ii

Carlos E. Solivrez



The the cover was drawn using the
Wolfram Demonstrations Project Ellipsoid by Jeff Bryant.


 ELECTROSTATICS AND MAGNETOSTATICS
OF POLARIZED ELLIPSOIDAL BODIES:
THE DEPOLARIZATION TENSOR METHOD

Carlos E. Solivrez

 ii Carlos E. Solivrez

This work is the English translation by the author of the first edition of his book in
Spanish, Propiedades electrostticas y magnetostticas de cuerpos elipsoidales:
formalismo del tensor depolarizacin. In the process of translation a few errors
were corrected and a new chapter, several problems and many appendices were
added concerning the actual calculation of the values of the depolarization tensor.

Free Scientific Information

Electronic editions by Carlos E. Solivrez

The author advocates free access to all scientific and technological information
obtained with full or partial government's support. This information should be
considered a commons belonging to the people who creates, supports and gives
meaning to it. From year 2012 onwards, these author publishes all his works in
Internet, including the present book, for its free use and distribution under a
Attribution-NonCommercial-ShareAlike 3.0 Unported (CC BY-NC-SA 3.0) license.


This work will be modified any time the author finds an error or considers
convenient to make further elaborations of its content to make it more useful or
easier to undersand. All valuable contributions of readers, specially errors, will be
duly acknowledged in the following version of the text.



1st. Spanish edition: Bariloche (Ro Negro, Argentina), June 28, 2012, 32.467
words, 147 pages.
1st. English edition: Bariloche (Ro Negro, Argentina), July 15, 2016, 39.030 words,
197 pages. ISBN 978-987-28304-0-3.

 Depolarization tensor method iii

Abstract

A study is made of the behaviour of ellipsoidal bodies, with no free currents and
charges, under uniform applied electrostatic and magnetostatic fields. The method
is valid for all sorts of solid homogenous isotropic or anisotropic materials:
dielectric, ferroelectric, diamagnetic, paramagnetic, ferromagnetic, conducting,
superconducting... Expressions are given, in all cases, for the body's internal and
external fields, as well as for the free electromagnetic energy and the torques
derived from it. Apart from the body's volume and the electromagnetic properties
of the material, these expressions depend only on the depolarization tensor n
determined by the two aspect ratios of the ellipsoid. Explicit expressions are given
for n both in field points internal and external to the body in terms of
elementary functions except for the triaxial ellipsoid where they are Legendres
elliptic functions. In the non-linear range the electric or magnetic polarization is an
implicit function of the applied field and the anisotropy of the material, while the
external field is an explicit function of the polarization. In the linear range both the
polarization and the internal field are explicit functions of the constant internal
value of n inside the body and the isotropic or anisotropic susceptibility tensor ,
and the external field is an explicit function of the polarization and n. In the
isotropic case the limit values = and = -1 (SI units) fully describe the
behaviour of conductor and superconductor ellipsoids. A discussion is made of
some common errors in the treatment of electromagnetic singularities and of the
behaviour of bodies of infinite extension and ellipsoidal cavities.

 iv Carlos E. Solivrez

 Depolarization tensor method v

Index
Abstract iii
Chapter 1: Fundamental concepts 1
Origin ................................................................................................................................................................ 1
History .............................................................................................................................................................. 1
Applications ................................................................................................................................................... 3
Requirements ................................................................................................................................................ 5
Limitations of the method ....................................................................................................................... 5
Dealing with singularities ........................................................................................................................ 5
Point charge type singularity ............................................................................................................ 6
Step discontinuities ............................................................................................................................... 9
Physical units and mathematical notation .................................................................................... 10
Organization of the book ....................................................................................................................... 11
Chapter 2: Ellipsoids in Electric and Magnetic Fields 15
Electric polarization ................................................................................................................................ 15
Basic equations .................................................................................................................................... 15
Permanent electric polarization ................................................................................................... 18
Induced electric polarization ......................................................................................................... 20
Induced electric polarization of two interacting atoms ................................................................. 21
Dielectrics .......................................................................................................................................................... 24
Magnetization ............................................................................................................................................ 26
Basic equations .................................................................................................................................... 26
Permanent magnetization ............................................................................................................... 28
Induced magnetization ..................................................................................................................... 29
Conductors .................................................................................................................................................. 30
Equivalent polarization .................................................................................................................... 33
Superconductors ....................................................................................................................................... 34
Magnetization model ..................................................................................................................................... 35
Surface conduction current model .......................................................................................................... 36
Summary of integro-differential equations .................................................................................. 38
Summary of induced polarization equations ............................................................................... 39
General formulation of the method .................................................................................................. 40
Solving the integro-differential equations by iteration ...................................................... 42
Chapter 3: The depolarization tensor: basic treatment 45
Definition ..................................................................................................................................................... 45
General properties of n .......................................................................................................................... 47
Symmetric tensor ................................................................................................................................ 47
Trace ......................................................................................................................................................... 48
Orthogonal transformations .......................................................................................................... 48
Symmetries ............................................................................................................................................ 49
n as a surface integral ............................................................................................................................. 50
Surface step discontinuity .................................................................................................................... 51
Surface charge density ...................................................................................................................... 51
Surface step discontinuity ............................................................................................................... 52
Calculation of n using electrostatic Gausss Law ......................................................................... 54
Sheet of constant thickness and infinite extension .............................................................. 54
Right circular cylinder of infinite length ................................................................................... 56
Sphere ...................................................................................................................................................... 58
Other properties of the internal depolarization tensor N ...................................................... 59
Integral expressions of the eigenvalues of N .......................................................................... 59
Diagonalization and inversion ....................................................................................................... 60
Infinite semiaxes and the inverse of N ....................................................................................... 60
vi Carlos E. Solivrez

N is invariant for similar ellipsoids ............................................................................................. 62

N(0) is determined by aspect ratios ........................................................................................... 63
Chapter 4: The depolarization tensor: advanced treatment 65
Calculations of n from gravitational potentials ........................................................................... 65
Triaxial ellipsoid .................................................................................................................................. 65
Internal depolarization tensor N ....................................................................................................... 67
Expressions with normal elliptic integrals E and F .............................................................. 67
N elliptic cylinder ................................................................................................................................ 68
N oblate spheroid ................................................................................................................................ 69
N prolate spheroid .............................................................................................................................. 74
Unified treatment of N for spheroids ......................................................................................... 77
Graphs of Na, Nb and Nc .................................................................................................................... 79
Na ............................................................................................................................................................................ 80
Nb ........................................................................................................................................................................... 80
Nc ............................................................................................................................................................................ 80
Combined graph .............................................................................................................................................. 84
External depolarization tensor ........................................................................................................... 85
Obtention of the external gravitational potential by Ivorys method .......................... 85
Near and far away point approximations to next .................................................................... 87
General expression of external n .................................................................................................. 87
Traces ................................................................................................................................................................... 90
Verificacion for the sphere ......................................................................................................................... 91
Elliptic cylinder ................................................................................................................................................ 92
Oblate spheroid ............................................................................................................................................... 95
Prolate spheroid .............................................................................................................................................. 96
Triaxial ellipsoid ............................................................................................................................................. 97
Chapter 5: Energy, forces and cavities 99
Thermodynamics of electrostatic and magnetostatic energy ............................................... 99
Basic concepts ...................................................................................................................................... 99
Thermodynamics of electromagnetism ................................................................................. 101
Anisotropy energy ................................................................................................................................. 103
Origin of torques exerted on a body .............................................................................................. 105
Torque exerted on an ellipsoid and the state of equilibrium ............................................. 106
Dielectrics ............................................................................................................................................ 106
Permanent polarization ............................................................................................................................ 106
Induced polarization .................................................................................................................................. 108
Anisotropic susceptibility ........................................................................................................................ 109
Magnetic materials .......................................................................................................................... 110
Permanent magnetization ....................................................................................................................... 110
Induced magnetization .............................................................................................................................. 111
Magnetic torque experiments ................................................................................................................ 112
Conductors .......................................................................................................................................... 112
Superconductors .............................................................................................................................. 113
Force on an ellipsoid in a non-uniform field ............................................................................. 114
Infinite and infinitesimal bodies ..................................................................................................... 114
Cavities ....................................................................................................................................................... 116
Standard treatment ......................................................................................................................... 116
Some specific cavities: or homoeoids ...................................................................................... 117
Thin shells and the dipole layer ................................................................................................. 118
Chapter 6: Selected problemas 121
Electric polarization ............................................................................................................................. 121
Problem 01: Autoconsistent electric polarization of two atoms ................................. 121
Problem 02: Shape or crystalline anisotropy? .................................................................... 121
Problem 03: Dielectric sphere .................................................................................................... 122

 Depolarization tensor method vii

Magnetic polarization .......................................................................................................................... 124

Problem 04: Deriving n from the magnetic vector potential ........................................ 124
Problem 05: Permanently magnetized infinite right circular cylinder .................... 124
Problem 06: Isotropic magnetic sphere ................................................................................. 126
Conductors ............................................................................................................................................... 126
Problem 07: Solve ellipsoidal conductors using equivalente polarization ............. 126
Problem 08: A conductor is a perfect dielectric ................................................................. 127
Problem 09: Polarization of a spherical conductor ........................................................... 127
Problem 10: External field of a spherical conductor ........................................................ 128
Problem 11: Directions for which P is parallel to E0 in triaxial ellipsoids ............... 129
Problem 12: Electric fields generated by sharp points .................................................... 129
Superconductors .................................................................................................................................... 130
Problem 13: Superconducting infinite cylinder and sheet ............................................ 130
Problem 14: Magnetized superconducting sphere ............................................................ 130
Problem 15: Superconductivity as perfect diamagnetism ............................................. 131
Problem 16: Magnetic moment of a superconducting ellipsoid .................................. 132
Problem 17: Description of Meissner Effect using magnetization currents ........... 133
Problem 18: Approximate solution for non ellipsoidal bodies .................................... 133
Depolarization tensor .......................................................................................................................... 133
Problem 19: Components of N in spherical symmetry .................................................... 133
Problem 20: Single eigenvalue may determine the type of spheroid ........................ 134
Problem 21: Eigenvalues and aspect ratios .......................................................................... 134
Problem 22: Sequence of ellipsoids over some aspect ratio curves .......................... 135
Problem 23: Eigenvalues of triaxial ellipsoids .................................................................... 135
Problem 24: Depolarization tensor of a very long right circular cylinder .............. 136
Problem 25: Depolarization tensor of a sphere .................................................................. 136
Problem 26: Alternative way of obtaining certain eigenvalues of N ......................... 136
Problem 27: Solving the prolate spheroid with Legendres elliptic integrals ....... 136
Problem 28: Spherical shell ......................................................................................................... 137
Problem 29. Unit vector normal to the ellipsoid's surface ............................................. 139
Problem 30: Surface step discontinuity of the depolarization tensor ...................... 139
Problem 31: Ellipsoidal conductor with net charge ......................................................... 140
Problem 32: Application of Ivorys method to the sphere ............................................. 140
Problem 33: Expression for far away fields obtained from next ................................... 141
Problem 34: Field on the external side of the surface of a polarized sphere ......... 142
Problem 35: Microscopic origin of depolarization tensors ........................................... 143
Energy, forces and torques ................................................................................................................ 143
Problem 36: Energy of a dielectric sphere ............................................................................ 143
Problem 37: Fermis contact interaction ............................................................................... 143
Problem 38: Torques on spheroids .......................................................................................... 144
Problem 39: Torque exerted over an anisotropic dielectric sphere .......................... 145
Problem 40: Torque on a magnetized disk ........................................................................... 145
Problem 41: Torque on a very long superconducting right circular cylinder ....... 146
Problem 42: Depolarization tensor of inifinitesimally thin shells .............................. 146
Problem 43: Applications of thin shells .................................................................................. 146
Appendix 1: Electromagnetic units 147
Appendix 2: Vectorial Operators 151
Appendix 3: Integral theorems of vector calculus 153
Laplacian of a point charges potential ........................................................................................ 153
Divergence, gradient and rotor theorems .................................................................................. 155
Extension of the divergence theorem ........................................................................................... 155
Point charge type singularities .................................................................................................. 156
 viii Carlos E. Solivrez

Step discontinuities ......................................................................................................................... 157

Generalized divergence theorem .......................................................................................................... 157
Generalized rotor theorem ...................................................................................................................... 158
Polarized bodies ........................................................................................................................................... 158
Appendix 4: Field of an electric dipole 161
General expression ............................................................................................................................... 161
Dipolar field's energy ........................................................................................................................... 162
Appendix 5: Simmetries of the susceptibility tensors of single crystals 165
Appendix 6: Dyadics 167
Appendix 7: Ellipsoids 169
Equations and main features ............................................................................................................ 169
Equations, sections, volume and surface ............................................................................... 169
Aspect ratios ....................................................................................................................................... 170
Normals to the surface and central distances ........................................................................... 171
Summary of properties .................................................................................................................. 173
Confocal ellipsoids ................................................................................................................................ 174
Physical-geometric interpretation of .................................................................................. 174
Corresponding points ..................................................................................................................... 175
Appendix 8: Useful integrals 179
ds
.................................................................................................................................. 179
( )
2
A+ s B + s

ds
................................................................................................................................. 180
( A + s ) (B + s )
3/2

Appendix 9: Legendres elliptic integrals 181
Definitions ................................................................................................................................................ 181
Notation used in the more important references .................................................................... 181
Special values and parametric graphs .......................................................................................... 182
Reduction to normal form of the elliptic integrals of interest ........................................... 184
Main references 187
Alphabetic index 189
About the author 193

 Depolarization tensor method 1

Chapter 1:
Fundamental concepts

Origin
The topics discussed in this book were developed by the author, with long
intervals, since the decade of 1970. The first ones were initially presented in the
theory of electromagnetism course for students of physics at the Balseiro Institute
(National Atomic Energy Comission National University of Cuyo, Argentina) in
combination with general technics for solving Maxwell's differential equations. In
the decade of 1990 a full revision was made in order to adapt the formulation to
the level of a first course on electricity and magnetism for engineers with basic
knowledge of vector analysis (Bariloche Campus, National University of Comahue,
Argentina). Only two small portions of this work1,2 were published in international
scientific magazines with peer review, because editors considered the full text
unsuitable for publication because of its length and pedagogical orientation. A
thorough revision was made of all the material thus gathered, revision that will
continue as long as readers are willing to make corrections and suggest
modifications,and the author is able to analize and incorporate them.
The depolarization tensor method provides an eficient and compact way of
solving the case of ellipsoidal bodies polarized by constant electric and magnetic
fields. It is thus well suited for technical applications, but also for research on
anisotropic properties of materials, for which no general method is, to the
knowledge of the author, presently available. From the point of view of the
foundations of electromagnetism, it is one of the simplest and clearest example of
the inextricable relationship between geometric and physical properties of
materials. This relationship is expressed by the boundary conditions that
determine the solution of the diferential equations, but this is never so clearly and
explicitly expressed as by the depolarization tensor. It is also a good illustration of
the power of integral methods for solving electromagnetic problems. The main
contribution of the author in this field, as shown by the current literature on the
subject, is the generalization of the use of this tensor for all static electromagnetic
problems, the treatment of anisotropic materials and the calculation of fields
outside the body.

History
The use of what is today identified as depolarization tensor is very old in the
analysis of permanently magnetized materials. The first to derive it explicity from
a potential was probably Maxwell3. To that end he used Poisson's proof (see eq.


1 Solivrez (1981), pp. 1363-1364.
2 Solivrez (2008), pp. 203-207.
3 Maxwell, pp. 66-69.
 2 Carlos E. Solivrez

2.2) that the calculation of the field created by a uniformly magnetized body is
mathematically equivalent to taking the directional derivative of the gravitational
attraction of an homogenous mass distribution with the same shape. The
mathematical analysis of the subject thus benefits from the numerous studies of
gravitational potential (proportional to the integral f defined by eq. 3.8) made
since Newton's times. It is in this field of physics that ellipsoidal bodies were and
still are intensively studied4.
Maxwell proved that ellipsoidal bodies are uniformly polarized when immersed
in a uniform static applied field (a constant vector in the region occupied by the
body). He points out for integral f (there called V5)
... the only cases with which we are acquainted in which V is a
quadratic function of the coordinates within the body are those in which
the body is bounded by a complete surface of the second degree, and the
only cases in which such a body is of finite dimensions is when it is an
ellipsoid.
The origin of this behaviour which is analized in Chapter 2 is that internal
fields, and the polarizations they produce, are proportional to second derivatives
of f. Soon afterwards Thomson and Tait proved ellipsoids to be the only finite
bodies with this property6.
The principal values of Maxwell's entity are frequently called demagnetizing or
demagnetization factors or coefficients, and the study of its history began as early
as 18967. The author of this book has not been able to determine who, when and
where rebaptized it as depolarization tensor, a name that rightly includes both
electric and magnetic phenomena8. In 1941, Stratton's classical text on
electromagnetism introduced the principal values of the depolarization tensor
inside dielectrics by solving Laplace's equation in ellipsoidal coordinates. He called
them depolarization factors9 but he neither mentions their tensorial character
(which Maxwell apparently did not detect), nor applies them to conducting or
magnetic materials. Two distinguished russian theoretical physicists, L. Landau
and E. Lifschitz, generalized in 1969 the use of these tensorial coefficients to the
case of dielectrics, conductors and superconductors in volume 8 of their
monumental Course on Theoretical Physics10, introducing them in a similar fashion

4 N. R. Libovitz, The mathematical development of the classical ellipsoids, International Journal of
Engineering Science vol. 36, pp. 1402-1420 (1998).
5 Maxwell, p. 67.
6 W. Thomson, W & P. G. Tait, P. G., Treatise on Natural Philosophy, vol. 2, Cambridge University
Press Cambridge (England), 2009 (new printing of the 1883 edition). P. Dive, Attraction des
ellipsodes homognes et rciproque d'un thorme de Newton, Bull. Soc. Math. France vol. 59,
pp. 128-140 (1931). W. Nikliborc, Eine Bemerkunguberdie Volumpotentiale, Math. Z. vol. 35,
625-631 (1932).
7 H. Du Bois, The Magnetic Circuit, Longmans (1896).
8 The name was used by Van Vleck, chapter IV, in 1932.
9 Stratton, pp. 206 and 213.
10 Landau and Lifchitz: conductors (pp. 7, 25, 28), dielectrics (pp. 42-45, 54), ferromagnets
(p. 163) and superconductors (p. 170).

 Depolarization tensor method 3

as Stratton11. They seem to be the first ones to point out the mathematical
equivalence of the behaviour of conductors and dielectrics with infinite
susceptibility12. In 1945 Stoner gave graphs and tables of the demagnetizing
factors for spheroids and general ellipsoids, without mentioning the tensor's
general properties. Almost simultaneously, Osborn gave a number of useful
formulas, tables and graphs for obtaining the values of the demagnetizing
coefficients. In 1966 Moskowitz and Della Torre analized some general properties
of the depolarization tensor13. In 1981 this author proved, for the magnetic case,
that this tensor also solves the anisotropic case and provides the non-uniform
value of the induced field outside the body14. In 2006 alternative methods were
proposed for the calculation of the tensor15. In 2008 this author discussed the
aplication to the electric case, including electrets and conductors16.

Applications
Elementary texts on electricity and magnetism solve only the behaviour of the
following fictive bodies (the equivalent ellipsoids are identified between
parenthesis):
point charge (ellipsoid with 3 equal and very small semi-axes);
sheet of uniform thickness and infinite extension (ellipsoid with 1 finite semi-
axis and 2 very large semi-axes);
cylinder of circular cross-section and infinite length (ellipsoid with 2 equal
semi-axes and 1 very large semi-axis);
volume of infinite extension (ill defined problem discussed in Chapter 5).
The sphere the finite body of highest simmetry is solved only in advanced
courses of electromagnetism by reducing Laplace's equation to a set of ordinary
differential equations with the canonical method of separation of variables in
spherical coordinates17. Solving triaxial ellipsoids (three different semi-axes) with
the same method requires advanced knowledge of metric properties and the use of
an uncommon and complex sistem of curvilinear coordinates18.
On the other hand, the depolarization tensor method requires only a basic
knowledge of vector analysis in order to solve the whole family of ellipsoids,
including the sphere, sheets of infinite extension, circular and elliptic cylinders of
infinite length. The internal fields may thus be expressed in terms of constant
matrices peculiar of each body, whose geometric part is a function of two
parameters which are the ratios of two semiaxes to the third (aspect ratios).

11 Landau and Lifchitz, p. 20.
12 Landau and Lifchitz, p. 40.
13 Moskowitz and Della Torre, pp. 739-744.
14 Solivrez (1981).

15 M. Beleggia, M. De Graef and Y. Millev, Demagnetization factors of the general ellipsoid: An

alternative to the Maxwell approach, Philosophical Magazine vol. 86, pp. 2451-2466, 2006.
16 Solivrez (2008).
17 Jackson, pp. 156-160; Reitz, pp. 93-94; Stratton, pp. 205-207.
18 Landau and Lifchitz, pp. 20-27; Stratton pp. 58-59, 211-213, 257-258.
 4 Carlos E. Solivrez

Only limited to the homogeneous case, the method is valid for any of the
following types of materials:
isotropic ones, like amorphous and policrystalline solids;
anisotropic ones, like single crystals;
permanent and induced electric (electrets, ferroelectrics...) and magnetic
(ferromagnets, antiferromagnets...) polarizations;
conductors (metals) and insulators (dielectrics);
superconductors.
The method allows to easily solve a large range of practical problems for
realistic finite bodies by using mostly elementary mathematical function and a
basic knowledge of vector analysis. The exception are triaxial ellipsoids which
due to the appearance of elliptic functions require the use of tables or software
like Mathematica or Maple.
Experimental determination of polarizations19 and of electric and magnetic
anisotropy constants are usually made by measuring forces and torques on
spheres, disks and cylinders placed in known uniform applied fields or appropiate
non-uniform ones20. Inversely, these forces and torques provide the value of the
field in which the bodies are immersed. Ellipsoids include or are good
approximations to those bodies21, also allowing an estimation of the errors caused
by departures from ideal shapes. They are also used in single crystals for the
calculation of lattice sums 22, the microscopic contributions of atomic and
molecular electric charges and dipole moments to the macroscopic field. There are
numerous technological methods and properties like magnetic resonance23,
electrostatic precipitation, magnetic coercivity of tapes and disks24 that can be
more easily interpreted when using ellipsoidal bodies. Also, some experiments
suggest that in certain cases ellipsoidal particles may appear spontaneously25.
Widely used anisotropic materials, seldom discussed in technical textbooks26
because of the difficulties of its mathematical treatment27, may be easily discussed
with the depolarization tensor method.
It is, therefore, surprising that the electromagnetic behaviour of ellipsoidal
bodies is absent from most textbooks on electricity and magnetism, being confined

19 In line with the use of the denomination depolarization tensor, a magnetization will be called
magnetic polarization in this book. Therefore, the term polarization refers both to the electric
and magnetic one.
20 H. Zijlstra, Experimental Methods in magnetism, vol. 2, North Holland Publishing Co., Amsterdam
(1967).
21 M. Beleggia et al., J. Phys. D: Appl. Phys. 39, pp. 891-899 (2006).
22 M. Widom, Shape-Adapted Ewald Summation.
23 J. H. Duyn and T. H. Barbara, Magn. Reson. Med., 72(1), pp. 1-3 (2014).

24 Q. F. Brown Jr. and A. H. Morrish, Phys. Rev. vol. 105, p. 1198 (1957).
25 G. R. Davies, The mechanisms of piezoelectricity and pyroelectricity in poly(vinylidene fluoride),
The Dielectric Society 1984 Meeting, Abstracts of Invited Papers.
26 A notable exception is the book by Landau and Lifchitz.

27 See, for instance, the Report of the Coulomb's Law Committee of the American Association of
Physics Teachers, Am. J. Phys. vol. 18, p. 1 (1950).

 Depolarization tensor method 5

to theoretical treatises and articles in specialized journals. This books intends to

fill this gap, collecting scattered formulas and properties, generalizing the formulas
and adding new ones whenever possible.

Requirements
The reader is required to be familiar with the differential and integral physical
laws relating electric and magnetic fields and potentials with their sources: electric
charge density and polarization, current density and magnetization. The required
knowledge of mathematics includes operations with matrices, derivatives, surface
and volume integrals, the combined application of vectorial operators and their
integral theorems. As previously mentioned, it is not necessary to solve Laplace's
equation, an unavoidable requisite unless the depolarization tensor method is
used.

Limitations of the method

The depolarization tensor method is valid only when:
The applied field is uniform over the whole body, but not necessarily elsewhere.
In the magnetic case the condition can be easily obtained with Helmholtz coils28.
The body's material must have a stable ellipsoidal shape which rules out
liquids) and uniform electric or magnetic properties in a macroscopic scale.
Otherwise, the material may be isotropic (amorphous or micro-policrystalline)
or anisotropic (single crystal or textured policrystalline29).
The applied field does not need to be fixed (unmodifiable by the introduction of
the body), except for simplicity in the calculation of the energy and the resulting
forces and torques. The method is valid for torque experiments whenever the
value of the applied field can be measured or calculated after the introduction of
the ellipsoidal body.

Dealing with singularities

Electromagnetic phenomena are fully described by Maxwell's equations and the
auxiliar constitutive equations which describe material properties. The standard
formulation makes use of vector analysis and techniques of resolution of partial
differential equations, but not of integro-differential equations like those given by
eqs. 2.31. The depolarization tensor formulation dispenses with the use of
differential equations, but makes intensive use of vector analysis and of integral
relationships like the divergence theorem, where special precautions must be
taken when dealing with singularities like point charges and jump discontinuities
across the body's surface:

28 W. Franzen, Generation of Uniform Magnetic Fields by Means of Air-Core Coils, Rev. Sci. Instr. vol.
33, pp. 933-938 (1962).
29 A policrystalline material is textured when its microcrystals have predominance of certain type
of faces. This often happens in laminated metallic materiales, because the lamination process
favours the appearence of certain crystal planes and disfavours others.
 6 Carlos E. Solivrez

point charge type singularity: 1/ r r ' cuando r r ', (1.1)
!
+
jump discontinuity: F(r ) F(r ) through)surface)S. (1.2)

These singularities appear repeatedly in surface and volume integrals, both for
the electric and magnetic case. As they arise from a simplified mathematical
representation of physical facts, to solve them is necessary to trace them back to
the basic physical phenomena they describe. This physical analysis is made in the
two following sections, while in Appendix 3 the integral theorems of vector
analysis are carefully reformulated in order to deal correctly with these
singularities, where it may not be valid, for instance, to interchange the order of
derivation and integration.
Scalar and vectorial magnitudes with constant values within a bounded region
play an essential role in the method. Electric and magnetic polarization, for
instance, have a finite and constant value inside and vanish outside the ellipsoidal
volume V. Applied electric and magnetic fields are constant only in region that
barely exceeds V, and may vary widely in the outside. The jump discontinuities of
polarization through the body's surface are the sources of the internal and external
induced field. The invariability of applied fields in V play a crucial role in the
determination of the forces and torques exerted on the ellipsoidal body (see
Chapter 5)
It is therefore convenient to have a common denomination for these kind of
fields, in order to clearly differentiate them from the constant vectors of
mathematics that are invariable over all space. For that reason, in this book a
vector is said to be uniform in a region V when it is constant in V (invariable
magnitude, direction and sense), but not necessarily outside.

Point charge type singularity

The origin of the point charge type singularity is the experimental Coulomb's
law (the value of constant k1 in different systems of units is discussed in Appendix
1),

rr'
F(r ) = k1qq' 3 = qE(r ), (1.3)
rr'
!
which gives the force experienced by a charge q in presence of a charge q'. The
experiment requires two different charges: the field's source q' and the charge q
where the force is applied (unless one is willing to invoke a "Mnchausen
!
effect"30). It is therefore necessary to differentiate between the field point r and
!
the source point r ' This requirement is ignored in the macroscopic formulation of
electromagnetism because otherwise fields would not be defined inside
continuous charge distributions. The price one has to pay for this simplification is
that the resulting macroscopic field differs from the microscopic field actually
experienced by atoms and molecules.

30 The mythical Mnchhausen baron raised himsel by pulling from his boot's straps.

 Depolarization tensor method 7

Coulomb's law gives origin to the definition of the differential electric field
! 3
!dE(r ) created at r by an infinitesimal charge !dq = (r ')d r ' at !r '. This field is
expresssed in terms of the volume charge density by

rr' rr' 3
dE(r ) = k1 3 dq' = k1 3
( r ')d r ', (1.4)
rr' rr'
!
where field and source points are explicit, as will be always done in all integrals
herein.
The total electric field created by the complete charge distribution is thus
given by

rr'
E(r ) = k1 3 (r ')d 3r '. (1.5)
V rr'
!
Expressions 1.3 and 1.5 do look similar, but they differ in a fundamental way.
! !
While for Coulomb's law it is not valid to take r = r ', the electric field expression
requires its validity for all volume and surface integrals. There are additional
problems because the volume charge density for point charges is not a well
behaved function, and the same happens with surface charge densities. These
singular densities may be given mathematical rigour by the use of distribution
theory, a method that will not be used here for the evaluation of integrals of this
kind because most engineers are not familiar with it. Instead, appropiate care will
be taken for the evaluation of singular integrands with standard methods, an
example of which is the calculation of the trace of the depolarization tensor in eqs.
3.15 and A3.7.
The simplest case of dealing with point charge type singularities is that of eq.
!
1.5. One may take a system of spherical coordinates with origin at point r so that
the integral becomes
2 R( ; )
! ! R ! ! !
E(r ) = k1 d sen d 2 ( , ,R) R2 dR con R = r ' r . (1.6)
0 0 R

The singularity turns out to be integrable because of the factor R2 in the differential
volume element. As will be discussed later on, this does not mean that such
singularities do not affect the validity of the standard divergence and curl
theorems of vector calculus.
Let's examine the first of these theorems. Coulomb's Law eq. 1.3 is the origin of
Gausss Law
! ! !

" E( r )i dS = 4 k1QS , (1.7)
S
where QS is the charge inside the closed surface S. According to the divergence
theorem eq. A3.8,
! ! ! ! !
E(r )i dS = i E(r )d r = 4 k Q
" = 4 k1 dV ,
3
1 S
(1.8)
S V V
 8 Carlos E. Solivrez

where V is the region bounded by S. The result is one of Maxwell's equations (see
eq. A1.3)
! ! ! ! ! !
i D(r ) = 0 i E(r ) = 4 k1 0 , so that i E(r ) = 4 k1 . (1.9)

The volume charge density of a point charge eq. 1.3 would then be

! ! ! !
1 q' r r'
= i E(r ) = i ! !3
4 k1 4 r r '


q' 1 ! ! 1 ! !
= i(r r ')+ ! ! 3 i(r r ') (1.10)
4 r! r! ' 3 r r'

! ! ! !
q' 3 3(r r ')i(r r ')
= = 0,
4 r! r! ' 3 r
! ! 5
r '


where use has been made of eq. A2.5.
The experimental validity of Gauss's Law has been thoroughly verified31, so it
follows that eq. 1.9 does not hold. Some authors try to "solve" the problem by
denying the validity of eq. 1.10 with dubious mathematical arguments32, but the
origin of the paradox is that the divergence theorem eq. A3.8 is not valid for
singular fields like the one given by eq. 1.3.
A similar problem arises with the calculation of the following expression, that
will be used later on for the calculation of the depolarization tensor trace:

d 3r '
i ! ! . (1.11)
V rr'

This expression is well known in physics, where it is usually solved by using
Dirac's delta function, that is, invoking distribution theory. Some authors of texts in
electromagnetism justify its value by juggling with the order of derivation and
integration, frequently making invalid manipulations. The topic is discussed in
Appendix 3 in order to make a valid derivation of eq. A3.7 with no recourse to
Dirac's delta function.
One may deal with singular fields in two different ways. The first one, which
preserves Maxwell's equations even for singular charge distributions, is to extend
the definition of charge (and current) densities in order to include point charges,
surface and line charge distributions (and the corresponding current
distributions). This requires distribution theory, a complex mathematical


31 See, for instance, Young and Freedman, p. 721.
32 Reitz, p. 45, ignores the basic rule of making all possible simplifications before taking limits, not
afterwards.

 Depolarization tensor method 9

formalism33 well known to physicists but seldom used by engineers. The other way
discussed in Appendix 3 is to extend the validity of the integral theorems of
vector analysis for the case of singular charge distributions like point charges. The
latter approach is used by some texts on electromagnetism, but usually not
explicity stated as a generalization of the standard theorems.

Step discontinuities
Step discontinuities present similar difficulties. In a well known text on
electromagnetism which the author used for his first studies of the subject, the
following integration by parts is made34 (here rewriten in our notation):
! ! !0 !
(r )i ( D(r ) D (r ))d r
! 3

V'
(1.12)
( ( ))
! ! ! ! ! !
!
= i D D0 d 3r (r ) i D(r ) D0(r ) d 3r , ( )
V' V'

! !
where is an electric potential, D, D0 are the displacement vectors in the
presence and absence of the material body, and use has been made of eq. A2.5. The
integration volume V ' includes both the dielectric body and the sources of . From
divergence theorem eq. A3.8, Jackson argues that, as the integration surface S' is

outside the body, it follows that there !D = D0 and

( ( ))
! !0 3 ! !0 2!
i D D d r = S' D d r = 0,
" D ( ) (1.13)
V'

where !d 2r is the differential element of area vector normal to the surface.
The standard divergence theorem may be used when the the first partial

derivatives of the integrand are continuous in V'35, but !D D0 has a step
discontinuity across the body's surface S. One may avoid the problem by dividing
the volume of integration in two by the surface S of step discontinuities, the
method used in the formulation of the generalized divergence theorem eq. A3.18.
Using this theorem eq. (1.12) gives

( ( ))
! !0 3 ! !0 2! ! + ! 2!
i D D d r = S'
" D D d r S
" D ( )
D d r 0, ( ) (1.14)
V'
because the last integrand does not vanish.
This sort of problems, which appear frequently in texts on electromagnetism,
usually remain undetected because such complex integrals are seldom computed
for real distributions of matter. This was not our case because it was quite easy to

33 For a simplified formulation based on L. Schwartz, Theorie des distributions (Hermann, Paris,
1950) see R. J. Gagnon, Am. J. Phys. vol. 38, pp. 879-891 (1979).
34 Jackson, p. 125.
35 Korn and Korn, p. 163.
 10 Carlos E. Solivrez

use the depolarization tensor method to compare the initial expression for an
ellipsoidal body with the final one, finding no match.

Physical units and mathematical notation

Texts used by physicists like Jackson's tend to use the Gauss system of units
because it's more suited for relativistic analysis, while engineers uniformly use
standard SI units. Instead of committing ourselves to one of the two, we use the
general system introduced by Jackson through constants k1, k2, k3, and ' defined
and tabulated in Appendix 1 for all commonly used systems of units36.
The mathematical notation used in this book uses the following conventions:
Vectors (fields, polarizations...) and tensors (depolarization, susceptibilies...) are
represented in two alternative ways:
1. As linear combinations of products of components and unit vectors such as

Variables are always written with italics, vectors are always
E(r ) = E z (r )z.
!
crowned with arrows and unit vectors with ^. The omission of the arrow

indicates the vector module, E = E. Second rank tensors are represented with
!
italics and bold letters, and its components are given in dyadic notation37 (see
Appendix 6)
! !
n(r ) = n (r )x x , (1.15)


thus making clear that its scalar product with a vector gives another vector. This
intuitive and compact notation simplifies the derivation of equations, but not its
resolution. For the latter it is more conveniento to use the following alternative
notation.
2. Vectors are alternatively represented as column or row matrices of dimensin
1 x3 or 3x1. Second rank tensors are then represented as square 3 x3 matrices,

as in eq. 3.66. The only disadvantage of this representation is that depends on

the chosen coordinate system, but so does the diagonal form (see below) of the
depolarization tensor.
V denotes both the region occupied by the ellipsoidal body (in integrals) and the
magnitude of its volume (in equations), concepts easily discriminated from the
context.

S is the boundary surface of region V (the body) and !s(r S ) the outgoing unit

vector normal to S at point !r S .
x, y, z are orthogonal cartesian coordinates.
Different coordinates are generically called x , where = x, y, z and xx = x, xy = y,
xz = z.

36 Jackson, pp. 613-618.
37 Morse and Feshbach, pp. 54-92.

 Depolarization tensor method 11

Unless explicitly stated otherwise, the cartesian coordinate system in use is the
one that diagonalizes the matrix representation of the depolarization tensor, the
ellipsoidal's body principal system of eq. A7.2.
1 if =
is Kronecker's delta function (or matrix) defined by = .
! 0 if
Therefore, the unit vectors corresponding to each of the orthogonal coordinate
axis fulfill the condition x i x = .
!

Symbol = x + y + z is the gradient operator nabla such that its
! x y z
application to a scalar function gives a vector. The divergence operator is
written as , the curl as x and the laplacian as or .

!r ' is a source point for the field (electric charges, polarization, electric
current...), which is always explicity distinguished from the field point where the
generated field acts upon.
d2r' is the scalar differential element of area in a surface integral.

!d 2r ' is the vectorial differential element of area in a surface integral.
d3r' is the scalar diferential element of volume in a volume integral.

Organization of the book

Three categories of materials have very similar behaviour both in the electric
and magnetic case: spontaneously polarized ones (like electrets, ferroelectrics and
ferromagnets); those polarized only in the presence of fields (like dielectrics,
diamagnets and paramagnets); those with vanishing internal fields (conductors
and superconductors). In order to facilitate comparisons, in Chapter 2 these cases
are separately discussed for electric and magnetic substances.
In permanently polarized materials a uniform applied field is necessary only for
maintaining and orienting the uniform polarization. The body's internal and
external fields are then fully characterized by the depolarization tensor n, both
with or without applied fields.
For substances that get polarized only in the presence of applied fields, the
linear approximacion is valid except in unusually high fields. In this approximation
the induced polarization is proportional to the uniform applied field through a
susceptibility that is a scalar in the isotropic case and a second rank tensor in the
anisotropic one. It is then found that the induced polarization and the internal field
both uniform for ellipsoidal bodies may be fully expressed in terms of n , the
susceptibility and the applied field. This result is well known for isotropic
materials, but has never been fully discussed for the anisotropic case38. In the
linear range a tensorial relationship the body's polarizability, function of n and
39may be stablished between the body's electric or magnetic moment and the
applied field, just as in the molecular case40.


38 See the discussion made by Landau and Lifchitz in pp. 58-61.
39 Landau and Lifchitz, p. 7, introduce the tensor but do not relate it to n.
40 Kittel, p. 459.
 12 Carlos E. Solivrez

According to the standard definition of macroscopic fields, there is no

polarization inside a conductor or a superconductor. However, a fictive
polarization (in this book called equivalent polarization) may be defined such that
the problem turns out to be equivalent to the case of induced polarization. The
equivalent polarizations gives both the right value of the internal fields and the
body's electric and magnetic moments. For conductors the corresponds to the limit
of a perfect dielectric, that of infinite electric susceptibility. For a superconductor,
it corresponds to the limit of perfect diamagnetism where the diamagnetic
susceptibility equals -1.
The integro-differential equations that determine induced fields and internal
polarizations are derived for all cases and collected in Table 1. An ansatz
corresponding to the uniform solution satisfies all these equations. At the end of
Chapter 2 an iterative method of solving the integro-differential equations is given,
which confirms the value of the uniform solutions.
All along Chapter 2 the depolarization tensor and some of its properties are
used without proof in order to show its convenience and way of use. Its definition
and elementary analysis starts at Chapter 3, where some expressions are given for
the tensor, as well as a simple method for determining its principal values in the
case of the sphere, the infinitely long circular cylinder and the infinite sheet of
constant thickness. Chapter 4 tackles the more complex case of the spheroids,
uniaxial ellipsoid and elliptic cylinder, as well as the study of the value of the
depolarization tensor outside the body for triaxial ellipsoids.
Chapter 5 gives a (too) brief introduction to the calculation of electromagnetic
energies, from which are derived the expressions for forces and torques exerted on
ellipsoidal bodies immersed in uniform applied fields. Infinite bodies and cavities
are discussed at the end of this chapter, showing the inconsistencies in their
standard analysis wich usually invokes a rigid behaviour of the polarization. Both
topics present difficulties rarely discussed in standard textbooks; they deserve a
more thorough analysis that is barely sketched here, probably a whole book.
The inclusion of Chapter 6, which gives detailed solutions to illustrative
exercises on all previously discussed topics, requires justification. This book is
neither a treatise on electromagnetism, nor one on the theory of polarized bodies.
It is an introduction to a practical method of calculation of static electromagnetic
properties of ellipsoidal bodies with no recourse to the resolution of Laplace's
equation, that is, to the techniques of resolution of partial differential equations.
Exercises have, therefore, a central role, but here they are not dispersed through
the text as is customary. Two main arguments justify grouping them in a single
chapter.
The first one is that the specificity of concrete problems distracts the reader
from the main line of reasoning. In the few ocasions where the author felt that the
problem is relevant to this reasoning, an actual case is discussed in the pertinent
place, but a link is algo given in Chapter 6. The necessity of this link is a
consequence of the second justification for the existence of this chapter.
This book is conceived as a complement to ordinary texts on electromagnetism
because of their customary lack of treatment of finite bodies, usually reserved to
courses with advanced knowledge of partial differential equations. Therefore
although the author does not recommend it it is expected that its use will be

 Depolarization tensor method 13

occasional and fragmentary, which requires special provisions. The given

organization makes the selection of exercises easier. So does the redundancy in the
discussion of central concepts and the provision of a detailed alfabetical index.
 14 Carlos E. Solivrez

 Depolarization tensor method 15

Chapter 2:
Ellipsoids in Electric and Magnetic Fields

This chapter discusses the uniform polarizations permanent, induced and

equivalent and the fields generated when homogeneous ellipsoidal bodies are
immersed in uniform applied fields. Results are expressed in terms of the
depolarization tensor n, leaving the solution of specific cases for Chapter 6.

Electric polarization
Basic equations
The discussion of the subject first made by the author in 200841 is here modified
in order to make its understanding easier. The electric polarization of an
ellipsoidal body V is usually a function of the position and generates an electric
! !
field E P! (r ) derivable from a potential42:

! ! ! !
! P(r ')i(r r ) 3 ! ! !
P (r ) = k1 ! ! 3
d r ', E P
( r ) = P
( r ). (2.1)
V r r

The formula is the extension of the potential of a point dipole to the case of a
continuous distribution, the macroscopic representation of a collection of neutral
atoms where the baricenters of the nuclear and electronic charges do not coincide.
Poisson was, apparently, the first to notice that

!

!
( ) 1
P (r ) = P i k1 ! ! d 3r ', (2.2)
V r r

The characterization of a uniformly polarized ellipsoidal body may thus be derived
from that of a uniformly charged one, a property that will be widely used in this
book.
As discussed in p. 7, the integrand in eq. 2.1 has an integrable singularity. It is
frequently said that the macroscopic electromagnetic fields are the average of the
microscopic fields, but it has been convincingly argued that one should take,

41 Solivrez (2008).
42 Reitz, p. 78 eq. 4-7.
 16 Carlos E. Solivrez

instead, the average of the potential43. The subject, not discussed in this book, is
mentioned only because the depolarization tensor may have an important role for
determining the connection between those two scales44. At the same time, a good
understanding of the origin of fundamental expressions as eq. 2.1 (and similar
magnetic ones) is the best guide for the interpretation and resolution of
mathematical singularities originated in mathematical idealizations of actual non-
singular distributions of charge.
The potential eq. 2.1 may be rewritten in two different ways. The first one will
be used for the discussion of an ellipsoidal body with permanent or induced
electric polarization. The second one, for conductors.
Using eqs. A2.2 and A2.5 the integrand of eq. 2.1 may be rewritten in the
following fashion :
! ! ! ! 1
! P(r ')i(r r ) 3 ! !
P (r ) = k1 ! ! 3 d r ' = k1
P( r ')i 3
! ! d r ',
V r r V r r
! ! ! ! (2.3)
P( r ') P(r ')
k1 i ! ! d 3r ' = k1 i ! ! d 3r '.
r r
V V r r

The last step requires a permutation of the order of derivation and integration, an
operation that may not be valid for singular integrands. The problem has been
studied for this kind of integrals in the gravitational case and proven to be
mathematically valid45.

The total electric field !E(r ) is the vectorial sum of the uniform applied field !E 0

and the induced field !E P :

! !
! ! ! ! ! P(r ') 3
E(r ) = E 0 P (r ) = E 0 + k1 i ! ! d r ' , (2.4)
V r r

equation valid in all space for all sorts of polarizations. This equation will be the
starting point for the treatment both of permanent and induced electric

polarizations. When !P is uniform even if it's a function !P(E int ) of the internal
electric field the last equation may be rewritten as

43 M.C. Vanwormhoudt, On the definition of macroscopic electric and magnetic fields, Physica vol.
42, pp. 439-446 (1969).
44 M. Widom, Shape-Adapted Ewald Summation. Dekker, pp. 141-144. Reitz, pp. 81-83. Jackson

pp. 115.
45 MacMillan, pp. 27-32.

 Depolarization tensor method 17

! ! ! 1 ! !
E(r ) = E 0 + k1 i ! ! d 3r ' P(E int ) . (2.5)
V r r

From eqs. 3.3 and A1.4, it follows that

! ! ! ! ! ! !
E(r ) = E 0 n(r )i P(E int ) r , (2.6)
0

valid both inside and outside V. The depolarization tensor n is defined by eq. 3.4,
being uniform only inside ellipsoidal bodies, as pointed out in page 2. In order to
solve this equation it is necessary to write separate expressions for the regions
inside and outside the body:

! ! ! ! !
E int = E 0 N i P(E int ) para r V ,
0
! ! ! ! ! ! ! (2.7)
E ext (r ) = E 0 next (r )i P(E int ) para r V ,
0

!
where E int is the uniform internal electric field, !E ext (r ) the external one, N the

internal depolarization tensor (uniform for ellipsoids) and !next (r ) the non-uniform

external one which is always a function of position !r . The eigenvalues of matrix N
are always positive, making the internal electric field (and the magnetic one, as will
be seen later for the ferromagnetic case) smaller in magnitude than the applied
one, fact that gave origin to the tensors name. Equations 2.7 are valid both for
permanent and induced polarizations, but each case is solved differently. When an
ellipsoidal body is uniformly polarizedstate not spontaneously acquired for the
permanent case the first equation shows that the resulting internal electric field
is also uniform.
Potential eq. 2.3 may be expressed in terms of polarization densities. This
requires a different use of identities A2.2 and A2.5:
! ! ! ! ! ! ! !
P(r ')i(r r ) ! ! 1 P(r ') 'i P(r ')
! ! 3 = P( r ')i' ! ! = 'i ! ! ! ! , (2.8)
r r r r r r r r


where ' operates on !r ' . Replacement of the integrand in eq. 2.3 then gives

! ! ! !
! P(r ') 3 'i P(r ') 3
P (r ) = k1 'i ! ! d r ' k1 ! ! d r '. (2.9)
V r r V r r

The first volume integral appears well suited for the application of divergence
theorem eq. A3.8, and most textbooks apply it without further ado, replacing it
 18 Carlos E. Solivrez

with a surface integral46. In fact, further analysis is required because !P has a step
discontinuity on the body's surface because is finite inside and vanishes outside.

That is, the derivatives of vector !P do not fulfill the standard conditions for the
application of the theorem because they may be different on both sides of the
body's surface47. It turns out that this does not affect the value of the divergence in
the volume integral and of the flux in the surface integral. The proof is sketched in
Appendix 3 (see the derivation of eq. A3.21).
It is thus obtained

! !
! P (r ') 3 P (r ')d 2r '
P (r ) = k1 ! ! d r ' + k1 "S r! r! ,
V r r (2.10)
! ! ! ! ! ! !
where P (r ) = i P(r ), P (r ) = P(r )i s(r ).


where !s(r ) is the unit vector normal to

surface S at point !r (see eq. A7.8). These
charges, as real as the conduction ones,
originate in the relative displacement of
the baricenters of nuclear and electronic
charges of atoms and molecules. The
expression will later used for solving the
case of conductors. As only uniform
polarizations will be considered in this Figure 1. Origin of the
book, P vanishes and the only depolarization effect.
contribution to the electric field is that
of the surface density of charge P. This explains the fact that inside the body the
magnitude of the electric field is smaller than in the outside, as shown in Figure 1.

The electric displacement vector !D(r ) is introduced in order to discriminate the
polarization charges, bound to atoms and molecules, from the free conduction
ones:
! ! ! ! ! !
D(r ) = 0E(r )+ P(r ). (2.11)

This constitutive equation for the electric case will be seldom used here, the
magnetic case being quite different.

Permanent electric polarization

The necessary condition for the presence of permanent electric polarization is
for some atoms or molecules of the material ferroelectric, electret or the like
below their critical temperature to have a finite electric dipole moment in the

absence of an applied electric field (!E 0 = 0 ). The state of polarization of these


46 See, for instance: Reitz, eq. 4-12 in p. 78.
47 J. Rey Pastor, P. Pi Calleja, and C. Trejo, Anlisis Matemtico vol. 1, p. 437 (1959, 4th edition).

 Depolarization tensor method 19

materials depends on the body's internal field !E int (r ) , but also on other
microscopic and macroscopic phenomena. The dominant ones among the former
are the atomic and molecular interactions, whose magnitude may outweight the
applied field. Among the latter the most important ones are the atomic and
molecular vibrations (temperature) and the existence of domains, local ordering of
electric moments that tends to minimize their interaction energy. The influence of
domains is a difficult topic because their size and statistical distribution are
critically dependent on the material's thermal history, their evolution consisting in
transitions between metaestable states48. This leads to irreversibility (hysteresis)
and a usually non-uniform polarization.
A reasonable approximation to uniform polarization may be obtained with a
controlled process involving a combined variation of temperature and the
application of a high field (saturation). Once a sufficiently uniform polarization has
been obtained, one has to determine its value. The most common way of doing this
is to suspend the body in a uniform field and measure the applied torque. The
relationship between field and torque involves the depolarization tensor (see the
corresponding section in page 105).
In this book no further details will be given of these topics, which are
thoroughly discussed in specialized books and scientific journals. What matters
here is that whatever the method used the necessary condition for a uniform
permanent polarization, although not a sufficiente one, is an ellipsoidal shape.
Edges and corners, for example, are examples of shapes that preclude a state of
uniform polarization even in extremely high applied fields.

When a uniform permanent polarization !P is stablished and its value is
determined, the depolarization tensor method can be used to calculate both the
internal and external electric field through eqs. 2.7. These equations, rewritten in
matrix notation, read

!
Eint = E0 N i P(Eint ) for r V ,
0
(2.12)
! ! !
E (r ) = E0 next (r )i P(Eint ) for r V .
ext

The external field !Eext cannot be determined without solving first the equation
for for Eint and then replacing its value in the second of eqs. 2.12. This requires the

knowledge of the function !P(E int ) , which in the isotropic case is usually taken to be
Langevin's function49. Problem 06 of chapter 5 gives a tentative graphical method
for this determination. In anisotropic materials the solution requires the separate

48 Brown (1963).
49 Dekker, pp. 138 and 192.
 20 Carlos E. Solivrez

determination of a number of functions determined by the single crystal

symmetries (see Appendix 5).

Induced electric polarization

In the absence of permanent polarization the induced polarization is usually
!
only a function of the applied field E 0 , the electric susceptibility e of the
!
magnetic material50 and the body's shape. This is so because the state of
polarization of any point of the body depends on the internal electric field, the
addition of the applied field and that generated by the rest of the body. The
induced field should, therefore, be determined in a self-consistent way, its value at
each point depending on the value at all other points. The mathematical expression
of this dependence is clearly expressed by the integral eq. 2.31.
The applied field, the polarization and the total electric field are in general not
parallel due to a combination of the anisotropy of the susceptibility (a material
property) and the shape anisotropy, a geometric property characterized by the
depolarization tensor, linear only in the case of ellipsoidal bodies. That is, the
linearity of the relationship between the polarization and the applied field is not a
mere consequence of the linearity of the field equations, as invoked by Landau and
Lifschitz to justify the introduction of the body's polarizability tensor. They claim
there that51 Since all the field equations are linear, it is evident In fact, there is a
complex relationship among the applied field components, the ensuing
polarization and the resulting total field, mediated by the material's properties and
the body's geometry. In the treatment made by standard texts on the theory of
electromagnetism these relationships are hidden in the boundary conditions of
differential equations. In the method described by this book both the anisotropy of
the material and the influence of geometry are explicitly expressed by the
susceptibility and depolarization tensors, while the interrelationship among fields
is expressed in terms of linear systems of equations among components, clearly
visible and soluble in matrix representation. A better understanding of this
abstract behaviour may be obtained from a concrete illustration. Two specific
examples are given to that end, the first one in Figure 2, the second in the following
section.
Figure 2 is a storyboard of the timed sequence of interactions of two polarizable
atoms subject, in vacuum, to an applied field. The same problem is next
mathematically solved using matrices, in a fashion closely resembling the
depolarization tensor method, to which it is an introduction.
The figure relies on the experimental fact that fields do not act instantaneously,
but propagate with the velocity c of electromagnetic waves. The applied field
arrives to atom 1 at time t1, inducing a dipole moment which acquires its full value
in a later time t2, when atom 2 is still unpolarized. At time t3 the applied field
arrives to atom 2, but not so the one created by the dipole moment induced on
atom 1. Molecule 2 acquires its initial dipole moment at time t4, but this value will
change when the induced field of atom 1 reaches its location. Shortly after that, at


50 Remember that this is a low field approximation, the most common but not universal case.
51 Landau and Lifchitz, p. 7.

 Depolarization tensor method 21

time t5, the modified dipole moment of atom 2 is fully stablished. But the process
does not end here, because atom 2 creates a new induced field that acts on atom 1,
modifying again its dipole moment at time t6. The succesive changes of
polarizations and fields take a certain time to reach a value within a small but non-
zero precission range. It is thus clearly seen how the values obtained are
necessarily self-consistent, because the dipole moment of each atom depends on
that of the other through the induced fields created by both.

Figure 2. Autoconsistent mutual polarization of two atoms.

Induced electric polarization of two interacting atoms

In order to find the equilibrium values of the induced electric dipole moments
and total field it will be assumed that both atoms are identical, isotropic and have
no permanent electric moments. This happens, for instance, for closed electronic
shells. The atoms, a distance d apart, are immersed in a uniform applied electric
field E0 and the induced fields will be calculated using the point dipole microscopic
model.
Under the influence of an applied uniform field E0, electric dipole moments p(1)
and p(2) are induced on the atoms at r(1) and r(2. In the linear range, for an isotropic
atom j its induced dipole moment is related to the microscopic electric field
experienced by
p(j) = E(r(j)), (2.13)

where constant is the same for both atoms. Field E(r(1)) is the addition of the
applied field E0 and the field E(2)(r(1)) generated at r(1) by the dipole moment of the
atom at r(2). In the rest of this section arguments of functions are always field
 22 Carlos E. Solivrez

points and upper indices identify source points. Using eq. A4.5 and the following
property of the interatomic distance52
(1) (2) (2) (1) (2) (1)
!d ( r ) = r r = d = d (r ) = (r (1) r (2) ), (2.14)
!
it turns out that a matrix m may be defined for both atoms, in terms of vector d ,
such that
E(1)(r(2) ) = k1m i p(1) , E(2)(r(1) ) = k1m i p(2) .
(2.15)

From eq. A4.5 it is obtanined

3d 2 d 2 3d d 3d x d z
1 x x y

m = 5 3d y d x 3d 2y d 2 3d y d z ,
d (2.16)
3d z d x 3d d 3d 2
d 2

z y z
!
where d = d = d x2 + d 2y + d 2z ,


d being the components of vector !d.
!
From eq. 2.13, it follows that the fields and dipole moments are determined by
the following system of equations:

E(r(1) ) = E0 + E(2)(r(1) ) = E0 + k1m i p(2) = E0 + k1m i E(r(2) ),

E(r(2) ) = E0 + E(1)(r(2) ) = E0 + k1m i p(1) = E0 + k1m i E(r(1) ),
! (2.17)

where it is clearly seen that the field on atom 1 (upper equation) explicitly
depends of the field on atom 2. As the field on atom 2 explicitly depends of the field
on atom 2 (lower equation), this self-consistency is expressed by the following pair
of matrix equations (that is, a system of 6 linear equations):

E(r(1) ) = E0 + k1 m i E(r(2) ), E(r(2) ) = E0 + k1 m i E(r(1) ).

! (2.18)

As the two atoms are identical, upon its exchange the system remains
unchanged. That is, E(r(1)) = E(r(2)). This symmetry argument may be verified by
substracting the second equation from the first obtaining

!
( ) (
E(r(1) ) E(r(2) ) = 1i E(r(1) ) E(r(2) ) = k1 m i E(r(2) ) E(r(1) ) ,

) (2.19)

where 1 is the unit matrix. That is

52 The notation, which appears to be excessively detailed, is well suited for the calculation of
lattice sums, the microscopic fields created in single crystals by ordered atoms and molecules.

 Depolarization tensor method 23

(1 k m) i (E(r
! 1
(1)
)
) E(r(2) ) = 0.
(2.20)

As the left member's square matrix is non-singular (see below), the column matrix
must vanish, verifying that

!E(r ) = E(r ) = E.
(1) (2)
(2.21)

Therefore, the original two equations system reduces to the single matrix equation
(notice the similarity with the second of eqs. 2.7)
E = E0 + k1 mi E, (2.22)
!
from which E may be obtained, as follows:

(1 k m) i E = E , ( )
1
0
E = 1 k1 m i E0 . (2.23)
! 1

Matrix 1-k1 m inverse is easily obtained in the coordinate system where it is

diagonal (its principal one), usually corresponding to the atomic system highest
symmetry. In this case the highest symmetry is an arbitrary rotation around the

atomic axis defined by vector !d . In fact, if one chooses a coordinate system such
that both atoms are on axis z, and the origin is arbitrarily chosen to be on atom 1, it
is obtained
1/d 3 0 0

d x = d y = 0, d z = d, m = 0 1/d 3 0 . (2.24)
0 0 2/d 3

Then
E
x

( )
1
E = E y = 1 k1 m i E0
E
z
(2.25)
1/1+ k /d 3 0 0 E0
1
x0
= 0 1/1+ k1 /d 3
0 i Ey .

0 0 1/1 2k1 /d 3 E z0

Choosing the proportionality constant in such a way that both members have
the same units (see eq. A1.3), one may define the atomic electric polarizability
tensor e by
!

( )
1
p = E = 0 e i E0 , where e = 1 k1 m . (2.26)
0

 24 Carlos E. Solivrez

The matrix C relating the resultant to the applied field, E = C E0, has the typical
form of the tensorial properties of uniaxial materials53,

c 0 0

C= 0 c 0 where c < c! ,

0 0 c!
(2.27)

in the single crystal's principal coordinate system. A simple estimation illustrates

that the difference in value of these two coefficients is easily measurable. k1 is of
order R3, R being the atomic radius54. In solids the interatomic distance d is or
order 2R. Therefore, k1 /d3 is about 0,1 which gives

k1
3
0,10, c 0,91, c 1,25,
!d (2.28)

the parallel coefficient being about 37% larger than the one perpendicular to the
interatomic axis.
The previous calculation corresponds to the local field of solid state physics, the
microscopic electric field experienced by an atom. This field does not coincide with
the macroscopic one described by Maxwell's equations, a suitable average of the
former or its potential over a single crystal's unit cell or a sufficiently large number
of atoms or molecules in the policrystalline or amorphous case. An important
feature of macroscopic fields, as discussed in page 6, is that any of the body's
internal points is both a field and a source point. The singularities thus created did
not arise in the previous treatment because point dipoles where used. A
consequence of this approximation is that matrix m corresponds to the exterior
depolarization tensor next, as shown by its zero trace (see eq. 3.15). This suggests
that it may not be feasible to obtain the internal depolarization tensor N as an
average of microscopic tensors like m, unless significative mathematical and
conceptual changes are made in the treatment. On the other hand, atoms are not
geometrical points, but no concept related to the depolarization tensor seems to
have been introduced for its (necessarily) quantum analysis.

Dielectrics
As in the case of permanent polarization, the starting point is eq. 2.12. In most
dielectrics applied fields are significantly smaller than the effective electric fields
of quantum interactions among electrons. It is then valid to keep only first order

terms in the Taylor's expansion of !P(E(r )) :

! ! ! !
P(r ) = 0 e i E(r ). (2.29)



53 Nye, p. 23 Table 3, see Appendix 5 in the present book.
54 Dekker, p. 135. C. Kittel, Introduction to Solid State Physics, Wiley and Sons, p. 229, 1979.

 Depolarization tensor method 25

There is no general agreement on the constant one should use for the definition of
electric susceptibility. The adimensional definition given above and in the book by
Stratton55 simplifies our main formulas, while Reitz convention introduces no
factor in eq. 2.2956. The reader should carefully verify the definition used by his
favourite author. With the above convention the following value of electric
permeability is obtained,
! ! ! ! ! !
D = 0E + P = 0E + 0 e i E = i E , where = 0(1+ e ). (2.30)

In homogeneous materials, the only ones studied in this book, the adimensional
electric susceptibility e is a uniform and symmetric rank 2 tensor57 which
!
reduces to a scalar in the cases of non textured policrystalline and amorphous
materials. In the anisotropic case some of its components may be related,
depending on the symmetries of the material's crystalline structure58.
From eq. 2.5, the total electric field due to the application of a uniform field to a
dielectric with the linear property eq. 2.29 satisfies the equation
! !
! ! !0 1 e i E(r ') 3
4
E(r ) = E + i ! ! d r ' . (2.31)
V rr'

This system of integro-diferential equations univocally determines the electric
field59 and, through eq. 2.29, the polarization. In the permanent polarization case
the internal field's uniformity was a consequence of a deliberately stablished
uniform polarization. Here, on the contrary, the uniform character of the field is a
consequence of eq. 2.31, from which follows that of the polarization.

A uniform internal electric field !E int provides a solution to eqs. 2.31 if the
following equations are satisfied (which are easily obtained in the same way as
eqs. 2.7):
! ! ! ! 0 !
E int = E 0 N i e i E int , P = e i E int for r V ,

(2.32)
! ext ! ! 0 ext ! !
E (r ) = E n (r )i P for r V .
0

In matrix representation it is easy to obtain Eint from the first equation from which
the values of P and Eext immediately follow:

55 Stratton, p. 12 eq. 8.
56 Reitz, p. 86.
57 Landau and Lifchitz, p. 58.
58 Nye, p. 23 Table 3: see Appendix 5 of the present book.

59 See the initial discussion in Solving the integro-differential equations by iteration.

 26 Carlos E. Solivrez

( )
1
Eint = 1+ N i e i E0 for r V ,
0 0
( )
1
P= e i Eint = e i 1+ N i e i E0 ,

0

( ) (2.33)
1
p=V P= e i E0 , e = V e i 1+ N i e ,

! !
Eext (r ) = E0 next (r )i P for r V ,
0

where A-1 is the inverse of matrix A. The relationship between the geometry of
ellipsoidal bodies and electric polarization P was stablished by Landau and
Lifschitz for the isotropic case60 through a generalization, which is not a proof, of
the results obtained for the sphere and the infinitely long cylinder:
The existence of such a relationship follows from the form of the
boundary conditions, as we saw above on the examples of the sphere and
the cylinder.
The geometric connection, previously extended by this author to the anisotropic
electric case61, is but a simple consequence of the method that allows the
introduction of e , the bodys electric polarizability tensor with units of volume:
!

( )
1
p = V P = 0 e i E0 , e = V e i 1+ N i e . (2.34)

This tensor was defined by Landau and Lifschitz62 only for conductors and
magnetic materials. Notice the non-casual similarity of the formula with eq. 2.26.
This tensor will be used in Chapter 5 for the calculation of the body's energy and
applied torques.

Magnetization

The behaviour of ellipsoidal bodies under applied uniform magnetic fields !H 0
the first work published by the author on the depolarization tensor method63
may be solved in a similar way as the electric case.

Basic equations
Unlike the electric case where the treatment of polarized matter is done using

the fundamental field !E in the permanent and induced magnetic cases use is

60 Landau and Lifchitz, p. 44.
61 Solivrez (2008), eq. 17.
62 Landau and Lifchitz, pp. 7 and 192.

63 Solivrez (1981).

 Depolarization tensor method 27

made of the auxiliar field !H. The magnetic induction !B will be used only in the
case of superconductors, where use must be made of the constitutive eq. A1.10,
! ! !
B = 0(H + 'M). (2.35)

The reason for using the magnetic field instead of the magnetic induction is that
all equations can then be derived as for the electric case because of the existence of
the following magnetic potential M 64:
!
! ! ! !
! ! ! ! ' M(r ')i(r r ') 3
4
HM (r ) = M (r ), where M (r ) = ! ! 3 d r '. (2.36)
V rr'

!
Besides, H plays a distinguished role in the study of permanente magnets. The
given equation is similar to eq. 2.1 of the electrostatic case. In the same fashion as
for the derivation of eq. 2.5, one obtains, upon addition of an applied uniform field,
that the total magnetic field is

! ! ! ' 1 !
H(r ) = H 0 + i ! ! d 3r ' M . (2.37)
4 V r r

This equation, the basis of the analysis of permanent and induced polarizations of
ellipsoidal bodies, is valid in all space independently of the source and spatial

dependence of !M(r ). When !M is uniform, case in which no argument is written,
the previous equation becomes

! ! !0 ' 1 !
4
H(r ) = H + i 3
! ! d r ' M . (2.38)
V r r

From eq. 3.3 it follows that
! ! !0 ! ! ! !
H(r ) = H 'n(r )i M( H) r . (2.39)

As previously done for the electric case, it is more convenient to rewrite this
equation in order to explicitly separate the body's internal and external regions:
! ! ! ! !
H int = H 0 'N i M( H int ) for r V ,
! ext ! !0 ext !
! ! int ! (2.40)
H ( r ) = H 'n ( r )i M( H ) for r V .

These equations are valid both for permanent and induced polarizations, but they
are solved differently in each case. Therefore, if an ellipsoidal body is uniformly

64 A detailed reading of pp. 189-194 of Reitz's book is recommended.
 28 Carlos E. Solivrez

magnetized (which requires a nontrivial preparation), the resulting internal

magnetic field is also uniform.
For the treatment of superconductors it is convenient to derive the magnetic
field from the surface density of current, case that bears no resemblance to that of
a surface density of charge (conductors). To that end it is convenient to use the
magnetic induction, derivable from the following vector potential65:
! ! ! !
! ! 0 ' M(r ')(r r ') 3 ! ! ! !
4
AM (r ) = ! ! 3 d r ', B( r ) = AM
(r ), (2.41)
V rr'

The integrand may now be transformed using eqs. A2.2 and A2.8, obtaining
! ! ! ! 1
! ! 0 ' M(r ')(r r ') 3 0 ' ! !
4
AM (r ) = ! ! 3 d r ' = M( r ') ' 3
! ! d r'
V rr' 4 V rr'
! ! ! ! (2.42)
' ' M(r ') ' M(r ')
= 0 ! ! d 3r ' 0 ' ! ! d 3r '.
4 V rr' 4 V rr'

The last volume integral may be transformed into a surface one by the curl
theorem eq. A3.10, where due atention should be paid to the step discontinuity
through the body's surface (see eq. A3.22). It is thus obtained
! ! ! !
! ! 0 ' J M (r ') 3 0 ' K M (r ) 2
4 S r! r! ' d r ',
AM (r ) = ! ! d r ' + 4 "
V rr' (2.43)
! ! ! ! ! ! ! ! !
where J M (r ) = M(r ), K M (r ) = M(r ) s(r ),


where !s(r ) is the unit vector normal to surface S at point !r (see eq. A7.8), J M is
!
the volume density of magnetization current and K M the surface density.
!
The previous equations show how magnetization may originate from currents
localized in the interior and on the surface of the body. The validity of this
interpretation for superconductors is discussed in the relevant section and will be
also used to show that they behave as perfect diamagnetic materials.

Permanent magnetization
The study of spontaneous magnetization and the phenomena of saturation and
hysteresis preceded that of ferroelectric materials, which took their name from
ferromagnets. As discussed in the ferroelectric case, a magnetization is in general
not only a function of the applied magnetic field, the material and its shape, but
also of its own buildup process.

65 Reitz, p. 190 eq. 9-11.

 Depolarization tensor method 29

The resulting equations 2.40 are similar to those obtained in page 18 for
permanent electric polarization. Writing them in matrix representation it is
obtained
!
Hint = H0 'N i M(Hint ) for r V ,
! ! ! (2.44)
Hext (r ) = H0 'next (r )i M(Hint ) for r V .

The scalar and isotropic version of the first equation, with M(H) proportional to
Brillouin's function66, is customarily used in the study of ferromagnetism67. The
applied field, the magnetization and the total field have in general different
directions. The same thing happens in isotropic materials for all but spherical
bodies, due to the geometrical anisotropy introduced by the depolarization tensor
n (see Problem 06).

Induced magnetization
Unlike the electric case, in the magnetic one there are two different types of
induced polarization: diamagnetic and paramagnetic. Diamagnetism is a
manifestion of Lenz' Law, a change in the orbital circulation of electrons that, as
happens with any current, generates a field that opposes the applied one.
Paramagnetism stems from the orientation of the permanent magnetic moments of
atoms without closed electronic shells. Both types are well described in the linear
range by
! ! 1 ! !
M(r ) = m i H(r ), (2.45)
'

where m is the magnetic susceptibility, a symmetric rank two tensor68. As in the

!
electric case, there is no general agreement on the definition of magnetic
susceptibility, and the reader should check the one used by his favourite author,
introducing the necessary additional constants. The definition given by eq. 2.45
simplifies the formulas eqs. 2.49. The magnetic susceptibility is a negative definite
tensor in the diamagnetic case and a ositive definite one in the paramagnetic case,
uniform for all the homogeneous materials studied in this book and a scalar for
policrystalline non-textured and amorphous materials. In the single crystal
anisotropic case its components have relationships determined by the symmetry
of the material's lattice structure69.

66 Dekker, pp. 454-455.
67 Dekker, p. 466.
68 Landau and Lifchitz, p. 58.

69 Nye, p. 23 Table 3. See Appendix 5 in the present book.

 30 Carlos E. Solivrez

For induced magnetizations eq. 2.37 reduces to

! !
! ! !0 1 m i H(r ') 3
4
H(r ) = H + i ! ! d r ' ,
rr' (2.46)
V

system of integro-differential equations that univocally determines the magnetic

field. When !H int is uniform inside the body case in which the spatial argument is
omitted the previous equation reduces to
!
! int ! 0 1 m i H int 3
4
H =H + i ! ! d r ' .
r r' (2.47)
V

From eq. 3.3 it is then obtained:
! ! ! ! 1 !
( ) !
H int = H 0 N i m i H int , H int = 1+ N i m i H 0 for r V ,
! ! ! ! ! !
H ext (r ) = H 0 'n(r )i M for r V ,
! 1 ! 1 !

'
1
'
(
M = m i H int = m i 1+ N i m i H 0 , ) (2.48)
! 1 !
!
( )
1
m = V M = m i H 0 , m = V m i 1+ N i m .
'

!m is the magnetic dipole moment and ! m is the ellipsoidal body's magnetic
polarizability tensor.
In matrix representation the previous equations are written

( ) !
1
Hint = 1+ N i m i H0 for r V ,
! ! !
Hext (r ) = H0 'n(r )i M for r V ,
1
( )
1
where M = m i 1+ N i m i H0 , (2.49)
'
1
( )
1
m = V M = m i H0 , m = V m i 1+ N i m .
'

Conductors
If a conducting body with zero net charge is placed in an applied uniform field
0
!E , a surface density of charge is induced such that, in the equilibrium state, the
internal field vanishes. The cancellation is not instantaneous but the culmination of
a dynamic process in which the body's surface electric charges are redistributed.
Electric forces alone do not suffice, as shown by Earnshaw's Theorem 70: electrons
move in order to be as far away as possible from each other, being stopped only by


70 Stratton, p. 116. The theorem stablishes that no charge may be in stable equilibrium under the
effect of only a macroscopic electric field.

 Depolarization tensor method 31

the atomic forces that prevent them from escaping from the body at normal
temperatures (at high temperatures thermoionic emission Edison Effect takes
place71). The vanishing of the internal electric field valid for conducting bodies of
any shape, not only ellipsoids is, therefore, an additional condition to the laws of
electrostatics, not a consequence of them:

!
E int = 0. (2.50)

The total electric field may be written in terms of the applied one and the
resulting but yet unknownsurface density of charge as72

! ! ! !
(r ') 2
S r! r'! d r',
E(r ) = E 0 k1 " (2.51)


where ! (r ) and !E(r ) are mutually determined. This equation may be written
solely in terms of the electric field by using the step discontinuity through the
body's surface eq. 3.35 (read the section Surface step discontinuity):

1 !+ ! !


!
k1 (r ) =
4
( !
)
E E i s(r ), where E = 0,

(2.52)

and !s(r ) is the unit vector normal to the ellipsoid's surface at point !r (see eq.
A7.8). Therefore
! ! !
! ! !0 1 E + (r ')i d 2r '
E(r ) = E
4 "
! ! , (2.53)
S rr'

a system of integro-differential equations that fully determines the electric field.

The use of the notation !E + stresses the step discontinuity through the body's
surface, crucial for the correct application of the integral theorems of vector
calculus.
! !
Inside the body E(r )= 0, equation that can be satisfied only if the second term

of the first member is a constant. This requirement is satisfied if !E + is uniform,
case in which eq. 3.29 gives
! !
1 E + i d 2r ' ! + !0
4 S r! r! '
" = N i E = E , (2.54)

where N is the body's internal depolarization tensor. This equation may be used to

obtain !E + whenever N has an inverse, which is true for the case of nonvanishing

71 Dekker, p. 220.
72 Reitz, p. 31.
 32 Carlos E. Solivrez

eigenvalues. If one or two of the eigenvalues are zero, a reduced inverse may be
defined which solves the problem, the links being given at page 60.
The external field is then obtained as follows:
! ! ! ! ! !
E int = E 0 N i E + = 0, E + = N 1 i E 0 for r V ,
! ext ! ! 0 ext ! ! + ! (2.55)
E ( r ) = E n ( r )i E for r V .

The electric dipole moment !p of the ellipsoidal conductor is

! ! ! 2
p="S (r ') r 'd r ', (2.56)

where, from eq. 2.52, the surface charge density is

! 0 ! ! 0 ! !
(r ) = s(r )i E + = s(r )i N 1 i E 0 . (2.57)


and !s(r ) is the outgoing unit vector normal to the body's surface at point !r .
Therefore
! 0 ! ! 2 1 ! 0 0 ! 2 ! 1 ! 0
S S
p= " r ' s(r ')d r ' i N i E = " r 'd r ' i N i E . (2.58)

The integral inside the parentesis may be evaluated by components using the
gradient theorem eq. A3.9, as follows:

! !
x
" S

x d 2r = x "


S
x d 2r =

(2.59)

( )
x x d 3r = V x x = V 1.
V

where !1 is the unit dyadic, the one that leaves unchanged any vector or dyadic.
Therefore
! !
p = 0 e i E 0 , where e = VN 1 , (2.60)

being the polarizability tensor of the conducting body.

! c
Although it will not be used in this book (but see problem Problem 31), it should

be noticed that the surface charge distribution ! (r ) of a conducting ellipsoid with
net charge Q may be expressed in the absence of applied field in terms of the
central distance of that point (see p. 172) as follows73:


73 Stratton, p. 209 eq. 12.

 Depolarization tensor method 33

!
! d(r )
(r ) = Q, (2.61)
4 abc

expression that should be compared with the density induced by a uniform field,
eq. 3.32. From there it follows that
!
d(r )d r = 4 abc.
"
2
(2.62)
S

In matrix representation one gets

!
Eint= E0 N i E + = 0, E + = N 1 i E0 for r V ,
! ! !
Eext (r ) = E0 next (r )i E + for r V (2.63)
0
p = c i E0 , where c = V N 1 .

The value of the dipole moment is the same obtained with the polarization model
eq. 6.29.
Except for the vanishing internal field , the equations for a conducting ellipsoid
+
are similar to those of a dielectric, where vector
!E corresponds to the uniform
polarization !P (see eq. 2.10). This suggests that the treatment for both types of
materiales may be unified by the introduction of a suitable equivalente
polarization for conductors, a topic discussed next.

Equivalent polarization
The behaviour of electrons in conductors may be considered as an extreme case
of polarization. In dielectrics the displacement of electrons under the effect of an
applied field is strictly limited by the quantum energies binding them to the nuclei,
which renders it microscopic, a fraction of the atomic or molecular diameter74. In
conductors, as explained by band theory75, electrons can easily migrate from one
atom to another. This may be interpreted as a macroscopic polarization, that of a
material with infinite electric susceptibility, as will be shown below.
As follows from equations 2.10, in the region where a polarization is uniform
the volume density of charge vanishes and its only effect comes from the step
discontinuity through the body's surface. One may, then, define the following

electric potential generated by an equivalent polarization Peq 76:
!

74 Dekker, p. 135.
75 Dekker, chapter 10 pp. 238-274.
76 Solivrez (2008), p. 207.
 34 Carlos E. Solivrez

! 2 ! ! !
! eq
( r ')d r ' Peq
(r ')i d 2r '
S r! r! = k1 "
eq (r ) = k1 " S r! r! . (2.64)


This potential gives rise to the electric field
! !
! ! !0 Peq i d 2r '
S r! r! ' ,
E(r ) = E k1 " (2.65)

which coincides with eq. 2.53 if

! ! !
E + = 4 k1Peq = Peq , (2.66)
0

where eq. A1.4 was used. The case of conductors may now be solved in the same

fashion as that of dielectrics if Peq is given a value such that the internal field
!
vanishes. This value, as shown in Problem 08 at page 127, corresponds to the
polarization induced in a material with infinite electric susceptibility.

Superconductors

If a magnetic induction !B 0 is applied to a superconducting body in an
appropriate range of temperature and field intensity surface currents are

generated such that the resulting internal value of !B vanishes. This Meissner effect
and the cancellation of the material's resistivity are the most notorious
consequences of a complex microscopic phenomena explained by the BCS theory
(initials of Bardeen, Cooper and Schrieffer) and phenomenologically described by
the London model. The latter makes use of currents confined within a finite but
very small distance of the body's surface77. From the macroscopic point of view
these currents may be characterized as zero thickness surface currents. The
depolarization tensor method may then be used to macroscopically characterize
surface currents in superconducting ellipsoidal bodies in a uniform magnetic
!
induction B 78. As in the case of conductors, an additional condition, the Meissner
effect, is required:

! ! ! ! ! !
( !
)
B(r ) = 0 H(r )+ 'M(r ) = 0 for r V . (2.67)

This condition the magnetic analogous of the infinite electric susceptibility of

conductors may be interpreted as a perfect diamagnetism that cancels any

internal magnetic field !H int . Alternately, one may assume that !H and !M vanish

77 Reitz, pp. 325-328.
78 Reitz, p. 319.

 Depolarization tensor method 35

inside the body and the resistivity being zero that the Meissner effect is the
consequence of a perdurable surface conduction current. Both models are used to
analyse the Meissner effect. Although they are mutually incompatible, the coherent
use of any of them fully describes the effect from the electromagnetic point of
view79. The depolarization tensor method will, therefore, be applied to these two
models, showing that both lead to exactly the same results.

Magnetization model
Equations 2.49 describing the behaviour of magnetizable ellipsoidal bodies in a

uniform field field !B 0 may be rewritten as follows:
!
! int ! int ! ! int B 0 !

B = 0 H + 'M , H = ( 0
!
)
'N i M, for r V ,
(2.68)
! ext ! ! 0 ext !
! !
B (r ) = B 0 'n (r )i M for r V .


The value of !M may be obtained from condition 2.67,

! ! ! ! 1 !


( )
B int = 0 H int + 'M = 0, M = H int ,
'
(2.69)

an equivalente magnetic polarization that corresponds to a perfect diamagnetic

susceptibility, as discussed in Problem 15. Replacing this value in the second of
equs. 2.66, it is obtained
!
! int B 0 ! ! ! ! ! !
H =
0
(
'N i M = 'M, B 0 = 0 'N i M 'M = 0 ' N 1 i M. (2.70) ) ( )

Rewriting the last member so that the tensor is positive definite (has an inverse)

! !
1
( )
1
M= 1 N i B0. (2.71)
0 '

In this interpretation the Meissner effect is the consequence of a magnetization

that cancels the internal magnetic field !H int . This magnetization gives origin to a

magnetic moment !m and an external magnetic induction, both functions of the
applied field:

! 1 ! !
! V
( 1
) ( )
1
m = V M = 1 N i B 0 = s i B 0 , s = V 1 N
0 ' 0 ' (2.72)
! ! ! ! ! ! 1 !

!
B ext (r ) = B 0 0 'n(r )i M = B 0 + n(r )i 1 N i B 0 . ( )
where s is the bodys polarizability tensor of the ellipsoidal superconductor.
!

79 Reitz, chapter 15 pp. 318-333.
 36 Carlos E. Solivrez

In matrix notation

( ) 1
( ) !
1
Bint = 0 Hint + 'M = 0, M = 1 N i B0 , for r V
0 '
! ! !
Bext (r ) = B0 0 'next (r )i M for r V , (2.73)
1
( )
1
where m = V M = s i B0 , s = V 1 N .
0 '

The same result may be derived from eq. 2.43, the magnetization surface
current discussed in Problem 17. In Problem 14 our equations are compared with
those given by Reitz (eq. 15-6) for the case of a superconducting sphere.

Surface conduction current model

In this model of superconductivity the Meissner effect is derived from a surface
conduction current. The difference between a conduction current that of electric
circuits from a magnetization current is that the latter is bound to single atoms
or molecules, so that no electrons are transferred between them80. Magnetization
currents are as real as conduction currents, but only the former are feasible in
insulating materials.

The total magnetic induction is the addition of the applied one, !B 0 , and that
!
generated by a surface conduction current K as given by eq. 2.43:
! ! ! ! ! ! ! ! !
B(r ) = B 0 + BK (r ), BK (r ) = A(r ),
! !
! ! 0 ' K M (r ') 2 ! ! ! ! ! (2.74)
A(r ) =
" !
4 S r r ' ! d r ', K M (r ) = M(r ) s(r ).

! !
where K(r ) is a vector field such that Meissners effect takes place. Its value can be
!
found by noting that any surface current K gives rise to a magnetic's induction step
discontinuity through the body's surface81,
! ! ! ! ! ! ! ! ! ! !


!
( )
s(r ) B + (r ) B (r ) = s(r ) B + (r ) = 0 'k3K(r ) as B (r ) = 0, (2.75)

where !s(r ) is the outgoing unit vector normal to the ellipsoid's surface S at point
! !
!r (see eq. A7.8) and the Meissner condition B (r ) = 0 has been used. Thus, taking
into account eqs. A1.11 and A7.8, the appropiate surface current is given by

80 See, for instance, Reitz, p. 188 Figure 9-2.
81 Stratton, p. 246 eq. 7.

 Depolarization tensor method 37

x y z
x + 2 y + 2 z
! ! k3 ! ! ! ! a 2
b c
K(r ) = s(r ) B + (r ), where s(r ) = . (2.76)
4 k2 x 2 y2 z2
+ +
a4 b 4 c 4

Replacement of this value in the expression of the total induction gives

! ! !
! ! !0 1 B + (r ') d 2r '
B(r ) = B
4 S r! r! ' ,
" (2.77)

the integro-differential equation that determines the magnetic induction for
!
ellipsoidal superconductor. Using as an Ansatz a uniform value B +, one obtains for
the second term
! ! ! !
! ! 1 B + d 2r ' 1 d 2r '
S r! r! ' = 4 B " S r! r! '
+
B K (r ) = "
4
(2.78)
1 ! 1 1 1 !
= B + ' ! ! d 3r ' = ! ! d 3r ' B + ,
4 V rr' 4 V r r '

where the curl theorem eq. A3.9 has been used. The last term which for the sake

of concision is written in terms of ! f (r ) (eqs. 3.8 and 3.4) may be transformed
according to the identity eq. A2.9:
! ! ! !
(
BK (r ) = f (r ) B + )
! ! ! !
( ) !
( ! !
) ( ) !
(
= B + i f (r ) f (r )i B + + f (r ) i B + B + i f (r ) )
(2.79)
! ! !

( ) !
= B + i f (r ) B + f (r ),

The expansion follows from the uniform character of !B + and the property eq. A2.3
of the Laplacian. Using eqs. 3.3 and A3.7 the last term may be rewritten as

! ! 1 !+ 1 !+
B K (r ) =
4
( ) d 3r '
B i ! !
V rr'
+
4
d 3r '
B ! !
V rr'
!+ !+ ! (2.80)
N i B B for r V
= ext ! ! + ! .
n (r )i B for r V

Therefore
! !

! ! ! 0 ! ! B 0 1 N i B + for r V
B(r ) = B + BK (r ) = !
(!+
) !
. (2.81)
ext ! !

B 0
+ n ( r )i B for r V

where !B + is fully determined by the Meissner condition
 38 Carlos E. Solivrez

! ! ! !
( ) ( )
1
B 0 1 N i B + = 0, B + = 1 N i B0. (2.82)

As shown in Problem 16 of page 132, the magnetic moment generated by this

density !K of surface conduction current is the same as that obtained from the
magnetizacion model eq. 2.73,

!
1 1
( )
1
m= V B+ = i B0 , s = V 1 N .
0 0 ' s
(2.83)

In matrix representation,

( )
1
Bint = Hint = M = 0, B+ = 1 N i B0
! !
Bext (r ) = B0 + n(r )i B+ ,
1
s i B0 , s = V (1 N ) ,
1
m=
0 '
(2.84)
x y z
! ! k3 ! x + 2 y + 2 z
! ! a2
b c
K(r ) = s(r ) B + , where s(r ) = .
4 k2 x 2 y2 z2
4+ 4+ 4
a b c

Apart from the difference in electrons localization, the only difference between
the magnetization and the surface current models of superconductivity is the

asignement of the internal values of !H and !M.

Summary of integro-differential equations

Material Equation Eq.

! !
! ! !0 1 e i E(r ') 3
4
dielectrics E(r ) = E + i ! ! d r ' . 2.31
V rr'

! !
diamagnets ! ! !0 1 m i H(r ') 3
4

H(r ) = H + i ! ! d r ' , 2.46
paramagnets V rr'

! ! !
!0 1 E + (r ')i d 2r ' !
conductors E
4 S
" !
rr'
! = 0 for r V . 2.53

! ! !
! ! !0 1 B + (r ') d 2r '
superconductors B(r ) = B
4
" S r! r! ' , 2.77

Table 1. Integro-differential equations derived in this chapter.

 Depolarization tensor method 39

Summary of induced polarization equations

Edep = N i e i Eint ,

( )
1
Eint = 1+ N i e E0 for r V ,
0 0
( )
1
P= e i Eint = e i 1+ N i e E0 ,

Dielectrics: 0 (2.33)
( )
1
p = V P = e i E0 , e = V e i 1+ N i e ,

! !
Eext (r ) = E0 next (r )i P for r V .
0


H = N i m i H
dep int

( !
)
1
Hint = 1+ N i m i H0 for r V ,
1
( )
1
M=
m i 1+ N i m i H0 ,
Magnetic materials: ' (2.49)
1
( )
1
m = V M = m i H0 , m = V m i 1+ N i m ,
'
ext ! ext ! !
H (r ) = H 'n (r )i M for r V .
0

+
E dep
= N i E
!
Eint = E0 N i E + = 0, E + = N 1 i E0 for r V ,
0
Conductors: p = c i E0 , c = V N 1 , (2.63)

ext ! ext ! + !
E (r ) = E n (r )i E for r V .
0

Superconductors (magnetization model):

( ) 1
( )
1
Bint = 0 Hint + 'M = 0, M = 1 N i B0 for V ,
0 '
1
( )
1
m = V M = s i B0 , s = V 1-N , (2.73)
0 '
ext ! ! !
B (r ) = B 0 'next (r )i M for r V .
0

 40 Carlos E. Solivrez

Superconductors (surface current model):

Bdep = (1 N)i B+

( )
1
Bint = Hint = 0 = M, B+ = 1 N i B0 ,
1
( )
1
m = s i B0 , s = V 1-N , (2.84)
0 '
! !
( )
1
Bext (r ) = B0 + next (r )i 1 N i B0 .

General formulation of the method

Along the sections of this chapter a study was made of the electromagnetic
behaviour of solid and homogeneous ellipsoidal bodies either isotropic or
anisotropic under an applied static and uniform electric or magnetic field. In
what follows a general formulation of the treatment is given in matrix
representation, showing that the ensuing integro-differential equations are
essentially identical in all cases and may all be solved, for the linear case, in exactly
the same way. The main discussion is devoted to induced polarizations because
permanent ones require specific analysis.

In all cases a vectorial field F electric field !E , magnetic field !H, or magnetic

induction !B is derived from a static applied one F0 through a vectorial
magnitude Q(F) electric polarization, magnetization or step discontinuity
through the body's surface which is a linear function of F. Q(F) fully describes
the state of uniform polarization of the body, which may be alternately
characterized by some surface density of charge or current. The linear relationship
between Q and F is either described by a susceptibility case of the linear but
perphaps anisotropic relationship Q= F or by a condition of cancellation of the
internal field. Volume integrals are involved in most cases, but conductors and
superconductors initially require surface integrals that may be transformed into
volume ones by definining appropriate equivalent polarizations. The ensuing
integro-differential equations, valid over all space, are separately solved for the
region occupied by the body and the external one. It is then obtained

!
Fint = F0 N i Q(Fint ) for r V ,
! ! ! (2.85)
Fext (r ) = F0 next (r )i Q(Fint ) for r V .

The depolarization tensor n, eq. 3.4, is uniform inside an ellipsoidal body (where
its position independent value is N). !Fint is therefore also uniform there, its value
being determined by the first equation, as well as that of !Q(Fint ). The value of the
external field then follows from the second of eqs. 2.85. The geometric
contribution is determined by N and next, while the material's properties are
characterized by Q.

 Depolarization tensor method 41

The explicit values of !Fint and Q are obtained in different ways according to the
material's properties. The first one is the use of the linear relationship

Q(F ) = i F ,
int int
(2.86)

where is a susceptibility. In this case

( ) ( )
1 1

Fint = 1+ N i i F0 , Q = i Fint = i 1+ N i i F0 . (2.87)

The second case applies when the internal field vanishes (see Table 1):

F N i Q = 0 or (1 N)i Q = F .
0 0
(2.88)

These relationships correspond to conductors and superconductors, where the

susceptibility obtains its largest value
1 1
Q = N i F or Q = (1 N) i F .
0 0
(2.89)

The surface conduction current model of superconductors is the only case for
which no single Q tensor may be defined for both the internal and external field.
The origin of this asymmetry is the difference between the constitutive equations
for the electric and the magnetic case, eqs. 2.11 and 2.35.
In the following table magnitudes F and Q are identified for all cases discussed,
as well as the linear functions and cancellations required for the validity of the
treatment. Although they do not satisfy eqs. 2.87 to 2.89, permanent polarizations
have been included in order to give a full panoramic view of the method.

Material F Q Condition Eq.

P
Ferroelectric E P = P(E ) 2.12
int
0

0
Dielectric E P P = e i Eint 2.33
0 e

+ 1
E = N i E
0
Conductor (surface charge) E E+ 2.63
Conductor (equivalent P) E 4 k1P = P = eEint 2.65
! !
!M = M(H )
int
Ferromagnetic H 'M 2.44
Diamagnetic and paramagnetic H 'M M = m i Hint 2.49
! m !
Superconductor (equivalent M) H 'M = 1 M = mHint 2.73
! m !
Superconductor (surface Qint = (N 1 1)i B+
B +
Bint=0 2.84
conduction current) ! Q ext
= B

Table 2. Identification of the magnitudes

in the general field equations for different kinds of materials.
 42 Carlos E. Solivrez

Solving the integro-differential equations by iteration

The dielectric case, eq. 2.31, may be taken as a prototype of the integro-
differential equations which determine the fields:
! ! !0 !
E( r ) = E + O( E), (2.90)

where O is a linear operator with the following properties (see eq. 3.3):
! !
! ! e i F(r ') 3
( )
O F(r ) = k1 i ! ! d r '
rr'
(2.92)
V
! ! ! !
( ) (
O k F(r ) = k O F(r ) , )
! ! ! ! ! ! ! !
( ) ( ) (
O F(r )+ G(r ) = O F(r ) + O G(r ) , ) (2.92)
! !

( ) !
O F 0 = 4 N i e i F 0 if r V .

The second member of the last equation holds true only when !F 0 is a uniform
vector throughout the body's volume V.

The solution for each of these equations was previously found by using a
uniform Ansatz, as discussed in the different sections. The resulting field coincides,
in all known cases, with that obtained by solving Laplace's equation in ellipsoidal
coordinates by the method of separation of variables82. The examples are only
illustrative and one should prove that eq. 2.90 has a unique solution, as sketched
below.
! !
Let F (1) and F (2) be two different solutions of eq. 2.90, that is
! (1) ! 0 ! (1) ! (2) ! 0 ! (2) ! (1) ! (2) ! !
F = F + O( F ), F = F + O( F ), that is F F = G = O(G). (2.93)

By iteration it is easily proved that

G
! = O n
(G) for n = 1,2,3..., (2.94)

equation that is satisfied only for !G = 0, proving that the solution is unique.
The inhomogeneous eq. 2.90 may be solved by iteration, succesively replacing
the argument of the operator by the full second member. The first replacement
gives

( ( ))
! ! ! !
E = E0 + O E0 + O E
!0 !0 ! !0 ! (2.95)
( ) ( ) () ( ) ()
1
0
= O E + O E + O E = O E + O E ,
1 2 2

=0

82 Full references to solutions are given in the section History of Chapter 1.

 Depolarization tensor method 43

where On is the nth power of O.

The second iteration gives

! 1 ! 1 ! !
=0
( )
E = O E 0 + O 2 O E 0 + O 2 E
=0
( ) ()
(2.96)
!0 !0 ! !0 !
( ) ( ) ( )
1 1 3
= O E + O
+2
E + O E = O E + O E .
4 4
( ) ( )
=0 =0 =0

In the infinite iterations limit one gets

! ! !
=0
( )
E = O E 0 + O E , ( )
(2.97)
! !
( ) ( )

O E 0 = 4k1N i e i E0 ,

where the operator series coincides with the Taylor's expansion

1
= 1 x + x 2 x 3 + x 4 + ... = (x) . (2.98)
1+ x =0

If the series converges and the last term of eq. 2.97 tends to 0, it is obtained
! ! !
( )
1
E(r ) = 1+ N i i E 0. (2.99)

The expression concides with eqs. 2.33 and 2.87, showing that the iterative
expansion converges and coincides with that obtained using a uniform Ansatz.
Although theoretical physicists frequently use series expansions of operators, the
method is seldom used by engineers. It is easier to give a rigorous justification of
this kind of expansion by using matrices83.
The same method may be used to solve the other three integro-differential
equations in Table 1. This iterative method, clumsier than the use of a uniform
Ansatz, is presented here only to stress the mathematical consistency of the use of
integro-differential equations. It might, perphaps, be used as an approximate
method for studying the case of non ellipsoidal bodies or equations 2.85 when Q is
not a linear function of F. The exploration of this subject, not made by the the
author, is left to the interested reader.

83 See, for instance: G. Goertzel and N. Tralli; Some Mathematical Methods of Physics; McGraw-Hill;
1960; chapters 1, 2 and 3, pp. 7-49.
 44 Carlos E. Solivrez

 Depolarization tensor method 45

Chapter 3:
The depolarization tensor: basic treatment

Definition
!
The depolarization tensor n(r ) arises from expressions like
! 2!
! ! ! ! ! ! F id r '
( I(r )i F ) , ( F i )( I(r )) , ( i I(r )F ) and "
S r! r'! ,
(3.1)
! d 3r ' !
where I(r ) = ! ! and F is uniform in V .
r r'
V

The integration is made over the volume V of an ellipsoidal body or its closed
!
surface S, and r is the position of a field point that may be either inside or outside

the body. !F is a uniform vector inside the region of integration the bodys
volume, including its interior surface that may be an electric or magnetic field or
a polarization. Table 1 gives the list of the integro-differential equations discussed
in this book where such expressions arise.
Only the first two expressions will be analyzed next as the third is very similar.
The case where n is derived from a surface integral is discussed at section n as a

surface integral. The proof is easier to follow if one writes !F and in terms of
components and unit vectors, as follows:

! (3.2)
= x , F = x F , where x i x = .
x
Then
!
! ! I(r )
( I(r )i F ) = x x i x F
x
x

! !
I(r ) I(r )
= x

(
x i
x ) F =
x
x , x x , x
F (3.3)

! !
I(r ) 2 I(r )
= x F =
x F.
x x , x x

 46 Carlos E. Solivrez

One may now define the components of the rank two depolarization tensor

!
! 1 2 I(r ) 1 2 d 3r '
n (r ) =
4 x x
=
4 x x
V
! ! ,
r r ' (3.4)

that in dyadic representation (see Appendix 6) may be written

! !
n(r ) = x n (r )x . (3.5)
,

For instance
! !
( F i )I(r! ) = F
! I(r ) 2 I(r )

x = F
x x , x x
x
! (3.6)
! ! !
= 4 F i x n(r ) x = 4 n(r )i F ,
,

as the depolarization tensor is symmetric (see eq. 3.12).

Therefore,

! !
! ! ! ! ! F i d 2r '
( )!
( ) (
F i I(r ) = i I(r )F = I(r )i F = " )
S r! r! '
! ! ! (3.7)
= 4 n(r )i F , where n(r ) = x n x .
,

The calculation of the depolarization tensor n is therefore reduced to the

!
calculation of the integral I(r ) eq. 3.3 for diferent kinds of ellipsoidal bodies. For a
solid ellipsoidal body this integral is, apart from a constant factor, either the
potential for a uniformly charged one, or the gravitational potential for a constant
mass density. In this chapter the first potential is used, as it is more familiar to
students of engineering, and the second potential the less familiar one with
more complex derivation is used in the following chapter to derive expressions
for n.
!
It is convenient to define an auxiliary function f (r ) such that eq. 3.4 reduces to

! ! !
! 2 f (r ) ! 1 dr ' (r )
4
n (r ) = , where f (r ) = ! ! = . (3.8)
x x V r r ' 4 k

In the last term is a gravitational or electrostatic potential, the constant mass

or charge density and k the gravitational or electrostatic constant determined by
the choice of system of units. That is, they are the factors such that

 Depolarization tensor method 47

1 ! d 3r '
(r ) = ! ! (3.9)
k V rr'

!
Tensor n(r ) is uniform inside ellipsoidal bodies, where it was originally
defined84. The author of this book was apparently the first one to define it outside
these bodies although restricted to the magnetic case85. In the literature of
magnetism its internal value is sometimes called demagnetization tensor (apart
from a possible denominator 4), its principal values being the well known
demagnetizing or demagnetization coefficients or factors86. The tensor is
indistinctly written here as a dyadic n or a 3x3 matrix n.
For applications it is necessary to explicitly identify the regions inside and
outside the ellipsoidal body. To that end N (dyadic) o N (matrix) denote the

uniform internal depolarization tensor and next(!r ) o next(!r ) the non-uniform
external one
! ! ! !
N = x x N for r V , next (r ) = x x n (r ) for r V . (3.10)


N N xy N xz
xx !
N = N xy N yy N yz for r V ,

N xz N yz N zz
(3.11)
n (r! ) ! !
nxy (r ) nxz (r )
xx ! ! ! !
ext !
n (r ) = nxy (r ) n yy (r ) n yz (r ) for r V .
! ! !
nxz (r ) n yz (r ) nzz (r )

The difference between both regions is emphasized by explicitly writing the
dependance on position for the external case and omitting it in the internal case, as
for constants or parameters.

General properties of n
Symmetric tensor
As the order of derivation may be freely exchanged87, n turns out to be a
symmetric tensor:
! !
n (r ) = n (r ), (3.12)

84 Landau and Lifchitz, p. 44. Maxwell, apparently the first to introduce it, dit not identify its
tensorial character.
85 Solivrez (1981).
86 Moskowitz and Della Torre, p. 739.
87 MacMillan, pp. 27-32.
 48 Carlos E. Solivrez

property that was previously used in eq. 3.6.

Trace
From eq. 3.4 the trace of n (sum of its diagonal elements) is
!
! ! 2 f (r ) ! 1 1
Tr n(r ) = n (r ) = = f (r ) = i ! ! d 3r ', (3.13)
x 2
4 V rr'

where is the Laplacian operator eq. A2.3. The value of this operation, frequently
used in the theory of electromagnetism, is an integrable singularity inside V and
vanishes outside the body. As the calculation is often made in a questionable way,
in Appendix 3 (see eq. A3.7) it is proved that88.
!
d 3r ' 4 for r V
! ! = ! . (3.14)
V r r ' 0 for r V

It follows that

Tr N = 1, Tr next = 0. (3.15)

In fact, the constants in eq. 3.8 were chosen in order to make the trace of N to be 1.

Orthogonal transformations
The transformation properties of matrix n are now analyzed under an
orthogonal change (rotations, reflections, inversions and all its combinations) of
coordinate system:
x '
R Rxy Rxz x
xx
x ' = R x , r' = y' = R yx R yy R yz i y = R i r. (3.16)

z'
Rzx Rzy Rzz z

These transformations are the basis of the use of symmetry properties in order to
simplify the calculation of many physical properties (see the section Symmetries,
and Problem 19).
Such operations are always represented by an orthogonal matrix R, that is, one
such that its inverse coincides with its transposed matrix:

R 1 = R t , R i R t = R t i R = 1, where R
t
= R . (3.17)

88 V. Hnizdo, Eur. J. Phys. vol. 21, pp. L1-L3 (2000) gives a rigurous but more complex proof.

 Depolarization tensor method 49

The inverse transformation is then

x
R R yx Rzx x '
xx
1 t
r = R i r' = R i r, y = Rxy R yy Rzy i y' , x = R x ' . (3.18)

z
Rxz R yz Rzz z'

In order to determine the components of n in the new coordinate system, one
should use definition eq. 3.8,
!
! 2 f (r )
n (r ) = , (3.19)
x x

and express the value of the new partial derivatives in terms of the original ones89.
In terms of the derivative operator:

x
= = R . (3.20)
x' x x' x

It is thus obtained for the new matrix N

N ' =
2 f
= R
x' x' ,
2 f
R = R i N i Rt
x x
( )
or N' = R i N i R t . (3.21)

where eq. 3.17 was used.
By using this transformation property one may stablish relationships among
components of the polarization tensor, topic discussed in the following section.
If one determines the eigenvalues of n in its principal coordinate system, the
value of n' in a coordinate system related to the former by eq. 3.21 is

n' = R i ni R t
R Rxy Rxz n (r! ) 0 0 R R yx Rzx
xx x !
xx
(3.22)
= R yx R yy R yz i 0 n y (r ) 0 i Rxy R yy Rzy .
!
Rzx Rzy Rzz 0 0 nz (r ) Rxz R yz Rzz


Symmetries
The relationships among the constant eigenvalues of N determined by the
body's symetries quite useful for applicationscan be found by using the
transformation properties eq. 3.22. The symmetries of the external depolarization
tensor next are more difficult to treat becauses its components usually are
cumbersome functions of position. This case is usually handled by using


89 Santal, pp. 299-303.
 50 Carlos E. Solivrez

appropiate curvilinear coordinate systems (see, for instance, the treatment of the
sphere in page 58).
When an orthogonal transformation R is a symmetry of the ellipsoidal body, the
value N' in the transformed coordinate system must be equal to the original one:

N' = R N R = N, or R N = N R,
t
(3.23)

From a practical point of view it is simpler to use the conmutation relationship

than the first one, as is illustrated next.
The case of spheroids (ellipsoids of revolution) is discussed now as an example.
If the axis of revolution is taken as coordinate axis z, the exchange of axes x and y (a
reflection plane) should leave N unchanged. Therefore, for

0 Ny 0 0 Nx 0
0 1 0

R= 1 0 0 ,R iN = N 0 0 = NiR = Ny 0 0 . (3.24)

x

0 0 1 0 0 N z 0 0 N z

It thus turns out that Nx = Ny and no further relationships are found using other
symmetry operations. The diagonal expression of N for spheroids is thus

N 0 0
e
N spheroid = 0 Ne 0 (3.25)

0 0 N p

Np being the polar eigenvalue corresponding to the spheroid's symmetry axis, an
Ne that of the two equal equatorial ones.
In the case of the sphere an exchange of axis y and z is also a symmetry,
which leads to the three eigenvalues being equal, as discussed in Problem 19. In
combination with the trace rule eqs. 3.15, one then obtains

Nx = Ny = Nz = N, Tr N = 3N = 1 and N = 1/3. (3.26)

The sphere is the only type of ellipsoid where the actual eigenvalues of N may
be determined by using only the trace rule and symmetry arguments. There are
only two other cases where this can be done without finding primitives of
integrals, but both require the knowledge of the integral expressions of N, topic
discussed in some sections of this chapter and the next.

n as a surface integral
The depolarization tensor n may be expressed as a surface integral in cases
involving surface densities of charge or current. To that end one should take the
!
scalar product of eq. 3.4 with a uniform vector !C :

 Depolarization tensor method 51

! ! ! ! !
n(r )i C = x n(r ) x i C = x n(r )C
, ,

1 ! (3.27)
1 d 3r ' 1
=
4 , x x
C
x ! ! =
4
' ! ! d 3
r ' iC .

V rr' V r r '


As !I(r ) (eq. 3.8) is a continuous function, the gradient theorem eq. A3.9 may be
used to obtain
1 !
d 2r '
S r! r! ' .
' ! ! d r ' = "
3
(3.28)
V rr'


As !C is an arbitrary vector, it follows that

!
! 1 d2 r '
n(r ) = ! ! ,
4 " S r r '
(3.29)

an expression previously given by Moskowitz and Della Torre90.

Surface step discontinuity

The depolarization tensor has a step discontinuity through the bodys surface,
whose characteristics are better understood by analizing a concrete case, here
taken to be the electric one. As discussed next, it turns out that in this case the
discontinuity is equivalent to a charge density. In the magnetic case where the
analysis is quite similar it corresponds to a surface current density. In Problem
30 the reader is challenged to find the expression for the surface discontinuity
without using the properties of electric or magnetic fields.

Surface charge density

For a uniform electric polarization !P whether it is spontaneous, induced or
equivalent the resulting electric field may always be written as (see eq. 2.9)

d 3r '
E P (r ) = P (r ) where P (r ) = k1 ' i P. (3.30)
V r r '
!
As I(r ) (eq. 3.8) is a continuous function, the gradient theorem eq. A3.9 may be

!
used to reduce the volume integral to the following surface integral over S:

90 Moskowitz and Della Torre, p. 740 eq. 17.
 52 Carlos E. Solivrez

d 3r ' d 2r '
P (r ) = k1 ' i P = k1
S r r ' i P
V r r '
(3.31)
P (r ')d 2r '
= k1 S r r ' , where P (r ') = s(r )i P.
!
The last integral expresses the potential created by a surface charge density P91.
Using eq. A7.8 it is obtained
xPx zPyPy
++ 2z
2 2
P (r ) = s(r )i P = a b c ,
(3.32)
2
x y z2
2

4
+ 4+ 4
! a b c

which is the electric charge density created by the step discontinuity on the bodys
surface, including the equivalent polarization case in the case of conductors. This
expression, useful for elllipsoidal conductors, is not given in any of the textbooks
used for writing of this book92. The validity of eq. 3.32 may be easily verified for
the special case of the sphere discussed in Problem 09.

Surface step discontinuity

Through a material interphase with a surface charge density of net (case not
considered in this book) or induced charge the component of an electric field
!
parallel to the surface is continuous, but the normal component is not. If s(rS ) is

!
the unit vector normal to the surface at point rS , the decomposition of any vector

in components parallel and normal to that surface is:
! ! ! ! ! ! ! !

( ) (
F = F" + F , F" = s i F s = ss i F , F = 1 ss i F ,

) (3.33)

where 1 is the unit dyadic of eq. A6.4.

The continuity of the electric field component tangential to the bodys surface is
then expressed in the following way93:
!+ ! ! !

(1 ss ) i ( E

) ( )(
E = E + E i 1 ss = 0. ) (3.34)

The discontinuity of the component normal to the body's surface S is94:

91 Reitz, p. 31 eq. 2-15.
92 Stratton the autor that discusses in more detail the electromagnetic properties of ellipsoidal
bodiesgives only the surface density for charged conductors in the absence of an applied
field, see eq. 12 in p. 209.
93 Reitz, p. 90.
94 Stratton, p. 188.

 Depolarization tensor method 53

!+ !

(E

)
E i s = 4 k1 .
(3.35)

The signs are chosen so that a component is positive when the vector points
outgoing from the positive side of the surface (+), chosen to be the one external to
the body.
For all the cases discussed in the previous chapter see eqs. 2.7 and A1.4 the
following relationship is fullfilled,


!
(
E + E = 4 k1 N next (r S ) i P, ) (3.36)

where !r S is a point on surface S.
From this identity and the expression of !s eq. 3.31, the field 's discontinuity
through the bodys surface, the depolarization tensor's step discontinuity, may be
written as


!
(
s i N next (r S ) i P = s i P. ) (3.37)

!
For the equation to be valid for arbitrary values of P whether it is a
spontaneous, induced or equivalent polarization it should be

S S ( ) ( ext ! S ) ! S ! S
! ! ext ! S
s(r )s(r )i N n (r ) = N n (r ) i s(r )s(r ). (3.38)

Any tensor T may be decomposed in the following fashion

( ) (
T = 1+ ss ss iT = ss iT + 1 ss T = T! +T ,

) (3.39)

in components such that when the scalar product with an arbitrary vector is taken,
the components parallel and normal to the surface are respectively given by each
term.

From eqs. 3.34 and 3.35 one gets

ext ! S
s i ( N next (r S )) + ( 1 ss ) i ( N next (r S )) = ss ,
! !
N n ( r ) = s (3.40)

which gives the value on the bodys surface of the external depolarization. The
following matrix expression is thus obtained, valid for all types of ellipsoids:
ext ! S ! !
S S
n (r ) = N s(r )i s(r ). (3.41)
From eq. A7.8,
! ! x S x S
s (r S )s (r S ) = . (3.42)
( x S )2 ( y S ) ( z S )2
2

+ +
a4 b4 c4

At an elementary level the expression may be verified for all ellipsoids where
the components of the external depolarization tensor may be expressed in terms of
 54 Carlos E. Solivrez

elementary functions, like the infinite cylinder and the sphere (see Problem 34).
Later on, it will be verified for all types of ellipsoids, including the triaxial one (see
eq. 4.66). The formula provides a simple approximation to the fields near the
surface of any ellipsoid. This includes the electric fields at sharp metallic tips, a
topic where explicit formulas are rarely given (see Problem 12).

Calculation of n using electrostatic Gausss Law

This section is devoted to the calculation of n from eq. 3.8, which requires the
!
expression of the auxiliary integral f (r ) . As stated there, apart from a
proportionality constant f coincides with the electric potential generated by
volume V when charged with uniform density 95:
!
! d 3r ' ! ! (r )
(r ) = k1 ! ! = 4 k1 f (r ), f (r ) = . (3.43)
V rr'
4 k1

This expression is useful for the shapes where the potential may be derived by
using electrostatic Gauss law, high symmetry cases which are frequently discussed
in first university courses on electricity and magnetism for physicist and
engineers. In order of increasing difficulty these shapes are:
a = b c: constant thickness sheet of infinite extension;
a b = c: right circular cylinder of infinite length;
a = b = c: sphere.

Sheet of constant thickness and infinite extension

The depolarization tensor of a sheet of constant
thickness e = 2c and infinite extension, as shown in
the figure, is calculated here. A cartesian coordinate
system is used such that its z axis is perpendicular to
the surface of the sheet, with origin on its middle
plane.
The simplest way to solve the equivalent
Figure 3. Infinite sheet.
electrostatic problem of eq. 3.43 is to evaluate the
electric field by application of electrostatic Gauss's Law, obtaining then the
potential by integration. This may be easily done due to the high symmetry of the
problem. As the sheet is invariant under arbitrary translations along the axes x, y,
the field may only be a function of z. As it is also invariant under arbitrary
rotations around the z axis, the field cannot have x and y components. The last
symmetry of interest is the reflection plane parallel to the sheet's surfaces through
its middle, making the field in one of its sides the specular image of that in the
other. Therefore
! ! ! !
E(z) = E(z), E( E(0) = 0.
r ) = E(z)z, (3.44)



95 A similar expression applies for f in the gravitational case discussed by Macmillan and Kellog,
where one should take k1 = 1 and to be the density of mass.

 Depolarization tensor method 55

Figure 4 shows the sheet's cross section in gray

and the two closed surfaces, S1 and S2, used for the
application of Gausss Law. These gaussian boxes are
right circular cylinders such that their vertical faces
are parallel to axis z and the horizontal ones, of area
A, are normal to it. The flux of the electric field
!
vanishes both on the base (where E = 0) and on the
Figure 4. Gaussian boxes
vertical face (where it is normal to the surface). This used for solving the infinite
reduces the calculation to the flux on the upper face, sheet.
where
! int "
i dS = 4 k1Q1 , gives E (z) A = 4 k1 Az, E (z) = 4 k1 z,
#
int int
E
S1
! ext " (3.45)

# = = = e.
ext ext
E i dS 4 k Q
1 2
, gives E (z) A 4 k1
Ae, E (z) 4 k1
S2
d(z)
From E(z) = and eq. 3.43 it is obtained
! dz
1 2 !
! z for r V ,
cte. 2 k1 z 2 for r V
(z) = ! , f (z) = 2 ,. (3.46)
cte. 2 k1
e z for r V 1 !
e z for r V
2

From the definition eq. 3.4 it follows that

0 0 0 0 0 0
for the sheet N = 0 0 0 , n = 0 0 0
ext .
(3.47)

0 0 1 0 0 0

Due to the inexistence of N-1 one should next analyze the relationship of the
fields thus obtained with those of the depolarization tensor method. Equations
2.33 and 2.49 show that the relevant matrix for the case of induced electric and
magnetic polarizations is
1
1 0 0 0 0 0
( )
1
1+ N == 0 1 0 + 0 0 0

0 0 1 0 0
(3.48)
1
1 0 0 1 0 0

= 0 1 0 = 0 1 0 ,


0 0 1+ 0 0 1/ (1+ )

where it is clear no problem arises. This does not happen for the case of
conductors and superconductors, eqs. 2.63, 2.73 and 2.84. In these cases one
 56 Carlos E. Solivrez

should go back to the original equation that was solved by the inversion of a
matrix. For instance, the first of eqs. 2.63 leads to the equation

N i E + = E0 , that is E+x = 0, E+y = 0, E+z = E0z , (3.49)

where E0 is the aplied electric field and E+ is the total electric field on the external
surface of the sheet due to the surface density of charge96 induced by E0. The
physical meaning of the previous equation is that, as regards the polarization
effects those that have origin in the redistribution of surface charges the only
component of the applied field to be taken into account is the one normal to the
sheet, the z component. In the reduced space of the z axis, matrix N has what may
be called a reduced inverse (this is better seen in the two-dimensional case
discussed at the end of next section).

Right circular cylinder of infinite length

As in the previous problem, the value of n will be
derived from the calculation of the electric field of a
uniformly charged right circular cylinder of radius b,
infinite length and uniform electric charge density . By
symmetry, the electric field lies on the plane normal to
the cylinder axis, is angle-independent and radial, that is,
normal to the cylinders surface. Its value, easily obtained Figure 5. Infinitely
using the electrostatic Gausss Law is97: long circular cylinder.

2 k1 r for r b
(r)
Er = = a2 , r= y2 + z2 . (3.50)
r 2 k for r b
1
r

Integrating the field with respect to r it is obtained

constant k1 r 2 for r b
(r) = , (3.51)

constant 2 k1 b2 lnr for r b

1 2 2 !
( y + z ) for r V
! 4
f (r ) = . (3.52)
1 2 2 2 !
b ln( y + z ) for r V
4
From eq. 3.4 it follows that

96 The reader should read in full detail the section Conductors in this book.
97 A. Halpern, 3000 Solved Problems in Physics, Schaum's Outline Series, McGraw-Hill, p. 406
problem 25.70; 1988.

 Depolarization tensor method 57

2 f 2 f 1 2 f 1 2 f
= 0, = , = , = 0 for x x , r V , (3.53)
x 2 y 2 2 z 2 2 x x

0 0 0
0 1/2 0
N= , (3.54)
0 0 1/2

which satisfies the trace rule eq. 3.15. This value of N coincides, apart from the
exchange of axis, with that obtained by symmetry arguments (see eqs. 3.25 and
3.73).
Outside the body
2 f 2 f
= = 0 if x = y,z,
x x x x
2 f b2 y 2 z 2 2 f b2 y 2 z 2
= , = , (3.55)
y2 2 r4 z 2 2 r 4
2 f 2 f b2 2 yz
= = .
y z z y 2 r4

In a spherical coordinate system where

x = r cos sin, y = r sin sin, z = r cos, (3.56)

= x + y = r sin ,
2 2

and R is the cylinderss radius, the matrix expressions for n are

1/2 0 0

Cylinder N = 0 1/2 0 , (3.57)

0 0 0

cos2 sin2 0
! R 2
n (r ) = 2 sin2
ext
cos2 0 , (3.58)
2
0
0 0

As there are no components along the cylinders z axis, the problem may be
solved in the reduced space of the x, y components where the reduced inverse N is

N' = 12 1 0 , N'1 = 2 1 0 , (3.59)
0 1 0 1

 58 Carlos E. Solivrez

using which the general matrix equations may be applied for all electrostatic and
magnetostatic cases.

Sphere
The sphere is the last body for which the
electrostatic potential may be obtained from the
electrostatic Gauss's Law. Let R be the radius of the
sphere, Q the total uniformly distributed charge and
take its center as the origin of coordinates. From the
symmetry of the body the electric field has to be radial,
! ! ! Figura 6. Sphere.
!E(r ) = E(r) r , the following function of the distance r to
the center of the sphere.
k1Q
3 r for r R,
E(r) = R (3.60)
k Q
1 for r > R.
r 2

As
! ! d r
E(r ) = E(r)r = (r) = (r), then (r) = E(r ')dr '. (3.61)
dr

Therefore
Q !
cte. k1 3 r 2 for r V
! 2R
(r ) = , (3.62)
cte. + k Q !
for r V
1
r
As
4 3 Q 3Q 1 R3
V= R , = = , f (r) = (r) = (r), (3.63)
3 V 4 R3 4 k1 3k1Q
!
it follows that the auxiliary function f , eq. 3.8, is given by

1 2
r for r R
! 6
f (r ) = 3 . (3.64)
R 1 for R r
3 r

From eq. 3.4 and

2 r 2 2(x 2 + y 2 + z 2 ) 2 r 1 3x x r 2
= = 2 , = , (3.65)
x x x x x x r5
!

where is Kronecker's delta, one gets the following expression. 98

98 Solivrez (2008), p. 205 eq. 12.

 Depolarization tensor method 59

1
0 0
3
1
For the sphere N =
1
0 0 = 1,
3 3

1
0 0
3
3x 2 r 2 (3.66)
3x y 3x z

r5 r5 r5
ext ! R3 3 y x 3 y2 r2 3 y z
n (r ) = .
3 r5 r5 r5

3z x 3z y 3z 2 r 2

r5 r5 r5

In dyadic notation
!! ! ! !! ! !
1 ext ! R3 3rr (r i r )1 V 3rr (r i r )1
N = 1, n (r ) = = , (3.67)
3 3 r5 4 r5

where V is the volume of the sphere. A general property of the depolarization

tensor for finite bodies is here put into evidence: its internal value is independent
of the body's volume V, while its external value is proportional to V. Notice that
Tr N = 1 and Tr next = 0, as previously discussed. Matrix N coincides with that
obtained in a simpler way from symmetry arguments (see Problem 19).

Other properties of the internal depolarization tensor N

Integral expressions of the eigenvalues of N
The proof of many important properties of the internal depolarization tensor N
may be easily derived from its general integral expressions. It is therefore
convenient to use the values given by eqs. 4.6, as reproduced next. N and N (0)
are different notations for the same integral. The values of N () with 0 are
necessary for the calculation of the depolarization tensor external to the body.

abc ds
2 0 a2 + s
Na = , where a b c, (3.68)
( ) ( )( )
3/2
b2 + s c 2 + s


abc ds
2 0 b2 + s
Nb = ,where a b c, (3.69)
( ) ( a + s )( c + s )
3/2
2 2

abc ds
2 0 c 2 + s
Nc = ,where a b c. (3.70)
( ) ( a + s )( b + s )
3/2
2 2

 60 Carlos E. Solivrez

The integrals are also highly symmetric as regards the semiaxes and it may be
easily proved that
Nb(a,b,c) = Na(b,a,c), Nc(a,b,c) = Na(c,b,a). (3.71)

That is, one single function gives the three eigenvalues, as long as they are sorted
in such a way that the condition a b c is fulfilled. As they are very efficient
methods for the numerical evaluation of integrals99, eqs. 3.71 may simplify the
numeral calculation of eigenvalues that are necessary for the general triaxial
elipsoids.

As the three integrals differ only in the single factor 1/(d2 + s) one might
assume that
1 1 1
if a b c, then 2 2 2 and Na Nb Nc . (3.72)
a +s b +s c +s

The conclusion is true, but the proof is flawed. This may be seen by using a similar
argument to prove that for constant a and c, the eigenvalues of N are a decreasing
function of b. As seen from Figure 14, Figure 15 and Figure 16, this is true for Nb,
but not for Na and Nc. The origin of the paradox is a combination of the symmetry
properties eq. 3.71 and the constraints imposed on the relative values of a, b and c
originated in the method used to derive the expressions of N in terms of elliptic
functions. In this case the symmetries eqs. 3.71 are lost and each eigenvalue is
given by a completely different function (see eqs. 4.9).

Diagonalization and inversion

As seen from eq. 4.6 the internal depolarization tensor N is diagonal in the
ellipsoid's principal system of coordinates, the one coinciding with its semiaxes.
Furthermore, from the integrals that define its eigenvalues (the diagonal matrix
elements or tensor's principal values) it is easily proved that they are never
negative and do not vanish for finite values of the semiaxes. Therefore, the N
matrix of all finite ellipsoids is positive definite and always has an inverse100 N-1.
The only exception are the cases when one or two semiaxes become infinite, where
the corresponding eigenvalues vanish, as discussed next. The calculation of the
internal and external fields for these cases is discussed at sections Sheet of constant
thickness and infinite extension and Right circular cylinder of infinite length.

Infinite semiaxes and the inverse of N

The constant thickness sheet of infinite extension and the infinitely long right
circular cylinder are two cases often discussed in introductory courses of
electromagnetism because of the simplicity of their treatment. They are not real
bodies, but they provide useful illustration of important properties and the fields
thus obtained are good aproximationes to those of a very long prolate spheirod
near the equator (for the cylinders case) and to a very short oblate spheroid near


99 Press, W. H. & Teukolsky, S. A. & Vetterling, W, T, & Flannery, B. P.; Numerical recipes in C;
Cambridge University Press (United Kingdom); 1992.
100 Goertzel and Tralli, p. 16.

 Depolarization tensor method 61

the axis (for the sheets case). Both cases can be taken to be the limits of spheroidal
ellipsoids where one axis (the symmetry one) or two axis (the equatorial ones) are
infinitely long. A third case will also be considered, the infinite right elliptic
cylinder, but only after the triaxial ellipsoids is discussed.
When semiaxis a becomes very large, the following limit is obtained for its
integral definition eq. 3.68:

abc ds
a 2
Na ( ) = lim
a +s
2
( ) ( a + s )( b + s )( c + s )
2 2 2

bc

ds (3.73)
a 2
= lim 3/2
= 0.
4/3 s
a + a2/3 b2 + s c 2 + s ( )( )

As the same thing happens for all coordinate axes, when the value of semiaxis
, the corresponding principal value N goes to 0. One should then check if the
other two principal values, eqs. 3.69 and 3.70, remain finite, which in our example
are Nb and Nc. For Nb,

abc ds
a 2
Nb ( ) = lim
b +s
2
( ) ( a + s )( b + s )( c + s )
2 2 2

bc

ds bc ds
(3.74)
= lim = ,
a 2 2 c 2 + s b2 + s
( )
3/2
s 2
(b + s ) ( )
3/2

1+ a2 c + s
2

where the integrand is finite and non-vanishing.
The same happens for Nc (),

abc ds
a 2
Nc ( ) = lim
c +s
2
( ) ( a + s )( b + s )( c + s )
2 2 2

bc

ds bc ds
(3.75)
= lim = .
a 2 2 b2 + s c 2 + s
( )
3/2
s 2
(c + s ) ( )
3/2

1+ a2 b + s
2

The last two cases correspond to the same type of integral, that will be solved
when discussing the infinitely long elliptic cylinder in next chapter.

The case b = c corresponds to the infinitely long right circular cylinder, where

b2 ds b2 b2
Nb ( ) = Nc ( ) = = = 12 . (3.76)

2 b2 + s
( )
2
(
2 b2 + s )
(b + )
2
 62 Carlos E. Solivrez

The next case for analysis is the sheet of constant thickness and infinite
extension. To that end consider an ellipsoid with finite semiaxis c, letting the other
two grow without limit. In order to preserve their ratio their values are written as
ka and kb, and the limit k is taken. Then

(ka)(kb)c
ds
lim Nc ( ) = lim
2
( c + s ) ( k a + s )( k b + s )
k k 3/2
2 2 2 2 2

abc ds c ds
2
= lim = (3.77)
(
2 c2 + s
)
3/2
2 s 2 s
(c + s )
k
3/2

a + k 2 b + k 2
2

c c
= = .
c2 + s c2 +

The infinite limit for the integral expressions for Nx() and Ny() is easily
proved to vanish. The same result is obtained at page 54 by using electrostatic
Gausss Law, where the field outside the body is also calculated showing it to be
independent of the ratio a/b, so that there is no elliptic sheet of infinite extensin
(although there is a right elliptic cylinder of infinite length, see eqs. 4.89 and 4.89).
The result is confirmed by taking appropiate limits in the cases of the elliptic
cylinder eq. 4.16 and the oblate spheroid eq. 4.29 with aspect ratio = 0.
In all these cases the inverse of N does not exist and one should solve the
equations that relate the components N, the applied field and the total field inside
the body, as illustrated for the sheet of constant thickness and infinite extension at
page 55 and for the right circular cylinder of infinite length al page 57.

N is invariant for similar ellipsoids

An important property of the internal depolarization tensor N is its
independence of the bodys volume, of the problem's scale. This is not evident from
a mere inspection of eq. 4.6. To prove the assertion consider a similar ellipsoid, one
whose semiaxes a, b, c differ from the original ones by a finite constant factor k:
a' = ka, b' = kb, c' = kc. The principal values of this similar ellipsoid turn out to be

k 3abc ds
2 k 2d 2 + s
N ( |ka,kb,kc) = , (3.78)
( ) ( k a + s )( k a + s )( k a + s )
2
2 2 2 2 2 2


apparently different. A change of integration variable s = k2s' does not modify the
limits of integration when = 0 and gives the identity

 Depolarization tensor method 63

k 3abc k 2ds'
2 0 k 2d 2 + k 2s'
N(0|ka,kb,kc) =
( ) (k a + k s')(k a + k s')(k a + k s')
2
2 2 2 2 2 2 2 2 2


(3.79)
abc ds'
2 0 d 2 + s'
= = N(0|a,b,c)
( ) ( )( )( )
2

a2 + s' a2 + s' a2 + s'


k 3abc k 2ds
2 0 k 2n2 + 2s'
N ' =

( k ) (k a + k s')(k b + k s')(k c + k s')
2 2 2 2 2 2 2 2 2

abc ds
2 0 n2 + s'
= = N .

( ) (a + s')(b + s')(c + s')

2 2 2

It follows that
N(0|a,b,c) = N(0|ka,kb,kc), where = x, y, z. (3.80)

It should be stressed that the property does not apply for the integrals with
0.
Therefore, the internal depolarization tensors of similar ellipsoids are equal.
This is of importance for the discussion of cavities in ellipsoidal bodies made in
page 116. It should be stressed here that the external depolarization tensor next
does not have this property; in fact, its value is proportional to the bodys volume
V.

N(0) is determined by aspect ratios

The principal values of N(0) are not a function of the three semiaxes, but of the
ratios of two of them to a third, their aspect ratios. This is a corollary of the
previous property when factor k in eqs. 3.80 is taken to be the inverse of the
largest semiaxis, in this book taken to be a.

1abc ds
N (0) =2
aaa0
( ) (
d /a + s a /a + s b /a + s c /a + s ) ( ) ( )
2 2 2 2

ds
= 12 , (3.82)
0 ( + s ) (1+ s )( + s )(
2

2 2
+s )
b c
where = x , y,z, = , = , x = 1, y = , z = .
a a

This is the convention used in the tabulations of N(0) for the triaxial ellipsoid, the
only type where the eigenvalues cannot be expressed in terms of elementary
functions.
 64 Carlos E. Solivrez

For further applications it is convenient to identify the values of aspect ratios

and for different categories of ellipsoids, as shown in Table 3, where

a b c, 1 . (3.83)

a b c Type of ellipsoid
c 101 0 sheet of constant thickness and infinite extension (see p. 54)
b b 0 0 right circular cylinder of infinite length (see p. 56)
b c < b 0 0 right elliptic cylinder of infinite length (see p. 58)
a a c < a 1 c/a oblate spheroid (see p. 69)
a b < a b b/a b/a prolate spheroid (see p. 74)
a a a 1 1 Sphere (see p. 58)
a b < a c < b b/a c/a triaxial ellipsoid (see p. 65)

Table 3. Values of and corresponding to different types of ellipsoids.

101 See eq. 3.77.

 Depolarization tensor method 65

Chapter 4:
The depolarization tensor: advanced treatment

Calculations of n from gravitational potentials

The gravitational potential of all ellipsoidal shapes have been solved both inside
and outside the body. The formulas obtained are useful not only for the calculation
of the depolarization tensor of the asymmetric triaxial ellipsoid but also for the
following shapes:
a > b > c: triaxial ellipsoid;
a b > c : right elliptic cylinder of infinite length;
a = b > c: oblate spheroid;
a > b = c: prolate spheroid.
The formulas given in the previous chapter for the infinite sheet, infinite
cylinder and sphere provide a verification of the validity of the formulas that will
be obtained next.

Triaxial ellipsoid
The triaxial ellipsoid is the shape where all the semiaxes
are different. Its principal coordinate system as in all
cases is the one such that axes x, y, z coincide with the
ordered semiaxes
a > b > c. (4.1)
As will be seen shortly, the components of the
depolarization tensor n are here given by elliptic integrals,
non-elementary trascendental functions unfamiliar for
Figure 7. Triaxial
many physicists and engineers. As their values are usually ellipsoid.
calculated from series expansions or taken from tables,
these functions are usually expressed in terms of a few canonical or normal ones
(Legendres elliptic functions), two of which appear here. Its definitions and
properties in the range of interest are given at Appendix 9. Nevertheless, as
pointed out in section Integral expressions of the eigenvalues of N, it might prove
simpler to use the single integral eq. 3.68 for the numerical calculation of internal
depolarization factors. For these reason the components of n will be expressed in
two different ways.
!
The calculation of V(r ), the internal gravitational potential of an homogeneous
solid ellipsoid of mass density can be made using either ellipsoidal102 or


102 Stratton, pp. 207-211.
 66 Carlos E. Solivrez

cartesian coordinates103. The latter, used here because it is the more concise
approach, was made in 1839 by the french mathematician Gustave Lejeune
Dirichlet104 and widely reproduced afterwards in books on potential theory. The
version of Dirichlets proof given by MacMillan105 is summarized next, where the
gravitational constant was chosen by this author to be unity106.
Taking as coordinate axes the principal ones of the ellipsoid, MacMillan changes
!
to a system of coordinates with origin at field point r . The radial part of the
!
volume integral turns out to have an upper limit which is a function of r . Using
symmetry arguments, he shows that several of the resulting integrals vanish,
leading to a quadratic expression in the field point coordinates,

! d 3r '
V(r ) = ! ! = C0 + C x x 2 + C y y 2 + C z z 2 . (4.2)
V r r '

The four coefficients are functions of the parameters a, b and c alone, and all
may be expressed in terms of a single integral and its derivatives respect of the
different semiaxes107,
1 C0 C0
C = , where = x , y,z, d x = a,d y = b,d z = c. (4.3)
d d d2

After a change of integration variable108 more symmetric expressions for Cx, Cy,
and Cz (see eq. 3.71) are obtained, as given below109. As all these coefficients are
needed to evaluate the external depolarization tensor next (see section General
expression of external n), it is convenient to use the auxiliary function f eq. 3.8:
!

!
f (r ) =
V(r ) abc
4
=
4
(
I0 + I x x 2 + I y y 2 + I z z 2 , ) (4.4)

where

ds ds
I0(0) = , I = ,
0 ( a + s )( b + s )( c + s )
2 2 2
0 ( d + s ) ( a + s )( b + s )( c + s )
2

2 2 2
(4.5)
where = x , y,z and d x = a, d y = b, d z = c.

103 MacMillan, pp. 45-62.
104 Stoner, p. 807.
105 MacMillan, section 32, pp. 45-49.
106 MacMillan, p. 24, eq. 20.1.
107 MacMillan, p. 48 eq. 32.10.

The sustitution sin = c / c + s replaces the polar angle by the new integration variable s,
108 2

and the upper limit of integration /2 by .
109 MacMillan, p. 49 eq. 32.13.

 Depolarization tensor method 67

Notice that the notation Ij (0) for the integrals comes about becauses it turns out
that they are a special case of the integrals Ij () discussed in section Obtention of
the external gravitational potential by Ivorys method.
From eqs. 4.4, 4.5 and 3.4 the depolarization factors N(0) are given by

ds

1
N (0) = 2 abc ,
0 ( d2 + s )( )(
a2 + s b2 + s c 2 + s )( ) (4.6)
where = x , y,z and d x = a, d y = b, d z = c.

The integrals may be checked for the depolarization tensor of the sphere (a = b
= c = R), where

ds
N R = 12 R3

0 R2 + s
(
R2 + s R2 + s R2 + s )( (4.7)
)( )( )

3/2
( )
1
( )
5/2
= 12 R3 R2 + s ds = 12 R3 23 R2 + s = ,
! 0 0 3

coincident with the values previously derived both by symmetry arguments and by
using Gausss Law (eq. 3.66 and Problem 19).

Internal depolarization tensor N

Expressions with normal elliptic integrals E and F
The integrals eqs. 4.4 are reduced to Legendre's incomplete elliptic integrals of
the first and second kind F(,k) and E(,k) (see Appendix 9) by the sustitution
a2 c2
s= 2
2
. (4.8)
tan(t ) sin(t )
It is thus obtained110:

N x (0) = Na (0) = ( E( ,k)+ F( ,k)) ,
( 1 2 ) 1 2

2 1 2
N y (0) = Nb (0) = 2 2 + E( ,k) F( ,k), (4.9)
1
2 2 2
(1 2 1 2 )( ) ( )
2
N z (0) = Nc (0) = E( ,k),

( 2
2 ) ( 2
2 ) 1 2

110 MacMillan, pp. 49-50; Osborn, eqs. 2.1 to 2.6; Stoner, eqs. 3.5.
 68 Carlos E. Solivrez

b c 1 2
where = , = , k = , 0 = cos 1
. (4.10)
a a 1 2 2

These principal values have the unit trace property eqs. 3.15. If one looks at the
denominators, formulas seem to be valid only for

1(b a), (b c)and 1 (c a).

This turns out to be false as may be seen from Figure 14, Figure 15 and Figure 16.
The reason is that, when taking limits, expressions eqs. 4.9 converge to the right
values.

N elliptic cylinder
The elliptic cylinder is the triaxial ellipsoid such
that one semiaxisa in our ordering is extremely
large. It includes, as a special case, the right circular
cylinder of infinite length (b = c in our convention).
From symmetry arguments the properties of the
cylinder are independent of coordinate x, and the
system reduces to an apparently two dimensional one.
Thus as follows also from taking the limit a in Figure 8. Elliptic cylinder.
eq. A7.2 the equation of the bodys surface is

y2 z2
2
+ 2 = 1. (4.11)
b c

The expressions of N ( )111 is given by eqs. 3.73,

N ec ( ) = 0. (4.12)
a
The two other terms are given by eqs. 3.74 and 3.75, where the value for Nc may
be obtained from that of Nb by permuting a with b. Solving the integral it is
obtained112


bc ds bc 2 c2 + s
2 b2 + s 3/2 c 2 + s
N ec ( ) = =
( )
b
2 b2 c 2 b2 + s

(4.13)
bc c2 +
= 1 .
b2 c 2 b2 +

111 All integrals I( ) are evaluated for 0 because otherwise the calculations will have to be
duplicated in the section devoted to the calculation of the External depolarization tensor.
112 Korn & Korn, p. 941, eq. 148.

 Depolarization tensor method 69

Permuting b and c gives

bc b2 + c 2 +
N ( ) = 0,
ec
a
N ( ) = 2 2
ec
b ,
b c b2 +
(4.14)
bc b2 + c 2 +
N c ( ) = 2 2
ec
.
b c c2 +

The depolarization factors are obtained113 by setting = 0 in the previous

equations, giving

c b
N ec (0) = 0, N ce (0) = , N ce (0) = . (4.15)
a b
b+c c
b+c

The values satisfy the unit trace rule eq. 3.15 and the order Na Nb Na. Notice
that these eigenvalues cannot be obtained from the graphs given by Figure 14,
Figure 15 or Figure 16.
For the right circular cylinder the last two eigenvalues reduce to the value
given by eq. 3.57. From here one may obtain the constant thickness sheet of
infinite extension. The limits to take are

c
Nbcts (0) = lim N ce (0) = lim = 0,
b b b b + c

b 1 (4.16)
N ccts (0) = lim N ce (0) = lim = lim = 1,
b c b b + c b c
1+
b

which coincide with the calculations made using Gausss Theorem, eq. 3.47.

N oblate spheroid
An oblate spheroid is, in our convention, the
ellipsoid of revolution such that its polar semiaxis c, its
axis of rotational symmetry,is smaller than the
equatorial ones: a = b > c. The equatorial eigenvalue Ne
of the oblate spheroid is obtained from eq. 4.6. As the
same integrals will be used later for the expression of Figure 9. Oblate spheroid.
the external tensor, the lower limit is taken, that for
the internal case should be made zero. Thus,

113 Osborn, eqs. 2.17 and 2.18. MacMillan, p. 71, use value of V with = 0.
 70 Carlos E. Solivrez

a 2c ds
N ( ) = N ( ) = N ( ) =
o o o
, (4.17)
( )
e a b
2 c 2 + s a2 + s 2

The primitive of this integral is given in Appendix 8, the constants being A = c 2,
B = a2, < B. Therefore, eq. A8.5 applies giving the following equatorial term.


a c 2
c2 + s 1 c +s 2
N e ( ) =
o
+ arctan 2 2
(
2 a2 c 2 a2 + s a2 c 2 )( ) ( 3/2
) a c
(4.18)
a2c c 2 + a2c c2 +
= 12 + 12 /2 arctan .

( a c )( a + ) (
2 2 2 3/2
a2 c 2 ) a2 c 2

Using
1 a2 c 2
tan 0 = = = tan , (4.19)
2 tan 0 c2 +

N oe ( ) = N oa ( ) = N ob ( )
a2c c 2 + a2c a2 c 2 (4.20)
= 12 + 12 arctan .

( a c )( a + ) (
2 2 2
a c
2 2
)
3/2
c +
2

Alternative expressions expressions for angle can be obtained as follows114 .

sin
tan = , sin = k a2 c 2 , cos = k c 2 + ,
cos
sin2 + cos2 = k 2 ( a2 + ) = 1, so that k = ( a2 + )
1/2
, (4.21)
a2 c 2 c2 +
therefore sin = < 1, cos = 2 < 1.
a2 + a +

114 MacMillan, p. 62 eq. 39.2, uses the expression in terms of arcsin.

 Depolarization tensor method 71

N oe ( ) = N oa ( ) = N ob ( )
a2c c 2 + a2c a2 c 2
= 12 + 12 arctan
( a c )( a + ) ( a c )
2 2 2 2 2
3/2
c2 +

a2c c 2 + a2c c2 + (4.22)

= 12 + 12 arccos
( a c )( a + ) ( a c )
2 2 2 2 2
3/2
a2 +

a2c c 2 + a2c a2 c 2
= 12 + 12 arcsin .
( a c )( a + ) ( a c )
2 2 2 2 2
3/2
a2 +

For = 0, the eigenvalues of N may be expressed in terms of the aspect ratio

as follows. Notice that = , so that either one may be used in the equation. The
second is chosen so that Figure 10 and Figure 12 have the same independent
variable.

N oe (0) = N oa (0) = N ob (0)

2 1 2
= 12 + 12 arctan
(1 ) ( 2
1 2
3/2
)

2
(4.23)
= 12 + 12 arccos( )
(1 ) (1 )
2 2
3/2

= 12
2
+ 12

(1 ) (1 2 )3/2
2
arcsin ( 1 ).2

Osborn and Maxwell use the expression in terms of arcsin115 instead of the simpler
one in terms of arccos.
The polar term may be obtained solving the definition integral eq. 4.6, whose
primitive is given in Appendix 8, where the constants take the values A = c 2, B = 1
so that < B. Therefore, eq. A8.11 applies giving

115 Osborn, eq. 2.19. Maxwell, p. 69 eqs. 438.11 and 438.12, where Np =-L/4, Ne = -M/4 = -
N/4, e2 = 1 2.
 72 Carlos E. Solivrez

a2c ds
N op ( ) = N oc ( ) = ,
2 c2 + s
( ) ( )
3/2
a +s
2

c2 + s
a2c a2c c 2 + s a2c
= 2 2 2 arctan (4.24)

(
c 2 + s a2 + s a c a + s ) (
a2 c 2 )( ) ( )
3/2 2
a c
2

a2c a2c c2 +
= + arctan 2 2 .

(a c )
2 2
c2 + ( a c
2 2
3/2

) a c 2
Using


(
tan /2 = 1/ tan( 0 ), ) (4.25)

and eqs. 4.21, the following expressions are obtained for Nc().

N op ( ) = N oc ( )

a2c a2c a2 c 2
= arctan
(a c )2 2
c2 + (a c )2 2
3/2
c2 +

a2c a2c c2 + (4.26)

= arccos
(a c )2 2
c +
2
(a c )2 2
3/2
a2 +

a2c a2 c 2
a2c
= arcsin 2 .
(a c )2 2
c2 + a2 c 2
3/2
( a +
)

For = 0, this eigenvalue of N may be expressed in terms of the aspect ratio as

follows:

N op (0) = N oc (0)

1 1 2
= arctan
(1 ) (1 )
2 2
3/2

1 (4.27)
= arccos( )
1 2
(1 2 )3/2

=
1
1 2


(1 2 )3/2
arcsin ( 1 ).2

 Depolarization tensor method 73

Notice, by inspection, that eqs. 4.23 and 4.27 fulfill the unit trace rule

Tr N = 2N eo + N po = 1. (4.28)

If 1 (very flat oblate spheroid), the eigenvalues may be approximated by116

Ne = Na = Nb 2 , N p = Nc 1 + 2 2 , (4.29)
4 2

which should be compared with the values given at section Sheet of constant
thickness and infinite extension.

Figure 10. Np and Ne as a function of for the oblate spheroid.

Figure 10 shows the values of the equatorial eigenvalue Ne and the polar one Np
for the oblate spheroid. The values = = 0 (a = b = ) correspond to the
constant thickness sheet of infinite extension where Np = 1, Ne = 0 (Table 3 and
eq. 3.47); the values of the sphere are = = 1 (eq. 3.66). Both functions will turn
out to be important in the discussion of the eigenvalues of N for the triaxial
ellipsoid (see section Graphs of Na, Nb and Nc) where they provide upper or lower
bounds of certain eigenvalues.


116 Osborn, eq. 2.21.
 74 Carlos E. Solivrez

N prolate spheroid
A prolate spheroid is an ellipsoid of revolution
(spheroid) such that in our conventionthe polar
semiaxis Na, its axis of rotational symmetry, is greater
than the equatorial ones: a > b = c.
The equatorial eigenvalues of N for the prolate
spheroid are obtained from eq. 4.6. As the same
integrals will be used later for the expression of the
external tensor, the lower limit is taken, that for the
internal case should be made zero. Using the upper
index p to identify the prolate spheroid, it is found Figure 11. Prolate spheroid.
that

ac 2 ds
N e ( ) = N b ( ) = N c ( ) =
p p p
. (4.30)

2 a2 + s c 2 + s 2
( )( )
The integral eq. 4.30 is discussed in Appendix 8, where the constants are A = a2,
B = c2, A > B, so that eq. A8.6 applies. It is thus found

N pe ( ) = N pb ( ) = N pc ( )


1 2 1+ s 1 2 1+ s 1 2
= 2 4 ln (4.31)

(
1 2 2 + s )( 1 2) ( )
3/2
1+ s + 1 2

ac 2 a2 + ac 2 a2 + a2 c 2
= 12 + 14 ln .

(a c ) (c + ) (
2 2 2
a c
2 2
3/2
) a +
2
+ a c
2 2

The value of the logarithm can be made positive (argument > 1) and this function
can be replaced by the hyperbolic ones by using the some of the following
identities:

a2 + a2 c 2
= ln
a2 + + a2 c 2
= 2ln
( )
a2 + + a2 c 2
ln . (4.32)
a2 + + a2 c 2 a2 + a2 c 2 c +
2

 Depolarization tensor method 75

The following relationship117 is used to obtain alternative expressions:

a2 +
if cosh = ,
c2 +

e = cosh + sinh = cosh + cosh 1 =
( 2 a2 + + a2 c 2 ). (4.33)

c2 +

N pe ( ) = N pb ( ) = N pc ( )
ac 2 a2 + ac 2 a2 + + a2 c 2
= 12 14 ln
(a c ) (c + ) (
2 2 2
a2 c 2
3/2
)
a2 + a2 c 2

ac 2 a2 + ac 2 a2 + + a2 c 2
= 12 12 ln (4.34)

(a c ) (c + ) (
2 2 2
a2 c 2
3/2
) c2 +

ac 2 a2 + ac 2 a2 +
= 12 12 cosh 1 .
(a c ) (c + ) (a c )
2 2 2 2 2
3/2
c2 +

The corresponding eigenvalues of N are found setting = 0 in eqs. 4.34 and

using the aspect ratio . Notice that = , so that either one may be used in the
equation, and the second is chosen so that Figure 10 and Figure 12 have the same
independent variable118. Then

N pe (0) = N pb (0) = N pc (0)

1 2 1+ 1 2
= 12 14 ln
(1 ) ( 2
1 2
3/2
) 1 1 2

= 1 1
12
2
1+ 1 2
ln
( ) (4.35)
2
(1 ) ( 2
1 2
3/2
)

1 2 1
= 12 12 cosh 1 .
(1 ) (1 )
2 2
3/2

117 Korn & Korn, p. 817 eqs. 21.2-37
118 Osborn, eq. 2.11. Macmillan, p. 63 eq. 4 with = 0.
 76 Carlos E. Solivrez

The integral expression for the polar term may be obtained from eq. 4.6 giving

ac 2 ds
N ( ) = N ( ) =
p p
(4.36)
( ) ( )
p a
2 a2 + s 3/2 c 2 + s

The primitive of this integral is given in Appendix 8. As A = a2, B = c 2, A > B,
eq. A8.12 should be used obtaining

a + s a c
ac 2 ac 2 2 2 2

1
N p ( ) = Na ( ) =
p p
+ ln
( )
a2 c 2 a2 + s 2 a2 c 2
( )
3/2 2
a + s + a c
2 2

(4.37)
ac 2 ac 2 a2 + + a2 c 2
= + 12 ln 2 .

(a c )
2 2
a2 + ( a c
2 2
3/2
)
a + a2
c 2

where the argument has been inverted to make positive the value of the natural
logarithm. The same alternatives as for eqs. 4.34 are valid here, giving

N pp ( ) = Nap ( )

ac 2 ac 2 a2 + + a2 c 2
= + 12 ln
(a c )
2 2
a2 + ( a2 c 2
3/2
) a + a c
2 2 2

ac 2 a2 + + a2 c 2
ac 2
= + ln (4.38)

(a c )
2 2
a2 + (
a2 c 2
3/2
)
c2 +

ac 2 ac 2 a2 +
= + cosh 1 .
(a c )
2 2
a +
2
(a c )
2 2
3/2
c2 +

The eigenvalues of N(0) in terms of aspect ratio are

N pp (0) = N pc (0)

2 2 1+ 1 2
= + 12 ln
(1 ) (
2
1 2
3/2

)1 1 2

2 2 1+ 1 2 (4.39)
= + ln
(1 ) (
2
1 2
3/2
)

2 2 1 1
= + cosh .
(1 ) (
2
1 2
3/2
)

 Depolarization tensor method 77

For very slender prolate spheroids the following approximations are valid119,

Ne (0) = Nb (0) = Nc (0) 1
2 (1+ )
2
12 2
ln(2/ ),
(4.40)

N p (0) = Na (0) + ln(2/ ), if 1.
2 2

The aspect ratio vanishes in the limit of infinitely large a, whereupon one gets
the eigenvalues of the right circular clinder of infinite length previously obtained
by using the electrostatic Gausss Law (see eqs. 3.57).

Figure 12. Np and Ne as a function of for the prolate spheroid.

Figure 12 shows the values of the polar eigenvalue Np and the equatorial one Ne
for the prolate spheroid. The values = 0 (a = ), Np = 0 correspond to the infinite
right circular cylinder where Ne = (Table 3 and eq. 3.66); the values = = 1,
Na = Nb = Nc = 1/3, to the sphere (eq. 3.66). Both functions will turn out to be
important in the discussion of the eigenvalues of N for the triaxial ellipsoid (see
section Graphs of Na, Nb and Nc) where they provide upper or lower bounds.

Unified treatment of N for spheroids

As pointed out in section Integral expressions of the eigenvalues of N, a single
integral like eq. 3.68 characterizes all eigenvalues of N. The same thing happens
with more specialized integrals, like the ones with two equal semiaxes that give
the spheroids eigenvalues. The reason for having different integrals for Na, Nb, Nc
is the convention a b c, wich underlies eq. 4.51. This subtle point is discussed
next.
Consider the integral expression for the equatorial eigenvalue of the oblate
spheroid eq. 4.17, where the domain of positive real numbers is divided in two,

119 Osborn, eq. 2.13.
 78 Carlos E. Solivrez

ds c

1
N = 2 , where 0 = 1,
o

(1+ s )
e
0
2
2+s a

(4.41)
ds c
N ? = 12 , where 1 = .
(1+ s ) a
2
0 2+s

The added ranges are expected to give a single continuous curve with
continuous slope. The curve for 0 1 the same given in Figure 10is the
dashed one in the left half of Figure 13. N?( ), not represented there, is an
increasing function of with minimum value 1/3 at = 1 and maximum value 1/2
at = . As that domain of the independent variable may not be represented
directly, it is convenient to make a change of variable to 1/, that corresponds to
the domain [1,0], its graph being the dashed curve on the right side of of Figure 13.
This resulting integral is

1 ds 4ds
0
N ? = 12 = 12
( )
2

( ) 1 + 2s 1+ 2s
2 2
0
1+ s +s
2 (4.42)

ds'2
1 a
= 12 , where s'= 2s, 1 = 0.
( ) c
2
0
2
+ s' 1+ s'

The last integral coincides with that for the equatorial eigenvalue of the prolate
spheroid, eq. 4.30. Its graphic previously given in Figure 12 is the dashed line
at the right of Figure 13. The definition = a/c seems to contradict its
identification with the parameter = c/a. As it was initially assumed that c > a, by
the convention a b c the exchange of names a c should be made. That is why
in the right side of Figure 13 the independent variable, that does not respect the
convention, is labeled as a/c.
Therefore

ds

1
N = 2 = N ep
? 2
(4.43)
( )
2
0 2+s 1+ s

is the equatorial eigenvalue of the prolate spheroid, eq. 4.30. The reader should
verify, in a similar fashion, that the extension to > 1 of the expression of the polar
eigenvalue of the oblate spheroid gives the polar eigenvalue of the prolate
spheroid, the continuous line in Figure 13.
The functions that give the dependence of the eigenvalues is very different for
the oblate (inverse trigonometric functions) and the prolate spheroid (logarithm
or inverse hyperbolic functions), as seen in eqs. A8.5 and A8.6. The origin of the
difference is the sign of = (1- 2), that is, wether c is smaller or greater than a.

 Depolarization tensor method 79

MacMillan120 has shown that the use of complex variable unifies both solutions
(see Problem 26)

Graphs of Na, Nb and Nc

Figure 13. The equatorial and polar eigenvalues
of spheroids are continuous curves.

For accurate calculations of the eigenvalues of N the best choice is to use

mathematical software that includes the evaluation of the incomplete elliptic
functions E(,k) and F(,k). On the other hand, a better understanting of their
behavior is obtained by the use of graphs. As discussed in section N(0) is
determined by aspect ratios, the eigenvalues are functions of the ratios of semiaxes
= b/a and = c/a. The best graphs known by the author, those of Osborn, depict
them as sets of functions of , where each of the 11 curves has fixed values of 0
1 at intervals of 0.1. These graphs for Na, Nb and Nc are reproduced next at full
page width.
As shown in the combined graph for the three eigenvalues, Figure 17, for any
fixed value of the eigenvalues are ordered as follows:

Na Nb Nc. (4.44)

Two finite eigenvalues may be equal only for the equatorial case of the prolate and
oblate spheroids (see Figure 12 and Figure 10). The lowest values of Na are given
by the curve for the polar eigenvalues of the prolate spheroids (see Figure 12) and
the largest values of Nc ( ,) for the polar eigenvalues of the oblate spheroids (see

120 MacMillan, p. 63 eq. 39.4.
 80 Carlos E. Solivrez

Figure 10). The in-between case Nb is bounded from below by Na and from above
by Nc.

Na
Function Na(, ) is shown in Figure 14, where its values are confined to the
grey shaded region, bounded by the upper curve Na(1, ) (oblate spheroid, see
Figure 10) and the lowest curve Na(,) (prolate spheroid, see Figure 12) as follows
from Table 3 the ordering of semiaxes convention eq. 3.83. Therefore

0 Na( 0, 0 ) Na(, 0) Na(1, 0 ) 1/3. (4.45)

For fixed 0, as shown by the constant curves, Na is an increasing function of ;

that is
if 2 > 1, Na (2, 0) > Na (1, 0). (4.46)

Each of the curves drawn correspond to a different value of the parameter =

b/a, the highest being the one for the oblate spheroid. The upper vertex Na(1,1) =
Nb(1,1) = Nc(1,1) = 1/3 corresponds to the sphere, the transition value from oblate
to prolate spheroids. The lowest vertex, Na(,0) = Na(1,0) = 0, characterizes the
constant thickness sheet of infinite extension (see Table 3).

Nb
Function Nb(, ) is shown in Figure 15, where its values are confined to the
grey shaded region, bounded by the lower curve Nb(1, ) (oblate spheroid, see
Figure 10)
and the upper curve Na(, ) (prolate spheroid, see Figure 12) as follows from Table
3 and the ordering of semiaxes convention eq. 3.83. Therefore

0 Nb(1, 0 ) Nb(, 0) Nb( 0, 0 ) 1/2. (4.47)

For fixed 0, as shown by the constant curves, Nb is an decreasing function of

; that is
if 2 > 1 , Na (2, 0) < Na (1, 0). (4.48)

Each of the curves drawn corresponds to a different value of the parameter

= b/a, the highest being the one for the prolate spheroid. The right hand vertex
Nb(1,1) = Na(1,1) = Nc(1,1) = 1/3 corresponds to the sphere, the common limit of
the oblate and prolate spheroids. The lowest left vertex, Nb(,0) = Nb(1,0) = 0,
characterizes the constant thickness sheet of infinite extension; the upper left
vertex, Nb(0,0), characterizes the right elliptic cylinder of infinite length (see Table
3). The vertical line = 0 corresponds to the the elliptic cylinders, whose values
cannot be identified from the graph, because all of them have = = 0 in this
parametrization.

Nc
Function Nc(, ) is shown in Figure 16, where its values are confined to the grey
shaded region, bounded by the upper curve Nc(1, ) (oblate spheroid, see Figure
10) and the lowest curve Na(, ) (prolate spheroid, see Figure 12) as follows from
Table 3 and the ordering of semiaxes convention eq. 3.83.

 Depolarization tensor method 81

Figure 14. Na (L/4 ) as a function of c/a ( ), where each curve corresponds
to a different value of the parameter b/a ( ) (taken from Osborn, Figure 1).

 82 Carlos E. Solivrez

Figure 15. Nb (M/4 ) as a function of c/a ( ), where each curve corresponds

to a different value of the parameter b/a ( ) (taken from Osborn, Figure 2).

 Depolarization tensor method 83

Figure 16. Nc (N/4 ) as a function of c/a ( ), where each curve corresponds
to a different value of the parameter b/a ( ) (taken from Osborn, Figure 3).

 84 Carlos E. Solivrez

Therefore
1/3 Nc(1, 0, 0 ) Nc(1,, 0) Nc(1,1, 0 ) 1. (4.49)

For fixed 0, as shown by the constant curves, Nc is an increasing function of ;

that is
if 2 > 1, Na (2, 0) > Na (1, 0). (4.50)

Each of the curves drawn correspond to a different value of the parameter

= b/a, the highest being the one for the oblate spheroid. The right vertex Nc(1,1)
= Nb(1,1) = Nc(1,1) = 1/3 corresponds to the sphere, the common limit of the oblate
and prolate spheroids. The lowest left vertex, Nc(0,0) = Nb(0,0) = 1/2, characterizes
the right elliptic cylinder of infinite length; the highest left vertex, Nc(1,0) = 1, the
constant thickness sheet of infinite extension (see Table 3).

Combined graph

Figure 17. Distribution of eigenvalues of N.

In Figure 17 the graphs for the three eigenvalues are piled up, identifying with
different colors the regions and boundary curves for each one121, where


121 A similar one was given in Stoner, Figure 1.

 Depolarization tensor method 85

eigenvalues Np correspond to the polar axis and Ne to the equatorial ones of the
spheroids. As an example, the segments of allowed values for each are shown for
= 0.25, illustrating the ordering given by eq. 3.72. The constant curves have
been omitted for simplicity.
The reader should exercise its understanding of the graphs with activities like
the one proposed in Problem 22 and incorporated into Figure 17. The ellipsoids
there analyzed are represented as seen from a direction at equal angles with the
three semiaxes, and the length of each (c is the vertical one) is represented as the
corresponding fraction of a. The value of is kept constant () and that of is ,
or 1. For Na, on the curve for the polar eigenvalue of the prolate spheroid, b is
equal to c (Nb = Nc) . When going up along the line = the b semiaxis increases
for Na until it is equal to a on the curve for the equatorial eigenvalue of the oblate
spheroid (Na = Nb). For Nb the b semiaxis decreases until is equal to c (Nb = Nc) on
the curve for the equatorial eigenvalue of the prolate spheroid. The behaviour is
again reversed for Nc, where b increases until it is again equal to a (Na = Nb) on the
curve for the polar eigenvalue of the oblate spheroid.

The range of the different eigenvalues is

0 Na 1/3, 0 Nb 1/2, 1/3 Nc 1, (4.51)

so that Nc is the only eigenvalue that may be univocally identified from its value in
the range (1/2,1] (see problem Problem 20).

External depolarization tensor

Obtention of the external gravitational potential by Ivorys method
!
In the previous sections the value of the gravitational potential V(r ) inside a
solid homogeneous general ellipsoid, eq. 4.2, was used to derive the values of the
internal depolarization tensor N. If a similar procedure were used for obtaining the
value outside the body, next, the expressions obtained would be extremely
cumbersome. More than a century ago, in 1809, the Scottish mathematician James
Ivory122 gave a concise way of deriving the potentials external values from the
internal ones using ellipsoids confocal with the bodys surface eq. A7.2. These
confocal ellipsoids are obtained from eq. A7.20, where > 0 is usually defined as
!
the largest algebraic root of that equation123, and its value is the sane for all r on

that surface124. These confocal ellipsoidal surfaces are larger than the body,
enclose it completely and have the same potential as certain corresponding points
inside the body (see section Confocal ellipsoids of Appendix 7). Only the final
expressions obtained using Ivorys method are given here. Readers interested in

122 James Ivory at English Wikipedia.
123 But see eq. Problem 20.
124 For a better understanding of the method the reader should start by reading section Confocal
ellipsoids and tackling afterwards Problem 32, its application for the obtention of the external
potential of a solid and homogeneous spherical mass.
 86 Carlos E. Solivrez

learning how to obtain them may look at the relevant sections of Macmillans
book125.
The potential V and the auxiliary function f eq. 3.8 are then given by126,
!

!
f (r ) =
V(r ) 1
4
(
= 2 N0( )+ Na ( ) x 2 + Nb ( ) y 2 + Nc ( ) z 2 , ) (4.52)

where the coefficients N (0 ) ( = x, y, z or a, b, c) are the eigenvalues of the internal

depolarization tensor, but also appear in the expression of the external one for
0. These coefficientes are given by the following integrals

abc ds
2 ( r! )
N0( ) = , (4.53)

( a + s )( b + s )( c + s )
2 2 2

abc ds
2 ( r! ) d 2 + s
N ( ) = ,
( )( )(
a2 + s b2 + s c 2 + s)( )
where d x = a,d y = b,d z = c, = x , y,z and (4.54)
x2 y2 z2
+ + = 1.
a2
+ b 2
+ c 2
+

The expressions of the integrals in terms of Legendres elliptic functions are

given at the end of Appendix 9127, where more details are given about these
functions. The corresponding values for Nj ( ) are

N0( ) =
abc
F ( ),k , ( ) (4.55)
a2
c 2

Na ( ) =
abc
(E (( ),k ) + F (( ),k )) ,

(
a2 c 2 a2 b2 ) (4.56)

125 MacMillan, sections 35-36, pp. 52-58.
126 MacMillan, p. 56, eq. 36.1.
127 See also Bartczak & Breiter & Jusiel; Ellipsoids, material points and material segments; Celestial
Mech. Dyn. Astr.; eq. 6.

 Depolarization tensor method 87

abc c 2 + abc a2 c 2
Nb ( ) = + (
E ( ),k )
( )( )( ) ( a b )( b c )
2 2 2 2
a2 + b2 + b2 c 2

abc
(
F ( ),k , )
(
a2 c 2 a2 b2 )
(4.56)
abc b2 +
Nc ( ) =
abc
(
E ( ),k , )
( a + )( c + ) ( b c ) ( a c ) ( b c )
2 2 2 2 2 2 2 2

a2 b2 x2 y2 z2 a2 c 2
k= , + + = 1, ( ) = arcsin .
a2 c 2 a2 + b2 + c 2 + a +
2

Notice that the only difference with the interior values of the potential eq. 4.4 is
!
that the lower limit 0 of the integrals is here replaced by (r ) , origin of the
notation N (0).

Near and far away point approximations to next

Expressions for the near or far away points are the simplest aproximations to
the external field of a uniformly polarized ellipsoid body, both when the
polarization is permanent or induced by an applied field.
For field points very near to the bodys surface the values eqs. 3.41 and 3.42
may be used, coinciding with eq. 4.66 for = 0:

x 2 /a4 x y /a2b2 x z /a2c 2

! ! !
w(r ) w(r ) w(r )
x y /a2b2 2 2
! y 2 /b4 y z /b c
next (r S ) = N(0) ! ! ! ,
w(r ) w(r ) w(r )
(4.57)
x z /a2c 2 y z /b2c 2 2
z /c 4
! ! !
w(r ) w(r ) w(r )

! x 2 y2 z2
where w(r ) = 4 + 4 + 4 .
a b c

For points far away from the uniformly polarized body it is expected that the
field will be the dipolar one (the extension of the second of eqs. 3.66) generated by
the dipole moment of the body, presumption that is confirmed by a calculation of
the limit value of and the resulting external tensor (see Problem 33).

General expression of external n

From eq. 3.4 the components of the depolarization tensor, either inside or
outside the ellipsoidal body, are given by
 88 Carlos E. Solivrez

!
! 2 f (r ) !
n ext
(r ) = , where r V . (4.58)
x x

For the purpose of taking first derivatives it is convenient to express f, eqs. 4.52
and 4.54, as the following single integral:

x2 y2 z2
+ + 1
!


!
f (r ) = 12 a + s b + s c + s
2 2 2

1
ds = Q(s, r )ds . (4.59)
( )( )( )
2
(r )
!
a +s b +s c +s
2 2 2
(
!
r )


!
Written in this fashion the integrand Q(s, r ) vanishes for the upper and lower limit
of the integral (see eq. A7.20), so that

! ! !

x ( r! )
Q(s, r )ds =
( r! )
Q(s, r )ds +
x ( r! ) x
Q(s, r )ds

( )
! ! !

= Q s, r ! + Q(s, r )ds = Q(s, r )ds (4.60)
( r ) x ! x ! x
(r ) (r )

2x
=
abc
2 ( r! ) (d 2 + s ) a2 + s b2 + s c 2 + s
ds = N ) x . ( )

( )( )( )
where eq. 4.54 has been used.
That is,
!
f (r )

x
( !
)
= N (r ) x , (4.61)

where N0( ) does not appear.
The calculation of the second derivatives of f gives
!
f (r )
=
x x x
!
N (r ) x ( ( ) )
! (4.62)
(!
)
= N (r ) + x

N

( ( )) (r )
x
,

requires the value of / x , which may be obtained from its implicit definition

eq. A7.20. The result will shed light on the meaning of the external tensor and
verify the bodys surface value of next eq. 3.41.

 Depolarization tensor method 89

From eq. A7.20,

x2 z2
y2
( )
1

+ 2 + 2 =
x a + b + c + x =x, y, z
2
x2 d 2 +

2x
( )
1
=
d +
2
+

x2 d 2 + x (4.63)

2x

( )
2
= x2 d 2 + x = x 1 = 0.
d +
2

2x
d 2 +
That is, = , (4.64)
x x 2 y z 2 2
+ +
( ) ( ) ( )
2 2 2
a2 + b2 + c2 +


which, as may be seen from eq. A7.8, defines the components of a vector normal to
the surface of the confocal ellipsoid A7.20.

The derivatives of N give

N ( ) N ( )
=
x x

abc ds
2 ( r! ) d + s
= (4.65)
( )( )(
a2 + s b2 + s c 2 + s x )( )

x 2
y2 z2
/ x + +
2
( ) ( ) ( )
d + a2 + c +
2 2 2
b +
2 2

= abc .

( d + ) ( a + )( b + )( c + )

2 2 2

From eqs. 4.61, 4.62 and 4.65 it is thus obtained

! N ( )
2 f (r )
!
n (r ) =
ext

x x
= N +
x
x ( )
! !
s (r , ) s (r , )
( )
= N abc .

a + b + c +
2 2
( 2
)( )( )
 90 Carlos E. Solivrez

Therefore128

! !
! s(r | ) s(r | )
n (r ) = N( ) abc
ext
,
( )(
a2 + b2 + c 2 + )( )

abc ds
2 d 2 + s
where N ( ) =

( ) ( a + s )( b + s )( c + s )
2 2 2 (4.66)

!
and s(r | ) is the unit vector normal to the surface of the confocal ellipsoid
!
eq. A7.20 at the external field point r :

x
! d +
2

s (r | ) = ,
x2 y2 z2
+ +
(a + ) (b + ) (c + )
2 2 2
2 2 2

2
x2 y2 z2
where 2 + 2 + 2 = 1. (4.67)
a + b + c +

Notice that the first term of next is a field point dependent extension of the
internal depolarization tensor N and proportional to its volume. In the second term
the numerator selects components of the polarization normal to the confocal
ellipsoid, while the denominator gives a distance of the order of r3 (see eq. A7.22).
For = 0 it is obtained the value of next at the surface of the body eq. 3.41.

Traces
In what follows it is verified if eq. 4.66 fulfills the zero trace rule eqs. 3.15.

( s
! !
)
2
2

(r | ) / s(r | )
! ext !
Trnext (r ) = n (r ) = N ( ) abc
. (4.68)


( a + )( b + )( c + )
2 2 2

128 Barczak & t & Jusiel; Ellipsoids, material points and material segments; Celestial Mech. Dyn. Astr.
; DOI 10.1007/s10569-006-9017-x. The value coincides with the Hessian of the gravitational
potential V, eqs. 9 and 10.

 Depolarization tensor method 91

From the definitions of N eqs. 4.54 it follows that

abc 1 1 1 ds
N ( ) = 2 a2 + s + b2 + s + c 2 + s
( a + s )( b + s )( c + s )
2 2 2

abc ( b2 + s )( c 2 + s ) + ( a2 + s )( c 2 + s ) + ( a2 + s )( b2 + s )

2
= ds
( )( )( )
3/2
a2 + s b2 + s c 2 + s

(4.69)

d 2
((
abc ds a + s b + s c + s
2
)(
2
)( ))
abc

= ds =
( )(
2 a2 + s b2 + s c 2 + s

3/2
)( )
a2 + s b2 + s c 2 + s ( )( )( )

abc
= .

a + b + c +
2 2
( 2
)( )( )
That is
abc
Tr N( ) = , (4.70)

( )(
a2 + b2 + c 2 + )( )
which, as its eigenvalues Na( ), Nb( ) and Nc( ),is of the order of r3 (see eq. A7.22)
and may be used to simplify its calculation.

From eq. A7.18,

!
s 2

(r | )

! 2
= 1. (4.71)
s(r | )

Upon replacement in the initial expression for the trace, it is verified that
ext !
Tr n (r ) = 0. (4.72)

Verificacion for the sphere

Equations 4.66 and 4.67 will next be verified using the solution found for the
sphere using the electrostatic Gausss Law eqs. 3.66 and 3.67. To that end the
different componentes of those equations are evaluated next.
The eigenvalues of N() are

abc ds
2 d 2 + s
Na ( ) = Nb ( ) = Nc ( ) =
( ) ( a + s )( b + s )( c + s )
2 2 2

(4.73)

R3 ds R3 2
( ) R3 2
( )
3/2 3/2
= = R +s = R + ,
2 R2 + s
( ) 3 3
5/2


 92 Carlos E. Solivrez

where a = b = c = R. The value of to be used is the solution of the equation

x2 y2 z2 r2
+ + = = 1, = r 2 R2 . (4.74)
a + b + c + R +
2 2 2 2

Therefore
R3 R3
Na ( ) = Nb ( ) = Nc ( ) = = 13 , (4.75)
( )
3/2
3 R +2 r3

its value being 1/3 for the spheres interior where = 0.

The last term of eq. 4.66 has in the the numerator

! ! x x r!
s(r | ) = s (r | )x = = , (4.76)
r r

and the denominator is

( a + )( b + )( c + ) = r ,
2 2 2 3
(4.77)
giving
!! !!
! 1 R3 3 rr 1 3 1 3r r
n (r ) = 3 3 1 R
ext
= R 1 . (4.78)
r r5 3 r 3 r5

This is the dyadic such that both for permanent or induced electric (eqs. 2.7 and
2.32a) and magnetic polarization (eqs. 2.40 and 2.48) gives the following external
dipolar fields contributions (see eqs. 3.66 and A4.1):
! ! ! !
! ! 3( p i r )r r 2p
E p! (r ) = k1 ,
r5
! ! ! (4.79)
! ! ' 3( mi r )r r 2m
Hm! (r ) = ,
4 r5

where eqs. A1.4 and A1.11 have been used.

Elliptic cylinder
The first term of eq. 4.66 may be obtained from eqs. 4.12, 4.13 and 4.14129:

N ec
a
( ) = 0,
bc b2 + c 2 + (4.80)
N ( ) = 2 2
ec
b ,
b c b2 +

129 These expressions coincide with those derived from the potential given by MacMillan at page
71.

 Depolarization tensor method 93

bc b2 + c 2 +
N ( ) = 2 2
ec
c . (4.80)
b c c2 +

Taking limit for a = in the second term of eq. 4.66 gives

abc
lim
a
( a + )( b + )( c + )
2 2 2

bc bc (4.81)
= lim = ,
2
( )( ) ( b + )( c + )
a 2 2

a2 b + c +
2
1+

y z
! b +
2 ! c +2
s y (r | ) = , s z (r | ) = . (4.82)
2 2
y z y2 z2
+ +
( ) ( ) ( ) (c + )
2 2 2 2
b2 + c2 + b2 + 2

As coordinate x is absent, the value of is given here by (see eq. A7.23 )

y2 z2
+ 2 =1 (4.83)
b + c +
2

From eq. 4.70,

Tr N ec ( ) = N ec
a
( )+ N ec
b
( )+ N ec
c
( )

bc b2 + c 2 + b2 + c 2 +
= 0+ + (4.84)
b2 c 2 b2 + c2 +
bc abc
= = lim .

( b2 + ) ( c 2 + ) a
(a2 + )(b2 + )(c 2 + )
For b = c the right circular cylinder is obtained. The value of is obtained from

y2 z2
+ = 1, so that = r 2 b2 . (4.85)
b + b +
2 2

The eigenvalues of N should be recalculated from eqs. 4.13 and 4.14,

bc ds b2 ds
N ec ( ) = =
( ) ( )
b
2 b2 + s 3/2
c2 + s 2 b2 + s 2

(4.86)

2 2 2
b 1 b 1b
= 12 = 2 = ,
b2 + s b2 + 2 r 2

where eq. 4.85 was used.
 94 Carlos E. Solivrez

The coefficient and the factors of matrix S are

x
b + = x ,
2 2
b b ! 2
= 2 , s (r | ) = (4.87)
b + r
2
y2 + z2 r

(b + )
2
2

so that for the right circular cylinder of inifinite length
! bc !
next (r ) = N( ) S(r | )
(
b2 + c 2 + )( )
0 0 0 0 0 0
2 2
b b
= 2 0 1/2 0 2 0 y /r 2
2
yz /r 2 (4.88)
r r
0 0 1/2 0 yz /r 2 z 2 /r 2
0 0 0
b2
= 2 0 1/2 y /r
2 2
yz /r 2 .
r
0 yz /r 2 1/2 z 2 /r 2

the very same zero trace eq. 3.58 obtained using Gausss Law.

Summarizing, for the right elliptic cylinder of infinite extension:

0 0 0
! bc ! ec
next (r ) = N( ) S(r , ), N = 0 N b 0 ,
( )(
b2 + c 2 + ) 0 0
Ncec

bc b2 + c 2 + bc b2 + c 2 +
N ( ) = 2 2
ec
b , N c ( ) = 2 2
ec
,
b c b2 + b c c2 +
x
0 0 0
! d +
2
2
S(r , ) = 0 s y s y s z , s (r! , ) =
,
(4.89)
y2 z2
0 s s
s 2z +
( ) (c + )
y z 2 2
b2 + 2

y2 z2
+ = 1, = y,z, d y = b, d z = c.
b2 + c 2 +

 Depolarization tensor method 95

Oblate spheroid
For oblate spheroids a = b, eq. 4.66 gives130
! !
! s(r , ) s(r , )
n (r ) = N( ) a c
ext 2
, (4.90)

a2 + c2 + ( )( )
where
x2 + y2 z2
+ 2 = 1, (4.91)
a + c +
2

N ( ) 0 0
e
N( ) = 0 Ne ( ) 0 . (4.92)

0 0 N p ( )

From eqs. 4.20 and 4.26,

N oe ( ) = N oa ( ) = N ob ( )
a2c c 2 + a2c a2 c 2 (4.93)
= 12 + 12 arctan .

( a c )( a + ) (
2 2 2
a2 c 2
3/2
) c +
2

N op ( ) = N oc ( )

a2c a2c a2 c 2 (4.94)

= arctan .

(a c )
2 2
c2 + ( a c
2 2
3/2
) c +
2

From eqs. and 4.67 and A7.27,

xx + yy zz

!
s (r , ) = a + c +
2 + 2
,
(x 2
+ y2 )+ z2
= 1. (4.95)
x2 + y2 z2 a2 + c2 +
+
(a + ) (c + )
2 2
2 2

Eqs. 4.20 and 4.26 reduce to the formulas for the infinite sheet (see eq. 3.47) for
a = b and to those of the sphere (see eq. 3.66) for a = b = c.

The trace eq. 4.70 is satisfied because

130 MacMillan, p. 62 eq. 39.2.
 96 Carlos E. Solivrez

Tr N( ) = N oa ( )+ N ob ( )+ N ob ( ) = 2N oe ( )+ N op ( )

a2c c 2 + a2c a2 c 2
= + arctan
( a c )( a + ) (
2 2 2
a2 c 2
3/2
) c +
2

a2c a2c a2 c 2 (4.96)

+ arctan
(a c )
2 2
c2 + (a c )
2 2
3/2
c2 +

a2c a2c c 2 + a2c aac

= = = .

(a c )
2 2
c +
2
( a c )( a + ) ( a + )
2 2 2 2
c +
2
(a2 + )(a2 + )(c 2 + )

Prolate spheroid
For prolate spheroids a > b = c, eq. 4.66 gives131
! !
! s(r , ) s(r , )
n (r ) = N ( ) ac
ext p 2
, (4.97)

a2 + c 2 + ( )( )
x2 y2 + z2
where + = 1, (4.98)
a + c +
2 2

N p ( ) 0 0
p
N p ( ) = 0 Nep ( ) 0 . (4.99)

0 0 Ne ( )
p


From eqs. 4.31 and 4.37,
N pp ( ) = N pa ( )

ac 2 ac 2 a2 + + a2 c 2 (4.100)
= + 12 ln 2 ,

(a c )
2 2
a2 + ( a c
2 2
3/2
) a + a c
2 2

N pe ( ) = N pb ( ) = N pc ( )
ac 2 a2 + ac 2 a2 + + a2 c 2 (4.101)
= 12 14 ln 2 .

(a c ) (c + ) (
2 2 2
a c
2 2
3/2
) a + a2
c 2

131 MacMillan, p. 62 eq. 39.2.

 Depolarization tensor method 97

Finally, from eqs. and 4.67 and A7.27,

xx yy + zz
+ 2
! a + c +
2
x2 y2 + z2
s (r , ) = , where + = 1. (4.102)
x2 y2 + z2 a2 + c 2 +
+
(a + ) (c + )
2 2
2 2

The trace eq. 4.70 is satisfied because

Tr N( ) = N pa ( )+ N pb ( )+ N pc ( ) = N pp ( + 2N pe ( )

ac 2 a2 + ac 2 a2 + + a2 c 2
= 12 ln 2
(a c ) (c + ) (
2 2 2
a2 c 2
3/2
)
a + a c
2 2

ac 2 ac 2 a2 + + a2 c 2 (4.103)
+ 12 ln 2
(a c )
2 2
a2 + ( a2 c 2
3/2
)
a + a c
2 2

ac 2 a2 + ac 2 ac 2 acc
= = = .

(a c ) (c + ) (a c )
2 2 2 2 2
a +
2
a + c +
2
( 2
) (a2 + )(c 2 + )(c 2 + )

Eqs. 4.37 and 4.31 reduce to the formulas for the infinite circular cylinder for
a , b = c, and to those of the sphere for a = b = c.

Triaxial ellipsoid
From eqs. A9.3, A9.5, A9.6 and A9.7 the eigenvalues of N() may be expressed in
terms of Legendre's elliptic funcions as follows.

abc
Na ( ) = ( E( ,k)+ F( ,k)) ,
a2 c 2 ( a2 b2 )
abc c 2 +
Nb ( ) =
(a2 + )(b2 + ) (b2 c 2 )
abc a2 c 2 abc
+ E( ,k) F( ,k),
(a2 b2 )(b2 c 2 ) a c2 2 (a2 b2 )
(4.104)
abc b + 2
abc
Nc ( ) = E( ,k),
(a2 + )(c 2 + ) (b2 c 2 ) a c 2 ( b2 c 2 )
2

a2 b2 a2 c 2
where a b c, k = , sin = ,
a2 c 2 a2 +
x2 y2 z2
+ + = 1.
a2 + b2 + c 2 +

 98 Carlos E. Solivrez

 Depolarization tensor method 99

Chapter 5:
Energy, forces and cavities

Thermodynamics of electrostatic and magnetostatic energy

Basic concepts
Thermodynamics studies the macroscopic exchanges of energy (work, heat,
internal and radiation energy...) between a material system and its surroundings,
and the values taken by its state variables in the process. To that end state
functions132 are defined such that for reversible processes its equilibrium
values are a function only of the set of final values of the state variables and are
independent of the path followed to obtain them133. Introductory courses of
thermodynamics discuss only the case of gases, where the state variables are
volume V, pressure P and temperature T. The first two are clearly insuficient for
elastic solids, and the electromagnetic variables have yet to be added.
The performance and interpretation of experiments with electric and magnetic
fields and polarizations require a careful formulation of its thermodynamic
behaviour that is seldom done in introductory texts on electromagnetism. As many
difficulties are found in the way, a short but general overview will be given in this
chapter in order to identify some important features of the problema and write
down without proofthe principal expressions required, identifying its source.
The author is not familiar with any book giving a full analysis of the
thermodynamics of electromagnetism, but there are several studies in
international journals of physics134.
The state variables required for the full thermodynamical characterization of
elastic solids are listed below. They are grouped in pairs ("cause" and "effect"),
through scalar and tensorial material's properties, by the differential relationship
that connects them in the linear range:
c
o temperature T and entropy S, related by the heat capacity c: dS = dT;
T

o mass m and gravitational field !g, related by the space metric;
o strain tensor e and stress tensor , related by the compliance tensor c of elastic
modulus: d = c de;

132 Landau and Lifschitz call them thermodinamic potentials, a name more evocative of its
properties. See pp. 46-96.
133 This characteristic is shared, for instance, by the electrostatic potential .
134 See, for instance, On Magnetic and Electrostatic Energy and The Thermodynamics of
Magnetization by E. A. Guggenheim, , Proc. Roy. Soc. (London) vol. A155, pp. 49101, 1936.
 100 Carlos E. Solivrez

o electric field !E and electric displacement !D, related by the electric permeability

tensor : !D = i E;

o magnetic field !H and magnetic induction !B, related by the magnetic

permeability tensor : !B = i H.
For the magnetic case equilibrium refers to the steady state where electric
currents are constant. It must also be stressed that the formulas given in this
chapter are not valid for dynamic phenomena such as those involved in the
interaction of electromagnetic radiation with matter, processes that are typically
adiabatic.
Irreversible processes are characterized by an irreversible increase of entropy
dS. If dQ is the quantity of heat transmitted in an elementary process, then

dQ = TdS. (5.1)

Conduction currents in metals and magnetic hysteresis are typical

electromagnetic irreversible (dissipative) processes. The atomic and molecular
currents giving rise to the magnetization of solids are not dissipative, nor is the
superconductor case. The theoretical calculation of electromagnetic energy
computes only the reversible work required to stablish a configuration of fields
and polarizations in an ideal non dissipative process. In real processes no matter
how slow you make them irreversible phenomena are always present. Thus,
work dW done on a real system partially dissipates in internal (case of the
electrical resistance) or external irreversible processes. Therefore

dW = dWrev TdSirrev, (5.2)

and the work computed in electromagnetic calculations is usually only the

reversible part dWrev.
The most important thermodynamic state function is the internal energy U that
characterizes the balance between the quantity of heat dQ delivered to the system
and the mechanical work dW done by the system on its surroundings,

dU = dQ - dW. (5.3)

For solids the work expression used for gases, dW = pdV, must be replaced by

dW = c de, (5.4)

to which must be added the contributions of electromagnetic energy.

Newcomers to electromagnetism try to calculate internal energies by using
forces to compute the external work required to build the system. This method,
originated in mechanics, is reinforced in the study of gases, where pressures are
introduced as surface distribution of forces. This is not the case in
electromagnetism, where the starting point for calculating the variations of energy
is not the concept of force or torque. On the contrary as will be seen in this
chapter the latter are evaluated as the rate of change of electromagnetic energy
respect to certain longitudes or angles.

 Depolarization tensor method 101

An important part of the problem is that the variation of electromagnetic energy

depends both on its spatial distribution and its time variation. If at this point the
reader argues that we are considering here only static fields and steady state
currents, some variations should be irrelevant, he should read again the
introduction to eq. 5.2. The electromagnetic energy contained in any distribution
of polarized matter is, by definition, that required to build that distribution from
an initial state where polarizations are zero. This necessarily requires a time
variation of polarizations and fields, process where induced currents make finite
contributions to the energy of the system and generate radiation, however slowly
the process is made. Moreover, fields are not limited to material bodies but extend
throughout all space.
According to the type of process a material system undergoes, it is convenient to
describe it using a peculiar state function different from U. This is done in a way
such that its differential expression does not contain the differential of the state
variable one wishes to keep constant. For instance, U is the more convenient state
function for the description of processes where strains can be fixed, letting
temperature T and stress to take the values that follow from the allowed heat
exchange. Of course, any process may be described by any state function, but the
theoretical analysis is simplified when using, ceteris paribus, an appropriate
thermodynamic potential.

Thermodynamics of electromagnetism
The measurement of electromagnetic properties of polarized bodies is
frequently, though not always, made in experiments where mechanical work is
done. In those cases the only discussed in this bookforces and torques are
related to the variation of parameters characterizing the configuration of fields and
polarizations. In a ceteris paribus fashion, several variables must be kept fixed
throughout the process.
The body's intrinsec geometric configuration the relative distribution of
polarizable matter is one of the variables that must be kep constant. This
configuration is in our case the ellipsoidal shape and semiaxes that determine
the fields's values, usually invariable when de, eq. 5.4, vanishes. As mentioned
before, it is then convenient to use of the internal energy U. In practice, strain's
control is not always possible, as happens in the phenomena of electrostriction135,
magnetostriction136, piezoelectricity 137 and piezomagnetism, that requiere special
treatment.
Electric and magnetic polarizations dependent on the microscopic behaviour
of atoms and molecules are functions of temperature T, a variable that will be
usually necessary to keep constant. As T is a natural variable of U, it is convenient
to define the following thermodynamic potential

F = U TS, (5.5)

135 Landau and Lifchitz, p. 55.
136 Landau and Lifchitz, p. 155.
137 Landau and Lifchitz, p. 73.
 102 Carlos E. Solivrez

called Helmholtz free energy, for which

dF = TdS dW TdS SdT = dW SdT. (5.6)

The external work done in an isothermal process is then

dW = dWrev TdSirrev. = dF. (5.7)

Any system tends to reorganize it elements in order to to minimize its energy in

the final state of equilibrium. When there are no external constraints, F varies
until reaching a minimum dF vanishes and the system ceases to do external work.
The reversible work delivered in the process of reacommodation is

rev rev
W = dW = (F T Sirrev. ), (5.8)

reason why F characterizes the maximum quantity of work that it is possible to

extract from a system138,139, concept that in chemistry is called afinity.
The internal energy U and the free energy F include an electromagnetic
contribution. For F its value Fem is the amount of reversible work necessary to
build up the final configuration of charges, electric currents and polarized matter.
For the linear case (whose description requires several anisotropy tensors of rank
two) its expression in terms of fields is140

1 ! ! ! ! 3 1 ! ! ! ! 3
Fem
e=const .
=
2 E( r )i D( r )d r +
2 ' B( r )i H(r )d r , (5.9)
V V

where the integral is taken over all the regions containing the sources of the fields,

the ones where ! i D and ! H are non-vanishing. There are no general
expressions for the non-linear case, where each specific situation has to be dealt
with. For conductors or superconductors the surface density of charge or current
should be made explicit through appropiate mathematical transformations.
Some authors extend integrals to all space, in accordance to the concept that
energy is contained in the fields. Although this seems to modify the value of Fem, it
is not so because there are partial cancellations of the contributions of the fields
inside the body with those outside it141. Special care should be taken with the
calculation of the work necessary to polarize matter with fields assumed to be
fixed, as sources provide o receive energy in order to keep fields constant when
matter is introduced, even for very slow process142.

138 Panofsky and Phillips, p. 90.
139 Landau and Lifchitz, pp. 52-55, 129-131.
140 Reitz, pp. 120 and 254.
141 Stratton, pp. 112-113, discusses this cancelation but gives no actual examples. See Problem 36.

142 See, for instance, V. Gilinsky and D. Holliday, Interaction Energy of a Dielectric in an Electrostatic
Field, Am. J. Phys. vol. 34, pp. 1134-1138 (1966).

 Depolarization tensor method 103

Fem does no include the contribution from bodys changes of shape, which is
usually describe using the thermodynamic potential called Gibbs free energy143

d = U TS + ed , (5.10)

where work includes the variations of e and stress takes the final value
corresponding to the isothermal process. The variation of Gibbs free energy can be
measured in actual experiments, but is more difficult to calculate than Helmholtz
free energy.
For the formulas given in this chapter processes are assumed to be isothermal
and Helmholtz free energy F is evaluated assuming a constant strain tensor e. For
instance, actual experiments should provide an efficient mean of fast transference
of any quantity of heat generated, so that the isothermal condition is fulfilled.
These two conditions are seldom fulfilled in irreversible phenomena as hysteresis
and in piezoelectric and piezomagnetic processes that modify significantly the
dimensiones of the crystal lattice, creating also additional fields. The use of the
depolarization tensor method for these kind of phenomena as well of others
peculiar of single crystales will probably requiere modifications or restrictions
that have not been studied by the author144.
Depending on the kind of anisotropy, it may be convenient to use coordinate
axes different than the ellipsoids semiaxes where N is diagonal. The goal is to
simplify the inverse matrices required for the case. This use of a coordinate system
fixed to the body should take into account the case of rotating samples depicted in
Figure 21, both for rotations and torques.

Solved problems related to the energy of polarizable ellipsoidal bodies

immersed in applied fields are solved in Chapter 6. They are only selected
examples of the way in which the depolarization tensor method may simplify the
anlisis of some experiments.

Anisotropy energy
A body under the action of an applied field experiences a torque if its energy
depends of the orientation of the field respect to the body. A sphere of untextured
0
policrystalline iron has minimum free energy when !M and !H are parallel. For any
other orientation, the bodys magnetic dipoles (that is, !M ) experience a torque
than tends to align them with the applied field. This does not generate a torque on
the body when the material is isotropic, so that the magnetization may freely
orientate with no expense of energy apart from disipative effects. The sphere will
remain in equilibrium at any orientation respect to the field because the
magnetization will follow the field, not the body.

143 Landau and Lifchitz, p. 62.
144 Nye, chapter X pp. 170-191, makes a detailed discussion of the equilibrium thermodynamics of
physical properties of single crystals. His figures 10.1a and 10.1b give a concise pero quite
illustrative summary of the problem.
 104 Carlos E. Solivrez

The direction and magnitude of polarization and therefore the free energy
may depend on the bodys orientation respect to the applied field. This anisotropy
of energy has two different origins, shape and cristalline effects.
Shape anisotropy of energy145 originates
from the effect of the bodys shape on the
internal field and polarization. For
ellipsoidal bodies the effect is fully described
by the internal depolarization tensor N. For
the family of ellipsoidal bodies, the only one
without shape anisotropy is the sphere
where the three eigenvalues are equal (see
eq. 3.66). For the rest, the energy required to

induce polarization along two different
directions within the ellipsoid is in general Figure 18. Deviation of the internal
different. field respect to the applied one
The general equations that describe the phenomena of induced polarization are
given in section 2.85 and reproduced below. The upper index that identifies the
internal field has been suppressed because it will be the only field of interest in
what follows.
F = F0 N i Q(F), Q(F) = i F,
(5.11)
( ) ( )
1 1
F = 1+ N i i F0 , Q = i Fint = i 1+ N i i F0 .

Lets analyze, then, the behaviour of an isotropic triaxial ellipsoidal body on the
plane xy where the principal values of the internal depolarization tensor are Na and
Nb. The field and the polarization are exclusively those induced by the fix applied

field!F 0 . In the bodys principal system of coordinates the components on the plane
of interest are
1
Fy = y Fy0 , Fz = z Fz0 , = . (5.12)
1+ N

The deviation angle of the induced field respect to the applied one (see Figure
18) can be found from the scalar product
! !
F i F 0 = cos F F 0 = Fy Fy0 + Fz Fz0
(5.13)
( ) ( ) ( )( )
2 2 2
= F 0 + z Fz0 = y sen2 + z cos2 F 0 .
y y
The modulus of the internal field differs from the applied one by a factor
depending on the orientation of the latter respect to the body:

( F ) + ( F )
2 2
F = F 2y + F 2z = y
0
y
0
z z z
= 2y sen2 + 2z cos2 F 0 . (5.14)



145 Brown (1963), p. 106.

 Depolarization tensor method 105

The deviation angle is then given by

! !
F i F 0 y sen + z cos
2 2

cos = = . (5.15)
F F 0 y sen + z cos
2 2 2 2

0
!F and !F are parallel only along the ellipsoids ejes principal axes, where its
magnitude differs by the factor . Due to shape anisotropy the minimum free
energy is obtained when the bodys largest or smaller semiaxis depending on
the kind of polarization, as discussed in next section is oriented along the
applied field. If this is not so, there is a torque that tends to orientate the body
along that direction.
Crystalline anisotropy energy is originated by the spatial order of the materials
atoms and molecules. A simple example that illustrates the phenomena is the two
polarizable molecules analyzed in the section Induced electric polarization of two
interacting atoms. The only materials with crystalline anisotropy are single
crystals or textured policrystalline ones. In the latter the small crystals have some
of its faces preferently oriented along certain direction, as happens in laminated
metals. Powders, untextured policrystalline materials and amorphous solids have
no macroscopic order and, therefore, no crystalline anisotropy energy.
For the case of induced polarization the only one discussed herethe
simmetries that give origin to crystalline anisotropy are reflected in the non-scalar
anisotropic susceptibility tensor. The main properties of this tensors are given in
Appendix 5. Its demonstration, not given there, can be made in similar way as for
the case of the depolarization tensor (see section Symmetries) but requires a
knowledge of crystalline structures that cannot be assumed for the readers of this
book.

Origin of torques exerted on a body

In this section it is discussed the origin of the moment exerted on an ellipsoid
outside the equilibrium orientation. The expression of the par resultante for each
case is discussed in other sections.
At Figure 19 an example is given of a common experiment where torque
appears. A prolate spheroid, ellipsoidal shape commonly used in experiments, is
suspended so that it can rotate around an axis normal to its symmetry axis x
(corresponding to the largest semiaxis a). The rotation axis z is outgoing from the
back of the figure. The minimum value of the isothermal and isobaric free energy
Fe is assumed to be obtained when the direction of the largest semiaxis a coincides
!
with that of the field F0 . If that semiaxis is initially an angle apart from the field,

a couple is exerted on the body, such that tends to diminish it.
The coupless origin is a force distributed over the bodys volume that, due to
the constraint imposed by the rotations axis, manifests as a moment of force along
that axis. In the given configuration, where the plane xz normal to the page is a
symmetry plane of the spheroid, the force over each differential element of volume
has no component normal to plane yz. The appearance of this force is a common
experimental fact for all potential energies as Fe. This is similar to what happens
 106 Carlos E. Solivrez

with a mass, where the that tends to bring it to the position of minimum
gravitational potential energy is the force distributed over all its volume that we
call weight. It is discussed next the relationship between the force couple and the
variation of the free energy.

Figure 19. Torque exerted on ellipsoid in applied field.

At the right of the figure! f (r ) is the force over the element of volume dV at !r ,
where r is the distance to the rotation axis z. Angle is the usual one of spherical
coordinates and the force has been decomposed in its tangential and radial
components f and fr. The work dW done on the body when it rotates d towards
the direction of stable equilibrium is

dW = d f r d 3r = d = dFe , (5.16)
V

where is the torque exerted on the body along its suspension axis (see, for
instance, eq. 5.23). Therefore
dF
= e . (5.17)
d

Torque exerted on an ellipsoid and the state of equilibrium

In what follows a discussion is made of the problem of equilibrium and the force
couples exerted on ellipsoidal bodies made with different kinds of materials and
subject to static electric and magnetic fields.

Dielectrics

Permanent polarization
The analysis is made here of the equilibrium configuration of a permanent
polarization in the absence of applied field. In such case the electrostatic
contribution to the energy, eq. 5.9, becomes (see eq. 2.12)

 Depolarization tensor method 107

V !P ! !
Fe =
2
(
E i 0E P + P = )
V

2 0
N (1 N )P 2 . (5.18)

As an illustration, the minimization of energy will be analized for the case of
spheroids, where

Fe =
V
2 0
( )
C e P 2e + C pPp2 , where
(5.19)
( )
C = Ne 1 Ne , N p = 1 2Ne , C p = 2Ne 1 2Ne ,
e
( )
where subindices p and e identify the polar and equatorial components or
eigenvalues, respectively and the trace rule eqs. 3.15 was used. Figure 20 shows
their values in a range that includes the right circular cylinder of infinite length
(point A) to the sphere (point B). The sheet of constant thickness and infinite
extension, not seen here, has the Np = 1. The vertical line at Ne = 1/3 is the
boundary of the regions of prolate and oblate spheroids.

Figure 20. N , C y C as functions of N .
! !

For oblate spheroids the largest coefficient is Cp so that free energy reaches its
maximum negative value when the polarization lies on the plane of largest
semiaxes, the equatorial ones. For prolate spheroids the largest coefficient is Ce
corresponding to the polar component (here the largest semiaxis), and the
polarization tends to align along that direction. For the sphere the polarization has
no prefered alignment because the coefficientes are the same for all components of
polarization.
 108 Carlos E. Solivrez

The total free energy is not the expression customarily used for this sort of

problem as it may be simplified eliminating the !E P i E P (see, below, the analogous
case of permanente magnetization), pero the obtained result is correct.

Induced polarization

According to Stratton146 , the torque exerted in the linear case by the fixed
0
field !E on an isotropic dielectric body when its orientation is not the one of
minimum free energy, is obtained from the following interaction energy,

! ! V V
( )
2
F' e = 12 P i E 0d 3r = 0 e Eint E0 = 0 e E0 ,
V
2 2
(5.20)
1
where = ,
1+ e N

where eq. 2.33 was used. The moment of distributed forces is then147 (see eq.
2.34):
! ! ! ! !
( )
1
= p E 0 , p = 0 e i E 0 , e = V e i 1+ N i e . (5.21)

For an isotropic dielectric

1
0 0
1+ e Na
1
( )
1
e = V e 1+ N i e = V e 0 0 . (5.22)
1+ e Nb
1
0 0
1+ e Nc

When the applied field has no z component, as in Figure 19, the only torque
component is
Nc Nb ! !
z = 0 2eV sin2 E 0 i E 0 . (5.23)

1+ e Nb 1+ e Nc ( )( )
The general case of spheroids is discussed at Problem 38. For triaxial ellipsoids
with semiaxes order a > b > c, the order of the principal values is Nc > Nb and the
force couple has positive sign, corresponding to a rotation that tends to align the
largest semiaxis with the field. This orientation is the one of minimum energy, as
may be verified from the expression of F 'e initially written.

146 Stratton, p. 113 eq. 51.
147 Compare with Stratton, p. 216 eq. 53.

 Depolarization tensor method 109

Anisotropic susceptibility
The electric susceptibility of uniaxial crystals has in its principal system of
coordinates, that in general does not coincide with that of the body the following
expression (see Appendix 5):


0 0

e = 0 0 . (5.24)

0 0 !

The microscopic origin of this anisotropy is not discussed here, but has a similar
origin as that of the two polarizable atoms analyzed at page 21, with the additional
complexity introduced by the regular order of a crystal lattice.
The force couple exerted on a spherical body of that material that is, without
shape anisotropy does not determine the two components but its difference, as
discussed next
From eqs. 2.33 and. 3.66, the components of the internal field and the
polarization in the principal system of coordinates of the susceptibility tensor are:

E 0 E 0
E int
= , P = E int
= . (5.25)

1+ /3
1+ /3


Because of the rotational symmetry of the susceptibility, the experiment must be
done on the plane that contains the symmetry axis and the applied field. When
descomposing all vectores in their components parallel and perpendicular to that
axis, the internal field is give by

E 0 E 0!
E int
= , E int
= . (5.26)

1+ /3 !
1+ ! /3

Upon replacement in the expression for the free energy eq. 5.20 it is obtained

1 ! !0 3 0
F 'e = P i E d r = V Eint E0
2 V 2

( )
( ) (5.27)
( ) ( )
2 2 2 2
0 0
0 E 0
E
= 0 V E 0
E .
= V
" "
+ " +
2 1+ /3 1+ " /3 2 1/ +1/3 1/ " +1/3

F 'e is minimum along the directions with maximum susceptibility. For >
!
the minimum is obtained when the symmetry axis is parallel to E 0 . For < ,
!
the minimum is obtained when the symmetry axis is normal to the field. Although
at first sight this appears indistinguisable from a shape anisotropy, it is not so (see
Problem 02). The calculation of the torque exerted by the applied field on the body
is done at Problem 39.
 110 Carlos E. Solivrez

Magnetic materials

Permanent magnetization
The magnitude of the spontaneous magnetization is usually constant at constant
temperature, independently of its orientation respect to the crystal axes148,149. One
may then compute the energy of spontaneously magnetized matter assuming
invariable the magnitude of the microscopic magnetic moments. This calculation
which does not include the energy necessary for building up the magnetic
atomic and molecular moments gives150,151

! ! ! !
Fm = 0 H int (r )i M(r )d 3r . (5.28)
2 V

For a uniformly magnetized ellipsoidal body eq. 2.44 gives

! int !
H = 'N i M. (5.29)
Therefore
0 ! ! 0 ' ! !
Fm = V H int i M = V M i N i M, (5.30)
2 2

expression that should be minimized with the condition of constant M. In the

principal system of coordinates of N the spherical components of !M are

M = M sen cos , M y = M sen sen , M z = M cos . (5.31)

x
Therefore
0 '
Fm = V N M 2
2 (5.32)
0 '

=
2
( )
V N x sen2 cos2 + N y sen2 sen2 + N z cos2 M 2 ,

should be minimized.
The minimum is easily found realizing that the parenthesis is the parametric
equation
X = N x sen cos , Y = N y sen sen , Z = N z cos , (5.33)

of the ellipsoid
X2 Y2 Z2
+ + = 1, (5.34)
( ) ( ) ( )
2 2 2
Nx Ny Nz



148 Kittel, p. 533.
149 Landau and Lifchitz, p. 146.
150 Chikazumi, p. 24 eq. 1.95.

151 Stratton, p. 130 eq. 63.

 Depolarization tensor method 111

whose square radius vector R152 has to be maximized. The minimum value of R2 is
obtained along the axis corresponding to the smallest denominator, that is, the
direction of the ellipsoids largest semiaxis (see eq. 4.6).
This property of permanently magnetizated prolate spheroids explains the
stability of the magnetization of a compasss needle . In this case there is no torque
applied on the body, but an irreversible rotation of magnetization (hysteresis) that
aligns it in the direction of minimum isothermal and isobaric free energy. Such a
body has a permanent magnetic dipole moment given among eqs. 2.49 and its
direction is that of the polar semiaxis. At constant temperature the magnetization
of the spheroid may be considered to be fixed because in the presence of low
magnetic fields like the terrestrial one, the induced magnetization is much lower
than the spontaneous or permanent one153. For such a case154 the free energy of
interaction with an applied field and the force couple are
! ! ! ! !
Fm = mi B 0 , = m B 0 . (5.35)


These expressions do not required further anlisis, as Fm is a minimum when !m
0
has the same direction and sense than!B , orientation generated by that vanishes
when both vectors are parallel.

Induced magnetization
For the calculation of the interaction energy of spontaneous electric or magnetic
polarizations it is not necessary to include the energy of its creation. It is not so for
induced polarizations, where the required expression of the free energy is155

0 ! ! ! ! 0 ! ! 1! !
Fm = H 0(r )i M(r )d 3r = V M i H 0 = mi B 0 , (5.36)
2 V 2 2

because both vectors are uniform (see page 6).

The torque exerted by the magnetic induction field over the ellipsoidal body is
then given by (see eq. 2.49)

! ! ! !
! 1
( )
1
= m B 0 , m = m i H 0 , m = V m i 1+ N i m . (5.37)
'

152 Radius vector is the distance of a surfaces point to its center.
153 Chikazumi, p. 439.
154 Jackson, p. 150. The factor 1/2 is not present when the moment and the field are rigid or
fixed.
155 Stratton pp. 128 eq. 51 and 154 eq. 7. An illuminating discussion of the subject is made by M. S.
Plesset and G. Venezian (Am. J. Phys. vol. 32, pp. 860-864, 1964), where the calculation is made
of the torque applied on an ellipsoidal body with induced magnetization. See also S. P. Puri, Am.
J. Phys. vol. 33, p. 472, 1965.
 112 Carlos E. Solivrez

Magnetic torque experiments

Figure 21 shows the main parts of
a goniometer, device that can measure
with good precission the force couples
exerted by an applied magnetic field
over a magnetized body. The
instrument is based on a calibrated
torsion wire that can be rotated in
order to return a reflected light beam
to its zero (position in the absence of
applied field). The glass tube blocks
air currents that may displace the
body. The paddles immersed in oil are
adjusted in size and shape in order to Figure 21. Elementary measurement
provide critical damping for a fast of magnetic force couples.
return to zero.

For the isotropic case the force couple reduces to

! ! ! 0 ! !
( )
1
= m B 0 = V mH 0 i 1+ mN i H0. (5.38)
'

In the ellipsoids principal system of coordinates the components are given by

0 2mV Nc Nb 0 0
x = H H , (5.39)

( m b )( (
' 1+ N 1+ N y z

m c ))
0 2mV Na Nc 0 0
y = H H , (5.40)

( m a )( (
' 1+ N 1+ N x z

m c ))
0 mV Nb Na 0 0
z = H H . (5.41)

( m a )( (
' 1+ N 1+ N x t

m b ))
Both for the paramagnetic (m > 0) and the diamagnetic (m < 0, | m | 1) case
denominators are positive numbers and the sign is determined by the difference of
principal values. When the applied field has only y, z components, the treatment is
similar to that given for the electric one, eq. 5.23. The torque tends then to align
the largest semiaxes (a in our convention) with the applied field, both in the
paramagnetic and diamagnetic case.

Conductors
The energy of a surface charge distribution on a conductor with net charge Q is
given by the next equation156.


156 Stratton, p. 107 eq. 16.

 Depolarization tensor method 113

1 ! 1 ! 1
Fe =
"
2 S 2
(r )d 2r = S " (r )d 2r = SQ = 0.
2
(5.42)
S

The result is obtained by computing the work done by bringing the charges
from infinite distances to a conductor at potential S until the total charge Q is
build up. The formula is not valid for the separation of charges over the surface of
an ellipsoid with zero net charge. Classical texts on electromagnetism, like those of
Stratton and Jackson, restrict their analysis to conducting bodies with non
vanishing net charge where the previously given shape-independent formula is
valid. The exception is that of Landau and Lifshitz, who compute the following
value for the interaction energy of a zero net charge conductor with an applied
electrostatic field:
! !
Fe = 12 p i E 0 , (5.43)

!
where p is the electric dipole moment of the body. For ellipsoidal conductors this
moment is given by eq. 2.60, so that in the principal system of coordinates

! ! V ! ! V
Fe = 12 E 0 i p = 0 E 0 i N 1 i E 0 = 0
2
1 0 2
E ,
2 N
( ) (5.44)

coinciding with eq. 2.14 of Landau & Lifshitz.
The expression for the force couple exerted over a conducting ellipsoid is the
one for a dielectric, eq. 5.21, in the limit of infinite susceptibility (see Problem 08).
It is thus obtained

V Nc Nb 0 0 V Na Nc 0 0 V Nb Na 0 0
x = 0 E y Ez , y = 0 Ex Ez , z = 0 E E . (5.45)
NbNc NaNc NaNb x y

When the three depolarization factors are equal that is, for spheres the
force couple vanishes, as corresponds to its shape isotropy. When all three are
different, as in triaxial ellipsoids, the force couple tends to align the largest
semiaxis with the field, as happens in the dielectric and magnetic case.

Superconductors
As in the case of conductors, the force couple exerted by the applied field on a
superconducting body may be derived from the case of induced magnetization,
eq. 5.37, using the condition of perfect diamagnetism m = 1 (see eq. 2.73).
!

! ! ! !
! 1
( )
1
= m B 0 , m = s i B 0 , s = V 1 N . (5.46)
0 '

A typical component is
V Nz N y 0 0
x = B y Bz , (5.47)

( )(
0 ' 1 N y 1 N z )
 114 Carlos E. Solivrez

where the other components are obtained from a cyclic permutation of indices.
For finite volume ellipsoids the denominator is always positive and the
behaviour of a superconductor body is the same as that of a magnetized one, with
the minimum energy obtained when the largest semiaxis is aligned with the
applied field. Degenerate ellipsoids as the sheet of infinite extension and the
ellyptic cilinders are not valid cases for the analysis of torques that there diverge
with volume.

Force on an ellipsoid in a non-uniform field

The depolarization tensor method has few advantages for the treatment of the
forces exerted by non-uniform fields on ellipsoids. Nevertheless, due to its
experimental relevance, the expressions are given of the forces exerted on
polarized dielectric and magnetic ellipsoids in quasi-uniform fields157,158:

! ! ! ! ! ! 0 ! 0 ! ! !
( ) !
(
f e = V P i E 0 (r ) = p i E 0 (r ) =

) ( )
E (r )i e i E 0(r ),
(5.48)
! ! !0 ! ! 1 !0 ! !0 !
( ) !
f m = V M i B (r ) = mi B (r ) = (
0 !

0 '
) ( )
B (r )i m i B (r ).

where the bodys polarizability tensors, eqs. 2.33 and 2.49, have been used.
The variation of the field inside the body should be small enough so that the
depolarization tensor method is valid within the acceptable range of error. At the
same time, the variation should be large enough so that the total force is
measurable. Experiments of this type159 make possible measurements of
diamagnetic susceptibility that are imposible with the torsion method (see
Magnetic torque experiments at page 112). This is so because of the small value of
this kind of susceptibility and the fact that torques are proportional to its square;
force, on the other hand, is directly proportional to the susceptibility.

Infinite and infinitesimal bodies

Equation 3.80 shows that similar ellipsoids have equal internal depolarization
tensors N. Therefore, if one makes all semiaxes grow without limit preserving their
relative values, N does not change. One must then be careful with the use of bodies
of infinite or indefinite extension, subterfuge often used for the introduction of
dielectric or magnetic material surrounding finite bodies, unless one specifies their
shape. These idealizations are justified only if they provide a good approximation
to a real situation, that usually involves finite bodies. The sphere of infinite
diameter can be used to approximate an isotropic dielectric environment, but one
has still to discuss the effect of the cavity that contains the finite body of interest.
The elliptic cylinder of infinite length provides a good approximation to the more


157 Landau and Lifchitz, p. 72 eq. 16.12.
158 Landau and Lifchitz, p. 144 eq. 34.8.
159 Faraday's method, see H. Zjilstra, Experimental Methods in magnetism, vol. 2, North Holland
Publishing, 1967, p. 94.

 Depolarization tensor method 115

complex calculation of the field near the equator of a triaxial ellipsoid, but there
are better aproximations than the sheet of infinite extension to the field near the
poles of a very flat oblate spheroid (see eq. 4.29).
Polarized matter creates fields in two different ways: by the space variation of
polarization inside the bodys volume and by the step discontinuity through the
bodys surface160. For uniform polarizations as the ones studied here, only the
second effect appears, as illustrated for the electric case by eq. 2.51,

! ! !0 !
(r ') 2
S r! r'! d r'.
E(r ) = E k1 "

The contribution to the field of surface element dS is dS/r 2 = d (notice that
the gradient increases in 1 the exponent of the denominator), where d is the
subtended element of solid angle. Therefore, the contribution of a surface sector to
the field at an internal point is proportional to the subtended solid angle. This is
illustrated by the infinite sheet and elliptic cylinder, where the depolarization
factors (principal values) vanish for the directions where the solid angle tends to 0
(the semiaxis tends to ). That is, very large distances do not ensure very small
contributions from the surface sector to the field, that depends on the solid angle
subtended by that sector. Although at large distances r the field of a finite polarized
body, whatever its shape, has a distance dependence of r -3 (see section Near and
far away point approximations to next), far away surface charges do not have this
behaviour.
From the mathematical point of view the problems with bodys of infinite
extension originate when the value of the integral

d 3r '
V
! !
rr'
(5.49)

is computed by extending V over all space. A careful analysis shows that the
integral is semi-convergent, that its value when volume V grows depends on the
peculiar way the infinite limit is taken161. An infinite number of different limits
may exist depending on the shape of V, as illustrated by the property eq. 3.80 of the
depolarization tensor.
The same error of assuming all infinite bodies to be equal has been made with
infinitesimal ones. The Fermi contact term is a little known contribution to the
energy of atomic electrons originated in its interaction with the nucleus. Its
calculation was made considering the nucleus to be a point charge, implicitly
spherical. The term contains an integral that this author proved to be of the
internal depolarization tensor type162, thus requiring a better specification of the
nuclear charge distribution. Similar errors are repeatedly made in the calculations


160 Reitz, p. 79 eqs. 4.13 and 4.14.
161 MacMillan, pp. 163-165.
162 C. E. Solivrez; The contact hyperfine interaction: an ill defined problem; J. Phys. C: Solid St. Phys.

vol. 13; pp. L1017- L1019; 1980.

 116 Carlos E. Solivrez

of microscopic fields in crystals when series of dipolars are numerically summed

withour analyzing its kind of convergence.

Cavities
Standard treatment
Cavities have had a distinguished role in the theory of electricity since Lord

Kelvin used them to define !E and !D 163. The problem is that they are treated as
ellipsoidal bodies with their same properties164, or as if the extraction of an
ellipsoidal piece of a uniformly polarized body dos not modify the uniform internal
fields. Stratton, without previous discussion, treats cavities as if they were bodies
of the same shape with the dielectric permittivity of vacuum165. Another reputed
author is more explicit when he states166:
Finally we consider the field inside a cavity in a material magnetized to
an intensity I (Fig. 1.16). The free pole distribution on the surface of the
cavity is the same as that on the surface of a solid body with the same
shape as the cavity, and with the same magnetization as the material
surrounding the cavity, except that the poles are of opposite sign. This
must be true, because if we superpose the body and the hole, we
have a uniformly magnetized solid without free poles.
As previously stated the argument in bold type is true for uniform polarization,
where the only contribution to the field comes from surface polarization
charges167. But this can only be equal to the polarization on the cavitys walls when
the polarization is fixed (rigid, as some authors call it), which is false.
Jackson explicitly solves the spherical cavity168 but he replaces in the solution
for the sphere the electric permittivity by its inverse with the pseudo-argument:
In fact, inspection of boundary conditions (4.56) shows that the results from the
cavity can be obtained from those for the sphere by the replacemente 1/ .
The alluded boundary conditions are the continuity of the tangential component of
the electric field through the cavitys surface and the step discontinuity of its
normal component due to the polarization charges eq. 3.35. In addition, he does
not specify the shape of the dielectric body where the cavity is made.
Van Vleck169, a reputed specialist on the properties of polarized matter, makes
use of cavities for the calculation of local fields, but his cavities are virtual, not real,
a mere artifact for promoting fruitful discussions. This is the case of the so called
Lorentz sphere, cavity used to estimate the internal or local electric field at the

163 See, for instance, Stratton, p. 214; Reitz, p. 82.
164 Stratton, pp. 206 and 213.
165 Stratton, p. 206 eq. 31.
166 Chikazumi, p. 16.
167 Reitz, p. 79 eqs. 4.13 and 4.14.
168 Jackson, pp. 114-115.

169 Van Vleck, chapter IV, analizes classsical calculations of fields inside cavities.

 Depolarization tensor method 117

molecules of a dielectric170, magnitude different from the macroscopic field in

Maxwells equations. The cavity is not empty but filled with molecules whose
field and polarization are to be computed individually according to their known
crystalline positions.
The analysis made at section Induced electric polarization of two interacting
atoms illustrates the mutual influence of different portions of matter on the final
polarization state. Polarization is not automatically uniform, nor it remains forever
fixed when this state is achieved. Shape efects are crucial and only closed single
surfaces of the second degree (ellipsoids) have been proved to be amenable of
uniform polarization. It is therefore clear that the following two aspects have to be
carefully analyzed:
1. The effect of the removal of material inside the ellipsoidal body on the
polarization state of the rest of the body.
2. The effect of the polarization charges or currents on the cavitys walls over the
fields inside the cavity and the remaining polarized matter.
It should also be stressed that a uniform applied and total field inside the body
is required, but this is a necessary condition and not a a sufficient one.

Some specific cavities: or homoeoids

A cavity is the result of the remotion of a volume V0 of polarized matter from a
larger one V. The simplest assumption is that both pieces have ellipsoidal shape, so
that they can, in principle, be uniformly polarized. The integral from which the
depolarization tensor is to be calculated (see the definition eq. 3.4) is

! d 3r ' d 3r ' d 3r ' ! !

I '(r ) = ! ! = ! ! ! ! = I(r ) I0(r ), (4.50)
V V0 r r' V r r' V0 r r'

where a well known mathematical property of integrals was used. Using the
definition eq. 3.4 one may derive from it the following mathematical relationship
! ! !
n'(r ) = n(r ) n0(r ), (5.51)

where n is the depolarization tensor of the ellipsoidal body of volume V and n0 the
one of V0. This seems to reduce the problem of fields to a an algebraic combination
of known cases, but it is not so. The superposition principle applies to rigid
distributions of charges, dipoles or currents, but not to induced polarizations.
Although this should be clear enough, it is best to solve some specific case to
illustrate the argument, what is done next.
Shells or homoeoids are one of the most symmetric case of cavities, when the
removed ellipsoidal region V0 is concentric to a similar ellipsoid with proportional
corresponding semiaxes. If the first of the equations below describes the bodys
surface and the second one the surface of the cavity, they determine an homoeoid
or shell:


170 Dekker, pp. 141-144.
 118 Carlos E. Solivrez

x 2 y2 z2 x 2 y2 z2
2
+ 2 + 2 = 1, 2 + 2 + 2 = k 2 , where 0 < k < 1. (5.52)
a b c a b c

Homoeoids have been intensively studied in the theory of gravitational171 and

electromagnetic potentials because they have the interesting property that the
potential is constant inside the shell, that is, internal fields vanish as inside
conductors but this may not apply to polarization fields.
Problem 28 solves the simple case of an spherical shell using the electrostatic
!
Gausss Law. The values of the auxiliary function f (r ) given by eq. 6.72 show that in
none of the three regions depicted by Figure 24 it is a quadratic function of
coordinates. Therefore, nowhere may a depolarization tensor be defined and,
presumably, no uniform internal field and polarization may be found. This does not
rule out the possibility that the field inside the cavity is uniform, it only shows that
the depolarization tensor method is not the appropiate one to make the
calculation.
If its cavities are small enough, the polarization of an allipsoidal body will have
small departures from that calculated from its depolarization tensor. The reason is
that the change in the field external to the cavity is proportional to its dipole
moment, that is to the cavitys volume. The field around the cavity, on the other
hand, would be very different from the one in its absence. The most common error
made with the treatment of cavities is the assumption that the polarization of the
surrounding matter remains inalterable, fixed, which is false without doubt.

Thin shells and the dipole layer

The lack of uniformity in the bodys interior does not appear in an
infinitesimally thin shell, where the body is reduced to a double sided surface, case
that is now analyzed. The case of the spherical homoeoid is analized here (see
problem Problem 28), where for the three different regions it is obtained

2 f1
= 0 for 0 r R1 , (5.53)
x x

2 f 2
x x

1 R3 3x 2 r 2 x y x z
1
+ R13 R13
3 3 r5 r5 r5 (5.54)
yx
3
1 R1 3 y 2 r 2 y z
= R13 5 + R13 5 for R1 r R2 ,
r 3 3 r5 r
zx z y
3
1 R1 3z 2 r 2
R13 5 R13 5 +
3 3
r r r5

171 A. Gray, On the Attractions of Spherical and Ellipsoidal Shells, Proc. Edinburgh Math. Soc. vol. 23,
pp. 91-100 (1913-1914).

 Depolarization tensor method 119

2 f 3
x x

3x 2 r 2 3x y 3x z

r5 r5 r5 (5.55)
3
R2 R1 3 y x
3
3 y2 r2 3 y z
= for r R2 .
3 r5 r5 r5

3z x 3z y 3z 2 r 2

r5 r5 r5

Taking the limit R2 R1 in eqs. 5.54 and 5.55 it is obtained

x2 xy xz

R12 R12 R12
yx y2 yz
ncav = 0, N = = s (r! S1 )s (r! S1 ), next = 0. (5.56)
R12 R12 2
R1

zx zy z2

R12 R12 R12

The expression, valid for all thin ellipsoidal homoeoids (see Problem 42), shows
that the depolarization tensor of a thin shell is reduced to the discontinuity
through the ellipsoids surface eq. 3.38. Its utility has yet to be asessed.


 120 Carlos E. Solivrez

 Depolarization tensor method 121

Chapter 6:
Selected problemas

Electric polarization
Problem 01: Autoconsistent electric polarization of two atoms
Two identical and isotropic atoms, with no permanent electric dipole moment,
are a distance d apart and immersed in a uniform electric applied field. Asumming
that the induced dipole moments are proportional to the total electric field at each
atom, calculate their values using the point dipole model.
Solution
See section Induced electric polarization of two interacting atoms.

Problem 02: Shape or crystalline anisotropy?

In a laboratory, a student made an experimental determination of the two
coefficients c c that relate the electric polarization with the applied field for an
!
ellipsoid:

P = c E 0 , P! = c!E!0 . (6.1)
0 0

The absent minded student that made the measurements forgot to register to
which of the two following bodies they correspond: 1) an isotropic dielectric
spheroid; 2) a uniaxial dielectric sphere (see Appendix 5 at page 165). A large
amount of time, work and resources were used, so it is highly desirable to find a
way to identify the right body. Is it possible to differentiate between the two cases
without repeating the measurements?
Solution
In the principal system of coordinates of a spheroid, the electric polarization of
an isotropic material with susceptibility placed in a uniform field has the
following components (see eq. 2.33)

P = E 0 = c E 0 , P! = E 0! = c!E 0! . (6.2)
1+ N 0 1+ N!
0

As the values N ,N of the spheroid may be easily computed, each equation

!
provide a value of ,
c c!
= , ! = . (6.3)
1 N c 1 N!c!

 122 Carlos E. Solivrez

If = ! = the problem is solved.

Otherwise the expressions for a uniaxial dielectric sphere have to be solved. In

the susceptibility tensors principal system the equations are

!
P = E 0 = c E 0 , P! = E 0! = c!E 0! . (6.4)
1+ /3 0 1+ ! /3
0
The eigenvalues are
c c!
= , ! = , (6.5)
1 c /3 1 c! /3

where the coefficients c cannot be equal. The susceptilitys eigenvalues have the
same order as the coefficients.
Therefore, the shape anisotropy of a spheroid has some characteristics that
resemble the dielectric anisotropy of a sphere, but others are different.

Problem 03: Dielectric sphere

Give, in SI units, the expressions of the internal and external electric field of a
dielectric sphere under an applied field. Compute the bodys electric dipole
moment and compare the results with those obtained by Panofsky and Phillips
(PP) by solving Laplaces differential equation172. Identify the depolarization factor
there given and verify if it has the right value.
Solution
For a sphere of radius a, in SI units, eqs. 2.33 and A1.14 give

3 3 0
E zint = E0 = E , (6.6)
(3 0 + 0 )/ 0 +2

where = 1 + is the dielectric constant.

From the potential PP, eq. 5-18, it is obtained
!
! 3E 0 (r ) 3 0
(r ) = z, E z = = E for r < a, (6.7)
+2 z +2

coincident con the previous equation.

For the depolarization tensor method the electric dipole moment is related to
the polarization by eqs. 2.33:

! ! 3 0 ! 0 3 0( 1) ! 0 ! ! 1 ! 0
P = 0 E zint = E = E , p = VP = 4 0a3 E , (6.8)
+2 +2 +2


172 Panofsky and Phillips, pp. 76-77.

 Depolarization tensor method 123

the same value given by PPs eq. 5-19.

The expression for the external field in the depolarization tensor method is
given by eqs. 2.33:

! ! ! 1 ! ! ! 1 ext ! ! 0
E ext (r ) = E 0 next (r )i P = E 0 3 n (r )i E , (6.9)
0 + 2 sphere

ext
where nsphere is given by eq. 3.66. Thus

3x 2 r 2 3x y 3x z

r5 r5 r5 0
! 1 3 y x 3 y2 r2 3 y z
Eext (r ) = E0 + a3 i0 ,
+2 r5 r5 r5
E 0
3z x 3z y 3z 2 r 2

r5 r5 r5
(6.10)
3 1 3x z 0
a + 2 r5 E

3 1 3 y z
= a E 0
+ 2 r5

E 0 + a3 1 3z r E 0
2 2

+ 2 r5

The potential given by PP, eq. 5-18, is

1 z 0
= a3 E z E 0 for r > a. (6.11)
+2 r 3

The fields components are therefore

Ex =

x
= a3
1 0 3
+2
zE
x
( )
r = a3
1 3x z 0
+ 2 r5
E ,

Ey =

y
= a3
1 0 3
+2
zE
y
( )
r = a3
1 3 y z 0
+ 2 r5
E , (6.12)

E z = = E 0 a3
z
1 0
+ 2
E
z
( )
zr 3 = E 0 + a3
1 3z 2 r 2 0
+ 2 r 5
E ,

coincident with the values eqs. 6.10.

The depolarization factor defined by por PPs eq. 5.20 (where factor 0 of the
!
used edition should be in the numerator, not in the denominator)
 124 Carlos E. Solivrez

! ! 3 0
E 0 E int E
0
E
+2 1 1
L = 0 ! = = = , (6.13)
P 3 3( 1) 3
0 0 E 0
+2

is the value of the three eigenvalue of the spheres unit depolarization tensor.

Magnetic polarization
Problem 04: Deriving n from the magnetic vector potential
Derive eq. for magnetic induction in terms of n starting from the vector

potential for uniformly magnetized matter!M : 173
! ! !
! ! 0 ' M (r r ') 3
4
A(r ) = ! ! 3 d r '. (6.14)
V rr'

Solution
Usando las equations A2.2 and 3.8 se obtiene

! ! 0 ' 1 ! ' ! !
! ! d r ' M =
0
A(r ) = 3
I(r ) M. (6.15)
4 V r r ' 4

From the identity eq. A2.9,

! ! ! ! 0 ' ! !
B(r ) = A(r ) =
4
I(r ) M ( )
! ! ! ! ! !
( ) ( ! !
) !
( ) !
(
I(r ) M = M i I(r ) I(r )i M + I(r ) i M M i I(r ) (6.16) ) ( )
! ! !

!
(
= M i I(r ) MI(r ). )
Using eqs. 3.3 and A3.7 it is obtained
! ! !
! ! 0 'M 0 'N i M for r V
B(r ) = ! ! ! , (6.17)

0 'next (r )i M for r V

!
which are eqs. 2.35 and 2.44 for H 0 = 0.

Problem 05: Permanently magnetized infinite right circular cylinder

A very long right circular cylinder of radius R has a permanent uniform
magnetization M forming an angle 0 with the symmetry axis. Evaluate the internal

173 Reitz, p. 191 eq. 9-12.

 Depolarization tensor method 125

and external magnetic field due to the magnetization and plot the external field
lines.
Solution
The internal and external magnetic fields are given by the matrix equations
2.44. The coordinate system is chosen so that the cylinders axis corresponds to z
and plane xz is parallel to vector M. From the first of eqs. 2.44,

H int 1/2 0 0 M x ' M sin 0 /2

x

Hint = H int
y
= ' N i M = ' 0 1/2 0 i 0 = 0 . (6.18)
int 0
Hz 0 0 M z 0

That is
' '
H xint = M sin 0 , H int y
= 0, H zint = 0, H = M , H! = 0, (6.19)
2 2

where in the last two equations the components are given in the directions parallel
and perpendicular to the cylinders axis. It should be noticed that the internal
magnetic fields component parallel to the cylinders axis vanishes because of the
shape anisotropy.
The external magnetic field is obtained the second of eqs. 2.44 and eq. 3.58:
! !
Hext (r ) = ' next (r )i M
H ext
x
cos2 sin2 0 M
ext ' R2
Hy = sin2 cos2 0 i 0 , (6.20)
ext 2 2

0 0 0 M"
Hz
H ext M cos2
x
ext ' R
2

Hy = 2
M sin2 .
ext 2 0
Hz

The equations of the field lines are174
x 2 + y 2 = k y, z = z0 . (6.21)

where k and z0 are. The geometry of these lines is
shown by Figure 22. The lighter gray
circumpherences are the equipotental lines, function
only of . The k values of the field lines, with typical
dipolar shape, are succesive multiples of 2, except for
the horizontal one for which k = 210. Lines are the Figure 22. Field lines of a
same for any value of z. uniformly magnetized cylinder.

174 Kemmer, p. 47 Problem 12.
 126 Carlos E. Solivrez

Problem 06: Isotropic magnetic sphere

An isotropic ferromagnetic sphere is placed in a
magnetic uniform field H0 with fixed sources.
Assuming the magnetization M(H) to be the
function given by the right graph where M s is the

saturation magnetization and M 0 the spontaneous

one determine the resulting values of the

magnetization and the magnetic field.

Solution Figure 23. Graphical solution for the
isotropic ferromagnetic sphere.
From eq. 2.44 it should be
'
H = H 0 M(H), (6.22)
3

where M(H) is the curve represented in Figure 23. In the same graph is
represented the line

H0 H H0
g(H) = , such that g(0) = g0 = , g(H 0 ) = 0. (6.23)
'/3 '/3

The desired values of M and H are given by the intersection of that line with the
function M(H), because at that point

H0 H '
g(H) = = M(H), that is, H = H 0 M(H). (6.24)
'/3 3

The external field is the given by

! ! ! ' ! ! !
H(r ) = H 0 next (r )M(r ), (6.25)
3

where the value of next is that of the sphere, eq. 3.66.

Conductors

Problem 07: Solve ellipsoidal conductors using equivalente polarization

Derive the equations that solve the case of an ellipsoidal conductor in a uniform
applied field using the equivalente polarization method.
Solution
Starting with the potential eq. 2.65, use the expression eq. 3.29 for the
depolarization tensor asumming that the equivalente polarization is uniform. It is
then obtained
! !
! ! !0 Pf i d 2r ' ! 0 ! !
S r r '
E(r ) = E k1 " ! ! = E 4 k1
n( r )i Pf . (6.26)

 Depolarization tensor method 127

!
The equivalente polarization Pf is determined by the vanishing internal field,

and the expression of the external field is obtained in the same way as in Problem
08.

Problem 08: A conductor is a perfect dielectric

Verify that ellipsoidal conductors behave as dielectrics with infinite scalar
susceptibility.
Solution
Combining eqs, 2.12 and 2.33, for a dielectric with scalar it is obtained

( )
1
Eint = E0 N i Peq , where Peq = 0 1+ N E0 for r V ,
0
(6.27)
! !
E (r ) = E0 next (r )i Peq for r V .
ext

Taking the limit in the principal system of coordinates of N,

( )
1

1+ N x 0 0

( ) ( )
1 1
1+ N = 0 1+ N y 0 ,

( )
1
0 0 1+ N z (6.28)

( ) 1
( )
1 1
lim 1+ N = lim = N 1, lim 1+ N = N 1 .
1
+ N


From the previous equation and eq. 2.66:

0
Peq = N 1 i E0 , Eint = E0
N i Peq = 0,
0
(6.29)
0
Eext (r ) = E0 next (r )i Peq , p = V Peq = V N 1 i E0 .
0
!
These equations are the same as 2.63 and 2.66 if the following assignation is made:

Peq = E + . (6.30)

0
Problem 09: Polarization of a spherical conductor
Compare the charge density induced on the surface of a spherical conductor,

eq. 2.57, that obtained by using the equivalent polarization Peq eq. 2.66.
!
 128 Carlos E. Solivrez

Solution175

From eq. 2.57, the surface charge density at point !r S is

0 ! ! !
! 1 ! rS
(r S ) = s(r S )i N 1 i E 0 , N = 1, s(r S ) = . (6.31)
3 R

From the expression of the electric dipole moment of the sphere, eq. 2.60,
! !S
! 3 0V ! 0 !S 1 pir
p= E , (r ) = . (6.32)
V R

The same expression is obtained from the last of eqs. 6.29 for the equivalent
polarization model. This expression for should be compared with that given by
eq. 3.32.

Problem 10: External field of a spherical conductor

Prove that the external field created by a spherical conductor immersed in a
uniform electric field is that of a point dipole and find its moment176 .
Solution
The field of a polarized ellipsoidal conductor is given by the first two of
eqs. 2.63, where the internal depolarization tensor is N = (1/3)1. That is
! ext ! ! 0 ext ! !0 !0
E ( r ) = E n ( r )i N 1
i E = ( 1 3next !
( r )) i E , (6.33)

where next for a sphere is given by the second of eqs. 3.67,

!! ! !
! R3 3rr (r i r )1
next (r ) = . (6.34)
3 r 5

Therefore
! ext ! ! 0 !( ! !0 ) 2 !0
3 3r r i E r E
E (r ) = E + R 5
. (6.35)
r

The electric dipole moment of the body is given by the last two of eqs. 2.63:

! 0V 1 ! 0 4 0R ! 0 1 3 ! 0
3

p= N iE = E = R E . (6.36)
k1

Upon replacemente in the expression for the external electric field it is obtained
! ! ! !
! ext ! ! 0 3r ( r i p ) r 2p
E (r ) E = k1 , (6.37)
r5


175 Solivrez (2008), p. 207.
176 Solivrez (2008), p. 207.

 Depolarization tensor method 129

which is the expression for a dipolar field (see eq. A4.1).

Problem 11: Directions for which P is parallel to E0 in triaxial ellipsoids

A metallic triaxial ellipsoid is placed in an arbitrary uniform field E0. Assuming
the three eigenvalues of N are known, determine:
a) The electric dipole moment p of the ellipsoid;
b) the directions along which the equivalent polarization Peq is parallel to the
applied field E0;
c) the angle between P and E0.
Solution
From eqs. 2.63 the electric dipole moment p in the ellipsoids principal system
of coordinates chosen so that a > b > c is

N 1 0 0 E x0
V 0V a
0 1 1 1
p = 0 N 1 i E0 = 0 N b1 0 i E y , where < < . (6.38)
N a
N b
N c
0 0 N c1 E z0

The angle between the equivalent polarization Peq = p/V and the field E0 is
obtained from the scalar product and norm of both vectors:
! !
P i E0 ! !0 0
( )
2
cos =
P E 0
, P i E = P E

0
= N

1
E 0
,

(6.39)
0
( ) ( )
2 2
P= N E , E = E .

1 0 0 0

Angle vanishes when the numerator and the denominator of the first equation
are equal, which happens only along the ellipsoids principal axes (see eq. 5.15 and
Figure 18) where

! ! 0 0
( )
2
P i E 0 = N 1 E0 , P = N1E0 , E 0 = E , = x , y,z. (6.40)

Problem 12: Electric fields generated by sharp points

Analize the fields generated in sharp portions of conductors submitted to
uniform electric fields. To that end calculate the field intensity on the intersection
of the surface of a very long prolate spheroid with symmetry axis x and compare it
with that on the spheroids equator.
Solution
The field on the bodys surface may be obtained from eqs. 3.41 and A7.8. For the
point (a,0,0) is obtained
 130 Carlos E. Solivrez

N 1 0 0
a
s(a,0,0) = x , next (a,0,0) = 0 Nb 0 . (6.41)

0 0 Nc

From eq. 2.63 the electric field at point (a,0,0) is
1i 0
E (0,0,c) = E n (0,0,c)i N E ,
ext 0 ext
(6.42)
where
next (0,0,c)i N 1
N 1 0 0 Na1 0 0 1
a 1 Na 0 0 (6.43)
= 0 Nb 0 i 0 N b1 0 = 0 1 0 .

0 0 Nc 0 0 N c1 0 0 1

Therefore, the electric field has the following components normal and paralled to
the x axis:
1 0
E xext (0,0,c) = E x , E ext
y
(0,0,c) = E zext (0,0,c) = 0. (6.44)
Nz

When the oblate spheroid becomes very thin a b = c and Na is very small. The x
component of the surface field becomes very large, normal and outgoing from the
body at point (a,0,0). A very small applied electric field, like the ones generated
during an electric storm, may thus generate a very large external field with the
value given by eq. 6.44.
The equatorial electric field, tangent to the surface, vanishes as the internal one
because of the continuity of the tangential component through any interphase.

Superconductors
Problem 13: Superconducting infinite cylinder and sheet
Show how to solve the infinite right circular cylinder of infinite length and the
sheet of constant thickness and infinite extension in both the magnetization and
surface current model. Discuss the existence of a reduced inverse of matrix N.
Solution
Use the same method applied for a conducting infinite sheet in page 56.

Problem 14: Magnetized superconducting sphere

Evaluate the magnetization and the internal and external fields generated by a
superconducting sphere under a uniform applied induction !B 0 z . Compare with the
expressions given by Reitz177.


177 Reitz, p. eq. 15-6.

 Depolarization tensor method 131

Solution
The expressions given by Reitz (in SI units) are:

! ! 3 0 ! 3 0
B = 0, H = M=
B z, B z for r a. (6.45)
20 20

! ! ! ! a3 1 a3
B(r ) = 0H(r ) = B 0 z 3 B 0 cos r 3 B 0 sen ,
r 2r
(6.46)
3 0
K M (a, , ) = B sen .
20

From eqs. 2.73, where N = (1/3)1, it is obtained in SI units (' = 1) for r a:

1
! ! ! ! 1 1 ! 3 !0 !
B int = 0 ( H int + M ) = 0, M = 1 B 0 = B = H int , (6.47)
0 3 20

as given by Reitz.
Equation 3.66 gives next for the sphere showing the it is a dipolar field (see eq.
A4.1) for r a,
! !
! ext ! ! 0 ext !
! ! 0 a3 r 2M 3(M i r! )r!
B (r ) = B 0n (r )i M = B 0
3 r5
! !
! 0 a3 B 0 3(B 0 i r)
r ! 0 a3 0 1 a3 0
=B + = B 3 B cos r 3 B sen ,
(6.48)
2 r3 r 2r
!0
as B = B z = B cos r B sen ,
0 0 0

as given by Reitz.
According to eqs 2.74 and A7.8 the magnetization surface current is

! ! ! ! ! 3 !0 3 0 3 0
K M (r ) = M(r ) s(r ) = B r = B z r = B sen , (6.49)
20 20 20

which completes the proof of perfect agreement.

Problem 15: Superconductivity as perfect diamagnetism

Verify that the equations describing superconductivity with the magnetization
model may also be obtained from the perfect diamagnetic case, that is with
susceptibility m = 1.

 132 Carlos E. Solivrez

Solution
Induced magnetization in ellipsoids is described by eqs. 2.49. For Hint and M,
where m = 1, it is

( ) 1 1
( )
1
Hint = 1 N i H0 , M = mHint = Hint , Bint = 0 Hint + 'M = 0. (6.50)
' '

The external field is then give by

ext ! !
H ( r ) = H0
'n( r )i M. (6.51)

The expressions are the same given by the magnetization model of

superconductivity, eqs. 2.73.

Problem 16: Magnetic moment of a superconducting ellipsoid

Evaluate the magnetic dipole moment of a superconducting triaxial ellipsoid
using the surface current model. Compare its value with that given by the
magnetization model.
Solution
From eq. A1.12, the magnetic dipole moment of a surface distribution of current
is
! k ! ! ! 2
2 S
m= 3 " r K(r )d r. (6.52)

! !
where K(r ) is current density vector whose value may be obtained from eqs. 2.75
and A1.11:
! ! 1 ! !
K(r ) = s(r ) B + , (6.53)
0 'k3

It is then obtained
! !

!
m=
1

"
20 ' S
!
r d 2r B + . ( )
(6.54)

The integrands double vector product may be expanded as follows,
! ! ! ! ! ! ! !


!
( ) ( ) ( )
r d 2r B + = r i B + d 2r r i d 2r B + , (6.55)

which replaced in the integral gives

! ! 1 ! 2! ! +

!
m=
1

"
20 ' S
( !
)
r i B + d 2r S r i d r B .
20 ' "
(6.56)

The first surface integral may be transformed into a volume integral by using the
gradient theorem eq. A3.9; the second, by using the divergence theorem eq. A3.8.
Thus

 Depolarization tensor method 133

! !+ 3 1 ! 3 !+
!
m=
1
20 '
r (
i B d r )
20 '
( i r )
d r B ,

V
! !
V (6.57)
(! +
) + !
where r i B = B , i r = 3, d r = V . 3

where use has been made of !B + . Using eq. 2.82 it is finally obtained

! !
! 1 1
( )
1
m= V B+ = s i B 0 , s = V 1 N . (6.58)
0 0 '


where s is the bodys polarizability tensor of the superconductor. This value of !m
!
coincides with that given by eq. 2.72.

Problem 17: Description of Meissner Effect using magnetization currents

Solve the Meissner Effect in an ellipsoidal superconductor using only the surface
magnetization current.
Solution
The solution follows the same steps as in the derivation surface current model
of superconductivity, but using eq. 6.53.

Problem 18: Approximate solution for non ellipsoidal bodies

Find an approximate solution for the field generated by a non ellipsoidal body in
one of the cases of Table 1.
Tentative solution
Use the iterative method discussed at the end of section Solving the integro-
differential equations by iteration at page 42. This is an open problem.

Depolarization tensor
Problem 19: Components of N in spherical symmetry
Add to the symmetry operations of spheroids (see eq. 3.25) a single one that
corresponds to a sphere. Use it to determine the three eigenvalues of N in
combination with the unit trace rule eq. 3.15. Which is the matrix representation of
N in a coordinate system different from the principal one?
Solution
A symmetry should be added that exchanges the spheroids rotational axis by
any of the two normal axes, for instance a rotation R in 90 around coordinate axis
x, as follows.
 134 Carlos E. Solivrez

1 0 0 N 0 0 N 0 0

R iN = 0 0 1 i 0 N 0 = 0 0 N!

0 1 0 0 0 N! 0 N 0
(6.59)
N 0 0 1 0 0
N 0 0



= NiR = 0 N 0 i 1 0 1 = 0 0 N .

0 0 N! 0 1 0 0 N! 0

from which N = N = N. If the unit trace rule eq. 3.15 is applied, it follows that
!
3N = 1, that is, N = 1/3.
As N is proportional to the unit matrix, its representation is invariable in any
coordinate system due to the transformation property eq. 3.23:

N' = R i( 13 1)i R t = 13 1 (6.60)

where the orthogonality relationship eq. 3.17 was used.

Problem 20: Single eigenvalue may determine the type of spheroid

An ellipsoid has an eigenvalue N = 0.16, the other two being equal. What is the
value of the latter? To what type of ellipsoid they pertain? Which is the ratio of the
smaller semiaxes to the largest?
Solution
The body is a spheroid with polar eigenvalue Np = 0.16 and two equal equatorial
eigenvalues Ne. From the trace rule eqs. 3.15, Np + 2Ne = 1; therefore,
Ne = (1 Np)/2 = 0.42, where Np < Ne. As the order of semiaxes is the inverse of the
order of eigenvalues, the polar semiaxes is greater then the equatorial ones,
corresponding to a prolate spheroid (see Figure 20). In the conventional order
a b c, Na = 0.16, Nb = Nc = 0.42. The ratio of semiaxes = may be obtained
from eq. 4.30 or from Figure 14; the latter being enough for the given precision. By
inspection it is easily seen that = 0.47, value that may be checked using Figure
15 for Nb = Nc = 0.42 and adjusted for higher precission using eq. 4.30.

Problem 21: Eigenvalues and aspect ratios

Using each of the following set of data, identify all the eigenvalues of N and the
aspect ratios of the ellipsoid to which it pertain.
a) N = 0.65; = 0.10.
b) N = 0.60; = 0.50.
c) N1 = N2 = 0.40.
Solutions
a) The range of values 0.5-1.0 pertains only to Nc. As seen from Figure 16, for
= 0.10 the value Nc = 0.65 is on the curve = 0.20. The other eigenvalues for
this aspect ratio are Nb = 0.32 (obtained from Figure 15) and Na = 0.03 (Figure

 !"#$%&'()&*($+,*"+-$',."*/$0, 94C,

9;^6,
2^,F,-(.(%&',*/(+?,/&##"+-,(+,*/(-,A&-"W,P/"'",9-,<,76E7,&+0,',<,76C7W,P/"'",I'$.,
J(?B'",9E,(*,(-,I$B+0,),,<,76=C6,K/",*P$,'".&(+(+?,"(?"+N&%B"-,&'",9;,<,76,=8,
\$2*&(+"0,I'$.,J(?B'",9C^,&+0,9+,<,7699,\J(?B'",9;^6,
A^,K/",$+%S,#$--(2%",-#/"'$(0,(-,*/",#'$%&*",$+",P(*/,9;,<,9-,<,76;7W,',<,),,<,76CG,
&+0,9+,<,76=76,

S+8(7$6&993&H$U5$"-$&8F&$77.1)8.#)&8B$+&)86$&,)1$-*&+,*.8&-5+B$)&
eN"',*/",%(+",),<,76CW,&+&%()",*/",A/&+?"-,$I,-/&#",$I,*/","%%(#-$(0,A$''"-#$+0(+?,
*$,9+W,9; &+0,9- &*,*/'"",0(II"'"+*,N&%B"-,$I,'6 !"N(-",&,-(.#%",?"$."*'(A,
'"#'"-"+*&*($+,*/&*,-/$P-,A%"&'%S,*/",A/&+?"-,(+,*/",'&*($,$I,-".(&_"-6,
H875*.8"&

' ) 2*% 2>% 2)% :77.1)8.#& &

76C7, 76C7, 769G, 76;9, 76;9, #'$%&*",-#/"'$(0,

,

76GC, 76C7, 76=9, 7649, 76;5, *'(&_(&%,

,

9677, 76C7, 76=;, 76=;, 76C9, $2%&*",-#/"'$(0,

,

S+8(7$6 9>3 :.<$"B,75$)&8F *+.,%.,7&$77.1)8.#)

[-(+?,*/",0&*&,?(N"+,2S,e-2$'+,&+0,M*$+"',A&%AB%",*/",#'(+A(#&%,N&%B"-,$I,V,I$',
&+,"%%(#-$(0,-BA/,*/&*,(*-,-".(&_"-,-&*(-IS,*/",'"%&*($+-/(#-,9,3,=,3,46,
H875*.8"&
K/",N&%B"-,A&+,2",$2*&(+"0,I'$.,*/",*&2%"-,&+0,?'&#/-,$I,e-2$'+",&+0,M*$+"'6,
F-,2$*/,B-",*/",A$+N"+*($+,+,+,;,+,-W,P",.&S,A/$$-",+,<,4W,;,<,=W,-,<9W,P(*/,&-#"A*,
'&*($- ),<,-y+,<,7W44444n,',<,;y+,<,7WEEEEE6,
J'$.,e-2$'+"Y-,K&2%",X,*/",I$%%$P(+?,&#'$_(.&*",N&%B"-,&'",$2*&(+"03,

-y+ <,7W4;=7= &+0,;y+ <,7WE5;9;3

, ,9+,<,7W9E9GEn,9;,<,7W=E=77n,9-,<7WCGE=;6,, \E6E9^,
, -y+,<,7W4;=7=,&+0,;y+,<,7WC899C3, ,
9+ <,7W9;5;7n,9; <,7W47;;7n,9- <7WC;E=76

J'$.,M*$+"'d-,?'&#/W,*/&*,?(N"-,%&'?"',"''$'-W,$+",$2*&(+-,9+,<,7W9Cn,9;,<,7W=8n,
9;,<,7WCC6,
 136 Carlos E. Solivrez

Problem 24: Depolarization tensor of a very long right circular cylinder

Compute the depolarization tensor, eq. 3.4, of a right circular cylinder of infinite
length using electrostatic Gausss Law.
Solution
See section Right circular cylinder of infinite length of chapter Chapter 3.

Problem 25: Depolarization tensor of a sphere

Compute the depolarization tensor of a sphere, eq. 3.4, using electrostatic
Gausss Law.
Solution
See section Sphere of chapter 3.

Problem 26: Alternative way of obtaining certain eigenvalues of N

Explore alternatives ways of obtaining the eigenvalues of N not given in this
book. See, for instance, MacMillans equations 32.10 and 39.4. Bear in mind the
discussion made at section Unified treatment of N for spheroids. Test your approach
deriving the depolarization factors of the two types of spheroids.
Solution
Open problem.

Problem 27: Solving the prolate spheroid with Legendres elliptic integrals
Show that the principal values for the prolate spheroids may be obtained from
the general expression in terms of normal of elliptic integrals eq. 4.9.
Solution
The prolate spheroid corresponds to the case k = 1, for which E and F are given
by eq. A9.1, and should be expressed in terms of = .

E( ,1) = sin , F( ,1) =

dt
cost
( ( ))
= ln tan /2+ /4 , cos = , (6.62)
0

where use has been made of the last of eqs. 4.9. Expressing the tangent in terms of
cosine , it is obtained

tan /2+ /4 =
sin /2+ /4
( ) ( )

cos /2+ /4
(
(6.63) )
sin( /2)cos( /4)+ cos( /2)sin( /4) sin( /2)+ cos( /2)
= = ,
cos( /2)cos( /4) sin( /2)sin( /4) cos( /2) sin( /2)

as sin(/4) = cos(/4). Taking into account the following trigonometric identity178 ,

178 Korn & Korn, p. 810 eq. 21.2.9.

 Depolarization tensor method 137

1 cos 1+ cos
sin( /2) = , cos( /2) = for 0< < , (6.64)
2 2 2

it is finally obtained
sin( /2)+ cos( /2)
(
tan /2+ /4 =
cos( /2) sin( /2)
)
1 cos 1+ cos 1+ cos 1 cos
+ +
2 2 2 2
= (6.65)
1+ cos 1 cos 1+ cos 1 cos
+
2 2 2 2

( ) = 1+
2
1+ cos + 1 cos 1 cos2 1+ 1 2
= = .

(
1+ cos 1 cos ) cos

Therefore
1+ 1 2
E( ,1) = 1 2 , F( ,1) = ln . (6.66)

As = , replacement of this values in the first of eqs. 4.9 gives

E( ,1)
Nx = + F( ,1)
(
1 2 ) 1 2 (
1 2 ) 1 2
(6.67)
2 1 1+ 1 2
= ln 1 .
1 2 1 2

The value coincides with that of eq. 4.30, obtained by direct integration of eq. 3.82,
which serves at the same time of verification of the expression in normal elliptic
integrals eq. A9.1.

2 1 2
Ny = 2 2 + 2 2
( 1+ 1 2
ln ,
)

1 2 ( )(
1 2 1 2 ) ( ) (6.68)
2
1 2
Nz = .

( 2
2
) ( 2
2 ) 1 2

Problem 28: Spherical shell

Explore the existence of a depolarization tensor for a spherical shell
(homoeoid), using Gausss Law.
Solution
Gausss theorem will be used to find the potential of the homoeoid of Figure 24
where the gray region is homogeneously charged matter. The depolarization
 138 Carlos E. Solivrez

tensor will then be calculated by derivation of the potential. The shell is assumed
to have constant volume density of charge . For the calculation of the flux of the
electric field the surfaces S1, S2 and S3 will be used, corresponding to the three
regions r < R1, R1 < r < R2 and R2 < r. From the symmetry of the problem the electric
field has only radial component E(r).
Over S1 the flux is
! !
S dS = 4 r E = 4 k1Q1 = 0, E = 0.
"
2
E (5.69)
1
Over S2 it is
! !
i dS = 4 r E = 4 k1Q2
"
2
E
S2

4 3 3
= 4 k1
3
r R1 , ( ) (6.70)

4 k1 R13
E= r 2 for R1 r R2 . Figure 24. Spherical homoeoid.
3 r
Over S3 it is obtained

! ! 4 3 3
S i dS = 4 r E = 4 k1Q3 = 4 k1 3 R2 R1 ,
" E 2
( )
2
(6.71)
4 k1 R23 R13
E= for R2 r.
3 r2

he electric potential is obtained by integration over r of the previous

T
expressions, from which it is obtained f (see eq. 3.43):

const. for r R1 ,
3
! (r) 1 1 2 R1 1
4 k1 4 k1
f (r ) = = E(r)dr = const. + r + r for R 1 r R2 , (6.72)
6 3

const. + R2 R1 r 1 for R r.
3 3

3 2

values that coincide with those given by MacMillan179. The potential is the
difference of the potentials of the spheres of radius R2 and R1. The double
derivatives of f (see eq. 3.8) do not define a valid (constant) internal
depolarization tensor.

179 MacMillan, p. 40.

 Depolarization tensor method 139

Problem 29. Unit vector normal to the ellipsoid's surface

Verify that vector s , eq. A7.8, is normal to the surface at the intersections with
the three cartesian axes and at every point of the ellipses determined by the
intersection with the planes defined by an arbitrary pair of coordinate axes.
Solution

Unit vector !s(r ) is given by
1

! x 2 y2 z2 2 x y z
s(r ) = 4 + 4 + 4 2 x + 2 y + 2 z . (6.73)
a b c a b c

At the intersection with any coordinate axis the ellipsoidal surface is trivially
normal to it. Less trivial is the condition for the elliptic section determined by a
pair or coordinate axes, like the plane xy discussed below. The corresponding
ellipsoids central section for z = 0 is

x2 y2
2
+ 2 = 1. (6.74)
a z
The equation of the straight line normal to the ellipse at point (x1, y1) is180

y y1 a2 y1
= . (6.75)
x x1 b2 x 1

Vector !s at that point is
1

! x 1 y1 2 x 1
2 2 2
y12
s(r1 ) = 4 + 4 2 x + 2 y , (6.76)
a b a b

whose components have the ratio
!
s y (r1 ) y12 a2
! = , (6.77)
s x (r1 ) b2 x 12

coincident with the slope of the straight line normal to the ellipse at that point.

Problem 30: Surface step discontinuity of the depolarization tensor

Prove eq. 3.38 without making reference to an electric or magnetic fields.
Solution
Hint: use eq. 3.29.

180 Korn and Korn, p. 46.
 140 Carlos E. Solivrez

Problem 31: Ellipsoidal conductor with net charge

Discuss the feasibility of using the depolarization tensor method for obtaining
the surface charge distribution on ellipsoidal conductors with net charge in an
applied electric field.
Solution
It is an open problem not solved by the author. The depolarization tensor
method seems to require zero net charges.

Problem 32: Application of Ivorys method to the sphere

As an illustration of Ivorys method given in section Obtention of the external
gravitational potential by Ivorys method, find the internal and external potential
of a homogeneous spherical mass.
Solution
Internal value
The potential inside the body is given by eqs. 4.4 and 4.5, so it is necessary to
evaluate the following integrals.

ds ds
( ) 2
1/2
= = 2 R2 + s = , (6.78)

0 ( a + s )( b + s )( c + s ) ( R + s )
2 2 2
0
2
3/2
0 R

ds ds
(R + s ) 1
3/2

= = 23 = 23
2
. (6.79)

0 (d
2
+s ) ( a + s )( b + s )( c + s ) ( R + s )
2 2 2
0
2
5/2
0
R3

The potential eq. 4.2 is therefore

! 1 1
(r ) = C0( )+ C x ( )x 2 + C y ( ) y 2 + C z ( )z 2 = M 3 r 2 , (6.80)
2R 2R

where M is the mass of the sphere. The expression, mutatis mutandis, coincides
with the potential of a uniformly charged sphere inside the body (eq. 3.62).
External value
The value of is determined by solving eq. A7.20, that for the sphere of radius R
becomes
x 2 + y2 + z2 r2
= = 1, giving = r 2 R2 . (6.81)
R 2
+ R 2
+

This value should be used in eqs. 4.52 and 4.54. The integrals to evaluate are

ds ds 2
= 2( s + R2 )
1/2
= = , (6.82)
( a + s )( b + s )( c + s ) (R + s )
r 2 R2

2 2 2 2
r R 2 2
3/2
r

 Depolarization tensor method 141

ds ds
(R + s ) 21
3/2

= = 23 =
2
. (6.83)


( d + s ) ( a + s )( b + s )( c + s )
2

2 2 2
r 2 R2 (R + s )
2
5/2
r 2 R2
3 r3

The potential eq. 4.2 at points external to the body is therefore

!
(r ) = C0( )+ C x ( )x 2 + C y ( ) y 2 + C z ( )z 2
1 2 x 2 + y 2 + z 2 4 R3 1 M 1 (6.84)
= abc 2 = 3 r= 4 r,
r 3 r3

where M is the mass of the sphere. The expression, mutatis mutandis, coincides
with the potential of a uniformly charged sphere outside the body (eq. 3.62).

Problem 33: Expression for far away fields obtained from next
Using Ivorys method give the expression of next at large distances from a
uniformly polarized triaxial elipsoid.
Solution
Far away from the body the radial distance r a and a2. Therefore, in the
equation A7.20 that determines it is valid to disregard the squared semiaxes in
the denominator so that

x2 y2 z2 r2
+ + 1, r 2 . (6.85)
a + b + c +
2 2 2

ext !
These value of should now be used in the expression for n (r ) , eqs. 4.66, with


abc ds
2 d 2 + s
N ( ) =
( ) ( a + s )( b + s )( c + s )
2 2 2

(6.86)

abc ds abc abc 1

3/2
= = s =2 ,
2 r2 s 5/2
3 r 3 r3

( a + )( b + )( c + ) = r ,
2 2 2 3
(6.87)

x
! d +
2
x
s (r | ) =
= . (6.88)
x2 y2 z2 r
+ +
(a + ) (b + ) (c + )
2 2 2
2 2 2

Therefore
 142 Carlos E. Solivrez

! abc ! ! ! !
next (r ) = N( ) s(r | ) s(r | )

( a + )( b + )( c + )
2 2 2

(6.89)
!!
1 Vr rr
2
= 1 3 5 ,
4 3 r 5 r

where V is the volume of the ellipsoid eq. A7.4. The tensor, equal to that of the
sphere eq. 3.66, corresponds to a dipolar field where the dipole moment is V times
the electric or magnetic polarization (see eq. 6.36).

Problem 34: Field on the external side of the surface of a polarized sphere
Verify the validity of eq. 3.41 giving the value of next on the surface of a sphere.
Solution
The quoted expression is
ext ! S !S !S
n (r ) = N s(r )s(r ), (6.90)

where, as for the sphere a = b = c = R,

1
!
!S x 2
y z x
2 2 2
y z rS !
s(r ) = 4 + 4 + 4 2 x + 2 y + 2 z = , where R= r S . (6.91)
a b c a b c R

From eqs. 3.66, N and the surface value of the external depolarization tensor for
the sphere of radius R are given by

1 1 x2 x y x z
0 0
3 3 R2 R2 R2

, next (r! S ) = yx 1 y2 y z
N=
1
0 0 2 2 . (6.92)
3 R 3 R2 R

1 zx z y 1 z2
0 0 2 2
3 R R 3 R2


In matrix notation !s(r )s(r ) is

x x2 x y x z

!S !S 1
s (r ) s (r ) = 2 y i
R
( x y z ) 1
= 2 yx
R
y2

y z . (6.93)
z zx z y z 2

Therefore, the two values coincide and
ext S S S
!n (r ) = N s(r ) s(r ). (6.94)

 Depolarization tensor method 143

Problem 35: Microscopic origin of depolarization tensors

Discuss the derivation of depolarization tensors from microscopic magnitudes
such as the tensor m discussed at page 24.
Solution
Apart from the argument given at page 24, it is an open problem of interest for
the discussion of the relationship between microscopic and macroscopic
electromagnetic fields.

Energy, forces and torques

Problem 36: Energy of a dielectric sphere
A dielectric sphere of radius R is uniformly polarized by a uniform applied field
0 P P
!E . Verify, integrating ! 0E i E over all space, that there is no contribution to the

total free energy from the depolarization field !E P .
Solution

The external field E Pext generated by the spheres polarization !P is the dipolar
!
one eqs. 6.10, A4.1 and 6.8:
! ! ! !
! ext ! 3( p i r) r r 2p ! ! 3 0( 1) ! 0
E P (r ) = k1 , where p = V P = V E . (6.95)
r5 +2

The contribution to the energy of this external field is (see eqs. A4.11 and A1.4):
2
0 ! ext ! ! ext ! 3 0 4 k1 p2 1 ! !
2
Fext
= E (r )i E P (r)d r = = V PiP (6.96)
e P
3 V 9 0
V'

The contribution to the energy of the internal depolarization field E Pint is (eqs.
!
2.12 and 3.66)

! ! ! ! int ! ! 2 !
E Pint = N i P = P, DP = 0E Pint + P = P,
0 3 0 3
(6.97)
! int ! int 3 V ! int ! int 1 ! !
Fe =
int 1
2
E P i DP d r = (
E i DP =
2 P
) 9 0
V P i P,
V

which cancels the external contribution.

Problem 37: Fermis contact interaction

For an atom, find the energy of interaction between the electronic current
density and the magnetization of the nucleus. Approximate the latter by a
uniformly magnetized triaxial ellipsoid. Compare with the standard expression of
Fermi s contact term UF and discuss the origin of the difference.
 144 Carlos E. Solivrez

! !
U F = 13 20 n i me , (6.98)

! !
where n is the nuclear magnetic moment and me the electronic magnetization

over the nucleus181.

Problem 38: Torques on spheroids

Compute the torque exerted over an isotropic dielectric spheroid immersed in a
uniform applied field. Is it possible to obtain the value of the dielectric
susceptibility only from this value?
Solution
The force couple is given by eqs. 5.21 that, in the spheroids principal system
with z as the symmetry axis, gives

0 N! N
x = V 2e E 0y E 0z ,
(
1+ N 1+ N
e ! )(
e )

0 N! N
y = V 2e E 0z E 0x , (6.99)
(
1+ N 1+ N
e ! )(
e )

N N
0
z = V E 0 E 0 = 0.
2

( )
e
1+ N 2
x y
e

The expression of the field components in spherical coordinates is

E 0x = E 0 sen cos , E 0y = E 0 sen sen , E 0z = E 0 cos ,

( ) k
( E ) sen2 sen ,
2 2
x = k E 0 sen cos sen = 0

2
= k ( E ) sen cos cos = ( E ) sen2 cos ,
0 k2
0
2
y
(6.100)
2

= ( E ) sen2 ( x sen y cos ) = ( E ) sen2 x cos( )+ y sen( ) ,
! k 0
2 k 0
2

2 2 2 2

0 N" N
k = V e
2
.

1+ N 1+ N
e " (e )( )

The torque is normal to the plane determined by the symmetry axis and the field,
and tends to turn the largest semiaxis towards the field. The value of k can be

181 C. E. Solivrez; The contact hyperfine interaction: an ill defined problem; J. Phys. Solid St. Phys.
vol. 13, L1017-L1019; 1980. Although the problem is of quantum nature, the discussion is
made in terms of classical electromagnetism.

 Depolarization tensor method 145

determined from the amplitude of the function ( ) . As all other parameters are
known, e may then be found by solving a second degree equation.
!
Problem 39: Torque exerted over an anisotropic dielectric sphere
A spherical and anisotropic dielectric is suspended in a uniform and fixed
electric field. Express the torque exerted over the sphere in terms of the principal
values of the electric susceptibility tensor.
Solution
For a sphere of anisotropic uniaxial dielectric (see Table 5) it is sufficient to
study the force couple on any plane containing the symmetry axis, using as
coordinates the principal system of the susceptibility tensor. z axis is taken to be
!
the symmetry of revolution one and plane yz as containing the applied field E 0 .
From eqs. 5.21 and 3.66 it is then obtained

! ! ! ! 0 ! !
( )
1
= p E 0 , p = e i E 0 , e = V e i 1+ 13 e , E 0 = E 0y y + E 0z z,


! 0 "
p= V
1+ 3
1 (
x
x +
y
y)+
z
1+ 3 "
1

z 0
(
i E y y + E z z
0
) (6.101)

0 E y "E z0
0

= V y + z .
1+ 13 1+ 13 "

The force couple is then

! 0 E y "E z0
0

= V
1+ 3
1
y + (
z E 0y y + E 0z z
1+ 3 "
1 )
(6.102)
0 " 0 "
= V E 0 0
E
x = V
E 0y E 0z x.

1+ 13 1+ 13 " y z (
3 1+ 13 1+ 13
" )( )

For an applied field in quadrant xy, the exerted torque makes the body align its
smallest susceptibility eigenvalue with the applied field.

Problem 40: Torque on a magnetized disk

A constant thickness circular disk of Fe-Si alloy is permanently magnetized in a
uniform way. The disk is suspended from its edge in a uniform and horizontal
magnetic field. Determine the value of its magnetization from the torque exerted
by the field on the disk. See the experimental configuration at Figure 21.
Solution
A circular disk of thickness much smaller than its diameter may be aproximated
by an oblate spheroid of polar semiaxis equal to half its thickness and an equatorial
semiaxis equal to its radius. From the discussion at page 111 about magnetization
of a compasss needle , the magnetization vector should lie on the equatorial plane.
 146 Carlos E. Solivrez

If the disk can rotate freely, The magnetization will align with the field so that the
free energy eq. 5.35 has it minimum value.
The coordinate system is chosen so that the suspension is parallel to the z axis,
the magnetization lies on the plane xy and the field lies along the direction x. Then
! !


!
( )
( ) = m B 0 = VM cos x + sin y B 0 x = VMB 0 sin z.
(6.103)

The value of M is obtained from the amplitude of the torque as a function of .

Problem 41: Torque on a very long superconducting right circular cylinder

Compute the torque exerted over a very long cylindrical superconductor
suspended normally to its symmetry axis in horizontal and uniform magnetic field.
Solution
The torque over a body of large volume is very large, and large bodies are very
difficult to keep at the temperature where they are superconductors. The problem
is unrealistic and without practical interest.

Problem 42: Depolarization tensor of inifinitesimally thin shells

Verify that all infinitesimally thin shells of ellipsoidal shape have a
depolarization tensor similar to that of the spherical one of eq. 5.56:
!S !S
N = s(r ) s(r ), n = 0.
ext
(6.104)

where !s(r S ) is the unit vector normal at point !r S of the bodys surface (see eq.
A7.8).
Solution
Use the superposition principle for the potential of ellipsoids with constant
density of charge.

Problem 43: Applications of thin shells

Explore the application of thin ellipsoidad shells for solving a practical problem.
Solution
Open answer.

 Depolarization tensor method 147

Appendix 1:
Electromagnetic units

The different systems of electromagnetic units are characterized by the

constants appearing in the basic laws of electromagnetism. While almost without
exception engineers use the International System of Units182 (SI), physicists often
use other systems more convenient for their particular field of studies. For that
reason these general constants are used in this book, following Jacksons ideas183.
A few other constants are next discussed that depend on the definition of certain
derived but relevant magnitudes, as electric and magnetic susceptibilities and
moments.
Coulombs Law
qq'
F = k1 2 , (A1.1)
! r

defines in some systems the unit of electric charge, through k1. The constant in
Gausss Law is a consequence of this definition,

i dS = 4 k1Q.
E (A1.2)
!S
Electric displacement is defined in terms of electric field and electric
polarization density,

D = 0E + P, (A1.3)
!
where 0 is vacuum permittivity. is an adimensional constant that makes electric
displacement a magnitude of the same species than electric polarization.
For all the systems of units here discussed it happens that

4 k1 0
= 1. (A1.4)

The electric dipole moment p of two point charges +q, -q at distance d is

182 Electromagnetic units at the Bureau International des Poids et Mesures (BIPM), SI maintenance
agency.
183 Jackson, pp. 611-621. The book does not analyze the relationships between E and B, P and M
brought by the theory of relativity, which surely stablishes more relationships between
constants, perphaps those given by eqs. A1.4 and A1.10.
 148 Carlos E. Solivrez

4 k1 0
p= qd = qd, (A1.5)

where eq. A1.4 was used.

The unit of electric current I is defined by the force per unit length exerted on
two parallel conductors a distance d apart:

F I I '
= 2k2 . (A1.6)
! l d

From Maxwells equations A1.13, constants k1 and k2 are related by the velocity
of light c, that of propagation of electromagnetic waves in vacuum,

k1
= c 2. (A1.7)
k
2
The unit of magnetic induction is defined by the magnetic part of the Lorentz

force exerted over a charge q moving at velocity !v :
! ! !
F = k3 q v B. (A1.8)

Ampres Law is the magnetic analogue of Gausss Law,

! ! 4 k2
"C i dl = k I.
B (A1.9)
3

The relationship between magnetic induction !B , magnetic field and
magnetization is

B = 0(H + 'M), (A1.10)
!
where 0 is vacuum magnetic permeabiliity. ' is an adimensional constant that
!
makes H and M magnitudes of the same species. The source of magnetic field H _ is
!
magnetized matter, reason why it is more often used in this book than B._

For the systems of units discussed here it happens thats

4 k2
= 1. (A1.11)
0k 23 '

The magnetic dipole moment of a plane coil of wire with electric current I that
encircles an area A is
m = k3 IA. (A1.12)

The constants appearing in Maxwells equations are the following:

 Depolarization tensor method 149

!
! !B
iD = ,
E = k3 ,
t
! (A1.13)
! ! ! D
B = 0, H = k3 ' J + ,
t

!
where is the volume density of electric charge and J the surface density of
electric current.
For the SI system the given constants have the following values:

1 7 2 kgm
3 0 107 kgm
k1 = = 10 c , k2 = = , k3 = 1,
4 0 C 2s 2 4 C2
(A1.14)
107 C 2s 2 kgm
0 = , = 1, 0 = 4 107 2 , ' = 1.
4 c kgm
2 3
C

The following table gives the values for all the common systems of
electromagnetic units: SI or rationalized MKS, ESU or electrostatic CGS, EMU or
electromagnetic CGS, Gauss and HL or Heaviside-Lorentz. Gauss system is the one
prefered by physicist for the study of relativistic transformations because it makes
explicit the speed of light c.

4 k1 0 4 k2
System k k k '
!1 !2 !3 !0 ! 0 0k 23 '

SI 1/40 0/4 1 184 1 410-7 1 1 1
ESU 1 1/c2 1 1 4 1/c2 4 1 1
EMU c2 1 1 1/c2 4 1 4 1 1
Gauss 1 1/c2 1/c 1 4 1 4 1 1
HL 1/4 1/4 c2 1/c 1 1 1 1 1 1

Table 4. The five common systems of electromagnetic units.

184 See eq. A1.14.
 150 Carlos E. Solivrez

 Depolarization tensor method 151

Appendix 2:
Vectorial Operators

The following properties of operator (nabla or del) are frequently used in this
book185.

= x + y + z . (A2.1)
! x y z
! !
1 r r'
! ! = ! ! 3 . (A2.2)
rr' rr'

2
= i = 2 . (A2.3)
x
!

F(r )i G(r )
(A2.4)
!
( ) ( ) (
= F(r )i G(r )+ G(r )i F(r )+ F(r ) G(r ) + G(r ) F(r ) . ) ( )


!
(
)
i f (r )F(r ) = f (r ) i F(r )+ f (r )i F(r ). (A2.5)

! ! ! ! ! ! ! ! ! ! ! !


( )
i F(r ) G(r ) = G(r )i F(r ) F(r ) G(r ). (A2.6)

! f ( r ) = 0. (A2.7)

! ! ! ! ! ! !


( !
) !
(
f (r )F(r ) = f (r ) F(r ) + f (r ) F(r ). ) ( ) (A2.8)

! ! ! ! ! ! ! ! ! !


( ) ( ) (
F G = G i F F i G + F iG G i F . ) ( ) ( ) (A2.9)

185 Korn and Korn, pp. 157-162.
 152 Carlos E. Solivrez

 Depolarization tensor method 153

Appendix 3:
Integral theorems of vector calculus

Laplacian of a point charges potential

Eq. A3.7 that defines the trace of the internal depolarization tensor uses the
value of the following expression:

!
f (r ') 3 !
i ! ! d r ' = I '(r ), (A3.1)
V rr'


which coincides with !I(r ) cuando ! f (r ') = 1.

For !r r ' the order of integrating and taking derivatives may be permuted.

!
f (r ') ! 1
! ! d 3r ' = f (r ') ! ! d 3r ',
V rr' V rr'
1 1 ! !
(r r ')
where ! ! = i ! ! = i ! ! 3 (A3.2)
rr' rr' rr'

! ! ! ! ! !
i(r r ') 1 ! ! 3 ( r r ')i( r r ') ! !
= ! ! 3 ! ! 3 i(r r ') = ! ! 3 + 3 ! ! = 0 r r ',
rr' 5

rr' rr' rr'

where eq. A2.5 was used. It is thus proved that the laplacian vanishes for !r r '.

The case where a field point coincides with a source point only possible
inside the material body is more difficult. It is not uncommon to find books on
electromagnetism with erroneous proofs that by mere chance give the right
value186. That is why a proof is given here that circumvents the common pitfalls.

The order of calculation of the integral and the derivatives may be exchanged
only when the integrand has integrable singularities187. It is therefore convenient

186 See, for instance Jackson, p. 13 and Reitz, p. 44, where they take the double derivative inside the
integral. V. Hnizdo, Eur. J. Phys. 21, pp. L1-L3 (2000) makes errors like using Gausss Law as
different from the divergence theorem.
187 Hnizdo, ; Generalized second-order partial derivatives of 1/r: European Journal of Physics, vol.
32, pp. 287-297; 2011.
 154 Carlos E. Solivrez

to divide the bodys volume in two parts: a spherical volume of radius R0 around
the singularity radius that at the end will tend to 0 and the rest of the body,
volume V' = V - V0, as illustrated at Figure 25.

That is, for !r fixed but arbitrary,
!
! f (r ') 3
I '(r ) = ! ! d r '
V rr'
! ! (A3.3)
f (r ') 3 f (r ') 3
= ! ! d r ' + ! ! d r '.
V V0 r r ' V0 r r '


Figure 25. Isolating the
As shown before, the laplacian of the first singularity inside the body.
integral vanishes so that the calculation is
reduced to
! !
f (r ') 3 f (r ') 3
! ! d r ' = ! ! d r '
V rr' V0 r r '
! (A3.4)
f (r ') 3 ! 1
= i ! ! d r ' = i f (r ')' ! ! d 3r '.
V0 r r ' V0 rr'


When!I '(r ) (eq. 3.8) is a continuous function with well behaved derivatives,

which happens for any continuos and differentiable function! f (r ') , the gradient
theorem eq. A3.9 may be used to reduce the last volume integral to a surface one
without singularities:

1 ! d 2r! '
! ! d 2r ' !
i f (r ')' ! ! d 3r ' = i " S f ( r ') ! ! =
" f ( r ') i ! !
V0 rr' 0
r r ' S0 rr'
(A3.5)
! ! 2!
! 1 2! ! (r r ')i d r ' 2
! d r'
=" S f (r ') r! r! ' i d r ' = "
S f (r ') R3 =" S f (r ') R2 .
0 0
0 0 0

!
where d 2r ' is the vectorial element of area.

The following limit is next taken:

! d 2r ' ! d 2r ' !
lim f (r ') 2 = f (r )lim " = 4 f (r ),
R0 0 "
2
(A3.6)
S0 R0 R0
0
S0 R0

where use was made of eq. A2.5, of the fact that the vectors in the scalar product
! ! !
( r r ')i d 2r ' are antiparallel, and the definition of solid angle. At all steps the
integrands have integrable singularities.
Therefore

 Depolarization tensor method 155

! ! !
f (r ')d 3r ' 4 f (r ) for r V
! ! = ! , (A3.7)
V r r ' 0 for r V

Divergence, gradient and rotor theorems

The divergence or Gauss-Ostrogradskys theorem stablishes the following
identity,188,189
3 2
i F( r )d r = S r )i d r ,
F( (A3.8)
!V

where r is the unit vector normal to surface S and outgoing from volume V. In its
!
standard formulation the theorem requires that the field F and its first derivatives
have no singularities in the regin V of integration and its boundary S. It is also
valid for fields with integrable singularities. Such conditions are not fulfilled for the
point charge field eq. 1.3.
The gradient theorem is a corollary of the divergence theorem190,191

f (r )d r =
f (r )d r .
3 2
(A3.9)
! V S

Another corollary is the curl theorem192,193:

! ! ! ! !
F(r )d r = "
F(r ) d r .
3 2
(A3.10)
V S

Extension of the divergence theorem

The standard formulation of the divergence theorem is insufficient for
applications to electromagnetism, where the field sources are often singularities
like point charges, lines of charge, surface densities of charge and current, and
step discontinuities of polarizations. The extension of the divergence theorem is
done next for two types od singularities; it may also be done for the line of
charge194, but it is of no interest here.

188 Korn and Korn, p. 163.
189 Kuntzmann, pp. 329-332.
190 Korn and Korn, p. 163.
191 Kuntzmann, p. 339.
192 Korn and Korn, p. 163.
193 Kuntzmann, p. 340.

194 Kuntzmann, p. 371.

 156 Carlos E. Solivrez

Point charge type singularities

The extension of the divergence theorem for point charges195, usually not
claimed as different from the original one196, is known in physics as the
electrostatic Gausss Law. It is applied to fields of the following kind:

q j (r rj )
F(r ) = . (A3.11)
3
j r rj
!
In order to apply the standard divergence theorem it is necessary to exclude from
the region V of integration an infinitesimal volume Vj around every singularity. The
exclusions are indicated here by replacing volume V by V ', replacing the original
integral by what mathematicians call principal value integral. The process is
similar to that previously depicted at Figure 25, except that one must include the
flux at the surfaces Sj that are the boundaries of each volume Vj. Therefore

3 2
q j (r rj ) i S j 2

i F(r )d r = F(r ) i d r
3 d r ', (A3.12)
V' S j Sj r rj

where the sums negative sign comes from the fact that the vectores are entrant to
the volume, not salient as required by the divergence theorem. Except for the
factor qj, all the integrals in the sum are equal because they subtend the same solid
angle197,

(r rj ) i dS j
S r r 3 =
S sen( )d d = 4 , (A3.13)
j j j

where a spherical coordinate system is used.

Resulta finalmente

F(r ) i d r 4 q .
i F(r )d r = (A3.14)
3 2
j
V' S j

In the formulation of the electrostatic Gausss theorem,!F is the electric field!E
and the equation is presented in the following way:
! ! 2! !
"S r )i d r = 4 k1
E( (r )d 3r + 4 k1 q j , (A3.15)
V ' j

where the relationship between!E and the volume charge density has been
used198. The most general expression of the relationship between the flux of the


195 Kuntzmann, p. 369.
196 Reitz, p. 35.
197 Kemmer, p. 71.

198 Reitz, p. 85 eq. 4-29.

 Depolarization tensor method 157

electric field and the distributions of free charge is obtained upon addition of the
surface densities of charge (step discontinuities discussed in next section) and
linear densities of charge (not done here). The polarization charges and currents
are discussed in the main text, case by case
Step discontinuities
Generalized divergence theorem
The generalization is given here of the divergence theorem for the case in which
the field has a step discontinuity through a surface199, that in this book will be the
bodys surface.
A field is said to have a step discontinuity
(also called jump discontinuity or discontinuity
of the first kind) through a surface when it
is continuous and finite in any neighbourhood
of but its limit values over both sides of
are different. Defining conventionally a
positive and a negative side of 200 (see
Figure 26) and identifying each limit with the
corresponding upper + or - index, one has
Figure 26. Step
discontinuity surface .
F + (r ) F (r ) 0. (A3.16)

The divergence theorem eq. A3.8 cannot be applied to a region than contains a
step discontinuity. The problem is not that the integrals involved cannot be
computed for such surfaces201, but the omission of the surface. Such integrals have
to be divided and computed separately for each of the regions that have as a
boundary. In the case of Figure 26 the divergence theorem should not be applied
for the whole of volume V, but to the two regions determined by . A
mathematically rigorous way of justifying this is to exclude an infinitesimal volume
V0 around , volume delimited by the surface S0 identified in the figure with
dashed lines.
The application of the theorem in the indicated way gives
3

V V0
i F( r )d r =
F( r )
S
i d S +
F(
S0
r ) i d S . (A3.17)

When the vector field vectorial is finite in any neighbourhood of , so it is the flux
on the edge of surface S0 tambin lo is and tends to 0 when the thickness of V0 goes
to 0. That is, the principal value integral should be computed, which will be
indicated in this book with the specification lim V0 0. Taking this limit does not
modify the volume integrals value.

199 Kuntzmann, pp. 371-372.
200 En este libro is la superficie del cuerpo and el valor positive del campo is el exterior a ella.
201 See, for instance, Morse and Feshbach, p. 34.
 158 Carlos E. Solivrez

In the aforesaid limit the previous equation gives

3 3

V
i F( r )d r = lim i F( r )d r
V0 0
V V0
+

= F( r
S
) i d S F ( r ) i d +

F ( r ) i d

+

= F( r ) i d S F ( r ) F ( r ) i d ( )
S
Therefore,
! ! 3 ! ! ! !+ ! ! ! !
i F( r )d r = S
" F( r )i dS
" F ( r ) F ( r ) i d . ( ) (A3.18)
V

where !d is a vector outgoing from the positive side of the surface. The equation
is the generalization of the divergence theorem for the case where the field has a
step discontinuity over a closed or open surface .
A more compact and elegant demonstration may be given by using
distribution theory, but this requires many other concedpts, as an extension of the
derivative operation202 which exceeds by far the requisite of a first course of vector
analysis imposed in this book.

Generalized rotor theorem

If in theorem A3.18 !F is replaced by !F c , where !c is an arbitrary constant
vector, one gets
! ! ! 3 ! ! ! 3
(
i F(r ) c d r = c i F(r )d r )
V V
! ! ! ! !+ ! ! ! ! !

="
S
( )
F(r ) c i dS " F ( r ) F ( r )

c (
id ) (A3.19)
! ! ! ! !+ ! ! !
!
( !
S dS F(r ) c i "
=c i" )
d F (r ) F (r ) , ( )


where eq. A2.6 and the cyclic property of the scalar triple product were used. As !c
is an arbitrary vector, it is obtained
! ! ! ! ! !+ ! ! ! !
F(r )d r = " ( F
F(r ) dS + " )
(r ) F (r ) d ,
3
(A3.20)
V S

Polarized bodies
!
All the bodies studied in this book are characterized by a polarization vector Q
with the following properties:
It is uniform and non-vanishing inside the body V;
it vanishes outside the body;

202 See, for instance, Farassat, F., Introduction to Generalized Functions with applications in
Aerodynamics and Aeroaccoustics, NASA, Langley Research Center, April 1996.

 Depolarization tensor method 159

it has a variable step discontinuity through the bodys closed surface S, the S
surface of the previous theorems;
If the integration region is V ' with frontier S ' that contains both V and , it is
obtained for eq. A3.18
! ! 3 ! ! 3
i F(r )d r = i F(r )d r
V' V
! ! ! ! ! ! ! ! ! (A3.21)
="S' F( r )i dS +
" F ( r )i d =
" F ( r )i d ,


because ! i F and !F vanish outside V.
In the same say, eq. A3.20 reduces to
! ! ! ! !
F(r )d r = "
F (r ) d .
3
(A3.22)
V
 160 Carlos E. Solivrez

 Depolarization tensor method 161

Appendix 4:
Field of an electric dipole

General expression
!
The electric field !E(r ) generated at point field r by a body with electric dipole

moment !p placed at the origin is

! ! ! ! ! !
3( p i r )r r 2 p
E(r ) = k1 . (A4.1)
r5

j
The field !E j generated by a dipole !p located at source point !r j is203
! ! ! ! !
!j !
E (r ) = k1
(
!
3d j(r ) d j(r )i p j d j(r)2 p j
,
)
j 5
d (r) (A4.2)
! ! ! !
3d j(r )d j(r ) d j(r)2 1 ! j j ! ! !j
= k1 i p , where d (r ) = r r ,
d j(r)5

and 1 is the unit dyadic such that for an arbitrary vector !C it is
! !
1i C = C. (A4.3)

The upper index indicates the source point and the argument the field point (see
p. 11).
In matrix representation the previous expression becomes

E(r ) = k1m j pj , (A4.4)
!

203 Reitz, p. 39 eq. 2-36.
 162 Carlos E. Solivrez

where
3d j (r )2 d j(r ) 2
3d jx (r ) d jy (r ) 3d jx (r )d jz (r )
x

d j(r )5 d j(r )5 d j(r )5

j 3d jy (r )d jx (r ) 3d jy (r ) 2 d j(r ) 2 3d jy (r )d jz (r ) .
m (r ) = (A4.5)
d j(r )5 d j(r )5 d j(r )5

3d j (r )d j (r ) 3d jz (r )d jy (r )

3d jz (r ) 2 d j(r ) 2
z x
j 5
d (r ) d j(r )5 d j(r )5
!

Dipolar field's energy

A polarizable sphere of radius R immersed in an applied uniform field !C 0
! !
generates outside a dipolar field C q (r ). Depending on the kind of material
(dielectric, conductor, diamagnetic, paramagnetic or superconductor) and applied
field (electric or magnetic) the constants and the dipolar moments will vary.
Nevertheles, in all cases, when the center of the sphere is the origin of the
coordinate system, the energy stored in the external field is proportional to the
following integral:
!q ! !q ! 3
(r )i C (r )d r ,
C (A4.6)
V'
where V' is the complementary region of V (the one occupied by the spherical
body) and
! q ! 3(q! i r! )r! r 2q!
C (r ) = . (A4.7)
r5
The integrand is then

!q ! !q !
C (r )i C (r ) =
! ! !
( ! ! ! !
3(q i r )r r 2q i 3(q i r )r r 2q )(
!
)
r 10 (A4.8)
2 ! ! 2 2 ! ! 2 ! !
9r (q i r ) 6r (q i r ) + q r
2 4
3(q i r )+ q 2
= 10
= .
r r6

In an spherical coordinate system with z axis parallel to !q

! q ! ! q ! 3 2
2 q ! 2
C (r )i C (r )d r = d sen d r C (r ) dr
V' 0 0 R

0 R
! 2 2
= 2 sen d r 4 3(q i r) (
+ q dr ) (A4.9)

4
0
(2

R
)
= 2 q 3cos +1 sen d r dr .
2

 Depolarization tensor method 163

As
0

(3cos +1)sen d = (3cos +1)d(cos )

2 2

0
R (A4.10)
r 3

0 1
= cos3 + cos = 4, r 4 dr = = 3,
3 3R
R

therefore
2
! q ! ! q ! 3 2 4 2 4 q2
C ( r )i C ( r )d r =
3 R3
q = 2 3 V , (A4.11)
V '

where V is the volume of the sphere and q the dipole moment norm.


 164 Carlos E. Solivrez

 Depolarization tensor method 165

Appendix 5:
Simmetries of the susceptibility tensors of single crystals
Electric and magnetic susceptibilities, as any tensorial property, have
simmetries that reflect those of the single crystal they characterize. As illustrated
by the depolarization tensor eq. 3.23, the relationship between components are a
consequence of the tensors transformation properties. The following table gives
the relationships for all crystalline systemas204 of the components of a single
crystals susceptibility tensor 205, where the number of independent ones are given
between parentheses in the first column.

cristalline principal matrix

clasification
system coordinates representation
0 0
isotropic
Cubic
or any one 0 0
(1)
anaxial 0 0


symmetry of
trigonal 0 0
revolution
tetragonal
uniaxial around axis of 0 0
hexagonal
maximum 0 0 !
(2)
symmetry
coordinate 0 0
Orthorhombic axes parallel to
biaxial 0 0
(3) the three
binary axis 0 0 !

1 coordinate 0 13
Monoclinic axis parallel to 11
biaxial 0 22 0
(4) the single
binary axis 13 0 33

12 31
Triclinic 11
biaxial none 12 22 23
(6)
31 23 33

Table 5. Simmetries of the susceptibility tensor.


204 Dekker, Captulo 1.
205 Nye, p. 23 Table 3.
 166 Carlos E. Solivrez

 Depolarization tensor method 167

Appendix 6:
Dyadics

A dyadic is a mathematical entity a linear combination of external products of

unit vectors such that its scalar product with a vector is a vector Dyadics may be
interpreted as vector operators such that its action (scalar product) on a vector
gives another vector. In this book dyadics are written in boldface italics. The
expression of n, the dyadic most often quoted here, is
! !
n(r ) = x j n jk (r )x k . (A6.1)
j ,k

A particular case of dyadic is the external product of two vectors:

!!
c = AB = A j x Bk x k . (A6.2)
j ,k

Such is the case of the dyadic !ss in eq. 3.33.

The scalar product of a vector with a dyadic may be taken in two different ways:
!
ni A = x j n jk x k i x l Al = x j n jk x k i x l Al
j ,k l j ,k ,l

= x j n jk kl Al = x j n jk Ak = x j n jk Ak ;
j ,k ,l j ,k j k
! (A6.3)
Ai n = A j x j i x k x l nkl = A j x j i x k x l nkl
j k ,l j ,k ,l

= A j jk nkl x l = A j n jl x l = Ak nkj x j = x j nkj Ak .

j ,k ,l j ,l j ,k j k

The symmetric unit dyadic, here written 1, is

1= x j x j = xx + y y + zz,
(A6.4)
j

such that
! !
1i A = x j x j i x k Ak = x j x j i x k Ak = x j jk Ak = x j A j = A;
j k j ,k j ,k j (A6.5)
! !
Ai 1= A.
In general ! !
ni A Ai n (A6.6)
 168 Carlos E. Solivrez

In the same fashion, exchanging the order of vectors modifies the value of external
products. For instance,

a = x j x k a jk x k x j a jk = x j x k akj = a t , (A6.7)
j ,k j ,k j ,k

where the last dyadic is the transposed of the first.

A dyadic a is symmetric when it is equal to its transposed dyadic, that is, when
its elements ajk are symmetric:

a = x j x k a jk = a t = x j x k akj si a jk = akj . (A6.8)

j ,k j ,k

The two ways of taking the scalar product coincide only for symmetric dyadics:
! !
ni A = x j n jk Ak = x j nkj Ak = Ai n. (A6.9)
j k j k

When a dyadic is represented by a square matrix here written in roman

boldface the two ways of taking the scalar product correspond to different
representations of the vectors:

a axy axz A
xx
x

If a = a yx a yy a yz , A = A y ,
A
azx azy azz z
a axy axz Ax axx + A y a yx + Az azx

(A )
xx
At i a = x
Ay Az i a yx a yy a yz = Ax axy + A y a yy + Az azy , (A6.10)

azx azy azz Ax axz + A y a yz + Az azz

a axy axz A axx Ax + axy A y + axz Az

xx x
a i A = a yx a yy a yz i A y = a yx Ax + a yy A y + a yz Az .

azx azy azz Az azx Ax + azy A y + azz Az

The trace of a dyadic, of special interest for the depolarization tensor, is defined
as the sum of its diagonal elements. That is

Tr a = a jj = axx + a yy + azz . (A6.11)

j

For dyadics defined as the external product of two vectores, the trace coincides
with the scalar product:
!! ! !
If a = AB, Tr a = a jj = A j B j = Ax B x + A y B y + Az Bz = Ai B. (A6.12)
j j

 Depolarization tensor method 169

Appendix 7:
Ellipsoids

Equations and main features

Equations, sections, volume and surface
In an arbitrary orthogonal system of coordinates the matrix equation of an
ellipsoidal surface206 is207
x
A
xx
Axy Axz

r t i A 1 i r = 1, where r = y , A = Axy A yy A yz , (A7.1)

z
Axz A yz Azz

where A is a positive definite symmetric matrix with eigenvalues 1/a2, 1/b2 and
1/c2. The coordinate system that diagonalizes A, here called the principal system,
is the one with origin at the center of the ellipsoid that coincides with its principal
axes208 . Equation A7.1 then becomes
x 2 y2 z2
2
+ 2 + 2 =1 (A7.2)
a b c
where a, b and c are the lengths of the ellipsoid semiaxes.
The parametric equations of the ellipsoid are209

x = acosusinv, y = bsinusinv, z = c cosv,

(A7.3)
where 0 u 2 , 0 v ,

where the definition of the angles u, v and their domains are identical to the angles
and of a spherical coordinate system, but not its relationships with the
!
componentes of the position vector r .

The intersection of the ellipsoid with a plane is always an ellipse210 that

becomes two overlapping circumferences for spheroids.

206 Those not familiar with actual tri-dimensional ellipsoids can profit from the interactive views
given by the Wolfram Demonstration Project Ellipsoids by Jeff Bryant.
207 Korn and Korn, pp. 74-82.
208 Korn and Korn, p. 43.

209 Ellipsoid in Wolfram MathWorld, eqs. 3-5.

 170 Carlos E. Solivrez

The ellipsoid's volume V is

4
V= abc. (A7.4)
! 3
The surface area S is

S = 2 c 2 + 2
ab
sen
(
E( ;k)sen2 + F( ;k)cos2 , )
(A7.5)
c a2(b2 c 2 )
donde a b c, cos = , k = 2 2 2 ,
2

a b (a c )

where F(,k) and E(, k) are the Legendre's incomplete elliptic integrals of the first

and second kind discussed in Appendix 9.

Aspect ratios
In daily life, a sphere is called a sphere regardless of the radius value, because
its apparent size varies with the distance to the observer. In mathematical
language it is said that all spheres are similar, the single parameter that
characterize them differs by an arbitrary multiplication factor k. When two
parameters characterize the shape, as in a rectangle of sides a and b, an observer
sees two similar ones (sides ka and kb) as having the same aspect. If a b, the
ratio = b/a is called the aspect ratio, where the equal sign characterizes the
square as a particular case. By definition an aspect ratio has the following range of
values,
0 1. (A7.6)

It may seem that the value 0 should be excluded, but it is convenient to include it in
order to characterize the degenerate rectangle with a = , two parallel lines.

In the case of ellipsoids, three parameters a, b, c characterize its shapes. If a a

b c, there are two aspect ratios,

0 = b/a 1, 0 = c/a 1. (A7.7)

The different shapes of ellipsoids according to its aspect ratios are identified in
Table 3. Notice that the aspect ratios of the ellipsoid do not suffice to tell apart a
right circular cylinder of inifinite length from an elliptic oner. The electrostatic and
magnetostatic properties of ellipsoidal bodies of the same aspect is exactly the
same because similar ellipsoids have the same depolarization tensor (see section
N(0) is determined by aspect ratios).

210 G. F. Childe, Singular Properties of the Ellipsoid and Associated Surfaces of the Nth Degree,
Macmillan, 1861, pp. 133-137.

 Depolarization tensor method 171

Normals to the surface and central distances

!
The unit vector !s(r ) normal to surface S at point !r may be obtained from the
gradient211 of the eq. A7.2:

x 2 y2 z2 x y z
2 + 2 + 2 x + 2 y + 2 z
! a b c a2
b c
s(r ) = = . (A7.8)
x 2 y2 z2 x2 2
y z 2
2 + 2 + 2 + 4+ 4
a b c 4
a b c

It may, for instance, be easily checked that this vector is normal to the surface of
the ellipsoid at the intersections with each of the three cartesian axes and on the
ellipses determined by the planes defined by any two cartesian axes212 (see
Problem 29).
Unit vector s has some useful properties, best derived when it is written in
! !
terms of the following normal vector s(r ) :
! !
! s(r ) ! ! x y z
s(r ) = ! ! , where s(r ) = 2 x + 2 y + 2 z.
(A7.9)
s(r ) a b c

!
Then, if r is the position vector of a point on the ellipsoidal surface, equation A7.2
is equivalent to the vector equation
x1 y1 z1

( )
r1 i s(r1 ) = x1 x + y1 y + z1 z i 2 x + 2 y + 2 z
a b c
(A7.10)
2 2 2
x1 y 1 z1
= 2 + 2 + 2 = 1 r1 S.
! a b c

The equation of the plane tangent to the ellipsoidal surface at point r1 is213
!
! ! ! x1 y1 z1
r i s(r1 ) = 2 x + 2 y + 2 z = 1, (A7.11)
a b c

relationship that may be easily checked to be true at the intersections of the three
principal axes with the ellipsoidal surface.

In a similar fashion, vectors r1 and s(r1 ) determine a family of parallel planes
! !
whose equation is
! ! !
r i r1 s(r1 ) = k, (A7.12)

211 Korn and Korn, p. 158 eq. 5.5-5.
212 Korn and Korn, p. 46.
213 Kemmer, p. 11 Problem 16 and p. 196.
 172 Carlos E. Solivrez

where k is constant for each plane and equal to its distance to the origin. The plane

through point r1 satisfies the equation
!
! ! ! ! ! ! ! !
r i r s(r1 ) = r1 r1 i s(r1 ) = 0, (A7.13)
1 1
! ! !
expression that shows that r1 , s(r1 ) and the center O of the ellipsoid lie on the

same plane.
In what follows a central plane of the ellipsoid will be any plane that contains its
center, the origin its principal system of coordinates. In the same fashion, a central
section of the ellipsoid will be any intersection of the surface with a central plane,
as illustrated in Figure 27.
The geometrical meaning of vector

! ) may be obtained from Figure 27. It
s( r
is there seen that the distance d j from
!
the ellipsoids center O to the the plane

tangent to S at point rj is the projection
!
of that vector on the direction of s j . This
!
value is here called the central distance
corresponding to the given point of the
ellipsoidal surface. The dot lines in the
figure correspond to the intersections of
the tangent planes and the plane that
Figure 27. Central distance and chord
determines the depicted central section. in an ellipsoids central section.

The general expression for the central distance d of a point !r on the ellipsoidal
surface is
! ! !
! ! ! r i s(r ) !
d(r ) = r cos = r i s(r ) = ! for r S. (A7.14)
s( r )

Using eqs. A7.8 and A7.10, the central distance turns out to be

1/2 1/2
! x 2 y2 z2 x 2 y2 z2 x 2 y2 z2 1
d(r ) = 2 + 2 + 2 4 + 4 + 4 = 4 + 4 + 4 = !
a b c a b c a b c s(r)
1 a2b2c 2 (A7.15)
= = ,
x12 y12 z12 b4c 4 x 12 + a4c 4 y 12 + a4b4 z 12
+ +
a4 b4 c4
! !
Equation A7.15 shows that the norm of vector s(r ) is the inverse of the central

distance of point !r on the ellipsoidal surface. A simple verification of the the
formula is to apply it to the sphere, where a = b = c = R = d.

Eq. A7.14 determines the angle between vector r1 of a point P1 on the ellipsoidal
!
surface and the normal to it at that point:

 Depolarization tensor method 173

! !
r1 i s(r1 )
cos1 = . (A7.16)
r1

The following dyadic is one of the components of the external depolarization
tensor eq. 4.66:
1
! ! x2 y2 z2 x x x x
s(r | )s(r | ) = + + 2 d 2 + d 2 + , ( )( )
( ) ( ) ( )
2 2
a2 + b2 + c 2 + , (A7.17)

where = x , y,z, d x = a,d y = b,d z = c.

This tensor has the property that when applied to a vector leaves only the
component normal to the ellipsoidal surface eq. A7.20 (see also section Surface
step discontinuity, in particular eq. 3.39).
The following property is responsible of the vanishing trace of the external
depolarization tensor eq. :
1
! ! x2 y2 z2 x 2
Tr s(r | )s(r | ) = + + 2
( ) ( ) ( ) (d )
2 2 2
a2 + b2 + c 2 +
2

+
1
(A7.18)
x2 y 2
z 2
x2 y z 2
2
= + + 2 2 + + 2
= 1.
( ) ( ) ( ) ( ) ( ) ( )
2 2 2 2
a2 + b2 + c 2 + a + b +
2
c +
2

Summary of properties
! !
The given properties of normal vector s(r ) are collected next.

! ! x y z ! ! !
s(r ) = 2 x + 2 y + 2 z : s(r ) S at r ;
a b c
! ! ! x 1
y 1
z 1 !
r i s(r1 ) = 2 x + 2 y + 2 z = 1 is the plane tangent to S at r1 ;
a b c
! ! ! x12 y 12 z12 !
r1 i s(r1 ) = 2 + 2 + 2 = 1 r1 S;
a b c
1 ! 1 a2b2c 2
! = d(r ) = =
s(r ) x12 y12 z12 b4c 4 x 12 + a4c 4 y 12 + a4b4 z 12 (A7.19)
+ +
a4 b 4 c 4
!
is the distance to the ellipsoid's center of the plane tangent at r ;
ss is the dyadic that projects from any vector the component
normal to the surface S of the ellipsoid.
 174 Carlos E. Solivrez

Confocal ellipsoids

Figure 28. Central section of two confocal ellipsoids.

For the calculation of the potential outside the body use is made of a family of
ellipsoids confocal with the bodys surface,
x2 y2 z2 !
+ + = 1, where 0, r V . (A7.20)
a + b + c +
2 2 2

Figure 28 shows the central section z = 0 of two such ellipsoids, where a = 5, b = 10

and = 20. Notice that the confocal ellipse is always rounder than the original one
and that the distance between surfaces is largest for the smallest semiaxes. is
!
usually determined as the largest or positive algebraic root of eq. A7.20, where r is
any field point outside the surface of the ellipsoidal body eq. A7.23. To the authors
knowledge no explicit expressions have been given for these roots, which in
general are calculated point by point214 by methods like the Newton-Raphson
iterations215.

Physical-geometric interpretation of
The unit of the confocal parameter is length2, which suggests its
interpretation as a squared distance. This assumption is confirmed when eq. A7.20
is solved for the only type of spheroid where an explicit solution may be easily
found, a sphere of radius R (see eq. 6.81). In this case

214 See, for instance, at Barczak & Breit & Jusiel, Ellipsoids, material points and material segments,
comment in the paragraph preceding eq. 70 in an un-numbered page.
215 Korn & Korn, section 20.2-1.

 Depolarization tensor method 175

= r R .
2 2
(A7.21)

For the general ellipsoid eq. A7.20, upper and lower bounds for may be found.
The squared distance from its center to any surfaces point satisfies the following
inequality,
c + r a + , so that r a r c .
2 2 2 2 2 2 2
(A7.22)
!
If r0 is a point on the bodys surface, its coordinates satisfy the equation

x02 y02 z02
2
+ 2 + 2 = 1, (A7.23)
a b c
its confocal ellipsoid being given by eq. A7.20. The maximum absolute value of
each variable is its semiaxis, when the two other variables are zero. For x, for
instance, it is then obtained


(
= x 2 x02 = a2 + a2 , ) (A7.24)

and a similar result is obtained is obtained for y and z suggesting that the value of
! !
may be r 2 r 20 . The conjecture is true when the components of r and r0 are

taken to have equal angles in their parametric expressions eqs. A7.3:

( )
x 02 = a2 cos2 u0 sin2 v0 , x 2 = a2 + cos2 u0 sin2 v0
y02 = b2 sin2 u0 sin2 v0 , y = ( b + ) sin u sin v ,
2 2 2
0
2
0
(A7.25)
z02 = c 2 cos2 v0 , z = ( c + ) cos v .
2 2 2

A pair of such vectors are depicted in Figure 28, showing that in general they
are not colinear with the elipsoids center (the origin of coordinates).
For corresponding points

r 2 r02 = x 2 x02 + y 2 y02 + z 2 z02

( )
= a2 + a2 cos2 u0 sin2 v0 + sin2 u0 sin2 v0 + cos2 v0 (A7.26)

(
= sin2 v0 + cos2 v0 = . )
That is,
! ! ! !
= r 2 r 20 = r i r r0 i r0 , (A7.27)

Corresponding points
Any two points satisfying eqs. A7.25 are called corresponding points of the
confocal pair of ellipsoids. A similar relationship and its properties are valid for
any two pair of confocal ellipsoids , , but this will not be needed here.
 176 Carlos E. Solivrez

! !
The correspondance of points r0 , r may also be expressed as

x0 x y y z z
= , 0= , 0= , (A7.28)
a a2
+ b b 2
+ c c 2
+

as follows from the fact that the angular parts of each pair of components are the
same. When thus expressed the correspondence provides a simple way of proving
the important properties that follow.
! ! ! !
If r0 , 0 are any two points on ellipsoid A7.23 and r , its corresponding points

on the confocal ellipsoid eq. A7.20, then

! ! ! !
r0 i = r i 0 . (A7.29)

! !
The correspondence between the components , , of vectors 0 and (eqs.

A7.28) and eq. A7.27 give

0 0
= , = , 0= ,
a a2 + b b2 + c c2 + (A7.30)
! ! ! !
where i 0 i 0 = .

The proof easily follows from the expansion of the scalar product and the
relationships eqs. A7.27, A7.28 and A7.30:

! ! a2 + b2 + c2 +
r0 i = x 0 + y0 + z0 = x 0 0 + y0 0 + z0 0 ,
a b c
! ! a2 + b2 + c2 +
r i 0 = x0 + y 0 + z 0 = x 00 + y0 0 + z0 0 (A7.31)
a b c
! !
= r0 i .

A corollary of eqs. A7.27 and A7.29 is

! ! ! !
r 0 = r0 , (A7.32)

as follows from
(r! ! ) i (r! ! ) ( ! r! ) i ( ! r! )
0 0 0 0
! ! ! ! ! ! ! ! ! ! ! ! ! ! ! !
= r i r r i 0 0 i r + 0 i 0 i + i r0 + r0 i r0 i r0
(A7.33)
! ! ! !
(
! ! ! !
) (
! ! ! !
= r i r r0 i r0 i 0 i 0 + 2 r0 i r i 0 )
= + 20 = 0.

 Depolarization tensor method 177

This property is the foundation of Ivorys method for the calculation of the
gravitational potential at points external to the body216.

216 MacMillan, pp. 54-57.
 178 Carlos E. Solivrez

 Depolarization tensor method 179

Appendix 8:
Useful integrals

ds

( )
2
A+ s B + s

The integrals that give the equatorial eigenvalues of both type of spheroids are
of this type. In order to solve it the first step is to reduce the exponent of A+s to 1
using the formula217
ds X 1 ds
X Y 2
=
( )

(
A B Y 2 A B ) X Y
, (A8.1)

where X = A + s, Y = B + s. (A8.2)

The primitives of the remaining integral are218

ds 2 X
X Y
=
BA
arctan if A < B,
B A
(A8.3)

ds 1 X A B
X Y
= ln
A B X + A B
if A > B. (A8.4)

Therefore,

ds X 1 X
=
(
+
) (
arctan if A < B. (A8.5)
)
3/2
X Y 2 BA Y BA B A

217 Korn & Korn, last line of p. 938 and eq. 154 of p. 941.
218 Korn & Korn, p. 941 eq. 149.
 180 Carlos E. Solivrez

ds X 1 X A B
=
(
12
)
ln if A > B,
( )
3/2
X Y 2 A B Y A B X + A B (A8.6)
where X = A + s,Y = B + s.

ds

( A + s ) (B + s )
3/2

This integral, also found for spheroids, may be reduced to the previous one
when integrating by parts. Using the same notation eq. A8.2,

ds ds
= . (A8.7)
( ) ( )
3/2
A+ s B+s X 3/2 Y

Upon integration by parts
udv = uv v du, where
1 ds (A8.8)
u = , du = 2 , v = 2X 1/2 , dv = X 3/2 ,
Y Y
it is obtained
ds 2 ds
udv = X 3/2
Y
=
X Y
2
X Y 2
. (A8.9)

Therefore
ds 2 ds
X 3/2
Y
=
X Y
2
X Y 2
. (A8.10)

Upon replacement from eqs. A8.5 and A8.6 it is obtained

ds 2 2 X 2 X
X =
(

) (
arctan if A < B, (A8.11)
)
3/2 3/2
Y X Y BA Y BA B A

ds 2 1 X A B
X = + ln if A > B.
3/2
Y ( A B) X ( A B
3/2
)
X + A B
(A8.12)

 Depolarization tensor method 181

Appendix 9:
Legendres elliptic integrals

Definitions
!F( ,k) and !E( ,k) are Legendre's incomplete elliptic integrals of the first
219

and second kind, respectively, considered to be a generalization of the well known

trigonometric functions. Though not often mentioned in the theory of
electromagnetism, they appear in common problems such as the computation of
the arc length of an ellipse (from which they got their name), the dependence of
the period of a pendulus on the amplitude of oscillation220 and the surface area of
an ellipsoid (see eq. A7.5).
The canonical expressions for the elliptic integrals are221:

E( ,k) 1 k 2 sin2 t dt 1 sin2 sin2 t dt = E( \ ),

0 0
(A9.1)
dt dt
F( ,k) = F( \ )
0 1 k 2 sin2 t 0 1 sin2 sin2 t

is the amplitude, k the elliptic modulus and the modular angle of these
integrals. Their values can be found in double entry tables like those of
Abramowitz and Stegun (pp. 613-618), and may be calculated for arbitrary values
of their two arguments by mathematical software like Mathematica and Maple.

Notation used in the more important references

Different authors use different conventions for the arguments, both in order and
meaning. The following table shows the relationship between the notation used in
this book and that of the authors more often quoted here. Before making any
calculations the reader should carefully check those used by its favourite or
available author or software, for which it is recommended to check one or more
values, as those given in Problem 21.

219 Complete elliptic integrals have = /2.

220 See, for instance. C. E. Solivrez, Comportamiento de un pndulo real sin rozamiento.

221 Korn & Korn, p. 833, eqs. 21.6-29a; p. 834, eqs. 21.6-29c and 21.6-29d. Weisstein, Eric W;

Elliptic Integral of the First Kind; MathWorld--A Wolfram Web Resource. Abramowitz & Stegun,
pp. 589 eqs. 17.2.8 and 17.2.6.
 182 Carlos E. Solivrez

Nx, Ny,
Author k E( ,k), F( ,k) pp.
Nz
Abramowitz E(\), F(\);
& Stegun
___222 sin 589-590
E(|m), F(|m)223
E(k,), F(k,);
Korn & Korn ___224 sin E(\), F(\)
833-836

MacMillan ___225 k E(,k), F(,k) 58-60
L/4,
Osborn M/4, cos cos sin E(k,), F(k,) 351-352
N/4
Stoner Da, Db, Dc k E(k,), F(k,) 807-808
EllipticE[,k2],
Mathematica ___ k
EllipticF[,k2],
Table 6. Notation used by several authors.

Although all arguments are here real, for a full analysis of the general properties
of elliptic integrals they should be taken to be complex numbers. In respect to their
interest in this book the expressions for the components of the depolarization
tensor of the triaxial ellipsoid, eqs. 4.9 and their arguments are real, bounded
and given by ratios of the ellipsoids semiaxes, as follows:

1 2
0 sin = 1 2 1, 0 k = = sin 1,
1 2 (A9.2)

b c
where = , = .
a a

Special values and parametric graphs

Some special values of the elliptic functions are given in the following table for
its use in the calculation of depolarization tensors. In the case of the constant
interior tensor N, the values of the arguments are connected in a one to one
relationship with those of and , as given in the expressin of the auxiliary
function f for the triaxial ellipsoid eq. 4.52. These values are also shown in the
table, identifying the type of ellipsoid they characterize.

222 Gives general properties and tables of values.
223 m = sin2 .
224 Gives general properties and graphs.
225 Gives only the expression of the potential.

 Depolarization tensor method 183

k Ellipsoids type E( , k) F( , k)
0 k 1 1 sphere 0226 0227
0 0 1 oblate spheroid 228 229

1 /2 prolate spheroid
sin
230
( (
ln tan /2+ /4 )) 231

infinite elliptic
/2 1 /2 0 0 1
cylinder232
infinite sheet of
/2 0 0 1233 0 /2 /2234
constant thickness
Table 7. Special values of the incomplete elliptic functions E( , k) and F( , k).

Figure 29 and Figure 30 show parametric graphs of E(\) and F(\). In both
cases the left side graph shows the dependence on for taken as a parameter
(there written ). The right side shows the dependence on , where each curve
corresponds to a diferent value of parameter .

Figure 29. Parametric graph of the incomplete elliptic function of the second kind E.235

226 Eq. A9.1.
227 Eq. A9.1.
228 Abramowitz & Stegun, eq. 17.4.23.

229 Abramowitz & Stegun, eq. 17.4.19.

230 Abramowitz & Stegun, eq. 17.4.25.
231 Abramowitz & Stegun, eq. 17.4.21.
232 MacMillan, p. 70.
233 See the comments on the infinite sheet in section Infinite semiaxes.

234 Abramowitz & Stegun, Fig. 17.2.

 184 Carlos E. Solivrez

Figure 30. Parametric graph of the incomplete elliptic function of the first kind F( \ ).236

As seen in Figure 29, E(\) is never negative. Its highest value in the intervals
determined by eqs. A9.2 is E(/2\0)= /2237. Its lowest value is attained at the
origin, where E(0\0) = 0. For constant , E(\) is a decreasing function of which
is almost constant for the lowest values of . For constant it is an increasing
function of , linear for = 0 (the line E(\0) = in Table 7).
As seen in Figure 30, F(\) is never negative and diverges at F(/2\ /2). For
fixed it is almost constant for low values ( in the figure). For = 0, is linear in
(the line E(\0) = in in Table 7).

Reduction to normal form of the elliptic integrals of interest

The integrals of interest for the calculation of the depolarization tensor are238

a2 b2 a2 c 2 x 2 y2 z2
For a b c, k = 2 2 , sin = , + + = 1, (A9.3)
a c a2
+ a2
+ b 2
+ c 2
+

235 Abramowitz & Stegun, p. 594, figures 17.6 and 17.7.
236 Abramowitz & Stegun, p. 594, figures 17.3 and 17.4.
237 This corresponds to the complete elliptic integral of the second kind. Its graph is given in
Abramowitz & Stegun, p. 592 Figure 17.2.
b +
2

238 MacMillan, pp. 58-60, where the expression for function dn in p. 60 should be dn v = ,
a +
2

as follows from its definition at p 58, and E() is E( ,k).

 Depolarization tensor method 185

ds 2
( = F( ,k), (A9.4)

a 2
+ s ) ( b2 + s ) ( c 2 + s ) a2 c 2

ds 2
( =
a2 c 2 ( a2 b2 )
E( ,k)+ F( ,k) , (A9.5)
+ s) ( b2 + s ) ( c 2 + s )
3/2
2
a

ds 2 c2 +
( =
(a2 + )(b2 + ) (b2 c 2 )
b2 + s ) (a2 + s )(c 2 + s )
3/2

(A9.6)
2 a c 2 2
2
+ E( ,k) F( ,k),

(a2 b2 )(b2 c 2 ) a2 c 2 ( a2 b2 )

ds
(
c2 + s ) (a2 + s )(b2 + s )
3/2

(A9.7)
2 b + 2
2
= E( ,k).

(a2 + )(c 2 + ) (b2 c 2 ) a c ( b2 c 2 )
2 2

 186 Carlos E. Solivrez

 Depolarization tensor method 187

Main references

The following list identifies the sources quoted more than once in this book.
This simplifies both quotations and its identification by the surnames of the first
two authors of each work. All other references are given at the bottom of the page.
Abramowitz, M. & Stegun, I. A. (editors); Handbook of Mathematical Functions;
Dover Publications; 1972 (10th printing).
Brown, Jr., W. F.; Magnetostatic Principles in Ferromagnetism; North Holland
Publishing; Amsterdam, 1962.
Brown, Jr., W. F.; Micromagnetics; Interscience Publishers; 1963.
Chikazumi, S.; Physics of Ferromagnetism; Oxford University Press; 1997.
Dekker, Adrianus J.; Solid State Physics; Prentice Hall; 1962 (6th printing).
Goertzel, G. & Tralli, N.; Some Mathematical Methods of Physics; McGraw-Hill;
1960.
Gray, A.; On the Attractions of Spherical and Ellipsoidal Shells; Proc. Edinburgh
Math. Soc. vol. 23; 1913-1914; pp. 91-100.
Jackson, J. D. ; Classical Electrodynamics; John Wiley and Sons; 1975 (2nd
edition). One of the three textbooks most quoted in this book, uses Gauss's
system of units.
Kemmer, N.; Vector Analysis: A physicist's guide to the mathematics of fields in
three dimensions; Cambridge University Press; 1977.
Kellogg, Oliver Dimon; Foundations of Potential Theory; Verlag von Julius
Springer; Berlin (Germany); 1929.
Kittel, C; Introduction to Solid State Physics; John Wiley and Sons; 1971 (4th
edition).
Korn, G. A. & Korn, T. M.; Mathematical Handbook for Scientists and Engineers;
McGraw-Hill; 1968.
Kuntzmann, J.; Mathmatiques de the Physique et de the Technique; Hermann;
1961.
Landau, L. D. & Lifchitz, E. M.; Electrodynamics of Continuous Media; Pergamon
Press; 1960.
MacMillan, W. D.; The Theory of the Potential; Dover Publications; 1958.
Maxwell, James Clerk; A Treatise on electricity and magnetism, vol. 2; Dover
Publications; 1954; sectiones 437 and 438.
Morse, P. M. & Feshbach, H.; Methods of Theoretical Physics vol. 1; McGraw-Hill;
1953.
Moskowitz, R. & Della Torre, E.; Theoretical Aspects of Demagnetization Tensors;
IEEE Transactions on Magnetics vol. 2, N 4; pp. 739-744; 1966.
Nye, J. F.; Physical Properties of Crystals; Oxford University Press; 1979.
Panofsky, W. K. H. & Phillips, M.; Classical electricity and magnetism; Addison-
Wesley; 1995.
Osborn, J. A.; Physical Review vol. 67; Demagnetizing Factors of the General
Ellipsoid; pp. 351-357; 1945.
Reitz, J. R. & Milford, F. J. & Christy, R. W.; Foundations of Electromagnetic
Theory; Addison-Wesley; 1960. One of the three main textbooks quoted in this
book, uses SI units.
 188 Carlos E. Solivrez

Santal, L. A.; Vectores y tensores con sus aplicaciones; EUDEBA; 1964 (3rd
edition).
Solivrez, Carlos E.; Magnetostatics of Anisotropic Ellipsoidal Bodies; IEEE
Transactions on Magnetics Mag. vol. 17, N 3; pp. 1363-1364 ;1981.
Solivrez, Carlos E.; Campos elctricos generados por elipsoides uniformemente
polarizados; Revista Mexicana de Fsica, vol. E54, N 2; pp. 203-207; 2008.
Stratton, J. A.; Electromagnetic Theory; McGraw-Hill; 1941. It deals with
ellipsoids by solving Laplace's equation in ellipsoidal coordinates (pp. 58-59).
The text discusses in detail conducting (pp. 207-211), dielectric (pp. 211-217)
and magnetic ellipsoids (pp. 257-258), including degenerate ones and
spheroids. One of the three main textbooks quoted in this book, uses SI units.
Van Vleck, J. H.; The Theory of Electric and Magnetic Susceptibilities; Oxford
University Press; 1932 (1st edition).
Weisstein, Eric W; MathWorld; Wolfram Web Resource.
Young, H. D.; & Freedman, R. A.; Sears and Zemanskys University Physics with
Modern Physics; Addison-Wesley; 2000 (10th edition).

 Depolarization tensor method 189

Alphabetic index

method, limitations: 2, 5, 103
microscopic origin: 24, 143
A N for similar ellipsoids: 63
Ns inverse: 60
aspect ratio: 63, 170 Ns, values: 3
auxiliary function f: 46 oblate spheroid: 69, 95
obtention from potentials: 46
principal coordinates system: 60
B prolate spheroid: 74, 96
representation in arbitrary cartesian
body of infinite extension: 114 coordinates: 49
sheet: 54
sheet of infinite extension: 62
C
sphere: 50, 58, 133, 136, 142
surface step discontinuity: 51
cavity: 116
symmetric tensor: 47
compass
symmetries of N: 50
stability of the needles magnetization: 111
trace: 48
conductor
triaxial ellipsoid: 65, 97, 135
as perfect dielectric: 127
dielectric: 24
discussion: 30
free energy: 106
surface charge density: 52
dielectric constant: 122
constitutive equation
Dirac's delta function: 8
electric case: 18
Dirichlet, gravitational potential of an ellipsoidal
magnetic case: 27
body: 66
dyadic
D definition and properties: 167
notation: 10
demagnetization coefficient: see depolarization
tensor
E
demagnetization factor: see depolarization tensor
demagnetization tensor: see depolarization tensor
Earnshaw's theorem: 30
demagnetizing coefficient: see depolarization
electric dipole moment
tensor
conductors: 33
demagnetizing factor: see depolarization tensor
induced: 25
depolarization: 63, 67
electric displacement vector: 18
depolarization factors: see depolarization tensor
electric field
depolarization tensor
as a surface integral: 31
bodys surface value: 53
as a volume integral: 16
cylinder, circular: 56
at sharp tips: 129
cylinder, elliptic: 68, 93
conductors: 32
definition: 46
general polarization case: 17
eigenvalue Na: 80, 81
induced polarization: 20, 25
eigenvalue Nb: 80, 82
integro-differential equation for conductors: 31
eigenvalue Nc: 80
integro-differential equation for dielectrics: 25
eigenvalues, graphs: 79
on conductors' surface: 31
eigenvalues, integral expressions: 67, 179
permanent polarization: 19
eigenvalues, order: 84
point dipole _: 161
expressions from which is derived: 46
electric permeability: 25
external value: 47, 85
electric polarization
history: 1
equivalent _ for conductors: 33
infinite semiaxes: 60
induced: 20
internal value: 47
permanent: 19
method, applications: 3, 4
electric potential
 190 Carlos E. Solivrez

polarization surface density: 18 atomic case: 21

polarization volume density: 15, 16 discussion: 20
electric susceptibility infinite right circular cylinder
apparent anisotropy: 121 magnetized: 125
definition: 25 integral theorems
electromagnetic units: 147 curl theorem: 155, 158
electrostriction: 101 divergence theorem: 155, 156, 158
ellipsoid gradient theorem: 155
normal vector: 139 singularities: 5
ellipsoid irreversible processes: 100
aspect ratios and type of ellipsoid: 64 Ivory's method: 85
central distance: 172
central plane: 172
central section: 172 J
confocal ellipsoids: 174
corresponding points: 175 jump discontinuity: see singularity
cylinder, circular: 54
cylinder, elliptic: 65
normal unit vector: 171, 173 K
oblate spheroid: 65
principal coordinate system: 169 Kronecker's delta: 11
prolate spheroid: 65
semiaxes order: 64
L
sheet: 54
similar ones: 62
laplacian of a point charge potentials: 153
sphere: 54
lattice sums: 4, 22
surface area: 170
local field: 24, 116
tangent plane: 171
triaxial ellipsoid: 65
volume: 170 M
elliptic integrals: 170, 181
notation: 181 macroscopic field: 6
energy magnetic
crystalline anisotropy: 105 polarization
potential _ of an ellipsoidal body: 66 induced: 30
equivalent polarization: see electric and magnetic magnetic dipole moment
polarization induced: 30
permanent: 29
superconductor: 35
F
surface current: 132
torque: 111
field point: 6, 11, 45
magnetic disk: 145
example: 21
magnetic field
force
as a volume integral: 27
electric case: 114
general magnetization case: 27
magnetic case: 114
induced magnetization case: 30
integro-differential equation for induced
G magnetization: 30
magnetic induction
Gauss-Ostrogradskys integral theorem: 155 constitutive equation: 27
Gibbs free energy: 103 integro-differential equation for the
superconducting case: 37
magnetization model: 36
H magnetic polarization
discussion: 26
Helmholtz free energy equivalent _ for superconductors: 35
definition: 101 induced: 29
linear case: 102 permanent: 28, 29
homoeoid magnetic scalar potential: 27
definition: 117 magnetic susceptibility
spherical _: 137 definition: 29
magnetic vector potential
magnetization: 28
I magnetization current: 28
magnetization: see magnetic polarization
induced electric polarization magnetostriction: 101

 Depolarization tensor method 191

mathematical notation: 10 superconductor: 130

matrix: see matrix representation state function: 99
matrix representation: 10 state variables: 99
Meissner effect: 34 step discontinuity: see singularity
microscopic field: 6 superconductivity: 34
as perfect diamagnetism: 131
superconductor
P magnetization model: 35
surface conduction current model: 36
piezoelectricity: 101 surface charge density
piezomagnetism: 101 conductor: 31
point charge: see singularity polarized dielectrics: 18
polarizability tensor surface current
atomic: 23 superconductor: 36
body's _: 20, 114 susceptibility tensor
conducting body: 32, 33 crystal symmetries: 165
dielectric body: 26
magnetic body: 30
superconducting body: 35, 133 T
principal coordinate system
definition: 11 textured policrystals: 5
thermodynamic potential: 99
thermodynamics: 99
R thermodynamics of electromagnetism: 101
torque
references: 187 anisotropic dielectric: 109
dielectric triaxial ellipsoid: 108
magnetized disk: 145
S spheroid: 144
shell: see homoeoid
singularity V
integrable: 15, 48
integral theorems: 156 vector
jump discontinuity: 6 notation: 10
point charge: 6, 7 uniform: 6
step discontinuity: 9, 157 unit: 10
source point: 6, 11 vectorial operators
example: 21 notation: 11
sources: see references properties: 151
sphere volume
conductor: 127, 128 magnitude: 10
dielectric: 122 region: 10
ferromagnetic: 126

 192 Carlos E. Solivrez

 Depolarization tensor method 193

About the author

Carlos Eduardo Solivrez, born in 1939 in San Salvador de Jujuy

(Argentina), has a Ph. D. in Physics from the Universidad Nacional de
the Plata (Argentina) and a Social Sciences diploma in Education from
Facultad Latinoamericana de Ciencias Sociales (FLACSO)
Universidad Autnoma de Madrid (Spain). He made graduate studies
on Theoretical Solid State Physics at the Berkeley campus of the
University of California. He was a research fellow at the Theoretical Physics
Departament of Oxford University, visiting professor at Fourier University
(Grenoble, France), staff member of the Centro Atmico Bariloche and teached at
several argentine universities, including Instituto Balseiro (Universidad Nacional
de Cuyo Centro Atmico Bariloche, Argentina). His special field of knowledge in
Physics is the theory of magnetic behaviour of quasi-molecular groups. Now
retired, he is presently devoted to the divulgation of scientific and technological
endeavours in Argentina through the wiki Enciclopedia de Ciencias y Tecnologas
en Argentina of which he is the founder and editor where links are given to his
writings.



 196 Carlos E. Solivrez

Depolarization tensor method 197

198 Carlos E. Solivrez