Guide To Geophysical Equations

A Students Guide to Geophysical Equations
The advent of accessible student computing packages has meant that geophysics students
can now easily manipulate datasets and gain rst-hand modeling experience essential
in developing an intuitive understanding of the physics of the Earth. Yet to gain a more
in-depth understanding of the physical theory, and to be able to develop new models and
solutions, it is necessary to be able to derive the relevant equations from rst principles.
This compact, handy book lls a gap left by most modern geophysics textbooks,
which generally do not have space to derive all of the important formulae, showing the
intermediate steps. This guide presents full derivations for the classical equations of
gravitation, gravity, tides, Earth rotation, heat, geomagnetism, and foundational seismology, illustrated with simple schematic diagrams. It supports students through the successive steps and explains the logical sequence of a derivation facilitating self-study
and helping students to tackle homework exercises and prepare for exams.
william lowrie was born in Hawick, Scotland, and attended the University of
Edinburgh, where he graduated in 1960 with rst-class honors in physics. He achieved a
masters degree in geophysics at the University of Toronto and, in 1967, a doctorate at the
University of Pittsburgh. After two years in the research laboratory of Gulf Oil Company
he became a researcher at the Lamont-Doherty Geological Observatory of Columbia
University. In 1974 he was elected professor of geophysics at the ETH Zrich (Swiss
Federal Institute of Technology in Zurich), Switzerland, where he taught and researched
until retirement in 2004. His research in rock magnetism and paleomagnetism consisted
of deducing the Earths magnetic eld in the geological past from the magnetizations of
dated rocks. The results were applied to the solution of geologic-tectonic problems, and
to analysis of the polarity history of the geomagnetic eld. Professor Lowrie has authored
135 scientic articles and a second edition of his acclaimed 1997 textbook Fundamentals
of Geophysics was published in 2007. He has been President of the European Union of
Geosciences (19879) and Section President and Council member of the American
Geophysical Union (20002). He is a Fellow of the American Geophysical Union and
a Member of the Academia Europaea.
A Students Guide
to Geophysical Equations
WILLIAM LOWRIE
Institute of Geophysics
Swiss Federal Institute of Technology
Zurich, Switzerland
cambridge university press

Cambridge, New York, Melbourne, Madrid, Cape Town,
Singapore, So Paulo, Delhi, Tokyo, Mexico City
Cambridge University Press
The Edinburgh Building, Cambridge CB2 8RU, UK
Published in the United States of America by Cambridge University Press, New York
www.cambridge.org
Information on this title: www.cambridge.org/9781107005846
William Lowrie 2011
This publication is in copyright. Subject to statutory exception
and to the provisions of relevant collective licensing agreements,
no reproduction of any part may take place without the written
permission of Cambridge University Press.
First published 2011
Printed in the United Kingdom at the University Press, Cambridge
A catalogue record for this publication is available from the British Library
Library of Congress Cataloguing in Publication data
Lowrie, William, 1939
A students guide to geophysical equations / William Lowrie.
p. cm.
Includes bibliographical references and index.
ISBN 978-1-107-00584-6 (hardback)
1. Geophysics Mathematics Handbooks, manuals, etc.
2. Physics Formulae Handbooks, manuals, etc. 3. Earth Handbooks, manuals, etc.
I. Title.
QC809.M37L69 2011
550.10 51525dc22
2011007352
ISBN 978 1 107 00584 6 Hardback
ISBN 978 0 521 18377 2 Paperback
Cambridge University Press has no responsibility for the persistence or
accuracy of URLs for external or third-party internet websites referred to
in this publication, and does not guarantee that any content on such
websites is, or will remain, accurate or appropriate.
This book is dedicated to Marcia
Contents
Preface
Acknowledgments
1
page xi
xiii
Mathematical background
1.1 Cartesian and spherical coordinates
1.2 Complex numbers
1.3 Vector relationships
1.4 Matrices and tensors
1.5 Conservative force, eld, and potential
1.6 The divergence theorem (Gausss theorem)
1.7 The curl theorem (Stokes theorem)
1.8 Poissons equation
1.9 Laplaces equation
1.10 Power series
1.11 Leibnizs rule
1.12 Legendre polynomials
1.13 The Legendre differential equation
1.14 Rodrigues formula
1.15 Associated Legendre polynomials
1.16 Spherical harmonic functions
1.17 Fourier series, Fourier integrals, and Fourier transforms
Further reading
Gravitation
2.1 Gravitational acceleration and potential
2.2 Keplers laws of planetary motion
2.3 Gravitational acceleration and the potential of a solid
sphere
2.4 Laplaces equation in spherical polar coordinates
2.5 MacCullaghs formula for the gravitational potential
Further reading
vii
1
1
1
4
8
17
18
20
23
26
28
32
32
34
41
43
49
52
58
59
59
60
66
69
74
85
viii
Contents
Gravity
3.1 The ellipticity of the Earths gure
3.2 The geopotential
3.3 The equipotential surface of gravity
3.4 Gravity on the reference spheroid
3.5 Geocentric and geographic latitude
3.6 The geoid
Further reading
The tides
4.1 Origin of the lunar tide-raising forces
4.2 Tidal potential of the Moon
4.3 Loves numbers and the tidal deformation
4.4 Tidal friction and deceleration of terrestrial and lunar
rotations
Further reading
Earths rotation
5.1 Motion in a rotating coordinate system
5.2 The Coriolis and Etvs effects
5.3 Precession and forced nutation of Earths rotation axis
5.4 The free, Eulerian nutation of a rigid Earth
5.5 The Chandler wobble
Further reading
Earths heat
6.1 Energy and entropy
6.2 Thermodynamic potentials and Maxwells relations
6.3 The melting-temperature gradient in the core
6.4 The adiabatic temperature gradient in the core
6.5 The Grneisen parameter
6.6 Heat ow
Further reading
Geomagnetism
7.1 The dipole magnetic eld and potential
7.2 Potential of the geomagnetic eld
7.3 The Earths dipole magnetic eld
7.4 Secular variation
7.5 Power spectrum of the internal eld
7.6 The origin of the internal eld
Further reading
Foundations of seismology
8.1 Elastic deformation
86
86
88
91
96
102
106
115
116
116
119
124
130
136
137
138
140
142
155
157
169
170
171
172
176
178
179
182
197
198
198
200
205
213
214
217
225
227
227
Contents
ix
8.2 Stress
8.3 Strain
8.4 Perfectly elastic stressstrain relationships
8.5 The seismic wave equation
8.6 Solutions of the wave equation
8.7 Three-dimensional propagation of plane P- and S-waves
Further reading
228
233
239
244
252
254
258
Appendix A Magnetic poles, the dipole eld, and current loops

Appendix B Maxwells equations of electromagnetism
References
Index
259
265
276
278
Preface
This work was written as a supplementary text to help students understand the
mathematical steps in deriving important equations in classical geophysics. It is
not intended to be a primary textbook, nor is it intended to be an introduction to
modern research in any of the topics it covers. It originated in a set of handouts, a
kind of do-it-yourself manual, that accompanied a course I taught on theoretical
geophysics. The lecture aids were necessary for two reasons. First, my lectures
were given in German and there were no comprehensive up-to-date texts in the
language; the recommended texts were in English, so the students frequently
needed clarication. Secondly, it was often necessary to explain classical theory
in more detail than one nds in a multi-topic advanced textbook. To keep such a
book as succinct as possible, the intermediate steps in the mathematical derivation
of a formula must often be omitted. Sometimes the unassisted student cannot ll
in the missing steps without individual tutorial assistance, which is usually in
short supply at most universities, especially at large institutions. To help my
students in these situations, the do-it-yourself text that accompanied my lectures explained missing details in the derivations. This is the background against
which I prepared the present guide to geophysical equations, in the hope that it
might be helpful to other students at this level of study.
The classes that I taught to senior grades were largely related to potential
theory and primarily covered topics other than seismology, since this was the
domain of my colleagues and better taught by a true seismologist than by a
paleomagnetist! Theoretical seismology is a large topic that merits its own
treatment at an advanced level, and there are several textbooks of classical
and modern vintage that deal with this. However, a short chapter on the
relationship of stress, strain, and the propagation of seismic waves is included
here as an introduction to the topic.
Computer technology is an essential ingredient of progress in modern geophysics, but a well-trained aspiring geophysicist must be able to do more than
xi
xii
Preface
apply advanced software packages. A fundamental mathematical understanding

is needed in order to formulate a geophysical problem, and numerical computational skills are needed to solve it. The techniques that enabled scientists to
understand much about the Earth in the pre-computer era also underlie much of
modern methodology. For this reason, a university training in geophysics still
requires the student to work through basic theory. This guide is intended as a
companion in that process.
Historically, most geophysicists came from the eld of physics, for which
geophysics was an applied science. They generally had a sound training in
mathematics. The modern geophysics student is more likely to have begun
studies in an Earth science discipline, the mathematical background might be
heavily oriented to the use of tailor-made packaged software, and some students
may be less able to handle advanced mathematical topics without help or
tutoring. To ll these needs, the opening chapter of this book provides a
summary of the mathematical background for topics handled in subsequent
chapters.
Acknowledgments
In writing this book I have beneted from the help and support of various
people. At an early stage, anonymous proposal reviewers gave me useful
suggestions, not all of which have been acted on, but all of which were
appreciated. Each chapter was read and checked by an obliging colleague.
I wish to thank Dave Chapman, Rob Coe, Ramon Egli, Chris Finlay, Valentin
Gischig, Klaus Holliger, Edi Kissling, Emile Klingel, Alexei Kuvshinov,
Germn Rubino, Rolf Sidler, and Doug Smylie for their corrections and suggestions for improvement. The responsibility for any errors that escaped scrutiny is, of course, mine. I am very grateful to Derrick Hasterok and Dave
Chapman for providing me with an unpublished gure from Derricks Ph.D.
thesis. Dr. Susan Francis, Senior Commissioning Editor at Cambridge
University Press, gave me constant support and friendly encouragement
throughout the many months of writing, for which I am sincerely grateful.
Above all, I thank my wife Marcia for her generous tolerance of the intrusion
of this project into our retirement activities.
xiii
1
1.1 Cartesian and spherical coordinates

Two systems of orthogonal coordinates are used in this book, sometimes
interchangeably. Cartesian coordinates (x, y, z) are used for a system with
rectangular geometry, and spherical polar coordinates (r, , ) are used for
spherical geometry. The relationship between these reference systems is shown
in Fig. 1.1(a). The convention used here for spherical geometry is dened as
follows: the radial distance from the origin of the coordinates is denoted r, the
polar angle (geographic equivalent: the co-latitude) lies between the radius
and the z-axis (geographic equivalent: Earths rotation axis), and the azimuthal
angle in the xy plane is measured from the x-axis (geographic equivalent:
longitude). Position on the surface of a sphere (constant r) is described by the
two angles and . The Cartesian and spherical polar coordinates are linked as
illustrated in Fig. 1.1(b) by the relationships
x r sin cos
y r sin sin
z r cos
(1:1)
1.2 Complex numbers

The numbers we most commonly use in daily life are real numbers. Some of
them are also rational numbers. This means that they can be expressed as the
quotient of two integers, with the condition that the denominator of the quotient
must not equal zero. When the denominator is 1, the real number is an integer.
Thus 4, 4/5, 123/456 are all rational numbers. A real number can also be
irrational, which means it cannot be expressed as the quotient of two integers.
1
(a)
(b)
z = r cos
y
x
r sin
y = r sin sin
x = r sin cos
Fig. 1.1. (a) Cartesian and spherical polar reference systems. (b) Relationships
between the Cartesian and spherical polar coordinates.
Imaginary
axis
z = x + iy
+y
r sin
r cos
+x
Real
axis
Fig. 1.2. Representation of a complex number on an Argand diagram.
Familiar examples are , e (the base of natural logarithms), and some square
roots, such as 2, 3, 5, etc. The irrational numbers are real numbers that do
not terminate or repeat when expressed as decimals.
In certain analyses, such as determining the roots of an equation, it is
necessary to nd the square root of a negative real number, e.g. (y2), where
y is real. The result is an imaginary number. The negative real number can be
written as (1)y2, and its square root is then (1)y. The quantity (1) is written
i and is known as the imaginary unit, so that (y2) becomes iy.
A complex number comprises a real part and an imaginary part. For example,
z = x + iy, in which x and y are both real numbers, is a complex number with a
real part x and an imaginary part y. The composition of a complex number can
be illustrated graphically with the aid of the complex plane (Fig. 1.2). The real
part is plotted on the horizontal axis, and the imaginary part on the vertical axis.
The two independent parts are orthogonal on the plot and the complex number z
1.2 Complex numbers
is represented by their vector sum, dening a point on the plane. The distance r
of the point from the origin is given by
r
p
x2 y2
(1:2)
The line joining the point to the origin makes an angle with the real (x-)axis,
and so r has real and imaginary components r cos and r sin , respectively. The
complex number z can be written in polar form as
z rcos i sin
(1:3)
It is often useful to write a complex number in the exponential form introduced

by Leonhard Euler in the late eighteenth century. To illustrate this we make use
of innite power series; this topic is described in Section 1.10. The exponential
function, exp(x), of a variable x can be expressed as a power series as in (1.135).
On substituting x = i, the power series becomes
i2 i3 i4 i5 i6

2!
3!
4!
5!
6!
i2 i4 i6
i3 i5
1
i

2!
4!
6!
3!
5!

2 4 6
3 5
(1:4)
1 i
2! 4! 6!
3! 5!
expi 1 i
Comparison with (1.135) shows that the rst bracketed expression on the
right is the power series for cos ; the second is the power series for sin .
Therefore
expi cos i sin
(1:5)
On inserting (1.5) into (1.3), the complex number z can be written in exponential
form as
z r expi
(1:6)
The quantity r is the modulus of the complex number and is its phase.
Conversely, using (1.5) the cosine and sine functions can be dened as the
sum or difference of the complex exponentials exp(i) and exp(i):
expi expi
2
expi expi
sin
2i
cos
(1:7)

A scalar quantity is characterized only by its magnitude; a vector has both
magnitude and direction; a unit vector has unit magnitude and the direction of
the quantity it represents. In this overview the unit vectors for Cartesian
coordinates (x, y, z) are written (ex, ey, ez); unit vectors in spherical polar
coordinates (r, , ) are denoted (er, e, e). The unit vector normal to a surface
is simply denoted n.
1.3.1 Scalar and vector products

The scalar product of two vectors a and b is dened as the product of their
magnitudes and the cosine of the angle between the vectors:
a b ab cos
(1:8)
If the vectors are orthogonal, the cosine of the angle is zero and
ab 0
(1:9)
The vector product of two vectors is another vector, whose direction is perpendicular to both vectors, such that a right-handed rule is observed. The magnitude
of the vector product is the product of the individual vector magnitudes and the
sine of the angle between the vectors:
ja bj ab sin
(1:10)
If a and b are parallel, the sine of the angle between them is zero and
ab0
(1:11)
Applying these rules to the unit vectors (ex, ey, ez), which are normal to each
other and have unit magnitude, it follows that their scalar products are
ex ey ey ez ez ex 0
e x ex e y ey e z ez 1
(1:12)
The vector products of the unit vectors are

ex ey ez
ey ez ex
ez ex ey
ex ex ey ey ez ez 0
(1:13)
A vector a with components (ax, ay, az) is expressed in terms of the unit vectors
(ex, ey, ez) as
a ax ex ay ey az ez
(1:14)
The scalar product of the vectors a and b is found by applying the relationships
in (1.12):

a b ax ex ay ey az ez bx ex by ey bz ez
(1:15)
ax bx ay by az bz
The vector product of the vectors a and b is found by using (1.13):

a b ax ex ay ey az ez bx ex by ey bz ez

ay bz az by ex az bx ax bz ey ax by ay bx ez
(1:16)
This result leads to a convenient way of evaluating the vector product of two
vectors, by writing their components as the elements of a determinant, as
follows:

ex ey ez

a b ax ay az
(1:17)
bx by bz
The following relationships may be established, in a similar manner to the
above, for combinations of scalar and vector products of the vectors a, b, and c:
a b c b c a c a b
(1:18)
a b c bc a ca b
(1:19)
a b c bc a ab c
(1:20)
1.3.2 Vector differential operations

The vector differential operator is dened relative to Cartesian axes (x, y, z)
as
r ex
ey ez
x
y
z
(1:21)
The vector operator determines the gradient of a scalar function, which may
be understood as the rate of change of the function in the direction of each of the
reference axes. For example, the gradient of the scalar function with respect to
Cartesian axes is the vector
r ex
ey ez
x
y
z
(1:22)
The vector operator can operate on either a scalar quantity or a vector. The
scalar product of with a vector is called the divergence of the vector. Applied
to the vector a it is equal to

ax ex ay ey az ez
r a ex ey ez
x
y
z
ax ay az
(1:23)
x
y
z
If the vector a is dened as the gradient of a scalar potential , as in (1.22), we
can substitute potential gradients for the vector components (ax, ay, az). This
gives

r r
(1:24)
x x
y y
z z
By convention the scalar product ( ) on the left is written 2. The resulting
identity is very important in potential theory and is encountered frequently. In
Cartesian coordinates it is
r2
2 2 2
x2 y2 z2
(1:25)
The vector product of with a vector is called the curl of the vector. The
curl of the vector a may be obtained using a determinant similar to (1.17):

ex
ey
ez

r a =x =y =z
(1:26)
ax
ay
az
In expanded format, this becomes

az ay
ax az
ay ax

ex

ey

ez
ra
y
z
z
x
x
y
(1:27)
The curl is sometimes called the rotation of a vector, because of its physical
interpretation (Box 1.1). Some commonly encountered divergence and curl
operations on combinations of the scalar quantity and the vectors a and b
are listed below:
r a r a r a
(1:28)
Box 1.1. The curl of a vector

The curl of a vector at a given point is related to the circulation of the vector
about that point. This interpretation is best illustrated by an example, in
which a uid is rotating about a point with constant angular velocity .
At distance r from the point the linear velocity of the uid v is equal to r.
Taking the curl of v, and applying the identity (1.31) with constant,
r v r w r wr r w rr
(1)
To evaluate the rst term on the right, we use rectangular coordinates (x, y, z):

wr r w ex ey ez
xex yey zez
x
y
z

w ex ex ey ey ez ez 3w
(2)
The second term is
w rr

y z
xex yey zez
x
y
z
x ex y ey z ez w
(3)
Combining the results gives

r v 2w
(4)
1
w r v
2
(5)
Because of this relationship between the angular velocity and the linear
velocity of a uid, the curl operation is often interpreted as the rotation of
the uid. When v = 0 everywhere, there is no rotation. A vector that
satises this condition is said to be irrotational.
r a b b r a a r b
(1:29)
r a r a r a
(1:30)
r a b ar b br a a rb b ra
r r 0
(1:31)
(1:32)
m3
z0
n3
z
2
3
x0
m1
1
2
n2
y0 m 2
x
n1
Fig. 1.3. Two sets of Cartesian coordinate axes, (x, y, z) and (x0, y0, z0), with
corresponding unit vectors (n1, n2, n3) and (m1, m2, m3), rotated relative to each other.
r r a 0
(1:33)
r r a rr a r2 a
(1:34)
It is a worthwhile exercise to establish these identities from basic principles,

especially (1.19) and (1.31)(1.34), which will be used in later chapters.

1.4.1 The rotation matrix
Consider two sets of orthogonal Cartesian coordinate axes (x, y, z) and (x0, y0,
z0) that are inclined to each other as in Fig. 1.3. The x0-axis makes angles (1, 1,
1) with each of the (x, y, z) axes in turn. Similar sets of angles (2, 2, 2) and
(3, 3, 3) are dened by the orientations of the y0- and z0-axes, respectively, to
the (x, y, z) axes. Let the unit vectors along the (x, y, z) and (x0, y0, z0) axes be (n1,
n2, n3) and (m1, m2, m3), respectively. The vector r can be expressed in either
system, i.e., r = r(x, y, z) = r(x0, y0, z0), or, in terms of the unit vectors,
r xn1 yn2 zn3 x0 m1 y0 m2 z0 m3
(1:35)
We can write the scalar product (r m1) as

r m1 xn1 m1 yn2 m1 zn3 m1 x0
(1:36)
The scalar product (n1 m1) = cos 1 = 11 denes 11 as the direction cosine of
the x0-axis with respect to the x-axis (Box 1.2). Similarly, (n2 m1) = cos 1 = 12
and (n3 m1) = cos 1 = 13 dene 12 and 13 as the direction cosines of the
x0-axis with respect to the y- and z-axes, respectively. Thus, (1.36) is equivalent to
x0 11 x 12 y 13 z
(1:37)
On treating the y0- and z0-axes in the same way, we get their relationships to the
(x, y, z) axes:
y0 21 x 22 y 23 z
z0 31 x 32 y 33 z
(1:38)
The three equations can be written as a single matrix equation

2 3 2
32 3
2 3
x0
11 12 13
x
x
4 y0 5 4 21 22 23 54 y 5 M4 y 5
z
z
z0
31 32 33
(1:39)
The coefcients nm (n = 1, 2, 3; m = 1, 2, 3) are the cosines of the interaxial

angles. By denition, 12 = 21, 23 = 32, and 31 = 13, so the square matrix M
is symmetric. It transforms the components of the vector in the (x, y, z)
coordinate system to corresponding values in the (x0, y0, z0) coordinate system.
It is thus equivalent to a rotation of the reference axes.
Because of the orthogonality of the reference axes, useful relationships exist
between the direction cosines, as shown in Box 1.2. For example,
11 2 12 2 13 2 cos2 1 cos2 1 cos2 1

1 2
x y2 z2 1
2
r
(1:40)
and
11 21 12 22 13 23 cos 1 cos 2 cos 1 cos 2 cos 1 cos 2 0
(1:41)
The last summation is zero because it is the cosine of the right angle between the
x0-axis and the y0-axis.
These two results can be summarized as

3
X
1;
mn
mk nk
(1:42)
0;
m 6 n
k1
1.4.2 Eigenvalues and eigenvectors

The transpose of a matrix X with elements nm is a matrix with elements mn
(i.e., the elements in the rows are interchanged with corresponding elements in
10
Box 1.2. Direction cosines

The vector r is inclined at angles , , and , respectively, to orthogonal
reference axes (x, y, z) with corresponding unit vectors (ex, ey, ez), as in Fig.
B1.2. The vector r can be written
r xex yey zez
(1)
where (x, y, z) are the components of r with respect to these axes. The scalar
products of r with ex, ey, and ez are
r
ez
ex
ey
Fig. B1.2. Angles , , and dene the tilt of a vector r relative to orthogonal
reference axes (x, y, z), respectively. The unit vectors (ex, ey, ez) dene the
coordinate system.
r ex x r cos
r ey y r cos
r ez z r cos
(2)
Therefore, the vector r in (1) is equivalent to

r r cos ex r cos ey r cos ez
(3)
The unit vector u in the direction of r has the same direction as r but its
magnitude is unity:
r
u cos ex cos ey cos ez lex mey nez
r
where (l, m, n) are the cosines of the angles that the vector r makes with
the reference axes, and are called the direction cosines of r. They are
useful for describing the orientations of lines and vectors.
(4)
11
The scalar product of two unit vectors is the cosine of the angle they form.
Let u1 and u2 be unit vectors representing straight lines with direction
cosines (l1, m1, n1) and (l2, m2, n2), respectively, and let be the angle
between the vectors. The scalar product of the vectors is

u1 u2 cos l1 ex m1 ey n1 ez l2 ex m2 ey n2 ez
(5)
Therefore,
cos l1 l2 m1 m2 n1 n2
(6)
The square of a unit vector is the scalar product of the vector with itself and is
equal to 1:
uu
rr
1
r2
(7)
On writing the unit vector u as in (4), and applying the orthogonality

conditions from (2), we nd that the sum of the squares of the direction
cosines of a line is unity:

(8)
lex mey nez lex mey nez l2 m2 n2 1
the columns). The transpose of a (3 1) column matrix is a (1 3) row matrix.

For example, if X is a column matrix given by
2 3
x
X 4y5
(1:43)
z
then its transpose is the row matrix X T, where
XT x
T
The matrix equation X MX = K, where

surface:
2
11
XT MX x y z 4 21
31
z
(1:44)
K is a constant, denes a quadric

12
22
32
32 3
x
13
23 54 y 5 K
z
33
The symmetry of the matrix leads to the equation of this surface:
(1:45)
12
fx; y; z 11 x2 22 y2 33 z2 212 xy 223 yz 231 zx K (1:46)

When the coefcients nm are all positive real numbers, the geometric expression of the quadratic equation is an ellipsoid. The normal direction n to the
surface of the ellipsoid at the point P(x, y, z) is the gradient of the surface. Using
the relationships between (x, y, z) and (x0, y0, z0) in (1.39) and the symmetry of
the rotation matrix, nm = mn for n m, the normal direction has components
f
211 x 12 y 13 z 2x0
x
f
221 x 22 y 23 z 2y0
y
f
231 x 32 y 33 z 2z0
z
(1:47)
and we write
nx; y; z rf ex
f
f
f
ey e z
x
x
x
(1:48)
nx; y; z 2x0 ex y0 ey z0 ez 2rx0 ; y0 ; z0 (1:49)

The normal n to the surface at P(x, y, z) in the original coordinates is parallel to
the vector r at the point (x0, y0, z0) in the rotated coordinates (Fig. 1.4).
The transformation matrix M has the effect of rotating the reference axes from
one orientation to another. A particular matrix exists that will cause the directions of the (x0, y0, z0) axes to coincide with the (x, y, z) axes. In this case the
normal to the surface of the ellipsoid is one of the three principal axes of the
tangent
plane
z
x
(x0,y0,z0)
r
er
n
P(x,y,z)
Fig. 1.4. Location of a point (x, y, z) on an ellipsoid, where the normal n to the
surface is parallel to the radius vector at the point (x0, y0, z0).
13
ellipsoid. The component x0 is then proportional to x, y0 is proportional to y, and

z0 is proportional to z. Let the proportionality constant be . Then x0 = x,
y0 = y, and z0 = z, and we get the set of simultaneous equations
11 x 12 y 13 z 0
21 x 22 y 23 z 0
(1:50)
31 x 32 y 33 z 0
which, in matrix form, is
2
11
4 21
31
12
22
32
32 3
x
13
23 54 y 5 0
33
z
(1:51)
The simultaneous equations have a non-trivial solution only if the determinant

of coefcients is zero, i.e.,

11

21

31
12
22
32

13
23 0
33
(1:52)
This equation is a third-order polynomial in . Its three roots (1, 2, 3) are

known as the eigenvalues of the matrix M. When each eigenvalue n is inserted
in turn into (1.50) it denes the components of a corresponding vector vn, which
is called an eigenvector of M.
Note that (1.51) is equivalent to the matrix equation
2
11
4 21
31
12
22
32
32 3
2
13
x
1
23 54 y 5 4 0
33
z
0
0
1
0
32 3
0
x
0 54 y 5 0
1
z
(1:53)
which we can write in symbolic form

M IX 0
(1:54)
The matrix I, with diagonal elements equal to 1 and off-diagonal elements 0, is

called a unit matrix:
2
1 0
I 40 1
0 0
3
0
05
1
(1:55)
14
1.4.3 Tensor notation

Equations describing vector relationships can become cumbersome when written in full or symbolic form. Tensor notation provides a succinct alternative way
of writing the equations. Instead of the alphabetic indices used in the previous
section, tensor notation uses numerical indices that allow summations to be
expressed in a compact form.
Let the Cartesian coordinates (x, y, z) be replaced by coordinates (x1, x2, x3)
and let the corresponding unit vectors be (e1, e2, e3). The vector a in (1.14)
becomes
X
a a1 e1 a2 e2 a3 e3
ai e i
(1:56)
i1;2;3
A convention introduced by Einstein drops the summation sign and tacitly

assumes that repetition of an index implies summation over all values of the
index, in this case from 1 to 3. The vector a is then written explicitly
a ai e i
(1:57)
Alternatively, the unit vectors can be implied and the expression ai is understood to represent the vector a. Using the summation convention, (1.15) for the
scalar product of two vectors a and b is
a b a1 b1 a2 b2 a3 b3 ai bi
(1:58)
Suppose that two vectors a and b are related, so that each component of a is a
linear combination of the components of b. The relationship can be expressed in
tensor notation as
ai Tij bj
(1:59)
The indices i and j identify components of the vectors a and b; each index takes
each of the values 1, 2, and 3 in turn. The quantity Tij is a second-order (or
second-rank) tensor, representing the array of nine coefcients (i.e., 32). A
vector has three components (i.e., 31) and is a rst-order tensor; a scalar property
has a single (i.e., 30) value, its magnitude, and is a zeroth-order tensor.
To write the cross product of two vectors we need to dene a new quantity,
the Levi-Civita permutation tensor ijk. It has the value +1 when a permutation
of the indices is even (i.e., 123 = 231 = 312 = 1) and the value 1 when a
permutation of the indices is odd (i.e., 132 = 213 = 321 = 1). If any pair of
indices is equal, ijk = 0. This enables us to write the cross product of two vectors
in tensor notation. Let u be the cross product of vectors a and b:
u a b a2 b3 a3 b2 e1 a3 b1 a1 b3 e2 a1 b2 a2 b1 e3 (1:60)
15
In tensor notation this is written

ui ijk aj bk
(1:61)
This can be veried readily for each component of u. For example,

u1 123 a2 b3 132 a3 b2 a2 b3 a3 b2
(1:62)
The tensor equivalent to the unit matrix dened in (1.55) is known as

Kroneckers symbol, ij, or alternatively the Kronecker delta. It has the values

1; if i j
ij
(1:63)
0; if i 6 j
Kroneckers symbol is convenient for selecting a particular component of a
tensor equation. For example, (1.54) can be written in tensor form using the
Kronecker symbol:

(1:64)
ij ij xj 0
This represents the set of simultaneous equations in (1.50). Likewise, the
relationship between direction cosines in (1.42) simplies to
mk nk mn
(1:65)
in which a summation over the repeated index is implied.
1.4.4 Rotation of coordinate axes

Let vk be a vector related to the coordinates xl by the tensor Tkl
vk Tkl xl
(1:66)
A second set of coordinates xn is rotated relative to the axes xl so that the

direction cosines of the angles between corresponding axes are the elements of
the tensor nl:
x0n nl xl
(1:67)
Let the same vector be related to the rotated coordinate axes xn by the tensor
T kn:
v0k T 0kn x0n
(1:68)
vk and vk are the same vector, expressed relative to different sets of axes.
Therefore,
v0k kn vn kn Tnl xl
(1:69)
16
Equating the expressions in (1.68) and (1.69) for vk gives

0 0
T kn
xn kn Tnl xl
(1:70)
Using the relationships between the axes in (1.67),

0 0
0
xn T kn
nl xl
T kn
(1:71)
0
nl kn Tnl
T kn
(1:72)
Therefore,
On multiplying by ml and summing,

0
ml nl T kn
ml kn Tnl
(1:73)
Note that in expanded form the products of direction cosines on the left are
equal to
ml nl m1 n1 m2 n2 m3 n3 mn
(1:74)
as a result of (1.42). Therefore the transformation matrix in the rotated coordinate system is related to the original matrix by the direction cosines between
the two sets of axes:
0
ml kn Tnl
T km
(1:75)
The indices m and k can be interchanged without affecting the result. The
sequence of terms in the summation changes, but its sum does not. Therefore,
0
kl mn Tnl
T km
(1:76)
This relationship allows us to compute the elements of a matrix in a new

coordinate system that is rotated relative to the original reference axes by angles
that have the set of direction cosines nl.
1.4.5 Vector differential operations in tensor notation

In tensor notation the vector differential operator in Cartesian coordinates
becomes
r ei
xi
(1:77)
The gradient of a scalar function with respect to Cartesian unit vectors (e1, e2, e3)
is therefore
r e1
e2
e3
ei
x1
x2
x3
xi
17
(1:78)
Several shorthand forms of this equation are in common use; for example,
ri ;i i
xi
(1:79)
The divergence of the vector a is written in tensor notation as

ra
a1 a2 a3 ai
i ai
x1 x2 x3 xi
The curl (or rotation) of the vector a becomes

a3 a2
a1 a3
a2 a1
e2
e3
r a e1

x2 x3
x3 x1
x1 x2
r ai ijk
ak
ijk j ak
xj
(1:80)
(1:81)
(1:82)

If the work done in moving an object from one point to another against a force is
independent of the path between the points, the force is said to be conservative.
No work is done if the end-point of the motion is the same as the starting point;
this condition is called the closed-path test of whether a force is conservative. In
a real situation, energy may be lost, for example to heat or friction, but in an
ideal case the total energy E is constant. The work dW done against the force F is
converted into a gain dEP in the potential energy of the displaced object. The
change in the total energy dE is zero:
dE dEP dW 0
(1:83)
The change in potential energy when a force with components (Fx, Fy, Fz)
parallel to the respective Cartesian coordinate axes (x, y, z) experiences elementary displacements (dx, dy, dz) is

dEP dW Fx dx Fy dy Fz dz
(1:84)
The value of a physical force may vary in the space around its source. For
example, gravitational and electrical forces decrease with distance from a
source mass or electrical charge, respectively. The region in which a physical
18
quantity exerts a force is called its eld. Its geometry is dened by lines
tangential to the force at any point in the region. The term eld is also used to
express the value of the force exerted on a unit of the quantity. For example, the
electric eld of a charge is the force experienced by a unit charge at a given
point; the gravitational eld of a mass is the force acting on a unit of mass; it is
therefore equivalent to the acceleration.
In a gravitational eld the force F is proportional to the acceleration a. The
Cartesian components of F are therefore (max, may, maz). The gravitational
potential U is dened as the potential energy of a unit mass in the gravitational
eld, thus dEP = m dU. After substituting these expressions into (1.84) we get

dU ax dx ay dy az dz
(1:85)
The total differential dU can be written in terms of partial differentials as
dU
U
U
U
dx
dy
dz
x
y
z
(1:86)
On equating coefcients of dx, dy, and dz in these equations:

ax
U
;
x
ay
U
;
y
az
U
z
(1:87)
These relationships show that the acceleration a is the negative gradient of a

scalar potential U:
a rU
(1:88)
Similarly, other conservative elds (e.g., electric, magnetostatic) can be derived

as the gradient of the corresponding scalar potential. According to the vector
identity (1.32) the curl of a gradient is always zero; it follows from (1.88) that a
conservative force-eld F satises the condition
rF0
(1:89)
1.6 The divergence theorem (Gausss theorem)

Let n be the unit vector normal to a surface element of area dS. The ux d of a
vector F across the surface element dS (Fig. 1.5) is dened to be the scalar product
d F n dS
(1:90)
If the angle between F and n is , the ux across dS is

d F dS cos
(1:91)
1.6 The divergence theorem
19
F
n
dS
dS n
Fig. 1.5. The ux of a vector F across a small surface dS, whose normal n is
inclined to the vector, is equal to the ux across a surface dSn normal to the vector.
dz
Fx
Fx + dFx
y
x
dy
x + dx
x
Fig. 1.6. Figure for computing the change in the ux of a vector in the x-direction
for a small box with edges (dx, dy, dz).
where F is the magnitude of F. Thus the ux of F across the oblique surface dS is

equivalent to that across the projection dSn (=dS cos ) of dS normal to F.
Consider the net ux of the vector F through a rectangular box with edges dx,
dy, and dz parallel to the x-, y-, and z-axes, respectively (Fig. 1.6). The area dSx
of a side normal to the x-axis equals dy dz. The x-component of the vector at x,
where it enters the box, is Fx, and at x + dx, where it leaves the box, it is Fx + dFx.
The net ux in the x-direction is
dx Fx dFx Fx dSx dFx dy dz
(1:92)
If the distance dx is very small, the change in Fx may be written to rst order as
dFx
Fx
dx
x
(1:93)
20
The net ux in the x-direction is therefore

dx
Fx
Fx
dx dy dz
dV
x
x
(1:94)
where dV is the volume of the small element. Similar results are obtained for the
net ux in each of the y- and z-directions. The total ux of F through the
rectangular box is the sum of these ows:
d dx dy dz

d

Fx Fy Fz
dV r FdV
x
y
z
(1:95)
(1:96)
We can equate this expression with the ux dened in (1.90). The ux through a
nite volume V with a bounding surface of area S and outward normal unit
vector n is
ZZZ
ZZ
r FdV
F n dS
(1:97)
V
This is known as the divergence theorem, or Gausss theorem, after the German
mathematician Carl Friedrich Gauss (17771855). Note that the surface S in
Gausss theorem is a closed surface, i.e., it encloses the volume V. If the ux of F
entering the bounding surface is the same as the ux leaving it, the total ux is
zero, and so
rF 0
(1:98)
This is sometimes called the continuity condition because it implies that ux is

neither created nor destroyed (i.e., there are neither sources nor sinks of the
vector) within the volume. The vector is said to be solenoidal.
1.7 The curl theorem (Stokes theorem)

Stokes theorem relates the surface integral of the curl of a vector to the
circulation of the vector around a closed path bounding the surface. Let the
vector F pass through a surface S which is divided into a grid of small elements
(Fig. 1.7). The area of a typical surface element is dS and the unit vector n
normal to the element species its orientation.
First, we evaluate the work done by F around one of the small grid elements,
ABCD (Fig. 1.8). Along each segment of the path we need to consider only the
1.7 The curl theorem
21
n
S
C
D
dS
C
Fig. 1.7. Conguration for Stokes theorem: the surface S is divided into a grid of
elementary areas dS and is bounded by a closed circuit C.
(x + dx, y + dy)
(x, y + dy)
D
dy
Fy
(x, y)
dx
(x + dx, y)
Fx
Fig. 1.8. Geometry for calculation of the work done by a force F around a small
rectangular grid.
vector component parallel to that segment. The value of F may vary with
position, so, for example, the x-component along AB may differ from the
x-component along CD. Provided that dx and dy are innitesimally small, we
can use Taylor series approximations for the components of F (Section 1.10.2).
To rst order we get
Fx
dy
y

Fy
dx
Fy BC Fy DA
x
Fx CD Fx AB
(1:99)
The work done in a circuit around the small element ABCD is the sum of the
work done along each individual segment:
xZdx
I
Fdl
ABCD
ydy
Z

Fy BC dy
Fx AB dx
x
Zx
Zy
Fx CD dx
xdx
Fy

DA
dy
ydy
(1:100)
22
xdx
Z
Fdl

Fx AB Fx CD dx
ABCD
ydy
Z

Fy BC Fy DA dy
(1:101)
Substituting from (1.99) gives
xdx
Z
ydy

Z
Fx
Fy
dy dx
dx dy

y
x
Fdl
ABCD
(1:102)
The mean-value theorem allows us to replace the integrands over the tiny
distances dx and dy by their values at some point in the range of integration:

I
Fy Fx

dx dy
(1:103)
Fdl
x
y
ABCD
The bracketed expression is the z-component of the curl of F

I
F d l r Fz dx dy
(1:104)
ABCD
The normal direction n to the small area dS = dx dy is parallel to the z-axis (i.e.,
out of the plane of Fig. 1.8), and hence is in the direction of ( F)z. Thus,
I
F d l r Fn dS
(1:105)
ABCD
The circuit ABCD is one of many similar grid elements of the surface S. When
adjacent elements are compared, the line integrals along their common boundary are equal and opposite. If the integration is carried out for the entire surface
S, the only surviving parts are the integrations along the bounding curve C
(Fig. 1.7). Thus
ZZ
I
r F n dS F d l
(1:106)
S
This equation is known as Stokes theorem, after the English mathematician

George Gabriel Stokes (18191903). It enables conversion of the surface integral
of a vector to a line integral. The integration on the left is made over the surface S
through which the vector F passes. The closed integration on the right is made
around the bounding curve C to the surface S; d l is an innitesimal element of this
23
boundary. The direction of dl around the curve is right-handed with respect to the
surface S, i.e., positive when the path is kept to the right of the surface, as in Fig. 1.7.
Note that the surface S in Stokes theorem is an open surface; it is like the
surface of a bowl with the bounding curve C as its rim. The integration of F
around the rim is called the circulation of F about the curve C. If the integral is
zero, there is no circulation and the vector F is said to be irrotational.
Comparison with the left-hand side shows that the condition for this is
rF0
(1:107)
As shown in Section 1.5, this is also the condition for F to be a conservative eld.

The derivations in this and the following sections are applicable to any eld that
varies as the inverse square of distance from its source. Gravitational acceleration is
used as an example, but the electric eld of a charge may be treated in the same way.
Let S be a surface enclosing an observer at P and a point mass m. Let dS be
a small element of the surface at distance r in the direction er from the mass
m, as in Fig. 1.9. The orientation of dS is specied by the direction n normal
to the surface element. With G representing the gravitational constant (see
Section 2.1), the gravitational acceleration aG at dS is given by
aG G
m
er
r2
(1:108)
Let be the angle between the radius and the direction n normal to the surface
element, and let the projection of dS normal to the radius be dSn. The solid angle
d with apex at the mass is dened as the ratio of the normal surface element
dSn to the square of its distance r from the mass (Box 1.3):
S
dS
m
aG
er
n
dSn
Fig. 1.9. Representation of the ux of the gravitational acceleration aG through a

closed surface S surrounding the source of the ux (the point mass m).
24
Box 1.3. Denition of a solid angle

A small element of the surface of a sphere subtends a cone with apex at the
center of the sphere (Fig. B1.3(a)). The solid angle is dened as the ratio of
the area A of the surface element to the square of the radius r of the sphere:
(a)
r d
(b)
d
dA
r
r
dA
r sin d
r sin
Fig. B1.3. (a) Relationship of the solid angle , the area A of an element
subtended on the surface of a sphere, and the radius r of the sphere. (b) The
surface of a sphere divided into rings, and each ring into small surface elements
with sides r d and r sin d.
A
r2
(1)
This denition can be used for an arbitrarily shaped surface. If the surface
is inclined to the radial direction it must be projected onto a surface normal
to the radius, as in Fig. 1.5. For example, if the normal to the surface A
makes an angle with the direction from the apex of the subtended cone, the
projected area is A cos and the solid angle subtended by the area is
A cos
r2
(2)
As an example, let the area on the surface of a sphere be enclosed by a small

circle (Fig. B1.3(b)). Symmetry requires spherical polar coordinates to
describe the area within the circle. Let the circle be divided into concentric
rings, and let the half-angle subtended by a ring at the center of the sphere be .
The radius of the ring is r sin and its width is r d. Let the angular position
of a small surface element of the ring be ; the length of a side of the element
is then r sin d. The area dA of a surface area element is equal to r2 sin
d d. The solid angle subtended at the center of the sphere by the element
of area dA is
d
r2 sin d d
sin d d
r2
(3)
25
This expression is also equivalent to the element of area on the surface of

a unit sphere (one with unit radius). Integrating in the ranges 0 2 and
0 0, we get the solid angle 0 subtended by a circular region of
the surface of a sphere dened by a half-apex angle 0:
Z0 Z2
0
sin d d 2 1 cos 0
(4)
0 0
The unit of measurement of solid angle is the steradian, which is analogous

to the radian in plane geometry. The maximum value of a solid angle is
when the surface area is that of the complete sphere, namely 4r2. The
solid angle at its center then has the maximum possible value of 4. This
result is also obtained by letting the half-apex angle 0 in (4) increase to its
maximum value .
dSn dS cos er ndS
r2
r2
r2
(1:109)
The ux dN of the gravitational acceleration aG through the area element is

m
er ndS
r2
(1:110)
cos dS
Gm d
r2
(1:111)
dN aG n dS G
dN Gm
If we integrate this expression over the entire surface S we get the total gravitational ux N,
ZZ
Z
N
aG n dS Gm d 4Gm
(1:112)
Now we replace this surface integral by a volume integration, using the divergence theorem (Section 1.6)
ZZZ
ZZ
r aG dV
aG n dS 4Gm
(1:113)
V
This is valid for any point mass m inside the surface S. If the surface encloses
many point masses we may replace m with the sum of the point masses. If mass
26
is distributed in the volume with mean density , a volume integral can replace
the enclosed mass:
ZZZ
ZZZ
r aG dV 4G
dV
(1:114)
V
ZZZ
r aG 4GdV 0
(1:115)
For this to be generally true the integrand must be zero. Consequently,

r aG 4G
(1:116)
The gravitational acceleration is the gradient of the gravitational potential UG as

in (1.88):
r rUG 4G
r2 UG 4G
(1:117)
(1:118)
Equation (1.118) is known as Poissons equation, after Simon-Denis Poisson

(17811840), a French mathematician and physicist. It describes the gravitational potential of a mass distribution at a point that is within the mass distribution. For example, it may be used to compute the gravitational potential at a
point inside the Earth.

Another interesting case is the potential at a point outside the mass distribution.
Let S be a closed surface outside the point mass m. The radius vector r from the
point mass m now intersects the surface S at two points A and B, where it forms
angles 1 and 2 with the respective unit vectors n1 and n2 normal to the surface
(Fig. 1.10). Let er be a unit vector in the radial direction. Note that the outward
normal n1 forms an obtuse angle with the radius vector at A. The gravitational
acceleration at A is a1 and its ux through the surface area dS1 is

Gm
dN1 a1 n1 dS1 2 r n1 dS1
(1:119)
r1

Gm
cos 1 dS1
(1:120)
dN1 2 cos 1 dS1 Gm
r1
r21
27
B
m
a1
a2
dS1
n1
dS2
er
n2
Fig. 1.10. Representation of the gravitational ux through a closed surface S that

does not enclose the source of the ux (the point mass m).
dN1 Gm d
(1:121)
The gravitational acceleration at B is a2 and its ux through the surface area dS2 is
dN2 a2 n2 dS2 Gm
cos 2 dS2
r22
dN2 Gm d
(1:122)
(1:123)
The total contribution of both surfaces to the gravitational ux is

dN dN1 dN2 0
(1:124)
Thus, the total ux of the gravitational acceleration aG through a surface S that

does not include the point mass m is zero. By invoking the divergence theorem
we have for this situation
Z
Z
r aG dV aG n dS 0
(1:125)
V
For this result to be valid for any volume, the integrand must be zero:
r aG r rUG 0
(1:126)
r2 UG 0
(1:127)
Equation (1.127) is Laplaces equation, named after Pierre Simon, Marquis de

Laplace (17491827), a French mathematician and physicist. It describes the
gravitational potential at a point outside a mass distribution. For example, it is
applicable to the computation of the gravitational potential of the Earth at an
external point or on its surface.
In Cartesian coordinates, which are rectilinear, Laplaces equation has the
simple form
28
2 UG 2 UG 2 UG
0
x2
y2
z2
(1:128)
Spherical polar coordinates are curvilinear and the curvature of the angular
coordinates results in a more complicated form:

1 2 UG
1
UG
1 2 UG
0 (1:129)
r
sin
r2 r
r2 sin
r2 sin2 2
1.10 Power series

A function (x) that is continuous and has continuous derivatives may be
approximated by the sum of an innite series of powers of x. Many mathematical functions e.g., sin x, cos x, exp(x), ln(1 + x) fulll these conditions of
continuity and can be expressed as power series. This often facilitates the
calculation of a value of the function. Three types of power series will be
considered here: the MacLaurin, Taylor, and binomial series.
1.10.1 MacLaurin series

Let the function (x) be written as an innite sum of powers of x:
fx a0 a1 x a2 x2 a3 x3 a4 x4 an xn
(1:130)
The coefcients an in this sum are constants. Differentiating (1.130) repeatedly

with respect to x gives
df
a1 2a2 x 3a3 x2 4a4 x3 nan xn1
dx
d 2f
2a2 3 2a3 x 4 3a4 x2 nn 1an xn2
dx2
d 3f
3 2a3 4 3 2a4 x nn 1n 2an xn3
dx3
(1:131)
After n differentiations, the expression becomes
d nf
nn 1n 2 . . . 3 2 1 an terms containing powers of x
dxn
(1:132)
Now we evaluate each of the differentiations at x = 0. Terms containing powers
of x are zero and
1.10 Power series

df
f0 a0 ;
a1
dx x0
2
3
d f
d f
2a
;
3 2a3
2
dx2 x0
dx3 x0
n
d f
nn 1n 2 . . . 3 2 1an n!an
dxn x0
29
(1:133)
On inserting these values for the coefcients into (1.130) we get the power
series for (x):

df
x2 d 2 f
x3 d 3 f

fx f0 x
dx x0 2! dx2 x0 3! dx3 x0
(1:134)

xn d n f

n! dxn x0
This is the MacLaurin series for (x) about the origin, x = 0. It was derived in the
eighteenth century by the Scottish mathematician Colin MacLaurin (1698
1746) as a special case of a Taylor series.
The MacLaurin series is a convenient way to derive series expressions for
several important functions. In particular,
x3 x5 x7
x2n1
1n1

3! 5! 7!
2n 1!
x2 x 4 x6
x2n2

cos x 1 1n1
2! 4! 6!
2n 2!
(1:135)
x2 x3
xn1
x
expx e 1 x

2! 3!
n 1!
x2 x3 x4
xn

ln1 x loge 1 x x 1n1
2
3
4
n
sin x x
1.10.2 Taylor series

We can write the power series in (1.134) for (x) centered on any new origin, for
example x = x0. To do this we substitute (x x0) for x in the above derivation.
The power series becomes

df
x x0 2 d 2 f
fx fx0 x x0
2!
dx xx0
dx2 xx0

x x0 3 d 3 f
x x0 n d n f

dx3 xx0
dxn xx0
3!
n!
(1:136)
30
This is called a Taylor series, after an English mathematician, Brooks Taylor

(16851731), who described its properties in 1712.
The MacLaurin and Taylor series are both approximations to the function
(x). The remainder between the true function and its power series is a measure
of how well the function is expressed by the series.
1.10.3 Binomial series

Finite series
An important series is the expansion of the function (x) = (a + x)n. If n is a
positive integer, the expansion of (x) is a nite series, terminating after (n + 1)
terms. Evaluating the series for some low values of n gives the following:
n 0 : a x0 1
n 1 : a x1 a x
n 2 : a x2 a2 2ax x2
(1:137)
n 3 : a x3 a3 3a2 x 3ax2 x3
n 4 : a x4 a4 4a3 x 6a2 x2 4ax3 x4
The general expansion of (x) is therefore
nn 1 n2 2
a x
12
nn 1 . . . n k 1 nk k
a x xn
k!
a xn an nan1 x
(1:138)
The coefcient of the general kth term is equivalent to

nn 1 . . . n k 1
n!
k!
k!n k!
(1:139)
This is called the binomial coefcient.

When the constant a is equal to 1 and n is a positive integer, we have the
useful series expansion
1 xn
n
X
n!
xk
k!
n

k
!
k0
(1:140)
Innite series
If the exponent in (1.140) is not a positive integer, the series does not terminate,
but is an innite series. The series for (x) = (1 + x)p, in which the exponent p is
not a positive integer, may be derived as a MacLaurin series:
1.10 Power series

df
dx
p1 xp1
x0
x0
31
d 2f
pp 11 xp2
pp 1
2
x0
dx x0
n
d f
pp1 . . . pn11xpn x0 pp 1 . . . p n 1
dxn x0
(1:141)
On inserting these terms into (1.134), and noting that (0) = 1, we get for the
binomial series
pp 1 2 pp 1p 2 3
x
x
12
123
pp 1 . . . p n 1 n
x
n!
1 xp 1 px
(1:142)
If the exponent p is not an integer, or p is negative, the series is convergent in the

range 1 < x < 1.
1.10.4 Linear approximations

The variations in some physical properties over the surface of the Earth are small
in relation to the main property. For example, the difference between the polar
radius c and the equatorial radius a expressed as a fraction of the equatorial radius
denes the attening , which is equal to 1/298. This results from deformation of
the Earth by the centrifugal force of its own rotation, which, expressed as a
fraction m of the gravitational force, is equal to 1/289. Both and m are less than
three thousandths of the main property, so 2, m2, and the product fm are of the
order of nine parts in a million and, along with higher-order combinations, are
negligible. Curtailing the expansion of small quantities at rst order helps keep
equations manageable without signicant loss of geophysical information.
In the following chapters much use will be made of such linear approximation. It simplies the form of mathematical functions and the usable part of the
series described above. For example, for small values of x or (x x0), the
following rst-order approximations may be used:
sin x x;
expx 1 x;
1 xp 1 px
cos x 1
ln1 x x
fx fx0 x x0
df
dx
(1:143)

xx0
32
1.11 Leibnizs rule

Assume that u(x) and v(x) are differentiable functions of x. The derivative of
their product is
d
dvx
dux
uxvx ux
vx
dx
dx
dx
(1:144)
If we dene the operator D = d/dx, we obtain a shorthand form of this equation:

Duv u Dv v Du
(1:145)
We can differentiate the product (uv) a second time by parts,

D2 uv DDuv u D2 v DuDv DvDu v D2 u
u D2 v 2DuDv v D2 u
(1:146)
and, continuing in this way,

D3 uv u D3 v 3Du D2 v 3 D2 u Dv v D3 u

D4 uv u D4 v 4Du D3 v 6 D2 u D2 v 4 D2 u Dv v D4 u
(1:147)
The coefcients in these equations are the binomial coefcients, as dened in
(1.139). Thus after n differentiations we have
Dn uv
n
X
k nk
n!
D u D v
k!n k!
k0
(1:148)
This relationship is known as Leibnizs rule, after Gottfried Wilhelm Leibniz

(16461716), who invented innitesimal calculus contemporaneously with
Isaac Newton (16421727); each evidently did so independently of the other.

Let r and R be the sides of a triangle that enclose an angle and let u be the side
opposite this angle (Fig. 1.11). The angle and sides are related by the cosine rule
u2 r2 R2 2rR cos
(1:149)
Inverting this expression and taking the square root gives

"
!
1 1
r
12
cos
u R
R
r
R
!2 #1=2
(1:150)
33
Fig. 1.11. Relationship of the sides r and R, which enclose an angle , and the side
u opposite the angle, as used in the denition of Legendre polynomials.
Now let h r=R and x = cos , giving

1=2 1
1 1
1 2xh h2
1 t1=2
u R
R
(1:151)
where t = 2xh h2. The equation can be expanded as a binomial series

1 3
1 3 5

2 2
2 2 2
1
1=2
2
1t
1 t
t
t3
2
12
123
!2
!3
t 13 t
135 t
1

2 12 2
123 2
!n
1 3 5 . . . 2n 1 t

(1:152)
1 2 3...n
2
The innite series of terms on the right-hand side of the equation can be written
1 t1=2
1
X
an t n
(1:153)
n0
The coefcient an is given by

an
1 3 5 . . . 2n 1
2n n!
(1:154)
Now, substitute the original expression for t,

1 2xh h2
1=2
1
X
n0
1

n X
an 2xh h2
an hn 2x hn
(1:155)
n0
This equation is an innite series in powers of h. The coefcient of each term in

the power series is a polynomial in x. Let the coefcient of hn be Pn(x). The
equation becomes
1

1=2 X
x; h 1 2xh h2
hn Pn x
n0
(1:156)
34
Equation (1.156) is known as the generating function for the polynomials Pn(x).
Using this result, and substituting h = r/R and x = cos , we nd that (1.151)
becomes
!n
1
1 1X
r
Pn cos
(1:157)
u R n0 R
The polynomials Pn(x) or Pn(cos ) are called Legendre polynomials, after the
French mathematician Adrien-Marie Legendre (17521833). The dening
equation (1.157) is called the reciprocal-distance formula. An alternative formulation is given in Box 1.4.

The Legendre polynomials satisfy an important second-order partial differential
equation, which is called the Legendre differential equation. To derive this
equation we will carry out a sequence of differentiations, starting with the
generating function in the form

1=2
1 2xh h2
(1:158)
Differentiating this function once with respect to h gives

3=2
x h 1 2xh h2
x h3
h
(1:159)
Differentiating twice with respect to x gives

3=2
h3
h 1 2xh h2
x
1
3
h x
2
3h2 5
3h2
2
x
x
1 2
5 2 2
3h x
(1:160)
(1:161)
Next we perform successive differentiations of the product (h) with respect to

h. The rst gives
h h
hx h3
h
h
(1:162)
35
Box 1.4. Alternative form of the reciprocal-distance formula

The sides and enclosed angle of the triangle in Fig. 1.11 are related by the
cosine rule
u2 r2 R2 2rR cos
(1)
Instead of taking R outside the brackets as in (1.150), we can move r outside

and write the expression for u as
"

2 #1=2
1 1
R
R
(2)
12
cos
u r
r
r
Following the same treatment as in Section 1.12, but now with h R=r and
x = cos , we get
1=2 1
1 1
1 t1=2
1 2xh h2
u r
r
(3)
where t = 2xh h2. The function (1 t)1/2 is expanded as a binomial series,

which again gives an innite series in h, in which the coefcient of hn is
Pn(x). The dening equation is as before:
1

1=2 X
hn Pn x
x; h 1 2xh h2
(4)
n0
On substituting h R=r and x = cos , we nd an alternative form for the

generating equation for the Legendre polynomials:
1 n
1 1X
R
Pn cos
(5)
u r n0 r
On repeating the differentiation, and taking (1.159) into account, we get

2
h
hx h32
x 2h3
h
h
h2
x h3 3hx h2 5 x 2h3
2x 3h3 3hx h2 5
Now substitute for 3, from (1.160), and 5, from (1.161), giving
(1:163)
36

2
1
1 2
2
2x

3h
3h
x

h
h2
h x
3h2 x2
(1:164)
Multiply throughout by h:
2

2
2
h 2 h 2x 3h
x h
h
x
x2

2

2x
3h
x h 2
x
x
x2
(1:165)
The second term on the right can be replaced as follows, again using (1.160) and
(1.161):
3h
1 2
3h2 3 2 2
x
x

2
1 2xh h2
x2
On substituting into (1.165) and gathering terms, we get

h

i 2
2
2x
h 2 h x h2 1 2xh h2
x2
x
h

2
2
h 2 h x2 1
2x
h
x2
x
(1:166)
(1:167)
(1:168)
The Legendre polynomials Pn(x) are dened in (1.156) as the coefcients of hn

in the expansion of as a power series. On multiplying both sides of (1.156) by
h, we get
h
1
X
hn1 Pn x
(1:169)
n0
We differentiate this expression twice and multiply by h to get a result that can
be inserted on the left-hand side of (1.168):
1
X
n 1hn Pn x
h
h
n0
(1:170)
1
X
2
h
nn 1hn Pn x
2
h
n0
(1:171)
Using (1.156), we can now eliminate and convert (1.168) into a second-order
differential equation involving the Legendre polynomials Pn(x),
37
Table 1.1. Some ordinary Legendre polynomials of low degree

n
Pn(x)
Pn(cos )
0
1
1
x

1 2
3x 1
2

1 3
5x 3x
2

1
35x4 30x2 3
8
1
cos

1
3 cos2 1
2

1
5 cos3 3 cos
2

1
35 cos4 30 cos2 3
8
2
3
4
1
X

hn
n0
1
X
n0
X
1
2
d 2 Pn x
dPn x
x 1
2x
nn 1hn Pn x (1:172)
dx2
dx
n0

hn

2
d 2 Pn x
dPn x
x 1
2x
x
0
nn 1Pn
dx2
dx
(1:173)
If this expression is true for every non-zero value of h, the quantity in curly
brackets must be zero, thus

d 2 Pn x
dPn x
2x
1 x2
nn 1Pn x 0
2
dx
dx
(1:174)
An alternative, simpler form for this equation is obtained by combining the rst
two terms:

d
2 dPn x
(1:175)
1x
nn 1Pn x 0
dx
dx
This is the Legendre differential equation. It has a family of solutions, each of
which is a polynomial corresponding to a particular value of n. The Legendre
polynomials provide solutions in potential analyses with spherical symmetry,
and have an important role in geophysical theory. Some Legendre polynomials
of low degree are listed in Table 1.1.
1.13.1 Orthogonality of the Legendre polynomials

Two vectors a and b are orthogonal if their scalar product is zero:
a b ax bx ay by az bz
3
X
i1
ai bi 0
(1:176)
38
By analogy, two functions of the same variable are said to be orthogonal if their
product, integrated over a particular range, is zero. For example, the trigonometric functions sin and cos are orthogonal for the range 0 2, because
Z2
Z2
sin cos d
2

1
1
sin2d cos2 0
2
4
0
(1:177)
The Legendre polynomials Pn(x) and Pl(x) are orthogonal over the range 1 x 1.
This can be established as follows. First, we write the Legendre equation in short
form, dropping the variable x for both Pn and Pl, and, for brevity, writing
d
Pn x P0n
dx
and
d2
Pn x P00n
dx2
(1:178)
Thus

1 x2 P00n 2xP0n nn 1Pn 0

1 x2 P00l 2xP0l ll 1Pl 0
(1:179)
(1:180)
Multiplying (1.179) by Pl and (1.180) by Pn gives

1 x2 Pl P00n 2xPl P0n nn 1Pl Pn 0

1 x2 Pn P00l 2xP0l Pn ll 1Pl Pn 0
(1:181)
(1:182)
Subtracting (1.182) from (1.181) gives

1 x2 Pl P00n Pn P00l 2x Pl P0n P0l Pn nn 1 ll 1Pl Pn
0
(1:183)
Note that

d
Pl P0n P0l Pn Pl P00n P0l P0n P0l P0n P00l Pn Pl P00n P00l Pn
dx
(1:184)
and

d

Pl P0n P0l Pn 2x Pl P0n P0l Pn
dx

d
1 x2 Pl P0n P0l Pn
dx
1 x2
(1:185)
39
Thus

d
1 x2 Pl P0n P0l Pn nn 1 ll 1Pl Pn 0
dx
(1:186)
Now integrate each term in this equation with respect to x over the range
1 x 1. We get

1x

Pl P0n
P0l Pn
1

x1
Z1
nn 1 ll 1
Pl Pn dx 0
x1
(1:187)
The rst term is zero on evaluation of (1 x ) at x = 1; thus the second term
must also be zero. For n l the condition for orthogonality of the Legendre
polynomials is
Z 1
Pn xPl xdx 0
(1:188)
2
x1
1.13.2 Normalization of the Legendre polynomials

A function is said to be normalized if the integral of the square of the function over
R 1
its range is equal to 1. Thus we must evaluate the integral x1 Pn x2 dx. We
begin by recalling the generating function for the Legendre polynomials given in
(1.156), which we rewrite for Pn(x) and Pl(x) individually:
1
X

1=2
hn Pn x 1 2xh h2
(1:189)

1=2
hl Pl x 1 2xh h2
(1:190)
n0
1
X
l0
Multiplying these equations together gives

1 X
1
X

1
hnl Pn xPl x 1 2xh h2
(1:191)
l0 n0
Now let l = n and integrate both sides with respect to x, taking into account
(1.188):
1
X
n0
Z1
h
Z1
2
Pn x dx
2n
x1
x1
dx
1 h2 2xh
(1:192)
40
The right-hand side of this equation is a standard integration that results in a

natural logarithm:
Z
dx
1
lna bx
(1:193)
a bx b
The right-hand side of (1.192) therefore leads to

1
dx
1
2
ln 1 h 2xh
1 h2 2xh 2h
x1
Z1
x1

1
ln 1 h2 2h ln 1 h2 2h
2h
(1:194)
and
Z1
x1

dx
1
1
2 1
2

ln
1

h
ln
1 h2 2xh h
2
2
1
ln1 h ln1 h
h
(1:195)
Using the MacLaurin series for the natural logarithms as in (1.135), we get
ln1 h h
h2 h3 h4
hn
1n1
2
3
4
n
ln1 h h
h2 h3 h4
hn
1n1

2
3
4
n
(1:196)
(1:197)
Subtracting the second equation from the rst gives

Z1
x1

1
dx
2
h3 h5
2X
h2n1

2
3
5
1 h 2xh h
h n0 2n 1
(1:198)
Inserting this result into (1.192) gives

1
X
n0
1
X
n0
0
h2n @
Z1
x1
Z1
h
Pn x2 dx
2n
x1
1
2X
h2n1
h n0 2n 1
1
2
A0
Pn x2 dx
2n 1
(1:199)
(1:200)
41
This is true for every value of h in the summation, so we obtain the normalizing
condition for the Legendre polynomials:
Z1
Pn x2 dx
x1
2
2n 1
(1:201)

1=2
It follows that n 12
Pn x is a normalized Legendre polynomial.

The Legendre polynomials can be easily computed with the aid of a formula
derived by a French mathematician, Olinde Rodrigues (17951851). First, we
dene the function

n
fx x2 1
(1:202)
Differentiating (x) once with respect to x gives
n

n1
df
d 2
x 1 2nx x2 1
dx dx
(1:203)
Multiplying the result by (x2 1) gives

2
d 2
n

n
x 1
x 1 2nx x2 1
dx
(1:204)
2
df
2nxf
x 1
dx
(1:205)
Now we use Leibnizs rule (1.144) to differentiate both sides of this equation
n + 1 times with respect to x. Writing D = d/dx as in Section 1.11,
Dn1 uv
n1
X
n 1! k n1k
v
D u D
k!n 1 k!
k0
(1:206)
On the left-hand side of (1.205) let u(x) = (x2 1) and v(x) = d/dx = D.
Applying Leibnizs rule, we note that after only three differentiations of (x2 1)
the result is zero and the series is curtailed.
On the right-hand side let u(x) = 2nx and v(x) = . Note that in this case the
series is curtailed after two differentiations.
Thus, using Leibnizs rule to differentiate each side of (1.205) n + 1 times,
we get
42

n 1n n
x2 1 Dn2 f 2xn 1Dn1 f 2
D f 2nx Dn1 f
12
2nn 1Dn f
(1:207)
On gathering terms and bringing all to the left-hand side, we have

2

x 1 Dn2 f 2x Dn1 f nn 1Dn f 0
(1:208)
Now we dene y(x) such that

yx Dn f
n
dn 2
x 1
n
dx
(1:209)
and we have

d 2y
x2 1
dx2
2x
dy
nn 1y 0
dx
(1:210)
On comparing with (1.174), we see that this is the Legendre equation. The
Legendre polynomials must therefore be proportional to y(x), so we can write
Pn x cn
n
dn 2
x 1
n
dx
(1:211)
The quantity cn is a calibration constant. To determine cn we rst write

n d n
dn 2
x

1
n x 1 n x 1 n
dxn
dx
(1:212)
then we apply Leibnizs rule to the product on the right-hand side of the
equation:
n
n X
dn 2
n!
dm
d nm
x

1
x 1n nm x 1n
n
m
dx
dx
m!n m! dx
m0
(1:213)
The successive differentiations of (x 1)n give

d
x 1n nx 1n1
dx
d2
x 1n nn 1x 1n2
dx2
d n1
x 1n nn 1n 2 . . . 321x 1 n!x 1
dxn1
dn
x 1n n!
dxn
(1:214)
43
Each differentiation in (1.214) is zero at x = 1, except the last one. Thus each
term in the sum in (1.213) is also zero except for the last one, for which m = n.
Substituting x = 1 gives
n

n
n
d 2
nd
n
x

1
x

1
2n n!
(1:215)
n
dxn
dx
x1
x1
Putting this result and the condition Pn(1) = 1 into (1.211) gives
n

n
d 2
x

1
cn 2n n! 1
Pn 1 cn
dxn
x1
(1:216)
where
cn
1
2n n!
(1:217)
Rodrigues formula for the Legendre polynomials is therefore

Pn x
1 dn
2n n! dxn
x2 1
n
(1:218)

Many physical properties of the Earth, such as its magnetic eld, are not
azimuthally symmetric about the rotation axis when examined in detail.
However, these properties can be described using mathematical functions that
are based upon the Legendre polynomials described in the preceding section. To
derive these functions, we start from the Legendre equation, (1.174), which can
be written in shorthand form as

1 x2 P 00n 2xP0n nn 1Pn 0
(1:219)
Now we differentiate this equation with respect to x:

d 00
d
d
P 2xP00n 2x P0n 2P0n nn 1 Pn 0
1 x2
dx n
dx
dx
(1:220)
On noting that we can equally write P 00n d=dxP0n and P0n d=dxPn , this
can be written alternatively as

1 x2
d 00
d
d
P 4x P0n nn 1 2 Pn 0
dx n
dx
dx
which can be written, for later comparison,
(1:221)
44

d 00
d
d
Pn 22x P0n nn 1 12 Pn 0
1 x2
dx
dx
dx
(1:222)
Next, we differentiate this expression again, observing the same rules and
gathering terms,

d 2 00
d 00
d 0
d2 0
d2
1x
P
2x
4x
P
Pn 0

4
1
2

P
P
n
n
n
n
dx2
dx2
dx2
dx
dx
2
(1:223)

d 2 00
d2 0
d2 0
d2
P

2x
P

4x
P

2

4

Pn 0
1 x2
dx2 n
dx2 n
dx2 n
dx2
(1:224)

1 x2
d 2 00
d2 0
d2
P

6x
P

6

Pn 0
n
n
dx2
dx2
dx2
(1:225)
which, as we did with (1.222), we can write in the form

1 x2
d 2 00
d2
d2
Pn 23x 2 P0n nn 1 23 2 Pn 0
2
dx
dx
dx
(1:226)
On following step-by-step in the same manner, we get after the third

differentiation

1 x2
d 3 00
d3 0
d3
P

2
4
x
P

3

Pn 0
n
n
dx3
dx3
dx3
(1:227)
Equations (1.222), (1.226), and (1.227) all have the same form. The higherorder differentiation is accompanied by systematically different constants. By
extension, differentiating (1.219) m times (where m n) yields the differential
equation

1 x2
d m 00
dm
dm
Pn 2m 1x m P0n nn 1 mm 1 m Pn 0
m
dx
dx
dx
(1:228)
Now let the mth-order differentiation of Pn be written as

dm
Qx
Pn x
m=2
dxm
1 x2
(1:229)
Substitution of this expression into (1.228) gives a new differential equation

involving Q(x). We need to determine both d m =dxm P0n and d m =dxm P00n , so
rst we differentiate (1.229) with respect to x:
dm 0
Q0
P

n
m=2
dxm
1 x2
!
m
Q
2x
m=21
2
1 x 2

m2=2

dm 0
P n 1 x2
1 x2 Q0 mxQ
m
dx
45
(1:230)
(1:231)
A further differentiation of (1.231) by parts gives

m2=2

d
1 x2
1 x2 Q0 mxQ
dx

m2=2 d

1 x2 Q0 mxQ
1 x2
dx

2 m2=21
m 2x 1 x
1 x2 Q0 mxQ

m2=2

1 x2 Q00 mxQ0 2xQ0 mQ
1 x2
d m 00
P
dxm n
(1:232)
n

d m 00
2 m2=2
P
1

x
1 x2 Q00 m 2xQ0 mQ m 2xQ0
n
dxm

mm 2x2 Q
(1:233)
1 x2

00
d m 00
mm 2x2 Q
2 m2=2
2
0
P
1

x
1

x
2mxQ
mQ
Q
dxm n
1 x2
(1:234)
Now we substitute (1.231) and (1.234) into (1.228). Unless the multiplier
(1 x2) (m + 2)/2 is always zero, Q must satisfy the following equation:

2

1 x2 Q00 2mx 1 x2 Q0 m 1 x2 Q mm 2x2 Q

2m 1x 1 x2 Q0 2mm 1x2 Q

nn 1 mm 1 1 x2 Q 0
(1:235)
The remainder of the evaluation consists of gathering and reducing terms; we

nally get

m2
Q0
1 x Q 2xQ nn 1
1 x2
2
00
(1:236)
46
The functions Q(x) involve two parameters, the degree n and order m, and are
written Pn,m(x). Thus

d2
d
m2
Pn;m x 0
P
1 x2
P

2x

n;m
n;m
dx2
1 x2
dx
(1:237)
This is the associated Legendre equation. The solutions Pn,m(x) or Pn,m(cos ),
where x = cos , are called associated Legendre polynomials, and are obtained
from the ordinary Legendre polynomials using the denition of Q in (1.229):

m=2 d m
Pn;m x 1 x2
Pn x
dxm
(1:238)
Substituting Rodrigues formula (1.218) for Pn(x) into this equation gives

m=2
n
1 x2
d nm 2
x 1
Pn;m x
n
nm
2 n!
dx
(1:239)
The highest power of x in the function (x2 1)n is x2n. After 2n differentiations
the result will be a constant, and a further differentiation will give zero.
Therefore n + m 2n, and possible values of m are limited to the range 0 m n.
1.15.1 Orthogonality of associated Legendre polynomials

For succinctness we again write Pn,m(x) as simply Pn,m. The dening equations
for the associated Legendre polynomials Pn,m and Pl,m are

0
m2
1 x Pn;m 2x Pn;m nn 1
Pn;m 0
1 x2

00

0
m2
2
1 x Pl;m 2x Pl;m ll 1
Pl;m 0
1 x2
2

00
(1:240)
(1:241)
As for the ordinary Legendre polynomials, we multiply (1.240) by Pl,m and

(1.241) by Pn,m:

00

0
m2
Pn;m Pl;m 0
1 x2 Pn;m Pl;m 2x Pn;m Pl;m nn 1
1 x2
(1:242)

47

00

0
m2
1 x2 Pl;m Pn;m 2x Pl;m Pn;m ll 1
Pn;m Pl;m 0
1 x2
(1:243)
On subtracting (1.243) from (1.242) we have

i
h
i

h
00

00
0

0
1 x2 Pn;m Pl;m Pl;m Pn;m 2x Pn;m Pl;m Pl;m Pn;m
nn 1 ll 1Pn;m Pl;m 0
(1:244)
Following the method used to establish the orthogonality of the ordinary

Legendre polynomials (Section 1.13.1), we can write this equation as
o

0

0
d n
1 x2
Pn;m Pl;m Pl;m Pn;m
nn 1 ll 1Pn;m Pl;m
dx
0
(1:245)
On integrating each term with respect to x over the range 1 x 1, we get
n
o1

0

0

1 x2
Pn;m Pl;m Pl;m Pn;m
x1
Z1
nn 1 ll 1
Pn;m Pl;m dx 0
(1:246)
x1
The rst term is zero on evaluation of (1 x2) at x = 1; thus the second term
must also be zero. Provided that n l, the condition of orthogonality of the
associated Legendre polynomials is
x1
Z
Pn;m xPl;m xdx 0
(1:247)
x1
1.15.2 Normalization of associated Legendre polynomials

Squaring the associated Legendre polynomials and integrating over 1 x 1
gives
x1
Z
x1

2
Pn;m x dx
2 n m!
2n 1 n m!
(1:248)
48
The squared functions do not integrate to 1, so they are not normalized. If each
polynomial is multiplied by a normalizing function, the integrated squared
polynomial can be made to equal a chosen value. Different conditions for this
apply in geodesy and geomagnetism.
The Legendre polynomials used in geodesy are fully normalized. They are
dened as follows:
Pm
n x

2n 1 n m! 1=2
Pn;m x
2
n m!
(1:249)
The Legendre polynomials used in geomagnetism are partially normalized (or

quasi-normalized). Schmidt in 1889 dened this method of normalization so that

Pm
n x
n m!
2
n m!
P0n x Pn;0 x;
1=2
Pn;m x;
m0
m 6 0
(1:250)
(1:251)
Integration of the squared Schmidt polynomials over the full range 1 x 1

gives the value 1 for m = 0 and 1/(2n + 1) for m > 0.
Some fully normalized Legendre polynomials and partially normalized
Schmidt polynomials are listed in Table 1.2.
Table 1.2. Some fully normalized associated Legendre polynomials and
partially normalized Schmidt polynomials of low degree and order
Pm
n cos ; Legendre,
fully normalized
Pm
n cos ; Schmidt,
partially normalized
cos
sin
1
3 cos2 1
2
3 sin cos
3 sin2
15 sin2 cos
15 sin3
cos
sin
1
3 cos2 1
2p
p3 sin cos
3
sin2
2
1
cos 5 cos2 3
2p
6
sin 5 cos2 1
4
p
15
15 sin2 cos
p2
10
sin3
4
1
1
0
1
1
cos 5 cos2 3
2
3
sin 5 cos2 1
2
49

Several geophysical potential elds for example, gravitation and geomagnetism satisfy the Laplace equation. Spherical polar coordinates are best suited
for describing a global geophysical potential. The potential can vary with
distance r from the Earths center and with polar angular distance and
azimuth (equivalent to co-latitude and longitude in geographic terms)
on any concentric spherical surface. The solution of Laplaces equation in
spherical polar coordinates for a potential U may be written (see Section
2.4.5, (2.104))
U

1 X
n
X
m
Bn
m
An rn n1 am
n cosm bn sinm Pn cos (1:252)
r
n0 m0
m
Here An, Bn, am
n , and bn are constants that apply to a particular situation. On the
surface of the Earth, or an arbitrary sphere, the radial part of the potential of a
point source at the center of the sphere has a constant value and the variation
over the surface of the sphere is described by the functions in and . We are
primarily interested in solutions outside the Earth, for which An is zero. Also we
can set the constant Bn equal to Rn+1, where R is the Earths mean radius. The
potential is then given by
1 X
n n1
X
m
R
m
am
U
n cosm bn sinm Pn cos
r
n0 m0
(1:253)
m
Let the spherical harmonic functions Cm
n ; and Sn ; be dened as
m
Cm
n ; cosm Pn cos
m
Sm
n ; sinm Pn cos
(1:254)
The variation of the potential over the surface of a sphere may be described by
these functions, or a more general spherical harmonic function Ym
n ; that
combines the sine and cosine variations:

m
Ym
n ; Pn cos
cosm
sinm

(1:255)
Like their constituent parts the sine, cosine, and associated Legendre
functions spherical harmonic functions are orthogonal and can be
normalized.
50
1.16.1 Normalization of spherical harmonic functions

m
Normalization of the functions Cm
n ; and Sn ; requires integrating the
squared value of each function over the surface of a unit sphere. The element of
surface area on a unit sphere is d = sin d d (Box 1.3) and the limits of
integration are 0 and 0 2. The integral is
ZZ
2
Cm
n ;
Z2
Cm
n ;
2
sin d d
0 0
Z2
cosmPm
n cos
2
sin d d (1:256)
0 0
Let x = cos in the associated Legendre polynomial, so that dx = sin d and

the limits of integration are 1 x 1. The integration becomes
8
Z1 >
< Z2
x1
>
:
0
9
>
Z1
=
2
m 2
2
m
cos md Pn x dx
Pn x dx
>
;
(1:257)
x1
Normalization of the associated Legendre polynomials gives the result in

(1.248), thus
ZZ
Cm
n ;
2

d

2
n m!
2n 1 n m!
(1:258)
The normalization of the function Sm

n ; by this method delivers the same
result.
Spherical harmonic functions make it possible to express the variation of a
physical property (e.g., gravity anomalies, g(, )) on the surface of the Earth as
an innite series, such as
g;
1 X
n
X
m
m m
am
n Cn ; bn Sn ;
(1:259)
n0 m0
m
The coefcients am
n and bn may be obtained by multiplying the function g;
m
m
by Cn ; or Sn ; , respectively, and integrating the product over the
surface of the unit sphere. The normalization properties give
51

ZZ
2n 1 n m!
g; Cm
n ; d
2
n m!
S

ZZ
2n
n

m
!
bm
g; Sm
n
n ; d
2
n m!
am
n
(1:260)
1.16.2 Zonal, sectorial, and tesseral spherical harmonics

The spherical harmonic functions Ym
n ; have geometries that allow graphic
representation of a potential on the surface of a sphere. Deviations of the potential
from a constant value form alternating regions in which the potential is larger or
smaller than a uniform value. Where the potential surface intersects the spherical
surface a nodal line is formed. The appearance of any Ym
n ; is determined by
the distribution of its nodal lines. These occur where Ym

n ; = 0. To simplify the
discussion we will associate a constant value of the polar angle with a circle of
latitude, and a constant value of the azimuthal angle with a circle of longitude.
The denition of the associated Legendre polynomials in (1.239) shows that
the equation Pm
n x = 0 has n m roots, apart from the trivial solution x = 1.
The variation of the spherical harmonic Ym
n ; with latitude thus has n m
nodal lines, each a circle of latitude, between the two poles. If additionally m =
0, the potential on the sphere varies only with latitude and there are n nodal lines
separating zones in which the potential is greater or less than the uniform value.
An example of a zonal spherical harmonic is Y02 ; , shown in Fig. 1.12(a).
The solution of Laplaces equation (1.253) shows that the variation in
potential around any circle of latitude is described by the function
m
am
n cosm bn sinm
(1:261)
There are 2m nodal lines where () = 0, corresponding to 2m meridians of

longitude, or m great circles. In the special case in which n = m, there are no
(a) zonal, Y20
(b) sectorial, Y55
(c) tesseral, Y54
Fig. 1.12. Appearance of (a) zonal, (b) sectorial, and (c) tesseral spherical
harmonics, projected on a meridian plane of the reference sphere.
52
nodal lines of latitude and the longitudinal lines separate sectors in which the
potential is greater or less than the uniform value. An example of a sectorial
spherical harmonic is Y55 ; , shown in Fig. 1.12(b).
In the general case (m 0, n m) the potential varies with both latitude and
longitude. There are n m nodal lines of latitude and m nodal great circles (2m
meridians) of longitude. The appearance of the spherical harmonic resembles a
patchwork of alternating regions in which the potential is greater or less than the
uniform value. An example of a tesseral spherical harmonic is Y45 ; , which
is shown in Fig. 1.12(c).
1.17 Fourier series, Fourier integrals, and

Fourier transforms
1.17.1 Fourier series
Analogously to the representation of a continuous function by a power series
(Section 1.10), it is possible to represent a periodic function by an innite sum
of terms consisting of the sines and cosines of harmonics of a fundamental
frequency. Consider a periodic function (t) with period that is dened in the
interval 0 t , so that (a) (t) is nite within the interval; (b) (t) is periodic
outside the interval, i.e., (t + ) = (t); and (c) (t) is single-valued in the
interval except at a nite number of points, and is continuous between these
points. Conditions (a)(c) are known as the Dirichlet conditions. If they are
satised, (t) can be represented as
ft
1
a0 X
an cosnt bn sinnt
2
n1
(1:262)
where = 2/ and the factor 12 in the rst term is included for reasons of
symmetry. This representation of (t) is known as a Fourier series. The
orthogonal properties of sine and cosine functions allow us to nd the coefcients an and bn of the nth term in the series by multiplying (1.262) by sin(nt)
or cos(nt) and integrating over a full period:
2
an
Z=2
ftcosntdt
t=2
bn
(1:263)
Z=2
ftsinntdt
t=2
1.17 Fourier series, integrals, and transforms
53
Instead of using trigonometric functions, we can replace the sine and

cosine terms with complex exponentials using the denitions in (1.7), i.e., we
write
expint expint
2
expint expint
sinnt
2i
cosnt
(1:264)
Using these relationships in (1.262) yields

bn
expint expint
2
2i
n0

1
1
X an ibn
X an ibn
expint
expint
(1:265)
2
2
n0
n0
ft
1
X
an
expint expint
The summation indices are dummy variables, so in the second sum we can
replace n by n, and extend the limits of the sum to n = ; thus
ft
1
X
an ibn
n0
1
X

0
X
an ibn
expint
expint
2
n1
an an ibn bn

expint
2
n1
(1:266)
If we dene cn as the complex number

cn
an an ibn bn

2
(1:267)
the Fourier series (1.262) can be written in complex exponential form as

ft
1
X
cn expint
(1:268)
n1
In this case the harmonic coefcients cn are given by

1
cn
Z=2
ftexpintdt
t=2
(1:269)
54
1.17.2 Fourier integrals and Fourier transforms

A Fourier series represents the periodic behavior of a physical property as an
innite set of discrete frequencies. The theory can be extended to represent a
function (t) that is not periodic and is made up of a continuous spectrum of
frequencies, provided that the function satises the Dirichlet conditions specied above and that it has a nite energy:
Z1
j ftj2 dt51
(1:270)
t1
The innite sum in (1.268) is replaced by a Fourier integral and the complex
coefcients cn are replaced by an amplitude function g():
Z1
ft
gexpitd
(1:271)
1
where g() is a continuous function, obtained from the equation

1
g
2
Z1
ftexpitdt
(1:272)
t1
The transition from Fourier series to Fourier integral is explained in Box 1.5.
The function g() is called the forward Fourier transform of (t), and (t) is
called the inverse Fourier transform of g(). Fourier transforms constitute a
powerful mathematical tool for transforming a function (t) that is known in the
time domain into a new function g() in the frequency domain.
1.17.3 Fourier sine and cosine transforms

A simple but important characteristic of a function is whether it is even or odd.
An even function has the same value for both positive and negative values of its
argument, i.e., (t) = (t). The cosine of an angle is an example of an even
function. The integral of an even function over a symmetric interval about the
origin is equal to twice the integral of the function over the positive argument.
The sign of an odd function changes with that of the argument, i.e., (t) = (t).
For example, the sine of an angle is an odd function. The integral of an odd
function over a symmetric interval about the origin is zero. The product of two
odd functions or two even functions is an even function; the product of an odd
function and an even function is an odd function.
55
Box 1.5. Transition from Fourier series to Fourier integral

The complex exponential Fourier series for a function (t) is
1
X
ft
cn expint
(1)
n1
where the complex coefcients cn are given by

Z=2
1
cn
ftexpintdt
(2)
t=2
In these expressions = 2/ is the fundamental frequency and is the

fundamental period. From one value of n to the next, the harmonic frequency
changes by = 2/, so the factor preceding the second equation can be
replaced by 1/ = /(2). To avoid confusion when we insert (2) into (1), we
change the dummy variable of the integration to u, giving
Z=2
cn
2
fuexpinudu
(3)
u=2
After insertion, (1) becomes

0
1
Z=2
1
X
B
C
ft
fuexpinuduA expint
@
2
n1
u=2
1
X
B
@
2
n1
Z=2
1
C
fuexpint uduA
(4)
u=2
We now dene the function within the integral as

Z=2
fuexpint udu
F
u=2
The initial Fourier series becomes
(5)
56
ft
1
1 X
F
2 1
(6)
We now let the incremental frequency become very small, tending in the
limit to zero; this is equivalent to letting the period become innite. The
index n is dropped because is now a continuous variable; the discrete sum
becomes an integral and the function f(t) is
Z1
1
f t
2
Fd
(7)
1
while the function F() from (5) becomes

Z1
fuexpit udu
(8)
u1
On inserting F() into (7) we get

2 1
3
Z
Z1
1
4
ft
fuexpit udu5d
2
1
u1
2 1
3
Z
Z1
1
4
fuexpiudu5 expitd
2
1
(9)
u1
The quantity in square brackets, on changing the variable from u back to t, is

1
g
2
Z1
ftexpitdt
(10)
t1
and the original expression can now be written

Z1
ft
gexpitd
(11)
1
The equivalence of these two equations is known as the Fourier integral

theorem.
57
Fourier series that represent odd or even functions consist of sums of sines or
cosines, respectively. In the same way, there are sine and cosine Fourier integrals
that represent odd and even functions, respectively. Suppose that the function (t)
is even, and let us replace the complex exponential in (1.272) using (1.5):
1
g
2
Z1
ftcost i sintdt
(1:273)
t1
The sine function is odd, so, if (t) is even, the product (t)sin(t) is odd, and the
integral of the second term is zero. The product (t)cos(t) is even, and we can
convert the limits of integration to the positive interval:
Z1
1
g
2
ftcostdt
t1
Z1
ftcostdt
(1:274)
t0
Thus, if (t) is even, then g() is also even. Similarly, one nds that, if (t) is
odd, g() is also odd.
Now we expand the exponential in (1.271) and apply the same conditions of
evenness and oddness to the products:
Z1
ft
gcost i sintd
1
Z1
gcostd
(1:275)
If we were to substitute (1.275) back into (1.274), the integration would be

preceded by a constant 2/, the product of the two constants in these equations.
Equations (1.275) and (1.274) form a Fourier-transform pair, and it does not
matter how the factor 2/ is divided between them. We will associate it here
entirely with the second equation, so that we have the pair of equations
Z1
gcostd
ft
0
2
g
(1:276)
Z1
ftcostdt
t0
58
The even functions (t) and g() are Fourier cosine transforms of each other.
A similar treatment for a function (t) that is odd leads to a similar pair of
equations in which the Fourier transform g() is also odd and
Z1
ft
gsintd
0
2
g
(1:277)
Z1
ftsintdt
t0
The odd functions (t) and g() are Fourier sine transforms of each other.
further reading
Boas, M. L. (2006). Mathematical Methods in the Physical Sciences, 3rd edn. Hoboken,
NJ: Wiley, 839 pp.
James, J. F. (2004). A Students Guide to Fourier transforms, 2nd edn. Cambridge:
Cambridge University Press, 135 pp.
2
Gravitation
2.1 Gravitational acceleration and potential

The Universal Law of Gravitation deduced by Isaac Newton in 1687 describes
the force of gravitational attraction between two point masses m and M separated by a distance r. Let a spherical coordinate system (r, , ) be centered on
the point mass M. The force of attraction F exerted on the point mass m acts
radially inwards towards M, and can be written
F G
mM
er
r2
(2:1)
In this expression, G is the gravitational constant (6.674 21 1011 m3 kg1 s2),

er is the unit radial vector in the direction of increasing r, and the negative sign
indicates that the force acts inwardly, towards the attracting mass. The gravitational acceleration aG at distance r is the force on a unit mass at that point:
aG G
M
er
r2
(2:2)
The acceleration aG may also be written as the negative gradient of a gravitational potential UG
aG rUG
(2:3)
The gravitational acceleration for a point mass is radial, thus the potential
gradient is given by
UG
M
G 2
r
r
UG G
59
M
r
(2:4)
(2:5)
60
Gravitation
In Newtons time the gravitational constant could not be veried in a laboratory

experiment. The attraction between heavy masses of suitable dimensions is weak
and the effects of friction and air resistance relatively large, so the rst successful
measurement of the gravitational constant by Lord Cavendish was not made until
more than a century later, in 1798. However, Newton was able to conrm the
validity of the inverse-square law of gravitation in 1687 by using existing
astronomic observations of the motions of the planets in the solar system.
These had been summarized in three important laws by Johannes Kepler in
1609 and 1619. The small sizes of the planets and the Sun, compared with the
immense distances between them, enabled Newton to consider these as point
masses and this allowed him to verify the inverse-square law of gravitation.

Johannes Kepler (15711630), a German mathematician and scientist, formulated his laws on the basis of detailed observations of planetary positions by
Tycho Brahe (15461601), a Danish astronomer. The observations were made
in the late sixteenth century, without the aid of a telescope. Kepler found that the
observations were consistent with the following three laws ( Fig. 2.1).
1. The orbit of each planet is an ellipse with the Sun at one focus.
2. The radius from the Sun to a planet sweeps over equal areas in equal intervals
of time.
P*
Q*
aphelion
S
b
P(r, )
perihelion
Fig. 2.1. Illustration of Keplers laws of planetary motion. The orbit of each planet
is an ellipse with the Sun at its focus (S); a, b, and p are the semi-major axis, semiminor axis, and semi-latus rectum, respectively. The area swept by the radius to a
planet in a given time is constant (i.e., area SPQ equals area SP*Q*); the square of
the period is proportional to the cube of the semi-major axis. After Lowrie ( 2007).
61
3. The square of the period is proportional to the cube of the semi-major axis of
the orbit.
The fundamental assumption is that the planets move under the inuence of a
central, i.e., radially directed force. For a planet of mass m at distance r from the
Sun the force F can be written
Fm
d 2r
frer
dt2
(2:6)
The angular momentum h of the planet about the Sun is

hrm
dr
dt
(2:7)
Differentiating with respect to time, the rate of change of angular momentum is

dh
d
dr
dr dr
d 2r
m
r
m

m r 2
dt
dt
dt
dt dt
dt
(2:8)
The rst term on the right-hand side is zero, because the vector product of a
vector with itself (or with a vector parallel to itself) is zero. Thus
dh
d 2r
rm 2
dt
dt
(2:9)
On substituting from ( 2.6) and applying the same condition, we have

dh
r frer frr er 0
dt
(2:10)
This equation means that h is a constant vector; the angular momentum of the
system is conserved. On taking the scalar product of h and r, we obtain

dr
rh r r m
(2:11)
dt
Rotating the sequence of the vectors in the triple product gives
rh m
dr
r r 0
dt
(2:12)
This result establishes that the vector r describing the position of a planet is
always perpendicular to its constant angular momentum vector h and therefore
denes a plane. Every planetary orbit is therefore a plane that passes through the
Sun. The orbit of the Earth denes the ecliptic plane.
62
Gravitation
2.2.1 Keplers Second Law

Let the position of a planet in its orbit be described by polar coordinates (r, )
with respect to the Sun. The coordinates are dened so that the angle is zero at
the closest approach of the planet to the Sun (perihelion). The angular momentum at an arbitrary point of the orbit has magnitude
h mr2
d
dt
(2:13)
In a short interval of time t the radius vector from the Sun to the planet
moves through a small angle and denes a small triangle. The area A of
the triangle is
1
DA r2 D
2
The rate of change of the area swept over by the radius vector is

dA
DA
1 2 D
lim
lim
r
Dt!0 Dt
Dt!0 2
dt
Dt
(2:14)
(2:15)
dA 1 2 d
r
dt
2 dt
(2:16)
dA
h
dt
2m
(2:17)
On inserting from ( 2.13) we get
Thus the area swept over by the radius vector in a given time is constant. This is
Keplers Second Law of planetary motion.
2.2.2 Keplers First Law

If just the gravitational attraction of the Sun acts on the planet (i.e., we ignore the
interactions between the planets), the total energy E of the planet is constant.
The total energy E is composed of the planets orbital kinetic energy and its
potential energy in the Suns gravitational eld:
2
2
1
dr
1
d
S
mr2
Gm E
m
2
dt
2
dt
r
(2:18)
The rst term here is the planets linear (radial) kinetic energy, the second term
is its rotational kinetic energy (with mr2 being the planets moment of inertia
63
about the Sun), and the third term is the gravitational potential energy. On
writing
dr dr d
dt d dt
(2:19)
and rearranging terms we get

2 2
2
dr
d
S
E
2 d
r
2G 2
d
dt
dt
r
m
(2:20)
Now, to simplify later steps, we make a change of variables, writing

u
1
r
(2:21)
Then

dr
d 1
1 du
2 du
2
r
d d u
u d
d
Substituting from ( 2.22) into ( 2.20) gives

2
d 2 du 2
d
S
E
r2
r2
2G 2
dt
d
dt
r
m
(2:22)
(2:23)
With the result of ( 2.13) we have

d h
r2
dt
m

d
1 h
h
r
u
dt
r m
m
On replacing these expressions, ( 2.23) becomes
2 2
2
h
du
E
2 h
u
2uGS 2
m
d
m
m
du
d
2
u2 2uGS
m2
Em
2 2
2
h
h
(2:24)
(2:25)
(2:26)
The rest of the evaluation is straightforward, if painstaking. First we add a

constant to each side,
du
d
2
u2 2uGS

2

2
m2
m2
Em
m2
GS
GS
h2
h2
h2
h2
(2:27)
64
Gravitation
du
d
2
m2
u GS 2
h
2

2
Em
m2
2 2 GS 2
h
h
(2:28)
Next, we move the second term to the right-hand side of the equation, giving

du
d
du
d
2

2
2
Em
m2
m2
2 2 GS 2 u GS 2
h
h
h
2

2

2
m2
2Eh2
m2
GS 2
1 2 2 3 u GS 2
h
GS m
h
(2:29)
(2:30)
Now, we dene some combinations of these terms, as follows:

u0 GS
m2
h2
(2:31)
2Eh2
G2 S2 m3
(2:32)
e2 1
Using these dened terms, ( 2.30) simplies to a more manageable form:

2
du
u20 e2 u u0 2
(2:33)
d
q
du
u20 e2 u u0 2
(2:34)
d
The solution of this equation, which can be tested by substitution, is
u u0 1 e cos
(2:35)
The angle is dened to be zero at perihelion. The negative square root in ( 2.34)
is chosen because, as increases, r increases and u must decrease. Let
p
1
h2
u0 GSm2
(2:36)
p
1 e cos
(2:37)
This is the polar equation of an ellipse referred to its focus, and is the proof of
Keplers First Law of planetary motion. The quantity e is the eccentricity of the
ellipse, while p is the semi-latus rectum of the ellipse, which is half the length of
a chord passing through the focus and parallel to the minor axis ( Fig. 2.1).
These equations show that three types of trajectory around the Sun are
possible, depending on the value of the total energy E in ( 2.18). If the kinetic
65
energy is greater than the potential energy, the value of E in ( 2.32) is positive,
and e is greater than 1; the path of the object is a hyperbola. If the kinetic energy
and potential energy are equal, the total energy is zero and e is exactly 1; the path
is a parabola. In each of these two cases the object can escape to innity, and the
paths are called escape trajectories. If the kinetic energy is less than the potential
energy, the total energy E is negative and the eccentricity is less than 1. In this
case (corresponding to a planet or asteroid) the object follows an elliptical orbit
around the Sun.
2.2.3 Keplers Third Law

It is convenient to describe the elliptical orbit in Cartesian coordinates (x, y),
centered on the mid-point of the ellipse, instead of on the Sun. Dene the x-axis
parallel to the semi-major axis a of the ellipse and the y-axis parallel to the semiminor axis b. The equation of the ellipse in Fig. 2.1 is
x2 y2
1
a2 b2
(2:38)
The semi-minor axis is related to the semi-major axis by the eccentricity e, so

that

b2 a2 1 e2
(2:39)
The distance of the focus of the ellipse from its center is by denition ae. The
length p of the semi-latus rectum is the value of y for a chord through the focus.
On setting y = p and x = ae in ( 2.38), we obtain
p2
ae2
1 2 1 e2
2
b
a

2
p 2 a 2 1 e2
(2:40)
(2:41)
Now consider the application of Keplers Second Law to an entire circuit of the
elliptical orbit. The area of the ellipse is ab, and the period of the orbit is T, so
dA ab
dt
T
(2:42)
h 2ab
m
T
(2:43)
Using (2.17),
66
Gravitation
From ( 2.36) and ( 2.43) we get the value of the semi-latus rectum,

1 h 2
1 2ab 2
p
GS m
GS
T
(2:44)
Substituting from ( 2.41) gives

4 2 a2 b2 4 2 a4

a 1 e2
1 e2
GST 2
GST 2
(2:45)
After simplifying, we nally get

T 2 4 2
a3
GS
(2:46)
The quantities on the right-hand side are constant, so the square of the period is
proportional to the cube of the semi-major axis, which is Keplers Third Law.
2.3 Gravitational acceleration and the potential

of a solid sphere
The gravitational potential and acceleration outside and inside a solid sphere
may be calculated from the Poisson and Laplace equations, respectively.
2.3.1 Outside a solid sphere, using Laplaces equation

Outside a solid sphere the gravitational potential UG satises Laplaces equation
( Section 1.9). If the density is uniform, the potential does not vary with the
polar angle or azimuth . Under these conditions, Laplaces equation in
spherical polar coordinates ( 2.67) reduces to

2 UG
0
(2:47)
r
r
r
This implies that the bracketed quantity that we are differentiating must be a
constant, C,
r2
UG
C
r
(2:48)
UG C
2
r
r
(2:49)
The gravitational acceleration outside the sphere is therefore
2.3 The potential of a solid sphere
67

UG
C
2 er
r
r
(2:50)
aG r4R
At its surface the gravitational acceleration has the value

UG
C
aG R
2 er
r
R
(2:51)
The boundary condition at the surface of the sphere is that the accelerations
determined outside and inside the sphere must be equal there. We use this to
derive the value of the constant C. On comparing ( 2.51) and ( 2.60) we have
C GM
(2:52)
On inserting for C in ( 2.50), the gravitational acceleration outside the sphere is

aG r4R G
M
er
r2
(2:53)
The gravitational potential outside the solid sphere is obtained by integrating (

2.53) with respect to the radius. This gives
UG r4R G
M
r
(2:54)
2.3.2 Inside a solid sphere, using Poissons equation

Inside a solid sphere with radius R and uniform density the gravitational
potential UG satises Poissons equation ( Section 1.8). Symmetry again
requires the use of spherical polar coordinates, and, because the density is
uniform, there is no variation of potential with the polar angle or azimuth .
Poissons equation in spherical polar coordinates reduces to
1 2 UG
4G
r
r
r2 r
(2:55)
On multiplying by r2 and integrating with respect to r, we get

2 UG
4Gr2
r
r
r
r2
UG 4
Gr3 C1
r
3
(2:56)
(2:57)
This equation has to be valid at the center of the sphere where r = 0, so the
constant C1 = 0 and
68
Gravitation
UG 4
Gr
r
3
UG
aG r5R
r
(2:58)

4
Gr er
3
(2:59)
This shows that the gravitational acceleration inside a homogeneous solid

sphere is proportional to the distance from its center. At the surface of the
sphere, r = R, and the gravitational acceleration is

4
GM
aG R GR er 2 er
3
R
(2:60)
where the mass M of the sphere is

4
M R3
3
(2:61)
To obtain the potential inside the solid sphere, we must integrate ( 2.58). This
gives
2
UG Gr2 C2
3
(2:62)
The constant of integration C2 is obtained by noting that the potential must be

continuous at the surface of the sphere. Otherwise a discontinuity would exist
and the potential gradient (and force) would be innite. Equating ( 2.54) and (
2.62) at r = R gives
2
GM
4
GR2 C2
GR2
3
R
3
(2:63)
C2 2GR2
(2:64)
The gravitational potential inside the uniform solid sphere is therefore given by
2
UG Gr2 2GR2
3
(2:65)

2
UG G r2 3R2
3
(2:66)
A schematic graph of the variation of the gravitational potential inside and

outside a solid sphere is shown in Fig. 2.2.
69
r /R
0
1
inside
sphere
outside
sphere
0.5
1
U G (R)
1.5
U G(r)
r=R
Fig. 2.2. Variation with radial distance r of the gravitational potential inside and
outside a solid sphere of radius R. The potential of the surface of the sphere is
UG(R).
2.4 Laplaces equation in spherical polar coordinates

In the above examples the sphere was assumed to have uniform density so that
only the radial term in Laplaces equation had to be solved. This is also the case
when density varies only with radius. In the Earth, however, lateral variations of
the density distribution occur, and the gravitational potential UG is then a
solution of the full Laplace equation
1 2 UG
1
UG
1 2 UG
0
r
sin
r2 r
r2 sin
r2 sin2 2
(2:67)
This equation is solved using the method of separation of variables. This is a

valuable mathematical technique, which allows the variables in a partial differential equation to be separated so that only terms in one variable are on one side
of the equation and terms in other variables are on the opposite side. A trial
solution for UG is
UG r; ; <r
(2:68)
Here , , and are all functions of a single variable only, namely r, , and ,
respectively. Multiplying ( 2.67) by r2 and inserting ( 2.68) for UG gives
2 < <
< 2
0
r
sin
2
r r sin
sin 2
(2:69)
70
Gravitation
On dividing throughout by we get

1 2 <
1
1 2
0
r
sin
< r r sin
sin2 2
(2:70)
Next we isolate the radial terms on the left-hand side of the equation, so that
1 2 <
1
1 2
r

sin

< r r
sin
sin2 2
(2:71)
The left-hand side of the equation is a function of r only, while the right-hand
side does not depend on r. Whatever the value of the left-hand side, the righthand side must always equal it. But r, , and are independent variables, so the
identity can exist only if the opposite sides of the equation are equal to the same
constant. Let this constant be K. For the opposite sides of ( 2.71) we get
1 2 <
r
K
< r r
(2:72)
1 2
K
sin

sin
sin2 2
(2:73)
If we multiply the last equation throughout by sin2, the variables can again be
separated:
sin
1 2
sin
K sin2

2
(2:74)
The variables on the opposite sides of ( 2.74) are independent, so each side must
be equal to the same constant, which we write temporarily as K2. Thus we can
replace equation ( 2.70) with three equations, consisting of ( 2.72) and the
following two:
sin
sin
K sin2 K2

(2:75)
1 2
K2
2
(2:76)
2.4.1 Azimuthal (longitudinal) solution

The constant K2 may be chosen to suit the conditions governing the gravitational potential. The function () describes the variation of the potential with
azimuth (longitude, in geographic terms). If we measure azimuthal uctuations
71
of the potential around a circle of constant polar angle (geographic co-latitude),

the same potential must result after a full circuit. This requires that the solution
for () be periodic, and that condition will be fullled if we let the constant
equal m2. For the right-hand side of ( 2.74) we get
1 2
m2
2
2
m2 0
2
(2:77)
(2:78)
This is the equation of simple harmonic motion, which has periodic solutions of
the form
am cosm bm sinm
(2:79)
2.4.2 Polar (latitudinal) solution for rotational symmetry

We rst consider solutions of Laplaces equation that have rotational symmetry
about the reference axis, which in the Earth is its axis of rotation. Since there is
no azimuthal variation of the potential in this situation, we can set m = 0. The
variation of potential with angle is described by
sin

sin
K sin2 0
sin
K 0
sin

1
1
sin2
K 0
sin
sin
(2:80)
(2:81)
(2:82)
If we write x = cos , then
1
x sin
(2:83)

2
1x
K 0
x
x
(2:84)
and (2.82) becomes
Comparison with (1.175) shows that this is equivalent to the Legendre differential equation, with n(n + 1) = K. If we make this choice of constant, we ensure
72
Gravitation
that the Laplace equation will have periodic solutions in polar angle (co-latitude),
namely the Legendre polynomials. The equation is

2 Pn x
1x
nn 1Pn x 0
(2:85)
x
x
and its solutions are
n Pn x Pn cos
(2:86)
2.4.3 Radial solution

With K = n(n + 1), the equation for the radial variation of the gravitational
potential becomes
1 2 <
r
nn 1
< r r
(2:87)
There will be a radial solution for each value of n, so we write it n, where

2 <n
nn 1<n 0
r
r r
(2:88)
Let n(r) be represented by the power series

<n r
1
X
ap rp
(2:89)
p0
Differentiating with respect to r gives

1
< X
pap rp1
r
p0
(2:90)
Multiplying by r2 and differentiating the product

1
< X
pap rp1
r
p0
(2:91)
1
2 < X
r
pp 1ap rp
r r
p0
(2:92)
r2
Inserting this result into ( 2.88) gives

1
X
pp 1ap rp nn 1
1
X
73
ap rp 0
(2:93)
ap rp pp 1 nn 1 0
(2:94)
p0
p0
1
X
p0
For this result to be true for any value of r, the expression in square brackets
must equal zero,
pp 1 nn 1 0
(2:95)
p2 p nn 1 0
(2:96)
That is,
Thus p can have the values p = n or p = (n + 1) and the radial variation of the
potential is given by
<n r An rn
Bn
rn1
(2:97)
where An and Bn are constants determined by the boundary conditions.
2.4.4 Solution of Laplaces equation for rotational symmetry

Combining the radial and polar variations, the gravitational potential for a mass
distribution that has rotational symmetry about an axis is

1
X
Bn
An rn n1 Pn cos
(2:98)
UG
r
n0
2.4.5 General solution of Laplaces equation

In the general case the potential may vary azimuthally about the reference axis.
The constant m is no longer zero and instead of ( 2.80) we have
sin
sin
K sin2 m2

sin

sin
K sin2 m2 0
(2:99)
(2:100)
As in the case with rotational symmetry, we substitute x = cos and obtain
74
Gravitation

m2
1 x2
K1
0
1 x2
x
x
(2:101)
If we again write n(n + 1) for the constant K,

1 x2

2
m2
0

2x
nn
1

1 x2
x2
x
(2:102)
This equation is equivalent to the associated Legendre equation (1.237), and the
functions are the associated Legendre polynomials:
m
Pm
n x Pn cos
(2:103)
The general solution of Laplaces equation for the gravitational potential in

spherical polar coordinates is obtained by combining the results of ( 2.79), (
2.97), and ( 2.103):

1 X
n
X
m
Bn m
n
UG
An r n1 an cosm bm
n sinm Pn cos
r
n0 m0
(2:104)
2.5 MacCullaghs formula for the gravitational potential

The yielding of the Earth to the deforming forces of its own rotation results in a
shape that is symmetric about the rotation axis and slightly attened at the poles.
The gure is classied as an ellipsoid of revolution, and, since it deviates only
slightly from a sphere, it may be called a spheroid. The equation and geometric
properties of a spheroid are summarized in Box 2.1.
The attening of the Earth is dened as the difference between the equatorial
radius and the polar radius, expressed as a fraction of the equatorial radius:
f
ac
a
(2:105)
The value of is known accurately from satellite geodesy ( Table 2.1) to be

= 1/298.252.
Let the Earth be represented by a spheroid with attening , and let the origin
of a Cartesian coordinate system (x, y, z) be at the center of mass of the spheroid
( Fig. 2.3). UG is the gravitational potential at an external point P at distance r
from the center of the Earth. For a continuous distribution of mass in a body we
can employ integral calculus to calculate its mass, moments of inertia, or the
location of its center of mass. However, it is instructive to regard the Earth as a
2.5 MacCullaghs formula
75
Box 2.1. The ellipsoid and spheroid

Let an ellipsoid with three unequal principal axes be referred to a set of
orthogonal Cartesian axes (x, y, z) such that the x-axis is oriented parallel to
the longest dimension of the ellipsoid and the z-axis parallel to its shortest
dimension ( Fig. B2.1(a)). The equation of the ellipsoid is
z
(a)
(b)
b
c
b
a
(c)
(d)
c
c
a
x
c
x
a
y
Fig. B2.1. (a) General ellipsoid with three unequal principal axes, a > b > c; (b)
elliptical cross-section through the center of an ellipsoid; b is the radius of a
circular section, inclined to the short axis c at an angle ; (c) prolate ellipsoid;
and (d) oblate ellipsoid.
x2 y2 z2
1
a2 b2 c2
(1)
where a, b, and c the intercepts of the ellipsoid with the x, y, and z reference
axes, respectively are the lengths of its principal axes. The volume of the
ellipsoid is
4
V abc
3
(2)
Each cross-section through the center of a triaxial ellipsoid is an ellipse,

except for two, which are circular sections. Dening the axes such that a > b
> c, the radius of a circular section is equal to the intermediate axis b and it is
inclined to the short axis c at an angle ( Fig. B2.1(b)) given by
76
Gravitation
a
tan
c
r
b2 c2
a2 b2
(3)
An ellipsoid of revolution is symmetric about one of its axes. If this is the

long x-axis, every axis in the yz plane is of equal length c. An ellipsoid with
this elongated shape is said to be prolate ( Fig. B2.1(c)). If the ellipsoid of
revolution is symmetric about its short z-axis, every axis in the xy plane is of
equal length a. An ellipsoid with this attened shape is said to be oblate (
Fig. B2.1(d)). An ellipsoid of revolution has only one circular section, which
lies in the (equatorial) xy plane of an oblate ellipsoid, or in the yz plane of a
prolate ellipsoid.
The equation of an oblate ellipsoid of revolution is
x2 y2 z2
21
a2
c
(4)
4
V a2 c
3
(5)
Its volume is
Every cross-section that includes the axis of rotational symmetry is an ellipse

with semi-major axis a and semi-minor axis c. These are related by the
ellipticity, f , dened as
ac
(6)
f
a
An oblate ellipsoid of revolution that is almost spherical in shape (i.e., the
axes a and c are almost equal) is called a spheroid. This is the closest
geometric approximation to the shape of the Earth; the ellipticity of a polar
section of the spheroid is called the attening.
collection of discrete point masses mi like the one at Q with Cartesian coordinates (xi, yi, zi). This point mass is distant ri from the center and ui from the
observation point at P. The gravitational potential at P can be written (compare
with ( 2.54)) as the sum of contributions from all the point masses in the body:
X mi
UG G
(2:106)
u
i
77
Table 2.1. Some useful geodetic parameters (source: Groten, 2004)

Parameter
Symbol
Geocentric gravitational constant

Mass of Earth: GE/G
Equatorial radius
Polar radius: a(1 )
Radius of equivalent sphere: (a2c)1/3
Flattening
Inverse attening
Dynamic form factor
Nominal mean angular velocity
Mean equatorial gravity
Acceleration ratio: 2a3/(GE)
Inverse acceleration ratio
Moment of inertia ratio for C
Moment of inertia ratio for B
Moment of inertia ratio for A
Dynamic ellipticity
Inverse dynamic ellipticity
14
Value
3 2
10 m s
1024 kg
km
km
km
103
GE
E
a
c
R
1/
J2
ge
m
1/m
C/(Ea2)
B/(Ea2)
A/(Ea2)
H
1/H
103
105 rad s1
m s2
103
103
3.986 004 418

5.973 7
6,378.136 7
6,356.752
6,371.000 4
3.352 865 9
298.252 31
1.082 635 9
7.292 115
9.780 327 8
3.461 391
288.901
0.330 701
0.329 622
0.329 615
3.273 787 5
304.513
(i ,i ,i )
z
(xi ,yi ,zi)
Q m
r
i
i
Units
ui
P
(x,y,z)
r
y
Fig. 2.3. Conguration for calculation of the gravitational potential of an ellipsoid,

considered as a distribution of discrete point masses mi.
Let the radius to the point mass at Q make an angle i with the radius to the
external point P. The reciprocal distance formula (1.157) for the Legendre
polynomials can be applied to the sides of the triangle OPQ:
1 n
1 1X
ri
Pn cos i
(2:107)
ui r n0 r
Substituting this into ( 2.106) gives for the gravitational potential of the body
78
Gravitation
( i , i ,i )
Q
( , , )
Fig. 2.4. Angle i bounded by straight lines OP, with direction cosines (, , ), and
OQ, with direction cosines (i, i, i).
UG G
X
i
mi
1 n
1X
ri
Pn cos i
r n0 r
(2:108)
Expanding the reciprocal distance formula gives an innite sequence of terms.

The ratio of successive terms depends on ri/r, which is less than 1 outside the
body. Moreover, if the shape of the body does not deviate much from a sphere,
higher-order terms are not signicant, so
UG G
1X
1X
1X
mi G 2
mi ri cos i G 3
mi r2i P2 cos i
r i
r i
r i
(2:109)
U 0 U1 U 2
Each term after the rst involves cos i, which can be computed ( Box 1.2,
equation (6)) from the direction cosines (, , ) of OP and the direction cosines
(i, i, i) of OQ, the lines bounding the angle i ( Fig. 2.4):
cos i i i i
(2:110)
The direction cosines of the two lines are as follows: for OP,
x
;
r
y
;
r
z
r
(2:111)
and for OQ,

i
xi
;
ri
yi
;
ri
zi
ri
(2:112)
Substituting into (2.110) gives

cos i
1
xxi yyi zzi
rri
(2:113)
Now we take a closer look at the individual terms in ( 2.109) for the potential.
For the case n = 0, potential U0:
U0 G
1X
GM
mi
r i
r
79
(2:114)
Comparison with ( 2.54) shows that U0 is the potential of a sphere at an external

point P.
U1 G
1X
mi ri cos i
r2 i
(2:115)
From ( 2.113) we obtain

1
ri cos i xxi yyi zzi
r
On substituting into ( 2.115) and gathering terms, we have
"
#
X
X
X
1
U1 G 3 x
m i xi y
m i yi z
m i zi
r
i
i
i
(2:116)
(2:117)
The origin of the coordinate system is at the center of mass of the body. The
center of mass is dened as the point about which the sums of the moments of
the point masses that make up the body are zero:
X
X
X
m i xi
m i yi
mi z i 0
(2:118)
i
Each sum on the right-hand side of ( 2.117) is zero, and consequently

U1 0
(2:119)

U2 G
1X
mi r2i P2 cos i
r3 i
(2:120)
On substituting for P2(cos ) from Table 1.1, we obtain

1 X
1 X
mi r2i 3 cos2 i 1 G 3
mi r2i 2 3 sin2 i
U2 G 3
2r i
2r i
(2:121)
"
U2 G
X
1 X
2mi r2i 3
mi r2i sin2 i
3
2r
i
i
#
(2:122)
The principal moments of inertia A, B, and C of a body about the x-, y-, and
z-axes, respectively, are dened in Box 2.2:
80
Gravitation
Box 2.2. Moments and products of inertia

The angular momentum h of a body rotating at angular velocity about an
axis is given by
h I
(1)
The quantity I is the moment of inertia of the body. It is a measure of the
distribution of its mass about an axis of rotation. For a point mass m at
perpendicular distance r from an axis of rotation the moment of inertia is
I mr2
(2)
If an extended body is made up of discrete particles with mass mi at distance

ri from the rotation axis, the moment of inertia is the sum of all the
contributions from all these particles:
X
mi r2i
(3)
I
i
Let the mass distribution of a body be described relative to three orthogonal

Cartesian coordinate axes. The moments of inertia A, B, and C about the
x-, y-, and z-axes, respectively, are
X

A
mi y2i z2i
i
B
C
X
i
X

mi z2i x2i
(4)

mi x2i y2i
Another property that affects the rotational behavior of a body is its product
of inertia about the axis of rotation. The products of inertia H, J, and K of a
body relative to the x-, y-, and z- reference axes are dened as
X
H
mi yi zi
i
m i z i xi
(5)
m i xi y i
Suppose that in a homogeneous body the zx plane is a plane of symmetry.

For every particle at (xi, yi) there is an equivalent particle at (xi, yi) that
cancels out its contribution to the product of inertia K, which is therefore
zero. If each pair of reference axes denes a plane of symmetry as in a
sphere, spheroid, or ellipsoid then all the products of inertia are zero. Nonzero products of inertia are expressions of the lack of symmetry of a
homogeneous body.

mi y2i z2i ;

mi z2i x2i ;
81

mi x2i y2i
(2:123)
Adding these moments of inertia gives
ABC2
mi r2i
(2:124)
Substituting into ( 2.122) gives

"
#
X
1
2
U2 G 3 A B C 3
mi r2i sin i
2r
i
(2:125)
Let the moment of inertia of the body about the line OP joining the center of the
ellipsoid and the point of observation be I ( Box 2.2). The distance of the point Q
from the line OP ( Fig. 2.3) is ri sin i and the moment of inertia I is given by
X
mi r2i sin2 i
(2:126)
I
i
The second-order term in the potential becomes

U2 G
1
A B C 3I
2r3
(2:127)
Combining the expressions for U0 and U2, the gravitational potential of the
spheroid at P is
UG G
M
A B C 3I
G
r
2r3
(2:128)
This is known as MacCullaghs formula (and dates from 1855).
2.5.1 Gravitational potential of a spheroid

The shape of the Earth deviates only slightly from a sphere and is best represented as a spheroid that is symmetric about the rotation axis. For an ellipsoid
the moment of inertia I in MacCullaghs formula can be expressed in terms
of the principal moments of inertia A, B, and C. The denition of I can be
expanded as
X
X
X
mi r2i sin2 i
mi r2i
mi r2i cos2 i
(2:129)
I
i
Because the sum of the squares of direction cosines is 1, we can write
82
Gravitation
X
mi r2i

mi x2i y2i z2i 2 2 2
(2:130)
Using the denitions of ri cos i ( 2.116) and the direction cosines (, , ) of

OP ( 2.111),
X
1X
mi r2i cos2 i 2
mi xxi yyi zzi 2
r
i
i
X
(2:131)
mi xi yi zi 2
i
Expanding the squared expression and taking the direction cosines outside the
sums gives
X
X
X
X
mi r2i cos2 i 2
mi x2i 2
mi y2i 2
mi z2i
i
mi xi yi 2
mi yi zi 2
mi zi xi
(2:132)
On combining ( 2.130) and ( 2.132), we have that the moment of inertia of the
ellipsoid about the line OP is
X
X
X

I 2
mi y2i z2i 2
mi z2i x2i 2
mi x2i y2i
i
2
X
i
mi xi yi 2
i
X
i
mi yi zi 2
mi zi xi
(2:133)
The rst three sums on the right are recognizable as the denitions of the
principal moments of inertia A, B, and C, while the nal three terms are
denitions of the products of inertia H, J, and K (see Box 2.2). Thus the
moment of inertia I about an axis with direction cosines (, , ) is related to the
principal moments and products of inertia by
I A2 B2 C2 2K 2H 2J
(2:134)
In an ellipsoid the xy, yz, and zx planes are planes of symmetry, so the
products of inertia are H = J = K = 0. The expression for I reduces in the case of
an ellipsoid to
I A2 B2 C2
Substituting this expression for I in MacCullaghs formula gives

!
A B C 3 A2 B2 C2
M
UG G G
2r3
r
(2:135)
(2:136)
83
= cos
= sin sin
sin
= sin cos
Fig. 2.5. Relationship between the direction cosines of a line and the angles and
that dene its direction.
The symmetry of the Earth about its rotation axis means that the moment of
inertia about any axis in the equatorial plane has the same value, i.e., A = B. For
the spheroidal Earth this results in
!

2A C 3A 2 2 3C2
M
UG G G
(2:137)
2r3
r
Now we revert from the direction cosines of OP to the direction of the line in
terms of the angles and , corresponding respectively to co-latitude and
longitude in geographic terms. These angles and the direction cosines are
related as in Fig. 2.5:
sin cos
sin sin
cos
Squaring and summing the direction cosines and gives

2 2 sin2 cos2 sin2 sin2
1 cos2
Replacing the direction cosines with the above expressions gives

2A C 3A 1 cos2 3C cos2
M
UG G G
r
2r3

M
1 3 cos2
UG G GC A
r
2r3
(2:138)
(2:139)
(2:140)
(2:141)
84
Gravitation
UG G
M
CA
G 3 P2 cos
r
r
(2:142)
This is the gravitational potential of an ellipsoid of revolution at an external point.
2.5.2 MacCullaghs formula and the gure of the Earth

The Earths shape deviates only slightly from a sphere, and is close to that of an
oblate spheroid. MacCullaghs formula is not an exact expression for the
gravitational potential of the Earth, because terms of higher order than U2
were omitted from ( 2.109). In order to express UG more exactly, we need to
use an innite series of potentials:
UG U0 U1 U2 U3
1
X
Un
(2:143)
n0
Each term of order n is proportional to (1/r)n and decreases in relative importance with increasing distance r. An alternative form for the gravitational
potential UG of the Earth at an external point is to write it as an innite series
of terms involving the Legendre polynomials and using Earths mass E and
equatorial radius a:
"
#
n
1
X
E
a
Jn
Pn cos
(2:144)
UG G
1
r
r
n2
The sum inside the square brackets modies the potential U0 of a sphere to
reect the real mass distribution in the Earth. The coefcients Jn describe the
relative importance of successive terms in the series. The sum begins at n = 2
because U1 = 0 when the coordinate system is centered at the Earths center of
mass, as in ( 2.119). Values for the coefcients Jn are obtained from satellite
geodesy. They are very small, of order 106, except for J2, which is about 1,000
times larger and has the value 1.082 103. J2 is called the dynamic form factor
of the Earth. The coefcient J3 has the value 2.54 106; it describes a slight
deviation from a spheroid, being more depressed at the south pole and elevated
at the north pole. This makes the Earth slightly pear-shaped. The coefcient J4 is
equal to 1.59 106 and is needed in order to obtain a more exact description
of the gravitational potential for a model Earth whose mass distribution is
symmetric about the equator.
Writing (2.144) to rst order:
"
#
2
E
a
1 J2
UG G
P2 cos
(2:145)
r
r
Further reading
85
This has to be equivalent to MacCullaghs formula for the spheroidal Earth. On

equating terms in ( 2.142) and ( 2.145), we get the result
2
E
a
CA
G J2
P2 cos G 3 P2 cos
r
r
r
(2:146)
where
J2
CA
Ea2
(2:147)
This result shows that the dynamic form factor J2 is dependent on the difference
between the principal moments of inertia, C and A. The polar attening of
Earths gure results from the centrifugal acceleration of its rotation. The
redistribution of mass nds expression as a difference between the principal
moments of inertia. This difference, in turn, affects how the Earth reacts to
external gravitational torques, which cause the rotation axis to precess about the
pole to the ecliptic. The difference between C and A even affects the free
rotation of the Earth, creating a longer-period wobble that is superposed on
the daily rotation.
further reading
Blakely, R. J. (1995). Potential Theory in Gravity & Magnetic Applications. Cambridge:
Lowrie, W. (2007). Fundamentals of Geophysics, 2nd edn. Cambridge: Cambridge
University Press, 381 pp.
Ofcer, C. B. (1974). Introduction to Theoretical Geophysics. New York: Springer,
385 pp.
Stacey, F. D. and Davis, P. M. (2008). Physics of the Earth, 4th edn. Cambridge:
3
Gravity
At any point on the Earth gravity acts in a direction normal to a surface on which
the potential of gravity is constant. This equipotential surface is the best-tting
geometric gure to mean sea-level on the Earth. Its shape is that of a slightly
attened spheroid, for which the radius at any point can be computed. The
potential of gravity on this spheroid the geopotential is computed by
combining the gravitational potential and the potential of the centrifugal acceleration due to Earths rotation. Gravity measurements are made with a high
degree of accuracy. In order to compute a theoretical value of gravity for
comparison at any latitude similar accuracy must be attained. Consequently,
each step in computing the formula for the reference gravity must be carried out
to second order in the attening f and related parameters.

Every cross-section of Earths spheroidal shape that includes both poles is an
identical ellipse, with equatorial semi-major axis a and polar semi-minor axis c,
which are related (Box 2.1) by the attening f through the equation c = a(1 f ).
In Cartesian coordinates the equation of the ellipse is
x2 z2
1
a2 c2
(3:1)
A position on the reference spheroid is specied by the polar angle and radius r,
dened relative to the axis of rotational symmetry and center of the spheroid,
repectively (Fig. 3.1). Consider a polar cross-section that includes the x- and
z-axes, so that x = r sin and z = r cos . By substituting into (3.1) we get the
equation of the elliptical section in polar coordinates:
r2 sin2 r2 cos2
1
a2
c2
86
(3:2)
87
c
r
a
R
Fig. 3.1. Polar cross-section of a spheroid with principal axes a and c (c < a),
compared with a sphere (dashed) with radius R and the same volume as the spheroid.
r2
cos2
sin2
2
a
1 f 2
!
1
(3:3)
On rearranging slightly, this becomes

r2
a2 1 f 2
cos2 1 f 2 sin2
(3:4)
The denominator can be expanded, giving

cos2 1 f 2 sin2 1 2f f 2 sin2 cos2
sin2 cos2 2f sin2 f 2 sin2
(3:5)
Noting that sin2 + cos2 = 1, we can rewrite this as

cos2 1 f 2 sin2 1 2f sin2 f 2 sin2 sin2 cos2
1 2f sin2 f 2 sin4 f 2 sin2 cos2

2
1 f sin2 f 2 sin2 cos2
(3:6)
By substituting into (3.4) and taking the square root, we get an equation for the
radius:
r
1f

1=2

2
a
1 f sin2 f 2 sin2 cos2
!1=2
1f
f 2 sin2 cos2
1
2
1 f sin2
1 f sin2
(3:7)
88
Gravity
Applying the binomial theorem twice to the last line and expanding to order f 2
gives an equation for the surface of a spheroid

r
1f
1 2 2
1f
2

1 f sin cos
(3:8)
2
a 1 f sin
2
1 f sin2
The expansions for the gravitational potential and for gravity on the reference
ellipsoid require the ratio a/r. Upon inverting (3.8) with the aid of the binomial
expansion we get, to order f 2,

a 1 f sin2
1

1 f 2 sin2 cos2
r
1f
2

1
1 f sin2 1 f 2 sin2 cos2 1 f f 2
2
1
1 f f 2 f sin2 f 2 sin2 f 2 sin2 cos2
(3:9)
2
a
1
1
1 f cos2 f 2 cos2 f 2 cos2 f 2 cos4
r
2
2

3
1
1 f 1 f cos2 f 2 cos4
2
2
(3:10)
For some purposes it sufces to know the equation of the ellipticity only to rst
order in f. This is derived in Box 3.1.

The main component of gravity is the gravitational acceleration aG towards the
center of the Earth. This component varies with latitude because of the varying
radius of the spheroid. The deviation from a spherical shape results from the
deforming effect of Earths rotation, which produces a centrifugal acceleration
ac directed perpendicular to and away from the axis of rotation (Fig. 3.2). This
component is proportional to the distance from the rotation axis, so it also varies
with latitude.
Gravity is the vector combination of the centrifugal and gravitational components, each of which is conservative and is the gradient of a scalar potential.
The potential of gravity Ug at a point on Earths surface, the geopotential, is the
sum of the gravitational potential UG and the centrifugal potential Uc at that
point,
Ug UG Uc
(3:11)
89
Box 3.1. First-order equation of a slightly attened spheroid

The equation in polar coordinates of an ellipse with semi-major axis a and
ellipticity f is, from (3.8),
r
1f
a 1 f sin2
This equation can be expanded using the binomial theorem:

1

r
1 f 1 f sin2
1 f 1 f sin2
a
(1)
(2)
Because f is equal to 1/298.252 (Table 2.1), the quantity f 2 is of the order of

105 and is for many purposes negligibly small. The binomial expansion
may be curtailed to rst order in f, giving

r
(3)
1 f f sin2 1 f 1 sin2
a
r
1 f cos2
a
(4)
It is often convenient to express the elliptical polar section in terms of the

Legendre polynomial P2(cos ). Rearranging the equation for P2(cos ) from
Table 1.1 gives
1
cos2 1 2P2 cos
3
(5)
By substituting into (4) above, we get

r
f 2
1 fP2 cos
a
3 3
(6)
Upon invoking the binomial expansion and ignoring terms of second order
and higher in f, this reduces to

r
f
2
(7)
1
1 fP2 cos
a
3
3
Let R be the radius of a sphere with the same volume as the spheroid
(Fig. 3.1). Then, omitting the factor 4/3 common to each volume, we have
R 3 a 2 c a 3 1 f
(8)
90
Gravity
Taking the cube root and using the binomial expansion to rst order gives

f
1=3
R a 1 f
(9)
a 1
3
Thus the equation for the radius of an elliptical polar section of the Earth in
terms of the Legendre polynomial P2(cos ), the attening f, and the mean
radius R of an equivalent sphere is

2
r R 1 fP2 cos
(10)
3
This is a useful rst-order approximation to the shape of the Earth.
x
ac
Fig. 3.2. Centrifugal acceleration ac at co-latitude , directed perpendicular to and

away from the axis of rotation.
3.2.1 Gravitational potential

To compute gravity on the reference spheroid it is necessary to determine the
geopotential to second order in the small quantities that dene it. Each of the
quantities f, m, and J2 is around 103 in size (Table 2.1), so their squares and
products are around 106. The gravitational potential (2.144) must be determined with the same denition, which means that it is inadequate to use only the
terms up to J2. If we assume that the mass distribution of the Earth is symmetric
about the equator, the term J3 can be omitted, but we need to include the term J4
for an accurate description of the gravitational potential. Up to the term J4 this
becomes
GE
UG
a

3
5

a
a
a
J2
P2 cos J4
P4 cos
r
r
r
91
(3:12)
3.2.2 Centrifugal potential

The centrifugal acceleration is the gradient of the centrifugal potential Uc,
ac rUc
(3:13)
Let x be the perpendicular distance from the rotation axis to a point on the
surface at latitude and let be the angular rate of rotation of the Earth
(Fig. 3.2). The centrifugal acceleration is equal to 2x, so, for a constant rate
of rotation, Uc varies only with x. Therefore
2 x
Uc
x
(3:14)
Integrating both sides with respect to x gives

1
Uc 2 x2 U0
2
(3:15)
The potential is zero at the axis of rotation, where x = 0, and the constant of
integration Uc = 0. The equation for the centrifugal potential in terms of the
polar angle is
1
1
Uc 2 x2 2 r2 sin2
2
2
(3:16)

In order to compute gravity accurately on the reference ellipsoid it is necessary
to develop the geopotential to second order in the small quantities f, m, and J2,
so we must use also the gravitational potential coefcient J4 whose magnitude is
around 106. The geopotential consists of the sum of the gravitational and
centrifugal potentials:
"
#
3
5
GE a
a
a
P2 cos J4
P4 cos
J2
Ug
a
r
r
r
2
1
r
2 a2
sin2
(3:17)
2
a
Taking the centrifugal term inside the bracketed expression gives
92
GE
Ug
a
Gravity
"
#

2
3
5
a
a
a
1 2 a3
r
P2 cos J4
P4 cos
sin2
J2
r
r
r
2 GE
a
(3:18)
3
The geopotential involves the ratios a/r, (a/r) , and (a/r) , which we develop
using (3.10). Note that the term in (a/r)3 is multiplied by J2 so it must be
evaluated only to rst order in f; the coefcient J4 is itself of order 106, so the
ratio (a/r)5 on the equipotential surface of gravity may be set equal to 1. Then

3
a
3
1
1 3 f 1 f cos2 f 2 cos4 1 3f cos2
(3:19)
r
2
2
For succinctness, let the last term inside the brackets in (3.18) be called . The
ratio r/a is obtained from (3.8), thus

2
1 2 a3
r
sin2
2 GE
a

(3:20)
1 2 a3 1 f 1 f sin2

2
2
GE
1 f sin2
1 1 f sin2
m
2
2
1 f sin2
(3:21)
Here m is the centrifugal acceleration ratio dened in Box 3.2, equation (3),
m
2 a3 1 f
GE
(3:22)
The denominator in (3.21) can be expanded using the binomial theorem; we

need do so only to rst order because of the factor m, which is similar in size to f.
The centrifugal term becomes

1
m1 f 1 2f sin2 sin2
2
(3:23)
Multiplying, and retaining only the terms of rst order in f, gives

1
m sin2 1 f 2f sin2
2
(3:24)
In the equation for the geopotential, the centrifugal term must be combined with
a term in J2P2(cos ), which has the form cos2, and a term in J4P4(cos ), which
93
Box 3.2. The acceleration ratio, m

The magnitudes of the gravitational and centrifugal components of gravity
can be directly compared at the equator where the vectors are directly
opposed to each other. The parameter m is dened as the ratio of the
centrifugal acceleration at the equator to the gravitational acceleration at the
equator:
m
2 a
2 a3
GE
GE=a2
(1)
The value of m dened in this way is 3.461 391 103 = 1/288.901.

An alternative, commonly used denition of m is the ratio of the equatorial
centrifugal acceleration to the gravitational acceleration on a sphere with the
same volume as the spheroid. The volume of a spheroid with equatorial
radius a and polar radius c is (4/3)a2c. The attening f relates a and c so
that c = a(1 f ). Let the radius of a sphere with the same volume be R;
its volume is (4/3)R3. On comparing the volumes and dropping the
common numerical factor, we have
R3 a2 c a3 1 f
(2)
The alternative denition of the acceleration ratio m is then

m
2 R3 2 a3 1 f
GE
GE
(3)
In this case the value of m is 3.449 786 103 = 1/289.873.

contains terms in both cos2 and cos4 (see Table 1.2). It is advantageous to
convert (3.24) to the same format:

1
m 1 cos2 1 f 2f 1 cos2
2

1
m 1 f 2f cos2 cos2 f cos2 2f cos4
2

1
m 1 f 1 3f cos2 2f cos4
2
(3:25)
(3:26)
Now we can return to (3.18). By writing the full expressions for P2(cos ) and
P4(cos ) from Table 1.1, and the ratios a/r from (3.10) and (a/r)3 from (3.19),
94
Gravity
and using (3.26) for the centrifugal term, we get the geopotential as a function of
cos2 and cos4:

3
1 f 1 32 f cos2 12 f 2 cos4

2
cos4 =2 7
GE 6
6 J2 1 31 f cos 9f
7

Ug
6
7
5
a 4 J4 3 30 cos2 35 cos4 =8

1
2
4
2 m 1 f 1 3f cos 2f cos
2
(3:27)
After gathering terms to get the coefcients that multiply cos2 and cos4, we
get the nal expression for the geopotential:
2
3
1 f 12 m 12 J2 38 J4

GE 6
7
J4 cos2 5
Ug
4 f 32 f 2 12 m 32 fm 32 1 f J2 15
4
1 2

a
4
2 f mf 92 fJ2 35
8 J4 cos
(3:28)
3.3.1 Relationship of J2, J4, f, and m

By denition, the geopotential must be constant on the equipotential surface.
However, the potential in (3.28) can vary with polar angle through the terms in
cos2 and cos4. This apparent contradiction implies that the coefcients of
these terms must be zero, i.e.,
1
3
3
3
15
f m f 2 fm 1 f J2 J4 0
2
2
2
2
4
(3:29)
1 2
9
35
f mf fJ2 J4 0
2
2
8
(3:30)
Since J4 is much smaller than J2, we can neglect it initially and write (3.29) to
rst order:
1
3
f m J2 0
2
2
1
J2 2f m
3
(3:31)
(3:32)
This value for J2 is now inserted into (3.30) to obtain a second-order equation
for J4:

8
1
9 2
1
f 2 mf f f m
35
2
2 3
3
4
4 2
fm f
7
5
95
J4
(3:33)
By inserting this expression back into (3.29) we eliminate J4 and get an equation
for J2:

2
1
5 4
4 2
2
(3:34)
1 f J2 f m f fm
fm f
3
3
2 7
5
2
1
3
1 f J2 f m f 2 fm
3
3
7
Applying the binomial theorem to rst order in f gives

2
1
3
2
J2
f m f fm 1 f
3
3
7
(3:35)
(3:36)
After multiplying and tidying up the terms, we get a second-order equation

for J2:

1
2
J2
2f m f 2 fm
(3:37)
3
7
3.3.2 Inferred increase of density with depth in the Earth

In Section 2.5.2 the dynamic form factor J2 is expressed in terms of the principal
moments of inertia. We can replace the equatorial radius a by the mean radius R,
so that to rst order
J2
CA CA

Ea2
ER2
(3:38)
By combining this result with (3.32) we obtain a relationship among the difference in the principal moments of inertia, the attening responsible for the
difference, and the centrifugal acceleration that causes the deformation:
CA 1
2f m
ER2
3
(3:39)
Equation (3.39) allows us to make an inference about the distribution of mass

inside the Earth. The Sun and Moon exert torques on the spheroidal shape of the
Earth that cause the rotation axis to precess about the pole to the ecliptic plane,
96
Gravity
Hollow cylinder
Hollow sphere
C = MR 2
C=
MR 2
Solid sphere
C=
MR 2
Fig. 3.3. Moments of inertia of a hollow cylinder, hollow sphere, and uniform solid
sphere about an axis of symmetry.
which is manifest in the precession of the equinoxes (see Section 5.3). The rate
of precession is determined by the dynamic ellipticity H, dened as
H
C A B=2 C A

C
C
(3:40)
The value of H is known quite accurately from astronomic observations. H is a

very small quantity of the same order as f and m (Table 2.1). Rewriting (3.39) gives

CA C
1
2f m
(3:41)
C
ER2 3

C
1 2f m
1

(3:42)
ER2 3
H
3
1
C ER2
3
(3:43)
Figure 3.3 shows the moments of inertia of some standard objects about an axis
of symmetry. With increasing distribution of the mass of the object closer to its
center, the factor preceding the product MR2 decreases from 1 for an openended hollow cylinder to 0.67 for a hollow spherical shell and 0.4 for a
homogeneous solid sphere. The numerical factor is 0.33 for the Earth, indicating that the density of the Earth is not uniform but increases towards its center,
i.e., the density of the Earth increases with depth.

The reference gure for standard calculations of gravity at a particular latitude is
the spheroid, or ellipsoid, of revolution. The acceleration due to gravity on the
reference spheroid has both a radial component gr and a polar component g,

g gr er g e
97
(3:44)
The polar component g is much smaller than the radial component gr, but it has
important effects. It deects the vertical from the radial direction at every point
on the Earth, except at the poles and on the equator. This deection results in a
difference between geocentric and geographic latitude; the maximum difference
is less than 0.2, but this has a large effect on measurements of gravity. The polar
component cannot be neglected, since this would be akin to assuming that
gravity acts in a radial direction at all points. To determine the theoretical
gravity on the reference spheroid we must combine expressions for the radial
and polar components:
!

1=2
1 g 2
2
2
g g r g
(3:45)
gr 1
2 gr
As we will see, the polar component g is of order f, so its effect on gravity is
proportional to f 2. To determine the variation of gravity on the reference
ellipsoid we will have to evaluate the radial component to second order as
well. This makes it necessary to express the shape of the spheroid and the
geopotential to second order in the small quantities f, m, and J2. We must also
use an expression for the gravitational potential up to the coefcient J4, which is
about the same size as the squares and products of these parameters.
3.4.1 Polar component of gravity

The polar component of gravity on the reference ellipsoid is the gradient of the
geopotential in the direction of increasing polar angle ,
(
"
#
)
3
5
1
GE a
a
a
1 2 2 2
J2
g
P2 cos J4
P4 cos r sin

r
a
r
r
r
2
(3:46)
The rst term is independent of and drops out of the differentiation. We can
take the centrifugal term inside the square brackets and use the denition of the
centrifugal ratio m as in (3.22):
"
6
4
GE
a
g 2 J2
P2 cos J4
P4 cos
a
r

1
m
r
2
sin

(3:47)
2 1 f a
98
Gravity
The Legendre polynomials P2(cos ) and P4(cos ) are listed in Table 1.1.
Differentiating them with respect to gives

3 cos2 1
P2 cos
3 cos sin
(3:48)
2

35 cos4 30 cos2 3
P4 cos

5
(3:49)
cos sin 7 cos2 3
2
On substituting these into (3.47) and simplifying, we obtain
"
#
6
4

GE
a
5
a
m
r
J4
7 cos2 3
g 2 sin cos 3J2
a
r
2
r
1f a
(3:50)
As explained above, we need to evaluate g only to rst order in f, so terms with
J4 and the products fJ2 and fm may be neglected. The ratios (a/r)4, (a/r)6, and r/a
may be set effectively equal to 1. We dene
g0
GE
a2
(3:51)
The polar component of gravity on the reference ellipsoid is therefore given to

rst order by
g g0 3J2 msin cos
(3:52)
Now we recall the relationship among J2, f, and m established in (3.32) and
substitute for J2, which gives the rst-order expression
g g0 f sin2
(3:53)
Note that g is positive for 90 and negative for 90 180, i.e., in each
hemisphere g acts in the direction from the pole to the equator.
3.4.2 Radial component of gravity

The radial component of gravity on the reference ellipsoid is obtained from the
gradient of the geopotential with respect to the radius r:

3
5
GE a
a
a
P2 cos J4
P4 cos

J2
gr
r
a
r
r
r
#)

2
1 2 a3
r
sin2
(3:54)
2 GE
a
GE
gr 2
a
"
4
6
2
a
a
a
P2 cos 5J4
P4 cos
3J2
r
r
r

m
r
2

sin
1f a
99
(3:55)
To simplify this cumbersome evaluation somewhat, we examine the four terms

inside the square brackets individually. We write g0 as in (3.51):
gr g0 T1 T2 T3 T4
(3:56)
For term T1, using the ratio a/r dened in (3.10), and neglecting terms of higher
order than f 2, the rst term in square brackets is
2

a
3
1
1 2f 1 f cos2 f cos4
r
2
2

2

3
1
f 2 cos4 1 f f cos2
(3:57)
2
2
Thus

T1 1 2f 3f 2 cos2
(3:58)
For term T2, the term in (a/r)4 is multiplied by J2, so we need only expand it to
order f :

4

a
3
1
2
4
1 4 f 1 f cos f cos
(3:59)
r
2
2
4
a
1 4f cos2
(3:60)
r
Using the expansion of the Legendre polynomial P2(cos ) given in Table 1.1,

3
T2 3J2 1 4f cos2 P2 cos J2 1 4f cos2 3 cos2 1
2
(3:61)

3
3
T2 J2 3 2f J2 cos2 18fJ2 cos4
(3:62)
2
2
For term T3, the term in (a/r)6 is multiplied by J4, which is of order 106, so
we can neglect products of J4 with f. Effectively we can set (a/r)6 equal to 1.
Using the expansion of P4(cos ),

5
T3 5J4 P4 cos J4 3 30 cos2 35 cos4
8
(3:63)
100
Gravity
For term T4, the ratio r/a is given by (3.8), and to second order this term is

m
1f
sin2
(3:64)
T4
m sin2 1 f sin2
2
1f
1 f sin
On converting the sines to cosines for compatibility with the other terms we obtain
T4 m1 f m1 2f cos2 mf cos4
Now we can insert these four terms into (3.56):
2
3
1 f 2 3f cos2

6 3
7
6 J2 3 3 2f J2 cos2 18fJ2 cos4 7
2
2
6
7
gr g0 6

7
4 58 J4 3 30 cos2 35 cos4
5
2
4
m1 f m1 2f cos mf cos
(3:65)
(3:66)
After gathering terms to form coefcients of cos2 and cos4, we have

2
3
1 32 J2 15
J4 m1 f
8

2 7
6
gr g0 4 f 2 3f 3 32 2f J2 75
4 J4 m1 2f cos 5

4
mf 18fJ2 175
8 J4 cos
(3:67)
J2 and J4 can be replaced by expressions in f and m, as in (3.37) and (3.33),
respectively. After expanding and grouping the terms, the radial gravity component becomes
2
3
1 f 32 m f 2 27
14 fm

6
72
2
2 7
(3:68)
gr g0 4 52 m f 13
2 f 7 fm cos 5
15

2
4
2 fm 11
2 f cos
3.4.3 Variation of reference gravity with geocentric latitude

Instead of using the polar angle to describe position on the reference ellipsoid,
it is customary to use the latitude. The geocentric latitude c is the complement
of , so cos = sin c, cos2 = sin2c, and

1
cos4 sin4 c sin2 c 1 cos2 c sin2 c sin2 2c
4
(3:69)
On substituting this change, the radial component of gravity on the spheroid as a

function of geocentric latitude is
101
3
1 f 32 m f 2 27
14 fm
2 7
6 5
2
39
7
gr g0 6
4 2 m f f 14 fm sin c 5
18 f15m 11f sin2 2c
(3:70)
Note that the polar component g (see (3.53)) referred to geocentric latitude is
unaltered:
g g0 f sin2 g0 f sin2c
(3:71)
Gravity on the reference gure of the Earth acts normal to the ellipsoidal
equipotential surface. It is computed by combining the radial and polar components as in (3.45):
2 !

2 !
1 g
1 2 2
3
gr 1 f sin 2c 1 f m
g gr 1
2 gr
2
2
(3:72)

1
g gr 1 f 2 sin2 2c
2
(3:73)
Thus the polar component affects only the sin2(2c) term in (3.70), and gravity
on the reference ellipsoid is given by
2
3
1 f 32 m f 2 27
14 fm
6 5
2 7
2
39
7
g g0 6
(3:74)
4 2 m f f 14 fm sin c 5
18 f15m 7f sin2 2c
Let the value of gravity at the equator, where sin c = sin(2c) = 0, be

3
27
ge g0 1 f m f 2 fm
2
14
(3:75)
Taking this out of the bracketed expression and using the binomial expansion to
rst order in f gives

1
3
27
g ge 1A sin2 c f15m7f sin2 2c 1 f m f 2 fm
8
2
14
(3:76)
2
where, for succinctness, A 52 m f f 2 39
14 fm. The coefcient of sin (2c)
is already of second order, so, when we multiply the terms, only the coefcient
A of sin2c is affected. It expands to
102
Gravity

5
39
3
27
m f f 2 fm 1 f m f 2 fm
2
14
2
14
5
39
5
15 2 3
2
2
m f f fm fm f m fm
2
14
2
4
2
5
15
17
m f m2 fm
2
4
14
(3:77)
The nal expression for the variation of gravity with geocentric latitude is

5
15
17
1
g ge 1
m f m2 fm sin2 c f15m 7f sin2 2c
2
4
14
8
(3:78)
3.4.4 Clairauts formula

The value of gravity at the poles, gp, is found by setting c = /2 = 90. To rst
order

5
gp ge 1
mf
(3:79)
2
Rearranging this equation gives
gp ge 5
mf
ge
2
(3:80)
This is the Clairaut formula for the difference between the gravity at the pole
and that at the equator, attributed to a French mathematician and astronomer,
Alexis Claude de Clairaut (17131765).

The latitude in the above formulae is the geocentric latitude c dened by the
radius from the Earths center to the point on the ellipsoid. However, the latitude
in common use is the geographic (or geodetic) latitude dened by the vertical
direction, which is normal to the surface of the reference ellipsoid and does not
pass through the Earths center (Fig. 3.4). There is a simple relationship between
the geocentric and geographic latitudes.
Let P be a point on the ellipsoid with geocentric latitude c and geographic
latitude (Fig. 3.5(a)). The angle between the radial and vertical directions at
P is c. The horizontal direction PH and the direction PN normal to the radius
103
ac
aG g
Fig. 3.4. Comparison of geocentric latitude c, dened by the radius of the

ellipsoidal Earth, and geographic (or geodetic) latitude , dened by the normal
direction to the surface of the ellipsoid. After Lowrie (2007).
(a)
(b)
r d
vertical
H
dr
N
horizontal
H
N
Fig. 3.5. (a) The difference ( c) between geographic latitude and geocentric
latitude c is the same as the angle between the horizontal and a plane perpendicular
to the radius. (b) Details of the construction of a small triangle whose sides PN and
PH contain the angle ( c).
at P form the same angle. Consider a small increase d in the polar angle for the
point P. The radius to the surface increases by a small amount dr, and there is an
angular displacement r d perpendicular to the radius. These increments displace the intersection of the radius with the surface along the ellipsoid. The three
displacements form a small triangle PNH (Fig. 3.5(b)), whose sides PN and PH
contain the angle ( c). In the triangle PNH
tan c
dr
r d
(3:81)
104
Gravity
On differentiating the equation of the ellipsoid (3.8) we have to rst order in f

1 dr a d
1f
r d r d 1 f sin2
a f2 sin cos

r 1 f sin2 2
(3:82)
f sin2
Because is the complement of c, we can replace sin(2) by sin(2c) and obtain
the result
tan c f sin2c
(3:83)
The difference = c is very small, because the tangent of the angle is less
than f,
c tan1 f 0:19
(3:84)
The small difference allows us to replace the tangent in (3.83) with the angle (in
radians), so that
c f sin2c
c f sin2c
(3:85)
(3:86)
3.5.1 Normal gravity on the reference ellipsoid

Measurements of gravity must be corrected for various factors, such as the
latitude of the measurement site, its altitude with respect to the reference
ellipsoid, and the surrounding topography. The corrected value must then be
compared with the theoretical value for the geographic latitude of the observation. The gravity formula in (3.78) gives the variation of gravity with geocentric
latitude. This must now be converted to a form that depends on geographic
latitude, which requires nding expressions for sin2c and sin2(2c) in terms of .
The gravity formula in (3.78) can be written
gn ge 1 b1 sin2 c b2 sin2 2c
(3:87)
On comparing (3.87) with (3.78), we note that the constant b1 contains terms of
both rst and second order in f and m, whereas b2 is entirely of second order.
This allows us to simplify the conversions.
105
From (3.86) we have c = and, because is a very small angle, we can

make the approximations sin() and cos() 1. The expressions for sin c
and cos c reduce to
sin c sin sin cos cos sin sin cos
(3:88)
cos c cos cos cos sin sin cos sin
(3:89)
The gravity formula contains the term sin2c, which we can now write as
sin2 c sin2 2 sin cos sin2 sin2
(3:90)
Next, we combine (3.88) and (3.89) to get an expression for sin2(2c), which is,
to rst order in ,
sin2c 2sin cos cos sin

2 sin cos 2 cos2 sin2 22 sin cos
sin2 2 cos2
(3:91)
Squaring, and again neglecting the term in ()2, gives

sin2 2c sin2 2 4 sin2cos2 sin2 2 2 sin4
(3:92)
In the gravity formula (3.87) this term is multiplied by the constant b2, which is
of second order in f and m. Thus, neglecting the small product b2 ,
b2 sin2 2c b2 sin2 2 2b2 sin4 b2 sin2 2
(3:93)
Equation (3.91) allows us to rewrite in (3.86),

f sin2c f sin2 f cos2 f sin2
(3:94)
Upon inserting this into (3.90) we get

sin2 c sin2 f sin2 2
(3:95)
Substituting (3.93) and (3.95) into (3.87) gives the gravity formula for geographic latitude :

(3:96)
gn ge 1 b1 sin2 f sin2 2 b2 sin2 2
gn ge 1 b1 sin2 b2 fb1 sin2 2
(3:97)
106
Gravity
The coefcient of sin2 is the same as that of sin2c in the gravity formula (3.87)
for geocentric latitude, but the coefcient of sin2(2) is modied to

1
5
15 2 17
b2 f b1 f15m 7f f m f m fm
8
2
4
14
(3:98)
1
f f 5m
8
On replacing b1 and b2 by the corresponding expressions in (3.78), we get the
normal gravity formula

(3:99)
gn ge 1 1 sin2 2 sin2 2
in which gn is the normal gravity at geographic latitude on the International
Reference Ellipsoid, ge is its value at the equator, and 1 and 2 are small
constants, given by
5
15
17
1 m f m2 fm
2
4
14

1 2
2 f 5fm
8
(3:100)
From (3.51) and (3.75) the value of gravity on the equator is given by

GE
3
27
ge 2 1 f m f 2 fm
a
2
14
3.6 The geoid

The real surface of the Earth is irregular and cannot be described by a simple
geometric form. It is replaced by a smooth equipotential surface of gravity,
chosen so that it agrees with mean sea-level far from land. This surface is called
the geoid. The distribution of density in the Earths crust is complex, with local
mass anomalies that inuence the geoid and cause it to undulate about a mean
shape. The mathematical reference gure for the Earth is a spheroid that has the
same volume and the same potential as the geoid.
A local excess of mass deects the direction of a plumb-line towards it and at
the same time increases the local value of gravity. In order to maintain a constant
potential, the equipotential surface must bulge upwards over the excess mass.
The shape of the bulge is determined by the condition that the equipotential
must lie normal to the direction of gravity and hence to the plumb-line. The
mass excess elevates the geoid above the spheroid (Fig. 3.6); conversely, a mass
3.6 The geoid
local
gravity
107
geoid
G
h
E
ellipsoid
mass
excess
Fig. 3.6. Elevation of the geoid above the reference ellipsoid due to an excess of
mass below the ellipsoid, and related local deections of the direction of gravity.
After Lowrie (2007).
decit depresses the geoid below the spheroid. The undulations of the geoid
with respect to the spheroid correlate with the gravity anomalies caused by the
inhomogeneity of density. The height of the geoid relative to the spheroid may
be calculated from an analysis of these gravity anomalies.
3.6.1 The potential of a geoid undulation

Let E be a point on the reference ellipsoid (idealized gravity equipotential) over
an anomalous mass. The effect of this mass is to raise the geoid (true gravity
equipotential) so that the point G corresponding to E is at a height h above the
ellipsoid (Fig. 3.6). The work done against gravity g changes the potential. If the
displacement h is small, the additional potential W due to the excess mass is
simply W = gh. Thus the height of the geoid above the spheroid is
h
W
g
(3:101)
Gravity observations are rst corrected for local topography and transient tidal
effects. The corrected value is then reduced to the reference surface by compensating for the altitude of the measurement station. A gravity anomaly is
computed by subtracting the theoretical gravity for the latitude of the measurement station. However, altitudes are specied relative to mean sea-level, so the
altitude adjustment reduces the gravity value to the geoid rather than the
ellipsoid. The gravity anomaly after corrections and reduction is specied at
the point G on the geoid, but the reference value is computed for the point E on
the ellipsoid (Fig. 3.6). The height difference corresponds to the geoid undulation, which must be taken into account in an accurate gravity survey.
The gravity anomaly g at the point G arises from two superposed effects.
The main effect is the gravitational attraction of the additional mass. This causes
108
Gravity
a vertical gravity anomaly g1 that can be calculated to rst order by assuming

the vertical and radial directions to be the same and differentiating the potential
W with respect to r,
Dg1
W
r
(3:102)
The second contribution g2 to the gravity anomaly is the effect of the distance
h between the geoid and spheroid. This can be computed in an analogous way to
the gravity free-air correction:
g
r

g
GE
g
2 2
r r
r
r
Dg2 h
(3:103)
(3:104)
On combining the two contributions, we get for the gravity anomaly of the
anomalous mass

W
W
Dg Dg1 Dg2
2
(3:105)
r
r
The geoid undulations h are much smaller than the Earths radius R, so it is
unimportant if this expression is evaluated on the spherical Earth rather than on
the actual spheroid. We can conveniently use the surface of the sphere r = R, in
which case

1 2
(3:106)
r W
Dg 2
r r
rR
3.6.2 Stokes formula for the height of the geoid

Suppose the height of the geoid is to be determined at a point P from gravity
anomalies on the Earths surface r = R. Let the spherical coordinates be dened
relative to a radial axis through the point P. For a point Q where gravity was
measured, is the polar angle relative to P and is the azimuth of Q on a circle
around P. The gravity anomalies on the spherical surface can then be expressed
as a sum of spherical harmonic functions, Ym
n ; (see Section 1.16):
Dg;
1 X
n
X
n0 m0
m
gm
n Yn ;
(3:107)
3.6 The geoid
109
Also, the potential W of the excess mass must be a solution of Laplaces

equation, so we can write
W
1 X
n
X
Bm Ym ;
n
rn1
n0 m0
(3:108)
Multiplying by r2 gives
r2 W
1 X
n
X
Bm Ym ;
n
n0 m0
rn1
(3:109)
Differentiating with respect to r gives

1 X
n
2 X
Bm Ym ;
n 1 n n n
rW
r
r
n0 m0
(3:110)
Upon inserting this expression into (3.106) and evaluating on the surface r = R,
we have
Dg;
1 X
n
X
n 1
n0 m0
m
Bm
n Yn ;
Rn2
(3:111)
Note that there is no term for n = 1 in this sum; also, the term for n = 0 is a
constant, which may be considered part of the overall potential, but is not of
interest for the anomalies. Thus the summation begins at n = 2. On comparing
the coefcients of Ym
n ; in (3.107) and (3.111), we have
Dgm
n n 1
Bm
n
Bm
n
Rn2
Rn2 m
g
n1 n
(3:112)
(3:113)
This expression can now be substituted into (3.108) for the potential,
n1
1 X
n
X
1
R
m
WR
Dgm
(3:114)
n Yn ;
n

1
r
n2 m0
Computation of the height of the geoid is simplied by introducing a zonal
approximation. The distribution of gravity anomalies Ym
n ; is replaced by
zonal harmonics, which are essentially the zeroth-order Legendre polynomials
Pn(cos ). Effectively, the gravity anomalies at co-latitude are summed over
longitude . Compared with (3.107), we make the replacement
110
Gravity
D
gn Pn cos
n
X
m
gm
n Yn ;
(3:115)
m0
As a result the gravity anomalies on the surface of the sphere are now represented by
1
X
Dg;
D
gn Pn cos
(3:116)
n2
In order to make use of the orthogonal properties of the Legendre polynomials

(see Section 1.15), we multiply both sides by Pn(cos ) and integrate over the
surface of the unit sphere. The element of surface area on the unit sphere (radius
r = 1) is d = sin d d (Box 1.3) and the limits of integration are 0 and
0 2. The integral is
ZZ
Dg; Pn cos d
1
X
Z2 Z
n2
Pn cos 2 sin d d
D
gn
0 0
(3:117)
Let cos = x, then sin d = dx, and, on integrating with respect to , we have
ZZ
Dg; Pn cos d 2
1
X
n2
Z1
Pn x2 dx 4
D
gn
x1
D
gn
2n 1
(3:118)
The last step uses the normalization of the Legendre polynomials (Section 1.13.2).
We can now obtain D
gn from (3.118) and insert it into (3.114) to nd the
potential W of the geoid elevation. Using (3.101), we get the height of the geoid
undulation:
h
1
R X
4g n2
ZZ
S

2n 1 R n1
Pn cos Dg; d
n1 r
(3:119)
The summation under the integration reduces to a function of the angle only,
which we designate F(). With this function the height of the geoid is
ZZ
R
FDg; dS
(3:120)
h
4g
S
This is known as Stokes formula for the height of the geoid.
3.6 The geoid
111
3.6.3 Evaluation of the function F()

The function F() in Stokes formula for the height of the geoid is the value,
on the surface of the Earth, of the function F(r,) in the integrand of (3.119),
given by
Fr;
n1
1
X
2n 1 R
n2
n1
Pn cos
(3:121)
In order to simplify this expression we use the reciprocal-distance denition of

the Legendre polynomials, in the alternative form developed in Box 1.4:
1 n
1
1 1X
R
1 R cos X
Rn
Pn cos
Pn cos
2
rn1
u r n0 r
r
r
n2
(3:122)
After altering the sequence, this allows us to write

1
X
Rn
1 1 R cos
P cos
n1 n
r
u r
r2
n2
(3:123)
Expanding the sum in (3.121) gives

Fr; 2
1 n1
X
R
n2
n1
1
R
Pn cos 3
Pn cos (3:124)
n1 r
n2
1
X
The rst term on the right is simply 2R times the left-hand side of (3.123).
To evaluate the second term on the right we note that
1
r2
Z1
r

dr
1
1
rn n 1 rn1
(3:125)
This relationship can be used to change the second expression on the right of
(3.124) to
n1
Z1 X
1
1
R
3
Rn1
3
Pn cos 2
Pn cos dr
rn
n1 r
r
n2
n2
1
X
3R
r2
Z1 X
1
Rn
r
Pn cos dr
rn1
n2
r
Now we can substitute from (3.123):
112
Gravity
n1

Z1
1
R
3R
r
R cos
3
Pn cos 2
1
dr
n1 r
r
u
r
n2
r
9
8
Z1
=
<
3R
r dr
2
r R cos log r1
r
;
r :
u
1
X
(3:126)
The integration on the right must be done in several steps because the denominator u is a function of r. We must rst rewrite the equation in a more tractable
form:
Z1
r dr
Z1
r
r dr
p
r2 2rR cos R2
Z1
r
r R cos R cos
q dr
r R cos 2 R2 sin2
(3:127)
Z1
r
r dr
Z1
r
r R cos dr
q
r R cos 2 R2 sin2
Z1
r
R cos dr
q
r R cos 2 R2 sin2
(3:128)
Next, we carry out each of these integrations separately: the rst part is simply
Z
r R cos dr
q
r R cos 2 R2 sin2
q
r R cos 2 R2 sin2 u (3:129)
For the second part we make use of the following standard integration:
Z

p
a
p dy a log y y2 b2
(3:130)
y2 b2
Letting y = r R cos , a = R cos , and b = R sin in this equation, the second
integration becomes
Z

R cos dr
q R cos log r R cos :
r R cos 2 R2 sin2
p
r2 2rR cos R2
R cos logr R cos u
Combining (3.128), (3.129), and (3.131) gives
(3:131)
3.6 The geoid

Z1
r dr
u R cos logr R cos u1
r
u
113
(3:132)
Upon inserting this result into (3.126) we get

n1
1
R
3R
3
Pn cos 2 u R cos logr R cos u
n1 r
r
n2
1
X
r R cos log r1

r
(3:133)
At the limits of the integration we cannot insert r = directly. However, for very
large r,

1=2
2R cos R2
1 2R cos
ur 1
2
r R cos
r 1
r
r
2
r
(3:134)
Now we substitute this result into (3.133) to get the upper limit of the bracketed
expression:
u R cos logr R cos u r R cos log r1
R cos R cos log2r R cos R cos log r

r R cos
R cos log 2
1
r
R cos log 2 1
(3:135)
Evaluating both limits in (3.133) gives

n1
1
R
3
Pn cos
n

1
r
n2

3R
r R cos u
2 R cos log 2 1 u r R cos log
r
r
(3:136)
1
X
Now we add this result to 2R times (3.123) to get the solution of (3.124):

1 1 R cos
Fr; 2R
u r
r2

3R
r R cos u
2 R cos u r R cos log
r
2r
(3:137)
114
Gravity
P
u/2
r =R
u/2
/2
Fig. 3.7. Geometry for calculation of the geoid height at a point P from gravity
measurements. G is a point on the surface of the Earth at which gravity was
measured.
The point P at which the geoid height is to be calculated and the point G at
which a gravity measurement is known lie on the surface of the Earth, where r =
R, as in Fig. 3.7. These points form an isosceles triangle with the center of the
Earth at O, so that u 2R sin(/2) and

r R cos u 1
1 cos 2 sin
sin
sin
2r
2
2
2
2
(3:138)
On substituting into (3.137), and noting that on the surface of the sphere F(r,)
becomes F(), we have

1
1 cos
F 2
2 sin=2

1 cos log sin

sin2
3 cos 2 sin
2
2
2
(3:139)

1
1 6 sin
5 cos
F
sin=2
2

3 cos log sin

sin2
(3:140)
2
2
The function F() is plotted in Fig. 3.8. It has a singularity at = 0, which must
be excluded from the computation. F() decreases rapidly with increasing angle
for < 30 but still has an appreciable value at large angles, which means that
distant gravity measurements can have an inuence on the calculated geoid
height.
Further reading
115
150
100
F( )
50
180
0
0
30
60
90
120
polar angle, ( )
150
180
Fig. 3.8. Variation with angular distance of the function F() in Stokes formula
for the height of the geoid.
further reading
Bullen, K. E. (1975). The Earths Density. London: Chapman and Hall, 420 pp.
Groten, E. (1979). Geodesy and the Earths Gravity Field. Bonn: Dmmler, 409 pp.
Hofmann-Wellenhof, B. and Moritz, H. (2006). Physical Geodesy, 2nd edn. Vienna:
Springer, 403 pp.
Torge, W. (1989). Gravimetry. Berlin: de Gruyter, 465 pp.
4
The tides
The gravitational attractions of the Moon and Sun deform the Earth, giving rise
to the periodic uctuations of the oceanic surface known as the marine tides.
The same forces also give rise to bodily tides in the solid Earth. The Moons
mass is much smaller than that of the Sun, but the lunar tidal effect is greater
than the Suns, because the Moon is much closer to the Earth. We rst analyze
the lunar tides, then take account of the solar tidal effects.

The lunar tidal forces arise from two sources: the gravitational attraction of the
Moon on the Earth, and the joint rotation of the Earth and Moon about their
common center of mass, which is called the barycenter. The barycenter moves
around the Sun along Earths orbit.
To nd the location of the barycenter of the EarthMoon system, let the
distance between Earth and Moon be rL, the mass of the Earth E, and the mass of
the Moon M. If the barycenter B is at distance d from the center of the Earth,
then, taking moments about B,
Ed MrL d
(4:1)
and hence
d
M
rL
EM
(4:2)
The mass-ratio of Moon and Earth M/E is equal to 0.0123, and the distance
between Earth and Moon is 384,400 km, so the distance d is 4,670 km; i.e., the
barycenter lies within the Earth. The center of the Earth moves around this point
with the same rotational angular velocity L as does the Moon (Fig. 4.1), and
describes a circle with radius d.
116
117
M
R
d
O
rL
F
Fig. 4.1. Geometry of Earth and Moon in the plane of the Moons orbit. The
barycenter of the rotation is at B; L is the rotation rate of the Moon about its axis
and about the Earth; is the Earths own rotation rate, assumed normal to the
Moons orbit.
Let the EarthMoon barycenter be at B and the center of the Earth at O; let the
Earths radius be R, the Moons mass be M, and the distance between the centers
of Earth and Moon be rL, as in Fig. 4.1. At the center of the Earth, the gravitational acceleration aO towards the Moon exactly balances the centrifugal
acceleration ac = L2 d of the Earths motion around the circle with radius d, thus
GM
2L d
r2L
(4:3)
The point F on the far side of the Earth is at distance rL + R from the Moon
and R + d from the barycenter. The gravitational acceleration at F towards the
Moon is balanced by the centrifugal acceleration away from the Moon, and the
net acceleration at F towards the Moon is
aF
GM
rL R2
2L R d
Applying the binomial expansion up to fourth order gives

GM
GMR
GMR2
2L d 2L R
aF
2 3 3
r2L
r4L
rL
(4:4)
(4:5)
The term L2d is again the centrifugal acceleration of a rotation about a circle
with radius d, and is directed away from the Moon. The centrifugal acceleration
L2R is also directed away from the Moon. It corresponds to motion of the point
F about a circle with radius R. This rotation displaces F to F in Fig. 4.1 and is a
component of the Earths rotation about its own axis. It does not contribute to
the lunar tidal acceleration. Omitting this term and using the result of (4.3), we
have for the tide-raising acceleration at F
118
The tides

GMR
GMR2
aF 2 3 3
r4L
rL
(4:6)
The negative sign indicates that the net acceleration at F is away from the Moon.
This causes a tide on the far side of the Earth from the Moon.
Similar arguments can be applied to the accelerations at N on the near side of the
Earth, which is at distance rL R from the Moon and R d from the barycenter.
The centrifugal acceleration of the common rotation augments the gravitational
acceleration of the Moon, and the net acceleration aN towards the Moon is
aN
GM
rL R2
2L R d
(4:7)
The binomial expansion leads to the following equation for the acceleration at N
towards the Moon:

GM
GMR
GMR2
2 3 3
(4:8)
aN
2L d 2L R
r2L
r4L
rL
As before, the bracketed term is the lunar gravitational attraction and the
centrifugal acceleration L2d is away from the barycenter. The centrifugal
acceleration L2R is now directed towards the Moon, as expected for a rotation
about the Earths axis. The tide-raising acceleration at N is

GMR
GMR2
aN 2 3 3
(4:9)
r4L
rL
This acceleration acts towards the Moon and is responsible for the tide on the
near side of the Earth.
The balance of the tidal forces is summarized in Fig. 4.2. The centrifugal
acceleration L2d away from the Moon is present at all points of the Earth.
Earth
ac
F
aF
ac
a O ac
O
aN
Moon
Fig. 4.2. Accelerations responsible for the lunar tides on the Earth: aF, aO, and aN
are the gravitational accelerations of the Moon at the furthest point (F), center of the
Earth (O), and nearest point (N) to the Moon; ac is the constant acceleration due to
the Earths rotation about the barycenter, excluding the component of this rotation
about Earths own axis.
119
It arises from the rigid-body rotation of the Earth about the barycenter (see
Lowrie (2007) for a graphical explanation).
Comparison of (4.6) and (4.9) shows that the tidal accelerations at F and N are
unequal. As a result, the lunar tide on the near side of the Earth is higher than
that on the far side. A more detailed analysis of the tidal components and the
direction of the tide-raising forces on the Earth is obtained by examining the
tidal potential.

The calculation of the potential of the Moons gravitational attraction at a point
in the Earth (Fig. 4.3) is similar to the development of MacCullaghs formula.
Spherical polar coordinates are centered at the center of the Earth. The lunar
potential is calculated for a point P in the Earth at distance r from Earths center.
The radius to P makes an angle with the direction to the Moon, and the
geometry has rotational symmetry about this axis. The lunar potential W at P is
inversely proportional to the distance u of P from the center of the Moon. The
reciprocal-distance formula introduces the Legendre polynomials to describe
the potential:
!
1 n
X
M
M
r
W G G
1
Pn cos
(4:10)
u
rL
rL
n1
Upon expanding the rst few terms in the summation we get
W G
M
Mr cos
Mr2 P2 cos
Mr3 P3 cos
G
G
G

2
3
rL
rL
r4L
rL
(4:11)
E
r
O x Q
u
M
rL
Fig. 4.3. Calculation of the lunar potential for a point P in the Earth at distance r
from Earths center and distance u from the Moon.
120
The tides
This equation is equivalent to a sum of individual potential terms, the rst few of
which give
W W0 W1 W 2 W3
(4:12)
4.2.1 Signicance of individual terms in the lunar potential

Potential W0
W0 G
M
rL
(4:13)
This rst term in the sum is a constant, so its gradient is zero:

a0 rW0 0
(4:14)
This potential does not play a role in the tidal deformation of the Earth.
Potential W1
W1 G
Mr cos
M
G 2 x
r2L
rL
(4:15)
Here we have dened OQ in Fig. 4.3 as x = r cos . The x-axis is along the
direction to the Moon. The gradient of the potential W1 gives

W1
GM
;
0;
0
(4:16)
a1 rW1
x
r2L
This acceleration acts in the direction of positive x, i.e., towards the Moon. It is
independent of the position coordinates (r, ) and is therefore constant throughout the body of the Earth. It does not contribute to the tide-raising forces but
balances the centrifugal acceleration of the EarthMoon rotation about their
common barycenter. An equal and opposite acceleration acts on the Moon and
holds it in orbit around the Earth.
Potential W2
W2 G
Mr2 P2 cos
r3L
(4:17)
This is the potential of the main tidal deformation. It is much larger than all
following terms and is regarded as the tidal potential, except in detailed
analyses. It is proportional to the second-order Legendre polynomial
(a)
121
(c)
(b)
P2
P2 + P3
P3
Fig. 4.4. Components of the lunar potential (not to scale): (a) main symmetric
deformation proportional to a second-order Legendre polynomial; (b) next-largest
component of deformation, proportional to a third-order Legendre polynomial; and
(c) superposition of these components that gives rise to the diurnal tidal inequality.
P2(cos ) and so has rotational symmetry about the EarthMoon axis and gives
equal tides on opposite sides of the Earth (Fig. 4.4(a)). For use in later discussions, let
A G
M
r3L
(4:18)
This enables us to write the tidal potential in the more compact form
W2 Ar2 P2 cos Ar2 P2
(4:19)
Potential W3
W3 G
Mr3 P3 cos
r4L
(4:20)
This potential describes a deformation with the symmetry of the third-order

Legendre polynomial P3(cos ). It is symmetric about the EarthMoon axis but
results in a tidal elevation on Earths near side and a tidal depression on Earths
far side (Fig. 4.4(b)). Together with W2 it describes the unequal diurnal tides
explained in Section 4.1 (Fig. 4.4(c)). W3 is the second-largest term in the tidal
deformation, but is much smaller than W2, as can be shown by forming the ratio
of the two potentials:

W2 r2 P2 cos
r4L
rL
P2
80
(4:21)
W3
r3 P3 cos
r
P3
r3L
This and higher-order terms in the tidal potential are usually disregarded except
in detailed evaluation of the tidal heights.
122
The tides
4.2.2 The lunar tide-raising acceleration

The tide-raising acceleration is equal to the gradient of the tidal potential, for
which we will use the dominant potential W2. Using polar coordinates (r, ) the
acceleration has a radial component ar given by
W2
M
G 3 r3 cos2 1
r
rL
Mr
G 3 1 3 cos2
2rL
ar
(4:22)
The transverse component a is

1 W2
M 1
G 3 r
3 cos2 1
r
rL 2
Mr
G 3 3 sin2
2rL
a
(4:23)
These accelerations cause tidal displacements that are vertical (i.e., radial) on the
EarthMoon axis at = 0 and = , as well as at an angular distance = /2
from the axis. At intermediate locations the tide-raising forces have a horizontal
as well as a radial component (Fig. 4.5).
4.2.3 The solar tide-raising acceleration

The tide-raising acceleration of the Sun can be described in a similar way to that
of the Moon. The dependence of the lunar tidal amplitude on the Moons mass
+/2
to the
Moon
/2
Fig. 4.5. Direction of the lunar tidal-raising force as a function of angular distance
from the EarthMoon axis.
123
Table 4.1. Rotational and orbital parameters of the Earth and Moon (sources:
Groten, 2004; McCarthy and Petit, 2004).
Parameter
Symbol
Units
Value
Mass of Sun
Heliocentric gravitational constant
Mass of Earth
Geocentric gravitational constant
Solar mass ratio, S/E
Mass of Moon
Selenocentric gravitational constant
Lunar mass ratio, M/E
Mean geocentric radius of the
Moons orbit
Mean heliocentric radius of Earths
orbit
Present rotation rate of the Earth
Moment of inertia of Earth about its
rotation axis
Angular momentum of EarthMoon
system
Earths mean radius
Moons mean radius
S
GS
E
GE
S
M
GM
L
rL
1030 kg
1014 m3 s2
1024 kg
1020 m3 s2
105
1022 kg
1012 m3 s2
108 m
1.988 92
3.986 004 418
5.973 7
1.327 124 4
3.329 46
7.347 7
4.902 799
0.012 300 034
3.844
rS
1011 m
1.495 874 4
0
C
105 rad s1
1037 kg m2
7.292 1
8.019
1034 kg m2 s1
3.435
R
RL
106 m
106 m
6.371 000 4
1.738
and distance from the Earth is contained in the factor A dened in (4.18), which
we will call AL for this comparison. The tidal effect of the Sun depends on a
similar factor AS, in which the mass S of the Sun replaces the lunar mass M, and
the EarthSun separation rS replaces the EarthMoon separation rL. At any
given point (r, ) on the Earth the ratio AL/AS expresses the relative effects of the
lunar and solar tide-raising accelerations:

aL AL GM=r3L M rS 3
2:2
(4:24)
aS A S
S rL
GS=r3S
The masses of Sun and Moon, and their distances from the Earth are listed in
Table 4.1. The ratio of the Suns mass to the Moons mass (S/M) is about
27,000,000. The ratio of the Suns distance to the Moons distance (rS/rL) is
389. However, in comparing the lunar and solar tidal effects the distance-ratio is
cubed, which attenuates the tidal effect of the Sun more than it does that of the
Moon. Consequently, the Sun is responsible for only about one third of the
observed tide, with two thirds being caused by the Moon.
124
The tides
The lunar and solar tidal accelerations depend on the relative phases of the
Sun and Moon. When they are aligned, on the same side of the Earth (known as
conjunction) or on opposite sides (opposition), their tidal accelerations reinforce
each other and give rise to extra-high spring tides. When the directions to Sun
and Moon are perpendicular, the tidal accelerations are in quadrature and tend
to cancel each other out partially, causing extra-low neap tides.
4.3 Loves numbers and the tidal deformation

When we think of the tides, we usually mean the observed semidiurnal rise and
fall of the ocean surface. The marine tide is an elastic response of the Earth as a
whole to the lunar deforming potential. However, the tide is measured with
respect to the solid Earth, which is also deformed by the lunar gravitation. The
observed tide is the difference. The marine and bodily tides are characterized by
global elastic constants called Loves numbers.
4.3.1 Tidal height

Let the elevation of the equipotential surface due to W2 at any particular point be
H0. The uplift takes place against the acceleration of gravity, so the work done
(gH0) is equal to the change in potential. The height of the elevation is given by
H0
W2
g
(4:25)
Tidal deformations are the elastic response of the Earth to the lunar deforming
forces. The redistribution of mass gives rise to an additional potential, which
must be taken into account in analyzing the tidal potential. In 1911, A. E. H.
Love, an English mathematician, reasoned that the extra potential U2 of the
deformation should be proportional to the deforming potential W2, i.e.,
U2 kW2
(4:26)
The proportionality constant k is a global value for the elastic response of

the Earth as a whole. The added potential enhances the total tidal potential to
(1 + k)W2 and increases the vertical tidal displacement to H1 (Fig. 4.6):
H1
W 2 U2
W2
1 k
g
g
(4:27)
The solid body of the Earth is involved in the tidal response. The potential of
the solid surface displacement is also proportional to the perturbing potential
4.3 Loves numbers and tidal deformation
125
(1 + k )W2
W2
U(R + H 0)
H0
H1
H2
U(R)
Fig. 4.6. Factors involved in computation of the height of the equilibrium tide on
an elastic Earth. W2 is the lunar tidal potential and k is Loves rst number.
W2, with proportionality constant h, so the height H2 of the bodily tide can be
expressed as
H2 h
W2
g
(4:28)
On combining the results, the height H of the equilibrium tide is seen with
reference to Fig. 4.6 to be
H H1 H2 1 k h
W2
H0
g
(4:29)
where
1kh
(4:30)
Here is the ratio of the observed vertical tidal height to the theoretical height
on a rigid Earth (k = h = 0). Empirical values can be obtained from direct
measurements of tidal height. However, restrictive conditions for direct tidal
observations must be observed. The body of water must be small enough that it
has a short reaction time to the perturbing potential and there is no phase lag.
The shape and bathymetry of the body of water must not amplify the tidal
effects. For these reasons enclosed bodies of water with natural periods less than
a day have been favored in direct measurements. These give a value 0.7.
4.3.2 Tidal gravity anomaly

The lunar tidal attraction affects measurements of gravity made on Earth,
necessitating a tidal correction. The tidal gravity anomaly derives from three
potentials that affect a gravimeter set up on the Earths surface: (1) the geopotential, (2) the lunar tidal potential, and (3) the potential of the tidal
126
The tides
deformation. For the rst of these it is adequate to substitute the Earths gravitational potential, while the second potential is the lunar deforming potential W2.
As explained in the previous section, the lunar tide corresponds to a mass
redistribution within the Earth, which has a potential kW2. We need to determine
the potential of this deformation outside the Earth on the measurement surface.
Equation (4.19) shows that the deformation potential kW2 is equal to
kAr2P2(cos ). This is a solution of Laplaces equation for a space in which r
can be zero, i.e., inside the Earth. We seek a solution that is valid outside the
Earth. In general, a potential satisfying Laplaces equation may be written

B
2
(4:31)
Ar 3 P2 cos
r
We separate this potential into two potentials for different realms:
i Ar2 P2 cos ; r 5 R
B
e 3 P2 cos ; r R
r
(4:32)
The rst part, i, is valid inside the Earth, where r can be zero; the second part,
e, is valid outside the Earth, where r can be innite. The two solutions vary
differently with radial distance. At the same azimuth from the symmetry axis
they are in the ratio

e B=r3
B 1
(4:33)
i
Ar2
A r5
The potential must be continuous at the Earths surface, i.e., e = i where r = R,
thus
B
R5
A
(4:34)
5
R
e
i
r
(4:35)
and
By applying this result to the lunar tidal deformation, we nd that its potential
inside the Earth is kW2, so its potential outside the Earth is kW2(R/r)5. Thus the
potential UT of the tidal gravity anomaly, as measured outside the Earth, is
UT G
5
E
R
W2 kW2
r
r
(4:36)
127
The rst term represents the gravity potential of the undeformed Earth, the
second term that of the Moon. The third term is the gravity potential associated
with the tidal deformation. The acceleration due to gravity is the radial gradient
due to UT:
5
UT
E
R
G 2 W2 k W2
(4:37)
gr
r
r
r
r
r
Each term must be evaluated at the surface of the solid Earth. The tidal displacement
of the solid surface (4.28) raises this to the position

H0
(4:38)
r R H2 R 1 h
R
The tidal elevation H0 is very small compared with the Earths radius, so we can
make use of the binomial expansion to rst order, by writing

H0 n
H0
1h
1 nh
(4:39)
R
R
On differentiating the rst term in (4.36) and using this simplication, we get

E
E
H0 2
G 2 1 h
G 2
R
r rR1hH0 =R
R

H0
(4:40)
gR 1 2h
R
Differentiating the second term and neglecting terms of order (H0/R)2 and
higher gives
2
W2
W2 Ar P2 cos 2
r rR1hH0 =R
r
r

H0
H0
1h
2g
R
R
H0
2gR
(4:41)
R
By applying the same rules to expand the third term in (4.37) we obtain
5
R
R5
k W2
kAP2 cos
r
r
r r3
R5
3kAP2 cos 4
r
(4:42)
128
3kAP2 cos
The tides

R5
R5
H0
3kAP
cos
1

4h
2
r4 rR1hH0 =R
R4
R

W2
H0
1 4h
3k
R
R
5
R
H0
k W2
3kgR
R
r
r
On combining the results of (4.40), (4.41), and (4.44) we have

H0
H0
H0
2gR
3kgR
gr gR 1 2h
R
R
R

H0
H0
H0
gr gR 1 2
2h
3k
R
R
R
(4:43)
(4:44)
(4:45)
(4:46)
The difference between g(r) and g(R) is the gravity anomaly g caused by the
lunar tide on the deformed Earth:

H0
3
1h k
(4:47)
Dg gr gR 2gR
R
2
If the Earth were rigid (k = h = 0) and unable to deform in response to the lunar
tidal forces, there would still be a tidal gravity anomaly, corresponding to the
gravitational attraction of the Moon
Dg0 2gR
H0
R
(4:48)
Thus,

3
Dg Dg0 1 h k g0
2
(4:49)
3
1h k
2
(4:50)
where
is the ratio of the observed tidal gravity anomaly on the deformed Earth to the
theoretical value for a rigid Earth. Direct measurements give 1.15.
The simultaneous solution of (4.30) and (4.50) using the measured values for
and yields values k 0.3 and h 0.6 for the Love numbers.
129
4.3.3 Tidal deection of the vertical

The horizontal component of the tide-raising acceleration (Fig. 4.5) produces a
horizontal tidal displacement. As before, the tidal potential W2 is enhanced by
the tidal bulge to (1 + k)W2. In 1912 T. Shida introduced the number l to account
for the potential of the horizontal tide, which, analogously to Loves number h,
is proportional to the deforming potential W2. The complete potential of the
horizontal tide is then
Wh 1 k lW2
(4:51)
The effect of the horizontal tide is to deect the vertical direction. The
deforming tidal potential W2 produces horizontal components of gravity g
and g in the directions of increasing polar angle and longitude , respectively. At the Earths surface r = R these are given by
1 Wh
R
1 Wh
g
R sin
g
(4:52)
The vertical direction is deected by amounts and corresponding to the

angles formed between the horizontal components of gravity and the radial
component:
g
g
g
tan
g
tan
(4:53)
The deections of the vertical of tidal origin are obtained by combining (4.51),
(4.52), and (4.53):
1 W2
gR
1
W2
1 k l
gR sin
1 k l
(4:54)
On a rigid Earth k = l = 0 and the deections of the vertical are

1 W2
gR
1
W2
0
gR sin
0
(4:55)
130
The tides
The quantity
1kl
(4:56)
represents the ratio of the observed deection of the vertical caused by the lunar
tide on an elastic Earth to the theoretical deection for a rigid Earth. Analysis of
the tidal deection of the vertical shows that Shidas number is a very small
quantity (l 0.08).
4.3.4 Satellite-derived values for k, h, and l

Satellite observations have replaced direct measurement as a means of determining the Love and Shida numbers. The tidal deformations of the geopotential
cause slight perturbations of satellite orbits. The observed satellite orbits are
compared with what would be expected for a model Earth. The models have to
incorporate some assumptions, namely that the Earth is spherical, non-rotating,
elastic, and isotropic. The elastic constants then vary only with depth, and may
be interpreted from observations of seismic travel times. The most widely used
is the Preliminary Reference Earth Model (PREM) (Dziewonski and Anderson,
1981). The satellite-derived values of Loves number and Shidas number for
the ellipsoidal tidal deformation are k = 0.2980, h = 0.6032, and l = 0.0839.
4.4 Tidal friction and deceleration of terrestrial

and lunar rotations
The tidal bulge of the Earth is, to a rst approximation, a prolate ellipsoid with
symmetry axis aligned with the EarthMoon axis. This conguration would
give high tides at positions directly under the Moon and on the opposite side of
the Earth. However, for several reasons the reaction of the Earth to the tidal
forces is delayed. This is partly because the response of the solid Earth to forces
on the timescale of the tides is not perfectly elastic. Also, the redistribution of
water in the oceans is hindered by its viscosity, as well as by the presence of
islands, bays, and uneven bottom topography. These interactions act as a frictional resistance that delays the tidal deformation. During the delay time the
Earths own rotation carries the tidal bulge forward. By the time the bulge has
reached its peak height the axis of the tidal bulge has advanced about 2.9 past
the EarthMoon axis (Fig. 4.7).
Suppose the excess mass in the tidal bulge at Q to be concentrated at a point.
The gravitational attraction of the Moon exerts a force F2 on this part of the
bulge. Similarly, a force F1 acts on the part of the bulge at P. Because Q is closer
4.4 Tidal friction and deceleration
131
E
2.9
P
F1
F2
Fig. 4.7. Relationship of the torque that decelerates the Earths rotation to the delay
of the lunar tidal bulge due to inelastic and frictional effects.
to the Moon than P, the force F2 is stronger than F1; also, the acute angle at Q is
larger than the acute angle at P, so the component of F2 normal to the axis of the
tidal bulge is larger than that of F1. The forces cause a torque on the spinning
Earth opposite to its direction of rotation. The frictional torque slows the Earths
rotation, causing the length of the day to increase by about 2.4 seconds per
century. To maintain constant angular momentum of the closed EarthMoon
system, the rates of rotation of the Moon about its axis and about the Earth also
decrease, and the EarthMoon separation increases. The Moons rotation rate
about its axis has decreased to the extent that it is now synchronous with its
rotation rate about the Earth. As a result an observer on Earth always seems to
see the same face of the Moon.
In fact, the maximum amount of the Moons surface visible at any time from
the Earth is about 40%, because the curvature of the Moons surface means that
the periphery of the lunar globe is not visible from Earth. However, the Moons
orbit is slightly elliptical, its axis is slightly tilted to the pole to its orbit around
the Earth, and due to Earths rotation an observer views the Moon from slightly
different angles at different times of day. These effects cause irregularities in the
Moons motion as viewed from Earth called librations that over time enable
us to see 59% of the Moons surface.
4.4.1 Angular momentum of the EarthMoon system

The dimensions and rates of rotation of the Earth and the Moon, their separation, and the location of their barycenter are shown schematically in Fig. 4.1, as
viewed from above the orbital plane of the Moon; the values of these parameters
are given in Table 4.1. The focus of the orbit is at the barycenter, which is at a
distance d from the mid-point of the Earth and at rL d from the mid-point of the
Moon. Let the moment of inertia of the Earth about its rotation axis be C and that
of the Moon about its axis be CL. The rotation axes are assumed to be
perpendicular to the orbital plane.
132
The tides
The angular momentum of the system consists of contributions from (1) the
Earth about its rotation axis, C; (2) the Moon about its rotation axis, CLL;
(3) the Earth about the barycenter, Ed2L; and (4) the Moon about the barycenter, M(rL d)2L. The sum of these terms is
h C CL L Ed 2 L MrL d2 L
(4:57)
It was shown in Section 3.3.2 that the moment of inertia of a sphere is proportional to its mass times the square of its radius. The proportionality constants for
most Earth-like planets are around 0.3, so the ratio of the angular momenta of
the Earth and Moon can be estimated:

CL L M RL 2 L
1 1 1

3:3 105
(4:58)
C
E R
81 13 27
In this comparison the lunar mass ratio is M/E = 0.0123 = 1/81, the equatorial
radius of the Moon is RL 1,738 km, that of the Earth is R = 6,378 km, and the
lunar sidereal rotation rate is 27.3 days. The very small value of the ratio shows
that the angular momentum of the Moons own rotation can be ignored in this
discussion.
From (4.2) the distance of the center of the Moon from the barycenter is
rL d
E
rL
EM
(4:59)
By inserting this and (4.2) into (4.57), we get the angular momentum of the
EarthMoon system:

2

2
M
E
2
2
h C EL rL
ML rL
(4:60)
EM
EM

EM
(4:61)
h C L r2L
EM
4.4.2 Slowing of terrestrial and lunar rotations

Equation (4.61) has implications for the rates of rotation of the Earth and Moon.
The gravitational attraction of the Earth on the Moon exactly balances the
centrifugal acceleration of the Moons orbital acceleration about the barycenter.
This provides the additional equation
GE
2L rL d
r2L
(4:62)

and, on substituting for (rL d) from (4.59), this becomes

GE
E
2
L rL
EM
r2L
133
(4:63)
and thus
GE M 2L r3L
(4:64)
This is, in fact, Keplers Third Law for the EarthMoon system. Now we square
both sides, getting
G2 E M2 4L r6L
(4:65)
Next we form the cube of (4.61),

h C3 3L r6L
EM
EM
3
(4:66)
Comparing (4.65) and (4.66) gives

h C3
G2 E M2 E3 M3
L
E M 3
(4:67)
Simplifying so that only the constant terms G, E, and M are on the right of the
equation, we have
L h C3
G2 E3 M3
EM
(4:68)
The lunar tidal friction acts as a brake on the Earths rotation, slowing it down
and increasing the length of the day by about 2.4 ms per century. The total
angular momentum of the system, h, is constant, as is the right-hand side of the
equation. Thus, if on the left-hand side of the equation is decreasing, the lunar
rotation L must also be decreasing. At the same time, in order to maintain
(4.64), the distance between the Earth and Moon, rL, must be increasing. At
present the increase amounts to about 3.7 cm per year.
4.4.3 Development of the EarthMoon separation

The tidal friction exerted by the Earth on the Moon has slowed the Moons
rotation until it is now synchronous with its orbital rotation around the Earth.
Eventually the lunar tidal friction will slow the Earths rotation so that it is also
synchronous with the Moons rotation. At that stage a terrestrial day, a lunar day,
and the month will all have the same length. Meanwhile the Moon will continue
134
The tides
to move further from the Earth. How far will the Moon be from the Earth when
the rotations are synchronous? We can answer this question by setting L = in
(4.68). For convenience we also normalize the rotation in terms of 0, the
present rate of rotation of the Earth:

h
3
G2 E 3 M3
(4:69)

3 4
0 C0 0
C 0 E M
Let the normalized rotation rate be n = /0 and the normalized angular
momentum be a = h/(C0), and let the expression on the right-hand side of the
equation be b. Both a and b are constants, so we have to solve an equation with
the form
na n3 b
(4:70)
This fourth-order equation in n has four roots, of which two are imaginary
and of no interest, and two are real. The real roots, obtained numerically or
graphically as in Box 4.1, are n = 0.213 and n = 4.92. The rst solution
Box 4.1. Synchronous rotation of Earth and Moon

Equation (4.67) for the synchronous rotation of the Earth about its axis, the
Moon about the Earth, and the Moon about its own axis can be written as
na n3 b
(1)
in which the normalized rotation rate is n = /0, and the constants a and b
are
a
h
C0
G2 E3 M3
M
C3 40 E
(2)
(3)
The numerical values of a and b are found by inserting the currently

accepted values of the relevant parameters (Table 4.1) into the dening
equation. This yields a = 5.8742 and b = 4.272. The equation becomes
n5:8742 n3 4:272
(4)
135
The real roots of this fourth-order equation can be found by evaluating

numerically the functions
F1 n 5:8742 n3
(5)
4:272
n
(6)
F2 n
and nding the values of n that give F1(n) = F2(n). Alternatively, the
functions can be plotted as in Fig. B4.1 and the points of intersection of the
curves determined.
The equation has only two real roots, which are n = 0.0213 and n = 4.92.
300
= 0.213
0
F1(n), F2(n)
200
F2(n)
100
F1 (n)
= 4.92
0
0.02 0.05
0
0.01
0.10
1.0
relative rotation rate, n = /0
10
Fig. B4.1. Graphical solution for , the synchronous rotation rate of the Earth
and Moon; 0 is the present rotation rate of the Earth.
corresponds to a rotation period of 47 days and an EarthMoon separation of 87

times Earths radius (rL = 87R). The present distance between the centers of the
Earth and Moon is 60 times Earths radius, so this solution gives the conditions
for a future synchroneity of the rotations. The second root gives a rotation
period of 4.9 hr and a lunar distance of 2.3 times Earths radius (rL = 2.3R),
corresponding to an earlier time in the Moons history. However, this solution is
unrealistic because it places the Moon within the Roche limit of the Earth, at
which position the Earths gravity would tear the Moon apart.
136
The tides
further reading
Lambeck, K. (1988). Geophysical Geodesy: The Slow Deformations of the Earth.

Oxford: Clarendon Press, 718 pp.
Melchior, P. (1966). The Earth Tides. Oxford: Pergamon Press, 458 pp.
5
Earths rotation
The Earth is not rigid and its rotation causes it to deform, attening at the poles
and bulging at the equator. The gravitational attractions of Sun and Moon on the
equatorial bulge result in torques on the Earth, which cause additional motions of
the rotation axis, known as precession and nutation. These motions occur relative
to a coordinate system xed in space, for example in the solar system. The
rotation axis is inclined to the pole to the ecliptic plane at a mean angle of
23.425; this angle is the obliquity of the axis. Precession is a very slow motion
of the tilted rotation axis around the pole to the ecliptic, with a period of 25,720 yr.
The nutation is superposed on this motion and consists of slight uctuations in the
rate of precession as well as in the obliquity.
The other planets also affect the Earths rotation, causing small but signicant
cyclical changes on a very long timescale. These are observable directly by
precise measurement of the position of the rotation axis using very-longbaseline interferometry (VLBI). The uctuations inuence the intensity of
solar radiation incident on the Earth and produce cyclical climatic effects that
are evident in sedimentary processes, where they are known as the
Milankovitch (or Milankovi) cycles. They correspond to retrograde precession
of the rotation axis (period ~ 26 kyr), changes in the angle of obliquity (period ~
41 kyr), prograde precession of Earths elliptical orbit (period ~ 100 kyr), and
variation of the ellipticity of the orbit (period ~ 100 kyr).
In addition to these phenomena, the Earths rotation is affected on a shorter
timescale by the planets mass distribution. When the instantaneous rotation
axis deviates from the axis of gure determined by the long-term rotation, a
cyclical motion of the rotation axis about its mean position arises. This is known
as the Chandler wobble. In contrast to the precession and nutation resulting
from external forces, the wobble results from the imbalance in mass distribution
with respect to the instantaneous rotation axis. It takes place in the Earths
coordinate system and is evident as small variations in latitude with a period of
435 days.
137
138
Earths rotation
(b)
(a)
r sin
A
Fig. 5.1. Rotation of a displacement vector r inclined at angle to the rotation axis.

The displacement of a body on the rotating Earth may be considered to have two
parts. The rst is a simple displacement relative to coordinate axes dened for
the Earth. The second arises from the rotation of the Earth relative to a xed set
of axes; these might be dened, for example, relative to the solar system.
5.1.1 Velocity
Consider an orthogonal spherical coordinate system with unit vectors (er, e, e).
Let r be a displacement vector that makes an angle with the axis of rotation
(Fig. 5.1(a)). If the Earth rotates about this axis with angular velocity relative to
xed axes, then, in an innitesimal time t, the vector r rotates through an angle
. This produces a rotational displacement r1 = (r sin ) e (Fig. 5.1(b)).
If, in the same time, r undergoes a local incremental change r, the total displacement relative to the xed coordinate system is
r r r1 r r sin e
(5:1)
Dividing throughout by the time increment t gives the relationship between a

velocity relative to the xed axes and the velocity in the rotating system:

r r

(5:2)
r sin
e
t t
t

dr
r
r
lim
r sin e
(5:3)
t!0
dt
t
t
The last term in (5.3) is equal to ( r), thus
139
dr r
w r
dt t
(5:4)
vf v w r
(5:5)
Thus, we have
where vf is the velocity relative to the xed axes, v is the velocity in the rotating
system, and ( r) is an additional velocity component due to the rotation of
the moving set of axes.
5.1.2 Acceleration
Equation (5.4) can be rewritten as
d
r
dt
w r
t
(5:6)
The expression in parentheses may be regarded as an operator acting on the

vector r. This allows us to express the acceleration as

d2
d dr
r
r
w
wr
(5:7)
dt2
dt dt
t
t
Evaluating the right-hand side step-by-step gives

d 2 r 2 r
r
2 w r w
w w r
dt2
t
t
t
(5:8)
If we assume that the angular velocity of the rotating system is constant, then

d 2 r 2 r
r
22 w
w w r
(5:9)
dt2
t
t
On rearranging terms, we get
2 r d 2 r
2 w w r 2w v
t2
dt
(5:10)
ar af aR aC
(5:11)
or
where ar = 2r/t2 is the acceleration experienced by a moving object in the

rotating system, and af = d2r/dt2 is the acceleration in the xed coordinate
system. The second acceleration on the right-hand side is aR = ( r r).
140
Earths rotation
(a)
(b)
eN
vN
eN
v
v = vN e N + vE eE
eE
DvN
eD
vE
DvE
eE
aC = 2 D( vE e N + vN e E)
Fig. 5.2. (a) Directions of the north (eN), east (eE), and vertically downward (eD)
unit vectors of orthogonal reference axes, and the horizontal velocity v, in relation
to the rotation vector . (b) Vectors in the horizontal plane, showing that the
Coriolis acceleration aC acts perpendicularly to the right of the direction of
motion v in the northern hemisphere.
Inspection of the direction and magnitude of aR shows that it is the familiar

centrifugal acceleration. The nal acceleration is
aC 2w v 2v w
(5:12)
aC is called the Coriolis acceleration; it has important consequences for moving

objects in a rotating framework.

Suppose that a body is moving with horizontal velocity v on the surface of the
Earth, which is rotating with angular velocity about the rotation axis (Fig. 5.2(a)).
The unit vectors along orthogonal axes parallel to the north, east, and vertically
downward directions at the position of the object are (eN, eE, eD) and dene a local
coordinate system. The horizontal velocity of the body has components (vN, vE, 0)
parallel to these axes. The angular velocity of rotation has a constant direction.
Transposed to the position of the moving body, it acts normal to the easterly
component and has a positive northerly component at all latitudes. However,
because eD is dened to be positive downward, the vertical component is negative
(upward) in the northern hemisphere and positive (downward) in the southern
hemisphere. Thus the components of the rotation vector in the northern hemisphere
are (N, 0, D). The velocity and rotation vectors are
v vN eN vE eE
(5:13)

w N eN D eD
141
(5:14)
Equation (5.12) can be evaluated by writing the vector cross product as a

determinant:

eN eE
eD

aC 2v w 2 vN vE
(5:15)
0
N 0 D
On evaluating the determinant, we get
aC 2vE D eN vN D eE vE N eD
(5:16)
In a geographic frame, the Coriolis acceleration has a component parallel to the

vertical axis eD and a component in the horizontal plane dened by eN and eE.
5.2.1 Vertical component: the Etvs effect

The last term in (5.16) describes the vertical component of the Coriolis
acceleration:
aE 2vE N eD
(5:17)
The formation of a vertical acceleration through the interaction of a horizontal

eastwest velocity with Earths rotation is known as the Etvs effect. It
modies the value of gravity measured from a moving platform, such as a
vehicle, ship, or aircraft. If the body has an eastward velocity component (i.e.,
vE is positive), aE acts in the direction of eD, i.e., upwards. Conversely, if
the velocity has a westward component, the Etvs acceleration is downward.
Its magnitude is dependent on the velocity and on the latitude through the
value of N, which is maximum at the equator and zero at the poles. For
example, in a ship moving westwards at 7 knots (13 km hr1) at latitude
30 N, the Etvs acceleration increases the measured gravity by about 45
mgal. This greatly exceeds the measurement sensitivity in a marine gravity
survey and necessitates a so-called Etvs correction to gravity measurements.
5.2.2 Horizontal component: the Coriolis effect

The rst two terms in (5.16) describe the horizontal component of the Coriolis
acceleration:
aH 2D vE eN vN eE
(5:18)
142
Earths rotation
Its direction is normal to the velocity of the moving body, as can be veried by
taking the scalar product of aH and v, which is zero:
aH v 2D vE eN vN eE vN eN vE eE 0
(5:19)
The angular velocity of rotation has a constant direction. Its vertical component
D is negative (upward) in the northern hemisphere and positive (downward) in
the southern hemisphere. As a result, the Coriolis acceleration acts to the right of
the direction of motion in the northern hemisphere, as can be seen by inspection
of Fig. 5.2(b); it acts to the left in the southern hemisphere. The Coriolis effect
causes deection of the motion of bodies, such as air masses, moving across the
surface of the Earth. In meteorology it gives rise to cyclonic and anticyclonic
wind systems.

The main components of the precession and nutation result from the gravitational torques of the Sun and Moon on the Earth. In addition, the Suns
attraction causes the Moons orbit to precess around the equator with a period
of 18.6 yr. This motion results in a contribution to the nutation of Earths
rotation axis, which will be considered later. We rst evaluate the precession
and nutation caused by the solar torque, then extend the analysis to the lunar
torque.
5.3.1 Effects of the torque due to the Suns attraction

As the Earth moves around its orbit it experiences a variable torque due to the
gravitational attraction of the Sun (Fig. 5.3(a)). For convenience assume that the
Sun is at the center of the elliptical orbit. The tilt of the rotation axis inclines
the northern hemisphere towards the Sun at the summer solstice and away from
it at the winter solstice. Consider the Suns attraction at the summer solstice
(Fig. 5.3(b)). The gravitational attraction F1 on the part of the equatorial bulge
closest to the Sun is greater than the attraction F2 on the opposite side. These
forces are not collinear: the center of action of F1 is above the ecliptic, whereas
that of F2 is below the ecliptic. The resulting torque T tries to reduce the tilt of the
rotation axis. This causes the angular momentum vector to precess (Fig. 5.4(a)).
The torque causes an incremental change in angular momentum, h, so that
the angular momentum vector is displaced (Fig. 5.4(b)). Successive positions of
the angular momentum vector lie on the surface of a cone whose axis is the pole
to the ecliptic. The gravitational torque acts about an axis parallel to the line of
equinox
T= 0
(a)
143
(b)
h
T
T
Sun
winter
summer
F2
to Sun
trace of
ecliptic
F1 > F2
T
summer
Fig. 5.3. (a) A torque of variable magnitude but constant direction is exerted by the
Sun on the spinning Earth as it moves around its orbit. (b) A section through the
inclined Earth in a plane normal to the ecliptic that includes the direction to the Sun,
showing how the solar torque arises from unequal gravitational attraction on the
equatorial bulge.
(a)
(b)
pole to
ecliptic
nutation
ssion
prece
h
3
2
1
Earth's
rotation
axis
successive
angular
momentum
vectors
T
F2
F1
or
t
equa
to
the
Sun
successive
positions of
line of
equinoxes
4
3
2
to the
Sun
1
torque
Fig. 5.4. (a) Precessional motion of the rotation axis about the pole to the ecliptic,
on which nutation of the axis is superposed. (b) Incremental displacements of the
angular momentum vector dene the surface of a cone whose axis is the pole to the
ecliptic. After Lowrie (2007).
equinoxes, which in turn is perpendicular to the rotation axis. As the angular

momentum vector creeps over the surface of the cone, the line of equinoxes
perpendicular to it moves around the ecliptic plane. The sense of motion is
retrograde, opposite to the direction of the Earths rotation.
Let x-, y-, and z-axes be dened as the orthogonal reference axes of the
Earths gure, with the z-axis parallel to the Earths spin and the xy plane
coincident with the equatorial plane (Fig. 5.5(a)). The spin vector is
s sez
(5:20)
144
Earths rotation
ez 0
ez
ez
(a)
ez 0
ey
(b)
z s
Earth
ex
Sun
equator
ey
ey 0
ex 0
equator
line of
equinoxes
ex 0
ecliptic
ey 1
ey 0
ex
Fig. 5.5. (a) Denition of orthogonal reference axes relative to the Earth (ex, ey, ez)
and to the ecliptic (ex0, ey0, ez0). (b) Rotations involved in the transformation of
vector components from Earth coordinates to the Suns coordinate system.
Now suppose that the reference axes are able to rotate with angular velocity
relative to a xed set of coordinates, so that it has components (x, y, z) along
the respective reference axes of the Earth (Fig. 5.5(b)). Thus
w x ex y ey z ez
(5:21)
Let the principal moments of inertia of the Earth about the reference axes be
A, B, and C, respectively. The Earths angular momentum is
h hx ex hy ey hz ez
(5:22)
The components (hx, hy, hz) are given by

hx Ax
hy By
hz C s z
(5:23)
where hz includes both the Earths own spin and the z-component of the rotating
coordinate system. The angular momentum is
h Ax ex By ey Cz sez
(5:24)
A torque T with components (L, M, N) along the respective reference axes

causes a change of angular momentum given by
T
h h w h
dt
t
(5:25)
The operator dened in (5.6) is used here to

Earths rotation.
Using the determinant of components

ex ey

w h x y
hx hy
145
take into account the effect of

ez
z
hz
(5:26)
we obtain for the cross product

w h y hz z hy ex z hx x hz ey x hy y hx ez (5:27)
Each of the x-, y-, and z-components of the motion described by (5.23) may now
be analyzed in turn. For example, the x-component is
L

hx
y hz z hy
t
(5:28)
For succinctness we use the short form _ x x =t in the following timedifferentiations. We assume that the principal moments of inertia (A, B, C) are
constant and that the changes in angular momentum result only from changes in
angular rotation. Using the expressions in (5.23) for the components of angular
momentum (hx, hy, hz), we get
L A_ x Cy z s By z
(5:29)
The equations of motion for the y- and z-components of the torque, M and N, are
obtained in similar fashion and give the following:
M B_ y Cx z s Az x
(5:30)
N C_ z s_ B Ax y
(5:31)
For the spheroidal Earth, the moments of inertia about all axes in the equatorial
plane are equal, thus A = B and (5.31) becomes
N C_ z s_
(5:32)
As explained above, the gravitational torque of the Sun acts parallel to the
line of equinoxes, and thus normal to the rotation axis. It has no component
along the rotation axis, i.e., N = 0. Thus,
_ z s_ 0
(5:33)
z s
(5:34)
and
146
Earths rotation
where is a constant rate of rotation. The remaining equations of motion can

now be written as
L A_ x Cy Ay z
(5:35)
M A_ y Cx Az x
(5:36)
The torque components L and M result from the gravitational attraction of the
Sun on the spheroidal Earth (Fig. 5.3) and vary with the orbital position of the
Earth, which is dened relative to the xed axes. The angular velocity components are dened relative to Earths reference axes, which are free to rotate. To
solve the equations of motion it is necessary to establish a relationship between
the xed and rotating coordinate systems. The Suns torque on the Earth must be
derived and its components L and M along the rotating axes resolved.
5.3.2 Comparison of vectors in the coordinate systems

of Earth and Sun
Let (ex0, ey0, ez0) be the orthogonal unit vectors of a solar coordinate system,
dened so that ez0 is the pole to the ecliptic, ex0 is parallel to the minor axis, and
ey0 is parallel to the major axis of Earths elliptical orbit. Let (ex, ey, ez) be
orthogonal unit vectors for the rotating Earth, such that ez is parallel to the spin
axis and ex lies along the intersection of the equatorial plane with the ecliptic,
i.e., the line of equinoxes (Fig. 5.5(a)). The angle between ez and ez0 is the
obliquity of the rotation axis, and the angle between ex and ex0 denes the
position of the line of equinoxes in the ecliptic plane.
The transformation of vector components from the Earths coordinates to the
Suns coordinate system can be achieved with two rotations (Fig. 5.5(b)). The
rst is a rotation of about the x-axis. This aligns the rotation axis with the pole
to the ecliptic, and brings ey into an intermediate orientation ey1 in the ecliptic.
The x-components of a vector are unchanged by this rotation. On comparing
vector components we see that
ey1 ey cos ez sin
ez0 ey sin ez cos
(5:37)
A second rotation of about the pole to the ecliptic aligns ex with ex0 and ey
with ey0. The ez0-components are not changed by this rotation, which gives the
equations
ex0 ex cos ey1 sin
ey0 ex sin ey1 cos
(5:38)
ez
(b) Earth
(a)
147
ey 0
d
ex 0
Su n
Fig. 5.6. (a) Denition of the angle between the Earths rotation axis ez and the
radial direction d to the Sun. (b) Denition of the angular orbital position of the
Earth and the reference axes ex0 and ey0 in the ecliptic plane.
Substituting from (5.37) into (5.38) gives

ex0 ex cos ey cos ez sin sin

ey0 ex sin ey cos ez sin cos
(5:39)
(5:40)
After arranging terms, we get a set of equations relating the unit vectors (ex0, ey0,
ez0) in the xed coordinate system to the unit vectors (ex, ey, ez) in the rotating
coordinate system:
ex0 ex cos ey cos sin ez sin sin
ey0 ex sin ey cos cos ez sin cos
(5:41)
ez0 ey sin ez cos
5.3.3 Computation of the Suns torque on the Earth

The Suns torque can be computed from the potential energy of the EarthSun
pair. Let the angle between the Earths rotation axis and the radial direction to
the Sun at distance d be (Fig. 5.6(a)). The gravitational potential UG of the
Earth at the Suns location is obtained from the MacCullagh formula
(Section 2.5),
UG G
M
CA
P2 cos
G
d
d3
(5:42)
Multiplying by the mass S of the Sun gives the potential energy UPE of the
gravitational interaction of Sun and Earth:
UPE G
ES
C AS
P2 cos
G
d
d3
(5:43)
148
Earths rotation
The gravitational torque of the Sun on the Earth is obtained by differentiating

the potential energy with respect to the angle ,
T
UPE
(5:44)
The rst term in (5.43) does not depend on , so

C AS
C AS 3 cos2 1
P2 cos G
(5:45)
T G
d3
d3
2
T 3G
C AS
cos sin
d3
(5:46)
The Suns torque on the equatorial bulge depends on the difference between
the principal moments of inertia (C A), which would not exist for a spherical
Earth. The torque depends on the angle between the rotation axis ez and the
radius vector d from the Earth to the Sun, which varies as the Earth moves
around its orbit. From Fig. 5.6(a) the following relationships are obtained:
d ez d cos
(5:47)
d ez d sin
(5:48)
The cross product (d ez) gives the correct sense of the torque of the Sun on the
Earth. We can now substitute for sin and cos in (5.46), obtaining
T 3G
C AS
d ez d ez
d5
(5:49)
5.3.4 Equations of solar-induced precession and nutation

Referring to Fig. 5.6(b), the radial vector d can be written
d d cos ex0 d sin ey0
(5:50)
If the Earth orbits the Sun with constant angular velocity p, then in time t the
radius vector moves through an angle = pt. Therefore

(5:51)
d d ex0 cos pt ey0 sin pt
The scalar product of d and ez is

d ez d cos ptex0 ez d sin pt ey0 ez
(5:52)
149
We now substitute the expressions for ex0 and ey0 from (5.39) and (5.40),
respectively, keeping in mind the following orthogonal relations between the
unit vectors:

ex ez ey ez 0;
ez ez 1
(5:53)
This gives

ex0 ez ex cos ey cos sin ez sin sin ez
sin sin

ey0 ez ex sin ey cos cos ez sin cos ez
sin cos
(5:54)
(5:55)
Inserting (5.54) and (5.55) into (5.52) gives

d ez d cos ptsin sin d sin ptsin cos
d sin sinpt
(5:56)
In order to determine the cross product

d ez d cos ptex0 ez d sin pt ey0 ez
(5:57)
we again make use of the orthogonality of the unit vectors:

ex ez ey ;
ey ez ex ; ez ez 0
(5:58)
By again substituting for ex0 and ey0 from (5.39) and (5.40) we get

ex0 ez ex cos ey cos sin ez sin sin ez

ex ez cos ey ez cos sin
ey cos ex cos sin

ey0 ez ex sin ey cos cos ez sin cos ez

ex ez sin ey ez cos cos
ey sin ex cos cos
and, on inserting these expressions into (5.57), we have

d ez d cos pt ey cos ex cos sin

d sin pt ey sin ex cos cos
(5:59)
(5:60)
(5:61)
This equation can be simplied further by making use of trigonometric

identities for the sine and cosine of the difference of two angles:
150
Earths rotation
d ez d cos sin ptcos cos ptsin ex

dcos ptcos sin ptsin ey

d ez d cos sinpt ex cospt ey
(5:62)
(5:63)
By combining the results for the scalar product (5.56) and cross product (5.63)
we get the nal expressions for the torque components L and M along the x- and
y-axes, respectively:
C AS 2
d sin cos sin2 pt
d5
C AS
3G
sin cos 1 cos2pt
2d 3
L 3G
C AS 2
d sin sinpt cospt
d5
C AS
sin sin2pt
3G
2d 3
(5:64)
M 3G
(5:65)
Upon inserting the equations for L and M into (5.35) and (5.36) we get
A_ x Cy Ay z 3G
A_ y Cx Ax z 3G
C AS
sin cos 1 cos2pt
2d 3
(5:66)
C AS
sin sin2pt
2d 3
(5:67)
5.3.5 Simplication of the equations of motion

The equations describe a forced harmonic motion, with the driving force
dependent on the sine and cosine of 2(pt ). It is easier to proceed with the
solution of the equations if we simplify them by comparing the magnitudes of
the terms on the left-hand side of each equation. This allows us to neglect terms
that are unimportant to rst order. Let the sine and cosine functions be represented by the real and imaginary parts of a complex number (Section 1.2) with
phase equal to 2(pt ); we can write it as exp[2i(pt )]. Each equation then
has the form
a_ b c2 exp2i pt
(5:68)
in which stands for either of the angular velocities x and y. The driving force
on the right-hand side of the equation is periodic with angular frequency 2p.
151
The solution of the equation must also be periodic, so we may expect that

j_ x j 2px and _ y 2py .
The rotation of the Earth about its axis has period 2/ = 1 day; the angular
velocity p of the Earth about the Sun has period 365 days, so = 365p. The
angular velocity components of the rotating coordinate system are much smaller
than the daily rotation rate of the Earth: x ~ y . On comparing the rst
and second terms on the left of (5.67) and (5.68) we see that the rst term can be
neglected because
j_ j 2p
(5:69)
Similarly, the magnitude of the third term may be neglected compared with the
second term because

y z 2
(5:70)
Thus Cx and Cy are the dominant terms on the left of the equations and
the other terms on the left may be neglected by comparison. This leads to
simpler equations of motion, such as
C AS
sin sin2pt
2d3
(5:71)

3GS C A
x
sin sin2pt
2d3
C
(5:72)

3GS C A
sin cos 1 cos2pt
2d3
C
(5:73)
Cx 3G
from which
Similarly,
y
The angular velocities of the rotating coordinate axes are related to the rates
of change with time of the angles and . It is evident by reference to Fig. 5.5(b)
that
x
;
t
y sin
;
t
z cos
(5:74)
The same parameters appear on the right of each equation of motion. We can
substitute

3GS C A
FS
(5:75)
2d3
C
152
Earths rotation
Using these relationships, the equations of motion become
FS sin sin2pt
t
(5:76)
FS cos FS cos cos2pt

t
(5:77)
5.3.6 Precession and nutation induced by the Sun

The angle denes the position of the line of equinoxes in the ecliptic plane.
Equation (5.77) shows that the rate of change of consists of two parts. The rst
term, FS cos , describes a motion of the x-axis the line of equinoxes around
the ecliptic plane, at a constant rate. The rotation axis (z-axis) moves accordingly, staying orthogonal to the x-axis. The rotation axis thus moves across the
surface of a cone whose axis is the pole to the ecliptic (Fig. 5.4(a)). This motion
is the precession of the rotation axis. The mean precession rate is 50.385 arcsec
per year, corresponding to a period of 25,720 yr. The term FS is negative (5.75),
so the precession is retrograde, i.e., the motion is in the opposite sense to Earths
rotation. The parameters that dene FS have constant values, all of which are
known except the moments of inertia, A and C. The ratio H dened by
H
CA
C
(5:78)
is the dynamic ellipticity of the Earth. It can be calculated from the observed rate
of precession and has the value 3.273 787 5 103 (1/305.457).
The term on the right of (5.76) describes a periodic uctuation in the obliquity
. This nodding motion is called the nutation in obliquity of the rotation axis.
A similar uctuation of the angle is shown by the second term on the right of
(5.77). This uctuation occurs in the plane of the ecliptic and is known as the
nutation in longitude. These forced nutations each have the same frequency, 2p,
corresponding to a period of half a year (183 days). They are called the semiannual nutations. Their amplitudes are very small and unequal, amounting to
only a few seconds of arc. Using for convenience the short form for timedifferentiations, we can write
_
sin2pt
FS sin
_ FS cos
cos2pt
FS cos
(5:79)
(5:80)
153
Squaring both sides and summing gives

_ FS cos 2
FS cos 2
2
_
FS sin 2
(5:81)
The equation of an ellipse with semi-major axes a and b is

x2 y2
1
a2 b2
(5:82)
On comparing (5.79) and (5.80) we see that the two forced nutations
combine to produce an elliptical motion of the rotation axis about its mean
position, superposed on the steady motion around the precession cone
(Fig. 5.4(a)).
5.3.7 Precession and nutation induced by the Moon

The Earths nearest neighbor, the Moon, is much smaller than the distant Sun,
but its gravitational effect also causes both precession and nutation of the
Earths rotation axis. The combined effects of Sun and Moon are known as
the lunisolar precession and nutation. The effects of the attraction of the Moons
mass M on Earths equatorial bulge are analyzed in the same way as the solar
torque, and we get equations that have the same form as (5.76) and (5.77). Using
subscript L to identify the lunar parameters, we get
_L FL sin L sin2pL t L
(5:83)
_ L FL cos L FL cos L cos2pL t L
(5:84)
Here the angles L and L locate the rotation axis relative to the Moons orbit,
and pL is the angular velocity of the Moon around the Earth. This gives a
nutation component with a period of half a month. Because the Moons orbit is
only slightly inclined to the ecliptic, the solar and lunar effects can be added as
scalars.
The constant FL depends on the mass M of the Moon and its distance dL from
the Earth:

3GM C A
FL
(5:85)
2dL3
C
It is interesting to compare this term for the lunar effect with the corresponding
term for the Suns inuence on the precession (using subscript S for the
respective solar parameters):
154
Earths rotation

3GM C A
3
FL
M
dS
2dL3
C

C

A
FS
d
S
L
3GS3
C
2dS
(5:86)
The masses of the Sun and Moon and their distances from the Earth are given in
Table 4.1. Inserting the appropriate values gives
3
FL
M
dS
2:2
(5:87)
S
FS
dL
The ratio is the same as that involved in comparing the tide-raising accelerations
of the Sun and Moon (Section 4.2.3), and the explanation of the result is the
same. The mass of the Moon is much smaller than that of the Sun, but the ratio
of their inuences depends on the cube of the distance ratio, so the Moon
accounts for about two thirds of the combined lunisolar precession and nutation,
and the Sun about one third.
5.3.8 Nutation due to precession of the Moons orbit

As a result of tidal friction the Moons spin rate about its own axis is the same as
its orbital angular velocity pL about the Earth. If the moment of inertia of the
Moon about its spin axis is IL, its mass M and radius RL (1,738 km), the spin
angular momentum is
hL IL pL kL MR2L pL
(5:88)
For the Moon kL is equal to 0.394. For a uniform sphere kL = 0.4. A smaller
value indicates that density increases with depth, e.g., for the Earth kE = 0.3308.
The orbital angular momentum is
hO Mr2L pL
where rL is the radius of the Moons orbit (384,400 km)
On comparing the spin and orbital angular momenta, we have
2
hL kL MR2L pL
RL
kL
hO
rL
Mr2L pL
(5:89)
(5:90)
Upon inserting appropriate values, it is evident that the Moons spin angular
momentum is much less than its orbital angular momentum.
The Moons orbit and its angular momentum vector are inclined at a small
angle (5.145) to the ecliptic plane. The Suns attraction results in a torque that
attempts to turn the inclined angular momentum vector normal to the ecliptic.
155
Similarly to the effect of the Sun on Earths angular momentum (Fig. 5.4(b)), the
solar torque causes the Moons orbit to precess about the pole to the ecliptic. The
effective inclination of the Moons orbit to the Earths rotation axis varies between
18.28 and 28.58 (i.e., 23.43 5.15) with a period of 18.6 yr, which results in a
corresponding component in the nutation of Earths rotation axis. The precession
of the Moons orbit causes the largest part of the nutation, with amplitudes of 9.2
arcsec in obliquity and 17.3 arcsec in longitude. The semi-annual nutation has
amplitudes of only 1.3 arcsec in longitude and 0.6 arcsec in obliquity.

External forces on the spinning Earth give rise to the forced nutation and
precession of the rotation axis. These were described by allowing the reference
axes of the Earth to rotate relative to the spin axis. The long-term average
rotation of the Earth gives it a spheroidal shape about the axis of gure. If a
symmetric body spins freely about its axis of symmetry, its orientation in space
remains xed. However, if some event displaces the spin axis from its mean
direction, the Earths instantaneous rotation is no longer about its axis of
symmetry. This results in a motion called the free nutation. It was predicted in
the eighteenth century by the Swiss mathematician Leonhard Euler, and is also
called Eulerian nutation. The use of the term nutation is an unfortunate misnomer as the motion does not involve nodding of the spin axis. In Eulerian
nutation the instantaneous rotation axis moves around the surface of a cone
whose axis is the axis of symmetry.
Let the reference axes be dened relative to the gure of the Earth so that the
z-axis agrees with the axis of symmetry and the x- and y-axes lie in the equatorial
plane (Fig. 5.7). The reference axes rotate along with the Earth, so the angular
z
( , , )
y
y
Fig. 5.7. Angular velocity components (x, y, ) and direction cosines (, , ) of

the displaced instantaneous rotation axis.
156
Earths rotation
velocity z about the z-axis is the same as the Earths spin . A displacement of
the instantaneous spin vector is represented by angular velocities x and y
about the equatorial axes. The instantaneous rotation vector is then
w x ex y ey z ez
(5:91)
Using as before A, B, and C for the principal moments of inertia about the x-, y-,
and z-axes, respectively, the angular momentum is given by
h Ax ex By ey Cz ez
(5:92)
In contrast to the forced motion of the rotation axis caused by solar and lunar
attraction, the motion of the rotation axis is in this case free of external torques.
Thus
T
d h h
w h 0
dt
t
(5:93)
Assuming that the Earth rotates as a rigid body, the equations of motion for each
of the reference axes can be developed as in the case of forced nutation
(see Section 5.3.1):
A_ x C By z 0
B_ y A Cx z 0
C_ z B Ax y 0
(5:94)
The symmetry of the Earths gure implies that the equatorial moments of
inertia are equal, A = B:
A_ x C Ay z 0
(5:95)
A_ y C Ax z 0
(5:96)
C_ z 0
(5:97)
The last equation requires that the angular velocity about the z-axis is constant:
z
Rewriting (5.95) and (5.96) gives

CA
y 0
_ x
A

CA
_ y
x 0
A
(5:98)
(5:99)
(5:100)
Differentiating (5.99) with respect to time t gives

CA
x
_ y 0
157
(5:101)
We can now substitute from (5.100) into (5.101), which gives an equation for
x:

CA 2 2
x
x 0
(5:102)
A
This equation represents a simple harmonic motion and has the solution

CA
x 0 cos
t
(5:103)
A
where 0 is the amplitude and the phase. By substituting this result into
(5.100) and solving for y we get

CA
t
(5:104)
sin
y
0
A
Equations (5.103) and (5.104) describe a periodic motion of the instantaneous spin axis about the axis of gure. It is called the free nutation (or Euler
nutation). Its period is

2
A
0
(5:105)
CA
The factor 2/ represents the daily rotation of the Earth, so the period of the free
nutation is A/(C A) days. The dynamic ellipticity obtained from the precession
period (5.78) indicates that this period is about 305 days (~10 months). However,
astronomers in the eighteenth and early nineteenth centuries were unable to
detect a motion of Earths axis with this period. The reason lies in the assumption
that the Earth rotates as a rigid body. In fact its elasticity allows it to deform
slightly as a result of the displacement of the instantaneous rotation axis from
the axis of gure, and this extends the period to 435 days (~14 months). The
observed motion is called the Chandler wobble.

The Chandler wobble is a somewhat irregular cyclical motion of the instantaneous rotation axis with a period of about 435 days and an amplitude of a few
158
Earths rotation
millisec of arc along Greenwich meridian
100
2008
Apr 5
2009
May 20
2007
Mar 12
2010
Jul 1
100
2006
Aug 24
2007
Sep 28
2009
Dec 6
2008
Nov 1
200
300
500
400
300
200
100
millisec of arc along 90 East
Fig. 5.8. The instantaneous rotation axis of the Earth exhibits a nearly circular
motion with period 435 days the Chandler wobble and an annual circular
motion. These motions are superposed on a slow drift of about 20 m per century
along longitude 80 W. Data source: International Earth Rotation and Reference
Systems Service.
tenths of a second of arc, approximately 1015 m (Fig. 5.8). The displacement

of the rotation axis from its mean position is thought to result from changes in
oceanic circulation and uctuations in atmospheric pressure. The displacement
of the instantaneous rotation axis from the axis of gure gives rise to an
asymmetry in the Earths shape. The moments of inertia A, B, and C about the
reference axes are no longer adequate to describe the inertia tensor. The
products of inertia H, J, and K are needed to express the asymmetry of the
mass distribution (see Box 2.2). Let the instantaneous rotation axis have a
direction specied by direction cosines (, , ) relative to the x-, y-, and
z-axes dened in Fig. 5.7. The moment of inertia I about the instantaneous
rotation axis is given by (2.134):
I A2 B2 C2 2K 2H 2J
On writing I11 = A, I22 = B, and I33 = C for the principal moments of inertia
and I12 = I21 = K, I13 = I31 = J, and I23 = I32 = H for the products of inertia
(Box 5.1), this equation becomes
I I11 2 I22 2 I33 2 2I12 2I23 2I31
159
(5:106)
The angular velocity has components (x, y, ). Using numerical subscripts

1, 2, and 3 for the x-, y-, and z-components, respectively, the angular momentum
h and angular velocity are related by the tensor equation
hi Iij j
(5:107)
where the symmetric inertia tensor Iij (Box 5.1) represents the elements of the
matrix
Box 5.1. The inertia tensor

Let a rigid body be composed of elementary particles with mass mi and
coordinates (xi, yi, zi) relative to an orthogonal Cartesian coordinate system.
Let the body rotate with angular velocity about an axis through the origin.
The linear velocity of a particle mi at distance ri from the origin is
vi w ri
(1)
The linear momentum of the particle is mivi and its contribution to the
angular momentum of the rotating body is
h i r i mi v i
The angular momentum of the body is
X
X
h
mi ri vi
mi ri w ri
i
(2)
(3)
Using the identity in (1.18), the vector cross product is

ri w ri wr2i ri w ri
(4)
On substituting this expression into (3), the angular momentum becomes

X
X
mi r2i
mi ri w ri
(5)
hw
i
The x-component hx is
X
X

mi x2i y2i z2i
mi xi x xi y yi z zi
hx x
i
(6)
160
Earths rotation
hx x
X
X

mi y2i z2i y
mi xi yi z
m i z i xi
(7)
Analogously, the y- and z-components, hy and hz, of the angular momentum

are, respectively,
X
X
X

hy x
mi yi xi y
mi z2i x2i z
m i yi z i
(8)
i
hz x
m i z i xi y
m i z i yi z
mi x2i y2i
(9)
Using the denitions of moments and products of inertia in Box 2.2, the
angular momentum components are
hx Ax Ky Jz
hy Kx By Hz
hz Jx Hy Cz
(10)
These equations relating the components of h and can be written as a

single matrix equation,
0 1 0
10 1
A K J
x
hx
@ hy A @ K B H A@ y A
(11)
J H C
hz
z
Using numerical subscripts 1, 2, and 3 for the x-, y-, and z-components,
respectively, the moments of inertia (diagonal elements) are represented by
I11 = A, I22 = B, and I33 = C. The products of inertia (non-diagonal elements)
are I12 = I21 = K, I13 = I31 = J, and I23 = I32 = H. The matrix equation is
then
0 1 0
10 1
h1
I11 I12 I13
1
@ h2 A @ I21 I22 I23 A@ 2 A
(12)
h3
I31 I32 I33
3
In tensor notation this equation is written succinctly as
hi Iij j
i 1; 2; 3; j 1; 2; 3
(13)
The symmetric, second-order tensor Iij, whose components are the moments
and products of inertia, is called the inertia tensor.

0
I11
Iij @ I21
I31
I12
I22
I32
1
I13
I23 A
I33
161
(5:108)
Equation (5.93) for the free motion of the displaced instantaneous rotation
axis becomes
h_i w hi 0
(5:109)
Upon inserting (5.107), we have for the rst term
h_i Iij j I_ij j Iij _ j

t
(5:110)
The x-, y-, and z-components of the cross product have the form
w h1 2 I3k k 3 I2k k
(5:111)
The components of the equation of motion become

h_1 2 I3k k 3 I2k k 0
h_2 3 I1k k 1 I3k k 0
h_3 1 I2k k 2 I1k k 0
(5:112)
By expanding these equations of motion separately, we obtain expressions for

each individual component.
For the x-component,
I11 _ 1 I12 _ 2 I13 _ 3 I_11 1 I_12 2 I_13 3
2 I31 1 2 I32 2 2 I33 3 3 I21 1 3 I22 2 3 I23 3 0
(5:113)
For the y-component,
I21 _ 1 I22 _ 2 I23 _ 3 I_21 1 I_22 2 I_23 3
3 I11 1 3 I12 2 3 I13 3 1 I31 1 1 I32 2 1 I33 3 0
(5:114)
For the z-component,
I31 _ 1 I32 _ 2 I33 _ 3 I_31 1 I_32 2 I_33 3
1 I21 1 1 I22 2 1 I23 3 2 I11 1 2 I12 2 2 I13 3 0
(5:115)
162
Earths rotation
5.5.1 Simplication of the equations of motion

Each of the equations of motion contains many terms, some of which are
effectively irrelevant because they are very small compared with other terms.
In order to obtain an analytical solution it is necessary to introduce some
approximations, as follows.
1. The angular velocities (1, 2) are small compared with the daily rotation .
We will retain 1 and 2 to rst order but neglect their products and higher
orders, i.e.,
21 22 1 2 0
2. The products of inertia (non-diagonal elements in the inertia tensor) are
small, and we may neglect their products with the velocities (1, 2), i.e.,
I13 1 I13 2 I12 1 I12 2 I23 1 I23 2 0
3. We may also assume that the products of inertia change very slowly with
time. In this case we may neglect further products with the velocities
(1, 2), i.e.,
I_13 1 I_13 2 I_12 1 I_12 2 I_23 1 I_23 2 0
4. We may assume that the principal moments of inertia A, B, and C do not
change with time, i.e., only the asymmetry in the mass distribution is
responsible for the wobble of the rotation axis. That is,
I_ii 0
If we now apply these assumptions to the equations of motion, most of the terms
drop out. For example, (5.115) reduces to
I33 _ 3 0
(5:116)
This leads to the same result as for the Euler precession of the rigid Earth,
namely that the angular velocity about the axis of gure is constant:
3
(5:117)
The remaining two equations of motion reduce to

I11 _ 1 I_13 3 2 3 I33 I22 23 I23 0
(5:118)
I22 _ 2 I_23 3 3 1 I11 I33 23 I13 0
(5:119)
163
These can now be rewritten with the more easily recognizable parameters for the
moments and products of inertia:
_ 0
A_ 1 2 C A 2 H J
(5:120)
_ 0
A_ 2 1 C A 2 J H
(5:121)
The displacement of the instantaneous axis of rotation from the z-axis is very
small, amounting to less than 0.25 arcsec. The direction cosines of the rotation
axis may therefore be written as (, , 1) and the angular velocities as (1 = ,
2 = ). Upon inserting these values into the equations of motion and dividing
throughout by , we get the simultaneous equations
A_ C A H J_ 0
(5:122)
A_ C A J H_ 0
(5:123)
Note that the product of inertia K, which describes asymmetry in the xy

plane, does not play a role in the wobble equations. Only asymmetries in the yz
and zx planes that include the rotation axis determine the wobble motion. This
will become evident when we compute the values of the products of inertia H
and J, which we will obtain from a comparison with the MacCullagh equation
for the gravitational potential of the non-spheroidal Earth.
5.5.2 Computation of the products of inertia

The Earth is deformed by the centrifugal force of its rotation, the main result
being its spheroidal shape. If the axis of rotation is displaced from the axis of
symmetry of a rigid Earth, the spheroid exhibits Euler nutation about the spin
axis without additional deformation (Fig. 5.9(a)). However, the body of an
elastic Earth can adjust its shape to the displaced spin axis by deforming further,
as illustrated in Fig. 5.9(b). Parts of the ellipsoid are elevated above the original
spheroid (regions e), while other parts are depressed below it (regions d).
The shape conforming to the elastic deformation caused by the Chandler
wobble is not symmetric with respect to the reference axes. This gives rise to
the products of inertia H and J.
At a point in the Earth specied by co-latitude and radial distance r the
distance from the rotation axis is r sin and the potential of the centrifugal
acceleration is
1
1
1
2 r2 sin2 2 r2 2 r2 cos2
2
2
2
(5:124)
164
Earths rotation
(a) rigid
(b) elastic
e
x
Fig. 5.9. (a) Displacement of the rotation axis of a rigid Earth results in Euler
nutation without additional deformation. (b) The elastic Earth adjusts its shape to
the displaced spin axis by deforming further, so that regions e lie above and
regions d lie below the elliptical section (dashed) of the rigid body.
Let the Cartesian coordinates of the point be (x, y, z). The direction cosines
(0, 0, 0) of the radius through the point at (r, ) are
x
0 ;
r
y
0 ;
r
z
r
(5:125)
If the direction cosines of the instantaneous rotation axis are (, , ), then is

approximately the angle between the two lines, and
cos 0 0 0
(5:126)
Inserting the values from (5.125) gives

r cos x y z
(5:127)
and, using this relationship in (5.124), we get the centrifugal potential

1
1
2 x2 y2 z2 2 x y z2
2
2

1
2 x 2 y 2 z 2
2

1
2 2 x2 2 y2 2 z2 2xy 2yz 2zx
2
(5:128)
(5:129)
This may be simplied as before by setting the second-order values 2 = 2

= = 0 and 2 = = 1. Then

1
2 x2 y2 2 zx y
2
(5:130)
The rst term here is the centrifugal potential due to rotation about the axis of
gure. The second term is the extra centrifugal potential 2 due to the displacement of the instantaneous rotation axis in the Chandler wobble,
2 2 zx y
165
(5:131)
The wobble potential is a second-order solution of Laplaces equation, because

r2 2 2

2
2
2
zx yz 0
x2 y2 z2
(5:132)
2 is a deforming potential and causes a corresponding deformation that has its

own gravitational potential i, which, as in the theory of the equilibrium tides, is
proportional to 2,
i k2 zx y
(5:133)
The constant of proportionality k is the rst Love number. The potential i is

a solution of Laplaces equation for a space in which r can be zero. In our case it
describes the wobble centrifugal potential within the Earth. We need a solution
that is valid outside the Earth. As shown in Section 4.3.2 for the tidal gravity
anomaly, the general solution of Laplaces equation may be written

B
Ar 3 P2 cos i e
r
2
(5:134)
where the rst part i is valid inside and the second part e outside a volume of
interest. The two solutions vary differently with radial distance r, but their ratio
for the Earth with radius R is
5
R
e
i
(5:135)
r
On substituting for i from (5.133), the potential of the deformation caused by
the wobble is
e
R5 2
k zx y
r5
(5:136)
On converting the Cartesian coordinates (x, y, z) to direction cosines (0, 0, 0)

of the line through the point of observation (5.125), we get the potential e of
the wobble deformation at an external point:
e
k2 R5 0 0 0
r3
(5:137)
166
Earths rotation
5.5.3 Comparison of the wobble potential with MacCullaghs

formula
The MacCullagh formula for the gravitational potential UG of a triaxial ellipsoid
with mass E at an external point is given by (2.128), repeated here:
UG G
E
A B C 3I
G
r
2r3
(5:138)
I is the moment of inertia about a radial line passing through the point of
observation. Substituting (5.106) for I with direction cosines (0, 0, 0) gives
E
r

!
ABC3 A20 B20 C20 2K0 0 2H0 0 2J0 0
G
2r3
UG G
(5:139)
The terms involving products of inertia describe contributions to the potential
from features that deviate from symmetry with respect to the xy, yz, and zx
planes. The potential of the deformation associated with the Chandler wobble
depends on the products of direction cosines 00 and 00. On comparing the
coefcients of these products in (5.137) and (5.139) we get the following
expressions for the products of inertia:
H
2 R5 k
3G
(5:140)
J
2 R5 k
3G
(5:141)
5.5.4 Period of the Chandler wobble

The products of inertia H and J in the equations of motion (5.122) and (5.123)
may now be replaced by the above expressions. The pair of simultaneous
equations becomes
A_ C A
3 R5 k
2 R5 k _
0
3G
3G
(5:142)
A_ C A
3 R5 k
2 R5 k
_ 0
3G
3G
(5:143)
On regrouping the terms in these equations we get

2 R5 k _
2 R5 k
A
C A
0
3G
3G

2 R5 k
2 R5 k
A
_ C A
0
3G
3G
167
(5:144)
(5:145)
Analogous equations (5.95) and (5.96) for the rigid Earth yielded the period of
the free, Eulerian nutation,

2
A
(5:146)
0
CA
Proceeding in the same manner, the solutions of the nutation equations for an
elastic Earth are reduced to a simple harmonic motion of the rotation axis with
period

2
A 2 R5 k=3G
(5:147)
C A 2 R5 k=3G
This is the period of the Chandler wobble. The numerator in (5.147) is larger
than that in (5.146) and the denominator is smaller than that in (5.146). Thus the
period of the Chandler wobble for the elastic Earth is longer than the period of
the Eulerian nutation for a rigid Earth. The difference in periods can be used to
compute a measure of the Earths elastic yielding.
5.5.5 Calculation of Loves number k from the period of the

Chandler wobble
Loves number k, which we encountered in the theory of the tides, is a measure
of the global yielding of the Earth to the deforming tidal forces. A similar
situation is encountered here: the elastic yielding of the Earth to the centrifugal
force related to the free nutation results in the lengthened period observed in the
Chandler wobble, which therefore depends on k.
The density distribution in the Earth is dependent on the ratio m between the
centrifugal acceleration and the gravitational attraction at the equator (Box 3.2):
m
2 a
2 a3
GE=a2
GE
(5:148)
Ignoring the small differences between the equatorial radius and mean radius,
and using for the Earths rotation, we can replace this denition of m by
168
Earths rotation
m
2 R3
GE
(5:149)
It follows that in (5.147) we can write

2 R5 mER2
3G
3

2
A kmER2 =3
C A kmER2 =3

2
A
1 kmER2 =3A
C A 1 kmER2 =3C A

1 kmER2 =3A
0
1 kmER2 =3C A
(5:150)
(5:151)
(5:152)
In (3.39) we established a relationship between the principal moments of

inertia A and C, the attening , and the centrifugal acceleration ratio m,
C A 2f m
ER2
3
(5:153)
and from (3.43) we know that the approximate values of A and C are
1
A C ER2
3
(5:154)
We can substitute these values into (5.152), which simplies to

1 1
m
1k
1 km1
0
2f m
(5:155)
This relationship can be expanded as a binomial series. Neglecting second-order

and higher powers of m and , we obtain to rst order

1 1
m
1
1
1k
1 km
km
1
0
2f m
0
2f m
(5:156)
This reduces further to

1 1
m
1k
0
2f m
By rearranging terms and solving for Loves number we get
(5:157)
Further reading

0 2f m
k 1
169
(5:158)
Upon inserting the known values for , m, 0, and we get k = 0.28, in good
agreement with the value obtained from the theory of the tides.
further reading
Lambeck, K. (1980). The Earths Variable Rotation: Geophysical Causes and
Consequences. Cambridge: Cambridge University Press, 464 pp.
Moritz, H. and Mueller, I. I. (1988). Earth Rotation: Theory and Observation. New York:
Ungar, 617 pp.
Munk, W. H. and MacDonald, G. J. F. (1975). The Rotation of the Earth: A Geophysical
Discussion. Cambridge: Cambridge University Press, 384 pp.
6
Earths heat
The early thermal history of the Earth is a matter of some speculation. Current
scientic consensus is that planet Earth formed by accretion of material with the
same composition as chondritic meteorites. Accretion, a process that generated
heat as colliding material gave up kinetic energy, led to differentiation of the
planetary constituents into concentric layers. When the temperature of the early
Earth reached the melting point of iron, the dense iron, accompanied by other
siderophile elements such as nickel and sulfur, sank towards the center of the planet
to form a liquid core. Meanwhile lighter elements rose to form an outer layer, the
primitive mantle. Further differentiation took place later, creating a chemically
different thin crust atop the mantle. Only the outer core is now molten, surrounding
a solid inner core of iron that solidied out of the core uid. Lighter elements left
behind in the core rise through the core uid and result in a composition-driven
convection in the outer core, which is in addition to thermal convection. Although
the short-term behavior of the mantle is like that of a solid, allowing the passage of
seismic shear waves, its long-term behavior is characterized by plastic ow, so heat
transport by convection or advection is possible. In the solid lithosphere and inner
core heat is transported dominantly by thermal conduction.
The physical states of the Earths mantle and core are well understood, but the
variation of temperature with depth is not well known. Direct access is impossible
and it is very difcult in laboratory experiments to achieve the temperatures and
pressures in the Earths deep interior. Consequently, some important thermodynamic parameters are inadequately known. Points on the melting-point curve can
be determined from experiments at high temperature and pressure. Convection
ensures that the temperature prole in the mantle and outer core is close to the
adiabatic temperature curve, which can be calculated. From these considerations
an approximate temperature prole in the Earths interior can be estimated
(Fig. 6.1). The temperatures in the mantle and outer core are close to the adiabatic
curve, little temperature change occurs in the solid inner core, and comparatively
rapid change occurs in the asthenosphere and lithosphere.
170
5000
Asthenosphere
(partial melting)
Temperature (C)
4000
171
geotherm
Lithosphere
solidus
3000
2000
1000
670
400
MANTLE
(solid silicate)
1000
2000
OUTER CORE
(liquid iron alloy)
3000
4000
INNER
CORE
(solid iron
alloy)
5000
6000
Depth (km)
Fig. 6.1. Models of the adiabatic temperature prole (geotherm, solid curve) and
the melting-point curve (solidus, dashed curve) in the Earths interior. Data sources:
tables in appendix G of Stacey and Davis (2008); for mantle solidus, Stacey (1992),
appendix G.

Analysis of the thermal conditions in the Earth is based upon the First and
Second Laws of Thermodynamics. The First Law is an application of the
conservation of energy to a thermodynamic system. It states that energy cannot
be created or destroyed in a closed system, but can only be transformed from
one form to another. In an open system, extra terms must be considered to allow
for the transfer of energy into or out the system (e.g., by the ow of matter). The
total energy, Q, of a closed system consists of its internal energy, U, and the
work, W, done in any external transfer of energy to the surroundings. The energy
balance is expressed by the equation
dQ dU dW
(6:1)
Heat added to (or removed from) a closed system is used to increase the internal
energy and to perform external work. For example, the gas molecules in a heated
balloon are more energetic, and, if it is able to expand, the volume, V, increases. The
external work dW due to the change in volume at constant pressure, P, is
dW P dV
(6:2)
and so from the First Law of Thermodynamics the energy equation is

dQ dU P dV
(6:3)
172
Earths heat
The Second Law of Thermodynamics asserts that the energy of an isolated

system tends to become uniformly distributed with the passage of time. The
concept of entropy, S, is used as a measure of the microscopic disorder in a
system at a particular temperature. The change dS in the entropy of a system
caused by a change in energy dQ at a temperature T is dened as
dS
dQ
T
(6:4)
On substituting this into the energy equation we get

T dS dU P dV
(6:5)
This important relation, uniting the First and Second Laws, is the central
equation of thermodynamics. It is important in the analysis of thermal conditions inside the Earth, because it denes adiabatic conditions.
An adiabatic thermodynamic process is one in which heat cannot enter or
leave the system, i.e., dQ = 0. The entropy of an adiabatic reaction remains
constant, because dS = dQ/T = 0. The adiabatic temperature gradient in the Earth
serves as an important reference for estimates of the actual temperature gradient
and for determining how heat is transferred.
6.2 Thermodynamic potentials and Maxwells relations

The thermodynamic state of a system can be expressed with the aid of scalar
functions called thermodynamic potentials. These are the internal energy, U, the
enthalpy, H, the Helmholtz energy, A, and the Gibbs free energy, G. Each
potential consists of a particular combination of the physical parameters pressure, temperature, volume, and entropy.
6.2.1 Thermodynamic potentials

Internal energy (U) has been described and dened above. A change in internal
energy at constant temperature and pressure is related to changes in volume and
entropy by
dU T dS P dV
(6:6)
Enthalpy (H) is a measure of the total energy of a system; it is a combination

of the internal energy and the product of the pressure and volume:
H U PV
By taking the differentials of both sides of the equation we get
(6:7)
6.2 Thermodynamic potentials
173
dH dU P dV V dP
(6:8)
The conservation of energy, expressed in (6.5), allows us to reduce this to

dH T dS V dP
(6:9)
The Helmholtz energy (A) is dened from the relationship between the
thermodynamic properties of macroscopic materials and their behavior on a
microscopic level through statistical mechanics. It is a measure of the work
obtainable from a closed thermodynamic system at constant temperature and
constant volume, and is dened as
A U TS
(6:10)
Taking the differentials of both sides gives

dA dU T dS S dT
(6:11)
dA P dV S dT
(6:12)
Using (6.5), this becomes
The Gibbs energy (G) is dened in a similar way to the Helmholtz energy, but
for constant pressure and temperature. It represents the maximum amount of
energy obtainable from a closed system (i.e., one isolated from its surroundings)
without increasing its volume, and is dened as
G A PV
(6:13)
The differentials give the equation

dG dA P dV V dP
(6:14)
Combining this with (6.12) gives

dG V dP S dT
(6:15)
6.2.2 Maxwells thermodynamic relations

Maxwells relations are a set of partial differential equations derived from the
denitions of the thermodynamic potentials that relate the parameters S, V, T,
and P. The relations depend on the mathematical equality between the second
derivatives of these parameters. This follows because the order of differentiation of a function F(x, y) of two variables x and y is not important:

F
2 F
2 F
F
x y x x y y x y x y
174
Earths heat
Maxwells thermodynamic relations are derived in Box 6.1 by applying this

condition to the different thermodynamic potentials. Summarized, they are
Box 6.1. Derivation of Maxwells thermodynamic relations

The internal energy, U, changes with V and S as in (6.6):
dU T dS P dV
(1)
dU can be written as a perfect differential using the partial derivatives of U

with respect to V and S:
U
S
dU

dS
U
V

dV
(2)
The coefcients of dV and dS in these expressions must be equivalent, thus

U
(3)
P
V S

T
U
S

(4)
V
P
2 U

S
T S
(5)
T
2 U
V T S
(6)
T
V

P

S V
S
(7)
This is one of the Maxwell thermodynamic relations. The three others are
obtained in a like manner.
The enthalpy, H, changes with P and T as in (6.9):
dH T dS V dP
(8)
dH can be written as a perfect differential using the partial derivatives of H

with respect to T and P:
6.2 Thermodynamic potentials

dH
H
S

H
dS
dP
P S
P
175
(9)
On equating the coefcients of dS and dP in these expressions, we have

T H=SP and V H=PS . Differentiating T with respect to P and
V with respect to S gives

T
V
(10)
P S
S P
The Helmholtz energy, A, changes with V and T as in (6.12):
dA P dV S dT
(11)
dA can be written as a perfect differential using the partial derivatives of A


A
A
dV
dT
(12)
dA
V T
T V
On equating the coefcients of dV and dT in these expressions, we have
P A=VT and S A=TV . Differentiating P with respect to T
and S with respect to V gives

P
S
(13)
T V
V T
The Gibbs energy, G, changes with P and T as in (6.15):
dG V dP S dT
(14)
dG can be written as a perfect differential using the partial derivatives of G


G
G
dP
dT
(15)
dG
P T
T P
On equating the coefcients of dP and dT in these expressions, we have
V G=PT and S G=TP . Differentiating V with respect to T
and S with respect to P gives

V
S

(16)
T P
P T
176
Earths heat

T
P

V S
S V

T
V
P S
S P

P
S
T V
V T

V
S

T P
P T
(6:16)
(6:17)
(6:18)
(6:19)
6.3 The melting-temperature gradient in the core

The ambient pressure has a strong inuence on the temperature at which the
inner core solidies from the core uid. At the inner-core boundary the pressure
is 330 GPa and the melting point of iron is around Tm = 5,000 K. If the latent
heat of fusion of iron is L, the amount of heat exchanged when a mass m melts is
dQ = mL and the change in entropy is
dS
dQ mL
T
Tm
(6:20)
Writing (6.17) in terms of full differentials, with T = Tm and substituting (6.20)

for dS, we have

dTm
dV
VL V S
(6:21)
dP S
dS P mL=Tm
where VL is the volume occupied by the mass of iron in a liquid state, and VS is
its volume in a solid state. We can write (6.21) as

dTm
Tm
VL VS
(6:22)
dP S mL
This is known as the ClausiusClapeyron equation for the change of state.
During solidication the density changes from L for the liquid to S for the
solid. The volume of a mass m of the material changes from VL = m/L before
the change of state to VS = m/S after the change of state, so that

1 dTm 1 1
1
(6:23)

Tm dP
L L S
6.3 The melting-temperature gradient
177
This equation must now be converted into a function of depth. The pressure
inside the Earth is assumed to be hydrostatic. Under these conditions an increase
in depth dz results in an increase in pressure dP solely because of the extra
material added to the vertical column. If the local gravity at depth z is g(z) and
the local density is L(z), the hydrostatic pressure increase is
dP gzL zdz
(6:24)
On substituting this into (6.23), we get an equation relating the increase in

melting temperature to increasing depth:

1 dTm g
L
1
(6:25)
S
Tm dz
L
The conditions in the core can be estimated from experiments and modeling.
The melting temperature and the latent heat of fusion of iron at the enormous
pressure in the core are not accurately known. For example, temperature
estimates lie within the range 5,0006,000 K. Some representative values of
physical properties in the core are given in Table 6.1. Using values for the
boundary between the inner and outer core in the modied ClausiusClapeyron
equation (6.25) the gradient of the melting temperature curve at that boundary is
dTm
1:4 K km1
dz
(6:26)
Table 6.1. Values of some physical parameters in the outer and inner core near
to the coremantle boundary (CMB) and inner-core boundary (ICB) (sources:
(1) Dziewonski and Anderson, 1981; (2) Stacey, 2007)
Outer core
at CMB
Outer core
at ICB
Inner core
at ICB
Source
m s2
kg m3
GPa
m2 s2
J K1 kg1
K
10.7
9,900
646
67.3
815
3,700
1.44
4.4
12,160
1,300
107
794
5,000
1.39
4.4
12,980
1,300
107
728
5,000
1.39
1
1
1
1
2
2
2
106 K1
18.0
10.3
9.7
9.6
Physical property
Units
Gravity, g
Density,
Bulk modulus, KS
= KS/
Specic heat, cP
Temperature, T
Grneisen parameter,
Volume expansion
coefcient,
Latent heat of melting,
L
105 J kg1
178
Earths heat
6.4 The adiabatic temperature gradient in the core

When heat is added to a material it causes an increase in temperature. The
specic heat of the material is the amount of heat needed to raise the temperature
of 1 kg of the material by 1 K; it can be dened for constant pressure, cP, or
constant volume, cV. For a mass m of the material the heat dQ required to raise
the temperature by dT at constant pressure is
dQ mcP dT
(6:27)
from which we get

Q
T

mcP
(6:28)
The increase in temperature causes the material to expand. The coefcient of

thermal expansion P is dened as the fractional increase in volume per degree
increase in temperature. This can be written

1 V
P
(6:29)
V T P
The change in energy due to the heat added can be expressed as a perfect
differential, giving

Q
Q
dQ
dT
dP
(6:30)
T P
P T
Using the denition of entropy, this becomes

Q
S
T dS
dT T
dP
T P
P T
(6:31)
Equation (6.28) can be used in the rst term on the right, and the Maxwell
relation from (6.19) can be used in the second term:

V
dP
(6:32)
T dS mcP dT T
T P
The condition for an adiabatic process, in which no heat is gained or lost by
the system, is that the entropy remains constant, dS = 0, so
mcP dT TVP dP
T
P
TVP TP
mcP
cP
(6:33)
(6:34)
179
This gives the adiabatic change of temperature with increasing pressure. Using
(6.24), we convert the change in pressure to a change in depth and obtain the
adiabatic temperature gradient,

T
gTP
(6:35)
cP
z S
The depth prole of the adiabatic temperature is important for understanding
conditions in the uid core. If the actual temperature prole deviates from the
adiabatic curve, this gives rise to convection currents, which redistribute the
temperature to maintain adiabatic conditions. The physical parameters in
Table 6.1 give an adiabatic temperature gradient in the uid core of

T
0:88 K km1
(6:36)
z S
at the coremantle boundary, and

T
0:29 K km1
z S
(6:37)
at the boundary with the inner core.

Comparison of these values with (6.26) shows that the melting temperature Tm
increases more rapidly with depth than does the adiabatic temperature. In the early
Earth, cooling from the surface, the melting temperature would have been reached
rst at the center. The core would have solidied from the bottom upwards, thus
giving rise to the present layering of uid outer core and solid inner core. Once the
inner core became solid, it could cool further only by conduction, whereas
convection continues to be the dominant process of heat transfer in the outer core.

The atoms of a metal are located at specic sites in a regular lattice, forming a
crystalline pattern that corresponds to the ambient conditions. Iron has a bodycentered cubic (b.c.c.) structure at room pressure and temperature, but, as the
pressure increases, the structure changes to a denser face-centered cubic (f.c.c.)
packing, and eventually to hexagonal close packing (h.c.p.). At the pressure
(330 GPa) and temperature (6,000 K) of the inner-core boundary iron is
believed to have the h.c.p. structure. On a microscopic level the atoms in the
iron lattice vibrate at a frequency given by the temperature. The atomic vibrations cannot take arbitrary values, but exhibit normal modes like classical
vibrations of a string. The quantized vibrations, or phonons, are responsible
180
Earths heat
for heat conduction in the solid and the long-wavelength phonons transport
sound. A change in the temperature of a solid causes a change in volume, which
alters the inter-atomic distances and thus the vibrational modes (phonon frequencies) of the crystal lattice. In solid-state physics this change is described by
the Grneisen parameter, . This is a dimensionless parameter, originally
dened to represent the dependence of a particular mode of lattice vibration
(phonon frequency) on a change of volume V. The microscopic denition of a
Grneisen parameter for a particular mode with frequency i is

ln i
(6:38)
i
ln V T
It is difcult to adapt this denition to measurable quantities, because to do so
requires detailed knowledge of the lattice dynamics. A more useful macroscopic denition of the Grneisen parameter relates it to thermodynamic properties such as the bulk modulus, KS, density, , specic heat, c, and coefcient of
thermal expansion, . The denition at constant pressure is
P K S
cP
(6:39)
The importance of in geophysics is due to its occurrence in equations that

describe the dependence of physical properties on temperature and pressure,
and therefore on depth. However, it is difcult to obtain values for the physical
properties that dene in laboratory experiments at high pressure and temperature that are representative for the conditions in the core. Conveniently, varies
only slowly with pressure and temperature. It changes noticeably at Earths
important internal boundaries, but between these does not change much over
large ranges of depth (Fig. 6.2).
Equation (6.35) for the adiabatic temperature gradient can be reformulated as
follows:

dT
gT P KS
(6:40)
cP
dz S
KS
Inserting the macroscopic denition of allows the temperature gradient to be
written as

dT
gT
(6:41)
dz S
KS
This equation can be rened further by using the velocities of seismic waves
through the Earth, which are determined by the elastic constants. The relations
Gr neisen parameter
0.8
1.2
0.4
181
1.6
0
400
670
2000
Depth
(km)
4000
6000
Fig. 6.2. Estimated variations of the Grneisen parameter in different regions of

the Earths interior. Data source: Stacey and Davis (2008), appendix G.
between the P-wave velocity and S-wave velocity and the bulk modulus KS,
rigidity , and density are developed in Section 8.5, giving
4
KS
3

KS
4
2 2
(6:42)
(6:43)
(6:44)
is called the seismic parameter and is well known as a function of depth in the
Earth because of the precise knowledge of seismic velocities on which Earth
models such as PREM (Dziewonski and Anderson, 1981) are founded. Using
this function, the equation for the adiabatic temperature gradient reduces to

dT
gT
(6:45)
dz S
6.5.1 Temperature and density in the Earth

Thermal convection is the main form of heat transport in the outer core and is
also important in the Earths mantle. It keeps the ambient temperature close to
the adiabatic temperature in these regions. Equation (6.41) for the adiabatic
gradient can be reformulated as a function of pressure instead of depth,
182
Earths heat
dT
gT
T
dz
dP
KS
KS
(6:46)
When the pressure increases, the volume normally decreases. In an elastic

material the fractional change in volume is proportional to the pressure change;
the proportionality constant is the bulk modulus, which under adiabatic conditions is denoted KS,

dP
dP
(6:47)
KS V
dV S
d S
On rearranging this relationship we obtain
dP d
KS
(6:48)
Substituting into the adiabatic equation gives

dT
dP
d
T
KS
(6:49)
Integrating both sides gives

ln

T2
ln 2
T1
1

T2
2
T1
1
(6:50)
(6:51)
In this way, knowing the Grneisen parameter for a particular domain allows the
variation of temperature to be estimated from the variation of density with
depth, which is well known.
6.6 Heat ow
When a straight conductor is heated so that one end is maintained at temperature
T1 and the other at a higher temperature T2 (Fig. 6.3), the amount of heat Q
owing out of the cooler end is inversely proportional to the length L of the
conductor, and directly proportional to its cross-sectional area A, the measurement time t, and the temperature difference between the ends:
DQ / A
T2 T1
Dt
L
(6:52)
We use this observation to dene the vertical ow of heat at the Earths surface.
6.6 Heat ow
183
T2 >T1
Q
A
T1
Fig. 6.3. The ow of heat Q along a conductor of length L and cross-section A, with
ends maintained at different temperatures T1 and T2 (T2 > T1).
6.6.1 The heat-ow equation

Let Cartesian axes be dened so that the z-axis is vertically downwards and the
x- and y-axes lie in the horizontal plane (Fig. 6.4). Consider the heat owing
vertically upwards along a very short conductor of cross-sectional area Az
normal to the z-direction and of length dz, such that its upper, cooler end at
depth z has temperature T and the lower, warmer end at z + dz has temperature
T + dT. Upon inserting these values into (6.52) and introducing a proportionality constant k we obtain a differential equation for the heat loss per unit time:
dQz
dT
kAz
dt
dz
(6:53)
The minus sign indicates that the heat ows in the direction of decreasing z (i.e.,
upwards). The proportionality constant is a material property of the conductor,
namely its thermal conductivity. The heat ow qz is dened as the heat crossing
unit area per second:
qz
1 dQz
dT
k
Az dt
dz
(6:54)
This gives the vertical heat ow along the z-axis; it is possible to dene
horizontal components along the x- and y-axes in a similar way, so in general
we can write the heat ow as a vector,
q k rT
(6:55)
6.6.2 The thermal-conduction equation

Returning to the one-dimensional situation, consider the heat owing vertically
upwards (along the z-axis) through a small rectangular box of sides x, y, and
184
Earths heat
qz
x
z
z
qz + dq z
T + dT
Az
z + dz
dy
dx
Fig. 6.4. Heat Qz + Qz ows vertically into the base Az of a small box with sides
x, y, and z, whereas the amount of heat that leaves the top of the box is Qz.
z with top surface at depth z, where the temperature is T (Fig. 6.4). The heat
ow through the top surface is qz, and the area of the surface normal to the ow
is Az = x y, so the total vertical loss of heat Qz in time t is
Qz qz Dx DyDt
(6:56)
At depth z + dz the heat entering the bottom end of the box is Qz + Qz, where
Qz DQz Qz
Qz
Dz
z
(6:57)
The amount of heat remaining in the box is the difference between the amounts
entering and leaving it; on substituting from the right-hand side of (6.56) we have
DQz
Qz
qz
Dz
DzDx DyDt
z
z
(6:58)
Now we substitute the denition of the heat ow from (6.54) to obtain the
amount of heat Qz retained in the box

T
2 T
DQz
(6:59)
DV Dt k 2 DV Dt
k
z
z
z
Let cP be the specic heat at constant pressure and the density of the material
in the box, and let the rise in temperature caused by the extra heat be T. The mass
of matter in the box is m = V, so, using the denition of specic heat,
DQz cP m DT cP DV DT
(6:60)
By equating this with (6.59) and deleting the factor V on each side, we get
6.6 Heat ow
cP
185
T
2 T
k 2
t
z
(6:61)
2
T
k T

t
cP z2
(6:62)
The combination of physical parameters in parentheses denes the thermal

diffusivity, ,
k
cP
(6:63)
The one-dimensional equation of heat conduction is therefore

T
2 T
2
t
z
(6:64)
This equation is one of the most important in geophysics. An equation with

identical form describes the process of diffusion, by which a net ux of randomly
moving particles that is proportional to the gradient in concentration of the particles
can take place. Consequently, the thermal-conduction equation is sometimes
called the heat-diffusion equation. Two specic examples of one-dimensional
heat conduction are described in the following sections: the penetration of external
heat into the Earth and the loss of heat from a cooling half-space.
By extension to the x- and y-directions, similar components are found, the
only difference being that the second-order differentiation is with respect to x
and y, respectively. The heat-conduction equation for three dimensions is
2

T
T 2 T 2 T
(6:65)

t
x2 y2 z2
or
T
r2 T
t
(6:66)
6.6.3 Penetration of solar heat in the Earth

Solar energy heats Earths surface in a quasi-cyclical fashion, with a high and a
low temperature each day, and a warmest and coldest month each year. The solar
heat is transported downwards by conduction and is able to penetrate some
distance into the Earth. The decay of temperature with depth below the surface
can be evaluated by solving the one-dimensional heat-conduction equation with
appropriate boundary conditions.
186
Earths heat
Let the z-axis again be the vertical direction. The temperature satisfying
(6.64) is a function of both depth and time: T = T(z, t). As in other cases, we
apply the method of separation of variables. The depth variation is described by
the function Z(z) and the time variation by (t). Then
Tz; t Zzt
(6:67)
This expression is inserted into the heat-conduction equation, and both sides are
then divided by the product Z(z)(t). We have
2 Z
2
t
z
(6:68)
1
1 2 Z
t
Z z2
(6:69)
Each side of this equation involves a different independent variable, thus

both sides equal the same constant. This allows us to separate the equation
into two parts. We must choose the constant to t the boundary conditions of
the stated problem. If the incident solar energy is a periodic function of time,
then the solution will also be periodic. The time dependence of the surface
temperature can be expressed by the real part of the complex function
exp(it):
T T0 cost T0 Reexpit
(6:70)
On comparing this with the left-hand side of (6.69), we see that the common
constant in this equation must equal i:
1
i
t
(6:71)
The time dependence of the temperature variation at depth is therefore

0 expit
(6:72)
Because both sides of (6.69) equal the same constant, the depth function
satises
1 2 Z
i
Z z2
2 Z
i Z0
z2
This has the form of a simple harmonic equation,
(6:73)
(6:74)
6.6 Heat ow
187
2 Z
n2 Z 0
z2
(6:75)
Z Z1 expinz Z0 expinz
(6:76)
with solution
On comparing (6.74) and (6.75) we have
in i
n2 i
(6:77)
(6:78)
As shown in Section 1.2, the complex number exp(i) can be written
Thus
and
expi cos i sin
(6:79)

i exp i
2
(6:80)

p
1
i exp i
cos
i sin
p 1 i
4
4
4
2
(6:81)
Equation (6.78) can now be written

r
in
1 i
2
(6:82)
Upon inserting this into (6.76), the variation of temperature with depth becomes
r

r
1 iz Z0 exp
1 iz
(6:83)
Z Z1 exp
2
2
In this problem of solar heating we are interested in the ow of heat downwards into the Earth, in the +z-direction. The temperature uctuation related to
solar heating decreases with increasing depth, thus dZ/dz must be negative. The
rst term in (6.83) increases exponentially with depth, so we exclude it by
setting Z1 = 0 and obtain
r
Tz; t Z0 exp
1 iz 0 expit
(6:84)
2
188
Earths heat
The initial conditions at the surface (depth z = 0, time t = 0) are that the
temperature is equal to T0. Thus Z00 = T0 and the solution to the heatconduction equation is
r
r

z exp i t
z
(6:85)
Tz; t T0 exp
2
2
The temperature variation with time and depth is the real part of this solution:

z
z
Tz; t T0 exp
cos t
(6:86)
d
d
We have simplied the result by using
r
2
d
(6:87)
This is a characteristic depth for the problem, often called the penetration
depth. It is the depth at which the temperature uctuation has decreased to 1/e of
its surface value. It depends both on the frequency of the uctuation and on the
material properties of the ground. The thermal diffusivity is dened on the basis
of the specic heat, density, and thermal conductivity, all of which vary with
temperature. Consequently the thermal diffusivity is temperature-dependent; in
common rocks it decreases with increasing temperature. Assuming representative values of the physical properties of some common near-surface rock types,
typical penetration depths can be calculated (Table 6.2). The penetration depth
of the daily temperature variation (period = 86,400 s, = 7.27 105 rad s1) is
around 18 cm; that of the annual uctuation (period = 3.15 107 s, = 1.99
107 rad s1) is around 3.5 m.
Table 6.2. Calculated penetration depths of solar energy in continental surface

rocks for daily and annual temperature uctuations (source: average values
from graphed data in Vosteen and Schellschmidt (2003))
Thermal property
Units
Mean value
Thermal conductivity, k
Specic heat, cP
Density,
Thermal diffusivity,
Penetration depth of daily uctuation
Penetration depth of annual uctuation
W m1 K1
J kg1 K1
kg m3
106 m2 s1
m
m
2.5
800
2,750
1.1
0.18
3.4
6.6 Heat ow
189
Note that the penetration depth d is not the maximum depth to which the
solar energy can penetrate, but merely the depth at which the amplitude sinks
to 1/e. The surface temperature change is felt well below the penetration
depth. At a depth of 5d the signal has attenuated to about 1% of the surface
value.
The attenuation of the surface temperature uctuation is accompanied by a
shift in phase of the signal. We can write (6.86) as

z
Tz; t T0 exp
(6:88)
cost t0
d
The time t0 represents a delay in the time at which the surface extreme values are
felt at depth z:
s

z
z
z
t0
p
(6:89)
d
2
2
Figure 6.5 shows the attenuation and phase shift of the temperature for a
hypothetical sedimentary rock, using the data in Table 6.2. The surface temperature is assumed to vary periodically between +10 C and 10 C. At depths
below about 1 m the daily surface change is barely discernible; the corresponding depth for the annual uctuation is about 19 m. At depth z = d (around 11 m
in this case) the phase shift of the annual variation with respect to surface values
is 180; i.e., when the surface temperature is at its peak, the temperature at this
depth is minimum.
(a) 10
(b) 10
0 cm
0m
1m
Temperature (C)
12 cm
20 cm
100 cm
5
0
2m
5m
10 m
5
10
10
0
12
Time of day
18
24
Jan
Feb
Mar
Apr
May
Jun
Jul
Aug
Sep
Oct
Nov
Dec
Temperature (C)
4 cm
Month of year
Fig. 6.5. Effect of surface solar heating on near-surface temperatures in a sedimentary

rock. Attenuation and phase shift of (a) daily and (b) annual temperature uctuations.
190
Earths heat
6.6.4 Cooling of a semi-innite half-space

The second application of the heat-conduction equation is to the outward
vertical ow of heat from the Earths interior as it cools from an initially hot
state. We assume a one-dimensional model consisting of a semi-innite halfspace that extends to innity in the (vertical) z-direction. Lateral components of
heat ow, such as result from modication by surface topography, are ignored.
The problem consists of determining the temperature distribution T(z, t) as a
function of depth z in the half-space at time t after it starts to cool.
Let the temperature of the upper surface be zero. The temperature in the
cooling half-space must satisfy the heat-conduction equation, and is obtained by
separation of the variables as in (6.69):
1 1 1 2 Z
t Z z2
(6:90)
In this instance we are studying not a uctuating temperature, but a steady

cooling process. Separating the variables as before, we set the separation
constant equal to n2:
1 1
n2
t
(6:91)
1 2 Z
n2
Z z2
(6:92)
The particular solution of the time-dependent part is

0 exp n2 t
(6:93)
and that of the spatial part is

Z An cosnz Bn sinnz
(6:94)
The boundary condition on the upper surface at z = 0 is T(0, t) = 0, which

requires An = 0. The general solution is a sum over all possible values of n:
Tz; t 0
1
X

exp n2 t Bn sinnz
(6:95)
n0
For a continuous temperature distribution the summation can be replaced by an

integral in which the constants 0 and Bn are combined in a continuous function B(n):
Z1
Tz; t
n0

exp n2 t Bnsinnzdn
(6:96)
6.6 Heat ow
191
Suppose that at t = 0 the cooling half-space has an initial temperature distribution T(z):
Z1
Bnsinnzdn
(6:97)
Tz; 0 Tz
n0
This is a Fourier integral equation, in which the amplitude function B(n) must
be determined. This is obtained by using the properties of Fourier sine transforms, which are explained briey in Section 1.17. The Fourier sine transform
allows us to write the amplitude function as
2
Bn
Z1
z0
2
Tzsinnzdz
Z1
T sinn d
(6:98)
In the nal expression the integration variable has been changed from z to to
avoid subsequent confusion when we insert the result back into (6.96). The
substitution gives
2 1
3
Z
Z1

2
T 4
exp n2 t sinnzsinn dn5d
(6:99)
Tz; t
n0
Now we can change the integrand by using the trigonometric relationship

2 sinnzsinn cosn z cosn z
giving
1
Tz; t
"
Z1
Z1
T
0
(6:100)
expn2 tcosn z:
n0
expn2 tcosn zdn d
(6:101)
Each of the integrals inside the square brackets has the same form, namely

R1
2
n0 exp n cosnudn, with = t and u = z or u = + z, respectively.
The integration of this function is shown in Box 6.2 to be
Z1
n0

1
exp n2 cosnudn
2
r
2
u
exp
(6:102)
Applying this solution to each integral in the square brackets in (6.101), with
= t, gives
192
Earths heat
Box 6.2. The cooling half-space integration

The cooling half-space solution requires evaluation of the integral
nZ1

exp n2 cosnudn
Y
(1)
n0
Note that, on differentiating with respect to u,

Y
nZ1

n exp n2 sinnudn
(2)
n0
Integrating (2) by parts with respect to n gives

1
Z1

exp n2
Y
u
u
exp n2 cosnudn Y (3)
sinnu
u
2
2
2
0
0
1 Y
u
lnY
Y u u
2
(4)
u2
lnY0
4
Here we have introduced Y0 as a constant of integration, and the solution to
the integration is

u2
Y Y0 exp
(5)
4
lnY
The constant Y0 is the value of the integral Y for u = 0. This constant

may be determined as follows:
Z1
Y0

exp x2 dx
x0
0
Y0 2 @
Z1
Z1

exp y2 dy
(6)
y0
1
10 1
Z

C
B
exp x2 dxA@
exp y2 dyA
x0
Z1 Z1
x0 y0
y0

exp x2 y2 dx dy
(7)
6.6 Heat ow
193
On changing to polar coordinates (r, ), we have x = r cos and y = r sin ,

and the element of area becomes dx dy = r dr d. The limits of integration
change from (0 x ; 0 y ) to (0 r ; 0 /2):
0
1
Z=2 Z1
Z=2 Z1

2
2
2
@
Y0
exp r r dr d
exp r r drAd (8)
0 r0
Z=2
2
Y0
0
1
Y0
2
r0

1
Z=2
exp r2
1

d
d
2
2
4
r0
0
(10)
By inserting this value into (5) we get the evaluated integral:

r

1
u2
Y
exp
4
2
1
Tz; t p
2 t
(9)
Z1
0
"
z2
T exp
4t
z 2
exp
4t
(11)
!#
d
(6:103)
If the cooling body has initially a uniform temperature T0, then T(z) = T0 and
the temperature distribution can be written
8
9
!
! >
>
Z1
Z1
2
2
<
=
T0
z
z
Tz; t p
d
d
exp
exp
>
4t
4t
2 t >
:
;
0
(6:104)
p
In the rst integration, on writing w z=2 t, we have
p
dw 1=2 td and the upper and lower limits of the integration change
p
p
to and z=2 t, respectively. Similarly, on writing v z=2 t in
the second integration, we get an equivalent expression for dv, but the integrap
tion limits become and z=2 t, respectively. Equation (6.104) becomes
194
Earths heat
8
>
T0 <
Tz; t p
>
:
Z1
Z1
p
wz=2 t
expw2 dw
p
vz=2 t
expv2 dv
9
>
=
>
;
(6:105)
The integration variables w and v in this equation are interchangeable, and can
be combined in a single integration, modifying the integration limits accordingly. This gives
8
9
8
9
p
p
z=2
z=2
>
>
>
>
Z t
Z t
<
=
<
=
T0
2T0
expw2 dw p
expw2 dw
Tz; t p
>
>
>
>
:
;
:
;
p
wz=2 t
w0
(6:106)
8
>
< 2
Tz; t T0 p
>
:
p
z=2
Z t
expw2 dw
w0
9
>
=
(6:107)
>
;
The expression in brackets is the error function (Box 6.3), dened as
Box 6.3. The error function

The error function is closely related to the bell-shaped normal distribution.
However, only positive values of the independent variable u are considered, so
the graph of the dening function is similar to the right half of a normal
distribution as in Fig. B6.3(a). Its equation is

2
fu p exp u2
(1)
The error function erf() is dened as the area under this curve from the origin
at u = 0 to the value u = :
2
erf p

exp u2 du
(2)
The complementary error function, erfc(), is dened as

2
erfc 1 erf p
Z1

exp u2 du
(3)
6.6 Heat ow
195
The value of erf() or erfc() for any particular value of may be obtained
from standard tables, or from a graph like Fig. B6.3(b).
(b) 1.0
(a) 1.2
1.0
F(u) = 2 exp(u 2/2)
0.8
0.8
F(u ) 0.6
erf()
0.6
erf()
0.4
0.4
erfc()
0.2
0.2
u =
1.0
2.0
1.0
2.0
3.0
Fig. B6.3. (a) The error function erf() is dened as the area under the normal
distribution curve from the origin at u = 0 to the value u = . (b) Graphs of the error
function erf() and complementary error function erfc().
2
erf p

exp u2 du
(6:108)
u0
Values of the error function are tabulated for any nite argument. The solution
for the temperature distribution as a function of time and depth in the cooling
half-space is therefore

z
Tz; t T0 erf p
(6:109)
2 t
This equation allows us to understand the heat ow measured over oceanic
crust.
6.6.5 Cooling of oceanic lithosphere

In plate-tectonic theory the oceanic lithosphere is formed at a ridge axis and is
transported away from the ridge by sea-oor spreading, cooling as it does so.
The age, or cooling-time t, of the lithosphere at any place is proportional to its
distance from the ridge axis, assuming a constant spreading rate. Two models
are in common use: a one-dimensional half-space model as described above,
196
Earths heat
and a plate model that considers the lithosphere to be a cooling boundary layer
with its top surface at sea-oor temperature, and with its base and the edge at the
spreading ridge at the temperature of the asthenosphere. The rst of these is
discussed further here.
The half-space model divides the lithosphere into narrow vertical columns,
initially at the same uniform temperature as the ridge material. When a block is
transported away from the ridge, it cools and emits a vertical heat ow;
horizontal heat conduction is ignored. In this simple model the temperature T
of an oceanic plate at a time t after forming at temperature T0 at the ridge is given
by an equation such as (6.109). The heat ow qz over oceanic crust of age t is
obtained from the vertical temperature gradient:
qz k
dT
dT d
k
dz
d dz
dT
d
2T0 d
T0 erf p
d
d
d
(6:110)

exp u2 du
u0

2T0
p exp 2

d
d
z
1
p p
dz dz 2 t
2 t
(6:111)
(6:112)
On combining these equations, we obtain the heat ow

T0
qz p exp 2
t
(6:113)
At the surface of the oceanic plate, = z = 0, and exp(2) = 1, so the heat ow

over crust of age t is given by
T0
qz p
t
(6:114)
The inverse-square-root dependence on age predicted by the half-space

model agrees well with observed oceanic heat-ow values (Fig. 6.6).
Heat-ow data in young sea oor, and where sediment cover is thin, are
systematically biased by hydrothermal circulation, which transports some of the
heat by advection. This can be compensated for by considering only sites that
have sufcient sediment cover and are far enough from basement outcrops that
hydrothermal circulation perturbations are minimal. In particular, Fig. 6.6
shows sites on young sea oor where detailed investigations (seismic imaging
Further reading
197
250
Heat-flow data:
hydrothermal filter
detailed site control
Heat flow (mW m 2 )
200
Cooling model:
plate
half-space
150
100
50
50
100
Age (Ma)
150
Fig. 6.6. Oceanic heat-ow data from all the oceans, plotted versus lithospheric
age. The data have been ltered to exclude sites where sediment thickness is less
than 325 m and those which are within 85 km of a seamount. Solid dots show
median heat ow for 2-Myr age bins; open squares represent high-quality data from
sites where the environment of the site is known from seismic imaging of the sea
oor and other geophysical investigations. The dashed and solid lines represent
heat ow for the half-space and plate cooling models, respectively. After Hasterok
(2010).
of the buried basement topography, closely spaced heat-ow measurements and

proles) have been carried out. The heat-ow values at these sites agree very
well with the predictions of both cooling models. For older oceanic lithosphere
the plate model ts the data more closely than the half-space model and appears
to be a better overall model.
further reading
Anderson, O. L. (2007). Grneisens parameter for iron and Earths core, in
Encyclopedia of Geomagnetism and Paleomagnetism, ed. D. Gubbins and
E. Herrero-Bervera. Dordrecht: Springer, pp. 366373.
Carslaw, H. S. and Jaeger, J. C. (2001). Conduction of Heat in Solids. Oxford: Clarendon
Press, 510 pp.
Jessop, A. M. (1990). Thermal Geophysics. Amsterdam: Elsevier, 306 pp.
ziik, M. N. (1980). Heat Conduction. New York: John Wiley & Sons, 687 pp.
Stein, C. A. (1995). Heat ow of the Earth, in Global Earth Physics: A Handbook of
Physical Constants, ed. T. J. Ahrens. Washington, DC: American Geophysical
Union, pp. 144158.
7
Geomagnetism
The existence of a magnetic force was known for centuries before William Gilbert
pointed out in 1600 that the Earth itself behaved like a huge magnet. Gradually
maps were made of the geomagnetic elements. Systematic investigation of
magnetic behavior was undertaken in the late eighteenth and early nineteenth
centuries. The French scientist Charles Augustin de Coulomb showed experimentally that forces of attraction and repulsion exist between the ends of long thin
magnetized rods, and that they obey rules similar to those determining the
interaction of electrical charges. A freely suspended magnet was observed to
align approximately northsouth; the north-seeking end became known as its
north pole, the opposite end as its south pole. The origin of magnetic force was
attributed to magnetic charges, which, through association, became known as
magnetic poles. Subsequently, it was shown that individual magnetic poles, or
monopoles, do not exist. All magnetic elds originate in electric currents. This is
true even at atomic dimensions; circulating (and spinning) electrical charges
impart magnetic properties to atoms. However, the concept of multiple pole
combinations (e.g., the dipole, quadrupole, and octupole) proved to be very useful
for describing the geometries of magnetic elds.

The most important eld geometry is that of a magnetic dipole. This was
originally imagined to consist of two equal and opposite magnetic poles that
lie innitesimally close to each other (Appendix A2). At distances several times
greater than the size of the source the eld of a very short bar magnet is very
nearly a dipole eld, as is the magnetic eld produced by an electric current in a
small plane loop. In an external magnetic eld B a magnetic dipole experiences
a torque that aligns it with the eld (Appendix A4). The torque is governed by
the relationship
198
199
= mB
(7:1)
In this equation m is the magnetic moment of the dipole, a measure of its

strength. For a current-carrying loop m is equal to the product of the current I
in the loop and its area A, and its direction en is that of the normal to the plane of
the loop (Appendix A4):
(7:2)
m IAen
The dimensions of magnetic moment are by this denition A m2; the dimensions of torque are N m; thus the SI unit of the magnetic eld B, the tesla, has the
dimensions N A1 m1.
The potential W of a dipole magnetic moment m at distance r from its center
and at an azimuthal angle between the dipole axis and the radial direction is
(Appendix A2)
m cos
W 0
(7:3)
4 r2
The constant 0 is the magnetic eld constant. It is dened in SI units to be
exactly 4 107 N A2 (alternatively designated henry m1). The dipole
potential is the most important component of the geomagnetic eld, representing more than 93% of its energy density.
The dipole magnetic eld B is the gradient of the dipole potential: B = W.
In spherical coordinates the eld has a radial component Br and an azimuthal
component B. These are

0 m cos
2m cos
0
(7:4)
Br
2
4
r 4 r
r3

1 0 m cos
m sin
B
0
(7:5)
r 4 r2
4 r3
For a dipole at the center of the spherical Earth, the azimuthal component of
the eld, B, is horizontal. Moreover, if the dipole is aligned with the Earths
axis, the angle is the complement of the magnetic latitude . The direction of
the eld makes an angle I with the horizontal called the inclination of the eld
(see Fig. 7.1(b) and Appendix A, Fig. A1). The inclination, magnetic colatitude, and magnetic latitude are related by
tan I
Br
2 cot 2 tan
B
(7:6)
This equation forms the basis of paleomagnetic determination of ancient paleolatitudes from the inclinations of remanent magnetizations measured in oriented rock samples.
200
Geomagnetism

The empirical laws that govern electricity and magnetism are summarized in
Maxwells equations (Appendix B). Analysis of the present geomagnetic eld
requires Gausss law and Ampres law.
Gausss law established that the net magnetic ux through any closed surface
is zero. This is equivalent to stating that there are no magnetic monopoles:
dipole sources such as current circuits, even at atomic scale, produce zero net
ux through a surrounding surface. The corresponding equation is
rB 0
(7:7)
Ampres law showed that an electric current produces a magnetic eld in the
surrounding space, and it relates the strength of the magnetic eld B to the
electric eld E that causes the current:
r B 0 E 0 0
E
t
(7:8)
The rst term on the right is the electric current associated with the ow of free
charges in a conductor and relies on Ohms law; the second term is the electric
displacement current that results from time-dependent motions of charges
bound to a parent atom. The parameter 0 is the magnetic eld constant, or
permeability of free space, and 0 is the electric eld constant, or permittivity of
free space; is the electrical conductivity of the medium.
In a region that is free of sources of the magnetic eld (such as the space just
above the Earths surface in which the eld is measured), we can assume that
there are no electric or displacement currents, thus
rB0
(7:9)
Consequently, the magnetic eld B can be written as the gradient of a scalar

potential, W:
B rW
(7:10)
On substituting for B in (7.7) the potential W of the Earths magnetic eld is seen
to satisfy Laplaces equation:
r2 W 0
(7:11)
7.2.1 The elds of internal and external origin

The geomagnetic potential at Earths surface arises from two sources. The most
important part of the eld originates in the Earths interior, and the rest
201
originates outside the Earth, e.g., from current systems in the ionosphere. Let Wi
be the potential of the eld of internal origin and We be the potential of the eld
of external origin. The total geomagnetic potential W at Earths surface is
W We Wi
(7:12)
The geomagnetic potential has to be conformable with Earths approximately

spherical geometry, so the solution of (7.11) requires spherical polar coordinates. The general solution of Laplaces equation is therefore as described in
Section 1.16. The variation of potential on a spherical surface is described by
spherical harmonic functions of the co-latitude and longitude . The variation
of potential with radial distance r consists of two parts. In a region where r can
be zero, the potential is proportional to rn. At Earths surface this condition
applies to the eld due to sources outside the Earth, so We must vary as rn. In a
region where r can be very large or innite, the potential is proportional to
1/rn+1. Outside the Earth and on its surface, this applies to the potential of the
eld of internal origin, so Wi must vary as 1/rn+1. These considerations lead to
the following denition for the potential We of the eld of external origin:
We R
1 X
n n
X
r
n1 m0
m
m
5R
Gm
n cosm Hn sinm Pn cos ; r
(7:13)
Similarly, the potential Wi of the eld of internal origin is

Wi R
1 X
n n1
X
R
n1 m0
m
m
gm
n cosm hn sinm Pn cos ; r 4 R
(7:14)
Terms with n = 0 are absent from these expressions because magnetic monopoles do not exist. At the Earths surface the expressions simplify to
We R
1 X
n
X
m
m
Gm
n cosm Hn sinm Pn cos
(7:15)
n1 m0
Wi R
1 X
n
X
m
m
gm
n cosm hn sinm Pn cos
(7:16)
n1 m0
In a convention adopted in 1939 by the scientic body that preceded the

modern International Association of Geomagnetism and Aeronomy (IAGA), it
was agreed to base the spherical harmonic functions in the magnetic potential on
the partially normalized Schmidt polynomials (Section 1.15.2). The coefcients
202
Geomagnetism
m
m
m
(gm
n , hn ) and (Gn , Hn ) are called the Gauss (or GaussSchmidt) coefcients of
the elds of internal and external origin, respectively. They have the dimensions
of magnetic eld and their magnitudes diagnose the relative importance of the
external and internal sources of the eld.
7.2.2 Determination of the Gauss coefcients

It is not possible to measure the geomagnetic potential directly, so the Gauss
coefcients are calculated from measurements of the northward (X ), eastward
(Y ), and vertically downward (Z ) components of the magnetic eld at or above
the Earths surface (Fig. 7.1(a)). These components are related to other geomagnetic elements, such as the horizontal eld (H ), total eld (T ), angle of
inclination (I ), and angle of declination (D), as illustrated in Fig. 7.1(b). The
eld components in spherical polar coordinates are

1 W
(7:17)
X B
r rR

1 W
(7:18)
Y B
r sin rR

W
Z Br
r rR
(7:19)
The differentiations, after evaluating on the Earths surface at r = R, result in

the following set of equations involving the unknown Gauss coefcients:
(a)
Br
geographic magnetic
North
North
N
H
(b)
D
I
B
E
B
V
Fig. 7.1. (a) Relationship between the north (X), east (Y), and vertical (Z)
components of the geomagnetic eld and the spherical polar components Br, B,
and B. (b) The eld may be described by the X, Y, and Z components, or by its
intensity (T), declination (D), and inclination (I). A magnetic compass aligns with
the horizontal component H, which is directed towards magnetic north.
1 X
n
X
n1 m0
203
m
m
m
gm
P cos
n Gn cosm hn Hn sinm
n
(7:20)
1 X
n
X
n1 m0
m m
m
m
gm
P cos
n Gn sinm hn Hn cosm
sin n
(7:21)
Z
1 P
n
P

m
n 1gm
n nGn cosm
n1 m0

m
m
n 1hm
n nHn sinmPn cos
(7:22)
Note that the Gauss coefcients have the same dimensions as the magnetic eld
B, namely tesla. The tesla is a large magnetic eld, so the geomagnetic eld
intensity and the Gauss coefcients are usually expressed in nanotesla (1 nT =
109 T). In the north and east components the Gauss coefcients occur as
m

m
and hm
gn Gm
n
n Hn , and therefore the horizontal components alone
do not allow separation of the external and internal parts. However, the Gauss
coefcients occur in a different combination in the vertical eld, and by virtue of
this the external and internal elds can be separated.
In theory, the summations are over an innite number of terms, but in practice
they are truncated after a certain degree N. The coefcients h0n and H0n do not
exist, because sin(m) = 0 for m = 0, and these terms make no contribution to the
potential. For n = 1 there are three coefcients for the internal eld (g01 ; g11 ; h11 )
and three for the external eld (G01 ; G11 ; H11 ). Similarly, there are ve of each for
n = 2, and in general 2(2n + 1) for degree n. The total number of coefcients Sn
up to and including order N for each part of the eld is
SN 21 1 22 1 23 1 2N 1
2 1 2 3 N N
(7:23)
The sum of the rst N natural numbers is N(N + 1)/2, so the number of
coefcients up to degree and order N of the internal eld is N(N + 2). The same
number is obtained for the external eld. Thus separation requires knowing the
eld values at a minimum of 2N(N + 2) stations.
From 1835 to 1841 Carl Friedrich Gauss and Wilhelm Weber organized the
semi-continuous (every 5 minutes, 24 hr/day) acquisition of data from up to 50
magnetic observatories distributed worldwide, albeit unevenly. Gauss in 1839
carried out the rst analysis of the geomagnetic eld up to degree and order 4, and
established that it is dominantly of internal origin; the coefcients of the external
204
Geomagnetism
eld are small compared with those of the internal eld, and may to a rst
approximation be neglected. The potential of the internal eld is given by (7.14).
Magnetic eld components have historically been measured and recorded at
geomagnetic observatories. A drawback of the data from observatories is their
uneven geographic distribution. A superior global coverage has been obtained
during the last decades with the addition of data from satellites. The coefcients
of the modern geomagnetic eld have now been evaluated reliably up to degree
and order 13. The data are updated and published regularly as the coefcients of
the International Geomagnetic Reference Field (IGRF). The coefcients up to
degree and order 3, corresponding to the dipole, quadrupole, and octupole
components of the eld at the Earths surface are listed in Table 7.1 for some
selected eld models. The terms with n = 1 describe a dipole eld; the higherorder terms with n 2 are referred to collectively as the non-dipole eld.
Table 7.1. Dipole (n = 1), quadrupole (n = 2), and octupole (n = 3) Gauss

Schmidt coefcients from some historical eld analyses. The coefcients DGRF
are for Denitive Geomagnetic Reference Fields that will not be modied
further. Details of the construction of the International Geomagnetic Reference
Field IGRF 2010 are given in Finlay et al. (2010).
Epoch and source
1835,
Gauss,
in 1839
g10
g11
h11
g20
g21
h21
g22
h22
g30
g31
h31
g32
h32
g33
h33
32,350
3,110
6,250
510
2,920
120
20
1,570
1885,
Schmidt,
in 1895
31,730
2,360
5,990
520
2,830
720
680
1,500
940
1,230
300
1,430
30
400
680
1922,
Dyson and
Furner (1923)
30,920
2,260
5,920
890
2,990
1,240
1,440
840
1,140
1,650
460
1,200
120
880
230
1965,
DGRF
30,334
2,119
5,776
1,662
2,997
2,016
1,594
114
1,297
2,038
404
1,292
240
856
165
1985,
DGRF
29,873
1,905
5,500
2,072
3,044
2,197
1,687
306
1,296
2,208
310
1,247
284
829
297
2010,
IGRF
29,496.5
1,585.9
4,945.1
2,396.6
3,026.0
2,707.7
1,668.6
575.4
1,339.7
2,326.3
160.5
1,231.7
251.7
634.2
536.8
205

The dominant component of the Earths surface magnetic eld is the dipole
component. The axis of the dipole is inclined to the rotation axis, thus it can be
separated into an axial dipole and two orthogonal equatorial dipoles. As we will
see, shifting these dipoles from the center of the Earth generates higher-order
components in the geomagnetic potential.
7.3.1 The geocentric axial dipole

Each term in the geomagnetic potential (7.14) represents the potential of a particular pole conguration. The potential described by the largest coefcient, g01 , is
W 01
R3 g01 0
R3 g01 cos
P1 cos
2
r
r2
(7:24)
Comparison with (7.3) shows that this is the potential at distance r from the midpoint of a magnetic dipole and at angle from the dipole axis. In Earth coordinates this is the potential at co-latitude of a geocentric dipole aligned with the
rotation axis and pointing to the north pole with magnetic moment m given by
m
4R3 0
g
0 1
(7:25)
The magnetic eld of an axial dipole is horizontal at the equator (see (7.4) and
(7.5)). Its value at Earths surface is

1 R3 g01 cos
B
g01 sin
(7:26)

r
r2
rR
At the equator this is equal to g01 .
7.3.2 The geocentric inclined dipole

The coefcients of degree n = 1 and order m = 1 also have an inverse-square
dependence on distance, so g11 and h11 too must represent dipoles. The combined
potential of the dipole terms is
2

R 0 0
W1 R
g1 P1 cos g11 cos h11 sin P11 cos
(7:27)
r
W1 R
2

R 0
g1 cos g11 cos sin h11 sin sin
r
(7:28)
206
Geomagnetism
geographic
pole
magnetic
pole
( 0 , 0 )
site
P( , )
=0
Greenwich
meridian
equator
Fig. 7.2. Angular relationships pertaining to the computation of the potential of an

inclined geocentric magnetic dipole.
Consider now the direction cosines of a line OP inclined at angle to the

reference axis and at angle to the reference axis = 0, as in Fig. 7.2. The
direction cosines (, , ) of OP are
sin cos
sin sin
(7:29)
cos
Suppose the axis of a magnetic dipole to be inclined at angle 0 to the z-axis and
at angle 0 to the reference axis = 0. The direction cosines (0, 0, 0) of the
dipole axis are
0 sin 0 cos 0
0 sin 0 sin 0
0 cos 0
(7:30)
If is the angle between OP and the dipole axis, and r the distance of P from the
dipole center, the magnetic potential at P is
W1
0 m
m
cos 0 2 0 0 0
4r2
4r
(7:31)
The components of the dipole moment m along the reference axes (Fig. 7.3) are
207
z
( 0 , 0 , 0 )
mz
m
0
y
my
mx
Fig. 7.3. Relationship between the Cartesian components and direction cosines of
a magnetic dipole m, which is inclined at angle 0 to the rotation axis and has an
azimuth 0 in the equatorial meridian.
mx m cos x m0
my m cos y m0
mz m cos 0 m0
The potential of the inclined dipole becomes

W1 0 2 mx my mz
4r
(7:32)
(7:33)
Using the relationships in (7.29), the potential of the inclined dipole is

(7:34)
W1 0 2 mz cos mx cos sin my sin sin
4r
On equating individual terms with the expression for the potential using Gauss
coefcients (7.28) it is evident that the coefcients g11 and h11 represent orthogonal dipoles in the equatorial plane. The equatorial dipole components are
mx
4R3 1
g
0 1
(7:35)
my
4R3 1
h
0 1
(7:36)
4R3 0
g
0 1
(7:37)
The axial component of the dipole is

mz
208
Geomagnetism
The points where the dipole axis intersects the Earths surface are called the
geomagnetic poles (Fig. 7.2). At these points the dipole magnetic eld is normal
to the surface. The geomagnetic poles are antipodal to each other, because they
lie at the opposite ends of the inclined axis. The co-latitude 0 of the pole is
equal to the tilt of the inclined axis. From (7.30) and (7.32)
q 4 q
2 2
(7:38)
m sin 0 m2x m2y R3 g11 h11
0
Together with the axial component, this denes the tilt 0 of the dipole axis,
which is also the co-latitude of its pole:
q
1 2 1 2
g1 h1
m sin 0
tan 0
(7:39)
m cos 0
g01
The components of the dipole moment in the equatorial plane, mx and my, dene
the longitude 0 of the pole. From (7.35) and (7.36)
tan 0
0 my h11
0 mx g11
(7:40)
The dipole magnetic moment m is obtained by squaring and summing mx, my,
and mz, giving
q
0 2 1 2 1 2
4
g1 g1 h1
(7:41)
m R3
0
Analysis of the geomagnetic eld for epoch 2010 (Finlay et al., 2010) locates
the north geomagnetic pole at 80.08 N, 287.78 E and the south geomagnetic
pole at 80.08 S, 107.78 E. The places where the total magnetic eld of the
Earth is normal to the surface are the magnetic dip poles. The total eld is
expressed by all the terms in (7.14). Because of the non-dipole components the
magnetic dip poles are not antipodal; also, because of secular variation
(Section 7.4) the pole locations change slowly with time. For epoch 2010, the
north dip pole was at 85.01 N, 227.34 E; the south dip pole was at 64.43 S,
137.32 E, which is outside the Antarctic Circle.
7.3.3 Axial dipole with axial offset

The terms with n = 2 are referred to as the quadrupole component of the eld.
However, one must keep in mind that the multipole expression of the magnetic
eld is a mathematical convenience that simply allows us to subdivide it for
convenient reference. That is, just as there are no physical magnetic dipoles
(a)
D
r
S
d
O
(b)
P
209
Fig. 7.4. (a) Geometry for calculation of the potential at P of an axial magnetic
dipole at D, displaced a distance d along the rotation axis from the Earths center at
O. (b) The similar case of an axial magnetic dipole displaced in the equatorial plane.
inside the Earth, there are also no quadrupoles; a complex system of electric
currents deep in the Earth causes the magnetic phenomena that we measure. The
n = 2 coefcients are responsible for an offset of the magnetic dipole from the
Earths center. This can be shown as follows.
Let the axial magnetic dipole be displaced a small distance d along the dipole
axis, as in Fig. 7.4(a). The position P is now a distance u from the center of the
dipole at D, and the line DP makes an angle with the dipole axis. The dipole
potential at P is now
W
0 m cos
4 u2
(7:42)
The line DP makes a small angle with the radius OP of length r. In the triangle
ODP = + , so
cos cos cos cos sin sin

d
u2 r2 d2 2rd cos r2 1 2 cos
r
(7:43)
(7:44)
In the triangle SDP, created by drawing DS perpendicular to OP,

sin
DS d sin
u
u
(7:45)
For a very small displacement d r, the distances r and u are almost equal, so
the following relationships are approximately true to rst order:
sin
d sin
r
cos cos
and
d
sin2
r
cos 1
(7:46)
(7:47)
210
Geomagnetism
The potential of the axially displaced dipole may now be written

0 m cos d=r sin2
W
4r2 1 2d=r cos
(7:48)
Using the binomial expansion and truncating it after the rst order in d/r,

m
d
d
W 0 2 cos sin2 1 2 cos
(7:49)
4r
r
r

0 m
d
d
2
2
cos sin 2 cos
(7:50)
W
4r2
r
r
W

0 m
md
cos 0 3 3 cos2 1
2
4r
4r
(7:51)
W
0 m 0
2md
P cos 0 3 P02 cos
4r2 1
4r
(7:52)
The rst term is the potential of a geocentric axial dipole; the second term is that
of a geocentric axial quadrupole. An axial displacement of the dipole is equivalent to introducing the quadrupole term. The two terms are equivalent to the g01
and g02 terms in (7.14) for the multipole expansion of the potential.
7.3.4 Axial dipole with equatorial offset

To determine the effect of displacing the center of the axial dipole in the
equatorial plane, we use the same approach as in the previous section. The
geometry is as in Fig. 7.4(b) and the potential at P is, as before,
W
0 m cos
4 u2
(7:53)
With reference to the triangle ODP, we now have = , so

cos cos cos cos sin sin

d
2
2
2
2
r 1 2 sin
u r d 2rd cos
2
r
(7:54)
(7:55)
For a very small displacement d r, the distances r and u are almost equal, so to
rst order
sin
d sin
r
and
cos 1
(7:56)
d
sin2
r
In the triangle SDP, created by drawing DS perpendicular to OP,

DS d
d
sin cos
sin
u
u
2
r
cos cos
211
(7:57)
(7:58)
Using the binomial expansion and truncating it after the rst order in d/r, the
potential of the equatorially displaced dipole (7.53) may now be written

,
0 m
d
d
2
sin
sin
cos
1

2
W
4r2
r
r
(7:59)

0 m
d
d
2

cos sin 1 2 sin
4r2
r
r

m
d
d
W 0 2 cos 2 sin cos sin2
(7:60)
4r
r
r
W
0 m
2md
md
cos 0 3 sin cos 0 3 sin2
2
4r
4r
4r
(7:61)
Reference to Table 1.2 shows that the angular dependence of each term can be
replaced by an associated Legendre polynomial, which gives
W
0 m 0
2md=3 1
md=3 2
P cos 0
P2 cos 0
P2 cos (7:62)
4r2 1
4r3
4r3
As before, the main term is the centered axial dipole. The additional terms result
from the equatorial displacement, and are equivalent to the terms governed by
coefcients g12 and g22 in (7.14).
7.3.5 Best-tting eccentric inclined dipole

The best t of a dipole to the observed magnetic eld is obtained with an
eccentric inclined dipole centered a few hundred kilometers from the center of
the Earth (Box 7.1). To compute the offset of the dipole it is necessary to use all
terms of degree and order n 2 in the multipole expansion of the potential.
Using the Gauss coefcients for IGRF 2010 (Table 7.1), the location of the besttting eccentric inclined dipole has displacements x0 = 400 km, y0 = 208 km, z0
= 210 km, r0 = 498 km; i.e., it lies north of the equator under the North Pacic
Ocean (Fig. 7.5) at 25 N, 153 E. The location of the eccentric dipole based on
Quaternary and Recent paleomagnetic data and deep-sea cores (Creer et al.,
212
Geomagnetism
Box 7.1. The eccentric dipole

The geomagnetic eld is dominantly that of a dipole. The question
naturally arises as to the location of the dipole that best ts the present
eld. Several methods of nding the optimum position have been
summarized by Lowes (1994). The most commonly used is a method that
was developed in 1934 by A. Schmidt, which yields the equations below
(Schmidt, 1934).
The tilt of the dipole axis is determined by the Gauss coefcients of rst
degree, n = 1. The best-tting dipole is not centered at the center of the
Earth but is displaced to a position with coordinates (x0, y0, z0), where z0 is
the shift of the dipole center along the rotation axis, x0 the shift in the
direction of the Greenwich meridian, and y0 the shift orthogonal to these
displacements. The displacements can be determined approximately using
all Gauss coefcients with n 2; a more exact solution requires the n = 3
coefcients as well. The following equations describe the location of the
dipole center in a spherical Earth with radius R for n 2:

x0 R L1 g11 E =3m2

y0 R L2 h11 E =3m2

z0 R L0 g01 E =3m2
2 2 2
m2 g01 g11 h11

E L0 g01 L1 g11 L2 h11 =4m2

p
L0 2g01 g02 g11 g12 h11 h12 3

p
L1 g11 g02 g01 g12 g11 g22 h11 h22 3

p
L2 h11 g02 g01 h12 h11 g22 g11 h22 3
The displacement r0 of the center of the eccentric inclined dipole from the
center of the Earth is
q
r0 x0 2 y0 2 z0 2
1973) was found to be offset by about 200 km in the same direction, suggesting
the existence of persistent non-axial components in the global eld.
It is important to remember that the multipole method of expressing the
geomagnetic potential is a mathematical convenience. In reality there are no
213
rotation
axis
Greenwich
meridian
x
z0
y0
m
x0
North
Pacific
equator
y
East
Fig. 7.5. The location of the best-tting eccentric dipole for IGRF 2010 is offset
into the northern hemisphere and the Pacic hemisphere. The orientation of the
dipole is not changed by the offset.
dipoles, quadrupoles, or other multipoles. However, these concepts provide a

convenient way of visualizing the geometry of parts of the eld. As noted
above, a displacement of a dipole from the center of coordinates creates higherorder terms in the multipole expansion. Thus it is possible to model the eld
with a moderate number of displaced dipoles. If each dipole corresponds to a
current loop, this type of model may be physically more realistic. However, it is
not practical for a mathematical description of the eld.

The Gauss coefcients are not constants but change slowly with time, a
phenomenon known as the secular variation of the eld. Both the dipole and
the non-dipole parts of the eld exhibit secular variations. The dipole secular
variations can be illustrated graphically by plotting the strength of the dipole
magnetic moment and the orientation of the dipole axis, expressed as the
latitude and longitude of a geomagnetic pole (Fig. 7.6). The timescale of dipole
secular variations is of the order of thousands of years. The strength of the
dipole magnetic moment has declined steadily over the past 150 years, during
which observatory measurements of the eld have been made. In the same time
interval, the tilt of the dipole axis changed little until about 1960, but has since
been decreasing. Similarly, the longitude of the geomagnetic pole was steady
until the middle of the twentieth century, but has since been decreasing; this
corresponds to a westward motion of the dipole axis around the rotation axis.
Longitude of pole ( E)
Geomagnetism
12
32
Tilt of axis ( )
Dipole moment (T)
214
31
30
1900
Year
2000
11
10
1900
2000
Year
292
290
288
1900
2000
Year
Fig. 7.6. Geomagnetic secular variations: the dipole magnetic moment, the tilt of
the dipole axis relative to the rotation axis, and the longitude of the geomagnetic
pole.
When the dipole component is subtracted from the total eld, the remainder
described by the Gauss coefcients with n 2 is called the non-dipole eld.
Maps of the non-dipole eld are characterized by large positive and negative
anomalies that can have amplitudes amounting to a large fraction of the dipole
eld. These anomalies have a cell-like appearance, and change position and
intensity with time. The non-dipole eld has a standing (stationary) part, which
exhibits intensity uctuations without signicant displacement, and a drifting
(mobile) part. The best-known feature is a westward drift of many of the
mapped cells at an average rate of about 0.3 per year.

The depth of the sources of the geomagnetic eld of internal origin can be
determined from the power spectrum of the Gauss coefcients. The power (or
energy density) n associated with the coefcients of degree n at the Earths
surface is given by Lowes (1966, 1974):
<n n 1
n
X
gm
n
2
2
hm
n
(7:63)
m0
The term of degree n in the geomagnetic potential varies with radial distance r as
r(n+1), so the strength of the eld varies as r(n+2). The power, or energy density,
is proportional to the square of the amplitude, and thus varies as r2(n+2). If the
m
coefcients gm
n and hn have been determined on the surface of a sphere of radius
r, the power spectrum on a surface of radius R closer to the center of the Earth is
found by augmenting the spectrum by the ratio (r/R)2(n+2). The process is called
downward continuation. The power spectrum on the surface of radius R is then
given by
215
2n2
r
<n R
<n r
R
(7:64)
2n2 X
n
2 m 2
r
<n R n 1
gm
hn
n
R
m0
(7:65)
The satellite MAGSAT measured the magnetic eld at an average altitude of

420 km, equal to a radial distance of r = 6,791 km. The large quantity of data
allowed harmonic analysis up to degree n = 63. The power spectrum based on
the Gauss coefcients derived from the MAGSAT data is shown in Fig. 7.7
(lower curve). The n = 1 dipole term lies disproportionately above the other
terms. On a semi-logarithmic plot the data form two almost linear segments,
above and below n = 14. The part of the spectrum with n 14 is attributed
mainly to sources in the core; the part with higher values of n arises from
sources mainly in the crust; the signal above n 50 was considered to be noise,
which averaged 0.091 nT2 per degree. The two parts of the spectrum overlap
around the break in slope.
The upper curve in Fig. 7.7 shows the data after downward continuation to
the Earths surface (radius R = 6,371 km). Note that the slope of the line for
1010
108
106
Wn (nT2)
104
at Earth s surface
102
optimized
100
102
0
at 420 km altitude
10
20
30
40
Degree n
50
60
Fig. 7.7. The energy intensity associated with each degree of the sphericalharmonic analysis of the geomagnetic eld, from measurements by the
MAGSAT satellite at altitude 420 km, after reduction to the Earths surface. Data
source: Cain et al. (1989).
216
Geomagnetism
core sources (n 14) is atter than that at altitude 420 km. This suggests that
if downward continuation is carried out to even deeper surfaces the slope
might become zero. For n > 15 the slope of the line becomes positive. This is
because downward continuation amplies preferentially higher frequencies,
including the noise inherent in the measured signal. When the noise is
removed, the downward-continued spectrum at the Earths surface is almost
at for n > 15 (the smooth curve in Fig. 7.7). The data after removing the
average noise (and without the dipole term) can be tted by a continuous
curve with equation
<n 9:66 108 0:286n 19:10:996n
(7:66)
7.5.1 Estimation of the source depth of the main eld

A method for estimating the approximate depth of the source layer of a magnetic
or gravity anomaly is to assume that the power spectrum is white at that level
(i.e., every part of the spectrum has the same amplitude). This can be applied to
the non-dipole core eld, for which
<n 9:66 108 0:286n
(7:67)
The power of a signal is dened to be the square of its amplitude. Thus the term
of degree n in the power spectrum has amplitude
p
Bn <n 3:108 104 0:535n
(7:68)
The ratio of the amplitudes of successive terms is
Bn1
0:535
Bn
(7:69)
The Gauss coefcients in the power spectrum of the internal eld are dened
from the solution of Laplaces equation given in (7.14). The amplitude of the nth
term in the potential varies with radial distance according to
n1
R
Wn / Bn
r
The ratio of successive terms in the potential is then

Wn1 Bn1 R
R
0:535
r
r
Wn
Bn
(7:70)
(7:71)
7.6 Origin of the internal eld
217
If the power spectrum becomes white, then all terms in the potential are equal,
Wn = Wn+1, and
r 0:535R
(7:72)
This result locates the source layer of the non-dipole terms (2 n 14) at a radial
distance of about 3,400 km. The radius of the core is 3,480 km, thus the source
depth of the non-dipole terms is in the outer core, close to the coremantle
boundary.
The power spectrum at the Earths surface, corrected for noise (solid line in
Fig. 7.7), is almost at above n = 15, signifying that the source layer of this part
of the spectrum is very close to the surface and hence can be associated with
crustal sources.
7.6 The origin of the internal eld

William Gilberts concept in 1600 of the Earth as a giant permanently magnetized sphere proved to be unrealistic in light of later knowledge of rock magnetic
properties and the internal structure of the Earth. The magnetic eld of a
geocentric axial dipole is horizontal at the magnetic equator, where its strength
Be on the surface r = R is
Be
0 m sin=2
03m
R3
4
4R
(7:73)
The magnetization M is equal to the magnetic moment m per unit volume, so

Be
0 4R3
M 0M
3
4R 3
3
(7:74)
The equatorial eld is equal to g01 (i.e., ~30,000 nT), which gives a mean
magnetization of 70 A m1. This greatly exceeds the magnetization of the
most common strongly magnetized rocks (M is about 1 A m1 in basalt).
Moreover, it does not take into account that the temperature inside the Earth
soon exceeds the Curie temperature of magnetic minerals, above which no
permanent magnetization is possible, so only the thin outer shell could be
permanently magnetized. This would require an even greater magnetization
than that calculated. Finally, the concept of a permanent magnet does not
account for the observed secular variation of the magnetic eld.
The experiments of Ampre and rsted in the early nineteenth century
showed that magnetism was caused by electric currents. It is reasonable to
assess whether the geomagnetic eld has an electromagnetic origin.
218
Geomagnetism
7.6.1 Electromagnetic model

Maxwells equations of electromagnetism (Appendix B) lead to an electromagnetic model for generation of the geomagnetic eld in the uid Earths
core. The electrical conductivity of the liquid-iron outer core is estimated to be
about 5 105 1 m1 (Stacey and Anderson, 2001), which makes it a good
conductor. Any free charges would rapidly dissipate, so the free charge density
in Coulombs law (Appendix B, part 1) is zero. A comparison of the magnitudes of the two terms on the right of Ampres law (Appendix B, part 2) for a
periodic variation with angular frequency = 2/ gives
jD=tj 0 jE=tj 0 jiEj 20
jEj
jE j
jJj
(7:75)
The electric eld constant is 0 = 8.854 1012 C2 N1 m2 and the approximate

conductivity of the core is = 5 105 1 m1. For a period longer than a year
(3.15 107 s), the ratio in (7.75) is less than 1024. Thus the displacement
current D/t can be ignored in the core. Maxwells equations for the core
become
rE 0
Coulombs law
r B 0 J
rB 0
rE
Ampres law
Gausss law
B
t
Faradays law
(7:76)
(7:77)
(7:78)
(7:79)
Taking the curl of both sides of (7.77) gives

r r B 0 r E
(7:80)
Substituting on the right from (7.79) gives

r r B 0
B
t
(7:81)
Using the vector identity of (1.34), the left-hand side can be expanded, giving
rr B r2 B 0
B
t
(7:82)
The rst term can be eliminated because of Gausss law, leaving

r2 B 0
B
t
(7:83)
219
B
1 2
r B m r2 B
t 0
(7:84)
This differential equation has the same form as the diffusion equation (6.66),
and the parameter m = 1/(0) is called the magnetic diffusivity.
The magnetic eld B must satisfy Gausss law, having a solution such as
B rW r A
(7:85)
In this solution the scalar potential W is the familiar solution of Laplaces

equation, whereas A is a vector potential that must be added because of the
vector identity that the divergence of the curl of a vector is always zero (see
(1.33)). The scalar potential can be used for a magnetic eld in a region that is
free of electric currents (such as the description of the geomagnetic eld using
Gauss coefcients). A vector potential is appropriate to describe a eld that
arises from electric currents. If we insert this solution into (7.84) we get
rW r A m r2 rW r A
t

W
2
m r W r A m r2 r A
r
t
t
(7:86)
(7:87)
Both sides of this equation have the same form as the thermal conductivity
equation, if each side is set to zero. The solutions depend on space and time, and
can be obtained by separating the variables with appropriate boundary
conditions.
In a three-dimensional problem this can be complicated, but we can get an
order-of-magnitude solution by considering a one-dimensional case. Let the
scalar equation depend only on x and t,
W
2 W
m 2
t
x
(7:88)
This is a magnetic equivalent of the heat-conduction equation (Section 6.6.2). A

possible solution is

x
W W0 sin 2n
expt=
(7:89)
L
The quantity L is a length that is characteristic of the problem. It may be
comparable to the size of the outer core, for example. The magnetic potential
W decays exponentially; the quantity is a relaxation time, over which the eld
sinks to 1/e of its initial value. Upon inserting the solution into (7.88) and taking
the fundamental mode of the distance dependence (n = 1) we get
220
Geomagnetism
1
4 2 m
W
W
L2
(7:90)
This gives the relaxation time in terms of other core parameters:
0 L2
4 2
(7:91)
The electrical conductivity of the core is approximately 5 105 1 m1 and

0 = 4 107 N A2, so, taking a characteristic length L = 2,000 km, the
relaxation time is 6.4 1010 s or about 2,000 yr. In a time equal to 5 an
exponential function sinks to less than 1% of its initial value, so the magnetic
eld generated by a purely electromagnetic model would disappear in about
10,000 years. Magnetizations in ancient rocks show that the Earth has had a
magnetic eld since the Pre-Cambrian, i.e., for times on the order of 109 yr, so
the electromagnetic model is inadequate. A satisfactory model must be capable
of sustaining a magnetic eld for this long.
A further mechanism is needed to regenerate the magnetic eld and prevent it
from diffusing away. This is provided by physical motion of the electrically
conducting core uid, which interacts with the magnetic eld lines in the core.
The mechanism is analogous to that of a dynamo, in which a coil of wire is
moved through the eld of a magnet to create an electric current in the wire. The
process of generating the geomagnetic eld by induction from the motion of the
conducting core uid is known as the dynamo model.
7.6.2 The magnetohydrodynamic model

When an electrical charge q moves with velocity v through a magnetic eld B, it
experiences the Lorentz force F, which is normal to the eld and to the direction
of motion (Appendix A3):
F qv B
(7:92)
In the case of the Earths core it gives rise to an additional electric eld EL given by
EL
F
vB
q
(7:93)
The total electric eld experienced by the material of the core is now Et = E +
EL, and for Ohms law we get
J Et E EL E v B
Ampres equation (7.77) becomes
(7:94)

r B 0 J 0 E v B
221
(7:95)
With the additional term we now proceed as for the electromagnetic model,
taking the curl of both sides of the equation:
r r B 0 r E r v B

B
2
rvB
rr B r B 0
t
(7:96)
(7:97)
The rst term is zero because of Gausss law; rearranging the other terms gives
B
m r2 B r v B
t
(7:98)
This is known as the magnetohydrodynamic induction equation. The constant m is the magnetic diffusivity, as before. As a result of the additional term
on the right, the magnetic eld no longer decays exponentially with time. The
rst term describes the tendency of the eld to decay by diffusion; the second
term provides additional energy to regenerate the eld from the interaction of
the eld with the motion of the conducting uid. The ratio of the terms on the
right is called the magnetic Reynolds number, Rm, dened as
Rm
jr v B j
jm r2 Bj
(7:99)
The magnetic Reynolds number is dened by analogy with uid mechanics,

where the Reynolds number is a property of a uid that determines the predominance of laminar ow or turbulent ow. At low Reynolds numbers viscous forces
are dominant, and the ow is laminar; at high Reynolds numbers inertial forces
result in turbulent ow, which is less stable and typied by random eddies. For a
magnetic Reynolds number Rm 1, the magnetic eld simply diffuses away by
ohmic dissipation as in the electromagnetic example discussed in the previous
section. If Rm 1, the magnetic-eld lines are carried along by the conducting
uid and the uid motion predominates in the generation of the eld.
We can use dimensional analysis to estimate the magnitude of Rm in the core.
The dimension of a gradient is [L]1, we can write [B] for the dimension of the
eld, and the magnetic diffusivity m = 1/(0). Thus
Rm
jr v Bj 0 L1 vB

m j r 2 B j
L2 B
Rm 0 vL
(7:100)
(7:101)
222
Geomagnetism
The quantities v and L are not known precisely. L is an unspecied length

assumed to be typical for a core motion; we may use the same value as before for
the core, i.e., L = 2,000 km. The velocity v of the conducting uid has
been estimated from the westward motion of eld features to be on the order
of 1020 km yr1, i.e., v 0.30.6 mm s1. This gives a magnetic Reynolds
number of about 250500. Even slower motions of the core give Rm 1, so to a
rst approximation we can ignore the diffusive term and write
B
rvB
t
(7:102)
This equation would be exactly true for a material with innite conductivity,
but the nite conductivity of the core means that there is some leakage of the
magnetic ux. However, the assumption of innite conductivity allows deeper
insight into the generation of the geomagnetic eld.
7.6.3 The frozen-ux theorem

Let S be a surface bounded by a closed loop L in an electrically conducting uid
at time t, and let B(t) be a magnetic eld cutting S (Fig. 7.8). If dS is an element
of the surface area, the magnetic ux 0 through S is
Z
(7:103)
0 Bt dS
S
Suppose that the conducting uid moves with velocity v. In a short time increment
t the loop is displaced through a small distance dx = v t. This denes a cylinder
of volume V with a total surface area A, made up of (1) the bottom surface with
nT
T, B(t + t)
LT
Q
n
dx
nQ
S, B(t)
L
dl
Fig. 7.8. Conguration for derivation of the frozen-ux theorem. At time t the
magnetic eld B(t) intersects a surface S moving with velocity v through a
conducting uid; at time t + t the eld has changed to B(t + t) and the surface
area has changed to T. Relative to the enclosed volume, the normal directions nT
and nQ to surfaces T and Q are outward; the normal direction n to the bottom
surface S is inward.
223
area S bounded by loop L, (2) the top surface with area T bounded by loop LT, and
(3) the side surfaces with area Q. During the elapsed time t the magnetic eld
itself changes to B(t + t). The ux 2 through the top surface T is
Z
2 Bt Dt dS
(7:104)
T
We can apply the divergence theorem (Section 1.6) and Gausss law for magnetism to the volume V cut by the eld lines of B. At any time
Z
Z
B dS r BdV 0
(7:105)
V
The integration on the left is the ux of the magnetic eld through all the
surfaces bounding the volume V. It can be written as the sum of the ux through
each end surface plus the ux through the side surface: thus, at time t + t,
Z
Z
Z
Bt Dt dS Bt Dt dS Bt Dt dS 0 (7:106)
S
The negative sign in the rst term is necessary because the normal direction to
each surface is outward, but we have dened the ux of the eld to be inward
across S and outward across T. On rearranging terms, the ux across the top
surface T is given by
Z
Z
Z
2 Bt Dt dS Bt Dt dS Bt Dt dS (7:107)
T
The change in ux has two causes: the rst is the change in the magnetic eld
with time, and the second is the change of surface area through which the eld
passes. If the time t is short, we can write the rst term on the right to rst
order as
Bt Dt Bt
Bt
Dt
t
(7:108)
Upon inserting this into (7.107) we have

Z
2
Z
Bt dS Dt
Bt
dS
t
The change in ux through the moving loop is
Z
Bt Dt dS
Q
(7:109)
224
Geomagnetism
Z
D 2 0 Dt
S
Bt
dS
t
Z
Bt Dt dS
(7:110)
The ux through the side surfaces must now be calculated. In time t the
displacement parallel to the local velocity vector of the uid is dx = v t.
Together with an incremental distance dl along the loop L, this displacement
denes an element of the surface Q with area
d S d l dx d l vDt
(7:111)
Thus the magnetic ux across the side surface Q is

Z
Z
Bt Dt dS Dt Bt Dt d l v
Q
(7:112)
We can change the variable of integration by using the vector identity in (1.18).
The surface integration over Q is converted into a linear integration along dl,
i.e., around the closed loop L:
Z
Z
Bt Dt dS Dt v Bt Dt d l
(7:113)
L
Now we again use (7.108) to replace B(t + t) by B(t) and its time-derivative:

Z
Z
Bt
Bt Dt dS Dt
v Bt
Dt d l
t
L
Q

Z
Z
Bt
v Bt d l Dt2
Dt
v
dl
t
L
(7:114)
By inserting this expression into (7.110) we obtain the change in ux in time t:

Z
Z
Z
Bt
Bt
2
v Bt d l Dt
v
dS Dt
dl
D Dt
t
t
L
(7:115)
On dividing throughout by t, we have

Z
Z
Z
D
Bt
Bt
dS v Bt d l Dt
d l (7:116)
v
Dt
t
t
S
Further reading
225
The rate of change of magnetic ux is the limit of this expression as t tends to

zero; the nal term disappears and
Z
Z
d
D
Bt
lim
dS v Bt d l
(7:117)
Dt0 Dt
dt
t
L
The integral around the closed loop L can be converted into an integral over the
open bounded surface S by applying Stokes theorem (Section 1.7):
Z
Z
v Bt dl r v Bt dS
(7:118)
L
The rate of change of magnetic ux through the closed loop L is therefore

Z
d
Bt
r v Bt dS
(7:119)
dt
t
S
If the electrical conductivity of the moving uid is innite, the approximation in

(7.102) applies, and the expression in brackets is zero. Therefore,
d
0
dt
(7:120)
and
Z
Bt dS constant
(7:121)
This result states that the magnetic ux in a uid with innite electrical
conductivity does not change as the uid moves. This is known as the frozenux (or frozen-in-ux) theorem. It was formulated in 1943 by H. Alfvn, a
Swedish physicist, for an electrically conductive plasma (such as the solar
wind). The theorem can be applied as an approximation for any conducting
uid with a high magnetic Reynolds number, such as the Earths liquid core. It
describes how, in an ideal case, magnetic eld lines are trapped by the high
conductivity and compelled to move with the uid. As a result, uid motions in
the core, in particular thermally and compositionally driven convection, provide
the energy source and feedback mechanism for a self-sustaining magnetic eld.
further reading
Campbell, W. H. (2003). Introduction to Geomagnetic Fields. Cambridge: Cambridge
226
Geomagnetism
Gubbins, D. and Herrero-Bervera, E. (2007). Encyclopedia of Geomagnetism and

Paleomagnetism. Dordrecht: Springer, 1,054 pp.
Merrill, R. T., McElhinny, M. W., and McFadden, P. L. (1996). The Magnetic Field of the
Earth: Paleomagnetism, the Core, and the Deep Mantle. San Diego, CA: Academic
Press, 527 pp.
8
Our knowledge of Earths internal structure has been obtained from detailed
analysis of the travel-times of seismic waves in the Earth. A standard model of
the layered interior PREM, the Preliminary Reference Earth Model
(Dziewonski and Anderson, 1981) that gives the variations with depth of
seismic velocities, density, pressure, and elastic parameters has been derived.
This chapter handles the dependence of seismic-wave velocities on the elastic
properties of the medium in which they are transmitted.
The propagation of a seismic wave takes place by innitesimal elastic displacements of the material it passes through. An elastic displacement is reversible, i.e.,
after the disturbing force has been removed the material returns to its original
condition. The elastic properties and density of the material determine the type of
wave that passes through it, and the speed with which the wave travels.
8.1 Elastic deformation

Elastic deformation is governed by Hookes law, which was formulated in the
seventeenth century on the basis of empirical observations. These are illustrated
by the deformation of a rod of length x and cross-sectional area A, which extends
by an amount x due to an applied force F (Fig. 8.1). In an elastic deformation
the fractional increase in length (x/x) is directly proportional to the applied
force F and inversely proportional to its cross-section A:
x F
/
x
A
(8:1)
Stress and strain are dened for a small volume of a continuous medium as
limiting cases when the volume shrinks to zero, i.e., when both the length x and
the cross-sectional area A become very small. The limit of the force per unit area
(F/A) is the stress, , which has the units of pressure (pascal):
227
228
x
x
F
A
Fig. 8.1. Extension of a rod of length x and cross-sectional area A due to an

applied force F.

F
lim
A!0 A
(8:2)
The limit of the fractional change in dimension (x/x) is the strain, , which is
dimensionless:

x
(8:3)
lim
x!0 x
Hookes law states that in an elastic deformation the stress and strain are
proportional to each other:
/
(8:4)
The law describes the initial deformation of a material; the stressstrain relationship is linear, and the behavior is said to be perfectly elastic. If the stress
increases continuously, the linearity breaks down, but the behavior is still elastic
and no permanent deformation results (Fig. 8.2). Eventually the limit of elastic
behavior is reached, permanent deformation results, and nally failure occurs.
The propagation of seismic waves takes place within the elastic range of
behavior.
8.2 Stress
The forces acting on an elastic body can be divided into body forces (e.g.,
gravity, centrifugal force) and surface forces (e.g., pressure, tension, and shear).
Imagine a small volume V bounded by a surface S within a continuous larger
body of uniform density . The body forces acting on V (including inertial
forces) produce acceleration of V and of the body as a whole. The material
surrounding V exerts inward forces on the surface S; to maintain equilibrium,
equal and opposite surface forces act outwards across S. They cause the small
volume to change shape and dene the state of stress in the body.
8.2 Stress
229
elastic
plastic
deformation
deformation
failure
elastic
limit
Stress
Hooke s
law
linear
range
Strain
Fig. 8.2. Hypothetical stressstrain relationship, showing the regions of elastic and
plastic deformation, and the linear range within which Hookes law holds.
x3
x 2
A3
x1
F3
x3
x1
x2
F1
F2
A1
A2
Fig. 8.3. Denitions of the quantities involved in calculating the components of

stress caused by force components F1, F2, and F3 acting on the sides of a small
rectangular box with surface areas A1, A2, and A3, respectively.
The denition of components of stress is illustrated for a small rectangular

box. Let F be a force with components F1, F2, and F3 referred to orthogonal
Cartesian coordinate axes x1, x2, and x3, respectively. F acts upon the surfaces of
a small rectangular box with sides parallel to the reference axes (Fig. 8.3). The
direction of each component of F is normal to one of the surfaces and tangential
to the other two. The orientation of each surface is specied by its outward
normal, and the respective areas are A1, A2, and A3.
The component of force F1 normal to the surface A1 produces a normal stress,
denoted 11. The components F2 and F3 tangential to the surface A1 result in
shear stresses 12 and 13. The three components of stress acting on the surface
A1 are dened as
230
x3
33
32
31
23
22
13
11
21
12
x2
x1
Fig. 8.4. Denition of the components of normal and shear stress.

11 lim
A1 !0

F1
;
A1

12 lim
A1 !0

F2
;
A1

13 lim
A1 !0
F3
A1

(8:5)
Similarly, the components of F acting on the surface A2 dene a normal stress

22 and shear stresses 21 and 23, while the components of F acting on the
surface A3 dene a normal stress 33 and shear stresses 31 and 32 (Fig. 8.4).
The nine components kn (k = 1, 2, 3; n = 1, 2, 3) form the elements of the stress
tensor, which in matrix form is
0
1
11 12 13
kn @ 21 22 23 A
(8:6)
31 32 33
In each case the rst index of a stress element identies the orientation of a
surface and the second index identies the component of force acting on the
surface.
8.2.1 Symmetry of the stress tensor

Let the sides of the small rectangular box have lengths x1, x2, and x3 parallel
to the reference axes (Fig. 8.5). For the box to be in static equilibrium, the sum
of the forces on the box (which would displace it) must be zero, and the sum of
the moments of the forces (which would rotate it) must also be zero. Consider
rst the balance of the moments acting on pairs of faces. The couple exerted
about a line through the center of the box parallel to the x3-axis by the shear
stresses on the faces normal to x1 (Fig. 8.5(a)) is (to rst order, neglecting the
second-order term in x12)
8.2 Stress
(a)
13 +
13
x1
23+
x2
11
21
13
x3
11 +
x3
(b)
x3
x1
x1
12
x1
231
x2
12 +
11
x1
12
x1
x1
x1
33 +
(c)
31 +
31
x3
x1
x3
x2
33
31
x3
23
21 +
x2
22 +
22
x1
x3
23
21
x2
22
x2
x2
x2
x2
x3
32+
32
x3
x3
x2
32
33
Fig. 8.5. Forces acting on the surfaces of a small rectangular box in the directions
of (a) the x1-axis, (b) the x2-axis, and (c) the x3-axis.

12
x1
x1
12 A1
12 x1 A1 12 x1 x2 x3
12
x1 A1
x1
2
2
12 V
(8:7)
A further couple is exerted about the x3-axis by the shear stresses on the faces
normal to x2 (Fig. 8.5(b)). This acts in the opposite sense to the rst couple and
(also to rst order) is equal to

21
x2
x2
x2 A2
(8:8)
21 A2
21 x2 A2 21 V
21
x2
2
2
The resulting couple about the x3-axis is the difference between (8.7) and (8.8).
For the box to be in equilibrium, the sum of the moments about the x3-axis must
be zero; therefore
12 21 V 0
(8:9)
This must be valid for any small volume V; therefore,

12 21
(8:10)
Similar evaluations of the moments about the x1- and x2-axes show, respectively, that 23 = 32 and 31 = 13. The equilibrium of moments acting on the
232
elementary volume requires the stress tensor to be symmetric (kn = nk), which
reduces the number of different elements in the matrix to six.
8.2.2 Equation of motion

Let the small box experience a displacement u = unen, where en is a unit vector in
the direction of displacement. The acceleration of the box as a result of all forces
acting on it is a = anen, where
an
2 un
t2
(8:11)
If the density of the material in the small box is and the volume of the box is V,
its mass m is equal to V. Let the body force per unit mass have components F1,
F2, and F3. The resultant force along the x1-axis is due to the normal stresses
acting on the surfaces with area A1 (Fig. 8.5(a)) and the shear stresses on the
surfaces with areas A2 (Fig. 8.5(b)) and A3 (Fig. 8.5(c)), respectively. The
resultant of the surface forces in the x1-direction is

11
21
A
x
x
11
1
11
1
21
2
21 A2
x1
x2

31
31
x3 31 A3
x3
11
21
31
x1 x2 x3
x2 x3 x1
x3 x1 x2
x
x
x3
1
2
11 21 31
(8:12)
V
x1
x2
x3
The equation of motion in the x1-direction as a result of the inertial, body, and
surface forces is

11 21 31
ma1 mF1
V
(8:13)
x1
x2
x3

11 21 31
(8:14)
a1 F1
x1
x2
x3
Similar expressions are obtained for the net forces along the x2- and x3-axes.
Using the summation convention (where the repeated index implies the sum for
k = 1, 2, and 3), we get the tensor equation
an Fn
kn
xk
(8:15)
8.3 Strain
233
If the body force per unit mass Fn can be neglected, we can write the acceleration as in (8.11), and this equation reduces to the homogeneous equation of
motion:
2 un kn
t2
xk
(8:16)
8.3 Strain
Let the vector x dene a point P in an arbitrary body and let Q be another point
of the body at an innitesimal distance y from P, as in Fig. 8.6. In a general
displacement of the body the point P is displaced to a new position P1 by the
vector u, and Q is displaced to Q1 by the vector v. If the difference between the
displacements is du, then
v u du u
u
u
u
y1
y2
y3
x1
x2
x3
(8:17)
Here y1, y2, and y3 are the components of y in the directions of the coordinates
x1, x2, and x3, respectively. In tensor notation
vk uk duk uk
uk
yn
xn
(8:18)
The relationship is not changed if we subtract the term 12 un =xk , and then add
it back again, giving
Q1
du
P1
u
v = u + du
u
y
P
x
O
Fig. 8.6. Illustration of a general displacement of points in a medium. The point P

is displaced to a new position P1 by the vector u and Q is displaced to Q1 by the
vector v.
234
v k uk

1 uk un
1 uk un
yn
yn
2 xn xk
2 xn xk
vk uk kn yn kn yn
(8:19)
(8:20)
The rst term on the right-hand side of this equation represents a rigid-body
translation of the entire body by the vector u. This takes place without internal
deformation of the body.
The second term on the right contains the tensor kn, whose elements are

1 uk un

kn
(8:21)
2 xn xk
Comparison with (1.27) and Box 1.1 shows that kn are the components of a
rotation about u = 0, i.e., the point P. The elements kk = 0 and kn = nk; the
tensor is antisymmetric and its diagonal elements are all zero:
2
3
0
12 13
kn 4 12
0
23 5
(8:22)
13 23 0
The product of this tensor with the relative position vector yn gives, in matrix
form,
3
2
32 3 2
12 y2 13 y3
0
12 13
y1
kn yn 4 12 0
(8:23)
23 54 y2 5 4 12 y1 23 y3 5
13 23 0
y3
13 y1 23 y2
The column matrix on the right-hand side of this equation has the same
components as the vector

e1
e2
e3

(8:24)
23 13 12 j y
y1
y2
y3
Here e1, e2, and e3 are unit vectors for the x1-, x2-, and x3-axes, respectively. The
vector represents a rotation, while y denotes the position of an arbitrary point
Q of the body relative to the point P, so y describes an innitesimal rigidbody rotation of the body about an axis through P. The direction of the rotation
axis is the vector with components (23, 13, 12). Following (8.21), this
can also be written

u3 u2
u1 u3
u2 u1
e1
e2
e3

(8:25)
j
x2 x3
x3 x1
x1 x2
8.3 Strain

e1

j =x1

u1
e2
=x2
u2
235

e3

=x3 r u

u3
(8:26)
The rigid-body rotation is a displacement of the entire body without deformation. Neither the translation u nor the rotation of the rigid body takes part in
the propagation of seismic waves.
The quantity kn in (8.20) is the strain tensor. It describes a deformation in
which different parts of the body are displaced relative to each other. As long
as these displacements are small, the deformation is elastic and the strains can
be described by a (3 3) strain matrix, whose general term is dened by
(8.19):

1 uk un
kn
(8:27)
2 xn xk
It is evident from this denition that interchanging the indices does not change
the general term; i.e., the strain matrix is symmetric (kn = nk). The diagonal
terms of the strain matrix (i.e., kk) describe normal strains, which correspond to
changes in elongation of the body; the non-diagonal terms describe shear
strains, which arise from angular distortion of the body.
8.3.1 Normal strain

Consider two points of a body that lie close to each other at the positions x1 and
(x1 + x1), respectively (Fig. 8.7(a)). If the body is stretched in the direction of
the x1-axis (Fig. 8.7(b)), the points are displaced by the small amounts u1 and
(u1 + u1), respectively. Using a MacLaurin or Taylor series, we can write
u1 u1 u1
u1
1 2 u1
x1
x1 2
x1
2 x21
x1 + x1
x1
(a)
u1
u1 + u1
(b)
x1 + u1
(x1 + x1)
+ (u1 + u1)
Fig. 8.7. Denition of normal strain for extension in the x1-direction.
(8:28)
236
If the displacements are innitesimally small, we can truncate the power series
at rst order, getting
u1
u1
x1
x1
(8:29)
The original separation of the two points was x1; after extension their separation
is (x1 + u1). The normal strain parallel to the x1-axis is the fractional change in
length resulting from an innitesimal displacement parallel to the x1-axis and is
denoted 11; thus,
11 lim
x1 !0
x1 u1 x1 u1
x1
x1
(8:30)
In a similar way, normal strains are dened for the x2- and x3-directions. If a
point at xk is displaced by an innitesimal amount to xk + uk, then there arise
normal strains kk, corresponding to
kk
uk
xk
(8:31)
The normal strains are not independent of each other in an elastic body.
Consider the change in shape of the bar in Fig. 8.8. When it is stretched parallel
to the x1-axis, it becomes thinner parallel to the x2-axis and parallel to the x3axis. The transverse strains 22 and 33 are of opposite sign to the extension 11,
but are proportional to it; so they can be expressed as
22 33

11 11
(8:32)
The constant of proportionality is Poissons ratio. The value of is constrained

to lie between 0 (no lateral contraction) and a maximum value of 0.5 for an
incompressible uid. In the Earths interior, has a value around 0.240.27. A
body that has = 0.25 is called an ideal Poisson body.
The normal strains result in a change of volume. The volume of the rectangular box in Fig. 8.5 is V = x1 x2 x3. As a result of innitesimal displacements
x1
(a)
x2
x1 + x1
(b)
F
x2 x2
Fig. 8.8. Illustration of the lateral contraction and the change in the angles between
the diagonals of a rectangular cross-section as a result of longitudinal extension.
8.3 Strain
237
u1, u2, and u3 the edges increase to x1 + u1, x2 + u2, and x3 + u3,
respectively. The fractional change in volume is
V x1 u1 x2 u2 x3 u3 x1 x2 x3
x1 x2 x3
V

x1 u1
x2 u2
x3 u3
(8:33)
1
x1
x2
x3
The limit of the fractional change in volume, for small V, is dened as the
dilatation, . As in (8.30) the limiting values of u1/x1, u2/x2, and u3/x3 are
the longitudinal strains 11, 22, and 33, respectively. Thus
V
1 11 1 22 1 33 1
V!0 V
lim
(8:34)
This expression for contains second- and third-order products of the strains
that can be neglected, thus
11 22 33
u1 u2 u3
x1 x2 x3
(8:35)
Taking u as the displacement vector, the dilatation is equivalent to

ru
(8:36)
Using tensor notation, and the summation convention implied by a repeated

index,
kk
uk
xk
(8:37)
8.3.2 Shear strain

The stress components (12, 23, 31) act obliquely on the surface of the
rectangular reference box (Fig. 8.4) and produce shear strains, which are
manifested as changes in the angular relationships between parts of a body.
These can also result from normal stresses. For example, the angles and
between the internal diagonals of a rectangular cross-section (Fig. 8.8), before
and after extension, respectively, are unequal; i.e., a longitudinal extension
gives rise to shear strain as well as normal strain.
Consider the two-dimensional distortion of a rectangle A0B0C0D0 by shear
stresses in the x1x2 plane (Fig. 8.9). Point A0 is displaced parallel to the x1-axis
by an amount u1 and parallel to the x2-axis by an amount u2. The shear strain
238
x 2-axis
C
(u 1 /x 2 ) x 2
u1
C0
D0
x2
2
A
u1
(u 2 /x 1 ) x 1
1
x1
u2
A0
u2
B0
x 1-axis
Fig. 8.9. Displacements accompanying two-dimensional shear strain in the x1x2

plane.
causes point D0, at a vertical distance x2 above A0, to be displaced by an

amount (u1/x2)x2 parallel to the x1-axis. This rotates side AD clockwise
through a small angle 2. For innitesimal displacements
2 tan 2
u1 =x2 x2 u1
x2
x2
(8:38)
Similarly, point B0, which is initially at a horizontal distance x from A0, is

displaced by the amount (u2/x1)x1 parallel to the x2-axis, causing AB to
rotate counterclockwise through a small angle 1 given by
1 tan 1
u2 =x1 x1 u2
x1
x1
The shear-strain component 12 is dened in (8.27):

1 u2 u1
1
1 2
12
2 x1 x2
2
(8:39)
(8:40)
Transposition of the indices 1 and 2 yields the shear-strain component 21,

which is identical to 12. The total distortion in the x1x2 plane is
1 2 12 21 212 221
(8:41)
239
The same argument leads to the denition of strain components 23 (=32) and
31 (=13) for angular distortions in the x2x3 and x3x1 planes, respectively. The
shear strains are therefore

1 u2 u1
12 21
2 x1 x2

1 u3 u2
(8:42)
23 32
2 x2 x3

1 u1 u3
31 13
2 x3 x1
They are expressed in tensor form by
kn nk

1 un uk
2 xk xn
(8:43)
The longitudinal and shear strains together form the symmetric strain matrix
0
1
11 12 13
(8:44)
kn @ 21 22 23 A
31 32 33
The elements of the matrix represent the strain tensor kn (k = 1, 2, 3; n = 1, 2, 3),
which, because of its symmetry, has six independent elements.

Hookes law describes perfectly elastic deformation, which occurs by means of
innitesimal strains. The components of strain are then linear functions of the
components of stress. The linear dependence allows the denition of elastic
moduli, each of which is a constant of proportionality between stress and strain.
Youngs modulus, the shear modulus, and the bulk modulus relate the different
elements of the stress and strain tensors for appropriate types of deformation.
Youngs modulus
Each normal stress kk is proportional to the corresponding normal strain kk.
Thus,
kk Ekk
(8:45)
The constant of proportionality, E, is Youngs modulus. The lateral contraction that

accompanies longitudinal extension is described by Poissons ratio, (see (8.32)).
240
Shear modulus (or rigidity modulus)

The shear strain kn (i.e., the total angular distortion) in a plane is proportional to
the shear stress in the plane, kn. Equation (8.41) denes the shear strain, so for
k n we have the relationship
kn 2kn
(8:46)
The constant of proportionality, , is the rigidity (or shear) modulus.

Bulk modulus (or incompressibility)
The bulk modulus, K, is a measure of the change of pressure needed to cause a
change of volume. A body under hydrostatic pressure p (dened as acting
inwards, equivalent to a negative normal stress) experiences a change of
volume. The fractional change in volume is the dilatation, , which is related
to the principal strains as in (8.34)(8.37). Under hydrostatic conditions there
are no shear stresses (kn = 0) and the normal stresses are equal (kk = p). The
dilatation is proportional to the pressure and the constant of proportionality is K.
Thus, we have the simple relationships
p K K
uk
K r u
xk
(8:47)
8.4.1 The Lam constants

A change of length in the x1-direction consists of the extension due to 11 and
contributions from the lateral contractions in the x2- and x3-directions that are
due to 22 and 33. The normal strain equals 11/E and, using (8.32), the lateral
contractions contribute 22/E and 33/E, respectively, to the longitudinal
strain. Thus, for the x1-direction
11
22
33

(8:48)
11
E
E
E
Similar equations are obtained for the x2- and x3-directions. On multiplying
each equation throughout by E, we get the set of equations
E11 11 22 33
E22 22 33 11
E33 33 11 22
(8:49)
Adding these equations gives

E11 22 33 11 22 33 1 2
(8:50)
E 11 22 33 1 2
(8:51)
241
This equation can be rewritten for 11:

11
E
22 33
1 2
(8:52)
We can obtain another expression for the sum (22 + 33) from the rst line of
(8.49):
22 33
E11 11
(8:53)
Substituting this expression for (22 + 33) into (8.52) gives

11
E
E11 11
1 2
(8:54)
11
E
E11 11
1 2
(8:55)
E
E
11
1 21
1
(8:56)
11
The coefcients of and 11 dene the Lam constants and , respectively:
E
1 21
(8:57)
E
1
(8:58)
The relationship between normal stress and normal strain in terms of the Lam
constants is
11 211
(8:59)
A similar result would be obtained by using any line in (8.49), so in general the
normal stresses and strains are related by
kk 2kk
(8:60)
The Lam constant is equivalent to the shear modulus. This can be shown by
establishing independently the relationship among Youngs modulus, the shear
modulus, and Poissons ratio (Box. 8.1), which leads to the same equation as
that in (8.58). The shear modulus is dened in (8.46) as the ratio of the shear
stress kn to the shear strain kn. Using the Kronecker-delta symbol, we can
therefore write the more general relationship
kn kn 2kn
(8:61)
242
Box 8.1. Relationship of the shear modulus, Youngs modulus,

and Poissons ratio
Consider a body with a square cross-section subject to normal stresses in
the x1x2 plane only (i.e., 33 = 0), as in Fig. B8.1.1(a). Let the area of
each side normal to the gure be A. Let p be the average of the normal
stresses 11 and 22, and let be the stress difference between p and each
normal stress. Therefore
11 p p 22
(1)
The outward stress difference along the x1-axis causes extension, whereas
the inward stress difference along the x2-axis causes contraction
(Fig. B8.1.1(b)). The change of shape of the cross-section results in angular
distortions internally. Thus the normal stresses give rise to both normal
strains and shear strains.
22
(a)
(b)
x2
11
x1
Fig. B8.1.1. (a) Normal stresses 11 and 22 in the x1x2 plane. (b) Deviatoric
stresses , equal to the difference between the normal stresses and their mean
value.
(A)
(a)
(b)
s
x2
s
(A)
s(1 + 11)
s(1 22)
x1
Fig. B8.1.2. (a) Undeformed square cross-section showing inward and

outward forces (A) due to deviatoric stresses. (b) Side lengths, normal
strains, and changes to the angles between intersecting diagonals as a result
of deviatoric stresses.
243
The outward force in the x1-direction is (A), which has a component

(A)/2 along the body diagonal (Fig. B8.1.2(a)). Likewise, the inward force
in the x2-direction has a component (A)/2 in the same direction. The
combined force parallel to the diagonal is 2(A). The area of a side normal
to the cross-section is A, so the area of a normal planar section that includes the
diagonal is 2A. The shear stress along the diagonal is therefore equal to .
The diagonals are initially at right angles to each other, but after
deformation their mutual orientation changes by an angle (Fig B8.1.2(b)),
which, as dened in Section 8.3.2, is the shear strain in the x1x2 plane.
Consider the angles and side lengths in the triangle BCD. If the original side
length of the square cross-section is s (Fig. B8.1.2(a)), the side along the
x-axis extends to s(1 + 11) while the side normal to this contracts to
s(1 + 22). The tangent of the angle BCD is DB/BC; thus,

s1 22 =2 1 22
tan
(2)
4 2
s1 11 =2 1 11
The trigonometric formula for the tangent of the difference of two angles
gives

tan =4 tan =2
1 tan =2
(3)
tan
4 2
1 tan =4 tan =2 1 tan =2
On equating the two expressions, we have
1 22 1 tan =2
1 11 1 tan =2
(4)
From (8.46), with 33 = 0 and replacing the normal stresses by the deforming
stress differences, we can write expressions for 11 and 22,
11
11
22

1
E
E
E
E
E
(5)
22
22
11
1
E
E
E
E
E
(6)
We now insert these expressions into (4). Note that the angle is very small,
so we can replace the tangent of the angle by the angle itself,
1 =E1 1 =2
1 =E1 1 =2
(7)
244

1
2 E
(8)
The shear modulus is the ratio of the shear stress to the shear strain; in this
case, the ratio of the deforming stress to the angular distortion :
(9)
From (8) we therefore have the following relationship among the shear
modulus , Youngs modulus E, and Poissons ratio :
E
21
(10)

In order to describe the propagation of a seismic wave in the Earth, some
simplifying assumptions must be made. First, the heterogeneity of the medium
is neglected. We assume that the medium is uniform and isotropic. This allows
us to use the homogeneous equation of motion derived in (8.16) to describe
particle displacements. Secondly, the medium is assumed to behave as a
perfectly elastic substance; only innitesimal displacements of the particles of
the medium are considered. The relationship between stress and strain is
governed by (8.61). The equation of motion becomes
2 un
kn 2kn
2
t
xk
(8:62)
Next we assume that the Lam parameters and do not vary with position, and
therefore can be treated as constants. This implies in effect that there are no
velocity gradients in the medium. On writing = nn and observing the
Kronecker delta, we have
2 un
nn
kn
2
2
t
xn
xk
(8:63)
Now we can insert the denitions of nn from (8.37) and kn from (8.43),

2 un
uk
un uk
2
(8:64)
xn xk
xk xk xn
t
2 un
uk
2 un
uk
2
2
t
xn xk
xk xn
xk
245
(8:65)
Note that the order of differentiation in the last term can be interchanged without
altering the meaning:
uk
2 uk
uk
xk xn xk xn xn xk
After gathering terms and simplifying, we have

2 un
uk
2 un
2
2
t
xk xn
xk
(8:66)
(8:67)
In symbolic form this equation is
2 u
rr u r2 u
t2
(8:68)
Now we recall the vector identity in (1.34) to obtain an expression for 2u:
r2 u rr u r r u
(8:69)
The homogeneous equation of motion becomes
2 u
rr u rr u r r u
t2
(8:70)
2 u
2rr u r r u
t2
(8:71)
This is the starting point for the treatment of elastic waves in an isotropic
homogeneous medium.
Minerals are individually anisotropic, their properties being controlled by
their crystal structure. However, in a large enough assemblage, random ordering
of the crystals makes a material macroscopically isotropic and justies the
assumption of this condition for the Earths interior. The assumption of homogeneity is unrealistic. For example, the density and elastic parameters that
control the passage of seismic disturbances change with depth and may also
vary laterally at a given depth. However, a heterogeneous medium can be
modeled acceptably by dividing it into smaller elements (e.g., parallel horizontal layers, or small blocks) and assuming homogeneous conditions in each
element. Real conditions can then be approximated by judicious choice of the
thickness, density, and elastic parameters of each element.
246
The assumption that seismic signals propagate by elastic displacements of the

medium is true only at some distance from the source. In an earthquake or
explosion the medium immediately surrounding the source is destroyed, particle displacements are large and permanent, and the deformation is anelastic.
However, the elastic conditions underlying (8.71) are applicable for the passage
of a seismic disturbance at a distance from its source.
In order to proceed further with the equations of motion for seismic body
waves we take separately the divergence and curl of both sides of (8.71). This
leads to the description of primary and secondary seismic waves.
8.5.1 Primary waves (P-waves)

First we take the divergence of both sides of (8.71):
2 r u
2r rru r r r u
t2
(8:72)
The vector identity (1.33) states that the divergence of the curl of any vector a is
zero, i.e., ( a) = 0. Thus the second term on the right is zero, and we get
2 r u
2r2 r u
t2
(8:73)
The dilatation , dened as the fractional change in volume, was shown in

(8.36) to equal the divergence of the displacement vector u; thus,
2
2r2
t2
(8:74)
2
2 r2
t2
(8:75)
where
2
(8:76)
On examining both sidesof(8.75) it is evident that has the dimensions of a velocity.

It is the velocity with which a change in volume (dilatation) propagates through
the medium. The disturbance propagates as a succession of compressions and
dilatations with velocity . The corresponding seismic wave is the primary wave,
or P-wave, so called because it is the rst arrival on the record of a seismic event.
The bulk modulus, Youngs modulus, and Poissons ratio can each be
expressed solely in terms of the Lam constants (Box 8.2). The relationship
between the bulk modulus and the Lam constants allows us to write (8.76) as
247
Box 8.2. Elastic parameters and the Lam constants

1. The bulk modulus, K
The bulk modulus describes volumetric shape changes of a material under
the effects of the normal stresses 11, 22, and 33. Hookes law for each
normal stress gives the equations
11 211
22 222
33 233
(1)
Adding these equations together gives

11 22 33 3 211 22 33
(2)
The dilatation is dened as

11 22 33
(3)
For hydrostatic conditions 11 = 22 = 33 = p. Substituting into (2) and rearranging gives

3p 3 2
(4)
The denition of the bulk modulus is K = p/. Therefore,

2
K
3
(5)
2. Youngs modulus, E
When a uniaxial normal stress is applied to a material, there results a
longitudinal extension or shortening that is proportional to the stress. The
constant of proportionality is Youngs modulus. Suppose that the applied
stress is along the x1-axis, so that yy = zz = 0. Hookes law applied to each
axis gives
11 211
0 222
0 233
Adding both sides of these equations gives
(6)
248
11 3 211 22 33 3 2
11
3 2
Inserting this into the rst line of (6) gives

11
11
211
3 2
After gathering and rearranging terms,

211
11 1
3 2

3 2
11
11
The denition of Youngs modulus is E = 11/11, so

3 2
E
(7)
(8)
(9)
(10)
(11)
(12)
3. Poissons ratio,
The denitions of the Lam constants in (8.57) and (8.58) give, respectively,
E
E
1 21 1 1 2
(13)
E
1
(14)
On combining these equations we obtain
2
1 2
(15)
In terms of the Lam constants, Poissons ratio is given by
(16)
t = t0
t = t0
249
t = t0 +
(a)
P-wave
(b)
S-wave
Fig. 8.10. Schematic illustration of (a) changes of volume and the angles between
intersecting diagonals during passage of a P-wave, and (b) the change of shape due
to shear during passage of an S-wave.

1
2
4
1
4

K
3
3
(8:77)
The velocity of the P-wave depends both on the bulk modulus (or incompressibility) and on the shear modulus. Thus a P-wave can propagate through a uid
phase in which the shear modulus is zero.
The propagation of a one-dimensional compression is illustrated in Fig. 8.10(a),
which shows an undeformed volume at time t0, the compressed volume at an
earlier time t0 t, and the dilated volume at a later time t0 + t. Changes of the
angles between the diagonals of the original square demonstrate that the deformation in the compressional wave also has a shearing aspect.
8.5.2 Secondary waves (S-waves)

Next, we proceed to take the curl of both sides of (8.71):
2 r u
2r rr u r r r u
t2
(8:78)
Again we use a vector identity to simplify the equation. The identity in (1.32)
states that the curl of the gradient of any scalar function f is zero, i.e., f = 0.
Thus the rst term on the right is zero. The remaining equation is
2 r u
r r r u
t2
Now we again use the vector identity (1.34), obtaining
(8:79)
250
2 r u
rr r u r2 r u
t2
(8:80)
The divergence of the curl of a vector is zero, therefore
2 r u
r2 r u
t2
(8:81)
2 r u
2 r2 r u
t2
(8:82)
where
2
(8:83)
The components of u are in the plane normal to the displacement u. The

disturbance propagates through the medium as a succession of shear displacements and travels with velocity . Because it depends on the shear modulus,
which is zero in liquids and gases, a shear wave can propagate only in solid
materials.
Comparison of (8.77) and (8.83) yields the seismic parameter , dened as
4
K
2 2
3
(8:84)
This parameter is important for determining the variation of density as well as

the adiabatic temperature gradient inside the Earth, which can be computed
because the P-wave and S-wave velocities are well known as functions of depth.
The S-wave velocity is less than the P-wave velocity . As a result the
seismic shear wave (or S-wave) is recorded at a seismic station later than the
P-wave, so it is also called the secondary wave. During the propagation of a onedimensional shear deformation (Fig. 8.10(b)), the shape of an originally square
cross-section at time t0 is distorted to a parallelogram at times t0 t and t0 + t.
The area of the parallelogram is, however, the same as that of the original
square. In three dimensions the shear wave propagates without change in
volume.
8.5.3 Displacement potentials

A theorem established by Helmholtz shows that a vector eld such as the
displacement vector u can be expressed in terms of both a scalar potential
and a vector potential , provided that the scalar eld is irrotational ( = 0)
and the vector eld is divergence-free ( = 0). Thus

u r r y
251
(8:85)
An irrotational displacement eld is one that has no shear components, whereas

a divergence-free displacement takes place without change of volume.
Consequently, in a seismic disturbance the potentials and correspond to
the displacements in P- and S-waves, respectively, and are obtained by solving
the corresponding wave equations.
P-waves
On taking the divergence of u and noting that ( ) = 0, we have
r u r2
(8:86)
Substituting into (8.73) with as the P-wave velocity gives

2 r2
2 r2 r2
t2
2
2
2
2
r 0
r
t2
(8:87)
(8:88)
This equation is always true if the expression in square brackets is zero. The
dening equation for the scalar potential of the P-wave displacement is
therefore
2
2 r2 0
t2
(8:89)
S-waves
Next, taking the curl of u, we have
r u r r r r y
(8:90)
Using the identities in (1.32) and (1.34), we get

r u rr y r2 y
(8:91)
On applying the condition that the vector potential be divergence-free ( = 0),

this becomes
r u r2 y
Substituting into (8.82) with as the S-wave velocity gives
(8:92)
252

2
2 r2 y
2 2
r
ry
t2
2
2
2
2
r y 0
r
t2
(8:93)
(8:94)
Here again the equation is true if the expression in square brackets is zero. This
leads to a dening equation for the vector potential of the S-wave
displacement:
2 y
2 r2 y 0
t2
(8:95)

The wavefront of a seismic wave is dened as a surface in which all particles
vibrate in phase with each other. Close to a point source in a homogeneous
medium, the wavefronts form spheres around the source, and the wave is called
a spherical wave. With increasing distance from the source the curvature of the
spherical wavefront decreases and eventually becomes at enough to be
regarded as a plane. The normal to the wavefront is the direction of propagation
of the wave, called the seismic ray path. Far from its source a seismic wave is
called a plane wave and it may be described using orthogonal Cartesian
coordinates.
8.6.1 One-dimensional solution for plane P-waves

For a plane P-wave propagating in the x1-direction the x2- and x3-axes are
perpendicular to each other in the plane of the wavefront. There is no change
in the x2- and x3-directions, so derivatives with respect to these coordinates are
zero. Equation (8.89) can then be written
1 2 2
2 t2 x21
(8:96)
In this equation is a function of both time and position. Invoking the method
of separation of variables, we can write
x1 ; t Xx1 Tt
(8:97)
Upon inserting this into the equation and dividing both sides by we get
1 2 T 1 2 X
k2
t2
X x21
2 T
253
(8:98)
Each side is a function of only one variable, so each side must equal the same
constant, which we write as k2. The negative sign is chosen so as to deliver
periodic solutions. We get the equations
1 2 T
k2
2 T t2
1 2 X
k2
X x21
(8:99)
Rearranging the equations gives

2 T
k2 2 T 0
t2
2 X
k2 X 0
x21
(8:100)
These are simple harmonic motions. If we dene = k, the separate solutions

for the dependence on time and position are
T T1 expit T2 expit
X X1 expik x1 X2 expik x1
(8:101)
k is called the wave-number and the angular frequency of the P-wave. The
general solution for a P-wave traveling along the x1-axis is obtained by combining the partial solutions:
x1 ; t A expit k x1 B expit k x1
C expit k x1 D expit k x1
(8:102)
The solution contains four arbitrary constants (A = T1X1, B = T2X2, C = T1X2,

D = T2X1), whose values in a given situation are determined by the boundary
conditions. If we consider only the real parts of the solutions (with new
constants A1 = A + B, A2 = C + D), we obtain
x1 ; t A1 cost k x1 A2 cost k x1
(8:103)
The two parts of the solution have phases (t + kx1) and (t kx1), respectively. The velocity with which a constant phase travels is called the phase
velocity. The propagation of a constant phase of the rst solution is governed by
254
the condition that (t + kx1) is constant. On differentiating with respect to time,

with and k held constant (and therefore also , because = /k), we get
dx1

dt
k
(8:104)
The negative sign indicates that this phase is a P-wave propagating with velocity
in the negative x1-direction. The second part of the solution can be treated
in the same way. It is seen to describe a P-wave propagating with velocity in
the positive x1-direction. The velocity is known as the phase velocity of the
wave.
8.6.2 One-dimensional solution for plane S-waves

Using (8.95), the equation for the vector potential of an S-wave traveling in the
direction of the x1-axis can be written for each component n as
1 2 n 2 n
x21
2 t2
(8:105)
This wave equation is solved as for P-waves, yielding solutions akin to (8.103).
For S-waves propagating with velocity , the wave-number is k and the
components of the vector potential are

n x1 ; t Bn1 cos t k x1 Bn2 cos t k x1
(8:106)
The solutions describe shear waves that travel in the negative and positive x1directions with wave-number k and phase velocity = /k.
8.7 Three-dimensional propagation of plane

P- and S-waves
The assumption that the plane wave is traveling along the x1-axis is too restrictive.
It is common usage in seismology (and other geophysical disciplines) to dene
Cartesian coordinates so that the vertical direction is the x3-axis and the horizontal
surface is the plane dened by the x1- and x2-axes. Box 8.3 shows how the onedimensional solutions can be extended to three dimensions. This is applicable to
both P-waves and S-waves. The solutions of the wave equation depend on the
velocity of the wave, which determines the wave-number. For P-waves we have
|k| = /, and for S-waves |k| = /.
8.7 Propagation of plane P- and S-waves
255
Box 8.3. Three-dimensional solution of the wave equation

Let e1, e2, and e3 be unit vectors corresponding to a set of Cartesian
coordinates x1, x2, and x3. The P-wave equation then becomes
1 2 2 2 2
2 t2 x21 x22 x23
(1)
and the solution by the method of separation of variables involves three

spatial components,
x1 ; x2 ; x3 ; t X1 x1 X2 x2 X3 x3 Tt
(2)
Inserting the solution and dividing throughout by , as for one-dimensional

propagation, gives
1 2 T
1 2 X 1
1 2 X 2
1 2 X3
k2
2
2
2
t
X1 x1
X2 x2
X3 x23
2 T
(3)
The constant k2 is equal to both the time-dependent and the spatially

dependent parts of the solution. Continuing as for the one-dimensional case,
by successively separating parts that depend on different coordinates on
opposite sides of the equality sign, we get for the time-dependent variation
1 2 T
k2
2 T t2
(4)
This is a simple harmonic motion with angular frequency = k. The

solution is
T T0 expit
(5)

1 2 X 1
1 2 X2
1 2 X3
2
k21
k
X1 x21
X2 x22
X3 x23
(6)

1 2 X2
1 2 X 3
k2 k21
k22
2
X2 x2
X3 x23
(7)

1 2 X3
k2 k21 k22 k23
2
X3 x3
(8)
The spatial variations are
256
Positive and negative values of k1, k2, k3, and satisfy these equations. We
choose a particular solution that corresponds to a wave traveling in the
direction of the positive reference axes:
x1 ; x2 ; x3 ; t 0 expik1 x1 expik2 x2 expik3 x3 expit
0 expit k1 x1 k2 x2 k3 x3
(9)
Note that k1x1 + k2x2 + k3x3= k x, where x is a position vector dened as

x x1 e1 x2 e2 x3 e3
(10)
and k is the wave-number vector, dened as

k k1 e1 k2 e2 k3 e3
(11)
whose magnitude is given by k2 = k12 + k22 + k32. The particular solution of the
wave equation is therefore
x; t 0 expit k x
(12)
8.7.1 P-wave propagation

The scalar potential of P-waves propagating in the direction of the wavenumber vector k is
x; t 0 expit k x
(8:107)
The P-wave displacement uP is the gradient of , and has components

uP r e1
e2
e3
x1
x2
x3
(8:108)
This can be written more succinctly using tensor notation:

uP en
expit kk xk i0 en kn expit kk xk

xn 0
(8:109)
uP i0 k expit k x
(8:110)
Now suppose that the P-wave is propagating in a vertical plane and dene the
x1-axis to coincide with the horizontal projection of the direction of propagation. The motions in the P-wave are conned to the x1x3 vertical plane, so there
8.7 Propagation of plane P- and S-waves
257
is no displacement in the horizontal x2-direction and differentiation with respect

to x2 gives zero. The P-wave-number is in this case
k k1 e1 k3 e3
(8:111)
and (8.110) becomes

uP k1 e1 k3 e3 i0 expit k1 x1 k3 x3
(8:112)
The direction of this displacement is the same as that of the ray path or wavenumber vector; i.e., the P-wave propagates as an alternation of compressions
and rarefactions along the direction of propagation.
8.7.2 S-wave propagation

The vector potential of S-waves propagating in the direction k has
components

(8:113)
n x; t 0n exp i t k x
where the S-wave-number is the vector
k k1 e1 k3 e3
(8:114)
The S-wave displacement uS is the curl of , and has components

3 2
1 3
2 1
e1
e2
e3

uS r
x2 x3
x3 x1
x1 x2
(8:115)
If we again consider propagation in the x1x3 vertical plane so that differentiation with respect to x2 gives zero, this equation reduces to

2
1 3
2

(8:116)
uS
e1
e2
e3
x3
x3 x1
x1
This can be rearranged as

2
2
1 3
e1
e3

uS
e2
x3
x1
x3 x1
(8:117)
The second bracketed term on the right describes displacements in the

direction of the x2-axis,

1 3
uSH
e2

(8:118)
x3 x1
258

uSH i 03 k1 01 k3 exp i t k x e2
(8:119)
The displacements are by denition in the horizontal plane and hence are always
normal to the direction of propagation. The horizontal component of a bodily
shear wave is known as the SH wave.
The rst bracketed term on the right of (8.117) describes a shear wave
conned to the vertical x1x3 plane and known as the SV wave. The 2
component of the vector potential in (8.113) is

(8:120)
2 02 exp i t k1 x1 k3 x3
The SV displacement is therefore

uSV 2 e1 2 e3
x
x3
1

k3 e1 k1 e3 i 02 exp i t k1 x1 k3 x3
(8:121)
The scalar product of the amplitude of the SV displacement vector uSV and the
wave-number k is

(8:122)
k3 e1 k1 e3 k1 e1 k3 e3 0
This conrms that the SV displacements, like the SH displacements, are normal
to the direction of propagation of the S-wave.
These results show that the displacements in the wavefront of a shear wave
can be resolved into two orthogonal motions: the SH-component is horizontal
and the SV-component is in the vertical plane containing the ray path.
further reading
Aki, K. and Richards, P. G. (2002). Quantitative Seismology, 2nd edn. Sausalito, CA:
University Science Books, 704 pp.
Bullen, K. E. (1963). An Introduction to the Theory of Seismology, 3rd edn. Cambridge:
Chapman, C. (2004). Fundamentals of Seismic Wave Propagation. Cambridge:
Lay, T. and Wallace, T. C. (1995). Modern Global Seismology. San Diego, CA:
Academic Press, 515 pp.
Shearer, P. M. (2009). Introduction to Seismology, 2nd edn. Cambridge: Cambridge
Udias, A. (2000). Principles of Seismology. Cambridge: Cambridge University Press,
490 pp.
Appendix A
Magnetic poles, the dipole eld,
and current loops
A1. The concept of magnetic poles and Gausss law

Coulomb carried out experiments with long magnetized needles and showed
that their ends exerted forces of attraction and repulsion on the ends of other
magnetized needles, similar to the forces between electrical charges. If freely
suspended, a magnet aligns in the Earths own magnetic eld so that one end is a
north-seeking pole (unfortunately shortened to north pole) and the other a southseeking pole. Magnetism originates in electric currents, but in some contexts the
concept of ctive magnetic poles can be useful. The force between the ends, or
poles, of two magnets is proportional to the product of the pole strengths and
inversely proportional to the square of the distance r between them. Between
two poles of strength p1 and p2 the force F is
F
0 p1 p2
er
4 r2
(A1)
where 0 is the magnetic eld constant, or permeability of free space; it is

dened to be exactly 4 107 N A2. The resemblance to Coulombs law for
electrical forces allows us to develop expressions for the magnetic potential and
ux. The magnetic eld may be dened as the force that acts on a unit magnetic
pole. With p1 = p and p2 = 1, the magnetic eld B of a pole p at distance r is
B
0 p
er
4r2
(A2)
where er is the radial direction. The magnetic potential of a single pole at

distance r is therefore
Z1
W
B er dr
0 p
4r
(A3)
The ux m of the magnetic eld B through a surface S surrounding the pole p is

259
260
Appendix A
Z
m
B n dS
(A4)
where n is the normal to the surface. Upon inserting the magnetic eld B of the
pole from (A2) and dening as the angle between n and the radial direction er,
the magnetic ux through a surface surrounding the pole p is
Z
m
0 p
cos dS
4r2
(A5)
Now we make use of the relationship between the solid angle d subtended
at distance r from an inclined surface element dS (Box 1.3), and obtain
Z4
m
0
0 p
d 0 p
4
(A6)
The total pole strength p enclosed by the surface S is therefore given by

p
1
1
m
0
0
Z
B n dS
(A7)
Because every magnet has two poles of equal and opposite strength, the sum of
all the poles in a volume is zero. The total magnetic ux through any closed
surface is therefore also zero. On applying the divergence theorem, we have
Z
m
Z
B n dS
r B dV 0
(A8)
For this to be true for an arbitrary volume

rB 0
(A9)
This result implies that magnetic monopoles cannot exist. It is known as Gausss
law after Carl Friedrich Gauss (17771855), who formalized it. The basic
magnetic eld is that of a dipole.
A2. The magnetic dipole
Two magnetic poles of equal strength but opposite sign, +p and p, are a
distance d apart (Fig. A1). The geometry has rotational symmetry about the
line AB joining the poles, the magnetic axis. The radius of length r from the
point M midway between the poles to the point P, where the magnetic potential
is to be determined, makes an angle with the magnetic axis. Let the distance of
P from the positive pole be r(+) and the distance from the negative pole be r().
Following (A3), the potential of the positive pole at P is
Appendix A
+p
r (+)
d/2
d/2
r( )
Br
P
I
261
Fig. A1. The geometry for calculation of the magnetic potential and the radial and
azimuthal elds of a pair of opposite and equal magnetic poles. In the limit, as the
separation of the poles tends to zero, the potential and elds are those of a magnetic
dipole.
0 p
4 r
(A10)
On applying the reciprocal-distance denition of the Legendre polynomials

(Section 1.12, Fig. 1.11) to the triangle AMP, this potential expands to
!
1 n
X
0 p
d
1
Pn cos
4 r
2r
n1
(A11)
Similarly, for the negative pole, the relations of the sides in the triangle BMP
give
!
1 n
X
0 p
0 p
d
1

Pn cos
2r
4 r
4 r
n1
(A12)
The combined potential of both magnetic poles at the point P is

(
)
1 n
0 p X
d
W
Pn cos Pn cos
4 r n1 2r
(A13)
From Rodrigues formula (Section 1.14) we nd that

Pn x
n
1
dn
1n n x2 1 1n Pn x
dx
2n n!
(A14)
The potential of the magnetic pole-pair is thus

1 n
0 p X
d
W
Pn cos 1n Pn cos
4 r n1 2r
(A15)
Each successive term is smaller than the previous term by the ratio d/(2r). The
rst terms are
262
Appendix A
2
0 pd
0 pd d
P
cos
P3 cos
1
4 r2
4 r2 2r
(A16)
A dipole is the constellation when the two poles are innitesimally close to each
other, so that d r. For innitesimal d/r we can ignore terms of higher than rst
order, so the magnetic potential of the dipole is given by the rst term in the
equation, which we can write
W
0 m cos
4 r2
(A17)
The quantity m = pd is called the magnetic moment of the dipole, for the
following reason. A dipole of length d, whose axis makes an angle with a
uniform magnetic eld B, experiences a force +pB on one pole and an opposite
force pB on the other pole. The perpendicular distance between the lines of
action of these forces is d sin , so the eld exerts a torque of magnitude
pdB sin in the direction normal to both the eld and the dipole.
pdB sin mB sin
(A18)
= mB
(A19)
The magnetic moment m of the dipole is a vector oriented along the dipole axis
from the negative to the positive pole.
A3. The Lorentz force
When an electrical charge q moves with velocity v through a magnetic eld B,
there arises a force F that is normal both to the eld and to the direction of
motion (Fig. A2(a)). This is the Lorentz force, which serves to dene the unit of
magnetic eld,
(a)
B
(b)
B
I
dl
v
F = q (v B)
dF = I (d l B)
Fig. A2. (a) The Lorentz force F on a charged particle moving with velocity v in a
magnetic eld B acts normal to both the velocity and the eld, resulting in a curved
trajectory (dashed line). (b) The BiotSavart law gives the increment of force d F
experienced by a short conductor of length d l carrying a current I in a magnetic eld
B. After Lowrie (2007).
Appendix A
263
F qv B
(A20)
With force measured in newtons (N), charge in coulombs (C), velocity in meters
per second (m s1), and electric current in amperes (A = C s1), the unit of
magnetic eld is the tesla, which has the dimensions N A1 m1.
Imagine the moving charge to be conned to move along a conductor of
length dl and cross-section A (Fig. A2(b)). Let the number of charges per unit
volume be N. The total charge inside the element of length dl is then NAq dl and
the Lorentz force acting on the element d l is
d F NAq dlv B
(A21)
The current v and the element dl of the conductor have the same direction, so we
can write
d F NAqvd l B
(A22)
The electric current I along the conductor is the total charge that crosses a
surface A per second; it is equal to NAqv. The force experienced by the element
d l of a conductor carrying a current I in a magnetic eld B is therefore
dF Id l B
(A23)
A4. Torque on a current loop in a magnetic eld

Using (A23), we can compute the force acting on each side of a small rectangular loop PQRS, which carries an electric current I in a magnetic eld B
(Fig. A3(a)). Let the lengths of the sides of the loop be a and b, respectively,
and let the x-axis be parallel to the sides of length a. The area A of the loop is equal
to ab; n is the direction normal to the plane of the loop. The magnetic eld B acts
(a)
B
S
F = IaB
a
P
(b)
Fx
F = IaB
Fx
F = IaB
x
Q
b sin
F = IaB
Fig. A3. (a) Forces on the sides a and b of a rectangular coil whose plane is
inclined at angle to a magnetic eld B. (b) Cross-section showing how the equal
and opposite, but not collinear, forces produce a torque on the coil. After Lowrie
(2007).
264
Appendix A
normal to the x-axis, making an angle with the direction n. A force Fx equal to
IbB cos acts on the side PQ in the direction of +x, and an equal and opposite
force Fx acts on the side RS in the direction of x; these forces are collinear and
cancel each other out. Forces equal to IaB act in opposite directions on the sides
QR and SP and the perpendicular distance between their lines of action is b sin
(Fig. A3(b)), so the magnitude of the torque experienced by the current loop is
IaBb sin IAB sin mB sin
(A24)
=mB
(A25)
The quantity m = IAn is a vector normal to the plane of the current loop.
Comparison with (A19) shows that it corresponds to the magnetic moment of
the current loop. At distances much greater than the dimensions of the loop, the
magnetic eld is that of a dipole at the center of the loop. Consequently,
magnetic behavior is more correctly explained by replacing ctive magnetic
dipoles by current loops. This is true even at atomic dimensions; circulating
(and spinning) electrical charges impart magnetic moments to atoms. The
denition of m in terms of a current-carrying loop shows that magnetic moment
has the dimensions of current times area, or ampere meter2 (A m2).
Appendix B
Maxwells equations of electromagnetism
In the early nineteenth century, experimental observations of electrical and

magnetic behavior led to the establishment of fundamental physical laws
governing electricity and magnetism. In 1873 the Scottish scientist James
Clerk Maxwell synthesized all known empirical laws of electricity and magnetism into a set of equations that describe electromagnetic phenomena. They
embody in succinct form the empirical laws of Coulomb, Ampre, Gauss, and
Faraday.
1. Coulombs law
Charles Augustin de Coulomb (17361806) discovered experimentally that the
force F between two electrical charges Q1 and Q2 is proportional to the product
of the charges and inversely proportional to the square of the distance r between
them. Let er be the unit vector from Q1 to Q2. In the international system (SI) of
units Coulombs law is
F
Q1 Q2
er
40 r2
(B1)
In this equation 0 is the electric eld constant, or the permittivity of free

space; it is equal to 8.854 187 817 1012 C2 N1 m2. If both charges are
positive or negative, the force between them is repulsive; if the charges have
opposite sign, the force is attractive.
The electric eld E is dened as the force that acts on a unit positive electrical
charge. If we let Q1 = Q and Q2 = 1, the electric eld of the charge Q at distance r is
E
Q
er
40 r2
(B2)
If the charge Q is positive, the eld acts outwards, in the direction of increasing
r. The electric potential at distance r is
265
266
Appendix B
Z1
U
E er dr
r
Q
40 r
(B3)
The ux of the electric eld E through a surface S surrounding the charge Q is

Z
Q
er ndS
40 r2
E n dS
S
(B4)
where n is the unit vector normal to the surface element dS. If is the angle
between n and the radial direction er, the scalar product of the unit vectors
equals cos , therefore
Z
Q
cos dS
40 r2
(B5)
We can use the denition of a solid angle (Box 1.3) to change the surface
integral to an integral over a solid angle around the charge Q:
Z
Q cos
dS
40 r2
Q 0 0
Z4
Q
Q
d
40
0
E n dS
(B6)
(B7)
If the electrical charge Q is distributed throughout a volume V with charge

density ,
Z
Q
dV
(B8)
We can apply Gausss divergence theorem to the right-hand side of (B7), which
becomes
Z
Z
dV 0
Z
E n dS 0
r E dV
(B9)
0 r EdV 0
(B10)
The volume V is arbitrary, so the integrand must always be zero. This gives
Coulombs law for the eld of free electrical charges with density distribution :
rE
(B11)
Appendix B
267
1.1. The effect of bound charges

In some materials, called dielectrics, electrical charges are not free, but are
bound to atoms in xed locations. An applied electric eld can cause the bound
charges to shift position (e.g., from one side of an atom to the other), with
positive and negative charges displaced in opposite directions. This results in an
electric polarization P. A charge QD accumulates on an arbitrary surface S
within a homogeneous dielectric material, equivalent to
Z
QD
P n dS
(B12)
The total charge QT carried by a polarizable material is the sum of the free
charge Q and the bound surface charge QD:
Z
QT Q QD
Z
Z
T dV 0 E n dS P n dS
(B13)
(B14)
Gausss theorem allows us to convert the surface integrals into volume integrals:
Z
Z
T dV 0
Z
r E dV
r P dV
(B15)
It follows that
r 0 E P T
(B16)
The electric displacement vector D is dened by

D 0 E P
(B17)
Coulombs law for a material that can be polarized electrically is therefore

r D T
(B18)
In a homogeneous dielectric material the electric polarization P is proportional to the electric eld E. In SI usage the proportionality constant is written as
the product of the permittivity 0 and the electric susceptibility . Thus
P 0 E
(B19)
D 0 E 0 E
(B20)
D 1 0 E 0 E
(B21)
The dimensionless quantity is the relative permittivity, or dielectric constant,

of the material. In a material that cannot be polarized = 1 and
268
Appendix B
D 0 E
(B22)
In this case, if the density of free charges is ,

rD
(B23)
2. Ampres law
Ampres law describes magnetic elds produced by electric currents.
Experiments begun in 1820 by Andr-Marie Ampre (17751836) and Hans
Christian rsted (17771851) showed that an electric current produces a
magnetic eld. Ampres experiments on a long, straight, electrical conductor
showed that the magnetic eld is in the plane normal to the conductor, and the
eld direction obeys a right-hand rule with respect to the current (i.e., the
directions of current and eld are indicated by the thumb and ngers, respectively). For example, the eld lines around a long straight conductor are
concentric circles (Fig. B1(a)). The strength of the magnetic eld outside the
conductor is proportional to the current I in the conductor and inversely proportional to the distance r from the conductor:
B/
I
r
(B24)
In general, if dl is an element of the closed path L around a conductor carrying a

current I in a magnetic eld B, Ampres law is
I
B d l 0 I
(B25)
The magnetic eld constant 0 ensures compatibility between the units of

electric current and magnetic eld. The integration can also be applied to a
(a)
(b)
B(r)
dl
r
L
J
Fig. B1. (a) The lines of magnetic eld B around a long straight conductor carrying
an electric current I are concentric circles. (b) For a path inside an electrical
conductor only the fraction of the current enclosed by the path causes the
magnetic eld B along the path.
Appendix B
269
path L inside an electrical conductor, at right angles to the ow of current

(Fig. B1(b)). In this case, not all the current is enclosed by the loop, and only the
fraction of the current passing through the loop causes the magnetic eld B. If J
is the electric current density (i.e., the current per unit cross-sectional area
normal to the ow), the amount of current enclosed by the loop is
Z
I
J n dS
(B26)
Equating this with (B25) gives

I
Z
B d l 0
J n dS
(B27)
We now use Stokes theorem to convert the left-hand side into a surface integral:
Z
Z
r B n dS 0
J n dS
(B28)
This must be true for any surface intersecting the current, thus
r B 0 J
(B29)
This is Ampres law for the magnetic eld produced by an electric current in a
conductor.
The current density J is proportional to the electric eld E. This follows from
Ohms law, which relates the current (I) and voltage (V) to the resistance (R) of a
circuit:
V IR
(B30)
The electric eld E is the voltage per unit distance along a circuit. In a straight
conductor of length L and cross-sectional area A the voltage V equals EL and
the current I equals JA. The resistance R of a conductor is proportional to its
length L and inversely proportional to its cross-sectional area A. The constant of
proportionality is the resistivity; its inverse is the conductivity, . Consequently
R = (1/)L/A and substitution into Ohms law gives

EL JA

L
A
(B31)
After simplifying, we get Ohms law in vector form:

J E
(B32)
By combining this with (B29), we get an alternative form of Ampres law:

r B 0 E
(B33)
270
Appendix B
This law applies to the magnetic effect produced by a current of free electrical
charges. However, bound electrical charges can also result in an electric current
and produce a magnetic eld.
2.1. The effect of displacement currents

In a dielectric material, the electrical charges are bound to atoms, but a timedependent change in their positions is equivalent to a displacement current ID.
The total electric current IT is the sum of the current I passing through the
material and the displacement current ID. Differentiating (B13) gives
QT Q QD
t
t
t
(B34)
Using (B26) and writing the volume density of the bound charges as D,
Z
Z
JT n dS
J n dS
S
Z
D dV
(B35)
Applying Gausss theorem to the rst two terms and using the result of (B18)
gives
Z
Z
r JT dV
r JdV
V
Z
r DdV
(B36)
The total current density, combining the free charges and bound charges, is
JT J
D
t
(B37)
Using the total current density in Ampres equation, we get

r B 0 J 0
D
t
(B38)
Finally, using Ohms law (B32) and the relation between the electric displacement vector and the electric eld (B22), Ampres law for a non-polarizable
medium is
r B 0 E 0 0
E
t
(B39)
3. Gausss law for magnetism

Early experimenters concluded that, unlike electrical charges, magnetic monopoles did not exist. Division of a magnet into smaller pieces always left a number
of magnets with two poles. All magnetic elds originate from electric currents,
Appendix B
271
dB
r
I
dl
er
Fig. B2. At distance r in a direction er from a short conductor of length d l carrying

a current I the magnetic eld d B is normal to both d l and er .
whether at macroscopic or at microscopic (atomic) level. Ampres investigations were extended by his contemporaries, Jean-Baptiste Biot (17741862)
and Flix Savart (17911841). Their empirical studies of the forces between
straight conductors carrying electric currents showed that the magnetic eld d B
at a distance r from a short conductor of length dl carrying a current I is given by
dB
0
Id l er
4r2
(B40)
The unit vector (direction) er is from the current element to the point of
observation (Fig. B2). The total eld of a current circuit at the point of
observation P is found by integrating (B40) around the circuit, which necessarily depends on the geometry of the circuit.
It follows that the magnetic eld is divergence-free. Taking the divergence of
(B40) gives
rdB

0 I
d l er
r
r2
4
(B41)
The length of the current element dl is constant with respect to the differentiation. The order of the differentiation can be changed, changing sign
accordingly, which gives
rdB

0 I
er
dl r 2
r
4
(B42)
The function of r to be differentiated is recognizable as

er
1
r
r
r2
(B43)
Substitution into (B42) leads to the curl of a gradient, which is always zero (see
(1.32)):
rdB

0 I
1
dl r r
0
4
r
(B44)
If this is true for every contribution d B to the eld, it must be true for the entire
eld. This yields Gausss law for magnetism:
272
Appendix B
rB 0
Let V be an arbitrary volume enclosed by a surface S in a magnetic eld B. The

net ux of the magnetic eld through the surface is obtained using Gausss
divergence theorem (Section 1.6):
Z
Z
B ndS
r BdV 0
(B45)
The net ux of the magnetic eld through the surface is always zero; the number
of eld lines entering the surface is the same as the number leaving the surface.
Hence magnetic eld lines always form complete loops; they do not begin or
end on charges as the electric eld does. This implies that magnetic monopoles do not exist. The elementary magnetic eld is that of a dipole.
3.1. The magnetic eld inside a magnetizable material

Just as bound charges affect the electric eld inside a dielectric, the magnetic
eld inside a magnetically polarizable material is modied by the internal
electric currents in the material. The atoms in crystalline materials occupy
xed positions in a regular lattice structure and their atomic magnetic moments
can be partially aligned by a magnetic eld. The net magnetic moment per
unit volume of the material is its magnetization, M. Consider a small volume
element with sides x, y, and z at the point (x, y, z) in a magnetizable
material (Fig. B3). A current I1 ows around the small loop with sides y and
z, causing a magnetization component Mx in the x-direction. The magnetic
moment of a current loop is the product of its area and the current in the loop
(Appendix A4):
mx Mx DV Mx Dx Dy Dz I1 Dy Dz
I1 Mx x
(B46)
(B47)
The magnetization is not necessarily uniform, so in the adjacent loop in the

y-direction it may equal (Mx + Mx) with a circulation current I2, where
I2 Mx DMx Dx

Mx
Mx
Dy Dx
y
(B48)
The net current at the interface between the loops is in the z-direction. Its
magnitude is the difference between I1 and I2:
I z I1 I 2
Mx
Dy Dx
y
(B49)
Appendix B
273
x
z I1
y
I2
I3
I4
Mx
Mx + Mx
y
Fig. B3. Production of magnetization components Mx and Mx + Mx in the

x-direction from currents I1 and I2 in adjacent small loops in the yz plane within
a magnetizable material.
If J is the current density in the material, the z-component of current must equal
Jz x y. The x-component of magnetization thus makes a contribution to the
current density in the z-direction equal to
Jz
Mx
y
(B50)
A similar argument can be applied to the current loops in the xz plane, which
carry currents I3 and I4, respectively, causing magnetization components My and
(My + My). Taking into account the sense of the currents around the small
loops, the net current in the z-direction from these loops is
Iz I4 I3
My
Dz Dx
x
(B51)
The corresponding contribution to the current density in the z-direction is

Jz
My
x
(B52)
The net z-component of the current density is found by combining (B50) and
(B52):
Jz
My Mx

r Mz
x
y
(B53)
By treating the current circulation in other pairs of the reference planes, the
other components of J can be obtained. The current density Jm associated with
the magnetization M is therefore
Jm r M
(B54)
274
Appendix B
Inside a magnetizable material we must modify Ampres law (B29) by adding

the extra current density associated with the magnetization. We then get
r B 0 J Jm 0 J r M
(B55)
On rearranging, we have

r
B
M
0

J
(B56)
Let an auxiliary vector H be dened as

B
M
0
(B57)
B 0 H M
(B58)
H has the same dimensions (A m ) as magnetization. Historically it has been

called the magnetizing eld, despite having the wrong dimensions. Inside an
isotropic, non-ferromagnetic material the magnetization M is proportional to H:
M H
(B59)
The constant of proportionality is the magnetic susceptibility, , which is a

dimensionless property of the material. The relationship between B and H is thus
B 0 H1 0 H
(B60)
The quantity = 1 + is the magnetic permeability of the material. In free space

and in materials that cannot acquire a magnetization the susceptibility is zero
and the permeability = 1, so
B 0 H
(B61)
4. Faradays law
In 1831 an English scientist, Michael Faraday (17911867), demonstrated that
a change in the magnetic ux m through a coil induced in the coil an electric
voltage V proportional to the rate of change of the ux. The direction of the
induced voltage was shown by Heinrich Lenz (18041865) to oppose the
change in ux through the coil. Thus
V
m
t
(B62)
The ux of the magnetic eld through a coil with surface area S is

Z
m
B n dS
S
(B63)
Appendix B
275
If E is the electric eld induced in the coil, and d l is an element of the wire in the
coil, the voltage induced in a path of length L (e.g., a circumference of the coil) is
Z
V
Edl
(B64)
With the aid of Stokes theorem the linear integral around the closed path L can
be converted into a surface integral over the area S enclosed by L:
Z
V
r E n dS
(B65)
Combining (B62), (B63), and (B65) gives

Z
r E n dS
S
Z
B n dS
(B66)
It follows that
rE
B
t
(B67)
This is Faradays law describing the generation of an electric eld from a

changing magnetic eld.
References
Cain, J. C., Wang, Z., Schmitz, D. R., and Meyer, J. (1989). The geomagnetic spectrum
for 1980 and corecrustal separation. Geophys. J., 97, 443447.
Creer, K. M., Georgi, D. T., and Lowrie, W. (1973). On the representation of the
Quaternary and Late Tertiary geomagnetic elds in terms of dipoles and quadrupoles. Geophys. J. R. Astron. Soc., 33, 323345.
Dyson, F. and Furner, H. (1923). The earths magnetic potential. Mon. Not. R. Astron.
Soc. Geophys. Suppl., 1, 7688.
Dziewonski, A. M. and Anderson, D. L. (1981). Preliminary Reference Earth Model
(PREM). Phys. Earth Planet. Inter., 25, 297356.
Finlay, C. C., Maus, S., Beggan, C. D. et al. (2010). International Geomagnetic
Reference Field: The Eleventh Generation. Geophys. J. Int., 183, 12161230.
Groten, E. (2004). Fundamental parameters and current (2004) best estimates of the
parameters of common relevance to astronomy, geodesy, and geodynamics.
J. Geodesy, 77, 724731.
Hasterok, D. P. (2010). Thermal State of Continental and Oceanic Lithosphere, Ph.D.
thesis, University of Utah, Salt Lake City, USA.
Lowes, F. J. (1966). Mean square values on sphere of spherical harmonic vector elds.
J. Geophys. Res., 71, 2179.
(1974). Spatial power spectrum of the main geomagnetic eld, and extrapolation to the
core. Geophys. J. R. Astron. Soc., 36, 717730.
(1994). The geomagnetic eccentric dipole: facts and fallacies. Geophys. J. Int., 118,
671679.
McCarthy, D. D. and Petit, G. (2004). IERS Conventions (2003), IERS Technical Note
No. 32. Frankfurt am Main: Verlag des Bundesamtes fr Kartographie und
Geodsie, 127 pp.
Schmidt, A. (1934). Der magnetische Mittelpunkt der Erde und seine Bedeutung.
Gerlands Beitrge zur Geophysik, 41, 346358.
276
References
277
Stacey, F. D. (1992). Physics of the Earth, 3rd edn. Brisbane: Brookeld Press, 513 pp.
(2007). Core properties, physical, in Encyclopedia of Geomagnetism and
Paleomagnetism, ed. D. Gubbins and E. Herrero-Bervera. Dordrecht: Springer,
pp. 9194.
Stacey, F. D. and Anderson, O. L. (2001). Electrical and thermal conductivities of FeNi
Si alloy under core conditions. Phys. Earth Planet. Inter., 124, 153162.
Stacey, F. D. and Davis, P. M. (2008). Physics of the Earth, 4th edn. Cambridge:
Vosteen, H.-D. and Schellschmidt, R. (2003). Inuence of temperature on thermal
conductivity, thermal capacity and thermal diffusivity for different types of rock.
Phys. Chem. Earth, 28, 499509.
Index
acceleration, 18
centrifugal, 88, 91, 117, 119, 140
Coriolis, 140, 141
Etvs, 141
gravitational, 23, 59, 66, 68
tide-raising, 117, 119, 122
adiabatic, 178
Ampres law, 200, 268270, 273
angular momentum, 142, 159
conservation of, 61
EarthMoon system, 132
lunar, 154
barycenter, 116, 119121, 131
binomial coefcient, 30
binomial series, 3031, 35
BiotSavart law, 271
bulk modulus, 180, 182, 240, 246, 247
Chandler wobble, 137, 157167
equations of motion, 161, 163
Loves number k, 167
period, 167
circulation, 20, 23
hydrothermal, 196
see also curl
Clairauts formula, 102
ClausiusClapeyron equation, 176
co-latitude, 1, 49, 83, 109, 199, 208
complex number, 2, 53, 150, 187
complex plane, 2
conservation of energy, 62, 171
continuity condition, 20
cooling model
half-space, 190195
oceanic lithosphere, 196
core thermal properties, 177
Coriolis acceleration, 141, 142
Coulombs law, 218, 265268
curl, 6, 7, 17
curl theorem, see Stokes theorem
deformation, 228, 246
elastic, 163, 227
tidal, 121, 124
dielectric constant, 267
diffusion equation, 185
diffusivity
magnetic, 219
thermal, 185, 188
dilatation, 237, 240, 246
dipole, 205, 206, 210, 260262
eccentric, 209213
eld, 199, 204
moment, 199, 206, 207, 208, 262
direction cosines, 9, 10, 15, 78
, 83, 164
Dirichlet conditions, 52
displacement
current, 200, 218, 270
electric vector, 267, 270
innitesimal, 138, 227, 244
P-wave, 256
S-wave, 257
tidal, 122, 124, 127, 129
divergence, 6, 17
theorem, 1820, 25, 223, 266
dynamic ellipticity, 96, 152
278
Index
EarthMoon system, 116, 119, 134

dimensions, 123
increase in separation, 135
synchronous rotation, 133
ecliptic plane, 61, 95, 137, 146, 152, 154
eigenvalue, 13
eigenvector, 13
ellipse, 64, 89
ellipsoid types, 75
ellipticity, see attening
enthalpy, 172, 174
entropy, 172, 178
Etvs gravity correction, 141
equations of motion
Chandler wobble, 162
Euler nutation, 156
homogeneous seismic, 233, 244
precession, 150
equipotential surface, 86, 92, 101, 124
error function, 194, 195
Faradays law, 274275
eld, 18
conservative, 18, 23
geomagnetic, see geomagnetic eld
magnetic, 262
magnetizing, 274
eld constant
electric, 265
magnetic, 199
attening, 31, 74, 76, 86, 90, 93
ux, 19
frozen-in, 225
gravitational, 25
heat, see heat ow
magnetic, 200, 223, 260
Fourier
integral, 54, 56, 57
series, 52, 55
transform, 5258, 191
frozen-ux theorem, 222225
Gauss
coefcients, 202, 203, 204, 207
law of magnetism, 200, 218, 260, 270274
theorem, see divergence theorem
geodetic parameters, 77
geoid, 106
height of undulation, see Stokes formula
geomagnetic eld, 203
dipole component, see dipole
279
elements, 202
models of origin, 217222
non-dipole, 204, 214
poles, 208
potential, 200, 201
power spectrum, 215
quadrupole component, 210
source depths, 216
geopotential, 86, 8894, 97, 98, 125
Gibbs energy, 173, 175
gravity
anomaly of geoid undulation, 107
anomaly of lunar tide, 125, 128
equatorial value, 106
normal, 101, 104
radial and polar components, 96100
Grneisen parameter, 180, 182
heat
conduction equation, 183185, 186, 190
ow, 183, 196
transport in the Earth, 170
Helmholtz energy, 173, 175
Hookes law, 227228, 239
inertia tensor, 159
internal energy, 172, 174
International Geomagnetic Reference Field,
204, 208, 212
J2 (dynamic form factor), 84, 9094
Keplers laws, 6066
Kronecker delta, 15, 241
Lam constants, 240, 241, 246, 247
Laplaces equation
geomagnetic eld, 200
gravitational potential, 66
spherical polar coordinates, 6974
latitude, 100, 102104
Legendre differential equation
associated, 46, 74
ordinary, 3437
Legendre polynomials, 3234, 37, 98, 110
associated, 4348, 51, 211
generating function, 34, 35, 39
normalization, 3941, 47
orthogonality, 3739, 46
reciprocal-distance formula, 34, 77, 111,
119, 261
280
Leibnizs rule, 32, 41

Levi-Civita permutation tensor, 14
librations, 131
line of equinoxes, 143, 146, 152
Lorentz force, 220, 262
Loves numbers, 124128, 130, 165, 167, 168
m (centrifugal acceleration ratio), 92, 93
MacCullaghs formula, 7481, 82, 85, 147, 166
MacLaurin series, 29, 40, 235
magnetic moment, 199, 262
magnetic pole, 208, 259
magnetic Reynolds number, 221
magnetization, 272
magnetohydrodynamic equation, 221
MAGSAT, 215
Maxwell
equations of electromagnetism, 218, 261, 265
thermodynamic relations, 173, 174
Milankovitch cycles, 137
moment of inertia, 62, 79, 80, 81, 152, 158
uniform sphere, 96, 132, 154
normal gravity formula, 106
nutation, 137
Euler (free), 155, 157
forced, 148, 153, 155
in longitude and obliquity, 152
obliquity of rotation axis, 137
Ohms law, 220
paleomagnetic equation, 199
permeability, 200, 274
phonons, 179
Poisson body, 236
Poissons equation, 2326, 67
Poissons ratio, 236, 239, 244, 248
potential, 18
centrifugal, 91, 163
Chandler wobble, 164
gravitational, 59, 6668, 76, 84
lunar tidal, 116, 119, 122
magnetic, 259, 262
tidal gravity anomaly, 126
vector, 219
power series, 3, 2831
power spectrum of geomagnetic eld, 214216
precession, 137, 142155
equations of motion, 150, 152
lunar orbit, 155
Index
solar-induced, 148, 153

Preliminary Reference Earth Model, 130,
181, 227
products of inertia, 80, 82, 158, 162, 163, 166
quadric surface, 11
quadrupole, 208
reciprocal-distance formula, 34, 35, 77, 111,
117, 119, 261
Rodrigues formula, 4143, 46, 261, 265
rotation
coordinate axes, 1516
uctuations of, 137
rigid-body, 119, 234
synchronous, 134
see also curl
rotation matrix, 812
rotational symmetry, 71, 73, 76, 116, 119, 260
scalar product, 4, 5, 11
Schmidt polynomials, 48, 201
secular variations, 213
seismic parameter, 250
seismic wave, 244
displacement potentials, 250252
propagation, 254258
P-waves, 246, 252254, 256
SH and SV waves, 258
S-waves, 250, 254, 257
wavefront, 252
semi-latus rectum, 64, 66
separation of variables, 69, 186, 190, 252, 255
shear modulus, 240, 244
Shidas number, 129, 130
solar heat penetration, 185189
solid angle, 23, 24, 260, 266
specic heat, 178, 180, 184
spherical harmonic functions, 4951, 108, 201
normalization, 50
zonal, sectorial, and tesseral, 51
spheroid, 74, 76, 86
polar equation, 88, 89
reference gure for Earth, 81, 86, 90, 96
steradian, 25
Stokes formula for geoid height, 108114
Stokes theorem, 2023, 225, 269, 275
strain, 228, 233235
normal, 235239
shear, 239
stress, 227, 228232
Index
stressstrain relationships, 239241

surface area element, 24, 50
susceptibility
electric, 267
magnetic, 274
Taylor series, 21, 30
temperature
adiabatic gradient, 170, 172, 179, 180, 250
melting-point gradient, 176, 179
tensor, 1415
inertia, 159
strain, 235, 239
stress, 230
thermal expansion coefcient, 178, 180
thermodynamic potentials, 172
tides
bodily, 125
deceleration of rotation, 131135
deection of vertical, 130
inequality, 119, 120, 121
lunar versus solar, 123
origin of, 116119
potential of, 116, 119121, 126
torque
frictional, 131
281
gravitational, 95, 137

magnetic, 198, 262, 264
solar, 142, 146, 147
trajectory types, 64
transformation
coordinate systems, 146
matrix, 12, 16
transpose of a matrix, 9
unit sphere, 25, 50
vector, 4
differential operator, 5, 16
identities, 5, 8
irrotational, 7, 23
product, 4, 5
solenoidal, 20
wave equation, 244254
solution, 253, 255
wave-number, 253, 254, 256,
257, 258
Youngs modulus, 239, 242, 248
zonal approximation, 109

Guide To Geophysical Equations

Uploaded by

Copyright:

Available Formats

Guide To Geophysical Equations

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Guide To Geophysical Equations

Uploaded by

Copyright:

Available Formats

A Students Guide to Geophysical Equations

cambridge university press

This book is dedicated to Marcia

Appendix A Magnetic poles, the dipole eld, and current loops

apply advanced software packages. A fundamental mathematical understanding

1.1 Cartesian and spherical coordinates

1.2 Complex numbers

Fig. 1.2. Representation of a complex number on an Argand diagram.

1.2 Complex numbers

It is often useful to write a complex number in the exponential form introduced

1.3 Vector relationships

1.3.1 Scalar and vector products

The vector products of the unit vectors are

1.3 Vector relationships

1.3.2 Vector differential operations

1.3 Vector relationships

Box 1.1. The curl of a vector

Combining the results gives

It is a worthwhile exercise to establish these identities from basic principles,

1.4 Matrices and tensors

We can write the scalar product (r m1) as

1.4 Matrices and tensors

The three equations can be written as a single matrix equation

The coefcients nm (n = 1, 2, 3; m = 1, 2, 3) are the cosines of the interaxial

1.4.2 Eigenvalues and eigenvectors

Box 1.2. Direction cosines

Therefore, the vector r in (1) is equivalent to

1.4 Matrices and tensors

On writing the unit vector u as in (4), and applying the orthogonality

the columns). The transpose of a (3 1) column matrix is a (1 3) row matrix.

The matrix equation X MX = K, where

K is a constant, denes a quadric

The symmetry of the matrix leads to the equation of this surface:

fx; y; z 11 x2 22 y2 33 z2 212 xy 223 yz 231 zx K (1:46)

nx; y; z 2x0 ex y0 ey z0 ez 2rx0 ; y0 ; z0 (1:49)

1.4 Matrices and tensors

ellipsoid. The component x0 is then proportional to x, y0 is proportional to y, and

The simultaneous equations have a non-trivial solution only if the determinant

This equation is a third-order polynomial in . Its three roots (1, 2, 3) are

which we can write in symbolic form

The matrix I, with diagonal elements equal to 1 and off-diagonal elements 0, is

1.4.3 Tensor notation

A convention introduced by Einstein drops the summation sign and tacitly

1.4 Matrices and tensors

In tensor notation this is written

This can be veried readily for each component of u. For example,

The tensor equivalent to the unit matrix dened in (1.55) is known as

in which a summation over the repeated index is implied.

1.4.4 Rotation of coordinate axes

A second set of coordinates xn is rotated relative to the axes xl so that the

Equating the expressions in (1.68) and (1.69) for vk gives

Using the relationships between the axes in (1.67),

On multiplying by ml and summing,

This relationship allows us to compute the elements of a matrix in a new

1.4.5 Vector differential operations in tensor notation

1.5 Conservative force, eld, and potential

The divergence of the vector a is written in tensor notation as

The curl (or rotation) of the vector a becomes

1.5 Conservative force, eld, and potential