2 Theory-Of-Seismic-Waves

Theory of seismic waves
Overview fronts and raypaths are introduced, as is the more gen-

The seismic method utilizes the propagation of waves eral Huygens' principle approach.
through the earth. To introduce the basic concepts of The two forms of the wave equation that had been
wave motion, we first discuss waves on a stretched derived earlier are related to two types of disturbances
string (§2.1.1) and introduce definitions of phase, fre- that can travel through the body of solids (§2.4). These
quency, wavelength, and other terms dealing with pe- involve changes in volume (P-waves) and rotations
riodicity. Because wave propagation depends upon the (S-waves). Discussion of potential functions, from
elastic properties of the rocks, we next discuss some which particle displacements and velocities can be de-
of the basic concepts of elasticity. (For more thorough rived, follows. At interfaces, both stresses and particle
treatments, see Saada, 1974, or Landau and Lifshitz, displacements must be continuous; these boundary
1986.) conditions are discussed in §2.4.4.
The size and shape of a solid body can be changed Surface waves are examined next. Rayleigh waves
by applying forces to the external surface of the body. are important because of the ground-roll noise that
These external forces are opposed by internal forces, they produce on seismic records. Love, Stoneley, and
which resist the changes in size and shape. As a result, tube waves are encountered occasionally.
the body tends to return to its original condition when Most seismic theory assumes that media are iso-
the external forces are removed. Similarly, a fluid re- tropic, that is, their properties are the same regardless
sists changes in size (volume) but not changes in of the direction of measurement. Anisotropy (§2.6) of
shape. This property of resisting changes in size or several types has been observed; however, anisotropic
shape and of returning to the undeformed condition effects are usually small. The most important excep-
when the external forces are removed is called elastic- tions requiring study are those of transverse isotropy
ity. A perfectly elastic body is one that recovers com- because of layering and fracturing.
pletely after being deformed. Many substances includ- Section 2.7 examines what happens to seismic body
ing rocks can be considered perfectly elastic without waves as they travel in the earth. Intensity decreases
appreciable error provided the deformations are because of geometrical spreading (divergence) and ab-
small, as they are in seismic surveys. sorption (and partitioning at interfaces; see chap. 3).
The theory of elasticity relates the forces that are Divergence is the most important factor affecting the
applied to the external surface of a body to the re- change of intensity for the first few kilometers, but
sulting changes in size and shape. The relations be- eventually absorption becomes dominant. Absorption
tween the applied forces and the deformations are increases approximately linearly with frequency and
most conveniently expressed in terms of the concepts hence changes the waveshape with distance. Various
of stress and strain. Strain, a change in shape or di- expressions for absorption are interrelated. Disper-
mensions, is generally proportional to the stress (force sion and the concepts of group and phase velocity are
per unit area) that produces it, as stated in Hooke's discussed, although dispersion is not an important
law. The constant of proportionality is called an elas- factor in seismic exploration.
tic constant, or modulus, and moduli for different Reflection and refraction are discussed in §2.7.5.
types of stress and strain are interrelated. Diffraction (§2.8), the scattering of waves at disconti-
Section 2.2 concerns seismic-wave motion. New- nuities, involves somewhat complex mathematics.
ton's second law of motion, that an unbalanced force However, the construction of diffraction wavefronts
on a mass produces an acceleration, is used to derive using Huygens' principle is fairly straightforward and
two forms of the wave equation. The wave equation is nonmathematical.
expressed in vector as well as the more conventional
scalar notation. Methods of including a source of dis- 2.1 Theory of elasticity
turbance and KirchhofFs theorem are also given in
this section. 2.1.1 Waves on a stretched string
Plane- and spherical-wave solutions to the wave As an introduction to seismic waves in three dimen-
equation are given next. Waves are disturbances that sions, we consider the one-dimensional wave in a
travel through the medium. The concepts of wave- stretched string because many basic concepts of wave
33
Downloaded from https:/www.cambridge.org/core. University of Arizona, on 07 Apr 2017 at 02:36:39, subject to the Cambridge Core terms of use, available at
https:/www.cambridge.org/core/terms. https://doi.org/10.1017/CBO9781139168359.003
34 THEORY OF SEISMIC WAVES
motion can be more simply illustrated in this way.

Parts of the following discussion will be treated later
in a broader context. y A
We assume an ideal case where the mass of the
string per unit length, |x, is negligibly small in compar-
ison with the tension, T, in the string, that the string
when at rest is along the x-axis, and that the displace-
ments, i|/, which are parallel to the .y-axis, are small in
comparison with the length of the string so that angles
aY and a2 are also small (fig. 2.1a). Because these
angles are not equal, the tension produces a net force
in the j-direction (the net force in the x-direction is
negligible) on an element of the string, Ax, equal to
j(sin a2 - sin c^) ~ T(tan a2 - tan a}) « j(d\\f/dx\x - (a)
d\\s/dx\Xi) « T A(di|;/dx). Newton's second law of mo- crest or peak
tion states that this force equals the product of the
mass |JL Ax and the acceleration 32\\t/dt2. Dividing both
sides by T0 Ax and taking the limit as Ax -> 0 gives the
one-dimensional wave equation:
d 2 i|//dx 2 = (2.1)
1/2
where V = (T/|X) (compare with eq. (2.45)). Equation trough
(2.1) shows that Fhas dimensions of distance/time, or Frequency
velocity. The wave equation relates variation in space
(the left side) with variation in time (the right side).
The general solution of eq. (2.1) (also called
d'Alembert's solution; see §2.2.5) is
i|i(s, 0 = ill, (x - Vi) + v|/2 (x + Vt\ (2.2)
where \\fY and \\t2 are arbitrary functions, \\JY is a distur-
bance moving in the positive x-direction with increase
of time, i|i2 a disturbance moving in the negative x-
direction, and V is the velocity of propagation along
the string (see the following). Wovenumber =
Fourier analysis (§9.1.2) shows that any waveform
(within reason) can be represented by a superposition
of harmonic (sinusoidal) waves, so we do not lose gen-
erality by confining our attention to harmonic waves.
Thus, we consider a harmonic solution of eq. (2.1) in
the form
= A cos [(2ir/X)(x - Vi)]. (2.3)
The waveform is harmonic with i|i varying between
+A and —A; A is the amplitude. If we look at the wave
passing afixedpoint in space (fig. 2.1b), period Tis the
time between successive repetitions of the waveform;
frequency v = \IT is the number of waves per unit
time. If we look at the waveform at some moment of (d)
time (fig. 2.1c) the distance between successive repeti-
tions of the waveform is the wavelength X and 1/X is Fig. 2.1 Waves on a stretched string, (a) Portion of the string
the wavenumber or number of waves per unit distance. showing the relation between displacement and tension; (b) rep-
Multiplying 1/rand 1/X by 2ir, we get the angular fre- resentation of the wave in time; (c) representation of the wave
in space; (d) the effect of change in mass/unit length.
quency a) = ITTIT = 2irv and the angular wavenumber
K = 2TT/X. Because v is the number of waves passing a
fixed point per unit time and each wave has length X,
velocity V must be given by the equation (2TT/X) (X - Vi) = K(X - Vi) = (KX - a>f), is called the
V = vX. (2.4) phase. In eq. (2.3), the phase is zero at the origin; at
times we add a fixed phase angle 70 so that the phase
The argument of the cosine in eq. (2.3), namely, becomes KX — at + y0.
THEORY OF E L A S T I C I T Y 35
Returning to the stretched string, if the mass/unit can vibrate in a number of patterns called modes or
length changes abruptly from \xl to JLJL2 at some point, eigenstates, the frequencies being harmonics (multi-
say, x = 0 (fig. 2.Id), certain boundary conditions ples) of the fundamental, that is, vt = nv0, n = 1, 2, 3,
(§2.4.4) must hold, namely, both the displacement and . . . . In each case, the ends of the string are nodes and
the ^-component of the tension in the string must be L = n\ll = (2n)\/4.
continuous, that is, neither changes in value as we go If the left end of the string is fixed and the right end
through the junction. These conditions can be ex- free, we set K2 = 0 and get R = -1. The end of the
pressed by the equations string is an antinode, L = A74, v0 = VIAL, and the
harmonics are v = (2n + l)v0 and L = (2n + 1)AV4.
=
^left bright* The two cases of a string fixed at one end only and
(2.5)
fixed at both ends are analogous to organ pipes closed
We take the incident wave as Ai cos (KXX - u>t) coming at one end only and closed at both ends (Logan, 1987;
from the left and the wave passing on to the right (the see also §13.3).
transmitted wave) as At cos (K2X - co/); however, we
cannot satisfy eqs. (2.5) with these two waves only and
we must postulate a reflected wave going to the left, 2.1.2 Stress
Ar cos (KXX + a)/). Substituting into eqs. (2.5), we find Stress is defined as force per unit area. Thus, when a
that the boundary conditions will be satisfied provided force is applied to a body, the stress is the ratio of the
force to the area on which the force is applied. If the
K K K (2.6a) force varies from point to point, the stress also varies,
lA l^r lAr J
and its value at any point is found by taking an infini-
Equations (2.6a) can be solved for Ar and A;. tesimally small element of area centered at the point
and dividing the total force acting on this area by the
R = ArIAt =(K2 - KI)/(K2 + Kj),
(2 6b) magnitude of the area. If the force is perpendicular to
- the area, the stress is said to be a normal stress (or
where R and T are called the reflection coefficient (or pressure). In this book, positive values correspond to
reflectivity) and the transmission coefficient, respec- tensile stresses (the opposite convention of signs is
tively (see also §3.2). sometimes used). When the force is tangential to the
If the string is fixed at x = 0, the effect is the same element of area, the stress is a shearing stress. When
as if |JL2 = oo; then T = 0, so no wave is transmitted, the force is neither parallel nor perpendicular to the
and R = + 1, which means that the reflected wave is element of area, it can be resolved into components
the same as the incident one except that the direction parallel and perpendicular to the element; hence, any
of travel is reversed. The two waves interfere (§2.3.2) stress can be resolved into component normal and
at the fixed end to produce perfect cancellation, shearing stresses.
hence, zero movement (node). If both ends are fixed, If we consider a small element of volume inside a
perfect cancellation must occur at both ends, so these stressed body, the stresses acting upon each of the six
are nodes. faces of the element can be resolved into components,
When a string fixed at both ends is vibrating at its as shown in fig. 2.2 for the two faces perpendicular to
lowest frequency, called the fundamental (v0), the dis- the x-axis. Subscripts denote the x-, y-, and z-axes,
placement has its maximum amplitude at the mid- respectively, and ayx denotes a stress parallel to the y-
point (antinode). The wave pattern is fixed, so the axis acting upon a surface perpendicular to the x-axis.
wave is said to be stationary, or standing. If the string When the two subscripts are the same (as with axx\
length is L, L = X/2 and v0 = Vl\ = VI2L. The string the stress is a normal stress; when the subscripts are
different (as with ayx), the stress is a shearing stress.
When the medium is in static equilibrium, the
stresses must be balanced. This means that the three
stresses, avx, <jyx, and crzx, acting on face OABC must
be equal and opposite to the corresponding stresses
shown on opposite face DEFG, with similar relations
for the remaining four faces. In addition, a pair of
shearing stresses, such as vyx, constitute a couple tend-
ing to rotate the element about the z-axis, the magni-
tude of the couple being
force X lever arm = (<jyx dy dz) dx
If we consider the stresses on the other four faces,
we find that this couple is opposed solely by the
Fig. 2.2 Components of stress on faces perpendicular to the couple due to the pair of stresses axy with magnitude
x-axis. (<rxy dx dz) dy. Because the element is in equilibrium,
the total moment must be zero; hence vxy = vyx. In

general, we must have
*„ = o> (2.7)
2.1.3 Strain
T
When an elastic body is subjected to stresses, changes
in shape and dimensions occur. These changes, which
are called strains, can be resolved into certain funda-
mental types.
Consider rectangle PQRS in the .xy-plane (see fig.
2.3). When the stresses are applied, let P move to P',
PP' having components u and v. If the other vertices
Q, R, and S have the same displacement as P, the rect-
angle is merely displaced as a whole by the amounts u
and v; in this case, there is no change in size or shape,
and no strain exists. However, if u and v are different
for the different vertices, the rectangle will undergo
changes in size and shape, and strains will exist. Fig. 2.3 Analysis of two-dimensional strain.
Let us assume that u = u(x, y) and v = v(x, y). Then
the coordinates of the vertices of PQRS and P'Q'R'S' the xy-plane is reduced when the stresses are applied,
are as follows: hence, is a measure of the change in shape of the me-
P(x, y)\ P'(x + u,y + v); dium; it is known as a shearing strain and will be
Q(x + dx, y): denoted by the symbol exy. The quantity (dv/dx —
du/dy)/2, which represents a rotation of the body
Q'\x + dx + u + — dx, y + v + — dx); about the z-axis, does not involve change in size or
* \ dx dx I
shape and hence is not a strain; we shall denote it by
S(pc, y + dy): the symbol 9Z.
S'(x + u + — dy, y + dy + v + ^- dy); Extending this analysis to three dimensions, we
\ dy dy ) write (u, v, w) as the components of displacement of a
R(x + dx, y + dy): point P(x, y, z). The elementary strains are thus
R'lx + dx + u + — dx + — dy,
\ dx dy Normal strains e = —,
xx
dv , dv \ dx
yy + dv
y + vH dx H dyv . _ dv
dx dy ) (2.8)
dy
In general, the changes in u and v are much smaller dw
than the quantities dx and dy; accordingly, we shall Szz
~ ~dz'
assume that the terms (du/dx), (du/dy), and so on are
small enough that powers and products can be ne- Shearing strains exy = syx = — + —,
glected. With this assumption, we see the following: dw dv
(2.9)
1. PQ increases in length by the amount
(du/dx) dx and PS by the amount (dv/dy) dy; _ _ du dw
hence du/dx and dv/dy are the fractional in- ~ Sx* ~ J Y
creases in length in the direction of the axes.
2. The infinitesimal angles o^ and 82 are equal In addition to these strains, the body is subjected to
to dv/dx and du/dy, respectively. simple rotation about the three axes given by
3. The right angle at P decreases by the amount dw/dy — dv/dz
bl + 82 = dv/dx + du/dy.
4. The rectangle as a whole has been rotated du/dz - dw/dx
counterclockwise through the angle (b{ - (2.10)
52)/2 = (dv/dx - du/dy)/2. 2 '
— du/dy
Strain is defined as the relative change (that is, the e =
fractional change) in a dimension or shape of a body.
The quantities du/dx and dv/dy are the relative in- Equations (2.10) can be written in vectorial form (see
creases in length in the directions of the x- and >>-axes, §15.1.2(a)andl5.1.2(c)):
and are referred to as normal strains. The quantity
dv/dx + du/dy is the amount by which a right angle in ©= e.k = vx (2.11)
THEORY OF ELASTICITY 37
Plastic yield change in volume per unit volume A is

Rupture
(2.12)
2.1.4 Hooke's law
In order to calculate the strains when the stresses are
known, we must know the relationship between stress
and strain. When the strains are small, this relation is
given by Hooke's law, which states that a given strain
is directly proportional to the stress producing it. The
strains involved in seismic waves are usually less than
10~8 except very near the source, so that Hooke's law
holds. When several stresses exist, each produces
strains independently of the others; hence, the total
strain is the sum of the strains produced by the indi-
(a)
vidual stresses. This means that each strain is a linear
function of all of the stresses and vice versa. This lin-
Rupture earity has important implications that will be utilized
later: It allows us to represent curved wavefronts as
a superposition of plane waves, for example, in /?-T
transforms (§9.1.5 and 9.11.1), to express a reflected
wavetrain as a superposition of individual reflections
(the convolutional model), and to justify many aspects
of seismic data processing.
In general, Hooke's law leads to complicated rela-
tions. Stress and strain can both be regarded as
second-order (3 X 3) matrices so that the Hooke's law
proportionality relating them is a fourth-order tensor.
Stress and strain can also be looked on as (1 X 6) ma-
trices (as in eq. (2.15)) and the Hooke's law propor-
tionality as a 6 X 6 matrix whose elements are elastic
constants (Landau and Lifshitz, 1986: 32-51). Sym-
metry considerations immediately reduce the number
of independent constants to 21. However, when the
medium is isotropic, that is, when properties do not
depend upon direction, it can be expressed in the fol-
lowing relatively simple form (Love, 1944: 102):
<x,7 = \A + 2u.s, (i = x, y z), (2.13)
a, = 2{L89 (i, j = x, y z; i # j). (2.14)
Fig. 2.4 Stress-strain-time relationships, (a) Stress versus These equations are often expressed as a matrix equa-
strain; (b) strain versus time. tion, <T = CE:
X+2|x X X 0 0 0
X X+2|x X 0 0 0
where £ = u\ + vj + wk is the vector displacement of
X X 0 0 0
point P(x, y), and i, j , k are unit vectors in the x-, y-,
z- directions, respectively. 0 0 0 [i 0 0
The changes in dimensions given by the normal 0 0 0 0 |x 0
strains result in volume changes when a body is 0
0 0 0 0 ix
stressed. The change in volume per unit volume is
called the dilatation and represented by A. If we start
with a rectangular parallelepiped with edges dx, dy, (2.15)
and dz in the unstrained medium, in the strained me- The equation is sometimes written e = So-, where S =
dium the dimensions are dx(l + sxx), dy(l + syy), and C"1. Components of C (or S) are sometimes called
dz(l + e2Z), respectively; hence the increase in volume stiffness (or compliance) components.
is approximately (sxx + eyy + ezz) dx dy dz. Because The quantities X and |x are known as Lames con-
the original volume was (dx dy dz), we see that the stants. If we write etj = ov/|x, it is evident that etj is
smaller the larger |JL is. Hence, |x is a measure of the

resistance to shearing strain and is often referred to (2.19)
X + jx
as the modulus of rigidity, incompressibility, or shear
modulus. X
<T = (2.20)
Although Hooke's law has wide application, it does 2(X + |uu)'
not hold for large stresses. When the stress is increased
beyond an elastic limit (fig. 2.4a), Hooke's law no k = \ (3X + (2.21)
longer holds and strains increase more rapidly. Strains
resulting from stresses that exceed this limit do not In nonviscous fluids, the shear modulus \x = 0, and
entirely disappear when the stresses are removed. hence k = X. Because we have not previously given a
With further stress, a plastic yield point may be specific name to X, we may call it the fluid ^compress-
reached at which plastic flow begins and the plastic ibility. By eliminating different pairs of constants
yielding may result in decreasing the strain. Some ma- among the three equations, many different relations
terials do not pass through a plastic flow phase can be derived expressing one of the five constants in
but rupture first. Rocks usually rupture at strains terms of two others (see problem 2.2).
~ 10-MO-4. The elastic constants are defined in such a way that
Some materials also have a time-dependent behav- they are positive numbers. As a consequence of this,
ior to stress (fig. 2.4b). When subjected to a steady a must have values between 0 and 0.5 (this follows
stress, such materials creep until eventually they rup- from eq. (2.20), because both X and |x are positive and
ture. Creep strain does not disappear if the stress is re- hence X/(X + |JL) is less than unity). Values range from
moved. 0.05 for very hard, rigid rocks to about 0.45 for soft,
poorly consolidated materials. Liquids have no resis-
2.1.5 Elastic constants tance to shear and hence for them jx = 0 and a = 0.5.
For most rocks, E, k, and |JL lie in the range from 20
Although Lame's constants are convenient when we to 120 GPa (2 X 1010 to 12 X 1010 N/m2), E generally
are using eqs. (2.13) and (2.14) other elastic constants being the largest and JJL the smallest of the three.
are also used. The most common are Young's modulus Tables of elastic constants of rocks have been given by
(E), Poissons ratio (a), and the bulk modulus (k) (the Birch (1966). (See also problem 2.4.)
symbol a is more or less standard for Poisson's ratio - Most of the preceding theory assumes an isotropic
the subscripts should prevent any confusion with the medium. In fact, rocks are usually in layers with
stressCTiy).To define the first two, we consider a me- different elastic properties, these properties often var-
dium in which all stresses are zero except axx. Assum- ying with direction. Nevertheless, in discussing wave
ing axx is positive (that is, a tensile stress), dimensions propagation, we generally ignore such differences and
parallel to vxx will increase and dimensions normal to treat sedimentary rocks as isotropic media; when one
GXX will decrease; this means that exx is positive (elon- does so, the results are useful and to do otherwise
gation in the x-direction) whereas eyy and ezz are nega- leads to extremely complex and cumbersome mathe-
tive. Also, we can show (see problem 2.1a) that eyy = matical equations, except for the case of transversely
szz. We now define E and a by the relations isotropic media, that is, media in which the properties
are the same in one plane but different along the nor-
E=ajexx, (2.16)
mal to the plane. Some rocks, especially shales, are
transversely isotropic, and more importantly, a series
of parallel beds, each of which is isotropic, but where
with the minus sign inserted to make a positive.
the properties vary from bed to bed, behaves as
To define the bulk modulus k, we consider a me-
though it is transversely isotropic (Postma, 1955;
dium acted upon only by a pressure 2P; this is equiva-
Uhrig and van Melle, 1955). Anisotropy is discussed
lent to the stresses
in §2.6.
Pressure 2P causes a decrease in the volume AT and a 2.1.6 Strain energy

dilatation A = AT/T; k is defined as the ratio of the
When an elastic medium undergoes deformation,
pressure to the dilatation that it causes, that is,
work is done and an equivalent amount of potential
k = -3VA, (2.18) energy is stored in the medium; this energy is inti-
mately related to elastic wave propagation.
with the minus sign inserted to make k positive. Some-
If the stress ax;c results in a displacement sxx, we as-
times the compressibility, \lk, is used as an elastic con-
sume that the stress is increased uniformly from zero
stant rather than the bulk modulus.
to dxx, and hence the average stress is <JXX/2. Thus,
By substituting the preceding values in Hooke's law,
we can obtain the following relations between E, a, E = work done per unit volume
and k and Lame's constants, X and JJL (see problems = energy per unit volume
2.1b and 2.1c):
WAVE EQUATIONS 39
Summing the effects of all the independent stresses Newton's second law of motion states that the un-
and using eqs. (2.13) and (2.14) gives (Love, 1944: balanced force equals the mass times the acceleration;
100) thus, we obtain the equation of motion along the x-
axis:
_ 1/ «,* + p — = unbalanced force in the x-direction on

2 ^ df
a unit volume
+ (Jy.
d(rxz
(2.24)
_ 1 dx dy dz
2[ I (XA + 2|xe . )e . . +li II
J
(i # y)
where p is the density (assumed to be constant). Simi-
A2 + |JL<' 6 L + eyy + 8
L) lar equations can be written for the motion along the
y- and z-axes.
(2.22)
Equation (2.24) relates the displacements to the
stresses. We can obtain an equation involving only dis-
Note that eq. (2.22) gives
placements by using Hooke's law to replace the
_ = XA- stresses with strains and then expressing the strains in
dE/dsxy = [x,sxy = cr^, terms of the displacements, using eqs. (2.8), (2.9),
(2.12), (2.13), and (2.14). Thus,
hence,
^. = ov ft j = x, y, z). (2.23) d2u dcr dcr dcr
+ xx _\ xz
p
dx dy dz
df dA de e^ dexz
^ + |JL
2.2 Wave equations = X + 2|JL — + IX
dx dx dy dz
2.2.7 Scalar wave equation
dx dx2 \dxdy dy2
Up to this point, we have been discussing a medium
in static equilibrium. We shall now remove this restric-
tion and consider what happens when the stresses are
not in equilibrium. In fig. 2.2, we now assume that the
Vdxdz
.
+
1
d
d (dll
dll dv dw
= X— + ^M + (Ji— — + — +
stresses on the rear face of the element of volume are dx dx \dx dy d z)
as shown in the diagram but that the stresses on the
front face are, respectively, = (x- + M; + (2.25)
where V2w is the Laplacian of u = d2u/dx2 + d2u/dy2 +

d2uldz2 (see eq. (15.14)). By analogy, we can write the
equations for v and w:
Because these stresses are opposite to those acting on
the rear face, the net (unbalanced) stresses are d2v
= (X (2.26)
dy
to* dx, dx,
dx dx dx d2w ,. ^ ,dA (2.27)
p = (X + | x > — + |x V 2 w.
d£2 dz
These stresses act on a face having an area (dy dz) and
affect the volume (dx dy dz); hence, we get for the net To obtain the wave equation, we differentiate these
forces per unit volume in the directions of the x-, y-, three equations with respect to x, y, and z, respec-
and z-axes the respective values tively, and add the results together. This gives
d2idu dv
dx dx dx
Similar expressions hold for the other faces; hence, we

dy dz
find for the total force in the direction of the x-axis
the expression that is,
dx dy dz
or in two ways in general: (a) include in the wave equa-

tion terms that represent the forces generating the
waves or (b) surround the point of observation P by a
a2 dt2 closed surface £f and regard the effect at P as being
(2.28) given by a volume integral throughout the interior of
where
tf to take into account sources inside £f plus a surface
a2 = (X + 2|m)/p. integral over £f to give the effect of sources outside tf
(see §2.2.4). To apply the first method, we note that
By subtracting the derivative of eq. (2.26) with re- eqs. (2.25), (2.26), and (2.27) are equivalent to New-
spect to z from the derivative of eq. (2.27) with respect ton's second law, and these three equations are com-
to y we get bined in eq. (2.31). Therefore, a source can be taken
into account by adding to the right-hand side of eq.
d2
dt 2
(dw dv\ _ (dw dv\ (2.31) the term pF, where F is the external nonelastic
force per unit mass (often called body force) that gives
dy dz) ~ ^ [dy dzV rise to the wave motion. Thus, eq. (2.31) becomes
that is,
p^jj = (X + \L) VA + jx V2£ + pF. (2.34)
2 2
(3 dt
(2.29) Taking the divergence and curl of eq. (2.34) and using
where
2 eq. (15.14) and problem 15.7 gives
P =
= a2 V2A + V • F, (2.35)
By subtracting appropriate derivatives, we obtain sim- dt2
ilar results for Qy and 6Z. Equations (2.28) and (2.29)
are different examples of the wave equation, which we ^=P2V2« F/2. (2.36)
can write in the general form dt2
These equations are difficult to solve as they stand.

2 2
(2.30) The solution is greatly simplified by using the Helm-
V dt
holtz separation method, which involves expressing
where V is a constant. both J and F in terms of new scalar and vector func-
tions. Thus, we write
2.2.2 Vector wave equation (2.37)
The wave equation can also be obtained using vector F = VY + V X (1, V • ft = 0. (2.38)
methods. Equations (2.25), (2.26), and (2.27) are
equivalent to the vector wave equation: Then, using problem 15.7, we obtain
A = V • £ = V2c(>,
2
p ^ = (X + IL)VA + M.V £. (2.31) 20 = V X £ = -V2\,
(2.39)
V F = V2Y,
If we take the divergence of eq. (2.31) and use eqs. V x F = -V 2 ft .
(2.12) and (15.14) we get eq. (2.28). Taking the curl of Substituting in eqs. (2.35) and (2.36), we get
eq. (2.31) and using eq. (2.11) and problem 15.7 gives
the vector wave equation for S-waves (see §2.4.1),
1 d2® V2 ((32V2X + ft —)2 = 0.

= V 2 0, (2.32)
P 2
dt 2 dt )
which is equivalent to the three scalar equations, Whenever §, \ , X or ft contain powers of x, y and z
1 -j2Q higher than the first, these equations can only be satis-
(2.33) fied for all values of x, y and z if the expressions inside
the parentheses are identically zero at all points. Be-
cause a linear function of x, y, and z corresponds to a
2.2.3 Wave equation including source term uniform translation and/or rotation of the medium,
we can ignore this possibility and write (Savarensky,
The foregoing discussion of the wave equation has 1975: 199)
made no mention of the sources of the waves, and in
fact, the equations discussed are only valid in a ^ = a2V2d> + Y, (2.40)
source-free region. Sources can be taken into account dt2
WAVE EQUATIONS 41
tained identical results for S-waves. Thus, the previous

(2.41)
a. equations refer to either P- or *S-waves.
2.2.4 Kirchhoff's theorem 2.2.5 Plane-wave solutions
Method (b) referred to in §2.2.3 is in fact an extension Let us consider first the case where i|/ is a function
of method (a). It uses the superposition concept only of x and t, so that eq. (2.30) reduces to
(which follows from the linearity expressed in Hooke's
law). We regard the wave motion at a point P as the (2.45)
V2 dt2 dx2
superposition of the waves from all sources R within
some volume T surrounding P plus the waves radiated Any function of (x — Vt),
by points Q on the surface £f surrounding the volume
(which takes into account any disturbances from =f(x- Vt), (2.46)
sources outside the volume). We adjust the times for is a solution of eq. (2.45) (see problem 2.5a) provided
these sources so that their effects all arrive at P at the that \\J and its first two derivatives are finite and con-
same instant t0. We take Y(x, y z, tR) in eq. (2.40) as tinuous. This solution (known as d'Alembert's solu-
the source density (body force/unit volume) inside £f tion) furnishes an infinite number of particular solu-
and specify <fr(x, y z, tQ) for each point on the surface tions (for example, e(x~Vt\ sin (x - Vt), (x - Vt)3,
tf, tR, and tQ being the retarded times (t0 — r/V), where where we must exclude points at which these functions
V is the velocity, and r is the distance between P and and their first three derivatives cease to exist or are
the sources R or Q, that is, r/V is the time for the wave discontinuous). The answer to a specific problem con-
to travel from R or Q to P. Thus, we specify the wave sists of selecting the appropriate combination of solu-
motion at different points at different times such that tions that also satisfies the boundary conditions for
the waves from all points arrive at P at the same in- the problem.
stant t0. The result, known as Kirchhoff's theorem (or A body wave is defined as a "disturbance" that trav-
formula) (Ewing, Jardetzky, and Press, 1957: 16), is els through the medium and carries energy (Logan,
1987: 230). In our notation, the disturbance i|i is a vol-
ume change when \\f = A and a rotation when \\f =
y z, t0) = jjf (^ 8.. Obviously, the disturbance in eq. (2.46) is traveling
4l dif,
d(\lr)
(2.42)
along the x-axis. We shall now show that it travels
with a speed equal to the quantity V
In fig. 2.5a the certain part of the wave has reached
point JP0 at time t0. If the coordinate of Po is x0, then
the value of i|i at Po is i|i0 = f(x0 - VQ. If this same
portion of the wave reaches Px at time t0 + A/, then
we have for the value of i|/ at Pl
where T] is the outward-drawn unit normal, and the
square brackets denote functions evaluated at point Q ill, = / ! * , + A x - V(to + At)}.
at time tQ = tQ - r/V; [<|>] is often referred to as a re-
tarded potential. If we assume that each source emits But, because this is the same portion of the wave that
spherical waves (§2.2.6) of the form (l/r)e-ja)(r/K'-r) (see was at Po at time t0, we must have i|/0 = i|/15 that is,
eqs. (2.55) and (2.56)), eq. (2.42) becomes (Savaren- x0 vt0 x0 x - V(t0 Af).
sky, 1975: 234)
Thus, the quantity Kis equal to Ax/At and is therefore
the speed with which the disturbance travels. The re-
ciprocal of velocity, l/V, is called slowness.
A function of (x + Vt), for example, i|> = g(x + Vt),
is also a solution of eq. (2.45). It denotes a wave trav-
eling in the negative x-direction. The general solution
of eq. (2.45),
g(x+ Vt), (2.47)
where in the integrand, represents two waves traveling along the x-axis in op-
posite directions with velocity V
(2.44) The quantity x ± Vt (or a constant times these ex-
pressions; see §2.1.1) is the phase. The surfaces on
and a) is the angular frequency (see §2.1.1). which the wave motion is the same, that is, the sur-
Because we started from eq. (2.40), eqs. (2.42) to faces on which the phase has the same value, are
(2.43) are valid for P-waves (see §2.4.1). However, we known as wavefronts. In the case we are considering,
could just as well have started from eq. (2.41) and ob- i|i is independent of y and z, and so the disturbance is
the same everywhere on a plane perpendicular to the

x-axis; the wavefront is therefore plane and the wave
is a plane wave. Note that the wave is traveling in the
direction normal to the wavefront; this holds for all
waves in isotropic media. A line denoting the direction
of travel of the wave energy is called a raypath.
Plane waves are easier to visualize and to treat
mathematically than more complicated waves. More-
over, curved wavefronts can be approximated as
closely as desired by a superposition of plane waves.
-k
It is convenient at times to have an expression for a
plane wave traveling along a straight line inclined at
an angle to each of the axes. Assume that the wave is
traveling along the x'-axis, which has direction co- Fig. 2.5 Illustrating the velocity of a wave.
sines (€, m, n) relative to the x-, y-, and z-axes (fig.
2.6). Then, at a point P on the x'-axis at a distance x'
from the origin, we have
x' = €JC + my + nz,
where the coordinates of P are {x, y, z). Then,
i|i = /(€v + my + nz - Vt)
+ g(€x + my + nz + Vt). (2.48)
2.2.6 Spherical-wave solutions

In addition to plane waves, we shall have occasion to
use another important type of wave, the spherical
wave, where the wavefronts are a series of concentric
spherical surfaces. We express eq. (2.30) in spherical
coordinates (r, 0, 4>), where 0 is the colatitude, and $
the longitude (see problem 2.6b).
ira 1 d .. d\b\ Fig. 2.6 Wave direction not along an axis.
+ sine —
r 2
[dr VTr) sin 6 dO \
1 d2\\s
2
(2.49)
sin 0
We consider only the special case when the wave mo-
tion is independent of 0 and 4>, hence is a function
only of r and t. Then we get the simplified equation
1 82\\>
(2.50)
V2 dt2 r2 dr \ dr)
A solution of the foregoing equation is
i|» = (\lr)f(r - Vt) (2.51)
(see eq. (2.46)). Obviously,
* = (l/r)g(r + Vt)
is also a solution and the general solution of eq. (2.50)
(see problem 2.5c) is
v|i = (l/r)/(r - Vt) + (\/r)g(r + Vt), (2.52) Fig. 2.7 Relation between spherical and plane waves.
in which the first term represents a wave expanding
outward from a central point and the second term a
wave collapsing toward the central point. has the same value at all points on the spherical sur-
When r and t are fixed, (r - Vt) is constant and face of radius r. The spherical surfaces are therefore
hence i|/ is constant. Thus, at the instant /, the wave wavefronts and the radii are rays. Obviously, the rays
G E N E R A L ASPECTS OF WAVES 43
are normal to the wavefronts as in the case of plane we can get either the cosine or sine form by taking the
waves. (This is not always the case in anisotropic real or imaginary part of i|/.
media.) The quantities (€, m, n) in eq. (2.54) represent the
As the wave progress outward from the center, the direction cosines of the ray. In problem 15.9a, we
radius increases by the amount V during each unit of show that £2 + m2 + n2 = 1. Although ordinarily each
time. Eventually, the radius becomes very large and of the cosines has a maximum value of unity, satis-
the portion of the wavefront near any particular point fying the wave equation requires only that the sum of
will be approximately plane. If we considerfig.2.7, we the squares be unity. If we admit pure imaginary num-
see that the error that we introduce when we replace bers, some of the "direction cosines" can be greater
the spherical wavefront PQR by the plane wavefront than unity. Let us take infig.2.6, dY = j0, 92 = 2~TT,
P'QR' is due to the divergence between the true direc- 93 = 2 TT — j9, 9 being real and positive; then
tion of propagation given by the direction of the ra-
dius and the assumed direction normal to the plane. € = cos j9 = cosh 9, m = 0,
By taking OQ very large or PR very small (or both), n = cos (ITT - j9) = sin j9 = j sinh 9,
we can make the error as small as desired. Because
€2 + m2 + n2 = cosh2 9 - sinh2 9 = 1,
plane waves are easy to visualize and also the simplest
Jj = ^Q-(MZ/V) sinh 6pjw[(x/F) cosh 6-t] (^ ^J\
to handle mathematically, we generally assume that
conditions are such that the plane-wave assumption This represents a plane wave traveling parallel to the
is valid. x-axis with velocity F/cosh 9 < V and amplitude
ê-(Wz/F)sinhe if w e h a( j taken 9j = —j9, this would give
a wave traveling in the x-direction with amplitude de-
23 General aspects of waves creasing upward in the negative z-direction. Because
the amplitude decreases exponentially with z, these
2.3.1 Harmonic waves
waves are called evanescent waves. We shall refer again
In §2.2.5 and 2.2.6, we discussed the geometrical as- to these waves in §2.7.5.
pects of waves, that is, how they depend on the space In exploration seismology, the range of frequencies
coordinates. We now consider the time dependence of recorded with appreciable energy is generally from
wave motion. about 2 to 120 Hz, and the dominant frequencies lie
The simplest time variation that a wave can have is in a narrower range from 15 to 50 Hz for reflection
harmonic (sinusoidal), equivalent to simple harmonic work and from 5 to 20 Hz for refraction work. Be-
motion. In general, waves are more complex than this, cause velocities generally range from 1.6 to 6.5 km/s,
but the methods of Fourier analysis (§ 15.2) allow us dominant wavelengths range from about 30 to 400 m
to represent almost any complex wave as a superposit- for reflection work and from 80 to 1300 m for re-
ion of harmonic waves. Harmonic waves, because of fraction.
their simplicity, can be regarded as the time equivalent
of plane waves in space.
2.3.2 Wave interference
Adding TT/2 to the phase in eq. (2.3) changes cosine
to sine, so harmonic waves can be written in either If two waves are superimposed, they interfere with
form. Some of the commonest forms are the fol- each other; the interference is constructive if they tend
lowing: to add and destructive if they tend to cancel. When the
i|i = A cos [(2TT/\)(X - Vt)] = A cos K(X - Vt)
two waves are harmonic and have the same frequen-
= A cos (KX — at) cies and wavelengths (hence the same velocities), their
= A cos 2TT(X/\ - vt) = A cos 2ir(jc/\ - tIT)
amplitudes sometimes add together and sometimes
= Acosu>(x/V- t), cancel (at least partially); thus, they form a new wave
of the same frequency and wavelength with different
(2.53) amplitude and phase-shifted. When several harmonic
i|/ = A cos K(€X + my + nz - Vt) 1 n -.. waves with different amplitudes, frequencies, and/or
(
= A sin [K(€X + my + nz - Vt) + TT/2], J > wavelengths are added together, the results are usually
very complex; constructive interference occurs when
I|I = {Air) cos K(r - Vt) + (Blr) cos K(r + Vt). (2.55)
the phases are nearly the same (e.g., §2.7.4 and 13.3),
Equation (2.53) represents a plane wave traveling in otherwise destructive interference results in at least
the +x-direction, eq. (2.54) a plane wave moving some attenuation. If the waves are not harmonic, they
along a straight line with direction cosines (€, m, «), can be resolved by Fourier analysis (§9.1 and 15.2)
and eq. (2.55) a spherical wave expanding from and into harmonic components that can then be added to
collapsing toward the origin. determine the nature of the interference.
Equation (15.45) enables us to combine the cosine If we add two harmonic waves of equal amplitudes
and sine expressions for a harmonic wave; thus, if we (A) and velocities but slightly different frequencies
write (see problem 2.7), the sum is B cos (KOX - co00, where
B ~ 2A cos (AKX — Aa>0, KO and o)0 being average
- t] =
(2.56) values, and AK and Aco half of the differences between
the values for the two waves. We regard B as the vari- wavefront at time t0 + At. Except on the envelope, the
able amplitude of the resultant wave; at a fixed point, elemental waves interfere destructively with each
B varies between ±2^4 at the rate of AO)/2TT times per other so that their effects cancel. When AB is plane,
second, that is, slowly in comparison with the wave and V constant, we need draw only two arcs and the
frequency (o0. This phenomenon is called beating. straight-line tangent to the two arcs defines the new
wavefront. Note that Huygens' principle gives only
phase information; it does not give amplitudes.
2.3.3 Huygens' principle
The solutions of the wave equation given by eqs. (2.47)
2.4 Body waves
and (2.52) are restricted to plane and spherical waves.
On the other hand, the Kirchhoff formula is valid for 2.4.1 P-waves and S-waves
any type of body wave. As expressed in eq. (2.42) (as-
suming no sources inside £P), it states that the effect at Up to this point, our discussion of wave motion has
a point P is the sum of effects that took place earlier been based upon eq. (2.30). The quantity i|/ has not
at all points on a surface Sf enclosing P, allowance be- been defined; we have merely inferred that it is some
ing made for the time for these effects to travel from disturbance that is propagated from one point to an-
&> to P. Thus, each point on &> behaves as though it other with speed V. However, in a homogeneous iso-
were a new wave source. tropic medium, eqs. (2.28) and (2.29) must be satisfied.
We can identity the functions A and 6, with i|/ and con-
To obtain Huygens' principle, we take Sf coincident
clude that two types of waves can be propagated in a
with that portion of the wavefront that we wish to take
homogeneous isotropic medium, one corresponding
into account in finding the effect at P, and then com-
to changes in the dilatation A and the other to one or
plete the closed surface by passing it through space
more components of the rotation given in eq. (2.11).
where the effect has not yet arrived so that 4> is zero
The first type is variously known as a dilatational,
over this part.
longitudinal, irrotational, compressional, or P-wave,
Huygens' principle is important in understanding
the latter name being given because this type is usually
wave travel and is often useful in drawing successive
the first (primary) event on an earthquake recording.
positions of wavefronts. Huygens' principle states that
The second type is referred to as the shear, transverse,
every point on a wavefront can be regarded as a new
rotational, or S-wave (because it is usually the second
source of waves. The physical rationale behind this is
major event observed on earthquake records). The P-
that each particle located on a wavefront has moved
wave has the velocity a in eq. (2.28) and the S-wave
from its equilibrium position in approximately the
the velocity p in eq. (2.29), that is,
same manner, that the elastic forces on neighboring
particles are thereby changed, and that the resultant
of the changes in force because of the motion of all
the points on the wavefront thus begins to produce
the motion that forms the next wavefront. In this way,
Huygens' principle helps explain how information (2.59)
about seismic disturbances is communicated in the
earth. Specifically, given the location of a wavefront where M is the P-wave modulus. Because the elastic
at a certain instant, future positions of the wavefront constants are always positive, a is always greater than
can be found by considering each point on the first p. Using eq. (2.20), we see that
wavefront as a new wave source. In fig. 2.8, AB is the
wavefront at the time t0 and we wish to find the wave-
front at a later time t0 + Af. During the interval Af, (2.60)
X + 2(JL 1- a
the wave will advance a distance V Af, V being the
velocity (which may vary from point to point). We se- (see fig. 2.9). As a decreases from 0.5 to 0, p/a in-
lect points on the wavefront, Pv P2, P3, and so on, creases from 0 to its maximum value, \I-\J2; thus, the
from which we draw arcs of radii V bd. Provided we velocity of the S-wave ranges from 0 to 70% of the
select enough points, the envelope of the arcs {A'B') velocity of the P-wave.
will define as accurately as we wish the position of the For fluids, jx is zero and hence p is also zero; there-
fore S-waves do not propagate in fluids. Using eq.
(2.21), we see that for a fluid, X = k; hence,
j\ p2 a = (k/p)m. (2.61)
The seismic velocity in actual rocks depends on
many factors, including porosity, lithology, cementa-
tion, depth, age, pressure regime, interstitial fluids,
etc., which are discussed in chap. 5. The velocity of
water-saturated sedimentary rocks is generally in the
Fig. 2.8 Using Huygens' principle to locate new wavefronts. 1.5 to 6.5 km/s range, increasing with loss of porosity,
BODY WAVES 45
a/p mum compression (maximum A) at the wavefront D;

5.0 4.0 3.0 2.5 2.0 1.7 1.4 particle velocity is zero at each of these points.
0.5
J We can visualize the plane-wave situation by imag-
5.0 ining that the radius in fig. 2.10 has become very large
0.4
so that the wavefronts are practically plane surfaces.
\ 3.0
s The displacements will be everywhere perpendicular
0.3 2.0 to these planes so that there will no longer be conver-
\
1.5 gence or divergence of the particles of the medium as
0.2 they move back and forth parallel to the direction of
1.0 propagation of the wave. Such a displacement is longi-
\
0.1 tudinal, which explains why P-waves are sometimes
\ called longitudinal waves. P-waves are the dominant
waves involved in seismic exploration. A plane P-wave
0.2 0.4 0.6
is illustrated in fig. 2.11a.
p/a
To determine the motion of a medium during the
Fig. 2.9 (3/a as a function of Poisson's ratio a and kl\L. passage of an S-wave, we return to eq. (2.29) and con-
sider the case where a rotation 0Z, which is a function
of x and t only, is being propagated along the x-axis.
We have
p 2 dt2 dx2
Because
_ dv du _ dv
z
~ dx dy~ dx
from eq. (2.10), we see that the wave motion consists
Fig. 2.10 Displacements for a spherical P-wave.
cementation, depth, and age. (Velocity versus depth

relations for three situations are shown in fig. 11.28.)
The velocity of P-waves in water is approximately 1.5
km/s. P-wave velocity is lowered, often markedly,
when a gas replaces water as the interstitial fluid. This
is especially important in the near-surface, generally
above the water table, where a low-velocity layer
(LVL, also called the weathered layer) typically has a
velocity in the 0.4 to 0.8 km/s range, occasionally as
low as 150 m/s, sometimes as high as 1.2 km/s.
Let us investigate the nature of the motion of the
medium corresponding to the two types of wave mo-
tion. Consider a spherical P-wave of the type given
by eq. (2.51). Figure 2.10 shows wavefronts drawn at
quarter-wavelength intervals, t being chosen so that (a) (b)
KFMS a multiple of TT/2. The arrows represent the di- Fig. 2.11 Motion during passage of plane body waves. (After
rection of motion of the medium at the wavefront. Earth, 2d ed., by F. Press and B. Siever, p. 424. Copyright 1974
The medium is undergoing maximum compression at by W. H. Freeman and Company; reprinted with permission.)
B (that is, the dilatation A is a minimum) and mini- (a) P-wave; (b) S-wave.
solely of a displacement v of the medium in the y-

direction, v being a function of both JC and t. Because
v is independent of y and z, the motion is the same
everywhere in a plane perpendicular to the x-axis;
thus, the case we are discussing is that of a plane
S-wave traveling along the x-axis (fig. 2.11b).
We can visualize the foregoing relations using fig.
2.12. When the wave arrives at P, it causes the medium
in the vicinity of P to rotate about the axis Z'Z" (par-
allel to the z-axis) through an angle e. Because we are
dealing with infinitesimal strains, s must be infinites-
imal and we can ignore the curvature of the displace-
ments and consider that points such as P' and P" are
displaced parallel to the >>-axis to the points Q' and Z"
Q". Thus, as the wave travels along the x-axis, the me-
dium is displaced transversely to the direction of
propagation, hence the name transverse wave. More-
over, because the rotation varies from point to point at
any given instant, the medium is subjected to varying
shearing stresses as the wave moves along; this ac- Fig. 2.12 Rotation of medium during passage of an S-wave.
counts for the name shear wave.
Because we might have chosen to illustrate 9^ in fig.
2.12 instead of 9Z, it is clear that shear waves have two
degrees of freedom, unlike P-waves, which have only
one - along the radial direction. In practice, S-wave and x(x> y> z, 0> which are solutions of the P- and S-
motion is usually resolved into components parallel wave equations, respectively, and which are so chosen
and perpendicular to the surface of the ground, these that u, v, w (or u, v, w) can be found bv differentiation.
are known respectively as SH- and SV-waves. When A simple example of such functions is the following:
the wave is traveling neither horizontally nor verti-
cally, the motion is resolved into a horizontal [SH] X = 0, V<\> = £ = (id + vj + wk),
component and a component in the vertical plane so that (2.62)
through the direction of propagation. Henceforth,
S-wave will mean S V-wave unless otherwise noted. u = —-, w =
dx dy dz
Because the two degrees of freedom of S-waves are
independent, we can have an S-wave that involves mo- This procedure is valid only if it corresponds with A
tion in only one plane, for example, 57/or SVmotion; being a solution of the P-wave equation. Because £ is
such a wave is said to be plane-polarized. We can also a solution and A = V • £ = V2c|), A is also a solution
have a wave in which the SH and S V motion have the (because derivatives of a solution are also solutions).
same frequency and a fixed phase difference; such a Setting x = 0 is equivalent to saying that S-waves do
wave is elliptically polarized. Polarization of 5-waves not exist and this choice of potential functions is suit-
is a factor in their exploration use (see §13.1). Note able for discussing wave motion in fluids.
that we cannot have a spherically symmetrical S-wave For wave motion in three-dimensional solids, $ and
(analogous to the F-wave illustrated in fig. 2.10). X can be defined so that
S-wave amplitude must vary with direction.
In the case of a medium that is not homogeneous (2 63)
and isotropic, it may not be possible to resolve wave -
motion into separate P- and S-waves. However, inho- This ensures that A and 0 are solutions of the P- and
mogeneities and anisotropy in the earth are small S-wave equations, respectively (see problem 2.9a).
enough that the assumption of separate P- and For two-dimensional wave motion in the xz-plane,
S-waves is valid for practical purposes. c|> and x can be defined by
X = ~Xh
2.4.2 Displacement and velocity potentials
(2 64)
Solutions of the wave equations such as those in eqs. JJ —- I_ I (± '
(2.48) and (2.52) furnish expressions for A and 9. dX dz' dz dx
However, often we need to know the displacements u, It is easy to show that eqs. (2.12) and (2.11) can be
v, w, or the velocities u — du/dt, v, w, and reference to expressed as
eqs. (2.8) to (2.12) shows that these are not easily
found given only values of A and 9 r This difficulty is A = V • £ = V2c(),
(2.65)
often resolved by using potential functions 4>(x, y> z, t) 2 0 = V X £= V 2 xj,
BODY WAVES 47
so that A and 0 are again solutions of the P- and By using the adiabatic law, logarithmic differentiation
S-wave equations (see problem 2.9b). of eq. (2.69) gives k = 7 ^ , and hence
Because the wave equations are still valid if both
sides are differentiated with respect to time t, it fol- a = (707p)1/2. (2.70)
lows that velocity potentials will be obtained in each
of the preceding cases if u, v, w, and £ are replaced 2.4.4 Boundary conditions
with u, v, w, and £.
When a wave arrives at a surface separating two media
having different elastic properties, it gives rise to re-
flected and refracted waves as described in §3.1.1. The
2.4.3 Wave equation in fluid media relationships between the various waves can be found
from the relations between the stresses and displace-
In fluids, only P-waves are propagated and we are gen- ments on the two sides of the interface. At the bound-
erally interested in pressure variations rather than dis- ary between two media, the stresses and displace-
placements or velocities, as in solid media. Equation ments must be continuous.
(2.62) can be expressed in terms of pressure 2P. We Two neighboring points R and S, which lie on op-
redefine 4> in the form posite sides of the boundary as shown in fig. 2.13 will
in general have different values of normal stress. This
V<|> = u\ + vj + wk, u = —, etc. (2.66) difference results in a net force that accelerates the
dt layer between them. However, if we choose points
In eq. (2.24), we set closer and closer together, the stress values must ap-
proach each other and in the limit when the two
Vxy = °yz = azx = 0, <rxx = uyy = vzz = -9; points coincide on the boundary, the two stresses must
hence, using eq. (2.24), we get be equal. If this were not so, the infinitesimally thin
layer at the boundary would be acted upon by a finite
p— = = acceleration along the x-axis, (2.67) force and hence have an acceleration that would ap-
dt2 dx proach infinity as the two points approach each other.
and similarly for the y- and z-axes. Adding the three Because the same reasoning applies to a tangential
components of acceleration gives stress, we see that the normal and tangential compo-
nents of stress must be continuous (cannot change
abruptly) at the boundary.
dt The normal and tangential components of displace-
ment must also be continuous. If the normal displace-
Ignoring the additive constant due to hydrostatic pres- ment were not continuous, one medium would either
sure (because we are interested only in pressure varia- separate from the other, leaving a vacuum in between,
tions), or else would penetrate into the other so that the two
media would occupy the same space. If the tangential
(2.68) displacement were not continuous, the two media
would move differently on opposite sides of the
if we consider only harmonic waves of the form boundary and one would slide over the other. Such
relative motion is assumed to be impossible and so
displacement must be continuous.
(see eq. (2.56)). Thus, both <\> and 9 satisfy the P-wave When one or both of the solid media are replaced
equation as in eq. (2.28), the velocity reducing to a = by a fluid or a vacuum, the boundary conditions are
(k/p)m in fluids (see eq. (2.61)). reduced in number (see problem 2.10),
In the case of a gas, k depends upon the way the
gas is compressed, isothermally or adiabatically (that 2.4.5 Waves from a spherical source
is, with no transfer of heat during the wave passage).
For sound waves in air, the compression is essentially The potential function <f> = (llr)f(t — rIV) is a solution
adiabatic so that the pressure and volume obey the to the wave equation when there is spherical symmetry
law, (see eq. (2.51)); hence, the radial displacement u(r, i) is
g>W = constant, 7 = cjcv « 1.4 for air, (2.69)

where c^ and cr are the specific heats at constant pres-
sure and volume, respectively (Hsieh, 1975: 54-5;
Lapedes, 1978). Equation (2.18) can be written
(using eq. (2.62) with the x-axis in the radial direc-
k = — = tion). For harmonic waves, the two terms have equal
AT/T dT ' importance at a distance r = \/2TT, but the first term
where ASP is the pressure change created by the wave. decays rapidly in importance at greater distances. The
/e'i
Fig. 2.14 Radial displacement involves shape distortion. When

the radial displacement decreases with distance from source S,
sector ABDC becomes thinner and approaches a rectangular
Fig. 2.13 Continuity of normal stress. plate.
second term is the far-field effect, whereas the near- Note that the lower limit of the integral means that
field effect depends on both terms. This distinction is t = 0 is the instant at which the wave first reaches the
important when calculating a far-field waveshape surface r0, uo(t) being zero before this.
from near-field recordings. Note that radial motion To carry the calculation further, we must know uo(t).
involves shape distortion (fig. 2.14) and therefore Let us approximate an explosion by the expression
shear strain.
Equation (2.71) can be used to derive the wave mo- t > 0, a > 0, (2.76)
tion created by symmetrical displacement of the me- = 0 t< 0.
dium outward from a point source. When the wave is
created by very high pressures, as in an explosion of
Then
dynamite, the wave equation is not valid near the
source because the medium does not obey Hooke's f(f) = -r0Ke-™° I teWo-'Mt
law there; this difficulty is usually resolved by sur-
Vk
rounding the source by a spherical surface of radius
r0 such that the wave equation is valid for r > r0, then V/r0 - a
specifying the displacement or pressure on this sur- We replace t in this expression by £ = t — (r — ro)/V
face due to the source. and eq. (2.73) becomes
Let us consider the case where the displacement
u(r, i) is to be found, given the displacement uo(t) of
u(r, t) = ^ =
the surface r = r0. We let £ = t - (r - ro)/Kand write dr r(V/r0 - a) [r0
0= ^ 0, r > r0 (2.72) _ Q- Vt,hro + — e~ a d (2.77)
= 0, r r J
then
f( VIY — fl\ \f I
1 \
u(r, i) = ^= - - (2.73) (2.78)
the latter equation giving the far-field solution.
At r = r0, £ = t and u(r, t) = uo(t), where uo(t) depends The fact that eqs. (2.77) and (2.78) are valid only
on the specific source: for £ > 0 (see eq. (2.72)) merely means that u(r, t) is
zero until t = (r - ro)/V, that is, until the disturbance
(2.74) reaches the point. At this instant, £ = 0 and u(r, t) =
k(rjr); hence the initial displacement is the same as
that of the surface r0 except that it is reduced by the
Using these values and multiplying both sides of eq. factor rjr, that is, u(r, t) falls off inversely as the dis-
(2.74) by the integrating factor eF'/r°, we get tance (see §2.7.1 and eq. (2.109). Moreover, u = 0 at
t = oo and also when (see eq. (2.77))
V(\/r0 - o + (VIr - a)e~at = 0,
= -r0 Vuo(t)ev<\ that is, when
fit) = -r0Ve-Vtlr° I uo(t)evt»»dt. (2.75) t = r - In

Jo V/r0 - a ror(a ~ VIf)
SURFACE WAVES 49
Provided Vlr0 > a > Vlr, this equation has a real posi- components exist (S7/-motion is parallel to the xy-
tive root and u(r, i) will vanish, that is, the displace- plane and so is not involved in the boundary condi-
ment must change sign. Because VI r0 is large in prac- tions) and adjust their amplitudes to satisfy the
tice and Vlr rapidly becomes small, in general the boundary conditions.
unidirectional pulse in eq. (2.76) gives rise to an oscil- Appropriate potentials are
latory wave. § = ^Q-mKZQJKix- VRt)9 ^ BQ-IKZQJK(X- VRt)^
=
By using different expressions for uo(t) in eq. (2.75)
or by specifying 0PO(O, the pressure at the cavity, we (2.82)
can investigate the wave motion for various spheri- where m and n must be real positive constants so that
cally symmetrical sources (see Blake, 1952; Savaren- the wave decreases in amplitude away from the sur-
sky, 1975: 243-55). By finding the limit as a in eq. face; VR is, of course, the velocity of the Rayleigh
(2.76) goes to zero (see problem 2.12), we get the result wave. Substituting 4> and \ v i n ecls- (2.79) and (2.80)
for a unit step, step (t); then the results for other inputs gives
can be found using convolution techniques (see
§15.4.1). m2 = (1 - K 2 /a 2 ), n2 = (1 - F 2 /p 2 ). (2.83)
Because m and n are real, VR < P < a, so that the
velocity of the Rayleigh wave is less than that of the
2.5 Surface waves S-wave.
2.5.1 General We next apply the boundary conditions. Using the
results of problem 2.11, we get for z = 0
The wave equations for P~ and S- waves in terms of the
potential functions of eq. (2.64) are
V2((> = (I/a 2 ) d2$ldt2, (P-wave) (2.79)
dz2 dxdz (2.84)
V2Xv = (1/P2) d\vldt\ (SF-wave) (2.80)
V2XH = (1/P2) VXHIW, (SH-vniYe) (2.81) dxdz dz2 dx2 =o.
where the S-wave potential has been replaced with the
functions \ v and xH corresponding to SV- and SH-
Substituting eq. (2.82) into the foregoing and setting
components. If we consider plane waves traveling in
z = 0 gives
the direction of the x-axis in an infinite homogeneous
medium, solutions of these equations are of the form [(X + 2\x)m2 - X]A + 2JHJJL0 = 0
ejK(x - vt)^ y — a o r p However, other solutions are pos-
and
sible when the infinite medium is divided into different
media. When the xy-plane separates two media, solu- -2jm^ + (n2 + \)B = 0.
tions of the form e±KZeJK(x ~ Vt) exist under certain con-
ditions. These solutions correspond to plane waves We can use eqs. (2.58), (2.59), and (2.83) to write the
traveling parallel to the x-axis with velocity Kand am- first result in the form
plitude decreasing exponentially with distance from (2p2 - V$A + 2j«p2£ = 0.
the .xy-plane (in a semiinfinite medium; see §2.5.2 to
2.5.4). Such waves are called surface waves because Eliminating the ratio BIA from the two equations
they are "tied" to the surface and diminish as they get gives
farther from the surface. (2 - V2l$2)(n2 + 1) = 4mn;
hence,
2.5.2 Rayleigh waves
V6R - 8P2K£ + (24 - 16p2/a2)p<%2
The most important surface wave in exploration seis- + 16(p 2 /a 2 - l)p 6 = 0. (2.85)
mology is the Rayleigh wave, which is propagated
along a free surface of a solid. Although a "free" sur-
Because the left side of eq. (2.85) is negative for VR =
face means contact with a vacuum, the elastic con-
0 and positive for VR = +p, a real root must exist be-
stants and density of air are so low in comparison
tween these two values, this root giving the Rayleigh
with values for rocks that the surface of the earth is
wave velocity VR. However, we cannot find this root
approximately a free surface. Ground roll is the term
without knowing p/a.
commonly used for Rayleigh waves.
For many rocks, a « lA, that is, (p/a)2 « Vz from
We take the free surface as the xy-plane with the
eq. (2.60). If we use this value, the three roots of eq.
z-axis positive downward. The boundary conditions
(2.85) are V\ = 4p2, 2(1 ± 1/A/3)P2. Because VRI$
(§2.4.4) require that azz = 0 = axz at z = 0 (see prob-
must be less than unity, the only permissible solution
lem 2.10), that is, two conditions must be satisfied,
and so we require two parameters that can be is
adjusted. Therefore, we assume that both P- and SV- V*R = 2(1 - 1/V3)P2, or VR = 0.919p.
We now find that VRla = 0.531, m = 0.848, n = 0.393, Medium (2):

and BIA = +1.468J; hence,
where mp n. are real, positive constants, and Vs is the

velocity. Substituting <(>., x, in eqs. (2.79) and (2.80),
Using eq. (2.64), we get for the displacements at the we find that
surface
m] = 1 - (Vsla)\ n] = 1 - (Vs/$)2 (i = 1, 2)
u = 0.423JKê ,JK(X~ VRt)
w = X V
0.620KAQ^ - K'\ (2.88)
Taking the real part of the solution (which corre- Because m. and n. are real, Vs must be less than the
sponds to a displacement at the source of cos a>t (see smaller of $x and P2.
eq. (15.45)), we obtain finally The boundary conditions required that wx = w2,
"i = W2> °"Ji = °"zJ2> °"*Ji = °"J2 at z = 0. The results
u = - 0 . 4 2 3 K , 4 sin K(X - VRt),
of problem 2.11 show that these conditions lead to
w = 0.620K,4 COS K(JC - VRi). (2.86) equations involving first and second derivatives of the
At a given point on the surface, a particle describes potentials, z being set equal to zero after the differen-
an ellipse in the vertical xz-plane, as shown in fig. tiation; consequently, all terms will have the factors
2
2.15a, the horizontal axis being about two-thirds the ejK(* - vst) a n d either K or K , and these factors will can-
vertical axis. The angle 0 is given by cel out, so we can ignore them; at the same time, the
exponential term in z will become unity. Moreover,
tan 6 - - wlu = 1.465 cot K(X - VRi). (2.87)
differentiation with respect to x and z is equivalent to
As t increases, cot K(X — VRi) and 0 increase, that multiplying the potentials by JK and ±mtK or ±n.K,
is, P moves around the ellipse in a counterclockwise respectively; by eliminating K, differentiation becomes
(retrograde) direction for a wave moving from left to equivalent to multiplying by j and ±m., ±nr The four
right. boundary conditions now give
Rayleigh-wave velocity as a function of Poisson's
mlAl - ]BX = -m2A2 - j£ 2 ,
ratio is shown in fig. 2.16. Because VR as given by eq.
(2.85) is independent of frequency, Rayleigh waves on )AX + nlBl = )A2 - n2B2,
the surface of a homogeneous medium do not exhibit
dispersion (see §2.7.4). Field observations (figs. 2.15b -.Mi)
and 2.15d) agree roughly with the type of motion = \ 2 (-1 + m22)A2 ]n2B2\
2
shown in fig. 2.15a, differences being attributed to the (n
Earth being layered and anisotropic rather than an (n\\ + \)B2].
ideal, homogeneous, isotropic medium. Measure- If we transfer all terms to the left-hand side, we
ments also show that Rayleigh waves are dispersive have a set of four homogeneous equations (§15.1.1);
(Dobrin, 1951). Rayleigh waves are low-velocity, low- these have a nontrivial solution only if the determi-
frequency waves with a spectrum that is not sharply nant of the coefficients vanishes. Setting the determi-
peaked, and hence involve a broad range of wave- nant equal to zero and using eqs. (2.58), (2.59), and
lengths. Because WK and HK determine the penetration (2.88), we get the following equation for Vs:
(penetration showing the exponential falloff predicted
by eq. (2.82) is illustrated in fig. 2.15c), there is a large VI [(P2 - Pi)' - (Pim2
variation of penetration for different frequency com- + 4 ^ ( | x 1 - |x2) [p2 (1 - mxnx) — px (1 — m2n2)}
ponents, most of the energy being confined to a zone + 4 ( ^ - ^ 2 ) 2 (1 + mxnx) (1 + m2n2) = 0.
one or two wavelengths thick. Because the elastic con- (2.89)
stants vary considerably near the surface, especially at
the base of the LVL (see §5.3.2), the velocity varies Equation (2.89) was first given by Stoneley (1924).
with wavelength, the waves are dispersive, and the Scholte (1947) studied the properties of the equation
shape of the wavetrain changes with distance. and found that a solution always exists when one of
the media is a fluid but, when both media are solids,
a solution exists only when $x « (32 and the ratios
2.5.3 Stoneley waves p/p 2 and i x / ^ fall within the narrow limits shown in
fig. 2.17. These waves are a type of generalized Ray-
If an infinite medium is divided by the .xy-plane into leigh waves and are usually called Stoneley waves.
two different semiinfinite media, four boundary con- Stoneley waves are often present in borehole seis-
ditions must be satisfied, so we need four parameters mic surveys (§2.5.5). Stoneley-wave frequencies are far
to adjust. We take the potential functions in the form: below those used in sonic logging (§5.4.3), but they
Medium (1): fall within the frequency range of VSPs (§13.4) and
Stoneley waves constitute an important source of co-
herent noise in VSP surveys.
40f
9
/
50 f
60 • /
70f
6
80 •
6
90f
Fig. 2.15 Rayleigh waves, (a) Cross-section showing motion on
the surface and diminishing with depth for a semiinfinite solid,
(b) Predicted motion of a particle on the surface of a semiinfinite
solid, (c) Actual motion of a particle on the surface (from How-
100' *
ell, 1959). (d) Rayleigh-wave motion from a small explosion as
recorded by buried geophones; motion changes from retrograde (d)
to prograde at about 40 ft (after Dobrin, 1951).
0.5 We must have j]2 real and positive so that e 712KZ -> 0
— ,
-<X4 a s z - > +°o; therefore, VL < P 2 . However, T^ is un-
0.4 x restricted because z is finite in the upper layer.
a 0.3
1 -05
Applying the boundary conditions, we have at z =
/?/oc
\ - 06 -h, <Jyz = 0= M ^ J i ' s o dvi/dz = ° f r o m ec l- (2-9)>
fJ
0.2 that is,
\
0.1
\ 07
0.2 0.4 0.6 0.8 1.0
At z = 0, CTji = a^z|2, vt = v2, so
VELOaTYRAHO
- B) = - j
Fig. 2.16 Rayleigh-wave velocity, VR, as a function of Poisson's and
ratio, <r.
A + B = C.
By setting a = e~2T11K/l, b = M'2'VlJLi'Tli> tnese

equations
/ become
/
/ aA - B = 0,
A - B + bC = 0, (2.91)
- ,4 + ^ - C = 0.
/ For these homogeneous equations to have a nontrivial

ft- solution, the determinant of the coefficients must van-
A, _A2 = 1
ish (see §15.1.1), that is,
a -1 0
1 -1 = 0,
1 +1 -1
1 ! 1 1 t 1
1.0
Pi/Pi or
Fig. 2.17 Conditions for the existence of a Stoneley wave; solu-

-b = (1 - fl)
tions exist within the shaded area. (After Scholte, 1947.)
= tanh
But tanh x is positive for all real values of x, also TI2
2.5.4 Love waves is real and positive; therefore, T^ must be imaginary,
that is, rij = j£, where t, is real. Because tanh jx =
Love waves (Love, 1911) are surface waves consisting
of ^//-motion parallel to an interface. They exist only j tan x, we now get
when a semiinfinite medium is overlain by an upper ^ tan iKh. (2.92)
layer of finite thickness terminating at a free surface.
We take the lower interface as the xy-plane and the From eq. (2.90), we have
free surface as the parallel plane z = —h; the SH-
motion is in the j^-direction. The only component of
displacement is v, so £ = vj in eq. (2.11), hence, v satis- so that ^ > p r Thus,
fies the wave equation and we can dispense with the
potential \H °f eQ- (2.81) and use v instead.
The boundary conditions (see problem 2.10) re- and the S-wave velocity must be higher in the deeper
quire that oryz = 0 at z = -h and that cryz and v be layer than in the surface layer, VL then being in be-
continuous at z = 0; thus, we need three parameters tween the two velocities.
to adjust and so we write the following: Because K = 2TT/\ = o>/J£, as the frequency in-
Medium (1): creases from zero, tan K£/Z increases and approaches
infinity; thus for eq. (2.92) to hold, as the frequency
vx = (AQ7""2 + 2te~7)1KZ)eJK(x -h < z ^ 0. increases, £ must approach zero and VL must approach
Medium (2): P r Conversely, as K approaches zero, £ approaches its
maximum value and VL approaches P2. Hence, at high
v 2 = Ce~V2KZ&K(x ~ VLt\ 0 ^ z ^ +00. frequencies, the Love-wave velocity approaches the
velocity of S-waves in the surface layer, and as the fre-
Substituting vx and v2 in the wave equations gives
quency approaches zero, the Love-wave velocity ap-
Ti 2 = 1 - proaches the S-wave velocity in the lower layer (Do-
(2.90)
brin, 1951).
SURFACE WAVES 53
The expression for V! can be written
on using eq. (2.91). Therefore,
}«x-fr- VLt)
-**-™ (2.93)
on taking the real part of the amplitude. We see that
Vj vanishes on planes where
K£(Z + h) = (r + l/2)ir (r integral) (2.94)
(recall that h is positive and z is negative in the upper
layer); these planes are called nodal planes (see §2.1.1).
Nodal planes are characteristic of normal-mode prop-
agation (§13.3) and indeed Love waves can be ex-
plained in terms of normal-mode propagation (Grant
and West, 1965: 81-5).
Fig. 2.18 Changes involved in passage of a tube wave.
2.5.5 Tube waves The change in volume AT is due to expansion both

Waves traveling in a fluid-filled borehole or on the along the axis and radially, that is,
walls of a borehole in the direction of the axis (tube f) W
waves) are of considerable interest in velocity surveys AY = irr2 — Az + (2irn/r)Az,
dz
in wells (§5.4.2), in vertical seismic profiling (§13.4),
and in sonic logging (§5.4.3). Because they have where ur is the change in the radius of the hole. Thus,
mainly only 1 degree of freedom (along the axis), their we get
amplitude decreases slowly with distance. Sometimes
several modes of tube waves are present and often the
mechanisms of their generation and the nature of their dz
motion are not clear. Tube waves have the potential Lamb (1960: §157) derived the following relation be-
of furnishing information about the elastic properties tween ur and & for an annulus of inner and outer radii
and permeability of the surrounding formations. r and R, where E, a, and |x are respectively Young's
Most tube-wave energy travels axially, but radial modulus, Poisson's ratio, and the shear modulus for
motion is also involved in some modes. A pressure the annulus material:
geophone or one hanging freely in the borehole will
sense the maximum tube-wave effects in the borehole uL = 9 (1 + a)(R2 + r2) - 2ar2
fluid, whereas a geophone clamped to the borehole
wall will sense much smaller motion.
If we let R —»°°, we obtain for a cylindrical hole in an
The classical tube wave is merely a P-wave propa-
infinite medium
gating in the fluid, the borehole wall expanding and
contracting as the pressure wave passes. We assume a urlr = 9(1 + &)/E = 9I2\L
homogeneous fluid in a cylindrical borehole penetrat-
ing a homogeneous isotropic medium (fig. 2.18). Us- (using problem 2.2). Substitution in eq. (2.96) gives
ing 9 for the pressure and w for the displacement,
9(
Newton's second law, net force = mass X acceleration, \k + J ~ dz'
applied to a volume element of the fluid, Y =
Ttr2 Az, is and substitution of this result in eq. (2.95) gives the
2
wave equation:
dw
— Azjirr 2 = -(PTrr2Az)
^ 22 = (M
2
^, KT2 = If 1 - f i r (2.97)
V
d
dz \V J df p \ k )
or
White (1965: 153-6; 1983: 139-91) discusses tube
39 _ d2w waves in greater detail.
(2.95)
Tz~ ~9Jt2' Cheng and Toksoz (1981) discuss two other tube-
From eq. (2.18), wave modes. One is a Stoneley wave (§2.5.3) propagat-
ing along the borehole wall and dying away exponen-
9 = -kA = -kAY/Y. tially in the formation surrounding the borehole; this
is the dominant tube-wave mode in VSP work. The

other tube-wave mode is pseudo-Rayleigh waves,
guided waves (§13.3) confined largely to the fluid, also
dying away exponentially in the surrounding forma-
tion. Both waves are dispersive (§2.7.4).
Cheng and Toksoz calculated dispersion curves for
both modes (fig. 2.19a). The Stoneley wave is slightly
dispersive with both group and phase velocities close
to 0.90^, where af is the P-wave velocity in the bore-
hole fluid (see eq. (2.61)). Pseudo-Rayleigh waves can-
not exist below a minimum frequency (where their ve-
locity equals the S-wave velocity of the surrounding
rock, (3r) and their group velocity passes through a
O7
minimum, which results in an Airy phase (see §13.3 o 10 20 30
and fig. 13.19). Several modes may exist (see eq. FREQUENCY (kHz)
(13.1)). Pseudo-Rayleigh waves are not a factor in or- (a)
dinary seismic work (fig. 2.19a shows a 10-kHz low-
frequency cutoff), but they are involved in sonic log-
ging. At higher frequencies, the velocities of both Sto-
neley and pseudo-Rayleigh waves approach the S-
wave velocity in the medium surrounding the
borehole.
Cheng and Toksoz calculated "synthetic micro-
seismograms" for various circumstances, one of which
is shown in fig. 2.19b; fig. 2.19c shows an observed TIME
waveform.
Hardage (1985) discusses the role of tube waves in
VSP surveys. Figure 2.20a shows prograde elliptical
motion in an axial plane. The radial motion is zero at
the center of the hole and maximum at the borehole
wall, where it is continuous (fig. 2.20b), but it decays
rapidly in the surrounding formation. The axial com-
ponent of motion is relatively constant in the fluid but
is discontinuous at the borehole wall where its ampli-
tude decreases by a factor as large as several hundred.
This explains why geophones should be clamped to
the borehole wall.
Tube waves are reflected at impedance changes, just
as other acoustic waves are (§3.2). When the borehole
cross-sectional area changes from ax to <z2, the reflec-
tion (R) and transmission (7) coefficients are (Har-
dage, 1985: 86-7)
2a,
T= (2.98)
a2 + a,
(compare with eqs. (3.14) and (3.15)). At the top of (C)
the borehole fluid and the bottom of the hole, R =
— 1 and +1, respectively. Tube waves are also reflected
at a geophone sonde and where casing changes. Fig-
ure 2.21 shows several reflected tube waves. Fig. 2.19 Stoneley and pseudo-Rayleigh waves in a fluid-filled
Tube waves can be generated by almost anything borehole. (From Cheng and Toksoz, 1981: 1045.) (a) Dispersion
that disturbs the borehole fluid. The most common curves for the Stoneley wave and the first two pseudo-Rayleigh
source is a Rayleigh wave passing over the top of the modes are shown for $Ja.f = 1 . 5 , where £ r is the S-wave velocity
borehole; thus, tube waves are uncommon in marine in the surrounding rock, and af is the P-wave velocity in the
borehole fluid (velocities normalized by dividing by the fluid ve-
VSP surveys and, in land surveys, lowering the bore-
locity), (b) Calculated signature for a broadband source show-
hole fluid level often lessens tube-wave generation. ing the pseudo-Rayleigh wave, Stoneley wave, and Airy phase;
Tube waves initially have the same spectrum as the a, = 5.94 km/s, Pr = 3.05 km/s, Pr = 2.30 g/cm3, a 7 = 1 . 8 3
generating source and their spectrum changes slowly km/s, pf = 1.20 g/cm3, hole radius = 10.2 cm (after Hardage,
because there is little absorption in the borehole fluid. 1985: 75). (c) Observed tube-wave signature.
ANISOTROPIC MEDIA 55
variation of seismic velocity with the direction in

DOWNGCMNG
TUBE WAVE which it is measured or with wave polarization
ELASTIC ROCK (§13.1.6). The general elasticity matrix relating stress
MATERIAL ov to strain ekl (the generalized form of the 6 X 6 ma-
trix in eq. (2.15)) can contain at most 21 independent
PARTICLE MOTION
constants because of symmetry considerations, but
Winterstein (1990: 1084-5) says that only 18 of these
can be truly independent. The number of independent
constants depends on the symmetry of the system
(Love, 1944: 99).
A number of different types of symmetry (symmetry
systems) can exist. Classically, eight systems are de-
fined (Love, loc. cit.; Landau and Lifshitz, 1986;
Saada, 1974), but some writers define subsystems as
well; for example, Winterstein (1990: 1083-5) lists 11
systems plus subsystems in discussing cracks. Aniso-
tropy types are associated with the symmetry systems.
At seismic wavelengths, however, the only anisotropy
types reported are transverse isotropy (hexagonal
symmetry), orthorhombic anisotropy, and mono-
clinic anisotropy.
Transverse isotropy involves elastic properties that
mm RADIAL DISPLACEMENT
are the same in any direction perpendicular to an axis
but are different parallel to this axis. Two important
types of transverse isotropy are observed: that with a
nearly vertical symmetry axis (thin-layer anisotropy)
and that with a nearly horizontal axis (azimuthal an-
isotropy) (Bush and Crampin, 1987). Transverse iso-
tropy is the most important type of anisotropy en-
countered; it is discussed further in §2.6.2.
Orthorhombic anisotropy is equivalent to a superpo-
sition of thin-layer anisotropy and azimuthal aniso-
SEDIMENT S^=5AXIAL DISPLACEMEN tropy. It arises because a vertical fracture system has
SECONDS
DOUBLE
CASING
500
IOOO
Fig. 2.20 Wave motion for a tube wave. (From Hardage, 1985: 3 1500
78, 79.) (a) Prograde elliptical motion of fluid particles during
passage of a tube wave (ellipticity is greater than shown here),
(b) Axial and radial displacements for hard formation, v = 82 2000
Hz; and (c) for soft formation, v = 74 Hz.
2500
Consequently, tube waves often have appreciable en-
ergy in the signal range even after considerable travel. CEMENT
2.6 Anisotropic media

2.6.1 Types of aniso tropy Fig. 2.21 VSP record showing several tube waves. Tube wave
(1) is generated at the base of the surface casing; (2) is generated
Anisotropy is a general term denoting variation of a at the surface by a Rayleigh wave; (3) is a reverberation of wave
physical property depending on the direction in which (2) between the well sonde and the surface; and (4) is a reflection
it is measured. Seismic anisotropy is evidenced by a from the bottom of the borehole. (From Hardage, 1985: 88.)
been superimposed on a horizontally layered system. refringence (Crampin, 1981). For horizontal thin-layer
VSP data from the Paris Basin have been interpreted anisotropy, the two waves are the qSP-w&ves (that is,
using an orthorhombic model (Bush and Crampin, quasi-S-waves having displacement parallel to the
1987; MacBeth, 1990). Layering anisotropy is usually symmetry axis) and &K-waves (displacement in radial
much stronger than fracture anisotropy so that the directions). For azimuthal asymmetry, they are some-
overall effect may be difficult to distinguish from thin- times called qSV- and SH-waves.
layer anisotropy. Monoclinic anisotropy can be pro- In anisotropic media, pure S- and P-waves may ex-
duced by superimposing tilted fractures on a layered ist only in certain directions. In transversely isotropic
medium (Schoenberg and Muir, 1989). Examples of media, SV- and P-modes of propagation are coupled
monoclinic anisotropy have been observed in the field (see §2.6.3). Wavefronts are not in general orthogonal
(Crampin, McGonigle, and Bamford, 1980; Win- to the directions of wave propagation. Phase velocity
terstein and Meadows, 1990). is velocity perpendicular to a surface of constant
The stress-strain relationships require 5 indepen- phase (a wavefront), and group velocity the velocity
dent elastic moduli for transverse isotropy, 9 for or- with which the energy travels (§2.7.4), is in a different
thorhombic anisotropy, and 13 for monoclinic anisot- direction (see fig. 2.23). The surfaces for SV-
ropy, compared with only 2 for the isotropic case. wavefronts may have cusps.
Anisotropy is often described by the fractional
2.6.2 Transverse isotropy difference between the maximum and minimum veloc-
ities for a given wave surface, i.e., (J£ax - ^ n ) / ^ a x ,
Taking the z-axis as the axis of symmetry, Love (1944: sometimes by the ratio of maximum and minimum ve-
160-1) showed that for transverse isotropy, Hooke's locities, VmJV^n.
law reduces to the following: Uhrig and van Melle (1955) give a table showing
anisotropy values of 1.2 to 1.4 for rocks at the surface
and 1.1 to 1.2 for sediments at depths of 2.1 to 2.4 km
(2.99) in west and central Texas. Stoep (1966) found average
values between 1.00 and 1.03 for Texas Gulf Coast
sediments. Segonzac and Laherrere (1959) obtained
values from 1.00 for sandstones to 1.08 to 1.12 for
limestones and 1.15 to 1.20 for anhydrites from the
(2.100)
northern Sahara.
where thefiveindependent constants are \ and JJLIP k± 2.6.3 Wave equation for transversely isotropic media
and |JL±, and jx*.
Layering and parallel fracturing tend to produce When media are not isotropic, the mathematics be-
transverse isotropy. A sequence of isotropic layers come more complex the more anisotropic the me-
(such as sedimentary bedding) produces thin-layer an- dium. However, the case of a transversely isotropic
isotropy for wavelengths appreciably larger than the medium can be treated without great difficulty. We
layer thicknesses (k > $d, where d is layer thickness; consider waves in the xz-plane, where the symmetry
see Ebrom et al., 1990). The symmetry axis is perpen- axis is along the z-axis. Derivatives with respect to y
dicular to the bedding with the velocities of P- and S- are zero, but 5-waves may involve motion in the y-
waves that involve motion parallel to the bedding direction. We substitute eqs. (2.99) and (2.100) into
larger than those involving motion perpendicular to eq. (2.24), and using eqs. (2.8) and (2.7), we get the
the bedding. The velocity parallel to the bedding is wave equations for transversely isotropic media:
greater because the higher-velocity members carry the
d2u _ dax dvxz
first energy, whereas for wave motion perpendicular
to the bedding, each member contributes in propor- dt2 dx dz
tion to the time taken to traverse it.
±
Nonhorizontal fracturing and microcracks produce dx dx dz\ dz r \dz dx)\
azimuthal anisotropy with a symmetry axis perpen-
dicular to the fracturing (fractures often are some- 11
dx2 dz2 dz dx
9
what parallel and vertical). The velocity of waves that
involve motion parallel to the fracturing (S{) is larger (2.101)
than that of waves with motion perpendicular to the d2v d2v . d2v
P = h a* , (2.102)
fracturing (S2). If the motion is neither parallel nor dt2 dx2 dz2
perpendicular to the fracturing, an S-wave splits into
two waves with orthogonal polarizations (fig. 2.22):
one (S(l) traveling at the Sx velocity, the other (SJ at
the S2 velocity; this is called shear-wave splitting or bi- (2.103)
EFFECTS OF THE MEDIUM ON WAVE PROPAGATION 57
these become
(V2 - a2€2 - a*2n2)£A

- [V2 - (a2 - a*2 + p*2)€2 - $*2n2]nB = 0,
(V2 - a2±n2 - a*2e2)nA
+ [V2 - (a2 - a*2 + p*2)«2 - p*2€2]€£ = 0.
Eliminating A and 2? gives the following quadratic
equation in V2:
Sj_: N 45° E [V2 - (a2 - a*2 + p*2)€2 - p*2«2]«

AT
S|| : N 45° W (V2 - a2€2 - a*2«2)€

2 2 2 2 2 2 2
_ - [V - (a , - a* + P* )M - p* l ]l
Fig. 2.22 S-wave propagation in a fractured medium with (V2 - a\n2 - a*2€2)«
cracks oriented N45°W. For an S-wave traveling parallel to frac-
turing, the velocity (S2) is slower for a component involving mo- or
tion perpendicular to the fracturing than for one involving mo-
tion parallel to the fracturing (velocity S{). [V2 - ( a 2 - a* 2 + p* 2 )€ 2 - p* 2 « 2 ]
Note the dependence in eqs. (2.101) and (2.103) on X (V2 - a2±n2 -

derivatives of both u and w; P- and S V-waves are said
to be coupled. The S//-wave governed by eq. (2.102) + [V2 - (a2 - a*2 + p* 2 > 2 - p*2€2]
is, however, independent of the other two. X(V2 - a2e2 - a*2n2)t2 = 0. (2.104)
We simplify the problem by assuming a plane wave
traveling in the xz-plane in the direction of increasing The solution has been given by Stoneley (1949), Grant
x and decreasing z, the angle between the raypath and and West (1965: 42), and White (1965: 46). The roots
the x-axis being 0. We now use the potential functions are always real and positive and approach a and p of
of eq. (2.64) in the form eqs. (2.58) and (2.59) as the anisotropy approaches
zero. When the wave is traveling vertically, € = 0, n =
x = 1, and V = a± or p* for vertically traveling P- or SH-
waves. When € = 1, n = 0, V = ay or p*, correspond-
ing to horizontally traveling P- or >S7/-waves. How-
where ever, when the wave is traveling at an angle to the ver-
tical, the roots are complicated functions of the elastic
£ = ( € * - nz)IV- t, = cos 6, n = sin i
constants and the motion is not separated into distinct
Then P- and S-waves.
_ d§ dx _ /jw 2.7 Effects of the medium on wave propagation

dx dz
2.7.1 Energy density and geometrical spreading
w = — - — = -\ — \(nA + Probably the single most important feature of any
dz dx wave is the energy associated with the motion of the
When we substitute these into eqs. (2.101) and (2.103), medium as the wave passes through it. Usually, we are
the following factors appear in every term and hence not concerned with the total energy of a wave but
can be ignored: jco/K (j<*>)2, and eK Equations (2.101) rather with the energy in the vicinity of the point
and (2.103) become where we observe it; the energy density is the energy
per unit volume.
- (X,, + 2^,,)€3 - (X± + 2\L*)€n2]A Consider a spherical harmonic P-wave for which
K - (X,, + 2^, - X± - \x*)Z2n - \x*rt]B = 0, the radial displacement for a fixed value of r is given
[9V2n - (X± + 2p,> 3 - (X± + 2^)£2n}A by
(JJL* - 2^J£n2 - ^*e3]B = 0.
u = A cos (a>t + 7),
Writing
where 7 is a phase angle. The displacement u ranges
ai = (X±
from - A to +A. Because displacement varies with
time, each element of the medium has a velocity, u =
Group velocity
Phase velocity
Wavefront at time t
Q Isotropic
Fast direction
Wavefront from P
Slow direction
Reflector Q Anisotropic
Velocity function
(a) (b)
-4000 -2000 2000 4000
Nonelliptical
Green River shale
(c)
Fig. 2.23 Wavefronts in anisotropic media, (a) Application of constant phase for a point source) in transversely isotropic me-
Huygens' principle to an anisotropic medium illustrates direc- dia are elliptical; however, P- and S V-wavefronts are not ellip-
tion and magnitude differences between phase and group veloci- tical except in special instances. VH > Vv with vertical axis of
ties, (b) Fermat's principle applied to a reflection for a coinci- symmetry. V^ is group velocity as a function of the angle with
dent source and receiver shows that a reflection may not occur the symmetry axis.
at a right angle to the reflector, (c) 5//-wavefronts (surfaces of
duldt, and an associated kinetic energy. The kinetic en- the energy is converted back and forth from kinetic to
ergy bEk contained within each element of volume potential forms, the total energy remaining fixed.
oTis When a particle is at zero displacement, the potential
energy is zero and the kinetic energy is a maximum,
and when the particle is at its extreme displacement,
the energy is all potential. Because the total energy
The kinetic energy per unit volume is equals the maximum value of the kinetic energy, the
energy density E for a harmonic wave is
E = ipco2A2 = 2ir2pv2 A2. (2.105)
This expression varies from zero to a maximum of Thus, we see that the energy density is proportional
to the first power of the density of the medium and to
The wave also involves potential energy resulting the second power of the frequency and amplitude of
from the elastic strains created during the passage of the wave. (See Braddick, 1965, for a different deriva-
the wave. As the medium oscillates back and forth, tion of eq. (2.105).)
E F F E C T S OF THE M E D I U M ON WAVE PROPAGATION 59
We are also interested in the rate of flow of energy

and we define the intensity as the quantity of energy
that flows through a unit area normal to the direction
of wave propagation in unit time. Take a cylinder of
infinitesimal cross-section, 8^, whose axis is parallel
to the direction of propagation and whose length is
equal to the distance traveled in the time, St. The total
energy inside the cylinder at any instant t is EVbt SSf;
at time / + bt all of this energy has left the cylinder
through one of the ends. Dividing by the area of the
end of the cylinder, 85^, and by the time interval, o7,
we get /, the amount of energy passing through unit
area in unit time:
Fig. 2.24 Dependence of intensity upon distance.
I=EV. (2.106)
For a harmonic wave, this becomes
/ = ipFa)2A2 = 2TT2PFV2A2. (2.107) The foregoing assumes constant velocity, whereas
velocity usually increases with depth, producing more
In fig. 2.24, we show a spherical wavefront diverg- rapid spreading. A factor of V2st is often used (§9.8),
ing from a center O. By drawing sufficient radii, we where Vs is the stacking velocity (§5.4.1). The term
can define two portions of wavefronts, $fx and £f2, of "spherical divergence" is still used in this situation
radii rx and r2, such that the energy that flows outward even though wavefronts may not be spherical.
through the spherical cap &)l in 1 second must be
equal to that passing outward through the spherical
cap ^ 2 in 1 second (because the energy is moving only 2.7.2 Absorption
in the radial direction). The flow of energy per second (a) General In the preceding section, we considered
is the product of the intensity and the area; hence, variations of the energy distribution as a function of
geometry. Implicit in the discussion was the assump-
tion that none of the wave energy was transformed
Because the areas SP, and £f2 are proportional to the into other forms. In reality, as the wave motion passes
square of their radii, we get through the medium, the elastic energy associated
with the wave motion is gradually absorbed by the me-
IJI, = SP,/£f2 = (r,/r2)2.
dium, reappearing ultimately in the form of heat. This
Moreover, it follows from eq. (2.106) that E is propor- process is called absorption and is responsible for the
tional to / and hence eventual complete disappearance of the wave motion
(see also §6.5). Toksoz and Johnston (1981) summa-
= EJEX = {rxlr2)\ (2.108)
rize much of the literature regarding absorption.
Thus, geometrical spreading causes the intensity and The measurement of absorption is very difficult,
the energy density of spherical waves to decrease in- mainly because it is not easy to isolate absorption
versely as the square of the distance from the source from other effects making up attenuation (see §6.5.2).
(Newman, 1973). This is called spherical divergence. Moreover, absorption varies with frequency, so that
For a plane wave, the rays do not diverge and hence it is not clear how laboratory measurements apply to
the intensity of a plane wave is constant. Figure 2.24 seismic wave travel in the earth.
could represent a cross-section of a cylindrical wave,
that is, a wave generated by a very long linear source, (b) Expressions for absorption. The decrease of am-
arcs S?l and ^ 2 being cylindrical wavefronts. Because plitude due to absorption appears to be exponential
the arcs are proportional to the radii, cylindrical diver- with distance for elastic waves in rocks. Thus, we can
gence causes the intensity to vary inversely as the ra- write for the decrease in amplitude because of ab-
dius. Thus, we can write sorption
= EJEX = (2.109) A = Aoe-v, (2.110)
where m = 0, 1, or 2 according as the wave is plane, where A and Ao are values of the amplitudes of a plane
cylindrical, or spherical. wavefront at two points a distance x apart, and r\ is
Ratios of intensity, energy, or power are usually ex- the absorption coefficient.
pressed in decibels, the value in dB being 10 log10 of Other measures of absorption are based on the de-
the intensity, energy, or power ratio. Because these crease in amplitude with time; to relate these to in, we
vary as the square of the amplitude, dB is also given assume a cyclic waveform:
as 20 log10 of the amplitude ratio. The natural log of
A = A0Q~ht cos 2irW, (2.111)
the amplitude ratio (in nepers) is also used (see prob-
lem 2.17). and make measurements at a fixed location; h is called
the damping factor. The logarithmic decrement {log dec) where xs is the distance to the source. The table shows
8 is defined by that losses by spreading are more important than
losses by absorption for low frequencies and short dis-
amplitude tances from the source. As the frequency and distance
5= )• (2-112)
amplitude 1 cycle later increase, absorption losses increase and eventually be-
It can be expressed in terms of the damping factor as come dominant.
The increased absorption at higher frequencies re-
8 = hT= h/v = 2lrh/u, (2.113) sults in change of waveshape with distance. Peg-leg
where T is the period; 8 is measured in nepers. Quality multiples (§6.3.2b) and possibly other mechanisms
factor Q can be defined as also produce waveshape changes. Figure 2.25 shows
the energy decreasing with distance and with fre-
Q = 27r/(fraction of energy lost per cycle) quency; the frequency-dependent attenuation is
(2.114) greater than expected from absorption alone.
where Ais is energy loss. Because energy is propor-
tional to amplitude squared, E = E0e~2ht and AE/E0 = 2.7.4 Dispersion; group velocity
2k A*. Setting At = T, we get AE/E0 = 2hT = 28 and
Velocities V, a, and p, which appear in §2.2 and subse-
Q = n/hT= TT/8. (2.115) quent sections, are phase velocities because they are
If n is the number of oscillations for the amplitude to the distances traveled per unit time by a point of con-
decrease by the factor e, then &nT = e, n = UhT, and stant phase (for example, a trough) of a simple wave
such as those in eqs. (2.53) to (2.55). This is not neces-
Q = irn. (2.116) sarily the velocity with which a pulse travels, called
Still another manner of expressing Q is Q = cot 4>, the group velocity. For the wavetrain shown in fig.
where of) is the loss angle. 2.26a, we can determine the group velocity Uby draw-
During one period, a wave travels one wavelength ing the envelope of the pulse (the double curve ABC,
so that if the loss of energy is due to absorption only, AB'C) and measuring the distance that the envelope
the attenuation factor is hT = y\\ (from eqs. (2.110) travels in unit time. The phase velocity V is given ap-
and (2.111)), and we can interrelate in, 8, and Q: proximately by the rate of advance of a distinct "phase
break," as indicated in figs. 2.26a and 2.26b, but to
Q = T^X = == T T / 8 . (2.117) find V accurately, we should decompose the pulse into
Absorption in the form given by eq. (2.110) appears its frequency components by Fourier analysis (see eq.
naturally in solutions of the type given in eq. (2.56) if (15.113)) and measure the speed of each component
we permit the elastic constants to be complex num-
bers. Real elastic constant values correspond to media Attenuation (dB)
without absorption and complex values imply expo- 18 27 36
nential absorption. Complex values of X and JUL result
in complex velocity values (see eqs. (2.58) and (2.59)).
If the \IV in eq. (2.56) is replaced with \IV +
then
0.5
which agrees with eq. (2.110).
2.7.3 Relative importance of absorption and spreading

To compare the loss by absorption with the loss of
intensity by geometrical spreading (see eq. (2.108)),
we have calculated the losses in going various dis-
1.5
tances from a point 200 m from the source assuming Divergence
j] = 0.15 dB/X. The results shown in table 2.1 were Divergence plus
calculated using the following relations: transmission loss 20 Hz 40 Hz 60 Hz
Absorption: 2.0
intensity loss in dB= 10 log10 (I0II)
= 20 log10 (AJA) Fig. 2.25 Loss of amplitude as a function of one-way trav-
= 0.3(x/X) = 0.3(x, - 200)/X eltime, based on measurements with a geophone clamped in
= 0.3v(x, - 200)/2000, borehole with the source at the surface. The curves labeled "di-
Spreading: vergence" and "divergence plus transmission loss" are calcu-
lated from sonic-log data allowing for loss of energy in transmis-
intensity loss in dB= 10 log10 (Io/I) sion through reflecting interfaces. The 20-, 40-, 60-Hz curves
= 20 log10 (x,/200), show attenuation at those frequencies. (Courtesy of SSC.)
EFFECTS OF THE MEDIUM ON WAVE PROPAGATION 61
Table 2.1 Energy losses by absorption and spreading frequency, we have inverse dispersion and the opposite
(j) = 0.15 dB/wavelength and V = 2.0 kmls) is true (as in fig. 2.26a).
Dispersion of body waves is a consequence of most
Distance from shotpoint (xs) theories proposed to account for absorption. Aki and
Frequency Richards (1980: 170-2) show that the assumption of
(") 1200 m 2200 m 4200 m 8200 m constant Q, which most data indicate is the situation
in solid earth materials, and no dispersion are incon-
Absorption lHz 0.075 dB 0.15 dB 0.3 dB 0.6 dB
sistent because they lead to noncausality. Their argu-
3 0.22 0.45 0.9 1.8
10 0.75 1.5 3 6
ment follows. Starting with a plane-wave impulse at
30 2.2 4.5 9 18
x = 0 = t and using eqs. (15.127) and (15.136), we
100 7.5 15 30 60 have
Spreading All 16 21 26 32
I|I(JC, 0 = §(t - x/K) <-> e-***":
Adding attenuation corresponding to constant Q
If phase velocity V is the same for all frequencies in (see eqs. (2.110) and (2117)), the right side becomes
the pulse, the pulse shape does not change and U =
e(-j<ox/K-M*/2KG). w e m u s t u s e th e absolute value of co to
V. However, if the velocity varies with frequency, the avoid having the amplitude increase with increasing x
different components travel with different speeds, the when co is negative). Using eq. (15.109) to revert to the
pulse shape changes, and U ^ V, that is, the medium time domain, we have
is dispersive. Consider the two harmonic waves shown
in fig. 2.26c that travel at slightly different velocities,
the higher-frequency one being faster than the lower- ty(Xt f) = Q-aKjx/V-x/2VQ)Qjw
frequency one. The "pulse" that results from their in-
+
1 e "*"*"'° w " d "
terference changes waveshape and travels at a velocity
different from either of them (greater than either in
this instance). If we write A((s))&{KX~(Ot) for a plane-wave
component traveling along the x-axis, K being K(OO), in
general, ^4(co) varies slowly whereas the phase (KJC — QO>(-A+JB)
oit) varies rapidly. When adjacent frequency compo-

nents are added together, the net result is usually ap- where A = x/2VQ, and B = (t - xlV). Integration
proximately zero because of destructive interference. gives
However, when the phase (KX — to/) varies slowly, con-
structive interference occurs in the vicinity of a point B2)
(x, t). The condition for this is X/2TTVQ
(2.119)
(x/2VQf + (t- xlVf
[ This function is shown in fig. 2.27; it has a maximum

dto value at t = x/Fand is roughly symmetrical. Observa-
Point (x, t) will move with velocity U, where U = tions show instead a sharper rise near x/F followed by
dx/dt (see §2.2.5). Hence, differentiation of the above a slower decay. The function of eq. (2.119) has a finite
equation gives value for / < 0, x # 0, and is therefore noncausal,
showing that our assumptions are inconsistent. We
u=dx= TdK^r1 = |d[ a) IK 1 conclude, therefore, that absorption necessarily re-
quires that V varies with co, that is, dispersion must
dt I dw J lda>|.F(a))Jj exist.
J co dF(co) Dispersion is important for several reasons, per-
LF(co) F^CO) dco haps the most important being that the energy of a
The derivative is small, so pulse travels with the velocity U (except where there is
appreciable absorption; see Brillouin, 1960: 98-100).
U~v+Ji=v+V^.
Dispersion of seismic body waves has not been defini-
d\ tively observed over the wide range of frequencies
dco dv
(2.118) from hertz to megahertz. Most rocks simply exhibit
little variation of velocity with frequency in the seis-
where V, to, v, X, dF/dco, dK/dv, and dVld\ are average mic frequency range. Ward and Hewitt (1977) found
values for the range of frequencies making up the the same velocity at 35 Hz as at 55 Hz in a mono-
principal part of the pulse. (See problem 2.7 for a frequency well survey to about 800 m. Futterman
more elementary derivation of eq. (2.118).) (1962) shows that the dispersion expected for seismic
When V decreases with frequency, we have normal body waves is small for usual situations. Dispersion is,
dispersion, and V is larger than U, where the envelope however, important in connection with surface waves
travels slower than the individual cycles; this is the (see §2.5) and channel waves (§13.3) as well as other
usual case with ground roll. When V increases with phenomena.
62
THEORY OF SEISMIC WAVES
AAAAA/VWWWWW
4
WVWvAAAAAAAAAAA
WWÂ!A/VWWW
Fig. 2.26 Comparison of group and phase velocities, (a) Defi- right form beats; the envelope travels with the group velocity
nition of group velocity U and phase velocity V; (b) arrival of a U = AW/AK and points of constant phase within the beats with
dispersive wave at different geophones; (c) two sine waves of the phase velocities Vx = <O,/K, and V2 = O)2/K2 (from Gerkens
slightly different frequency and velocity traveling from left to 1989: 37).
2.7.5 Reflection and refraction; Snell's law

downward a distance V2 At. By drawing arcs with cen-
Whenever a wave encounters an abrupt change in the ter A' and lengths equal to Vx At and V2 At, and then
elastic properties, as when it arrives at a surface sepa- drawing the tangents from R to these arcs, we locate
rating two beds, part of the energy is reflected and re- the new wavefronts, RS and RT in the upper and
mains in the same medium as the original energy; the lower media. The angle at S is a right angle and A'S =
balance of the energy is refracted into the other me- Vx At = B'R; therefore, the triangles A'B'R and A'SR
dium with an abrupt change in the direction of propa- are equal, with the result that the angle of incidence Q{
gation occurring at the interface. Reflection and re- is equal to the angle of reflection OJ; this is the law of
fraction are fundamental in exploration seismology reflection. For the refracted wave, the angle at T is a
and we shall discuss these in some detail. right angle and we have
We can derive the familiar laws of reflection and
V2 At = A'R sin 6,
refraction using Huygens' principle. Consider a plane
wavefront AB incident on a plane interface as in fig. and
2.28 (if the wavefront is curved, we merely take A and
VxAt = A'R sin 6 P
B sufficiently close together that AB is a plane to the
required degree of accuracy; however, see also §2.8.1); hence,
AB occupies the position A'B' when A arrives at the
surface; at this instant, the energy at Br still must sin sin 8
2 (2.120)
travel the distance B'R before arriving at the interface. = P-
If B'R = Vx At, then A* is the time interval between Angle 02 is called the angle of refraction and eq. (2.120)
the arrival of the energy at A' and at R. By Huygens' is the law of refraction, also known as Snell's law. The
principle, during time At, the energy that reached A' angles are usually measured between the raypaths and
will have traveled either upward a distance Vx At or a normal to the interface, but these angles are the
DIFFRACTION 63
sin 8C = VJV2. (2.121)

For angles of incidence greater than 6c, it is impossible
to satisfy Snell's law (using real angles) because sin 62
cannot exceed unity and total reflection occurs. This
does not mean that 100% of the energy is reflected
as P-waves, however, because converted S-waves (see
§3.1.1) and evanescent waves (see §2.3.1) are gen-
erated.
Noting the method used to derive eq. (2.57) in
§2.3.1, we write Snell's law for the case 0! > 9C (see fig.
2.29) in the form
Fig. 2.27 Illustrating eq. (2.119). sin 92 = (V2IVX) sin §x = sin (iir - j8)
= cos j9 = cosh 0 = €,
same as those between the interface and the wave- n = cos 02 = sin j0 = j sinh 0;
fronts in isotropic media. The laws of reflection and hence, eq. (2.56) becomes
refraction can be combined in the single statement: at
\U = y^Q-(oiz/V)sinhQQJ<x}[(x/V)coshe-t] /O 192^
an interface, the quantity/? = (sin ^)IVi has the same
value for the incident, reflected, and refracted waves. If we take 0 negative in fig. 2.29, the only change is in
The quantity p is called the raypath parameter. It will the sign of the first exponential on the right-hand side.
be shown in §3.1.1 that Snell's law also holds for wave Thus, just as in the case of eq. (2.57), evanescent
conversion from P- to S-waves (and vice versa) upon waves can exist on both sides of the interface and their
reflection or refraction. The generalized form of amplitudes decrease as we go away from the interface.
Snell's law (eq. (3.1)) will be understood in future ref- The rate of attenuation is proportional to sinh 0,
erences to Snell's law. which has its maximum value at the grazing angle,
When the medium consists of a number of parallel 0! = ^rr. The introduction of imaginary angles to
beds, Snell's law requires that the quantity p have the satisfy Snell's law for angles exceeding the critical
same value everywhere for all reflected and refracted angle means that the reflection coefficient (§3.2) will
rays resulting from a given initial ray. be complex and phase shifts will occur (see problem
The foregoing derivation assumed a planar surface 3.6b) that will be complicated functions of the angle
and therefore specular reflection. If the surface in- of incidence.
cludes bumps of height d, reflected waves from them Snell's law is very useful in determining raypaths
will be ahead of those from the rest of the surface by and arrival times and in deriving reflector position
Id. These can be neglected where 2d/\ < lA (the "Ray- from observed arrival times, but it does not give infor-
leigh" criterion), i.e., when d < X/8. Most interfaces mation about the amplitudes of the reflected and
satisfy this criterion for ordinary seismic waves. For transmitted waves. This subject is taken up in chap. 3.
oblique reflection, the criteria are less stringent and
reflection can be regarded as specular from relatively 2.8 Diffraction
rough surfaces.
When V2 is less than Vv 02 is less than 6,. However, 2.8.1 Basic formulas
when V2 is greater than Vv 02 reaches 90° when dx = In discussing reflection and refraction, we stated that
sin-l(Vl/V2). For this value of 0 P the refracted ray is when an interface is curved we merely have to select a
traveling along the interface. The angle of incidence portion sufficiently small that it can be considered a
for which 02 = 90° is the critical angle, 0c; obviously, plane. However, such a simplification is not always
possible, for example, when the radius of curvature of
an interface is less than a few wavelengths or the re-
flector is terminated by a fault, pinchout, unconfor-
mity, etc. In such cases, the simple laws of reflection
and refraction are no longer adequate because the en-
ergy is diffracted rather than reflected or refracted. Be-
cause seismic wavelengths are often 100 m or more,
many geologic features give rise to diffractions.
The mathematical treatment of diffraction is com-
plex and we shall give only a brief summary of a sim-
plified treatment due to Trorey (1970). We shall as-
sume a coincident source and receiver (see Trorey,
1977, for the noncoincident case) and constant veloc-
ity. Using <$> in place of i|i, we take the Laplace trans-
Fig. 2.28 Reflection and refraction of a plane wave. form of the wave equation, eq. (2.30), obtaining
assume a constant reflection coefficient over if, so that

c in eq. (2.123) is constant. Clearly, Mr = 0 = <l> over
the hemisphere, hence the contribution to the integral
over the hemisphere is zero. We can also set 4> = 0
over the portions of the plane z = h except for the
portion if whose effect we wish to evaluate.
We now substitute eq. (2.123) in eq. (2.124), noting
that r in eq. (2.123) is now r0 infig.2.30a because the
source is now at the image point Of; hence
dr z h
5
dz r r
m h
dz r0 dt) r3'
= d®dr» = c
Fig. 2.29 Imaginary angles of reflection and refraction.
dr\ dr0 dr\ V J\ r0
V2c|> = V 2 d> = = 4
where <I>(x, y z, s) is the the Laplace transform of V
where we have set r0 = r after differentiating. The re-
<$>(x, y, z, i) (see §15.3), and the double-headed arrow sult is
indicates equivalence in different domains. Note that
we are assuming that $ and d$/dt are zero at t = 0 for
2<rr<l> = ch\\ - + —\ dif. (2.125)
all x, y z.
The solution of this equation for a point source at
the origin is This surface integral can be transformed into a con-
tour integral as follows. Infig.2.30b, the element of
= (c/r)Q~sr/v, (2.123)
area is p dp d0 in polar coordinates; because r2 = p2
where r is the distance from the source to the point of + h2, p dp = r dr, and, therefore,
observation, and V is the wave velocity. (This can be
verified by direct substitution, noting that r2 = x2 + 2TT<D = ch f f e - 2 " " ^ 3 + —\dr d0. (2. 126)
y2 + z2, dr/dx = x/r, etc.) In general, c should include )Jr \r Vr2)
the Laplace transform of the input waveform at the
source, but in effect we have taken the transform to be If we integrate the first term by parts with respect to
unity, that is, the source is c 8(0 (eq. (15.180)). The r, we obtain
results for other types of sources can be found by
time-domain convolution (see eqs. (15.201) and
(15.202)).
In a source-free region, the P-wave potential func-
tion 4> is given by eq. (2.42) with Y = 0; hence, we can Substituting in eq. (2.126), we get
get another expression for $ by taking the Laplace
transform of eq. (2.42), the result being 4TT<I> = chS) \(l/r2)Q-2sr^ - (l/r2)e-2sr2/v] dB.
(2.127)
.rVd'X] di
(2.124) If the z-axis does not cut if, we can get r = % for points
sr/v
on the boundary, giving
The factor e~ arises because $ in the integrand of
eq. (2.42) is evaluated at time t = t0 - rIV, whereas O fi (2.128)
is the transform of $(x, y z, 0 (see eq. (15.187)).
where A traverses the boundary of if in the counter-
2.8.2 Diffraction effect of part of a plane reflector clockwise direction.
We shall now calculate the diffraction effect of an area If the z-axis cuts if (seefig.2.30c), r{ = h = constant
if that is part of a plane reflector z = h (seefig.2.30a), ineq. (2.127), so that
both source and detector being at the origin. We en-
close the origin with a hemisphere of infinite radius = (cl2h)Q-2sh/v -
with center (0, 0, /*), the base of which is the plane z = (2.129)
h. In order to apply eq. (2.124), we replace the source
with its image at (0, 0, 2/z), thus making the hemi- where A again traverses the boundary in the counter-
sphere a source-free region. We ignore absorption and clockwise direction. If if includes the entire .xy-plane,
DIFFRACTION 65
(c)
Fig. 2.30 Diffraction effect of a plane area &}. (After Trorey,

1970.) (a) Calculation using a surface integral; (b) calculation
using a line integral where the origin is not over the area; (c)
calculation when the origin is over the area; (d) evaluating the
line integral when £ is multivalued.
i.e., we have a continuous plane reflector, £ = °o and curring at a point or along a line, we are greatly sim-
the integral vanishes. The first term in this equation plifying the actual phenomena. In fact, both reflec-
thus represents the simple reflection from this plane tions and diffractions are the resultants of energy that
and the integral represents the diffracted wave. Com- return from all parts of the surface. From this point
parison of eqs. (2.128) and (2.129) shows that the of view a reflection is merely a special type of diffrac-
diffracted wave is given by the same expression in tion, a point of view that has interesting practical ap-
both cases. plications (see §9.12.2).
An important point to note is that both the reflec-
tion and diffraction terms in eqs. (2.128) and (2.129)
2.8.3 Time-domain solution for diffraction
are derived from the integral in eq. (2.125), where the
integration is carried over the entire surface. When we We now obtain the time-domain solution of eqs.
use rays and think of reflection and diffraction as oc- (2.128) and (2.129). The inverse transform of the re-
flection term on the right-hand side of eq. (2.129) fact B and D are at infinity, so that 9 increases from
gives the impulse (c/2h) h(t — 2h/V), that is, a repeti- —^TT to +ÎT as the point of integration A traverses the
tion of the input at the source after a delay of 2hl V, boundary in a clockwise direction. The result (see
which is the two-way traveltime from the source to the problem 2.19) is
plane surface, with the amplitude falling off inversely
_ 2(chlTTV>)(\l?)(tvt)
as the distance. Thus, the reflection has the same
waveshape as the source. The diffraction terms can be (t2 + tj- t%f - £T
found as follows: We write t = 2%/V, which is the two- (2.132)
way traveltime from the source to the variable point (t > O,
on the boundary. Equation (2.128) now becomes t2)
= o (t
ch ch
—2 ^dt. (2.130)
«v§ t dt where t = 2%V, tr = 2rlV, and ty = 2yJV.
This value of <\>(t) gives the diffracted wave recorded
We must pay careful attention to the limits of inte- at point P(0, 0, 0) as the result of an impulse, c 8(0,
gration because £, and hence t, is in general a multival- applied at the same point. If the input is cgif) instead
ued function of 6; for example, when 9 = 83 in fig. of c 5(/), O will have the factor G{s) and the response
2.30d, £ can have any of the values £m, £„, or ^. To becomes 4>(0 * g(t) (see eq. (15.195).
avoid difficulties, the integral is calculated as the point Equation (2.132) gives the diffraction effect whether
of integration goes from A to B (9 from §x to 02), then P is off the plane, as in fig. 2.31, or over the plane;
from B to C, C to D, and finally D to A, the proper because y0 changes sign as P passes over the edge, the
values of £ (and t) being used along each segment of diffracted wave undergoes a 180° phase shift as P pas-
the path. Along a given portion of the path, say, be- ses over the edge. Moreover, if we write D for the value
tween t = tx and / = t2 (t2 > tx), we have of $ for the diffracted wave observed when P is infini-
tesimally close to the edge and to the left of it, the
total effect observed when P is the same distance to
the right of the edge will be R - D, R being the value
of the reflection term in eq. (2.129). Because <)>(/) is
Comparing the right-hand integral with continuous,
eq. (15-178a), we conclude that (2.131)
R- D=D or (2.133)
<K0 = 0 (t < tx), Thus, the maximum amplitude of the diffraction from
= (ch/TrV2t2)(dd/dt) (tx<t< t2\ a half-plane is half the amplitude of the reflected wave
= 0 it > t2). (as observed far from the edge). Figure 2.32 shows
what is expected from a half-plane based on eq.
The derivative, d9/df, is finite except when £ is con-
(2.132). As the edge of the reflector is approached, the
stant, such as where W is bounded by an arc of a circle
diffraction gains in amplitude whereas (R — D) de-
with its center at the origin for which case dt = 0. In
creases in amplitude until at the edge D = l-R and the
this special case, eq. (2.130) gives
where 9,. and 9y fix the end points of the circular arc,
and t0 is the two-way traveltime to the arc. The inverse
transform is
| = (chlirV2?^,
<> - d)b(t - t0).
When dbldt is finite, we can get the time-domain
solution for the diffracted wave by dividing the bound-
ary of the area V so that the two-way traveltime t be-
tween P and the boundary is a single-valued function
of 9 in each part, calculating 4> for each part from eq.
(2.131), then summing the various (|>'s to get $ for the
diffracted wave.
2.8.4 Diffraction effect of a half-plane

A simple illustration of the method is the calculation
of the diffraction effect in the important case of a hori-
zontal half-plane at a depth h, the edge being parallel
to the x-axis and a distance y0 from it. By referring to Fig. 2.31 Calculation of the diffraction from a half-plane.
fig. 2.31, the edge of the half-plane is BD, where in
Surface
K=1.6km/s 800 m
(a)
r IT ir L IT jE }r
: I- !': H f- \~
: t: |:
1.0 ••c-r-r
UyH ;jULLi.
TTT
-; i ••!:
1 1.3 .J i r.
'- » - • -4—•-•
• " ! I J . :
' - I- !
; != i£ ;•
1.6 ! l~ I-" I-
Fig. 2.32 Seismic response of a half-plane. (After Trorey, sources and geophones. The arrowhead indicates the location of
1970.) (a) Model and (b) computed seismic record for coincident the edge of the half-plane.
I i I I I 1 I I I
/ = t0 + 2A'
Fig. 2.33 Diffracted wavefronts for a faulted bed.
sum is -R. The phase reversal of the diffraction before 2.3 (a) Firing an air gun (§7.4.3) in water creates a
the edge is reached (the backward branch of the pressure transient a small distance away with peak
diffraction) from that beyond the edge (the forward pressure of 5 atmospheres (5 X 105 Pa). If the com-
branch) is evident in fig. 2.32. pressibility of water is 4.5 X 10~10/Pa, what is the peak
energy density?
(b) If the same wave is generated in rock with X = |x =
3 X 1010 Pa, what is the peak energy density? Assume
2.8.5 Using Huygens' principle to construct diffracted a symmetrical P-wave with exx = eyy = ezz, etj = 0 for
wavefronts
2.4 To illustrate the interrelationship and magnitude
The surface integral in eq. (2.125) shows that the of the elastic constants, complete Table 2.3. Note that
diffraction effect at a point is the sum of effects arising these values apply to specific specimens; the values for
from the entire diffracting surface. This suggests the rocks range considerably, especially as porosity and
use of Huygens' principle to construct diffracted pressure change.
wavefronts, and this is the case for points more than
2.5 (a) Verify that i|i = fix - Vt) and i|i = g(x + Vt)
a few wavelengths away from the diffracting source.
in eqs. (2.46) and (2.47) are solutions of the one-
Figure 2.33 illustrates this construction for a faulted
dimensional wave equation, eq. (2.45). {Hint: Let £ =
reflector. We assume a plane wavefront AB incident
x — Vt and show that
normally on the faulted bed CO, the position of the
wavefront when it reaches the surface of the bed at
t = t0 being COD. At t = t0 + At, the portion to the = 4f
dx d£
right of O has advanced to position GH, whereas the
portion to the left of O has been reflected and has etc.)
reached position EF. We might have constructed (b) Verify that i|/ = f(€x + my + nz - Vt) + g{€x +
wavefronts EF and GH by selecting a large number of my + nz + Vt) in eq. (2.48) is a solution of the plane-
centers in CO and OD and drawing arcs of length wave equation, eq. (2.30).
V&t; EF and GH would then be determined by the (c) Using the same technique, show that i|/ = (l/r)/(r
envelopes of these arcs. However, for the portion EF, - Vt) + {\lr)g{r + Vt) in eq. (2.52) satisfies the wave
there would be no centers to the right of O to define equation in spherical coordinates, eq. (2.50).
the envelope, whereas for portion GH there would be 2.6 (a) Show that the wave equation, eq. (2.30), can
no centers to the left of O to define the envelope. be written in cylindrical coordinates {x = r cos 6, y =
Thus, O marks the transition point between centers r sin G, and z = z; see fig. 2.34a) as
that give rise to the upward-traveling wavefront EF
and centers that give rise to the downward-traveling
wavefront GH; arc FFG with center O is the diffracted
1 dty J _ a ^ d^
dr2 r dr r2 d6 2 dz2
wavefront originating at O and connecting the two
wavefronts, £Fand GH. The diffracted wavefront also (b) Verify that eq. (2.49) is the wave equation in spher-
extends into the geometrical shadow area GN and ical coordinates by substituting the following coordi-
into region FM. nate transformation (see fig. 2.34b) into eq. (2.30):
The characteristics of diffractions in various situa-
tions are discussed further in §6.3.1. x = r sin 0 cos c|>,
y = r sin 9 sin ()>,
z = r cos 0.
(For an easier solution, see problem 15.8.)
Problems
2.7 A pulse composed of two different frequency
2.1 (a) If uxx is the only nonzero normal stress, that components, co0 ± Aoo, can be represented by factors
is, if <jyy = azz = 0, use Hooke's law, eq. (2.13), to show involving the sum and differences of the frequencies.
that strain eyy - szz, and verify eq. (2.20) for Pois- If they have equal amplitudes, we can write for the
son's ratio. two components
(b) By adding the three equations in part (a) for crxx,
A cos {KXX — A cos {K2X — o)2t),
<ryy9 and crzz, derive eq. (2.19) for Young's modulus.
(c) Pressure 0> is equivalent to stresses vxx = uyy = where ooj = co0 + Aw, a)2 = co0 — Aco, K0 = 2TT/X.O =
(izz = —3P. Substituting these relationships into eq. u*JV, K2 ~ K0 + AK « (o)0 + Ao))/K and K2 ~ K0 —
(2.13), derive eq. (2.21) for the bulk modulus. AK « (CD0 - Aco)/K
2.2 The entries in table 2.2 express the quantities at (a) Show that the pulse is given approximately by the
the heads of the columns in terms of the pairs of elas- expression 2& cos (KOJC - (o0/), where 2& = 2A cos AK{X
tic constants or velocities at the left ends of the rows. - (ACO/AK)/}.
The first three entries in the ninth row are eqs. (2.19) (b) Why do we regard 2& as the amplitude? Show that
to (2.21). Starting with these and eqs. (2.58) and the envelope of the pulse is the graph of Sft plus its
(2.59), derive the other relations in the table. reflection in the x-axis.
Table 2.2 Relations between elastic constants and velocities (isotropic media)
Young's Poisson's Bulk Shear Lame P-wave S-wave Velocity

modulus, ratio, modulus, modulus, constant, velocity, velocity, ratio,
E a k |x X a P
£(l-a) I"2 | E im
(1 - 2 a
3(1 - 2a) 2(1 + a) (1 + a)(l - 2a) (1 + a)(l - 2a)p 2(1 + a)p 2(1 - a)
(£*)
• -
6k
E
9k- E
~A p(9k - E)
E)
3k
E-2 Vr
3A:(1 - a ) 3k 1 - 2a
(a, A:) 3A:(1 - 2<x)
2 ( l + a ) p(l + a) 2p\l+a/ 2(1 - a )
1/2 / \ 1/2
2ji\ / 1 - a 1 -2a
O, + <r)
2(1 - a)
3(1 - 2a) P / \ 1 " 2a
a)(l - 2g) 1 -2a

P\ 2a 2(1 - a)
3k
- 2|x/3
2(3/C+JJl) 4(i/3/
3(k - X) 1/ A : - \
(feX) k
W^> 3k- - X 2p 2lA: - 2X/3
X
2(X + M.)
p(a 2 - 2p 2 )
(a) (b)
Fig. 2.34 Coordinate systems, (a) Cylindrical coordinates; (b)

spherical coordinates.
(c) Show that the envelope moves with the group ve- where the subscripts denote partial differentiation
locity U, where (these are respectively the normal and tangential dis-
placements, normal and tangential stresses).
2.12 A source of seismic waves produces a step dis-
AK <1K dX do) placement on a spherical cavity of radius r0 enclosing
(seefig.2.26). the source of the form
2.8 The magnitudes of period T, frequency v, wave- step0 (0 = 0 (/ < 0),
length X, and angular wavenumber K are important = k (t > 0).
in practical situations. Calculate T, X, and K for the
situations shown in table 2.4. Show from eq. (2.77) that the displacement is given by
2.9 (a) Show that the potential function £ = V(<)> +
d\ldz) - V2xk of eq. (2.63) requires that A and 8Z be
solutions of the P- and S-wave equations, eqs. (2.28)
and (2.29), respectively. (Hint: Recall that <\> and x are
solutions of the wave equation and calculate V • £ and Is the motion oscillatory? What is the final (perma-
V x { using eqs. (15.12) and (15.13).) nent) displacement at distance r?
(b) Show that the dilatation and rotation (eqs. (2.65)) 2.13 Show that for harmonic waves of the form c|> =
can be derived from £ in eq. (2.64) using eq. (15.14) (Air) cos co(r/K - t), the two terms in eq. (2.71), which
and the results of problem 15.7. decay at different rates, are of equal importance at dis-
2.10 Justify on physical grounds the following bound- tance r = X/2TT.
ary conditions for different combinations of media in 2.14 Equations (2.86) and (2.87) for a Rayleigh wave
contact at an interface: are valid at the surface z = 0 for a = 1/4.
(a) Solid-fluid: normal stress and displacement are (a) Show that at depths z ^ 0 , the expressions for the
continuous, tangential stress in the solid vanishes at x- and z-displacements u and w are
the interface. u = KA(-Q~0MSKZ + 0.577e-°-393KZ)sin K(X - VRt\
(b) Solid-vacuum: normal and tangential stresses in w = K^(-0.848e-°-848K2+1.468e-°393KZ)
the solid vanish at the interface. X cos K (x - VRt).
(c) Fluid-fluid: normal stresses and displacements
are continuous. (b) What are the values of u, w, and 9 when z = 1/2K?
(d) Fluid-vacuum: normal stress in the fluid vanishes When z = 1/K?
at the interface. (c) Is the motion retrograde for all values of z? (Hint:
2.11 Using eqs. (2.8), (2.9), (2.12), (2.13), (2.14), and Note that for the motion to change direction, the am-
(2.64), show that the boundary conditions at the xy- plitude of either u or w must pass through zero.)
plane separating two semiinfinite solids require, for a (d) What are the values of VR, the Rayleigh-wave veloc-
wave in the xz-plane, the continuity of the following ity, when a = 0.4 and when a = 0.2? What are the
potential functions: corresponding values of the constants in part (a)?
2.15 Assume three geophones so oriented that one re-
cords only the vertical component of a seismic wave,
XV2<t> + 2jx(c|>zz - X J , another records only the horizontal component in the
+ Xzz ~ XXXX direction of the source, and the third only the hori-
REFERENCES 71
Table 2.3 Example of magnitudes of elastic constants and velocities
Water Stiff mud Sandstone Limestone Granite
Young's modulus, E (X 109 Pa) 16 54 50

Bulk modulus, k (X 109 Pa) 2.1
Rigidity modulus, |JL (X 109 Pa)
Lame's \ constant (X 109 Pa)
Poisson's ratio, a 0.5 0.43 0.34 0.25 0.20
Density, p (g/cm3) 1.0 1.5 1.9 2.5 2.7
P-wave velocity, a (km/s) 1.5 1.6
S-wave velocity, p (km/s)
Table 2.4 Magnitudes ofT, A, and K
a (km/s) For v = 15 Hz For v = 60 Hz
\(m) K
1
(m" ) T(s) \(m) K (m-1)
Weathering (mm.) 0.1

Weathering (avg.) 0.5
Water 1.5
Poorly consolidated sands-shales 2.0
at 0.75 km
Tertiary elastics at 3.00 km 3.3
Porous limestone 4.3
Dense limestone 5.5
Salt 4.6
Anhydrite 6.1
zontal component at right angles to this. Assume a What percentage of the energy is lost over three cy-
simple waveshape and draw the response of the three cles? What is the value of/??
geophones for the following cases: 2.19 Using eq. (2.131), the general equation for the
(a) A P-wave traveling directly from the source to diffraction from a plane surface, verify eq. (2.132) for
the geophones. the diffraction effect of a half-plane.
(b) A P-wave reflected from a deep horizon.
(c) An S-wave generated by reflection of a P-wave at
an interface. References
(d) A Rayleigh wave generated by the source.
(e) A Love wave. Aki, K., and P. G. Richards. 1980. Quantitative Seismology:
Theory and Methods. San Francisco: W. H. Freeman.
Compare the relative magnitudes of the components
for short and long offsets. Birch, F. 1966. Compressibility; elastic constants. In Handbook
of Physical Constants, S. P. Clark, ed., GSA Memoir 97.
2.16 (a) A tube wave has a velocity of 1.05 km/s. The
fluid in the borehole has a bulk modulus of 2.15 X 109 Blake, F. C. 1952. Spherical wave propagation in solid media. J.
Pa and density 1.20 g/cm3. The wall rock has a = 0.25 Acoust. Soc. Amer., 24: 211-15.
and p = 2.5 g/cm3. Calculate |JL and a for the wall rock, Braddick, H. J. J. 1965. Vibrations, Waves, and Diffractions. New
(b) Repeat for VT = 1.20 km/s and 1.30 km/s. What York: McGraw-Hill.
do you conclude about the accuracy of this method Brillouin, L. 1960. Wave Propagation and Group Velocity. New
for determining |x? York: Academic Press.
2.17 The natural logarithm of the ratio of amplitudes Bush, I., and S. Crampin. 1987. Observations of EDA and PTL
is measured in nepers. Show that 1 neper = 8.686 dB. anisotropy in shear-wave VSP. Expanded Abstracts, 57th An-
nual International Meeting of the Society of Exploration Geo-
2.18 A refraction seismic wavelet assumed to be es- physicists, pp. 646-59. Tulsa: Society of Exploration Geophysi-
sentially harmonic with frequency 40 Hz is found to cists.
have amplitudes of 5.00 and 4.57 mm on traces 2.50
Cheng, C. H., and M. N. Toksoz. 1981. Elastic wave propaga-
and 3.00 km from the source. Assuming a velocity of tion in fluid-filled borehole and synthetic acoustic logs. Geo-
3.20 km/s, constant subsurface conditions, and ideal physics, 46: 1042-53.
recording conditions, what is the ratio of the ampli- Crampin, S. 1981. A review of wave propagation in anisotropic
tudes on a given trace of the first and fourth cycles? and cracked elastic media. Wave Motion, 3: 343-91.
Crampin, S., R. McGonigle, and D. Bamford. 1980. Estimating Press, F , and R. Siever. 1978. Earth, 2d ed. San Francisco:
crack parameters from observations of P-wave velocity aniso- W H. Freeman.
tropy. Geophysics, 45: 345-60.
Saada, A. S. 1974. Elasticity: Theory and Applications. Oxford:
Dobrin, M. B. 1951. Dispersion in seismic waves. Geophysics, Pergamon Press.
16: 63-80.
Savarensky, A. 1975. Seismic Waves. Moskow: MIR.
Ebrom, D. A., R. H. Tatham, K. K. Sekharan, J. A. McDonald,
and G. H. F. Gardner. 1990. Dispersion and anisotropy in lam- Schoenberg, M., and F. Muir. 1989. A calculus for finely layered
inated versus fractured media: An experimental comparison. media. Geophysics, 54: 581-9.
Expanded Abstracts, 60th Annual International Meeting of the Scholte, J. C. 1947. The range of existence of Rayleigh and Sto-
Society of Exploration Geophysicists, pp. 1416-19. Tulsa: Soci- neley waves. Royal Astron. Soc. Monthly Notices Geophys.
ety of Exploration Geophysicists. Supp, Ser. A, 106: 416-28.
Ewing, W. M., W. S. Jardetzky, and F. Press. 1957. Elastic Waves Segonzac, P. D., and J. Laherrere. 1959. Application of the con-
in Layered Media. New York: McGraw-Hill. tinuous velocity log to anisotropy measurements in Northern
Futterman, W. I. 1962. Dispersive body waves. J. Geophys. Res., Sahara: Results and consequences. Geophys. Prosp, 7: 202-17.
67:5279-91. Stoep, P. M. 1966. Velocity anisotropy measurements in wells.
Geophysics, 31: 900-16.
Gerkens, J. C. d'Arnaud. 1989. Foundation of Exploration Geo-
physics. Amsterdam: Elsevier. Stoneley, R. 1924. Elastic waves at the surface of separation of
Grant, F. S., and G. F. West. 1965. Interpretation Theory in Ap- two solids. Proc. Roy. Soc. (London), A-106: 416-28.
plied Geophysics. New York: McGraw-Hill. Stoneley, R. 1949. The seismological implications of aeolotropy
Hardage, B. A. 1985. Vertical Seismic Profiling, Part A: Prin- in continental structures. Monthly Notices, Roy. Astron. Soc.
Geophys. Supp, 5: 343-53.
ciples, 2d ed. London: Geophysical Press.
Howell, B. 1959. Introduction to Geophysics. New York: Toksoz, M. N., and D. H. Johnston. 1981. Seismic Wave Attenu-
ation, SEG Geophysical Reprint Series 2. Tulsa: Society of Ex-
McGraw-Hill.
ploration Geophysicists.
Hsieh, J. S. 1975. Principles of Thermodynamics. New York:
Trorey, A. W. 1970. A simple theory for seismic diffractions.
McGraw-Hill.
Lamb, H. 1960. Statics. New York: Cambridge University Press.
Trorey, A. W. 1977. Diffractions for arbitrary source-receiver lo-
Landau, L. D., and E. M. Lifshitz. 1986. Theory of Elasticity, cations. Geophysics, 42: 1177-82.
3d ed. Oxford: Pergamon Press.
Uhrig, L. F , and F A. van Melle. 1955. Velocity anisotropy in
Lapedes, D. N., ed. 1978. McGraw-Hill Dictionary of Physics stratified media. Geophysics, 20: 774-9.
and Mathematics. New York: McGraw-Hill.
Ward, R. W, and M. R. Hewitt. 1977. Monofrequency borehole
Logan, J. D. 1987. Applied Mathematics. New York: John Wiley. traveltime survey. Geophysics, 42: 1137-45.
Love, A. E. H. 1911. Some Problems of Geodynamics. New White, J. E. 1965. Seismic Waves - Radiation, Transmission, and
York: Dover. Attenuation. New York: McGraw-Hill.
Love, A. E. H. 1944. A Treatise on the Mathematical Theory of White, J. E. 1983. Underground Sound - Application of Seismic
Elasticity. New York: Dover. Waves. Amsterdam: Elsevier.
MacBeth, C. 1990. Inversion of shear-wave polarizations for an- Winterstein, D. F. 1990. Velocity anisotropy terminology for
isotropy using three-component offset VSPs. Expanded Ab- geophysicists. Geophysics, 55: 1070-88.
stracts, 60th Annual International Meeting of the Society of Ex-
ploration Geophysicists, pp. 1404-6. Tulsa: Society of Winterstein, D. F , and M. A. Meadows. 1990. Shear-wave po-
larizations and subsurface stress directions at Lost Hills Field.
Exploration Geophysicists.
Expanded Abstracts, 60th Annual International Meeting of the
Newman, P. 1973. Divergence effects in a layered earth. Geo- Society of Exploration Geophysicists, pp. 1431-4. Tulsa: Society
physics, 38: 481-8. of Exploration Geophysicists.
Postma, G. W. 1955. Wave propagation in a stratified medium.

2 Theory-Of-Seismic-Waves

Загружено:

Сведения о документе

Оригинальное название

Авторское право

Доступные форматы

Поделиться этим документом

Поделиться или встроить документ

Параметры публикации

Этот документ был вам полезен?

Это неприемлемый материал?

Авторское право:

Доступные форматы

2 Theory-Of-Seismic-Waves

Загружено:

Авторское право:

Доступные форматы

Theory of seismic waves

Overview fronts and raypaths are introduced, as is the more gen-

motion can be more simply illustrated in this way.

the total moment must be zero; hence vxy = vyx. In

Plastic yield change in volume per unit volume A is

smaller the larger |JL is. Hence, |x is a measure of the

Pressure 2P causes a decrease in the volume AT and a 2.1.6 Strain energy

_ 1/ «,* + p — = unbalanced force in the x-direction on

where V2w is the Laplacian of u = d2u/dx2 + d2u/dy2 +

Similar expressions hold for the other faces; hence, we

or in two ways in general: (a) include in the wave equa-

These equations are difficult to solve as they stand.

1 d2® V2 ((32V2X + ft —)2 = 0.

tained identical results for S-waves. Thus, the previous

2.2.4 Kirchhoff's theorem 2.2.5 Plane-wave solutions

the same everywhere on a plane perpendicular to the

2.2.6 Spherical-wave solutions

a/p mum compression (maximum A) at the wavefront D;

Fig. 2.10 Displacements for a spherical P-wave.

cementation, depth, and age. (Velocity versus depth

solely of a displacement v of the medium in the y-

g>W = constant, 7 = cjcv « 1.4 for air, (2.69)

Fig. 2.14 Radial displacement involves shape distortion. When

fit) = -r0Ve-Vtlr° I uo(t)evt»»dt. (2.75) t = r - In

We now find that VRla = 0.531, m = 0.848, n = 0.393, Medium (2):

where mp n. are real, positive constants, and Vs is the

By setting a = e~2T11K/l, b = M'2'VlJLi'Tli> tnese

/ For these homogeneous equations to have a nontrivial

Fig. 2.17 Conditions for the existence of a Stoneley wave; solu-

The expression for V! can be written

on using eq. (2.91). Therefore,

2.5.5 Tube waves The change in volume AT is due to expansion both

is the dominant tube-wave mode in VSP work. The

variation of seismic velocity with the direction in

2.6 Anisotropic media

(V2 - a2€2 - a*2n2)£A

Sj_: N 45° E [V2 - (a2 - a*2 + p*2)€2 - p*2«2]«

S|| : N 45° W (V2 - a2€2 - a*2«2)€

Note the dependence in eqs. (2.101) and (2.103) on X (V2 - a2±n2 -

_ d§ dx _ /jw 2.7 Effects of the medium on wave propagation

-4000 -2000 2000 4000

Green River shale

We are also interested in the rate of flow of energy

which agrees with eq. (2.110).

2.7.3 Relative importance of absorption and spreading

oit) varies rapidly. When adjacent frequency compo-

[ This function is shown in fig. 2.27; it has a maximum

2.7.5 Reflection and refraction; Snell's law

sin 8C = VJV2. (2.121)

assume a constant reflection coefficient over if, so that

Fig. 2.30 Diffraction effect of a plane area &}. (After Trorey,

2.8.4 Diffraction effect of a half-plane

Fig. 2.33 Diffracted wavefronts for a faulted bed.

Young's Poisson's Bulk Shear Lame P-wave S-wave Velocity

a)(l - 2g) 1 -2a

Fig. 2.34 Coordinate systems, (a) Cylindrical coordinates; (b)

Table 2.3 Example of magnitudes of elastic constants and velocities

Water Stiff mud Sandstone Limestone Granite

Young's modulus, E (X 109 Pa) 16 54 50

Table 2.4 Magnitudes ofT, A, and K

a (km/s) For v = 15 Hz For v = 60 Hz

Weathering (mm.) 0.1

Вам также может понравиться

Sj_: N 45° E [V2 - (a2 - a2 + p2)€2 - p*2«2]«