Ge162 Seismology (2004)

1. Introduction (2 hours)

1.1 Earthquake Phenomena (Frequency Spectrum)

1.2 Source of an Earthquake (Fault, Crustal Deformation etc)
1.3 Seismic Waves (P, S, and Surface Waves)
1.4 Seismograph

2. Concepts in Classical Seismology (5 hours)

2.1 Structure of the Earth

2.1.1 Ray theory and Snell's Law
2.1.2 Seismic Ray
2.1.3 Travel-Time Curve
2.1.4 Gross Structure of the Earth

2.2 Location, Magnitude and Mechanism of Earthquakes

(+Practice Session)
2.3 Seismicity of the Earth

2.3.1 Global Seismicity (and Earthquake Mechanism),

California Seismicity
2.3.2 Depth Variation of Seismicity
2.3.3 Temporal Variation of Seismicity
2.3.4 Magnitude-Frequency Relationship

3. Seismic Waves (7 hours)

3.1 Review of Theory of Elasticity

3.2 Wave Equation
3.3 Seismic Body Waves
3.4 Ray Theory
3.5 Reflection and Refraction
3.6 Seismic Surface Waves (+Practice Session)
3.7 Normal Mode Theory

1
4. Seismic Tomography (2 hours)

4.1 Herglotz-Wiechert (+Practice Session)

4.2 General Tomography (+Practice Session)

5. Theory of Seismic Source (4 hours)

5.1 Static Source

5.1.1 Single Force
5.1.2 Force Couples
5.1.3 Elastic Dislocation and Moment Tensor
5.1.4 Summary of Seismic Source Representation
5.1.5 Stress Relaxation

5.2 Elastodynamic Source

5.2.1 Single Force
5.2.2 Force Couples
5.2.3 Elastic Dislocation and Moment Tensor
5.2.4 Stress Relaxation

6. Retrieval of Seismic Source Parameters (+Practice Session) (2 hours)

6.1 Body Waves

6.2 Long-Period Waves

7. Summary of Seismic Source Parameters (1 hour)

Seismic Moment, Radiation Pattern, Source Finiteness, Rupture Speed,

Directivity, Static and Dynamic Stress Drops, Energy

8. Physics of Earthquakes (2.5 hours)

8.1 Scaling Relations

8.2 Physics of Earthquakes
8.3 Earthquake as a Complex System

9. Earthquakes and Plate Motion (0.5 hour)

2
Primary Text Books

1. Lay, T., and T. C. Wallace, Modern Global Seismology, Academic Press, San Diego,
1-517, 1995.

2. Bullen, K. E., and B. A. Bolt, An Introduction to the Theory of Seismology, 4 edition,

Press Syndicate of the University of Cambridge, Cambridge, 1-499, 1985.

3. Udias, Agustin, Principles of Seismology, Cambridge University Press, Cambridge, 1-

475, 1999.

Other References

1. Richter, C. F., Elementary Seismology, W. H. Freeman, San Francisco, 1-768, 1958.

2. Scholz, C. H., The Mechanics of Earthquakes and Faulting, Cambridge University

Press, New York, 1-439, 1990.

3. Bullen, K. E., An Introduction to the Theory of Seismology, 3 edition, Cambridge Univ.

Press, Cambridge, 1-381, 1963.

4. Kasahara, K., Earthquake Mechanics, Cambridge University Press, Cambridge, 1-248,

1981.

5. Aki, K., and P. G. Richards, Quantitative Seismology, W. H. Freeman, San

Francisco, 1-932, 1980.

6. Aki, K., and P.G. Richards, Quantitative Seismology, 2nd Edition, 685 pp., University
Science Books, Sausalito, 2002.

7. Dahlen, F. A. and Tromp, J., Theoretical Global Seismology, Princeton University

Press, Princeton, 1025, 1998.

3
8. Shearer, P., Introduction to Seismology, Cambridge University Press, New York, 260,
1999.

9. Stein, S., and M. Wysession, An Introduction to Seismology, Earthquakes and Earth

Structure, 498 pp., Blackwell, Malden, 2003.

4
Ge 162 Seismology

1. Introduction

1.1 Frequency Spectrum of Earthquake Phenomena

Phenomena associated with earthquakes occur over a broad frequency range as

shown in Figure 1.1.

Fig. 1.1

5
1.2 Source of an Earthquake

Faulting and Crustal Deformation

An earthquake is a failure process in the Earth's crust (sometimes in the mantle,

too). As the stress in the crust builds up, it eventually exceeds the strength of crustal
rock, and failure occurs. The result is faulting (Figure 1.2), which causes deformation of
the crust (Figure 1.3).

6
 Fig. 1.3
The pattern of crustal deformation can be studied in detail using geodetic methods
(traditional ground-based method, GPS, SAR) and seismological methods.

Failure (Fracture vs. Frictional Sliding) (Figure 1.4)

The failure process during an earthquake is often illustrated as fracture of rocks.

If a piece of rock is subjected to stress (force), eventually fracture occurs. Although this
is qualitatively correct, it is probably more appropriate to view earthquake faulting as
frictional sliding. In this case, the sliding surface corresponds to an earthquake fault. A
fault is formed by a long-term geological process, and represents a weak zone. In a
sense, the major difference between fracture and frictional sliding is whether there is a
pre-existing weak zone (plane) or not.
A simple experiment on frictional sliding exhibits many important characteristics
of earthquakes; i.e. loading (stress accumulation), sudden slip, repetition of slip events.
This general behavior is called stick slip.

7
Fracture and Frictional Sliding

Stress (τ)

Stick-slip
Stable sliding

Displacement (x) time

time

Figure 1.4

Strain, Stress, Stress-Strain Relation (Hooke's law)

Strain is a measure of deformation of a deformable (continuous) medium, and

represents displacement per unit length. Stress is a measure of force. It is measured by
force per unit area. In seismology, shear strain and shear stress are most important.
The relation between stress (force) and strain (deformation) can be best illustrated
by using a spring. (In fact a spring is a very useful analog of Earth's crust, and we can
explain many important relationships in seismology using a spring.) Suppose we apply a
force F to stretch a spring, and the spring is stretched by Δl. Then F and Δl are related
by the Hooke's law,

F=ksΔl (1-1)

Here ks is called the spring constant. A similar relation

8
 σ = με (1-2)

holds between stress σ and strain ε . Here μ is called the elastic constant. If the rock
is harder, then μ is larger. We have different elastic constants for shear deformation and
volumetric deformation (volume change). In general, for simple deformable bodies, like
rocks, metals etc, we need two elastic constants μ and k, rigidity and incompressibility
(also called bulk modulus), for shear and volumetric deformations, respectively. For
most seismological problems, these two elastic constants, and the density ρ are most
important. For most crustal rocks, the representative values are:

μ =3x1011 dyne/cm2=0.3 Mbars=30 GPa

k=5x1011 dyne/cm2=0.5 Mbars=50 GPa
ρ =2.7 g/cm3=2,700 kg/m3

Critical strain (strength) of Earth's crust

An important question is "How strong is the Earth's crust?". Many geodetic and
seismological studies have demonstrated that the change in strain (deformation)
associated with an earthquake ranges from 3x10-5 to 3x10-4, or, in terms of stress, this
corresponds to 10 to 100 bars (i.e., 1 to 10 MPa, or 10 to 100 atmospheric pressure). If
we try to break an intact piece of rock, we normally need a few kbar stress. This suggests
that an earthquake occurs on a pre-existing weak plane (fault), and frictional sliding
appears to be a more appropriate model for an earthquake.

Ground Motion

When an earthquake occurs, the ground shakes. The motion of the ground is
given by the displacement u(t) as a function of time, t, in 3 directions, usually, UD, NS,
and EW. If we take the time derivative of u(t), we get the velocity of ground motion
v(t ) = u (t ) , and if we differentiate it again, we get ground-motion acceleration
a (t ) = v(t ) = u (t ) .
The ground motions near the source of an earthquake are measured with
geological, geodetic and seismological methods (Figure 1.5).

9
 For large earthquakes,

u is 1 to 20 m
v is 10 cm/sec to 3 m/sec
a is 0.1 to 20 m/sec2 (10 m/sec2 is about 1 g)

These are useful numbers in seismology to remember.

1.3 Seismic Waves

The disturbance caused by a faulting in Earth's crust propagates as elastic waves.

These waves are called seismic waves. In a large homogeneous medium without

10
boundary, two types of waves exist. The first wave is mainly due to volume change, and
is called compressional wave or P wave, and the second type of wave is caused by shear
deformation, and is called shear wave or S wave. The P and S wave velocities, α and β,
respectively, are given by

k + (4 / 3) μ μ
α= and β = (1-3)
ρ ρ

For most solids, k=(5/3) μ , so that α = 3β = 1.732β . Although this is an

approximate relation, it is a good approximation and useful in observational seismology.
In the shallow part of Earth's crust α=5 to 6.5 km/sec, and β=3 to 3.8 km/sec.
In real world, we have the surface of Earth, and also the velocity changes as a
function of depth. In general, the velocity increases with depth. When P and S waves
propagate in a medium with a free surface and layers of different velocities, complex
reflection and refraction occur and P and S waves interact to generate surface waves.
This is an interesting mathematical problem, but here we just introduce two types of
surface waves, Rayleigh waves and Love waves. These waves primarily propagate along
Earth's surface so that the amplitude decays more slowly with distance than P and S
waves. Hence, at large distances, surface waves are more dominant on seismograms
(recordings of seismic waves). Also, these waves exhibit dispersion (velocity varies with
the wave period) so that the appearance of these waves are usually distinct. The particle
motion of Rayleigh waves is vertical and in the direction of wave path, and that of Love
waves is horizontal and in the direction normal to the path. In the shallow crust, the
velocity of Rayleigh waves CR, ranges approximately from 2.5 to 3.0 km/sec, and that of
Love waves, CL, from 2.8 to 3.5 km/sec. Two examples are shown in Figure 1.6.

11
12
Fig. 1.6 Seismograms of the 1995 Kobe, Japan, and the 1999 Izmit, Turkey,
earthquakes.

As shown above, the velocities of these waves are different, i.e., in general α >
β > CL > CR. Thus, at a station some distance from the source, the P wave arrives first,
which is followed by the S wave. Then the large amplitude Love wave and Rayleigh
wave arrive.
Seismologists study these seismic waves in detail to determine the earthquake
source parameters (the size, type of faulting etc), and the structure of the Earth.
One useful relation is that between the S-P time, t S− P , (time interval between P
and S waves) and the distance, Δ. Since t S− P is given by

13
 Δ Δ Δ ⎛α ⎞
tS − P = − = ⎜ − 1⎟ (1−4)
β α α⎝β ⎠

from which

α
Δ= tS − P (1-5)
⎛α ⎞
⎜ β − 1⎟
⎝ ⎠

For the shallow part of the crust (i.e., Δ<1000 km), α is about 6 km/sec, and α / β is
1.732, so that this relation gives

Δ ≈ 8t S− P (t S− P in sec and Δ in km) (1-6)

For example, if t S− P is 50 sec, the distance is about 400 km.

1.4 Seismograph (LW (Lay and Wallace), pp 173-199)

The instrument that measures ground motion caused by an earthquake is a

seismometer or a seismograph. During an earthquake, everything moves so that it is
difficult to measure the ground motion accurately. We need a reference point from which
we can measure the motion of the ground. Seismologists use a pendulum as a reference.
Since most seismographs use a pendulum as a reference of the position, here we
briefly discuss the principle (Figure 1.7).

14
Pendulum and Simple Mechanical Seismograph

Consider a simple pendulum (a string with a small mass hanging from it). Hold
one end of the string and let the mass swing in a vertical plane. The natural period of the
pendulum, T0, is given by,

15
 l
T0 = 2π (1-7)
g

where g is the acceleration of gravity, 9.8 m/sec2. Thus if l=1 m, then the period is about
2 sec.
Suppose we have an earthquake, and the ground starts shaking horizontally.
Since you are standing on the ground (i.e., fixed to the ground), you will be shaken with
the ground. If the ground shakes very gradually, say at period T >>T0, then the mass will
move with you so that you cannot use it as a reference point (i.e., everything moves in the
same way). In this case, this pendulum is not good as a seismometer. However, if the
ground shakes very rapidly, e.g., T<<T0, then the mass tends to stay at the same place, if
not completely. This is the very principle of a horizontal seismometer, a seismometer
that measures horizontal motion. You can measure the motion of ground with respect to
the mass which is approximately stationary. Thus, if T<<T0, this pendulum is a good
seismometer, and if you record the motion of the ground (i.e., you) with respect to the
mass, you can have a seismogram.
If we use a spring with a mass hanging vertically, we can measure the vertical
ground motion with the same principle.
In the real seismometer, we need to attach a device to magnify the motion and
damp out the resonance (damper), but the basic principle is the same. The Wiechert
seismograph and the Wood-Anderson seismograph used in California are all of this type.
It is difficult to record very long-period ground motions with these simple
mechanical seismographs, because it is difficult to build a stable pendulum with a very
long natural period.

The response of a mechanical seismograph can be derived from the following

equation (see the figure below).

x + 2hω 0 x + ω 02 x = −Vy (1-8)

where x is the motion of a reference point (i.e., mass) of the seismograph with respect to
the ground, and y is the ground motion displacement. h is the damping constant,
ω 0 = k / m is the natural angular frequency of the seismograph ( ω0 = 2π / T0 , T0 is the
natural period), and V is the static magnification. The response of a mechanical
seismograph is completely described by these 3 constants.

16
 Vx x

For a unit harmonic input

y = exp(iω t ) (1-9)

the output is

Vω 2
x= exp(iω t ) (1-10)
−ω 2 + 2ihω 0ω + ω 02

then, the response is given by

Vω 2
Hˆ (ω ) = (1-11)
−ω 2 + 2ihω 0ω + ω 02

The amplitude response | Hˆ (ω ) | is plotted in Figure 1.8.

Other Types of Seismographs

17
 In order to increase the sensitivity of the instrument, and improve the response at
long-period, various developments have been made. These seismographs use an electro-
magnetic sensor (moving coil etc), and a galvanometer. The examples are: Galitzin,
Benioff short-period, Benioff long-period, Press-Ewing, and Benioff strain seismographs
(Figure 1.8).

Modern Broad-band Seismograph and Strong-motion Seismograph

18
 Until recently seismograms written on paper were the standard data for most
seismological research and routine reporting. While these analog records are still useful
for various research purposes, these instruments are limited in two respects. First, since
the movement of the pendulum is mechanically limited by the physical size of the
instrument, it is not possible to record very large ground motions; i.e., the dynamic range
is limited. Second, as long as recordings are made on paper (analog recording), the
dynamic range is limited by resolution in visually reading the records, which is normally
1/1000, i.e., 60 db.
To remove these limitations, modern seismographs adopt a force balance
mechanism and digital recording system. In the force balance mechanism, the output
signal from the transducer is amplified and fed back to a device that holds the mass at the
original unperturbed position (Figure 1.9). The strength of the signal (usually measured
in voltage) is proportional to ground motion. With some filters in the feedback circuit, it
is possible to make the output proportional to acceleration, velocity or displacement of
ground motion, at least over a certain frequency band.
In this type of instruments, there is virtually no displacement of the mass, and the
dynamic range can be increased. Also, with an appropriate feedback system, the
response can be adjusted relatively easily.

19
 The response of the standard broad-band instruments used in seismology (often
called the VBB system) is approximately flat for a ground-motion velocity over a wide
frequency band (e.g. 7 Hz to 0.0033 Hz (300 sec)) (Figures 1.10 and 1.11). The broad-
band instruments usually have a 140 db (107) dynamic range.

20
21
 Seismographs designed to record very strong ground motion are called strong-
motion seismographs and are used in earthquake engineering. Modern strong-motion
seismographs have a force balance mechanism with voltage output proportional to
ground-motion acceleration.
Recent developments in solid state electronics made it possible to build stable
force balance seismographs; they are now widely used in the world for research and
routine monitoring.

22
Ge 162

2. Concepts in Classical Seismology

2.1 Structure of the Earth

2.1.1 Ray theory and Snell's Law (LW, pp70-91)

In the early days of seismology, the structure of the Earth was determined mainly
by using ray theory. When the wavelength of seismic waves is sufficiently short (i.e., if
the period is sufficiently short), we can treat seismic waves as a geometrical ray, just as
we do in geometrical optics.
In ray theory, the most fundamental is Snell's Law, which is illustrated in Figure
1.

23
 Snell’s Law

i1 sin i1 sin i2
=
v1 v2
v1

v2
i2

i
sin i( z )
= p = const
v( z )
r sin i (r ) r sin i (r )
v(z) =p =p
z v(r ) v( r )

Fig. 1Figure 1

Suppose a ray is incident from a medium with a wave speed v1 on a medium with
a speed v2 . Let the incident and emergent angles be i1 and i2 . Then the Snell's law is
given by

sin i1 sin i2
= (1)
v1 v2

This can be shown easily from the geometry of the two triangles OAB and OBC
shown below.

24
 If the wave speed changes continuously with z (i.e., depth) as v(z), then (1) can be
written as

sin i
= p =constant for a given ray (2)
v

p is called the ray parameter.

For a spherical geometry as shown in Figure 1, the Snell's law can be written as

rk sin ik rk +1 sin ik +1
= (3)
vk vk +1

and

25
 r sin i (r )
=p (4)
v( r )

corresponding to (1) and (2), respectively.

2.1.2 Seismic Ray (LW, pp. 200-217)

Figure 2 shows various seismic rays in the Earth's interior. P and S waves are
denoted by P and S, respectively. Other symbols are:

K: P wave ray in the core

p: P wave ray before reflection near the source (only for a deep source)
s: S wave ray before reflection near the source (only for a deep source)
c: Reflection at the core
P': PKP

26
27
 P, S, Pdif, Sdif pP, sS, PP, SS PcP, ScS ScP, PcS

PKP SKS P’P’ (PKPPKP) S’S’ (SKSSKS)

Storchak, D., Schweitzer, J., and Bormann, P., The IASPEI Standard Phase List, Seismological Research Letters,
74, 761-772, 2003.

28
Some Examples

Figure 3 and Figure 4 show some examples.

29
 Figure 5 shows ray paths in the Earth's interior for 3 representative velocity structures,
and the corresponding travel time curves.

30
31
 Figure 6 shows 3 regions in the Earth's interior and the corresponding travel time
curves.

Figure 7 shows the crustal structure for ocean and continent, and corresponding
travel time curves.

32
33
2.1.3 Travel-Time Curve (LW pp.213-217)

Figure 8 shows the travel times observed at many stations and reported to the
International Seismological Center (ISC).

34
35
Figure 9 shows the same with the phase names labeled.

36
 For a spherically symmetric structure, the distance, Δ , traveled by a seismic ray
with a ray parameter p, and the travel times, T , can be computed easily. Referring to the
figure below,

dr v2 p2
= cos i = 1 − sin 2 i = 1 − 2 (5)
ds r

ds sin i 1 vp
dΔ = = 2 dr (6)
r r v2 p2
1− 2
r

37
Then,

∫
r0
r
T =2 dr (7)
rp v r 2 − v2 p2

∫
r0
vp
Δ=2 dr (8)
rp r r − v2 p2
2

where r0 is the radius of Earth and rp is the radial distance of the deepest point of the
ray. These integrals are fundamental in the seismological ray theory (e.g., Herglotz
Wiechert Method).

2.1.4 Gross Structure of the Earth (LW, pp. 26-33)

In the early days of seismology, the structure of the Earth was determined from
the travel-time data, as shown in the upper figure of Figure 10. This figure shows one of
the standard laterally homogeneous model, which is used for various seismological
studies.
The bottom figure in Figure 10 shows an example of 3-D Earth structure
determined by more recent studies using seismic body waves, surface waves and normal-
mode data (Helffrich and Wood, 2001). Tomographic methods are used for determining
these structures.

38
References

Helffrich, G. R., and B. J. Wood, The earth's mantle, Nature, 412, 501-507, 2001.
Anderson, D. L., Top-down tectonics, Science, 293, 2016-2018, 2001.

39
Ge 162 Plate Motion and Great Earthquakes

Earthquakes occur in the Earth's crust and mantle due to stresses caused by global

plate motion. The actual pattern of stress distribution is probably very complex, but we

expect that the activities of great and large earthquakes must reflect the global plate

motion.

40
 The world greatest earthquakes occur at subduction zones (e.g., 1960 Chilean

earthquake, and the 1964 Alaskan earthquake), but not every subduction zone has

experienced a great earthquake (e.g., the Marianas, the Tonga-Kermadec). It is possible

that the length of earthquake catalog is too short to be representative of long-term

seismicity. With this caveat in mind, we investigate the level of seismic activity and

plate motion. Ideally, the seismic activity along a subduction zone should be defined by

the energy release per unit length along the subduction zone, and unit time, i.e.,

41
 1 L T
e=
LT ∫ ∫
0 0
ER dldt

where L and T are the length of the subduction zone and the time period involved,

respectively.

Unfortunately, the available seismic record is too short to compute this. So, we

take the magnitude, M w , of the largest earthquake that occurred in a particular

subduction zone as a parameter that represents e for that subduction zone. Then , it is

reasonable to assume that

Mw ∝V

where V is the convergence rate. However, the plot of M w versus V does not show any

obvious trend. This suggests that other factors may be controlling seismicity. Another

potentially relevant parameter is the age, T, of the subducting plate. However, no

obvious negative correlation is seen between M w and T.

42
 Then, we can try a 3-parameter regression between M w , V and T. The result is

shown in the following figure. The horizontal axis shows the observed M w and the

vertical axis shows M w predicted by the regression relation.

43
(Ruff, L., and H. Kanamori, Seismicity and the subduction process, Phys. Earth Planet.
Inter., 23, 240-252, 1980)

If this regression is valid, this provides a useful method for assessing the seismic

potential of subduction zones for which no great earthquake has occurred. This pattern

suggests that the subduction zones where a relatively young plate is subducting at a

44
relatively fast rate are more likely to have great earthquakes, and those with an old plate

subducting at a moderate rate are less likely to have great earthquakes. The end-member

subduction zones are the Chilean type and the Mariana type, shown below.

(Uyeda, S., and H. Kanamori, Back-arc opening and the mode of subduction, J. Geophys.
Res., 84 (B3), 1049-1061, 1979)

45
 Another interesting implication of this correlation is the seismic potential of the

Pacific Northwest (i.e., Oregon-Washington coast). The Juan de Fuca plate is subducting

beneath the states of Oregon and Washington. The background seismicity there is very

low, as shown below, and until mid 1980's, it was generally believed that the seismic

potential in the Pacific Northwest is low (i.e., great earthquakes are unlikely). However,

the age of the Juan de Fuca plate is very young, about 10 My, and it is subducting at a

rate of 3 cm/year. Thus, in view of the regression relation shown above, one would

expect a large, M w =8.5 to 9, earthquake there. This suggestion motivated the interest of

geologists who started extensive investigation for finding palaeo-seismological evidence.

Geological evidence for regional submergence and evidence for large tsunami which

occurred in 1700 [Satake et al., 1996] now seem to have convinced most people, which

seems to have led to upgrading of building code in the area. This is a good example in

which seismological study, even if it is poorly constrained, can be useful if it is followed

up by investigations from different disciplines.

46
(Heaton, T., and H. Kanamori, Seismic potential associated with subduction in the
northwestern United States, Seismol. Soc. Am. Bull., 74 (3), 933-941, 1984)

47
(see, Atwater, B. F., and others, Summary of coastal geologic evidence for past great
earthquakes at the Cascadia subduction zone, Earthquake Spectra, 11, 1-18, 1995)

48
(R. S. Yeats, Living with Earthquakes in the Pacific Northwest, Oregon State University
Press, 1998)

References

Brune, J., Seismic moment, seismicity, and rate of slip along major fault zones, J.
Geophys. Res., 73, 777-784, 1968.

Satake, K., Shimazaki, K., Tsuji, Y. and Ueda, K., Time and size of a giant earthquake in
Cascadia inferred from Japanese tsunami records of January 1700, Nature, 379, 246-249,
1996.

49
Ge 162

2.2 Location, Magnitude, and Mechanism of Earthquakes

Locating Earthquakes (LW, pp. 217-235)

Referring to Figure 1, the arrival time of P wave at station i can be written as,

ti = f (ro , ri , v(r )) + to (1)

where the function f ( ) gives the travel time between the source and station i.

50
Here, ro is the hypocenter location (i.e., the location of the beginning of an earthquake),

ri is the location of the i-th station, v(r ) is the wave speed which is in general a function

of the position, and to is the origin time. For a homogeneous medium,

[( xi − xo ) 2 + ( yi − yo ) 2 + ( zi − zo ) 2 ]1/ 2
f (ro , ri , v(r )) = (2)
v

We want to determine 4 unknowns, ro = ( xo , yo , zo ) and to , from N observations,

t1 , t2 , t3 , t4 , ...t N .

51
 This problem is a nonlinear problem even for the simplest case for a
homogeneous medium.
To solve a nonlinear problem like this, we start from a first approximation

ro0 = ( xo0 , yo0 , zo0 ) and to0 (3)

Then writing

xo = xo0 + δ xo etc, and to = to0 + δ to (4)

and taking the first order terms in δ xo etc and δ to , we set up linear equations for δ xo etc

and δ to as,

∂f (ri ) ∂f (ri ) ∂f ( ri )
ti − ti0 = δ x0 + δ y0 + δ z0 + δ t0 (5)
∂xo 0 ∂yo 0 ∂zo 0

(i=1, 2. 3, .....N), where ti0 is the arrival time at station i computed for the first

approximation (i.e., ti0 = to0 +(travel time computed for the first approximation)).

This problem can be solved by the method of least squares. By iterating this, we
can determine ro = ( xo , yo , zo ) and to which best fit the observed travel times. More
details will be discussed in the practice session.

52
Earthquake Magnitude (LW, pp. 379-385)

Traditionally, magnitude scales are used to indicate the size of an earthquake.

The magnitude M is determined from the amplitude of the observed seismic waves. In
general,

M = log( A) + f (Δ) (6)

where A is the amplitude of the observed seismic waves (body waves, surface waves, or
unspecified), and f (Δ ) is an amplitude attenuation curve as a function of distance
determined for specific type of waves.
For example, in case of the traditional local magnitude M L , the amplitude A is the
amplitude of the Wood Anderson seismogram in mm, and f (Δ) is given by a table, or by
a nomogram such as that shown in Figure 2.

53
 In case of the surface-wave magnitude, M S , A is the ground-motion amplitude of

20 sec (period) surface waves in μ observed at a distance of Δ deg. Then,

M S = log A + 1.656 log(Δ ) + 1.818 (7)

Although the magnitude is a useful parameter, it is difficult to attach some

specific physical quantity to it. Empirically, M S is related to the total wave energy, ER ,
radiated from an earthquake by

54
 logER = 1.5M S + 4.8 (joule) (8)

In more recent studies, we use another scale, M W , which is determined from

seismic moment , M 0 . The seismic moment M 0 is given by

M 0 = μ DS (9)

where D is the fault offset, S is the fault area, and μ is the rigidity of the crust
surrounding the fault. The unit of M 0 is N-m. Unlike other magnitude scales, M W
represents a specific earthquake source parameter, the overall static size of an earthquake
given by M 0 . The relation between M w and M 0 is given by

log M 0 = 1.5M w + 9.1 (10)

This will be discussed in more detail later.

Earthquake Mechanism (LW, pp.346-356)

As shown in Figure 3, the fault motion on a vertical strike slip fault would
produce compressional and dilatational quadrants in the Earth's crust.

55
 The pattern of compression and dilatation can be detected by the first motion of P
waves. In the compressional quadrant P wave is up and in the dilatational quadrant, it is
down. As viewed from above, the sense of the first motion (up or down, or compression
or dilatation) alternates in quadrant. The planes separating the compressional quadrant
and the dilatational quadrant are called the nodal planes. The fault plane coincides with

56
one of the nodal planes. The other nodal plane is called the auxiliary plane. With this
method alone, we cannot distinguish the fault plane from the nodal plane, i.e., the fault
plane can be either one of the nodal planes.
Thus, from the observations of P wave first-motion data from many stations
surrounding the source we can determine the geometry of the faulting. The pattern thus
determined is usually referred to as "Mechanism of Earthquake". Faults with different
types (e.g., strike slip fault, thrust fault, and normal fault) produce different radiation
patterns of P waves. This can be easily seen, if we consider a small sphere surrounding
the source. This sphere is called the focal sphere (Figure 3).
The radiation pattern is three dimensional, and the surface of the focal sphere is
divided into quadrants of compression and dilatation. (This can be best understood using
a worn-out tennis ball with compressional quadrants painted dark.)
We need to show this three-dimensional pattern on the focal sphere on a piece of
paper. Since the pattern is point symmetric with respect to the center of the focal sphere,
we need to show only the pattern on a hemisphere. It is conventional to show the lower
focal hemisphere, but in rare cases, the upper hemisphere or the side hemisphere is
shown. We use a standard projection method, most commonly the equal-area
stereographic projection, to project the lower hemisphere to a flat horizontal plane
(Figure 4).

57
 Some examples are shown in Figure 5. The projected diagram is called the
mechanism diagram. As shown in Figure 5, the stereographic mechanism diagram is
intuitive for understanding the geometry of faulting (more details in the practice session).

58
 Figure 6 shows the mechanism of large earthquakes along the Circum-Pacific
belt. Most of them are low-angle thrust mechanisms which are consistent with subduction
of the Nazca and the Pacific plates beneath the South American, the North American, and
the Eurasian plates. The normal fault events represents tensional failure within the
oceanic plate upon bending caused by subduction.

59
Ge 162 Practice Session 1 Locating Earthquakes

60
Data

Table-1 is the travel time data obtained using "STP" which is a SCEC (Southern
California Earthquake Center) tool to extract earthquake data (parameter data and
waveform data). (For details of STP, see the SCEC Web site). The data are rearranged
in (x, y) coordinate (in km) with the Pasadena station (Latitude=34.1484°, Longitude=-
118.1711° ) as the origin. The origin time of the arrival times is arbitrary. Table-1 gives
only the first 10 stations. The actual data are in loc_dat_1 (all the data and program files
are in a FTP site on ftp.gps.caltech.edu /home/ftp/pub/hiroo/ge162.dir).

Table 1.
Earthquake-1
Reference Station (Origin of (x,y)) PAS 34.1484 -118.1711

Station Lat. Long. Elev. x(km) y(km) arrival t.(sec)

MIK 34.1370 -118.1260 235.0000 4.160 -1.264 1.453
GSA 34.1368 -118.1283 165.0000 3.947 -1.286 1.453
KIK 34.1504 -118.1016 168.0000 6.409 0.224 1.596
PAS 34.1484 -118.1711 257.0000 0.000 0.000 1.615
GVR 34.0497 -118.1200 141.0000 4.718 -10.947 1.620
RUS 34.0505 -118.0799 37.0000 8.420 -10.856 1.753
RIO 34.1047 -117.9796 79.0000 17.670 -4.831 2.778
MWC 34.2237 -118.0529 1696.0000 10.891 8.359 2.731
LGB 33.9756 -118.1491 6.0000 2.033 -19.167 3.417
TCC 33.9947 -118.0140 195.0000 14.514 -17.038 2.731
GR2 34.1183 -118.2994 316.0000 -11.837 -3.331 3.134
...... .................. ...................... .................... ................. ................. ...........

Locate the earthquake taking the following steps. Ideally, you should write your
own program to carry out 2 to 6, but if you find it difficult to do so, you can use a simple
program, eqloc.f . To do 7, you will need to use this program. In case you use this
program, try to follow the steps taken in the program.

61
1. Use the first approximation, (0.0, 0.0, -10.0, 0.0).

2. Compute the travel times and the partial derivatives in a homogeneous medium with
v=6 km/s which is a good average for the shallow crust. Refer to equation (1) to (5) in
class note 2.2.

∂f (ri ) ∂f (ri ) ∂f (ri )

ti0 , , and (i=1, 2. 3, .....N) (1)
∂xo 0 ∂yo 0 ∂zo 0

3. Set up the equation for the least-square solution.

⎛ ∂f (r1 ) ∂f (r1 ) ∂f (r1 ) ⎞

⎜ 1⎟
⎜ ∂xo 0 ∂yo 0 ∂zo 0 ⎟
⎜ ⎟
⎜ ∂f (r2 ) ∂f (r2 ) ∂f (r2 )
1⎟
⎛ t1 − t10 ⎞
⎜ ∂xo 0 ∂yo 0 ∂zo 0 ⎟ ⎜ ⎟
t − t0
⎜ ⎟ ⎛ δ x0 ⎞ ⎜ 2 20 ⎟
⎜ ∂f (r3 ) ∂f (r3 ) ∂f (r3 ) ⎟⎜ ⎟ ⎜ t3 − t3 ⎟
⎜ 1 ⎟ ⎜ δ y0 ⎟ ⎜ ⎟
=⎜ . ⎟ (2)
⎜ ∂xo 0 ∂yo 0 ∂zo 0 ⎟ ⎜ δ z0 ⎟
⎜ ⎜ ⎟
. . . . ⎟ ⎜⎜ δ t ⎟⎟ ⎜ . ⎟
⎜ ⎟⎝ 0 ⎠
⎜ . . . .⎟ ⎜ . ⎟
⎜ ⎜ t − t0 ⎟
. . . .⎟ ⎝ N N⎠
⎜ ⎟
⎜ ∂f (rN ) ∂f (rN ) ∂f (rN ) ⎟
⎜⎜ 1⎟
⎟
⎝ ∂xo 0 ∂yo 0 ∂zo 0 ⎠

62
which we write as

Am = d (3)

where A is an Nx4 matrix

⎛ ∂f (r1 ) ∂f (r1 ) ∂f (r1 ) ⎞

⎜ 1⎟
⎜ ∂xo 0 ∂yo 0 ∂zo 0 ⎟
⎜ ⎟
⎜ ∂f (r2 ) ∂f (r2 ) ∂f (r2 )
1⎟
⎜ ∂xo 0 ∂yo 0 ∂zo 0 ⎟
⎜ ⎟
⎜ ∂f (r3 ) ∂f (r3 ) ∂f (r3 ) ⎟
1⎟
A = ⎜ ∂xo ∂yo 0 ∂zo 0 (4)
⎜ 0 ⎟
⎜ . . . .⎟
⎜ ⎟
⎜ . . . .⎟
⎜ . . . .⎟
⎜ ⎟
⎜ ∂f (rN ) ∂f (rN ) ∂f (rN ) ⎟
⎜⎜ 1⎟
⎟
⎝ ∂xo 0 ∂yo 0 ∂zo 0 ⎠

and m and d are column vectors containing the parameters to be determined and the
data, respectively, i.e.,

63
 ⎛ t1 − t10 ⎞
⎜ 0 ⎟

⎛ δ x0 ⎞ ⎜ t2 − t2 ⎟
⎜ ⎜ t3 − t30 ⎟
⎜ δ y0 ⎟⎟ ⎜ ⎟
m= , and d =⎜ . ⎟ (5)
⎜ δ z0 ⎟
⎜⎜ ⎟⎟ ⎜ . ⎟
⎝ δ t0 ⎠ ⎜ ⎟
⎜ . ⎟
⎜t − t0 ⎟
⎝ N N⎠
4. Determine m .

The normal equation is,

AT Am = AT d (6)

If AT A is not singular, the formal solution is given by

m = ( AT A) −1 AT d (7)

and the error estimates are determined by the variance of the data and the diagonal
elements of the inverse matrix of the normal equation (6 ). Usually, we write the
uncertainty in mi by Δmi , and compute it by

∑ (t − t ) /( N − N ) ,
N

Δmi = cii j
c 2
j p i=1, 2, 3 (8)
j =1

64
where N p is the number of parameters (here 4), cii are the diagonal elements of ( AT A) −1 ,

and t cj are the computed arrival times for station j.

5. Obtain the 2nd approximation by:

x0 = 0 + δ x0

y0 = 0 + δ y0 (9)

z0 = −10. + δ z0

t0 = 0 + δ t0

6. Iterate 2, 3, 4, 5.

A simple program eqloc.f is in /home/ftp/pub/hiroo. This program uses 2 input

files.

i_eqloc and c_eqloc

i_eqloc
loc_dat_1 : Name of the travel-time data file
0.0 0.0 -10.0 0.0 : 1st approx.
c_eqloc
10 : maximum number of interations
half_space
: name of the structure
s-cal.pvel

half_space
half_space

65
1 : number of layers
9999. 6.0 : layer thickness, α
s-cal.pvel
southern cal. P structure, with a slightly low surface velocity
5
1.0 4.0
3. 5.5
23.4 6.3
5. 6.8
9999.0 7.8

7. If the second line of c_eqloc is replaced by s-cal.pvel, it will use a more general
subroutine which computes the travel times etc for a layered model given by s-cal.pvel.
Try eqloc.f with s-cal.pvel. Output of eqloc is in o_eqloc.

8. Vary the 1st approximation to see how the solution is affected.

9. Remove the data for which ti is less than 5 sec, and locate the event.

Example output
o_eqloc

10
southern california P structure, with a
1.000 4.000
3.000 5.500
23.400 6.300
5.000 6.800
9999.000 7.800
loc_dat_1
0.000 0.000 -10.000 0.000
Earthquake-1
PAS 34.148 -118.171
x0= 6.778 0.319
y0= -4.360 0.378

66
z0= -10.676 1.159
th0= 0.017 0.054
34.148 -118.171 -10.000 0.000
x0= 6.790 0.316
y0= -4.186 0.381
z0= -8.184 1.079
th0= 0.027 0.054
34.109 -118.098 -10.676 0.017
x0= 6.782 0.308
y0= -4.172 0.367
z0= -7.645 1.180
th0= 0.030 0.048
34.111 -118.098 -8.184 0.027
x0= 6.782 0.307
y0= -4.167 0.365
z0= -7.533 1.217
th0= 0.032 0.047
34.111 -118.098 -7.645 0.030

...........
station data

Station Dist.(km) Azm.(deg) ti ti(computed) Res.

MIK 3.911 317.927 1.453 1.552 -0.099
GSA 4.040 315.481 1.453 1.563 -0.110
KIK 4.404 355.193 1.596 1.594 0.002
PAS 7.958 301.591 1.615 1.983 -0.368
GVR 7.088 196.979 1.620 1.876 -0.256
RUS 6.888 166.284 1.753 1.853 -0.100
RIO 10.909 93.540 2.778 2.377 0.401
MWC 13.179 18.203 2.731 2.703 0.028
LGB 15.734 197.612 3.417 3.083 0.334
TCC 15.014 149.045 2.731 2.975 -0.244
GR2 18.637 272.606 3.134 3.525 -0.391
WLT 17.605 129.626 3.714 3.367 0.347
USC 20.145 239.757 3.804 3.756 0.048
WTT 23.123 219.086 4.858 4.218 0.640
PEM 21.884 73.425 3.989 4.026 -0.037

...........
0.3273 RMS of residuals

Ge 162 Practice Session 2 P-wave First-Motion Mechanism

67
Data
Table-1 is the phase data obtained using "STP" which is a SCEC tool to extract
earthquake data (parametric data and waveform data). (For details of STP, see the SCEC Web
site). The data contain station names, first motion data (C or D), quality (Q, ignore in this
problem), azimuth, and take-off angle. The azimuth and take-off angle, ih, (measured from
downward vertical) are computed using a standard southern California structure. Table-1 lists
only selected 10 stations. The actual data are in mech_dat_1 in
/home/ftp/pub/hiroo/ge162.dir/practice2.dir.

Table 1. P-wave first-motion data

station C or D Q Azimuth (°) ih (°)
BVH D 2 340.332 164.798
SMS D 2 231.855 131.689
DJJ C 2 310.759 131.679
USC C 2 114.973 123.098
LAX D 2 190.413 116.500
HLL D 2 11.487 115.842
PAS D 2 63.638 103.096
NOT C 2 320.286 101.493
MWC D 2 59.358 96.635
LKL D 2 39.880 92.397
.................................................................

Determine the mechanism of this earthquake taking the following steps.

1. Compute the radial distance on a mechanism diagram using the equal-area projection,

ri = 2 sin(ihi / 2) (equal-area projection) (1)

68
 Remember that if ih is larger than 90°, then 180° must be added to the azimuth (i.e., the
station must be plotted in the opposite azimuth.), and ih must be changed to 180°-ih.

2. Plot the first-motion (filled circle for compression and open circle for dilatation) on a
mechanism diagram. (Plot the data for the 10 stations listed in Table-1 manually, or with your
own program).

3. A program mplotr_2.f is provided in practice_2.dir to plot the first-motion data on a

mechanism diagram. mplotr_2.f takes 2 input files, i_mplotr_2 and c_mplotr_2. The
output is a postscript file p_mplotr_2.

i_mplotr_2
Earthquake(mech)-1 : Job ID
f : fault(f) or moment tensor(m)
85. 180. 350. : dip, rake, and fault strike
mech_dat_1 : file name of the data

(For plotting only the first-motion data, the 2nd and 3rd lines are irrelevant.)
c_mplotr_2
7.0 0.1 10. 5.0 0 0 1 0 : these parameters control the type and
style of the plot
(change plot_opt parameter (2nd from the last) only. 1 for the first-motion
data only (in this case, the fault parameters are ignored), 3 for the data +
nodal lines)

(radius, sizef, delt1c, delt2c, shading p, sv or sh, ndens,

plot_opt., projection)

69
4. The final step is to determine the mechanism by drawing two orthogonal nodal planes so that
they divide the compressional and dilatational stations. Usually there are always some
inconsistent stations, but try to find the best solution. Many methods have been developed, but
here try a few mechanisms. mplotr_2.f draws 2 nodal lines corresponding to the fault
mechanism given by dip, δ , rake, λ , and fault strike, φ f (run mplotr_2 with plot_opt=3 in

c_mplotr_2). These fault parameters are defined in the figure below.

70
5. Assume that the fault strike is -45° from N. Plot the mechanism diagrams for (1) right-
lateral vertical strike slip, (2) vertical dip slip (north-east side down), (3) thrust fault
dipping 20° NE, (4) normal fault dipping 45° NE, and (4) an oblique-slip mechanism
(e.g., add some right-lateral component to (3)). (This problem has nothing to do with

71
the first-motion data given in mech_dat_1. You can just draw a sketch of mechanism
diagrams, or run mplotr_2.f with
c_mplotr_2
7.0 0.1 10. 5.0 0 50 4 0
.)
Ge162

2.3. Seismicity of the Earth (LW, pp.434-477)

2.3.1 Global Seismicity

Earthquakes occur mainly:

1. Along trenches (subduction zones),

2. Along ridge-transform systems,

3. In continental interiors.

Figures 1 and 2 show seismicity in the world and California, and Figures 3 and 4 show

the mechanisms of large earthquakes in the world and California.

72
73
 The world largest earthquakes occur along subduction zones (e.g., the 1960 Chilean

earthquake, Mw=9.5, the 1964 Alaskan earthquake, Mw=9.2). More than 75 % of the seismic

energy release takes place there. Most of these events represent slip on the interface between a

subducting oceanic plate and an overriding plate (Figure 3). In these zones deep focus

earthquakes occur to a depth of about 700 km.

74
75
 Most earthquakes along ridge-transform systems are shallow and relatively small (10% in

energy release). The events on ridges have normal-fault mechanism, and those along the

transform boundaries have strike-slip mechanisms. Transform fault events are generally larger;

occasionally the magnitude reaches 8 (e.g., the 1906 San Francisco earthquake, Mw≅8 ). No deep

focus earthquake has been found beneath ridge-transform systems.

The distribution of the events in continental interiors (15 % in energy release) is

very diffuse. Partly because of their proximity to major population centers, large intra-

76
continental events are often very devastating (e.g., the 1976 Tanshang earthquake, the

1988 Armenian earthquake, the 2001 India earthquake)

2.3.2 Depth Variation (Figure 5)

Most earthquakes occur at depths shallower than 60 km. However, some

significant earthquakes occur at depths of as deep as 700 km, especially in subduction

zones (e.g. 1970 Colombia, d=653 km, Mw=8.1; 1954 Spain, d=640 km, M=7.8, 1994

Bolivia, d=635 km, Mw=8.3). The largest recorded deep focus earthquake is the 1994

Bolivia earthquake.

77
 A pronounced peak in the energy release is seen at a depth of 600 km, just before

the seismic activity dies out.

2.3.3 Temporal Variation of Seismicity (Figure 6)

Figure 6 shows the temporal variation of seismicity. The energy release during the

period from 1952 to1965 dominates. This peak is a result of five large subduction-zone

earthquakes in the Pacific (1952 Kamchatka, 1957 Aleutian Is., 1960 Chile, 1964 Alaska,

78
1965 Aleutian Is.). The energy release rate is not uniform in time, and fluctuates on a

time scale of at least 100 years.

The average energy release is about 4.5x1024 ergs/year, which is approximately

1/4 of the energy released in volcanic eruptions, and 0.05 % of the terrestrial heat flow.

Figures 7, 8, and 9 show the distribution of great earthquakes, the temporal

distribution of damaging earthquakes, and the temporal distribution of damaging tsunami.

79
80
81
2.3.4 Magnitude-Frequency Relation (Gutenberg-Richter Relation)

Figure 10 shows the distribution of the number, N, of earthquakes equal to, or

larger than, magnitude M. In general, this distribution is expressed as

log N ( M ) = a − bM

82
 The results obtained for many regions indicate that the value of b (called b value)

is approximately equal to 1.

Figure 9 shows that approximately 1 earthquake with M ≥ 8 occurs every year.

2.3.5 Aftershocks

83
 After a large earthquake (main shock), many smaller earthquakes occur near the

epicenter of the earthquake. The decay of aftershock activity follows the Omori's law

given by

K
n(t ) =
t +c

where n(t ) is the number of aftershocks larger than a given magnitude per unit time. A

modified Omori's law is given by

K
n(t ) =
(t + c) p

where p is a constant, which is usually slightly larger than 1.

3.1 Review of Elasticity Theory

3.1.1 Stress

In the theory of elasticity we consider two types of forces.

Body Force

The body force f is defined by a force per unit mass in a medium (Figure 1).

Then, the body force per unit volume is ρ f where ρ is the density of the medium.

84
 Then the force acting on a volume element dV is

ρ fdV

which can be integrated over a volume V as

∫ ρ f dV
V

e.g. body force due to gravity: ρ f = ρ g (downward)

Surface Force

85
 The surface force is the force distributed over a surface of the body, either internal
or external. Usually it is defined by the force per unit area, f S . Then the force acting on a

surface element dS is f S dS .

Stress
Consider a deformed elastic body in equilibrium (see Figure 1). Let dS be a surface
element at P which divides the medium on the + side and the - side. A unit normal vector
n ( n = 1 ) is taken from the - side to the + side. In equilibrium, the force, F+ , exerted by

the + side on dS should be balanced by the force, F− exerted by the - side on dS , i.e.,

F+ + F− = 0 .

The stress at P acting on dS from the + side is then defined by

F+
f n = lim
dS →0 dS

f n is a vector (often called a stress vector, or traction) and its dimension is force/area.

We take a Cartesian coordinate system ( x1 , x2 , x3 ) and write x1 , x2 , and x3

components of f n by ( f1n , f 2 n , f3n ) . Note that f n is a function of not only the location

of P but also the orientation of n . Hence, in order to specify the stress at P uniquely, we
need two vectors n and f n .

This situation can be understood more easily in the simple example shown in
Figure 2.

86
 Consider an elastic beam with one end, AB, clamped at the wall. Then, apply a
force F uniformly on the surface S at the other end BC. Let us consider the stress at P. In
(a), we consider the stress acting on S1 that is perpendicular to the axis of the beam. In
this case, it is obvious that

( f1n , f 2 n ) = ( F / S1 , 0)

In (b), we consider S2 which is parallel to the axis. Suppose we cut the beam in two

parts along S2. The beam will be still in equilibrium without change in shape. That is, there

is no force acting on S2. Hence, in this case,

( f1n , f 2 n ) = (0, 0)

Stress Tensor

87
 For simplicity, we consider a 2-dimensional problem. The section shown in
Figure 3 depicts a 2-D medium extending to infinity in x3 direction (perpendicular to the
face of the paper).

Consider a surface element dS1 normal to the x1 axis. We call the medium on the
+ x1 side M+ and that on the - x1 side, M- (Figure 4). Let σ 11 and σ 21 be the x1 and

x2 components of the stress exerted by M+ on dS1 . Then −σ 11 and −σ 21 are x1 and x2

components of stresses exerted by M-, respectively.

88
 Note that the first subscript denotes the component, and the second, the direction
of the normal to the surface. σ ij ' s with any i and j can be defined similarly.

Consider a triangular beam with an infinitesimally small cross section BOA .

Referring to Figure 5, we define the following.

n1 : x1 component of n , n1 = cos θ
n2 : x2 component of n , n2 = sin θ

σ 1n : x1 component of the stress acting on AB

σ 2 n : x2 component of the stress acting on AB

S: AB x unit length in x3 direction

S1: OB x unit length in x3 direction
S2: OA x unit length in x3 direction

89
Then the forces acting on BOA are given as follows.

x1 component x2 component
force on S1 −σ 11S1 −σ 21S1

force on S2 −σ 12 S 2 −σ 22 S 2

force on S σ 1n S σ 2n S

Total − σ 11S1 − σ 12 S2 + σ 1n S − σ 21S1 − σ 22 S2 + σ 2 n S

In equilibrium, the total force should vanish. Since S1 = S cos θ = Sn1 , and
S 2 = S sin θ = Sn2 , we obtain

σ 1n = σ 11n1 + σ 12 n2
σ 2 n = σ 21n1 + σ 22 n2 (1)
or, in matrix notation,

90
 ⎛ σ 1n ⎞ ⎛ σ 11 σ 12 ⎞ ⎛ n1 ⎞
⎜ ⎟=⎜ ⎟⎜ ⎟ (2)
⎝ σ 2 n ⎠ ⎝ σ 21 σ 22 ⎠ ⎝ n2 ⎠

Consider an infinitesimally small square around P (Figure 6). In equilibrium, the

total moment around P should vanish. This condition leads to

σ 12 = σ 21 (3)

The above analysis can be extended easily to a 3-dimensional

problem, and we can derive,

⎛ σ 1n ⎞ ⎛ σ 11 σ 12 σ 13 ⎞ ⎛ n1 ⎞
⎜ ⎟ ⎜ ⎟⎜ ⎟
⎜ σ 2 n ⎟ = ⎜ σ 21 σ 22 σ 23 ⎟ ⎜ n2 ⎟ (4)
⎜σ ⎟ ⎜σ ⎟⎜ ⎟
⎝ 3n ⎠ ⎝ 31 σ 32 σ 33 ⎠ ⎝ n3 ⎠

with the symmetry relations σ ij = σ ji . Here, σ in (i=1,2,3) are the x1 , x2 , and x3

components of the stress acting on dS (at P) whose unit normal is n .

The matrix ( σ ij ) above is called the stress tensor at P. (It can be shown that ( σ ij ) is a

tensor.) Because of the symmetry σ ij = σ ji , it has only 6 independent elements. Once we

91
know the stress tensor ( σ ij ) at P, we can calculate, using (4), the stresses on any surface

at P whose unit normal is n (Figure 7).

In the above equilibrium analysis, we ignored body forces compared with surface
forces. This is justified because if we consider a small volume around P with a linear
dimension da, then the total body force is proportional to da3 while the total surface force
is proportional to da2. Hence as da → 0, the body forces can be ignored.

Normal, Shear, Principal Stresses

We resolve the stress acting on dS into two components, one parallel to n , σ nn ,

and the other perpendicular to n , σ nt (see Figure 8). σ nn is called the normal stress, and

σ nt is the shear stress. In general, σ nn ≠ 0 and σ nt ≠ 0 . However, at any point P, it is

always possible to choose a Cartesian coordinate system ( x1' , x2' , x3' ) for which

σ i ' j ' = 0 (i ' ≠ j ') . In other words, for this new coordinate system ( σ i ' j ' ) is a diagonal

matrix. The non-zero diagonal components, σ 1'1' , σ 2'2' , and σ 3'3' are called the principal

92
stresses, and x1' , x2' , and x3' axes, the principal axes. It can be shown that

σ 1'1' + σ 2'2' + σ 3'3' = σ 11 + σ 22 + σ 33 .

Equations of Motion

First, consider a 1-dimensional problem illustrated in Figure 9. Figure 9 shows an

elastic rod with cross section S.

The equation of motion for a small part of it, ABCD (length dx1 ) is given by,

93
 ρ Sdx1u1 = ρ f1Sdx1 + σ 11 ( x1 + dx1 ) S − σ 11 ( x1 ) S

where ρ is the density, u1 is the displacement, and f1 is the body force. Expanding
σ 11 ( x1 + dx1 ) around x1 , and retaining only the first order terms in dx1 , we obtain,

∂σ 11
ρ u1 = ρ f1 + (5)
∂x1

This is the equation of motion in 1-dimension.

A similar analysis can be made for a 3-dimensional geometry, and we obtain:

3 ∂σ ij
ρ ui = ρ fi + ∑ (i=1,2,3) (6)
j =1 ∂x j

These are the equations of motion expressed in terms of the stress components.

Tensor Notation

In tensor notation, if any suffix occurs twice in a single term, it is to be put equal
to 1, 2, and 3 in turn and the results are to be added. For example,

3
aii = a11 + a22 + a33 = ∑ aii
i =1

a1 j b j 3 = a11b13 + a12b23 + a13b33 = a1l bl 3

3
aip bpj = ∑ aik bkj
k =1

94
Also, we use ,j to denote differentiation by x j . For example,

∂ui
= ui , j
∂x j

Also note that,

∂ul ∂u1 ∂u2 ∂u3

ul ,l = = + + = u1,1 + u2,2 + u3,3
∂xl ∂x1 ∂x2 ∂x3

We also use the Kronecker’s delta

⎧1 if i = j
δ ij = ⎨
⎩0 if i ≠ j

Note that δ ii = 3 .

In the tensor notation, (6) can be written as

ρ ui = ρ fi + σ il ,l (7)

Boundary Conditions

From the definition of the stress, it is evident that the normal and the tangential
stresses should be continuous across any surface. In particular, at the free surface, there is
no force acting on it; hence the normal and the shear stresses should vanish there. If the
free surface is perpendicular to x3 axis,

95
 σ 31 = σ 32 = σ 33 = 0

Within the interior of continuum, the displacement ui should be continuous.

3.1.2 Strain

The displacement in continuum consists of three parts,

1. Translation (rigid body)

2. Rotation (rigid body)
3. Deformation

In the theory of elasticity, we are primarily concerned with “Deformation”.

Referring to Figure 10, let P and Q represent two points dx apart in an

elastic medium, and u and u + du be their displacements.

If u = u + du , i.e., du = 0 , for any pair of points, u simply represents

96
rigid body translation. Thus du is considered to represent rotation and deformation. To
the first order,

⎛ ∂u1 ∂u1 ∂u1 ⎞

⎜ ⎟
∂x ∂x2 ∂x3 ⎟
⎛ du1 ⎞ ⎜ 1 ⎛ dx1 ⎞
⎜ ⎟ ⎜ ∂u2 ∂u2 ∂u2 ⎟ ⎜ ⎟
⎜ du 2 ⎟ = ⎜ ∂x ∂x2
⎟ dx2 ⎟
∂x3 ⎟ ⎜⎜
(8)
⎜ du ⎟ ⎜ 1 ⎟
⎝ 3⎠ ⎜ ⎝ dx3 ⎠
∂u ∂u3 ∂u3 ⎟
⎜⎜ 3 ⎟
⎝ ∂x1 ∂x2 ∂x3 ⎟⎠

Denoting the matrix ( ui , j ) by U and ( u j ,i ) by UT (transpose of U), we can write U as

1 1
U = (U + U T ) + (U − U T ) (9)
2 2

1 1
We denote (U + U T ) by D and (U − U T ) by R. D is symmetric and R is anti-
2 2
symmetric. We will show that D represents deformation, and R represents rigid-body
rotation.

1
The elements of D, (ui , j + u j ,i ) = eij can be interpreted as follows.
2
If only e11 ≠ 0 ,

⎛ du1 ⎞ ⎛ e11 0 0 ⎞ ⎛ dx1 ⎞

⎜ ⎟ ⎜ ⎟⎜ ⎟
⎜ du2 ⎟ = ⎜ 0 0 0 ⎟ ⎜ dx2 ⎟
⎜ du ⎟ ⎜ 0 0 0 ⎟ ⎜ dx ⎟
⎝ 3⎠ ⎝ ⎠⎝ 3 ⎠

We then have, du1 = e11dx1 , du2 = 0 , and du3 = 0 . This means that the line element dx1 in

97
x1 direction is stretched by du1 = e11dx1 in x1 direction. Hence, e11 represents extension
(or contraction if e11 <0) per unit length in x1 direction. e22 and e33 can be interpreted
similarly.
Next consider e12 , and e21 (= e12 ).

⎛ du1 ⎞ ⎛ 0 e12 0 ⎞ ⎛ dx1 ⎞

⎜ ⎟ ⎜ ⎟⎜ ⎟
⎜ du2 ⎟ = ⎜ e21 0 0 ⎟ ⎜ dx2 ⎟
⎜ du ⎟ ⎜ 0 0 0 ⎟⎠ ⎜⎝ dx3 ⎟⎠
⎝ 3⎠ ⎝

i.e., du1 = e12 dx2 , du2 = e21dx1 , and du3 = 0 . As shown in Figure 11, the angle between x1

and x2 axes which is originally π / 2 becomes θ = π / 2 − 2e12 after the deformation.

Thus, e12 is equal to half the angle change between the x1 axis and x2 axis, and is called
the shear strain. e13 and e23 can be interpreted similarly.

98
 D is called the strain tensor (it can be shown that ( eij ) is a tensor).

e11 , e22 , and e33 represent extension or contraction, and e12 , e13 , and e23 represent shear.

We now show that R represents rigid body rotation. R can be written as

⎛ 1 1 ⎞
⎜ 0 (u1,2 − u2,1 ) (u1,3 − u3,1 ) ⎟
2 2
⎜ ⎟
1 1
R = ⎜ (u2,1 − u1,2 ) 0 (u2,3 − u3,2 ) ⎟
⎜2 2 ⎟
⎜ ⎟
⎜⎜ 1 (u3,1 − u1,3 ) 1
(u3,2 − u2,3 ) 0 ⎟⎟
⎝2 2 ⎠

Define ω1 , ω 2 , and ω3 by

ω1 = −r23 (= r32 ) = −(1/ 2)(u2,3 − u3,2 )

ω 2 = −r31 (= r13 ) = −(1/ 2)(u3,1 − u1,3 )
ω3 = −r12 (= r21 ) = −(1/ 2)(u1,2 − u2,1 )

Then,

⎛ 0 −ω3 ω2 ⎞
⎜ ⎟
R = ⎜ ω3 0 −ω1 ⎟
⎜ −ω ω1 0 ⎟⎠
⎝ 2

Consider the case where ω1 = ω 2 = 0 and ω3 is non-zero.

99
Then, we have

⎛ du1 ⎞ ⎛ 0 −ω3 0 ⎞ ⎛ dx1 ⎞

⎜ ⎟ ⎜ ⎟⎜ ⎟
⎜ du2 ⎟ = ⎜ ω3 0 0 ⎟ ⎜ dx2 ⎟
⎜ du ⎟ ⎜ 0 0 0 ⎟⎠ ⎜⎝ dx3 ⎟⎠
⎝ 3⎠ ⎝

i.e., du1 = −ω3dx2, du2 = ω3dx1, and du3 = 0 . As shown in Figure 12, this displacement

represents counter-clockwise rotation by ω3 around the x3 axis. ω1 and ω 2 can be

interpreted similarly. Thus R represents rigid body rotation.

3.1.3 Stress-Strain Relations and Elastic Constants

The relation between stress and strain is the extension of the Hooke’s law for a
spring. (i.e., F = k Δl where F is the force, Δl is the length change and k is the spring
constant.)
We assume that the material is isotropic and perfectly elastic. If the medium is

100
perfectly elastic, the stress should be expressed as a homogeneous linear function of
strain. Since there are six independent stress components and strain components, in
general there can be 6x6=36 constants. However, if the material is isotropic, we can show
that there are only two independent constants.
Consider an elastic parallelepiped shown in Figure 13. Apply a normal stress σ 11
in x1 direction. The extension in x1 direction is e11 (Figure 13). In the linear theory, e11 is
proportional to σ 11

1
e11 = σ 11
E

1
The constant of proportionality is written as , and is called the Young’s
E
modulus. Note that e11 is non-dimensional, so that E has the dimension of stress. Under
this stress, there will be contraction in x2 and x3 directions that is proportional to e11 .
Since the material is isotropic,

ν ν
e22 = −ν e11 = − σ 11 and e33 = −ν e11 = − σ 11
E E

101
ν is called the Poisson’s ratio.
If we apply σ 11 , σ 22 , and σ 33 in x1 , x2 and x3 directions simultaneously, then by
superposition,

σ 11 ν
e11 = − (σ 22 + σ 33 )
E E
σ 22 ν
e22 = − (σ 11 + σ 33 ) (10)
E E
σ 33 ν
e33 = − (σ 11 + σ 22 )
E E

Adding the three equations in (10) we obtain

(1 − 2ν )
Δ= Σ (11)
E

where

Δ = e11 + e22 + e33 (12)

and

Σ = σ 11 + σ 22 + σ 33

Consider deformation of a small parallelepiped. The initial volume is

V0 = dx1dx2 dx3 and the volume after deformation is

V = (1 + e11 )dx1 (1 + e22 )dx2 (1 + e33 )dx3

102
 Then, to the first order, the volume change is

dV = V − V0 = V0 (e11 + e22 + e33 )

Hence,

Δ = dV / V0

Thus, Δ represents relative change in the volume, and is called the volumetric
strain or dilatation.
If σ 11 = σ 22 = σ 33 = σ , then from (11), we obtain

E
σ= Δ = kΔ
3(1 − 2ν )
where
E
k= (13)
3(1 − 2ν )

103
is called the bulk modulus or incompressibility.

Substituting (11) into (10), and solving for σ 11 , σ 22 , and σ 33 , we obtain

νE E
σ 11 = Δ+ e11
(1 − 2ν )(1 + ν ) (1 + ν )
νE E
σ 22 = Δ+ e22
(1 − 2ν )(1 + ν ) (1 + ν )
νE E
σ 33 = Δ+ e33
(1 − 2ν )(1 + ν ) (1 + ν )

We introduce the Lame’s elastic constants λ and μ by

νE
λ= (14)
(1 − 2ν )(1 + ν )

E
μ= (15)
2(1 + ν )

Then,

σ 11 = λΔ + 2μ e11
σ 22 = λΔ + 2μ e22 (16)
σ 33 = λΔ + 2μ e33

Next, we consider shear stress and shear strain. From Figure 15, we see that the
shear strain e12 is caused by the shear stress σ 12 :

σ 12 = 2Ge12

where G is called the shear modulus.

104
 Similarly,

σ 13 = 2Ge13
σ 23 = 2Ge23

It can be shown (homework problem) that G is equal to the Lame’constant μ . Hence,

σ 12 = 2μ e12
σ 13 = 2 μ e13 (17)
σ 23 = 2μ e23

Equations (16) and (17) give the stress-strain relations in isotropic media. (16) and (17)
can be written collectively as

σ ij = λΔδ ij + 2μ eij (18)

105
Note that, although we have introduced five elastic constants E , ν , k , λ , and μ above,
there are only two independent constants. If we choose E and ν as the basic constants
then

E νE E
k= , λ= , μ= (19)
3(1 − 2ν ) (1 + ν )(1 − 2ν ) 2(1 + ν )

If we choose λ and μ as the basic constants, then

λ 3λ + 2μ 2
ν= , E= μ, k = λ + μ (20)
2(λ + μ ) λ+μ 3

Some Relations in Vector Analysis

In the theory of elasticity, the following definitions and relations are often used.
Here, φ ( x1 , x2 , x3 ) is a scalar function and u (u1 , u2 , u3 ) and v (v1 , v2 , v3 ) are vectors.
We assume that these functions are continuous and differentiable.

1) grad φ is a vector with the components

⎛ ∂φ ∂φ ∂φ ⎞
⎜ , , ⎟ (21)
⎝ ∂x1 ∂x2 ∂x3 ⎠

We often use an operator

⎛ ∂ ∂ ∂ ⎞
∇≡⎜ , , ⎟
⎝ ∂x1 ∂x2 ∂x3 ⎠

106
and write it as ∇φ .

2) div u is a scalar defined by

∂u1 ∂u2 ∂u3

+ + = ul ,l = ∇ ⋅ u (22)
∂x1 ∂x2 ∂x3

3) curl u (rot u ) is a vector defined by

⎛ ∂u3 ∂u2 ∂u1 ∂u3 ∂u2 ∂u1 ⎞

⎜ − , − , − ⎟ = ∇ xu (23)
⎝ ∂x2 ∂x3 ∂x3 ∂x1 ∂x1 ∂x2 ⎠

4) Laplacian ∇ 2φ is a scalar defined by

∂ 2φ ∂ 2φ ∂ 2φ
∇ 2φ = + + = φ,ll (24)
∂x12 ∂x 22 ∂x32

5) Vector Laplacian ∇ 2u is a vector defined by

(∇ 2u1 , ∇ 2u2 , ∇ 2u3 ) = (u1,ll , u2,ll , u3,ll ) (25)

We use this only in the Cartesian coordinate.

6) ∇ 2u = graddivu − curlcurlu = ∇∇ ⋅ u − ∇x∇xu (26)

For a non-Cartesian coordinate system, this should be considered as the definition of

∇ 2u .

7) curlgradφ ≡ 0

8) divcurlu ≡ 0

107
9) If curlu = 0 , u is called an irrotational vector, and can be written as u = gradφ .
φ is the scalar potential.

10) If divu =0, then u is called a solenoidal vector, and can be written as u = curlv .
v is the vector potential.

11) Any vector field u can be decomposed into an irrotational field u I and a
solenoidal field u II , i.e.
u = u I + u II

where curlu I =0 and divu II =0. Using 9) and 10) u can be written as

u = gradφ + curlv (27)

Relations 6), 7), and 8) can be easily verified. Proof of 9), 10) and 11) requires some
knowledge of Potential Theory.

3.1.4 Equation of Motion Expressed in Terms of Displacement

From (18) and the expression for eij , we obtain

σ ij , j = λΔ , jδ ij + μ (ui , jj + u j ,ij )
= λΔ ,i + μ (∇ 2ui + Δ ,i ) = (λ + μ )Δ ,i + μ∇ 2ui

where we assume that the medium is homogeneous, i.e., λ and μ are constants.

Substituting this into (7), we obtain

ρ ui = ρ fi + (λ + μ )Δ ,i + μ∇ 2ui

or in a vector form,

108
 ρ u = ρ f + (λ + μ )graddivu + μ∇ 2u (28)

Using (26), this can be rewritten as,

ρ u = ρ f + (λ + 2μ )graddivu − μ curlcurlu (29)

or

ρ u = ρ f + (λ + 2μ )∇ 2u + (λ + μ )curlcurlu (30)

Equations (28), (29) and (30) are among the most fundamental equations in seismology.

If we decompose u into irrotational field u I (curlu I = 0) and solenoidal field

u II (divu II = 0) , we obtain for u I , using (30)

ρ u I = ρ f + (λ + 2μ )∇ 2u I (31)

For u II , we obtain, using (28)

ρ u II = ρ f + μ∇ 2u II (32)

(31) and (32) are the wave equations for the irrotational field and the solenoidal field,
respectively.

109
Ge 162 Problem #3

1. Show that the Lame's elastic constant μ=E/2(1+ν) (E: Young's modulus, ν: Poisson's
ratio) is actually equal to the shear modulus G defined by

σ ij = 2Geij (i ≠ j)

Follow the steps described below for a 2-dimensional problem.

1) Consider a rectangular parallelepiped ABCD shown in Figure 1, and apply a normal

stress -σ uniformly on BC and AD, and +σ on AB and DC.

110
 Consider equilibrium of a beam with the square cross section abcd in the
parallelpiped. The square abcd is now deformed into a parallelogram a'b'c'd' as shown in
Figure 2.

Show that

ob' (1 + ν )
=1 − σ (1)
ob E

oa' (1 + ν )
=1 + σ (2)
oa E

2) Now consider equilibrium of a triangular beam with the cross section aob. The normal
stress is -σ on oa and σ on ob.

Show that

σt=σ, and σn=0,

where σt and σn are the shear and normal stresses on ab, respectively (see Figure 1).

3) Referring to the square abcd, this deformation can be viewed as shear deformation due
to the shear stress σt .
Show that the corresponding shear strain et is given by

et = (π / 4) − (1/ 2)∠d' a' b' ( ∠d' a' b' is θ in Figure 2)

(3)

4) From (1), (2) and (3),

ob ' E − (1 + ν )σ
tan((π / 4) − et ) = = (4)
oa ' E + (1 + ν )σ

111
Assuming that et is small (i.e. you can put sinet=et , coset=1), obtain the expression for G
by relating et to σt .

Ge 162, Problem #2

Strain at a Point, Principal Strain

Consider a 2-D problem shown in Fig. 1.

A line element PQ becomes P'Q' after deformation. The strain components for
this deformation are given by e11, e22, and e12.

Determine the unit elongation in the direction of PQ and the shear strain for the
directions of PQ and PT (PT is perpendicular to PQ). Follow the steps given below.

Let the displacement of P and Q be u ( x ) and u + du , respectively. Observe that

vector QQ" represents du (du1 , du2 ) , where du1 and du2 are the x1 and x2 components of
du , and are given by QR and RQ" in the figure, respectively.

112
1. Referring to the geometry shown in Fig. 1, show that the difference in the length
between PQ" and PQ (elongation in PQ direction) is given by:

dl = du1 cos α + du2 sin α

Then dividing this by PQ, show that the unit elongation in the direction of PQ is
given by:

eα = e11 cos2 α + e22 sin 2 α + 2e12 sin α cos α

∂u1 ∂u dx
( Use the relations like du1 = dx1 + 1 dx2 and 1 = cos α etc.)
∂x1 ∂x2 PQ

2. Referring to the geometry shown in Fig. 1, show that the angle through which PQ is
rotated is given by:

du2 cos α − du1 sin α ∂ u2 ∂u

= cos 2 α + (e22 − e11 ) cos α sin α − 1 sin 2 α
PQ ∂ x1 ∂ x2

3. The line segment PT makes an angle α+π/2 with the x1 axis. Using the result obtained
above, show that the rotation of PT is given by:

∂ u2 2 ∂u
sin α − (e22 − e11 ) cos α sin α − 1 cos 2 α
∂ x1 ∂ x2

4. Using the results of 2 and 3, show that the shear strain for the directions of PQ and PT
(i.e. 1/2 of the change in angle between PQ and PT) is given by:

es = e12 (cos 2 α − sin 2 α ) + (e22 − e11 )sin α cosα

From this result we see that there are two values of α, differing by π/2, for which
shear strain vanishes. They are given by:

113
 2e12
tan 2α =
e11 − e22

The corresponding strains eα are called the principal strains.

Ge 162 Problem #1

Stress at a point, principal axes and stresses

Consider a 2-D problem shown in Fig. 1.

Consider a (infinitesimally) small prism BOA at P with the three sides parallel to
the x3 axis. Let n and t be the unit vectors normal and parallel to BA respectively (the
directions are shown in the figure.). In class, we showed that the x1 and x2 components of
stress acting on plane BA are given by,

⎛ σ 1n ⎞ ⎛ σ 11 σ 12 ⎞ ⎛ n1 ⎞
⎜ ⎟=⎜ ⎟⎜ ⎟
⎝ σ 2 n ⎠ ⎝ σ 21 σ 22 ⎠ ⎝ n2 ⎠

1. Show that the normal stress and shear stress on the plane BA are given by,

σ nn = σ 11 cos 2 θ + σ 22 sin 2 θ + 2σ 12 cosθ sin θ

(1)

114
 σ nt = σ 12 (cos 2 θ − sin 2 θ ) + (σ 22 − σ 11 ) cosθ sin θ

Show that if the angle θ is chosen such that

2σ 12
tan 2θ =
σ 11 − σ 22
the shear stress vanishes.

The directions of n and t for this θ are called the principal directions.
If we take x1' and x2' axes in n and t directions, they are the principal axes. The normal
stresses on the plane normal to x1' and x2' axes are the principal stresses.

If x1 and x2 axes were the principal axes, (1) becomes

σ nn = σ 11 cos 2 θ + σ 22 sin 2 θ

σ nt = (σ 22 − σ 11 ) cos θ sin θ

2. Refer to Fig. 2.

Show that the X and Y coordinates of point P is given by σ nn and σ nt ,

respectively. The circle shown in Fig. 2 is called Mohr's circle which is used for
graphical representation of the normal stress and shear stress on plane AB. Note that x1
and x2 axes are the principal axes.

115
3.2 Wave Equation and Seismic Waves

Wave Equations

Following (31) and (32) of 3.1.4, and ignoring the body force, we obtain for the
irrotational field u I ( curlu I = 0 ),

u I = α 2∇ 2u I (1)

where
λ + 2μ
α= (2)
ρ

For the solenoidal field u II ( divu II = 0 ),

u II = β 2∇ 2u II (3)

where

116
 μ
β= (4)
ρ

Equations (1) and (3) are three-dimensional wave equations which are of fundamental
importance in seismology.

3.3 Seismic P Wave (Compressional Wave) and S Wave (Shear Wave)

We first examine the property of u I . Here we use a coordinate system (x, y, z) instead of
(xl, x2, x3), and denote the x, y, and z components of displacement by u, v, and w, respectively.

Let us consider a plane elastic wave propagating in the x direction, that is, a wave in
which u I is a function of x and t. Since all derivatives with respect to y and z are zero, we have
from curlu I = 0 ,

∂w ∂v
= 0 , and =0
∂x ∂x

which give v = w =0. (Actually v and w are constant, but constant displacement is not
important in wave propagation problems and they are set equal to 0.) Therefore, only non-zero
displacement component is u . This means, u I represents a wave in which the particle motion is
in the direction of propagation. The propagation velocity is, from (2),

117
 λ + 2μ
α=
ρ

Because of this particle motion, this wave is also called longitudinal wave (Fig.1).

Fig. 1
Similarly, for u , we have from divu = 0 ,
II II

∂u
=0
∂x

which gives u = 0 . Then the non-zero components are v and w . This means, u II represents a
wave in which the particle motion is confined on the plane perpendicular to the propagation
direction; the propagation velocity is, from (4),

μ
β=
ρ

118
 Because of this particle motion, this wave is called a shear wave (Fig. 2). Since
divu II = 0 , it does not involve volume change.

In seismology, the longitudinal wave is often called P wave, and the shear wave
is called S wave.

Fig. 2

3.4 Ray Theory

In the above, we assumed that the medium is homogeneous (in addition to being
isotropic). That is the elastic constants λ and μ do not vary spatially. The equations
(28), (29), and (30) in 3.1.4 were all derived with this assumption.

Once we obtain a 1 -D wave equation in the form,

119
 u ( x, t ) = c 2u ′′( x, t ) (5)

we can obtain the solution,

u ( x, t ) = f ( x − ct ) + g ( x + ct ) (6)

where f(ξ) and g(ξ) are twice differentiable arbitrary functions of ξ. f and g
represent a plane wave propagating in positive and negative x directions,
respectively, with velocity c. For a wave propagating in the positive x direction, we
can consider a wave front x=ct+constant. The line perpendicular to the wave front
determines the path along which the wave front propagates. This line (or curve, in
general) is called a ray. Thus, in case of a homogeneous medium, we can use rays to
describe wave propagation completely (ray theory). Ray theory is more intuitive than
wave theory, and has been used very extensively in seismology.

However, in the real medium, λ and μ are usually a function of x, y, and z. The
equation of motion is consequently far more complex than (28), (29), or (30) in 3.1.4, and
we cannot obtain simple wave equations; consequently we cannot use the ray theory (rays
cannot be defined rigorously). Fortunately, however, if the medium is only weakly
heterogeneous, we can define a “ray” approximately, and use the ray theory. This is a
common practice in seismology. The question is “what is considered weakly heterogeous
?“ This problem can be discussed in detail in the Appendix to this section. Here we only
discuss this condition qualitatively.

Suppose that the wave velocity c(x) is a function of position x. Then the change in
velocity c over a wave length λ is given by

Δc = λ c ′

120
 If this is much smaller than c itself, we can consider that the medium is “weakly
heterogeneous”. The condition then can be stated as

Δc / c = λ c′ / c = 2π c′ / ω << 1

Ignoring 2π , we can say that if

ω >> c′ (7)

then we can use the ray theory to discuss wave propagation problem in a heterogeneous
medium. In this sense, the ray theory represents a high-frequency approximation.

As shown in the Appendix, a harmonic wave in a l-D homogeneous medium is

given by,

u ( x, t ) = exp[iω (t − x / c)] (8)

while in a 1-D weakly heterogeneous medium, it is given by

1 ⎡ ⎛ dx ⎞ ⎤
u ( x, t ) ∝ exp ⎢iω ⎜ t − ∫ ⎟⎥ (9)
ρc ⎣ ⎝ c( x) ⎠ ⎦

Equation (9) shows that the wavefront propagates at the local velocity c(x), and the
amplitude varies as 1/ ρ c .
A similar relation can be obtained for a 3-D medium. The most important result
here is that a “wavefront” can be defined and it propagates at the local velocity, which
leads to the well-known “Snell’s Law” (see the Appendix, for more details).
If a ray is incident from a medium with velocity c1 into a medium with velocity
c2 , then the incidence angle i1 and the angle of emergence i2 are related by (Figure 3),

121
 sin i1 sin i2
= (10)
c1 c2

In general, if the velocity varies vertically as c(z), then, for a given ray on a vertical plane, we
have (Figure 3)

sin i ( z )
= p =const.
c( z )

Fig. 3
Ge 162 Appendix to 3.4

Wave Propagation in a Slightly Heterogeneous Medium

- Solution with the WKBJ Method-

1. One-Dimensional Case

The equation of motion

∂ 2u ∂ ⎛ ∂u ⎞
ρ 2 = ⎜E ⎟
∂t ∂x ⎝ ∂x ⎠

122
 ∂E
does not reduce to the simple wave equation because ≠ 0 in general. We obtain
∂x

∂ 2u ∂ 2u ∂E ∂u
ρ = E +
∂t 2 ∂x 2 ∂x ∂x

Introducing a new variable v by

u = E −1/ 2 v

and ignoring the term containing E ′′ , we obtain

∂ 2v ∂ 2 v 1 E ′2
ρ = E + v
∂t 2 ∂x 2 4 E

We consider a harmonic wave

v( x, t ) = V ( x)eiω t

Then, we have

d 2V ⎡ 2 1 ⎛ E ′ ⎞ ⎤
2

+ ⎢k + ⎜ ⎟ ⎥ V = 0
dx 2 ⎢⎣ 4 ⎝ E ⎠ ⎦⎥

where

ω E
k= and c =
c ρ

If

123
 2
⎛ E′ ⎞
⎜ ⎟ << k , or ω >> c′
2

⎝E⎠

Then

d 2V
2
+ k 2V = 0 (A-1)
dx

Note that if k = k0 = const , then the solution is

V = e ± ik0 x = e ∫
± i k0 dx

If k is a slowly varying function of x, we can solve this equation with the WKBJ method.

We seek a solution in the form

(
V = exp ±i ∫ κ dx )
Substituting this into (1), we find that κ must satisfy the equation

−κ 2 ± iκ ′ + k 2 = 0 (A-2)

1-st Approximation

Since

| κ ′2 |<< κ 2

we obtain for the first approximation of κ by

κ (1)2 = k 2 or κ (1) = k

2nd Approximation

124
 The 2nd approximation can be obtained by using κ (1) for the second term of (A-
2). Then, the 2nd approximation

( )
1/ 2
k′ ⎞
1/ 2
⎛
= ⎜⎛ k 2 ± iκ (1)′ ⎟⎞
1/ 2
κ (2)
= k ± ik ′
2
= k ⎜1 ± i 2 ⎟ (A-3)
⎝ ⎠ ⎝ k ⎠

k′
If, ω >> c′ , then << 1 , and
k2

i k′
κ (2) = k ±
2k

i k′ ⎞ ⎤ 1
∴
⎡ ⎛
V ( x) = exp ⎢ ±i ∫ ⎜ k ±
⎣ ⎝
⎟ dx =
2 k ⎠ ⎥⎦ k
exp ±i ∫ kdx ( )

1 1 ⎡ ⎛ dx ⎞ ⎤
∴ u ( x, t ) = V exp(iω t ) = exp ⎢ ±iω ⎜ t ± ∫ ⎟ ⎥ (A-4)
E ρω c ⎣ ⎝ c ⎠⎦

This is the WKBJ solution. This solution is valid under the condition ω >> c′ .

2. Vertically Heterogeneous Medium

We consider a SH wave, i.e., the displacement vector is given by (0, v( x, z ), 0) .

125
Then the equation of motion is given by

∂ 2v ⎛ ∂ 2v ∂ 2v ⎞ ∂v
ρ = μ ⎜ 2 + 2 ⎟ + μ′
∂t 2
⎝ ∂x ∂z ⎠ ∂z

As is in the 1-D case, we put

v( x, t ) = μ −1/ 2 eikxV ( z )eiω t

Then, we obtain

d 2V ⎡ 2 1 ⎛ μ′ ⎞ ⎤
2

+ ⎢kβ ( z ) − k + ⎜ ⎟ ⎥ V = 0
2

dz 2 ⎢⎣ 4 ⎝ μ ⎠ ⎥⎦

where

ω μ
kβ = and β =
β ρ

We introduce χ ( z ) by

126
 χ 2 ( z ) = k β2 ( z ) − k 2

If,
2
1 ⎛ μ′ ⎞
χ ( z ) >> ⎜ ⎟
2
(A-5)
4⎝ μ ⎠

then,

d 2V
2
+ χ 2 ( z )V = 0 (A-6)
dz

For (A-5) to be satisfied, at least ω >> β ′ must be satisfied. If k β2 ( z ) − k 2 =0, then this
condition is never satisfied.

Then the solution of (A-6) is

V ( z) =
1
χ
(
exp ±i ∫ χ ( z )dz )
Then,

v ( x, z , t ) =
1 1
μ (k β − k 2 )1/ 4
2 ⎢⎣ ( )
exp ⎡i ω t + kx ± ∫ k β2 − k 2 dz ⎤
⎥⎦

The wave front (i.e., plane of constant phase) is defined by

kx ± ∫ k β2 − k 2 dz = const

∴
dz k
=±
dx kβ − k 2
2

127
at z = z0

k
= const
kβ − k 2
2

dz 1 k β
sin i = = = = k
d x+d z
2 2
kβ − k
2 2 kβ ω
1+
k2

∴
sin i ( z )
= const for a given ray.
β ( z)

However, if k β = k or i = π / 2 , which occurs at the turning point of a ray, the

condition (A-5) cannot be satisfied and the WKBJ method breaks down.

Ge 162 Problem #4

Reflection and Transmission of SH wave at a Plane Boundary

128
 An incident SH wave S1, reflected SH wave Sr and refracted (transmitted) SH
wave S2 can be written as follows. (The displacement has only z component, w.)

⎡ ⎛ sin j1 cos j1 ⎞⎤
w1 = exp ⎢iω ⎜ x− y − t ⎟⎥ (1)
⎣ ⎝ β1 β1 ⎠⎦

⎡ ⎛ sin j1 cos j1 ⎞⎤
wr = R exp ⎢iω ⎜ x+ y − t ⎟⎥ (2)
⎣ ⎝ β1 β1 ⎠⎦

⎡ ⎛ sin j2 cos j2 ⎞⎤
w2 = T exp ⎢iω ⎜ x− y − t ⎟⎥ (3)
⎣ ⎝ β2 β2 ⎠⎦

where

sin j1 sin j2 1
= =
β1 β2 c

129
Here, the amplitude of the incident wave is assumed to be 1. R and T are reflection and
transmission coefficients respectively. c is the phase velocity along the boundary. (The
above relations are solutions of the wave equation in each layer.)

Using c, the above equations can be written as,

⎡ ⎛ x cos j1 ⎞⎤
w1 = exp ⎢iω ⎜ − y − t ⎟⎥ (4)
⎣ ⎝c β1 ⎠⎦

⎡ ⎛ x cos j1 ⎞⎤
wr = R exp ⎢iω ⎜ + y − t ⎟⎥ (5)
⎣ ⎝c β1 ⎠⎦

⎡ ⎛ x cos j2 ⎞⎤
w2 = T exp ⎢iω ⎜ − y − t ⎟⎥ (6)
⎣ ⎝c β2 ⎠⎦

Two boundary conditions must be satisfied at the boundary: one for displacement and the
other for stress (traction).

Show that these conditions are given by

1+ R = T

μ1 μ μ
− cos j1 + 1 cos j1 R = − 2 cos j2T
β1 β1 β2

where μ1 and μ 2 are the rigidity.

Solving these equations, show that

130
 − μ 2 β1 cos j2 + μ1 β 2 cos j1
R= (7)
μ1 β 2 cos j1 + μ 2 β1 cos j2

2 μ1 β 2 cos j1
T= (8)
μ1 β 2 cos j1 + μ 2 β1 cos j2

Determine the reflection and transmission coefficients for normal incidence (e.g., j1 =0).

− μ 2 β1 + μ1 β 2
R= (9)
μ1 β 2 + μ 2 β1

2μ1 β 2
T= (10)
μ1 β 2 + μ 2 β1

If β 2 > β1 , sin j2 exceeds 1 for j1 larger than the critical angle. In this case,

cos j2 = (1 − sin 2 j2 )1/ 2

becomes purely imaginary, i.e., cos j2 = ib (b is real).

Referring to (6), briefly describe the behavior of the "transmitted" wave S2 for this case.

Also, the reflection coefficient R becomes complex. Determine the amplitude and
phase (with respect to the incident wave) of the reflected wave.

3.5 Reflection and Refraction (LW, pp. 96-104)

When the elastic constant, say velocity, changes discontinuously in the

medium, the condition (7) cannot be satisfied for any wavelength, because | c′ |
becomes infinity there. Thus, in this case the ray theory cannot be used, and the wave
equation must be solved with the appropriate boundary conditions (continuity of

131
displacements and traction) on the discontinuity surfaces. The problem, in general,
becomes very complicated, and complex reflection, refraction and energy coupling
between P and S waves take place. The simplest case in which a homogeneous
medium is bounded at a plane boundary by another homogeneous medium having
different elastic property (Fig. 4) has been studied by many investigators.

First let us consider a plane S wave incident at the boundary from medium 1
to 2. Let β1 and β 2 be S wave velocities in media 1 and 2, respectively. We assume

β1 < β 2 . It is usually convenient to decompose an S wave into SH and SV

components. SH wave is an S wave whose particle motion is parallel to the boundary,
and the particle motion of SV wave is on the plane including the ray and
perpendicular to the boundary (this plane is called the plane of incidence) (Fig. 4).

132
 Since SH wave does not have a component of particle motion perpendicular
to the boundary, the situation is relatively simple. As shown in Fig. 5a, when SH
wave is incident at the boundary with incident angle, j1 , reflected SH wave, Sr, and

refracted SH wave, S2, result. The angle of emergence, j2 , is related to j1 by

sin j1 sin j2
=
β1 β2

133
This relation is similar to Snell’s law (10) of 3.4. Thus, we see that, even in this case,

134
Snell’s law can be used to determine the emergence angle. The amplitude ratios
S r / S1 and S 2 / S1 are functions of j1 and, of course, β 2 / β1 . When j1 exceeds

⎛β ⎞
j1,c = sin −1 ⎜ 1 ⎟
⎝ β2 ⎠

(2)

no refracted wave results (Fig. 5b). In this case, part of the wave energy is trapped
along the boundary, and total reflection occurs. At the same time phase shift occurs
on reflection. The angle j1,c is called the critical angle.

When P wave is incident on the boundary, the situation is more complicated. We

let α1 and α 2 be the P wave velocities in media 1 and 2, respectively. We assume α 2 >
α1 > β 2 > β1 . In this case, in addition to refracted and reflected P waves, refracted and
reflected SV waves result (Fig. 5c). The following relation holds,

sin i1 sin i2 sin j1 sin j2

= = = (3)
α1 α2 β1 β2

where i1 and i2 are angles of incidence and emergence of the P wave, and j1 and
j2 are angles of reflection and emergence of the SV wave. When i1
exceeds

135
 ⎛α ⎞
i1,c = sin −1 ⎜ 1 ⎟
⎝ α2 ⎠

no refracted P wave results, and part of the wave energy is trapped along the boundary. The
angle i1,c is the critical angle.

When SV wave is incident at the boundary, the situation is even more complicated. In
this case, refracted and reflected P and SV waves result (Fig. 5d). As before, the following
relations hold
sin j1 sin j2 sin i2 sin i1
= = = (5)
β1 β2 α2 α1

where j1 and j2 are angles of incidence and emergence of SV waves, and i1

and i2 are angles of reflection and refraction of P waves. When j1 exceeds

⎛β ⎞
j1,c1 = sin −1 ⎜ 1 ⎟ (6)
⎝ α2 ⎠

P wave no longer propagates into

medium 2. When j1 exceeds

⎛β ⎞
j1,c2 = sin −1 ⎜ 1 ⎟ (7)
⎝ α1 ⎠

no P wave is reflected into medium 1.

When j1 exceeds

136
 ⎛β ⎞
j1,c3 = sin −1 ⎜ 1 ⎟ (8)
⎝ β2 ⎠

then no SV wave refraction

takes place. In these cases,
part of the energy is trapped
along the boundary, and
complicated phase shift
occurs. The angles j1,c1 , j1,c2

and j1,c3 are the critical

angles.
When one medium is vacuum, the
boundary becomes a free surface, and only
reflections are to be considered. It is not
difficult to see from Fig. 5 that there is no
critical angle for incident SH and P waves,
but for incident SV wave

⎛β ⎞
j1,c2 = sin −1 ⎜ 1 ⎟ (9)
⎝ α1 ⎠

becomes critical angle at which no P wave reflection occurs.

Ge 162 Problem #5

137
Dispersion of Love Waves

We discussed the simplest case of plane Love wave in a single homogeneous

layer overlying a homogeneous half-space. The phase velocity c of Love waves can be
determined by the characteristic equation

⎛ω ⎞ μ S
tan ⎜ S1 H ⎟ = 2 2 (1)
⎝c ⎠ μ1S1

c2 c2
where S1 = − 1 , and S 2 = 1 − , and β1 < c < β 2 .
β12 β 22

Equation (1) can be solved graphically to determine the phase velocity c for given
μ1 , μ 2 , β1, β2, and H. The roots of (1) are given by the intersection of the two curves
corresponding to RHS (right-hand side) and LHS of (1) both of which are functions of
phase velocity.

For the case

H=35 km
μ1 =3x1011 dyne/cm2, β1 =3.5 km/sec
μ 2 =7x1011 dyne/cm2, β 2 =4.6 km/sec

a) Compute the values of RHS and LHS of (1) for phase velocities
c=3.52, 3.6, 3.8, 4.0, 4.2, and 4.5 km/sec, and for periods T=20, 35, 50, and 80 sec.

b) Plot the values computed above (vertical axis) as a function of phase velocity c
(horizontal axis). Draw the two curves and find the phase velocity from the intersection
of the two curves (this gives the phase velocity of the fundamental mode).

c) Compare these phase velocities c(T) with the dispersion curve for Love waves for the
continental region (see Fig. 14 in class note 3.6).

Ge 162 Practice Session 3 Surface Wave Phase and Group Velocities

138
 Figure 1 shows the Rayleigh wave trains from an earthquake (1/2/2002, M=7.3) in the
Vanuatu Is., southwest Pacific, recorded at two TriNet stations, PAS (Pasadena) and NEE
(Needles) (The SAC files, van2.pas.lhz.sac and van2.nee.lhz.sac are in
/home/ftp/pub/hiroo/ge162.dir/practice3.dir, but they are not necessary to do this problem).
These two stations and the epicenter are almost on a great circle, and the distance between the
two stations, Δ, is 331 km. The unit of the time scale on the horizontal axis is sec (i.e., 2100 to
3100 sec).

Fig. 1

139
 Figures 2 and 3 show the original (top trace) and band-pass filtered records (at periods of
about 20, 30, 40, 50, and 60 sec) of the seismograms shown in Figure 1.

Fig. 2 (PAS)

140
 Fig. 3 (NEE)

The band-pass filtered records at T=60 sec are of marginal quality. Thus, we use the data
from 20 to 50 sec.

1. Determine the group velocity, U, between PAS and NEE, at periods of 20, 30, 40, and 50 sec,
by measuring the arrival times of the wave train. (Equation 57 in class handout), and plot the
results on a U-T diagram, and compare it with that for a simple crustal structure shown at the
end.

The determination of the phase velocities, c, is a bit more difficult. Figures 4 and 5 show
the plots of the harmonic components of the records at PAS and NEE. The top trace is the
original and the six traces below it are the harmonic components at periods of 20.08, 30.12,
40.96, 51.20, 60.24, and 73.14 sec.

141
 In principle, the phase velocities can be determined from the phase arrival times of the
harmonic components using equation 51 in class handout. In this example, the records at PAS
and NEE have the same starting time. The difficulty is that, if only harmonic components at
discrete periods are given, it is not possible to determine the integer N. In other words, all the
peaks look exactly the same. In this exercise, we assume that the phase velocities in southern
California are approximately known within a certain range, and we determine N for each period
so that the measured phase velocities fall in this range. (Note: If all the harmonics are given, as
is the case in the real situation, we need to determine N only for one period.)

Fig. 4

142
 Fig. 5
We assume that the phase velocities in southern California are within the
following ranges.

T (sec) c (km/sec)
20 3.1-3.9
30 3.0-4.3
40 3.3-4.5
50 3.3-4.5

2. Determine the phase velocities between PAS and NEE using the pairs of harmonic
waves shown below at periods of 20.08, 30.12, 40.97, and 51.20 sec, and plot them on a
c-T diagram, and compare it with the phase velocity dispersion curve for a simple
structure given at the end.

143
144
145
146
147
148
 These dispersion curves are computed for a simple crustal structure:

Thickness P-velocity S-velocity Density

(km) (km/sec) (km/sec) (gm/cm3)

4.0 5.5 3.18 2.6

23.4 6.3 3.64 2.67
5.0 6.8 3.93 2.8
100. 8.0 4.62 3.2

Note:

149
 In this exercise, we use only 2 stations and the path is chosen to be a great circle.
In practice, we use multiple stations and the wave is assumed to be a plane wave and the
propagation azimuth and the phase velocities are simultaneously determined. Using a
high-density network like TriNet, the accuracy can be improved significantly, and the
method can be used for tomographic inversion.
3.6 Seismic Surface Waves (LW, pp. 116-153)

Rayleigh Wave

As we discussed in 3.5, when an SV wave is incident at a free surface, P and SV

waves are, in general, reflected. However, when the incidence angle exceeds the
critical angle, no P wave reflection occurs and part of the wave energy is trapped
along the free surface. This situation suggests existence of surface waves whose
energy is confined near the free surface. We consider here one of the simplest cases.
Consider a homogeneous elastic medium whose P wave velocity, S wave
velocity, and density are α , β , and ρ , respectively. This medium is bounded by a
plane surface, which we take as the x-y plane. We let the medium be in z <0 (Fig. 6).
Let us consider a harmonic surface wave propagating in x direction. Since we are
considering a surface wave as a result of the coupling between P (irrotational) and SV
(solenoidal) waves, we will look for a solution u in a form,

u = u I + u II (1)

where u I is the irrotational field satisfying

curlu I = 0 and u I = α 2∇ 2u I (2)

150
and u II is the solenoidal field satisfying

divu II = 0 and u II = β 2∇ 2u II (3)

Since we consider only P and SV type motions propagating in x direction, u I and u II have only
x and z components (u I , wI ) and (u II , wII ) which do not depend on y. We write u I , wI , u II ,

and wII in a form,

151
 ⎛ x⎞
iω ⎜ t − ⎟
u I = f1 ( z )e ⎝ c⎠

⎛ x⎞
iω ⎜ t − ⎟
w = h1 ( z )e
I ⎝ c⎠
(4)
⎛ x⎞
iω ⎜ t − ⎟
u II = f 2 ( z )e ⎝ c⎠

⎛ x⎞
iω ⎜ t − ⎟
w = h2 ( z )e
II ⎝ c⎠

At this point, the phase velocity c is unknown, and we investigate whether we can find c which

satisfies all the equations and the boundary conditions for surface waves.

From curlu I = 0 , and divu II = 0 , we have

∂u I ∂wI ∂u II ∂wII
− = 0, + =0
∂z ∂x ∂x ∂z

from which

1c ′ 1c ′
h1 ( z ) = − f1 ( z ) and f2 ( z) = h2 ( z ) (5)
iω iω

immediately follow.

Substituting u I into (2), and wII into (3), we have

2
⎛ω ⎞ ⎛ c2 ⎞
f1′′ − ⎜ ⎟ ⎜1 − 2 ⎟ f1 = 0 (6)
⎝c⎠ ⎝ α ⎠

152
 and

2
⎛ω ⎞ ⎛ c2 ⎞
h2′′ − ⎜ ⎟ ⎜ 1 − h =0
2 ⎟ 2
(7)
⎝c⎠ ⎝ β ⎠

⎛ c2 ⎞
respectively. If ⎜1 − 2 ⎟ < 0 , then h2 becomes a periodic function which does not
⎝ β ⎠
decay in the medium. We therefore require

⎛ c2 ⎞
⎜ 1 − 2 ⎟
>0 (8)
⎝ β ⎠

Then, we have

ω ω
± Sα z ± Sβ z
f1 = Ae c
, h2 = Be c
(9)

c2 c2
Sα = 1 − , Sβ = 1 −
α2 β2

where A and B are constants,

The solution with the minus sign gives an infinite amplitude in the medium at z = −∞
which is physically implausible; the plus sign must be taken. Thus, we have

153
 ⎛ ⎞
⎛ ωc Sα z S β ω
S β z ⎞ iω ⎜ t − ⎟
x

u ( x, z , t ) = ⎜ Ae + Be c
⎟e
⎝ c⎠
(10)
⎝ i ⎠
ω ω ⎛ ⎞
⎛ Sα S β z ⎞ iω ⎜ t − ⎟
x
Sα z
w( x, z , t ) = ⎜ − Ae c
+ Be c
⎟e
⎝ c⎠
(11)
⎝ i ⎠
These solutions must satisfy the stress-free boundary conditions at the free surface z = 0, i.e.,

⎛ ∂u ∂w ⎞ ∂w
0 = σ zz = λ ⎜ + ⎟ + 2μ , at z=0 (12)
⎝ ∂x ∂z ⎠ ∂z

⎛ ∂u ∂w ⎞
0 = σ zx = μ ⎜ + ⎟, at z=0 (13)
⎝ ∂z ∂x ⎠

Substituting (10) and (11) into (12) and (13), and setting z = 0, we have

⎛ c2 ⎞ c2
−i ⎜ 2 − 2 ⎟ A + 2 1 − 2 B = 0 (14)
⎝β ⎠ β

c2 ⎛ c2 ⎞
2 1− A − i ⎜ 2 − ⎟B = 0 (15)
α2 ⎝ β2 ⎠

For non-trivial A and B to exist, we must have

2 1/ 2 1/ 2
⎛ c2 ⎞ ⎛ c2 ⎞ ⎛ c2 ⎞
⎜ 2 − ⎟ = 4 ⎜1 − 2 ⎟ ⎜ 1 − 2 ⎟
(16)
⎝ β2 ⎠ ⎝ α ⎠ ⎝ β ⎠

or

154
 c6 c4 2 ⎛ 24 16 ⎞ ⎛ β2 ⎞
−8 + ⎜ 2 − − 16 ⎜1 − 2 ⎟ = 0 (17)
α 2 ⎟⎠
c
β6 β4 ⎝β ⎝ α ⎠

⎛ β2 ⎞
Since LHS of this equation becomes − 16 ⎜1 − 2 ⎟ < 0 and 1 at c = 0 and c = β ,
⎝ α ⎠
respectively, (17) has a real root at
c
0< <1 (18)
β

This satisfies the condition (8) assumed previously. Thus, we have proved that the surface
wave given by (10) and (11) exists, and that it propagates with a velocity smaller than the shear
wave velocity β . This kind of wave is called Rayleigh wave, and the propagation velocity c is
called the phase velocity. Combining (14), (10), and (11), we have

⎛ ωc Sα z ⎛ 1 c 2 ⎞ ωc Sβ z ⎞ iω ⎛⎜⎝ t − cx ⎞⎟⎠
u ( x, z , t ) = A ⎜ e − ⎜1 − 2 ⎟
e ⎟e (19)
⎝ ⎝ 2β ⎠ ⎠
⎛ ω
Sα z 1 ⎛ 1 c 2 ⎞ ωc Sβ z ⎞ iω ⎜⎝ t − c ⎟⎠
⎛ x⎞

w( x, z , t ) = −iA ⎜ − Sα e c
+ ⎜1 − ⎟e ⎟⎟ e (20)
⎜ Sβ ⎝ 2 β 2 ⎠
⎝ ⎠

The amplitude ratio at z = 0 becomes

1 c2
⎛u⎞ 2 β2
⎜ ⎟ = (21)
⎝ w ⎠ z =0 1 ⎛ 1 c2 ⎞
− Sα + ⎜1 − ⎟
Sβ ⎝ 2 β 2 ⎠

and the phase of w is advanced by π / 2 with respect to u,

155
 When α = 3β (this corresponds to the Poisson’s solid for which Poisson’s ratio ν =

c2 2 2
0.25), (17) has three real roots of , namely, 4, 2 + , and 2 − . The first two do not
β 2
3 3
satisfy the condition (8), and therefore do not yield a surface wave. From the last value, we have

c = 0.92 β (22)

which satisfies (8). In this case, from (9), Sα = 0.85 , and S β = 0.39 . The particle motion at the
surface for simple harmonic waves can be obtained by taking the real part of (19) and (20):

u ( x, t ) = 0.42 A cos ω ( t − x / c ) (23)

w( x, t ) = 0.62 A sin ω ( t − x / c ) (24)

and the amplitude ratio becomes 0.68. Equations (23) and (24) show that the orbital motion is
elliptic and counter-clockwise (for a wave propagating in positive x direction) (Fig.6).
The decay of the amplitude with depth is governed by factors like

ω z ω z
Sα z 2π Sα Sβ z 2π S β
ec =e λ
and ec =e λ

where λ is the wave length. Thus short wave length components are more quickly attenuated
than long wave length components.
In the above, the medium is assumed homogeneous. Waves like Rayleigh waves also
exist in heterogeneous (usually only in z direction) medium. In this case, the phase velocity c is
in general a function of wave period (or wave length), and therefore the propagation becomes
dispersive. Curves which relate the phase velocity to the period are called phase velocity

156
dispersion curves. Calculation of such dispersion curves for a vertically heterogeneous medium
is usually made numerically,

Love Wave

As we saw in Fig. 5, SH wave is totally reflected when the incidence angle exceeds the

157
 ⎛β ⎞
critical angle j1,c = sin −1 ⎜ 1 ⎟ ( β 2 > β1 ) . Also, it is totally reflected at a free surface, regardless
⎝ β2 ⎠
of the incident angle. Therefore if we have a layer underlain by a half space (Fig. 7), and assume
that β 2 > β1 , we can consider a SH wave bouncing between the free surface and the boundary
without major loss of energy into the half space. This situation suggests propagation of wave
energy which is trapped within the layer.

Let us consider propagation of this kind of wave. We take the free surface and the
boundary to coincide with the plane z=0 and z=-H respectively. The half space occupies z < − H .
As before we consider a wave propagating in positive x direction. Since we consider an SH field,
the displacement vector has only y component. We let v1 and v2 be the displacements in the

layer and the half space, and look for the solutions in a form

158
 ⎛ x⎞
iω ⎜ t − ⎟
v1 = f1 ( z )e ⎝ c⎠

(25)

⎛ x⎞
iω ⎜ t − ⎟
v2 = f 2 ( z )e ⎝ c⎠

(26)

v1 and v2 satisfy the SH wave equation (3) in the layer and the half space respectively. Thus,

f1 and f2 must satisfy

ω 2 ⎛ c2 ⎞
f1′′ + 2 ⎜
− 1⎟ f1 = 0 (27)
c ⎝ β1 2
⎠
′′ ω 2 ⎛ c2 ⎞
f 2 − 2 ⎜1 − 2 ⎟ f 2 = 0 (28)
c ⎝ β2 ⎠

The solutions of these equations are

ω ω
f1 = A cos S1 z + B sin S1 z (29)
c c
ω
S2 z
f 2 = Ce c (30)

where

159
 c2 c2
S1 = − 1 and S 2 = 1 − (31)
β12 β 22

c2
For f2 to vanish at z → −∞ , 1 − > 0 , and we take only the term with positive sign in the
β 22
exponent of f2. At this point S1 can be either real or imaginary. f1 and f2 must satisfy the
boundary conditions at the free surface and at the boundary. At the free surface

∂v1
0 = σ zy = μ1 at z=0 (32)
∂z
At the boundary

∂v1 ∂v
v1 = v2 and μ1 = μ 2 2 at z=-H (33)
∂z ∂z

where μ1 and μ 2 are the rigidity of the layer and the half space, respectively. Substituting (29)
and (30) into (32) and (33) leads to

B=0 (34)
ω
ω − S2 H
A cos S1 H − Ce c
=0 (35)
c
ω
ω − S2 H
Aμ1S1 sin S1 H − C μ 2 S 2 e c
=0 (36)
c

For non-trivial A and C to exist, we must have

160
 ω μ 2 S2
tan S1 H = (37)
c μ1 S1

This is the characteristic equation. If this equation is satisfied, the surface wave exists, and
the phase velocity c is determined from (37). Since S2 is required to be real, this equation implies
that S1 is also real. Therefore, from (31),

β1 < c < β 2

If ω → ∞ , or λ ( = 2π c / ω ) → 0 then (37) gives c = β1 , and ω → 0 , or λ → ∞ , then (37) gives

c = β 2 . Since RHS of (37) varies continuously from ∞ to 0, as c changes from β1 to β 2 , and

ω
( β 2 / β1 )
2
LHS changes from 0 to tan − 1 H as c changes from β1 to β 2 , there is at least
β2
one solution of (37) between β1 and β 2 for any value of ω . For large values of ω

(ω > β π /
2 ( β 2 / β1 )
2
)
− 1 H , more than one solution exists (Fig. 8).

Thus, the existence of the surface wave is proved. This kind of

wave is called Love wave. As is evident from the previous discussion, Love wave is of purely
SH type, and is dispersive
(i.e., c depends on ω ).

As we can see from (35), (29) and (30), the amplitudes in the layer and the half space
become

161
 ω
f1 = A cos S1 z (38)
c
ω
ω S2 ( z + H )
f 2 = A cos S1 He c (39)
c

It can be shown that for a fixed ω , the solution (mode) with the lowest value of c (c = c1 in Fig.

8), gives an amplitude function which does not have a zero crossing (node). This mode is called

fundamental mode. When more than one solution exists, the modes with higher phase velocities

than c1 have an amplitude function with zero crossings (nodes) (Fig. 8). These modes are called

overtones or higher modes.

We can consider this type of wave for more complex medium, for example, a medium

which has more than one layer or continuous velocity variation with depth. The computation of

the dispersion curves for such complex media must be made numerically,

162
163
Dispersion Curve and Structure

Dispersion curves c(ω ) or c(T ) are important when we use surface waves to determine

the structure of the medium. As we discussed earlier, both Love and Rayleigh waves are

dispersive in layered media, and the dispersion curves are determined by the structure.

Figures 9, 10, and 11 show several examples of dispersed Love and Rayleigh waves.

164
 Fig. 9

165
 Fig. 10a

166
 Fig. 10b

167
 Fig. 11

Consider a layer-over-half space (Figure 12). Let the thickness of the layer be H, and the

shear velocity in the layer and the half space be β1 and β 2 , respectively ( β 2 > β1 ). At short

period, the Love wave energy will be in the layer and the phase velocity approaches β1 . At long

period, Love wave energy penetrates into the half space. At very long period, the velocity

approaches β 2 . Thus, the phase velocity dispersion curve would look like the one shown in

Figure 12.

168
 Fig. 12

Then, how would the shape of the dispersion curve change as H varies? If H decreases,

the Love wave would "feel" the higher speed half space at a shorter period. Thus, the shape of

the dispersion curves would change as shown. For a more complex multi-layered structure, a

similar pattern is expected. Thus, we can determine the structure from the shape of the

dispersion curve.

169
 We can make the same argument for Rayleigh waves except that the phase velocities at

the long-period and the short-period ends are 0.92 β 2 and 0.92 β1 (for Poisson solid),

respectively.

Figures 13 and 14 show the examples of phase velocity dispersion curves for oceanic,

continental and shield structures.

Fig. 13

170
 Fig. 14

The structures and the data are given in the following tables.

Models

Ocean

Thickness P velocity S-velocity Density

(km) (km/sec) (km/sec) (g/cm3 )

5.0 1.52 0.0 1.03

1.0 2.10 1.00 2.10
5.0 6.41 3.70 2.84
49. 7.82 4.61 3.34

171
 160. 8.17 4.30 3.44
100. 8.49 4.60 3.53

Continent

Thickness P velocity S-velocity Density

(km) (km/sec) (km/sec) (g/cm3 )

15.0 5.57 3.36 2.65

18.0 6.50 3.74 2.87
17.0 7.775 4.36 3.33
25.0 7.83 4.39 3.35
35.0 7.92 4.44 3.37
40.0 8.04 4.49 3.41
50.0 8.19 4.56 3.45
50.0 8.35 4.64 3.49
50.0 8.50 4.72 3.53

Shield

Thickness P velocity S-velocity Density

(km) (km/sec) (km/sec) (g/cm3 )

6.0 5.64 3.47 2.70

10.5 6.15 3.64 2.80
18.7 6.60 3.85 2.85
80.0 8.10 4.72 3.30
100.0 8.20 4.54 3.44
100.0 8.30 4.51 3.53

Data

Rayleigh Wave

Period O C S

172
(sec) (km/sec) (km/sec) (km/sec)

55 4.02 3.98 4.15

50 4.02 3.96 4.14
45 4.02 3.94 4.11
40 4.02 3.90 4.09
35 4.02 3.83 4.02
30 3.99 3.73 3.94
25 3.98 3.63 3.82

Love Wave

Period O C S
(sec) (km/sec) (km/sec) (km/sec)

70 4.56 4.45 4.62

60 4.55 4.42 4.59
50 4.52 4.34 4.52
40 4.49 4.20 4.42
30 4.45 4.03 4.29
20 4.40 3.88 4.05

173
Phase and Group Velocity of Dispersive Waves

Surface waves are usually dispersive. That is, the phase velocity c is a function of
angular frequency ω . In the treatment of dispersive waves we need to distinguish phase
velocity c(ω ) and group velocity U (ω ) .
In order to understand the propagation of a dispersive wave train, we consider a
wave packet made up of many harmonic wave trains like

g ( x, t ) = A0 exp [i (kx − ω t + const) ] (40)

The phase velocity is given by

c =ω /k (41)

174
175
 Consider a case in which all components are in phase at x=0, t=0. Then, we have
constructive interference at t=0, and destructive interference elsewhere leading to
negligible disturbance. Thus, we have an impulse at x=0. At some later time t and at a
distance x, a disturbance will be observable if any of the wave trains are in phase over a
frequency band. The condition for reinforcement is

kx − ω t = const.

within the frequency band. Differentiating with respect to k, we obtain

⎛ dω ⎞
x −⎜ ⎟t = 0 (42)
⎝ dk ⎠

176
The ratio x/t gives the velocity with which the disturbance (wave group) at the frequency
band propagates, and is called the group velocity U. From (42),

dω dc
U= =c+k (43)
dk dk

In general U is also a function of ω .

Determination of Phase Velocity c(ω )

Let a surface wave train at x = x1 be

+∞ +∞
g ( x1 , t ) =
∫
−∞
gˆ ( x1 , f ) exp(2π ift )df =
∫ −∞
| gˆ ( x1 , f ) | exp [i (2π ft + φ1 ) ] df (44)

where

+∞

gˆ ( x, f ) =| gˆ ( x, f ) | exp [iφ ( f ) ] =
∫ −∞
g ( x, t ) exp(−2π ift )dt (45)

| gˆ ( x, f ) | and φ ( f ) are the amplitude and phase spectrum of g ( x, t ) .

If the wave train propagates in x direction without changing the amplitude, then at
x = x2 ,

+∞
g ( x2 , t ) =
∫ −∞
| gˆ ( x1 , f ) | exp {i [ 2π f (t − ( x2 − x1 ) / c( f )) + φ1 ]} f
+∞
=
∫ −∞
| gˆ ( x1 , f ) | exp ⎡⎣i ( 2π ft + φ2 ) ⎤⎦ df

177
 (46)
and

φ2 = φ1 − 2π f ( x2 − x1 ) / c( f ) (47)

If the wave is non-dispersive (i.e., c = c0 = constant), then

g ( x2 , t ) = g ( x1 , t − x / c0 ) (48)

That is, the wave propagates without changing its waveform. In a dispersive medium, c is
a function of frequency, and the waveform changes with propagation.
Since, adding 2π N (N is an integer) to φ2 does not change (46), (47) is actually

φ2 = φ1 − 2π f ( x2 − x1 ) / c( f ) + 2π N (49)

Using (49), we can determine the phase velocity from the two wave trains measured at
distances x1 and x2 by measuring the phases φ1 and φ2 at these distances. Solving (49)
for c(f), we obtain,

f ( x2 − x1 )
c( f ) = (50)
N − (φ2 − φ1 ) / 2π
or,

( x2 − x1 )
c(T ) = (51)
NT + (−φ2T / 2π + φ1T / 2π )

where (−φ2T / 2π ) − (−φ1T / 2π ) is the phase travel time.

178
 In the above, the records at the two stations are assumed to have a common origin
time. In actual computations, we may use the records (seismograms) starting at different
times; at t = t1 at x = x1 and at t = t2 at x = x2 . Then, if we let ψ 1 , and ψ 2 be the Fourier

phases measured from these records, then we have

φ2 − φ1 = ψ 2 − ψ 1 − 2π f (t2 − t1 ) (52)

Substituting this in (50), we obtain

f ( x2 − x1 )
c( f ) = (53)
N − (ψ 2 −ψ 1 ) / 2π + f (t2 − t1 )

If the propagation path is very long, and involves antipolar or polar passages, a small
correction is necessary to correct for the polar phase shift. In this case,

f ( x2 − x1 )
c( f ) = (54)
N + m / 4 − (ψ 2 − ψ 1 ) / 2π + f (t2 − t1 )

where m is the number of polar or antipolar passages.

Determination of Group Velocity U (ω )

Once the phase velocity is determined as a function of ω (or, frequency, f), then
the group velocity U (ω ) can be computed from phase velocity using (43), i.e.,

179
 1 dk 1 ω dc(ω )
= = − 2 (55)
U (ω ) dω c(ω ) c (ω ) dω

Another way of determining the group velocity is to use narrow band-pass

filtering of the seismograms.
Let a propagating wave be given by,

+∞ +∞

∫ ∫
1 1
g ( x, t ) = gˆ ( x, ω ) exp[i(ω t − kx)]dω = | gˆ ( x, ω ) |exp[i (ω t − kx + φ )]dω
2π −∞ 2π −∞

(56)

If we consider a contribution over a narrow frequency band Δω around ω 0 , we obtain

from (56)

ω0 +Δω

∫
1
g ( x, t ) ≈ | gˆ ( x, ω 0 ) | cos(ω t − kx + φ )dω
π ω0 −Δω

⎛ ⎛ Δω ⎞⎞
⎜ sin ⎜ (t − k0′ x + φ0′ ⎟ ⎟
Δω ⎝ 2 ⎠ ⎟ cos(ω t − k x + φ )
= | gˆ ( x, ω 0 ) | ⎜ (57)
π ⎜ Δ ω ⎟
0 0 0

⎜ (t − k ′
0 x + φ ′
0 ) ⎟
⎝ 2 ⎠

where k0 and φ0 are the values of k and φ at ω = ω 0 , respectively, and k0′ and φ0′ are

dk dφ
the values of and at ω = ω 0 , respectively. Equation (57) gives a wave train
dω dω
which propagates at a speed

180
 1 dω
U (ω 0 ) = = (58)
k0′ dk ω =ω0

Thus, if we measure the propagation velocity of a band-pass filtered record, we can

determine the group velocity.

Ge 162

3.7 Normal Mode Theory (LW, pp. 154-172)

Since the Earth is a bounded medium, the wave propagation problem can be
treated as a normal-mode problem.

181
 The basic physics can be understood with a simple 1-D problem.

1. 1-D Problem

Consider free oscillations of a homogeneous elastic rod shown below.

The equation of motion is

182
 ∂ 2u ∂ 2u
ρ = E
∂ t2 ∂ x2

where ρ is the density and E is the Young's modulus.

For harmonic motion,

u ( x, t ) = y ( x)eiω t (1)

then,

y′′ + k 2 y = 0 , k 2 = ρω 2 / E (2)

The solution is

y=Acoskx+Bsinkx

If both ends of the rod are free (i.e., y′ =0 at x=0 and L), then B=0 and kn=nπ/L
(n=integer). Then, the eigen functions are

yn=Acos(nπx/L)

and the eigen frequencies are

ω n = E / ρ kn

Here, n determines the number of nodes.

183
 The eigen functions yn are orthogonal, i. e.,

L
∫0
ρ yn ym dx ∝ δ nm (3)

(this relation can be derived directly from (2), if yn and ym satisfy the homogeneous
boundary conditions ayn + yn′ = 0 and cym + ym′ = 0 at x=0 and L.)

Example

If a stretched string is plucked at a point, a localized disturbance is produced. The

disturbance propagates in the string and the time history of the disturbance is given as
shown in the following figure.

184
This is a seismogram. This displacement can be viewed as superposition of normal
modes as follows.
Normal modes for a string with length L, clamped at both ends are given by

yn ( x) = sin(nπ x / L)

The displacement of the string u ( x, t ) is given by the superposition of yn ( x) with weights

An, and can be written as

u ( x, t ) = ∑ A sin(nπ x / L) cos ω t
n =1
n n

where ω n = nπ c / L are the eigen frequencies, and c is the velocity of wave traveling in
the string.

The initial (at t=0) shape of the displacement u ( x, 0) is then given by

185
 ∞

u ( x, 0) = ∑ A sin(nπ x / L)
n =1
n

which can be considered as a Fourier series of u ( x, 0) .

If we assume that the initial disturbance is given by a delta function at x = xs ,

u ( x, 0) = δ ( x − xs ) , then An above can be determined by multiplying u ( x, 0) by
sin(nπx/L) and integrating it from 0 to L. Since

∫ ∫ δ (x − x )sin(nπx / L)dx = sin(nπx / L),

L L
L
sin(nπ x / L) sin(mπ x / L)dx = δ and
0 2 nm 0
s s

we obtain

An =(2/L)sin(nπxs/L)

Then the displacement is given by

∞

∑ sin(nπ x / L) sin(nπ x / L) cos ω t

2
u ( x, t ) = s n
L n =1

The following figure shows u ( x, t ) for x=xs=L/4, and is essentially the same as the
seismogram shown above. This equation can be viewed as superposition of normal
modes, sin(nπ x / L) cos ω n t , with the amplitudes (2 / L) sin(nπ xs / L) shown as
"spectrum".

186
 1.2

1
(L/2) x spectral amplitude

0.8

0.6

0.4

0.2

0
0 1 2 3 4 5 6 7 8
n

2. 2-D and 3-D Problems

The above result can be easily extended to 2-D and 3-D problems. For example,
free oscillations of a homogeneous rectangular membrane can be expressed by eigen
functions of the form,

ynm = A sin(mπ x / Lx ) sin(nπ y / Ly )

187
Here, m and n determine the number of nodes in x and y directions, respectively (see the
attached figure.).

For free oscillations of a homogeneous elastic rectangular block, the eigen

functions can have a form
n yl = A cos(lπ x / L x )cos(m πy / L y )cos(nπz / Lz )
m

188
the order numbers l, m and n determine the number of nodes in x, y and z directions,
respectively.

3. Free Oscillations of the Earth --- General ---

The free oscillations of the Earth can be formulated similarly as free oscillations
of an elastic sphere. The only difference is that it is more convenient to use a spherical
coordinate system (r, θ, ϕ) instead of the Cartesian coordinate system (x, y, z).

For a sphere with a laterally homogeneous structure, the eigen functions can be
written as,

n ylm = ARn (r )Θl (θ )Φ m (φ ) (4)

For a homogeneous elastic sphere, Rn(r) are Bessel functions, Θl(θ) are (associated)
Legendre functions and Φm(ϕ) are trigonometric functions. The three indices l, m, and n
determine the number of nodes in θ direction (meridional direction), ϕ direction
(longitudinal direction) and r direction (radial direction).
In general eigen frequencies depend on l, m, and n and can be written as n ω ml .
However, for a spherically symmetric, non-rotating Earth model, eigen frequencies do
not depend on m.

189
 Free oscillations of the Earth can be thought of as extensions of Love and
Rayleigh waves propagating many times around the Earth. When Love waves propagate
many times around the Earth, torsional free oscillations are set up. Similarly, when
Rayleigh waves propagate around the Earth many times, spheroidal oscillations are set
up. Torsional and spheroidal oscillations (modes) with order numbers l, m, and n are
written as n Tlm and n Sml and are shown in the attached figures and tables.

190
191
192
Spheroidal Modes, Bolivian Earthquake

193
 0S0 after 400,000 sec

4. Free Oscillations of the Earth --- Some Details ---

The equations of motion can be written using displacements and stresses

expressed in spherical coordinates (r, θ, ϕ). (Appendix 1)

Torsional Oscillations

We seek the solution in a form

194
 ⎛ ⎞
⎜ 0 ⎟
⎛ ur ⎞ ⎜ ⎟
⎜ ⎟ ⎜ 1 ∂ Yl m ⎟ iω t
⎜ uθ ⎟ = y1 (r ) ⎜ sin θ ∂φ ⎟ e (5)
⎜u ⎟ ⎜ ⎟
⎝ φ⎠ ⎜ ∂ Yl m ⎟
⎜ − ⎟
⎝ ∂θ ⎠

Yl m (θ , φ ) = Pl m (cosθ )( A cos mφ + B sin mφ ) m=0, 1, 2, ....., l (6)

1 ∂ Yl m ∂ Yl m
Tlm = (0, ,− ) (7)
sin θ ∂φ ∂θ

are called the vector spherical harmonics. It is easy to show that Tlm
make up an orthogonal system:

2π π 4π l (l + 1)(l + m)!
∫ ∫
0 0
(Tlm ⋅ Tlm′ ′ ) sin θ dθ dφ =
ε m (2l + 1)(l − m)!
δ ll ′δ mm′ (8)

ε m = 1 if m = 0
ε m = 2 if m ≠ 0

Substituting this into the equation of motion, we have

⎛ d 2 y1 2 dy1 ⎞ d μ ⎛ dy1 y1 ⎞ ⎛ 2 l (l + 1) ⎞
μ⎜ 2
+ ⎟+ ⎜ − ⎟ + ⎜ω ρ − μ ⎟ y1 = 0 (9)
⎝ dr r dr ⎠ dr ⎝ dr r ⎠ ⎝ r2 ⎠

if we put

⎛ dy (r ) y (r ) ⎞
y2 ( r ) = μ ( r ) ⎜ 1 − 1 ⎟
⎝ dr r ⎠

(9) can be written as

195
 dy1 (r ) 1 1
= y1 (r ) + y2 (r )
dr r μ
(10)
dy2 (r ) ⎛ (l 2 + l − 2) ⎞ 3
=⎜ 2
μ − ω 2 ρ ⎟ y1 (r ) − y2 (r )
dr ⎝ r ⎠ r

The boundary conditions are

σ rr = 0, σ rθ = 0, σ rφ = 0

at r=r1 (the Earth's surface) and r=r2 (core-mantle boundary).

Note that, using y2(r), we can write

1 ∂ Yl m ιω t ∂Ym
σ rθ = y2 (r ) e and σ rφ = − y2 (r ) l eιω t
sin θ ∂φ ∂θ

Thus, y2(r) gives the radial factor of stresses.

When μ=const, the above equation can be reduced to

d 2 y1 2 dy1 ⎛ ω 2 ρ l (l + 1) ⎞
+ +⎜ − ⎟ y1 = 0 (11)
dr 2 r dr ⎝ μ r2 ⎠

Equation (11) is one of the variations of the Bessel's differential equation. The general
solution is

1
y1 (r ) ≈ Z 1 (kr ) (12)
r l+ 2

ω
k= (13)
μ/ρ

where Zν(ξ) represents a Bessel function.

196
 In order to satisfy the boundary conditions, an appropriate kind of Bessel function
should be taken and k must take certain discrete values n k . Since (12) depends on l,
these values depend on l, and we write these values as n k l . Using (13), the eigen
frequencies n ω l can be given by

μ
ω l = n kl
ρ
n

Note that, since (11) does not contain m, the eigen values do not depend on m. Thus,
there are 2l+1 eigen functions belonging to n ω l , i.e. there is degeneracy of 2l+1 degree.
For a radially heterogeneous sphere where μ=μ(r), the solution of (9) cannot be
given by Bessel functions, but numerical integration of (10) with appropriate boundary
conditions yields eigen functions similar to (12), and eigen values n ω l . Since (10) does
not contain m, the eigen values do not depend on m.
For large l, the asymptotic expansion of Pl n gives

2 ⎡ 1 m π⎤
Pl n ≈ e mπ i ml cos ⎢ (l + )θ + π − ⎥ (14)
π l sin θ ⎣ 2 2 4⎦

Thus, uφ represents a wave propagating in θ direction. For l>>1, and l>>m,

∂Ylm ∂Yl m
≈ l Ylm and ≈ m Yl m
∂θ ∂φ

Thus, uφ >> uθ , which represents "torsional" or "Love wave" motion.

Spheroidal Oscillation

We seek the solution in a form

197
 ⎛ ⎞
⎜ 0 ⎟
⎛ ur ⎞ ⎛ Yl m ⎞ ⎜ ⎟
⎜ ⎟ ⎜ ⎟ iω t ⎜ ∂ Yl m ⎟ iω t
⎜ uθ ⎟ = y1 (r ) ⎜ 0 ⎟ e + y3 (r ) ⎜
∂θ ⎟e (15)
⎜ ⎟ ⎜ ⎟ ⎜ ⎟
⎝ uφ ⎠ ⎝ 0 ⎠ ⎜ 1 ∂ Yl ⎟
m

⎜ sin θ ∂φ ⎟
⎝ ⎠

where

S1,ml = (Ylm , 0 ,0)

∂Yl m 1 ∂Yl m
S2,m l = (0, , )
∂θ sinθ ∂φ

are vector spherical harmonics which satisfy the orthogonality condition

2π π 2π π 4π l (l + 1)(l + m)!
∫ ∫
0 0
(Sm2,l ⋅ Sm2,l' ' ) sin θ dθ dφ = ∫
0 ∫
0
(S1,ml ⋅ S1,ml' ' ) sin θ dθ dφ =
ε m (2l + 1)(l − m)!
δ ll ′δ mm′

Using the expression (15), the equation of motion for a radially heterogeneous
sphere can be reduced to a set of ordinary differential equations for y1 and y3. It can be
shown that y1 and y3 satisfy the orthogonality condition,

r2
∫ r1
ρ r 2 [ y1,n y1,n ' + l (l + 1) y3,n y3,n ' ]dr = Cδ nn '

and the eigen values n ω l do not depend on m.

For l>>1 and l>>m, ur and uθ become much larger than uφ , and the

displacement field represents "spheroidal" or "Rayleigh wave" motion.

198
199
Cartesian, Cylindrical, and Spherical Coordinates

200
201
202
203
Appendix 2

204
205
Appendix 3

206
207
Ge 162

4.1 Herglotz-Wiechert Method (LW, pp. 236-240)

Assume that T(Δ) and p(Δ) are continuous functions of Δ, as shown in Figures 1
and 2.

Fig.1 Fig. 2

dΔ r
Then, from Figure 3, tan i = r . Since p = sin i ,
dr v

208
 Fig. 3

p
dΔ = dr (1)
2
⎛r⎞
r ⎜ ⎟ − p2
⎝v⎠

∫
r0
p
Δ( p) = 2 dr (2)
2
⎛r⎞
rp
r ⎜ ⎟ − p2
⎝v⎠

where the suffix 0 denotes the values at the surface (i.e., r = r0 ) and rp is the radial
distance of the deepest point of a ray.
r
Introducing a new variable, η =
v

209
 η0

∫
p dr
Δ( p) = 2 dη (3)
p
r η 2 − p 2 dη

r
Δ(p) is known from observation, and η (r ) = , or v(r ) is a unknown function of r which we
v(r )
r
wish to determine. (Note that since p = sin i , p = η at the deepest point of the ray where
v
i = π / 2 .)

Mathematically, this is a Volterra's integral equation of the second kind. The

solution can be given by,

⎡ Δ1
⎡⎛ ⎞ ⎤ ⎤

∫
2
1 p ⎛ p ⎞
r1 = r0 exp ⎢ − ln ⎜ ⎟ + ⎜ ⎟ − 1 ⎥d Δ ⎥
⎢ (11)
⎢ π ⎢⎝ η1 ⎠ ⎝ η1 ⎠ ⎥ ⎥
⎣⎢ 0 ⎣ ⎦ ⎦⎥

The suffix 1 is used to denote values of variables at the level r1 , and let Δ1 be the value
of Δ for the ray whose deepest point is at r1 .

----------------------------------------------------------------------------------------------------------------
1
To solve this, multiply (3) by and integrate with p from η1 to η0 (see Figure
p − η12
2

4).

210
 Fig. 4

The suffix 1 is used to denote values of variables at the level r1 , and let Δ1 be the value
of Δ for the ray whose deepest point is at r1 .

η0 η0 η0

∫ ∫ ∫
1 dp p dr
Δ( p) dp = 2 dη (4)
η1
p 2 − η12 η1
p 2 − η12 p
r η 2 − p 2 dη

η0 η0

∫
⎛ p⎞ −1 dΔ ⎛ p⎞
LHS= Δ( p) cosh ⎜ ⎟ − cosh −1 ⎜ ⎟dp (5)
⎝ η1 ⎠ η1 η1
dp ⎝ η1 ⎠

⎛η ⎞
Since Δ(η0 ) = 0 and cosh −1 ⎜ 1 ⎟ = cosh −1 (1) = 0 ,
⎝ η1 ⎠

∫
0
⎛ p⎞
LHS= − cosh −1 ⎜ ⎟d Δ (6)
Δ1 ⎝ η1 ⎠

To evaluate the integral of RHS, we change the order of integration (see Figure
5),

211
 Fig. 5

η0 η

∫ ∫
1 dr 1 p
RHS= 2 dη dp (7)
r dη p 2 − η12 η 2 − p2
η1 η1

Introducing θ , by

p 2 − η12
sin θ =
2

η 2 − η12

we obtain,
π
η

∫ ∫
1 p 2
π
dp = dθ = (8)
p −η
2 2
η −p2 2 2
η1 1 0

Then,

∫
r0
1 ⎛r ⎞
RHS= π dr = π ln ⎜ 0 ⎟ (9)
r1
r ⎝ r1 ⎠
Hence.

Δ1 Δ1
⎡⎛ ⎞ ⎤

∫ ∫
2
⎛r ⎞ ⎛ p⎞ p ⎛ p⎞
π ln ⎜ 0 ⎟ = cosh −1 ⎜ ⎟d Δ = ln ⎜ ⎟ + ⎜ ⎟ − 1 ⎥d Δ
⎢ (10)
⎝ r1 ⎠ 0 ⎝ η1 ⎠ 0
⎢⎝ η1 ⎠
⎣ ⎝ η1 ⎠ ⎥
⎦

or,

212
 ⎡ Δ1
⎡⎛ ⎞ ⎤ ⎤

∫
2
⎢ 1 p ⎛ p⎞
r1 = r0 exp − ln ⎜ ⎟ + ⎜ ⎟ − 1 d Δ ⎥
⎢ ⎥ (11)
⎢ π ⎢⎝ η1 ⎠ ⎝ η ⎠ ⎥ ⎥
⎣⎢ ⎦ ⎦⎥
1
0 ⎣
------------------------------------------------------------------------------------------------------------

If p(Δ) is known, RHS of (11) can be evaluated for a given η1 to determine r1 ( η1

r
is the apparent velocity at Δ= Δ1 .). Once r1 is determined, v1 is given by v1 = 1 . More
η1
details are in the practice session.
Ge 162 Practice Session 4 Herglotz-Wiechert Method

The Herglotz-Wiechert (H-W) method is an elegant 'inversion' method which determines

the velocity distribution in the Earth from a given p-Δ curve obtained from the observation. It is
essential that a p-Δ curve is given continuously as a function of Δ. If a low-velocity zone exists
and a shadow zone is produced, then the travel time curve is segmented. In this case the p-Δ
curve is not continuous and the H-W method cannot be used. In general, the H-W method is
applicable to a structure without low-velocity zones. This is the limitation of the method.

Figure 1 gives a p-Δ curve.

213
 From equations (11) in class note, we obtain,

⎡ Δ1
⎡⎛ ⎞ ⎤ ⎤

∫
2
⎢ 1 p ⎛ p⎞
r1 = r0 exp − ln ⎜ ⎟ + ⎜ ⎟ − 1 d Δ ⎥
⎢ ⎥ (1)
⎢ π ⎢ η ⎝ η1 ⎠ ⎥ ⎥
⎢⎣ 0 ⎣⎝ 1 ⎠ ⎦ ⎥⎦
where

r
η (r ) = (2)
v( r )

Here, r1 is the radial distance to the bottoming point of the ray reaching the distance Δ1 ,
and

214
 r1
η1 = (3)
v(r1 )

Since i = π / 2 at the bottoming point, the ray parameter for the ray reaching Δ = Δ1 is
given by

r1 r
p1 = sin i1 = 1 ≡ η1 (4)
v(r1 ) v(r1 )

Thus, if a p-Δ curve is given, then for a given Δ1 , we can determine η1 . Then,
we can integrate (1) to obtain r1 . Once r1 is determined, then v(r1 ) can be computed from

r1
v(r1 ) = (5)
η1

Doing this for all Δ1 , we can determine v(r ) as a function of r.

1. The following table shows the travel times (2nd column) which are consistent with the ISC
travel-time curve given in class handout. Then the travel time data are numerically differentiated
to compute the ray parameter p which is shown on the 3rd column. The unit of p is sec (i.e., the
unit of the distance is converted to radian in this computation.). The following table lists the data
up to Δ = 30° , and a more complete table up to Δ = 98° is given in i_p-delta_2 in
/home/ftp/pub/hiroo/ge162.dir/practice_4.dir.
Plot the travel time and p in i_p-delta_2 as a function of distance, and make sure that the
travel times are consistent with the ISC data. Also, check the consistency between t and p. An
enlarged travel-time curve is shown below. (Spot checks at Δ = 30°, 60° ,90° are sufficient.)

Table

dist(deg) t(sec) p(sec)

0.00000 0.00000 1145.91516
2.00000 35.38889 828.24194
4.00000 63.90000 813.59973
6.00000 92.22500 807.87048

215
 8.00000 120.30000 802.14056
10.00000 147.98334 788.77106
12.00000 175.25000 773.49268
14.00000 201.87500 753.43927
16.00000 227.92500 736.25037
18.00000 253.14999 710.46747
20.00000 276.72501 627.38831
22.00000 297.45001 572.95752
24.00000 317.10001 555.76868
26.00000 336.14999 532.85071
28.00000 354.47501 518.52673
30.00000 372.45001 509.93188
........ ......... .........

216
217
2. Then, compute r1 for Δ1=0 ° to 98 ° at 2 ° intervals by carrying out the integral given
by (1)*.

3. Then, from r1 thus computed and η1 , determine v(r1 ) .

4. Plot v(r1 ) as a function of r1 , and compare the result with the P-wave structure shown
in class handout (section 2.1, a numerical table, jeffreys_model, for the Jeffreys model is
in practice_4.dir).

* Preferably, you should write a simple integration program (trapezoidal rule is

adequate), but if you find it difficult, a simple program hw3.f is provided in
practice_4.dir.

Note: The trapezoidal rule is given by,

∫ ∑( f
L
f ( x)dx ≈ i −1 + f i )Δx / 2 , where Δx = L /( N − 1) , and fi = f ((i − 1) * Δx)
0
i=2

Ge162

4.2 Tomography (LW, pp. 240-249)

The structure of the Earth is heterogeneous in 3 dimensions. To determine such a

3-dimensional structure, various tomographic methods are used. The basic principle can
be illustrated as follows.
Consider a square area as shown in the figure.

218
 Then, assume that seismic sources (X) and stations (square) are distributed on the
sides. The velocity is a function of space. Divide the whole area into cells (i ,j). The
wave speed in cell (i ,j) is vij (or slowness sij =1/ vij ) which we wish to determine from
observations. If we assume that the ray paths are straight (this assumption is not valid in
general for heterogeneous media). The travel time from source p to station q can be
written as

t pq = ∑ lijpq / vij = ∑ lijpq sij (1)

i, j i, j

where lijpq is the path length in the (i ,j) cell for the p-q source-station ray. If the
geometry and paths are fixed, as in this case, the problem is linear, but the problem is in
general nonlinear, and the problem is usually linearized, as is done in the earthquake
location problem.
If the problem is linear, or is linearized, (1) can be solved by the method of least
squares or some other inversion methods.
We renumber the cells sequentially such that

k = i + ( j − 1)nx

where nx is the number of cells in a row. Then (1) can be written as

t pq = ∑ lkpq sk (2)
k

219
 Simple examples are given in the following.

If we place the sources p=1, 2, 3, and 4, and the receivers q=1, 2, 3, and 4, as
shown, we can have the ray geometry as shown in the following figure.

Then equation (2) can be written as

220
 ⎛ l111 l211 l311 . . . . . .⎞ ⎛ t11 ⎞
⎜ ⎟ ⎜ ⎟
⎜ l112 l212 l312 ⎟ ⎜ t12 ⎟
⎜ l113 l213 l313 ⎟ ⎜ t13 ⎟
⎜ ⎟ ⎜ ⎟
⎜ . ⎟ ⎛ s ⎞ ⎜ t14 ⎟
⎜ . ⎟ ⎜ 1 ⎟ ⎜ t21 ⎟
⎜ ⎟ ⎜ s2 ⎟ ⎜ ⎟
⎜ ⎟⎜ ⎟ ⎜ . ⎟
⎜ ⎟ ⎜ s3 ⎟ ⎜ . ⎟
⎜ ⎟ ⎜ s4 ⎟ ⎜ ⎟
⎜ . ⎟⎜ ⎟ ⎜ . ⎟
⎜ . ⎟ ⎜ s5 ⎟ = ⎜ . ⎟
⎜ ⎟⎜ s ⎟ ⎜ ⎟
⎜ . ⎟⎜ 6 ⎟ ⎜ . ⎟
⎜ ⎟ ⎜ s7 ⎟ ⎜ ⎟
⎜ . ⎟⎜ s ⎟ ⎜ . ⎟
⎜ . ⎟ 8 ⎜ . ⎟
⎜ ⎟ ⎜⎝ s9 ⎟⎠ ⎜ ⎟
⎜ . ⎟ ⎜ . ⎟
⎜ . ⎟ ⎜ . ⎟
⎜ ⎟ ⎜ ⎟
⎜ . ⎟ ⎜ . ⎟
⎜l ⎟ ⎜ ⎟ (3)
⎝ 144 l244 l344 . . .⎠ ⎝ t44 ⎠

This problem can be solved with the method of least squares in the same way as
that used in the earthquake location problem.

Example 1.

s1 = s3 = s5 = s7 = s9 = 2 , s2 = s4 = s6 = s8 = 1 (checker board).

221
 The solution is given as follows.

checker_b
cross_hole_simple
cell s Δs
1: 1.892 0.011
2: 1.218 0.015
3: 1.890 0.011
4: 0.892 0.010
5: 2.219 0.010
6: 0.890 0.010
7: 1.891 0.011
8: 1.219 0.015
9: 1.890 0.011
RMS of residual 0.01160

Actually, for this geometry the inverse problem is fairly ill-posed, and even if no
noise is applied, the solution is unstable (i.e., the standard error is large), as shown above.
If random noise of up to 10% is added to t pq , the error becomes very large, as
shown below.

222
checker_b
cross_hole_simple
cell s Δs
1: 2.149 3.187
2: 0.646 4.242
3: 2.091 3.190
4: 1.110 2.745
5: 2.083 2.750
6: 0.687 2.743
7: 1.478 3.176
8: 2.104 4.222
9: 1.538 3.179
RMS of residual 3.31218

This instability arises from the inadequate source-receiver geometry.

Example 2.

The source-receiver geometry is now changed to that shown in the following

figure.

223
 Then the solution becomes more stable as shown below.

No- noise case

checker_b
two_sides
cell s Δs
1: 1.999 0.001
2: 1.001 0.001
3: 1.999 0.001
4: 1.001 0.001
5: 2.000 0.001
6: 1.001 0.001
7: 1.999 0.001
8: 1.001 0.001
9: 1.999 0.001
RMS of residual 0.01479

With random noise of up to 10 %:

checker_b
two_sides
cell s Δs
1: 2.046 0.138
2: 0.922 0.154
3: 1.831 0.138
4: 0.856 0.154
5: 1.814 0.144
6: 1.222 0.154
7: 2.011 0.138
8: 1.062 0.154
9: 2.203 0.138
RMS of residual 2.90495

224
Caveats

Although tomographic methods provide interesting 3-D structures of the Earth,

the following caveats are in order.

1. The results depend on the initial parameterization of the cells and rays.
2. In a heterogeneous medium, the rays are not straight, and ray bending must be
considered.
3. The effect of finite wave length must be considered in resolution.
4. The solutions are often regularized (damped) to avoid instability.

Several recent examples are shown in the following

Global Depth Slices

225
226
J. Ritsema [written communication, 2002]

227
Spectral Structure of the Structural Heterogeneity in Earth

228
A. M. Dziewonski, Global seismic tomography: past, present and future, in
Problems in Geophysics for the New Millennium, Editrice Compositori, 2000.

229
"Slab" Structure in Deep Interior

Ritsema, J., and van Heijst, H. J., Seismic imaging of structural heterogeneity in Earth's
mantle: evidence for large-scale mantle flow, Science Progress, 83 (3), 243-259, 2000.

230
"Slab" Structures

231
232
Fukao, et al., Reviews of Geophysics, 39, 291, 2001.
Ge162 Practice Session 5 (Optional) Simple Tomography

From class note, a tomography problem given below can be formulated by

233
 ⎛ l111 l211 l311 . . . . . .⎞ ⎛ t11 ⎞
⎜ ⎟ ⎜ ⎟
⎜ l112 l212 l312 ⎟ ⎜ t12 ⎟
⎜ l113 l213 l313 ⎟ ⎜ t13 ⎟
⎜ ⎟ ⎜ ⎟
⎜ . ⎟ ⎛ s ⎞ ⎜ t21 ⎟
⎜ . ⎟ ⎜ 1 ⎟ ⎜ t22 ⎟
⎜ ⎟ ⎜ s2 ⎟ ⎜ ⎟
⎜ ⎟⎜ ⎟ ⎜ . ⎟
⎜ ⎟ ⎜ s3 ⎟ ⎜ . ⎟
⎜ ⎟ ⎜ s4 ⎟ ⎜ ⎟
⎜ . ⎟⎜ ⎟ ⎜ . ⎟
⎜ . ⎟ ⎜ s5 ⎟ = ⎜ . ⎟
⎜ ⎟⎜ s ⎟ ⎜ ⎟
⎜ . ⎟⎜ 6 ⎟ ⎜ . ⎟
⎜ ⎟ ⎜ s7 ⎟ ⎜ ⎟
⎜ . ⎟⎜ s ⎟ ⎜ . ⎟
⎜ . ⎟ 8 ⎜ . ⎟
⎜ ⎟ ⎜⎝ s9 ⎟⎠ ⎜ ⎟
⎜ . ⎟ ⎜ . ⎟
⎜ . ⎟ ⎜ . ⎟
⎜ ⎟ ⎜ ⎟
⎜ . ⎟ ⎜ . ⎟
⎜l ⎟ ⎜ ⎟ (3)
⎝ 166 l266 l366 . . .⎠ ⎝ t66 ⎠

As we did in the practice session for earthquake location, this problem can be solved
using the method of least squares. (3) can be written as

Am = d (4)

where A is an 18x9 matrix and m and d are column vectors containing the parameters to
be determined and the data, respectively.
The normal equation is,

AT Am = AT d (5)

234
If AT A is not singular, the formal least-squares solution is given by

m = ( AT A) −1 AT d (6)

and the error estimates are determined by the variance of the data and the diagonal
elements of the inverse matrix of the normal equation (6 ). Usually, we write the
uncertainty in mi by Δmi , and compute it by

∑ (t
N

Δmi = cii o
( kl ) − t(ckl ) ) 2 /( N − N p ) , i=1, 2, 3,...,9 (7)
( kl ) =1

where N p is the number of parameters (here 9) and cii are the diagonal elements of

( AT A) −1 .

1. Determine the slowness si, i=1, 2,3,.....9, and the associated errors.
The matrix elements and the data are in o_lsq_mat_ts_0.1 and o_lsq_rhs_ts_0.1 in
/home/ftp/pub/hiroo/ge162.dir/practice5.dir, respectively. In this computation, the

235
length of the side is 10 km, and errors up to 10% are added to the data. (This is the same
geometry as Example 2 in class note. The numerical values of the standard errors may be
slightly different, because of the difference in the definition.)
Actually, in the real problems, the computations of the ray paths and the path
lengths in each cell are most difficult. In this problem, they are computed and the values
are put in o_lsq_mat_ts_0.1. (there are small round-off errors of the order of 0.3 %.)
Check the values for the first 3 rows of o_lsq_mat_ts_0.1.
Try to write your own program to do this problem, but if you find it difficult, you
can use the program tomo2.f provided in practice_5.dir. It takes an input file i_tomo2
which contains the names of the files for the matrix elements and the data. It also
requires c_tomo2, but this should not be changed (it has a constant for regularization of
the matrix inversion.).

2. Drop the data for source #6, and try the same.

3. Drop the data for sources #5 and #6, and try the same.

236
o_lsq_mat_ts_0.1
checker_b

two_sides
9 18 0.1000
⎛ 10.02 9.99 10.02 0.00 0.00 0.00 0.00 0.00 0.00⎞
⎜ 10.56 5.28 0.00 0.00 5.28 10.56 0.00 0.00 0.00 ⎟
⎜ ⎟
⎜ 9.05 0.00 0.00 3.03 12.01 3.03 0.00 0.00 9.05 ⎟
⎜ ⎟
⎜ 0.00 5.28 10.56 10.56 5.28 0.00 0.00 0.00 0.00 ⎟
⎜ 0.00 0.00 0.00 10.02 9.99 10.02 0.00 0.00 0.00 ⎟
⎜ ⎟
⎜ 0.00 0.00 0.00 10.56 5.28 0.00 0.00 5.28 10.56 ⎟
⎜ 0.00 0.00 9.05 3.03 12.01 3.03 9.05 0.00 0.00 ⎟
⎜ ⎟
⎜ 0.00 0.00 0.00 0.00 5.28 10.56 10.56 5.28 0.00 ⎟
⎜ 0.00 0.00 0.00 0.00 0.00 0.00 10.02 9.99 10.02 ⎟
⎜ ⎟
⎜ 10.02 0.00 0.00 9.99 0.00 0.00 10.02 0.00 0.00 ⎟
⎜ ⎟
⎜ 10.56 0.00 0.00 5.28 5.28 0.00 0.00 10.56 0.00 ⎟
⎜ 9.05 3.03 0.00 0.00 12.01 0.00 0.00 3.03 9.01 ⎟
⎜ ⎟
⎜ 0.00 10.56 0.00 5.28 5.28 0.00 10.56 0.00 0.00 ⎟
⎜ 0.00 10.02 0.00 0.00 9.99 0.00 0.00 10.02 0.00 ⎟
⎜ ⎟
⎜ 0.00 10.56 0.00 0.00 5.28 5.28 0.00 0.00 10.56 ⎟
⎜ 0.00 3.03 9.05 0.00 12.01 0.00 9.05 3.03 0.00 ⎟
⎜ ⎟
⎜ 0.00 0.00 10.56 0.00 5.28 5.28 0.00 10.56 0.00 ⎟
⎜ 10.02 ⎟⎠
⎝ 0.00 0.00 10.02 0.00 0.00 9.99 0.00 0.00

237
o_lsq_rhs_ts_0.1
checker_b
two_sides
9 18 0.1000

238
⎛ 47.28 ⎞
⎜ ⎟
⎜ 44.36 ⎟
⎜ 67.07 ⎟
⎜ ⎟
⎜ 44.95 ⎟
⎜ 39.69 ⎟
⎜ ⎟
⎜ 43.15 ⎟
⎜ 60.62 ⎟
⎜ ⎟
⎜ 51.73 ⎟
⎜ ⎟
⎜ 51.76 ⎟
⎜ 49.93 ⎟
⎜ ⎟
⎜ 49.19 ⎟
⎜ 68.50 ⎟
⎜ ⎟
⎜ 45.47 ⎟
⎜ 36.96 ⎟⎟
⎜
⎜ 51.15 ⎟
⎜ ⎟
⎜ 61.99 ⎟
⎜ 46.84 ⎟
⎜⎜ ⎟
⎜⎝ 53.53 ⎟⎠⎟

239
Ge 162

5. Earthquake Source Theory

5.1 Static Source

5.1.1 Static Displacement Field Due to a Single Point Force in Infinite Homogeneous
Medium (LW, pp. 323-331)

1. Description of the Problem

Consider an isotropic infinite homogeneous elastic medium with density ρ and elastic
constants λ and μ.
We apply a force at point O, and want to determine the displacement u at point P.
We assume that the outer boundary at infinity is constrained, i.e., u =0 at infinity.

240
2. Qualitative Solution

Suppose we apply a force F to a small sphere of radius a at the origin. We

consider the displacement on the plane containing the force. It is not difficult to imagine
that the resulting displacement field is given by the arrows shown in the following figure.
The magnitude of u may be approximately determined as follows.
Consider a sphere with radius r. In equilibrium, the body force F acting
at the center of this sphere must be balanced by stresses acting on the surface. The figure
suggests that the stress on the surface of this sphere is compressional at C, shear at A, and
half shear and half compression at B.

Let σ be the magnitude of this stress. Then, to the first approximation

F = 4π r 2σ (1)

Let u(r) be the magnitude of the displacement at r, then, from the geometry shown in the
figure, du/dr gives the magnitude of the strain at r.

du F
−E ≈σ ≈ (2)
dr 4π r 2

241
(The negative sign is taken because u(r) decreases as r increases.) Here E˜ represents an
appropriate elastic constant, and is of the order of λ or μ . Integrating (4), and using
u (r → ∞) = 0 , we obtain

F
u (r ) ≈ (3)
4π Er

Actually, a rigorous derivation shows (in the next section) that, if we define
the polar coordinates as shown in the figure,

F
ur = sin θ
4πμ r
(4)
F α
uθ = (1 − ) cos θ
4πμ r 2

where

λ+μ
α=
λ + 2μ

242
 Since for most solids, λ = μ and α =2/3, we see that (3) is a good approximation
of (4) and that the directions of the displacement shown in the figure are consistent with
(4).

3. Elasto-static Solution of the Problem

Definition

a. Point Force of Strength F

When we consider a force in continuum, we introduce force per unit volume, or

force per unit mass. Let f be a force per unit mass, then ρ f gives force per unit
volume.
Consider a small volume δ V, and apply ρ f to this volume. Then

F = lim ( ρ f δ V )
δ V →0

defines a point force. F =F is the strength of this force.

b. Three-Dimensional Delta Function

Three-Dimensional Delta Function δ (r) can be deined by

δ (r ) = 0 , r ≠ 0, ∫ δ (r)dV = 1
V

where V is a volume which includes the origin r=0. A convenient expression for this is

1 2
δ (r ) = − ∇ (1/ r) (5)
4π

Direct differentiation shows that

∇ 2 (1/ r) = 0 , if r ≠ 0 .

243
We can show that

∫ ∇ (1/ r)dV = −4π ,

V
2
V contains the origin. (6)

Here, following the expression for the Gauss' theorem, ∫

V
∇ 2 (1/ r )dV is understood to
mean ∫ ∇(1/ r ) ⋅ ndS .
S

Although this looks like a very simple problem, it is not that easy. We have to start
with elasto-static equation of equilibrium. From equation (29) of 3.1, we have

ρ f + (λ + 2μ )graddivu − μ curlcurlu = 0 (7)

where f is the force per unit mass. Consider a point force of magnitude F at the origin.

ρ f = Faδ (r ) (8)

where a is the unit vector in the direction of the force. By using (5) we have

⎛ a ⎞ ⎡ ⎛ a ⎞ ⎛ a ⎞⎤
ρ f = −F∇2 ⎜ ⎟ = − F ⎢graddiv ⎜ ⎟ − curlcurl ⎜ ⎟⎥ (9)
⎝ 4π r ⎠ ⎣ ⎝ 4π r ⎠ ⎝ 4π r ⎠ ⎦

We look for a solution in the form

u = graddivAp − curlcurlAs (10)

This form may appear somewhat artificial, but it is suggested from the fact that any
displacement field can be represented by a sum of solenoidal and irrotational fields and
that the forcing term (9) is given in this form. Substituting (9) and (10) into (7), we have

244
 ⎛ a ⎞ ⎛ a ⎞
− Fgraddiv ⎜ ⎟ + Fcurlcurl ⎜ ⎟ + (λ + 2μ )graddiv(graddivAp − curlcurlAp )
⎝ 4π r ⎠ ⎝ 4π r ⎠
− μ curlcurl(graddivAs − curlcurlAs ) = 0

In the third term curlcurlAs is replaced by curlcurlAp , because its divergence vanishes
anyway. Similarly, graddivAp in the fourth term is replaced by graddivAs . Using the
relation

graddivAp − curlcurlAp = ∇ 2 Ap

and the similar relation for As , we obtain

⎛ a ⎞ ⎛ a ⎞
graddiv ⎜ − F + (λ + 2 μ )∇ 2 Ap ⎟ + curlcurl ⎜ F − μ∇ 2 As ⎟ = 0 (11)
⎝ 4π r ⎠ ⎝ 4π r ⎠

This equation is satisfied if,

a
(λ + 2μ )∇ 2 Ap = F (12)
4π r

a
μ∇ 2 As = F (13)
4π r

If we put Ap = Ap a and As = As a , (12) and (13) can be reduced to Poisson's equation,

F F
∇ 2 Ap = and ∇ 2 As = (14)
4π (λ + 2 μ )r 4πμr

2
Since ∇ 2 r = , we have
r

245
 F
Ap = r (15)
8π ( λ + 2μ )

and

F
As = r (16)
8πμ

Using (10)

u = graddivAp − graddivAs + ∇ 2 As (17)

Substituting (15) and (10) into (17), we have for the i-th component of displacement
for a unit force (F=1) in j-th direction, uij ,

1 ∂ ∂r 1 ∂ ∂r 1 2
uij = − + δ ij ∇r
8π (λ + 2μ ) ∂ xi ∂ x j 8πμ ∂ xi ∂ x j 8πμ
(18)
1 ⎛ λ + μ ∂ 2r ⎞ 1
= ⎜⎜ δ ij ∇ r −
8πμ ⎝
2

λ + 2μ ∂ xi∂ x j
⎟⎟ =
8πμ
(δ ij∇ 2r − α r,ij ) = u ij
⎠

where

λ+μ
α= (2/3 for most solids) (19)
λ + 2μ

Equation (18) gives the solution of our problem. Note that uij = uij (symmetric). uij is
often called the Somigliana Tensor. We write all the components explicitly,

246
 1 ⎡2 ⎛ 1 x12 ⎞ ⎤ 1 ⎛ x1 x2 ⎞
u11 = ⎢ − α ⎜ − 3 ⎟⎥ , u12 = ⎜α ⎟
8πμ ⎣ r ⎝ r r ⎠⎦ 8πμ ⎝ r 3 ⎠
1 ⎛ x1 x3 ⎞ 1 ⎡2 ⎛ 1 x22 ⎞ ⎤
u13 = ⎜α ⎟, u22 = ⎢ − α ⎜ − 3 ⎟⎥ (20)
8πμ ⎝ r 3 ⎠ 8πμ ⎣ r ⎝ r r ⎠⎦
1 ⎛ x2 x3 ⎞ 1 ⎡2 ⎛ 1 x32 ⎞ ⎤
u23 = ⎜α ⎟, u33 = ⎢ − α ⎜ − 3 ⎟⎥
8πμ ⎝ r 3 ⎠ 8πμ ⎣ r ⎝ r r ⎠⎦

Let us consider the simplest case. A force with a magnitude F is applied in x1

direction. We take the polar coordinates (r , θ , φ ) as shown in the figure.

Then

⎛ ur ⎞ ⎛ sin θ cos φ sin θ sin φ cosθ ⎞ ⎛ u11 ⎞

⎜ ⎟ ⎜ ⎟⎜ ⎟
⎜ uθ ⎟ = ⎜ cosθ cos φ cosθ sin φ − sin θ ⎟ ⎜ u12 ⎟ (21)
⎜ u ⎟ ⎜ − sin φ cos φ 0 ⎟⎠ ⎜⎝ u31 ⎟⎠
⎝ φ⎠ ⎝

On x1 − x3 plane, φ =0,

F
ur = sin θ u11 + cos θ u31 = sin θ (22)
4πμ r

247
 F ⎛ α⎞
uθ = cos θ u11 − sin θ u31 = ⎜1 − ⎟ cos θ (23)
4πμ r ⎝ 2 ⎠

These are the results given by (4).

5.1.2 Static Displacement Field Due to Force Couples

248
Single Couple

If we apply a force at (ξ1, ξ 2, ξ3 ) instead of at the origin, the displacement at

( x1, x2, x3 ) can be still given by (18) with

r 2 = ( x1 − ξ1 ) 2 + ( x2 − ξ 2 ) 2 + ( x3 − ξ3 ) 2 ≡ ( xi − ξi ) 2 .

(1/2)dξ2

(ξ1, ξ2, ξ3)

1 1
If we apply a force F in x1 direction at ξ 2 + dξ 2 , and F in - x1 direction at ξ 2 − dξ 2 ,
2 2

the displacement at ( x1, x2, x3 ) is

1 1
Fui1 (ξ1, ξ 2 + dξ 2 , ξ3 , x1, x2, x3 ) − Fui1 (ξ1, ξ 2 − dξ 2 , ξ3 , x1, x2, x3 )
2 2
∂u 1
= F i dξ 2 + o(d 2ξ 2 )
∂ξ 2
(24)

249
Since r 2 = ( x1 − ξ1 ) 2 + ( x2 − ξ 2 ) 2 + ( x3 − ξ3 ) 2 ≡ ( xi − ξi ) 2 ,

∂r ∂r
=− (25)
∂ξi ∂ xi

(we often write this as r ,i = −r,i .)

therefore, from (18)

∂ uij ∂uj
=− i (26)
∂ξ k ∂ xk

Thus, the displacement for this force couple is given by

∂ ui1
−F dξ 2 + o(d 2ξ 2 ) (27)
∂ x2

If we make dξ 2 → 0 , F → ∞ , so that Fδξ 2 → M (finite), the displacement due to this

single couple (we will denote it by Ui ) can be written as

∂ui1
Ui = M = Mui1,2 (27')
∂ξ 2
or

∂ui1
Ui = −M = − Mui1,2 (27'')
∂x2

Thus, the displacement due to this single couple placed at the origin can be obtained by
replacing F by M in (20), taking the derivative of (20) with respect to x2, and changing
the sign.

250
 ⎡ x2
M ⎛ x2 x12 x2 ⎞ ⎤
U1 = − ⎢ −2 − α −
⎜ 3 + 3 ⎟⎥ (28)
8πμ
⎣ r
3
⎝ r r 5 ⎠⎦
M ⎡ x1 x x2 ⎤
U2 = − α ⎢ 3 − 3 1 52 ⎥ (29)
8πμ ⎣ r r ⎦
M ⎡ x1 x2 x3 ⎤
U3 = − α ⎢ −3 5 ⎥ (30)
8πμ ⎣ r ⎦

Similarly if we apply a single couple as shown in the following figure, we obtain

∂ui2
Ui = M = Mui2,1
∂ξ1

or

∂ui2
Ui = −M = − Mui2,1
∂x1

and the x1, x2, x3 components of the displacement can be given by

M ⎡ x2 x2 x12 ⎤
U1 = − α −3 5 ⎥ (31)
8πμ ⎢⎣ r 3 r ⎦

M ⎡ x1 ⎛ x1 x22 x1 ⎞ ⎤
U2 = − ⎢ −2 3 − α ⎜ − 3 + 3 5 ⎟ ⎥ (32)
8πμ ⎣ r ⎝ r r ⎠⎦

M ⎡ x1 x2 x3 ⎤
U3 = − α ⎢ −3 5 ⎥ (33)
8πμ ⎣ r ⎦

251
Double Couple

It is evident that for a double couple as shown in the figure, the displacements can
be given by

∂ui2
Ui = M = M (ui1,2 + ui2,1 ) (34)
∂ξ1

or

∂ui2
U i = −M = − M (ui1,2 + ui2,1 ) = − M (u1,2
i
+ u2,1
i
) (35)
∂x1

From the sum of (28) and (31), (29) and (32), and (30 and (33), we obtain

M ⎛ x2 ⎞ 2α M ⎛ x2 x12 x2 ⎞ M x2 ⎡ ⎛ x12 ⎞ ⎤
U1 = ⎜2 ⎟− ⎜ 3 − 3 ⎟ = ⎢1 − α ⎜ 1 − 3 ⎟⎥ (36)
8πμ ⎝ r 3 ⎠ 8πμ ⎝r r 5 ⎠ 4πμ r 2 r ⎣ ⎝ r 2 ⎠⎦

M ⎛ x1 ⎞ 2α M ⎛ x1 x22 x1 ⎞ M x1 ⎡ ⎛ x22 ⎞ ⎤
U2 = ⎜2 ⎟− ⎜ 3 − 3 ⎟ = ⎢1 − α ⎜ 1 − 3 ⎟⎥ (37)
8πμ ⎝ r 3 ⎠ 8πμ ⎝r r 5 ⎠ 4πμ r 2 r ⎣ ⎝ r 2 ⎠⎦

M ⎛ x1 x2 x3 ⎞ M ⎛ x1 x2 x3 ⎞
U3 = ⎜ 6α ⎟= ⎜ 3α ⎟ (38)
8πμ ⎝ r ⎠ 4πμ r 2 ⎝
5
r3 ⎠

252
 For double couples with different orientations, similar expressions can be obtained.

Using the polar coordinates (r, θ, φ),

the displacement components, ur, uθ, uφ are given by

253
 M ⎛ α⎞ 2
ur = ⎜ 1 + ⎟ sin θ sin 2φ (39)
4πμ r ⎝ 2 ⎠
2

M ⎛1 α ⎞
uθ = ⎜ − ⎟ sin 2θ sin 2φ (40)
4πμ r 2 ⎝ 2 2 ⎠
M
uφ = (1 − α ) sin θ cos 2φ (41)
4πμ r 2

⎛⎛ α ⎞ 2 ⎞
⎜ ⎜1 + 2 ⎟ sin θ sin 2φ ⎟
⎛ ur ⎞ ⎜⎝ ⎠ ⎟
⎜ ⎟ M ⎜⎛ 1 α ⎞ ⎟
Or, ⎜ uθ ⎟ = 4πμ r 2 ⎜ ⎜ 2 − 2 ⎟ sin 2θ sin 2φ ⎟ (42)
⎜u ⎟ ⎜⎝ ⎠ ⎟
⎝ φ⎠ ⎜ (1 − α ) sin θ cos 2φ ⎟
⎜⎜ ⎟⎟
⎝ ⎠

On the x1 - x2 plane, θ = π / 2 , uθ = 0 , and

⎛ α⎞
ur ≈ ⎜ 1 + ⎟ sin 2φ
⎝ 2⎠
uφ ≈ (1 − α ) cos 2φ

The azimuthal variation of ur and uφ is shown in the following figure.

254
The displacement field on a circle x12 + x22 = const is also shown.

Single Couple and Double Couple with Arbitrary Orientation

If the force is in k = (k1 , k2 , k3 ) direction and the arm is in l = (l1 , l2 , l3 ) direction,

then

∂uik
U iSC =
∂sl

∂
Here, is the derivative in l direction, and l is taken positive in the direction towards
∂sl
∂ ∂
the force oriented in positive k direction. Then, since = lq and uik = uip k p ,
∂sl ∂ξ q

U iSC = uip ,q k p lq (43)

Then, the displacement due to a unit double couple is

255
 U iDC = uip , q (k p lq + kq l p ) (44)

Ge 162 Problem #6 Double Couple and Dipole

Using the expressions discussed in 5.1.2 (i.e., uiSC = uip ,q k p lq ), we can write the
displacement due to a force dipole as uiDP = uip , q k p lq where k and l are the unit
vectors in the direction of force and the arm, respectively. k and l are parallel
but they can be of either opposite or the same direction, depending on the
orientation of the force (i.e., inward or outward).
Show that the force system with two dipoles (a) shown below is equivalent
to the double couple (b) (all the dipoles and the couples are with unit moment.).

Ge 162

5.1.3 Elastic Dislocation

1. Qualitative Description of the Problem

256
 Consider a fault shown in the figure. The displacement on one side is u+ and the
other side is u-, so that the displacement discontinuity is Δu=u+-u- . This is called fault
offset and is often denoted by D. Since we already know the displacement field due to a
point force and force couples, we will first try to produce, by applying forces or force
couples to the medium without dislocation, a displacement field similar to that caused by
faulting.

Since an earthquake represents transition from one equilibrium state to another,

there should be no net force and no net moment. Thus, a single force or a single couple is
not appropriate to represent an earthquake. A double couple is an appropriate force
system to represent an earthquake.
Now, we try to find a double couple applied in a medium without dislocation
which yields a displacement field similar to that produced by a fault (i.e. dislocation).
As shown in the figure, to produce D over a small distance of Δw , in the
medium without dislocation, we need to apply a force couple with one force
approximately equal to

257
 u+ − u−
F =μ S (1)
Δw

where S is the fault area.

We reduce Δw while maintaining D = u + − u − ,

lim F Δw = μ (u + − u − ) S = μ DS
Δw→∞

which means a moment

M = μ DS (2)

is required to produce the offset D over S.

To produce D over an extended area S, we can distribute double couples on S as
shown in the figure. In this case, the effects of the vertical forces are cancelled within S,
and the overall force system can produce a dislocation-like displacement field.

2. Definition of Elastic Dislocation

First we need to give a precise definition of "Elastic Dislocation" . Create a thin

cavity in an elastic medium. Let one side of this cavity be Σ + and the other be Σ − .
Apply force on these surfaces to move the Σ + side by u + and Σ − side by u − Then,

258
make the gap very small so that Σ + and Σ − form an open surface Σ . Then weld the two
surfaces and remove the force. Now we have a displacement discontinuity

Δu = u + − u −

across Σ . This is called the elastic dislocation. Note that, for a dislocation created this
way, the traction must be continuous across Σ .
If u + and u − are tangential to Σ , Δu is called the shear dislocation.

3. Description of the Problem

Consider an infinite homogeneous (isotropic) elastic medium in equilibrium.

Then create a dislocation Δu on Σ .
This dislocation causes deformation throughout the medium which is in new
equilibrium state.
We want to determine the displacement field u in the medium caused by this
dislocation.

259
4. Elasto-Static Theory of Dislocation

We use Volterra's dislocation theory to show

u k (Q) = ∫ Δu τ ν dS
Σ
k
i ij j (3)

where uk(Q) is the k-th component of displacement at Q ; τ ijk is the ij component of

stress on Σ due to a unit force in k-th direction at Q (the unit of τ ijk should be

understood
as stress/force=1/area ); Δui is the displacement discontinuity (dislocation) across Σ. i.e.
Δui =ui (on Σ + )-ui (on Σ − ). Also,

Σ + +Σ − : Closed surface defining the dislocation surface.

Σ : Open surface formed by Σ + and Σ − when they are
made coincident with each other.
ν : unit vector normal to Σ pointing from the
negative side to the positive side.
n+ =ν on Σ + , n − = ν on Σ −

Equation (3) will be proved in the next section.

260
 Let us consider one of the simplest cases.
Suppose we have a very small planar shear dislocation surface dΣ at P which is
perpendicular to x2 axis. We consider a displacement discontinuity in x1 direction.

Then Δu1 ≠ 0 , Δu2=Δu3=0 and ν = (0,1,0).

In this case (3) gives

261
 u k (Q) = d Σ[Δu1 (τ 11k ν 1 + τ 12k ν 2 + τ 13k ν 3 )
+ Δu2 (τ 21k ν 1 + τ 22
k
ν 2 + τ 23k ν 3 )
(4)
+ Δu3 (τ ν + τ ν + τ ν )]
k
31 1
k
32 2
k
33 3

= d ΣΔu1τ 12k

Note that
τ k = μ (u1,2
12
k
+ u2,1
k
)

Referring to (34) of 5.1.2, this means that (τ 12k / μ ) = (u1,2

k
+ u2,1
k
) is the k-th component of

displacement, U k , at Q due to a unit double couple placed on the x1 − x2 plane at P.

Thus,

u k (Q) = dΣΔu1 (μU k ) = (μΔu1dΣ)U k (5)

This means that u k (Q) is equal to the k-th component of displacement at Q due to a
double couple of moment M = μΔu1dΣ placed at P. Since this relation evidently holds
for any dΣ, we can conclude that a shear dislocation Δus over a surface S is equivalent to
a double couple whose total moment is

M=
∫ μΔu d Σ =μΔu S
S
s s (6)

262
5. Volterra's Dislocation Theory
We first introduce the following notation for the stress strain relation:

τ ij = λδ ij uk ,k + μ (ui , j + u j ,i ) = cijpq u p ,q (7)

where

cijpq = λδ ijδ pq + μ (δ ipδ jq + δ jpδ iq ) (8)

Then the following symmetry relations hold.

cijpq = c jipq = cijqp = c pqij (9)

Note:
Among 81 components of cijpq, most of them are zero. Nonzero components are

c1111=c2222=c3333= λ + 2μ
c1122=c2211=c1133=c3311=c2233=c3322= λ
c1212=c1221=c2112=c2121=c1313=c1331= μ
c3113=c3131=c2323=c2332=c3223=c3232= μ

263
Then, the equation of equilibrium can be written as:

ρ f i + τ ij , j = 0, or ρ fi + cijpq u p ,qj = 0 (10)

This notation is convenient for the derivation given in the following.

Consider an elastic medium (homogeneous, isotropic) bounded by a closed
surface S1. Within this medium consider a closed surface consisting of Σ + and Σ - . Σ +
and Σ- are made coincident with each other to make an open surface Σ .
We let V be the volume inside of S1, and outside of Σ + + Σ - .
We let ui be a solution of equation of equilibrium

cijpq u p ,qj = − ρ fi (11)

ui is continuous throughout V. On Σ + , ui = ui+ and on Σ - , ui = ui− ; ui+ − ui− = Δui .

Also, ui satisfies the boundary conditions, either ui=0 (rigid) or cijpqu p,q n j = 0 (free) on S1,
where n is a unit vector normal to S1 (positive outward).
Let vi be another solution of (10) and represent the displacement field due to a
body force ρ g (per unit volume). vi is continuous everywhere and satisfies the
boundary condition on S1 (either vi =0 (rigid) or cijpq v p ,q n j = 0 (free) ).

264
 cijpq v p, qj = − ρgi (12)

Multiplying (11) by vi and (12) by ui , subtracting and integrating over the volume V, we
obtain

∫V
(cijpq u p ,qj vi − cijpq v p ,qj ui )dV = − ∫ (ρ f i vi − ρ gi ui )dV
V
(13)

Note that

(cijpq u p , q vi − cijpq v p ,q ui ), j = (cijpq u p ,qj vi − cijpq v p , qj ui ) + (cijpq u p , q vi , j − cijpq v p , q ui , j )

Owing to the symmetry of cijpq, the second term of RHS vanishes. Hence,

∫ V
(cijpq u p , q vi − cijpq v p ,q ui ), j dV = − ∫ (ρ fi vi − ρ gi ui )dV
V
(14)

Applying the Gauss’ theorem to LHS,

265
 ∫S1 +Σ + +Σ −
(cijpq u p ,q vi n j − cijpq v p ,q ui n j )dS = − ∫ (ρ fi vi − ρ g i ui )dV
V
(15)

Due to the boundary conditions on S1, the integral on S1 vanishes. Since vi is continuous
on Σ

∫Σ + +Σ −
cijpq u p ,q vi n j dS = ∫ + cijpq u +p ,q vi n +j dS + ∫ − cijpq u −p ,q vi n −j dS
Σ Σ

= ∫ vi (cijpq u n + cijpq u n )dS

+
p ,q
+
j
−
p ,q
−
j
Σ

In equilibrium, the traction on the Σ + side cijpqu +p ,q n +j should be balanced by that on the Σ -

side, cijpqu −p,q n −j . Hence this integral vanishes.

On the other hand

266
 ∫
Σ + +Σ −
cijpq v p ,q ui n j dS = ∫ + cijpq v p , q ui+ (−ν j )dS + ∫ − cijpq v p , q ui− (+ν j )dS
Σ Σ

= − ∫ cijpq v p , q Δuiν j dS
Σ

Hence,

− ∫ cijpq v p ,q Δuiν j dS = ∫ (ρ fi vi − ρ gi ui )dV (16)

Σ V

We put ρ fi = 0, and, for gi, consider a unit force in k-th direction at point Q:

ρ gi = δ ik δ (Q)

Then (16) becomes

u k (Q) = ∫ Δuiτ ijkν j dS (17)

Σ

This is the Volterra's relation. τ ijk is the stress due to a unit force in k-th direction at Q,
and its unit should be stress/force = 1/area.

6. Moment Tensor

267
 Applying the Velterra's relation

u k ( y ) = ∫ Δuiτ ijkν j dS = ∫ Δui cijpq u kp , qν j dS

Σ Σ

to a small surface element ΔS over which ν , Δu and up,q

k
are constant, we obtain

u k ( y ) = cijpq u kp ,q Δuiν j ΔS (18)

= Mpq ukp,q (19)

where

M pq = cijpq Δuiν j ΔS (20)

Mpq is a symmetric tensor called the moment tensor. Then (19) can be written as

⎛ M 11 ⎞
⎜ ⎟
⎜ M 22 ⎟
⎛ u1 ⎞ ⎛ u1,1
1
u12,2 1
u3,3 1
u1,2 + u12,1 u1,3
1
+ u3,1
1
u2,3 + u3,2 ⎞
1 1

⎜ 2⎟ ⎜ 2 1 ⎟
⎜ M 33 ⎟
⎜ u ⎟ = ⎜ u1,1 u2,2 + u2,1 + u3,1 u12,3 + u3,2
2 2 2 2 2 2
u3,3 u1,2 u1,3 ⎟⎜ M ⎟ (21)
⎜ u 3 ⎟ ⎜ u1,1 1 ⎟⎜ 12 ⎟
⎝ ⎠ ⎝
3 3
u2,2 3
u3,3 3
u1,2 + u2,1
3 3
u1,3 + u3,1
3
u12,3 + u3,2 ⎠⎜ M ⎟
⎜⎜ M ⎟⎟
13

⎝ 23 ⎠

268
Using the explicit expressions for cijpq, (20) can be written as,

M 11 = ΔS [( λ + 2μ ) Δu1ν 1 + λΔu2ν 2 + λΔu3ν 3 ]

M 22 = ΔS[λΔu1ν 1 + ( λ + 2μ ) Δu2ν 2 + λΔu3ν 3 ]
M 33 = ΔS [λΔu1ν 1 + λΔu2ν 2 + ( λ + 2 μ ) Δu3ν 3 ]
(22)
M 12 = ΔS μ (Δu2ν 1 + Δu1ν 2 )
M 13 = ΔS μ (Δu3ν 1 + Δu1ν 3 )
M 23 = ΔS μ (Δu2ν 3 + Δu3ν 2 )

For shear dislocation, i.e., ν ⊥ Δu ,

M 11 = 2ΔS μΔu1ν 1
M 22 = 2ΔS μΔu2ν 2
M 33 = 2ΔS μΔu3ν 3
(23)
M 12 = ΔS μ (Δu2ν 1 + Δu1ν 2 )
M 13 = ΔS μ (Δu3ν 1 + Δu1ν 3 )
M 23 = ΔS μ (Δu2ν 3 + Δu3ν 2 )

which is often written as

M=μΔS ( Δu Tν + ν T Δu ) (23')

By taking the limit, ΔS => 0 ( ΔuiΔS is kept finite), we obtain a point dislocation
source that is represented by a moment tensor Mpq.
As a special case, consider a case where

269
 ν = (0,1,0), and Δu = (Δu1 ,0,0) .

Then, from (22)

M12 = ΔSμΔu1 and M pq = 0 ( p ≠ 1, q ≠ 2)

and, from (21)

u1 = (u1,2
1
+ u12,1 ) M 12
u 2 = (u1,2
2
+ u2,1
2
) M 12
u 3 = (u1,2
3
+ u2,1
3
) M 12

This is the case considered earlier.

Equation (21) can be used to determine the moment tensor from the observed
displacement uk, if the Somigliana tensor uik for the medium is known.
Suppose we observe the displacement l u k (l = 1,2,3,....., L) at L locations. Then,
we can compute l upk ,q by using the known Somigliana tensor. Then (21) can be written
as

( u ) = ( U )(M )
l
k
l pq (24)

where ( lU ) is the matrix on the RHS of (21) which can be computed by the known

Somigliana tensor. Combining (24) for l=1,2,3, ....L, we obtain a set of linear equations

270
with 6 unknowns Mpq and Lx3 data, and, for L>2 , the equations can be solved (unless the
problem is ill-posed) to obtain Mpq.

Ge 162 Problem #7 Moment Tensor and Fault Parameters

Consider a fault model as sketched in Figure 1. Only the foot-wall block is shown.
(ξ1, ξ2, ξ3) is the fault coordinate system (+ξ1 is in the direction of fault strike), and (x1,
x2, x3) is the geographical coordinate system (x1: North; x2: West; x3: Up).
The top surface of the block coincides with the surface of the Earth. D is the slip
vector of the hanging-wall block (not shown). δ is the dip angle of the fault plane
measured downward from the horizontal (i.e. from -ξ2 direction.). λ is the slip angle
(rake) measured counter-clockwise on the fault plane from the horizontal line as shown in
the figure (i.e. from +ξ1 direction.). ν is the unit vector normal to the fault plane. Let ΔS
be the fault area.
Place this fault model in a homogeneous elastic medium with the rigidity μ at the
origin of the Cartesian coordinates as shown in Figure 2. The strike of the fault, φ, is
measured clockwise on the free surface from x1 axis in Figure 2. We assume that ΔS is
small so that the fault can be considered a point source.

1) Determine the ξ1, ξ2, ξ3 components of the displacement vector D (magnitude D)

and the unit vector ν .

2) Determine the x1, x2, x3 components of the displacement vector D (magnitude D) and
the unit vector ν .

3) Determine the moment tensor elements Mij (i, j =1, 2, 3) for this fault model. Write
the result in terms of scalar moment M0=μDΔS, δ, λ, and φ. (The results are given in
5.1.4.) (Note that this is a shear dislocation. )

271
4) Determine the moment tensors for the Hector Mine earthquake, the Chi-Chi Taiwan
earthquake, and the Bhuj India earthquake. The fault parameters for these events are
given below ( δ (dip in degree), λ (rake in degree), φ f (strike in degree), M 0 (Seismic
Moment in N-m)).

1999, Hector Mine 85, 179, 336, 5.9x1019

1999, Chi-Chi, Taiwan 27, 82, 26, 4.1x1020
2001, Bhuj, India 50, 50, 65, 3.6x1020

Ge 162

5.1.4 Summary of Static Source Representation

272
1. Green's Function

Apply unit point force in k direction at ξ = (ξ1 , ξ 2 , ξ3 ) . Then, the i-th component
of the displacement at x = ( x1 , x2 , x3 ) is the Green' function Gik ( x , ξ ) .

In a homogeneous whole space, this is given by the Somigliana tensor, uik .

2. Single force at a point

The displacement for a single force F ( F 1 , F 2 , F 3 ) is then

ui = F k Gik (1)

3. Distributed force

The displacement due to a distributed force in V0 can be written as.

ui ( x ) = ∫ ρ f k (ξ )Gik ( x , ξ )dV (3)

V0

4. Force Couples

For a unit single couple on the x1 − x2 plane, the displacement is

uiSC = Gi1,2 (3)

For a double couple,

uiDC = Gi1,2 + Gi2,1 (4)

273
 In a homogeneous whole space, (3) and (4) can be also written as
ui
SC
= −Gi1,2 and uiDC = −(Gi1,2 + Gi2,1 ) .

If the force is in k = (k1 , k2 , k3 ) direction and the arm is in l = (l1 , l2 , l3 ) direction,

then

uiSC = Gip ,q k p lq (5)

This can be shown as follows. Let Gi( f ) be the i-th component of displacement
∂
due to a unit force in k direction, and be the derivative in l direction, then
∂sl

∂Gi( f ) ∂
uiSC = = lq (Gip k p ) = Gip , q k p lq (5)
∂sl ∂xq

Then, the displacement due to a unit double couple is

uiDC = Gip ,q (k p lq + kq l p ) (6)

For a source represented by force couples and dipoles with the strength (moment)
given by M kj ,

u i = M kj Gik , j (7)

5. Dislocation source

Referring to the figure below. the displacement due to an elastic dislocation Δu

on the surface Σ can be written as

274
 i

u i = ∫ τ ipq Δu pν q dS (8)
Σ

τ ipq = c pqkl Gki ,l (9)

u i = ∫ c pqkl Gki ,l Δu pν q dS = ∫ mkl Gki ,l dS (10)

Σ Σ

where

mkl = c pqkl Δu pν q (11)

In the limit of point dislocation (i.e., | Δu |→ ∞ , Σ ≡ ΔS → 0 and | Δu | ΔS → finite

275
 u i = M kl Gki ,l (12)

or, exchanging the source and the observation points,

ui = M kl Gik ,l (12')
where

M kl = c pqkl Δu pν q ΔS = M lk (13)

This is the same as the expression for the displacement field due to force couples (7).
Thus, a dislocation source is equivalent to force couples.

The explicit form of M lk is:

M 11 = ΔS ⎡⎣( λ + 2 μ ) Δu1ν 1 + λΔu2ν 2 + λΔu3ν 3 ⎤⎦

M 22 = ΔS ⎡⎣λΔu1ν 1 + ( λ + 2μ ) Δu2ν 2 + λΔu3ν 3 ⎤⎦
M 33 = ΔS ⎡⎣ λΔu1ν 1 + λΔu2ν 2 + ( λ + 2μ ) Δu3ν 3 ⎤⎦
(14)
M 12 = μΔS ( Δu1ν 2 + Δu2ν 1 )
M 13 = μΔS ( Δu1ν 3 + Δu3ν 1 )
M 23 = μΔS ( Δu2ν 3 + Δu3ν 2 )

In case of shear dislocation (i.e., Δu ⊥ ν ), then

M 11 = 2 μΔS Δu1ν 1
M 22 = 2μΔS Δu2ν 2
M 33 = 2 μΔS Δu3ν 3
(15)
M 12 = μΔS ( Δu1ν 2 + Δu1ν 2 )
M 13 = μΔS ( Δu1ν 3 + Δu3ν 1 )
M 23 = μΔS ( Δu2ν 3 + Δu3ν 2 )

276
or,

M= μΔS (Δu :ν ) (15')

The following figure illustrates the force couples represented by each element of a
moment tensor.

277
6. Fault parameters and Moment tensor elements.

The fault parameters (δ, λ, and φf ) defined in the figure below can be transformed
to moment tensor elements by ( μ DS is assumed to be 1),

M11 = −sin 2 φ f sin2δ sin λ − sin 2φ f cos λ sin δ

M22 = − cos2 φ f sin2δ sin λ + sin 2φ f cos λ sin δ
M33 = sin2δ sin λ
1 (7)
M12 = − sin 2φ f sin2δ sin λ − cos2φ f cos λ sin δ
2
M13 = cosφ f cosδ cos λ + sin φ f sin λ cos2δ
M23 = − sin φ f cosδ cos λ + cos φ f sin λ cos2δ

278
(The following part is not complete yet)
6. General Representation

Suppose that the medium is in equilibrium in the beginning (A). Then a failure
occurs in V0 (B), and displacement ui and stress σ ij are produced in the medium.

279
The equation for static equilibrium,

σ ij , j = 0 (16)

holds everywhere including V0 . In the region outside of V0 , ui and σ ij are related by

elastic constitutive relations; i.e., stress can be computed from ui . However, the elastic
constitutive relation does not hold in the failure zone V0 . Thus, σ ij cannot be computed
from ui . If we assume hypothetical elastic constants in V0 , then, we can compute the
stress in V0 from ui . We denote this stress by σ ijH (i.e., Hooke's stress). σ ijH is different
from σ ij . Then we can write (16) as

σ ijH − (σ ijH − σ ij ) = σ ijH − mij = 0 (17)

where

mij = (σ ijH − σ ij ) (18)

mij vanishes outside of V0 .

The form of equation (17) suggests that − mij , j can be considered as equivalent
body force. Thus, the displacement can be written as

280
 ui =
∫ ( −m
V0
kl ,l Gik )dV (19)

Using the Gauss theorem,

∫ ( m G ) dV = ∫ mkl ,l Gik dV + ∫ mkl Gik,l dV = ∫ mkl Gik nl dS = 0

k
kl i ,l
V0 V0 V0 S0

(S0 can be taken slightly outside of V0). Then, (19) can be written as

ui =
∫
V0
mkl Gik ,l dV (20)

For a point source,

∫ V0
mkl dV → M kl (21)

and

ui = M kl Gik ,l (22)

which is the same as (12').

mij defined by (18) can be considered as a general source representation, and is

called the seismic moment tensor.
5.1.5 Stress Relaxation

1. Description of the Problem

281
 Consider an elastic medium which is in equilibrium under stress σ ij (r ) . Let S be
an open surface defined in this medium of which the stress is σ ij(0) (r ) . Relax the stress
from σ ij(0) to σ ij(1) by creating a cut on S and letting the material on either side
of S to move. Let Δu be the resulting dislocation (displacement discontinuity). Δu is
not necessarily parallel to S. We want to determine Δu , and the difference in strain
energy ΔW between the two states, before and after the crack formation.

2. Qualitative Solution of the Problem

Let us consider a planar crack S under uniform shear stress σ 0 .

If the stress is relaxed on the crack surface to σ 1 (= const), there will be a
displacement offset Δu across the crack. In general, this is not uniform on S. Let D be
the average of Δu on S. If the representative dimension of the crack is L, then the
change in ε would be

Δε ≈ D / ( L / 2 )

282
This strain change is caused by the change in the stress Δσ = σ 0 − σ 1 , which we call
stress drop. Thus,

Δσ = μΔε = cμ D / ( L / 2 ) (1)

where c is a non-dimensional constant which depends upon the shape of the crack; it is
apparent from the above argument that c ~ 1.

The simplest case is a circular crack. If a is the radius of the crack, we can show
that L = 2a and c = 7π /16 ( λ = μ is assumed.).

Δσ = (7π /16) μ D / a (2)

The change in the strain energy ΔW is given by Sσ D . This is analogous to the change
in the strain energy of a spring (spring constant k) which is stretched from x0 to x1 . The
change in the strain energy is

ΔW = (1/ 2)k ( x02 − x12 ) = (1/ 2)k ( x0 + x1 )( x0 − x1 )

where (1/ 2)k ( x0 + x1 ) is the average force and ( x0 − x1 ) is the displacement.

Thus, for the circular crack

ΔW = Sσ D = (16 / 7 μ )a 3Δσσ (3)

where σ = (σ 0 + σ 1 ) / 2 is the average stress.

If we use the dislocation theory, this crack can be represented by a distribution of

double couples whose total moment M0 is

M 0 = μ SD = (16 / 7)a 3Δσ (4)

Next, we consider a crack as shown by the figure.

283
 σ0

σ0

The crack is a strip with a width of 2w and of infinite length. We consider a

uniform shear stress parallel to this strip. Because of the geometrical similarity between
this crack and a strike slip fault, we call this crack a strike-slip crack. In mechanics, this
is called Mode III crack. The argument similar to that made for the circular crack leads to

Δε ≈ D / w and Δσ = μΔε = cμ ( D / w) , c ≈ o(1)

we can prove that c = 2 / π . If the crack is not infinitely long but of finite length L, the
above relation is not exact. However, when L is very long, it is a good approximation.

For practical applications, we often take one half of the crack as shown by the
hatched portion in the figure. For this crack,

Δσ = (2 / π ) μ ( D / w)

ΔW = Lwσ D = (π / 2μ ) w2 LΔσσ (5)

M 0 = μ wLD = (π / 2) w2 LΔσ (6)

Another useful geometry is shown below.

284
 σ0

σ0

The crack extends in the direction perpendicular to the uniform shear stress. In
this case we can derive

4(λ + μ ) μ
Δσ = ( D / w)
π (λ + 2 μ )

We call this crack a dip-slip crack. In mechanics, this is called Mode II crack. For the
hatched portion,

π (λ + 2 μ ) 2
ΔW = Lwσ D = w LΔσσ
4(λ + μ ) μ
and

π (λ + 2 μ ) 2
M 0 = μ LwD = w L Δσ
4(λ + μ )

These relations are summarized in the following table.

Circular Strike Slip Dip Slip

(λ = μ ) (Mode III) (Mode II)
Srtess drop, (7π /16) μ D / a (2 / π ) μ ( D / w) 4(λ + μ ) μ
Δσ ( D / w)
π (λ + 2 μ )

285
Strain Energy, (16 / 7 μ )a 3 Δσσ (π / 2μ )w2 LΔσσ π (λ + 2 μ ) 2
w LΔσσ
ΔW = Sσ D 4(λ + μ ) μ
Moment, (16 / 7)a 3 Δσ (π / 2)w2 LΔσ π (λ + 2 μ ) 2
w LΔσ
M 0 = μ SD 4(λ + μ )

Relations between stress drop, strain energy change, offset, dimension, and moment for static
cracks. Dimensions of the fault are a radius, L length, w width; initial stress is σ 0 ; final stress is σ 1 ;
stress drop is Δσ = (σ 0 − σ 1 ) ; average stress is σ = 1/ 2(σ 0 + σ 1 ) ; average dislocation is D .

3. Elasto-Static Theory of Crack

Here, we consider only a strike-slip crack (Mode III). In this case, the
displacement field has a component only in the direction parallel to the crack. Therefore,
the problem becomes 2-dimensional.

Consider a crack extending in x3 of which the cross-section is elliptic. We will

later reduce this ellipse to a infinitesimally thin strip.

286
 Initially the crack is not formed. We introduce initial displacement field

U (0) = (0, 0, U 3(0) )

where U 3(0) = A0 x2 . Since the stress corresponding to this displacement is
∂U 3(0)
σ0 = μ = μ A0 = const , we have a uniform initial shear stress σ 0 = μ A0 . Let U be
∂x2
the displacement after the crack is formed. Because of the symmetry of the problem, U
has only x3 component. i.e.,

U = (0, 0, U 3 ) (7)

where U 3 does not depend on x3 , and

divU = 0 (8)

Thus, the equation of equilibrium

(λ + 2μ ) graddivU − μ curlcurlU = 0 (9)

gives

curlcurlU = 0 (10)

If we put
A = curlU (11)

then, from (10),

curlA = 0 (12)

Also from (11),

divA = 0 (13)

287
From (12) we can put

A = − gradφ (14)

where φ is a scalar potential. Substituting (14) into (13)

∇ 2φ = 0 (15)

From (11) and (14),

⎛ ∂U 3 ∂U ⎞ ⎛ ∂φ ∂φ ∂φ ⎞
⎜ , − 3 , 0⎟ = ⎜ − ,− ,− ⎟ (16)
⎝ ∂x2 ∂x1 ⎠ ⎝ ∂x1 ∂x2 ∂x3 ⎠

Let us consider a case where the stress on the crack is completely released,
i.e., τ 32 = 0 on S.

∂U 3
μ =0 on S
∂x2

By (16), this condition can be restated as

∂φ
=0 on S
∂x1

∂φ
Since = 0 (see (16)), φ = φ ( x2 ) . In the limit of a very thin crack x2 = 0 on S; thus,
∂x3
φ = const on S. Since φ is a potential, we can put, for the boundary condition

φ =0, on S (17)

288
Thus, our problem is now reduced to a problem of finding a solution of (15), with a
boundary condition (17). Note that this problem is equivalent to the well-known electro-
static problem of finding an electro-static field due to a perfect conductor placed in
vacuum.

------------------------------------------------------------------------------------------------------------
-
To solve this, we introduce elliptic coordinates (ξ1 , ξ 2 ) by

x1 = a cosh ξ1 cos ξ 2 and x2 = a sinh ξ1 sin ξ 2 (18)

Then
w = x1 + ix2 = a cosh(ξ1 + iξ 2 ) = a cosh(ζ ) (19)

where ζ = ξ1 + iξ 2 . ξ1 = const gives an ellipse with a major axis a cosh ξ1 , and a minor
axis a sinh ξ1 . Thus, the elliptic crack is given by ξ1 = const, and the limiting case ξ1 =0
represents an infinitesimally thin crack of width 2a.
We write

φ = φ0 + φ ′ (20)

where φ0 represents the initial field, and φ ′ , the perturbation due to the crack. Since the
initial field is U 3(0) = A0 x2 ,

φ0 = − A0 x1 = − A0 a cosh ξ1 cos ξ 2 (21)

φ = − A0 a cosh ξ1 cos ξ 2 + φ ′ (22)

Since ∇ 2φ = 0 , and ∇ 2φ0 = 0 (from (21)),

∇ 2φ ′ = 0 (23)

Since φ = 0 on S ( ξ1 = const), φ ′ must have the form

φ ′ = F (ξ1 ) cos ξ 2 (24)

289
From ∇ 2φ ′ = 0 , we have (use Laplacian for the elliptic coordinates),

d 2F
=F
dξ12

∴ F (ξ1 ) = A exp(−ξ1 ) + B exp(+ξ1 ) (25)

Because F (ξ1 ) must be finite for ξ1 → ∞ , we take only the first term

φ = ( A exp(−ξ1 ) − A0 a cosh ξ1 ) cos ξ 2 (26)

From the condition that φ = 0 at ξ1 = 0 ,

A = A0 a
1
φ = − A0 a sinh ξ1 cos ξ 2 = − A0 a(sinh ζ + (sinh ζ )* )
2 (27)
= − A0 a Re(sinh ζ ) = − A0 a Re(cosh ζ − 1) = − A0 Re( w − a )
2 1/ 2 2 2 1/ 2

Since, from (16)

∂φ ∂ (−U 3 ) ∂φ ∂ (−U 3 )
= and =−
∂x1 ∂x2 ∂x2 ∂x1

From the Cauchy-Riemann relation, if φ is the real part of analytic function

− A0 ( w2 − a 2 )1/ 2 of w = x1 + ix2 , as shown above, then −U 3 is the imaginary part of
− A0 ( w2 − a 2 )1/ 2 . Thus

σ0
U 3 = A0 Im( w2 − a 2 )1/ 2 = Im( w2 − a 2 )1/ 2 (28)
μ

290
is the solution of the problem.

------------------------------------------------------------------------------------------------------------
-

With the solution,

σ0
U 3 = A0 Im( w2 − a 2 )1/ 2 = Im( w2 − a 2 )1/ 2 ( w = x1 + ix2 ) (28)
μ

first, consider the displacement field on x1 − x3 plane ( plane of the crack ).

In this case x2 = 0 , and w = x1 .

σ0
U3 = Im( x12 − a 2 )1/ 2
μ

U3 = 0 , x1 > a (29)
∴
σ 0 2 2 1/ 2
U3 = (a − x1 ) , x1 ≤ a
μ

This is shown in the figure.

σ0a/μ

The maximum displacement U3max is

291
 ⎛σ ⎞
U 3max = ⎜ 0 ⎟ a
⎝ μ ⎠

and the average displacement is

1 a σ0 2
(a − x12 )1/ 2 dx1 = (πσ 0 / 4μ )a
2a ∫− a μ
U3 = (30)

The average dislocation is

D = 2U 3 = (πσ 0 / 2 μ ) a (31)

For a partial stress drop in which stress drops from σ 0 to σ 1 , σ 0 in (31) should simply
be replaced by Δσ = σ 0 − σ 1 . If we replace a in (31) by width w , (31) leads to the
relation given in the table.

Next, let us consider the displacement on x2 − x3 plane. Putting x1 = 0 , we have,

from (28)

σ0 ⎛σ ⎞
U3 = Im(− x22 − a 2 )1/ 2 = ± ⎜ 0 ⎟ ( x22 + a 2 )1/ 2 (32)
μ ⎝ μ ⎠

+ sign is taken for x2 > 0, and - sign for x2 < 0.

292
 σ0x2/μ

σ0a/μ

-σ0a/μ

⎛σ ⎞
Since the initial field is ⎜ 0 ⎟ x2 , the actual displacement caused by the crack is
⎝ μ ⎠
⎛σ ⎞
obtained by subtracting ⎜ 0 ⎟ x2 from (32):
⎝ μ ⎠

⎛σ ⎞ ⎛σ ⎞
U 3 = U 3 − ⎜ 0 ⎟ x2 = ⎜ 0 ⎟ ⎡⎣ ±( x22 + a 2 )1/ 2 − x2 ⎤⎦ (33)
⎝ μ ⎠ ⎝ μ ⎠

σ0a/μ

-σ0a/μ

The x2 coordinate where the displacement decays to half is given by

X 1/ 2 = ( 3 / 4 ) a or a = ( 4 / 3) X 1/ 2 (34)

293
The three-dimensional displacement is shown in the figure above.

If we use this model, we can estimate the depth of the fault from the observed
decay rate of the displacement field by using (34).

In the following examples, U 3 ( x1 = 0) is plotted as a function of x2 . It is readily

seen that X 1/ 2 = 10 km for Tango, 7 km for Tottori and 5 km for the San Francisco
earthquakes. From these values, w ~ 13 km, 9 km and 7 km are obtained for these
earthquakes. However, because of the incompleteness of the data set for the San
Francisco earthquake, the estimate of w for this event is very uncertain.

294
295
4. Stress

Next, let us consider the change in the stress. Since U1 = 0 , U 2 = 0 , U 3 ( x1 , x2 ) ,

∂U 3 ∂U 3
only nonzero components of stress are τ 32 = μ and τ 31 = μ .
∂x2 ∂x1
Using (28), we have on x2 − x3 plane (i.e., x1 = 0)

1 1
τ 31 = 0 , and τ 32 = σ 0 x2 =σ0 (35)
(x + a )
2
2
2 1/ 2
(1 + (a / x2 ) 2 )1/ 2

296
----------------------------------------------------------------------------------------------------------

Note: (28) gives the solution as an imaginary part of an analytic function. The
derivatives of U 3 with respect to x1 and x2 can be obtained using the following relation.
Consider an analytic function Ψ ( w) = P( x1 , x2 ) + iQ( x1 , x2 ) of w = x1 + ix2 . Then

∂Ψ ∂P ∂Q ∂Ψ ∂P ∂Q ∂P ∂Q
= +i , and = +i = −i +
∂w ∂x1 ∂x1 ∂w ∂ (ix2 ) ∂ (ix2 ) ∂x2 ∂x2

∂Q ∂Ψ ∂Q ∂Ψ
Hence, is the imaginary part of , and is the real part of .
∂x1 ∂w ∂x2 ∂w

----------------------------------------------------------------------------------------------------------

Since τ 11 ≡ 0 and τ 12 ≡ 0 , τ 31 = 0 obtained above means that the x2 − x3 plane can

be considered as a stress free surface. Thus, it is convenient to take this plane as a free
surface in modeling a shallow fault. In this case, we consider one half of the crack as
shown by the hatched part as the actual fault plane.

The sress τ 32 changes as a function of x2 on x2 − x3 plane as shown in the following

figure.

297
 σ0

On x1 − x3 plane ( x2 = 0), we have

τ 32 = 0 , | x1 |< a
(36)
1
τ 32 = σ 0 x1 , | x1 |> a
( x − a 2 )1/ 2
2
1

As shown in the figure τ 32 has a singularity at x1 = a of the order of ( x1 − a) −1/ 2 (i.e.,

square-root singularity). This singularity is typical of this type of thin cracks. Near,
x1 = a , i.e., at x1 = a + ε (ε << a) , and

1 a
τ 32 = σ 0 x1 ≈ σ 0 ε −1/ 2 (36')
( x1 − a) ( x1 + a )
1/ 2 1/ 2
2

298
5.2 Elasto-Dynamic Source ---Summary of 5.2.1. and 5.2.2. ---

5.2.1 Displacement Field due to a Point Force and Force Couples

(Time-dependent case)

The derivation is more complex than that for the static case, but is similar. The
displacement field consists of near-field and far-field. The field consists of
"compressional" and "shear", propagating at P and S wave velocities, respectively. The
near-field term yields displacement between P and S waves, as shown in the figure on
page 5.

The following relation holds.

uij (t ;τ ) = uij (−τ ; −t )

5.2.2 Force Couples

The derivation is essentially similar to that for the static case. In general, the
displacement field consists of near-field and far-field.

Far field:

299
 M ′(τ )
u≈
4πρ v 3 r

Near field

M (τ )
u≈
4πρ v 2 r 2

The displacement is schematically shown on page 7

.

The most important relation is given by equation 17', which gives the radiation
pattern and the amplitude of far-field P and S waves. The radiation pattern is exactly the
same as that for the static field.
For a shear faulting, S wave is much larger than P wave.

3
| max ur | ⎛ vs ⎞ 1
=⎜ ⎟ ≈
| max uφ | ⎜⎝ v p ⎟⎠ 5

5.2 Elasto-Dynamic Source

5.2.1 Displacement Field due to a Point Force and Force Couples

(Time-dependent case)

Single Force

The method is essentially the same as that used for static problem.

The equation of motion is:

300
 ∂ 2u
ρ = ρ f + (λ + 2μ ) graddivu − μ curlcurlu (1)
∂t 2

A point force is given by:

⎛ a ⎞
ρ f = Faδ (r )h(t ) = − Fh(t )∇ 2 ⎜ ⎟
⎝ 4π r ⎠
(2)
⎡ ⎛ a ⎞ ⎛ a ⎞⎤
= − Fh(t ) ⎢ graddiv ⎜ ⎟ − μ curlcurl ⎜ ⎟⎥
⎣ ⎝ 4π r ⎠ ⎝ 4π r ⎠ ⎦

We seek a solution in the form

u = graddivAp − μ curlcurlAs (3)

Then, we obtain,

Fh(t ) ∂ 2 Ap
(λ + 2μ )∇ 2 Ap = a+ρ (4)
4π r ∂t 2

Fh(t ) ∂2 A
μ∇ 2 As = a + ρ 2s (5)
4π r ∂t

Putting

Ap = Ap a and As = As a (6)

1 ∂ Ap
2
Fh(t )
∇ 2 Ap = a+ 2 (7)
4π (λ + 2μ )r v p ∂t 2

Fh(t ) 1 ∂ 2 As
∇ 2 As = a+ 2 (8)
4πμ r vs ∂t 2

where

301
 v p = (λ + 2 μ ) / ρ and vs = μ / ρ (9)

are P- and S-wave velocities, respectively.

Note that, the solution of an inhomogeneous scalar wave equation,

1 ∂ 2φ ( x, y, z , t )
∇ φ ( x, y , z , t ) − 2
2
= − g ( x, y , z , t ) (10)
c ∂t 2

is given by,

1 g (ξ ,η , ζ , t ± R / c)
φ ( x, y , z , t ) =
4π ∫
V R
dξ dη dζ (11)

where R 2 = ( x − ξ ) 2 + ( y − η ) 2 + ( z − ζ ) 2

Using this relation, and carrying out the integration,

we obtain

⎡1 ∞
1 ∞
⎤
∫ ∫
F
Ap (r , t ) = ⎢r h(t ± r / v p ± ν )ν dν − h(t ± ν )ν dν ⎥ (12)
4πρ ⎣ 0 r 0 ⎦

302
 ⎡1 ∞
1 ∞
⎤
∫ ∫
F
As (r , t ) = ⎢r h(t ± r / vs ± ν )ν dν − h(t ± ν )ν dν ⎥ (13)
4πρ ⎣ 0 r 0 ⎦

Then, for a unit force (F=1),

1 ⎪⎧ ∂ 2 ⎡1 ∞
⎤ 1 h(t ± r / vs ) ⎪⎫
uij = ⎨
4πρ ⎩⎪ ∂xi ∂x j ⎢r
⎣ ∫0
(h(t ± r / v p ± ν ) − h(t ± r / vs + ν ))ν dv ⎥ + 2
⎦ vs r
δ ij ⎬
⎭⎪

(14)

This solution gives both outgoing and converging waves. For outgoing waves, we
take the minus sign, and rewriting it

∂ 2 r −1
1 r / vs
1 1 ⎛ xi x j ⎞
∫
xi x j
ui =
j
ν h(t −ν )dν + h (t − r / v ) + ⎜ δ − ⎟ h(t − r / vs )
4πρ ∂xi ∂x j 4πρ v 2p r r 2 4πρ vs2 r ⎝ r2 ⎠
p ij
r / vp

(15)

uij is called Love tensor.

It is easy to show that if a δ function force is applied at Q ( y1, y2, y3 ) at time τ ,

then the displacement at P ( x1, x2, x3 ) at time t is given by,

1 ∂ 2 r −1 r / vs
1 1 ⎛ xx ⎞
∫
xi x j
uij = νδ (t − τ −ν )dν + δ (t − τ − r / v p ) + 2 ⎜ ij
δ − i 2 j ⎟ δ (t − τ − r / vs )
4πρ ∂xi ∂x j r / vp 4πρ v p r r
2 2
4πρ vs r ⎝ r ⎠

(15')

303
where

r 2 = ( xi − yi ) 2

The 2nd and the 3rd terms decay as 1/r and represent propagating P and S waves,
respectively. These terms are called far-field. The first term decays as 1/ r 2 , and
represents displacement between P and S wave arrivals. This term is called near-field.

The time history is schematically shown in the figure below.

Reciprocity

It is evident that

uij (t ;τ ) = u ij (t ;τ ) .

However, since t and τ appear as t − τ ,

uij (t ;τ ) ≠ uij (τ ; t )

but

uij (t ;τ ) = uij (−τ ; −t )

5.2.2 Force Couples

304
 The displacement field for single couples and double couples can be
obtained by differentiating (15) with respect to appropriate coordinates.

Qualitative Derivation

For a point force F(t) placed at Q, the displacement at P due to the 2nd or 3rd
term is

u ∝ F (τ ) / r

where τ = t − r / v (v: wave velocity).

For a force at Q’ (short distance a

away from Q),

F (τ − a / v) /(r + a)

Then for a couple

u ∝ F (τ ) / r − F (τ − a / v) /( r + a ) = F (τ ) / r − (1/ r )(1 − a / r )( F (τ ) − (a / v) F ′(τ ))

= aF (τ ) / r 2 + aF ′(τ ) / rv = M (τ ) / r 2 + M ′(τ ) / rv

305
The first term decays as 1/ r 2 ; hence, at large r, we have M ′(τ ) / rv which is called far
field. At short distance, the first term, M (τ ) / r 2 , dominates, which is called near field.
1
Recovering the term , we obtain
4ππ v 2

M (τ ) M ′(τ )
u≈ + (16)
4πρ v r
2 2
4πρ v 3 r

The following figure illustrates the contributions of near-field and far-field

displacements.

The actual displacement field is more complicated because we need to include the
contribituion of the 1st term in (15') too. However, at far field, the second term in (16)
essentially gives the displacement field which propagates at either P or S wave.

306
 The figure below shows the near-field (+far-field) displacement observed for the
1999 Chi-Chi, Taiwan earthquake.

Elasto-dynamic theory

Here, we consider only the far-field term. If we place a double couple with a
moment M (t ) on x1 − x2 plane as shown in the figure,

307
we have at large r (far-field),

2
1 ⎛v ⎞
ui = 2 ⎜ s ⎟ γ iγ 1γ 2 M ′(t − r / v p )
4πμ v p r ⎜⎝ v p ⎟⎠
(17)
1
+ ( −2γ iγ 1γ 2 + γ 2δ i1 + γ 1δ i 2 ) M ′(t − r / vs )
4πμ vs r

where γ i = xi / r .

If we take the polar coodinates (r , θ , φ ) ,

308
 γ 1 = sin θ cos φ , γ 2 = sin θ sin φ , γ 3 = cosθ ,

Then

⎛ ur ⎞ ⎛ sin 2 θ sin 2φ ⎞
⎜ ⎟ 1 ⎜ ⎟
⎜ uθ ⎟ = 4πρ rv 3 M ′(t − r / v p ) ⎜ 0 ⎟
⎜ uφ ⎟ p ⎜ 0 ⎟
⎝ ⎠ ⎝ ⎠ (17')
⎛ 0 ⎞
1 ⎜ ⎟
+ M ′(t − r / vs ) ⎜ 1/ 2sin 2θ sin 2φ ⎟
4πρ rvs
3
⎜ sin θ cos 2φ ⎟
⎝ ⎠

Thr first term on RHS represents P wave and the second term, S wave.
On x1 − x2 plane, θ = π / 2 and

1
ur = M ′(t − r / v p ) sin 2φ
4πρ rv 3p

uθ = 0 (18)

309
 1
uφ = M ′(t − r / vs ) cos 2φ
4πρ rvs3

The amplitude distribution is sketched in the following figures.

Note that the pattern (radiation pattern) is identical to that for the static case. The
amplitude ratio

3
| max ur | ⎛ vs ⎞ 1
=⎜ ⎟ ≈
| max uφ | ⎜⎝ v p ⎟⎠ 5

Also the time history of the displacement is given by the time derivative of the moment
time function. It is also important that the amplitude decays as 1/ r in contrast to 1/ r 2
for the static field.

5.2.3 Radiation from a Double-Couple Source

The attached four figures show the waveforms computed for the following:

310
Fig. 1 and Fig. 2 show the responses in a whole space (α=6 km/sec, β=3.5 km/sec, ρ=2.6
g/cm3). The "depth" is 10 km. This means that the displacements are computed at a
level 10 km above the source. The mechanism is a N-S striking vertical strike slip with
and the seismic moment is 1027 dyne-cm. The source time function is a triangle with a
rise-time of 0.2 sec and fall-off time of 0.2 sec.

Fig. 1.
Station azimuth=90°. The transverse components are shown at 20 km distance
intervals. P-wave arrivals are aligned.

311
312
Fig. 2.
Station azimuth=45°. The vertical components are shown at 20 km distance
intervals. P-wave arrivals are aligned.

313
314
Fig. 3. and Fig. 4 show the responses in a southern California crustal model given below.

Thickness (km) α (km/sec) β (km/sec) ρ (g/cm3)

5.5 5.5 3.18 2.4

10.5 6.3 3.64 2.67
19.0 6.7 3.87 2.8
7.8 4.5 3.0

The depth is 11 km. The mechanism is a vertical strike slip and the seismic
moment is 1027 dyne-cm. The source time function is a 1/2 cycle cosine with half-width
of 0.2 sec.

315
Fig. 3. Station azimuth = 90°. The transverse components are shown at 20 km distance
intervals.

316
317
Fig. 4.
Station azimuth =45°. The vertical components are shown at 20 km distance
intervals.

318
319
5.2.3 Elastic Dislocation ---- Summary ----

The equivalence between an elastic dislocation and a double couple shown for the
static problem holds for the dynamic case, too. A point dislocation given by D(t ) over
an infinitesimally small area is equivalent to a double couple with moment

M (t ) = μ D(t ) S

Moment tensor, M kl (ξ , t ) , can be defined in a similar manner.

The displacement due to a point double couple is given by

ui ( x , t ) = M kl (ξ , t ) * Gik ,l ( x , ξ , t )

+∞
where * denotes convolution (i.e., h(t ) = f (t ) * g (t ) means h(t ) = ∫ f (t − τ ) g (τ )dτ ).
−∞
In the frequency domain,

uˆi ( x , ω ) = Mˆ kl (ξ , ω )Gˆ ik ,l ( x , ξ , ω )

This is formally equivalent to the expression we obtained for the static case. In practice,
both the time-domain and the frequency-domain formulations are used for inversion. The
time-domain representation is most commonly used in body-wave inversion, and the
frequency-domain representation is more commonly used in inversion studies of normal
modes.

320
5.2.3 Elastic Dislocation

1. Equivalence Between Double Couple and Dislocations

The equivalence between elastic dislocation and a double couple we showed for
the static case holds for the dynamic case. In particular, the displacement field due to a
point elastic dislocation D on S, can be computed with a double couple with a moment of
M 0 = μ DS . For a finite source, we use a distribution of double couples.

To prove this, we use the Volterra's theory (next section):

+∞
u k ( y, t ) =
∫ dτ
∫ Δu ( x,τ )τ ν dS (1)
k
i ij j
−∞ Σ

where

τ ijk = λδ ij ulk,l (−τ ; −t ) + μ (uik, j (−τ ; −t ) + u kj ,i (−τ ; −t ))

= λδ ij ulk,l (t ;τ ) + μ (uik, j (t ;τ ) + u kj ,i (t ;τ ))

(by reciprocity). τ ijk is the i-j component of stress at time −τ on Σ

due to a δ function force applied in k direction at Q at time -t.

321
 Thus, for a small planar dislocation d Σ at P which is perpendicular to the x2
axis, and on which u1 , has discontinuity, Δu1 (τ )

+∞
u k ( y, t ) =
∫
−∞
Δu1 (τ )τ 12k d Σdτ (2)
where

τ 12k = μ (u1,2
k
(t ;τ ) + u2,1
k
(t ;τ )) = − μ (u1k ,2 (t ;τ ) + u2k ,1 (t ;τ ))

322
It is easy to see that τ 12k / μ is the k th component of displacement at Q at time t due to a
δ function double couple applied at τ on x1 − x2 plane.

Then it is clear from (2) that this dislocation is equivalent to a double couple of
moment μ d ΣΔu1 (τ ) .

2. Volterra's Dislocation Theory (Dynamic Case)

We use the same geometry and notation as those used for the static case.

The equations of motion for ui and vi are

cijpq u p ,qj ( x , t ) − ρ ui ( x , t ) = − ρ fi ( x , t ) (3)

cijpq v p , qj ( x , t ) − ρ vi ( x , t ) = − ρ gi ( x , t ) (4)

fi = gi = 0 and ui = vi = 0 for t< − T (5)

Define

v p ( x , t ) = v p ( x , −t ) (6)

gi ( x , t ) = gi ( x , −t ) (7)

then

gi ( x , t ) = v p ( x , t ) = 0 for t>T (8)

For v p and gi , we have,

cijpq v p , qj ( x , t ) − ρ vi ( x , t ) = − ρ gi ( x , t ) (9)

323
Forming (3) x vi - (9) x ui and integrating it with t and V

+∞

∫ dt ∫ ((3)xv − (9)xu )dV

−∞ V
i i (10)

Using the symmetry relations for cijpq and the relation

∂
(vi ui − ui vi ) = vi ui − ui vi
∂t

We have from (10)

+∞ +∞
∂
∫ ∫ −∞
dt
S1 +Σ + +Σ −
dS ⎡⎣ cijpq (vi u p ,q − ui v p ,q )n j ⎤⎦ −
∫
−∞
dt
∂t
(vi ui − ui vi )
(11)
+∞
=
∫ dt ∫ dV ρ (u g −vf )
−∞ V
i i i

The second term of LHS vanishes because

+∞
∂
∫
+∞
(vi ui − ui vi ) = (vi ui − ui vi ) −∞
dt
−∞ ∂t
= vi (+∞)ui (+∞) − ui (+∞)v (+∞ ) − vi (−∞ )ui (−∞ ) + ui (−∞ )v (−∞ ) = 0

(owing to (5) and (8) ).

As in the static case, we introduce

fi = 0 , and ρ gi ( x , t ) = δ kiδ ( x1 − y1 )δ ( x2 − y2 )δ ( x3 − y3 )δ (t + s )
then,

ρ gi ( x , t ) = δ kiδ ( x1 − y1 )δ ( x2 − y2 )δ ( x3 − y3 )δ (−t + s ) (12)

324
Substituting these into (11),

+∞
u k ( y, s) =
∫ dt ∫ Δu ( x, t )τ ν dS (13)
k
i ij j
−∞ Σ

where τ ijk = cijpq v p ,q ( x , t ) is for ρ gi ( x , t ) given by (12). From the definition, τ ijk is the ij
component of stress at time -t due to a δ function force applied in k direction at Q and at
time -s.

Changing s to t and t to τ we have

+∞
u k ( y, t ) =
∫ dt ∫ Δu ( x,τ )τ ν dS (14)
k
i ij j
−∞ Σ

This is the Volterra's relation used in (1).

3. Seismic Moment Tensor

In (14), the more explicit expression for τ ijk is

τ ij (−τ ; −t ) = cijpq u kp.q ( x , −τ ; −t ) = cijpq u kp.q ( x , t − τ )

Substituting this in (14),

+∞
u k ( y, t ) =
∫
−∞
dτ
∫c
Σ
ijpq u kp.q ( x , t − τ )Δui ( x ,τ )ν j dS (15)

For a point source ( dS = d Σ → 0 , Δui → ∞ , with Δui d Σ = constant),

325
 +∞
u k ( y, t ) =
∫
−∞
dτ cijpq u kp.q ( x , t − τ )Δui ( x ,τ )ν j d Σ (16)

Introducing M pq by

M pq ( x ,τ ) = cijpq Δui ( x ,τ )ν j d Σ (17)

we obtain,

+∞
u k ( y, t ) =
∫
−∞
M pq ( x ,τ )u kp.q ( x , t − τ )dτ (18)

M pq is the seismic moment tensor.

The above integral is convolution of M pq and u kp.q . Hence taking the Fourier transform of
this, we obtain,

uˆ k ( y , ω ) = Mˆ pq ( x , ω )uˆ kp.q ( x , ω ) (19)

where uˆ k ( y, ω ) etc are the Fourier transform of u k ( y, t ) etc, respectively.

5.2.4 Stress Relaxation Model and Cracks (Dynamic Case)

1. Description of the Problem

Consider an elastic medium which is in equilibrium under stress. Consider an

open surface S. For simplicity, assume that the shear stress on this surface is
uniform and σ 0 . At time t = 0 relax the stress on this surface from σ 0 to σ 1 . Then
the static stress drop is Δσ s = σ 0 − σ 1 .

326
 As discussed earlier, the resulting average dislocation D and Δσ are related by

Δσ s = cμ D / a (1)

where a is the representative dimension of the crack and c is a non-dimensional "shape

factor" which is of the order of 1. Thus, if we plot D as a function of time, it would look
like the curve shown below.

For t<0 D =0. At t = 0 when the stress is relaxed the side of the crack starts
moving. After time τ , the medium is in another equilibrium state in which dislocation

327
 D = aΔσ s / cμ
is produced.
The behavior between t = 0 and t = τ depends on how stress is relaxed on S. If
stress relaxation takes place smoothly, the change in D may be also relatively smooth.
The time constant τ depends on σ 0 and Δσ . To solve this problem rigorously using
elasto-dynamics is difficult. In most cases no analytic solution is available, and we have
to resort to numerical methods.

2. Qualitative Analysis of the Problem

Infinite Instantaneous Crack

First let us consider the simplest case. Consider an infinite homogeneous elastic
medium under uniform shear stress σ 0 . At time t = 0, relax this stress over an infinite
plane S (parallel to the applied shear stress ) instantaneously. In order to see the
displacement for t ≥ 0 this problem can be replaced by the following problem.

Consider an homogeneous half space bounded by a plane S.

Apply a uniform shear stress σ 0 at t = 0 over the entire surface of S instantaneously.

This is equivalent to relaxing the uniform shear stress σ 0 over S in an infinite medium.
Then at time t, the point P on S will be moved to P'. Let this displacement be u(t).

328
 The disturbance applied to the surface propagates into the medium with shear
velocity β . At time t, it propagates as far as β t . Thus, the instantaneous strain is

ε = u (t ) / β t

Since this is caused by the applied shear stress σ 0 ,

σ 0 = με = μ u (t ) / β t
∴
u (t ) = (σ 0 / μ ) β t (2)

This gives the displacement for infinite instantaneous crack. From (2)

u (t ) = (σ 0 / μ ) β = constant (3)

Note that the particle velocity is proportional to the initial stress, σ 0 .

This problem can be solved analytically.

The equation of motion for x>0 is

329
 ∂ 2u 1 ∂ 2u
= (4)
∂x 2 β 2 ∂t 2

The solution is u ( z ) = F ( z ) where z = x − β t .

The boundary condition at x=0 is

∂u
μ +σ0 = 0 , t>0
∂x

∴ μ F ′( z ) = −σ 0 , F ( z ) = −σ 0 z / μ = −(σ 0 / μ )( x − β t )

∴ u = (σ 0 / μ )( β t − x) , t > x/β

which gives (2).

3. Finite Instantaneous Crack

Next let us consider a strike-slip crack shown in the figure.

330
At t = 0, uniform shear stress σ 0 is applied in x3 direction instantaneously over a surface
( x2 = 0 , | x1 |≤ a ). Let us consider the displacement at point O, the middle point of the
crack. For small t (before the effect of the edge reaches this point), the displacement of
this point should be the same as for infinite crack. Thus, the initial velocity should be
given by σ 0 β / μ . The effect of the edge arrives at this point at time t = a / β , and
eventually the motion stops when a final equilibrium state is achieved. This is given by
the solution of the static problem, i.e., u = σ 0 a / μ . Hence, the displacement time
function may be schematically given by the following figure.

331
 One useful functional form which approximates this function is

u (t ) = (σ 0 a / μ )(1 − exp(− β t / a )) (5)

This problem can be solved numerically. The equation of motion in the medium
is

1 ∂ 2u ∂ 2u
= (6)
β 2 ∂t 2 ∂x12

and the boundary condition is

∂u
μ +σ0 = 0 ( x2 = 0 , | x1 |≤ a , t>0) (7)
∂x2

The numerical solution is shown in the figure below and the displacement at the middle
point is compared with the approximate solution given above. Note that the numerical
solution shows a slight overshoot.

(Burridge, R., Phil. Trans. Roy. Soc. London, 265, 353-381, 1969)

332
4. Finite Propagating Crack

Let us consider the same geometry as before. Instead of applying the stress
instantaneously, we apply a propagating stress,

σ = σ 0 H (t ± x1 / V ) , | x1 |≤ a (8)

where H(t) is the Heaviside step function and V is the rupture velocity. V is usually
slightly smaller than β . In this case, the final value of u is the same as before.
However, the point O starts feeling the edge effect even in the beginning, because at t =
0, the two edges are very close to this point. Thus, the motion is decelerated compared
with the previous example. The effect of the edge | x1 |= a will reach the point O at time
a (1/ β + 1/ V ) ≈ 2a / β after the stress application. Thus, the average velocity would be

σ 0a / μ
u≈ ≈ σ 0 β / 2μ (9)
2a / β

The schematic displacement time function may look like the one shown in the figure.

333
A useful functional form would be

u (t ) = (σ 0 a / μ )(1 − exp(− β t / 2a )) (10)

A result of numerical computation is shown in figure below (this is

computed for a slightly different geometry). Note that, in this case there is no overshoot
and that the agreement with (10) is very good.

(Hanson, M. E., A. R. Sanford, and R. J. Shaffer, J. Geophys. Res., 76,

3375-3383, 1971)

334
 The above description, however, is very simplified. In a real propagating crack,
the initial velocity could be faster than σ 0 β / 2 μ due to the stress concentration near the
crack tip. Equation (10) should be considered only approximate.

5. Comparison with the data

Regardless of all the details, the particle-motion velocity is governed by the

relation,

u ≈ σ 0β / μ

If σ 0 =100 bar, β =3 km/sec, and μ =3x1011 dyne/cm2, then

u ≈ 1 m/sec

The observation of u is difficult because of the complexity of faulting and the

propagation effects, but the ground-motion velocity observed very close to a fault can be
used as a good proxy of u at least approximately. The following figure shows several
examples of observed ground-motion velocity very close to the source. This comparison
suggests that the above estimates are approximately correct, i.e., the magnitude of the
driving stress is of the order of about 100 bar. However, the result from the 1999 Chi-
Chi, Taiwan, earthquake stands out, suggesting that ground-motion velocity significantly
larger than 1 m/sec can occur under certain conditions.

335
6. Frictional Stress Release

In actual earthquake faulting, the stress on the fault plane is released against
frictional stress σ f opposing the motion.

336
 It is also possible that, at a certain time, say τ 0 , the fault motion slows down due
to some locking mechanism. In this case, the time history of the stress drop on the fault
surface may be schematically given by the curve shown in the figure above.

Actual stress release pattern can be very complex as shown by figure (b) below.

(a) (b)

Since the details of the stress release are presently unknown, it is not meaningful
to consider overly detailed models.

If the stress release is simple as shown on the left, then the fault motion

337
is driven by σ 0 e = σ 0 − σ f ( σ f = σ 1 ) which is called the effective (tectonic) stress.
The particle velocity of the side of the crack (fault) will be given by σ 0 e β / μ or
σ 0 e β / 2μ depending upon whether the stress application is instantaneous or not
(or in general cσ 0 e β / μ , c=constant).

Thus, if we can determine the approximate time function of the source dislocation
function, we can determine the effective tectonic stress σ 0 e from the initial slope and the
stress drop Δσ s from the final dislocation (static field).

It is important to note that since only σ 0 e and Δσ s appear in these expressions

we cannot determine the actual tectonic stresses σ 0 and σ 1 , and the frictional stress, σ f ,
by seismological methods alone.

7. Static Stress Drop and Dynamic Stress Drop

Referring to the figure shown above, the static stress drop of an earthquake, Δσ s ,
is defined by

Δσ s = σ 0 − σ 1 (11)

where σ 0 and σ 1 are the initial and the final stresses, respectively. This definition is
straightforward and unambiguous.
The dynamic stress drop, Δσ d , is the stress that drives fault motion.
Unfortunately, the definition of dynamic stress drop is not universal, and is
ambiguous. For the simplest stress release pattern shown by figure (a) above, it is the
same as the effective tectonic stress and is equal to the static stress drop, i e.,

Δσ d = σ 0 e (= σ 0 − σ f ) = Δσ s (= σ 0 − σ 1 ) (12)

In this case, the dynamic stress drop is unambiguous.

338
 However, for more complex stress release patterns in which friction changes as a
function of slip as shown by figure (b) above, we can define Δσ d by

Δσ d = σ 0 − σ f (13)

where
1 D
D ∫0
σf = σ f ds (14)

(s: slip).

This definition is similar to (12 ) and reasonable. In general, σ f ≠ Δσ s . Thus, it

is important to distinguish Δσ d and Δσ s .

Ge 162 Practice Session 6, Retrieval of Source Parameters

Figure 1 shows the ground-motion displacement, U (t ) , of the Jan. 26, 2001,

Bhuj, India, earthquake recorded at ESK (Eskdalemuir) in Scotland.

Estimate the seismic moment M 0 , and the approximate duration of the source (as
viewed from ESK). It is also possible to constrain the depth, H, but, for simplicity, we
fix H, except in the last step where the effect of H on the waveform will be examined.

339
 India 1/26/2001 Station ESK Displacement

Displacement, cm

Time, sec

Fig. 1

The following is the general principle.

1) Waveform in the whole space.

As discussed in class, the far-field waveform in the Haskell model can be

approximated by a trapezoidal function

T (t ; τ 0 , tc ) (1)

where τ 0 and tc are the two time constants of the source. The area under T (t ; τ 0 , tc ) is
unity. Then, the P-wave form in a whole space is given by

340
 M 0 Rθφ
u1 (t ) = T (t − r / α ; τ 0 , tc ) (2)
4πρ rα 3

2) The source and receiver effects and the geometrical spreading.

In the real earth, the structure near the source and the receiver complicates the
wave form. Here, we use a simple half space. Then, we need to include the near-source
reflections pP ans sP, and the effect of the free surface near the station. Here, the near-
source effect is approximated by

CS (t ) = δ (t ) + a1δ (t − t1 ) + a2δ (t − t2 ) (3)

The first, second, and third terms represent, the direct P, pP, and sP, respectively. a1 , a2 ,
t1 , and t2 depend on the radiation pattern, the depth, and the structure. In particular,
changing the depth has a large effect on the waveform.

The receiver effect is simply given by a scalar factor, CR . (For SH wave, CR =2.)

The geometrical spreading factor is also simplified by a scalar factor g, and 1/r in
(2) is replaced by g/ RE where RE is the radius of Earth.

Including these, the wave form is now modified to,

rg
u2 (t ) = CR u1 (t ) * CS (t ) (4)
RE

Even for this simplified problem, computation of (4) is not that simple because of
the complexity of CS (t ) = 1 + a1δ (t − t1 ) + a2δ (t − t2 ) . A simple program pwsyn2g_asc.f is
provided in /home/ftp/pub/hiroo/ge162.dir/practice_6.dir. The output of this program is
u2 (t ) .
Run this program for various τ 0 , tc , and the depth H.. The constants, a1 , a2 , t1 ,
and t2 are computed in the program from the fault parameters and the depth. This
program computes the displacement in cm for a unit moment ( M 0 =1020 N-m).

341
(Another program pwsyn2g.f which does the same and outputs the results in SAC format
is also provided.)

3) Observed Waveform

To simulate the observed waveform at the station, we need to add the effects of
attenuation. The attenuation function, F(t; t*), for t * = 1sec is shown in Figures 2, and the
file is in practice_6.dir (o_futtm.asc (ASCII file) and o_futtm (SAC file)).

Fig. 2

The synthetic ground-motion displacement can be computed by

342
 u3 (t ) = u2 (t ) ∗ F (t ) (5)

This can be directly compared with the observed record, U (t ) , shown in Figure 1.

4) Determination of M 0

Once the waveforms are matched satisfactorily, the amplitude ratio of the
observed waveform

amplitude(U (t ))
amplitude(u3 (t ))

gives the seismic moment, M 0 , of the earthquake in the unit of 1020 N-m.
Since many approximations have been made, the waveforms cannot be matched
completely, but try to match the overall waveform and the amplitude of the first 1 cycle.

Here, we use only one station, and the mechanism (i.e., radiation pattern) is
assumed. However, if we have more than one station, we can determine the mechanism
by matching the amplitudes at different stations with the same M 0 .

In general, matching the displacement record is much easier than matching the
velocity record, V (t ) , shown in Figure 3, because high-frequency components which are
harder to model have been removed in the displacement record.

343
 India 1/26/2001 Station ESK Velocity

Velocity, cm/sec

Time, sec

Fig. 3

The relevant parameters needed for the computation are in the following.

Bhuj, India, Earthquake

O.T. 03:16:41.0, 1/26/2001

Lat. 23.40°, Long. 70.32°, Depth (H), 10-40 km
Dip( δ )= 64°, rake( λ )=60°, strike( φ f )=66°

Eskdalemuir

Lat.=55.3167°, Long.=3.2050°, Elevation=242m

Epicentral Distance ( Δ )=61°, Azimuth=321°, Takeoff Angle (ih)=18°

344
Source Crustal Structure (Half Space)

α =6.0 km/sec, β =3.46 km/sec, ρ =2.67 g/cm3

For the mechanism and the takeoff angle given above, the radiation pattern factor for P-
wave at the source

Rθφ =0.914 (this is computed in the program)

Geometrical spreading factor computed for a standard Earth model

g =0.35

Steps for Practice

1. Compute u2 (t ) for

H=10 km, τ 0 =8 sec, tc =12 sec

using pwsyn2g_asc.f. Plot, o_Pamp.asc (direct P wave), o_pPamp.asc (pP wave),

o_sPamp.asc (sP wave), and o_Ptotalamp.asc (sum of all). These are ASCII files with
simple 2-column t vs. u2 (t ) data.

Observe, how P, pP and sP phases interact. Note that the far-field waveform in a
whole space discussed in class should be a one-sided pulse, while the observed U (t ) is
two sided. The interaction between these phases is the primary reason why the observed
displacement is two-sided.

2. Convolve o_Ptotalamp.asc with F(t; t*=1 sec) to compute u3 (t ) . Program convg3m.f

is provided, but you may want to write a simple program for convolution.*

345
3. Compare u3 (t ) with U (t ) to determine M 0 . The starting time is arbitrary so that you
can shift the waveform arbitrarily in comparison.

4. Try 1, 2, and 3, for different combinations of τ 0 , and tc (fix H) and find the best
solution.

5. Compute u3 (t ) using a source time function shown below, and compare it with U (t ) .
Vary H from 5 to 35 km, and see the difference in the synthetic waveforms. (This source
time function is not a trapezoid, which means that the source model is different from the
Haskell model.)

Suggested source time function

0.0 sec 0.0

2.0 0.0
4.0 0.2
8.0 1.0
22.0 0.0
100.0 0.0

6. (Optional)

Compute the velocity v3 (ti ) = [u3 (ti +1 ) − u3 (ti ) ] / Δt for the best model, and
compare it with V(t).

Explanation of the Programs

pwsyn2g.f

pwsyn2g.f takes input file i_pwsyn2g. Only the parameters in bald face need to be
changed. (do not change c_pwsyn2g.)

i_pwsyn2g

346
line 1 station id

line 2 rho, α, dt, *, *, * (density in g/cm3, P-velocity in km/s(S-

velocity is assumed to be α/1.732), dt of data (0.05 sec for VBB)

line 3 fif, dip, rake (fault strike, dip and rake)

line 4 ih, fis, H(depth), g, * (take-off angle, station azimuth,

depth, g factor)

line 5 flag for the choice of source time function (do not change)

line 6 nt (# of points where time function is defined_

line 7 t1, y(t1) (time, amplitude (amplitude is arbitrary; the

amplitude is eventually normalized in the program.)

line 8 t2, y(t2)

etc

Example

ESK
2.67 6.0 0.05 0.01 0.01 0.01
66. 64. 60.
18. 321. 10. 0.35 1.0
1
6
0.0 0.0
2.0 0.0
10.0 1.0
14.0 1.0
22.0 0.0
100.0 0.0

347
The output file o_pwsyn2g contains the parameters used and other computed parameters.

Output files o_Pamp.asc, o_pPamp.asc, o_sPamp.asc, and o_Ptotalamp.asc contain the

time series for P, pP, sP, and the sum in ASCII format.

convg3m.f

This program computes convolution of x(t ) and y (t ) . The result is

z (t ) = x(t ) ∗ y (t ) .

i_convg3m

348
line 1 file name of x(t )
line 2 file name of y (t )
line 3 file name of z (t )

Example

i_convg3m

o_Ptotalamp.asc
o_futtm.asc
o_Ptotalamp_q.asc

* convolution of xi (i = 1, 2, , l ) and y j ( j = 1, 2, , m) is

⎛ l ⎞
zk = ⎜ ∑ xi yk −i ⎟ Δt k = 1, 2, , l + m
⎝ i =1 ⎠
(if k − i ≤ 0 , then yk −i =0)

Ge 162

349
6. Retrieval of Seismic Source Parameters

6.1 Far Field Body Waves

1. Introduction

Far-field body waves are widely used to determine the source mechanism (fault
plane solution), seismic moment, and rupture patterns (complexity). The method is
conceptually simple, but the actual procedure involves many steps. These steps include
the calculation of the following:

1. Source time function (Dislocation time history)

2. Source finiteness function (rupture function), T (t ;τ 0 , tc )
3. Radiation pattern, R
4. Response of source and receiver structures, CS (t ) and CR (t )
5. Geometrical spreading function, g (Δ, h)
6. Attenuation (Q operator), q (t )
7. Instrument response, I (t )

The displacement at far-field is then given by,

M 0 Rg (Δ, h)
u ( Δ, t ) = T (t ;τ 0 , tc ) * CS (t ) * CR (t ) * q(t ) * I (t )
4πρ vc3 RE

where M 0 is the seismic moment, ρ is the density, vc is P- or S-wave velocity, and RE

is the radius of Earth.

These steps are described in detail in the following.

2. Radiation Pattern from a Point Double Couple

The radiation pattern from a point double couple has been already given with
respect to the coordinates fixed to the double couple.

350
 For actual applications it is convenient to express the radiation pattern by using a
geographical coordinate system. For a fault model shown in the figure, the displacement
at point P can be given by:

1 (1)
ur ( r , t ) = R P M (t − r / α )
4πρ rα 3

1 (2)
uθ (r , t ) = Rθ M (t − r / β )
4πρ r β 3

351
 1 (3)
uφ (r , t ) = Rφ M (t − r / β )
4πρ r β 3

where
R P = sR (3cos 2 ih − 1) − qR sin 2ih − pR sin 2 ih (4)
3 1
Rθ = sR sin 2ih + qR cos 2ih + pR sin 2ih (5)
2 2
φ
R = − qL cos ih − pL sin ih (6)
qL = −(cos λ cos δ ) sin φ + (sin λ cos 2δ ) cos φ (7)
pL = (sin λ sin δ cos δ ) sin 2φ + (cos λ sin δ ) cos 2φ (8)
sR = sin λ sin δ cos δ (9)
qR = sin λ cos 2δ sin φ + cos λ cos δ cos φ (10)
pR = cos λ sin δ sin 2φ − sin λ sin δ cos δ cos 2φ (11)

If the azimuth of the fault strike ( x1 axis) φ f and the azimuth of the station (point
P) φs are measured clockwise from N, then φ (measured counter-clockwise from ξ1
axis) in the above formula should be given by

φ = φ f − φs (12)

By using (1) to (12) we can compute the far-field displacements of body waves
from a fault of arbitrary geometry. The station coordinates are given by (r , ih , φ ) rather
than the conventional (r , θ , φ ) .

352
3. Source Finiteness-- Haskell Model

One of the useful kinematic source models is the Haskell(1964) model. In this
model, a seismic source is represented by a rectangular fault with length L and width W.
The dislocation (fault offset), D, on the fault plane is uniform in space, and its temporal
variation is given by a linear ramp function with a rise time τ 0 .
The rupture propagation is assumed instantaneous in the width direction.
Lengthwise, the rupture propagates from one end of the fault to the other with a
uniform rupture speed V. In this sense, the rupture propagation is one-dimensional
unilateral.

Assume that the rupture propagates from A to B over a length L. If the seismic
moment per unit fault length is m, the moment per line element dl is mdl.

The far-field time function due to this line element is given by a box-car function of
width τ 0 and height mdl / τ 0 .

353
If AP is very large compared with AB, the displacement at P (station) can be given by

R ⎛ r l l cos Θ ⎞
mdls ⎜ t − − + ⎟
4πρ rvc
3
⎝ vc V vc ⎠

where r = AP, vc is either P or S wave velocity, R, the radiation pattern and V is the
rupture velocity. s(t) is defined above, and is called the local dislocation rate function.
Therefore, for a source propagating from A to B, we have

⎛ r l l cos Θ ⎞
∫
L
R
u (r , t ) = m s⎜t − − + ⎟ dl (13)
4πρ rvc3 0 ⎝ vc V vc ⎠

Introducing the following variables,

τ = t − r / vc , tc = L / V − L cos Θ / vc , t1 = ltc / L , dl = Ldt1 / tc (14)

we obtain,

∫
tc
R M0 (15)
u ( r ,τ ) = s (τ − t1 ) dt1
4πρ rvc3 tc 0

where M 0 = mL is the total seismic moment.

Introducing a function r (t1 ) , as shown in the figure,

354
we obtain

∫
tc
RM 0
u ( r ,τ ) = r (t1 ) s (τ − t1 ) dt1
4πρ rvc3 0

+∞
(16)

∫
RM 0
= r (t1 ) s (τ − t1 ) dt1
4πρ rvc3 −∞

which is convolution of r(t) and s(t). r(t) is called the rupture function.

The above convolution yields a trapezoidal function T (t ; t1, t2 )

given by the following figure.

if t1 > t2

Note that

+∞ t1 + t2

∫−∞
T (t ; t1 , t2 )dt =
∫ 0
T (t ; t1 , t2 )dt = 1 (17)

355
Thus, the pulse width at the station in the azimuth Θ is

L L
w(Θ) = τ 0 + − cos Θ (18)
V vc

If τ 0 is small, the pulse shape at Θ = 0, Θ = π /2 and Θ = π may be schematically

shown by the following figure.

Note that S is always equal to 1.

The amplitude spectrum of T (t ; τ 0 , tc ) is given by

sin(ωτ 0 / 2) sin(ω tc / 2)
Tˆ (ω ; τ 0 , tc ) = (19)
(ωτ 0 / 2) (ω tc / 2)

Note:

356
 In applying the above results to body-wave radiation in the real earth, another
factor needs to be considered. For example, if AB is horizontal, we let the take-off angle
of the ray be ih , and the azimuth of the station from AB measured on the earth’s surface
be Θ . Then the phase velocity along the Earth’s surface is vc / sin ih . Then we replace
r / vc and vc in the argument of s(t) by T and vc / sin ih respectively, where T is the travel
time. Hence we have, instead of (14) and (15),

∫
tc
RM 0
u ( r ,τ ) = s (τ − t1 ) dt1 (20)
4πρ rvc3tc 0

where,

τ = t − T , and tc = L / V − L cos Θ sin ih / vc (21)

For example, directivity observed for the 1999 Landers earthquake is illustrated in
the following figure.

357
358
 Directivity has strong influence on the distribution of strong ground motion as
shown for the 1994 Northridge earthquake.

359
360
4. Response Function at the Source and Receiver

If the source is shallow, the reflections from the free surface (and other structures)
complicates the source wave as shown by the following figure.

361
 A similar situation occurs near the station too. These effects are usually included
as the source and receiver response functions.

5. Geometrical Spreading of Body-Wave Energy

5.1 Energy Density and Energy Flux

In an elastic medium with density ρ , the elastic energy per unit volume is given
by
ε = ρu 2 (22)

where u is the displacement and u is the particle velocity.

Consider a plane wave propagating in x direction with velocity v.
Then, the energy flux per uinit area per unit time is given by

φE = ρ u 2v (23)

5.2. Radiated Energy from a Source

Consider an elastic wave radiating from a double couple source at O.

Consider a small sphere with radius r around O.

Then elastic wave energy emitted per unit time from the source within a ring
defined by ih and ih + dih is

ER = 2π RE sin ih rdih ρ hu 2 vh (24)

where RE is the radius of the Earth, and

M
u= RP
4πρ rvc
3

362
 P
Here the subscript h signifies the values of the variables at the source, and R is the
radiation pattern.

5.3 Energy arriving at station P

From the figure above, the energy arriving at station P (per unit time, per unit area
of wave front) is

ρ0u02 v0 (25)

where subscript 0 signifies the value of the variables at the station.

5.4. Geometrical Spreading Factor g (Δ, h) / RE

Since the total energy emitted in a ring defined by ih and ih + dih arrives in a ring
defined by Δ and Δ + d Δ , we obtain from (23), (24), and (25),

RE d Δ cos i0 2π RE sin Δρ0u02 v0 = ER (26)

from which we obtain

MR P 1 ρ h vh sin ih dih MR P 1
u0 = = g ( Δ, h) (27)
4πρ h vh3 RE ρ 0v0 sin Δ cos i0 d Δ 4πρ h vh3 RE
where
ρ h vh sin ih dih
g ( Δ, h) = (28)
ρ0v0 sin Δ cos i0 d Δ

g (Δ, h) / RE is called the geometrical spreading factor.

6. Effect of Anelasticity

363
 Usually, the anelasticity of the medium is given in terms of quality factor Q. If
we write the quality factor of P and S waves by Qα and Qβ , the amplitude attenuation of
P and S waves of angular frequency ω during the propagation can be given by

⎛ ω ds ⎞ ⎛ ω ds ⎞
exp ⎜ − ∫ ⎟ exp ⎜− ∫ ⎟
⎝ 2 Qα α ⎠ , and ⎝ 2 Qβ β ⎠ (29)

where the integral is taken along the ray path. Since Qα and Qβ are functions of depth
and therefore of the path length, s, we may write these

⎛ ω ⎞ ⎛ ω ⎞
exp ⎜ − tα* ⎟ , and exp ⎜ − t β* ⎟ (30)
⎝ 2 ⎠ ⎝ 2 ⎠

where

∫ ∫
ds ds
tα* = , and t β* = (31)
Qαα Qβ β

Empirically, tα* and t β* do not vary significantly with the distance, and tα* = 0.5 to 1 sec
and t β* = 3 to 4 sec for distance ranges Δ > 35 degrees.

In order to maintain causality of P and S pulses, the phase spectrum needs to be

modified too.
In time domain, we can represent the effect of attenuation by “Futterman Q
operator”

q(t ; tα* ) and q (t ; t β* ) (32)

These functions are shown below for tα* =1 sec and t β* =4 sec.

364
7. Instrument Response

Since seismographs have their own frequency response, they modify the
waveforms. In the following, the impulse response of the VBB and WWSSN-LP
instruments are shown.

365
366
367
8. Simplified Expressions for a Point Source

The following simplified expressions are not exact, but are useful for understanding
the basic principles for retrieval of source parameters.

8.1 Retrieval of Seismic Moment M 0

We approximate a finite source by a point source with a moment rate function

M 0 (t ) . The area under M 0 (t ) is the seismic moment M 0 and the width of M 0 (t ) is
approximately equal to the rupture time τ 0 = L / V where L is the length scale of the
source and V is the rupture speed.

In the following, we use the vertical component of P wave for illustration purposes.
We assume that the observed seismogram has been corrected for the instrument, I(t),
attenuation, q(t), and the response of the source strucuture, CS (t ) , and the response of the
receiver structure is given by a scalar CR . Then, from the first equation in this section,
we can write the vertical component of P wave, u p (t ) , as

M 0 (t ) R g
u p (t ) = CR (33)
4πρ hα h3 RE

From which,

4πρ hα h3 RE
M 0 (t ) = u p (t ) (34)
RgCR

Integrating,

4πρ hα h3 RE τp
M0 =
RgCR ∫
0
u p (t )dt (35)

368
where τ p is the duration of the P wave record.

Using (35), we can estimate M 0 from seismograms.

8.2 Determination of Radiated Energy from P waves.

Since S waves are severely attenuated during propagation to teleseismic distances,

we usually use P waves to estimate the radiated energy in P waves, Eα , and then estimate
the total radiated energy, ER . If we consider a small sphere around the source, and write
the P wave on the spherical surface by U p (t ) , the total energy radiated by P waves is
given by,

2
τp τp ⎡ R ⎤
Eα =
∫ dS ∫ ρ hα hU (t )dt =
∫ dS ∫ ρ hα h ⎢
2
M ( t ) ⎥ dt
⎣ 4πρ hα h r
p 3 0
S 0 S 0 ⎦ (36)
⎡ 1 ⎤ τp
= ρ hα h ⎢
∫ R dS ∫
2
3 ⎥
M 02 (t )dt
⎣ 4πρ hα h r ⎦ S 0

Substituting (34),

2
16π ⎛R ⎞ 1 1 τp
Eα =
15
ρ hα h ⎜ E ⎟ 2 2
⎝ g ⎠ R CR ∫0
u 2p (t )dt (37)

where

369
 1 4
4π r 2 ∫
S
R 2 dS ≡ R 2 =
15
(38)

for P waves is used.

Using (37), we can estimate Eα from the observed record at a station. For a point
source, we can show that

5
3⎛α ⎞
Eβ = ⎜ ⎟ Eα ≈ 23Eα (39)
2⎝ β ⎠

and

ER = Eα + Eβ = 24 Eα (40)

Ge 162 Problem #8 Interpretation of Far-field Body Wave

Figure 1 shows the P-wave seismograms (WWSSN long-period response)

observed for the May 2, 1983 Coalinga, California, earthquake. The numbers attached to
each seismogram is the amplitude.

370
 Fig. 1

In order to determine the source parameters (seismic moment, source duration),

we must remove the effects of

1. Radiation pattern, R
2. Response of source and receiver structures, CS (t ) and CR (t )
3. Geometrical spreading, g (Δ, h)
4. Attenuation (Q operator), q (t )
5. Instrument response, I (t )

Fortunately, Hartzell and Heaton (1983) have removed all of these effects, and
obtained a trapezoidal displacement at point P, as shown in Figure 2.

371
 Fig. 2

Fig. 3

As we discussed in class, the displacement u (r , t ) at far-field due to a Haskell

type source is given by

372
 +∞

∫
RM 0 RM 0
u ( r ,τ ) = r (t1 ) s (τ − t1 ) dt1 = T (τ ;τ 0 , tc ) (1)
4πρ rvc3 −∞ 4πρ rvc3

where ρ is the density, vc is P- or S-wave velocity, M 0 is the seismic moment, and

τ = t − r / vc . Note that the amplitude of T (τ ;τ 0 , tc ) is 1/ tc , and the area under it is 1. tc is
given by

L L cos Θ
tc = −
V vc

where Θ is the angle between the rupture direction and the ray.

1. Check the dimension of the right-hand-side of equation (1) to make sure that it is the
dimension of "length".

2. We assume that the medium is homogeneous with α = 6 km/sec, β =3.5 km/sec and
ρ =2.6 g/cm3, and the radiation pattern R is equal to 1. We also assume that the fault
plane AB and P are on the same plane. The take-off angle of the ray to P is 30 degrees
(note that this is different from Θ ), and the distance is r=1000 km. The rupture
propagation is assumed to be one dimensional unilateral (A to B) with a rupture speed
V=2.5 km/sec.

(1) Determine the rise time τ 0 and tc from Figure 2.

(2) Determine the seismic moment M 0 .
(3) Estimate the fault length L.
(4) Assuming that the fault width W=L/3, estimate the average dislocation.
(5) Estimate the particle velocity (average) of one side of the fault.
(6) Estimate the static stress drop (use the strike-slip geometry for simplicity).
(7) Sketch the (trapezoidal) waveform that would be observed at point P1, P2, and P3, all
at a distance of 1000 km, but in three different azimuths, as shown in Figure 3.

373
(Note: The solution obtained by Hartzell and Heaton is slightly different from the above.)

6.2 Excitation of Surface Waves and Free Oscillations

1. One Dimensional Problem

We first consider a one-dimensional problem shown below.

∂ 2u ∂ 2u
ρ = E +ρf (1)
∂t 2 ∂x 2

where f is the force per unit mass.

Consider a harmonic excitation ρ f ( x, t ) = F ( x)eiω t .

Putting u ( x, t ) = U ( x)eiω t we have

d 2U
− ρω 2U = E +F (2)
dx 2

Let yi , i = 1,2... be the eigen functions of the homogeneous equation of (2) i.e.,

2
d yi
− ρω i2 yi = E
dx 2 (3)

Then yi 's constitute an orthogonal system. That is,

374
 L
∫ ρy y dx = c δ
0 n l n,l
(4)

Then we can expand U in yi , i.e.,

∞
U = ∑ ai yi
i =1 (5)

where

1 L
c ∫0
ai = ρ yiUdx
(6)

Multiplying (2) by yn and integrating it from x=0 to x=L, we obtain

−c ω an = E ∑ al ∫0 yl′′y n dx + ∫0Fy n dx
L L
2

Using (3) and (4),

L
−c ω 2 an = −cω 2n an + ∫0F yn dx

1 L
∴ an = − 2 ∫0Fy n dx
c(ω − ω n )
2
(7)

∞ ⎡ ⎤ L
yn
∴ U = ∑ ⎢− 2 ⎥ ∫0F yn dx
n =1 ⎣ c(ω − ω n ) ⎦
2
(8)

∞ ⎡ ⎤ L
yn
∴ u = eiωt ∑ ⎢ − 2 ⎥ ∫0 Fyn dx
n =1 ⎣ c(ω − ω n ) ⎦
2
(9)

Now let us consider excitation by a step function force H(t). Since

375
 1 +∞ 1 iωt
2π ∫−∞ iω
H(t) = e dω
(10)

we have from (9)

∞
⎡ y L ⎤ 1 +∞ ⎡ e iω t ⎤
∴ u(t) = ∑ ⎢ − n ∫0Fy n dx ⎥
⎦ 2 π ∫− ∞ 2 ⎥dω
⎢ −
n =1
⎣ c ⎣ i ω (ω − ω n ) ⎦
2
(11)

The poles are at ω = 0, ω = ω n + iε , ω = −ω n + iε .

Since any physical system must have at least small attenuation, these poles are slightly
above the real axis (see Morse and Feshbach, p. 1334).

Carrying out the integration along the path given above, we have, for t > 0,
∞
⎤ (1 − cos ω n t )
∑ ∫ ⎡ L
u (t ) = ⎢ yn Fyn dx ⎥ (12)
n =1 ⎣ 0 ⎦ cnω n2

The first term

376
 ∞

∑ ∫ ⎡ ⎤ 1
L
u (t ) = ⎢ yn Fyn dx ⎥ (13)
⎦ cnω n
2
n =1 ⎣ 0

gives the permanent deformation, and the second term

∞
⎤ cos ω nt
∑ ∫ ⎡ L
u (t ) = − ⎢ yn Fyn dx ⎥ (14)
⎦ cnω n
2
n =1 ⎣ 0

gives the free oscillation.

2. Torsional Oscillation

For torsional oscillations of a sphere, the eigen functions and the force are
vectors.

(15)
⎛ ⎞
⎜ 0 ⎟
⎜ ⎟ ⎛ Fr ⎞
⎜ 1 ∂Yl m ⎟ ⎜ ⎟
n yl = y1, n ⎜
m
⎟ , F = ⎜ Fθ ⎟
⎜ sin θ ∂ φ ⎟ ⎜F ⎟
⎜ ⎝ φ⎠
∂Yl m ⎟
⎜ − ⎟
⎝ ∂θ ⎠

The expression for step function excitation can be derived in a manner similar to the one-
dimensional case.

⎤ (1 − cos n ωlmt )
u (r , t ) = ∑ ⎡ m
⎢ n yl
l ,m,n ⎣
∫V
F ⋅ n ylm dV ⎥
⎦ n cl n ωl
m m2
(16)

where
c =
∫ρ ylm ⋅ n ylm dV
m
n l n
V

377
3. Spheroidal Oscillation

⎛ ⎞
⎜ 0 ⎟
⎛ Yl m ⎞ ⎜ ⎟
⎜ ⎟ ⎜ ∂ Y m
⎟
n yl = y1, n ⎜ 0 ⎟ + y3, n ⎜ (17)
m l

∂θ ⎟
⎜ 0 ⎟ ⎜ ⎟
⎝ ⎠ ⎜ 1 ∂Yl ⎟
m

⎜ sin θ ∂φ ⎟
⎝ ⎠

Then

⎤ (1 − cos n ωlmt )
u (r , t ) = ∑ ⎡ m
⎢ n yl
l ,m,n ⎣
∫ V
F ⋅ n ylm dV ⎥
⎦ n cl n ωl
m m2
(18)

Again
c =
∫ρ ylm ⋅ n ylm dV (19)
m
n l n
V

4. Displacement due to a point Moment Tensor

Displacement due to force couples can be obtained by differentiating (16) and

(18) in appropriate directions. Then the displacement due to a point moment tensor M
can be written as

1 − cos n ωlmt
u (r , t ) = ∑
l ,m,n
⎡⎣( M : ε ) n ylm (r , t ) ⎤⎦
n cl n ω l
m m2
(20)

where ε is the strain tensor computed at the source for each l, m, and n.

5. Comparison with observations

Several examples are shown in the following.

1. 1994 Bolivia Deep earthquake

378
379
2. 2001 Bhuj India earthquake

380
381
3. 1999 Russia-China Border earthquake

382
383
7. Summary of Seismic Source Parameters

1. Fault Geometry

Usually given by the following three parameters: φ f =fault strike; δ = dip

angle, λ = slip angle (rake).

2. Fault Dimension

L and W or a (radius). Area S.

(S = LW) (S = π a2)

3. Dislocation (Fault Offset), D

In general, D(r , t )
Usually only the average D is used

384
 D = Δu

4. Seismic Moment (Moment)

M 0 = μ SD

5. Seismic Moment Tensor

M pq = Cijpq Δuiν j ΔS (in general)

M = μΔS (Δu :ν ) (shear dislocation)

This can be related to the fault parameters (5.1.4).

6. Stresses

Initial stress* σ0
Final stress* σ1
Static stress drop Δσ s = (σ 0 − σ 1 ) ∝ D
Frictional stress* σ f
Effective tectonic stress (Dynamic stress drop) Δσ d = σ 0 − σ f ∝ D
Average stress* σ = (σ 0 + σ 1 ) / 2

* This cannot be determined with seismological methods

7. Energy

Radiated Energy ER ∝ ρ v ∫ ∫ u 2 dtdS

385
 Potential Energy Change*
(Strain Energy + Gravitational Energy) ΔW = σD S
Frictional Energy Loss* EF = σ f DS
Fracture Energy* EG

ΔW = ER + EF + EG = (Radiated Energy) + (Non-Radiated Energy)

* This cannot be determined with seismological methods

8. Rupture Mode and Rupture Speed

Unilateral, Bilateral, Radial, Two-Dimensional

V, usually 75 to 90% of S velocity
Directivity

386
387
388
389
9. Complexity

Multiple Events, "Fractal" structure at short wave length?

390
391
10. Energy (or Moment) Magnitude

M w = (log M 0 − 16.1) /1.5 ( M 0 in dyne-cm)

M w = (log M 0 − 9.1) /1.5 ( M 0 in N-m)

---------------------------------------------------------------------------------------------------------------------------------
From the energy budget, the radiated energy, ER , is given by

ER = ΔW − EF = DS (σ − σ f )

As mentioned above, the initial stress σ 0 , the final stress σ 1 , and the frictional stress σ f
cannot be determined directly with seismological methods, and ER above cannot be
determined. However, note that

σ 0 + σ1 σ 0 − σ1 Δσ s
ER = DS ( − σ f ) = DS ( + (σ 1 − σ f )) = DS ( + (σ 1 − σ f ))
2 2 2

392
If we assume that σ 1 is equal to σ f (i.e., the stress on the fault plane is equal to the
average frictional stress when fault motion stops), then,

1 Δσ s Δσ s
ER = DS Δσ s = DS μ = M0
2 2μ 2μ

For most large earthquakes, Δσ s / 2μ ≈1/(2 x104 ) , and we have, approximately,

ER = M0 /(2x104)

Once ER is determined we can define a magnitude scale M w using the Energy-

Magnitude relation backward,

log ER = 1.5M w + 11.8 ( ER in ergs)

or,

M w = (log M 0 − 16.1) /1.5 ( M 0 in dyne-cm)

------------------------------------------------------------------------------------------------------------

393


394
Ge 162

8. Physics of Earthquakes

8.1 Scaling Relations

For understanding the overall physics of earthquakes without going into details, it
is useful to investigate scaling relations between several macroscopic source parameters.

1. Moment versus Fault Area

395
 [closed circles: interplate, open circles: intraplate, Kanamori, H., and D. L.
Anderson, Theoretical basis of some empirical relations in seismology, Bull. Seis. Soc.
Amer., 65 (5), 1073-1095, 1975]

By definition,

M 0 ∝ Δσ s a 3 circular
Δσ s w L rectangular
2

For a circular fault,

396
 3 ⎛ 16 ⎞
log M 0 = log S + log ⎜ π −3/ 2 Δσ s ⎟
2 ⎝7 ⎠

3
If Δ σ s is constant, then, log M 0 ≈ log S . From the attached figure, we see,
2

Δ σ s ~ 60 bars on the average

Δ σ s ~ 30 bars interplate (plate boundary)
Δ σ s ~ 100 bars intraplate

2. Seismic Moment vs. Source Dimension

The scaling relation shown above has been extended to small earthquakes as
shown below.

397
[Modified from Abercrombie, R., and P. Leary, Source parameters of small earthquakes
recorded at 2.5 km depth, Cajon Pass, Southern California: Implications for earthquake
scaling, Geophys. Res. Lett., 20, 1511-1514, 1993]
The L ∝ M 01/ 3 scaling is seen, and this is generally interpreted as evidence for
constant stress drop, but the fairly large scatter in Δσ s should be noted.

8.2 Physics of Earthquakes

The heterogeneity of properties and structures of fault planes seems to have a

profound influence on dynamics of faulting. Although we cannot resolve every detail of
fault zone heterogeneities, we want to interpret macroscopic seismological data in terms
of the overall processes occurring on a fault plane.

8.2.1 Energy Budget

As we discussed in 5.1.5, the total potential energy (strain energy + gravitational

energy) change in earthquakes is given by

1
ΔW = (σ 0 + σ 1 ) DS (1)
2

Then the energy budget can be written as

ΔW = ER + EF + EG (2)

where ER and EG are the radiated energy and the fracture energy, respectively, and EF

398
is the frictional energy. In this model, separation of EG and EF is somewhat arbitrary
(both represent non-radiated energy, i.e., dissipation), but, EG , and EF are commonly
defined by the hatched and cross-hatched areas shown in the figure below, respectively,
i.e.,

EF = σ f0 DS (3)

However, here the fracture energy, EG , is the energy dissipated during rupture
over a volume surrounding the fault zone.

(In this diagram, the energies are interpreted as those per unit area.)

The radiated energy, ER , is what we can measure from the radiated seismic
waves, as shown in 6.1, but because of the practical difficulty in measuring it accurately,
ER has not been fully used in seismology for the purpose of understanding the physics of
earthquakes.
Only recently, it became possible to measure ER accurately enough so that we
can investigate the physics of earthquakes in terms of energy budget.

8.2.2 The Radiation Efficiency

399
 The ratio of the fracture energy, EG , to the radiated energy, ER , determines the
characteristics of fracture, or an earthquake. Alternatively, the radiation efficiency
defined by

ηR = ER /( ER + EG ) (4)

can be used for the same purpose. Referring to the figure above,

ηR = ER /(Δσ s DS / 2) = ER /(Δσ s M 0 / 2μ ) = 2( ER / M 0 ) /(Δσ s / μ ) (5)

Thus, we can determine η R , from the observed macroscopic parameters,

e = ER / M 0 , and the static strain drop, Δε = Δσ s / μ .

.
8.2.3 Observations

General Observation

The results for large earthquakes are summarized in the following figure.

400
Radiation efficiency η R = ER /( ER + EG ) as a function of M w . The different symbols
show different types of earthquakes as described in the legend. Most earthquakes have
radiation efficiencies greater than 0.25, but tsunami earthquakes and two of the deep
earthquakes (the Bolivia earthquake and the Russia-China earthquake) have small
radiation efficiencies. (Venkataraman and Kanamori, 2004)

Except for the very large deep focus earthquake, the 1994 Bolivian earthquake,
and tsunami earthquakes, the radiation efficiency, η R , is larger than 0.25, which means
that the fracture energy for most large earthquakes, regardless of their tectonic
environment, are comparable or less than ER .
Deep Focus Earthquake

401
 It is difficult to accurately determine the size of the fault plane, S, for deep focus
earthquakes. However, for the 1994 Bolivian earthquake (Mw=8.3, depth=635 km), the
largest deep-focus earthquake ever recorded, the source parameters could be determined
well enough to investigate the energy budget.
As we discussed in 5.1.5, we cannot determine ΔW itself, but we can estimate its
lower bound,

ΔW0 = Δσ s DS / 2 (6)

from Δσ s , D , and S .
The result for the 1994 Deep Bolivian earthquake (Mw=8.3) showed that
ΔW0=1.4x1018 J and ER =5x1016 J, which is only 3 % of ΔW0 ; the difference ΔW0 –
ER=1.35x1018 J, was not radiated, and must have been deposited near the focal region,
probably in the form of thermal energy. This energy 1.35x1018 J is comparable to the
total thermal energy released during large volcanic eruptions such as the 1980 Mount
Saint Helens eruption. The thermal energy must have been released in a relatively small
focal region, about 50x50 km2, within a matter of about 1 min. The mechanical part of the
process, i.e. the earthquake observed as seismic waves, is only a small part of the whole
process. Thus, the Bolivia earthquake should be more appropriately viewed as a thermal
process rather than a mechanical process.
With this much of non-radiated energy, the temperature in the focal region must
have risen significantly. The actual temperature rise, ΔT, depends on the thickness of the
fault zone, which is not known, but if it is of the order of a few cm, the temperature could
have risen to above 10,000 °C.

Shallow Earthquakes
Although the situation for shallow earthquakes may be different from that for
deep focus earthquakes, a simple calculation shows that if σf is comparable to Δσs, about
10 MPa, the effect of shear heating is significant. If the thermal energy is contained

402
within a few cm around the slip plane during seismic slip, the temperature can easily rise
to 100 to 1000 °C.
We consider a gross thermal budget during faulting under a frictional stress σf.
Let S and D be the fault area and the displacement offset respectively. Then the total
heat generated during faulting is Q=σf DS. If we assume that the heat is distributed
during seismic faulting within a layer of thickness w around the rupture plane, the
average temperature rise ΔT is given by

ΔT=Q/CρSw=σf D/Cρw (7)

The figure below shows ΔT as a function of magnitude. If a fault zone is dry (no
fluid), melting may occur and friction may drop. If fluids exist in a fault zone, fluid
pressurization could occur.

403
404
 The key question is how thick the fault slip zone is. Geologists have examined
many old fault zones which were formed at depths and were brought to the surface by
long-term uplift (i.e., exhumed faults). Some fault zones have a very narrow (about 1
mm) distinct slip zone where fault slips seem to have occurred repeatedly. The
Punchbowl fault, California, implies that earthquake ruptures were not only confined to
the ultracataclasite layer, but also largely localized to a thin prominent fracture surface.
They suggest that mechanisms that are consistent with extreme localization of slip, such
as thermal pressurization of pore fluids, are most compatible with their observations. In
other cases, several narrow slip zones were found but evidence shows that each slip zone
represents a distinct slip event (i.e., an earthquake). Thus, geological evidence suggests a
narrow slip zone, at least for some faults, but this question will remain debatable.

405
(Richard Sibson, Written communication, 2000)

406
 If a fault zone is narrow and rough, and if the material in the fault zone behaves as
viscous fluid, it is also possible that elastohydrodynamic lubrication plays an important
role in reducing friction for large events. An interesting consequence of this is that as the
slip and slip velocity increase, the hydrodynamic pressure within a narrow zone becomes
large enough to widen the gap thereby suppressing high-frequency ground motion caused
by the fault asperities rubbing against each other. During the recent Chi-Chi, Taiwan,
earthquake, the observed ground-motion near the northern end of the fault was extremely
large (> 2.5 m/s, the largest ever recorded), but short period acceleration was not
particularly strong so that the damage to ordinary structures by shaking was minor. This
could be a manifestation of the high-speed lubrication effects. However, since this is the
only earthquake for which such large slip and slip velocity were instrumentally observed,
whether this is indeed a general behavior or not is yet to be seen.

State of Stress
The results obtained for large earthquakes suggest that the average stress level
along mature faults where large earthquakes occur must be low because of the dominant
thermal effects such as frictional melting and fluid pressurization, or of
elastohydrodynamic lubrication. Because of melting or pressurization, a fault zone is
self-organized into a low stress state. That is, even if the stress was high in the early
stage of fault evolution, it would eventually settle in a low stress state after many large
earthquakes. This state of stress is consistent with the generally held view that the
absence of heat flow anomaly along the San Andreas fault suggests a shear strength of
about 200 bars or less . The stress in the crust away from active mature faults can be high
as has been shown by many in-situ measurements of stress. The stress difference is
large, and a kbar type stress may be involved in small earthquakes, but the events are in
general so small that it is hard to determine the stress parameters accurately. What is
important, though, is that as long as the length of the fault is small, the state of stress in

407
the fault zone would not affect the regional stress drastically. However, as the fault
grows to some length (e.g. Japanese intra-plate earthquakes like Tango, Tottori, Nobi
etc), then some sort of self-organization occurs and the fault settles at a stress level
somewhat higher than that on more active plate boundaries.

8.3. Earthquake as a Complex System

Large-magnitude earthquakes are rare events. To a very good approximation, the

rate of occurrence of earthquakes falls exponentially as a function of magnitude, as
shown in the following figure which shows the distribution of the number, N, of
earthquakes equal to, or larger than, magnitude M for the world (a) and southern
California (b). Approximately 1 earthquake with M ≥ 8 occurs every year somewhere in
the world, and approximately 1 earthquake with M ≥ 5.5 occurs every year somewhere in
southern California.

408
409
 In general, this distribution is expressed as

log N(M)=a–bM (8)

where a and b are constants.

The results obtained for many regions indicate that the value of b (called b value)
is approximately equal to 1. This relation is called the Gutenberg-Richter relation,
Ishimoto-Iida relation, or simply the magnitude-frequency relation. Since

M ∝ log ER /1.5

410
equation (1) means

N ∝ ER− b /1.5 (9)

that is, the relation between N and ER is given by a power law. If b=1 and

ER ∝ r 3 (r: size)

Then,

( N / N ') = (r / r ') −2 (10)

This relation suggests "self-similarity" in 2-D.

The observation that b value is constant and close to 1 has attracted many
researcher’s attention. This relationship can be interpreted as a result of complex
interaction between many elements in a system which has a large number of degrees of
freedom. Such systems are often illustrated by a mechanical slider block model (shown
below) a sand-pile model and a percolation model.

411
References

Fracture Mechanics and Friction

Aki, K., Characterization of barriers on an earthquake fault, J. Geophy. Res., 84, 6140-
6148, 1979.

Ben-Zion, Y., and D. J. Andrews, Properties and implications of dynamic rupture along a
material interface, Bull. Seismol. Soc. Am., 88, 1085-1094, 1998.

Brace, W. F., and J. D. Byerlee, Stick slip as a mechanism for earthquakes, Science, 153,
990-992, 1966.

Byerlee, J. D., Friction, overpressure and fault normal compression,, Geophysical

Research Letters, 17, 4,741-4,750, 1990.

Dieterich, J. H., Modeling of rock friction 1. Experimental results and constitutive

equations, J. Geophys. Res., 84, 2161-2168, 1979a.

Dieterich, J. H., Modeling of rock friction 2. Simulation of preseismic slip, J. Geophys.

Res., 84, 2169-2175, 1979b.

412
Dmowska, R., and J. R. Rice, Fracture theory and Its seismological applications, in
Theories in Solid Earth Physics, edited by R. Teisseyre, pp. 187-255, PWN-Polish
Publishers, Warzawa, 1986.

Eshelby, J. D., The determination of the elastic field of an ellipsoidal inclusion and
related problems, Proceedings of the Royal Soc. London, 241, 376-396, 1957.

Heaton, T., Evidence for and implications of self-healing pulses of slip in earthquake
rupture, Physicsof the Earth and Planetary Interiors, 64, 1-20, 1990.

Ida, Y., Cohesive force across the tip of a longitudinal-shear crack and Griggith's specific
surface energy, J. Geophys. Res., 77, 3796-3805, 1972.

Ide, S., and M. Takeo, Determination of constitutive relations of fault slip based on
seismic wave analysis, J. Geophys. Res., 102, 27,379-27,391, 1997.

Kanamori, H., Mechanics of Earthquakes, Ann. Rev. Earth & Planetary Sciences, 22,
207-237, 1994.

Kostrov, B. V., Seismic moment and energy of earthquakes, and seismic flow of rock
(translated to English), Izv. Earth Physics, 1, 23-40, 1974.

Lachenbruch, A. H., Frictional heating, fluid pressure, and the resistance to fault motion,
J. Geophys. Res., 85, 6097-6112, 1980.

Lachenbruch, A. H., and J. H. Sass, Heat flow and energetics of the San Andreas fault
zone, J. Geophys. Res., 85, 6185-6222, 1980.

Lawn, B., Fracture of Brittle Solids - Second Edition, Cambridge University Press,
Cambridge, 1-378, 1993.

Li, V. C., Mechanics of shear rupture applied to earthquakes zones, in Fracture

Mechanics of Rock, edited by B. Atkinson, pp. 351-428, Academic Press, London, 1987.

Mase, C. W., and L. Smith, Pore-fluid pressures and frictional heating on a fault surface,
Pure and Applied Geophysics, 122, 583-607, 1985.

413
Mase, C. W., and L. Smith, Effects of frictional heating on the thermal, hydrologic, and
mechanical response of a fault, J. Geophys. Res., 92, 6249-6272, 1987.

Melosh, J., Acoustic fluidization: a new geologic process?, J. Geophys. Res., 84, 7513-
7520, 1979.

Melosh, H. J., Dynamical weakening of faults by acoustic fluidization, Nature, 379, 601-
606, 1996.

Rabinowicz, E., Friction and Wear of Materials, 2 edition, John Wiley & Sons, Inc., New
York, 1-315, 1995.

Scholz, C. H., The mechanics of earthquake faulting, Cambridge University Press, New
York, 1-438, 1990.

Weertman, J., Unstable slippage across a fault that separates elastic media of different
elastic constants, J. Geophys. Res., 85, 1455-1461, 1980.

Energy Budget

Brune, J. N., T. L. Henyey, and R. F. Roy, Heat flow, stress, and the rate of slip along the
San Andreas fault, California, J. Geophys. Res., 74, 3821-3827, 1969.

Dahlen, F. A., The balance of energy in earthquake faulting, Geophys. J. R. astr. Soc., 48,
239-261, 1977.

Knopoff, L., Energy release in earthquakes, Geophysic. Jour., 1, 44-52, 1958.

Kostrov, B. V., Seismic moment and energy of earthquakes, and seismic flow of rock
(translated to English), Izv. Earth Physics, 1, 23-40, 1974.

Orowan, E., Mechanism of seismic faulting in rock deformation, Geol. Soc. Am. Mem.,
79, 323-345, 1960.

Savage, J. C., and J. B. Walsh, Gravitational energy and faulting, Bull. Seismol. Soc. Am.,
68, 1613-1622, 1978.

414
Melting, Fluid Pressurization, Lubrication

Brodsky, E. E., and H. Kanamori, The Elastohydrodynamic Lubrication of Faults, J.

Geophys. Res., J. Geophys. Res., 106, 16,357-16,374

Cardwell, R. K., D. S. Chinn, G. F. Moore, and D. L. Turcotte, Frictional heating on a

fault zone with finite thickness, Geophys. J. Roy. Astron. Soc., 52, 525-530, 1978.

Jeffreys, H., On the mechanics of faulting, Geol. Mag., 79, 291-295, 1942.

Kanamori, H., T. H. Anderson, and T. H. Heaton, Frictional melting during the rupture of
the 1994 Bolivian Earthquake, Science, 279, 839-842, 1998.

Kanamori, H., and T. H. Heaton, Microscopic and macroscopic physics of earthquakes,

AGU Monograph Series 120, "GeoComplexity and the Physics of Earthquakes", J.
B. Rundle, D. L. Turcotte, and W. Klein, Eds., 147-163, American Geophysical
Union, Washington D. C., 2000.

Lachenbruch, A. H., Frictional heating, fluid pressure, and the resistance to fault motion,
J. Geophys. Res., 85, 6097-6112, 1980.

McKenzie, D. P., and J. N. Brune, Melting on fault planes during large earthquakes,
Geophys. J. R. astr. Soc., 29, 65-78, 1972.

Richards, P. G., Dynamic motions near an earthquake fault: a three-dimensional solution,

Bull. Seismol. Soc. Am., 66, 1-32, 1976.

Sibson, R. H., Interactions between temperature and fluid pressure during earthquake
faulting -- A mechanism for partial or total stress relief, Nature, 243, 66-68, 1973.

Sibson, R. H., Generation of pseudotachylite by ancient seismic faulting, Geophys. J. R.

astr. Soc., 43, 775-794, 1975.

Sibson, R. H., Kinetic shear resistance, fluid pressures and radiation efficiency during
seismic faulting, Pure and Applied Geophysics, 115, 387-400, 1977.

415
Sibson, R. H., Power dissipation and stress levels on faults in the upper crust, J. Geophys.
Res., 85, 6239-6247, 1980.

Fault-zone Structure

Chester, F. M., and J. S. Chester, Ultracataclasite structure and friction processes of the
Punchbowl fault, San Andreas system, California, Tectonophysic, 295, 199-221, 1998.

Others

Ando, M., Source mechanisms and tectonic significance of historical earthquakes along
the Nankai trough, Japan, Tectonophysics, 27, 119-140, 1975.

Aviles, C. A., C. H. Scholz, and J. Boatwright, Fractal analysis applied to characteristic

segments of the San Andreas fault, J. Geophys. Res., 92, 331-344, 1987.

Beroza, G. C., and W. L. Ellsworth, Properties of the seismic nucleation phase,

Tectonophysics, 261, 209-227, 1996.

Brudy, M., M. D. Zoback, K. Fuchs, F. Rumell, and J. Baumgartner, Estimation of the

complete stress tensor to 8 km depth in the KTB scientific drill holes: Implications for
crustal strength, J. Geophys. Res., 102, 18,453-18,475, 1997.

Ellsworth, W. L., and G. C. Beroza, Seismic evidence for an earthquake nucleation

phase, Science, 268, 851-855, 1995.

Kanamori, H., The state of stress in the Earth's lithosphere, in Phys. Earth's Int., Course
LXXVIII, edited by A. M. Dziewonski and E. Boschi, pp. 531-554, North-Holland Pub.
Co., Amsterdam, 1980.

Wesnousky, S. G., The Gutenberg-Richter or Characteristic Earthquake Distribution,

Which is it?, Bull. Seismol. Soc. Am., 84, 1940-1959, 1994.

416
Ge 162 9. Plate Motion and Great Earthquakes

Earthquakes occur in the Earth's crust and mantle due to stresses caused by global

plate motion. The actual pattern of stress distribution is probably very complex, but we

expect that the activities of great and large earthquakes must reflect the global plate

motion.

417
 The world greatest earthquakes occur at subduction zones (e.g., 1960 Chilean

earthquake, and the 1964 Alaskan earthquake), but not every subduction zone has

experienced a great earthquake (e.g., the Marianas, the Tonga-Kermadec). It is possible

that the length of earthquake catalog is too short to be representative of long-term

seismicity. With this caveat in mind, we investigate the level of seismic activity and

plate motion. Ideally, the seismic activity along a subduction zone should be defined by

the energy release per unit length along the subduction zone, and unit time, i.e.,

418
 1 L T
e=
LT ∫ ∫
0 0
ER dldt

where L and T are the length of the subduction zone and the time period involved,

respectively.

Unfortunately, the available seismic record is too short to compute this. So, we

take the magnitude, M w , of the largest earthquake that occurred in a particular

subduction zone as a parameter that represents e for that subduction zone. Then , it is

reasonable to assume that

Mw ∝V

where V is the convergence rate. However, the plot of M w versus V does not show any

obvious trend. This suggests that other factors may be controlling seismicity. Another

potentially relevant parameter is the age, T, of the subducting plate. However, no

obvious negative correlation is seen between M w and T.

419
 Then, we can try a 3-parameter regression between M w , V and T. The result is

shown in the following figure. The horizontal axis shows the observed M w and the

vertical axis shows M w predicted by the regression relation.

420
(Ruff, L., and H. Kanamori, Seismicity and the subduction process, Phys. Earth Planet.
Inter., 23, 240-252, 1980)

If this regression is valid, this provides a useful method for assessing the seismic

potential of subduction zones for which no great earthquake has occurred. This pattern

suggests that the subduction zones where a relatively young plate is subducting at a

421
relatively fast rate are more likely to have great earthquakes, and those with an old plate

subducting at a moderate rate are less likely to have great earthquakes. The end-member

subduction zones are the Chilean type and the Mariana type, shown below.

(Uyeda, S., and H. Kanamori, Back-arc opening and the mode of subduction, J. Geophys.
Res., 84 (B3), 1049-1061, 1979)

422
 Another interesting implication of this correlation is the seismic potential of the

Pacific Northwest (i.e., Oregon-Washington coast). The Juan de Fuca plate is subducting

beneath the states of Oregon and Washington. The background seismicity there is very

low, as shown below, and until mid 1980's, it was generally believed that the seismic

potential in the Pacific Northwest is low (i.e., great earthquakes are unlikely). However,

the age of the Juan de Fuca plate is very young, about 10 My, and it is subducting at a

rate of 3 cm/year. Thus, in view of the regression relation shown above, one would

expect a large, M w =8.5 to 9, earthquake there. This suggestion motivated the interest of

geologists who started extensive investigation for finding palaeo-seismological evidence.

Geological evidence for regional submergence and evidence for large tsunami which

occurred in 1700 [Satake et al., 1996] now seem to have convinced most people, which

seems to have led to upgrading of building code in the area. This is a good example in

which seismological study, even if it is poorly constrained, can be useful if it is followed

up by investigations from different disciplines.

423
(Heaton, T., and H. Kanamori, Seismic potential associated with subduction in the
northwestern United States, Seismol. Soc. Am. Bull., 74 (3), 933-941, 1984)

424
(see, Atwater, B. F., and others, Summary of coastal geologic evidence for past great
earthquakes at the Cascadia subduction zone, Earthquake Spectra, 11, 1-18, 1995)

425
(R. S. Yeats, Living with Earthquakes in the Pacific Northwest, Oregon State University
Press, 1998)

References

Brune, J., Seismic moment, seismicity, and rate of slip along major fault zones, J.
Geophys. Res., 73, 777-784, 1968.

Satake, K., Shimazaki, K., Tsuji, Y. and Ueda, K., Time and size of a giant earthquake in
Cascadia inferred from Japanese tsunami records of January 1700, Nature, 379, 246-249,
1996.

426
427