Essentials of Geophysics

Chapter 1
The Earth in the Solar

System
1.1 Solar System Formation, Accretion, and the
Early Thermal State of the Earth
To understand the composition and early evolution of the Earth it is necessary
to consider as far back as the formation of our solar system. Solar system
formation was a complex process that is not well understood because of the
lack of data and the vast physical and chemical complexities of the process.
However, there are certain key parameters that we do know. As discussed in a
later lecture, we know from the study of meteorites the age of the solar system
and its initial composition. And, comparatively speaking, we know much about
the nature of the present-day solar system. In addition, we have observations of
old and young stars that inform us about the life cycle of the sun. The goal is to
use all the information in combination with the laws of physics and chemistry
to ll in the blanks between the initial state and the present state of the solar
system, and to consider what this means for the constitution and initial state
of the Earth.
1.2 Rotation and Angular Momentum

All the planets revolve in the same direction around the sun, and in practically
the same plane. For the most part they also rotate in the same direction about
their own axes, although there are notable exceptions, such as Venus. The gravitational collapse of molecular clouds is widely believed to lead to star formation
and it is likely that our solar system condensed from a collapsed, rotating cloud
of gas and dust. Rotating disks of material are ubiquitous in space, occurring
all the way from planetary to galactic scales. A rotating disk is the signature of
a self-gravitating system that has contracted in radius and amplied its angular
3
CHAPTER 1. THE EARTH IN THE SOLAR SYSTEM
velocity in order to preserve its total angular momentum. In a rotating protostar the gravitational attraction everywhere will be towards the center of mass.
But the centrifugal force will be directed normal to the axis of rotation. The
resolved force vector will move gas and dust nearer to the median plane as the
cloud contracts. This process leads to the disk shape, which dissipates energy
and minimizes collisions.
One of the more interesting boundary conditions is the present distribution
of angular momentum. Consider a planet of mass m that orbits a central body
of mass M , whose position with respect to the central body can be described
by a vector r.
The orbital angular momentum (L) of the planet can be written
L = mr2 = mr2
d
,
dt
(1.1)
where r is distance, m is the mass, is the angular velocity (=d/dt), and is

the angle with respect to a xed direction in the orbit plane. It can be shown
that
r2
L
dS
d
=
=2 ,
dt
m
dt
(1.2)
where S is the area swept out by r. Then

dS
L
=
,
dt
2m
(1.3)
which is a statement of Keplers second Law of Motion: the line between

a planet and the sun sweeps out equal areas in equal periods of time. Equation
(1.3) is a statement of conservation of angular momentum. The total planetary energy (E), which is the sum of the kinetic and potential contributions1 ,
E=
mGM
1
= constant
mv 2 +
2
r
(1.4)
where
2
v =
dr
dt
d
+ r
dt
2
(1.5)
is the planetary velocity, is also conserved. It is possible to rewrite (1.1) in

terms of the mass of the sun and doing so yields
L = mr2 = (GMs )1/2 mr1/2 .
(1.6)
By integrating (1.6) over all the planets we nd that while the sun contains
99.8% of the mass of the solar system, it has only about 1% of the angular
1 Notice that this treatment does not follow the conventional denition of gravitational
potential as used in later chapters, U = GM
.
r
1.3. THE SUN
momentum. About 60% of the angular momentum of the solar system is associated with the orbit of Jupiter alone. Most models suggest that the protosun
was rotating more rapidly than at present. Helioseismological results show that
deeper parts of the sun rotate faster than the surface. The deep solar interior,
which has not yet been probed, may hold the record of that bodys relic rotation. Solar system evolution models must show how the protosuns angular
momentum gets transported outward. Most models invoke magnetic and gravitational torques that spin down the sun and spin up the planets. Magnetizations
of meteorites are consistent with this idea. The transfer of angular momentum
could have contributed to the chemical fractionation of the solar system, since
an outwardly migrating magnetic eld would aect the ionized plasma but not
condensed particles, which couple to the eld only by viscous drag. Thus higher
temperature condensates would remain in the inner part of the solar system
and more volatile constituents would be transferred outward. In fact this is
observed.
1.3
The Sun
Stellar Evolution: The Hertzsprung-Russell Diagram

A common method of characterizing stars is the Hertzsprung-Russell (H-R)
diagram, which is a plot of absolute magnitude or luminosity versus eective
(blackbody) temperature. It is traditional to plot eective temperature from
high to low on the abscissa, and luminosity from dim to bright along the ordinate. For two stars with the same eective temperature, more light will come
from the larger star than the smaller; hence the largest stars are at the top of
an H-R diagram.
As each star proceeds through its life cycle it moves around on the H-R
diagram. While we cant observe the life cycle of a single star, we can search
through the current snapshot of our galaxy and nd stars at all stages of
evolution. Making an ensemble H-R plot reveals that many stars fall along a
single line called the main sequence. Stars on the main sequence are in a
relatively steady state of hydrogen burning in their cores, as is the present sun.
An average G-type (yellow) star like our sun is thought to have a lifetime (i.e.
a residence time on the main sequence) of about 10 billion years.
The T-Tauri Stage

The conspicuous absence of gas between the planets in the solar system must be
explained in any model of solar system formation. Before a new star reaches the
main sequence it goes through a pre-main sequence evolution of gravitational
collapse from a protostellar nebula. Our best information about this stage
comes from studying a class of young stars called T Tauri stars. T Tauri
stars are thought to be still contracting and evolving, and typically less than
one million years old. They are typically 0.2 to 2 solar masses in size, and they
show evidence of strong magnetic activity. Some T Tauri stars have spectra that
include forbidden lines, which occur in low-density gas and are the signature
of a gaseous nebula. Rapid uctuations in ultraviolet and x-ray emissions are
common. They also tend to show strong infrared emission and have spectra
with silicon lines indicating that they are surrounded by dust clouds.
T Tauri stars are associated with strong solar winds and high luminosities.
It is thought that our sun probably passed through a T Tauri stage in its early
evolution, and that the volatile elements in the inner solar system were blown
away during this stage.
1.4
Planetary Formation
Condensation and Cooling

The most widely accepted cosmogonical (formation) theory is that of V. Safronov,
who was the rst to hypothesize that the solar system initially accreted from
a nebular cloud that evolved from a sphere to a disk. While details of solar
system formation models dier, a common premise is that the planets formed
from particle growth in an initially tenuous dust-gas nebula. The mechanism to
trigger the initial collapse of the nebula has been argued and hypotheses range
from uniform gravitational collapse, to galactic spiral density waves, to catastrophic suggestions such as a supernova in the solar neighborhood. A supernova,
though a low probability event, is supported by the discovery of micro-diamonds
in cosmic dust. These imply that the solar system environs achieved high pressures due to passage of severe shock waves that would accompany only an event
of this intensity. The problem with the supernova hypothesis is that it would
imply that solar system formation is not a common phenomenon, which runs
contrary to current thought.
There are a number of scenarios for planetary growth. It is possible that
the planets accumulated from small moon-sized bodies, called planetesimals,
by infrequent encounters. Or instead accumulation may have occurred from
groups of bodies that collectively became gravitationally unstable. It is not clear
whether planetary accumulation occurred in a gaseous or gas-free environment.
In a gaseous nebula temperatures tend to be homogeneous, but as gas clears
due to the solar wind and condensation into dust grains the opacity of the
nebula decreases signicantly. During this time the system establishes a large
temperature gradient.
It is generally accepted that the planets accreted from a nebula with a composition similar to that of the sun, i.e., made mostly of hydrogen. The slowlyrotating nebula had a pressure and temperature distribution that decreased radially outward. The density of the nebula was probably not very great. Model
estimates of typical pressures in the vicinity of Earths orbit generally fall in
the range of 10-100 Pa but these are not very well constrained. The disk must
have cooled primarily by radiation, condensing out dust particles that were
initially of composed of refractory elements. These high temperature conden-
1.4. PLANETARY FORMATION
sates rst appear at temperatures of 1600 1750K and consist of silicates

oxides and titanates of calcium and aluminum, such as Al2 O3 , CaTiO3 , and
Ca2 Al2 Si2 O7 and refractory metals such as those in the platinum group. These
minerals are found in white inclusions in the most primitive class of meteorites,
the Type III carbonaceous chondrites, discussed in a later lecture. Metallic
iron condenses out next, followed by the common silicate materials forsterite
(an olivine) and enstatite (a pyroxene). Iron sulde (troilite; FeS) and hydrous
minerals condense at temperatures of 700 -800K. Volatile materials, most notably H2 O and CO2 condense out at 300 -400K. Planets that contain these
substances were accumulated from material that condensed in this temperature
range, which provides some clue about the early thermal structure of the solar
nebula.
Time scales for the condensation of gas to dust, of accumulation of dust to
planetesimals, and of accretion of planetesimals to planets and moons are also
not well constrained. If cooling occurred slowly in comparison to other processes
then planets would have formed during the cooling process and could have
accreted inhomogeneously. If instead cooling occurred rapidly, then the planets
would have formed from cold, generally homogeneous material. Homogeneous
accretion models are favored, with planetary dierentiation thought to be mostly
accomplished in the early stages after accretion.
Accretion
The process or processes that were responsible for the accumulation of dust and
small particles into planetesimals is a matter of debate. Sticking mechanisms
such as electrostatic attraction and vacuum welding have been suggested. But
as material accumulates, more planetesimal surface area is available for adding
more material so the process accelerates. When planetesimals reach sizes of
order 102 km gravitational attraction begins to dominate and accretion becomes
dominated by that force. In the planetesimal accretion stage collisional velocities
are a key consideration. If relative velocities between planetesimals are too low,
then planetesimals will fall into nearly concentric orbits. Collisions will be low
probability events and planets will not grow. Whereas if relative velocities
between planetesimals are too high, fragmentation rather than accumulation
will occur, and again planets wont grow. Safronov used scaling arguments
concerning energy dissipation during collisions and an assumed size distribution
of planetesimals to suggest that the mutual gravitation causes relative velocities
to be somewhat less than the escape velocities of the largest bodies. By his
estimation the system should regulate itself in a way to favor the growth of
large planetesimals. If this idea holds in a general sense, then solar systems
should form with a relatively small number of large planetary bodies rather
than with many small bodies. Monte Carlo simulations bear this idea out.
1.5
Early Thermal State of the Earth
Accretional Heating
As planetesimals accrete into moons and planets a signicant amount of energy
must be released, much of which will be converted to heat. Various theories
place the time for the accretion of Earth from 105 -108 years, which was very
rapid indeed in comparison to the age of the solar system. If accretion occurred
rapidly, then not much cooling could have occurred between collisions.
To determine the amount of heating associated with accretion, it is necessary
to take an inventory of the various sources of energy in the system. These include
the kinetic energy of impacting projectiles, the potential energy of infalling
material to the planetary surface and the thermal energy. For simplicity we
will begin by assuming that accretion occurs suciently rapidly such that the
process is adiabatic, i.e. with no heat lost from the accreting planets surface.
The total energy per unit mass of accreted material is simply a sum of the
change in kinetic and potential contributions2 :
Cp T =
GM
1 2
v vp2 +
2
R
(1.7)
where T is the temperature change, Cp is specic heat, v is the absolute

velocity of the approaching projectile, vp is the planetary velocity, (v)2 =
2
v
vp2 is the relative impact velocity, G is the universal constant of gravitation,
M is the mass of the planet, R is the planetary radius and
GM
= gR,
R
(1.8)
where g is the gravitational acceleration at the planetary surface. It is reasonable

to assume that the impact process is not perfectly ecient and that only a
fraction h of the total energy will be converted to heat. Taking this into account
and substituting ( .5) we may write
1 2
2
Cp T = h
(1.9)
v vp + gR .
2
This expression provides an upper limit of the increase in temperature that
could occur during accretion. In practice the potential energy term dominates
(1.9). But this expression isnt very realistic because it doesnt allow for cooling.
So we next consider the additional complication that heat is lost from the
system by cooling at the surface. It is possible to write a balance between the
gravitational potential energy of accretion, the heat lost by radiation, and the
thermal energy associated with heating of the body. This causes the problem
to become time dependent:
GM (r)
dr = T 4 (r) Tb4 dt + Cp [T (r) Tb ] dt
r
(1.10)
2 Notice that this treatment does not follow the conventional denition of gravitational
potential as used in later chapters, U = GM
.
r
1.5. EARLY THERMAL STATE OF THE EARTH
where M (r) is the mass of the accumulating planet, is the density of accreting
material, is emissivity, is the Stefan-Boltzmann constant, Tb is the radiation
equilibrium (blackbody) temperature, and t is time. In reality there will also be
energy associated with latent heats of melting and vaporization that are ignored
here. Temperature increases associated with the accretion of the the terrestrial
planets from numerical solutions to (1.10) require rapid accretion times, 103
to 104 years for Earth, to exceed the melting temperature. These time scales
are less than suggested by accretion models and would suggest that accretional
heating is not very important for Earth or the other terrestrial planets. But
it is necessary to consider in the most realistic sense possible the importance
of radiation in ridding the planet of heat. Radiative temperature loss goes
as T 4 and so is highly ecient in the sense that the planetary surface cools
quickly. But if an impact site becomes buried by ejecta from fall-back or from
nearby impacts, the surface would be covered. In this situation the outer part
of the planet is hotter than the interior and thermal convection is prohibited.
The only way to rid the planet of heat is to conduct it to the surface where
it can be radiated away. Conduction is a much less ecient heat transport
process and so accretional heat would be retained longer if that mechanism
dominated. If accretional energy is buried deeply enough to prohibit thermal
radiation from the surface, then temperature increases of order 2000 can be
attained for planets that accrete in times suggested by models (106 -107 years).
But even if accretion did cause the near surface of the Earth to melt the process
does not explain the earliest heating of the Earths deep interior, which occurred
through the process of dierentiation.
Dierentiation
From the Earths moment of inertia (C/M R2 ), which will be discussed later,
we know that the Earth (and other terrestrial planets) have a radially stratied
internal density structure. The implied increases in density with depth are
greater than would be associated with simple self-compression due to an increase
of pressure with depth. This leaves compositional changes, and to a lesser extent
phase changes, to explain the observations. If the Earth accreted cold, then
there must have been a process of internal dierentiation to produce its radially
stratied density structure. Dierentiation from a homogeneous initial state to
a structure with a distinct core and mantle involves a change in gravitational
potential energy. The release of this energy was likely to have been an important
source of heat in some planetary bodies. It is believed that dierentiation would
have occurred early in planetary evolution after a period of radioactive heating
or in the last stages of impact accretion in which the temperature required
to melt iron is achieved at shallow depth. Molten iron separates out from its
silicate matrix and is denser than its surroundings and sinks by gravitational
settling. It is reasonable to assume that the separation and sinking time is short
compared to the time of heating. Also, the process is taking place in the interior
so to rst order surface heat loss may be neglected.
Under these assumptions it is possible to estimate the increase in temper-
10
Table 1.5: Temperature increase due to core formation

Planet
Core radius (km)
Energy released (J)
Mean temp increase

T ( K)
Earth
Venus
Mars
Mercury
Moon
3485
?
1400 2100
1840
< 400
1.5 1031
?
2 1029
2 1029
< 1 1027
2300
?
300-330
700
10
ature associated with core formation. We may calculate the change in gravitational potential energy associated with the instantaneous dierentiation of a
planet from a homogeneous state to a nal state with a core and mantle. We
shall assume that the total mass in the system remains constant. In addition,
we will neglect contributions from other eects such as phase changes, the latent
heat of melting, rotational kinetic energy (due to the change in moment of inertia), and strain energy. The gravitational potential energy () for a spherical
planet in hydrostatic equilibrium in which density is simply a function of radius
may be written
Gm
dm
r
=
0
(1.11)
where m = 4/3r3 is the mass of accreting spherical body, and dm = 4r2 dr.
Substituting (1.8) we nd
g(r)rdm.
(1.12)
We then re-arrange once again to integrate over the radius so that
= 4
g(r)(r)r3 dr.
(1.13)
In practice = (r) is determined from an empirically-derived equation of

state that relates density to pressure (i.e. depth). Equation (1.13) must be
evaluated numerically. Now assume that the change in gravitational potential
energy will be fully converted to heat. Then
= Cp T
or T =
.
Cp
(1.14)
Table 1.5 shows the mean temperature increase associated with instantaneous core formation for the terrestrial planets based on (1.13) and (1.14). Note
that for the Earth the increase in temperature is expected to have been great
enough to have produced extensive melting. So shortly after accretion the Earth
1.5. EARLY THERMAL STATE OF THE EARTH
11
would have been largely molten and vigorously convecting in the interior as a
consequence of dierentiation. For Venus the size of the core isnt known but if
it is similar to Earth (given that planets similar radius and mass), then Venus
also would have experienced signicant early melting when it formed its core.
Melting also probably occurred on Mercury. But for Mars and the Moon the
temperature increase is not great enough for melt generation, even taking into
account the considerable uncertainties in core radii. Core formation could not
have been a signicant heat source early in the evolution of these bodies.
Formation of the Moon

We discussed above the role of impacts in the Earths early heat budget from
the illustrative calculation of temperature increase due to accretional heating.
But after accretion there will continue to be impact infall as the planets sweep
up asteroidal debris. This is quite apparent from looking at the 4.6 BY-old
lunar highlands, which are saturated with impact craters formed during the
terminal bombardment. It is now thought that a massive post-accretional
impact was responsible for the formation of the Moon. The origin of the Moon
has been a long-debated topic. While moons around planets are common in the
solar system, Earths moon is somewhat unusual given its large size compared
to the primary. One might wonder then, whether special circumstances were
associated with lunar origin.
Traditional models for lunar formation included co-accretion (the Moon
formed near the Earth), capture (the Moon strayed too near to Earth and
became trapped in orbit), and ssion (the Moon formed by spinning o the
Earth during an early rapid rotational period). All of these models had serious
problems in explaining important features like the Moons bulk composition,
the angular momentum of the Earth-Moon system, etc.
The theory that is currently is favored is the giant impact hypothesis,
which has gained support from numerical simulations and is consistent with the
features above. In this scenario, shortly after accretion the Earth received a
glancing impact from a Mars-sized asteroidal body. Smoothed particle hydrodynamic simulations from independent groups at Harvard and the University of
Arizona have the same general features: The mantles of both the early Earth
and the impactor melted and vaporized and the core of the impacting body
wrapped around Earths core. Mantle material from Earth and the projectile
that was ejected re-condensed in orbit to form the Moon. This hypothesis is
able to explain the puzzling lack of iron in the Moon. If this event did indeed
occur then the Earth would have been largely melted by the event. Such a catastrophic occurrence must factor in to scenarios for the post-accretional evolution
of the Earth.
12
1.6
Radioactive Decay
Radioactivity was discovered by Henri Becquerel in 1896 and it ultimately had

profound implications for the evolution of the Earth. By that time sedimentary
layering in outcrops was somewhat understood, at least to the point where
it was known that observed sedimentary strata must have taken hundreds of
millions of years to accumulate. At that point the only known energy sources
available to the sun and Earth, i.e. the energy associated with gravitational
collapse, allowed a maximum age of 25 Ma. For about 3 decades geologists
debated whether to accept this age. The argument became moot due to the
discovery of radioactivity. From the point of view of the evolution of the Earth,
the discovery of radioactivity had two eects:
(1) it removed the short-term age limit of the Earth by providing a mechanism for long-term heating (here we are referring to the internal heat that drives
mantle convection and thus plate motions); and
(2) it provided a means of determining absolute dates for rocks.
The stability of elements with respect to decay is related to the relative
numbers of protons and neutrons. If the numbers of these are not approximately
equal then material is prone to decay. Elements with the same number of protons
but a dierent number of neutrons are called isotopes.
Radioactive decay occurs because some of the mass of an atom is held in
binding energy. If there is too much binding energy (and quantum mechanics
is required to assess what exactly constitutes too much) then the nucleus will
decay spontaneously to lower the energy state. Induced nuclear reactions, say
by bombarding large atoms with neutrons, may also be used to achieve the lower
energy state. Radioactive decay can occur by three classes of mechanisms:
alpha decay the escape of a helium nucleus,
beta decay the escape of an electron or positron, or
gamma decay the emission of gamma radiation.
Radioactive decay is described with a simple rate law. The change in the
total number N of radioactive particles over the time interval dt is therefore:
dN
= N,
dt
(1.15)
where the minus sign indicates that activity decreases with time, and N represents the average number of particles that decay per second. The constant is
based on the probability of a particular decay mechanism operating in an atom
of a given element. We may rewrite
dN
= dt,
N
(1.16)
and integrating both sides of (1.16) we nd

ln N = t + c,
(1.17)
1.6. RADIOACTIVE DECAY
13
where the constant c is found in the limit where t 0 to be ln No . Taking the

exponential of both sides we may write
N (t) = No et ,
(1.18)
where No is the initial number of radioactive particles. Equation 1.18 is the rate
law of radioactive decay. The half-life, T1/2 , which represents the time it takes
for half of the number of particles to decay, is found by setting N/No = 1/2
such that
N
1
= = eT1/2
No
2
T1/2 =
ln 2
0.69315
.
(1.19)
Note that the half-life represents an alternative way of expressing the decay
constant . Nuclear binding energies are very large and nuclei are so small that
radioactive decay rates are not signicantly aected by physical conditions on
Earth such as pressure and temperature. However, half-lives can be changed
slightly by changes in bonding energy. For example, solar wind studies have
shown that radioactive beryllium decays at slightly dierent rates on the sun
and Earth.
In principle, the experimentally-demonstrated accuracy of the simple expression (1.19) allows for the determination of the absolute ages of billion-year-old
rocks. However, in practice the initial concentration of the radioactive parent
element No is very often not known. We can more easily measure the concentration of the daughter product (D ), which is simply
D = No N .
(1.20)
We may substitute (1.18) for N to nd
D = No No et = No 1 et .
(1.21)
We want to eliminate No so we divide by (1.18) which gives
No 1 et
1 et
D
=
=
.
t
N
No e
et
(1.22)
D
= et 1.
N
(1.23)
or
Equation (1.23) can be used directly in the determination of ages if there is

no initial non-radiogenic daughter component, or if that initial component can
be estimated.
14
1.7
Radiometric Dating
The Rubidium-Strontium System

To illustrate the radiometric dating technique, consider the decay of the unstable
isotope of rubidium, 87 Rb, into the stable isotope of strontium, 87 Sr. This
system is particularly simple because the parent element only decays into one
type of daughter element, unlike 40 K, say, which decays into both 40 Ar and 40 Ca
. The Rb-Sr system is useful for dating old rocks because the decay constant
and half-life T1/2 for the Rb-Sr system are well suited for the purpose:
= 1.42 1011 yr1 ,
T1/2 = 48.8 109 yr .
(1.24)
Only a fraction of the rubidium that was present in the solar nebula has so far
decayed. If t is the time since some melting event reset the isotope ratios to
their high-temperature values, then by (1.18), the current amount of 87 Rb is
reduced from its initial amount 87 Rb0 by:
87
Rb =87 Rb0 e87 t .
(1.25)
The current amount of strontium, 87 Sr, is therefore increased from its initial
amount, 87 Sr0 , by:
87
Sr
=87 Sr0 +87 Rb0 87 Rb

=87 Sr0 +87 Rb(e87 t 1) .
(1.26)
(1.27)
To proceed with the dating, one uses a mass spectrometer to measure the
amounts of 87 Sr and 87 Rb present in each sample. Since dierent parts of a rock
will contain dierent concentrations of the unknown quantity 87 Sr0 , we must
normalize against another stable isotope with similar chemistry that occurs in
proportional concentrations, like 86 Sr. Dividing (1.9) by 86 Sr yields:
87
87
87
Sr
Rb 87 t
Sr
1) .
(1.28)
+
=
(e
86 Sr
86 Sr
86 Sr
0
The presence of initial daughter abundances also requires more than one
measurement of the parent/daughter ratio to obtain an age. Samples that have
dierent 87 Rb/86 Sr ratios can be plotted versus 87 Sr/86 Sr using (1.10). The
87
Rb/86 Sr ratio varies naturally from one mineral to another. For example it is
typically higher in plagioclase than in pyroxene, so a spread in the samples is
obtained by mineral separation. When plotted, the two ratios fall on a straight
line called an isochron (meaning equal time), which by 1.28 has a slope of
(et 1) t and a y-intercept of (87 Sr/86 Sr)0 . If the decay constant of the
radioactive parent is known, the isochron yields the age, t, of the rock.
The most useful decay systems for radiometric dating are Rubidium-Strontium
(Rb-Sr), Samarium-Neodymium (Sm-Nd), Potassium-Argon (K-Ar), ThoriumLead (Th-Pb), and the two Uran-ium-Lead (U-Pb) systems. As illustrated
1.7. RADIOMETRIC DATING
15
above, in order for a parent-daughter system to be useful, a non-radiogenic reference isotope of the daughter must be present for comparison. In addition,
the decay constant of the parent must be accurately known. The accuracy of
radiometric dating also depends on the extent to which the rock under study
has been a chemically closed system with respect to the parent and daughter
elements. If it has not been a closed system then the daughter/parent ratio
will not be solely due to radioactive decay, and the time information will be
corrupted.
The Uranium-Lead System

The U-Pb system is especially useful because only measurements of Pb are
required, and Pb tends to be reliable because it is not too mobile in rock. Also,
because of decades of nuclear research the decay constants for uranium are very
accurately known. Zircon crystals are resistant to uranium diusion and are
commonly used for this dating scheme. There are four isotopes of Pb: 204 P b,
206
P b, 207 P b, 208 P b. Only 204 P b does not have a radioactive progenitor, and
the decay schemes for the other three isotopes are:
238
U 206 Pb ,
235
207
232
208
Th
238
= 1.55 1010 yr1 ,
T1/2 = 4.5 By
(1.29)
Pb ,
235
= 9.85 1010 yr1 ,
T1/2 = 0.7 By
(1.30)
Pb ,
232
= 4.95 1011 yr1 ,
T1/2 = 14 By .
(1.31)
Using (1.8) and referencing to the non-radiogenic

206
238
206
Pb
U 238 t
Pb
=
+ 204
e
1 ,
204 Pb
204 Pb
Pb
0
207
207
235
Pb
U 235 t
Pb
=
+
e
1 .
204 Pb
204 Pb
204 Pb
0
204
Pb we nd
(1.32)
(1.33)
Now take the ratio of 207 Pb/204 Pb to 206 Pb/204 Pb:

207
207
Pb
Pb
204
235
204 Pb
Pb 0
U e235 t 1
,
=
206
206
238 U e238 t 1
Pb
Pb
204 Pb
204 Pb
0
(1.34)
and rewrite into an isochron equation:

207
Pb
=M
204 Pb
206
Pb
204 Pb
+B,
(1.35)
where the slope and y-intercept are:

U e235 t 1
= 0.613 ,
238 U e238 t 1
207
235
M=
B=
Pb
204 Pb
206
Pb
204 Pb
= 4.46 .(1.36)
0
16
The age information is contained in the slope, M , using only isotopes of Pb.
The value of 235 U/238 U is 1/137.88, and this ratio is very nearly constant in
all natural materials. To determine the initial lead ratios the standard practice
is to look to meteorites. Iron meteorites have virtually no uranium. The least
radiogenic lead found anywhere is in the Canyon Diablo meteorite. This is
dened to be primordial lead, the initial lead ratio in the solar nebula.
The Age of Earth

Using the decay constants for uranium, the 235 U/238 U ratio given above, and
the initial lead ratios from Canyon Diablo this gives a date of (4.54 0.03) 109
years. This is the time when isotopically homogeneous lead was isolated from
the solar nebula in various bodies with dierent U/Pb ratios. This is the best
estimate the age of the Earth. It is interesting that Earth and the meteorites
fall on the same lead-lead isochron, which is evidence that lead and uranium
were both isotopically homogeneous in the solar nebula before the accretion into
the planets.
The radiometric dating of crustal and oceanic rocks from Earths surface
indicates that most of the surface is less than 100 million years old. The oldest
rock was dated by Sam Bowring of MIT. It is an igneous rock and the age is
4.03 109 years. The surface of Earth is much younger than that of the other
solid planets for several reasons.
Erosion is ecient on Earth because of the abundance of liquid water and
because of the presence of a biosphere. Earth also renews its surface continuously through the action of plate tectonics, where new crustal material comes
to the surface at mid-ocean ridges and old crustal material plunges below the
surface at subduction zones. Given the great activity occurring on Earths surface, it is not at all surprising that terrestrial rocks are not the oldest rocks in
the solar system.
1.8
Radioactivity as a Heat Source
Radioactivity is an important source of heating in the early solar system and

in addition represents the major source of long-term heating of the Earth (and
other terrestrial planets). The important heat-producing radionuclides form two
classes: long-lived nuclides and short-lived nuclides.
Long-Lived Nuclides
The long-lived radionuclides, of which the most important are 238 U, 235 U,
232
Th and 40 K, are a primary source of heat over the span of Earth history. They
provide heat which drives present-day mantle convection. Long-lived radioactive
elements have combinations of valence states and ionic radius that prevent them
from being easily accommodated into the crystal lattices of the most common
silicate rocks. They are examples of lithophile elements, which preferentially
1.9. METEORITES AND THE BULK COMPOSITION OF THE EARTH 17

concentrate in the liquid phase; the majority of them are incorporated into the
rst few percent of a melt. For this reason, a signicant fraction of the Earths
radioactivity is concentrated in the continental crust.
Short-Lived Nuclides
The short-lived radionuclides may have been an important source of heat
responsible for the early melting of meteorites. They might also have provided
an early heat source for planets, depending on the time between nucleosynthesis
and planetary accretion. The most abundant of the short-lived radionuclides
is 26 Al, which decays with a half-life of 720,000 years to 26 Mg. The evidence
for 26 Al being an important source of heat in the early history of the solar
system comes from excess amounts of 24 Mg found in CAIs in the Allende meteorite. The isotope 26 Mg was enriched relative to the most common isotope
24
Mg compared to solar abundance. The heat-producing ability of this isotope
is such that solid objects a few km or greater would have been heated to melting if they formed with the ratio of 26 Al/27 Al implied to have been present in
Allende.
A major question is: how could 26 Al be incorporated fast enough into early
solar system objects to melt them? With such a short half-life, radioactive decay
begins to produce heat after a cosmically short period of time. The isotope 26 Mg
is only produced by decay of 26 Al, and 26 Al is only produced in supernovae.
This suggests that our solar system might have formed close to a supernova.
Another piece of information in support of the supernova hypothesis is the fact
that very small diamonds have been found in some meteorites. On Earth diamond forms at great depths due to very high pressures which contract carbon to
the closely-packed state, characterized by all covalent bonds. In space the pressures required to form diamond can only be achieved in a supernova. If the solar
system formed near a supernova it would solve the problem of the mechanism
that caused the protosolar cloud to collapse, as the shock waves that emanate
from supernovae would provide a natural mechanism for compression. However,
supernovae are rare events and if it is necessary to invoke the participation of
one it would imply that the formation of our solar system was a chance event.
1.9 Meteorites and the Bulk Composition of the

Earth
The chemical and mineralogical information contained in the various types of
meteorites yield some of our most important clues about the nature of the
early solar system and the early Earth. A wide range of information about
the chemical and physical state derives from analysis of the various classes of
meteorites. We review the various types of meteorites and their implications for
the Earths bulk composition.
18
1.10
Chondrites
Chondrites form the most abundant class of meteorites and as a group represent
primordial objects that are chemically similar to the sun. They are so named because they contain chondrules3 , which are primitive, glassy, silicate globules
up to a few millimeters in size. Chondrules were magnetized as independent
grains, and have never been found in a terrestrial rock. It is believed that the
chondrules condensed out of the protoplanetary nebula before being incorporated into a matrix that consists of crystalline silicate minerals and sometimes
grains or laments of iron-nickel alloy. The magnetization, spherical shape and
ne crystalline structure indicate that they were melted and very rapidly cooled,
perhaps on time scales as short as minutes. Chondrules have been preserved
because they were incorporated into parent bodies that were not large enough
to undergo dynamic processes that would modify or destroy them.
A comparison of elemental abundance in a chondrite versus elemental abundance in the suns photosphere, as determined by spectroscopy, yields an astonishing correspondence. The only elements that dont match well are the
most volatile elements, which tend to escape incorporation into a meteorite as
it cools, and lithium, which is depleted in the sun due to nuclear reactions.
The name chondrite has come to refer more broadly to any meteorite with a
chemical composition that is similar to that of the sun, and there are a number of subclasses, dened on the basis of their chemistry and their degree of
metamorphism, i.e., the modication of their structure and mineralogy due to
temperature and pressure.
Carbonaceous Chondrites
An important fall occurred in Chihuahua, Mexico, in 1969, when a large meteor was observed to come into the atmosphere many pieces. The rst piece
was found near a house in the small village of Pueblito de Allende. Following standard practice, all of the meteorite fragments that were recovered from
that fall are collectively named Allende. The Allende fall occurred just as the
Apollo program was swinging into full gear, and it gave scientists who were
preparing for the arrival of moon rocks an opportunity to practice techniques
for analysis of chemical composition on an extraterrestrial sample. Because of
its unique chemical constitution and the fact that there was plenty of it to go
around, analysis of the Allende meteorite has taught us much about the early
solar system.
Allende is a member of an important sub-class of chondrites referred to as
the carbonaceous chondrites. It is a Class III carbonaceous chondrite that
is the most primitive (i.e. least altered by heating or other metamorphism) of
that class and much of what we know about the bulk composition of the Earth
is based on that single meteorite.
3 From
the Greek word oo or chondros, meaning grain
1.11. SECONDARY PROCESSING
1.11
19
Secondary Processing
If we want to use chondrites and other meteorites as a probe of the early solar system, we need to have a good idea of the types of processes that modied these meteorites from their original forms. Igneous meteorites, like the
howardite-eucrite-diogenite suite (HED), have denitely experienced secondary processing. However, they do provide unique information about the
evolution of Earth and terrestrial planets, particularly basalt generation and
core formation. Chondrites may be the least altered class of meteorites, but
they too have undergone secondary processing. Most chondrites have experienced thermal metamorphism. The result is changes in texture and mineralogy, and possibly chemical composition. The temperatures necessary to cause
metamorphism are in the 400 to 1000C range for the relatively low pressures
encountered in small parent bodies. Possibly important heat sources are the
decay of short-lived radionuclides, electromagnetic induction, and accretion of
material. The least metamorphosed type III chondrites (like Allende) probably
carry the most information about the early solar system, but even these have
been eected somewhat by secondary thermal processing. Along with heat, the
chemical reactivity of water has played an important role in the secondary processing of some of the most compositionally primitive meteorites. This process,
called aqueous alteration, tends to replace the preaccretionary lithology with
new mineralizations, although the bulk chemistry is apparently preserved.
Along with internal processing, meteorites can be changed by exogenic processes like collisions. Violent impacts produce shock metamorphism of individual mineral grains, and also produce rocks called breccias that contain mixtures
of dierent previous rocks, just like the breccias found on the lunar highlands.
The study of breccias has provided information on the accretional growth and
processing of parent bodies. The eects of shock metamorphism have been seen
in all major groups of meteorites. It appears that collision-induced high speed
impacts took place before, during, and after the initial accretion and dierentiation of the parent bodies.
Meteorites contain information related to their long exposures to galactic
cosmic rays, solar radiation and the solar wind. It is possible to determine
how long a meteorite existed free of its parent body before it impacted Earth by
examining the cosmic ray damage. Noble gases are the most volatile elements in
meteorites, but they are nonetheless present in measurable quantities in virtually
all meteorites. Trapped noble gases are either solar or planetary. The
solar noble gases are actually implanted solar-wind or solar-are material, and
provide relatively direct information about the sun. The planetary noble gases
have elemental abundances similar to those found in Earths atmosphere.
1.12
Achondrites
The achondrites are igneous meteorites that lack water-bearing (hydrous) or

oxidized minerals. This class of meteorites includes the eucrites, diogenites
20
and howardites. Because they are basaltic in composition they are believed to
come from parent bodies that were large enough to have dierentiated (melted)
to produce a crust. The main belt asteroid Vesta is the best compositional
analog for the eucrites.
Meteorites from Mars

Shergottites are named after a meteorite that fell in 1865 in Shergotty, which
is located in the northeastern Indian state of Bihar, which borders Nepal. Unlike
the eucrites, their pyroxene and plagioclase mineralogy is strikingly similar to
terrestrial basalts. They also have small amounts of the hydrous mineral amphibole kaersutite, whereas eucrites show no evidence of water in their minerals, and
they have some magnetite, which contains iron in oxidized form (Fe+3 ), whereas
eucrites contain only reduced iron. Their pyroxene crystals are elongated and
arranged horizontally the way such crystals would accumulate after settling to
the bottom of a magma chamber. Such igneous rocks are called cumulates.
Two other types of cumulate meteorites have hydrous and oxidized minerals like
the shergottites, these are the nakhlites, which contain the dark-green to black
pyroxene mineral augite, and a unique meteorite that fell in Chassigny, France
called chassignite, which contains mostly olivine. Together these three types
of meteorites are referred to as the SNC (pronounced snick) meteorites. The
parent body for the SNC meteorites is conjectured to be Mars.
Meteorites from the Moon

Some meteorites have breccias with white clasts in darker matrix, like lunar
rocks. These lunar origin of some meteorites was established without contention because we have many lunar samples which provided close geochemical
and petrological matches to the meteorite samples. The identication of lunar
meteorites opened the door to dynamical studies that subsequently established
the possibility that meteorites can viably be deposited at Earth from Mars.
Ureilites
These meteorites contain olivine and pigeonite, and the matrix contains graphite
or diamond. The carbon content suggests a link with carbonaceous chondrites.
1.13
Irons and Stony-Irons
Iron and stony iron meteorites make up several percent of the meteorite population. They constitute 4% of the meteorites that fall to Earth but are the most
common nd since they look so much dierent than the Earths crustal rocks.
Stony iron meteorites consist of roughly equal parts of rock and iron, with
the rock component consisting of olivine, the most common mantle material,
and minor amounts of other silicate phases. Stony irons make up only about
1% of meteorites that fall to Earth.
1.13. IRONS AND STONY-IRONS
21
The metallic component of these meteorites is predominantly iron with nickel

in solid solution averaging usually about 10% but sometimes up to 20%. There
are also smaller amounts of sulde, graphite and occasionally silicate inclusions.
Within the iron there are two metal phases: the body-centered cubic () form
kamacite (5.5% nickel) and the face-centered cubic () taenite (variable, but
usually > 27% nickel)
These phases occur because iron and nickel form a solid solution when mixed
and are not completely miscible as they begin to cool. The iron and nickel are
structurally similar but not identical. At high temperatures they exchange
freely because the crystal lattice is expanded. But when cooling sets in their
slight dierences produce lattices with slightly dierent structures. At a point
the total energy of the system is minimized by segregating the elements into 2
separate lattices: one rich in iron and the other poor. To minimize the mismatch
where the lattices connect, newly formed lattices form in preferred orientations
called exsolution lamellae.
Approximately 75% of iron meteorites exhibit a crystal pattern called Widmanstatten structure, which is the term used for these exsolution lamellae.
The pattern is observed by taking a meteorite that is cut and polished and
dipping it in acid. Because this pattern forms when an iron-nickel alloy crystallizes, it is an indication that some asteroids were at least partially melted after
they formed. In fact, the details of the pattern tell the cooling history of the
meteorite parent body from which it was derived. From the iron-nickel phase
diagram we can see the evolution of the relative amounts of iron and nickel
that crystallized as the iron cooled. And from the variations in composition
across Widmanstatten structure boundaries it is possible to constrain the rate
of cooling.
Iron Meteorites and Planetary Cores

In the iron-nickel system, equilibrium is maintained at temperatures above
650C. Below 350 C crystal structures are frozen in. So the Widmanstatten structures yield the cooling rate in this temperature range. In the meteorites wide diusion boundaries correspond to slow cooling and narrow diusion
boundaries correspond to rapid cooling. In the meteorites there are iron-nickel
crystals that grown with lengths of up to several centimeters, which correspond
to rates of 0.4 40 /Ma. So the Widmanstatten structures yield cooling rates
of many millions of years. For a cooling periods in this range radii of meteorite
parent bodies in the range 100200 km are implied.
Identication of a signicant metallic group within the meteorites made it
the natural presumed component of dense planetary cores. It is known from
the masses of the planets that the interiors are (in most cases), after correcting
for the eect of self-compression, denser than rock. An iron-nickel alloy with
approximate meteoritic proportions is the leading candidate for the dense component. Shock wave experiments indicate that seismic wave velocities in the
Earths deep interior are consistent with a predominantly iron composition and
thus support the contention.
22
Table 1.14: Mean densities and moments of inertia for the

terrestrial planets
Planet
Bulk Density (kg m3 )
C/M R2
Mercury
Venus
Earth
Moon
Mars
5420
5250
5515
3340
3940
??
0.34 (inferred)
0.3335
0.391
0.366
The other piece of evidence for iron planetary cores comes from the process
of nucleosynthesis. Iron has the highest binding energy for nucleon and is thus
highly stable and produced in abundance in stellar evolution. The equilibrium
process (aka e-process) in stellar thermonuclear reactions breaks apart silicon
atoms and re-arranges them to convert silicon to heavier and more stable nuclei.
The most stable and thus most abundant element produced in the e-process is
iron.
1.14
The Terrestrial Planets
The table below shows a comparison of the mean densities and moment of
inertia factors (C/M R2 ) for the terrestrial planets. The mean densities are
uncompressed so they correct for self-compression and represent zero-pressure
densities. The moment of inertia, which will be derived later in the semester, is
a measure of the the extent to which mass is concentrated toward the center of a
body with a value of 0.4 representing a uniform, homogeneous sphere and lesser
values representing mass increasingly concentrated toward the center. This
parameter reects both self-compression and the presence of increasingly dense
material toward the center.
Note that Venus and Earth are quite similar, consistent with their similarity
in radius and mass. Mars has a lower bulk density and a higher moment of
inertia core indicating less iron and a smaller core. Mercury is denser and while
the moment of inertia hasnt been measured, it is thought based on the mass
that its core is nearly 80% of the planets radius. Mercury thus likely has a
large iron core. The Moon has a bulk density similar to that of Earths mantle
and a moment of inertia just slightly less than that of a uniform sphere. The
implied upper limit of core size is 350 km (out of a radius of 1738 km). Thus
the Moon has very little iron in the interior.
1.15
One-dimensional Earths structure
To a large extent, the radial stratication as inferred from seismology and constrained by astronomical and meteorite data represents the stable structure that
1.15. ONE-DIMENSIONAL EARTHS STRUCTURE
23
results from dierentiation processes as discussed by Maria Zuber. In addition

to using this structure for the introduction of some useful terminology this radial
structure serves as an important reference model for many geophysical investigations. The average properties are now fairly well known; it is fair to say
that the uncertainties in the average values are insignicant as compared to
the local and regional deviations from the reference value. Constraining and
understanding the aspherical variations in physical properties is the objective
of many geophysical studies. It involves the study of anomalies relative to a
reference values (e.g., gravity and heat ow anomalies). The deviations from
the reference values contain information pertinent to geodynamical processes.
Radial stratication of the Earth from seismic data

The essential data used to derive the depth variation of seismic wave speed in
these early days of seismology were the travel times of dierent types of seismic
waves. These can be determined from seismograms, which are records (analog
or digital) of ground motion due to earthquakes or man-made explosions (e.g.,
nuclear tests).
Some history: Until late last century most research relevant to seismology
was in fact done by physicists and mathematicians who loved to study elastic wave propagation (famous names that also contributed to seismology are
Navier, Poisson, Gauss, Rayleigh). Therefore, most of the theory was already available at around the time of birth of observational seismology (late
last century; installation of the rst seismometers for systematic earthquake
monitoring).
Following a rapid succession of discoveries in the early years of this century,
the major subdivision of the Earth in concentric shells was established about 60
years ago, in the mid- to late thirties, by the pioneering work of Jereys and
Bullen and by Gutenberg and Richter. The following dates give some idea
about the pace of developments:
1892- First record of earthquake (Japan)
1906- Oldham demonstrates existence of core from seismic data
1909- Mohorovicic discovers seismic interface that marks the crust-mantle
boundary (MOHO)
1912- Gutenberg estimates depth to core mantle boundary
1936- Lehman discovers existence of Inner Core
1939- early knowledge summarized in rst 1D Earth models, the famous
Jereys-Bullen tables (which are still surprisingly accurate)
1948- Bullen uses wave speed information + estimates of Earths moment
of inertia to determine average density as a function of increasing depth.
24

After that, main focus on determining 3D structure. Mid seventies: pioneering work at MIT (K. Aki, M. N. Toks
oz) and Harvard (A. Dziewonski)
results in rst realistic 3D models of seismic propagation speed in Earths
mantle.
How does this work? Just a brief account will do here: modern seismometers
measure the three spatial components of this ground motion over in a wide
frequency band and with a large dynamic range. What is important in these
records is (1) the large variation in frequency (body waves and surface waves)
and (2) the arrival of distinct seismic phases (such as P and S waves). For
imaging purposes the dierence in frequency controls the amount of detail that
can be investigated with certain seismic data (resolution) whereas the dierent
phases sample dierent parts of the Earths interior. Whats important for now
is how the travel times measured from such records can tell us something about
Earths radial structure.
Let us assume for the time being that we know where and when an earthquake occurred (for instance from local damage reports). One can then construct
so-called record sections, in which the seismograms are sorted according to the
distance of the station to the earthquake epicentre (location at the surface).
This distance (epicentral distance) is typically denoted by a [ ]. If many
data are available one can determine best-tting travel-time curves for dierent
seismic phases. (Note: in practice we do not know the earthquake location and
origin time (hypocentre) and the determination of source location and spatial
variations in wave speed are intimately connected, with some nasty trade-os).
The variation of the travel time as a function of distance (T() [s]) can
be used to construct models of the variation of wave speed with depth. Using
some simple physics one can also show that the knowledge of the wavespeed as
function of depth (in combination with constraints such as the average density
and the moment of inertia (I 0.33 M R2 for the Earth) can be used to deduce
the radial variation in density. The relevant equation, the Adams-Williamson
equation, is valid in regions where density is controlled by adiabatic compression
and will be derived later.
In general, the wave speed increases with increasing depth in the mantle due
to the eect of increasing
pressure on the bulk modulus () and rigidity ().
(Note: P wave speed = + 43 /; S wave speed = /).

An abrupt increase in both P and S wave speed occurs at a depth of about
10-40 km. This seismic discontinuity marks the crust-mantle interface and was
discovered by Mohorovicic (hence Moho discontinuity). The denition (and
thus the mapping) of the seismological Moho is primarily based on seismic wave
speeds and does not necessarily coincide with a petrologic moho (See Anderson,
1988, for discussion). This leads to a subdivision in terms of crust-mantle-core;
the subdivision in terms of crust-lithosphere-asthenosphere-mantle is based on
thermal or mechanical properties. The latter nomenclature is, unfortunately,
not unique and dierent scientists may mean dierent things, so beware!)
Between about 400 and 1000 km in depth the increase in wave speeds is larger
than expected from adiabatic compression. This observation (rst by Birch,
1.15. ONE-DIMENSIONAL EARTHS STRUCTURE
25
1952) has major implications for mantle dynamics and will be discussed in more
detail later. The increase in now generally believed to be due to isochemical
phase changes in the mantle silicates.
In the outer core, the P-wave speed decreases abruptly and this, in fact,
causes the shadow zone for P- waves that was used as one of the major arguments by Oldham (1906) and Gutenberg (1912) to infer from seismological data
the existence of a core and the depth to the core-mantle-boundary (CMB). The
core behaves as a uid (pointed out by Jereys in 1926 on the basis of tidal data,
i.e. later than the discovery of the CMB!) even on short time scales so that shear
waves cannot propagate through the (outer) core. The inner core (discovered
by Inge Lehman (Denmark) in 1936) is solid; the seismological evidence for this
will be discussed later. (Question: temperature increases monotonically with
increasing depth, so why is there an alternation between depth intervals where
rock is solid and where it behaves as a liquid?)
Composition
Despite centuries of geologic prospecting we have limited access to most parts of
the Earth. consequently our understanding of its bulk composition must come
from inference based on remote observation as well as the record from meteorites
and the solar atmosphere. The dominant non-volatile constitutents in the sun
are silicon [Si], magnesium [Mg] and iron [Fe]. The meteorites similarly are
dominated by these elements and their oxides.
Some of the internal boundaries of the Earth represent phase changes that
separate regions of the same bulk composition, others coincide with a change
in chemistry. Seismology helps but can often not uniquely determine variations
in temperature, composition, or density. There are often several mineral assemblages that have the same density or elastic properties. Using the cosmic
abundance of elements and results of laboratory experiments at high pressures
and temperatures with rock forming minerals and their analogs we have, however, arrived at the following broad picture of the average composition in each
of the concentric shells:
(Recall that the most abundant non-volatile elements in the solar system are
magnesium [Mg], iron [Fe], and silicon [Si].)
The crust, which is only 0.5% of the volume of the mantle, is rich in SiO2
and Al2 O3 (hence the old name SiAl) + CaO, Na2 O.
The composition of the mantle is roughly given by the following series of
solid solutions, with primary constituents SiO2 and MgO (hence the old name
SiMa; also MaFic, i.e. magnesium and iron (Fe) rich):
Important are the following two pairs:
(Mg , Fe1 )O + SiO2
2(Mg , Fe1 )O + SiO2
(Mg , Fe1 )SiO3 (orthopyroxene)

(Mg , Fe1 )2 SiO4 (olivine)
with = Mg/(Mg + Fe) approximately 0.9 (names used for olivine: =

1: forsterite; = 0: fayalite). The magnesium number is dened as 100
26
Mg/(Mg + Fe) and is often used in discussions of mantle rheology (a typical

value is 90). Another important upper mantle mineral is garnet, also a (Mg,Fe)
silicate. With increasing pressure in the mantle olivine transforms to spinel
(same composition) and post-spinel. Pyroxene + garnet transform to perovskite
in the lower mantle.
There is some debate whether the relative amount of iron increases in the
lower mantle (perhaps with a small, 2%, increase in intrinsic density (i.e. in
addition to the eect of adiabatic compression): such a small density contrast
may not be signicant from a dynamic point of view) and there is a growing consensus that the composition of the lower mantle is similar to that of the upper
mantle and transition zone. However, due to the higher pressures the crystallographic structure is dierent, more compact. The most important constituents
in the lower mantle are:
(Mg, Fe)O (Magnesio)w

ustite
(Mg, Fe)SiO3 (Perovskite)
There is also some debate about the exact ratio of pyroxene to olivine. The
solar abundance would favor a Mg/Si ratio close to 1, which would predict a
predominance of pyroxene, but the presence of Fe allows for more olivine and
the ratio that matters is the (Mg+Fe)/Si ratio. The thermodynamics of the
phase transformations in the two silicate systems (olivine and pyroxene) are
very important for our understanding of mantle dynamics, in particular for the
question as to whether material can ow across the transition zone into the
lower mantle. Because of their importance for mantel dynamics, phase changes
will be discussed in more detail later in this course.
The major outstanding problems pertinent to the mantle include the scale
of mantle convection, the eectiveness of mixing, and the survival of separate
reservoirs of compositional heterogeneity as inferred from isotope data. We will
come back to these exciting topics towards the end of the course.
The core (32% of the mass of the Earth; how much of its volume?) largely
consist of iron. Fe is the only heavy element with large enough solar abundance
to be a suitable candidate to form the heavy element required to explain the
large density of the Earths core. However, if the core would consist entirely
of metallic iron, its density would be higher than required from the moment of
inertia and the mean density of the Earth. There must, therefore, be a light
alloying element. Several of the light elements that are abundant in the solar
system are either too volatile or are insoluble with metallic iron. The most
likely candidate is oxygen (O) (which is insoluble with metallic iron at low
pressures and is thus not found in iron meteorites), although some would argue
for Sulfur (S) (which is found in iron meteorites), or Silicon (Si). (Stacey uses
80% O + 20% S, but S does not seem to be required). This is still an area of
active research; because of the extreme physical conditions there are hardly any
experimental data lot of theoretical (thermodynamics) studies of Equations
Of State (EOS) (See Anderson, 1988, for an introduction).
1.16. LATERAL HETEROGENEITY IN THE MANTLE
27
The inner core (IC) has a radius of about 1220 km the IC comprises less
than 1%(!) of the volume of the Earth, but it represents about 1.7% of the
Earths mass. Because of the small size of the IC, the uncertainty of its density
is relatively large. But within the uncertainty, the IC may simply be a frozen
version of the OC, and thus heavier since the lighter elements are selectively
rejected, with as major constituents Fe2 O, FeNiO, pure Fe or a Fe-Ni alloy.
Some compositional layering of the core is probable. (Anderson (1988) is a
good reference for core composition).
Because of the recent developments in understanding the structure of the
inner core (anisotropy, net rotation relative to the Earths mantle) and the
importance of the core for the Earths magnetic eld, the core will be discussed
in detail during this course (including reading assignments). (Some outstanding
problems: what maintains the IC-OC boundary? OC-lower mantle boundary?
What produces the energy to drive core convection that produces the magnetic
eld? Is there convection in the solid IC?).
1.16
Lateral heterogeneity in the mantle
Introduction
The outer core behaves as a liquid and its low viscosity cannot sustain lateral
variations in density and can thus be regarded as homogeneous for many practical purposes. The mantle, however, is probably heterogeneous at all length
scales. To a large extent this heterogeneity can be attributed to convective
circulation in the mantle and the recycling of (oceanic) lithosphere. The most
important dynamic processes will be discussed later. This heterogeneity causes
observed seismic data to dier from predictions from a simple radially stratied
reference model. Later in this course I will show how certain imaging techniques can be used to interpret these travel time residuals to map the aspherical
structure of the Earths interior.
Plate Tectonics forms an integral aspect of this convective system and a
relevant discussion of plate tectonics thus involves more than the essentially
kinematic concepts that were outlined in the 1960-ies. Therefore, plate tectonics, and its relationship with mantle convection, will be discussed in the second
half of this course. However, to facilitate communication it is useful to introduce some jargon and dene important concepts, processes, and their resulting
structures.
Plate boundaries as belts of high seismic activity

Once travel time table are established from a redundant seismological data set
they can be used to locate earthquakes. The pattern of seismicity is now well
established and denes the locations of the plate boundaries. Note that the
distribution of earthquakes within plate boundaries (or plate boundary zones)
varies, which gives important information about the deformation in certain re-
28
gions. On the basis of seismic belts about 12 major plates have been recognized.
Important aspects of kinematic plate-tectonic theory are that (1) the deformation within the plates (intraplate) is neglected; all deformation is assumed to
take place between plates (interplate) (if need be plates are subdivided into
smaller units until this condition is met). In other words, plates act as stress
guides, which is important for the understanding of the driving forces of plate
tectonics; (2) once new oceanic lithosphere is created it forms part of a rigid
plate; the plate may or may not contain continents; (3) in order to conserve the
total area of the Earths surface: the rate of lithosphere creation equals rate
of lithosphere destruction. The seismograms also contain information about
the physical rupture mechanism of earthquakes and can thus be used in the
classication of dierent types of plate boundary.
divergent plate boundaries Associated with normal faults (plates move
away from each other). Characteristic in Mid Ocean Ridge (MOR) systems; locations of sea oor spreading where new oceanic crust is being
created rather passively by means of decompression melting. Such boundaries are sometimes referred to as accreting margins but this is confusing
since an important mechanism of continental growth is by means of accretion of allochthonous terranes (accretionary wedges, accreted terranes)
onto Precambrian shields, for instance the western cordillera in Canada
and eastern Australia.
convergent plate boundaries Associated with reverse faults (plates move
towards each other). Sometimes referred to as destructive margins. Characteristic in circum-Pacic belt. Often accompanied with mountain building, subduction, deep seismicity, and arc volcanism. Diagnostic bathymetric feature: deep sea trenches.
transcurrent plate boundaries Associated with strike slip faults. Important in particular in association with ocean oor spreading transform
faults. There is a separate classication of transform faults based on
whether they connect ridges or trenches, but thats not so exciting. Most
common are ridge-ridge transforms. These are locations of many earthquakes since slip rate is twice as fast as spreading rate!
Lithosphere, asthenosphere, subduction zones

We will use the term lithosphere4 for the mechanically strong outer shell of
the earth, which contains the crust and part of the mantle. It is a thermal
boundary layer (TBL) through which heat is lost to the surface by conduction.
There is also a thermal denition, in which the base of the lithosphere coincides
with the 1300C isotherm (up to 100 km deep for oceans; at least twice that
much for some (parts of) continents), but the mechanically strong (elastic) part
of the lithosphere that can transmit stresses and support surface loads (e.g.,
4 o
or lithos=stone [Greek]
1.16. LATERAL HETEROGENEITY IN THE MANTLE
29
sea mounts) on geologic time scales is about half the TBL thickness (coinciding
with the 650 C isotherm). The cooling of oceanic lithosphere (+ relationship
between lithospheric thickness and heat ow with square root of its age since
formation at the MOR) is one of the classical aspects of plate tectonics that
will be discussed in the course of this term. In contrast, deformation in the
asthenosphere5 occurs more freely by means ductile creep on geologic time
scales. The asthenosphere may lubricate but does not participate directly in
tectonic motion, although it could accommodate some of the return ow. Heat
loss in asthenosphere and deeper mantle is controlled by convection.
As oceanic lithosphere moves away from the MOR it will cool and thicken
and become more dense due to thermal contraction. Eventually, it will become
gravitationally unstable and can, in principle, sink into the Earths interior at
ocean trenches (note: initiation of subduction is not trivial since even gravitationally unstable parts of the lithosphere can be retained at the surface by the
strength of the lithosphere). We will use the term subduction zone rather
loosely for the geographical region where plate convergence results in the descent of one plate beneath the other, typically the more mafic (Mg+Fe rich)
oceanic beneath the more acid, granitic continental plate. The result of this
process is a slab of subducted former oceanic lithosphere that sinks into
the mantle. The negative buoyancy of the downgoing slab is, in fact, one of the
most important forces of plate motion. The depth extent of slabs is still debated,
but that does not according to these denitions inuence the meaning of
a subduction zone. In the upper mantle and transition zone the slabs can be
delineated by earthquakes (the so called Wadati-Benio zones), the slab is said
to be seismogenic down to about 670 km depth (the actual depth to the deepest
earthquakes can be smaller depending on the thermal structure of the slab).
There is now convincing evidence to support the aseismic continuation of slabs
into the lower mantle.
or a-sthenes= not-strong [Greek]
Chapter 2
The Earths Gravitational

eld
2.1 Global Gravity, Potentials, Figure of the Earth,
Geoid
Introduction
Historically, gravity has played a central role in studies of dynamic processes in
the Earths interior and is also important in exploration geophysics. The concept
of gravity is relatively simple, high-precision measurements of the gravity eld
are inexpensive and quick, and spatial variations in the gravitational acceleration
give important information about the dynamical state of Earth. However, the
study of the gravity of Earth is not easy since many corrections have to be
made to isolate the small signal due to dynamic processes, and the underlying
theory although perhaps more elegant than, for instance, in seismology
is complex. With respect to determining the three-dimensional structure of the
Earths interior, an additional disadvantage of gravity, indeed, of any potential
eld, over seismic imaging is that there is larger ambiguity in locating the source
of gravitational anomalies, in particular in the radial direction.
In general the gravity signal has a complex origin: the acceleration due to
gravity, denoted by g, (g in vector notation) is inuenced by topography, aspherical variation of density within the Earth, and the Earths rotation. In
geophysics, our task is to measure, characterize, and interpret the gravity signal, and the reduction of gravity data is a very important aspect of this scientic
eld. Gravity measurements are typically given with respect to a certain reference, which can but does not have to be an equipotential surface. An important
example of an equipotential surface is the geoid (which itself represents deviations from a reference spheroid).
31
CHAPTER 2. THE EARTHS GRAVITATIONAL FIELD
32
The Gravity Field

The law of gravitational attraction was formulated by Isaac Newton (1642-1727)
and published in 1687, that is, about three generations after Galileo had determined the magnitude of the gravitational acceleration and Kepler had discovered
his empirical laws describing the orbits of planets. In fact, a strong argument
for the validity of Newtons laws of motion and gravity was that they could be
used to derive Keplers laws.
For our purposes, gravity can be dened as the force exerted on a mass m due
to the combination of (1) the gravitational attraction of the Earth, with mass M
or ME and (2) the rotation of the Earth. The latter has two components: the
centrifugal acceleration due to rotation with angular velocity and the existence
of an equatorial bulge that results from the balance between self-gravitation and
rotation.
The gravitational force between any two particles with (point) masses M
at position r0 and m at position r separated by a distance r is an attraction
along a line joining the particles (see Figure 2.1):
F = F = G
Mm
,
r2
(2.1)
or, in vector form:

F = G
Mm
Mm
(r r0 ) = G
r .
3
r r0
r r0 2
(2.2)
Figure 2.1: Vector diagram showing the geometry of the gravitational attraction.
where
r is a unit vector in the direction of (r r0 ). The minus sign accounts for
the fact that the force vector F points inward (i.e., towards M ) whereas the unit
vector
r points outward (away from M ). In the following we will place M at
the origin of our coordinate system and take r0 at O to simplify the equations
(e.g., r r0 = r and the unit vector
r becomes
r) (see Figure 2.2).
G is the universal gravitational constant: G = 6.673 1011 m3 kg1
s2 (or N m2 kg2 ), which has the same value for all pairs of particles. G must
not be confused with g, the gravitational acceleration, or force of a unit
2.1. GLOBAL GRAVITY, POTENTIALS, FIGURE OF THE EARTH, GEOID33
Figure 2.2: Simplied coordinate system.

mass due to gravity, for which an expression can be obtained by using Newtons
law of motion. If M is the mass of Earth:
F = ma = mg = G
and
g = g = G
M
.
r2
M
F
Mm
= G 2
rg=
r
2
m
r
r
(2.3)
(2.4)
The acceleration g is the length of a vector g and is by denition always

positive: g > 0. We dene the vector g as the gravity eld and take, by
convention, g positive towards the center of the Earth, i.e., in the r direction.
The gravitational acceleration g was rst determined by Galileo; the magnitude of g varies over the surface of Earth but a useful ball-park gure is g= 9.8
ms2 (or just 10 ms2 ) (in S.I. Syst`eme International dUnites units). In
his honor, the unit often used in gravimetry is the Gal. 1 Gal = 1 cms2 = 0.01
ms2 103 g. Gravity anomalies are often expressed in milliGal, i.e., 106 g or
microGal, i.e., 109 g. This precision can be achieved by modern gravimeters.
An alternative unit is the gravity unit, 1 gu = 0.1 mGal = 107 g.
When G was determined by Cavendish in 1778 (with the Cavendish torsion
balance) the mass of the Earth could be determined and it was found that the
Earths mean density, 5, 500 kgm3 , is much larger than the density of rocks
at the Earths surface. This observations was one of the rst strong indications
that density must increase substantially towards the center of the Earth. In
the decades following Cavendish measurement, many measurements were done
of g at dierent locations on Earth and the variation of g with latitude was
soon established. In these early days of geodesy one focused on planet wide
structure; in the mid to late 1800s scientists started to analyze deviations of
the reference values, i.e. local and regional gravity anomalies.
Gravitational potential
By virtue of its position in the gravity eld g due to mass M , any mass m has
gravitational potential energy. This energy can be regarded as the work
W done on a mass m by the gravitational force due to M in moving m from
rref to r where one often takes rref = . The gravitational potential U is
the potential energy in the eld due to M per unit mass. In other words, its
the work done by the gravitational force g per unit mass. (One can dene U as
34
either the positive or negative of the work done which translates in a change of
sign. Beware!). The potential is a scalar eld which is typically easier to handle
than a vector eld. And, as we will see below, from the scalar potential we can
readily derive the vector eld anyway.
(The gravity eld is a conservative eld so just how the mass m is moved
from rref to r is not relevant: the work done only depends on the initial and
nal position.) Following the denition for potential as is common in physics,
which considers Earth as a potential well i.e. negative we get for U:
U=
rref

g dr =
rref
GM
r dr = GM
r2
1
GM
dr =
r2
r
(2.5)
Note that
r dr = dr because
r and dr point in opposite directions.
Figure 2.3: By denition, the potential is zero at innity and decreases towards
the mass.
U represents the gravitational potential at a distance r from mass M . Notice

that it is assumed that U () = 0 (see Figure 2.3).
The potential is the integration over space (either a line, a surface or a
volume) of the gravity eld. Vice versa, the gravity eld, the gravity force per
unit mass, is the spatial derivative (gradient) of the potential.
g=
GM
r=
r2
r
GM
r

=
U = gradU U
r
(2.6)

Intermezzo 2.1 The gradient of the gravitational potential
We may easily see this in a more general way by expressing dr (the incremental
distance along the line joining two point masses) into some set of coordinates,
using the properties of the dot product and the total derivative of U as follows
(by our denition, moving in the same direction as g accumulates negative
potential):
dU
g dr
gx dx gy dy gz dz
(2.7)
By denition, the total derivative of U is given by:

dU
U
U
U
dx +
dy +
dz
x
y
z
(2.8)
Therefore, the combination of Eq. 2.7 and Eq. 2.8 yields:

g=
U U U
,
,
x y z
= grad U U
(2.9)
One can now see that the fact that the gravitational potential is dened to be
negative means that when mass m approaches the Earth, its potential (energy)
decreases while its acceleration due to attraction the Earths center increases.
The slope of the curve is the (positive) value of g, and the minus sign makes
sure that the gradient U points in the direction of decreasing r, i.e. towards the
center of mass. (The plus/minus convention is not unique. In the literature one
often sees U = GM/r and g = U .)
Some general properties:
The gradient of a scalar eld U is a vector that determines the rate and
direction of change in U . Let an equipotential surface S be the surface of
constant U and r1 and r2 be positions on that surface (i.e., with U1 = U2 =
U ). Then, the component of g along S is given by (U2 U1 )/(r1 r2 ) = 0.
Thus g = U has no components along S : the eld is perpendicular to
the equipotential surface. This is always the case, as derived in Intermezzo
2.2.
Since uids cannot sustain shear stress the shear modulus = 0, the
forces acting on the uid surface have to be perpendicular to this surface
in steady state, since any component of a force along the surface of the
uid would result in ow until this component vanishes. The restoring
forces are given by F = mU as in Figure 2.4; a uid surface assumes
an equipotential surface.
For a spherically symmetric Earth the equipotential would be a sphere and
36
Figure 2.4: F = mU provides the restoring force

that levels the sea surface along an equipotential surface.
g would point towards the center of the sphere. (Even in the presence of
aspherical structure and rotation this is a very good approximation of g.
However, if the equipotential is an ellipsoid, g = U does not point to
r = 0; this lies at the origin of the denition of geographic and geocentric
latitudes.)
Using gravity potentials, one can easily prove that the gravitational acceleration of a spherically symmetric mass distribution, at a point outside
the mass, is the same as the acceleration obtained by concentrating all
mass at the center of the sphere, i.e., a point mass.
This seems trivial, but for the use of potential elds to study Earths
structure it has several important implications:
1. Within a spherically symmetric body, the potential, and thus the
gravitational acceleration g is determined only by the mass between
the observation point at r and the center of mass. In spherical coordinates:
G
g(r) = 4 2
r
(r )r2 dr
(2.10)
This is important in the understanding of the variation of the gravity

eld as a function of radius within the Earth;
2. The gravitational potential by itself does not carry information about
the radial distribution of the mass. We will encounter this later when
we discuss more properties of potentials, the solutions of the Laplace
and Poisson equations, and the problem of non-uniqueness in gravity
interpretations.
3. if there are lateral variations in gravitational acceleration on the surface of the sphere, i.e. if the equipotential is not a sphere there must
be aspherical structure (departure from spherical geometry; can be
in the shape of the body as well as internal distribution of density
anomalies).

Intermezzo 2.2 Geometric interpretation of the gradient
Let C be a curve with parametric representation C( ), a vector function. Let
U be a scalar function of multiple variables. The variation of U , conned to the
curve C, is given by:
dC( )
d
[U (C(t))] = U (C(t))
dt
dt
(2.11)
d
[U (C( ))] will be zero.
Therefore, if C is a curve of constant U , dt

Now let C( ) be a straight line in space:
C( ) = p + at
(2.12)
then, according to the chain rule (2.11), at t0 = 0:
d

[U (C( ))]
= U (p) a
dt
t=t0
(2.13)
It is useful to dene the directional derivative of U in the direction of a at

point p as:
DA U (p) = U (p)
a
a
(2.14)
From this relation we infer that the gradient vector U (p) at p gives the
direction in which the change of U is maximum. Now let S be an equipotential
surface, i.e. the surface of constant U . Dene a set of curves Ci ( ) on this
surface S. Clearly,
d
dCi

(t0 ) = 0
[U (Ci ( ))]
= U (p)
dt
dt
t=t0
(2.15)
for each of those curves. Since the Ci ( ) lie completely on the surface S, the
dCi
(t0 ) will dene a plane tangent to the surface S at point p. Therefore, the
dt
gradient vector U is perpendicular to the surface S of constant U . Or: the
eld is perpendicular to the equipotential surface.
In global gravity one aims to determine and explain deviations from the
equipotential surfaces, or more precisely the dierence (height) between equipotential surfaces. This dierence in height is related to the local g. In practice
one denes anomalies relative to reference surfaces. Important surfaces are:
Geoid the actual equipotential surface that coincides with the average sea level
(ignoring tides and other dynamical eects in oceans)
(Reference) spheroid : empirical, longitude independent (i.e., zonal) shape
of the sea level with a smooth variation in latitude that best ts the geoid
(or the observed gravity data). This forms the basis of the international
gravity formula that prescribes g as a function of latitude that forms the
reference value for the reduction of gravity data.
38
Hydrostatic Figure of Shape of Earth : theoretical shape of the Earth if

we know density and rotation (ellipsoid of revolution).
We will now derive the shape of the reference spheroid; this concept is very
important for geodesy since it underlies the denition of the International Gravity Formula. Also, it introduces (zonal, i.e. longitude independent) spherical
harmonics in a natural way.
2.2 Gravitational potential due to nearly spherical body

How can we determine the shape of the reference spheroid? The attening of
the earth was already discovered and quantied by the end of the 18th century. It was noticed that the distance between a degree of latitude as measured, for instance with a sextant, diers from that expected from a sphere:
RE d, with RE the radius of the Earth, 1 and 2 two dierent
RE (1 2 ) =
latitudes (see Figure 2.5).
Figure 2.5: Ellipticity of the Earth measured by the

distance between latitudes of the Earth and a sphere.
In 1743, Clairaut1 showed that the reference spheroid can also be computed
directly from the measured gravity eld g. The derivation is based on the
computation of a potential U (P ) at point P due to a nearly spherical body, and
it is only valid for points outside (or, in the limit, on the surface of) the body.
The contribution dU to the gravitational potential at P due to a mass element dM at distance q from P is given by
dU =
G
dM
q
(2.16)
Typically, the potential is expanded in a series. This can be done in two

ways, which lead to the same results. One can write U (P ) directly in terms
of the known solutions of Laplaces equation (2 U = 0), which are spherical
harmonics. Alternatively, one can expand the term 1/q and integrate the resulting series term by term. Here, we will do the latter because it gives better
1 In
his book, Th
eorie de la Figure de la Terre.
2.2. GRAVITATIONAL POTENTIAL DUE TO NEARLY SPHERICAL BODY39
Figure 2.6: The potential U of the aspherical body iscalculated at point P , which is external to the mass M = dM ;
OP = r, the distance from the observation point to the
center of mass. Note that r is constant and that s , q, and
are the variables. There is no rotation so U (P ) represents
the gravitational potential.
understanding of the physical meaning of the terms, but we will show how these
terms are, in fact, directly related to (zonal) spherical harmonics. A formal
treatment of solutions of spherical harmonics as solutions of Laplaces equation
follows later. The derivation discussed here leads to what is known as MacCullaghs formula2 and shows how the gravity measurements themselves are used
to dene the reference spheroid. Using Figure 2.6 and the law of cosines we can
write q 2 = r2 + s2 2rs cos so that
G
dU =
21 dM

s
2
r 1 + r 2 sr cos
(2.17)
We can use the Binomial Theorem to expand this expression into a power
series of (s/r). So we can write:

1+
s 2
r
s
r
21
cos
and for the potential:

U (P ) =
dU
V
2 After
James MacCullagh (18091847).
1 s 2
2 r
s
3 s 2 2
cos + h.o.t.
+
cos +
r
2 r
s
1 s 2
= 1+
cos +
3 cos2 1
r
2 r
+ h.o.t.
(2.18)
= 1
40

s
1 s 2
G
cos +
3 cos2 1 dM
1+
r
r
2 r

G
G
=
s cos dM
dM 2
r
r

s2 3 cos2 1 dM
3
2r
(2.19)
In Equation 2.19 we have ignored the higher order terms (h.o.t). Let us
rewrite eq. (2.19) by using the identity cos2 + sin2 = 1:
U (P ) =
G
r

dM
G
r2

s cos dM
G
r3
s2 dM +
3G
2 r3
s2 sin2 dM (2.20)
Intermezzo 2.3 Binomial theorem
(a + b)n
1
n(n 1)an2 b2
2!
an + nan1 b +
1
n(n 1)(n 2)an3 b3 + . . .
3!
for | ab | < 1. Here we take b =
s 2
r
cos and a = 1.
(2.21)

Intermezzo 2.4 Equivalence with (zonal) spherical harmonics
Note that equation (2.19) is, in fact, a power series of (s/r), with the multiplicative factors functions of cos():
U (P )
G
r
0
s
1

2
+ cos
Pl (cos )
l=0
s 1
r
l
s
r
dM
3
2
cos2
1
2
s 2
r
dM
(2.22)
In spectral analysis there are special names for the factors Pl multiplying (s/r)l
and these are known as Legendre polynomials, which dene the zonal surface
spherical harmonicsa .
We will discuss spherical harmonics in detail later but here it is useful to point
out the similarity between the above expression of the potential U (P ) as a power
series of (s/r) and cos and the lower order spherical-harmonics. Legendre
polynomials are dened as
Pl () =
1 dl (2 1)l
2l l!
dl
(2.23)
with some function. In our case we take = cos so that the superposition of
the Legendre polynomials describes the variation of the potential with latitude.
At this stage we ignore variations with longitude. Surface spherical harmonics
that depend on latitude only are known as zonal spherical harmonics. For
l = 0, 1, 2 we get for Pl
P0 (cos )
(2.24)
P1 (cos )
(2.25)
P2 (cos )
cos
1
3
cos2
2
2
(2.26)
which are the same as the terms derived by application of the binomial theorem.
The equivalence between the potential expression in spherical harmonics and
the one that we are deriving by expanding 1/q is no coincidence: the potential
U satises Laplaces equation and in a spherical coordinate system spherical
harmonics are the general solutions of Laplaces equation.
a Surface
spherical harmonics are at the surface of a sphere what a Fourier

series is to a time series; it can be thought of as a 2D Fourier series which can be
used to represent any quantity at the surface of a sphere (geoid, temperature,
seismic wave speed).
We can get insight in the physics if we look at each term of eq. (2.20)
separately:

1. G
is essentially the potential of a point mass M at O.
dM = GM
r
r
This term will dominate for large r; at a large distance the potential due
42
to an aspherical density distribution is close to that of a spherical body

(i.e., a point mass in O).

2. s cos dM represents a torque of mass distance,
which
also underlies

the denition of the center of mass rcm = r dM/ dM . In our case,
we have chosen O as the center of mass and rcm = 0 with respect to O.
Another way to see that this integral must vanish is to realize that the
integration over dM is essentially an integration over between 0 and
2 and that cos = cos( 2 ). Integration over takes s cos back
and forth over the line between O and P (within the body) with equal
contributions from each side of O, since O is the center of mass.

3. s2 dM represents the torque of a distance squared and a mass, which
underlies the denition of the moment of inertia (recall that for a homogeneous sphere with radius R and mass M the moment of inertia is 0.4
MR2 ). The moment of inertia is dened as I = r2 dM . When talking
about moments of inertia one must identify the axis of rotation. We can
understand the meaning of the third integral by introducing a coordinate
2
system
y, z so that s =(x, y, z), s2 = x2 + y 2 + z 2 so that
2 s 2dM =
2 x,
2
2
2
2
2
2
(y + z ) dM
(x + y + z ) dM = 1/2[

+ (x + z ) dM + (x + y ) dM ]
and by realizing that (y 2 + z 2 ) dM, (x2 + z 2 ) dM and (x2 + y 2 ) dM
are the moments of inertia around the x-, y-, and z-axis respectively. See
Intermezzo 2.5 for more on moments of inertia.
With the moments of inertia dened as in the box we can rewrite the third
term in the potential equation
4.
G
r3
s2 dM =
G
(A + B + C)
2r3
(2.27)
s2 sin2 dM . Here, s sin projects s on a plane perpendicular to OP and

this integral thus represents the moment of inertia of the body around OP .
This moment is often denoted by I.
Eq. (2.20) can then be rewritten as

U (P ) =
GM
G
3 (A + B + C 3I)
r
2r
(2.28)
which is known as MacCullaghs formula.

At face value this seems to be the result of a straightforward and rather
boring derivation, but it does reveal some interesting and important properties
of the potential and the related eld. Equation (2.20) basically shows that in
absence of rotation the gravitational attraction of an irregular body has two
contributions; the rst is the attraction of a point mass located at the center of
gravity, the second term depends on the moments of inertia around the principal
axes, which in turn depend completely on the shape of the body, or, more
precisely, on the deviations of the shape from a perfect sphere. This second

term decays as 1/r3 so that at large distances the potential approaches that of
a point mass M and becomes less and less sensitive to aspherical variations in
the shape of the body. This simply implies that if youre interested in small
scale deviations from spherical symmetry you should not be to far away from
the surface: i.e. its better to use data from satellites with a relatively low
orbit. This phenomenon is in fact an example of up (or down)ward continuation,
which we will discuss more quantitatively formally when introducing spherical
harmonics.
44
Intermezzo 2.5 Moments and products of inertia

A moment of inertia of a rigid body is dened with respect to a certain axis of
rotation.
2
2
2
I=m
1 r21 + m2 r2 + m3 r3
+ . . . =
I = r dM
For discrete masses:

and for a continuum:
mi ri2
The moment of inertia is a tensor quantity

rT ) dM
r 2 (I r
MI =
(2.29)
Note: we revert to matrix notation and manipulation of tensors. I is a secondorder tensor.

zT )a projects the vector
(I r
rT ) is a projection operator: for instance, (I z
z. This is very useful in the general

a on the (x,y) plane, i.e., perpendicular to
expression for the moments of inertia around dierent axis.

I=
and
1
0
0

r2 I =
and
0
1
0
0
0
1

(2.30)
x2 + y 2 + z 2
0
0
r
rT
r 2
=
=
0
x2 + y 2 + z 2
0
T
T
r
r rr= r r
x

x y
y
z
0
0
x2 + y 2 + z 2

=
x2
yx
zx

(2.31)
xy
y2
zy
xz
yz
z2

(2.32)
So that:

r 2 (I r
rT ) =
y2 + z 2
yx
zx
xy
x2 + z 2
zy
xz
yz
x2 + y 2

(2.33)
The diagonal elements are the familiar moments of inertia around the x, y,
and z axis. (The o-diagonal elements are known as the products of inertia,
which vanish when we choose x, y, and z as the principal axes.)
Moment of Inertia around x-axis
around y-axis
around z-axis

Ixx = (y 2 + z 2 ) dM = A
Iyy = (x2 + z 2 ) dM = B
Izz = (x2 + y 2 ) dM = C
We can pursue the development further by realizing that the moment of

inertia I around any general axis (here OP ) can be expressed as a linear combination of the moments of inertia around the principal axes. Let l2 , m2 , and
n2 be the squares of the cosines of the angle of the line OP with the x-, y-, and
z-axis, respectively. With l2 + m2 + n2 = 1 we can write I = l2 A + m2 B + n2 C
(see Figure 2.7).
Figure 2.7: Denition of direction cosines.

So far we have not been specic about the shape of the body, but for the
Earth it is relevant to consider rotational geometry so that A = B = C. This
leads to:
I = A + (C A)n2
(2.34)
Here, n = cos with the angle between OP and the z-axis, that is is the
co-latitude. ( = 90 , where is the latitude).
I = A + (C A) cos2
(2.35)
and
U (P ) =
GM
G
+ 3 (C A)
r
r
1
3
cos2
2
2

(2.36)
It is customary to write the dierence in moments of inertia as a fraction J2

of M a2 , with a the Earths radius at the equator.
C A = J2 M a2
(2.37)
so that
GJ2 M a2
GM
+
U (P ) =
r
r3
3
1
cos2
2
2

(2.38)
J2 is a measure of ellipticity; for a sphere C = A, J2 = 0, and the potential

U (P ) reduces to the expression of the gravitational potential of a body with
spherical symmetry.
46
Intermezzo 2.6 Ellipticity terms

Lets briey return to the equivalence with the spherical harmonic expansion.
If we take = cos (see box) we can write for U (P )
U (P )
=
=
+
U (r, )

GM
a
P1 (cos )
[J0 P0 (cos ) + J1
r
r
a 2
P2 (cos )]
J2
r
(2.39)
The expressions (2.20), rewritten as (2.38), and (2.39) are identical if we dene
the scaling factors Jl as follows. Since P0 (cos ) = 1, J0 must be 1 because
GM/r is the far eld term; J1 = 0 if the coordinate origin coincides with the
center of mass (see above); and J2 is as dened above. This term is of particular
interest since it describes the oblate shape of the geoid. (The higher order terms
(J4 , J6 etc.) are smaller by a factor of order 1000 and are not carried through
here, but they are incorporated in the calculation of the reference spheroid.)
The nal step towards calculating the reference gravity eld is to add a
rotational potential.
Let =
z be the angular velocity of rotation around the z-axis. The
choice of reference frame is important to get the plus and minus signs right. A
particle that moves with the rotating earth is inuenced by a centripetal force
Fcp = ma, which can formally be written in terms of the cross products between
the angular velocity and the position vector as m ( s). This shows
that the centripetal acceleration points to the rotation axis. The magnitude of
the force per unit mass is s 2 = r 2 cos . The source of Fcp is, in fact, the
F
F
gravitational attraction g (ge + mcp = g). The eective gravity ge = g mcp
(see Figure 2.8). Since we are mainly interested in the radial component (the
tangential component is very small) we can write ge = g r 2 cos2 .
Figure 2.8: The gravitational attraction produces the centripetal force due to
the rotation of the Earth.
In terms of potentials, the rotational potential has to be added to the grav-

itational potential Ugravity = Ugravitation + Urot , with

1
1
Urot = r 2 cos2 dr = r2 2 cos2 = r2 2 sin2
2
2
(2.40)
(which is in fact exactly the rotational kinetic energy (K = 21 I 2 = 21 mr2 2 ) per

unit mass of a rigid body 21 2 r2 = 21 v 2 , even though we used an approximation by ignoring the component of ge in the direction of varying latitude d.
Why? Hint: use the above diagram and consider the symmetry of the problem)
The geopotential can now be written as

3
GM
G
1
1
cos2
U (r, ) =
+ 3 J2 M a2
r2 2 sin2
(2.41)
r
r
2
2
2
which describes the contribution to the potential due to the central mass, the
oblate shape of the Earth (i.e. attening due to rotation), and the rotation
itself.
We can also write the geopotential in terms of the latitude by substituting
(sin = cos ):

G
3
1
1
GM
2
+ 3 (C A)
r2 2 cos2
sin
(2.42)
U (r, ) =
r
r
2
2
2
We now want to use this result to nd an expression for the gravity potential
and acceleration at the surface of the (reference) spheroid. The attening is
determined from the geopotential by dening the equipotential U0 , the surface
of constant U .
Since U0 is an equipotential, U must be the same (U0 ) for a point at the
pole and at the equator. We take c for the polar radius and a for the equatorial
radius and write:
U0,pole = U (c, 90) = U0,equator = U (a, 0)
Upole
Uequator
G
GM
+ 3 J2 M a2
c
c
G
1
GM
3 J2 M a2 a2 2
a
2a
2
(2.43)
(2.44)
(2.45)
(2.46)
and after some reordering to isolate a and c we get

ac
3
1
3 J2 M a2
1 a 2
f
+
= J2 + m
a
2
M a2
2 GM/a2
2
2
(2.47)
Which basically shows that the geometrical attening f as dened by the

relative dierence between the polar and equatorial radius is related to the
ellipticity coecient J2 and the ratio m between the rotational (a 2 ) to the
48
gravitational (GM a2 ) component of gravity at the equator. The value for the
attening f can be accurately determined from orbital data; in fact within a
year after the launch of the rst articial satellite by the soviets this
value could be determined with much more accuracy than by estimates given
by many investigators in the preceding centuries. The geometrical attening
is small (f = 1/298.257 1/300) (but larger than expected from equilibrium
attening of a rotating body). The dierence between the polar and equatorial
radii is thus about RE f = 6371km/300 21 km.
In order to get the shape of the reference geoid (or spheroid) one can use the
assumption that the deviation from a sphere is small, and we can thus assume
the vector from the Earths center to a point at the reference geoid to be of the
form
rg r0 + dr = r0 (1 + ) or, with r0 = a , rg a(1 + )
(2.48)
It can be shown that can be written as a function of f and latitude as given

by: rg a(1 f sin2 ) and (from binomial expansion) rg2 a2 (1 + 2f sin2 ).
Geoid anomalies, i.e. the geoid highs and lows that people talk about are
deviations from the reference geoid and they are typically of the order of several
tens of meters (with a maximum (absolute) value of about 100 m near India),
which is small (often less than 0.5%) compared to the latitude dependence of
the radius (see above). So the reference geoid with r = rg according to (2.48)
does a pretty good job in representing the average geoid.
Finally, we can determine the gravity eld at the reference geoid with a shape
as dened by (2.48) calculating the gradient of eqn. (2.42) and substituting the
position rg dened by (2.48).
In spherical coordinates:

U 1 U
,
= U =
r r

2 12
2
1 U
U
U
+
= |g| =
r
r
r
(2.49)
(2.50)
because r1 U
is small.
So we can approximate the magnitude of the gravity eld by:

GJ2 M a2 3
1
GM
2
sin
g = 2 3
r 2 cos2
r
r4
2
2

and, with r = rg = a 1 f sin2
GM
3GJ2 M a2
a2 (1 f sin2 )2
a4 (1 f sin2 )4
a 2 (1 f sin2 ) cos2
3
1
sin2
2
2
(2.51)
(2.52)

or, with the approximation (binomial expansion) given below Eqn. (2.48)
g()
=
=
=
=

GM
3
1
2
2
2
sin
(1 + 2f sin ) 3J2
m(1 sin )
a2
2
2

GM
3
9
2
(1 + J2 m) + 2f J2 + m sin
a2
2
2

2f (9/2)J2 + m
GM
3
2
(1 + J2 m) 1 +
sin
a2
2
1 + (3/2)J2 m

GM
3
(2.53)
(1 + J2 m) 1 + f sin2
a2
2
Eqn. (2.53) shows that the gravity eld at the reference spheroid can be
expressed as some latitude-dependent factor times the gravity acceleration at
the equator:

3
GM
1 + J2 m
geq ( = 0) = 2
(2.54)
a
2
Information about the attening can be derived directly from the relative
change in gravity from the pole to the equator.
gpole = geq (1 + f ) f =
gpole geq
geq
(2.55)
Eq. 2.55 is called Clairauts theorem3 . The above quadratic equation for
the gravity as a function of latitude (2.53) forms the basis for the international
gravity formula. However, this international reference for the reduction of gravity data is based on a derivation that includes some of the higher order terms.
A typical form is
g = geq (1 + sin 2 + sin2 2)
(2.56)
with the factor of proportionality and depending on GM , , a, and f .

The values of these parameters are being determined more and more accurate by
the increasing amounts of satellite data and as a result the international gravity
formula is updated regularly. The above expression (2.56) is also a truncated
series. A closed form expression for the gravity as function of latitude is given
by the Somigliana Equation4

1 + k sin2
g() = geq
.
(2.57)
1 e2 sin2
This expression has now been adopted by the Geodetic Reference System
and forms the basis for the reduction of gravity data to the reference geoid
(or reference spheroid). geq = 9.7803267714 ms2 ; k = 0.00193185138639 ;
e = 0.00669437999013.
3 After
4 After
Alexis Claude Clairaut (17131765).
C. Somigliana.
50
2.3
The Poisson and Laplace equations
The gravitational eld of the Earth is caused by its density. The mass distribution of the planet is inherently three-dimensional, but we mortals will always
only scratch at the surface. The most we can do is measure the gravitational
acceleration at the Earths surface. However, thanks to a fundamental relationship known as Gausss Theorem5 , the link between a surface observable and
the properties of the whole body in question can be found. Gausss theorem is
one of a class of theorems in vector analysis that relates integrals of dierent
types (line, surface, volume integrals). Stokess, Greens and Gausss theorem
are fundamental in the study of potential elds. The theorem due to Gauss
relates the integral over the volume of some property (most generally, a tensor
T) to a surface integral. It is also called the divergence theorem. Let V be a
volume bounded by the surface S = V (see Figure 2.9). A dierential patch
of surface dS can be represented by an outwardly pointing vector with a length
corresponding to the area of the surface element. In terms of a unit normal
vector, it is given by n
dS.
V
ds = n|ds|
Figure by MIT OCW.

Figure 2.9: Surface enclosing a volume. Unit normal vector.
Gausss theorem (for generic stu T) is as follows:

T dV =
n
T dS.
(2.58)
Lets see what we can infer about the gravitational potential within the
Earth using only information obtained at the surface. Remember we had
g=
5 After
GM
r2
and g = U.
Carl-Friedrich Gauss (17771855).
(2.59)
2.3. THE POISSON AND LAPLACE EQUATIONS
51
Suppose we measure g everywhere at the surface, and sum the results. What
we get is the ux of the gravity eld

g dS.
(2.60)
V
At this point, we can already predict that if S is the surface enclosing the
Earth, the ux of the gravity eld should be dierent from zero, and furthermore, that it should have something to do with the density distribution within
the planet. Why? Because the gravitational eld lines all point towards the
center of mass. If the ux was zero, the eld would be said to be solenoidal
. Unlike the magnetic eld the gravity eld is essentially a monopole. For the
magnetic eld, eld lines both leave and enter the spherical surface because
the Earth has a positive and a negative pole. The gravitational eld is only
solenoidal in regions not occupied by mass.
Anyway, well start working with Eq. 2.60 and see what we come up with.
On the one hand (we use Eq. 2.58 and Eq. 2.59)6 ,

U dV =
2 U dV.
(2.61)
gn
dS =
g dV =
V
On the other hand (we use the denition of the dot product and Eq. 2.59,
and dene gn as the component of g normal to dS):

GM
dV .
(2.62)
gn dS = 4r2 2 = 4G
gn
dS =
r
V
V
V
Weve assumed that S is a spherical surface, but the derivation will work for
any surface. Equating Eq. 2.61 and 2.62, we can state that
2 U (r) = 4G(r)
Poissons Equation
(2.63)
and in the homogeneous case

2 U (r) = 0
Laplaces Equation
(2.64)
The interpretation in terms of sources and sinks of the potential elds and
its relation with the eld lines is summarized in Figure 2.10:
Poissons equation is a fundamental result. It implies
1. that the total mass of a body (say, Earth) can be determined from measurements of U = g at the surface (see Eq. 2.62), and
2. no information is required about how exactly the density is distributed
within V
6 Note that the identity 2 U = U is true for scalar elds, but for a vector eld V we
should have written 2 V = ( V) ( V).
52
Figure 2.10: Poissons and Laplaces equations.
If there is no potential source (or sink) enclosed by S Laplaces equation

should be applied to nd the potential at a point P outside the surface S that
contains all attracting mass, for instance the potential at the location of a
satellite. But in the limit, it is also valid at the Earths surface. Similarly, we
will see that we can use Laplaces equation to describe Earths magnetic eld
as long as we are outside the region that contains the source for the magnetic
potential (i.e., Earths core).
We often have to nd a solution for U of Laplaces equation when only the
value of U , or its derivatives |U | = g are known at the surface of a sphere. For
instance if one wants to determine the internal mass distribution of the Earth
from gravity data. Laplaces equation is easier to solve than Poissons equation.
In practice one can usually (re)dene the problem in such a way that one can
use Laplaces equation by integrating over contributions from small volumes dV
(containing the source of the potential dU , i.e., mass dM ), see Figure 2.11 or
by using Newtons Law of Gravity along with Laplaces equation in an iterative
way.
Figure 2.11: Applicability of Poissons and Laplaces equations.
See Intermezzo 2.7.
2.4. CARTESIAN AND SPHERICAL COORDINATE SYSTEMS
53
Intermezzo 2.7 Non-uniqueness

One can prove that the solution of Laplaces equation can be uniquely determined if the boundary conditions are known (i.e. if data coverage at the surface
is good); in other words, if there are two solutions U1 and U2 that satisfy the
boundary conditions, U1 and U2 can be shown to be identical. The good news
here is that once you nd a solution for U of 2 U = 0 that satises the BCs
you do not have to be concerned about the generality of the solution. The bad
news is (see also point (2) above) that the solution of Laplaces equation does
not constrain the variations of density within V . This leads to a fundamental
non-uniqueness which is typical for potentials of force elds. We have seen this
before: the potential at a point P outside a spherically symmetric body with total mass M is the same as the potential of a point mass M located in the center
O. In between O and P the density in the spherical shells can be distributed in
an innite number of dierent ways, but the potential at P remains the same.
2.4
Cartesian and spherical coordinate systems
In Cartesian coordinates we write for 2 (the Laplacian)

2 =
2
2
2
+
+
.
x2
y 2
z 2
(2.65)
For the Earth, it is advantageous to use spherical coordinates. These are

dened as follows (see Figure 2.12):
x = r sin cos
y = r sin sin
(2.66)
z = r cos
Figure 2.12: Denition of r,

and in the spherical coordinate
system.
where = 0 = co-latitude, = 0 2 = longitude.
It is very important to realize that, whereas the Cartesian frame is described
r, and
by the immobile unit vectors x,
y
and
z, the unit vectors
are
dependent on the position of the point. They are local axes. At point P ,
r
points in the direction of increasing radius from the origin, in the direction of
increasing colatitude and
in the direction of increasing longitude .
54
One can go between coordinate axes by the transformation
r
x
sin cos sin sin cos

= cos cos cos sin sin y

z
sin
cos
0
(2.67)
Furthermore, we need to remember that integration over a volume element

dx dy dz becomes, after changing of variables r2 sin dr d d. This may be
remembered by the fact that r2 sin is the determinant of the Jacobian matrix,
i.e. the matrix obtained by lling a 33 matrix with all partial derivatives of
Eq. 2.66. After some algebra, we can write the spherical Laplacian:

2
1
1
U
U
U
1
= 0.(2.68)
r2
+ 2
sin
+ 2 2
2 U = 2
r r
r
r sin
r sin 2
2.5
Spherical harmonics
We now attempt to solve Laplaces Equation 2 U = 0, in spherical coordinates.

Laplaces equation is obeyed by potential elds outside the sources of the eld.
Remember how sines and cosines (or in general, exponentials) are often solutions
to dierential equations, of the form sin kx or cos kx, whereby k can take any
integer value. The general solution is any combination of sines and cosines of
all possible ks with weights that can be determined by satisfying the boundary
conditions (BCs). The particular solution is constructed by nding a linear
combination of these (basis) functions with weighting coecients dictated by
the BCs: it is a series solution. In the Cartesian case they are Fourier Series.
In Fourier theory, a signal, say a time series s(t), for instance a seismogram, can
be represented by the superposition of cos and sin functions and weights can be
found which approximate the signal to be analyzed in a least-squares sense.
Spherical harmonics are solutions of the spherical Laplaces Equation: they
are basically an adaption of Fourier analysis to a spherical surface. Just like
with Fourier series, the superposition of spherical harmonics can be used to
represent and analyze physical phenomena distributed on the surface on (or
within) the Earth. Still in analogy with Fourier theory, there exists a sampling
theorem which requires that sucient data are provided in order to make the
solution possible. In geophysics, one often talks about (spatial) data coverage,
which must be adequate.
We can nd a solution for U of 2 U = 0 by the good old trick of separation
of variables. We look for a solution with the following structure:
U (r, , ) = R(r)P ()Q()
(2.69)
Lets take each factor separately. In the following, an outline is given of how
to nd the solution of this elliptic equation, but working this out rigorously
requires some more eort than you might be willing to spend. But lets not try
to lose the physical meaning what we come up with.
2.5. SPHERICAL HARMONICS
55
Radial dependence: R(r)

It turns out that the functions satisfying Laplaces Equation belong to a special
class of homogeneous7 harmonic8 functions. A rst property of homogeneous
functions that can be used to our advantage is that in general, a homogeneous
function can be written in two dierent forms:
U1 (r, , )
U2 (r, , )
rl Yl (, )
(l+1)
1
Yl (, )
r
(2.70)
(2.71)
This, of course, gives the form of our radial function:

R(r) =
rl
1
l+1

(2.72)
The two alternatives R(r) = rl and R(r) = (1/r)l+1 describe the behavior
of U for an external and internal eld, respectively (in- and outside the mass
distribution). Whether to use R(r) = rl and R(r) = (1/r)l+1 depends on the
problem youre working on and on the boundary conditions. If the problem
requires a nite value for U at r = 0 than we need to use R(r) = rl . However if
we require U 0 for r then we have to use R(r) = (1/r)l+1 . The latter
is appropriate for representing the potential outside the surface that encloses
all sources of potential, such as the gravity potential U = GM r1 . However,
both are needed when we describe the magnetic potential at point r due to an
internal and external eld.
Longitudinal dependence: Q()

Substitution of Eq. 2.69 into Laplaces equation with R(r) given by Eq. 2.72,
and dividing Eq. 2.69 out again yields an equation in which - and -derivatives
occur on separate sides of the equation sign. For arbitrary and this must
mean:
d2
d2 Q
= constant,
(2.73)
which is best solved by calling the constant m2 and solving for Q as:
Q() = A cos m + B sin m.
(2.74)
Indeed, all possible constants A and B give valid solutions, and m must be a
positive integer.
7A
homogenous function f of degree n satises f (tx, ty, tz) = tn f (x, y, z).
denition, a function which satises Laplaces equation is called harmonic.
8 By
56
Latitudinal dependence: P ()
The condition is similar, except it involves both l and m. After some rearranging, one arrives at
d
sin
d

d
sin P () + [l(l + 1) sin2 m2 ]P () = 0.
d
(2.75)
This equation is the associated Legendre Equation. It turns out that

the space of the homogeneous functions has a dimension 2l +1, hence 0 m l.
If we substitute cos = z, Eq. 2.75 becomes
(1 z 2 )

m2
d
d2
P (z) = 0.
P
+
l(l
+
1)
P
(z)
2z
dz 2
dz
1 z2
(2.76)
Eq. 2.76 is in standard form and can be solved using a variety of techniques.
Most commonly, the solutions are found as polynomials Plm (cos ). The associated Legendre Equation reduces to the Legendre Equation in case m = 0. In
the latter case, the longitudinal dependence is lost as also Eq. 2.74 reverts to a
constant. The resulting functions Pl (cos ) have a rotational symmetry around
the z-axis. They are called zonal functions.
It is possible to nd expressions of the (associated) Legendre polynomials
that summarize their behavior as follows:
Pl (z) =
Plm (z) =
1 dl 2
(z 1)l
2l l! dz l
m
(1 z 2 ) 2 dl+m 2
(z 1)l ,
l!2l
dz l+m
(2.77)
(2.78)
written in terms of the (l + m)th derivative and z = cos . It is easy to make a

small table with these polynomials (note that in Table 2.1, we have used some
trig rules to simplify the expressions.) this should get you started in using
Eqs. 2.77 or 2.78.
l
Pl (z)
Pl ()
cos
1
2
2 (3z
1
3
2 (5z
1)
3z)
1
4 (3 cos 2
1
8 (5 cos 3
+ 1)
+ 3 cos )
Table 2.1: Legendre polynomials.

Some Legendre functions are plotted in Figure 2.13.
57
1
0.8
0.6
0.4
P(cos)
0.2
0
0.2
0.4
0.6
P0
P1
P2
P3
0.8
1
0
0.5
1.5
2.5
Figure 2.13: Legendre polynomials.
Spherical harmonics
The generic solution for U is thus found by combining the radial, longitudinal
and latitudinal behaviors as follows:

rl
m
m
1
l+1 [Am
(2.79)
U (r, , ) =
l cos m + Bl sin m]Pl (cos )
r
These are called the solid spherical harmonics of degree l and order m.
The spherical harmonics form a complete orthonormal basis. We implicitly
assume that the full solution is given by a summation over all possible l and m
indices, as in:

l

rl
m
m
1
l+1 [Am
U (r, , ) =
l cos m+Bl sin m]Pl (cos )(2.80)
l=0 m=0
The constants need to be determined from the boundary conditions. Because

the spherical harmonics form a complete orthonormal basis, an arbitrary real
function f (, ) can be expanded in terms of spherical harmonics by
f (, ) =
l

m
m
[Am
l cos m + Bl sin m]Pl (cos ).
(2.81)
l=0 m=0
m
The process of determining the coecients Am
l and Bl is analogous to
that to determine the coecients in a Fourier series, i.e. multiply both sides

m
of Eq. 2.81 by cos m Plm
(cos ) or sin m Pl (cos ), integrate, and use the
orthogonality relationship out comes Am
l . For unequal data distributions,
the coecients may be found in a least-squares sense.
58
Visualization
It is important to visualize the behavior of spherical harmonics, as in Figure
2.14.
Figure 2.14: Some spherical harmonics.

Some terminology to remember is that on the basis of the values of l and m
one identies three types of harmonics.
The zonal harmonics are dened to be those of the form Pl0 (cos ) =
Pl (cos ). The superposition of these Legendre polynomials describe variations with latitude; they do not depend on longitude. Zonal harmonics
vanish at l small circles on the globe, dividing the spheres into latitudinal
zones.
m
The sectorial harmonics are of the form sin(m)Pm
(cos ) or
m
cos(m)Pm (cos ). As they vanish at 2m meridians (longitudinal lines, so
m great circles), they divide the sphere into sectors.
The tesseral harmonics are those of the form sin(m)Plm (cos )

or cos(m)Plm (cos ) for l = m. The amplitude of a surface spherical
59
harmonic of a certain degree l and order m vanishes at 2m meridians of

longitude and on (l m) parallels of latitude.
Intermezzo 2.8 Cartesian vs spherical representation
If you work on a small scale with local gravity anomalies (for instance in exploration geophysics) it is not ecient to use (global) basis functions on a sphere
because the number of coecients that youd need would simply be too large.
For example to get resolution of length scales of 100 km (about 1 ) you need to
expand up to degree l=360 which with all the combinations 0 < m < l involves
several hundreds of thousands of coecients (how many exactly?). Instead you
would use a Fourier Series. The concept is similar to spherical harmonics. A
Fourier series is just a superposition of harmonic functions (sine and cosine functions) with dierent frequencies (or wave numbers k = 2/, the wavelength):
gz = constant sin(kx x)ekx z = constant sin
2x
x
2z
(2.82)
(For a 2D eld the expression includes y but is otherwise be very similar.) Or,
in more general form
gz =

n=0
an cos
nx
x
+ bn sin
nx
x
nz
(2.83)
(compare to the expression of the spherical harmonics). In this expression the

up- and downward continuation of the 1D or 2D harmonic eld is controlled by
an exponential form. The problem with downward continuation becomes immediately clear from the following example. Suppose in a marine gravity expedition
to investigate density variation in the sediments beneath the sea oor, say, at 2
km depth, gravity measurements are taken at 10 m intervals on the sea surface
(x0 species the size of the grid at which the measurements are made). Upon
downward continuation, the signal associated with the smallest wavelength allowed by such grid spacing would be amplied by a factor of exp(2000/10) =
exp(200) 10273 . (The water does not contain any concentrations of mass
that contribute to the gravity anomalies and integration over the surface enclosing the water mass would add only a constant value to the gravity potential but
that is irrelevant when studying anomalies, and Laplaces equation can still be
used.) So it is important to lter the data before the downward continuation
so that information is maintained only on length scales that are not too much
smaller than the distance over which the downward continuation has to take
place.
In other words the degree l gives the total number of nodal lines and the
order m controls how this number is distributed over nodal meridians and nodal
parallels. The higher the degree and order the ner the detail that can be
represented, but increasing l and m only makes sense if data coverage is sucient
to constrain the coecients of the polynomials.
A dierent rendering is given in Figure 2.15.
An important property follows from the depth dependence of the solution:
From eqn. (2.80) we can see that (1) the amplitude of all terms will decrease
with increasing distance from the origin (i.e., the internal source of the potential)
2.6. GLOBAL GRAVITY ANOMALIES
61
So in terms of a surface spherical harmonic potential U (l) on the unit circle,

we get the following equations for the eld in- and outside the mass distribution:
U in (r, l) =
r l
U (l)
a
a l+1
U (l)
U out (r, l) =
r
(2.87)
For gravity, this becomes:

gin (r, l) =
gout (r, l) =
l l1
r
r U (l)
al
1
r
al+1 (l + 1) l+1 U (l)
r
(2.88)
What is the gravity due to a thin sheet of mass of spherical harmonic degree
l? Lets represent this as a sheet with vanishing thickness, and call (l) the mass
density per unit area. This way we can work at constant r and use the results
for spherical symmetry. We know from Gausss law that the ux through any
surface enclosing a bit of mass is equal to the total enclosed mass (times 4G).
So constructing a box around a patch of surface S with area dS, enclosing a bit
of mass dM , we can deduce that
gout gin = 4G(l)
(2.89)
On this shell give it a radius a, we can use Eqs. 2.88 to nd gout =

U (l)(l + 1)/a and gin = U (l)l/a, and solve for U (l) using Eq. 2.89 as U (l) =
4G(l)a/(2l + 1). Plugging this into Eqs. 2.88 again we get for the gravity inand outside this mass distribution
g in (r, l) =
g out (r, l) =
4Gl
rl1
(l) l1
2l + 1
a
4Gl(l + 1)
al+2
(l) l+2
2l + 1
r
(2.90)
Length scales
Measurements of gravitational attraction are as we have seen useful in the
determination of the shape and rotational properties of the Earth. This is important for geodesy. In addition, they also provide information about aspherical
density variations in the lithosphere and mantle (important for understanding
dynamical processes, interpretation of seismic images, or for nding mineral deposits). However, before gravity measurements can be used for interpretation
several corrections will have to be made: the data reductions plays an important
role in gravimetry since the signal pertinent to the structures we are interested
in is very small.
62
Lets take a step back and get a feel for the dierent length scales and
probable sources involved. If we use a spherical harmonic expansion of the
eld U we can see that its the up- or downward continuation of the eld and
its dependence on r and degree l that controls the behavior of the solution at
dierent depths (or radius) (remember Eq. 2.87).
With increasing r from the source the amplitude of the surface harmonics
become smaller and smaller, and the decay in amplitude (spatial attenuation)
is stronger for the higher degrees l (i.e., the small-scale structures).
Table 2.2 gives and idea about the relationship between length scales, the
probable source regions, and where the measurements have to be taken.
wavelength
Source region
Measurement:
how, where?
Representation
Coordinate system
short wavelength
( < 1000 km or l > 36)
shallow:
crust, lithosphere
close to source: surface,
sea level, low orbit
satellites, planes
values at grid points;
2D Fourier series
cartesian
long wavelength
( > 1000 km or l < 36)
probably deep
(lower mantle) but shallower
source cannot be excluded
Larger distance from origin
of anomalies; perturbations
of satellite orbits
spherical harmonics
spherical
Table 2.2: Wavelength ranges of gravity anomalies
The free-air gravity anomaly

Lets assume that the geoid height N with respect to the spheroid is due to an
anomalous mass dM . If dM represents excess mass, the equipotential is warped
outwards and there will be a geoid high (N > 0); conversely, if dM represents
a mass deciency, N < 0 and there will be a geoid low.
We can represent the two potentials as follows: the actual geoid, U (r, , )
is an equipotential surface with the same potential W0 as the reference geoid U0 ,
only
U (r, , ) = U0 (r, , ) + U (r, , )
(2.91)
We dene the free-air gravity anomaly as the gravity g(P ) measured at point
P minus the gravity at the projection Q of this point onto the reference geoid
at r0 , g0 (Q). Neglecting the small dierences in direction, we can write for the
magnitudes:
g = g(P ) g0 (Q)
In terms of potentials:
(2.92)
63
Mass Excess
Reference Spheroid
Mass Deficit
Geoid - Equipotential
Gravity
Figure by MIT OCW.

Figure 2.16: Mass decit leads to geoid undulation.

U0 (P ) =
=
U0 (Q) +

dU0
N
dr r0
U0 (Q) g0 N
(2.93)
(Remember that g0 is the magnitude of the negative gradient of U and therefore

appears with a positive sign.) We knew from Eq. 2.91 that
U (P ) =
=
U0 (P ) + U (P )
U0 (Q) g0 N + U (P )
(2.94)
But also, since the potentials of U and U0 were equal, U (P ) = U0 (Q) and
we can write
g0 N = U (P )
(2.95)
This result is known as Bruns formula. Now for the gravity vectors g and
g0 , they are given by the familiar expressions
g
= U
g0
= U0
(2.96)
and the gravity disturbance vector g = g g0 can be dened as the

dierence between those two quantities:
64
Figure 2.17: Derivation. Note that in this gure, the

sign convention for the gravity is reversed; we have
used and are using that the gravity is the negative
gradient of the potential.
g = U
g = g g0 =
r
(2.97)
On the other hand, from a rst-order expansion, we learn that

g0 (P ) = g0 (Q) +

g(P ) = g0 (Q) +

dg0
N
dr r0

dU
dg0
N
dr r0
dr
(2.98)
(2.99)
Now we dene the free-air gravity anomaly as the dierence of the gravitational accelartion measured on the actual geoid (if youre on a mountain youll
need to refer to sea level) minus the reference gravity:
g = g(P ) g0 (Q)
This translates into

g

dg0
U
N

dr r0
r
(2.100)

=
=
=
g

GM
U
N

2
r
r
r

0
U
2 GM
N
r0
r02
r
2
U
g0 N
r0
r
2
U
U
r0
r
d
dr
65
(2.101)
(2.102)
So at this arbitrary point P on the geoid, the gravity anomaly g due to

the anomalous mass arises from two sources: the direct contribution dgm due
to the extra acceleration by the mass dM itself, and an additional contribution
dgh that arises from the fact that g is measured on height N above the reference
spheroid. The latter term is essentially a free air correction, similar to the
one one has to apply when referring the measurement (on a mountain, say) to
the actual geoid (sea level).
Note that Eq. (2.95) contains the boundary conditions of 2 U = 0. The
geoid height N at any point depends on the total eect of mass excesses and deciencies over the Earth. N can be determined uniquely at any point (, ) from
measurements of gravity anomalies taken over the surface of the whole Earth
this was rst done by Stokes (1849) but it does not uniquely constrain
the distribution of masses.
Gravity anomalies from geoidal heights

A convenient way to determine the geoid heights N (, ) from either the potential eld anomalies U (, ) or the gravity anomalies g(, ) is by means of
spherical harmonic expansion of N (, ) in terms of U (, ) or g(, ).
Its convenient to just give the coecients of Eq. 2.86 since the basic expressions are the same. Lets see how that notation would work for eq. (2.86):

m
GM
UA
Al
(2.103)
=
Blm
UB
a
Note that the subscripts A and B are used to label the coecients of the
cos m and sin m parts, respectively. Note also that we have now taken the
factor GM al as the scaling factor of the coecients.
We can also expand the potential U0 on the reference spheroid:

m
GM
U0,A
Al
=
(2.104)
U0,B
0
a
(Note that we did not drop the m , even though m = 0 for the zonal harmonics
used for the reference spheroid. We just require the coecient Alm to be zero
for m =
0. By doing this we can keep the equations simple.)
66
The coecients of the anomalous potential U (, ) are then given by:

GM
UA
Alm )
(Am
l
=
(2.105)
UB
Blm
a
We can now expand g(, ) in a similar series using eq. (2.101). For the
U , we can see by inspection that the radial derivative as prescribed has the
following eect on the coecients (note that the reference radius r0 = a from
earlier denitions):
dU
l+1
dr
r0
(2.106)
and the other term of Eq. 2.101 brings down

2
2
U
r0
a
(2.107)
As a result, we get

gA
gB

=
=

m
l+1
(Am
l Al )
Blm
a

m
(Am
l Al )
g0 (l 1)
Blm
GM
a
(2.108)
(2.109)
The proportionality with (l 1)g0 means that the higher degree terms are
magnied in the gravity eld relative to those in the potential eld. This leads to
the important result that gravity maps typically contain much more detail than
geoid maps because the spatial attenuation of the higher degree components is
suppressed.
Using Eq. (2.95) we can express the coecients of the expansion of N (, )
in terms of either the coecients of the expanded anomalous potential

m
GM
NA
(Am
l Al )
(2.110)
=
g0
NB
Blm
a
which, if we replace g by g and by assuming that g g gets the following form

NA
Alm )
(Am
l
=a
(2.111)
NB
Blm
or in terms of the coecients of the gravity anomalies (eqns. 2.109 and 2.111)

a
gA
NA
Alm )
(Am
l
=
=a
(2.112)
NB
gB
Blm
(l 1)g0
The geoid heights can thus be synthesized from the expansions of either the
gravity anomalies (2.112) or the anomalous potential (2.111). Geoid anomalies
2.7. GRAVITY ANOMALIES AND THE REDUCTION OF GRAVITY DATA67

have been constructed from both surface measurements of gravity (2.112) and
from satellite observations (2.111). Equation (2.112) indicates that relative to
the gravity anomalies, the coecients of N (, ) are suppressed by a factor of
1/(l 1). As a result, shorter wavelength features are much more prominent on
gravity maps. In other words, geoid (and geoid height) maps essentially depict
the low harmonics of the gravitational eld. A nal note that is relevant for the
reduction of the gravity data. Gravity data are typically reduced to sea-level,
which coincides with the geoid and not with the actual reference spheroid. Eq.
8 can then be used to make the additional correction to the reference spheroid,
which eectively means that the long wavelength signal is removed. This results
in very high resolution gravity maps.
2.7 Gravity anomalies and the reduction of gravity data

The combination of the reduced gravity eld and the topography yields important information on the mechanical state of the crust and lithosphere. Both
gravity and topography can be obtained by remote sensing and in many cases
they form the basis of our knowledge of the dynamical state of planets, such
as Mars, and natural satellites, such as Earths Moon. Data reduction plays
an important role in gravity studies since the signal caused by the aspherical
variation in density that one wants to study are very small compared not only
to the observed eld but also other eects, such as the inuence of the position
at which the measurement is made. The following sum shows the various components to the observed gravity, with the name of the corresponding corrections
that should be made shown in parenthesis:
Observed gravity = attraction of the reference spheroid, PLUS:
eects of elevation above sea level (Free Air correction), which should
include the elevation (geoid anomaly) of the sea level above the reference
spheroid
eect of normal attracting mass between observation point and sea level
(Bouguer and terrain correction)
eect of masses that support topographic loads (isostatic correction)
time-dependent changes in Earths gure of shape (tidal correction)
eect of changes in the rotation term due to motion of the observation
otv
point (e.g. when measurements are made from a moving ship. (E
os
correction)
eects of crust and mantle density anomalies (geology or geodynamic
processes).
68
Only the bold corrections will be discussed here. The tidal correction is
small, but must be accounted for when high precision data are required. The
application of the dierent corrections is illustrated by a simple example of a
small density anomaly located in a topography high that is isostatically compensated. See series of diagrams.
Free Air Anomaly

So far it has been assumed that measurements at sea level (i.e. the actual geoid)
were available. This is often not the case. If, for instance, g is measured on the
land surface at an altitude h one has to make the following correction :
dgFA = 2
hg
r
(2.113)
For g at sea level this correction amounts to dgFA = 0.3086h mgal or

0.3086h 105 ms2 (h in meter). Note that this assumes no mass between
the observer and sea level, hence the name free-air correction. The eect of
ellipticity is often ignored, but one can use r = Req (1 f sin2 ). Note: per
meter elevation this correction equals 3.1 106 ms2 3.1 107 g: this is
on the limit of the precision that can be attained by eld instruments, which
shows that uncertainties in elevation are a limiting factor in the precision that
can be achieved. (A realistic uncertainty is 1 mgal).
Make sure the correction is applied correctly, since there can be confusion
about the sign of the correction, which depends on the denition of the potential. The objective of the correction is to compensate for the decrease in gravity
attraction with increasing distance from the source (center of the Earth). Formally, given the minus sign in (2.114), the correction has to be subtracted, but
it is not uncommon to take the correction as the positive number in which case
it will have to be added. (Just bear in mind that you have to make the measured
value larger by adding gravity so it compares directly to the reference value at
the same height; alternatively, you can make the reference value smaller if you
are above sea level; if you are in a submarine you will, of course, have to do the
opposite).
The Free Air anomaly is then obtained by the correction for height above
sea level and by subtraction of the reference gravity eld
gFA = gobs dgFA g0 () = (gobs + 0.3086h 103 ) g0 ()
(2.114)
(Note that there could be a component due to the fact that the sea level ( the
geoid) does not coincide with the reference spheroid; an additional correction
can then be made to take out the extra gravity anomaly. One can simply apply
(2.114) and use h = h + N as the elevation, which is equivalent to adding a
correction to g0 () so that it represents the reference value at the geoid. This
correction is not important if the variation in geoid is small across the survey
region because then the correction is the same for all data points.)
2.7. GRAVITY ANOMALIES AND THE REDUCTION OF GRAVITY DATA69
Bouguer anomaly
The free air correction does not correct for any attracting mass between observation point and sea level. However, on land, at a certain elevation there will
be attracting mass (even though it is often compensated - isostasy (see below)).
Instead of estimating the true shape of, say, a mountain on which the measurement is made, one often resorts to what is known as the slab approximation
in which one simply assumes that the rocks are of innite horizontal extent.
The Bouguer correction is then given by
dgB = 2Gh
(2.115)
where G is the gravitational constant, is the assumed mean density of crustal

rock and h is the height above sea level. For G = 6.67 1011 m3 kg3 s2 and
= 2, 700 kgm3 we obtain a correction of 1.1106 ms2 per meter of elevation
(or 0.11 h mgal, h in meter). If the slab approximation is not satisfactory, for
instance near the top of mountains, on has to apply an additional terrain
correction. It is straightforward to apply the terrain correction if one has
access to digital topography/bathymetry data.
The Bouguer anomaly has to be subtracted, since one wants to remove the
eects of the extra attraction. The Bouguer correction is typically applied after
the application of the Free Air correction. Ignoring the terrain correction, the
Bouguer gravity anomaly is then given by
gB = gobs dgFA g0 () dgB = gFA dgB
(2.116)
In principle, with the Bouguer anomaly we have accounted for the attraction
of all rock between observation point and sea level, and gB thus represents
the gravitational attraction of the material below sea level. Bouguer Anomaly
maps are typically used to study gravity on continents whereas the Free Air
Anomaly is more commonly used in oceanic regions.
Isostasy and isostatic correction

If the mass between the observation point and sea level is all that contributes
to the measured gravity one would expect that the Free Air anomaly is large,
and positive over topography highs (since this mass is unaccounted for) and
that the Bouguer anomaly decreases to zero. This relationship between the two
gravity anomalies and topography is indeed what would be obtained in case
the mass is completely supported by the strength of the plate (i.e. no isostatic
compensation). In early gravity surveys, however, they found that the Bouguer
gravity anomaly over mountain ranges was, somewhat surprisingly, large and
negative. Apparently, a mass deciency remained after the mass above sea level
was compensated for. In other words, the Bouguer correction subtracted too
much! This observation in the 19th century lead Airy and Pratt to develop
the concept of isostasy. In short, isostasy means that at depths larger than
a certain compensation depth the observed variations in height above sea level
70
no longer contribute to lateral variations in pressure. In case of Airy Isostasy

this is achieved by a compensation root, such that the depth to the interface
between the loading mass (with constant density) and the rest of the mantle
varies. This is, in fact Archimedes Law, and a good example of this mechanism
is the oating iceberg, of which we see only the top above the sea level. In the
case of Pratt Isostasy the compensation depth does not vary and constant
pressure is achieved by lateral variations in density. It is now known that both
mechanisms play a role.
s.l.
w
c
s.l.
h
w
Figure by MIT OCW.

Figure 2.18: Airy (left) and Pratt (right) isostasy.
The basic equation that describes the relationship between the topographic
height and the depth of the compensating body is (see Figure 2.19):
H=
c h
m c
(2.117)
2.8. CORRELATION BETWEEN GRAVITY ANOMALIES AND TOPOGRAPHY.71
Figure 2.19: Airy isostasy.

Assuming Airy Isostasy and some constant density for crustal rock one can
compute H(x, y) from known (digital) topography h(x, y) and thus correct for
the mass deciency. This results in the Isostatic Anomaly. If all is done correctly the isostatic anomaly isolates the small signal due to the density anomaly
that is not compensated (local geology, or geodynamic processes).
2.8 Correlation between gravity anomalies and

topography.
The correlation between Bouguer and Free Air anomalies on the one hand and
topography on the other thus contains information as to what level the topography is isostatically compensated.
In the case of Airy Isostasy it is obvious that the compensating root causes
the mass deciency that results in a negative Bouguer anomaly. If the topography is compensated the mass excess above sea level is canceled by the mass
deciency below it, and as a consequence the Free Air Anomaly is small; usually,
it is not zero since the attracting mass is closer to the observation point and is
thus less attenuated than the compensating signal of the mass deciency so that
some correlation between the Free Air Anomaly and topography can remain.
Apart from this eect (which also plays a role near the edges of topographic
features), the Free air anomaly is close to zero and the Bouguer anomaly large
and negative when the topography is completely compensated isostatically (also
referred to as in isostatic equilibrium).
In case the topography is NOT compensated, the Free air anomaly is large
and positive, and the Bouguer anomaly zero.
(This also depends on the length scale of the load and the strength of the
supporting plate).
Whether or not a topographic load is or can be compensated depends largely
on the strength (and the thickness) of the supporting plate and on the length
scale of the loading structure. Intuitively it is obvious that small objects are not
compensated because the lithospheric plate is strong enough to carry the load.
This explains why impact craters can survive over very long periods of time!
(Large craters may be isostatically compensated, but the narrow rims of the
72
crater will not disappear by ow!) In contrast, loading over large regions, i.e.
much larger than the distance to the compensation depth, results in the development of a compensating root. It is also obvious why the strength (viscosity)
of the plate enters the equation. If the viscosity is very small, isostatic equilibrium can occur even for very small bodies (consider, for example, the oating
iceberg!). Will discuss the relationship between gravity anomalies and topography in more (theoretical) detail later. We will also see how viscosity adds a
time dependence to the system. Also this is easy to understand intuitively; low
strength means that isostatic equilibrium can occur almost instantly (iceberg!),
but for higher viscosity the relaxation time is much longer. The ow rate of the
material beneath the supporting plate determines how quickly this plate can
assume isostatic equilibrium, and this ow rate is a function of viscosity. For
large viscosity, loading or unloading results in a viscous delay; for instance the
rebound after deglaciation.
2.9
Flexure and gravity.
The bending of the lithosphere combined with its large strength is, in fact, one
of the compensation mechanisms for isostasy. When we discussed isostasy we
have seen that the depth to the bottom of the root, which is less dense than
surrounding rock at the same depth, can be calculated from Archimedes
Principle: if crustal material with density c replaces denser mantle material
with density m a mountain range with height h has a compensating root with
thickness H
H=
hc
m c
(2.118)
This type of compensation is also referred to as Airy Isostasy. It does

not account for any strength of the plate. However, it is intuitively obvious
that the depression H decreases if the strength (or the exural rigidity) of the
lithosphere increases. The consideration of lithospheric strength for calculates
based on isostasy is important in particular for the loading on not too long a
time scale.
An elegant and very useful way to quantify the eect of exure is by considering the exure due to a periodic load. Lets consider a periodic load due to topography h with maximum amplitude h0 and wavelength : h = h0 sin(2x/).
The corresponding load is then given by

2x
V (x) = c gh0 sin
(2.119)
so that the exure equation becomes

D
d4 w
+ (m c )gh = c gh0 sin
dx4
2x

(2.120)
2.9. FLEXURE AND GRAVITY.
73
The solution can be shown to be

h0 sin 2x
2x
!
w(x) =
2
4 = w0 sin
m
D
c 1 + c g
(2.121)
From eq. (2.121) we can see that for very large exural rigidity (or very large
elastic thickness of the plate) the denominator will predominate the equation
and the deection will become small (w 0 for D ); in other words,
the load has no eect on the depression. The same is true for short wavelengths, i.e. for 2(D/c g)1/4 . In contrast, for very long wave lengths
( 2(D/c g)1/4 ) or for a very weak (or thin) plate the maximum depression
becomes
w0
c h 0
m c
(2.122)
which is the same as for a completely compensated mass (see eq. 2.118). In
other words, the plate has no strength for long wavelength loads.
The importance of this formulation is evident if you realize that any topography can be described by a (Fourier) series of periodic functions with dierent
wavelengths. One can thus use Fourier Analysis to investigate the depression or
compensation of any shape of load.
Eq. (2.121) can be used to nd expressions for the inuence of exure on
the Free Air and the Bouguer gravity anomaly. The gravity anomalies depend
on the exural rigidity in very much the same way as the deection in (2.121).
Free-air gravity anomaly:

2bm /
e
2x
gfa = 2c G 1
h
sin
2
4
0
D
1 + (m
c )g
(2.123)
Bouguer gravity anomaly:

gB =
2c Ge2bm /
2
4 h0 sin
D
1 + (m
c )g
2x

(2.124)
where bm is the depth to the Moho (i.e. the depressed interface between c and
m ) and the exponential in the numerator accounts for the fact that this interface
is at a certain depth (this factor controls, in fact, the downward continuation).
The important thing to remember is the linear relationship with the topography h and the proportionality with D1 . One can follow a similar reasoning
as above to show that for short wavelengths the free air anomaly is large (and
positive) and that the Bouguer anomaly is almost zero. This can be explained
by the fact that the exure is then negligible so that the Bouguer correction successfully removes all anomalous structure. However, for long wavelength loads,
the load is completely compensated so that after correction to zero elevation,
74
the Bouguer correction still feels the anomalously low density root (which is
not corrected for). The Bouguer anomaly is large and negative for a completely
compensated load. Complete isostasy means also that there is no net mass
dierence so that the free air gravity anomaly is very small (practically zero).
Gravity measurements thus contain information about the degree of isostatic
compensation.
The correlation between the topography and the measured Bouguer anomalies can be modeled by means of eq. (2.124) and this gives information about
the exural rigidity, and thus the (eective!) thickness of the elastic plate. The
diagram below gives the Bouguer anomaly as a function of wave length (i.e.
topography was subjected to a Fourier transformation). It shows that topography with wavelengths less than about 100 km is not compensated (Bouguer
anomaly is zero). The solid curves are the predictions according to eq. 2.124
for dierent values of the exural parameter . The parameters used for these
theoretical curves are m = 3400 kgm3 , c = 2700 kgm3 , bm = 30 km,
= [4D/(m c )g]1/4 = 5, 10, 20, and 50 km. There is considerable scatter
but a value for of about 20 seems to t the observations quite well, which,
with E= 60 GPa and = 0.25, gives an eective elastic thickness h 6 km.
1.0
-gB / 2cGh0
0.8
0.6
0.4
United States
= 50 km
= 20 km
= 10 km
= 5 km
0.2
0
0
200
400
600
800
1000
(km)
Figure by MIT OCW.
Figure 2.20: Bouguer anomalies and topography.
Post-glacial rebound and viscosity
So far we have looked at the bending or exure of the elastic lithosphere to
loading, for instance by sea mounts. To determine the deection w(x) we used
the principle of isostasy. In order for isostasy to work the mantle beneath the
lithosphere must be able to ow. Conversely, if we know the history of loading,
or unloading, so if we know the deection as a function of time w(x, t), we can
75
investigate the ow beneath the lithosphere. The rate of ow is dependent on

the viscosity of the mantle material. Viscosity plays a central role in understanding mantle dynamics. Dynamic viscosity can be dened as the ratio of the
applied (deviatoric) stress and the resultant strain rate; here we mostly consider
Newtonian viscosity, i.e., a linear relationship between stress and strain rate.
The unit of viscosity is Pascal Second [Pa s].
A classical example of a situation where the history of (un)loading is suciently well known is that of post-glacial rebound. The concept is simple:
1. the lithosphere is depressed upon loading of an ice sheet (viscous mantle
ow away from depression make this possible)
2. the ice sheet melts at the end of glaciation and the lithosphere starts
rebound slowly to its original state (mantle ow towards the decreasing
depression makes this possible). The uplift is well documented from elevated (and dated) shore lines. From the rate of return ow one can
estimate the value for the viscosity.
Start of Glaciation
Load Causes
Subsidence
Load
Elastic Lithosphere
Viscous Mantle
C
Ice Melts at End

of Glaciation
Two remarks:
Subsequent Slow
Rebound of Lithosphere
Figure by MIT OCW.
1. the dimension of the load determines to some extend the depth over which
the mantle is involved in the return ow the comparison of rebound
history for dierent initial load dimensions gives some constraints on the
variation of viscosity with depth.
2. On long time scales the lithosphere has no strength, but in sophisticated
modeling of the post glacial rebound the exural rigidity is still taken into
76
300
Uplift in central Fennoscandia
(mouth of Angerman R.)
Uplift (m)
200
Model
100
Ice Melting
0
10
Thousands of Years BP
Figure 2.23. Uplift in central Fennoscandia calculated for a constant
viscosity (1021 Pas) mantle (green line) and geological observations
(circle) from the nothern Gulf of Bothnia.
Figure by MIT OCW.
account. Also taken into account in recent models is the history of the
melting and the retreat of the ice cap itself (including the changes in
shore line with time!). In older models one only investigated the response
to instantaneous removal of the load.
Typical values for the dynamic viscosity in the Earths mantle are 1019 Pa s
for the upper mantle to 1021 Pa s for the lower mantle. The lithosphere is even
stier, with a typical viscosity of about 1024 (for comparison: water at room
temperature has a viscosity of about 103 Pa s; this seems small but if youve
ever dived of a 10 m board you know its not negligible!)
Things to remember about these values:
1. very large viscosity in the entire mantle
2. lower mantle (probably) more viscous than upper mantle,
3. the dierence is not very large compared to the large value of the viscosity
itself.
An important property of viscosity is that it is temperature dependent; the
viscosity decreases exponentially with increasing temperature as = 0 e30T /Tm ,
where Tm is the melting temperature and -30 is an empirical, material dependent
value.
77
This temperature dependence of viscosity explains why one gets convection

beneath the cooling lithosphere. As we have discussed before, with typical
values for the geothermal gradient (e.g., 20 Kkm1 ) as deduced from surface
heat ow using Fouriers Law the temperature would quickly rise to near the
solidus, the temperature where the rock starts to melt. However, we know from
several observations, for instance from the propagation of S- waves, that the
temperature is below the solidus in most parts of the mantle (with the possible
exception in the low velocity zone beneath oceanic and parts of the continental
lithosphere). So there must be a mechanism that keeps the temperature down,
or, in other words, that cools the mantle much more eciently than conduction.
That mechanism is convection. We saw above that the viscosity of the lithosphere is very high, and upper mantle viscosity is about 5 orders of magnitude
lower. This is largely due to the temperature dependence of the viscosity (as
mentioned above): when the temperature gets closer to the solidus (Tm ) the
viscosity drops and the material starts to ow.
Temperature (oC)
0
1000
5 Myr
Depth (km)
25 Myr
100 Myr
100
Craton
200
Figure by MIT OCW.
Dry solidus
Chapter 3
The Magnetic Field of the

Earth
Introduction
Studies of the geomagnetic eld have a long history, in particular because of its
importance for navigation. The geomagnetic eld and its variations over time
are our most direct ways to study the dynamics of the core. The variations
with time of the geomagnetic eld, the secular variations, are the basis for the
science of paleomagnetism, and several major discoveries in the late fties
gave important new impulses to the concept of plate tectonics. Magnetism also
plays a major role in exploration geophysics in the search for ore deposits.
Because of its use as a navigation tool, the study of the magnetic eld has
a very long history, and probably goes back to the 12thC when it was rst
exploited by the Chinese. It was not until 1600 that William Gilbert postulated
that the Earth is, in fact, a gigantic magnet. The origin of the Earths eld has,
however, remained enigmatic for another 300 years after Gilberts manifesto
De Magnete. It was also known early on that the eld was not constant in
time, and the secular variation is well recorded so that a very useful historical
record of the variations in strength and, in particular, in direction is available
for research. The rst (known) map of declination was published by Halley
(yes, the one of the comet) in 1701 (the chart of the lines of equal magnetic
variation, also known as the Tabula Nautica).
The source of the main eld and the cause of the secular variation remained
a mystery since the rapid uctuations seemed to be at odds with the rigidity
of the Earth, and until early this century an external origin of the eld was
seriously considered. In a breakthrough (1838) Gauss was able to prove that
almost the entire eld has to be of internal origin. Gauss used spherical
harmonics and showed that the coecients of the eld expansion, which he
determined by tting the surface harmonics to the available magnetic data at
that time (a small number of magnetic eld measurements at intervals of about
30 along several parallels - lines of constant latitude), were almost identical to
79
CHAPTER 3. THE MAGNETIC FIELD OF THE EARTH
80
the coecients for a eld due to a magnetized sphere or to a dipole. In fact, he

also showed from a spectral analysis that the best t to the observed eld was
obtained if the dipole was not purely axial but made an angle of about 11 with
the Earths rotation axis.
An outstanding issue remained: what causes the internal eld? It was clear
that the temperatures in the interior of the Earth are probably much too high
to sustain permanent magnetization. A major leap in the understanding of the
origin of the eld came in the rst decade of the twentieth century when Oldham
(1906) and Gutenberg (1912) demonstrated the existence of a (outer) core with
a very low viscosity since it did not seem to allow shear wave propagation (
rigidity =0). So the rigidity problem was solved. From the cosmic abundance
of metallic iron it was inferred that metallic iron could be the major constituent
of the (outer) core (the seismologist Inge Lehmann discovered the existence of
the inner core in 1936). In the 40s Larmor postulated that the magnetic eld
(and its temporal variations) were, in fact, due to the rapid motion of highly
conductive metallic iron in the liquid outer core. Fine; but there was still the
apparent contradiction that the magnetic eld would diuse away rather quickly
due to ohmic dissipation while it was known that very old rocks revealed a remnant magnetic eld. In other words, the eld has to be sustained by some, at
that time, unknown process. This lead to the idea of the geodynamo (Sir
Bullard, 40-ies and 50-ies), which forms the basis for our current understanding
of the origin of the geomagnetic eld. The theory of magneto-hydrodynamics
that deals with magnetic elds in moving liquids is dicult and many approximations and assumptions have to be used to nd any meaningful solutions. In
the past decades, with the development of powerful computers, rapid progress
has been made in understanding the eld and the cause of the secular variation.
We will see, however, that there are still many outstanding questions.
Dierences and similarities with Gravity

Similarities are:
The magnetic and gravity elds are both potential elds, the elds are
the gradient of some potential V , and Laplaces and Poissons equations
apply.
For the description and analysis of these elds, spherical harmonics is the
most convenient tool, which will be used to illustrate important properties
of the geomagnetic eld.
In both cases we will use a reference eld to reduce the observations of
the eld.
Both elds are dominated by a simple geometry, but the higher degree
components are required to get a complete picture of the eld. In gravity,
the major component of the eld is that of a point mass M in the center
of the Earth; in geomagnetism, we will see that the eld is dominated
81
by that of an axial dipole in the center of the Earth and approximately
aligned along the rotational axis.
Dierences are:
In gravity the attracting mass m is positive; there is no such thing as
negative mass. In magnetism, there are positive and negative poles.
Figure 3.1:
In gravity, every mass element dM acts as a monopole; in contrast, in
magnetism isolated sources and sinks of the magnetic eld H dont exist
( H = 0) and one must always consider a pair of opposite poles. Opposite poles attract and like poles repel each other. If the distance d between
the poles is (innitesimally) small dipole.
Gravitational potential (or any potential due to a monopole) falls of as
1 over r, and the gravitational attraction as 1 over r2 . In contrast, the
potential due to a dipole falls of as 1 over r2 and the eld of a dipole as
1 over r3 . This follows directly from analysis of the spherical harmonic
expansion of the potential and the assumption that magnetic monopoles,
if they exist at all, are not relevant for geomagnetism (so that the l = 0
component is zero).
The direction and the strength of the magnetic eld varies with time
due to external and internal processes. As a result, the reference eld
has to be determined at regular intervals of time (and not only when
better measurements become available as is the case with the International
Gravity Field).
The variation of the eld with time is documented, i.e. there is a historic record available to us. Rocks have a memory of the magnetic eld
through a process known as magnetization. The then current magnetic
eld is frozen in a rock if the rock sample cools (for instance, after eruption) beneath the so called Curie temperature, which is dierent for
dierent minerals, but about 500-600C for the most important minerals
such as magnetite. This is the basis for paleomagnetism. (There is no
such thing as paleogravity!)
82
3.1
The main eld
From the measurement of the magnetic eld it became clear that the eld has
both internal and external sources, both of which exhibit a time dependence.
Spherical harmonics is a very convenient tool to account for both components.
Lets consider the general expression of the magnetic potential as the superposition of Legendre polynomials:
Vm (r, , ) = a

l

l=1 m=0
l+1
a
r r l
a
[glm cos m + hlm sin m]
m
[g l
cos m +
m
h l
sin m]

Plm (cos ),(3.1)
or, assuming Einsteins summation convention (implicit summation over

repeated indices), we can write:

r l
a l+1
Vm (r, , ) = a Ilm
Plm (cos )
+ Elm
r
a
(3.2)
where Ilm and Elm are the amplitude factors of the contributions of the internal
and external sources, respectively. (Note that, in contrast to the gravitational
potential, the rst degree is l = 1, since l = 0 would represent a monopole,
which is not relevant to geomagnetism.)
3.2
The internal eld
The internal eld has two components: [1] the crustal eld and [2] the core eld.
The crustal eld

The spatial attenuation of the eld as 1 over distance cubed means that the
short wavelength variations at the Earths surface must have a shallow source.
Can not be much deeper than mid crust, since otherwise temperatures are too
high. More is known about the crustal eld than about the core eld since we
know more about the composition and physical parameters such as temperature
and pressure and about the types of magnetization. Two important types of
magnetization:
Remanent magnetization (there is a eld B even in absence of an ambient eld). If this persists over time scales of O(108 ) years, we call this
permanent magnetization. Rocks can acquire permanent magnetization when they cool beneath the Curie temperature (about 500-600 for
most relevant minerals). The ambient eld then gets frozen in, which is
very useful for paleomagnetism.
Induced magnetization (no eld, unless induced by ambient eld).
3.2. THE INTERNAL FIELD
83
No mantle eld
Why not in the mantle? Firstly, the mantle consists mainly of silicates and the
average conductivity is very low. Secondly, as we will see later, elds in a low
conductivity medium decay very rapidly unless sustained by rapid motion, but
convection in the mantle is too slow for that. Thirdly, permanent magnetization
is out of the question since mantle temperatures are too high (higher than the
Curie temperature in most of the mantle).
The core eld

The temperatures are too high for permanent magnetization. The eld is caused
by rapid (and complex) electric currents in the liquid outer core, which consists
mainly of metallic iron. Convection in the core is much more vigorous than in
the mantle: about 106 times faster than mantle convection (i.e, of the order of
about 10 km/yr).
Outstanding problems are:
1. the energy source for the rapid ow. A contribution of radioactive decay of Potassium and, in particular, Uranium, can - at this stage - not
be ruled out. However, there seems to be increasing consensus that the
primary candidate for providing the driving energy is gravitational energy
released by downwelling of heavy material in a compositional convection caused by dierentiation of the inner core. Solidication of the
inner core is selective: it takes out the iron and leaves behind in the outer
core a relatively light residue that is gravitationally unstable. Upon solidication there is also latent heat release, which helps maintaining an
adiabatic temperature gradient across the outer core but does not eectively couple to convective ow. The lateral variations in temperature in
the outer core are probably very small and the role of thermal convection
is negligible. Any aspherical variations in density would be annihilated
quickly by convection as a result of the low viscosity.
2. the details of the pattern of ow. This is a major focus in studies of the
geodynamo.
The knowledge about ow in the outer core is also restricted by observational
limitations.
the spatial attenuation is large since the eld falls of as 1 over r3 . As a
consequence eects of turbulent ow in the core are not observed at the
surface. Conversely, the downward continuation of small scale features in
the eld will be hampered by the amplication of uncertainties and of the
crustal eld.
the mantle has a small but non-zero conductivity, so that rapid variations
in the core eld will be attenuated. In general, only features of length
scales larger than about 1500 km (l < 12, 13) and on time scales longer
84

than 1 to 5 year are attributed to core ow, although this rule of thumb
is ad hoc.
The core eld has the following characteristics:
1. 90% of the eld at the Earths surface can be described by a dipole inclined at about 11 to the Earths spin axis. The axis of the dipole intersects the Earths surface at the so-called geomagnetic poles at about
(78.5 N, 70 W) (West Greenland) and (75.5 S, 110 E). In theory the angle between the magnetic eld lines and the Earths surface is 90 at the
poles but owing to local magnetic anomalies in the crust this is not necessarily the case in real life.
The dipole eld is represented by the degree 1 (l = 1) terms in the harmonic expansion. From the spherical harmonic expansion one can see
immediately that the potential due to a dipole attenuates as 1 over r2 .
2. The remaining 10% is known as the non-dipole eld and consists of

a quadrupole (l = 2), and octopole (l = 3), etc. We will see that at
the core-mantle-boundary the relative contribution of these higher degree
components is much larger!
Note that the relative contribution 90%10% can change over time as
part of the secular variation.
3. The strength of the Earths magnetic eld varies from about 60,000 nT at
the magnetic pole to about 25,000 nT at the magnetic equator. (1nT =
1 = 101 Wb m2 ).
4. Secular variation: important are the westward drift and changes in the
strength of the dipole eld.
5. The eld is probably not completely independent from the mantle. Coremantle coupling is suggested by several observations (i.e., changes of the
length of day, not discussed here), by the statistics of eld reversals, and
by the suggested preferential reversal paths.
3.3. THE EXTERNAL FIELD
85
Intermezzo 3.1 Units of confusion

The units that are typically used for the dierent variables in geomagnetism are
somewhat confusing, and up to 5 dierent systems are used. We will mainly use
the Syst`
eme International dUnites (S.I.) and mention the electromagnetic units
(e.m.u.) in passing. When one talks about the geomagnetic eld one often talks
about B, measured in T (Tesla) (= kg1 A1 s2 ) or nT (nanoTesla) in S.I., or
Gauss in e.m.u. In fact, B is the magnetic induction due to the magnetic eld
H, which is measured in Am1 in S.I. or Oersted in e.m.u. For the conversions
from the one to the other unit system: T = 104 G(auss) 1nT = 105 G =
1 (gamma). B = 0 H with 0 the magnetic permeability in free space; 0 =
4107 kgmA2 s2 [=NA2 = H(enry) m1 ], in S.I., and 0 = 1 G Oe in
e.m.u. So, in e.m.u., B = H, hence the liberal use of B for the Earths eld.
The magnetic permeability is a measure of the ease with which the eld H
can penetrate into a material. This is a material property, and we will get back
to this when we discuss rock magnetism.
In the next table, some of the quantities are summarized together with their
units and dimensions. There are only 4 so-called dimensions we need. These are
(with their symbol and standard units) mass [M (kg)], length [L (m)], time
[T (s)] and current [I (Amp`ere)].
Quantity
force
charge
electric eld
electric ux
electric potential
magnetic induction
magnetic ux
magnetic potential
permittivity of vacuum
permeability of vacuum
resistance
resistivity
3.3
Symbol
F
q
E
E
VE
B
B
Vm
0
0
R
Dimension
MLT2
IT
MLT3 I1
ML3 T3 I1
ML2 T3 I1
MT2 I1
ML2 T2 I1
MLT2 I1
M1 L3 T4 I2
MLT2 I2
ML2 T3 I2
ML3 T3 I2
S.I. Units
Newton (N)
Coulomb (C)
N/C
N/C m2
Volt (V)
Tesla (T)
Weber (Wb)
Tm
C2 /(N m2 )
Wb/(A m)
Ohm ()
m
The external eld
The strength of the eld due to external sources is much weaker than that of the
internal sources. Moreover, the typical time scale for changes of the intensity
of the external eld is much shorter than that of the eld due to the internal
source. Variations in magnetic eld due to an external origin (atmospheric,
solar wind) are often on much shorter time scales so that they can be separated
from the contributions of the internal sources.
The separation is ad hoc but seems to work ne. The rapid variation of the
external eld can be used to study the (lateral variation in) conductivity in the
Earths mantle, in particular to depth of less than about 1000 km. Owing to
the spatial attenuation of the coecients related to the external eld and, in
particular, to the fact that the rapid uctuations can only penetrate to a certain
depth (the skin depth, which is inversely proportional to the frequency), it is
86
dicult to study the conductivity in the deeper part of the lower mantle.
3.4 The magnetic induction due to a magnetic

dipole
Magnetic elds are fairly similar to electric elds, and in the derivation of the
magnetic induction due to a magnetic dipole, we can draw important conclusions
based on analogies with the electric potential due to an electric dipole. We will
therefore start with a brief discussion of electric dipoles.
On the other hand, our familiarity with the gravity eld should enable us to
deduce dierences and similarities of the magnetic eld and the gravity eld as
well. In this manner, we will start with the eld due to a magnetic dipole the
simplest conguration in magnetics in a straightforward analysis based on
experiments, and subsequently extend this to the eld induced by higher-order
poles: quadrupoles, octopoles, and so on. The equivalence with gravitational
potential theory will follow from the fact that both the gravitational and the
magnetic potential are solutions to Laplaces equation.
The electric eld due to an electric dipole

The law obeyed by the force of interaction of point charges q (in vacuum) was
established experimentally in 1785 by Charles de Coulomb. Coulombs Law
can be expressed as:
F = Ke
1 q0 q
q0 q
r=

r,
r2
40 r2
(3.3)
where
r is the unit vector on the axis connecting both charges. This equation
is completely analogous with the gravitational attraction between two masses,
as we have seen. Just as we dened the gravity eld g to be the gravitational
force normalized by the test mass, the electric eld E is dened as the ratio of
the electrostatic force to the test charge:
E=
F
,
q0
(3.4)
or, to be precise,
E = lim
q0 0
F
,
q0
(3.5)
Now imagine two like charges of opposite sign +p and p, separated by a

distance d, as in Figure 3.2. At a point P in the equatorial plane, the electrical
elds induced by both charges are equal in magnitude. The resulting eld is
antiparallel with vector m. If we associate a dipole moment vector m with this
3.4. THE MAGNETIC INDUCTION DUE TO A MAGNETIC DIPOLE
87
conguration, pointing from the negative to the positive charge and whereby
|m| = dp, the eld strength at the equatorial point P is given by:
E = Ke
|m|
1 |m|
=
.
r3
40 r3
(3.6)
Figure 3.2:
Next, consider an arbitrary point P at distance r from a nite dipole with
moment m. In gravity we saw that the gravitational eld g (the gravitational
force per unit mass), led to the gravitational potential at point P due to a mass
element dM given by Ugrav = GdM r1 . We can use this as an ad hoc analog
for the derivation of the potential due to a magnetic dipole, approximated by
a set of imaginary monopoles with strength p. To get an expression for the
magnetic potential we have to account for the potential due to the negative
(p) and the positive (+p) pole separately.
With A some constant we can write

1
1
1
1
r+ r
(3.7)
= Ad
Vm = A
r+
r
d
and for small d

1
1 1
1
d r+
r
d r
(3.8)
eq. (3.7) becomes:

Vm = Ad

1
r
(3.9)
d (1/r) is the directional derivative of 1/r in the direction of d. This expression

can be written as the directional derivative in the direction of r by projecting
the variations in the direction of d on r (i.e., taking the dot product between d
or m and r):

1
1
1
(3.10)
=
cos = 2 cos
r
d r
r r
88
Intermezzo 3.2 Magnetic field induced by electrical

current
There is no such thing as a magnetic charge or mass or monopole that
would make a magnetic force a law similar to the law of gravitational or electrostatic attraction. Rather, the magnetic induction is both dened and measured
as the force (called the Lorentz forcea ) acting on a test charge q0 that travels
through such a eld with velocity v.
F = q0 (E + v B)
(3.11)
On the other hand, it is observed that electrical currents induce a magnetic

eld, and to describe this, an equation is found which resembles the electric
induction to to a dipole. The idea of magnetic dipoles is born. In 1820, the
French physicists Biot and Savart measured the magnetic eld induced by an
electrical current. Laplace cast their results in the following form:
dB(P ) =
r
0 dl
i
.
4
r2
(3.12)
An innitesimal contribution to the magnetic induction dB due to a line segment l through which ows a current i is given by the cross product of that
line segment (taken in the direction of the current ow) and the unit vector
connecting the dl to point P .
For a point P on the axis of a closed circular current loop with radius R, the
total induced eld B can be obtained as:
B=
0 R2
0 |m|
=
.
i
2 r 3
2 r 3
(3.13)
In analogy with the electrical eld, a dipole moment m is associated with the
current loop. Its magnitude if given as |m| = R2 i, i.e. the current times the
area enclosed by the loop. m lies on the axis of the circle and points according
in the direction a corkscrew moves when turned in the direction of the current
(the way you nd the direction of a cross product). Note how similar Eq. 3.13
is to Eq. 3.6: the simplest magnetic conguration is that of a dipole.
The denition of electric or gravitational potential energy is work done per
unit charge or mass. In analogy to this, we can dene the magnetic potential
increment as:
dV = B dl
B = V
(3.14)
What is the potential at a point P due to a current loop? Using Eq. 3.12, we
can write Eq. 3.14 as:
0
dV (P ) = i
4
dl
r
dl.
r2
(3.15)
Working this out (this takes a little bit of math) for a current loop small in
diameter with respect to the distance r to the point P and introducing the
magnetic dipole moment m as done above, we obtain for the magnetic potential
due to a magnetic dipole:
V (P ) =
a After
0 m
r
.
4 r 2
Hendrik A. Lorentz (18531928).
(3.16)
3.5. MAGNETIC POTENTIAL DUE TO MORE COMPLEX CONFIGURATIONS89

Since Umonopole 1/r this expression means that the potential due to a
dipole is the directional derivative of the potential due to a monopole. (Note
that is the angle between the dipole axis d and OP (or r) and thus represents
the magnetic co-latitude.)
Just as the Newtonian potential was proportional to GdM , the constant A
must be proportional to the strength of the poles, or to the magnitude of the
magnetic moment m = |m| A = Cpd = Cm. We have, in S.I. units,
Vm =
0 m
0 m r
r
0 m cos
=
=
4r2
4r3
4 r2
(3.17)
3.5 Magnetic potential due to more complex congurations

Laplaces equation for the magnetic potential
In gravity, the simplest conguration was the gravity eld due to a point mass, or
gravitational monopole. After that we went on and proved how the gravitational
potential obeyed Laplaces equation. The solutions were found as spherical
harmonic functions, for which the l = 0 term gave us back the gravitational
monopole.
Magnetic monopoles have not been proven to exist. The simplest geometry
therefore is the dipole. If we can prove that the magnetic eld obeys Laplaces
equation as well, we will again be able to obtain spherical harmonic solutions,
and this time the l = 1 term will give us back the dipole formula of Eq. 3.16.
It is easy enough to establish that for a closed surface enclosing a magnetic
dipole, just as many eld lines enter the surface as are leaving. Hence, the
total magnetic ux should be zero. At the north magnetic pole, your test
dipole will be attracted, whereas at the south pole it will be repelled, and vice
versa. Remember how this was untrue for the ux of the gravity eld: an apple
falls toward the Earth regardless if it is at the north, south or any other pole.
Mathematically speaking, in contrast to the gravity eld, the magnetic eld is
solenoidal. We can write:

B =
B dS = 0.
(3.18)
S
Using Gausss Divergence Theorem just like we did for gravity, we nd that
the magnetic induction is divergence-free and with Eq. 3.14 we obtain that
indeed
2 V = 0.
(3.19)
This equation is known in magnetics as Gausss Law; we will encounter it

again as a special case of the Maxwell equations. We have previously solved Eq.
90
3.19. The solutions are spherical harmonics, so we know that the solution for
an internal source is given by (for r a):
V =a

l l+1

a
Plm (cos )[glm cos m + hlm sin m].
r
m=0
(3.20)
l=1
Intermezzo 3.3 Solenoidal, potential, irrotational

A solenoidal vector eld B is divergence-free, i.e. it satises satises
B = 0
(3.21)
for every vector B. If this is true, then there exists a vector eld A such that
BA
(3.22)
This follows from the vector identity

B = ( A) = 0.
(3.23)
This vector eld A is a potential eld. For a function satisfying Laplaces

equation is solenoidal (and also irrotational). An irrotational vector eld
T is one for which the curl vanishes:
T = 0.
(3.24)
Reduction to the dipole potential

The potential due to a dipole is obtained from Eq. 3.20 by setting l = 1 and
taking the appropriate associated Legendre functions:
VD =
a3 0
[g cos + g11 cos sin + h11 sin sin ].
r2 1
(3.25)
This is valid in Earth coordinates, with the z-axis the rotation axis.The
coecients are (g10 , g11 , h11 ) and they represent one axial (g10 ) and two equatorial
components of the eld. (with g11 taken along the Greenwich meridian). At any
point (r, , ) outside the source of the eld, i.e, outside the core, the dipole
eld can be composed as the sum of these three components.
Earlier, we had obtained Eq. 3.16, which we can write, still in geographical
coordinates, as:
0 1 x
y
z
V =
.
(3.26)
m
+
m
+
m
x
y
z
4 r2
r
r
r
Later, we will see how in a special case, we can take the dipole axis to be
the z-axis of our coordinate system (geomagnetic coordinates or axial dipole
Figure 3.3:
assumption). Then, there is no longitudinal variation of the potential, mz is
the only nonzero component and the only coecient needed is g10 . Comparing
Eqs. 3.25 and 3.26 we see the equivalence of the Gauss coecients glm and
hm
l with the Cartesian components of the magnetic dipole vector:
3 1
m
= 4
0 a g1
x
3 1
(3.27)
my = 4
0 a h1
m = 4 a3 g 0 .
z
Obtaining the magnetic eld from the potential

Weve seen in Eq. 3.14 that the magnetic induction is the gradient of the magnetic potential. It is certainly more convenient to express the eld in spherical
coordinates. To this end, we remind the reader of the spherical gradient operator:
1
1
=
r
+
+
.
(3.28)
r
r
r sin
In other words, the three components of the magnetic induction in terms of
the magnetic potential are given by:
Br
V
r
V
r
V
r sin
(3.29)
We remind that
r points in the direction of increasing distance from the
origin (outwards from the Earth), in the direction of increasing (that is,
southwards) and
eastwards. See Figure 3.4.
92
Figure 3.4:
Geographic and geomagnetic reference frames

It is useful to point out the dierence between geocentric (or geographic) and
geomagnetic reference frames.
In a geomagnetic reference frame, the dipole axis coincides with the z coordinate axis. Since the dipole eld is axially symmetric, it is now symmetric
around the z-axis. This implies that there is no longitudinal variation: no dependence. The components of the eld can be described by zonal spherical
harmonics in the upper hemisphere, eld lines are entering the globe, and
they are leaving in the lower hemisphere. Only one Gauss coecient is necessary: G10 or Mz describe the dipole completely (curled letters used for dipole
reference frame). See Figure 3.5(A).
In Figure 3.5(B), the dipole is placed at an angle to the coordinate axis. To
describe the eld, we need more than one spherical harmonic: a zonal and a
sectoral one. Longitudinal -variation is introduced: m =
0. We need three
Gauss coecients to describe the dipole behavior: g10 , g11 and h11 . Of course the
dipole itself hasnt changed: its magnitude is now [(g10 )2 + (g11 )2 + (h11 )2 ](1/2) =
(G10 )2 . Compared to the dipole reference frame, all we have done is a spherical
harmonic rotation, resulting in a redistribution of the magnitude of the dipole
over three instead of one Gauss coecients.
The angle the magnetic induction vector makes with the horizontal is called
the inclination I. The angle with the geographic North is the declination D.
In a dipole reference frame the declination is indentically zero.
Lets use Eqs. 3.16 and 3.29 to calculate the components of a dipole eld in
the dipole reference frame for a few special angles.
V =
0 |m| cos
,
4
r2
(3.30)
Figure 3.5: Geographic and geomagnetic reference frames and how to represent
the dipole with spherical harmonics coecients in both references systems.
from which follows that
0 |m|
B
= 4
r 3 2 cos = 2B0 cos

r
0 |m|
B = 4 r3 sin = B0 sin
B
= 0.
(3.31)
B0 = (0 |m|)/(4a3 ) = 3.03105T (= 0.303 Gauss) at the surface of the

Earth.
So for the magnetic North Pole, Equator and South Pole, respectively, we
get the eld strengths summarized in Table 3.1.
So the eld at the magnetic equator is half that at the magnetic poles, and at
the North Pole it points radially inward, but outward at the geographic South
Pole.
94
North Pole
Equator
South Pole
0
/2
0
Br
2B0
0
2B0
B
0
B0
0
B
0
0
0
Table 3.1: Field strengths at dierent latitudes in terms of the eld strength at the equator.
In geomagnetic studies, one often uses Z = Br and H = B and E = B .
The magnetic North Pole is in fact close to the geographic South Pole! The
expression for inclination is often given as:
tan I =
Z
cos
=2
= 2 tan1 = 2 cot = 2 tan m .
H
sin
(3.32)
with m the magnetic latitude (m = 90 ) at which the eld line crosses the
point r = a.
Figure 3.6:
3.6
Power spectrum of the magnetic eld
The power spectrum Il at degree l of the eld B is dened as the scalar product
Bl Bl averaged over the surface of the sphere with radius a. In other words,
the denition of Il is:
1
Il =
4a2
2
0
Bl Bl a2 sin dd,
(3.33)
which, with, as weve seen Bl equal to

l
a l+1
m
Bl = a
(glm cos m + hm
l sin m)Pl (cos )
r
m=0
(3.34)
3.6. POWER SPECTRUM OF THE MAGNETIC FIELD
95
From this, the power spectrum for a particular degree l is given by

Il = (l + 1)
l
[(glm )2 + (hlm )2 ].
(3.35)
m=0
The root mean square

(r.m.s) eld strength at the Earths surface for
degree l is dened as Bl = Il and the total r.m.s. eld is given by

B=
12
Il

=

l=1
l=1
(l + 1)
l
21
2
[(glm )2 + (hm
l ) ]
(3.36)
m=0
Il can be plotted as a function of degree l:
10
9
Extrapolation
to CMB
Log10 (Power)
7
6
5
Core Field
4
3
2
Crustal Field
Noise Level
0
-1
0
16
24
32
40
48
56
64
72
Harmonic Degree
Figure by MIT OCW.
Figure 3.7:
The power spectrum consists of two regimes. Up to degree l = 14-15 there

is a rapid roll-o of the mean square eld with degree l . The eld is obviously
predominated by the lower degree terms, and such a so called red spectrum is,
of course, consistent with the spherical harmonic expansion for internal sources,
see Eq. 3.20. This part of the spectrum is due to the core eld. To be more
96
precise, the core eld dominates the spectrum up to degree l = 14-15. At higher
degrees the core eld is obscured by a at white-ish spectrum where the power
no longer seems to depend on the degree, or, alternatively, the wavelength of
the causative anomalies. This part of the spectrum must be due to sources close
to the observation points; it relates to the crustal eld. We will see that this
eld is important even, or, in particular, when one wants to study the magnetic
eld at the surface of the outer core, the core mantle boundary (CMB).
3.7
Downward continuation
In order to study core dynamics or the geodynamo one wants to know what the
magnetic eld is close to its source, i.e., at the CMB.
From Eq. 3.20, we deduce that
Vl (r)
B = Vl (r)
Vl (a)
a l+1
r
Bl (a)(l + 1)
(3.37)
a l+2
r
(3.38)
(3.39)
So for the power spectrum, logically,

Il (r) = Il (a)
a 2l+4
r
(3.40)
with Il (a) the power at Earths surface (r = a).

Lets look at some numbers to illustrate the eect of downward continuation
(see Table 3.2). Consider the (r.m.s.) eld strength at the equator at both the
Earths surface (r = a) and at the CMB (r = 0.54a (so that a/r = 1.82) (Use
eq. (3.36)).
dipole (l=1)
quadrupole (l=2)
octopole (l=3)
Surface (nT)
42,878
8,145 (19% of dipole)
6,079 (14% of dipole)
CMB (nT)
258,493
89,367 (35% of dipole)
121,392 (47% of dipole)
Table 3.2: Field strengths at the surface and the core mantle boundary.
In other words, if the spectrum of the core eld is red at the Earths surface, it is more pink-ish at the CMB because the higher degree components are
preferentially amplied upon downward continuation. However, the amplitude
of the higher degree components (the value of the related Gauss coecients)
is small and, consequently, the relative uncertainty in these coecients large.
Upon downward continuation these uncertainties are of course also amplied, so that at the CMB the higher degree components are large but uncertain
and the observational constraints for them are increasingly weak! We can now
3.8. SECULAR VARIATION
97
also understand why the crustal eld poses a problem if one wants to study the
core eld at the CMB for degrees l > 14: these high degree components will be
strongly amplied upon downward continuation and for high harmonic degrees
the core eld at the CMB will be contaminated with the crustal eld!
3.8
Secular variation
Secular variation is loosely used to indicate slow changes with time of the geomagnetic eld (declination, inclination, and intensity) that are (probably) due
to the changing pattern of core ow. The term secular variation is commonly
used for variations on time scales of 1 year and longer. This means that there is
some overlap with the temporal eects of the external eld, but in general the
variations in external eld are much more rapid and much smaller in amplitude
so that confusion is, in fact, small. From measurements of the components H,
Z, and E, at regular time intervals one can also determine the time derivatives
t glm = g lm and t hlm = h lm , and, if need be, also the higher order derivatives.
The values of glm and hm
l averaged over a particular time interval along with the
m
time derivatives g l and h m
l determine the International Geomagnetic Reference
eld (IGRF), which is published in map and tabular form every 5 year or so.
Temporal variations in the internal eld are modeled by expanding the Gauss
coecients in a Taylor series in time about some epoch te , e.g.,
gem (t)
gem (te )

g
(t te )
t te

2 g (t te )2
+ higher-order terms
2!
t2 te
(3.41)
Most models include only the rst two terms on the right-hand side, but
sometimes it is necessary to include the third derivative term as well, for distance, in studies of magnetic jerks.
Similar to the mean square of the surface eld, we can dene a mean square
value of the variation in time of the eld at degree l:

Il = (l + 1)
[(g lm )2 + (h lm )2 ]
(3.42)
and the relaxation time l for the degree l component as
l =
12
Il
Il
(3.43)
There are at least three important phenomena:

1. Change in the strength of the dipole. We can infer that for the dipole,
the coecients g lm and h lm are all of opposite sign than those of the main
eld. This indicates a weakening of the dipole eld. From the numbers in
98
Table 7.1 of Stacey and eq. (3.43) we deduce that the relaxation time of
the dipole is about 1000 years; in other words, the current rate of change
of the strength of the dipole eld is about 8% per 100yr. Note that this
represents a snapshot of a possibly complex process, and that it does
not necessarily mean that we are headed for a eld reversal within 1000yr.
2. Change in orientation of the main eld: the orientation of the best tting
dipole seems to change with time, but on average, say over intervals of
several tens of thousands of years, it can be represented by the eld of
an axial dipole. For London, in the last 400 year or so the change in
declination and inclination describes a clockwise, cyclic motion which is
consistent with a westward drift of the eld.
3. Westward drift of the eld. The westward drift is about 0.2 a1 in some
regions. Although it forms an obvious component of the secular variation
in the past 300-400 years, it may not be a fundamental aspect of secular
variation for longer periods of time. Also there is a strong regional dependence. It is not observed for the Pacic realm, and it is mainly conned
to the region between Indonesia and the Americas.
Cause of the secular variation

The slow variation of the eld with time is most likely due to the reorganization
of the lines of force in the core, and not to the creation or destruction of eld
lines. The variation of the strength and direction of the dipole eld probably
reect oscillations in core ow. The westward drift has been attributed to either
of two mechanisms:
1. dierential rotation between core and the mantle
2. hydromagnetic wave motion: standing waves in the core that slowly migrate westward, but without dierential motion of material.
Like many issues in this scientic eld, this problem has not been resolved
and the cause of the secular variations are still under debate.
3.9
Source of the internal eld: the geodynamo
Introduction
Over the centuries, several mechanisms have been proposed, but it is now the
consensus that the core eld is caused by rapid and complex ow of highly
conductive, metallic iron in the outer core. We will not give a full treatment
of the complex issues involved, but to provide the reader with some baggage
with which it is easier to penetrate the literature and to follow discussions and
presentations.
3.9. SOURCE OF THE INTERNAL FIELD: THE GEODYNAMO
99
Maxwells Equations
Maxwells Equations describe the production and interrelation of electric and
magnetic elds. A few of them weve already seen (in various forms). In this
section, we will give Maxwells Equations in vector form but derive them from
the integral forms which were based on experiments.
Two results from vector calculus will be used here. The rst we already know:
it is Gausss theorem or the divergence theorem. It relates the integral
of the divergence of the eld over some closed volume to the ux through the
surface that bounds the voume. The divergence measures the sources and sinks
within the volume. If nothing is lost or created within the volume, there will be
no net ux through its surface!

T dV =
n
T dS
(3.44)
V
A second important law is Stokes theorem. This law relates the curl of
a vector eld, integrated over some surface, to the line integral of the eld over
the curve that bounds the surface.

T dl
(3.45)
T dS =
S
1. The magnetic field is solenoidal

We have already seen that magnetic eld lines begin and end at the magnetic dipole. Magnetic charges or monopoles do not exist. Hence,
all eld lines leaving a surface enclosing a dipole, reenter that same surface. There is no magnetic ux (in the absence of currents and outside
the source of the magnetic eld):

B = B dS = 0
(3.46)
Rewriting this with Eq. 3.44 gives Maxwells rst law:
B= 0
(3.47)
2. Electromagnetic induction
An empirical law due to Faraday says that changes in the magnetic ux
through a surface induce a current in a wire loop that denes the surface.

d
d
B =
B dS = E dl,
(3.48)
dt
dt
which can be rewritten using Eq. 3.45 to give Maxwells second law:
E=
B
t
(3.49)
100
3. Displacement current
Weve seen that a time-dependent magnetic ux induces an electric eld.
The reverse is also true: a time-dependent electric ux induces a magnetic
eld. But a current by itself was also responsible for a magnetic eld (see
Eq. 3.12). Both eects can be combined into one equation as follows:

d
0 i + 0 E = B dl
dt
(3.50)
d
The term i is the conductive regular current. The term 0 dt
E , where 0
is the permittivity of free space and has units of farads per meter = A2 s4 kg
m3 , also has the dimensions of a current and is termed displacement
current. Instead of current i we will now speak of the current density
vector J (per unit of surface and perpendicular to the surface) so that
S J ds = i. We can also also use the denition of the electric ux
(analogous to Eq. 3.46) and write the electric displacement vector
as D = 0 E. Then, again using Stokes Law (Eq. 3.45) and dening
B = 0 H, we can write Maxwells third equation:
H = J+
D
t
(3.51)
4. Electric flux in terms of charge density

Remember how we obtained the ux of the gravity eld in terms of the
mass density. In contrast, the ux of the magnetic eld was for a closed
surface enclosing a dipole. For a closed surface enclosing a charge distribution, the ux through that surface will be related to the electrical charge
density contained in the volume! This is a manifestation of the potential
(rather than solenoidal) nature of the electric eld.
We write

q
E dS = ,
0
S
(3.52)
which, with the help of Gausss theorem (Eq. 3.19) transforms easily to
Maxwells fourth law:
D = E
(3.53)
Its interesting to note that, in the absence of conduction or displacement

current, the magnetic eld is both irrotational and solenoidal (divergencefree): B = 0 and B = 0. In that case, there is actually a theorem
that says that B should be harmonic, satisfying 2 B = 0. Hence, the
Maxwell equations imply Laplaces equation: they are more general.
101
Ohms Law of Conduction

A last important law is due to Ohm: it describes the conduction of current
in an electromagnetic eld. Experimentally, it had been veried that a
force called the Lorentz force was exerted on a charge moving in an electric
and magnetic eld, according to:
F = q(E + v B)
(3.54)
This can be transformed into Ohms law which is obeyed by all materials for which the current depends linearly upon the applied potential
dierence. Here is the conductivity, in 1/(Ohm m).
J = (E + v B).
(3.55)
Intermezzo 3.4 Scalar potential for the magnetic field

For a formal derivation of the relationship between the eld and the potential
we have to consider two of Maxwells Equations
H
J + t D
(3.56)
(3.57)
where H is the magnetic eld, B, the induction, J the electric current density
and t D the electric displacement current density. We will use this in the discussion of the geodynamo, but for the study of the magnetic eld outside the core
we make the following approximations. Ignoring electromagnetic disturbances
such as lightning, and neglecting the conductivity of Earths mantle, the region
outside the Earths core (and in the atmosphere up to about 50 km) is often
considered an electromagnetic vacuum, with J = 0 and t D = 0, so that the
magnetic eld is rotation free ( H = 0). This means that H is a conservative
eld in the region of interest and that a scalar potential exists of which H is the
(negative) gradient (but watch out for normalization constants well actually
dene the potential starting from the magnetic induction B rather than from
the eld H by saying that B = Vm ), where B = 0 H. With (3.57) it follows
that such a potential potential must satisfy Laplaces equation (2 Vm = 0 )
so that we can use spherical harmonics to describe the potential and that we
can use up- and downward continuation to study the behavior of the eld at
dierent positions r from Earths center.
The Magnetic Induction Equation

In geomagnetism an important simplication is usually made, known as the
magnetohydrodynamic (MHD) approximation: electrons move according
to Ohms law (steady state), which means that t D = 0. Now,
H = (E + 0 v H)
(3.58)
102
We apply the rotation operator to both sides and use the vector rule
() = () 2 (). With the help of this and the Maxwell equations, Eq.
3.58 can be rewritten as:
1
H
= (v H) +
2 H
t
0
(3.59)
This rephrasing of Maxwells equations is probably one of the most important equations in dynamo theory, the magnetic induction equation. We
can recognize identify the two terms on the right hand side as related to ow
(advection) and one due to diusion. In other words, the temporal change of
the magnetic eld is due to the inow of new material, which induces new eld,
plus the variation of the eld when its left to decay by Ohmic decay.
It is interesting to discuss the two end-member cases corresponding to this
equation, when either of the two terms goes to zero, that is.
1. Infinite conductivity: the frozen flux
Suppose that either the ow is very fast (large v) or that the conductivity
is very large (or both) so that the advection term dominates in eq.
(3.59).
t H = (v H)
(3.60)
It is important to realize that H and v are so-called Eulerian variables:

they specify the magnetic and velocity elds at xed points in space:
H = H(r, t) and v(r, t). The partial derivative is not connected to a
physical body. Now lets consider any area S bounded by a line C. The
surface moves about with the velocity eld v. Consider the ux integrals
of both sides of Eq. 3.60:

S
H n dS =
t

S
(v H) n dS
(3.61)
Using Stokes theorem and the non-commutativity of the vector product,

we obtain:

H (v dl) = 0
(3.62)
H n dS +
C
S t
Using a relationship known as Reynolds theorem, we can transform Eq.
3.62 into:

d
H n dS = 0
(3.63)
dt S
This equation is called the frozen-ux equation. For any surface moving through a highly conductive uid, the magnetix ux B always stays
103
constant. Note that the derivative is a material derivative: it describes

the variation of the ux through a moving surface while it is moving! The
eld lines do not move with respect to the owing material: there is no
change in the electromagnetic eld within a perfect conductor. This is
one of the fundamental approximations used to make problems in dynamo theory tractable and it underlies many computational and theoretical developments in geomagnetism. While it simplies the very complex
magneto-hydrodynamic theory, it is now known that it is probably not
correct. In particular, if one wants to describe eects on a somewhat
longer time scale, say longer than several tens of years, one has to account
for diusion. However, for the description of relatively fast processes the
application of the frozen-ux approximation is appropriate.
2. No flow: diffusion (decay) of the field
Suppose that either there is no ow (v = 0) or that the conductivity
is very low. In both cases the diusion term in eq. (3.59) ((0 )1 2 H)
will control the temporal variation in H. Eectively, eq. (3.59) can be
rewritten as the (vector) diusion equation
t H = (0 )1 2 H
(3.64)
which means that H (=|H|) decays exponentially with time at a rate

(0 )1 = 1 , where is the decay time of the eld. The decay time
increases with conductivity , but unless we consider a superconductor,
the eld will decay. For the earth, this case would represent the situation
that the main eld is due to some primordial eld and that no core ow
is involved. For realistic numbers, the geomagnetic eld would cease to
exist after several tens of thousands of years.
This is a very important conclusion, since it means that the magnetic eld
has to be sustained! because otherwise it dies out relatively quickly (on
the geological time scale). This is one of the primary requirements of a
geodynamo: it has to sustain itself ! (by means of scenario 2)
The diusion equation also shows that the depth to which the ambient eld
can penetrate into conducting material is a function of frequency. This is
an important concept if one wants to use uctuating elds to constrain
the conductivity or if one studies the propagation of changes in core eld
through the conducting mantle and crust.
Consider a magnetic eld that varies over time with a certain frequency
(in practice we would use a Fourier analysis to look at the dierent
frequencies), diusing into a half-space with constant conductivity . It
is straightforward to show that a solution of the vector diusion equation
is
z
H = H0 e ei(t )
(3.65)
104
with
2
0
12
(3.66)
the skin depth, the depth at which the amplitude of the eld has decreased to 1/e of the original value. The skin depth is large for low frequency signals and/or low conductivity of the half-space. Rapidly uctuating elds do not penetrate into the material. (I gave the example
of swimming in still pond: on a winter day you would - probably - not
consider to go swimming even on a nice day with a day time temp of, say,
20 C, whereas you might on a summer day with the same temperature.
The point is that the short period uctuations controled by night-day cycles do not penetrate deep into the water; the temperature of the water
that makes you decide whether or not to go for a swim depend more on
the long period variations due to the changing seasons.) The rapid uctuations of the external eld are used to study in the upper mantle (z
< 1000 km) and the core eld is used to study in the lower mantle.
Geodynamo
So we have a large volume of highly conducting liquid (metallic iron) that moves
rapidly in the Earths outer core. The basic idea behind the geodynamo is that
the rapid motion of part of the liquid in an ambient magnetic eld generates a
current that induces a secondary magnetic eld which is largely carried along in
the uid low (frozen ux) and which reinforces the original eld. In principle,
this concept can be illustrated by Faradays disk generator.
Excess of the light constituent in the outer core is released at the inner
core boundary by progressive freezing out of the inner core. The resulting
buoyancy drives compositional convection in the outer core, and the combination
of convection and rotation produces the complex motion needed for self-excited
dynamo action. The rotation eectively stretches the poloidal eld into toroidal
eld lines (the -eect). Most geodynamo models require a strong toroidal eld,
about 0.01 T (or 100 Gauss), even though this eld cannot be observed at the
Earths surface. These toroidal eld lines are warped up or down due to the
radial convective ow (assuming frozen ux): as a result of the Coriolis force
this results in helical motion, which, in fact, recreates a poloidal component from
a toroidal one (this is know as the -eect). The rotation controls the motion in
such a way that the dipole eld is stronger than any other poloidal component
and, averaged over a sucient time, coincides with the Earths rotation axis.

Intermezzo 3.5 Spheroidal, toroidal, poloidal
The following will help understanding the terminology. An arbitrary vector eld
T on the surface of a unit sphere can be represented in terms of three scalar
elds U , V and W as follows:
r 1 W,
u=
rU + 1 V
(3.67)
where the operator 1 is the dimensionless surface gradient on the unit sphere,
formed by taking the projection of the real gradient onto the plane tangent
to the surface.
rU + 1 V is said to be spheroidal (i.e. having both

A vector eld of the form
radial and tangential components) whereas one of the form

r 1 W is said
to be toroidal. A toroidal eld is purely tangential. It resembles a torus which
is purely circular about the z-axis of a sphere (i.e., follows lines of latitude).
The curl of a toroidal eld (its rotation), is, by denition, poloidal. A poloidal
eld resembles a magnetic multipole which has a component along the z-axis of
a sphere and continues along lines of longitude.
105
106
3.10
Crustal eld and rock magnetism
From spherical harmonic analysis it is clear that short wavelength magnetic

anomalies must have a shallow origin. Within the outer few km of the Earth
are rocks with minerals that have ferromagnetic properties. The study of these
rocks and their magnetization has two important applications in geophysics:
These rocks distort the magnetic eld due to the core and the induced
local eld can be used to investigate crustal structures. (Note that we
have seen that the downward continuation of the crustal eld obscures
the higher degree components of the core eld at the CMB).
Some of these rocks exhibit permanent magnetization (= remanent magnetization with very long, i.e. > 108 yr, relaxation times) and eectively
provide an invaluable record of the history of the magnetic eld and the
relative motion of tectonic units paleomagnetism.
Before discussing paleomagnetism we need to know some of the basics of
rock magnetism in order to study the local eld. In particular:
What are the possible sources of magnetization and what are the conditions that result in a strong, stable magnetization?
What are the important rock types and minerals?
What are the essential aspects of sample preparation before any accurate
paleomagnetic measurements can be done? (I will discuss this only briey.)
The physics of the magnetization of an assemblage of rocks is not simple.
Traditionally, the French have played a major role in research of magnetism,
and L. Neel was awarded a Nobel prize for his pioneering theoretical work on
rock magnetism.
3.11
Magnetization
Strength of magnetization: permeability and susceptibility

Lets start from one of Maxwells equations. Remember that
H = Jmac +
D,
t
(3.68)
where H is the magnetic eld strength and Jmac and D are the macroscopic
current and displacement current densities, respectively. Lets forget about the
displacement current density for a moment (like in the magnetohydrodynamic
assumption). The displacement current is usually spread out over large areas
and hence D and also t

D can be neglected.
In a vacuum, the only current one needs to worry about is the macroscopic
current. In real materials (such as rocks), the atoms and molecules that make
3.11. MAGNETIZATION
107
up the substance are like little magnetic dipoles with a dipole moment m: a
hydrogen atom, for instance is little more than the primitive current loop we
started the denition of the magnetic dipole moment with. The magnetization
of a substance is dened as the volume density of all these little dipole vectors:

m
V
.
(3.69)
M = lim
V 0 V
One can express the density of microscopic molecular currents Jmol through
this magnetization vector as follows:
M = Jmol .
(3.70)
Hence we can rewrite Eq. 3.68 with both the macroscopic and microscopic
current densities as follows (using B = 0 H):
B = 0 Jmac + 0 M,
which leads to

M = Jmac .
0
(3.71)
(3.72)
Comparison with Eq. 3.68 shows that the eld strength in a material is given
with respect to the macroscopic currents as:
H=
B
M,
0
(3.73)
It is customary to associate the magnetization not with the magnetic induction, but with the eld strength. It is assumed that this relationship exists:
M = X H.
(3.74)
The Greek capital letter chi (X) is used for the magnetic susceptibility tensor.
It relates the three components of the internal magnetization to the applied eld,
hence it is a second-order tensor (a matrix). This is usually a complex function,
as X may depend on anything temperature, grain size, H, strain and so
on. It can be negative, too. If X is a tensor, then H and M need not be
collinear. Usually, the assumption of magnetic isotropy is made, and Eq. 3.74
is approximated by a scalar relationship:
M = H.
(3.75)
Now H and M are collinear, but the magnitudes and sense are regulated by
the value and sign of . Because both H and M have dimensions of the eld,
is a dimensionless constant: the magnetic susceptility.
108
So Eq. 3.73 reduces to

H=
B
B
=
.
0 (1 + )
0
(3.76)
A quantity was dened which is called the relative permeability of the

substance. In vacuum, = 1, = 0 and M = 0 no bound charges exist in
free space.
When a magnetic body is placed in an external magnetic eld H the density
of the eld lines inside the body depends on the strength of H and on the
magnetization M induced by H. Thus, the magnetic susceptibility indicates
the ease with which a magnetic body can be magnetized in an external eld.
Types of magnetization
The magnetization of a mineral is controlled by intrinsic (i.e., material dependent) magnetic moments of electrons spinning about their axes (spin dipole
moments) or the motion of electrons in their orbits about the atomic nuclei
(orbital dipole moments). There are several types of spin interactions that give
rise to dierent magnetic eects. The following is a brief summary.
Remanent or Induced? (K
onigsberger ratio)
When we talk about magnetization, we can broadly identify two types:
1. Induced magnetization, MI , which occurs only if an ambient eld H is
present and decays rapidly if this external eld is removed. This induced
eld is very important in ore exploration.
2. Remanent magnetization, MR , which is the part of initial magnetization
that remains after the external eld disappears or changes in character.
MR forms the record of the past eld and is the type of magnetization
that makes paleomagnetism work.
In natural rocks, the ratio between the two is known as the K
onigsberger
ratio Q
Q=
|MR |
|MI |
(3.77)
Rocks with high Q tend to be magnetically stable and are good recorders
of the ancient geomagnetic eld. (I just remark at this stage that the strength
of |MR | does not only depend on the composition (more precisely the type
of magnetic minerals) but also on the grain size, and whether the magnetic
particles consist of single or multiple grains). Rock types with high Q are most
of the mac rocks, such as basalt and gabbro, and also granite. Limestone, for
instance, typically has a very small K
onigsberger ratio.
3.11. MAGNETIZATION
109
Examples of induced magnetization are diamagnetism and paramagnetism.

The elds are weak and decay rapidly when the external eld is removed (by
means of diusion).
Diamagnetism
All materials tend to repel the magnetic eld lines of force so that the density
of eld lines within the body ( |Bi |) is smaller than outside (|Bo |):
Bi = 0 (1 + )|H| < Bo = 0 |H| (1 + ) < 1 < 0
(3.78)
This eect, which is controlled by orbital dipole moments can be explained

by Lenzs Law, which states that the eld produced by a conductor moving
into a magnetic eld tends to oppose the external eld. Here, the induced eld
produced by the electron orbits tends to oppose the external eld. Even though
all materials are diamagnetic, in some the eect is completely overshadowed by
much stronger eects such as ferrimagnetism.
Diamagnetism is a weak eect || < 106 so that Binduced Bfree space .
Owing to diusion, the original state is quickly restored when the external
eld H would be removed unless the conductivity is very large (superconductors) so that the diusion (which scales as 1/, see magnetic induction
equation) can be neglected.
Paramagnetism
Intrinsic paramagnetism is relevant for only a small class of materials, but
most magnetic minerals are paramagnetic above the curie temperature (see
below). Paramagnetic minerals tend to concentrate the lines of force so that
the induced internal eld B is larger than the external eld H.
Bi = 0 (1 + )|H| > Bo = 0 |H| (1 + ) > 1 > 0
(3.79)
Electrons spin in opposite directions giving rise to spin dipole moments in

opposite directions. The spins are arranged in pairs so that the net eect of
110
the magnetic dipoles is zero. However, if the number of orbital electrons is odd
there is a small net magnetic moment that can be aligned with the external eld.
This is known as paramagnetism. Atoms with an even number of electrons tend
to be diamagnetic and those with an odd number tend to be both paramagnetic
and diamagnetic. Like diamagnetism, the eect is small: || < 104 so that
Binduced Bfree space .
The susceptibilities of para- and diamagnetic minerals are virtually independent of the ambient eld (which gives a linear behavior in the above diagrams) but paramagnetism is strongly temperature dependent because thermal
uctuations tend to disorient the alignment with the applied eld (competition
between thermal energy that tends to destroy alingment and magnetic energy
that tends to create it). The Curie Law states that paramagnetic susceptibility
is inversely proportional to the absolute temperature.
Ferro- and ferrimagnetism

This type of magnetism can result in a remanent magnetization which remains even after the ambient eld H changes at a later time. The susceptibility
of ferro- and ferri- magnetic minerals depends strongly on the applied external
eld, but in general
0. In ferromagnetic minerals the electron spins line
up spontaneously. Perfect line-up occurs only in a few metals, such as metallic
iron in the Earths core, and alloys; in other minerals, such as magnetite, the
line-up is not complete which results in ferrimagnetism.
Which minerals are important?

The most important rock-forming minerals with magnetic properties are
magnetite Fe3 O4 = Fe23+ Fe2+ O4
hematite
Fe2 O3 = Fe23+ O3 (can be formed from magnetite by oxidation)
ilmenite
FeTiO3
The magnetic properties of these magnetic minerals and the continuous series of solid solutions between them can be conveniently displayed in ternary
diagram of the FeO TiO2 Fe2 O3 system:
Fe3 O4 = FeO + Fe2 O3 ;
FeTiO3 = FeO + TiO2 ;
ospinel Fe2 TiO4 = FeO + FeTiO3
ulv
The titanomagnetite solid solution series A from magnetiteulv
ospinel consists of strongly magnetic cubic oxides. The titanohematite series B (from
3.11. MAGNETIZATION
A
Ferromagnetism
111
Antiferromagnetism
Spin-canted
Antiferromagnetism
D
Ferrimagnetism
Zero
Spontaneous Magnetic Moment
Figure by MIT OCW.
ilmenitehematite) consists of weakly magnetic rhombohedral minerals. Some
important properties of the minerals in the ternary diagram are: Curie temperatures decrease from right to left; generally, the susceptibility increases from
right to left. Even though, for instance, magnetite has a lower susceptibility
than, say ulv
ospinel , it is more suitable for paleomagnetic studies, because uit
maintains its remanence up to much higher temperatures.
Intermezzo 3.6 Magnetic hysteresis
When a ferromagnetic body (or an assemblage of asymmetric single grains)
is placed in an ambient eld the magnetization will initially increase linearly
with the strength of the ambient eld. This linear part of the magnetization is
reversible and the sample will return to its original state when the ambient eld
is removed. The initial susceptibility is then dened as = (dM/dH)| t=0 . If the
external eld increases in strength saturation will occur, for instance because no
more magnetic domains within a rock can be aligned, and the magnetization M
rotates into the direction of the ambient eld H. This part of the process in nonreversible: if the ambient eld is removed, there may be some relaxation (the
reversible part) but a remanent magnetization MR will remain. If the external
eld would change directions a hysteresis the magnetization would follow a
hysteresis curve. The strength of the magnetization depends on grain size and
temperature: the larger the single grains and the lower the temperature, the
wider the hysteresis curve and the stronger (harder) the magnetization. See
Figure 3.8.
Magnetic domains and inuence of grain size

In bulk magnetic material, say, magnetite, the regions of spontaneous magnetization (magnetic domains) are arranged in patterns to form paths of magnetic
112
TiO2
Anatase Rutile
Brookite
1/3 FeTi2O5
Ferropseudobrookite
1/2 FeTiO3 Ilmenite
1/3 Fe2TiO4
..
Ulvospinel
Pse
-22
3 oC
udo
bro
-40
oki
te s
eri
es
Tit
140
ano
-15
xF
3
hem
-5
eTi
O atite
3 (1140
320
Tit
x)F series
ano
e2 O
x F ma
315
3
e2 T gn
iO etite
460
4 (1x)F series
578 o
e3 O
C
4
FeO
..
Wustite
1/3 Fe3O4
Magnetite
1/3 Fe2TiO5
Pseudobrookite
500
680 o
C
1/2 Fe2O3
Hematite ()
Maghemite ()
Figure by MIT OCW.

ux closure with neighboring domains, so that no net eect is observed. This
is known as the demagnetized state. The size of the single domains in such agglomerates depends on susceptibility and are larger in hematite than in, say, the
titanomagnetites (see below). When placed in an ambient eld H the boundaries between the domains may migrate in such a way as to enlarge the domains
that are favorably oriented with respect to the external eld, and this may cause
magnetic remanence. The process of cell boundary migration is (energetically
speaking) an easier process than the realignment of the direction of magnetization and, as a consequence, large multidomain grains are magnetically soft
(i.e. not stable over long periods of time and sensitive to later changes in the
orientation and strength of the ambient eld).
In contrast, single domain grains have the same direction of spontaneous
magnetization throughout, and if they are not in contact with neighboring grains
they cannot form closed loops and they can more easily be magnetized to saturation.
At high temperatures (or for very small single grains) the kinetic energy, or
the thermal agitation, of the system can be too large for any kind of cooperative
3.11. MAGNETIZATION
113
M
Saturation Magnetization
Ms
Isothermal Remanent
Magnetization
Mrs
Coercive Force
Hcr
Hc
Remanent
Coercivity
Figure by MIT OCW.

Figure 3.8: Magnetic hysteresis.
process to be eective and the electron spins are not aligned but cancel out. In
this state the rocks are paramagnetic, and magnetization can only be induced
by an external eld. If the assemblage is given a magnetization at time t0 the
(induced) magnetization decays a et /0 where 0 , the relaxation time, varies
directly as grain volume (d3 ) and inversely as temperature T . In other words,
the relaxation time becomes shorter exponentially with increasing temperature.
When the relaxation time 0 is very large (say 108 year) the remanence is said
to be permanent. The dependence of the relaxation time on temperature and
the grain size of single grain particles can be illustrated schematically:
When the rock sample cools, the relaxation time and the susceptibility
(Curies Law) increase (the width of the hysteresis curve increases) and there
comes a point, at T = Tc , the Curie temperature (after Pierre Curie), at which
the thermal agitation is no longer large enough to prevent alignment of the magnetic moments. At this point the assemblage of grains acquires a magnetization
in the direction of the ambient eld, which upon further cooling becomes frozen
in. This is known as ThermoRemanent Magnetization (TRM).
The transition from a state of paramagnetism to TRM is not instantaneous
and typically occurs over a certain temperature range. The process of acquiring TRM consists of a series of partial thermo remanent magnetizations, each
114
+ + + +
+ + + - - -
- - - -
- - - + + +
Domain Wall
+ + + +
- - - -
- - - -
+ + + +
A
B
Figure by MIT OCW.
Figure 3.9: Magnetic domains in grains.
acquired in a given temperature range, say, T1 < T < T2 , and each preserving a
record of the ambient eld in that time interval. Reheating to T1 would not destroy that part of the remanent magnetization, but reheating to T2 would. This,
in fact, underlies the principle of thermal cleaning. For most rocks containing
a single, magnetic constituent the partial TRM acquired at about 50 below the
Curie point is dominant. If the assemblage consists of a series of minerals with
dierent Curie points the initial remanence can easily be overprinted with secondary components reecting the eld direction at some later time. For several
minerals along the solid solution lines, the ternary system described above gives
values of the Curie temperature Tc above which ferro- or ferrimagnetism is
not possible. In general the Curie temperature decreases from right to left, i.e.,
hematite to ilmenite and from magnetite to ulvospinel. For hematite Tc =680,
for magnetite Tc =580, and for metallic iron (core!) Tc =770 .
It is important to realize that the strongly magnetic minerals do thus not
necessary result in a strong and stable (= long lasting) magnetization since their
Curie temperatures can be so small that the original magnetization can easily
3.12. OTHER TYPES OF MAGNETIZATION
115
Figure 3.10: Temperature dependence of Curie temperature.

be altered during later thermal events (for some magnetic minerals this can
happen at room temperature). Conversely, the weakly magnetic minerals may,
in fact, produce very stable magnetization. Weak and strong refers to the value
(small, large) of the magnetic susceptibility; stability refers to the time scales
over which the magnetization can last, and this is more a function of grain size
and the composition-dependent Curie and blocking temperatures.
3.12
Other types of magnetization
TRM ThermoRemanent Magnetization

In terms of later overprints, it is good to realize that the direction of TRM
can be reset if the temperature is raised to above the relevant Curie temperature during some later thermal event. Upon cooling a new TRM is frozen
in. Often, however, reheating will also reset the dating clocks since the closure
temperatures for minerals typically used for radio-isotope dating are often lower
than the Curie temperatures for the iron oxides. (This is generally true for the
minerals involved in KAr and Rb-Sr dating hornblende (530 ), muscovite
(350 ), biotite (280 ), apatite (350 ), but U-Pb dating is more borderline
with the closure temperatures for several important minerals exceeding 600 ,
for instance Zircon at 750). As a consequence, the data can still be used for
paleomagnetism; the magnetization just relates to a later thermal event.
DRM Detrital Remanent Magnetization
When igneous rock is eroded and the magnetic constituent is deposited in
suciently still water the magnetic grains that carry TRM from previous events
may become aligned in the ambient eld. Since this type of magnetization is
in fact based on previous TRM it can be a stable and strong magnetization
which can be useful, provided that the time of deposition can be determined
accurately. There are, however, several complications, for instance, the change
in any lineation direction due to compaction may result in an underestimation
if the inclination I and this would underestimate the paleolatitudes.
116
PTRMs
1.0
100
200
Saturation Magnetization
J/J0
Paramagnetism
400
0.6
Curie point
500
0.4
550
0.2
575
100
200
300
Temperature
400
500
600
(oC)
Figure by MIT OCW.

Figure 3.11: Saturation magnetization.
CRM Chemical Remanent Magnetization
Some ferro- or ferri- magnetic minerals may be formed long after the rocks
rst cooled below the Curie temperature. For instance, magnetite may oxidize to hematite which then can settle as cement in the matrix between the
grains. Conversely, magnetite may be formed from hematite in a reducing environment. Upon the chemical reactions and renewed deposition the minerals
are re-magnetized in the then ambient eld. In these examples the CRM is
very strong. Since the acquisition of CRM does not involve any reheating up to
the Curie temperature (which is typically above the blocking temperature for
radioisotope dating!) it is not possible to put a time tag on and it can not be
used for paleomagnetism.
IRM = Isothermal Remanent Magnetization
Exposure of a magnetized sample to very strong eld even without increasing
the temperature (for long enough time) can result in the re-arrangement of part
of the initial TRM. IRM is characterized by a very narrow hysteresis curve so
that the magnetization is relatively weak. An important source for IRM are
lightning strikes.
VRM Viscous Remanent Magnetization
Results from the slight reorganization of the magnetic moment in some grains
if the grain is exposed to an ambient eld for a very long time. The external
eld can diuse into the sample, in particular if both and are large.
3.13
Magnetic cleaning procedures
For paleomagnetic purposes, the overprints by the other types of magnetization

are thus a nuisance; if the secondary remanence is soft their eects can be
Temperature of acquisition of
remanence (oC)
300
0.8
3.14. PALEOMAGNETISM
117
removed by techniques collectively known as magnetic cleaning. Cleaning is

based on the principle that the soft components are destroyed while the strong
original TRM or DRM is preserved. This can not always be guaranteed and, as
a result, the intensity of the eld after cleaning may be unreliable.
Alternating eld cleaning (a.f. cleaning). Makes use of hysteresis properties of ferro- and ferri- magnetic minerals. The strength of a particular
magnetization is determined by the width of the hysteresis curve, which
represents the eld of opposed direction that would have to be applied to
reduce the remanence to zero. The eld is often referred to as the coercive force Hc (see hysteresis curve). In a.f. cleaning one exposes the eld
sample to an alternating eld with diminishing strength and the sample
is rotated in all directions. This process wipes out all components with a
strength Hc that is less than the maximum eld applied. Of course, this
only works if the primary component is stronger than that! This process
must be performed in a laboratory set up where one compensated for the
external Earth eld in order to avoid remagnitization along HE .
Thermal cleaning. Makes use of Curies law and Curie temperatures. The
eld sample is heated to a particular temperature to destroy magnetization
of minerals with Curie temperatures smaller than that value. This is, of
course, only useful if the component youre interested in has a higher Curie
temperature than the temperature that is applied.
Chemical cleaning: leaching or dissolving certain components of the rock,
such a the hematite rich matrix in a sand stone.
In general, one applies progressively stronger cleaning agents to the sample
until the remaining eld is stable and no longer changes in direction. The
intensity of the original TRM/DRM will be obscured in the process!
3.14
Paleomagnetism
A primary objective of paleomagnetism is to determine the history of the geomagnetic eld (for a variety of purposes!) and what is needed is (1) hard
(+stable) magnetization and (2) information about the time at which that magnetization was acquired. So, of the above mechanisms only TRM and DRM are
useful since they have hard magnetization and the timing of acquisition of the
RMs can be determined (radio-isotope dating and stratigraphy, respectively).
Field practice: orientation of the sample

One of the primary objectives is the measurement of the inclination and declination. One has to know where these angles measured in a lab refer to.
Before the rock sample is taken from the eld one, therefore, has to make sure
that its orientation in space at the sample site is known exactly. The current
118
North direction must be marked on the sample, and also the horizontal level.
If it is clear that the sample is taken from a rock setting that has been deformed one must also mark the relation between the reference horizontal and
the original horizontal, or the orientation of the stratication (in sedimentary
rocks). For example, if a sample is taken from marine strata that have been
folded, one must measure the dip of the bedding plane. One must then also
apply techniques that are well developed in structural geology to correct for the
later deformation and retrieve the original inclination.
Figure 3.12: Paleomagnetic orientations.

Ideally, many samples are taken in a sedimentary or volcanic sequence so that
the measurements represent, in fact, a time span of several thousands of years.
In this way the eects of secular variation can be averaged out and one can apply
the so called axial dipole assumption, see below. Core drilling perpendicular
to the stratication is an eective way to do that, but the original azimuth
required to determine the declination may not be preserved, for instance in a
core of a deep marine sediment, so that only the inclination can be used.
Presentation of inclination and declination

The angular information can be used in two ways.
Bauer plots: these are plots (named after Bauer) of the angle of inclination
vs. declination. We have used them to illustrate the secular variation for
London.
Virtual Geomagnetic Pole (VGP) plots.
From the measured inclination I and declination D we can calculate (1)
the magnetic co-latitude (i.e., the angular distance from the paleo-pole to
the sampling site) and (2) the position of the paleomagnetic pole and the
virtual (or apparent) geomagnetic pole (VGP). Lets see how we can do
that.
From the derivation of the potential of an axial dipole we know that the
magnetic co-latitude m can be calculated from the basic equation for paleo-
119
1400
Inclination
1300 1500
200
60oN
1200
300
400
1100
100
1900
0
500
1000
800
70oN
1700
1800
20oW
900
600
700
0o
20oE
Declination
Figure by MIT OCW.
Figure 3.13: Bauer plot.
magnetism:
tan I = 2 cot m = 2tanm
(3.80)
where m is the magnetic latitude (m = 90 - m ). We can also calculate the

location of the pole by measuring out the angular distance m in the direction
of D from North.
This can be done by means of stereographic projections, but we can also use
the sine and cosine rules on a sphere (see Intermezzo).
120

Intermezzo 3.7 Cosine and sine rules on a sphere
Let a, b, and d be the angular distances between 3 points and A, B, and D the
angles as indicated in Figure 3.15.
The cosine rule for a and A is:
cos a = cos b cos d + sin b sin d cos A
(3.81)
and the sine rule:

sin b
sin d
sin a
=
=
sin A
sin B
sin D
(3.82)
(Note that we used (d,D) instead of the (c,C) to be consistent with the use of
D for declination.)
Figure 3.14: Cosine and sine rules on a sphere.

Let , , be the longitude, co-latitude, and latitude of a point, and subscripts p and s indicate the paleopole and the sample site, respectively (m and
m as above). Then we can apply the cosine rule for D:
cos p = cos m cos s + sin m sin s cos D
(3.83)
p = arccos{cos m cos s + sin m sin s cos D}
(3.84)
0 < < 180
(3.85)
and
with
The application of the sine rule gives the longitude of the paleopole relative
to the longitude of the sample site:
sin m
sin(p s )
p s
sin p
sin D
= arcsin
121
(3.86)

sin D sin m
sin p
(Note that arcsin is double valued: one always has to check that one uses the
correct solution).
Figure 3.15: Location of paleopole.

The coordinate (p ,p ) is the apparent position of the pole for a magnetic
eld if the eld is dipolar. The pole determined from a single site is called
the Virtual Geomagnetic Pole (VGP). An important result from paleomagnetism is that the non-dipole eld almost completely averages out over extended
periods of time (10,000 yr +) and that the average of all the VGPs from rocks
of about the same age cluster together about the rotation pole. The is summarized in the Axial Dipole Assumption: if we consider suciently long time
intervals the average VGP represents the paleopole or the paleogeographic
pole.
It was soon realized that the paleogeographic poles did not coincide with the
current rotation pole and that the poles determined from samples of a dierent
age do not necessarily overlap. This was initially interpreted as evidence that
the poles must move, hence the (obsolete) name Polar Wander. We now know
that the poles are xed and that the pole paths describe the relative motion of
the landmass relative to the pole. A rotation of both the pole and the sample
site (continent) about a rotation pole at the present-day equator at and angle
of 90 from (p ,p ) gives the situation at the time of magnetization (assuming
normal direction of the eld). For each separate tectonic unit, measurements
from samples that are magnetized at dierent times determines give a series of
VGPs which dene a pole path , or apparent polar wander (APW) path.
122
Image removed due to copyright concerns.

See paleogeographic map on the book:
Stacey, F. D. Physics of the Earth. 3rd ed.
Brisbane: Brookfield Press, 1992. ISBN: 0646090917.
Figure 3.16:
Note that on VGP diagrams one plots the VGP position relative to a xed
landmass but in reality the poles are the same and the landmass moves!
The observation that the pole paths do not agree for dierent continents
demonstrates that the continents must have moved relative to one another
plate tectonic motion. Conversely, the dierence in pole paths can be used
to reconstruct these movements, and this is one of the basic principles of plate
reconstruction.
Ambiguities in determination of VGPs and in reconstructing past plate motion

The position of the VGP can thus be determined from measurements of the
inclination and declination. There are, however, two important ambiguities.
1. The position of the VGP (p ,p ) is invariant for variation in longitude of
the sample site. The landmass can be anywhere along a small circle at
an angular distance of m about the paleopole. Since many earth systems
(for instance, climate) exhibit a high degree of rotational symmetry and,
thus, primarily zonal variation, this ambiguity is dicult to solve.
2. The declination in the sense used above is poorly dened. In addition to
rotation about the paleopole (which does not eect the calculation of the
VGPs) the sample site may have been subject to rotation around a vertical
axis nearby or through the land mass. In the latter case the uncertainty in
3.15. FIELD REVERSALS
123
N
o
90
NA
(,)
90o
Figure by MIT OCW.

Figure 3.17: Apparent polar wander paths.
the true declination means that the VGP can in principle be anywhere
along a small circle at an angular distance of m about the sample site.
The true paleopole can then be found by the intersection of many such
small circles and VGP estimates. The same applies to data from drill
cores since one does not know the declination.
Once the paleopole is known from many observations, this ambiguity disappears, and from the known pole one can determine the magnetic (co-)latitude,
or, in other words, the N-S motion of the sample site and the relative rotation
about any axis other than the paleopole.
3.15
Field Reversals
For Earth, a reversal can be dened as a globally observed change in sign of

the gaussian coecient g10 that is stable over long (> 5kyr) periods of time.
Whats the relevance of studying eld reversals?
The alternating elds are recorded as magnetic anomalies in newly created
ocean oor. This provides us with a fantastic dating device (the magnetic
reversal time scale) and a very powerful tool to track past plate motions.
It may give us important clues and constraints for our understanding of
core dynamics, the very origin of the main eld, and the possible mechanisms of core-mantle coupling.
124
The magnetic eld creates a protective shield for radiation from space
(solar wind, for instance) and changes in intensity of the eld may also
change the eectiveness of this protection. Major events in the evolutionary process have been linked to variations in paleo-intensity.
Bow Shock
Magnetopause
Earth Radii
Solar Wind
10
20
40
30
Van Allen
Belts
Magnetic Equator
Magnetosheath
Figure by MIT OCW.

Figure 3.18: The magnetosphere.
For navigation purposes it is important to know whats going on with the
eld, for instance whether it is going through a more rapid-than-average
change.
Discovery of eld reversals: a historical note

TRM was well known in the 19th century. Evidence for eld reversals was rst
reported in 1906 when the French physicist Bernard Brunhes discovered that
the direction of magnetization in a lava ow and the adjacent baked clay was
opposite of that of the current main eld. Brunhes concluded that the eld
must have reversed, but that explanation was not generally accepted. Recall
that at that time the origin of the main eld was still a mystery. In fact, it was
in the same year 1906 that Oldham demonstrated the existence of a liquid core
from seismological observations, and that discovery triggered the development
125
of geodynamo theory as we now know it. Later, in 1929 Matuyama published

similar observations as Brunhes, but these also attracted little attention. This
continued until well into the 50ies, by which time there was ample observational
evidence for eld reversals.
However, the necessity for eld reversals was debated and a major proponent
of an alternative explanation was the French physicist Neel who showed that
under certain conditions ferri-magnetic minerals could, in principle, be selfreversing. Self-reversal can happen in a situation where you have a solid solution
of magnetic minerals (or two at least) with dierent Curie temperatures (think
of the hematite-ilmenite series). Then, when the sample cools the mineral with
the highest Curie temperature, say mineral A, freezes in the ambient main eld
while the minerals with lower Curie temperatures are still in paramagnetic state.
If these paramagnetic minerals are very close to mineral A, the remanent eld of
A can induce a eld in the other minerals that may be opposite of the direction
of the main eld. This secondary eld becomes frozen in when mineral B cools
below its own Curie temperature, etc. etc. In case the mineral with the lower
Curie temperature has a larger susceptibility (and this is indeed the case for
the minerals along the solid solution lines in the ternary diagram!) the reversed
eld may dominate the combination of the ambient eld and the induced eld
in mineral A, and the net eect would be opposite in direction of the main eld
without requiring a reversal of that main eld. Even though it has been shown
experimentally that self-reversal is possible, it requires very special physical
conditions and it is NOT a plausible mechanism to explain the by then (mid
50-ies) overwhelming evidence that reversals are quite common.
Study of baked clays (or thermal aureoles of igneous bodies in general) at
many sites world wide demonstrated that the direction of magnetization in the
igneous body of the same age and the contact aureole were often the same but
did not coincide with that of the adjacent non-metamorphosed rocks. It was
found that the rocks showing reversals grouped in specic age intervals, and this
argument became even more convincing when accurate radiometric techniques
such as KAr dating were applied. The reversals appeared to be rather global
phenomena and could be traced in volcanic rocks as well as in marine sediments.
There appeared to be no chemical dierence between the rocks the exhibited
normal and reversed orientations.
By the time of the early 60-ies several important discoveries associated with
sea oor spreading resulted in the general acceptance of eld reversals.
Outstanding questions
1. What happens to the dipolar eld during reversals? Does the dipole eld
decay to almost zero before it is regenerated in the opposite direction, or
is the dipole eld perhaps weakened but still prominent during reversals?
The measurement of paleointensity of the main eld is dicult and prone
to large uncertainties (see also notes on magnetic cleaning). Yet, it is
now well established that during reversals the intensity of the eld reduces
by at least a factor of 3. Recall that in the present-day eld the dipole
126

makes up for about 90% of the eld and that the non-dipole component
is only about 10% . So although the eld reduces to about 30% of its
original strength, the dipolar component cannot be neglected during reversals. These numbers are, however, not yet well constrained. Some data
(and numerical models) suggest a reduction during reversal to about 10%
of the original strength, which would either imply that the dipole component has become negligible or that if there still is a dipolar eld also
the non-dipole components must signicantly reduce in strength. The observational evidence is problematic because [1] the non-dipole eld at the
surface is so much weaker than it is at the CMB and [2] because a reversal
is a relatively short phenomenon (say 5,000 yr) and it is hard to get the
time resolution in volcanic or sedimentary strata.
2. A closely related question is: can a non-dipole eld produce continuous

VGP paths during reversals? Are VGP paths random or are there are
preferred longitudes along which the VGPs move from the one pole to the
other, and what are the implications of the answer to this question for our
understanding of core dynamics and core-mantle coupling?
Mechanisms of reversals
The mechanisms of eld reversals are still subject of spirited debate, but it is
generally accepted that they are controlled by core dynamics. Some important
constraints (and, indeed, the need for core-mantle coupling) can be inferred
directly from general properties of the geodynamo and from studying the (statistical) behavior of the eld during and in between reversals.
1. It is important to realize that the geodynamo does not care about the
sense of the eld. The magnetic induction equation H/t = (v
H) + (40 )1 2 H is invariant to a change in sign of H.
We have argued before that in the core we can largely ignore the diusion
term, and it is clear that the (sign) of the changes in magnetic eld depend
entirely on how the turbulent ow eld v interacts with the magnetic eld
H.
2. The typical relaxation time of core processes is much longer than the time
interval in which most reversals take place. If the eld is not sustained by
dynamo action, for instance if the velocity v in the above equation goes
to zero, the eld will die out in several 104 years. The rate of reversals
must mean that the core eld is actively destructed and regenerated. Part
of dynamo theory is the complex interaction of and conversion between
the poloidal and (unseen) toroidal eld. Dierential (radius dependent)
rotation in the outer core stretches the poloidal eld into a toroidal eld;
radial ow owing to compositional convection distorts the purely toroidal
eld (and the coriolis force adds extra complexity in the form of helical
motion) which gives a component along the poloidal eld. It is conceivable
127
that turbulence produces local changes in the direction of the helicity of

the toroidal eld and thus the poloidal component, which may trigger a
reversal.
3. Reversals are not simply extreme cases of secular variation even though
the distinction is vague. Growth, decay, and drift of the non-dipole eld
and of the axial dipole are continuous, albeit possibly random, processes.
Reversals, however, occur over time scales that are longer than the typical secular variations, but much shorter than the average time interval
between reversals (> 105 yr). In other words, statistically speaking reversals do not occur as frequently as one might expect from the continuous
(secular) uctuations of the main eld. It resembles a chaotic system that
switches back and forth between two quasi-stable states. This behavior
is nicely illustrated by measurements of the inclination, declination, and
intensity during a reversal that characterizes the lower boundary of the
Jaromillo event (a short N event in the Reversed Matuyama epoch documented near a Creek in Jaromillo, Mexico:
90
Inclination (degrees)
Declination (degrees)
90
360
270
180
30
0
-30
-60
-90
Intensity (Am-1)
90
60
0.1
0.01
Depth (metres)
9.5
9.0
0
10.0
8
10 11
Relative Age (thousands of years)

Figure 3.20. Detailed record of a reversal from a rapidly deposited deep-sea sediment
core. Samples were "cleaned" in a 10-2 T alternating field before measurement. This
reversal marks the lower boundary of the Jaramillo polarity event.
Figure by MIT OCW.
128
3.16 Qualitative arguments that explain the need

for core-mantle coupling
There are compelling reasons to believe that the mantle and the dynamic processes in the mantle have a strong inuence on reversals (and thus perhaps on
core dynamics in general). It is probable that core ow is too turbulent and
chaotic to be able to organize systematic patterns in eld reversals. Reversals
seem to be a random process in the sense that [1] N and R directions of the
Earths magnetic moment occur equally often and [2] the probability that a
reversal happens in the next t years is independent of the time since the last
reversal. However, as mentioned above, the average time interval between reversals is much larger than expected from the continuous uctuations in the
eld. Moreover, in the last 5 year several papers have been published that presented evidence for preferred reversal paths of VGPs along meridians through
the Americas and through east Asia. The selection of the same VGP path over
and over again seems to require that the core has some memory about what
happened many thousands of years ago, which is not very likely.
In contrast, the mantle, with its high viscosity has such memory since
the mantle structure does not change much over periods that are smaller than
1 Myr. For this reason, investigators have invoked a strong inuence of (or
coupling with) the mantle. This coupling could be mechanical (topography on
core mantle boundary) but that is not very likely. Alternatively, the coupling
could be electro-magnetical; in this mechanism ux out of the core would be
concentrated in regions where the conductivity of the lowermost mantle is lowest
(see dynamo equation above!) and swept away from regions of high conductivity.
Several investigators have related the patches of high and low mantle conductivity to lateral variation in the thickness of the so called D layer and this
structure is likely to be inuenced by ow in the entire mantle. This depicts a
complex system where mantle convection controls the lateral variation in thickness of the D layer (thin beneath major downwellings (slabs) Americas and
east Asia and thicker beneath upwellings plumes (central Pacic, Hawaii))
and thus if D is anomalously conductive lateral variation in mantle conductivity that produces torques on the core and creates windows for ux bundles
from the core into the mantle. The research in this exciting eld is very much
ongoing and no conclusive models has been developed yet.
3.17 Reversals: time scale, sea oor spreading,

magnetic anomalies
We have talked in some detail about the implications of the reversals of the
magnetic eld for our understanding of core dynamics and core-mantle coupling. This is a eld of active research: rapid progress is being made, but many
outstanding questions remain to be answered. About 30-35 years ago (rst half
of the 60-ies) the reversals of the main eld played a central role in the devel-
3.17. REVERSALS: TIME SCALE, SEA FLOOR SPREADING, MAGNETIC ANOMALIES129

opment of the concept of sea oor spreading, which resulted in the acceptance
of plate tectonics.
To appreciate the line of thoughts it is good to realize that by the end of
the 50-ies there was no agreement on the issue of eld reversals, even though
an increasing number of scientists had accepted the idea. At the same time,
continental drift was still hotly debated. Since the idea was rst clearly postulated by Alfred Wegener in 1912 it had been advocated by many scientists, for
example the South African DuToit, but, in particular, in the US the concept
met with continued resistance. A central issue was the driving force and the
mechanism of continental drift; one did not believe it possible that continents
would plow their way through the mac and strong oceanic lithosphere.
1. Initially, the evidence for reversals was mainly based on land samples
of volcanic rocks. It was recognized that rocks with the same magnetic
polarity grouped into distinct time intervals, and the boundaries of these
time windows could be dated accurately with radiometric methods. The
rst Geomagnetic Polarity Time Scale (GPTS) was based on KAr and
40
Ar/39 Ar dating of rocks less than 5 Ma old. Based on the rst time
scales one dened epochs as relatively long periods of time of mostly
one polarity, and events as relatively short excursions or reversals of the
eld within the epochs. The rst (= most recent) epochs are named after
some of the pioneers in geomagnetism (Brunhes, Matuyama, Gauss, and
Gilbert); events are typically named after the location where it was either
rst discovered or where it is best represented (Jaromilla creek, Mexico;
Olduvai gorge, Tanzania). The current epoch is referred to as normal (or
negative since the magnetic moment direction is opposite of the angular
moment direction). As more data became available the time tables were
rened and it became clear that the distinction between epochs and events
is not so useful.
2. In the early 60-ies, marine expeditions and geomagnetic surveys resulted

in the rst detailed maps of magnetic anomalies at the ocean oor. A
stripe pattern formed by alternating strength of the magnetic eld (the
anomalies were made visible after subtracting the values of the reference
eld (IGRF)) was documented for the Pacic sea oor o the coast of
Oregon, Washington, and British Columbia (Ra & Mason, 1961), and for
the Atlantic just south of Iceland. The origin of this pattern was, however,
not known. Explanations included the lateral variation in susceptibility
130

owing to variations in composition or changes in stress due to buckling
(folding) of the oceanic lithosphere (it was known since the large 1906
earthquake in San Francisco that changes in the stress eld can inuence
magnetization).
Image removed due to copyright concerns.

See Figure 3.7 on Geol. Soc. Am. Bull. 72 (1961): 1267-1270.
3. At about the same time, Dietz (1961) and Hess (1962) postulated the
concept of sea oor spreading. Not as an explanation for the magnetic
anomalies, but rather as a way out of the long standing problem that
continental drift seemed unlikely because continents cannot move through
the oceanic lithosphere. Dietz and Hess realized that the oldest parts of
oceanic sea oor were less than about 200 Ma old, which is more than an
order of magnitude less than the age of the oldest continental rocks. This
suggests that oceanic lithosphere is being recycled more eciently than
continental material. These scientists proposed that sea oor is generated
at mid oceanic ridges (where pillow lavas had been dredged and high heat
ow measured) and then spreads away carrying the continents with them.
So the continents move along with the ocean oor, not through it!!
In 1963 Vine and Matthews and, independently, Morley, put these ideas
and developments (1-3) together and correctly interpreted the magnetic
anomalies at the sea oor: new ocean oor is created at mid oceanic ridges
the basalt cools below the curie temperature and freezes in the direction
3.18. MAGNETIC ANOMALY PROFILES
131
of the then current magnetic eld (normal (N) or reversed (R)) the N-R
pattern spreads away from the ridge. In this way the ocean oor acts as
a tape recorder of the polarity of the Earths magnetic eld in the past.
Initially, the time table based on land volcanics was correlated with the
stripe pattern in the oceans, but this correlation was restricted to the
past 5 Ma. Later, paleontological data (the fossil record) was used (along
with radiometric calibration of land volcanics)) to extend the reversal time
scale to ocean oor of Jurassic age (160 Ma). This was rst done in the
south Atlantic where the time scale was extended to 80 Ma by assuming
that the rate of sea oor spreading was constant in that time interval,
an assumption that was later justied). With the increasing amount of
data, the time scale is continually being updated. The time scale is very
important for two main reasons:
Using the GPTS allows the calculation of sea oor spreading rates
by measuring the distance between identied lineations.
Identication of the magnetic pattern of the sea oor and matching
with the time scale can be used to date a piece of oceanic crust.
3.18
Magnetic anomaly proles
In order to exploit the time scales (for instance for points (a) and (b) above) one
must be able to identify and interpret magnetic anomalies, or, more specically
for marine surveys, magnetic proles due to a succession of magnetic anomalies
with reverse polarity. In the interpretation one usually tries to nd a combination of N and R blocks that best matches the observed prole. This is easier
said than done. A visual inspection of any of the proles used as illustrations in,
for instance, it can be demonstrated that the relationship between the simplied
block models and the corresponding magnetic proles is far from trivial. The
magnetic proles look more complex than you would expect from the simple
block models, and this is true not only for real data but also for the synthetic
(computer generated) proles.
In the rst place, the real sea oor is not magnetized in the form of regular
geometric bodies. The geological processes at work in a mid oceanic ridge
environment are complex and so are the resulting patterns of magnetization. In
particular, the creation of new ocean oor in the so called emplacement zone
is an irregular process that is not necessarily symmetric across the ridge axis.
Also the spreading rate has an eect in that fast spreading (Pacic rise) results in
wide zones with the same magnetization which facilitates identication, whereas
slow spreading (Atlantic) produces narrow anomalies with poor separation of
the anomalies, which complicates the analysis.
But even for a regular alternation of N and R polarity, say in a computer
model, the magnetic proles can be complicated. This is mainly due to the
dipole character of the magnetic eld as opposed to the monopoles in gravity,
which means that the orientation of the anomaly plays an important role.
132
Before we look at this in more detail, lets dene what we mean by a magnetic anomaly in the context of marine surveys. At sea one uses proton precession magnetometers (see, for instance, Garland, 1979) that are towed behind
the ship. These instruments make use of the precession of hydrogen atoms after
a strong (articial) ambient eld generated by a current in a coil surrounding a
contained with many hydrogen atoms (e.g., water!) is switched o. When the
initial eld that aligns the hydrogen dipole moments is switched o the hydrogen atoms precess with a frequency that depends on the strength of the ambient
eld. The precession induces a current in the coil, which can be measured. Precession magnetometers measure the magnitude of the total magnetic eld, not
its direction (which means that they do not have to be orientated themselves!)
and have a sensitivity of about 1 nT. The magnetic anomalies are determined
by subtracting the IGRF from the observed values.
In vector notation, the total eld B is given by the contributions of the main
eld (the core eld) BE and the magnetized body B:
B = BE + B
(3.87)
But the scalar magnetometer measures

B = |B| = BE + A
(3.88)
with A the magnetic anomaly.

B
B
= |B| = (B B) 2 = {BE BE + 2BE B + B B} 2

=
2
{BE
+ 2BE B + B }
(3.89)
1
2
Typically, B is of the order of several hundreds of nT whereas BE is of the

order of 20,000 to 60,000 nT. Therefore we can safely neglect B 2 so that
1
2
E B|BE |1 } 21
B = {BE
+ 2BE .B} 2 = BE {1 + 2B
(3.90)
E in the direction of BE . The second term is small so

with the unit vector B
that this expression can be expanded in a binomial series
E B
B BE {1 + BE B|BE |1 } = BE + B
(3.91)
and we dene the magnetic anomaly A, see (3.88) as

E B
A = B BE = B
(3.92)
In other words, the only components of the local eld B that we measure
is the one parallel to the direction of the ambient (core eld). A vector diagram
shows that this is a good approximation since B BE . (For a dipole, the eld
rr3 }.)
is given by the gradient of the potential B = V = (0 /(4)1 m

Lets consider some examples in which BE is approximated by an axial
dipole. Assume an innitely long block (lineation) that is magnetized in the
133
direction of the ambient Earth eld. The magnetization direction can be decomposed into components parallel and perpendicular to the strike of the block.
It can be shown that the component | to the strike does not contribute to a
magnetic anomaly since the innitely small dipoles that make up the anomalous
structure all lie head-to-toe and cancel out except for the ones near the end
(which innitely far away (in other words, the eld lines never leave the body).
In this geometry, the magnetized body only has magnetization in the directions
to the strike (say, in the x and z direction). Consequently, a N-S striking
E B = 0 since
anomaly magnetized at the equator will always have A = B
B is orthogonal to BE (or, more precisely, there is no non-zero component of
E . In contrast, lineations magnetized at the pole and
B in the direction of B
E |B so that there will be a strong signal centered
measured at the pole have B
over the anomaly.
The eect of the orientation (which theoretically speaking gives rise to a
phase shift) presents itself as a skewness of the anomalies (see diagram) which
can be corrected for (for instance by reducing the anomaly to the pole, see
diagram).
In addition to these distortions due to the phase shift theres the eect
of amplitude modulation. Amplitude modulation consists, in fact, of two
contributions:
1. usually, the magnetized body (at the ocean oor) is several km beneath
the vessel. The observed magnetic eld thus represents the upward continuation of the eld at the source. As a result of the 1/r3 decay of the eld
there is signicant spatial attenuation of the high frequency components.
Sharp contrasts will be observed as more gradual transitions.
2. Since the magnetized body can be thought to consist of dipoles a homogeneously magnetized lineation will only display a magnetic anomaly near
its ends; in between, the plus and minus poles cancel out (same argument
as above) and no anomalous eld is measured. This results in an eect
known as sagging. See diagram on next page.
134
S
a
N c
d S
Inducing magnetic field
Bz > 0
Anomalous field of
magnetized body
Component of anomalous
field parallel to inducing
field
a + b+ c +d
a
b
c
Bz < 0
Fig. 3.24. Explanation of the origin of the magnetic anomaly of an infinitely
long vertical prism in terms of the pole distributions on top, bottom and side
surfaces, when the magnetic field (or magnetization) is inclined.
Figure by MIT OCW.
135
Bz
Bz
Bz
Bz
Figure 3.25. The effect of block width on the shape of the magnetic anomaly
over a vertically magnetized crustal block.
Figure by MIT OCW.
136
Present-day
Field
Ridge Axis
J B
B J
Figure. 3.26. Explanation of the shape of a magnetic profile across an oceanic

spreading center: (A) the anomalies of individual oppositely magnetized crustal
blocks on one side of the ridge, (B) overlap of the individual anomalies, (C) the
effect for the opposite sequence of blocks on the other side of the ridge, and
(D) the complete anomaly profile.
Figure by MIT OCW.
Chapter 4
Seismology
4.1
Historical perspective
1678 Hooke Hookes Law F = c u (or = E)

1760 Mitchell Recognition that ground motion due to earthquakes is related to wave
propagation
1821 Navier Equation of motion
1828 Poisson Wave equation
P & S-waves
1885 Rayleigh Theoretical account surface waves
Rayleigh & Love waves
1892 Milne First high-quality seismograph begin of observational period
1897 Wiechert Prediction of existence of dense core (based on meteorites Fe-alloy)
1900 Oldham Correct identication of P, S and surface waves
1906 Oldham Demonstration of existence of core from seismic data
1906 Galitzin First feed-back broadband seismograph
1909 Mohorovi
ci
c Crust-mantle boundary
1911 Love Love waves (surface waves)
1912 Gutenberg Depth to core-mantle boundary : 2900 km
1922 Turner location of deep earthquakes down to 600 km (but located some at 2000 km,
and some in the air...)
1928 Wadati Accurate location of deep earthquakes
Wadatai-Benio zones
1936 Lehman Discovery of inner core
1939 Jereys & Bullen First travel-time tables
1D Earth model
1948 Bullen Density prole
1977 Dziewonski & Toks
oz First 3D global models
1996 Song & Richards Spinning inner core?
Observations :
1964
1960
1978
1980
ISC (International Seismological Centre) travel times and earthquake locations

WWSSN (Worldwide Standardized Seismograph Network) (analog records)
GDSN (Global Digital Seismograph Network) (digital records)
IRIS (Incorporated Research Institutes for Seismology)
137
CHAPTER 4. SEISMOLOGY
138
4.2
Introduction
With seismology1 we face the same problem as with gravity and geomagnetism;
we can simply not oer a comprehensive treatment of the entire subject within
the time frame of this course. The material is therefore by no means complete.
We will discuss some basic theory to show how expressions for the propagation of
elastic waves, such as P and S waves, can be obtained from the balance between
stress and strain. This requires some discussion of continuum mechanics. Before
we do that, lets look at a very brief and incomplete overview of the historical
development of seismology. Modern seismology is characterized by alternations
of periods in which more progress is made in theory development and periods
in which the emphasis seems to be more on data collection and the application
of existing theory on new and often better quality data. Its good to realize
that observational seismology did not kick o until late last century (see section
4.1). Prior to that seismology was eectively restricted to the development
of the theory of elastic wave propagation, which was a popular subject for
mathematicians and physicists. For some important dates, see attachment above
table (this historical overview is by no means complete but it does give an idea
of the developments of thoughts). Lay & Wallace (1995) give their view on
the current swing of the research pendulum in the following tables (with source
related issues listed on the left and Earth structure topics on the right) :
Classical Research Objectives
A. Source location
(latitude, longitude, depth)
B. Energy release
(magnitude, seismic moment)
C. Source type
(earthquake, explosion, other)
D. Faulting geometry
(area, displacement)
E. Earthquake distribution
A. Basic layering
(crust, mantle, core)
B. Continent-ocean dierences
C. Subduction zone geometry
D. Crustal layering, structure
E. Physical state of layers
(uid, solid)
Table 4.1: Classical Research objectives in seismology.

We will discuss some classical concepts and also discuss some of the more
current topics. Before we can do this we have to deal with some basic theory.
In principle, what we need is a formulation of the seismic source, equations to
describe elastic wave propagation once motion has started somewhere, and a
theory for coupling the source description to the solution for the equations of
motion. We will concentrate on the former two problems. The seismic waves
1 From the Greek words o (seismos), earthquake and oo (logos), knowledge. In
that sense, earthquake seismology is superuous.
4.2. INTRODUCTION
139
Current Research Objectives

A. Slip distribution on faults
B. Stresses on faults
and in the Earth
C. Initiation/termination
of faulting
D. Earthquake prediction
E. Analysis of landslides,
volcanic eruptions, etc
A. Lateral variations
(crust, mantle, core?)
B. Topography on internal
boundaries
C. Anelastic properties
of the interior
D. Compositional/thermal
interpretations
E. Anisotropy
Table 4.2: Current research objectives in seismology (after Lay & Wallace
(1995))
basically result from the balance between stress and strain, and we will therefore
have to introduce some concepts of continuum mechanics and work out general
stress-strain relationships.
Intermezzo 4.1 Some terminology

For most of the derivations we will use the Cartesian coordinate system and
denote the position vector with either x = (x1 , x2 , x3 ) or r = (x, y, z). The
displacement of a particle at position x and time t is given by u = (u1 , u2 , u2 ) =
u(x, t), this is the vector distance from its position at some previous time t0
(Lagrangian description of motion). The velocity and acceleration of the particle
are given by u
= u/t and u
= 2 u/ 2 t, respectively. Volume elements are
denoted by V and surface elements by S. Body (or non-contact) forces, such
as gravity, are written as f and tractions by t. A traction is the stress vector
representing the force per unit area across an internal oriented surface S within
a continuum, and this is, in fact, the contact force F per unit area with which
particles on one side of the surface act upon particles on the other side of the
surface.
= c2 2 u,
A general form of a wave equation is 2 u/ 2 t = c2 2 u/ 2 x or u
which is a dierential equation describing the propagation of a displacement
disturbance u with speed c.
We will see that the fundamental theory of wave

propagation is primarily
based on two equations : Newtons second law ( F = ma = m 2 u/ 2 t) and
Hookes constitutive law F = cu (stating that the extension of an elastic material results in a restoring force F, with c the elastic (spring) constant (not wave
speed as in the box above!). In one dimension, Hookes law can also be formulated as the proportionality between stress and strain , with proportionality
140
factor E is Youngs modulus : = E. We will see that this linear relationship
between stress and strain does not hold
in 2D or 3D, in which case we need the
so-called generalized Hookes Law. For
F = ma we have to consider both the
non-contact body forces, such as gravity that works on a certain volume, as well
as the contact forces applied by the material particles on either side of arbitrary
and imaginary internal surfaces. The latter are represented by tractions (stress
vectors). We therefore have to look in some detail at the denitions of stress
and strain.
4.3
Strain
The strain involves both length and angular distortions. To get the idea, lets
consider the deformation of a line element l1 between x and x + x.
Due to the deformation position x is displaced to x + u(x) and x + x to x +

x + u(x + x) and l1 becomes l2 .
The strain in the x direction, xx , can then be dened as

xx =
l2 l1
u(x + x) u(x)
=
l1
x
(4.1)
If we assume that x is small we can linearize the problem around the reference
state u(x) by using a Taylor expansion on u(x + x) :

u
u
u(x + x) = u(x) +
x + O(x2 ) u(x) +
x
(4.2)
x
x
so that
xx =
u(x)
x
1
=
2
u(x) u(x)
+
x
x

(4.3)
which represents the normal strain in the x direction. Similar relationships

can be derived for the normal strain in the other principal directions and also for
the shear strain xy and xz (etc), which involve the rotation of line elements
within the medium.
The general form of the strain tensor ij is
ij
=
=

1 ui
1 u(xi ) u(xj )
uj
=
+
+
2
xj
xi
2 xj
xi

ui
1 uj
= ji
+
2 xi
xj
(4.4)
4.4. STRESS
141
with normal strains for i = j and shear strains for i = j. (In this discussion
of deformation we do not consider translation and/or rotation of the material
itself). Equation (4.4) shows that the strain tensor is symmetric, so that there
the maximum number of dierent coecients is 6.
4.4
Stress
Stress is force per unit area, and the principle unit is Nm2 (or Pascal : 1Nm2 = 1Pa).
Similar to strain, we can also distinguish between normal stress, the force
F per unit area that is perpendicular to the surface element S, and the shear
stress, which is the force F per unit area that is parallel to S (see Fig. 4.1).
The force F acting on the surface element S can be decomposed into three
components in the direction of the coordinate axes : F = (F1 , F2 , F3 ). We
further dene a unit vector n
normal to the surface element S. The length of
n
is, of course, |
n| = 1.
For stress we dene the traction as a vector that represents the total force
per unit area on S. Similar to the force F, also the traction tt can be decomposed into t = (t1 , t2 , t3 ) = t1 x1 + t2 x2 + t3 x3 . The traction t represents the
total stress acting on S.
In order to obtain a more useful denition of the traction t in terms of
elements of the stress tensor consider a tetrahedron. Three sides of the tetrahedron are chosen to be orthogonal to the principal axes in the sense that si
is orthogonal to xi ; the fourth surface, S, has an arbitrary orientation. The
stress working on each of the surfaces of the tetrahedron can be decomposed
into components along the principal axes of the coordinate system. We use the
following notation convention : the component of the stress that works on the
plane x1 in the direction of xi is 1i , etc.
Figure 4.1: Stress balancing in the stress tetrahedron.

If the system is in equilibrium then a force F that
works on S must be cancelled
by forces acting on the other three surfaces :
Fi = ti S 1i s1 2i s2
3i s3 = 0 so that ti S = 1i s1 + 2i s2 + 3i s3 . We know that the
expression we are after should not depend on our choice of s nor on S (since
142
the former were just chosen and the latter is arbitrary). This is easily achieved
by realizing that S and S are related to each other : si is nothing more than
the orthogonal projection of S onto the plane perpendicular to the principal
, the normal to S, and
axis xi : si = cos i S , with i the angle between n
xi . But cos i is in fact simply ni so that si = ni S. Using this we get :
ti S = 1i n1 S + 2i n2 S + 3i n3 S
(4.5)
ti = 1i n1 + 2i n2 + 3i n3
(4.6)
or
Thus : the ith component of the traction vector t is given by a linear combination
of stresses acting in the ith direction on the surface perpendicular to xj (or
parallel to nj ), where j = 1, 2, 3;
ti = ji nj
(4.7)
Conversely, an element ji of the stress tensor is dened as the ith component

of the traction acting on the surface perpendicular to the j th axis (xj ) :
ij = ti (xj )
(4.8)
The 9 components ji of all tractions form the elements of the stress tensor.
It can be shown that in absence of body forces the stress tensor is symmetric
ij = ji so that there are only 6 independent elements :
11 12 13
11 12 13
ij = 21 22 12 =
22 12
(4.9)
31 32 13
13
The normal stresses are represented by the diagonal elements (i=j) and the
shear stresses are the o diagonal elements (i = j). We can diagonalize the
stress tensor by changing our coordinate system in such a way that there are
no shear stresses on the surfaces perpendicular to any of the principal axes (see
Intermezzo 4.2). The stress tensor then gets the form of

0
0
0
1 0
11
0 = 0 2 0
(4.10)
ij = 0 22
0
0 33
0
0 3
Some cases are of special interest :
uni-axial stress : only one of the principal stresses is non-zero, e.g.
1 = 0, 2 = 3 = 0
plane stress : only one of the principal stresses is zero, e.g. 1 = 0,
2 , 3 = 0
4.5. EQUATIONS OF MOTION, WAVE EQUATION, P AND S-WAVES 143

pure shear : 3 = 0, 1 = 2
isotropic (or, hydrostatic) stress : 1 = 2 = 3 = p (p = 31 (1 +
2 + 3 )) so that the deviatoric stress, i.e. the deviation from hydrostatic
stress is written as :
0
0
1 p
0
2 p
0
ij
=
0
0
3 p
(4.11)
Equations of motion, wave equation, P and

S-waves
4.5
With the above expression for the (symmetric) strain tensor (Eq. 4.4) and the
denitions of the stress tensor ij and the traction ti , we can formulate the basic
expression for the equation of motion :

Fi
=
V

fi dV +
=
V
ti dS
(4.12)
fi dV +

ij nj dS =
2 ui
dV = mai
t2
If we apply Gauss divergence theorem2 , this can be rewritten as

V
2 ui
2 dV
t
2 ui
t2

ij
fi +
dV
=
xj
V
= fi +
(4.13)
ij
xj
which is Naviers equation (also known as Cauchys law of motion from

1827). For many practical purposes in seismology it is appropriate to ignore
body forces so that the equation of motion is simplied to :
2 ui
ij
=
2
t
xj
or ui = ij,j
(4.14)
2 Gauss divergence theorem states that in the absence of creation or destruction of matter,
the density within a region of space V can change only by having it ow into or away from
the region through its boundary S :
t. dS =
S
.t dV
V
144
Intermezzo 4.2 Diagonalization of a matrix

Many problems in (geo)physics can be simplied if we can diagonalize a matrix.
Under certain conditions (almost always satised in geophysics), for any square
matrix A of dimension n, there exists a n n matrix X that diagonalize A :
X1 AX
= diag(1 , ..., n )
1
0
0
0
0
0
2
...
...
...
...
0
...
0
...
...
...
...
n1
0
0
0
0
0
n
(4.15)
This means that there exists a coordinate system in which A is diagonal. Diagonalizing A corresponds to nding this coordinate system and the values of
the diagonal elements of A in this coordinate system. We can rewrite the last
equation as follows :
AX = X
(4.16)
(A I)X = 0
(4.17)
or
I is the Identity matrix. The i (i = 1, ..., n) are called the eigenvalues of A

and the columns of X are formed by n eigenvectors. Diagonalizing a matrix is
equivalent to nding its eigenvalues and eigenvectors. This is called an eigenvalue problem. Finding the eigenvalues can easily be done by solving the system
of n linear equations and n unknowns (the i ) formed by Eq. 4.17. This has a
non-trivial solution if the determinant is zero (this is called Cramers rule) :
|A I| = 0
(4.18)
The eigenvectors can then be found by replacing the eigenvalues in the system of
linear equations formed by Eq. 4.17. If all eigenvalues are dierent, the n eigenvectors are linearly independent and orthogonal. Otherwise, the eigenvalues are
said to be degenerate and the number of independent eigenvectors is given by the
number of independent eigenvalues. In the case of n independent eigenvalues,
the eigenvectors can form a new orthogonal basis and they are called principal
axes. If we change the coordinate system and use the system dened by the
principal axes, matrix A becomes diagonal and its elements are given by the
eigenvalues.
In the case of the stress tensor, equation 4.17 takes the form :
( I)n = 0
(4.19)
The three eigenvalues (also called principal stresses and represented by the
scalar ) are thus found by solving :

11

| I| =
21
31
12
22
32
13
23
33

=0

(4.20)
This will give three values for (1 , 2 and 3 ). In the coordinate system
formed by the three principal axes ni , the stress tensor is diagonal, as expressed
in Eq. 4.10.
4.5. EQUATIONS OF MOTION, WAVE EQUATION, P AND S-WAVES 145

Note that body forces such as gravity cannot always be ignored in what is
known as low-frequency seismology. For instance, gravity is an important
restoring force for some of Earths free oscillations. We can also introduce a
body force term to describe the seismic source.
Weve derived Eq. 4.14 using index notation. Lets state it in vector form.
The acceleration is proportional to the divergence of the stress tensor (see Intermezzo 4.3) :
u=
(4.21)
Equation (4.14) represents, in fact, three equations (for i=1,2,3) but there are
more than three unknowns (the 6 independent elements of the stress tensor ij
plus density . In this general form the equation of motion does not have a
unique solution. Also, we have introduced forces and tractions but we not yet
specied how the material reacts to the applied (non-)contact forces. We need
some physics to help us out. Specically, we need to know the relationship
between stress and strain, i.e. a constitutive relationship.
Intermezzo 4.3 Divergence of a tensor

We know how to dene the divergence of a vector. The divergence of a tensor
is simply the generalization to higher dimensions of the divergence of a vector
(remember that a vector is nothing more than a tensor of dimension 1).
The divergence of a vector v is a scalar denoted by .v and given by :
.v =
vi
i
xi
v1
v2
v3
+
+
x1
x2
x3
(4.22)
Similarly, the divergence of a dimension 2 tensor is a vector whose components

are given by :
(.)j =
ij
i
xi
1j
2j
3j
+
+
x1
x2
x3
(4.23)
And we can further generalize : the divergence of a n-dimension tensor is a

tensor of dimension n-1 obtained in a way similar to Eq. 4.23.
In one-dimension this relationship is given, as mentioned before, by = E

(or i = Ei , where E is the Youngs modulus, which is the ratio of uniaxial
stress to strain in the same direction, i.e. a measure of the resistance against
extension. A simple example demonstrates that in more dimensions this scalar
proportionality breaks down. Imagine an elastic band : if one stretches this
band in one direction, say the x1 direction, than the band will extend in that
direction. In other words there will be strain e11 due to stress 11 . However, the
strap will also thin in the x2 and x3 directions; so e22 = e33 = 0 even though
22 = 33 = 0.
146
Clearly, a simple scalar relationship between the stress and strain tensors
is invalid : ij = Eij . Somehow we must express the elements of the stress
tensor as a linear combination of the elements of the strain tensor. This linear
combination is given by a 4th order tensor cijkl of elastic constants :
ij = Cijkl kl
(4.24)
This form of the constitutive law for linear elasticity is known as the generalized Hookes law and C is also known as the stiness tensor. Substitution of
eq (4.24) in (4.14) gives the wave equation for the transmission of a displacement
disturbance with wave speed dependent on density and the elastic constants
in Cijkl in a general elastic, homogeneous medium (in absence of body forces) :

uk
uk
2 ui
Cijkl
= Cijkl
ui = 2 =
= Cijkl uk,lj
(4.25)
t
xj
xl
xj xl
In three dimensions, a fourth order tensor contains 34 = 81 elements. What
did we gain by doing all this? After all, we mentioned above that we needed to
introduce a constitutive relationship in order to solve the wave equation (Eq.
4.14) since the number of equations was less than the number of unknowns.
Now we have arrived at a situation (Eq. 4.25) in which we have 3 equations
to solve for 82 unknowns (density + 81 elastic moduli), so the introduction of
physics does not seem to have helped us at all! The situation improves once
we consider the intrinsic symmetry of the tensors involved. The symmetry
of the stress and strain tensors leads to symmetry of the elasticity tensor :
Cijkl = Cijlk = Cjilk . This reduces the number of independent elements in
Cijkl to 66=36. It can also be demonstrated (with less trivial arguments)
that Cijkl = Cklij , which further reduces the number of independent elements
in Cijkl to 21. This represents the most general (homogeneous) anisotropic
medium (anisotropy in this context means that the relationship between stress
and strain is dependent on the direction i).
By restricting the complexity of the medium we can further reduce the number of independent elements of the elasticity tensor. For instance, one can investigate special cases of anisotropy by allowing directional dependence in a plane
perpendicular to certain symmetry axes only. We will come back to this later.
The simplest case is a homogeneous, isotropic medium (i.e. no directional
dependence of elastic properties), and it can be shown (see, e.g., Malvern (1969))
that in this situation the general form of the 4th order (linear) elasticity tensor
is
Cijkl = ij kl + (ik jl + il jk )
(4.26)
where and are the only two independent elements; and are known as
Lames (elastic) constants (or moduli), after the French mathematician G. Lame.
(The Kronecker (delta) function ij = 1 for i = j and ij = 0 for i = j).
Substitution of Eq. (4.26) in (4.24) gives for the stress tensor
ij = Cijkl kl = ij kk + 2ij = ij + 2ij
(4.27)
4.6. P AND S-WAVES
147
with the cubic dilation, or volume change. This form of Hookes law was
rst derived by Navier (1820-ies). The Lame constant is known as the shear
modulus or rigidity : it is a measure of the resistance against shear or torsion
of the medium. The shear modulus is large for very sti material, but is small
for media with low viscosity ( = 0 for water or for liquid metallic iron in the
outer core). The other Lame constant, , does not have much (general) physical
meaning by itself, but denes important elastic parameters in combination with
the shear modulus . Of most interest for us right now is the denition of ,
the bulk modulus or incompressibility : = + 2/3. The bulk modulus is
a measure of the resistance against volume change : = P/, with P the
pressure and the cubic dilatation, and is large when the change in volume is
small even for large (hydrostatic) pressure. The minus sign is necessary to keep
> 0 since < 0 when P > 0. For isotropic media other important elastic
parameters, such as the Poissons ratio, i.e. the ratio of lateral contraction to
longitudinal extension, and Youngs modulus can also be expressed as linear
combinations of and (or and ). We can readily see that the stress tensor
consists of terms representing (resistance to) either changes in volume or shear
(or torsion).
stress : eects of volume change + torsion (or shear) of material
This is a fundamental result and it underlies, what we will see below, the formulation of wave propagation in terms of compressional (dilatational) P and
transversal (shear) S-waves.
With the above constitutive relationships we can now derive the equation
that describes wave propagation in a homogeneous, isotropic medium
uk
2 ui
= ( + )
+ 2 ui
t2
xi xk
(4.28)
which represents a system of three equations (for i=1,2,3) with three unknowns
(, , ). Note that for practical purposes in seismology these parameters are not
really constant; in Earth they are functions of position r and vary signicantly,
in particular with depth.
4.6
P and S-waves
There are several ways to demonstrate that solutions of the equation of motion
essentially consist of a dilatational and a rotational term, the P and S-waves,
respectively. Using vector notation the equation of motion is written as
u = ( + )( u) + 2 u
(4.29)
or, by making use of the vector identity

2 u = ( u) ( u),
(4.30)
148
we can write the equation of motion as :
u = ( + 2)( u) ( u)
dilatational
rotational
(4.31)
which is a system of three partial dierential equations for a general displacement eld u through an unbounded, homogeneous, and isotropic medium.
In general, it is dicult to solve this system directly for the displacement u.
Typically, one tries to decompose the general wave equation into separate equations that relate to P- and S-wave propagation. One approach is to eliminate directly any rotational contributions to the displacement by taking the divergence
of Eq. (4.31) and using the property that for a vector eld a, ( a) = 0.
Similarly we can eliminate the dilatational contributions by taking the rotation
of (4.31) and using the identity that, for a scalar eld , = 0. Assuming
no body force f , we get :
Taking the divergence leads to
2 ( u)
= ( + 2)2 ( u)
t2
(4.32)
or, with u = ,
2
= 2 2
t2
(4.33)
which is a scalar wave equation that describes the propagation of a volume

change through the medium with wave speed

=
+ 2
=
+ 4/3
(4.34)
In general = (r), = (r), = (r) = (r)

Taking the rotation leads to
2 ( u)
= ( + 2) ( u) ( u)
t2
(4.35)
which, with ( u) = 0 and the vector identity as used above (and

again using ( a) = 0), leads to :
2 ( u)
= 2 2 ( u)
2t
(4.36)
4.6. P AND S-WAVES
149
This is a vector wave equation that describes the transmission through a

medium of a rotational disturbance u with wave speed

=
(4.37)
In general = (r), = (r) = (r)

The dilatational and rotational components of the displacement eld are
known as the P and S-waves, and and are the P and S-wave speed, respectively.
Another (more elegant) way to see that solutions of the wave equation are
in fact P and S-waves is by realizing that any vector eld can be represented by
a combination of the gradient of some scalar potential and the curl of a vector
potential. This decomposition is known as Helmholtzs Theorem and the
potentials are often referred to as Helmholtz Potentials. Using Helmholtzs
Theorem we can write for the displacement u
u = +
(4.38)
. = 0
(4.39)
and
with a rotation-free scalar potential (i.e. = 0) and the divergencefree vector potential. Substitution of (4.38) into the general wave equation (4.31)
(and applying the vector identity (4.30)) we get :
+ [2 ]
=0
[( + 2)2 ]
(4.40)
which is a third-order dierential equation3 . Equation (4.40) can be satised

by requiring that both
=0
( + 2)2
(4.41)
which is a scalar wave equation for the propagation of the rotation-free displacement eld with wave speed

+ 2
+ 4/3
=
=
(4.42)
and
=0
2
(4.43)
3 Strictly speaking this is not the way to formulate the problem. The need to solve thirdorder dierential equations could have been avoided if the problem was set up in a dierent
way by making use of what is known as Lam
es theorem. This also involves Helmholtz
potentials. See, for instance, Aki & Richards, Quantitative Seismology (1982) p. 67-69.
This mathematical correctness is, however, not required for a basic understanding of the
decomposition in P and S terms
150
which is a vector wave equation for the propagation of the divergence-free displacement eld with wave speed

=
(4.44)
Comparing Eq. 4.33 and 4.41, we can identify with the volume change ( u
is called the cubic dilatation). Similarly, can be identied with the rotational
component of the displacement eld by comparing Eq. 4.36 and 4.42.
It is often much easier to solve the wave equations (4.41) and (4.43) than to
solve the equation of motion directly for u, and from the solution for the potentials the displacement u can then determined directly by Eq. (4.38). Note that
even though P and S-waves are often treated separately, the total displacement
eld comprises both wave types.
Lets now consider a Cartesian coordinate system with z oriented downward,
x parallel with the plane of the paper, and y out of the paper. Well make the
x-z plane the special plane of the problem. Because /y
y = 0, we can write :
=
x+
z
x
z
(4.45)
and

/x /y

y
= x
x
/z
z
(4.46)
Therefore,
ux
uy
uz
x
z
x
z
x
z
y
+
z
x
(4.47)
The displacement direction from is in the x-z plane and it is compressional

is the P -wave potential. P wave propagation is thus rotation-free and has
no components perpendicular to the direction of wave propagation, k : it is
a longitudinal wave with particle motion in the direction of k. In contrast,
the particle motion associated with the purely rotational S-wave is in a plane
perpendicular to k : transverse particle motion can be decomposed into vertical
polarization, the so-called SV wave, and horizontal polarization, the so-called
SH-wave (see Fig. 4.2) The displacement uy from the SV -wave potential is in
the same plane. In this formulation, uy could just as well have been called the
SH -wave potential with displacement direction perpendicular to the x-z-plane.
4.7. FROM VECTOR TO SCALAR POTENTIALS POLARIZATION 151
Figure 4.2: P and S waves : partical motion and propagation

direction.
4.7
From vector to scalar potentials Polarization
Using the Fourier transform, we show that the vector decomposition with
and can be reduced into three equations with the scalar potentials , SV
and SH (waves are typically described by oscillatory functions, i.e. complex
exponentials. It is therefore natural to move the analysis to the frequency
domain, i.e. to use Fourier transforms). We will write u(r, t) for the time
and space domain displacement, and u(r, ) for the displacement in space and
frequency. The transformation to the frequency domain is done by means of the
(temporal) Fourier transform, which is dened as :

u(r, )
u(r, t)eit dt
(4.48)
+
u(r, )eit d
(4.49)
u(r, t)
1
2
It is easy to see how the time derivative in a partial dierential equation (PDE)
brings out a factor of i. This can be veried using the PDE obtained in section
4.6 :
2u
= ( + 2)( u) ( u)
t2
(4.50)
The separation of the equation of motion into two parts was done in section
4.6. It can also be done in the frequency domain : using Eq. 4.49 and Eq. 4.38,
Eq. 4.50 (the equation of motion) becomes :
2 =
2
2 u
2
We thus easily get :
(4.51)
(4.52)
152
(A)
(B)
Figure by MIT OCW.
Figure 4.3: Successive stages in the deformation of a block of material by Pwaves and SV-waves. The sequences progress in time from top to bottom and
the seismic wave is travelling through the block from left to right. Arrow marks
the crest of the wave at each satge. (a) For P-waves, both the volume and the
shape of the marked region change as the wave passes. (b) For S-waves, the
volume remains unchanged and the region undergoes rotation only.
2 2 =
2 2 SV =
2
SH
2
2 SV
(4.53)
SH
We now have ordinary dierential equations (ODEs), also known as Helmholtz

equations, which are much easier to solve than PDEs. (NB one can readily
see that would lead to ik and 2 to k 2 ; therefore k2 = f rac 2 2 ,
k2 = f rac 2 2 .)
4.8. SOLUTION BY SEPARATION OF VARIABLES
4.8
153
Solution by separation of variables
In a way, weve solved the wave equation by realizing that we could reduce it to
an ordinary dierential equation using the Fourier transform. So we knew the
solution would be a complex exponential in the time variable (it is a natural
way of describing a wave). We will now justify the validity of this approach by
attempting to solve the following partial dierential equation :
c2 2 =
t2
(4.54)
without resorting to the Fourier transform.

If we propose a solution by separation of variables :
= X(x)Y (y)Z(z)T (t)
(4.55)
and plug Eq. 4.55 into Eq. 4.54, we obtain :

1 d2 X
1 d2 Y
1 d2 Z
1 d2 T
+
+
=0
X dx2
Y dy 2
Z dz 2
c2 T dt2
(4.56)
The partial derivatives are regular derivatives now : we went from a PDE to
ordinary dierential equations (ODE). Each of these terms needs to be constant.
We can pick these constants ( 2 for T , and kx2 , ky2 and kz2 for the spatial functions) but they will not be independent (they are linked to one another through
Eq. 4.56). If we pick , kx and ky , then kz is not independent anymore and
satises :
kz2 =
2
kx2 ky2
c2
(4.57)
With those constants, it is easy to show that X, Y , Z and T are oscillatory

functions :
X
Y
exp(ikx x)
exp(iky y)
Z
T
exp(ikz z)
exp(it)
(4.58)
We have obtained solutions to the wave equation. Of course any linear

combination of particular solutions leads to the general solution, and also we
need to pick the sign in Eq. 4.58 (from the boundary conditions). Relation
4.57 is called a dispersion relation. kx , ky and kz can be seen as the cartesian
component of a vector k and can be written as an oscillatory function of the
type
exp(i(k r t))
(4.59)
These waves are called plane waves and k is the direction of wave propagation.
154
4.9
Plane waves
Weve called functions of the type exp(i(k r t)) plane waves. Lets look at
a few characteristics.
Traveling waves Lets rst notice that plane waves are of the general form
describing traveling waves :
(x, t) = f (x ct) + g(x + ct)
(4.60)
with f and g arbitrary functions, provided that they are twice dierentiable
with regard to space and time (and that the second derivatives are continuous).
= c2 2 . This is referred to as dAlemberts
After all, they need to solve
solution. The function f (x ct) represents a disturbance propagating in the
positive x direction with speed c. The function g(x+ct) represents a disturbance
propagating in the negative x axis : this part of the solution will be ignored
in the following, but it must be taken into account when dealing with wave
interference.
Wavelength
With k = 2/, the spatial part can be manipulated as follows:
2
ei
= ei x ei2N = ei
(x+N )
(4.61)
to show that is indeed the wave length after this distance, the displacement
pattern repeats itself.
t = t0
t = t'
X1 = X0
g(X1 - Ct0)
g
g
X1 = X 0
X1
g(X0 - Ct)
g(X1 - Ct')
X1
X0 X'
t0
t'
Figure by MIT OCW.
Figure 4.4: Plane waves : propagating disturbances.
Phase
With increasing time t the argument of function f does not change provided
that x also increases (hence the propagation in the positive x axis). In other
words, if the argument remains constant it means that the shape dened by
function f translates through space. The argument of f , x ct, is referred to as
the phase; one can dene the wave front as the propagating function for a given
4.9. PLANE WAVES
155
value of the phase. That c is the phase velocity is easily obtained by considering a
constant phase at times t and t (xct = x ct c = (x x)/(t t) dx/dt
speed).
Wavefront
A wavefront is a surface through all points of equal phase, i.e. a surface
connecting all points at the same travel time T from the source (see Fig. 4.5).
In other words, at the wave front, all particles move in phase. Rays are the
normals to the wave fronts and they point in the direction of wave propagation.
The use of rays, ray paths, and wave fronts in seismology has many similarities
with optics, and is called geometric ray theory.
Figure 4.5: Seismic wavefronts.
Plane waves have plane wave fronts. The function remains unchanged for
all points on the plane perpendicular to the wave vector : indeed, on such a
plane, the dot product k r is constant.
At distances suciently far from the source body waves can be model-led
as plane waves. As a rule of thumb : observer must be more than 5 wave
lengths away from source to apply far eld or plane wave approximation.
Closer to the source one would need to consider spherical waves. Note that a
seismogram corresponds to the recording of u = u(r0 , t) at a xed position r0 ;
i.e. the displacement as a function of time that records the passage of a wave
group past r0 .
Polarization direction
The polarization direction is dierent from the propagation direction. As
already mentionned in sections 4.8 and 4.7, all waves propagate in the direction
of their wave vector k. The P -wave displacement () is parallel with the k.
The SV -displacement ( (
y)) is perpendicular to this, in the x z plane,
and the SH-displacement is out of the plane.
156
To indicate explicitly the propagation in the direction of or perpendicular to

wave vector k, one sometimes also writes
for P-waves:
for S-waves:
(r, t) = An k ei(krt)
(r, t) = Bn k ei(krt)
(4.62)
Low- and high-frequency seismology

The variables used to describe the harmonic components are related as follows;
Angular frequency
Wavelength
Wavenumber
Frequency
Period
= kc
= cT = 2/k
k = /c
f = /2 = c/
T = 1/f = /c = 2/
Seismic waves have frequencies f ranging roughly from about 0.3 mHz to
100 Hz. The longest period considered in seismology is that associated with
fundamental free oscillations of the earth : T 59 min. For a typical wave
speed of 5 km/s this involves signal wavelengths between 15,000 km and 50m.
A loose subdivision in seismological problems is based on frequency, although
the boundaries between these elds are vague (and have no physical meaning) :
low frequency seismology
high frequency seismology
exploration seismics:
4.10
f <20 mHz
50 mHz < f <10 Hz
f > 10 Hz
> 250 km
0.5 km < < 100 km
< 500 m
Some remarks
1. The existence of P and S-waves was rst demonstrated by Poisson (in

1828). He also showed that P and S-type waves are, in fact, the only solutions of the wave equations for an unbounded medium (a whole space), so
that u = + provides the complete solution for the displacement
in an elastic, isotropic and homogeneous medium. Later we will see that
if the medium is not unbounded, for instance a half-space with perhaps
some stratication, there are more solutions to the general equation of
motions. Those solutions are the surface (Rayleigh and Love) waves.
2. Since > 0 and 0 > : P-waves propagate faster than shear
waves! See Fig. 4.6.
3. It can be shown that independent propagation of the P and S-waves is only
guaranteed for suciently high frequencies (the so-called high-frequency
approximation, high frequency in the sense that spatial variations in
4.11. NOMENCLATURE OF BODY WAVES IN EARTHS INTERIOR 157

elastic properties occur over much larger distances than the wavelength
of the waves involved) underlies most (but not all) of the theory for body
wave propagation).
4. The three components of the wave eld (P, SV, and SH-waves, see section
4.7 for more details) can be recorded completely with three orthogonal
sensors. In seismometry one uses a vertical component [Z] sensor along
with two horizontal component sensors. In the eld the latter two are
oriented along the North-South [N] and East-West [E] directions, respectively. Fig. 4.7 is an example of such a three-component recording; we
will come back to this in more detail later in the course.
Vp
Vs
12
Wave speed (kms1)
10
1000
2000
3000
Depth (km)
4000
5000
6000
Figure 4.6: P and S wave speed in the ak135 Earth model.
4.11
Nomenclature of body waves in Earths interior
At this stage it is useful to introduce the jargon used to describe the dierent
types of body wave propagation in Earths interior. We will get back to several
wave propagation issues in more detail after we have discussed the basics of
ray theory and the construction and use of travel time curves. There are a few
simple basic rules, but there are also some inconsistencies :
Capital letters are used to denote body wave propagation (transmission)
through a medium. For example, P and S for the compressional and shear
waves, respectively, K and I for outer and inner core propagation of compressional waves (K for German Kerne; I for Inner core), and J for shear
wave propagation in the Inner Core (no denitive observations of this seismic phase, although recent research has produced compelling evidence for
its existence).
158
Figure 4.7: Example of a three-component seismic record
Lower case letters are either used to indicate either reections (e.g., c
for the reection at the CMB, i for the reection at the ICB, and d for
reections at discontinuities in the mantle, with d standing for a particular
depth (e.g., 410 or 660 km), or upward propagation of body waves
before they are reected at Earths surface (e.g., s for an upward traveling
shear wave, p for an upward traveling P wave). Note that this is always
used in combination of a transmitted wave : for example, the phase pP
indicates a wave that travels upward from a deep earthquake, reects at
the Earths surface, and then travels to a distant station.
Figure 4.8: Nomenclature of body waves
4.12. MORE ON THE DISPERSION RELATION
4.12
159
More on the dispersion relation
We have already introduced the concept of dispersion (Eq. 4.57). Searching for
a solution by separation of varibles, we have seen that the solution to the wave
equation is an exponential both in the time and space domain. We had, however,already shown the oscillatory behavior of the solution in the time domain
by using the time Fourier transform. In this section, we go one step further.
Predicting that the solution will be a complex exponential in the spatial domain
as well, we will investigate what insight the spatial Fourier-transform will bring
us. Time and space are linked through the wave equation (it is a PDE) the
linkage between them is by the dispersion relation which we are deriving here.
As denition for the spatial Fourier transform and its inverse, we take

(k, ) =
(r, )eikr d3 r
(4.63)
V
and
1
(r, ) =
(2)3
(k, )eikr d3 k
(4.64)
The integrations are over all of physical space V (dxdydz) and all of wave
vector space K (dkx dky dkz ), respectively. The dot product kr = kq xq with the
Einstein summation convention. Remember also that kp2 = kp kp = |k| = k 2 .We
need the Laplacian of , this is given by :

1
2
2
=
=
(k, )eikq xq i2 kp2 d3 k
(4.65)
xp xp
(2)3 K
Comparison with Eq. 4.54 leads to (call or now c) :
k 2 +

2

=
0
or
|k|
=

c2
(4.66)
We can quickly convert this dispersion relation into something youre all familiar
with : with k = 2/ and f = /(2), we get f = c : the frequency of a wave
times its wave lengths gives the propagation speed. We will discuss this in more
detail below.
The complete solution to the wave equation is thus given by inverse transformation of (r, ) as follows :
1
(r, t) =
(2)4
+
+
+

(kx , ky , , z)ei(krt) dkx dky d
(4.67)
There are three independent quantities involved here (not four) : kx , ky and
, and their relationship is given by the dispersion equation. In other words,
2
1/2
2
2
k r = kx x + ky y + z
k
(4.68)
x
y
c2
160
Its important to see Eq. 4.67 as what it is : a superposition (integral) of plane

waves with a certain wave vector and frequency, each with its own amplitude.
The amplitude is a coecient which will have to be determined from the initial
or boundary conditions.
We thus have seen that the dispersion equation can be obtained either by
solving the wave equation by separation of variables or by introducing the time
and spatial Fourier transforms.
4.13
The wave eld Snells law
In this section, well use plane wave displacement potentials to solve a simple problem of wave propagation. Not only will we understand why and how
reections, refractions and phase conversions happen, but well also derive an
important relation for plane waves in planar media known as Snells law.
Lets start with a plane P -wave incident on the free surface, making an angle
with the normal i. We can identify the P -wave with its wave vector. In our
case, we know that

kx = sin i and kz = cos i
(4.69)
Two kinds of boundary conditions are used in seismology there are the
kinematic ones, which put constraints on the displacement, and the dynamic
ones, which constrain the stresses or tractions. The free surface needs to be
traction-free. We remember that the traction vector was given by dotting the
stress tensor into the normal vector representing the plane on which we are
computing the tractions : ti = ij nj . For a normal vector in the positive
z-direction, the traction becomes :
t(u,
z) = (xz , yz , zz )
(4.70)
For isotropic materials, we have seen the following denition for the stress tensor :

ui
uj
ij = ( u)ij +
+
(4.71)
xj
xi
Tractions due to the P wave
We know that the displacement is given by the gradient of the P -wave displacement potential (see Eq. 4.47) :

u = =
, 0,
x
z
Therefore the required components of the stress tensor are :
(4.72)
4.13. THE WAVE FIELD SNELLS LAW
2
xz
161
(4.73)
xz
yz
(4.74)
2
2 + 2 2
z
(4.75)
zz
Tractions due to the SV wave

The displacement is given as the rotation of the potential (see Eq. 4.47) :

u=
, 0,
z
x
(4.76)
For the stress tensor, we nd :

2 2
x2
z 2

(4.77)
xz
xz
(4.78)
2
2
xz
(4.79)
zz
Tractions due to the SH wave

The SH wave, as weve seen, has only one component in this coordinate
system :
u = (0, uy , 0)
(4.80)
and the stress tensor components are given by
xz
yz
zz
0
uy
z
0
(4.81)
(4.82)
(4.83)
Comparing Eqs. 4.75 and 4.79, we see how P and SV waves are naturally
coupled. In this plane-wave plane-layered case, the P -wave had energy only
in the x- and z-component, and so did SV . Upon reection and refraction,
energy can be transferred from the incoming P -wave to a reected P -wave and
a reected SV -wave. No SH waves can enter the system they have all their
energy on the y-component.
Analogously to Eq. 4.69, we can represent the incoming P , the reected P
and the reected SV wave by the following slownesses :
162
P re
SV re

cos i
sin i
, 0,

sin i
cos i
, 0,

cos j
sin j
, 0,
inc
(4.84)
(4.85)
(4.86)
Thus the total P -potential is made up from the incoming and reecting P wave, and the shear-wave potential is given by the reected SV -wave. All of
them, of course, have the plane wave form, so that we can write :
inc
re
re

cos i
sin i
x
zt
A exp i

sin i
cos i
B exp i
x+
zt

sin j
cos j
x+
zt
C exp i
(4.87)
(4.88)
(4.89)
As pointed out before, there are no kinematic boundary conditions on the

free surface. The displacement of the free surface is unconstrained, and above
it there is no displacement at all. The dynamic boundary conditions, however,
are non-trivial. The tractions must vanish on the free surface : so xz = yz =
zz = 0 at z = 0. It is easy to see that, with z = 0, the sum of the three plane
wave displacement potentials will be of the type

sin i
A exp i
xt
+

sin i
B exp i
xt

sin j
xt
C exp i
Hence, for this sum to be zero for all x and t, we need :

sin i
sin j
sin i
=
=
p
(4.90)
Thus, for plane waves in plane-layered media, the whole system of rays is
characterized by a common horizontal slowness. This is true for the whole wave
eld of reected and transmitted (refracted) waves. Eq. 4.90 is known as Snells
law and p is called the ray parameter. In the following paragraph, a more
general principle called Fermats principle is used to prove Snells Law.
4.14. FERMATS PRINCIPLE AND SNELLS LAW
4.14
163
Fermats Principle and Snells law
An important principle in optics is Fermats principle, which governs the geometry of ray paths. This principle states that a wave propagating from position
A to position B follows a path of stationary time. The principle of stationary
time plays a fundamental role in high frequency seismology. Note that stationary time does not necessarily mean minimum time; it can also be a maximum
time.
Figure 4.9: The principle of stationary time.

Consider Fig. 4.9. A ray leaves point P that is in a medium with wave
speed c1 and travels to point Q in a medium with wave speed c2 . What path
will the ray take to Q? Since the wave speeds in the media are constant the ray
path in each medium is a straight line, so that in this simple case the geometry
is completely dened by the positions of P , Q, and the point x where the ray
crosses the interface.
The travel time on an arbitrary path between P and Q is given by

b2 + (c x)2
a2 + x2
e
d
+
=
+
(4.91)
tP Q =
c1
c2
c1
c2
For the path to be a stationary time path (i.e. time is maximum or minimum)
we simply set the spatial derivative of the travel time to zero :
dT
cx
x

=0=
dx
c1 a2 + x2
c2 b2 + (c x)2
(4.92)
and note that

x
= sin i1
2
a + x2
and
cx

= sin i2
2
b + (c x)2
(4.93)
This gives Snells law :

sin i1
sin i2
=
p
c2
c1
p is called the ray parameter.
(4.94)
164
One can expand on this simple geometry and consider many more layers,
but the result is the same : the ray parameter p is constant along the entire ray!
As a ray enters material of increasing velocity, the ray is deected toward the
horizontal; if the ray enters material with lower velocity, the ray is deected to
the vertical. In seismology the angle between the ray and the vertical is referred
to as the angle of incidence (also, take-o angle).
4.15
Ray geometries of the wave eld
For most applications we have to deal with a complex wave eld : in each layer
of a stratied medium there can be 6 dierent body wave groups : the up- and
down-going P, SV, and SH-waves. The propagation of such a wave eld through
a stratied medium (a stack of horizontal layers or spherical shells in which the
wave speed is constant) is controlled by Snells law (Fermats Principle) and
boundary conditions.
The wave eld is determined by reections, refractions, and phase conversions; for instance, a down-going P wave can reect at an interface and part
of its energy can be transmitted to the other side, and part of its energy can
(or often has to be) converted to SV-wave energy (see Fig. 4.10).
Figure 4.10: Ray conversions at interfaces.

The incidence angles of the reected and refracted waves that compose this
complex wave eld are controlled by an extended form of Snells law. For this
example, Snells law is :
sin j1
sin i2
sin j2
sin i1
=
=
=
p
1
1
2
2
(4.95)
This generalization of Snells law shows an important concept that the whole
system of seismic waves produced by reection and transmission of plane waves
in a stratied medium is characterized by the value of their common horizontal
slowness, or the ray parameter p. It can also be used directly to determine the
angles for critical reection and refraction. The ray parameter is constant not
only for a single ray, but for the entire wave eld generated by reection and
refraction of an incoming P or S-wave.
4.16. TRAVEL TIME CURVES AND RADIAL EARTH STRUCTURE
4.16
165
Travel time curves and radial Earth structure
We have been developing some basic theory and concepts of body wave seismology. One of the major objectives of seismology is to extract structural
information about Earths structure from the observed data, the seismograms.
We will discuss some rather classical techniques to do this.
Snells Law
We derived Snells law for a at Earth :
sin i2
sin in
sin i1
=
= ... =
= constant p, the ray parameter
c1
c2
cn
(4.96)
The ray parameter is constant along the entire ray path, and is the same for
all rays (reections, refractions, conversions) associated with the same incoming
ray. The ray parameter plays a very important role in seismology.
Snells Law shows that the ray parameter is inversely proportional to velocity,
or proportional to 1/velocity, which is the slowness. In seismology it is often
more convenient to use slowness instead of wave speed. One signicant advantage
of the slowness vector is that it can be added vectorially, whereas this is not
always justied (in our context) for the velocity.
s = (s1 , s2 , s3 ) = s1 x1 + s2 x2 + s3 x3
(4.97)
The vector summation for velocity can give practical problems : consider,
for instance, the plane wave that propagates in the direction k. The apparent
velocity c1 measured at the surface (from observations at several stations) is
larger than the true velocity c : with i the angle of incidence, c1 = c/ sin i > c,
so that c = c1 + c3 .
From Fig. 4.11 we can easily derive two other important relationships :
sin i =
dt
ds
sin i
dt
c
1
=
=c
=
p=
=
dx1
dx1
c1
c
dx1
c1
Figure 4.11: Derivation of Snells law.
(4.98)
166
1. The ray parameter p is 1/cx , which is referred to as the horizontal slowness!

2. the ray parameter is simply the derivative of the travel time T with horizontal distance. This will prove to be of major importance (and convenience!).
For a spherical earth we can derive a relationship for the ray parameter that
is similar to Eq. (4.98), the only dierence being the scale factor r :
p=r
sin i
v(r)
(4.99)
where r is the radius to any point along the ray path, and v(r) the wave speed
at that radius. It can also be shown that (with the angular distance)
p=
(4.100)
Figure 4.12: Ray parameter in spherical geometry

Notice the similarity between the denition of the ray parameter as the
spatial derivative of travel time for the at (Eq. 4.98) and spherical earth
(Eq. 4.100)! Beware : For a at earth the unit of ray parameter is s/km (or
s/m), for the spherical earth it is either s/rad or just s or s/deg, so even though
the denitions are completely equivalent there are dierences in units!
With the denition for the ray parameter in a spherical Earth (Eq. 4.99)
we can also get a simple expression that relates p to the minimum radius (or
maximum depth) along the ray path : this point is known as the turning or
bottoming point of the ray. A turning ray is the spherical Earth equivalent
of the head wave (see Fig. 4.12).
rmin
rmin sin 90
=
=p
v(rmin )
v(rmin )
(4.101)
Under the assumption of a reference earth model for seismic wave speeds we
can determine the horizontal distance traveled by the ray (e.g., from 4.98) and
the depth to the turning point (from Eq. 4.101) once we know the ray parameter.
Before showing how the ray parameter can be determined from observed data,
let me mention another important concept based on the ray parameter :
4.16. TRAVEL TIME CURVES AND RADIAL EARTH STRUCTURE
167
Travel time curves

Eq. (4.98) indicates that the ray parameter, i.e. the horizontal slowness, can be
determined from seismic data by determining the dierence in travel time of a
phase arrival at two adjacent stations. Ideally one uses an array of instruments
to do this accurately.
Figure 4.13: Determination of the ray with a seismometer array

In other words, one can determine the value of the ray parameter directly from
the travel time curve, which represents the variation of travel time as a function
of distance : T (X) or T (). A travel time curve can be constructed by arranging observed records of ground motion due to the same explosion or earthquake
as a function of distance. In such a record section the travel time curve of a
particular phase is just the curve that connect onset times of that phase in all
records. One could also construct a travel time curve by using many measurements, phase picks, of the travel time of particular phases, say the P-phase, at
dierent distances from the source. Seismologists try to nd simple models of
radial variations of wave speed that produce travel time curves consistent with
the observed data. Theoretical travel time curves in this sense are thus best
tting curves determined from some reference model of seismic wave speeds.
Well known models for the Earths depth dependent structure are the Preliminary Reference Earth Model (PREM) by Dziewonski & Anderson
(1981), and the more recent iasp91 model (Kennett & Engdahl, 1991). Typically, this tting is not done by trial and error but by means of inversion of
either the travel times or the travel time curves. A classical approach that is
discussed in most text books is the one rst applied by Herglotz and Wiechert
in the beginning of this century. They were the rst to invert travel time data
for simple radially stratied models of seismic wave speed, and their technique
has been used for decades. The rst comprehensive model and the corresponding travel time tables was published by Jereys & Bullen (1939/1940). In fact
their model, known as the JB model, is still being used for routine earthquake
location by the International Seismological Centre in the U.K.
The ray parameter of a seismic wave (group) arriving at a certain distance
can be thus be determined from the slope of the travel time curve. The straight
line tangent to the travel time curve at can be written as a function of the
168
intercept time and the slope p :
p=
T
T
T () = +
= + p
(4.102)
and this equation forms the basis of what is known as the p method.
Figure 4.14: Determination of the ray parameter from the traveltime curve
The (local) slope of the travel time curve contains important information about
the horizontal slowness, and thus about the wave speed, and the intercept time
, the zero oset time, contains information about the layer thickness. This
property is exploited in exploration seismics, where we typically deal with travel
time curves that consist of segments of straight lines (see Fig. 4.14).
Another piece of information that can be obtained from travel time curves
is contained in the second derivative of the travel time curve with distance, or
the variation of ray parameter with distance p/. This quantity controls the
amplitude of the arrivals. To see this, consider a situation (that we will discuss
in more detail below) in which rays with dierent incident angles at the source
(and receiver) are somehow focused to travel to the same seismographic station
so that the amplitude increases. In that case, p = 0 but = 0 so
p
2T
=
(4.103)
In other words, the larger p/, the more energy arrives at a small distance
range , and the higher the amplitude. In real life the amplitude of seismic
waves is always nite, and this reveals, in fact, one of the shortcomings of
ray theory. If rays are assumed to be innitesimally narrow the theoretical
amplitude can go to innity, but in practice the amplitude remains nite as a
result of the interference of the waves that arrive at the same time.
4.17
Radial Earth structure
In a spherical earth we typically encounter three important situations that are

characterized by the geometry of the ray paths, the travel time curves T (), the
variation of the ray parameter with distance p(), and the (p) curves. In the
following, just imagine what happens if you shoot rays from an earthquake
source at the surface to increasing distances. In other words, you start of with
4.17. RADIAL EARTH STRUCTURE
169
Figure 4.15: Case 1 : Wave speed monotonously increases with depth.

a large take-o angle and you analyze what happens when you decrease this
angle (i.e. let the ray dive steeper into the Earth).
1. The situation that applies to most depth ranges in the Earths interior is
that of a steady increase in seismic wave speed (see Fig. 4.15) so that:
Ray paths : the rays sample progressively deeper regions in the Earth,
T () : and arrive at progressively larger distances.
p() : the slope of the travel time curves decrease monotonically with
increasing distance (i.e. the ray parameter decreases for waves traveling to larger distances), so there are no signicant changes in amplitude (other than those due to geometrical spreading!) (p/ < 0).
the intercept time decreases with increasing ray parameter (decreasing distance!)
A look at the travel time curves suggests that this situation is indeed very
common and describes the overall character of the curves pretty well.
Figure 4.16: Case : The presence of a low-velocity zone.

2. The rst important deviation from this situation is when there is a decrease
in wave speed with increasing depth or decreasing radius (see Fig. 4.16).
This gives rise to some interesting eects.
Ray paths : The rays will still sample progressively deeper regions
when the ray parameter decreases, but the pattern is more complex.
Initially (i.e. above the depth where the wave speed drops) the behavior is the same as in the general situation above. However, when
the ray parameter decreases further the rays interact with the low
velocity zone. (A sucient condition for the low velocity zone is
that v/r < v/r.) The decrease in wave speed results in the deection of the ray toward the vertical and the rays do not turn within
the low velocity zone; they only reect back to the Earths surface to
170
be recorded by seismometers when the wave speed increases again.
The corresponding waves arrive signicantly farther away from the
source than the ones with only a slightly less ray parameter. (You
can also say that here we have a situation where p 0 but = 0
so that the amplitude is zero.) Initially, some rays may reect at the
top of the base of the low velocity zone so that energy is projected
to shorter distances with a further decrease in ray parameter (incidence angle), but eventually, the eect of the low wave speed zone
is no longer felt and the rays sample deeper regions and behave in a
manner similar to the general situation.
In terms of ray geometry : there will be a region in the Earths interior
that is not sampled.
T () : The travel time curve will reveal a shadow zone, a region
where (according to our simplied ray theory based on the high
frequency approximation) no phases arrive. There will be a small
distance where two phases can arrive : the wave reected from the
base of the low velocity zone and the direct arrival which is the wave
that turns beneath the LVZ.
p() : Initially, p will decrease with increasing distance (p/ < 0),
and p() is continuous. When p decreases so that the ray is refracted
through the LVZ two things happen :
(a) the p() curve is no longer continuous since the ray dened by
the incrementally smaller p arrives at a dierent distance, and
(b) with decreasing p the distance initially decreases because of the
reection (p/ > 0). If p decreases even further the normal
behavior is established again (p/ < 0).
Amplitude : The amplitude is zero in the shadow zone (the p
curve is horizontal), but becomes large for arrivals at a distance
just outside the shadow zone corresponding to rays that bottom just
beneath the LVZ (the p curve is vertical).
The two most important regions in the Earth where this happens are the
low velocity layer beneath oceanic lithosphere and at the transition from
the mantle to the outer core (for P-waves).
Figure 4.17: Case 3 : A sharp increase in wave speed with depth.
4.17. RADIAL EARTH STRUCTURE
171
3. The second important deviation from the normal situation is when there
is a region where the wave speed increases rapidly with increasing depth :
v/r >> 1 (see Fig. 4.17). Lets for the discussion assume that the
increase in wave speed occurs instantly, i.e. that there is a seismic discontinuity in v/r (the function v(r) itself is of course continuous;
this situation is also known as a rst order discontinuity), but you must
realize that similar eects occur when the gradient in wave speed is steep.
Ray paths : For large incidence angles the rays turn above the discontinuity. These form the direct rays. When the incidence angle
(or, equivalently, the ray parameter) decreases the rays will reect at
the interface. The ray with the smallest ray parameter that does not
reect is called the grazing ray. The rays that are reected from
the interface form arrivals at shorter distances those corresponding
to the grazing ray. This leads to a situation where there is a distance
range where we have arrivals of both the direct and the reected
waves. The situation is slightly more complicated because when the
ray parameter continues to decrease, there is a critical angle where
the rays no longer reect but refract into the deeper earth. From
that point onward, the behavior of the rays is as one would expect
from the normal situation, and the rays go to larger distances. The
reection will cause the ray paths to cross which causes a caustic
and results in large amplitudes of the phase arrivals.
T () : The corresponding travel time curve is complicated. In the
distance range between the arrival of the waves associated with the
grazing ray and the critical ray there are, in fact, three arrivals : the
direct phase propagating through the medium above the interface,
the reected phase, and the refracted wave that propagates in part
in the medium beneath the interface. This distance range it, therefore, called the triplication range because there are, in fact, three
arrivals.
p() : For large ray parameters the behavior is as in the standard situation; a gradual increase in distance with decreasing p (p/ < 0).
When p becomes smaller than that of the grazing ray the reection
causes the distance to decrease with decreasing p (p/ > 0), but
when p decreases further and becomes smaller than the for the critical
ray the distance increases again (p/ < 0).
Amplitude : there are two points in the p() curve where p/ becomes very large (in ray theory the slope can go to innity!). These
two points correspond to the ray parameter for the grazing and critically refracted rays, respectively. Consequently, the amplitude of the
phase arrival will be large on either end of the triplication range.
172
Final remarks
It is clear that the p curves are the only curves associated with travel time
curves that are continuous in all circumstances, and this is a very attractive
property in, for instance, inversion of travel time information for Earths structure. In fact, this curve also plays a central role in the computation of synthetic
seismograms with the so-called WKBJ approximation.
A signicant body of research is based on the arrival times of rst arriving,
direct phases such as P. In triplication zones there are typically more than two
arrivals; there can be as many as 5 when triplication zones due to discontinuities
at dierent depths overlap. The identication problem is aggravated due to the
eect of the caustics on the amplitude : near the cusps in the travel time curve
the later arriving triplication phases have signicantly higher amplitude than
the rst arrival and for small signal to noise ratio in the data (for instance when
theres a small earthquake) the rst arrival that can be identied in the record
can, in fact, be a later arriving phase. This causes substantial scatter in the
arrival time data in these distance ranges.
The diculty of phase identication in the triplications due to upper mantle
discontinuities and the related uncertainty in the geometry of the ray paths
involved has important implications for the imaging of upper mantle structure,
which is more dicult than the imaging of lower mantle structure, and for the
accurate location of earthquake hypocenters using these data.
In seismological literature one encounters the terms regional and teleseismic distances. The precise boundary between these distances is not well dened. It basically refers to the distance ranges where eects of an upper mantle
low velocity layer and the discontinuities are (regional) or are not (teleseismic)
signicant. Regional distance is the distance where the associated rays bottom
in the upper mantle and transition zone (i.e. above 660 km depth) and this is
about 25 to 30 , with exact values dependent on the reference Earth model
used. Teleseismic arrivals refer to arrivals beyond the triplication range and
refer to turning rays in the lower mantle.
When waves pass through caustic (i.e. the arrivals on the receding branches
of the travel time curves, for instance the P KPAB phase and the reections o
a seismic discontinuity) the wave form will be distorted due to a 90 phase shift
in the phase term i(r, t). This will cause additional complications in picking
the arrival time by hand. A better way is to generate synthetic waveforms that
have the same phase shifts and apply cross correlation techniques.
4.18. SURFACE WAVES
4.18
173
Surface waves
Introduction
We have seen before that the solutions of the equations of motion in an unbounded, homogeneous, isotropic medium are remarkably simple and that the
total displacement eld due to a stress imbalance is completely accounted for
by propagating P and S-waves. We also discussed how this body wave eld becomes increasingly complex in the presence of interfaces, for instance the Earths
(free) surface, and the rst order seismic discontinuities such as the Moho, the
410 and 660 km discontinuities, the CMB, and the ICB. The total P- and Sdisplacement eld is then composed of up and downgoing SV and SH waves and
their interaction is controlled by the reection and transmission coecients and
by the boundary conditions at the interfaces.
In a bounded medium there is another important class of seismic waves,
the surface waves; these are caused by the interaction of body waves with the
free surface. Specically, the interaction of the P-SV eld with the free surface
results in Rayleigh waves (after Lord Rayleigh, 1842-1919) whereas the interaction of the SH wave eld with the free surface combines with internal layering
to produce Love waves (after mathematician A. E. H. Love, 1843-1940, who
predicted the existence of these waves in 1911). Later we will see that both
the body waves and the surface waves can be represented by and are equivalent with a superposition of the normal modes of free oscillation of the
Earth and it is important to be aware of the intimate relationship between these
seemingly separate descriptions of wave propagation in the Earths interior (see
Table 4.3). All body waves propagating in the Earths interior have counterparts
in both propagating surface waves or standing free oscillations. However, each
representation has distinct advantages for studying specic problems related to
Earths structure and the seismic source.
Body waves
P-SV waves
SH waves
Surface waves
Rayleigh waves
Love waves
Free oscillations
Speroidal modes
Toroidal modes
Table 4.3: Body waves, surface waves and free oscillation equivalencies.
General properties of surface waves

Surface waves propagate along the Earths surface. This seems like a rather
trivial statement but it has important implications for the amplitude of surface
waves.
The cylindrical expansion of the wave front of the waves along the Earths
surface implies that the energy of surface waves decreases as 1 over r, with r the
distance between the source and the position of the wave front. The amplitude
of surface
waves, related to the square root of the energy, therefore falls of as 1
over r. In contrast, the geometrical spreading of body waves in the Earths
174
interior implies that the energy decays as 1 over r2 so that the amplitude of
body waves decays as 1 over r. As a result of the dierence in geometrical
spreading, the amplitude of surface waves is typically much larger than that of
body waves, in particular at larger distances from the source. (The distance
from source to receiver is typically referred to as the epicentral distance).
Another implication of horizontal wave propagation and energy conservation
is that surface waves are evanescent, i.e., the amplitude decays with increasing
depth and goes to zero for very large depths. As a rule of thumb: the (fundamental mode of) surface waves are most sensitive at a depth z = /3, with the
wave length, and their sensitivity becomes very small for z > . For example,
at a period of T = 100s, the wavelength is about 450 km. Those waves are
most sensitive in the upper 180 km of the mantle (where the shear wave speed
is about 4.5 km/s).
The fact that the amplitude of surface waves decays with depth as 1 over
means that long wave length (or low frequency) waves are more sensitive to
deeper structure than high frequency waves. In combination with the fact that,
in general, the wave speed changes with depth, this explains why surface waves
are dispersive: surface waves of dierent frequency propagate with dierent
wave speeds.
Due to the dispersion, the wave form will change with increasing distance
from the source so that it becomes less clear what is meant if one talks about
the velocity of surface waves; to understand dispersion it will be necessary to
consider two denitions of propagation velocity: group and phase velocity.
The surface waves are typically of substantially lower frequency than the
body waves. Owing to the low frequency (sometimes in the same range as the
eigenfrequencies of man made constructions) and their large amplitude, surface
waves typically cause most of the earthquake damage to buildings.
Rayleigh waves
Interference between P and SV waves near the free surface4 causes a type of
displacement known as Rayleigh waves. Since the SV wave speed is smaller
than the P wave speed there is an angle of incidence for an incoming SV wave
that produces a critically refracted P wave, which propagates horizontally along
the interface (see Fig. 4.18)
In other words, P-wave energy is trapped along the surface in a natural way,
i.e., it does not require any particular wave speed variations at depth (Rayleigh
waves can, in principle, exist in a half space). To conserve energy the amplitude
of the horizontally propagating P wave must decrease with depth and vanish at
some point, i.e., a critically refracted P wave is an evanescent wave.
4 The boundary condition at the free surface is that the traction on that surface vanishes.
It is convenient to take n3 as the direction normal to the Earths surface, so that 13 = 23 =
33 = 0 and T3 = i3 ni = 0
4.18. SURFACE WAVES
175
Free Surface
i1
, , p
i2
j1
PR
PI
Free Surface
X1
SvR
X3
SvI
SvR
Figure by MIT OCW.
Figure 4.18: Free-surface interactions of an incident P and S wave.
Intermezzo 4.4 Evanescent waves

From analysis of a displacement potential it can be shown that the amplitude
A(z) of a horizontally propagating, critically refracted P-wave decays with increasing depth.
Consider the potential
(4.104)
with k the wave number vector and p and the horizontal and vertical components of the P-wave slowness. From the vector properties of the slowness it
2 = 1/2 . The horizontal slowness p (the ray parameter!),
follows that p2 +
is constant for the entire wave eld generated by the incoming SV wave, which
has a wave speed < . In the case that p = 1/c > 1/ then

=
1
p2 = i
2
p2
1
= i
2
(4.105)
so that
= B(z)ei(pxt) e z
Pn
j2
X3
= A(z)ei(krt) = A(z)ei(px+ zt)
X1
(4.106)
A similar expression can be given for the SV-wave, with instead of .

The fact that the argument of the exponential component of the amplitude
factor is real has important implications for the admissible wave speeds. Since
the wave number = kz is related to |k| = 2/, with the wavelength,
it also follows that the amplitude decay with depth is larger for small wave
lengths than for long wave lengths, and this is of fundamental importance for
the understanding of the dispersion of surface waves. (NB the horizontally
propagating, evanescent P-wave must interfere everywhere with SV-waves; this
can be achieved if there is an incoming SV-waveeld but for Rayleigh waves
the evanescent P-wave interferes with a horizontally propagating, and thus also
evanescent, SV-wave.)
Along the interface the critically refracted P-wave exists simultaneously with
the incident SV-wave; in fact, the evanescent P-waves alone do not satisfy the
stress-free boundary conditions and they cannot propagate along the interface
without coupling to SV. The interference of P and SV-wave produces a particle
, , p
176
PR
X1
SV
SVR
SVI
X1
P
ic = sin -1
X3
X3
Figure by MIT OCW.
Figure 4.19: Evanescent waves; left evanescent P wave; right evanescent S wave.
Amplitude decays exponentially with increasing distance from the interface.
motion in the x z plane that is retrograde at shallow depth, but changes to

prograde at larger depth (see Fig. 4.20). This is similar to the particle motion
in ocean waves.
Direction of wave propagation
Figure by MIT OCW.
Figure 4.20: Elliptical particle motion for Rayleigh wave propagation.
The Rayleigh wave can thus be observed at both the vertical (in the direction of z) and horizontal (radial, i.e., in the direction of x) components of the
displacement eld (see also Fig. 4.21).
Love waves
Another type of surface wave, the Love wave, is formed by interaction of
the SH-waveeld and the free surface. In contrast to the critically refracted
waves that interfere to produce Rayleigh waves, there is no critical refraction of
SH-waves (angle of incidence = angle of reection) and in order to satisfy the
boundary conditions there must be total reection of the SH-waves at the free
surface. SH energy can thus not be trapped near the surface in a half space. In
order for Love waves to exist SH energy has to be reected back to the surface
4.18. SURFACE WAVES
177
Love wave
Rayleigh wave
Figure by MIT OCW.
Figure 4.21: Love and Rayleigh wave displacement.

by a wave speed gradient at some depth; there must be a layer over a half
space with the shear wave speed in the layer lower than in the half space. If
the shear wave speed increases with depth a wave guide is formed in which
rays are multiply reected between the free surface and the turning points of
the rays. In general, some energy may leak into the half space (if the form of
SH body waves), unless the incoming SH-ray strikes the reecting interface at
(post) critical angles so that eectively - a head wave is formed and all
energy is trapped within the wave guide (see Fig. 4.22). The headwave is also
evanescent, and its amplitude decreases in with increasing depth beneath the
layer (see box).
Figure 4.22: Trapped waves in the crust.

Since Love waves are interfering SH-waves, the particle motion is purely
horizontal, in the x2 , or y, direction. Wave guides formed by a low-wave speed
layer over a faster half space occur naturally in the Earth; the wave speed in the
crust is larger than that in the mantle beneath the Moho, and at larger depths
there can be a low velocity zone in particular beneath oceanic lithosphere
that can cause ecient Love-wave propagation. Love waves are observed only
on the transverse component (parallel to x2 ) of the displacement eld.
Propagation speed
From looking at data we can make an important observation: Love waves arrive
before Rayleigh waves. Love waves propagate intrinsically faster than Rayleigh
waves, see below, but the dierence is not large enough to explain the observed
advance of the Love wave arrival. Since Love waves involve only horizontal
displacement whereas Rayleigh waves are composed of P-waves and vertically
178
polarized SV-waves, the observed advance of the Love waves suggests a form of
seismic anisotropy with faster wave propagation in the horizontal plane than
in the vertical direction (a situation known as transverse isotropy).
It can be shown, using the information given in the box below, that for
horizontally propagating waves to be evanescent they must travel with a propagation velocity c that is always smaller than the compressional wave speed ,
c = 1/p < , and also smaller than the shear wave speed , c = 1/p < . If
1/p the amplitude of the surface waves no longer decays with depth and
conservation of energy is then achieved by the leaking of energy into the half
space in the form of body waves (SV in the case of Rayleigh waves and SH in
the case of Love waves). If this happens one speaks of leaky modes.
So Rayleigh waves always propagate with a speed that is lower than the
shear wave speed. For a half space with shear wave speed 1 , the propagation
speed of the Rayleigh wave is about 0.91 . (In the Earth the situation is more
complicated because of the radial variation of both P and S-wave speed: if the
wave speed gradually increases with depth from c = 1 at the surface to c = 2
in the half space: 0.91 < cRayleigh < 0.92 ). We will see below that the surfacewave propagation speed depends on the wave length, and thus on frequency, of
the wave (dispersion). For Love waves it is slightly dierent. Here its the
head wave that is evanescent; for high-frequency waves (short wavelengths) the
evanescent head wave hardly penetrates into the half space (suppose a shear
wave speed of 2 ) so that the propagation speed is dominated by SH-propagation
in the layer over the half space (propagation speed c = 1 ). For longer period
Love waves, the head wave is sensitive to as much larger depth range and the
propagation speed gets closer to the shear wave speed in the half space (2 ).
Thus: 1 < cLove < 2 .
4.19
Sensitivity kernels
For evanescent waves such as Rayleigh and Love waves we have seen that long
wavelength waves penetrate deeper into the half space than short-wavelength
waves. As a rule of thumb, at a depth of 0.4 the amplitude is reduced to
1/e of its value at the surface, and wave propagation is inuenced by structure
anywhere in this depth interval. How exactly structure in a certain depth interval inuences a wave of a particular frequency is described by a sensitivity
kernel. They represent the maximum partical motion at a certain depth as a
function of frequency, which can be computed from a reference Earth model. A
few examples are given below.
These kernels are a sort of Greens functions and they are typically convolved with (a model of) Earth structure in order to synthesize observables such
as waveforms. (Note: we have seen someting like this before: in travel time tomography I mentioned that one solves the system of equations given by in
matrix notation Am = d, with m the model vector and d the data vector.
The matrix A contains the kernels and is therefore sometimes referreed to as
the sensitivity matrix. In the case of travel-time tomography the kernels, the
4.20. EXCITATION OF SURFACE WAVES
179
elements of A are simply the path length of a ray in a certain block.)
4.20
Excitation of surface waves
Figure 4.23: Phase speed sensitivity kernels.

Fig. 4.23 can be used to understand in qualitative sense the excitation of
surface waves by earthquakes. In general, the position of the earthquake (i.e.
the depth in our case of depth-dependent media) determines which modes can
be excited. A fundamental mode has no displacement deeper than a certain
depth; by reciprocity, a source (assume a white spectrum of the source so that it
can in principle excite all frequencies) that is located at those large depth
will not cause displacement of that fundamental mode at the surface.
4.21
Dispersion: phase and group velocity
The dependence of the depth of penetration on the period is described by the

sensitivity kernels. If the wave speed is constant in the half space the waves
associated with dierent kernels travel with the same wave speed and thus arrive
180
at the same time at a receiver at some distance from the source. But if, as is
the case in Earth, the P and S-wave speed changes with depth, the longer
period waves arrive at a dierent time than the shorter period waves. In Earth,
the propagation speed of Rayleigh waves is thus frequency-dependent, and the
waveform changes with increasing or decreasing distance from the source. This
frequency dependence of propagation speed is called dispersion. Love waves are
always dispersive since they cannot exist unless there is a layer over a half space,
with the shear wave speed in the half space larger than in the overlying layer.
As a result of dispersion the surface waveform changes with varying distance
from the source, and it is clear that one can no longer describe the wave propagation with a single wave speed. We describe the propagation velocity of the part
of the waveform that remains constant, such as the onset of the phase arrrival,
a peak, or a trough (see discussion of plane waves) with the phase velocity
c = /k. Wave packages with dierent frequencies travel at dierent velocities
and their interference results in a phenomenon known as beating (see Interm):
the propagation velocity of the envelope, which is related to the energy, of the
resulting wave train is called the group velocity U.
Peaks or troughs in the wave form, or the onset of a particular phase arrival
in the seismogram, all propagate with the phase velocity. In fact, we have
seen this before when we discussed travel time curves of the body waves, which
depend on the phase velocity. The phase velocity can thus be measured directly
from travel time curves (recall that the horizontal slowness p can be determined
from the slope of the travel time curve at a certain distance).
In Fig. 4.25 the dashed lines through A, B, etc. are travel time curves for
those phases. But note that the frequency of those phases change with distance,
so that the waveform changes. For instance, with increasing distance, the rst
arriving phase (A) is composed of waves with larger frequencies (because they
sample deeper).
The group velocity is constant for a given frequency (d = 0). Thus
the group velocity of surface waves of a particular frequency denes a straight
line through the origin and through the signal of that particular frequency on
records of ground motion at dierent distances. The group velocity decreases
as the frequency increases. As a result, high frequency phases become less
and less pronounced with increasing distance from the source (or time in the
seismogram).
The group velocity is very important: the energy in surface waves propagates
mainly in the constructively interfering wave packets, which move with the group
velocity.
Narrow-band ltering can isolate the wave packets with specic central frequencies (see Fig.4.26), and the group velocity for that frequency can then be
determined by simply dividing the path length along the surface by the observed
travel time. This technique can be used for the construction of dispersion curves
(see Sec. 4.22).
4.22. DISPERSION CURVES
181
Intermezzo 4.5 Group velocity

Consider two harmonic waves with the same amplitude but slightly dierent
frequencies (1 and 2 ), wave numbers k1 and k2 , and phase velocities k1 =
1 /c1 and k2 = 2 /c2 (see Fig. 4.24). These waves combine to give the total
displacement
u(x, t) = cos(k1 x 1 t) + cos(k2 x 2 t).
(4.107)
If we dene as the average between 1 and 2 so that 1 + = =

2 , and k1 + k = k = k2 k, with and k k, insert
it into (4.107) and apply the cosine rule 2 cos x cos y = cos(x + y) + cos(x y),
we obtain
u(x, t) = 2 cos(kx t) cos(kx t)
(4.108)
This is the product of two cosines, the second of which varies much more slowly
than the rst. The second cosine modulates the amplitude of the rst. The
propagation speed of this envelope is given by U () = /k). In the limit as
0 and k 0
U () =
dc
dc
d
= c+k
=c
dk
dk
d
(4.109)
The group velocity is related to interference of waves with slightly dierent phase
velocities; in other words U depends on c and on how c varies with frequency (or
wavelength or wave number). In the earth dc/d > 0 so that the group velocity
is typically smaller than the phase velocity.
fA= 16 Hz
CA= 5.45 km/sec
1.5/5.45 = 0.275 sec
A'
x = 0 km
fB= 18 Hz
CB= 5 km/sec
x = 1.5 km
sin t - x
C
1.5/5 = 0.3 sec

B'
U = 1.5 km/0.5 sec = 3 km/sec
0.5 sec
A+B
0.5 sec
A' + B'
0.5
1 sec
0.5
Figure by MIT OCW.
Figure 4.24: Two harmonic waves with the same amplitude but slightly dierent frequencies. The resulting beating is visible in the lowermost trace.
4.22
Dispersion curves
We have seen that the radial variation of shear wave speed causes dispersion
of the surface waves. This means that the observed surface wave dispersion
1 sec
182
Figure 4.25: Group velocity windows and phase veclocity curves.

contains structural information about the radial variation of seismic properties.
A plot of the group or phase velocity as a function of frequency is called a
dispersion curve. Their diagnostic value of 1D structure has been explored in
great detail. Typically, the curves produced from observed records are matched
with standard curves computed from an assumed reference Earth model that can
have a structure that is characteristic for a certain type of upper mantle (e.g.,
old/young continents, old/young oceans, etc.). Such analyses have produced the
rst maps of the thickness of oceanic lithosphere which revealed the increase in
thickness with increasing age of the lithosphere (or distance from the ridge),
and also underlie the discovery of the Low Velocity Zone (LVZ) at a depth of
about 100 to 200 km beneath most oceans and beneath the younger parts of
continents. Fig. 4.27 shows a variety of typical dispersion curves for dierent
tectonic provinces.
4.23
Seismology: free oscillations
Like any bounded medium, the Earth can ring like a bell and after occurrence of a big earthquake it can oscillate in normal modes with discrete
(eigen)frequencies. Normal modes of the Earth were predicted to exist in the
4.23. SEISMOLOGY: FREE OSCILLATIONS
183
Figure 4.26: Frequency-band ltering of seismograms.

early part of the 19th century when mathematicians (Poisson, Rayleigh) studied elastic wave propagation extensively. However, in absence of sensitive longperiod seismometers the normal models of free oscillation of the Earth remained
undetected until the Benio strain seismometer recorded the low-frequency signal due to a great earthquake in Kamchatka (1952). With the global network of
highly sensitive broad-band seismometers many (many more than 1500) normal
modes have now been observed and identied.
The tone of the ringing contains information about the structure of the
Earths interior. Since the entire Earth is involved in the free oscillations, the
normal modes are more sensitive to average properties and whole-earth structure
than to local anomalies. Of particular relevance is also that the low-frequency
waves have to do work against gravity so that records of the modes contain information about the density distribution within the Earth. For these reasons the
normal modes have played a central role in the development of global reference
models for seismic properties.
A second important implication of normal modes is that the displacement of
184
6.0
Oceanic Love
Group Velocity (km/sec)
5.0
Mantle Love
(G Phase)
4.0
Continental Rayleigh
Sedimentary Love
3.0
Mantle Rayleigh
Continental
Love
2.0
Oceanic Rayleigh
Sedimentary
Rayleigh
1.0
10
20
30
40 50
100
200
500
1000
Period (sec)
Figure by MIT OCW.
Figure 4.27: Dispersion curves for dierent tectonic provinces.

any number of normal modes can be summed as a Fourier series, with certain
weights for the dierent frequencies, in order to construct synthetic seismograms
(a technique known as mode summation). In fact, body and surface-wave
propagation can be simulated by superposition of a sucient number of fundamental and higher modes. In the discussion of surface waves we considered
a at Earth and an innite half space (overlain, in case of Love waves, by a
low wave speed wave guide). This is only useful to derive some fundamental
properties, in particular at relatively short periods (T < 200s), but for long
period surface waves , which penetrate deep into the Earths interior and for
the interference of waves that have propagated along the circumference of the
Earth, one must take sphericity into account. The surface waves were characterized by their frequency and wave number k. We did not consider boundaries
of the medium other than the free surface, and the frequency was taken as the
independent variable: for each frequency there are only certain discrete wave
numbers k = kn () for which the boundary conditions could be satised. Instead we could have formulated the problem in terms of discrete eigenfrequencies
= n (k) with k the independent variable. This formalism makes more sense
for the discussion of free oscillations of the Earth, since the medium is bounded.
In the spherical geometry the horizontal wave number k is xed at certain
discrete values by the nite lateral extend of the medium. One often uses the
angular wave number l instead of k, with l zero or a positive integer (see Fig.
4.28).
185
Figure by MIT OCW.
Figure 4.28: Standing waves in a spherical Earth.
Normal modes and overtones

To get some insight in the problem, lets consider the simple situation of vibrations of a string held xed at either end. The motions in the string must obey
the 1D wave equation, with c the phase velocity:
2u
1 2u
=
x2
c2 t2
The general solution of this equation is
x
(4.110)
u(x, t) = Aei(t+ c ) + Bei(t c ) + Cei(t+ c ) + Dei(t c )
(4.111)
The constants A D can be determined from the boundary conditions, i.e.

the xed end points: u(0, t) = u(L, t) = 0. The rst gives A = B and C = D.
The condition at x = L then gives

L
it
it
=0
(4.112)
)2i sin
(Ae Ce
c
which has nontrivial solutions for L/c = (n + 1), n = 0, 1, 2, 3, .
These discrete frequencies, labeled n , are called the eigenfrequencies of this
bounded system. The corresponding displacements, Eq. (4.110), are the eigenfunctions or normal modes of the system and are of the form u = exp(in t) sin(n x/c).
The fundamental mode is given for n = 0, and has no internal nodes (where
u = 0) within the system; n > 0 corresponds to higher modes or overtones,
which have n internal nodes. It is important to realize that the motion of each
of the modes occurs without horizontal motion of the nodes: they are standing waves and the modes themselves dont propagate horizontally. However,
constructive interference of the coexisting vibrations corresponds to traveling
waves. We have previously said that P and S-waves are the complete solutions
to the wave equation, and it can be shown that the normals modes of free oscillations are, in fact, not fundamentally dierent from the body waves. Normal
modes can be used to describe body wave propagation. Indeed, any propagating disturbance can be represented by an innite weighted sum of the eigen
frequencies (Fourier series!) so that normal mode summation can be used to
simulate propagating waves such as body waves and surface waves:
u(x, t) =

An ein t + Bn en t
n=0
(4.113)
186
1
X1
L
X3
n=0
n=1
n=2
Figure by MIT OCW.
Figure 4.29: A string under tension. Fundamental mode is given by n = 0; n

= 1, 2, ... are the over tones.
Power spectrum
The individual modes can, in general, not be observed directly from the seismograms. Free oscillations are studied with spectral techniques. If one was to
take a Fourier transform of a suciently long record of ground motion, typically
many hours or even days, one gets a power spectrum that reveals the distinct
eigenfrequencies of the Earths free oscillations (see Fig. 4.30).
Nomenclature of normal modes
Normal modes of free oscillation are just the solutions of the wave equation in
a spherical coordinate system and the nomenclature of the modes is therefore
based on spherical harmonics. Recall that the gravity and magnetic potentials were, in fact, summations of modes with dierent coecients (Gaussian
coecients in the case of the magnetic potential). The expression of mode summation is similar to the spherical harmonic expressions used when we discussed,
for instance, the geoid and the magnetic eld with two dierences: (1) the normalization of the harmonic coecients are typically specic to each application
(seismology, gravity, geomagnetism), but dont worry about that now, and (2)
instead of doing the summation from m = 0 to l with two (Gaussian) coe-
RADIAL MODES
0S 0
187
TOROIDAL MOTIONS
1S 0
0T2
0T3
SURFACE PATTERNS
0S 2
0S 3
0S 4
RADIAL PATTERNS
n=0
Fundamental
n=1
First Overtone
n=2
Second Overtone
n =3
Third Overtone
Figure by MIT OCW.
Figure 4.30: Surface and nodal patterns of free oscillations.

cients, in seismology one typically uses a notation that sums from m = l to +l:
in both cases there are 2l + 1 coecients (this is called a 2l + 1 degeneracy).
There are two basic types of free oscillation (1) spheroidal modes, which
are analogous to the P-SV-system and the Rayleigh waves and have a component
of motion parallel to the radius from the Earths center; and (2) toroidal or torsional modes involving shear motions parallel to the Earths surface, analogous
to SH and Love waves. Spheroidal modes involve expansion and contraction
of (parts of) the Earth, whereas toroidal modes involve dierential rotation of
parts of the globe. Gravity does not inuence the toroidal motion but longperiod spheroidal oscillations do involve signicant work against gravity; observation of these modes can therefore yield information about the Earths gross
density structure.
The toroidal and spheroidal modes are labeled n Tl and n Sl , respectively,
where n indicates the number of nodes along the radius of the Earth5 Torsional
modes are only sensitive to shear wave speed; spheroidal modes are sensitive
5 The latter would be true if the Earth was homogeneous and uniform; in reality it is more
complicated. The behavior of normal modes in the Earth is complicated by stratication, the
existence of a uid outer core, by the rotation of the sphere, and, of course, by deviations
from sphericity (3D structure + anisotropy).
188
to compressional and shear wave speed and density n is the overtone number
and l (the angular order or degree or wave number) indicates the number of
nodal planes on the surface (see Fig. 4.31).
0T2
1T2
0S2
0S3
Figure by MIT OCW.
Figure 4.31: Dierent toroidal modes (0 T2 , 1 T2 ; top) and spheroidal modes

(0 S2 , 0 S3 ; bottom).
For example, the mode 0 T2 corresponds to alternating twisting of the entire
upper and lower hemisphere of the spherical body; the mode 1 T2 corresponds
to similar twisting of the center of the sphere, but now with twisting in the
reverse direction of the outer part of the sphere (see Fig. 4.32). The modes
with n = 0 sense the gross mantle structure, and the modes with increasing n
are, in general, sensitive to elastic properties at dierent depths in the sphere.
For toroidal modes, the poles have no motion, counting as the l = 1 term. The
mode 0 T1 cannot exist. Spheroidal modes with l = 0 have no nodal planes at
the surface and are therefore sometimes called radial modes. The mode 0 S0
involves expansion and contraction of the sphere as a whole; mode 0 S2 has two
equatorial bands of zero displacement, 0 S3 has three nodal lines etc. (see Fig.
4.32).
Mode
0 S0
0 S2
0 S15
0 S30
0 S45
0 S60
0 S150
1 S2
1 S10
2 S10
Period (s)
1277.52
3223.25
426.15
262.09
193.91
153.24
66.90
1470.85
465.46
415.92
Mode
0 T2
0 T10
0 T20
0 T30
0 T40
0 T50
0 T60
1 T2
1 T10
2 T40
Period (s)
2636.38
618.97
360.03
257.76
200.95
164.70
139.46
756.57
381.65
123.56
Table 4.4: Oscillation periods of some normal modes.

Table 4.4 gives the periods of some of the observed modes. The normal mode
with the longest period is the spheroidal mode 0 S2 , with a period of about 54
189
minutes. In the last 4 decades many modes have been identied. This also is
a game of matching the observed spectra with model predictions, identifying
the modes, using that to improve the reference Earth models, and the improved
starting models may then allow the identication of previously unknown modes.
Normal mode splitting: aspherical Earths structure

We have used the notation of modes in terms of S and T and the degree l
and the overtone number n, for instance 0 S2 . Just as in the use of spherical
harmonics to describe the gravity and magnetic elds we also have the order m
in seismology. (As a reminder: there are l nodal lines at the surface: there are
m nodal lines along great circles (m=0 gives the zonal harmonics) and there
are thus l m nodal lines along latitude. For l = m: tesseral harmonics). For
each angular degree l there are 2l + 1 values for m. In a spherically symmetric,
non-rotating body the 2l + 1 modes have the same eigenfrequency, the modes
correspond to a single peak in the spectrum the overlapping peaks are known
as multiplets and this redundancy is the reason why the superscript m is
usually ignored in the notation. However, the 5 dierent modes that constitute
0 S2 have dierent angular moments and when the body is rotating the 2l + 1
peaks, or singlets do not exactly overlap any more. This phenomenon is known
as the splitting of the modes. The split modes have eigenfrequencies that are
very close together so that interference occurs.
Splitting can be caused by rotation, but also by aspherical Earths structure
such as lateral variation in isotropic seismic properties (due to dynamic processes
in the mantle) or by seismic anisotropy. Conversely, the analysis of splitting
in the power spectra can give invaluable information about 3D structure and
anisotropy.
Chapter 5
Geodynamics
5.1
Heat ow
Thermally controlled processes within Earth include volcanism, intrusion of

igneous rocks, metamorphism, convection within the mantle and outer core,
and plate tectonics. The global heat ow can be measured by measuring the
temperature gradient everywhere at the surface of the Earth. This gives us an
estimate of the mean rate of heat loss of the Earth, which can be broken up into
various components (Table 7.3 in Fowler):
Continents
Oceans
Area
(km2 )
201
309
Conductive cooling
Hydrothermal circulation
Total Earth
510
Heat Flow
(mWm2 )
58
100
[66]
[34]
83
Heat Loss
(1012 W)
11.5
30.4
[20.3]
[10.1]
41.9
The amount of heat lost through the ocean basins in enormous! up to 73%!
(The oceans cover about 60% of the Earths surface). This was a famous paradox
before the discovery of plate tectonics. It was well known that the abundance
of radioactive elements (which are a source of heat through radioactive decay)
in the ocean basins was much lower than that in the continents. So what causes
the signicantly higher heat ow in the oceans? With the discovery of plate
tectonics it was realized that most of the heat loss occurs through the cooling
and creation of oceanic lithosphere. The mean rate of plate generation therefore
depends on the balance between the rate of heat production within the Earth
and the rate of heat loss at the surface.
In this course we will address some of the basic concepts of heat ow and
Earths thermal structure, and we will discuss in some detail the cooling of
oceanic lithosphere and the implications of Earth thermal structure for mantle
convection.
191
CHAPTER 5. GEODYNAMICS
192
Heat sources
There are several possibilities for the source of heat within the earth:
1. Original or primordial heat; this is the release of heat due to the
cooling of the Earth. The amount of heat released by this process can be
estimated by calculating the heat released by a change in temperature of
1 at constant pressure. This depends on the specic heat, CP which
is the energy that is needed to heat up 1 kg of material by 1 (i.e., its a
material property).
We can do a quick calculation to nd out how much heat would be released
by dropping the temperature of the mantle by 1 C (Lets for now ignore
latent heat due to phase changes):
Mantle; for silicates:CP = 7.1 102 Jkg1 C1 ; the mass of the
mantle is about 4.11024 kg
Core; for iron: CP = 4.6 102 Jkg1 C1 ; the mass of the core is
about 1.91024 kg
For T = 1 C this gives E = 3.71027J. In absence of any other sources
for heat production, the observed global heat ux of 4.21013W can thus
be maintained by a cooling rate of 4.21013 [W] divided by 3.71027 [J]=
1.11014 Cs1 .
In other words, since the formation of Earth, 4.5 Ga ago, the average
temperature would have dropped by T 1, 500C. Note that the actual
cooling rate is much lower because there are sources of heat production.
2. Gravitational potential energy released by the transfer of material from
the surface to depths. Imagine dropping a small volume of rock from the
crust to the core. The gravitational potential energy released would be:
E = gh, with g 10ms2 and h = 3 106 m
silicates 3 103 kgm3 and iron 7 103 kgm3 , so that = 4 103
kgm3 .
E 1.2 1011 Jm3 .
The present-day heat ux would thus be equivalent to dropping a volume
of about 350 m3 every second. This is equal to dropping a 22 m thick
surface layer every million years.
So even if a small amount of net dierentiation were taking place within
the earth, this would be a signicant source of heat!
3. Radioactive decay: for an order of magnitude calculation, see Stacey 6.3.1.
The bottom line is that for the Earth a very signicant fraction of heat
loss can be attributed to radioactive decay (primarily of Uranium (U),
Thorium (Th) and Potassium (K). More, in fact, than can be accounted
for by heat production of the MORB source.
5.1. HEAT FLOW
193
Heat transfer
The actual cooling rate of the Earth depends not only on these sources of heat,
but also on the eciency at which heat is transferred to and lost at the Earths
surface.
How does heat get out of the system?
Conduction this will be discussed below in the context of the cooling of
oceanic lithosphere.
Convection For example, in the mantle and core.
Radiation most of the heat that the Earth receives from external sources
(i.e. the Sun) is radiated out.
Radiation
The net eect is that the Earth is cooling at a small rate (of the order of
50-100C per Ga!) (See Stacey (1993), p. 286.)
Figure 5.1:
194
5.2
Heat ow, geothermal gradient, diusion
The rate of heat ow by conduction across a thin layer depends on

1. the temperature contrast across the layer (T )
2. the thickness of the layer (z)
3. the ease with which heat transfer takes place (which is determined by
the thermal conductivity k). The thinner the layer and the larger the
temperature contrast (i.e., the larger the gradient in temperature), the
larger the heat ow.
In other words, the heat ow q at a point is proportional to the temperature
gradient at that point. This is summarized in Fouriers Law of conduction:
q = kT k
T
z
(5.1)
where the minus sign indicates that the direction of heat ow is from high
to low tempertaures (i.e., in the opposite dirtection of z if z is depth.). (For
simplicity we talk here about a 1D ow of heat, but Fouriers Law is also true
for a general 3D medium).
We can use this denition to formulate the conduction (or diusion) equation, which basically describes how the temperature per unit volume of material
changes with time. This change depends on
1. the amount of heat that ows in or out of the system which is described
by the divergence of heat ow
2. the amount of heat produced within the volume (denoted by the density
of heat sources A)
3. the coupling between this change in heat and a change in temperature
(which is controlled by the specic heat)
The thermal diusion equation is given by:
CP
T
= q + A
t
(5.2)
Or: the change in heat content with time equals the divergence of the heat
ow (into and out of the volume) and the generation of heat within the volume.
Combined with Fouriers Law the diusion equation can be written as
CP
T
= (kT ) + A = k2 T + A
t
(5.3)
In a situation of steady-state the diusion equation transforms to the expression of the geotherm, the variation of temperature with depth in the Earth:
k2 T + A = CP
T
A
= 0 2 T =
t
k
(5.4)
5.2. HEAT FLOW, GEOTHERMAL GRADIENT, DIFFUSION
195
If there is no heat production (by radioactive decay), i.e., A = 0, then

the temperature increases linearly with increasing depth. If A = 0 then the
temperature/depth prole is given by a second-order polymomial in z. In other
words, the curvature of the temperature-depth prole depends on the amount
of heat production (and the conductivity).
Figure 5.2: Heat production causes nonlinear geotherms.

A typical value for the geotherm is of order 20 Kkm1 , and with a value
for the conductivity k = 3.0 Wm1 K1 this gives a heat ow per unit area of
about 60 mWm2 (which is close to the global average, see table above). If the
temperature increases according to this gradient, at a depth of about 60 km a
temperature of about 1500 K is reached, which is close to or higher than the
melting temperature of most rocks. However, we know from the propagation
of shear waves that the Earths mantle behaves as a solid on short time scales
( > 0). So what is going on here? Actually, there are two things that are
important:
1. At some depth the geothermal gradient is no longer controlled by conductive cooling and adiabatic compression takes over. The temperature
gradient for adiabatic compression (i.e., the change of temperature due
to a change of pressure alone, without exchange of heat with its invironment) is much smaller than the gradient in the conducting thermal
boundary layer.
2. With increasing pressure the temperature required for melting also increases. In fact it can be shown that with increasing depth in Earths
mantle, the actual temperature increases (from about 0 C at the surface to about 3,500 1000 C at the core-mantle boundary CMB) but
the melting temperature Tm increases even more as a result of the increasing pressure. Consequently, at increasing depth in the mantle the
ratio of T over Tm (the homologous temperature) decreases. At even
larger depth, in Earths core, the temperature continues to increase, but
the melting temperature for pure iron drops (pure chemical compounds
such as pure iron typically have a lower melting temperature then
196
most mixtures such as silicate rock) so that the actual temperaute ex

ceeds the melting temperature and the material is in liquid state. Even
tough the mantle is solid it behaves as highly viscous uid so that owis possible over very long periods of time.
5000
5000
4000
Solidus
3000
3000
Temperature
2000
2000
Temperature (oC)
Temperature (K)
4000
1000
1000
0
0
0
2000
D''
6000
4000
Depth (km)
6371
5150
2891
400
670
Inner
Core
Outer Core
Lower Mantle
L = Lithosphere (0-80 km)
A = Asthenosphere (80-220 km)
D'' = Lower-Mantle D'' Layer
400, 670 = Phase Transitions
Figure by MIT OCW.
Figure 5.3: Geotherms in the Earth.
If we ignore heat production by radioactive decay we can simplify the con-
duction equation to
CP
k
T
T
2 T = 2 T
= k2 T
=
CP
t
t
(5.5)
with the thermal diusivity

=
k
CP
(5.6)
We will look at solutions of the diusion equation when we discuss the cooling
of oceanic lithosphere after its formation at the mid oceanic ridge. Before we
do that lets look at an important aspect of the diusion equation.
5.3. THERMAL STRUCTURE OF THE OCEANIC LITHOSPHERE
197
From a dimensional analysis of the diusion equation

T
= 2 T
t
(5.7)
we see that the diusivity has the dimension

of length2 time1 . We can
now dene a diusion length L as L = t.

If a temperature change occurs at some time t0 , then aftera characteristic
time interval it will have propagated over a distance L = through the
medium with diusivity . Similarly, it takes a time l2 / for a temperature
change to propagate over a distance l.
5.3
Thermal structure of the oceanic lithosphere
Introduction
The thermal structure of the oceanic lithosphere can be constrained by the
observations of:
1. Heat ow
2. Topography (depth of the ocean basins)
3. Gravity (density depends inversely on temperature)
4. Seismic velocities ( = (T ), = (T )); in particular, surface waves are
sensitive to radial variations in wave speed and surface wave dispersion
is one of the classical methods to constrain the structure of oceanic (and
continental) lithosphere.
In the following we address how the heat ow and the depth of ocean basins
is related to the cooling of oceanic lithosphere.
The conductive cooling of oceanic lithosphere when it spreads away from the
mid-oceanic ridge can be described by the diusion equation
T
= 2 T + A
t
(5.8)
We will simplify this equation by (1) ignoring the heat production by radiocative decay, so that A = 0 (this is reasonable for the oceanic lithosphere
since the basalts do not contain a signicant fraction of major radio-isotopes
Uranium, Potassium, and Thorium)1 , and (2) by assuming a 2D geometry so
that we can ignore the variations in the y direction. The latter assumption is
justied for regions away from fracture zones. With these simplications the
diusion equation would reduce to a
2
T
2T
T
(5.9)
= 2 T =
+
t
x2
z 2
198
with z the depth below the surface and x the distance from the ridge. The
variation in temperature in a direction perpendicular to the ridge (i.e., in the
spreading direction x) is usually much smaller than the vertical gradient. In that
case, the heat conduction in the x direction can be ignored, and the cooling of
a piece of lithosphere that moves along with the plate, away from the ridge, can
be described by a 1D diusion equation:
T
=
t
2T
z 2
(5.10)
(i.e., the observer, or the frame of reference, moves with the plate velocity
u = x/t). Note that, in this formulation, the time plays a dual role: it is used
as the time at which we describe the temperature at some depth z, but this also
relates to the age of the ocean oor, and thus to the distance x = ut from the
ridge axis).
Ridge
x=0
SURFACE
T = Ts
LITHOSPHERE
ISOTHERM
Tm
ASTHENOSPHERE
y
Figure by MIT OCW.
Figure 5.4: The cooling of oceanic lithosphere.
Ridge
T = Ts
T = Tm
u
q
q
t=0
x
t= u
u
q
t = t1
t = t2
Figure by MIT OCW.
Figure 5.5: Bathymetry changes with depth.
199
The assumption that the oceanic lithosphere cools by conduction alone is

pretty good, except at small distances from the ridge where hydrothermal circulation (convection!) is signicant. We will come back to this when we discuss
heat ow. There is a still ongoing debate as to the success of the simple cooling
model described below for large distances from the ridge (or, equavalently, for
large times since spreading began). This is important since it relates to the
scale of mantle convection; can the cooling oceanic lithosphere be considered as
the Thermal Boundary Layer (across which heat transfer occurs primarily by
conduction) of a large scale convection cell or is small scale convection required
to explain some of the observations discussed below? See the recent Nature
paper by Stein and Stein, Nature 359, 123129, 1992.
Cooling of oceanic lithosphere: the half-space model

The variation of temperature with time and depth can be obtained from solving
the instant cooling problem: material at a certain temperature Tm (or T0 in
Turcotte and Schubert) is instantly brought to the surface temperature where
it is exposed to surface temperature Ts (see cartoons below; for a full derivation,
see Turcotte and Schubert).
y=0
T0
y=0
t = 0
Ts
T0
y=0
T0
t = 0+
Ts
t>0
Figure by MIT OCW.
Figure 5.6: The heating of a halfspace

Diusion, or relaxation to some reference state, is described by error functions2 , and the solution to the 1D diusion equation (that satises the appropriate boundary conditions) is given by
z
T (z, t) = Tz (t) = Ts + (Tm Ts ) erf
2 t
(5.11)
or
z
T (z, t) Ts = (Tm Ts ) erf
2 t
(5.12)
with T (z, t) the temperature within the cooling boundary layer, Ts and Tm
the temperature at the surface and in the mantle, respectively, the thermal
2 So
called because they are integrations of the standard normal distribution.
200
diusivity3 , = k/Cp (k is the thermal conductivity and Cp the specic

heat), and the error function operating on some argument dened as
2
erf() =
eu du
(5.13)
The so called complementary error function, erfc, is dened simply as erfc() =

1 erf(). The values of the error function (or its complement) are often presented in table form4 . Figure 5.7 depicts the behavior of the error function:
when the argument increases the function value creeps asymptotically to a
value erf = 1.
1.0
0.8
erf
0.6
0.4
erfc
0.2
0
Figure by MIT OCW.
Figure 5.7: Error function and complimentary error function.

Lets look at the temperature according to (5.12) for dierent boundary
conditions. For large values of z the solution of the diusion equation becomes
T (, t) = Tm ; at the surface, z = 0 so that T (0, t) = Ts , and= after a very long
3 The thermal diusivity has the dimension of distance2 /time; a typical value for is 1
mm2 /s. The square root of the product t is porportional to the diusion length L t.
If the temperaure changes occur over a characteristic time interval t they will propagate a
distance of the order of L. Similarly, a time l2 / is required for temperature changes to
propagate distance l.
4 Type help erfin MatlabTM
201
time, T (z, ) = Ts , i.e., the whole system has cooled so that the temperature
is the same as the surface temperature everywhere.
For the Earth we can set Ts = 0 C so that T Tm erf(), = z/(2 t)

for most practical purposes; but the above formulas are readily applicable to
other boundary layer problems (for instance to the cooling of the lithosphere on
Venus where the surface temperature is much than that at Earth).
Examples of the geothermal gradient as a function of lithospheric age are
given in the diagram below (from Davies & Richards, Mantle Convection
J. Geol., 100, 1992).
Temperature (oC)
0
1000
5 Myr
Depth (km)
25 Myr
100 Myr
Dry solidus
100
Craton
200
Figure by MIT OCW.
Figure 5.8: Cratonic and oceanic geotherms.
Figure 5.9 (from Turcotte & Schubert, 1982) shows a series of isotherms
(lines of constant temperature (i.e, T (z, t) Ts = constant) for Tm Ts =
1300C; it shows that the depth to the isotherms as dened by (2) are hyperbola.
From this, one can readily see that if the thermal lithosphere
is bounded by
isotherms, the thickness of the lithosphereincreases as t. For back-of-theenvelope calculations you can use D 2.3 t for lithospheric thickness. (For
= 1 mm2 /s and t = 62.8 Ma, which is the average age of all ocean oceanic
202
t (Myr)
0
50
100
150
200
y (km)
400
600
50
800
1000
100
150
Figure by MIT OCW.
Figure 5.9: Oceanic geotherms.

lithosphere currently at the Earths surface, D 104 km). This thickening
occurs because the cool lithosphere reduces the temperature of the underlying
material which can then become part of the plate. On the diagram the open
circles depict estimates of lithospheric thickness from surface wave dispersion
data. Note that even though the plate is moving and the resultant geometry
is two dimensional the half space cooling model works for an observer that is
moving along with the plate. Beneath this moving reference point the plate is
getting thicker and thicker.
203
Intermezzo 5.1 lithospheric thickness from surface wave

dispersion
In the seismology classes we have discussed how dispersion curves can be used
to extract information about lithospheric structure from the seismic data. The
thickness of the high wave speed lid, the structure above the mantle low velocity zone, as determined from surface wave dispersion across parts of oceanic
lithosphere of dierent age appears to plot roughly between the 900 and 1100 C
isotherms, or at about T = 0.8 0.9Tm (see Figure 5.10). So the seismic lithosphere seems to correspond roughly to thermal lithosphere. In other words; in
short time scales i.e. the time scale appropriate for seismic wave transmission
(sec - min) most of the thermal lithosphere may act as an elastic medium,
whereas on the longer time scale the stress can be relaxed by steady state creep,
in particular in the bottom half of the plate. However, a word of caution is in order since this interpretation of the dispersion data has been disputed. Anderson
and co-workers argue (see, for instance, Anderson & Regan, GRL, vol. 10, pp.
183-186, 1983) that interpretation of surface wave dispersion assuming isotropic
media results in a signicant overestimation of the lid thickness. They have
investigated the eects of seismic anisotropy and claim that the fast isotropic
LID extends to a much cooler isotherm, at T 450 600 C, than the base of
the thermal lithosphere.
Heat ow
If we know the temperature at the surface we can deduce the heat ow by
calculating the temperature gradient:
=
=
T
T
= k
z
z
k(Tm Ts )
k
erf()

(Tm Ts ) erf() =
2 t
2 t
(5.14)
with
erf()
eu du
2
2
2
eu du = e
(5.15)
so that
q = k
(Tm Ts ) 2
e
t
(5.16)
204
Age (Ma)
100
Depth (km)
300oC
50
900oC
600oC
1200oC
100
1325oC
Figure by MIT OCW.
Figure 5.10: Elastic thickness.

For the heat ow proper we take z = 0 (q is measured at the surface!) so
that = 0 and
q = k
1
(Tm Ts )
q
t
t
(5.17)
with k the conductivity (do not confuse with , the diusivity!). The important result is that according to the half-space cooling model the heat ow
drops of as 1 over the square root of the age of the lithosphere. The heat ow
can be measured, the lithospheric age t determined from, for instance, magnetic
anomalies, and if we assume values for the conductivity and diusivity, Eq.
(5.17) can be used to determine the temperature dierence between the top and
the bottom of the plate
t
Tm Ts = q
(5.18)
k
Parsons & Sclater did this (JGR, 1977); assuming k = 3.13 JK1 m1 s1 ,
CP = 1.17103 Jkg1 K1 , and = 3.3103 kgm3 and using q = 471 mWm2
as the best t to the data they found: Tm Ts = 1350 275 C.
205
Comparison to observed heat ow data:
qs (hfu)
50
100
150
t (106 yr)
Figure by MIT OCW.
Figure 5.11: .
Near the ridge crest the observed heat ow is signicantly lower than the
heat ow predicted from the cooling half-space model. In old oceanic basins the
heat ow seems to level o at around 46 mW/m2 , which suggests that beyond
a certain age of the lithosphere the rate of conductive cooling either becomes
smaller or the cooling is partly o set by additional heat production. Possible
sources of heat which could prevent the half-space cooling are:
1. radioactivity (A is not zero!)
2. shear heating
3. small-scale convection below plate
4. hot upwelings (plumes)
206
Intermezzo 5.2 Plate-cooling Models

There are two basic models for the description of the cooling oceanic lithosphere,
a cooling of a uniform half space and the cooling of a layer with some nite
thickness. The former is referred to as the half-space model (rst described
in this context by Turcotte and Oxburgh, 1967); The latter is also known as the
plate model (rst described by McKenzie, 1967).
Both models assume that the plate moves as a unit, that the surface of the
lithosphere is at an isothermal condition of 0 C, and that the main method of
heat transfer is conduction (a good assumption, except at the ridge crest). The
major dierence (apart from the mathematical description) is that in the halfspace model the base of the lithosphere is dened by an isotherm (for instance
1300 C) so that plate thickness can grow indenitely whereas in the plate model
the plate thickness is limited by some thickness L.
The two models give the same results for young plates near the ridge crest, i.e.
the thickness is such that the bottom of the lithosphere is not yet sensed.
However, they dier signicantly after 50 Myr for heat ow predictions and 70
Myr for topography predictions. It was realized early on that at large distances
from the ridge (i.e., large ages of the lithosphere) the oceans were not as deep
and heat ow not as low as expected from the half-space cooling model (there
does not seem to be much thermal dierence between lithosphere of 80 and 160
Myr of age). The plate model was proposed to get a better t to the data, but
its conceptual disadvantage is that it does not explain why the lithosphere has
a maximum thickness of L. The half space model makes more sense physically
and its mathematical description is more straightforward. Therefore, we will
discuss only the half space cooling model, but we will also give some relevant
comparisons with the plate model.
5.4
Thermal structure of the oceanic lithosphere
Bathymetry
The second thermal eect on the evolution of the cooling lithosphere is its
subsidence or the increase in ocean depth with increasing age. This happens
because when the mantle material cools and solidies after melting at the MOR
it is heavier than the density of the underlying mantle. Since we have seen that
the plate thickens with increasing distance from the MOR and if the plate is
not allowed to subside this would result in the increase in hydrostatic pressure
at some reference depth. In other words the plate would not be in hydrostatic
equilibrium. But when the lithosphere subsides, denser material will be replaced
by lighter water so that the total weight of a certain column remains the same.
The requirement of hydrostatic equilibrium gives us the lateral variation in
depth to the ocean oor. Application of the isostatic principle gives us the
correct ocean oor topography.
Let w and m represent the density of water (w 1000 kg/m3 ) and the
mantle/asthenosphere (m 3300 kg/m3 ), respectively, and (z, t) the density
as a function of time and depth within the cooling plate. The system is in
hydrostatic equilibrium when the total hydrostatic pressure of a column under
the ridge crest at depth w + zL is the same as the pressure of a column of the
x2
x1
Ridge
207
WATER
yL
LITHOSPHERE
m
|x |
t2 = u2
|x |
t1 = u1
ASTHENOSPHERE
Figure by MIT OCW.
Figure 5.12: Oceanic isostasy.

same width at any distance from the MOR:
zL
(z, t) dz
(5.19)
zL
w(w m ) + ( m ) dz = 0
(5.20)
m (w + zL ) = ww +
0
or
According to (5.20) the mass deciency caused by w(w m ) (which is

less than 0!) is balanced by the dierence between m (>0!) integrated over
the (as yet unknown) lithospheric thickness. The lithosphere is thus heavier
than the underlying half space! (Assuming, as we do here, that the lithosphere
has the same composition as the asthenosphere). This increase in density is due
to cooling; the relationship between the change in density due to a change in
temperature is given by
d = dT m
(T Tm )
(z, t) = m + (Tm T (z, t))
(5.21)
(5.22)
with the coecient of thermal expansion, so that

zL
w(w m ) + m (Tm T ) dz = 0
0
With T = T (z, t), this gives (verify!!)
(5.23)
208
zL
z
dz
w(w m ) = m
(Tm Ts ) (Tm Ts ) erf
2 t
0
zL
z
= m (Tm Ts )
1 erf
dz
2 t
0
zL
= m (Tm Ts )
erfc
0
2 t
dz
(5.24)
we can change the integration boundary from zL to because at the base

of the lithosphere T Tm and m so that we can take the compensation
a t any depth beneath the base of the cooling lithosphere
(and erfc integrated
from zL to is very small). If we also use = z/(2 ) (see above) than we

can write
w(w m ) = m (Tm Ts )
0
z
erfc
2 t

= 2m (Tm Ts ) t
erfc()
dz
(5.25)
now use
w(t) =
erfc(q) dq =
2m (Tm Ts )
(w m )
to get
(5.26)
with w(t) the depth below the ridge crest; if the crest is at depth w0 (5.26)
becomes
2m (Tm Ts ) t
w(t) = w0 +
(5.27)
(w m )
So from the half-space cooling model it follows that the depth to the sea oor
Sclater
increases as the square root of age! Using = 3.2105 C1 , Pars ons &
(1977) found from the t to bathymetry data gives: w(t) = 2500 + +350 t [m],
for t < 70 Ma.
There is still a lively debate about the details of the parameters that give
the best t to the model, see, for instance the papers by Stein & Stein (Nature,
1992) and McNutt (Reviews of Geophysics, 1995). But despite the ongoing
209
discussions it is fair to say that these models have been very successful in predicting heat ow, topography, gravity, and have thus played a major role in
the understanding of the evolution of oceanic lithosphere with time. The typical game that is played by such successful theoretical models, in particular ones
that are so simple (= easy + fast to compute) as the cooling models is to predict
the rst order behavior of a certain process and take out that trend from the
observed data. In this case, the residual signal is then analyzed for deviations
from the simple conduction model. The addition of heat to the system (for instance by plumes) could cause anomalous topography (thermal topography)
whereas the eect of deep dynamic processes in the Earths mantle can cause
dynamic topography. Removing the eects of conduction alone thus helps to
isolate the structural signal due to other processes. This is likely to continue,
perhaps with the new model (GHD1) by Stein & Stein (1992) instead of that
by earlier workers; since regional dierences are often larger than the residual
between observed and predicted heat ow and depth curves one could question
how useful a (set of) simple model(s) is (are). For instance, if one allows the
thermal expansion coecient as a free parameters in the inversions, one might
also look into allowing lateral variation of this coecient. Davies and Richards
argue that the success of the cooling models in predicting the topography and
heat ow over almost the entire age range of oceanic lithosphere (they attribute
the deviations to the choice of the wrong sites for data which is rather questionable) indicates that conductive cooling is the predominant mode of heat loss
of most of the lithosphere (about 85% of the heat lost from the mantle ows
through oceanic lithosphere), which suggests that the lithosphere is the boundary layer of a convective system with a typical scale length dened by the plates
(plate-scale ow). They follow up on a concept tossed up by Brad Hager that
the oceanic lithosphere organizes the ow in the deeper mantle. It is for arguments such as these that question as to whether or not the topography levels
o after, say, 80Ma, in not merely of interest to statisticians. It is quite clear
that the details of the bathymetry of the oceans still contain signicant keys to
the understanding of dynamic processes in the deep interior of the Earth. One
of the remarkable aspects of the square root of time variation of ocean depth is
that it does a very good job in describing the true bathymetry, even at distances
pretty close to the MORs. This indicates that conduction alone is likely to be
the predominant mode of heat loss, even close to the MOR. The absence of any
substantial dynamic topography near the ridge crest suggests that the active,
convection related upwellings are not signicant. The upwelling is passive: the
plates are pulled apart (mainly as the result of the gravitational force, the slab
pull, acting on the subducting slabs) and the asthenospheric mantle beneath
the ridges ows to shallower depth to ll the vacancy. In doing so the material
will cross the solidus, the temperature at which rock melts (which decreases
with decreasing depth) so that the material melts. This process is known as decompression melting (see Turcotte & Schubert, Chapter 1), which results in
a shallow magma chamber beneath the MOR instead of a very deep plume-like
conduit.
210
Temperature (oC)
Depth (km)
1000
0
50
1200
1400
Depth at which partial

melting begins
Temperature of
ascending mantle
rock
Solidus
100
Figure by MIT OCW.
Figure 5.13: Pressure-release melting.
5.5
Bending, or exure, of thin elastic plate
Introduction
We have seen that upon rifting away from the MOR the lithosphere thickens (the
base of the thermal lithosphere is dened by an isotherm, usually Tm 1300C)
and subsides, and that the cooled lithosphere is more dense than the underlying
mantle. In other words, it forms a gravitationally unstable layer. Why does it
stay atop the asthenosphere instead of sinking down to produce a more stable
density stratication? That is because upon cooling the lithosphere also acquires
strength. Its weight is supported by its strength; the lithosphere can sustain
large stresses before it breaks. The initiation of subduction is therefore less
trivial than one might think and our understanding of this process is still far
from complete.
The strength of the lithosphere has important implications:
1. it means that the lithosphere can support loads, for instance by seamounts
2. the lithosphere, at least the top half of it, is seismogenic
3. lithosphere does not simply sink into the mantle at trenches, but it bends
or exes, so that it inuences the style of deformation along convergent
plate boundaries.
Investigation of the bending or exure of the plate provides important information about the mechanical properties of the lithospheric plate. We will see
that the nature of the bending is largely dependent on the exural rigidity,
which in turn depends on the elastic parameters of the lithosphere and on the
elastic thickness of the plate.
5.5. BENDING, OR FLEXURE, OF THIN ELASTIC PLATE
211
An important aspect of the derivations given below is that the thickness

of the elastic lithosphere can often be determined from surprisingly simple observations and without knowledge of the actual load. In addition, we will see
that if the bending of the lithosphere is relatively small the entire mechanical
lithosphere behaves as an elastic plate; if the bending is large some of the deformation takes place by means of ductile creep and the part of the lithosphere
that behaves elastically is thinner than the mechanical lithosphere proper.
5.5.1
Basic theory
To derive the equations for the bending of a thin elastic plate we need to
1. apply laws for equilibrium:
sumof the forces is zero and the sum of all
moments is zero:
F = 0 and
M =0
2. dene the constitutive relations between applied stress and resultant
strain
3. assume that the deection w L, the typical length scale of the system,
and h, the thickness of the elastic plate L. The latter criterion (#3) is
to justify the use of linear elasticity.
Figure 5.14: Deection of a plate under a load.

In a 2D situation, i.e., there is no change in the direction of y, the bending
of a homogeneous, elastic plate due to a load V (x) can be described by the
fourth-order dierential equation that is well known in elastic beam theory in
engineering:
D
d4 w
d2 w
+ P 2 = V (x)
4
dx
dx
(5.28)
with w = w(x) the deection, i.e., the vertical displacement of the plate,
which is, in fact, the ocean depth(!), D the exural rigidity, and P a horizontal
force.
The exural rigidity depends on elastic parameters of the plate as well as on
the thickness of the plate:
D=
Eh3
12(1 2 )
(5.29)
212
with E the Youngs modulus and the Poissons ratio, which depend on the
elastic moduli and (See Fowler, Appendix 2).
The bending of the plate results in bending (or ber) stresses within the
plate, xx ; depending on how the plate is bent, one half of the plate will be in
compression while the other half is in extension. In the center of the plate the
stress goes to zero; this denes the neutral line or plane. If the bending is not
too large, the stress will increase linearly with increasing distance z away from
the neutral line and reaches a maximum at z = h/2. The bending stress is
also dependent on the elastic properties of the plate and on how much the plate
is bent; xx elastic moduli z curvature, with the curvature dened as the
(negative of the) change in the slope d/dx(dw/dx):
xx =
Eh3 d2 w
z
1 2 dx2
(5.30)
y
xx
comp
h
y=0
ext
xx
Figure by MIT OCW.
Figure 5.15: Curvature of an elastic plate.

This stress is important to understand where the plate may break (seismicity!) with normal faulting above and reverse faulting beneath the neutral
line.
The integrated eect of the bending stress is the bending moment M ,
which results in the rotation of the plate, or a plate segment, in the x z plane.
2
M=
xx z dz
(5.31)
h
2
Equation (5.28) is generally applicable to problems involving the bending of

a thin elastic plate. It plays a fundamental role in the study of such problems
as the folding of geologic strata, the development of sedimentary basins, the
post-glacial rebound, the proper modeling of isostasy, and in the understanding
213
of seismicity. In class we will look at two important cases: (1) loading by sea
mounts, and (2) bending at the trench.
Before we can do this we have to look a bit more carefully at the dynamics of
the system. If we apply bending theory to study lithospheric exure we have to
realize that if some load V or moment M causes a deection of the plate there
will be a hydrostatic restoring force owing to the replacement of heavy mantle
material by lighter water or crustal rock. The magnitude of the restoring force
can easily be found by applying the isostasy principle and the eective load is
thus the applied load minus the restoring force (all per unit length in the y
direction): V = Vapplied wg with w the deection and g the gravitational
acceleration. This formulation also makes clear that lithospheric exure is in fact
a compensation mechanism for isostasy! For oceanic lithosphere = m w
and for continental exure = m c . The bending equation that we will
consider is thus:
D
d4 w
d2 w
+ P 2 + wg = V (x)
4
dx
dx
(5.32)
Loading by sea mounts

Lets assume a line load in the form of a chain of sea mounts, for example
Hawaii.
Figure 5.16: Deection of an elastic plate under a line load.

Let V0 be the load applied at x = 0 and V (x) = 0 for x = 0. With this
approximation we can solve the homogeneous form of (5.32) for x > 0 and take
the mirror image to get the deection w(x) for x < 0. If we also ignore the
horizontal applied force P we have to solve
D
d4 w
+ wg = 0
dx4
(5.33)
The general solution of (5.33) is

x
w(x) = e
x
x
x
x
x
A cos + B sin
+ e C cos + D sin
(5.34)
214
with the exural parameter, which plays a central role in the extraction
of structural information from the observed data:
4D
g
14
(5.35)
Courtesy of Annual Reviews Inc. Used with permission.
Figure 5.17: .
The constants A D can be determined from the boundary conditions. In
this case we can apply the general requirement that w(x) 0 for x so that
A = B = 0, and we also require that the plate be horizontal directly beneath
x = 0: dw/dx = 0 for x = 0 so that C = D: the solution becomes
w(x) = Ce
x
cos
x
x
+ sin
(5.36)
From this we can now begin to see the power of this method. The deection
w as a function of distance is an oscillation with period x/ and with an exponentially decaying amplitude. This indicates that we can determine directly
from observed bathymetry proles w(x), and from equations (5.36) and (5.29)
we can determine the elastic thickness h under the assumption of values for the
elastic parameters (Youngs modulus and Poissons ratio). The exural parameter has a dimension of distance, and denes, in fact, a typical length scale of
the deection (as a function of the strength of the plate).
The constant C can be determined from the deection at x = 0 and it can
be shown (Turcotte & Schubert) that C = (V0 3 )/(8D) w0 , the deection
215
x/
w/w0
x0/
xb/
0.5
Figure by MIT OCW.
Figure 5.18: A deection prole.

beneath the center of the load. The nal expression for the deection due to a
line load is then
V0 3 x
x
x
x0
(5.37)
e cos + sin
w(x) =
8D
Lets now look at a few properties of the solution:

The half-width of the depression can be found by solving for w = 0. From
(5.37) it follows that cos(x0 /) = sin(x0 /) or x0 / = tan1 (1)
x0 = (3/4 + n), n = 0, 1, 2, 3 . . . For n = 0 the half-width of the
depression is found to be 3/4.
The height, wb , and location, xb , of forebulge nd the optima of the
solution (5.37). By solving dw/dx = 0 we nd that sin(x/) must be zero
x = n, and for those optima w = w0 en , n = 0, 1, 2, 3 . . . For the
location of the forebulge: n = 1, xb = and the height of the forebulge
wb = w0 e or wb = 0.04w0 (very small!).
216
Important implications: The exural parameter can be determined from

the location of either the zero crossing or the location of the forebulge. No need
to know the magnitude of the load! The depression is narrow for small , which
means either a weak plate or a small elastic thickness (or both); for a plate with
large elastic thickness, or with a large rigidity the depression is very wide. In
the limit of very large D the depression is innitely wide but the amplitude w0 ,
is zero no depression at all! Once is known, information about the central
load can be obtained from Eq. (5.37)
Note: the actual situation can be complicated by lateral variations in thickness h, fracturing of the lithosphere (which inuences D), compositional layering
within the elastic lithosphere, and by the fact that loads have a nite dimensions.
Flexure at deep sea trench
With increasing distance from the MOR, or with increasing time since formation
at the MOR, the oceanic lithosphere becomes increasingly more dense and if the
conditions are right5 this gravitational instability results in the subduction of the
old oceanic plate. The gravitational instability is signicant for lithospheric ages
of about 70 Ma and more. We will consider here the situation after subduction
itself has been established; in general the plate will not just sink vertically into
the mantlebut it will bend into the trench region.
-V0
wb
x=0
-M0
x = x0
x = xb
Figure by MIT OCW.
Figure 5.19:
This bending is largely due to the gravitational force due to the negative
buoyancy of the part of the slab that is already subducted M0 . For our modeling
5 Even for old oceanic lithosphere the stresses caused by the increasing negative buoyancy
of the plate are not large enough to break the plate and initiate subduction. The actual cause
of subduction initiation is still not well understood, but the presence of pre-existing zones of
weakness (e.g. a fracture zone, thinned lithosphere due to magmatic activity e.g. an island
arc) or the initiation of bending by means of sediment loading have all been proposed (and
investigated) as explanation for the triggering of subduction.
217
we assume that the bending is due to an end load V0 and a bending moment
M0 applied at the tip of the plate. As a result of the bending moment the slope
0 at x = 0 (note the dierence with the seamount example where this
dw/dx =
slope was set to zero!). The important outcome is, again, that the parameter
of our interest, the elastic thickness h, can be determined from the shape of the
plate, in vertical cross section, i.e. from the bathymetry prole w(x)!, in the
subduction zone region, without having to know the magnitudes of V0 and M0 .
We can use the same basic equation (5.33) and the general solution (5.34)
(with A = B = 0 for the reason given above)
x
x
x
w(x) = e C cos + D sin
(5.38)
but the boundary conditions are dierent and so are the constants C and
D. At x = 0 the bending moment6 is M0 and the end load V0 . It can be
shown (Turcotte & Schubert) that the expressions for C and D are given by
C = (V0 + M0 )
2
2D
and D =
M0 2
2D
(5.39)
so that the solution for bending due to an end load and an applied bending
moment can be written as
2 ex/
x
x
w(x) =
(V0 + M0 ) cos M0 sin
(5.40)
2D
We proceed as above to nd the locations of the rst zero crossing and the
fore bulge, or outer rise.
w(x) = 0 tan(x0 /) = 1 + V0 /M0
dw/dx = 0 tan(xb /) = 1 2M0 /V0
(5.41)
(5.42)
In contrast to similar solutions for the sea mount loading case, these expressions for x0 and xb still depend on V0 and M0 . In general V0 and M0
are unknown. They can, however, be eliminated, and we can show the dependence of w(x) on x0 and xb , which can both be estimated from the nathymetry
prole. A perhaps less obvious but elegant way of doing this is to work out
tan(1/(x0 xb )). Using sine and cosine rules (see Turcotte & Schubert, 3.17)
one nds that
xb x0
tan
=1
(5.43)
so that x0 xb = (/4 + n), n = 0, 1, 2, 3, . . . For n = 0 one nds that

= 4(x0 xb )/, so that the elastic thickness h can be determined if one can
measure the horizontal distance between x0 and xb .
6 At this moment, it is important that you go back to the original derivation of the plate
equation in Turcotte & Schubert and realize they obtained their results with denite choices
as to the signs of applied loads and moments hence the negative signs.
218
2
1
100
-100
x (km)
200
300
w (km)
-1
-2
-3
-4
Figure by MIT OCW.
Figure 5.20:
After a bit of algebra one can also eliminate to nd the deection w(x)
as a function of wb , x0 , and xb . The normalized deection w/wb as a function
of normalized distance (x x0 )/(xb x0 ) is known as the Universal Flexure
Prole.
x x0
x x0
w(x)
sin
(5.44)
= 2e 4 exp
4 xb x0
wb
4 xb x0
In other words, there is a unique way to bend a laterally homogeneous elastic
plate so that it goes through the two points (x0 , 0) and (xb , wb ) with the condition that the slope is zero at x = xb . The example of the Mariana trench shown
in Figure 5.20 demonstrates the excellent t between the observed bathymetry
and the prediction after Eq. (5.44) (for a best tting elastic thickness h as
determined from the exural parameter calculated from equation (5.43).
Bending stress and seismicity

Many shallow earthquakes occur in near the convergent margin. Both in the
overriding plates as well as in the subducting plate. The latter can be attributed
to the bending stresses in the plate. The bending stress is given by Eq. (5.30).
Earthquakes are most likely to occur in the region where the bending stress is
largest (thats the place where the elastic plate is most likely to fail if there are
5.6. THE UPPER MANTLE TRANSITION ZONE
219
no pre-existing inhomogeneities such as transform faults). To nd the horizontal

location where the stress is largest we must solve
dxx
d d2 w
d3 w
=0
(5.45)
=
0
=
dx
dx dx2
dx3
This gives the location x where the stress is a maximum (or minimum!) and
substitution in (5.30), with the exural parameter D determined as above from
the bathymetry prole, then gives the amplitude of the maximum stress. If this
stress exceeds the strength of the plate, failure will occur. The mechanism of
the earthquake depends on the location relative to the neutral stress plane.
5.6
The upper mantle transition zone
Derivation of density variation with depth :

Adams-Williamson Equation
How about density? Can the radial variation in density and the elastic moduli
be constrained independently from the travel time curves? Indirectly, yes! This
was rst shown by Adams and Williamson in 1923. Here, we will only give the
basic principles and, in particular, discuss its implications for our understanding
of the Earths physical state.
The fundamental result I want you to remember is that the Adams-Williamson
equation relates the gradient in density to radial variations in seismic wave speed
(through the seismic parameter) and the mass of the Earth, which quantities
are assumed to be known, but that this result only applies to homogeneous
regions of the same physical phase.
From the travel time curves we can determine radial variations of P and
S-wave speed, (r) and (r).
2
+ 4/3
(5.46)
(5.47)
which can be combined to get what is known as the seismic parameter

4
= 2 2 =
3
(5.48)
where , , , , , and are all functions of radius:(r), (r), (r), (r),

(r), and (r).
The seismic parameter is also known as the bulk sound velocity, as the
counterpart of the shear velocity . (Notice that the incompressibility in
these equations is, in fact, the adiabatic incompressibility or bulk modulus
S because the time scale of any change in due to changes in temperature T
220
are much larger than the transit time of a seismic wave.) The aim is to show
that not only the density-normalized shear and bulk moduli can be determined,
but also the density itself (and thus and ).
In general, variations in density can be due to changes in pressure (dP ),
temperature (dT ), composition (dc) and physical phase (d), which can be
written (in gradient form) as:
d
=
dr
dP
+
dr
dT
+
dr
dc
+
dr
d
dr
(5.49)
For a homogeneous medium (same composition and phase throughout) this

equation simplies to;
d
=
dr
dP
+
dr
dT
dr
(5.50)
For the sake of the argument I will concentrate on the eect of adiabatic
compression, i.e., there is no variation of density with temperature.
d
dr
dP
dr
(5.51)
This assumption seems reasonable for most of the convecting mantle, and
leads to the original Adams-Williamson equation. For thermal boundary layers
such as the lithosphere and the lowermost mantle (D), and - in case of layered
convection a TBL between the upper and the lower mantle, an additional
gradient term has to be taken into account, and this modication has been
applied by Birch in his famous 1952 paper (see Fowler 4.3, and Stacey 5.3.1).
For adiabatic self-compression the increase in pressure that results from the
descent from radius r + dr to radius r is due to the weight of the overlying shell
with thickness r, so that the pressure gradient can be written as:
dP
= g,
dr
with
g=G
m(r)
r2
(5.52)
The other term in Eq. (5.51), the pressure derivative of the density, can be
evaluated in terms of the adiabatic bulk modulus S :
S =
dP
dP
increase in pressure
=
=
dV /V
fractional change in volume
d
(5.53)
Substitution of (5.52) and (5.53) in (5.51) and using (5.48) gives the AdamsWilliamson equation:
d
=
dr
Gm(r)
Gm(r)
=
r2
r2
(5.54)
221
which relates the density gradient to the known seismic parameter and the
gravitational attraction of the mass m(r). Rewrite for m(r)
m(r) = 4
0
(a)a da = MEarth 4
REarth
(a)a2 da
(5.55)
shows that m(r) is, in fact, the mass of the Earth less the mass of the shell
between point r and the radius of the Earth REarth . The mass of the Earth
is assumed to be know from astronomical data and is an important constraint
on the density gradient. So the only unknown in (5.55) is the density (a)
between r and REarth . We can nd a solution of (5.54) by working from the
Earths surface to larger depths: at the surface, the density of crustal rock is
fairly well known so that the density gradient can be determined for the crust.
This gradient is then used to estimate the density at the base of the crust,
which is then used to calculate the mass of the crustal shell. In this way we
can carry on the dierentiation and integration to larger depths. As already
mentioned, and explicit in (5.55), any solution of (5.54) must agree with the
total mass of the Earth, as well as with the moment of inertia, which forms the
second independent constraint on the solution. This process can only be applied
in regions where d/dr = d/dr = 0, and in this form one must also require
/T = 0, but as mentioned above there are approximations to (5.54)
that take small deviations from adiabatic compression into account.
Application of the (modied) Adams-Williamson equation by, amongst others, Bullen resulted in pretty good density models for the Earth.
The upper mantle transition zone

In 1952, Birch realized that both the density gradient and the wave speed gradients in the Earth mantle between 200 and 900 km in depth are larger than
expected from adiabatic compression only, see the abstract of his famous paper
(attached). This means that either dc/dr = 0 or d/dr = 0, or both. The mantle region where the density and wave speed gradient are larger than predicted
from adiabatic compression alone is loosely referred to as the (upper mantle)
transition zone7 .
There is no consensus yet on which one applies to the Earth but it is now
clear that d/dr =
0 is a sucient condition and is probably the most important
factor to explain the excess density. Birch suggested that phase transformations
in the Mg, Fe silicate system (Mg, Fe)2 SiO4 (olivine, spinel) and (Mg, Fe)SiO3
(pyroxene) could explain the increase in density required to explain the nonadiabatic parts of the density and wave speed gradients. We now know that
phase transformations do indeed occur in this mineral system at depths of about
410 and 660 km. Initially the sharpness of the interface as deduced from the
7 This is the original denition of the transition zone. Later, it became common to use the
term transition zone in a more restricted to indicate the mantle region between the 410 and
660 km discontinuities. In terms of mantle processes (convection, slab behavior) the original
denition is more useful.
222
Density (gcm-3)
0
10
11
12
13
1000
Depth (km)
2000
3000
4000
5000
6000
Figure by MIT OCW.
Figure 5.21: The density of the Earth according to model

ak135.
reection and phase conversion of high frequency seismic waves was used as
evidence for compositional layering and against eects of a phase change. However, experimental rock mechanics in the late eighties demonstrated that phase
changes can occur over suciently narrow depth ranges to explain the seismic
observations, see attached phase diagrams by Ito and Takahashi (JGR, 1989).
The phase changes in the (Mg, Fe) silicates play an important role in mantle dynamics because the pressure induced phase changes are also temperature
dependent! This means that phase changes can occur at dierent depth depending on the temperature of the medium in which the transformation occurs.
The temperature dependence is governed by the value of the Clapeyron slope
dP/dT of the boundaries between the stability elds of Olivine (Ol), Spinel (Sp),
and Perovskite/Magnesiow
ustite (Pv+Mw) in the P T diagram. The phase
diagrams by Ito & Takahashi at 1100C and 1600C illustrate that the phase
223
Figure 5.22: Phase diagrams of mineral transformations in

the mantle.
change occurs at smaller pressure if the temperature increase; i.e., the Clapeyron slope for the transition from Sp Pv+Mw is negative! Its a so called
endothermic phase change: upon phase transformation the material looses heat
and cools down. In contrast, Ol Sp transition that marks the phase change
at about 410 km depth has a positive Clapeyron slope and is exothermic, i.e.
there is a release of latent heat upon transformation and the material will warm
up.
What does this mean for dynamics and plate driving forces? In the diagram
I have given schematically the stability elds of Ol, Sp, and Pv+Mw, and the
boundaries between them (i.e. the Clapeyron slopes). If one would descend into
the mantle along an average mantle geotherm one would cross the Clapeyron
slope where Ol and Sp coexist at a pressure that corresponds to a depth of about
410 km and the phase line between Sp and Pv+Mw at a pressure corresponding
to about 660 km depth. Consider now the situation that a slab of cold, former
oceanic lithosphere subducts into the mantle and crosses the stability elds of
224
the silicates (assume for simplicity that the composition of the slab is the same as
the mantle which is not the case). Within the slab the phase transformation
from Ol Sp will occur at a shallower depth than in the ambient mantle.
This means that for depths just less than 410 km the density within the slab
is locally higher than in the ambient mantle, and this, in fact, gives rise to an
extra negative buoyancy force that helps the slab to subduct (it is an important
plate driving force). At 660 km the dynamical eect of the phase change is the
opposite. Inside and in the direct vicinity of the slab the phase boundary will
be depressed; consequently, the density in the slab is less than the density of
the ambient mantle which creates a buoyancy force that will resist the further
penetration of the slab.
Figure 5.23: Eects of phase transformations on downgoing

slabs.
From the diagram it is clear that the steeper the Clapeyron slope, the
stronger the dynamic eects. On the one hand, a lot of laboratory research
is focused on estimating the slopes of these phase boundaries in experimental
conditions. On the other hand, and that brings us back to seismology, seismologists attempt to estimate the topography of the seismic discontinuity and
thus constrain the clapeyron slope and asses the dynamical implications. Important classes of seismological data that have the potential to constrain both
the sharpness of and the depth to the discontinuities are reections and phase
(mode) transformations. An example of a useful reection is the underside reection of the P KIKP P KIKP , or P KPDF P KPDF , or just P P , at the 660
km discontinuity.
Since the paths of the P P phase and the P660

P are almost similar except
for near the reection point, the dierence in travel time gives direct information
about the depth to the interface. Another example is the use of SS underside
reections S660 S. Apart from proper phase identication (usually one applies
stacking techniques to suppress signal to noise) the major problem with such
techniques is that one has to make assumptions about upper mantle structure
225
between the Earths surface and the discontinuity, and these corrections are not
always reliable. The time dierence between the reections at the surface and
the discontinuity contains information about the depth to the interface, whereas
the frequency content of both the direct and the reected phase gives information about the sharpness of the interface. Also phase conversions can be used!
This line of research is still very active, and there is some consensus only about
the very long wave length variations in depth to the seismic discontinuities.

Essentials of Geophysics

Uploaded by

Copyright:

Available Formats

Essentials of Geophysics

Uploaded by

Document Information

Original Description:

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Essentials of Geophysics

Uploaded by

Copyright:

Available Formats

Chapter 1

The Earth in the Solar

1.2 Rotation and Angular Momentum

CHAPTER 1. THE EARTH IN THE SOLAR SYSTEM

where r is distance, m is the mass, is the angular velocity (=d/dt), and is

where S is the area swept out by r. Then

which is a statement of Keplers second Law of Motion: the line between

is the planetary velocity, is also conserved. It is possible to rewrite (1.1) in

1.3. THE SUN

Stellar Evolution: The Hertzsprung-Russell Diagram

The T-Tauri Stage

CHAPTER 1. THE EARTH IN THE SOLAR SYSTEM

Condensation and Cooling

1.4. PLANETARY FORMATION

sates rst appear at temperatures of 1600 1750K and consist of silicates

CHAPTER 1. THE EARTH IN THE SOLAR SYSTEM

Early Thermal State of the Earth

where T is the temperature change, Cp is specic heat, v is the absolute

where g is the gravitational acceleration at the planetary surface. It is reasonable

1.5. EARLY THERMAL STATE OF THE EARTH

CHAPTER 1. THE EARTH IN THE SOLAR SYSTEM

Table 1.5: Temperature increase due to core formation

Core radius (km)

Energy released (J)

Mean temp increase

We then re-arrange once again to integrate over the radius so that

In practice = (r) is determined from an empirically-derived equation of

1.5. EARLY THERMAL STATE OF THE EARTH

Formation of the Moon

CHAPTER 1. THE EARTH IN THE SOLAR SYSTEM

Radioactivity was discovered by Henri Becquerel in 1896 and it ultimately had

and integrating both sides of (1.16) we nd

1.6. RADIOACTIVE DECAY

where the constant c is found in the limit where t 0 to be ln No . Taking the

We may substitute (1.18) for N to nd

We want to eliminate No so we divide by (1.18) which gives

Equation (1.23) can be used directly in the determination of ages if there is

CHAPTER 1. THE EARTH IN THE SOLAR SYSTEM

The Rubidium-Strontium System

T1/2 = 48.8 109 yr .

Rb =87 Rb0 e87 t .

=87 Sr0 +87 Rb0 87 Rb

1.7. RADIOMETRIC DATING

The Uranium-Lead System

= 1.55 1010 yr1 ,

= 9.85 1010 yr1 ,

= 4.95 1011 yr1 ,

Using (1.8) and referencing to the non-radiogenic

Now take the ratio of 207 Pb/204 Pb to 206 Pb/204 Pb:

and rewrite into an isochron equation:

where the slope and y-intercept are:

CHAPTER 1. THE EARTH IN THE SOLAR SYSTEM

The Age of Earth

Radioactivity as a Heat Source

Radioactivity is an important source of heating in the early solar system and

1.9. METEORITES AND THE BULK COMPOSITION OF THE EARTH 17

1.9 Meteorites and the Bulk Composition of the

CHAPTER 1. THE EARTH IN THE SOLAR SYSTEM

the Greek word oo or chondros, meaning grain