See

discussions, stats, and author profiles for this publication at: https://www.researchgate.net/publication/266738365

Divine Proportions: Rational Trigonometry to

Universal geometry

Book · October 2005

CITATIONS READS

46 7,358

1 author:

Norman Wildberger
UNSW Sydney
75 PUBLICATIONS 587 CITATIONS

SEE PROFILE

Some of the authors of this publication are also working on these related projects:

The hyperbolic equilateral hyperbola View project

Rational trigonometry and applications View project

All content following this page was uploaded by Norman Wildberger on 22 July 2015.

The user has requested enhancement of the downloaded file.

 Title0.ai 15/08/2005 2:28:33 PM

Divine Proportions

CM

MY

CY

CMY

K
 Title1.ai 14/08/2005 4:28:20 PM

Divine Proportions
Rational Trigonometry to
C Universal geometry
M

CM

MY

N J Wildberger
CY

CMY

K
Assoc/Prof N J Wildberger
School of Mathematics
UNSW
Sydney 2052
Australia

email: n.wildberger@unsw.edu.au

webpages: http://web.maths.unsw.edu.au/~norman/

Copyright 2005 Wild Egg Pty Ltd

Typeset in Latex
Cover design by Alex Snellgrove
Printed on 100% acid-free paper by BPA Print Group
Printed in Australia

National Library of Australia Cataloguing-in-Publication Data

Wildberger, Norman John,

Divine proportions : rational trigonometry to universal
geometry

Bibliography.
Includes Index.
ISBN 0 9757492 0 X.
1. Trigonometry. 2. Geometry. I. Title.

516.2

Mathematics Subject Classification: 51XX, 14XX

Published in Australia by Wild Egg Pty Ltd

http://wildegg.com
To my parents, John and Trudy Wildberger
 Preface

Divine Proportions: Rational Trigonometry to Universal Geometry introduces a

remarkable new approach to trigonometry and Euclidean geometry, with dramatic
implications for mathematics teaching, industrial applications and the direction of
mathematical research in geometry.

The key insight is that geometry is a quadratic subject. So rational trigonometry

replaces the quasi-linear notions of distance and angle with the related, but more
elementary, quadratic concepts of quadrance and spread, thus allowing the development
of Euclidean geometry over any field. This text covers the key definitions and results of
this new theory in a systematic way, along with many applications including Platonic
solids, projectile motion, Snell’s law, the problems of Snellius-Pothenot and Hansen,
and three-dimensional volumes and surface areas.

The message of this book is controversial, but it will be hard to deny the power of its
content. Teachers and students will benefit from a simpler and cleaner theory, which
takes less than half of the usual time to learn. Engineers, surveyors and scientists now
have a fascinating new set of tools that increase accuracy and reduce computation time
for geometric problems. Mathematicians finally have a logically coherent framework for
metrical geometry, which opens up exciting areas for investigation in algebraic
geometry, number theory, combinatorics and special functions.

Even though the content is in some sense elementary, this text is intended for a
mathematically mature audience. The novelty will provide some challenge even for
professional mathematicians.

With Divine Proportions it should become clearer why so many students are turned off
mathematics by trigonometry. The current curriculum attempts to arm them with
machine guns to hunt rabbits. Most students sense intuitively that this is not what
they need, but until now there has been no reasonable alternative to focus the minds of
educators. This book finally addresses the failure of modern geometry to win the
minds of young people, and provides the mathematical foundation for a dynamic and
elegant new approach to teaching trigonometry and geometry.

N J Wildberger
Sydney, 2005

vii
viii PREFACE

Acknowledgements

It is a pleasure to warmly thank Walter Bloom, Michael Duffy, Kimmo Eriksson,

Shuxiang Goh, Jack Hall, Chris Halse-Rogers, Vinh Le and John Wildberger for careful
readings of portions of the manuscript and for suggesting valuable improvements.

Helen Armstrong, Peter Brown, Daniel Chan, Peter Donovan, Hendrik Grundling,
Bruce Harvey, Rick Hill, Michael Hirschhorn, Krista Ruffolo, George Szekeres and
Martin Wildberger contributed useful comments and discussions. Dolores Knox, Thai
Son Huang, Ming Huo, Alex Ma and Ross Moore helped with layout, diagrams and
computer programming. Alex Snellgrove created the cover design. Many thanks to
them all for making this a more sensible, readable and attractive book.

This work owes a debt to the efforts of many expositors and researchers of geometry.
Some of the texts which have been particularly helpful to me have been Introduction to
geometry [Coxeter], Geometry Revisited [Coxeter-Greitzer], The Elements [Euclid], the
Concise Encyclopedia of Mathematics [Weisstein], Surveying Problems and Solutions
[Shepherd], Metric Affine Geometry [Snapper-Troyer], Trigonometric Delights [Maor],
and last but not least, my favourite 20th century mathematics text, 100 Great
Problems of Elementary Mathematics [Dorrie].

Without wonderful developments in computer software, this project could not have
been completed in the given time frame. I would like to thank Donald Knuth for his
invention of TeX, the many developers of LaTeX, MacKichan Software for Scientific
Workplace, the makers of CorelDraw and Geometric SketchPad, and the people at
Microsoft for a dependable operating system.

Thanks go to the Australian Research Council for financial support. The School of
Mathematics at UNSW provided a supportive and friendly work environment, and I
thank all my colleagues who offered encouragement despite the heretical nature of the
project. Finally I would like to thank Kim and Ali, for moral support during the
writing of this book.
 Contents

Introduction xv

I Preliminaries 1
1 Overview 3
1.1 Introducing quadrance and spread . . . . . . . . . . . . . . . . . . . . . . 3
1.2 Laws of rational trigonometry . . . . . . . . . . . . . . . . . . . . . . . . . 9
1.3 Why classical trigonometry is hard . . . . . . . . . . . . . . . . . . . . . . 11
1.4 Why rational trigonometry is easier . . . . . . . . . . . . . . . . . . . . . . 13
1.5 Comparison example . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
1.6 Ancient Greek triumphs and difficulties . . . . . . . . . . . . . . . . . . . 17
1.7 Modern ambiguities . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19

2 Background 21
2.1 Fields . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
2.2 Proportions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
2.3 Identities and determinants . . . . . . . . . . . . . . . . . . . . . . . . . . 27
2.4 Linear equations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30
2.5 Polynomial functions and zeroes . . . . . . . . . . . . . . . . . . . . . . . 31
2.6 Quadratic equations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32

3 Cartesian coordinate geometry 35

3.1 Points and lines . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35
3.2 Collinear points and concurrent lines . . . . . . . . . . . . . . . . . . . . . 38
3.3 Parallel and perpendicular lines . . . . . . . . . . . . . . . . . . . . . . . . 40
3.4 Parallels and altitudes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41
3.5 Sides, vertices and triangles . . . . . . . . . . . . . . . . . . . . . . . . . . 43
3.6 Quadrilaterals . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45
3.7 Affine combinations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46
3.8 Perpendicular bisectors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 50

4 Reflections 51
4.1 Affine transformations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51
4.2 Lineations and reflection sequences . . . . . . . . . . . . . . . . . . . . . . 53

ix
x CONTENTS

II Rational trigonometry 57
5 Quadrance 59
5.1 Quadrances of triangles and quadrilaterals . . . . . . . . . . . . . . . . . . 59
5.2 Triple quad formula . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 63
5.3 Pythagoras’ theorem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 65
5.4 Quadrance to a line . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 66
5.5 Quadrea . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 68
5.6 Archimedes’ formula . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 70
5.7 Quadruple quad formula . . . . . . . . . . . . . . . . . . . . . . . . . . . . 70

6 Spread 73
6.1 Spreads of triangles and quadrilaterals . . . . . . . . . . . . . . . . . . . . 73
6.2 Cross . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 74
6.3 Twist . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 75
6.4 Ratio theorems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 76
6.5 Complementary spreads . . . . . . . . . . . . . . . . . . . . . . . . . . . . 79
6.6 Spread law . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 80
6.7 Cross law . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 81
6.8 Spreads in coordinates . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 84
6.9 Vertex bisectors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 85

7 Triple spread formula 89

7.1 Triple spread formula . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 89
7.2 Triple cross formula . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 92
7.3 Triple twist formula . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 93
7.4 Equal spreads . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 94
7.5 Spread reflection theorem . . . . . . . . . . . . . . . . . . . . . . . . . . . 95
7.6 Examples using different fields . . . . . . . . . . . . . . . . . . . . . . . . 96
7.7 Quadruple spread formula . . . . . . . . . . . . . . . . . . . . . . . . . . . 98

8 Spread polynomials 101

8.1 Combining equal spreads . . . . . . . . . . . . . . . . . . . . . . . . . . . . 101
8.2 Spread polynomials . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 102
8.3 Special cases . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 104
8.4 Explicit formulas . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 106
8.5 Orthogonality . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 109
8.6 Composition of spread polynomials . . . . . . . . . . . . . . . . . . . . . . 110
8.7 Cross polynomials . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 112

9 Oriented triangles and turns 113

9.1 Oriented sides, vertices and triangles . . . . . . . . . . . . . . . . . . . . . 113
9.2 Turns of oriented vertices . . . . . . . . . . . . . . . . . . . . . . . . . . . 114
9.3 Signed areas . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 116
CONTENTS xi

III Universal Geometry 119

10 Triangles 121
10.1 Isosceles triangles . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 121
10.2 Equilateral triangles . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 125
10.3 Right triangles . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 126
10.4 Congruent and similar triangles . . . . . . . . . . . . . . . . . . . . . . . . 127
10.5 Solving triangles . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 129

11 Laws of proportion 131

11.1 Triangle proportions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 131
11.2 Quadrilateral proportions . . . . . . . . . . . . . . . . . . . . . . . . . . . 133
11.3 Two struts theorem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 134
11.4 Stewart’s theorem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 136
11.5 Median quadrance and spread . . . . . . . . . . . . . . . . . . . . . . . . . 137
11.6 Menelaus’ and Ceva’s theorems . . . . . . . . . . . . . . . . . . . . . . . . 139

12 Centers of triangles 143

12.1 Perpendicular bisectors and circumcenter . . . . . . . . . . . . . . . . . . 143
12.2 Formulas for the circumcenter . . . . . . . . . . . . . . . . . . . . . . . . . 146
12.3 Altitudes and orthocenter . . . . . . . . . . . . . . . . . . . . . . . . . . . 147
12.4 Formulas for the orthocenter . . . . . . . . . . . . . . . . . . . . . . . . . 149
12.5 Incenters . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 150

13 Isometries 155
13.1 Translations, rotations, reflections . . . . . . . . . . . . . . . . . . . . . . 155
13.2 Classifying isometries . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 157

14 Regular stars and polygons 159

14.1 Regular stars . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 159
14.2 Order three stars . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 160
14.3 Order five stars . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 161
14.4 Order seven stars . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 162
14.5 Regular polygons . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 163

15 Conics 167
15.1 Centers of conics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 167
15.2 Circles and ribbons . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 169
15.3 Parabolas . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 172
15.4 Quadrolas . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 174
15.5 Grammolas . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 175
15.6 Intersections with lines . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 176

16 Geometry of circles 177

16.1 Diameters and chords . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 177
16.2 Spreads in a circle . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 178
16.3 Parametrizing circles . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 181
xii CONTENTS

17 Quadrilaterals 183
17.1 Cyclic quadrilaterals . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 183
17.2 Circumquadrance formula . . . . . . . . . . . . . . . . . . . . . . . . . . . 185
17.3 Cyclic quadrilateral quadrea . . . . . . . . . . . . . . . . . . . . . . . . . . 187
17.4 Ptolemy’s theorem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 190
17.5 Four point relation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 191

18 Euler line and nine point circle 193

18.1 Euler line . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 193
18.2 Nine point circle . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 194

19 Tangent lines and tangent conics 197

19.1 Translates and Taylor conics . . . . . . . . . . . . . . . . . . . . . . . . . . 197
19.2 Tangent lines . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 199
19.3 Higher order curves and tangents . . . . . . . . . . . . . . . . . . . . . . . 202
19.4 Folium of Descartes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 203
19.5 Lemniscate of Bernoulli . . . . . . . . . . . . . . . . . . . . . . . . . . . . 207

IV Applications 211
20 Triangle spread rules 213
20.1 Spread ruler . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 213
20.2 Line segments, rays and sectors . . . . . . . . . . . . . . . . . . . . . . . . 214
20.3 Acute and obtuse sectors . . . . . . . . . . . . . . . . . . . . . . . . . . . 216
20.4 Acute and obtuse triangles . . . . . . . . . . . . . . . . . . . . . . . . . . 218
20.5 Triangle spread rules . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 219

21 Two dimensional problems 221

21.1 Harmonic relation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 221
21.2 Overlapping triangles . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 222
21.3 Eyeball theorem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 224
21.4 Quadrilateral problem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 225

22 Three dimensional problems 227

22.1 Planes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 227
22.2 Boxes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 228
22.3 Pyramids . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 230
22.4 Wedges . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 231
22.5 Three dimensional Pythagoras’ theorem . . . . . . . . . . . . . . . . . . . 232
22.6 Pagoda and seven-fold symmetry . . . . . . . . . . . . . . . . . . . . . . . 233

23 Physics applications 235

23.1 Projectile motion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 235
23.2 Algebraic dynamics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 238
23.3 Snell’s law . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 239
23.4 Lorentzian addition of velocities . . . . . . . . . . . . . . . . . . . . . . . . 242
CONTENTS xiii

24 Surveying 243
24.1 Height of object with vertical face . . . . . . . . . . . . . . . . . . . . . . 243
24.2 Height of object with inaccessible base . . . . . . . . . . . . . . . . . . . . 244
24.3 Height of a raised object . . . . . . . . . . . . . . . . . . . . . . . . . . . . 245
24.4 Regiomontanus’ problem . . . . . . . . . . . . . . . . . . . . . . . . . . . . 246
24.5 Height from three spreads . . . . . . . . . . . . . . . . . . . . . . . . . . . 247
24.6 Vertical and horizontal spreads . . . . . . . . . . . . . . . . . . . . . . . . 248
24.7 Spreads over a right triangle . . . . . . . . . . . . . . . . . . . . . . . . . . 249
24.8 Spherical analogue of Pythagoras’ theorem . . . . . . . . . . . . . . . . . . 250

25 Resection and Hansen’s problem 251

25.1 Snellius-Pothenot problem . . . . . . . . . . . . . . . . . . . . . . . . . . . 251
25.2 Hansen’s problem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 255

26 Platonic solids 259

26.1 Tetrahedron . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 260
26.2 Cube . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 261
26.3 Octahedron . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 262
26.4 Icosahedron . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 263
26.5 Dodecahedron . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 264

27 Rational spherical coordinates 265

27.1 Polar spread and quadrance . . . . . . . . . . . . . . . . . . . . . . . . . . 265
27.2 Evaluating π2 /16 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 269
27.3 Beta function . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 270
27.4 Rational spherical coordinates . . . . . . . . . . . . . . . . . . . . . . . . . 271
27.5 Surface measure on a sphere . . . . . . . . . . . . . . . . . . . . . . . . . . 276
27.6 Four dimensional rational spherical coordinates . . . . . . . . . . . . . . . 277
27.7 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 278

A Rational polar equations of curves 279

B Ellipson 287

Theorems with pages and Important Functions 289

xiv CONTENTS
 Introduction

This book revolutionizes trigonometry, re-evaluates and expands Euclidean geometry,

and gives a simpler and more natural approach to many practical geometric problems.
This new theory unites the three core areas of mathematics–geometry, number theory
and algebra–and expels analysis and infinite processes from the foundations of the
subject. Learning trigonometry and geometry should be easier than it currently is, and
Divine Proportions attempts to develop a complete and precise alternative framework
from which educators can draw.

The concise format is suitable for mathematicians, as well as mathematically inclined

surveyors, engineers and physical scientists. Mathematics teachers, undergraduate
mathematics majors, gifted high school students and dedicated amateurs with a strong
interest in geometry and good skills at algebraic manipulation should also be able to
follow the development. Others should wait for the author’s next book on the subject.

Advanced mathematical knowledge, such as linear algebra, number theory and group
theory, is generally not needed. Indeed the geometry in this book often provides
motivation for these subjects. Some familiarity with elementary aspects of fields is
useful. The chapters on physics and rational spherical coordinates require calculus.

Rational trigonometry

The new form of trigonometry developed here is called rational trigonometry, to

distinguish it from classical trigonometry, the latter involving cos θ, sin θ and the many
trigonometric relations currently taught to students. An essential point of rational
trigonometry is that quadrance and spread, not distance and angle, are the right
concepts for metrical geometry (i.e. a geometry in which measurement is involved).

Quadrance and spread are quadratic quantities, while distance and angle are almost,
but not quite, linear ones. The quadratic view is more general and powerful. At some
level, this is known by many mathematicians. When this insight is put firmly into
practice, as it is here, a new foundation for mathematics and mathematics education
arises which simplifies Euclidean and non-Euclidean geometries, changes our
understanding of algebraic geometry, and often simplifies difficult practical problems.

xv
xvi INTRODUCTION

Quadrance measures the separation of points, and spread measures the separation of
lines. It turns out that

quadrance = (distance)2
2
spread = ( sin (angle))

although the actual definitions used in this text are independent of distance, angle and
the trigonometric functions. They are ultimately very simple, based on finite
arithmetic and algebra as taught in schools.

New laws now replace the Cosine law, the Sine law, and the dozens of other
trigonometric formulas that often cause students difficulty. The most important new
laws are the Triple quad formula, the Spread law, the Cross law and the Triple spread
formula. Pythagoras’ theorem, restated in terms of quadrances, also plays a key role.
The derivation of these rules from first principles is straightforward, involving some
moderate skill with basic algebra. The usual trigonometric functions, such as cos θ and
sin θ, play no role at all.

Rational trigonometry deals with many practical problems in an easier and more
elegant fashion than classical trigonometry, and often ends up with answers that are
demonstrably more accurate. In fact rational trigonometry is so elementary that almost
all calculations may be done by hand. Tables or calculators are not necessary, although
the latter certainly speed up computations. It is a shame that this theory was not
discovered earlier, since accurate tables were for many centuries not widely available.

Universal geometry

Because rational trigonometry uses only arithmetic and algebra, it allows the
development of Euclidean geometry in a consistent and general way in an arbitrary
field. This is universal geometry.

Historically metrical geometry has been difficult to develop outside the decimal
numbers, due largely to the transcendental nature of angle. Once liberated from a
dependence on analysis, the scope is much wider, and the opportunities to effectively
harness the power of modern mathematical software, such as MAPLE, MAGMA,
MUPAD, MATHEMATICA, MATLAB and others, increase dramatically.

Many classical geometrical subjects are here reformulated into this more general
framework, such as Heron’s, or Archimedes’ formula, similar triangles, parallelograms,
Pons Asinorum, centroids, orthocenters, theorems of Menelaus and Ceva, Stewart’s
theorem, circumcircles, Brahmagupta’s formula for cyclic quadrilaterals, regular
polygons, the Euler line, the nine point circle and reflection properties of conics.
 xvii

Some topics, such as vertex bisectors and triangle incenters, become dependent on
number theoretic considerations. Many more areas await the attention of researchers.

Universal geometry deals only with geometrical concepts and results which are
common to all fields. The usual notions of convexity and betweenness, for example,
hold only in the rational and decimal number fields and some closely related ones, so
they are not as fundamental as say perpendicularity, which applies to all fields. Over
finite fields a host of new geometrical configurations can now be studied, with many
number theoretic and combinatorial implications.

Universal geometry does not require a prior understanding of classical geometry. It is a

complete, independent system that stands logically separate from the existing
treatment of Euclid. It neither contains classical geometry nor is contained by it,
although there are many close connections. One of the main features of this new
theory is its logical precision and clarity–at least in principle!

A subtle obstacle to beginning a serious study of geometry is the fact that many of the
main concepts are already familiar from ordinary life. The temptation arises to dismiss
the need for precise definitions, because everyone surely already knows what a point,
line and triangle are. But without very clear and careful statements of the key
concepts, mathematics generally ends up relying on doubtful arguments.

There are three rocks on which most attempts at developing Euclidean geometry
founder. These are

• the ambiguity of defining the ‘continuum’, or the ‘real number line’

• the problem of stating precisely what an ‘angle’ is
• the difficulty in making the jump from two to three dimensions.

The ‘real number line’ involves philosophizing about infinite sets, and confusion with
issues of computability. Angles require hand-waving about ‘rotations’, or ‘lengths of
circular arcs’. ‘Three dimensional geometry’ usually involves pictures and physically
plausible arguments without proper mathematical basis.

All three are serious obstacles. The first two are overcome in this book, the third is
not. To avoid the logical deficiencies with ‘real numbers’, universal geometry works
over an arbitrary field. Furthermore the presentation given here avoids any mention of
‘infinite sets’. To avoid angles, rational trigonometry deals with spreads.

As for the step from two to three dimensions, in principle this can be done, but it
requires considerably more work. With universal geometry, one begins to appreciate
the large gulf between these dimensions, both number theoretically and in terms of
concepts and proofs required.

All theorems in this book will be presented only in the planar two dimensional case.
The applications part of the book however does use three dimensions, so is in parts
guilty of the usual logical obscurity.
xviii INTRODUCTION

Universal geometry naturally leads to a rebirth of the traditional idea of algebraic

geometry as a study of the metrical aspects of curves and varieties, in the spirit of the
classical investigations of the ancient Greeks, as well as the work of Fermat, Newton,
Euler and their contemporaries. This is a more rigid subject than modern algebraic
geometry, and possibly just as rich.

For example, many of the most important and beautiful properties of conics, going
back to Apollonius and Archimedes, occur in a metrical setting. In universal geometry
there are many unanswered questions even about conics. For more general curves and
varieties, this way of thinking opens up a large potential area for investigation, and
unifies aspects of algebraic and differential geometry. Tangent conics and higher degree
generalizations can be used to classify points on curves and for more accurate
approximations in practical problems.

This view of algebraic geometry does not elevate algebraically closed fields to a
position of dominance, and also reaffirms the ordinary intuition that curves and
varieties are often most natural in an affine rather than a projective setting. In fact the
metrical approach extends also to projective varieties.

The foundations of Euclidean geometry established here can be generalized to

non-Euclidean geometries as well, including spherical (or elliptic) geometry, and
hyperbolic geometry too. Universal geometry leads to a yet much broader vista, in
which Euclidean and non-Euclidean geometries merge in a spectacular way to form
chromogeometry. These topics are more advanced, and will be developed elsewhere.

As well as providing new directions for mathematics, universal geometry offers the
possibility for new insights into physics. One reason is that physics is also largely a
quadratic subject. Another is that Einstein’s special theory of relativity fits naturally
into this framework.

Organization of the book

Part I gives first of all an overview of rational trigonometry and how it differs from
classical trigonometry. A brief statement of the main definitions and basic laws will
allow the reader to get an initial feeling for the subject. (This material is developed
carefully in Part II). A preliminary section clarifies terminology and reviews those
aspects of basic mathematics that will be needed.

Cartesian planar geometry is introduced in a systematic manner. Fundamental

concepts are clearly laid out, and the basic results of planar coordinate geometry are
established over a general field. This material is mostly classical, but the definitions
are often novel. Another chapter develops the theory of rotations in points and
reflections in lines, along with the associated lineations, and includes important
formulas that are often absent from geometry courses.

Part II is the heart of the book, and develops rational trigonometry over a general
field, with characteristic two, and sometimes three, excluded for technical reasons.
 xix

The notions of quadrance, spread and quadrea are introduced, as well as the concepts of
cross and twist between lines. The Triple quad formula, Pythagoras’ theorem, Spread
law, Cross law and Triple spread formula are derived, as well as more complicated
extensions involving four quadrances and spreads. The important spread polynomials
naturally appear and have many interesting properties, valid over a general field.

Another chapter introduces oriented triangles and the concept of turn between two
lines, closely related to the twist. The oriented version allows a notion of signed areas
of triangles and n-gons. This completes the basic material on rational trigonometry.

Part III uses rational trigonometry to develop the fundamentals of universal geometry.
Isosceles, equilateral and right triangles are studied, along with the procedures for
solving general triangles. Then the laws of proportion for triangles and quadrilaterals
are derived, along with Stewart’s theorem, Menelaus’ theorem and Ceva’s theorem.
Special lines and centers of triangles are discussed, such as medians, altitudes and
vertex bisectors, along with circumcenters, orthocenters and incenters.

An introduction to general conics includes some surprises, as the usual definitions do

not always generalize well to universal geometry. Circles and parabolas appear, as well
as less familiar conics called ribbons, quadrolas and grammolas. The geometry of circles
is studied. Some basic facts about quadrilaterals, especially cyclic ones, are included,
as well as the Four point relation going back to Euler. The notions of a tangent line
and a tangent conic are introduced, and illustrated with the folium of Descartes and
the lemniscate of Bernoulli.

Every theorem in this book holds over an arbitrary field, with characteristic two
excluded. Although sometimes parallel to existing theorems, and given the same
names, the results are considerably more general. Note that no familiarity with
classical trigonometry or geometry is assumed. The reader need not know what a circle
or an ellipse is, what similar triangles are, or even what a line is. Indeed the reader
should put aside fixed preconceptions in these directions.

Part IV shows how to apply the theory to a wide variety of practical problems using
the decimal numbers, which in practice often means the rational numbers. There are
applications to physics, surveying, including the problems of Snellius-Pothenot and
Hansen, two and three-dimensional situations, and Platonic solids.

The final chapter requires some knowledge of basic calculus, and derives the new
rational polar and spherical coordinates, with applications to the Beta function,
volumes and surface areas of spheres, hyperboloids and other related situations. Some
brief discussion shows how to extend this framework to four dimensional space and
beyond. One Appendix contains a list of rational polar equations of some classical
curves in the plane, and another briefly introduces the beautiful ellipson.

In this book, an equation of equality between previously defined objects generally uses
the equal sign =, while an equation of assignment of a name or a value to a previously
defined object generally uses the assignment sign ≡. This follows E. Bishop’s famous
dictum that "Meaningful distinctions deserve to be maintained." Occasionally both =
and ≡ are used in the same equation, hopefully without undue confusion.
xx INTRODUCTION

All definitions are highlighted in bold. Italics are usually reserved for emphasis.
Theorems are given descriptive names, and begin with capitals. Especially important
material, such as a major Definition, Theorem or Problem, is enclosed in a gray box.
To further structure the content, the ends of Proofs and Solutions are marked by the
symbol while the ends of Examples and Exercises are marked by the symbol ¦.

In universal geometry formulas are particularly significant, and proofs by calculation

necessarily play an increasingly important role as you delve further into the subject. A
computer with a modern algebra package, such as those mentioned earlier, becomes an
indispensable tool into higher investigations, as in the approach of [Zeilberger] (see also
D. Zeilberger’s website for the computer package RENE).

While much effort has been made to avoid ambiguities and to give careful proofs, there
are undoubtedly places in this text where this goal has not been reached. I hope
nevertheless that the framework is sufficiently solid for others to take the subject
further with confidence.

To the reader

Mathematics is a conservative discipline, and it is not easy to acknowledge that

traditional thinking might involve elements of misunderstanding.

In this context, the following analogy may be useful. In the Roman period, which saw
the beginnings of classical trigonometry, arithmetic used Roman numerals (such as the
page numbers in this introduction). Cities were built, students were taught, and an
empire was administered, with an arithmetic that was cumbersome and hard to learn,
at least when compared to the one we now use built from the Arabic-Hindu numerical
system. Today we understand that the difficulty with arithmetic in Roman times was
largely due to the awkward conceptual framework.

Much the same holds, in my opinion, for classical trigonometry–it has been such a
hurdle to generations of students not because of the essential intractability of the
subject, but rather because the basic notions used to study it for the last two thousand
years are not the right ones.

By the time you have finished this book, you should be comfortable with the fact that
geometry is a quadratic subject, requiring quadratic mathematics. Using more or less
linear ideas, such as distance and angle, may be initially appealing but is ultimately
inappropriate. With the natural approach of rational trigonometry, many more people
should be able to appreciate the rich patterns of geometry and perhaps even
experience the joy of mathematical discovery.

This book asks something from you, the reader–an openness to fresh ideas, attention
to detail, and a willingness to explore. In return, it promises to transform your
understanding of mathematics, to empower you with knowledge previously hidden, and
to shed a new light on the beauty and unity of the world around us.
 Part I

Preliminaries

1
 1

Overview

1.1 Introducing quadrance and spread

Trigonometry is the measurement of triangles. In classical trigonometry, measurement

uses distance and angle, while in rational trigonometry measurement uses quadrance
and spread. To appreciate the difference, let’s first consider a specific triangle in the
decimal number plane from both points of view.

Classical measurements

The triangle with side lengths

d1 ≡ 4 d2 ≡ 7 d3 ≡ 5 A3
has respective angles (approximately) q3
7
θ1 ≈ 33.92◦ θ2 ≈ 102.44◦ θ3 ≈ 43.64◦ . 4
q1 q2
A1
There are numerous classical relations between 5 A2
these six quantities, such as the Sums of angles law

θ1 + θ2 + θ3 = 180◦ ,

along with the Cosine law, the Sine law, and others. Typically these laws involve the
trigonometric functions cos θ, sin θ and tan θ, and implicitly their inverse functions
arccos x, arcsin x and arctan x, all of which are difficult to define precisely without
calculus.

3
4 1. OVERVIEW

Rational measurements

In rational trigonometry the most important measurements associated with

the same triangle are the quadrances of the sides,

Q1 = 16 Q2 = 49 Q3 = 25 A3
s3
and the spreads at the vertices 49
16
s1 = 384/1225 s2 = 24/25 s3 = 24/49.
s1 s2
Let’s informally explain the meaning of these A1
25 A2
terms, and show how these numbers were obtained.

Quadrance

Quadrance measures the separation of two points. An easy definition is that quadrance
is distance squared. Of course this assumes that you already know what distance is. A
point in the decimal number plane can be specified by its x and y coordinates with
respect to a fixed pair of rectangular axes, and the usual definition of the distance
|A1 , A2 | between the points A1 ≡ [x1 , y1 ] and A2 ≡ [x2 , y2 ] is
q
2 2
|A1 , A2 | ≡ (x2 − x1 ) + (y2 − y1 ) .

So the quadrance Q (A1 , A2 ) between the points is

2 2
Q (A1 , A2 ) ≡ (x2 − x1 ) +(y2 − y1 ) .

From this point of view, quadrance is the more fundamental quantity, since it does not
involve the square root function. The relationship between the two notions is perhaps
more accurately described by the statement that distance is the square root of
quadrance.

In diagrams, small rectangles along the sides of a triangle indicate that quadrance, not
distance, is being measured, a convention maintained throughout the book.
Occasionally when this is inconvenient, a quadrance is enclosed in a rectangular box.

Spread

Spread measures the separation of two lines. This turns out to be a much more subtle
issue than the separation of two points. Given two intersecting lines such as l1 and l2
in Figure 1.1, we would like to define a number that quantifies how ‘far apart’ the lines
are spread. Historically there are a number of solutions to this problem.
1.1. INTRODUCING QUADRANCE AND SPREAD 5

The most familiar is the notion of angle, which roughly speaking can be described as
follows. Draw a circle of radius one with center the intersection A of l1 and l2 . Let this
circle intersect the lines l1 and l2 respectively at points B1 and B2 . Then ‘define’ the
angle θ between the lines to be the length of the circular arc between B1 and B2 , as in
the first diagram in Figure 1.1.

If the lines are close to being parallel, the angle is close to zero, while if the lines are
perpendicular, the angle turns out to be, after a highly non-trivial calculation going
back to Archimedes, a number with the approximate value 1. 570 796 326 . . . .

l1 l1
B1 B1
q
l2 l2
1 1
B2 B2
A A

Figure 1.1: Separation of two lines: two dubious approaches

Immediately one observes some difficulties with this ‘definition’.

First of all there are two possible choices for B1 , and also
two possible choices for B2 . There are then four possible pairs l1
[B1 , B2 ] to consider. Each such pair divides the circle into two B1
circular arcs. In general there are eight possible circular arcs l2
to measure, and four possible results of those measurements. 1
B2
Furthermore, defining the length of a circular arc is not A
at all straightforward, a point that will be returned to later.
B2
There are other approaches to the question of how to measure
B1
the separation of two lines. For example, with the circle of
radius one as above, and the same choice of points B1 and B2 ,
you could consider the length |B1 , B2 | as the main object of interest, as in the second
diagram in Figure 1.1. When the lines are close to parallel, with the right choices of √ B1
and B2 , |B1 , B2 | is close to zero, and when the lines are perpendicular |B1 , B2 | is 2,
independent of the choices.

However in general there are still two different values that can be obtained. This is less
than optimal, but it is already an improvement on the first method. An even better
choice is the quadrance Q (B1 , B2 ), but the two-fold ambiguity in values remains.

A completely different tack would be to give up on the separation between lines, and
measure only the separation between rays. But a line is a more elementary,
fundamental and general notion than a ray, so this would be a form of capitulation.
6 1. OVERVIEW

l1
B
Q l2
R
C
A
s

Figure 1.2: Separation of two lines: the right approach

The right idea is remarkably simple, and is shown in Figure 1.2. Take any point B 6= A
on one of the lines, and then let C be the foot of the altitude, or perpendicular, from B
to the other line. Then let’s temporarily define the spread s (l1 , l2 ) between the two
lines l1 and l2 to be the ratio of quadrances

Q (B, C) Q
s (l1 , l2 ) ≡ = . (1.1)
Q (A, B) R

This number s ≡ s (l1 , l2 ) is independent of the choice of first line, or the choice of the
point B on it. It is a unique number, somewhere between 0 and 1, which measures the
separation of two lines unambiguously. Note that the circle, which played an important
role in the previous constructions, is now essentially irrelevant.

In diagrams, spreads are placed adjacent to small straight line segments joining the
relevant lines, instead of the usual circular arcs used to denote angles. There are four
possible places to put this spread, each equivalent, as in Figure 1.3.

l2

s s
l1
s s

Figure 1.3: Four possible labellings

Why do we not consider the spread to be the ratio of lengths |B, C| / |A, B|? This
becomes clearer when the spread is expressed in terms of the coordinates of the lines.
1.1. INTRODUCING QUADRANCE AND SPREAD 7

Spread in terms of coordinates

A line in the plane can be specified by an equation in the form y = mx + b, or the more
general form ax + by + c = 0. The latter is preferred in this text, for a variety of good
reasons, among them being that it includes vertical lines, generalizes well to projective
geometry, and is better suited for the jump to linear algebra and higher dimensions.

The equation ax + by + c = 0 of a line l is not unique, as you can multiply by an

arbitrary non-zero number. This means that it is actually the proportion a : b : c that
specifies the line, and later on we define the line in terms of this proportion.

Two lines l1 and l2 with respective equations a1 x + b1 y + c1 = 0 and a2 x + b2 y + c2 = 0

are parallel precisely when
a1 b2 − a2 b1 = 0
and perpendicular precisely when

a1 a2 + b1 b2 = 0.

The spread between them is unchanged if the lines are moved while remaining parallel,
so you may assume that the lines have equations a1 x + b1 y = 0 and a2 x + b2 y = 0 with
intersection the origin A ≡ [0, 0]. Let’s calculate the spread between these two lines.

A point B on l1 is B ≡ [−b1 , a1 ]. An arbitrary point on l2 is of the form

C ≡ [−λb2 , λa2 ]. The quadrances of the triangle ABC are then

Q (A, B) = b21 + a21 (1.2)

¡ ¢
Q (A, C) = λ2 b22 + a22 (1.3)
Q (B, C) = (b1 − λb2 )2 + (λa2 − a1 )2 . (1.4)

Pythagoras’ theorem asserts that the triangle ABC has a right vertex at C precisely
when
Q (A, C) + Q (B, C) = Q (A, B) .
This is the equation
¡ ¢ 2 2
λ2 b22 + a22 + (b1 − λb2 ) + (λa2 − a1 ) = b21 + a21 .

After expansion and simplification you get

¡ ¡ ¢¢
2λ a1 a2 + b1 b2 − λ a22 + b22 = 0.

Thus λ = 0 or
a1 a2 + b1 b2
λ= . (1.5)
a22 + b22
But λ = 0 precisely when the lines are perpendicular, in which case (1.5) still holds. So
substitute (1.5) back into (1.4) and use some algebraic manipulation to show that

(a1 b2 − a2 b1 )2
Q (B, C) = .
a22 + b22
8 1. OVERVIEW

Thus

2
Q (B, C) (a1 b2 − a2 b1 )
s (l1 , l2 ) = = 2 .
Q (A, B) (a1 + b21 ) (a22 + b22 )

Now it should be clearer why the idea of a spread as a ratio of quadrances is a good
one–the final answer is a rational expression in the coordinates of the two lines. If we
had chosen to define the spread as a ratio of lengths, then square roots would appear
at this point. Square roots are to be avoided whenever possible.

Calculating a spread

To calculate the spread s1 in the triangle A1 A2 A3 with side lengths |A2 , A3 | ≡ 4,

|A1 , A3 | ≡ 7 and |A1 , A2 | ≡ 5, let B be the foot of the altitude from A2 to the line
A1 A3 , and let a ≡ |A1 , B| and b ≡ |A2 , B| as in Figure 1.4.

7-a A3
B
a b 4

A1
5 A2

Figure 1.4: Calculating a spread

In the right triangles A1 A2 B and A3 A2 B, Pythagoras’ theorem gives

a2 + b2 = 25 and (7 − a)2 + b2 = 16.
Take the difference between these equations to get
14a − 49 = 9
so that
a = 58/14 = 29/7.
Then
2
b2 = 25 − a2 = 25 − (29/7) = 384/49
and so
Q (A2 , B) 384/49 384
s1 = = = .
Q (A1 , A2 ) 25 1225

Exercise 1.1 Verify also that s2 = 24/25 and s3 = 24/49. ¦

1.2. LAWS OF RATIONAL TRIGONOMETRY 9

1.2 Laws of rational trigonometry

This section introduces the main concepts and basic laws of rational trigonometry. So
far the discussion has been in the decimal number plane. The algebraic nature of
quadrance and spread have the important consequence that the basic definitions and
laws may be formulated to hold with coefficients from an arbitrary field. That involves
rethinking the above discussion so that no mention is made of distance or angle.

The laws given below are polynomial, and can be derived rigorously using only algebra
and arithmetic, in a way that high school students can follow. This is shown in Part II
of the book. If you work with geometrical configurations whose points belong to a
particular field, then the solutions to these equations also belong to this field.

So these laws are valid in great generality, and it is now possible to investigate
Euclidean geometry over general fields. This is the beginning of universal geometry.
The following definitions thus hold in a general field.

A point A is an ordered pair [x, y] of numbers. The quadrance Q (A1 , A2 ) between

points A1 ≡ [x1 , y1 ] and A2 ≡ [x2 , y2 ] is the number

Q (A1 , A2 ) ≡ (x2 − x1 )2 +(y2 − y1 )2 .

A line l is an ordered proportion ha : b : ci, representing the equation ax + by + c = 0.

The spread s (l1 , l2 ) between lines l1 ≡ ha1 : b1 : c1 i and l2 ≡ ha2 : b2 : c2 i is the
number

2
(a1 b2 − a2 b1 )
s (l1 , l2 ) ≡ .
(a21 + b21 ) (a22 + b22 )

For distinct points A1 and A2 the unique line passing through them both is denoted
A1 A2 . Given three distinct points A1 , A2 and A3 , define
the quadrances

Q1 ≡ Q (A2 , A3 ) A2
Q2 ≡ Q (A1 , A3 )
Q3 s2
Q3 ≡ Q (A1 , A2 )
Q1
and the spreads A1 s 3 A3
s1
s1 ≡ s (A1 A2 , A1 A3 ) Q2
s2 ≡ s (A2 A1 , A2 A3 )
s3 ≡ s (A3 A1 , A3 A2 ) .

Here are the five main laws of rational trigonometry.

10 1. OVERVIEW

Triple quad formula The three points A1 , A2 and A3 are collinear (meaning they all
lie on a single line) precisely when
2 ¡ ¢
(Q1 + Q2 + Q3 ) = 2 Q21 + Q22 + Q23 .

Pythagoras’ theorem The lines A1 A3 and A2 A3 are perpendicular precisely when

Q1 + Q2 = Q3 .

Spread law For any triangle A1 A2 A3 with non-zero quadrances

s1 s2 s3
= = .
Q1 Q2 Q3

Cross law For any triangle A1 A2 A3 define the cross c3 ≡ 1 − s3 . Then

(Q1 + Q2 − Q3 )2 = 4Q1 Q2 c3 .

Triple spread formula For any triangle A1 A2 A3

¡ ¢
(s1 + s2 + s3 )2 = 2 s21 + s22 + s23 + 4s1 s2 s3 .

These formulas are related in interesting ways, and deriving them is reasonably
straightforward. There are also numerous alternative formulations of these laws, as
well as generalizations to four quadrances and spreads, which become important in the
study of quadrilaterals. The difference between the two sides of the Triple quad
formula is used to define the quadrea A of a triangle, which is the single most
important number associated to a triangle. There is also an oriented version of the
theory, with signed areas of triangles and turns and coturns of oriented vertices.

You also need to understand the spread polynomials. If l1 , l2

and l3 are lines making spreads s (l1 , l2 ) = s (l2 , l3 ) ≡ s, then
l3
the Triple spread formula shows that the spread r ≡ s (l1 , l3 )
is either 0, in which case l1 and l3 are parallel, or 4s (1 − s).
l2
The polynomial function S2 (s) ≡ 4s (1 − s) figures
prominently in chaos theory, where it is called the logistic map. r ss
l1
Generalizing this, Sn (s) is the spread made by n successive spreads
of s, which can be formulated in terms of repeated reflections of lines in lines. It turns
out that the spread polynomial Sn (s) is of degree n in s, and in the decimal number
field is closely related to the classical Chebyshev polynomial Tn (x).

The remarkable properties of these polynomials extend to general fields, suggesting

new directions for the theory of special functions. There are also analogous cross
polynomials. These are the basic ingredients for rational trigonometry.
1.3. WHY CLASSICAL TRIGONOMETRY IS HARD 11

1.3 Why classical trigonometry is hard

For centuries students have struggled to master angles, trigonometric functions and
their many intricate relations. Those who learn how to apply the formulas correctly
often don’t know why they are true. Such difficulties are to an extent the natural
reflection of an underlying ambiguity at the heart of classical trigonometry. This
manifests itself in a number of ways, but can be boiled down to the single critical
question:

What precisely is an angle??

The problem is that defining an angle correctly requires calculus. This is a point
implicit in Archimedes’ derivation of the length of the circumference of a circle, using
an infinite sequence of successively refined approximations with regular polygons. It is
also supported by the fact that The Elements [Euclid] does not try to measure angles,
with the exception of right angles and some related special cases. Further evidence can
be found in the universal reluctance of traditional texts to spell out a clear definition of
this supposedly ‘basic’ concept.

Exercise 1.2 Open up a few elementary and advanced geometry books and see if this
claim holds. ¦

Let’s clarify the point with a simple example. The rectangle ABCD in Figure 1.5 has
side lengths |A, B| = 2 and |B, C| = 1. What is the angle θ between the lines AB and
AC in degrees to four decimal places? (Such accuracy is sometimes needed in
surveying.)

D C

q
A 2 B

Figure 1.5: A simple problem?

Without tables, a calculator or calculus, a student has difficulty in answering this

question, because the usual definition of an angle (page 5) is not precise enough to
show how to calculate it. But how can one claim understanding of a mathematical
concept without being able to compute it in simple situations? If the notion of an
angle θ cannot be made completely clear from the beginning, it cannot be fundamental.
12 1. OVERVIEW

Of course there is one feature about angles that is useful–they transform the
essentially nonlinear aspect of a circle into something linear, so that for example the
sum of adjacent angles can be computed using only addition, at least if the angles are
small. However the price to be paid for this convenience is high.

One difficulty concerns units. Most scientists and engineers work with degrees, the
ancient Babylonian scale ranging from 0◦ to 360◦ around a full circle, although in
Europe the grad is also used–from 0 to 400. Mathematicians officially prefer radian
measure, in which the full circle gets the angle of 6. 283 185 . . . , also known as 2π. The
radian system is theoretically convenient, but practically a nuisance. Whatever scale is
chosen, there is then the ambiguity of adding ‘multiples of a circle’ to any given angle,
and of deciding whether angles are to have signs, or to be always positive.

Angles in the range from 90◦ to 180◦ are awkward and students invariably have
difficulty with them. For example the angle 156◦ is closely related to the angle 24◦ ,
and a question involving the former can be easily converted into a related question
about the latter. There is an unnecessary duplication of information in having separate
measurements for an angle and its supplement. For angles in the range from 180◦ to
360◦ similar remarks apply. Indeed, such angles are rarely necessary in elementary
geometry.

Another consequence of relying on angles is a surprisingly limited range of

configurations that can be analysed in classical trigonometry using elementary
techniques. Students are constantly given examples that deal essentially with
90◦ /60◦ /30◦ or 90◦ /45◦ /45◦ triangles, since these are largely the only ones for which
they can make unassisted calculations.

Small wonder that the trigonometric functions cos θ, sin θ and tan θ and their inverse
functions cause students such difficulties. Although pictures of unit circles and ratios
of lengths are used to ‘define’ these in elementary courses, it is difficult to understand
them correctly without calculus. For example, the function tan θ is given, for θ in a
suitable range, by the infinite series
1 2 17 7 62 9
tan θ ≈ θ + θ3 + θ5 + θ + θ + ···
3 15 315 2835
while the inverse function arccos x is given by
1 1 3 5 7 35 9
arccos x ≈ π − x − x3 − x5 − x − x + ···
2 6 40 112 1152
and in the case of tan θ the coefficients defy a simple closed-form expression. This kind
of subtlety is largely hidden from beginning students with the reliance on the power of
calculators.

If the foundations of a building are askew, the entire structure is compromised. The
underlying ambiguities in classical trigonometry revolving around the concept of angle
are an impediment to learning mathematics, weaken its logical integrity, and introduce
an unnecessary element of approximation and inaccuracy into practical applications of
the subject.
1.4. WHY RATIONAL TRIGONOMETRY IS EASIER 13

1.4 Why rational trigonometry is easier

The key concepts of rational trigonometry are simpler, and mathematically more
natural, than those of classical trigonometry. Quadrance is easier to work with than
distance (as most mathematicians already know) and a spread is more elementary than
an angle. The spread between two lines is a dimensionless quantity, and in the rational
or decimal number fields takes on values between 0 and 1, with 0 occurring when lines
are parallel and 1 occurring when lines are perpendicular. Forty-five degrees becomes a
spread of 1/2, while thirty and sixty degrees become respectively spreads of 1/4 and
3/4. What could be simpler than that?

Another advantage with spreads is that the measurement is taken between lines, not
rays. As a consequence, the two range of angles from 0◦ to 90◦ and from 90◦ to 180◦
are treated symmetrically. For example, the spreads associated to 24◦ and 156◦ are
identical (namely s ≈ 0.165 434 . . .). If one wishes to distinguish between these two
situations in the context of rays, a single binary bit of additional information is
required, namely the choice between acute (ac) and obtuse (ob). The Triangle spread
rules described in the applications part of the book deal with this additional
information.

There are many triangles that can be analysed completely by elementary means using
rational trigonometry, giving students exposure to a wider range of examples.

The straightforwardness of rational trigonometry is also evident from the polynomial

form of the basic laws, which do not involve any transcendental functions, rely only on
arithmetical operations, and are generally quadratic in any one variable. As a
consequence, tables of values of trigonometric functions, or modern calculators, are not
necessary to do trigonometric calculations. Computations for simple problems can be
done by hand, more complicated problems can use computers more efficiently.

With the introduction of rational polar and spherical coordinates in calculus, this
simplicity can be put to work in solving a wide variety of sophisticated problems.
Computations of volumes, centroids of mass, moments of inertia and surface areas of
spheres, paraboloids and hyperboloids become in many cases more elementary. This
simplicity extends to higher dimensional spaces, where the basic algebraic relations
reduce the traditional reliance on pictures and argument by analogy with lower
dimensions.

Rational trigonometry works over any field. So the difficulties inherent in the decimal
and ‘real number’ fields can be avoided. It is not necessary to have a prior model of
the continuum before one begins geometry. Furthermore many calculations become
much simpler over finite fields, which can be a significant advantage.

Even when extended to spherical trigonometry and hyperbolic geometry, the theory’s
basic formulas are polynomial, as will be shown in a future volume. The concepts of
rational trigonometry also have advantages when working statistically, but this point
will not be developed here.
14 1. OVERVIEW

1.5 Comparison example

Here is an illustration of how rational trigonometry allows you to solve explicit

practical problems, and why this method is superior to the classical approach. There is
nothing particularly special about this example or the numbers appearing in it.

Problem 1 The triangle A1 A2 A3 has side lengths |A1 , A2 | ≡ 5, |A2 , A3 | ≡ 4 and

|A3 , A1 | ≡ 6. The point B is on the line A1 A3 with the angle between A1 A2 and A2 B
equal to 45◦ . What is the length d ≡ |A2 , B|?

A3
6
B 4
b d
A1 a 45 A2
5

Figure 1.6: Classical version

Classical solution

Let the angles at A1 and B be α and β respectively, as shown in Figure 1.6. The
Cosine law in the triangle A1 A2 A3 gives
42 = 52 + 62 − 2 × 5 × 6 × cos α
so that using a calculator
3
α = arccos ≈ 41. 4096◦ .
4
Since the sum of the angles in A1 A2 B is 180◦ ,
β ≈ 180◦ − 45◦ − 41. 4096◦ ≈ 93. 5904◦ .
Now the Sine law in A1 A2 B states that
sin α sin β
=
d 5
so that again using the calculator
5 sin 41. 4096◦
d≈ ≈ 3. 3137.
sin 93. 5904◦
1.5. COMPARISON EXAMPLE 15

Rational solution

To apply rational trigonometry, first convert the initial information about lengths and
angles into quadrances and spreads. The three quadrances of the triangle are the
squares of the side lengths, so that Q1 = 16, Q2 = 36 and Q3 = 25.

The spread corresponding to the angle 45◦ is 1/2, as can be easily seen from Figure 1.7.

B
2
1

A 1
2 C
1

Figure 1.7: Spread of 1/2

Let s be the spread between the lines A1 A2 and A1 A3 , and let r be the spread between
the lines BA1 and BA2 . Let Q ≡ Q (A2 , B). This yields Figure 1.8 involving only
quantities from rational trigonometry.

A3
36
B
r 16
Q

A1 s 1 A2
2

25

Figure 1.8: Rational version

Let’s follow the use of the basic laws to first find s, then r, and then Q. Use the Cross
law in A1 A2 A3 to get

(25 + 36 − 16)2 = 4 × 25 × 36 × (1 − s)

so that s = 7/16.

Use the Triple spread formula in A1 A2 B to obtain for r the quadratic equation
µ ¶2 µ ¶
7 1 49 1 2 7 1
+ +r =2 + +r +4× × × r.
16 2 256 4 16 2
16 1. OVERVIEW

This simplifies to
1
r2 − r + =0
256
so that
1 3√
r= ± 7.
2 16
For each of these values of r, use the Spread law in A1 A2 B
r s
=
25 Q
and solve for Q, giving values
√
Q1 = 1400 − 525 7
√
Q2 = 1400 + 525 7.
To convert these answers back into distances, take square roots
p
d1 = Q1 ≈ 3. 3137 . . .
p
d2 = Q2 ≈ 264.056 . . . .
Clearly the answer is d1 . Where does d2 come from? The initial information (apart
from the picture) describes two different possibilities. The second one is that the line
A2 B makes a spread of 1/2 with A1 A2 as shown in Figure 1.9, with the intersection B
off the page in the direction shown, since s = 7/16 is less than 1/2.

A3
B
36
16
Q2
A1 7/16 ½
25 A2

Figure 1.9: Alternate possibility

Comparison

Clearly the solution

√ using rational trigonometry is more accurate, and reveals that the
irrational number 7 is intimately bound up in this problem, which is certainly not
obvious from the initial data, nor apparent from the classical solution.

Perhaps it appears that the classical solution is however shorter. That is because the
main computations of arccos 3/4, sin α and sin β were done by the calculator.
1.6. ANCIENT GREEK TRIUMPHS AND DIFFICULTIES 17

The rational solution can be found entirely by hand. To be fair, it should be added
that classical trigonometry is capable of getting the exact answers obtained by the
rational method, but a more sophisticated approach is needed. One can simulate
rational trigonometry from within classical trigonometry–by never calculating angles!

The rational method has solved two problems at the same time. This phenomenon is
closely related to the fact that spreads, unlike angles, do not distinguish between the
concepts of acute and obtuse, although in the Applications part of the book these
concepts do appear in the context of the Triangle spread rules.

Since rational trigonometry is so much simpler than the existing theory, why was it not
discovered a long time ago? In fact bits and pieces of rational trigonometry have
surfaced throughout the history of mathematics, and have also been used by physicists.

It is well known, for example, that many trigonometric identities and integrals can be
easily verified and solved by appropriate rational substitutions and/or use of the
complex exponential function. The ubiquitous role of squared quantities in Euclidean
geometry is also familiar, as is the insight of linear algebra to avoid angles as much as
possible by using dot products and cross products. Modern geometry is very aware of
the central importance of quadratic forms. In Einstein’s theory of relativity the
quadratic ‘interval’ plays a crucial role.

Perhaps these different clues have not been put together formerly because of the
strength of established tradition, and in particular the reverence for Greek geometry.

1.6 Ancient Greek triumphs and difficulties

The serious study of geometry began with the ancient Greeks, whose mathematical
heritage is one of the most remarkable and sublime achievements of humanity.
Nevertheless, they encountered some difficulties, some of which persist today and
interfere with our understanding of geometry.

A Pythagorean dilemma

The Pythagoreans believed that the workings of the universe could ultimately be
described by proportions between whole numbers, such as 4 : 5. This was not necessarily
a mistake–the future may reveal the utility of such an idea. After early successes in
applying this principle to music and geometry, however, they discovered a now famous
dilemma.
√ The length of the diagonal to the length of the side of a square, namely
2 : 1, could not be realized as a proportion between two whole numbers.

Unfortunately, this difficulty was something of a red herring. Had they stuck with
their original beliefs in the workings of the Divine Mind, and boldly concluded that the
squares of the lengths ought to be more important than the lengths themselves, then
the history of mathematics would look quite different.
18 1. OVERVIEW

Circles and trigonometric functions

Points, lines and circles are the most important objects in Euclidean geometry. The
ancient Greeks were much enamoured with the elegance of the circle, and chose to give
the compass equal importance to the straightedge in their geometry. Following their
lead, mathematics courses still rely heavily on the circle and its properties to introduce
the basic concepts of classical trigonometry. Most high school students will be familiar
with the unit circle used to define both angles, as on page 5, as well as the
trigonometric ratios cos θ, sin θ and tan θ.

This thinking is unfortunate. A circle is a more complicated object than a line. In

Cartesian coordinates it is described by a second degree equation, while a line is
described by a first degree equation. Defining arc lengths of curves other than line
segments is quite sophisticated, even for arcs of circles. So it does not make
mathematical sense to treat circles on a par with lines, or to attempt to use circles to
define the basic measurement between lines.

In rational trigonometry points and lines are more fundamental than circles. The
essential formulas involve only points and lines, while circles appear later as particular
kinds of conics in universal geometry, where they can be studied purely in the rational
context without the use of trigonometric functions. Even rotations can be investigated
solely within universal geometry.

The essential roles of cos θ, sin θ and tan θ arise rather when considering uniform
motion around a circle, which has more to do with mechanics than geometry.
Trigonometric functions also occur in harmonic analysis and complex analysis, where
they are however secondary to the complex exponential function, whose theory is
simpler and more fundamental.

And what about all those engineering problems with spherical or cylindrical
symmetry? Rational trigonometry solves many of these problems in a more direct
fashion, as shown by the last chapter in this book. The reality is that the
trigonometric functions are overrated, and their intensive study in high schools is for
many students an unnecessary complication.

Euclid’s ambiguities

To develop a general and powerful geometry, it is critical to ensure that the basic
definitions are completely unambiguous, and that circular reasoning plays no role.

Unfortunately, it is too easy to loosely convey a geometric idea with a picture.

Students can seemingly learn concepts–such as parallel and perpendicular lines,
triangles, quadrilaterals, tangent lines to a circle, angles and many others–by being
shown a sufficient number of illustrative examples. However, when they want to use
these ideas in proofs or precise calculations, the lack of clear definitions becomes a
burden, not a convenience.
1.7. MODERN AMBIGUITIES 19

One of the appeals of mathematics is that one can understand completely the basic
concepts and then logically derive the entire theory, step by step, using only valid
logical arguments. When students meet mathematics in which the logical structure is
not clearly evident, their confidence and understanding suffers. This applies to all
students, even–perhaps especially–to weaker ones.

The extraordinary work of Euclid, The Elements, which set the tone of mathematics
for two thousand years, is guilty of ambiguity in its initial definitions, despite the
otherwise generally high standard of logic and precision throughout. The historical
dominance of Euclid’s work, with its supposedly axiomatic framework, has coloured
mathematical thinking to this day, not altogether in a positive way.

According to Euclid, a point is that which ‘has no parts or magnitude’, a line is

‘length without breadth’, and perpendicularity of lines is defined as ‘lying evenly
between each other’. For someone who already has an intuitive idea of what a point,
line and perpendicular mean, perhaps by having been shown many relevant pictures
from an early age, these definitions may seem reasonable, but to someone without prior
experience of geometry they are surely quite unintelligible.

For many centuries it appeared that there was no alternative to the geometry of
Euclid. But in the seventeenth century the work of Fermat, Descartes and others
opened up a dramatic new possibility. The Cartesian idea of defining a point as an
ordered pair of numbers, and then analysing curves by algebraic manipulations of their
defining equations is very powerful. It reduces the foundations of geometry to that of
arithmetic and algebra, and allows the formulation of many geometrical questions over
general fields.

Rational trigonometry adopts the logical and precise Cartesian approach of Descartes,
but it adds a metrical aspect with the notions of quadrance and spread, still in the
purely algebraic setting. This allows Euclidean geometry to be put on a firm and
proper foundation.

1.7 Modern ambiguities

In the nineteenth century some of the difficulties and limitations with Euclid’s
geometry became more recognized, and for this reason and others, the twentieth
century saw a slow but steady decline in the importance of Euclid as the basis of
education in geometry. Unfortunately, there was no successful attempt to put into
place an alternative framework.

One of those who realized the importance of the task was D. Hilbert, whose
Foundations of Geometry [Hilbert] was an attempt to shore up the logical deficiencies
of Euclid. This approach was not universally accepted, and the thought arose that
perhaps the task was intrinsically impossible, and that a certain amount of ambiguity
could not be removed from the subject.
20 1. OVERVIEW

Twentieth century mathematics largely resigned itself to an acceptance of vagueness in

the foundations of mathematics. It became accepted wisdom that any treatment of
Euclidean geometry had to incorporate the philosophical aspects of the ‘construction
of the continuum’ and Cantor’s theory of ‘infinite sets’. Distance, angle and the
trigonometric functions required the usual framework of modern analysis, along with
its logical deficiencies.

A common feeling was that the best strategy was to adopt an axiomatic system,
involving formal manipulations of underlying concepts that are not given meaning. It
was often supposed that this is what Euclid had in mind.

But this is not what Euclid had in mind. When he stated his ‘axioms’, Euclid wanted
only to clarify which facts he was going to regard as obvious, before deriving all other
facts using deduction. The meaning of these basic facts was never in question. Euclid’s
work in this sense is quite different from formalist systems which came to be modelled
on it.

When we begin a study of universal geometry, these preconceived twentieth century

notions must be put aside. It is possible to start from the beginning, and to aspire to
give a complete and precise account of geometry, without any missing gaps. The
logical framework should be clearly visible at all times. The basic concepts should be
easy to state to someone beginning the subject. Pulling in theorems from the outside
is not allowed. The ‘continuum’ need not be understood. ‘Infinite set theory’ and its
attendant logical difficulties–deliberately swept under the carpet by modern
mathematics–should be avoided.

With such an approach the logical coherence and beauty of mathematics become easier
to verify and appreciate directly, without appeals to authority and convention. To
build up mathematics properly, axioms are not necessary. You do not have to engage
in philosophy to do geometry, nor require sophisticated modern logic to understand
foundations. Clear thinking, careful definitions and an interest in applications should
suffice.

So let’s begin the story, starting with an informal review of the basic assumed
knowledge from algebra and arithmetic. Then let’s build up trigonometry and
geometry, one step at a time.
 2

Background

2.1 Fields

A field is a number system involving the operations of addition and multiplication

obeying particular laws. It is possible to read this book without any knowledge of the
theory of fields by working only with rational numbers–expressions of the form a/b
where a and b are integers with b non-zero, with the convention that a1 /b1 = a2 /b2
precisely when a1 b2 − a2 b1 = 0, and with the usual arithmetical operations.

In this book, the term field refers not to a set, but rather to the specification of a type
of number. Thus for example we do not need the ‘infinite set’ of ‘all’ rational numbers
in order to specify what a rational number is, and to define addition and multiplication
between two of them. This understanding fits naturally with common usage outside of
pure mathematical circles, and avoids metaphysical speculations.

A field F contains numbers 0, 1 and −1. By defining 2 ≡ 1 + 1, −2 ≡ (−1) + (−1),

3 ≡ 1 + 1 + 1 and so on, the integer a is associated to an element in F. Thus also any
rational number expression a/b has a meaning in F, provided that b 6= 0. This
convention allows polynomials with integer coefficients to be defined over any field.

The characteristic of a field F is the least natural number n ≥ 1 such that n = 0 in

F , if it exists. This book adopts the convention that characteristic two fields are
excluded, since many formulas have denominators involving 2 or 4. Characteristic three
fields are occasionally problematic.

For simplicity the main fields used as examples in this book are the field of rational
numbers, the field of integers modulo an odd prime p, denoted Fp , the field of decimal
numbers, and the field of complex numbers. Of course other fields may be used too.

There are unfortunately serious logical difficulties with the decimal number system,
and even more remarkable confusions with the field of so-called ‘real numbers’.

21
22 2. BACKGROUND

The reader is warned that the views expressed on the following two pages represent the
author’s position, which is currently a minority view.

The field of decimal numbers

When the term decimal number is used in this book, it refers to a number such as
π 2 /16 ≈ 0.616 850 275 . . .
with a decimal expansion that is specified by an algorithm,
√ computer program or
function. All the constants of mathematics, such as 2, π, e, γ and so on, are decimal
numbers in this sense, as are any arithmetical expressions formed by them, even those
using infinite sequences and series, provided of course that these sequences and series
are specified in a finite way.

Arithmetic with decimal numbers is thus intimately connected with the theory of
computation. Unfortunately it is difficult in practice, and perhaps impossible in theory,
to consistently determine when two algorithms generate the same decimal number.

This is a very serious deficiency, and implies that there is no effective notion of equality
in the theory. So there is no general procedure to tell whether a given arithmetical
statement involving decimal numbers is correct. Consider for example the formula
µ ¶2 µ ¶
1 1 1 1 3 1 1 1
− + − + ··· = + + + ···
1 2 3 4 8 12 22 32
which is essentially a statement of equality between two algorithms. How could one
check the validity of this equation without recognizing that the decimal number
represented by both sides is π 2 /16?

The topic of decimal numbers is a source of considerable confusion and ambiguity in

mathematics. A proper development of the subject, which might have been a major
agenda item for twentieth century mathematics, has yet to be taken sufficiently
seriously. It is too difficult to be attempted here.

As an illustration of the problems that arise, it is generally accepted lore that the
decimal numbers are countable and not complete. However with a careful examination
of the definitions involved, it turns out that the reverse is true–the field of decimal
numbers is both complete and not countable!
√
Despite such confusions, it is convenient to refer to numbers such as 2, π, e and so on,
temporarily putting aside the conceptual difficulties. So the decimal number field will
be flagged as an ‘informal’ field, awaiting a proper treatment.

A complex number will here be of the form a + bi, with a and b decimal numbers, and
the usual operations based on i2 = −1. Algebraic numbers have decimal expansions
given by algorithms (such as Newton’s method), so they fit into this framework. The
fundamental theorem of algebra holds in the field of complex numbers, meaning that
any polynomial with complex coefficients has a zero.
2.1. FIELDS 23

The field of ‘real numbers’

Even more logically unsound than the field of decimal numbers is the field of so-called
‘real numbers’. Believers in ‘real numbers’, which currently includes the majority of
mathematicians, assert that the decimal numbers of the previous section should be
called computable decimal numbers, to be distinguished from ‘non-computable decimal
numbers’. These latter ‘numbers’, which play the role of Leprechauns in modern
mathematics, supposedly have infinite decimal expansions which are not determined
by an algorithm, formula or computer program. In other words, these are decimal
expansions which by definition cannot be described explicitly by finite beings such as
ourselves, or our computers.

There are consequently no known examples of ‘non-computable decimal numbers’.

Actual mathematical or scientific computations never involve ‘non-computable decimal
numbers’. Defining addition and multiplication of ‘non-computable decimal numbers’
is necessarily even more ambiguous than for computable decimal numbers (see
[Wildberger]), and not surprisingly standard texts uniformly ignore the issue.

Nonetheless, according to the adherents, the ‘real number field’–comprising both the
computable decimal numbers together with the ‘non-computable decimal
numbers’–provides the correct model of the continuum, and forms the basis of
geometry, as well as much of mathematics. This is a curious position, to put it mildly!

In this book, ‘non-computable decimals’ play no role. Readers who are adamant about
their existence may replace, in their minds, any reference to the field of decimal
numbers with the field of ‘real numbers’. No computational outcomes are in the least
affected.

The field Fp

For p an odd prime, the field Fp will be considered to consist of expressions of the form
a/b, where a and b are integers with b not divisible by p, with the convention of equality
a1 /b1 = a2 /b2
precisely when a1 b2 − a2 b1 is divisible by p. In practice that means that a/b is
unchanged if a multiple of p is added or subtracted from either a or b. It is then a
standard fact that any such number is equal to one of the form a/1 ≡ a, for a one of
the p numbers 0, 1, 2, · · · , p − 1. Arithmetic now follows the usual rules for ordinary
fractions, and so the usual congruence notation becomes unnecessary.

Example 2.1 Here are some illustrations of arithmetic in F7 .

3+5=8=1
3 × 6 = 18 = 4
34 19 6 5 38 3
+ = + = = = 2. ¦
53 17 4 3 12 5
24 2. BACKGROUND

Square and spread numbers

Fix a field F . The term number refers to an element of F.

A number a in F is a square number (or just a square) precisely when it is of the

form a = r2 for some number r in F . Clearly 0 and 1 are square numbers in any field.
Whether or not −1 is a square number is an interesting question which strongly
influences the behavior of arithmetic in a field, and has consequences for geometry too.

Example 2.2 In the field Fp with p elements, there are exactly (p + 1) /2 distinct
square numbers. Here is a list for small values of p.

p Squares in Fp
3 0, 1
5 0, 1, 4
7 0, 1, 2, 4
11 0, 1, 3, 4, 5, 9
13 0, 1, 3, 4, 9, 10, 12
17 0, 1, 2, 4, 8, 9, 13, 15, 16
19 0, 1, 4, 5, 6, 7, 9, 11, 16, 17

Note that in F11 , the number −1 = 10 is not a square, while in F13 , the number
−1 = 12 is a square. ¦

In the field Fp , if p is a prime of the form p = 4l + 1 then −1 is a square, while if

p = 4l + 3 then the number −1 is not a square.

Definition A number s in a field F is a spread number precisely when s (1 − s) is

a square. Clearly 0 and 1 are spread numbers in any field.

Example 2.3 In the decimal number field the square numbers are precisely the
positive numbers, and the spread numbers are precisely those s satisfying 0 ≤ s ≤ 1. ¦

Example 2.4 In the field Fp there are either (p + 1) /2 or (p + 3) /2 spread numbers,

depending on whether p is of the form 4l + 1 or 4l + 3 respectively. Here is a list for
small values of p.
p Spread numbers in Fp
3 0, 1, 2
5 0, 1, 3
7 0, 1, 3, 4, 5
11 0, 1, 2, 3, 6, 9, 10
13 0, 1, 4, 6, 7, 8, 10
17 0, 1, 2, 6, 7, 9, 11, 12, 16
19 0, 1, 2, 4, 8, 9, 10, 11, 12, 16, 18 ¦
2.2. PROPORTIONS 25
¡ ¢
Exercise 2.1 Show that for any number u with u2 6= −1, the number s ≡ u2 / 1 + u2
is a spread number, and that every spread number s 6= 1 is of this form. ¦

Exercise 2.2 Show that for any numbers x and y satisfying x2 6= −1 and y 2 6= −1,
the number
2
(x − y)
s≡
(1 + x2 ) (1 + y 2 )
¡ ¢
is a spread number. If xy 6= −1 then find a number u so that s = u2 / 1 + u2 . ¦

Definition A number s in a field F is a square-spread number precisely when it

is both a square number and a spread number.

Example 2.5 Here is a list of square-spread numbers in Fp for small values of p.

p Square-spread numbers in Fp
3 0, 1
5 0, 1, 4
7 0, 1, 4
11 0, 1, 3, 9
13 0, 1, 4, 10
17 0, 1, 2, 9, 16
19 0, 1, 4, 9, 11, 16 ¦

Exercise 2.3 (Harder) How many square-spread numbers are there in Fp ? ¦

2.2 Proportions

The ancient Greeks believed that the relative sizes of two objects was a more
fundamental notion than the absolute size of either of them. This idea has largely gone
out of fashion in mathematical circles, but there is something to be said for it, as
appreciated by modern physicists.

Definition A 2-proportion is an expression of the form a : b where a and b are

numbers, not both zero, with the convention that

a : b = λa : λb

for any non-zero number λ. If the context is clear, a 2-proportion will be called
simply a proportion.

The convention for equality may be restated as

a1 : b1 = a2 : b2 ⇔ a1 b2 − a2 b1 = 0.
26 2. BACKGROUND

Example 2.6 In the rational number field

12
3:7= :4
7
since
12
3×4− × 7 = 0. ¦
7

One difference between fractions and proportions is rather important for geometry.
The fraction a/b makes sense only if b is non-zero, while the proportion a : b is valid as
long as one of a or b is non-zero.

Another difference between fractions and proportions is that the concept of proportion
can be extended to three or more numbers. This is quite natural when dealing with
the relative sizes of more than two objects.

Definition A 3-proportion is an expression of the form a : b : c where a, b and c

are numbers, not all zero, with the convention that

a : b : c = λa : λb : λc

for any non-zero number λ. If the context is clear, a 3-proportion will be called
simply a proportion.

The convention of equality may be restated as

a1 : b1 : c1 = a2 : b2 : c2

precisely when

a1 b2 − a2 b1 = 0 b1 c2 − b2 c1 = 0 c1 a2 − c2 a1 = 0.

Exercise 2.4 Show that in general one has to check all three conditions to ensure
that two 3-proportions are equal. ¦

Definition For a natural number n, an n-proportion is an expression of the form

a1 : a2 : · · · : an

where a1 , a2 , · · · , an are not all zero, with the convention that

a1 : a2 : · · · : an = λa1 : λa2 : · · · : λan

for any non-zero number λ.

Note that there are (n − 1) n/2 corresponding conditions for equality between two
n-proportions.
2.3. IDENTITIES AND DETERMINANTS 27

2.3 Identities and determinants

The rational approach to geometry inevitably involves polynomial and rational

identities, some of which become rather involved.

Example 2.7 The familiar Binomial theorem gives the expansion of (x1 + x2 )n for
n a natural number and x1 , x2 elements of a field, using the integers
µ ¶
n n (n − 1) · · · (n − k + 1)
≡ .
k k (k − 1) · · · 1
The statement is
µ ¶ µ ¶
n n n−1 n
(x1 + x2 ) = xn1 + x x2 + · · · + x1 xn−1 + xn2 . ¦
1 1 n−1 2

Example 2.8 Another useful identity is the following special case of the extended
Binomial theorem for three variables
2
(x1 + x2 + x3 ) = x21 + x22 + x23 + 2 (x1 x2 + x1 x3 + x2 x3 ) .
Both sides are symmetric functions of the variables x1 , x2 and x3 , meaning that if
any two of the indices are interchanged, the expression remains the same. ¦

Example 2.9 The most important identity in geometry, and also possibly in number
theory, is Fibonacci’s identity
¡ ¢¡ ¢
(x1 y2 − x2 y1 )2 + (x1 x2 + y1 y2 )2 = x21 + y12 x22 + y22 .

It was probably known to Diophantus, but the first recorded proof is in Fibonacci’s
Liber quadratorum of 1225. ¦

Example 2.10 An extension of Fibonacci’s identity is Cauchy’s identity

(x1 y2 − x2 y1 )2 + (y1 z2 − y2 z1 )2 + (z1 x2 − z2 x1 )2 + (x1 x2 + y1 y2 + z1 z2 )2
¡ 2 ¢¡ ¢
= x1 + y12 + z12 x22 + y22 + z22
which plays a major role in projective trigonometry. ¦

Exercise 2.5 Check the Fibonacci and Cauchy identities, as well as the following
variant of Fibonacci’s identity
¡ ¢¡ ¢
(x1 y2 + x2 y1 )2 + (x1 x2 − y1 y2 )2 = x21 + y12 x22 + y22 . ¦

Exercise 2.6 Check the following identity, which occurs in the proof of the Spread
ratio theorem (page 77)
(y1 − y2 ) (x3 − x1 ) − (y1 − y3 ) (x2 − x1 ) = (y1 − y3 ) (x3 − x2 ) − (y2 − y3 ) (x3 − x1 ) . ¦
28 2. BACKGROUND

Zero denominator convention

In this book the following Zero denominator convention is used.

A statement involving a rational identity will be assumed to be empty when a choice

of variables creates a denominator of zero.

For example the rational identity

(x1 y2 − x2 y1 )2 (x1 x2 + y1 y2 )2
+ =1
(x21 + y12 ) (x22 + y22 ) (x21 + y12 ) (x22 + y22 )

which follows from Fibonacci’s identity in the previous section will be considered a
true statement for any values of the variables, even those that create a situation where
one of the factors x21 + y12 or x22 + y22 is zero, since in this case the identity is considered
to be an empty statement, so still true. This removes the need to prohibit special
cases, but means that the reader must be particularly careful, when faced with a
rational expression, to ask

When is the denominator zero?

In addition, some care is needed when such identities form part of an extended logical
argument–one must remember the conditions that resulted in zero denominators, even
after the denominators may have disappeared.

Determinants and anti-symmetric polynomials

Determinants are particularly useful polynomial expressions which occur frequently in

geometry. This section introduces convenient notation which extends the idea of a
determinant to create more general anti-symmetric polynomials.

Example 2.11 A common determinant is

¯ ¯
¯x1 y1 z1 ¯
¯ ¯
¯x2 y2 z2 ¯ = x1 y2 z3 − x1 y3 z2 + x2 y3 z1 − x3 y2 z1 + x3 y1 z2 − x2 y1 z3 .
¯ ¯
¯x3 y3 z3 ¯

The six terms in this expression are obtained from the first by performing all six
permutations of the indices i = 1, 2 and 3, and multiplying by the sign of the
permutation. Note the convention that the position of the variables in the terms do not
move, only the indices do. The above determinant will also be written as [x1 y2 z3 ]−
3 . ¦
2.3. IDENTITIES AND DETERMINANTS 29

Example 2.12 A special case of the previous is

¯ ¯
¯x1 y1 1¯
¯ ¯
¯x2 y2 1¯ = x1 y2 − x1 y3 + x2 y3 − x3 y2 + x3 y1 − x2 y1 .
¯ ¯
¯x3 y3 1¯

This expression will also be written as [x1 y2 ]−

3 . ¦

More generally, for any natural numbers m1 , m2 and m3 define the anti-symmetric
polynomial
¯ m1 ¯
¯x y1m2 z1m3 ¯¯
¯ 11
[xm1 m2
1 y2 z3 ]3 ≡ ¯¯xm
m3 −
2 y2m2 z2m3 ¯¯ .
¯xm1 y3m2 z3m3 ¯
3

Then
m3 −
[xm1 m2
1 y2 z3 ]3 = xm1 m2 m3 m1 m2 m3 m1 m2 m3
1 y2 z3 − x1 y3 z2 + x2 y3 z1
−xm 1 m2 m3 m1 m2 m3 m1 m2 m3
3 y2 z1 + x3 y1 z2 − x2 y1 z3

is obtained by taking the alternating sum of monomials, starting with xm 1 m2

1 y2 z3 ,
m3

and successively performing interchanges on the (sub)indices 1, 2 and 3, in the order

2 ←→ 3 1 ←→ 2 2 ←→ 3 1 ←→ 2 2 ←→ 3.
More generally, for any sequence of variables
x1 , x2 , x3 , y1 , y2 , y3 , z1 , z2 , z3 , w1 , w2 , w3 , · · ·
with (sub)indices in the range {1, 2, 3}, and any polynomial p in these variables, define
−
[p]3 to be the alternating sum of the six terms obtained from p by applying these same
interchanges to the indices, in this same order.

Example 2.13
£ ¤−
x1 x52 y2 3 = x1 x52 y2 − x1 x53 y3 + x2 x53 y3 − x3 x52 y2 + x3 x51 y1 − x2 x51 y1 . ¦

Example 2.14
[x1 ]−
3 = x1 − x1 + x2 − x3 + x3 − x2 = 0. ¦

Exercise 2.7 Prove the identity

£ 2 ¤− − −
x1 y2 3 + [x1 x2 y2 ]3 = (x1 + x2 + x3 ) [x1 y2 ]3 . ¦

This notational device can be extended to larger index sets. For example, for a
polynomial p involving terms with indices in {1, 2, 3, 4}, define [p]−
4 to be the
alternating sum of the 24 terms obtained from p by applying all possible permutations
of the four indices, with each term multiplied by the sign of the corresponding
permutation.
30 2. BACKGROUND

Example 2.15
£ 2 ¤−
x1 x2 y3 4 = x21 x2 y3 − x21 x2 y4 + x21 x3 y4 − x21 x4 y3 + x21 x4 y2 − x21 x3 y2
+x22 x3 y1 − x22 x4 y1 + x23 x4 y1 − x23 x4 y2 + x22 x4 y3 − x22 x3 y4
+x23 x2 y4 − x24 x2 y3 + x24 x3 y2 − x24 x3 y1 + x24 x2 y1 − x23 x2 y1
+x23 x1 y2 − x24 x1 y2 + x24 x1 y3 − x23 x1 y4 + x22 x1 y4 − x22 x1 y3 . ¦

2.4 Linear equations

The following facts should be familiar over the rational or decimal numbers, and indeed
they hold in any field. Consider two linear equations in variables x and y of the form
a1 x + b1 y + c1 = 0
a2 x + b2 y + c2 = 0
where a1 and b1 are not both zero, and also a2 and b2 are not both zero. A unique
solution [x, y] exists precisely when
¯ ¯
¯a1 b1 ¯
¯
a1 b2 − a2 b1 = ¯ ¯ 6= 0.
a2 b2 ¯
In this case
¯ ¯ ¯ ¯
b1 c2 − b2 c1 ¯c1 b1 ¯ ¯a1 b1 ¯
x= ¯
= −¯ ¯ / ¯ ¯
a1 b2 − a2 b1 c2 b2 ¯ ¯a2 b2 ¯
¯ ¯ ¯ ¯
c1 a2 − c2 a1 ¯a1 c1 ¯ ¯a1 b1 ¯
y= ¯
= −¯ ¯ / ¯ ¯.
a1 b2 − a2 b1 a2 c2 ¯ ¯a2 b2 ¯

On the other hand suppose that a1 b2 − a2 b1 = 0, so that a1 : b1 = a2 : b2 . If

a1 : b1 : c1 = a2 : b2 : c2
then there is more than one solution, while if
a1 : b1 : c1 6= a2 : b2 : c2
then there is no solution.

Another common situation is the case of three homogeneous equations in variables x, y

and z
a1 x + b1 y + c1 z = 0
a2 x + b2 y + c2 z = 0
a3 x + b3 y + c3 z = 0.
There is always the solution [x, y, z] = [0, 0, 0]. Another solution exists precisely when
¯ ¯
¯a1 b1 c1 ¯
¯ ¯
¯a2 b2 c2 ¯ = 0.
¯ ¯
¯a3 b3 c3 ¯
2.5. POLYNOMIAL FUNCTIONS AND ZEROES 31

2.5 Polynomial functions and zeroes

Polynomial functions with integer coefficients have a meaning in any field. Such a
polynomial p (x) has a zero at the number x0 precisely when p (x0 ) = 0. For example
the spread polynomial
¡ ¢2
S5 (x) ≡ x 16x2 − 20x + 5 (2.1)
turns out to control five-fold symmetry in any field, and its zeroes relate to the
existence of regular pentagons.

Example 2.16 Over the decimal numbers the function S5 (s) may be (partially)
visualized in the usual way by plotting some of the values [x, S5 (x)] as in Figure 2.1.

1.4

1.2

0.8

0.6

0.4

0.2

-0.5 0.5 1
-0.2

-0.4

¡ ¢2
Figure 2.1: Spread polynomial S5 (x) ≡ x 16x2 − 20x + 5 ¦

In a finite field a polynomial may be conveniently represented by a table of values.

¡ ¢2
Example 2.17 In F19 the polynomial S5 (x) ≡ x 16x2 − 20x + 5 has values
x 0 1 2 3 4 5 6 7 8 9
S5 (x) 0 1 10 13 1 5 17 6 10 0
x 10 11 12 13 14 15 16 17 18
. ¦
S5 (x) 10 1 10 14 3 15 0 7 10

Recall the important Factor theorem, which asserts that if a polynomial function
p (x) has a zero at x0 , then there is a natural number k less than or equal to the degree
of p such that
k
p (x) = (x − x0 ) q (x)
for some polynomial q (x) with q (x0 ) 6= 0. The number k is the multiplicity of the
zero x0 .

Example 2.18 In F19 the polynomial S5 (x) has zeroes 0, 9 and 16, the former with
multiplicity one and the latter two with multiplicity two. In fact
S5 (x) = 9x (x − 9)2 (x − 16)2 . ¦
32 2. BACKGROUND

2.6 Quadratic equations

Geometry is a quadratic subject, and so is intimately connected with the subject of

quadratic equations. This section reviews the familiar procedure for solving a
quadratic equation, along with an important result on pairs of quadratic equations.
Recall that we are working in a given field F , not of characteristic two.

Completing the square

A quadratic equation in x is a polynomial equation of degree two in x of the form

ax2 + bx + c = 0
where a and b are numbers with a 6= 0. The normal form of a quadratic equation is
(x − p)2 = q
for some numbers p and q. The quadratic equation ax2 + bx + c = 0 can always be
rewritten in normal form by the process of completing the square, as follows.

Step 1. Take c to the other side, and divide both sides by the non-zero number a
b c
x2 + x = − .
a a

Step 2. Add the square of one half the coefficient of x to both sides

b b2 b2 c
x2 + x + 2 = 2 − .
a 4a 4a a

Step 3. Factor the left side and simplify the right to obtain
µ ¶2
b b2 − 4ac
x+ = .
2a 4a2

This has a solution in x precisely when b2 − 4ac is a square, say r2 , in which case
−b ± r
x= .
2a

Note that the usual ‘quadratic formula’ is subsumed by this derivation. Over a general
field square roots do not always exist, and if they do they are generally not unique.
The usual situation with the decimal numbers exploits the existence of a distinguished
(positive) square root of a positive number, and is a good example of ‘symmetry
breaking’ in mathematics.
2.6. QUADRATIC EQUATIONS 33

Pairs of quadratic equations

Definition Two quadratic equations

(x − p1 )2 = q1 (x − p2 )2 = q2

in x are compatible precisely when they have a common solution.

Here is our first theorem, and although entirely algebraic, it anticipates aspects of
rational trigonometry and universal geometry.

Theorem 1 (Quadratic compatability) The quadratic equations

2 2
(x − p1 ) = q1 (x − p2 ) = q2

are compatible precisely when

³ ´2
(p1 − p2 )2 − (q1 + q2 ) = 4q1 q2 .

In this case, if p1 6= p2 then there is a unique common solution

p1 + p2 (q1 − q2 )
x= − .
2 2 (p1 − p2 )

Proof. If p1 = p2 then the two equations

(x − p1 )2 = q1 (x − p2 )2 = q2

are compatible precisely when q1 = q2 , which is equivalent to the condition that

2
(q1 + q2 ) = 4q1 q2 .

Suppose otherwise that p1 6= p2 . If the two equations are compatible with common
solution x, then take the difference between the equations to obtain

2 (p1 − p2 ) x = p21 − p22 − q1 + q2

so that
p1 + p2 (q1 − q2 )
x= − . (2.2)
2 2 (p1 − p2 )
Since this x satisfies the second equation,
µ ¶2
p1 + p2 (q1 − q2 )
− − p2 = q2 .
2 2 (p1 − p2 )
34 2. BACKGROUND

Now deduce the following equivalent equations

³ ´2
2 2
(p1 − p2 ) − (q1 − q2 ) = 4 (p1 − p2 ) q1
(p1 − p2 )4 − 2 (p1 − p2 )2 (q1 − q2 ) + (q1 − q2 )2 = 4 (p1 − p2 )2 q1
4 2 2
(p1 − p2 ) + 2 (p1 − p2 ) (q1 + q2 ) + (q1 − q2 ) = 0
³ ´2
(p1 − p2 )2 − (q1 + q2 ) = 4q1 q2 .

Conversely if
³ ´2
2
(p1 − p2 ) − (q1 + q2 ) = 4q1 q2 .
then retrace the steps and use the symmetry between the two original equations to
show that (2.2) is a common solution.

Note that if p1 = p2 in the theorem, then the two quadratic equations are compatible
precisely when they are identical, and so in this case there are two common solutions.

Example 2.19 In the rational number field the quadratic equations

(x − 4)2 = 9 (x − 6)2 = 1
are compatible since
³ ´2
2
(4 − 6) − (9 + 1) = 36 = 4 × 9 × 1.

As 4 6= 6, the unique common solution is

4+6 (9 − 1)
x= − = 7. ¦
2 2 (4 − 6)

Exercise 2.8 Show that the quadratic equations

a1 x2 + b1 x + c1 = 0
a2 x2 + b2 x + c2 = 0
are compatible precisely when
¯ ¯2 ¯ ¯¯ ¯
¯a1 c1 ¯¯ ¯a b1 ¯¯ ¯¯b1 c1 ¯¯
¯ = ¯¯ 1
¯a2 c2 ¯ a2 b2 ¯ ¯b2 c2 ¯
in which case
a2 c1 − a1 c2
x= . ¦
a1 b2 − a2 b1
Exercise 2.9 Show that the equations
2 2
(x − p1 ) = q (x − p2 ) = q
are compatible precisely when either p1 = p2 or p1 − p2 = ±2q. In the latter case show
that the unique common solution is
p1 + p2
x= . ¦
2
 3

Cartesian coordinate geometry

This foundational chapter introduces careful and novel definitions of the basic building
blocks of planar Cartesian geometry, and establishes fundamental facts that will be
used repeatedly throughout the rest of the book. This includes notions of points, lines,
triangles, quadrilaterals, parallel, perpendicular, affine combinations and reflections.
Throughout we work in, or over, a fixed field F, not of characteristic two, whose
elements are called numbers. Diagrams generally illustrate the situation in the rational
and decimal number fields, or occasionally in a finite field.

3.1 Points and lines

Definition A point A ≡ [x, y] is an ordered pair of numbers. The numbers x and y

are the coordinates of A.

Example 3.1 Figure 3.1 shows the points [2, 3] and [5, 2] in the rational or decimal
number fields.

4 [2,3]
3 [5,2]
2
1

-2 -1 0 1 2 3 4 5 6

Figure 3.1: Two points over the rational or decimal fields ¦

35
36 3. CARTESIAN COORDINATE GEOMETRY

Over the field Fp , a point may be represented by a box in a square array of boxes, with
horizontal and vertical positions numbered from 0 to p − 1. Alternative labellings will
also be used; for example over F11 the horizontal and vertical positions could be
labelled with the numbers −5, −4, · · · , 4, 5. Sometimes such a diagram will be repeated
periodically in both directions to illustrate a larger view.

Example 3.2 Figure 3.2 shows the points [3, 7] and [6, 1] in the field F11 .

10
9
[3,7] 8
7
6
5
4
3
2 [6,1]
1
0
0 1 2 3 4 5 6 7 8 9 10

Figure 3.2: Two points over F11 ¦

Definition A line l ≡ ha : b : ci is a 3-proportion, enclosed in pointed brackets, with

the property that a and b are not both zero.

The alternative notation ¿ À

a : b
l≡
c
is also useful, as it clarifies the distinguished role played by c, while ha : b : ci
anticipates the development of projective rational trigonometry, where all three
coefficients are treated equally.

Definition The line l ≡ ha : b : ci is a null line precisely when

a2 + b2 = 0.

Otherwise l is a non-null line.

If the line l ≡ ha : b : ci is a null line then a and b must both be non-zero and
2
(b/a) = −1, so that −1 is a square.

Conversely if −1 is a square, say −1 = i2 for some number i, then there are only two
solutions to the equation x2 = −1, namely i and −i. Thus either b = ia or b = −ia,
and since a is non-zero the coefficients a, b and c of l can be divided by a to obtain
l = h1 : i : di or l = h1 : −i : di for some d.
3.1. POINTS AND LINES 37

Definition The point A ≡ [x, y] lies on the line l ≡ ha : b : ci, or equivalently the
line l passes through the point A, precisely when

ax + by + c = 0.

Note carefully that a line is not a set of points. This allows us to circumvent the
logical difficulties with ‘infinite set theory’. In diagrams a line is illustrated by
exhibiting the points, or some of the points, which lie on it.

A line l ≡ ha : b : ci is central or standard precisely when it passes through the

origin O ≡ [0, 0]. This is equivalent to the condition that c = 0.

Example 3.3 In the rational or decimal number fields, Figure 3.3 shows the line
h1 : 3 : −11i, which passes through the points [2, 3] and [5, 2].

4 [2,3]
3 [5,2]
2
1 1:3:-11
-2 -1 0 1 2 3 4 5 6

Figure 3.3: The line h1 : 3 : −11i ¦

Example 3.4 In the field F11 , Figure 3.4 shows the line h10 : 5 : 1i, which passes
through the points [3, 7] and [6, 1].

10
9
[3,7] 8
7
6
5
4
3
2 [6,1]
1
0
0 1 2 3 4 5 6 7 8 9 10

Figure 3.4: The line h10 : 5 : 1i in F11 ¦

38 3. CARTESIAN COORDINATE GEOMETRY

3.2 Collinear points and concurrent lines

Theorem 2 (Line through two points) For any two distinct points A1 and A2 ,
there is a unique line l ≡ A1 A2 passing through them both. If A1 ≡ [x1 , y1 ] and
A2 ≡ [x2 , y2 ] then

l = A1 A2 = hy1 − y2 : x2 − x1 : x1 y2 − x2 y1 i .

Proof. If A1 ≡ [x1 , y1 ] and A2 ≡ [x2 , y2 ] are distinct points, then

l ≡ hy1 − y2 : x2 − x1 : x1 y2 − x2 y1 i

passes through them both, since

(y1 − y2 ) x1 + (x2 − x1 ) y1 + x1 y2 − x2 y1 = 0
(y1 − y2 ) x2 + (x2 − x1 ) y2 + x1 y2 − x2 y1 = 0.

On the other hand if m ≡ ha : b : ci is a line passing through both A1 and A2 then

ax1 + by1 + c = 0
ax2 + by2 + c = 0.

Take the difference of these two equations to get

a (x2 − x1 ) + b (y2 − y1 ) = 0

so that
a : b = y1 − y2 : x2 − x1 .
Thus
m = hy1 − y2 : x2 − x1 : di
for some number d, and since A1 lies on m,

(y1 − y2 ) x1 + (x2 − x1 ) y1 + d = 0.

Conclude that
d = x1 y2 − x2 y1
so that
m = hy1 − y2 : x2 − x1 : x1 y2 − x2 y1 i .

Definition Three or more points which lie on a common line are collinear. Three
or more lines which pass through a common point are concurrent.
3.2. COLLINEAR POINTS AND CONCURRENT LINES 39

Theorem 3 (Collinear points) The points [x1 , y1 ], [x2 , y2 ] and [x3 , y3 ] are
collinear precisely when

x1 y2 − x1 y3 + x2 y3 − x3 y2 + x3 y1 − x2 y1 = 0.

Proof. If all three points are identical then the identity is automatic. Otherwise
assume by symmetry that [x1 , y1 ] 6= [x2 , y2 ]. From the previous theorem, [x3 , y3 ] lies on
the unique line passing through [x1 , y1 ] and [x2 , y2 ] precisely when
(y1 − y2 ) x3 + (x2 − x1 ) y3 + x1 y2 − x2 y1 = 0.
This is the required condition.

Recall that
¯ ¯
¯x1 y1 1¯¯
¯
x1 y2 − x1 y3 + x2 y3 − x3 y2 + x3 y1 − x2 y1 = ¯¯x2 y2 1¯¯ = [x1 y2 ]−
3 .
¯x3 y3 1¯

Theorem 4 (Concurrent lines) If the lines ha1 : b1 : c1 i, ha2 : b2 : c2 i and

ha3 : b3 : c3 i are concurrent then

a1 b2 c3 − a1 b3 c2 + a2 b3 c1 − a3 b2 c1 + a3 b1 c2 − a2 b1 c3 = 0.

Proof. If the three lines pass through the point [x, y] then [x, y, 1] is a non-zero
solution to the homogeneous linear system
a1 x + b1 y + c1 z = 0
a2 x + b2 y + c2 z = 0
a3 x + b3 y + c3 z = 0
so that by Section 2.4,
¯ ¯
¯a1 b1 c1 ¯¯
¯
¯a2 b2 c2 ¯¯ = a1 b2 c3 − a1 b3 c2 + a2 b3 c1 − a3 b2 c1 + a3 b1 c2 − a2 b1 c3 = 0.
¯
¯a3 b3 c3 ¯

Exercise 3.1 Show that the converse does not hold. ¦

Exercise 3.2 Show that in the field Fp there are exactly p + 1 lines passing through
every point, and that every line passes through exactly p points. ¦

Exercise 3.3 Show that if there are null lines, then there are exactly two null lines
passing through any given point. ¦
40 3. CARTESIAN COORDINATE GEOMETRY

3.3 Parallel and perpendicular lines

Definition The lines l1 ≡ ha1 : b1 : c1 i and l2 ≡ ha2 : b2 : c2 i are parallel precisely

when
a1 b2 − a2 b1 = 0
or equivalently when a1 : b1 = a2 : b2 .

Theorem 5 (Point on two lines) If the lines l1 and l2 are not parallel, then there
is a unique point A ≡ l1 l2 which lies on them both. If l1 ≡ ha1 : b1 : c1 i and
l2 ≡ ha2 : b2 : c2 i then
∙ ¸
b1 c2 − b2 c1 c1 a2 − c2 a1
A ≡ l1 l2 = , .
a1 b2 − a2 b1 a1 b2 − a2 b1

Proof. The point [x, y] lies on l1 ≡ ha1 : b1 : c1 i and l2 ≡ ha2 : b2 : c2 i precisely when
a1 x + b1 y + c1 = 0
a2 x + b2 y + c2 = 0.
Since l1 and l2 are not parallel, a1 b2 − a2 b1 6= 0, so as in Section 2.4 the unique
solution to this system of equations is
b1 c2 − b2 c1
x=
a1 b2 − a2 b1
c1 a2 − c2 a1
y= .
a1 b2 − a2 b1

The point A ≡ l1 l2 = l2 l1 is the intersection of the two lines l1 and l2 . Alternatively,

the lines l1 and l2 intersect at A ≡ l1 l2 .

Now comes the single most important definition in all of geometry. It colours the
entire subject. In a subsequent volume, the geometry presented here in this book,
based on this definition, will be called blue geometry, and other colours also appear.

Definition The lines l1 ≡ ha1 : b1 : c1 i and l2 ≡ ha2 : b2 : c2 i are perpendicular

precisely when
a1 a2 + b1 b2 = 0
or equivalently when a1 : b1 = −b2 : a2 .

Exercise 3.4 Show that if the lines l1 and l2 are parallel, with the line l3 parallel
(respectively perpendicular) to l1 , then l3 is parallel (respectively perpendicular) to l2 .
¦
3.4. PARALLELS AND ALTITUDES 41

Exercise 3.5 Show that if the lines l1 and l2 are perpendicular, and the lines l2 and l3
are perpendicular, then l1 and l3 are parallel. ¦

Exercise 3.6 Show that if −1 = i2 then any two null lines of the form h1 : i : di and
h1 : i : ei are both parallel and perpendicular, as are any two null lines of the form
h1 : −i : di and h1 : −i : ei. ¦

Exercise 3.7 Show that if two lines l1 and l2 are both parallel and perpendicular,
then they are both null lines. ¦

3.4 Parallels and altitudes

Theorem 6 (Parallel to a line) For any point A ≡ [x, y] and any line l ≡ ha : b : ci
there is a unique line k, called the parallel through A to l, which passes through A
and is parallel to l, namely
k = ha : b : −ax − byi .

Proof. If l ≡ ha : b : ci then any line k parallel to l must have the form k ≡ ha : b : di

for some number d, passing through A ≡ [x, y] precisely when
ax + by + d = 0.
Thus such a line k is uniquely determined, and has the form
k = ha : b : −ax − byi .

Theorem 7 (Altitude to a line) For any point A ≡ [x, y] and any line
l ≡ ha : b : ci there is a unique line n, called the altitude from A to l, which passes
through A and is perpendicular to l, namely

n = h−b : a : bx − ayi .

Proof. If l ≡ ha : b : ci then any line n perpendicular to l must have the form

n ≡ h−b : a : di for some number d, passing through A ≡ [x, y] precisely when
−bx + ay + d = 0
Thus such a line n is uniquely determined, and has the form
n = h−b : a : bx − ayi .
42 3. CARTESIAN COORDINATE GEOMETRY

Exercise 3.8 Show that for any points A ≡ [x, y], A1 ≡ [x1 , y1 ] and A2 ≡ [x2 , y2 ] with
A1 and A2 distinct, the parallel through A to A1 A2 is the line
k ≡ hy1 − y2 : x2 − x1 : (y2 − y1 ) x + (x1 − x2 ) yi . ¦

Exercise 3.9 Show that for any points A ≡ [x, y], A1 ≡ [x1 , y1 ] and A2 ≡ [x2 , y2 ] with
A1 and A2 distinct, the altitude from A to A1 A2 is the line
n ≡ hx1 − x2 : y1 − y2 : (x2 − x1 ) x + (y2 − y1 ) yi . ¦

Theorem 8 (Foot of an altitude) For any point A ≡ [x, y] and any non-null line
l ≡ ha : b : ci, the altitude n from A to l intersects l at the point
∙ 2 ¸
b x − aby − ac −abx + a2 y − bc
F ≡ , .
a2 + b2 a2 + b2

Proof. By the Altitude to a line theorem, the altitude

from A ≡ [x, y] to l ≡ ha : b : ci is n ≡ h−b : a : bx − ayi.
These two lines are not parallel since by assumption
a2 + b2 6= 0. By the Point on two lines theorem (page 40) A
∙ 2 ¸ n
b x − aby − ac −abx + a2 y − bc
F ≡ nl = , .
a2 + b2 a2 + b2
l F
The intersection F of the altitude n from A to l with the line l
is the foot of the altitude. Note that if l is a null line then the altitude n
from a point A to l is parallel to l, so there is no foot of the altitude.

Example 3.5 In the field F11 , the altitude from A ≡ [7, 6] to the line l ≡ h10 : 5 : 1i
(gray boxes) is n ≡ h9 : 4 : 1i (black circles) and the foot of the altitude is F ≡ [2, 9], as
shown in Figure 3.5.

10
9
8
7
6
5
4
3
2
1
0
0 1 2 3 4 5 6 7 8 9 10

Figure 3.5: Altitude to a line in F11 ¦

3.5. SIDES, VERTICES AND TRIANGLES 43

Example 3.6 In the field F13 , the altitude from A ≡ [5, 2] to the null line
l ≡ h3 : 2 : 1i (gray boxes) is the null line n ≡ h6 : 4 : 1i (black circles), as shown in
Figure 3.6. These two lines are both perpendicular and parallel.

12
11
10
9
8
7
6
5
4
3
2
1
0
0 1 2 3 4 5 6 7 8 9 10 11 12

Figure 3.6: Altitude to a null line in F13 ¦

Exercise 3.10 Show that the foot F ≡ [x0 , y0 ] of the altitude from A ≡ [x, y] to the
non-null line passing through [x1 , y1 ] and [x2 , y2 ] has coordinates
(x2 − x1 )2 x − (y1 − y2 ) (x2 − x1 ) y − (y1 − y2 ) (x1 y2 − x2 y1 )
x0 ≡ 2 2
(x2 − x1 ) + (y1 − y2 )
(y1 − y2 )2 y − (y1 − y2 ) (x2 − x1 ) x − (x2 − x1 ) (x1 y2 − x2 y1 )
y0 ≡ 2 2 . ¦
(x2 − x1 ) + (y1 − y2 )

3.5 Sides, vertices and triangles

The following definitions are more precise than usual.

Definition A side A1 A2 ≡ {A1 , A2 } is a set with A1 and A2 points. The line of

the side A1 A2 is the line A1 A2 . The side A1 A2 is a null side precisely when A1 A2 is
a null line.

Note that A1 and A2 are automatically distinct by the convention that repetitions are
not allowed in a set.

Two sides A1 A2 and A3 A4 are parallel (respectively perpendicular) precisely when

A1 A2 is parallel (respectively perpendicular) to A3 A4 .

Definition A vertex l1 l2 ≡ {l1 , l2 } is a set with l1 and l2 intersecting lines. The

point of the vertex l1 l2 is the point l1 l2 . The vertex l1 l2 is a null vertex precisely
when l1 or l2 is a null line, and is a right vertex precisely when l1 and l2 are
perpendicular.
44 3. CARTESIAN COORDINATE GEOMETRY

Definition A triangle A1 A2 A3 ≡ {A1 , A2 , A3 } is a set with A1 , A2 and A3

non-collinear points.

The points A1 , A2 and A3 are the points of the triangle A1 A2 A3 , and the lines
l1 ≡ A2 A3 , l2 ≡ A1 A3 and l3 ≡ A1 A2 are the lines of the triangle. The sides A1 A2 ,
A2 A3 and A1 A3 are the sides of the triangle, and the vertices l1 l2 , l2 l3 and l1 l3 are the
vertices of the triangle.

The point A1 is opposite the side A2 A3 , and so on, and the line l1 is opposite the
vertex l2 l3 , and so on. With this terminology a triangle has exactly three points,
necessarily non-collinear, and exactly three lines, necessarily non-concurrent. A triangle
also has exactly three sides, and exactly three vertices.

Definition A triangle A1 A2 A3 is a right triangle precisely when it has a right

vertex. A triangle A1 A2 A3 is a null triangle precisely when one or more of its lines
is a null line.

In the rational or decimal number fields, there are no null triangles, since there are no
null lines.

Example 3.7 In the field F13 the points A1 ≡ [2, 8], A2 ≡ [9, 9] and A3 ≡ [10, 0] form a
triangle A1 A2 A3 with lines l1 ≡ A2 A3 = h9 : 1 : 1i (gray boxes), l2 ≡ A1 A3 = h9 : 9 : 1i
(open circles) and l3 ≡ A1 A2 = h7 : 3 : 1i (solid circles) shown in Figure 3.7.

12
11
10
A2
A1 9
8
7
6
5
4
3
2
1 A3
0
0 1 2 3 4 5 6 7 8 9 10 11 12

Figure 3.7: A triangle in F13 ¦

Exercise 3.11 Show that a triangle has at most one right vertex. ¦

Exercise 3.12 Show that it is impossible for a triangle to have three null lines. ¦

Exercise 3.13 (Harder) How many triangles are there over Fp ? How many of these
are null triangles? ¦
3.6. QUADRILATERALS 45

3.6 Quadrilaterals

Definition A quadrilateral A1 A2 A3 A4 is a list [A1 , A2 , A3 , A4 ] of four distinct

points, no three of which are collinear, with the conventions that

A1 A2 A3 A4 ≡ A2 A3 A4 A1 and A1 A2 A3 A4 ≡ A4 A3 A2 A1 .

The points A1 , A2 , A3 and A4 are the points

of the quadrilateral A1 A2 A3 A4 , and the
lines l12 ≡ A1 A2 , l23 ≡ A2 A3 , l34 ≡ A3 A4 and l14 ≡ A1 A4 are the A3
lines of the quadrilateral. The lines l13 ≡ A1 A3 and l24 ≡ A2 A4 A2
are the diagonal lines (or just diagonals) of the quadrilateral.

The sides A1 A2 , A2 A3 , A3 A4 and A1 A4 are the sides of the A4

quadrilateral, and the vertices l12 l23 , l23 l34 , l34 l14 and l14 l12 are
A1
the vertices of the quadrilateral. The sides A1 A3 and A2 A4 are the diagonal sides of
the quadrilateral.

Two vertices which contain a common line are adjacent, otherwise vertices are
opposite. Two sides which contain a common point are adjacent, otherwise sides are
opposite.

Note that the notation for quadrilaterals is different than for triangles. These notions
generalize in an obvious way to defining n-gons A1 A2 · · · An , except that only any
three consecutive points are required to be non-collinear.

A parallelogram is a quadrilateral A1 A2 A3 A4 with the property that both pairs of

opposite sides are parallel, so that A1 A2 and A3 A4 are parallel, and A2 A3 and A1 A4
are parallel, as in Figure 3.8. A rectangle is a quadrilateral with the property that
any pair of adjacent sides are perpendicular. Every rectangle is a parallelogram.

A rhombus is a parallelogram with the property that the diagonals are perpendicular,
as in the second parallelogram of Figure 3.8.

A3 A3
A2 A2

A1 A4 A1 A4

Figure 3.8: Parallelogram and rhombus

Exercise 3.14 Give an example where the diagonals of a quadrilateral are parallel. ¦
46 3. CARTESIAN COORDINATE GEOMETRY

3.7 Affine combinations

Definition For any points A1 ≡ [x1 , y1 ] and A2 ≡ [x2 , y2 ] and any two numbers λ1
and λ2 satisfying λ1 + λ2 = 1, the affine combination λ1 A1 + λ2 A2 is the point

λ1 A1 + λ2 A2 ≡ [λ1 x1 + λ2 x2 , λ1 y1 + λ2 y2 ] .

Theorem 9 (Affine combination) Every point lying on the line A1 A2 is a unique

affine combination λ1 A1 + λ2 A2 for some numbers λ1 and λ2 with λ1 + λ2 = 1, and
conversely any affine combination of this form lies on A1 A2 .

Proof. Suppose that A1 ≡ [x1 , y1 ] and A2 ≡ [x2 , y2 ] are distinct, and that
A3 ≡ λ1 A1 + λ2 A2 for some numbers λ1 and λ2 satisfying λ1 + λ2 = 1. Then the
identity

(y1 − y2 ) (λ1 x1 + λ2 x2 ) + (x2 − x1 ) (λ1 y1 + λ2 y2 ) + x1 y2 − x2 y1

= (λ1 + λ2 − 1) (x2 y1 − x1 y2 )

together with the Line through two points theorem (page 38) shows that A3 lies on
A1 A2 .

Conversely suppose that A ≡ [x, y] lies on A1 A2 , so that

(y1 − y2 ) x + (x2 − x1 ) y + x1 y2 − x2 y1 = 0. (3.1)

Since A1 and A2 are distinct, one of (x2 − x1 ) and (y2 − y1 ) are non-zero. Suppose
without loss of generality that x2 − x1 6= 0. Then set
x − x2
λ≡
x1 − x2
so that
x = λx1 + (1 − λ) x2 .
Substitute this value of x into (3.1) and factor, to obtain

(y − λy1 − (1 − λ) y2 ) (x2 − x1 ) = 0.

Thus
y = λy1 + (1 − λ) y2
and so
A = λA1 + (1 − λ) A2 .
3.7. AFFINE COMBINATIONS 47

For the uniqueness, if

λA1 + (1 − λ) A2 = µA1 + (1 − µ) A2

then the difference between the left and right hand sides yields

(λ − µ) (x1 − x2 ) = 0
(λ − µ) (y1 − y2 ) = 0.

Since by assumption A1 6= A2 , you may conclude that λ = µ.

Example 3.8 Figure 3.9 shows the affine combination (1/3) A1 + (2/3) A2 for two
points A1 , A2 in the rational or decimal number fields.

1
3 A1 + 3 A2
2
A2

A1

Figure 3.9: An affine combination over the rationals ¦

Definition For distinct points A1 ≡ [x1 , y1 ] and A2 ≡ [x2 , y2 ], the point

∙ ¸
1 1 x1 + x2 y1 + y2
M ≡ A1 + A2 = ,
2 2 2 2

is the midpoint of the side A1 A2 .

In the rational or decimal number fields, the midpoint of a side

will often be illustrated by small bars (perhaps more than one). A1 M A2
Exercise 3.15 Suppose that the numbers α1 , α2 and α3 are
each distinct from 1, and satisfy α1 + α2 + α3 = 1. Show that for any three points
A1 , A2 and A3
µ ¶
α2 α3
α1 A1 + (1 − α1 ) A2 + A3
1 − α1 1 − α1
µ ¶
α3 α1
= α2 A2 + (1 − α2 ) A3 + A1
1 − α2 1 − α2
µ ¶
α1 α2
= α3 A3 + (1 − α3 ) A1 + A2
1 − α3 1 − α3

which allows an unambiguous meaning to the expression α1 A1 + α2 A2 + α3 A3 . ¦

48 3. CARTESIAN COORDINATE GEOMETRY

A median of a triangle A1 A2 A3 is a line m passing through a point of the triangle

and the midpoint of the opposite side.

Exercise 3.16 Show that if the field has characteristic three, then all the medians are
parallel, while otherwise they intersect at a point G, called the centroid of the
triangle, as in Figure 3.10. Furthermore if A1 ≡ [x1 , y1 ], A2 ≡ [x2 , y2 ] and A3 ≡ [x3 , y3 ]
then ∙ ¸
1 1 1 x1 + x2 + x3 y1 + y2 + y3
G = A1 + A2 + A3 = , .
3 3 3 3 3

A2

M3 M1
G

A1 M2 A3

Figure 3.10: Medians and centroid ¦

Theorem 10 (Thales’ theorem) Suppose that A1 , A2 and A3 are three distinct

points. For any number λ 6= 0 set

B2 ≡ (1 − λ) A1 + λA2
B3 ≡ (1 − λ) A1 + λA3 .

Then B2 B3 is parallel to A2 A3 .

Proof. Since A2 and A3 are distinct, so are B2 and B3 .

If A1 ≡ [x1 , y1 ], A2 ≡ [x2 , y2 ] and A3 ≡ [x3 , y3 ] then

B2 = [(1 − λ) x1 + λx2 , (1 − λ) y1 + λy2 ]

A3
B3 = [(1 − λ) x1 + λx3 , (1 − λ) y1 + λy3 ] . B3
Then by the Line through two points theorem (page 38)
A1
A2 A3 = hy2 − y3 : x3 − x2 : x2 y3 − x3 y2 i A2 B2
B2 B3 = hλ (y2 − y3 ) : λ (x3 − x2 ) : ci

for some number c. These two lines are parallel.

3.7. AFFINE COMBINATIONS 49

Theorem 11 (Parallelogram center) If A1 A2 A3 A4 is a parallelogram, then the

midpoints of the diagonal sides A1 A3 and A2 A4 coincide.

Proof. Suppose that A1 ≡ [x1 , y1 ], A2 ≡ [x2 , y2 ], A3 ≡ [x3 , y3 ] and A4 ≡ [x4 , y4 ]. Then

since

A1 A2 = hy1 − y2 : x2 − x1 : x1 y2 − x2 y1 i
A3 A4 = hy3 − y4 : x4 − x3 : x3 y4 − x4 y3 i

are parallel, it follows that

(y1 − y2 ) (x4 − x3 ) − (y3 − y4 ) (x2 − x1 ) = 0. (3.2)

Similarly, since A2 A3 and A1 A4 are parallel,

(y2 − y3 ) (x1 − x4 ) − (y4 − y1 ) (x3 − x2 ) = 0. (3.3)

Write (3.2) and (3.3) as a pair of linear equations in x4 and y4

(y1 − y2 ) x4 + (x2 − x1 ) y4 − (y1 − y2 ) x3 − (x2 − x1 ) y3 = 0 (3.4)

(y3 − y2 ) x4 + (x2 − x3 ) y4 + (x3 − x2 ) y1 + (y2 − y3 ) x1 = 0. (3.5)

The points A1 , A2 and A3 are not collinear, so by the Collinear points theorem (page
39) ¯ ¯
¯y1 − y2 x2 − x1 ¯
¯ ¯
¯y3 − y2 x2 − x3 ¯ = − (x1 y2 − x1 y3 + x2 y3 − x3 y2 + x3 y1 − x2 y1 ) 6= 0.

Solve the equations (3.4) and (3.5) to get

A2 A3
x4 = x1 − x2 + x3
y4 = y1 − y2 + y3

so that
1 1 1 1 A1 A4
A1 + A3 = A2 + A4 .
2 2 2 2

Note in particular that the two diagonals of a parallelogram are not parallel, as they
intersect at their common midpoint. This point is the center of the parallelogram.

Exercise 3.17 If A1 A2 A3 A4 is a parallelogram and M12 , M23 , M34 and M14 are the
midpoints of the sides A1 A2 , A2 A3 , A3 A4 and A1 A4 respectively, then show that
M12 M23 M34 M14 is a parallelogram, with the same center as A1 A2 A3 A4 . ¦
50 3. CARTESIAN COORDINATE GEOMETRY

3.8 Perpendicular bisectors

Definition The perpendicular bisector of the side A1 A2 is the altitude from the
midpoint M of A1 A2 to the line A1 A2 .

Theorem 12 (Perpendicular bisector) If A1 ≡ [x1 , y1 ] and A2 ≡ [x2 , y2 ] are

distinct points then the side A1 A2 has perpendicular bisector
¿ À
x2 − x21 + y22 − y12
p ≡ x1 − x2 : y1 − y2 : 2 .
2

Proof. Since
A1 A2 = hy1 − y2 : x2 − x1 : x1 y2 − x2 y1 i
any line perpendicular to A1 A2 has the form

p ≡ hx1 − x2 : y1 − y2 : ci
for some number c. Then p passes through the midpoint p
∙ ¸
x1 + x2 y1 + y2 A1
M≡ ,
2 2

of A1 A2 precisely when M A2

(x1 − x2 ) (x1 + x2 ) (y1 − y2 ) (y1 + y2 )

+ + c = 0.
2 2
Thus
x22 − x21 + y22 − y12
c=
2
and so ¿ À
x22 − x21 + y22 − y12
p = x1 − x2 : y1 − y2 : .
2
 4

Reflections

This chapter introduces affine transformations, and in particular the reflections in lines
and the associate lineations, which reflect lines in lines. These are the most important
symmetries in planar geometry. Formulas for reflections and lineations are derived.
The results are not all easy, and they are too often taken for granted in courses on
geometry. Beginners might skim the proofs of this chapter and return to them later.

4.1 Affine transformations

Definition A transformation is a function which inputs and outputs points.

Small Greek letters will generally be used for transformations. The effect of the
transformation τ on the point A is denoted τ (A). The transformation ι defined by
ι (A) ≡ A for any point A is the identity. The convention for composition is

(τ 2 ◦ τ 1 ) (A) ≡ (τ 2 τ 1 ) (A) ≡ τ 2 (τ 1 (A)) .

A transformation τ is invertible precisely when there is a transformation υ such that

τ υ = υτ = ι. Then υ ≡ τ −1 is the inverse of τ . A transformation τ fixes a point A
precisely when τ (A) = A.

Definition A transformation τ is affine precisely when for any points A1 and A2 ,

and any two numbers λ1 and λ2 satisfying λ1 + λ2 = 1,

τ (λ1 A1 + λ2 A2 ) = λ1 τ (A1 ) + λ2 τ (A2 ) .

51
52 4. REFLECTIONS

Definition For a point A define ρA , the rotation in A, to be the transformation

defined by
ρA (B) ≡ 2A + (−1) B.

Then ρA (B) lies on AB by the Affine combination theorem (page 46), and
A = (1/2) B + (1/2) ρA (B), so that A is the midpoint of BρA (B).

Exercise 4.1 Show that for any point A, the rotation ρA is an invertible affine
transformation, with ρ−1
A = ρA . ¦

Exercise 4.2 For any two points A1 and A2 define the successive rotations
A3 ≡ ρA2 (A1 ), A4 ≡ ρA3 (A2 ) and so on, as well as A0 ≡ ρA1 (A2 ), A−1 ≡ ρA0 (A1 )
and so on. Show that for any integers k and l, ρAk (Al ) = A2k−l .

A-1 A0 A1 A2 A3 A4

Figure 4.1: Succesive rotations in points ¦

Definition If l is a non-null line and F the foot of the altitude from a point A to l,
then the transformation that sends A to B ≡ ρF (A) is the reflection in the line l
and is denoted σ l .

Theorem 13 (Reflection of a point in a line) If l ≡ ha : b : ci is a non-null line

and A ≡ [x, y], then
"¡ ¢ ¡ ¢ #
b2 − a2 x − 2aby − 2ac −2abx + a2 − b2 y − 2bc
σ l (A) = , .
a2 + b2 a2 + b2

Proof. The foot F of the altitude from A ≡ [x, y] to l ≡ ha : b : ci is, by the Foot of an
altitude theorem (page 42),
∙ 2 ¸
b x − aby − ac −abx + a2 y − bc
F = , .
a2 + b2 a2 + b2
By definition σ l (A) ≡ σ F (A) ≡ 2F + (−1) A so that
"¡ ¢ ¡ ¢ #
b2 − a2 x − 2aby − 2ac −2abx + a2 − b2 y − 2bc
σ l (A) = , .
a2 + b2 a2 + b2
4.2. LINEATIONS AND REFLECTION SEQUENCES 53

Exercise 4.3 Show that the reflection σ l in a non-null line l is an invertible affine
transformation, with σ −1
l = σl . ¦

B
F
A
sl
n

Figure 4.2: Reflection of a point in a line ¦

4.2 Lineations and reflection sequences

Definition A lineation is a function which inputs and outputs lines.

If τ is an invertible affine transformation, then define the associated lineation Σ by

the rule
Σ (A1 A2 ) ≡ τ (A1 ) τ (A2 ) .
By the Affine combination theorem (page 46), this is well defined.

Theorem 14 (Rotation of a line in a point) For any point A ≡ [x, y], the
rotation ρA has associated lineation ΣA given by

ΣA (ha : b : ci) = ha : b : −2xa − 2yb − ci . ¦

Proof. Suppose that A1 ≡ [x1 , y1 ] and A2 ≡ [x2 , y2 ] are distinct points, so that by the
Line through two points theorem (page 38),
A1 A2 = hy1 − y2 : x2 − x1 : x1 y2 − x2 y1 i = ha : b : ci .
By the definition of ρA
ρA (A1 ) = 2A + (−1) A1 = [2x − x1 , 2y − y1 ]
ρA (A2 ) = 2A + (−1) A2 = [2x − x2 , 2y − y2 ] .
Then again by the Line through two points theorem,
ρA (A1 ) ρA (A2 ) = h−y1 + y2 : −x2 + x1 : (2x − x1 ) (2y − y2 ) − (2x − x2 ) (2y − y1 )i
= h−y1 + y2 : −x2 + x1 : 2x (y1 − y2 ) + 2y (x2 − x1 ) + x1 y2 − x2 y1 i
= h−a : −b : 2xa + 2yb + ci
= ha : b : −2xa − 2yb − ci .
54 4. REFLECTIONS

Theorem 15 (Reflection of a line in a line) For any non-null line l ≡ ha : b : ci,

the reflection σ l has associated lineation Σl given by
¿ ¡ 2 ¢ ¡ 2 ¢ À
a − b2 a1 + 2abb1 : 2aba
¡ 1− a¢ − b2 b1
Σl (ha1 : b1 : c1 i) = .
2aca1 + 2bcb1 − a2 + b2 c1

Proof. Suppose l ≡ ha : b : ci is non-null, so that a2 + b2 6= 0, and that A1 ≡ [x1 , y1 ]

and A2 ≡ [x2 , y2 ] are distinct, with
A1 A2 = hy1 − y2 : x2 − x1 : x1 y2 − x2 y1 i ≡ ha1 : b1 : c1 i .
Then by the Reflection of a point in a line theorem
"¡ ¢ ¡ ¢ #
b2 − a2 x1 − 2aby1 − 2ac −2abx1 + a2 − b2 y1 − 2bc
σ l (A1 ) = ,
a2 + b2 a2 + b2
"¡ ¢ ¡ ¢ #
b2 − a2 x2 − 2aby2 − 2ac −2abx2 + a2 − b2 y2 − 2bc
σ l (A2 ) = , .
a2 + b2 a2 + b2

Use the Line through two points theorem and simplify to see that
Σl (ha1 : b1 : c1 i) = σ l (A1 ) σ l (A2 )
¿ ¡ 2 ¢ ¡ 2 ¢ À
a − b2 a1 + 2abb1 : 2aba 2
¡ 21 − 2a¢ − b b1 .
=
2aca1 + 2bcb1 − a + b c1
The line l2 ≡ Σl (l1 ) is the reflection of l1 in the line l. The lineation Σl is the
reflection lineation in l.

l2
l1
Sl

Figure 4.3: Reflection of a line in a line

Exercise 4.4 Show that Σl (l1 ) = l1 precisely when l1 is perpendicular to l. ¦

Exercise 4.5 Show that Σ−1

l = Σl for any non-null line l. ¦

Exercise 4.6 Show that the formula for the reflection Σl in the previous theorem
makes sense even if l is a null line. Show that in this case Σl (l1 ) = l for any line l1 . ¦
4.2. LINEATIONS AND REFLECTION SEQUENCES 55

Theorem 16 (Reflection) If l1 and l2 are non-null lines, then

σ l2 σ l1 = σ Σl2 (l1 ) σ l2 .

l2
l1
sl2sl1(A) Sl2(l 1)

sl2(A) sl1(A)
A

Figure 4.4: Reflection theorem

Proof. Suppose that l1 ≡ ha1 : b1 : c1 i and l2 ≡ ha2 : b2 : c2 i are non-null lines and
A ≡ [x, y]. Then by the Reflection of a line in a line theorem, Σl2 (l1 ) ≡ l3 is
­¡ 2 ¢ ¡ ¢ ¡ ¢ ®
a2 − b22 a1 + 2a2 b2 b1 : 2a2 b2 a1 − a22 − b22 b1 : 2a2 c2 a1 + 2b2 c2 b1 − a22 + b22 c1 .

Now use the Reflection of a point in a line theorem to show that both
Ã" ¡ ¢ ¡ ¢ #!
b21 − a21 x − 2a1 b1 y − 2a1 c1 −2a1 b1 x + a21 − b21 y − 2b1 c1
σ l2 (σ l1 (A)) = σ l2 ,
a21 + b21 a21 + b21

and
Ã" ¡ ¢ ¡ ¢ #!
b22 − a22 x − 2a2 b2 y − 2a2 c2 −2a2 b2 x + a22 − b22 y − 2b2 c2
σ l3 (σ l2 (A)) = σ l3 ,
a22 + b22 a22 + b22

are equal to the point

∙ ¸
p q
,
(a21 + b21 ) (a22 + b22 ) (a21 + b21 ) (a22 + b22 )
where

p = (a1 a2 + b1 b2 + a2 b1 − a1 b2 ) (a1 a2 + b1 b2 − a2 b1 + a1 b2 ) x
¡ ¢ ¡ ¢
−2 (a1 b2 − a2 b1 ) (a1 a2 + b1 b2 ) y + 4a2 b1 b2 c1 + 2a1 a22 − b22 c1 − 2a2 a21 + b21 c2
q = (a1 a2 + b1 b2 + a2 b1 − a1 b2 ) (a1 a2 + b1 b2 − a2 b1 + a1 b2 ) y
¡ ¢ ¡ ¢
+2 (a1 b2 − a2 b1 ) (a1 a2 + b1 b2 ) x + 4a1 a2 b2 c1 − 2b1 a22 − b22 c1 − 2b2 a21 + b21 c2 .

Thus σ l2 (σl1 (A)) = σ Σl2 (l1 ) (σl2 (A)).

It is useful to note that the Reflection theorem can be restated in the form

σ Σl2 (l1 ) = σl2 σ l1 σ −1

l2 = σ l2 σ l1 σ l2 .
56 4. REFLECTIONS

Theorem 17 (Lineation) If l1 and l2 are non-null lines, then

Σl2 Σl1 = ΣΣl2 (l1 ) Σl2 .

Proof. Since the reflection Σl (m) of the line m in the line l is determined by the
reflections σ l (A1 ) and σ l (A2 ) of any two distinct points A1 and A2 lying on m, this is
a direct consequence of the previous theorem.

Definition A two-sided sequence of lines · · · , l−2 , l−1 , l0 , l1 , l2 , · · · is a reflection

sequence precisely when for any integer k

Σlk (lk−1 ) = lk+1 .

Theorem 18 (Cyclic reflection) Suppose that · · · , l−2 , l−1 , l0 , l1 , l2 , · · · is a

reflection sequence of lines. Then for any integers k and j

Σlk (lj ) = l2k−j . ¦

Proof. We will prove that

Σlk (lk+n ) = lk−n
for fixed k by induction on n = 0, 1, 2, · · · . For n = 0 and n = 1 the statement is clear,
since Σ−1
lk = Σlk . Assume it is true for all values up to n ≥ 1. Then since

lk+n+1 = Σlk+n (lk+n−1 )

the Lineation theorem shows that

¡ ¢
Σlk (lk+n+1 ) = Σlk Σlk+n (lk+n−1 )
= ΣΣlk (lk+n ) (Σlk (lk+n−1 ))
= Σlk−n (lk−n+1 )
= lk−n−1 = lk−(n+1) .

So the statement is also true for n + 1. So it is true for any positive n, and so true for
any integer n, since the following statements are equivalent

Σlk (lk+n ) = lk−n and Σlk (lk−n ) = lk+n .

Now set k + n = j to get

Σlk (lj ) = l2k−j .
 Part II

Rational trigonometry

57
 5

Quadrance

This chapter begins with the concept of the quadrance between points, and then
examines null lines, midpoints, the Triple quad formula, Pythagoras’ theorem, the
quadrea of a triangle, and a generalization of the classical Heron’s formula called–in
this text–Archimedes’ formula. Perpendicular bisectors and quadrance from a point
to a line are discussed. Archimedes’ function and the Quadruple quad formula are
defined. All these topics hold in an arbitrary field F not of characteristic two (from
now on this will not necessarily be repeated).

5.1 Quadrances of triangles and quadrilaterals

Definition The quadrance Q (A1 , A2 ) between the points A1 ≡ [x1 , y1 ] and

A2 ≡ [x2 , y2 ] is the number
2 2
Q (A1 , A2 ) ≡ (x2 − x1 ) + (y2 − y1 ) .

Sometimes Q (A1 , A2 ) will be called the quadrance from A1 to A2 , or the quadrance

of the side A1 A2 . Clearly

Q (A1 , A2 ) = Q (A2 , A1 ) .

In the rational or decimal number fields Q (A1 , A2 ) is always positive, and is zero
precisely when A1 = A2 . This is not necessarily the case for other fields.

Example 5.1 In the complex number field with A1 ≡ [0, 0] and A2 ≡ [1, i]

Q (A1 , A2 ) = 12 + i2 = 0. ¦

59
60 5. QUADRANCE

Theorem 19 (Null line) If A1 and A2 are distinct points, then A1 A2 is a null line
precisely when Q (A1 , A2 ) = 0.

Proof. If A1 ≡ [x1 , y1 ] and A2 ≡ [x2 , y2 ] are distinct points, then by the Line through
two points theorem (page 38)
A1 A2 = hy1 − y2 : x2 − x1 : x1 y2 − x2 y1 i .
This is a null line precisely when
2 2
(y1 − y2 ) + (x2 − x1 ) = 0
which is exactly the condition that Q (A1 , A2 ) = 0.

Theorem 20 (Midpoint) If A1 A2 is a non-null line, then there is a unique point A

lying on A1 A2 which satisfies

Q (A1 , A) = Q (A, A2 ) .

This is the midpoint M ≡ (1/2) A1 + (1/2) A2 of the side A1 A2 . Furthermore

Q (A1 , M ) = Q (M, A2 ) = Q (A1 , A2 ) /4.

Proof. If A1 ≡ [x1 , y1 ] and A2 ≡ [x2 , y2 ] are distinct points, then by the Affine
combination theorem (page 46) any point on the line A1 A2 has the form
A ≡ λA1 + (1 − λ) A2 = [λx1 + (1 − λ) x2 , λy1 + (1 − λ) y2 ]
for some number λ. The condition Q (A1 , A) = Q (A, A2 ) is then
2 2
((λ − 1) x1 + (1 − λ) x2 ) + ((λ − 1) y1 + (1 − λ) y2 )
= (λx1 + (−λ) x2 )2 + (λy1 + (−λ) y2 )2 .
Rewrite this as ³ ´³ ´
(1 − λ)2 − λ2 (x2 − x1 )2 + (y2 − y1 )2 = 0.
By assumption A1 A2 is a non-null line, so by the previous Null line theorem,
2 2
(x2 − x1 ) + (y2 − y1 ) 6= 0. Thus λ = 1/2, and A is the midpoint
M ≡ (1/2) A1 + (1/2) A2 = [(x1 + x2 ) /2, (y1 + y2 ) /2]
of the side A1 A2 . Then
µ ¶2 µ ¶2
x1 − x2 y1 − y2
Q (A1 , M ) = + = Q (M, A2 )
2 2
= Q (A1 , A2 ) /4.
5.1. QUADRANCES OF TRIANGLES AND QUADRILATERALS 61

Exercise 5.1 (Harder) Show that if l1 ≡ ha1 : b1 : c1 i, l2 ≡ ha2 : b2 : c2 i and

l3 ≡ ha3 : b3 : c3 i then
¡ 2 ¢
a1 + b21 (a1 b2 c3 − a1 b3 c2 + a2 b3 c1 − a3 b2 c1 + a3 b1 c2 − a2 b1 c3 )2
Q (l1 l2 , l1 l3 ) = . ¦
(a1 b2 − a2 b1 )2 (a1 b3 − a3 b1 )2

Definition For a triangle A1 A2 A3 , the numbers Q1 ≡ Q (A2 , A3 ), Q2 ≡ Q (A1 , A3 )

and Q3 ≡ Q (A1 , A2 ) are the quadrances of the triangle, with Q1 the quadrance of
the side A2 A3 , or the quadrance opposite the vertex l2 l3 , and similarly for the
other quadrances.

A2

Q3
Q1
A1 A3
Q2

Figure 5.1: Quadrances of a triangle

This convention will generally be followed for A1 , A2 and A3 any three points, even if
they are collinear.

Definition For a quadrilateral A1 A2 A3 A4 , the numbers Q12 ≡ Q (A1 , A2 ),

Q23 ≡ Q (A2 , A3 ), Q34 ≡ Q (A3 , A4 ) and Q14 ≡ Q (A1 , A4 ) are the quadrances of
the quadrilateral. The numbers Q13 ≡ Q (A1 , A3 ) and Q24 ≡ Q (A2 , A4 ) are the
diagonal quadrances of the quadrilateral.

With this notation a quadrilateral has four quadrances and two diagonal quadrances.

A2 Q 23
A3

Q 12 Q 34

A1 Q 14 A4

Figure 5.2: Quadrances of a quadrilateral

62 5. QUADRANCE

Exercise 5.2 Show that if the characteristic of the field is not three, then the centroid
G of a triangle A1 A2 A3 with midpoints M1 , M2 and M3 of the sides satisfies
4Q (A1 , M1 )
Q (A1 , G) =
9
Q (A1 , M1 )
Q (G, M1 ) =
9
Q (A1 , G) = 4Q (G, M1 ) . ¦

Exercise 5.3 Show that in the complex number field, given complex numbers Q1 , Q2
and Q3 , not all zero, there is always a triangle with these numbers as quadrances. ¦

Theorem 21 (Parallelogram quadrance) In a parallelogram A1 A2 A3 A4 the

quadrances of opposite sides are equal.

Proof. Suppose that A1 ≡ [x1 , y1 ], A2 ≡ [x2 , y2 ], A3 ≡ [x3 , y3 ] and A4 ≡ [x4 , y4 ]. From

the Parallelogram center theorem (page 49), the midpoints of A1 A3 and A2 A4
coincide, so that
x1 + x3 = x2 + x4
y1 + y3 = y2 + y4 .
Thus
x1 − x2 = x4 − x3
y1 − y2 = y4 − y3 .
But then
Q (A1 , A2 ) = (x2 − x1 )2 + (y2 − y1 )2
2 2
= (x4 − x3 ) + (y4 − y3 )
= Q (A3 , A4 ) .

Exercise 5.4 Show that the diagonal sides A1 A3 and A2 A4 of a parallelogram

A1 A2 A3 A4 have equal quadrances precisely when the parallelogram is a rectangle. ¦

Exercise 5.5 Suppose that the parallelogram A1 A2 A3 A4 has quadrances

Q ≡ Q (A1 , A2 ) = Q (A3 , A4 ) and P ≡ Q (A2 , A3 ) = Q (A1 , A4 ) and that
R ≡ Q (A1 , A3 ) and T ≡ Q (A2 , A4 ) are the quadrances of the diagonal sides. Show
that
R + T = 2 (Q + P ) . ¦

Exercise 5.6 Show that all four quadrances of a parallelogram are equal precisely
when the parallelogram is a rhombus. ¦
5.2. TRIPLE QUAD FORMULA 63

5.2 Triple quad formula

This key result has ramifications and generalizations throughout the subject.

Theorem 22 (Triple quad formula) Suppose that A1 , A2 and A3 are points with
Q1 ≡ Q (A2 , A3 ), Q2 ≡ Q (A1 , A3 ) and Q3 ≡ Q (A1 , A2 ). Then
2 ¡ ¢
(Q1 + Q2 + Q3 ) = 2 Q21 + Q22 + Q23

precisely when A1 , A2 and A3 are collinear.

Proof. First verify the important polynomial identity

2 ¡ ¢ 2
(Q1 + Q2 + Q3 ) − 2 Q21 + Q22 + Q23 = 4Q1 Q2 − (Q1 + Q2 − Q3 ) .

Assume that A1 ≡ [x1 , y1 ], A2 ≡ [x2 , y2 ] and A3 ≡ [x3 , y3 ]. Then

Q1 = (x3 − x2 )2 + (y3 − y2 )2 A3
2 2
Q2 = (x3 − x1 ) + (y3 − y1 ) A2
2 2 A1 Q1
Q3 = (x2 − x1 ) + (y2 − y1 ) .
Q3 Q2
Rewrite this in the form
Q1 = a21 + b21
Q2 = a22 + b22
2 2
Q3 = (a2 − a1 ) + (b2 − b1 )
where a1 ≡ x3 − x2 , b1 ≡ y3 − y2 , a2 ≡ x3 − x1 and b2 ≡ y3 − y1 , so that
Q1 + Q2 − Q3 = 2 (a1 a2 + b1 b2 ) .
Then
4Q1 Q2 − (Q1 + Q2 − Q3 )2
³¡ ¢¡ ¢ ´
2
= 4 a21 + b21 a22 + b22 − (a1 a2 + b1 b2 )
= 4 (a1 b2 − a2 b1 )2
2
= 4 (x1 y2 − x1 y3 + x2 y3 − x3 y2 + x3 y1 − x2 y1 )
where Fibonacci’s identity (page 27) was used to go from the second line to the third.
Since 4 6= 0, the Collinear points theorem (page 39) shows that this is zero precisely
when the three points A1 , A2 and A3 are collinear.
64 5. QUADRANCE

Definition Archimedes’ function A (a, b, c) for numbers a, b and c is defined by

¡ ¢
A (a, b, c) ≡ (a + b + c)2 − 2 a2 + b2 + c2 .

Note that A (a, b, c) is a symmetric function of a, b and c. This motivates and justifies
the following definition.

Definition A set {a, b, c} of numbers is a quad triple precisely when

A (a, b, c) = 0.

Exercise 5.7 Show that

2
A (a, b, c) = 4ab − (a + b − c)
¡ ¢
= 2 (ab + bc + ca) − a2 + b2 + c2
2
= 4 (ab + bc + ca) − (a + b + c)
¯ ¯
¯ 2a a + b − c¯¯
=¯¯
a+b−c 2b ¯
¯ ¯
¯0 a b 1¯
¯ ¯
¯a 0 c 1¯
= −¯ ¯ ¯. ¦
¯
¯ b c 0 1¯
¯1 1 1 0¯

Exercise 5.8 Show that if three quantities a, b and c satisfy one of the relations
a ± b = ±c, then A ≡ a2 , B ≡ b2 and C ≡ c2 form a quad triple {A, B, C}. ¦

Exercise 5.9 Show that in general not every quad triple is of the form
n o
2
a2 , b2 , (a + b)

for some numbers a and b. ¦

Exercise 5.10 Show that as a quadratic equation in Q3 , the Triple quad formula is

Q23 − 2 (Q1 + Q2 ) Q3 + (Q1 − Q2 )2 = 0. ¦

Exercise 5.11 Show that the two quadratic equations in x

2
(x − p1 ) = q1
(x − p2 )2 = q2
n o
are compatible precisely when q1 , q2 , (p1 − p2 )2 is a quad triple. ¦
5.3. PYTHAGORAS’ THEOREM 65

5.3 Pythagoras’ theorem

This most famous of geometrical theorems becomes here more general, extending to an
arbitrary field, not of characteristic two.

Theorem 23 (Pythagoras’ theorem) Suppose that the triangle A1 A2 A3 has

quadrances Q1 ≡ Q (A2 , A3 ), Q2 ≡ Q (A1 , A3 ) and Q3 ≡ Q (A1 , A2 ). Then

Q1 + Q2 = Q3

precisely when A1 A3 and A2 A3 are perpendicular.

Proof. Suppose the points are A1 ≡ [x1 , y1 ], A2 ≡ [x2 , y2 ] and A3 ≡ [x3 , y3 ] .

Then by the Line through two points theorem (page 38)
A2
A1 A3 = hy1 − y3 : x3 − x1 : x1 y3 − x3 y1 i
Q3
and Q1
A2 A3 = hy2 − y3 : x3 − x2 : x2 y3 − x3 y2 i . A1 A3
These lines are perpendicular precisely when Q2

(y1 − y3 ) (y2 − y3 ) + (x3 − x1 ) (x3 − x2 ) = 0.

Now

Q1 + Q2 − Q3 = (x2 − x3 )2 + (y2 − y3 )2 + (x1 − x3 )2 + (y1 − y3 )2

2 2
− (x2 − x1 ) − (y2 − y1 )
¡ ¢
= 2 x23 − x2 x3 − x1 x3 + x2 x1 + y1 y2 − y1 y3 − y2 y3 + y32
= 2 ((y1 − y3 ) (y2 − y3 ) + (x3 − x1 ) (x3 − x2 )) .

Since 2 6= 0, A1 A3 and A2 A3 are perpendicular precisely when Q1 + Q2 = Q3 .

Exercise 5.12 Show that if A1 A2 A3 is a right triangle with right vertex at A3 , then it
is impossible for either Q1 or Q2 to be zero. Give an example to show that in some
fields it is possible for Q3 to be zero. ¦

Example 5.2 In the field F11 the right triangle A1 A2 A3 where

A1 ≡ [3, 7] A2 ≡ [10, 3] A3 ≡ [6, 1]

is shown in Figure 5.3, along with its lines.

66 5. QUADRANCE

10 A1 A2 = 3:8:1
9
A1 8 A2 A3 = 8:6:1
7
6
5 A1 A3 = 10:5:1
4
3
A3 2 A2
1
0
0 1 2 3 4 5 6 7 8 9 10

Figure 5.3: A right triangle in F11

You may check that A1 A3 and A2 A3 are perpendicular lines. The quadrances of
A1 A2 A3 are Q1 = 9, Q2 = 1 and Q3 = 10, and indeed Q1 + Q2 = Q3 . ¦

5.4 Quadrance to a line

Theorem 24 (Equal quadrance to two points) Suppose that p is the

perpendicular bisector of the side A1 A2 . Then every point A lying on p satisfies

Q (A, A1 ) = Q (A, A2 )

and conversely every point A satisfying this equation lies on p.

p
A

A1

M A2

Figure 5.4: Equal quadrance to two points

Proof. Suppose that A lies on p. If p = A1 A2 then A1 A2 is a null line, so by the Null

line theorem Q (A, A1 ) = Q (A, A2 ) = 0. If p 6= A1 A2 then A1 A2 is a non-null line. In
this case if A lies on A1 A2 , then A is the midpoint of A1 A2 , so by the Midpoint
5.4. QUADRANCE TO A LINE 67

theorem (page 60) Q (A, A1 ) = Q (A, A2 ). Otherwise A does not lie on A1 A2 . Then
AM A1 and AM A2 both have right vertices at the midpoint M of A1 A2 , so from
Pythagoras’ theorem
Q (A, A1 ) = Q (A, M ) + Q (M, A1 )
Q (A, A2 ) = Q (A, M ) + Q (M, A2 ) .
By the Midpoint theorem Q (M, A1 ) = Q (M, A2 ), so Q (A, A1 ) = Q (A, A2 ).

Conversely suppose that A is a point with Q (A, A1 ) = Q (A, A2 ) ≡ Q. If

Q (A1 , A2 ) = 0, then by the Null line theorem (page 60) A1 A2 is a null line. In that
case the quadrances of {A, A1 , A2 } are Q, Q and 0, so that they satisfy the Triple quad
formula. Thus the three points are collinear, and A lies on A1 A2 , which is itself the
perpendicular bisector p of A1 A2 .

Otherwise suppose that Q (A1 , A2 ) 6= 0, so that A1 A2 is a non-null line. Let F be the

foot of the altitude from A to the line A1 A2 . Then by Pythagoras’ theorem
Q (F, A1 ) = Q − Q (A, F ) = Q (F, A2 )
so that by the Midpoint theorem F is indeed the midpoint M of the side A1 A2 , and A
lies on the perpendicular bisector p of A1 A2 .

Theorem 25 (Quadrance to a line) For a point A ≡ [x, y] and a non-null line

l ≡ ha : b : ci, the quadrance from A to the foot F of the altitude from A to l is

(ax + by + c)2
Q (A, F ) = .
a2 + b2

Proof. From the Foot of an altitude theorem (page 42) the foot of the altitude from A
to l is ∙ 2 ¸
b x − aby − ac −abx + a2 y − bc
F ≡ , .
a2 + b2 a2 + b2
The required quadrance is then
µ 2 ¶2 µ ¶2
b x − aby − ac −abx + a2 y − bc
Q (A, F ) = −x + −y
a2 + b2 a2 + b2
a2 (ax + by + c)2 b2 (ax + by + c)2
= 2 + 2
(a2 + b2 ) (a2 + b2 )
(ax + by + c)2
= .
a2 + b2

Definition The quadrance Q (A, l) from the point A to the non-null line l is
the quadrance from A to the foot F of the altitude from A to l.
68 5. QUADRANCE

5.5 Quadrea

Definition For three points A1 , A2 and A3 with quadrances Q1 ≡ Q (A2 , A3 ),

Q2 ≡ Q (A1 , A3 ) and Q3 ≡ Q (A1 , A2 ), the quadrea A of the set {A1 , A2 , A3 } is the
number ¡ ¢
2
A ≡ A (Q1 , Q2 , Q3 ) = (Q1 + Q2 + Q3 ) − 2 Q21 + Q22 + Q23 .

The quadrea of {A1 , A2 , A3 } is also denoted A ({A1 , A2 , A3 }) and is a symmetric

function of the three points. In Chapter 9 the quadrea A of a triangle will be shown to
be sixteen times the square of the signed area of an associated oriented triangle.

Theorem 26 (Quadrea) The quadrea A of {A1 , A2 , A3 }, for A1 ≡ [x1 , y1 ],

A2 ≡ [x2 , y2 ] and A3 ≡ [x3 , y3 ], is

A = 4 (x1 y2 − x1 y3 + x2 y3 − x3 y2 + x3 y1 − x2 y1 )2 .

In particular the quadrea A is a square, and {A1 , A2 , A3 } is a triangle precisely when

A is non-zero.

Proof. The formula for A was derived in the course of the proof of the Triple quad
formula (page 63) and shows that A is a square. By the Collinear points theorem
(page 39)
x1 y2 − x1 y3 + x2 y3 − x3 y2 + x3 y1 − x2 y1 6= 0
precisely when A1 , A2 and A3 are not collinear, that is when they form a triangle.
Since 4 6= 0, this condition is equivalent to A 6= 0.

Theorem 27 (Right quadrea) If a right triangle A1 A2 A3 has quadrances Q1 , Q2

and Q3 , and right vertex at A3 , then it has quadrea

A = 4Q1 Q2 .

Proof. Use the asymmetric form

2
A = 4Q1 Q2 − (Q1 + Q2 − Q3 )
of the quadrea together with Pythagoras’ theorem.

Exercise 5.13 In the rational or decimal number fields, show that a variable triangle
A1 A2 A3 with two fixed quadrances Q1 and Q2 has maximum possible quadrea A
precisely when it has a right vertex at A3 . ¦
5.5. QUADREA 69

Exercise 5.14 Show that

A (Q1 , Q2 , Q3 ) = Q1 (Q2 + Q3 − Q1 ) + Q2 (Q3 + Q1 − Q2 ) + Q3 (Q1 + Q2 − Q3 ) . ¦

Exercise 5.15 Show that

A (Q1 , Q2 , Q3 ) = (Q3 + Q1 − Q2 ) (Q1 + Q2 − Q3 )

+ (Q1 + Q2 − Q3 ) (Q2 + Q3 − Q1 )
+ (Q2 + Q3 − Q1 ) (Q3 + Q1 − Q2 ) . ¦

Exercise 5.16 (Harder) Show that if the triangle A1 A2 A3 has lines

l1 ≡ ha1 : b1 : c1 i, l2 ≡ ha2 : b2 : c2 i and l3 ≡ ha3 : b3 : c3 i then the quadrea is

4 (a1 b2 c3 − a1 b3 c2 + a2 b3 c1 − a3 b2 c1 + a3 b1 c2 − a2 b1 c3 )4
A= 2 2 2 . ¦
(a2 b3 − a3 b2 ) (a3 b1 − a1 b3 ) (a1 b2 − a2 b1 )

The next result is due to V. Le.

Theorem 28 (Triangle quadrea) Suppose Q1 , Q2 and Q3 are not all zero and are
each the sum of two squares. Then a triangle A1 A2 A3 with these quadrances exists
precisely when A =A (Q1 , Q2 , Q3 ) is a non-zero square.

Proof. If such a triangle exists, then by the Quadrea theorem A (Q1 , Q2 , Q3 ) is a

non-zero square. Conversely suppose that there exists r 6= 0 such that

A (Q1 , Q2 , Q3 ) = 4Q1 Q2 − (Q1 + Q2 − Q3 )2 = r2 .

Then choose A3 ≡ [0, 0] and A1 ≡ [a1 , b1 ] such that a21 + b21 = Q2 6= 0. Now define

R ≡ Q1 + Q2 − Q3

and
a1 R − b1 r
a2 ≡
2Q2
b1 R + a1 r
b2 ≡ .
2Q2

Then check that A2 ≡ [a2 , b2 ] satisfies both Q (A2 , A3 ) = Q1 and Q (A2 , A1 ) = Q3 .

Exercise 5.17 (V. Le) In Fp suppose that Q1 , Q2 and Q3 are numbers such that
A ≡A (Q1 , Q2 , Q3 ) is a non-zero square. Show that i) if p = 4l + 3 then there are
exactly 2p2 (p + 1) triangles with quadrances Q1 , Q2 and Q3 and ii) if p = 4l + 1 then
there are exactly 2p2 (p − 1) triangles with quadrances Q1 , Q2 and Q3 . ¦
70 5. QUADRANCE

5.6 Archimedes’ formula

Heron’s formula of Euclidean geometry over the decimal number field is extended to a
general field, and now attributed (more properly) to Archimedes.

Theorem 29 (Archimedes’ formula) Suppose that a triangle A1 A2 A3 has square

quadrances of the form Q1 ≡ d21 , Q2 ≡ d22 and Q3 ≡ d23 for some numbers d1 , d2 and
d3 . Then the quadrea is

A = (d1 + d2 + d3 ) (d1 + d2 − d3 ) (d2 + d3 − d1 ) (d3 + d1 − d2 ) .

Proof. Proceed as follows

2 ¡ ¢2
A = 4Q1 Q2 − (Q1 + Q2 − Q3 ) = 4d21 d22 − d21 + d22 − d23
¡ ¡ ¢¢ ¡ ¡ ¢¢
= 2d1 d2 − d21 + d22 − d23 2d1 d2 + d21 + d22 − d23
³ ´³ ´
2 2
= d23 − (d1 − d2 ) (d1 + d2 ) − d23
= (d3 − d1 + d2 ) (d3 + d1 − d2 ) (d1 + d2 − d3 ) (d1 + d2 + d3 ) .

5.7 Quadruple quad formula

The next two theorems extend the Triple quad formula to four points.

Definition The Quadruple quad function Q (a, b, c, d) is defined by

³ ¡ ¢´2
Q (a, b, c, d) ≡ (a + b + c + d)2 − 2 a2 + b2 + c2 + d2 − 64abcd.

Theorem 30 (Two quad triples) Suppose that {a, b, x} and {c, d, x} are both
quad triples. Then
Q (a, b, c, d) = 0.
If a + b 6= c + d then
(a − b)2 − (c − d)2
x= .
2 (a + b − c − d)
5.7. QUADRUPLE QUAD FORMULA 71

Proof. If {a, b, x} and {c, d, x} are quad triples then

(x − a − b)2 = 4ab
(x − c − d)2 = 4cd.

The Quadratic compatibility theorem (page 33) asserts that then a, b, c and d must
satisfy the condition
³ ´2
2
(a + b − c − d) − 4 (ab + cd) = 64abcd

which can be rewritten more symmetrically as

³ ¡ ¢´2
2
(a + b + c + d) − 2 a2 + b2 + c2 + d2 = 64abcd.

The same theorem also asserts that if a + b 6= c + d then

(a + b) + (c + d) 4ab − 4cd
x= −
2 2 (a + b − c − d)
(a − b)2 − (c − d)2
= .
2 (a + b − c − d)

¡ ¢
The expression (a + b + c + d)2 − 2 a2 + b2 + c2 + d2 appears in a theorem of
Descartes as well as in F. Soddy’s circle theorem. This is discussed in [Coxeter, pages
14, 15], which also has an amusing related verse by Soddy.

Theorem 31 (Quadruple quad formula) Suppose that A1 , A2 , A3 and A4 are

collinear points and that Qij ≡ Q (Ai , Aj ) for all i, j = 1, 2, 3 and 4. Then

Q (Q12 , Q23 , Q34 , Q14 ) = 0.

Furthermore

(Q12 − Q23 )2 − (Q34 − Q14 )2

Q13 =
2 (Q12 + Q23 − Q34 − Q14 )
2 2
(Q23 − Q34 ) − (Q12 − Q14 )
Q24 =
2 (Q23 + Q34 − Q12 − Q14 )

provided the denominators are not zero.

Proof. If A1 , A2 , A3 and A4 are collinear points, then {Q12 , Q23 , Q13 } and
{Q14 , Q13 , Q34 } are both quad triples. Apply the Two quad triples theorem to see that
Q (Q12 , Q23 , Q34 , Q14 ) = 0, and to obtain the stated formula for Q13 . The formula for
Q24 is analogous.
72 5. QUADRANCE

Theorem 32 (Brahmagupta’s identity) Suppose that Q12 ≡ d212 , Q23 ≡ d223 ,

Q34 ≡ d234 and Q14 ≡ d214 for some numbers d12 , d23 , d34 and d14 . Then

Q (Q12 , Q23 , Q34 , Q14 )

= (d12 − d14 + d23 + d34 ) (d12 + d14 + d23 − d34 ) (d14 − d12 + d23 + d34 )
× (d12 + d14 − d23 + d34 ) (d12 + d14 + d23 + d34 ) (d12 − d14 − d23 + d34 )
× (d12 − d14 + d23 − d34 ) (d23 − d14 − d12 + d34 ) .

Proof. Make the substitutions Qij ≡ d2ij for all i and j to turn the expression
³ ¡ ¢´2
2
(Q12 + Q23 + Q34 + Q14 ) − 2 Q212 + Q223 + Q234 + Q214 − 64Q12 Q23 Q34 Q14

into a difference of squares. This is then the product of the expression

¡ 2 ¢2 ¡ ¢
d12 + d223 + d214 + d234 − 2 d412 + d423 + d414 + d434 + 8d12 d23 d34 d14
= (d12 − d14 + d23 + d34 ) (d12 + d14 + d23 − d34 )
× (d14 − d12 + d23 + d34 ) (d12 + d14 − d23 + d34 )

and the expression

¡ 2 ¢2 ¡ ¢
d12 + d223 + d214 + d234 − 2 d412 + d423 + d414 + d434 − 8d12 d23 d34 d14
= (d12 + d14 + d23 + d34 ) (d12 − d14 − d23 + d34 )
× (d12 − d14 + d23 − d34 ) (d23 − d14 − d12 + d34 ) .

The first of the two expressions involved in the proof of Brahmagupta’s identity
corresponds in the decimal number system to sixteen times the square of the area of a
convex cyclic quadrilateral with side lengths d12 , d23 , d34 and d14 . The other
corresponds to an analogous result for a non-convex cyclic quadrilateral with these side
lengths (see [Robbins]). These connections with cyclic quadrilaterals will become
clearer in Chapter 17 with the Cyclic quadrilateral quadrea theorem (page 187).

Exercise 5.18 Show that if {a, b, x} and {b, c, x} are quad triples and a 6= c then

a + c − 2b
x= . ¦
2

Exercise 5.19 (Harder) Is there a Quintuple quad formula? Generalize. ¦

 6

Spread

This chapter introduces the spread between non-null lines in terms of the coefficients of
the lines, and derives an alternative formulation as the ratio of two quadrances. The
related notions of the cross and twist between two lines are also introduced.

Spread and cross are seen to be essentially equivalent, but spread is here given pride of
place. This becomes particularly useful in elliptic and hyperbolic geometries. The
Spread law and Cross law are derived, along with some formulas allowing us to
calculate spreads from coordinates of points. The subtle issue of bisectors of a vertex is
examined.

6.1 Spreads of triangles and quadrilaterals

Definition The spread s (l1 , l2 ) between the non-null lines l1 ≡ ha1 : b1 : c1 i and
l2 ≡ ha2 : b2 : c2 i is the number

(a1 b2 − a2 b1 )2
s (l1 , l2 ) ≡ .
(a21 + b21 ) (a22 + b22 )

If one or both of l1 , l2 is a null line, then the spread s (l1 , l2 ) is an empty expression
and any statement involving it is considered an empty statement. The spread is
otherwise well-defined, in the sense that if the coefficients of one of the proportions are
multiplied by a non-zero number, then the spread remains unchanged.

Also s (l1 , l2 ) = s (l2 , l1 ), and s (l1 , l2 ) = 0 precisely when l1 and l2 are parallel. The
spread s (l1 , l2 ) is unchanged if l1 is replaced with a line parallel to it, or if l2 is
replaced by a line parallel to it.

73
74 6. SPREAD

Definition Given a triangle A1 A2 A3 , the numbers s1 ≡ s (A1 A2 , A1 A3 ),

s2 ≡ s (A2 A1 , A2 A3 ) and s3 ≡ s (A3 A1 , A3 A2 ) are the spreads of the triangle, with
s1 the spread of the vertex l2 l3 , or the spread opposite the side A2 A3 , and
similarly for the other spreads.

A2 A2
A3
s2
s2 s3

A1 s1 s 3 A3 s1 s4
A1 A4

Figure 6.1: Spreads of a triangle and quadrilateral

More generally if l1 , l2 and l3 are any three lines, the usual convention is that the three
spreads formed by the pairs of lines are denoted s1 ≡ s (l2 , l3 ), s2 ≡ s (l1 , l3 ) and
s3 ≡ s (l1 , l2 ).

Definition Given a quadrilateral A1 A2 A3 A4 , the numbers s1 ≡ s (A1 A2 , A1 A4 ),

s2 ≡ s (A2 A1 , A2 A3 ), s3 ≡ s (A3 A2 , A3 A4 ) and s4 ≡ s (A4 A1 , A4 A3 ) are the spreads
of the quadrilateral.

6.2 Cross

Definition The cross c (l1 , l2 ) between the non-null lines l1 ≡ ha1 : b1 : c1 i and
l2 ≡ ha2 : b2 : c2 i is the number
2
(a1 a2 + b1 b2 )
c (l1 , l2 ) ≡ .
(a21 + b21 ) (a22 + b22 )

A cross involving a null line is an empty concept. Also c (l1 , l2 ) = c (l2 , l1 ), and
c (l1 , l2 ) = 0 precisely when l1 and l2 are perpendicular. The cross is unchanged if
either of the lines is replaced by a line parallel to it.

In a triangle, the conventions are similar to the ones for spread, so that c1 is the cross
at the vertex l2 l3 and so on. In a diagram in the rational or decimal number fields,
the cross between intersecting lines will be written inside the small triangle formed by
the two lines and a small line segment joining them, as in Figure 6.2.
6.3. TWIST 75

A2

c2

A1 c1 c3 A 3

Figure 6.2: Crosses of a triangle

Given three lines l1 , l2 and l3 , the usual convention is that the three crosses formed by
the pairs of lines are denoted c1 ≡ c (l2 , l3 ), c2 ≡ c (l1 , l3 ) and c3 ≡ c (l1 , l2 ).

6.3 Twist

Definition The twist t (l1 , l2 ) between the non-perpendicular lines l1 ≡ ha1 : b1 : c1 i

and l2 ≡ ha2 : b2 : c2 i is the number
2
(a1 b2 − a2 b1 )
t (l1 , l2 ) ≡ .
(a1 a2 + b1 b2 )2

A2

t2

A1 t 1 t 3 A3

Figure 6.3: Twists of a triangle

The twist is not defined if l1 and l2 are perpendicular, as in this case a1 a2 + b1 b2 = 0.

The twist is always a square number. Also t (l1 , l2 ) = t (l2 , l1 ), and t (l1 , l2 ) = 0
precisely when l1 and l2 are parallel. If l1 and l2 are non-parallel non-null lines then
t (l1 , l2 ) = s (l1 , l2 ) /c (l1 , l2 ) .
Note that if one or both of l1 and l2 are null lines, then the twist between them may
still be defined, although in this case neither the spread nor the cross is defined. The
twist between two lines may be indicated by a double line segment, as in Figure 6.3.
76 6. SPREAD

6.4 Ratio theorems

Theorem 33 (Spread plus cross) For non-null lines l1 and l2 , let s and c be
respectively the spread and cross between them. Then

s + c = 1.

Proof. This follows from the definitions of spread and cross, together with Fibonacci’s
identity
¡ ¢¡ ¢
(a1 b2 − a2 b1 )2 + (a1 a2 + b1 b2 )2 = a21 + b21 a22 + b22 .
Over the rational or decimal number field, this theorem implies that both the spread
and cross always take values in the interval [0, 1], since they are both always positive.

Theorem 34 (Spread number) For non-null lines l1 and l2 the spread s ≡ s (l1 , l2 )
is a spread number, and every spread number is obtained as a spread between two
lines.

Proof. If l1 ≡ ha1 : b1 : c1 i and l2 ≡ ha2 : b2 : c2 i then the definitions of the spread

s ≡ s (l1 , l2 ) and cross c ≡ c (l1 , l2 ), together with the Spread plus cross theorem, gives

s (1 − s) = sc
2 2
(a1 b2 − a2 b1 ) (a1 a2 + b1 b2 )
=
(a1 + b1 ) (a2 + b2 ) (a21 + b21 ) (a22 + b22 )
2 2 2 2
µ ¶2
(a1 b2 − a2 b1 ) (a1 a2 + b1 b2 )
= .
(a21 + b21 ) (a22 + b22 )

Since this is a square, s is a spread number.

Conversely, suppose that s is a spread number, so that s (1 − s) ≡ r2 for some number

r. If s = 1, then it is the spread between the two lines h0 : 1 : 0i and h1 : 0 : 0i.
Otherwise, consider the two lines l1 = h0 : 1 : 0i and l2 = hr : 1 − s : 0i. The spread
between them is, from the definition,

r2
s (l1 , l2 ) = ³ ´
1 × r2 + (1 − s)2
s (1 − s)
= = s.
1−s
6.4. RATIO THEOREMS 77

Theorem 35 (Spread ratio) Suppose that A1 A2 A3 is a non-null right triangle

with a right vertex at A3 , and quadrances Q1 , Q2 and Q3 . Then the spread at the
vertex A1 is
s1 ≡ s (A1 A2 , A1 A3 ) = Q1 /Q3 .

s 1 = Q 1/Q 3 A2
Q3
Q1

s1
A1 Q2 A3

Figure 6.4: Spread ratio theorem

Proof. Assume the points are A1 ≡ [x1 , y1 ], A2 ≡ [x2 , y2 ] and A3 ≡ [x3 , y3 ]. Then
A1 A2 = hy1 − y2 : x2 − x1 : x1 y2 − x2 y1 i
A1 A3 = hy1 − y3 : x3 − x1 : x1 y3 − x3 y1 i
A2 A3 = hy2 − y3 : x3 − x2 : x2 y3 − x3 y2 i
so that
((y1 − y2 ) (x3 − x1 ) − (y1 − y3 ) (x2 − x1 ))2
s1 = ³ ´³ ´.
(y1 − y2 )2 + (x2 − x1 )2 (y1 − y3 )2 + (x3 − x1 )2
Since A1 A3 and A2 A3 are perpendicular
(y1 − y3 ) (y2 − y3 ) + (x3 − x1 ) (x3 − x2 ) = 0.
The following identity appeared in Exercise 2.6
(y1 − y2 ) (x3 − x1 ) − (y1 − y3 ) (x2 − x1 )
= (y1 − y3 ) (x3 − x2 ) − (y2 − y3 ) (x3 − x1 ) .
So the numerator of s1 is
2
((y1 − y2 ) (x3 − x1 ) − (y1 − y3 ) (x2 − x1 ))
= ((y1 − y3 ) (x3 − x2 ) − (y2 − y3 ) (x3 − x1 ))2
+ ((y1 − y3 ) (y2 − y3 ) + (x3 − x1 ) (x3 − x2 ))2
³ ´³ ´
= (y1 − y3 )2 + (x3 − x1 )2 (y2 − y3 )2 + (x3 − x2 )2

where Fibonacci’s identity is used in the last step. Hence

s1 = (Q2 Q1 ) / (Q3 Q2 ) = Q1 /Q3 .
78 6. SPREAD

Theorem 36 (Cross ratio) Suppose that A1 A2 A3 is a non-null right triangle with

a right vertex at A3 , and quadrances Q1 , Q2 and Q3 . Then the cross at the vertex A1
is
Q2
c1 ≡ c (A1 A2 , A1 A3 ) = .
Q3

Proof. Use the Spread plus cross theorem, the Spread ratio theorem
and Pythagoras’ theorem to get
c1 ≡ c (A1 A2 , A1 A3 )
c1 = Q 2/Q 3 A2
= 1 − s (A1 A2 , A1 A3 )
Q1 Q3
= 1− Q1
Q3
Q2
= . c1
Q3
A1 Q2 A3

Theorem 37 (Twist ratio) Suppose that A1 A2 A3 is a right triangle with a right

vertex at A3 , and quadrances Q1 , Q2 and Q3 . Then the twist t1 at the vertex A1 is
Q1
t1 ≡ t (A1 A2 , A1 A3 ) = .
Q2

Proof. If Q3 6= 0, this follows immediately from t1 = s1 /c1 together with the Spread
ratio theorem s1 = Q1 /Q3 and the Cross ratio theorem c1 = Q2 /Q3 .

However if Q3 = 0 then neither s1 nor c1 are defined. In this case Pythagoras’ theorem
shows that Q1 = −Q2 . It then suffices to show that t1 = −1. By the Null line theorem
(page 60) l3 is a null line, so is of the form h1 : i : di for some number i satisfying
i2 = −1, and some number d.

The lines l2 and l3 are not perpendicular, since otherwise l1 would be parallel to l3 . So
if l2 ≡ ha : b : ci, then a + bi 6= 0, and the definition of the twist gives
2
(ai − b)
t1 = t (l2 , l3 ) = = −1.
(a + bi)2

Exercise 6.1 Show that if the spread s, cross c and twist t are defined between two
lines, then
t = s/c = s/ (1 − s) = (1 − c) /c s = t/ (1 + t) c = 1/ (1 + t) . ¦
6.5. COMPLEMENTARY SPREADS 79

6.5 Complementary spreads

Two spreads s1 and s2 are complementary precisely when s1 + s2 = 1.

Theorem 38 (Complementary spreads) If the non-null lines l1 and l2 are

perpendicular, then for any non-null line l3 the spreads s1 ≡ s (l2 , l3 ) and
s2 ≡ s (l1 , l3 ) are complementary.

Proof. Suppose that l1 and l2 are perpendicular non-null lines. If l3 is parallel to one
of these lines, then it is perpendicular to the other, in which case s1 and s2 are 0 and 1
in some order, and so complementary.
Otherwise any two of the lines intersect, and
since spreads are unchanged when lines are replaced by parallel
lines, assume that the lines l1 , l2 and l3 are not concurrent.
Let A1 ≡ l2 l3 , A2 ≡ l1 l3 and A3 ≡ l1 l2 , and suppose that the s2
quadrances of A1 A2 A3 are Q1 , Q2 and Q3 . Then since A1 A2 A3 l 2 l3
has a right vertex at A3 , the Spread ratio theorem states that s1
s1 = Q1 /Q3 and s2 = Q2 /Q3 . l1
Thus
s1 + s2 = (Q1 + Q2 ) /Q3 = 1
by Pythagoras’ theorem.

Exercise 6.2 Show that the converse does not hold. ¦

Theorem 39 (Perpendicular spreads) If the non-null lines l1 and l2 are

perpendicular, and the non-null lines m1 and m2 are perpendicular, then

s (l1 , m1 ) = s (l2 , m2 ) .

Proof. Apply the Complementary spreads theorem to

the lines m1 , m2 and l1 to get
s (l1 , m1 ) + s (l1 , m2 ) = 1.
Apply it to the lines l1 , l2 and m2 to get m2 m1
s (l1 , m2 ) + s (l2 , m2 ) = 1.
Subtract these equations to conclude that s l2 s l1
s (l1 , m1 ) = s (l2 , m2 ) .
80 6. SPREAD

6.6 Spread law

Theorem 40 (Spread law) Suppose three points A1 , A2 and A3 form non-zero

quadrances Q1 ≡ Q (A2 , A3 ), Q2 ≡ Q (A1 , A3 ) and Q3 ≡ Q (A1 , A2 ). Define the
spreads s1 ≡ s (A1 A2 , A1 A3 ), s2 ≡ s (A2 A1 , A2 A3 ) and s3 ≡ s (A3 A1 , A3 A2 ). Then
s1 s2 s3
= = .
Q1 Q2 Q3

Proof. If A1 , A2 and A3 are collinear, then s1 = s2 = s3 = 0 and the theorem is

immediate. Otherwise the three points form a triangle A1 A2 A3 , which is not null,
since the quadrances are non-zero. Suppose that D is the foot of the altitude from A1
to A2 A3 . Define the quadrances

R1 ≡ Q (A1 , D) R2 ≡ Q (A2 , D) R3 ≡ Q (A3 , D)

as illustrated in either of the diagrams in Figure 6.5.

A1 A1

Q2 Q2
Q3 R1
R1 Q3
A3 A2 A3
R3 D R2 Q 1 A2 R 2 D
Q1 R3

Figure 6.5: Spread law

Since A1 A2 D and A1 A3 D are right triangles, with right vertices at D, the Spread ratio
theorem asserts that
s2 = R1 /Q3 s3 = R1 /Q2 .
Solve for R1 to get
R1 = Q3 s2 = Q2 s3
so that
s2 s3
= .
Q2 Q3
In a similar manner
s1 s2
= .
Q1 Q2
6.7. CROSS LAW 81

6.7 Cross law

The following result contains both Pythagoras’ theorem and the Triple quad formula
as special cases.

Theorem 41 (Cross law) Suppose three points A1 , A2 and A3 form quadrances

Q1 ≡ Q (A2 , A3 ), Q2 ≡ Q (A1 , A3 ) and Q3 ≡ Q (A1 , A2 ), and define the cross
c3 ≡ c (A3 A1 , A3 A2 ). Then
2
(Q1 + Q2 − Q3 ) = 4Q1 Q2 c3 .

Proof. If the three points A1 , A2 and A3 are collinear, then c3 = 1, so then the
statement follows from the Triple quad formula.

Otherwise the three points form a triangle A1 A2 A3 . By the definition of the cross c3 ,
A1 A3 and A2 A3 are not null lines, so let D be the foot of the altitude from A1 to the
line A2 A3 . As in Figure 6.5 define the quadrances

R1 ≡ Q (A1 , D) R2 ≡ Q (A2 , D) R3 ≡ Q (A3 , D) .

Since A1 A2 D and A1 A3 D are right triangles, with right vertices at D, use Pythagoras’
theorem to get

Q3 = R1 + R2
Q2 = R1 + R3 .

By the Cross ratio theorem (page 78)

c3 = R3 /Q2 .

Solve sequentially for R3 , R1 and then R2 to get

R3 = Q2 c3
R1 = Q2 (1 − c3 )
R2 = Q3 − Q2 (1 − c3 ) .

Since A2 , A3 and D are collinear, apply the Triple quad formula to the three
quadrances Q1 , R2 and R3 , yielding

(Q1 + R3 − R2 )2 = 4Q1 R3 .

Substitute the values of R3 and R2 , to get

2
(Q1 + Q2 − Q3 ) = 4Q1 Q2 c3 .
82 6. SPREAD

The next result is a consequence of the Cross law, and gives a convenient way to
quickly calculate the spreads of a triangle from the quadrances.

Theorem 42 (Quadrea spread) Suppose three points A1 , A2 and A3 form

quadrances Q1 , Q2 and Q3 as usual, and spread s3 ≡ s (A3 A1 , A3 A2 ). Then the
quadrea of {A1 , A2 , A3 } is
A = 4Q1 Q2 s3 .

Proof. From the Spread plus cross theorem and the Cross law

4Q1 Q2 s3 = 4Q1 Q2 (1 − c3 )
2
= 4Q1 Q2 − (Q1 + Q2 − Q3 )
¡ ¢
= (Q1 + Q2 + Q3 )2 − 2 Q21 + Q22 + Q23
= A.

Exercise 6.3 Use the Quadrea spread theorem to show that if a triangle A1 A2 A3 has
quadrances Q1 , Q2 and Q3 , spreads s1 , s2 and s3 , and quadrea A, then

s1 s2 s3 A
= = = . ¦
Q1 Q2 Q3 4Q1 Q2 Q3

Exercise 6.4 Suppose the null triangle A1 A2 A3 has quadrances Q1 , Q2 and Q3 , with
Q3 ≡ 0. Show that the quadrea of the triangle is
2
A = − (Q1 − Q2 )

and that
2
s3 = − (Q1 − Q2 ) /4Q1 Q2 . ¦

Exercise 6.5 Suppose the null triangle A1 A2 A3 has quadrances Q1 6= 0 and

Q2 = Q3 ≡ 0. Show that the quadrea is

A = −Q21 . ¦

Exercise 6.6 (Triangle inequality) In the rational or decimal number fields, show
that if a triangle has quadrances Q1 , Q2 and Q3 , then

2
(Q1 + Q2 − Q3 ) ≤ 4Q1 Q2 . ¦
6.7. CROSS LAW 83

Examples

Example 6.1 In the decimal number field, the triangle A1 A2 A3 with quadrances

Q1 ≡ 16 Q2 ≡ 36 Q3 ≡ 9

has quadrea
¡ ¢
A = A (16, 36, 9) = (16 + 36 + 9)2 − 2 162 + 362 + 92 = 455.

So by the Quadrea spread theorem

A 455 455
s1 = = =
4Q2 Q3 4 × 36 × 9 1296
A 455 455
s2 = = =
4Q1 Q3 4 × 16 × 9 576
A 455 455
s3 = = = . ¦
4Q1 Q2 4 × 16 × 36 2304

Example 6.2 The Quadrea theorem (page 68) shows that there is no triangle over the
rational numbers with quadrances

Q1 ≡ 16 Q2 ≡ 36 Q3 ≡ 9

since in this field the quadrea A ≡ 455 is not a square. ¦

Example 6.3 In the field F11 the number A ≡ 455 = 4 is a square, and every number
is the sum of two squares, so by the Triangle quadrea theorem (page 69) there is a
triangle with quadrances

Q1 ≡ 16 = 5 Q2 ≡ 36 = 3 Q3 ≡ 9.

The spreads of this triangle are

A 4
s1 = = =9
4Q2 Q3 4×3×9
A 4
s2 = = =1
4Q1 Q3 4×5×9
A 4
s3 = = = 3. ¦
4Q1 Q2 4×5×3

Example 6.4 Over the complex numbers, the triangle A1 A2 A3 with quadrances

Q1 ≡ 1 Q2 ≡ 2 Q3 ≡ i

has quadrea
2
A ≡ A (1, 2, i) = (1 + 2 + i) − 2 (1 + 4 − 1) = 6i.
So by the Quadrea spread theorem

s1 = 3/4 s2 = 3/2 s3 = 3i/4. ¦

84 6. SPREAD

6.8 Spreads in coordinates

The next theorems give the spread and cross in terms of points lying on lines.

Theorem 43 (Spread from points) Suppose that the non-null lines l1 and l2
intersect at the point A3 ≡ [x3 , y3 ], and that A1 ≡ [x1 , y1 ] is any other point on l1
and A2 ≡ [x2 , y2 ] is any other point on l2 . Then the spread s between l1 and l2 is

((y1 − y3 ) (x3 − x2 ) − (y2 − y3 ) (x3 − x1 ))2

s= ³ ´³ ´.
(x1 − x3 )2 + (y1 − y3 )2 (x2 − x3 )2 + (y2 − y3 )2

Proof. The Line through two points theorem (page 38) shows that
l1 = hy1 − y3 : x3 − x1 : x1 y3 − x3 y1 i
l2 = hy2 − y3 : x3 − x2 : x2 y3 − x3 y2 i .
Now use the definition of the spread.

l2
A2=[x 2,y 2]

A1=[x 1,y 1]
s
c l1
A3=[x 3,y 3]

Figure 6.6: Spread and cross from points

Theorem 44 (Cross from points) Suppose that the non-null lines l1 and l2
intersect at the point A3 ≡ [x3 , y3 ] and that A1 ≡ [x1 , y1 ] is any other point on l1 and
that A2 ≡ [x2 , y2 ] is any other point on l2 . Then the cross c between l1 and l2 is

((y1 − y3 ) (y2 − y3 ) + (x3 − x1 ) (x3 − x2 ))2

c= ³ ´³ ´.
(x1 − x3 )2 + (y1 − y3 )2 (x2 − x3 )2 + (y2 − y3 )2

Proof. This is the same argument as the previous proof.

Exercise 6.7 Formulate the corresponding result for twists. ¦

6.9. VERTEX BISECTORS 85

6.9 Vertex bisectors

Definition A bisector of the non-null vertex l1 l2 is a line h which passes through

l1 l2 and satisfies s (l1 , h) = s (l2 , h).

Theorem 45 (Vertex bisector) A bisector of the non-null vertex l1 l2 exists

precisely when s (l1 , l2 ) is a square. In this case there are exactly two bisectors, and
they are perpendicular. If l1 ≡ ha1 : b1 : c1 i and l2 ≡ ha2 : b2 : c2 i, and
a1 b2 + a2 b1 = 0 then the two bisectors of l1 l2 always exists and are

ha1 b2 − a2 b1 : 0 : b2 c1 − b1 c2 i and h0 : a1 b2 − a2 b1 : c2 a1 − c1 a2 i .
¡ ¢¡ ¢
6 0 and a21 + b21 a22 + b22 ≡ r2 , then the two bisectors of l1 l2 are
If a1 b2 + a2 b1 =
given by
¿ À
(a1 a2 −
¡ b21 b2 +¢r) (a1 b2¡− a2 b1 )¢: (a1 b2 + a2 b1 ) (a1 b2 − a2 b1 )
a1 + b21 b2 c2 − a22 + b22 b1 c1 + r (b2 c1 − b1 c2 )

and ¿ À
(a1 a2 −
¡ b21 b2 −¢r) (a1 b2¡− a2 b1 )¢: (a1 b2 + a2 b1 ) (a1 b2 − a2 b1 ) .
a1 + b21 b2 c2 − a22 + b22 b1 c1 − r (b2 c1 − b1 c2 )

Proof. If s (l1 , h) = s (l2 , h) ≡ s then the Equal spreads theorem (page 94) shows that
s (l1 , l2 ) = 4s (1 − s), since l1 and l2 are not parallel. The Spread number theorem
(page 76) states that s (1 − s) is a square, so that s (l1 , l2 ) is also a square.

Conversely suppose that l1 ≡ ha1 : b1 : c1 i and l2 ≡ ha2 : b2 : c2 i, and that s ≡ s (l1 , l2 )

is a square. Since
(a1 b2 − a2 b1 )2
s= 2
(a1 + b21 ) (a22 + b22 )
this is equivalent to the condition that
¡ 2 ¢¡ ¢
a1 + b21 a22 + b22 = r2

for some r. If h ≡ ha : b : ci then the equation s (l1 , h) = s (l2 , h) is

2 2
(a1 b − ab1 ) (a2 b − ab2 )
= 2 . (6.1)
(a21+ b21 ) (a2 + b2 ) (a2 + b22 ) (a2 + b2 )

Verify the polynomial identity

2¡ ¢ 2¡ ¢
(a1 b − ab1 ) a22 + b22 − (a2 b − ab2 ) a21 + b21
¡¡ ¢ ¢
= − (a1 b2 − a2 b1 ) a2 − b2 (a1 b2 + a2 b1 ) − 2ab (a1 a2 − b1 b2 ) . (6.2)
86 6. SPREAD

Since l1 and l2 are not parallel, a1 b2 − a2 b1 6= 0, so use (6.2) to replace (6.1) with the
equivalent condition

a2 (a1 b2 + a2 b1 ) − 2ab (a1 a2 − b1 b2 ) − b2 (a1 b2 + a2 b1 ) = 0. (6.3)

Suppose that a1 b2 + a2 b1 = 0. Then (6.3) reduces to

ab (a1 a2 − b1 b2 ) = 0.

If then a1 a2 − b1 b2 = 0, then the variant of Fibonacci’s identity of Exercise 2.5,

¡ ¢¡ ¢
(a1 a2 − b1 b2 )2 + (a1 b2 + a2 b1 )2 = a21 + b21 a22 + b22

would imply that ¡ 2 ¢¡ ¢

a1 + b21 a22 + b22 = 0
which is impossible since neither l1 nor l2 is a null line. So instead ab = 0, and h is
either of the form h1 : 0 : d1 i or h0 : 1 : d2 i for unique values of d1 and d2 , determined
by the fact that h passes through
∙ ¸
b1 c2 − b2 c1 c1 a2 − c2 a1
l1 l2 = , .
a1 b2 − a2 b1 a1 b2 − a2 b1
This gives the two possibilities

h1 ≡ ha1 b2 − a2 b1 : 0 : b2 c1 − b1 c2 i

or
h2 ≡ h0 : a1 b2 − a2 b1 : c2 a1 − c1 a2 i .
Note that s is automatically a square in this case, and in fact
µ ¶2
a1 b2 − a2 b1
s= .
a1 a2 − b1 b2

Suppose now that a1 b2 + a2 b1 6= 0. In this case (6.3) shows that neither a nor b can be
zero, since if one of them is zero, so is the other, which is impossible as ha : b : ci is a
line. So by an appropriate non-zero scaling of a, b and c you can take

b ≡ a1 b2 + a2 b1 6= 0.

In (6.3) complete the square in a using the variant of Fibonacci’s identity to get

(a − (a1 a2 − b1 b2 ))2 = (a1 a2 − b1 b2 )2 + (a1 b2 + a2 b1 )2

¡ ¢¡ ¢
= a21 + b21 a22 + b22 = r2 .

So the possibilities for h are

h1 ≡ ha1 a2 − b1 b2 + r : a1 b2 + a2 b1 : d1 i

or
h2 ≡ ha1 a2 − b1 b2 − r : a1 b2 + a2 b1 : d2 i .
6.9. VERTEX BISECTORS 87

The numbers d1 and d2 are determined by the fact that h passes through l1 l2 . A
calculation shows that the possibilities for h can then be written
¿ À
(a1 a2 −
¡ b21 b2 +¢r) (a1 b2¡− a2 b1 )¢: (a1 b2 + a2 b1 ) (a1 b2 − a2 b1 )
h1 =
a1 + b21 b2 c2 − a22 + b22 b1 c1 + r (b2 c1 − b1 c2 )

or ¿ À
(a1 a2 −
¡ b21 b2 −¢r) (a1 b2¡− a2 b1 )¢: (a1 b2 + a2 b1 ) (a1 b2 − a2 b1 )
h2 = .
a1 + b21 b2 c2 − a22 + b22 b1 c1 − r (b2 c1 − b1 c2 )

These two lines are perpendicular since

((a1 a2 − b1 b2 ) + r) ((a1 a2 − b1 b2 ) − r) + (a1 b2 + a2 b1 )2 = 0

by the variant of Fibonacci’s identity and the definition of r.

Definition The vertex l1 l2 is a square vertex precisely when it has bisectors.

Example 6.5 Working over the rational number field, suppose that l1 ≡ h2 : 1 : −1i
and l2 ≡ h11 : 2 : −4i. Then
¡ 2 ¢¡ ¢ 2
a1 + b21 a22 + b22 = 625 = (25)

is a square, so the vertex l1 l2 has bisectors. Set r ≡ 25, and observe that
a1 b2 + a2 b1 6= 0. Then
¿ À
(a1 a2 −
¡ b21 b2 ±¢r) (a1 b2¡− a2 b1 )¢: (a1 b2 + a2 b1 ) (a1 b2 − a2 b1 )
a1 + b21 b2 c2 − a22 + b22 b1 c1 ± r (b2 c1 − b1 c2 )

evaluate to give bisectors h1 ≡ h1 : −3 : 1i and h2 ≡ h21 : 7 : −9i. These lines are

perpendicular and you may check that

s (l1 , h1 ) = s (l2 , h1 ) = 49/50

s (l1 , h2 ) = s (l2 , h2 ) = 1/50. ¦

Example 6.6 Over the rational number field, suppose that l1 ≡ h3 : 1 : 5i and
l2 ≡ h−3 : 1 : 2i. Then s (l1 , l2 ) = 9/25 is a square, so the vertex l1 l2 has bisectors. In
this case a1 b2 + a2 b1 = 0, so the bisectors are given by the theorem as

h1 ≡ h6 : 0 : 3i = h2 : 0 : 1i and h2 ≡ h0 : 6 : 21i = h0 : 2 : 7i .

You may check that

s (l1 , h1 ) = s (l2 , h1 ) = 1/10

s (l1 , h2 ) = s (l2 , h2 ) = 9/10. ¦
88 6. SPREAD

Theorem 46 (Equal quadrance to two lines) If l1 l2 is a square vertex, then any

point A lying on a vertex bisector h of l1 l2 satisfies

Q (A, l1 ) = Q (A, l2 )

and conversely any point satisfying this equation lies on one of the vertex bisectors of
l1 l2 .

Proof. Suppose that l1 l2 is a square vertex and that A is a point on one of the vertex
bisectors h of l1 l2 . Suppose that A0 ≡ l1 l2 and that F1 and F2 are the feet of the
altitudes from A to l1 and l2 respectively as in Figure 6.7.

l2
F2 A h

s l1
A0 s F1

Figure 6.7: Equal quadrance to two lines

Then both A0 AF1 and A0 AF2 are right triangles, and by assumption

Q (A, F1 ) Q (A, F2 )
s (l1 , h) = = s (l2 , h) = .
Q (A, A0 ) Q (A, A0 )

Thus
Q (A, l1 ) = Q (A, F1 ) = Q (A, F2 ) = Q (A, l2 ) .
Conversely if A 6= l1 l2 and Q (A, l1 ) = Q (A, l2 ), then with F1 and F2 the feet of the
altitudes from A to l1 and l2 respectively, you have

Q (A, F1 ) = Q (A, F2 ) .

Let A0 ≡ l1 l2 and h ≡ AA0 . Then

Q (A, F1 ) Q (A, F2 )
s (l1 , h) = = = s (l2 , h)
Q (A, A0 ) Q (A, A0 )

so that h is indeed a vertex bisector of l1 l2 . In particular l1 l2 is a square vertex by the

Vertex bisector theorem.
 7

Triple spread formula

This chapter introduces the Triple spread formula, the analogue in rational
trigonometry to the fact that the sum of the angles in a triangle in the decimal number
field is 180◦ . It differs from the Triple quad formula only by a single cubic term. There
are a number of related results, including the Triple cross formula and the Triple twist
formula, and an extension to four spreads. A particular application is the Equal
spreads theorem, in which the logistic map of chaos theory makes an appearance.

7.1 Triple spread formula

Theorem 47 (Triple spread formula) Suppose that l1 , l2 and l3 are non-null

lines, with s1 ≡ s (l2 , l3 ), s2 ≡ s (l1 , l3 ) and s3 ≡ s (l1 , l2 ). Then
¡ ¢
(s1 + s2 + s3 )2 = 2 s21 + s22 + s23 + 4s1 s2 s3 .

Proof. If at least two of the lines are parallel, then one of the spreads is zero, and the
other two spreads must be equal. In this case the formula holds automatically. If the
three lines are concurrent, then replace one of the lines with a line parallel to it, to
obtain a triangle A1 A2 A3 with spreads s1 , s2 and s3 and quadrances Q1 , Q2 and Q3 .

From the Spread law, there is a non-zero number D such that

s1 s2 s3 1
= = ≡ . (7.1)
Q1 Q2 Q3 D
The Cross law can be expressed as
2
(Q1 + Q2 − Q3 ) = 4Q1 Q2 (1 − s3 ) .

89
90 7. TRIPLE SPREAD FORMULA

Rewrite this as
¡ ¢
(Q1 + Q2 + Q3 )2 = 2 Q21 + Q22 + Q23 + 4Q1 Q2 s3 . (7.2)
Use (7.1) to replace Q1 by s1 D, Q2 by s2 D and Q3 by s3 D in (7.2), and then divide by
D2 . The result is
2 ¡ ¢
(s1 + s2 + s3 ) = 2 s21 + s22 + s23 + 4s1 s2 s3 .

Definition The Triple spread function S is defined by

¡ ¢
S (a, b, c) ≡ (a + b + c)2 − 2 a2 + b2 + c2 − 4abc.

Note that S (a, b, c) is a symmetric function of a, b and c. This motivates and justifies
the following definition, in analogy with the definition of a quad triple (page 64).

Definition A set {a, b, c} of numbers is a spread triple precisely when

S (a, b, c) = 0.

Exercise 7.1 In analogy with Exercise 5.7, show that

S (a, b, c) = A (a, b, c) − 4abc
¡ ¢
= 2 (ac + bc + ab) − a2 + b2 + c2 − 4abc
2
= 4 (ab + bc + ca) − (a + b + c) − 4abc
= 4 (1 − a) (1 − b) (1 − c) − (a + b + c − 2)2
¯ ¯
¯0 a b 1¯
¯ ¯
¯a 0 c 1¯
= −¯ ¯ ¯. ¦
¯
¯ b c 0 1¯
¯1 1 1 2¯

Exercise 7.2 Demonstrate the following identity, of importance for the Triangle
spread rules (page 219)
S (a, b, c) = ((a + b − c) c + (c − a + b) (c − b + a)) (1 − c)
−c (c − (a + b)) (1 − (a + b)) . ¦

Exercise 7.3 Show that if {a, b, 0} is a spread triple then a = b. Show that if {a, b, 1}
is a spread triple then a + b = 1. ¦

Exercise 7.4 Show that in the rational or decimal number fields

0 ≤ S (a, a, a)
precisely when a ≤ 3/4. ¦
7.1. TRIPLE SPREAD FORMULA 91

Exercise 7.5 Show that for any numbers u and v with u2 6= −1 and v 2 6= −1,
( )
2
u2 v2 (u − v)
, ,
1 + u2 1 + v 2 (1 + u2 ) (1 + v 2 )
is a spread triple. Give an example where not every spread triple is of this form. ¦

Exercise 7.6 Show that as a quadratic equation in s3 , the Triple spread formula is
s23 − 2s3 (s1 + s2 − 2s1 s2 ) + (s1 − s2 )2 = 0.
Show that it can also be written as
2
(s3 − s1 + s2 ) = 4 (1 − s1 ) s2 s3
(s3 − s2 + s1 )2 = 4s1 (1 − s2 ) s3
(s3 − s1 − s2 )2 = 4s1 s2 (1 − s3 )
and that in normal form it is

2
(s3 − (s1 + s2 − 2s1 s2 )) = 4s1 s2 (1 − s1 ) (1 − s2 ) . ¦

The next result shows how to determine a spread triple if one of the elements and the
ratio of the other two is known.

Theorem 48 (Spread from ratio) Suppose that {s1 , s2 , s3 } is a spread triple with
s2 , s3 both non-zero and s1 /s2 ≡ a. Then a (1 − s3 ) is a square, and if a (1 − s3 ) ≡ r2
for some r, then
s2 = s3 / (a + 1 ± 2r) .

Proof. Use the previous exercise to write the Triple spread formula in the form
2
(s3 − s1 − s2 ) = 4s1 s2 (1 − s3 ) .
By assumption s2 is non-zero, so divide by s22 to get
µ ¶2
s3
− a − 1 = 4a (1 − s3 ) .
s2
Thus a (1 − s3 ) is a square, so that
a (1 − s3 ) ≡ r2
for some r. Then
s3 /s2 = a + 1 ± 2r
is non-zero, so that
s2 = s3 / (a + 1 ± 2r) .
92 7. TRIPLE SPREAD FORMULA

7.2 Triple cross formula

A corresponding relationship to the Triple spread formula exists for three crosses.

Theorem 49 (Triple cross formula) Suppose that l1 , l2 and l3 are non-null lines,
with c1 ≡ c (l2 , l3 ), c2 ≡ c (l1 , l3 ) and c3 ≡ c (l1 , l2 ) . Then
2
(c1 + c2 + c3 − 1) = 4c1 c2 c3 .

Proof. Use the Spread plus cross theorem to substitute s1 = 1 − c1 , s2 = 1 − c2 and

s3 = 1 − c3 into the Triple spread formula written as
(s1 + s2 + s3 − 2)2 = 4 (1 − s1 ) (1 − s2 ) (1 − s3 ) .

Example 7.1 In the complex number field, the Cross law shows that a triangle with
quadrances Q1 ≡ 1, Q2 ≡ −1 and Q3 ≡ i has crosses
2
(Q2 + Q3 − Q1 ) 3
c1 = =1+ i
4Q2 Q3 4
(Q3 + Q1 − Q2 )2 3
c2 = =1− i
4Q3 Q1 4
2
(Q1 + Q2 − Q3 ) 1
c3 = = .
4Q1 Q2 4
Both sides of the Triple cross formula become 25/16. ¦

Exercise 7.7 Show that as a quadratic equation in c3 , the Triple cross formula in
normal form is
2
(c3 + c1 + c2 − 2c1 c2 − 1) = 4c1 c2 (1 − c1 ) (1 − c2 )
or also incorporating the spreads,
2
(c3 − (c1 c2 + s1 s2 )) = 4c1 c2 s1 s2 . ¦

Exercise 7.8 Following Exercises 5.7 and 7.1, show that

¯ ¯
¯ 1 c1 c2 1¯¯
¯
¯c 1 c3 1¯¯
(c1 + c2 + c3 − 1) − 4c1 c2 c3 = − ¯¯ 1
2
. ¦
c c
¯ 2 3 1 1¯¯
¯1 1 1 2¯

Exercise 7.9 Show that in the rational and decimal number fields, the crosses c1 , c2
and c3 of any triangle satisfy
c1 c2 c3 ≤ 1/64 and 3/4 ≤ c1 + c2 + c3 ≤ 5/4. ¦
7.3. TRIPLE TWIST FORMULA 93

7.3 Triple twist formula

The corresponding relation between the three twists of a triangle is somewhat more
complicated. It will be seen to be closely connected to the addition of velocities in
Einstein’s special theory of relativity (page 242). A simpler version of this formula is
suggested by Exercise 7.12 below, and realized by the Triple turn formula (page 115).

Theorem 50 (Triple twist formula) Suppose that l1 , l2 and l3 are lines, with the
twists t1 ≡ t (l2 , l3 ), t2 ≡ t (l1 , l3 ) and t3 ≡ t (l1 , l2 ). Then

(t1 + t2 + t3 − t1 t2 t3 )2 = 4 (t1 t2 + t1 t3 + t2 t3 + 2t1 t2 t3 ) .

Proof. Suppose first that all the lines are non-null lines. Then use the Triple cross
formula
2
(c1 + c2 + c3 − 1) = 4c1 c2 c3
and the relation c = 1/ (1 + t) between a cross and twist to get
µ ¶2
1 1 1 4
+ + −1 = .
1 + t1 1 + t2 1 + t3 (1 + t1 ) (1 + t2 ) (1 + t3 )

Expand this to get

t21 + t22 + t23 + t21 t22 t23 = 2t1 t2 + 2t2 t3 + 2t1 t3

+2t1 t2 t3 (t1 + t2 + t3 ) + 8t1 t2 t3

and rewrite as

(t1 + t2 + t3 − t1 t2 t3 )2 = 4 (t1 t2 + t1 t3 + t2 t3 + 2t1 t2 t3 ) .

If one of the lines, say l3 , is a null line, then by the proof of the Twist ratio theorem
t1 = t2 = −1 and the required formula becomes the equation 4 = 4.

Exercise 7.10 Rewrite the Triple twist formula as a quadratic equation in t3 in

normal form. ¦

Exercise 7.11 Show that the Triple twist formula can be rewritten as
2 2
(t1 + t2 − t3 − t1 t2 t3 ) = 4t1 t2 (1 + t3 ) . ¦

Exercise 7.12 Show that if three numbers r1 , r2 and r3 satisfy

r1 + r2 + r3 = r1 r2 r3

and t1 ≡ r12 , t2 ≡ r22 and t3 ≡ r32 , then {t1 , t2 , t3 } satisfy the Triple twist formula. ¦
94 7. TRIPLE SPREAD FORMULA

7.4 Equal spreads

Theorem 51 (Equal spreads) If l1 , l2 and l3 are non-null lines with

s (l1 , l2 ) = s (l2 , l3 ) ≡ s as in either of the diagrams, then s (l1 , l3 ) = 0 or

s (l1 , l3 ) = 4s (1 − s) .

l3

l2 l2 s
s
r s r l1
l1 s
l3

Figure 7.1: Equal spreads theorem

Proof. If s (l1 , l3 ) ≡ r then from the Triple spread formula

2 ¡ ¢
(2s + r) = 2 2s2 + r2 + 4s2 r
which becomes
r (r − 4s (1 − s)) = 0.
Thus r = 0 or r = 4s (1 − s).

In the decimal number field the quadratic polynomial function

S2 (s) ≡ 4s (1 − s)
is the logistic map, and is important in chaos theory. This is an example of a spread
polynomial (see page 102). The graph of S2 (s) for s near the range 0 ≤ s ≤ 1 is given
in Figure 7.2.

1.4

1.2

0.8

0.6

0.4

0.2

-0.5 0.5 1
-0.2

-0.4

Figure 7.2: Logistic map S2 (s) ≡ 4s (1 − s)

7.5. SPREAD REFLECTION THEOREM 95

7.5 Spread reflection theorem

Recall the definition of the reflection lineation Σl for a non-null line l on page 54.

Theorem 52 (Spread reflection) If l1 , l2 and l are concurrent non-null lines with

l1 and l2 distinct, then Σl (l1 ) = l2 precisely when

s (l1 , l) = s (l2 , l) .

C R
l2 Q F
B
l Q
l1 n

Figure 7.3: Spread reflection theorem

Proof. Suppose that l1 , l2 and l are non-null lines intersecting at A, with l1 and l2
distinct. The reflection Σl (l1 ) of the line l1 in the line l was defined in terms of the
reflection σ l in l of points. Suppose that l2 = Σl (l1 ). If B 6= A lies on l1 , then by
definition l2 is the line passing through A and C ≡ σ l (B).

If n is the altitude from B to l, with foot F , then define the quadrances

Q ≡ Q (B, F ) = Q (F, C) and R ≡ Q (A, F )
as in Figure 7.3. Then by the Spread ratio theorem (page 77)
s (l1 , l) = Q/ (Q + R) = s (l2 , l) .
Conversely suppose that s (l1 , l) = s (l2 , l) ≡ s. Then s 6= 1, since otherwise l1 and l2
would both be perpendicular to l, which is impossible. Thus the two twists t (l1 , l) and
t (l, l2 ) are both equal to t ≡ s/ (1 − s) . If B is any point on l1 other than A, F the foot
of the altitude n from B to l and C ≡ nl2 , then by the Twist ratio theorem (page 78)
Q (B, F ) Q (F, C)
t= = .
Q (F, A) Q (F, A)
Thus
Q (B, F ) = Q (F, C) .
But then by the Midpoint theorem, (page 60) F is the midpoint of BC, so that
C = σ l (B) . Thus l2 = Σl (l1 ).
96 7. TRIPLE SPREAD FORMULA

7.6 Examples using different fields

To gain confidence in the formulas of this chapter, let’s verify them in particular cases
involving different fields. This will also bring out some of the rich number theoretical
possibilities in the subject.

Example 7.2 Over a general field, consider the triangle A1 A2 A3 where

A1 ≡ [1, 1] A2 ≡ [5, 2] A3 ≡ [3, −1] .
The quadrances are Q1 = 13, Q2 = 8 and Q3 = 17 and so the quadrea is
2 ¡ ¢
A = (13 + 8 + 17) − 2 132 + 82 + 172 = 400.
Use the Quadrea spread theorem to obtain the spreads s1 = 25/34, s2 = 100/221 and
s3 = 25/26, so that the crosses are c1 = 9/34, c2 = 121/221 and c3 = 1/26. Check the
Spread law, the Triple spread and Triple cross formulas as follows
25/34 100/121 25/26 25
= = =
13 8 17 442
225 625 ¡ ¢
(s1 + s2 + s3 )2 = = 2 s21 + s22 + s23 + 4s1 s2 s3
48 841
2 1089
(c1 + c2 + c3 − 1) = = 4c1 c2 c3 .
48 841
2 2
Since 221 = 13 × 17, 442 = 2 × 13 × 17 and 48 841 = (13) × (17) , these formulas are
valid in any field not of characteristic 13 or 17. If in addition the field is not of
characteristic 3 or 11, the crosses are all non-zero, and so the twists are t1 = 25/9,
t2 = 100/121 and t3 = 25. The Triple twist formula is then
902 500
(t1 + t2 + t3 − t1 t2 t3 )2 = = 4 (t1 t2 + t1 t3 + t2 t3 + 2t1 t2 t3 ) . ¦
1089
Example 7.3 In the field F11 consider the triangle A1 A2 A3 where
A1 ≡ [1, 1] A2 ≡ [3, 4] A3 ≡ [9, 5] .
The lines of the triangle are l1 ≡ h10 : 6 : 1i (small black circles), l2 ≡ h1 : 9 : 1i (large
open circles) and l3 ≡ h8 : 2 : 1i (gray boxes) as shown in Figure 7.4.

10
9
8
7
6
5
4
3
2
1
0
0 1 2 3 4 5 6 7 8 9 10

Figure 7.4: A triangle in F11

7.6. EXAMPLES USING DIFFERENT FIELDS 97

The quadrances are Q1 = 4, Q2 = 3 and Q3 = 2, so the quadrea is

2
A = (4 + 3 + 2) − 2 (16 + 9 + 4) = 1.

The Quadrea spread theorem gives the spreads s1 = 6, s2 = 10 and s3 = 3. The Spread
law takes the form
6/4 = 10/3 = 3/2 = 7.

The Triple spread formula is

¡ ¢
(s1 + s2 + s3 )2 = 9 = 2 s21 + s22 + s23 + 4s1 s2 s3 . ¦

Example 7.4 In the complex number field, consider the triangle A1 A2 A3 where

A1 ≡ [0, 0] A2 ≡ [3, −i] A3 ≡ [1 + i, 2i] .

The quadrances are

Q1 = (−2 + i)2 + (3i)2 = −6 − 4i

2 2
Q2 = (1 + i) + (2i) = −4 + 2i
2
Q3 = (−3) + i2 = 8.

and the quadrea is

³ ´
2 2 2 2
A = (−6 − 4i − 4 + 2i + 8) − 2 (−6 − 4i) + (−4 + 2i) + (8)
= −192 − 56i.

Compute the spreads, using for example the Quadrea spread theorem,
41 19 43 27
s1 = 40 + 20 i s2 = 52 − 104 i s3 = − 199
130 −
16
65 i

and verify the Spread law

41
40+ 19
20 i
43
− 27 i − 199 − 16
65 i
= 52 104 = 130
−6 − 4i −4 + 2i 8
199 2
=− − i.
1040 65
The Triple spread formula is

398 38 577 ¡ ¢
(s1 + s2 + s3 )2 = − + i = 2 s21 + s22 + s23 + 4s1 s2 s3 .
4225 135 200
The Triple twist formula is

1293 148 389 312

(t1 + t2 + t3 − t1 t2 t3 )2 = − + i = 4 (t1 t2 + t1 t3 + t2 t3 + 2t1 t2 t3 ) . ¦
4372 281 1457 427
98 7. TRIPLE SPREAD FORMULA

7.7 Quadruple spread formula

The next theorems extend the Triple spread and Triple cross formulas to four lines.
They should be compared to the results of Section 5.7.

Theorem 53 (Two spread triples) Suppose that {a, b, x} and {c, d, x} are both
spread triples. Then a, b, c and d satisfy the condition
µ 2 ¡ ¢ ¶2
(a + b + c + d) − 2 a2 + b2 + c2 + d2
−4 (abc + abd + acd + bcd) + 8abcd
= 64abcd (1 − a) (1 − b) (1 − c) (1 − d) .

Furthermore if a + b − 2ab 6= c + d − 2cd then

(a − b)2 − (c − d)2
x= .
2 (a + b − c − d − 2ab + 2cd)

Proof. Suppose that {a, b, x} and {c, d, x} are both spread triples. Then the Triple
spread formula gives quadratic equations in x which may be written as
2
(x − a − b + 2ab) = 4ab (1 − a) (1 − b)
(x − c − d + 2cd)2 = 4cd (1 − c) (1 − d) .

Then the Quadratic compatibility theorem (page 33) states that a, b, c and d must
satisfy
µ ¶2
(a + b − 2ab − c − d + 2cd)2
−4 (ab (1 − a) (1 − b) + cd (1 − c) (1 − d))
= 64abcd (1 − a) (1 − b) (1 − c) (1 − d) .

This compatibility condition may be rewritten more symmetrically as

µ ¡ ¢ ¶2
(a + b + c + d)2 − 2 a2 + b2 + c2 + d2
−4 (abc + abd + acd + bcd) + 8abcd
= 64abcd (1 − a) (1 − b) (1 − c) (1 − d) .

The same theorem also asserts that if a + b − 2ab 6= c + d − 2cd then

(a + b − 2ab) + (c + d − 2cd) 4ab (1 − a) (1 − b) − 4cd (1 − c) (1 − d)
x= −
2 2 (a + b − c − d − 2ab + 2cd)
(a − b)2 − (c − d)2
= .
2 (a + b − c − d − 2ab + 2cd)
7.7. QUADRUPLE SPREAD FORMULA 99

Definition The Quadruple spread function R (a, b, c, d) is defined for numbers

a, b, c and d by
µ ¡ ¢ ¶2
(a + b + c + d)2 − 2 a2 + b2 + c2 + d2
R (a, b, c, d) ≡
−4 (abc + abd + acd + bcd) + 8abcd
− 64abcd (1 − a) (1 − b) (1 − c) (1 − d) .

Theorem 54 (Quadruple spread formula) Suppose l1 , l2 , l3 and l4 are non-null

lines with sij = s (li , lj ) for all i, j = 1, 2, 3 and 4. Then

R (s12 , s23 , s34 , s14 ) = 0.

Furthermore

(s12 − s23 )2 − (s34 − s14 )2

s13 =
2 (s12 + s23 − s34 − s14 − 2s12 s23 + 2s34 s14 )
2 2
(s23 − s34 ) − (s12 − s14 )
s24 =
2 (s23 + s34 − s12 − s14 − 2s23 s34 + 2s12 s14 )

provided the denominators are non-zero.

l3
l2
s 34
s 23
s 24 l1
s 13
s14
s 12 l4

Figure 7.5: Four lines and their spreads

Proof. Since {s12 , s23 , s13 } and {s14 , s34 , s13 } are both spread triples, and
{s23 , s34 , s24 } and {s12 , s14 , s24 } are also both spread triples, the formulas follow from
the Two spread triples theorem.

The Quadruple spread formula expresses s13 and s24 in terms of the four spreads s12 ,
s23 , s34 and s14 . Note that by symmetry it can also be restated to express s12 and s34
in terms of the four spreads s13 , s14 , s23 and s24 , or to express s14 and s23 in terms of
the four spreads s12 , s13 , s24 and s34 .
100 7. TRIPLE SPREAD FORMULA

Theorem 55 (Quadruple cross formula) Suppose l1 , l2 , l3 and l4 are non-null

lines with cij ≡ c (li , lj ) for all i, j = 1, 2, 3 and 4. Then

R (c12 , c23 , c34 , c14 ) = 0.

Proof. The polynomial identity

¡ ¢
(a + b + c + d)2 − 2 a2 + b2 + c2 + d2
− 4 (abc + abd + acd + bcd) + 8abcd
= 4abcd + 4 (1 − a) (1 − b) (1 − c) (1 − d)
− (a + b + c + d − 2)2

has the property that the right hand side is unchanged if each of the numbers a, b, c
and d is replaced by 1 − a, 1 − b, 1 − c and 1 − d respectively. That means the same is
true for the left hand side, so that the Quadruple spread formula

R (s12 , s23 , s34 , s14 ) = 0

remains true if each spread sij is replaced by the corresponding cross cij = 1 − sij .

Note: This has exactly the same form as the Quadruple spread formula! This is an
unexpected symmetry, not occurring in the analogous Triple spread and Triple cross
formulas.

Exercise 7.13 Show that if {a, b, x} and {b, c, x} are spread triples with a 6= c and
b 6= 1/2 then
a + c − 2b
x= . ¦
2 (1 − 2b)

Exercise 7.14 Suppose l1 , l2 , l3 and l4 are non-null lines with cij ≡ c (li , lj ) for all
i, j = 1, 2, 3 and 4. Then show that

(c12 − c23 )2 − (c34 − c14 )2

c13 =
2 (c12 + c23 − c34 − c14 − 2c12 c23 + 2c34 c14 )
2 2
(c23 − c34 ) − (c12 − c14 )
c24 = . ¦
2 (c23 + c34 − c12 − c14 − 2c23 c34 + 2c12 c14 )

Exercise 7.15 (Harder) Is there a Quintuple spread formula? Generalize. ¦

 8

Spread polynomials

Spread polynomials are the rational equivalents of the Chebyshev polynomials of the
first kind, to which they are closely related. However they are defined in any field,
suggesting an underlying ‘universal’ phenomenon behind aspects of the theory of
special functions. In particular there is an interesting appearance of orthogonality in
finite fields. The analogous cross polynomials are also discussed.

8.1 Combining equal spreads

Theorem 56 (Three equal spreads) Suppose that the lines l0 , l1 , l2 and l3 satisfy
s (l0 , l1 ) = s (l1 , l2 ) = s (l2 , l3 ) ≡ s, and that l0 and l2 are not parallel. Then
2
s (l0 , l3 ) = s or s (l0 , l3 ) = s (3 − 4s) .

Proof. By the Equal spreads theorem (page 94), s (l0 , l2 ) = 4s (1 − s).

If r ≡ s (l0 , l3 ), then {4s (1 − s) , s, r} is a spread triple so that
2 l3
(4s (1 − s) + r − s) = 16s (1 − s) r (1 − s) .
l2
Simplify and factor to obtain
³ ´ s l1
r − s (3 − 4s)2 (r − s) = 0 s
r s
l0
so that r = s or
2
r = s (3 − 4s) .

101
102 8. SPREAD POLYNOMIALS

Exercise 8.1 Show that if the lines l0 , l3 and l4 form spreads s (l3 , l4 ) = s and
2
s (l0 , l3 ) = s (3 − 4s) , then

s (l0 , l4 ) = 4s (1 − s) or s (l0 , l4 ) = 16s (1 − s) (1 − 2s)2 . ¦

8.2 Spread polynomials

The above pattern gives rise to an important family of polynomials. The following
recursive definition shows that these polynomials are defined with integer coefficients,
so have an unambiguous meaning in any field, not of characteristic two.

Definition The spread polynomial Sn (s) is defined recursively by

S0 (s) ≡ 0 and S1 (s) ≡ s

and the rule

Sn (s) ≡ 2 (1 − 2s) Sn−1 (s) − Sn−2 (s) + 2s.

The coefficient of sn in Sn (s) is a power of four, so the degree of the polynomial Sn (s)
is n.

Theorem 57 (Recursive spreads) The polynomials Sn (s) have the property that
{Sn−1 (s) , s, Sn (s)} is a spread triple for any n ≥ 1 and any number s.

Proof. Fix a number s and use induction on n. For n = 1 the statement is immediate.
For a general n ≥ 1, {Sn−1 (s) , s, Sn (s)} is a spread triple precisely when
2 ¡ 2 ¢
(Sn−1 (s) + s + Sn (s)) = 2 Sn−1 (s) + s2 + Sn2 (s) + 4sSn−1 (s) Sn (s) (8.1)

while {Sn (s) , s, Sn+1 (s)} is a spread triple precisely when

¡ ¢
(Sn (s) + s + Sn+1 (s))2 = 2 Sn2 (s) + s2 + Sn+12
(s) + 4sSn (s) Sn+1 (s) . (8.2)

The difference between (8.1) and (8.2) can be rearranged and factored as

(Sn+1 (s) − Sn−1 (s)) (Sn+1 (s) − 2 (1 − 2s) Sn (s) + Sn−1 (s) − 2s) = 0.

Thus (8.2) follows from (8.1) if

Sn+1 (s) = 2 (1 − 2s) Sn (s) − Sn−1 (s) + 2s.

This is the induction step of the proof.

8.2. SPREAD POLYNOMIALS 103

Theorem 58 (Consecutive spreads) Suppose that l0 and l1 are intersecting

non-null lines with s (l0 , l1 ) ≡ s, and define ln+1 ≡ Σln (ln−1 ) for n = 1, 2, 3, · · · . Then
for any n = 0, 1, 2, · · ·
s (l0 , ln ) = Sn (s) .

Proof. First note by the Spread reflection theorem (page 95) that

s = s (l0 , l1 ) = s (l1 , l2 ) = s (l2 , l3 ) = · · · .

To show that s (l0 , ln ) = Sn (s) for all n = 0, 1, 2, · · · proceed by induction on n, with

the cases n = 0 and 1 immediate.

For a general n ≥ 1 assume that s (l0 , lk ) = Sk (s) for all k ≤ n. Let s (l0 , ln+1 ) ≡ r.
Then apply the Triple spread formula to the spreads made by the lines l0 , ln and ln+1
to deduce that {Sn (s) , s, r} is a spread triple.

But by the Recursive spreads theorem

{Sn (s) , s, Sn−1 (s)} and {Sn (s) , s, Sn+1 (s)}

are both spread triples. Thus since a quadratic equation has at most two solutions,
r = Sn+1 (s) or r = Sn−1 (s) . If r = Sn+1 (s) then the induction is complete.

If r = Sn−1 (s) then

r = s (l0 , ln−1 ) = s (l0 , ln+1 )
so l0 is a bisector of the vertex ln−1 ln+1 , as is ln . By the Vertex bisector theorem (page
85), l0 and ln are then either parallel or perpendicular, so that Sn (s) is either 0 or 1.
Let’s consider both cases.

If Sn (s) = 0 then {0, s, r} is a spread triple, and by the Recursive spreads theorem, so
is {0, s, Sn+1 (s)} . Then by Exercise 7.3

r = s = Sn+1 (s) .

If Sn (s) = 1 then {1, s, r} is a spread triple, as is {1, s, Sn+1 (s)}. Then by the same
exercise,
1 = r + s = Sn+1 (s) + s
so that again r = Sn+1 (s).

Thus if r = Sn−1 (s) then r = Sn+1 (s). So s (l0 , ln+1 ) = Sn+1 (s) and the induction is
complete.

Definition The order of the spread s is the least natural number n ≥ 1 such that
sn ≡ Sn (s) = 0, if it exists.
104 8. SPREAD POLYNOMIALS

Definition If the order of the spread s is n, then the spread sequence of s is the
list
[0, s1 , s2 , · · · , sn−1 ]
where sk ≡ Sk (s), for k = 1, · · · , n − 1.

Example 8.1 In F13 the non-zero spread numbers are 1, 4, 6, 7, 8 and 10. The order of
s ≡ 6 is 12, and the spread sequence of s is
[0, 6, 10, 7, 4, 8, 1, 8, 4, 7, 10, 6] .
Since this spread sequence contains all the spreads, the other spread sequences can be
easily determined from this one by stepping through the sequence in a circular fashion.
This is a consequence of the Cyclic reflection theorem (page 56). (The situation is
analogous to that of subgroups of the cyclic group with twelve elements). For example,
the spread sequence for s ≡ 10 is
[0, 10, 4, 1, 4, 10] . ¦

Example 8.2 In F19 the non-zero spread numbers are 1, 2, 4, 8, 9, 10, 11, 12, 16 and 18.
The order of s ≡ 8 is 20 and the spread sequence of s is
[0, 8, 4, 2, 9, 10, 11, 18, 16, 12, 1, 12, 16, 18, 11, 10, 9, 2, 4, 8] .
As in the previous example, all other spread sequences can be determined easily from
this one by stepping cyclically through it. For example the spread sequence for s ≡ 4 is
[0, 4, 9, 11, 16, 1, 16, 11, 9, 4]
while that for s ≡ 9 (particularly important for five-fold symmetry) is
[0, 9, 16, 16, 9] . ¦

8.3 Special cases

The first few spread polynomials are

S0 (s) = 0
S1 (s) = s
S2 (s) = 4s − 4s2 = 4s (1 − s)
2
S3 (s) = 9s − 24s2 + 16s3 = s (3 − 4s)
S4 (s) = 16s − 80s2 + 128s3 − 64s4 = 16s (1 − s) (1 − 2s)2
¡ ¢2
S5 (s) = 25s − 200s2 + 560s3 − 640s4 + 256s5 = s 5 − 20s + 16s2
S6 (s) = 36s − 420s2 + 1792s3 − 3456s4 + 3072s5 − 1024s6
2 2
= 4s (1 − s) (1 − 4s) (3 − 4s)
S7 (s) = 49s − 784s2 + 4704s3 − 13 440s4 + 19 712s5 − 14 336s6 + 4096s7
¡ ¢2
= s 7 − 56s + 112s2 − 64s3 .
8.3. SPECIAL CASES 105

The spread polynomials make sense in any field. In the decimal number field, here are
the graphs of S1 (s), S2 (s), S3 (s), S4 (s) and S5 (s), both separately and together, in
the interval [0, 1]. The function S2 (s) is called the logistic map in chaos theory.

1.4
1.2
1.2 1.4
1 1.2
1
0.8 1
0.8
0.6 0.8
0.6
0.6
0.4 0.4 0.4
0.2 0.2 0.2

-0.5 0.5 1 -0.5 0.5 1 -0.5 0.5 1

-0.2 -0.2 -0.2

-0.4 -0.4
-0.4

1.4
1.4
1.2
1.2 1.2

1 1
1

0.8 0.8 0.8

0.6 0.6 0.6

0.4 0.4 0.4
0.2 0.2 0.2

-0.5 0.5 1 -0.5 0.5 1 -0.5 0.5 1

-0.2 -0.2 -0.2
-0.4 -0.4 -0.4

Example 8.3 Here is a table of values of spread polynomials in the field F5 . Recall
that the spread numbers in this field are 0, 1 and 3.
s 0 1 2 3 4
S0 (s) 0 0 0 0 0
S1 (s) 0 1 2 3 4
S2 (s) 0 0 2 1 2
S3 (s) 0 1 0 3 1
S4 (s) 0 0 2 0 2
S5 (s) 0 1 2 3 4
S6 (s) 0 0 0 1 0
S7 (s) 0 1 2 3 4
S8 (s) 0 0 2 0 2
S9 (s) 0 1 0 3 1
S10 (s) 0 0 2 1 2
S11 (s) 0 1 2 3 4
S12 (s) 0 0 0 0 0

The pattern repeats with period 12, that is Sn (s) = Sn+12 (s) for all n and for all s.
However for values of s which are spread numbers, Sn (s) = Sn+4 (s) for all n. ¦
106 8. SPREAD POLYNOMIALS

Example 8.4 Here is a table of values of spread polynomials in the field F7 . Recall
that the spread numbers in this field are 0, 1, 3, 4 and 5.

s 0 1 2 3 4 5 6 s 0 1 2 3 4 5 6
S0 (s) 0 0 0 0 0 0 0 S13 (s) 0 1 2 5 4 3 6
S1 (s) 0 1 2 3 4 5 6 S14 (s) 0 0 6 4 1 4 6
S2 (s) 0 0 6 4 1 4 6 S15 (s) 0 1 1 3 4 5 0
S3 (s) 0 1 1 5 4 3 0 S16 (s) 0 0 6 0 0 0 6
S4 (s) 0 0 6 1 0 1 6 S17 (s) 0 1 2 3 4 5 6
S5 (s) 0 1 2 5 4 3 6 S18 (s) 0 0 0 4 1 4 0
S6 (s) 0 0 0 4 1 4 0 S19 (s) 0 1 2 5 4 3 6
S7 (s) 0 1 2 3 4 5 6 S20 (s) 0 0 6 1 0 1 6
S8 (s) 0 0 6 0 0 0 6 S21 (s) 0 1 1 5 4 3 0
S9 (s) 0 1 1 3 4 5 0 S22 (s) 0 0 6 4 1 4 6
S10 (s) 0 0 6 4 1 4 6 S23 (s) 0 1 2 3 4 5 6
S11 (s) 0 1 2 5 4 3 6 S24 (s) 0 0 0 0 0 0 0
S12 (s) 0 0 0 1 0 1 0 S25 (s) 0 1 2 3 4 5 6

The pattern repeats with period 24, that is Sn (s) = Sn+24 (s) for all n and for all s.
For values of s which are spread numbers, Sn (s) = Sn+8 (s) for all n. ¦

Exercise 8.2 (S. Goh) Show that in the field Fp , if k = (p − 1) (p + 1) /2 then

Sn (s) = Sn+k (s)
for all n and for all s. ¦

Exercise 8.3 In the rational or decimal number fields, show that the spread
polynomials satisfy 0 ≤ Sn (s) ≤ 1 for 0 ≤ s ≤ 1 and all n. ¦

Exercise 8.4 (Uses Chebyshev polynomials) In the decimal number field, show
that if Tn (x) is the nth Chebyshev polynomial of the first kind then
¡√ ¢
Sn (s) = 1 − Tn2 1 − s
1 − Tn (1 − 2s)
= . ¦
2

8.4 Explicit formulas

Theorem 59 (Spread polynomial formula) Suppose that 4s (1 − s) = r2 and

that the field contains a number i satisfying i2 = −1. Then for any n = 0, 1, 2, · · ·

2 − (1 − 2s + ir)n − (1 − 2s − ir)n
Sn (s) = .
4
8.4. EXPLICIT FORMULAS 107

Proof. Use induction on n. For n = 0 and n = 1 the formula in the theorem agrees
with the values S0 (s) = 0 and S1 (s) = s. Assume that the formula holds for values up
to and including n. Then set
C ≡ 1 − 2s + ir D ≡ 1 − 2s − ir
so that
Sn+1 (s) ≡ 2 (1 − 2s) Sn (s) − Sn−1 (s) + 2s
µ ¶
2 − (1 − 2s + ir) C n−1 − (1 − 2s − ir) Dn−1
= 2 (1 − 2s)
4
µ n−1 n−1
¶
2−C −D
− + 2s
4
1 C n−1 ¡ ¢
= − 1 − 8s + 8s2 + 2ir − 4irs
2 4
Dn−1 ¡ ¢
− 1 − 8s + 8s2 − 2ir + 4irs .
4
But the assumption on r means that
2
(1 − 2s + ir) = 1 − 8s + 8s2 + 2ir − 4irs = C 2
2
(1 − 2s − ir) = 1 − 8s + 8s2 − 2ir + 4irs = D2
so that indeed
1 C n+1 Dn+1
Sn+1 (s) = − −
2 4 4
n+1 n+1
2 − (1 − 2s + ir) − (1 − 2s − ir)
= .
4

Note that even if the field does not contain a number i satisfying i2 = −1, a quadratic
extension of it does, so the formula in the previous theorem works quite generally.

Exercise 8.5 Show that in the field Fp , Sp (s) = s for all s. ¦

Exercise 8.6 Use the Spread polynomial theorem and the Binomial theorem to show
that if a ≡ 1 − 2s and b ≡ 4s (1 − s) then
µ ¶ µ ¶
n n−2 n n−4 2
1 − 2Sn (s) = an − a b+ a b − ···
2 4

ending either in (−b)l if n = 2l or na (−b)l if n = 2l + 1. ¦

The next exercise is motivated by Problem 222 in [Shklarsky-Chentzov-Yaglom].

Exercise 8.7 (Harder) Show that if n = 2l + 1 is an odd natural number then

µµ ¶ µ ¶ µ ¶ ¶2
n l n l−1 n l−2 2 l l
Sn (s) = s (1 − s) − (1 − s) s + (1 − s) s − · · · + (−1) s . ¦
1 3 5
108 8. SPREAD POLYNOMIALS

Exercise 8.8 (S. Goh) Show that S1 (s), S2 (s), S3 (s), S4 (s) · · · can be written as
¯ ¯
¯ ¯ ¯2 − 4s −1 1¯¯
¯ ¯ ¯2 − 4s 1¯ ¯
¯s¯ , ¯ ¯ ¯ 2 − 4s 1¯¯ ,
¯ −2s s¯ , ¯ −1
¯ 0 −2s s¯
¯ ¯
¯2 − 4s −1 0 1¯¯
¯
¯ −1 2 − 4s −1 1¯¯
¯ ··· . ¦
¯ 0 −1 2 − 4s 1¯¯
¯
¯ 0 0 −2s s¯

Exercise 8.9 (M. Hirschhorn, S. Goh) Show that for any natural number n

n−1
X µ ¶
n 2n − 1 − k
Sn (s) = s (−4s)n−k−1 . ¦
n−k k
k=0

Define the scaled spread polynomials Sn∗ (r) = 4Sn (r/4). As observed by M.
Hirschhorn, they can be evaluated as follows. Write Pascal’s triangle so that the usual
‘rows’ are appearing diagonally, and each entry is the sum of the number above it and
the number diagonally above it to its left, as in the left-most array below.
⎡ ⎤
1 ⎡ ⎤
⎢ 1 ⎥ 1
⎢ ⎥ ⎢ 1 ⎥
⎢1 1 ⎥ ⎢ ⎥
⎢ ⎥ ⎢3 1 ⎥
⎢ 2 1 ⎥ ⎢ ⎥
⎢ ⎥ ⎢ 4 1 ⎥
⎢1 3 1 ⎥ ⎢ ⎥
⎢ ⎥ ⎢5 5 1 ⎥
⎢ 3 4 1 ⎥ ⎢ ⎥.
⎢ ⎥ ⎢ 9 6 1 ⎥
⎢1 6 5 1 ⎥ ⎢ ⎥
⎢ ⎥ ⎢7 14 7 1 ⎥
⎢ 4 10 6 1 ⎥ ⎢ ⎥
⎢ ⎥ ⎢ 16 20 8 1 ⎥
⎢1 10 15 7 1 ⎥ ⎣ ⎦
⎣ ⎦ .. .. ..
.. .. .. . . .
. . .
Now slide Pascal’s array down two steps, add the result to the original array, and
disregard the first column, to get the spread array, as in the right-most array above.
Every entry outside of the first column of the spread array is still the sum of the entry
above it and the entry diagonally above it to the left. The first column of the spread
array contains all the odd numbers, and the second column contains all the non-zero
squares. Then those rows of the spread array beginning with the square numbers in the
second column give the coefficients of the scaled spread polynomials, beginning with r,
then r2 and so on. For example the row beginning with the third square 9 yields
S3∗ (r) = 9r − 6r2 + r3 .

Exercise 8.10 (M. Hirschhorn) Show that a generating function for the spread
polynomials is
X∞
sx (1 + x)
Sn (s) xn = 3 . ¦
n=1 (1 − x) + 4sx (1 − x)
8.5. ORTHOGONALITY 109

8.5 Orthogonality

This is a brief introduction to a phenomenon that requires more study. In the field F5
the array of values
X4
Sm (k) Sn (k) ≡ am,n
k=0

where m and n are indexed by 0, 1, 2, 3 and 4 is

m\n 0 1 2 3 4
0 0 0 0 0 0
1 0 0 0 4 2 .
2 0 0 4 0 3
3 0 4 0 1 2
4 0 2 3 2 3

In the field F7 the array of values

6
X
Sm (k) Sn (k) ≡ am,n
k=0

where m and n are indexed by 0, 1, · · · , 6 is

m\n 0 1 2 3 4 5 6
0 0 0 0 0 0 0 0
1 0 0 0 0 0 3 1
2 0 0 0 0 3 0 5 .
3 0 0 0 3 0 4 1
4 0 0 3 0 4 0 1
5 0 3 0 4 0 0 1
6 0 1 5 1 1 1 5

Exercise 8.11 (Harder) In the field Fp show that for 1 ≤ m, n ≤ p − 2

p−1
X
Sm (k) Sn (k) = 0
k=0

unless m + n = p ± 1. What are the values of the sum in these latter cases? ¦

Exercise 8.12 (Uses Chebyshev polynomials) Show that over the decimal
number field, if m 6= n then
Z 1
(Sm (s) − 1/2) (Sn (s) − 1/2)
p ds = 0
0 s (1 − s)

while if m = n the integral equals π/4. Hint: See Exercise 8.4 and Example 27.1. ¦
110 8. SPREAD POLYNOMIALS

8.6 Composition of spread polynomials

Define the composition Sn ◦ Sm of spread polynomials Sn and Sm by the rule

(Sn ◦ Sm ) (s) = Sn (Sm (s)) .

Theorem 60 (Spread composition) For any natural numbers n and m

Sn ◦ Sm = Snm .

Proof. Since the spread polynomials have integer coefficients, it suffices to prove this
over the rational number field. Suppose that s is a spread number, so that by the
Spread number theorem (page 76) there are lines l0 and l1 with s (l0 , l1 ) = s. Consider
the reflection sequence
· · · , l−2 , l−1 , l0 , l1 , l2 , · · · (8.3)
characterized by the condition that for any integer n

Σln (ln−1 ) = ln+1 .

The Consecutive spreads theorem (page 103) states that for any natural number k

Sk (s) = s (l0 , lk )

and in particular for any natural numbers n and m,

Snm (s) = s (l0 , lnm ) .

Suppose that Sm (s) = s (l0 , lm ) ≡ r. Now consider the two sided subsequence

· · · , l−2m , l−m , l0 , lm , l2m , · · · (8.4)

of (8.3). The Cyclic reflection theorem (page 56) shows that

Σlm (l0 ) = l2m Σl2m (lm ) = l3m

and so on, so that more generally

Σlkm (ljm ) = l(2k−j)m .

This means that (8.4) is itself a reflection sequence with common spread r. But then

Snm (s) = s (l0 , lnm ) = Sn (r) = Sn (Sm (s)) = (Sn ◦ Sm ) (s) .

Since (Sn ◦ Sm ) (s) and Snm (s) agree for any spread numbers s, and there are more
spread numbers than the degree nm, the two polynomials are equal.
8.6. COMPOSITION OF SPREAD POLYNOMIALS 111

It follows that the values of Sn for prime values of n are particularly basic, and to
calculate Sn for large values of n this theorem is more effective than the recursive
definition. Another consequence is a pleasant commutativity and associativity of
composition for spread polynomials; that is

Sn ◦ Sm = Sm ◦ Sn
(Sn ◦ Sm ) ◦ Sk = Sn ◦ (Sm ◦ Sk ) .

Example 8.5 Using

S3 (s) = s (3 − 4s)2
and
¡ ¢2
S5 (s) = s 5 − 20s + 16s2
you find that
¡ ¢2 ³ ¡ ¢2 ´2
S15 (s) = S3 (S5 (s)) = s 5 − 20s + 16s2 3 − 4s 5 − 20s + 16s2
¡ ¢2 ¡ ¢2
= s (3 − 4s)2 5 − 20s + 16s2 1 − 32s + 224s2 − 448s3 + 256s4 .

In the same vein

S30 (s) = S2 (S15 (s)) = 4S15 (s) (1 − S15 (s))

¡ ¢2 ¡ ¢2
= 4s (1 − s) (3 − 4s)2 5 − 20s + 16s2 1 − 32s + 224s2 − 448s3 + 256s4
2¡ ¢2 ¡ ¢2
× (1 − 4s) 1 − 12s + 16s2 1 − 96s + 416s2 − 576s3 + 256s . ¦

Exercise 8.13 (S. Goh) Show that for every k = 1, 2, 3, · · · there is a polynomial
Φk (s) with integer coefficients of degree φ (k), where φ is Euler’s totient function, such
that for any n = 1, 2, 3, · · ·
Y
Sn (s) = Φk (s) .
k|n

The first few values of these ‘spread-cyclotomic’ polynomials are

Φ1 (s) = s Φ2 (s) = 4 (1 − s)
Φ3 (s) = (3 − 4s)2 Φ4 (s) = 4 (1 − 2s)2
¡ ¢2
Φ5 (s) = 5 − 20s + 16s2 Φ6 (s) = (1 − 4s)2
¡ ¢2 ¡ ¢2
Φ7 (s) = 7 − 56s + 112s2 − 64s3 Φ8 (s) = 4 1 − 8s + 8s2 .
¡ ¢2
So for example since also Φ12 (s) = 1 − 16s + 16s2 ,

S12 (s) = Φ1 (s) Φ2 (s) Φ3 (s) Φ4 (s) Φ6 (s) Φ12 (s)

2 2 2 ¡ ¢2
= 16s (1 − s) (3 − 4s) (1 − 2s) (1 − 4s) 1 − 16s + 16s2 . ¦
112 8. SPREAD POLYNOMIALS

8.7 Cross polynomials

Definition The cross polynomials Cn (c) are defined recursively by

C0 (c) ≡ 1 and C1 (c) ≡ c

and the rule

Cn (c) ≡ 2 (2c − 1) Cn−1 (c) − Cn−2 (c) + 2 (1 − c) .

The properties of the cross polynomials are analogous and closely related to those of
the spread polynomials. They will be developed here in the form of a number of
exercises.

Exercise 8.14 Show that the polynomials Cn (c) have the property that for any n ≥ 1
and any c, the numbers Cn−1 (c) , c and Cn (c) satisfy the Triple cross formula. ¦

Exercise 8.15 Verify that the first few cross polynomials are
C0 (c) = 1
C1 (c) = c
2
C2 (c) = 1 − 4c + 4c2 = (1 − 2c)
C3 (c) = 9c − 24c2 + 16c3 = c (3 − 4c)2
¡ ¢2
C4 (c) = 1 − 16c + 80c2 − 128c3 + 64c4 = 1 − 8c + 8c2
¡ ¢2
C5 (c) = c 5 − 20c + 16c2
¡ ¢2
C6 (c) = (1 − 2c)2 1 − 16c + 16c2 . ¦

Exercise 8.16 Show that for any natural numbers n and m

Cn (Cm (c)) = Cnm (c) . ¦

Exercise 8.17 Show that if c + s = 1 then Cn (c) + Sn (s) = 1. ¦

Exercise 8.18 Show that if n is odd then Cn (t) = Sn (t), and that if n is even then
Cn (t) = 1 − Sn (t). ¦

Exercise 8.19 Show that for any n

2
(Cn (t) − Sn (1 − t)) = C2n (t)
and
4Sn (t) Cn (1 − t) = S2n (t) .
Generalize. ¦

Exercise 8.20 Show how the coefficients of the scaled cross polynomials
Cn∗ (r) = 4Cn (r/4) can be read off from the rows of the spread array. ¦
 9

Oriented triangles and turns

There is an alternative way to develop rational trigonometry, where the fundamental

−−−−−→
object is not a triangle A1 A2 A3 , but rather an oriented triangle A1 A2 A3 . In this book,
which emphasizes geometry, the unoriented approach takes precedence, but the
oriented version is useful in applications to calculus, and is simpler in some situations.

This chapter introduces the more refined notions of oriented sides and vertices,
oriented triangles and oriented n-gons. This allows a definition of signed area, so that
the quadrea A of a triangle becomes sixteen times the square of the signed area, and
allows the extension of quadrea to n-gons.

The twist of a vertex can be written as the square of the turn of one of the associated
oriented vertices. The Triple turn formula then turns out to be linear in any of the
three turns. Instead of turn polynomials one finds turn rational functions, which are
well-known. Bretschneider’s formula for the quadrea of a quadrilateral is established.

9.1 Oriented sides, vertices and triangles

−−−→
Definition An oriented side A1 A2 is an ordered pair [A1 , A2 ] of distinct points.
−→
An oriented vertex l1 l2 is an ordered pair [l1 , l2 ] of distinct intersecting lines.

−−−−−→
Definition An oriented triangle A1 A2 A3 is a list [A1 , A2 , A3 ] of distinct
non-collinear points, with the convention that
−−−−−→ −−−−−→ −−−−−→
A1 A2 A3 ≡ A2 A3 A1 ≡ A3 A1 A2 .

113
114 9. ORIENTED TRIANGLES AND TURNS

The points A1 , A2 and A3 are the points, and the lines l1 ≡ A2 A3 , l2 ≡ A1 A3 and
−−−−−→ −−−→ −−−→ −−−→
l3 ≡ A1 A2 are the lines of the oriented triangle A1 A2 A3 . Also A1 A2 , A2 A3 and A3 A1
−−−−−→ −→ −→ −→
are the oriented sides of the oriented triangle A1 A2 A3 , while l1 l2 , l2 l3 and l3 l1 are
−−−−−→
the oriented vertices of A1 A2 A3 .
−−−→ −→
An oriented side A1 A2 determines a side A1 A2 , and similarly an oriented vertex l1 l2
−−−−−→
determines a vertex l1 l2 . An oriented triangle A1 A2 A3 determines a triangle A1 A2 A3 ,
and to every triangle A1 A2 A3 there are correspondingly two oriented triangles, namely
−−−−−→ −−−−−→
A1 A2 A3 and the opposite oriented triangle A3 A2 A1 .

−−−−−−−→
Definition An oriented quadrilateral A1 A2 A3 A4 is a cyclical list [A1 , A2 , A3 , A4 ]
of distinct points, meaning that
−−−−−−−→ −−−−−−−→ −−−−−−−→ −−−−−−−→
A1 A2 A3 A4 ≡ A2 A3 A4 A1 ≡ A3 A4 A1 A2 ≡ A4 A1 A2 A3

and such that no three consecutive points are collinear (including A3 , A4 , A1 and
A4 , A1 , A2 ).

Define lij ≡ Ai Aj for i 6= j. The points A1 , A2 , A3 and A4 are the points of the
oriented quadrilateral, and the lines l12 , l23 , l34 and l41 are the lines of the oriented
−−−→ −−−→ −−−→ −−−→
quadrilateral. Also A1 A2 , A2 A3 , A3 A4 and A4 A1 are the oriented sides of the
−−−−−−−→ −−−→ −−−→ −−−→ −−−→
oriented quadrilateral A1 A2 A3 A4 , while l12 l23 , l23 l34 , l34 l41 and l41 l12 are the oriented
−−−−−−−→
vertices of A1 A2 A3 A4 .
−−−−−−−−−→
These definitions extend to an oriented 5-gon A1 A2 A3 A4 A5 and beyond in the obvious
−−−−−−−−→
fashion. As with triangles, an oriented n-gon A1 A2 · · · An has associated to it the
n-gon A1 A2 · · · An , and to any such n-gon there are correspondingly two oriented
−−−−−−−−→ −−−−−−−−→
n-gons, namely A1 A2 · · · An and its opposite oriented n-gon An An · · · A1 .

9.2 Turns of oriented vertices

Definition If l1 ≡ ha1 : b1 : c1 i and l2 ≡ ha2 : b2 : c2 i are non-perpendicular lines

−→
then the turn of the oriented vertex l1 l2 is the number
³−→´ a1 b2 − a2 b1
u l1 l2 ≡ u (l1 , l2 ) ≡ .
a1 a2 + b1 b2

Comparing with the definition of twist t (l1 , l2 ), observe that

t (l1 , l2 ) = (u (l1 , l2 ))2 .

Note also that u (l1 , l2 ) = −u (l2 , l1 ) so that order matters with turns.
9.2. TURNS OF ORIENTED VERTICES 115

If the triangle A1 A2 A3 has points A1 ≡ [x1 , y1 ], A2 ≡ [x2 , y2 ] and A3 ≡ [x3 , y3 ], then by

the Line through two points theorem (page 38) the lines l1 ≡ A2 A3 and l2 ≡ A1 A3 are

l1 = hy2 − y3 : x3 − x2 : x2 y3 − x3 y2 i
l2 = hy1 − y3 : x3 − x1 : x1 y3 − x3 y1 i .

The turn u (l1 , l2 ) can therefore be written as

(y2 − y3 ) (x3 − x1 ) − (y1 − y3 ) (x3 − x2 )

u (l1 , l2 ) = .
(y2 − y3 ) (y1 − y3 ) + (x3 − x2 ) (x3 − x1 )

The following lovely result is closely related to the Triple twist formula (page 115), but
is clearly more elementary. This is an example where the oriented theory is simpler
than the unoriented one.

Theorem 61 (Triple turn formula) Suppose that l1 , l2 and l3 are lines, no two
perpendicular, with the turns u1 ≡ u (l2 , l3 ), u2 ≡ u (l3 , l1 ) and u3 ≡ u (l1 , l2 ). Then

u1 + u2 + u3 = u1 u2 u3 .

Proof. Suppose that l1 ≡ ha1 : b1 : c1 i, l2 ≡ ha2 : b2 : c2 i and l3 ≡ ha3 : b3 : c3 i. Then

a2 b3 − a3 b2
u1 =
a2 a3 + b2 b3
a3 b1 − a1 b3
u2 =
a3 a1 + b3 b1
a1 b2 − a2 b1
u3 = .
a1 a2 + b1 b2
The polynomial identity

(a2 b3 − a3 b2 ) (a3 a1 + b3 b1 ) (a1 a2 + b1 b2 )

+ (a3 b1 − a1 b3 ) (a2 a3 + b2 b3 ) (a1 a2 + b1 b2 )
+ (a1 b2 − a2 b1 ) (a2 a3 + b2 b3 ) (a3 a1 + b3 b1 )
= (a1 b2 − a2 b1 ) (a2 b3 − a3 b2 ) (a3 b1 − a1 b3 )

implies that
u1 + u2 + u3 = u1 u2 u3 .

Note that the Triple turn formula is linear in any one of the variables u1 , u2 and u3 , so
that for example if u1 and u2 are known then
u1 + u2
u3 = .
u1 u2 − 1
116 9. ORIENTED TRIANGLES AND TURNS

Exercise 9.1 (Triple coturn formula) If l1 ≡ ha1 : b1 : c1 i and l2 ≡ ha2 : b2 : c2 i

are non-parallel lines then define the coturn o (l1 , l2 ) to be

a1 a2 + b1 b2
o (l1 , l2 ) = .
a1 b2 − a2 b1
Show that
o1 o2 + o2 o3 + o3 o1 = 1. ¦

Exercise 9.2 Show that if u ≡ u (l0 , l1 ) and v ≡ u (l1 , l2 ) then

u+v
w ≡ u (l0 , l2 ) = .
1 − uv
Deduce that if u = v then
2u
w= . ¦
1 − u2
Exercise 9.3 Carrying on from the previous exercise, suppose that
u (l0 , l1 ) = u (l1 , l2 ) = · · · = u (ln−1 , ln ) ≡ u
for some reflection sequence of lines l0 , l1 , · · · , ln . For a natural number n, define the
turn (rational) function Un by the rule Un (u) ≡ u (l0 , ln ). Show that
U1 (u) = u
2u
U2 (u) =
1 − u2
3u − u3
U3 (u) =
1 − 3u2
4u − 4u3
U4 (u) =
1 − 6u2 + u4
5u − 10u3 + u5
U5 (u) = .
1 − 10u2 + 5u4
State and prove the general form for Un (u). ¦

9.3 Signed areas

−−−−−→
Definition The signed area a of the oriented triangle A1 A2 A3 with points
A1 ≡ [x1 , y1 ], A2 ≡ [x2 , y2 ] and A3 ≡ [x3 , y3 ] is the number
¯ ¯
³−−−−−→´ x y − x y + x y − x y + x y − x y ¯x y1 1¯
1 ¯¯ 1 ¯
= ¯x2 y2 1¯¯ .
1 2 2 1 2 3 3 2 3 1 1 3
a ≡ a A1 A2 A3 ≡
2 2 ¯x y 1¯
3 3
9.3. SIGNED AREAS 117

This formula makes sense even if the points are collinear, in which case it gives the
value 0. Note that the convention for ordering the six terms involved in a is different
than in previous chapters. If A is the quadrea of A1 A2 A3 then by the Quadrea
theorem (page 68)
A = 16a2 .

−−−−−−−−→
Theorem 62 (Signed area) For an oriented n-gon A1 A2 · · · An and a point P , the
sum ³−−−−→´ ³−−−−→´ ³−−−−−−−→´ ³−−−−−→´
a ≡ a P A1 A2 + a P A2 A3 + · · · + a P An−1 An + a P An A1

is independent of P. If Ai ≡ [xi , yi ] for i = 1, 2, · · · , n then

x1 y2 − x2 y1 + x2 y3 − x3 y2 + · · · + xn−1 yn − xn yn−1 + xn y1 − x1 yn
a= .
2

Proof. Suppose that P ≡ [x0 , y0 ] and Ai ≡ [xi , yi ] for i = 1, 2, · · · , n. Note that

³−−−−→´ 1 1 1 1
a P A1 A2 = x0 (y1 − y2 ) + y0 (x2 − x1 ) + x1 y2 − x2 y1 .
2 2 2 2
Thus in the sum
³−−−−→´ ³−−−−→´ ³−−−−−−−→´ ³−−−−−→´
a ≡ a P A1 A2 + a P A2 A3 + · · · + a P An−1 An + a P An A1

all the terms involving x0 and y0 cancel, and what is left is

x1 y2 − x2 y1 + x2 y3 − x3 y2 + · · · + xn−1 yn − xn yn−1 + xn y1 − x1 yn
a= .
2

Definition The expression a in the above theorem

³−−−−−− is the´signed area of the
−−−−−−−−→ −−→
oriented n-gon A1 A2 · · · An and is denoted a A1 A2 · · · An .

−−−−−−−→
In particular the signed area of the oriented quadrilateral A1 A2 A3 A4 with points
A1 ≡ [x1 , y1 ], A2 ≡ [x2 , y2 ], A3 ≡ [x3 , y3 ] and A4 ≡ [x4 , y4 ] is
³−−−−−−−→´ x y − x y + x y − x y + x y − x y + x y − x y
1 2 2 1 2 3 3 2 3 4 4 3 4 1 1 4
a ≡ a A1 A2 A3 A4 = . (9.1)
2
−−−−−−−→
Exercise 9.4 Show that the signed area of the oriented quadrilateral A1 A2 A3 A4 with
points
A1 ≡ [k, 0] A2 ≡ [l, 0] A3 ≡ [l, n] A4 ≡ [k, m]
is
a = (m + n) (l − k) /2. ¦
118 9. ORIENTED TRIANGLES AND TURNS

−−−−−−−−→
Exercise 9.5 Show that for any oriented n-gon A1 A2 · · · An
³−−−−−−−−→´ ³−−−−−→´ ³−−−−−→´ ³−−−−−−−→´
a A1 A2 · · · An = a A1 A2 A3 + a A1 A3 A4 + · · · + a A1 An−1 An . ¦

−−−−−−−−→
Exercise 9.6 Show that if 3 ≤ m < n, then for any oriented n-gon A1 A2 · · · An
³−−−−−−−−→´ ³−−−−−−−−→´ ³−−−−−−−−−−−−−→´
a A1 A2 · · · An = a A1 A2 · · · Am + a A1 Am Am+1 · · · An . ¦
³−−−−−−−−−−→´ ³−−−−−−−−→´
Exercise 9.7 Show that a An An−1 · · · A1 = −a A1 A2 · · · An . ¦

Definition For n ≥ 4 the quadrea A of an n-gon A1 A2 · · · An is

A ≡ 16a2
−−−−−−−−→
where a is the signed area of the oriented n-gon A1 A2 · · · An .

This is well-defined by the previous exercise, and extends the definition of quadrea of a
triangle.

Theorem 63 (Bretschneider’s formula) Suppose that A1 A2 A3 A4 is a

quadrilateral with Qij ≡ Q (Ai , Aj ) for all i, j = 1, 2, 3 and 4. Then the quadrea A of
A1 A2 A3 A4 is
2
A = 4Q13 Q24 − (Q12 + Q34 − Q23 − Q14 ) .

Proof. Suppose that A1 ≡ [x1 , y1 ], A2 ≡ [x2 , y2 ], A3 ≡ [x3 , y3 ] and A4 ≡ [x4 , y4 ], so

that
2 2
Qij = (xj − xi ) + (yj − yi )
for all i, j = 1, 2, 3 and 4. Then
2 2
(Q12 + Q34 − Q23 − Q14 ) = 4 (x1 x2 + x3 x4 − x1 x4 − x2 x3 + y1 y2 + y3 y4 − y1 y4 − y2 y3 )
while from (9.1) and the definition of quadrea,
A = 4 (x1 y2 − x2 y1 + x2 y3 − x3 y2 + x3 y4 − x4 y3 + x4 y1 − x1 y4 )2 .
Add these two expressions and factor to obtain the surprising relation
(Q12 + Q34 − Q23 − Q14 )2 + A
³ ´³ ´
2 2 2 2
= 4 (x1 − x3 ) + (y1 − y3 ) (x2 − x4 ) + (y2 − y4 )
= 4Q13 Q24 .
Thus
A = 4Q13 Q24 − (Q12 + Q34 − Q23 − Q14 )2 .
 Part III

Universal Geometry

119
 10

Triangles

Universal geometry parallels Euclidean geometry, except that all theorems in the
subject are necessarily valid over a general field, excluding characteristic two. In this
chapter, basic and useful facts about isosceles, equilateral and right triangles are
derived, and congruent and similar triangles are defined.

10.1 Isosceles triangles

Definition A triangle is isosceles precisely when at least two of its quadrances are
equal.

Theorem 64 (Null isosceles) An isosceles triangle has either zero or two null lines.

Proof. As in Exercise 3.12, a triangle cannot have three null lines, for by the Null line
theorem (page 60) this would imply that all three quadrances are zero, in which case
the quadrea is zero, which is impossible by the Quadrea theorem (page 68).

Suppose that A1 A2 A3 is isosceles, with quadrances Q1 = Q2 ≡ Q and Q3 , and with

exactly one null line. If this line is A1 A3 or A2 A3 then Q = 0 which means that both
A1 A3 and A2 A3 are null which is a contradiction. Otherwise A1 A2 is null, so Q3 = 0
and Q 6= 0. The Cross law (page 81) then states that
(Q + Q − 0)2 = 4Q2 c3
from which c3 = 1, and so A1 A3 and A2 A3 are parallel, which is impossible for a
triangle.

121
122 10. TRIANGLES

Theorem 65 (Pons Asinorum) In a non-null triangle Q1 = Q2 precisely when

s1 = s2 .

Proof. This follows from the Spread law

s1 s2 s3
= = .
Q1 Q2 Q3

Theorem 66 (Isosceles triangle) Suppose a non-null triangle A1 A2 A3 has

quadrances Q1 = Q2 ≡ Q and Q3 , and corresponding spreads s1 = s2 ≡ s and s3 .
Then
Q3 = 4Q (1 − s)
and µ ¶
Q3 Q3
s3 = 1− .
Q 4Q

A3

Q s3 Q
s s
A1 Q3 A2

Figure 10.1: Isosceles triangle theorem

Proof. Combine the Equal spreads theorem (page 94)

s3 = 4s (1 − s)

and the Spread law

s s3
=
Q Q3
to get
Q3 = 4Q (1 − s) .
Then
s = 1 − Q3 /4Q and 1 − s = Q3 /4Q
so that µ ¶
Q3 Q3
s3 = 1− .
Q 4Q
10.1. ISOSCELES TRIANGLES 123

Theorem 67 (Isosceles median) Suppose that A1 A2 A3 is non-null and isosceles,

with quadrances Q1 = Q2 ≡ Q and Q3 , and corresponding spreads s1 = s2 ≡ s and
s3 , with l a line passing through A3 . The following are equivalent

i) l is the altitude from A3 to A1 A2 .

ii) l is the median passing through the midpoint M3 of A1 A2 .
iii) l is a bisector of the vertex of A1 A2 A3 at A3 which intersects A1 A2 .

A3

Q r r Q
l
s s
A1 M3 A2

Figure 10.2: Isosceles median theorem

Proof. i)⇐⇒ ii) If l = A3 M3 is the median through A3 , then by the Equal quadrance
to two points theorem (page 66), A3 is on the perpendicular bisector of A1 A2 , so l is
the altitude from A3 to A1 A2 . Conversely if l is the altitude from A3 to A1 A2 , with
foot F lying on A1 A2 , then Pythagoras’ theorem shows that
Q (A1 , F ) = Q − Q (A3 , F ) = Q (A2 , F ) .
Thus by the Midpoint theorem (page 60), F = M3 and so l is the median passing
through M3 .

i)⇐⇒ iii) If l is the altitude from A3 to A1 A2 then it meets A1 A2 at the foot F. By

Pons Asinorum and the Complementary spreads theorem (page 79)
s (A3 A1 , A3 F ) = s (A3 A2 , A3 F ) .
So l is then a bisector of the vertex of A1 A2 A3 at A3 which intersects A1 A2 .

Conversely, suppose l is a bisector of the vertex at A3 which intersects A1 A2 at a point

B with s (A3 A1 , A3 B) = s (A3 A2 , A3 B) ≡ r. Then by the Spread law in the triangles
A1 A3 B and A2 A3 B
r s r
= = .
Q (A1 , B) Q (A3 , B) Q (A2 , B)
So Q (A1 , B) = Q (A2 , B) and by the Midpoint theorem B = M3 . Therefore by the first
part of the proof, l is also the altitude from A3 to A1 A2 .
124 10. TRIANGLES

Theorem 68 (Isosceles reflection) Suppose a non-null triangle A1 A2 A3 has

quadrances Q1 = Q2 ≡ Q and Q3 , and that the altitude from A3 to A1 A2 with foot
M3 makes (necessarily) equal spreads

s (A3 A1 , A3 M3 ) = s (A3 A2 , A3 M3 ) ≡ r.

Then
Q3 = 4rQ.

A3

Q r r Q
A1 M3 A2

Q3

Figure 10.3: Isosceles reflection theorem

Proof. In the right triangle A1 A3 M3 the Spread ratio theorem (page 77) states that
Q (A1 , M3 )
r= .
Q
By the Isosceles median theorem, M3 is the midpoint of the side A1 A2 , so that the
Midpoint theorem (page 60) shows that

Q (A1 , M3 ) = Q3 /4.

Thus
Q3 = 4rQ.

Exercise 10.1 If a triangle has crosses c1 = c2 ≡ c and c3 , and twists t1 = t2 ≡ t and

t3 , then show that
2 2
c3 = (2c − 1) t3 = 4t/ (1 − t) . ¦

Exercise 10.2 Show that in the notation of the Isosceles triangle theorem, the
quadrea of the isosceles triangle A1 A2 A3 is

A = 16Q2 s (1 − s) = Q3 (4Q − Q3 ) . ¦

Exercise 10.3 What happens to the previous theorems if isosceles triangles with two
null lines are considered? ¦
10.2. EQUILATERAL TRIANGLES 125

10.2 Equilateral triangles

Definition A triangle is equilateral precisely when all three of its quadrances are
equal.

By Pons Asinorum, this is equivalent to all three spreads being equal.

Theorem 69 (Equilateral triangle) In an equilateral triangle A1 A2 A3 , each

spread has the value 3/4, and each cross has the value 1/4. If all quadrances are equal
to Q, then the quadrea is A = 3Q2 .

Proof. If s1 = s2 = s3 ≡ s then the Triple spread formula becomes

¡ ¢
(3s)2 = 2 3s2 + 4s3
so 3s2 = 4s3 . Since s = 0 is impossible, deduce that s = 3/4. Thus by the Spread plus
cross theorem each cross has value 1 − 3/4 = 1/4. Use the definition of the quadrea to
find that ¡ ¢
A = (3Q)2 − 2 3Q2 = 3Q2 .

Exercise 10.4 Suppose the characteristic of the field is not three. In an equilateral
triangle A1 A2 A3 with all quadrances Q, show that the quadrance from A3 to the
centroid G is Q/3, that the quadrance from G to the midpoint M3 of A1 A2 is Q/12,
and that the quadrance from A3 to the midpoint M3 is 3Q/4, as in Figure 10.4.

A3

Q/ 3
M2 M1 3Q/ 4
G
Q/ 12
A1 A2
Q/ 4 M 3 Q/ 4
Q

Figure 10.4: Medians in an equilateral triangle ¦

Exercise 10.5 Show that an equilateral triangle with common quadrance Q = 1

exists precisely when 3 is a non-zero square number. ¦
126 10. TRIANGLES

10.3 Right triangles

This is the second most important result about right triangles (after Pythagoras’
theorem).

Theorem 70 (Right midpoint) Suppose that M is the midpoint of the side A1 A2

of the triangle A1 A2 A3 . Then the vertex at A3 is a right vertex precisely when

Q (M, A1 ) = Q (M, A3 ) .

A2
M

A1 A3

Figure 10.5: Right midpoint theorem

Proof. Suppose the points are A1 ≡ [x1 , y1 ], A2 ≡ [x2 , y2 ] and A3 ≡ [x3 , y3 ] so that
M = [(x1 + x2 ) /2, (y1 + y2 ) /2]. Then Q (M, A1 ) = Q (M, A3 ) precisely when
³ ´
2 2 2 2
(x2 − x1 ) + (y2 − y1 ) /4 = (x3 − (x1 + x2 ) /2) + (y3 − (y1 + y2 ) /2) .

Expand to obtain the condition

x23 − x3 x1 − x3 x2 + x1 x2 + y32 − y3 y1 − y3 y2 + y1 y2 = 0
which can be rewritten as
(x3 − x1 ) (x3 − x2 ) + (y1 − y3 ) (y2 − y3 ) = 0.
By the Cross from points theorem (page 84) this is exactly the condition that A1 A3 is
perpendicular to A2 A3 .

Suppose the right triangle A1 A2 A3 has quadrances Q1 , Q2 and Q3 , and spreads s1 , s2

and s3 , with s3 ≡ 1. Then any two of the five quantities Q1 , Q2 , Q3 , s1 and s2
determine the other three quantities unambiguously, with the exception of s1 and s2 .

If you know two of the quadrances, the third follows from Pythagoras’ theorem, so that
s1 and s2 can then be determined from the Spread ratio theorem (page 77) to be
s1 = Q1 /Q3 s2 = Q2 /Q3 .
10.4. CONGRUENT AND SIMILAR TRIANGLES 127

If you know one quadrance and one spread, say s1 , then by the Complementary
spreads theorem (page 79) s1 + s2 = 1 gives s2 , and then the equations

Q1 = Q3 s1 Q2 = Q3 s2

allow you to determine the other two quadrances. So with rational trigonometry there
are a wealth of right triangles that can be completely analysed by elementary means.

A2
Q3 s2

Q1
s1
A1 Q2 A3

Figure 10.6: Right triangle

Exercise 10.6 Suppose that A1 A2 A3 is a right triangle with a right vertex at A3 ,

quadrances Q1 , Q2 and Q3 , and that F is the foot of the altitude from A3 to A1 A2 .
Show that

Q (A1 , F ) = Q21 /Q3 Q (A2 , F ) = Q22 /Q3 Q (A3 , F ) = Q1 Q2 /Q3 . ¦

10.4 Congruent and similar triangles

Definition Two triangles A1 A2 A3 and B1 B2 B3 are congruent precisely when they

have identical quadrances.

Note that this information does not tell you which sides of the two triangles have equal
quadrances.

Exercise 10.7 In the rational number field, show that the triangle with points
A1 ≡ [2, 3], A2 ≡ [3, 1] and A3 ≡ [10, 2] is congruent to the triangle with points
B1 ≡ [0, 1], B2 ≡ [−2, 2] and B3 ≡ [5, 6]. ¦

Definition Two triangles A1 A2 A3 and B1 B2 B3 are similar precisely when they

have identical spreads.

Since the quadrances of a triangle determine the spreads, congruent triangles are
similar.
128 10. TRIANGLES

Exercise 10.8 In the rational number field, show that the triangle with points
A1 ≡ [−4, 3], A2 ≡ [6, 5] and A3 ≡ [−1, 1] is similar to, but not congruent to, the
triangle with points B1 ≡ [4, −7], B2 ≡ [−2, −11] and B3 ≡ [2, 9]. ¦

Exercise 10.9 Show that if two triangles are similar, then there is a non-zero
number λ such that the quadrances of one triangle are all λ times the quadrances of
the other. ¦

Exercise 10.10 In F3 show that any two triangles are similar, and that any triangle
is both isosceles and right. ¦

Exercise 10.11 In F5 show that any two non-null triangles are similar, and are
isosceles. Show that up to similarity there are two types of null triangles. ¦

Theorem 71 (Median triangle) Suppose that A1 A2 A3 is a triangle with M1 , M2

and M3 the midpoints of the sides A2 A3 , A1 A3 and A1 A2 respectively. Then
M1 M2 M3 is similar to A1 A2 A3 , and

Q (M1 , M2 ) = Q (A1 , A2 ) /4.

A2

Q 3 /4 s 2 Q 1 /4
M3 M1
Q 3 /4 s2 Q 1 /4
s1 s1
A1 A3
Q 2 /4 M2 Q 2 /4

Figure 10.7: Median triangle theorem

Proof. From Thales’ theorem (page 48) M1 M2 is parallel to A1 A2 , M2 M3 is parallel

to A2 A3 and M1 M3 is parallel to A1 A3 . Thus M1 M2 M3 has the same spreads as
A1 A2 A3 , so these triangles are similar. Also A1 M3 M1 M2 is a parallelogram, so by the
Parallelogram quadrance theorem (page 62) and the Midpoint theorem (page 60)
Q (M1 , M2 ) = Q (A1 , M3 ) = Q (A1 , A2 ) /4.

The triangle M1 M2 M3 is the median triangle of A1 A2 A3 .

10.5. SOLVING TRIANGLES 129

10.5 Solving triangles

Generally speaking if you know three of the six possible quadrances and spreads of a
triangle then you can determine a small number of possibilities for the other unknown
quantities. The exception is knowing three spreads, which determines the triangle only
up to similarity.

Three quadrances

The quadrances Q1 , Q2 and Q3 of a non-null triangle A1 A2 A3 determine the quadrea

A, and then the spreads s1 , s2 and s3 by the Quadrea spread theorem (page 82).

Two quadrances and a spread

If two quadrances and a spread, say s3 , of a non-null triangle are known, then the
Cross law
(Q1 + Q2 − Q3 )2 = 4Q1 Q2 (1 − s3 )
gives a quadratic equation for the third unknown quadrance. If the unknown
quadrance is Q3 , then this is already in normal form. If the unknown quadrance is Q1
or Q2 , then first complete the square.

For each value of the third quadrance you can determine the other spreads by the
Spread law.

Example 10.1 In the rational number field a triangle A1 A2 A3 has s3 ≡ 81/130,

Q1 ≡ 5 and Q3 ≡ 17. Then use the Cross law to get
49 98
(Q2 − 12)2 = 4 × 5 × Q2 × = × Q2 .
130 13
Rearrange to find
(13Q2 − 72) (Q2 − 26) = 0
so that
i) Q2 = 72/13 or ii) Q2 = 26.
i) If Q2 = 72/13, then the Spread law is

s1 s2 81/130
= =
5 72/13 17

so that
s1 = 81/442 s2 = 2916/14 365 s3 = 81/130.
This is illustrated to scale on the left in Figure 10.8.
130 10. TRIANGLES

A3
81
130
A3 26
72 5
5 81 13
130
A2 A1 A1
17 A2 17

Figure 10.8: Two possibilities

ii) If Q2 = 26, then the Spread law is

s1 s2 81/130
= =
5 26 17
so that
s1 = 81/442 s2 = 81/85 s3 = 81/130.
This is illustrated to scale on the right in Figure 10.8. ¦

One quadrance and two spreads

If one quadrance and two spreads, say s1 and s2 are known, then the Triple spread
formula gives a quadratic equation for the third spread s3 . For each of the generally
two solutions s3 , the two unknown quadrances may be found using the Spread law.

Exercise 10.12 Suppose that in the complex number field a triangle A1 A2 A3 has
quadrance Q3 ≡ −3, and spreads s1 ≡ 4/15 + 2i/15 and s2 ≡ 1/24 + 7i/24. Then show
that either
Q1 = 8 − 8i Q2 = 11 + 2i
or

Q1 = 10 872/11 881 + 2232i/11 881 Q2 = 7929/11 881 − 7578i/11 881. ¦

Three spreads

Knowledge of the three spreads of a triangle determines the triangle only up to

similarity. Although the quadrances are not determined, the Spread law
s1 s2 s3
= =
Q1 Q2 Q3
determines the proportion Q1 : Q2 : Q3 .
 11

Laws of proportion

This chapter introduces useful results when triangles and quadrilaterals are augmented
by additional lines. These lead to generalizations of the classical theorems of Stewart,
Menelaus and Ceva.

11.1 Triangle proportions

Theorem 72 (Triangle proportions) Suppose that A1 A2 A3 is a triangle with

quadrances Q1 , Q2 and Q3 , corresponding spreads s1 , s2 and s3 , and that D is a
point lying on the line A1 A2 distinct from A1 and A2 . Define the quadrances
R1 ≡ Q (A1 , D) and R2 ≡ Q (A2 , D), and the spreads r1 ≡ s (A3 A1 , A3 D) and
r2 ≡ s (A3 A2 , A3 D). Then
r1 s1 R1 Q1 R1
= = .
r2 s2 R2 Q2 R2

A3
r1 r 2
Q2 Q1
R3
s1 s2
A1 R1 D R2 A2

Figure 11.1: Triangle proportions theorem

131
132 11. LAWS OF PROPORTION

Proof. Define also R3 ≡ Q (A3 , D) as in Figure 11.1. In DA2 A3 use the Spread law to
get
s2 r2
= .
R3 R2
In DA1 A3 use the Spread law to get
s1 r1
= .
R3 R1

Thus
s2 R2 s1 R1
R3 = =
r2 r1
and rearrange to obtain
r1 s1 R1
= .
r2 s2 R2
Since
s1 Q1
=
s2 Q2
this can be rewritten as
r1 Q1 R1
= .
r2 Q2 R2

Special cases
• If r1 = r2 then the line DA3 bisects the vertex of A1 A2 A3 at A3 . The theorem
then reduces to
R1 s2 Q2
= = .
R2 s1 Q1
Recall that in general there are two possible vertex bisectors.

• If s1 = s2 then the triangle A1 A2 A3 is isosceles. The theorem then reduces to

r1 R1
= .
r2 R2

• If R1 = R2 then the line DA3 is a median of A1 A2 A3 . The theorem then reduces

to
r1 s1 Q1
= = .
r2 s2 Q2

Exercise 11.1 Show that with the same notation as in the theorem, if s3 = 1 then
µ ¶−1
r2 R1
s1 = 1+ . ¦
r1 R2
11.2. QUADRILATERAL PROPORTIONS 133

11.2 Quadrilateral proportions

Theorem 73 (Quadrilateral proportions) Suppose that ABCD is a

quadrilateral with spreads s1 ≡ s (AB, AD) and s2 ≡ s (CB, CD) at A and C
respectively. Define the spreads r1 ≡ s (BA, BD) and r2 ≡ s (BC, BD), and the
quadrances R1 ≡ Q (A, D) and R2 ≡ Q (C, D). Then

r1 s1 R1
= .
r2 s2 R2

In the rational or decimal number fields, both diagrams in Figure 11.2 illustrate the
theorem.

B r1
B
r1 r 2
r2
R
A C A
s1 R s2 s1
D
R1
R1 R2 s2
R2
D C

Figure 11.2: Quadrilateral proportions theorem

Proof. Define R ≡ Q (B, D). Use the Spread law in ABD to obtain
r1 s1
=
R1 R

and in BCD to obtain

r2 s2
= .
R2 R
Solve for R
s1 R1 s2 R2
R= =
r1 r2
and rearrange to obtain
r1 s1 R1
= .
r2 s2 R2
134 11. LAWS OF PROPORTION

11.3 Two struts theorem

This theorem illustrates a quadratic aspect of barycentric coordinates.

Theorem 74 (Two struts) Suppose that A1 A2 A3 is a triangle and that the points
B1 and B2 are distinct from A1 , A2 and A3 and lie on the lines A2 A3 and A1 A3
respectively, with A1 B1 and A2 B2 intersecting at the point C. Suppose the
quadrances
P1 ≡ Q (A1 , B2 ) P2 ≡ Q (B2 , A3 )
R1 ≡ Q (A2 , B1 ) R2 ≡ Q (B1 , A3 )
N1 ≡ Q (A1 , C) N2 ≡ Q (C, B1 )
M1 ≡ Q (A2 , C) M2 ≡ Q (C, B2 )
are all non-zero, and define the corresponding numbers

p ≡ P1 /P2 r ≡ R1 /R2
n ≡ N1 /N2 m ≡ M1 /M2 .

Then {r, rp, mp} and {p, rp, nr} are both quad triples.

A2
R1
M1 B1
C N2
R2
N1 M2

A1 A3
P1 B2 P2

Figure 11.3: Two struts theorem

Proof. Suppose the quadrances of the triangle A1 A2 A3 are Q1 , Q2 and Q3 as usual.

Use the Triangle proportions theorem in the triangles A1 A2 A3 and A1 A2 B2 , together
with the line A1 B1 , to get
s (A1 B1 , A1 A2 ) Q2 R1 P1 M1
= =
s (A1 B1 , A1 A3 ) Q3 R2 Q3 M2
so that
Q2 m
= . (11.1)
P1 r
11.3. TWO STRUTS THEOREM 135

Since {Q2 , P1 , P2 } is a quad triple, the Triple quad formula gives

2 ¡ ¢
(Q2 + P1 + P2 ) = 2 Q22 + P12 + P22 .

Divide by P12 and use (11.1) to obtain

µ ¶2 µ ¶
m 1 m2 1
+1+ =2 2
+1+ 2 .
r p r p

Now remove denominators to get

2 ¡ ¢
(mp + rp + r) = 2 m2 p2 + r2 p2 + r2 .

This is the statement that {r, rp, mp} is a quad triple. The situation with {p, rp, nr} is
similar.

Note that the ratio p for example does not determine the point B2 uniquely on the line
A1 A3 , and in fact there are generally two possible choices. Similarly given r there are
two possible choices for B1 on A2 A3 . This might suggest that given p and r there
should be four possible values for m, and also for n. However the theorem shows that
only two occur in each case.

Example 11.1 Suppose that p ≡ 1/4 and r ≡ 9/64. Then the theorem gives two
quadratic equations, the one for m factoring as

(64m − 9) (64m − 81) = 0

and the one for n factoring as

(36n − 25) (36n − 121) = 0. ¦

Exercise 11.2 Suppose that r = p in the above theorem. Show that n and m must
both be solutions to the quadratic equation (x − r − 1)2 = 4r. ¦

Exercise 11.3 Suppose that m = r in the above theorem. Show that then p = 1/4. ¦

Exercise 11.4 In the notation of the above theorem, suppose that A1 B1 and A2 B2
are bisectors of the vertices of A1 A2 A3 at A1 and A2 respectively. Suppose that the
spreads of the triangle A1 A2 A3 are s1 , s2 and s3 as usual. Show that m and n satisfy
respectively the quadratic equations
µ ¶2 ¡ ¢
s1 + s3 2 s21 + s23
m+ =
s2 s2
µ ¶2 ¡ 2 2 2¢
s2 + s3 2 s2 + s3
n+ = . ¦
s1 s21
136 11. LAWS OF PROPORTION

11.4 Stewart’s theorem

The decimal number version of this result was published by Matthew Stewart in 1745.

Theorem 75 (Stewart’s theorem) Suppose that A1 A2 A3 is a non-null triangle

with quadrances Q1 , Q2 and Q3 , with D a point lying on A1 A2 . Define

R1 ≡ Q (A1 , D) R2 ≡ Q (A2 , D) R3 ≡ Q (A3 , D) .

Then
R2 (R3 + R1 − Q2 )2 = R1 (R3 + R2 − Q1 )2 .

A3

Q2 Q1
R3
s1 r s2
A1 R1 D R2 A2

Figure 11.4: Stewart’s theorem

Proof. If R3 6= 0 then define r ≡ s (DA1 , DA3 ). Then the Cross law in the triangles
A1 A3 D and A2 A3 D yields

(R3 + R1 − Q2 )2 = 4R1 R3 (1 − r)
2
(R3 + R2 − Q1 ) = 4R2 R3 (1 − r) .

Combine these two to get the result. If R3 = 0 then use Exercise 6.4 and the Spread
law to get
s1 (R1 − Q2 )2 R2 Q1 Q1
= 2 =
s2 (R2 − Q1 ) R1 Q2 Q 2

from which the result follows.

Exercise 11.5 Suppose that Q1 = Q2 ≡ Q in Stewart’s theorem. Show that

(Q − R3 )2 = R1 R2 . ¦

Exercise 11.6 Suppose that Q1 = Q2 = Q3 ≡ Q in Stewart’s theorem. Show that Q

equals
2R3 − R1 − R2 or (2R3 + R1 + R2 ) /3. ¦
11.5. MEDIAN QUADRANCE AND SPREAD 137

11.5 Median quadrance and spread

Theorem 76 (Median quadrance) Suppose a triangle A1 A2 A3 has quadrances

Q1 , Q2 and Q3 , with Q1 6= 0. Let P1 denote the quadrance from A1 to the midpoint
M1 of A2 A3 . Then
2 (Q2 + Q3 ) − Q1
P1 = .
4

A2

Q 1 /4
Q3
M1
P1 Q 1 /4
A1 Q2 A3

Figure 11.5: Median quadrance theorem

Proof. By Stewart’s theorem,

µ ¶2 µ ¶2
Q1 Q1 Q1 Q1
P1 + − Q2 = P1 + − Q3
4 4 4 4
or
Q1 (Q3 − Q2 ) (4P1 + Q1 − 2Q2 − 2Q3 ) = 0.
By assumption Q1 6= 0. If Q3 − Q2 = 0 then the triangle A1 A2 A3 is isosceles and so
from the Equal quadrance to two points theorem (page 66), A1 M1 is the perpendicular
bisector of the side A2 A3 . Then Pythagoras’ theorem implies that
Q1 2 (Q2 + Q3 ) − Q1
P1 = Q2 − = .
4 4
Otherwise
4P1 + Q1 − 2Q2 − 2Q3 = 0
so that also
2 (Q2 + Q3 ) − Q1
P1 = .
4

Exercise 11.7 If P2 and P3 are defined analogously to P1 , show that

3
P1 + P2 + P3 = (Q1 + Q2 + Q3 ) . ¦
4
138 11. LAWS OF PROPORTION

Theorem 77 (Median spreads) Suppose the triangle A1 A2 A3 has quadrances

Q1 , Q2 and Q3 , and spreads s1 , s2 and s3 , with M1 the midpoint of the non-null side
A2 A3 . If A1 M1 is a non-null line then
4s2 s3
s (M1 A2 , M1 A1 ) =
2 (s2 + s3 ) − s1
s1 s2
s (A1 A2 , A1 M1 ) =
2 (s2 + s3 ) − s1
s1 s3
s (A1 A3 , A1 M1 ) = .
2 (s2 + s3 ) − s1

Proof. As in Figure 11.6 define the quadrance P1 ≡ Q (A1 , M1 ) 6= 0 and the spreads
r1 ≡ s (M1 A1 , M1 A2 ) u1 ≡ s (A1 A2 , A1 M1 ) v1 ≡ s (A1 A3 , A1 M1 ) .

A2

r1 M
1
u1
v1 P1
A1 A3

Figure 11.6: Median spreads

Since Q (A2 , M1 ) = Q1 /4, the Spread law in A1 A2 M1 gives

r1 s2 u1
= = .
Q3 P1 Q1 /4
Use the Median quadrance theorem and the Spread law in A1 A2 A3 to get
4s2 Q3
r1 =
2 (Q2 + Q3 ) − Q1
4s2 s3
= .
2 (s2 + s3 ) − s1
Thus
r1 Q1 s1 r1 s1 s2
u1 = = = .
4Q3 4s3 2 (s2 + s3 ) − s1
Similarly from the Spread law in A1 A3 M1
r1 Q1 s1 r1 s1 s3
v1 = = = .
4Q2 4s2 2 (s2 + s3 ) − s1
11.6. MENELAUS’ AND CEVA’S THEOREMS 139

11.6 Menelaus’ and Ceva’s theorems

The results in this section include rational analogues of the famous theorems of
Menelaus (about 100 A.D.) and Ceva (1678). These theorems properly belong to affine
geometry, but they can also be viewed metrically.

Theorem 78 (Menelaus’ theorem) Suppose that l is a non-null line not parallel

to any of the lines of the triangle A1 A2 A3 , and intersecting A2 A3 , A1 A3 and A1 A2 at
the points B1 , B2 and B3 respectively. Define the quadrances

R1 ≡ Q (A2 , B1 ) P1 ≡ Q (B1 , A3 )
R2 ≡ Q (A3 , B2 ) P2 ≡ Q (B2 , A1 )
R3 ≡ Q (A1 , B3 ) P3 ≡ Q (B3 , A2 ) .

Then
R1 R2 R3 = P1 P2 P3 .

B1
r1 R1
P1
A2
B3 P3
R3 r3

A1 P2 r 2 B 2 R2 A3
l

Figure 11.7: Menelaus’ theorem

Proof. If A1 A2 A3 is a null triangle then both sides of the equation are zero.
Otherwise define the spreads between l and the lines A2 A3 , A1 A3 and A1 A2 to be
respectively r1 , r2 and r3 , as in Figure 11.7. Then use the Spread law in the triangles
B1 B2 A3 , B2 B3 A1 and B3 B1 A2 to get

r1 /r2 = R2 /P1 r2 /r3 = R3 /P2 r3 /r1 = R1 /P3 .

Multiply these three equations to obtain

R1 R2 R3 = P1 P2 P3 .
140 11. LAWS OF PROPORTION

Exercise 11.8 Show that the converse of Menelaus’ theorem does not hold: the
relation R1 R2 R3 = P1 P2 P3 does not necessarily imply that the points B1 , B2 and B3
are collinear. ¦

Theorem 79 (Alternate spreads) Suppose that A0 is a point distinct from the

points of a non-null triangle A1 A2 A3 , with the spreads

r1 ≡ s (A1 A2 , A1 A0 ) p1 ≡ s (A1 A3 , A1 A0 )
r2 ≡ s (A2 A3 , A2 A0 ) p2 ≡ s (A2 A1 , A2 A0 )
r3 ≡ s (A3 A1 , A3 A0 ) p3 ≡ s (A3 A2 , A3 A0 ) .

Then
r1 r2 r3 = p1 p2 p3 .

A2

p2 r2
N2

A0 p3
r1 N1 N3
p1 r3
A1 A3

Figure 11.8: Alternate spreads theorem

Proof. If A0 lies on one of the lines of the triangle then both sides of the required
equation equal zero. Otherwise, define the quadrances

N1 ≡ Q (A0 , A1 )
N2 ≡ Q (A0 , A2 )
N3 ≡ Q (A0 , A3 ) .

Then the Spread law applied to the triangles A0 A2 A3 , A0 A3 A1 and A0 A1 A2 yields

r2 /p3 = N3 /N2 r3 /p1 = N1 /N3 r1 /p2 = N2 /N1 .

Multiply these three equations to get

r1 r2 r3 = p1 p2 p3 .
11.6. MENELAUS’ AND CEVA’S THEOREMS 141

Theorem 80 (Ceva’s theorem) Suppose that A0 is a point distinct from the

points A1 , A2 and A3 of a non-null triangle A1 A2 A3 , and that the non-null lines
A0 A1 , A0 A2 and A0 A3 meet A2 A3 , A1 A3 and A1 A2 respectively at the points B1 , B2
and B3 . Define the quadrances

R1 ≡ Q (A2 , B1 ) P1 ≡ Q (B1 , A3 )
R2 ≡ Q (A3 , B2 ) P2 ≡ Q (B2 , A1 )
R3 ≡ Q (A1 , B3 ) P3 ≡ Q (B3 , A2 ) .

Then
R1 R2 R3 = P1 P2 P3 .

A2
P3 R1
B3 B1
A0
R3 P1

A1 A3
P2 B2 R2

Figure 11.9: Ceva’s theorem

Proof. Let s1 , s2 and s3 be the usual spreads of the triangle A1 A2 A3 . Define the
spreads
r1 ≡ s (A1 A2 , A1 A0 ) p1 ≡ s (A1 A3 , A1 A0 )
r2 ≡ s (A2 A3 , A2 A0 ) p2 ≡ s (A2 A1 , A2 A0 )
r3 ≡ s (A3 A1 , A3 A0 ) p3 ≡ s (A3 A2 , A3 A0 ) .
as in the Alternate spreads theorem. Then use the Triangle proportions theorem (page
131) with the triangle A1 A2 A3 and the respective lines A1 B1 , A2 B2 and A3 B3 to
obtain
r1 s2 R1
=
p1 s3 P1
r2 s3 R2
=
p2 s1 P2
r3 s1 R3
= .
p3 s2 P3
Multiply these three equations and use the Alternate spreads theorem to get
R1 R2 R3 r1 r2 r3
= = 1.
P1 P2 P3 p1 p2 p3
142 11. LAWS OF PROPORTION

Exercise 11.9 Show that the Alternate spreads theorem extends to a quadrilateral
A1 A2 A3 A4 , so that if A0 is any point distinct from A1 , A2 , A3 and A4 , and the spreads
are as shown in Figure 11.10, then

r1 r2 r3 r4 = p1 p2 p3 p4 .

Generalize further to an n-gon.

A2 p3 A3
r2
p2 r3

r1 A0 p4
p1 r4
A1 A4

Figure 11.10: Alternate spreads (quadrilateral) ¦

Exercise 11.10 Show that the converse of Ceva’s theorem does not hold: the relation
R1 R2 R3 = P1 P2 P3 does not necessarily imply that the lines A1 B1 , A2 B2 and A3 B3 are
concurrent. ¦

Exercise 11.11 What happens when the conditions of the theorems in this chapter
are relaxed to allow null triangles and null lines? ¦
 12

Centers of triangles

In this chapter the circumcenter, orthocenter and incenters of a triangle are studied.
Typically the existence of incenters depends on number theoretic considerations, and if
incenters exist, then there are four which play a symmetrical role.

12.1 Perpendicular bisectors and circumcenter

Theorem 81 (Circumcenter) For any triangle A1 A2 A3 the three perpendicular

bisectors of the sides are concurrent at a point C.

Proof. Since two lines of a triangle are never parallel, the respective perpendicular
bisectors always intersect. If C is the intersection of the perpendicular bisectors p1 and
p2 of A2 A3 and A1 A3 respectively, then it follows from the Equal quadrance to two
points theorem (page 66) that Q (C, A2 ) = Q (C, A3 ) and Q (C, A1 ) = Q (C, A3 ).

But then Q (C, A1 ) = Q (C, A2 ) so that C must lie on the perpendicular bisector p3 of
A1 A2 .

Definition The circumcenter C of the triangle A1 A2 A3 is the common

intersection of the three perpendicular bisectors of the sides of the triangle. The
circumquadrance K of the triangle A1 A2 A3 is the common quadrance

K ≡ Q (C, A1 ) = Q (C, A2 ) = Q (C, A3 ) .

143
144 12. CENTERS OF TRIANGLES

Theorem 82 (Extended spread law) Suppose a non-null triangle A1 A2 A3 has

quadrances Q1 , Q2 and Q3 , corresponding spreads s1 , s2 and s3 , and
circumquadrance K. Then
s1 s2 s3 1
= = = .
Q1 Q2 Q3 4K

Proof. Suppose that C is the circumcenter of A1 A2 A3 , and that M1 , M2 and M3 are

the midpoints of the sides A2 A3 , A1 A3 and A1 A2 respectively. Since by the Median
triangle theorem (page 128) M1 , M2 and M3 are distinct, you may assume without loss
of generality that C is distinct from M1 and M2 , and does not lie on M1 M2 . Define
the quadrances
R1 ≡ Q (M1 , C) R2 ≡ Q (M2 , C)
as in Figure 12.1.

s3
A3
Q 2 /4
Q 1 /4
M2 K M1
Q 2 /4
R2 R1 Q 1 /4
K C K
A1 A2
s1 Q 3 /4 M3 Q 3 /4 s2

Figure 12.1: Circumcenter of a triangle

Since A3 M2 is perpendicular to M2 C and A3 M1 is perpendicular to M1 C, the

Perpendicular spreads theorem (page 79) gives

s (CM1 , CM2 ) = s (A3 M1 , A3 M2 ) = s3 .

The Median triangle theorem (page 128) states that

Q (M1 , M2 ) = Q3 /4.

From Thales’ theorem (page 48), M1 M2 is parallel to A1 A2 , so that

s (M2 M1 , M2 A3 ) = s1 and s (M1 M2 , M1 A3 ) = s2 .

Use the Complementary spreads theorem (page 79) to see that

s (M2 M1 , M2 C) = 1 − s1 and s (M1 M2 , M1 C) = 1 − s2 .

12.1. PERPENDICULAR BISECTORS AND CIRCUMCENTER 145

Now solve the triangle M1 M2 C, since you know one quadrance and all three spreads.
Use the Spread law to get
Q3 (1 − s1 )
R1 =
4s3
Q3 (1 − s2 )
R2 = .
4s3
Pythagoras’ theorem in the right triangle A3 M1 C gives
Q1 Q3 (1 − s1 ) Q1
K = R1 + = + .
4 4s3 4
From the Spread law in A1 A2 A3
Q3 Q1
=
s3 s1
so that
Q1 (1 − s1 ) Q1 Q1
K= + = .
4s1 4 4s1
By symmetry
s1 s2 s3 1
= = = .
Q1 Q2 Q3 4K

Exercise 12.1 Using the notation of the previous proof, show that
s (A3 C, A3 A2 ) = 1 − s1 s (A3 C, A3 A1 ) = 1 − s2
and
s (CA3 , CM1 ) = s1 = s (CA2 , CM1 ) . ¦

Exercise 12.2 Show that if the triangle A1 A2 A3 has quadrances Q1 , Q2 and Q3 ,

spreads s1 , s2 and s3 , quadrea A and circumquadrance K, then
Q1 Q2 Q3
A= = 64K 2 s1 s2 s3 . ¦
K
Exercise 12.3 Show that with the notation of the previous exercise,
s1 + s2 = (Q1 + Q2 ) /4K and s1 + s2 + s3 = (Q1 + Q2 + Q3 ) /4K. ¦

Exercise 12.4 Show that if A2 A3 is a null line, then the circumcenter of A1 A2 A3 lies
on A2 A3 . Deduce that if A1 A3 and A2 A3 are both null lines, then the circumcenter is
A3 . ¦

Exercise 12.5 Verify the polynomial identity

³ ¡ ¢´
(x1 + x2 + x3 )2 − 2 x21 + x22 + x23 (x1 + x2 + x3 ) − 8x1 x2 x3
= (x1 + x2 − x3 ) (x2 + x3 − x1 ) (x3 + x1 − x2 ) .

Deduce that Q1 + Q2 + Q3 = 8K precisely when the triangle A1 A2 A3 is a right

triangle. ¦
146 12. CENTERS OF TRIANGLES

12.2 Formulas for the circumcenter

Theorem 83 (Circumcenter formula) If A1 ≡ [x1 , y1 ], A2 ≡ [x2 , y2 ] and

A3 ≡ [x3 , y3 ], then the circumcenter C of the triangle A1 A2 A3 is
"£ ¤− £ ¤− £ ¤− £ ¤− #
x21 y2 3 + y12 y2 3 x1 x22 3 + x1 y22 3
C= − , − .
2 [x1 y2 ]3 2 [x1 y2 ]3

Proof. By the Circumcenter theorem, it suffices to find the intersection of the

perpendicular bisectors p3 and p1 of the sides A1 A2 and A2 A3 , which by the
Perpendicular bisector theorem (page 50) are
­ ¡ ¢ ®
p3 = x1 − x2 : y1 − y2 : x22 − x21 + y22 − y12 /2
­ ¡ ¢ ®
p1 = x2 − x3 : y2 − y3 : x23 − x22 + y32 − y22 /2 .
Use the Point on two lines theorem (page 40) to deduce that p1 and p3 intersect at the
point C ≡ [x0 , y0 ] where
µ 2 ¶
x1 y2 − x21 y3 + x22 y3 − x23 y2 + x23 y1 − x22 y1
+y12 y2 − y12 y3 + y22 y3 − y32 y2 + y32 y1 − y22 y1
x0 ≡
2 (x1 y2 − x1 y3 + x2 y3 − x3 y2 + x3 y1 − x2 y1 )
and µ ¶
x1 x22 − x1 x23 + x2 x23 − x3 x22 + x3 x21 − x2 x21
+x1 y22 − x1 y32 + x2 y32 − x3 y22 + x3 y12 − x2 y12
y0 ≡ .
2 (x1 y2 − x1 y3 + x2 y3 − x3 y2 + x3 y1 − x2 y1 )
Express C using the conventions for anti-symmetric polynomials (page 29).

Theorem 84 (Affine circumcenter) If the quadrances of the triangle A1 A2 A3 are

Q1 , Q2 and Q3 , and the quadrea is A, then the circumcenter C can be written as the
affine combination
C = γ 1 A1 + γ 2 A2 + γ 3 A3
where

γ1 ≡ Q1 (Q2 + Q3 − Q1 ) /A
γ2 ≡ Q2 (Q3 + Q1 − Q2 ) /A
γ3 ≡ Q3 (Q1 + Q2 − Q3 ) /A.

Proof. This is a straightforward but laborious verification using the previous theorem.
Note that Exercise 5.14 (page 69) shows that γ 1 + γ 2 + γ 3 = 1.
12.3. ALTITUDES AND ORTHOCENTER 147

12.3 Altitudes and orthocenter

Definition An altitude of a triangle A1 A2 A3 is an altitude from a point of the

triangle to the line of the opposite side.

Theorem 85 (Orthocenter) For any non-null triangle A1 A2 A3 the three altitudes

are concurrent at a point O.

Proof. In the special case when the triangle is a right triangle, the theorem is
immediate, since then the altitudes intersect at the point of the right vertex.

Since the lines A1 A3 and A2 A3 of the triangle are not parallel, the altitudes n1 from
A1 to A2 A3 and n2 from A2 to A1 A3 are not parallel, so intersect at some point O.
Suppose that F1 and F2 are the feet of the altitudes n1 and n2 lying on A2 A3 and
A1 A3 respectively. If O lies either on A1 A3 or A2 A3 then it must coincide with one of
the feet, in which case A1 A2 A3 has a right vertex at A3 .

Otherwise assume that O does not lie on A1 A3 or A2 A3 . Suppose that the quadrances
of A1 A2 A3 are Q1 , Q2 and Q3 , and the corresponding spreads are s1 , s2 and s3 . It
suffices to show that A1 A2 and OA3 are perpendicular.
A2

F1
r
O
A3
A1 F2

Figure 12.2: Two altitudes intersecting

By properties of right triangles

s (A1 A3 , A1 F1 ) = 1 − s3 s (A2 A3 , A2 F2 ) = 1 − s3
s (A1 A2 , A1 F1 ) = 1 − s2 s (A2 A1 , A2 F2 ) = 1 − s1 .

Use the Spread ratio theorem (page 77) to get

Q (A1 , F2 ) = (1 − s1 ) Q3 Q (A2 , F1 ) = (1 − s2 ) Q3
Q (A3 , F1 ) = (1 − s3 ) Q2 Q (A3 , F2 ) = (1 − s3 ) Q1 .
148 12. CENTERS OF TRIANGLES

Use the Spread law in A1 OF2 and A2 OF1 to obtain

Q (O, F2 ) = (1 − s1 ) (1 − s3 ) Q3 /s3
Q (O, F1 ) = (1 − s2 ) (1 − s3 ) Q3 /s3 .

Apply Pythagoras’ theorem

Q (O, A3 ) = Q (O, F1 ) + Q (F1 , A3 )

= (1 − s2 ) (1 − s3 ) Q3 /s3 + (1 − s3 ) Q2
= (1 − s2 ) (1 − s3 ) Q3 /s3 + (1 − s3 ) s2 Q3 /s3
= (1 − s3 ) Q3 /s3 .

Thus OA3 is a non-null line and so define the spread r ≡ s (A1 A2 , OA3 ). From the
Spread ratio theorem
s (A3 O, A3 F1 ) = 1 − s2 and s (A3 O, A3 F2 ) = 1 − s1 . (12.1)

Now apply the Triple spread formula to the three lines A1 A2 , A1 A3 and A3 O to see
that {r, s1 , 1 − s1 } is a spread triple. Similarly {r, s2 , 1 − s2 } is a spread triple. The
Two spread triples theorem (page 98) shows that if

1 − 2s1 (1 − s1 ) 6= 1 − 2s2 (1 − s2 ) (12.2)

then
2 2
(2s1 − 1) − (2s2 − 1)
r= = 1.
2 (1 − 1 − 2s1 (1 − s1 ) + 2s2 (1 − s2 ))
But the condition (12.2) occurs precisely when either s1 = s2 or s1 = 1 − s2 , in which
case the triangle is either an isosceles or a right triangle respectively. The case of a
right triangle has already been considered.

If s1 = s2 ≡ s then (12.1) shows that

s (A3 O, A3 F1 ) = s (A3 O, A3 F2 )

and so A3 O is a vertex bisector of A1 A2 A3 . If A3 O is parallel to A1 A2 , then both A3

and O lie on the perpendicular bisector of A1 A2 , the latter since
s (A1 O, A1 A2 ) = s (A2 O, A2 A1 ) = 1 − s. Then O must coincide with A3 , and this case
has already been excluded. Thus A3 O intersects A1 A2 , and so the Isosceles median
theorem (page 123) shows it to be the altitude from A3 to A1 A2 .

Definition The orthocenter O of the triangle A1 A2 A3 is the common intersection

of the altitudes of the triangle.

If a triangle A1 A2 A3 is a null triangle with exactly one line a null line, say A1 A2 , then
the altitudes from A1 and A2 to the opposite sides are still defined, as is the
orthocenter.

The next theorem provides another computational proof of the previous one.
12.4. FORMULAS FOR THE ORTHOCENTER 149

12.4 Formulas for the orthocenter

Theorem 86 (Orthocenter formula) If A1 ≡ [x1 , y1 ], A2 ≡ [x2 , y2 ] and

A3 ≡ [x3 , y3 ], then the orthocenter O of the triangle A1 A2 A3 is
" £ 2 ¤− £ 2 ¤− #
[x1 x2 y2 ]− −
3 + y1 y2 3 [x1 y1 y2 ]3 + x1 x2 3
O= , .
[x1 y2 ]−
3 [x1 y2 ]−
3

Proof. The altitudes n1 and n2 from A1 and A2 respectively are given by the Altitude
to a line theorem (page 41) as
n1 = hx3 − x2 : y3 − y2 : x1 (x2 − x3 ) + y1 (y2 − y3 )i
n2 = hx3 − x1 : y3 − y1 : x2 (x1 − x3 ) + y2 (y1 − y3 )i .
The intersection of these lines is the point O ≡ [x0 , y0 ] where
µ ¶
x1 x2 y2 − x1 x3 y3 + x2 x3 y3 − x3 x2 y2 + x3 x1 y1 − x2 x1 y1
+y1 y22 − y1 y32 + y2 y32 − y3 y22 + y3 y12 − y2 y12
x0 ≡
x1 y2 − x1 y3 + x2 y3 − x3 y2 + x3 y1 − x2 y1
and µ ¶
x1 y1 y2 − x1 y1 y3 + x2 y2 y3 − x3 y3 y2 + x3 y3 y1 − x2 y2 y1
+x21 x2 − x21 x3 + x22 x3 − x23 x2 + x23 x1 − x22 x1
y0 ≡ .
x1 y2 − x1 y3 + x2 y3 − x3 y2 + x3 y1 − x2 y1
Use the conventions for anti-symmetric polynomials to rewrite O as stated. Since x0
and y0 are quotients of anti-symmetric polynomials, they are symmetric in the indices
1, 2 and 3, showing that in fact O is the intersection of all three altitudes.

Theorem 87 (Affine orthocenter) Suppose the quadrances of the triangle

A1 A2 A3 are Q1 , Q2 and Q3 as usual, and the quadrea is A. Then the orthocenter O
can be written as the affine combination

O = β 1 A1 + β 2 A2 + β 3 A3

where

β1 ≡ (Q3 + Q1 − Q2 ) (Q1 + Q2 − Q3 ) /A
β2 ≡ (Q1 + Q2 − Q3 ) (Q2 + Q3 − Q1 ) /A
β3 ≡ (Q2 + Q3 − Q1 ) (Q3 + Q1 − Q2 ) /A.

Proof. This is a straightforward but laborious verification using the previous theorem.
Note that Exercise 5.15 (page 69) shows that β 1 + β 2 + β 3 = 1.
150 12. CENTERS OF TRIANGLES

12.5 Incenters

Theorem 88 (Incenter) Suppose that in the triangle A1 A2 A3 the vertices at A1

and A2 are bisected by the lines h1 and h2 respectively. Then h1 and h2 intersect at a
point I, and A3 I is a bisector of the vertex at A3 .

Proof. By the Spread reflection theorem (page 95) the line l2 ≡ A1 A3 is the reflection
of the line l3 ≡ A1 A2 in the bisector h1 , and l1 ≡ A2 A3 is the reflection of l3 in the
bisector h2 . If h1 and h2 are parallel, then it follows that l2 and l1 are parallel, which
they are not, since they intersect at A3 . So h1 and h2 intersect, say at the point I.

The Equal quadrance to two lines theorem (page 88) shows that

Q (I, l2 ) = Q (I, l3 )
Q (I, l1 ) = Q (I, l3 )

so that Q (I, l1 ) = Q (I, l2 ). That means that A3 I is a bisector of the vertex at A3 .

A3
I2 I1
I0
A1 A2

I3

Figure 12.3: Incenters of a triangle

From the Vertex bisector theorem (page 85), if a vertex has a bisector, then it has
exactly two bisectors and they are perpendicular. So there are four possible choices for
the ordered pair [h1 , h2 ] where h1 is a bisector of the vertex at A1 and h2 is a bisector
of the vertex at A2 . This gives in general four intersection points I0 , I1 , I2 and I3 of
the possible bisectors, the incenters of A1 A2 A3 . If one incenter exists, so do all four.

Each of these four incenters lies on a bisector of each of the three vertices of A1 A2 A3 .
By the Equal quadrance to two lines theorem the quadrances from a particular
incenter Ii to each of the three lines of the triangle are equal, and have the common
value Ri , the associated inquadrance of the triangle A1 A2 A3 . In general there are
four inquadrances, R0 , R1 , R2 and R3 .

The following theorem is surprisingly tricky to prove, especially when compared to the
12.5. INCENTERS 151

usual decimal number result involving angles. One should remember that the version
here covers the case of all four incenters at once.

Theorem 89 (Incenter spread) Suppose that I is an incenter of A1 A2 A3 , with

s (A1 I, A1 A2 ) ≡ u1 , s (A2 I, A2 A3 ) ≡ u2 and s (A3 I, A3 A1 ) ≡ u3 . Then

s (IA1 , IA2 ) = 1 − u3 . (12.3)

A2
u
l3 u2 2
h2
l1
h1 I h3
u1 u3
A1 u1 u3 A3
l2

Figure 12.4: Incenter spread theorem

Proof. (Using a computer) To simplify matters, suppose that the coordinates of all
points have been translated so that I ≡ [0, 0], and that A1 ≡ [x1 , y1 ] and A2 ≡ [x2 , y2 ].
The general case follows the same lines. Suppose that l1 , l2 and l3 are the usual lines of
the triangle A1 A2 A3 , and that the vertex bisectors through I and A1 and A2 are
respectively h1 and h2 . Then using the Line through two points theorem (page 38)

l3 = A1 A2 = hy1 − y2 : x2 − x1 : x1 y2 − x2 y1 i
h1 = IA1 = hy1 : −x1 : 0i
h2 = IA2 = hy2 : −x2 : 0i .

Since l2 is the reflection of l3 in h1 , the Reflection of a line in a line theorem (page 54)
gives
¿ ¡ 2 ¢ ¡ 2 ¢ À
y1 − x21 (y1 − y2 ) − 2y1 x1 (x 2
¡ 22 − x12)¢: −2y1 x1 (y1 − y2 ) − y1 − x1 (x2 − x1 )
l2 =
− y1 + x1 (x1 y2 − x2 y1 )

Similarly
¿ ¡ 2 ¢ ¡ 2 ¢ À
y2 − x22 (y2 − y1 ) − 2y2 x2 (x
¡
2
1 − x2 )¢: −2y2 x2 (y2 − y1 ) − y2 − x2 (x1 − x2 )
l1 = .
− y22 + x22 (x2 y1 − x1 y2 )

Now use the Point on two lines theorem (page 40) to find A3 ≡ l1 l2 ≡ [x3 , y3 ].
152 12. CENTERS OF TRIANGLES

The result is
¡ 2 ¢
x1 y2 − x22 y1 + y12 y2 − y1 y22 (x1 y2 − x2 y1 )
x3 = ³ ´
(x2 − x1 )2 + (y2 − y1 )2 (x1 x2 + y1 y2 )
¡ ¢
x1 x2 − x21 x2 + x1 y22 − x2 y12 (x1 y2 − x2 y1 )
y3 = ³ 2 ´ .
(x2 − x1 )2 + (y2 − y1 )2 (x1 x2 + y1 y2 )

Then use a computer to calculate s (A3 I, A3 A1 ) using the Spread from points theorem
(page 84). The result is
2
(x1 x2 + y1 y2 )
s (A3 I, A3 A1 ) = u3 = .
(x21 + y12 ) (x22 + y22 )
By Fibonacci’s identity
2
(x1 y2 − x2 y1 )
1 − u3 = = s (IA1 , IA2 ) .
(x21 + y12 ) (x22 + y22 )

Note that since the triangle A1 A2 I has spreads u1 , u2 and 1 − u3 , the Triple spread
theorem shows that {u1 , u2 , 1 − u3 } is a spread triple.

Exercise 12.6 Verify that for any a, b and c

¡ ¢
S (a, b, 1 − c) = 2 (a + b + c) − 2 a2 + b2 + c2 − 1 − S (a, b, c)

and hence deduce that

S (a, b, 1 − c) = S (a, 1 − b, c) = S (1 − a, b, c) . ¦

Exercise 12.7 With notation as in the previous theorem, show that if u1 6= 1/2 then

u1 − 2u2 − 2u3 + 2u22 + 2u23

r1 = s (A1 I, A2 A3 ) = . ¦
2u1 − 1

Example 12.1 Suppose that the field is F13 , and that A1 ≡ [3, 7] and A2 ≡ [−2, −4]
with I ≡ [0, 0] . Then proceeding as in the previous theorem, A3 = [5, 7] and

u1 = 6 u2 = 7 u3 = 4.

You may then check that {6, 7, −3}, {−5, 7, 4} and {6, −6, 4} are all spread triples.
The spread between A1 I = h2 : 1 : 0i and A2 A3 = h4 : 1 : −1i is

(2 × 1 − 4 × 1)2 6 − 2 × 7 − 2 × 4 + 2 × 72 + 2 × 42
r1 = 2 1 2 2
=8=
(2 + 1 ) (4 + 1 ) 2×6−1
as stated in the previous exercise. While r3 may be determined in a similar way, some
care must be taken with r2 since 7 = 1/2. ¦
12.5. INCENTERS 153

Theorem 90 (Inquadrance) Suppose that I is an incenter of A1 A2 A3 , and that

s (A1 I, A1 A2 ) ≡ u1 , s (A2 I, A2 A3 ) ≡ u2 and s (A3 I, A3 A1 ) ≡ u3 . Suppose the
quadrances of A1 A2 A3 are Q1 , Q2 and Q3 as usual, and the circumquadrance is K.
Then the inquadrance from I to each of the lines A1 A2 , A2 A3 and A1 A3 is

R = 16Ku1 u2 u3 .

Proof. Since I lies on a vertex bisector of each of the three vertices of A1 A2 A3 , the
quadrance from I to the three lines of that triangle are equal. From the previous
theorem s (IA1 , IA2 ) = 1 − u3 , so the Spread law in the triangle A1 A2 I gives
u1 Q3
Q (A2 , I) = .
1 − u3
The quadrance from I to the line A1 A2 is then
u1 u2 Q3
R = u2 Q (A2 , I) =
1 − u3
4u1 u2 u3 Q3 4Q3
= = u1 u2 u3
4u3 (1 − u3 ) s3
= 16Ku1 u2 u3

by the Extended spread law (page 144).

Exercise 12.8 Suppose that A1 A2 A3 has incenters I0 , I1 , I2 and I3 , with the

respective inquadrances R0 , R1 , R2 and R3 . If A is the quadrea of the triangle
A1 A2 A3 then show that µ ¶2
A
R0 R1 R2 R3 = . ¦
16

Example 12.2 The following example works in the field of rational complex
numbers–complex numbers a + bi where a and b are rational numbers. Suppose that
A1 A2 A3 is a triangle with
A1 ≡ [1, 1 + i] and A2 ≡ [1 − i, −2i]

and with incenter O ≡ [0, 0]. Then using the proof of the Incenter spread theorem
∙ ¸
16 20 46 10
A3 = − + i, − i .
39 117 117 39

The quadrea of A1 A2 A3 is
99 328 400 768
A=− + i
13 689 4563
and the circumquadrance is
25 25
K= − i.
64 96
154 12. CENTERS OF TRIANGLES

The spreads u1 , u2 and u3 are

32 56
u1 = − − i
195 195
4 32
u2 = − + i
195 195
9
u3 = .
5
The equations of the three vertex bisectors meeting at I are

h1 ≡ A1 I = h1 + i : −1 : 0i
h2 ≡ A2 I = h2i : 1 − i : 0i
h3 ≡ A3 I = h46 − 30i : 48 − 20i : 0i .

The other vertex bisectors through A1 , A2 and A3 respectively are perpendicular to

these, so by the Altitude to a line theorem (page 41) they can be determined to be

j1 = h1 : 1 + i : −1 − 2ii
j2 = h1 − i : −2i : 4 + 2ii
j3 = h72 − 30i : −69 + 45i : 40 − 60ii .

The other incenters are

∙ ¸
1 2 4
I1 = h1 j2 = h1 j3 = − − i, − i
3 3 3
∙ ¸
1 2 1
I2 = h2 j1 = h2 j3 = + i, 1 + i
3 3 3
∙ ¸
5 1 5
I3 = h3 j1 = h3 j2 = − + i, − i
2 2 2

and the inquadrances are

16 8
R0 = − i
39 13
688 8
R1 = + i
351 13
92 136
R2 = − − i
351 117
2287 211
R3 = − i.
156 26
These values may be used to check some of the formulas of this section. ¦
 13

Isometries

Transformations which preserve quadrance are particularly important for geometry.

This chapter studies and classifies these isometries, including translations, central
reflections and central rotations. Solutions to the equation q 2 + r2 = 1 naturally
become involved.

13.1 Translations, rotations, reflections

Definition A transformation σ is an isometry precisely when it is both invertible

and
Q (A1 , A2 ) = Q (σ (A1 ) , σ (A2 ))
for any points A1 and A2 . An isometry that fixes the origin O ≡ [0, 0] is an
orthogonality.

Exercise 13.1 Show that if σ is an isometry, then so is the inverse function σ −1 , and
that if σ 1 and σ 2 are isometries, then so is the composition σ1 ◦ σ 2 = σ 1 σ 2 . ¦

Exercise 13.2 Show that for any point A the rotation ρA in A (page 52) is an
isometry. Show that for any non-null line l the reflection σ l in l (page 52) is an
isometry. ¦

Example 13.1 For numbers u and v the transformation τ u,v defined by

τ u,v ([x, y]) ≡ [x + u, y + v]
is a translation. The inverse of τ u,v is τ −u,−v , and for any points A1 and A2
Q (A1 , A2 ) = Q (τ u,v (A1 ) , τ u,v (A2 ))
so τ u,v is an isometry. If B ≡ [u, v] then τ u,v is also denoted by τ B . ¦

155
156 13. ISOMETRIES

Example 13.2 For any numbers q and r satisfying q 2 + r2 = 1, the transformation

σ q,r defined by
σ q,r ([x, y]) ≡ [qx + ry, rx − qy]
is a central reflection. It is an orthogonality, since its inverse is itself, it fixes the
origin, and for any two points A1 ≡ [x1 , y1 ] and A2 ≡ [x2 , y2 ]
Q (σ q,r (A1 ) , σ q,r (A2 ))
= Q ([qx1 + ry1 , rx1 − qy1 ] , [qx2 + ry2 , rx2 − qy2 ])
2 2
= (q (x2 − x1 ) + r (y2 − y1 )) + (r (x2 − x1 ) − q (y2 − y1 ))
¡ ¢³ 2 2
´
= q 2 + r2 (x2 − x1 ) + (y2 − y1 )
= Q (A1 , A2 ) . ¦

Exercise 13.3 Show that if l is a non-null central line (passing through O ≡ [0, 0]),
then the reflection σ l in l is a central reflection, and that every central reflection is of
the form σ l for some unique central line l. ¦

Example 13.3 For any numbers q and r satisfying q 2 + r2 = 1, the transformation

ρq,r defined by
ρq,r ([x, y]) ≡ [qx − ry, rx + qy]
is a central rotation. The inverse of ρq,r is ρq,−r , and for any two points
A1 ≡ [x1 , y1 ] and A2 ≡ [x2 , y2 ]
¡ ¢
Q ρq,r (A1 ) , ρq,r (A2 )
= Q ([qx1 − ry1 , rx1 + qy1 ] , [qx2 − ry2 , rx2 + qy2 ])
= (q (x2 − x1 ) + r (y1 − y2 ))2 + (r (x2 − x1 ) − q (y1 − y2 ))2
¡ ¢³ ´
= q 2 + r2 (x2 − x1 )2 + (y2 − y1 )2
= Q (A1 , A2 )
so since ρq,r fixes the origin, it is an orthogonality. ¦

Exercise 13.4 Show that an isometry cannot be simultaneously both a central

rotation and a central reflection. Which of the two is the identity transformation ι? ¦

Exercise 13.5 Show that for numbers q and r satisfying q 2 + r2 = 1, and numbers t
and u satisfying t2 + u2 = 1,
σ t,u σ q,r = ρv,w
where
v = qt + ru and w = qu − rt
while
ρq,r ρt,u = ρt,u ρq,r = ρl,m
where
l = qt − ru and m = qu + rt. ¦

In Chapter 16, the Unit circle theorem describes all ordered pairs [q, r] satisfying
q 2 + r2 = 1, and the previous exercise is related to multiplication of complex numbers.
13.2. CLASSIFYING ISOMETRIES 157

13.2 Classifying isometries

Theorem 91 (Isometries preserve collinearity) Suppose that σ is an isometry.

Then A1 , A2 and A3 are collinear points precisely when σ (A1 ), σ (A2 ) and σ (A3 ) are
collinear points.

Proof. It follows from the Quadrea theorem (page 68) that three points A1 , A2 and
A3 forming quadrances Q1 ≡ Q (A2 , A3 ) , Q2 ≡ Q (A1 , A3 ) and Q3 ≡ Q (A1 , A2 ) are
collinear precisely when A ≡ A (Q1 , Q2 , Q3 ) = 0. Now since an isometry preserves
quadrances, this is equivalent to the condition that σ (A1 ), σ (A2 ) and σ (A3 ) are
collinear.

Theorem 92 (Specifying isometries) An isometry is completely determined by

its action on the points O ≡ [0, 0], I1 ≡ [1, 0] and I2 ≡ [0, 1].

Proof. Suppose that σ 1 and σ2 are both isometries which agree on the three points
O ≡ [0, 0], I1 ≡ [1, 0] and I2 ≡ [0, 1]. Then the isometry σ ≡ σ −1
2 σ 1 fixes all three
points, and it suffices to show that σ must be the identity isometry ι, by showing that
σ fixes all other points too.

For a point A ≡ [x, y], suppose that σ (A) ≡ [u, v]. Then the quadrances from A and
σ (A) to the points O, I1 and I2 must be respectively equal, so that

x2 + y 2 = u2 + v 2 (13.1)
2 2 2 2
(x − 1) + y = (u − 1) + v (13.2)
2 2 2 2
x + (y − 1) = u + (v − 1) . (13.3)

Now (13.1) minus (13.2) gives 2x − 1 = 2u − 1, so that x = u, and similarly (13.1)

minus (13.3) leads to y = v. Thus σ (A) = A and so σ = ι.

Theorem 93 (Orthogonality) An orthogonality is either a central rotation or a

central reflection.

Proof. Suppose that µ is an orthogonality with µ (I1 ) ≡ D1 ≡ [q, r] and µ (I2 ) ≡ D2 .

Then since Q (O, I1 ) = 1,
Q (O, D1 ) = q 2 + r2 = 1.
158 13. ISOMETRIES

The quadrances in triangle OI1 I2 are Q1 = Q2 = 1 and Q3 = 2 so by Pythagoras’

theorem, OI1 is perpendicular to OI2 . Since the isometry µ preserves quadrances, the
quadrances in triangle OD1 D2 are then also Q1 = Q2 = 1 and Q3 = 2, so that OD1 is
perpendicular to OD2 . That means D2 is a multiple of [r, −q], and since Q (O, D2 ) = 1
the only two possibilities are
D2 = [r, −q]
or
D2 = [−r, q] .
By the Specifying isometries theorem, this information then determines µ. In the first
case µ agrees with the central reflection σ q,r , and in the second case µ agrees with the
central rotation ρq,r .

Theorem 94 (Classification of isometries) An isometry σ can be uniquely

expressed as
σ = τµ
where µ is an orthogonality and τ is a translation.

Proof. Suppose that the isometry σ satisfies σ (O) ≡ B. If τ ≡ τ B then τ −1 σ fixes O,

so is an orthogonality µ, and σ = τ µ. To show uniqueness, suppose also that σ = τ 0 µ0
for some translation τ 0 and some orthogonality µ0 . Then

B = σ (O) = (τ 0 µ0 ) (O) = τ 0 (O) .

Since a translation is determined by its effect on one point, τ 0 = τ . But then

µ0 = τ −1 σ = µ. So the expression for σ is unique.

Exercise 13.6 Show that the rotation ρO in the origin O ≡ [0, 0] is a central rotation,
and that ρO = ρ−1,0 . Show that for any point A, ρA = τ B ρO where

B ≡ ρA (O) . ¦

Exercise 13.7 Show that if l ≡ ha : b : ci then σ l = τ B σ k where

∙ ¸
−2ac −2bc
B≡ 2 , = σ l (O)
a + b2 a2 + b2

and where σ k is the central reflection in the line k ≡ ha : b : 0i, given by

"¡ ¢ ¡ ¢ #
b2 − a2 x − 2aby −2abx + a2 − b2 y
σ k ([x, y]) = , . ¦
a2 + b2 a2 + b2
 14

Regular stars and polygons

This chapter introduces regular stars and polygons, which illustrate cyclical and
dihedral symmetry in the plane, and occur often in applications over the decimal
numbers. The existence of regular stars of a given order n is a number theoretical
issue, analysed here for the cases n = 3, 5 and 7. Regular polygons are constructed
from regular stars by successive reflections of a point on one of the lines of the star,
motivating an unorthodox labelling of vertices. The case of the regular pentagon
receives special attention. The lineations Σl defined in Chapter 4 play a major role.

14.1 Regular stars

Definition A regular star of orderP n is a list of n distinct mutually non-parallel

lines [l0 , l1 , · · · , ln−1 ] such that lk+1 = lk (lk−1 ) for all k, with the convention that
lk ≡ lk+n for all k, and the conventions that

[l0 , l1 , · · · , ln−2 , ln−1 ] ≡ [l1 , l2 , · · · , ln−1 , l0 ]

[l0 , l1 , · · · , ln−2 , ln−1 ] ≡ [ln−1 , ln−2 , · · · , l1 , l0 ] .

Any spread s of
the form s (li , lj ), with li and lj lines in the star, is a spread of the
star. If s ≡ s (l1 , l2 ) then the spreads of the star are all of the form
Sk (s) for some k, and so contained in the spread sequence for s.

In particular Sn (s) = 0, and from the Cyclic

reflection theorem (page 56) Sn (r) = 0 for any spread r of the star.
Order seven star
159
160 14. REGULAR STARS AND POLYGONS

14.2 Order three stars

Theorem 95 (Order three star) A regular star of order three exists precisely
when the number 3 is a non-zero square.

Proof. Suppose that a regular star [l0 , l1 , l2 ] of order three exists. Then
s (l0 , l1 ) = s (l1 , l2 ) = s (l2 , l0 ) ≡ s for some non-zero number s. Since l0 , l1 and l2 are
distinct concurrent lines, no two can be parallel, so that the Three equal spreads
theorem (page 101) shows that
2
0 = s (l0 , l0 ) = S3 (s) = s (3 − 4s) .

Thus s = 3/4 is non-zero and by the Spread number theorem (page 76) is a spread
number, so that
s (1 − s) = 3/16
is a square. Thus 3 is both non-zero and a square.

Conversely suppose that 3 ≡ r2 for some non-zero number r. Then consider

l0 ≡ h0 : 1 : 0i l1 ≡ hr : 1 : 0i l2 ≡ h−r : 1 : 0i .

Since
r2 3 (2r)2
s (l0 , l1 ) = s (l0 , l2 ) = = = 2 = s (l2 , l3 )
r2 + 1 4 (r2 + 1)
it follows from the Spread reflection theorem (page 95) that [l0 , l1 , l2 ] is a regular star
of order three.

Example 14.1 In the fields F3 , F5 , F7 , F17 and F19 , the number 3 is not a non-zero
square, so there are no regular stars of order three. ¦

Example 14.2 In the field F11 , 3 = 52 , so an example of a regular star of order three
is [l0 , l1 , l2 ], where

l0 ≡ h0 : 1 : 0i l1 ≡ h5 : 1 : 0i l2 ≡ h6 : 1 : 0i . ¦

Example 14.3 In the field F13 , 3 = 42 , so an example of a regular star of order three
is [l0 , l1 , l2 ], where

l0 ≡ h0 : 1 : 0i l1 ≡ h4 : 1 : 0i l2 ≡ h9 : 1 : 0i . ¦

Example 14.4 In the rational number field, 3 is not a square, so regular stars of
¡√ ¢2
order three do not exist. However in the decimal number field 3 = 3 , so an
example of a regular star of order three is [l0 , l1 , l2 ], where
­√ ® ­ √ ®
l0 ≡ h0 : 1 : 0i l1 ≡ 3:1:0 l2 ≡ − 3 : 1 : 0 . ¦
14.3. ORDER FIVE STARS 161

14.3 Order five stars

Theorem 96 (Order five star) A regular star of order five exists precisely when
there is a non-zero number r satisfying the conditions i) r2 = 5 and ii) 2 (5 − r) is a
square.

Proof. Suppose that a regular star [l0 , l1 , l2 , l3 , l4 ] of order five exists with
s (l1 , l2 ) ≡ s 6= 0. By the Consecutive spreads theorem (page 103)
¡ ¢2
S5 (s) = s 5 − 20s + 16s2 = 0. (14.1)
Thus s is non-zero and satisfies the quadratic equation 5 − 20s + 16s2 = 0, so that also
s 6= 1. Complete the square to obtain
(s − 5/8)2 = 5/64.
A solution s exists precisely when 5 is a square, so that s = (5 + r) /8 for some number
r satisfying r2 = 5. If r = 0 then 5 = 0 so that s = 0 which is impossible. Since s is a
spread number,
s (1 − s) = (5 + r) (3 − r) /64 = (10 − 2r) /64
is a square, so that 2 (5 − r) is a square. Thus r is non-zero and satisfies i) and ii).

Conversely suppose that there is some non-zero number r with the properties that i)
r2 = 5 and ii) 2 (5 − r) = v 2 for some number v. Then you may check that
l0 ≡ h0 : 1 : 0i l1 ≡ hv : 3 − r : 0i l2 ≡ hv (3 − r) : 2 − 2r : 0i
l3 ≡ hv (r − 3) : 2 − 2r : 0i l4 ≡ h−v : 3 − r : 0i
defines a regular star [l0 , l1 , l2 , l3 , l4 ] of order five.

Example 14.5 In the fields F3 , F5 , F7 , F13 and F17 the equation S5 (s) = 0 has no
non-zero solutions, and so there are no regular stars of order 5. ¦

Example 14.6 In the field F11 the equation S5 (s) = 0 has exactly two non-zero
solutions s = 7 = (5 + 7) /8 and s = 8 = (5 + 4) /8. But the spread numbers in F11 are
0, 1, 2, 3, 6, 9 or 10, so there are no regular stars of order 5. Equivalently the numbers
2 (5 − 7) = 7 and 2 (5 − 4) = 2 are not squares. ¦

Example 14.7 In the field F19 the equation S5 (s) = 0 has exactly two non-zero
solutions s = 9 = (5 + 10) /8 and s = 16 = (5 + 9) /8. These are spread numbers, or
equivalently the numbers 2 (5 − 10) = 9 and 2 (5 − 9) = 11 are squares. An example of
a regular star [l0 , l1 , l2 , l3 , l4 ] which has these spreads is
l0 ≡ h0 : 1 : 0i l1 ≡ h2 : 1 : 0i l2 ≡ h5 : 1 : 0i
l3 ≡ h−5 : 1 : 0i l4 ≡ h−2 : 1 : 0i ¦

A Go board has dimensions 19 × 19, the smallest square grid for which five fold
symmetry exists in the above sense.
162 14. REGULAR STARS AND POLYGONS

14.4 Order seven stars

The analysis of the previous two sections does not easily continue for higher order
stars. However one can still make statements of the following kind.

Theorem 97 (Order seven star) A regular star of order seven exists precisely
when there is a non-zero number s for which

7 − 56s + 112s2 − 64s3 = 0

and such that s (1 − s) is a square.

Proof. Suppose that a regular star [l0 , l1 , · · · , l6 ] exists with s (l0 , l1 ) ≡ s 6= 0. By the
Consecutive spreads theorem
¡ ¢2
S7 (s) = s 7 − 56s + 112s2 − 64s3 = 0. (14.2)
Thus s is non-zero, satisfies the cubic equation 7 − 56s + 112s2 − 64s3 = 0, and because
it is a spread number, satisfies the condition that s (1 − s) is a square.

Conversely suppose that s is a non-zero spread number satisfying (14.2). There are two
lines l0 and l1 with spread s (l0 , l1 ) = s, and suppose that they intersect at the point P.
Using the rule lk+1 = Σlk (lk−1 ) the reflection sequence
· · · l−1 , l0 , l1 , l2 , · · ·
may be obtained, and since S7 (s) = 0 you know that l7 = l0 , since it is both parallel to
l1 and also passes through P. Thus ln+7 = ln for any n. If any two of the lines
l0 , l1 , · · · , l6 are the same, then the reflection sequence satisfies ln+k = ln for any n for
some positive integer 1 ≤ k < 7. But then since k and 7 are relatively prime, it follows
that l0 = l1 , which is impossible. Thus [l0 , l1 , · · · , l6 ] is a regular star of order seven.

Example 14.8 In the fields F3 , F5 , F7 , F11 , F17 , F19 and F23 the equation
¡ ¢2
S7 (s) = s 7 − 56s + 112s2 − 64s3 = 0 has no non-zero solutions, and so there are no
regular stars of order 7. In the field F13 the equation S7 (s) = 0 has exactly three
non-zero solutions s = 2, s = 5 and s = 11. But the spread numbers in F13 are
0, 1, 4, 6, 7, 8 or 10, so there are no regular stars of order 7. ¦

Example 14.9 In the field F29 the equation S7 (s) = 0 has exactly three non-zero
solutions s = 6, s = 7 and s = 25. These are spread numbers, and the regular star
[l0 , l1 , · · · , l6 ] where
l0 ≡ h0 : 1 : 0i l1 ≡ h15 : 1 : 0i l2 ≡ h11 : 1 : 0i l3 ≡ h20 : 1 : 0i
l4 ≡ h9 : 1 : 0i l5 ≡ h18 : 1 : 0i l6 ≡ h14 : 1 : 0i
has spreads 0, 6, 25 and 7. ¦
14.5. REGULAR POLYGONS 163

Example 14.10 In the rational number field (14.2) has no solutions, so there are no
regular stars of order seven. In the decimal number field regular stars of order seven
exist, as the equation S7 (s) = 0 has non-zero solutions

γ 1 ≈ 0.188 255 099 . . . γ 2 ≈ 0.611 260 467 . . . γ 3 ≈ 0.950 484 434 . . . . ¦

14.5 Regular polygons

Suppose that [l0 , l1 , · · · , ln−1 ] is a regular star of order n with common intersection P.
Adopt the convention that lk+n ≡ lk for any integer k, so that · · · l−1 , l0 , l1 , · · · is a
reflection sequence of lines. Then by the Cyclic reflection theorem (page 56), for any
integers k and l
Σlk (lj ) = l2k−j .
The Reflection theorem (page 55) implies that

σ l2k−j = σlk σ lj σ lk .

In particular
σ l2 = σ l1 σ l0 σ l1 σl3 = σ l2 σ l1 σ l2
and so on. From this and the fact that σ −1
l = σ l for all l, it follows that

σ l2 σ l0 = σ l2 σ l1 σ l2 σ l1 = σl3 σ l1

and so on.

Exercise 14.1 Extend the argument to show that for any integers j, k and m

σ lj σ lk = σ lj+m σlk+m . ¦

Choose a point A0 on l0 distinct from P. Define the sequence of points

A0 , A2 ≡ σ l1 (A0 ) , A4 ≡ σ l3 (A2 ) , A6 ≡ σ l5 (A4 ) , · · ·

and so on, so that A2k lies on l2k by induction. Using the above relations you find that

A4 = σl3 σ l1 (A0 ) = σ l2 σl0 (A0 ) = σ l2 (A0 )

A6 = σl5 σ l2 (A0 ) = σ l3 σl0 (A0 ) = σ l3 (A0 )
A8 = σl7 σ l3 (A0 ) = σ l4 σl0 (A0 ) = σ l4 (A0 )

and so on. Thus for any positive integer k,

A2k = σ lk (A0 ) .

In particular
A2n = σ ln (A0 ) = σ l0 (A0 ) = A0 .
164 14. REGULAR STARS AND POLYGONS

The n-gon A0 A2 · · · A2(n−1) constructed this way is called regular. Note that
¡ ¢
Q (P, A0 ) = Q (P, A2 ) = · · · = Q P, A2(n−1) .

In the case of n odd, the points A0 , A2 , · · · , A2(n−1) are all distinct because they lie on
distinct lines.

In the case of n = 2m even, the points Ak and Ak+n lie on the same line lk , and
Ak+n = σlk+m (Ak ) 6= Ak since σ lk+m acts on the points lying on lk as the rotation in
P, which fixes only P. Thus in this case the points A0 , A2 , · · · , A2(n−1) are also distinct.

Example 14.11 If [l0 , l1 , l2 , l3 , l4 ] is a regular star of order five, then for any point A0
on l0 distinct from the common intersection P of the lines, one obtains by this
construction a regular 5-gon, or pentagon, A0 A2 A4 A6 A8 as shown in Figure 14.1.

l3 l2
A2
l4 l1
A4
A0
P l0

A6 A8

Figure 14.1: Regular pentagon ¦

Example 14.12 If [l0 , l1 , l2 , l3 , l4 , l5 ] is a regular star of order six, then for any point
A0 on l0 distinct from the common intersection P of the lines, one obtains by this
construction a regular 6-gon, or hexagon, A0 A2 A4 A6 A8 A10 as shown in Figure 14.2.

l3 l2
l4 A2
A4
l5 l1
A0
A6 l0
P

A10
A8

Figure 14.2: Regular hexagon ¦

If A0 A2 · · · A2(n−1) is a regular n-gon of order n constructed from a regular star

[l0 , l1 , · · · , ln−1 ] of order n, then a triangle of three successive points such as A0 A2 A4 ,
A2 A4 A6 , · · · , A2(n−1) A0 A2 is a corner triangle of the n-gon.
14.5. REGULAR POLYGONS 165

Theorem 98 (Polygon triangle) Suppose that A0 A2 · · · A2(n−1) is a regular n-gon

constructed from a regular star [l0 , l1 , · · · , ln−1 ] of order n with common intersection
point P. Suppose that s ≡ s (l0 , l1 ) = s (l1 , l2 ) = · · · and that Q (P, A2k ) ≡ Q for all k.
Then the quadrances and spreads of the corner triangle A0 A2 A4 are

R ≡ Q (A0 , A2 ) = Q (A2 , A4 ) = 4sQ

D ≡ Q (A0 , A4 ) = 16s (1 − s) Q

and

s (A0 A4 , A0 A2 ) = s (A4 A0 , A4 A2 ) = s
s (A2 A0 , A2 A4 ) = 4s (1 − s) .

l3
l4 l2
A4 R A2
D l1
Q R
Q s
s
l0
P Q A0

Figure 14.3: Polygon triangle theorem

Proof. Since s (l0 , l1 ) = s (l1 , l2 ) ≡ s, the Isosceles reflection theorem (page 124) shows
that
R ≡ Q (A0 , A2 ) = Q (A2 , A4 ) = 4sQ.
Similarly since
s (l0 , l2 ) = s (l2 , l4 ) = S2 (s) = 4s (1 − s)
it follows that

D ≡ Q (A0 , A4 ) = 4 × 4s (1 − s) Q = 16s (1 − s) Q.

Now in A0 A2 A4 use the Isosceles triangle theorem (page 122) to get

s (A0 A4 , A0 A2 ) = s (A4 A0 , A4 A2 ) = 1 − D/ (4R) .

= 1 − (1 − s) = s.

Then use the Spread law to see that

s (A2 A0 , A2 A4 ) = 4s (1 − s) .
166 14. REGULAR STARS AND POLYGONS

Exercise 14.2 Extend this theorem to show that if a regular n-gon A0 A2 · · · A2(n−1)
is constructed from a regular star [l0 , l1 , · · · , ln−1 ] of order n with common
intersection P and with s ≡ s (l0 , l1 ) = s (l1 , l2 ) = · · · as above, and that
Q (P, Ak ) ≡ Q for all k, then

Q (A0 , A2k ) = 4Sk (s) Q

for all k = 0, 1, · · · , n − 1. This gives a geometric interpretation of Sk (s). ¦

Exercise 14.3 Show that in the decimal number system the non-trivial zeroes of
¡ ¢2
S5 (s) = s 5 − 20s + 16s2 = 0
are
¡ √ ¢ ¡ √ ¢
α ≡ 5 − 5 /8 ≈ 0.345 491 . . . and β ≡ 5 + 5 /8 ≈ 0.904 508 . . . .
With notation as in Figure 14.4, show that if Q (A, B) ≡ R is the common quadrance
of the sides, then
Q (P, A) = (1 − α) R/β = R/4α Q (P, M ) = (1 − α) R/4α
Q (B, F ) = (1 − α) R Q (B, E) = 4 (1 − α) R = βR/α.

B
R
C a
M
b a
R a 1-a
b
P F
A
D
E

Figure 14.4: Spreads in a pentagon ¦

Exercise 14.4 Using the notation of the previous exercise, show that the proportion
β : α is equal to both s (P A, P B) : s (P B, P E) and Q (B, E) : Q (A, B). ¦

Exercise 14.5 (Harder) Show that in F19 the five points

A0 ≡ [2, −4] A2 ≡ [1, 0] A4 ≡ [2, 4] A6 ≡ [7, −3] A8 ≡ [7, 3]
form a regular pentagon A0 A2 A4 A6 A8 with center O ≡ [0, 0]. Show that the
intersections of the diagonals of this pentagon yield five new points, also forming a
regular pentagon, and that by repeating this operation a total of eighteen times, the
original pentagon is recovered. ¦
 15

Conics

After points and lines and combinations of them, the next natural objects of study in
geometry are conics, given by quadratic equations. In the setting of universal
geometry, there are interesting families of conics which can be defined metrically, such
as circles, ribbons, parabolas, quadrolas and grammolas. This raises perhaps the first
serious problem of metrical algebraic geometry: How to classify conics? No answer is
given here. Much remains to be discovered about these intriguing curves.

15.1 Centers of conics

Definition A conic q ≡ hd : e : f : a : b : ci is a 6-proportion, enclosed in pointed

brackets, such that at least one of d, e or f is non-zero.

The conic q ≡ hd : e : f : a : b : ci will sometimes also be written in the form

* d : e : f +
q= a : b .
c

Definition A point X ≡ [x, y] lies on the conic q ≡ hd : e : f : a : b : ci, or

equivalently q passes through X, precisely when

dx2 + exy + f y 2 + ax + by + c = 0.

This is an equation of the conic q, as is any equivalent equation derived from it using
the usual rules for manipulating equations. A conic q is standard precisely when it
passes through the origin O ≡ [0, 0]. A conic q is empty if no point lies on q.

167
168 15. CONICS

In studying conics, looking for symmetry is useful. Recall from page 52 that if
C ≡ [u, v], then the rotation in C is defined by ρC ([x, y]) ≡ [2u − x, 2v − y].

Definition A point C ≡ [u, v] is a center of the conic q ≡ hd : e : f : a : b : ci

precisely when
dx2 + exy + f y 2 + ax + by + c
is unchanged by the substitutions x → 2u − x and y → 2v − y.

Example 15.1 The point C ≡ [1, 2] is a center for the conic h1 : 2 : −1 : −6 : 2 : 1i

since
(2 − x)2 + 2 (2 − x) (4 − y) − (4 − y)2 − 6 (2 − x) + 2 (4 − y) + 1
= x2 + 2xy − y 2 − 6x + 2y + 1. ¦

Theorem 99 (Conic center) The conic q ≡ hd : e : f : a : b : ci has a unique center

C precisely when e2 − 4df 6= 0, in which case
∙ ¸
2af − be 2bd − ae
C= 2 , .
e − 4df e2 − 4df

If e2 − 4df = 0 and e : 2f : b = 2d : e : a, then there is more than one center. If

e2 − 4df = 0 and e : 2f : b 6= 2d : e : a, then there is no center.

Proof. The condition for [u, v] to be a center of q ≡ hd : e : f : a : b : ci is

d (2u − x)2 + e (2u − x) (2v − y) + f (2v − y)2 + a (2u − x) + b (2v − y) + c
= dx2 + exy + f y 2 + ax + by + c
which simplifies to
x (2du + ev + a) + y (eu + 2f v + b) − (2du + ev + a) u − (eu + 2f v + b) v = 0.
This is satisfied precisely when the coefficients of x and y are zero, that is when
eu + 2f v + b = 2du + ev + a = 0.
By page 30 this pair of linear equations in u and v has a unique solution precisely
when e2 − 4df 6= 0, in which case
2af − be
u =
e2 − 4df
2bd − ae
v = .
e2 − 4df
If e2 − 4df = 0 then there is more than one solution precisely when
e : 2f : b = 2d : e : a, and no solution otherwise.
15.2. CIRCLES AND RIBBONS 169

Exercise 15.1 Find all centers of the conics with equations x2 + 2xy + y 2 = 0 and
x2 + 2xy + y 2 + x = 0. ¦

Exercise 15.2 Show that if C1 and C2 are distinct centers for a conic q, then every
point lying on C1 C2 is a center for q. ¦

Definition A conic q is centered precisely when it has a center, and is a central

conic precisely when it has O ≡ [0, 0] as a center.

15.2 Circles and ribbons

Definition A circle c is a conic whose equation in X ≡ [x, y] has the form

Q (X, C) = K

for some fixed point C and some fixed number K.

It will be seen below that the point C and the number K are determined by the circle
c. The point C is the unique center of c, and the number K is the quadrance of c. A
circle c is central precisely when its center is O ≡ [0, 0]. A circle c is null precisely
when its quadrance is K ≡ 0.

Example 15.2 In the field F11 , Figure 15.1 shows the eleven central circles, in
particular the circles of quadrance one (circles) and quadrance two (gray squares).

5 6 8 1 7 4 3 4 7 1 8 6
4 8 10 3 9 6 5 6 9 3 10 8
3 1 3 7 2 10 9 10 2 7 3 1
2 7 9 2 8 5 4 5 8 2 9 7
1 4 6 10 5 2 1 2 5 10 6 4
0 3 5 9 4 1 0 1 4 9 5 3
-1 4 6 10 5 2 1 2 5 10 6 4
-2 7 9 2 8 5 4 5 8 2 9 7
-3 1 3 7 2 10 9 10 2 7 3 1
-4 8 10 3 9 6 5 6 9 3 10 8
-5 6 8 1 7 4 3 4 7 1 8 6
-5 -4 -3 -2 -1 0 1 2 3 4 5

Figure 15.1: Central circles in F11

170 15. CONICS

The number in any position is the quadrance of the circle on which it lies. Each of the
ten non-null circles has exactly 12 points lying on it, and the null circle has only one
point lying on it. Figure 15.2 is a view of the central circle of quadrance one on a
larger scale.

Figure 15.2: Circle of quadrance one in F11 ¦

Exercise 15.3 In the field F13 , Figure 15.3 shows the thirteen central circles, in
particular the circles of quadrance one (circles) and quadrance two (gray boxes), as
well as the central null circle (black squares). Each of the twelve non-null circles has
exactly 12 points lying on it, while the null circle has 25 points lying on it.

6 7 9 0 6 1 11 10 11 1 6 0 9 7
5 9 11 2 8 3 0 12 0 3 8 2 11 9
4 0 2 6 12 7 4 3 4 7 12 6 2 0
3 6 8 12 5 0 10 9 10 0 5 12 8 6
2 1 3 7 0 8 5 4 5 8 0 7 3 1
1 11 0 4 10 5 2 1 2 5 10 4 0 11
0 10 12 3 9 4 1 0 1 4 9 3 12 10
-1 11 0 4 10 5 2 1 2 5 10 4 0 11
-2 1 3 7 0 8 5 4 5 8 0 7 3 1
-3 6 8 12 5 0 10 9 10 0 5 12 8 6
-4 0 2 6 12 7 4 3 4 7 12 6 2 0
-5 9 11 2 8 3 0 12 0 3 8 2 11 9
-6 7 9 0 6 1 11 10 11 1 6 0 9 7
-6 -5 -4 -3 -2 -1 0 1 2 3 4 5 6

Figure 15.3: Central circles in F13

Figure 15.4 shows on the left a larger view of the central circle of quadrance one, and
on the right a larger view of the central null circle.
15.2. CIRCLES AND RIBBONS 171

Figure 15.4: Circles of quadrance one and zero in F13 ¦

Theorem 100 (Circle uniqueness) The center and quadrance of a circle c are
unique.

Proof. Suppose that C ≡ [u, v] is a fixed point and K a number. With X ≡ [x, y], the
equation
Q (X, C) = K
is
(x − u)2 + (y − v)2 = K
which defines the circle
­ ®
c ≡ 1 : 0 : 1 : −2u : −2v : u2 + v 2 − K .

With the proportion scaled so that the first and third entries are 1, the coordinates u
and v of the center C are immediate from the fourth and fifth entries, and then the
quadrance K is determined from the last entry.

Exercise 15.4 Show that q ≡ hd : e : f : a : b : ci is a circle precisely when e = 0 and

d = f 6= 0. ¦

Exercise 15.5 Show that in the field F13 every point lying on the unique central null
circle c lies either on the line h2 : 3 : 0i or the line h3 : 2 : 0i, and conversely that every
point lying on one of these lines also lies on c. Generalize. ¦

Exercise 15.6 Recall from page 67 that Q (X, l) denotes the quadrance from the
point X to the line l. A ribbon r is a conic whose equation in X ≡ [x, y] has the form

Q (X, l) = K

for some fixed non-null line l and some fixed number K. Show that l and K are unique,
and that every point lying on l is a center for r. ¦
172 15. CONICS

15.3 Parabolas

Definition A parabola p is a conic whose equation in X ≡ [x, y] has the form

Q (X, F ) = Q (X, l)

for some fixed point F and some fixed non-null line l not passing through F.

It will be seen below that the point F and the line l are determined by p. The point F
is the focus and the line l the directrix of p.

Example 15.3 If a 6= 0 then the parabola p with focus F ≡ [a, 0] and directrix
l ≡ h1 : 0 : ai has equation
(x − a)2 + y 2 = (x + a)2
or y 2 = 4ax. Thus p = h0 : 0 : 1 : −4a : 0 : 0i. ¦

Example 15.4 In the field F11 two views of the parabola p with equation

y 2 = 4x

are pictured in Figure 15.5 (squares), along with the focus F ≡ [1, 0] (dot) and the
directrix l ≡ h1 : 0 : 1i (gray boxes). Note that the parabola still deserves its name.

5
4
3
2
1
0
-1
-2
-3
-4
-5
-5 -4 -3 -2 -1 0 1 2 3 4 5

Figure 15.5: Parabola y 2 = 4x in F11 : two views

Exercise 15.7 Show that a parabola p does not have a center. In particular a
parabola cannot be a circle. ¦

Exercise 15.8 Show that a parabola has a unique axis, namely a line l for which the
reflection σ l preserves the equation of the conic. ¦

Exercise 15.9 Show that there is a unique point lying on both a parabola and its
axis, called the vertex. ¦
15.3. PARABOLAS 173

Theorem 101 (Parabola uniqueness) The focus and directrix of a parabola p are
unique.

Proof. If F ≡ [u, v] and l ≡ ha : b : ci are a corresponding focus and directrix for p

then using the Quadrance to a line theorem (page 67) the conic has equation
(ax + by + c)2
(x − u)2 + (y − v)2 = .
a2 + b2
After expansion this gives
* 2 2 +
¡ 2 b 2:¢−2ab : a ¡ 2 ¢
p= −2ac − 2 ¡ a + b ¢ ¡u : −2bc¢− 2 a + b2 v .
a2 + b2 u2 + v 2 − c2

Let’s see how to recover the point F and the line l from this proportion. Since
ha : b : ci is a line at least one of a and b is non-zero, and the proportion p determines
which is non-zero. Suppose that a 6= 0, so that without loss of generality a = 1 and
p = hD : E : 1 : A : B : Ci, with D, E, F, A, B and C now determined uniquely.

The number b can be recovered from −2b = E, and then there are three equations for
u, v and c, namely
¡ ¢
−2c − 2 1 + b2 u = A (15.1)
¡ ¢
−2bc − 2 1 + b2 v = B (15.2)
¡ ¢¡ ¢
1 + b2 u2 + v 2 − c2 = C. (15.3)
Use (15.1) and (15.2) to solve for u and v in terms of c
−2c − A
u =
2 (1 + b2 )
−2bc − B
v =
2 (1 + b2 )
so that (15.3) becomes
³ ´ ¡ ¢ ¡ ¢
(2c + A)2 + (2bc + B)2 − 4 1 + b2 c2 = 4 1 + b2 C.

Upon expansion however this becomes linear in c, namely

¡ ¢
A2 + B 2 + 4 (A + Bb) c = 4 1 + b2 C
giving a unique value of c provided that 4 (A + Bb) = 4A − 2BE 6= 0. Replace A, B
and E with the respective values above and simplify to see that this condition is
equivalent to ¡ ¢
1 + b2 (u + bv + c) 6= 0.
But 1 + b2 6= 0 because l is non-null, and u + bv + c 6= 0 since F does not lie on l, so
there is a unique value of c, and hence also of u and v.
174 15. CONICS

15.4 Quadrolas

Definition A quadrola q is a conic whose equation in X ≡ [x, y] has the form

A (Q (X, F1 ) , Q (X, F2 ) , K) = 0 (15.4)

for some distinct fixed points F1 and F2 , and some fixed non-zero number K, such
that Q (F1 , F2 ) is neither 0 nor K.

¡ ¢
Recall that A (x1 , x2 , x3 ) ≡ (x1 + x2 + x3 )2 − 2 x21 + x22 + x23 is Archimedes’ function.

The points F1 and F2 are foci, and the number K the associated quadrance of the
quadrola.

Exercise 15.10 (Harder) Show that if −1 is not a square then the foci and
quadrance of a quadrola are unique. ¦

Exercise 15.11 Show that the equation (15.4) is of degree two in x and y, so that a
quadrola is indeed a conic. ¦

Example 15.5 If a 6= 0 then the quadrola q with foci F1 ≡ [a, 0] and F2 ≡ [−a, 0] and
quadrance K has equation
¡ ¢ ¡ ¢
4 K − 4a2 x2 + 4Ky 2 = K K − 4a2 .
Note that Q (F1 , F2 ) = 4a2 . In the decimal number field if 4a2 < K then this is an
ellipse, if 0 < K < 4a2 then this is a hyperbola, and if K < 0 then this is an ‘empty
ellipse’. For a general field, if −1 = i2 then q also has foci G1 ≡ [0, ia] and
G2 ≡ [0, −ia] with associated quadrance L ≡ K − 4a2 . ¦

Example 15.6 In F11 suppose that F1 ≡ [1, 0], F2 ≡ [−1, 0] and K ≡ 1. Then two
views of the quadrola q ≡ h7 : 0 : 5 : 0 : 0 : 1i (large squares) are shown in Figure 15.6.

5
4
3
2
1
0
-1
-2
-3
-4
-5
-5 -4 -3 -2 -1 0 1 2 3 4 5

Figure 15.6: Quadrola 7x2 + 5y 2 + 1 = 0 in F11 : two views

Exercise 15.12 Show that a quadrola is a centered conic. ¦

Exercise 15.13 (Harder) Show that a quadrola has two axes (see Exercise 15.8). ¦
15.5. GRAMMOLAS 175

15.5 Grammolas

Definition A grammola g is a conic whose equation in X ≡ [x, y] has the form

Q (X, l1 ) + Q (X, l2 ) = K

for some fixed non-null lines l1 and l2 which are neither parallel nor perpendicular,
and some non-zero number K.

Then l1 and l2 are the diagonal lines, and K is the quadrance of the grammola g.

Exercise 15.14 (Harder) Show that the diagonal lines and quadrance of a
grammola are unique. ¦

Example 15.7 Suppose that l1 ≡ ha : −1 : 0i and l2 ≡ ha : 1 : 0i with a2 6= ±1, and

let g be the grammola with diagonal lines l1 and l2 and quadrance K 6= 0. By the
Quadrance to a line theorem (page 67) the equation for g is

(ax − y)2 (ax + y)2

+ = K.
a2 + 1 a2 + 1

This simplifies to
¡ ¢
2a2 x2 + 2y 2 = K a2 + 1 . ¦

Example 15.8 In the decimal number field the grammola g with equation
2x2 + 8y 2 = 5 (the case a ≡ 1/2 and K ≡ 1 of the previous example) is shown in
Figure 15.7 along with the diagonal lines with equations x − 2y = 0 and x + 2y = 0.
Note that the diagonal lines
h √intersect
√ i
the grammola, which is in this case an ellipse, at
the distinguished points ± 25 , ± 45 .

y 2

-2.5 -1.25 0 1.25 2.5

x
-1

-2

Figure 15.7: Grammola 2x2 + 8y 2 = 5 ¦

176 15. CONICS

Example 15.9 In the field F11 the grammola g with equation 2x2 + 8y 2 = 5 is shown
in Figure 15.8 (large squares) along with its diagonal lines l1 ≡ h1 : −2 : 0i and
l2 ≡ h1 : 2 : 0i (small squares). Note that the diagonal lines also pass through the four
points [±2, ±1] lying on the grammola.

5
4
3
2
1
0
-1
-2
-3
-4
-5
-5 -4 -3 -2 -1 0 1 2 3 4 5

Figure 15.8: Grammola 2x2 + 8y 2 = 5 in F11 ¦

Exercise 15.15 Show that a grammola is a centered conic. ¦

Exercise 15.16 Show that in F11 the grammola 2x2 + 8y 2 = 5 is not a quadrola. ¦

Exercise 15.17 Give an example of a quadrola which is not a grammola. ¦

Exercise 15.18 (Harder) Show that a grammola with diagonal lines l1 and l2 has
two axes precisely when the vertex l1 l2 is square, in which case the axes are bisectors
of l1 l2 . ¦

15.6 Intersections with lines

Definition A line l intersects a conic q at a point A precisely when A lies on both

l and q. A line which intersects a conic q at two distinct points is a chord of the
conic.

Theorem 102 (Conic line intersection) If a line l intersects a conic q in three or

more distinct points, then every point lying on l lies on q.

Proof. This is a consequence of the fact that a quadratic equation has at most two
zeroes in a field.
 16

Geometry of circles

This chapter investigates basic properties of circles, particularly the useful Subtended
spread theorem and some of its consequences, and shows also how to parametrize
points lying on circles. This includes both the unit circle and the projective circle. An
exercise shows how to multiply complex numbers in rational polar form.

16.1 Diameters and chords

Definition A line which intersects a circle c and passes through the center C of c is
a diameter of the circle.

Definition For a triangle A1 A2 A3 with circumcenter C and circumquadrance K,

the circle c with center C and quadrance K is the circumcircle of A1 A2 A3 .

Theorem 103 (Right diameter) Suppose that the line A1 A2 of the triangle
A1 A2 A3 is a diameter of the circumcircle c. Then A1 A2 A3 has a right vertex at A3 .

Proof. This is a consequence of the Right midpoint theorem (page 126), since if C is
the circumcenter, Q (A1 , C) = Q (A2 , C) = Q (A3 , C) , implying that A3 is a right
vertex.

Exercise 16.1 Show that a diameter of a non-null circle is a chord of the circle. ¦

Exercise 16.2 Show that the vertex formed by two distinct diameters of a fixed
non-null circle is always a square vertex. ¦

177
178 16. GEOMETRY OF CIRCLES

Theorem 104 (Circle chord) A chord of a non-null circle is a non-null line.

Proof. Suppose that the non-null circle c has center C ≡ [u, v] and quadrance K 6= 0,
and that the distinct points A1 ≡ [x1 , y1 ] and A2 ≡ [x1 + a, y1 + b] are both points of
intersection of c and a chord l of c. Then
2 2
(x1 − u) + (y1 − v) = K
2 2
(x1 + a − u) + (y1 + b − v) = K.

Take the difference between these two equations to get

2a (u − x1 ) + 2b (v − y1 ) = a2 + b2 .

Now by the Null line theorem (page 60), l = A1 A2 is a null line precisely when
Q (A1 , A2 ) = a2 + b2 = 0. In this case b = ia 6= 0 for some number i satisfying i2 = −1,
so that
(u − x1 ) + i (v − y1 ) = 0.
But then
(x1 − u)2 + (y1 − v)2 = 0 = K
which contradicts the fact that c is non-null. Thus a chord l of a non-null circle is
necessarily a non-null line.

Exercise 16.3 Show that a line l intersects a non-null circle c in at most two points.
¦

16.2 Spreads in a circle

The next result follows from the Extended spread law (page 144), and is fundamental
when dealing with circles. The classical version over the decimal numbers must
distinguish between two separate cases, depending on which ‘side’ of the chord the
third point is on, and must also be careful about the relative positions of the triangle
and the center of the circle. The universal version here is more robust.

Theorem 105 (Subtended spread) Suppose that A1 and A2 are two distinct
points lying on a non-null circle c with center C, and define the spread
s ≡ s (CA1 , CA2 ). Suppose that A3 is any other point lying on c, with
s3 ≡ s (A3 A1 , A3 A2 ). Then s3 is independent of the position of A3 , and furthermore

s = S2 (s3 ) ≡ 4s3 (1 − s3 ) .
16.2. SPREADS IN A CIRCLE 179

c A3
s3
A1 C
s

A2

Figure 16.1: Subtended spreads theorem

Proof. Suppose that the quadrance of c is K 6= 0. From the previous exercise, A1 , A2

and A3 are non-collinear. By the Extended spread law

s3 1
= (16.1)
Q3 4K

where Q3 ≡ Q (A1 , A2 ) 6= 0 by the Circle chord theorem combined with the Null line
theorem (page 60). Thus the spread s3 is constant independent of the position of A3
on the circle c.

Combine (16.1) with the Isosceles triangle theorem (page 122) in A1 A2 C to get
µ ¶
Q3 Q3
s = 1−
K 4K
= 4s3 (1 − s3 ) ≡ S2 (s3 ) .

Definition The spread s3 determined by the side A1 A2 on the circle c is the spread
subtended by A1 A2 .

Theorem 106 (Equal products) Suppose that A1 , A2 , A3 and A4 are distinct

points lying on a circle c such that A1 A4 and A2 A3 intersect at a point B. Then

Q (A1 , B) Q (B, A4 ) = Q (A2 , B) Q (B, A3 ) .

Proof. From the Subtended spread theorem,

s (A2 A1 , A2 A3 ) = s (A4 A1 , A4 A3 ) ≡ s1
s (A1 A2 , A1 A4 ) = s (A3 A2 , A3 A4 ) ≡ s2 .
180 16. GEOMETRY OF CIRCLES

By collinearity

s (A2 A1 , A2 A3 ) = s (A2 A1 , A2 B) = s1
s (A4 A1 , A4 A3 ) = s (A4 B, A4 A3 ) = s1
s (A1 A2 , A1 A4 ) = s (A1 A2 , A1 B) = s2
s (A3 A2 , A3 A4 ) = s (A3 B, A3 A4 ) = s2 .

Thus the Spread law in the triangles A1 A2 B and BA4 A3 gives

s1 Q (A1 , B) Q (B, A3 )
= =
s2 Q (A2 , B) Q (B, A4 )

so that
Q (A1 , B) Q (B, A4 ) = Q (A2 , B) Q (B, A3 ) .

Note that in the rational or decimal number field, the theorem refers to both of the
diagrams in Figure 16.2.

A4 B A2 A4

A3
B

A1
A1 A3
A2

Figure 16.2: Equal products theorem

Exercise 16.4 (Van Schouten’s theorem) Suppose that A1 A2 A3 is an equilateral

triangle with circumcircle c, and that B is any point on c. Show that

{Q (A1 , B) , Q (A2 , B) , Q (A3 , B)}

is a quad triple. ¦

Exercise 16.5 (Brahmagupta’s theorem) Suppose that the quadrilateral ABCD

has points lying on a non-null circle c, with perpendicular diagonals AC and BD
intersecting at P . Show that the altitude from P to AD passes through the midpoint
M of BC. ¦

Exercise 16.6 Suppose that A and B are distinct points lying on a diameter of a
circle c with center C, and that Q (A, C) = Q (B, C). Show that for any point X lying
on c, the quantity Q (A, X) + Q (B, X) is constant independent of X. ¦
16.3. PARAMETRIZING CIRCLES 181

16.3 Parametrizing circles

Definition The projective circle cP is the circle with center C ≡ [0, 1/2] and
quadrance K ≡ 1/4. The unit circle cU is the circle with center O ≡ [0, 0] and
quadrance K ≡ 1.

The equation of cP is
µ ¶2
2 1 1
x + y− =
2 4
and the four points [0, 0], [0, 1] and [±1/2, 1/2] lie on cP .

The equation of cU is
x2 + y 2 = 1
and the four points [±1, 0] and [0, ±1] lie on cU . Recall by Section 13.1 that points
lying on cU parametrize central rotations as well as central reflections.

Theorem 107 (Projective circle) For any number λ satisfying λ2 6= −1, the point
∙ ¸
λ 1
,
1 + λ2 1 + λ2
lies on cP . Apart from [0, 0] every point lying on cP is of this form.

Proof. It is easy to check that for any number λ satisfying λ2 6= −1, the point
∙ ¸
λ 1
,
1 + λ2 1 + λ2
lies on cP . If A ≡ [x, y] is any point on cP other than [0, 0], then y 6= 0, so let λ ≡ x/y.
Then x = λy, and because A lies on cP
µ ¶2
2 1 1
(λy) + y − = .
2 4
This simplifies to the quadratic
¡ ¢
y y + yλ2 − 1 = 0
¡ ¢
so that λ2 =
6 −1 and y = 1/ 1 + λ2 . Thus
∙ ¸
λ 1
A= ,
1 + λ2 1 + λ2
has the required form.
182 16. GEOMETRY OF CIRCLES

Exercise 16.7 Show that lines passing through O ≡ [0, 0] and points lying on the
projective circle cP are in one to one correspondence. ¦

Theorem 108 (Unit circle) For any number λ satisfying λ2 6= −1, the point
∙ ¸
1 − λ2 2λ
2 ,
1 + λ 1 + λ2
lies on cU . Apart from [−1, 0] every point lying on cU is of this form.

Proof. The point [x, y] lies on cP precisely when

µ ¶2
2 1 1
x + y− − = x2 + y 2 − y = 0.
2 4

The point [2y − 1, 2x] lies on cU precisely when

¡ ¢
(2y − 1)2 + (2x)2 − 1 = 4 y 2 − y + x2 = 0.

Thus [x, y] lies on cP precisely when [2y − 1, 2x] lies on cU . The parametrization of the
previous theorem then gives the parametrization
∙ ¸ ∙ ¸
2 2λ 1 − λ2 2λ
− 1 , = ,
1 + λ2 1 + λ2 1 + λ2 1 + λ2
of the points on cU , excluding [−1, 0].

Exercise 16.8 (Rational polar form of complex numbers) Identify [x, y] with
the complex number z = x + iy. Then show that every complex number z lying on
the unit circle with equation x2 + y 2 = 1, except for z = −1, can be written as
2
(1 + iλ)
z=
1 + λ2
for a unique decimal number λ, and conversely any such number lies on the unit
circle. Show that iλ is the intersection of the imaginary axis and the ¡line passing
¢
2
through z and −1. If (r, λ) denotes the complex number r (1 + iλ) / 1 + λ2 then
show that µ ¶
λ1 + λ2
(r1 , λ1 ) (r2 , λ2 ) = r1 r2 , .
1 − λ1 λ2
Explain why the formula of Exercise 9.2 (page 116) makes its appearance here. ¦
 17

Quadrilaterals

This chapter discusses some basic facts about cyclic quadrilaterals, Ptolemy’s theorem,
Brahmagupta’s formula, and the Four point relation, essentially due to Euler.

17.1 Cyclic quadrilaterals

Definition A quadrilateral A1 A2 A3 A4 is cyclic precisely when all the points Ai lie

on a circle c.

Theorem 109 (Cyclic quadrilateral) The quadrilateral A1 A2 A3 A4 with

A1 ≡ [x1 , y1 ], A2 ≡ [x2 , y2 ], A3 ≡ [x3 , y3 ] and A4 ≡ [x4 , y4 ] is cyclic precisely when
£¡ 2 ¢ ¤−
x1 + y12 x2 y3 4 = 0.

Proof. (Using a computer) The condition that A1 A2 A3 A4 be cyclic is

Q (C, A1 ) = Q (A4 , C)
where C is the circumcenter of A1 A2 A3 as given in the Circumcenter theorem (page
143). Expand this and rearrange to obtain the stated formula.

Suppose that A1 A2 A3 A4 is a cyclic quadrilateral with all points Ai lying on the circle
c of quadrance K. By the Subtended spread theorem (page 178), the spread subtended
by the side Ai Aj from any point A on c is independent of A, and will be denoted by
rij , so that for example r12 ≡ s (A3 A1 , A3 A2 ) = s (A4 A1 , A4 A2 ).

183
184 17. QUADRILATERALS

The spreads r12 , r23 , r34 and r14 are the subtended spreads of the cyclic
quadrilateral, and the spreads r13 and r24 are the diagonal subtended spreads.

r24
Q 23 A3
A2 A3
A2 r13 r12
r34 r14
r14
Q 13
Q 12 Q 34
Q 24 r23 r23
A1 r34 r12 A4
A4 r24 r13
A1 Q 14

Figure 17.1: Quadrances and subtended spreads in a cyclic quadrilateral

Quadrances, diagonal quadrances, subtended spreads and diagonal subtended spreads

of a quadrilateral are shown in Figure 17.1. The diagonal subtended spreads coincide
with the usual spreads of A1 A2 A3 A4 , and if the circle c has quadrance K then by the
Extended spread law
Qij = Q (Ai , Aj ) = 4Krij
for all i 6= j. It is therefore useful to consider the subtended spreads rij as basic
measurements in a cyclic quadrilateral, since they are independent of scaling.

Theorem 110 (Cyclic quadrilateral spreads) The subtended spreads r12 , r23 , r34
and r14 of a cyclic quadrilateral satisfy the Quadruple spread formula (page 99)

R (r12 , r23 , r34 , r14 ) = 0.

Furthermore
2 2
(r12 − r23 ) − (r14 − r34 )
r13 =
2 (r12 + r23 − r14 − r34 − 2r12 r23 + 2r14 r34 )
(r12 − r14 )2 − (r23 − r34 )2
r24 =
2 (r12 + r14 − r23 − r34 − 2r12 r14 + 2r23 r34 )

provided that the denominators are not zero.

Proof. Both {r12 , r23 , r13 } and {r14 , r34 , r13 } are spread triples, since these are the
spreads of lines meeting at A2 and A4 respectively. Since r13 is common to both, the
Two spread triples theorem (page 98) shows that R (r12 , r23 , r34 , r14 ) = 0, and that r13
can be expressed as stated. The argument with r24 is similar, using the two spread
triples {r12 , r14 , r24 } and {r23 , r34 , r24 }.
17.2. CIRCUMQUADRANCE FORMULA 185

A cyclic quadrilateral with subtended spreads r12 , r23 , r34 and r14 is solvable
precisely when the denominators in the above theorem are both nonzero.

17.2 Circumquadrance formula

Define the circumquadrance K of a cyclic quadrilateral A1 A2 A3 A4 to be the

quadrance of the circle on which the points lie. This is also the circumquadrance of
any three of the points of the quadrilateral. The next theorem gives a generally
quadratic formula for K in terms of the quadrances.

Theorem 111 (Quadrilateral circumquadrance) If a cyclic quadrilateral

A1 A2 A3 A4 lying on a circle c has quadrances Q12 , Q23 , Q34 and Q14 , then the
quadrance K of the circle c satisfies the equation

pK 2 − 2qK + r = 0

where
p = Q (Q12 , Q23 , Q34 , Q14 )

q = (Q12 Q14 − Q23 Q34 ) (Q23 Q14 − Q12 Q34 ) (Q12 + Q23 − Q34 − Q14 )
+ (Q12 Q14 − Q23 Q34 ) (Q34 Q14 − Q12 Q23 ) (Q12 − Q23 + Q34 − Q14 )
+ (Q12 Q23 − Q34 Q14 ) (Q23 Q14 − Q12 Q34 ) (Q12 − Q23 − Q34 + Q14 )

and
r = (Q12 Q14 − Q23 Q34 ) (Q23 Q14 − Q12 Q34 ) (Q12 Q23 − Q34 Q14 ) .

Proof. (Using a computer) Note that p is the Quadruple quad function (page 70)
applied to Q12 , Q23 , Q34 and Q14 . By the Cyclic quadrilateral spreads theorem, the
four subtended spreads r12 , r23 , r34 and r14 of A1 A2 A3 A4 satisfy the Quadruple spread
formula
R (r12 , r23 , r34 , r14 ) = 0
where R is the Quadruple spread function.

Now when the relations

rij = Qij /4K
are substituted, this becomes an equation involving K and the quadrances
Q12 , Q23 , Q34 and Q14 . After cancellation of a factor of 256K 6 , this has the form
pK 2 − 2qK + r = 0
with coefficients p, q and r that are polynomial expressions in Q12 , Q23 , Q34 and Q14 .
These coefficients can be rearranged to take the stated forms.
186 17. QUADRILATERALS

Exercise 17.1 (Using a computer) Show that in the notation of the above theorem
2 2
(2q) − 4pr = 4Q12 Q23 Q34 Q14 (Q12 + Q23 − Q34 − Q14 )
× (Q12 − Q23 + Q34 − Q14 )2 (Q12 − Q23 − Q34 + Q14 )2 . ¦

Example 17.1 In the field F11 consider the existence of a cyclic quadrilateral with
quadrances
Q12 ≡ 5 Q23 ≡ 3 Q34 ≡ 6 Q14 ≡ 6.
The constants p, q and r can be evaluated to be 3, 0 and 10 respectively, so the
equation for the circumquadrance K of the quadrilateral is
3K 2 + 10 = 0
with solutions K = 2 and 9. On the central circle with quadrance K ≡ 2, the
quadrilateral A1 A2 A3 A4 has the above quadrances, where
A1 ≡ [2, 3] A2 ≡ [9, 3] A3 ≡ [1, 10] A4 ≡ [8, 9] .
On the central circle with quadrance K ≡ 9, the quadrilateral B1 B2 B3 B4 also has the
above quadrances, where
B1 ≡ [4, 2] B2 ≡ [3, 0] B3 ≡ [8, 0] B4 ≡ [9, 7] .
Both examples are shown in Figure 17.2.

10 A3 10
9 A4 9
8 8
7 7 B4
6 6
5 5
4 4
3 A1 A2 3
2 2 B1
1 1
0 0 B2 B3
0 1 2 3 4 5 6 7 8 9 10 0 1 2 3 4 5 6 7 8 9 10

Figure 17.2: Cyclic quadrilaterals on circles of quadrance 2 and 9

For the quadrilateral A1 A2 A3 A4 the lines are l12 = h0 : 7 : 1i, l23 = h7 : 8 : 1i,
l34 = h2 : 3 : 1i and l14 = h1 : 10 : 1i while the diagonal lines are l13 = h8 : 9 : 1i and
l24 = h8 : 5 : 1i. The spreads of the quadrilateral A1 A2 A3 A4 are s1 = 6, s2 = 9, s3 = 6
and s4 = 9. The subtended spreads are r12 = 2, r23 = 10, r34 = 9 and r14 = 9, with
diagonal subtended spreads r13 = 9 and r24 = 6. You may verify that r12 , r23 , r34 and
r14 satisfy the Quadruple spread formula (page 99), and that r13 and r24 may be
determined from them as in that theorem. The quadrea of A1 A2 A3 A4 is A = 4.

The quadrilateral B1 B2 B3 B4 may be similarly analyzed. Its quadrea is A = 9. ¦

17.3. CYCLIC QUADRILATERAL QUADREA 187

17.3 Cyclic quadrilateral quadrea

Theorem 112 (Cyclic quadrilateral quadrea) Suppose a solvable cyclic

quadrilateral A1 A2 A3 A4 has quadrances Q12 , Q23 , Q34 and Q14 with

Q (Q12 , Q23 , Q34 , Q14 ) (Q12 Q14 − Q23 Q34 ) (Q23 Q14 − Q12 Q34 ) (Q12 Q23 − Q34 Q14 ) 6= 0.

Then its quadrea A satisfies the quadratic equation

A2 − 2mA + p = 0

where
¡ ¢
m = (Q12 + Q23 + Q34 + Q14 )2 − 2 Q212 + Q223 + Q234 + Q214
p = Q (Q12 , Q23 , Q34 , Q14 ) .

Proof. (Using a computer) The following proof involves some seemingly miraculous
computations requiring a computer. It would be preferable to have a simpler and more
direct argument.

Suppose the circumquadrance of the cyclic quadrilateral A1 A2 A3 A4 is K. From the

Quadrilateral circumquadrance theorem (page 185) you know that K satisfies the
equation
pK 2 − 2qK + r = 0 (17.1)
where p, q and r are known functions of the quadrances Q12 , Q23 , Q34 and Q14 , with
p = Q (Q12 , Q23 , Q34 , Q14 ) 6= 0 and r 6= 0 by assumption. Thus K 6= 0. Apply the
Cyclic quadrilateral spreads theorem and the relation Qij = 4Krij to obtain

(Q12 − Q23 )2 − (Q14 − Q34 )2

Q13 = 2K
(4K (Q12 + Q23 − Q14 − Q34 ) − 2Q12 Q23 + 2Q14 Q34 )
2 2
(Q12 − Q14 ) − (Q23 − Q34 )
Q24 = 2K .
(4K (Q12 + Q14 − Q23 − Q34 ) − 2Q12 Q14 + 2Q23 Q34 )

Use Bretschneider’s formula (page 118) to write the quadrea A of the quadrilateral
A1 A2 A3 A4 as
2
A = 4Q13 Q24 − (Q12 + Q34 − Q23 − Q14 ) .
Replace Q13 and Q24 and simplify, to get an expression A = f (K) /g (K) where f and
g are polynomials in K generally of degrees one and two respectively, with coefficients
that depend on Q12 , Q23 , Q34 and Q14 . Use the expression (17.1) to replace K 2 in
g (K) with the linear expression (2qK − r) /p in K. After simplification this gives

aK + b
A=
cK + d
188 17. QUADRILATERALS

with a, b, c and d polynomials in Q12 , Q23 , Q34 and Q14 . Equivalently

Ad − b
K= .
a − Ac
Substitute this into (17.1) to get a quadratic equation of the form
¡ ¢
w1 w2 A2 − 2mA + p = 0
in A, where m and p are as in the statement of the theorem and where
w1 = Q (Q12 , Q23 , Q34 , Q14 ) (Q14 Q34 − Q12 Q23 )2 (Q23 Q34 − Q12 Q14 )2
w2 = (Q23 − Q14 − Q12 + Q34 )2 (Q14 − Q12 − Q23 + Q34 )2 (Q12 − Q14 − Q23 + Q34 )4 .
Now w2 6= 0, for otherwise one of Q13 and Q24 would be zero, contradicting the Circle
chord theorem (page 178). So if w1 6= 0 then the theorem follows.

Given quadrances Q12 , Q23 , Q34 and Q14 satisfying w1 6= 0, there are in general then
two possible quadreas A1 and A2 for a cyclic quadrilateral with these as its
quadrances. By properties of the zeroes of a quadratic equation,

2 ¡ ¢
(A1 + A2 ) /2 = (Q12 + Q23 + Q34 + Q14 ) − 2 Q212 + Q223 + Q234 + Q214
A1 A2 = Q (Q12 , Q23 , Q34 , Q14 ) .

In the first equation you see again the expression of Descartes discussed on page 71.
The second equation is a rational form of Brahmagupta’s formula. It shows that the
quantity Q (Q12 , Q23 , Q34 , Q14 ) appearing in Brahmagupta’s identity (page 72) is the
product of the two quadreas of the cyclic quadrilaterals with the given quadrances.

Example 17.2 In Example 17.1 over F11 the cyclic quadrilateral A1 A2 A3 A4 had
quadrances Q12 ≡ 5, Q23 ≡ 3, Q34 ≡ 6 and Q14 ≡ 6. So then m = 1 and p = 3 and the
quadrea A satisfies A2 − 2A + 3 = 0, with zeroes 4 and 9 as obtained there. ¦

−−−−−−−→
Theorem 113 (Cyclic signed area) The oriented cyclic quadrilateral A1 A2 A3 A4
with points £ ¡ ¢ ¡ ¢ ¡ ¢¤
Ai = 2λi / 1 + λ2i , 1 − λ2i / 1 + λ2i
³−−−−−−−→´
i = 1, 2, 3 and 4, lying on the unit circle, has signed area a A1 A2 A3 A4 equal to

2 (λ1 − λ3 ) (λ2 − λ4 ) (λ1 + λ3 − λ2 − λ4 + λ1 λ2 λ3 + λ1 λ3 λ4 − λ2 λ3 λ4 − λ1 λ2 λ4 )

¡ ¢¡ ¢¡ ¢¡ ¢ .
1 + λ21 1 + λ22 1 + λ23 1 + λ24

Proof. (Using a computer) A direct calculation.

17.3. CYCLIC QUADRILATERAL QUADREA 189

Exercise 17.2 Show that if λ3 = λ4 , for example, then the signed area of the previous
−−−−−→
theorem reduces to that of the oriented triangle A1 A2 A3 , namely
³−−−−−→´ 2 (λ − λ ) (λ − λ ) (λ − λ )
1 2 2 3 1 3
a A1 A2 A3 = ¡ 2
¢¡ 2
¢¡ 2
¢ . ¦
1 + λ1 1 + λ2 1 + λ3

Example 17.3 In the decimal number field, suppose that a quadrilateral A1 A2 A3 A4

has side lengths
|A1 , A2 | ≡ 5 |A2 , A3 | ≡ 8 |A3 , A4 | ≡ 2 |A1 , A4 | ≡ 10
so that the associated quadrances are Q12 = 25, Q23 = 64, Q34 = 4 and Q14 = 10.
Then using the notation of the Quadrilateral circumquadrance theorem and the Cyclic
quadrilateral quadrea theorem, the polynomial expressions p, q, r and m take on the
values
p = 19 490 625 q = 582 390 000
r = 16 964 640 000 m = 7775.
Thus the circumquadrance K of A1 A2 A3 A4 satisfies the quadratic equation
0 = 19 490 625K 2 − 1164 780 000K + 16 964 640 000
= 50 625 (7K − 176) (55K − 1904)
so that
K = 176/7 or K = 1904/55.
The corresponding radii of the circumcircles are the square roots of these quantities,
approximately
5. 014 265 or 5. 883 721.
The quadrea A of A1 A2 A3 A4 satisfies the quadratic equation
A2 − 15 550A + 19 490 625 = (A − 14 175) (A − 1375) = 0
so that
A = 14 175 or A = 1375.
The corresponding signed areas are one quarter the square roots of these quantities,
approximately
±29. 764 702 or ±9. 270 248.
These two situations are realized in Figure 17.3, drawn to scale in appropriate units.

A2
A2
A1 A4
A1
C A3
C
A3
A4

Figure 17.3: Two cyclic quadrilaterals ¦

190 17. QUADRILATERALS

17.4 Ptolemy’s theorem

Ptolemy was one of the great astronomers of antiquity, and his classic text The
Almagest was for many centuries to astronomy what Euclid’s The Elements was to
geometry.

Theorem 114 (Ptolemy’s theorem) Suppose that A1 A2 A3 A4 is a solvable cyclic

quadrilateral with subtended spreads rij for i, j = 1, 2, 3 and 4. Then

{r12 r34 , r14 r23 , r13 r24 }

is a quad triple.

Proof. (Using a computer) Since A1 A2 A3 A4 is solvable, the Cyclic quadrilateral

spreads theorem (page 184) gives expressions for r13 and r24 in terms of r12 , r23 , r34
and r14 . Use Archimedes’ function A and some pleasant factoring to get

1 R (r12 , r23 , r34 , r14 )

A (r12 r34 , r14 r23 , r13 r24 ) =
16 (r12 + r23 − r14 − r34 − 2r12 r23 + 2r14 r34 )2
B (r12 , r23 , r34 , r14 )
×
(r12 + r14 − r23 − r34 − 2r12 r14 + 2r23 r34 )2

where R is the Quadruple spread function (page 99) and where

B (a, b, c, d) ≡ 4 (a + c − b − d)
³ ´
× (a − c)2 (b + d) (b + d − 1) − (b − d)2 (a + c) (a + c − 1)
−R (a, b, c, d) .

The same theorem asserts that the subtended spreads r12 , r23 , r34 and r14 satisfy the
Quadruple spread formula, so that

R (r12 , r23 , r34 , r14 ) = 0.

Thus
A (r12 r34 , r14 r23 , r13 r24 ) = 0

and {r12 r34 , r14 r23 , r13 r24 } is a quad triple.

Example 17.4 In F11 the quadrilateral A1 A2 A3 A4 of Example 17.1 has subtended

spreads r12 = 2, r23 = 10, r34 = 9 and r14 = 9, and diagonal subtended spreads r13 = 9
and r24 = 6. Thus Ptolemy’s theorem states that {2 × 9, 9 × 10, 9 × 6} = {7, 2, 10} is a
quad triple, which is true since (7 + 2 − 10)2 = 1 = 4 × 7 × 2. ¦
17.5. FOUR POINT RELATION 191

17.5 Four point relation

The Triple quad formula concerned three points on a line, and the associated three
quadrances. Here is a generalization to four points in the plane, forming six
quadrances. This formula goes back essentially to Euler, who was interested in the
volume of a tetrahedron as a function of its six side lengths. The treatment here
borrows from [Dorrie].

Definition The Euler function E is

¯ ¯
¯ 2P1 P1 + P2 − Q3 P1 + P3 − Q2 ¯¯
¯
E (Q1 , Q2 , Q3 , P1 , P2 , P3 ) ≡ ¯¯P1 + P2 − Q3 2P2 P2 + P3 − Q1 ¯¯ .
¯P1 + P3 − Q2 P2 + P3 − Q1 2P3 ¯

Theorem 115 (Four point relation) Suppose that the triangle A1 A2 A3 has
quadrances Q1 , Q2 and Q3 , and that B is any point with quadrances P1 ≡ Q (A1 , B),
P2 ≡ Q (A2 , B) and P3 ≡ Q (A3 , B). Then

E (Q1 , Q2 , Q3 , P1 , P2 , P3 ) = 0.

A2

Q3 P2
Q1
B
P1 P3 A3
A1 Q2

Figure 17.4: Four point relation

Proof. Suppose that B ≡ [x0 , y0 ] and that

A1 ≡ [x0 + x1 , y0 + y1 ]
A2 ≡ [x0 + x2 , y0 + y2 ]
A3 ≡ [x0 + x3 , y0 + y3 ] .

So
2 2
P1 = x21 + y12 and Q1 = (x3 − x2 ) + (y3 − y2 )
and similarly for the other indices.
192 17. QUADRILATERALS

Thus

P1 + P2 − Q3 = x21 + y12 + x22 + y22 − (x1 − x2 )2 − (y1 − y2 )2

= 2 (x1 x2 + y1 y2 ) .

After removing a factor of two, the required determinantal equation is then

¯ ¯
¯ x21 + y12 x1 x2 + y1 y2 x1 x3 + y1 y3 ¯¯
¯
¯x1 x2 + y1 y2 x22 + y22 x2 x3 + y2 y3 ¯¯ = 0.
¯
¯x1 x3 + y1 y3 x2 x3 + y2 y3 x23 + y32 ¯

This can be verified directly. Alternatively the matrix in the determinant above is the
product of the matrices ⎛ ⎞⎛ ⎞
0 x1 y1 0 0 0
⎝0 x2 y2 ⎠ ⎝x1 x2 x3 ⎠
0 x3 y3 y1 y2 y3
and so the determinant is 0.

Exercise 17.3 Show that

E (Q1 , Q2 , Q3 , P1 , P2 , P3 )
µ ¶
4P1 P2 P3 + (P2 + P1 − Q3 ) (P2 + P3 − Q1 ) (P1 + P3 − Q2 )
=2 . ¦
−P1 (P2 + P3 − Q1 )2 − P2 (P1 + P3 − Q2 )2 − P3 (P2 + P1 − Q3 )2

Exercise 17.4 Show that

¯ ¯
¯0 P1 P2 P3 1¯¯
¯
¯P1 0 Q3 Q2 1¯¯
¯
E (Q1 , Q2 , Q3 , P1 , P2 , P3 ) = ¯¯P2 Q3 0 Q1 1¯¯ . ¦
¯P3 Q2 Q1 0 1¯¯
¯
¯1 1 1 1 0¯

Exercise 17.5 Show that as a quadratic equation in P3 , the Four point relation can
be written in terms of Archimedes’ function A as
µ ¶2
(Q1 − Q2 ) (P2 − P1 )
P3 − P1 − P2 + Q3 − Q1 − Q2 +
Q3
A (Q1 , Q2 , Q3 ) A (P1 , P2 , Q3 )
= . ¦
4Q23

Exercise 17.6 Use the Four point relation to re-derive the formula K = Q1 Q2 Q3 /A
(page 145) for the circumquadrance of a triangle with quadrea A and quadrances
Q1 , Q2 and Q3 . ¦
 18

Euler line and nine point circle

This chapter establishes universal analogues of the classical Euler line and nine point
circle. The proofs illustrate two quite different approaches–the traditional synthetic
one using a sequence of deductions, and the more modern computational one using
algebraic manipulation of coordinates.

18.1 Euler line

Theorem 116 (Euler line) If the characteristic is not three, then the centroid G,
circumcenter C and orthocenter O of a triangle A1 A2 A3 are collinear.

Proof. If A1 ≡ [x1 , y1 ], A2 ≡ [x2 , y2 ] and A3 ≡ [x3 , y3 ] then from Exercise 3.16, the
Circumcenter theorem (page 143) and the Orthocenter theorem (page 147), the
centroid G, circumcenter C and orthocenter O of A1 A2 A3 are given by the expressions
∙ ¸
x1 + x2 + x3 y1 + y2 + y3
G = ,
3 3
"£ ¤− £ 2 ¤− £ ¤− £ ¤− #
x1 y2 3 + y1 y2 3 x1 x22 3 − x21 x2 3
2
C = − , −
2 [x1 y2 ]3 2 [x1 y2 ]3
" £ 2 ¤− £ 2 ¤− #
[x1 x2 y2 ]− −
3 − y1 y2 3 [x1 y1 y2 ]3 + x1 x2 3
O = − , − .
[x1 y2 ]3 [x1 y2 ]3

Now use Exercise 2.7 (page 29) to deduce that (2/3) C + (1/3) O = G. The Affine
combination theorem (page 46) shows that either all three points coincide, or G lies on
OC.

193
194 18. EULER LINE AND NINE POINT CIRCLE

Exercise 18.1 Show that the three points G, C and O coincide precisely when
A1 A2 A3 is equilateral. ¦

If A1 A2 A3 is not equilateral, then the line e ≡ CO is the Euler line of the triangle.

18.2 Nine point circle

If M1 , M2 and M3 are the midpoints of the sides of the triangle A1 A2 A3 , then

M1 M2 M3 is also a triangle, and the circumcircle n and circumcenter N of M1 M2 M3
are respectively the nine point circle and the nine point center of the triangle
A1 A2 A3 .

Figure 18.1 shows the reason for the terminology, since in the decimal number field the
nine point circle also passes through the feet F1 , F2 and F3 of the altitudes of A1 A2 A3 ,
as well as the midpoints P1 , P2 and P3 of the respective sides A1 O, A2 O and A3 O. Also
shown in this Figure is the Euler line e.

A1 M3 F3
A2
P1 C
P2
G
N
M2
O M1

F1
F2 P3

A3

Figure 18.1: Nine point circle and Euler line

Exercise 18.2 (Using a computer) Suppose the triangle A1 A2 A3 has circumcenter

C, orthocenter O and midpoints of the sides M1 , M2 and M3 . If A1 ≡ [x1 , y1 ],
A2 ≡ [x2 , y2 ] and A3 ≡ [x3 , y3 ], then show that the circumcenter N of M1 M2 M3 is the
point
"£ ¤− £ 2 ¤− £ 2 ¤− £ 2 ¤− #
x21 y2 3 + 2 [x1 x2 y2 ]−
3 − y1 y2 3 x1 x2 3 + 2 [x1 y1 y2 ]−
3 − x1 y2 3
N= − , − .
4 [x1 y2 ]3 4 [x1 y2 ]3

Deduce that N is the midpoint of OC, and hence lies on the Euler line e. ¦
18.2. NINE POINT CIRCLE 195

Theorem 117 (Nine point circle) The nine point circle n of a non-null triangle
A1 A2 A3 passes through the midpoints M1 , M2 and M3 of the sides, the feet F1 , F2
and F3 of the altitudes, and the midpoints P1 , P2 and P3 of the respective sides
A1 O, A2 O and A3 O, where O is the orthocenter.

A3
P3 F1
F2 M1
O
M2
N
P2
P1 A2
F3
A1 M3

Figure 18.2: Nine point circle theorem

Proof. By Thales’ theorem (page 48) applied to both A1 A2 A3 and A1 A2 O, the sides
M1 M2 and P1 P2 are both parallel to A1 A2 , and both have quadrance equal to Q3 /4.
The same argument, applied to both A1 A3 O and A2 A3 O, shows that the sides M2 P1
and M1 P2 are both parallel to A3 O, and have quadrance equal to Q (A3 , O) /4.

Hence P1 P2 M1 M2 is a rectangle. Similarly P1 P3 M1 M3 and P2 P3 M2 M3 are rectangles.

These three rectangles have the property that any two share a common diagonal side.
Thus the centers of these rectangles coincide at a point N. By Exercise 5.4, the
Parallelogram center theorem (page 49) and the Midpoint theorem (page 60), the
quadrance from N to any of the points Mi or Pj are equal, for i, j = 1, 2 and 3. Thus
these six points lie on a circle n centered at N , which is the circumcircle of M1 M2 M3 .

Since M3 P3 is a diagonal of the circle n and s (F3 M3 , F3 P3 ) = 1, the Right midpoint

theorem (page 126) shows that F3 also lies on n. Thus by symmetry the three feet F1 ,
F2 and F3 all lie on n.

Example 18.1 In the field F17 the triangle A1 A2 A3 with points

A1 ≡ [2, 3] A2 ≡ [10, 5] A3 ≡ [6, 12]

has lines l1 = h2 : 6 : 1i, l2 = h7 : 12 : 1i and l3 = h12 : 3 : 1i, with l3 a null line. The
circumcenter is C = [5, 8], the orthocenter is O = [8, 4], and the centroid is G = [6, 1].
The midpoints of the sides are M1 = [8, 0], M2 = [4, 16] and M3 = [6, 4]. The nine
point circle n has center N = [15, 6] and quadrance zero, so is a null circle.
196 18. EULER LINE AND NINE POINT CIRCLE

The Euler line is h3 : 15 : 1i. The feet of the altitudes are F1 = [0, 14] and F2 = [13, 15],
but there is no foot F3 , as the altitude to the null line l3 does not intersect it. The
midpoints of the sides A1 O, A2 O and A3 O are P1 = [5, 12], P2 = [9, 13] and P3 = [7, 8].

This is all shown in Figure 18.3, where the midpoints M1 , M2 and M3 , the feet F1 and
F2 , and the midpoints P1 , P2 and P3 all lie on the nine point circle n (circles), and O,
N , G and C all lie on the Euler line (black boxes). Note that N also lies on the
nine-point circle, so the circle still deserves its name!

16 M2
15 F2
14 F1
13 P2
12 P1 A3
11
10
9
8 C P3
7
6 N
5 A2
A2
4 M3 O
3 A1
2
1 G
0 M1
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16

Figure 18.3: Nine point circle in F17 ¦

Exercise 18.3 (Eight point circle) Show that if a quadrilateral A1 A2 A3 A4 has

perpendicular diagonals then the midpoints M12 , M23 , M34 and M14 of the sides and
the feet F12 , F23 , F34 and F14 of the altitudes from the respective midpoints to the
opposite sides all lie on a circle. See Figure 18.4.

A2

F12 M23
F23
M12
A3
C

M 34
F34
A1
M 41 F41 A4

Figure 18.4: Eight point circle ¦

 19

Tangent lines and tangent

conics

The notion of a tangent line to a conic (or indeed to any algebraic curve) does not
require derivations, derivatives or indeed any calculus, and applies to general fields.
Generalizing leads to the idea of a tangent conic. In this way classical properties of
conics can be extended to universal geometry, and then applied to metrical algebraic
geometry. For example, since conics come in different types, you may classify points on
curves according to the geometric properties of the tangent conic. Tangent conics to
the folium of Descartes and the lemniscate of Bernoulli are studied.

In this chapter the translational structure of the plane becomes important, particularly
the isometries τ B defined in Chapter 13. Some calculations require a computer.

19.1 Translates and Taylor conics

Recall from Chapter 15 that the conic q ≡ hd : e : f : a : b : ci is standard precisely

when it passes through the origin O ≡ [0, 0], or equivalently when c = 0. The standard
conic q ≡ hd : e : f : a : b : 0i is singular precisely when a = b = 0.

Definition If q ≡ hd : e : f : a : b : 0i is a non-singular standard conic, then the line

T (1) q ≡ ha : b : 0i is the tangent line to q at O.

The idea is now to generalize this notion to points lying on arbitrary conics. Recall
from Chapter 13 that the translate of X ≡ [x, y] by B ≡ [u, v] is the point

τ B ([x, y]) ≡ [x + u, y + v] .

197
198 19. TANGENT LINES AND TANGENT CONICS

Theorem 118 (Translate of a conic) Suppose that B ≡ [u, v]. Then the point
X ≡ [x, y] lies on the conic q ≡ hd : e : f : a : b : ci precisely when the translate
τ B ([x, y]) lies on the conic q1 ≡ hd1 : e1 : f1 : a1 : b1 : c1 i, where

d1 ≡ d e1 ≡ e f1 ≡ f

and
a1 ≡ a − 2du − ev b1 ≡ b − 2f v − eu
2 2
c1 ≡ du + euv + f v − au − bv + c.

Proof. The coefficients of q1 are chosen so that the following polynomial identity holds
2 2
d (x − u) + e (x − u) (y − v) + f (y − v) + a (x − u) + b (y − v) + c
= d1 x2 + e1 xy + f1 y 2 + a1 x + b1 y + c1 .

Then τ B ([x, y]) ≡ [x + u, y + v] lies on q1 precisely when [x, y] lies on q.

Definition The conic q1 ≡ τ B (q) defined in this theorem is the translate of q by

B.

Example 19.1 The circle c ≡ h1 : 0 : 1 : 0 : 0 : −1i has equation

x2 + y 2 − 1 = 0.

If B ≡ [2, 3], then the point [x, y] lies on c precisely when τ B ([x, y]) = [x + 2, y + 3] lies
on the circle τ B (c) with equation
2 2
(x − 2) + (y − 3) − 1 = 0.

Expand to see that τ B (c) = h1 : 0 : 1 : −4 : 6 : 12i. ¦

Example 19.2 The conic q ≡ h1 : −2 : −1 : 1 : 0 : −5i has equation

x2 − 2xy − y 2 + x − 5 = 0 (19.1)

and A ≡ [3, 1] lies on q. Then τ −1

A (q) is a standard conic, and its equation is obtained
by replacing in (19.1) each x by x + 3, and each y by y + 1. This simplifies to

x2 − 2xy − y 2 + 5x − 8y = 0

so that τ −1
A (q) = h1 : −2 : −1 : 5 : −8 : 0i. ¦

Definition If the point A lies on the conic q then the standard conic τ −1
A (q) is the
Taylor conic of q at A. It is also denoted by qA .
19.2. TANGENT LINES 199

19.2 Tangent lines

Suppose that q is a conic passing through the point A ≡ [x0 , y0 ]. Suppose that
* d : e0 : f0 +
0
qA ≡ a0 : b0
0

is the Taylor conic of q at A. The point A is a non-singular point of q precisely

when at least one of a0 and b0 are non-zero. In this case the line

l ≡ T (1) qA ≡ ha0 : b0 : 0i

is the tangent line to qA at O. The translate τ A (l) is then the tangent line to q at
(1)
A, and is denoted TA q. Thus

³ ´ ³ ´
(1)
TA q = τ A T (1) qA = τ A T (1) τ −1
A (q) .

So to find the tangent line to a conic q at some non-singular point A lying on it, first
translate q by applying τ −1
A , then take the linear part, and then translate this line back
by applying τ A . This ensures that the tangent line passes through A.

Example 19.3 From Example 19.2, the conic q ≡ h1 : −2 : −1 : 1 : 0 : −5i passes

through A ≡ [3, 1] and
qA = h1 : −2 : −1 : 5 : −8 : 0i .
Thus A is a non-singular point of q (since one of 5 and −8 is non-zero). The tangent
line to qA at O is T (1) qA = h5 : −8 : 0i. The tangent line to q at A thus has equation
(1)
5 (x − 3) − 8 (y − 1) = 0, so it is TA q = h5 : −8 : −7i. The conic q and the tangent
line to q at A are illustrated in the decimal number field in Figure 19.1.

10
y 8
6
4
2

-10 -8 -6 -4 -2 2 4 6 8 10
-2 x
-4
-6
-8
-10

Figure 19.1: Tangent line to a conic ¦

200 19. TANGENT LINES AND TANGENT CONICS

Theorem 119 (Tangent to a conic) Suppose the point A ≡ [x0 , y0 ] is a

non-singular point lying on the conic q ≡ hd : e : f : a : b : ci. Then the tangent line to
q at A is
(1)
TA q = ha + 2dx0 + ey0 : b + ex0 + 2f y0 : ax0 + by0 + 2ci .

Proof. Since q has equation 0 = dx2 + exy + f y 2 + ax + by + c, the equation for qA is

2 2
0 = d (x + x0 ) + e (x + x0 ) (y + y0 ) + f (y + y0 )
+a (x + x0 ) + b (y + y0 ) + c
= dx2 + exy + f y 2 + (a + 2dx0 + ey0 ) x + (b + 2f y0 + ex0 ) y.

By assumption at least one of the coefficients of x or y is non-zero, so the tangent line

T (1) qA has equation

0 = (a + 2dx0 + ey0 ) x + (b + 2f y0 + ex0 ) y.

(1)
The equation of the tangent line TA q is then

0 = (a + 2dx0 + ey0 ) (x − x0 ) + (b + ex0 + 2f y0 ) (y − y0 )

= (a + ey0 + 2dx0 ) x + (b + ex0 + 2f y0 ) y + ax0 + by0 + 2c

where the constant term is simplified by the fact that [x0 , y0 ] lies on q.

Example 19.4 The tangent line £ to the ¤ parabola p ≡ h0 : 0 : 1 : −4 : 0 : 0i with

equation y 2 = 4x at the point y02 , 2y0 is
­ ® ­ ®
−4 : 4y0 : −4y02 = 1 : −y0 : y02 .

All points lying on p are non-singular. ¦

Theorem 120 (Tangent to a circle) Every point A lying on a non-null circle c is

non-singular. If C is the center of c then the tangent line to c at A is the altitude
from A to AC, and is a non-null line.

Proof. Suppose that A ≡ [x0 , y0 ] lies on the circle c. If C ≡ [x1 , y1 ] is the center of c,
then the equation for c is

(x − x1 )2 + (y − y1 )2 = (x0 − x1 )2 + (y0 − y1 )2

which yields
­ ®
c = 1 : 0 : 1 : −2x1 : −2y1 : 2y0 y1 + 2x0 x1 − x20 − y02 .
19.2. TANGENT LINES 201

By the previous Tangent to a conic theorem

(1) ­ ¡ ¢®
TA c = −2x1 + 2x0 : −2y1 + 2y0 : −2x1 x0 + −2y1 y0 + 2 2y0 y1 + 2x0 x1 − x20 − y02
­ ®
= x1 − x0 : y1 − y0 : x20 + y02 − x1 x0 − y1 y0 .
2 2
This is a non-null line since (x0 − x1 ) + (y0 − y1 ) is the quadrance of c, which is
non-zero. The Line through two points theorem (page 38) gives

AC = hy1 − y0 : x0 − x1 : x1 y0 − x0 y1 i
(1) (1)
which is indeed perpendicular to TA c. Thus TA c is the altitude from A to AC.

Theorem 121 (Parabola reflection) Suppose k is the tangent line to a parabola p

at a point A lying on it. Then

s (k, m) = s (k, n)

where m ≡ AF with F the focus of p and n is the altitude from A to the directrix l.

Proof. (Using a computer) If F ≡ [x1 , y1 ] and l ≡ ha : b : ci are the focus and

directrix of a parabola p, then the equation for p is w (x, y) = 0 where
¡ ¡ ¢ ¢
w (x, y) = b2 x2 − 2abxy + a2 y 2 − 2ac + 2 a2 + b2 x1 x
¡ ¡ ¢ ¢ ¡ ¢¡ ¢
− 2bc + 2 a2 + b2 y1 y + a2 + b2 x21 + y12 − c2 .

Suppose that A ≡ [x0 , y0 ] is a point lying on the parabola. Then by the Tangent to a
conic theorem, the tangent line to p at A is
¿ ¡ ¢ ¡ ¢ À
b¡2 x0 −¡aby0 − ¢a2 +¢ b2 x¡1 − ac¡ : −abx0¢ + a
¢
2
y0 − a¡2 + b2 y¢1¡− bc ¢
k≡ .
−x0 ac + a2 + b2 x1 − y0 bc + a2 + b2 y1 − c2 + a2 + b2 x21 + y12

The Line through two points theorem (page 38) gives

m ≡ AF = hy1 − y0 : x0 − x1 : x1 y0 − x0 y1 i .

The Altitude to a line theorem (page 41) shows the altitude from A to l to be

n ≡ h−b : a : bx0 − ay0 i .

Having the equations of the three lines k, m and n, use the definition of the spread to
obtain an expression for s (k, m) − s (k, n) in terms of x0 and y0 . Using a computer,
this difference in spreads can be shown to be a rational function of x0 and y0 , with
w (x0 , y0 ) as a factor. Since A lies on p, this difference is then zero.

Exercise 19.1 Check this argument more directly for the parabola p with equation
y 2 = 4ax. ¦
202 19. TANGENT LINES AND TANGENT CONICS

Exercise 19.2 Show that for the parabola p with equation y 2 = 4x the tangents to p
at any two distinct points A1 and A2 lying on p intersect, say at a point B. Show
further that if F is the focus of p then
2 2
(Q (A1 , B)) : (Q (A2 , B)) = Q (A1 , F ) : Q (A2 , F ) .

The situation is illustrated over F11 in Figure 19.2, where F ≡ [1, 0], A1 ≡ [1, 2],
A2 ≡ [4, −4] and B ≡ [−2, −1]. Then both sides of the above equation are 5 : 1.

5
4
3
2
1
0
-1
-2
-3
-4
-5
-5 -4 -3 -2 -1 0 1 2 3 4 5

Figure 19.2: Two tangents to y 2 = 4x in Z11 ¦

Exercise 19.3 (Quadrola reflection—Harder) Show that the tangent line to a

quadrola q at a point A lying on it makes equal spreads with the lines AF1 and AF2 ,
where F1 and F2 are a pair of foci of q. ¦

19.3 Higher order curves and tangents

The notation and definitions for conics and their tangents generalize to higher order
curves.

A cubic curve q is a proportion of the form hg : h : i : j : d : e : f : a : b : ci

representing the equation z (x, y) = 0 where

z (x, y) ≡ gx3 + hx2 y + ixy 2 + jy 3 + dx2 + exy + f y 2 + ax + by + c

and where i is just another coefficient and does not have its usual special meaning. If
A ≡ [x0 , y0 ] lies on the curve q, meaning that z (x0 , y0 ) = 0, then the polynomial
w (x, y) obtained by replacing x by x + x0 and y by y + y0 in z (x, y) has no constant
term. Then

w (x, y) = g1 x3 + h1 x2 y + i1 xy 2 + j1 y 3 + d1 x2 + e1 xy + f1 y 2 + a1 x + b1 y = 0

is the equation of the Taylor curve qA to q at A. Then A is non-singular, or more

accurately first order non-singular, precisely when at least one of a1 or b1 is
(1)
non-zero. In this case the line TA q = τ A (ha1 : b1 : 0i) is the tangent line to q at A.
19.4. FOLIUM OF DESCARTES 203

Furthermore A is second order non-singular precisely when at least one of d1 , e1 , f1

(2)
is non-zero. In this case TA q = τ A (hd1 : e1 : f1 : a1 : b1 : 0i) is the tangent conic to
q at A.

Note that the two notions of non-singular are completely independent. The tangent
(1) (2)
line TA q is also tangent to the tangent conic TA q at A.

Two cubic curves q1 and q2 intersect at A precisely when A lies on them both. In
this case they are tangent at A precisely when A is a non-singular point of both and
(1) (1)
TA q1 = TA q2 . The two curves are second order tangent at A precisely when A is
(2) (2)
a second order non-singular point of both and TA q1 = TA q2 .

These concepts generalize to higher degree curves and higher order tangents. The next
two sections apply this kind of analysis using tangent conics to two famous curves, and
suggest further questions arising from these notions.

19.4 Folium of Descartes

Suppose the field is not of characteristic three. The folium of Descartes q is a cubic
curve with equation
x3 + y 3 + 3xy = 0 (19.2)

Example 19.5 In the decimal number field the curve is shown in Figure 19.3.
y 3

2.5

t = -2 1.5

0.5
0
-3 -2.5 -2 -1.5 -1 -0.5 0 0.5 1 1.5 2 2.5 3
-0.5
t = 1/2 x
-1
t = -1/2
-1.5
t= 2 -2

-2.5

-3

Figure 19.3: Folium of Descartes

By setting y = tx and substituting, (19.2) becomes

¡ ¢
x2 x + t3 x + 3t = 0.
Thus either x = 0, in which case y = 0, or [x, y] has the form
∙ ¸
3t 3t2
− ,− . (19.3)
1 + t3 1 + t3
204 19. TANGENT LINES AND TANGENT CONICS

Conversely, for every t for which t3 6= −1, the point (19.3) lies on the folium, giving a
parametrization of the curve, including the point [0, 0] when t = 0.

Example 19.6 In the field F13 , Figure 19.4 shows the folium q with the various values
of the parameter t. There are only 10 points, since t = 4, 10 and 12 satisfy t3 = −1 and
so do not contribute.

6 3
5 1
4
3 2
2 9
1 6
0 0
-1
-2 11
-3
-4 5
-5 7
-6 8
-6 -5 -4 -3 -2 -1 0 1 2 3 4 5 6

Figure 19.4: The folium of Descartes in F13 ¦

Fix a point A lying on the folium q corresponding to a value t. Translate by τ −1

A to get
a standard cubic q1 with equation
µ ¶3 µ ¶3 µ ¶µ ¶
3t 3t2 3t 3t2
0= x− + y− +3 x− y−
1 + t3 1 + t3 1 + t3 1 + t3
or
9t 9t2 2
0 = x3 + y 3 − x 2
+ 3xy − y (19.4)
1 + t3 1 + t3
¡3 ¢ 2 ¡ 3 ¢
9 t −2 t 9 2t − 1 t
− x+ y.
(t + 1)2 (t2 − t + 1)2 (t + 1)2 (t2 − t + 1)2
Then A is singular precisely when t = 0, since this is the only value for which both
coefficients of x and y are zero, since the field is not of characteristic three.
(1)
Suppose that t 6= 0. To find the tangent line TA q to q at A, take the linear terms of q1
and translate back using τ A to get
¡ ¢ µ ¶ ¡ ¢ µ ¶
9 t3 − 2 t2 3t 9 2t3 − 1 t 3t2
− 2 2 x+ + 2 2 y+ = 0.
(t + 1) (t2 − t + 1) 1 + t3 (t + 1) (t2 − t + 1) 1 + t3
(1)
Simplify this equation for TA q to get

¡ ¢ ¡ ¢
t 2 − t3 x + 2t3 − 1 y + 3t2 = 0. (19.5)
19.4. FOLIUM OF DESCARTES 205

Returning to (19.4), the existence of the quadratic term 3xy ensures that every point
A lying on q is second order non-singular, even when t = 0. To find the tangent conic
(2)
TA q to q at A, take the terms of degree two or less in q1 and translate back using τ A
to get
µ ¶2 µ ¶µ ¶ µ ¶2
9t 3t 3t 3t2 9t2 3t2
0 = − x + + 3 x + y + − y +
1 + t3 1 + t3 1 + t3 1 + t3 1 + t3 1 + t3
¡3 ¢ 2 µ ¶ ¡ 3 ¢ µ ¶
9 t −2 t 3t 9 2t − 1 t 3t2
− 2 2 x+ + 2 2 y+ .
(t + 1) (t2 − t + 1) 1 + t3 (t + 1) (t2 − t + 1) 1 + t3
(2)
After simplification you get for TA q the equation

¡ ¢ ¡ ¢2 ¡ ¢
0 = 3t 1 + t3 x2 − 1 + t3 xy + 3t2 1 + t3 y 2 + 9t2 x + 9t4 y + 9t3 . (19.6)

When t = 0 this conic is a pair of lines, with equation xy = 0.

Example 19.7 In the field F13 , Figure 19.5 shows the point A ≡ [1, −2], which
corresponds to the parameter t = 11 and lies on the folium q (large open squares).
Using (19.5) the tangent line to q at A (gray boxes) has equation 2x + 3y + 4 = 0.
Using (19.6) the tangent conic to q at A (small black squares) has equation
3x2 + 3xy + 7y 2 + 10x + y + 6 = 0. Notice that the tangent line intersects the folium at
two points, and the tangent conic intersects the folium at four points.

6
5
4
3
2
1
0
-1
-2
-3
-4
-5
-6
-6 -5 -4 -3 -2 -1 0 1 2 3 4 5 6

Figure 19.5: Tangent line and tangent conic at [1, −2] ¦

Example 19.8 (Using some linear algebra) In the decimal number field, to
determine whether the tangent conic is an ellipse, parabola or hyperbola in the usual
decimal number sense, consider the associated quadratic form to (19.6). After
removing the factor 1 + t3 , this has matrix
∙ ¡ ¢ ¸
¡ 3t ¢ − 1 + t3 /2
M≡ .
− 1 + t3 /2 3t2
206 19. TANGENT LINES AND TANGENT CONICS
¡ ¢
Then det M = (−1/4) t6 − 34t3 + 1 . The quadratic equation x2 − 34x + 1 = 0 has
√ p
3
√
solutions x = 17 ± 12 2 which are reciprocals. Thusp when t = 17 ± 12 2 the
3
√
tangent
p conic is a parabola, while for t between 17 − 12 2 ≈ 0.308 . . . and
3
√
17 + 12 2 ≈ 3. 238 . . . the tangent conic is an ellipse, and otherwise a hyperbola.

The graphs of tangent lines and tangent conics to the folium of Descartes are shown
below, illustrating cases where the tangent conic is elliptic, hyperbolic and parabolic.

y 4 y 4

2 2

0 0
-4 -2 0 2 4 -4 -2 0 2 4

x x

-2 -2

-4 -4

Figure 19.6: Elliptic and hyperbolic tangent conics at t = 2 and t = −2

y 4

0
-4 -2 0 2 4

-2

-4

p
3
√
Figure 19.7: Parabolic tangent conic at t = 17 + 12 2 ¦

Exercise 19.4 (Harder) Suppose that A and B are distinct points on the folium q
corresponding respectively to the parameters t and u. Show that the tangent conic
(2) 3
TA q to q at A passes through B if (tu) = 1.
(1)
In the special case when u = t−1 , show that the spread between the tangent line TA q
(2)
at A and the tangent line to the tangent conic TB q at A equals the spread between
(1) (2)
the tangent line TB q at B and the tangent line to the tangent conic TA q at B. ¦
19.5. LEMNISCATE OF BERNOULLI 207

19.5 Lemniscate of Bernoulli

A Cassini oval is a particular form of quartic curve, whose equation in P ≡ [x, y] is

Q (P, F1 ) Q (P, F2 ) = K

for some distinct points F1 and F2 and a non-zero number K. A particular case is the
lemniscate of Bernoulli b, where F1 ≡ [a, 0], F2 ≡ [−a, 0] and K ≡ a4 , giving the
equation
¡ 2 ¢2 ¡ ¢
x + y 2 = 2a2 x2 − y 2 .

This equation is even, in the sense that it involves only x2 ≡ X and y 2 ≡ Y. Setting
a2 ≡ A, it becomes the associated equation in X and Y

(X + Y )2 = 2A (X − Y ) .

To find a parametrization of this associated equation, suppose that Y = tX, so that

2
X 2 (1 + t) = 2AX (1 − t) .

If X = 0 then the choices for Y are Y = 0 and Y = −2A. Otherwise [X, Y ] is of the
form " #
2A (1 − t) 2At (1 − t)
2 , 2
(1 + t) (1 + t)
where t 6= −1.

Suppose B = [x0 , y0 ] is a point on the lemniscate b, so that the equation for the Taylor
curve bB is
³ ´2 ³ ´
2 2 2 2
(x + x0 ) + (y + y0 ) = 2A (x + x0 ) − (y + y0 )

which expands to

0 = x4 + 2x2 y 2 + y 4 + 4x0 x3 + 4y0 x2 y + 4x0 xy 2 + 4y0 y 3

¡ ¢ ¡ ¢
+2 3x20 + y02 − A x2 + 8x0 y0 xy + 2 x20 + 3y02 + A y 2
¡ ¢ ¡ ¢
+4x0 x20 + y02 − A x + 4y0 x20 + y02 + A y.

Take linear terms to get the equation of the tangent line T (1) bB
¡ ¢ ¡ ¢
0 = x0 x20 + y02 − A x + y0 y02 + x20 + A y

(1)
and translate back to get the equation of the tangent line TB b to b at B

¡ ¢ ¡ ¢ ¡ ¢
0 = x0 x20 + y02 − A x + y0 y02 + x20 + A y − A x20 − y02 .
208 19. TANGENT LINES AND TANGENT CONICS

The equation of the tangent conic T (2) bB is

¡ 2 ¢ ¡ ¢
0 = 3x0 + y02 − A x2 + 4x0 y0 xy + x20 + 3y02 + A y 2
¡ ¢ ¡ ¢
+2x0 x20 + y02 − A x + 2y0 x20 + y02 + A y.

(2)
Translate back to get the equation of the tangent conic TB b to b at B

¡ 2 ¢ ¡ ¢
0 = 3x0 + y02 − A x2 + 4x0 y0 xy + x20 + 3y02 + A y 2
¡ ¢ ¡ ¢ ¡ ¢
−4x0 y02 + x20 x − 4y0 y02 + x20 y + 3A x20 − y02 .

Note that
2 ¡ ¢¡ ¢ ¡ ¢
(4x0 y0 ) − 4 3x20 + y02 − A x20 + 3y02 + A = 4 −8Ax20 + 8Ay02 + A2

which is zero precisely when

x20 − y02 = A/8.

This equation defines the discriminant conic of the lemniscate. By the Conic center
(2)
theorem (page 168) the tangent conic TB b is central unless B lies on the intersection
of the lemniscate and this discriminant conic. The discriminant conic is a quadrola.

Exercise 19.5 Show that the lemniscate and the discriminant conic are inverses of
each other: this means that along any line passing through the origin O, the products
of the quadrances from O to the lemniscate and to the discriminant conic are constant.
¦

Lemniscate in the rational number field

Example 19.9 Over the rational number field, suppose that a ≡ 1, so that the
lemniscate b has equation
¡ 2 ¢2 ¡ ¢
x + y 2 = 2 x2 − y 2 .

In Figure 19.8 both the lemniscate and the discriminant conic x2 − y 2 = 1/8, which in
this
£ √case is√a usual
¤ hyperbola, are shown. The two curves intersect at the points
± 5/4, ± 3/4 .
£√ √ ¤
The tangent line and tangent conic at the point 5/4, 3/4 ≈ [0.559 . . . , 0.433 . . .]
are also shown. The latter is the parabola
D √ √ √ E
1 : 2 15 : 15 : −4 5 : −4 3 : 3 .
19.5. LEMNISCATE OF BERNOULLI 209

y
1

-2 -1 1 x 2

-1

-2

Figure 19.8: Parabolic tangent conic

The tangent line and tangent conic at the points [0.4, 0.347 . . .] and [0.7, 0.345 041 . . .]
are shown in Figures 19.9 and 19.10 respectively.
2
y
1

-3 -2 -1 1 x 2 3

-1

-2

Figure 19.9: Hyperbolic tangent conic

2
y
1

-3 -2 -1 1 x 2 3

-1

-2

Figure 19.10: Elliptical tangent conic ¦

210 19. TANGENT LINES AND TANGENT CONICS

Lemniscate in F11

Example 19.10 In the field F11 , Figure 19.11 shows two views of the lemniscate (gray
¡ ¢2 ¡ ¢
squares) with equation x2 + y 2 = 2 x2 − y 2 , and the discriminant conic
(diamonds) with equation x2 − y 2 = 7. These curves intersect at the points [±1, ±4].

5
4
3
2
1
0
-1
-2
-3
-4
-5
-5 -4 -3 -2 -1 0 1 2 3 4 5

Figure 19.11: Lemniscate and discriminant conic: two views

The left diagram in Figure 19.12 shows the tangent line h4 : 7 : 1i (small circles) and
tangent conic h3 : 10 : 1 : 7 : 6 : 9i (small black squares) to the lemniscate (gray
squares) at the point [1, 4]. The tangent conic is a parabola, with directrix h10 : 5 : 1i
(stars) and focus [9, 9] = [−2, −2] (large circle).

The right diagram in Figure 19.12 shows the tangent line (small circles) and tangent
conic h4 : 0 : 3 : 0 : 1 : 3i (small black squares) to the lemniscate (gray squares) at the
point [0, 3]. The tangent conic in this case is a grammola, with diagonal lines h3 : 2 : 4i
and h1 : 3 : 6i, and quadrance K ≡ 9.

5 5
4 4
3 3
2 2
1 1
0 0
-1 -1
-2 -2
-3 -3
-4 -4
-5 -5
-5 -4 -3 -2 -1 0 1 2 3 4 5 -5 -4 -3 -2 -1 0 1 2 3 4 5

Figure 19.12: Tangent conics at [1, 4] and [0, 3] ¦

 Part IV

Applications

211
 20

Triangle spread rules

This chapter introduces concepts for working over the rational and decimal number
fields. It shows how to practically construct a spread ruler, how to define rays and
sectors, and gives the important Triangle spread rules that augment rational
trigonometry in these particular fields, and which are particularly useful for practical
applications. The arguments and definitions are generally informal.

20.1 Spread ruler

The spread ruler shown in Figure 20.1 allows you to measure spreads between two
lines in a similar way that a protractor measures angles between two rays.

.5 .6 .7 .8 .9 .95 .99 1 .99 .95 .9 .8 .7 .6 .5

.4 .4
.3 .3
.2 .2
.1 .1
.05
.05
.02 .02
.01 .01
.005 .005

Figure 20.1: Spread ruler

213
214 20. TRIANGLE SPREAD RULES

Making a spread ruler is perhaps more straightforward than making a protractor.

Consider the right triangle OAB with
spread s at O, with Q (O, A) ≡ 1 and Q (A, B) ≡ Q. Then by B
Pythagoras’ theorem and the Spread ratio theorem (page 77)
Q +1
Q Q
s=
Q+1 s
so that
s
O 1 A
Q= .
1−s
It follows that if O ≡ [0, 0] and A ≡ [1, 0], then the position of B, given s, is
h p i
B = 1, s/ (1 − s) .

Exercise 20.1 Show how to use the symmetry between s and 1 − s to restrict
necessary values of s to the range [0, 1/2] . ¦

Here are some approximate values for the construction of a spread ruler.
sp 0.05 0.1 0.2 0.25 0.3 0.4 0.5
s/ (1 − s) 0.230 0.333 0.5 0.577 0.655 0.816 1.0

20.2 Line segments, rays and sectors

The definitions of this section hold for the decimal or rational number fields, and rely
on properties of positive numbers. For these fields, the terms side and line segment
will be used interchangeably. The point A lies on the line segment A1 A2 precisely
when
A = λ1 A1 + λ2 A2
for some numbers λ1 , λ2 ≥ 0 satisfying λ1 + λ2 = 1. Such a point A is interior to the
line segment precisely when λ1 , λ2 > 0. The notion ‘A lies on A1 A2 ’ is more general
than ‘A is an element of A1 A2 ’, since A1 A2 ≡ {A1 , A2 } has only two elements.

Two line segments A1 A2 and B1 B2 overlap precisely when there is a point which is
interior to both, and are adjacent precisely when there is no point which is interior to
both and exactly one point which is an element of both.

Suppose that three collinear points A1 , A2 and A3 form quadrances Q1 ≡ Q (A2 , A3 ),

Q2 ≡ Q (A1 , A3 ) and Q3 ≡ Q (A1 , A2 ). Then {Q1 , Q2 , Q3 } is a quad triple, so that
given Q1 and Q2 , the third quadrance Q3 is obtained from the triple quad formula
(Q3 − Q1 − Q2 )2 = 4Q1 Q2 .
For a general field this is as much as one can say. However over the decimal or rational
number fields one knows that quadrances are always positive, so that the solutions are
p
Q3 = Q1 + Q2 ± 2 Q1 Q2 .
20.2. LINE SEGMENTS, RAYS AND SECTORS 215

Then you can determine, in terms of the relative positions of A1 A3 and A2 A3 , just
which of these two possibilities occurs.

Collinear quadrance rules Suppose that Q1 ≡ Q (A2 , A3 ), Q2 ≡ Q (A1 , A3 ) and

Q3 ≡ Q (A1 , A2 ) are the quadrances formed by three collinear points A1 , A2 and A3 .
Then
√
1. If A1 A3 and A2 A3 are adjacent then Q3 = Q1 + Q2 + 2 Q1 Q2
√
2. If A1 A3 and A2 A3 are overlapping then Q3 = Q1 + Q2 − 2 Q1 Q2 .

−−−→ ←−−−
A ray A1 A2 , also written A2 A1 , is an ordered pair [A1 , A2 ] of distinct
points, with the convention that
−−−→ −−−→
A1 A2 = A1 A3

precisely when
A2
A3 = λ1 A1 + λ2 A2 A1
for some numbers λ1 and λ2 satisfying λ1 + λ2 = 1 and λ2 ≥ 0.
This notion treats A2 and A3 symmetrically. The point A1 is the base point of the
−−−→ −−−→
ray A1 A2 . A point B lies on the ray A1 A2 precisely when

B = λ1 A1 + λ2 A2
−−−→ −−−→
with λ1 + λ2 = 1 and λ2 ≥ 0. Two rays A1 A2 and B1 B2 are parallel precisely when
A1 A2 is parallel to B1 B2 .
←−−−−→ n←−−− −−−→o
A sector α ≡ A2 A1 A3 is a set A2 A1 , A1 A3 of non-parallel rays with a common
←−−−−→
base point A1 . The point B lies on the sector A2 A1 A3 precisely when

B = λ2 A2 + λ1 A1 + λ3 A3

with λ1 + λ2 + λ3 = 1 and λ2 , λ3 ≥ 0. Figure 20.2 shows (some of) the points B lying
←−−−−→
on α = A2 A1 A3 .

A2

A3

A1

←−−−−→ n←−−− −−−→o

Figure 20.2: The sector A2 A1 A3 = A2 A1 , A1 A3
216 20. TRIANGLE SPREAD RULES

←−−−−→ −−−→ −−−→

A sector α ≡ A2 A1 A3 determines two rays A1 A2 and A1 A3 , together with two lines
A1 A2 and A1 A3 . The spread s (α) of the sector α is the spread between these two
lines, so that
³←−−−−→´
s (α) = s A2 A1 A3 ≡ s (A1 A2 , A1 A3 ) .

20.3 Acute and obtuse sectors

←−−−−→
The sector α ≡ A2 A1 A3 is acute type, abbreviated as (ac), precisely when

Q (A1 , A2 ) + Q (A1 , A3 ) ≥ Q (A2 , A3 )

and obtuse type, abbreviated as (ob), precisely when

Q (A1 , A2 ) + Q (A1 , A3 ) ≤ Q (A2 , A3 ) .

The sector α is a right sector precisely when s (α) = 1. By Pythagoras’ theorem, a

sector is a right sector precisely when it is both acute and obtuse.

Exercise 20.2 Show that these definitions are indeed well-defined. ¦

A general sector determines both a spread and a type. These two pieces of information
can be usefully recorded together when referring to sectors. Figure 20.3 shows two
sectors, the left with an (acute) spread of s = 0.625 (ac) and the right with an (obtuse)
spread of 0.845 (ob).

A2

B2
B3
A3 B1
A1

Figure 20.3: Acute and obtuse sectors

−−−→ −−−→ −−−→

If three mutually non-parallel rays A0 A1 , A0 A2 and A0 A3 have the common base point
←−−−−→
A0 , then there are two possible relations between the two sectors β 3 ≡ A1 A0 A2 and
←−−−−→
β 1 ≡ A2 A0 A3 . They overlap precisely when there is a point B which lies on both
−−−→ −−−→ −−−→
sectors but not on any of the rays A0 A1 , A0 A2 or A0 A3 . They are adjacent precisely
−−−→
when the only points which lie on both sectors lie on the ray A0 A2 . These two
situations are respectively shown in Figure 20.4.
20.3. ACUTE AND OBTUSE SECTORS 217

b3 b3 A1
A1 A2
b1 b1
A2

A3
A3 A0
A0

Figure 20.4: Overlapping and adjacent sectors

←−−−−→
If β 2 ≡ A3 A0 A1 then the three sectors β 1 , β 2 and β 3 may have the property that one
of them overlaps with each of the other two, while those other two are adjacent, as in
either of the diagrams in Figure 20.4. Another possibility is that any two of them are
adjacent, as in Figure 20.5.

A2
b3

b1 A1
A0
b2
A3

Figure 20.5: Three adjacent sectors

←−−−−→
A triangle A1 A2 A3 gives rise to three distinguished sectors, namely α1 ≡ A2 A1 A3 ,
←−−−−→ ←−−−−→
α2 ≡ A3 A2 A1 and α3 ≡ A1 A3 A2 . The spreads of these sectors are then the usual
spreads of the triangle. If the spreads of the three sectors are s1 ≡ s (α1 ), s2 ≡ s (α2 )
and s3 ≡ s (α3 ) then the Triple spread formula asserts that {s1 , s2 , s3 } is a spread
triple, so that
2
(s3 − (s1 + s2 − 2s1 s2 )) = 4s1 s2 (1 − s1 ) (1 − s2 ) . (20.1)

When viewed as a quadratic equation in s3 in the decimal number field, the two
solutions can be labelled the little spread rl = rl (s1 , s2 ) and the big spread
rb ≡ rb (s1 , s2 ) where

p
rl (s1 , s2 ) = s1 + s2 − 2s1 s2 − 2 s1 s2 (1 − s1 ) (1 − s2 )
p
rb (s1 , s2 ) = s1 + s2 − 2s1 s2 + 2 s1 s2 (1 − s1 ) (1 − s2 ).
218 20. TRIANGLE SPREAD RULES

20.4 Acute and obtuse triangles

The point B lies on the triangle A1 A2 A3 precisely when

B = λ1 A1 + λ2 A2 + λ3 A3
for some numbers λ1 , λ2 , λ3 ≥ 0 satisfying λ1 + λ2 + λ3 = 1. Such a point B is interior
to the triangle if λ1 , λ2 , λ3 > 0.

A triangle A1 A2 A3 is acute if all three of its sectors are acute. Otherwise it is obtuse.
If the quadrances of the triangle are Q1 , Q2 and Q3 as usual, then the sector with base
point A1 (or just the sector at A1 ) is acute precisely when Q2 + Q3 ≥ Q1 , and
similarly for the other sectors. So the triangle is acute precisely when
Q1 + Q2 ≥ Q3 Q2 + Q3 ≥ Q1 Q3 + Q1 ≥ Q2 . (20.2)

Observe that if the corresponding spreads of the triangle are s1 , s2 and s3 , then by the
Spread law and the fact that all the quadrances and spreads
³←−−−−are´positive, the sector at
→
A3 is acute, or alternatively the spread of the sector s A1 A3 A2 is acute, precisely
when either
Q1 + Q2 ≥ Q3 or s1 + s2 ≥ s3 .

Exercise 20.3 Show that if a triangle A1 A2 A3 has spreads s1 , s2 and s3 , then any
two of the following inequalities implies the third, and implies the triangle is acute.
s1 ≥ |s2 − s3 | s2 ≥ |s3 − s1 | s3 ≥ |s1 − s2 | . ¦

Exercise 20.4 Show that a triangle can have at most one obtuse sector. ¦

Problem 2 Show that A1 A2 A3 is acute precisely when the circumcenter C lies on

the triangle.

Solution. Suppose that the quadrances of A1 A2 A3 are Q1 , Q2 and Q3 as usual,

and that the quadrea is A. The Affine circumcenter
theorem (page 146) shows that C is the affine combination A3
C = γ 1 A1 + γ 2 A2 + γ 3 A3
where
γ1 ≡ Q1 (Q2 + Q3 − Q1 ) /A C
γ2 ≡ Q2 (Q1 + Q3 − Q2 ) /A A2
γ3 ≡ Q3 (Q1 + Q2 − Q3 ) /A. A1
So C is in the interior of A1 A2 A3 precisely when the three
inequalities for acuteness are satisfied.
20.5. TRIANGLE SPREAD RULES 219

20.5 Triangle spread rules

The following rules apply to the rational and decimal number fields, and those closely
related to them. They provide an important guide to dealing with acute and obtuse
sectors in practical applications of rational trigonometry.

Triangle spread rules Suppose that s1 , s2 and s3 are the respective spreads of the
←−−−−→ ←−−−−→ ←−−−−→
three sectors α1 ≡ A2 A1 A3 , α2 ≡ A3 A2 A1 and α3 ≡ A1 A3 A2 of a triangle A1 A2 A3 .
Then

1. The spread s3 is equal to rb (s1 , s2 ) precisely when s1 and s2 are both acute
2. The spread s3 is obtuse precisely when s1 and s2 are acute and s1 + s2 ≤ 1.

This is summarized in the following table, which gives the value of s3 , depending on s1
and s2 .
s3 s1 (ac) s2 (ac) s1 (ac) s2 (ob)
s1 + s2 ≤ 1 rb (s1 , s2 ) (ob) rl (s1 , s2 ) (ac)
s1 + s2 ≥ 1 rb (s1 , s2 ) (ac) rl (s1 , s2 ) (ac)

Problem 3 Demonstrate the validity of these rules.

Solution (Rule 1). Recall that the Triple spread formula, as a quadratic equation
in s3 , has normal form
2
(s3 − (s1 + s2 − 2s1 s2 )) = 4s1 s2 (1 − s1 ) (1 − s2 ) .

Suppose that
p
s3 = rl (s1 , s2 ) = s1 + s2 − 2s1 s2 − 2 s1 s2 (1 − s1 ) (1 − s2 ). (20.3)

If s1 is acute then s3 + s2 ≥ s1 , so that

p
s2 − s1 s2 ≥ s1 s2 (1 − s1 ) (1 − s2 ).

Use the fact that any spread s satisfies 0 ≤ s ≤ 1 to see that both sides are positive, so
the inequality is maintained when both sides are squared. Thus

s2 (1 − s1 ) ≥ s1 (1 − s2 )

which is equivalent to
s2 ≥ s1 .
220 20. TRIANGLE SPREAD RULES

Similarly if s2 is acute then

s1 ≥ s2 .
Thus if both s1 and s2 are acute then s1 = s2 ≡ s, in which case by (20.3)

s3 = 2s − 2s2 − 2s (1 − s) = 0.

This is impossible, so you may conclude that if s1 and s2 are acute, then
p
s3 = s1 + s2 − 2s1 s2 + 2 s1 s2 (1 − s1 ) (1 − s2 ) = rb (s1 , s2 ) .

Conversely if s3 = rb (s1 , s2 ) then

p
s3 + s2 = s1 + 2s2 (1 − s1 ) + 2 s1 s2 (1 − s1 ) (1 − s2 ) ≥ s1

so s1 is acute, and similarly s2 is acute.

Solution (Rule 2). Recall from Exercise 7.2 (page 90) that the Triple spread
formula S (s1 , s2 , s3 ) = 0 can be rewritten as the equation

s3 (s3 − (s1 + s2 )) (1 − (s1 + s2 ))

= ((s1 + s2 − s3 ) s3 + (s3 − s1 + s2 ) (s3 − s2 + s1 )) (1 − s3 ) . (20.4)

Now s1 and s2 are acute precisely when

s3 − s1 + s2 ≥ 0 and s3 − s2 + s1 ≥ 0

respectively, while s3 is obtuse precisely when

s3 − (s1 + s2 ) ≥ 0.

Any spread s of a triangle satisfies 0 < s ≤ 1. So if s1 and s2 are acute and s1 + s2 ≤ 1,

then s3 must be obtuse, since otherwise the right hand side of (20.4) is strictly positive
while the left hand side is negative.

Conversely suppose s3 is obtuse. Then s1 and s2 are acute by Exercise 20.4. If

s1 + s2 > 1 then the left hand side of (20.4) is strictly negative, so that

((s1 + s2 − s3 ) s3 + (s3 − s1 + s2 ) (s3 − s2 + s1 )) = s3 s1 + s3 s2 + 2s1 s2 − s21 − s22 < 0.

But then
2
s3 (s1 + s2 ) < (s1 − s2 )
which is impossible since
2 2
s3 (s1 + s2 ) ≥ (s1 + s2 ) > (s1 − s2 ) .

Thus s1 + s2 ≤ 1.
 21

Two dimensional problems

This chapter gives some geometrical applications in the decimal number plane.

21.1 Harmonic relation

Problem 4 Distinct points A1 , A2 and A3 lie on a line l, with B1 , B2 and B3 points

on the respective altitudes from A1 , A2 and A3 to l, such that B1 , B2 and A3 are
collinear, as are A1 , B2 and B3 , as in Figure 21.1. Define Q1 ≡ Q (A1 , B1 ),
Q2 ≡ Q (A2 , B2 ) and Q3 ≡ Q (A3 , B3 ). Show that {1/Q1 , 1/Q2 , 1/Q3 } is a quad
triple.

B3
B1
B2 Q3
Q1
Q2

A1 P3 A2 P1 A3
P2

Figure 21.1: Harmonic configuration

Solution. Define P1 ≡ Q (A2 , A3 ), P2 ≡ Q (A1 , A3 ) and P3 ≡ Q (A1 , A2 ). Then by

221
222 21. TWO DIMENSIONAL PROBLEMS

the Twist ratio theorem (page 78)

Q2 Q1
t (A3 A2 , A3 B2 ) = = (21.1)
P1 P2
Q2 Q3
t (A1 A2 , A1 B2 ) = = . (21.2)
P3 P2
By the Triple quad formula {P1 , P2 , P3 } forms a quad triple, so that
(P1 + P3 − P2 )2 = 4P1 P3 .
Divide both sides by P22 and substitute using (21.1) and (21.2) to get
µ ¶2
Q2 Q2 Q22
+ −1 =4
Q1 Q3 Q1 Q3
or µ ¶2
1 1 1 1 1
+ − =4 .
Q1 Q3 Q2 Q1 Q3
This is the statement that {1/Q1 , 1/Q2 , 1/Q3 } is a quad triple.

21.2 Overlapping triangles

Problem 5 Two triangles ABC and ABD share a side AB as shown in Figure 21.2,
with quadrances as indicated. What are the quadrances Q (A, E), Q (B, E), Q (C, E)
and Q (D, E)?

C
D
169 81
64
E
100 s3
s1 s2
A 25 B

Figure 21.2: Overlapping triangles

Solution. Introduce the spreads s1 , s2 and s3 of the sectors of ABE as shown. Use
the Cross law in ABC, together with the fact that 81 + 25 > 64, to get
2
(81 + 25 − 64) 176
s1 = 1 − = (ac).
4 × 25 × 81 225
21.2. OVERLAPPING TRIANGLES 223

Similarly in ABD, since 169 + 25 > 100,

(169 + 25 − 100)2 2016

s2 = 1 − = (ac).
4 × 169 × 25 4225

Now use the Triple spread formula in ABE to obtain the quadratic equation
975 136 215 296
s23 − s3 + = 0. (21.3)
950 625 2313 441
Since s1 and s2 are acute, and
176 2016 47 888
s1 + s2 = + = >1
225 4225 38 025
the Triangle spread rules (page 219) show that the correct solution to (21.3) is

s3 = rb (s1 , s2 ) (ac)
487 568 10 528 √
= + 154 (ac).
950 625 316 875
Then the Spread law in ABE gives
µ ¶
176/225 2016/4225 1 487 568 10 528 √
= = + 154 .
Q (B, E) Q (A, E) 25 950 625 316 875
This yields the values
34 556 382 2238 516 √
Q (A, E) = − 154
525 625 525 625
56 649 307 3669 666 √
Q (B, E) = − 154.
525 625 525 625
Now the Collinear quadrance rules show that since AC and AE are overlapping,
p
Q (C, E) = Q (A, C) + Q (A, E) − 2 Q (A, C) Q (A, E)
111 662 307 7758 666 √
= − 154
525 625 525 625
and similarly
p
Q (D, E) = Q (B, D) + Q (B, E) − 2 Q (B, D) Q (B, E)
18 789 082 1476 384 √
= + 154.
525 625 525 625
Note, perhaps surprisingly, that the square roots involved work out pleasantly,
meaning that all expressions of the form
q
√
a + b 154
√
which occur turn out to be expressible in the simpler form c + d 154, with c and d
rational numbers.
224 21. TWO DIMENSIONAL PROBLEMS

21.3 Eyeball theorem

This result is described in [Gutierrez].

Problem 6 Suppose that two circles have centers C1 and C2 , respective quadrances
K1 and K2 , and that tangents from each center to the other circle are drawn,
intersecting the two circles in points A, B and E, F respectively, as in Figure 21.3.
Show that
Q (A, B) = Q (E, F ) .

K1
A E K2
s
M C2
C1 Q F
B

Figure 21.3: Eyeball theorem

Solution. Suppose that Q (C1 , C2 ) ≡ Q and define the spread s ≡ s (C1 A, C1 C2 ).

Then use the Spread ratio theorem (page 77) to find that

K2
s= .
Q

Let M denote the midpoint of the side AB, so that in the right triangle AM C1

Q (A, M )
s= .
K1
From this
K1 K2
Q (A, M ) =
Q
so that by the Midpoint theorem (page 60)

4K1 K2
Q (A, B) = .
Q

This is symmetric in K1 and K2 , so it also equals Q (E, F ).

21.4. QUADRILATERAL PROBLEM 225

21.4 Quadrilateral problem

Problem 7 A quadrilateral A1 A2 A3 A4 has quadrances Q12 ≡ 65, Q34 ≡ 26, and

Q14 ≡ 49, and diagonal quadrances Q13 ≡ 61 and Q24 ≡ 100 as in Figure 21.4. Find
Q ≡ Q23 .

A2
Q

A3
65 100

26
61
A1 A4
49

Figure 21.4: A quadrilateral problem

Solution 1: (Using Euler’s function). The Four point relation (page 191) states
that
E (100, 65, 49, 61, 26, Q) = 0.
This becomes the quadratic equation
(Q − 114)2 = (80)2
←−−−−→
with solutions 34 and 194. But 65 + 49 > 100 so A2 A1 A4 is acute, and then so is
←−−−−→
A2 A1 A3 . Thus 65 + 61 > Q, so that Q = 34.

Solution 2: (Using spreads and the Triangle spread rules). Let the
intersection of the diagonal lines A1 A3 and A2 A4 be C. From the Cross law and the
definitions of acute and obtuse,
³←−−−−→´ (49 + 61 − 26)
2
25
s A4 A1 A3 = 1 − = (ac).
4 × 49 × 61 61
Similarly
³←−−−−→´ ³←−−−−→´
s A2 A1 A4 = 64/65 (ac) s A1 A4 A2 = 16/25 (ac)
³←−−−−→´ ³←−−−−→´
s A1 A4 A3 = 25/26 (ac) s A1 A2 A4 = 784/1625 (ac).

This yields Figure 21.5, also showing the unknown spreads

³←−−−−→´ ³←−−−→´
x ≡ s A2 A1 A3 and z ≡ s A1 CA4 .
226 21. TWO DIMENSIONAL PROBLEMS

A2
784 Q
1625

A3
C
z

x
A1 25/61 16/25 A4
64/65 25/26

Figure 21.5: Spreads known and unknown

Then the Triple spread formula applied to the spread triple {25/61, 16/25, z} gives the
quadratic equation
µ ¶2 µ ¶2
801 144
z− = .
1525 305
Now
1601
25/61 + 16/25 = >1
1525
so use the Triangle spread rules in A1 A4 C to see that
801 144 1521
z= + = (ac).
1525 305 1525
Although you could now similarly solve for x using the Triple spread formula in
A1 A2 C, another approach is to apply the Two spread triples theorem (page 98). Since
{x, 1521/1525, 784/1625} and {x, 25/61, 64/65} are both spread triples, and the sector
←−−−→
A1 CA2 is obtuse,
¡ 1521 ¢
784 2
¡ ¢2
1525 − 1625 − 2561 − 64
65
x = ¡ ¢
2 × 1521 784 25 64 1521 784 25 64
1525 + 1625 − 61 − 65 − 2 × 1525 × 1625 + 2 × 61 × 65
1849
= (ac).
3965
Now use the Cross law in A1 A2 A3 ,
µ ¶
2 1849
(Q − 65 − 61) = 4 × 61 × 65 × 1 −
3965
to obtain
(Q − 126)2 = (92)2 .
Since x is acute, 65 + 61 ≥ Q, so the solution must be
Q = 126 − 92 = 34.
 22

Three dimensional problems

This chapter illustrates applications of rational trigonometry and universal geometry

to practical problems involving three-dimensional space over the decimal numbers.
Giving a careful and reasonably complete introduction to three-dimensional geometry
is not easy, which is one of the reasons why no-one has done it yet. Instead, the usual
physical arguments and description by pictures will be adopted, which is of course
logically unsatisfying.

22.1 Planes

The notions of parallel and perpendicular lines extend to three-dimensional space.

Rather briefly, a plane Π is given by a linear equation in the coordinates [x, y, z] of
space, with the plane passing through the non-collinear points A, B and C denoted
ABC. We’ll assume that for the decimal number field most of the results of two
dimensional geometry developed thus far hold in any plane in three-dimensional space.

Two planes are parallel if they do not intersect. A line n is perpendicular to a plane
Π if it is perpendicular to every line lying on Π. In such a case n is a normal to Π.
Any two lines perpendicular to a plane Π are themselves parallel.

Define the spread S (Π1 , Π2 ) between the planes Π1 and Π2 to be the spread s (n1 , n2 )
between respective normals n1 and n2 . Two planes Π1 and Π2 are perpendicular
precisely when S (Π1 , Π2 ) = 1; this is equivalent to the condition that one of the planes
contains (or passes through) a normal to the other.

The spread between a line l and a plane Π intersecting at a point A is defined to be

the spread between l and the line m formed by intersecting Π with the plane through l
and the normal n to Π at A.

227
228 22. THREE DIMENSIONAL PROBLEMS

22.2 Boxes

A box is assumed to be rectangular, meaning that any two of its faces which meet are
perpendicular.

Problem 8 The horizontal sides of a box have quadrances 3 and 4, while the vertical
side has quadrance 5. Find the quadrances of the long diagonals, the spread that they
make with the base, and the possible spreads between two long diagonals.

Solution. Label the vertices of the box as shown, with

G
Q (A, B) = 3 Q (B, C) = 4 Q (C, G) = 5. F

Then by Pythagoras’ theorem

H E
Q (A, C) = Q (A, B) + Q (B, C) = 3 + 4 = 7
P
5
and so also
C 4
B
Q (A, G) = Q (A, C) + Q (C, G) = 7 + 5 = 12.
3
Thus the quadrance of the long diagonal side AG is 12
D A
and by symmetry the other long diagonal sides BH, DF
and CE also have quadrance 12. The spread that any of these long diagonals makes
with the base (the plane containing A, B, C and D) is

Q (C, G) 5
s (AC, AG) = = .
Q (A, G) 12

If P is the center of the box then the quadrance from P to any vertex is one quarter
the quadrance of a long diagonal side, hence 3. The spread between the two diagonals
AG and BH, which intersect at P , is then equal to the spread s (P A, P B) in the
equilateral triangle AP B with equal quadrances 3, which by the Equilateral triangle
theorem (page 125) is 3/4.

The spread between the two diagonals AG and DF is the spread s (P A, P D) in the
isosceles triangle DP A with quadrances 3, 3 and 4. By the Isosceles triangle theorem
(page 122) this is µ ¶
4 4 8
s (P A, P D) = 1− = .
3 4×3 9
Similarly µ ¶
5 5 35
s (P A, P E) = 1− = .
3 4×3 36
The three possibilities for spreads between diagonals are 3/4, 8/9 and 35/36.
22.2. BOXES 229

Exercise 22.1 Show more generally that if the quadrances of a box are P, Q and R
then the three spreads formed by pairs of long diagonals are

4P (Q + R)
2
(P + Q + R)
4Q (R + P )
(P + Q + R)2
4R (P + Q)
. ¦
(P + Q + R)2

Exercise 22.2 Show that in Problem 8 the spread between the plane ABP and the
line P F is 20/27. ¦

Problem 9 The top V of a flagpole subtends a spread of 0.12 at a point A which is

a distance of 70 due south, and a spread of 0.19 at a point B which is due west of the
flagpole. Calculate the distance |A, B| from A to B.

Solution. This problem is given in terms of distance,

so first convert the information into rational trigonometry.
If the base of the flagpole is C then Q (A, C) = (70)2 = 4900. V
In the right triangle ACV the spread at A is 0.12, so
the spread at V is 1 − 0.12 = 0.88, and the Spread law gives
0.12 3 7350
Q (C, V ) = × Q (A, C) = × 4900 = .
0.88 22 11
B .19
In the right triangle BCV the spread at B is 0.19, so C
the spread at V is 1 − 0.19 = 0.81, and the Spread law gives .12 2
(70)
0.81 A
Q (C, B) = × Q (C, V )
0.19
81 7350 595 350
= × = .
19 11 209
Use Pythagoras’ theorem to get
595 350 1 619 450
Q (A, B) = 4900 + = .
209 209
So far no approximations have been introduced. To calculate the distance from A to
B, take the square root of the quadrance, to get
√
35 276 298
|A, B| =
209
which is approximately 88.02.
230 22. THREE DIMENSIONAL PROBLEMS

22.3 Pyramids

A pyramid consists of a rectangular base with an apex directly above the center of the
base.

Problem 10 A square ABCD with quadrance 10 is the base of a pyramid. The

quadrance from the center P of the base to the apex V, directly above it, is 18. Find
the spread s (V A, V C), and the spread between the planes ABV and BCV.

Solution. The triangle AV C is isosceles with P the

midpoint of the side AC, and V P
bisects the vertex at V . In the right triangle ABC, use V
Pythagoras’ theorem to see that Q (A, C) = 20, so that

Q (A, P ) = Q (A, C) /4 = 5.
18 F C
D
Use Pythagoras’ theorem in the
right triangle AP V to see that Q (A, V ) = 23, so that P 10
A B
s (V A, V P ) = 5/23. 10

By symmetry s (V C, V P ) = 5/23, so now use the Equal spreads theorem (page 94) to
get µ ¶
5 5 360
s (V A, V C) = 4 × × 1− = .
23 23 529

To determine the spread between the planes ABV and BCV , find the foot F of the
altitude from A to V B, which by symmetry is also the foot of the altitude from C to
V B. Then the plane AF C is perpendicular to V B, so that the spread S between the
planes ABV and BCV is equal to the spread r between the lines AF and F C. The
isosceles triangle ABV has quadrances 23, 23 and 10, so by the Isosceles triangle
theorem (page 122)
µ ¶
10 10 205
s (V A, V B) = 1− =
23 4 × 23 (23)2
and thus
Q (A, F ) = Q (A, V ) s (V A, V F ) = 205/23.
Similarly
Q (C, F ) = 205/23.
So in the isosceles triangle AF C
µ ¶
20 20 1656
r = s (F A, F C) = 1− = = S.
205/23 4 × (205/23) 1681
22.4. WEDGES 231

22.4 Wedges

A wedge is formed by two intersecting planes, often with one of the planes horizontal.

Problem 11 Suppose an inclined plane has a spread of S with the horizontal plane.
An insect climbing up the plane walks on a straight line which makes a spread of r
with the line of greatest slope. At what spread to the horizontal does the insect climb
on this path?

D C

E B
rs
s

Figure 22.1: Path on a wedge

Solution. Denote by AC a line making the maximum possible spread s ≡ S with the
horizontal, and AD the path of the insect, as in Figure 22.1. The spread between AC
and AD is r, and you need to find the spread s (AE, AD) . There are no units in the
problem, so assume that Q (B, C) = Q (E, D) = 1. From the right triangle ABC

Q (B, C) 1
Q (A, C) = = .
s s

From the right triangle ACD, with right vertex at C,

Q (A, C) 1
Q (A, D) = = .
1−r s (1 − r)

Thus the right triangle ADE gives

Q (D, E)
s (AE, AD) = = s (1 − r) = S (1 − r) .
Q (A, D)
232 22. THREE DIMENSIONAL PROBLEMS

22.5 Three dimensional Pythagoras’ theorem

Problem 12 Suppose that three points B1 , B2 and B3 in space are distinct from a
point C and that the three lines CB1 , CB2 and CB3 are mutually perpendicular. Let
A be the quadrea of the triangle B1 B2 B3 , and A1 , A2 and A3 the quadreas of the
triangles CB2 B3 , CB1 B3 and CB1 B2 respectively. Show that

A = A1 + A2 + A3 .

B2

Q1
P2
Q3
B3 P3
C

Q2 P1

B1

Figure 22.2: Three-dimensional Pythagoras

Solution. Let Q1 , Q2 and Q3 denote the quadrances of the triangle B1 B2 B3 , with

P1 = Q (B1 , C), P2 = Q (B2 , C) and P3 = Q (B1 , C). Since the triangles
CB2 B3 , CB1 B3 and CB1 B2 are right triangles,
Q1 = P2 + P3 Q2 = P1 + P3 Q3 = P1 + P2 .

The quadrea A of B1 B2 B3 is
2
A = 4Q1 Q2 − (Q1 + Q2 − Q3 )
= 4 (P2 + P3 ) (P1 + P3 ) − 4P32
= 4 (P2 P3 + P1 P3 + P1 P2 ) .
But by the Right quadrea theorem (page 68)
A1 = 4P2 P3 A2 = 4P1 P3 A3 = 4P1 P2 .
Thus A = A1 + A2 + A3 .

Exercise 22.3 Show that any triangle B1 B2 B3 forming part of such a right
tetrahedron is acute, and given such a triangle there are in general exactly two such
tetrahedra. ¦
22.6. PAGODA AND SEVEN-FOLD SYMMETRY 233

22.6 Pagoda and seven-fold symmetry

Problem 13 A retired engineer decides to build the roof of a pagoda with a base of
a regular 7-gon, with the quadrance of each side 2, and the apex V above the center
C of the regular 7-gon at a quadrance of 1 from the base. The roof then consists of
seven identical isosceles triangles. What should the quadrances and spreads of these
triangles be?

A4
V A2
s1 r1
r
A6 1 r1
s r A0
C

A8
2 A12

A10

Figure 22.3: A seven-sided pagoda

Solution. Suppose the regular 7-gon is A0 A2 A4 A6 A8 A10 A12 as in Figure 22.3. The
lines A C, A2 C,
³0← · · · , A12 C form a regular star of order seven, so the spread
−−−→´
s ≡ s A0 CA2 must satisfy
¡ ¢2
S7 (s) = s 7 − 56s + 112s2 − 64s3 = 0.
Of the three approximate solutions,
0.188 255 0.611 260 0.950 484
the relevant one is
s ≈ 0.611 260 (ac).
Define the spreads of the sectors
³←−−−→´ ³←−−−→´
r ≡ s CA0 A2 = s CA2 A0
³←−−−→´ ³←−−−→´
r1 ≡ s V A0 A2 = s V A2 A0
³←−−−→´
s1 ≡ s A0 V A2 .
234 22. THREE DIMENSIONAL PROBLEMS

Use the Isosceles triangle theorem (page 122) with A0 A2 C to get

s = S2 (r) = 4r (1 − r)

so that √
1± 1−s
r= .
2
This gives the possibilities

r ≈ 0.188 25 r ≈ 0.811 75

and the Triangle spread rules show that the relevant one is

r ≈ 0.811 75 (ac).

Use the Spread law in A0 A2 C to see that

rQ (A0 , A2 )
Q (A0 , C) = Q (A2 , C) =
s
0.811 745
≈ × 2 ≈ 2.655 9.
0.611 260
Then use Pythagoras’ theorem in A0 CV to obtain

Q (A0 , V ) ≈ 2.655 9 + 1 = 3.655 9.

Apply the Isosceles triangle theorem to A0 A2 V to get

4Q (A0 , V ) (1 − r1 ) = Q (A0 , A2 ) = 2

so that
r1 ≈ 0.863 24
and
s1 = S2 (r1 ) = 4r1 (1 − r1 ) ≈ 0.472 24.
The triangle A0 A2 V thus has approximate quadrances 2, 3.655 9 and 3.655 9, and
respective approximate spreads 0.472 2, 0.863 2 and 0.863 2.

Exercise 22.4 Show that the spread S between the planes V A0 A2 and V A2 A4 is
approximately 0.224 4. ¦
 23

Physics applications

Some applications to physics are given, including maximizing the trajectory of a

projectile, a derivation of Snell’s law, and a rational formulation of Lorentzian addition
of velocities in Einstein’s special theory of relativity. An example of algebraic
dynamics over a finite field is discussed. Some basic calculus will be assumed here.

23.1 Projectile motion

The motion of a projectile is a parabola, and if the projectile begins at the origin with
velocity −
→
v ≡ [a, b] as in Figure 23.1, then its position at time t is given by
∙ ¸
gt2
at, bt −
2
where g is the acceleration due to gravity.

s
v=[a,b ]

[0,0] x

Figure 23.1: Projectile motion

235
236 23. PHYSICS APPLICATIONS

√
Problem 14 Given that the initial speed v ≡ a2 + b2 is fixed, what spread s from
the vertical results in the projectile traveling the farthest horizontally before it comes
to ground again at a point [x, 0] for some x?

Solution. The projectile comes to ground at time t, where

gt2
bt − =0
2
so that either t = 0 or t = 2b/g. Using rational trigonometry, quadrance is preferred
over distance, so the question is what value of A ≡ a2 and B ≡ b2 , subject to the
condition A + B = v 2 ≡ V , results in the horizontal quadrance

2 4AB
x2 = (at) =
g2

being maximized? This is then the problem of maximizing the product AB of two
numbers A and B given their sum V . The maximum occurs when A = B = V /2,
giving a maximum horizontal quadrance of

V2
x2 = .
g2

So the projectile should be fired at a spread of s = 1/2 from the vertical.

Problem 15 Suppose that the projectile is fired from the origin on a hill represented
by the line l through the origin making a spread of r with the vertical as in Figure
23.2. Given that the initial speed v is fixed, what spread s from the vertical results in
a maximal horizontal displacement after landing?

s v=[a,b ]
r

[0,0] x

Figure 23.2: Projectile fired on a hill

23.1. PROJECTILE MOTION 237
¡ ¢
Solution. The hill is determined by the equation x2 = r x2 + y 2 and so the
projectile intercepts the hill when
³ ¡ ¢2 ´
2 2
(at) = r (at) + bt − gt2 /2 .
This yields that t = 0, or t satisfies the quadratic equation
µ ¶2
2b 4a2 (1 − r)
t− = .
g g2 r
Thus r
2b 2a 1 − r
t= ± .
g g r
To maximize the horizontal displacement, you need to maximize at, or equivalently
f (a, b) ≡ ab ± a2 c
by choosing a and b subject to the constraint
g (a, b) ≡ a2 + b2 = V (23.1)
and where c is the constant r
1−r
c≡ .
r
This can now be done by converting it to a one-dimensional calculus problem, but it is
also interesting to apply the method of Lagrange. At a relative maximum the gradients
∇f = (b ± 2ac, a) ∇g = (2a, 2b)
should be proportional, implying that
(b ± 2ac) b − a2 = 0.
Rearrange and square to eliminate the ambiguity of the sign
¡ ¢2
4a2 b2 c2 = a2 − b2
and substitute using (23.1) to get
¡ ¢ ¡ ¢2
4a2 V − a2 c2 = 2a2 − V .
This quadratic equation in a2 can be written
µ ¶2
2 V V 2 c2 V 2 (1 − r)
a − = = .
2 4 (1 + c2 ) 4
Thus
V ¡ √ ¢
a2 = 1± 1−r
2
and the spread s between the initial direction and the vertical is
√
a2 1± 1−r
s= = .
V 2
But this is equivalent to
r = 4s (1 − s) = S2 (s)
so that the projectile’s initial direction should bisect the vertex formed by the hill and
the vertical. Note that there are two solutions, one downhill and the other uphill.
238 23. PHYSICS APPLICATIONS

23.2 Algebraic dynamics

Recently mathematicians have begun investigating dynamics in finite fields. Here is a

particularly simple case modelled on the usual projectile motion under constant
negative acceleration due to gravity. Whether such an example has any possible
physical significance is unclear, but it seems interesting from a mathematical
perspective.

Example 23.1 Suppose that in F11 a particle starts at time t = 0 with position
p0 ≡ [0, 0], velocity v0 ≡ [1, 3] and has constant acceleration at ≡ [0, −1] for times
t = 0, 1, 2, 3, · · · . Suppose that subsequent positions and velocities are determined for
future times by the equations

pt+1 ≡ pt + vt
vt+1 ≡ vt + at .

This results in the following positions and velocities, which then repeat.

Time 0 1 2 3 4 5
Position [0, 0] [1, 3] [2, 5] [3, 6] [4, 6] [5, 5]
Velocity [1, 3] [1, 2] [1, 1] [1, 0] [1, 10] [1, 9]

Time 6 7 8 9 10 11
Position [6, 3] [7, 0] [8, 7] [9, 2] [10, 7] [0, 0]
Velocity [1, 8] [1, 7] [1, 6] [1, 5] [1, 4] [1, 3]
£ ¤
The position at time t is t, 5t2 − 2t . The trajectory contains exactly those points
lying on the curve with equation x2 + 4x + 2y = 0, which turns out to be a parabola
(black circles) in the sense of Chapter 15. The directrix is the line l ≡ h0 : 1 : 3i (gray
boxes) and the focus is F ≡ [9, 7] (open box) as shown in Figure 23.3. Notice, perhaps
surprisingly, that the vertex of this parabola is the point [9, 2].

10
9
8
7
6
5
4
3
2
1
0
0 1 2 3 4 5 6 7 8 9 10

Figure 23.3: Trajectory in F11 ¦

23.3. SNELL’S LAW 239

23.3 Snell’s law

Problem 16 Suppose a particle travels from the point A ≡ [0, a] to the point
B ≡ [c, −b], where a, b > 0, via some variable point D ≡ [x, 0] on the horizontal axis
as in Figure 23.4. If the particle has speed v1 in the region y ≥ 0, and speed v2 in the
region y < 0, what choice of D minimizes the total time taken?

A
s1
v1
a

O D c-x C
x

v2 b
s2
B

Figure 23.4: Snell’s Law

Solution. The basic formula relating distance d, time t and speed v is

v = d/t. (23.2)

This is not a formula involving universal geometry, as distance is involved.

Nevertheless, let’s proceed some way in the classical framework before switching over
to rational trigonometry. With |O, D| = |x| and |D, C| = |c − x|, the times t1 and t2
taken to travel from A to D (in a straight line) and from D to B respectively are
√
|A, D| a2 + x2
t1 = =
v1 v
q 1
2
|D, B| (c − x) + b2
t2 = = .
v2 v2
The total time t taken is then
q
√
a2 + x2 (c − x)2 + b2
t= + .
v1 v2
This is a function of x, since a, b, v1 and v2 are constants. You could now use calculus
to find the value of x at which this function attains a maximum or minimum.
240 23. PHYSICS APPLICATIONS

√
To do so, the derivative of x is required. Instead, let’s reconsider the problem from
the viewpoint of rational trigonometry.

Since Q ≡ d2 is a rational concept, it makes sense to square (23.2), obtaining

V = Q/T

where V ≡ v 2 and T ≡ t2 . Since Q (A, D) = a2 + x2 and Q (D, B) = (c − x)2 + b2 , the

squared times T1 and T2 taken to travel from A to D and from D to B respectively are

a2 + x2
T1 = (23.3)
V1
2
(c − x) + b2
T2 = . (23.4)
V2

Now t = t1 + t2 , so Exercise 5.8 shows that {T, T1 , T2 } is a quad triple. All three
quantities depend on a variable x and the aim is to choose x so as to minimize T. The
following argument deals with this general situation.

Suppose that {T, T1 , T2 } is a quad triple, so that

2
(T1 + T2 − T ) = 4T1 T2 (23.5)

and that all three quantities T, T1 and T2 depend on a variable x. Take differentials to
obtain µ ¶
dT1 dT2 dT d (T1 T2 )
2 (T1 + T2 − T ) + − =4 . (23.6)
dx dx dx dx
To maximize or minimize T, set
dT
= 0.
dx
Square (23.6) to get
µ ¶2 µ ¶2
2 dT1 dT2 dT1 dT2
(T1 + T2 − T ) + = 4 T2 + T1 .
dx dx dx dx

Now substitute (23.5) so that

µ ¶2 µ ¶2
dT1 dT2 dT1 dT2
T1 T2 + = T2 + T1 .
dx dx dx dx

Upon expansion, rearrangement and cancellation of an extraneous factor T1 − T2 , this

becomes the following general formula for a maximum or minimum

µ ¶2 µ ¶2
dT1 dT2
T2 = T1 . (23.7)
dx dx
23.3. SNELL’S LAW 241

Now to return to the case at hand, apply (23.7) to (23.3) and (23.4) where

dT1 2x
=
dx V1
dT2 2 (x − c)
= .
dx V2

You get
³ ´
¡ 2 ¢
(c − x)2 + b2 4x2 a + x2 4 (c − x)
2
× 2 = ×
V2 V1 V1 V22
or ¡ 2 ¢
2
V2 (c − x) a + x2
= × .
V1 (c − x)2 + b2 x2
But the spreads s1 and s2 made by the lines AD and DB respectively with the vertical
are
x2
s1 = 2
a + x2
and
(c − x)2
s2 = 2 .
(c − x) + b2
This yields Snell’s Law–The time taken is minimized when

V2 s2
= .
V1 s1

The rational solution presented here avoids differentiation of the square root function
and uses only derivatives of linear and quadratic functions.

This analysis also suggests a view of physics in which not only the square of distance,
but also the squares of speed and time play a larger role. Such ideas were introduced
in Einstein’s theory of relativity in 1905. In fact Einstein showed that neither the
square of distance nor the square of time was ultimately of significance, but in suitable
units only the difference between them. The square of mass also figures prominently.

In retrospect one can speculate that if rational trigonometry had been developed prior
to the twentieth century, then the value of Einstein’s revolutionary ideas would have
been recognized more readily, and indeed they might have been anticipated earlier.
Universal geometry and relativity theory naturally have common aspects.

Perhaps there is the potential to take this further, as current formulations of special
(and general) relativity rely on square root functions, and from the point of view of
universal geometry this is not optimal. The next section shows how to eliminate this
dependence in one special situation.
242 23. PHYSICS APPLICATIONS

23.4 Lorentzian addition of velocities

If a train travels along a track with speed v1 and a bullet is fired from the train in the
same direction with speed v2 with respect to the train, then in Newtonian mechanics
the speed v of the bullet with respect to the ground is the sum of the two speeds

v = v1 + v2 . (23.8)

Thus the respective squares V, V1 and V2 of the speeds v, v1 and v2 form a quad triple,
in other words

2
(V1 + V2 − V ) = 4V1 V2 . (23.9)

In Einstein’s special theory of relativity, (23.8) needs to be modified to

v1 + v2
v= (23.10)
1 + v1 v2
where units have been chosen so that the speed of light is c = 1. Square both sides of
(23.10) and rearrange to get
¡ ¢
v 2 1 + 2v1 v2 + v12 v22 = v12 + 2v1 v2 + v22

or ¡ ¢
v 2 − v12 − v22 + v 2 v12 v22 = 2v1 v2 1 − v 2 .
Then square both sides again to get

2 2
(V1 + V2 − V − V V1 V2 ) = 4V1 V2 (1 − V ) . (23.11)

Note that for small values of V, V1 and V2 this is approximated by (23.9). Furthermore
(23.11) can be rewritten as the symmetric expression

2
(V + V1 + V2 − V V1 V2 ) = 4 (V V1 + V V2 + V1 V2 − 2V V1 V2 )

which is a form quite close to the Triple twist formula (page 93).
 24

Surveying

In this chapter classical problems in surveying are solved using rational trigonometry,
such as finding heights of objects from a variety of measurements, and Regiomontanus’
problem of determining the maximum spread subtended by a window. Some of the
examples are parallel to ones from [Shepherd], allowing a comparison between rational
and classical methods. As an application of one of the formulas obtained, the
important spherical analogue of Pythagoras’ theorem is derived.

24.1 Height of object with vertical face

Problem 17 An observer at A measures the vertical spread s to the point B directly

above C. The quadrance Q (A, C) ≡ P is known. What is the vertical quadrance
Q ≡ Q (B, C)?

Solution. The Complementary spreads theorem (page 79)

shows that the spread at B is 1 − s, so the Spread law gives
sP B
Q= .
1−s
R
Q
Example 24.1 Suppose the quadrance from A to C is 100 A
and the spread at A is measured with a theodolite to be
s
s ≡ 0.587. Then µ ¶ P C
0.587
Q= × 100 = 142.131. ¦
0.413

243
244 24. SURVEYING

24.2 Height of object with inaccessible base

Problem 18 The points A1 , A2 and C are horizontal and in a line, and the point A3
is vertically above C, as in either of the diagrams in Figure 24.1. The spreads s1 and
s2 in A1 A2 A3 are measured, and the quadrance Q3 ≡ Q (A1 , A2 ) is known. What is
the vertical quadrance Q ≡ Q (A3 , C)?

A3
s3 s3
Q2 A3
Q Q2
Q1 Q1
Q
s1 s2 s1 s2
A1 Q3 A2 R2 C A1 R1 C R 2 A2
R1 Q3

Figure 24.1: Height from two spread readings

Solution. Let s3 ≡ s (A3 A1 , A3 A2 ). The Triple spread formula as a quadratic

equation in s3 is

(s3 − (s1 + s2 − 2s1 s2 ))2 = 4s1 s2 (1 − s1 ) (1 − s2 ) .

For each of the two solutions the triangle A1 A2 A3 may be solved using the Spread law
for the quadrance Q1 , since Q3 is known. Then the right triangle A2 A3 C with right
vertex C may be solved, using s2 and Q1 , to obtain Q.
³←−−−−→´
Example 24.2 Suppose that Q3 ≡ 25 and that s1 ≡ s A2 A1 A3 ≡ 0.2352 (ac) and
³←−−−−→´
s2 ≡ s A1 A2 A3 ≡ 0.3897 (ob) as in the first of the diagrams in Figure 24.1. The
Triple spread formula becomes (s3 − 0.4416)2 = 0.1711. Use the Triangle spread rules,
and the fact that s1 + s2 ≤ 1, to get
√
s3 = rl (s1 , s2 ) (ac) = 0.4416 − 0.171 1 (ac) ≈ 0.0280 (ac).

Then apply the Spread law in A1 A2 A3 to get

0.0280/25 ≈ 0.2352/Q1

from which Q1 ≈ 211.5. Then in the right triangle A2 A3 C

Q = s2 Q1 ≈ 0.3897 × 211.5 ≈ 82.4. ¦

24.3. HEIGHT OF A RAISED OBJECT 245

24.3 Height of a raised object

Problem 19 The points A1 , A2 and C are horizontal and in a line. There are two
points D and A3 vertically above the point C as in Figure 24.2. The spreads s1 and s2
in triangle A1 A2 A3 are measured, as is the spread r ≡ s (A2 C, A2 D). The quadrance
Q3 between A1 and A2 is known. What is the vertical quadrance R ≡ Q (A3 , D)?

A3
s3
R
Q2 Q1 R2
D
R1
s2 r
s1 Q3
A1 A2 P C

Figure 24.2: Height of a raised object

Solution. Define the quadrances R1 ≡ Q (C, D), R2 ≡ Q (C, A3 ) and P ≡ Q (A2 , C).
Use the Triple spread formula and the Triangle spread rules in A1 A2 A3 to find s3 .
Then the Spread law in A1 A2 A3 gives
s1 Q3
Q1 = .
s3
In the right triangle A2 A3 C
s1 s2 Q3
R2 = s2 Q1 = (24.1)
s3
and
P = (1 − s2 ) Q1 .
Then in the right triangle A2 DC
rP (1 − s2 ) rQ1
R1 = =
1−r 1−r
so that also
s1 (1 − s2 ) rQ3
R1 = . (24.2)
s3 (1 − r)
Now {R, R1 , R2 } is a quad triple so solve
2
(R − R1 − R2 ) = 4R1 R2

with the Collinear quadrance rules (page 215) to obtain R.

246 24. SURVEYING

24.4 Regiomontanus’ problem

Regiomontanus, whose name was Johann Müller, lived from 1436 to 1476, and
published mathematical and astronomical books. In his most famous work On
Triangles of Every Kind, he mentions the following extremal problem.

Problem 20 (Regiomontanus’ problem) In Figure 24.3, what value of the

quadrance P will maximize the spread s subtended by the window BD? The
positions of the points B, D and C on the vertical line are known and fixed, so the
quadrances Q, Q1 and Q2 can be taken as given.

B
1- s 2
Q
Q
R D 2
1-s 1
s Q1
s1
s2 A P C

Figure 24.3: Regiomontanus’ problem

Solution. Pythagoras’ theorem gives R = P + Q1 , while from the Spread ratio

theorem
Q2
s2 = .
P + Q2
From the Spread law in ABD
s 1 − s2
= .
Q R
Combine these equations to get
QP
s=
(P + Q1 ) (P + Q2 )
Q
= .
Q1 + Q2 + P + (Q1 Q2 /P )
Now choose P so that this expression is maximized, or equivalently so that
Q1 Q2
P+
P
is minimized. With the product of two summands constant, the sum is minimum when
the summands are equal, so that P 2 = Q1 Q2 . Thus P must be the geometric mean of
Q1 and Q2 .
24.5. HEIGHT FROM THREE SPREADS 247

24.5 Height from three spreads

Problem 21 A triangle A1 A2 A3 is horizontal, the point B is directly above A3 , and

D is a third point lying on A1 A2 , as in Figure 24.4. The vertical spreads

r1 ≡ s (A1 A3 , A1 B) r2 ≡ s (A2 A3 , A2 B) r3 ≡ s (DA3 , DB)

are known, as are the quadrances P1 ≡ Q (A1 , D) and P2 ≡ Q (A2 , D). Find the
vertical quadrance H ≡ Q (A3 , B).

A3
r1 r3 Q1
Q 2 P3 r2

A1 P1 D P2 A2

Figure 24.4: Height from three spreads

Solution. Suppose the quadrances of A1 A2 A3 are Q1 , Q2 and Q3 as usual. Let

P3 ≡ Q (A3 , D). From the right triangle A1 A3 B
H
r1 =
H + Q2
so that
(1 − r1 ) H
Q2 = .
r1
Similarly from the right triangles A2 A3 B and DA3 B
Q1 = (1 − r2 ) H/r2 and P3 = (1 − r3 ) H/r3 .
Now in the triangle A1 A2 A3 use Stewart’s theorem (page 136) to get
2 2
P2 (P3 + P1 − Q2 ) = P1 (P3 + P2 − Q1 ) .
Substitute for Q2 , Q1 and P3 , to get for H the quadratic equation

µ µ ¶ ¶2 µ µ ¶ ¶2
1 1 1 1
P2 H − + P1 = P1 H − + P2 .
r3 r1 r3 r2
248 24. SURVEYING

24.6 Vertical and horizontal spreads

Problem 22 The points A1 , A2 and A3 form a horizontal triangle with quadrances

Q1 , Q2 and Q3 , and spreads s1 , s2 and s3 as usual. The point B is directly above the
point A3 . What is the relationship between the vertical spreads r1 ≡ s (A1 A3 , A1 B)
and r2 ≡ s (A2 A3 , A2 B)?

A3
r1 Q s3
2 Q1 r2
s1 s2
A1 A2

Figure 24.5: Vertical and horizontal spreads

Solution. Suppose that H ≡ Q (A3 , B) as in Figure 24.5. From the right triangle
A1 A3 B
H r1
=
Q2 1 − r1
and similarly from the right triangle A2 A3 B

H r2
= .
Q1 1 − r2

Use the Spread law in the triangle A1 A2 A3 and the previous equations to get

s1 Q1 r1 (1 − r2 )
= = .
s2 Q2 (1 − r1 ) r2

This can also be written as either

s1 (1 − r1 ) s2 (1 − r2 )
=
r1 r2
or
s1 s2
− = s1 − s2 .
r1 r2
24.7. SPREADS OVER A RIGHT TRIANGLE 249

24.7 Spreads over a right triangle

Problem 23 (Spreads over a right triangle) Suppose that the points A1 , A2 and
A3 form a horizontal right triangle with right vertex at A3 , and that B is directly
above the point A3 as in Figure 24.6. What is the relationship between the spreads
s ≡ s (BA1 , BA2 ), r1 ≡ s (A1 A3 , A1 B) and r2 ≡ s (A2 A3 , A2 B)?

B
s
H

A3
r1 Q 2 Q1
r2

A1 Q3 A2

Figure 24.6: Spreads over a right triangle

Solution. Let the quadrances of A1 A2 A3 be Q1 , Q2 and Q3 , and let H ≡ Q (A3 , B).

By Pythagoras’ theorem
Q (A1 , A2 ) = Q3 = Q1 + Q2
Q (A1 , B) = Q2 + H
Q (A2 , B) = Q1 + H.
From the Cross law in A1 A2 B
2
((Q1 + H) + (Q2 + H) − (Q1 + Q2 )) = 4 (Q1 + H) (Q2 + H) (1 − s) .

Thus ultimately independent of the triangle A1 A2 A3 ,

µ ¶µ ¶
H H
1−s= = r1 r2 . (24.3)
Q1 + H Q2 + H

Exercise 24.1 (Harder) Suppose A1 A2 A3 is an equilateral triangle, and that B is

directly above the circumcenter C of A1 A2 A3 . Show that if
q ≡ s (BA1 , BA2 ) = s (BA2 , BA3 ) = s (BA1 , BA3 )
and S is the spread between any two of the planes A1 A2 B, A2 A3 B and A1 A3 B, then
(1 − Sq)2 = 4 (1 − S) (1 − q) . ¦
250 24. SURVEYING

24.8 Spherical analogue of Pythagoras’ theorem

From (24.3) follows a remarkable and important formula. Suppose that the points
A1 , A2 and A3 form a horizontal right triangle with right vertex at A3 , and that O is
directly above the point A3 . Define the spreads q1 ≡ s (OA1 , OA3 ), q2 ≡ s (OA2 , OA3 )
and q ≡ s (OA1 , OA2 ) as in Figure 24.7. Then q1 and q2 are complementary to the
spreads r1 and r2 in Figure 24.6.

The use of the small letter q here and in the previous exercise anticipates projective
trigonometry, where the quadrance between two ‘projective points’ is defined to be the
spread between the associated lines through the origin.

q
O
q1 q 2
H
A3

A1 A2

Figure 24.7: Spherical Pythagoras’ theorem

Using (24.3),

q = 1 − r1 r2
= 1 − (1 − q1 ) (1 − q2 ) .

So

q = q1 + q2 − q1 q2 .

This is the spherical or (elliptic) analogue of Pythagoras’ theorem. Its pivotal role in
projective trigonometry will be explained more fully in a subsequent volume. Note
that if q1 and q2 are small then this is approximated by the usual planar form of
Pythagoras’ theorem.
 25

Resection and Hansen’s

problem

The problems of Snellius-Pothenot and Hansen are among the most famous of
surveying problems, and are also of importance in navigation. The Snellius-Pothenot,
or resection, problem has a number of solutions, and the one presented here uses
Euler’s Four point relation. Hansen’s problem is illustrated with a specific example,
and an exercise shows its connection with a somewhat notorious problem of elementary
Euclidean geometry.

25.1 Snellius-Pothenot problem

The problem of resection was originally stated and solved by Snellius (1617) and then
by Pothenot (1692).

A2
Problem 24
Q3 P2
The quadrances Q1 , Q2 and Q3 of A1 A2 A3 are known. r 3 r1 Q 1
The spreads r1 ≡ s (BA2 , BA3 ), r2 ≡ s (BA1 , BA3 ) B r2
and r3 ≡ s (BA1 , BA2 ) are measured. Find P1 P3 A3
P1 ≡ Q (B, A1 ), P2 ≡ Q (B, A2 ) and P3 ≡ Q (B, A3 ). A1 Q2

The problem cannot be solved if B lies on the circumcircle c of A1 A2 A3 , since in that

case the Subtended spread theorem (page 178) shows that any point on c yields the
same values for r1 , r2 and r3 . Here is a procedure to find P1 and P2 , which works
provided B is not on c, using the Four point relation (page 191).

251
252 25. RESECTION AND HANSEN’S PROBLEM

Solution. Take the circumcircle c3 of A1 A2 B and let H, called Collin’s point, be

the intersection of c3 with A3 B which is distinct from B.

Define the quadrances R1 ≡ Q (H, A1 ), R2 ≡ Q (H, A2 ) and R3 ≡ Q (H, A3 ) . By the

Subtended spread theorem, the spreads s (A1 H, A1 A2 ), s (A2 H, A2 A1 ) and
s (HA1 , HA2 ) are respectively r1 , r2 and r3 . Let v1 ≡ s (HA1 , HA3 ) and
v2 ≡ s (HA2 , HA3 ). This is shown in Figure 25.1.

A2
H r2
r3
v2 R2
v1
P2
Q3 r1 Q1
R1
r 2 Br
c3 P1 3
r1

A1 A3
Q2

Figure 25.1: Snellius-Pothenot problem

Use the Spread law in A1 A2 H to get

R1 = r2 Q3 /r3 and R2 = r1 Q3 /r3 . (25.1)

The Four point relation applied to the triangle A1 A2 A3 with the additional point H is

E (Q1 , Q2 , Q3 , R1 , R2 , R3 ) = 0.

By Exercise 17.5, this is the quadratic equation in R3 given by

µ ¶2
(Q1 − Q2 ) (R2 − R1 )
R3 − R1 − R2 + Q3 − Q1 − Q2 +
Q3
A (Q1 , Q2 , Q3 ) A (R1 , R2 , Q3 )
=
4Q23
where A is Archimedes’ function. After substituting for the values of R1 and R2 from
(25.1), this becomes the equation
2
(R3 − C) = D

where
(Q1 + Q2 + Q3 ) (r1 + r2 + r3 ) − 2 (Q1 r1 + Q2 r2 + Q3 r3 )
C=
2r3
and
r1 r2 A (Q1 , Q2 , Q3 )
D= .
r3
25.1. SNELLIUS-POTHENOT PROBLEM 253

For either of the two solutions to this equation, the Cross law in A1 A3 H gives

(R1 + R3 − Q2 )2
v1 = 1 −
4R1 R3

while the Cross law in A2 A3 H gives

(R2 + R3 − Q1 )2
v2 = 1 − .
4R2 R3

Then the Spread laws in A1 BH and A2 BH give the required values

v1 R1 v1 Q3
P1 = =
r2 r3
v2 R2 v2 Q3
P2 = = .
r1 r3

Example 25.1 Suppose that the triangle A1 A2 A3 has points

A1 ≡ [1, 1] A2 ≡ [5, 2] A3 ≡ [3, −1]

with quadrances
Q1 = 13 Q2 = 8 Q3 = 17.
If B is taken to be the point [4, 5] then

r1 = 81/370 r2 = 196/925 r3 = 169/250.

The three values r1 , r2 and r3 will be taken as measurements, and the location of B
otherwise considered unknown. Then from (25.1)

R1 = r2 Q3 /r3 and R2 = r1 Q3 /r3

gives
R1 = 33320/6253 and R2 = 34425/6253.
Now use the Four point relation

E (Q1 , Q2 , Q3 , R1 , R2 , R3 ) = 0

to get the quadratic equation

µ ¶2 µ ¶2
46 216 2 520
R3 − =
6 253 481

with solutions

i) R3 = 13456/6253 or ii) R3 = 78976/6253.

254 25. RESECTION AND HANSEN’S PROBLEM

i) If R3 = 13456/6253 then

2
(R1 + R3 − Q2 ) 169
v1 = 1 − =
4R1 R3 170
2
(R2 + R3 − Q1 ) 169
v2 = 1 − =
4R2 R3 425
so that
v1 R1 v1 Q3
P1 = = = 25
r2 r3
v2 R2 v2 Q3
P2 = = = 10.
r1 r3
ii) If R3 = 78976/6253 then

(R1 + R3 − Q2 )2 33 124
v1 = 1 − =
4R1 R3 52 445
(R2 + R3 − Q1 )2 474 721
v2 = 1 − =
4R2 R3 524 450
so that
v1 R1 v1 Q3 9 800
P1 = = =
r2 r3 617
v2 R2 v2 Q3 14 045
P2 = = = .
r1 r3 617
The first of these cases correctly yields the quadrances to the initial point B ≡ [4, 5]. ¦

The two solutions obtained in the previous Example correspond to the two points B
and B 0 that make the same spreads r1 , r2 and r3 with the reference triangle A1 A2 A3 .
The relation between these two points may be described by the following known result
(see [Wells, page 258]).

Let B be a point not on the lines A1 A2 , A2 A3 and A1 A3 , and let B3 , B1 and B2 be the
reflections of B in the lines A1 A2 , A2 A3 and A3 A1 respectively. Let c1 , c2 and c3 be
the respective circumcircles of the triangles A2 A3 B1 , A1 A3 B2 and A1 A2 B3 . Then c1 , c2
and c3 intersect in a unique point B 0 .

The map that sends B to B 0 in the above result is not a bijection. If B is any point on
the circumcircle of A1 A2 A3 , then it turns out that B 0 is always the orthocenter of
A1 A2 A3 .

Exercise 25.1 Use the Triangle spread rules to identify the correct choice of R3 in the
previous Example. ¦

Exercise 25.2 Find another solution to the resection problem, not using the Four
point relation. ¦
25.2. HANSEN’S PROBLEM 255

25.2 Hansen’s problem

Problem 25 (Hansen’s problem) Two known points A and B with known

quadrance Q ≡ Q (A, B) are sighted from two variable points C and D. The four
spreads s (DA, DB), s (DB, DC), s (CA, CB) and s (CA, CD) are measured from the
points C and D. The positions of C and D are to be determined, in the sense that
the quadrances Q (A, C), Q (B, C), Q (A, D) and Q (B, D) are to be found.

This problem was solved by Hansen (1795-1884), a German astronomer, but according
to [Dorrie] also by others before him. The treatment presented here will be illustrated
by a particular example. The general case follows the same lines. Assume the
quadrance between the fixed points A and B is Q (A, B) ≡ 26.

Suppose that the following spreads are known

³←−−→´ ³←−−→´
s ADB = 361/425 (ac) s BDC = 169/250 (ac)
³←−−→´ ³←−−→´
s BCA = 441/697 (ac) s ACD = 121/410 (ac).

This information is shown to scale in Figure 25.2, along with the intersection E of AC
and BD.

5 B
y C
441
4 697
121

3 E 410

2 169
250
361
425
D
1 x
A
-1 0 1 2 3 4 5 6

-1

Figure 25.2: Hansen’s problem I

Solution. Apply the Triangle spread rules to the three sectors with base D. Note
that this new application inverts the type. Since

361/425 + 169/250 = 6483/4250 ≥ 1

256 25. RESECTION AND HANSEN’S PROBLEM

and both spreads are acute,

³←−−→´ µ ¶
361 169
s ADC = rb , (ob)
425 250
361 169 361 169
= + −2× ×
425r 250 425 250
361 169 64 81 121
+2 × × × (ob) = (ob).
425 250 425 250 170
Similarly apply the Triangle spread rules to the sectors with base C. Since

441/697 + 121/410 = 6467/6970 ≤ 1

and both spreads are acute,

³←−−→´ µ ¶
441 121
s BCD = rb , (ac)
697 410
441 121 441 121
= + −2× ×
697r 410 697 410
441 121 256 289 169
+2 × × × (ac) = (ac).
697 410 697 410 170
Now apply the Triangle spread rules to CDE. In this case, no inversion of type takes
place. Since
169/250 + 121/410 = 4977/5125 ≤ 1
and both spreads are acute,
³←−−→´ µ ¶
169 121
s DEC = rb , (ob)
250 410
169 121 169 121
= + −2× ×
250r 410 250 410
169 121 81 289 1024
+2 × × × (ob) = (ob).
250 410 250 410 1025
³←−−→´ ³←−−→´
The spreads s DAC and s DBC may now be determined using the same
procedure, but an alternative is to use the Two spread triples theorem (page 98) and
the function
(a − b)2 − (c − d)2
P (a, b, c, d) ≡ .
2 (a + b − c − d − 2ab + 2cd)
Then ³←−−→´ µ ¶
1024 361 121 121 121
s DAC = P , , , =
1025 425 170 410 697
and ³←−−→´ µ ¶
1024 441 169 169 169
s DBC = P , , , = .
1025 697 250 170 425
This information is now summarized in Figure 25.3, with the unknown spreads x and y
to be determined.
25.2. HANSEN’S PROBLEM 257

5 B 169
y 425 C 169
170
441
4 697
121
410
3 1024
E
1025

2 169
250
361
425
D
1 x 121 121
697 170
A
-1 0 1 2 3 4 5 6

-1

Figure 25.3: Hansen’s problem II

The Alternate spreads theorem, extended to a quadrilateral as in Exercise 11.9, gives

the formula
169 121 361 441 169 121
× × ×x= × × ×y
425 410 425 697 250 697
so that
x 11 025
a≡ = .
y 14 801
Now in the notation of the Spread from ratio theorem (page 91), with s = 1024/1025,
2
a (1 − s) = 441/606841 = (21/779)

so set
r ≡ 21/779.
Then
y = s/ (a + 1 ± 2r)
and
x = ya.
Substitute to get the possibilities
∙ ¸
441 361
[x, y] = ,
1066 650
or ∙ ¸
112 896 92 416
[x, y] = , .
256 537 156 425
The first of these corresponds to the picture above. The Spread law in ABD gives

361/425 361/650
= .
26 Q (A, D)
258 25. RESECTION AND HANSEN’S PROBLEM

Thus Q (A, D) = 17. The Spread law in ABC gives

441/697 441/1066
=
26 Q (B, C)
so that Q (B, C) = 17. The Spread law in ADC gives
121/410 121/170
=
17 Q (A, C)
so that Q (A, C) = 41. The Spread law in BCD gives
169/250 169/170
=
17 Q (B, D)
so that Q (B, D) = 25. This establishes the four required quantities.

The example was chosen with A ≡ [0, 0], B ≡ [1, 5], C ≡ [5, 4] and D ≡ [4, 1], and the
validity of each of these computations may thereby be checked.

Exercise 25.3 (Rational version of a notorious problem) The triangle A1 A2 A3

represented to scale in Figure 25.4 is isosceles with Q (A1 , A3 ) = Q (A2 , A3 ) ≡ 58 and
Q (A1 , A2 ) ≡ 36. Also known are the spreads
³←−−−−→´ ³←−−−−→´
s A1 A2 B2 ≡ 49/170 (ac) s A2 A1 B1 ≡ 64/185 (ac)

A3

B2 B1
x y

64 49
185 170
A1 A2

Figure 25.4: A notorious problem

Use the analysis of this section to determine the spreads

³←−−−−→´ ³←−−−−→´
x = s A2 B2 B1 y = s A1 B1 B2 .

[The answer is
9834 496
x = (ac)
25 778 545
28 654 609
y = (ac).] ¦
112 212 490
 26

Platonic solids

Rational trigonometry can be used to understand aspects of the five Platonic solids:
the (regular) tetrahedron, cube, octahedron, icosahedron and dodecahedron. A more
complete investigation involves projective trigonometry, the rational analogue of
spherical trigonometry which will be explained in a future volume.

This chapter computes the face spread S of each Platonic solid, namely the spread
between adjacent faces, as well as some related results. Curiously, the face spreads
turn out to be rational numbers in all five cases.

Figure 26.1: The five Platonic solids

259
260 26. PLATONIC SOLIDS

26.1 Tetrahedron

The tetrahedron has four points, six sides and four faces, each an equilateral triangle.
To determine the face spread, suppose that each side of a tetrahedron ABCD has
quadrance Q, with M the midpoint of the side AB as in Figure 26.2. Then by
Pythagoras’ theorem Q (C, M ) = Q (D, M ) = 3Q/4.

The isosceles triangle CM D therefore has quadrances 3Q/4, 3Q/4 and Q, so the
Isosceles triangle theorem (page 122) shows that

µ ¶
Q Q 8
s ≡ s (M C, M D) = 1− = .
3Q/4 3Q 9

D
Q
Q C

3Q/ 4
A s 3Q/ 4

Q/ 4 Q
M
Q/ 4 B

Figure 26.2: Tetrahedron

This is equal to the face spread S ≡ S (ABC, ABD), as can be seen by applying the
Perpendicular spreads theorem (page 79) to the plane DCM . The ‘three-dimensional
sector’ formed by these faces towards the interior of the tetrahedron is in a natural
sense acute.

Exercise 26.1 Show that the quadrance from one point of the tetrahedron to the
centroid of the opposite face is 2Q/3. Show that the quadrance from one point of the
tetrahedron to the center P of the tetrahedron is 3Q/8. ¦

Exercise 26.2 Show that the face spread S = 8/9 is the same as the spread
s (P A, P B), where P is the center of the tetrahedron. ¦

Exercise 26.3 By comparing the spread s = 8/9 with the appropriate zero of S5 (s),
show that it is possible to arrange five solid tetrahedrons sharing a common side. Show
that it is not possible to arrange six solid tetrahedrons sharing a common side. ¦
26.2. CUBE 261

26.2 Cube

The cube has eight points, twelve sides and six faces, each a square. Clearly the spread
made by adjacent faces is S = 1.

Let’s consider the problem of determining the possible spreads made by two lines from
the center of a cube to two points of the cube.

Suppose that a cube has each side of quadrance Q, and center P with points labelled
as in Figure 26.3.

G C

H D
P
F B
K L
E A
Q

Figure 26.3: Cube

Since Q (P, K), Q (K, L) and Q (A, L) are all equal to Q/4, use Pythagoras’ theorem to
get

Q (P, L) = Q/4 + Q/4 = Q/2

Q (P, A) = Q (P, L) + Q (A, L) = 3Q/4
Q (A, C) = Q (A, B) + Q (B, C) = 2Q.

The Isosceles triangle theorem applied to AP C, with quadrances 3Q/4, 3Q/4 and 2Q,
shows that
2Q 1
s (AP, AC) = 1 − =
4 (3Q/4) 3
µ ¶
2Q 2Q
s (P A, P C) = 1−
3Q/4 4 (3Q/4)
8 1
= ×
3 3
8
= .
9

Note that since ACF H is a tetrahedron, this latter formula recovers the result of
Exercise 26.2.
262 26. PLATONIC SOLIDS

26.3 Octahedron

The octahedron has six points, twelve sides and eight faces, each an equilateral
triangle. To determine the face spread S, suppose that the common quadrance of a
side is Q, and let M be the midpoint of the side BE as in Figure 26.4, so that CM
and AM are both perpendicular to BE, and Q (A, M ) = Q (M, C) = 3Q/4.

Then the isosceles triangle ACM has quadrances 3Q/4, 3Q/4 and 2Q, so using the
Isosceles triangle theorem

µ ¶
2Q 2Q 8
S ≡ s (M A, M C) = 1− = .
3Q/4 3Q 9

A M D
C
B

Figure 26.4: Octahedron

While the spread between adjacent faces of the tetrahedron and octahedron thus agree,
the former is acute, while the latter is obtuse.

To see the equality directly, observe that the six midpoints of the sides of a
tetrahedron form an octahedron. This octahedron can also be obtained by slicing off at
each vertex of the tetrahedron a smaller corner tetrahedron as in Figure 26.5. The
corner tetrahedron so sliced off shares adjacent faces with the central octahedron, so
the face spreads are the same.

Figure 26.5: Slicing corners off a tetrahedron

26.4. ICOSAHEDRON 263

26.4 Icosahedron

The icosahedron has twelve points, thirty sides and twenty faces, each an equilateral
triangle. To determine the face spread S, suppose that V is a point of the icosahedron
with adjacent points A, B, C, D and E forming a regular pentagon as in Figure 26.6.
Suppose that the common quadrance of a side is Q, and that M is the midpoint of the
side V E, so that DM and M A are perpendicular to V E, and that
Q (D, M ) = Q (M, A) = 3Q/4.

A
C V

M
E
G D

Figure 26.6: Icosahedron

Recall from Exercise 14.3 that Q (A, D) = βQ/α where

¡ √ ¢ ¡ √ ¢
α ≡ 5 − 5 /8 ≈ 0.345 491 . . . and β ≡ 5 + 5 /8 ≈ 0.904 508 . . . .
Apply the Isosceles triangle theorem to ADM with sides 3Q/4, 3Q/4 and βQ/α and
some pleasant simplification to get

µ ¶
βQ/α βQ/α
S ≡ s (M D, M A) = 1−
3Q/4 3Q
µ ¶
4β β 4
= 1− = .
3α 3α 9

Exercise 26.4 Using the same diagram, show that

√
10 − 2 5
s (M D, M G) =
15
and √
10 + 2 5
s (M A, M G) = .
15
Hence deduce that à √ !
5+ 5
Q (A, G) = Q. ¦
2
264 26. PLATONIC SOLIDS

26.5 Dodecahedron

The dodecahedron has twenty points, thirty sides and twelve faces, each a regular
pentagon. To determine the face spread S, suppose that each side of the dodecahedron
has quadrance Q. Three sides meet at every point.

If the point V has adjacent points A, B and C then ABC is an equilateral triangle
with quadrances βQ/α, since this is the quadrance of a diagonal side of a regular
pentagon of quadrance Q, as in Exercise 14.3. Furthermore the spread r ≡ s (V A, V B)
is equal to β, since this is the spread between adjacent lines of a regular pentagon.
This is shown in Figure 26.7.

A
Q
rQ B
P
bQ a F V
P Q
C

Figure 26.7: Dodecahedron

Now suppose that F is the foot of the altitude from A to BV , and so by symmetry
also the foot of the altitude from C to BV. Then using the right triangle AF V , the
quadrance P ≡ Q (A, F ) = Q (C, F ) is
P = rQ (A, V ) = βQ.
In the isosceles triangle AF C the quadrances are then βQ, βQ and βQ/α. Use the
Isosceles triangle theorem and some pleasant simplification to obtain the face spread

µ ¶
βQ/α βQ/α
S ≡ s (F A, F C) = 1−
βQ 4βQ
4α − 1 4
= = .
4α2 5

To summarize: the face spreads of the regular tetrahedron, cube, octahedron,

icosahedron and dodecahedron are respectively

8/9 1 8/9 4/9 4/5 .

 27

Rational spherical coordinates

One of the important traditional uses of angles and the transcendental trigonometric
functions cos θ and sin θ is to establish polar coordinates in the plane, and spherical
and cylindrical coordinates in three-dimensional space. This simplifies problems with
rotational symmetry in advanced calculus, mechanics and engineering.

This chapter shows how to employ rational analogues to accomplish the same tasks,
with examples chosen from some famous problems in the subject. The rational
approach employs conventions that generalize well to higher dimensions.

27.1 Polar spread and quadrance

For a point A ≡ [x, y] in Cartesian coordinates, introduce the polar spread s and the
quadrance Q by

¡ ¢
s ≡ x2 / x2 + y 2
Q ≡ x2 + y 2 .

Then [s, Q] are the rational polar coordinates of the point A ≡ [x, y]. The spread s
is defined between OA and the y axis. This convention

• corresponds to the usual practice in surveying and navigation

• integrates more smoothly with higher dimensional generalizations
• is natural for human beings, for whom up is more interesting than right.

265
266 27. RATIONAL SPHERICAL COORDINATES

P =[x,y]

Q
s

O x

Figure 27.1: Rational polar coordinates

The rational polar coordinates s and Q determine x and y up to sign, so determine A

uniquely in the first quadrant. This quadrant is better described by the respective
signs of x and y, so call it also the (++)-quadrant.

To specify a general point A, the rational coordinates s and Q need to be augmented

with two additional bits of information–the signs of x and y respectively. Now

x2 = sQ (27.1)
y2 = (1 − s) Q.

Take differentials of these two relations to obtain

2x dx = Q ds + s dQ
2y dy = −Q ds + (1 − s) dQ.

Thus in the (++)-quadrant

¯ ¯
¯Q s ¯¯
4xy dx dy ¯
= ¯ ds dQ
−Q 1 − s¯
¯ ¯
¯Q s¯
= ¯¯ ¯ ds dQ = Q ds dQ. (27.2)
0 1¯

For future reference, note that the determinant is evaluated by adding the first row to
the second to get a diagonal matrix. In the (++)-quadrant, use (27.1) to obtain
p
xy = s (1 − s) Q

so the element of area is

1
dx dy = p ds dQ. (27.3)
4 s (1 − s)
27.1. POLAR SPREAD AND QUADRANCE 267

Example 27.1 The area a of the central circle of quadrance K is, by symmetry,
Z KZ 1 Z 1
1 1
a=4 p ds dQ = K p ds.
0 0 4 s (1 − s) 0 s (1 − s)

This is not an integral which can be evaluated explicitly using basic calculus,
motivating the definition of the number

Z 1
1
π= p ds. (27.4)
0 s (1 − s)

So the area of the central circle of quadrance K is πK. ¦

Exercise 27.1 Use the substitutions s ≡ r2 and s ≡ 1/t to show that

Z 1 Z ∞
dr dt
π=2 √ = √ .
1−r 2 t t−1
0 1
¡ ¢
Then use the substitutions r ≡ 2u/ 1 + u2 and v ≡ 1/u to show that
Z 1 Z ∞
du dv
π=4 =4 . ¦
0 1 + u2 1 1 + v2

Example 27.2 A lemniscate of Bernoulli has Cartesian equation

¡ 2 ¢2
x + y 2 = x2 − y 2 (27.5)

and polar equation

r2 = cos 2θ.

0.25

0.125

0
-1 -0.5 0 0.5 1

x
-0.125

-0.25

Figure 27.2: Lemniscate of Bernoulli

Replace x2 and y 2 in (27.5) using (27.1) to get

Q2 = sQ − (1 − s) Q
= (2s − 1) Q.
268 27. RATIONAL SPHERICAL COORDINATES

So either of
Q = 2s − 1 or s = (Q + 1) /2

is a rational polar equation of the lemniscate. Rational polar equations of some

other classical curves are described in Appendix 1. For the lemniscate the polar spread
varies in the range 1/2 ≤ s ≤ 1, so the area is
Z 1 Z 2s−1
1
a = 4 p dQ ds
1/2 0 4 s (1 − s)
Z 1 Z 1/4
2s − 1 1
= p ds = √ du = 1. ¦
1/2 s (1 − s) 0 u

R∞ 2
Example 27.3 The integral I = 0 e−x dx is difficult to evaluate using only the
calculus of one variable. Using rational polar coordinates, the idea is as follows, where
the integral is over the (++)-quadrant.
Z ∞ Z ∞
2 −x2 2
I = e dx e−y dy
Z0 ∞ Z ∞
0
−(x2 +y 2 )
= e dx dy
0 0
Z ∞Z 1
e−Q
= p ds dQ
0 0 4 s (1 − s)
Z ∞ Z 1
1
= e−Q dQ p ds
0 0 4 s (1 − s)
£ ¤∞
= −e−Q Q=0 × π/4
= π/4
√
so that I = π/2. ¦

The rotationally invariant measure dµ on the circle of quadrance Q = r2 is, since

dQ = 2r dr, determined by the equation

dµ dQ
dx dy = dµ dr = .
2r
Compare this with (27.3) to see that

r
dµ = p ds.
2 s (1 − s)

It follows that the quarter of the central circle of radius r in the (++)-quadrant has
measure πr/2, and the full circle has measure 2πr.
27.2. EVALUATING π2 /16 269

27.2 Evaluating π 2 /16

The unit quarter circle has area π/4, so a squared area of

π2 /16 ≈ 0.616 850 275 068 . . . .
To evaluate this constant, we follow ideas of Archimedes. Approximate a quarter circle
successively by first one, then two, then four isosceles triangles, and so on, each time
subdividing each triangle into two by a vertex bisector, as shown in Figure 27.3.

Q1 Q2
Q2
1 Q0 1 1 Q2
1 Q1
Q2
1 1 1

Figure 27.3: Approximations to a quarter circle

By the Quadrea spread theorem (page 82), the quadrea of an isosceles triangle with
Q1 = Q2 ≡ 1 and spread s3 ≡ s is A = 4s. After n divisions there are 2n congruent
isosceles triangles, each with spread sn at the common point, and hence each with
quadrea 4sn . This gives for the resulting (2n + 2)-gon a total quadrea of
2
An = (2n ) × 4sn , and so a squared area of a2n = An /16 = 22n−2 sn . Now since
√
1 − 1 − sn
sn+1 =
2
it follows that
¡ √ ¢
a2n+1 = 22n sn+1 = 22n−1 1 − 1 − sn
³ p ´
= 22n−1 1 − 2−n+1 22n−2 − a2n
p
= 22n−1 − 2n 22n−2 − a2n .
Surprisingly, this recurrence relation yields a pleasant form for the general term a2n , as
indicated by the following computations.
a20 = 2−2 = 0.25
p
a21 = 2−1 − 20 2−2 − 2−2 = 2−1 = 0.5
p √
a22 = 21 − 21 20 − 2−1 = 2 − 2 ≈ 0.585 786
q q
√ √
a23 = 23 − 22 22 − 2 + 2 = 8 − 4 2 + 2 ≈ 0.608 964
s µ q ¶ r q
√ √
a24 5 3 4
= 2 − 2 2 − 8 − 4 2 + 2 = 32 − 16 2 + 2 + 2 ≈ 0.614 871

Exercise 27.2 Show that this pattern continues, giving a closed expression for a2n . ¦
270 27. RATIONAL SPHERICAL COORDINATES

27.3 Beta function

Following Euler, for decimal numbers p > 0 and q > 0 define the Beta function, or
Beta integral, Z 1
B (p, q) ≡ sp−1 (1 − s)q−1 ds.
0
There is a standard expression for the Beta function in terms of the Gamma
function defined for t > 0 by
Z ∞ Z ∞
2
−u t−1
Γ (t) ≡ e u du = 2 e−x x2t−1 dx.
0 0
Integration by parts and direct calculation shows that
Γ (t + 1) = t Γ (t)
Γ (1) = 1.
This implies that
Γ (n) = (n − 1)!
for any positive integer n ≥ 1.

Use rational polar coordinates to rewrite the following integral over the (++)-quadrant
Z ∞ Z ∞
2 2
Γ (p) Γ (q) = 4 e−x x2p−1 dx e−y y 2q−1 dy
0 0
Z ∞ Z ∞
−(x2 +y2 ) 2(p−1) 2(q−1)
= e x y 4xy dx dy
0 0
Z ∞ Z 1
= e−Q (sQ)p−1 ((1 − s) Q)q−1 Q ds dQ
0 0
Z ∞ Z 1
q−1
= e−Q Qp+q−1 dQ sp−1 (1 − s) ds
0 0
= Γ (p + q) B (p, q)
where (27.2) was used to go from the second to the third line. Thus

Γ (p) Γ (q)
B (p, q) = . (27.6)
Γ (p + q)

Values of the Beta function are particularly useful in calculations involving rational
polar or spherical coordinates. Note that in particular
B (1/2, 1/2) = π = (Γ (1/2))2
so that, recovering the computation of Example 27.3,
Z ∞
2 √
Γ (1/2) = 2 e−x dx = π.
0
27.4. RATIONAL SPHERICAL COORDINATES 271

27.4 Rational spherical coordinates

Represent a point in three-dimensional space by A ≡ [x, y, z], and define

the rational spherical coordinates [s, q, R] of A by
z
¡ ¢
s ≡ x2 / x2 + y 2 P
¡ ¢ ¡ ¢
q ≡ x2 + y 2 / x2 + y 2 + z 2
q
R ≡ x2 + y 2 + z 2 . R
O
s
Geometrically Q y
if A12 = [x, y, 0] is the perpendicular projection of x
P12
A onto the x − y plane, then s is the polar spread between
OA12 and the y axis, while the second polar spread
q is the spread between OA and the z axis. Then R is the
three-dimensional quadrance, and Q ≡ x2 + y 2 = qR.

Then
x2 = sqR y 2 = (1 − s) qR z 2 = (1 − q) R (27.7)
so that x, y and z are determined, up to sign, by [s, q, R]. Take differentials to obtain
2x dx = qR ds + sR dq + sq dR
2y dy = −qR ds + (1 − s) R dq + (1 − s) q dR
2z dz = 0 ds − R dq + (1 − q) dR.
Thus in the (+ + +)-octant, where the signs of x, y and z are all positive,
¯ ¯
¯ qR sR sq ¯¯
¯
8xyz dx dy dz = ¯¯−qR (1 − s) R (1 − s) q ¯¯ ds dq dR
¯ 0 −R 1−q ¯
¯ ¯
¯qR sR sq ¯
¯ ¯
= ¯¯ 0 R q ¯¯ ds dq dR = qR2 ds dq dR
¯0 0 1¯
where the determinant is evaluated by adding the first row to the second, and then the
second row to the third, to obtain a diagonal matrix.

In the (+ + +)-octant, combine the equations of (27.7) to obtain

p
xyz = R3/2 q s (1 − s) (1 − q)
so the element of volume is
√
R
dx dy dz = p ds dq dR. (27.8)
8 s (1 − s) (1 − q)
272 27. RATIONAL SPHERICAL COORDINATES

Example 27.4 The volume v of the central sphere of quadrance K ≡ k 2 (k ≥ 0) is

eight times the volume in the (+ + +)-octant. It is thus
Z KZ 1Z 1 √
R
v = 8 p ds dq dR
0 0 0 8 s (1 − s) (1 − q)
Z 1 Z 1 Z K√
ds dq
= p √ R dR
0 s (1 − s) 0 1−q 0
" 3 #K
h p i1 2R 2 4πK 2
3
4πk3
= π −2 1 − q = = .
q=0 3 3 3
R=0

y
x

Figure 27.4: Volume of a sphere ¦

Example 27.5 An ‘ice cream cone’ lies above the cone z 2 = x2 + y 2 , and inside the
projective sphere x2 + y 2 + z 2 = z centered at [0, 0, 1/2] with quadrance 1/4. Write the
cone as q = 1/2 and the sphere as R = 1 − q, so that the volume v is
Z 1 Z 1/2 Z 1−q √
R
v = 4 p dR dq ds
0 0 0 8 s (1 − s) (1 − q)
Z ∙ 3/2 ¸1−q
π 1/2 1 2R
= √ dq
2 0 1−q 3 R=0
Z ∙ ¸1/2
π 1/2 π q2 π
= (1 − q) dq = q− = .
3 0 3 2 q=0 8

y
x

Figure 27.5: Volume of an ice cream cone ¦

27.4. RATIONAL SPHERICAL COORDINATES 273

Example 27.6 To find the volume v of the spherical cap inside the sphere
x2 + y 2 + z 2 = K ≡ k2 (k ≥ 0) and lying above the plane z = d ≥ 0, where d ≤ k, use
rational cylindrical coordinates [s, Q, z]
Z 1 Z K−d2 Z √K−Q
1
v = 4 p dz dQ ds
0 0 d 4 s (1 − s)
Z K−d2 ³p ´ h iK−d2
= π K − Q − d dQ = π −2 (K − Q)3/2 /3 − Qd
0 Q=0
π¡ 3 ¢ π 2
= d − 3dk 2 + 2k 3 = (k − d) (2k + d) .
3 3

z
k
z=d

k k
y
x

Figure 27.6: Volume of spherical cap ¦

Example 27.7 The volume of the spherical ring remaining when a cylinder with axis
the z-axis is removed from the central sphere of quadrance K ≡ k 2 (k ≥ 0), leaving a
solid bounded by the planes z = d and z = −d, where d ≤ k, is
Z 1 Z K Z √K−Q Z K p
1
v = 8 p dz dQ ds = 2π K − Q dQ
0 K−d2 0 4 s (1 − s) K−d2
h iK 4π 3
= 2π −2 (K − Q)3/2 /3 = d .
Q=K−d2 3
Curiously, this is independent of the quadrance K of the sphere.

z =d

y
x k k
z = -d

Figure 27.7: Volume of spherical ring ¦

274 27. RATIONAL SPHERICAL COORDINATES

Example 27.8 To find the volume v above the paraboloid z = x2 + y 2 and below the
plane z = r ≥ 0
Z 1Z rZ r Z r
1 πr2
v=4 p dz dR ds = π (r − R) dR = .
0 0 R 4 s (1 − s) 0 2

As discovered
√ by Archimedes, this is one half of the volume of the cylinder of height r
and radius r. ¦

Example 27.9 The moment Mxy of the upper hemisphere of the unit sphere of
density 1 and mass M ≡ 2π/3 with respect to the xy−plane is
Z 1Z 1Z 1 √
z R
Mxy = 4 p ds dq dR
0 0 0 8 s (1 − s) (1 − q)
p
where z = (1 − q) R. Thus
Z 1
π π
Mxy = ×1× R dR =
2 0 4
and the centroid has z coordinate z ≡ Mxy /M = 3/8, so is [0, 0, 3/8].

z
1

3/8

1 1
y
x

Figure 27.8: Center of mass of upper hemisphere ¦

Example 27.10 The moment of inertia of the solid unit ball around the z axis is
Z 1Z 1Z 1 √
qR R
Iz = 8 p ds dq dR
0 0 0 8 s (1 − s) (1 − q)
Z 1 Z 1
q
= π R3/2 dR √ dq
0 0 1 −q
µ ¶
2 1 2 4 8π
= π × × B 2, =π× × =
5 2 5 3 15

since from (27.6)

µ ¶ ¡ ¢ ¡ ¢ ¡1¢
1 Γ (2) Γ 12 Γ 12 Γ 2 4
B 2, = ¡ ¢ = ¡ ¢= ¡1¢ = . ¦
2 Γ 52 3
2 × Γ 32 3
2 × 1
2 ×Γ 2
3
27.4. RATIONAL SPHERICAL COORDINATES 275

Example 27.11 The volume v of the hyperbolic cap shown in Figure 27.9, above the
top sheet of the hyperboloid z 2 − x2 − y 2 = K ≡ k 2 (k ≥ 0) and below the plane
z = d ≥ 0, where d ≥ k, is
Z 1 Z d2 −K Z d Z d2 −K ³ p ´
1
v = 4 √
p dz dQ ds = π d − Q + K dQ
0 0 Q+K 4 s (1 − s) 0
h id2 −K π
= π Qd − 2 (Q + K)3/2 /3 = (k − d)2 (2k + d) .
Q=0 3
This is the same formula as the volume of a spherical cap in Example 27.6!

z =d

y
x

Figure 27.9: Volume of hyperbolic cap ¦

Example 27.12 The volume of the hyperbolic ring shown in Figure 27.10 inside a
cylinder with axis the z-axis and outside the hyperboloid of one sheet
x2 + y 2 − z 2 = K ≡ k2 (k ≥ 0) bounded by the planes z = d and z = −d is
Z 1 Z d2 +K Z √
Q−K Z d2 +K p
1
v = 8 p dz dQ ds = 2π Q − K dQ
0 K 0 4 s (1 − s) K
h id2 +K 4π 3
3/2
= 2π 2 (Q − K) /3 = d .
Q=K 3
Curiously, this is independent of K, and is the same as the volume of the spherical ring
in Example 27.7!

z =d

k k y
x
z = -d

Figure 27.10: Volume of hyperbolic ring ¦

276 27. RATIONAL SPHERICAL COORDINATES

27.5 Surface measure on a sphere

For a fixed value K of R, the polar spreads s and q parametrize that part of the
surface of the sphere of quadrance K contained in the (+ + +)-octant. To describe the
full sphere these two spreads must be augmented by three additional bits of
information, namely the signs of x, y and z.

The rotationally invariant surface measure dν on the sphere R ≡ r2 is, since

dR = 2r dr, determined by

dx dy dz = dν dr = dν dR /2r.

Compare this with (27.8) to get

√
R 1
p ds dq dR = √ dν dR.
8 s (1 − s) (1 − q) 2 R

Thus

R ds dq
dν = p .
4 s (1 − s) (1 − q)

Example 27.13 The total surface area a of the sphere of quadrance K ≡ k2 is

Z 1Z 1
K
a = 8 p ds dq
0 0 4 s (1 − s) (1 − q)
Z 1 Z 1
ds dq
= 2K p √
0 s (1 − s) 0 1−q
= 2K × π × 2 = 4πK = 4πk 2 . ¦

Example 27.14 The surface area a of the spherical cap of the sphere
x2 + y 2 + z 2 = K ≡ k 2 (k ≥ 0) lying above the plane z = d ≥ 0, where d ≤ k, as shown
in Figure 27.6, is
Z (K−d2 )/K Z 1
K
a = 4 p ds dq
0 0 4 s (1 − s) (1 − q)
h i(K−d2 )/K
= πK −2 (1 − q)1/2
q=0
µ ¶
d
= 2πk2 1 − .
k
The linear dependence of this expression on d is one of the most remarkable properties
of a sphere, and is responsible for the fact that an egg slicer subdivides a sphere into
strips of constant surface area. This fact is also important for harmonic analysis on a
sphere, and for the representation theory of the rotation group. ¦
27.6. FOUR DIMENSIONAL RATIONAL SPHERICAL COORDINATES 277

27.6 Four dimensional rational spherical coordinates

For a point A ≡ [x, y, z, w] in four dimensional space define

¡ ¢
s ≡ x2 / x2 + y 2
¡ ¢ ¡ ¢
q ≡ x2 + y 2 / x2 + y 2 + z 2
¡ ¢ ¡ ¢
r ≡ x2 + y 2 + z 2 / x2 + y 2 + z 2 + w2
T ≡ x2 + y 2 + z 2 + w2 .

Then T is the four-dimensional quadrance, and r is the third polar spread between
OA and the new (fourth) w−axis. Then
x2 = sqrT
y2 = (1 − s) qrT
z2 = (1 − q) rT
w2 = (1 − r) T. (27.9)
Take differentials and follow the established pattern to get
¯ ¯
¯ qrT srT sqT sqr ¯¯
¯
¯−qrT (1 − s) rT (1 − s) qT (1 − s) qr¯
16 xyzw dx dy dz dw = ¯¯ ¯ ds dq dr dT
¯ 0 −rT (1 − q) T (1 − q) r ¯¯
¯ 0 0 −T 1−r ¯
¯ ¯
¯qrT srT sqT sqr ¯
¯ ¯
¯ 0 rT qT qr ¯¯
= ¯¯ ds dq dr dT
¯ 0 0 T r ¯¯
¯ 0 0 0 1 ¯
= qr2 T 3 ds dq dr dT.
In the (+ + ++)-octant, (27.9) yields
p
xyzw = s1/2 q r3/2 T 2 (1 − s) (1 − q) (1 − r)
so the element of content (four dimensional version of volume) is

√
rT
dx dy dz dw = p ds dq dr dT.
16 s (1 − s) (1 − q) (1 − r)

Example 27.15 The central sphere of quadrance K ≡ k 2 has content

Z KZ 1Z 1Z 1 √
rT
c = 16 p ds dq dr dT
0 0 0 0 16 s (1 − s) (1 − q) (1 − r)
Z 1 Z 1 Z 1 √ Z K
ds dq r dq
= p √ √ T dT
0 s (1 − s) 0 1−q 0 1−r 0
µ ¶
3 1 K2
= π×2×B , × .
2 2 2
278 27. RATIONAL SPHERICAL COORDINATES

But µ ¶ ¡3¢ ¡1¢ ¡1¢ ¡1¢

1
3 1 Γ 2 Γ 2 2Γ 2 Γ 2 π
B , = = =
2 2 Γ (2) 1 2
so that
π2 K 2 π2 k4
c= = . ¦
2 2

If dν denotes spherical surface measure on the unit 3-sphere determined by

dx dy dz dw = dν dT /2

then (since T = 1) √
r
dν = p ds dq dr.
8 s (1 − s) (1 − q) (1 − r)

Exercise 27.3 Use this to show that the surface volume of the unit 3-sphere is 2π 2 . ¦

It should now be clear how to extend rational spherical coordinates to higher

dimensions. In n-dimensional space, rational spherical coordinates involve (n − 1)
polar spreads, and one quadrance. The basic relations are algebraic, and so do not
require an understanding and visualization of projections.

27.7 Conclusion

Congratulations on having made it this far–hopefully without too much cheating!

This book is only a beginning, and much remains to be done. Hundreds of classical
results of Euclidean geometry may be generalized to the universal setting. A coherent
and precise framework for three-dimensional geometry should be created. The number
theoretical and combinatorial implications of metrical geometry over finite fields
requires investigation, as do the spread polynomials along with other related special
functions. Researchers should ponder the opportunities in regarding algebraic
geometry as an essentially metrical theory. Many additional applications should be
developed and tested, both in applied and pure mathematics. Physicists might enjoy
speculating about the implications for their subject.

Rational analogues of spherical and hyperbolic geometries will be described in a future

book, along with the remarkable synthesis of Euclidean and non-Euclidean geometries
called chromogeometry.

But perhaps the most exciting possibility of all is to re-evaluate the mathematics
taught (and not taught) in schools and colleges, and to think about ways of presenting
to young people this simpler and more logical approach to trigonometry and geometry.
 Appendix A

Rational polar equations of

curves

Recall that the relationships between the Cartesian coordinates [x, y] and the rational
polar coordinates [s, Q] of spread and quadrance are given by

¡ ¢
s = x2 / x2 + y 2 Q = x2 + y 2
x2 = sQ y 2 = (1 − s) Q.

In this Appendix, some well known curves are listed, together with the usual Cartesian
and polar forms, as well as new rational polar forms involving s and Q. As in the case
of both Cartesian and polar coordinates, rational polar coordinates will have the most
pleasant form only when the position of the curve is suitably chosen. For example,
both the Cartesian and polar equations of the ellipse become more complicated if the
ellipse is rotated and/or translated.

With rational polar coordinates it often becomes convenient to express s as a function

of Q, not the other way around as the usual polar situation might suggest.
Surprisingly, many diverse curves seem to have rational polar equations of a somewhat
similar form, typically a quadratic equation in s. This occurs particularly frequently
when the equation of the curve is even. This phenomenon should be explained.

It is important to note that the rational polar forms of these curves have an enormous
advantage over the usual polar forms for pure mathematics–they allow extensions of
these curves to general fields. In the examples below, we adopt the notational
convention that A ≡ a2 and B ≡ b2 . The Cartesian and polar forms for these classical
curves are taken from A catalog of special plane curves [Lawrence] and A book of
curves [Lockwood].

279
280 APPENDIX A. RATIONAL POLAR EQUATIONS OF CURVES

The derivations of the rational polar equations are left to the reader; they are often
interesting. Of course there are many additional curves to investigate.

Line The line has Cartesian equation y = ax and polar equation tan θ = a. Using
rational polar coordinates its equation is

1
s= .
1+A

Ellipse The ellipse has Cartesian equation

x2 y 2
+ 2 =1
a2 b
and polar equation
ab
r= p .
b cos θ + a2 sin2 θ
2 2

Using rational polar coordinates its equation is

A (Q − B)
s= .
Q (A − B)

y
1

-2 -1 0 1 x 2

-1

Figure A.1: Ellipse

Hyperbola The hyperbola has Cartesian equation

x2 y 2
− 2 =1
a2 b
and polar equation
ab
r= p .
b2 cos2 θ − a2 sin2 θ
Using rational polar coordinates its equation is

A (Q + B)
s= .
Q (A + B)
 281

4
y
2

-4 -2 0 2 x 4

-2

-4

Figure A.2: Hyperbola

Parabola The parabola has Cartesian equation

y 2 = 4ax

and polar equation

2a
= 1 − cos θ.
r

y2
1

0
1 2 3 4
x
-1

-2

Figure A.3: Parabola

Using rational polar coordinates its equation is

(1 − s)2 Q = 16As
or µ ¶2
Q + 8A 16A (Q + 4A)
s− = .
Q Q2

Cardioid The cardioid has polar equations of the form

r = 2a (1 + cos θ)
r = 2a (1 − cos θ)

with the following respective graphs.

282 APPENDIX A. RATIONAL POLAR EQUATIONS OF CURVES

2 2

1 1

0
0 1 2 3 4 4 3 2 -1
-1 -1

-2 -2

Figure A.4: Two cardioids

Both cases are covered by the rational polar equation

µ ¶2
Q + 4A Q
s− = .
4A A

This results in the ‘symmetric cardioid’ shown in Figure A.5.

-4 -2 0 2 4
-1

-2

Figure A.5: Symmetric cardioid

Limacon The limacon is a generalization of the cardioid, and has polar equation

r = 2a cos θ + b.

Using rational polar coordinates its equation is

µ ¶2
Q+B BQ
s− = .
4A 4A2

If B = 4A then this reduces to a cardioid. If B = A then this is the trisectrix with

rational polar equation
 283

µ ¶2
(Q + A) Q
s+ = .
4A 4A

Figure A.6 shows a graph of a trisectrix.

1.5

0.5

0 0.5 1 1.5 2 2.5 3

-0.5

-1

-1.5

Figure A.6: Trisectrix

Astroid The astroid has Cartesian equation

2 2 2
x3 + y3 = a3

or
¡ 2 ¢3
x + y 2 − a2 + 27a2 x2 y 2 = 0.

-1 x 1

-1

Figure A.7: Astroid

Using rational polar coordinates its equation is

µ ¶2
1 1 (A − Q)3
s− = − .
2 4 27AQ2
284 APPENDIX A. RATIONAL POLAR EQUATIONS OF CURVES

Eight curve The eight curve, or lemniscate of Gerono, has Cartesian equation
¡ ¢
x4 = a2 x2 − y 2

and polar equation

r2 = a2 sec4 θ cos 2θ.

Using rational polar coordinates its equation is

µ ¶2
A (A − Q) A
s− = .
Q Q2

Bullet nose The bullet nose has Cartesian equation

a2 b2
− =1
x2 y 2

and polar equation

r2 sin2 θ cos2 θ = a2 sin2 θ − b2 cos2 θ.

y 4

0
-1 -0.5 0.5 x1
-2

-4

Figure A.8: Bullet nose

Using rational polar coordinates its equation is

µ ¶2 2
A+B+Q (A + B + Q) − 4AQ
s− = .
2Q Q2

Deltoid The deltoid has the Cartesian equation

¡ 2 ¢2 ¡ ¢ ¡ ¢
x + y 2 − 8ax x2 − 3y 2 + 18a2 x2 + y 2 = 27a4 .
 285

y
2

-1 0 1 2 3
x
-1

-2

Figure A.9: Deltoid

Using rational polar coordinates its equation is

¡ 2 ¢2
2Q + 18AQ − 27A2
s (3 − 4s) = .
64Q3 A

The three-fold symmetry of the curve is reflected in the appearance of the third spread
polynomial S3 (s) = s (3 − 4s)2 .

Hippopede The hippopede, or horse fetter, (Proclus, 75 BC) has Cartesian equation
¡ 2 ¢2 ¡ ¢¡ ¢
x + y 2 +4b2 b2 − a2 x2 + y 2 = 4b4 x2

and polar equation ¡ ¢

r2 = 4b2 a2 − b2 sin2 θ .

1
y

-2 -1 0 1 2
x

-1

Figure A.10: Horse fetter

Using rational polar coordinates its equation is

Q
s= − A + B.
4B
286 APPENDIX A. RATIONAL POLAR EQUATIONS OF CURVES

Lemniscate of Bernoulli The lemniscate of Bernoulli has Cartesian equation

¡ 2 ¢2 ¡ ¢
x + y 2 = 2 x2 − y 2

and polar equation

r2 = 2 cos 2θ.

0.4
y
0.2

-1 -0.5 0.5 x 1

-0.2

-0.4

Figure A.11: Lemniscate of Bernoulli

Using rational polar coordinates its equation is

Q+2
s= .
4

Folium of Descartes The folium of Descartes, has Cartesian equation

x3 + y 3 + 3xy = 0.

3
y
2

-3 -2 -1 1 2 3
x
-1

-2

-3

Figure A.12: Folium of Descartes

This curve has a more complicated rational polar equation; it is

¡ ¢2 ¡ ¢
(2s − 1)2 s2 − s + 1 Q2 −18s (1 − s) 1 − 3s + 3s2 Q+81s2 (1 − s)2 = 0.
 Appendix B

Ellipson

The ellipson consists of all points [x, y, z] inside, or on, the unit cube 0 ≤ x, y, z ≤ 1
satisfying the Triple spread formula
2 ¡ ¢
(x + y + z) = 2 x2 + y 2 + z 2 + 4xyz.

Figure B.1: The ellipson

This surface resembles an inflated tetrahedron, and indeed intersects the unit cube at
precisely a tetrahedron, with points [0, 0, 0] , [1, 1, 0] , [1, 0, 1] and [0, 1, 1], and volume
1/3. Its cross sections in any of the coordinate plane directions is otherwise always an
ellipse, tangent to the unit square.

287
288 APPENDIX B. ELLIPSON

For example, the cross sections corresponding to z = 0.1, 0.2, 0.4 and 0.8 are shown in
the x − y plane in Figure B.2.

0.8

0.6
y
0.4

0.2

0
0 0.2 0.4 x 0.6 0.8 1

Figure B.2: Elliptical cross sections

On the other hand, a cross section parallel to a face of the tetrahedron, such as the
plane x + y + z = 2, yields the interesting curve shown in Figure B.3.

Figure B.3: Oblique cross section

In suitable coordinates, such a curve has an equation of the form

√ √
3b2 a 6 − 6a2 − a3 6 − 6b2 + 5/6 = 0.

In rational polar coordinates this becomes

2 (36Q − 5)2
s (3 − 4s) = 3
(6Q)
which bears some similarity to the equation defining the deltoid. The three-fold
symmetry is reflected by the fact that the left hand side is S3 (s).

Exercise B.1 (Harder, requires calculus) Show that the ellipson has volume

π 2 /16 ≈ 0.616 850 275 · · · . ¦

 Theorems with pages and
Important Functions

1. Quadratic compatibility (33) 33. Spread plus cross (76)

2. Line through two points (38) 34. Spread number (76)
3. Collinear points (39) 35. Spread ratio (77)
4. Concurrent lines (39) 36. Cross ratio (78)
5. Point on two lines (40) 37. Twist ratio (78)
6. Parallel to a line (41) 38. Complementary spreads (79)
7. Altitude to a line (41) 39. Perpendicular spreads (79)
8. Foot of an altitude (42) 40. Spread law (80)
9. Affine combination (46) 41. Cross law (81)
10. Thales’ theorem (48) 42. Quadrea spread (82)
11. Parallelogram center (49) 43. Spread from points (84)
12. Perpendicular bisector (50) 44. Cross from points (84)
13. Reflection of a point in a line (52) 45. Vertex bisector (85)
14. Rotation of a line in a point (53) 46. Equal quadrance to two lines (88)
15. Reflection of a line in a line (54) 47. Triple spread formula (89)
16. Reflection (55) 48. Spread from ratio (91)
17. Lineation (56) 49. Triple cross formula (92)
18. Cyclic reflection (56) 50. Triple twist formula (93)
19. Null line (60) 51. Equal spreads (94)
20. Midpoint (60) 52. Spread reflection (95)
21. Parallelogram quadrance (62) 53. Two spread triples (98)
22. Triple quad formula (63) 54. Quadruple spread formula (99)
23. Pythagoras’ theorem (65) 55. Quadruple cross formula (100)
24. Equal quadrance to two points (66) 56. Three equal spreads (101)
25. Quadrance to a line (67) 57. Recursive spreads (102)
26. Quadrea (68) 58. Consecutive spreads (103)
27. Right quadrea (68) 59. Spread polynomial formula (106)
28. Triangle quadrea (69) 60. Spread composition (110)
29. Archimedes’ formula (70) 61. Triple turn formula (115)
30. Two quad triples (70) 62. Signed area (117)
31. Quadruple quad formula (71) 63. Bretschneider’s formula (118)
32. Brahmagupta’s identity (72) 64. Null isosceles (121)

289
290 THEOREMS WITH PAGES AND IMPORTANT FUNCTIONS

65. Pons Asinorum (122) 94. Classification of isometries (158)

66. Isosceles triangle (122) 95. Order three star (160)
67. Isosceles median (123) 96. Order five star (161)
68. Isosceles reflection (124) 97. Order seven star (162)
69. Equilateral triangle (125) 98. Polygon triangle (165)
70. Right midpoint (126) 99. Conic center (168)
71. Median triangle (128) 100. Circle uniqueness (171)
72. Triangle proportions (131) 101. Parabola uniqueness (173)
73. Quadrilateral proportions (133) 102. Conic line intersection (176)
74. Two struts (134) 103. Right diameter (177)
75. Stewart’s theorem (136) 104. Circle chord (178)
76. Median quadrance (137) 105. Subtended spread (178)
77. Median spreads (138) 106. Equal products (179)
78. Menelaus’ theorem (139) 107. Projective circle (181)
79. Alternate spreads (140) 108. Unit circle (182)
80. Ceva’s theorem (141) 109. Cyclic quadrilateral (183)
81. Circumcenter (143) 110. Cyclic quadrilateral spreads (184)
82. Extended spread law (144) 111. Quadrilateral circumquadrance (185)
83. Circumcenter formula (146) 112. Cyclic quadrilateral quadrea (187)
84. Affine circumcenter (146) 113. Cyclic signed area (188)
85. Orthocenter (147) 114. Ptolemy’s theorem (190)
86. Orthocenter formula (149) 115. Four point relation (191)
87. Affine orthocenter (149) 116. Euler line (193)
88. Incenter (150) 117. Nine point circle (195)
89. Incenter spread (151) 118. Translate of a conic (198)
90. Inquadrance (153) 119. Tangent to a conic (200)
91. Isometries preserve collinearity (157) 120. Tangent to a circle (200)
92. Specifying isometries (157) 121. Parabola reflection (201)
93. Orthogonality (157)

Important Functions
2 ¡ ¢
A (a, b, c) = (a + b + c) − 2 a2 + b2 + c2 .
¡ ¢
S (a, b, c) = (a + b + c)2 − 2 a2 + b2 + c2 − 4abc.
³ ¡ ¢´2
Q (a, b, c, d) = (a + b + c + d)2 − 2 a2 + b2 + c2 + d2 − 64abcd.
µ ¡ ¢ ¶2
(a + b + c + d)2 − 2 a2 + b2 + c2 + d2
R (a, b, c, d) =
−4 (abc + abd + acd + bcd) + 8abcd
− 64abcd (1 − a) (1 − b) (1 − c) (1 − d) .

E (Q1 , Q2 , Q3 , P1 , P2 , P3 )
µ ¶
4P1 P2 P3 + (P2 + P1 − Q3 ) (P2 + P3 − Q1 ) (P1 + P3 − Q2 )
=2 .
−P1 (P2 + P3 − Q1 )2 − P2 (P1 + P3 − Q2 )2 − P3 (P2 + P1 − Q3 )2
 Bibliography

[Coxeter] H. S. M. Coxeter, Introduction to Geometry (2nd ed.),

John Wiley and Sons, New York, 1989.
[Coxeter-Greitzer] H. S. M. Coxeter and S. L. Greitzer, Geometry
Revisited, Random House, New York, 1967.
[Dorrie] H. Dorrie, 100 Great Problems of Elementary
Mathematics: Their history and solution, translated by
D. Antin, Dover, New York, 1965.
[Euclid] Euclid, The Elements (in three volumes), Dover, New
York, 1956.
[Gutierrez] A. Gutierrez, ‘Eyeball theorems’, in The Changing
Shape of Geometry, Editor Chris Pritchard, Cambridge
University Press, Cambridge, 2003.
[Hilbert] D. Hilbert, Foundations of Geometry, Revised by P.
Bernays, Open Court, LaSalle, Illinois, 1971.
[Lawrence] J. D. Lawrence, A Catalog of Special Plane Curves,
Dover, New York, 1972.
[Lockwood] E. H. Lockwood, A Book of Curves, Cambridge
University Press, London, 1961.
[Maor] E. Maor, Trigonometric Delights, Princeton University
Press, Princeton, 1998.
[Robbins] D. A. Robbins, ‘Areas of Polygons Inscribed in a Circle’,
The American Mathematical Monthly Volume 102
Number 6 (1995) 523—530.
[Shepherd] F. A. Shepherd, Surveying Problems and Solutions,
Edward Arnold, London, 1968.
[Shklarsky-Chentzov-Yaglom] D. O. Shklarsky, N. N. Chentzov and I. M. Yaglom, The
USSR Olympiad Problem Book, W. H. Freeman and
Company, San Francisco, 1962.

291
292 BIBLIOGRAPHY

[Snapper-Troyer] E. Snapper and R. J. Troyer, Metric Affine Geometry,

Academic Press, New York, 1971.
[Weisstein] E. W. Weisstein, CRC Concise Encyclopedia of
Mathematics, Chapman and Hall, Boca Raton, 1999.
[Wells] D. Wells, The Penguin Dictionary of Curious and
Interesting Geometry, Penguin, London, 1991.
[Wildberger] N. J. Wildberger, ‘Real Fish, Real Numbers, Real Jobs’,
Mathematical Intelligencer Volume 21 Number 2 (1999)
4—7.
[Zeilberger] D. Zeilberger (with S. B. Ekhad), ‘Plane Geometry: An
elementary school textbook (ca. 2050)’, Mathematical
Intelligencer Volume 21 Number 3 (1999) 64-70.
 Index

Acceleration, 235 Bretschneider’s formula, 118, 187

Acute Bullet nose, 284
sector, 216, 219
triangle, 218 Cantor, 20
Adjacent sectors, 216 Cardioid, 281
Affine Cartesian
combination, 46, 146, 218 approach of Descartes, 19
transformation, 51, 53 coordinate geometry, 35
Affine circumcenter theorem, 146, 218 Cassini oval, 207
Affine combination theorem, 46, 60 Cauchy’s identity, 27
Affine orthocenter theorem, 149 Center
Algebraic dynamics, 238 of a circle, 169
Almagest, The, 190 of a conic, 168
Alternate spreads theorem, 140—142 of a parallelogram, 49
Altitude Central
foot of, 42 circle, 169, 170, 186
from a point to a line, 41 line, 37
of a triangle, 147 reflection, 156—158
Altitude to a line theorem, 41 rotation, 156—158, 181
Anti-symmetric polynomial, 29 Centroid, 193
Archimedes, 5, 11, 269, 274 of a hemisphere, 274
Archimedes’ formula, 70 of a triangle, 48
Archimedes’ function, 64, 190, 192, 252 Ceva, 139
Area, 72, 269 Ceva’s theorem, 141
signed, 68, 116, 188 Chaos theory, 105
surface, 276 Characteristic of a field, 21, 193, 203
Astroid, 283 Chebyshev polynomials, 10, 106
Chord, 176—178
Beta function/integral, 270, 274, 277 Chromogeometry, xviii
Binomial theorem, 27 Circle, 18, 169, 177
Bisector center of, 169
of a vertex, 85, 88, 269 central, 169
perpendicular, 50 chord of, 177
Blue geometry, 40 diameter, 177
Box, 228 diameter of, 177
Brahmagupta’s formula, 188 null, 169—171
Brahmagupta’s identity, 72, 188 projective, 181
Brahmagupta’s theorem, 180 quadrance of, 169

293
294 INDEX

unit, 181 Cosine law, 3, 14

Circle chord theorem, 178 Cross, 74, 124
Circle uniqueness theorem, 171 polynomial, 112
Circumcenter, 143, 146, 193 polynomial (scaled), 112
Circumcenter formula, 146 Cross from points theorem, 84
Circumcenter theorem, 143 Cross law, xvi, 10, 81, 222, 225, 249,
Circumcircle, 177, 189, 194, 251 253
Circumquadrance, 143 Cross ratio theorem, 78
Classification of isometries theorem, Cube, 261
158 Cubic curve, 202
Collin’s point, 252 Cyclic quadrilateral, 72, 183, 185
Collinear points, 38 solvable, 185
Collinear points theorem, 39 subtended spreads of, 184
Collinear quadrance rules, 215 Cyclic quadrilateral quadrea theorem,
Complementary spreads theorem, 79 187, 189
Completing the square, 32 Cyclic quadrilateral spreads theorem,
Complex number, 22, 62, 153, 182 184, 190
Composition Cyclic quadrilateral theorem, 183
of central reflections, 156 Cyclic reflection theorem, 56
of central rotations, 156 Cyclic signed area theorem, 188
of spread polynomials, 110
Decimal number, 22
of transformations, 51
Deltoid, 284
Concurrent lines, 38
Descartes, 19, 71
Concurrent lines theorem, 39
folium of, 203
Congruent triangles, 127
Determinants, 28, 192, 206, 266, 271
Conic, 167, 197, 235
Diagonal
center of, 168
lines of a quadrola, 175
centered, 169
of a quadrilateral, 45
central, 169
side of a quadrilateral, 45
discriminant, 208
Diameter of a circle, 177
empty, 167 Diophantus, 27
equation of, 167 Directrix of a parabola, 172, 173, 201
singular, 197 Discriminant conic, 208, 210
tangent line to, 199 Dodecahedron, 264
Taylor, 198
translate of, 198, 205 Eight curve, 284
Conic center theorem, 168 Eight point circle, 196
Conic line intersection theorem, 176 Einstein’s theory of relativity, xviii,
Consecutive spreads theorem, 103, 110, 241, 242
161 Elements, The, 11, 19
Content (4-dimensional volume), 277 Ellipse, 174, 175, 205, 279, 280
Coordinates imaginary, 174
Cartesian, 19 Ellipson, 287
of a point, 35 cross sections, 288
rational cylindrical, 273 volume, 288
rational polar, 265 Empty conic, 167
rational spherical, 271 Equal products theorem, 179
INDEX 295

Equal quadrance to two lines theorem, Hansen, 255

88 Hansen’s problem, 255
Equal quadrance to two points Harmonic relation, 221
theorem, 66, 123, 137, 143 Height
Equal spreads theorem, 85, 94, 101, 230 from three spreads, 247
Equilateral triangle, 125 of a raised object, 245
Equilateral triangle theorem, 125 of object with inaccessible base,
Euclid, 11, 19 244
Euler, 111, 191, 270 of object with vertical face, 243
Euler function, 191, 225, 251, 253 Hemisphere
Euler line, 194 centroid of, 274
Euler line theorem, 193 moment of, 274
Even equation, 207, 279 Heron’s theorem, 70
Extended spread law, 144, 179 Hexagon, 164
Eyeball theorem, 224 Hilbert, D., 19
Hippopede, 285
Face spread, 259, 264 Hirschhorn, M., 108
of cube, 261 Horse fetter, 285
of dodecahedron, 264 Hyperbola, 174, 205, 280
of icosahedron, 263 Hyperbolic cap, 275
of octahedron, 262 Hyperbolic geometry, 13
of tetrahedron, 260 Hyperbolic ring, 275
Factor theorem, 31
Fermat, 19 Ice cream cone, 272
Fibonacci, 27 Icosahedron, 263
Fibonacci’s identity, 27, 87, 152 Identity transformation, 51
Field, 21 Incenter of a triangle, 150
Fifth spread polynomial, 31 Incenter spread theorem, 151
Finite field, 23 Incenter theorem, 150
Fixes a point, 51 Inquadrance of a triangle, 150
Focus Inquadrance theorem, 153
of a parabola, 172, 173, 210 Intersection
of a quadrola, 174, 202 of a line and a conic, 176
Folium of Descartes, 203, 286 of two curves, 203
Foot of an altitude, 6, 42, 43, 196, 230, of two lines, 40
264 Invariant measure
Foot of an altitude theorem, 42 on a circle, 268
Four point relation, 191, 225, 251 on a sphere, 276
Invertible transformation, 51
Gamma function, 270 Isometries preserve collinearity
Go board, 161 theorem, 157
Goh, S., 106, 108, 111 Isometry, 155
Grammola, 175, 176, 210 Isosceles median theorem, 123, 124, 148
diagonal lines of, 175 Isosceles reflection theorem, 124, 165
quadrance of, 175 Isosceles triangle, 121
Greeks (ancient), 17 Isosceles triangle theorem, 122, 124,
Gutierrez, A., 224 165, 179, 228, 234, 260—264
296 INDEX

Le, V., 69 Nine point circle, 194

Lemniscate of Bernoulli, 207, 267, 286 Nine point circle theorem, 195
Lemniscate of Gerono, 284 Non-singular point
Leprechauns, 23 of first order, 202
Limacon, 282 of second order, 203
Line, 9, 36, 37, 280 Normal form
Cartesian equation of, 9 of a quadratic equation, 32
null, 36 Normal to a plane, 227
of a quadrilateral, 45 Null
of a side, 43 circle, 169, 171
of a triangle, 44, 114 line, 36
reflection in a, 52 triangle, 44
Line segment, 214 Null isosceles theorem, 121
adjacent, 214 Null line theorem, 60, 66, 78, 121, 178
overlap, 214
Line through two points theorem, 38 Obtuse
Linear equations, 30, 49, 168 sector, 216, 219
Lineation, 53, 159 triangle, 218
Lineation theorem, 56 Octahedron, 262
Lines Octant, 271, 277
concurrent, 38 Order five star theorem, 161
intersection of, 40 Order seven star theorem, 162
parallel, 40 Order three star theorem, 160
perpendicular, 40 Oriented
Logistic map, 10, 94, 105 n-gon, 114, 188
quadrilateral, 114
Müller, Johann, 246 side, 113
MAGMA, xvi triangle, 113
MAPLE, xvi vertex, 113
MATHEMATICA, xvi Origin, 7, 37, 155, 208, 235, 250
MATLAB, xvi Orthocenter, 148, 149, 193
Median of a triangle, 48, 123, 128, 132 Orthocenter formula, 149
Median quadrance theorem, 137, 138 Orthocenter theorem, 147, 193
Median spreads theorem, 138 Orthogonality
Median triangle theorem, 128, 144 isometry, 155
Menelaus, 139 of spread polynomials, 109
Menelaus’ theorem, 139 Orthogonality theorem, 157
Midpoint of a side, 47 Overlapping sectors, 216
Midpoint theorem, 60, 66, 67, 95, 123,
124, 195, 224 Parabola, 172, 205, 281
Moment of hemisphere, 274 algebraic dynamics, 238
Moment of inertia, 274 axis of, 172
Multiplicity of a zero, 31 directrix of, 172, 201, 210, 238
MUPAD, xvi focus of, 172, 201, 210, 238
projectile motion, 235
N-gon, 45, 269 vertex of, 172, 238
oriented, 114, 188 Parabola reflection theorem, 201
Nine point center, 194 Parabola uniqueness theorem, 173
INDEX 297

Parallel cross, 112

lines, 40 fifth spread, 31, 161
planes, 227 spread, 101, 102, 104
sides, 43 symmetric, 27
to a line through a point, 41 third spread, 160, 285, 288
Parallel to a line theorem, 41 Pons Asinorum, 122, 123, 125
Parallelogram, 45, 49, 62, 128 Pothenot, 251
center of, 49 Proclus, 285
Parallelogram center theorem, 49, 195 Projectile motion, 235
Parallelogram quadrance theorem, 62 Projective circle, 181
Parametrizing parametrizing, 181
the projective circle, 181 Projective circle theorem, 181
the unit circle, 182 Projective trigonometry, 27, 259
Pascal’s triangle, 108 Proportions, 17, 25, 36, 130, 131
Pentagon, 31, 164, 263, 264 Ptolemy, 190
Perpendicular Ptolemy’s theorem, 190
bisector, 50 Pyramid, 230
line to a plane, 227 Pythagoras’ theorem, xvi, 10, 65, 126,
lines, 40 137, 145, 148, 158, 214, 216,
planes, 227 228—230, 234, 246, 249
sides, 43 in three dimensions, 232
Perpendicular bisector theorem, 50 spherical case, 250
Perpendicular spreads theorem, 79, 260 Pythagoreans, 17
Pi, 267
Pi squared over 16, 22, 269, 288 Quad triple, 64, 134, 180, 190, 214,
Plane, 227, 230, 231, 234, 260 221, 240
normal to, 227 Quadrance, 4, 9, 59, 243, 246, 255, 265,
parallel, 227 271, 279
spread between, 227 from a point to a line, 67
Point, 9, 35, 37 of a circle, 169
collinear, 38 of a grammola, 175, 176
coordinates of, 35 of a quadrola, 174
interior to a line segment, 214 of a side, 61
interior to a triangle, 218 of a triangle, 61
lying on a ray, 215 opposite a vertex, 61
non-singular, 199 Quadrance to a line theorem, 67
of a quadrilateral, 45 Quadrant (++), 266
of a triangle, 44, 114 Quadratic compatibility theorem, 33,
of a vertex, 43 71, 98
Point on two lines theorem, 40, 42, 146, Quadratic equation, 32, 33, 98, 103,
151 237, 247, 252, 253, 279
Polar spread, 265 compatibility, 33, 64
second, 271 completing the square, 32
third, 277 normal form, 32, 91, 92, 129, 219
Polygon triangle theorem, 165 Quadratic forms, 17
Polynomial Quadrea, 68, 70, 82, 83, 96, 113, 117,
anti-symmetric, 29 118, 124, 145, 153, 269
298 INDEX

of a quarter circle, 269 Reflection of a line in a line theorem, 54

of an n-gon, 118 Reflection of a point in a line theorem,
Quadrea spread theorem, 82, 269 52
Quadrea theorem, 68 Reflection theorem, 55, 163
Quadrilateral, 45 Regiomontanus (Johann Müller), 246
cyclic, 183 Regiomontanus’ problem, 246
diagonal quadrance, 61 Regular
diagonal side, 45 7-gon, 233
oriented, 114 n-gon, 164
quadrance, 61 pentagon, 166, 263, 264
side, 45 polygon, 163
vertex, 45 Regular star, 159
Quadrilateral circumquadrance of order five, 161
theorem, 185, 187 of order seven, 162
Quadrilateral proportions theorem, 133 of order three, 160
Quadrola, 174, 176, 202, 208 spread of, 159
focus, 174 Relativity theory of Einstein, 17, 242
focus of, 202 RENE, xx
quadrance, 174 Resection problem, 251
reflection property, 202 Rhombus, 45, 62
Quadruple cross formula, 100 Ribbon, 171
Quadruple quad formula, 71, 185 Right
Quadruple quad function, 70, 185 sector, 216
Quadruple spread formula, 99, 100, triangle, 44
184, 185 vertex, 43, 68
Quadruple spread function, 99, 190 Right diameter theorem, 177
Right midpoint theorem, 126, 177, 195
Rational cylindrical coordinates, 273 Right quadrea theorem, 68
Rational numbers, 21 Romans, xx
Rational polar coordinates, 265, 279 Rotation
Rational polar equation, 279 central, 156
of a lemniscate, 268 composition, 156
of a lemniscate of Bernoulli, 267 composition of, 156
of a line, 280 in a point, 52
Rational spherical coordinates, 265, 271 Rotation of a line in a point theorem,
Ray, 215 53
base point of, 215
Rectangle, 45, 62, 195 Second polar spread, 271
Rectangular box, 228 Sector, 215, 255
Recursive spreads theorem, 102, 103 acute, 216
Reflection adjacent, 216
central, 156 base point of, 215
composition of, 156 obtuse, 216
in a line, 52 overlapping, 216
property of parabola, 201 right, 216
property of quadrola, 202 spread of, 216
sequence, 56 type, 216
INDEX 299

Seven-fold symmetry, 162, 233 ruler, 213

Side, 43, 113 second polar, 271
adjacent, 45 sequence, 104, 159
midpoint of, 47 subtended, 179, 184
of a quadrilateral, 45 third polar, 277
of a triangle, 44, 114 triple, 90
opposite, 45 Spread composition theorem, 110
Signed area, 116, 188, 189 Spread from points theorem, 84
Signed area theorem, 117 Spread from ratio theorem, 91, 257
Similar triangles, 127 Spread law, xvi, 10, 16, 80, 246, 248,
Sine law, 3, 14 253, 258
Singular conic, 197 Spread number theorem, 76, 85, 110
Snell’s law, 239 Spread plus cross theorem, 76
Snellius, 251 Spread polynomial formula, 106
Snellius-Pothenot problem, 251 Spread ratio theorem, 77
Soddy, F., 71 Spread reflection theorem, 95
Solvable cyclic quadrilateral, 185, 187 Spread triple, 90, 91, 98, 101, 102, 148,
Special theory of relativity, xviii, 241 152
Specifying isometries theorem, 157 Square
Spherical cap number, 24, 36, 68, 69
surface area, 276 vertex, 87, 177
volume, 273 Square-spread number, 25, 85
Spherical coordinates, 271 Standard
Spherical ring, 273 conic, 167
Spherical surface measure, 278 line, 37
Spherical trigonometry, 13 Star (regular), 159, 163
Pythagoras’ theorem, 250 Stewart’s theorem, 136, 247
Spread, 4, 9, 73, 279 Stewart, Matthew, 136
array, 108 Subtended spread theorem, 178, 251
between planes, 227 Surface area, 276
big, 217, 219 Surface volume of 3-sphere, 278
complementary, 79 Surveying problems, 243
diagonal subtended, 184 Symmetric function, 27
face, 259
little, 217, 219 Tangent
number, 24 conic, 203, 205
of a quadrilateral, 74 curves, 203
of a regular star, 159 line to a circle, 200
of a sector, 216 line to a conic, 199, 202
of a triangle, 74 line to a standard conic, 197
of a vertex, 74 second order, 203
opposite a side, 74 Tangent to a circle theorem, 200
order of, 103 Tangent to a conic theorem, 200
over a right triangle, 249 Taylor conic, 198
polar, 265 Taylor curve, 202, 207
polynomial, 10, 102 Tetrahedron, 260, 261
polynomial (scaled), 108 Thales’ theorem, 48, 128, 144, 195
 300 INDEX

Third polar spread, 277 219, 223, 287

Three spreads theorem, 101 normal form, 91
Transformation, 51, 155 Triple spread function, 90
affine, 51, 53 Triple turn formula, 115
composition of, 51 Triple twist formula, 93, 115, 242
Translate Trisectrix, 282
by a point, 155, 197 Turn, 114
of a conic, 198 Twist, 75, 114, 124
Translate of a conic theorem, 198 Twist ratio theorem, 78, 95, 222
Triangle, 44, 113 Two quad triples theorem, 70, 71
acute, 218 Two spread triples theorem, 98, 148
altitude, 147 Two struts theorem, 134
centroid, 48 Type of a sector, 216
circumcenter, 143
circumquadrance, 143 Uniform motion, 18
congruent, 127 Unit circle, 181
equilateral, 125 parametrizing, 182
incenter, 150 Unit circle theorem, 182
inequality, 82 Universal geometry, 9
inquadrance, 150 Van Schouten’s theorem, 180
isosceles, 121, 269 Velocity of a projectile, 235
median, 48, 128 Vertex, 43
null, 44 adjacent, 45
obtuse, 218 of a parabola, 172
orthocenter, 148 of a quadrilateral, 45
perpendicular bisector, 143 of a triangle, 44, 114
point, 44 opposite, 45
quadrea, 68 oriented, 113
right, 44 right, 43, 44, 68, 126
side, 44 square, 87, 177
signed area, 116 Vertex bisector theorem, 85
similiar, 127 Vertical and horizontal spreads, 248
vertex, 44 Volume
Triangle proportions theorem, 131 of 3-sphere, 277
Triangle quadrea theorem, 69 of hyperbolic cap, 275
Triangle spread rules, 90, 219, 223, 234, of hyperbolic ring, 275
255 of ice cream cone, 272
Triangles of Every Kind, 246 of sphere, 272
Trigonometric functions, 3, 18 of spherical cap, 273
Trigonometry, 3 of spherical ring, 273
classical, 3
rational, 4 Wedge, 231
Triple coturn formula, 116
Triple cross formula, 92, 93, 96, 100 Zeilberger, D., xx
Triple quad formula, xvi, 10, 63, 67, 81, Zero, 31
191 multiplicity of, 31
Triple spread formula, xvi, 10, 89, 217, Zero denominator convention, 28

View publication stats