Treatise of Plane Geometry Through Geometric Algebra - R G Calvet

TREATISE OF PLANE GEOMETRY
THROUGH GEOMETRIC ALGEBRA
Ramon González Calvet

TREATISE OF PLANE GEOMETRY THROUGH GEOMETRIC ALGEBRA
The geometric algebra,

initially discovered by Hermann
Grassmann (1809-1877) was
reformulated by William Kingdon
Clifford (1845-1879) through the
synthesis of the Grassmann’s
extension theory and the
quaternions of Sir William Rowan
Hamilton (1805-1865). In this way
the bases of the geometric algebra
were established in the XIX
century. Notwithstanding, due to
the premature death of Clifford, the
vector analysis −a remake of the
quaternions by Josiah Willard
Gibbs (1839-1903) and Oliver
Heaviside (1850-1925)− became,
after a long controversy, the
geometric language of the XX century; the same vector analysis whose beauty attracted the
attention of the author in a course on electromagnetism and led him -being still
undergraduate- to read the Hamilton’s Elements of Quaternions. Maxwell himself already
applied the quaternions to the electromagnetic field. However the equations are not written
so nicely as with vector analysis. In 1986 Ramon contacted Josep Manel Parra i Serra,
teacher of theoretical physics at the Universitat de Barcelona, who acquainted him with the
Clifford algebra. In the framework of the summer courses on geometric algebra which they
have taught for graduates and teachers since 1994, the plan of writing some books on this
subject appeared in a very natural manner, the first sample being the Tractat de geometria
plana mitjançant l’àlgebra geomètrica (1996) now out of print. The good reception of the
readers has encouraged the author to write the Treatise of plane geometry through
geometric algebra (a very enlarged translation of the Tractat) and publish it at the Internet
site http://campus.uab.es/~PC00018, writing it not only for mathematics students but also
for any person interested in geometry. The plane geometry is a basic and easy step to enter
into the Clifford-Grassmann geometric algebra, which will become the geometric language
of the XXI century.
Dr. Ramon González Calvet (1964) is high school teacher of mathematics since 1987,
fellow of the Societat Catalana de Matemàtiques (http://www-ma2.upc.es/~sxd/scma.htm)
and also of the Societat Catalana de Gnomònica (http://www.gnomonica.org).
TREATISE OF PLANE GEOMETRY
THROUGH GEOMETRIC ALGEBRA
Dr. Ramon González Calvet
Mathematics Teacher
I.E.S. Pere Calders, Cerdanyola del Vallès
I
To my son Pere, born with the book.
 Ramon González Calvet ( [email protected] )

This is an electronic edition by the author at the Internet site
http://campus.uab.es/~PC00018. All the rights reserved. Any electronic or
paper copy cannot be reproduced without his permission. The readers are
authorised to print the files only for his personal use. Send your comments
or opinion about the book to [email protected] .
ISBN: 84-699-3197-0
First Catalan edition: June 1996

First English edition: June 2000 to June 2001
II
PROLOGUE
The book I am so pleased to present represents a true innovation in the field of the
mathematical didactics and, specifically, in the field of geometry. Based on the long
neglected discoveries made by Grassmann, Hamilton and Clifford in the nineteenth
century, it presents the geometry -the elementary geometry of the plane, the space, the
spacetime- using the best algebraic tools designed specifically for this task, thus making
the subject democratically available outside the narrow circle of individuals with the
high visual imagination capabilities and the true mathematical insight which were
required in the abandoned classical Euclidean tradition. The material exposed in the
book offers a wide repertory of geometrical contents on which to base powerful,
reasonable and up-to-date reintroductions of geometry to present-day high school
students. This longed-for reintroductions may (or better should) take advantage of a
combined use of symbolic computer programs and the cross disciplinary relationships
with the physical sciences.
The proposed introduction of the geometric Clifford-Grassmann algebra in high
school (or even before) follows rightly from a pedagogical principle exposed by
William Kingdon Clifford (1845-1879) in his project of teaching geometry, in the
University College of London, as a practical and empirical science as opposed to
Cambridge Euclidean axiomatics: “ ... for geometry, you know, is the gate of science,
and the gate is so low and small that one can only enter it as a little child”. Fellow of the
Royal Society at the age of 29, Clifford also gave a set of Lectures on Geometry to a
Class of Ladies at South Kengsinton and was deeply concerned in developing with
MacMillan Company a series of inexpensive “very good elementary schoolbook of
arithmetic, geometry, animals, plants, physics ...”. Not foreign to this proposal are Felix
Klein lectures to teachers collected in his book Elementary mathematics from an
advanced standpoint1 and the advice of Alfred North Whitehead saying that “the hardest
task in mathematics is the study of the elements of algebra, and yet this stage must
precede the comparative simplicity of the differential calculus” and that “the
postponement of difficulty mis no safe clue for the maze of educational practice” 2.
Clearly enough, when the fate of pseudo-democratic educational reforms,
disguised as a back to basic leitmotifs, has been answered by such an acute analysis by
R. Noss and P. Dowling under the title Mathematics in the National Curriculum: The
Empty Set?3, the time may be ripen for a reappraisal of true pedagogical reforms based
on a real knowledge, of substantive contents, relevant for each individual worldview
construction. We believe that the introduction of the vital or experiential plane, space
and space-time geometries along with its proper algebraic structures will be a
substantial part of a successful (high) school scientific curricula. Knowing, telling,
learning why the sign rule, or the complex numbers, or matrices are mathematical
structures correlated to the human representation of the real world are worthy objectives
in mass education projects. And this is possible today if we learn to stand upon the
shoulders of giants such as Leibniz, Hamilton, Grassmann, Clifford, Einstein,
Minkowski, etc. To this aim this book, offered and opened to suggestions to the whole
world of concerned people, may be a modest but most valuable step towards these very
good schoolbooks that constituted one of the cheerful Clifford's aims.
1
Felix Klein, Elementary mathematics from an advanced standpoint. Dover (N. Y., 1924).
2
A.N. Whitehead, The aims of education. MacMillan Company (1929), Mentor Books (N.Y.,
1949).
3
P. Dowling, R. Noss, eds., Mathematics versus the National Curriculum: The Empty Set?. The
Falmer Press (London, 1990).
III
Finally, some words borrowed from Whitehead and Russell, that I am sure
convey some of the deepest feelings, thoughts and critical concerns that Dr. Ramon
González has had in mind while writing the book, and that fully justify a work that
appears to be quite removed from today high school teaching, at least in Catalunya, our
country.
“Where attainable knowledge could have changed the issue, ignorance has
the guilt of vice”2.
“The uncritical application of the principle of necessary antecedence of
some subjects to others has, in the hands of dull people with a turn for
organisation, produced in education the dryness of the Sahara”2.
“When one considers in its length and in its breadth the importance of this
question of the education of a nation's young, the broken lives, the defeated
hopes, the national failures, which result from the frivolous inertia with
which it is treated, it is difficult to restrain within oneself a savage rage”2.
“A taste for mathematics, like a taste for music, can be generated in some
people, but not in others. ... But I think that these could be much fewer than
bad instruction makes them seem. Pupils who have not an unusually strong
natural bent towards mathematics are led to hate the subject by two
shortcomings on the part of their teachers. The first is that mathematics is
not exhibited as the basis of all our scientific knowledge, both theoretical
and practical: the pupil is convincingly shown that what we can understand
of the world, and what we can do with machines, we can understand and do
in virtue of mathematics. The second defect is that the difficulties are not
approached gradually, as they should be, and are not minimised by being
connected with easily apprehended central principles, so that the edifice of
mathematics is made to look like a collection of detached hovels rather than
a single temple embodying a unitary plan. It is especially in regard to this
second defect that Clifford's book (Common Sense of the Exact Sciences) is
valuable.(Russell)” 4.
An appreciation that Clifford himself had formulated, in his fundamental paper upon
which the present book relies, relative to the Ausdehnungslehre of Grassmann,
expressing “my conviction that its principles will exercise a vast influence upon the
future of mathematical science”.
Josep Manel Parra i Serra, June 2001
Departament de Física Fonamental

Universitat de Barcelona
4
W. K. Clifford, Common Sense of the Exact Sciences. Alfred A. Knopf (1946), Dover (N.Y.,
1955).
IV
« On demande en second lieu, laquelle des deux qualités doit être préférée
dans des élémens, de la facilité, ou de la rigour exacte. Je réponds que cette
question suppose una chose fausse; elle suppose que la rigour exacte puisse
exister sans la facilité & c’est le contraire; plus une déduction est
rigoureause, plus elle est facile à entendre: car la rigueur consiste à reduire
tout aux principes les plus simples. D’où il s’ensuit encore que la rigueur
proprement dit entraîne nécessairement la méthode la plus naturelle & la
plus directe. Plus les principles seront disposés dans l’ordre convenable,
plus la déduction sera rigourease; ce n’est pas qu’absolument elle ne pût
l’être si on suivonit une méthode plus composée, com a fait Euclide dans ses
élémens: mais alors l’embarras de la marche feroit aisément sentir que cette
rigueur précaire & forcée ne seroit qu’improprement telle. »5
[“Secondly, one requests which of the two following qualities must be

preferred within the elements, whether the easiness or the exact rigour. I
answer that this question implies a falsehood; it implies that the exact rigour
can exist without the easiness and it is the other way around; the more
rigorous a deduction will be, the more easily it will be understood: because
the rigour consists of reducing everything to the simplest principles.
Whence follows that the properly called rigour implies necessarily the most
natural and direct method. The more the principles will be arranged in the
convenient order, the more rigorous the deduction will be; it does not mean
that it cannot be rigorous at all if one follows a more composite method as
Euclid made in his elements: but then the difficulty of the march will make
us to feel that this precarious and forced rigour will only be an improper
one.”]
Jean le Rond D'Alembert (1717-1783)
5
«Elémens des sciences» in Encyclopédie, ou dictionaire raisonné des sciences, des arts et des
métiers (París, 1755).
V
PREFACE TO THE FIRST ENGLISH EDITION
The first edition of the Treatise of Plane Geometry through Geometric Algebra
is a very enlarged translation of the first Catalan edition published in 1996. The good
reception of the book (now out of print) encouraged me to translate it to the English
language rewriting some chapters in order to make easier the reading, enlarging the
others and adding those devoted to the non-Euclidean geometry.
The geometric algebra is the tool which allows to study and solve geometric
problems through a simpler and more direct way than a purely geometric reasoning, that
is, by means of the algebra of geometric quantities instead of synthetic geometry. In
fact, the geometric algebra is the Clifford algebra generated by the Grassmann's outer
product in a vector space, although for me, the geometric algebra is also the art of
stating and solving geometric equations, which correspond to geometric problems, by
isolating the unknown geometric quantity using the algebraic rules of the vectors
operations (such as the associative, distributive and permutative properties). Following
Peano6:
“The geometric Calculus differs from the Cartesian Geometry in that

whereas the latter operates analytically with coordinates, the former
operates directly on the geometric entities”.
Initially proposed by Leibniz7 (characteristica geometrica) with the aim of

finding an intrinsic language of the geometry, the geometric algebra was discovered and
developed by Grassmann8, Hamilton and Clifford during the XIX century. However, it
did not become usual in the XX century ought to many circumstances but the vector
analysis -a recasting of the Hamilton quaternions by Gibbs and Heaviside- was
gradually accepted in physics. On the other hand, the geometry followed its own way
aside from the vector analysis as Gibbs9 pointed out:
“And the growth in this century of the so-called synthetic as opposed to

analytical geometry seems due to the fact that by the ordinary analysis
geometers could not easily express, except in a cumbersome and unnatural
manner, the sort of relations in which they were particularly interested”
6
Giuseppe Peano, «Saggio di Calcolo geometrico». Translated in Selected works of Giuseppe
Peano, 169 (see the bibliography).
7
C. I. Gerhardt, G. W. Leibniz. Mathematical Schriften V, 141 and Der Briefwechsel von
Gottfried Wilhelm Leibniz mit Mathematiker, 570.
8
In 1844 a prize (45 gold ducats for 1846) was offered by the Fürstlich Jablonowski'schen
Gessellschaft in Leipzig to whom was capable to develop the characteristica geometrica of
Leibniz. Grassmann won this prize with the memoir Geometric Analysis, published by this
society in 1847 with a foreword by August Ferdinand Möbius. Its contents are essentially those
of Die Ausdehnungslehre (1844).
9
Josiah Willard Gibbs, «On Multiple Algebra», reproduced in Scientific papers of J.W. Gibbs,
II, 98.
VI
The work of revision of the history and the sources (see J. M. Parra10) has
allowed us to synthesise the contributions of the different authors and completely
rebuild the evolution of the geometric algebra, removing the conceptual mistakes which
led to the vector analysis. This preface has not enough extension to explain all the
history11, but one must remember something usually forgotten: during the XIX century
several points of view over what should become the geometric algebra came into
competition. The Gibbs' vector analysis was one of these being not the better. In fact,
the geometric algebra is a field of knowledge where different formulations are possible
as Peano showed:
“Indeed these various methods of geometric calculus do not at all

contradict one another. They are various parts of the same science, or rather
various ways of presenting the same subject by several authors, each
studying it independently of the others.
It follows that geometric calculus, like any other method, is not a
system of conventions but a system of truth. In the same way, the methods
of indivisibles (Cavalieri), of infinitesimals (Leibniz) and of fluxions
(Newton) are the same science, more or less perfected, explained under
different forms.”12
The geometric algebra owns some fundamental geometric facts which cannot be
ignored at all and will be recognised to it, as Grassmann hoped:
“For I remain completely confident that the labour which I have

expanded on the science presented here and which has demanded a
significant part of my life as well as the most strenuous application of my
powers will not be lost. It is true that I am aware that the form which I have
given the science is imperfect and must be imperfect. But I know and feel
obliged to state (though I run the risk of seeming arrogant) that even if this
work should again remained unused for another seventeen years or even
longer, without entering into the actual development of science, still the
time will come when it will be brought forth from the dust of oblivion, and
when ideas now dormant will bring forth fruit. I know that if I also fail to
gather around me in a position (which I have up to now desired in vain) a
circle of scholars, whom I could fructify with these ideas, and whom I could
stimulate to develop and enrich further these ideas, nevertheless there will
come a time when these ideas, perhaps in a new form, will arise anew and
will enter into living communication with contemporary developments. For
truth is eternal and divine, and no phase in the development of truth,
however small may be the region encompassed, can pass on without leaving
10
Josep Manel Parra i Serra, «Geometric algebra versus numerical Cartesianism. The historical
trend behind Clifford’s algebra», in Brackx et al. ed., Clifford Algebras and their Applications
in Mathematical Physics, 307-316, .
11
A very complete reference is Michael J. Crowe, A History of Vector Analysis. The Evolution
of the Idea of a Vectorial System.
12
Giuseppe Peano, op. cit., 168.
VII
a trace; truth remains, even though the garment in which poor mortals clothe
it may fall to dust.”13
As any other aspect of the human life, the history of the geometric algebra was
conditioned by many fortuitous events. While Grassmann deduced the extension theory
from philosophic concepts unintelligible for authors such as Möbius and Gibbs,
Hamilton identified vectors and bivectors -the starting point of the great tangle of vector
analysis- using a heavy notation14. Clifford had found the correct algebraic structure15
which integrated the systems of Hamilton and Grassmann. However due to the
premature death of Clifford in 1879, his opinion was not taken into account16 and a long
epistolary war was carried out by the quaternionists (specially Tait) against the
defenders of the vector analysis, created by Gibbs17, who did not recognise to be
influenced by Grassmann and Hamilton:
“At all events, I saw that the methods which I was using, while
nearly those of Hamilton, were almost exactly those of Grassmann. I
procured the two Ed. of the Ausdehnungslehre but I cannot say that I found
them easy reading. In fact I have never had the perseverance to get through
with either of them, and have perhaps got more ideas from his
miscellaneous memoirs than from those works.
I am not however conscious that Grassmann's writings exerted any
particular influence on my Vector Analysis, although I was glad enough in
the introductory paragraph to shelter myself behind one or two distinguished
names [Grassmann and Clifford] in making changes of notation which I felt
would be distasteful to quaternionists. In fact if you read that pamphlet
carefully you will see that it all follows with the inexorable logic of algebra
from the problem which I had set myself long before my acquaintance with
Grassmann.
I have no doubt that you consider, as I do, the methods of Grassmann
to be superior to those of Hamilton. It thus seemed to me that it might [be]
interesting to you to know how commencing with some knowledge of
Hamilton's method and influenced simply by a desire to obtain the simplest
algebra for the expression of the relations of Geom. Phys. etc. I was led
essentially to Grassmann's algebra of vectors, independently of any
influence from him or any one else.”18
13
Hermann Gunther Grassmann. Preface to the second edition of Die Ausdehnungslehre (1861).
The first edition was published on 1844, hence the "seventeen years". Translated in Crowe, op.
cit. p. 89.
14
The Lectures on Quaternions was published in 1853, and the Elements of Quaternions
posthumously in 1866.
15
William Kingdon Clifford left us his synthesis in «Applications of Grassmann's Extensive
Algebra».
16
See «On the Classification of Geometric Algebras», unfinished paper whose abstract was
communicated to the London Mathematical Society on March 10th, 1876.
17
The first Vector Analysis was a private edition of 1881.
18
Draft of a letter sent by Josiah Willard Gibbs to Victor Schlegel (1888). Reproduced by
Crowe, op. cit. p. 153.
VIII
Before its beginning the controversy was already superfluous19. Notwithstanding
the epistolary war continued for twelve years.
The vector analysis is a provisional solution20 (which spent all the XX century!)
adopted by everybody ought to its easiness and practical notation but having many
troubles when being applied to three-dimensional geometry and unable to be
generalised to the Minkowski’s four-dimensional space. On the other hand, the
geometric algebra is, by its own nature, valid in any dimension and it offers the
necessary resources for the study and research in geometry as I show in this book.
The reader will see that the theoretical explanations have been completed with
problems in each chapter, although this splitting is somewhat fictitious because the
problems are demonstrations of geometric facts, being one of the most interesting
aspects of the geometric algebra and a proof of its power. The usual numeric problems,
which our pupils like, can be easily outlined by the teacher, because the geometric
algebra always yields an immediate expression with coordinates.
I'm indebted to professor Josep Manel Parra for encouraging me to write this
book, for the dialectic interchange of ideas and for the bibliographic support. In the
framework of the summer courses on geometric algebra for teachers that we taught
during the years 1994-1997 in the Escola d’estiu de secundària organised by the
Col·legi Oficial de Doctors i Llicenciats en Filosofia i Lletres i en Ciències de
Catalunya, the project of some books on this subject appeared in a natural manner. The
first book devoted to two dimensions already lies on your hands and will probably be
followed by other books on the algebra and geometry of the three and four dimensions.
Finally I also acknowledge the suggestions received from some readers.
Cerdanyola del Vallès, June 2001
19
See Alfred M. Bork «“Vectors versus quaternions”—The letters in Nature».
20
The vector analysis bases on the duality of the geometric algebra of the three-dimensional
space: the fact that the orientation of lines and planes is determined by three numeric
components in both cases. However in the four-dimensional time-space the same orientations
are respectively determined by four and six numbers.
IX
CONTENTS
First Part: The vector plane and the complex numbers
1. The vectors and their operations. (June 24th 2000, updated March 17th 2002)
Vector addition, 1.- Product of a vector and a real number, 2.- Product of two vectors,
2.- Product of three vectors: associative property, 5. Product of four vectors, 7.- Inverse
and quotient of two vectors, 7.- Hierarchy of algebraic operations, 8.- Geometric algebra
of the vector plane, 8.- Exercises, 9.
2. A base of vectors for the plane. (June 24th 2000)

Linear combination of two vectors, 10.- Base and components, 10.- Orthonormal bases,
11.- Applications of the formulae for the products, 11.- Exercises, 12.
3. The complex numbers. (August 1st 2000, updated July 21st 2002)
Subalgebra of the complex numbers, 13.- Binomial, polar and trigonometric form of a
complex number, 13.- Algebraic operations with complex numbers, 14.- Permutation of
complex numbers and vectors, 17.- The complex plane, 18.- Complex analytic
functions, 19.- The fundamental theorem of algebra, 24.- Exercises, 26.
4. Transformations of vectors. (August 4th 2000, updated July 21st 2002)

Rotations, 27.- Reflections, 28.- Inversions, 29.- Dilatations, 30.- Exercises, 30
Second Part: The geometry of the Euclidean plane
5. Points and straight lines. (August 19th 2000, updated September 29th 2000)
Translations, 31.- Coordinate systems, 31.- Barycentric coordinates, 33.- Distance
between two points and area, 33.- Condition of alignment of three points, 35.- Cartesian
coordinates, 36.- Vectorial and parametric equations of a line, 36.- Algebraic equation
and distance from a point to a line, 37.- Slope and intercept equations of a line, 40.-
Polar equation of a line, 40.- Intersection of two lines and pencil of lines, 41.- Dual
coordinates, 43.- The Desargues theorem, 47.- Exercises, 50.
6. Angles and elemental trigonometry. (August 24th 2000, updated July 21st 2002)
Sum of the angles of a polygon, 53.- Definition of trigonometric functions and
fundamental identities, 54.- Angle inscribed in a circle and double angle identities, 55.-
Addition of vectors and sum of trigonometric functions, 56.- Product of vectors and
addition identities, 57.- Rotations and De Moivre's identity, 58.- Inverse trigonometric
functions, 59.- Exercises, 60.
7. Similarities and single ratio. (August 26th 2000, updated July 21st 2002)
Direct similarity, 61.- Opposite similarity, 62.- The theorem of Menelaus, 63.- The
theorem of Ceva, 64.- Homothety and single ratio, 65.- Exercises, 67.
8. Properties of the triangles. (September 3rd 2000, updated July 21st 2002)
Area of a triangle, 68.- Medians and centroid, 69.- Perpendicular bisectors and
circumcentre, 70.- Angle bisectors and incentre, 72.- Altitudes and orthocentre, 73.-
Euler's line, 76.- The Fermat's theorem, 77.- Exercises, 78.
X
9. Circles. (October 8th 2000, updated July 16th 2002)
Algebraic and Cartesian equations, 80.- Intersections of a line with a circle, 80.- Power
of a point with respect to a circle, 82.- Polar equation, 82.- Inversion with respect to a
circle, 83.- The nine-point circle, 85.- Cyclic and circumscribed quadrilaterals, 87.-
Angle between circles, 89.- Radical axis of two circles, 89.- Exercises, 91.
10. Cross ratios and related transformations. (October 18th 2000, updated July 21st
2002)
Complex cross ratio, 92.- Harmonic characteristic and ranges, 94.- The homography
(Möbius transformation), 96.- Projective cross ratio, 99.- The points at the infinity and
homogeneous coordinates, 102.- Perspectivity and projectivity, 103.- The projectivity as
a tool for theorems demonstration, 108.- The homology, 110.- Exercises, 115.
11. Conics (November 12th 2000, updated July 21st 2002)

Conic sections, 117.- Two foci and two directrices, 120.- Vectorial equation, 121.- The
Chasles' theorem, 122.- Tangent and perpendicular to a conic, 124.- Central equations
for the ellipse and hyperbola, 126.- Diameters and Apollonius' theorem, 128.- Conic
passing through five points, 131.- Conic equations in barycentric coordinates and
tangential conic, 132.- Polarities, 134.- Reduction of the conic matrix to a diagonal
form, 136.- Using a base of points on the conic, 137.- Exercises, 137.
Third Part: Pseudo-Euclidean geometry
12. Matrix representation and hyperbolic numbers. (November 22nd 2000, updated
May 31st 2002)
Rotations and the representation of complex numbers, 139.- The subalgebra of the
hyperbolic numbers, 140.- Hyperbolic trigonometry, 141.- Hyperbolic exponential and
logarithm, 143.- Polar form, powers and roots of hyperbolic numbers, 144.- Hyperbolic
analytic functions, 147.- Analyticity and square of convergence of the power series,
150.- About the isomorphism of Clifford algebras, 152.- Exercises, 153.
13. The hyperbolic or pseudo-Euclidean plane (January 1st 2001, updated July 21st
2002)
Hyperbolic vectors, 154.- Inner and outer products of hyperbolic vectors, 155.- Angles
between hyperbolic vectors, 156.- Congruence of segments and angles, 158.-
Isometries, 158.- Theorems about angles, 160.- Distance between points, 160.- Area on
the hyperbolic plane, 161.- Diameters of the hyperbola and Apollonius' theorem, 163.-
The law of sines and cosines, 164.- Hyperbolic similarity, 167.- Power of a point with
respect to a hyperbola with constant radius, 168.- Exercises, 169.
Fourth Part: Plane projections of tridimensional spaces
14. Spherical geometry in the Euclidean space. (March 3rd 2001, updated August 25th
2001)
The geometric algebra of the Euclidean space, 170.- Spherical trigonometry, 172.- The
dual spherical triangle, 175.- Right spherical triangles and Napier’s rule, 176.- Area of a
spherical triangle, 176.- Properties of the projections of the spherical surface, 177.- The
XI
central or gnomonic projection, 177.- Stereographic projection, 180.- Orthographic
projection, 181.- Spherical coordinates and cylindrical equidistant (Plate Carré)
projection, 182.- Mercator's projection, 183.- Peter's projection, 184.- Conic projections,
184.- Exercises, 185.
15. Hyperboloidal geometry in the pseudo-Euclidean space (Lobachevsky's

geometry). (April 13th 2001, updated August 21st 2001)
The geometric algebra of the pseudo-Euclidean space, 188.- The hyperboloid of two
sheets, 190.- The central projection (Beltrami disk), 191.- Hyperboloidal
(Lobachevskian) trigonometry, 196.- Stereographic projection (Poincaré disk), 198.-
Azimuthal equivalent projection, 200.- Weierstrass coordinates and cylindrical
equidistant projection, 201.- Cylindrical conformal projection, 202.- Cylindrical
equivalent projection, 203.- Conic projections, 203.- On the congruence of geodesic
triangles, 205.- Comment about the names of the non-Euclidean geometry, 205.-
Exercises, 205.
16. Solutions of the proposed exercises. (April 28th 2001 and May 27th 2001, updated
July 20th 2002)
1. The vectors and their operations, 207.- 2. A base of vectors for the plane, 208.- 3. The
complex numbers, 209.- 4. Transformations of vectors, 213.- 5. Points and straight
lines, 214.- 6. Angles and elemental trigonometry, 223.- 7. Similarities and single ratio,
226.- 8. Properties of the triangles, 228.- 9. Circles, 236.- 10. Cross ratios and related
transformations, 240.- 11. Conics, 245.- 12. Matrix representation and hyperbolic
numbers, 250.- 13. The hyperbolic or pseudo-Euclidean plane, 251.- 14. Spherical
geometry in the Euclidean space, 254.- 15. Hyperboloidal geometry in the pseudo-
Euclidean space (Lobachevsky's geometry), 260.
Bibliography, 266.
Index, 270.
Chronology, 275.
XII
TREATISE OF PLANE GEOMETRY THROUGH GEOMETRIC ALGEBRA 1
FIRST PART: THE VECTOR PLANE AND THE COMPLEX NUMBERS
Points and vectors are the main elements of the plane geometry. A point is
conceived (but not defined) as a geometric element without extension, infinitely small,
that has position and is located at a certain place on the plane. A vector is defined as an
oriented segment, that is, a piece of a straight line having length and direction. A vector
has no position and can be translated anywhere. Usually it is called a free vector. If we
place the end of a vector at a point, then its head determines another point, so that a
vector represents the translation from the first point to the second one.
Taking into account the distinction between points and vectors, the part of the
book devoted to the Euclidean geometry has been divided in two parts. In the first one
the vectors and their algebraic properties are studied, which is enough for many
scientific and engineering branches. In the second part the points are introduced and
then the affine geometry is studied.
All the elements of the geometric algebra (scalars, vectors, bivectors, complex
numbers) are denoted with lowercase Latin characters and the angles with Greek
characters. The capital Latin characters will denote points on the plane. As you will see,
the geometric product is not commutative, so that fractions can only be written for real
and complex numbers. Since the geometric product is associative, the inverse of a
certain element at the left and at the right is the same, that is, there is a unique inverse
for each element of the algebra, which is indicated by the superscript −1. Also due to the
associative property, all the factors in a product are written without parenthesis. In order
to make easy the reading I have not numerated theorems, corollaries nor equations.
When a definition is introduced, the definite element is marked with italic characters,
which allows to direct attention and helps to find again the definition.
1. THE VECTORS AND THEIR OPERATIONS
A vector is an oriented segment, having length and direction but no position, that
is, it can be placed anywhere without changing its orientation. The vectors can represent
many physical magnitudes such as a force, a celerity, and also geometric magnitudes
such as a translation.
Two algebraic operations for vectors are defined, the addition and the product,
that generalise the addition and product of the real numbers.
Vector addition
The addition of two vectors u + v

is defined as the vector going from the
end of the vector u to the head of v when
the head of u contacts the end of v (upper
triangle in the figure 1.1 ). Making the
construction for v + u, that is, placing the
end of u at the head of v (lower triangle
in the figure 1.1) we see that the addition
vector is the same. Therefore, the vector
addition has the commutative property:
Figure 1.1
2 RAMON GONZALEZ CALVET
u+v=v+u
and the parallelogram rule follows: the addition of two vectors is the diagonal of the
parallelogram formed by both vectors.
The associative property follows from
this definition because (u+v)+w or
u+(v+w) is the vector closing the
polygon formed by the three vectors as
shown in the figure 1.2.
The neutral element of the
vector addition is the null vector, which
has zero length. Hence the opposite
vector of u is defined as the vector −u
with the same orientation but opposite
direction, which added to the initial Figure 1.2
vector gives the null vector:
u + ( −u) = 0
Product of a vector and a real number
One defines the product of a vector and a real number (or scalar) k, as a vector
with the same direction but with a length increased k times (figure 1.3). If the real
number is negative, then the direction is the opposite. The geometric definition implies
the commutative property:
ku=uk Figure 1.3

Two vectors u, v with the same
direction are proportional because there
is always a real number k such that v =
k u , that is, k is the quotient of both
vectors:
k = u −1 v = v u −1
Two vectors with different

directions are said to be linearly
independent.
Product of two vectors
The product of two vectors will be called the geometric product in order to be
distinguished from other vector products currently used. Nevertheless I hope that these
other products will play a secondary role when the geometric product becomes the most
used, a near event which this book will forward. At that time, the adjective «geometric»
will not be necessary.
The following properties are demanded to the geometric product of two vectors:
1) To be distributive with regard to the vector addition:
u(v+w)=uv+uw
2) The square of a vector must be equal to the square of its length. By

definition, the length (or modulus) of a vector is a positive number and it is
noted by | u |:
u2 = | u |2
3) The mixed associative property must exist between the product of vectors
and the product of a vector and a real number.
k(uv)=(ku)v= kuv
k(lu)=(kl)u= klu
where k, l are real numbers and u, v vectors. Therefore, parenthesis are not
needed.
These properties allows us to deduce the product. Let us suppose that c is the
addition of two vectors a, b and calculate its square applying the distributive property:
c=a+b
c2 = ( a + b )2 = ( a + b ) ( a + b ) = a2 + a b + b a + b2
We have to preserve the order of the factors because we do not know whether the
product is commutative or not.
If a and b are orthogonal vectors, the Pythagorean theorem applies and then:
a⊥b ⇒ c2 = a2 + b2 ⇒ ab+ba=0 ⇒ ab=−ba
That is, the product of two perpendicular vectors is anticommutative.

If a and b are proportional vectors then:
a || b ⇒ b = k a, k real ⇒ ab=aka=kaa=ba
because of the commutative and mixed associative properties of the product of a vector
and a real number. Therefore the product of two proportional vectors is commutative. If
c is the addition of two vectors a, b with the same direction, we have:
|c|=|a|+|b|
c2 = a2 + b2 + 2 | a | | b |
ab=|a||b| angle(a, b) = 0
But if the vectors have opposite directions:
|c|=|a|−|b|
c2 = a2 + b2 − 2 | a | | b |
ab=−|a||b| angle(a, b) = π
How is the product of two vectors with any directions? Due to the distributive
property the product is resolved into one product by the proportional component b|| and
another by the orthogonal component b⊥:
a b = a ( b|| + b⊥ ) = a b|| + a b⊥
The product of one vector by the proportional component of the other is called
the inner product (also scalar product) and noted by a point · (figure 1.4). Taking into
account that the projection of b onto a is proportional to the cosine of the angle between
both vectors, one finds:
a · b = a b|| = | a | | b | cos α
Figure 1.4
The inner product is always a real
number. For example, the work made by a
force acting on a body is the inner product of
the force and the walked space. Since the
commutative property has been deduced for
the product of vectors with the same
direction, it follows also for the inner
product:
a·b=b·a
The product of one vector by the orthogonal component of the other is called the
outer product (also exterior product) and it is noted with the symbol ∧ :
a ∧ b = a b⊥
The outer product represents the area of the parallelogram formed by both vectors
(figure 1.5):
 a ∧ b =  a b⊥ = absin α
Figure 1.5
Since the outer product is a product of
orthogonal vectors, it is anticommutative:
a∧b=−b∧a
Some example of physical

magnitudes which are outer products are the
angular momentum, the torque, etc.
When two vectors are permuted, the
sign of the oriented angle is changed. Then the cosine remains equal while the sine
changes the sign. Because of this, the inner product is commutative while the outer
product is anticommutative. Now, we can rewrite the geometric product as the sum of
both products:
ab=a·b+a∧b
Figure 1.6
From here, the inner and outer
products can be written using the geometric
product:
ab+ba
a·b =
2
ab−ba
a ∧b=
2
In conclusion, the geometric product of two proportional vectors is commutative

whereas that of two orthogonal vectors is anticommutative, just for the pure cases of
outer and inner products. The outer, inner and geometric products of two vectors only
depend upon the moduli of the vectors and the angle between them. When both vectors
are rotated preserving the angle that they form, the products are also preserved (figure
1.6).
How is the absolute value of the product of two vectors? Since the inner and
outer product are linearly independent and orthogonal magnitudes, the modulus of the
geometric product must be calculated through a generalisation of the Pythagorean
theorem:
ab=a·b+a∧b ⇒  a b 2 = a · b2 +a ∧ b2
 a b 2 =  a 2  b 2 ( cos2α + sin2α ) =  a 2  b 2
That is, the modulus of the geometric product is the product of the modulus of
each vector:
 a b = a  b 
Product of three vectors: associative property
It is demanded as the fourth property that the product of three vectors be

associative:
4) u(vw)=(uv)w=uvw
Hence we can remove parenthesis in multiple products and with the foregoing
properties we can deduce how the product operates upon vectors.
We wish to multiply a vector a by a product of two vectors b, c. We ignore the
result of the product of three vectors with different orientations except when two
adjacent factors are proportional. We have seen that the product of two vectors depends
only on the angle between them. Therefore the parallelogram formed by b and c can be
rotated until b has, in the new orientation, the same direction as a. If b' and c' are the
vectors b and c with the new orientation (figure 1.7) then:
b c = b' c'
a ( b c ) = a ( b' c' )
Figure 1.7
and by the associative property:
a ( b c ) = ( a b' ) c'
Since a and b' have the same

direction, a b' = ab is a real
number and the triple product is a
vector with the direction of c' whose
length is increased by this amount:
a ( b c ) = ab c'
It follows that the modulus of the product of three vectors is the product of their
moduli:
a b c = abc
On the other hand, a can be Figure 1.8

firstly multiplied by b, and after this we
can rotate the parallelogram formed by
both vectors until b has, in the new
orientation, the same direction as c
(figure 1.8). Then:
( a b ) c = a'' ( b'' c ) = a'' bc
Although the geometric

construction differs from the foregoing
one, the figures clearly show that the
triple product yields the same vector, as expected from the associative property.
( a b ) c = a'' bc = cb a'' = c b'' a'' = c ( b a )
That is, the triple product has the property:
abc=cba
which I call the permutative property: every vector can be permuted with a vector
located two positions farther in a product, although it does not commute with the
neighbouring vectors. The permutative property implies that any pair of vectors in a
product separated by an odd number of vectors can be permuted. For example:
abcd=adcb=cdab=cbad
The permutative property is characteristic of the plane and it is also valid for the
space whenever the three vectors are coplanar. This property is related with the fact that
the product of complex numbers is commutative.
Product of four vectors
The product of four vectors can be deduced from the former reasoning. In order
to multiply two pair of vectors, rotate the parallelogram formed by a and b until b' has
the direction of c. Then the product is the parallelogram formed by a' and d but
increased by the modulus of b and c:
a b c d = a' b' c d = a'  b c d = b c a' d
Now let us see the special case when a = c and b = d. If both vectors a, b have the same
direction, the square of their product is a
positive real number:
a || b ( a b ) 2 = a2 b 2 > 0
If both vectors are perpendicular, we must

rotate the parallelogram through π/2 until b'
has the same direction as a (figure 1.9). Then
a' and b are proportional but having opposite
signs. Therefore, the square of a product of
two orthogonal vectors is always negative:
Figure 1.9
a⊥b ( a b )2 = a' b' a b = a'bab = −a2 b2 <0
Likewise, the square of an outer product of any two vectors is also negative.
Inverse and quotient of two vectors
The inverse of a vector a is that vector whose multiplication by a gives the unity.
Only the vectors which are proportional have a real product. Hence the inverse vector
has the same direction and inverse modulus:
a −1 = aa −2 ⇒ a −1 a = a a −1 = 1
The quotient of vectors is a product for an inverse vector, which depends on the
order of the factors because the product is not commutative:
a −1 b ≠ b a −1
Obviously the quotient of proportional vectors with the same direction and sense
is equal to the quotient of their moduli. When the vectors have different directions, their
quotient can be represented by a parallelogram, which allows to extend the concept of
vector proportionality. We say that a is proportional to c as b is to d when their moduli

are proportional and the angle between a and c is equal to the angle between b and d1:
a c −1 = b d −1 ⇔ ac−1 =bd−1 and α(a, c) = α(b, d)
Then the parallelogram formed by a and b is similar to that formed by c and d,

being α(a, c) the angle of rotation from the first to the second one.
The inverse of a product of several vectors is the product of the inverses with the
exchanged order, as can be easily seen from the associative property:
( a b c ) −1 = c −1 b −1 a −1
Hierarchy of algebraic operations
Like the algebra of real numbers, and in order to simplify the algebraic notation,
I shall use the following hierarchy for the vector operations explained above:
1) The parenthesis, whose content will be firstly operated.
2) The power with any exponent (square, inverse, etc.).
3) The outer and inner product, which have the same hierarchy level but must
be operated before the geometric product.
4) The geometric product.
5) The addition.
As an example, some algebraic expressions are given with the simplified

expression at the left hand and its meaning using parenthesis at the right hand:
a∧bc∧d=(a∧b)(c∧d)
a2 b ∧ c + 3 = ( ( a2 ) ( b ∧ c ) ) + 3
a+b·cde=a+((b·c)de)
Geometric algebra of the vectorial plane
The set of all the vectors on the plane together with the operations of vector
addition and product of vectors by real numbers is a two-dimensional space usually
called the vector plane V2. The geometric product generates new elements (the complex
numbers) not included in the vector plane. So, the geometric (or Clifford) algebra of a
vectorial space is defined as the set of all the elements generated by products of vectors,
for which the geometric product is an inner operation. The geometric algebra of the
Euclidean vector plane is usually noted as Cl2,0(R) or simply as Cl2. Making a
parallelism with probability, the sample space is the set of elemental results of a certain
1
William Rowan Hamilton defined the quaternions as quotients of two vectors in the way that
similar parallelograms located at the same plane in the space represent the same quaternion
(Elements of Quaternions, posthumously edited in 1866, Chelsea Publishers 1969, vol I, see p.
113 and fig. 34). In the vectorial plane a quaternion is reduced to a complex number. The
quaternions were discovered by Hamilton (October 16th, 1843) before the geometric product by
Clifford (1878).
random experiment. From the sample space Ω, the union ∪ and intersection ∩ generate
the Boole algebra A(Ω), which includes all the possible events. In the same manner, the
addition and geometric product generate the geometric algebra of the vectorial space.
Then both sample and vectorial space play similar roles as generators of the Boole and
geometric (Clifford) algebras respectively.
Exercises
1.1 Prove that the sum of the squares of the diagonals of any parallelogram is equal to
the sum of the squares of the four sides. Think about the sides as vectors.
1.2 Prove the following identity:
( a · b )2 − ( a ∧ b )2 = a2 b2
1.3 Prove that: a∧bc∧d=a·(bc∧d)=(a∧bc)·d
1.4 Prove that: a ∧ b c ∧ d + a ∧ c d ∧ b+ a ∧ d b ∧ c = 0
1.5 Prove the permutative property resolving b and c into the components which are
proportional and orthogonal to the vector a.
1.6 Prove the Heron’s formula for the area of the triangle:
A = s (s − a ) ( s − b ) ( s − c )
where a, b and c are the sides and s the semiperimeter:
a + b + c
s=
2
2. A BASE OF VECTORS FOR THE PLANE
Linear combination of two vectors
Every vector w on the plane is always a linear combination of two independent

vectors u and v:
w=au+bv a, b real
Because of this, the plane has dimension equal to 2. In order to calculate the coefficients
of linear combination a and b, we multiply w by u and v at both sides and subtract the
results:
u w = a u2 + b u v and w u = a u2 + b v u ⇒ uw−wu=b(uv−vu)
v w = a v u + b v2 and w v = a u v + b v2 ⇒ wv−vw=a(uv−vu)
to obtain:
w∧v u∧w
a= b=
u∧v u∧v
The resolution of a vector as a linear combination of two independent vectors is a

very frequent operation and also the foundation of the coordinates method.
Base and components
Any set of two independent vectors {e1 , e2} can be taken as a base of the vector
plane. Every vector u can be written as linear combination of the base vectors:
u = u1 e1 + u2 e2
The coefficients of this linear combination u1, u2 are the components of the vector in this
base. Then a vector will be represented as a pair of components:
u = ( u1, u2 )
The components depend on the base, so that a change of base leads to a change of the
components of the given vector.
We must only add components to add vectors:
v = ( v1, v2 )
u + v = ( u1 + v1 , u2 + v2)
The expression of the geometric product with components is obtained by means of

the distributive property:
u v = ( u1 e1 + u2 e2 ) ( v1 e1 + v2 e2 ) = u1 v1 e12 + u2 v2 e22 + u1 v2 e1 e2+ u2 v1 e2 e1
u v = u1 v1 | e1 |2 + u2 v2 | e2 |2 +( u1 v2 + u2 v1 ) e1 · e2 + ( u1 v2 − u2 v1 ) e1 ∧ e2
Hence the expression of the square of the vector modulus written in components is:
| u |2 = u2 = u12 | e1 |2 + u22 | e2 |2 + 2 u1 u2 e1 · e2
Orthonormal bases
Any base is valid to describe vectors using

components, although the orthonormal bases, for Figure 2.1
which both e1 and e2 are unitary and perpendicular
(such as the canonical base shown in the figure 2.1),
are the more convenient and suitable:
e1 ⊥ e2 e1 = e2 = 1
For every orthonormal base :
e12 = e22 = 1 e1 e2 = − e2 e1
The product e1 e2 represents a square of unity area. The square power of this product is
equal to −1:
( e1 e2 )2 = e1 e2 e1 e2 = − e1 e1 e2 e2 = −1
For an orthonormal base, the geometric product of two vectors becomes:
u v = u1 v1 + u2 v2 + ( u1 v2 − u2 v1 ) e1 e2
Note that the first and second summands are real while the third is an area. Therefore it
follows that they are respectively the inner and outer products:
u · v = u1 v1 + u2 v2 u ∧ v = ( u1 v2 − u2 v1 ) e1 e2
Also, the modulus of a vector is calculated from the self inner product:
u2 = u12 + u22
Applications of the formulae for the products
The first application is the calculation of the angle between two vectors:
u1 v1 + u 2 v 2 u1 v 2 − u 2 v 1
cos α = sin α =
u v u v
The values of sine and cosine determine a unique angle α in the range 0<α<2π.
The angle between two vectors is a sensed magnitude having positive sign if it is
counterclockwise and negative sign if it is clockwise. Thus this angle depends on the
order of the vectors in the outer (and geometric) product. For example, let us consider the
vectors (figure 2.2) u and v:
u = (−2, 2 ) u =2 2 v=(4,3) v =5
1 7
cos α (u, v ) = cos α (v, u ) = − sin α (u, v ) = −sin α (v, u ) = −
5 2 5 2
α (u, v) = 4.5705... α( v, u ) = 1.7127...
Also we may take the angle α( u, v ) = 4.5705...−2π = −1.7127... The angle so obtained is
always that going from the first to the second vector, being unique within a period.
Other application of the outer product is the calculus of areas. Using the
expression with components, the area
(considered as a positive real number) Figura 2.2
of the parallelogram formed by u and v
is:
A = u ∧ v = u1 v2 − u2 v1= 14
When calculating the area of

any triangle we must only divide the
outer product of any two sides by 2.
Exercises
2.1 Let (u1, u2) and (v1, v2) be the

components of the vectors u and v in
the canonical base. Prove geometrically that the area of the parallelogram formed by both
vectors is the modulus of the outer product u ∧ v = u1 v2 − u2 v1 .
2.2 Calculate the area of the triangle whose sides are the vectors (3, 5), (−2, −3) and their
addition (1, 2).
2.3 Prove the permutative property using components: a b c = c b a.
2.4 Calculate the angle between the vectors u = 2 e1 + 3 e2 and v = −3 e1 + 4 e2 in the

canonical base.
2.5 Consider a base where e1 has modulus 1, e2 has modulus 2 and the angle between both
vectors is π/3. Calculate the angle between u = 2 e1 + 3 e2 and v = −3 e1 + 4 e2 .
2.6 In the canonical base v = (3, −5). Calculate the components of this vector in a new
base { u1, u2 } if u1 = (2, −1) and u2 = (5, −3).
3. THE COMPLEX NUMBERS
Subalgebra of the complex numbers
If {e1, e2} is the canonical base of the vector plane V2, its geometric algebra is
defined as the vector space generated by the elements {1, e1, e2, e1e2} together with the
geometric product, so that the geometric algebra Cl2 has dimension four. The unitary area
e1e2 is usually noted as e12. Due to the associative character of the geometric product, the
geometric algebra is an associative algebra with identity. The complete table for the
geometric product is the following:
1 e1 e2 e12
_______________________
1  1 e1 e2 e12

e1  e1 1 e12 e2

e2  e2 −e12 1 −e1

e12  e12 −e2 e1 −1
Note that the subset of elements containing only real numbers and areas is closed
for the product:
1 e12
___________
1  1 e12

e12  e12 −1
This is the subalgebra of complex numbers. e12 , the imaginary unity, is usually noted as i.
They are called complex numbers because their product is commutative like for the real
numbers.
Binomial, polar and trigonometric form of a complex number
Every complex number z written in the binomial form is:
z = a + b e12 a, b real
where a and b are the real and imaginary components respectively. The modulus of a
complex number is calculated in the same way as the modulus of any element of the
geometric algebra by means of the Pythagorean theorem:
z2 =  a + b e122 = a2 + b2
Since every complex number can be written as a product of two vectors u and v
forming a certain angle α:
z = u v = u v ( cos α + e12 sin α ) Figure 3.1
we may represent a complex number as

a parallelogram with sides being the
vectors u and v. But there are infinite
pairs of vectors u' and v' whose product
is the complex z provided that:
uv = u'v' and α = α'
All the parallelograms having

the same area and obliquity represent a
given complex (they are equivalent) independently of the length and orientation of one
side.(figure 3.1). The trigonometric and polar forms of a complex number z specifies its
modulus zand argument α :
z = z ( cos α + e12 sin α ) = z α
A complex number can be written using the exponential function, but firstly we
must prove the Euler’s identity:
exp(α e12) = cos α + e12 sin α α real
The exponential of an imaginary number is

defined in the same manner as for a real Figure 3.2
number:
n
exp(α e12 ) = lim 1 + α e12 
n → ∞ n 
As shown in the figure 3.2 (for n = 5), the limit

is a power of n rotations with angle α,/n or
equivalently a rotation of angle α.
Now a complex number written in
exponential form is:
z = z exp(α e12 )
Algebraic operations with complex numbers
Each algebraic operation is more easily calculated in a form than in another

according to the following scheme:
addition / subtraction ↔ binomial form

product / quotient ↔ binomial or polar form
powers / roots ↔ polar form
The binomial form is suitable for the addition because both real components must
be added and also the imaginary ones, e.g.:
z = 3 + 4 e12 t = 2 − 5 e12
z + t = 3 + 4 e12 + 2 − 5 e12 = 5 − e12
If the complex numbers are written in another form, they must be converted to the
binomial form, e.g.
z = 23π / 4 t = 4π / 6
 3π 3π   π π
z + t = 2 3π / 4 + 4 π / 6 = 2  cos + e12 sin  + 4  cos + e12 sin 
 4 4   6 6

= 2  −
2
+ e12
2  3
+4
 
1
+ e12  = − 2 + 2 3 + e12 ( 2+2 )
 2 2   2 2
In order to write the result in polar form, the modulus must be calculated:
 z + t 2 = ( − 2 + 2 3 )2 + ( 2 + 2 )2 = 20 + 4 ( 2 − 6 )
 z + t  = 2 5 + 2 − 6 = 3.9823...
and also the argument from the cosine and sine obtained as quotient of the real and
imaginary components respectively divided by the modulus:
− 2+2 3 2+2
cos α = sinα = α = 1.0301...
2 5+ 2 − 6 2 5+ 2 − 6
z + t = 3.98231.0301
When multiplying complex numbers in binomial form, we apply the distributive

property taken into account that the square of the imaginary unity is −1, e.g.:
(−2 + 5 e12 ) ( 3 −4 e12 ) = −6 + 8 e12 + 15 e12 − 20 ( e12 )2 = −6 + 20 + 23 e12 = 14 + 23 e12
The exponential of an addition of arguments (real or complex) is equal to the

product of exponential functions of each argument. Applying this characteristic property
to the product of complex numbers in exponential form, we have:
z t = z exp(α e12 ) t exp( β e12 ) = z t exp[ (α + β ) e12 ]
from where the product of complex numbers in polar form is obtained by multiplying
both moduli and adding both arguments.:
zα tβ = z t α + β
One may subtract 2π to the

resulting argument in order to keep it
between 0 and 2π. The product of
two complex numbers z and t is the
geometric operation consisting in the
rotation of the parallelogram
representing the first complex
number until it touches the Figure 3.3
parallelogram representing the second
complex number. When they contact
in a unitary vector v (figure 3.3), the parallelogram formed by the other two vectors is the
product of both complex numbers:
z=uv t=vw v2 = 1
zt=uvvw=uw
This geometric construction is always possible because a parallelogram can be

lengthened or widened maintaining the area so that one side has unity length.
The conjugate of a complex number (symbolised with an asterisk) is that number
whose imaginary part has opposite sign:
z = a + b e12 z* = a − b e12
The geometric meaning of the

conjugation is a permutation of the
vectors whose product is the complex
number (figure 3.4). In this case, the
inner product is preserved while the
outer product changes the sign. The
product of a complex number and its
conjugate is the square of the modulus: Figure 3.4
z z* = u v v u = u2 = z2
The quotient of complex numbers is defined as the product by the inverse. The
inverse of a complex number is equal to the conjugate divided by the square of the
modulus:
z * a − b e12
z −1 = = 2
z
2
a + b2
Then, let us see an example of quotient of complex numbers:
− 2 + 5 e12 (− 2 + 5 e12 )(3 + 4 e 12 ) − 26 + 7 e12

= =
3 − 4 e12 3 2 + (− 4 )
2
25
With the polar form, the quotient is obtained by dividing moduli and subtracting
arguments:
z  z 
α
=  

t β  t α − β
The best way to calculate the power of a complex number with natural exponent is
through the polar form, although for low exponents the binomial form and the Newton
formula is often used, e.g.:
( 2 − 3 e12 )3 = 23 − 3 · 22 e12 + 3 · 2 e122 − e123 = 8 − 12 e12 − 6 + e12 = 2 − 11 e12
From the characteristic property of the exponential function it follows that the
modulus of a power of a complex number is the power of its modulus, and the argument
of this power is the argument of the complex number multiplied by the exponent:
( z α
)n
= z
n
nα
This is a very useful rule for large exponents, e.g.:
( 2 + 2 e12 )1000 = ( 2 2 π/4 )1000 = [( 2 2 )1000]250 π = 215000 = 21500
When the argument exceeds 2π, divide by this value and take the remainder, in order to
have the argument within the period 0<α <2π.
Since a root is the inverse operation of a power, its value is obtained by extracting
the root of the modulus and dividing the argument by the index n. But a complex number
of argument α may be also represented by the arguments α +2 π k. Their division by the
index n yields n different arguments within a period, corresponding to n different roots:
n zα = n z k = 0, ... n −1
(α + 2π k ) / n
For example, the cubic roots of 8 are 3 8 0 = { 2 0 , 2 2π / 3 , 2 4π/3 } . In the complex plane, the
n-th roots of every complex number are located at the n vertices of a regular polygon.
Permutation of complex numbers and vectors
The permutative property of the vectors is intimately related with the commutative
property of the product of complex numbers. Let z and t be complex numbers and a, b, c
and d vectors fulfilling:
z=ab t=cd
Then the following equalities are equivalent:
zt=tz ⇔ abcd=cdab
A complex number z and a vector c do not commute, but they can be permuted by
conjugating the complex number:
z c = a b c = c b a = c z*
Every real number commute with any vector. However every imaginary number
anticommute with any vector, because the imaginary unity e12 anticommute with e1 as
well as with e2:
zc=−cz z imaginary
The complex plane
In the complex plane, the complex numbers are represented taking the real
component as the abscissa and the imaginary component as the ordinate. The vectorial
plane differs from the complex plane in the fact that the vectorial plane is a plane of
absolute directions whereas the complex plane is a plane of relative directions with
respect to the real axis, to which we may assign any direction. As explained in more detail
in the following chapter, the unitary complex numbers are rotation operators applied to
vectors. The following equality shows the ambivalence of the Cartesian coordinates in the
Euclidean plane:
e1 ( x + y e12 ) = x e1 + y e 2
Due to a careless use, often the complex numbers have been improperly thought as
vectors on the plane, furnishing the confusion between the complex and vector planes to
our pupils. It will be argued that this has been very fruitful, but this argument cannot
satisfy geometers, who search the fundamentals of the geometry. On the other hand, some
physical magnitudes of a clearly vectorial kind have been taken improperly as complex
numbers, specially in quantum mechanics. Because of this, I’m astonished when seeing
how the inner and outer products transform in a special commutative and anti-
commutative products of complex numbers. The relation between vectors and complex
numbers is stated in the following way: If u is a fixed unitary vector, then every vector a
is mapped to a unique complex z fulfilling:
a=uz with u2 = 1
Also other vector b is mapped to a complex number t:
b=ut
The outer and inner products of the vectors a and b can be written now using the
complex numbers z and t:
1 1 1
a∧b= ( a b − b a ) = ( u z u t − u t u z ) = u2 ( z* t − t* z ) =
2 2 2
1
= ( z* t − t* z ) = ( zR tI − zI tR ) e12
2
1 1 1
a·b= ( a b + b a ) = ( u z u t + u z u t ) = ( z* t + t* z ) = zR tR + zI tI
2 2 2
where zR, tR, zI, tI are the real and imaginary components of z and t. These products have
been called improperly scalar and exterior products of complexes. So, I repeat again that
complex quantities must be distinguished from vectorial quantities, and relative directions
(complex numbers) from absolute directions (vectors). A guide for doing this is the
reversion, under which the vectors are reversed while the complex number are not1.
Complex analytic functions
The complex numbers are a commutative algebra where we can study functions as
for the real numbers. A function f(x) is said to be analytical if its complex derivative
exists:
lim f (z ) − f (z 0 )
f(z) analytical at z0 ⇔ ∃
z → z0 z − z0
This means that the derivative measured in any direction must give the same result. If f(x)
= a + b e12 and z = x + y e12 , the derivatives following the abscissa and ordinate directions
must be equal:
∂a ∂b ∂a ∂b
f' (z ) = + e12 = − e12 +
∂x ∂x ∂y ∂y
whence the Cauchy-Riemann conditions are obtained:
∂a ∂b ∂a ∂b
= and =−
∂x ∂y ∂y ∂x
Due to its linearity, now the derivative in any direction are also equal. A consequence of
these conditions is the fact that the sum of both second derivatives (the Laplacian)
vanishes, that is, both components are harmonic functions:
∂ 2a ∂ 2a ∂ 2b ∂ 2b
+ = + =0
∂x 2 ∂y 2 ∂x 2 ∂y 2
1
A physical example is the alternating current. The voltage V and intensity I in an electric circuit
are continuously rotating vectors. The energy E dissipated by the circuit is the inner product of
both vectors, E = V · I. The impedance Z of the circuit is of course a complex number (it is
invariant under a reversion). The intensity vector can be calculated as the geometric product of the
voltage vector multiplied by the inverse of the impedance I = V Z−1 . If we take as reference a
continuously rotating direction, then V and I are replaced by pseudo complex numbers, but
properly they are vectors.
The values of a harmonic function (therefore the value of f(z)) within a region are
determined by those values at the boundary of this region. We will return to this matter
later. The typical example of analytic function is the complex exponential:
exp( x + y e12 ) = exp( x )(cos y + e12 sin y )
which is analytic in all the plane. The logarithm function is defined as the inverse function
of the exponential. Since z = zexp(e12 ϕ) where ϕ is the argument of the complex, the
principal branch of the logarithm is defined as:
log z = log z + e12ϕ 0 ≤ ϕ < 2π
Also ϕ + 2πk (k integer) are valid arguments for z yielding another branches of the
logarithm2. In Cartesian coordinates:
x
log( x + y e12 ) = log x 2 + y 2 + e12 arccos 0 ≤ ϕ <π
x2 + y2
 x 
log( x + y e12 ) = log x 2 + y 2 + e12  arccos +π  π ≤ ϕ < 2π
 x2 + y2 
 
At the positive real half axis, this logarithm is not analytic because it is not continuous.
Now let us see the Cauchy’s theorem: if a function is analytic in a simply
connected domain on the complex plane, then its integral following a closed way C
within this domain is zero. If the analytic function is f (z ) = a + b e12 then the integral is:
∫ f (z ) dz = ∫ (a + b e ) (dx + dy e ) = ∫ (a dx − b dy ) + e ∫ (a dy + b dx )
C C
12 12
C
12
C
Since C is a closed way, we may apply the Green theorem to write:
 ∂b ∂a   ∂a ∂b 
= − ∫∫  +  dx dy + e12 ∫∫  −  dx dy = 0
D 
∂x ∂y  D 
∂x ∂y 
where D is the region bounded by the closed way C. Since f(z) fulfils the analyticity
conditions everywhere within D, the integral vanishes.
From here the following theorem is deduced: if f(z) is an analytic function in a
simply connected domain D and z1 and z2 are two points of D, then the definite integral
between these points has a unique value independently of the integration trajectory,
which is equal to the difference of the values of the primitive F(z) at both points:
dF (z )
f (z ) dz = F (z 2 ) − F (z1 ) f (z ) =
z2
∫ z1
if
dz
2 n
The logarithm is said to be multi-valued. This is also the case of the roots z.
If f(z) is an analytic function (with a unique value) inside the region D bounded by
the closed path C, then the Cauchy integral formula is fulfilled for a counterclockwise
path orientation:
1 f (z )
∫z−z dz = f (z 0 )
2π e12 C 0
Obviously, the integral does not vanish because the integrand is not analytic at z0 .
However, it is analytic at the other points of the region D, so that the integration path from
its beginning to its end passes always through an analyticity region, and by the former
theorem the definite integral must have a constant value, independently of the fact that
both extremes coincide. Now we integrate following a circular path z = z 0 + r exp(e12ϕ ) ,
where the radius r is a real constant and the angle ϕ is a real variable. The evaluation of
the integral gives f(z0):
2π
1
2π ∫ f (z
0
0 + r exp(e12ϕ )) dϕ = f (z 0 )
because we can take any radius and also the limit r → 0. The consequence of this theorem
is immediate: the values of f(z) at a closed path C determine its value at any z0 inside the
region bounded by C. This is a characteristic property of the harmonic functions, already
commented above.
Let us rewrite the Cauchy integral formula in a more suitable form:
1 f (t )
2π e12 ∫ t − z dt = f (z )
C
The first and successive derivative with respect to z are:
1 f (t ) n! f (t )
∫ (t − z ) 2
dt = f' (z ) ∫ (t − z ) n +1
dt = f n
(z )
2π e 12 C
2π e12 C
For z = 0 we have:
1 f (t ) n! f (t )
∫ dt = f (0) ∫t n +1
dt = f n
(0)
2π e12 C
t 2π e12 C
Now we see that these integrals always exist if f(z) is analytic, that is, all the derivatives
exist at the points where the function is analytic. In other words, the existence of the first
derivative (analyticity) implies the existence of those with higher order.
The Cauchy integral formula may be converted into a power series of z:
f (t ) f (t ) dt f (t ) ∞  z 
k
1 1 1
f (z ) = ∫C t − z dt = 2π e12 ∫C t (1 − z / t ) = 2 π e12 ∫ ∑   dt
2 π e12 C
t k =0  t 
Rewriting this expression we find the Taylor series:

1 ∞
f (t ) ∞
f k (0) k
f (z ) = ∑ zk ∫ dt = ∑ z
2 π e12 k =0 C t k +1 k =0 k!
The only assumption made in the deduction is the analyticity of f(z). So this series is
convergent within the largest circle centred at the origin where the function is analytic
(that is, the convergence circle touches the closest singular point). The Taylor series is
unique for any analytic function. On the other hand, every analytic function has a Taylor
series.
Instead of the origin we can take a series centred at another point z0. In this case,
following the same way as above, one arrives to the MacLaurin series:
∞
f k
(z 0 )
f (z ) = ∑ ( z − z 0 )k
k =0 k!
For instance, the Taylor series (z0 = 0) of the exponential, taking into account the fact that
all the derivatives are equal to the exponential and exp(0) = 1, is:
z2 z3
exp(z ) = 1 + z + + + ...
2 ! 3!
The exponential has not any singular point. Then the radius of convergence is infinite.
In order to find a convergent series for a function which is analytic in an annulus
although not at its centre (for r1 <z − z0< r2 as shown in figure 3.5), we must add
powers with negative exponents, obtaining the Lauren series:
∞
f (z ) = ∑ a (z − z ) Figure 3.5
k
k 0
k = −∞
The Lauren series is unique, and coincident

with the McLaurin series if the function has
not any singularity at the central region. The
coefficients are obtained in the same way as
above:
1 f (t ) dt
ak =
2 π e12 ∫
C t k +1
where the path C encloses the central circle. When f(x) is not analytic at some point of this
central circle (e.g. z0), the powers of negative exponent appear in the Laurent series. The
coefficient a−1 is called the residue of f. The Lauren series is the addition of a series of
powers with negative exponent and the McLaurin series:
∞ −1 ∞
f (z ) = ∑ a (z − z ) ∑ a (z − z ) + ∑ a (z − z )
k k k
k 0 = k 0 k 0
k = −∞ k = −∞ k =0
and is convergent only when both series are convergent, so that the annulus goes from the
radius of convergence of the first series (r1) to the radius of convergence of the McLaurin
series (r2).
Let us review singular points, the points where an analytic function is not defined.
An isolated singular point may be a removable singularity, an essential singularity or a
pole. A point z0 is a removable singularity if the limit of the function at this point exists
and, therefore, we may remove the singularity taking the limit as the value of the function
at z0. A point z0 is a pole of a function f(z) if it is a zero of the function 1/f(z). Finally, z0 is
an essential singularity if both limits of f(z) and 1/f(z) at z0 do not exist.
The Lauren series centred at a removable singularity has not powers with negative
exponent. The series centred at a pole has a finite number of powers with negative
exponent, and that centred at an essential singularity has an infinite number of powers
with negative exponent. Let us see some examples. The function sin z / z has a removable
singularity at z = 0:
sin z z2 z4 z6
=1− + − + ...
z 3! 5 ! 7 !
So the series has only positive exponents.

The function exp(1/z) has an essential singularity at z = 0 . From the series of
exp(z), we obtain a Lauren series with an infinite number of powers with negative
exponent by changing z for 1/z. Also the radius of convergence is infinite:
1 1 1 1
exp  = 1 + + 2
+ + ...
z z 2! z 3! z 3
Finally, the function 1/ z2 (z −1) has a pole at z = 0 . Its Lauren series:
= − 2 (1 + z + z 2 + z 3 + z 4 ...) = − 2 − − 1 − z − z 2 ...
1 1 1 1
z (z − 1)
2
z z z
is convergent for 0<z<1 since the function has another pole at z = 1.

To see the importance of the residue, let us calculate the integral of a function
through an annular way from its Lauren series:
∫ f (z ) dz = ∑ a ∫ (z − z ) k
k 0 dz
C k = −∞ C
For k≥0 the integral is zero because (z − z0)k is analytic in the whole domain enclosed by
C. For k<−2 the integral is also zero because the path is inside a region where the powers
are analytic:
lim  (z − z 0 )k −1 
z2
∫ (z − z )
k
0 dz =   =0
C
z1 → z 2  k − 1 
z1
However k = −1 is a special case. Taking the circular path z = r exp(e12ϕ ) ( r1 < r < r2 )
we have:
2π
dz
∫ z − z 0 = e12 ∫0 dϕ = 2π e12
C
So that the residue theorem is obtained:
∫ f (z ) dz = 2π e
C
12 a −1
where a−1 is the coefficient of the Lauren series centred at the pole. If the path C encloses
some poles, then the integral is proportional to the sum of the residues:
∫ f (z ) dz = 2π e ∑ residues
C
12
Let us see the case of the last example. If C is a path enclosing z = 0 and z = 1 then the
residue for the first pole is 1 and that for the second pole –1 (for a counterclockwise path)
so that the integral vanishes:
dz  1 1 1  dz dz
∫ z (z − 1) = ∫  z − 1 − z − z
C
2
C
2 

dz = ∫
C
−∫ =0
z −1 C z
The fundamental theorem of algebra
Firstly let us prove the Liouville’s theorem: if f(x) is analytic and bounded in the
whole complex plane then it is a constant. If f(x) is bounded we have:
f(x)< M
The derivative of f(x) is always given by:
1 f (t )
f' (z ) = ∫ (t − z ) 2
dt
2 π e 12 C
Following the circular path t − z = r exp(e12ϕ ) we have:
2π
1
f' (z ) = ∫ f (r exp(e ϕ )) exp(− e ϕ ) dϕ
12 12
2π r 0
Using the inequality ∫ f ( z ) dz ≤ ∫ f ( z ) dz , we find:
2π 2π
1 1 M
f' (z ) ≤ ∫ f (r exp(e12ϕ )) dϕ ≤ ∫ M dϕ = 2π r
2π r 0
2π r 0
Since the function is analytic in the entire plane, we may take the radius r as large as we
wish. In consequence, the derivative must be null and the function constant, which is the
proof of the theorem.
A main consequence of the Liouville’s theorem is the fundamental theorem of
algebra: any polynomial of degree n has always n zeros (not necessarily different):
p(z) = a0 + a1 z + a2 z2 + ... + an zn = 0 ⇒ ∃ zi i ∈{1, .., n} p(zi) = 0
where a0, a1, a2, ... an are complex coefficients. For real coefficients, the zeros are whether
real or pairs of conjugate complex numbers. The proof is by supposing that p(z) has not
any zero. In this case f(z) = 1/p(z) is analytic and bounded (because p(z)→ 0 for z→ ∞)
in the whole plane. From the Liouville’s theorem f(z) and p(z) should be constant
becoming in contradiction with the fact that p(z) is a polynomial. In conclusion p(z) has at
least one zero.
According to the division algorithm, the division of the polynomial p(z) by z − b
decreases the degree of the quotient q(z) by a unity, and yields a complex number r as
remainder:
p(z) = (z − b) q(z) + r
The substitution of z by b gives:
p(b) = r
That is, the remainder of the division of a polynomial by z − b is equal to its numerical
value for z = b . On the other hand, if b is a zero zi, the remainder vanishes and we have an
exact division:
p(z) = (z − zi) q(z)
Again q(z) has at least one zero. In each division, we find a new zero and a new factor, so
that the polynomial completely factorises with as many zeros and factors as the degree of
the polynomial, which ends the proof:
n
p( z ) = a n ∏ ( z − z i ) n = degree of p(z)
i =1
For example, let us calculate the zeros of the polynomial z3 − 5 z2 + 8 z − 6. By the

Ruffini method we find the zero z = 3:
 1 −5 8 −6

3  3 −6 6
_______________________
1 −2 2  0
The zeros of the quotient polynomial z2 −2z −2 are obtained through the formula
of the equation of second degree:
2 ± 4 − 4 ⋅1⋅ 2
z= =1± − 1 = 1 ± e12
2 ⋅1
yielding a pair of conjugate zeros. Then the factorisation of the polynomial is:
z3 − 5 z2 + 8 z − 6 = ( z − 3 ) ( z − [1 + e12 ] ) ( z − [1 − e12 ] )
Exercises
3.1 Multiply the complex numbers z = 1 + 3 e12 and t = −2 + 2 e12. Draw the geometric
figure of their product and check the result found.
3.2 Prove that the modulus of a complex number is the square root of the product of this
number by its conjugate.
3.3 Solve the equation x4 −1 = 0. Being of fourth degree, you must obtain four complex
solutions.
3.4 For which natural values of n is fulfilled the following equation?
( 1 + e12 )n + ( 1 − e12 )n = 0
3.5 Find the cubic roots of −3 + 3 e12 .
3.6 Solve the equation:
z2 + ( −3 + 2 e12 ) z + 5 − e12 = 0
3.7 Find the analytical extension of the real functions sin x and cos x.
3.8 Find the Taylor series of log(1 + exp(− z )) .

1
3.9 Calculate the Lauren series of 2 in the annulus 1<z − 2 < 4.
z + 2z − 8
∞
1
3.10 Find the radius of convergence of the series ∑ n and its analytic function.
n =1 4 ( z + 1)
n
sin z
3.11Calculate the Lauren series of 2 and the annulus of convergence.
z
3.12 Prove that if f(z) is analytic and does not vanish then it is a conformal mapping.
4. TRANSFORMATIONS OF VECTORS
The transformations of vectors are mappings from the vector plane to itself. Those
transformations preserving the modulus of vectors, such as rotations and reflections, are
called isometries and those which preserve angles between vectors are said to be
conformal. Besides rotations and reflections, the inversions and dilatations are also
conformal transformations.
Rotations Figure 4.1
A rotation through an angle α is the

geometric operation consisting in turning a
vector until it forms an angle α with the
previous orientation. The positive direction of
angles is counterclockwise (figure 4.1). Under
rotations the modulus of any vector is
preserved. According to the definition of
geometric product, the multiplication of a
vector v by a unitary complex number with
argument α produces a vector v' rotated
through an angle α with regard to v.
v' = v 1α = v ( cos α + e12 sin α )
This algebraic expression for rotations, when applied to a real or complex number
instead of vector, modifies its value. However, real numbers are invariant under rotations
and the parallelograms can be turned without changing the complex number which they
represent. Therefore this expression for rotations, although being useful for vectors, is not
valid for complex numbers. In order to remodel it, we factorise the unitary complex
number into a product of two complex numbers with half argument. According to the
permutative property, we can permute the vector and the first complex number whenever
writing the conjugate:
α α α α
v' = v 1α = v 1α/2 1α/2 = 1−α/2 v 1α/2 = ( cos − e12 sin ) v ( cos + e12 sin )
2 2 2 2
The algebraic expression for rotations now found preserves complex numbers:
z' = 1−α/2 z 1α/2 = z
Let us calculate the rotation of the vector 4 e1 through 2π / 3 by multiplying it by the

unitary complex with this argument:
 2π 2π   1 3
v' = 4 e1  cos + e12 sin  = 4 e1  − + e12  = −2 e1 + 2 3 e 2

 3 3   2 2 
On the other hand, using the half angle π/3 we have:

28 RAMON GONZÁLEZ CALVET
 π π  π π
v' =  cos − e12 sin  4 e1  cos + e12 sin 
 3 3  3 3
1 3 1 3
=  − e12  4 e1  + e12
 
 = −2 e1 + 2 3 e 2
2 2  2 2 
With the expression of half angle, it is not necessary that the complex number has unity
modulus because:
z =z1α/2 z −1 =1−α/2 z−1
Then, the rotation through an angle α can be written as:
v' = z −1 v z
The composition of two successive rotations implies the product of both complex
numbers, whose argument is the addition of the angles of both rotations.
v'' = 1 −β / 2 v' 1β / 2 = 1 −β / 2 1 −α / 2 v 1α / 2 1β /2 = 1− ( α+β)/2 v 1 (α+β)/2
Reflections
Figure 4.2
A reflection of a vector in a direction is
the geometric transformation which preserves
the component having this direction and
changes the sign of the perpendicular
component (figure 4.2). The product of
proportional vectors is commutative and that
of orthogonal vectors is anti-commutative.
Because of this, the reflected vector v' may be
obtained as the multiplication of the vector v
by the unitary vector u of the reflection axis at
the left and right hand sides:
v' = u v u = u ( v|| + v⊥ ) u = u v|| u + u v⊥ u = v|| − v⊥ with u2 = 1
where v|| and v⊥ are the components of v proportional and perpendicular to u respectively.
Instead of the unitary vector u, any vector d having the axis direction can be
introduced in the expression for reflections whenever we write its inverse at the left side
of the vector:
d vd
v' = 2
= d −1 v d
d
Although the reflection does not change the absolute value of the angle between
vectors, it changes its sign. Under reflections, real numbers remain invariant but the
complex numbers become conjugate because a reflection generates a symmetric
parallelogram (figure 4.3) and changes the sign of the imaginary part:
z = a + b e12 a, b real
d ( a + b e12 ) d Figure 4.3

z' = 2
d
d 2 a − d 2 b e12
= 2
= a − b e12 = z *
d
For example, let us calculate the reflection

of the vector 3 e1 + 2 e2 with regard to the
direction e1 − e2. The resulting vector will
be:
1 1
v' = ( e1 − e2 ) ( 3 e1 + 2 e2 ) ( e1 − e2 ) = ( e1 − e2 ) ( 1 − 5 e12 ) = − 2 e1 − 3 e2
2 2
Inversions
The inversion of radius r is the geometric transformation which maps every vector
v on r2 v −1 , that is, on a vector having the same direction but with a modulus equal to r2 /
v:
v' = r2 v −1 r real Figure 4.4
This operation is a generalisation of

the inverse of a vector in the geometric
algebra (radius r = 1). It is called inversion
of radius r, because all the vectors with
modulus r, whose heads lie on a
circumference with this radius, remain
invariant (figure 4.4). The vectors whose
heads are placed inside the circle of radius r
transform into vectors having the head
outside and reciprocally.
The inversion transforms complex
numbers into proportional complex numbers with the same argument (figure 4.5):
v' = r2 v −1 w' = r2 w −1 z=vw

r4 z
z' = v' w' = r4 v −1 w −1 = r4 v w v −2 w −2 = 2
z
30 RAMON GONZÁLEZ CALVET
Dilatations
The dilatation is that

Figure 4.5
geometric transformation which
enlarges or shortens a vector, that is,
it increases (or decreases) k times the
modulus of any vector preserving its
orientation. The dilatation is simply
the product by a real number k
v' = k v k real
The most transformations of

vectors that will be used in this book
are combinations of these four
elementary transformations. Many
physical laws are invariant under some of these transformations. In geometry, from the
vector transformations we define the transformations of points on the plane, indispensable
for solving geometric problems.
Exercises
4.1 Calculate with geometric algebra what is the composition of a reflection with a
rotation.
4.2 Prove that the composite of two reflections in different directions is a rotation.
4.3 Consider the transformation in which every vector v multiplied by its transformed v' is
equal to a constant complex z2. Resolve it into elementary transformations.
4.4 Apply a rotation of 2π/3 to the vector –3 e1 + 2 e2 and find the resulting vector.
4.5 Find the reflection of the former vector in the direction e1 + e2 .

SECOND PART: THE GEOMETRY OF THE EUCLIDEAN PLANE
A complete description of the Euclidean plane needs not only directions (given by
vectors) but also positions. Points represent locations on the plane. Then, the plane R2 is
the set of all the points we may draw on a fictitious sheet of infinite extension.
Recall that points are noted with capital Latin letters, vectors and complex
numbers with lowercase Latin letters, and angles with Greek letters. A vector going from
the point P to the point Q will be symbolised by PQ and the line passing through both
points will be distinguished as PQ .
Translations
A translation v is the geometric operation which moves a point P in the direction

and length of the vector v to give the point Q:
+ : R2 × V2 → R2
(P, v) → Q = P + v
Then the vector v is the oriented segment going from P to Q. The set of points together
with vectors corresponding to translations is called the affine plane1 (R2 , V2, +). From this
equality a vector is defined as a subtraction of two points, usually noted as PQ :
v = Q − P = PQ
The sign of addition is suitable for translations, because the composition of translations
results in an addition of their vectors:
Q=P+v R=Q+w R=(P+v)+w=P+(v+w)
5. POINTS AND STRAIGHT LINES
Coordinate systems Figure 5.1
Given an origin of coordinates O,

every point P on the plane is described by the
position vector OP traced from the origin. The
coordinates (c1, c2) of a point are the
components of the postion vector in the chosen
vector base {e1, e2}:
OP = c1 e1 + c2 e2
Then every point is given by means of a pair of
1
The affine plane does not imply a priori an Euclidean or pseudo-Euclidean character.
coordinates:
P = (c1, c2)
For example, the position vector shown in the figure 5.1 is:
OP = 2 e1 + 3 e2
and hence the coordinates of P are (2, 3).

A coordinate system is the set {O; e1, e2}, that is, the origin of coordinates O
together with the base of vectors. The coordinates of a point depend on the coordinate
system to which they belong. If the origin or any base vector is changed, the coordinates
of a point are also modified. For example, let us calculate the coordinates of the points P
and Q in the figure 5.2. Both coordinate systems have the same vector base but the origin
is different:
S = {O; e1, e2} S' = {O'; e1, e2}
Since OP = 2 e1 + 3 e2 and O'P = 2 e2 ,

the coordinates of the point P in the
coordinate systems S and S' are
respectively:
P = (2, 3)S = (0, 2)S'
Analogously:
OQ = − 4 e1 + e2 and O'Q = − 6 e1
Q = (−4, 1)S = (−6, 0)S' Figure 5.2
The coordinate axes are the straight lines passing through the origin and having
the direction of the base vectors. All the points lying on a coordinate axis have the other
coordinate equal to zero.
If a point Q is obtained from a point P by the translation v then:
Q=P+v ⇔ OQ = OP + v
that is, we must add the components of the translation vector v to the coordinates of the
point P in order to obtain the coordinates of the point Q:
(q1, q2) = (p1, p2 ) + (v1, v2)
For example, let us apply the translation v = 3 e1 − 5 e2 to the point P = (−6, 7):
Q = P + v = (−6, 7 ) + (3, −5) = (−3, 2)
On the other hand, given two points, the translation vector having these points as
extremes is found by subtracting their coordinates. For instance, the vector from P = (2, 5)
to Q = (−3, 4) is:
PQ = Q − P = (−3, 4 ) − ( 2, 5) = (−5, −1 ) = −5 e1 − e2
Remember the general rule for operations between points and vectors:
coordinates = coordinates + components
components = coordinates − coordinates
Barycentric coordinates
Why is the plane described by complicated concepts such as translations and the
coordinate system instead of using points as fundamental elements? This question was
studied and answered by Möbius in Der Barycentrische Calculus (1827) and Grassmann
in Die Ausdehnungslehre (1844). From the figure 5.3 it follows that:
OR = c1 e1 + c2 e2 = c1 OP + c2 OQ
When writing all vectors as difference of

points and isolating the generic point R,
we obtain:
R = ( 1 − c1 − c2 ) O + c1 P + c2 Q
That is, a coordinate system is a set of Figure 5.3

three non-aligned points {O, P, Q} so that
every point on the plane is a linear
combination of these three points with
coefficients whose addition is equal to the unity. The coefficients (1 − c1 − c2 , c1, c2) are
usually called barycentric coordinates, although they only differ from the usual
coordinates in a third dependent coordinate.
Distance between two points and area
The distance between two points is the modulus of the vector going from one
point to the other:
d(P, Q) = PQ = Q − P
The distance has the following properties:

1) The distance from a point to itself is zero: d(P, P) = PP = 0 and the reverse
assertion: if the distance between two points is null then both points are coincident:
d(P, Q) = 0 ⇒ PQ= 0 ⇒ Q−P=0 ⇒ P=Q
2) The distance has the symmetrical property:

d(P, Q) = PQ = QP = d(Q, P)
3) The distance fulfils the triangular

inequality: the addition of the lengths of any
two sides of a triangle is always higher than
or equal to the length of the third side (figure
5.4).
d(P, Q) + d(Q, R) ≥ d(P, R)
that is:
Figure 5.4
PQ +QR ≥ PR = PQ +QR
The proof of the triangular inequality is based on the fact that the product of the
modulus of two vectors is always higher than or equal to the inner product:
PQ QR ≥ PQ · QR =PQQR cos α
Adding the square of both vectors to the left and right hand sides, one has:
PQ2 + QR2 + 2 PQQR ≥ PQ2 + QR2 + 2 PQ · QR
(PQ + QR)2 ≥ (PQ + QR )2
PQ + QR ≥ PQ + QR
For any kind of coordinate system the distance is calculated through the inner
product:
d2(P, Q) = PQ2 = PQ · PQ
The oriented area of a parallelogram is the outer product of both non parallel sides.
If P, Q and R are three consecutive vertices of the parallelogram, the area is:
A = PQ ∧ QR
Then, the area of the triangle with these vertices is the half of the parallelogram area.
1
A= PQ ∧ QR
2
If the coordinates are given in the system formed by three non-linear points {O, X,
Y}:
P = ( 1 − xP − yP ) O + xP X + yP Y
Q = ( 1 − xQ − yQ ) O + xQ X + yQ Y
R = ( 1 - xR - yR ) O + xR X + yR Y
it is easy to prove that:
1 − xP − yP xP yP
PQ ∧ QR = 1 − x Q − y Q xQ y Q OX ∧ OY
1 − xR − yR xR yR
Then the absolute value of the area of the triangle PQR is equal to the product of the
absolute value of the determinant of the coordinates and the area of the triangle formed by
the three base points:
OX ∧ OY
A =  1 − x Q − y Q xQ yQ 
2
Condition of alignment of three points
Three points P, Q, R are said to be aligned, that is, they lie on a line, if the vectors
PQ and PR are proportional, and therefore commute.
P, Q, R aligned ⇔ PR = k PQ ⇔ PQ ∧ PR = 0 ⇔
PQ PR − PR PQ = 0 ⇔ PQ PR = PR PQ ⇔ PR = PQ −1 PR PQ
The first equation yields a proportionality between components of vectors:
yQ − y P yR − yP
=
xQ − x P xR − xP
The second equation means that the area of the triangle PQR must be zero:
1 − xQ − yQ xQ yQ = 0
that is, the barycentric coordinates of the three points are linearly dependent. According to
the previous chapter, the last equality means that the vector PR remains unchanged after a
reflection in the direction PQ. Obviously, this is only feasible when PR is proportional to
PQ, that is, when the three points are aligned.
Cartesian coordinates
The coordinates (x, y) of any system having an orthonormal base of vectors are
called Cartesian coordinates from Descartes2. The horizontal coordinate x is called
abscissa and the vertical coordinate y ordinate. In this kind of bases, the square of a vector
is equal to the sum of the squares of both components, leading to the following formula
for the distance between two points:
P = (xP, yP) Q = (xQ, yQ)
d (P, Q ) = (x
Q − xP ) + ( yQ − y P )
2 2
For example, the triangle with vertices P = (2, 5), Q = (3, 2) and R = (−4, 1) have
the following sides:
PQ = Q − P = ( 3, 2 ) − ( 2, 5 ) = ( 1, −3) = e1 − 3 e2 PQ = 10
QR = R − Q = (−4, 1 ) − ( 3, 2 ) = (−7, −1 ) = −7 e1 − e2 QR = 50
RP = P − R = ( 2, 5 ) − (−4, 1 ) = ( 6, 4 ) = 6 e1 + 4 e2 PR = 40
The area of the triangle is the half of the outer product of any pair of sides:
1 1
A= PQ ∧ QR = ( e1 − 3 e2 ) ∧ (−7 e1 − e2 ) = −11 e12 A= 11
2 2
Vectorial and parametric equations of a line
The condition of alignment of three points is the starting point to deduce the
equations of a straight line. A line is determined whether by two points or by a point and a
direction defined by its direction vector, which is not unique because any other
proportional vector will also be a direction vector for this line. Known the direction vector
v and a point P on the line, the vectorial equation gives the generic point R on the line:
r: {v, P} P = (xp, yP) R = (x, y)
PR = k v
Separating coordinates, the parametric equation of the line is obtained:
(x, y) = (xp, yp) + k (v1, v2)
Here the position of a point on the line depends on the real parameter k, usually identified
2
If the x-axis is horizontal and the y-axis vertical, an orthonormal base of vectors is called
canonical.
with the time in physics.

Knowing two points P, Q on the line, a direction vector v can be obtained by
subtraction of both points:
r: {P, Q} v=Q−P
In this case, the vectorial and parametric equations become:
PR = k PQ ⇔ R=P+k(Q−P)=P(1−k)+kQ
The parameter k indicates in which

proportion the point R is closer to P than to
Q. For example, the midpoint of the
segment PQ can be calculated with k = 1/2.
The figure 5.5 shows where the point R is
located on the line PQ as a function of the
parameter k. In this way, all the points of
the line are mapped to the real numbers
incorporating the order relation and the
topologic properties of this set to the line3.
Figure 5.5
Algebraic equation and distance
from a point to a line
If the point R belongs to the line, the vector PR and the direction vector v are
proportional and commute:
PR v = v PR ⇔ PR v − v PR = 0 ⇔
PR ∧ v = 0 ⇔ PR = v −1 PR v
The last equation is the algebraic equation and

shows that the vector PR remains invariant
under a reflection in the direction of the line.
That is, the point R only belongs to the line
when it coincides with the point reflected in
this line. Separating components, the two-point Figure 5.6
equation of the line arises:
x − xP y − yP x − xP y − yP
= ⇔ =
v1 v2 xQ − xP yQ − yP
3
In the book Geometria Axiomàtica (Institut d'Estudis Catalans, Barcelona, 1993), Agustí
Reventós defines a bijection between the points of a line and the real numbers preserving the order
relation as a simplifying axiom of the foundations of geometry. I use implicitly this axiom in this
book.
where v = ( v1, v2 ).
If the point R does not belong to the line, the reflected point R' differs from R. This
allows to calculate the distance from R to the line as the half of the distance between R
and R' (figure 5.6). The vector going from R to R' is equal to PR' − PR .
1 1 1 −1
d( R, r ) = R' − R  = PR' − PR=  v PR v − PR 
2 2 2
We will obtain an easier expression for the distance by extracting the direction vector as
common factor:
1 −1 1 −1 PR ∧ v
d( R, r ) = v ( PR v − v PR ) = v  PR v − v PR=
2 2 v
In this formula, the distance from R to the

line r is the height of the parallelogram
formed by the direction vector and (figure
5.7).
Similar line equations can also be
deduced using the normal (perpendicular)
vector n. Like for the direction vector, the
normal vector is not unique because any
other proportional vector may also be a Figure 5.7
normal vector. The normal vector
anticommutes with the direction vector of
the line, and therefore with the vector PR,
P being a given point and R a general point on the line:
n PR = −PR n ⇔ n PR + PR n = 0 ⇔ n · PR = 0 ⇔ PR = − n −1 PR n
The last equation is also called the algebraic equation and shows that the vector
PR is reversed under a reflection in the perpendicular of the line. Taking components, the
general equation of the line is obtained:
n1 (x − xP) + n2 (y − yP) = 0 ⇔ n1 x + n2 y + c = 0
where n = (n1, n2) and c is a real constant.

When the point R does not lie on the
line, the point R' reflected of R in the
perpendicular does not belong to the line. Let
PR'' be the reversed of PR', or equivalently
R'' be the symmetric point of R with regard to
P (figure 5.8). Then, the distance from R to
the line r is the half of the distance from R to
R'':
1 Figure 5.8
d( R, r ) = PR'' − PR 
2
1 1 PR · n
= − n −1 PR n − PR = − n −1 ( PR n + n PR ) =
2 2 n
This expression is the modulus of the projection of the vector PR upon the perpendicular
direction. Both formulas of the distance are equivalent because the normal and direction
vectors anticommute:
nv=−vn ⇔ nv+vn=0 ⇔ n·v=0
Written with components:
n1 v1 + n2 v2 = 0
From each vector the other one is easily obtained by exchanging components and altering
a sign:
n = ( n1 , n2 ) = ( v2 , −v1 )
As an application, let us calculate the equation of the line r passing through the
points (2, 3) and (7, 6). A direction vector is the segment having as extremes both points:
v = (7, 6) − (2, 3) = (5, 3)
and an equation for this line is:
x −2 y −3
=
5 3
From the direction vector we calculate the perpendicular vector n and the general equation
of the line:
n = (3, −5) 3 (x − 2) − 5 (y − 3) = 0 ⇔ 3x−5y+9=0
Now, let us calculate the distance from the point R = (−2, 1) to the line r. We choose the
point (2, 3) as the point P on the line:
PR = R − P = (−2, 1) − (2, 3) = (−4, −2)
PR ∧ v PR · n − 4 ⋅ 3 + ( −2) ⋅ ( −5) 2
d( (−2, 1), r ) = = = =
v n 34 17
For another point R = (−3, 0 ), we will find a null distance indicating that this point lies on
r.
Slope and intercept equations of a line
If the ordinate is written as a function of the abscissa, the equation so obtained is

the slope-intercept equation:
y=mx+b
Note that this expression cannot describe the vertical lines whose equation is x = constant.
The coefficient m is called the slope because it is a measure of the inclination of the line.
For any two points P and Q on the line we have:
yP = m xP + b
yQ = m xQ + b
Subtracting both equations:
yQ − yP = m (xQ − xP)
we see that the slope is the quotient of

ordinate increment divided by the abscissa
increment (figure 5.9):
Figure 5.9
yQ − y P
m=
xQ − x P
This quotient is also the trigonometric tangent of the oriented angle between the line and
the positive abscissa semiaxis. For angles larger than π/2 the slope becomes negative.
b is the ordinate intercepted at the origin, the y-intercept:
x=0 ⇒ y=b
If we know the slope and a point on a line, the point-slope equation may be used:
y − yP = m (x − xP)
Also, one may write for the intercept equation of a line:
x y
+ =1
a b
where a and b are the x-intercept and y-intercept of the line with each coordinate axis.
Polar equation of a line
In this equation, the distance from a fixed point F to the generic point P on the line
is a function of the angle α with regard to the perpendicular direction (figure 5.10). If d is
the distance from F to the line, then:
d
FP =
cos α Figure 5.10
This equation allows to relate easily the

straight line with the circle and other conic
sections.
Intersection of two lines and pencil of

lines
The calculation of the intersection

of two lines is a very usual problem. Let us
suppose that the first line is given by the point P and the direction vector u and the second
one by the point Q and the direction vector v. Denoting by R the intersection point, which
belongs to both lines, we have:
R=P+ku=Q+lv ⇔ k u − l v = Q − P = PQ
In order to find the coefficients k and l, the vector PQ must be resolved into a linear
combination of u and v, what results in:
k = PQ ∧ v (u ∧ v) −1 l = − u ∧ PQ (u ∧ v) −1
The intersection point is also obtained by directly solving the system of the
general equations of both lines:
 n1 x + n 2 y + c = 0

n1' x + n 2' y + c' = 0
The pencil of lines passing through this point is the set of lines whose equations
are linear combinations of the equations of both given lines:
(1 − p) [ n1 x + n2 y + c ] + p [ n'1 x + n'2 y + c' ] = 0
where p is a real parameter and

− ∞ ≤ p ≤ ∞ . Each line of the pencil
determines a unique value of p,
independently of the fact that the
general equation for a line is not
unique
If the equation system has a
unique point as solution, all the lines
of the pencil are concurrent at this
point. But in the case of an
incompatible system, all the lines of Figure 5.11
the pencil are parallel, that is, they
intersect in a point at infinity, which gives generality to the concept of pencil of lines.
If the common point R is known or given, then the equation of the pencil is written
as:
[ ( 1 − p ) n1 + p n'1 ] ( x − xR ) + [ ( 1 − p ) n2 + p n'2 ] ( y − yR ) = 0
For 0<p<1, the normal and direction vectors of the p-line are respectively comprised
between the normal and direction vectors of both given lines4. In the other case the vectors
are out of this region (figure 5.11)
For example, calculate the equation of the line passing through the point (3, 4) and
the intersection of the lines 3 x + 2 y + 4 = 0 and 2 x − y + 3 = 0. The equation of the
pencil of lines is:
(1−p)(3x+2y+4)+p(2x−y+3)=0
The line of this pencil to which the point (3, 4) belongs must fulfil:
(1−p)(3⋅3+2⋅4+4)+p(2⋅3−4+3)=0
yielding p = 21/16 and the line:
27 x − 31 y + 43 = 0
Which is the meaning of the coefficient of linear combination p? Let us write the
pencil of lines P determined by the lines R and S as:
P=(1−p)R+pS
which implies the same relation for the normal vectors:
nP = ( 1 − p) nR + p nS
Taking outer products we obtain:
nP ∧ nR p nP ∧ nR
= ⇒ p=
nP ∧ nS p − 1 nP ∧ ( nR − nS )
These equalities are also valid for direction vectors whenever they have the same modulus
than the normal vectors. When the modulus of the normal vectors of the lines R and S are
equal we can simplify:
p sin (RP )
=
1 − p sin (PS )
In this case, the value p = 1/2 corresponds to the bisector line of R and S. We see from this
expression and the foregoing ones that each value of p corresponds to a unique line.
4
This statement requires that the angle from the direction vector to the normal vector be a positive
right angle.
Dual coordinates
The duality principle states that any theorem relating incidence of lines and points
implies a dual counterpart where points and lines have been exchanged. The phrase «two
points determine a unique line passing through these points» has the dual statement «two
lines determine a unique point, intersection of these lines». All the geometric facts have an
algebraic counterpart, and the duality is not an exception. With the barycentric coordinates
every point R on the plane can be written as linear combination of three non aligned points
O, P and Q:
R = ( 1 − p − q ) O + p P + q Q = (p, q)
where p and q are the coordinates and O the origin.

These three points determine three non parallel lines A, B, and C in the following
way:
A = PQ B = QO C = OP
Then the direction vectors vA = PQ, vB = OP, vC = QO are related by:
v A + v B + vC = 0
Note that vB and −vC are the base of the vectorial plane.
Any line D on the plane can be written as linear combination of these lines:
D=(1−b−c)A+bB+cC
This means that the general equation (with point coordinates) of D is a linear combination
of the general equations of the lines A, B and C. I call b and c the dual coordinates of the
line D. In order to distinguish them from point coordinates, I shall write D = [b, c]. The
choice of the equation for each line must be unambiguous and therefore the normal vector
in the implicit equations for A, B and C will be obtained by turning the direction vector
over π/2 counterclockwise.
Let us see some special cases. If the dual coordinate b is zero we have:
D=(1−c)A+cC
which is the equation of the pencil of lines passing through the intersection of the lines A
and C, that is, the point P. Then P = [0, c] (for every c). Analogously, c = 0 determines the
pencil of lines passing through the intersection of the lines A and B, which is the point Q,
and then Q = [b, 0] (for every b). The origin of coordinates is the intersection of the lines
B and C and then O = [b, 1 − b] (for every b). Compare the dual coordinates of these
points:
O = [b, 1−b] P = [0, c] Q = [b, 0] ∀ b, c
with the point coordinates of the lines A, B and C:

A = (p, 1− p) B = (0, q) C = (p, 0) ∀ p, q
Let us see an example: calculate the dual Cartesian coordinates of the line 2x + 3y
+ 4 = 0. The points of the Cartesian base are O = (0, 0), P = (1, 0) and Q = (0, 1). Then the
lines of the Cartesian base are A: −x − y +1= 0, B: x = 0, C: y = 0. We must solve the
identity:
2 x + 3 y + 4 ≡ a' ( − x − y + 1) + b' x + c' y ∀ x, y
x ( 2 + a' − b' ) + y ( 3 + a' − c' ) + 4 − a' ≡ 0
whose solution is:
a' = 4 b' = 6 c' = 7
Dividing by the sum of the coefficients we obtain:
2x +3y +4 4 6 7
≡ ( − x − y +1)+ x+ y
17 17 17 17
from where the dual coordinates of this line are obtained as [b, c] = [6/17, 7/17]. Let us
see their meaning. The linear combination of both coordinates axes is a line of the pencil
of lines passing through the origin:
6 7
x+ y=0 or 6x+7y=0
13 13
This line intersects the third base line −x −y + 1 = 0 at the point (7, −6), whose pencil of
lines is described by:
6 7 
a ( − x − y + 1) + ( 1 − a )  x + y = 0
 13 13 
Then 2 x + 3 y + 4 = 0 is the line of this pencil determined by a = 4/17.

On the other hand, how may we know whether three lines are concurrent and
belong to the same pencil or not? The answer is that the determinant of the dual
coordinates must be zero:
D=(1−b−c)A+bB+cC
E = ( 1 − b' − c' ) A + b' B + c' C
F = ( 1 − b'' − c'' ) A + b'' B + c'' C

1− b − c b c
D, E and F concurrent ⇔ 1 − b' − c' b' c' = 0
1 − b'' − c'' b'' c''
When it happens, we will say that the lines are linearly dependent, and we can write
anyone of them as linear combination of the others:
F=(1−k)D+kE
We can express a point with an equation for dual coordinates in the same manner
as we express a line with an equation for point coordinates. For example, the point (5, 3)
is the intersection of the lines x = 5, y = 3 whose dual coordinates we calculate now:
x − 5 ≡ a' ( 1 − x − y ) + b' x + c' y ⇒ a' = −5 b' = −4 c' = −5
From where a = 5/14, b = 4/14 and c = 5/14. The dual coordinates of x = 5 are [4/14,
5/14]. Analogously:
y − 3 ≡ a' ( 1 − x − y ) + b' x + c' y ⇒ a' = −3 b' = −3 c' = −2
From where a = b = 3/8, c = 2/8. The dual coordinates of y = 3 are [3/8, 2/8]. The linear
combinations of both lines are the pencil of the point (5, 3), which is described by the
parametric equation:
[b, c] = ( 1 − k )  4 ,
5
+k
3 2
 8 , 8 
14 14 
By removing the parameter k, the general equation is obtained:
b − 4 / 14 c − 5 / 14
= ⇔ 12 b + 10 c − 7 = 0
5 −6
That is, the dual direction vector of the point (5, 3) is [5, −6] and the dual normal vector is
[6, 5]. In the dual plane, an algebra of dual vectors can be defined. A dual direction vector
for a point may be obtained as the difference between the dual coordinates of two lines
whose intersection be the point. Then, there exist dual translations of lines:
dual + : L × W → L L ={plane lines} W = {dual vectors}
(A, w) → B = A + w
That is, we add a fixed dual vector w to the line A in order to obtain another line B.
Let us prove the following theorem: all the points whose dual direction vectors are
proportional are aligned with the centroid of the coordinate system .
The proof begins from the dual continuous equation for a point P:
b − b0 c − c 0
=
v1 v2
which can be written in a parametric form:

P: [b, c] = (1 − k ) [b0 , c0 ] + k [b0 +`v1 , c 0 + v 2 ]
This equation means that P is the intersection point of the lines with dual coordinates [b0,
c0] and [b0 +v1, c0 +v2]. Then P must be obtained by solving the system of equations of
each line for the point coordinates p, q (x, y if Cartesian):
 (1 − b0 − c 0 ) A + b0 B + c 0 C = 0

(1 − b0 − c 0 − v1 − v 2 ) A + (b0 + v1 ) B + (c 0 + v 2 ) C = 0
By subtraction of both equations, we find an equivalent system:
(1 − b0 − c 0 ) A + b0 B + c 0 C = 0

 − (v1 + v 2 ) A + v1 B + v 2 C = 0
Now, if we consider a set of points with the same (or proportional) dual direction vector,
the first equation changes but the second equation remains constant (or proportional). That
is, the first line changes but the second line remains constant. Therefore all points will be
aligned and lying on the second line:
− (v1 + v 2 ) A + v1 B + v 2 C = 0
However in which matter do two points differ whether having or not a proportional dual
direction vector? We cannot say that two points are aligned, because we need three points
at least. Let us search the third point X, rewriting the foregoing equality:
v1 v
(B − A) + 2 (C − A) = 0
v1 + v 2 v1 + v 2
Now the variation of the components of the dual vector generates the pencil of the lines B
− A and C − A. That is, the intersection of both lines is the point X searched:
B − A = 0
X: 
C − A = 0
Then, two points have a proportional dual direction vector if they are aligned with the
point X, intersection of the lines B − A and C − A. However, note that the addition of
coefficients of each line is zero instead of one, so that one dual coordinate of each line is
infinite:
B − A = [∞, c ] C − A = [b, ∞]
I call X the point at the dual infinity or simply the dual infinity point. The dual infinity
point has finite coordinates and a very well defined position. Then the points with
proportional dual vector are always aligned with the dual infinity point, and I shall say
that they are parallel points in a dual sense, of course. In order to precise which point is X
let us take in mind that the lines B − A and C − A are the medians of the triangle OPQ (the
point base). Hence the dual infinity point is the centroid of the base triangle OPQ. With
Cartesian coordinates O = (0, 0), P = (1, 0), Q = (0, 1) and X = (1/3, 1/3). This ends the
proof.
Summarising, we can say that two parallel points are aligned with the dual infinity
point, which is the dual statement of the fact that two parallel lines meet at the infinity,
that is, there is a line at the infinity, in the usual sense. This is the novelty of the projective
geometry in comparison with the Euclidean geometry. In fact, the problem arises because
the point coordinates take infinite values, but the line L at the infinity is a well defined line
with finite dual coordinates [1/3, 1/3]:
1 1 A + B + C
L= , =
 3 3 3
This means that the line L at the infinity belongs to the pencil of the lines (A + B ) /2 and C
. But both lines are parallel because their direction vectors are proportional:
v A + v B = −v C
That is, the lines (A + B ) /2 and C meet at a point located at an infinite distance from the
origin of coordinates. Since any other parallel line meets them also at the infinity, this
argument does not suffice. However we can understand that the line L also belongs to the
pencil of the lines A and (B + C ) /2, which are also parallel with another direction, that is,
they meet at another point of the infinity. Any other pencil we take has always its point of
intersection located at the infinity. Then L only has points located at the infinity and
because of this it is called the line at the infinity. Summarising we can say that the line at
the infinity is the centroid of the three base lines in spite of the incompatibility of its
equation:
L=
A+ B +C (− x − y + 1) + x + y = 0 1
=0
⇒ ⇒
3 3 3
Moreover, we may interpret the parallel lines as those lines aligned (in a dual sense) with
the line at the infinity:
E = [bE, cE] F = [bF, cF]
bF − b E b F − 1 / 3
E || F ⇔ =
cF − cE cF − 1 / 3
This is a useful equality because it allows us to know whether two lines are parallel from
their dual coordinates.
Many of the incidence theorems (and also their dual theorems) can be solved by
means of line equations or dual coordinates. A proper example is the proof of the
Desargues theorem.
The Desargues theorem

Given two triangles ABC and A'B'C', let P be the intersection of the prolongation
of the side AB with the side A'B', Q the intersection of BC with B'C', and R the intersection
of CA with C'A'. The points P, Q and R are aligned if and only if the lines AA', BB' and
CC' meet at the same point.
Proof ⇒ The hypothesis states that the lines AA', BB' and CC' intersect at the same
point O (figure 5.12):
O = a A + ( 1 − a ) A'
O = b B + ( 1 − b ) B'
O = c C + ( 1 − c ) C'
with a, b and c being real. Equating

the first and second equations we
obtain:
Figure 5 12
a A + ( 1 − a ) A' = b B + ( 1 − b ) B'
which can be rearranged as:
a A − b B = − ( 1 − a ) A' + ( 1 − b ) B'
Dividing by a − b the sum of the coefficients becomes the unity, and then the equation
represents the intersection of the lines AB and A'B', which is the point P:
a b a −1 b −1
P= A− B= A' − B'
a−b a−b a−b a−b
By equating the second and third equations for the point O, and the third and first ones,
analogous equations for Q and R are obtained:
b c b −1 c −1
Q= B− C= B' − C'
b−c b−c b−c b−c
c a c −1 a −1
R= C− A= C' − A'
c−a c−a c−a c−a
Now we must prove that P, Q and R are aligned, that is, fulfil the equation:
R=dP+(1−d)Q with d real
With the substitution of the former equations into the last one we have:
c a  a b   b c 
C− A=d  A− B + ( 1 − d )  B− C
c−a c−a a − b b−a  b − c b−c 
Arranging all the terms at the left hand side, the expression obtained must be identical to
zero because A, B and C are non aligned points:
 da a   db b (1 − d )   c c (1 − d ) 
− a − b − c − a  A +  a − b − b − c  B +  c − a + b − c  C ≡ 0
     
This implies that the three coefficients must be null simultaneously yielding a unique
value for d,
a−b
d=
a−c
fact which proves the alignment of P, Q and R, and gives the relation for the distances
between points:
QR QP −1 = ( A'O A'A −1 − B'O B'B −1 ) ( A'O A'A-1 − C'O C'C −1 ) −1
Proof ⇐ We will prove the Desargues theorem in the other direction following the
same algebraic way but applying the duality, that is, I shall only change the words. O, A,
B, C, P, Q and R will be now lines (figure 5.13).
The hypothesis states that the
points AA', BB' and CC' belong to the
same line O, that is, O belongs to the
pencil of the lines A and A', but also
to the pencil of the lines BB' and CC':
O = a A + ( 1 − a ) A'
O = b B + ( 1 − b ) B'
Figure 5.13
O = c C + ( 1 − c ) C'
with a, b and c being real

coefficients. Equating the first and
second equations we obtain:
a A + ( 1 − a ) A' = b B + ( 1 − b ) B'
which can be rearranged as:
a A − b B = − ( 1 − a ) A' + ( 1 − b ) B'
Dividing by a − b the sum of the coefficients becomes the unity, and then the equation
represents the line passing through the points AB and A'B' (belonging to both pencils),
which is the line P:
a b a −1 b −1
P= A− = A' − B'
a−b a−b a−b a−b
By equating the second and third equations for the line O, and the third and first ones,
analogous equations for Q and R are obtained:
b c b −1 c −1
Q= B− C= B' − C'
b−c b−c b−c b−c
c a c −1 a −1
R= C− A= C' − A'
c−a c−a c−a c−a
Now we must prove that the lines P, Q and R belong to the same pencil, that is, fulfil the
equation:
R=dP+(1−d)Q with d real
With the substitution of the former equations into the last one we have:
c a  a b   b c 
C− A=d  A− B + ( 1 − d )  B− C
c−a c−a  a − b b − a   b − c b − c 
Arranging all the terms at the left hand side, the expression obtained must be identical to
zero because A, B and C are independent lines (not belonging to the same pencil):
 da a   db b (1 − d )   c c (1 − d ) 
− a − b − c − a  A +  a − b − b − c  B +  c − a + b − c  C ≡ 0
     
This implies that the three coefficients must be null simultaneously yielding a unique
value for d,
a−b
d=
a−c
fact which proves that P, Q and R are lines of the same pencil.
Exercises
5.1 Let A = (2, 4), B = (4, −3) and C = (2, −5) be three consecutive vertices of a
parallelogram. Calculate the fourth vertex D and the area of the parallelogram.
5.2 Prove the Euler’s theorem: for any four points A, B, C and D the product AD BC +
BD CA + CD AB vanishes if and only if A, B and C are aligned.
5.3 Consider a coordinate system with vectors { e1, e2 } where e1= 1, e2= 1 and
the angle formed by both vectors is π/3.
a) Calculate the area of the triangle ABC being A = (2, 2), B = (4, 4), C = (4, 2).
b) Calculate the distance between A and B, B and C, C and A.
5.4 Construct a trapezoid whose sides AB, BC, CD and DA are known and
AB is parallel to CD. Study for which values of the sides does the trapezoid exist.
5.5 Given any coordinate system with points {O, P, Q} and any point R with
coordinates (p, q) in this system, show that:
Area RPQ Area ORQ Area OPR

1− p − q = p= q=
Area OPQ Area OPQ Area OPQ
5.6 Let the points A = (2, 3), B = (5, 4), C = (1, 6) be given. Calculate the distance
from C to the line AB and the angle between the lines AB and AC.
5.7 Given any barycentric coordinate system defined by the points {O, P, Q} and the
non aligned transformed points {O', P', Q'}, an affinity (or an affine
transformation) is defined as that geometric transformation which maps each point
D to D' in the following way:
D = ( 1 − x − y ) O + x P + y Q→ D' = ( 1 − x − y ) O' + x P' + y Q'
Prove that:
a) An affinity maps lines onto lines.
b) An affinity is equivalent to a linear mapping of the coordinates, that is, the
coordinates of any transformed point are linear functions of the coordinates of the
original point.
c) An affinity preserves the coordinates expressed in any other set of independent
points different of the given base:
D = ( 1 − b − c ) A + b B + c C→ D' = ( 1 − b − c ) A' + b B' + c C'

d) An affinity maps a parallelogram to another parallelogram, and hence, parallel
lines to parallel lines.
e) An affinity preserves the ratio DE DF −1 for any three aligned points D, E and
F.
5.8 If {A, B, C} is a base of lines and {A', B', C'} their transformed lines -the lines of
each set being independent-, consider the geometric transformation which maps
each line D to D' in the following way:
D = ( 1 − b − c ) A + b B + c C→ D' = ( 1 − b − c ) A' + b B' + c C'
a) Prove that every pencil of lines is mapped to another pencil of lines.

b) The dual coordinates of D' are linear functions of the dual coordinates of D.
c) This transformation preserves the coefficients which express a line as a linear
combination of any three non concurrent lines, that is, in the foregoing mapping
{A, B, C} have not to be necessarily the dual coordinate base and can be any other
set of independent lines.
d) Parallel points are mapped to parallel points.
e) For any three concurrent lines P, Q and R, the single ratio of the dual vectors
PQ PR−1 is preserved.
f) Using the formula of the cross ratio of a pencil of any four lines (ABCD) as a
function of their direction vectors5:
v A ∧ vC v B ∧ v D
( ABCD ) =
v A ∧ v D v B ∧ vC
show that it is preserved.
5.9 Calculate the dual coordinates of the lines x – y + 1 = 0 and x – y + 3 = 0 . See

that they are aligned in the dual plane with the line at the infinity, whose dual
coordinates are [1/3, 1/3], and therefore are parallel.
5.10 Calculate the dual equations of the points (2, 1) and (–3, –1) and their direction
vectors. See that they are parallel points. Hence prove that they are aligned with
the dual infinity point, the centroid of the coordinate system (1/3, 1/3).
5
This formula is deduced in the chapter devoted to the cross ratio.
6. ANGLES AND ELEMENTAL TRIGONOMETRY
Here, the basic identities of the elemental trigonometry are deduced in close
connection with basic geometric facts, a very useful point of view for our pupils.
Sum of the angles of a polygon
Firstly let us see the special case

of a triangle. For any triangle ABC the
following identity holds:
CA AB −1 AB BC −1 BC CA −1 = 1
Let α, β and γ be the exterior angles

between the sides CA and AB, AB and
BC, BC and CA respectively (figure 6.1). Figure 6.1
Applying the definition of geometric
quotient, the modulus of all the sides are
simplified and only the exponentials of the arguments remain:
exp(α e12 ) exp(β e12 ) exp(γ e12 ) = exp[ (α + β + γ ) e12 ] = 1
The three angles have the same orientation, which we suppose positive, and are
lesser than π. Hence, since the exponential is equal to the unity, the addition of the three
angles must be equal to 2π:
α + β + γ = 2π
The interior angles, those formed by AB and AC, BC and CA, CA and CB are
supplementary of α, β, γ (figure 6.1). Therefore the sum of the angles of a triangle is equal
to π:
(π − α ) + (π − β ) + (π − γ ) = π
This result is generalised to any polygon from the following identity:
AB BC −1 BC ... YZ −1 YZ ZA−1 ZA AB −1 = 1
Let α, β, γ ... ω be the exterior angles formed by the sides ZA and AB, AB and BC,
BC and CD, ..., YZ and ZA, respectively. After the simplification of the modulus of all
vectors, we have:
exp[ (α + β + γ + ... + ω ) e12 ] = 1
Let us suppose that the orientation defined by the vertices A, B, C ... Z is

counterclockwise, although the exterior angles be not necessarily all positive1. Translating
1
That is, it is not needed that the polygon be convex.
them to a common vertex, each angle is placed close each other following the order of the
perimeter and summing one turn:
α + β + γ + ... + ω = 2π
The interior angles formed by the sides AB and ZA, BC and BA, CD and CB,... ZA and ZY
are supplementary of α, β, γ, ... ω. Therefore the sum of the angles of a polygon is:
( π − α ) + ( π − β ) + ( π − γ ) + ... + ( π − ω ) = n π − 2 π = ( n − 2 )π
n being the number of sides of the polygon. The deduction for the clockwise orientation of
the polygon is analogous with the only difference that the result is negative.
Definition of trigonometric functions and fundamental identities
Let us consider a circle with radius r (figure 6.2). The extreme of the radius is a
point on the circumference with coordinates (x, y). The arc between the positive X
semiaxis and this point (x, y) has an oriented length s, positive if counterclockwise and
otherwise negative. Also, the X-axis,
the arc of circumference and the
radius delimit a sector with an
oriented area A. An oriented angle Figure 6.2
α is defined as the quotient of the arc
length divided by the radius2:
s
α=
r
Since the area of the sector is

proportional to the arc length and the
area of the circle is 2πr, it follows
that:
α r2 2A
A= ⇔ α=
2 r2
The trigonometric functions3, sine, cosine and tangent of the angle α, are respectively
defined as the ratios:
y x y
sin α = cos α = tg α =
r r x
and the cosecant, secant and cotangent as their inverse fractions:
2
With this definition it is said that the angle is given in radians, although an angle is a quotient of
lengths and therefore a number without dimensions.
3
Also called circular functions due to obvious reasons, to be distinguished from the hyperbolic
functions.
r 1 r 1 x 1
csc α = ≡ sec α = ≡ cot α = ≡
y sin α x cos α y tg α
being r related with x and y by the Pythagorean theorem:
r2 = x2 + y2
Then the radius vector v is:
v = x e1 + y e2 = r ( cosα e1 + sinα e2 )
From these definitions the fundamental identities follow:
sin α 1
tg α ≡ sin 2α + cos 2 α ≡ 1 1 + tg 2 α ≡ ≡ sec 2 α
cos α cos 2 α
If we take the opposite angle –α instead of α, the sign of y is changed while x and r are
preserved, so we obtain the parity relations:
sin(− α ) ≡ −sin α cos(− α ) ≡ cos α tg(− α ) ≡ − tg α
The sine and tangent are odd functions while the cosine is an even function.
Look at the figure 6.2: β = π/2−α is the complementary angle of α . If α is higher
than π/2, the angle β becomes negative. On the other hand, if the angle α is negative then
β is higher than π/2. The trigonometric ratios for β give the complementary angle
identities:
x y x 1
sin β = ≡ cos α cos β = ≡ sin α tg β = ≡
r r y tg α
Angle inscribed in a circle and double angle identities
Let us draw any diameter PQ and any

radius OA (figure 6.3) in a circle with centre O. Figure 6.3
Since OP is also a radius, the triangle POA is
isosceles and the angles OPA and PAO, which
will be denoted as α, are equal. Because the
addition of the three angles is equal to π, the
angle AOP is π − 2α . The angle QOA is
supplementary of AOP, and therefore is equal to
2α, the double of the angle APQ. Let us draw
from A a segment perpendicular to the diameter
and touching it at the point N. By the definition of
sine we have:
NA NA PA
sin 2α = =
OA PA OA
The first quotient is sinα for the triangle PNA. If M is the midpoint of the segment
PA, then PA = 2 MA and the second quotient is equal to 2 cosα for the triangle
MOA:
MA
sin 2α = 2 sin α ≡ 2 sin α cos α
OA
Through an analogous way we obtain cos 2α :
ON ( PN − PO ) PA
cos 2α = = ≡ 2 cos 2 α − 1 ≡ cos 2 α − sin 2α
OA PA OA
Also this result may be obtained from the second fundamental identity. In order to obtain
the tangent of the double angle we make use of the first fundamental identity:
sin 2α 2 sin α cos α 2 tg α

tg 2α ≡ ≡ ≡
cos2α cos α − sin α 1 − tg 2 α
2 2
Finally, let us draw any other segment PB (figure 6.3). The angle BPQ will be
denoted as β. By the same arguments as above the angle BOQ is 2β. While the angle BPA
is α + β, the angle BOA is 2α +2β : an angle inscribed in a circumference is equal to the
half of the central angle (angle whose vertex is the centre of the circumference) which
intercepts the same arc of circle. Consequently, all the angles inscribed in the same circle
and intercepting the same arc are equal independently of the position of the vertex on the
circle.
Addition of vectors and sum of trigonometric functions
Let us consider two unitary vectors forming the angles α and β with the e1
direction (figure 6.4):
u = e1 cos α + e 2 sinα v = e1 cos β + e 2 sin β
u + v = e1 (cos α + cos β ) + e 2 (sin α + sin β )

Figure 6.4
The addition of both vectors, u + v , is the
diagonal of the rhombus which they form, whence
it follows that the long diagonal is the bisector of
the angle α − β between both vectors. The short
diagonal cuts the long diagonal perpendicularly
forming four right triangles. Then the modulus of
u + v is equal to the double of the cosine of the
half of this angle:
α−β
u + v = 2 cos
2
Moreover, the addition vector forms an angle (α + β) /2 with the e1 direction:
α−β  α+β α+β

u + v = 2 cos  e1 cos + e 2 sin 
2  2 2 
By identifying this expression for u + v with that obtained above, we arrive at two
identities, one for each component:
α+β α−β
cos α + cos β ≡ 2 cos cos
2 2
α+β α−β
sin α + sinβ ≡ 2 sin cos
2 2
In a similar manner, but using a subtraction of unitary vectors, the other pair of
identities are obtained:
α+β α−β
cos α − cos β ≡ −2 sin sin
2 2
α+β α−β
sin α − sin β ≡ 2 cos sin
2 2
The addition and subtraction of tangents are obtained through the common
denominator:
sin α sinβ sin α cos β ± cos α sin β sin (α ± β )

tg α ± tg β ≡ ± ≡ ≡
cos α cosβ cos α cos β cosα cosβ
Product of vectors and addition identities
Let us see at the figure 6.4, but now we calculate the product of both vectors:
v u = ( e1 cos β + e 2 sin β ) ( e1 cos α + e 2 sin α ) =
= cos α cos β + sin α sin β + e12 (sin α cos β − cos α sin β )
Since u and v are unitary vectors, their product is a complex number with unitary
modulus and argument equal to the oriented angle between them:
v u = cos(α − β ) + e12 sin(α − β )
The identification of both equations gives the trigonometric functions of the angles
difference:
sin(α − β ) ≡ sinα cos β − cos α sin β
cos(α − β ) ≡ cos α cos β + sin α sin β
Taking into account that the sine and the cosine are odd and even functions respectively,
one obtains the identities for the angles addition:
sin(α + β ) ≡ sinα cos β + cos α sin β
cos(α + β ) ≡ cos α cos β − sin α sin β
Rotations and De Moivre’s identity
If v' is the result of turning the vector v over an angle α then:
v' = v (cos α + e12 sin α )
To repeat a rotation of angle α by n times is the same thing as to turn over an angle nα :
v'' = v ( cos α + e12 sinα ) = v ( cos nα + e12 sin nα )

n
cos nα + e12 sin nα ≡ ( cos α + e12 sinα )

n
⇒
This is the De Moivre’s identity4, which allows to calculate the trigonometric

functions of multiplies of a certain angle through the binomial theorem. For example, for n
= 3 we have:
cos 3α + e12 sin 3α ≡ (cos α + e12 sin α )

3
≡ cos 3 α − 3 cos α sin 2α + e12 ( 3 cos 2 α sin α − sin 3α )
Splitting the real and imaginary parts one obtains:
cos 3α ≡ cos 3 α − 3 cos α sin 2α sin3α ≡ 3 cos 2α sinα − sin 3α
And dividing both identities one arrives at:
3 tg α − tg 3 α
tg 3α ≡
1 − 3 tg 2 α
4
With the Euler’s identity, the De Moivre’s identity is:
exp( nα e12 ) ≡ [exp(α e12 )]

n
Inverse trigonometric functions
The arcsine, arccosine and arctangent are defined as the inverse functions of the
sine, cosine and tangent. They are multivalued functions and the principal values are taken
in the following intervals:
π π
y = arcsin x ⇔ x = sin y and − ≤ y≤
2 2
y = arccos x ⇔ x = cos y and 0 ≤ y ≤ π
π π
y = arctg x ⇔ x = tg y and − < y<
2 2
y = arccot x ⇔ x = cot y and 0 < y < π
From the definitions the parity of the inverse functions follow immediately:
arcsin x ≡ −arcsin (− x ) arccos x ≡ arccos(− x )
arctg x ≡ − arctg (− x )
Through the fundamental identities for the circular functions one obtains the
following identities5 for the inverse functions:
[ ]
arcsin x ≡ arccos 1 − x 2 ≡ arctg
x
1− x2
[ ]

arccos x ≡ arcsin 1 − x 2 ≡ arctg
1− x2 
x 


x  1 
arctg x ≡ arcsin ≡ arccos 
1+ x2  1+ x2 
where the brackets indicate that the identity only holds for positive values.
From the complementary angles identities we have:
π π
arcsin x ≡ − arccos x arctg x ≡ − arc cot x
2 2
5
The only inverse circular function predefined in the language Basic is the arctangent ATN(X), so
these identities allows us to program the arcsine and arccosine.
Exercises
6.1 Prove the law of sines, cosines and tangents: if a, b and c are the sides of a triangle
respectively opposite to the angles α, β and γ, then:
a b c
= = (law of sines)
sin α sin β sin γ
c2 = a2 + b2 − 2 a b cos γ (law of cosines)
α+β
a + b tg
= 2
(law of tangents)
a − b α−β
tg
2
Hint: use the inner and outer products of sides.
6.2 Prove the following trigonometric identities:
α−β β −γ γ −α
cos(α − β ) + cos(β − γ ) + cos(γ − a ) ≡ 4 cos cos cos −1
2 2 2
sin(α − β ) + sin(β − γ ) + sin(γ − α ) ≡ −4 sin sin sin
2 2 2
6.3 Express the trigonometric functions of 4α as a polynomial of the trigonometric

functions of α .
6.4 Let P be a point on a circle arc whose extremes are the points A and B. Prove that the
sum of the chords AP and PB is maximal when P is the midpoint of the arc AB.
6.5 Prove the Mollweide’s formulas for a triangle:
α−β α−β
a + b cos a − b sin
= 2 = 2
c γ c γ
sin cos
2 2
6.6 Deduce also the projection formulas:
a = b cos γ + c cos β b = c cos α + a cos γ c = a cos β + b cos α
6.7 Prove the half angle identities:
α 1 − cos α α 1 + cos α α 1 − cos α 1 − cos α sinα

sin ≡± cos ≡± tg ≡ ± ≡ ≡
2 2 2 2 2 1 + cos α sinα 1 + cos α
7. SIMILARITIES AND SINGLE RATIO
Two geometric figures are similar if they have the same shape. If the orientation of both
figures is the same, they are said to be directly similar. For example, the dials of clocks are
directly similar. On the other hand, two figures can have the same shape but different
orientation. Then they are said to be oppositely similar. For example our hands are oppositely
similar. However these intuitive concepts are insufficient and the similarity must be defined
with more precision.
Direct similarity (similitude)
If two vectors u and v on the plane

form the same angle as the angle between the
vectors w and t, and they have proportional Figure 7.1
lengths (figure 7.1), then they are
geometrically proportional:
α (u, v ) = α (w, t )
u w 
 ⇒ u v −1 = w t −1
=
v t 

that is, the geometric quotient of u and v is equal to the quotient of w and t, which is a complex
number. This definition of a geometric quotient is also valid for vectors in the space provided
that the four vectors lie on the same plane. In this case, the geometric quotient is a quaternion, as
Hamilton showed.
The geometric proportionality for vectors allows to define the similarity of triangles.
Two triangles ABC and A'B'C' are said to be directly similar and their vertices and sides denoted
with the same letters are homologous if:
AB BC −1 = A'B' B'C' −1
that is, if the geometric quotient of two sides of the first triangle is equal to the quotient of the
homologous sides of the second triangle. Arranging the vectors of this equation one obtains the
equality of the quotients of the homologous sides:
AB−1 A'B' = BC −1 B'C'
One can prove easily that the third quotient of homologous sides also coincides with the
other quotients:
AB−1 A'B' = BC −1 B'C' = CA−1 C'A' = r
The similarity ratio r is defined as the quotient of every pair of homologous sides, which
is a complex constant. The modulus of the similarity ratio is the size ratio and the argument is
the angle of rotation of the triangle A'B'C' with respect to the triangle ABC.
A' B'
r= exp[α ( AB, A' B' ) e12 ] Figure 7.2
AB
The definition of similarity is

generalised to any pair of polygons in the
following way. Let the polygons ABC...Z and
A'B'C'...Z' be. They are said to be directly
similar with similarity ratio r and the sides
denoted with the same letters to be homologous
if:
r = AB −1 A'B' = BC −1 B'C' = CD −1 C'D' = ... = YZ −1 Y'Z' = ZA −1 Z'A'
One of these equalities depends on the others and we do not need to know whether it is
fulfilled. Here also, the modulus of r is the size ratio of both polygons and the argument is the
angle of rotation. The fact that the homologous exterior and interior angles are equal for directly
similar polygons (figure 7.2) is trivial because:
B'A' B'C' −1 = BA BC −1 ⇒ angle A'B'C' = angle ABC
C'B' C'D' −1 = CB CD −1 ⇒ angle B'C'D' = angle BCD etc.
The direct similarity is an equivalence relation since it has the reflexive, symmetric and
transitive properties. This means that there are classes of equivalence with directly similar
figures.
A similitude with r=1 is called a displacement, since both polygons have the same
size and orientation.
Opposite similarity
Two triangles ABC and A'B'C' are

oppositely similar and the sides denoted with
the same letters are homologous if:
AB BC −1 = ( A'B' B'C' −1 )* = B'C' −1 A'B'
The former equality cannot be arranged in a

quotient of a pair of homologous sides as we
have made before. Because of this, the
similarity ratio cannot be defined for the Figure 7.3
opposite similarity but only the size ratio,
which is the quotient of the lengths of any two
homologous sides. An opposite similarity is always the composition of a reflection in any line
and a direct similarity.
AB BC −1 = v −1 A'B' B'C' −1 v ⇔ BC −1 v −1 B'C' = AB −1 v −1 A'B' ⇔
BC −1 ( v −1 B'C' v ) = AB −1 ( v −1 A'B' −1 v ) = r
where r is the ratio of a direct similarity whose argument is not defined but depends on the
direction vector v of the reflection axis. Notwithstanding, this expression allows to define the
opposite similarity of two polygons. So two polygons ABC...Z and A'B'C'...Z' are oppositely
similar and the sides denoted with the same letters are homologous if for any vector v the
following equalities are fulfilled:
AB −1 ( v −1 A'B' −1
v ) = BC −1 ( v −1 B'C' v ) = .... = ZA −1 ( v −1 Z'A' v )
that is, if after a reflection one polygon is directly similar to the other. The opposite similarity is
not reflexive nor transitive: if a figure is oppositely similar to another, and this is oppositely
similar to a third figure, then the first and third figures are directly similar. Then there are not
classes of oppositely similar figures.
An opposite similarity with r=1 is called a reversal, since both polygons have the
same size both opposite orientations.
The theorem of Menelaus
For every triangle ABC (figure 7.4), Figure 7.4

three points D, E and F lying respectively on
the sides BC, CA and AB or their prolongations
are aligned if and only if:
AF FB −1 BD DC −1 CE EA −1 = −1
Proof ⇒ Let us suppose that D, E and F are

aligned on a crossing straight line. Let us
denote by p, q and r the vectors with origin at
the vertices A, B and C and going
perpendicularly to the crossing line. Then every pair of right angle triangles having the
hypotenuses on a common side of the triangle ABC are similar so that we have:
BF −1 AF = q −1 p CD −1 BD = r −1 q AE −1 CE = p −1 r
Multiplying the three equalities one obtains:
AE −1 CE CD −1 BD BF −1 AF = 1
Taking the complex conjugate expression and changing the sign of BF, CD and AE, the theorem
is proved in this direction:
AF FB −1 BD DC −1 CE EA −1 = −1
Proof ⇐ Being F a point on the line AB, D a point on the line BC and E a point on the line CA
we have:
F = a A + (1 − a) B D = b B + (1 − b) C E = c C + (1 − c) A
with a, b and c real. Then:
AF = (1 − a) AB FB = a AB BD = (1 − b) BC
DC = b BC CE = (1 − c) CA EA = c CA
That the product of these segments is equal to −1 implies that:
(1 − a )(1 − b )(1 − c ) (1 − a )(1 − b )

= −1 ⇒ c=
abc 1− a − b
The substitution of c in the expression for E gives:
(1 − a )(1 − b ) ab
E= C− A
1− a − b 1− a − b
1− a
= [(1 − b) C + b B] − b [a A + (1 − a ) B]
1− a − b 1− a − b
1− a b
= D− F
1− a − b 1− a − b
That is, the point E belongs to the line FD, in other words, the point D, E and F are
aligned, which is the prove:
1− a
E=dD+(1−d)F with d =
1− a − b
The theorem of Ceva

Figure 7 5
Given a generic triangle ABC, we draw
three segments AD, BE and CF from each vertex
to a point of the opposite side (figure 7.5). The
three segments meet in a unique point if and
only if:
BD DC −1 CE EA −1 AF FB −1 = 1
In order to prove the theorem, we

enlarge the segments BE and CF to touch the line which is parallel to the side BC and passes
through A. Let us denote the intersection points by M and N respectively. Then the triangle BDO
is directly similar to the triangle MAO, and the triangle CDO is also directly similar to the
triangle NAO. Therefore:
BD DO −1 = MA AO −1
DO DC −1 = AO AN −1
The multiplication of both equalities gives:
BD DC −1 = MA AN −1
Analogously, the triangles MEA and CEB are similar as soon as NFA and BFC. Hence:
CE EA −1 = BC AM −1 AF FB −1 = AN BC −1
The product of the three equalities yields:
BD DC −1 CE EA −1 AF FB −1 = MA AN −1 BC AM −1 AN BC −1 = 1
The sufficiency of this condition is proved in the following way: let O be the point of
intersection of BE and CF, and D' the point of intersection of the line AO with the side BC. Let
us suppose that the former equality is fulfilled. Then:
BD' D'C −1 = BD DC −1
Since both D and D' lie on the line BC, it follows that D = D'.
Homothety and single ratio
A homothety with centre O and ratio k is the geometric transformation which dilates the
distance from O to any point A in a factor k:
Figure 7.6
OA' = OA k
O is the unique invariant point of the

homothety. If k is real number, the homothety is
said to be simple (figure 7.6), otherwise is called
composite. If k is a complex number, it may always
be factorised in a product of the modulus and a
unitary complex number:
k = k z with z = 1
The modulus of k is the ratio of a simple homothety with centre O, and z indicates an
additional rotation with the same centre. Then, a composite homothety is equivalent to a simple
homothety followed by a rotation. Direct similarities and homotheties are different names for
the same transformations (exercise 7.5). Also homotheties with k=1 and displacements are
equivalent.
The single ratio of three points A, B, and C is defined as:
( A B C ) = AB AC −1
The single ratio is a real number when the three points are aligned, and a complex
number in the other case. Under a homothety, the single ratio of any three points remains
invariant. Let us prove this. Because any vector is dilated and rotated by a factor k, we have:
A'B' = OB' − OA' = OB k − OA k = AB k
A'C' = AC k ⇒ A'C' −1 = k −1 AC −1
( A' B' C' ) = AB k k −1 AC −1 = AB AC −1 = ( A B C )
If the single ratio is invariant for a geometric transformation, then it transforms triangles
into directly similar triangles and hence also polygons into similar polygons:
−1 −1 −1
( A B C ) = ( A' B' C' ) ⇒ AB AC = A'B' A'C' ⇒ AC A'C' = AB −1 A'B'
Therefore, the homothety always transforms triangles into directly similar triangles as
shown in the figure 7.6. It follows immediately that the homothety preserves the angles between
lines. It is the simplest case of conformal transformations, the geometric transformations which
preserve the angles between lines. Since a line
may also be transformed into a curve, and for
the general case a curve into a curve, the
conformal transformations are those which
preserve the angle between any pair of curves,
that is, the angle between the tangent lines to
both curves at the intersection point. A
transformation is directly or oppositely Figure 7.7
conformal whether it preserves or changes the
sense of the angle between two curves (figure
7.7). This condition is equivalent to the
conservation of the single ratio of any three
points at the limit of accumulation:
lim (A B C ) = lim ( A' B' C' ) directly conformal

B, C → A B' , C' → A'
lim (A B C ) = lim ( A' B' C' ) * oppositely conformal

B, C → A B' , C' → A'
Exercises
7.1 Let T be the triangle with vertices (0, 0), (2, 0), (0, 1) and T' that with vertices (2,0), (5, 1)
and (4, 2). Find which vertices are homologous and calculate the similarity ratio. Which is
the size ratio? Which is the angle of rotation of the triangle T' with respect to T?
7.2 The altitude perpendicular to the hypotenuse divides a right triangle in two smaller right
triangles. Show that they are similar and deduce the Pythagorean theorem.
7.3 Every triangle ABC with not vanishing area has a circumscribed circle. The line tangent to
this circle at the point B cuts the line AC at the point M. Prove that:
MA MC −1 = AB2 BC −2
7.4 Let ABC be an equilateral triangle inscribed inside a circle. If P is any point on the arc BC,
show that PA = PB + PC.
7.5 Given two directly similar triangles ABC and A'B'C', show that the centre O of the
homothety that transform one triangle into another is equal to:
O = A − AA' ( 1 − AB −1 A'B' ) −1
7.6 Draw any line passing through a fixed point P which cuts a given circle. Let Q and Q' be the
intersection points of the line and the circle. Show that the product PQ PQ' is constant for
any line belonging to the pencil of lines of P.
7.7 Let a triangle have sides a, b and c. The bisector of the angle formed by the sides a and b
divides the side c in two parts m and n. If m is adjacent to a and n to b respectively, prove
that the following proportion is fulfilled:
m n
=
a b
8. PROPERTIES OF TRIANGLES
Area of a triangle
Since the area of a parallelogram is obtained as the outer product of two

consecutive sides (taken as vectors, of course), the area of a triangle is the half of the outer
product of any two of its three sides:
1 1 1
aPQR = PQ ∧ PR = QR ∧ QP = RP ∧ RQ
2 2 2
If the vertices P, Q and R are counterclockwise oriented, the area is a positive

imaginary number. Otherwise, the area is a
negative imaginary number. Note that the
fundamental concept in geometry is the
oriented area1. The modulus of the area Figure 8.1
may be useful in the current life but is
insufficient for geometry. From now on I
shall only regard oriented areas.
Writing the segments of the former
equation as differences of points we arrive
to:
1
aPQR = (P ∧ Q + Q ∧ R + R ∧ P )
2
which is a symmetric expression under cyclic permutation of the vertices. The position
vector P goes from an arbitrary origin of coordinates to the point P. Then, P ∧ Q is the
double of the area of the triangle OPQ. Analogously Q ∧ R is the double of the area of the
triangle OQR and R ∧ P is the double of the area of the triangle ORP. Therefore the
former expression is equal to (figure 8.1):
aPQR = aOPQ + aOQR + aORP
For the arrangement of points shown in the figure 8.1 the area of OPQ is positive,
and the areas of OQR and ORP are negative, that is, the areas of ORQ and OPR are
positive. Therefore, one would intuitively write, taking all the areas positive, that:
aPQR = aOPQ − aORQ − aOPR
When one considers oriented areas, the equalities are wholly general and
independent of the arrangement of the points2.
1
The integral of a function is also the oriented area of the region enclosed by the curve and the X-
axis.
2
About this topic, see A. M. Lopshitz, Cálculo de las áreas de figuras orientadas, Rubiños-1860
(1994).
Medians and centroid
The medians are the segments going from each vertex to the midpoint of the
opposite side. Let us prove that the three medians meet in a unique point G called the
centroid (figure 8.2). Since G is a point on the median passing through P and (Q + R)/2:
Q+R
G = k P + (1 − k ) k real
2
G lies also on the median passing through Q and (P + R) / 2:
P+R
G = m Q + (1 − m ) m real
2
Equating both expressions we find:
Q+R P+R
k P + (1 − k ) = m Q + (1 − m )
2 2
 1 m 1 k   k m
P k − +  + Q  − − m + R − +  = 0
 2 2 2 2   2 2
A linear combination of independent points can vanish only if every coefficients are null,
a condition which leads to the following
system of equations:
Figure 8.2
 1 m
k − 2 + 2 = 0
 1 k
 − −m=0
2 2
 − k + m =0
 2 2
The solution of this system of

equations is k =1/3 and m =1/3, indicating
that the intersection of both medians are
located at 1/3 distance from the midpoints. The substitution into the expression of G
gives:
P+Q + R
G=
3
This expression for the centroid is symmetric under permutation of the vertices. Therefore
the three medians meet at the same point G, the centroid3.
3
The medians are a special case of cevian lines (lines passing through a vertex and not
coinciding with the sides) and this statement is also proved by means of Ceva’s theorem.
Perpendicular bisectors and circumcentre
The three perpendicular bisectors of the sides of a triangle meet in a unique point
called the circumcentre, the centre of the circumscribed circle. Every point on the
perpendicular bisector of PQ is equidistant from P and Q. Analogously every point on the
perpendicular bisector of PR is equidistant from P and R. The intersection O of both
perpendicular bisectors is simultaneously equidistant from P, Q and R. Therefore O also
belongs to the perpendicular bisector of QR and the three bisectors meet at a unique point.
Since O is equally distant from the three vertices, it is the centre of the circumscribed
circle. Let us use this condition in order to
calculate the equation of the circumcentre:
Figure 8.3
OP2 = OQ2 = OR2 = d2
where d is the radius of the circumscribed

circle. Using the position vectors of each
point we have:
(P − O)2 = (Q − O)2 = (R − O)2
The first equality yields:
P2 − 2 P · O + O2 = Q2 − 2 Q · O + O2
By simplifying and arranging the terms containing O at the left hand side, we have:
2 ( Q − P ) · O = Q2 − P2
2 PQ · O = Q2 − P2
From the second equality we find an analogous result:
2 QR · O = R2 − Q2
Now we introduce the geometric product instead of the inner product in these equations:
PQ O + O PQ = Q2 − P2
QR O + O QR = R2 − Q2
By subtraction of the second equation multiplied on the right by PQ minus the first
equation multiplied on the left by QR, we obtain:
PQ QR O − O PQ QR = PQ R2 − PQ Q2 − Q2 QR + P2 QR
By using the permutative property on the left hand side and simplifying the right hand
side, we have:
PQ QR O − QR PQ O = P2 QR + Q2 RP + R2 PQ
2 ( PQ ∧ QR ) O = P2 QR + Q2 RP + R2 PQ
Finally, the multiplication by the inverse of the outer product on the left gives:
O = ( 2 PQ ∧ QR ) −1 ( P2 QR + Q2 RP + R2 PQ )
= − ( P2 QR + Q2 RP + R2 PQ ) ( 2 PQ ∧ QR ) −1
a formula able to calculate the coordinates of the circumcentre. For example, let us
calculate the centre of the circle passing through the points:
P = ( 2, 2 ) Q = ( 3, 1 ) R = ( 4, −2 )
P2 = 8 Q2 = 10 R2 = 20
QR = R − Q = e1 − 3 e2 RP = P − R = − 2 e1 + 4 e2 PQ = Q − P = e1 − e2
2 PQ ∧ QR = − 4 e12
e12
O = − ( 8 ( e1 − 3 e2 ) + 10 ( −2 e1 + 4 e2 ) + 20 ( e1 − e2 ) )
4
= − e1 − 2 e2 = ( −1, −2 )
In order to deduce the radius of the circle, we take the vector OP:
OP = P − O = P + ( P2 QR + Q2 RP + R2 PQ ) ( 2 PQ ∧ QR ) −1
and extract the inverse of the area as a common factor:
OP = ( 2 P PQ ∧ QR + P2 QR + Q2 RP + R2 PQ ) ( 2 PQ ∧ QR ) −1 =
= [ 2 P ( P ∧ Q + Q ∧ R + R ∧ P ) + P2 QR + Q2 RP + R2 PQ ] ( 2 PQ ∧ QR ) −1 =
= [ P ( P Q − Q P + Q R − R Q + R P − P R ) + P2 ( R − Q ) + Q2 ( P − R ) +
+ R2 ( Q − P ) ] ( 2 PQ ∧ QR ) −1
The simplification gives:
OP = ( P Q R − P R Q + P R P − P Q P + Q2 P − Q2 R + R2 Q − R2 P ) ( 2 PQ ∧ QR ) −1
= − (Q − P) (R − Q) (P − R) ( 2 PQ ∧ QR ) −1 = − PQ QR RP ( 2 PQ ∧ QR ) −1
Analogously:
OQ = − QR RP PQ ( 2 PQ ∧ QR ) −1 OR = − RP PQ QR ( 2 PQ ∧ QR ) −1
The radius of the circumscribed circle is the length of any of these vectors:
PQ QR RP PQ QR RP
OP = = = =
2 PQ ∧ QR 2 sin QRP 2 sin RPQ 2 sin PQR
where we find the law of sines.
Angle bisectors and incentre
The three bisector lines of the angles of a Figure 8.4

triangle meet in a unique point called the incentre.
Every point on the bisector of the angle with
vertex P is equidistant from the sides PQ and PR
(figure 8.4). Also every point on the angle
bisector of Q is equidistant from the sides QR and
QP. Hence its intersection I is simultaneously
equidistant from the three sides, that is, I is unique
and is the centre of the circle inscribed into the
triangle.
In order to calculate the equation of the
angle bisector passing through P, we take the sum
of the unitary vectors of both adjacent sides:
PQ PR QP QR
u= + v= +
PQ PR QP QR
The incentre I is the intersection of the angle bisector passing through P, whose
direction vector is u, and that passing through Q, with direction vector v:
I=P+ku=Q+mv k, m real
Arranging terms we find PQ as a linear combination of u and v:
k u − m v = Q − P = PQ
The coefficient k is:
PQ ∧ v PQ RP PQ ∧ QR
k= =
u∧v PQ ∧ QR RP + QR ∧ RP PQ + RP ∧ PQ QR
Since all outer products are equal because they are the double of triangle area, this
expression is simplified:
PQ RP
k=
RP + PQ + QR
Then, the centre of the circumscribed circle is:
PQ RP  PQ PR 
I =P+ku=P+  + 
PQ + QR + RP  PQ PR 
By taking common denominator and simplifying, we arrive at:
P QR + Q RP + R PQ
I=
QR + RP + PQ
For example, let us calculate the centre of the circle inscribed inside the triangle with
vertices:
P = (0, 0) Q = (0, 3) R = (4, 0)
PQ = 3 QR = 5 RP = 4
5 (0, 0) + 4 (0, 3) + 3 (4, 0) (12, 12 )

I= = = (1, 1)
5+4+3 12
In order to find the radius, firstly we must obtain the segment IP:
QP RP + RP PQ
IP =
QR + RP + PQ
The radius of the inscribed circle is the distance from I to the side PQ:
IP ∧ PQ RP ∧ PQ
d (I , PQ ) = =
PQ PQ + QR + RP
whence the ratio of radius follows:
radius of circumscribed circle 1 PQ QR RP

=
radius of inscribed circle 2 PQ + QR + RP
Altitudes and orthocentre
The altitude of a side is the segment perpendicular to this side (also called base)
which passes through the opposite vertex. The three altitudes of a triangle intersect on a
unique point called the orthocentre. Let us prove this statement calculating the
intersection H of two altitudes. Since H belongs to the altitude passing through the vertex
P and perpendicular to the base QR (figure 8.5), its equation is:
H = P + z QR z imaginary
Figure 8.5
because the product by an imaginary
number turns the vector QR over π/2, that
is, z PQ has the direction of the altitude. H
also belongs to the altitude passing through
Q and perpendicular to the base RP. Then
its equation is:
H = Q + t RP t imaginary
By equating both expressions:
P + z QR = Q + t RP
we arrive at a vector written as a linear combination of two vectors but with imaginary
coefficients:
z QR − t RP = PQ
In this equation we must resolve PQ into components with directions perpendicular to the
vectors QR and RP. The algebraic resolution follows the same way as for the case of real
linear combination. Let us multiply on the right by RP:
z QR RP − t RP2 = PQ RP
and on the left:
RP z QR − RP t RP = RP PQ
The imaginary numbers anticommute with vectors:

2
− z RP QR + t RP = RP PQ
By adding both equalities we arrive at:
z ( QR RP − RP QR ) = PQ RP + RP PQ
PQ RP + RP PQ PQ · RP
z= =
QR RP − RP QR QR ∧ RP
Analogously one finds t:
PQ · QR
t=
QR ∧ RP
Both expressions differ from the coefficients of real linear combination in the fact
that the numerator is an inner product. The substitution in any of the first equations gives:
H = P + z QR = P + ( PQ RP + RP PQ ) ( QR RP − RP QR ) −1 QR =
The vector QR anticommutes with the outer product, an imaginary number:
H = P − ( PQ RP + RP PQ ) QR ( QR RP − RP QR ) −1 =
Extracting the area as common factor, we obtain:
H = [ P ( QR RP − RP QR ) − ( PQ RP + RP PQ ) QR ] ( QR RP − RP QR ) −1
By using the fact that PQ = Q − P, etc., we arrive at:
H = ( P QR RP − Q RP QR − P PQ QR + R PQ QR ) ( QR RP − RP QR ) −1
= ( P2 QR + Q2 RP + R2 PQ + P QR P + Q RP Q + R PQ R ) ( QR RP − RP QR ) −1
= ( P P · QR + Q Q · RP + R R · PQ ) ( QR ∧ RP ) −1
This formula is invariant under cyclic permutation of the vertices. Therefore all the
altitudes intersect on a unique point,
the orthocentre.
The equation of the orthocentre Figure 8.6
resembles that of the circumcentre. In
order to see the relationship between
both, let us draw a line passing through
P and parallel to the opposite side QR,
another one passing through Q and
parallel to RP, and a third line passing
through R and parallel to PQ (figure
8.6). Let A be the intersection of the
line passing through Q and that passing
through R, B be the intersection of the
lines passing through R and P respectively, and C be the intersection of the lines passing
through P and Q respectively. The triangle ABC is directly similar to the triangle PQR
with ratio −2:
1 1 1
PQ = − AB QR = − BC RP = − CA
2 2 2
and P, Q and R are the midpoints of the sides of the triangle ABC:
B+C C+A A+ B
P= Q= R=
2 2 2
Therefore the altitudes of the triangle PQR are the perpendicular bisectors of the
sides of the triangle ABC, and the orthocentre of the triangle PQR is the circumcentre of
ABC. In order to prove with algebra this obvious geometric fact, we must only deduce
from the former relations the following equalities and substitute them into the orthocentre
equation:
B3 − B C 2 + C B2 − C 3
P P · QR =
8
C − C A + A C 2 − A3
3 2
Q Q · RP =
8
A − A B + B A2 − B 3
3 2
R R · PQ =
8
By adding the three terms, the cubic powers vanish. On the other hand, the area of
the triangle PQR is four times smaller than the area of the triangle ABC:
BC ∧ CA
QR ∧ RP =
4
H = ( −B C2 + C B2 − C A2 + A C2 − A B2 + B A2 ) ( 2 BC ∧ CA ) −1 =
= − ( A2 BC + B2 CA + C2 AB ) ( 2 BC ∧ CA ) −1
This is just the equation of the circumcentre of the triangle ABC.
Euler's line
The centroid G, the circumcentre O and the orthocentre H of a triangle are always
aligned on the Euler's line. To prove this fact, observe that the circumcentre and
orthocentre equations have the triangle area as a "denominator", while the centroid
equation does not, but we can introduce it:
P+Q + R
G= = ( P + Q + R ) QR ∧ RP ( 3 QR ∧ RP ) −1
3
Introducing the equality:
P Q −Q P +Q R − R Q + R P − P R
QR ∧ RP = P ∧ Q + Q ∧ R + R ∧ P =
2
the centroid becomes:
G = ( P + Q + R ) ( P Q − Q P + Q R − R Q + R P − P R ) ( 6 QR ∧ RP ) −1
=(PPQ−PQP+PRP−PPR+QPQ−QQP+QQR−QRQ+
R Q R − R R Q + R R P − R P R ) ( 6 QR ∧ RP ) −1
= ( − P2 QR + P QR P − Q2 RP + Q RP Q − R2 PQ + R PQ R ) ( 6 QR ∧ RP ) −1
from where the relation between the three points G, H and O follows immediately:
H +2O
G=
3
Hence the centroid is located between the orthocentre and the circumcentre, and its
distance from the orthocentre is double of its distance from the circumcentre.
The Fermat's theorem
The geometric algebra allows to prove through a very easy and intuitive way the
Fermat's theorem.
Over every side of a triangle ABC we draw an equilateral triangle (figure 8.7). Let
T, U and S be the vertices of the
Figure 8.7
equilateral triangles respectively
opposite to A, B and C. Then the
segments AT, BU and CS have the same
length, form angles of 2π/3 and intersect
on a unique point F, called the Fermat's
point. Moreover, the addition of the
three distances from any point P to each
vertex is minimal when P is the Fermat's
point, provided that any of the interior
angles of the triangle ABC be higher
than 2π/3.
Firstly, we must demonstrate that
BU is obtained from AT by means of a rotation of 2π/3, which will be represented by the
complex number t:
2π 2π
t = cos + e12 sin
3 3
AT t = (AC + CT ) t = AC t + CT t
By construction, the vector AC turned over 2π/3 is the vector CU, and CT turned over
2π/3 is BC, so:
AT t = CU + BC = BU
Analogously, one finds CB = BU t and AT = CS t . That is, the vectors CS, BU and AT
have the same length and each one is obtained from each other by successive rotations of
2π/3.
Let us see that the sum of the distances from P to the three vertices A, B and C is
minimal when P is the Fermat’s point. We must prove firstly that the vectorial sum of PA
turned 4π/3, PB turned 2π/3 and PC is constant independently of the point P. (figure 8.8).
That is, for any two points P and P’ it is always true that:
PA t2 + PB t + PC = P'A t2 + P'B t + P'C
fact which is proved by arranging all the terms in one side of the equation:
PP' ( t2 + t + 1 ) = 0
This product is always zero since t2 + t + 1 = 0. Hence, there is a unique point Q such
that:
PA t2 + PB t + PC = QC
Figure 8.8
For any point P, the three segments form a
broken line as that shown in figure 8.8.
Therefore, by the triangular inequality we
have:
PA + PB + PC ≥ QC
When P is the Fermat’s point F, these

segments form a straight line. Then, the
addition of the distances from F to the three
vertices is minimal provided that no angle of
the triangle be higher than 2π/3:
FA + FB + FC = QC ≤ PA + PB + PC
In the other case, someone of the vectors FA t2, FB t or FC has a sense opposite to the
others, so that its length is subtracted from the others and their sum is not minimal.
Exercises
8.1 The Napoleon’s theorem. Over each side of a generic triangle draw an equilateral
triangle. Prove that the centres of the three equilateral triangles also form an
equilateral triangle.
8.2 Prove the Leibniz’s theorem. Let P be any point on the plane and G the centroid of a
triangle ABC. Then the following equality holds:
1
3 PG2 = PA2 + PB2 + PC2 − ( AB2 + BC2 + CA2 )
3
8.3 Let A, B and C be three given points on the plane. Then any point G on the plane can
be expressed as a linear combination of these three points (G is also considered as the
centre of masses located at A, B and C with weights a, b and c 4 ).
4
See August Ferdinand Möbius, Der Barycentrische Calcul, Leipzig, 1827, p. 17 (facsimile
edition of Georg Olms Verlag, Hildesheim,1976).
G=aA+bB+cC with a + b + c = 1
Prove that:
a) a, b, c are the fractions of the area of the triangle ABC occupied by the
triangles GBC, GCA and GAB respectively.
b) the geometric locus of the points P on the plane such that:
a PA2 + b PB2 + c PC2 = k
is a circle with centre G (Apollonius’ lost theorem).
8.4 Over every side of a convex quadrilateral ABCD, we draw the equilateral triangles
ABP, BCQ, CDR and DAS. Prove that:
a) If AC = BD then PR is perpendicular to QS.
b) If PR = QS then AC is perpendicular to BD.
8.5 Show that the length of a median of a triangle ABC passing through A can be
calculated from the sides as:
AB 2 + AC 2 BC 2
m A2 = −
2 4
8.6 In a triangle ABC we draw the bisectors of the angles A and B. Through the vertex C
we draw the lines parallel to each angle bisector. Let us denote the intersections of
each parallel line with the other bisector by D and E. Prove that if the line DE is
parallel to the side AB then the triangle ABC is isosceles.
8.7 The bisection of a triangle (proposed by Cristóbal Sánchez Rubio). Let P and Q be
two points on different sides of a triangle such that the segment PQ divides the
triangle in two parts with the same area. Calculate the segment PQ in the following
cases:
a) PQ is perpendicular to a given direction.
b) PQ has minimum length.
c) PQ passes through a given point inside the triangle.
9. CIRCLES
Algebraic and Cartesian equations
A circle with radius r and centre F is the locus of the points located at a distance r
from F, so its equation is:
d(F, P) =FP= r ⇔ FP2 = r2
which can be written as:
(P − F)2 = r2 ⇒ P2 − 2 P · F + F2 − r2 = 0
By introducing the coordinates of P = (x, y) and F = (a, b) one obtains the analytic
equation of the circle:
x2 + y2 − 2 a x − 2 b y + a2 + b2 − r2 = 0
For example, the circle whose equation is:
x2 + y2 + 4 x − 6 y + 9 = 0
has the centre F = (−2, 3) and radius 2.

There always exists a unique circle passing through any three non aligned points.
In order to obtain this circle, one may substitute the coordinates of the points on the
Cartesian equation of the circle, arriving at a system of three linear equations of three
unknown quantities a, b and c:
 x1 2 + y1 2 + a x1 + b y1 + c = 0
 2 2
 x2 + y 2 + a x2 + b y 2 + c = 0
x 2 + y 2 + a x + b y + c = 0
 3 3 3 3
This is the classic way to find the equation of the circle, and from here the centre of the
circle.
However, a more direct way is the calculation of the circumcentre of the triangle
whose vertices are the three given points (see the previous chapter). Then the radius is the
distance from the centre to any vertex.
Intersections of a line with a circle
The coordinates of the intersection of a line with a circle are usually calculated by
substitution of the line equation into the circle equation, which leads to a second degree
equation. Let us now follow an algebraic way. Let F be the centre of the circle with radius
r, and R and R’ the intersections of the line passing through two given points P and Q
(figure 9.1):
Figure 9.1
 FR 2 = r 2

 R = P + k PQ
From the first equality we obtain:
(R − F )2 = R2 − 2 R · F + F2 = r2
The substitution of R given by the second

equality yields:
( P + k PQ )2 − 2 ( P + k PQ ) · F + F2 = r2
P2 + 2 k P · PQ + k2 PQ2 − 2 P · F − 2 k PQ · F + F2 − r2 = 0
By arranging the powers of k one obtains:
k2 PQ2 + 2 k ( P · PQ − F · PQ ) + P2 − 2 F · P + F2 − r2 = 0
Taking into account that PF = F − P one finds:
k2 PQ2 − 2 k PF · PQ + PF2 − r2 = 0
This equation has solution whenever the discriminant be positive:
4 ( PF · PQ )2 − 4 PQ2 ( PF2 − r2 ) ≥ 0
Introducing the identity (PF · PQ )2 − (PF ∧ PQ )2 = PQ2 PF2 the discriminant becomes:
4 (PF ∧ PQ )2 + 4 PQ2 r2 ≥ 0
The solution of the second degree equation for k is:
k = PF · PQ −1 ± r 2 PQ − 2 + (PF ∧ PQ −1 )
2
When r = PF ∧ PQ/PQ, the line is tangent to the circle. In this case the height of the
parallelogram formed by the vectors PF and PQ is equal to the radius of the circle.
If u denote the unitary direction vector of the line PQ (figure 9.1):
PQ
u=
PQ
then both intersection points are written as:

R = P + kPQ = P + u  PF · u − r 2 + (PF ∧ u ) 
2
 
R' = P + k' PQ = P + u  PF · u + r 2 + (PF ∧ u ) 
2
 
Power of a point with respect to a circle
Both intersection points R and R' have the following property: the product of the
vectors going from a given point P to the intersections R and R' of any line passing
through P with a given circle is constant:
PR PR' = ( PF · u )2 − r2 − (PF ∧ u )2 Figure 9.2
= PF2 − r2 = PT2
where T is the contact point of the tangent

line passing through P (figure 9.2). The
product PR PR’, which is independent of
the direction u of the line and only
depends on the point P, is called the
power of the point P with respect to the
given circle. The power of a point may be
calculated through the substitution of its coordinates into the circle equation. If its centre
is F = (a, b) and P = (x, y), then:
FP2 − r2 = ( x − a )2 + ( y − b )2 − r2 = x2 + y2 − 2 a x − 2 b y + a2 + b2 − r2
The power is equal to zero when P belongs to the circle, positive when P lies outside the
circle, and negative when P lies inside the circle.
Polar equation
We wish to describe the distance

from a point P to the points R and R' on the
circle as a function of the angle α between
the line PR and the diameter (figure 9.3).
The vectors PR and PR' are:
PR = u  PF · u − r 2 − PF ∧ u 
2

 
PR' = u  PF · u + r 2 + PF ∧ u 
2
 Figure 9.3
 
We arrive at the polar equation by taking the modulus of PR and PR', which is the
distance to the circle:
PR = PF cos α − r 2 − PF 2 sin 2α
Figure 9.4
PR' = PF cos α + r 2 − PF 2 sin 2α
If P is a point on the circle then PF = r

(figure 9.4) and the equation is simplified:
PR = 2 r cos α
Inversion with respect to a circle
The point P' is the inverse point of P

with respect to a circle of centre F and radius r if the vector FP’ is the inverse vector of
FP with radius k:
FP' = k2 FP −1 ⇔ FP FP' = k2
Obviously, the product of the

vectors FP and FP' is only real and Figure 9.5
positive when the points P and P’ are
aligned with the centre F and are located
at the same side of F. The circle of centre
F and radius k is called the inversion
circle because its points remain invariant
under this inversion. The inversion
transforms points located inside the
inversion circle into outside points and
reciprocally.
To obtain geometrically the
inverse of an inside point P, draw firstly
(figure 9.5) the diameter passing through P; after this draw the perpendicular to this
diameter from P which will cut the circle in the point T; finally draw the tangent with
contact point T. The intersection P' of this tangent with the prolongation of the diameter is
the inverse point of P. Note that the right triangles FPT and FTP' are oppositely similar:
FP FT −1 = FP' −1 FT ⇒ FP' FP = FT2 = k2
To obtain the inverse of an outside point, make the same construction but in the
opposite sense. Draw the tangent to the circle passing through the point (P' in the figure
9.5) and then draw the perpendicular to the diameter FP' passing through the contact point
T of the tangent. The intersection P of the perpendicular with the diameter is the searched
inverse point of P'.
Every line not containing the centre of inversion is transformed into a circle
passing through the centre (figure 9.6) and reciprocally. To prove this statement let us take
the polar equation of a straight line:
d
FP = Figure 9.6
cos α
Then the inverse of P with centre F has the

equation:
k2 k 2 cos α
FP' = =
FP d
which is the polar equation of a circle passing

through F and with diameter k2 / d . α is the angle
between FP and FD, the direction perpendicular to
the line; therefore, the centre O of the circle is located on FD.
Let us prove that an inversion transforms a circle not passing through its centre
into another circle also not passing through its centre. Consider a circle of centre F and
radius r which will be transformed under an inversion of centre P and radius k. Then any
point R on the circle is mapped into another point S according to:
k2 k2
PS = =
PR PF cos α ± r 2 − PF 2 sin 2α
where the polar equation of the circle is used. The multiplication by the conjugate of the
denominator gives:
PS =
(
k 2 PF cos α m r 2 − PF 2 sin 2α )
PF − r
2 2
which is the polar equation of a circle with centre G and radius s:
PS = PG cos α m s 2 − PG 2 sin 2α
k 2 PF k 2 PF k2 r k2 r
where PG = = s= =
PF 2 − r 2 PT 2 PF 2 − r 2 PT 2
PF2 − r2 = PT2 is the power of the point P with respect to the circle of centre F.
Then we see that the distance from the centre of inversion to the centre of the circle and
the radius of this circle changes in the ratio of the square of the radius of inversion divided
by the power of the centre of inversion with respect to the given circle.
What is the power of the centre of inversion P with respect to the new circle?:
k 4 (PF 2 − r 2 ) k4
PU 2 = PG 2 − s 2 = 4
=
PT PT 2
That is, the product of the powers of the centre of inversion with respect to any circle and
its transformed is constant and equal to the fourth power of the radius of inversion:
PT2 PU2 = k4
In the equation of the transformed circle the symbol m appears instead of ± . It

means that the sense of the arc from R to R' is opposite to the sense of the arc from S to S'.
The nine-point circle
The Euler’s theorem states that the midpoints of the sides of any triangle, the feet
of the altitudes and the midpoints from each vertex to the orthocentre lie on a circle called
the nine-point circle.
Let any triangle PQR whose orthocentre is H be. Now we evaluate the expression:
(P − H + Q − R )2 − (P − H − Q + R )2 = 4(P − H ) · (Q − R ) = 0
which vanishes because the segment PH lying on the altitude passing through P is
perpendicular to the base QR. Analogously:
(P − H + Q − R )2 − (P + H − Q − R )2 = 4(Q − H ) · (P − R ) = 0
and
(P + H − Q − R )2 − (P − H − Q + R )2 = 4(R − H ) · (Q − P ) = 0
Hence: (P + Q − R − H )2 = (P − Q + R − H )2 = (P − Q − R + H )2
Since the sign may be opposite inside the square we have:
(P + Q − R − H )2 = (P − Q + R − H )2 = (P − Q − R + H )2 =
= (− P − Q + R + H ) = (− P + Q − R + H ) = (− P + Q + R − H )
2 2 2
Now we introduce the point N = (P + Q + R + H )/4 to obtain:

2 2 2 2 2 2
 R+H  Q+H  Q+ R  P+Q  P+ R  P+H
N −  = N −  = N −  = N −  = N −  = N − 
 2   2   2   2   2   2 
That is, N is the centre of a circle (figure 9.7) passing through the three midpoints of the
sides (P + Q)/2, (Q + R)/2 and (R + P)/2 and the three midpoints of the vertices and the
orthocentre (P + H)/2, (Q + H)/2 and (R + H)/2.
Since we can write:
1  P + H  1 Q + R
N=  +  
2 2  2 2 
it follows that (P+H)/2 and (Q+R)/2 are

opposite extremes of the same diameter. Figure 9.7
Then the vectors J−(P+H)/2 and
J−(Q+R)/2, which are orthogonal:
 P+H  Q + R
J − ·J − =0
 2   2 
are the sides of a right angle which

intercepts a half circumference of the nine
point circle. Therefore, this angle is
inscribed and its vertex J also lies on the
nine-point circle. Also, we may go
through the algebraic way to arrive at the
same conclusion. Developing the former inner product:
Q+R P+H P+H Q+R

J2 −J · − ·J + · =0
2 2 2 2
Introducing the centre N of the nine-point circle, we have:

2
Q + R Q + R
J2 −2J ·N +2N · −  =0
2  2 
and after adding and subtracting N2 we arrive at:

2
(J − N )2 −  Q + R − N  =0
 2 
which shows that the length of JN is the radius of the nine-point circle.
From the formulas for the centroid and orthocentre one obtains the centre N of the
nine-point circle:
3G + H
= (P QR P + Q RP Q + R PQ R )(4 PQ ∧ QR )
−1
N=
4
Figure 9.8
that is, the point N also lies on the Euler’s line. The
relative distances between the circumcentre O, the
centroid G, the nine-point circle centre N and the
orthocentre H are shown in the figure 9.8.
Cyclic and circumscribed quadrilaterals
First, let us see a statement valid for every quadrilateral: the area of a quadrilateral
ABCD is the outer product of the diagonals AC and BD. The prove is written in one line.
Since we can divide the quadrilateral in two triangles we have:
1 1 Figure 9.9
Area = AB ∧ BC + AC ∧ CD
2 2
1
= (AC ∧ BC + AC ∧ CD )
2
1
= AC ∧ BD
2
1
= ACBDsin(AC, BD) e12
2
The quadrilaterals inscribed in a circle are called cyclic quadrilaterals. They fulfil
the Ptolemy’s theorem: the product of the lengths of both diagonals is equal to the
addition of the products of the lengths of opposite sides (figure 9.9):
AC BD = AB CD + BC DA
A beautiful demonstration makes use of the cross ratio and it is proposed in the first
exercise of the next chapter.
For the cyclic quadrilaterals, the sum of opposite angles is equal to π, because they
are inscribed in the circle and intercept opposite arcs. Another interesting property is the
Brahmagupta formula for the area. If the lengths of the sides are denoted by a, b, c and d
and the semiperimeter by s then:
a+b+c+d
Area = (s − a ) (s − b) (s − c ) (s − d ) s=
2
With this notation, the proof is written more briefly. The law of cosines applied to each
triangle of quadrilateral gives:
 AC 2 = a 2 + b 2 − 2 a b cos λ

 AC = c + d − 2 c d cos( π − λ )
2 2 2
Equating both equations taking into account that the cosines of supplementary angles have
opposite sign, we have:
a 2 + b 2 − 2 a b cos λ = c 2 + d 2 + 2 c d cos λ
a 2 + b 2 − c 2 − d 2 = 2 ( a b + c d ) cos λ
a 2 + b2 − c 2 − d 2
= cos λ
2(ab+cd )
The area of the quadrilateral is the sum of the areas of the triangles ABC and CDA:
ab cd
Area = sin λ + sin( π − λ )
2 2
Both sines are equal; then we have:
Area =2 ( a b + c d )2
sin λ =
2 (a b + c d )2 
1−
(a 2
+ b 2 − c 2 − d 2 ) 
2
 4(ab+cd ) 
2
4 4  
4(ab+cd )2 − ( a 2 + b 2 − c 2 − d 2 )2
= = (s − a )(s − b )(s − c )(s − d )
16
Another type of quadrilaterals are those Figure 9.10

where a circle can be inscribed, also called
circumscribed quadrilaterals. For these
quadrilaterals the bisectors of all the angles
meet in the centre I of the inscribed circle,
because it is equidistant from all the sides.
When tracing the radii going to the tangency
points, the quadrilateral is divided into pairs of
opposite triangles (figure 9.10), and hence one
deduces that the sums of the lengths of
opposite sides are equal:
AB+CD= x + y + z + t = y + z + t + x =BC+DA
If a quadrilateral has inscribed and circumscribed circle, then we may substitute this
condition (a + c = b + d) in the Brahmagupta formula to find:
Area = a b c d = AB BC CD DA
Since the sum of the angles of any quadrilateral is 2π and either of the four small
quadrilaterals of a circumscribed quadrilateral has two right angles (figure 9.10), it
follows that the central angles are supplementary of the vertices:
CBA ≡ UBT = π − TIU ADC ≡ WDV = π − VIW
If the quadrilateral is also cyclic then the angles CBA and ADC are inscribed and intercept
opposite arcs of the outer circle, so that the segments joining opposite tangency points are
orthogonal:
CBA + ADC = π ⇒ TIU + VIW = π ⇒ TV⊥UW
Angle between circles
Let us calculate the angle between a circle centred at O with radius r and another
one centred at O' with radius r'. If O' − O< r + r' then they intersect. The points of
intersection P must fulfil the equations for both circles:
 (P − O )2 = r 2  P2 − 2 P ·O + O2 = r2
 ⇒  2
(P − O' ) = r'
2
 P − 2 P · O' + O' = r'
2 2 2
Adding both equations we obtain:
2 P 2 − 2 P · (O + O' ) + O 2 + O' 2 = r 2 + r' 2
It is trivial that the angle of intersection is the supplementary of the angle between both
radius, its cosines being equal:
OP · O'P P 2 − P · (O + O' ) + O · O'

cos α = =
OP OP' r r'
The substitution of the first equality into the second gives:
r 2 + r' 2 − (O − O' )
2
cos α =
2 r r'
The cosine of a right angle is zero; therefore two circles are orthogonal if and only if:
r 2 + r' 2 = (O − O' )
2
which is the Pythagorean theorem.
Radical axis of two circles
The radical axis of two circles is the locus of the points which have the same
power with respect to both circles:
OP 2 − r 2 = O'P 2 − r' 2
Let X be the intersection of the radical axis with the line joining both centres of the circles.
Then:
OX 2 − r 2 = O'X 2 − r' 2 ⇒ OX 2 − O'X 2 = r 2 − r' 2
(OX + O'X ) · (OX − O'X ) = r 2 − r' 2
 O + O' 
2X −  · OO' = r − r'
2 2
 2 
Since OX and O'X are proportional to OO', the inner and geometric products are
equivalent and we may isolate X:
O + O' + ( r 2 − r' 2 ) OO' −1

X=
2
Now let us search which kind of geometric figure is the radical axis. Arranging as before
the terms, we have:
 O + O' 
2 P −  · OO' = r − r'
2 2
 2 
Introducing the point X we arrive at:
2 (P − X ) · OO' = 0
That is, the radical axis is a line passing through X and perpendicular to the line joining
both centres. If the circles intersect, then the radical axis is the straight line passing
through the intersection points.
The radical centre of three circles is the intersection of their radical axis. Let O, O'
and O'' be the centre of three circles with radii r, r' and r''. Let us denote with X (as above)
the intersections of the radical axis of the first and second circles with the line OO', and
with Y the intersection of the radical axis of the second and third circles with the line
joining its centres O'O''. The radical centre P must fulfil the former equation for both
radical axis:
 O' 2 − O 2 + r 2 − r' 2
 (P − X ) · OO' = 0  P · OO' = X · OO' = 2
 ⇒ 
(P − Y ) · O'O'' = 0 − + r' 2 − r'' 2
2 2
 P · O'O'' = Y · O'O'' = O'' O'
 2
Now we have encountered the coefficients of linear combination of an imaginary

decomposition. This process is explained in page 74 for the calculation of the orthocentre.
However there is a difference: we wish not to resolve the vector into orthogonal
components but from the known coefficients we wish to reconstruct the vector P. Then:
P · O'O'' P · OO'
P= OO' − O'O''
OO' ∧ O'O'' OO' ∧ O'O''
The substitution of the known values of the coefficients results in the following formula
for the radical centre which is invariant under cyclic permutation of the circles:
P = (2 OO' ∧ O'O'' )
−1
((O'' 2
− r'' 2 ) OO' + (O 2 − r 2 ) O'O'' + (O' 2 − r' 2 ) O''O )
Exercises
9.1 Let A, B, C be the vertices of any triangle. The point M moves according to the
equation:
MA2 + MB2 + MC2 = k
Which geometric locus does the point M describe as a function of the values of the real
parameter k?
9.2 Prove that the geometric locus of the points P such that the ratio of distances from P to
two distinct points A and B is a constant k is a circle. Calculate the centre and the radius of
this circle.
9.3 Prove the Simson’s theorem: the feet of the perpendiculars from a point D upon the
sides of a triangle ABC are aligned if and only if D lies on the circumscribed circle.
9.4 The Bretschneider's theorem: let a, b, c and d be the successive sides of a

quadrilateral, m and n its diagonals and α and γ two opposite angles. Show that the
following law of cosines for a quadrilateral is fulfilled:
m 2 n 2 = a 2 c 2 + b 2 d 2 − 2 a b c d cos(α + γ )
9.5 Draw three circles passing through each vertex of a triangle and the midpoints of the
concurrent sides. Then join the centre of each circle and the midpoint of the opposite side.
Show that the three segments obtained in this way intersect in a unique point.
9.6 Prove that the inversion is an opposite conformal transformation, that is, it preserves
the angles between curves, but changes their sign.
10. CROSS RATIOS AND RELATED TRANSFORMATIONS
Complex cross ratio
The complex cross ratio of any four points A, B, C and D on the plane is defined
as the quotient of the following two single ratios:
( A B C D ) = ( A C D ) ( B C D ) −1 = AC AD −1 BD BC −1
If the four points are aligned, then the cross ratio is a real number, otherwise it is a
complex number. Denoting the angles CAD and CBD as α and β respectively, the cross
ratio is written as:
AC AD BD BC
( ABC D)= 2 2 Figure 10.1
AD BC
AC BD
= exp[(α − β ) e12 ]
AD BC
If the four points are located in this order

on a circle (figure 10.1), the cross ratio is a
positive real number, because the inscribed angles
α and β intercept the same arc CD and hence are
equal, so:
AC BD
( ABC D)=
AD BC
But if one of the points C or D is located between A and B (figure 10.2), then α and β
have distinct arcs CD with opposite orientation. Since α and β are the half of these arcs, it
follows that α −β = π and the cross ratio is a
negative real number: Figure 10.2
AC BD
( ABC D)= −
AD BC
Analogously, if the four points are aligned

and C or D is located between A and B, the cross
ratio is negative1.
Other definitions of the cross ratio differ
from this one given here only in the order in
which the vectors are taken. This question is
equivalent to study how the value of the cross ratio changes under permutations of the
points. For four points, there are 24 possible permutations, but only six different values for
the cross ratio. The calculations are somewhat laborious and I shall only show one case.
1
A straight line is a circle with infinite radius.
The reader may prove any other case as an exercise. I also indicate whether the
permutations are even or odd, that is, whether they are obtained by an even or an odd
number of exchanges of any two points from the initial definition ( A B C D ):
(1) (ABCD)=(BADC)=(CDAB)=(DCBA)=r even

1
(2) (BACD)=(ABDC)=(DCAB)=(CDBA)= odd
r
(3) ( A C B D ) = ( B D A C ) = ( C A D B ) = ( D B C A ) = 1 − r odd
1
(4) (CABD)=(DBAC)=(ACDB)=(BDCA)= even
1− r
r −1
(5) (BCAD)=(ADBC)=(DACB)=(CBDA)= even
r
r
(6) (CBAD)=(DABC)=(ADCB)=(BCDA)= odd
r −1
Let us prove, for example the equalities (3):
(3) ( A C B D ) = AB AD −1 CD CB −1 = ( AC + CB ) AD −1 ( CB + BD ) CB −1 =
= AC AD −1 CB CB −1 + AC AD −1 BD CB −1 + CB AD −1 CB CB −1 + CB AD −1 BD CB −1
In the last term, under the reflection in the direction CB, the complex number AD −1 BD
becomes conjugate, that is, these vectors are exchanged:
= AC AD −1 + AC AD −1 BD CB −1 + CB AD −1 + BD AD −1
= ( AC + CB + BD ) AD −1 − AC AD −1 BD BC −1 = AD AD −1 − r = 1 − r
Let us see that this value is the same for all the cross ratios of the case (3):
( B D A C ) = BA BC −1 DC DA −1 = AB CB −1 CD AD −1 = AB AD −1 CD CB −1
=(ACBD)
where in the third step the permutative property has been applied. Also by using this
property one obtains:
( C A D B ) = CD CB −1 AB AD −1 = CD AD −1 AB CB −1 = AB AD −1 CD CB −1
=(ACBD)
( D B C A ) = DC DA −1 BA BC −1 = BA DA −1 DC BC −1 = AB AD −1 CD CB −1
=(ACBD)
Harmonic characteristic and ranges
From the cross ratio, one can find a quantity independent of the symbols of the
points and only dependent on their locations. I call this quantity the harmonic
characteristic -because of the obvious reasons that follow now- denoting it as [A B C D].
The simplest way to calculate it (but not the unique) is through the alternated addition of
all the permutations of the cross ratio, condensed in the alternated sum of the six different
values:
(ABCD)−(BACD)−(ACBD)+(CABD)+(BCAD)−(CBAD)=
1 1 r −1 r
=r− − (1 − r ) + + −
r 1− r r r −1
(2 r − 1 ) ( r + 1 ) ( r − 2 )
=
r (1 − r)
When two points are permuted, this sum still changes the sign. The square of this
sum is invariant under every permutation of the points, and is the suitable definition of the
harmonic characteristic:
[ A B C D ] = [( A B C D ) − ( B A C D ) − ( A C B D ) + (C A B D ) + ( B C A D ) − ( A B C D )]
2
(2 r − 1 )2 ( r + 1 )2 ( r − 2 )2
[ A B C D] =
r 2 (1 − r)
2
The first notable property of the harmonic characteristic is the fact that it vanishes for a
harmonic range of points, that is, when the cross ratio (A B C D) takes the values 1/2, −1
or –2 dependent on the denominations of the points. The second notable property is the
fact that the harmonic characteristic of a degenerate range of points (two or more
coincident points) is infinite.
Perhaps, the reader believes that other harmonic characteristics may be obtained in
many other ways, that is, by using other combinations of the permutations of the cross
ratio. However, other attempts lead whether to the same function of r (or a linear
dependence) or to constant values. For example, we can take the sum of all the squares,
which is also invariant under any permutation of the points to find:
(A B C D) 2 + (B A C D)2 + (A C B D)2 + (C A B D)2 + (B C A D)2 + (C B A D)2
= r2 +
1
+ (1 − r ) 2
+
1
+
(r − 1)2 + r 2 = [ A B C D ] + 21
r2 (1 − r )2 r2 (r − 1)2 2
On the other hand, M. Berger (vol. I, p. 127) has used a function that is also
linearly dependent on the harmonic characteristic:
4(r2 − r +1 )
3
[ A B C D]
=1+
27 r ( 1 − r )
2
2
27
Other useless possibilities are the sum of all the values of the permutations of the cross
ratio, which is identical to zero, or their product giving always a unity result.
Four points A, B, C and D so ordered on a line are said to be a harmonic range if
the segments AB, AC and AD are in harmonic progression, that is, the inverse vectors
AB−1, AC −1 and AD −1 are in arithmetic progression:
AB −1 − AC −1 = AC −1 − AD −1 ⇔ AC AD −1 BD BC −1 = 2
The cross ratio of a harmonic range is 2, but if this order is altered then it can be –1 or 1/2.
Note that this definition also includes harmonic ranges of points on a circle. On the other
hand, all the points of a harmonic range always lie on a line or a circle.
In order to construct the fourth point D forming a harmonic range with any three
given points A, B and C (ordered in this way on a line) one must follow these steps (figure
10.3):
1) Draw the line passing through A, B and C. Trace two any distinct lines R and S also
different from ABC and passing through A.
2) Draw any line passing through B. This
line will cut the line R in the point P and the Figure 10.3
line S in the point X.
3) Draw the line passing through C and X.
The intersection of this line with R will be
denoted by Q. Draw also the line passing
through C and P. This line intersects the
line S in the point Y.
4) Now trace the line passing through Y and
Q, which will cut the line ABC in the
searched point D, which completes the
harmonic range.
For a clearer displaying I have
drawn some segments instead of lines, although the construction is completely general
only for lines.
If A, B and C are not aligned, the point D which completes the harmonic range lies
on a circle passing through the three given points (figure 10.4). In order to determine D
one must make an inversion (which preserves the cross ratio if this is real, as shown
below) with centre on the circle ABC, which transforms this circle into a line. In this line
we determine the point D’ forming a harmonic range with A’, B’, C’ and then, by means
of the same inversion, the point D is obtained. In the drawing method the radius and the
exact position of the centre of inversion do not play any role whenever it lie on the circle
ABC. Then we shall draw any line outside the circle, project A, B and C into this line,
make the former construction and obtain D by projecting D’ upon the circle.
The cross ratio becomes conjugated under circular inversion. In order to prove this
statement, let us see firstly how the single ratio is transformed. If A', B', C' and D' are the
transformed points of A, B, C and D under an inversion with centre O and any radius, we
have:
A'C' A'D' −1 = (OC' − OA' ) (OD' − OA' ) −1 = (OC −1 − OA −1) (OD −1 − OA −1) −1
= OC −1 ( OA − OC ) OA −1 [ OD −1 ( OA − OD ) OA −1 ] −1
= OC −1 CA OA −1 [ OD −1 DA OA −1 ] −1 = OC −1 CA OA −1 OA DA −1 OD
= OC −1 AC AD −1 OD
Analogously we have: Figure 10.4
B'D' B'C' −1 = OD −1 BD BC −1 OC
The product of both expressions is the cross

ratio:
( A' B' C' D' ) = A'C' A'D' −1 B'D' B'C' −1
= OC −1 AC AD −1 BD BC −1 OC
= OC −1 ( A B C D ) OC = ( A B C D )*
So, the cross ratio of four points on a line or a circle, which is real, is preserved, otherwise
it becomes conjugated.
Finally let us see that the single ratio is the limit of the cross ratio for one point at
infinity:
(∞ B C D) = lim AC AD −1 BD BC −1 = BD BC −1 = ( B D C )
A→ ∞
Analogously ( A ∞ C D ) = ( A C D ), etc.
The homography (Möbius transformation)
The homography is defined as the geometric transformation that preserves the

complex cross ratio of four points on the plane. In order to specify a homography one
must gives only the images A', B' and C' of three points A, B, and C because the fourth
image D' is determined by the conservation of their cross ratio:
( A' B' C' D' ) = ( A B C D )
A'C' A'D' −1 B'D' B'C' −1 = AC AD −1 BD BC −1
Arranging vectors we write:
A'D' −1 B'D' = A'C' −1 AC AD −1 BD BC −1 B'C' = A'C' −1 AC BC −1 B'C' AD −1 BD
After conjugation we find:

B'D' A'D' −1 = BD AD −1 B'C' BC −1 AC A'C' −1
B'D' A'D' −1 = BD AD −1 z where z = BC −1 B'C' A'C' −1 AC
D'B' D'A' −1 = DB DA −1 z ⇒ ( D' B' A' ) = ( D B A ) z
Hence the homography multiplies the single ratio by a complex number z. The modulus of
the single ratio increases proportionally to the modulus of z and the angle BDA is
increased in the argument of z:
z = z exp[α e12 ]
D'B' DB
= z
D'A' DA
angle B'D'A' = angle BDA + α
Then a homography can also be determined by specifying the images A' and B' of two
points A and B, and a complex number z. The image D' of D is calculated by isolation in
the former equation:
( D'A' + A'B' ) D'A' −1 = DB DA −1 z
A'B' D'A' −1 = DB DA −1 z − 1
D'A' = ( z DB DA −1 − 1 ) −1 A'B'
D' = A' + ( 1 − z DB DA −1 ) −1 A'B'
The homography preserves the complex cross ratio of any four points.
Analogously to the former equality, we find for any three points P, Q, R :
P' = A' + ( 1 − z PB PA −1 ) −1 A'B'
Q' = A' + ( 1 − z QB QA −1 ) −1 A'B'
R' = A' + ( 1 − z RB RA −1 ) −1 A'B'
D'Q' = [ ( 1 − z QB QA −1 ) −1 − ( 1 − z DB DA −1 ) −1 ] A'B' =
= z ( −DB DA −1 + QB QA −1 ) ( 1 − z DB DA −1 ) −1 ( 1 − z QB QA −1 ) −1 A'B'
D'R' = z ( −DB DA −1 + RB RA −1 ) ( 1 − z DB DA −1 ) −1 ( 1 − z RB RA −1 ) −1 A'B'
P'Q' = z ( −PB PA −1 + QB QA −1 ) ( 1 − z PB PA −1 ) −1 ( 1 − z QB QA −1 ) −1 A'B'

P'R' = z ( −PB PA −1 + RB RA −1 ) ( 1 − z PB PA −1 ) −1 ( 1 − z RB RA −1 ) −1 A'B'
In the cross ratio, many factors are simplified:
D'Q' D'R' −1 P'R' P'Q' −1 = ( −DB DA −1 + QB QA −1 ) ( −DB DA −1 + RB RA −1 ) −1
( −PB PA −1 + RB RA −1 ) ( −PB PA −1 + QB QA −1 ) −1
Writing the first factor as product of factors we obtain:

−1 −1 −1 −1
−DB DA + QB QA = − ( DA + AB ) DA + ( QA + AB ) QA =
= −AB DA −1 + AB QA −1 = AB DA −2 ( −DA QA2 + DA2 QA ) QA −2 =
= AB DA −2 DA ( −QA + DA ) QA QA −2
= AB DA −1 QD QA −1
In the same way each of the other three factors can be written as product of factors:
−1 −1 −1 −1
−DB DA + RB RA = AB DA RD RA
−1 −1 −1 −1
−PB PA + RB RA = AB PA RP RA
−1 −1 −1 −1
−PB PA + QB QA = AB PA QP QA
Then the cross ratio will be:
D'Q' D'R' −1 P'R' P'Q' −1 = ( AB DA −1 QD QA −1 ) ( RA RA −1 DA AB −1 )
( AB PA −1 RP RA −1 ) ( QA QP −1 PA AB −1 )
where the brackets are not needed but only show the factors they come from. Applying the
permutative property and simplifying, we arrive at:
D'Q' D'R' −1 P'R' P'Q' −1 = DQ DR −1 PR PQ −1
This proves that the cross ratio of every set of four point is preserved under a
homography. Four points lying on a line or a circle have a real cross ratio. Then the
homography is a circular transformation since it transforms circles into circles in a general
sense (taking the lines as circles with infinite radius).
The product of two circular inversions is always a homography because the cross
ratio becomes conjugate under inversion and thus it is preserved under an even number of
inversions. Let us calculate the homography resulting from a composition of two
inversions. Let A be any point, A' be the transformed point under a first inversion with
centre O and radius r, and A" be the transformed point of A' under a second inversion with
centre P and radius s, which is consequently the transformed point of A under the
homography:
OA' = r2 OA −1 PA'' = s2 PA' −1
From where it follows that:
s2 PA'' −1 = PA' = OA' − OP = r2 OA −1 − OP
PA'' −1 = s −2 ( r2 OA −1 − OP )
PA'' = s2 (r2 OA −1 − OP ) −1 = s2 [( r2 −OP OA ) OA −1] −1 = s2 OA (r2 −OP OA ) −1
This equation allows to determine the image A" of any point A having the centre O and P
and the radius r and s as data. For instance, let us calculate the homography corresponding
to:
O = ( 0, 0 ) r=2 P = ( 3, 0 ) s=3
OP = 3 e1
In order to specify the homography we determine the images of three points:
A = ( 0, 2 ) B = ( 1, 1 ) C = ( 1, 0 )
OA = 2 e2 OB = e1 + e2 OC = e1
OP OA = 6 e12 OP OB = 3 + 3 e12 OP OC = 3
− 27 e1 + 18 e 2 − 9 e1 + 18 e 2
PA'' = PB'' = PC'' = 9 e1
13 5
 12 18   6 18 
A'' =  ,  B'' =  ,  C'' = (12, 0)
 13 13  5 5 
Projective cross ratio
The projective cross ratio of four Figure 10.5

points A, B, C and D with respect to a
point X is defined as the cross ratio of
the pencil of lines XA, XB, XC and XD,
which is the cross ratio of the
intersection points of these lines with
any crossing line (figure 10.5). As it will
be shown, the projective cross ratio is
independent of the crossing line The
intersection points A’, B’, C’ and D’ are
the projections with centre X of the points A, B, C and D upon the crossing line. Due to the
alignment of these projections, the projective cross ratio is always a real number:
{ X, A B C D } = ( A' B' C' D' ) = A'C' A'D' −1 B'D' B'C' −1
The projected points A', B', C' and D' are dependent on the crossing line and this
expression is not suitable for calculations. Note that the segment A'C' is the base of the
triangle XA'C', B'D' is the base of the triangle XB'D', etc. Since the altitudes of all these
triangles are the distance from X to the crossing line, their bases are proportional to their
areas, so we can write:
XA' ∧ A'C' XB' ∧ B'D '

{ X , A B C D} =
XA' ∧ A'D' XB' ∧ B'C'
The area of a triangle is the half of the outer product of any two sides. For example
for the triangle XA'C' we have:
XA' ∧ A'C' = XA' ∧ ( A'X + XC' ) = XA' ∧ XC'
Then, the cross ratio becomes:
XA' ∧ XC' XB' ∧ XD'

{ X , A B C D} =
XA' ∧ XD' XB' ∧ XC'
The simplification of the modulus gives:
sin A'XC' sin B'XD' sin AXC sin BXD

{ X , A B C D} = =
sin A'XD' sin B'XC' sin AXD sin BXC
because the angles A'XC' and AXC are identical, B'XD' and BXD are also identical, etc.
Now it is obvious that the projective cross ratio does not depend on the crossing line, but
only on the angles between the lines of the pencil. If we introduce the modulus of XA, XB,
etc., we find the outer products of these vectors:
XA ∧ XC XB ∧ XD
{ X , A B C D} =
XA ∧ XD XB ∧ XC
Now, the arbitrary crossing line of the pencil has disappeared in the formula, which has
become suitable for calculations with Cartesian coordinates2. When the points A, B, C and
D are aligned, the complex and projective cross ratios have the same value independently
of the point X, because the segments AC, BD, AD and BC have the same direction, which
can be taken as the crossing line:
2
The reader will find an application of this formula for the projective cross ratio to the error
calculus in the electronic article "Estimation of the error committed in determining the place from
where a photograph was taken", Ramon González, Josep Homs, Jordi Solsona (1999)
http://www.terra.es/personal/rgonzal1 (translation of the paper published in the Butlletí de la
Societat Catalana de Matemàtiques 12 [1997] 51-71).
{X, A B C D} = ( A B C D )
If A, B, C and D lie on a circle, the complex cross ratio is equal to the projective
cross ratio taken from any point X on that circle. To prove this statement, make an
inversion with centre X. Since the circle ABCDX passes through the centre of inversion, it
is transformed into a line, where the points A', B', C' and D' (transformed of A, B, C and D
by the inversion) lie with the same cross ratio. Now observe that X, A and A’ are aligned,
also X, B and B’, etc. So the projective cross ratio is equal to the complex cross ratio.
From the arguments given above, it follows that the cross ratio of a pencil of lines
is the quotient of the outer products of their direction (or normal) vectors independently of
their modulus:
v ∧ v XC v XB ∧ v XD n XA ∧ n XC n XB ∧ n XD
{ X , A B C D} = XA =
v XA ∧ v XD v XB ∧ v XC n XA ∧ n XD n XB ∧ n XC
Instead of XA, XB, XC and XD, I shall denote by A, B, C and D the pencil of lines
passing through X. Then the projective cross ratio {X, ABCD} is written as (ABCD). Let
us suppose that these lines have the following dual coordinates expressed in the lines base
{O, P, Q}:
A = [xA , yA] B = [xB , yB] C = [xC , yC] D =[xD , yD]
Then the direction vector of the line A is:
v A = (1 − x A − y A )v O + x A v P + y A v Q
But the direction vectors of the base fulfil:
vO + v P + vQ = 0
Hence:
v A = (− 1 + 2 x A + y A ) v P + (− 1 + x A + 2 y A ) v Q
and analogously:
v B = (− 1 + 2 x B + y B ) v P + (− 1 + x B + 2 y B ) v Q
v C = (− 1 + 2 x C + y C ) v P + (− 1 + x C + 2 y C ) v Q
v D = (− 1 + 2 x D + y D ) v P + (− 1 + x D + 2 y D ) v Q
The outer product of vA and vC is:

v A ∧ v C = ( x C − x A − ( y C − y A ) + 3 ( x A y C − x C y A )) v P ∧ v Q
which leads to the fact that the cross ratio only depends on the dual coordinates:
( A B C D) = =
( x C − x A − ( y C − y A ) + 3 ( x A y C − x C y A )) ( x D − x B − ( y D − y B ) + 3 ( x B y D − x D y B ))
=
( x D − x A − ( y D − y A ) + 3 ( x A y D − x D y A )) ( x C − x B − ( y C − y B ) + 3 (x B y C − x C y B ))
(x A − 1 / 3, y A − 1 / 3) ∧ (x C − 1 / 3, y C − 1 / 3) ( x B − 1 / 3, y B − 1 / 3) ∧ ( x D − 1 / 3, y D − 1 / 3)
=
(x B − 1 / 3, y B − 1 / 3) ∧ (x C − 1 / 3, y C − 1 / 3) ( x A − 1 / 3, y A − 1 / 3) ∧ ( x D − 1 / 3, y D − 1 / 3)
(x A − x C , y A − y C ) ∧ (x C − 1 / 3, y C − 1 / 3) ( x B − x D , y B − y D ) ∧ ( x D − 1 / 3, y D − 1 / 3)
=
(x B − x C , y B − y C ) ∧ (x C − 1 / 3, y C − 1 / 3) ( x A − x D , y A − y D ) ∧ ( x D − 1 / 3, y D − 1 / 3)
And taking into account the fact that the dual points corresponding to the lines of the
pencil are aligned in the dual plane, it follows that:
( A B C D ) = (x A − x C , y A − y C ) (x B − x C , y B − y C ) (x B − x D , y B − y D ) (x A − x D , y A − y D )−1
−1
= AC BC −1 BD AD −1 = AC AD −1 BD BC −1
where the four dual vectors are proportional. That is: the projective cross ratio of a pencil
of lines is the cross ratio of the corresponding dual points. Obviously, by means of the
duality principle the dual proposition must also hold: the cross ratio of four aligned points
is equal to the projective cross ratio of the dual pencil of the corresponding lines on the
dual plane.
The points at the infinity and homogeneous coordinates
The points at the infinity play the same role as the points with finite coordinates in
the projective geometry. Under a projectivity they can be mapped mutually without
special distinction. In order to avoid the algebraic inconsistencies of the infinity, the
homogeneous coordinates are defined as proportional to the barycentric coordinates with
a no fixed homogeneous constant. The following expressions are equivalent and represent
the same point:
(x, y) = (1 − x − y, x, y) = k (1 − x − y, x, y)
Then for instance:
(2, 3) = (−4, 2, 3 ) = (−20, 10, 15)

A point has finite coordinates if and only if the sum of the homogeneous
coordinates are different from zero. In this case we can normalise the coordinates
dividing by this sum in order to obtain the barycentric coordinates, and hence the
Cartesian coordinates by omitting the first coordinate. On the other hand, the sum of the
homogeneous coordinates sometimes vanishes and then they cannot be normalised (the
normalisation implies a division by zero yielding infinite values). In this case, we are
regarding a point at the infinity, whose algebraic operations should be performed with
the finite values of the homogeneous coordinates3. For example:
(2, −1, −1) = (2 ∞, −∞, −∞) = (∞, ∞)
This is the point at the infinity in the direction of the bisector of the first quadrant. On the
other hand, points at the infinity are equivalent to directions and hence to vectors. For
example:
v = 2 O − P − Q = −OP−OQ = −2 e1 −2 e2
Every vector proportional to v also represents this direction and the point at the infinity
given above.
Perspectivity and projectivity
A perspectivity is a central projection of a range of points on a line upon another

line. Since the cross ratio of a pencil of lines is independent of the crossing line, it follows
from the construction (figure 10.6) that a
perspectivity preserves the projective cross
ratio:
{O, A B C D} = {O, A' B' C' D'}
The composition of two perspectivities with

different centres also maps ranges4 of points
to other ranges preserving their cross ratio.
This is the most general transformation
which preserves alignment and the projective Figure 10.6
cross ratio. A projectivity (also projective
transformation) is defined as the geometric transformation which maps aligned points to
aligned points preserving their cross ratio. A theorem of projective geometry states that
any projectivity is expressible as the sequence of not more than three perspectivities. If the
projectivity relates ranges on two distinct lines, two perspectivities suffice.
3
In fact, the calculations are equally feasible with ∞ if we understand it strictly as a number.
Then the equality 2∞ = ∞ should be wrong. We could simplify by ∞, e.g.: (2 ∞, −∞, 1) = (2, −1,
0) is a finite point and (2 ∞, −∞, −∞) = (2, −1, −1) a point at the infinity.
4
A range of points is defined as a set of aligned points.
A projectivity is determined by giving three aligned points and their corresponding

images, which must be also aligned. On the other hand, a projectivity is determined by
giving four points no three of which are collinear (a quadrilateral) and their non-aligned
images (another quadrilateral). Since the projectivity preserves the alignment and the
incidence, we must take a linear transformation of the coordinates. However, the affinity
is already a linear transformation of coordinates but only of the independent coordinates.
The more general linear transformation must account for all the three barycentric
coordinates. Thus the projectivity is given by a 3×3 matrix:
1 − x' − y'   h00 h01 h02  1 − x − y

    
 x'  = k  h10 h11 h12   x  with det hij ≠ 0
 y'  h h21  
h22   y 
   20 
The coefficients hij being given, the product of the matrices on the right hand side
does not warrantee a set of coordinates with sum equal to 1. This means that the matrix hij
of a perspectivity is defined except by the proportionality constant k (which depends on
the coordinates), and the transformed point is obtained with homogeneous coordinates
instead of barycentric ones. Nevertheless, this matrix maps collinear points to collinear
points. This follows immediately from the matrices equalities:
 k (1 − x A' − y A' ) l (1 − x B' − y B' ) m (1 − x C' − y C' )  h00 h01 h02 

   
 k x A' l x B' m x C'  =  h10 h11 h12  ⋅
 k y A' l y B' m y C'  h h21 h22 
   20
1 − x A − y A 1 − x B − y B 1 − xC − yC 
 
 xA xB xC 
 yA yB yC 
 
A set of points are collinear if the determinant of the barycentric coordinates vanishes.
Since det hij ≠ 0, the implication of the alignment is bi-directional:
1 − x A' − y A' 1 − x B' − y B' 1 − x C' − y C' 1 − x A − y A 1 − x B − y B 1 − xC − yC

klm x A' x B' x C' = xA xB xC det hij
y A' y B' y C' yA yB yC
As an example, let us transform the origin of coordinates under the following

projectivity:
1 − 1 2 1 1
    
 2 3 1 0 =  2 ⇒ (0, 0) → (1, 2, 1) =  1 , 2 , 1  =  1 , 1 
 1 2 0 0 1 4 4 4 2 4
    
Analogously the transformed point of (1, 0) is:

 1 − 1 2   0   − 1
    
 2 3 1 1 =  3  ⇒ (1, 0) → (− 1, 3, 2) =  3 , 1 
 1 2 0 0  2  4 2
    
Also we may transform the point (0, ∞) at the infinity, which has the homogeneous
coordinates (−1, 0, 1):
 1 − 1 2   − 1  1 
    
 2 3 1   0  =  − 1 ⇒ (0, ∞ ) → (1, − 1, − 1) = (1, 1)
 1 2 0   1   − 1
    
This case exemplifies how points at the infinity can be mapped under a projectivity to
points with finite coordinates and vice versa. As a comparison, the affinity always maps
points at the infinity to points at the infinity (although it does not leave them invariant).
That is, the affinity always maps parallel lines to parallel lines, while under a projectivity
parallel lines can be mapped to concurrent lines and vice versa5. Parallel lines are
concurrent in a point at the infinity, which does not play any special role in the
projectivity.
Let us see that four non collinear points (a quadrilateral) and their non collinear
images (another quadrilateral) determine a unique projectivity. Let us suppose that the
images of the points (0, 0), (0, 1), (1, 0) and (1, 1) are (a1, a2), (b1, b2), (c1, c2) and (d1, d2)
where no three of which are collinear. The barycentric coordinates of these points are:
(0, 0) = (1, 0, 0) (1, 0) = (0, 1, 0) (0, 1) = (0, 0, 1)
(a1, a2) = (a0, a1, a2) with a0 = 1− a1 − a2 etc.
The substitution into the matrix equality produces a system of three equations with three
unknowns k, l and m:
 d 0   k a0 l b0 m c 0   − 1
    
 d 1  =  k a1 l b1 m c1   1 
d  k a l b2 m c 2   1 
 2  2
Since the points (0, 0), (1, 0) and (0, 1) are not collinear, the matrix is regular and the
system has a unique solution, which proves that a projectivity is determined by the image
of the base quadrilateral. If the vertices of the initial quadrilateral are not the base points,
the matrix is calculated as the product of the inverse matrix of the projectivity which maps
the base quadrilateral to the initial quadrilateral multiplied by the matrix of the projectivity
which maps the point base to the final quadrilateral.
Let us calculate the projectivity which maps the quadrilateral with vertices (1, 1)-
(2, 3)-(3, 1)-(1, 0) to the quadrilateral (4, 1)-(6, 0)-(4, 0)-(6, 1). The projectivity which
5
Of course, the affinity is a special kind of projectivity.
maps the base quadrilateral (0, 0)-(1, 0)-(0, 1)-(1/3, 1/3) to the first quadrilateral is given
by:
0  − k −4l − 3 m  1

    
1 =  k 2l 3 m  1
0  k 3l m  1
  
5 1 1
whose solution is: k= l=− m=
4 2 4
Taking the values k = 5, l = −2 m =1, we find the matrix:
 − 5 8 − 3  7 5 6 
  1  
 5 − 4 3  whose inverse matrix is  5 5 0 
 5 −6 1  20  
   − 5 5 − 10 
The projectivity which maps the base quadrilateral to the final quadrilateral is:
 − 6   − 4k −5l − 3 m   1
    
 6  =  4k 6l 4 m   1
 1   k 0 0  1
  
whose solution is: k = 1, l =1, m = −1 and the square matrix:
− 4 − 5 3 
 
 4 6 − 4
 1 0 0 

Then, the projectivity which maps the initial quadrilateral to the final one is the product:
 − 4 − 5 3   7 5 6   − 68 − 30 − 54 
    
 4 6 − 4   5 5 0  =  78 30 64 
 1 0 0   − 5 5 − 10   7 5 6 

For example, the point (2, 5/3) is transformed into:
 − 68 − 30 − 54   − 8   94 
      62 2
 78 30 64   6  =  − 124  =  , − 
 7  13 13 
 5 6   5   4 
A projectivity may be also defined in an alternative way: let ABCD be a

quadrilateral and A'B'C'D' another quadrilateral. Then the mapping:
D=aA+bB+cC → D' = a' A' + b' B' + c' C'
is a projectivity if the coefficients a', b', c' are homogeneous linear functions of a, b, c. For
example in the foregoing example the quadrilateral (1, 1)-(2, 3)-(3, 1)-(1, 0) is mapped to
the quadrilateral (4, 1)-(6, 0)-(4, 0)-(6, 1). Then we have:
1 1 5
(1, 0) = (1 − b − c ) (1, 1) + b ( 2, 3) + c (3, 1) ⇒ b=− c= a =1− b − c =
2 4 4
(6, 1) = (1 − b' − c' ) ( 4, 1) + b' (6, 0) + c' ( 4, 0) ⇒ b' = 1 c' = −1 a' = 1 − b' − c' = 1
Taking into account that for a = 1, b = 0, c = 0 → a' =1, b' = 0, c' = 0 and analogously for
each of the first three points, the linear mapping of coefficients is a diagonal matrix which
can be written simply:
4
a' = a b' = −2 b c' = − 4 c
5
If we wish to calculate the transformed point of (2, 5/3) we calculate firstly a, b and c:
 5 1 1 1
P =  2,  = (1 − b − c ) (1, 1) + b (2, 3) + c (3, 1) ⇒ b= c= a=
 3 3 3 3
4 2 4
⇒ a' = b' = − c' = −
15 3 3
Normalising the coefficients, we find the transformed point P':
P' = −
4
(4, 1) + 10 (6, 0) + 20 (4, 0) =  62 , − 2 
26 26 26  13 13 
This example shows the normalisation. In general, we say that the following
transformation is a projectivity:
D = (1− x − y) O + x P + y Q → D' = (1 − x' − y' ) O' + x' P' + y' Q'
lx my
x' = y' =
k (1 − x − y ) + l x + m y k (1 − x − y ) + l x + m y
Let us prove that a projectivity defined in this algebraic way leaves invariant the
projective cross ratio (as evident by construction from perspectivities). Under a
projectivity, four collinear points A, B, C and D are transformed into another four collinear
points A', B', C' and D' . Their cross ratio is the quotient of differences of any coordinate
(it is not needed that x and y are Cartesian coordinates) because these differences for x are
proportional to the differences for y due to the alignment:
l [k ( x C − x A ) + (k − m ) ( x A y C − x C y A )]
x' C − x' A =
[k (1 − x C − y C ) + l x C + m y C ] [k (1 − x A − y A ) + l x A + m y A ]
Let us suppose that the collinear points A, B, C and D lie on the line having the equation y
= s x + t, where the constants s and t play the role of a “slope” and “ordinate intercept”
although the x-axis and the y-axis be not orthogonal:
l ( x C − x A ) [k − (k − m ) t ]
x' C − x' A =
[k (1 − x C − y C ) + l x C + m y C ] [k (1 − x A − y A ) + l x A + m y A ]
Now we see that a projectivity changes every single ratio due to the denominators, but the
cross ratio is preserved because these denominators are cancelled in this case:
(x C' − x A' ) (x D' − x B' ) (x C − x A ) (x D − x B )

( A' B' C' D' ) = = = (A B C D)
(x D' − x A' ) (x C' − x B' ) (x D − x A ) (x C − x B )
The projectivity as a tool for theorems demonstration
One of the most interesting applications of the projectivity is the demonstration of

theorems of projective nature owing to the fact that the projectivity preserves alignment
and incidence. For example, in the figure 10.3 I have displayed without proof the method
of construction of the fourth point which
completes a harmonic range from the other
three. Now let us apply to this figure a
projectivity which maps the point A to the
infinity. In this case the lines ABCD, R and S
become parallel (figure 10.7) and the cross ratio
becomes a single ratio:
B'D'
( A B C D ) = ( ∞ B' C' D' ) = =2
B'C'
Figure 10.7
By similarity of triangles one sees that B'D' =2
B'C', so that the cross ratio is 2 and
proves that the points A, B, C and Figure 10.8
D in the figure 10.3 are a harmonic
range.
Another example is the
proof of the Pappus’ theorem: the
lines joining points with distinct
letters belonging to two distinct
ranges ABC and A'B'C' intersect in
three collinear points L, M and N
(figure 10.8). Applying a
perspectivity to make both ranges
parallel, now we see that the
triangles ABL and A'B'L are
similar so that:
A'B' AB A'B' AB
L= A+ B' and also L= B+ A'
AB + A'B' AB + A'B' AB + A'B' AB + A'B'
We may gather both equations obtaining an expression more adequate to our purposes
which is function of the midpoints:
A'B' A+ B AB A' + B'

L= +
AB + A'B' 2 AB + A'B' 2
Analogously for M and N we obtain:
A'C' A+C AC A' + C'

M= +
AC + A'C' 2 AC + A'C' 2
B'C' B+C BC B' + C'

N= +
BC + B'C' 2 BC + B'C' 2
These expressions fulfil the following equality what implies that the three points are
aligned:
AB + A'B' BC + B'C'
M= L+ N
AC + A'C' AC + A'C'
To prove this, begin from
( AC + A'C' ) M = ( AB + A'B' ) L + ( BC + B'C' ) N ⇒
A'C' ( A + C ) + AC ( A' + C' ) = A'B' ( A + B ) + AB ( A' + B' ) + B'C' (B + C ) + BC (B' + C' )
Arranging A, B and C on the left hand side and A', B' and C' on the right hand side
and taking into account that AC=AB+BC and A'C'=A'B'+B'C' we obtain
the following vector equality:
− B'C' AB + A'B' BC = − AB B'C' + BC A'B'
which is an identity taking into account that the lines ABC and A'B'C' are parallel and all
vectors have the same direction and sense.
The linear relation between L, M and N is very useful to calculate the cross ratio
(LMNR). In fact the coefficients of the midpoints are the relative altitudes of the point with
respect to both parallel lines. Since this cross ratio is also the quotient of the relative
altitudes, we have:
 B'C' A'B'  − AC
 − 
 BC + B'C' AB + A'B'  AC + A'C' AC
(LMNR ) =  = = (OABC )
− AB  B'C' A'C'  AB
 − 
AB + A'B'  BC + B'C' AC + A'C' 
The single ratio is the cross ratio corresponding to O = ∞. This identity of cross ratios is
preserved under a projectivity and it is also valid for the upper scheme on the figure 10.8
when O is a point with finite coordinates. Analogously:
 B'C' A'B'  A'C'

 − 
 BC + B'C' AB + A'B'  AC + A'C' A'C'
(LMNR' ) =  = = (OA'B'C' )
A'B'  B'C' A'C'  A'B'
 − 
AB + A'B'  BC + B'C' AC + A'C' 
The homology
The points A', B' are homologous of the points A and B with respect to the centre
O and the axis of homology given by the point F and the vector v (figure 10.9) if:
1) The centre O does not lie on the axis.
2) Every pair of homologous points and the homology centre are aligned, that is,
O, A and A' are aligned and also O, B and B'.
3) Every pair of homologous lines AB and A'B' intersects in a point on the
homology axis.
A homology6 is determined by an axis, a
centre O and a pair of homologous points A and Figure 10.9
A'. To obtain the homologous of any point B, we
draw the line AB, whose intersection with the
homology axis will be denoted as Z. Then we
draw the line A'Z and the intersection with the
line OB is the homologous point B'. By
construction, the homology axis and the lines
AB, A'B' and OZ are concurrent in Z. The
projective cross ratio of this pencil of lines is
independent of the crossing line where it is
measured. Let G and H be the intersections of
the homology axis with the lines OA and OB
respectively. Then we have:
{ Z, O A G A' } = { Z, O B H B' }
6
The word homology is used by H. Eves (A Survey of Geometry, p. 105) with a different
meaning. He names as homology a composite homothety, that is, a single homothety followed
by a rotation.
Since the four points of each set are collinear, their projective cross ratio is identical to
their complex cross ratio:
( O A G A' ) = ( O B H B' )
That is, the cross ratio of the four points O, B, H and B' is a real constant r for any point B
and is called homology ratio. Hence we can write:
( O B H B' ) = OH OB' −1 BB' BH −1 = r ≠ 1, 0
This equation leads to a simpler definition of the homology: B' is the homologous point of
B with respect a given axis and centre O (outside the axis) with ratio r if O, B and B' are
aligned and the cross ratio ( O B H B' ) is equal to r, being H the intersection point of the
line OB with the homology axis. Hence, a homology is determined by giving a centre O,
an axis and the homology ratio. If the homology ratio is the unity we have two degenerate
cases: whether B lies on O and B' is any point, or B' lies on H (on the homology axis) and
B is any point, that is, there is not one to one mapping, a not useful transformation. Then
we must impose a ratio distinct of the
Figure 10.10
unity. On the other hand, a null
homology ratio implies that BB' = 0 and
the homology becomes the identity.
In the special case r = 2 (when
the points A and A′ are harmonic
conjugates with respect O and H) the
transformation is called a harmonic
homology.
The first limit line is defined as
the set of points L on the plane having as
homologous points those at infinity
(figure 10.10):
( O L H ∞ ) = OH O∞ −1 L∞ LH −1 = r
The limit of the quotient of both distances tending to the infinity is the unity:
O∞ −1 L∞ = lim OX −1 LX = 1
X →∞
Then: OH LH −1 = r ⇒ LH OH −1 = r −1
Because all the points H lie on the homology axis, all the points L form a line, the limit
line.
The second limit line is defined as the points that are homologous of those at
infinity:
( O ∞ H L' ) = OH OL' −1 ∞L' ∞H −1 = r

OH OL' −1 = r ⇒ OL' OH −1 = r −1
Therefore the distance from the second limit line to the homology axis is equal to the
distance from the first limit line to the homology centre (figure 10.10), and both lines are
located whether between or outside both elements according to the value of r:
OL' = LH = r −1 OH
Let us calculate the homologous point B' for every point B on the plane. Since H is
a changing point on the homology axis, we write it as a function of a fixed point F also on
the axis and its direction vector v (figure 10.9). At one time, the point H is the intersection
of the line OB and the homology axis:
H = O + b OB = F + k v k, b real
b OB − k v = OF ⇒ b = OF ∧ v ( OB ∧ v ) −1
H = O + OF ∧ v ( OB ∧ v ) −1 OB
Now we may calculate the segments OH and BH:
OH = OF ∧ v ( OB ∧ v ) −1 OB
BH = BO + OF ∧ v ( OB ∧ v ) −1 OB = [ OF ∧ v ( OB ∧ v ) −1 − 1 ] OB =
= ( OF ∧ v − OB ∧ v ) ( OB ∧ v ) −1 OB = BF ∧ v ( OB ∧ v ) −1 OB
By substitution in the cross ratio we obtain:
OH OB' −1 BB' BH −1 = OH BH −1 BB' OB' −1 = OF ∧ v ( BF ∧ v ) −1 BB' OB' −1 = r
BB' OB' −1 = r v ∧ BF ( v ∧ OF ) −1
and isolating OB' as a function of OB and other known data:
( −OB + OB' ) OB' −1 = r v ∧ BF ( v ∧ OF ) −1 −OB OB' −1 = r v ∧ BF ( v ∧ OF ) −1 − 1
OB' −1 = OB −1 [ 1 − r v ∧ BF ( v ∧ OF ) −1 ] = OB −1 ( v ∧ OF − r v ∧ BF ) ( v ∧ OF ) −1
Inverting the vectors we find:
OB' = OB v ∧ OF ( v ∧ OF − r v ∧ BF ) −1 = OB ( 1 − r v ∧ BF ( v ∧ OF ) −1 ) −1
This equation allows to calculate the homologous point B' of any point B using the data of
the homology: the centre O, the ratio r, and a direction vector v and a point F of the axis.
The homology preserves the projective cross ratio of any set of points. Let A', B',
C', D' and E' be the transformed points of A, B, C, D and E under the homology. Then:
{ E, A B C D } = { E', A' B' C' D' }
Let us prove this statement. Being A' and E' homologous points of A and E we have:
OA' = OA ( 1 − r v ∧ AF ( v ∧ OF ) −1 ) −1
OE' = OE ( 1 − r v ∧ EF ( v ∧ OF ) −1 ) −1
Thus the segment E'A' is:
E'A' = OA' − OE'
E'A' = OA ( 1 − r v ∧ AF ( v ∧ OF ) −1 ) −1 − OE ( 1 − r v ∧ EF ( v ∧ OF ) −1 ) −1 =
= [ OA ( 1 − r v ∧ EF ( v ∧ OF ) −1 ) − OE ( 1 − r v ∧ AF ( v ∧ OF ) −1 ) ]
[ 1 − r v ∧ AF ( v ∧ OF ) −1 ] −1 [ 1 − r v ∧ EF ( v ∧ OF ) −1 ] −1 =
= [ EA + r ( −OA v ∧ EF + OE v ∧ AF ) ( v ∧ OF ) −1 ]
[ 1 − r v ∧ AF ( v ∧ OF ) −1 ] −1 [ 1 − r v ∧ EF ( v ∧ OF ) −1 ] −1 =
= [ EA ( 1 − r ) + r ( OA v ∧ OE − OE v ∧ OA ) ( v ∧ OF ) −1 ]
[ 1 − r v ∧ AF ( v ∧ OF ) −1 ] −1 [ 1 − r v ∧ EF ( v ∧ OF ) −1 ] −1 =
= [ EA ( 1 − r ) + r v OA ∧ OE ( v ∧ OF ) −1 ]
[ 1 − r v ∧ AF ( v ∧ OF ) −1 ] −1 [ 1 − r v ∧ EF ( v ∧ OF ) −1 ] −1
For the other vectors analogous expressions are obtained:
E'B' = [ EB ( 1− r ) + r v OB ∧ OE ( v ∧ OF ) −1 ]
[ 1 − r v ∧ BF ( v ∧ OF ) −1 ] −1 [ 1 − r v ∧ EF ( v ∧ OF ) −1 ] −1
E'C' = [ EC ( 1− r ) + r v OC ∧ OE ( v ∧ OF ) −1 ]
[ 1 − r v ∧ CF ( v ∧ OF ) −1 ] −1 [ 1 − r v ∧ EF ( v ∧ OF ) −1 ] −1
E'D' = [ ED ( 1− r ) + r v OD ∧ OE ( v ∧ OF ) −1 ]
[ 1 − r v ∧ DF ( v ∧ OF ) −1 ] −1 [ 1 − r v ∧ EF ( v ∧ OF ) −1 ] −1
Let us calculate the outer product E'A' ∧ E'C':

E'A' ∧ E'C' = [ EA ∧ EC ( 1 − r )2 + ( EA ∧ v OC ∧ OE + v ∧ EC OA ∧ OE ) r ( 1 − r )
( v ∧ OF ) −1 ] [ 1 − r v ∧ AF ( v ∧ OF ) −1 ] −1 [ 1 − r v ∧ CF ( v ∧ OF ) −1 ] −1
[ 1 − r v ∧ EF ( v ∧ OF ) −1 ] −2
By substitution of the following identity, which may be proved by means of the geometric
algebra (see exercise 1.4):
EA ∧ v OC ∧ OE + v ∧ EC OA ∧ OE = EA ∧ v EC ∧ OE + v ∧ EC EA ∧ OE =
EA ∧ EC v ∧ OE
E'A' ∧ E'C' = [ EA ∧ EC ( 1 − r )2 + EA ∧ EC v ∧ OE r ( 1 − r ) ( v ∧ OF ) −1 ]
[1− r v ∧ AF (v ∧ OF ) −1 ] −1 [1− r v ∧ CF ( v ∧ OF ) −1 ] −1 [1− r v ∧ EF ( v ∧ OF ) −1 ] −2
Extracting common factor:
E'A' ∧ E'C' = EA ∧ EC [ ( 1 − r )2 + v ∧ OE r ( 1 − r ) ( v ∧ OF ) −1 ]
[1− r v ∧ AF (v ∧ OF ) −1 ] −1 [1− r v ∧ CF ( v ∧ OF ) −1 ] −1 [1− r v ∧ EF ( v ∧ OF ) −1 ] −2
In the projective cross ratio, all the factors are simplified except the first outer product:
E'A' ∧ E'C' E'B' ∧ E'D' EA ∧ EC EB ∧ ED

=
E'A' ∧ E'D' E'B' ∧ E'C' EA ∧ ED EB ∧ EC
This proves that the projective cross ratio of any four points A, B, C and D with respect to
a centre of projection E is equal to the projective cross ratio of the homologous points A',
B', C' and D' with respect to the homologous centre of projection E'. It means that the
homology is a special kind of projectivity where a line is preserved, the axis of homology.
When the centre of projection is a point H on the homology axis, H and H' are coincident
and we find in the former equality the initial condition of the homology again:
H = H' { H, A' B' C' D' } = { H, A B C D }
The following chapter is devoted to the conics and the Chasles’ theorem, which
states that the locus of the points from where the projective cross ratio of any four points
is constant is a conic passing through these four points. Since the homology preserves the
projective cross ratio, it transform conics into conics. Hence, a conic may be drawn as the
homologous curve of a circle. Depending on the position of this circle there are three
cases:
1) The circle does not cut the limit line L (whose homologous is the line at
infinity). In this case the homologous curve is an ellipse, because very next points on the
circle have also very next homologous points and therefore the homologous curve must be
closed.
2) The circle touches the limit line L. Then its homologous curve is the parabola,
because the homologous of the contact point is a point at infinity. The symmetry axis of
the parabola has the direction of the line passing through the centre of homology and the
contact point.
3) The circle cuts the limit line L. Then the homologous curve is the hyperbola
because the homologous points of both intersections are two points at infinity. The lines
passing through the centre of homology and both intersection points have the direction of
the asymptotes.
Exercises
10.1 Prove the Ptolemy’s theorem: For a quadrilateral inscribed in a circle, the product of
the lengths of both diagonals is equal to the sum of the products of the lengths of the pairs
of opposite sides.
10.2 Find that the point D forming a harmonic range with A, B and C is given by the
equation:
D = A + AB ( 1 − 2 AC −1 BC ) −1
10.3 Show that the homography is a directly conformal transformation, that is, it preserves
angles and their orientations.
10.4 Prove that if A', B', C' and D' are the homologous of A, B, C and D, and H is a point
on the homology axis, then the equality { H, A' B' C' D' } = { H, A B C D } holds. That is,
show directly that:
HA' ∧ HC' HB' ∧ HD' HA ∧ HC HB ∧ HD

=
HA' ∧ HD' HB' ∧ HC' HA ∧ HD HB ∧ HC
10.5 A special conformal transformation of centre O and translation v is defined as the

geometric operation which, given any point P, consists in the inversion of the vector OP,
the addition of the translation v and the inversion of the resulting vector again.
OP' −1 = ( OP −1 + v ) −1
Prove the following properties of this transformation:

a) It is an additive operation with respect to the translations: the result of
applying firstly a transformation with centre O and translation v, and later a
transformation with the same centre and translation w is identical to the result
of applying a unique transformation with translation v + w.
b) It preserves the complex cross ratio and thus it is a special case of
homography.
10.6 Prove that if a homography keeps invariant three or more points on the plane, then it
is the identity.
10.7 An antigraphy is defined as that transformation which conjugates the complex cross
ratio of any four points:
( A B C D ) = ( A' B' C' D' )*
Then an antigraphy is completely specified by giving the images A', B', C' of any three
points A, B, C. Prove that:
a) The antigraphy is an opposite conformal transformation, that is, it changes the
orientation of the angles but preserving their absolute values.
b) The composition of two antigraphies is a homography.
c) A product of three inversions is an antigraphy.
d) If an antigraphy has three invariant points, then it is a circular inversion.
10.8 Let A, B, C and A', B', C' be two sets of independent points. A projectivity has
been defined as the mapping:
D=aA+bB+cC → D' = a' A' + b' B' + c' C'
where the coefficients a', b', c' are homogeneous linear functions of a, b, c. Prove that
collinear points are mapped to collinear points.
10.9 Prove the following theorem: let the sides of a hexagon ABCDEF pass alternatively
through the points P and Q. Then the lines AD, BE and CF joining opposite vertices meet
in a unique point. Hint: draw the hexagon in the dual plane.
11. CONICS
Conic sections
The conic sections (or conics) are the intersections of a conic surface with any
transversal plane (figure 11.1). The proper conics are the curves obtained with a plane not
passing through the vertex of the cone. If the plane contains the vertex we have improper
conics, which reduce to a pair of straight lines or a point. In general, I shall regard only
proper conics, taking into account that the other case can be usually obtained as a limit
case.
Let us consider the two spheres inscribed in the cone which are tangent to the
Figure 11.1
plane of the conic. The spheres touch this plane at the points F and F', the foci of the conic
section. On the other hand, both spheres touch the cone surface at two tangency circles
lying in planes perpendicular to the cone axis. The directrices of the conic are defined as
the intersections of these planes with the plane of the conic.
Let P be a point on the conic. Since it lies on the plane of the conic, the segment
going from P to the focus F is tangent to the sphere. Since it belongs to the cone surface,
the generatrix passing through P is also tangent to the sphere. PF and PS are tangent to the
same sphere so that their lengths are equal: PF=PS. Also PF' and PS' are tangent to
the other sphere so that their lengths are also equal: PF'=PS'. On the other hand the
segment SS' of any directrix has constant length, what implies that the addition of
distances from any point P on the conic to both foci is constant:
PF + PF' = PS + PS' = SS' = constant
This discussion slightly changes for a plane which intercepts the upper and lower cones.
In this case, the distances from P to the upper sphere and to the upper branch of the curve
must be considered negative1. If the plane is parallel to a generatrix of the cone these
distances become infinite. This unified point of view for all conics using negative or
infinite distances when needed will be the guide of this chapter. I shall not separate
equations but only specify cases for distinct conics.
A lateral view of figure 11.1 is given in the figure 11.2. In this figure the point P
has been drawn in both extremes of the conic (and S and S' also twice), although we must
consider any point on the conic section. As indicated previously PF=PS. But PS
is proportional to the distance from P to the upper plane containing S and this distance is
also proportional to the distance from P to the directrix r (figure 11.2):
sinα
PF = PS = d (P , r )
sinδ
The quotient of the sines of both angles is called the eccentricity e of the conic:
sinα
e= >0 Figure 11.2
sinδ
Then a conic can be also defined as

the geometric locus of the points P
whose distance from the focus F is
proportional to the distance from the
directrix r (figure 11.3).
FP = e d (P, r )
FP is called the focal radius. Let T

be the point of the directrix closest to
the focus (figure 11.3). Then the main
axis of symmetry of the conic is the line
which is perpendicular to the directrix and
passes through the focus. Under a
reflection with respect the main axis of
symmetry, the conic is preserved. The
oriented distance from any point P to the
directrix r is expressed, using the vector
FT, as:
Figure 11.3
1
To consider sensed distances is not a trouble but an advantage. For example, the distance from
a point to a line can be also taken as an oriented distance, whose sign indicates the half-plane
where the point lies.
FT · PT
d (P , r ) =
FT
Defined in this way, the distance from P to r is positive when P is placed on the half-plane
that contains the focus of both in which the directrix divides the plane, and negative when
P lies on the other half-plane. The equation of the conic becomes:
FT · PT
FP = e d (P, r ) = e
FT
FP FT = e FT · PT = e FT · (FT − FP ) = e (FT 2 − FT · FP )
Let us denote by β the angle between the main axis FT and the focal vector FP:
FP FT = e (FT 2 − FT FP cos β )
FP FT (1 + e cos β ) = e FT 2
From here one obtains the polar equation of a conic:
FT
FP =
1 + e cos β
According to the value of the eccentricity there are three types of proper conics:
1) For 0 < e < 1, the inclination α of the plane of the conic is lower than the
inclination δ of the cone generatrix and the denominator is always positive.
Then the focal radius is always positive and the points P form a closed curve
on the focal half-plane, an ellipse.
2) For e = 1, α = δ , the plane of the conic is parallel to a cone generatrix. Since
the denominator vanishes for β = π, the points P form an open curve called
parabola. Except for this value of β, the radius is always positive and the
curve lies on the focal half-plane.
3) For e > 1, α >δ , the plane of the conic intercepts both upper and lower cones.
The denominator vanishes twice so that it is an open curve with two branches
called hyperbola (figure
11.4). The lowest positive
Figure 11.4 angle that makes the
denominator zero is the
asymptote angle βa :
 1
β a = arccos − 
 e
The maximum eccentricity of the

hyperbolas as cone sections is obtained
with α =π/2:
1
1< e ≤
sinδ
The focal radius is positive (P lies on the focal branch) for the ranges of β :
0 < β < βa and 2π – βa < β < 2π
and negative (P lies on the non focal branch) for the range:
βa < β < 2π – βa
A negative focal radius means that we must take the sense opposite that determined by the
angle (figure 11.4). For example, a radius –2 with an angle 7π/6 is equivalent to a radius 2
with an angle of π/6. So the vertex T' of the non focal branch has an angle π (not zero as it
would appear) and a negative focal radius.
Two foci and two directrices Figure 11.5
There are usually two spheres

touching the cone surface and the conic
plane, and hence there are also two foci. For
the case of the ellipse, both tangent spheres
are located in the same cone. For the case of
the hyperbola, each sphere is placed in each
cone so that each focus is placed next to
each branch. In the case of the parabola,
one sphere and the corresponding focus is
placed at the infinity. For the improper
Figure 11.6
conics, the tangent spheres have null radius
and the foci are coincident with the vertex of
the cone, which is the crossing point of both
lines.
As the figure 11.1 shows for a proper
conic, both directrices r and r' are parallel, and
both foci F and F' are located on the axis of
symmetry of the conic. Above we have already
seen that the addition of oriented distances
from both foci to any point P is constant. Let
us see the relation with the distance between
the directrices. If P is any point on the conic
then:
FP + F'P = e d (P, r ) + e d (P, r' ) = e d (r, r' )
The sum of the oriented distances from any point P to any two parallel lines (the
directrices in our case) is constant2. Then the addition of both focal radii of any point P
2
Note that this statement is only right for oriented distances but not for positive distances.
on the conic is the product of the eccentricity multiplied by the oriented distance between
both directrices. P can be any point, for example Q, the extreme of the conic:
FP + F'P = FQ + F'Q = F'Q' + F'Q = QQ'
whence it follows that the addition of both focal radii is QQ' , the major diameter.
Specifying:
1) For the case of the ellipse (figure 11.5), both focal radii are positive and also
the major diameter.
2) For the case of the parabola, one directrix is the line at the infinity and one
focus is a point at the infinity on the main axis of symmetry. Hence the major
diameter has an infinite value.
3) For the case of the hyperbola (figure 11.6), the focal radius of a point on the
non focal branch is negative, its absolute value being higher than the other
focal radius, which is positive, yielding a negative major diameter.
Vectorial equation
The polar equation of a conic may be written having as parameter the focal
distance FQ instead of the distance from the focus to the directrix FT:
1+ e
FP = FQ
1 + e cos β
Let Q' be the vertex of the conic closest to the focus F'. The distance from the focus F to
Q' is found for β = π:
1+ e
FQ' = FQ
1− e
For the ellipse FQ' is a positive distance; for the hyperbola it is a negative distance and
for the parabola Q' is at infinity. Then the distance between both foci is:
2e
FF' = FQ' − F'Q' = FQ' − FQ = FQ
1− e
On the other hand the major diameter is:
2 FQ
QQ' = FQ + FQ' =
1− e
By dividing both equations, an alternative definition of the eccentricity is obtained:
FF'
e=
QQ'
The eccentricity is the ratio of the distance between both foci divided by the major
diameter. For the case of ellipse, both distances are positive. For the case of the hyperbola,
both distances are negative, so the eccentricity is always positive. When e = 0 both foci
are coincident in the centre of a circle. Note that the circle is obtained when we cut the
cone with a horizontal plane. In this case, the directrices are the line at the infinity.
The vectorial equation of a conic is obtained from the polar equation and contains
the radius vector FP. Since FP forms with FQ an angle β (figure 11.3), FP is obtained
from the unitary vector of FQ via multiplication by the exponential with argument β and
by the modulus of FP yielding:
1+ e
FP = FQ (cos β + e12 sinβ )
1 + e cos β
On the other hand, from the directrix property, one easily finds the following
equation for a conic:
FP2 FT2 = e2 ( FT2 − FT · FP )2
F, T and e are parameters of the conic, and P = (x, y) is the mobile point. Therefore from
this equation we will also obtain a Cartesian equation of second degree. For example, let
us calculate the Cartesian equation of an ellipse with eccentricity ½ and focus at the point
(3, 4) and vertex at (4,5):
e = 1/2 F = ( 3, 4 ) Q = ( 4, 5 ) P = ( x, y )
1+ e
FT = FQ = 3 e1 + 3 e 2 T = F + FT = ( 6, 7 )
e
FP = ( x − 3 ) e1 + ( y − 4 ) e2
Then the equation of the conic is:
1
[ ( x − 3 )2 + ( y − 4 )2 ] 18 = [ 18 − ( 3 (x − 3) + 3 ( y − 4 ) ) ]2
4
After simplification:
7 x2 − 2 x y + 7 y2 − 22 x − 38 y + 31 = 0
The Chasles’ theorem
According to this theorem3, the

projective cross ratio of any four given
points A, B, C and D on a conic with
respect to a point X also on this conic is
constant independently of the choice of Figure 11.7
3
Michel Chasles, Traité des sections coniques, Gauthier-Villars, Paris, 1865, p. 3.
the point X (figure 11.7):

{ X, A B C D } = { X', A B C D }
To prove this theorem, let us take into Figure 11.8

account that the points A, B, C, D and X must
fulfil the vectorial equation of the conic. Let
us also suppose, without loss of generality,
the main axis of symmetry having the
direction e1 (this supposition simplifies the
calculations):
FQ = FQ e1
Now on α, β, γ, δ and χ will be the angles

which the focal radii FA, FB, FC, FD and
FX form with the main axis with direction
vector FQ (figure 11.8). Then:
 e cos α + e 2 sinα e1 cos χ + e 2 sinχ 

XA = FA − FX = FQ (1 + e )  1 − 
 1 + e cos α 1 + e cos χ 
Introducing a common denominator, we find:
(1 + e ) [e1 (cos α − cos χ ) + e 2 (sinα − sinχ + e sinα cos χ − e cos α sinχ )]

XA = FQ
(1 + e cos α ) (1 + e cos χ )
From XA and the analogous expression for XC, and after simplification we obtain:
e (1 + e ) (sinγ cos α − sinα cos γ + sinχ cos γ − sinγ cos χ + sinα cos χ − sinχ cos α )
2
XA ∧ XC = FQ 122
(1 + e cos α )(1 + e cos γ )(1 + e cos χ )
(1 + e )2 [ sin(γ − α ) + sin (χ − γ ) + sin (α - χ ) ]

= FQ 2 e12
(1 + e cos α )(1 + e cos γ )(1 + e cos χ )
Using the trigonometric identities of the half-angles, the sum is converted into a product
of sines (exercise 6.2):
( 1 + e )2  sin γ − α  sin χ − γ  sin α − χ  
XA ∧ XC = − 4 FQ 2   2   2   2  e
( 1 + e cos α ) ( 1 + e cos γ ) ( 1 + e cos χ ) 12
In the same way the other outer products are obtained. The projective cross ratio is their
quotient, where the factors containing the eccentricity or the angle χ are simplified:
γ −α δ −β AFC BFD
sin sin sin sin
XA ∧ XC XB ∧ XD 2 2 = 2 2
{X , A B C D } = =
XA ∧ XD XB ∧ XC δ − α γ − β AFD BFC
sin sin sin sin
2 2 2 2
since γ −α is the angle AFC, etc. Therefore, the projective cross ratio of four points A, B,
C and D on a conic is equal to the quotient of the sines of the focal half-angles, which do
not depend on X, but only on the positions of A, B, C and D, fact which proves the
Chasles’ theorem. This statement is trivial for the case of the circumference, because the
inscribed angles are the half of the central angles. However the inscribed angles on a conic
vary with the position of the point X and they differ from the half focal angles. In spite of
this, it is a notable result that the quotient of the sines of the inscribed angles (projective
cross ratio) is equal to the quotient of the half focal angles. For the case of the hyperbola I
remind you that the focal radius of a point on the non focal branch is oriented with the
opposite sense and the focal angle is measured with respect this orientation.
Tangent and perpendicular to a conic
The vectorial equation of a conic with the

major diameter oriented in the direction e1 (figure Figure 11.9
11.9) is:
(1 + e ) FQ
FP = ( e1 cos α + e 2 sinα )
1 + e cosα
The derivation with respect the angle α gives:
d FP ( 1 + e ) FQ
= [− e sinα + e2 ( e + cos α ) ]
dα ( 1 + e cos α )2 1
This derivative has the direction of the line tangent to the conic at the point P, its unitary
vector t being:
− e1 sinα + e 2 ( e + cos α )
t=
1 + e 2 + 2 e cos α
The unitary normal vector n is orthogonal to the tangent vector:
e1 ( e + cos α ) + e 2 sinα
n=
1 + e 2 + 2 e cos α
In the same way, t and n are obtained as functions of the angle β (figure 11.9) from the
vectorial equation for the focus F':
(1 − e ) FQ
F'P = ( e1 cos β + e 2 sinβ )
1 − e cosβ
− e1 sinβ + e 2 (− e + cos β )
t=
1 + e 2 − 2 e cos β Figure 11.10
e1 ( − e + cos β ) + e 2 sinβ
n=
1 + e 2 − 2 e cos β
Let us see now that the normal

vector has the direction of the bisector of
the angle between both focal radii (figure
11.10). In order to prove this statement, let
us consider an infinitesimal displacement
of a point P on the conic (figure 11.11).
The sum of the focal radii of any conic is constant, whence it follows that the addition of
their differentials are null:
FP + F'P = constant ⇒ 0 = dFP + dF'P
The differential of the focal

radius4 is obtained by differentiating the Figure 11.11
square of the focal vector:
d FP2 = 2 FP · d FP = 2 FP dFP
d FP · FP
d FP = −
FP
Summing the differentials of both

moduli we obtain:
d FP · FP d F'P · F'P  FP F'P 

0 = d FP + d F'P = − − = − d FP ·  + 

FP F'P  FP F'P 
since the differentials of the focal vectors are equal: dP = d FP = d F'P. In conclusion the
differential vector (or the tangent vector) is orthogonal to the bisector of both focal
vectors. The figure 11.11 clearly shows that both right triangles are opposite because they
share the same hypotenuse d FP, and the legs dFP and dF'P have the same length.
The reflection axis of both triangles, which is the shorter diagonal of the rhombus, is the
bisector line of both focal vectors.
In the case of the parabola (figure 11.12), the focus F' is located at the infinity and
F'P is parallel to the main axis of symmetry. In the case of the hyperbola (figure 11.13)
we must take into account that the radius of a point on the non focal branch is negative, so
4
Note that it differs from the modulus of the differential of the focal vector: dFP≠d FP.
it has the sense from P to the focus F'. The normal direction is the bisector of both
oriented focal vectors.
These geometric features are the widely known optic reflection properties of the
conics: parallel beams reflected by a parabola are concurrent at the focus. Also the beams
emitted by a focus of an ellipse and reflected in its perimeter are concurrent in the other
focus.
Figure 11.12 Figure 11.13
Central equations for the ellipse and hyperbola
We search a polar equation to describe a conic using its centre as the origin of the
radius. Above we have seen that the eccentricity is the ratio of the distance between both
foci and the major diameter:
FF'
e=
QQ'
If O is the centre of the conic then it is the

midpoint of both foci:
F + F'
O=
2
OF FQ
e= 1− e = Figure 11.14
OQ OQ
For the case of the ellipse all the quantities are positive. For the case of the
hyperbola both OF and OQ are negative with FQ positive. Let the angle QOP be
γ (figure 11.14). Then:
OP sin γ = FP sin α
OP cos γ = OF + FP cos α

For the ellipse all the moduli are positive whereas for the hyperbola OP and
OF are always negative (the centre is placed between both branches). For the parabola
they become infinite. By substitution of the polar equation into these equalities we have:
OP sinγ = FQ
(1 + e ) sinα
= OQ
(1 − e )sinα
2
1 + e cos α 1 + e cos α
OP cosγ = OF +
FQ (1 + e ) cosα 
= OQ e +
(1 − e 2 )cos α  e + cosα
 = OQ
1 + e cos α  1 + e cos α  1 + e cos α
Summing the squares of both former equalities we find:
 sin 2 γ  OQ
OP 2  + cos 2 γ  = OQ 2 ⇒ OP =
1 − e
2
 sin 2 γ
+ cos 2 γ
1 − e2
The direction of OP with respect to OQ is given by the angle γ and we may add it
multiplying by the unitary complex with argument γ to obtain the central vectorial
equation:
OQ (cos γ + e12 sinγ )

OP =
sin 2 γ
+ cos 2 γ
1 − e2
Let R be the point P for γ =π/2. Then OR lies on the secondary axis of symmetry
of the conic, which is perpendicular to OQ, the main axis of symmetry. ORis the minor
half-axis. The eccentricity relates both:
OR2 = ( 1 − e2 ) OQ2
OR = 1 − e 2 OQ
OR = 1 − e 2 OQ e12
Using the minor half-axis, the equation of the ellipse becomes:
1  sinγ 
OP =  OQ cos γ + OR 
 
sin 2 γ  1 − e2 
+ cos 2 γ
1− e 2
Hence it follows:
OP 2 OP 2
2
sin 2
γ + 2
cos 2 γ = 1
OR OQ
Taking OP and OQ as the coordinates axis and O the origin of coordinates, we have the
canonical equation of a conic:
x = OP cos γ y = OP sin γ
x2 y2
+ =1
OQ 2 OR 2
This equation of a conic is specific of this coordinate system and has a limited
usefulness. If another system of coordinates is used (the general case) the equation is
always of second degree but not so beautiful. For the ellipse both half-axis are positive
real numbers. However for the hyperbola, the minor half-axis OQ is imaginary and
OQ2<0 converting the sum of the squares in the former equation in a difference of real
positive quantities.
Diameters and Apollonius’ theorem
If e<1 (ellipse) we may introduce the angle θ in the following way:
cos γ Figure 11.15

cos θ =
sin 2 γ
+ cos 2 γ
1− e 2
Then the central equation of an ellipse

becomes5:
OP = OQ cos θ + OR sin θ
The angle θ has a direct geometric

interpretation if we look at a section of a
cylinder (figure 11.15):
cos γ cos φ
cos θ =
sin γ + cos 2 γ cos 2 φ
2
From where it follows the relationship between the inclination of the cylindrical section
and the eccentricity:
cos φ = 1 − e 2
5
G. Peano (Gli elementi di calcolo geometrico [1891] in Opere Scelte, vol. III, Edizioni Cremone,
[Roma, 1959], p.59) gives this central equation for the ellipse as function of the angle θ. He also
used the parametric equations of the parabola, hyperbola, cycloid, epicycloid and the spiral of
Archimedes.
Then θ is an angle between axial planes of the cylinder.

The equation of the ellipse may be written using another pair of axis turned by a
cylindrical angle χ (figure 11.16):
OQ' = OQ cos χ + OR sinχ OR' = −OQ sin χ + OR sinχ
Each pair of OQ' and OR' obtained in this way are conjugate central radii and
twice them are conjugate diameters. They are intersections of the plane of the ellipse with
two axial planes of the cylinder that form
a right angle. From the definition it is
obvious that:
OQ' 2 + OR' 2 = OQ2 + OR2
The inverse relation between the

conjugate radii are:
OQ = OQ' cos χ − OR' sin χ
OR = OQ' sin χ + OR' cos χ Figure 11.16
Then, the central radius of any point P is:
OP = ( OQ' cos χ − OR' sin χ ) cos θ + ( OQ' sin χ + OR' cos χ ) sin θ
= OQ' cos( θ −χ ) + OR' sin( θ −χ )
Changing the sign of the cylindrical angle we obtain another point P' on the ellipse owing
to the parity of trigonometric functions:
OP' = OQ' cos(θ −χ ) − OR' sin(θ −χ ) = OQ' cos(χ −θ ) + OR' sin(χ −θ )
Then the chord PP' is parallel to the radius OR' (figure 11.16):
PP' = OP' − OP = −2 OR' sin( θ − χ )
Now, let us mention the Apollonius theorem: a diameter of a conic is formed by

all the midpoints of the chords parallel to its conjugate diameter. To prove this statement,
see that the central radius of the midpoint has the direction OQ':
OP + OP'
= OQ' cos(θ − χ )
2
The properties of the conjugate diameters are also applicable to the hyperbola, but
taking into account that OR2 = (1 − e2) OQ2 < 0, that is, the minor half-axis OR is an
imaginary number. In this case OR has the same direction than OQ:
OR = 1 − e 2 OQ e12 = e 2 − 1 OQ
And also the angle θ is an imaginary number, and the trigonometric functions are turned
into hyperbolic functions of the real argument ψ = θ / e12 :
cos γ
cos θ = = coshψ ≥ 1 e >1
sin γ
2
+ cos 2 γ
1− e 2
Taking into account the relations between trigonometric and hyperbolic functions:
ψ ψ sinhψ
cos = coshψ sin =
e12 e12 e12
the relation of the hyperbolic angle ψ with the Cartesian coordinates is:
x y
coshψ = sinhψ = e12
OQ OR
(
OP = ± (OQ coshψ + OR e12 sinhψ ) = ± OQ coshψ + e 2 − 1 sinhψ OQ e12 )
Let us define OS, which is taken
usually as the minor half-axis of the Figure 11.17
hyperbola although this definition is
not exact, as (figure 11.17):
OS = e 2 − 1 OQ e12
Then the equation of the hyperbola is:
OP = OQ coshψ + OS sinhψ
When the hyperbolic angle tends to

infinity we find the asymptotes:
OQ + OS
ψ→∞ OP → ± expψ
2
OQ − OS
ψ→−∞ OP → ± exp(− ψ )
2
Two radii are conjugate if they are turned through the same hyperbolic angle ϕ (figure
11.17):
OQ' = OQ cosh ϕ + OS sinhϕ

OS' = OQ sinhϕ + OS cosh ϕ
Then the equality OQ' 2 − OS' 2 = OQ2 − OS2 holds and the central radius of any point of
the hyperbola is:
OP = ± (OQ' cosh(ψ − ϕ ) + OS' sinh(ψ − ϕ ))
Also it is verified that the diameter is formed by the midpoints of the chords parallels to
the conjugate diameter.
A complete understanding of the central equation of the hyperbola is found in the
hyperbolic prism in a pseudo-Euclidean space. The equality OQ' 2 − OS' 2 = OQ2 − OS2 is
the condition of hyperbolic prism of constant radius. Then the hyperbolas are planar
section of this hyperbolic prism.
Conic passing through five points
Five non collinear points determine a conic. Since every conic has a Cartesian
second degree equation:
a x2 + b y2 + c x y + d x + e y + f = 0
where there are five independent parameters, the substitution of the coordinates of five
points leads to a linear system with five equation, with a hard solving. A briefer way to
obtain the Cartesian equation of the conic passing through these points is through the
Chasles’ theorem. Let us see an example:
A = (1, 1) B = (2, 3) C = (1, −1) D = (0, 0) E = (−1, 3)
EA = (2, −2) EB = (3, 0) EC = (2, −4) ED = (1, −3)
The projective cross ratio of the four points A, B, C and D on the conic is:
EA ∧ EC EB ∧ ED − 4 (− 9 ) 3
r = {E , A B C D} = = =
EA ∧ ED EB ∧ EC − 4 (− 12 ) 4
If the point X = (x, y) then:
XA = (1−x, 1−y) XB = (2−x, 3−y) XC = (1−x, −1−y) XD = (−x, −y)
According to the Chasles’ theorem, the projective cross ratio is constant for any point X
on the conic:
3 XA ∧ XC XB ∧ XD (2 x − 2 ) (3x − 2 y )
= = ⇒ 0 = 12 x 2 − 3 y 2 − x y − 9 x + y
4 XA ∧ XD XB ∧ XC ( x − y ) (4 x − y − 5)
Another way closest to geometric algebra is as follows. Let us take A, B and C as a

point base of the plane and express D and X in this base
D = d A A + d B B + dCC with d A + d B + dC = 1
X = x A A + x B B + xC C with x A + x B + xC = 1
The projective cross ratio is the quotient of the areas of the triangles XAC, XBD,
XAD and XBC:
x A x B xC x A x B xC
1 0 0 0 1 0
XA ∧ XC XB ∧ XD 0 0 1 dA dB dC x B (x A d C − x C d A )
r= = =
XA ∧ XD XB ∧ XC xA xB xC xA xB xC (x B d C − xC d B ) x A
1 0 0 0 1 0
dA dB dC 0 0 1
which yields the following equation:
0 = r x C x A d B + (1 − r ) x A x B d C − x B x C d A
The fifth point E also lying on the conic fulfils this equation:
0 = r eC e A d B + (1 − r ) e A e B d C − e B eC d A
which results in a simplified expression for the cross ratio:
d A dC
−
e A eC
r=
d B dC
−
e B eC
The substitution of the cross ratio r in the equation of the conic gives:
d A e A x B x C (d B eC − d C e B ) + d B e B x C x A (d C e A − d A eC ) + d C eC x A x B (d A e B − d B e A ) =
d A eA d B eB d C eC
d A xA d B xB d C xC = 0
eA x A eB x B eC x C
Conic equations in barycentric coordinates and tangential conic
The Cartesian equation of a conic:
a x2 + b y2 + c x y + d x + e y + f = 0
is written in barycentric coordinates as a bilinear mapping:
 d e 
 f + f + f 
 2 2  1 − x − y
d c+d +e  
(1 − x − y x y)  + f a+d + f + f  x =0
2 2  
e c+d +e  y 
 + f + f b+e+ f 
2 2 
Let us calculate now the dual conic of a given conic. This is defined as the locus of
the points which are dual of the lines tangent to the conic. Then the conic is the envelope
of the tangents. Let us differentiate the Cartesian equation:
δx (2 a x + c y + d ) + δy (2 b y + c x + e ) = 0
where δ indicates the ordinary differential in order to avoid confusion with the coefficient
d. The equation of the line touching the conic at (x0, y0) is:
(2 a x 0 + c y 0 + d ) (x − x 0 ) + (2 b y 0 + c x 0 + e )( y − y 0 ) = 0
(2 a x 0 + c y 0 + d ) x + (2 b y 0 + c x 0 + e ) y − 2 a x 02 − 2 b y 02 − 2 c x 0 y0 − d x0 − e y0 = 0
(2 a x 0 + c y 0 + d ) x + (2 b y 0 + c x 0 + e ) y + d x 0 + e y 0 + 2 f =0
(d x 0 + e y 0 + 2 f )(1 − x − y ) + ((2 a + d ) x 0 + (c + e ) y 0 + d + 2 f ) x +
+ ((2 b + e ) y 0 + (c + d ) x 0 + e + 2 f ) y = 0
If we denote by t, u and v the dual coordinates, the dual homogeneous coordinates t', u'
and v' are linear functions of the barycentric point coordinates:
 t'   2 f d +2 f e+2 f  1 − x 0 − y 0 
u'  =  d + 2 f 2a+2d +2 f

c+d +e+2 f  x0

   
 v'   e + 2 f c+d +e+2 f 
2b+2e+2 f   y0 

Now we see that this is twice the matrix of the conic equation (the matrix of a conic is
defined except by a common factor). Let M be the matrix of the conic, X the matrix of the
point coordinates, and U the matrix of the dual (homogeneous or normalised) coordinates.
U=MX
M −1 U = X
XT = UT M −1
because like M, the inverse matrix M −1 is also symmetric and does not change under
transposition. The substitution in the equation of the conic XT M X = 0 gives:
UT M −1 U = 0
That is, the matrix of the dual conic equation is the inverse of the matrix of the point
conic. Every tangent line of the conic is mapped onto a point of the dual conic and vice
versa. Let us consider for instance the ellipse:
16 16 16  1 − x − y
x
2
 y
2
  
  +   =1 ⇔ (1 − x − y x y ) 16 15 16   x =0
4 2 16 16 12   
  y 
The inverse matrix is:

−1
16 16 16   − 19 / 16 1 1/ 4 
   
16 15 16  = 1 −1 0 
16 16 12   1/ 4 0 − 1 / 4 
  
So the equation of the dual conic is:
 − 19 16 4  t 
 
[ t u v ]  16 − 16 0  u = 0
 4 0 − 4   v 

where we can substitute indistinctly homogeneous or normalised coordinates. If we take t

= 1 − u − v then we obtain the Cartesian equation:
− 67 u 2 − 31 v 2 − 78 u v + 70 u + 46 v − 19 = 0
A concrete case is the tangent line at (0, 2) with equation y = 2. The dual coordinates of
this line are obtained as follows:
 2 1
y − 2 ≡ t' (1 − x − y ) + u' x + v' y ⇒ [t' u' v' ] = [− 2, − 2, − 1] =  , 
 5 5
Let us check that this dual point lies on the dual conic:
 − 19 16 4  2 
 
[2 2 1]  16 − 16 0  2 = 0
 4 0 − 4  1

Polarities
A correlation is defined as the geometric transformation which maps collinear

points onto a pencil of lines, in other words, a linear mapping of points to lines (dual
points), which may be represented with a matrix M:
a'  1 − x − y
b'  = M  x

   
 c'   y 
 
where a', b' and c' are homogeneous dual coordinates.

A polarity is a correlation whose square is the identity. This means that by
applying it twice, a point is mapped onto itself and a line is mapped onto itself. Let us
consider the line [a, b, c]:
1 − x − y a' 
  b'  = 0
[a b c]  x =0 ⇒ [a b c] M −1
 
 y   c' 
 
On the other hand:
1 − x' − y'  a 
  b  = 0
[ a' b' c' ]  x' =0 ⇒ [ a' b' c' ] M −1
 
 y'   c 
 
since [ a'' b'' c'' ]= k [ a b c ] where k is a homogeneous constant. Whence it follows

that the matrix M is symmetric for a polarity.
Under a polarity a point is mapped to its polar line. This point is also called the
pole of the line. The polarity transforms the pole into the polar and the polar into the pole.
In general the polar does not include the pole, except by a certain subset of points on the
plane. The set of points belonging to their own polar line (self-conjugate points) fulfil the
equation:
1 − x − y 
 
(1 − x − y x y ) M  x  = 0
 y 
 
This is just the equation of a point conic. The set of lines passing through their own poles
(self-conjugate lines) fulfil the equation:
Figure 11.18
 
a
[a b c] M −1 b  = 0
 c 
which is just the equation of its dual

(tangential) conic. Then, a polarity has an
associated point and tangential conics, and
every conic defines a polarity. However
depending on the eigenvalues of the matrix, there are polarities that do not have any
associated conic.
The polar of a point outside its associated conic is the line passing through the
points of contact of the tangents drawn from the point (figure 11.18). The reason is the
following: the polarity maps the tangent lines of the associated conic to the tangency
points, so its intersection (the pole) must be transformed into the line passing through the
tangency points.
Reduction of the conic matrix to a diagonal form
Since the matrix of a polarity is symmetric, we can reduce it to a diagonal form.

Let D be the diagonal matrix and B the exchange base matrix. Then:
M = B −1 D B
 d1 0 0
 
D= 0 d2 0
0 0 d 3 

There exist three eigenvectors of the conic matrix. These eigenvectors are also
characteristic vectors of the dual conic matrix with inverse eigenvalues:
M −1 = B −1 D −1 B
 1 
 0 0 
 d1 
 1  1
D −1 = 0 0  d i Vi = M Vi ⇔ M −1 Vi = Vi
 d2  di
 0 1 
 0
 d 3 
The eigenvectors are three points (equal for both punctual and dual planes) that I will call
the eigenpoints of the polarity. Let X be a point on the conic given in Cartesian
coordinates. Then:
XT M X = XT B −1 D B X = VT D V
where V = B X are the coordinates of the point in the base of eigenpoints. If we name with
t, u, v the eigencoordinates, then the equation of the conic is simply:
d1 t 2 + d 2 u 2 + d 3 v 2 = 0
The eigenpoints of a polarity do not never belong to the associated conic since they have
the eigencoordinates (1,0,0), (0,1,0) and (0,0,1) and di ≠ 0. On the other hand, the conic
only exist if the eigenvalues have different signs (one positive and two negative or one
negative and two positive). This fact classifies the polarities in two classes: those having
an associated conic, and those not having an associated conic. In fact, if we state a level, a
certain value of the matrix product, there can be a conic or not. That is, there is a family of
conics for different levels and there is a threshold value, from where the conic does not
exist.
Using a base of points on the conic
If instead of Cartesian coordinates we use the coordinates of a base of three points

A, B and C lying on the conic, the coefficients on the matrix diagonal vanish. Then the
equation of the conic is:
 0 c12 c13   x A 
1   
(x A xB x C )  c12 0 c 23   x B  = c 23 x B x C + c 31 x C x A + c12 x A x B = 0
2 c
 13 c 23 0   x C 
With the same base, the equation of the conic passing through D and whose projective
cross ratio {X, A B C D} is r is:
0 = r x C x A d B + (1 − r ) x A x B d C − x B x C d A
From where we have:
 0 (1 − r ) d C r dB   xA 
  
(x A xB x C )  (1 − r ) d C 0 − d A   xB  = 0
 rd − dA 0   x C 
 B
This conic matrix is not constant, and has the determinant:
det = −2 r (1 − r ) d A d B d C
so the matrix with constant coefficients is:
 0 (1 − r ) d C r dB 
1  
C=  (1 − r ) d C 0 − dA 
3 − r (1 − r ) d d d
C 
A B
 r dB − dA 0 
Exercises
11.1 Prove that the projective cross ratio of any four points on a conic is equal to the ratio
of the sines of the cylindrical half-angles. That is, if:
OA = OQ cos α + OR sin α OB = OQ cos β + OR sin β , etc.

γ −α δ −β
sin sin
2 2
Then { X , A B C D} =
δ −α γ −β
sin sin
2 2
11.2 Prove that an affinity transforms a circle into an ellipse, and using this fact prove
the Newton’s theorem: Given two directions, the ratio of the product of distances from
any point on the plane to both intersections with an ellipse along the first direction and
the same product of distances along the other direction is independent of the location of
this point on the plane.
11.3 Prove that a line touching an ellipse is parallel to the conjugate diameter of the
diameter whose end-point is the tangency point. Prove also that the parallelogram
circumscribed around an ellipse has constant area (Apollonius’ theorem).
11.4 Prove the former theorem applied to a hyperbola: the parallelogram formed by two
conjugate diameters has constant area.
11.5 Given a pole, trace a line passing through the pole and cutting the conic. Prove that
the pole, and the intersections of this line with the conic and with the polar form a
harmonic range.
11.6 Calculate the equation of the point and tangential conic passing through the points
(1, 1), (2, 3), (1, 4), (0, 2) and (2, 5).
11.7 The Pascal’s theorem: Let A, B, C, D, E and F be six distinct points on a proper
conic, Let P be the intersection of the line AE with the line BF, Q the intersection of the
line AD with CF, and R the intersection of the line BD with CE. Show that P, Q and R are
collinear.
11.8 The Brianchon’s theorem. Prove that the diagonals joining opposite vertices of a
hexagon circumscribed around a proper conic are concurrent. Hint: take the dual plane.
THIRD PART: PSEUDO-EUCLIDEAN GEOMETRY
This part is devoted to the hyperbolic plane, where vectors and numbers have a
pseudo Euclidean modulus, that is, a modulus of a Minkowski space. The bidimensional
geometric algebra already includes the Euclidean and pseudo Euclidean planes. In fact,
the geometric algebra does not make any special distinction between both kinds of
planes. On the other hand, the plane geometric algebra can be represented by real 2×2
matrices, which helps us to define some concepts with more precision.
12. MATRIX REPRESENTATION AND HYPERBOLIC NUMBERS
All the associative algebras with neutral elements for the addition and product
can be represented with matrices. The representations can have different dimensions,
but the most interesting is the minimal representation which is an isomorphism.
Rotations and the representation of complex numbers
We have seen that a rotation of a vector v over an angle α is written in geometric

algebra as:
v' = v ( cos α + e12 sin α )
Separating the components:
v1' = v1 cos α − v2 sin α
v2' = v1 sin α + v2 cos α
and writing them in matrix form, we have:
 cos α sinα 
(v1' v 2' ) = (v1 v 2 )  
 − sinα cos α 
Identifying this equation with the first one:
 cos α sinα 
cos α + e12 sinα =  
 − sinα cos α 
we obtain the matrix representation for the complex numbers:
1 0  0 1  a b
1 =   e12 =   z = a + b e12 =  
0 1  − 1 0 − b a
Now we wish to obtain also the matrix representation for vectors. Note that the
matrix form of the rotation of a vector gives us the first row of the vector representation,
which may be completed with a suitable second row:
 v1' v 2'   v1 v 2   cos α sinα 

 =  
 v 2' − v1'   v 2 − v1   − sinα cos α 
from where we arrive to the matrix representation for vectors:
1 0  0 1 v v2 
e1 =   e 2 =   v = v1 e1 + v 2 e 2 =  1 
 0 − 1 1 0  v2 − v1 
These four matrices are a basis of the real matrix space M2×2(R) and therefore the
geometric algebra of the vectorial plane V2 is isomorphic to this matrix space:
Cl (V 2 ) ≡ Cl 2,0 = M 2×2 (R )
The moduli of a complex number and a vector are related with the determinant of the
matrix:
2  a b 2 2 2 v v2 
z = a 2 + b 2 = det   v = v1 + v 2 = − det  1 
− b a  v2 − v1 
Let us think about a heterogeneous element being a sum of a complex and a

vector. Its geometric meaning is unknown for us but we can write it with a matrix:
 a + v1 b + v2 
z + v = a + b e12 + v1 e1 + v 2 e 2 =  
 − b + v2 a − v1 
The determinant of this matrix
det (z + v ) = a 2 + b 2 − v1 − v 2
2 2
reminds us the Pythagorean theorem in a pseudo Euclidean plane.
The subalgebra of the hyperbolic numbers
In the matrix representation of the geometric algebra, we can see that the
diagonal matrices are a subalgebra:
a + b 0 
a + b e1 =  
 0 a − b 
The product of elements of this kind is commutative (and also associative). Because of
this, we may talk about «numbers»:
( a + b e1 ) ( c + d e1 ) = a c + b d + ( a d + b c ) e1
Using the Hamilton's notation of pairs of real numbers, we write:
( a, b ) ( c, d ) = ( a c + b d , a d + b c )
From the determinant we may define their pseudo Euclidean modulus z:
det (a + b e1 ) = a 2 − b 2 = a + b e1
2
The numbers having constant modulus lie on a hyperbola. Because of this, they are
called hyperbolic numbers. Due to the fact that e1 cannot be a privileged direction on the
plane, any other set of elements having the form:
a+ku
with a and k real, and u being a fixed unitary vector, is also a set of hyperbolic numbers.
By defining the conjugate of a hyperbolic number as that number with the
opposite vector component:
( a + b e1 )* = a − b e1
we see that the square of the modulus of a hyperbolic number is equal to the product of
this number by its conjugate:
z2 = z z*
Hence the inverse of a hyperbolic number follows:
z*
z −1 = 2
z
Since the modulus is the square root of the determinant, and the determinant of a matrix
product is equal to the product of determinants, it follows that the modulus of a product
of hyperbolic numbers is equal to the product of moduli:
zt = z t
The elements (1 + e1 ) / 2 and (1 − e1 ) / 2 , which are idempotents, and their multiples have
a null modulus and no inverse. They form two ideals, that is, the product of any
hyperbolic number by a multiple of an idempotent (or any other multiple) yields also a
multiple of the idempotent.
Hyperbolic trigonometry
Let us consider the locus of the points (hyperbolic numbers) located at a constant
distance r from the origin (hyperbolic numbers with constant determinant equal to r2),
which lie on the hyperbola x2 − y2 = r2 (figure 12.1). I shall call r the hyperbolic radius
following the analogy with the circle. The extreme of the radius is a point on the
hyperbola with coordinates (x, y). The arc between the positive X half axis and this point
(x, y) has an oriented length s. On the other hand, the radius, the hyperbola and the X-axis
delimit a sector with an oriented area A. The hyperbolic angle (or argument) ψ is defined
as the quotient of the arc length divided by the radius1:
s Figure 12.1
ψ=
r
It follows from this definition that

the oriented area A is2:
ψ r2 2A
A= ⇒ ψ=
2 r2
The hyperbolic sine, cosine and

tangent are defined as the
following quotients:
y
sinh ψ =
r
x y
coshψ = tghψ =
r x
These definitions yield the three fundamental identities of the hyperbolic trigonometry:
sinh ψ 1
tghψ ≡ cosh 2 ψ − sinh 2ψ ≡ 1 1 − tgh 2 ψ ≡
coshψ cosh 2 ψ
Now we search an explicit expression of the hyperbolic functions in terms of elemental

functions such as polynomials, exponential, logarithm, etc. The differential of the arc
length (being real) is related with the differentials of the coordinates by the pseudo
Pythagorean theorem:
 d (sinh ψ )   d (coshψ ) 
2 2
ds = dy − dx
2 2 2
⇒ 1 =   −  
 dψ   dψ 
That is, we have the following system with one differential equation:
1
Note that x2 − y2 = r2 > 0 while s2 < 0. However I overcome this trouble taking in these
definitions r and s and also the area A as real numbers but being oriented, that is, with sign.
2
The formula of the sector area is obtained by an analogous argument from Archimedes: the
total area is the addition of areas of the infinitesimal triangles with altitude equal to r and base
equal to ds . r being constant, the area of the hyperbola sector is A = r s / 2 . Obviously the
radius is orthogonal to each infinitesimal piece of arc of the hyperbola. This question and the
concept and calculus of areas are studied with more detail in the following chapter.
 1 = cosh 2 ψ − sinh 2ψ

1 = (sinh' ψ ) − (cosh 'ψ )
2 2
whose solution, according to the initial conditions given by the geometric definition
(sinh 0 = 0 and cosh 0 = 1), is:
expψ − exp(− ψ ) expψ + exp(− ψ )

sinh ψ = coshψ =
2 2
Hyperbolic exponential and logarithm
Within the hyperbolic numbers one can define and study functions and develop a
hyperbolic analysis, with the aid of matrix functions. In the matrix algebra, the
exponential function of a matrix A is defined as:
∞
An
exp( A) = ∑
n =0 n!
The matrices which represent the hyperbolic numbers (e1 direction) are diagonal. Then
their exponential matrix has the exponential of each element in the diagonal:
x + y 0   exp( x + y ) 0 
exp( x + y e1 ) = exp  =  
 0 x− y  0 exp( x − y )
By extracting the common factor and introducing the hyperbolic functions we arrive at:
 cosh y + sinh y 0 
exp( x + y e1 ) = exp( x )  
 0 cosh y − sinh y 
which is the analogous of the Euler's identity:
exp( x + y e1 ) = exp( x )(cosh y + e1 sinh y )
From this exponential identity, we can find the logarithm function:
x+ y
log( x 2 − y 2 ) + e1 arg tgh  = log( x 2 − y 2 ) + 1 log
1  y 1 e
log( x + y e1 ) = 
2 x 2 2 x− y 
for –x < y < x . This condition is the set of hyperbolic numbers with positive
determinant, which is called a sector3. The two ideals generated by the idempotents
(1 + e1 ) / 2 and (1 − e1 ) / 2 separate two sectors of hyperbolic numbers, one with
3
According to the relativity theory, the region of the space-time accessible to our knowledge
must fulfil the inequality c2 t2 − x2 ≥ 0, being x the space coordinate, t the time and c the light
celerity.
positive determinant and real modulus and another with negative determinant and
imaginary modulus.
The characteristic property of the exponential function is:
exp( z + t ) = exp( z ) exp( t )
z, t being hyperbolic numbers. Taking exponentials with unity determinant:
exp((ψ + χ ) e1 ) ≡ exp(ψ e1 ) exp(χ e1 )
and splitting the real and vectorial parts, we obtain the addition identities:
cosh(ψ + χ ) ≡ coshψ cosh χ + sinh ψ sinh χ
sinh (ψ + χ ) ≡ sinh ψ cosh χ + coshψ sinh χ
Also through the equality:
exp(nψ e1 ) ≡ ( exp(ψ e1 ) )
n
the analogous of Moivre’s identity is found:
cosh(nψ ) + e1 sinh (nψ ) ≡ (coshψ + e1 sinh ψ )

n
For example, for n = 3 it becomes:
cosh 3ψ = cosh 3 ψ + 3 coshψ sinh 2 ψ
sinh 3ψ = 3 cosh 2 ψ sinhψ + sinh 3 ψ
Polar form, powers and roots of hyperbolic numbers
The exponential allows us to write any hyperbolic number in polar form. For
example:
13 3
z = 13 + 5 e1 z = 13 2 − 5 2 = 12 arg z = arg tgh = log
5 2
 3   3  3 
z = 12 exp e1 log  = 12  cosh log  + e1 sinh  log   = 12 3
 2   2  2  log
2
The product of hyperbolic numbers, like that of complex numbers, is found by the
multiplication of the moduli and the addition of the arguments, while the division is
obtained as the quotient of moduli and difference of arguments. The division is not
possible when the denominator is a multiple of an idempotent (modulus zero).
The power of a hyperbolic number has modulus equal to the power of the
modulus, and argument the addition of arguments. Here the periodicity is absent, and
distinct arguments always correspond to distinct numbers. The roots must be viewed
with more detail. For example, there are four square roots of 13 + 12 e1:
13 + 12 e1 = (3 + 2 e1 ) = (2 + 3 e1 ) = (− 3 − 2 e1 ) = (− 2 − 3 e1 )
2 2 2 2
that is, the equation of second degree z2 −13 −12 e1 = 0 has four solutions. Recently G.
Casanova4 has proved that n2 is the maximum number of hyperbolic solutions
(including real values) of an algebraic equation of nth degree. In the case of the equation
of second degree we can use the classical formula whenever we know to solve the
square roots. For example, let us consider the equation:
z 2 − 5 z + 5 + e1 = 0
whose solutions, according to the second degree formula, are:
5 ± 25 − 4 ⋅ 1 ⋅ (5 + e1 ) 5 ± 5 − 4 e1
z= =
2 ⋅1 2
Now, we must calculate all the square roots of the number 5 − 4 e1 . In order to find
them we obtain its modulus and argument:
−4
5 − 4 e1 = 5 2 − (− 4 ) = 3 arg( 5 − 4 e1 ) = arg tgh
2
= − log 3
5
The square root has half argument and the square root of the modulus:
[ ( )
5 − 4 e1 = 3 − log 3 = 3 cosh − log 3 + e1 sinh − log 3 = 2 − e1
2
( )]
There are four square roots of this number:
2 − e1 , − 2 + e1 , 1 − 2 e1 , − 1 + 2 e1
and so four solutions of the initial equation:
5 + 2 − e1 7 − e1 5 − 2 + e1 3 + e1
z1 = = z2 = =
2 2 2 2
5 + 1 − 2 e1 5 − 1 + 2 e1
z3 = = 3 − e1 z4 = = 2 + e1
2 2
Let us study the second degree equation with real coefficients using the matrix
representation:
4
Gaston Casanova, Advances in Applied Clifford Algebras 9, 2 (1999) 215-219.
a z2 + b z + c = 0 a, b and c real and z = x + e1 y
Dividing the second degree equation by a we obtain the equivalent equation:
b c
z2 + z+ =0
a a
which is fulfilled by the number z, but also by its matrix representation. So, according to
the Hamilton-Cayley theorem, it is the characteristic polynomial of the matrix of z,
whose eigenvalues are the elements on the diagonal x + y, x − y (whenever they are
distinct, that is, z is not real). Since b is the opposite of the sum of eigenvalues, and c is
their product, we have:
z 2 − 2x z + x 2 − y 2 = 0
which gives the equalities:
b c b2 − 4 a c
x=− = x2 − y2 ⇒ y2 =
2a a 4 a2
and the solutions:
− b ± e1 b 2 − 4 a c
z=
2a
The two real solutions must be added to these values, obtaining the four expected.
Anyway, the equation has only solutions if the discriminant is positive. Let us see an
example:
z2 + 3 z + 2 = 0
− 3+1 − 3−1 − 3 +e1 − 3 − e1

z1 = = −1 z2 = = −2 z3 = z4 =
2 2 2 2
On the other hand, we may calculate for instance the cubic root of 14−13 e1. The
modulus and argument of this number are:
 13 
14 − 13 e1 = 14 2 − 13 2 = 27 arg (14 − 13 e1 ) = arg tgh −  = − log 27
 14 
From where it follows that the unique cubic root is:
3 14 − 13 e1 = 3 − log 27 = 3 − log 3 = 2 − e1
3
In order to study with more generality the number of roots, let us consider again
the matrix representation of a hyperbolic number:
a + b 0 
a + b e1 =  
 0 a − b 
Now, it is obvious that there is only a unique root with odd index, which always exists
for every hyperbolic number:
n a + b 0 
n a + b e1 =   n odd
 0
n
a − b 
On the other hand, if a + b > 0 and a − b > 0 (first half-sector) there are four roots with
even index, one on each half-sector (I follow the anticlockwise order as usually):
n even
n a + b 0  n a + b 0  nd
  ∈1st half-sector   ∈2 half-sector
 0 n
a − b   0 − n a − b 
 
− n a + b 0  rd − n a + b 0  th
  ∈3 half-sector   ∈4 half-sector
 0 − a − b 
n  0 n
a − b 
 
If the number belongs to a half-sector different from the first, some of the elements on
the diagonal is negative and there is not any even root. This shows a panorama of the
hyperbolic algebra far from that of the complex numbers.
Hyperbolic analytic functions
Which conditions must fulfil a hyperbolic function f(z) of a hyperbolic variable z

to be analytic? We wish that the derivative be well defined:
f (z + ∆z ) − f (z )
f' (z ) = lim
∆z → 0 ∆z
that is, this limit must be independent of the direction of ∆z. If f(z) = a + b e1 and the
variable z = x + y e1 , then the derivative calculated in the direction ∆z = ∆x is:
∂a ∂b
f' (z ) = + e1
∂x ∂x
while the derivative calculated in the direction ∆z = e1 ∆y becomes:
∂a ∂b
f' (z ) = e1 +
∂y ∂y
Both expressions must be equal, which results in the conditions of hyperbolic

analyticity:
∂a ∂b ∂a ∂b
= and =
∂x ∂y ∂y ∂x
Note that the exponential and logarithm fulfil these conditions and therefore they are
hyperbolic analytic functions. More exactly, the exponential is analytic in all the plane
while the logarithm is analytic for the sector of hyperbolic numbers with positive
determinant.
By derivation of both identities one finds that the analytic functions satisfy the
hyperbolic partial differential equation (also called wave equation):
∂ 2a ∂ 2a ∂ 2b ∂ 2b
− = − =0
∂x 2 ∂y 2 ∂x 2 ∂y 2
Now, we must state the main integral theorem for hyperbolic analytic functions:
if a hyperbolic function is analytic in a certain domain on the hyperbolic plane, then its
integral following a closed way C within this domain is zero. If the hyperbolic function
is f (z ) = a + b e1 then the integral is:
∫ f (z ) dz = ∫ (a + be )(dx + dy e ) = ∫ (a dx + b dy ) + e ∫ (a dy + b dx )
C C
1 1
C
1
C
Since C is a closed path, we may apply the Green theorem to write:
 ∂b ∂a   ∂a ∂b 
= ∫∫  −  dx ∧ dy + e1 ∫∫  −  dx ∧ dy = 0
D 
∂x ∂y  D 
∂x ∂y 
where D is the region bounded by the closed way C. Since f(z) fulfils the analyticity
conditions everywhere within D, the integral vanishes.
From here other theorems follow like for the complex analysis, e. g.: if f(z) is a
hyperbolic analytic function in a domain D simply connected and z1 and z2 are two
points on D then the definite integral:
∫ f (z ) dz
z1
z2
between these points has a unique value independently of the integration path.
Let us see an example. Consider the function f(z) = 1 / ( z − 1 ). The function is
only defined if the inverse of z − 1 exists, which implies z − 1≠0. Of course, this
function is not analytic at z = 1, but neither for the points:
z − 12 = 0 ⇔ ( x − 1)2 − y2 = 0 ⇔ ( x + y − 1) ( x − y − 1) = 0
The lines x + y = 1 and x − y = 1 break the analyticity domain into two sectors. Let us
calculate the integral:
( 5, 3 )
dz
∫
( 5 , −3 )
z −1
through two different trajectories within a sector. The first one is a straight path given
by the parametric equation z = 5 + t e1 (figure 12.2):
( 5, 3 )
dz
3
dt (4 − t e1 ) dt
3
∫
( 5, −3)
z − 1
=∫
−3
4 + t e1
e1 = ∫
−3 16 − t 2
e1
Ought to symmetry the integral of the odd

function is zero:
3 3
4 dt e 4+t 
=∫ e =  1 log = e1 log 7
− 3 16 − t
2 1
2 4 − t  − 3
The second path (figure 12.2) is the hyperbola

going from the point (5, −3) to (5, 3): Figure 12.2
( 5, 3 )
dz (z * −1) dz
( 5, 3 )
(z * −1) dz
( 5, 3 )
∫
( 5, −3)
= ∫ = ∫
z − 1 ( 5, −3) (z − 1) (z * −1) ( 5, −3) det (z ) − 2 Re (z ) + 1
Introducing the parametric equation of this path, z = 4 ( cosh t + e1 sinh t) we have:
log 2
(4 cosh t − 4e1 sinh t − 1) (4sinht + 4 e1 sinh t ) log 2
e1 (4 − cosh t ) − 4 sinh t
= ∫
− log 2
16 − 8 cosh t + 1
dt = 4 ∫
− log 2
17 − 8 cosh t
dt
Due to symmetry, the integral of the hyperbolic sine (an odd function) divided by the
denominator (an even function) is zero. Then we split the integral in two integrals and
find its value:
− 8 exp(t ) + 2 
log 2 log 2 log 2
15 dt dt e1 
= e1 ∫ + e1 ∫ = log + e1 log 2 = e1 log 7
2 − log 2 17 − 8 cosh t − log 2
2 2  − 8 exp(t ) + 32  − log 2
Now we see that the integral following both paths gives the same result, as indicated by
the theorem. In fact, the analytical functions can be integrated directly by using the
indefinite integral:
( 5, 3 )
( )
( 5, 3 )
dz 1  y 
∫ = [ log(z − 1) ] ( 5, −3) =  log ( x − 1) − y 2 + e1 arg tgh
( 5, 3 ) 2

( 5, −3)
z −1 2  x − 1  ( 5, −3)
( 5, 3 )
e x −1+ y 
=  1 log  = e1 log 7
2 x − 1 − y  ( 5, −3)
Analyticity and square of convergence of the power series
A matrix function f(A) can be developed as a Taylor series of powers of the

matrix A:
∞
x + y 0 
f ( A) = ∑ a n A n A = x + y e1 =  
n =0  0 x− y 
The series is convergent when all the eigenvalues of the matrix A are located within the
radius r of convergence of the complex series:
∑a
n =0
n zn < ∞ for |z|<r
which leads us to the following conditions:
|x+y|<r and |x−y|<r
Therefore, the region of convergence of a Taylor series of hyperbolic variable is a

square centred at the origin of coordinates with vertices (r, 0)-(0, r)-(−r, 0)-(0, −r). Note
that from both conditions we obtain:
det ( x + y e1 ) = x 2 − y 2 < r 2 .
which is a condition similar to that for complex numbers, that is, the modulus must be
lower than the radius of convergence. However this condition is not enough to ensure
the convergence of the series. Let us see, for example, the function f(z) = 1/(1 − z):
∞
1
f (z ) = = ∑ zn
1 − z n =0
The radius of convergence of the complex series is r = 1, so that the square of the
convergence of the hyperbolic series is (1, 0)-(0, 1)-(−1,0)-(0,−1). Otherwise, we have
formerly seen that 1/(z −1) (and hence f(z) ) is not analytic at the lines x + y −1 = 0 and x
− y − 1= 0 belonging to the boundary of the square of convergence. Now we find a
phenomenon which also happens within the complex numbers: at some point on the
boundary of the region of convergence of the Taylor series, the function is not analytic.
Another example is the Riemann’s function:
∞ ∞
1
ζ (z ) = ∑ z
= ∑ e − z log n
n =1 n n =1
Now, we take z instead of real being a hyperbolic number:
∞
ζ ( x + y e1 ) = ∑ e − x log n (cosh( y log n ) − e1 sinh( y log n ))
n =1
∞
 n y + n − y n y − n − y  ∞  − x + y 1 − e1 1 + e1 
= ∑ n − x  − e1  = ∑  n + n −x− y 
n =1  2 2  n =1  2 2 
The series is convergent only if −x +y < −1 and −x −y < −1 or:
1−x<y<x−1
which is the positive half-sector beginning at (x, y) = (1, 0). You can see that it is an
analytic function within this domain. We can rewrite the Riemann’s function in the
form:
1 + e1 1 − e1
ζ ( x + y e1 ) = ζ (x + y ) + ζ (x − y )
2 2
and then define the Riemann’s zeta function extended to the other sector taking into
account the complex analytic continuation when needed for ζ(x + y) or ζ(x − y):
∞
2 π z 1 t z −1
ζ (1 − z ) = ζ (z ) Γ(z ) cos and ζ (z ) = 0 < Re(z ) < 1
Γ(z ) ∫0 e t − 1
dt
(2 π ) z 2
Now a very interesting theorem arises: every analytic function f(z) can be always
written in the following form:
1 + e1 1 − e1
f ( x + y e1 ) = f (x + y ) + f (x − y )
2 2
Let us prove this statement calculating the partial derivatives of each part a and b of the
analytic function:
∂a 1  df ( x + y ) df ( x − y ) ∂b 1  df ( x + y ) df ( x − y )
= + = −
∂x 2  d ( x + y ) d ( x − y )  ∂x 2  d ( x + y ) d ( x − y ) 
∂a 1  df ( x + y ) df ( x − y ) ∂b 1  df ( x + y ) df ( x − y ) 
= − = +
∂y 2  d ( x + y ) d ( x − y )  ∂y 2  d ( x + y ) d ( x − y ) 
These derivatives fulfil the analyticity condition provided that the derivatives of the real
function at x + y and x − y exist. On the other hand, this expression is a method to get
the analytical continuation of any real function. For example, let us construct the
analytical continuation of f(x) = cos x :
1 + e1 1 − e1
cos( x + y e 1 ) = cos( x + y ) + cos( x − y )
2 2
1 e
= [ cos( x + y ) + cos( x − y ) ] + 1 [ cos( x + y ) − cos( x − y ) ]
2 2
= cos x cos y − e1 sin x sin y

The other consequence of this way of analytic continuation is the fact that if f(z)
loses analyticity for the real value x0 then it is neither analytic at the lines with slope 1
and −1 passing through (x0, 0). Since we have supposed that the function is analytic
except for a certain real value, the former equality holds although the analyticity is lost
at this point. But then the function cannot be analytic at (x0 + t, − t ) nor (x0 − t, t )
because:
1 + e1 1 − e1
f ( x 0 + t − t e1 ) = f (x 0 + 2 t ) + f (x 0 )
2 2
1 + e1 1 − e1
f ( x 0 − t + t e1 ) = f (x 0 ) + f (x 0 − 2 t )
2 2
For example, look at the function f(z) = 1/z, which is not analytic for the real value x =
0. Then it will not be analytic at the lines y = x and y = −x. We can see it through the
decomposition of the function:
1 1 x − y e1 1 + e1 1 1 − e1 1
= = 2 = +
z x + y e1 x − y 2
2 x+ y 2 x−y
Now we may return to the question of the convergence of the Taylor series. A
function can be only developed in a power series in the neighbourhood of a point where
it is analytic. The series is convergent till where the function breaks the analyticity, so
that the lines y = x − x0 and y = −x + x0 (being x0 the real value for which f(x) is not
analytic) are boundaries of the convergence domain. On the other hand, due to the
symmetry of the powers of z, the convergence domain forms a square around the centre
of development. This implies that there are not multiply connected domains. The
Gruyère cheese picture of a complex domain is not possible within the hyperbolic
numbers, because if the function is not analytic at a certain point, then it is not analytic
for all the points lying on a cross which passes through this point. The hyperbolic
domains are multiply separated, in other words, formed by disjoined rectangular regions
without holes.
About the isomorphism of Clifford algebras
Until now, I have only used the geometric algebra generated by the Euclidean
plane vectors (usually noted as Cl2,0(R)). This algebra already contains the hyperbolic
numbers and hyperbolic vectors. However an isomorphic algebra Cl1,1(R) generated by
hyperbolic vectors (time-space) is more used in relativity:
2
e0 = 1
2
e1 = −1 e0 e1 = − e1 e0 (e0 e1 )2 = 1
The isomorphism is:
Cl2,0(R) ↔ Cl1,1(R)
1 ↔ 1
e1 ↔ e01
e2 ↔ e0
−e12 ↔ e1
A hyperbolic number has the expression z = x + y e01 and a hyperbolic vector v = v0 e0 +

v1 e1.
This description perhaps could be satisfactory for a mathematician or a physicist,
but not for me. I think that in fact both algebras only differ in the notation but not in
their nature, so that they are the same algebra. Moreover, they are equal to the matrix
algebra, which is the expected algebra for a space of multiple quantities (usually said
vectors). The plane geometric algebra is the algebra of the 2×2 real matrices5.
Cl2,0(R) = Cl1,1(R) = M2×2(R)
This distinct notation only expresses the fact that the plane geometric algebra is equally
generated (in the Grassmann’s sense) by Euclidean vectors or hyperbolic vectors. The
vector plane (Euclidean or hyperbolic) is the quotient space of the geometric algebra
divided by an even subalgebra (complex or hyperbolic numbers). Just this is the matter
of the next chapter.
Exercises
12.1 Calculate the square roots of 5 + 4 e1.
12.2 Solve the following equation: 2 z 2 + 3 z − 17 + 3 e1 = 0
12.3 Solve directly the equation: z2 − 6 z + 5 = 0
12.4 Find the analytical continuation of the function f(x) = sin x .
12.5 Find cosh 4ψ and sinh 4ψ as functions of coshψ and sinhψ.
12.6 Construct the analytical continuation of the real logarithm and see that it is
identical to the logarithm found from the hyperbolic exponential function.
e1
∫z dz following a straight path z = t e1 and a circular path z

2
12.7 Calculate the integral
− e1
= cos t + e1 sin t, and see that the result is identical to the integration via the primitive.
12.8 Prove that if f(z) is a hyperbolic analytic function and does not vanish then it is a
hyperbolic conformal mapping.
5
In comparison, the 2×2 complex matrices (Pauli matrices) are a representation of the algebra
of the tridimensional space.
13. THE HYPERBOLIC OR PSEUDO-EUCLIDEAN PLANE
Hyperbolic vectors
The elements not belonging to the subalgebra of the hyperbolic numbers form a
quotient space1, but not an algebra:
 0 v 2 − v 21 
v = v 2 e 2 + v 21 e 21 =  
 v 2 + v 21 0 
In other words, their linear combinations are in the space, but the products are
hyperbolic numbers. These elements play the same role as the Euclidean vectors with
respect to the complex numbers. Because of this, they are called hyperbolic vectors.
Like the hyperbolic numbers, the hyperbolic vectors have also a pseudo-Euclidean
determinant:
det (v 2 e 2 + v 21 e 21 ) = v 21 − v 2
2 2
Following the analogy with Euclidean vectors:
det (w1 e1 + w2 e 2 ) = −(w1 e1 + w2 e 2 ) = − w

2 2
the determinant of the hyperbolic vectors is also equal to its square with opposite sign:
det (v 2 e 2 + v 21 e 21 ) = −(v 2 e 2 + v 21 e 21 ) = − v
2 2
whence the modulus of a hyperbolic vector can be defined:
v = v 22 − v 21
2
Like for Euclidean vectors, the inverse of a hyperbolic vector is equal to this vector
divided by its square (or square of the modulus):
v 2 e 2 + v 21 e 21
(v 2 e2 + v 21 e21 )−1 = 2 2
v 2 − v 21
As before, there is not any privileged direction in the plane. Then every subspace
whose elements have the form:
1
Every Euclidean vector is obtained from e1 through a rotation and dilation given by the
complex number with the same components: v1 e1 + v2 e2 = e1 (v1 + v2 e12 ) so that the complex
and vector planes are equivalent. This statement also holds in hyperbolic geometry: every
hyperbolic vector is obtained from e2 through a rotation and dilation given by the hyperbolic
number with the same components: v2 e2 + v21 e21 = e2 (v2 + v21 e1 ) or in relativistic notation: v0
e0 + v1 e1 = e0 (v0 + v1 e01 ) so that both hyperbolic planes are equivalent. That reader accustomed
to the relativistic notation should remember the isomorphism: e0 ↔ e2 , e1 ↔ e21 , e01 ↔ e1 .
vw w + v21 e21
where w is a unitary Euclidean vector perpendicular to the unitary Euclidean vector u, is

also a subspace of hyperbolic vectors complementary of the hyperbolic numbers with
the direction of u.
The hyperbolic vectors have the following properties, which you may prove:
1) The product of two hyperbolic vectors is always a hyperbolic number. This
fact is shown by the following table, where the products of all the elements
of the geometric algebra are summarised:
× hyp. numbers hyp. vectors

hyp. numbers hyp. numbers hyp. vectors
hyp. vectors hyp. vectors hyp. numbers
2) The conjugate of a product of two hyperbolic vectors x and y is equal to the

product with these vectors exchanged:
( x y )* = y x
3) The product of three hyperbolic vectors x, y and z fulfils the permutative

property:
xyz=zyx
When a hyperbolic number n and a hyperbolic vector x are exchanged, the

hyperbolic number becomes conjugated according to the permutative
property:
n x = x n*
4) The modulus of a product of hyperbolic vectors is equal to the product of

moduli:
vw = v w
This property follows immediately from the fact that the determinant of a
product of matrices is equal to the product of determinants of each matrix.
Inner and outer products of hyperbolic vectors
For any two hyperbolic vectors r and s such as:
r = r2 e 2 + r21 e 21 s = s 2 e 2 + s 21 e 21
their inner and outer products are defined by means of the geometric (or matrix) product
in the following way:
rs + sr
r·s= = r2 s 2 − r21 s 21
2
rs − sr
r∧s= = ( r2 s 21 − r21 s 2 ) e1
2
Then the product of two vectors can be written as the addition of both products:
rs=r·s+r∧s
If the outer product of two hyperbolic vectors vanishes, they are proportional:
r2 r12
r∧s=0 ⇔ = ⇔ r || s
s 2 s12
Two hyperbolic vectors are said to be orthogonal if their inner product vanishes:
r2 s 21
r⊥s ⇔ r·s=0 ⇔ r2 s 2 − r21 s 21 = 0 ⇔ =
r21 s 2
When the last condition is fulfilled, we see both vectors being symmetric with respect to
any quadrant bisector. When the ratio of components is equal to ±1 (directions of
quadrant bisectors), the vectors have null modulus and are self-orthogonal. For ratios
differing from ±1, one vector has negative determinant and the other positive
determinant, so that they belong to different sectors. That is, there is not any pair of
orthogonal vectors within the same sector.
Angles between hyperbolic vectors
The oriented angle α between two Euclidean vectors u and v (of the form x e1 +
y e2) is obtained from the complex exponential function:
 u·v
uv  cos α = u v u ∧ v e 21
exp(α e12 ) = ⇔  ⇔ α = arctg + (π e12 )
u v u ∧ v e 21 u· v
sinα =
 u v
Then, the oriented hyperbolic angle ψ between two hyperbolic vectors u and v (of the
form x e2 + y e21) is defined as:
 u·v
 coshψ =
uv  u v
exp(ψ e1 ) = ⇔  ⇔
u v u ∧ v e1
sinhψ =
 u v
 u ∧ v e1  1  u · v + u ∧ v e1 
ψ = arg tgh  + (π e12 ) = log  + (π e12 )
 u·v  2  u · v − u ∧ v e1 
The parenthesis for π e12 indicates that this angle is added to the arctangent or not
depending on the quadrant (Euclidean plane) or half sector (hyperbolic plane).
Let us analyse whether these expressions are suitable. The hyperbolic cosine is
always greater or equal to 1. It cannot describe the angle between two vectors belonging
to opposite half sectors, which have a negative inner product. So we must keep the
complex analytic continuation of the hyperbolic sine and cosine:
sinh( x + e12 y ) ≡ sinh x cos y + e12 cosh x sin y
cosh( x + e12 y ) ≡ cosh x cos y + e12 sinh x sin y
to get sinh(ψ + e12 π ) ≡ −sinhψ and cosh(ψ + e12 π ) ≡ − coshψ with ψ real.
2
Therefore opposite hyperbolic vectors form a circular angle of π radians .
Let us consider the angle between vectors lying on different sectors. The
modulus of one vector is real while that of the other is imaginary what implies that the
hyperbolic sine and cosine of the angle are imaginary. Using again the complex analytic
continuation of the hyperbolic sine and cosine we find these imaginary values:
 π
sinhψ ± e12  ≡ ± e12 coshψ ψ real
 2
 π
coshψ ± e12  ≡ ± e12 sinhψ
 2
Then, which is the angle between two orthogonal vectors? Since they have a null
inner product while their outer product
equals in real value the product of both
moduli and the hyperbolic cosine is ±e12 ,
they form an angle of π e12 /2 or 3π e12 /2,
that is, orthogonal hyperbolic vectors form a
circular right angle:
1 π
ψ ⊥ = log(− 1) + (π e12 ) = e12 + (π e12 )
2 2
The analytic continuations of the

hyperbolic trigonometric functions must be
consistent with their definitions given in the Figure 13.1
page 142. For example, in the lowest half
2
The so called “antimatter” is not really any special kind of particles, but matter with the
energy-momentum vector (E, p) lying on the negative half sector instead of the positive one
(usual matter). Also we may wonder whether matter having (E, p) on the imaginary sector exists
since, from a geometric point of view, there is not any obstacle. It is known that the particle
dynamics for c<V<∞ is completely symmetric to the dynamics for 0≤V<c, but with an
imaginary mass. In this case, the head of the energy-momentum vector is always located on the
hyperbola having the y-axis (p axis) as principal axis of symmetry. According to Einstein’s
formula, for V = ∞ we have E = 0 and p = m c, that is, the particle has a null energy! The
technical question is how can a particle trespass the light barrier?
sector, the modulus of the hyperbolic radius is imaginary. The abscissa x and ordinate y
divided by r, the modulus of the radius vector, is also imaginary so the angle is a real
quantity plus a right angle and the hyperbolic sine and cosine are exchanged. The values
of the angles consistent with the analytic continuation are displayed in the figure 13.1.
Congruence of segments and angles
Two segments (vectors) are congruent (have equal length) if their determinants
are equal, in other words, when one segment can be obtained from the other through an
isometry. Segments having equal length always lie on the same sector.
Two hyperbolic angles are said to be congruent (or equal) if they have the same
moduli, that is, if they intercept arcs of hyperbola with the same hyperbolic length.
A triangle is said to be isosceles if it has two congruent sides. In colloquial
language we talk about “sides with equal length”, even “equal sides”. Let us see the
isosceles triangle theorem: the angles adjacent to the base of an isosceles triangle are
equal. The proof uses the outer product. The triangle area is the half of the outer product
of any two sides, as proved below in the section on the area. Suppose that the sides a
and b have the same modulus. Then:
a c sinh β = b c sinh α ⇒ β = α
Isometries
In the Euclidean plane, an isometry is a geometric transformation that preserves

the modulus of vectors and complex numbers. Now I give a more general definition: a
geometric transformation is an isometry if it preserves the determinant of every element
of the geometric algebra. In fact, the isometries are the inner automorphism of matrices:
A' = B −1 A B ⇒ det A' = det A
When B represents a Euclidean vector, the

isometry is a reflection in the direction of B. If
a complex number of argument α is
represented by B, the transformation is a
rotation of angle 2α .
We wish now to obtain the isometries
for hyperbolic vectors. The hyperbolic rotation
of a hyperbolic vector (figure 13.2) is obtained
through the product by an exponential having
unity determinant:
Figure 13.2
v' 2 e 2 + v' 21 e 21 = ( v 2 e 2 + v 21 e 21 )(coshψ + sinhψ e1 )
Writing the components, we have:
 v' 2 = v 2 coshψ + v 21 sinhψ


v' 21 = v 2 sinhψ + v 21 coshψ
which is the Lorentz transformation3 of the relativity. When a vector is turned through a
positive angle ψ, its extreme follows the hyperbola in the direction shown by the
arrowheads in the figure 13.2, as is deduced from the components. So, they indicate the
geometric positive sense of hyperbolic angles (a not trivial question). Since the points of
intersections with the axis corresponds to ψ = 0 plus multiples of π e12 /2, the signs of
the hyperbolic angles are determined: positive in the first and third quadrant, and
negative in the second and fourth quadrant.
We can also write the hyperbolic rotation as an inner automorphism of matrices
by using the half argument identity for hyperbolic trigonometric functions:
2
 ψ ψ
coshψ + e1 sinhψ ≡  cosh + e1 sinh 
 2 2
and the permutative property:
 ψ ψ   ψ ψ 
v' 2 e 2 + v' 21 e 21 =  cosh − sinh e1  ( v 2 e 2 + v 21 e 21 )  cosh + sinh e1 
 2 2   2 2 
If z = exp(e1ψ /2), then the rotation is written as:
v' = z −1 v z
Now it is not needed that z be unitary and can have any modulus. Also, we see that the
hyperbolic rotation leaves the hyperbolic numbers invariant.
Analogously to reflections in the Euclidean plane, the hyperbolic reflection of a
hyperbolic vector v with respect to a direction given by the vector u is:
v' = u −1 v u = u −1 ( v|| + v⊥) u = v|| − v⊥
because the proportional vectors commute and those which are orthogonal anticommute
(also in the hyperbolic case,
because the inner product is Figure 13.3
zero). Every pair of orthogonal
directions (for example u|| and u⊥
in figure 13.3) are always seen
by us as having symmetry with
respect to the quadrant bisectors.
Since the determinant is
preserved, a vector and its
reflection always belong to the
same sector, the hyperbolic angle
between each vector and the
direction of reflection being
equal with opposite sign.
3
The hyperbolic vector in the space-time is ct e0 + x e1 where x is the space coordinate, t the
time and c the light celerity. The argument of the hyperbolic rotation ψ is related with the
relative velocity V of both frames of reference through ψ = arg tgh V/c.
Finally, note that the extremes of a vector and its transformed vector under a
hyperbolic reflection (or any isometry) lie on an equilateral hyperbola.
Theorems about angles
The sum of the oriented angles of a hyperbolic triangle is minus two right
circular angles. To prove this theorem, draw a line parallel to a side and apply the Z-
theorem (figure 13.4): they will form
an angle between opposite directions,
which has the value −π e12. The minus Figure 13.4
sign is caused by the fact that,
following the positive orientation of
angles, a right angle is subtracted each
time an asymptote is trespassed.
Look at the figure 13.5 where
an angle inscribed in an equilateral
hyperbola is drawn. A diameter
divides the inscribed angle into the
angles α and β. Then we draw two
radius from the origin to both extremes of the angle. All the radius have the same
length, so that the upper and lowest triangle are isosceles and have two equal angles.
Since the sum of the angles of each hyperbolic triangle is −π e12, the third angle is
−π e12−2α and −π e12−2β respectively. So the supplementary angles are 2α and 2β ,
and hence the theorem follows: the
inscribed angle on a hyperbola x2 − y2 =
r2 is the half of the central angle
(intercepted arc) so that it is constant
and independent of the location of its
vertex.
For example, take the points (5,
3) and (5, −3) as extremes of an
inscribed angle with the vertex placed
at the negative branch, for example at
(−4, 0) or (−5, −3). Even you can take
the vertex at the positive branch, e. g. at Figure 13.5
(4, 0). Anyway its value is log 2. Now
calculate the central angle or
intercepted arc (with vertex at the origin) and see that its value is log 4, twice the
inscribed angle.
Distance between points
The distance between two points on the hyperbolic plane is defined as the
modulus of the vector going from one point to the another:
d ( P, Q ) = PQ = (x Q − x P ) − (yQ − y P )
2 2
In the hyperbolic plane, the distance is a real or imaginary positive number (depending
on the sector), which has the following properties:
1) d (P, Q ) = d (Q , P )
2) If the vectors PQ, QR and PR lie on the positive half sector then they fulfil
the triangular inequality:
d (P, R ) ≥ d (P, Q ) + d (Q, R )
The prove is obtained by means of the inner product:
PR 2 = (PQ + QR ) = PQ 2 + QR 2 + 2 PQ · QR
2
= PQ 2 + QR 2 + 2 PQ QR coshψ ≥ PQ 2 + QR 2 + 2 PQ QR
The extraction of the square root yields the triangular inequality:
PR ≥ PQ + QR
So the sum of two sides of a triangle is smaller than the third side whenever they are
taken in the positive half sector. For example, let us apply this statement to the triangle
with vertices A = (5, 3), B = (1, 0) and C = (10, 1). Taking the modulus of the sides
positive, we have:
BC = 80 BA = 7 AC = 21
BC ≥ BA + AC ⇒ 80 ≥ 7 + 21
This result must be commented with more detail. Firstly, we see that the straight
path has the highest length among the possible paths between two given points. In the
Minkowski’s space-time this fact causes the twin paradox. Since the pseudo-Euclidean
length of the path is the proper time for each person, the brother who followed the
straight path -going in an inertial frame- has aged more than the brother who followed
another path -subjected to accelerations-.
Area on the hyperbolic plane
Now I give a general geometric definition of area valid for both Euclidean and
hyperbolic planes. If a parallelogram has orthogonal sides, then the modulus of its area
is equal to the product of the lengths of the orthogonal sides. In Euclidean geometry a
parallelogram with orthogonal sides is called a rectangle. In the hyperbolic plane, two
sides are orthogonal if their directions are seen by us as symmetric with respect to the
direction of the quadrant bisector. So, I have preferred to avoid the word rectangle in
this case, while the term parallelogram is still valid within this context.
Which is the suitable algebraic expression to calculate the area of any

parallelogram? Suppose that it is the outer product. If a and b are the sides of the
parallelogram then:
A = a ∧ b = a b sinhψ (a , b )
Let us see the consistence of this expression. First at all, recall that the angle
between orthogonal hyperbolic vectors is a right circular angle:
π
ψ ⊥ = ± e12
2
so that the area of a parallelogram with orthogonal sides is the product of their lengths
as expected:
 π  π
A = a b sinh ± e12  = a b sin = a b
 2  2
For any parallelogram not having orthogonal sides, the area is the product of the
base (one side) for the altitude (the projection of the other side onto the direction
orthogonal to the base) and this product is only given by the outer product because its
anticommutativity removes the proportional projection:
A = a ∧ b = (a || + a ⊥ ) ∧ b = a ⊥ b
The expression for the area in Cartesian

components is equal to that for the
Euclidean plane. It means that we can
calculate areas graphically in the usual way,
fact that has allowed till now to define the
hyperbolic trigonometric functions from the
area scanned by the hyperbola radius in the
Euclidean plane, in spite of not being their
proper plane. However, as I have explained
in the footnote at the page 142, we must Figure 13.6
perceive the pseudo-Euclidean nature of the
area in the hyperbolic plane. The figure 13.6
displays how are the radii perpendicular to the hyperbola because the bisector of the
radius and the tangent vector is parallel to the quadrant bisector. Also it may be proved
analytically by means of differentiation of the equation x2 − y2 = r2:
2 x dx − 2 y dy = 0 ⇔ r · dr = 0
showing that the radius vector and its differential are orthogonal for the hyperbola with
constant radius. On the other hand, observe that the triangles drawn in the figure 13.6
are isosceles and have two equal angles approaching the right angle value at the limit of
null area.
Diameters of the hyperbola and Apollonius’ theorem
At the page 129, I postponed an interpretation of the central equation for the
hyperbola analogous to the cylindrical angles for the ellipse because it is properly of
pseudo-Euclidean nature. Here I develop this interpretation.
If P is any point on a hyperbola with major and minor half-axis OQ and OS
(figure 11.17 reproduced below) then its central equation is:
sinh χ
OQ cosh χ + OS
e2 − 1
OP = ±
sinh 2 χ
cosh 2 χ −
e2 − 1
Since OQ and OS are orthogonal, the square

of this equation fulfils:
OP 2 OP 2
cosh 2
χ + sinh 2 χ = 1
Figure 11.17
OQ 2 OS 2
Both half-axis are related with the

eccentricity through:
OS 2 = (1 − e 2 ) OQ 2
Note that OS2<0 since it is the square of an ordinate on the hyperbolic plane.
Introducing the hyperbolic angle ψ in the following way:
cosh χ
coshψ = Figure 13.7
sinh 2 χ
cosh 2 χ −
e2 − 1
now the equation of the hyperbola

becomes:
OP = ± (OQ coshψ + OS sinh ψ )
Observe in the figure 13.7 that

any hyperbola can be obtained as an
intersection of a transverse plane with the
equilateral hyperbolic prism. The plane
of the acute hyperbola (with 1 < e < 2 )
forms an angle φ with respect to the horizontal, and hence:
cosh χ cos φ
coshψ =
cosh χ cos 2 φ − sinh 2 χ
2
So the eccentricity e of the hyperbola is related with the obliquity φ of the transverse
plane through the relationship:
cos φ = e 2 − 1 with 1< e < 2
The Apollonius’ conjugate diameters of any hyperbola are the intersections of

the transverse plane with a pair of orthogonal axial planes; in other words, two radii are
conjugate (figure 11.17) if their projections onto the horizontal plane are turned through
the same hyperbolic angle ϕ :
OQ' = OQ cosh ϕ + OS sinhϕ
OS' = OQ sinhϕ + OS cosh ϕ
Our Euclidean eyes see the horizontal projections as symmetric lines with respect to the
quadrant bisector. However, they are actually orthogonal because:
OQ' 2 − OS' 2 = OQ2 − OS2
and, therefore, can be taken as a new system of orthogonal coordinates. Even we can
draw a new picture with the new diameters on the Cartesian axis.
The central equation of the hyperbola using the rotated axis is:
OP = ± (OQ' cosh(ψ − ϕ ) + OS' sinh(ψ − ϕ ))
which shows that a hyperbolic rotation of the coordinate axis has been made with respect
to the principal diameter of the hyperbola.
The law of sines and cosines
Since the modulus of the area is identical on the Euclidean and hyperbolic
planes, a parallelogram is divided by its diagonal in two triangles of equal area. This
statement is somewhat subtle since the Euclidean congruence of triangles is not valid in
the hyperbolic plane. I shall return to this question later. Now we only need to know
that the area of a hyperbolic triangle is the half of the outer product of any two sides.
Following the perimeter of a triangle, let a, b, and c be its sides respectively
opposite to the angles α, β and γ. Then the angles formed by the oriented sides are
supplementary of the angles of the triangle and:
a ∧b=b∧c=c∧a ⇒ − a b sinh γ = − b c sinh α = − c a sinh β
a b c
= =
sinh α sinh β sinh γ
which is the law of sines.

From a + b + c = 0 , we have:
a 2 = (− b − c ) = b 2 + c 2 + 2 b · c a 2 = b 2 + c 2 − 2 b c cosh α
2
⇒
which is the law of cosines. And also:
b 2 = a 2 + c 2 − 2 a c cosh β
Figure 13.8
c 2 = a 2 + b 2 − 2 a b cosh γ
When applying both theorems, we must

take care with the sides having
imaginary length and the signs of the
angles and trigonometric functions.
As an application of the law of
sines and cosines, consider the
hyperbolic triangle with vertices A=(5,
3), B=(1, 0), C=(10, 1), whose sides
belong to the real sector (figure 13.8):
AB = B − A = −4 e 2 − 3 e 21 AB = (− 4)2 − (− 3)2 = 7
BC = C − B = 9 e 2 + e 21 BC = 9 2 − 12 = 80
CA = A − C = −5 e 2 + 2 e 21 CA = (− 5)2 − 2 2 = 21
26
BC 2 = CA 2 + AB 2 − 2 CA AB cosh α cosh α = −
7 3
33
CA 2 = AB 2 + BC 2 − 2 AB BC cosh β cosh β =
4 35
47
AB 2 = BC 2 + CA 2 − 2 BC CA cosh γ cosh γ =
4 105
α = −1.3966... − π e12 β = 0.8614... γ = 0.5352...
I have obtained their signs from the definition of the angles α = BAC, β = CBA, γ = ACB
and the geometric plot (figure 13.8). Note that α + β + γ = −π e12 and they fulfil the law
of sines:
BC CA AB
= =
sinh α sinh β sinh γ
because sinhα = sinh(−1.3966+ πe12) = −sinh(−1.3966) = sinh 1.3966.

Consider now another triangle A = (2, 4), B = (1,0) and C = (6, 1), having sides
on real and imaginary sectors (figure 13.9):
AB = B − A = − e 2 − 4 e 21
Figura 13.9
AB = (− 1) 2
− (− 4 ) = 15 e12
2
BC = C − B = 5 e 2 + e 21
BC = 5 2 − 12 = 24
CA = A − C = −4 e 2 + 3 e 21
CA = (− 4 )2 −32 = 7
16 e12
BC 2 = CA 2 + AB 2 − 2 CA AB cosh α cosh α =
105
− e12
CA 2 = AB 2 + BC 2 − 2 AB BC cosh β cosh β =
6 10
23
AB 2 = BC 2 + CA 2 − 2 BC CA cosh γ cosh γ =
2 42
From the last hyperbolic cosine, which is real, we find the hyperbolic sine for γ, which
is a positive angle as shown by the plot:
19
sinh γ =
2 42
which we may use in the law of sines:
24 7 15 e12 19 e12 19 e12

= = ⇒ sinh α = − sinh β = −
sinh α sinh β 19 / 2 42 105 6 10
Recalling that for ψ real:
 π  π
sinhψ ± e12  ≡ ± e12 coshψ coshψ ± e12  ≡ ± e12 sinhψ
 2  2
the angles α, β and γ follow:

π π
α = −1.2284... − e12 β = 0.0527... − e12 γ = 1.1757
2 2
Observe that the addition of the three angles is −π e12, as expected. According to the
definition α = BAC, β = CBA, γ = ACB, let you see the consistence with the geometric
plot. The angle γ has positive sign as shown by the figure 13.9. The bisector of α
parallel to the quadrant bisector divides it in two angles, one real and the other complex
(with imaginary part π e12 /2). The algebraic addition of both angles is α.. Taking into
account that the second has opposite orientation and must be subtracted and
predominates over the first, the negative value of α is explained.
Hyperbolic similarity
Two triangles ABC and A'B'C' are said to be directly similar4 and their vertices
and sides denoted with the same letters are homologous if:
AB BC −1 = A'B' B'C' −1 ⇒ AB−1 A'B' = BC −1 B'C'
One can prove easily that the third quotient of homologous sides also coincides
with the other quotients:
AB−1 A'B' = BC −1 B'C' = CA−1 C'A' = r
The similarity ratio r is defined as the quotient of every pair of homologous

sides, which is a hyperbolic number. The modulus of the similarity ratio is the size ratio
and the argument is the angle of rotation of the triangle A'B'C' with respect to the
triangle ABC.
A' B'
r= exp[α ( AB, A' B ' ) e1 ]
AB
The definition of similarity is generalised to any pair of polygons in the

following way. The polygons ABC...Z and A'B'C'...Z' are said to be directly similar with
similarity ratio r and the sides denoted with the same letters to be homologous if:
r = AB −1 A'B' = BC −1 B'C' = CD −1 C'D' = ... = YZ −1 Y'Z' = ZA −1 Z'A'
Here also, the modulus of r is the size ratio of both polygons and the argument is
the angle of rotation. The fact that the homologous exterior and interior angles are equal
for directly similar polygons is trivial because:
−1
B'A' B'C' = BA BC −1 ⇒ angle A'B'C' = angle ABC
C'B' C'D' −1 = CB CD −1 ⇒ angle B'C'D' = angle BCD etc.
4
A direct similarity is also called a similitude and an opposite similarity sometimes an
antisimilitude.
The direct similarity is an equivalence relation since it has the reflexive,

symmetric and transitive properties. This means that there are classes of equivalence
with directly similar figures.
A similitude with r=1 is a displacement, since both polygons have the same
size and orientation.
Two triangles ABC and A'B'C' are oppositely similar and the sides denoted with
the same letters are homologous if:
AB BC −1 = ( A'B' B'C' −1 )* = B'C' −1 A'B'
where the asterisk denotes the hyperbolic conjugate. The former equality cannot be
arranged into quotients of pairs of homologous sides as done before. Because of this,
the similarity ratio cannot be defined for the opposite similarity but only the size ratio,
which is the quotient of the lengths of any two homologous sides. An opposite
similarity is always the composition of a reflection in any line and a direct similarity.
AB BC −1 = v −1 A'B' B'C' −1 v ⇔ BC −1 v −1 B'C' = AB −1 v −1 A'B' ⇔
BC −1 ( v −1 B'C' v ) = AB −1 ( v −1 A'B' −1 v ) = r
where r is the ratio of a direct similarity whose argument is not defined but depends on
the direction vector v of the reflection axis. Notwithstanding, this expression allows to
define the opposite similarity of two polygons. So two polygons ABC...Z and A'B'C'...Z'
are oppositely similar and the sides denoted with the same letters are homologous if for
any hyperbolic vector v the following equalities are fulfilled:
AB −1 ( v −1 A'B' −1
v ) = BC −1
( v −1 B'C' v ) = .... = ZA −1 ( v −1 Z'A' v )
that is, if after a reflection one polygon is directly similar to the other. The opposite
similarity is not reflexive nor transitive and there are not classes of oppositely similar
figures.
An opposite similarity with r=1 is called a reversal, since both polygons have
the same size but opposite orientations.
Obviously the plot of similar figures on the hyperbolic plane breaks our
Euclidean intuition about figures with the same form. The algebraic definition is very
rigorous and clear, but we must change our visual illusions. So I recommend the
exercise 13.6.
This chapter is necessarily unfinished because the geometry of the hyperbolic
plane can be developed and studied with the same extension and profundity as for the
Euclidean plane. Then, it is obvious that many theorems should follow in the same way
as Euclidean geometry. As a last example I will define the power of a point with respect
to a hyperbola.
Power of a point with respect to a hyperbola with constant radius
The locus of the points placed at a fixed distance r from a given point O is an
equilateral hyperbola centred at this point with equation:
OP = r ( x − x O )2 − ( y − y O )2 = r 2
The power of a point with respect to this Figure 13.10

hyperbola is the product of both oriented
distances from P to the intersections R and R'
of a line passing through P with the hyperbola.
The power of a point is constant for every line
of the pencil of lines of P. In the proof the
inscribed angle theorem is used (figure 13.10):
the angles S'RR' and S'SR' are equal so the
triangles SPR' and RPS' are oppositely similar.
Then:
PR' PS −1 = PR −1 PS' ⇒ PR PR' = PS' PS
Developing the product of distances on the line

passing through the centre of the hyperbola we find:
PS PS' = (PO + OS ) (PO + OS' ) = PO 2 + OS OS' = ( x P − x O ) − ( y P − y O ) − r 2

2 2
that is, the power of a point is obtained by substitution of the coordinates of P on the
hyperbola equation.
Exercises
13.1 Let A = (2, 2), B = (1, 0) and C = (5, 3) be the vertices of a hyperbolic triangle.
Calculate all the sides and angles and also the area.
13.2 Turn the vector 2 e2 + e21 through an angle ψ = log 2. Do a reflection in the
direction 3 e2 − e21 . Make also an inversion with radius 3.
13.3 Find the direction and normal vectors of the line y = 2 x + 1 and calculate the
perpendicular line passing through the point (3, 1).
13.4 Calculate the power of the point P = (−7, 3) with respect to the hyperbola x2 − y2 =
16 using the intersections of the lines y = 3, y = −3x −18 and that passing through the
centre of the hyperbola y = −3x / 7. See that in all cases the power of P is equal to the
value found by substitution in the Cartesian equation.
13.5 In this chapter I have deduced the law of sines and cosines. Therefore a law of
tangents should be expected. Find and prove it.
13.6 Check that the triangle A = (0, 0), B = (5, 0) and C = (5, 3) is directly similar to the
triangle A' = (0, 0), B' = (25, −15), C' = (16, 0). Find the similarity ratio and the rotation
and dilation of the corresponding homothety. Draw the triangles and you will astonish.
FOURTH PART: PLANE PROJECTIONS OF TRIDIMENSIONAL SPACES
The complete study of the geometric algebra of the tridimensional spaces falls
out the scope of this book. However, due to the importance of the Earth charts and of
the Lobachevsky’s geometry, the first one being more practical and the second one
more theoretical, I have written this last section. In order to make the explanations
clearer, the tridimensional geometric algebra has been reduced to the minimal concepts,
enhancing the plane projections.
The geometric quality of being Euclidean or pseudo-Euclidean is not the
signature + or − of a coordinate, but the fact that two coordinates have the same or
different signature, in other words, it is a characteristic of a plane. For instance, a plane
with signatures + + is equivalent, from a geometric point of view, to another with − − .
Therefore, only two kinds of three-dimensional spaces exist: the room space where all
the planes are Euclidean (signatures + + + or − − −), and the pseudo-Euclidean space,
which has one Euclidean plane and two orthogonal pseudo-Euclidean planes (signatures
+ − − or + + − ).
14. SPHERICAL GEOMETRY IN THE EUCLIDEAN SPACE
The geometric algebra of the Euclidean space
A vector of the Euclidean space is an oriented segment in this space with

direction and sense, although it can represent other physical magnitudes such as forces,
velocities, etc. The set of all the segments (geometric vectors) together with their
addition (parallelogram rule) and the product by real numbers (dilation of vectors) has a
structure of vector space, symbolised with V3. Every vector in V3 is of the form:
v = v1 e1 + v 2 e 2 + v 3 e3
where ei are three unitary perpendicular vectors, which form the base of the space. If we
define an associative product (geometric or Clifford product) being a generalisation of
that defined for the plane in the first chapter of this book, we will arrive to:
ei2 = 1 and ei e j = −e j ei for i ≠ j
In general, the square of a vector is the square of its modulus and perpendicular vectors
anticommute whereas proportional vectors commute.
From the base vectors one deduces that the geometric algebra generated by the
space V3 has eight components:
Cl (V3 ) = Cl 3, 0 = 1, e1 , e 2 , e3 , e 23 , e31 , e12 , e123
Let us see with more detail the product of two vectors:
v w = (v1 e1 + v 2 e 2 + v 3 e 3 ) (w1 e1 + w2 e 2 + w3 e3 ) = v1 w1 + v 2 w2 + v 3 w3
+ (v 2 w3 − v 3 w2 ) e 23 + (v 3 w1 − v1 w3 ) e31 + (v1 w2 − v 2 w1 ) e12

The product (or quotient) of two vectors is said a quaternion1. The quaternions are the
even subalgebra of Cl3, 0 that generalise the complex numbers to the space. Splitting a
quaternion in the real and bivector parts, we obtain the inner (or scalar) product and the
outer (or exterior) product respectively:
v · w = v1 w1 + v 2 w2 + v 3 w3
v ∧ w = (v 2 w3 − v 3 w2 ) e 23 + (v 3 w1 − v1 w3 ) e31 + (v1 w2 − v 2 w1 ) e12
The bivectors are oriented plane surfaces and indicate the direction of planes in the
space. Who be acquainted with the vector analysis will say that both vectors and
bivectors are the same thing. This confusion was originated by Hamilton2 himself, and
continued by the founders of vector analysis, Gibbs and Heaviside. However, vectors
and bivectors are different things just as physicists have experienced and know long
time ago. The proper vectors are called usually “polar vectors” while the pseudo-vectors
that actually are bivectors are usually called “axial vectors”. The following magnitudes
are vectors: of course a geometric segment, but also a velocity, an electric field, the
momentum, etc. On the other hand, the oriented area is, of course, a bivector, but also
the angular momentum, the angular velocity and the magnetic field. As a criterion to
distinguish both kind of magnitudes one uses the reversal of coordinates, which changes
the sense of vectors while leaves bivectors invariant.
The product of two bivectors yields a real number plus a bivector. Both parts can
be separated as the symmetric and antisymmetric product. The symmetric product is a
real number and its negative value will be denoted here with the symbol • while the
antisymmetric product is also a bivector and will be noted here with the symbol × :
1
v•w=− (v w + w v ) = v 23 w23 + v 31 w31 + v12 w12
2
1
v×w=− (v w − w v )
2
= (v 31 w12 − v12 w31 ) e 23 + (v12 w23 − v 23 w12 ) e31 + (v 23 w31 − v 31 w23 ) e12
v w = −v • w − v × w
Let us see what happens with the outer product of three vectors. According to
the extension theory of Grassmann, the product u ∧ v ∧ w is the oriented volume
generated by the surface represented by the bivector u ∧ v when it is translated
parallelly along the segment w:
1
Hamilton discovered the quaternions in October 16th 1843 and defined them as quotients of
two vectors. From this definition he deduced the properties of the product of quaternions. I
recommend you the reading of the initial chapters of the Elements of Quaternions because of its
pedagogic importance.
2
This confusion is due to the fact that vectors and bivectors are dual spaces of the algebra Cl3, 0.
However, this duality does not exist at higher dimensions, although there is also duality among
other spaces.
ux vx wx
u ∧ v ∧ w = uy vy w y e123
uz vz wz
Finally, let us see how is the product of three vectors u, v and w. The vector v
can be resolved into a component coplanar with u and v and another component
perpendicular to the plane u-v:
u v w = u v || w + u v ⊥ w
Now let us analyse the permutative property. In the plane we found u v w − w v u = 0. In

the space the permutative property becomes3:
u v w − w v u = u v ⊥ w − w v ⊥ u = v ⊥ (− u w + w u )
= −2 v ⊥ u ∧ w = − 2 v ∧ u ∧ w = 2 u ∧ v ∧ w
I take the same algebraic hierarchies as in chapter 1: all the abridged products
must be operated before the geometric product, convention which is coherent with the
fact that in many algebraic situations, the abridged products must be developed in sums
of geometric products.
Spherical trigonometry
In this section the relations for the sides and angles of the spherical triangles are
deduced. I will take for convenience the sphere having unity radius, although the
trigonometric identities are equally valid for a sphere of any radius.
Let us consider any three points A,
B and C on the sphere with unity radius
centred at the origin (figure 14.1). Then
A = B = C = 1 . The angles formed by
each pair of sides will be denoted by α, β
and γ, and the sides respectively opposite
to these angles will be symbolised by a, b
and c respectively. Then a is the arc of the
great circle passing through the points B
and C, that is, the angle between these
vectors:
Figure 14.1
sin a = B ∧ C
3
From this result it follows that u ∧ v ∧ w = (u v w − u w v + v w u − v u w + w u v − w v u) / 6 .
In other words, the outer product is a fully antisymmetric product. However although being
beautiful, it is not so useful for geometric algebra as the permutative property u ∧ v ∧ w = (u v w
− w v u)/2.
Note that in this equality the sine is positive and therefore a≤π. So, the spherical
trigonometry is deduced for strict triangles, those having a, b, c≤π.
Also, α is the angle between the sides b and c of the triangle, that is, the angle
between the planes passing through the origin, A and B, and the origin, A and C. Since
the direction of a plane is given by its bivector, which can be obtained through the outer
product, we have:
(A ∧ B) × (A ∧ C )
sin α =
A∧B A∧C
Now we write the products of the numerator using the geometric product:
− ( A B − B A) ( A C − C A) + ( A C − C A)( A B − B A)
sin α =
8 A∧B A∧C
− A B A C + A B C A + B A2 C − B A C A + A C A B − A C B A − C A2 B + C A B A
sin α =
8 A∧B A∧C
We extract the vector A as common factor at the left, but without writing it because
A = 1:
− B A C + B C A + A B C − A −1 B A C A + C A B − C B A − A C B + A −1 C A B A
sin α =
8 A∧ B A∧C
Applying the permutative property to the suitable pairs of products, we have:
6 A ∧ B ∧ C + 2 A −1 A ∧ B ∧ C A A∧ B ∧C
sin α = =
8 A∧ B A∧C A∧ B A∧C
since the volume A ∧ B ∧ C is a pseudoscalar, which commutes with all the elements of
the algebra. Now the law of sines for spherical triangles follows:
sin α sin β sin γ A∧ B ∧C

= = = (I)
sin a sin b sin c A∧ B B ∧C C ∧ A
2
Let us see the law of cosines. Since e 23 = e31
2
= e122 = −1 then:
cos α =
(A ∧ B) • (A ∧ C ) = (A ∧ B) • (A ∧ C )
A∧ B A∧C sin c sin b
1
sin b sin c cos α = − [( A B − B A) ( A C − C A) + ( A C − C A)( A B − B A)]
8
=−
1
8
( A B A C − A B C A − B A 2 C + B A C A + A C A B − A C B A − C A 2 B + C A B A)
Now taking into account that:
4 A · B A · C = ( A B + B A) ( A C + C A) = A B A C + A B C A + B A 2 C + B A C A
But also:
4 A · B A · C = ( A C + C A) ( A B + B A) = A C A B + A C B A + C A 2 B + C A B A
and adding the needed terms, we find:
sin b sin c cos α = −

1
8
(8 A · B A · C − 2 A B C A − 2 B A 2 C − 2 A C B A − 2 C A 2 B )
Extracting common factors and using A2 = B2 =1, we may write:
1
sin b sin c cos α = − (8 A · B A · C − 2 A (B C + C B ) A − 2 (B C + C B )) = − A · B A · C + B · C
8
sin b sin c cos α = − cos c cos b + cos a
cos a = cos b cos c + sin b sin c cos α (II)
which is the law of cosines for sides. The substitution of cos c by means of the law of
cosines gives:
cos a = cos b (cos a cos b + sin a sin b cos γ ) + sin b sin c cos α
cos a (1 − cos 2 b ) = cos b sin a sin b cos γ + sin b sin c cos α
and the simplification of sin b:
cos a sin b = cos b sin a cos γ + sin c cos α
The substitution of sin c = sin a sin γ / sin α yields:
sin a sin γ cos α

cos a sin b = cos b sin a cos γ +
sin α
Dividing by sin a :
cot a sin b = cos b cos γ + sin γ cot α (III)

The dual spherical triangle
The perpendicular to the plane containing each side of a spherical triangle cuts
the spherical surface in a point. The three points A', B' and C' obtained in this geometric
way form the dual triangle. The algebraic way to calculate them is the duality operation,
which maps bivectors into the perpendicular vectors. The dual of any element is
obtained as the product by the pseudoscalar unity −e123, which commutes with all the
elements of the algebra:
B ∧C C∧A A∧B
A' = −e123 B' = − e123 C' = − e123
B ∧C C∧A A∧B
The inner product of two vectors yields:
B' · C' =
(C ∧ A) • ( A ∧ B ) = − ( A ∧ B ) • ( A ∧ C ) ⇔ cos a' = − cos α
C ∧ A A∧ B A∧B A∧C
which shows that the angle a' and α are supplementary, and so also the other angles:
a' = π − α b' = π − β c' = π − γ
It is trivial that the dual of the dual triangle is the first triangle, and hence:
α' = π − a β' = π−b γ' = π − c
We may apply the laws of sines and cosines to the dual triangle. The law of sines
is self-dual and may be written in a more symmetric form:

= = = (I)
sin a sin b sin c A' ∧ B' ∧ C'
On the other hand, the law of cosines yields a new result when applying duality:
− cos α' = cos γ' cos β' − sin β' sin γ' cos a'
Removing the marks, since this law must be valid for any spherical triangle, we find the
law of cosines for angles:
cos α = − cos γ cos β + sin β sin γ cos a (IV)
Also, applying the equality (III) to the dual triangle we find:
cot α sin β = − cos β cos c + sin c cot a (V)
The five Bessel’s equalities (I to V) allow to solve every spherical triangle knowing any
three of its six elements.
Right spherical triangles and Napier’s rule
For the case of a right angle spherical triangle, the five Bessel’s formulas are
reduced to a simpler form, and then they may be remembered with the help of the
Napier’s pentagon rule (figure 14.2). Draw the angles and the sides of the triangle
following the order of the
perimeter removing the right
angle and writing instead of
the legs (the sides adjacent to
the right angle) the
complementary arcs. Then
follow the Pentagon rules:
1) The cosine of every
element is equal to the
product of the cotangents
Figure 14.2
of the adjacent elements.
2) The cosine of every
element is equal to the product of the sines of the nonadjacent elements.
For example:
π  π 
cos a = sin − b  sin − c  = cos b cos c
2  2 
π  π 
also: cos − b  = cot γ cot − c  ⇒ sin b = cot γ tg c
 2   2 
This rule is applied to the right side triangles in the same way: remove the right side and
write the complementary of the adjacent angles.
Area of a spherical triangle
A lune is a two-sided polygon on the sphere defined by two great circles. The
area of a lune is proportional to the angle α
between both great circles. For an angle π/2
its area is π, therefore the area of a lune with
angle α is 2α. Now let us consider in the
sphere shown by figure 14.3 the three lunes
having the angles α, β and γ: Then:
 s + t = 2α

 s + u' = 2 β
 s + v = 2γ Figure 14.3

where u' is the area of the antipodal triangle

of u. Both triangles u and u' have the same angles and area and the system can be
rewritten:
s + t = 2α

s + u = 2 β
s + v = 2γ

Adding the three equations, we find:
3 s + t + u + v = 2 (α + β + γ )
The four triangles s, t, u and v fill an hemisphere:
s+t+u+v=2π
so the area of the triangle s is the spherical excess, that is, the addition of the three
angles minus π:
s =α + β +γ − π
Properties of the projections of the spherical surface
No chart of the spherical surface preserving the scale of distances everywhere

exists, that is, we cannot depict any map with distances proportional to those measured
on the sphere. The distortion is the variation of the scale and the angles. Usually there is
a line of zero distortion, where the scale is constant.
Since the scale of distances is never preserved for all the points, projections with
other interesting properties have been searched in cartography. A projection is said to be
conformal if it preserves in the map the angles between great circles on the sphere. A
projection is said to be equivalent if it preserves the area, that is if the figures on the
sphere are projected into figures on the map having the same area. A projection is said
to be equidistant if the scale of distances is preserved, not everywhere but on the line
perpendicular to the line of zero distortion, or radially outwards from a point of zero
distortion.
The more general concept of projection is any one-to-one mapping of any point
(x, y, z) on the sphere into a point (u, v) on the plane, although the main types of
projections are azimuthal, cylindrical and conic. An azimuthal projection is a standard
projection into a plane, which may be considered touching the sphere. In the tangency
point there is zero distortion and the bearings or azimuths from this point are correctly
shown. A cylindrical projection is a projection into a cylindrical surface around the
sphere that will be unrolled. A conic projection is a projection into a conical surface
tangent to the sphere that will also be unrolled.
Now, I review the main and more used projections beginning with the azimuthal
projections.
The central or gnomonic projection
Let us consider the sphere with unity radius (figure 14.4) centred at the origin.
Any point on the sphere has the coordinates (x, y, z) fulfilling:
x2 + y2 + z2 = 1
Every point on the upper

hemisphere is projected into
another point on the plane z = 1
using as centre of projection the
centre of the sphere. Let u and v
be the Cartesian coordinates on
the projection plane. Taking
similar triangles the following
relations are found:
x y
u= v=
z z Figure 14.4
from where we obtain:
u v 1
x= y= z=
u + v +1
2 2
u + v +1
2 2
u + v2 +1
2
The differential of the arc length is obtained through the differentiation of the above
relationships:
ds = dx e1 + dy e 2 + dz e3
ds 2 = dx 2 + dy 2 + dz 2 =
( 1 + v )du
2 2
− 2 u v du dv + ( 1 + u 2 ) dv 2
(1 + u 2
+ v2 )
2
The geodesics of the sphere are the great circles, which are the intersections with planes
passing through its centre. These planes cut the projection plane in straight lines, which
are the projections of the geodesics. In other words, any great circle is projected into a
line on the projection plane. Taking as equation of the line:
v=ku+l
with k and l constant, the substitution in ds gives:
1+ k2 + l2
ds = du
u 2 (1 + k 2 ) + 2 k l u + 1 + l 2
By integration we arrive to the following primitive:
u (1 + k 2 ) + k l
s = arctg + const
1+ k2 + l2
The arc length between two points on this great circle is the difference of this
primitive between both points. However it is more advantageous to write it using
cosines instead of tangents by means of the trigonometric identity:
1 + tg s A tg s B
cos(s B − s A ) ≡
1 + tg 2 s A 1 + tg 2 s B
After removing k and l using the equation of the line, we arrive to:
1 + u A uB + v A vB
cos(s B − s A ) =
2 2 2 2
1 + uA + vA 1 + uB + vB
(u A e1 + v A e 2 + e3 ) · (u B e1 + v B e2 + e3 )
=
u A e1 + v A e 2 + e3 u B e1 + v B e 2 + e3
a trivial result because the arc length is the angle between the position vectors of both
points, and also of the proportional vectors going to the projection plane. However, the
interest of this result is its analogy with the result found for the hyperboloidal surface.
From this value of the cosine, we may obtain the sine of the arc:
(u A v B − u B v A )2 + (u A − v A )2 + (u B − v B )2
sin(s B − s A ) = 1 − cos (s B − s A ) =
2
2 2 2 2
1 + uA + vA 1 + uB + vB
(u A e1 + v A e2 + e3 ) ∧ (u B e1 + v B e 2 + e3 )
=
u A e1 + v A e 2 + e3 u B e1 + v B e 2 + e3
which is also a trivial result, since the sine of the angle is proportional to the modulus of
the outer product.
Let us see the area function. The differential of the area is easily obtained taking
into account that it is a bivector and using the outer product of the differentials of the
coordinates:
du ∧ dv
dA = (dx ∧ dy )2 + (dy ∧ dz )2 + (dz ∧ dx )2 =
(1 + u 2
+ v2 )
3/ 2
This result shows that the central projection is not equivalent and the distortion
increases with the distance to the origin.
Let us consider a plane passing through the centre of the sphere, which cuts its
surface in the great circle determined by the equation system:
 x2 + y2 + z2 =1

 a x + b y + z = 0 a , b real
Then the angle between two great circles is the angle between both central planes:
(a e 23 + b e31 + e12 ) • (a' e23 + b' e31 + e12 ) a a' + b b' + 1

cos α = =
a e 23 + b e31 + e12 a' e 23 + b' e31 + e12 a 2 + b 2 + 1 a' 2 + b' 2 + 1
The sides of a spherical triangle are great circles; therefore the central projection
of a spherical triangle is a triangle whose sides are straight lines.
Stereographic projection
In the stereographic projection the point of view is placed on the spherical

surface. As before, every point (x, y, z) on the sphere with unity radius centred at the
origin fulfils the equation:
x2 + y2 + z2 =1
We project the spherical surface into

the plane z = 0 locating the centre of
projection at the pole (0, 0, −1) (figure
14.5). The upper hemisphere is
projected inside the circle of unity
radius while the lowest hemisphere is
projected outside. If u and v are the
Cartesian coordinates on the projection
plane, we have by similar triangles:
Figure 14.5
x y
= z +1 = z +1
u v
Using the equation of the sphere one arrives to:
2u 2v 2
x= y= z= −1
1+ u2 + v2 1+ u2 + v2 1 + u + v2
2
from where the differential of the arc length is obtained:
4 ( du 2 + dv 2 )
ds 2 = dx 2 + dy 2 + dz 2 =
(1 + u 2
+ v2 )
2
Now we see that this projection is not equidistant and the distortion increases with the
distance to the origin. The factor 4 indicates that the lengths at the origin of coordinates
on the plane are the half of the lengths on the sphere. Taking instead of the plane z = 0,
the plane z = 1 this factor becomes 1. Then, we can state correctly the scale of the chart
(for example, a polar chart).
The geodesic lines (great circles) are intersection of the sphere surface with
planes passing through the origin, which have the equation:
z=a x+b y
The substitution by the stereographic coordinates yields:

Figure 14.6
( u + a) 2
+ ( v + b) = a + b + 1
2 2 2
which is the equation of a circle centred at (−a,

−b) with radius r = a 2 + b 2 + 1 . Observe that
this radius is the hypotenuse of a right angle
triangle having as legs the distance to the
origin and the unity. That is, the great circles
on the sphere are shown in the stereographic
projection as circles that intersect the circle x2 +
y2 =1 in extremes of diameters (figure 14.6).
Usually only the projection of the upper
hemisphere is used, so that the great circles are represented as circle arcs inside the
circle x 2 + y 2 ≤ 1 . The angle α between two of these circles, as explained in the page
89, is obtained from their radii r and r' and centres O and O' through:
r 2 + r' 2 − (O − O' )
2
a a' + b b' + 1
cos α = =
2 r r' a + b 2 + 1 a' 2 + b' 2 + 1
2
Just this is the angle between the planes a x + b y – z = 0 and a' x + b' y – z = 0, that is,
the angle between the two great circles represented by the projected circles. Therefore,
the stereographic projection is a conformal projection of the spherical surface.
If we calculate the differential of area we find:
4 du ∧ dv
dA =
(1 + u 2
+ v2 )
2
Now we see that this projection is not equivalent since distortion of areas increases with
the distance from the origin (as commented above, the factor 4 becomes 1 projecting
into the plane z =1).
Orthographic projection
The orthographic projection of the sphere is a parallel projection (the point of

view is placed at the infinity). If we make a projection parallel to the z-axis upon the
plane z = 0, the Cartesian coordinates on the map are identical to x, y and we have:
x dx + y dy
z = 1 − x2 − y2 dz = −
1 − x2 − y2
dx 2 + dy 2 + 2 x y dx dy
ds 2 = dx 2 + dy 2 + dz 2 =
1− x2 − y2
1 + x2 + y2
dA = dx ∧ dy
1− x2 − y2
Introducing the distance r to the origin of coordinates and the angle ϕ with respect to the
x-axis, we have:
x = r cos ϕ y = r sin ϕ
dr 2 + r 2 dϕ 2 + 2 r dr dϕ
ds 2 =
1− r2
An example of orthographic projection is the Earth image that appears in many TV

news.
Spherical coordinates and cylindrical equidistant (Plate Carrée) projection
For a sphere with unity radius, the spherical coordinates4 are related with the
Cartesian coordinates by means of:
x = sin θ cos ϕ
y = sin θ sin ϕ
z = cos θ
Then the differential of arc length is:
ds = (cos θ cos ϕ dθ − sinθ sinϕ dϕ ) e1 + (cos θ sin ϕ dθ + sinθ cosϕ dϕ ) e 2 − sinθ dθ e3
Introducing the unitary vectors eθ and eϕ

as:
eθ = cosθ cos ϕ e1 + cosθ sin ϕ e2 − sinθ e3
eϕ = −sin ϕ e1 + cos ϕ e 2
the differential of arc length in spherical

coordinates becomes:
ds = dθ eθ + sin θ dϕ eϕ
ds 2 = dθ 2 + sin 2θ dϕ 2 Figure 14.7
4
For geographical coordinates, θ is the colatitude and ϕ the longitude.
Note that eθ and eϕ are orthogonal vectors, since their inner product is zero. At θ = 0,
ds depends only on dθ and there is a pole. Then θ is the arc length from the pole to the
given point (figure 14.7), while ϕ is the arc length over the equator (θ = π/2, sinθ = 1).
The meridians (ϕ = constant) are geodesics, but the parallels (θ = constant) are not.
Exceptionally, the equator is also a geodesic.
In the Plate Carrée projection u = ϕ and v = π/ 2 − θ (latitude) so that the
meridians and parallels are shown in a squared graticule. This projection is equidistant
for any meridian whereas the distortion in the parallels increases, as well as for every
cylindrical projection, as we separate from the equator. Let us see the area:
dA = sin θ dθ ∧ dϕ = cos v du ∧ dv
The ratio of the real area with respect to the represented area on the chart is equal to the
cosine of the latitude and therefore the projection is not equivalent.
Mercator's projection
The Mercator's projection5 is defined as the cylindrical conformal projection. If

we wish to preserve the angles between curves, we must enlarge the meridians by the
same amount as the parallels are enlarged in a cylindrical projection, that is, by the
factor 1/sin θ (the secant of the geographical latitude):
− dθ θ
dv = ⇒ v = − log tg
sinθ 2
The differential of arc length is:
4 exp(2 v )
ds 2 = dθ 2 + sin 2θ dϕ 2 = sin 2θ (du 2 + dv 2 ) = (du 2
+ dv 2 )
(exp(2v ) + 1) 2
2 exp( v )
ds = du 2 + dv 2
exp(2v ) + 1
where the distances are increased in an amount independent of the direction and
proportional to the inverse of the sine of θ. The differential of area is:
4 exp(2 v )
dA = sinθ dθ ∧ dϕ = du ∧ dv
(exp(2v ) + 1)2
5
The difference between the U.T.M. (Universal Transverse Mercator) projection and the
Mercator planisphere is not geometric but geographic: in the U.T.M. the cylinder of projection
is tangent to a meridian instead of the equator. All the Earth has been divided in zones of 6º of
longitude, where the cylinder of projection is tangent to the central meridian (3º, 9º, 15º ... ).
Peters’ projection
The Peters’ projection is the cylindrical equivalent projection. If we wish to

preserve area, we must shorten the meridians in the same amount as the parallels are
enlarged in the cylindrical projection, what
yields:
dv = −sin θ dθ ⇒ v = cos θ
ds 2 = dθ 2 + sin 2θ dϕ 2 = (1 − v 2 ) du 2 +
dv 2
1− v2
dA = sin θ dθ ∧ dϕ = du ∧ dv
which displays clearly the equivalence of the

projection. Observe that v = cos θ means the
sphere is projected following planes
perpendicular to the cylinder of projection
(figure 14.8).
Figure 14.8
Conic projections
These projections are made into a cone surface tangent to the sphere (figure
14.9). Because the cone surface unrolled is a plane circular sector, they are often used to
display middle latitudes, while the azimuthal projections are mainly used for poles. In a
conic projection a small circle (a parallel) is shown as a circle with zero distortion. The
characteristic parameter of a conic projection is the constant of the cone n = cos θ0,
being θ0 the angle of inclination of the generatrix of the cone and also the angle from
the axis of the cone to any point of tangency with the sphere. Since the graticule of the
conic projections is radial, to use the radius r and the angle χ is more convenient:
dr = f (θ ) dθ dχ = n dϕ
The differentials of arc length and area for a conic projection are:
dr 2 sin 2θ
ds 2 = dθ 2 + sin 2θ dϕ 2 = + dχ 2
[ f (θ )]2 n2
sin θ
dA = sinθ dθ ∧ dϕ = dr ∧ dχ
n f (θ )
Let us see as before the three special cases: equidistant, conformal and
equivalent projections. The differential of area for polar coordinates r, χ is
dA = r dr ∧ dχ . If the projection is equivalent, we must identify both dA to find :
d  sin θ  1 θ 2 θ
= f (θ ) ⇒ f (θ ) = cos and r= sin
dθ  n f (θ ) n 2 n 2
r2 n
1−
n 4 r 2 dχ 2
ds 2 = dr 2 +
nr 2
n Figure 14.9
1−
4
If the projection is equidistant, the meridians

have zero distortion so dθ = dr / n and:
1  2 r 
ds 2 = 2
 dr + sin 2   dχ 2 
n  n 
If the projection is conformal then

ds 2 ∝ dr 2 + r 2 dχ 2 so:
sin 2θ
ds 2 = ( dr 2 + r 2 dχ 2 )
n2r2
Solving the differential equation:
sinθ
dθ = dr
nr
with the boundary condition tg θ 0 = r0 as shown by the figure 14.9 we find the
Lambert’s conformal projection:
n
 θ 
 tg 
r = tg θ 0  2 
 θ0 
 tg 
 2 
Exercises
14.1 We have seen the gnomonic projection of great circles being always straight lines.
Then, why does the shadow of a gnomon of a sundial follow a hyperbola on a plane
surface instead of a line during a day? Why does this hyperbola become a straight line
the March 21 and September 23?
14.2 Three points on the sphere are projected by means of the stereographic projection
into a circle of unity radius with the coordinates A = (−0.5, 0.5), B = (0, −2/3) and C =
(2/3, 0). Calculate the sides, angles and area of the triangle which they form.
14.3 Built at Belfast, the Titanic begun its first and last travel in Southampton the April
10th 1912. After visiting Cherbourg the Titanic weighed anchor in the Cork harbour
with bearing New York the 11th April. Calculate the length of the shortest trajectory
from Fastnet (Ireland) at 51º30’N 9º35’W to Sandy Hook (New York) at 40º30’N
74ºW. From January 15th to July 15th the ships had to follow the orthodrome (great
circle) from Fastnet to the point 42º30’N 47ºW and from this point to Sandy Hook
passing at twenty miles from the floating lighthouse of Nantucket. Calculate the length
of the route the Titanic should have followed. Take as an averaged radius of the Earth
6366 km. The Titanic sank the 15th April at the position 41º46’N 50º14’W. Is this point
on the obliged track?
Figure 14.10
14.4 The figure 14.11 is a photograph of the Hale-Bopp comet. The orientation of the
camera is unknown, but the W of Cassiopeia constellation appears in the photograph.
The declination D of a star is the angular distance from the celestial equator to the star
(measured on the great circle passing through the celestial pole and the star). The right
ascension A is the arc of celestial equator measured eastward from the vernal equinox
(one of the intersections of the celestial equator with the ecliptic, also called Aries
point) to the foot of the great circle passing through the star and the pole. The right
ascension and declination of the stars are constant while those of the comets, planets,
the Sun and the Moon are variable. The data of Cassiopeia constellation are:
star Magnitude A D
α Cassiopeia (Schedir) 2.2 0h 40min 6.4s = 10.0267º 56º 29’ 57” N
β Cassiopeia (Caph) 2.3 0h 8min 48.1s = 2.2004º 59º 6’ 40” N
γ Cassiopeia (Tsih) 2.4 0h 56min 16.9s = 14.0704º 60º 40’ 44” N
δ Cassiopeia (Ruchbah) 2.7 1h 25min 21.2s = 21.3383º 60º 11’ 57” N
In your system of coordinates, take the x-axis as the Aries point and the z-axis as the
north pole. Then D = π/2−θ and A=ϕ .
a) Calculate the focal distance of the photograph, that is, the distance from the point of
view to the plane of the photograph. Knowing that the negative was universal
(24×36 mm), calculate the focal distance of the camera.
b) Calculate the orientation (right ascension and declination) of the photographic
camera.
c) Calculate the coordinates of the Hale-Bopp comet.
Figure 14.11
15. HYPERBOLOIDAL GEOMETRY IN THE PSEUDO-EUCLIDEAN SPACE

(LOBACHEVSKY’S GEOMETRY)
The geometric algebra of the pseudo-Euclidean space
A vector of the pseudo-Euclidean space is an oriented segment in this space with

direction and sense. The set of all segments (vectors) together with their addition
(parallelogram rule) and the product by real numbers (dilation of vectors) has a structure
of vector space, symbolised here with W3. Every vector in W3 is of the form:
v = v1 e1 + v 2 e 2 + v 3 e3
where ei are three unitary perpendicular vectors, which constitute the base of the space.
The modulus of a vector is:
2
v = − v12 − v 22 + v 32
It determines the geometric properties of the space, very different from the Euclidean
space. Now we define an associative product (geometric or Clifford product) being a
generalisation of those defined for the Euclidean and hyperbolic planes. Imposing the
condition that the square of the modulus be equal to the square of the vector, we find:
2
v = v2
e12 = −1 e 22 = −1 e32 = 1 and ei e j = −e j ei for i ≠ j
From the base vectors one deduces that the geometric algebra generated by the
space W3 has eight components:
Cl (W3 ) = Cl1, 2 = 1, e1 , e 2 , e3 , e 23 , e31 , e12 , e123
Let us see with more detail the product of two vectors:
v w = (v1 e1 + v 2 e 2 + v 3 e3 ) (w1 e1 + w2 e 2 + w3 e 3 ) = −v1 w1 − v 2 w2 + v 3 w3
+ (v 2 w3 − v 3 w2 ) e 23 + (v 3 w1 − v1 w3 ) e31 + (v1 w2 − v 2 w1 ) e12
I shall call the product (or quotient) of two vectors a tetranion. The tetranions are the
even subalgebra of Cl1, 2 that generalises the complex and hyperbolic numbers to the
pseudo-Euclidean space. Splitting a tetranion in the real and bivector parts, we obtain
the inner (or scalar) product and the outer (or exterior) product respectively:
v · w = − v1 w1 − v 2 w2 + v 3 w3
v ∧ w = (v 2 w3 − v 3 w2 ) e 23 + (v 3 w1 − v1 w3 ) e31 + (v1 w2 − v 2 w1 ) e12

Here also, the bivectors are oriented plane surfaces indicating the direction of planes in
the pseudo-Euclidean space. As before, vectors and bivectors are different things. This
fact has been experienced also by physicists: in the Minkowski’s space, the
electromagnetic field is a bivector while the tetrapotential is a vector. On the other hand,
the oriented area is, of course, a bivector. As a criterion to distinguish both kind of
magnitudes one also uses the reversal of coordinates, which changes the sense of
vectors while leaves bivectors invariant.
Two vectors are said to be orthogonal if their inner product vanishes:
v⊥w ⇔ v·w=0
So the outer product is the product by the orthogonal component and the inner product
is the product by the proportional component:
v · w = v w|| v ∧ w = v w⊥
The product of two bivectors yields a tetranion. The real and bivector parts can
be separated as the symmetric and antisymmetric product. Since:
2
e 23 = e31
2
=1 and e122 = −1
the symmetric product is a real number whose negative value will be denoted here with
the symbol •, whereas the antisymmetric product, denoted here with the symbol ×, is
also a bivector:
1
v•w=− (v w + w v ) = −v 23 w23 − v 31 w31 + v12 w12
2
1
v×w=− (v w − w v )
2
= −(v 31 w12 − v12 w31 ) e 23 − (v12 w23 − v 23 w12 ) e31 + (v 23 w31 − v 31 w23 ) e12
v w = −v • w − v × w
The outer product of three vectors has the same expression as for Euclidean
geometry and this is a natural outcome of the extension theory: the product u ∧ v ∧ w is
the oriented volume generated by the surface represented by the bivector u ∧ v when it
is translated parallelly along the segment w:
ux vx wx
u ∧ v ∧ w = uy vy w y e123
uz vz wz
Finally, let us see how the product of three vectors u, v and w is. The vector v
can be resolved into a component coplanar with u and v and another component
perpendicular to the plane u-v:
u v w = u v || w + u v ⊥ w
Now let us analyse the permutative property. In both Euclidean and hyperbolic planes
we found u v w − w v u = 0. In the pseudo-Euclidean space the permutative property
becomes:
u v w − w v u = u v ⊥ w − w v ⊥ u = v ⊥ (− u w + w u )
= −2 v ⊥ u ∧ w = − 2 v ∧ u ∧ w = 2 u ∧ v ∧ w
I take the same algebraic hierarchies as in the former chapter: all the abridged
products must be operated before the geometric product, convention adequate to the fact
that in many algebraic situations, the abridged products must be developed in sums of
geometric products.
The hyperboloid of two sheets
According to Hilbert (Grundlagen der Geometrie, Anhang V) the complete

Lobachevsky’s “plane” cannot be represented by a smooth surface with constant
curvature as proposed by Beltrami. But this
result only concerns surfaces in the Euclidean
space. The surface whose points are placed at a
fixed distance from the origin in a pseudo-
Euclidean space (the two-sheeted hyperboloid) is
the surface searched by Hilbert which realises
the Lobachevsky’s geometry1. It is known that it
has a characteristic distance like the radius of the
sphere. Since all the spheres are similar, we need
only study the unitary sphere. Likewise, all the
hyperboloidal surfaces z2−x2−y2 = r2 are similar Figure 15.1
and the hyperboloid with unity radius (figure
15.1):
z2 − x2 − y2 =1
where (x, y, z) are Cartesian although not Euclidean2 coordinates, suffices to study the
whole Lobachesky’s geometry.
1
The reader will find a complete study of the hyperboloidal surface in Faber, Foundations of
Euclidean and Non-Euclidean Geometry, chap. VII. “The Weierstrass model”.
2
For me, Cartesian coordinates are orthogonal coordinates in a flat space (with null curvature),
independently of the Euclidean or pseudo-Euclidean nature of the space. Lines and planes
(linear relations between Cartesian coordinates) can always be drawn in both spaces, although
the pseudo-Euclidean orthogonality is not shown as a right angle, and the pseudo-Euclidean
distance is not the viewed length. The incompleteness of Cartesian coordinates was already
criticised by Leibniz in a letter to Huygens in 1679 proposing a new geometric calculus: «Car
premierement je puis exprimer parfaitement par ce calcul toute la nature ou définition de la
figure (ce que l’algèbre ne fait jamais, car disant que x2 + y2 aeq. a2 est l’equation du cercle, il
The fact that the hyperboloidal surface owns the Lobachevsky’s geometry will
be evident through the different projections here reviewed. If we followed the same
order as in the former chapter, we should firstly deal with the Lobachevskian
trigonometry. However, the concept of arc length on the hyperboloid is much less
intuitive than on the sphere. Therefore we must see the concept and determination of the
arc length before studying the hyperboloidal trigonometry.
The central projection (Beltrami disk)
In this projection the centre of the hyperboloid is the centre of projection. Thus,
we project every point on the upper sheet of the hyperboloid z2−x2−y2 = 1 into another
point on the plane z = 1 taking the origin x = 0, y = 0, z = 0 as centre of projection
(figure 15.1). Let u and v be Cartesian coordinates on the plane z =1, which touches the
hyperboloid in the vertex. By similar triangles3:
x y
u= v=
z z
From where we obtain:
u v 1
x= y= z=
1− u − v
2 2
1− u − v
2 2
1− u2 − v2
The differential of the arc length upon the hyperboloid is:
ds = dx e1 + dy e 2 + dz e3
ds 2 = −dx 2 − dy 2 + dz 2 = −
(1 − v )du
2 2
+ 2 u v du dv + ( 1 − u 2 )dv 2
(1 − u 2
− v2 )
2
which is the negative value of the usual metric of the Lobachevsky’s surface in the
Beltrami disk. The negative sign indicates that the arc length is comparable with lengths
in the x-y plane. The whole upper sheet of the hyperboloid is projected inside the
Beltrami’s circle u 2 + v 2 ≤ 1 with unity radius. However, if we project the one-sheeted
hyperboloid z 2 − x 2 − y 2 = −1 into the plane z = 1 we find the same metric but
positive, what indicates the lengths being comparable to those measured on the z-axis.
All the upper half of the one-sheeted hyperboloid is projected outside the circle of unity
radius, giving rise to a new geometry not studied up till now.
faut expliquer par la figure ce que c’est x et y.» (Josep Manel Parra, “Geometric Algebra versus
numerical cartesianism” in F. Brackx, R. Delanghe, H. Serras, Clifford Algebras and Their
Applications in Mathematical Physics).
3
Now to consider that the legs of both right angle triangles are horizontal and vertical, and
therefore perpendicular is enough for the deduction. Notwithstanding, in a pseudo-Euclidean
space one must use the algebraic definition of similitude: a triangle with sides a and b is directly
similar to a triangle with sides c and d if and only if a c −1 = b d −1. This condition implies that
both quotients of the modulus are equal and both arguments also.
Let us consider a plane passing through the origin of coordinates (central plane).
Its intersection with the two-sheeted hyperboloid is the hyperbola determined by the
equations system:
 z2 − x2 − y2 =1

 z=ax+by a , b real
We search the Frenet's trihedron. By differentiation we find a linear differential

equations system :
 z dz = x dx + y dy

 dz = a dx + b dy
where dx and dy can be isolated:
y −bz az−x
dx = dz dy = dz
a y −bx a y −bx
Then the differential of the arc length of this hyperbola is:
 y −bz az−x 
ds = dx e1 + dy e 2 + dz e3 =  e1 + e 2 + e3  dz
a y −bx a y −bx 
1 − a 2 − b2
and its square: ds 2 = −dx 2 − dy 2 + dz 2 = dz 2
(a y − b x ) 2
Now we obtain the unitary vector t tangent to the hyperbola:
dx dy dz ( y − b z ) e1 + ( a z − x ) e 2 + ( a y − b x ) e3
t= e1 + e 2 + e3 =
ds ds ds 1 − a 2 − b2
Observe that t and ds have imaginary components since a plane only cuts the two-
sheeted hyperboloid if a2 + b2 > 1. Anyway, the geometric vector can be taken with real
components, although then its modulus is −1.
The derivative of the tangent vector with respect to the arc length is not only
proportional but equal to the normal vector n of the hyperbola:
dt dt
n= = − ( x e1 + y e 2 + z e3 ) ⇒ κ= =1
ds ds
because the position vector of every point on the two sheet hyperboloid has unitary
modulus. Hence the curvature κ of the hyperbola is constant and equal to 1. All the
hyperbolas being intersections of the hyperboloid with planes passing through the origin
and a given point on the hyperboloid have the same radius of curvature r:
−1
 dt 
r=  = −( x e1 + y e 2 + z e3 )
 ds 
Two consequences are deduced from this fact: firstly their common radius of curvature
is perpendicular to the surface, and secondly, these hyperbolas are geodesics of the
hyperboloid. Because all the hyperbolas lie on planes passing through the origin, their
central projections are straight lines. In other words, every geodesic hyperbola is
projected into a segment on the Beltrami disk. Taking as equation of the line:
v=ku+l
with k and l constant, the substitution in ds gives4:
ds 1+ k2 − l2
= du
e123 − u 2 (1 + k 2 ) − 2 k l u + 1 − l 2
By integration we arrive at the following primitive:
k l + 1+ k2 − l2
u (1 + k 2 ) + k l
u+
s 1 1+ k2
= arg tgh + const = log + const
e123 1+ k2 − l2 2 k l − 1+ k2 − l2
u+
1+ k2
The line v = k u + l cuts the Beltrami’s circle u 2 + v 2 = 1 at the points A and D

whose coordinates are the solution of the system of the equations:
v =k u +l − k l − 1+ k2 − l2 − k l + 1+ k2 − l2
 2 ⇒ uA = uD =
u + v = 1
2
k2 +1 k2 +1
Thus the primitive may be rewritten:

Figure 15.2
s 1 u − uA
= log + const
e123 2 u − uD
Then, the arc length between two points B and C on this

geodesic hyperbola is the difference of this primitive
between both points:
sC − s B 1 (u − u A ) (u D − u B ) 1
= log C = log( A B C D )
e123 2 (u D − u A ) (u C − u B ) 2
4
As said before, the arc length s on the two-sheeted hyperboloid has imaginary values. I’m
sorry by the inconvenience of that reader accustomed to take real values. However the
coherence of all the geometric algebra force to take account of the imaginary unity e123. This
also explains why Lobachevsky found the hyperboloidal trigonometry by choosing imaginary
values for the sides in the spherical trigonometry.
that is, the half of the logarithm of the cross ratio of the four points (figure 15.2) on the
Beltrami disk.
However it is more advantageous to write it using cosines instead of tangents by
means of the trigonometric identity:
sB s
1 − tgh tgh C
sC − s B e123 e123
cosh ≡
e123 sB s
1 − tgh 2 1 − tgh 2 C
e123 e123
After removing k and l using the equation of the line in the primitive, we arrive at:
sC − s B 1 − u B uC − v B vC
cosh =
e123 2 2 2
1 − u B − v B 1 − uC − vC
2
where the expression is real for all the points on the hyperboloid, whose projections lie
on the Beltrami disk u2 + v2 ≤ 1. This expression is equivalent to write:
sC − s B (u B e1 + v B e2 + e3 ) · (u C e1 + v C e 2 + e3 )
cosh =
e123 (u B e1 + v B e 2 + e3 ) (u C e1 + v C e 2 + e3 )
Since the vector u A e1 + v A e 2 + e 3 is proportional to the position vector, we have:
sC − s B (x B e1 + y B e 2 + z B e3 ) · (x C e1 + y C e 2 + z C e3 )
cosh = cos(s C − s B ) =
e123 (x B e1 + y B e 2 + z B e3 ) (x C e1 + y C e 2 + z C e3 )
which is the expected expression to calculate angles between vectors in the pseudo-
Euclidean space. This result must be commented in more detail. Note that every pair of
vectors going from the origin to the two-sheeted hyperboloid always lie on a central
plane having hyperbolic nature, since it cuts the cone z 2 − x 2 − y 2 = 0 of vectors with
zero length. Because of this, the angle measured on this plane is hyperbolic.
A hyperboloidal triangle is the region of the hyperboloid bounded by three
geodesic hyperbolas. The length of a side of the triangle can be calculated by integration
of the arc length of the hyperbola according to the last result. On the other hand, the
angle of a vertex of a hyperboloidal triangle is the angle between the tangent vectors of
the hyperbolas of each side at the vertex:
− ( y − b z )( y − b' z ) − (a z − x )(a' z − x ) + (a y − b x )(a' y − b' x )

cos α = t ⋅ t ' =
1 − a 2 − b2 1 − a' 2 − b' 2
1 − a a' − b b'
=
1 − a − b2
2
1 − a' 2 − b' 2
Note that every pair of tangent vectors lie on a plane of Euclidean nature, a plane
parallel to a central plane not cutting the cone of zero length. So the angles measured on
this plane are circular. We arrive at the same conclusion by proving that the absolute
value of the cosine (and also its square) is lesser than the unity:
(1 − a a' − b b' )2
<1
(1 − a 2
− b 2 ) (1 − a' 2 − b' 2 )
2 2
The denominator is positive (although both squares are negative because a 2 + b 2 > 1 )
and it can pass to the right hand side of the inequality without changing the sense.
Removing the denominator and after simplification:
− 2 a a' − 2 b b' + 2 a a' b b' < −a 2 − b 2 − a' 2 − b' 2 + a 2 b' 2 + a' 2 b 2
and arranging terms we find an equivalent inequality:
a 2 + a' 2 − 2 a a' + b 2 + b' 2 − 2 b b' < a 2 b' 2 + a' 2 b 2 − 2 a a' b b'
(a − a' )2 + (b − b' )2 < (a b' − b a' )2

Just this is the condition for the existence of the intersection point where two central
planes and the hyperboloid meet as I shall show now. The planes with equations a x + b
y = z and a' x + b' y = z are projected into the lines a u + b v = 1 and a' u + b' v = 1 on
the Beltrami disk. Then the point of intersection of both planes and the hyperboloidal
surface is projected into the point on the Beltrami disk given by the system of equations:
 a u + b v =1

a' u + b' v = 1
b' − b − a' + a
whose solution is: u= v=
a b' − a' b a b' − a' b
This point lies inside the Beltrami circle (and the point of intersection of the
hyperboloid and both planes exists) if and only if:
u2 + v2 =
(a − a' )2 + (b − b' )2 <1
(a b' − b a' )2
which is the condition above obtained.
Note that the expression for the angle between geodesic hyperbolas coincides
with the angle between the bivectors of the planes containing these hyperbolas:
(a e 23 + b e31 + e12 ) • (a' e 23 + b' e31 + e12 )

cos α =
a e 23 + b e 31 + e12 a' e 23 + b' e31 + e12
It is not a surprising result, because it happens likewise for the sphere: the angle
between two maximum circles is the angle between the central planes containing them.
The planes passing through the origin only cut the two-sheeted hyperboloid if a2
+ b >1. These planes also cut the cone of zero length z2 − x2 − y2 = 0 so they are
2
hyperbolic and the angles measured on them are hyperbolic arguments. On the other
hand, the planes with a2 + b2 <1 do not cut the hyperboloid neither the cone and have
Euclidean nature. In the Lobachevskian trigonometry, the planes containing the sides of
a hyperboloidal triangle are always hyperbolic, while the planes touching the
hyperboloid and containing the angles between sides are always Euclidean, and hence
the angles are circular. Resuming, real modulus of a bivector (0<−a2 −b2 +1) indicates
an Euclidean plane and imaginary modulus (0>−a2 −b2 +1) a hyperbolic plane.
Now we already have all the formulas to deduce the hyperboloidal trigonometry.
Hyperboloidal (Lobachevskian) trigonometry
The Lobachevskian trigonometry is the study of the trigonometric relations for

the sides and angles of geodesic triangles on two-sheeted hyperboloids in the pseudo-
Euclidean space. We will take for
convenience the hyperboloid with
unity radius z2 − x2 − y2 =1,
although the trigonometric identities
are equally valid for a hyperboloid
with any radius.
Consider three points A, B and
C on the unitary hyperboloid (figure
15.3). Then A = B = C = 1 . The
angles formed by each pair of sides
will be denoted by α, β and γ, and the
real value of the sides respectively
opposite to these angles will be
symbolised by a, b and c respectively.
Figure 15.3
Then a is the real value of the arc of
the geodesic hyperbola passing
through the points B and C, that is, the hyperbolic angle between these vectors:
B ∧C
sinh a =
e123
As said above, the planes intersecting the hyperboloid have imaginary modulus. So we
must divide by the pseudoscalar imaginary unity, which is also the volume element. In
this equality the hyperbolic sine is taken positive and also a.
α is the circular angle between the sides b and c of the triangle, that is, the angle
between the planes passing through the origin, A and B, and the origin, A and C
respectively. Since the direction of a plane is given by its bivector, which can be
obtained through the outer product, we have:
(A ∧ B) × (A ∧ C )
sin α = −
A∧B A∧C
where the minus sign is due to the imaginary modulus of the outer products of the
denominators. Now we write the products of the numerator using the geometric product:
− ( A B − B A)( A C − C A) + ( A C − C A)( A B − B A)
sin α = −
8 A∧B A∧C
− A B A C + A B C A + B A2 C − B A C A + A C A B − A C B A − C A2 B + C A B A
sin α = −
8 A∧B A∧C
We extract the vector A as common factor at the left, but without writing it because
A = 1:
− B A C + B C A + A B C − A −1 B A C A + C A B − C B A − A C B + A −1 C A B A
sin α = −
8 A∧ B A∧C
Applying the permutative property to the suitable pairs of products, we have:
6 A ∧ B ∧ C + 2 A −1 A ∧ B ∧ C A A∧ B ∧C
sin α = − =−
8 A∧ B A∧C A∧ B A∧C
since the volume A ∧ B ∧ C is a pseudoscalar, which commutes with all the elements of
the algebra. Now the law of sines for hyperboloidal triangles follows:

= = = (I)
sinh a sinh b sinh c A ∧ B B ∧ C C ∧ A e123
2
Let us see the law of cosines. Since e 23 = e 31
2
= − e122 = 1 then:
cos α =
(A ∧ B) • (A ∧ C ) = − (A ∧ B) • (A ∧ C )
A∧ B A∧C sinh c sinh b
1
sinh b sinh c cos α = (( A B − B A) ( A C − C A) + ( A C − C A)( A B − B A))
8
=
1
( A B A C − A B C A − B A 2 C + B A C A + A C A B − A C B A − C A 2 B + C A B A)
8
Now taking into account that:

4 A · B A · C = ( A B + B A ) ( A C + C A) = A B A C + A B C A + B A 2 C + B A C A
But also:
4 A · B A · C = ( A C + C A) ( A B + B A) = A C A B + A C B A + C A 2 B + C A B A
and adding the needed terms, we find:
sinh b sinh c cos α =

1
8
(8 A · B A · C − 2 A B C A − 2 B A 2 C − 2 A C B A − 2 C A 2 B )
Extracting common factors and using A2 = B2 =1, we may write:
1
sinh b sinh c cos α = (8 A · B A · C − 2 A (B C + C B ) A − 2 (B C + C B )) = A · B A · C − B · C
8
sinh b sinh c cos α = cosh c cosh b − cosh a
cosh a = cosh b cosh c − sinh b sinh c cos α (II)
which is the law of cosines for sides. The substitution of cosh c by means of the law of
cosines gives:
cosh a = cosh b (cosh a cosh b − sinh a sinh b cos γ ) − sinh b sinh c cos α
cosh a (1 − cosh 2 b ) = − cosh b sinh a sinh b cos γ − sinh b sinh c cos α
and the simplification of −sinh b:
cosh a sinh b = cosh b sinh a cos γ + sinh c cos α
The substitution of sinh c = sinh a sin γ / sin α yields:
sinh a sin γ cos α

cosh a sinh b = cosh b sinh a cos γ +
sin α
Dividing by sinh a :
coth a sinh b = cosh b cos γ + sin γ cot α (III)
Figure 15.4
Stereographic projection (Poincaré disk)
We project the hyperboloidal surface into the

plane z = 0 (figure 15.4) taking as centre of
projection the lower pole (0, 0, −1). The upper sheet
is projected inside the circle of unity radius while the
lower sheet is projected outside. Let u and v be the
Cartesian coordinates of the projection plane. By similar triangles we have:
x y
= z +1 = z +1
u v
where (x, y, z) are the coordinates of a point on the hyperboloid. Using its equation
z 2 − x 2 − y 2 = 1 , one arrives at:
2u 2v 2
x= y= z= −1
1− u2 − v2 1 − u2 − v2 1 − u − v2
2
from where the differential of the arc length is obtained:
4 ( du 2 + dv 2 )
ds 2 = − dx 2 − dy 2 + dz 2 = −
(1 − u 2
− v2 )
2
The geodesic hyperbolas are intersection of the hyperboloid with central planes having
the equation:
z=a x+b y Figure 15.5
The substitution by the stereographic coordinates

yields:
( u − a )2 + ( v − b )2 = a 2 + b 2 − 1
which is the equation of a circle centred at (a, b)
with radius r = a 2 + b 2 − 1 . That is, the geodesic
hyperbolas on the hyperboloid are shown as circles
in the stereographic projection (figure 15.5). The
central planes cutting the two-sheeted hyperboloid
fulfil a 2 + b 2 > 1 , so that the radius is real and the centres of these circles are always
placed outside the Poincaré disk.
The angle α between two circles is obtained through:
r 2 + r' 2 − (O − O' )
2
a a' + b b' − 1
cos α = =
2 r r' a 2 + b 2 − 1 a' 2 + b' 2 − 1
Just this is the angle between two geodesic hyperbolas (between the tangent vectors t
and t’) found above. Therefore, the stereographic projection (Poincaré disk) is a
conformal projection of the hyperboloidal surface.
On the other hand, the geodesic circles are always orthogonal to the Poincaré's
circle u 2 + v 2 = 1 ( r' =1, O' = (0, 0) ) because:
r2 +1− O2
cos α = =0
r
The differential of area is easily obtained through the modulus of the differential
bivector:
4 du ∧ dv
dA = (dx ∧ dy )2 − (dy ∧ dz )2 − (dz ∧ dx )2 =
(1 − u 2
− v2 )
2
When doing an inversion of the

Poincaré disk centred at a point lying on
the limit circle, another projection of the Figure 15.6
Lobachevsky's geometry is obtained: the
Poincaré’s half plane. Here the geodesics
are semicircles orthogonal to the base line
of the half plane. Since the inversion is a
conformal transformation, the upper half
plane is also a conformal projection of the
hyperboloid. The figure 15.6 displays the
projection of a hyperboloidal triangle and
its sides in the Poincaré’s upper half plane.
Azimuthal equivalent projection
This projection preserves the area and is similar to the azimuthal equivalent
projection of the sphere usually used in the polar maps of the Earth.
The differential of the area in the pseudo-Euclidean space is a bivector whose
square is:
dA 2 = (dx ∧ dy ) − (dy ∧ dz ) − (dz ∧ dx )

2 2 2
The substitution of the hyperboloid equation:
x y
z2 − x2 − y2 =1 ⇒ dz = dx + dy
z z
1
yields: dA 2 = 2
(dx ∧ dy )2
z
Every azimuthal projection has plane coordinates u and v proportional to x and y,

being the proportionality constant only function of z:
u = x f (z ) v = y f (z )
Then by differentiation of these equalities and substitution of dz, which is a linear

combination of dx and dy we arrive at:
 z2 −1 
du ∧ dv =  f 2 + f f '  dx ∧ dy
 z 
Identifying both area differentials we find the following differential equation:
z2 −1 1
f2+ f f' =
z z
I introduce the auxiliary function g = f 2 :
z2 −1 1
g+ g' =
2z z
Rewriting the equation with differentials, we arrive at an exact differential:
2
−1  z2 −1 
(z g − 1) dz + z dg = 0 ⇔ d  g − z  = 0
2  2 
2 (z + const ) 2 (z + const )
g (z ) = ⇒ f (z ) =
z2 −1 z2 −1
Next to the pole x and u are coincident and also y and v. Then f(1)=1, which implies the
integration constant be −1 and then5:
2 (z − 1) 2 2 2
f (z ) = = ⇒ u=x v= y
z2 −1 z +1 z +1 z +1
Weierstrass coordinates and cylindrical equidistant projection
These are a set of coordinates similar to

the spherical coordinates, but on the
hyperboloid surface in the pseudo-Euclidean
space.
For a hyperboloid with unity radius, the
Weierstrass coordinates are: Figure 15.7
x = sinh ψ cos ϕ
y = sinh ψ sin ϕ
z = coshψ
Then the differential of arc length is:
5
As a comparison in the central projection f(z) = 1/z while for the stereographic projection f(z) =
1 / (z + 1).
ds = (coshψ cos ϕ dψ − sinhψ sinϕ dϕ ) e1 + (coshψ sin ϕ dψ + sinhψ cosϕ dϕ ) e 2 + sinhψ dψ e 3
Introducing the unitary vectors eψ and eϕ as:
eψ = coshψ cos ϕ e1 + coshψ sin ϕ e 2 + sinhψ e3
eϕ = −sin ϕ e1 + cos ϕ e 2
the differential of arc length in hyperboloidal coordinates becomes:
ds = dψ eψ + sinh ψ dϕ eϕ ds 2 = −( dψ 2 + sinh 2ψ dϕ 2 )
Note that eψ and eϕ are orthogonal vectors, since their inner product is zero. At ψ = 0,
ds only depends on dψ , so that there is a pole. Then ψ is the arc length from the pole to
the given point on the hyperboloid surface (figure 15.7), while ϕ is the arc length over
the equator, the parallel determined by sinhψ = 1 and ψ = log(1+ 2 ). The meridians (ϕ
= constant) are geodesics, while the parallels (ψ = constant) including the equator are
not geodesics because their planes do not pass through the origin of coordinates.
In the cylindrical projections, the hyperboloid is projected onto a cylinder
passing through its equator, whose axis is the z axis. The chart is the Euclidean map
obtained unrolling the cylinder6. Using ϕ and ψ as the coordinates u and v of the chart,
the hyperboloid surface is projected into a rectangle with 2π width and infinite height.
The area on this chart is:
dA = sinhψ dϕ ∧ dψ = sinh v du ∧ dv
This projection is equidistant for the meridians (ϕ = constant) and for the equator (ψ =
log(1+ 2 )).
Cylindrical conformal projection
If we wish to preserve the angles between curves, we must enlarge the meridians
by the same amount as the parallels are enlarged in a cylindrical projection, that is by a
factor 1/sinh ψ:
dψ ψ
dv = ⇒ v = log tgh
sinhψ 2
The differential of arc length and area are:
6
Every plane tangent to the two-sheeted hyperboloid has Euclidean nature. So the
Lobachevsky’s surface can be only projected onto a Euclidean chart, and the cylinder of
projection is not hyperbolic and does not belong properly to the pseudo-Euclidean space.
− 4 exp(2 v )
ds 2 = −( dψ 2 + sinh 2ψ dϕ 2 ) = −sinh 2ψ (du 2 + dv 2 ) = (du 2
+ dv 2 )
(1 − exp(2 v ) ) 2
4 exp(2 v )
dA = sinhψ dϕ ∧ dψ = du ∧ dv
(1 − exp(2 v ) )2
Cylindrical equivalent projection
If we wish to preserve area, we must shorten the meridians in the same amount
as the parallels are enlarged in the cylindrical projection, what yields:
dv = sinhψ dψ ⇒ v = coshψ
ds 2 = −( dψ 2 + sinh 2ψ dϕ 2 ) = −(v 2 − 1) du 2 −
1
dv 2
v −1
2
dA = sinhψ dϕ ∧ dψ = du ∧ dv
which clearly displays the equivalence Figure 15.8

of the projection. Observe that v =
cosh ψ means the hyperboloid is
projected following planes
perpendicular to the cylinder of
projection (figure 15.8).
Conic projections
These projections are made

into a cone surface tangent to the
hyperboloid. Here to take the cone
z 2 − x 2 − y 2 = 0 is the most natural
choice, although as before, the cone of
projection has Euclidean nature and does not belong properly to the pseudo-Euclidean
space, where it should be the cone of zero length. The cone surface unrolled is a circular
sector. In a conic projection a parallel is shown as a circle with zero distortion. The
characteristic parameter of a conic projection is the constant of the cone n = cos θ0,
being θ0 the Euclidean angle of inclination of the generatrix of the cone. The relation
with the hyperbolic angle ψ0 of the parallel touching the cone (which is represented
without distortion in the projection) and θ0 is:
1
n = cos θ 0 =
1 + tgh 2 ψ 0
Since the graticule of the conic projections is radial, is more convenient to use
the radius r and the angle χ :
dr = f (ψ ) dψ dχ = n dϕ
The differentials of arc length and area for a conic projection are:
 dr 2 sinh 2ψ 
ds 2 = −( dψ 2 + sinh 2ψ dϕ 2 ) = −  + dχ 2 
 [ f (ψ )]
2 2
n 
sinhψ
dA = sinhψ dϕ ∧ dψ = dr ∧ dχ
n f (ψ )
Let us see as before the three special cases: equidistant, conformal and
equivalent projections. The differential of area for polar coordinates r, χ is
dA = r dr ∧ dχ . If the projection is equivalent, we must identify both dA to find:
d  sinhψ  1 ψ 2 ψ
 n f (ψ )  = f (ψ ) ⇒ f (ψ ) = cosh and r= sinh
dψ   n 2 n 2
 n r2 
 1+ 
n 4 r 2 dχ 2 
ds 2 = −  dr 2
+
 n r2 n 
1 + 
 4 
If the projection is equidistant, the meridians have zero distortion so dψ = dr / n and:
1  2 r 
ds 2 = − 2
 dr + sinh 2   dχ 2 
n  n 
If the projection is conformal then ds 2 ∝ dr 2 + r 2 dχ 2 so:
sinh 2ψ
ds 2 = − ( dr 2 + r 2 dχ 2 )
n2r2
Then we solve the differential equation:
sinhψ
dψ = dr
nr
with the boundary condition tghψ 0 = r0 to find the conformal projection:

n
 ψ 
 tgh 
r = tghψ 0  2 
 ψ0 
 tgh 
 2 
On the congruence of geodesic triangles
Two geodesic triangles on the hyperboloid having the same angles also have the
same sides and are said to be congruent. This follows immediately from the rotations in
the pseudo-Euclidean space, which are expressed by means of tetranions in the same
way as quaternions are used in the rotations of Euclidean space. However, this falls out
of the scope of this book and will not be treated. The reader must perceive, in spite of
his Euclidean eyes7, that all the points on the hyperboloid are equivalent because the
curvature is always the unity and the surface is always perpendicular to the radius. So
the pole (vertex of the hyperboloid) is not any special point, and any other point may be
chosen as a new pole provided that the new axis are obtained from the old ones through
a rotation.
Comment about the names of the non-Euclidean geometry
Finally it is apparent that the different kinds of non-Euclidean geometry have

been often misnamed. The Lobachevskian surface was improperly called the
“hyperbolic plane” in contradiction with the fact that it has constant curvature and
therefore is not flat. Certainly, in the Lobachevsky’s geometry the hyperbolic functions
are widely used, but in the same way as the circular functions work in the spherical
trigonometry, as Lobachevsky himself showed. In the pseudo-Euclidean plane the
geometry of the hyperbola is properly realised, the hyperbolic functions being also
defined there. Thus the pseudo-Euclidean plane should be named properly the
hyperbolic plane, while the Lobachevsky’s geometry might be alternatively called
hyperboloidal geometry. The plane models of the Lobachevsky’s geometry (Beltrami
disk, Poincaré disk and half plane, etc.) must be understood as plane projections of the
hyperboloid in the pseudo-Euclidean space, which is its proper space, and then the
Lobachevskian trigonometry may be called hyperboloidal trigonometry.
On the other hand, the space-time was the first physical discovery of a pseudo-
Euclidean plane, but one no needs to appeal to relativity because, as the lector has
viewed in this book, both signatures of the metrics (Euclidean and pseudo-Euclidean)
are included in the plane geometric algebra and also in the algebras of higher
dimensions.
Exercises
15.1. According to the Lobachevsky’s axiom of parallelism, at least two parallel lines
pass through an exterior point of a given line. The two lines approaching a given line at
7
In fact, eyes are not Euclidean but a very perfect camera where images are projected on a
spherical surface. The principles of projection are likewise applicable to the pseudo-Euclidean
space. Our mind accustomed to the ordinary space (we learn its Euclidean properties in the first
years of our life) deceives us when we wish perceive the pseudo-Euclidean space.
the infinity are called «parallel lines» while those not cutting and not approaching it are
called «ultraparallel» or «divergent lines».
Lobachevsky defined the angle of parallelism Π(s)
as the angle which forms the line parallel to a given
line with its perpendicular, which is a function of
the distance s between the intersections on the
perpendicular. Prove that Π(s) = 2 arctg(exp(−s))
using the stereographic projection.
15.2. Find the equivalent of the Pythagorean

theorem in the Lobachevskian geometry by taking a
right angle triangle. Figure 15.9
15.3. A circle is defined as the curve which is the intersection of the hyperboloid with a
plane not passing through the origin with a slope less than the unity. Show that:
a) The curve obtained is an ellipse.
b) The distance from a point that we call the centre to every point of this ellipse
is constant, and may be called the radius of the circle.
c) This ellipse is projected as a circle in the stereographic projection.
15.4. A horocycle is the intersection of the hyperboloid and a plane with unity slope
( a 2 + b 2 = 1 ). Show that the horocycles are projected as circles touching the limit circle
x 2 + y 2 = 1 in the stereographic projection. Find the centre of a horocycle.
15.5 a) Show that the differential of area in the Beltrami projection is:
du ∧ dv
dA =
(1 − u 2
− v2 )
3/ 2
b) By integrating dA prove that

the area between two lines
forming an angle ϕ and the
common “parallel” line (figure
15.10-a) is equal to π − ϕ.
c) Prove that the area of every
triangle (figure 15.10-b) is equal
Figure 15.10
to its angular defect π − α − β −
γ.
15.6 Use the azimuthal conformal projection to show that the area of a circle
(hyperboloidal segment) with radius ψ is equal to 2 π (cosh ψ − 1). This result is
analogous to the area of a spherical segment 2 π (1 − cos θ ).
16. SOLUTIONS OF THE PROPOSED EXERCISES
1. The vectors and their operations
1.1 Let a, b be the different sides of the parallelogram and c, d both diagonals. Then:
c=a+b d=a−b
c2 + d2 = ( a + b )2 + ( a − b )2 = a2 + 2 a · b + b2 + a2 − 2 a · b + b2 =
= 2 a2 + 2 b2
which are the sum of the squares of the four sides of the parallelogram.
(a b + b a )2 (a b − b a )2
1.2 (a · b ) − (a ∧ b ) =
2 2
+
4 4
a b a b + b a b a + a b2 a + b a 2 b a b a b + b a b a − a b2 a − b a 2 b
= −
4 4
= a 2 b2
1.3 Let us prove that a ∧ b c ∧ d = ( a ∧ b c ) · d :
4a∧b c∧d=(ab−ba)(cd−dc)=abcd−abdc−bacd+badc=
=abcd−dbac−bacd+dabc=(abc−bac)d+d(abc−bac)=
=2(abc−bac)·d=4(a∧b c)·d
In the same way the identity a ∧ b c ∧ d = a · ( b c ∧ d ) is proved.
1.4 Using the identity of the previous exercise we have:
a∧b c∧d+a∧c d∧b+a∧d b∧c=
= a·(b c∧d)+a·(c d∧b)+a·(d b∧c)=
= a·(b c∧d+c d∧b+d b∧c)=
=a·(bcd−bdc+cdb−cbd+dbc−dcb)=a·0=0
where the summands are simplified taking into account the permutative property.
1.5 Let us denote the components proportional and perpendicular to the vector a with ||
and ⊥. Then using the fact that orthogonal vectors anticommute and those with the same
direction commute, we have:
a b c = a ( b|| + b⊥ ) ( c|| + c⊥ ) = ( b|| − b⊥ ) a ( c|| + c⊥ ) = ( b|| − b⊥ ) ( c|| − c⊥ ) a =
= ( b|| c|| − b|| c⊥ − b⊥ c|| + b⊥ c⊥ ) a = ( c|| b|| + c⊥ b|| + c|| b⊥ + c⊥ b⊥ ) a =
= ( c|| + c⊥ ) ( b|| + b⊥ ) a = c b a
1.6 Let a, b, c be the sides of a triangle and h the altitude corresponding to the base b.
Then the altitude divides the triangle in two right triangles where the Pythagorean theorem
may be applied:
b = a2 − h2 + c2 − h2
Isolating the altitude as function of the sides we obtain:
h =a 2 2
−
(a
+ b2 − c2 )
2 2
4 b2
Then the square of the area of the triangle is:
b2h 2 4 a 2b2 − ( a 2 + b2 − c 2 )
2
− a 4 − b4 − c 4 + 2 a 2b2 + 2 a 2c 2 + 2 b2c 2
A =
2
= =
4 16 16
1
Introducing the semiperimeter s = (a+b+c) we find:
2
A = s (s − a ) ( s − b ) ( s − c )
2. A base of vectors for the plane
2.1 Draw, for example, two vectors of the first quadrant and mark u1, u2, v1 and v2 on the
Cartesian axis. Then calculate the area of the parallelogram from the areas of the
rectangles:
A = ( u1 + v1 ) ( u2 + v2 ) − u1 u2 − v1 v2 − 2 v1 u2 = u1 v2 − u2 v1
2.2 The area of a triangle is the half of the area of the parallelogram formed by any two
sides. Therefore:
A = ( 3 e1 + 5 e2 ) ∧ ( −2 e1 − 3 e2 ) = ( −9 + 10 ) e12 = e12 ⇒ A = 1
2.3 a b c = ( a1 e1 + a2 e2 ) ( b1 e1 + b2 e2 ) ( c1 e1 + c2 e2 )
= [ a1 b1 + a2 b2 + e12 ( a1 b2 − a2 b1 ) ] ( c1 e1 + c2 e2 )
= e1 (a1 b1 c1 + a2 b2 c1 + a1 b2 c2 − a2 b1 c2) + e2 (a1 b1 c2 + a2 b2 c2 − a1 b2 c1 + a2 b1 c1)

= ( c1 e1 + c2 e2 ) [ b1 a1 + b2 a2 + e12 ( b1 a2 − b2 a1 ) ] =
= ( c1 e1 + c2 e2 ) ( b1 e1 + b2 e2 ) ( a1 e1 + a2 e2 ) = c b a
2.4 u v = ( 2 e1 + 3 e2 ) ( −3 e1 + 4 e2 ) = 6 + 17 e12
u v = 325 α(u, v) =1.2315=70º33'36"
2.5 For this base we have:
e1 · e2 = | e1 | | e2 | cos π/3 = 1 | e1 ∧ e2 | = | e1 | | e2 | | sin π/3 | = 3
Applying the original definitions of the modulus of a vector one finds:
u2 = ( 2 e1 + 3 e2 )2 = 4 e12 + 9 e22 + 12 e1 · e2 = 4 + 36 + 12 = 52 ⇒ u = 52
v2 = (−3 e1 + 4 e2 )2 = 9 e12 + 16 e22 − 24 e1 · e2 = 9 + 64 −24 = 49 ⇒ v = 49
u v = ( 2 e1 + 3 e2 ) (−3 e1 + 4 e2 ) = −6 e12 + 12 e22 + 8 e1 e2 −9 e2 e1 =
= 42 − e1 · e2 + 17 e1 ∧ e2 = 41 + 17 e1 ∧ e2 ⇒ u v = 2548
u·v 41 u∧v 17 3
cos α = = sinα = ± =
uv 2548 uv 2548
α(u, v) = 0.6228=35º 41' 5" oriented with the sense from e1 to e2 .
When e1 and e2 are not perpendicular, e1 e2 ≠ − e2 e1 and e12 has not the meaning of a pure
area but the outer product e1 ∧ e2 with an area of 3 . Also observe that u v = u v .
2.6 In order to find the components of v for the new base {u1, u2}, we must resolve the
vector v into a linear combination of u1 and u2 .
v = c1 u1 + c2 u2
v ∧ u2 3 (− 3) − (− 5) 5 u1 ∧ v 2 (− 5) − (− 1) 3
c1 = = = −16 c2 = = =7
u1 ∧ u 2 −1 u1 ∧ u 2 −1
so in the new base v = (−16, 7).
3. The complex numbers
3.1 z t = ( 1 + 3 e12 ) ( −2 + 2 e12 ) = −2 −6 + ( 2 −6 ) e12 = −8 −4 e12

3.2 Every complex number z can be written as product of two vectors a and b, its modulus
being the product of the moduli of both vectors:
2 2 2
z=ab ⇒ |z|=|a||b| ⇒ z = a b = a b b a = z z*
4
3.3 The equation x −1 = 0 is solved by extraction of the fourth roots:
x2 = 1 ⇒ x1 = 1, x2 = −1
4
x =1
x2 = −1 ⇒ x3 = e12 , x4 = −e12
3.4 Passing to the polar form we have:
( 2π /4 ) n
(
= − 2 −π / 4 ) n
2n nπ /4 = − 2n − n π/4 ⇒ 2n n π/4 = 2n − n π/4 +π
Therefore the arguments must be equal except for k times 2π:
nπ nπ nπ
=− +π + 2 k π ⇒ =π + 2 k π ⇒ n=2+4k
4 4 2
3 3 3
3.5 The three cubic roots of –3 + 3 e12 are 18 π / 4 , 18 11 π / 12 , 18 19 π / 12 .
3.6 Using the formula of the equation of second degree we find: z1 = 2−3 e12, z2 = 1+e12.
3.7 We suppose that the complex analytic extension f has a real part a of the form:
a = sin x K(y) with K(0) = 1
Applying the first Cauchy-Riemann condition, we find the imaginary part b of f :
∂a ∂b
= cos x K ( y ) = ⇒ b = cos x ∫ K ( y ) dy
∂x ∂y
Applying the second Cauchy-Riemann condition:
∂a ∂b
= sin x K' ( y ) = − = sin x ∫ K ( y ) dy
∂y ∂x
we arrive at a differential equation for K(y) whose solution is the hyperbolic cosine:
K' ( y ) = ∫ K ( y ) dy 
 ⇒ K(y) = cosh y
K (0) = 1 
Hence the analytical extension of the sine is:

sin(x + e12 y) = sin x cosh y + e12 cos x sinh y
The analytical extensions of the trigonometric functions may be also obtained from the
exponential function. In the case of the cosine, we have:
exp( e12 z ) + exp( − e12 z ) exp(− y )(cos x + e12 sin x ) + exp( y )(cos x − e12 sin x )
cos z = =
2 2
cos z = cos x cosh y − e12 sin x sinh y
3.8 Let us calculate some derivatives at z = 0:
f (z ) = log(1 + exp(− z )) f(0) = log 2
− exp(− z ) −1 1
f' (z ) = = f' (0) = −
1 + exp(− z ) exp(z ) + 1 2
exp(z ) 1
f'' (z ) = f'' (0) =
(exp(z ) + 1) 2
4
exp(z ) 2 exp(2 z )
f''' (z ) = − f''' (0) = 0
(exp(z ) + 1) 2
(exp(z ) + 1)3
exp(z ) 2 exp(2 z ) 6 exp(3z )
f IV
(z ) = − −
4 exp( 2 z )
+ f IV
(0) = − 1
(exp(z ) + 1) 2
(exp(z ) + 1) 3
(exp(z ) + 1) 3
(exp(z ) + 1) 4
8
Hence the Taylor series is:
z z2 z4
f (z ) = log 2 − + − + ...
2 4 8
Observe that f (e12 π ) = log(0) is divergent. This is the singularity nearest the origin.
Therefore the radius of convergence of the series is π.
3.9 Separating fractions:
1 1 1
f (z ) = = −
z + 2 z − 8 6 (z − 2 ) 6 (z + 4 )
2
and developing the second fraction:

1
1 
2 3
1 1 6 z − 2  z − 2  z − 2
= = = 1 − +  −  + ... 
z+4 6+ z−2 z −2 6 6  6   6  
1+  
6
n (z − 2 )
∞ n
1
we find: f (z ) = − ∑ (− 1)
6( z − 2 ) n = 0 6 n+2
This Lauren series is convergent in the annulus 0 <z − 2< 6, which contains the
required annulus 1 <z − 2< 4 .
3.10 Taking into account that (1 − x) −1 = 1 + x + x2 + ... , the function defined by the series
is:
∞
1 1 1
f (z ) = ∑ n = −1=
n =1 4 ( z + 1) 1 4z +3
n
1−
4 (z + 1)
Now we see that at z = −3/4 the function has a pole. Therefore the radius of convergence
of this series (centred at z = −1) is −3/4 −( −1)=1/4.
3.11 From the successive derivatives of the sines calculated at z = 0, one obtains the
Taylor series for sin z:
z3 z5 ∞
z 2 n +1
− ... = ∑ (− 1)
n
sin z = z − +
3! 5 ! n =0 (2 n + 1)!
so the Lauren series for the given function is:
sin z ∞ z 2 n −1 1 z z3
= ∑ (− 1) n
= − + − ...
z2 n =0 (2 n + 1)! z 3! 5!
Since the pole z = 0 is the unique singularity (see that the analytical extension of the real
sine in the exercise 3.7 has no singularities), the annulus of convergence is 0<z< ∞ .
3.12 If f(z) is analytic then the derivative at each point is unique and we can write for two
different directions dz1 and dz2:
df1 = dz1 f'(z) df2 = dz2 f'(z)
According to the relationship between the complex and vectorial planes, we can multiply
by e1 at the left in order to turn the complex differentials into vectors:
e1 e1
df = da + db e12 → df = da e1 + db e 2 dz = dx + dy e12 → dz = dx e1 + dy e 2
Now let us calculate the geometric product of the vector differentials:
df1 df2 = dz1 f'(z) dz2 f’(z) = dz1 dz2 [f’(z)]* f'(z) =f'(z)2 dz1 dz2
Sincef'(z)2 is real, both geometric products have the same argument, α(df1, df2) = α(dz1,
dz2), and hence the transformation is conformal.
4. Transformations of vectors
4.1 If d is the direction vector of the reflection, the vector v' reflected of v with respect to
this direction is given by:
v' = d −1 v d
If v'' is obtained from v' by a rotation through an angle α, and z is a unitary complex
number with argument α /2 then:
α α
v'' = z −1 v' z z = cos + e12 sin
2 2
Joining both equations:
v'' = z −1 d −1 v d z = c −1 v c
That is, one obtains a reflection with respect another direction with vector c = d z, the
product of the direction vector of the initial reflection and the complex number with half
argument.
4.2 Let w'' be the transformed vector of w by two consecutive reflections with respect
different directions u and v:
w'' = v −1 w' v = v −1 u −1 w u v
Then we can write:
w'' = z −1 w z z=uv z being a complex number
so that it is equivalent to a rotation with an angle equal to the double of that formed by
both direction vectors.
4.3 If the product of each transformed vector v' by the initial vector v is equal to a
complex number z2 (v' and v always form a constant angle) then:
v v' = z2
v' = v −1 z2 = z* v −1 z = z −1 | z | 2 v −1 z = | z | 2 ( z −1 v z ) −1
which represents an inversion with radius | z |2 followed by a rotation with an angle equal
to the double of the argument of z. As the algebra shows, both elemental transformations
commute.
5. Points and straight lines
5.1 If A, B, C and D are located following this order on the perimeter of the parallelogram,
then AB=DC:
AB = DC ⇒ B−A=C−D ⇒ D=C−B+A
D = (2, −5) − (4, −3) + (2, 4) = (0, 2)
The area is: AB ∧ BC = ( 2 e1 −7 e2 ) ∧ ( −2 e1 −2 e2 ) = −18 e12 = 18
5.2 The Euler’s theorem:
AD BC + BD CA + CD AB = ( D −A ) ( C −B ) + ( D −B ) ( A −C ) + ( D −C ) ( B −A )
=CA−AC+AB−BA+BC−CB=2(C∧A+A∧B+B∧C)=
= 2 ( B − A ) ∧ ( C − B ) = 2 AB ∧ BC
This product only vanishes if A, B and C are collinear. On the other hand we see that the
product is the oriented area of the triangle ABC.
5.3 a) The area of the triangle ABC is the outer product of two sides:
AB = B − A = (4, 4) − (2, 2) = 2 e1 + 2 e2 AC = C − A = ( 4, 2) − (2, 2) = 2 e1
AB ∧ AC = ( 2 e1 + 2 e2 ) ∧ 2 e1 = 4 e2 ∧ e1
AB ∧ AC = 4 e 2 ∧ e1 = 4 e 2 e1 sin 60 o = 4 3
b) The distance from A to B is the length of the vector AB, etc:
AB2 = ( 2 e1 + 2 e2 )2 = 4 e12 + 4 e22 + 8 e1 · e2 = 4 + 16 + 16 cos π/3 = 28
AC2 = ( 2 e1 )2 = 4 BC2 = (−2 e2 )2 = 4 e22 = 16
d( A, B ) =  AB = 28 d( B, C ) = BC = 4 d( C, A) =  CA = 2
5.4 A side of the trapezoid is a vectorial sum of the other three sides:
AD = AB + BC + CD
AD2 = (AB +BC +CD )2 = AB2 +BC2 +CD2 +2 AB · BC + 2 AB · CD + 2 BC · CD
= AB2 + BC2 + CD2 + 2 AB · BC − 2ABCD+ 2 BC · CD
where the fact that AB and CD be vectors with the same direction and contrary sense has
been taken into account. Arranging terms:
AD2 − AB2 − BC2 − CD2 + 2 ABCD = 2 AB · BC + 2 BC · CD
AD2 − AB2 − BC2 − CD2 + 2 ABCD = 2 BC · ( AB + CD )
Without loss of generality we will suppose that AB>CD:
AB + CD = AB−CD
Since the angle ABC is the supplement of the angle formed by the vectors AB and BC, we
have:
AD2 −AB2 −BC2 −CD2 + 2 ABCD = −2 BC ( AB−CD ) cos ABC
whence we obtain the angle ABC:
− AD 2 + AB 2 + BC 2 + CD 2 − 2 AB CD
cos ABC =
2 BC ( AB − CD )
The trapezoid can only exist for the range −1< cos ABC <1, that is:
2 BC ( AB−CD ) > −AD2 + AB2 + BC2 + CD2 −2 ABCD >
> −2 BC (AB−CD)
5.5 R=(1−p−q)O+pP+qQ ⇒ RP = ( 1 − p − q ) OP + q QP
⇒ RP ∧ PQ = ( 1 − p − q ) OP ∧ PQ
whence it follows that: Area RPQ = ( 1 − p − q ) Area OPQ
5.6 The direction vector of the straight line r is AB:
AB = B − A = (5, 4) − (2, 3) = 3 e1 + e2 AC = C − A = (1, 6) − (2, 3) = − e1 + 3 e2
The distance from the point C to the line r is:
AC ∧ AB (− e1 + 3 e 2 ) ∧ (3 e1 + e 2 )
d (C , r ) = = = 10
AB 10
The angle between the vectors AB and AC is deduced by means of the sine and cosine:
AB ⋅ AC AB ∧ AC
cos α = =0 e12 sinα = = e12
AB AC AB AC
Therefore, α = π/2. The angle between two lines is always comprised from −π/2 to π/2
because a rotation of 2π around the intersection point does not alter the lines. When the
angle exceeds these boundaries, you may add or subtract π.
5.7 a) Three points D, E, F are aligned if they are linearly dependent, that is, if the
determinant of the coordinates vanishes.
D = (1 − xD − yD ) O + xD P + yD Q
E = (1 − xE − yE ) O + xE P + yE Q
F = (1 − xF − yF ) O + xF P + yF Q
1 − xD − yD xD yD
1 − xE − yE xE yE = 0
1 − xF − yF xF yF
where the barycentric coordinate system is given by the origin O and points P, Q (for
example the Cartesian system is determined by O = (0, 0), P = (1, 0) and Q = (0, 1)). The
transformed points D', E' and F' have the same coordinates expressed for the base O', P'
and Q'. Then the determinant is exactly the same, so that it vanishes and the transformed
points are aligned. Therefore any straight line is transformed into another straight line.
b) Let O', P' and Q' be the transformed points of O, P and Q by the given affinity:
O' = (o1 , o 2 ) P' = ( p1 , p 2 ) Q' = (q1 , q 2 )
and consider any point R with coordinates ( x, y ):
R = ( x, y ) = ( 1 − x − y ) O + x P + y Q
Then R', the transformed point of R, is:
R' = ( 1 − x − y ) O' + x P' + y Q' = ( 1 − x − y ) ( o1 , o2 ) + x ( p1 , p2 ) + y ( q1 , q2 ) =
= ( x ( p1 − o1 ) + y (q1 − o1 ) + o1 , x ( p2 − o2 ) + y ( q2 - o2 ) + o2 ) = ( x', y' )
where we see that the coordinates x' and y' of R' are linear functions of the coordinates of
R:
x' = x ( p1 − o1 ) + y ( q1 − o1 ) + o1
y' = x ( p2 − o2 ) + y ( q2 - o2 ) + o2
In matrix form:
 x'   p1 − o1 q1 − o1   x   o1 
  =    +  
 y'   p 2 − o 2 q 2 − o 2   y   o 2 
Because every linear (and non degenerate) mapping of coordinates can be written in this
regular matrix form, now we see that it is always an affinity.
c) Let us consider any three non aligned points A, B, C and their coordinates:
A = ( 1 − x A − y A ) O + x A P + y AQ
B = ( 1 − x B − y B ) O + x B P + y BQ
C = ( 1 − xC − y C ) O + xC P + yC Q
In matrix form:
 A  1 − x A − y A xA y A  O 
    
 B  = 1 − x B − y B xB yB   P 
 C  1 − x − y xC y C   Q 
   C C
A certain point D is expressed with coordinates whether for { O, P, Q } or { A, B, C }:
D = (1 − b − c ) A + b B + c C = (1 − x D − y D ) O + x D P + y D Q
In matrix form:
O   A
   
D = (1 − xD − yD xD y D ) P  = ( 1 − b − c b c) B 
Q  C 
   
1 − x A − y A xA yA 
 
(1 − xD − yD xD y D ) = ( 1 − b − c b c ) 1 − x B − y B xB yB 
1 − x − y xC y C 
 C C
which leads to the following system of equations:
 x D = ( 1 − b − c) x A + b x B + c xC

 y D = ( 1 − b − c) y A + b y B + c yC
An affinity does not change the coordinates x, y of A, B, C and D, but only the point base -
{O', P', Q'} instead of {O, P, Q}-. Therefore the solution of the system of equations for b
and c is the same. Then we can write:
D' = ( 1 − b − c ) A' + b B' + c C'
d) If the points D, E, F and G are the consecutive vertices in a parallelogram then:
DE = GF ⇒ G = D − E + F
The affinity preserves the coordinates expressed in any base {D, E, F}. Then the
transformed points form also a parallelogram:
G' = D' − E' + F' ⇒ D'E' = G'F'
e) For any three aligned points D, E, F the single ratio r is:
DE DF −1 = r ⇒ DE = r DF ⇒ E=(1−r)D+rF
The ratio r is a coordinate within the straight line DF and it is not changed by the affinity:
−1
E' = ( 1 − r ) D' + r F' ⇒ D'E' D'F' =r
5.8 This exercise is the dual of the problem 2. Then I have copied and pasted it changing
the words for a correct understanding.
a) Three lines D, E, F are concurrent if they are linearly dependent, that is, if the
determinant of the dual coordinates vanishes:
D = (1 − xD − yD ) O + xD P + yD Q
E = (1 − xE − yE ) O + xE P + yE Q
F = (1 − xF − yF ) O + xF P + yF Q
1 − xD − yD xD yD
1 − xE − yE xE yE = 0
1 − xF − yF xF yF
where the dual coordinate system is given by the lines O, P and Q. For example, the
Cartesian system is determined by O = [0, 0] (line −x −y +1=0), P = [1, 0] (line x = 0) and
Q = [0, 1] (line y = 0). The transformed lines D', E' and F' have the same coordinates
expressed for the base O', P' and Q'. Then the determinant also vanishes and the
transformed lines are concurrent. Therefore any pencil of lines is transformed into another
pencil of lines.
b) Let O', P' and Q' be the transformed lines of O, P and Q by the given transformation:
O' = [o1 , o 2 ] P' = [ p1 , p 2 ] Q' = [q1 , q 2 ]
and consider any line R with dual coordinates [x, y]:
R = [x, y] = ( 1 − x − y ) O + x P + y Q
Then R', the transformed line of R, is:
R' = ( 1 − x − y ) O' + x P' + y Q' = ( 1 − x − y ) [ o1 , o2 ] + x [ p1 , p2 ] + y [ q1 , q2 ] =
= [ x ( p1 − o1 ) + y (q1 − o1 ) + o1 , x ( p2 − o2 ) + y ( q2 − o2 ) + o2 ] = [ x', y' ]

where we see that the coordinates x' and y' of R' are linear functions of the coordinates of
R:
x' = x ( p1 − o1 ) + y ( q1 − o1 ) + o1
y' = x ( p2 − o2 ) + y ( q2 - o2 ) + o2
In matrix form:
 x '   p1 − o1 q1 − o1   x   o1 
 y ' =  p − o q 2 − o 2   y  + o 
   2 2    2
Because any linear mapping (non degenerate) of dual coordinates can be written in this
regular matrix form, now we see that it is always an affinity.
c) Let us consider any three non concurrent lines A, B, C and their coordinates:
A = ( 1 − x A − y A ) O + x A P + y AQ
B = ( 1 − x B − y B ) O + x B P + y BQ
C = ( 1 − xC − y C ) O + xC P + yC Q
In matrix form:
 A 1 − x A − y A xA y A  O 
 B  = 1 − x − y xB y B  P
   B B  
C  1 − x C − y C xC y C  Q 
A certain line D is expressed with dual coordinates whether for {O, P, Q } or {A, B, C }:
D = ( 1 − b − c) A + b B + c C = ( 1 − x D − y D ) O + x D P + y D Q
In matrix form:
O   A
D = [1 − x D − y D xD y D ] P = [1 − b − c b c ]  B 
 
   
Q  C 
1 − x A − y A xA yA
[1 − x D − y D xD y D ] = [1 − b − c b c ] 1 − x B − y B xB yB 
 
1 − x C − y C xC y C 
which leads to the following system of equations:
 x D = (1 − b − c ) x A + b x B + c x C

 y D = (1 − b − c ) y A + b y B + c y C
An affinity does not change the

Figure 16.1
coordinates x, y of A, B, C and D,
but only the lines base -{O', P', Q'}
instead of {O, P, Q}-. Therefore the
solution of the system of equations
for b and c is the same. Then we can
write:
D' = (1 − b − c ) A' + b B' + c C'
d) The affinity maps parallel points

into parallel points (points aligned
with the point (1/3, 1/3), point at the
infinity in the dual plane). If the
lines D, E, F and G are the
consecutive vertices in a dual
parallelogram (figure 16.1) then:
DE = GF ⇒ G = D − E + F
Where DE is the dual vector of the intersection point of the lines D and E, and GF the
dual vector of the intersection point of G and F. Obviously, the points EF and DG are also
parallel because from the former equality it follows:
EF = DG
The affinity preserves the coordinates expressed in any base {D, E, F}. Then the
transformed lines form also a dual parallelogram:
G' = D' − E' + F' ⇒ D'E' = G'F'
e) For any three concurrent lines D, E, F the single dual ratio r is:
DE DF −1 = r ⇒ DE = r DF ⇒ E=(1−r)D+rF
The ratio r is a coordinate within the pencil of lines DF and it is not changed by the
affinity:
−1
E' = ( 1 − r ) D' + r F' ⇒ D'E' D'F' =r
f) Let us consider four concurrent lines A, B, C and D with the following dual coordinates
expressed in the lines base {O, P, Q}:
A = [xA , yA] B = [xB , yB] C = [xC , yC] D = [xD , yD]

Then the direction vector of the line A is:
v A = ( 1 − x A − y A ) vO + x A v P + y A vQ
But the direction vectors of the base fulfils:
vO + v P + vQ = 0
Then:
v A = ( −1 + 2 x A + y A ) v P + ( − 1 + x A + 2 y A ) v Q
And analogously:
v B = (− 1 + 2 x B + y B ) v P + (− 1 + x B + 2 y B ) v Q
v C = (− 1 + 2 x C + y C ) v P + (− 1 + x C + 2 y C ) v Q
v D = (− 1 + 2 x D + y D ) v P + (− 1 + x D + 2 y D ) v Q
The outer product is:
v A ∧ v C = ( x C − x A − ( y C − y A ) + 3 ( x A y C − x C y A )) v P ∧ v Q
Then the cross ratio only depends on the dual coordinates but not on the direction of the
base vectors (see chapter 10):
( ABCD ) = =
( x C − x A − ( y C − y A ) + 3 ( x A y C − x C y A )) ( x D − x B − ( y D − y B ) + 3 ( x B y D − x D y B ))
=
( x D − x A − ( y D − y A ) + 3 ( x A y D − x D y A )) ( x C − x B − ( y C − y B ) + 3 (x B y C − x C y B ))
Hence it remains invariant under an affinity. Each outer product can be written as an outer
product of dual vectors obtained by subtraction of the coordinates of each line and the
infinite line:
 1 1  1 1
v A ∧ v C = 3  x A − , y A −  ∧  x C − , y C −  = 3 LA ∧ LC
 3 3  3 3
where LA = A − L is the dual vector going from the line L at the infinity to the line A, etc.
Then we can write this useful formula for the cross ratio of four lines:
( ABCD ) = LA ∧ LC LB ∧ LD
LA ∧ LD LB ∧ LC
This exercise links with the section Projective cross ratio in the chapter 10.
5.9 To obtain the dual coordinates of the first line, solve the identity:
x – y – 1 ≡ a' ( – x – y + 1) + b' x + c' y ⇒ a' = −1 b' = 0 c' = −2
1 2
⇒ a= b=0 c= ⇒ [0, 2/3]
3 3
For the second line:
x – y + 3 ≡ a' ( – x – y + 1) + b' x + c' y ⇒ a' = 3 b' = 4 c' = 2
3 4 2
⇒ a= b= c= ⇒ [4/9, 2/9]
9 9 9
Both lines are aligned in the dual plane with the line at the infinity (whose dual
coordinates are [1/3, 1/3]) since the determinant of the coordinates vanishes:
1/ 3 0 2/3
2/9 4/9 2/9 = 0
1/ 3 1/ 3 1/ 3
Therefore they are parallel (this is also trivial from the general equations).
5.10 The point (2, 1) is the intersection of the lines x – 2 = 0 and y – 1 = 0, whose dual
coordinates are:
x–2=0 ⇒ [1/5, 2/5]
y–1=0 ⇒ [1/2, 0]
The difference of dual coordinates gives a dual direction vector for the point:
v = [1/2, 0] – [1/5, 2/5] = [3/10, –4/10]
and hence we obtain the dual continuous and general equations for the point:
b − 1/ 2 c
= ⇔ 4 b + 3c = 2
3 −4
The point (–3, –1) is the intersection of the lines x + 3 = 0 and y + 1 = 0, whose dual
coordinates are:
x+3=0 ⇒ [4/10, 3/10]
y+1=0 ⇒ [1/4, 2/4]
The difference of dual coordinates gives a dual direction vector for the point:
w = [1/4, 2/4] – [4/10, 3/10] = [–3/20, 4/20]
and hence we obtain the dual continuous and general equations for the point:
b − 1/ 4 c − 2 / 4
= ⇔ 8b + 6c = 5
−3 4
Note that both dual direction vectors v and w are proportional and the points are parallel in
a dual sense. Hence the points are aligned with the centroid (1/3, 1/3) of the coordinate
system.
6. Angles and elemental trigonometry
6.1 Let a, b and c be the sides of a triangle taken anticlockwise. Then:
a+b+c=0 ⇒ c=−a−b ⇒ c2 = ( a + b )2 = a2 + b2 + 2 a · b
c2 = a2 + b2 + 2 a b cos(π − γ ) = a2 + b2 − 2 ab cos γ (law of

cosines)
The area s of a triangle is the half of the outer product of any pair of sides:
2s=a∧b=b∧c=c∧a
ab sin(π − γ ) e12 = bc sin(π − α) e12 = ca sin(π − β ) e12
sinγ sinα sinβ

= = (law of sines)
c a b
From here one quickly obtains:
a + b sinα + sinβ a − b sinα − sinβ

= =
a sinα b sinβ
By dividing both equations and introducing the identities for the addition and subtraction
of sines we arrive at:
α+β α −β
a + b 2 sin cos
sinα + sinβ 2 2
= =
a − b sinα − sinβ α+β α −β
2 cos sin
2 2
α+β
a + b tg
= 2
(law of tangents)
a − b α−β
tg
2
6.2 Let us substitute the first cosine by the half angle identity and convert the addition of
the two last cosines into a product:
α−β β −α β − 2γ + α
cos(α − β ) + cos(β − γ ) + cos(γ − α ) = 2 cos 2 − 1 + cos cos
2 2 2
Let us extract common factor and convert the addition of cosines into a product:
α−β  α−β α − 2γ + β 
= 2 cos  cos + cos  −1
2  2 2 
α−β α −γ γ −β
= 2 cos cos cos −1
2 2 2
= 2 cos cos cos −1
2 2 2
The identity for the sines is proved in a similar way.
6.3 Using the De Moivre’s identity:
cos 4α + e12 sin 4α ≡ (cos α + e12 sin α )

4
After developing the right hand side we find:
sin 4α ≡ 4 cos3α sinα − 4 cosα sin3α cos 4α ≡ cos4α − 6 cos2α sin2α + sin4α
4 tg α − 4 tg 3 α
And dividing both identities: tg 4α =
1 − 6 tg 2 α + tg 4 α
6.4 Let α, β and γ be the angles of the triangle PAB with vertices A, B and P respectively.
The angle γ embracing the arc AB is constant for any point P on the arc AB. By the law of
sines we have:
PA PB PC
= =
sinα sinβ sinγ
The sum of both chords is:
sinα + sinβ
PA + PB = PC
sinγ
Converting the sum of sines into a product we find:
α+β α−β π −γ α−β

2 sin cos 2 sin cos
PA + PB = PC 2 2 = PC 2 2
sinγ sinγ
Since γ is constant, the maximum is attained for cos(α /2 − β /2)=1, that is, when the
triangle PAB is isosceles and P is the midpoint of the arc AB.
6.5 From the sine theorem we have:
α+β α−β α−β

a + b 2 sin cos cos
sinα + sinβ 2 2 = 2
= =
c sinγ γ γ γ
2 sin cos sin
2 2 2
Following the same way, we also have:
α+β α−β α+β

a − b 2 cos sin sin
sinα − sinβ 2 2 = 2
= =
c sinγ γ γ γ
2 sin cos cos
2 2 2
6.6 Take a as the base of the triangle and draw the altitude. It divides a in two segments
which are the projections of the sides b and c on a:
a = b cos γ + c sinβ and so forth.
6.7 From the double angle identities we have:
α α α α
cos α ≡ cos 2 − sin 2 ≡ 1 − 2 sin 2 ≡ 2 cos 2 − 1
2 2 2 2
From the last two expressions the half-angle identities follow:
α 1 − cos α α 1 + cos α
sin ≡± cos ≡±
2 2 2 2
Making the quotient:

1 − cos α 1 − cos α sinα

tg α ≡ ± ≡ ≡
1 + cos α sinα 1 + cosα
7. Similarities and single ratio
7.1 First at all draw the triangle and see that the homologous vertices are:
P = ( 0, 0 ) P' = ( 4, 2 )
Q = ( 2, 0 ) Q' = ( 2, 0 )
R = ( 0, 1 ) R' = ( 5, 1 )
PQ = 2 e1 QR = −2 e1 + e2 RP = −e2
P'Q' = −2 e1 + 2 e2 Q'R' = 3 e1 + e2 R'P' = −e1 + e2
r = P'Q' PQ −1 = Q'R' QR −1 = R'P' RP −1 = −1 − e12 = 2 5π / 4
The size ratio is r = 2 and the angle between the directions of homologous sides is
5π/4.
7.2 Let ABC be a right triangle being C the right angle. The altitude CD cutting the base
AB in D splits ABC in two right angle triangles: ADC and BDC. In order to simplify I
introduce the following notation:
AB = c BC = a CA = b AD = x DB = c − x
The triangles CBA and DCA are oppositely similar because the angle CAD is common and
the other one is a right angle. Hence:
b x −1 = ( c b −1 )* = b −1 c ⇒ b2 = c x
The triangles DBC and CBA are also oppositely similar, because the angle DBC is
common and the other one is a right angle. Hence:
a ( c − x ) −1 = ( c a −1 )* = a −1 c ⇒ a2 = c ( c − x )
Summing both results the Pythagorean theorem is obtained:
a2 + b2 = c ( c − x ) + c x = c2
7.3 First at all we must see that the triangles ABM and BCM are oppositely similar. Firstly,
they share the angle BMC. Secondly, the angles MBC and BAM are equal because they
embrace the same arc BC. In fact, the limiting case of the angle BAC when A moves to B
is the angle MBA. Finally the angle BCM is equal to the angle ABM because the sum of
the angles of a triangle is π. The opposite similarity implies:
MA AB −1 = BC −1 MB ⇒ MA = BC −1 MB AB
MB AB −1 = BC −1 MC ⇒ MC −1 = AB MB −1 BC −1
Multiplying both expressions we obtain:
MA MC −1 = BC −1 MB AB2 MB −1 BC −1 = AB2 BC-2
7.4 Let Q be the intersection point of the segments PA and BC. The triangle PQC is
directly similar to the triangle PBA and oppositely similar to the triangle BQA:
BA PA −1 = QC PC −1 ⇒ QC = BA PA −1 PC ⇒ QC = BA PA−1 PC
BQ BA −1 = PA −1 PB ⇒ BQ = PA −1 PB BA ⇒ −1
BQ = PA PB BA
Summing both expressions: BC = BQ + QC = BA ( PB + PC ) PA −1
The three sides of the triangle are equal; therefore: PA = PB + PC
7.5 Let us firstly calculate the homothety ratio k, which is the quotient of homologous
sides and the similarity ratio of both triangles:
AB −1 A'B' = k
Secondly we calculate the centre of the homothety by isolation from:
OA' = OA k
OA' − OA = OA ( 1 − k )
AA' ( 1 − k ) −1 = OA
Since the segment AA' is known, this equation allows us to calculate the centre O :
O = A − AA' ( 1 − k ) −1 = A − AA' ( 1 − AB −1 A'B' ) −1
7.6 Draw the line passing through P and the centre

of the circle. This extended diameter cuts the circle Figure 16.2
in the points R and R'. See that the angles R'RQ and
QQ'R' are supplementary because they intercept
opposite arcs of the circle. Then the angles PRQ
and R'Q'P are equal. Therefore the triangle QRP
and R'Q'P are oppositely similar and we have:
PR −1 PQ = PR' PQ' −1 ⇒ PQ PQ' = PR PR'
Since PR and PR' are determined by P and the

circle, the product PQ PQ' (the power of P) is constant independently of the line PQQ'.
7.7 The bisector d of the angle ab divides the triangle abc in two triangles, which are
obviously not similar! However we may apply the law of sines to both triangles to find:
m a n b
= =
sin ad sin dm sin db sin nd
The angles nd and dm are supplementary and sin nd = sin dm. On the other hand, the
angles ad and db are equal because of the angle bisector. Therefore it follows that:
m n
=
a b
8. Properties of the triangles
8.1 Let A, B and C be the vertices of the given triangle with anticlockwise position. Let S,
T and U be the vertices of the three equilateral triangles drawn over the sides AB, BC and
CA respectively. Let P, Q and R be the centres of the triangles ABS, BCT and CAU
respectively. The side CU is obtained from AC through a rotation of 2π/3:
2π 2π
CU = AC t with t = 12π / 3 = cos + e12 sin
3 3
CR is 2/3 of the altitude of the equilateral triangles ACU ; therefore is 1/3 of the
diagonal of the parallelogram formed by CA and CU:
CA + CU CA ( 1 − t )
CR = =
3 3
The same argument applies to the other equilateral triangles:
BC ( 1 − t ) AB ( 1 − t )
BQ = AP =
3 3
From where P, Q and R as functions of A, B and C are obtained:
AB ( 1 − t ) BC ( 1 − t ) CA ( 1 − t )
P = A+ Q=B+ R=C +
3 3 3
Let us calculate the vector PQ:
( C − B − B + A)( 1 − t )
PQ = Q − P = B − A +
3
A+ B +C
Introducing the centroid: G=
3
PQ = AB + BG ( 1 − t ) = AG – BG t
Analogously:
QR = BG – CG t RP = CG – AG z
Now we apply a rotation of 2π/3 to PQ in order to obtain QR:
PQ t = ( AG – BG t ) t = AG t – BG t2
Since t is a third root of the unity (1 + t + t2 = 0) then:
PQ t = AG t + BG + BG t = AG t + BG t + CG t + BG − CG t
= 3 GG + BG – CG t = BG – CG t = QR
Through the same way we find:
QR t = RP RP t = PQ
Therefore P, Q and R form an equilateral triangle with centre in G, the centroid of the
triangle ABC:
P + Q + R A + B + C AB + BC + CA
= + (1 − t ) = G
3 3 3
8.2 The substitution of the centroid G of the triangle ABC gives:

3 PG = 3  P −
2 A+ B + C
2
= 3 P 2 − 2 P · (A + B + C ) +
(A + B + C ) 2
 3 
 3
A + B + C + 2 A· B + 2 B ·C + 2C · A
2 2 2
= 3 P 2 − 2 P · (A + B + C ) +
3
2 (− A − B 2 − C 2 + A · B + B · C + C · A)
[ ]
2
= ( A − P ) + (B − P ) + (C − P ) +
2 2 2
= PA 2 + PB 2 + PC 2 −
(B − A) + (C − B ) + ( A − C )
2 2 2
3
AB + BC + CA 2
2 2
= PA 2 + PB 2 + PC 2 −
3
8.3 a) Let us calculate the area of the triangle GBC:
GB ∧ BC = [ − a A + ( 1 − b ) B − c C ] ∧ BC = [ − a A + ( a + c ) B − c C ] ∧ BC =
GB ∧ BC
= ( a AB + c CB ) ∧ BC = a AB ∧ BC ⇒ a=
AB ∧ BC
The proof is analogous for the triangles GCA and GAB.
b) Let us develop PG2 following the same way as in the exercise 8.2:
PG2 = [P − ( a A + b B + c C ) ]2 = P2 − 2 P · ( a A + b B + c C ) + ( a A + b B + c C )2
= ( a + b + c ) P2 − 2 a P · A − 2 b P · B − 2 c P · C + a2 A2 + b2 B2 + c2 C 2 +
+2abA·B+2bcB·C+2caC·A
= a (A − P )2 + b ( B − P )2 + c ( C − P )2 + a ( a − 1 ) A2 + b ( b − 1) B2
+ c ( c − 1 ) C2 + 2 a b A · B + 2 b c B · C + 2 c a C · A
= a PA2 + b PB2 + c PC2 − a ( b + c ) A2 − b ( c + a ) B2 − c ( a + b ) C 2
+2abA·B+2bcB·C+2caC·A
= a PA2 + b PB2 + c PC2 − a b ( B − A )2 − b c ( C − B )2 − c a ( A − C )2
= a PA2 + b PB2 + c PC2 − a b AB2 − b c BC2 − c a CA2
The Leibniz’s theorem is a particular case of the Apollonius’ lost theorem for a = b = c =
1/3.
8.4 Let us consider the vertices A, B, C and D ordered clockwise on the perimeter. Since
AP is AB turned π/3, BQ is BC turned π/3, etc, we have:
z = cos π/3 + e12 sin π/3
AP = AB z BQ = BC z CR = CD z DS = DA z
PR · QS = ( PA + AC + CR ) · ( QB + BD + DS )
= [ ( CD − AB ) z + AC ] · [ ( DA − BC ) z + BD ]
= [ ( −AC + BD ) z + AC ] · [ ( −AC − BD ) z + BD ] =
= [ ( −AC + BD ) z ] · [ ( −AC − BD ) z ] + [ (−AC + BD ) z ] · BD
+ AC · [ ( −AC − BD ) z + BD ]
The inner product of two vectors turned the same angle is equal to that of these
vectors before the rotation. We use this fact for the first product. Also we must develop
the other products in geometric products and permute vectors and the complex number z:
z + z*
PR · QS = ( −AC + BD ) · ( −AC − BD ) + ( −AC BD − BD AC − AC2 + BD2 )
2
AC − BD
2 2
+ AC · BD =
2
Therefore AC = BD ⇔ PR · QS = 0. The statement b) is proved through an

analogous way.
8.5 The median is the segment going from a vertex to the midpoint of the opposite side:
2 2
B+C   AB AC  AB 2 AC 2 AB · AC
m 2A =  − A =  +  = + +
 2   2 2  4 4 2
AB 2 AC 2 2 AB · AC − AB 2 − AC 2 AB 2 AC 2 BC 2
= + + = + −
2 2 4 2 2 4
8.6 If E is the intersection point of the bisector of B with the line parallel to the bisector of
A, then the following equality holds:
 BA BC   AB AC 
E = B + b  + =C +a
  AB + 

 BA BC   AC 
where a, b are real. Arranging terms we obtain a vectorial equality:
 AB AC   BA BC 
− a  + +b
  +  = BC

 AB AC   BA BC 
The linear decomposition yields after simplification:
BC CA
a=−
AB + BC + CA
In the same way, if D is the intersection of the bisector of A with the line parallel to the
bisector of B we have:
D = A + c  + =C +d  + 
 AB AC   BA
 BC 
where c and d are real. Arranging terms:
− c  + +d


 BA +  = CA

 AB AC   BC 
After simplification we obtain:
BC CA
d =−
AB + BC + CA
Now we calculate the vector ED:

 2 AB BC AC  BC CA
ED = D − E =  − + 
 AB BC AC  AB + BC + CA
Then the direction of the line ED is given by the vector v:
2 AB BC AC  2 1   1 1 
v= − + = AB  +  + BC  −
  + 

AB BC AC  AB AC   BC AC 
When the vector v has the direction AB, the second summand vanishes, AC = BC
and the triangle becomes isosceles.
8.7 Let us indicate the sides of the triangle ABC with a, b and c in the following form:
a = BC b = CA c = AB
Suppose without loss of generality that P lies on the side BC and Q on the side AC .
Hence:
P = k B + (1 − k ) C Q = l A + (1 − l ) C
CP = k CB = − k a CQ = l CA = l b
where k and l are real and 0 < k, l < 1. Now the segment PQ is obtained:
PQ = k a + l b
Since the area of the triangle CPQ must be the half of the area of the triangle ABC, it
follows that:
1
CP ∧ CQ = CB ∧ CA ⇒ (− k a ) ∧ (l b ) = − 1 a ∧ b ⇒ kl =
1
2 2 2
The substitution into PQ gives:
1
PQ = k a + b
2k
a) If u is the vector of the given direction, PQ is perpendicular when PQ · u = 0 which

results in:
 1 
k a + b · u = 0
 2k 
whence one obtains:

b·u b·u a ·u
k= − PQ = a − +b −
2a ·u 2a ·u 2b ·u
In this case the solutions only exist when a · u and b · u have different signs. However we
can also choose the point P (or Q) lying on the another side c, what gives an analogous
solution containing the inner product c · u . Note that if a · u and b · u have the same sign,
then c · u have the opposite sign because:
a+b+c=0 ⇒ c·u=−a·u−b·u
what warrants there is always a solution.
b) Let us calculate the square of PQ:

1
PQ 2 = k 2 a 2 + 2 b 2 + a ⋅ b
4k
By equating the derivative to zero, one obtains the value of k for which PQ2 is minimum:
b
k=
2 a
Then we obtain the segment PQ and its length:
b a
PQ = a +b PQ = a b + a·b
2 a 2 b
c) Every point may be written as linear combination of the three vertices of the triangle
the sum of the coefficients being equal to the unity:
R = x A + y B + (1 − x − y ) C ⇔ CR = x CA + y CB
where all the coefficients are comprised between 0 and 1:
CR ∧ CB CA ∧ CR
0 < x, y, 1 − x − y < 1 x= y=
CA ∧ CB CA ∧ CB
Now the point R must lie on the segment PQ, that is, P, Q and R must be aligned. Then
the determinant of their coordinates will vanish:
0 k 1− k 0 k 1− k
1 1
det (P, Q , R ) = l 0 1− l = 0 1− =0
2k 2k
x y 1− x − y x y 1− x − y
1± 1−8x y 1 1m 1−8x y
2xk2 − k + y =0 ⇒ k= and l= =
4x 2k 4y
There is only solution for a positive discriminant:
1>8xy
The limiting curve is an equilateral hyperbola on the plane x-y. Since the triangle may be
obtained through an affinity, the limiting curve for the triangle is also a hyperbola
although not equilateral.
If the extremes P and Q could move along the prolongations of the sides of the
triangle without limitations, this would be the unique condition. However in this problem
the point P must lie between C and B, and Q must lie between C and A. It means the
additional condition:
1 1 1 1± 1−8x y
< k <1 ⇔ < l <1 ⇔ < <1
2 2 2 4x
From the first solution (root with sign +) we have:
2 x −1< 1− 8 x y < 4 x −1
only existing for x >1/4. For 1/4 < x < 1/2 the left hand side is negative so that the unique
restriction is the right hand side. By squaring it we obtain:
1 1
1 − 8 x y < (4 x − 1)
2
⇒ 0 < 2 x + y −1 for <x<
4 2
For x >1/2 the right hand side is higher than 1 so that the unique restriction is the left hand
side:
1
(2 x − 1)2 < 1 − 8 x y ⇒ x + 2 y −1< 0 for x>
2
For the second solution (root with sign −) we have:
2 x −1< − 1− 8 x y < 4 x −1
only existing for x < 1/2. For 1/4 < x < 1/2 the right hand side is positive so that the
unique restriction is the left hand side:
1 1
(2 x − 1)2 > 1 − 8 x y ⇒ x + 2 y −1> 0 for <x<
4 2
For x < 1/4 both members are negative and after squaring we have:
1
(2 x − 1)2 > 1 − 8 x y > (4 x − 1)2 ⇒ x + 2 y −1> 0 > 2 x + y −1 for x <
4
All these conditions are plotted in the figure 16.3.
Figure 16.3
It is easily proved that the limiting lines touch the hyperbola at the points (1/4, 1/2)
and (1/2, 1/4). For both triangles painted with light grey there is a unique solution. In the
region painted with dark grey there are two solutions, that is, there are two segments
passing through the point R and dividing the triangle in two parts with equal area. By
means of an affinity the plot in the figure 16.3 is transformed into any triangle (figure
16.4). The affinity is a linear transformation of the coordinates, which converts the
limiting lines into the medians of the triangle intersecting at the centroid and the
hyperbola into another non equilateral hyperbola:
Figure 16.4
Observe that the medians divide the triangle ABC in six triangles. If a point R lies outside
the shadowed triangles, we can always support the extremes of the segment PQ passing
through R in another pair of sides. This means that the division of the triangle is always
possible provided of the suitable choice of the sides. On the other hand, if we consider all
the possibilities, the points neighbouring the centroid admit three segments.
9. Circles
9.1 ( A − M )2 + ( B − M )2 + ( C − M )2 = k
3 M2 − 2 M · ( A + B + C ) + A2 + B2 + C2 = k
Introducing the centroid G = ( A + B + C ) / 3:
k 2 ( A 2 + B 2 + C 2 + A · B + B · C + C · A)
M 2 − 2 M ·G + G 2 = −
3 9
( M − G ) 2 = k − ( B − A) + (C − B ) + ( A − C )
2 2 2
3 9
k AB 2 + BC 2 + CA 2
GM 2 = − = r2
3 9
M runs on a circumference with radius r centred at the centroid of the triangle when k is
higher than the arithmetic mean of the squares of the three sides:
AB 2 + BC 2 + CA 2
k≥
3
9.2 Let A and B be the fixed points and P any point on the searched geometric locus. If the
ratio of distances from P to A and B is constant it holds that:
PA = k PB ⇔ PA2 = k2 PB2 with k being a real positive number
A2 − 2 A · P + P2 = k2 ( B2 − 2 B · P + P2 )
( 1 − k2 ) P2 − 2 ( A − k2 B ) · P = k2 B2 − A2
2 P · ( A − k 2 B) k 2 B 2 − A2
P2 − =
1− k2 1− k2
A − k 2B
We indicate by O the expression: O = . Adding O2 to both members we obtain:
1− k2
k 2 B 2 − A 2 (A − k 2 B )
2
OP = 2
+
1− k2 (1 − k 2 )2
k 2 ( B 2 − 2 A · B + A2 ) k 2 AB 2
OP 2 = =
(1 − k )
2 2
(1 − k )
2 2
Therefore, the points P form a circumference centred at O, which is a point of the line AB,
with radius:
k AB
OP =
1− k2
9.3 Let us draw the circumference

circumscribed to the triangle ABC
(figure 16.5) and let D be any point
on this circumference. Let P, Q and R
be the orthogonal projections of P on
the sides AB, BC and CA
respectively. Note that the triangles
ADP and CDR are similar because of
DAB = π − BCD = DCR . The
similarity of both triangles is written
as:
Figure 16.5
−1 −1
DC DA = DR DP
The triangles ADQ and BDR are also similar because of the equality of the inscribed
angles CAD = CBD:
DB DA −1 = DR DQ −1
Since the angles RDC = PDA and RDB = QDA, the bisector of the angle BDQ is
also bisector of the angle CDP and RDA, being its direction vector:
DA DR DB DQ DC DP
v= + = + = +
DA DR DB DQ DC DP
Let P', Q' and R' be the reflected points of P, Q and R with respect to the bisector:
DR' = v −1 DR v
Multiplying by DA we have:
DR' DA = v −1 DR v DA = DR DA
which is an expected result since the reflection of R with respect the bisector of the angle
RDA yields always a point R' aligned with D and A. Analogously for B and C we find that
the points D, B and Q' are aligned and also the points D, C and P':
DQ' DB = v −1 DQ v DR DQ −1 DA = v −1 DR v DA = DR DA

DP' DC = v −1 DP v DR DP −1 DA = v −1 DR v DA = DR DA
Now we see that the points R', Q' and P' are the transformed of A, B and C under
an inversion with centre D:
DR' DA = DQ' DB = DP' DC
Since D lies on the circumference passing through A, B and C, the inversion transforms
them into aligned points and the reflection preserves this alignment so that P, Q and R are
aligned.
9.4 If m = a + b and n = b +c then:
m 2 n 2 = (a + b ) (b + c ) = (a 2 + a b + b a + b 2 )(b 2 + b c + c b + c 2 )
2 2
= a 2b2 + a 2b c + a 2c b + a 2c 2 + a b3 + a b2c + a b c b + a b c 2 +
+ b a b 2 + b a b c + b a c b + b a c 2 + b 4 + b3c + b 2 c b + b 2 c 2
Taking into account that d = −(a + b + c) then:
d 2 = (a + b + c ) = a 2 + b 2 + c 2 + a b + b a + b c + c b + c a + a c
2
so we may rewrite the former equality:
m2n 2 = a 2c 2 + b2d 2 + a 2b c + a 2c b + a b c b + a b c 2 + b a b c + b a c 2
We apply the permutative property to some terms:
m2n 2 = a 2c 2 + b2d 2 + a 2b c + a b c a + a b c b + a b c 2 + b a b c + c a b c
in order to arrive at a fully symmetric expression:
m2n2 = a 2c2 + b2d 2 − d a b c − a b c d
Now we express the product of each pair of vectors using the exponential function of their
angle:
m2n2 = a 2c2 + b2d 2 − a b c d (exp((2π − γ − α ) e12 ) + exp((2π − β − δ ) e12 ))
Adding the arguments of the exponential, simplifying and taking into account that α + β +
γ + δ = π.
m2n2 = a 2c2 + b2d 2 − 2 a b c d cos(α + γ )

9.5 Let A, B and C be the midpoints of the sides of any triangle PQR :
Q+R R+P P+Q

A= B= C=
2 2 2
The centre of the circumferences RAB, PBC and QCA will be denoted as D, E and F
respectively. The circumference PBC is obtained from the circumscribed circumference
PQR by means of a homothety with centre P. Then if O is the circumcentre of the triangle
PQR, the centre of the circumference PBC is located at half distance from R:
P+O Q +O
E= F=
2 2
PQ OR PQ OR PQ
EA = + FB = − + EF =
2 2 2 2 2
Let Z be the intersection of EA with FB. Then:
Z = a A + ( 1 − a ) E = b B + ( 1 − b) F
Arranging terms we find:
a EA − b FB = EF
The linear decomposition gives:
EF ∧ FB EA ∧ EF
a= −b=
EA ∧ FB EA ∧ FB
EF ∧ FB FA ∧ FB A E P+Q + R+O
Z= A+ E= + =
EA ∧ FB EA ∧ FB 2 2 4
which is invariant under cyclic permutation of the vertices P, Q and R. Then the lines EA,
FB and DC intersect in a unique point Z. On the other
hand, the point Z lies on the Euler’s line (figure 9.8): Figure 9.8
3G + O
Z=
4
9.6 We must prove that the inversion changes the single ratio of three very close points by
its conjugate value, because the orientation of the angles is changed. Let us consider an
inversion with centre O and radius r and let the points A', B', C' be the transformed of A,
B, C under this inversion. Then:
OA' = r2 OA−1 OB' = r2 OB −1 OC' = r2 OC −1
A'B' = OB' − OA' = r2 (OB −1 − OA −1) = r2 OA −1 (OA − OB) OB −1

= − r2 OA −1 AB OB −1
In the same way: A'C' = − r2 OA −1 AC OC −1

Let us calculate the single ratio of the transformed points:
( A', B', C' ) = A'B' A'C' −1 = OA −1 AB OB −1 OC AC −1 OA =
= OB −1 AB AC −1 OC = ( OA + AB ) −1 AB AC −1 ( OA + AC )
When B and C come near A, AB and AC tend to zero and the single ratio for the inverse
points becomes the conjugate value of that for the initial points:
lim ( A' , B' , C' ) = OA −1 AB AC −1 OA = AC −1 AB = ( A, B, C ) *

B, C → A
10. Cross ratios and related transformations
10.1 Let ABCD be a quadrilateral inscribed in a circle. We must prove the following
equality:
BC AD + AB CD=BD AC
which is equivalent to:
BC AD AB CD
+ =1
BD AC BD AC
Now we identify these quotients with cross ratios of four points A, B, C and D on a circle:
(BACD)+(BCAD)=1
equality that always holds because if ( A B C D ) = r then:
1 r −1
+ =1
r r
When the points do not lie on a circle, the cross ratio is not the quotient of moduli
but a complex number:
BC AD AB CD
exp[e12 (CBD − CAD )] + exp[e12 ( ABD − ACD )] = 1
BD AC BD AC
According to the triangular inequality, the modulus of the sum of two complex
numbers is lower than or equal to the sum of both moduli:
BC AD AB CD
+ ≥1 ⇒ BC AD + AB CD ≥ BD AC
BD AC BD AC
10.2 If A, B, C and D form a harmonic range their cross ratio is 2:
AC AD −1 BD BC −1 = 2
In order to isolate D we must write BD as addition of BA and AD:
AC AD −1 ( BA + AD ) BC −1 = 2
AC AD −1 BA BC −1 = 2 − AC BC −1
AD −1 = AC −1 ( 2 − AC BC −1 ) BC BA −1 = ( 2 AC −1 BC −1 ) BA −1
AD = AB ( 1 − 2 AC −1 BC ) −1
10.3 The homography preserves the cross ratio:
A'C' A'D' −1 B'D' B'C' −1 = AC AD −1 BD BC −1
which may be rewritten as:
A'C' A'D' −1 ( B'A' + A'D' ) ( B'A' + A'C' ) −1 = AC AD −1 ( BA + AD ) ( BA + AC ) −1
When C and D approach to A, the vectors AC, AD, A'C' and A'D' tend to zero. So the
single ratio of three very close points remains constant:
lim A'C' A'D' −1 = lim AC AD −1

C, D → A C, D → A
and therefore the angle between tangent vectors of curves. So the homography is a
directly conformal transformation.
10.4 If F is a point on the homology axis we must prove that:
{ F, A B C D } = { F, A' B' C' D' }
As we have seen, a point A' homologous of A is obtained through the equation:
OA' = OA [ 1 − r v ∧ FA ( v ∧ FO ) −1 ] −1
where O is the centre of homology, F a point on the axis, v the direction vector of the axis
and r the homology ratio. Then the vector FA' is:
FA' = FO + OA' = FO + OA [ 1 − r v ∧ FA ( v ∧ FO ) −1 ] −1
= [ FO ( 1 − r v ∧ FA ( v ∧ FO ) −1 ) + OA ] [ 1 − r v ∧ FA ( v ∧ FO ) −1 ] −1
= [ FA − r FO v ∧ FA ( v ∧ FO ) −1 ] [ 1 − r v ∧ FA ( v ∧ FO ) −1 ] −1
Analogously:
FB' = [ FB − r FO v ∧ FB ( v ∧ FO ) −1 ] [ 1 − r v ∧ FB ( v ∧ FO ) −1 ] −1
FC' = [ FC − r FO v ∧ FC ( v ∧ FO ) −1 ] [ 1 − r v ∧ FC ( v ∧ FO ) −1 ] −1
FD' = [ FD − r FO v ∧ FD ( v ∧ FO ) −1 ] [ 1 − r v ∧ FD ( v ∧ FO ) −1 ] −1
The product of two outer products is a real number so that we must only pay attention to
the order of the first loose vectors FA and FO, etc:
FA' ∧ FC' = [ FA ∧ FC − r ( FA ∧ FO v ∧ FC + FO ∧ FC v ∧ FA ) ( v ∧ FO ) −1 ]
[ 1 − r v ∧ FA ( v ∧ FO ) −1 ] −1 [ 1 − r v ∧ FC ( v ∧ FO ) −1 ] −1
Using the identity of the exercise 1.4, we have:
FA' ∧ FC' = [ FA ∧ FC − r FA ∧ FC v ∧ FO ( v ∧ FO ) −1 ]
[ 1 − r v ∧ FA ( v ∧ FO ) −1 ] −1 [ 1 − r v ∧ FC ( v ∧ FO ) −1 ] −1
= FA ∧ FC ( 1 − r ) [ 1 − r v ∧ FA ( v ∧ FO ) −1 ] −1 [ 1 − r v ∧ FC ( v ∧ FO ) −1 ] −1
In the projective cross ratio all the factors except the first outer product are simplified:
FA' ∧ FC' FB' ∧ FD' FA ∧ FC FB ∧ FD

=
FA' ∧ FD' FB' ∧ FC' FA ∧ FD FB ∧ FC
10.5 a) Let us prove that the special conformal transformation is additive:
OP' = ( OP −1 + v ) −1 ⇒ OP' −1 = OP −1 + v
OP'' = ( OP' −1 + w ) −1 = ( OP −1 + v + w ) −1
b) Let us extract the factor OP from OP':
OP' −1 = ( 1 + v OP ) OP −1 ⇒ OP' = OP ( 1 + v OP ) −1
Then we have:
OA' = OA ( 1 + v OA ) −1 OB' = OB ( 1 + v OB ) −1
OC' = OC ( 1 + v OC ) −1 OD' = OD ( 1 + v OD ) −1
A'C' = OC' − OA' = OC ( 1 + v OC ) −1 − OA ( 1 + v OA ) −1 =
= [ OC ( 1 + v OA ) − OA ( 1 + v OC ) ] ( 1 + v OC ) −1 ( 1 + v OA ) −1
= AC ( 1 + v OC ) −1 ( 1 + v OA ) −1
Analogously:
B'D' = BD ( 1 + v OB ) −1 ( 1 + v OD ) −1
A'D' = AD ( 1 + v OA ) −1 ( 1 + v OD ) −1
B'C' = BC ( 1 + v OB ) −1 ( 1 + v OC ) −1
From where it follows that the complex cross ratio is preserved and it is a special case of
homography:
A'C' A'D' −1 B'D' B'C' −1 = AC ( 1 + v OC ) −1 ( 1 + v OD ) AD −1
BD ( 1 + v OD ) −1 ( 1 + v OC ) −1 BC −1 = AC AD −1 BD BC −1
10.6 If the points A, B and C are invariant under a certain homography, then for any other
point D' the following equality is fulfilled:
( A B C D ) = ( A B C D' )
AC AD −1 BD BC −1 = AC AD' −1 BD' BC −1
The simplification of factors yields:
AD −1 BD = AD' −1 BD' ⇒ AD −1 ( BA + AD ) = AD' −1 ( BA + AD' ) ⇒

−1 −1
⇒ AD BA = AD' BA ⇒ AD −1 = AD' −1 ⇒ D = D'
10.7. a) From the definition of antigraphy we have:
A'C' A'D' −1 B'D' B'C' −1 = ( AC AD −1 BD BC −1 )* = BC −1 BD AD −1 AC
If C and D approach A, then BC, BD → BA:
lim A' C ' A' D' −1 = lim AD −1 AC

C' , D' → A C, D → A
So the antigraphy is an opposite conformal transformation.

b) The composition of two antigraphies is always a homography because twice the
conjugation of a complex number is the identity, so that the complex cross ratio is
preserved.
c) The inversion conjugates the cross ratio. Consequently an odd number of inversions is
always an antigraphy.
d) If an antigraphy has three invariant points, then all the points lying on the
circumference passing through these points are invariant. Let us prove this statement: if A,
B and C are the invariant points and D belongs to the circumference passing through these
points, then the cross ratio is real:
( A B C D ) = ( A B C D )* = ( A B C D' )
from where D = D' as proved in the exercise 10.6. If D does not belong to this
circumference, D' ≠ D and the cross ratio ( A B C D ) becomes conjugate. A geometric
transformation that preserves a circumference and conjugates the cross ratio can only be
an inversion because there is a unique point D' for each point D fulfilling this condition.
The centre and radius of inversion are those for the circumference ABC.
10.8 Consider the projectivity:
D=aA+bB+cC → D' = a' A' + b' B' + c' C'
If A', B' and C' are the images of A, B and C then the coefficients a', b', c' must be
proportional to a, b, c:
ka lb mc
a' = b' = c' =
k a+lb+mc ka+lb+mc ka+lb+mc
where the values are already normalised. Three points are collinear if the determinant of
the barycentric coordinates is zero, what only happens if the initial points are also
collinear:
aD bD cD
k l m aE aE aE
a' D b' D c' D
aF aF aF
a' E a' E a' E = =0
(k a D + l bD + m c D ) (k a E + l bE + m c E )(k a F + l bF + m c F )
a' F a' F a' F
10.9 In the dual plane the six sides of the hexagon are six points. Since the sides AB, CD
and EF (or their prolongations) pass through P and BC, DE and FA pass through Q, this
means that the dual points are alternatively aligned in two dual lines P and Q. By the
Pappus’ theorem the dual points AD, BE and CF lie on a dual line X, that is, these lines
joining opposite vertices of the hexagon are concurrent in the point X. Moreover, as
proved for the Pappus’ theorem the cross ratios fulfil the equalities in the left hand side:
(BE AD FC P) = (PQ AB FE CD) ⇔ {X, B A F P} = {P, Q A F C}
(BE AD FC Q) = (PQ DE CB FA) ⇔ {X, B A F Q} = {Q, P D C F}
Since the cross ratio of a pencil of lines is equal to the cross ratio of the dual points, the
equalities in the right hand side follow immediately.
11. Conics
11.1 The starting point is the central equation of a conic:
OA = OQ cos α + OR sin α OB = OQ cos β + OR sin β
OC = OQ sin γ + OR sin γ OD = OQ cos δ + OR sin δ
OX = OQ cos χ + OR sin χ
XA = OA − OX = OQ ( cos α − cos χ ) + OR ( sin α − sin χ )
XC = OC − OA = OQ ( cos γ − cos χ ) + OR ( sin γ − sin χ )
XA ∧ XC = OQ ∧ OR [ (cos γ − cos χ ) (sin α − sin χ ) − (cos α − cos χ ) (sin γ − sin χ)]
= OQ ∧ OR [ cos γ sin α − cos γ sin χ − cos χ sin α − cos α sin γ
+ cos α sin χ + cos χ sin γ ]
= OQ ∧ OR [ sin( α − γ ) + sin( γ − χ ) + sin( χ − α ) ]
α −γ γ −χ χ −α
= − 4 OQ ∧ OR sin sin sin
2 2 2
α −γ β −δ
sin sin
XA ∧ XC XB ∧ XD 2 2
=
XA ∧ XD XB ∧ XC α −δ β −γ
sin sin
2 2
11.2 The central equation of an ellipse of centre O with semiaxis OQ and OR is:
OP = OQ cos θ + OR sin θ ⇒ P = O ( 1 − cosθ − sin θ ) + Q cos θ + R sin θ
There is always an affinity transforming an ellipse into a circle:
O' = (0, 0) Q' = (1, 0) R' = (0, 1)
P' = O' ( 1 − cos θ − sin θ ) + Q' cos θ + R' sin θ = ( cos θ , sin θ )
Let us now consider any point E and a circle. A line passing through E cuts the circle in
the points A and B; another line passing also through E but with different direction cuts
the ellipse in the points C and D. Since the power of a point E with respect to the circle is
constant we have:
EA EB = EC ED EA EB −1 EB2 = EC ED −1ED2
The affinity preserves the single ratio of three aligned points:
EA EB −1 = E'A' E'B' −1 EC ED −1 = E'C' ED' −1
Then:
E'A' E'B' −1EB2 = E'C' E'D' −1ED2

2 2
E'A' E'B' E'B' ED
= 2 2
E'C' E'D' EB E'D'
and this quotient is constant for two given directions because the preservation of the single
ratio implies that the distances in each direction are enlarged by a constant ratio:
EA EB
=
E'A' E'B'
11.3 If a diameter is formed by the midpoints of the chords parallel to the conjugate
diameter, then it is obvious that the tangent to the point where the length of the chord
vanishes is also parallel to the conjugate diameter. This evidence may be proved by
derivation of the central equation:
OP = OQ cos θ + OR sin θ
d OP
= − OQ sinθ + OR cos θ
dθ
That is, OP and dOP / dθ (having the direction of the tangent to P) are conjugate radius.
The area a of the parallelogram circumscribed to the ellipse is the outer product of both
conjugate diameters, which is independent of θ :
a = 4 (OQ cosθ + OR sinθ ) ∧ ( −OQ sinθ + OR cosθ ) = 4 OQ ∧ OR
11.4 The area of a parallelogram formed by two conjugate diameters of the hyperbola is
also independent of ψ:
a = 4 (OQ cosh ψ + OR sinh ψ ) ∧ ( OQ sinh ψ + OR cosh ψ ) = 4 OQ ∧ OR
11.5 Let us prove the statement for a circle. The intersections R and R' of a circumference
with a line passing through P are given by (fig. 9.3):
PR = PF cos α − r 2 − PF 2 sin 2α
PR' = PF cos α + r 2 − PF 2 sin 2α
The point M located between R and R' forming harmonic range with P, R and R' fulfils the
condition:
PM RR'
( P R M R' ) = =2
PR' RM
2 PR PR'
Then: PM =
PR + PR'
The numerator is the power of P (constant for any line passing through P) and after
operations we obtain the polar equation of the geometric locus of the points M:
PT 2
PM =
PF cos α
which is the polar line. When α is the angle TPF (figure 9.2) we havePM = PT, that
is, the polar passes through the points T of tangency to the circumference.
Now, by means of any projectivity (e.g. a homology) the circle is transformed into
any conic. Since it preserves the projective cross ratio, the points M also form an aligned
harmonic range in the conic, so the polar also passes through the touching points of the
tangents drawn from P.
11.6 Let the points be denoted by:
A = (1, 1) B = (2, 3) C = (1, 4 ) D = (0, 2 ) E = (2, 5)
Then: EA ∧ EC = −3 EB ∧ ED = −4 EA ∧ ED = −5 EB ∧ EC = −2
The cross ratio is:
EA ∧ EC EB ∧ ED 6
{E , A B C D} = =
EA ∧ ED EB ∧ EC 5
Now a generic point of the conic will be X = (x, y):
XA ∧ XC = −3 x + 3 XB ∧ XD = x − 2 y + 4
XA ∧ XD = − x − y + 2 XB ∧ XC = − x − y + 5
Applying the Chasles’ theorem we find the equation of the conic:

XA ∧ XC XB XD (− 3 x + 3) (x − 2 y + 4 ) 6
= =
XA ∧ XD XB ∧ XC (− x − y + 2 ) (− x − y + 5) 5
After simplifying we arrive at the point equation:
7 x2 + 2 y2 − 6 x y + x − 4 y = 0
which may be written in the form:
 0 1/ 2 − 2  1 − x − y
  
(1 − x − y x y ) 1 / 2 8 − 9 / 2  x =0
− 2 −9/2 − 2   y 
 
The inverse matrix of the conic is:

−1
 0 1/ 2 −2   145 − 40 − 55 
  1  
1 / 2 8 − 9 / 2 =  − 40 16 4 
− 2 −9/2 90 
 − 2   − 55 4 1 
So the equation of the tangential conic is:
 145 − 40 − 55  1 − u − v 
 
[1 − u − v u v ]  − 40 16 4   u  = 0
 − 55 4 1   v 

241 u 2 + 256 v 2 + 488 u v − 370 u − 400 v + 145 = 0
11.7 If we take A, B and C as the base of the points, the coordinates in this base of every
point X will be expressed as:
X = (x A , x B , x C ) x A + x B + xC = 1
If P is the intersection point of the lines AE and BF, then A, P and E are aligned and also
B, P and F, conditions which we may express in coordinates:
1 0 0 0 1 0
(eC f A , e B f C , eC f C )
pA pB pC = 0 pA pB pC = 0 ⇒ ( p A , p B , pC ) =
eC f A + e B f C + eC f C
eA eB eC fA fB fC
If Q is the intersection point of the lines AD and CF, then A, Q and D are aligned and also
C, Q and F, conditions which we may express in coordinates:
1 0 0 0 0 1
(d B f A, d B f B , dC f B )
qA qB qC = 0 qA qB qC = 0 ⇒ (q A , q B , qC ) =
dB f A + d B f B + dC f B
dA dB dC fA fB fC
If R is the intersection point of the lines BD and CE, then B, R and D are aligned and also
C, R and E, conditions which we may express in coordinates:
0 1 0 0 0 1
(d A e A , d A e B , d C e A )
rA rB rC = 0 rA rB rC = 0 ⇒ (rA , rB , rC ) =
d A e A + d A eB + d C e A
dA dB dC eA eB eC
The points P, Q and R will be aligned if and only if the determinant of their coordinates
vanishes. Setting aside the denominators, we have the first step:
fA eB 1 1 1
1
pA pB pC eC f A eB f C eC f C fC eC fC eC dC
f dC 1 1 1
qA qB qC ∝ d B f A dB fB dC f B ∝ A 1 ∝
fB dB fB eB dB
rA rB rC d Ae A d AeB dC eA eB dC 1 1 1
1
eA dA fA eA dA
In the second step each row has been divided by a product of two coordinates. In the third
step each column has been divided by a coordinate. And finally, after transposition and
exchange of the first and third columns (turning the elements 90º in the matrix) and
multiplying each column by a product of three coordinates we obtain:
1 1 1
pA pB pC fA fB fC d AeA d B eB d C eC
1 1 1
qA qB qC ∝ ∝ dA fA dB fB dC fC = 0
eA eB eC
rA rB rC 1 1 1 eA f A eB f B eC f C
dA dB dC
The last determinant is zero if and only if the point F lies on the conic passing through A,
B, C, D and E because then it fulfils the conic equation:
d AeA d B eB d C eC
d AxA d B xB d C xC = 0
eA x A eB x B eC x C
11.8 The Brianchon’s theorem is the dual of the Pascal’s theorem. In the dual plane, the
point conic is represented by the tangential conic, the lines tangent to the point conic are
represented by points lying on the tangential conic and the diagonals of the circumscribed
hexagon are represented by the points of intersection P, Q and R of the former exercise.
So the algebraic deduction is the same but in the dual plane.
12. Matrix representation and hyperbolic numbers
12.1 Let us calculate the modulus and argument:
1 4
5 + 4 e1 = 5 2 − 4 2 = 3 arg(5 + 4 e1 ) = arg tgh = log 3
2 5
Then:
( ( )
5 + 4 e1 = 3 log 3 = 3 cosh log 3 + e1 sinh log 3 = 2 + e1
2
( ))
The four square roots are:
2 + e1 1 + 2 e1 − 2 − e1 − 1 − 2e1
− 3 ± 9 − 4 ⋅ 2 ⋅ (− 17 + 3 e1 )
12.2 2 z 2 + 3 z − 17 + 3 e1 = 0 ⇒ z=
2⋅2
We must calculate the square root of the discriminant:
145 − 24 e1 = 143 log (11 / 13) = 12 − e1

2
So the four solutions are:
9 − e1 − 15 + e1 − 1 − 6 e1
z1 = z2 = z3 = z 4 = −1 + 3 e1
4 4 2
12.3 Let us apply the new formula for the second degree equation:
6 ± 36 − 4 ⋅ 1 ⋅ 5 6 ± 4
z2 − 6 z + 5 = 0 z= = and
2 ⋅1 2
6 ± e1 36 − 4 ⋅ 1 ⋅ 5 6 ± 4 e1
z= =
2 ⋅1 2
1 + e1 1 − e1
12.4 sin( x + y e1 ) = sin( x + y ) + sin( x − y )
2 2
= sin x cos y + e1 cos x sin y
12.5 From the analogous of Moivre’s identity we have:

(coshψ + e1 sinh ψ )4 ≡ cosh 4ψ + e1 sinh 4ψ
cosh 4ψ ≡ cosh 4 ψ + 6 cosh 2 ψ sinh 2ψ + sinh 4ψ
sinh 4ψ ≡ 4 cosh 3 ψ sinhψ + 4 coshψ sinh 3ψ
12.6 The analytical continuation of the real logarithm is:
1 + e1 1 − e1
log( x + y e1 ) = log( x + y ) + log( x − y )
2 2
x+ y
log(x 2 − y 2 ) + 1 log
1 e
=
2 2 x−y
It may be rewritten in the form:
y
= log x 2 − y 2 + e1 arg tgh
x
12.7 For the straight path z = t e1 we have:
e1 1 1
t 3  2e
∫−e = 1∫ = = 1
2 2
z dz e t dt e1  
1 −1  3  −1 3
For the circular path z = cos t + e1 sin t we have:

e1 π /2
∫ z dz = ∫ (cos t + e sin t ) (− sin t + e1 cos t ) dt

2 2
1
− e1 −π / 2
π /2
 2  2  2e
= cos t + cos 3 t + e1  sint − sin 3 t  = 1
 3  3  −π / 2 3
Using the indefinite integral, we find also the same result:
e1 e1
 z3  2 e1
∫−e = 3 = 3
2
z dz
1
  − e1
12.8 The proof is analogous to the exercise 3.12: turn the hyperbolic numbers df, dz into
hyperbolic vectors by multiplying them at the left by e2.
13. The hyperbolic or pseudo-Euclidean plane

13.1 If the vertices of the triangle are A = (2, 2), B = (1, 0) and C = (5, 3) then:
AB = (−1, −2) BC = (4, 3) CA = (−3, −1)
AB = 3 e12 BC = 7 CA = 8
From the cosine theorem we have:

9
AB 2 = BC 2 + CA 2 − 2 BC CA cosh γ ⇒ cosh γ = ⇒ γ ≅ 0.6264
2 14
taking into account that γ = ACB is a positive angle. From the sine theorem we have:
BC CA AB 7 8 3 e12
= = ⇒ = =
sinh α sinh β sinh γ sinh α sinh β 5
2 14
whence the angles α and β follow:
5 e12 π
sinhα = − ⇒ α ≅ −0.2027 − e12
2 6 2
5 e12 π
sinhβ = − ⇒ β ≅ −0.4236 − e12
21 2
The plot helps to choose the right sign of the angles. Anyway a wrong choice would be
revealed by the cosine theorem:
BC 2 = CA 2 + AB 2 − 2 CA AB cosh α
CA 2 = AB 2 + BC 2 − 2 AB BC cosh β
13.2 The rotation of the vector is obtained with the multiplication by a hyperbolic
number:
5 3 
v' = v zα = (2 e 2 + e 21 ) (cosh log 2 + e1 sinh log 2 ) = (2 e 2 + e 21 )  + e1 
4 4 
13 11
= e 2 + e 21
4 4
The reflection is obtained by multiplying on the right and on the left by the direction
vector of the reflection and its inverse respectively:
13 e 2 − 11 e 21
v' = d −1 v d = (3 e 2 − e 21 )
−1
(2 e 2 + e 21 ) (3 e 2 − e 21 ) = 1 (3 e 2 − e21 )(7 − 5 e1 ) =
8 4
The inversion with radius r = 3 is:

v' = r 2 v −1 = 9 (2 e 2 + e 21 ) = 3 (2 e 2 + e 21 ) = 6 e 2 + 3 e 21
−1
13.3 The two-point equation of the line y = 2 x + 1 is:
x +1 y
=
1 2
which gives us the direction vector:
v = e1 + 2 e 2
The direction vector is the normal vector of the perpendicular line. The general equation
of a line in the hyperbolic plane is:
n · PR = 0 ⇒ n x (x − x P ) − n y ( y − y P ) = 0
Note the minus sign since this line lies on the hyperbolic plane. The substitution of the
components of the direction vector of the first line and the coordinates of the point R =
(3, 1) through which the line passes results in the equation:
1 ⋅ ( x − 3) − 2 ⋅ ( y − 1) = 0 ⇒ x − 2 y −1= 0
The fact that both lines and the first quadrant bisector intersect in a unique point is
circumstantial. Only it is required that the bisectors of both lines be parallel to the
quadrant bisectors.
2 2
13.4 The line y =3 intersects the hyperbola x – y = 16 in the points R = (−5, 3) and R' =
(5, 3). Then the power of P = (−7, 3) is:
PR PR' = 2 e 2 · 12 e 2 = 24
The line y = – 3 x – 18 cuts the hyperbola in the points T = (–5, –3) and T' = (–17/2,
15/2). Then the power of P is:
 3e 9e 
PT PT' = (2 e 2 − 6 e 21 ) ·  − 2 + 21  = −3 + 27 = 24
 2 2 
The line y = –3 x / 7 cuts the hyperbola in the points:
 7 10 3 10   7 10 3 10 
S =  − ,  and S' =  ,− 
 5 5   5 5 
The power is:
  7 10   3 10     7 10   3 10  
PS PS' =   − + 7  e 2 +  − 3  e 21  ·   + 7  e 2 +  − − 3  e 21 
 5   
   5    5   5  
147 27
= − = 24
5 5
Also the substitution of the coordinates gives 24:
x P2 − y P2 − 16 = 49 − 9 − 16 = 24
13.5 From the law of sines we find:
a sinhα a + b sinhα + sinhβ

= ⇒ =
b sinhβ a − b sinhα − sinhβ
Introducing the identity for the addition and subtraction of sines we arrive at the law of
tangents:
α+β
a + b tgh
= 2
a − b α − β
tgh
2
13.6 The first triangle has the vertices A = (0, 0), B = (5, 0), C = (5, 3) and the sides:
AB = 5 e2 BC = 3 e21 CA = −5 e2 − 3 e21
The second triangle has the vertices A' = (0, 0), B' = (25, −15), C' = (16, 0) and the
sides:
A'B' = 25 e2 − 15 e21 B'C' = −9 e2 + 15 e21 C'A' = −16 e2
Both triangles are directly similar since:
AB−1 A'B' = BC−1 B'C' = CA−1 C'A' = 5 − 3 e1 = r
which is the similarity ratio r. The size ratio is the modulus of the similarity ratio:
r = 52 − 32 = 4
that is, the second triangle is four times larger than the first. The angle of rotation
between both figures is the argument of the similarity ratio:
 3
ψ = arg(5 − 3 e1 ) = arg tgh −  = − log 2 = −0.6931...
 5
Now plot the vertices of each triangle in the hyperbolic plane and break your Euclidean
illusions about figures with the same shape.
14. Spherical geometry in the Euclidean space

14.1 The shadow of a gnomon follows a hyperbola on a plane (a quadrant) because the
Sun describes approximately a parallel around the North pole during a day. The altitude
of this parallel from the celestial equator is the declination, which changes slowly from
one day to another. The March 21 and September 23 are the equinoxes when the Sun
follows the equator. Since it is a great circle, its central projection is a straight line
named the equinoctial line.
14.2 From the stereographic coordinates we find the Cartesian coordinates of the points:
2 2 1 12 e 2 5 e 3 12 5
A = − e1 + e 2 + e3 B=− + C= e1 + e 3
3 3 3 13 13 13 13
Their outer products are equal to the cosines of the sides:
25
cos a = B · C = a = 1.4223
169
19
cos b = C · A = − b = 2.0797
39
19
cos c = A · B = − c = 2.0797
39
Now we calculate the bivectors of the planes containing the sides:
22 e 23 + 10 e31 + 24 e12 1160

A∧B = A∧ B =
39 39
− 60 e 23 + 60 e31 + 144 e12 27936

B ∧C = B ∧C =
169 169
− 10 e 23 − 22 e31 + 24 e12 1160

C ∧ A= C∧A =
39 39
The cosines of the angles between planes are obtained through the analogous of scalar
product for bivectors:
(C ∧ A) · ( A ∧ B ) 136
cos α = − =− α = 1.6882
C∧A A∧B 1160
( A ∧ B ) · (B ∧ C ) 2736
cos β = − =− β = 2.0721
A∧B B ∧C 1160 ⋅ 27936
(B ∧ C ) · (C ∧ A) 2736
cos γ = − =− γ = 2.0721
B ∧C C∧A 27936 ⋅ 1160
The area of the triangle is the spherical excess:
area = α + β + γ − π = 2.6908
14.3 Let us calculate the Cartesian coordinates of Fastnet (point F), the point in the
middle Atlantic (point A) and Sandy Hook (point S). Identifying the geographical
longitude with the angle ϕ and the colatitude with the angle θ we have:
x = sinθ cos ϕ y = sinθ sinϕ z = cos θ
F = (θ = 38º 30' , ϕ = −9 º 35' ) = 0.6138 e1 − 0.1036 e 2 + 0.7826 e 3
A = (θ = 47 º 30' , ϕ = −47 º ) = 0.5028 e1 − 0.5392 e 2 + 0.6756 e 3
S = (θ = 49º 30' , ϕ = −74 º ) = 0.2096 e1 − 0.7309 e 2 + 0.6494 e3
The length of the arcs is obtained from the inner products of the position vectors:
cos FS = F · S = 0.7126 FS = 44º33' = 4950 km
cos FA = F · A = 0.8932 FA = 26º 43' = 2969 km
cos AS = A · S = 0.9382 AS = 20º14' = 2250 km
Now we see that the track is 269 km longer than the shortest path between Fastnet and
Sandy Hook.
Let us investigate whether the Titanic followed the obliged track, that is whether
the point T where the tragedy happened lies on the line AS, by calculating the
determinant of the three points:
T = (θ = 48º14' , ϕ = −50º14' ) = 0.4771 e1 − 0.5733 e 2 + 0.6661 e3
0.5028 0.4771 0.2096

A ∧ T ∧ S = − 0.5392 − 0.5733 − 0.7309 e123 = −0.0050 e123
0.6756 0.6661 0.6494
The negative sign indicates that the

orientation of the three points is
clockwise (figure 16.6). In other words,
the Titanic was shipping at the South of
the obliged track. We can calculate the
distance from T to the track AS taking the Figure 16.6
right angle triangle. By the Napier’s rule we have:
A∧T ∧ S 0.0050
sin d = sinσ sin ST = = = 0.0144 d = 49' = 92 km
A∧S 0.3461
14.4 After removing the margins and subtracting the coordinates of the centre of the
photograph in the bitmap file, I have obtained a pair of coordinates u' and v' in pixels,
which are proportional to u and v:
star u' v'

α Cassiopeia 173 −139
β Cassiopeia 349 −135
γ Cassiopeia 199 24
δ Cassiopeia 89 93
Hale-Bopp comet −240 −320
On the other hand, the right ascension A and declination D of these stars are
known data. From the spherical coordinates and taking θ = 90º − D and ϕ = A one
obtains their Cartesian equatorial coordinates in the following way:
x = cos D cos A y = cos D sin A z = sin D
The Aries point has the equatorial coordinates (1,0,0). Using these formulas I have
found the coordinates:
star x y z
α Cassiopeia 0.543519 0.096098 0.833878
β Cassiopeia 0.512996 0.019711 0.858216
γ Cassiopeia 0.475011 0.119054 0.871889
δ Cassiopeia 0.462917 0.180840 0.867759
A photograph is a central projection. So the arch s between two stars A and B is related
with their coordinates (u, v) on the projection plane by:
1 + u A uB + v A vB
cos s AB =
2 2 2 2
1 + uA + vA 1 + uB + vB
The focal distance f is the distance between the projection plane (the photograph) and
the centre of projection. When the focal distance is unknown we only can measure
proportional coordinates u' and v' instead of u and v, and the foregoing formula
becomes:
f 2 + u' A u' B + v' A v' B

cos s AB =
2 2 2 2
f 2 + u' A + v' A f 2 + u' B + v' B
which is a second degree equation on f 2:
(f 2 2
+ u' A + v' A
2
)( f 2 2 2
)
+ u' B + v' B cos 2 s AB = ( f 2 + u' A u' B + v' A v' B )
2
0 = f 4 (1 − cos 2 s AB ) + f 2
[ 2 ( u' A (
u' B + v' A v' B ) − u' A + v' A + u' B + v' B cos 2 s AB
2 2 2 2
) ]
+ ( u' A u' B + v' A v' B ) − u' A + v' A
2
( 2 2
)( u' B
2 2
)
+ v' B cos 2 s AB
All the coefficients of the equation are known, because cos sAB = A · B is calculated
from the Cartesian equatorial coordinates obtained from the right ascension and
declination of both stars. The focal distances so obtained are:
star f
α -β Cassiopeia 2012.7
α -γ Cassiopeia 2010.6
α -δ Cassiopeia 2017.2
β -γ Cassiopeia 2003.0
β -δ Cassiopeia 2019.1
γ -δ Cassiopeia 2049.1
mean value 2018.6
The width of the photograph is 750 pixels and the height 1166. If 750 pixels
corresponds to 24 mm and 1166 to 36 mm, we find a focal distance of the camera equal
to 64.6 mm and 62.3 mm respectively. However the author of the photograph indicated
me that his camera has a focal distance of 58 mm, what implies that the original image
was cut a 7 % in the photographic laboratory. This is a customary usage in photography,
so we do not know the enlargement proportion of a paper copy and we must calculate
the focal distance directly from the photograph, which is always somewhat higher than
that calculated from the focal distance of the camera. On the other hand, the calculus of
the focal distance is very sensitive to the errors and truncation of decimals. The best
performance is to write a program to calculate the mean focal distance from the original
data or to make us the paper copies without cutting the image. Known the focal
distance, we are able to calculate the normalised coordinates u and v:
star u v
α Cassiopeia 0.085703 −0.068860

β Cassiopeia 0.172892 −0.066878

γ Cassiopeia 0.098583 0.011889
δ Cassiopeia 0.044090 0.046072

Hale-Bopp comet −0.118894 −0.158526
and from here the inner product between stars and the comet:
1 + u A uH + v A vH
cos( A, H ) =
1 + u A2 + v A2 1 + u H2 + v H2
1 + uB uH + vB vH
cos(B, H ) =
1 + u B2 + v B2 1 + u H2 + v H2
The unknown equatorial coordinates of the comet H are calculated from a system of
three equations:
 A · H = cos( A, H )

 B · H = cos(B, H )
 H 2 =1

Let us introduce the Clifford product instead of the inner product and multiply the first
equation by B and the second one by A on the right:
AH +H A AH B+H AB
 = cos( A, H ) ⇒ = cos( A, H ) B
2 2
 B H +H B B H A+ H B A
 = cos(B, H ) ⇒ = cos(B, H ) A
 2 2
 H 2 =1

Using the permutative property (in the space u v w – w v u = 2 u ∧ v ∧ w), the

difference of both equations is equal to:
A ∧ H ∧ B + H A ∧ B = cos( A, H ) B − cos(B, H ) A
or equivalently: − H ∧ A ∧ B + H A ∧ B = cos( A, H ) B − cos(B, H ) A
The outer product by H is the geometric product by its perpendicular component H⊥ , so

we have:
H || A ∧ B = cos( A, H ) B − cos(B, H ) A
where we can isolate the coplanar component:
H || = (cos( A, H ) B − cos(B, H ) A) ( A ∧ B )
−1
Observe that this result can be only obtained and expressed through the
geometric product, and not through the more usual inner and outer products. The
component perpendicular to the plane AB is proportional to the dual vector of the outer
product of this vectors. Known the coplanar component we may now fix the modulus of
the perpendicular component because H2=1:
e123 A ∧ B
H⊥ = ± 1 − H ||2
A∧B
There are two solutions, but according to the statement of the problem only one
of them is valid. Since each pair of stars gives two values of H (the position vector of
the comet), we may distinguish the true solution because it has proper values for
different pairs of stars. The mean value with the standard deviation so obtained is:
H = (0.661606 ± 0.000741) e1 + (0.236258 ± 0.001298) e 2 + (0.711659 ± 0.000302 ) e3
from where the following equatorial coordinates of the Hale-Bopp comet are obtained:
AH = 1h 18min 36s D H = 45º 22'12' '
In the ephemeris (http://www.xtec.es/recursos/astronom/hb/ephjan97.txt) for the comet

we find:
Date Right ascension (A) Declination (D)
1997 March 28 01h 09min 18.03s +45º 36’ 30.6”
1997 March 29 01h 19min 06.07s +45º 25’ 15.6”
1997 March 30 01h 28min 44.92s +45º 10’ 51.9”
So I conclude that this photograph was taken the March 29 of 1997.

On the other hand, I have also calculated the orientation O of the camera (of the
line passing through the centre of the photograph):
O = (0.513505 ± 0.000673) e1 + (0.201570 ± 0.000471) e 2 ± (0.834075 ± 0.000360 ) e3
AO = 1h 25min 44s DO = 56 0 31'11"
15. Hyperboloidal geometry in the pseudo-

Euclidean space (Lobachevsky’s geometry)
15.1 The circle being projection of a parallel line (geodesic

hyperbola) shown in the figure 15.9 passes through the
point (1, 0) which implies the equation:
(u − 1)2 + (v − b)2 = b 2
According to this equation, the intersection of this line with the v-axis has the
( )
coordinates 0, b − b − 1 (the other intersection point falls outside the Poincaré
2
disk). The inverse of the slope at this point with opposite sign is the tangent of the angle
of parallelism: Figure 15.9
 du  1
tg Π (s ) = −  = b 2 − 1 ⇒ cos Π (s ) =
 dv  u = 0 b
Now let us calculate the distance s from the origin to this point. Its hyperbolic cosine is
the inner product of the unitary position vectors of the origin and the point of
intersection:
 b − b2 − 1  b  b
cosh s = e 3 ·   e + e 
3 =
 − b2 + 1 + b b2 − 1  2
b 2 − 1  b2 − 1
 
b +1 b −1 1 − cos Π (s ) Π (s )
s = log exp(− s ) = = = tg
b −1 b +1 1 + cos Π (s ) 2
so we find the Lobachevsky’s formula for the angle of parallelism:
Π (s ) = 2 arctg(exp(− s ))
15.2 The law of cosines in hyperboloidal trigonometry is:
cosh a = cosh b cosh c − sinh b sinh c cos a
For a right angle triangle α = π/2 :
cosh a = cosh b cosh c
In the limit of small arcs it becomes the Pythagorean theorem:
  
+ O (a 4 ) = 1 + + O (b 4 ) 1 + + O (c 4 )
a2 b2 c2
1+
2  2  2 
a 2 + O (a 4 ) = b 2 + c 2 + O (b 2 c 2 , b 4 , c 4 )
15.3 A «circle» is given by the equation system:

 z2 − x2 − y2 =1

a x + b y + c = z
a) The intersection of a quadric (hyperboloid) with a plane is always a conic. Since the
curve is closed, it must be an ellipse (for horizontal planes it is a circle).
b) We hope that the centre of the ellipse be the intersection of the plane with its
perpendicular line passing through the origin (this occurs also in spherical geometry).
The plane has a bivector:
a e 23 + b e31 − e12
The vector perpendicular, the dual vector is:
a e1 + b e 2 + e 3
because the product of the vector and bivector is equal to a pure volume element:
(a e 23 + b e31 − e12 ) (a e1 + b e 2 + e3 ) = ( a 2 + b 2 − 1) e123
The axis of a circle is the line passing through the origin and perpendicular to the plane
containing this circle. The centre of the circle is the intersection point of the plane and
axis of this circle, given by the equation system:
a x + b y + c = z
 ac bc c
 x= y=z ⇒ x= y= z=
 a b 1 1 − a 2 − b2 1 − a 2 − b2 1 − a − b2
2
The distance from any point of the circle to its centre is constant:
2 2
 c   ac   bc 
z − 2 
−x − 2 
−y− 2 
=
 1− a − b 
2
 1− a − b 
2
 1− a − b 
2
2 c2 c 2 (1 − a 2 − b 2 ) c2
1− + = 1−
1− a 2 − b2 (1 − a 2
− b2 )
2
1− a 2 − b2
Since c > 1 and a + b < 1 this value is always negative, what means that the radius is
2 2
a distance comparable with distances on the Euclidean plane x-y. Taking the real value,
the radius of a circle is:
c2
r= −1
1 − a 2 − b2
However this is not the radius measured on the hyperboloid and obtained from its
projections. The intersection of the axis with the hyperboloid gives its hyperboloidal
centre:
z 2 − x 2 − y 2 = 1
 a b 1
 x= y=z ⇒ x= y= z=
 a b 1 1− a − b
2 2
1− a − b
2 2
1 − a 2 − b2
The hyperboloidal radius is the arc length on the hyperboloid from any point of the
circle to its hyperboloidal centre, which should be constant. Making the inner product of
the position vectors of a point on the circle and its centre, we find:
 a b z 
coshψ = ( x e1 + y e 2 + z e3 ) ·  e +
2 1
e +
2 2
e =
2 3
1 − a − b 1− a − b 1− a − b
2 2 2

−a x−b y+ z c
=
1− a − b
2 2
1 − a − b2
2
So the hyperboloidal radius is:
c
ψ = arg cosh
1 − a − b2
2
The figure 16.7 shows a lateral view of a

plane with a = 0 and equation b x + c = z.
The points P and P' lie on the circle. O is
the centre of the circle on its plane and H
is the centre on the hyperboloid. Then OH
is the axis of the circle, which is Figure 16.7
perpendicular (in a pseudo-Euclidean way)
to the circle plane, that is, their bisector has a unity slope. Note that according to a
known property of the hyperbola, O must be the midpoint of the chord PP'. The plane
radius is the distance PO or P'O, while the hyperboloidal radius is the arc length PH or
P'H.
c) The coordinates of the stereographic projection are:
2u 2v 1 + u2 + v2
x= y= z=
1− u2 − v2 1 − u2 − v2 1− u2 − v2
Let us make the substitution of these coordinates in the equation of the circle plane:
2au+2bv 1 + u2 + v2
+ c =
1 − u2 − v2 1 − u2 − v2
After simplification we arrive to an equation of a circle:

2 2
 a   b  a 2 + b2 + c 2 − 1
 u −  +  v −  =
 c + 1  c + 1 (c + 1)2
15.4 A horocycle is a circle having a + b = 1 . Therefore its equation in the

2 2
stereographic projection is:

2 2
c2  a   b 
= u −  + v − 
(c + 1)2  c + 1  c + 1
Let us search the intersection points (if they exist) with the limit circle u + v = 1 .
2 2
Then we solve the system of both equations and find that it has a unique solution:
u=a v=b
Since they meet in a unique point and a + b = 1 , the horocycle projection is tangent
2 2
to the limit circle. The centre of the horocycle is the intersection of the hyperboloid with
its axis. However this axis has unity slope, so the intersection lies at the infinity.
15.5 By differentiation of the coordinates in the Beltrami projection we find:
dx =
( 1 − v ) du + u v dv
2
dy =
(1 − u ) dv + u v du
2
(1 − u − v )
2 2 3/ 2
(1 − u − v ) 2 2 3/ 2
u du + v dv
dz =
(1 − u 2
− v2 )
3/ 2
The differential of area is a bivector and we search its modulus:
du ∧ dv
dA = (dx ∧ dy )2 − (dy ∧ dz )2 − (dz ∧ dx )2 =
(1 − u 2
− v2 )
3/ 2
Note the outer products of the differentials

of coordinates so dx ∧ dy = − dy ∧ dx .
Now we must integrate the differential of
area of a doubly asymptotic triangle with
angle ϕ (figure 15.10a). We change to
polar coordinates r and θ in the Beltrami
disk:
v
r = u2 + v2 tg θ =
u Figure 15.10
cos(ϕ / 2 )
Since the equation of the line is r = the integral becomes:
cos θ
cos (ϕ / 2 ) cos (ϕ / 2 )
ϕ /2 ϕ /2 ϕ /2 ϕ /2
cos θ
r dr ∧ dθ  1  cos θ cos θ
A= ∫ ∫ (1 − r ) 2 3/ 2
= ∫  
−ϕ / 2  1 − r  0
2
dθ = ∫
2 ϕ
dθ − ∫ dθ =
0 −ϕ / 2 −ϕ / 2
1 − cos − sin θ
2 −ϕ / 2
2
ϕ /2
 
 sinθ 
= arcsin −θ = π −ϕ
 ϕ 
sin
 2  −ϕ / 2
This expression is applied to each asymptotic triangle surrounding the central

triangle (figure 15.10b). Since the area of a triangle having the three vertices in the limit
circle is 3π−2π = π, the area of the central triangle follows:
A= π −α − β −γ
15.6 In the azimuthal projection, the radius r of a circle centred at the origin is:
2 2
u=x v= y r 2 = u 2 + v 2 = 2 (z − 1)
z +1 z +1
Since the projection is equivalent, we can evaluate the area directly in the plane:
A = π r 2 = 2 π ( z − 1) = 2 π (coshψ − 1)
where ψ is the Weierstrass coordinate, that is, the radius of the circle on the
hyperboloid.
BIBLIOGRAPHY
R. ABLAMOWICZ, P. LOUNESTO, J. M. PARRA eds. Clifford Algebras with numeric and

symbolic computation applications. Birkhäuser (Boston, 1996).
D. V. ALEKSEEVSKIJ, E. B. VINBERG, A. S. SOLODOVNIKOV. I.Geometry of Spaces
of Constant Curvature in Geometry vol. II, Spaces of Constant Curvature, ed. by E. B. Vinberg.
Encyclopaedia of Mathematical Sciencies 29, Springer Verlag (New York, 1993).
William E. BAYLIS (editor). Clifford (Geometric) Algebras. Birkhäuser (1996).
Marcel BERGER.. Geometry, vol. I-II. Springer-Verlag (Berlin, Heidelberg,1987).
Roberto BONOLA. Non-Euclidean Geometry. Dover Publications Inc. (New York, 1955).
Alfred M. BORK. « “Vectors Versus Quaternions”—The letters in Nature», American Journal
of Physics, 34 (1966) 202-211.
F. BRACKX, R. DELANGHE, H. SERRAS eds. Clifford Algebras and Their Applications in
Mathematical Physics. Kluwer (Dordrecht, 1993).
Lev M. BUGAYEVSKIY, John P. SNYDER. Map projections. A reference manual. Taylor &
Francis (London, 1995).
David A. BRANNAN, Matthew F. ESPLEN, Jeremy J. GRAY. Geometry. Cambridge
University Press (Cambridge, 1999).
Gaston CASANOVA. L’algèbre vectorielle. Presses Universitaires de France (París, 1976).
G. CASANOVA. «Fonctions analytiques hyperboliques», Advances in Applied Clifford
Algebras, 9 (1999) 93.
G. CASANOVA. «Roots of algebraic equations and Clifford algebra», Advances in Applied
Clifford Algebras, 9 (1999) 215.
William Kingdon CLIFFORD. «Applications of Grassmann’s Extensive Algebra» American
Journal of Mathematics Pure and Applied, 1 (1878) 350-358, reproduced in Mathematical
papers by William Kingdon Clifford, 266-276. Chelsea Publishing Company (New York, 1968).
W. K. CLIFFORD. «On the classification of geometric algebras», Mathematical papers by
William Kingdon Clifford, 397-401. Chelsea Publishing Company (New York, 1968).
Julian Lowell COOLIDGE. A History of the Conic Sections and Quadric Surfaces. Dover
Pub. Inc. (New York, 1968).
Harold S. M. COXETER. Introduction to Geometry. John Wiley & Sons, Inc. (NewYork,
1961).
H. S. M. COXETER. The Real Projective Plane, 2nd ed. Cambridge Univ. Press (Cambridge,
1955).
Michael J. CROWE. A History of Vector Analysis. The Evolution of the Idea of a Vectorial
System. Dover Publications, Inc. (New York, 1994).
N. V. EFIMOV. Higher geometry. Mir Publishers (Moscow, 1980).
Howard EVES, A Survey of Geometry. Allyn and Bacon, Inc. (London, 1965).
Richard L. FABER. Foundations of Euclidean and Non-Euclidean Geometry. Marcel Dekker,
Inc. (New York, 1983).
Werner FENCHEL. Elementary Geometry in Hyperbolic Space. De Gruyter Studies in
Mathematics 11, de Gruyter (New York, 1989).
Henry G. FORDER. The Calculus of Extension. Cambridge University Press (London, 1941).
Rudolf FUCKE, Konrad KIRCH, Heinz NICKEL. Darstellende Geometrie. VEB
Fachbuchverlag (Leipzig, 1987).
C. I. GERHARDT.«Leibniz an Huygens» in Der Briefwechsel von Gottfried Wilhelm Leibniz
mit Mathematiker, 570.Georg Olms Verlag (Hildesheim, 1987).
C. I. GERHARDT. «Characteristica geometrica. Analysis geometrica propria. Calculus situs»
in G. W. Leibniz. Mathematische Schriften V, 141-183. Georg Olms Verlag (Darmstadt, 1971).
Josiah Willard GIBBS. «On multiple algebra». Proceedings of the American Association for
the Advancement of Science XXXV (1886) 37-66. Reproduced together with the following
letters to Nature in Scientific works of J. W. Gibbs, II, Longman, Green (London, 1906).
J. W. GIBBS. «On the role of quaternions in the algebra of vectors». Nature, XLIII (1891)
511-513.
J. W. GIBBS. «Quaternions and the Ausdehnungslehre». Nature, XLIV (1891) 79-82.
J. W. GIBBS. «Quaternions and the algebra of vectors». Nature, XLVII (1893) 463-464.
J. W. GIBBS. «Quaternions and vector analysis». Nature, XLVIII (1893) 364-367.
J. W. GIBBS, Edwin Bidwell WILSON (author). Vector analysis, a text-book for the use of
students of mathematics and physics founded upon the lectures of J. Willard Gibbs. Yale
bicentennial publications, C. Scribner’s Sons (New York, 1901).
Ramon GONZALEZ, Josep HOMS, Jordi SOLSONA. Butlletí de la Societat Catalana de
Matemàtiques 12, 1 (1997) 51-71 (English translation «Estimate of the error committed in
determining the place from where a photograph has been taken» at
http://www.terra.es/personal/rgonzal1/fotoart2.htm ).
Hermann GRASSMANN. Die Ausdehnungslehre, first edition (Szczecin, 1844). Spanish
translation Teoría de la extensión1 by Emilio Óscar Roxín in the collection Historia y filosofía
de la ciencia, Espasa-Calpe (Buenos Aires, 1947).
H. GRASSMANN. A new branch of mathematics: The “Ausdehnungslehre” of 1844 and other
works. Translated by Lloyd C. Kannenberg. Open Court (Chicago, 1995).
H. GRASSMANN. Extension Theory. Translation of Die Ausdehnungslehre of 1862 by Lloyd
C. Kannenberg. American Mathematical Society (Providence, Rhode Island, 2000).
H. GRASSMANN. «The position of the Hamiltonian quaternions in the extension theory».
Translation of «Der Ort der Hamilton’schen Quaternionen in der Ausdenungslehre»,
Mathematik Annalen 12 (1877) 375-386, by Lloyd C. Kannenberg in A new branch of
mathematics: The “Ausdehnungslehre of 1844 and other works, 525-538 and editorial notes
539-542.
V. GUSEV, V. LITVINENKO, A. MORDKOVICH. Solving Problems in Geometry. Mir
Publishers (Moscow, 1988).
rd
William Rowan HAMILTON. Elements of Quaternions (1869). Ed. by Charles Jasper Joly, 3
edition, Chelsea Publishing Company (New York, 1969).
Oliver HEAVISIDE. «The elements of vectorial algebra and analysis» (13th Nov 1891),
Electromagnetic Theory, The Electrician Series (1891-1898), The Electrician printing and
publishing Company Ltd. (London).
David HESTENES. Space-Time Algebra. Gordon and Breach (New York, 1966).
D. HESTENES. New Foundations for Classical Mechanics. Ed. by Alwyn Van der Merwe. D.
Reidel Publishing Company (Dordrecht, 1986).
D. HESTENES, Garret SOBCZYK. Clifford Algebra to Geometric Calculus. A Unified
Language for Mathematics and Physics, Reidel Publishing Company (Dordrecht, 1984).
D. HESTENES. «Mathematical viruses» in Clifford Algebras and their Applications in
Mathematical Physics, 3-16, edited by A. Micali, R. Boudet and J. Helmstetter. Kluwer
Academic Publishers (Dordrecht, 1992).
David HILBERT. Grundlagen der Geometrie. Mit Supplementen von Paul Bernays. Teubner
(Stuttgart, 1977).
David HILBERT, S. COHN-VOSSEN. Geometry and the Imagination. Chelsea Publishing
Company (New York, 1952).
International Cartographic Association. Basic Cartography for students and technician.
(Over Wallop, 1984).
Birger IVERSEN. Hyperbolic Geometry. London Mathematical Society Student Texts, 25.
Cambridge University Press (Cambridge, 1992).
George A. JENNINGS. Modern Geometry with Applications. Springer-Verlag (New York,
1994).
Felix KLEIN. Vorlesungen über nicht-euklidische geometrie. Verlag von Julius Springer
1
In this edition, the date in the preface by H. Grassmann was misprinted as 1884 instead of
1844.
(Berlin, 1928).
Daniel LEHMANN, Rudolph BKOUCHE. Initiation à la géométrie. Presses universitaires de
France (Paris, 1988).
Pertii LOUNESTO. Clifford Algebras and Spinors. Cambridge University Press (Cambridge,
1997).
George E. MARTIN. The Foundations of Geometry and the Non-Euclidean Plane. Springer-
Verlag (New York, 1975).
Edwin E. MOISE. Elementary Geometry from an Advanced Standpoint, Addison-Wesley
Publishing Company, 2nd ed. (Reading, Massachusetts, 1974)
Junjiro NOGUCHI. Introduction to Complex Analysis. Translations of Mathematical
Monographs, 168. American Mathematical Society (Providence, Rhode Island, 1998).
Josep Manel PARRA i SERRA. «Clifford algebras. Towards a common language for
mathematicians and physicists. Some suggested notations», Advances in Applied Clifford
Algebras 2 (1992) 145.
J. M. PARRA I SERRA. «Geometrical algebra versus numerical Cartesianism. The historical
trend behind Clifford’s algebra», in Clifford Algebras and their Applications in Mathematical
Physics, 307-316. F. Brackx, R. Delanghe, H. Serras eds., Kluwer, Dordrecht (Deinze, 1993).
J. M. PARRA I SERRA. «Teaching Clifford Algebras as Physiscal Mathematics», in Clifford
Algebras and their Applications in Mathematical Physics, 291-304. V. Dietrich, K. Habetha, G.
Janks eds. Kluwer, Dordrecht (Aachen, 1996).
Giuseppe PEANO. Calcolo geometrico secondo l’Ausdehnungslehre di Hermann Grassmann.
Fratelli Boca Editori (Torino, 1888).
G. PEANO. «Saggio di Calcolo geometrico», Atti della Reale Accad. Delle Scienze di Torino
XXXI (1895-96) 952-975. Collected in Opere Scelte, III, 167, ed. Cremone, (Roma, 1959).
Also collected and translated by Hubert C. Kennedy in Selected Works of Giuseppe Peano, 169-
188. George Allen & Unwin Ltd. (London, 1973).
Dan PEDOE. Geometry, A Comprehensive Course. Dover Publications, Inc., (New York,
1988).
George POLYA, Gordon LATTA. Complex variables. John Wiley & Sons, Inc., (New York,
1974).
Murali RAO, Henrik STETKÆR. Complex Analysis, an invitation. World Scientific
(Singapore, 1991).
Hans REICHARDT. Gauss und die Anfänge der nicht-euklidischen Geometrie. Mit
Originalarbeiten von J. Bolyai, N. I. Lobatschewski und F. Klein. Teubner-Archiv zur
Mathematik, 4. BSB B. G. Teubner Verlaggesellschaft (Leipzig, 1985).
Marcel RIESZ. Clifford Numbers and Spinors. The Institute of Fluid Mechanics and Applied
Mathematics, Univ. Maryland (1958). Reprinting E. Folke Bolinder, P. Lounesto eds. Kluwer
(Dordrecht, 1993).
Boris A. ROSENFELD. A History of Non-Euclidean Geometry. Evolution of the Concept of a
Geometric Space. Studies in the History of Mathematics and Physical Sciences, 12. Springer-
Verlag (New York, 1988).
Patrick J. RYAN. Euclidean and non-Euclidean geometry. An analytic approach. Cambridge
University Press (Cambridge, 1986).
A. A. SAZANOV. El universo tetradimensional de Minkowski. Editorial Mir (Moscú, 1990).
I. SHARIGUIN. Problems in plane geometry. Mir Publishers (Moscow, 1988).
Hans SCHWERDTFEGER. Geometry of Complex Numbers. Dover Publications, Inc. (New
York, 1979).
John STILLWELL. Geometry of Surfaces. Universitext, Springer-Verlag (New York, 1992).
The teaching of trigonometry in schools. A report prepared for the Mathematical Association.
G. Bell & Sons, Ltd. (London, 1962).
C. ZWIKKER. The advanced geometry of plane curves and their applications. Dover Pub. Inc.
(New York, 1963).
INTERNET BIBLIOGRAPHY
Personal pages
Rafal Ablamowicz http://math.tntech.edu/rafal/

William E. Baylis http://www.cs.uwindsor.ca/users/b/baylis/
Ramon González Calvet. http://www.terra.es/personal/rgonzal1
David Hestenes. http://modelingnts.la.asu.edu/
Jaime Keller http://www.tuwien.ac.at/cms/jkeller/
Pertii Lounesto. http://www.helsinki.fi/~lounesto/
Josep Manel Parra i Serra. http://hermes.ffn.ub.es/~jmparra
Software pages
CABRI http://www.cabri.com
CINDERELLA homepage http://www.cinderella.de/en/index.html
application to geometric algebra http://carol.wins.uva.nl/~leo/cinderella/
CLICAL (Clifford Calculator) http://www.helsinki.fi/~lounesto/CLICAL.htm
CLU http://www.perwass.de/cbup/start.html
GAIGEN http://carol.wins.uva.nl/~fontijne/gaigen/about.html
GEOMA http://nklein.com/products/geoma/
GLUCAT http://glucat.sourceforge.net/
MAPLE V package http://math.tntech.edu/rafal/cliff5/
NONEUCLID http://math.rice.edu/~joel/NonEuclid/
REDUCE packages http://hermes.ffn.ub.es/~jmparra
Journal pages
Advances in Applied Clifford Algebras http://www.clifford.org/journals/jadvclfa.html
Subject pages
European Mathematical Information Service http://www.emis.de/

The Geometry Center http://www.geom.umn.edu/
The geometry of the sphere http://math.rice.edu/~pcmi/sphere/
Geometric Calculus: Research and Development http://modelingnts.la.asu.edu/
The Geometric Algebra Research Group (Cambridge) http://www.mrao.cam.ac.uk/~clifford/
Clifford Research Group (Gent) http://cage.rug.ac.be/~fb/crg/
INDEX
A bisectors,
addition identities, 57 of the sides, 70
affine plane, 31 of the angles, 72
affinity, 51 Bretschneider’s theorem 91, 238 (9.4)
algebra Brianchon’s theorem, 138, 248, (11.8)
fundamental theorem, 24
geometric, of the C
Euclidean space, 170 Cartesian equation of a circle, 80
pseudo-Euclidean space, 188 Cauchy-Riemann conditions, 19
algebraic equation of a line, 37 Cauchy’s
algebraic equations of a circle, 80 theorem, 20
alignment of three points, 35 integral formula, 21
altitudes of a triangle, 73 central
analytic equation for conics, 126
continuation, 151 projection
complex functions, 19 of the sphere, 177
hyperbolic functions, 147 of the hyperboloid, 191
analyticity conditions centroid, 69
complex, 19 Ceva’s theorem, 64
hyperbolic, 148 cevian lines, 69
angle Chasles' theorem, 122
between circles, 89 circle
between Euclidean vectors, 12 of the nine points, 85
between hyperbolic vectors, 156 inscribed in a triangle, 72, 73
of parallelism, 206, 260 (15.1) circumscribed to a triangle, 70, 73
inscribed in a circle, 55 circles, 80, 236
antigraphy, 116 radical axis, 89
Apollonius' theorem radical centre, 90
theorem for the ellipse, 128 circular functions, 54
theorem for the hyperbola, 163 circumcentre, 70
lost theorem, 79, 229 (8.3) circumscribed quadrilateral, 88
area, 33 Clifford, vi, viii, 8
of a parallelogram, 4 Clifford algebras
of a triangle, 35, 68 isomorphism of, 152
of a spherical triangle, 176 complex analytic functions, 19
of Lobachevskian triangle, 206, complex numbers, 13
263 (15.6) binomial form, 13
on the hyperbolic plane, 161 polar and trigonometric form, 13
alignment of three points, 35 matrix representation of, 139
analytic functions, 19 algebraic operations, 14
angle inscribed in a circle, 55 complex plane, 18
associative property, 5 conformal
azimuthal equivalent projection of the cylindrical projection, 202
hyperboloid, 200 transformations, 66
congruence
B of hyperbolic segments, 158
base of vectors for the plane, 10 congruence of Lobachevskian triangles,
barycentric coordinates, 33 205
Beltrami’s disk, 191 conic sections, 117
binomial form of a complex number, 15 central equations, 126
in barycentric coordinates, 132 distortion, 177

directrices, 120 double angle identities, 55
foci, 120 dual coordinates, 43
matrix of, 136 dual spherical triangle, 175
passing through five points, 131 duality principle, 43
perpendicular to
tangent to, 124 E
tangential conic, 132 eccentricity, 118
vectorial equation, 121 ellipse,
conical projections central equation, 126
of the sphere, 184 Apollonius’ diameters, 128
of the hyperboloid, 203 equations
conjugate of a circle, 80
diameters, 129, 163 of a line, 36-40
of a complex number, 16 of conics, 119, 121, 126
of a hyperbolic number, 141 equivalent cylindrical projection
convergence of the sphere, 184
of complex powers series, 23 of the hyperboloid, 203
of hyperbolic powers series, 150 Euler's line, 76
radius of, 22 Euclidean space, 170
square of, 150 excess, spherical, 177
coordinate systems, 31 exponential
coordinates complex, 14
barycentric, 33 hyperbolic, 143
Cartesian, 36 exterior product, 4
dual, 43
homogeneous, 102 F
correlation, 134 Fermat's theorem, 77
cosines law focal distance
Euclidean, 60, 223 (6.1) of a conic, 121
hyperbolic, 164 of a photograph, 186, 258 (14.4)
hyperboloidal, 198 foci of a conic, 120
spherical, 174, 175 function of Riemann, 150
cross ratio
complex, 92 G
projective, 99 geometric algebra
cylindrical projections of the vectorial plane, 8
of the sphere, 182 of the pseudo-Euclidean space, 188
of the hyperboloid, 201 Gibbs, vi, viii
cyclic quadrilaterals, 87 gnomonic projection, 177
Grassmann, vii,viii
D
De Moivre's identity, 58 H
Desargues’ theorem, 47 Hamilton, viii, 8, 171
diameters harmonic
of the ellipse, 128 characteristic, 94
of the hyperbola, 163 ranges, 94
dilatations, 30 Heron’s formula, 9, 208 (1.6)
direct similarity, 61 hierarchy of algebraic operations, 8
distance between two points Hilbert, 190
on the Euclidean plane, 33 homogeneous coordinates, 102
on the hyperbolic plane, 160 homography, 96
distance from a point to a line, 37 homologous sides, 61
homology, 110 lines on the Euclidean plane, 31

homothety, 65 linear combination of two vectors, 10
horocycle, 206, 263 (15.4) Liouville’s theorem, 24
hyperbola Lobachevskian
central equation, 126 geometry, 188
hyperbolic trigonometry, 196
analyticity, 150 logarithm
analytic functions, 147 complex, 20
cosine, 142 hyperbolic, 143
exponential and logarithm, 143
numbers, 140 M
polar form, 144 McLaurin series, 22
powers and roots, 144 matrix representation, 139
plane medians, 69
distance between points Menelaus’ theorem, 63
160 Mercator's projection, 183
area on , 161 Möbius, viii, 33, 78
sector, 142 transformation, 96
similarity, 167 modulus
sine, 142 of a complex number, 13
tangent, 142 of a hyperbolic number, 141
trigonometry, 141 Mollweide’s formulas, 60
vectors, 154
hyperboloid of two sheets, 190 N
hyperboloidal Napier’s rule, 176
arc, 193 Napoleon’s theorem, 78
defect, 206, 263 (15.5) Newton’s theorem, 138
triangle, 196 nine-point circle, 85
trigonometry, 196 non-Euclidean geometry, 205
I O
imaginary unity, 13 opposite similarity, 62
incentre, 72 orthocentre, 73
infinity, points at, 102 orthogonal hyperbolic vectors, 157
inner product orthographic projection of the sphere, 181
of hyperbolic vectors, 155 orthonormal bases, 11
intercept equation of a line, 40 outer product
intersection of Euclidean vectors, 4
of a line with a circle, 80 of hyperbolic vectors, 155
of two lines, 41
inverse P
of a vector, 7 parametric equations of a line, 36
trigonometric functions, 59 Pascal’s theorem, 138, 248, (11.7)
inversions of vectors, 29 Pauli’s matrices, 153
inversion with respect to a circle, 83 Peano, vi, vii, 128
isometries pencil of lines, 41
hyperbolic, 158 permutation of complex and vector, 17
isomorphism of Clifford algebras, 152 permutative property
isosceles triangle theorem, 158 of Euclidean vectors, 6, 9, 207 (1.5)
of hyperbolic vectors, 155
L perspectivity, 103
Lauren series, 22 Peter's projection, 184
Leibniz, vi, 190 Plate Carré projection, 182
line at the infinity, 47 Poincaré’s
disk, 198 quaternions, 8, 171

half-plane, 200 quotient
points, 31 of two vectors, 7
alignment, 31 of complex numbers, 16
at the infinity, 102
on the Euclidean plane, 214 R
polar equation of a line, 40 radical axis of two circles, 89
polar equation of a circle, 82 radical centre of three circles, 90
polar form of a complex number, 14 ratio
polarities, 134 complex cross, 92
poles, 23 of similarity, 61
polygon projective cross, 99
sum of the angles, 53 single, 65
power of a point reflections of vectors, 28
with respect to a circle, 82 residue theorem, 24
with respect to a hyperbola, 168 reversal, 168
product Riemann’s function, 150
of a vector and a real number, 2 rotations of vectors, 27, 139
of two vectors, 2
of three vectors, 5 S
of four vectors, 7 sector of a hyperbola, 142
inner series
of hyperbolic vectors, 155 Lauren, 22
outer McLaurin, 22
of hyperbolic vectors, 155 Taylor, 21
projection similarity
of the spherical surface direct, 61
central or gnomonic, 177 opposite, 62
conical, 184 hyperbolic, 167
Mercator’s, 183 Simson’s theorem, 91, 237, (9.3)
Peter's projection, 184 sines law
Plate Carré’s, 182 Euclidean, 60, 223 (6.1)
orthographic, 181 hyperbolic, 164
stereographic, 180 hyperboloidal, 197
of the hyperboloid spherical, 173
central (Beltrami’s), 191 single ratio, 65
conical, 203 singularities, 23
stereographic, 198 slope-intercept equation of a line, 40
azimuthal equivalent, 200 space
cylindrical equidistant, 201 Euclidean, 170
cylindrical conformal, 202 pseudo-Euclidean, 188
cylindrical equivalent, 203 spherical triangle
projective cross ratio, 99 excess of, 177
projectivity, 103 dual, 175
pseudo-Euclidean area of, 176
plane, 154 spherical trigonometry, 172
space, 188 spherical coordinates, 182
Ptomely’s theorem, 115, 240, (10.1) stereographic projection
of the sphere, 180
Q of the hyperboloid, 198
quadrilaterals strict triangles, 173
cyclic, 87 sum of the angles
circumscribed, 88 of a polygon, 53
of a hyperbolic triangle, 160 spherical, 172

sum of trigonometric functions, 56
V
vector
T addition, 1
tangents law 60, 169, 254 (13.5) product, 2
tangent to a conic, 124 transformations, 213
Taylor series, 22 vectorial equation of a line, 36
tetranions, 188
theorem W
of the algebra, fundamental, 24 Weierstrass’
of Apollonius, 128, 138, 163, coordinates, 201
(11.3) model, 190
lost, 78, 229, (8.3)
of Brianchon, 138, 248, (11.8)
of Cauchy, 20
of Ceva, 64
of Chasles, 122
of Desargues, 47
of Fermat, 77
of Liouville, 24
of Menelaus, 63
of Napoleon, 78
of Newton, 138
of Pascal, 138, 248, (11.7)
of Ptolemy, 115, 240, (10.1)
of Simson, 91, 237, (9.3)
of the residue, 24
of the isosceles triangle, 158
theorems
projective demonstration, 108
about hyperbolic angles, 160
transformations
of vectors, 213
special conformal, 115, 242, (10.5)
translations, 31
triangle
dual spherical, 175
triangular inequality, 34
trigonometric
addition identities, 57
double angle identities, 55
form of a complex number, 14
functions, 54
fundamental identities, 54
inverse functions, 59
sum of functions, 56
product of vectors
trigonometry
circular, 53
hyperbolic, 141
hyperboloidal (Lobachevskian),
196
CHRONOLOGY OF THE GEOMETRIC ALGEBRA
1679 Letters of Leibniz to Huygens on the characteristica geometrica.

1799 Publication of Om Directionens analytiske Betegning by Caspar Wessel with scarce
diffusion.
1805 Birth of William Rowan Hamilton at Dublin.
1806 Publication of Essai sur une manière de représenter les quantités imaginaires dans les
constructions géométriques by Jean Robert Argand.
1809 Birth of Hermann Günther Grassmann at Stettin.
1818 Death of Wessel
1822 Death of Argand.
1827 Publication of Der barycentrische Calcul by Möbius at Leipzig.
1831 Birth of James Clerk Maxwell at Edinburgh.
1831 Birth of Peter Guthrie Tait.
1839 Birth of Josiah Willard Gibbs at New Haven.
1843 Discovery of the quaternions by Hamilton.
1844 Publication of Die Lineale Ausdehnungslehre (first edition), where Grassmann presents the
anticommutative product of geometric unities (outer product).
1845 Birth of William Kingdon Clifford at Exeter.
1847 Publication of Geometric Analysis with a foreword by Möbius, memoir with which
Grassmann won the prize to whom developed the Leibniz’s characteristica geometrica.
1850 Birth of Oliver Heaviside at London.
1853 Publication of Lectures on Quaternions where Hamilton introduces the nabla operator
(gradient).
1862 Publication of the second edition of Die Ausdehnungslehre.
1864 Publication of A dynamical theory of the electromagnetic field by Maxwell, where he defines
the divergence and the curl.
1865 Death of Hamilton.
1866 Posthumous publication of Hamilton’s Elements of Quaternions.
1867 Publication of Elementary Treatise on Quaternions by Tait.
1873 Publication of Introduction to Quaternions by Kelland and Tait.
1873 Maxwell publishes the Treatise on Electricity and Magnetism where he writes the equations
of electromagnetism with quaternions.
1877 Publication of the Grassmann’s paper «Der Ort der Hamilton’schen Quaternionen in der
Ausdehnungslehre».
1877 Death of Grassmann.
1878 Publication of the paper «Applications of Grassmann's Extensive Algebra» by Clifford where
he makes the synthesis of the systems of Grassmann and Hamilton.
1879 Death of Maxwell.
1879 Death of Clifford.
1880 Publication of Lipschitz’ Principes d’un calcul algebraic.
1881 Private printing of Elements of Vector Analysis by Gibbs.
1886 Publication of Lipschitz’ Untersuchungen uber die Summen von Quadraten.
1886 Publication of the Gibbs’ paper «On multiple algebra».
1888 Publication of Peano’s Calcolo geometrico secondo l’Ausdehnungslehre di H. Grassmann
preceduto dalle operazione della logica deduttiva.
1891 Oliver Heaviside publishes «The elements of vectorial algebra and analysis» in The
Electrician Series.
1895 Publication of Peano’s «Saggio di Calcolo Geometrico».
1901 Death of Tait.
1901 Wilson publishes the Gibbs’ lessons in Vector Analysis.
1903 Death of Gibbs.
1925 Death of Heaviside.

1926 Wolfgang Pauli introduces his matrices to explain the electronic spin.
1928 Publication of the paper «The Quantum Theory of Electron», where Paul A. M. Dirac defines
a set of 4×4 anticommutative matrices built from the Pauli’s matrices.
This comparative diagram of the life and works of the authors of (or related with)
the geometric algebra visualises and summarises the chronology. The XIX century may be
properly called the century of the geometric algebra. Note the premature death of Clifford,
which caused the delay in the development of the geometric algebra along the XX
century.
1800 1820 1840 1860 1880 1900

| | | | | |
Wessel**************
Argand*************************
Hamilton *******************************
Grassmann ***********************************
Maxwell *************************
Tait ************************************
Gibbs *********************************
Clifford ******************
Heaviside **************************************
Essai sur ... |

Der barycentrische Calcul |
Die Ausdehnungslehre |
Lectures on Quaternions |
Elements on Quaternions |
Elementary Treatise on Quaternions |
Treatise on Electricity and Magnetism |
Applications of Grassmann's Extensive Algebra |
Elements of Vector Analysis |
The Electrician Series |
Vector Analysis |

Treatise of Plane Geometry Through Geometric Algebra - R G Calvet

Uploaded by

Copyright:

Available Formats

Treatise of Plane Geometry Through Geometric Algebra - R G Calvet

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Treatise of Plane Geometry Through Geometric Algebra - R G Calvet

Uploaded by

Copyright:

Available Formats

What is the book about?

What is the book about?

What are some of the key figures mentioned in the text?

What are some of the key figures mentioned in the text?

TREATISE OF PLANE GEOMETRY

THROUGH GEOMETRIC ALGEBRA

Ramon González Calvet

Ramon González Calvet

The geometric algebra,

THROUGH GEOMETRIC ALGEBRA

Dr. Ramon González Calvet

 Ramon González Calvet ( [email protected] )

First Catalan edition: June 1996

Josep Manel Parra i Serra, June 2001

Departament de Física Fonamental

[“Secondly, one requests which of the two following qualities must be

Jean le Rond D'Alembert (1717-1783)

“The geometric Calculus differs from the Cartesian Geometry in that

Initially proposed by Leibniz7 (characteristica geometrica) with the aim of

“And the growth in this century of the so-called synthetic as opposed to

“Indeed these various methods of geometric calculus do not at all

“For I remain completely confident that the labour which I have

Ramon González Calvet

Cerdanyola del Vallès, June 2001

First Part: The vector plane and the complex numbers

2. A base of vectors for the plane. (June 24th 2000)

4. Transformations of vectors. (August 4th 2000, updated July 21st 2002)

Second Part: The geometry of the Euclidean plane

11. Conics (November 12th 2000, updated July 21st 2002)

Third Part: Pseudo-Euclidean geometry

Fourth Part: Plane projections of tridimensional spaces

15. Hyperboloidal geometry in the pseudo-Euclidean space (Lobachevsky's

FIRST PART: THE VECTOR PLANE AND THE COMPLEX NUMBERS

1. THE VECTORS AND THEIR OPERATIONS

The addition of two vectors u + v

Product of a vector and a real number

ku=uk Figure 1.3

Two vectors with different

Product of two vectors

1) To be distributive with regard to the vector addition:

2) The square of a vector must be equal to the square of its length. By

a⊥b ⇒ c2 = a2 + b2 ⇒ ab+ba=0 ⇒ ab=−ba

That is, the product of two perpendicular vectors is anticommutative.

But if the vectors have opposite directions:

Some example of physical

In conclusion, the geometric product of two proportional vectors is commutative

ab=a·b+a∧b ⇒  a b 2 = a · b2 +a ∧ b2

Product of three vectors: associative property

It is demanded as the fourth property that the product of three vectors be

Since a and b' have the same

On the other hand, a can be Figure 1.8

( a b ) c = a'' ( b'' c ) = a'' bc

Although the geometric

( a b ) c = a'' bc = cb a'' = c b'' a'' = c ( b a )

That is, the triple product has the property:

Product of four vectors

a b c d = a' b' c d = a'  b c d = b c a' d

If both vectors are perpendicular, we must

Inverse and quotient of two vectors

vector proportionality. We say that a is proportional to c as b is to d when their moduli