BAB 6

Persamaan Gerak: Kontinua Homogen Isotropik

Dari studi tentang alam, muncullah kelompok persamaan diferensial parsial yang saat ini paling
banyak diselidiki secara menyeluruh dan mungkin paling penting dalam struktur umum
pengetahuan manusia, yaitu persamaan fisika matematika.
Sergei L. Sobolev dan Olga A. Ladyzenskaya (1969) Persamaan diferensial parsial dalam
Matematika (editor: Alek-sandrov dkk.)
Preliminary Remarks
Having formulated system (4.4.5)—a system of equations to describe the behaviour of an elastic
continuum—we wish to write Cauchy’s equations of motion explicitly in the context of the stress-
strain equations for such a continuum. This way, we commence our study of wave phenomena in
an elastic continuum.
We begin by choosing the simplest type of elastic continuum, namely an isotropic homogeneous
one, and, hence, we derive the corresponding equations of motion, which lead to the wave
equations. In the process of formulating these equations, we learn about the existence of the two
types of waves that can propagate in isotropic continua. Furthermore, we obtain the expressions for
the speed of these waves as functions of the properties of the continuum.
We begin this chapter by combining Cauchy’s equations of motion (2.8.1) with constitutive
equations (5.12.5). This formulation results in the derivation of the wave equations. To gain insight into
these equations, we study them in the context of plane waves and displacement potentials. We also
investigate the solutions of the wave equations, including solutions in various spatial dimensions and
nondifferentiable solutions. We conclude this chapter with examples of extensions of the standard
form of the wave equation that take into account aspects of anisotropy and of inhomogeneity.
6.1. Wave Equations
6.1.1. Equation of motion
To derive the wave equation, assume that a given three-dimensional contin- uum is isotropic and
homogeneous. Thus, we consider the corresponding stress-strain equations given by expression
(5.12.5), namely,

where λ and µ are constants. We also consider Cauchy’s equations of motion (2.8.1), namely,
e wish to combine stress-strain equations (6.1.1) with equations of motion (6.1.2) to get the
equations of motion in an isotropic homogeneous continuum. In other words, we substitute
expression (6.1.1) into equa- tions (6.1.2) to obtain

Now, we wish to express the right-hand side of equations (6.1.3) in terms of the displacement vector,
u . Invoking the definition of the strain tensor, given in expression (1.4.6), we can rewrite equations
(6.1.3) as

Using the property of Kronecker’s delta, we obtain

Using the linearity of the differential operators, we can rewrite these equa- tions as

where, in the first summation, for the summation indices, we let k = j . Using the equality of
mixed partial derivatives, we obtain

We can use vector calculus to concisely state equations (6.1.4). Con- sider the right-hand side of these
equations. The first summation term is the divergence of u , namely, ∇ u , while the second summation
term is the Laplacian, namely, ∇2 . Consequently, we can rewrite equations (6.1.4) as

Noticing that the first matrix on the right-hand side involves the gradi-
ent operator, we can concisely state the three equations shown in expres- sion (6.1.5) as

This is the equation of motion that applies to isotropic homogeneous con- tinua.
To proceed from expression (6.1.5) to expression (6.1.6), we assume that u is expressed in
Cartesian coordinates, which allow us to use the equality given by (∇2u)i = ∇2ui . In general, the
ith component of ∇2u is not equal to the Laplacian of the ith component of u .1 However— even
though, for convenience, we use Cartesian coordinates in the above derivation—equation (6.1.6),
above, identity (6.1.7) and equation (6.1.8), below, are valid in curvilinear coordinates, provided
that a proper form of the vector Laplacian is used.
We wish to write equation (6.1.6) in a form that allows us to express it in terms of the dilatation and the
rotation vector, in accordance with their definitions stated in Chapter 1. Using the vector identity given
by
 and letting a = u , we can rewrite equation (6.1.6) as

Equation (6.1.8) contains information about the deformations expressed in terms of the divergence
and the curl operators. Recalling the definitions of the dilatation and the rotation vector, given by
expressions (1.4.18) and respectively, we can immediately write

Equation (6.1.9) describes the propagation of the deformations in terms of both dilatation and the
rotation vector in an isotropic homogeneous con- tinuum. It describes the propagation related to both
the change in volume and the change in shape. The divergence operator is associated with the
change of volume while the curl operator is associated with the change in shape.
6.1.2. Wave equation for P waves
To gain insight into the types of waves that propagate in an isotropic homo- geneous continuum, we
wish to split equation (6.1.9) into its parts, which are associated with the dilatation and with the

rotation vector.2
To obtain the wave equation for P waves, we take the divergence of equation (6.1.9). Since in a
homogeneous continuum λ and µ are con- stants, we can write

The factor of µ disappears since for all Ψ . Considering the factor of λ + 2µ and invoking the
definition of the Laplacian, we can write

Consequently, equation (6.1.10) becomes

Let us consider the left-hand side of equation (6.1.11). In a homoge- neous continuum, the
mass density, ρ , is a constant. Hence—in view on

the linearity of the differential operators we can take ρ outside of the di- vergence. Also—in view of the
equality of mixed partial derivatives we can interchange time and space derivatives. Thus, we
get

where, on the left-hand side, we used again definition (1.4.18). Rearranging, we obtain
 which is the wave equation whose wave function is given by dilatation,

Equation (6.1.12) is the wave equation for P waves. As shown in Sec- tion 6.5

is the propagation speed. In view of Section 5.12, the presence of both Lamé’s parameters in
expression (6.1.13) suggests that P waves subject the continuum to both a change in volume and a
change in shape.
In view of definition (1.4.18), P waves are sometimes referred to as di- latational waves. Also, since
the dilatation, ϕ , is the relative change in volume, they are sometimes referred to as pressure waves.
Furthermore, since the speed of P waves is always greater than the speed of S waves, which are
discussed below, in earthquake observations, P waves are some- times referred to as primary
waves.
6.1.3. Wave equation for S waves
To obtain the wave equation for S waves, we take the curl of equation (6.1.9) and write

The factor of λ + 2µ disappears since ∇ ∇ϕ = 0 for all ϕ . Recalling definition (1.5.2) and
considering the constancy of the mass density, ρ — in view of the linearity of the differential
operators as well as the equality

of mixed partial derivatives—we get

Invoking vector-calculus identity (6.1.7) and letting a = Ψ , we can write equation (6.1.15) as

In view of definition (1.5.2) and the vanishing of the divergence of a curl, the first term in
brackets disappears. Hence, we obtain

which is the wave equation whose wave function is given by the rotation vector,
Equation (6.1.16) is the wave equation wave for S waves. As shown in Section 6.5,
 is the propagation speed. In view of Section 5.12, the presence of the single
Lamé’s parameter, namely, µ , in expression (6.1.17), suggests that S waves subject the continuum to a
change in shape. Also, due to the vanishing of rigidity in fluids, we can conclude that the propagation of S
waves is limited to solids.
In view of definition (1.5.2), S waves are sometimes referred to as ro- tational waves. Since the
rotation vector is given by Ψ = ∇ u , we con- clude that ∇ Ψ = 0 . If the divergence of a vector field
vanishes, this vector field is volume-preserving; hence, S waves are sometimes referred to as the
equivoluminal waves. In English, the justification for the letter S is due to the fact that these waves
are often referred to as shear waves. Also, due to the fact that the speed of S waves is always
smaller than the speed of P waves, in earthquake observations, S waves are sometimes referred to
as secondary waves.
6.1.4. Physical interpretation
Examining equations (6.1.12) and (6.1.16) in view of definitions (1.4.18) and (1.5.2), respectively,
we see that both of them contain the three compo- nents of the displacement vector, u . However, in
view of definition (1.4.6), we see that equation (6.1.12) contains only partial derivatives given by
∂ui/∂xi . In view of definition (1.5.1), we see that equation (6.1.16) con- tains only ∂ui/∂xj , for which
i /= j . Recalling expression (1.5.3), we write

This Jacobian matrix can be viewed as a second-rank tensor associated with the deformation
gradient. Equation (6.1.12) contains only the main- diagonal entries, equation (6.1.16) only the
offdiagonal ones. Let us in- terpret the role of the terms that do not appear explicitly in a given
wave equation; in other words, the role of the main-diagonal entries for equa- tion (6.1.16) and of
the offdiagonal for equation (6.1.12).
In equation (6.1.12), ∂ui/∂xi , which constitute ϕ , describe the restor- ing force that results in
wave propagation. Terms ∂u i /∂x j with i = j , which are not part of this equation, prevent any local
rotation during prop- agation; in other words, they ensure that ∇ u = 0 .
In equation (6.1.16), the terms that constitute Ψ = ∇ u describe the restoring force, and the other
ones prevent any change in volume; in other words, they ensure that ∇ u = 0 .
To justify these statements, let us consider Figure 1.4.1. After deforma- tion, the horizontal edge
can be written as vector
and the vertical edge as

Examining expression (6.1.18), we see that the horizontal edge is rotated counterclockwise in the
X1X2-plane by

where we assume ∂u1/∂X1 1 . Similarly, by expression (6.1.19), it fol- lows that the vertical edge is
rotated clockwise by ∂u1/∂X2 .
Let us recall Figure 1.4.2. If ∇ u = [∂u2/∂X1 ∂u1/∂X2 , 0 , 0 ] = 0 , the two angles are equal to
one another, which implies that the diagonal of the deformed rectangle is aligned with the diagonal
of the undeformed rectangle. Thus, the deformed rectangle is not rotated. In general, ∇ × u = 0
means that the deformation does not result in rotation.
Let us illustrate the meaning of ∇ u = 0 . The area of the deformed rectangle is the magnitude of
the cross product of vectors (6.1.18) and (6.1.19), which is

which is the only nonzero component of the cross product. Considering only the first-order terms
in partial derivatives, we rewrite this area as

If ∂u1/∂X1 + ∂u2/∂X2 = 0 , the above expression reduces to ∆X1∆X2 , which implies that there is no
change in area. In general, ∇ u = 0 means that the deformation does not result in change of
volume.
6.2. Plane Waves
In general, equations (6.1.4) are complicated partial differential equations. This shows that even in
isotropic homogeneous continua, the description of wave phenomena constitutes a serious
mathematical problem. We can simplify these equations by introducing certain abstract mathematical
enti- ties that allow us to describe particular aspects of wave phenomena. While studying wave
propagation in homogeneous media, we can consider plane waves. These are the waves for which the
components of the displacement vector are functions of the direction of propagation only.
To gain insight into the concept of plane waves, let us revisit equa- tions (6.1.12) and (6.1.16). Let
the plane waves propagate along the x1- axis. Thus, in view of the properties of plane waves and
following expres- sion (2.4.3), we write the displacement vector as
Since all the partial derivatives of u with respect to x2 and x3 vanish, equa- tions (6.1.4) become

After algebraic manipulations, we can write

Taking the derivative of equation (6.2.1) with respect to x1 , we obtain

Using expression (6.2.4) in equation (6.2.5), we obtain

which is a plane-wave form of equation (6.1.12). Examining equations (6.2.4) and (6.2.6), we
recognize that the displacement and the direction of propa- gation are parallel to one another, which
is the key property of P waves in isotropic continua. This property is also shown in Exercise 9.4.
Now, consider equations (6.2.2) and (6.2.3). Recall expression (1.5.2), which in this case
becomes

Taking the derivative of equations (6.2.3) and (6.2.2) with respect to x1 , and writing them as a
vector, we obtain

Using the equality of mixed partial derivatives and expression (6.2.7), we obtain

which is a plane-wave form of equation (6.1.16). Examining equations (6.2.7) and (6.2.8), we
recognize that the displacements and the direction of prop- agation are orthogonal to each other,
which is the key property of S waves in isotropic continua. This property is also shown in
Exercise 9.6.
Plane waves are an approximation that allows us to study, in homo- geneous media, a wavefield
that results from a distant source. Notably, in Chapter 10, we use plane waves to study reflection
and transmission of waves at an interface separating two anisotropic homogeneous half-spaces. For
close sources, we can construct a wavefield as a superposition of plane waves. In such an approach,
there is a constructive interference in the re- gions where the plane waves coincide and a destructive
interference outside of these regions.
While studying inhomogeneous media, the behaviour of seismic waves cannot be conveniently
described using plane waves and their superposi- tion. For such studies, we introduce in Section
6.11.4 another abstract mathematical entity—a seismic ray, which belongs to the realm of asymp- totic
 methods and provides us with a different perspective to study seismic wavefields.

6.2. Displacement Potentials

6.2.1. Helmholtz’s decomposition
In Sections 6.1.2 and 6.1.3, we derived the wave equations for P and S waves, respectively. To derive
the wave equation for P waves, which we ex- pressed in terms of scalar function ϕ , we took the
divergence of Cauchy’s equations of motion. To derive the wave equation for S waves, which we
expressed in terms of vector function Ψ , we took the curl of these equa- tions. Herein, we obtain
equations that correspond to P and S waves by using Helmholtz’s method of separating a vector
function into its scalar and vector potentials. We obtain these equations by inserting the potentials into
Cauchy’s equations of motion.
According to Helmholtz’s theorem3, a differentiable function u (x,t) can be decomposed into

where P and S = [S1, S2, S3] are called the scalar and vector potentials, respectively. Following the
definitions of the gradient and curl operators, we can explicitly write the components of u as

which constitute a system of differential equations. This system does not have a unique solution. It
is common to consider also another equation; namely,
∇· S (x,t) = 0 . (6.3.2)
since we can always find P and S that satisfy the system composed of equations (6.3.1) and (6.3.2).4
Introducing equation (6.3.2) does not result in a unique determination of S . It only reduces possible
choices of this vector, as shown below.
We use expressions (6.3.1) and (6.3.2) in Sections 6.3.3 and 6.3.4. In the next section, we
justify our introducing equation (6.3.2).
6.2.2. Gauge transformation
Let us consider equation (6.3.2) in the context of equation (6.3.1). We are allowed to set ∇ S = 0 since,
in view of properties of the vector operators, S used in equation (6.3.1) is determined up to a
gradient, ∇ f , where f (x) is any differentiable function. In mathematical physics, changing S by
adding ∇ f to it is called a gauge transformation. Let

S˜ = S + ∇ f . (6.3.3)
Taking the curl of both sides of equation (6.3.3), using the linearity of the differential operator and
 the vanishing of the curl of a gradient, we obtain

∇ × S˜ = ∇× (S + ∇ f ) = ∇ × S .
Examining this result in the context of expression (6.3.1), we see that the same u is obtained using

either S or S + ∇ f . We can use this freedom of choice to set ∇ · S˜ = 0 . This is tantamount to

finding f such that ∇2 f =−∇ · S . To reach this conclusion, we took the divergence of both sides
of equation (6.3.3) to get

∇ · S˜ = ∇· (S + ∇ f (x)) = ∇ · S + ∇2 f = 0 . (6.3.4)

Since both S˜ and S result in the same u , we have justified our adding equation (6.3.2) to
system (6.3.1).5
In particular, it is interesting to notice by examining equation (6.3.4), that if f is a harmonic

function—in other words, if f is a solution of ∇2 f = 0 —then, ∇ · S˜ = ∇ · S . Consequently, in

view of equation (6.3.3), we can always add to S the gradient of a harmonic function.
6.2.3. Equation of motion
To study the equations of motion in terms of the displacement potentials, we insert expression
(6.3.1) into equation (6.1.6) and write

∂
2 (∇P + ∇×
S) · + S)]+ µ∇2 + ∇× S) .
ρ = (λ + µ) ∇[∇
(∇P ∇× (∇P

∂t 2
Using the vanishing of the divergence of a curl and the definition of the Laplacian, we obtain

Using the linearity of the differential operators and the fact that in a homo- geneous continuum ρ , λ and
µ are constants, as well as using the equality of mixed partial derivatives, we can rewrite this
equation as

Rearranging, we obtain

which is the equation of motion for isotropic homogeneous continua in terms of the scalar and
vector potentials.

6.2.4. P and S waves

Looking at equation (6.3.5), we see that it is satisfied if

and

which, in view of equations (6.1.12) and (6.1.16), appear to be associated with P and S waves,
respectively. A rigorous analysis of this result is as- sociated with Lamé’s theorem.6 Equations
(6.3.6) and (6.3.7) are wave equations whose wave functions are the scalar and vector potentials,
re- spectively. Motivated by this observation, we wish to study the relation of the scalar and vector
potentials to the two wave equations whose wave functions are given by the dilatation and the
rotation vector; namely equa- tions (6.1.12) and (6.1.16), respectively.
As in Section 6.1.2, let us take the divergence of equation (6.3.5). Using the vanishing of the
divergence of a curl and the definition of the Laplacian, we obtain

Using the linearity of the differential operator and the equality of mixed partial derivatives,
we can rewrite equation (6.3.8) as

To relate the scalar potential, P , to the dilatation, ϕ , let us take the diver- gence of expression
(6.3.1). Using the vanishing of the divergence of a curl and recalling definition (1.4.18) as well as
the definition of the Laplacian, we obtain
ϕ := ∇· u = ∇· ∇P ≡ ∇2P . (6.3.10)
In other words, the dilatation is equal to the Laplacian of the scalar poten- tial. Using expression
(6.3.10), we can rewrite equation (6.3.9) as

which is equation (6.1.12), as expected. Thus, we conclude that the Lapla- cian of the scalar
potential, P , satisfies the wave equation for P waves.
As in Section 6.1.3, let us take the curl of equation (6.3.5). Using the vanishing of the curl of a
gradient, we get

Recalling identity (6.1.7) and letting a denote the term in parentheses, we can rewrite this
equation as

Using the linearity of the differential operators and the equality of mixed partial derivatives, we
 can rewrite this equation as

In view of equation (6.3.2), ∇ · S = 0 , equation (6.3.12) becomes

Again using the linearity of the differential operator and the equality of mixed partial
derivatives, we can write equation (6.3.13) as

To relate the vector potential, S , to the rotation vector, Ψ , let us take the curl of expression (6.3.1).
Using the vanishing of the curl of a gradient and recalling definition (1.5.2), we obtain
Ψ := ∇× u = ∇ × ∇ × S .
Following identity (6.1.7) and letting a = S , we get
Ψ = ∇ × ( ∇ × S) = ∇( ∇ · S) − ∇2S .
In view of equation (6.3.2), ∇ · S = 0 , we obtain
Ψ = −∇2S . (6.3.15)
In other words, the rotation vector is equal to the negative Laplacian of the vector potential. Using
expression (6.3.15), we can rewrite equation (6.3.14) as

which is equation (6.1.16), as expected. Thus, we conclude that under

condition (6.3.2) the Laplacian of the vector potential, S , satisfies the wave equation for S waves.
This derivation of equations (6.3.11) and (6.3.16) is analogous to the method for obtaining
Maxwell’s equations in the electromagnetic theory using the vector and scalar potentials. There are,
however, several distinc- tions between the origins of potentials for the wave equations in elasticity and
electromagnetism.7

6.3. P and S Waves in Terms of Displacements

6.3.1. Introductory comments
8 Examining expression (6.1.8), namely,

and the formulations of wave equations in Sections 6.1 and 6.3, which result in equations (6.3.11)
and (6.3.16), we see that expression (6.4.1) is an equation of motion stated in terms of the
displacement vector, u , and expressions (6.3.11) and (6.3.16) are wave equations stated in terms of
 derivatives of u . In this section, we formulate equations for the P and S waves in terms of the
displacement vectors themselves, and we show that the irrotational and equivoluminal
displacements imposed by derivatives of u are the sufficient and necessary conditions for the
existence of the P and S waves, respectively.
6.3.2. Sufficient conditions for P and S waves
P waves
If the displacement is irrotational, which means that
∇× u = 0 , (6.4.2)
equation (6.4.1) becomes

In such a case, identity (6.1.7) —where we let a ≡ u —becomes

∇2u = ∇(∇· u) . (6.4.4)

Hence, equation (6.4.3) becomes

which is analogous to expression (6.3.11): the equation for P waves, but stated in terms of u , not
∇ u . Hence, under condition (6.4.2), equa- tion (6.4.1) implies equation (6.4.5). In other words, given
equation (6.4.1), expression (6.4.2) is a sufficient condition for equation (6.4.5).

S waves
If the displacement is equivoluminal, which means that
∇· u = 0 , (6.4.6)
equation (6.4.1) becomes

In such a case, identity (6.1.7) becomes

∇2u = − ∇ × ( ∇ × u) .
(6.4.8)
Hence, equation (6.4.3) becomes
(6.3.16): the not ∇ u . Hence, equation (6.4.9). In other words, given
which is equation for S under condition equation (6.4.1), expression (6.4.6) is a
analogous ×
to waves, but stated (6.4.6), equa- tion sufficient condition for equation
expression in terms of u , (6.4.1) implies (6.4.9).
6.3.3. Necessary (6.4.2) and equation (6.4.9) to the one presented in Section 6.1.2, we
conditions for P (6.4.6), we have implies condition take the divergence of equation (6.4.1),
and S waves to show that (6.4.6). which, due to ∇ · ∇ × u = 0 ,
To assert the equation (6.4.5) Following an
necessity of implies condition approach analogous becomes
conditions (6.4.2) and
we
Using using identity (6.4.4) and the rewrite · u = 0 .
linearity of differential operators, it as
Examining this equation, we see that either ∇ · u = 0 or the result of the differential operator in
parentheses acting on ∇ · u is zero. Denoting ∇ · u by ϕ and denoting the operator in parentheses by

QP , we express these alternatives as

To present our argument for the necessary condition, we invoke equa- tion (6.4.9), without
considering its derivation. Taking the divergence of this equation and using the linearity of

differential operators together with the equality of mixed partial derivatives, we obtain QS ϕ = 0 , where

QS is the differential operator in expression (6.4.12), below. However, if ϕ is a general solution for

QS ϕ = 0 , it cannot be a general for QP ϕ = 0 , since these operators differ by a constant factor in front
of the temporal deriva- tive. Notably, in view of expression (6.5.6), below, we see that the general
solutions are different from one another; in a single spatial dimension, they are f (x µ ρ t) + g(x + µ ρ
t) and f (x (λ + 2µ) ρ t) + g(x
(λ 2µ)/ρ t) , respectively, and, as shown in Exercise 5.18, for any isotropic Hookean solid, µ > 0 and
λ > 2µ/3 , which means that the two waves have different propagation speeds. Thus, we must
choose the first alternative, ϕ = 0 , which does not exhibit any inconsistency with equa- tion (6.4.9).
This means that equation (6.4.9) implies condition (6.4.6):
∇ u = 0 , and, hence, it is a necessary condition.
Following an approach analogous to the one presented in Section 6.1.3, we take the curl of equation
(6.4.1), which, due to ∇ ∇(∇ u) = 0 , becomes

Using identity (6.4.8), we rewrite equation (6.4.11) as

 Examining this equation, we see that either ∇ u = 0 or the result of the operator in parentheses acting
on ∇ u is zero. We express these alterna- tives as

Taking the curl of equation (6.4.5), we obtain QPΨ = 0 . This result dis- agrees with the second of
alternatives (6.4.13); thus, we must choose the first alternative, which is consistent with equation
(6.4.5). This means that equation (6.4.5) implies condition (6.4.2): it is a necessary condition.
To complete this section, using definitions (5.12.2), let us rewrite equa- tions (6.4.5) and (6.4.9) as

respectively, which is consistent with the notation in Chapter 7, where we discuss equations of
motion in anisotropic inhomogeneous continua; equa- tions (6.4.14) and (6.4.15) are special cases
of such equations.
6.5. Solutions of Wave Equation for Single Spatial Dimension
6.5.1. d’Alembert’s approach
To gain further insights into the physical meaning of equations (6.1.12) and (6.1.16), we study the
solution of their generic form, where we do not specify if the wave function corresponds to P waves
or to S waves.

Consider the initial-value problem given by

where u = u (x,t) is the wave function and v is a constant. Let the initial conditions be stated by

The following method of solving the wave equation was introduced in 1746 by d’Alembert10 and
further elaborated upon by Euler, with important contributions from Daniel Bernoulli and Lagrange.11
It is based on the following two lemmas, which allow us to obtain the solution by integration.
LEMMA 6.5.1. Equation

The form ∂ 2u (y, z) /∂y∂z = 0 is a normal form of the hyperbolic dif- ferential equation, where
y and z are referred to as the natural coordinates; informally speaking, we have factorized the
differential to obtain a solution by integration, as shown below. For y and z given by constants, y = x +
vt and z = x vt are straight lines in the xt-plane, which are known as the characteristics of the wave
equation. Expressions y (x,t) and z (x,t) satisfy the characteristic equation of wave equation (6.5.1),
as shown in Exercise 6.19.12
LEMMA 6.5.2. For equation

the only form of the solution is

u (y, z) = f (y) + g (z) , (6.5.5)
where f and g are twice-differentiable arbitrary functions.
Details of the derivation of Lemma 6.5.2 are shown in Exercise 6.2.
Combining Lemma 6.5.1 and Lemma 6.5.2, we can state the following corollary.
COROLLARY 6.5.3. Following Lemma 6.5.1 and Lemma 6.5.2, and using coordinates (6.5.4), we can
write the only form of the solution of equa- tion (6.5.1) as
u (x,t) = f (x + vt) + g (x − vt) , (6.5.6)
where f and g are arbitrary twice-differentiable functions.
Solution (6.5.6) allows arbitrary twice-differentiable functions f and g . Further constraints must
be imposed on functions f and g if we wish to obtain a particular solution.
Herein, we wish to obtain a particular form of solution (6.5.6) that sat- isfies the constraints
provided by initial conditions (6.5.2). Inserting ex- pression (6.5.6) into system of equations
(6.5.2), we can write
( γ (x)
f (x) + g (x) =
v f J (x) − vgJ (x) = η (x) , (6.5.7)
where we used the chain rule with f J and gJ denoting the derivatives with respect to arguments (x + vt)
and (x vt) , respectively, and evaluated the results at t = 0 . This system of equations can be solved
explicitly for f (x) and g (x) . Integrating both sides of the second equation of this system, we obtain

while subtracting the second equation from the first one gives us
Inserting expressions (6.5.9) and (6.5.10) into solution (6.5.6), we write

Using the fact that reversing the limits of integration changes the sign of the integral, we obtain

which is the solution of the initial-value problem given by equations (6.5.1) and (6.5.2).
Let us interpret the physical meaning of solution (6.5.11). If we view x as the position variable
and t as the time variable, solution (6.5.11) de- scribes propagation of u in the one-dimensional x-
space. Solution u (x,t) is completely determined by the differential equation and the initial condi- tions,
which describe the solution at the initial time, u (x, 0) = γ (x) , and the velocity of displacement of u at
that instant, η (x); in other words, at point x , γ displaces with velocity η .
Herein, the one-dimensional space is homogeneous and infinite. To examine inhomogeneous,
finite and semifinite cases, we need to consider conditions at the boundary, as exemplified in
Section 6.8, below.13
To examine the concept of propagation, let us consider solution (6.5.6) with g = 0; namely,

We wish to examine the propagation of a given point that belongs to f . Such a point corresponds to
a particular value of f . In view of expres- sion (6.5.12), we see that a particular value of u (x,t)
remains the same if the value of x + vt stays the same. According to conditions (6.5.2), at time t =
0 , we have u (x, 0) = γ (x) . Let us consider location x0 , at that time. At this location and at that time,
we have γ (x0) . Let us follow this value of γ . At time t1 , we have γ (x1 + vt1) . To follow the same value,
we require the constancy of the argument; in other words, x0 = x1 + vt1 . Solving for x1 , we get x1 =
x0 − vt1 , which means that the value of γ that was at x0 moved to x0 vt1; it moved left along the x-axis
by distance vt1 . We conclude that constant v in equation (6.5.1) is the propagation speed.
Returning to solution (6.5.6) and following an argument analogous to the one presented above but
with f = 0 , we conclude that f and g move in opposite directions. In general, f and g are different
from one an- other; they are explicitly stated in terms of the two initial conditions by expressions
(6.5.9) and (6.5.10). Examining these expressions, we note that the two functions propagating in the
opposite directions are the same if η (x) = 0; in other words, if there is no initial velocity of
displacement. In such a case, the amplitude of displacement at any location and instant is the
arithmetic mean of the two identical waves travelling in the opposite directions.14
The initial-value problem given by equations (6.5.1) and (6.5.2) does not contain any source; in
other words, there is no action of external force. We could incorporate such a force in the initial-value
problem by adding a term to equation (6.5.1) to write it as

where f is a function representing the source.15

If the solution exhibits a spherical symmetry—which means that the ini- tial conditions, γ(x) and

η(x) , are spherically symmetric, since the wave equation itself, v2∇2u = ∂ 2u/∂t2 , is symmetric—
d’Alembert’s approach can be extended to higher spatial dimensions, as illustrated in Exercises 6.3 and
 6.4. In such a case, the general solution is
1 1
u (r,t) = f (r + vt) + g (r − vt) ,

where the first term on the right-hand side represents a spherically sym- metric wave that propagates
towards the origin and the second one away from it; r is the distance from the origin, it is the magnitude

of r , and it is tantamount to x2 + x2 + x2 , in Cartesian coordinates.

In this manner, the extension discussed in Exercises 6.3 and 6.4 remains formally a problem of a single
spatial dimension. The amplitude of the wave propagating towards the origin increases with distance,
as the wave- front approaches the origin; the amplitude of the wave propagating away decreases.
The change of amplitude is proportional to r−1 for both two and three dimensions, which is a
consequence, respectively, of a wavefront in two spatial dimensions represented by a circle, which is a
one-dimensional entity, and in three dimensions by a two-dimensional sphere. Note that the form of
the Laplacian for a function that is independent of the angle is the same for both polar and spherical
coordinates, as illustrated in Exercise 6.3.
6.5.2. Directional derivative
To gain further insight into wave equation (6.5.1) and solution (6.5.6), let us rewrite this equation
using directional derivatives. We write

where the term in parentheses is a differential operator, which we can rewrite as a composition
of two differential operators; namely,16

Using the scalar product, we can write each operator as

where the terms in parentheses have the form of directional derivatives with directions [v, 1] and [v, 1]
. Equation (6.5.15) means that u is constant in these directions. In other words, u is constant along the
lines whose slopes are

Solving this ordinary differential equation, we get

x = ±vt +C,
where C is the integration constant. Solving for C , we write
C = x ∓ vt.
 Since a function of a constant is a constant, the solution of equation (6.5.15) can be written as
u (x,t) = f (x + vt) + g (x − vt) ,
which is solution (6.5.6); it states that f and g are constant along
In view of equation (6.5.15), we can formulate the solution in a dif- ferent way. Function u
remains unchanged along directions [v, 1] and [v, 1] , which correspond to lines that are described by
their perpendiculars; namely, [1, v] and [1, v] , respectively. Hence, the arguments of f and g are
and respectively. [x,t] · [1, v] = x + vt
[x,t] · [1, −v] = x − vt,
6.5.3. Well-posed problem
We are interested in knowing whether or not the solution of the initial- value problem given by
equations (6.5.1) and (6.5.2) is unique. Also, since this solution results from the initial conditions,
we wish to know whether or not it depends smoothly on the initial data. In other words, we wish
to verify that the dependence is such that a small change in input affects the solution by a small
amount only. We refer to such a solution as a stable solution. A problem consisting of equations
that result in a unique and stable solution is called a well-posed problem. Wave equation (6.5.1)
together with conditions (6.5.2) constitute a well-posed problem as we see below.
It is important to note that many mathematical physics questions do not constitute well-posed
problems; they are ill-posed problems. However, this classical nomenclature does not imply that a
well-posed problem is physically more realistic than an ill-posed problem. For instance, inverse
problems, which are of great interest in seismology, often do not possess unique solutions.
Let us demonstrate the uniqueness of solution (6.5.11). We demon- strate it explicitly by
following the construction of our solution derived in Section 6.5.1.
Consider equation (6.5.1), namely, Following Lemma 6.5.1 and Lemma 6.5.2, we obtained
Corollary 6.5.3, according to which
u (x,t) = f (x + vt) + g (x − vt) (6.5.17)
is the only form of the solution of equation (6.5.16). Using initial condi- tions (6.5.2), we obtain
system (6.5.8), which allows us to uniquely solve for f and g in terms of γ and η . Consequently, we
can write solution (6.5.17) in terms of γ and η to obtain expression (6.5.11). Thus, in view of the
con- struction of solution (6.5.11), we conclude that the solution is unique.
Similarly, examining the construction that led to solution (6.5.11), as described in Section 6.5.1,
we see that γ and η are linear combinations of the continuous and differentiable functions f and g
and of their first derivatives. Hence, functions γ and η , which are the initial conditions, are also
continuous. Since the solution depends linearly on γ and on the integral of η , we conclude that a
small change in γ or η results in a small change in u . Hence the solution, u (x,t) , is stable.
We can also show the uniqueness of the solution of the wave equa- tion by studying the energy of
wave function u (x,t) at a given instant in time. Herein, by analogy to the above section, we discuss
the energy of a wave function in a single spatial dimension, where u = u (x,t) . The presented method,
however, can be easily extended to higher dimensions, thereby showing that, in general, the wave
equation with appropriate ini- tial conditions is a well-posed problem. Furthermore, we could also
define the wave-function energy for equations in dissipative and dispersive me- dia; this aspect of
wave phenomena, however, is beyond the scope of this book.17
Considering equation (6.5.1), let us define the wave-function energy to be

In the study of partial differential equations, it is common to define energy as an integral whose
integrand is composed of squares of first derivatives of a function, as we did in expression (6.5.18)
for function u .
To interpret the physical meaning of this definition, we first note that, as written, the physical
units of E are the product of velocity squared and distance, which are not units of energy.
However, if we multiply E by unit mass per unit length, [kg/m] , the units of E are the product
of velocity squared and mass, which are the units of energy. Also, we can write E as the sum of
two integrals, namely, (∂u/∂t)2 /2 dx and
v2 (∂u/∂x)2 /2 dx . We can view the former integral as corresponding
to the kinetic energy of displacement. Since v2/2 is a constant, we can view the latter integral as
corresponding to the potential energy; in particular, it corresponds to the the strain energy that is
associated with deformation u .
Differentiating expression (6.5.18) with respect to time and using the fact that limits of
integration are fixed, we get

To study this integral, we rewrite it as two integrals, namely,

Solving the second integral by parts, we ge

We assume that η , which appears in condition (6.5.2), has compact sup- port. This means that η (x) is
nonzero on a closed and bounded interval, but is zero everywhere else. Thus, ∂u/∂t vanishes at
both positive and negative infinity of the space variable, x . In other words, there is no dis- placement
velocity infinitely far from the neighbourhood of x = 0 . For the integrated term, (∂u/∂t)(∂u/∂x)|
−∞ , to vanish, we must also assume ∂u/∂x to be finite at both positive and negative infinity of x . If
we as- sume that function γ has compact support, ∂u/∂x = 0 at x = ∞ since the propagation speed, v ,
is finite. This implies that, at an infinite distance, u (x,t) = 0 , for all t < ∞ . Using these
assumptions, we can write

Inserting this expression into expression (6.5.19), we get

Examining the integrand in view of wave equation (6.5.1), we see that the term in brackets
vanishes. Hence, we obtain

which implies that E is constant.

Equation (6.5.20) states that for wave equation (6.5.1), the wave-function energy defined by
expression (6.5.18) is conserved.18 We can state it as the following theorem.
THEOREM 6.5.4. If u (x,t) is a solution of wave equation (6.5.1) together with initial conditions
(6.5.2) that are given by functions with compact sup- port, then energy defined by expression (6.5.18)
is conserved. In other words, E (t) = E (0) for all t .
In Exercise 6.5, we prove an analogous theorem for a boundary-value, rather than an initial-value,
problem.
We can use Theorem 6.5.4 to show that the solution of the problem given by equations
(6.5.1) and (6.5.2) is unique.19 Since equation (6.5.1).
is linear, a solution can be composed of a difference of two arbitrary so- lutions. Each arbitrary
solution must obey the initial conditions. Exam- ining the first of conditions (6.5.2), we see that—at
t = 0 —each solution is γ (x) . This means that—at t = 0 —the solution composed of a differ- ence of
two solutions is zero. If the difference of two arbitrary solutions is zero, these two solutions are
equal to one another; in other words, the solution is unique at t = 0 . Now, using Theorem 6.5.4, we
wish to verify this uniqueness for t > 0 .
Following definition (6.5.18) and using the fact that, at t = 0 , the so-
lution composed of a difference of two solutions is u (x,t) = 0 , we see that the corresponding E (0)
= 0 . Then, Theorem 6.5.4 states that E (t) = E (0) = 0 , for all t . Invoking definition (6.5.18), we
 can explicitly write

where t1 is an arbitrary time, t1 (0, ∞) . Since (∂u/∂t)2 , v2 and (∂u/∂x)2

are positive for all t and x , for the integral to vanish we require that

for any x and for arbitrary t . This implies that

where we used the fact that v is a finite velocity of propagation. Hence, we can say that solution u is
constant. Since u is unique at t = 0 while being constant for all t (0, ∞) , it must be unique for all t .
Thus we can conclude with the following corollary of Theorem 6.5.4.
COROLLARY 6.5.5. Solution u (x,t) of wave equation (6.5.1) together with conditions (6.5.2) is
unique.
The uniqueness of the solution of a hyperbolic partial differential equa- tion, such as the wave
equation, is linked intimately to its initial conditions. This subject is studied within the realm of
Cauchy’s problems. 20
6.5.4. Causality, propagation speed, sharpness of signal
Let us continue to examine solution (6.5.11), namel.

For given (x,t) , the value of solution u depends on the values of γ at points x vt and on the values of
η on interval [x vt, x + vt] . Hence, this interval is the domain of dependence of u (x,t) , as illustrated
in Figure 6.5.1. To examine the effect of a point source, at x0 , on a location, say x1 , let us consider
this figure. Point x0 belongs to the domains of dependence of points (x,t) that lie in a triangular
region whose apex is at x0 and whose sides have the slopes of ∓1/v . This region is the range of
influence of x0 on solutions u (x,t) . The signal generated at x0 at t = 0 does not reach x1 until time t1 .
Prior to that instant, the source has no effect at x1 . Thus, we conclude that the process is causal
and the propagation speed of the signal is finite; its magnitude is (x1 − x0) /t1 . There is another
important consequence of the range of influence. At x1 , the effect of the signal is observed not only
at t1 but also afterwards. This means that, in general, a sharp signal generated at x0 does not
propagate as a sharp signal—its effect persists after t1 . It is interesting to note that for the acoustic
case, where u represents pressure, the constant value of u is not audible since our hearing relies on
the change in pressure. Thus, we would hear only the initial arrival of the signal—a sharp-signal
effect. For the elastic case, where u represents displacement, the constant value is observable.
 The domain of dependence and the range of influence can be viewed as the time reverses of one
another. To consider the domain of dependence is to ask about the origin of the phenomena observed
at a given instant and

Fig. 6.5.1 The domain of dependence and the range of influence for the wave equation in a single spatial
dimension. Consider the initial-value problem given in expressions (6.5.1) and (6.5.2), and its solution
(6.5.21) at point (x,t) . The point of γ(x) located initially at x + vt reaches xs , from the right, at time ts ; the
point located at x vt reaches the same location from the left. The value of the solution, u , that
corresponds to xs and ts is the sum of the values of γ at the points initially at x + vt and x − vt , and of
the integral of η whose limits are x − vt and x + vt . Also, at time t1 , the source at x0 influences the
values of the solution between x1J and x1 . The point of γ(x) located initially at x0 reaches x + vt and x
vt at time t1 . The limits of the integral that contributes to the solution are x vt and x + vt . Furthermore, we
can conclude that the speed of propagation of a signal is finite, since before time t1 elapses, point x1 is
not affected by the source. The slopes of the sides of the triangle are ∓1/v .
location. To consider the range of influence is to ask about the regions influenced by an event.21
As stated above, in general, a sharp signal generated at a point does not propagate as a sharp
signal. It does so in particular cases, however. By setting the initial displacement velocity, η , to
zero,
we rewrite solu- tion (6.5.21) as a common example of η = 0 is the case of a taut string that is pulled,
and then released. Herein, the value of the solution depends on points x vt only. Hence, if there is no
initial displacement velocity, the effect of the signal does not persist after the signal passes;
consequently, sharp signals can propagate.
 In Section 6.6, we see that the solution of the wave equation changes with the spatial dimension
of the problem being considered; unlike in the cases of one and two dimensions, three-dimensional
media allow for the propagation of sharp signals. Such a change of behaviour is a particu- lar
property of the wave equation; solutions of the heat equation and the steady-state equation,
commonly known as Laplace’s equation, do not ex- hibit such changes in physical interpretation
due to dimensions.22
6.6. Solution of Wave Equation for Two and Three Spatial Dimensions
6.6.1. Introductory comments
Having obtained the solution of the wave equation in one spatial dimension in Section 6.5, we wish to
investigate the solutions of the wave equation in two and three spatial dimensions. Let us consider

where ∇2 is the Laplacian, and the corresponding initial conditions; namely,

and
September 5, 2015 22:19 World Scientific Book - 9in x 6in BookOneSecSec
September 5, 2015 22:19 World Scientific Book - 9in x 6in BookOneSecSec