb 4b2 t
5.4 A Time-Harmonic Load on a Plate
Let us consider an elastic plate on which a time-harmonic point load is applied. The
dynamic deflectionof the plate is governed by the deflection equation with the
nonhomogeneous loading term,
4
@ w @4w @4w @2w
D þ 2 þ þ qh 2 ¼ PdðxÞdðyÞ expðþixtÞ ð5:4:1Þ
@x 4 @x @y
2 2 @y 4 @t
where P is the magnitude of the load (but is not of the same dimension as that of the
impulsive load in the previous section). The convergence condition at infinity,

@ 2 w @ 3 w
ffi ¼ @w
wjpxffiffiffiffiffiffiffiffi ¼ ¼
2 þy2 !1
@x pxffiffiffiffiffiffiffiffiffi
2 þy2 !1 @x2 pxffiffiffiffiffiffiffiffi ffi
2 þy2 !1 @x3 pxffiffiffiffiffiffiffiffiffi ¼0
2 þy2 !1
ð5:4:2Þ
@w @ 2 w @ 3 w @ 2 w
¼ ¼ ¼ ¼ 0
@y pffiffiffiffiffiffiffiffiffi
x2 þy2 !1 @y2 pxffiffiffiffiffiffiffiffi ffi
2 þy2 !1 @y3 pxffiffiffiffiffiffiffiffi ffi
2 þy2 !1 @x@ypxffiffiffiffiffiffiffiffi ffi
2 þy2 !1
is also assumed. The double Fourier transform defined by Eqs. (5.3.5) and (5.3.6) is
applied to the deflection equation (5.4.1). We then have the simple differential
equation,
~
d2 w
~ ¼ Q expðþixtÞ
þ b4 ðn2 þ g2 Þ2 w ð5:4:3Þ
dt2
and its particular solution is given by
Q
~ ¼
w expðþixtÞ ð5:4:4Þ
b ðn þ g2 Þ2 x2
4 2
where β and Q are defined by Eq. (5.3.8).

The formal Fourier inversion is given by the double integral
Zþ1 Zþ1
Q expðinx igyÞ
w¼ expðþixtÞ dndg ð5:4:5Þ
ð2pÞ2 b4 ðn2 þ g2 Þ2 x2
1 1
5.4 A Time-Harmonic Load on a Plate 149
Introducing the polar coordinates defined by Eqs. (5.3.11) and (5.3.12) in the
previous section, the double Fourier inversion integral can be rewritten as
Z1 Z2p
Q u
w¼ 2
expðþixtÞ expfiur sinðu þ hÞgdudu ð5:4:6Þ
ð2pÞ ðbuÞ4 x2
0 0
The inner integral with respect to the angular variable φ is evaluated with the aid of
the formulas (5.3.14) and (5.3.16), and the above Eq. (5.4.6) is reduced to the single
integral,
Z1
Q u
w¼ expðþixtÞ pffiffiffiffi 4 J0 ðurÞdu ð5:4:7Þ
2pb4 u4 ð x=bÞ
0
where the radial distance r is x2 þ y2 .
The last integral with respect to the variable u is evaluated by applying the
complex integral theory. As was done in the former time-harmonic problem in
Sect. 5.2 and earlier too, we introduce a small negative imaginary part to the
frequency, and then consider the complex integral
Z
z ð1Þ
U¼ pffiffiffiffi 4 H0 ðzrÞdz ð5:4:8Þ
z4 ð x=bÞ
C
ð1Þ
where H0 ð:Þ is Hankel function of the first kind. The closed loop C is shown in
Fig. 5.2. Since the Hankel function has a logarithmic singularity and its branch
point is at the origin, one branch cut along the negative real axis is introduced. The
integrand has four poles, whereof two are in the upper z-plane due to the intro-
duction of the imaginary part to the frequency. They are marked in the figure and
these two poles are specified with argument
pffiffiffiffi pffiffiffiffi pffiffiffiffi pffiffiffiffi
x x x x
z¼ ¼ expðþpiÞ; z ¼ þi ¼ expðþpi=2Þ ð5:4:9Þ
b b b b
It should be understood that the argument of z along the positive real axis is zero,
but that along the negative real axis is þp. Then, we apply Cauchy’s theorem to the
complex integral Φ in Eq. (5.4.8). The integral along the large semi-circle vanishes
as its radius tends to infinity, and that along the small semi-circle also vanishes as
its radius tends to zero. Thus, the integral along the real axis is converted to the sum
of two residues at the poles. That is
Im( z )
C
arg( z ) = +π
z = +i ω / β
R →∞
z=− ω /β
F
D E A B
Re( z )
ε →0 arg( z ) = 0
branch cut
Fig. 5.2 A closed loop C for the complex integral Φ
Z
R!1
n o
u ð1Þ ð1Þ
p ffiffiffi
ffi 4
H0 ðurÞ H0 ðurepi Þ du
u4 ð x=bÞ
e!0 ð5:4:10Þ
b2 ð1Þ r x pi=2 b2 ð1Þ r x pi
¼ 2pi H0 e þ H0 e
4x b 4x b
The formulas for the Hankel function (Watson 1966, pp. 75 and 78),
ð1Þ ð2Þ ð1Þ 2i

H0 ðxepi Þ ¼ H0 ðxÞ; H0 ðxepi=2 Þ ¼ K0 ðxÞ ð5:4:11Þ
p
are very useful for arranging Eq. (5.4.10). Then, we have the integration formula for
our use,
Z1 pffiffiffiffi pffiffiffiffi
u b2 pi ð2Þ r x r x
pffiffiffiffi 4 J0 ðruÞdu ¼ H0 þ K0 ð5:4:12Þ
u ð x=bÞ
4 2x 2 b b
0
Applying this formula to Eq. (5.4.7), the time-harmonic response of the plate, i.e.
the time-harmonic Green’s function, is given by
Q pi ð2Þ r x r x
wðx; y; tÞ ¼ 2
H0 þ K0 expðþixtÞ ð5:4:13Þ
4pb x 2 b b
where r is the radial distance from the load.

Table 5.1 Green’s functions for beam and plate
Deflection equation Source p(x, t) Green’s function
4 2 pffiffih i
1D beam EI @@xw4 þ qA @@tw2 ¼ pðx; tÞ PdðxÞdðtÞ wðx; tÞ ¼ Qa pt ppzffiffi2 fSðzÞ CðzÞg þ sin p4 ð2z2 þ 1Þ ;

1=4
5.4 A Time-Harmonic Load on a Plate
EI P z ¼ apxffiffiffiffiffi
a ¼ qA ; Q ¼ qA 2pt
h n
pffiffiffi o
pffiffiffi i
PdðxÞ expðþixtÞ wðx; tÞ ¼ pQffiffiffi3 i exp þi xt ax jxj þ exp ax jxj þ ixt
4a x
4
2D plate @ w @4w @4w PdðxÞdðyÞdðtÞ Q r
D 2 wðx; y; tÞ ¼ 4pb 2 si ; r ¼ x2 þ y2
þ þ 4b2 t
@x4 @x2 @y2 @y4 n
pffiffiffi
pffiffiffio
PdðxÞdðyÞ expðþixtÞ ð2Þ
@2w wðx; y; tÞ ¼ 4pbQ2 x pi2 H0 r bx þ K0 r bx expðþixtÞ;
þqh 2 ¼ pðx; y; tÞ pffiffiffiffiffiffiffiffiffiffiffiffiffiffi
@t r ¼ x2 þ y2

1=4
D P
b ¼ qh ; Q ¼ qh
151
The reader will find that the deflection has two components: One is a time-
ð2Þ pffiffiffiffi
harmonic wave which is given by the product H0 ðr x=bÞ expðþixtÞ. The other
pffiffiffiffi
is a simple time-harmonic vibration K0 ðr x=bÞ expðþixtÞ whose amplitude
decays exponentially with the distance. The same nature can be found in the 1D
beam response in Eq. (5.2.12). Table 5.1 shows the summary for the deflection
Green’s functions.
Exercises
(5:1) Derive the wave velocity from the time-harmonic responses (5.2.9), (5.2.12)
and (5.4.13) and show that all velocities depend on the frequency.
(5:2) Using the generating function for Bessel function (5.3.14), derive the integral
representation for the Bessel function as
Zþp Zþp
1 1
Jn ðzÞ ¼ expðiz sin u þ inuÞdu ¼ cosðnu z sin uÞdu ðaÞ
2p p
p 0
(5:3) The asymptotic formula for the Hankel function is given by
rffiffiffiffiffi
ð1Þ 2
H0 ðzÞ expfþiðz p=4Þg; z ! 1 ðbÞ
pz
Explain why the integral along the semi-circle BCD in Fig. 5.2 vanishes for
the complex integral (5.4.8).
Appendix
See Table 5.1.
References
Erdélyi A (ed) (1954) Tables of integral transforms, vol I and II. McGraw-Hill, New-York
Cambridge
Chapter 6
Cagniard-de Hoop Technique
The success of the integral transform method hinges on the evaluation of inversion
integrals. It is not always easy to find a suitable integration formula. If we cannot
find any suitable formula, the inversion is left in its integral form and some
numerical techniques must be applied for the evaluation. However, in the case of
double inversion, such as Laplace and Fourier inversions, if we could convert the
first Fourier inversion integral to a definition integral of Laplace transform, the next
Laplace inversion can be carried out by inspection without evaluating its Laplace
inversion integral. For example, when we have the double transformed function
Fðn; sÞ and its Fourier inversion integral
Zþ1

f ðx; sÞ ¼ Fðn; sÞ expðinxÞdn ð6:1Þ
1
if we could apply some mathematical techniques to the integral and convert the
integral to the definition form of Laplace transform,
Z1
f ðx; sÞ ¼ Gðx; tÞ expðstÞdt ð6:2Þ
a
i.e. the Laplace transform integral, the Laplace inversion is easily carried out by
inspection and its inversion is given by

Gðx; tÞ; t [ a
f ðx; tÞ ¼ ¼ Hðt aÞGðx; tÞ ð6:3Þ
0; t\a
The most important and substantial part of this technique is to convert the
Fourier inversion integral to the form of the Laplace transform,
Zþ1 Z1
Fðn; sÞ expðinxÞdn ¼ Gðx; tÞ expðstÞdt ð6:4Þ
1 a

DOI 10.1007/978-3-319-17455-6_6
154 6 Cagniard-de Hoop Technique
This conversion is carried out with aids of the theory of complex integrals and is
called “Cagniard-de Hoop technique.” Cagniard (1962) is the author of this tech-
nique, but his original technique is to map the inversion integral onto the integral in
another complex t-plane. De Hoop (1961) has modified the technique to use the same
complex plane without mapping. The present chapter explains the technique devel-
oped by de Hoop, but uses the name “Cagniard-de Hoop technique.” The essential
idea is to convert the first inversion integral to the definition integral of the second
integral transform. This technique is applicable to other combinations of two integral
transforms, not limited to Laplace-Fourier transforms. In addition, it will be worth to
cite textbooks (Achenbach 1973; Fung 1970; Graff 1975; Miklowitz 1978), which
treat elastic waves and explain the Cagniard-de Hoop technique in some depths.
The first section in the present chapter treats a very simple problem in order to
demonstrate the Cagniard-de Hoop technique in details. In the following sections,
2D and 3D Green’s functions (dyadic) for the coupled elastodynamic equations are
discussed and some additional explanations for the Cagniard-de-Hoop technique
are described.
6.1 2D Anti-plane Deformation
As the first example of the Cagniard-de Hoop technique, the simplest elastody-
namic problem is discussed. We consider an elastic half space and take the
Cartesian coordinate (x, y, z) as shown in Fig. 6.1 where the z-axis is normal to the
paper plane. The surface of the solid is denoted by y ¼ 0 and its interior y [ 0. An
impulsive anti-plane shear load directed to the negative z-direction is applied on the
surface and is expressed by

ryz y¼0 ¼ Pz dðxÞdðtÞ ð6:1:1Þ
Fig. 6.1 An impulsive anti-

plane shear load on the
σ yz = Pzδ ( x)δ (t )
y =0
surface of a semi-infinite y=0 x
elastic solid
μ, ρ
y
6.1 2D Anti-plane Deformation 155
where Pz is the load magnitude. Since the load of infinite length is directed to the
anti-plane z-direction, an anti-plane deformation is induced. The equation of motion
is given by
@rxz @ryz @ 2 uz
þ ¼q 2 ð6:1:2Þ
@x @y @t
where q is the density. Hooke’s law for two shear stresses is given by
@uz @uz
rxz ¼ l ; ryz ¼ l ð6:1:3Þ
@x @y
where l is the shear rigidity. We also employ the quiescent condition at an initial
time,

@uz
uz jt¼0 ¼ ¼0 ð6:1:4Þ
@t t¼0

@uz @uz
uz jpffiffiffiffiffiffiffiffiffi ¼ ¼ ¼0 ð6:1:5Þ
ffi
x2 þy2 !1 @y pxffiffiffiffiffiffiffiffi
ffi
2 þy2 !1
Equations (6.1.1)–(6.1.5) constitute the present elastodynamic problem.

Substituting Hooke’s law of Eq. (6.1.3) into the equation of motion (6.1.2), we
obtain the typical wave equation for the anti-plane displacement uz,
@ 2 uz @ 2 uz 1 @ 2 uz
þ 2 ¼ 2 2 ð6:1:6Þ
@x 2 @y cs @t
where the shear wave velocity cs is defined by

cs ¼ l=q ð6:1:7Þ
In order to solve our elastodynamic problem, Laplace and Fourier transforms are
employed; Laplace transform with respect to the time t,
Z1

0
and Fourier transform with respect to the space variable x,

Zþ1 Zþ1
2p
1 1
With the aid of the quiescent and convergence conditions, the displacement
equation (6.1.6) and Hooke’s law (6.1.3) are transformed to
d 2 uz n 2 o
n þ ðs=c s Þ2
uz ¼ 0 ð6:1:10Þ
dy2
duz
xz ¼ inluz ;
r yz ¼ l
r ð6:1:11Þ
dy
After solving the transformed displacement equation (6.1.10), the displacement and
stress, which satisfy the convergence condition at infinity, are given by
uz ¼ Aðn; sÞ expðas yÞ;

yz ¼ las Aðn; sÞ expðas yÞ
r ð6:1:12Þ
Here Aðn; sÞ is an unknown coefficient to be determined by the loading condition,

and the radical is defined by
as ¼ n2 þ ðs=cs Þ2 ; Reðas Þ [ 0 ð6:1:13Þ
The boundary condition is also transformed, to

yz
r ¼ Pz ð6:1:14Þ
y¼0
Substituting the second equation in Eq. (6.1.12) into the above transformed
boundary condition, the coefficient is determined as
Pz
Aðn; sÞ ¼ ð6:1:15Þ
las
and thus, the displacement in the transformed domain is explicitly determined as
Pz
uz ¼ expðas yÞ ð6:1:16Þ
las
Now, we shall consider the inversion. The formal Fourier inversion for the
displacement is given by the infinite integral
Zþ1
Pz 1
uz ¼ expðas y inxÞdn ð6:1:17Þ
2pl as
1
The above Eq. (6.1.17) is the Laplace transformed displacement, but in the form
of Fourier inversion integral. If we could convert the integral into the definition
form of Laplace transform integral, its Laplace inversion can be performed by
inspection (i.e. the Cagniard-de Hoop technique).
Let us start to apply the technique. First of all, in order to eliminate the Laplace
transform parameter s, leaving it only in the argument of the exponential function in
the integrand, a variable transform from n to 1 is introduced as
n ¼ s1 ð6:1:18Þ
Here we assume that the Laplace transform parameter s is a positive real con-
stant. The integral in Eq. (6.1.17) is rewritten as
Zþ1 n qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi o
Pz 1
uz ¼ pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi exp s y 12 þ 1=c2s þ i1x d1 ð6:1:19Þ
2pl 12 þ 1=c2s
1
where the radical must satisfy the radiation condition,

qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
Re 12 þ 1=c2s [ 0 ð6:1:20Þ
Examining the integral in Eq. (6.1.19), the Laplace transform parameter is

included only in the argument of the exponential function as a simple multiplier. If
we could change the argument to a simple product, such as st where t is a new
variable, the integral will have the form of the definition integral of the Laplace
transform. So, the variable transform from 1 to the new variable t is introduced as
t¼y 12 þ 1=c2s þ i1x ð6:1:21Þ
However, its inverse gives multiple values for 1

itx y t2 ðx2 þ y2 Þ=c2s
1ðÞ ¼ ð6:1:22Þ
s
x2 þ y 2
Due to the multi-valuedness, we are puzzled as to which one is the suitable

inversion for 1. In order to solve this puzzle, we examine the inverse as a function
of the real parameter t. We separate the real and imaginary parts of the inverse as
ðÞ
1s ¼ X iY, where
t2 ðx2 þ y2 Þ=c2s
y tx
X¼ ; Y¼ ð6:1:23Þ
x2 þ y2 x2 þ y2
and eliminate the parameter t. An equation for the hyperbolic curve in (X, Y) plane
is derived as
2 2
X Y 1
¼ 2 ð6:1:24Þ
y x cs
ðÞ
Thus, the inverse 1s has the form of two semi-hyperbolic curves in the complex 1-
ðÞ ðþÞ
plane. When x [ 0, 1s is the left half and 1s is the right half of the hyperbola in
the lower 1-plane, as shown in Fig. 6.2. On the other hand, when x\0, the
ðÞ
hyperbolic curves 1s are in the upper 1-plane. The connected two semi-hyperbolic
curves CAB shown in the figure is called Cagniard’s path.
If we could connect the real line with the edges of two semi-hyperbolas in the
complex plane, a closed loop for the complex integral is formed; and then Cauchy’s
integral theorem can be applied. Thus, we consider the complex integral U, whose
integrand is the same as that of the Fourier inversion integral in Eq. (6.1.19),
I n qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi o
1 1
U¼ pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi exp s y 12 þ 1=c2s þ i1x d1 ð6:1:25Þ
2p 12 þ 1=c2s
C
Fig. 6.2 Cagniard’s path

The integrand has two branch points at
1 ¼ i=cs ð6:1:26Þ
In order to guarantee the radiation condition of Eq. (6.1.20), two branch cuts, which
are discussed in Sect. 1.3.2 in Chap. 1, are introduced along the imaginary axis in
the complex 1-plane as shown in Fig. 6.3. When the real parameter t varies from
pffiffiffiffiffiffiffiffiffiffiffiffiffiffi ðþÞ
x2 þ y2 =cs to þ1, the inverse 1s moves on the semi-hyperbola from A to B,
ðÞ
and the other inverse 1s moves on the other semi-hyperbola from A to C. These
two semi-hyperbolas are connected at the saddle point A,
ix pffiffiffiffiffiffiffiffiffiffiffiffiffiffi
1saddle ¼ pffiffiffiffiffiffiffiffiffiffiffiffiffiffi at t ¼ x2 þ y2 =cs ð6:1:27Þ
cs x2 þ y2
This saddle point is always smaller in magnitude than the branch point 1 ¼ i=cs .
The other edges at the infinite t must be connected with the real line. The edge B is
connected to the line at the positive infinity F with the large arc. The edge C is also
connected to the line at the negative infinity D with the large arc, as shown in
Fig. 6.3. Then, the closed loop for the complex integral is formed by the loop
CABFEDC. The closed loop C(−) in the lower 1-plane is employed in the case of
positive x and the loop C(+) in the upper plane is in the case of negative x. In either
Fig. 6.3 Closed loop C ðÞ for the complex integral U

case, no singular point is included in the closed loop. The integral along the large
arc vanishes as its radius tends to infinity, since the radiation condition of
Eq. (6.1.20) is kept. Thus, the integral along the real axis is converted to that along
the two semi-hyperbolas, CA and AB, due to Cauchy’s integral theorem.
The integral along the hyperbola is one of the parametric integrals and its
parameter is the variable t. The identity between two integrals along the real axis
and along the hyperbolas is thus given by
1 1
pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi exp s y 12 þ 1=c2s þ i1x d1
2p 12 þ 1=c2s
1
8 9
Z1
1 < 1 d1 1 d1 =
¼ pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi expðstÞdt
2p pffiffiffiffiffiffiffiffiffi : 12 þ 1=c2s dt ðþÞ 12 þ 1=c2s dt ðÞ
;
1¼1s 1¼1s
x2 þy2 =cs
ð6:1:28Þ
The integrand in the right hand side of Eq. (6.1.28) can be simplified. Explicit
expressions for the radical and the gradient are derived from the definition of
Eq. (6.1.21) and its inverse Eq. (6.1.22). They are
1 x2 þ y2
ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
r ffi¼ pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
ðÞ
2 yt ix t2 ðx2 þ y2 Þ=c2s
1s þ 1=c2s
ð6:1:29Þ
ðÞ pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
d1s yt ix t2 ðx2 þ y2 Þ=c2s
¼ pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
dt ðx2 þ y2 Þ t2 ðx2 þ y2 Þ=c2s
Substituting these into the integral in the right hand side of Eq. (6.1.28), the
parametric integral along the hyperbola is simplified as
8 9
Z1
1 < 1 d1 1 d1 =
pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi expðstÞdt
2p pffiffiffiffiffiffiffiffiffi : 12 þ 1=c2s dt ðþÞ 12 þ 1=c2s dt 1¼1ðÞ ;
1¼1s s
x2 þy2 =cs
Z1
1 1
¼ pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi expðstÞdt ð6:1:30Þ
p pffiffiffiffiffiffiffiffiffi t2 ðx2 þ y2 Þ=c2s
x2 þy2 =cs
Thus, we have converted the Fourier inversion integral to the integral form of the
Laplace transform,
1 1
pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi exp s y 12 þ 1=c2s þ i1x d1
2p 1 þ 1=cs
2 2
1
Z1 ð6:1:31Þ
1 1
¼ pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi expðstÞdt
p pffiffiffiffiffiffiffiffiffi t ðx2 þ y2 Þ=c2s
2
x2 þy2 =cs
Substituting the right hand side of the above equation into the Laplace transformed
displacement in Eq. (6.1.19), we have the Laplace transformed displacement in the
form of Laplace transform integral,
Z1
Pz 1
uz ðx; y; sÞ ¼ pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi expðstÞdt ð6:1:32Þ
pl pffiffiffiffiffiffiffiffiffi t ðx2 þ y2 Þ=c2s
2
x2 þy2 =cs
This integral is just the form of Laplace transform, but its lower limit is not zero.
The integral states that the original function is vanishing before t ¼ x2 þ y2 =cs
and after this time the function has the form of the integrand. Then, we can easily
anticipate the original displacement function before Laplace transform, i.e. Laplace
inversion of Eq. (6.1.32),

uz ðx; y; tÞ ¼ L1 uz ðx; y; sÞ

Pz < pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi ; t [ x2 þ y2 =cs
1
t ðx þy Þ=cs
2 2 2 2 ð6:1:33Þ
¼ pffiffiffiffiffiffiffiffiffiffiffiffiffiffi
pl :
0; t\ x2 þ y2 =cs
Utilizing Heaviside’s unit step function, the displacement is expressed in the

compact form,
Pz 1 pffiffiffiffiffiffiffiffiffiffiffiffiffiffi
uz ðx; y; tÞ ¼ pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi H cs t x2 þ y2 ð6:1:34Þ
pl t2 ðx2 þ y2 Þ=c2s
This is our final result for the double inversion. The reader should notice that we did
not use any integration formula for Laplace and Fourier inversions but we did the
inversion. This is the Cagniard-de Hoop technique!
The conditional for the step function gives the circular region disturbed by the
transient SH-wave,
ðcs tÞ2 [ x2 þ y2 ð6:1:35Þ

and its edge is a circular (cylindrical in 3D) wave front with the center at the source
point,
ðcs tÞ2 ¼ x2 þ y2 ð6:1:36Þ
6.2 2D In-plane Deformation
We consider a 2D transient response of an semi-infinite elastic solid. Take the

coordinate system (x, y) so that the surface of the solid is at y ¼ 0 and its interior in
y [ 0 as shown in Fig. 6.4. An impulsive point load is applied on the surface and is
expressed by

ryx y¼0 ¼ Px dðxÞdðtÞ; ryy y¼0 ¼ Py dðxÞdðtÞ ð6:2:1Þ
where Pj ; j ¼ x; y are the components of the load. The in-plane deformation of the
elastic solid is governed by the equations of motion,
@rxx @ryx @ 2 ux
þ ¼q 2
@x @y @t
ð6:2:2Þ
@rxy @ryy @ 2 uy
þ ¼q 2
@x @y @t
where q is density. Hooke’s law for the plane strain is given by
Fig. 6.4 An impulsive load

on the surface of a semi-
σ yy = Pyδ ( x)δ (t )
y =0
infinite elastic solid
σ yx = Pxδ ( x)δ (t ) x
y =0 y=0
λ , μ, ρ
y
6.2 2D In-plane Deformation 163

@ux @uy @ux
rxx ¼ k þ þ 2l
@x @y @x

@ux @uy @uy
ryy ¼ k þ þ 2l ð6:2:3Þ
@x @y @y

@ux @uy
rxy ¼ ryx ¼ l þ
@y @x
where k and l are Lame’s constants. We employ the quiescent condition at an

initial time,

@ui
ui jt¼0 ¼ ¼ 0; i ¼ x; y ð6:2:4Þ
@t t¼0
and the convergence (radiation) condition at infinity

@ui @ui
ui jpxffiffiffiffiffiffiffiffi
ffi
2 þy2 !1 ¼ ¼ ffi ¼ 0; i ¼ x; y ð6:2:5Þ
ffi
x2 þy2 !1
Equations from (6.2.1) to (6.2.5) constitute the impulsive Lamb’s problem

(Fung 1970) for the 2D semi-infinite elastic solid. Substituting Hooke’s law into the
equations of motion, the displacement equations are obtained as
@ 2 ux @ 2 ux @ 2 uy 1 @ 2 ux
c2 þ þ ðc 2
1Þ ¼
@x2 @y2 @x@y c2s @t2
ð6:2:6Þ
@ 2 ux @ 2 uy @ 2 uy 1 @ 2 uy
ðc 1Þ
2
þ 2 þ c2 2 ¼ 2 2
@x@y @x @y cs @t
where c is the velocity ratio defined by Eq. (3.8) in Chap. 3. We apply the double
integral transform: Laplace transform with respect to the time t,
Z1

0
and Fourier transform with respect to the space variable x,
Zþ1 Zþ1
2p
1 1
With use of the quiescent and radiation conditions, the displacement equations are
transformed to the coupled ordinary differential equations with constant
coefficients,
d 2 ux 2 duy
fc 2 2
n þ ðs=c s Þ gu x inðc 2
1Þ ¼0
dy2 dy
ð6:2:9Þ
du d 2 uy
inðc2 1Þ x þ c2 2 fn2 þ ðs=cs Þ2 guy ¼ 0
dy dy
Hooke’s law and the boundary condition are also transformed to
1 1 du
xy ¼ r
r yx ¼ x in uy
l l dy
ð6:2:10Þ
1 duy
yy ¼ inðc2 2Þux þ c2
r
l dy
and

yx
r ¼ Px ; yy
r ¼ Py ð6:2:11Þ
y¼0 y¼0
The solution for the displacement equation (6.2.9) is given by
in
ux ¼ Aðn; sÞ expðad yÞ þ Bðn; sÞ expðas yÞ
ad
ð6:2:12Þ
in
uy ¼ Aðn; sÞ expðad yÞ Bðn; sÞ expðas yÞ
as
where Aðn; sÞ and Bðn; sÞ are unknown coefficients, and the two radicals are defined
by
aj ¼ n2 þ ðs=cj Þ2 ; j ¼ d; s ð6:2:13Þ
with the radiation condition Reðaj Þ [ 0. Substituting Eq. (6.2.12) into Eq. (6.2.10),
the stress components are obtained as
1 a2 þ n2
yx ¼ 2inAðn; sÞ expðad yÞ s
r Bðn; sÞ expðas yÞ
l as
ð6:2:14Þ
1 a2 þ n2
yy ¼ s
r Aðn; sÞ expðad yÞ þ 2inBðn; sÞ expðas yÞ
l ad
Applying the boundary condition of Eq. (6.2.11) to the above stresses, we obtain
the following simple algebraic equations for the unknown coefficients,
a2s þ n2
2inAðn; sÞ Bðn; sÞ ¼ Px =l
as
ð6:2:15Þ
a2 þ n2
s Aðn; sÞ þ 2inBðn; sÞ ¼ Py =l
ad
Their solutions are given by
2inad as ðPx =lÞ þ ad ða2s þ n2 ÞðPy =lÞ

Aðn; sÞ ¼
Rðn; sÞ
ð6:2:16Þ
as ða2s þ n2 ÞðPx =lÞ 2inad as ðPy =lÞ
Bðn; sÞ ¼
Rðn; sÞ
where Rðn; sÞ is Rayleigh equation, defined by
Rðn; sÞ ¼ ða2s þ n2 Þ2 4ad as n2 ð6:2:17Þ
Substitution of Eq. (6.2.16) into Eq. (6.2.12) gives the displacement in the double
transformed domain. The transformed displacement can also be expressed as the
sum of the dilatational and shear wave contributions,
Px nðdÞ o P n o
Ixx ðn; y; sÞ þ Ixx
y ðdÞ
ux ¼
ðsÞ
ðn; y; sÞ þ Ixy ðn; y; sÞ þ Ixy
ðsÞ
ðn; y; sÞ
l l
Px nðdÞ o P n o
Iyx ðn; y; sÞ þ Iyx ðn; y; sÞ þ
y ðdÞ

y ¼
u ðsÞ
Iyy ðn; y; sÞ þ Iyy
ðsÞ
ðn; y; sÞ
l l
ð6:2:18Þ
ðkÞ
where the superscripts “d” and “s” for k in Iij stand for the dilatational and shear
wave contributions, respectively. The first subscript i stands for the direction of the
displacement component and the second j for the component of the load. More
explicit expressions for each of the contributions are
Ixx
ðdÞ 2n2 as
ðn; y; sÞ ¼ expðad yÞ
Rðn; sÞ
Ixx
ðsÞ as ða2s þ n2 Þ
ðn; y; sÞ ¼ expðas yÞ
Rðn; sÞ
ð6:2:19Þ
Ixy
ðdÞ inðas þ n Þ
2 2
Rðn; sÞ
Ixy
ðsÞ 2inad as
Rðn; sÞ
x , and
for u
Iyx
ðdÞ 2inad as
Rðn; sÞ
Iyx
ðsÞ inða2s þ n2 Þ
Rðn; sÞ
ð6:2:20Þ
Iyy
ðdÞ ad ða2s þ n2 Þ
Rðn; sÞ
Iyy
ðsÞ 2n2 ad
Rðn; sÞ
y .
for u
We have obtained explicit expressions for the transformed displacement com-
ponents. Our next task is to invert the displacement into the real domain ðx; y; tÞ.
However, it might be impossible to invert the displacement as a whole. The
inversion must be carried out for each wave contribution one by one. As all dila-
ðdÞ
tational wave contributions given by Iij have the same exponential function
ðsÞ
expðad yÞ and the shear wave contributions Iij does the function expðas yÞ, we
invert these two wave contributions separately.
(1) Inversion of the dilatational wave contribution
As a representative for the dilatational wave contribution, we consider the
ðdÞ
inversion of Ixx . The formal Fourier inversion integral with respect to the
parameter n is given by
Zþ1
ðdÞ 1 2n2 as
Ixx ðx; y; sÞ ¼ expðad y inxÞdn ð6:2:21Þ
2p Rðn; sÞ
1
In order to eliminate the Laplace transform parameter s in the integrand except

the exponential function, the variable transform from n to the new variable 1 is
introduced as
n ¼ ðs=cd Þ1 ð6:2:22Þ
The inversion integral is rewritten as
Zþ1 n pffiffiffiffiffiffiffiffiffiffiffiffiffi o
ðdÞ 1 ðdÞ
Ixx ðx; y; sÞ ¼ Fxx ð1Þ exp ðs=cd Þ y 12 þ 1 þ ix1 d1 ð6:2:23Þ
2p
1
where
ðdÞ 212 12 þ c2
Fxx ð1Þ ¼ ð6:2:24Þ
Rð1Þ
and its denominator, which is called the Rayleigh equation/function, is redefined by

pffiffiffiffiffiffiffiffiffiffiffiffiffipffiffiffiffiffiffiffiffiffiffiffiffiffiffi
Rð1Þ ¼ ð212 þ c2 Þ2 412 12 þ 1 12 þ c2 ð6:2:25Þ
Note that the Rayleigh equation Rð1Þ differs from Rðn; sÞ in Eq. (6.2.17).
Subsequently, we consider the complex integral U whose integrand is the same
as that in Eq. (6.2.23),
I n pffiffiffiffiffiffiffiffiffiffiffiffiffi o
1 ðdÞ
U¼ Fxx ð1Þ exp ðs=cd Þ y 12 þ 1 þ ix1 d1 ð6:2:26Þ
2p
C ðÞ
pffiffiffiffiffiffiffiffiffiffiffiffiffi pffiffiffiffiffiffiffiffiffiffiffiffiffiffi
The closed loops C are discussed here. The two radicals 12 þ 1 and 12 þ c2
in the integrand have each two branch points at 1 ¼ i; ic, respectively. Thus,
four branch cuts are introduced along the imaginary axis in the complex 1-plane as
shown in Fig. 6.5. Only the lower half plane is shown since the path and singular
point are symmetric about the real axis. The Rayleigh equation has two pure
imaginary roots, 1 ¼ icR , where cR is the velocity ratio of the dilatational wave cd
to the Rayleigh wave cR ,
cR ¼ cd =cR ð6:2:27Þ
This velocity ration cR is always greater than c, since the following inequality holds
for the isotropic media:
cR \cs \cd ð6:2:28Þ
As to the argument of the exponential function, we hope to transform it to the

product of the new (time) variable t and the transform parameter s. So, we introduce
the new variable t as
pffiffiffiffiffiffiffiffiffiffiffiffiffi
t ¼ y 12 þ 1 þ ix1 =cd ð6:2:29Þ
Solving for the variable 1, we have two solutions, the so-called “Cagniard’s path”,
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
ðÞ
ixðcd tÞ y ðcd tÞ2 ðx2 þ y2 Þ
1d ¼ ð6:2:30Þ
x2 þ y 2
C (−) x>0
−ix
ς=
Im(ς ) x2 + y 2
Re(ς )
ς = −i
ς d( + )
ς ( −)
d
ς = −iγ
ς = −iγ R
Fig. 6.5 Closed loop C(−) and Cagniard’s path for the dilatational wave contribution
ðÞ
These solutions 1d take the form of two symmetrical semi-hyperbolas in the
complex 1-plane as the real parameter t varies from x2 þ y2 =cd to infinity. The
two semi-hyperbolas are connected to each other at the saddle point
ðdÞ ix pffiffiffiffiffiffiffiffiffiffiffiffiffiffi
1saddle ¼ pffiffiffiffiffiffiffiffiffiffiffiffiffiffi at t ¼ x2 þ y2 =cd ð6:2:31Þ
x þy
2 2
This connected hyperbola is called Cagniard’s path as shown in Fig. 6.5. Fortu-
nately, the saddle point is not on any branch cut. Then, we can determine the closed
loop in the complex 1-plane.
When the Cagniard’s path is in the lower plane, the closed loop is composed of
an infinite line along the real axis, Caginiard’s path and two large arcs which
connect the line to the Cagniard’s path. When the Caginiard’s path is in the upper
plane, the closed loop is composed of the similar ones in the lower plane and is
symmetric with the lower loop about the real axis. The lower loop C(−), and the
branch cuts and points are shown in Fig. 6.5. The closed loop C(−) is employed
when the space variable x is positive, and the loop C(+) when it is negative, due to
the convergence at the large arc. In any case, positive or negative x, i.e. the upper or
lower circuit in the complex plane, the closed loop does not include any singular
point. Then we can apply Cauchy’s integral theorem to the complex integral U.
Since the integral along the large arc vanishes as its radius tends to infinity, we can
convert the integral along the real axis to that along the Cagniard’s path. That is
Zþ1 n pffiffiffiffiffiffiffiffiffiffiffiffiffi o
ðdÞ 1 ðdÞ
Ixx ðx; y; sÞ ¼ Fxx ð1Þ exp ðs=cd Þ y 12 þ 1 þ ix1 d1
2p
1
Zþ1 ( )
ðþÞ ðÞ
1 ðdÞ ðþÞ d1d ðdÞ ðÞ d1d
¼ Fxx ð1d Þ Fxx ð1d Þ expðstÞdt
2p pffiffiffiffiffiffiffiffiffi dt dt
x2 þy2 =cd
ð6:2:32Þ
If we understand that the integration variable t as the real time variable, the integral
has just the form of Laplace transform integral, but with the shifted starting time
pffiffiffiffiffiffiffiffiffiffiffiffiffiffi ðdÞ
x2 þ y2 =cd . As the Laplace inversion of Ixx ðx; y; sÞ is the original function
before Laplace transform, the inversion is carried out by inspection on the last line
of Eq. (6.2.32). Utilizing the step function, the original function is given by
( )
h i ðþÞ ðÞ pffiffiffiffiffiffiffiffiffiffiffiffiffiffi
1 ðdÞ 1 ðdÞ ðþÞ d1d ðdÞ ðÞ d1d
L Ixx ðx; y; sÞ ¼ Fxx ð1d Þ Fxx ð1d Þ H t x2 þ y2 =cd
2p dt dt
ð6:2:33Þ
ðdÞ
Consequently, the double inversion for Ixx ðn; y; sÞ is given by
( )
ðþÞ ðÞ pffiffiffiffiffiffiffiffiffiffiffiffiffiffi
ðdÞ 1 ðdÞ ðþÞ d1d ðdÞ ðÞ d1d
Ixx ðx; y; tÞ ¼ Fxx ð1d Þ Fxx ð1d Þ H t x2 þ y2 =cd
2p dt dt
ð6:2:34Þ
The same Cagniard-de Hoop technique can be applied to the other dilatational wave
ðdÞ
contributions Iij . Thus, we have the unified expression for the dilatational wave
contributions as
( )
ðdÞ 1 ðdÞ ðþÞ d1d ðdÞ ðÞ d1d
Iij ðx; y; tÞ ¼ Fij ð1d Þ Fij ð1d Þ H t x2 þ y2 =cd
2p dt dt
ð6:2:35Þ
ðdÞ
where Fij ð1Þ are given by
ðdÞ 12 þ c2
212 ðdÞ i1ð212 þ c2 Þ
Fxx ð1Þ¼ ; Fxy ð1Þ ¼
Rð1Þ Rð1Þ
pffiffiffiffiffiffiffiffiffiffiffiffiffipffiffiffiffiffiffiffiffiffiffiffiffiffiffi pffiffiffiffiffiffiffiffiffiffiffiffiffi ð6:2:36Þ
ðdÞ 2i1 1 þ 1 1 þ c
2 2 2
ðdÞ ð212 þ c2 Þ 12 þ 1
Fyx ð1Þ ¼ ; Fyy ð1Þ ¼
Rð1Þ Rð1Þ
(2) Inversion of the shear wave contribution

We consider the double inversion for the shear wave contribution, which has the
exponential function expðas yÞ. As a representative, we consider the inversion of
Ixx
ðsÞ as ða2s þ n2 Þ
ðn; y; sÞ ¼ expðas yÞ ð6:2:37Þ
Rðn; sÞ
Its formal Fourier inversion is given by
Zþ1
ðsÞ 1 as ða2s þ n2 Þ
Ixx ðx; y; sÞ ¼ expðas y inxÞdn ð6:2:38Þ
2p Rðn; sÞ
1
Introducing the variable transform defined by Eq. (6.2.22), the inversion integral is
rewritten as
Zþ1 n pffiffiffiffiffiffiffiffiffiffiffiffiffiffi o
ðsÞ 1 ðsÞ
Ixx ðx; y; sÞ ¼ Fxx ð1Þ exp ðs=cd Þ y 12 þ c2 þ i1x d1 ð6:2:39Þ
2p
1
where
ðsÞ ð212 þ c2 Þ 12 þ c2
Fxx ð1Þ ¼ ð6:2:40Þ
Rð1Þ
and the Rayleigh equation Rð1Þ is defined by Eq. (6.2.25).

We shall consider the complex integral U whose integrand is the same as that in
Eq. (6.2.39),
I n pffiffiffiffiffiffiffiffiffiffiffiffiffiffi o
1 ðsÞ
U¼ Fxx ð1Þ exp ðs=cd Þ y 12 þ c2 þ i1x d1 ð6:2:41Þ
2p
C ðÞ
The closed loop CðÞ is discussed here. Four branch points at 1 ¼ i; ic are
found from the two radicals, and corresponding branch cuts are introduced along
the imaginary axis in the 1-plane. The denominator, i.e. the Rayleigh equation, has
two symmetric zeros at 1 ¼ icR . These zeros give the simple poles for the inte-
grand and the poles are on the imaginary axis (on the branch cut), but is greater in
magnitude than any of the branch points. The branch points, cuts and poles are
shown in Fig. 6.6a, b.
As to the Cagniard’s path, we introduce a new parameter t, defined by
t ¼ y 12 þ c2 þ i1x =cd ð6:2:42Þ
Its inversion gives the Cagniard’s path in the complex 1-plane,

qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
ixðcd tÞ y ðcd tÞ2 c2 ðx2 þ y2 Þ
1ðÞ ¼ ð6:2:43Þ
s
x2 þ y 2
The saddle point at t ¼ x2 þ y2 =cs is
ðsÞ icx
1saddle ¼ pffiffiffiffiffiffiffiffiffiffiffiffiffiffi ð6:2:44Þ
x2 þ y2
Comparing the magnitude of the saddle point with those of two branch points, we
find that the saddle point is always smaller than the branch points 1 ¼ ic of the
shear wave, but greater than the branch points 1 ¼ i of the dilatational wave, if the
inequality
cjxj pffiffiffiffiffiffiffiffiffiffiffiffiffi
pffiffiffiffiffiffiffiffiffiffiffiffiffiffi [ 1 ) y\jxj c2 1 ð6:2:45Þ
x2 þ y2
holds. Thus, we have two different closed loops for the complex integral.
When the saddle point is smaller in magnitude than the branch points 1 ¼ i, the
Cagniard’s path does not cross any branch cut; the closed loop is then composed of the
line along the real axis, the Cagniard’s path, and two large arcs which connect the line
with the Cagniard’s path. This closed loop is similar to that in the case of the dilata-
tional wave contribution and the Cagniard’s path is denoted by the path I in Fig. 6.6a.
On the other hand, when the saddle point is greater in magnitude than the branch
point of the dilatational wave, i.e. the inequality of Eq. (6.2.45) holds, the
Cagniard’s path crosses the branch cut. The Cagniard’s path has to be deformed
along the imaginary axis, so that the closed loop does not include any branch points
and cuts. This deformed Cagniard’s path is shown as the path II in Fig. 6.6b, where
the regular Cagniard’s path is deformed by two short lines along the dilatational
branch cut and a small circle around the dilatational branch point. Thus, the closed
loop for this case is composed of the line along the real axis, the deformed
Cagniard’s path II, and two large arcs. Then, two closed loops are considered for
the complex integral Φ and its choice depends on the inequality in Eq. (6.2.45).
(a)
γx
<1 x>0
x2 + y 2
Im(ς )
Re(ς )
ς s( − ) ς = −i
ς s( + )
ς = −iγ
ς = −iγ R
(b) γx
>1 x>0
x2 + y 2
Im(ς ) Re(ς )
ε →0
ς = −i
ς s( − ) ς s( + )
ς = −iγ
ς = −iγ R
Fig. 6.6 Two closed loops for the 2D Cagniard’s technique. a Loop with Cagniard’s path I.
b Loop with deformed Cagniard’s path II
When the inequality of Eq. (6.2.45) does not hold, we employ the closed loop
with the Cagniard’s path I, and the Cauchy’s integral theorem is applied to the
complex integral Φ in Eq. (6.2.41). The integral along the real axis is converted to
that along the Cagniard’s path I, since no singular point is included in the loop and
the integral along the large arc vanishes. That is,
ðsÞ 1 ðsÞ
Ixx ðx; y; sÞ ¼ Fxx ð1Þ exp ðs=cd Þ y 12 þ c2 þ i1x d1
2p
1
Z1 ( )
ðþÞ ðÞ
1 ðsÞ ðþÞ d1s ðsÞ ðÞ d1s
¼ Fxx ð1s Þ Fxx ð1s Þ expðstÞdt
x2 þy2 =cs
ð6:2:46Þ
Then, the Laplace inversion is carried out by inspection and the double inversion
ðsÞ
for Ixx is given by
ðsÞ
Ixx ðx; y; tÞ
( ) !
1
ðþÞ ðÞ pffiffiffiffiffiffiffiffiffiffiffiffiffiffi c j xj
ðsÞ ðþÞ d1s ðsÞ ðÞ d1s
¼ Fxx ð1s Þ Fxx ð1s Þ H t x þ y =cs H 1 pffiffiffiffiffiffiffiffiffiffiffiffiffiffi
2 2
2p dt dt x2 þ y2
ð6:2:47Þ
where the last step function is the conditional for the inequality.
When the inequality of Eq. (6.2.45) holds, we have to employ the loop with the
Cagniard’s path II. In addition to the regular hyperbolic path, the deformed
Cagniard’s path II has one small circle around the branch point 1 ¼ i and two
short lines along the branch cut. We perform the complex integral along the closed
loop shown in Fig. 6.6b. The integral around the small circle vanishes as its radius
tends to zero, but the line integral along the branch cut does not vanish and must be
discussed. As was discussed in Sect. 1.3.2 in Chap. 1, the radical for the dilatational
wave, 12 þ 1, has different arguments, depending on the side of the cut. How-
ever, the radical 12 þ c2 for the shear wave does not change its argument since
the saddle point is smaller in magnitude than the branch points ic. The sign and
argument of the two radicals on the deformed Cagniard’s path II are summarized in
Table 6.1, where the new radicals are introduced as
pffiffiffiffiffiffiffiffiffiffiffiffiffi pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
bd ¼ g2 1 ; bs ¼ c 2 g2 ð6:2:48Þ
Using the new notations, the complex integral U in Eq. (6.2.45) is carried out
along the closed loop with the deformed Cagniard’s path II. Since no singular point
is included in the loop and the integral along the large arc vanishes, the line integral
Table 6.1 Argument and Radicals Path AB: 1 ¼ ig Path DE: 1 ¼ ig
magnitude of the radical and pffiffiffiffiffiffiffiffiffiffiffiffiffi pffiffiffiffiffiffiffiffiffiffiffiffiffi pffiffiffiffiffiffiffiffiffiffiffiffiffi
Rayleigh functions 1 þ 1 þi g2 1 ¼ þibd
2 i g2 1 ¼ ibd
pffiffiffiffiffiffiffiffiffiffiffiffiffiffi pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
12 þ c2 þ c2 g2 ¼ þbs þ c2 g2 ¼ þbs
Rð1Þ ðc2 2g2 Þ2 þ 4ig2 bd bs ðc2 2g2 Þ2 4ig2 bd bs
pffiffiffiffiffiffiffiffiffiffiffiffiffi pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
bd ¼ g2 1; bs ¼ c2 g2
along the real axis is converted to that along the Cagniard’s path II. The deformed
Cagniard path is composed of two lines along the imaginary axis with different
argument for the radicals and the regular Cagniard’s path. Since the integral along
\
the small circle BCD vanishes as its radius tends to zero, the Fourier inversion
integral, i.e. the integral along the real axis, is converted to the three integrals
ðsÞ 1 ðsÞ
2p
1
Zþ1 ( )
ðþÞ ðÞ
x2 þy2 =cs
Z n pffiffiffiffiffiffiffiffiffiffiffiffiffiffi o
1 ðsÞ
þ Fxx ð1Þ exp ðs=cd Þ y 12 þ c2 þ i1x d1
2p
AB
1 ðsÞ
2p
DE
ð6:2:49Þ
The last two integrals along the branch line are rearranged further. With the use of
Table 6.1, the two branch line integrals are unified as
1 ðsÞ
Fxx ð1Þ exp ðs=cd Þ y 12 þ c2 þ i1x d1
2p
AB
1 ðsÞ
2p ð6:2:50Þ
DE
pffiffiffiffiffiffiffi
cx
Zx2 þy2
1
¼ fxxðsÞ ðgÞ expfðs=cd Þðbs y þ xgÞgdg
2p
1
where the new integrand is
8g2 ðc2 2g2 Þbd b2s

fxxðsÞ ðgÞ ¼ ð6:2:51Þ
ðc2 2g2 Þ4 þ 16g4 b2d b2s
In order to transform the last integral in Eq. (6.2.50) to the form of Laplace transform
integral, we introduce the variable transform from g to the new variable t as
t ¼ ðbs y þ xgÞ=cd ð6:2:52Þ

Solving for g, we have

xðcd tÞ y c2 ðx2 þ y2 Þ ðcd tÞ2 ð6:2:53Þ
gH ¼
x2 þ y 2
Then, the last integral in Eq. (6.2.50) is transformed to that of the Laplace
transform,
pffiffiffiffiffiffiffi
cx
Zx2 þy2
1
fxxðsÞ ðgÞ expfðs=cd Þðbs y þ xgÞgdg
2p
1
ffi ð6:2:54Þ
x2 þy2 =cs
Z
1 dgH
¼ fxxðsÞ ðgH Þ expðstÞdt
2p pffiffiffiffiffiffiffiffi dt
xþy c2 1 =cd
Finally, we have converted the Fourier inversion integral to the form of Laplace
transform integral as
ðsÞ 1 ðsÞ
2p
1
Zþ1 ( )
ðþÞ ðÞ
x2 þy2 =cs
pffiffiffiffiffiffiffiffiffi
x2 þy2 =cs
Z
1 dgH
þ fxxðsÞ ðgH Þ expðstÞdt
2p pffiffiffiffiffiffiffiffi dt
xþy c2 1 =cd
ð6:2:55Þ
The Laplace inversion is carried out by inspection. The branch line integral of
Eq. (6.2.54) appears only when the conditional of Eq. (6.2.45) holds. Using the step
function for this conditional, the Laplace inversion is finally expressed by
ðsÞ
Ixx ðx; y; tÞ
!" ( )
cx 1
¼ H pffiffiffiffiffiffiffiffiffiffiffiffiffiffi 1 Fxx ð1s Þ Fxx ð1s Þ H t x2 þ y2 =cs
x þy
2 2 2p dt dt
pffiffiffiffiffiffiffiffiffiffiffiffiffi!#
1 ðsÞ dgH pffiffiffiffiffiffiffiffiffiffiffiffiffiffi x þ y c2 1
þ fxx ðgH Þ H x þ y =cs t H t
2 2
2p dt cd
ð6:2:56Þ
Comparing this expression with that of Eq. (6.2.47), we see that the above
equation includes the full conditional. Thus, the final result for the double inver-
sions of the shear wave contribution is given by
ðsÞ
Ixx ðx; y; tÞ
( )
1
¼ Fxx ð1s Þ Fxx ð1s Þ H t x2 þ y2 =cs
2p dt dt
! pffiffiffiffiffiffiffiffiffiffiffiffiffi!
1 ðsÞ dgH cx pffiffiffiffiffiffiffiffiffiffiffiffiffiffi x þ y c2 1
þ fxx ðgH Þ H pffiffiffiffiffiffiffiffiffiffiffiffiffiffi 1 H x2 þ y2 =cs t H t
2p dt x2 þ y 2 cd
ð6:2:57Þ
The same inversion technique is applied to another double inversion for the shear
wave contribution and we have the unified expression for the shear wave contri-
bution as
ðsÞ
Iij ðx; y; tÞ
( )
1
¼ Fij ð1s Þ Fij ð1s Þ H cs t x2 þ y2
2p dt dt
!
1 ðsÞ dgH cjxj pffiffiffiffiffiffiffiffiffiffiffiffiffiffi pffiffiffiffiffiffiffiffiffiffiffiffiffi
þ fij ðgH Þ H pffiffiffiffiffiffiffiffiffiffiffiffiffiffi 1 H x2 þ y2 cs t H cd t jxj y c2 1
2p dt x2 þ y2
ð6:2:58Þ
where the integrands are given by

pffiffiffiffiffiffiffiffiffiffiffiffiffiffi pffiffiffiffiffiffiffiffiffiffiffiffiffipffiffiffiffiffiffiffiffiffiffiffiffiffiffi
ðsÞ ð212 þ c2 Þ 12 þ c2 ðsÞ 2i1 12 þ 1 12 þ c2
Fxx ð1Þ ¼ ; Fxy ¼ð1Þ
Rð1Þ Rð1Þ
pffiffiffiffiffiffiffiffiffiffiffiffiffi ð6:2:59Þ
ðsÞ i1ð212 þ c2 Þ ðsÞ 212 12 þ 1
Fyx ð1Þ ¼ ; Fyy ð1Þ ¼
Rð1Þ Rð1Þ
8g2 ðc2 2g2 Þbd b2s 4gðc2 2g2 Þ2 bd bs

fxxðsÞ ðgÞ ¼ ; fxyðsÞ ðgÞ ¼
RH ðgÞ RH ðgÞ
ð6:2:60Þ
8g3 ðc2 2g2 Þbd bs 4g2 ðc2 2g2 Þ2 bd
fyxðsÞ ðgÞ ¼ ; fyyðsÞ ðgÞ ¼
RH ðgÞ RH ðgÞ
and
RH ðgÞ ¼ ðc2 2g2 Þ4 þ 16g4 b2d b2s ð6:2:61Þ


We have just inverted the two wave contributions and thus the displacement
response is given by
Px n ðdÞ ðsÞ
o P n
y ðdÞ ðsÞ
o
ux ¼ Ixx ðx; y; tÞ þ Ixx ðx; y; tÞ þ Ixy ðx; y; tÞ þ Ixy ðx; y; tÞ
l l
ð6:2:62Þ
Px n ðdÞ ðsÞ
o P n
y ðdÞ ðsÞ
o
uy ¼ Iyx ðx; y; tÞ þ Iyx ðx; y; tÞ þ Iyy ðx; y; tÞ þ Iyy ðx; y; tÞ
l l
When we express the displacement in terms of Green’s dyadic Gij ðx; y; tÞ,
ux ¼ Px Gxx ðx; y; tÞ þ Py Gxx ðx; y; tÞ

ð6:2:63Þ
uy ¼ Px Gyx ðx; y; tÞ þ Py Gyy ðx; y; tÞ
the dyadic is given by
1 n ðdÞ ðsÞ
o
Gij ðx; y; tÞ ¼ Iij ðx; y; tÞ þ Iij ðx; y; tÞ ð6:2:64Þ
l
A more detailed expression is
Gij ðx; y; tÞ
"( )
ðþÞ ðÞ
1 ðdÞ ðþÞ d1 ðdÞ ðÞ d1
¼ Fxx ð1d Þ d Fxx ð1d Þ d Hðcd t rÞ
2pl dt dt
( )
ðþÞ ðÞ
þ Fij ð1s Þ Fij ð1s Þ Hðcs t rÞ
dt dt
ðsÞ dgH pffiffiffiffiffiffiffiffiffiffiffiffiffi
þ fij ðgH Þ Hðcjxj rÞHðr cs tÞH cd t jxj y c2 1
dt
ð6:2:65Þ
where the radial distance r ¼ x2 þ y2 is introduced.
It should be noticed that the step function Hðcd t rÞ shows the disturbed cir-
cular region produced by the dilatational wave and Hðcs t rÞ does the region
disturbed by the shear wave. These two circular waves are emanating from the
source point. The triple product of the step functions
Hðcjxj rÞHðr cs tÞH cd t jxj y c2 1
gives the region disturbed by the shear wave which is induced by the precursor
dilatational wave. We call this region “von Schmidt wave”. Its wave region is
shown in Fig. 6.7.
cd t − | x |
y=
γ 2 −1
cs t = r
y = γ 2 −1 | x |
(γ | x |= r )
cd t = r
y
Fig. 6.7 von Schmidt wave region
6.3 3D Dynamic Lamb’s Problem
The transient response of a fully 3D elastic half space is discussed here. We take the
3D coordinate system (x, y, z). The surface is denoted by z ¼ 0 and the interior by
z [ 0 as shown in Fig. 6.8. The 3D deformation of the isotropic elastic solid is
governed by equations of motion,
@rxx @ryx @rzx @ 2 ux

þ þ ¼q 2
@x @y @z @t
@rxy @ryy @rzy @ 2 uy
þ þ ¼q 2 ð6:3:1Þ
@x @y @z @t
@rxz @ryz @rzz @ 2 uz
þ þ ¼q 2
@x @y @z @t
and Hooke’s law for the isotropic elastic solid,

@ux @uy @uz @ux
rxx ¼ k þ þ þ 2l
@x @y @z @x

@ux @uy @uz @uy
ryy ¼ k þ þ þ 2l
@x @y @z @y

@ux @uy @uz @uz
rzz ¼ k þ þ þ 2l
@x @y @z @z
6.3 3D Dynamic Lamb’s Problem 179
Fig. 6.8 A suddenly applied σ zz z =0

= Pzδ ( x)δ ( y ) H (t ) x
load Pi on the surface of a
semi-infinite elastic solid
σ zy z =0
= Pyδ ( x)δ ( y ) H (t ) z=0
σ zx = Pxδ ( x)δ ( y ) H (t )
z =0
λ, μ, ρ

@ux @uy
rxy ¼ ryx ¼ l þ
@y @x

@uz @uy
ryz ¼ rzy ¼ l þ
@y @z

@ux @uz
rzx ¼ rxz ¼ l þ ð6:3:2Þ
@z @x
where k and l are Lame’s constants.

We assume that a point load is suddenly applied at the coordinate origin on the
surface z ¼ 0, i.e.
rzx jz¼0 ¼ Px dðxÞdðyÞHðtÞ

rzy z¼0 ¼ Py dðxÞdðyÞHðtÞ ð6:3:3Þ
rzz jz¼0 ¼ Pz dðxÞdðyÞHðtÞ
where Pj, j = x, y, z are components of the load. Note that the loading time function is
Heaviside’s unit step function, not Dirac’s delta function. This is because that we get the
simpler solution for the dynamic response. The quiescent condition at an initial time

@ui
ui jt¼0 ¼ ¼ 0; i ¼ x; y; z ð6:3:4Þ
@t t¼0

@ui @ui @ui
ui jpffiffiffiffiffiffiffiffiffiffiffiffiffiffi ¼ ¼ ¼ ¼ 0; i ¼ x; y; z
2x þy 2 þz !1
2
@x pffiffiffiffiffiffiffiffiffiffiffiffiffiffi
x2 þy2 þz2 !1 @y pxffiffiffiffiffiffiffiffiffiffiffiffiffiffi
2 þy2 þz2 !1 @z pxffiffiffiffiffiffiffiffiffiffiffiffiffiffi
2 þy2 þz2 !1
ð6:3:5Þ
are employed.
Substituting Hooke’s law of Eq. (6.3.2) into the equation of motion (6.3.1), the
displacement equations are given by
2
@ 2 ux @ 2 ux @ 2 ux @ uy @ 2 uz 1 @ 2 ux
c2
þ þ þ ðc 2
1Þ þ ¼
@x2 @y2 @z2 @x@y @x@z c2s @t2
2
@ uy
2
2 @ uy
2
@ uy
2
@ ux @ uz
2
1 @ 2 uy
þ c þ þ ðc 2
1Þ þ ¼ ð6:3:6Þ
@x2 @y2 @z2 @x@y @y@z c2s @t2
2
@ uz @ uz
2 2
2 @ uz
2
@ ux @ uy
2
1 @ 2 uz
þ þ c þ ðc 2
1Þ þ ¼
@x2 @y2 @z2 @x@z @y@z c2s @t2
The present 3D elastodynamic problem is also discussed by the integral trans-

form method. With use of the quiescent and convergence conditions, we apply
Laplace transform with respect to the time, defined by
Z1

0
and the double Fourier transform with respect to two space variables x and y,
defined by
Zþ1 Zþ1
~
f ðn; gÞ ¼ f ðx; yÞ expðþinx þ igyÞdxdy;
1 1
ð6:3:8Þ
Zþ1 Zþ1
1 ~f ðn; gÞ expðinx igyÞdndg
f ðx; yÞ ¼
ð2pÞ2
1 1
to the displacement equations (6.3.6) and Hooke’s law of Eq. (6.3.2). The dis-
placement equations are transformed to the coupled ordinary differential equations,
d2 ~
ux
2 ~ d ~uz
fc 2 2
n þ g 2
þ ðs=c s Þ g
u x ngðc 2
1Þ ~

u
y inðc 2
1Þ ¼0
dz2 dz
d2 ~
uy
2 ~ d ~uz
fn 2
þ c 2 2
g þ ðs=c s Þ g
u y ngðc 2
1Þ ~

u
x igðc 2
1Þ ¼ 0 ð6:3:9Þ
dz2 dz
d2~u
d~u
d ~uy
c2 2z fn2 þ g2 þ ðs=cs Þ2 g~uz inðc2 1Þ x igðc2 1Þ ¼0
dz dz dz
The stress components to be used for the boundary condition are transformed as
follows:
1 d u~
~zx ¼ x in~uz
r
l dz
1 d ~uy
r~ ¼ ig~uz ð6:3:10Þ
l zy dz
1 d ~u
~zz ¼ c2 z inðc2 2Þ~ux igðc2 2Þ~uy
r
l dz
The boundary condition on the surface is also transformed to
1 P
~ ¼ j ;
r j ¼ x; y; z ð6:3:11Þ
l zj z¼0 ls
The general solution for the displacement equations (6.3.9) is obtained as
in
~
ux ¼
Aðn; g; sÞ expðad zÞ þ Bðn; g; sÞ expðas zÞ
ad
ig
~
uy ¼ Aðn; g; sÞ expðad zÞ þ Cðn; g; sÞ expðas zÞ
ð6:3:12Þ
ad
i
~
uz ¼ Aðn; g; sÞ expðad zÞ fnBðn; g; sÞ þ gCðn; g; sÞg expðas zÞ

as
The stress components are
1 1 2
~zx ¼ 2inAðn; g; sÞ expðad zÞ

r ðas þ n2 ÞBðn; g; sÞ þ ngCðn; g; sÞ expðas zÞ
l as
1 1
r~ ¼ 2igAðn; g; sÞ expðad zÞ ngBðn; g; sÞ þ ða2s þ n2 ÞCðn; g; sÞ expðas zÞ

l zy as
1 a2 þ n2 þ g2
~zz ¼ s
r Aðn; g; sÞ expðad zÞ þ 2ifnBðn; g; sÞ þ gCðn; g; sÞg expðas zÞ
l ad
ð6:3:13Þ
where Aðn; g; sÞ; Bðn; g; sÞ and Cðn; g; sÞ are unknown coefficients to be determined
by the boundary condition. The radicals are defined by
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
aj ¼ n2 þ g2 þ ðs=cj Þ2 ; Reðaj Þ [ 0; j ¼ d; s ð6:3:14Þ
where the conditional Reðaj Þ [ 0 guarantees the convergence condition at infinity.

Applying the boundary condition (6.3.11) to the stresses of Eq. (6.3.13), we obtain
the following simple algebraic equations for the coefficients, Aðn; g; sÞ; Bðn; g; sÞ
and Cðn; g; sÞ,
1 2
Px
2inAðn; g; sÞ ðas þ n2 ÞBðn; g; sÞ þ ngCðn; g; sÞ ¼
as ls
1
Py
2igAðn; g; sÞ ngBðn; g; sÞ þ ða2s þ n2 ÞCðn; g; sÞ ¼ ð6:3:15Þ
as ls
a2s þ n2 þ g2 Pz
Aðn; g; sÞ þ 2ifnBðn; g; sÞ þ gCðn; g; sÞg ¼
ad ls
The coefficients are determined as
2inad as Px 2igad as Py ad ða2s þ n2 þ g2 Þ Pz

Aðn; g; sÞ ¼ ð6:3:16Þ
sRðn; g; sÞ l sRðn; g; sÞ l sRðn; g; sÞ l

n2 ða2s þ n2 þ g2 4ad as Þ 1 Px
Bðn; g; sÞ ¼
sas Rðn; g; sÞ sas l
ð6:3:17Þ
ngða2s þ n2 þ g2 4ad as Þ Py 2inad as Pz
þ þ
sas Rðn; g; sÞ l sRðn; g; sÞ l
ngða2s þ n2 þ g2 4ad as Þ Px
Cðn; g; sÞ ¼
sas Rðn; g; sÞ l
2 2 ð6:3:18Þ
g ðas þ n þ g 4ad as Þ
2 2
1 Py 2igad as Pz
þ þ
sas Rðn; g; sÞ sas l sRðn; g; sÞ l
where Rayleigh equation is defined by
Rðn; g; sÞ ¼ ðn2 þ g2 þ a2s Þ2 4ad as ðn2 þ g2 Þ ð6:3:19Þ
Then, the transformed displacement components are given by
Px n~I ðdÞ ~I ðsÞ ðn; g; z; sÞ þ ~I ðSHÞ ðn; g; z; sÞ

o
~
ux ¼
ðn; g; z; sÞ þ
l xx xx xx
n
Py ~ðdÞ o
I ðn; g; z; sÞ þ ~I xy ðn; g; z; sÞ
ðsÞ
þ ð6:3:20aÞ
l xy
Pz n~ðdÞ o
I xz ðn; g; z; sÞ þ ~I xz ðn; g; z; sÞ
ðsÞ
þ
l
Px n~ o
ðn; g; z; sÞ þ ~I yx ðn; g; z; sÞ
ðdÞ
I yx ðsÞ
~
uy ¼

l
Py n~ I ðdÞ ~I ðsÞ ðn; g; z; sÞ þ ~I ðSHÞ ðn; g; z; sÞ
o
þ ðn; g; z; sÞ þ ð6:3:20bÞ
l yy yy yy
Pz ~n o
þ I ðdÞ ðn; g; z; sÞ þ ~I ðsÞ
yz ðn; g; z; sÞ
l yz
Px n~ðdÞ o
I zx ðn; g; z; sÞ þ ~I zx ðn; g; z; sÞ
ðsÞ
~uz ¼
l
Py n~ðdÞ o
I zy ðn; g; z; sÞ þ ~I zy ðn; g; z; sÞ
ðsÞ
þ ð6:3:20cÞ
l
Pz n~ðdÞ o
I zz ðn; g; z; sÞ þ ~I zz ðn; g; z; sÞ
ðsÞ
þ
l
where
~ ðdÞ
I xx 2n2 as
ðn; g; z; sÞ ¼ expðad zÞ
sRðn; g; sÞ
~ ðsÞ
I xx n2 ða2s þ n2 þ g2 4ad as Þ
ðn; g; z; sÞ ¼ expðas zÞ
sas Rðn; g; sÞ
~
I ðSHÞ 1
xx ðn; g; z; sÞ ¼ expðas zÞ
sas
~ ðdÞ
I xy 2ngas
ðn; g; z; sÞ ¼ expðad zÞ ð6:3:21aÞ
sRðn; g; sÞ
~ ðsÞ
I xy ngða2s þ n2 þ g2 4ad as Þ
sas Rðn; g; sÞ
~ ðdÞ
I xz inða2s þ n2 þ g2 Þ
sRðn; g; sÞ
~ ðsÞ
I xz 2inad as
sRðn; g; sÞ
~
I ðdÞ 2ngas
yx ðn; g; z; sÞ ¼ expðad zÞ
sRðn; g; sÞ
~
I ðsÞ ngða2s þ n2 þ g2 4ad as Þ
yx ðn; g; z; sÞ ¼ expðas zÞ
sas Rðn; g; sÞ
~
I ðdÞ 2g2 as
yy ðn; g; z; sÞ ¼ expðad zÞ
sRðn; g; sÞ
~
I ðsÞ g2 ða2s þ n2 þ g2 4ad as Þ ð6:3:21bÞ
yy ðn; g; z; sÞ ¼ expðas zÞ
sas Rðn; g; sÞ
~ ðSHÞ
I yy 1
sas
~
I ðdÞ igða2s þ n2 þ g2 Þ
yz ðn; g; z; sÞ ¼ expðad zÞ
sRðn; g; sÞ
~
I ðsÞ 2igad as
yz ðn; g; z; sÞ ¼ expðas zÞ
sRðn; g; sÞ
~ ðdÞ
I zx 2inad as
sRðn; g; sÞ
~ ðsÞ
I zx inða2s þ n2 þ g2 Þ
sRðn; g; sÞ
~ ðdÞ
I zy 2igad as
sRðn; g; sÞ
ð6:3:21cÞ
~ ðsÞ
I zy igða2s þ n2 þ g2 Þ
sRðn; g; sÞ
~ ðdÞ
I zz ad ða2s þ n2 þ g2 Þ
sRðn; g; sÞ
~ ðsÞ
I zz 2ad ðn þ g Þ
2 2
sRðn; g; sÞ
In the above equations, the superscript k in ~I ij indicates the wave type. The
ðkÞ
superscript “d” stands for the dilatational wave (P-wave), “s” for the vertically
polarized shear wave (SV-wave), and “SH” for the horizontally polarized shear
wave (SH-wave). The meaning of the subscripts i and j are the same as those in the
2D problem. Note that the SH-wave contribution appears only in the horizontal
components and that no SH-wave contribution is included in the vertical compo-
nent which is normal to the surface.
Now, we shall invert each wave component separately.
(1) Inversion of P wave contribution: ~I ij ðn; g; z; sÞ

ðdÞ
As a representative, we consider the inversion of ~I xx ðn; g; z; sÞ. Its formal

ðdÞ
Fourier double inversion with respect to the parameters n and g is given by
Zþ1 Zþ1
ðdÞ 1 2n2 as
Ixx ðx; y; z; sÞ ¼ expðad z inx igyÞdndg ð6:3:22Þ
ð2pÞ2 sRðn; g; sÞ
1 1
We introduce the variable transform from ðn; gÞ to the new variables (p, q) defined
by
n ¼ ðs=rÞðxp yqÞ; g ¼ ðs=rÞðyp þ xqÞ ð6:3:23Þ

where the horizontal distance defined by r ¼ x2 þ y2 is introduced. This variable
transform corresponds to the rotation of the axes ðn; gÞ. The equivalent area element
is given by dndg ¼ s2 dpdq. The double integral in Eq. (6.3.22) is converted to
another double integral,
Zþ1 Zþ1
ðdÞ 1 2ðx2 p2 þ y2 q2 2xypqÞbs
Ixx ðx; y; z; sÞ ¼ 2
expfsðbd z þ irpÞgdpdq
ð2prÞ Rðp; qÞ
1 1
ð6:3:24Þ
where
Rðp; qÞ ¼ ðb2s þ p2 þ q2 Þ2 4bd bs ðp2 þ q2 Þ ð6:3:25Þ

qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
bj ¼ p2 þ q2 þ 1=c2j ; Reðbj Þ [ 0; j ¼ d; s ð6:3:26Þ
Note that the Rayleigh equation R(p, q) is different from the former Rayleigh
equation Rðn; g; sÞ.
The integrand in Eq. (6.3.24) includes both odd and even functions of the new
variable q. As the integration with respect to q is carried out over the whole range
from 1 to þ1, the integral whose integrand has an odd power (or the anti-
symmetric function of q) vanishes. Equation (6.3.24) is then slightly simplified, to
Z1 Zþ1
ðdÞ ðdÞ
Ixx ðx; y; z; sÞ ¼ dq Fxx ðp; qÞ expfsðbd z þ irpÞgdp ð6:3:27Þ
0 1
where the new notation for the integrand is introduced as
ðdÞ ðx2 p2 þ y2 q2 Þbs

Fxx ðp; qÞ ¼ ð6:3:28Þ
ðprÞ2 Rðp; qÞ
The inner integral in Eq. (6.3.27) is in a convenient form for the application of
Cagniard-de Hoop technique; thus we apply the technique to the inner integral with
respect to the variable p.
Now, we shall consider the complex integral U whose integrand is the same as
that of the inner integral in Eq. (6.3.27),
Z
ðdÞ
U¼ Fxx ðp; qÞ expfsðbd z þ irpÞgdp ð6:3:29Þ
C
In order to transform the argument of the exponential function to the simple product
of the new variable t and the Laplace transform parameter s, the new variable t is
introduced as
t ¼ bd z þ irp ð6:3:30Þ
Solving for p, the Cagniard’s path is given by

pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
ðÞ irt z t2 ðq2 þ 1=c2d ÞR2
pd ¼ ð6:3:31Þ
R2
and its gradient is
ðÞ pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
dpd zt ir t2 ðq2 þ 1=c2d ÞR2
¼ pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi ð6:3:32Þ
dt R2 t2 ðq2 þ 1=c2d ÞR2
where the 3D radial distance R is defined by

R¼ x2 þ y2 þ z 2 ð6:3:33Þ
The integrand has four branch points at

qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
p ¼ i q2 þ 1=c2d ; i q2 þ 1=c2s ð6:3:34Þ
From these branch points, four branch cuts are introduced along the imaginary axis
in the complex p-plane as was discussed in Sect. 1.3.2 in Chap. 1. Since the saddle
pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
point of the Cagniard’s path at t ¼ R q2 þ 1=c2d ,
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
ðdÞ ir
psaddle ¼ q2 þ 1=c2d ð6:3:35Þ
R
is smaller in magnitude than any of the branch points, the Cagniard’s path does not
cross any cut. In addition, the Rayleigh poles p ¼ i q2 þ 1=c2R , which are
derived from Rðp; qÞ ¼ 0, are on the imaginary axis and is greater in magnitude
than the saddle point and all branch points. Then, the closed loop C is composed of
the infinite line along the real p-axis, the Cagniard’s path, and two large arcs which
connect the Cagniard’s path to the line on the real axis.
The closed loop C is shown in Fig. 6.9. The complex integral U in Eq. (6.3.29) is
carried out along the closed loop. Since no singular point is included in the loop and
the integrals along the large arc vanish, the integral along the real axis is converted
to that along the Cagniard’s path, i.e.
Zþ1
ðdÞ
Fxx ðp; qÞ expfsðbd z þ irpÞgdp
1
Z1 ( ) ð6:3:36Þ
ðþÞ ðÞ
ðdÞ ðþÞ dp ðdÞ ðÞ dp
¼ Fxx ðpd ; qÞ d Fxx ðpd ; qÞ d expðstÞdt
ffi dt dt
R 2q þ1=cd 2
Closed loop C and Cagniard’s path
Im( p )
Re( p )
pd( − ) −i q2 + 1/ cd2
−i q 2 + 1/ cs2 pd( + )
−i q 2 + 1 / cR2
Fig. 6.9 Closed loop and Cagniard’s path for the complex integral U
Substituting the above integral into the inner integral in Eq. (6.3.27), we have
Z1 Zþ1 ( )
ðþÞ ðÞ
ðdÞ ðdÞ ðþÞ dpd ðdÞ ðÞ dp
Ixx ðx; y; z; sÞ ¼ dq Fxx ðpd ; qÞ Fxx ðpd ; qÞ d expðstÞdt
ffi dt dt
0 R q þ1=cd
2 2
ð6:3:37Þ
As the inner integral is just in the form of Laplace transform integral, we exchange
the order of integration so that the outer integral can be in the form of Laplace
transform. In discussing the supporting region for the double integral as shown in
Fig. 6.10, the exchange is symbolically carried out as
ðt=RÞ2 1=c2d
Z1 Z1 Z1 Z
dq gðt; qÞ expðstÞdt ¼ expðstÞdt gðt; qÞdq
0
ffi R=cd 0
R q þ1=cd
2 2
ð6:3:38Þ
Fig. 6.10 Supporting region t

for the double integral in
Eq. (6.3.38)
t t = R q 2 + 1/ cd2
R / cd
q = (t / R) 2 − 1/ cd2
where g(t, q) is an arbitrary non-singular function. Then, the double integral in

Eq. (6.3.37) is converted to the Laplace transform integral as
ðt=RÞ2 1=c2d ( )
Z1 Z ðþÞ ðÞ
ðdÞ ðdÞ ðþÞ dp ðdÞ ðÞ dp
Ixx ðx; y; z; sÞ ¼ expðstÞdt Fxx ðpd ; qÞ d Fxx ðpd ; qÞ d dq
dt dt
R=cd 0
ð6:3:39Þ
We have just arrived at the form of Laplace transform integral, and its integrand is
in the form of the finite integral with respect to the variable q. The Laplace
inversion is carried out by inspection and it results in the following integral:
ðt=RÞ2 1=c2d ( )
Z ðþÞ ðÞ
ðdÞ ðdÞ ðþÞ dpd ðdÞ ðÞ dpd
Ixx ðx; y; z; tÞ ¼ Hðt R=cd Þ Fxx ðpd ; qÞ Fxx ðpd ; qÞ dq
dt dt
0
ð6:3:40Þ
We can easily find that the disturbed region by the dilatational wave is given by the
operation of the step function Hðt R=cd Þ and it is the semi-sphere R cd t with
radius cd t.
The triple inversion for the dilatational wave contribution has been carried out
successfully. The other dilatational wave contributions can be inverted by the same
technique developed here. The unified expression for the dilatational wave con-
tribution is given by
ðt=RÞ2 1=c2d ( )
Z ðþÞ ðÞ
ðdÞ ðdÞ ðþÞ dp ðdÞ ðÞ dp
Iij ðx; y; z; tÞ ¼ Hðt R=cd Þ Fij ðpd ; qÞ d Fij ðpd ; qÞ d dq
dt dt
0
ð6:3:41Þ
ðÞ
where pd are defined by Eq. (6.3.31). The integrands are given by
ðdÞ ðx2 p2 þ y2 q2 Þbs

Fxx ðp; qÞ ¼
ðprÞ2 Rðp; qÞ
ðdÞ xyðp2 q2 Þbs
Fxy ðp; qÞ ¼ ð6:3:42aÞ
ðprÞ2 Rðp; qÞ
ðdÞ ixpðb2s þ p2 þ q2 Þ
Fxz ðp; qÞ ¼
2p2 rRðp; qÞ
ðdÞ xyðp2 q2 Þbs

Fyx ðp; qÞ ¼
ðprÞ2 Rðp; qÞ
ðdÞ ðy2 p2 þ x2 q2 Þbs
Fyy ðp; qÞ ¼ ð6:3:42bÞ
ðprÞ2 Rðp; qÞ
ðdÞ iypðb2s þ p2 þ q2 Þ
Fyz ðp; qÞ ¼
2p2 rRðp; qÞ
ðdÞ ixpbd bs
Fzx ðp; qÞ ¼
p2 rRðp; qÞ
ðdÞ iypb b
Fzy ðp; qÞ ¼ 2 d s ð6:3:42cÞ
p rRðp; qÞ
ðb2s þ p2 þ q2 Þbd
FzzðdÞ ðp; qÞ ¼
2p2 Rðp; qÞ
(2) Inversion of SV wave contribution: ~I ij ðn; g; z; sÞ

ðsÞ
We consider the inversion of the typical SV-wave contribution,
~ ðsÞ
I xx n2 ða2s þ n2 þ g2 4ad as Þ
ðn; g; z; sÞ ¼ expðas zÞ ð6:3:43Þ
sas Rðn; g; sÞ
The formal double Fourier inversion is given by
Zþ1 Zþ1
ðsÞ 1 n2 ða2s þ n2 þ g2 4ad as Þ
Ixx ðx; y; z; sÞ ¼ expðas z inx igyÞdndg
ð2pÞ2 sas Rðn; g; sÞ
1 1
ð6:3:44Þ
The variable transform defined by Eq. (6.3.23) is introduced. Examining the odd
and non-symmetric nature with respect to the variable q, we obtain the following
simpler form for the double integral,
Z1 Zþ1
ðsÞ ðsÞ
Ixx ðx; y; z; sÞ ¼ dq Fxx ðp; qÞ expfsðbs z þ irpÞgdp ð6:3:45Þ
0 1
where the integrand is given by
ðsÞ ðx2 p2 þ y2 q2 Þðb2s þ p2 þ q2 4bd bs Þ

Fxx ðp; qÞ ¼ ð6:3:46Þ
2ðprÞ2 bs Rðp; qÞ
Now, we consider the complex integral U whose integrand is the same as that of
the inner integral in Eq. (6.3.45),
Z
ðsÞ
U ¼ Fxx ðp; qÞ expfsðbs z þ irpÞgdp ð6:3:47Þ
C
In order to determine the closed loop C, the Cagniard’s path is examined. Introducing
the time variable t as
t ¼ bs z þ irp ð6:3:48Þ
the Cagniard’s path is given by

irt z
t2 ðq2 þ 1=c2s ÞR2
¼pðÞ
s ð6:3:49Þ
R2
pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
and its saddle point at the time t ¼ R q2 þ 1=c2s is
ðsÞ r
psaddle ¼ i q2 þ 1=c2s ð6:3:50Þ
R
The integrand has the four branch points given by Eq. (6.3.34) and branch cuts
are introduced along the imaginary axis as shown in Fig. 6.11a, b. Comparing the
magnitude of the branch points with that of the saddle point, we see that there are
two cases: the saddle point is on the branch cut or it is not; in other words, the
Cagniard’s path crosses the branch cut or it does not.
When the saddle point is smaller in magnitude than the branch point of the
dilatational wave,
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
r
q2 þ 1=c2s \ q2 þ 1=c2d ð6:3:51Þ
R
(a) r
q 2 + 1/ cs2 < q 2 + 1/ cd2
R
Im( p )
Re( p )
ps( − ) −i q 2 + 1/ cd2
−i q 2 + 1/ cs2 ps( + )
−i q 2 + 1 / cR2
(b) γ |x|
>1
x2 + y 2
Im( p ) Re( p )
ε →0
−i q 2 + 1/ cd2
ps( − ) ps( + )
−i q 2 + 1/ cs2
−i q 2 + 1 / cR2
Fig. 6.11 Two closed loops for the 3D Cagniard’s technique. a Loop with Cagniard’s path I.
b Loop with deformed Cagniard’s path II
the Cagniard’s path does not cross any branch cut. Then, the closed loop C is
composed of the infinite line along the real axis, the Cagniard’s path, and two large
arcs which connect the line to the Cagniard’s path. This closed loop is similar to
that in the case of the dilatational wave and is shown in Fig. 6.11a. Employing this
closed loop, we apply Cauchy’s theorem to the complex integral U in Eq. (6.3.47).
Since the loop does not include any singular point, the integral along the real axis is
converted to that along the Cagniard’s path, i.e.
Zþ1
ðsÞ
Fxx ðp; qÞ expfsðbs z þ irpÞgdp
1
Z1 (
ðþÞ ðÞ
) ð6:3:52Þ
ðsÞ ðþÞ dps ðsÞ ðÞ dps
¼ Fxx ðps ; qÞ Fxx ðps ; qÞ expðstÞdt
pffiffiffiffiffiffiffiffiffiffiffiffi dt dt
R q þ1=cs
2 2
When the saddle point is larger in magnitude than the branch point of the
dilatational wave,
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
r
q2 þ 1=c2s [ q2 þ 1=c2d ) q\ ðr=cs Þ2 ðR=cd Þ2 =z ð6:3:53Þ
R
the Cagniard’s path crosses the branch cut. In order to avoid the crossing, the path is
deformed along the cut, such as the Cagniard’s path II in Sect. 6.2. Our deformed
Cagniard’s path is composed of two symmetric semi-hyperbolas, two short lines
along the cut, and a small circle around the the branch point of the dilatational
wave. We name the deformed path the Cagniard path II, too. Figure 6.11b shows
the Cagniard path II and the closed loop C. The closed loop C has two large arcs
which connect the line with the Cagniard path II. Then, we apply Cauchy’s theorem
to the complex integral U with this closed loop.
No singular point is included in the loop. The integral along the small circle
\ \
(BCD) vanishes as its radius tends to zero, and those along the large arcs (IJ and
\
FG) also vanish as the radius tends to infinity. Then, the integral along the real axis
!
(IHG) is converted to the sum of two integrals. One is that along the regular
! !
Cagniard’s path ( AJ and EF ) and the other is the sum of two line integrals along
! !
the branch cut (AB and DE ). Thus the integral along the real axis is converted to
the sum of two integrations as
Zþ1
ðsÞ
1
Z1 ( )
ðþÞ ðÞ
¼ Fxx
ðps ; qÞ Fxx ðps ; qÞ
expðstÞdt
pffiffiffiffiffiffiffiffiffiffiffiffi dt dt ð6:3:54Þ
R q2 þ1=c2s
0 1
Z Z
B C ðsÞ
þB@ þ CF ðp; qÞ expfsðbs z þ irpÞgdp ¼ 0
A xx
! !
AB DE
Following the method in Sect. 1.3.2 in Chap. 1, we examine the argument of the
! !
radicals along the branch cut, AB and DE in Fig. 6.11b. Due to the radiation
!
condition Reðbj Þ [ 0, the argument of the radical bd on AB , which is the left side
!
of the cut, is þp=2 and that on the right side DE is p=2. The argument of the
radicals and the integration variable along the cut are summarized as follows:
! r
On AB : p ¼ i1; q2 þ 1=c2d \ 1 \ q2 þ 1=c2s
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi R ð6:3:55Þ
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
bd ¼ þi 12 ðq2 þ 1=c2d Þ; bs ¼ þ q2 þ 1=c2s 12
! r
On DE : p ¼ i1; q2 þ 1=c2d \ 1 \ q2 þ 1=c2s
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi R ð6:3:56Þ
bd ¼ i 12 ðq2 þ 1=c2d Þ ; bs ¼ þ q2 þ 1=c2s 12
The two line integrals along the branch cut are calculated and are unified as follows:
0 1
Z Z
B C ðsÞ
B þ CF ðp; qÞ expfsðbs z þ irpÞgdp
@ A xx
! !
AB DE
ðr=RÞ q2 þ1=c2s
R
ðsÞ
¼ Fxx ðp; qÞ p¼i1 expfsðcs z þ 1rÞgðid1Þ
ffi
2 q þ1=cd
2
bd ¼þicd
bs ¼cs ð6:3:57Þ
ðr=RÞ q2 þ1=c2s
R
ðsÞ
þ Fxx ðp; qÞ p¼i1 expfsðcs z þ 1rÞgðid1Þ
ffi bd ¼icd
2 q þ1=cd2
bs ¼cs
ðr=RÞ q2 þ1=c2s
R ðsÞ
¼ fxx ð1; qÞ expfsðcs z þ r1Þgd1
ffi
2 q þ1=cd2
where a new notation for the integrand is introduced as

8 9
< =
fxxðsÞ ð1; qÞ ¼ ðiÞ Fxx
ðsÞ
ðp; qÞ p¼i1 Fxx
ðsÞ
ðp; qÞ p¼i1 ð6:3:58Þ
: bd ¼icd bd ¼þicd ;
bs ¼cs bs ¼cs
and
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
cd ¼ 12 ðq2 þ 1=c2d Þ; cs ¼ q2 þ 1=c2s 12 ð6:3:59Þ
In order to reduce the last integral in Eq. (6.3.57) to the form of Laplace
transform, the variable transform from 1 to the time t is introduced as
t ¼ cs z þ 1r ð6:3:60Þ
Its inverse is
rt z ðq2 þ 1=c2s ÞR2 t2
1H ¼ ð6:3:61Þ
R2
The integral along the branch cut is thus converted to that of Laplace transform,
ðr=RÞ
Z q þ1=cs
2 2
fxxðsÞ ð1; qÞ expfsðcs z þ r1Þgd1

ffi
2q þ1=cd 2
pffiffiffiffiffiffiffiffiffiffiffiffi ð6:3:62Þ
Zq þ1=cs
R 2 2
d1H
¼ fxxðsÞ ð1H ; qÞ expðstÞdt
ffi pffiffiffiffiffiffiffiffiffiffiffiffi
ffi dt
z 1=cs 1=cd þr
2 2 2 q þ1=cd
2
Substituting the above equation into Eq. (6.3.54), we have the inner integral in the
form of Laplace transform integral,
Zþ1
ðsÞ
1
Z1 ( )
ðþÞ ðÞ
pffiffiffiffiffiffiffiffiffiffiffiffi dt dt ð6:3:63Þ
R 2 q þ1=cs
2
Zq þ1=cs
R 2 2
d1H
þ fxxðsÞ ð1H ; qÞ expðstÞdt
ffi dt
z 1=cs 1=cd þr
2 2 2 q þ1=cd
2
Comparing this equation with Eq. (6.3.52), we readily see that the last integral in
the above equation appears only when the conditional of Eq. (6.3.53) holds. Using
the step function, we can express the inner integral in the unified form as
Zþ1
ðsÞ
1
Z1 ( )
ðþÞ ðÞ
R 2q þ1=cs 2
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi Zq þ1=cs
R 2 2
d1
þH ðr=cs Þ2 ðR=cd Þ2 qz fxxðsÞ ð1H ; qÞ H expðstÞdt
ffi dt
z 1=cs 1=cd þr
2 2 2 q þ1=cd2
ð6:3:64Þ
We substitute the above equation into Eq. (6.3.45) to get
Z1 Z1 ( )
ðþÞ ðÞ
ðsÞ ðsÞ ðþÞ dps ðsÞ ðÞ dps
Ixx ðx; y; z; sÞ ¼ dq Fxx ðps ; qÞ Fxx ðps ; qÞ expðstÞdt
0 R 2 q þ1=cs2
Z1 qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
þ H ðr=cs Þ2 ðR=cd Þ2 qz dq
0
Zq þ1=cs
R 2 2
d1
fxxðsÞ ð1H ; qÞ H expðstÞdt
ffi dt
z 2 1=cs 1=cd þr
2 2 q þ1=cd
2
ð6:3:65Þ
To perform the Laplace inversion, the order of integration must be interchanged.

Discussing the supporting region for the double integral such as in Fig. 6.12a, we
exchange the order of the integrations in the first double integral as
2
Z1 Z1 Z1 ðt=RÞ
Z 1=cs
2
dq gðt; qÞdt ¼ dt gðt; qÞdq ð6:3:66Þ

0
pffiffiffiffiffiffiffiffiffiffiffiffi R=cs 0
R 2q þ1=cs 2
The supporting region for the second double integral is also shown in Fig. 6.12b.
After some examinations, we obtain the exchange formula
Z1 qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi Zq þ1=cs
R 2 2
2 2
H ðr=cs Þ ðR=cd Þ qz dq hðt; qÞdt
0
ffi
z 1=cs 1=cd þr
2 2 2 q þ1=cd
2
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi 2
1
ðtz 1=c2s 1=c2d Þ ðr=cd Þ2
ZR=cs r
Z
¼ dt hðt; qÞdq ð6:3:67Þ
ffi 0
r=cd þz 1=cs 1=cd
2 2
ffi pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi 2
R2
1=c2s 1=c2d
1
z Z r
Z
þ dt hðt; qÞdq
R=cs
2
ðt=RÞ 1=c2s
where g(t, q) and h(t, q) in Eqs. (6.3.66) and (6.3.67) are arbitrary non-singular
integrands.
(a) t
t t = R q 2 + 1/ cs2
R / cs
q = (t / R) 2 − 1 / cs2
(b) t
R2
1/ cs2 − 1/ cd2
z
t = R q 2 + 1/ cs2
t
R / cs
t
t = r / cd + z 1/ cs2 − 1/ cd2
r / cd + z 1/ cs2 − 1/ cd2
q
q1 q2
q=0
1
q= (r / cs ) 2 − ( R / cd ) 2
z
q1 = (t / R) 2 − 1/ cs2 q2 =
1
r (t − z 1 / c 2
s )
− 1 / cd2 − (r / cd ) 2
Fig. 6.12 Supporting region for the double integral in a Eq. (6.3.66) and b Eq. (6.3.67)
Applying the exchange formulas to the double integral in Eq. (6.3.65), we have
the Laplace transform integral for the Laplace transformed component,
Z 1=cs ( )
2
Z1 ðt=RÞ 2
ðþÞ ðÞ
ðsÞ ðsÞ ðþÞ dps ðsÞ ðÞ dps
Ixx ðx; y; z; sÞ ¼ expðstÞdt Fxx ðps ; qÞ Fxx ðps ; qÞ dq
dt dt
R=cs 0
1
ZR=cs r
Z
d1H
þ expðstÞdt fxxðsÞ ð1H ; qÞ dq
ffi dt
r=cd þz 1=cs 1=cd
2 2 0
ffi pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi 2
R2
1=c2s 1=c2d
1
z Z r
Z
d1H
þ expðstÞdt fxxðsÞ ð1H ; qÞ dq
pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi dt
R=cs 2
ðt=RÞ 1=c2s
ð6:3:68Þ
Finally, the Laplace inversion is carried out by inspection, and the SV-wave con-
tribution is given by
ðsÞ
Ixx ðx; y; z; tÞ
Z 1=cs ( )
2
ðt=RÞ 2
ðþÞ ðÞ
¼ Hðt R=cs Þ Fxx ðps ; qÞ Fxx ðps ; qÞ dq
dt dt
0
1
r
Z
d1
þ H t r=cd z 1=c2s 1=c2d HðR=cs tÞ fxxðsÞ ð1H ; qÞ H dq
dt
0
1
2 qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi r Z
R d1
þ H ðt R=cs ÞH 1=c2s 1=c2d t fxxðsÞ ð1H ; qÞ H dq
z pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi dt
ðt=RÞ2 1=c2s
ð6:3:69Þ
Examining the operation of the step functions ahead of the integral, the first term
with Hðt R=cs Þ shows a semi-spherical region disturbed by SV-wave. The second
term
which has the product of the two step functions,
H t r=cd z 1=cs 1=cd HðR=cs tÞ, shows a region of von-Schmidt wave.
2 2
2 pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
The last term with H ðt R=cs ÞH Rz 1=c2s 1=c2d t does not show the wave
nature, but appears only in the case of the 3D deformation (this non-wave front is
discussed by Gakenheimer and Miklowitz (1969)). These wave regions are denoted
by regions A and B in Fig. 6.13.
cd t = r + z γ 2 − 1
z = r γ 2 −1
cs t = R
cd t = R
R 2 γ 2 − 1 = cd t ⋅ z
Fig. 6.13 Wave fronts and disturbed regions A and B
To the other SV-wave contributions, the same inversion technique is applied and
we obtain the unified expression for SV-wave contributions as
ðsÞ
Iij ðx; y; z; tÞ
Z 1=cs ( )
2
ðt=RÞ 2
ðþÞ ðÞ
ðsÞ dps ðsÞ dps
¼ Hðt R=cs Þ Fij ðpðþÞ
s ; qÞ Fij ðpðÞ s ; qÞ dq
dt dt
0
1
r
Z
ðsÞ d1
þ H t r=cd z 1=c2s 1=c2d HðR=cs tÞ fij ð1H ; qÞ H dq
dt
0
1
2 qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi r Z
R ðsÞ d1
þ H ðt R=cs ÞH 1=c2s 1=c2d t fij ð1H ; qÞ H dq
z pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi dt
ðt=RÞ2 1=c2s
ð6:3:70Þ
ðÞ
where ps and 1H are given by Eqs. (6.3.49) and (6.3.61) respectively and
ðsÞ ðx2 p2 þ y2 q2 Þðb2s þ p2 þ q2 4bd bs Þ

Fxx ðp; qÞ ¼
xy ðp2 q2 Þðb2s þ p2 þ q2 4bd bs Þ
ðsÞ
Fxy ðp; qÞ ¼ ð6:3:71aÞ
ðsÞ x ipbd bs
Fxz ðp; qÞ ¼
p2 r Rðp; qÞ
ðsÞ xy ðp2 q2 Þðb2s þ p2 þ q2 4bd bs Þ

Fyx ðp; qÞ ¼
ðy2 p2 þ x2 q2 Þðb2s þ p2 þ q2 4bd bs Þ
ðsÞ
Fyy ðp; qÞ ¼ ð6:3:71bÞ
ðsÞ y ipbd bs
Fyz ðp; qÞ ¼
p2 r Rðp; qÞ
ðsÞ x ipðb2s þ p2 þ q2 Þ
Fzx ðp; qÞ ¼
2p2 r Rðp; qÞ
ðsÞ y ipðb2s þ p2 þ q2 Þ
Fzy ðp; qÞ ¼ ð6:3:71cÞ
2p2 r Rðp; qÞ
b ðp 2
þ q 2
Þ
FzzðsÞ ðp; qÞ ¼ d 2
p Rðp; qÞ
ðx1Þ2 ðyqÞ2
fxxðsÞ ð1; qÞ ¼ ðiÞ
2ðprÞ2 cs
( )
c2s 12 þ q2 4icd cs c2s 12 þ q2 þ 4icd cs

ðc2s 12 þ q2 Þ2 þ 4icd cs ð12 q2 Þ ðc2s 12 þ q2 Þ2 4icd cs ð12 q2 Þ
xy 12 þ q2
fxyðsÞ ð1; qÞ ¼ ðiÞ 2
2ðprÞ cs
( )

x
fxzðsÞ ð1; qÞ ¼ 2 1cd cs
p
(r )
1 1
þ
ð6:3:72aÞ
xy 1 þq2 2
fyxðsÞ ð1; qÞ ¼ ðiÞ
2ðprÞ2 cs
( )

xy ðy1Þ2 ðxqÞ2
fyyðsÞ ð1; qÞ ¼ ðiÞ 2
2ðprÞ cs
( )

y
fyzðsÞ ð1; qÞ ¼ 2 1cd cs
p
(r )
1 1
þ
ð6:3:72bÞ
x
fzxðsÞ ð1; qÞ ¼ ðþiÞ 2 1ðc2s 12 þ q2 Þ
( 2p r )
1 1

y
fzyðsÞ ð1; qÞ ¼ ðþiÞ 2 1ðc2s 12 þ q2 Þ
( 2p r )
1 1

cd ð12 q2 Þ
fzzðsÞ ð1; qÞ ¼ 2
(p )
1 1
þ
ð6:3:72cÞ
(3) Inversion of SH wave contribution: ~I ii

ðSHÞ
ðn; g; z; sÞ
Two SH-wave contributions are identical
~ 1
ðn; g; z; sÞ ¼ ~I yy ðn; g; z; sÞ ¼
ðSHÞ
I xx ðSHÞ
expðas zÞ ð6:3:73Þ
sas
The formal double Fourier inversion is given by
Zþ1 Zþ1
ðSHÞ 1 1
Iii ðx; y; z; sÞ ¼ expðas z inx igyÞdndg ð6:3:74Þ
ð2pÞ2 sas
1 1
Now, we introduce the variable transform defined by Eq. (6.3.23),
Z1 Zþ1 !
1 1 n qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi o
ðSHÞ
Iii ðx; y; z; sÞ ¼ dq pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi exp s z p2 þ q2 þ 1=c2s þ irp dp
2p2 p2 þ q2 þ 1=c2s
0 1
ð6:3:75Þ
The Cagniard-de Hoop technique is applied to the inner integral. Introducing the
variable transform from p to the new variable t,
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
t ¼ z p2 þ q2 þ 1=c2s þ irp ð6:3:76Þ
the Cagniard’s path is given by

irt z t2 ðq2 þ 1=c2s ÞR2
pðÞ
s ¼ ð6:3:77Þ
R2
We have exact expressions for the radical and the gradient,

qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
tz ir t2 ðq2 þ 1=c2s ÞR2
p2 þ q2 þ 1=c2s ðÞ
¼ ð6:3:78Þ
p¼ps R2
ðÞ pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
dps tz ir t2 ðq2 þ 1=c2s ÞR2
¼ pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi ð6:3:79Þ
dt R2 t2 ðq2 þ 1=c2s ÞR2
Following the former discussion for the Cagniard-de Hoop technique, we apply
Cauchy’s integral theorem to the complex integral,
Z !
1 1 n qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi o
U¼ 2 pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi exp s z p2 þ q2 þ 1=c2s þ irp dp
2p p2 þ q2 þ 1=c2s
c
ð6:3:80Þ
and find that the integral along the real axis is converted to that along the Cagn-
iard’s path. Since

1 dp 1 dp 2
pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi p¼pðþÞ pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
ffi ¼ pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
p þ q þ 1=cs dt
2 2 2 Ps p þ q þ 1=cs dt p¼pðÞ
2 2 2 t ðq2 þ 1ÞR2
2
s
ð6:3:81Þ
the double inversion integral in Eq. (6.3.75) is transformed to

0 1
Z1 Z1
ðSHÞ 1 B 1 C
Iii ðx; y; z; sÞ ¼ dq @ qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
ffiA expðstÞdt
p2 pffiffiffiffiffiffiffiffiffiffiffiffi t2 q2 þ 1=c2s R2
0 R 2q þ1=cs 2
ð6:3:82Þ
The order of integration is also exchanged as

2
0 1
Z1 ðt=R
ZÞ 1=cs
2
ðSHÞ 1 B 1 C
Iii ðx; y; z; sÞ ¼ expðstÞdt @ qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
ffiAdq
p2 t q þ 1=cs R
2 2 2 2
R=cs 0
ð6:3:83Þ
It is very lucky that the inner integral with respect to the variable q can be evaluated
exactly as
2
0 1
ðt=R
Z Þ 1=cs
2
B 1 C p
@ qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
ffiAdq ¼ 2R ð6:3:84Þ
0
t2 q2 þ 1=c2s R2
Thus, the SH-wave contribution is reduced to the simple Laplace transform integral,
Z1
ðSHÞ 1
Iii ðx; y; z; sÞ ¼ expðstÞdt ð6:3:85Þ
2pR
R=Cs
The Laplace inversion is carried out by inspection and we have the final result,
ðSHÞ 1
Iii ðx; y; z; tÞ ¼ H ðt R=Cs Þ; i ¼ x; y ð6:3:86Þ
2pR
Consequently, we have obtained the SH-wave contribution in the closed form.

The three wave contributions have just been inverted by the Cagniard-de Hoop
technique. The displacement response is expressed in terms of the Green’s dyadic
Gij(x, y, z, t) as
ux ¼ Px Gxx ðx; y; z; tÞ þ Py Gxy ðx; y; z; tÞ þ Pz Gxz ðx; y; z; tÞ

uy ¼ Px Gyx ðx; y; z; tÞ þ Py Gyy ðx; y; z; tÞ þ Pz Gyz ðx; y; z; tÞ ð6:3:87Þ
uz ¼ Px Gzx ðx; y; z; tÞ þ Py Gzy ðx; y; z; tÞ þ Pz Gzz ðx; y; z; tÞ
where the dyadic components are given by
1 n ðdÞ ðsÞ
o
Gxx ðx; y; z; tÞ ¼ Ixx ðx; y; z; tÞ þ Ixx ðx; y; z; tÞ þ Ixx
SH
ðx; y; z; tÞ
l
1 n ðdÞ ðsÞ
o
Gxy ðx; y; z; tÞ ¼ Ixy ðx; y; z; tÞ þ Ixy ðx; y; z; tÞ ð6:3:88aÞ
l
1 n ðdÞ ðsÞ
o
Gxz ðx; y; z; tÞ ¼ Ixz ðx; y; z; tÞ þ Ixz ðx; y; z; tÞ
l
1 n ðdÞ ðsÞ
o
Gyx ðx; y; z; tÞ ¼ Iyx ðx; y; z; tÞ þ Iyx ðx; y; z; tÞ
l
1 n ðdÞ ðsÞ ðSHÞ
o
Gyy ðx; y; z; tÞ ¼ Iyx ðx; y; z; tÞ þ Iyx ðx; y; z; tÞ þ Iyy ðx; y; zÞ ð6:3:88bÞ
l
1n d o
Gyz ðx; y; z; tÞ ¼ Iyz ðx; y; z; tÞ þ Iyz
S
ðx; y; z; tÞ
l
1 n ðdÞ ðsÞ
o
Gzx ðx; y; z; tÞ ¼ Izx ðx; y; z; tÞ þ Izx ðx; y; z; tÞ
l
1 n ðdÞ ðsÞ
o
Gzy ðx; y; z; tÞ ¼ Izy ðx; y; z; tÞ þ Izy ðx; y; z; tÞ ð6:3:88cÞ
l
1 n ðdÞ ðsÞ
o
Gzy ðx; y; z; tÞ ¼ Izz ðx; y; z; tÞ þ Izy ðx; y; z; tÞ
l
Exercises
(6:1) In Sect. 6.2, if the applied load is a semi-infinite
distribution such as
p ; x [ 0
ryz y¼0 ¼ dðtÞ 0
ðaÞ
0; x\0
the transformed boundary condition of Eq. (6.1.14) is replaced with

yz ¼ p0 dþ ðnÞ
r ðbÞ
y¼0
where dþ ðÞ is Heisenberg’s delta function. Verify the above equation (b).
(6:2) In Sect. 6.3, when the load is distributed uniformly in a quarter region
ðx [ 0; y [ 0Þ on the surface, how do you change the mathematical
expression for the boundary condition (6.3.3)?
References
Achenbach JD (1973) Wave propagation in elastic solids. North-Holland, New York

Cagniard L (1962) Reflection and refraction of progressive seismic waves (translated by Flinn ED,
Dix CH). McGraw-Hill, New York
De-Hoop AT (1961) The surface line source problem. Appl Sci Res B 8:349–356
Fung YC (1970) Foundation of solid mechanics (Japanese translation by Ohashi Y et al (1965)
Bai-Fu-Kan, Tokyo). Prentice-Hall, New Jersey
Gakenheimer DC, Miklowitz J (1969) Transient excitation of an elastic half space by a point load
traveling on the surface. J Appl Mech (Trans ASME Ser E) 36:505–515
Graff KF (1975) Wave motion in elastic solids. Clarendon Press, Oxford
Miklowitz J (1978) The theory of elastic waves and waveguides. North-Holland, New York
Chapter 7
Miscellaneous Green’s Functions
This last chapter presents five Green’s functions and one application technique of
the complex integral. The first and second sections consider the 2D static Green’s
dyadic for an orthotropic elastic solid, and for an inhomogeneous elastic solid. The
third section discusses the Green’s function for torsional waves in an anisotropic
solid. The fourth and fifth sections are concerned with Green’s function for SH
waves. One is wave reflection at a moving boundary and the other is wave scat-
tering by a rigid inclusion in an inhomogeneous solid. All the Green’s functions are
obtained in the closed form by applying the method of integral transform.
Especially, in the fourth section which discusses wave reflection, a conversion
formula between two different Laplace inversion integrals is developed so that we
can treat the moving boundary problem. In the fifth section which discusses the
wave scattering, a summation formula is derived for the Fourier series of the
Schlömlich type. It enables us to obtain the closed expression for the wave scat-
tering problem. The last section shows an excellent application technique of the
complex integral. It reduces a semi-infinite integral, which includes the product of
two Bessel functions, to a finite integral that is very suitable for numerical
computations.
7.1 2D Static Green’s Dyadic for an Orthotropic

Elastic Solid
An anisotropic elastic solid whose orthogonal Young’s moduli differ from each
other is called an “orthotropic” solid. In 2D in-plane deformation, Hooke’s law for
the orthotropic elastic solid is given by
2 3 2 32 3
rxx C11 C12 0 exx
4 ryy 5 ¼ 4 C12 C22 0 54 eyy 5 ð7:1:1Þ
rxy 0 0 C66 exy

DOI 10.1007/978-3-319-17455-6_7
206 7 Miscellaneous Green’s Functions
where Cij are elastic moduli. The strain components in the state of plane strain are
defined by

@ux @uy 1 @ux @uy
exx ¼ ; eyy ¼ ; exy ¼ þ ð7:1:2Þ
@x @y 2 @y @x
The equilibrium equations are
@rxx @rxy
þ ¼ Qx dðxÞdðyÞ
@x @y
ð7:1:3Þ
@rxy @ryy
þ ¼ Qy dðxÞdðyÞ
@x @y
where Qi is the magnitude of a body force.

Substituting Hooke’s law into the equilibrium equations, we obtain the dis-
placement equations as

@ 2 ux C66 @ 2 ux C66 @ 2 uy
C11 2 þ þ C12 þ ¼ Qx dðxÞdðyÞ
@x 2 @y2 2 @x@y
2 ð7:1:4Þ
C66 @ ux C66 @ 2 uy @ 2 uy
C12 þ þ þ C 22 ¼ Qy dðxÞdðyÞ
2 @x@y 2 @x2 @y2
Since we are concerned with Green’s dyadic, a particular solution corresponding to

the body force is explored. As the elastic medium is of infinite extent, the con-
vergence condition at infinity is assumed as

ffi @ui @ui
ui j x2 þy2 !1 ¼ ffi ¼ @y pffiffiffiffiffiffiffiffiffi ¼ 0; i ¼ x; y ð7:1:5Þ
x2 þy2 !1 x2 þy2 !1
We employ the double Fourier transform with respect to two space variables as
defined by
Z1 Z1
2p
1 1
Z1 Z1
2p
1 1
7.1 2D Static Green’s Dyadic for an Orthotropic Elastic Solid 207
The displacement equations (7.1.4) are transformed to the algebraic equations for
the transformed displacement components,
fC11 n2 þ ðC66 =2Þg2 g~ux þ ðC12 þ C66 =2Þng~uy ¼ Qx

ð7:1:8Þ
ðC12 þ C66 =2Þngu~x þ fðC66 =2Þn2 þ C22 g2 g~uy ¼ Qy
Explicit expressions for the displacement components are obtained in the trans-
formed domain,
Qx n2 þ bg2 Qy ðc þ 1Þng
~

ux ¼ þ
C22 ðg2 þ p21 n2 Þðg2 þ p22 n2 Þ C22 ðg2 þ p21 n2 Þðg2 þ p22 n2 Þ
ð7:1:9Þ
Qx ðc þ 1Þng Qy an2 þ g2
~

uy ¼ þ
C22 ðg2 þ p21 n2 Þðg2 þ p22 n2 Þ C22 ðg2 þ p21 n2 Þðg2 þ p22 n2 Þ
where the eigenvalues, pj ðj ¼ 1; 2Þ, are
sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
pffiffiffiffiffi sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
pffiffiffiffiffi
p1 ab cðc þ 2Þ þ 2 bc ab cðc þ 2Þ 2 bc
¼ ð7:1:10Þ
p2 4b 4b
and the elastic modulus parameters, a; b; c, are introduced as
C11 C22 C12

a¼ ; b¼ ; c¼ ð7:1:11Þ
C66 =2 C66 =2 C66 =2
Inspecting the transformed displacement in Eq. (7.1.9), we find that two

inversion formulas are necessary for the full inversion. They are
~
I 1 ðn; g; a; bÞ ¼ an2 þ bg2 ~I ðn; gÞ ¼ ng
; 2
ðg2 þ p21 n2 Þðg2 þ p22 n2 Þ ðg2 þ p21 n2 Þðg2 þ p22 n2 Þ
ð7:1:12Þ
If we get the inversions for these two expressions, the displacement in the actual
space can be expressed as
Qx Qy
ux ¼ þ I1 ðx; y; 1; bÞ ðc þ 1ÞI2 ðx; yÞ
C22 C22
ð7:1:13Þ
Qx Qy
uy ¼ ðc þ 1ÞI2 ðx; yÞ þ I1 ðx; y; a; 1Þ
C22 C22
(1) Fourier inversion with respect to the parameter g

The formal Fourier inversion with respect to the parameter g is given by the
integrals
Zþ1
I1 ðn; y; a; bÞ ¼ 1 an2 þ bg2
expðigyÞdg
2p ðg2 þ p21 n2 Þðg2 þ p22 n2 Þ
1
ð7:1:14Þ
Zþ1
I2 ðn; yÞ ¼ 1 ng
expðigyÞdg
2p ðg þ p1 n Þðg2 þ p22 n2 Þ
2 2 2
1
These two integrals can be evaluated by the application of some formulas, but, in
order to show the application of Jordan’s lemma, we consider the complex integrals
in the g-plane.
Let us consider the complex integral whose integrand is the same as that of the
corresponding integral in Eq. (7.1.14),
I
1 an2 þ bg2
U1 ¼ expðigyÞdg
2p ðg2 þ p21 n2 Þðg2 þ p22 n2 Þ
L
I ð7:1:15Þ
1 ng
U2 ¼ expðigyÞdg
2p ðg þ p1 n Þðg2 þ p22 n2 Þ
2 2 2
L
The integrand has the four simple poles at g ¼ ipj jnj; ðj ¼ 1; 2Þ as shown in
Fig. 7.1. The closed loop L is composed of a straight line along the real axis and a
Fig. 7.1 Closed loop L for

Fourier inversion integral
semi-circle with infinite radius. Due to the convergence on the semi-circle, we have
to choose the lower loop when y > 0 and the upper loop when y < 0. Since no other
singular point exists in the loop, the integral along the real axis is converted to the
sum of the residues at the two poles. Thus, we have for Ij ; j ¼ 1; 2
I1 ðn; y; a; bÞ ¼ a bp21 1 a bp22 1

exp ðp 1 jnjjyjÞ þ expðp2 jnjjyjÞ
2p1 ðp1 p2 Þ jnj
2 2 2p2 ðp1 p2 Þ jnj
2 2

I2 ðn; yÞ ¼ þ sgnðyÞ i i
expðp1 jnjjyjÞ expðp2 jnjjyjÞ
2ðp1 p2 Þ n
2 2 n
ð7:1:16Þ
where the sign function is defined by

þ1; y[0
sgnðyÞ ¼ ð7:1:17Þ
1; y\0
(2) Fourier inversion with respect to the parameter n

Applying the formal Fourier inversion integral with respect to the parameter n to
Eq. (7.1.16), we have
Zþ1
a bp21 1 1
I1 ðx; y; a; bÞ ¼ expðp1 jnjjyj inxÞdn
2p1 ðp21 p22 Þ 2p jnj
1
Zþ1
a bp22 1 1
þ expðp2 jnjjyj inxÞdn
2p2 ðp21 p22 Þ 2p jnj
1
8 ð7:1:18Þ
Zþ1
sgnðyÞ < 1 i
I2 ðx; yÞ ¼ þ expðp1 jnjjyj inxÞdn
2ðp1 p2 Þ
2 2 : 2p n
1
9
Zþ1 =
1 i
expðp2 jnjjyj inxÞdn
2p n ;
1
Inspecting the above equations, the two necessary inversion integrals are extracted as
Zþ1
1 1
I1j ¼ exp pj jnjjyj expðinxÞdn;
2p jnj
1
ð7:1:19Þ
Zþ1
1 i
I2j ¼ exp pj jnjjyj expðinxÞdn;
2p n
1
The second integral I2j is reduced to a semi-infinite integral and can be evaluated by
the application of the formula (Erdélyi 1954, vol. I, pp. 72, 2),
Z1
1 1 b
expðaxÞ sinðbxÞdx ¼ tan ð7:1:20Þ
x a
0
That is
Zþ1
1 i
I2j ¼ exp pj jnjjyj expðinxÞdn
2p n
1
Z1
11 ð7:1:21Þ
¼ exp pj njyj sinðnxÞdn
pn
0

1 x
¼ tan1
p pj jyj
The first integral I1j has the first order singularity at n ¼ 0, and thus this integral
cannot be evaluated in this form. The singular behavior is the same as that in the 2D
static plane problem in Chap. 3. We differentiate the integral with respect to the
variable x and jyj, respectively, to obtain
Zþ1
@I1j 1 in
¼ exp pj jnjjyj expðinxÞdn
@x 2p jnj
1
ð7:1:22Þ
Zþ1
@I1j 1 pj jnj
¼ exp pj jnjjyj expðinxÞdn
@jyj 2p jnj
1
Reducing the above integrals to semi-infinite ones, we get
Z1
@I1j 1
¼ exp pj njyj sinðnxÞdn
@x p
0
ð7:1:23Þ
Z1
@I1j pj
¼ exp pj njyj cosðnxÞdn
@jyj p
0
and then we apply the formulas (Erdélyi 1954, vol. I, pp. 14, 72), to get
Z1
pj jyj
exp pj njyj cosðnxÞdn ¼
x2 þ p2j y2
0
ð7:1:24Þ
Z1
x
exp pj njyj sinðnxÞdn ¼
x2 þ p2j y2
0
The two integrals with different derivative are expressed in terms of algebraic
functions,
@I1j 1 x @I1j p2j jyj

¼ 2 ; ¼ ð7:1:25Þ
@x p x þ p2j y2 @jyj p x2 þ p2j y2
In order to obtain the formula for I1j , we integrate the derivatives with respect to
each space variable,
1
1

I1j ¼ log x2 þ p2j y2 þ Cx ðyÞ; I1j ¼ log x2 þ p2j y2 þ Cy ðxÞ
2p 2p
ð7:1:26Þ
Since the two expressions for I1j must be same, the integration “constant” functions
Cx and Cy also must be a simple constant, without any space variable. Thus, we
have for I1j
1

I1j ¼ log x2 þ p2j y2 þ const: ð7:1:27Þ
2p
Summarizing the above discussion for the inversion integrals, we obtain the
simple formulas,
Zþ1

1 1 1
I1j ¼ exp pj jnjjyj expðinxÞdn ¼ log x2 þ p2j y2
2p jnj 2p
1
ð7:1:28Þ
Zþ1
1 i 1 x
I2j ¼ exp pj jnjjyj expðinxÞdn ¼ tan1
2p n p pj jyj
1
where the arbitrary constant for the integral I1j is neglected since it does not produce
any stress and strain. Substituting these formulas into Eq. (7.1.18), we obtain the
closed expressions for the double Fourier inversions I1 and I2 as
a bp21 2 a bp22
I1 ðx; y; a; bÞ ¼ þ log x þ p 2 2
y log x2 þ p22 y2
4pp1 ðp1 p2 Þ
2 2 1
4pp2 ðp1 p2 Þ
2 2

sgnðyÞ 1 x 1 x
I2 ðx; yÞ ¼ þ tan tan
2pðp21 p22 Þ p1 jyj p2 jyj
ð7:1:29Þ
where the sign function sgn(.) is defined by Eq. (7.1.17). Thus, the displacement
components in Eq. (7.1.13) are expressed in the closed form as

Qx 1 bp21 2 1 bp22 2
ux ðx; yÞ ¼ þ log x þ p1 y
2 2
log x þ p2 y
2 2
4pC22 ðp21 p22 Þ p1 p2

Qy x x
ðc þ 1Þ tan1 tan1
2pC22 ðp21 p22 Þ p1 y p2 y

Qx 1 x 1 x
uy ðx; yÞ ¼ ðc þ 1Þ tan tan
2pC22 ðp21 p22 Þ p1 y p2 y

Qy a p21 2 a p22 2
þ log x þ p1 y
2 2
log x þ p2 y
2 2
4pC22 ðp21 p22 Þ p1 p2
ð7:1:30Þ
Furthermore, if we introduce the notation of the Green’s dyadic Gij , the displace-
ment can be rewritten as
ux ðx; yÞ ¼ Qx Gxx ðx; yÞ þ Qy Gxy ðx; yÞ

ð7:1:31Þ
uy ðx; yÞ ¼ Qx Gyx ðx; yÞ þ Qy Gyy ðx; yÞ
where the dyadic components are given by

1 1 bp21 2 1 bp22 2
Gxx ðx; yÞ ¼ þ log x þ p 2 2
y log x þ p 2 2
y
4pC22 ðp21 p22 Þ p1 1
p2 2

1 x x
Gxy ðx; yÞ ¼ Gyx ðx; yÞ ¼ ðc þ 1Þ tan1 tan1
2pC22 ðp21 p22 Þ p1 y p2 y

1 a p21 2 a p22 2
Gyy ðx; yÞ ¼ þ log x þ p1 y
2 2
log x þ p2 y
2 2
4pC22 ðp21 p22 Þ p1 p2
ð7:1:32Þ
It should be noticed that the symmetry Gxy ¼ Gyx is held in spite of the ortho-
tropic nature. The logarithmic singularity at the source point and at the infinity is
the same as that for isotropic media in Chap. 3. These singular behaviors are
inherent in 2D plane elasticity.
7.2 2D Static Green’s Dyadic for an Inhomogeneous Elastic Solid 213
7.2 2D Static Green’s Dyadic for an Inhomogeneous

Elastic Solid
Materials whose material parameters, such as elastic moduli and density, are
varying with the space point are called “inhomogeneous” materials. Many mathe-
matical models for the inhomogeneity are proposed. The simplest model for the
inhomogeneity is a uniaxial exponential function for the material parameters. This
model is mathematically simple and tractable, but not fully correct since the
material parameters vanish or diverge at infinity. In spite of the unrealistic nature of
the exponential model, this type of the inhomogeneity is frequently used for stress
analysis due to its simplicity.
Let us consider the 2D plane-strain deformation of an inhomogeneous isotropic
elastic solid. Hooke’s law for an inhomogeneous solid is the same as that for a
homogeneous one, but its elastic moduli ðk; lÞ are functions of the space variables,
@ux @uy
rxx ¼ ðk þ 2lÞ þk
@x @y
@ux @uy
ryy ¼ k þ ðk þ 2lÞ ð7:2:1Þ
@x @y

@ux @uy
rxy ¼ ryx ¼ l þ
@y @x
As we are concerned with Green’s dyadic corresponding to a point source, the 2D

equilibrium equations with a point body force are given by
@rxx @ryx
þ ¼ Bx dðxÞdðyÞ
@x @y
ð7:2:2Þ
@rxy @ryy
þ ¼ By dðxÞdðyÞ
@x @y
where the body force is placed at the coordinate origin ð0; 0Þ. The inhomogeneity
which we adopt here is a uniaxial exponential function and thus, Lame’s constants
ðk; lÞ are assumed to be exponential functions with one space variable y,
kðyÞ ¼ k0 expfkðy=hÞg; lðyÞ ¼ l0 expfkðy=hÞg ð7:2:3Þ
where ðk0 ; l0 Þ are moduli at the coordinate origin, k is an inhomogeneity parameter,

and h is a reference length. Since two Lame’s constants have the same exponential
function, Poisson ratio v is constant throughout the medium.
Substituting Hooke’s law of Eq. (7.2.1) with the inhomogeneity of Eq. (7.2.3)
into the equilibrium equations (7.2.2), we obtain displacement equations with
constant coefficients,

@ 2 ux @ 2 ux @ 2 uy k @ux @uy Bx
c 2
þ 2 þ ðc 1Þ2
þ þ ¼ dðxÞdðyÞ
@x 2 @y @x@y h @y @x lðyÞ

@ 2 ux @ 2 uy @ 2 uy k @ux @uy By
ðc2 1Þ þ 2 þ c2 2 þ ðc2 2Þ þ c2 ¼ dðxÞdðyÞ
@x@y @x @y h @x @y lðyÞ
ð7:2:4Þ
where the constant material parameter c is defined by

sffiffiffiffiffiffiffiffiffiffiffiffiffiffi sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
k þ 2l k0 þ 2l0 2ð1 mÞ
c¼ ¼ ¼ ð7:2:5Þ
l l0 1 2m
In order to obtain a particular solution corresponding to the nonhomogeneous body

force term, we apply the double Fourier transform with respect to two space
variables,
Zþ1 Zþ1
f ðxÞ expðþinxÞdx; f ðxÞ ¼
2p
1 1
ð7:2:6Þ
Zþ1 Zþ1
f ðyÞ expðþigyÞdy; f ðyÞ ¼
2p
1 1
The convergence condition at infinity is also imposed on the displacement,

ffi @ui @ui
ui j x2 þy2 !1 ¼ ffi ¼ @y pffiffiffiffiffiffiffiffiffi ¼ 0; i ¼ x; y ð7:2:7Þ
x2 þy2 !1 x2 þy2 !1
With the aid of the integration formulas which include the delta function,
Zþ1 Zþ1
dðyÞ 1
dðxÞ expðþinxÞdx ¼ 1; expðþigyÞdy ¼ ð7:2:8Þ
lðyÞ lð0Þ
1 1
the displacement equations (7.2.4) are transformed to the algebraic equations for the
displacement components,
Bx
fc2 n2 þ g2 þ igðk=hÞg~ux þ fðc2 1Þng þ inðk=hÞg~uy ¼
lð0Þ
ð7:2:9Þ
By
fðc2 1Þng þ inðc2 2Þðk=hÞg~ux þ fn2 þ c2 g2 þ igc2 ðk=hÞg~uy ¼
lð0Þ
where
lð0Þ ¼ l0 ð7:2:10Þ
Solving Eq. (7.2.9) for the transformed displacement components, we get
Bx n2 þ c2 ðg2 þ j2 Þ By ðc2 1Þg ijðc2 3Þ

~

ux ¼ þ n ð7:2:11aÞ
lð0Þ c Dðn; gÞ
2 lð0Þ c2 Dðn; gÞ
Bx ðc2 1Þg þ ijðc2 3Þ By c2 n2 þ g2 þ j2

~

uy ¼ n þ ð7:2:11bÞ
lð0Þ c2 Dðn; gÞ lð0Þ c2 Dðn; gÞ
where the new inhomogeneity parameter j is introduced as
j ¼ k=ð2hÞ ð7:2:12Þ
and the denominator Dðn; gÞ and g are given by
Dðn; gÞ ¼ ðn2 þ g2 þ j2 Þ2 þ 4ð1 2=c2 Þj2 n2 ð7:2:13Þ
g ¼ g þ ij ð7:2:14Þ
In order to apply the inversion integral, we factorize the denominator as
Dðn; gÞ ¼ ðn ix1 Þðn þ ix1 Þðn ix2 Þðn þ ix2 Þ ð7:2:15Þ
where the eigenvalues xj ; j ¼ 1; 2 are

pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
x1 ¼ g2 þ j2 q2 jp; x2 ¼ g2 þ j2 q2 þ jp ð7:2:16Þ
and
pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
p¼ 1 2=c2 ; q¼ 2ð1 1=c2 Þ; q[p ð7:2:17Þ
The formal Fourier inversion with respect to the parameter ξ is given by the
integrals,
Zþ1
Bx 1 n2 þ c2 ðg2 þ j2 Þ
~
ux ¼ þ 2 expðinxÞdn
c lð0Þ 2p ðn ix1 Þðn þ ix1 Þðn ix2 Þðn þ ix2 Þ
1
Zþ1
By 1 ðc2 1Þg ijðc2 3Þ
n expðinxÞdn
2
1
ð7:2:18aÞ
Zþ1
Bx 1 ðc2 1Þg þ ijðc2 3Þ
~
uy ¼ 2 n expðinxÞdn
1
Zþ1
By 1 c2 n2 þ g2 þ j2
þ expðinxÞdn
c2 lð0Þ 2p ðn ix1 Þðn þ ix1 Þðn ix2 Þðn þ ix2 Þ
1
ð7:2:18bÞ
The above integrals can be evaluated by the complex integral theory. So, we apply
Cauchy’s theorem (Jordan’s lemma) to the integral in the complex n-plane. As an
example, the first integral in Eq. (7.2.18a) is discussed. We represent it by the
complex integral Uxx ,
Z
1 n2 þ c2 ðg2 þ j2 Þ
Uxx ¼ expðinxÞdn ð7:2:19Þ
2p ðn ix1 Þðn þ ix1 Þðn ix2 Þðn þ ix2 Þ
C
where the closed loop C is composed of a semi-circle with infinite radius and a
straight line along the real axis (see Fig. 7.2). In order to guarantee the convergence
on the large semi-circle, we have to employ the lower loop for x > 0, and the upper
for x < 0. The integrand has four poles; two of them are in the upper plane and other
two are in the lower plane. Each closed loop includes two poles. Applying the
Jordan’s lemma to the complex integral Uxx in Eq. (7.2.19), the integral along the
real axis is converted to the sum of two residues. When we employ the lower loop
for x [ 0, the complex integral yields
Zþ1
Uxx ¼ expðinxÞdn
2p ðn ix1 Þðn þ ix1 Þðn ix2 Þðn þ ix2 Þ ð7:2:20Þ
1
¼ 2piResðix1 Þ þ 2piResðix2 Þ
The detailed calculation for the residues is
2piResðix1 Þ þ 2piResðix2 Þ

2pi n2 þ c2 ðg2 þ j2 Þ
¼ expðinxÞ
2p ðn ix1 Þðn ix2 Þðn þ ix2 Þ n¼ix1 ð7:2:21Þ

2pi n2 þ c2 ðg2 þ j2 Þ
þ expðinxÞ
2p ðn ix1 Þðn ix2 Þðn þ ix1 Þ n¼ix2
Fig. 7.2 Closed loop C for

the complex integral Uxx Im(ξ )
+iω2
+iω1 x<a
−∞ +∞ Re(ξ )
−∞ +∞
x>a
−iω1
−iω2
Arranging the above equation, the inversion integral along the real axis is evaluated
exactly as
Zþ1
expðinxÞdn
2p ðn ix1 Þðn þ ix1 Þðn ix2 Þðn þ ix2 Þ
1
( ) ð7:2:22Þ
c2 1 c2 þ 1 pffiffiffiffiffiffiffiffiffiffiffiffi
ffi
sinhðjpxÞ þ pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi coshðjpxÞ ex g þj q
2 2 2
¼
4jp 4 g2 þ j2 q2
Similarly, the other integrals in Eq. (7.2.18) can be evaluated and we have
( )
Bx c2 þ 1 c2 1
~x ¼ þ 2
u pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi coshðjpxÞ þ sinhðjpjxjÞ exp jxj g2 þ j2 q2
4c lð0Þ þj q
g 2 2 2 jp
By igðc2 1Þ þ jðc2 3Þ
þ 2 pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi sinhðjpjxjÞ exp jxj g2 þ j2 q2
4c lð0Þ jp g þj q
2 2 2
ð7:2:23aÞ
Bx igðc2 1Þ jðc2 3Þ
~
uy ¼ þ p ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
ffi sinhðjpjxjÞ exp jxj
g2 þ j2 q2
4c2 lð0Þ jp g2 þ j2 q2
( )
By c2 þ 1 c2 1
þ 2 pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi coshðjpxÞ sinhðjpjxjÞ exp jxj g2 þ j2 q2
4c lð0Þ
g þj q
2 2 2 jp
ð7:2:23bÞ
Our next task is to evaluate the inversion integral with respect to the parameter g.
The formal inversion of the displacement is expressed in terms of the fundamental
inversion integrals Ii ðx; yÞ; i ¼ 0; 1; 2,
2
Bx c2 1 c þ1 sinhðjpjxjÞ
ux ¼ þ expðjyÞ coshðjpxÞI 0 ðx; yÞ þ I 1 ðx; yÞ
4lð0Þ c2 c2 1 jp

By c2 1 sinhðjpjxjÞ c2 3
þ expðjyÞ I 2 ðx; yÞ þ j I0 ðx; yÞ
4lð0Þ c2 jp c2 1
ð7:2:24aÞ

Bx c2 1 sinhðjpjxjÞ c2 3
uy ¼ þ expðjyÞ I 2 ðx; yÞ j I 0 ðx; yÞ
4lð0Þ c2 jp c2 1

B y c2 1 c2 þ 1 sinhðjpjxjÞ
þ expðjyÞ coshðjpxÞI 0 ðx; yÞ I 1 ðx; yÞ
4lð0Þ c2 c2 1 jp
ð7:2:24bÞ
where the fundamental integrals Ii are defined by

Zþ1 exp jxj g2 þ j2 q2 igy
1
I0 ðx; yÞ ¼ pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi dg ð7:2:25Þ
2p g2 þ j2 q2
1
Zþ1
1
I1 ðx; yÞ ¼ exp jxj g2 þ j2 q2 igy dg ð7:2:26Þ
2p
1

Zþ1 exp jxj g2 þ j2 q2 igy
1
I2 ðx; yÞ ¼ ig pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi dg ð7:2:27Þ
2p g2 þ j2 q2
1
is defined by Eq. (7.2.14).

and the variable g
Observing and examining the above three Eqs. (7.2.25)–(7.2.27), we find the
relation among the fundamental integrals as
@I0 ðx; yÞ @I0 ðx; yÞ
I1 ðx; yÞ ¼ ; I2 ðx; yÞ ¼ ð7:2:28Þ
@jxj @y
Since the two integrals I1 and I2 can be derived from I0, it is enough to evaluate the
only one integral I0. For the integral I0 in Eq. (7.2.25), we introduce the variable
transform from g to the new variable 1,
1 ¼ g ¼ g þ ij ð7:2:29Þ
Equation (7.2.25) is transformed to the integral along the complex line from 1 þ
ij to þ1 þ ij,

Z
þ1þij
exp jxj 12 þ j2 q2 i1y
1
I0 ðx; yÞ ¼ pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi d1 ð7:2:30Þ
2p 1 2 þ j2 q 2
1þij
The above integral resembles the formula (Erdélyi 1954, p. 17, (27)),
Zþ1
1
pffiffiffiffiffiffiffiffiffiffiffiffiffiffi exp a x2 þ c2 expðibxÞdx ¼ 2K0 c a2 þ b2 ð7:2:31Þ
x2 þ c2
1
where K0 ð:Þ is the modified Bessel function of zeroth order. However, this formula
cannot be applied directly since the integration path for our integral of Eq. (7.2.30)
is not real. In order to apply the formula, we need to transform the integral to one
along the real path.
Let us consider the complex integral U along the rectangular closed loop
ABCDA shown in Fig. 7.3,

Z exp jxj 12 þ j2 q2 i1y
1
U¼ pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi d1 ð7:2:32Þ
2p 1 2 þ j2 q 2
ABCDA
Im(ς )
branch cut
+iκ q
−∞ + iκ +iκ +∞ + iκ
C B
Re {ς 2
+ (κ q) 2
}> 0
D A
−∞ +∞ Re(ς )
branch cut
−iκ q
Fig. 7.3 Transform of integration path from the complex line to the real line
The integrand has two branch points at 1 ¼ ijq. Two branch cuts along the
imaginary axis are also introduced as shown in the figure. Fortunately, neither of
! ! !
two branch cuts cross the integration lines DA and BC , especially the line BC
since qð¼1=ð1 mÞÞ [ 1. No other singular point is included in the closed loop
ABCDA and the integrals along the vertical straight lines, AB and CD, vanish at
infinity, since the real part of the radical is positive. Thus, the integral along the
! !
complex line CB is converted to one along the real axis DA . That is

Z exp jxj 12 þ j2 q2 i1y Z exp jxj 12 þ j2 q2 i1y
1 1
pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi d1 ¼ pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi d1
2p 1 2 þ j2 q 2 2p 1 2 þ j 2 q2
! !
CB DA
ð7:2:33Þ
Consequently, the integral (7.2.30) along the complex line is converted to that along
the real path,

Zþ1 exp jxj 12 þ j2 q2 i1y
1
I0 ðx; yÞ ¼ pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi d1 ð7:2:34Þ
2p 1 2 þ j2 q 2
1
Therefore, we can apply the integration formula Eq. (7.2.31) to the above integral
and have the exact expression for the integral I0 ,
1
I0 ðx; yÞ ¼ K0 jjjq x2 þ y2 ð7:2:35Þ
p
Substituting this Eq. (7.2.35) into Eq. (7.2.28), the other two integrals, I1 and I2 , are
exactly evaluated as
jjjqjxj
I1 ¼ pffiffiffiffiffiffiffiffiffiffiffiffiffiffi K1 jjjq x2 þ y2
p x2 þ y2

pffiffiffiffiffiffiffiffiffiffiffiffiffiffi ð7:2:36Þ
jjjqy
I2 ¼ pffiffiffiffiffiffiffiffiffiffiffiffiffiffi K1 jjjq x2 þ y2
p x2 þ y2
Finally, we substitute Eqs. (7.2.35) and (7.2.36) into Eq. (7.2.24) and rewrite the
displacement in terms of Green’s dyadic Gij ,
ux ¼ Bx Gxx ðx; yÞ þ By Gxy ðx; yÞ

ð7:2:37Þ
uy ¼ Bx Gyx ðx; yÞ þ By Gyy ðx; yÞ
where the exact expressions for the dyadic are
c2 1
Gxx ðx; yÞ ¼ expðjyÞ
4pc2 lð0Þ
2
c þ1 x sinhðjpxÞ
2 coshðjpxÞK0 ðjjjqr Þ þ jjjq K1 ðjjjqr Þ
c 1 r jp
ð7:2:38aÞ

c2 1 sinhðjpxÞ y c2 3
Gxy ðx; yÞ ¼ expðjyÞ jjjq K1 ðjjjqr Þ þ j 2 K0 ðjjjqr Þ
4pc2 lð0Þ jp r c 1
ð7:2:38bÞ

c2 1 sinhðjpxÞ y c2 3
Gyx ðx; yÞ ¼ expðjyÞ jjjq K1 ð jjjqr Þ j K0 ð jjjqr Þ
4pc2 lð0Þ jp r c2 1
ð7:2:38cÞ
c2 1
Gyy ðx; yÞ ¼ expðjyÞ
4pc2 lð0Þ
2
c þ1 x sinhðjpxÞ
2 coshðjpxÞK0 ðjjjqr Þ jjjq K1 ðjjjqr Þ
c 1 r jp
ð7:2:38dÞ
Here the radial distance r from the source is defined by

r¼ x2 þ y2 ð7:2:39Þ
7.2.1 2D Kelvin’s Solution for Homogeneous Media
Inspecting the dyadic in Eq. (7.2.38), we can find that there is only one inhomo-
geneity parameter, i.e. jð¼j=ð2hÞÞ. When this parameter vanishes, the medium is
homogeneous. Thus, taking the limit j ! 0 as
sinhðjpxÞ
x; cosh ðjpxÞ 1 ð7:2:40Þ
jp
1
K0 ðjjjqrÞ logðjjjqrÞ; jjjqK1 ðjjjqrÞ ð7:2:41Þ
r
we have the dyadic for the homogeneous medium, i.e. the 2D Kelvin’s solution,

x2
1
Gxx ðx; yÞ ¼ ðc 2
1Þ ð7:2:42aÞ
4pc2 l0 r
c2 1 xy
Gxy ðx; yÞ ¼ Gyx ðx; yÞ ¼ ð7:2:42bÞ
4pc2 l0 r 2

y2
1
Gyy ðx; yÞ ¼ ðc 2
1Þ ð7:2:42cÞ
4pc2 l0 r
where l0 is a uniform rigidity throughout the medium. The reader will find that this
dyadic is the same as the 2D static dyadic of Eq. (3.3.33) in Sect. 3.3.
7.3 Green’s Function for Torsional Waves in a Monoclinic

Material
The axisymmetric torsional deformation of an elastic solid is governed by the

equation of motion
1 @ðr 2 rrh Þ @rzh @ 2 uh

þ ¼ q qBh ð7:3:1Þ
r 2 @r @z @t2
and Hooke’s law

rrh C44 C45 erh
¼ ; C45 ¼ C54 ð7:3:2Þ
rzh C54 C55 ezh
where uh uh ðr; z; tÞ is the torsional/circumferential displacement, rrh and rzh are

torsional stresses and erh and ezh are strains. When C44 ¼ C55 ¼ l and
C45 ¼ C54 ¼ 0, the elastic solid is isotropic. On the other hand, when C44 6¼ C55
and C45 ¼ C54 6¼ 0, the solid is anisotropic and is called “monoclinic.” The
monoclinic material is one of anisotropic solids. As Green’s function for the iso-
tropic solid has been obtained in Sect. 3.7 in Chap. 3, we shall obtain the Green’s
function for the monoclinic material.
Two strains are defined by

1 @uh uh 1 @uh
erh ¼ ; ezh ¼ ð7:3:3Þ
2 @r r 2 @z
7.3 Green’s Function for Torsional Waves in a Monoclinic Material 223
We employ the same circumferential ring body force as that in Sect. 3.7,
dðr aÞ
Bh ¼ Q dðzÞdðtÞ ð7:3:4Þ
r
where Q is the magnitude of the body force and dð:Þ is Dirac’s delta function.
Substituting Eq. (7.3.2) into the equation of motion with Eq. (7.3.4), we have the
displacement equation,
@ 2 uh 1 @uh uh @ 2 uh b @uh 2 @ uh
2
1 @ 2 uh
þ þ 2b þ þ a
@r 2 r @r r2 @r@z r @z @z2 c2 @t2
Q dðr aÞ
¼ 2 dðzÞdðtÞ ð7:3:5Þ
c r
where
pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
a¼ C44 =C55 ; b ¼ C45 =C55 ; c¼ C55 =ð2qÞ ð7:3:6Þ
Our task in the present section is to obtain a particular solution corresponding to

the nonhomogeneous term of the ring force. In order to reduce the partial differ-
ential equation to the ordinary one, we apply the double integral transform: Laplace
transform with respect to the time t
Z1

0
and Fourier transform with respect to the axial length z,
Zþ1 Zþ1
f ðzÞ expðþifzÞdz , f ðzÞ ¼ ð7:3:8Þ
2p
1 1
Assuming the quiescent condition at an initial time,

@uh
uh jt¼0 ¼ ¼0 ð7:3:9Þ
@t t¼0

@uh @uh
uh jprffiffiffiffiffiffiffiffi
ffi ¼ ¼ ffi ¼0 ð7:3:10Þ
@r pffiffiffiffiffiffiffiffi
ffi @z prffiffiffiffiffiffiffiffi
2 þz2 !1
r2 þz2 !1 2 þz2 !1
we apply the double transform to the displacement equation (7.3.5). It follows that

uh
d2^ 1 dûh 1 ibf Q dðr aÞ
þ 2ibf 2þ þ ðafÞ þ ðs=cÞ ûh ¼ 2
2 2
dr 2 r dr r r c r
ð7:3:11Þ
In order to obtain the simpler Bessel equation, we assume the solution as a product
of two unknown functions. They are assumed as
ûh ¼ gðrÞUðrÞ ð7:3:12Þ
Substituting the above equation into Eq. (7.3.11), we have

0
1 1 g ibfg
g U 00 þ U 0 2 U þ 2ðg0 ibfgÞU 0 þ þ g00 2ibfg0 U
r r r
Q dðr aÞ
gfðafÞ2 þ ðs=cÞ2 gU ¼ 2 ð7:3:13Þ
c r
where the prime (.)′ denotes the differentiation with respect to the radial variable
r. Inspecting the above equation, if we assume
g0 ibfg ¼ 0 ) gðrÞ ¼ expðþibfrÞ ð7:3:14Þ
we have the nonhomogeneous Bessel equation for another unknown U(r),
1 1 Q dðr aÞ
U 00 þ U 0 U fðcfÞ2 þ ðs=cÞ2 gU ¼ 2 expðibfrÞ ð7:3:15Þ
r r c r
where the parameter γ is the combination of two anisotropy parameters,

c¼ a 2 b2 ð7:3:16Þ
Equation (7.3.15) is in the form of Bessel differential equation for the unknown U
(r). In order to obtain the particular solution, we apply the Hankel transform with
the order 1,
Z1 Z1
~f ðnÞ ¼ rf ðrÞJ1 ðnrÞdr , f ðrÞ ¼ n~f ðnÞJ1 ðnrÞdn ð7:3:17Þ
0 0
to Eq. (7.3.15), and then have the simple algebraic equation for the transformed
~
function. The transformed function UðnÞ is given by
~ Q J1 ðanÞ
UðnÞ ¼ 2 expðibfaÞ 2 ð7:3:18Þ
c n þ ðcfÞ2 þ ðs=cÞ2
Its formal Hankel inversion is given by
Z1
Q n
UðrÞ ¼ 2 expðibfaÞ J1 ðanÞJ1 ðrnÞdn ð7:3:19Þ
c n2 þ ðcfÞ2 þ ðs=cÞ2
0
The product of two Bessel functions is replaced with its integral form. From the
addition theorem of the Bessel function (Watson 1966, p. 358), we can easily derive
the integral representation for the product as
Zp
1
J1 ðanÞJ1 ðrnÞ ¼ J0 n r 2 þ a2 2ar cos u cos udu ð7:3:20Þ
p
0
This equation is the same as Eq. (3.7.13) in Chap. 3. We substitute the above
integral representation into Eq. (7.3.19) and exchange the order of integration. It
follows that

Zp Z1 J0 n r 2 þ a2 2ar cos u
Q
UðrÞ ¼ expðibfaÞ cos udu n dn
pc2 n2 þ ðcfÞ2 þ ðs=cÞ2
0 0
ð7:3:21Þ
The simple integration formula (Erdélyi 1954, vol. II, pp. 23, 12)
Z1
n
J0 ðbnÞdn ¼ K0 ðabÞ ð7:3:22Þ
n þ a2
2
0
where K0 ð:Þ is the modified Bessel function of the first kind, is applied to
Eq. (7.3.21). We have the single integral representation for UðrÞ,
Zp pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiqffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
Q
UðrÞ ¼ 2 expðibfaÞ K0 r 2 þ a2 2ar cos u ðcfÞ2 þ ðs=cÞ2 cos udu
pc
0
ð7:3:23Þ
The product of Eq. (7.3.12) has just been determined by Eqs. (7.3.14) and (7.3.23).
Thus the double transformed torsional displacement is given by
Zp pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiqffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
Q
uh ¼
^ expfþibfðr aÞg K0 r 2 þ a2 2ar cos u ðcfÞ2 þ ðs=cÞ2 cos udu
pc2
0
ð7:3:24Þ
Let us start for the inversion of the double transform. Firstly, we apply the
Fourier inversion integral with respect to the parameter f to the above equation and
exchange the order of integration. It yields
Zp
Q
uh ¼ 2 cos udu
pc
0
Z1 pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiqffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
1
K0 r 2 þ a2 2ar cos u ðcfÞ2 þ ðs=cÞ2 cos½ffz bðr aÞg df
p
0
ð7:3:25Þ
Applying the integration formula (Erdélyi 1954, p. 56, (43))
Z1
p
K0 a x2 þ b2 cosðcxÞdx ¼ pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi exp b a2 þ c2 ð7:3:26Þ
2 a2 þ c 2
0
to Eq. (7.3.25), we have

qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
Zp exp ðs=cÞ r 2 þ a2 2ar cos u þ fz bðr aÞg2
Q
uh ¼ qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi cos udu
2pcc2
0 r 2 þ a2 2ar cos u þ fz bðr aÞg2
ð7:3:27Þ
Before going to the Laplace inversion, we make the change of variable for the
integral, u ! t,
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
1
t¼ r 2 þ a2 2ar cos u þ fz bðr aÞg2 ð7:3:28Þ
c
The new variable t is a simple variable, not the real time at the present stage. But, at
the next stage we understand it as the real time. With the change of the variable,
Eq. (7.3.27) is simplified as
ZR2 =c (sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi)

Q 1 R22 ðctÞ2 ðctÞ2 R21
uh ¼ expðstÞdt ð7:3:29Þ
4pcc ar ðctÞ2 R21 R22 ðctÞ2
R1 =c
where
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
R1 ¼ ðr aÞ2 þ fz bðr aÞg2 ; R2 ¼ ðr þ aÞ2 þ fz bðr aÞg2
ð7:3:30Þ
Equation (7.3.29) is the Laplace transformed displacement and the integral is just in
the form of the Laplace transform. Therefore, the original function of the dis-
placement is its integrand and the Laplace inversion can be carried out by
inspection. Finally, we have the exact closed form for the displacement. That is the
Green’s function for the torsional ring source in the monoclinic material,
(sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi)
Q 1 R22 ðctÞ2 ðctÞ2 R21
uh ¼ Hðct R1 ÞHðR2 ctÞ ð7:3:31Þ
4pcc ar ðctÞ2 R21 R22 ðctÞ2
The reader will find that this equation can be reduced to Eq. (3.7.26) in Chap. 3
when the material is isotropic, a ! 1; b ! 0 ðc ! 1Þ; c ! cs .
7.4 Reflection of a Transient SH-Wave at a Moving

Boundary
As an example of a moving boundary problem, wave reflection at a moving edge is

discussed. The simplest single SH-wave is considered and a conversion formula
between two Laplace transforms is developed so that the integral transform method
can be applicable to the moving boundary problem.
Let us consider a semi-infinite elastic solid and define Cartesian coordinates
(x, y) as shown in Fig. 7.4. The initial position of a moving edge of the half space is
at x ¼ l and the edge moves toward the positive x-direction with velocity V. A point
and stationary source of SH-wave is placed at the coordinate origin and is assumed
as an impulsive body force with magnitude Q. The SH-wave field produced by the
impulsive source is governed by the nonhomogeneous displacement equation,
@ 2 uz @ 2 uz 1 @ 2 uz Q
þ 2 ¼ 2 2 2 dðxÞdðyÞdðtÞ ð7:4:1Þ
@x 2 @y cs @t cs
Fig. 7.4 Moving edge of a y t=0 t>0

semi-infinite elastic solid
SH-source
Vt
Q
x
x=0 x=l
Moving boundary
where uz is the anti-plane displacement and cs is SH-wave velocity. The stress

components follow Hooke’s law as
@uz @uz
rzx ¼ l ; rzy ¼ l ð7:4:2Þ
@x @y
where l is rigidity.
We assume the stress-free condition at the moving edge x ¼ Vt þ l,
rxz jx¼Vtþl ¼ 0 ð7:4:3Þ
The radiation condition at infinity

@uz @uz
uz jpxffiffiffiffiffiffiffiffi
ffi
2 þy2 !1 ¼ ¼ ffi ¼0 ð7:4:4Þ
ffi
x2 þy2 !1
and the quiescent condition at an initial time

@uz
uz jt¼0 ¼ ¼0 ð7:4:5Þ
@t t¼0
are also assumed. Equations (7.4.1)–(7.4.5) constitute the present elastodynamic

problem.
The wave field can be decomposed into two parts: the incident wave uinc and the
reflected wave urefl: . The total displacement (wave field) is the sum of the two,
uz ðx; y; tÞ ¼ uinc ðx; y; tÞ þ urefl ðx; y; tÞ ð7:4:6Þ

7.4 Reflection of a Transient SH-Wave at a Moving Boundary 229
where uinc and urefl are a particular and the general solution of the non-homoge-
neous wave equation (7.4.1), respectively.
(1) Incident wave
The incident wave is generated by the impulsive source and radiates from the
source point. The incident wave is given as the particular solution of the nonho-
mogeneous wave equation for the displacement,
@ 2 uinc @ 2 uinc 1 @ 2 uinc Q

þ ¼ 2 dðxÞdðyÞdðtÞ ð7:4:7Þ
@x2 @y2 c2s @t2 cs
The particular solution has already been obtained in Sect. 2.4, i.e.
Q 1
uinc ðx; y; tÞ ¼ qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi H cs t x2 þ y2 ð7:4:8Þ
2pcs
ðcs tÞ2 ðx2 þ y2 Þ
However, we do not use this solution for the latter analysis, instead, the Fourier
transformed solution is employed. We apply the triple integral transform to
Eq. (7.4.7): Laplace transform with respect to the time,
Z1 Z
1
f ðsÞ ¼ f ðtÞ expðstÞdt; f ðtÞ ¼ f ðsÞ expðþstÞds ð7:4:9Þ
2pi
0 BrðsÞ
where “Br(s)” in the inverse transform denotes the Bromwich line in the complex
s-plane, and the double Fourier transform with respect to the space variables x and y,
Zþ1 Zþ1
2p
1 1
Zþ1 Zþ1
2p
1 1
The nonhomogeneous wave equation (7.3.7) is reduced to the simple algebraic

equation for the transformed displacement,
Q
fn2 þ g2 þ ðs=cs Þ2 g~uinc ¼ ð7:4:12Þ
c2s
and the particular solution in the transformed domain is then given by
Q 1
~uinc ¼ ð7:4:13Þ
c2s n2 þ g2 þ ðs=cs Þ2
Its formal Fourier inversion with respect to the parameter n is given by
Zþ1
Q 1
~uinc ¼ expðinxÞdn ð7:4:14Þ
2pc2s n þ
2
g2 þ ðs=cs Þ2
1
and is easily evaluated by the application of the formula (2.1.22), to give
Q
~uinc ¼ expðas jxjÞ ð7:4:15Þ
2c2s as
where the radical as is defined by

as ¼ g2 þ ðs=cs Þ2 ; Reðas Þ [ 0 ð7:4:16Þ
Furthermore, the formal Laplace inversion for the incident wave of Eq. (7.4.15) is
given by the Bromwich integral,
Z
Q 1 1
~uinc ¼ expðas jxj þ stÞds ð7:4:17Þ
2c2s 2pi as
BrðsÞ
where BrðsÞ denotes the Bromwich line from cs i1 to cs þ i1 in the complex

ðincÞ
s-plane. The transformed stress r ~xz produced by the incident wave is also
obtained in the form of Laplace inversion integral,
Z
ðincÞ Q 1
~xz
r ¼ 2 sgnðxÞ expðas jxj þ stÞds ð7:4:18Þ
2cs 2pi
BrðsÞ
where sgn(x) is the sign function defined by Eq. (7.1.17).

In the subsequent discussion for the stress-free condition, the integral form of
Eq. (7.4.18) is convenient since we develop the conversion formula for two dif-
ferent Laplace transforms. Therefore, we shall stop the discussion for the incident
wave.
(2) Reflected wave

The reflected wave is the general solution of the homogeneous displacement
equation,
@ 2 urefl @ 2 urefl 1 @ 2 urefl

þ ¼ 2 ð7:4:19Þ
@x 2 @y 2 cs @t2
The reflected wave is generated at the moving edge and runs toward the negative x-
direction. So, we transform the Eq. (7.4.19) to the moving coordinate system
ðz; y; tÞ where the moving coordinate z is defined by
z ¼ Vt þ l x ð7:4:20Þ
Due to this coordinate transform, the displacement of the reflected wave has the
different set of three variables, as
urefl: ðx; y; tÞ urefl: ðz; y; tÞ ð7:4:21Þ
The wave equation (7.3.19) is transformed to
@ 2 urefl 2M @ 2 urefl 1 @ 2 urefl @ 2 urefl

m2 2 þ ¼0 ð7:4:22Þ
@z2 c2 @z@t cs @t2 @y2
where Mach number M and its related parameter m are defined by

M ¼ V=cs ; m¼ 1 M2 ð7:4:23Þ
In order to solve the differential equation (7.4.22), we apply the double integral
transform: Laplace transform with respect to the time variable in the moving
coordinate system. It is defined by
Z1 Z

1
f ðpÞ ¼ f ðtÞ expðptÞdt; f ðtÞ ¼ f
ðpÞ expðptÞdp ð7:4:24Þ
2pi
0 BrðpÞ
Fourier transform with respect to the space variable y is defined by
Zþ1 Zþ1
2p
1 1
It should be noted that the Laplace transform is not the same as that in Eq. (7.4.9),
but that the Fourier transform is the same as that in Eq. (7.4.11). This is because we
have introduced the moving coordinate system along the x-axis, not along the
y-axis. Thus, it should be understood that the time variable t in the moving coor-
dinate system is slightly different from the time in the fixed coordinate system since
the variable z includes the time variable.
Using the quiescent condition at the initial time and the convergence condition at
infinity, y ! 1, the wave equation (7.4.22) is transformed to the ordinary dif-
ferential equation,
d2~ d~u
n o
m2 refl
2ðp=c s ÞM refl
g 2
þ ðp=c s Þ2
~u
refl ¼ 0 ð7:4:26Þ
dz2 dz
This solution which satisfies the convergence condition at z ! þ1 is given by

Mp z
~u
refl ¼ Aðg; pÞ exp bs þ ð7:4:27Þ

mcs m
where Aðg; pÞ is an unknown coefficient and the radical bs is defined by

sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
2ffi
p
bs ¼ g2 þ ; Reðbs Þ 0 ð7:4:28Þ
mcs
Then, the formal Laplace inversion of the reflected wave is given by the Bromwich
integral,
Z
1 Mp z
~
urefl ¼ Aðg; pÞ exp bs þ expðptÞdp ð7:4:29Þ
2pi mcs m
BrðpÞ
where BrðpÞ denotes the Bromwich line from cp i1 to cp þ i1 in the complex

p-plane. In order to obtain the stress, we change the moving coordinate z back to the
original one, z ¼ Vt þ l x, to get
Z
1 Mp Vt þ l x
~
urefl ¼ Aðg; pÞ exp bs þ þ pt dp ð7:4:30Þ
2pi mcs m
BrðpÞ
We then substitute this expression into the first of Eq. (7.4.2). The stress produced
by the reflected wave is given by
Z
ðreflÞ 1 1 Mp 1 Mp
~xz
r ¼ b Aðg; pÞ exp bs þ ðVt þ l xÞ þ pt dp
2pi m s mcs m mcs
BrðpÞ
ð7:4:31Þ
The unknown coefficient is determined by the stress-free condition on the

moving edge. The stress is also the sum of two wave contributions, the incident and
the reflected waves. Thus, the Fourier transformed boundary condition is given by
the sum of each wave contribution,
ðincÞ ðreflÞ
~xz ¼ r
r ~xz ~xz
þr ¼ 0; at x ¼ Vt þ l ð7:4:32Þ
Substituting Eqs. (7.4.18) and (7.4.31) into the above condition, we obtain
Z
Q 1
expfas ðVt þ lÞ þ stgds
2c2s 2pi
BrðsÞ
Z
1 1 Mp
þ bs Aðg; pÞ expðptÞdp ¼ 0 ð7:4:33Þ
2pi m mcs
BrðpÞ
This is the integral equation for the unknown coefficient Aðg; pÞ. The second
integral on the left-hand side is in the form of Laplace inversion integral with
respect to the parameter p, however, the first integral is not (in the form of Laplace
inversion integral). We thus have two different Laplace inversion integrals in a
single equation and it is not possible to apply any one of the Laplace transforms to
reduce it to an algebraic equation for the unknown. Thus, the inversion integral in
the first term must be converted to that with respect to the parameter p, i.e. the
matching of the inversion integral.
Let us consider the complex integral UA whose integrand is the same as that in
the first integral in Eq. (7.4.33),
Z
1
UA ¼ expðas lÞ expfðs as VÞtgds ð7:4:34Þ
2pi
C
where the radical as is defined by Eq. (7.4.16). The integration loop C is discussed
now. The integrand has two branch points s ¼ ics g and corresponding two branch
cuts are introduced along the imaginary axis in the complex s-plane. These are shown
in Fig. 7.5. If we introduce the variable transform from s to the new variable p as
p ¼ s as V ð7:4:35Þ
its inverse is given by

qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
1
s=cs ¼ p=c s M ðmgÞ2 þ ðp=cs Þ2 ð7:4:36Þ
m2
In order to determine the multiple sign ðÞ, we examine Eq. (7.4.35). We have the
following asymptotic relation between two variables:
p
s ; jpj ! 1 ð7:4:37Þ
1M
Fig. 7.5 Closed loop C for Im( s )

the complex integral UA in B
Eq. (7.4.34) γ s + i∞ p = γ p + i∞
icsη
γs Re( s )
−icsη s( p)
γ s − i∞ p = γ p − i∞
This relation must be hold for the inverse of the variable transform, i.e. we have to
choose the positive sign in Eq. (7.4.36). Then, the suitable inversion for the variable
transform is
1
s=cs ¼ p=c s þ M ðmgÞ2 þ ðp=cs Þ2 ð7:4:38Þ
m2
When the new variable p varies along the Bromwich line BrðpÞ in the complex p-
plane, i.e.
BrðpÞ: cp i1 ! cp þ i1 ð7:4:39Þ
where cp is a proper constant, the inverse s given by Eq. (7.4.38) moves along the
bumped curve ACB in Fig. 7.6, but it is almost straight at infinity. At infinity, the
inverse function takes the asymptotic form
8
> c þ i1
< p ; p ! cp þ i1
M
s c1 ð7:4:40Þ
>
: p i1 ; p ! cp i1
1M
Fig. 7.6 Closed loop C for the complex integral UB
So, if we take the real part of the Bromwich line BrðsÞ from cs i1 to cs þ i1 as
cp
cs ¼ ð7:4:41Þ
1M
the bumped curve can be connected to the edge of the Bromwich line BrðsÞ and
then, these two curves constitute a closed loop ACBA as shown in the figure. Thus,
we employ this closed loop C for the complex integral UA in Eq. (7.4.34).
Fortunately, the real part of the two edges is sufficiently large and the two
Bromwich lines
BrðpÞ: cp i1\p\cp þ i1
ð7:4:42Þ
BrðsÞ: cs i1\s\cs þ i1
have sufficiently large real parts so that all singular points of the integrand are in the
left side of the Bromwich line BrðsÞ. Then, Cauchy’s integral theorem can be
applied to the complex integral. Since the closed loop C(ACBA) does not include
any singular point, the integral along BrðsÞ is converted to a parametric integral
along the bumped curve. That is,
cZ cp þi1
s þi1 Z
1 1 ds
expðas lÞ expfðs as VÞtgds ¼ expðas lÞ expðptÞdp
2pi 2pi dp s¼sðpÞ
cs i1 cp i1
ð7:4:43Þ
where sðpÞ is the inverse given by Eq. (7.4.38).

Consequently, we could convert the Laplace inversion integral with respect to

the parameter s to that with respect to the parameter p and thus, Eq. (7.4.33) is
rewritten in the form of the single Laplace inversion integral with respect to the
parameter p, i.e.
Z
1 1 Mp
bs Aðg; pÞ expðptÞdp
2pi m mcs
BrðpÞ
Z
Q 1 ds
¼ 2 expðas lÞ expðptÞdp ð7:4:44Þ
2cs 2pi dp s¼sðpÞ
BrðpÞ
Both sides of the above equation have the same form of Laplace inversion integral
with respect to the single parameter p. Applying the Laplace transform with respect
to the time t to both sides, we obtain the simple algebraic equation for the unknown
coefficient,

1 Mp Q ds
bs Aðg; pÞ ¼ 2 expðas lÞ ð7:4:45Þ
m mcs 2cs dp s¼sðpÞ
Thus, the coefficient is determined as

Q m ds
Aðg; pÞ ¼ expða s lÞ ð7:4:46Þ
2c2s bs ðMpÞ=ðmcs Þ dp s¼sðpÞ
Using the relations derived from Eqs. (7.4.35) and (7.4.38), the radical and the
gradient are given by
1
as ¼ fMðp=cs Þ þ mbs g ð7:4:47Þ
m2
dsðpÞ Mðp=mcs Þ þ bs
¼ ð7:4:48Þ
dp m2 bs
The coefficient is then rewritten in the explicit form,

Q 1 bs þ Mðp=mcs Þ Mðp=cs Þ þ mbs
Aðg; pÞ ¼ exp l ð7:4:49Þ
2c2s mbs bs Mðp=mcs Þ m2
Lastly, substituting Eq. (7.4.49) into Eq. (7.4.27), the double transformed dis-
placement produced by the reflected wave is given by

Q 1 bs þ Mðp=mcs Þ Mðp=cs Þ þ mbs Mp z
u
~ refl ¼ exp l exp b þ

2c2s mbs bs Mðp=mcs Þ m2 s
mcs m
ð7:4:50Þ
(3) Cagniard’s technique

We have obtained the displacement produced by the reflected wave, but in the
transformed domain. As the incident wave has already been obtained in the explicit
form by Eq. (7.4.8), our next task is to perform the double inversion for the
displacement in Eq. (7.4.50). The double inversion is carried out by applying the
Cagniard-de Hoop technique. The formal Fourier inversion with respect to the
parameter g is given by
Zþ1
Q 1 bs þ Mðp=mcs Þ Mðp=cs Þðl zÞ bs ðz þ lÞ
u
refl ¼ exp igy dg

4pc2s mbs bs Mðp=mcs Þ m2 m
1
ð7:4:51Þ
Since the radical bs ¼ g2 þ ðp=mcs Þ2 has the Laplace transform parameter p, we
introduce the variable transform from g to the new variable 1 as
g ¼ ðp=mcs Þ1 ð7:4:52Þ
Equation (7.4.51) is rewritten as
Zþ1 pffiffiffiffiffiffiffiffiffiffiffiffiffi
Q 1 12 þ 1 þ M Mðl zÞ z þ l pffiffiffiffiffiffiffiffiffiffiffiffi
ffi
u
refl ¼ pffiffiffiffiffiffiffiffiffiffiffiffiffi pffiffiffiffiffiffiffiffiffiffiffiffiffi exp ðp=mcs Þ þ 12 þ 1 þ i1y d1

4pmc2s 12 þ 1 12 þ 1 M m m
1
ð7:4:53Þ
Now, let us consider the complex integral UB whose integrand is the same as that
of Eq. (7.4.53),
Z pffiffiffiffiffiffiffiffiffiffiffiffiffi
1 1 12 þ 1 þ M Mðl zÞ z þ l pffiffiffiffiffiffiffiffiffiffiffiffi
ffi
UB ¼ pffiffiffiffiffiffiffiffiffiffiffiffiffi pffiffiffiffiffiffiffiffiffiffiffiffiffi exp ðp=mcs Þ þ 1 þ 1 þ i1y d1
2
2p 12 þ 1 12 þ 1 M m m
C
ð7:4:54Þ
The closed loop C is discussed here. The integrand has two branch points at 1 ¼ i,
and two branch cuts are thus introduced along the imaginary axis in the complex 1-
plane. The vanishing point of the denominator, 12 þ 1 M ¼ 0, gives two poles
at 1 ¼ im on the imaginary axis; but they are smaller in magnitude than the branch
points.
The Cagniard’s path is determined by the variable transform from 1 to the new
variable t as

1 Mðl zÞ z þ l pffiffiffiffiffiffiffiffiffiffiffiffi
ffi
t¼ þ 12 þ 1 þ i1y ð7:4:55Þ
mcs m m
Solving for 1, the Cagniard’s path is given by

qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
imyfm2 cs t Mðl zÞg ðz þ lÞ fm2 cs t Mðl zÞg2 fðz þ lÞ2 þ ðmyÞ2 g
1ðÞ
s ¼
ðz þ lÞ2 þ ðmyÞ2
ð7:4:56Þ
The saddle point of the Cagniard’s path is

my
1saddle ¼ i qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi ð7:4:57Þ
at
1
t¼ Mðl zÞ þ ðz þ lÞ2 þ ðmyÞ2 ð7:4:58Þ
m2 cs
If we take the closed loop composed of the Cagniard’s path, the straight line along
the real axis and two large arcs which connect the straight line with the Cagniard’s
path, the pole 1 ¼ im is inside of the loop when
mjyj zþl
m\ qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi ) jyj [ ð7:4:59Þ
M
but outside when
mjyj zþl
m [ qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi ) jyj\ ð7:4:60Þ
M
The location of the pole depends on this inequality. The contribution from the pole
at 1 ¼ im is included only when the inequality of Eq. (7.4.59) holds. Thus, we
determine the closed loop C as ABCDOEA shown in Fig. 7.6. Applying Cauchy’s
theorem to the complex integral UB in Eq. (7.4.54), the line integral along the real
axis is converted to the sum of the integral along the Cagniard’s path and the
residue at the pole. That is
1 1 12 þ 1 þ M Mðl zÞ z þ l pffiffiffiffiffiffiffiffiffiffiffiffi
ffi
pffiffiffiffiffiffiffiffiffiffiffiffiffi pffiffiffiffiffiffiffiffiffiffiffiffiffi exp ðp=mcs Þ þ 12 þ 1 þ i1y d1
2p 12 þ 1 12 þ 1 M m m
1
2 ( pffiffiffiffiffiffiffiffiffiffiffiffiffi ) ( pffiffiffiffiffiffiffiffiffiffiffiffiffi ) 3
Z1
1 1
4 pffiffiffiffiffiffiffiffiffiffiffiffi 12 þ 1 þ M d1 1 12 þ 1 þ M d1 5ept dt
¼ ffi pffiffiffiffiffiffiffiffiffiffiffiffiffi pffiffiffiffiffiffiffiffiffiffiffiffiffi pffiffiffiffiffiffiffiffiffiffiffiffiffi
2p pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi 12 þ 1 12 þ 1 M dt 1¼1ðþÞ 12 þ 1 12 þ 1 M dt 1¼1ðþÞ
s s
MðlzÞþ ðzþlÞ2 þðmyÞ2
m 2 cs
" pffiffiffiffiffiffiffiffiffiffiffiffiffi #
2pi 1 þ im 12 þ 1 þ M Mðl zÞ z þ l pffiffiffiffiffiffiffiffiffiffiffiffi
ffi zþl
pffiffiffiffiffiffiffiffiffiffiffiffiffi pffiffiffiffiffiffiffiffiffiffiffiffiffi exp ðp=mcs Þ þ 12 þ 1 þ i1y H jyj
2p 12 þ 1 12 þ 1 M m m
1¼im
M
ð7:4:61Þ
Rearranging the second residue term in the right hand side, we obtain the somewhat
simpler form,
1 1 12 þ 1 þ M Mðl zÞ z þ l pffiffiffiffiffiffiffiffiffiffiffiffiffi
pffiffiffiffiffiffiffiffiffiffiffiffiffi pffiffiffiffiffiffiffiffiffiffiffiffiffi exp ðp=mcs Þ þ 12 þ 1 þ i1y d1
2p 12 þ 1 12 þ 1 M m m
1
2 ( pffiffiffiffiffiffiffiffiffiffiffiffiffi ) ( pffiffiffiffiffiffiffiffiffiffiffiffiffi ) 3
Z1
1 4 1 12 þ 1 þ M d1 1 12 þ 1 þ M d1 5ept dt
¼ p ffiffiffiffiffiffiffiffiffiffiffiffi
ffi p ffiffiffiffiffiffiffiffiffiffiffiffi
ffi
2p pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi 12 þ 1 12 þ 1 M dt 1¼1ðþÞ 12 þ 1 12 þ 1 M dt 1¼1ðþÞ
s s
MðlzÞþ ðzþlÞ2 þðmyÞ2
m2 c s

2M p 2Ml zþl
þ exp þ my H jyj
m mcs m M
ð7:4:62Þ
Substituting this equation into Eq. (7.4.53), we obtain the Laplace transformed
displacement as the sum of the Laplace transform integral and the simple expo-
nential function,
2 ( pffiffiffiffiffiffiffiffiffiffiffiffiffi ) 3
1 12 þ 1 þ M d1
6 pffiffiffiffiffiffiffiffiffiffiffiffiffi pffiffiffiffiffiffiffiffiffiffiffiffiffi 7
Z1 6 1 2þ1 1 2 þ 1 M dt 7
Q 6 1¼1
ðþÞ
7 pt
u
¼ 6 ( )
s
7e dt
refl 2
4pmcs 6 p ffiffiffiffiffiffiffiffiffiffiffiffi
ffi 7
4 1 1 2 þ 1 þ M d1 7
5
MðlzÞþ 2
ðzþlÞ þðmyÞ 2 pffiffiffiffiffiffiffiffiffiffiffiffiffi pffiffiffiffiffiffiffiffiffiffiffiffiffi
m2 cs 1 þ 1 1 þ 1 M dt 1¼1ðþÞ
2 2
s
MQ p 2Ml zþl
þ 2 2 exp þ my H jyj
m cs mcs m M
ð7:4:63Þ
The first term is just in the form of Laplace transform integral and is easily inverted
by inspection. The second term is the exponential function and the simple Laplace
inversion formula,
L1 ½expðapÞ ¼ dðt aÞ ð7:4:64Þ

is applied. Thus, the Laplace inversion is easily carried out and we get the final form
for the displacement produced by the reflected wave as
0 qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi1
Q Mðl zÞ þ ðz þ lÞ2 þ ðmyÞ2
urefl ¼ H @t A
4pmc2s m 2 cs
2( pffiffiffiffiffiffiffiffiffiffiffiffiffi ) ( pffiffiffiffiffiffiffiffiffiffiffiffiffi ) 3
1 1 2 þ 1 þ M d1 1 1 2 þ 1 þ M d1
4 pffiffiffiffiffiffiffiffiffiffiffiffiffi pffiffiffiffiffiffiffiffiffiffiffiffiffi pffiffiffiffiffiffiffiffiffiffiffiffiffi pffiffiffiffiffiffiffiffiffiffiffiffiffi 5
12 þ 1 12 þ 1 M dt 1¼1ðþÞ 12 þ 1 12 þ 1 M dt 1¼1ðþÞ

s s
MQ 1 2Ml zþl
2 2d t þ my H jyj
m cs mcs m M
ð7:4:65Þ
Fortunately, we can simplify the above equation and get an exact closed form
solution. Substituting Eq. (7.4.56) into Eq. (7.4.55), we have for the radical
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
pffiffiffiffiffiffiffiffiffiffiffiffiffi ðz þ lÞfm2 cs t Mðl zÞg iðmyÞ fm2 cs t Mðl zÞg2 fðz þ lÞ2 þ ðmyÞ2 g
12 þ 1 ¼ n o
ð7:4:66Þ
and thus we have the simple expression for the complex valued function as
( pffiffiffiffiffiffiffiffiffiffiffiffiffi ) ( pffiffiffiffiffiffiffiffiffiffiffiffiffi )
1 12 þ 1 þ M d1 1 12 þ 1 þ M d1
pffiffiffiffiffiffiffiffiffiffiffiffiffi pffiffiffiffiffiffiffiffiffiffiffiffiffi pffiffiffiffiffiffiffiffiffiffiffiffiffi pffiffiffiffiffiffiffiffiffiffiffiffiffi
12 þ 1 12 þ 1 M dt 1¼1ðþÞ 12 þ 1 12 þ 1 M dt 1¼1ðþÞ
s s
2
2m cs 2
T M R ðmyÞ2 2 2
¼ pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi 2
T R T 2Mðz þ lÞT þ M 2 R2 ðmyÞ
2 2 2
ð7:4:67Þ
where
T ¼ m2 cs t Mðl zÞ; R¼ ðz þ lÞ2 þ ðmyÞ2 ð7:4:68Þ
Then, Eq. (7.4.65) is simplified as the exact closed form solution for the reflected
wave,
cs m 1 T 2 M 2 R2 ðmyÞ2
urefl ¼ H ðT RÞ pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
Q 2p T 2 R2 T 2 2Mðz þ lÞT þ M 2 R2 ðmyÞ2
ð7:4:69Þ
M 2Ml zþl
d mðcs tÞ þ my H jyj
m m M
(4) Wave fronts

Disturbed regions and wave fronts are easily obtained. They are given by the
conditionals of the step and delta functions in Eq. (7.4.69). Returning to the original
fixed coordinate from the moving coordinate, we find that the step function attached
to the first term in Eq. (7.4.69) gives the disturbed region by the reflected wave:
0 qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi1
2 2
Mðl zÞ þ ðz þ lÞ2 þ ðmyÞ2
Wrefl: : H @t A ) cs t 2lM x 2l þ y2
m2 cs m2 m2
ð7:4:70Þ
This is a circular region, but not a full circle since the edge is moving and the
conditional for x is imposed as x Mcs t þ l. This reflected wave is shown in Fig. 7.7.
The second term includes the product of delta and step functions.
l ⎛ 2 ⎞ y
cs t = ⎜ + 1⎟
M ⎝ m2 ⎠
M
B C
A Q
cs t
2l / m 2 x
O
r = 3l / M
xedge = 2l (1 + 1 / m 2 )
2
Fig. 7.7 Wave fronts for incident and reflected waves at the time cs t ¼ Ml m2 þ1

1 2Ml zþl
Wflat : d t þ my H jyj
mcs m M
ð7:4:71Þ
2Ml
) d cs t 2 y H ðMjyj x Mcs t 2lÞ
m
Examining the product, we find that the delta function gives a line singular wave
and the step function restricts its line segment BC (in the figure). Thus, the wave is
given as
2Ml
Flat wave: Wflat : cs t ¼ y þ ð7:4:72Þ
m2
x 2l
Supporting region: jyj [ þ cs t ð7:4:73Þ
M
The incident wave has already been given by Eq. (7.4.8) and it is the circular
region. Its front is given by

Winc : H cs t x2 þ y2 ) ðcs tÞ2 x2 þ y2 ; x\Mðcs tÞ þ l ð7:4:74Þ
Figure 7.7 shows the typical wave front shape for the incident and reflected waves.
7.5 Wave Scattering by a Rigid Inclusion

in an Inhomogeneous Elastic Solid
The problem of wave scattering is one of the traditional elastodynamic problems.

However, as far as the author knows, no exact closed form solution for transient
(impulsive) waves is known for isotropic homogeneous elastic and acoustic
materials (Friedlander 1958; Pao and Mow 1973). Only one exact closed form
solution (Green’s function) was obtained for SH-waves in an inhomogeneous
elastic solid (Watanabe 1982). The present section shows the exact closed form of
Green’s function for the wave scattering problem in a radially inhomogeneous
elastic solid.
Let us consider a radially inhomogeneous elastic solid with a borehole and take
the cylindrical coordinate ðr; h; zÞ. We assume that the shear modulus l and the
density q are functions of the radial distance r from the coordinate origin as
l l0 ðr=aÞk ; q q0 ðr=aÞk2 ð7:5:1Þ
where l0 and q0 are the reference values at r ¼ a and k is an inhomogeneity

parameter. Upon this assumption, the shear wave velocity is the linear function of
the radial distance as
7.5 Wave Scattering by a Rigid Inclusion … 243
Rigid inclusion (r ,θ )
θ =0
θ = ±π
a
SH-wave source
Fig. 7.8 A circular rigid inclusion and a SH-wave source in an inhomogeneous elastic solid
pffiffiffiffiffiffiffiffi pffiffiffiffiffiffiffiffiffiffiffiffi
cs ¼ l=q ¼ cs0 ðr=aÞ; cs0 ¼ l0 =q0 ð7:5:2Þ
A rigid circular (cylindrical) inclusion with radius a is placed at the coordinate

origin and SH-wave source is also placed at ðr ¼ b; h ¼ 0Þ as shown in Fig. 7.8.
Assuming the anti-plane deformation, i.e. SH-wave problem, the equation of
motion and Hooke’s law are given by
@rrz 1 1 @rhz @ 2 uz
þ rrz þ þ qBz ¼ q 2 ð7:5:3Þ
@r r r @h @t
@uz 1 @uz
rrz ¼ l ; rhz ¼ l ð7:5:4Þ
@r r @h
where uz and rrz ; rhz are the anti-plane displacement and its associated stress
components. The body force Bz is assumed as the point/line wave source,
dðr bÞ
Bz ¼ B0 dðhÞdðtÞ ð7:5:5Þ
r
where B0 is the magnitude of the source and dð:Þ is Dirac’s delta function.
As the rigid circular inclusion is inserted in the hole, the elastic solid occupies its
outer region a\r\1 and the anti-plane displacement is fixed at the boundary, i.e.
the fixed boundary condition for the elastic solid, is
uz jr¼a ¼ 0; 0 t\1 ð7:5:6Þ
In addition to the above boundary condition, we employ the radiation condition at

infinity as

@uz
uz jr!1 ¼ ¼0 ð7:5:7Þ
@r r!1
The above equations from (7.5.1) to (7.5.7) constitute our wave scattering
problem. Substituting Eqs. (7.5.4) and (7.5.5) into the equation of motion (7.5.3),
we have the single displacement equation as
0
@ 2 uz l 1 @uz 1 @ 2 uz q @ 2 uz q dðr bÞ
þ þ þ 2 2 ¼ B0 dðhÞdðtÞ ð7:5:8Þ
@r 2 l r @r r @h l @t 2 l r
where the prime denotes the differentiation with respect to the radial variable as
dl
l0 ¼ ð7:5:9Þ
dr
Recalling Eqs. (7.5.1) and (7.5.2), the displacement Eq. (7.5.8) is rewritten as
@ 2 uz k þ 1 @uz 1 @ 2 uz 1
a2 @ 2 uz B0
a2 dðr bÞ
þ þ ¼ dðhÞdðtÞ
@r 2 r @r r 2 @h2 c2s0 r @t2 c2s0 r r
ð7:5:10Þ
Now, we start to solve the above differential equation. The variable transform for
the radial variable
z ¼ logðr=aÞ , r ¼ a expðzÞ ¼ aez ð7:5:11Þ
is introduced. Equation (7.5.10) is reduced to the nonhomogeneous differential

equation with constant coefficients as
2 2 2
@ 2 uz @uz @ 2 uz a @ uz a dðaez bÞ
þ k þ ¼ B0 dðhÞdðtÞ ð7:5:12Þ
@z 2 @z @h 2 cs0 @t 2 cs0 aez
Then, we apply the Laplace transform with respect to the time t and the Fourier
finite transform with respect to the circumferential angle θ. They are defined in
Chap. 1 as
Z1
Laplace transform: f ðsÞ ¼ f ðtÞ expðstÞdt ð7:5:13Þ
0
Zp
1 Xþ1
Fourier finite transform: fn ¼ f ðhÞ expðþinhÞdh , f ðhÞ ¼ fn expðinhÞ
2p n¼1
p
ð7:5:14Þ
Applying the double transform with the quiescent condition at an initial time,

@uz
uz jt¼0 ¼ ¼0 ð7:5:15Þ
@t t¼0
Equation (7.5.12) is reduced to the simple ordinary differential equation,

2
d 2 uzn duzn 2 a dðaez bÞ
2
þ k fn 2
þ ðas=c s0 Þ gu zn ¼ B 0 ð7:5:16Þ
dz dz cs0 aez
The homogeneous solution which satisfies the radiation condition (7.5.7) is easily
obtained as
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
uðhÞ
zn ¼ Cn ðsÞ exp z n 2 þ ðas=c Þ2 þ ðk=2Þ2 þ k=2
s0 ð7:5:17Þ
where Cn ðsÞ is an unknown coefficient to be determined by the boundary condition.

In order to obtain a particular solution corresponding to the wave source B0 , we
apply the Fourier transform with respect to the variable z to the ordinary differential
equation (7.5.16). The transform is defined by
Z1 Z1
f ðzÞ expðþifzÞdz , f ðzÞ ¼ ð7:5:18Þ
2p
1 1
We have the simple algebraic equation for the triple transformed displacement,
n o
a 2
f2 þ ikf þ n2 þ ðas=cs0 Þ2 u^ðpÞ
zn ¼ B0 expfþif logðb=aÞg ð7:5:19Þ
bcs0
where the source (nonhomogeneous) term is evaluated as
2 Z1
a dðaez bÞ
B0 expðþifzÞdz
cs0 aez
1
2 Z1
a dðr bÞ dr ð7:5:20Þ
¼ B0 expfþif logðr=aÞg
cs0 r r
0
2
a
¼ B0 expfþif logðb=aÞg
bcs0
Thus, we have the displacement in the triple transformed domain,

2
a expfþif logðb=aÞg
uðpÞ
^ zn ¼ B0 ð7:5:21Þ
bcs0 ðf þ ik=2Þ2 þ n2 þ ðas=cs0 Þ2 þ ðk=2Þ2
Now, let us consider the inversion. Firstly, we apply the Fourier inversion
integral with respect to the parameter f defined by the second of Eq. (7.5.18). It is
2 Z1
a 1 exp½iffz logðb=aÞg
uðpÞ
zn ¼ B0 df ð7:5:22Þ
bcs0 2p ðf þ ik=2Þ2 þ n2 þ ðas=cs0 Þ2 þ ðk=2Þ2
1
If we make the change of variable from f to g as g ¼ f þ ik=2, the integrand in

Eq. (7.5.22) is simplified, but, its path of integration is the complex line from
1 þ ik=2 to þ1 þ ik=2, not a real path.
2
a
uðpÞ
zn ¼ B0 exp½ðk=2Þfz logðb=aÞg
bcs0
Z
1þik=2 ð7:5:23Þ
1 exp½igfz logðb=aÞg
dg
2p g2 þ n2 þ ðas=cs0 Þ2 þ ðk=2Þ2
1þik=2
In order to have a real line integral, we consider the complex integral whose
integrand is the same as that in Eq. (7.5.23). That is
Z
U¼ dg ð7:5:24Þ
L
The integration loop L is a rectangular form, as shown in Fig. 7.9. It is composed of

! ! ! !
two parallel horizontal lines AB and CD , and two finite vertical lines DA and BC .
The edges of the horizontal line tend to infinite, i.e. 1 þ ik=2. Examining the
singularity of the integrand, we find two poles at
g ¼ i n2 þ ðas=cs0 Þ2 þ ðk=2Þ2 ð7:5:25Þ
which are outside of the loop L and the integrand tends to zero at the infinity
g ! 1 þ ik=2. Then, we apply the Cauchy theorem to the complex integral U.
! !
Since the integrals along the two vertical lines DA and BC vanish and no singular
!
point is included in the loop, the integral along the upper horizontal line AB is
!
converted to that along the lower horizontal line DC , i.e. the real axis in the
g-plane. Thus we have the conversion formula for our integral as
Im(η )
+iqn ( s )
+iλ / 2
−∞ + iλ / 2 +∞ + iλ / 2
Re(η )
−iλ / 2
qn ( s ) = n 2 + (as / cs 0 ) 2 + (λ / 2) 2
−iqn ( s )
Fig. 7.9 Integration loop L for the complex integral U
Z
1þik=2
dg
1þik=2
Z1
¼ dg ð7:5:26Þ
1
The infinite integral in the right hand side of the above equation is reduced to the
semi-infinite integral
Z1 Z1
1 exp½igfz logðb=aÞg 1 cos½gfz logðb=aÞg
2 2
dg ¼ dg
2p g þ n þ ðas=cs0 Þ þ ðk=2Þ
2 2 p g2 þ n2 þ ðas=cs0 Þ2 þ ðk=2Þ2
1 0
ð7:5:27Þ
and the simple integration formula (Erdélyi 1954, vol. I, p. 8, (11))
Z1
1 p
cosðxyÞdx ¼ expðaj yjÞ ð7:5:28Þ
x2 þa 2 2a
0
is applied. Thus, we can evaluate our integral as
Z
1þik=2
1 exp½igfz logðb=aÞg 1
dg ¼ expfqn ðsÞjz logðb=aÞjg
2p g2 þ n2 2
þ ðas=cs0 Þ þ ðk=2Þ 2 2qn ðsÞ
1þik=2
ð7:5:29Þ
where
qn ðsÞ ¼ n2 þ ðas=cs0 Þ2 þ ðk=2Þ2 ð7:5:30Þ
Substituting Eq. (7.5.29) into Eq. (7.5.23), the double transformed particular
solution is given by

B0 a 2 1
uðpÞ
zn ¼ exp½ðk=2Þfz logðb=aÞg expfqn ðsÞjz logðb=aÞjg
2 bcs0 qn ðsÞ
ð7:5:31Þ
Summing the homogeneous and particular solutions, Eqs. (7.5.17) and (7.5.31), we
have the full solution for the differential equation (7.5.16). That is
uzn ¼ uzn
ðhÞ ðpÞ
þ uzn
¼ Cn ðsÞ exp½zfqn ðsÞ þ k=2g

B0 a 2 1
þ exp½ðk=2Þfz logðb=aÞg expfqn ðsÞjz logðb=aÞjg
2 bcs0 qn ðsÞ
ð7:5:32Þ
In order to determine the coefficient Cn ðsÞ, we apply the boundary condition,

Eq. (7.5.6). The double transformed boundary condition with the change of variable
is given by

uzn z¼0 ¼ 0 ð7:5:33Þ
Substituting Eq. (7.5.32) into the above equation, we have for the coefficient,

B0 a 2 1
Cn ðsÞ ¼ exp½fðk=2Þ qn ðsÞg logðb=aÞg ð7:5:34Þ
2 bcs0 qn ðsÞ
and thus the displacement in the double transformed domain is completely

determined,

B0 a 2
uzn ¼ exp½ðk=2Þfz logðb=aÞg
2 bcs0

1 1
expfqn ðsÞjz logðb=aÞjg exp½qn ðsÞfz þ logðb=aÞg
qn ðsÞ qn ðsÞ
ð7:5:35Þ
For the simplification in the subsequent analysis, we return the variable z to the
original radial distance r. The transformed solution is rewritten as

B0 a 2 b k=2 expfqn ðsÞjlogðr=bÞjg exp qn ðsÞ logðbr=a2 Þ
uzn ¼
2 bcs0 r qn ðsÞ qn ðsÞ
ð7:5:36Þ
We have just obtained the solution in the double transformed domain. Our
subsequent work is to invert the solution into the actual ðr; h; tÞ space. As the first
inversion, we consider its Laplace inversion. The Laplace inversion is symbolically
written as

2
B0 a 2 b k=2 1 expfqn ðsÞjlogðr=bÞjg 1 exp qn ðsÞ logðbr=a Þ
uzn ¼ L L
2 bcs0 r qn ðsÞ qn ðsÞ
ð7:5:37Þ
Fortunately, since qn ðsÞ is the simple square root function given by Eq. (7.5.30), we
can apply the inversion formula (Erdélyi 1954, vol, I, p. 248, (24)),

1
L1 pffiffiffiffiffiffiffiffiffiffiffiffiffiffi exp b s2 þ a2 ¼ Hðt bÞJ0 a t2 b2 ð7:5:38Þ
s2 þ a2
where H(.) is Heaviside’s unit step function and J0(.) is Bessel function of the
zeroth order. It follows that
k=2 qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiqffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
B0 a b
uzn ¼ H ð s j logðr=bÞ j ÞJ 0 n2 þ ðk=2Þ2 s2 log2 ðr=bÞ
2 b2 cs0 r
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiqffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi

H s logðbr=a2 Þ J0 n2 þ ðk=2Þ2 s2 log2 ðbr=a2 Þ
ð7:5:39Þ
where the dimensionless time, τ = cs0t/a, is introduced. The next is the inversion of the
Fourier finite transform, i.e. Fourier series. Applying the inversion formula in
Eqs. (7.5.14)–(7.5.39), we have for the displacement
k=2
B0 a b
uz ¼
2 b2 cs0 r
" qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiqffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
1 X þ1
H ðs jlogðr=bÞjÞ J0 n2 þ ðk=2Þ2 s2 log2 ðr=bÞ expðinhÞ
2p n¼1
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiqffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi #
1 Xþ1
H s logðbr=a2 Þ J0 n2 þ ðk=2Þ2 s2 log2 ðbr=a2 Þ expðinhÞ
2p n¼1
ð7:5:40Þ
Since both coefficients in the two series include only even power of the index n, the
Fourier series is reduced to the semi-infinite sum as
k=2
B0 a b
uz ¼
2p b2 cs0 r
" qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiqffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
X
1
H ðs jlogðr=bÞjÞ en J0 n2 þ ðk=2Þ2 s2 log2 ðr=bÞ cosðnhÞ
n¼0
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiqffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi #
X
1
2
H s logðbr=a Þ 2
en J0 n2 þ ðk=2Þ s2 log ðbr=a2 Þ cosðnhÞ 2
n¼0
ð7:5:41Þ
where

1=2; n¼0
en ¼ ð7:5:42Þ
1; n1
Formally we have just arrived at the exact solution for the scattering problem,
however, this solution does not show wave propagation phenomena clearly. In
order to obtain the solution which shows the wave phenomena, i.e. wave front
shape, we have to discuss the Fourier series. Let us consider the Fourier series
defined by
X
1 qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
S¼ J0 a n2 þ b2 cosðnhÞ ð7:5:43Þ
n¼0
where a and b are positive constants. We replace the Bessel function with its
integral form (Erdélyi 1954, vol. I, p. 28, (42)),

qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi Za cos b a2 n2
2
J 0 a n 2 þ b2 ¼ pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi cosðnnÞdn ð7:5:44Þ
p a2 n2
0
and change the order of the summation and integration. Then, expanding the
integration range, two summations are unified as

Za cos b a2 n2 X1
1
S¼ pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi dn en ½cosfnðh þ nÞg þ cosfnðh nÞg
p a2 n2 n¼0
0

pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi ð7:5:45Þ
Zþa cos b a2 n2 X1
1
¼ pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi dn en cosfnðn hÞg
p a2 n2 n¼0
a
The Fourier series in the integrand is just the form of Fourier expansion of an
infinite periodic array of the delta function (see Exercises (1.1) and (1.2)). That is
X
1 X
þ1
en cosfnðn hÞg ¼ p dðn h þ 2mpÞ ð7:5:46Þ
n¼0 m¼1
We substitute the above equation into the last line in Eq. (7.5.45) and change the
order of integration and summation. It follows that

þ1 Z cos b a2 n2
þa
X
S¼ pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi dðn h þ 2mpÞdn ð7:5:47Þ
m¼1 a2 n2
a
Examining the supporting region for the delta function, we can evaluate the integral
with use of the simple formula (1.2.3) in Chap. 1. It results
X
þ1 cos b a2 ðh 2mpÞ2
S¼ H ða jh 2mpjÞ qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi ð7:5:48Þ
m¼1 a2 ðh 2mpÞ2
The above equation looks like the sum of the infinite terms. But, due to the
operation of the step function, this summation is truncated to the finite sum. Thus,
we have just reduced the infinite sum of the Fourier series to the finite sum of
simple algebraic function. Further, if we expand the range of the circumferential/
polar angle h from the finite range ðp\h\ þ pÞ to the infinite range
ð1\h\ þ 1Þ and rewrite the angle
h h 2mp ð7:5:49Þ
The summation with the index m in Eq. (7.5.48) is reduced to the single term. That is

cos b a2 h2
S ¼ Hða hÞ pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi ; 1\h\ þ 1 ð7:5:50Þ
a2 h2
Consequently, we have obtained the very simple expression for the Fourier
series. That is the summation formula,

X
1 qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi cos b a2 h2
J0 a n2 þ b2 cosðnhÞ ¼ Hða hÞ pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi ; 1 \ h \ þ1
n¼0 a2 h2
ð7:5:51Þ
Substituting this formula into Eq. (7.5.41), we have the closed form for the
displacement
k=2
B0 a b
uz ¼
2p b2 cs0 r
2 qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
q ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi cos ðk=2Þ s2 log2 ðr=bÞ h2
6
6
4H ðs jlogðr=bÞjÞHð s log ðr=bÞ jhjÞ
2 2
s2 log2 ðr=bÞ h2
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi3
q ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi cos ðk=2Þ s2 log2 ðbr=a2 Þ h2
7
H s logðbr=a2 Þ Hð s2 log2 ðbr=a2 Þ jhjÞ qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi 7
5
s2 log2 ðbr=a2 Þ h 2
ð7:5:52Þ
The product of two step functions in the above equation can be unified and then we
have the final form for the displacement, i.e. the Green function for the scattering
problem,
k=2
B0 a b
uz ¼ 2
2p b cs0 r
2 qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi cos ðk=2Þ s2 log2 ðr=bÞ h2
6
6H s log2 ðr=bÞ þ h2 qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
4
s2 log2 ðr=bÞ h2
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi3
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi cos ðk=2Þ s2 log2 ðbr=a2 Þ h2 7
H s log2 ðbr=a2 Þ þ h2 qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi 7
5
s2 log2 ðbr=a2 Þ h2
ð7:5:53Þ
The first term in the bracket shows the direct wave from the source and its wave
front is given by
s¼ log2 ðr=bÞ þ h2 ) r=b ¼ exp s2 h2 ð7:5:54Þ
The second term shows the scattering wave and its front is given by
s¼ log2 ðbr=a2 Þ þ h2 ) r=a ¼ ða=bÞ exp s2 h2 ð7:5:55Þ
The both equations show the spiral curves in the ðr; hÞ-plane and the time-devel-
opment of these wave fronts can be found in the paper (Watanabe 1982).
7.6 An Excellent Application of Cauchy Complex Integral
The present section does not show any Green’s function, but a very excellent
application of the Cauchy complex integral is described. This application technique
is originated by Mal (1970), when he considered the wave scattering by a crack.
The dynamic crack problem is reduced to the Fredholm integral equation whose
kernel is in the form of semi-infinite integral with the product of two Bessel
functions,
2
Kðx; yÞ ¼ n ðx=cÞ n J0 ðxnÞJ0 ðynÞdn;
2
x\y ð7:6:1Þ
0
where J0 ð:Þ is the Bessel function of the first kind. As we have no suitable inte-
gration formula for this semi-infinite integral, we have to evaluate it numerically.
However, the integrand shows very slow convergence and its numerical evaluation
takes much CPU time for sufficient accuracy. If we could transform the semi-
infinite integral to finite integrals, we can evaluate the kernel without regard to the
convergence and accuracy. Mal (1970) has reduced the integral to the finite one
such as
ð2Þ
Kðx; yÞ ¼ þi ðx=cÞ2 n2 J0 ðxnÞH0 ðynÞdn; x\y ð7:6:2Þ
0
ð2Þ
where H0 ð:Þ is Hankel function of the second kind. After Mal, many researchers
have employed his evaluation technique for the dynamic crack and punch problems.
But, they (including Mal) did not describe its mathematics in detail. Thus, applying
Im(ξ )
branch cut integration path
O A B
−ω / c Re(ξ )
+ω / c
branch cut
Fig. 7.10 Integration path in the complex ξ-plane
our former discussion on the branch cut in Sect. 1.3.2 in Chap. 1, his technique is
explained in some depths and then one can learn the power of the Cauchy complex
integral.
As the kernel function of Eq. (7.6.1) is the same in the form, regardless of the
positive and negative time factors, we assume the positive time factor expðþixtÞ
and the radiation/convergence condition at infinity
Re n2 ðx=cÞ2 0 ð7:6:3Þ
Due to these assumptions, the branch cuts in the complex n-plane are introduced as
was discussed in Sect. 1.3.2 in Chap. 1. The cut and integration path for the semi-
infinite integral of the kernel are shown in Fig. 7.10, where the cut is the bold line

!
and the integration path is the semi-infinite line OAB. Since the integration path is
on the positive real axis, we decompose the integral into two parts: one is the finite
! !
integral on OA ð0 n\x=cÞ and the other on AB ðx=c\n\1Þ. As was discussed
!
in Sect. 1.3.2, the square root function on the path OA takes the positive imaginary
values as
n2 ðx=cÞ2 ¼ þi ðx=cÞ2 n2 ; 0 n\x=c ð7:6:4Þ
!
and that on the path AB does the positive real value as
n2 ðx=cÞ2 ¼ þ n2 ðx=cÞ2 ; x=c\n\1 ð7:6:5Þ
7.6 An Excellent Application of Cauchy Complex Integral 255
So, the kernel K(x, y) of Eq. (7.6.1) is decomposed into two integrals as
2
Kðx; yÞ ¼ þi ðx=cÞ n n J0 ðxnÞJ0 ðynÞdn
2
0
ð7:6:6Þ
þ n2 ðx=cÞ2 n J0 ðxnÞJ0 ðynÞdn
x=c
The first integral in the above equation is in the form of the finite integral, but the
!
second one on the path AB is still in that of the semi-infinite integral. We have to
transform the latter integral to the finite ones. This is the main purpose of the
present section.
Since the relation between Bessel and Hankel functions exists,
1 n ð1Þ ð2Þ
o
ð1=2Þ
J0 ðZÞ ¼ H0 ðZÞ þ H0 ðZÞ ; H0 ðZÞ ¼ J0 ðZÞ iY0 ðZÞ ð7:6:7Þ
2
the latter integral in Eq. (7.6.6) is separated into two semi-infinite integrals as
I¼ n2 ðx=cÞ2 n J0 ðxnÞJ0 ðynÞdn
x=c
1 2 ð1Þ
¼ n ðx=cÞ n J0 ðxnÞH0 ðynÞdn
2
ð7:6:8Þ
2
x=c
1 ð2Þ
þ n2 ðx=cÞ2 n J0 ðxnÞH0 ðynÞdn
2
x=c
We consider two complex integrals whose integrands are the same as those in the
above equation, respectively, and are defined by
I qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
1 ð1Þ
U1 ¼ f2 ðx=cÞ2 f J0 ðxfÞH0 ðyfÞdf ð7:6:9Þ
2
C1
I qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
1 ð2Þ
U2 ¼ f2 ðx=cÞ2 f J0 ðxfÞH0 ðyfÞdf ð7:6:10Þ
2
C2
We assume the real and imaginary parts of the complex variable f are n and g,
respectively, i.e. f ¼ n þ ig. Both integrands have two branch points at f ¼ x=c
for the square root function, and one branch point at the origin f ¼ 0 for the
η = Im(ζ )
E1
Loop Γ1
+ω / c
−ω / c F1
A1 B1 C1 D1 ξ = Re(ζ )
P O A2 B2 C2 D2
F2
Loop Γ 2
E2
Fig. 7.11 Integration contours Γ1 and Γ1 for the complex integrals Φ1 and Φ2
logarithmic nature of the Hankel function. In order to examine the convergence at

infinity, we apply the asymptotic expression of the Bessel function (Watson 1966,
pp. 198–199) to the product of Bessel and Hankel functions and have

ð1Þ 1
J0 ðxfÞH0 ðyfÞ pffiffiffiffiffi ½expfþiðy xÞfg i expfþiðy þ xÞfg ð7:6:11Þ
jfj!1 pf xy

ð2Þ 1
J0 ðxfÞH0 ðyfÞ pffiffiffiffiffi ½expfiðy xÞfg þ i expfiðy þ xÞfg ð7:6:12Þ
jfj!1 pf xy
We learn that the product of Bessel and Hankel functions shows the exponential
ð1Þ
decay at infinity. Since y > x, the product J0 ðxfÞH0 ðyfÞ converges in the upper
ð2Þ
ζ-plane, i.e. Im(ζ) > 0, and J0 ðxfÞH0 ðyfÞ does in the lower ζ-plane, i.e. Im(ζ) < 0.
Two branch cuts for the square root function are the same as those in Fig. 7.10, and
the branch cut for the Hankel function is introduced along the negative real axis
starting from the origin, i.e. the dotted line OP in Fig. 7.11. So, we take the closed
loop Γ1 in the first quadrant and Γ2 in the fourth quadrant as shown in the figure.
These two loops are composed of line paths and circular arcs: They are
! \ ! \ ! \
C1 :A1 B1 þ B1 C1 þ C1 D1 þ D1 E1 þ E1 F1 þ F1 A1 ð7:6:13Þ
! \ ! \ ! \
C2 :A2 B2 þ B2 C2 þ C2 D2 þ D2 E2 þ E2 F2 þ F2 A2 ð7:6:14Þ
The complex integral is decomposed into the integrals along these path segments.
0 1
Z Z Z Z Z Z Cqffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
1B
B C ð1Þ
U1 ¼ B þ þ þ þ þ C f2 ðx=cÞ2 f J0 ðxfÞH0 ðyfÞdf
2@ A
! \ ! \ ! \
A1 B1 B1 C1 C1 D1 D1 E1 E1 F1 F1 A1
ð7:6:15Þ
0 1
1B
Z Z Z Z Z Z Cqffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
B C ð2Þ
U2 ¼ B þ þ þ þ þ C f2 ðx=cÞ2 f J0 ðxfÞH0 ðyfÞdf
2@ A
! \ ! \ ! \
A 2 B2 B2 C2 C2 D2 D2 E2 E2 F2 F2 A2
ð7:6:16Þ
\ \
The radii of the quarter- and semi-circular paths, Bj Cj and Fj Aj ; j = 1, 2, are very
small and tend to zero as the limit. Thus, the integrals on these small circular paths
vanish when we take their zero limit of the radius. On the other hand, due to the
exponential decay of the product of Bessel and Hankel functions, the integral on the
\
quarter-circle Dj Ej ; j ¼ 1; 2 vanishes as the radius of the quarter circle tends to
infinity. Consequently, since no singular point such as the pole is included in each
loop, the complex integral around the closed loop vanishes, i.e. Uj ¼ 0, and thus the
!
integral on the line paths Cj Dj ; j ¼ 1; 2 is transformed to the sum of two integrals on
! !
the paths Aj Bj and Ej Fj ,
1 2 ð1Þ
f ðx=cÞ f J0 ðxfÞH0 ðyfÞdf
2
2
!
C1 D1
0 1
ð7:6:17Þ
Z Z Cqffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
1B
B C ð1Þ
¼ B þ C f2 ðx=cÞ2 f J0 ðxfÞH0 ðyfÞdf
2@ A
! !
A1 B1 E1 F1
Table 7.1 Variable transforms on the line path

Path Integration variable f Square root function Product of Bessel and Hankel
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi ð1=2Þ
and its range functions J 0 ðxfÞH0 ðyfÞ
f2 ðx=cÞ2
f¼n ð1Þ
J 0 ðxnÞH0 ðynÞ
A1 B1 þi ðx=cÞ2 n2
0\n\x=c
f¼n ð1Þ
J 0 ðxnÞH0 ðynÞ
C1 D1 n2 ðx=cÞ2
x=c\n\1
E1 F1 f ¼ þig; þ i ¼ eþpi=2 þi g2 þ ðx=cÞ2
2ip I0 ðxgÞK0 ðygÞ
0\g\1
f¼n ð2Þ
J 0 ðxnÞH0 ðynÞ
A2 B2 i ðx=cÞ2 n2
0\n\x=c
f¼n ð2Þ
J 0 ðxnÞH0 ðynÞ
C2 D2 n2 ðx=cÞ2
x=c\n\1
E2 F2 f ¼ ig; i ¼ epi=2 i g2 þ ðx=cÞ2
þ 2ip I0 ðxgÞK0 ðygÞ
0\g\1
1 2 ð2Þ
f ðx=cÞ f J0 ðxfÞH0 ðyfÞdf
2
2
!
C2 D2
0 1
ð7:6:18Þ
Z Z Cqffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
1B
B C 2 ð2Þ
¼ B þ C f ðx=cÞ f J0 ðxfÞH0 ðyfÞdf
2
2@ A
! !
A2 B2 E2 F2
The values of the square root function and integration variable on each line path are
tabulated in Table 7.1, where the product of Bessel and Hankel functions is derived
by using the formulas
ð1Þ 2i ð2Þ 2i
J0 ðixÞ ¼ I0 ðxÞ; H0 ðixÞ ¼ K0 ðxÞ; H0 ðixÞ ¼ þ K0 ðxÞ ð7:6:19Þ
p p
where I0(.) and K0(.) are the modified Bessel function of the first and second kinds.
Then, we can rewrite Eqs. (7.6.17) and (7.6.18). They are
1 ð1Þ
n2 ðx=cÞ2 n J0 ðxnÞH0 ðynÞdn
2
x=c
1 ð1Þ ð7:6:20Þ
¼ þi ðx=cÞ2 n2 n J0 ðxnÞH0 ðynÞdn
2
0
i
þ g2 þ ðx=cÞ2 g I0 ðxgÞK0 ðygÞdg
p
0
1 2 ð2Þ
n ðx=cÞ n J0 ðxnÞH0 ðynÞdn
2
2
x=c
1 2 ð2Þ ð7:6:21Þ
¼þ i ðx=cÞ n þ n J0 ðxnÞH0 ðynÞdn
2
2
0
i
g2 þ ðx=cÞ2 g I0 ðxgÞK0 ðygÞdg
p
0
Adding two Eqs. (7.6.20) and (7.6.20), we have two finite integrals for the semi-
infinite integral in Eq. (7.6.8) as
n2 ðx=cÞ2 n J0 ðxnÞJ0 ðynÞdn
x=c
1 ð1Þ
¼ 2
i ðx=cÞ n n J0 ðxnÞH0 ðynÞdn
2
ð7:6:22Þ
2
0
1 ð2Þ
þ i ðx=cÞ2 n2 þ n J0 ðxnÞH0 ðynÞdn
2
0
This equation is simplified by using the formula (7.6.7) for the Hankel function. It
yields
2
n ðx=cÞ n J0 ðxnÞJ0 ðynÞdn
2
x=c
ð7:6:23Þ
2
¼ nJ0 ðynÞ þ ðx=cÞ n Y0 ðynÞ J0 ðxnÞdn
2
The semi-infinite integral is just transformed to the finite integral and thus we can
replace the second integral in the right hand side of Eq. (7.6.6) with the above finite
integral. Therefore, the kernel function of the semi-infinite integral is reduced to the
finite one, i.e.
ð2Þ
Kðx; yÞ ¼ þi ðx=cÞ2 n2 J0 ðxnÞH0 ðynÞdn; y[x ð7:6:24Þ
0
This is the same as Eq. (7.6.2).

When we evaluate the kernel function, the change of variable from n to u,
n ¼ ðx=cÞu ð7:6:25Þ
is introduced and the numerical computation is carried out for the simplified
integral,
Z1 pffiffiffiffiffiffiffiffiffiffiffiffiffi
ð2Þ
Kðx; yÞ ¼ þiðx=cÞ 1 u2 J0 ðxxu=cÞH0 ðxyu=cÞdu; y[x ð7:6:26Þ
0
As we have the good subroutines for Bessel functions, the numerical integration for
the unit interval is a very easy task. This completes the application of the complex
integral.
Exercise
(7:1) Decompose the integrand in Sect. 7.1, as
!
n2 1 1 1
¼ 2
g2 þ p21 n 2
g2 þ p22 n2 p1 p22 g2 þ p21 n2 g2 þ p22 n2
! ðaÞ
g2 1 p21 p22
¼ þ
g2 þ p21 n 2
g2 þ p22 n2 p21 p22 g2 þ p21 n2 g2 þ p22 n2
and carry out the inversion integral (7.1.14) with the aid of the tabulated
integration formulas.
References
Erdélyi A (ed) (1954) Tables of integral transforms, vols I and II. McGraw-Hill, New York
Friedlander FG (1958) Sound pulses. Cambridge University Press, Cambridge
Mal AK (1970) Interaction of elastic waves with a Griffith crack. Int J Eng Sci 8:763–776
Pao Y-H, Mow C-C (1973) Diffraction of elastic waves and dynamic stress concentrations. Crane
Russak, New York
Watanabe K (1982) Scattering of SH-wave by a cylindrical discontinuity in an inhomogeneous
elastic medium. Bull JSME 25(205):1055–1060
Watson GN (1966) A treatise on the theory of Bessel functions, vol 2. Cambridge University
Press, Cambridge
Index
A C
Achenbach, J.D., 154 Cagniard, L., 154
Acoustic fluid, 126, 129 Cagniard’s
Acoustic wave, 121–123, 127, 129, 130 — path, 158, 167–169, 171, 174, 186, 187,
Adiabatic change, 122, 125 190–192, 201, 238
Amplitude function, 38, 39, 52, 63 — path I, 171, 172, 191
Anisotropic elastic solid, 205 — path II, 171–173, 191, 192
Anti-plane Cagniard-de Hoop technique, 154, 157, 161,
— deformation, 79, 154, 155 169, 185, 201, 202, 237
— displacement, 228, 243 Calculus, 8, 48
— shear load, 154 Cartesian coordinate
— system, 77
B Cauchy’s (integral) theorem, 149, 192, 216,
Beam 238
1D —, 152 Complex frequency, 40, 42, 44, 54, 55, 59, 142
Bending rigidity, 139, 145 Complex integral, 1, 29, 40, 42, 43, 54–56, 59,
Bessel function, 5, 29, 50, 59, 82, 83, 101, 112, 60, 65, 132, 133, 142–144, 149, 150,
133, 147, 152, 205, 219, 225, 249, 250, 154, 158, 159, 167, 169–173, 185–187,
253, 256, 258, 260 190, 192, 201, 205, 208, 216, 217, 219,
Body force 233–235, 237, 238, 246, 247, 253–257,
— potential, 127, 129 260
Boundary condition, 156, 164, 181, 203, 243, Continuity equation, 122–124, 127
245, 248 Convergence condition, 5, 11, 33, 35, 36, 38,
Branch cut, 1, 11–21, 23–27, 30, 32, 53, 54, 39, 45, 46, 48, 49, 51, 52, 54, 55, 68,
56, 59, 60, 62, 63, 65, 66, 132, 149, 70, 71, 73, 80, 90, 96, 110, 140, 142,
159, 167, 168, 170, 171, 173, 186, 145, 148, 155, 156, 179–181, 206, 214,
190–192, 220, 233, 237, 254, 256 223, 232, 254
Branch line, 174 Convolution integral, 64, 73, 74, 76, 88, 108,
Branch point, 11, 14–16, 20, 23–25, 53–55, 59, 136
132, 149, 159, 167, 170, 171, 173, 186, Critical velocity, 144
190, 192, 220, 233, 237, 255
Bromwich integral, 230, 232 D
Bromwich line, 2, 229, 230, 232, 234, 235 Deflection
Bumped curve, 234, 235 — divergence, 144

DOI 10.1007/978-3-319-17455-6
262 Index
Deflection (cont.) Fourier

dynamic —, 141, 148 — cosine series, 3
dynamic plate —, 148 — cosine transform, 5, 62
— equation, 139, 140, 142, 145, 148 — inversion, 4, 8, 33, 36, 39, 40, 42–44,
— response, 139 46–48, 50, 53, 58, 59, 62, 63, 65, 69,
De-Hoop, A.T., 154 71, 81, 82, 84, 85, 90–92, 94, 100, 101,
Delta function, 7, 8, 30, 34–36, 45, 73, 107, 103, 104, 111, 114, 131, 141, 142, 146,
113, 115, 116, 142, 214, 241, 242, 251 148, 149, 153, 156, 157, 161, 166, 170,
Denominator, 142, 144, 167, 171, 215, 237 174, 175, 189, 200, 208, 209, 211, 215,
Density, 77, 122, 123, 125, 127, 145, 155, 162, 226, 230, 237, 246
213, 242 — sine series, 3
Dilatational wave, 78, 86, 166–169, 171, 173, — sine transform, 4, 5
177, 184, 188, 190–192 Fourier transform
Dirac’s delta function, 6–8, 30, 74, 110, 139, complex —, 3
179, 223, 243 double —, 45, 49, 52, 79, 90, 130, 145,
Dispersive, 141 148, 180, 206, 214, 229
Displacement finite —, 3, 30
— equation, 78–80, 87, 89, 90, 96, 97, 107, triple —, 68, 71, 97
116, 156, 163, 164, 180, 181, 206, 207, Fredholm integral equation, 253
213, 214, 223, 224, 227, 231, 244 Fresnel integral, 141
Distribution, 6–8 Fundamental functions, 80, 86
Doppler effects, 121 Fung, Y.C., 154, 163
Double transform, 34, 52, 140, 224, 226, 245
Dyadic G
Green’s —, 77, 79, 86–90, 95, 105, 106, Gakenheimer, D.C., 197
108, 116, 177, 202, 205, 206, 212, 213, Gradshteyn, I.S., 28, 29, 58, 62, 91, 94
220 Graff, K.F., 154
Green’s dyadic, 77, 79, 86–90, 95, 105, 106,
E 108, 116, 177, 202, 205, 206, 212–220
Eigenvalues, 142, 207, 215 Green’s function
Elastic 2D —, 63, 74
— beam, 139 3D —, 68, 70, 73, 74, 154
— moduli, 206, 213
— modulus, 207 H
— solid, 77, 79, 109, 154, 162, 163, 178, Hankel function, 58, 59, 61, 63, 65, 88, 89,
179, 205, 213, 222, 228, 242–244 136, 149, 150, 152, 253, 255–259
Elasticity equations, 77, 78 Hankel transform, 5, 29, 224
Elastodynamic Heaviside’s (unit), 6, 37, 86, 113, 161, 179,
— equations, 154 249
— problem, 154, 155, 180, 228, 242 Heisenberg’s delta function, 6, 8, 10, 203
3D —, 180 Helmholtz equation, 52
Equation of motion, 155, 180, 222, 223, 243, Hooke’s law, 77, 78, 155, 164, 213, 243
244 Hydro-static pressure, 122
Equation of state, 122, 123, 125 Hyperbola
Equilibrium equation semi —, 15, 16, 158, 168
2D —, 213 Hyperbolic curves
Erd’elyi, A., 9, 28, 29, 37, 50, 51, 58, 70, 72, semi —, 15, 16, 158
83–85, 104, 112, 115, 140, 141, 147,
210, 211, 219, 225, 226, 247, 250 I
Ewing, W.M., 11 Ichimatsu, 29
Impulsive
F — body force, 79, 96, 110
Fixed condition, 243 — point load, 139, 145, 162
Fixed coordinate system, 232 — (point) source, 40
Index 263
— response, 88, 136, 145 O

2D response, 49 Orthotropic, 205, 212
Incident wave, 228, 229, 237, 242 Out-going wave, 44, 63
Inclusion, 205, 242, 243
Incoming wave, 44 P
Inhomogeneity, 213, 221, 242 Pao, Y.-H., 242
Inhomogeneous (isotropic elastic) solid, 205, Parametric integral, 160, 235
213 Particle velocity, 122
Inhomogeneous media, 221 Plane strain, 162
Initial condition, 145 Plate, 145, 148
In-plane Poisson ratio, 77, 78, 109, 213
— deformation, 79, 90, 162, 205 Poles, 40, 218, 257
— displacement, 79 Press, F., 11
Isotropic, 178, 212 Pulse, 7
J Q
Jardetzky, W.S., 11 Quiescent condition, 33, 35, 70, 96, 140, 179,
Jordan’s lemma, 42, 43, 208, 216 228, 245
K R
Kelvin’s solution Radiation condition, 157, 159, 163, 228, 245
2D —, 95, 221 Radiation wave, 40, 44
Kernel, 1, 34, 253, 260 Rayleigh
Kronecker’s delta, 86 — equation/function, 167
— poles, 186
L — wave, 167
Lamb’s problem, 163, 178 Reference
Lame’s constants, 77, 163, 213 — length, 129, 213
Laplace — state, 122, 123, 126, 130
— equation, 44, 48, 68, 70 Reflected wave, 228, 231, 232, 237, 241
— inversion, 2, 153 Reflection, 121, 205, 227
— transform, 2, 5 Refraction, 121
Laplacian Residue, 42, 142, 238, 239
— operator, 128 Reynolds number, 129
Logarithmic singularity, 149, 212 Riemann sheets, 13, 14
Rigid motion, 95
M Ring source, 110, 227
Mach numbers, 129, 231 Ryzhik, I.M., 28
Magnus, 29
Mal, A.K., 253 S
Miklowitz, J., 154 Saddle point, 159, 168, 173, 190, 192, 238
Monoclinic, 222, 227 Scattering, 205, 242
Moriguchi, 29 SH-wave, 161, 184, 202, 227, 228, 243
Moving Shear rigidity, 157
— boundary, 205, 227 Shear wave, 86, 96, 155, 165, 173, 242
— coordinate, 231, 232 Sign function, 47, 85, 212
— edge, 232 Sine integrals, 147
Mow, C.-C., 242 Specific heats, 122
Static hydro-pressure, 123
N Step functions, 1, 6, 8, 10, 74, 135, 177, 252
Navier-Stokes equations, 122 Strain, 77, 78, 95, 211
Navier equations, 78 Stress, 77, 122, 181, 211, 230, 243
Newtonian fluid, 121 Stress-free, 228, 232
264 Index
Subsonic, 132 V
SV-wave, 184, 189, 197 Velocity
— gradients, 127
T — potential, 127, 128
Time-harmonic — ratio, 78, 109
— function, 88 Viscosity, 122, 126
— Green’s function, 33 Viscous fluid, 121, 127
— (point) load, 148 von Schmidt wave, 177, 178
— response, 44
— solution, 89 W
— source, 38, 51, 73, 87 Water waves, 121
— vibration, 18, 38 Watson, G.N., 29
— wave, 135, 151 Wave equation
Titchmarsh, E.C., 29 2D —, 49, 51
Torque 3D —, 70, 98
point —, 109, 110 Wave fronts, 107, 135, 197, 198, 241, 253
— source, 113 Wave nature, 40, 141, 197
Torsion problem, 109 Wave reflection, 205, 227
Transient response Wave source, 33, 79, 129, 243, 245
2D —, 162
Triple integral transform, 49, 80, 111, 145, 229 Y
Young’s moduli, 205
U Young’s modulus, 77
Udagawa, K., 29
Uniform flow, 129, 135

Integral Transform Techniques For Green's Function: Kazumi Watanabe

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Integral Transform Techniques For Green's Function: Kazumi Watanabe

Uploaded by

Copyright:

Available Formats

Lecture Notes in Applied and Computational Mechanics 76

Peter Wriggers, Universität Hannover, Hannover, Germany

This series aims to report new developments in applied and computational

More information about this series at http://www.springer.com/series/4623

ISSN 1613-7736 ISSN 1860-0816 (electronic)

Library of Congress Control Number: 2013940095

Springer Cham Heidelberg New York Dordrecht London

Printed on acid-free paper

Springer International Publishing AG Switzerland is part of Springer Science+Business Media

Hikoshima Island, Japan, January 2015 Kazumi Watanabe

Yonezawa, Japan, January 2013 Kazumi Watanabe

1 Deﬁnition of Integral Transforms and Distributions . . . . . . . . . . . 1

2 Green’s Functions for Laplace and Wave Equations . . . . . . . . . . 33

3 Green’s Dyadic for an Isotropic Elastic Solid . . . . . . . . . . . . . . . . 77

3.7 Torsional Source. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 109

4 Acoustic Wave in a Uniform Flow . . . . . . . . . . . . . . . . . . . . . . . . 121

5 Green’s Functions for Beam and Plate. . . . . . . . . . . . . . . . . . . . . 139

6 Cagniard-de Hoop Technique . . . . . . . . . . . . . . . . . . . . . . . . . . . 153

7 Miscellaneous Green’s Functions . . . . . . . . . . . . . . . . . . . . . . . .. 205

1.1 Integral Transforms

© Springer International Publishing Switzerland 2015 1

where the integration path from c i1 to c þ i1 is called the “Bromwich line.”

f ðtÞ ¼ L1 ½f ðsÞ ð1:1:6Þ

(1) Finite Fourier transform (Fourier series): f ðxÞ; x 2 ½p; p

where the index “n” is an integer.

(1.3) Fourier sine series

(2) Complex Fourier transform: f ðxÞ; x 2 ð1; 1Þ

(2.1) Fourier transform with a non-uniform parameter.

where hðnÞ is a monotonically increasing/decreasing function, and the lower and

g ¼ hðnÞ , n ¼ h1 ðgÞ ð1:1:12Þ

n ¼ h1 ðgÞ ) g ¼ hðnÞ; dg ¼ h0 ðnÞdn ð1:1:15Þ

Then, Eq. (1.1.14) is rewritten as

(3.1) Fourier cosine transform

(3.2) Fourier sine transform

(4) Hankel transform: f ðxÞ; x 2 ½0; 1Þ

f ðxÞ ) Oðjxjm Þ; m [ 1 ð1:1:20Þ

for the original function. However, if we employ the Hankel transform, it is

1.2 Distributions and Their Integration Formulas

For modeling engineering phenomena, many mathematical functions are used.

Hð0þ Þ ¼ þ1; Hð0 Þ ¼ 0 ð1:2:2Þ

Then, we have to understand that the step function is not deﬁned at x ¼ a.

Fig. 1.1 Heaviside’s unit

(2) Dirac’s delta function: dðx aÞ

Using this simple integration formula, we can obtain an integral representation

Its formal Fourier transform is

hðnÞ ¼ pdðnÞ þ h1 ðnÞ ð1:2:10Þ

We remember the integration formula that gives 1=2 in each semi-inﬁnite

h1 ðnÞ ¼ i ð1:2:15Þ

hðnÞ ¼ pdðnÞ þ i ð1:2:18Þ

Then, Heisenberg’s delta function has two deﬁnitions as

As a byproduct, we can obtain a new formula. Two operations, addition and

1.3 Branch Cut and Argument of Square Root Functions

on the root function.

Since the conditional is equivalent to

Since Y 2 \ 0; x must be smaller than a,

If the argument h is deﬁned within p h þ p, the argument of the root function

0 < arg z − z0 < π / 2

0 > arg z − z0 > −π / 2

upper Riemann sheet

lower Riemann sheet

This square rootpfunction appears frequently in many dynamic problems. The

x4 ða2 b2 Þx2 a2 b2 ¼ ðx2 a2 Þðx2 þ b2 Þ 0 ð1:3:17Þ

Solving for x2 , we have

We remember the integration formula that gives 1=2 in each semi-inﬁnite

This square rootpfunction appears frequently in many dynamic problems. The

Zþ1 x!þ1 Zþ1