Birkhäuser Advanced Texts

Basler Lehrbücher

Hiroshi Maehara
Horst Martini

Circles,
Spheres and
Spherical
Geometry
Birkhäuser Advanced Texts Basler Lehrbücher

A series of Advanced Textbooks in Mathematics

Series Editors
Steven G. Krantz, Washington University, St. Louis, USA
Shrawan Kumar, University of North Carolina at Chapel Hill, Chapel Hill, USA

This series presents, at an advanced level, introductions to some of the fields of

current interest in mathematics. Starting with basic concepts, fundamental results
and techniques are covered, and important applications and new developments
discussed. The textbooks are suitable as an introduction for students and non–
specialists, and they can also be used as background material for advanced courses
and seminars.
Hiroshi Maehara • Horst Martini

Circles, Spheres
and Spherical Geometry
Hiroshi Maehara Horst Martini
Faculty of Education Faculty of Mathematics
University of the Ryukyus University of Technology
Nakagami Gun Nishihara Cho, Okinawa, Chemnitz, Germany
Japan

ISSN 1019-6242 ISSN 2296-4894 (electronic)

Birkhäuser Advanced Texts Basler Lehrbücher
ISBN 978-3-031-62775-0 ISBN 978-3-031-62776-7 (eBook)
https://doi.org/10.1007/978-3-031-62776-7

© The Editor(s) (if applicable) and The Author(s), under exclusive license to Springer Nature Switzerland
AG 2024
This work is subject to copyright. All rights are solely and exclusively licensed by the Publisher, whether
the whole or part of the material is concerned, specifically the rights of translation, reprinting, reuse
of illustrations, recitation, broadcasting, reproduction on microfilms or in any other physical way, and
transmission or information storage and retrieval, electronic adaptation, computer software, or by similar
or dissimilar methodology now known or hereafter developed.
The use of general descriptive names, registered names, trademarks, service marks, etc. in this publication
does not imply, even in the absence of a specific statement, that such names are exempt from the relevant
protective laws and regulations and therefore free for general use.
The publisher, the authors and the editors are safe to assume that the advice and information in this book
are believed to be true and accurate at the date of publication. Neither the publisher nor the authors or
the editors give a warranty, expressed or implied, with respect to the material contained herein or for any
errors or omissions that may have been made. The publisher remains neutral with regard to jurisdictional
claims in published maps and institutional affiliations.

This book is published under the imprint Birkhäuser, www.birkhauser-science.com by the registered
company Springer Nature Switzerland AG
The registered company address is: Gewerbestrasse 11, 6330 Cham, Switzerland

If disposing of this product, please recycle the paper.

To Keiko and Angelika
Preface

Circles and spheres have been successfully studied in ancient Greece. The Greek
geometers saw the circles and spheres as perfect figures. By a metaphysical
reasoning, Aristotle (BC384-BC322) asserted that the orbits of planets are circles.
Though the notions of ellipses and parabolas were known due to the famous book on
conic sections by Apollonius of Perga (BC240-BC190), the opinion that the planet
orbits must be circles continued to survive, partly supported by the famous name of
Aristotle.
However, in the sixteenth century, Johannes Kepler (1571–1630) confirmed that
the orbits of planets must be ellipses. Additionally, Kepler posed a famous problem
on spheres: What is the densest packing of unit spheres (or solid unit balls) in three
dimensions? The famous Kepler conjecture says that the face-centered cubic lattice
packing yields this densest packing, and this was proved by Thomas C. Hales [371]
in 2005.
Isaac Newton (1643–1727) and his friend David Gregory (1659–1708) argued
about the maximum number of unit balls that can touch simultaneously a given unit
ball. Newton claimed that 12 is the maximum number, while Gregory asserted 13.
Which one is right? This problem is known as the problem of thirteen balls, or the
Gregory-Newton problem. It was settled in favor of Newton in 1953 by K. Schütte
and B. L. van der Waerden [716].
In the so-called Japanese temple geometry, geometers from the Edo-period
(1600–1868) also studied circles and spheres. Descartes’ circle theorem and its
three-dimensional version, namely Soddy’s formula, were found independently by
them. The Japanese mathematicians dedicated a large variety of related results to
temples and shrines as Sangaku (artistically painted wooden panels); see Hidetoshi
Fukagawa and Dan Pedoe [284], Hidetoshi Fukagawa and John F. Rigby [285] as
well as Hidetoshi Fukagawa and Tony Rothman [286].
Spherical geometry, or the geometry of spherical space, is the two-dimensional
geometry of the surface of a ball, as non-Euclidean geometry having important
applications in astronomy, geodesy, navigation, and further scientific fields, also
in form of higher dimensional analogues (see [131], [809] and, also for historical
background, [678]).

vii
viii Preface

In this book we deal with the geometry of circles and spheres, and also with
results in spherical spaces. As prerequisites for reading it, the authors recommend
only Linear Algebra, Advanced Calculus, and (for Chap. 8) elementary Probability
Theory. In addition, some results from Graph Theory are also used in this book.
They are completely introduced when it becomes necessary.
We begin the first chapter with the planar inversion on circles and its three-
dimensional analogue. Inversions are convenient transformations when one studies
families of circles or spheres. For example, suppose that mutually tangent three big
balls lie on the floor and touch the ceiling that is parallel to the floor. Imagine that
in the room enclosed by these three big balls (and also between the floor and the
ceiling), a number of small balls of various sizes are piled up in the following way:
The first ball is tangent to the floor and the three big balls; the second ball is tangent
to the first one and the three big balls; the next ball is tangent to the previous one
and the three big balls; and so on. Then one question is: how many small balls could
be piled up in such a way? Considering the inversion on a sphere with center at the
contact point of any two big balls, we can see that exactly four small balls could be
piled up in such a way. See also the literature on Soddy’s hexlet in Chap. 1 of this
book.
Bend formulas, as treated in Chap. 2, are related to the topics studied by Japanese
mathematicians in the Edo-period. Relations among the radii of Steiner cycles,
Soddy’s hexlet, and a pair of linked four cycles of spheres are studied in Chap. 2,
besides Descartes’ circle theorem and Soddy’s formula.
In Chap. 3, the orthogonal-circle-representation theorem (OCR theorem) is
introduced and proved. This theorem characterizes a special type of planar graphs
that can be represented by a family of circles satisfying certain conditions. From this
theorem, the coin-graph theorem (also called Koebe-Andreev-Thurston theorem)
and Steinitz’s theorem concerning the graph of a convex 3-polytope follow easily.
Chapters 4, 5, and 7 refer to spherical geometry, where figures on a sphere, such
as geodesic curves, spherical caps, spherical triangles, spherical polygons, spherical
ellipses, etc., are studied. Many formulas, e.g., the spherical cosine law, the spherical
sine law, Euler’s formula for the spherical excess, etc., are derived by Cesàro’s
method (a method using stereographic projections).
In Chap. 6, the problem of thirteen balls is considered. This is done by using a
result on planar graphs, together with a few results from spherical geometry.
In Chap. 8, problems from geometric probability on the sphere are studied, thus
referring to random point arrangements and random polygons.
In Chap. 9, intersection graphs of (also random) spherical caps are studied.
Chapter 10 concerns four-point sets (quartets) on a sphere that determine the
radius of the sphere.
Spheres and balls in high dimensions are considered in Chap. 11. Problems like
the determination of the number of unit balls that can simultaneously touch a ball of
small radius in n-dimensional space are considered there.
In Chap. 12, Cayley-Menger matrices are introduced, and orthogonal-sphere
systems as well as tangent-sphere systems are investigated. Further on, Ptolemy’s
theorem is generalized.
Preface ix

Casey’s theorem and a generalization of it are proved in Chap. 13. (Note

that Casey’s theorem is an extension of Ptolemy’s theorem, where the points in
Ptolemy’s theorem are replaced by circles, and the distance between points is
replaced by the common tangent distance between circles.)
At the end of each chapter, we present exercises of different levels of difficulty,
and at the end of the book, solutions to exercises selected from this set are given.
Each chapter ends with notes that support further reading and, in some cases,
bring the text close to the research front. Depending on the respective situation in the
specialist literature, these notes are very heterogeneous in nature. In some chapters,
they bring the text indeed close to the research front (reflecting also research papers),
and in other cases they are relatively short, citing only related monographs, or
monographs and surveys.
Topics that are obviously close to the contents of our book are, e.g., packing and
covering problems for circles/spheres and related topological results (such as the
Borsuk-Ulam theorem). However, we did not take them into consideration since they
would have exceeded the scope of this book. (For packing and covering problems,
we refer to the very recent extension [262] of L. Fejes Tóth’s famous monograph,
and the Borsuk-Ulam theorem is excellently discussed by J. Matoušek [561].)
We hope that our book finds interested readers who will share our enthusiasm for
the subfield of geometry covered by it.

Okinawa, Japan Hiroshi Maehara

Chemnitz, Germany Horst Martini
2024
Contents

1 Inversion and Stereographic Projection. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1

1.1 Inversion of the Plane . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.2 Euler’s Triangle Theorem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
1.3 Steiner Cycles . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
1.4 Inversions of 3-Space . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
1.5 Soddy’s Hexlet . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
1.6 Stereographic Projection . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
1.7 Appendix: Poncelet’s Porism . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
1.8 Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
1.9 Notes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
1.9.1 Notes on Inversion in the Plane and in 3-Space . . . . . . . . . . . 15
1.9.2 Notes on Euler’s Triangle Theorem. . . . . . . . . . . . . . . . . . . . . . . . 17
1.9.3 Notes on Steiner Chains and Steiner’s Porism . . . . . . . . . . . . 18
1.9.4 Notes on Soddy’s Hexlet . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20
1.9.5 Notes on the Stereographic Projection. . . . . . . . . . . . . . . . . . . . . 21
1.9.6 Notes on Poncelet’s Porism . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26
2 Bend Formulas . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31
2.1 Bend Formulas for Steiner Cycles . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31
2.2 Bend Formula for Soddy’s Hexlet . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35
2.3 Descartes’ Circle Theorem. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37
2.4 Soddy’s Formula . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38
2.5 Linked 4-Cycle Pairs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40
2.6 Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42
2.7 Notes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43
2.7.1 Notes on Descartes’ Circle Theorem and
Soddy’s Formula . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43
2.7.2 Notes on Apollonian Circle Packings . . . . . . . . . . . . . . . . . . . . . 44
2.7.3 Notes on Apollonius’ Touching Problem . . . . . . . . . . . . . . . . . . 46

xi
xii Contents

3 Graphs and Circle-Systems. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49

3.1 Graphs and Planar Graphs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49
3.2 Quadrangulations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 54
3.3 Orthogonal-Circle Representations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 56
3.4 3-Connected Graphs and Radial Graphs . . . . . . . . . . . . . . . . . . . . . . . . . . . . 63
3.5 The Coin Graph Theorem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 65
3.6 The Theorem of Steinitz . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 67
3.7 Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 68
3.8 Notes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 69
3.8.1 Notes on the Coin Graph Theorem . . . . . . . . . . . . . . . . . . . . . . . . 69
3.8.2 Notes on the Theorem of Steinitz . . . . . . . . . . . . . . . . . . . . . . . . . . 71
4 Spherical Geometry I . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 75
4.1 Geodesic Segments. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 75
4.2 Cylindrical Projections . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 77
4.3 Spherical Polygons . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 80
4.4 The Inscribed Angle Theorem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 83
4.5 The Polar Set . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 85
4.6 Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 88
4.7 Notes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 88
4.7.1 Notes on Spherical Geometry . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 88
4.7.2 Notes on Cylindrical Projections and Related Topics. . . . . 90
4.7.3 Notes on Spherical Polygons. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 94
5 Spherical Geometry II . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 99
5.1 The Cesàro Triangle . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 99
5.2 Edge-Lengths of Cesàro Triangles . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 101
5.3 Spherical Cosine Law and Sine Law . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 103
5.4 The Triangle Comparison Theorem for Spheres . . . . . . . . . . . . . . . . . . . . 105
5.5 Triangles with Two Fixed Edge-Lengths . . . . . . . . . . . . . . . . . . . . . . . . . . . . 109
5.6 The Isoperimetric Theorem for Quadrilaterals. . . . . . . . . . . . . . . . . . . . . . 111
5.7 Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 113
5.8 Notes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 114
5.8.1 Notes on Spherical Trigonometry and the
Respective Cosine and Sine Laws . . . . . . . . . . . . . . . . . . . . . . . . . 114
5.8.2 Notes on Triangle Comparison Theorems . . . . . . . . . . . . . . . . . 118
6 The Problem of Thirteen Balls . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 123
6.1 The Lemma on Proper Diagonals . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 123
6.2 Major Triangles. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 125
6.3 On a Triangulation of a Quadrilateral . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 126
6.4 The Problem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 128
6.5 Solution . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 129
6.6 Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 135
6.7 Notes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 135
6.7.1 Notes on the Problem of Thirteen Balls . . . . . . . . . . . . . . . . . . . 135
Contents xiii

7 Spherical Geometry III . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 143

7.1 Spherical Ellipses . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 143
7.2 Lexell’s Theorem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 145
7.3 Equilateral Spherical Triangles . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 146
7.4 Spherical Polygons Inscribed in a Cap . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 147
7.5 Regular Spherical Polygons . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 150
7.6 Appendix . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 151
7.6.1 Orthogonal Projections of a Spherical Ellipse . . . . . . . . . . . . 151
7.6.2 Proof of Lemma 7.1 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 153
7.7 Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 154
7.8 Notes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 154
7.8.1 Notes on Spherical Ellipses . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 155
7.8.2 Notes on Lexell’s Theorem. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 156
8 Geometric Probability on the Sphere. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 159
8.1 Random Points on a Sphere. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 159
8.2 Random Spherical Triangles . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 160
8.3 Random Quartets . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 162
8.4 Wendel’s Theorem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 164
8.5 Crofton’s Formula . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 165
8.6 Random Points on a Hemisphere . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 167
8.7 Santaló’s Chord Theorem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 169
8.8 Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 170
8.9 Notes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 170
8.9.1 Notes on Random Points and Triangles on the Sphere . . . 170
8.9.2 Notes on Crofton’s Formulae . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 175
9 Intersection Graphs of Spherical Caps. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 181
9.1 Connectivity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 181
9.2 On the Intersection of a Few Equal Caps . . . . . . . . . . . . . . . . . . . . . . . . . . . 183
9.3 Intersection Graphs of Equal Caps . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 186
9.4 Random Spherical Caps. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 190
9.5 Chebyshev’s Inequality and a Lemma. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 192
9.6 Asymptotic Probability . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 193
9.7 Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 195
9.8 Notes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 196
9.8.1 Notes on Intersection Graphs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 196
9.8.2 Notes on Chebyshev’s Inequality . . . . . . . . . . . . . . . . . . . . . . . . . . 198
10 Quartets on a Sphere . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 201
10.1 Metric Spaces. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 201
10.2 The Vertices of a Convex Quadrilateral . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 203
10.3 Unispherical Quartets . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 205
10.4 An Equation for the Radius . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 206
10.5 A Planar Quartet That Is also Spherical . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 210
10.6 Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 211
xiv Contents

10.7 Notes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 211

10.7.1 Notes on Four-Point Systems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 212
11 Higher Dimensions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 213
11.1 Figures in High Dimensions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 213
11.2 Moser’s Paradox . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 215
11.3 The Volume of a Ball. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 216
11.4 Two Lemmas . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 220
11.5 Kissing a Small Ball . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 223
11.6 Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 226
11.7 Notes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 226
11.7.1 Notes on Cubes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 226
11.7.2 Notes on Simplices . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 230
11.7.3 Notes on the Volume of (Unit) Balls. . . . . . . . . . . . . . . . . . . . . . . 236
12 The Cayley-Menger Determinant. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 239
12.1 The Cayley-Menger Matrix . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 239
12.2 Odd Integral Distances . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 242
12.3 Orthogonal-Sphere Systems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 243
12.4 Tangent-Sphere Systems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 244
12.5 A Generalization of Ptolemy’s Theorem . . . . . . . . . . . . . . . . . . . . . . . . . . . . 248
12.6 Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 250
12.7 Notes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 250
12.7.1 Notes on Cayley-Menger Determinants . . . . . . . . . . . . . . . . . . . 251
13 Casey’s Theorem. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 261
13.1 Bicolored Sets of Circles. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 261
13.2 Inversion Invariants . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 263
13.3 Isometric Radii Change . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 266
13.4 Proof of Casey’s Theorem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 269
13.5 A Theorem on Five Circles . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 269
13.6 An Extension of Casey’s Theorem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 271
13.7 Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 272
13.8 Notes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 273
13.8.1 Notes on Ptolemy’s Theorem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 273
13.8.2 Notes on Casey’s Theorem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 276
14 Solutions to the Selected Exercises. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 279

Bibliography . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 291
Index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 325
Chapter 1
Inversion and Stereographic Projection

Mathematics is not about numbers, equations, computations, or

algorithms: it is about understanding.
— William Paul Thurston

An inversion with respect to a circle in .R2 (or with respect to a sphere in .R3 ) is a
one-to-one transformation of .R2 (or .R3 ) that interchanges the inside of the circle
(sphere) and the outside of the circle (sphere). Inversions have nice properties,
and they turn out to be convenient when we study arrangements of circles, or
spheres. According to Coxeter [182], the inversion was invented by L. J. Magnus
in 1831. Using inversions, Steiner’s porism and Soddy’s hexlet are introduced. The
stereographic projection of a sphere is also obtained by restricting the domain of an
inversion.

1.1 Inversion of the Plane

Let C be a circle in the plane .R2 with radius r and center .p. Let

ϕ : R2 \ {p} → R2 \ {p}
.

be the map that sends .x ∈ R2 \ {p} to a point .x ' on the ray .−

→ satisfying
px

|px| · |px ' | = r 2 ,

.

where .|xy| denotes the length of the line segment connecting .x to .y, namely, .|xy| =
‖x − y‖. This map .ϕ is called the inversion of .R2 with respect to the circle C. Note
that .ϕ(p) is not defined. By the inversion with respect to a circle C, the inside and
the outside of C are interchanged, and each point on C is fixed. The center .p of C is
called the center of the inversion .ϕ and the squared radius of C is called the power
of the inversion. If we apply .ϕ again, then each point goes back to the original point,

© The Author(s), under exclusive license to Springer Nature Switzerland AG 2024 1

H. Maehara, H. Martini, Circles, Spheres and Spherical Geometry,
Birkhäuser Advanced Texts Basler Lehrbücher,
https://doi.org/10.1007/978-3-031-62776-7_1
2 1 Inversion and Stereographic Projection

Fig. 1.1 .Δpqx ∼ Δpyq ' l.

...
..
...
..
..
... .
.........
y ..
... .................
.. ........
.........
.......
....●
......... ..
......... ...
.............
K ...
................ x
.............................
.......................
........ .... ..
...
..
...... .....
.....●....... ...
..... ............ ........ ..
.... ...........
. ... ... ...
.
. ..
....... ... ... ..
.. .. . ......
.. ................. ..
p .●.........................................................................................●
.....
.........................................................................
...........................................●
..
..
...
...
q .
...
.
.
.
q
..
..
..
.
. ...
.... .... ..
.... . ...
...... ...... ..
......... .... ..
..................................... ..
..
..
...

that is, .ϕ(ϕ(x)) = x for every .x ∈ R2 \ {p}. Thus any inversion is an involution and
a one-to-one correspondence.
Theorem 1.1 (Circle-to-Circle Correspondence) Via an inversion .ϕ of .R2 with
center .p,
(i) every circle that passes through .p goes to a line that does not pass through .p,
(ii) every circle that does not pass through .p goes to a circle that does not pass
through .p, and
(iii) every line that passes through .p goes to itself.
Proof Let .r 2 be the power of .ϕ. (i) Let K be a circle that passes through .p, and let
'
.pq be a diameter of K. Put .q = ϕ(q). Let us show that .ϕ(K) is a line l that passes

through .q perpendicularly to the line .pq ' , see Fig. 1.1. For a point .x (/= p, q) on
'

K, let .y be the point where the line .px meets the line l. Since ./ pxq and ./ pq ' y are
right angles, .Δpxq and .Δpq ' y are similar. Hence

|pq| : |px| = |py| : |pq ' |,

.

from which we have .|px| · |py| = |pq| · |pq ' | = r 2 , and we can see that .ϕ(x) = y,
namely .ϕ(K) = l.
(ii) Now, let K be a circle that does not pass through .p. For a point .x on K,
let .y be the other intersection point of the line .px and K. By the power theorem,
' '
.|px| · |py| = c (constant). If we put .x = ϕ(x), then .|px| · |px | = r . Hence
2

|px ' |
r 2 = |px| · |px ' | = c
. .
|py|

Thus, .x ' is a point on the ray .−

→ that satisfies .|px ' | = (r 2 /c)|py|. This implies that
py
if .y moves along the circle K, then .x ' moves along the figure that is homothetic
to K with center .p and ratio .r 2 /c. Therefore, .ϕ(K) is a circle that does not pass
through .p.
1.1 Inversion of the Plane 3

(iii) is obvious. ⨆
⨅
From the above proof of (i), we have the following corollary.
Corollary 1.1 Let .ϕ be an inversion of the plane with center .p. If K is a circle that
passes through .p, then .ϕ(K) is a line that is parallel to the tangent line of K at .p.
A pair of circles (or a circle and a line) in the plane is said to cross each other if
they share two different points. If they share exactly one point, then they are tangent
to each other. By an inversion, a pair of mutually tangent circles are mapped to
either a pair of tangent circles, or a circle and its tangent line, or a pair of parallel
lines. For a pair of mutually crossing circles, the angle formed by the tangent lines
of the circles at one of the intersection points is called the crossing angle of the pair
of circles. Obviously, due to the axial symmetry of the whole figure the crossing
angles at each of the two intersection points are equal. If the crossing angle of a pair
of crossing circles .K1 , K2 is a right angle, then .K1 , K2 are said to be orthogonally
crossing to each other, denoted by .K1 ⊥ K2 . If .Ki has center .pi and radius .ri for
.i = 1, 2, then by the Pythagorean theorem we have

K1 ⊥ K2 ⇔ |p1 p 2 |2 = r12 + r22 .

.

Theorem 1.2 The crossing angle of a pair of mutually crossing circles .K1 , K2 is
invariant under any inversion. In other words, for any inversion .ϕ of the plane, the
crossing angle of .ϕ(K1 ), ϕ(K2 ) is equal to the crossing angle of .K1 , K2 .
Proof Let .q be an intersection point of .K1 , K2 , and let .li be the tangent line of .Ki
at .q for .i = 1, 2. Let .p be the center of the inversion .ϕ.
(i) The case .p = q: By Corollary 1.1, .ϕ(Ki ) is a line parallel to .li . Hence the
crossing angle of .ϕ(K1 ) and .ϕ(K2 ) is equal to the crossing angle of .l1 and .l2 .
(ii) The case .p /= q: Since .li and .Ki share only one point .q, .ϕ(li ) and .ϕ(Ki )
share only one point, that is, they are tangent. Hence the crossing angle of
.ϕ(l1 ), ϕ(l2 ) is equal to the crossing angle of .ϕ(K1 ), ϕ(K2 ).

(ii-a) Suppose that .p lies on .l1 . Then .ϕ(l1 ) = l1 , and .ϕ(l2 ) is a circle whose tangent
line at .p is parallel to .l2 . Hence the crossing angle of .ϕ(l1 ) and .ϕ(l2 ) is equal
to the crossing angle of .l1 and .l2 .
(ii-b) Suppose .p lies neither on .l1 nor on .l2 . Then .ϕ(li ) is a circle passing through
.p and its tangent line at .p is parallel to .li . Hence the crossing angle of

.ϕ(l1 ), ϕ(l2 ) is equal to the crossing angle of .l1 , l2 .

⨆
⨅
Lemma 1.1 Let .K1 , K2 be disjoint circles in .R2 with centers .q 1 , q 2 , respectively.
If .q 1 /= q 2 , then there is a circle .Γ with center on the line .q 1 q 2 such that it
orthogonally crosses both .K1 , K2 .
Proof Let .ri be the radius of .Ki , and suppose that .r1 ≥ r2 . We may put .q 1 =
(0, 0), q 2 = (a, 0), a > 0. Since .K1 , K2 are disjoint, we have either .a > r1 + r2
(when each circle lies outside the other), or .a < r1 − r2 (when .K2 lies inside .K1 ),
4 1 Inversion and Stereographic Projection

.... .... .... ......... .... .... .... ....

.. ....
.........
..............................................
....... ..
. ....
........
................. ......................... ... ...
...... Γ .... ..
.. ..
. ..
...
....... .
....... K 1
.....
.... Γ K ....... .....
.
K 1
.........
.
. .
.. ......
...
..
..... .... .. .................2 .....
.... .... ........ .... . . .... ..
. .. ..................... ...
... ........ .... ...... ...........
..... ... .
..........
...
. ....... ...
..
..
... ... .... .. .... ... ..
.... .. .. . .
..
..
.........................................................●
.
... ..
..
................................................................●
.
.
.. ... ...
.
.........................................
.............................................●
..
..
....
..
............................................................●
q 2
... ..... ....
.. .
.............................................................................●
...........●
.
..
.
............................................................................
..
..
... q 1
..
. ..
... ....
.
.
.
.
s
.
... ..
... .. q 2 .... .
.
. ..
..
...
..
q
...
.... 1
..
.
.. ... .
.
.
..
.
.
. s ....
...
... .
. .
.
.......
. ....
.
.....
.
...
.. ............................ .
. ... ...
... K .. . ..
2 ......................................... .......
..
..
..
..
.... ... .... ........ ... ....
... .......
....
..... .... ....
..... ...
........ ....... ........ ..... ...
.
........... ....... ........... ........ ..... ....
....................................... ....................................... . .... .....
.... ...
. .... .... ... .... ..
a > r1 + r 2 a < r1 − r 2 .... ......... .... .... .... .

Fig. 1.2 .Γ ⊥ Ki , i = 1, 2

see Fig. 1.2. Thus, .|a − r1 | > r2 . Let .Γ be a circle with center .s = (b, 0) and radius
r. The circle .Γ orthogonally crosses both .K1 , K2 if and only if
⌠
r 2 + r12 = b2
. . (1.1)
r 2 + r22 = (b − a)2

Let us regard this as a simultaneous equation on .b, r. From (1.1) we have .b2 − r12 =
(b − a)2 − r22 , and hence .b = (a 2 + r12 − r22 )/(2a). Since

⎛ a2 + r 2 − r 2 ⎞⎛ a 2 + r 2 − r 2 ⎞
.b2 − r12 = 1 2
− r1 1 2
+ r1
2a 2a
⎛ (a − r )2 − r 2 ⎞⎛ (a + r )2 − r 2 ⎞
1 1
= 2 2
> 0,
2a 2a
/
we have .r = b2 − r12 . Thus, the simultaneous Eq. (1.1) has a unique solution
.b, r > 0. Hence there is a circle with center on the line .pq which orthogonally

crosses both .K1 , K2 . ⨆

⨅

Theorem 1.3 For any pair of disjoint circles .K1 , K2 on the plane, there is a point
x on the line connecting the centers of .K1 , K2 , such that for any inversion .ϕ with
.

center .x, .ϕ(K1 ) and .ϕ(K2 ) are concentric circles.

Proof Let .q i be the center of .Ki . By Lemma 1.2, there is a circle .Γ that
orthogonally crosses both .K1 , K2 , with center on the line .q 1 q 2 . Let .x, y be the
intersection points of .Γ and the line .q 1 q 2 . Let C be any circle with center .x, and
.ϕ be the inversion of the plane with respect to C. By this inversion .ϕ, the line .q 1 q 2

goes to itself, and .ϕ(Γ ) becomes a line l perpendicular to .q 1 q 2 . By Theorem 1.2,

each of the circles .ϕ(K1 ), ϕ(K2 ) orthogonally crosses the lines .l = ϕ(Γ ) and .q 1 q 2 .
In this case, the center of .ϕ(Ki ) must be the intersection point of l and the line .q 1 q 2
for each .i = 1, 2. Hence .ϕ(K1 ), ϕ(K2 ) are concentric circles. ⨆
⨅
1.2 Euler’s Triangle Theorem 5

Fig. 1.3 Peaucellier linkage A ......

..............................
............. ..........
......... ... ..........
......... ........
x .........
...
. .....
......... ........
.
y
..........
..........
..........
....
.... .
.............
.... ....
y
.. .............
.. .
.... ..
Q
......
. ... ....
........
...... .
............
..
............ ...
y .. ........
.
.........
.........
.......
P .
.
..
. ... ............. y
..........
....
....
.. ....

O .....................................................
... ........................................
.. ..........
..........
...... ....
....
...
... . ................................................................................................
...
...
... z ...
..
.
x ............ ...
...

z
...
...
...
... ...
.. B
... ...
... ..
C
........
.......

Example 1.1 (Peaucellier Linkage) The Peaucellier linkage is a planar straight

line mechanism consisting of two rods (line segments) of length x and four rods of
length .y(< x) connected as in Fig. 1.3. The joint O is fixed on the plane, and the
angle formed by any two rods connected at a joint can vary freely.
In this apparatus, three points .O, P , Q lie on the line that perpendicularly bisects
AB, and since .P , Q lies on the circle with center A and radius y, we have, by the
power theorem, .|OP | · |OQ| = constant. By considering the special case when P
and Q coincide, we can see that .|OP | · |OQ| = x 2 − y 2 . Thus, when P “draws”
a figure F , then Q “draws” the figure .ϕ(F ), where .ϕ is the inversion of the plane
with center O and power .x 2 − y 2 . If we connect P to a fixed point C on the plane,
by a rod of length .z = |OC|, then, since P is restricted on a circle with center C
that passes through O, the point Q moves on a fixed line in the plane.
Concerning a planar linkage in which a vertex (link) draws a line, and a spatial
linkage in which a link draws a plane, see H. Rademacher and O. Toeplitz [653,
Chapter 16], and [397, Chapter 5] of D. Hilbert and S. Cohn-Vossen.

1.2 Euler’s Triangle Theorem

The next theorem was discovered by the Romanian mathematician G. Ţiţeica in

1908, and rediscovered by R. A. Johnson [430, p.75] in 1916, see also [806, p.125]
and, for a very recent contribution, [541].
Theorem 1.4 (The Ţiţeica-Johnson Theorem) If three mutually crossing circles
of the same radius r meet at a point, then the circle passing through the remaining
three intersection points also has radius r, see Fig. 1.4.
This theorem is also called the three-circles theorem.
Proof We may put .r = 1. Suppose that the three unit circles meet at the origin O,
and let .w1 , w 2 , w 3 be the centers of these unit circles. Then .‖w i ‖ = 1 (i = 1, 2, 3),
and the three intersection points .x, y, z other than O are given by

x = w1 + w2 , y = w2 + w3 , z = w3 + w1 .
.
6 1 Inversion and Stereographic Projection

Fig. 1.4 The Ţiţeica-Johnson

theorem
z
x

If we put .p = w1 + w2 + w3 , then we have .‖p − x‖ = ‖p − y‖ = ‖p − z‖ = 1,

and hence the unit circle with center .p passes through .x, y, z. ⨆
⨅
This theorem is used by G. Pólya [641, Chapter 10] to show how an idea occurs
to one who tries to solve a problem, and it is also listed in the book [199, Chapter 6
§3] by D. J. Davis and R. Hersh.
The three-circles theorem has many applications and extensions. A promising
direction for future research is given by its extension to normed planes. It turns out
that this theorem holds even in arbitrary such planes, where only strict convexity of
the unit circle should be demanded. This generalization has immediate connections
to orthocenters (following the important concept of isosceles orthogonality) and
to Feuerbach nine-point circles (which, in general, become six-point circles) of
triangles. These and many further related results can be found in [44], [551], and
[556]. Also in strictly convex normed planes, the four occurring circle centers plus
the four occuring intersection points give the 8 images of a parallel projection of
the vertices of some 3-cube. To increase the number of participating circles, one can
study analogous 2-projections of higher dimensional cubes, yielding planar Clifford
configurations of circles (see [552]). These are derived again for arbitrary normed
planes with strictly convex unit circles.
We come now to another theorem on circles, more precisely on the radii of in-
and circumcircles of triangles.
Theorem 1.5 (Euler’s Triangle Theorem) Let .R, r denote the radii of the
circumscribed circle and the inscribed circle, respectively, of a triangle .Δxyz in
the plane, and let d be the distance between the centers of these circles. Then
.d = R(R − 2r) holds.
2

Remark 1.1 If we replace the inscribed circle of .Δxyz by an escribed circle, then
d 2 = R(R + 2r) holds instead.
.

Proof If .d = 0, then .Δxyz is an equilateral triangle as easily seen. In this case, we

know that .R = 2r. Hence .d 2 = R(R − 2r) holds. So in the following we assume
that .d > 0. Let C and K be the inscribed circle and the circumscribed circle of
.Δxyz, and .p, q be the centers of .C, K, respectively. Let .u, v be the intersection

of the line .pq and the circumscribed circle K such that .‖u − p‖ ≤ ‖v − p‖, see
Fig. 1.5.
1.2 Euler’s Triangle Theorem 7

Fig. 1.5 .d 2 = R(R − 2r) v K

x
q z
p
C

y u

Let .ϕ be the inversion with respect to the circle C. The images .ϕ(xy), .ϕ(yz),
.ϕ(zx) of the lines .xy, yz, zx are circles of the same diameter r and they meet at
.p. Since .ϕ(K) is a circle that passes through .ϕ(x), ϕ(y), .ϕ(z), the circle .ϕ(K) has

also diameter r by Theorem 1.4. On the other hand, from

‖u − p‖ = ‖u − q‖ − ‖q − p‖ = R − d,
.

‖v − p‖ = ‖v − q‖ + ‖q − p‖ = R + d

we have .‖ϕ(u) − p‖ = r 2 /(R − d), ‖ϕ(v) − p‖ = r 2 /(R + d). Hence

r2 r2 2r 2 R
‖ϕ(u) − ϕ(v)‖ =
. − = 2 .
R−d R+d R − d2

Since .‖ϕ(u) − ϕ(v)‖ is equal to the diameter of .ϕ(K) (which is r), we have

2r 2 R
r=
. ,
R2 − d 2

from which .d 2 = R(R − 2r) follows. ⨆

⨅
To study incircles and circumcircles of triangles in normed planes is also still
a promising topic. Many related results on their geometry can be found in the
expository papers [557, Section 7] and [23].
Remark 1.2 For a tetrahedron T in .R3 , let .r, R be the radii of the inscribed sphere
of T and the circumscribed sphere of T , respectively, and d be the distance between
the centers of these spheres. Then .d 2 ≤ (R + r)(R − 3r) holds. This is known as
Grace-Danielsson inequality, see, e.g., [314]. No elementary proof of this inequality
is known so far.
The following theorem is a special case of Poncelet’s porism (cf. our Section 1.7,
and see also, e.g., [710] and [75].)
8 1 Inversion and Stereographic Projection

Theorem 1.6 Let .Δabc be a triangle in the plane, C be its inscribed circle, and
K be its circumscribed circle. If a triangle .Δxyz is circumscribed to C, and two
vertices .x, y lie on the circle K, then the remaining vertex .z also lies on K.
Proof Let .p be the center of C. We may suppose that the radius of C is 2. Let
.l, g, h be lines determined by the edges of the triangle .Δabc. By the inversion .ϕ
with respect to the circle C, .ϕ(l), ϕ(g), ϕ(h) become unit circles passing through
.p. Since .ϕ(K) is a circle passing through the three intersection points of the unit

circles .ϕ(l), ϕ(g), ϕ(h), it follows from Theorem 1.4 that .ϕ(K) is also a unit circle.
Now, since the lines .xy, yz, zx are tangent to C, their images .ϕ(xy), .ϕ(yz),
.ϕ(zx) are also unit circles, and .ϕ(xy) passes through the two points .ϕ(x), ϕ(y). By

Theorem 1.4, the circle that passes through .ϕ(x), ϕ(y), ϕ(z) is a unit circle. Note
that .ϕ(K) is a unit circle passing through .ϕ(x), ϕ(y), and it is different from the unit
circle .ϕ(xy). Since there are only two unit circles that pass through .ϕ(x), ϕ(y), the
unit circle that passes through .ϕ(x), ϕ(y), ϕ(z) must coincide with .ϕ(K). Hence
.ϕ(z) lies on .ϕ(K), and hence .z lies on K. ⨆
⨅

1.3 Steiner Cycles

Let .α, β be mutually disjoint circles in the plane. A cyclic sequence of mutually
non-crossing circles .σ1 , σ2 , . . . , σn that satisfy the condition
“each circle .σi is tangent to four circles .α, β, σi−1 , σi+1 ”

(where suffices are taken modulo n, that is, .σ0 = σn , σn+1 = σ1 ) is called a Steiner
cycle or Steiner n-cycle for .(α, β). Figure 1.6 shows a Steiner 6-cycle for .(α, β).
Remark 1.3 Generally, the assumption “mutually noncrossing” is not necessary
for defining a Steiner cycle, but to make our argument simple, we put such an
assumption.
Theorem 1.7 (Steiner’s Porism) Let .α, β be a pair of disjoint circles. If there is a
Steiner cycle for .(α, β), then for any circle .σ that is tangent to both .α, β, and that
encloses none of .α, β, there is a Steiner cycle that contains .σ as one of its members;
and every Steiner cycle for .(α, β) consists of the same number of circles.

Fig. 1.6 A Steiner 6-cycle

σ1
β σ6
α
σ2 σ5
σ3 σ 4
1.4 Inversions of 3-Space 9

Proof In the case that .α, β are concentric circles, the theorem is obvious. If .α, β are
not concentric, then there is an inversion .ϕ of the plane that maps .α, β to a pair of
concentric circles. We may suppose that .ϕ(α) lies inside .ϕ(β). If there is a Steiner
cycle for .(α, β), then it is sent by .ϕ to a Steiner cycle for .(ϕ(α), ϕ(β)). Hence
there is a Steiner cycle for .(ϕ(α), ϕ(β)), and .ϕ(σ ) is tangent to both .ϕ(α), ϕ(β),
containing none of them in its interior. Hence there is a Steiner cycle containing
.ϕ(σ ) as a member, and the number of members in any Steiner cycle for .(ϕ(α), ϕ(β))

is the same. Now, applying the inversion .ϕ again, the theorem follows. ⨆
⨅

1.4 Inversions of 3-Space

An inversion of .R3 is defined similarly to the planar case. Let S be a sphere in .R3
with center .p and radius r. The map .ϕ : R3 \ {p} → R3 \ {p} that sends a point
' −
→
.x /= p to a point .x on the ray .px such that

|px| · |px ' | = r 2

.

is called the inversion of .R3 with respect to the sphere S. The point .p is called the
center of the inversion .ϕ, and .r 2 is called the power of .ϕ. By the inversion .ϕ, the
interior and the exterior of S are interchanged, and every point of S is fixed. The
inversion .ϕ is a one-to-one correspondence.
Theorem 1.8 Under an inversion of .R3 with center .p,
(i) every sphere that passes through .p goes to a plane that does not pass through
.p,

(ii) every sphere that does not pass through .p goes to a sphere that does not pass
through .p, and
(iii) every plane that passes through .p goes to itself.
The proof is analogous to the proof of Theorem 1.1.
Similarly to Theorem 1.3, the following holds.
Theorem 1.9 For any pair of disjoint spheres .σ1 , σ2 in .R3 , there is a point .x on
the line connecting the centers of .σ1 , σ2 such that for any inversion .ϕ with center .x,
.ϕ(σ1 ) and .ϕ(σ2 ) are concentric spheres.

Proof Note that the restriction of an inversion of .R3 with center .p to a plane that
passes through .p is an inversion of the plane with center .p.
Now let H be a plane that contains the line l connecting the centers of .σi , i =
1, 2. Then .H ∩ σi , i = 1, 2 are disjoint circles in the plane H . By Theorem 1.3,
there is a point .x on the line l such that for the inversion .ϕ of .R3 with center .x, the
circles .ϕ(σi ) ∩ H, i = 1, 2, are concentric circles on H . Note also that the point .x
on l does not depend on the choice of the plane H containing l. Hence .ϕ(σ1 ), ϕ(σ2 )
are concentric spheres. ⨆
⨅
10 1 Inversion and Stereographic Projection

1.5 Soddy’s Hexlet

A cyclic sequence .σ1 , σ2 , . . . , σn of n spheres which are mutually noncrossing and

satisfy that

σi and σj are tangent ⇔ i − j ≡ ±1 (mod n),

.

is called an n-cycle of spheres. For a 3-cycle .τ1 , τ2 , τ3 of spheres, a cycle of 6

spheres

σ1 , σ2 , . . . , σ6
.

is called a hexlet (or Soddy’s hexlet) for .τ1 , τ2 , τ3 if each .σi is tangent to .τi , i =
1, 2, 3.
Theorem 1.10 (Soddy’s Hexlet) Let .τ1 , τ2 , τ3 be a 3-cycle of spheres. For any
sphere S that is tangent to these three spheres, there is a hexlet for .τ1 , τ2 , τ3 that
contains S as a member.
Proof Let .p be the contact point of .τ1 , τ2 , and let .ϕ be an inversion of .R3 with
center .p. Then .ϕ(τ1 ), ϕ(τ2 ) form a pair of parallel planes, and .ϕ(τ3 ), ϕ(S) are
mutually tangent spheres, both tangent to the pair of parallel planes. We may
suppose that .ϕ(τ1 ), ϕ(τ2 ) are the planes .z = −1, z = 1 in .R3 , respectively, and that
.ϕ(τ3 ), ϕ(S) are unit spheres with center .O = (0, 0, 0), q 1 = (2, 0, 0), respectively.

Let .q j = (2 cos j3π , 2 sin j3π , 0) for .j = 1, 2, 3, 4, 5, see Fig. 1.7. Then five unit
spheres with centers .q j , j = 1, 2, 3, 4, 5, and .ϕ(S) form together a 6-cycle of unit
spheres all tangent to .ϕ(τi ), i = 1, 2, 3. By applying .ϕ to these five spheres and
.ϕ(S), we get a hexlet for .τ1 , τ2 , τ3 that contains S as one of its members. ⨆
⨅

Fig. 1.7 Centers .q j of 5 unit

spheres on the plane .z = 0 ..
...
.. q 2 q 1
.
...
...
... ...
... ● .●..... ...
... ... ... ...
... .. .. ...
... ... .... ......... .... ...... .... .... ......... .... ..
. .. .. ...
... .... .... ... ... . ..
. . ...
... ... ... .. ...
... . .. .. . . .
. ..
...
.
........... .... .... ........... .... .... ......... .. ....
...
... q 3 ● ... ● . ● .. ...
...
... O
..
...
.. .. .
....
.. ..
... .... ... ϕ(S) ....
..
.
. ...
...
... .. .... ......... .... . ..
. . .... ......... .... . ... ...
... ...
... ...
... ...
... ● ● ...
...
...
...
q 4 q 5
...
...
...
... ...
z=0 ...
...
........................................................................................................................................................................................
...
..
1.6 Stereographic Projection 11

1.6 Stereographic Projection

Let .S be the sphere of radius r centered at the origin O in .R3 , and let .p ∈ S, and
∗ ∗
.p be its antipodal point (that is, the line segment .pp is a diameter of .S). Let H be
∗
the plane in .R that is tangent to .S at .p . For .x ∈ S \ {p}, let .x̄ be the intersection
3

point of the line .px and the plane H . Then the map

ϕ : S \ {p} → H ; x ⍿→ x̄
.

is called the stereographic projection of .S from .p, see Fig. 1.8. The stereographic
projection .ϕ is a bijection from .S \ {p} to H . As observed from Fig. 1.8, .Δpxp∗ ∼
Δpp∗ x̄ (similar). Hence .|px| : |pp∗ | = |pp∗ | : |px̄|, and

|px| · |px̄| = |pp∗ |2 = (2r)2 .

. (1.2)

Thus, the stereographic projection .ϕ is the restriction of the inversion of .R3 with
center .p and power .(2r)2 to .S \ {p}.
Since a circle on .S is the intersection of .S and a sphere (or a plane) in .R3 , it
is mapped by the inversion to the intersection of H and a sphere (or a plane) by
Theorem 1.8. Hence the next theorem follows.
Theorem 1.11 (Circle-to-Circle Correspondence) The stereographic projection
.ϕ : S \ {p} → H is a bijection. If K is a circle on .S that passes through .p, then
.ϕ(K) is a line on H . If K is a circle that does not pass through .p, then .ϕ(K) is a

circle on H . ⨆
⨅
By a tangent line l of a circle K at a point .x ∈ K, we mean a tangent line of K
at .x in the plane containing K. If K lies on the sphere .S, then this tangent line l and
.S intersect in one point, and hence l is also a tangent line of .S at .x. If a circle K on
∗ ∗
.S passes through .p and .p /∈ K, then the tangent line of K at .p is contained in the

Fig. 1.8 Stereographic

projection
12 1 Inversion and Stereographic Projection

plane H , and since l and .ϕ(K) share only one point .p∗ , l is also a tangent line of
.ϕ(K) at .p¯∗ = p .
∗

For a pair of intersecting circles in the plane, the angle formed by the tangent
lines of the two circles at an intersection point is called the crossing angle of
the circles. Similarly, for a pair of circles on a sphere .S, their crossing angle is
defined as the angle formed by the two tangent lines at an intersection point. If two
circles are tangent, then their crossing angle is 0. When two circles cross each other,
their crossing angle at one intersection point and the crossing angle at the other
intersection point are equal.
Theorem 1.12 (Invariance of Angles) The stereographic projection .ϕ : S\{p} →
H preserves the size of angles; that is, the crossing angle of two intersecting circles
on .S is equal to the crossing angle of their images on H .
Proof Let .K1 , K2 be two circles on .S intersecting at .q ∈ S \ {p}. Note that for each
.i = 1, 2, there is a circle .Ki' that passes through .p and is tangent to .Ki at .q. Then
the crossing angle of .K1 , K2 is equal to the crossing angle of .K1' , K2' . Note also that
'
.ϕ(Ki ) and .ϕ(K ) are tangent at .ϕ(q). Hence, the crossing angle of .ϕ(K1 ), ϕ(K2 )
i
is equal to the crossing angle of .ϕ(K1' ), ϕ(K2' ). Therefore, it is enough to show that
for two intersecting lines on H , the angle formed by them is equal to the crossing
angle formed by their correponding circles (the inverse images of the lines) on .S.
Let .l1 , l2 be two crossing lines on H , and let .K1 , K2 be the corresponding circles on
.S, respectively. Then, .Ki is the intersection of .S and the plane determined by .p and
' '
.li . Let .l be the line through .p and parallel to .li . Then .l is the tangent line of .Ki .
i i
Hence the crossing angle of .K1 , K2 is equal to the angle formed by .l1' , l2' , which is
equal to the angle formed by .l1 , l2 . ⨆
⨅

1.7 Appendix: Poncelet’s Porism

The following theorem is called Poncelet’s porism for circles.

Theorem 1.13 Let .C, K be disjoint circles in the plane such that C lies inside K.
Suppose that there is an n-gon .Γ satisfying that
(i) all its edges are tangent to C and
(ii) all its vertices lie on K.

Then, for any point .P1 on K, there is an n-gon that contains .P1 as a vertex and also
satisfies (i) and (ii).
Proof The following proof is based on [710]. Let .R, r be the radii of .K, C, and
−−→
.O, Q be their centers, respectively. Let A be the intersection of the ray .QO and
−→
K, see Fig. 1.9. Let us use polar coordinates with pole O and polar axis .OA. For a
point .X = (R, θ ) ∈ K, let .Y = (R, τ ) ∈ K be the point such that .θ < τ < θ + π
and XY is tangent to C. Then the argument .τ of Y is a continuous function of .θ :
.τ = f (θ ).
1.7 Appendix: Poncelet’s Porism 13

Fig. 1.9 X
.θ= / AOX, τ = / AOY
X
N
Y M
Y
A
Q a O R

' ' '

dθ = f (θ ). Let .X = (R, θ + Δθ ) and .Y = (R, τ + Δτ ) be a
First, let us find . dτ
' '
point such that .X Y is tangent to C. Let M be the contact point of XY and C, and
N be the intersection point of XY and .X' Y ' . Then .ΔXN X' and .ΔY ' NY are similar.
⎛

'
(
|Y Y | |N Y |
Hence . |X ' X| = |N X' | . Since .XX' = RΔθ, Y Y ' = RΔτ , we have
(

dτ Δτ YY' |Y Y ' | |NY | |MY |

. = lim = lim = lim = lim = .
dθ Δθ→0 Δθ Δθ→0 |X ' X| Δθ→0 |NX ' | |MX|
⎛

Δθ→0
XX '

|MY |
Thus, .f ' (θ ) = dτ
dθ = |MX| . Let us compute .|MX|, |MY |. Put .a = |OQ|. Since
./ XOQ = π − θ , we have

|QX|2 = R 2 + a 2 − 2Ra cos(π − θ ) = R 2 + a 2 + 2Ra cos θ.

.

Similarly, we have .|QY |2 = R 2 + a 2 + 2Ra cos τ . Hence

|NX|2 = |QX|2 − r 2 = R 2 − r 2 + a 2 + 2Ra cos θ

.

|NY |2 = |QY |2 − r 2 = R 2 − r 2 + a 2 + 2Ra cos τ.

//
Put .G(t) = 1 1 + R2Ra cos t
2 −r 2 +a 2 . Note here that .G(t) is a positive periodic function

with period .2π defined in .−∞ < t < ∞. Using this function, we get
√ √
|NX| =
. R 2 − r 2 + a 2 /G(θ ), |NY | = R 2 − r 2 + a 2 /G(τ ).
|N Y |
Hence we have .f ' (θ ) = |N X| = G(θ)
G(τ ) . Since .τ = f (θ ), we can write this in the
following way:

G(f (θ ))f ' (θ ) = G(θ ).

. (1.3)
14 1 Inversion and Stereographic Projection

Define .J (θ ) by
⎰ f (θ)
J (θ ) =
. G(t)dt.
θ

If .F (t) is the primitive function of .G(t), then .J (θ ) = F (f (θ )) − F (θ ). Hence

dJ (θ )
. = F ' (f (θ )) f ' (θ ) − F ' (θ ) = G(f (θ ))f ' (θ ) − G(θ ) = 0
dθ

by (1.3), and hence .J (θ ) = ω, a constant independent from .θ . Thus,

for any .θ, the line segment connecting the points .(R, θ) and .(R, f (θ)) is tangent to C, and
⎰ f (θ )
.J (θ)= θ G(t)dt = ω.

Now, let .X1 , X2 , . . . , Xn be the vertices of the n-gon .Γ in anti-clockwise order, and
let .Xi = (R, θi ), i = 1, 2, . . . , n. Then, .f (θi ) = θi+1 , i = 1, 2, . . . , n − 1. Since
./ X1 OX2 + / X2 OX3 + · · · + / Xn OX1 = 2π , we have

. (θ2 − θ1 ) + (θ3 − θ2 ) + · · · + (θn − θn−1 ) + (f (θn ) − θn ) = 2π.

Hence .f (θn ) = θ1 + 2π , and therefore

⎰ θ1 +2π
. G(t)dt = J (θ1 ) + J (θ2 ) + · · · + J (θn ) = nω.
θ1

Since .G(t) is a periodic function with period .2π ,

⎰ θ+2π
. G(t)dt = nω for every θ. (1.4)
θ

k
◜ ◞◟ ◝
Let .P1 = (R, η), and denote the iterated composition .f ◦ f ◦ · · · ◦ f of the function
f by .f k . Then
⎰ f n (η)
. G(t)dt = J (η) + J (f (η)) + J (f 2 (η)) + · · · + J (f n−1 (η)) = nω.
η

Since .G(t) is a positive function, (1.4) implies that .f n (η) = η + 2π. Therefore, the
point .(R, f n (η)) coincides with .P1 = (R, η), and hence the n-gon with vertices

P1 = (R, η), P2 = (R, f (η)), P3 = (R, f 2 (η)), . . . , Pn = (R, f n−1 (η))

.

satisfies (i) and (ii). ⨆

⨅
1.9 Notes 15

1.8 Exercises

1.1 Let ϕ be an inversion of the plane R2 with center O. For a point q /= O, let C
be a circle that passes through q and ϕ(q). Prove that ϕ(C) = C.
1.2 Prove that by an inversion of the plane, a pair of mutually tangent circles is
sent to either (i) a pair of mutually tangent circles, or (ii) a circle and its tangent
line, or (iii) a pair of parallel lines.
1.3 Let A, B be two circles in the plane. Prove that there is an inversion of the
plane that sends A, B to circles of the same size. (Consider mutually tangent
two circles that are also tangent to both A and B.)
1.4∗ Suppose that A, B are disjoint circles in the plane, and the circle B lies in the
interior of the circle A. Prove that the locus of the centers of such circles that
are internally tangent to A and externally tangent to B is an ellipse.
1.5 Let σ1 , σ2 , σ3 , σ4 be mutually tangent four spheres in R3 , and suppose
that their contact points are all different. Prove that the number of spheres
(including planes) that are tangent to all these four spheres is exactly two.
1.6 Prove that for each of the following conditions (1), (2), (3), there is a hexlet
for a cycle of three spheres that satisfies the condition.
(1) Just one element in the hexlet is a plane.
(2) Two elements in the hexlet are planes.
(3) One element (sphere) in the hexlet encloses all other five elements in the
hexlet.
1.7∗ Is it possible to draw more than five circles in the plane so that each circle is
tangent to exactly five other circles?

1.9 Notes

Due to the existing literature, the following notes are very extensive. We report on
main themes from Chap. 1, which are: inversion in the plane and in 3-space, Euler’s
triangle theorem, Steiner chains and Steiner’s porism, Soddy’s famous hexlet, the
stereographic projection and, as an appendix, also Poncelet’s porism.

1.9.1 Notes on Inversion in the Plane and in 3-Space

As a non-linear transformation in planar elementary geometry, the inversion in

circles yields many elegant applications and insights, and the study of the inversive
plane and higher dimensional analogues can create demanding geometric exercises.
Regarding the history (at least with the background of non-Euclidean geometries),
we refer to Rosenfeld’s monograph [676]: e.g., the contributions of Apollonius are
16 1 Inversion and Stereographic Projection

given there on pp. 114–116, and various further contributors are discussed in other
book parts.
We give here a selected family of basic books and booklets that deal completely,
or in form of whole chapters, in a comprehensive way with this interesting mapping
and its properties and applications. These references reflect different viewpoints,
motivations, approaches, and applications, namely referring to axiomatic and
elementary geometry (giving aid for solving related problems from circle geometry,
like the Apollonian one, and presenting theorems important in foundations of
geometry), inversive planes and spaces, theory of circular tufts and conformal
geometry, projections (e.g., the stereographic one), investigations around Möbius
transforms, complex numbers, (models of various) non-Euclidean geometries and,
last not least, teaching aspects. As common features of all these selected references,
one obviously sees a broad and geometric introduction of the inversion itself, and
that at least one of the mentioned aspects and application fields is discussed there.
Clearly, following all these criteria, the selection given now cannot be complete, but
we try our best.

Books and Book Chapters

Some older references (mainly from the first half of the twentieth century) are
Hadamard [353, Chapters V, VI and VII of the Complements to Book III], the first
chapter of Coolidge’s basic book [177], Johnson’s widely cited monograph [430,
pp. 43–57 and pp. 96–109], the famous book [179, Part II of Chapter 3] of Courant
and Robbins, and the booklet [698] of Schmidt.
Altshiller-Court contributed with two related books: in Chapter 9 of [28] he
introduces inversion, and there are also interesting applications; and in [29, Chapter
VIII] he discusses comprehensively the three-dimensional version. We mention here
the paper [92] in which problems around inversion are extended to 3-space, see also
Section 5.8 in Borceux [103]. Later the following related references came to be:
Yaglom [822, Chapters 1 to 4 and the supplement], Pedoe’s booklet [627, Chapter
I], the monograph [369, §39 and §40] due to Hajós, Coxeter’s classical book [182,
Chapter 6], Schwerdtfeger’s monograph [721, §2 and Chapter II], the chapters
on inversion and projection in Perfect’s book [630], Bakelman’s booklet [52],
Guggenheimer’s monograph [344, Chapters 8 and 9], Coxeter and Greitzer [186,
Chapter 5], especially (and not only) Chapters II and VI of Pedoe’s comprehensive
book [625], Greenberg’s excursion [322, Chapters 7 and 9] into non-Euclidean
geometries, and Ewald’s basic monograph [251, Chapter 6].
Further on, we should cite Giering’s book [299, Chapters 18, 19, and 20],
Section 9.5 and great parts of Chapter 10 in Berger’s comprehensive double
monograph [75] (see also Chapter II of his more recent book [78] where also
inversion in space is explicitly discussed), Chapter 3 (from 3.6 on) of Eves [250],
Chapter 5 on inversive geometry in [124], Hartshorne’s basic book [382, Section 37]
and applications in the sections following there, and the book [91] of Blair in which
also Liouville’s theorem on conformal maps and the question of when a convex body
1.9 Notes 17

remains convex under a sphere inversion are comprehensively discussed (see also
[92]). Furthermore, [103, Sections 5.7 to 5.11] of Borceux should be mentioned,
particularly Section 5.8 where spatial inversion is studied. Chapter 4 as well as
further parts with applications in Aumann’s book [48] should be added; this book
broadly discusses circle geometry, like also the book [644], where inversion and
applications of it yield the topic of Chap. 7.
In addition we refer to the broad representations [822], [823], and [820] of
Yaglom from the encyclopedic collection [17]. Also the surveys [811] of Wilker and
[665] of Reventós Tarrida treating inversive geometry should be mentioned here.

A Few Applications

Finally we want to mention two interesting applications of inversion: mechanical

linkages and fractals.
Peaucellier’s inversor, a mechanical instrument invented in 1873 to draw the
inverse of a given curve, is explained in many important books; we mention
here Johnson’s monograph [430, Sections 67 to 72], the famous book [397, §40]
of Hilbert and Cohn-Vossen, Schmidt’s booklet [698, Chapter VI], Coxeter and
Greitzer [186, Section 5.3], the end of Section 38 in the book [250] of Eves, and
Aumann [48, Section 4.4]. The mechanical models described in the classical book
[193] include two- and three-dimensional linkages, among them also the Peaucellier
linkage. The spherical analogue of Peaucellier’s inversor was studied by Goldberg
[307], and Reed [661] used the linkage to prove that certain oblique sections of a
torus are circular. In some of the references above also the related inversor of Hart
is presented.
It is well-known that inversion in circles is also important for the construction
of fractals, see Mandelbrot’s basic monograph [543]. We mention some further
results in this direction. Clancy and Frame [165] study fractal limit sets of groups of
inversions in the plane, and Gdawiec [296] extends, in view of related infinite circle
inversion fractals, circles to star-shaped sets. Similarly, Ramiŕez et al. [654] extend
the used circle inversions to inversions in unit circles of certain normed planes
and apply this extended transformation to famous fractals, such as the Sierpinski
triangle and the Fibonacci fractal. Boreland and Kunze [104] present mathematical
foundations to fractal constructions based on circle inversions, showing also a strong
connection to the theory of iterated function systems.

1.9.2 Notes on Euler’s Triangle Theorem

Euler’s triangle theorem says that for any triangle the radius R of the circumcircle,
the radius r of the incircle and the distance d between the centers of these two circles
satisfy the equality .d 2 = R(R − 2r). This theorem implies the Euler inequality
.R ≥ 2r, with equality iff the considered triangle is equilateral. Euler’s theorem
18 1 Inversion and Stereographic Projection

was obtained already by Chapple in 1746 (see [156, page 123]) and is certainly
much older. There are many books and articles where Euler’s triangle theorem
and/or his inequality are discussed or applied to get further inequalities. As examples
we mention Johnson’s book [430, p. 186], Altshiller-Court [29, Chapter 3, §152],
as well as Alsina and Nelsen [24, p. 56]. In Chapter 6 of the monograph [723]
by H. Sedrakyan and N. Sedrakyan problems and inequalities are selected which
have to do with circle radii. Among them one can also find a discussion of the
circum-inradius inequality for triangles. And in the glossary of the comprehensive
book [228], where almost 50 terms, theorems, formulas, functions, equations, and
techniques are collected, today commonly associated with Euler, also the theorem
discussed here is listed (see page 300 there).
For all triangles inscribed in a circle, Pambuccian and Schacht [613] give a
synthetic proof of Euler’s inequality for Hilbert’s absolute geometry, characteriz-
ing equilateral triangles by the equality case. And Carlitz [141] collects several
inequalities which are particularly related to Euler’s inequality. An additional
characterization of equilateral triangles is given by their area.
And in [764] a proof of Euler’s inequality in the hyperbolic and spherical case
is given.

1.9.3 Notes on Steiner Chains and Steiner’s Porism

A finite sequence of circles each being tangent to two fixed and non-intersecting
circles and to two further circles of that sequence is called a Steiner chain of circles
or a Steiner cycle. The so-called Steiner’s porism can also be formulated as follows:
if two such given circles allow such a chain, then they even admit an infinite number,
and each of these chains contains the same number of circles. In addition, any circle
tangent to the two given circles and surrounding either none or both of them is a
member of such a chain.

Books and Book Chapters

Often in combination with an interesting variety of related results on circles and

also on spheres in higher dimensions (e.g., the corresponding use of inversion,
the arbelos, and Soddy’s hexlet), theorems about Steiner chains and, in particular,
Steiner’s porism can be found in the books by Coolidge [177, pp. 31–37], Johnson
[430, Chapter VI], Coxeter and Greitzer [186, pp. 123–126, 175–176, 180], Pedoe
[625, Section 25], Eves [250, pp. 134–135], Honsberger [401, Chapter 6], Ogilvy
[601, pp. 51–54], Wells [806, pp. 244–245], Barnes [60, Chapter 7], and Aumann
[48, Chapter 9]. Also at the end of Chapter 1 of Yaglom’s book [822] Steiner
chains and related configurations are discussed. In addition we mention here the
book [644]; in its Chapter 7 Steiner chains and the Apollonius problem are nicely
presented.
1.9 Notes 19

Steiner chains and Soddy’s hexlet belong to the famous objects that were also
found by the Japanese Wasan masters, predating their discovery in the West. More
precisely, Steiner chains were studied already in the Japanese temple geometry
(Wasan) by Ajima Naonobu (1732–1798), i.e., earlier than by Jakob Steiner (1796–
1863). Ozone [608] studies Steiner chains as given in Wasan, and he discusses
elegant ways to study and prove them. Regarding the occurence of these two notions
in the Japanese temple geometry, we also cite here Fukagawa and Pedoe [284,
Chapters 1 and 9]. The interested reader might also look at Akopyan’s nice booklet
[13, Section 6.8] “without words, but figures”.

Some Research Results

Besides these representations, the following articles might be additionally quoted.

Generalizing former results, Iwata [424] gives an analytic condition for getting a
configuration of hyperspheres in n-space presenting an analogue of Steiner chains.
In the two papers [669] and [137] three-dimensional analogues of Steiner’s porism
are given. In the first paper, a computer algebra system is used, and in the second
contribution the analogue of Steiner’s idea is slightly generalized, showing also a
surprising connection to Platonic solids. In the expository paper [665] of Reventós
Tarrida Steiner’s porism is elegantly derived via inversion.
Yiu [827] studies rational Steiner chains in which R and r (the radius of the
larger resp. smaller given circle) as well as d (the distance between the midpoints
of both circles) are given by rational numbers. It is easy to see that rational Steiner
m-chains exist only for .m = 3, 4, 6. Using previous results, the author presents
a parametrization for all triples .(R, r, d) with .R = 1 corresponding to a rational
Steiner chain. Protasov [649] gives a nice discussion of Steiner’s porism, and Baralić
et al. [56] extend the notion of Steiner chain beginning with more than two circles.
Also for this situation they find an analogue to the Steiner porism, and they pose a
couple of interesting open questions.
In [524] Maehara presents interesting bend formulas for Steiner chains of circles
and extends this also to sphere systems in higher dimensions.
Musin [591] studies configurations of hyperspheres in n-dimensional space,
getting far-reaching high-dimensional extensions of classical results concerning
Steiner’s porism (in the plane) and Soddy’s hexlet (in 3-space). To obtain these
results, he uses kissing arrangements of hyperspheres, corresponding packings and
spherical codes.
For an odd prime power q, Hungerbühler and Kusejko [417] study Steiner’s
porism in finite Miquelian Möbius planes constructed over the pair of finite fields
GF(q) and GF(.q 2 ). The authors derive properties of common tangent circles for two
given concentric circles and obtain a finite version of Steiner’s porism for concentric
circles. Using the quadratic residue theorem in GF(q), conditions on the length of
a Steiner chain are derived. Also extensions to arbitrary pairs of non-intersecting
circles are given.
20 1 Inversion and Stereographic Projection

The paper [719] of Schwartz and Tabachnikov deals with the 1-parameter family
of Steiner chains of cardinality k, say, as they are described by Steiner’s porism
for two fixed given circles. E.g., they ask: what do the members of such a Steiner
chain of cardinality k have in common? They show that the first .k − 1 moments
of their curvatures remain constant within a 1-parameter family (the case .k = 3 is
established by the circle theorem of Descartes). Using also spherical designs, first a
more general result is proved, and from this the main results on Steiner chains are
derived. In addition, the authors succeed in extending one of their main results on
Steiner chains to spherical and hyperbolic geometries.
Motivated by invesitigations around Poncelet’s porism, Heidari and Honari [390]
derive an extension of Steiner’s porism to n-dimensional space. Steiner chains for n
given spheres are introduced, and by the three-inversion theorem it is shown that
there exists a 1-parameter subgroup of Möbius transformations which preserves
these n spheres. The authors obtain a corresponding 1-parameter family of Steiner
chains by applying the elements of this subgroup to a Steiner chain.

1.9.4 Notes on Soddy’s Hexlet

The famous hexlet of Soddy is a special three-dimensional analogue of the Steiner

porism, and it was perhaps first observed in the Japanese temple geometry (see [284,
Chapter 9] (Sangaku tablets from 1822 in the Kanagawa prefecture). Soddy, who
received the Nobel prize in 1921 for discovering isotopes, rediscovered it (cf. [747]).

Books and Book Chapters

As classical geometric configuration of spheres, Soddy’s hexlet is more detailed

mentioned or studied in various geometry books, namely those of Coxeter [182,
Section 1.59], Pedoe [625, Section 40.3], Honsberger [401, Chapter 6], Ogilvy
[601, Sections 5.3 and 5.4], Eves [250, p. 152], Barnes [60, Chapter 7], as well
as Fukagawa and Pedoe [284, Chapter 9]. In the geometry dictionary [806], Soddys
hexlet is discussed on page 232.

Some Research Results

The envelope of a varying sphere touching three fixed spheres is a Dupin cyclide.
Coxeter [181] studies two families of spheres which touch any three members of the
other family and two from its own family, thus forming rings as it is the case with
Soddy’s hexlet. He considers the case when the cyclide is of such a shape that there
is a closed ring of m spheres, every of them touching the next. Each sphere of the
same family belongs to such a ring if there is one ring. And if this one goes round the
1.9 Notes 21

cyclide d times, then it is called a p-ring with .p = m/d. Using stereographic projec-
tions of inspheres of cells of certain 4-polytopes, Coxeter proves the following nice
result of Steiner: if one considered family has p-rings, then the other has q-rings,
with the equality .1/p + 1/q = 1/2. Soddy’s hexlet is the subcase .p = 3, q = 6.
In [811] Wilker developed an elegant method to represent n-balls in n-
dimensional inversive space. He applies this method in [812] to study Soddy’s
hexlet, e.g. numerical relations between curvatures of participating spheres. He
shows, for example, that the sum of the curvatures of opposite spheres of the
hexlet is equal to twice the sum of the curvatures of the original three tangent
spheres. Suitably defining curvature for the spherical and hyperbolic 3-space,
Wilker succeeds in extending his results also to these non-Euclidean cases. The
paper [669] of Roanes-Lozano was discussed already above, when referring to
Steiner chains of circles. Variants related to Soddy’s hexlet are studied. And also
Musin’s paper [591] was already presented there, due to his successful use of
spheical codes when studying analogues of Steiner’s porism and Soddy’s hexlet.
Maehara and Oshiro [542] define an m-cycle of balls to be a cyclic sequence of
m three-dimensional balls, non-overlapping, where every two consecutive balls are
tangent to each other, and an (.m, p)-link be a pair of an m-cycle and a p-cycle that
form a non-splittable link with no two balls overlapping. It is shown that a (.3, p)-
link exists only when .p ≥ 6, and in any (.3, 6)-link every ball from the 3-cycle
is tangent to all balls of the 6-cycle. Further on, a (.4, 4)-cycle exists, and in any
such cycle every ball is tangent to all balls in the other 4-cycle, too. And also in the
paper [524] variants of Steiner chains, Soddy’s hexlet and further sphere systems
are investigated.

1.9.5 Notes on the Stereographic Projection

In the classical sense, the stereographic projection is a perspective projection of the

2-sphere in 3-space, whose center of projection is the north pole of sphere, onto the
tangent plane of it through the south pole. This projection is conformal (i.e., angles
between curves are preserved; e.g., circles become circles), but it is not isometric or
equiareal.

Some History and Applications

Its history goes back to ancient Greek astronomers who projected down the celestial
sphere for studying movements of stars and planets in the plane. Important names
occuring here are Hipparchus, Apollonius, Archimedes, and Ptolemy, but still many
things about this early step are controversial. In form of the planispheric astrolabe
the Byzantine and Arabic astronomers used and developed the related ideas further,
and their knowledge reached Western Europe in the eleventh and twelfth century.
Contributions and applications due to Mercator, Harriot, and Halley followed; the
22 1 Inversion and Stereographic Projection

latter gave the first proof for the conformality. The current name “stereographic
projection” is due to d’Aguilon (1613), see [835]. Further material about the
history is collected by Rosenfeld in [676, pp. 121–130, 242, 342]. Classifying
the cartographic mapping procedures, Schröder [712] gives also a lot of historical
information, particularly also around stereographic projection. In the second edition
of [79] Berggren gives a more thorough exposition on using the stereographic
projection in the astrolabe, an important step.
Today stereographic projection is famous as an elegant tool to visualize elegantly
mathematical phenomena, often astonishing. In mathematics, this refers, e.g., to
models of non-Euclidean geometries, compactification in topology, complex analy-
sis and the Riemann sphere, projective geometry, differential geometry, arithmetic
geometry, polytope theory (Schlegel diagrams), simplifications of integrals involving
trigonometric functions and so on. In neighbouring disciplines we have it in
cartography, geodesy, crystallography, photography and optics (e.g., fisheye lenses
and mappings of spherical panoramas), geology, planetary science and so on.

Books and Book Chapters

Let us start with book chapters where stereographic projection is comprehensively

introduced and applications are shown, first within mathematics and then in a few
other fields.
Classical geometry books containing such comprehensive sections or chapters
are Klein [455, pp. 50–51 and pp. 294–300], Johnson [430, pp. 106–109], Hilbert
and Cohn-Vossen [397, §36], Altshiller-Court [28, pp. 261–262], Coxeter [182,
Section 16.2], see also the survey Rosenfeld [674, Section 4.1]. And also books
mainly devoted to other themes (like inversions) contain a lot of material on
stereographic projection; see, e.g., Schmidt’s book [698].
In the second part of Borsuk’s monograph [105] on n-dimensional analytical
geometry, stereographic projection is used to study projective and Möbius n-
dimensional space. Similar in nature is Blaschke’s book [94], where in Chap. 2 the
natural way from stereographic projection to inversions and linkages is shown.
Yaglom [822, pp. 74–78] introduces and uses stereographic projection for
proving that any circular transformation of the plane can be realized by an inversion
followed by a similarity.
Milman and Parker [575, pp. 301–305] introduce the stereographic projection for
studying mainly the Klein and Poincaré disk models. Similarly, in a classical sense
and using it for models of non-Euclidean geometry, Prasolov and Tikhomirov [646,
pp. 93–96] introduce the projection type. The latter books are nicely written also
for teaching aspects on the undergraduate level. Even more elementary is the known
book [186] of Coxeter and Greitzer, successfully integrating it into a network of
related notions.
Stereographic projection yields also an elegant way to get the formulas for
spherical trigonometry from those of planar geometry. The crystallographer and
mineralogist G. Cesàro developed this way; we refer to Donnay’s book [220] for
1.9 Notes 23

more details. For this theme we refer also to Chapter 8 of the more recent, nice book
[131] of van Brummelen.
In the long Chapter 18 of his monograph [299] Giering introduces the stereo-
graphic projection more generally, for quadrics and in the projective setting, also
with coordinate representations. The projection of spheres he treats as a subcase.
(The same repeats with inversions, in the next chapter.) On an analogous level,
Pedoe [625, pp. 387–390] studies stereographic projection, too. Here we mention
also Schwerdtfeger’s book [721] on Möbius transformations, complex numbers,
and models of non-Euclidean geometries; in Part (3) of Chap. 1 there is a nice
introduction to stereographic projection, with interesting examples.
Bigalke [87] introduces in Section 5.5 the stereographic projection, giving
various properties, and in Section 9.1 he comprehensively discusses the construction
of so-called pole figures of crystals generated by their facet normals, and how
stereographic projection and its conformality is used in that framework. Basic
constructions are carefully presented, like also the related Wulff net, and finally
crystallographic results are represented, like the corresponding point groups.
Further applications in crystallography are given.
Berger discusses in [75, Part II], more precisely in the subsections 18.1.4, 18.1.8.6,
and 18.10.2, various sides of stereographic projection, e.g.: definitions, basic
properties, cartography, and the Möbius group. In his more recent book [78] he
discusses at the beginning of Section II.5 and in Section III.2 the conformal group
and applications of the Möbius group.
Section 6.3 of the textbook [458] of Knörrer is one of the most careful intro-
ductions and representations of stereographical projections. It uses group theory
and Riemann’s number ball, and in Section 6.7 the author uses a four-dimensional
stereographic projection to study the fibers of the Hopf map.
Borceux [103, Section 5.9], introduces the main properties of stereographic
projection, and also Aumann [48, Section 7.4] gives a nice introduction to it,
enumerating main properties and supporting this by wonderful figures. A nice
presentation of properties of stereographic projection is also given in Section 5.4
of Glaeser’s book [302].

Some Research Results: Differential Geometry

The following references are given in thematic order, and within these themes
chronologically. We give first a few related publications from differential geometry.
We begin with Euler who studies in [246] properties of angle-preserving maps;
in particular, he develops the differential geometry of the stereographic projection.
In his nice book [815], Wilson starts with Euclidean geometry, switching then to
spherical and hyperbolic geometry and taking several versions of the Gauss-Bonnet
theorem as guideline; finally this theorem is also discussed for closed surfaces. At
the end of Chap. 2 he presents Möbius geometry based on stereographic projection.
The science of invisibility combines Einstein’s general relativity and Maxwell’s
principles of electromagnetism; Leonhardt and Philbin have a part in its trans-
24 1 Inversion and Stereographic Projection

formation from fiction to science, e.g. using analogies between mechanics and
optics, and the geometry of curved spaces. Their book [501] gives the mathematical
foundations, mainly differential geometry, and shows nicely how highly theoretical
concepts can become useful in electrical and optical engineering. Due to Maxwell’s
role in this book, in the “optical part” also his fisheye lens is discussed, and how it
is deduced from the stereographic projection.
Cecil and Ryan present in [149] a thorough treatment of the differential
geometry of hypersurfaces in real, complex, and quaternionic space forms. They
lay special emphasis on isoparametric and Dupin hypersurfaces in real space
forms and Hopf hypersurfaces in complex space forms. The book shows at several
places how important the stereographic projection in this framework is. After its
introduction in Chap. 2, this projection type is used in Chap. 4 regarding Möbius
geometry, isoparametric and Dupin hypersurfaces, and in Chap. 5 with respect to
the classification of taut hypersurfaces in Euclidean space. The first two chapters
have much overlap with the book [148] of the same authors, where stereographic
projection is also needed.

Cartography and Geodesy

Not being far from differential geometry, we continue with references related to
cartography and geodesy.
In the booklet [697] of Scheffers the basic knowledge from differential geometry
is presented. More precisely, the analytic derivations of projections of the sphere
onto the plane are treated, and this refers to the mercator, stereographic and equal-
area projections.
A main contribution of the classic book [466] of König and Weise is the
successful use of general complex vector coordinates for the spheroid, to apply then
a direct cartographic transformation of this interesting object such that methods
of complex function theory come into the game. In Chapter VII the stereographic
projection, the conical projection, and the general arc projection of the spheroid are
comprehensively discussed.
Although mentioned already above (regarding history), we cite here once more
Schröder’s nice little book [712]. Stereographic projection is also historically
discussed.
Feeman exposes in [255] the mathematical ideas involved in creating and
analyzing (geographical) maps. His booklet seems to be the first more recent
book in which the famous problem of mapping (portions of) the Earth in a very
understandable and, at the same time, comprehensive style. Chapter 9 deals with
conformal mappings and, of course, particularly with the stereographic projection.
1.9 Notes 25

Classical Geometry, Foundations of Geometry

We continue with references in the spirit of classical geometry (also foundations of

geometry) and non-Euclidean geometries.
Hartl [380] gives an elementary proof of the n-dimensional generalization of the
classical theorem that two complex quadrics, having a conic section in common,
have a second conic section in common. He applies this result to classical results on
the n-dimensional stereographic projection.
Benz [73] develops Möbius sphere geometry for an arbitrary real inner product
space X (i.e., a real vector space X of dimension at least 2 and not necessarily finite-
dimensional, with a fixed scalar product, positive definite). Extending X suitably, he
also introduces (in section 9) stereographic projection.
The book [667] of Richter-Gebert, wonderfully emphasizing pearls from clas-
sical projective geometry and bringing many results from various non-Euclidean
geometries and related topics, is interesting for students, mathematicians, computer
scientists, and physicists. Many of the presented results are nicely illustrated. In Part
III also Möbius transformations and the stereographic projection are treated.
Popko’s nice book [643] treats a large variety of topics based on the sphere in
Euclidean 3-space. (Platonic) polyhedra, spherical polygons, packing problems and
kissing spheres are examples. In Appendix B also stereographic projection and its
basic properties are treated.
Swart and Torrence [766] describe advances in photographic technology (e.g.,
Google Maps street view) and their mathematical background on the level of
geometric projections. All that we see is imprinted on a sphere, where we are in
its center. To get an ordinary (plane) picture, suitable projections are used. One of
them is stereographic projection, on which the article is (not only, but) mainly based.
In the papers [770] and [734] two classical results are reproved by using in a
refreshing way stereographic projection: in the first case it is the famous Apollonius
problem (on circles touching three given ones), and in the second Lexell’s theorem
(on equiareal spherical triangles with a fixed given side). The reader should note
that the used stereographic projection refers to Möbius and Laguerre planes.
Ionascu [419] deals with the analogue of the circle of Apollonius in spherical
geometry which is the pre-image via stereographic projection of an algebraic curve
of degree three. And this curve has two connected components, where each is the
reflection of the other through the center of the sphere.

Further Topics

The final four references are from different areas.

Wilker [813] uses stereographic projection for presenting the classical Hopf
fibration. In his approach the participating 3-sphere is clearly the union of two solid
tori with common boundary, that the fibers are pairwise linked circles.
In his nice, expository paper [468], Kollár presents the most important results,
problems, tools and ideas from birational geometry. So he surveys also results on
26 1 Inversion and Stereographic Projection

rational and rigid hypersurfaces, and he shows the connection with the classification
of varieties developed by Enriques and others. Explaining the rationality of quadric
hypersurfaces, he starts with the historically essential stereographic projection.
In order to calculate lasing modes and optical power flow in multisection lasers
with open boundaries, O’Connor et al. [599] introduce a formalism which is
based on a projection of the complex-valued electric field and its spatial derivative
onto a suitably extended complex plane. In single-section lasers, a laser mode is a
loxodrome on the extended complex plane, and in a multisection laser, loxodromes
for individual sections of the laser are derived. A natural visualization of such
constructions is given by the stereographic projection of the extended complex plane
onto the Riemann sphere.
The book [298] of Geiges on the importance of geometry in celestial mechanics
deals mainly with geometric interpretations of the Kepler problem. In Chap. 8 the
author treats the needed differential geometry and spherical geometry, and the role
of stereographic projection is clarified, too.

1.9.6 Notes on Poncelet’s Porism

Poncelet published in his Traité des propriétés projectives des figures (1822) the
following important theorem from projective geometry, which is called Poncelet’s
porism: If for two smooth (real) conics in the plane there exists an m-sided polygon
which is inscribed in one conic and circumscribed about the other, then there are
infinitely many such polygons, and each point of the first conic is vertex of one of
them. The interscribed m-gon is often called Poncelet polygon. It seems that the
first solution to Poncelet’s porism was proposed by W. Chapple [156], before it
became known as Poncelet’s theorem. Giving here an overview to what happened
mathematically around this theorem, we are in a comfortable situation: there exist
various excellent comprehensive surveys and even books around this theorem such
that in great parts it suffices to cite them.

Some History

The history of the porism, of variants, modifications, generalizations, and appli-

cations of it can be read in the comprehensive surveys [113, Lecture 8] by Bos,
[115] due to Bos, Kers, Oort, and Raven (this article contains also new results),
and the two papers [208] and [209] by Del Centina. Also the interesting paper [718]
(with a discovery referring to Poncelet’s prisoner-of-war time in Russia) should
be mentioned here. And we add also Martini’s survey [546]; in its Section 2.4
Poncelet’s porism is (based on older sources) updated till 1993.
1.9 Notes 27

Books and Book Chapters

There are even two books completely dedicated to Poncelet’s porism and related
topics. The first one [271] by Flatto gives two short proofs of Poncelet’s and
Cayley’s theorem on polygons inscribed in a conic and circumscribed about another
one, and this is done in the complex projective plane as well as the real affine
plane. The needed tools are clarified, i.e., a short course on analytic geometry
in the projective plane, Riemann surfaces, tori, and elliptic curves and functions
is given. The real case for two ellipses does not need these preparations. After
the proofs, relations to billiards and other topics are presented. A supplement of
Tabachnikov is added, interestingly offering another look at Poncelet’s porism and
billiards. The second book [225], due to Dragović and Radnović, is based on the
possibility to prove Poncelet’s theorem with the help of elliptic curves depending on
the representation of an elliptic curve as the double cover of the outer participating
conic, with four ramification points. (An analogous covering regarding the inner
conic is possible, too.) Based on this, in a comprehensive way the authors want
to show what current research directions are related to the framework around the
Poncelet porism (namely, dynamics of integrable billiards, algebraic geometry of
hyperelliptic Jacobians, and classical projective geometry of pencils of quadrics).
But a historical view is also given, underlining basic ideas and the way to the recent
state of the art.
Of course, Poncelet’s porism can also be found in classical geometry books,
partially with proofs. Examples here are the book II of Berger [75, Section 16.6],
that of Prasolov and Tikhomirov [646, pp. 178–184], and Berger’s more recent
book [78, Section IV.8]. Here we also mention Schoenberg’s book [710] whose
14th chapter is dedicated to Poncelet’s porism and Steiner cycles.

Expository Chapters and Articles

But Poncelet’s porism is also very popular and suitable for collections (in books
or expository papers) of really nice mathematical results from different fields of
mathematics or geometry; see, particularly, e.g. [300] and [737], or the final chapter
of the popular book [283] of Fuchs and Tabachnikov. And in Chapter 4 of the
collection [787] of lectures (meant together with its first part), Ueno et al. present
three lectures about Poncelet’s theorem. An appendix, consisting of two parts, adds
a proof using elementary plane geometry, whereas the second lecture contains an
analytic proof. Posamentier and Geretschläger [644] treat the porism in Chap. 3.

Some Research Results

We continue now with selected research results which are not mentioned in the wide
surveys cited above.
28 1 Inversion and Stereographic Projection

Ronda and Valdés [673] study the relation between Kruppa equations and
Poncelet’s porism to show via some suitable parametrization how the classical
theory of projective conics yields new insights and results on the 3D reconstruction
from two images taken with uncalibrated cameras. Depending on the relation
between Schwarzenberger bundles and Poncelet curves, Vallès [794] gives a nice
and short proof of Poncelet’s porism. Mirman [576] considers the subcase of two
given circles and uses a recurrent procedure based on a biquadratic equation to find
out whether there is an m-sided Poncelet polygon interscribed between these circles.
This procedure has interesting applications for Hankel determinants. Extensions to
given ellipses and hyperbolas are also presented.
The next results cited here are closely related to billiards. Considering a Poncelet
m-gon P , Levi and Tabachnikov [502] study the system of extensions of the
sides of P to lines, thus getting an arrangement of lines which has .m(m + 1)/2
intersection points. Schwartz showed that these points interestingly lie on a family
of conics, and the authors continue by relating Schwartz’ arguments to billiards
inside ellipses. Dragović and Radnović [224] study higher-dimensional analogues
of Poncelet-like problems by synthetic methods. Observing the billiard nature of
some classical constructions and configurations, the billiard algebra is developed.
This is a group structure on the set of lines simultaneously tangent to .n − 1 quadrics
from a given confocal family in Euclidean n-space. Successfully using this tool, the
authors obtain various results. Mozgawa [587] defines bar billiards for two nested
conics .C1 , C2 having the Poncelet porism property and proves that for any oval C,
in a suitably small neighborhood to the inner conic .C2 , there exist ovals .Cin and
.Cout inside and outside of C such that the pairings .(C, Cin ) and .(C, Cout ) have the

Poncelet property. Continuing these studies, Cieślak, Martini and Mozgawa [164]
study the rotation index of bar billiards and show that it is rational iff the two circles
(considered here instead of the conics above) have the Poncelet porism property.
Billiard knots are knots describable as non-self-intersecting periodic orbits in a
three-dimensional billiard. Pecker [623] proves that for any elliptic right cylinder
D every type of knot can be realized as the trajectory of a ball in D. To show this,
Jacobi’s proof of Poncelet’s theorem by means of elliptic functions is used.
Halbeisen and Hungerbühler [370] consider chains of conics .C0 , . . . , Cm−1 such
that for each .0 ≤ i ≤ m−2 there is a Poncelet triangle .Tj with vertices on .Cj whose
sides are tangent to .Cj +1 . They show that each chain can be closed by adding an
appropriate conic .Cm+1 and confirm the existence of particular closed chains where
the contact points of .Ti with .Ci+1 serve as vertices of the subsequent triangle .Ti+1 .
Further related interesting results are elegantly derived.
Kasner’s theorem concerns what he termed the inscribed and diagonal pentagons
constructed from a given pentagon. The extension of this theorem to Poncelet
polygons was repeatedly investigated. Tabachnikov [769] discusses connections
between Kasner’s and Poncelet’s work, showing also a link to dynamical systems
via the pentagram map. It is shown that also nowadays there are various related
research activities.
Garcia et al. [293] study loci of triangle centers over variants of two-well known
triangle porisms: the bicentric and confocal families. In a natural way this leads
1.9 Notes 29

to the general version of Poncelet’s porism. It is confirmed that the loci of certain
triangle centers and associated points remain conics and/or circles. And in [292]
the authors study one-dimensional families of Poncelet triangles “between” a circle
C (outside) and an ellipse E, calling them Brocard porisms. The corresponding
Brocard angle is invariant, and the Brocard points are stationary at the foci of E.
There is also a Brocard circle passing through the two foci of E and the midpoint of
C, and there is a second Brocard triangle whose vertices are obtained as intersections
of the symmedians with the Brocard circle. The corresponding one-dimensional new
family of triangles is inscribed to the Brocard circle and circumscribed about another
ellipse. Continuing this generation procedure, an infinite, converging sequence of
porisms created by certain nested objects is obtained. The authors derive many
geometric results about this continuous family of porisms.
Chapter 2
Bend Formulas

It is clear that the chief end of mathematical study must be to

make students think.
— John Wesley Young

If a Steiner cycle .σ1 , . . . , σn for .(α, β) is given, the following question occurs: what
can we say about the relation between the radii of the circles? Such problems were
considered in Japanese mathematics (Wasan) in the Edo-period (1600–1868), see
the booklet of H. Fukagawa and D. Pedoe [284]. In this chapter, we discuss such
relations on radii (or their reciprocals) for Steiner cycles, Soddy’s hexlet, mutually
tangent four circles in the plane (Descartes’ circle theorem), mutually tangent five
spheres in .R3 (Soddy’s formula), and others.

2.1 Bend Formulas for Steiner Cycles

A family of circles in the plane is called a normal family if for every circle in the
family, all other circles lie in the same side (inside or outside) of the circle. For
example, in Fig. 2.1, (a) is a normal family, but (b) and (c) are not. Note that the
family of circles that forms a Steiner cycle for a pair of circles is a normal family
since we imposed in the definition of Steiner cycle the condition that the circles are
mutually non-crossing.
Denote the radius of a circle .σ by .r(σ ). For a normal family .F of circles, the
bend .b(σ ) of a circle .σ ∈ F is defined as follows:
⌠
−1/r(σ ) if all other circles lie inside σ
b(σ ) =
.
1/r(σ ) otherwise.

© The Author(s), under exclusive license to Springer Nature Switzerland AG 2024 31

H. Maehara, H. Martini, Circles, Spheres and Spherical Geometry,
Birkhäuser Advanced Texts Basler Lehrbücher,
https://doi.org/10.1007/978-3-031-62776-7_2
32 2 Bend Formulas

(a) (b) (c)

Fig. 2.1 Three families of circles

Theorem 2.1 Let .σ1 , σ2 , . . . , σn be a Steiner cycle for .(α, β). If .n = km for some
k ≥ 2, then for every .j, 1 ≤ j ≤ m,
.

⎲
k−1
k ( )
. b(σj +im ) = cot2 (π/n) b(α) + b(β)
2
i=0

holds.
Example 2.1 In the Steiner 6-cycle shown in Fig. 1.6, we have .cot2 (π/6) = 3 and
.6 = 2 · 3. Hence, by Theorem 2.1,

b(σ1 ) + b(σ4 ) = 3(b(α) + b(β))

.

b(σ2 ) + b(σ5 ) = 3(b(α) + b(β))

b(σ3 ) + b(σ6 ) = 3(b(α) + b(β))
b(σ1 ) + b(σ3 ) + b(σ5 ) = 92 (b(α) + b(β))
b(σ2 ) + b(σ4 ) + b(σ6 ) = 29 (b(α) + b(β)).

From this simultaneous equation it follows that the radii of .σi , i = 1, 2, 3, 4, can be
determined by the radii of .α, β, σ5 , σ6 .
We prepare the proof of Theorem 2.1 by two lemmas. Let us denote by .fp the
inversion of .R2 with respect to the unit circle centered at .p ∈ R2 . So, for a circle
.σ , .fp (σ ) is its image. Note that if .F is a normal family of circles and no circle in

.F passes through .p, then the family .{fp (σ ) : σ ∈ F} is also a normal family of

circles.
Lemma 2.1 Let .{σ1 , σ2 , . . . , σn } be a normal family of n unit circles, and suppose
that the centers .x i of .σi satisfy
⌠
x 1 + x 2 + · · · + x n = O (the origin),
.
‖x 1 ‖ = ‖x 2 ‖ = · · · = ‖x n ‖ = δ ≥ 1.
2.1 Bend Formulas for Steiner Cycles 33

Fig. 2.2 .σi has the diameter

.‖bi− ai ‖
σi

bi
●
xi
ai
p

Then, for a point .p that lies on none of .σi , we have

b(fp (σi )) = ‖x i − p‖2 − 1 (i = 1, 2, . . . , n) and

.

⎲
n
b(fp (σi )) = n(‖p‖2 + δ 2 − 1).
i=1

Proof Let us denote .fp (σi ) by .σi∗ . Let .a i , bi be the intersection points of the line
.px i and the circle .σi such that .‖a i − p‖ < ‖b i − p‖, see Fig. 2.2. Then the diameter

of .σi∗ is equal to .‖f (bi ) − f (a i )‖.

(i) The case that .p lies in the exterior of .σi : Since

‖a i − p‖ = ‖x i − p‖ − 1, ‖fp (a i ) − p‖ · ‖a i − p‖ = 1,
.

we have .‖fp (a i ) − p‖ = 1/(‖x i − p‖ − 1), .‖fp (bi ) − p)‖ = 1/(‖x i − p‖ + 1).

Hence

2r(σi∗ ) = ‖fp (a i ) − fp (bi )‖ = ‖fp (a i ) − p‖ − ‖fp (bi ) − p‖

.

1 1 2
= − = ,
‖x i − p‖ − 1 ‖x i − p‖ + 1 ‖x i − p‖2 − 1

from which .b(σi∗ ) = ‖x i − p‖2 − 1 follows.

(ii) The case that .p lies in the interior of .σi : Since

‖a i − p‖ = 1 − ‖x i − p‖, ‖fp (a i ) − p‖ · ‖a i − p‖ = 1,
.

we have .‖fp (a i ) − p‖ = 1/(1 − ‖x i − p‖) = −1/(‖x i − p‖ − 1). Similarly,

‖fp (bi ) − p‖ = 1/(‖x i − p‖ + 1). Hence
.

2r(σi∗ ) = ‖fp (a i ) − fp (bi )‖ = ‖fp (a i ) − p‖ + ‖fp (bi ) − p‖

.

−1 1 −2
= + = .
‖x i − p‖ − 1 ‖x i − p‖ + 1 ‖x i − p‖2 − 1
34 2 Bend Formulas

Noting that, in this case, .σi∗ encloses all other circles .σj∗ , we have .b(σi∗ ) =
−1/r(σi∗ ) = ‖x i − p‖2 − 1 by the definition of “bend”.
From (i) and (ii), we always have .b(σi∗ ) = ‖x i − p‖2 − 1. Thus,

⎲
n ⎲
n ⎲
n
. b(σi∗ ) = ‖x i − p‖2 − n = nδ 2 + n‖p‖2 − n − 2 xi · p
i=1 i=1 i=1

(⎲
n
)
= n(‖p‖ + δ − 1) − 2
2 2
x i · p = n(‖p‖2 + δ 2 − 1).
i=1

⨆
⨅
Lemma 2.2 Let .τ1 , τ2 be a pair of concentric circles in the plane such that .r(τ1 ) =
δ + 1, r(τ2 ) = δ − 1 (δ > 1). Let .p be a point that lies on neither of them. Then

2
. b(fp (τ1 )) + b(fp (τ2 )) = (‖p‖2 + δ 2 − 1).
δ2 −1

Proof We may suppose that the center of the concentric circles is the origin O. To
make our argument clear, let us suppose that .p lies inside .τ2 . (In other cases, proofs
follow almost similarly.) Then the diameter of .fp (τ2 ) is computed as

1 1 2(δ − 1)
. + = .
δ − 1 − ‖p‖ δ − 1 + ‖p‖ (δ − 1)2 − ‖p‖2

Similarly, the diameter of .fp (τ1 ) is computed as

2(δ + 1)
. .
(δ + 1)2 − ‖p‖2

Since .fp (τ2 ) encloses the other circle, we have

(δ + 1)2 − ‖p‖2 (δ − 1)2 − ‖p‖2

b(fp (τ1 )) + b(fp (τ2 )) =
. −
δ+1 δ−1
⎛ 1 1 ⎞
= (δ + 1) − (δ − 1) + ‖p‖2 −
δ−1 δ+1
2
= 2 (‖p‖2 + δ 2 − 1).
δ −1
⨆
⨅
Proof of Theorem 2.1 By Theorem 1.3, there is a point .p ∈ R2 such that
.fp (α), fp (β) are concentric circles, .r(fp (α)) < r(fp (β)). Then all .fp (σi ), i =

1, . . . , n, have the same radius. We may suppose that the center of the concentric
2.2 Bend Formula for Soddy’s Hexlet 35

circles .fp (α), fp (β) is the origin O. Since the formula of the theorem is indepen-
dent on the unit of length, we may suppose that all circles .fp (σi ), i = 1, . . . , n, are
unit circles. Let .δ be the distance between O and the center .x i of .fp (σi ). Since the
points .x i , i = 1, 2, . . . , n, form the vertices of a regular n-gon, we have

. sin(π/n) = 1/δ,

and the radii of .fp (α), fp (β) are .δ − 1, δ + 1, respectively. Note also that for every
j (1 ≤ j ≤ m) the k points .x j +im , i = 0, 1, . . . , k − 1, are the vertices of a regular
.

k-gon, and we have

⎲
k−1
. x j +im = O.
i=0
( ) ( ) ( )
Since .fp fp (σi ) = σi , fp fp (α) = α, fp fp (β) = β, we have

⎲
k−1
. b(σj +im ) = k(‖p‖2 + δ 2 − 1),
i=0

2
b(α) + b(β) =
. (‖p‖2 + δ 2 − 1),
δ2 −1

by Lemmas 2.1 and 2.2. Now, since .δ = 1/ sin(π/n) implies .δ 2 − 1 = cot2 (π/n),
the equality of the theorem follows. ⨆
⨅

2.2 Bend Formula for Soddy’s Hexlet

Similarly to a circle-case, a family .F of spheres is called a normal family if for any

.σ ∈ F all other spheres lie in the same side (inside or outside) of .σ . The radius of
.σ is denoted by .r(σ ). The bend .b(σ ) of a sphere .σ in a normal family .F is defined

(similarly to the circle-case) in the following way:

⌠
−1/r(σ ) if all other spheres lie inside σ
b(σ ) =
.
1/r(σ ) otherwise.

Theorem 2.2 Let .σ1 , σ2 , . . . , σ6 be a hexlet for a 3-cycle .τ1 , τ2 , τ3 . Then we have
( )
b(σ1 ) + b(σ4 ) = 2 b(τ1 ) + b(τ2 ) + b(τ3 )
.
( )
b(σ2 ) + b(σ5 ) = 2 b(τ1 ) + b(τ2 ) + b(τ3 )
36 2 Bend Formulas

( )
b(σ3 ) + b(σ6 ) = 2 b(τ1 ) + b(τ2 ) + b(τ3 )
( )
b(σ1 ) + b(σ3 ) + b(σ5 ) = 3 b(τ1 ) + b(τ2 ) + b(τ3 )
( )
b(σ2 ) + b(σ4 ) + b(σ6 ) = 3 b(τ1 ) + b(τ2 ) + b(τ3 ) .

Remark 2.1 From this it follows that if .b(τ1 ), b(τ2 ), b(τ3 ) and one of .b(σi ) are
rational numbers, then all bends of the hexlet are rational numbers.
Let us state here the sphere-version of Lemma 2.1, whose proof follows from the
proof of Lemma 2.1 by replacing “circle” with “sphere”.
Lemma 2.3 Let .{σ1 , σ2 , . . . , σn } be a normal family of spheres in .R3 , and suppose
that the centers .x i of .σi satisfy
⌠
x 1 + x 2 + · · · + x n = O (the origin),
.
‖x 1 ‖ = ‖x 2 ‖ = · · · = ‖x n ‖ = δ ≥ 1.

Then for the inversion .fp of .R3 with respect to a unit sphere with center .p lying on
none of .σi , we have

b(fp (σi )) = ‖x i − p‖2 − 1 (i = 1, 2, . . . , n), and

.

⎲
n
b(fp (σi )) = n(‖p‖2 + δ 2 − 1).
i=1

⨆
⨅
Proof of Theorem 2.2 Let .p ∈ R3
be the contact point of .τ1 and .τ2 , and let
.fp be the inversion of .R3 with respect to the unit sphere centered at .p. Put
.ξ = fp (τ1 ), η = fp (τ2 ), τ
∗ = f (τ ). Then .ξ, η are a pair of parallel planes, and
p 3
∗ ∗
.τ is a sphere that is tangent to these two planes. Put .σ = fp (σi ), i = 1, 2, . . . , 6.
i
∗ ∗
Note that .σ1 , . . . , σ6 are spheres, each being tangent to the planes .ξ, η and to the
sphere .τ ∗ . The point .p lies between the planes .ξ, η and ouside of .τ ∗ , σi∗ . Since the
equality of Theorem 2.2 that we are going to prove is independent of the choice of
the unit of length, we may suppose that each .σi∗ and .τ ∗ are unit spheres. We may
further suppose that the center of .τ ∗ is the origin O, and the centers .x i of .σi∗ lie on
the plane .z = 0. Then .‖x i ‖ = 2. So, by putting .δ = 2 and applying Lemma 2.3, we
have

⎲
2
. b(fp (σj∗+2i )) = 3(‖p‖2 + 3) (in the case of k = 3),
i=0

⎲
1
b(fp (σj∗+2i )) = 2(‖p‖2 + 3) (in the case of k = 2).
i=0
2.3 Descartes’ Circle Theorem 37

If we put .p = (x, y, z), then .−1 < z < 1. We may suppose .z ≥ 0. Then, since

1 1 2
.2r(fp (τ ∗ )) = − =
‖p‖ − 1 ‖p‖ + 1 ‖p‖2 − 1
1
2r(fp (ξ )) =
z−1
1
2r(fp (η)) = ,
z+1

we have .b(fp (τ ∗ )) + b(fp (ξ )) + b(fp (η)) = ‖p‖2 + 3. Thus the equality of the
theorem holds. ⨅
⨆
Remark 2.2 Frederick Soddy found a hexlet in 1936 without using inversions, and
he proved, concerning the bends .b1 , . . . , b6 of the six spheres, the equality .b1 +b4 =
b2 + b5 = b3 + b6 . However, this equality was found more than 100 years earlier
in Japanese temple geometry, see the booklet of H. Fukagawa and D. Pedoe [284,
Ch. 10, Problem 9.10].

2.3 Descartes’ Circle Theorem

We continue with a further famous theorem on circle arrangements.

Theorem 2.3 (Descartes’ Circle Theorem) If .{α1 , α2 , α3 , α4 } is a normal family
of mutually tangent circles in the plane, then

⎛⎲
4 ⎞2 ⎲
4
. b(αi ) =2 b(αi )2
i=1 i=1

holds, where .b(αi ) is the bend of the circle .αi .

Proof Let .p be the contact point of .α3 and .α4 . Then .fp (α3 ) and .fp (α4 ) are parallel
lines, and .fp (α1 ) and .fp (α2 ) are mutually tangent circles, both tangent to the two
parallel lines. Since the formula in the theorem is independent of the unit of length,
we may suppose that .fp (α1 ), fp (α2 ) are unit circles with centers .(−1, 0) and .(1, 0),
respectively. Put .p = (x, y). We may suppose .0 ≤ y < 1. Then, by Lemma 2.1,

b(α1 ) = (x + 1)2 + y 2 − 1 = x 2 + 2x + y 2 ,
.

b(α2 ) = (x − 1)2 + y 2 − 1 = x 2 − 2x + y 2 .
38 2 Bend Formulas

In the pair of parallel lines .fp (αi ), i = 3, 4, suppose that .p is nearer to .fp (α3 ).
Then .2r(α3 ) = 1/(1 − y), 2r(α4 ) = 1/(1 + y), hence

b(α3 ) = 2(1 − y), b(α4 ) = 2(1 + y).

.

∑
Therefore, . 4i=1 b(αi ) = 2(x 2 + y 2 ) + 4. On the other hand,

⎲
4
. b(αi )2 = 2(x 2 + y 2 )2 + 8x 2 + 8 + 8y 2
i=1

= 2(x 2 + y 2 )2 + 8(x 2 + y 2 ) + 8.

Hence the formula in the theorem holds. ⨆

⨅

2.4 Soddy’s Formula

We start with a lemma.

Lemma 2.4 Let .x i ∈ R2 , i = 1, 2, 3, be the vertices of an equilateral triangle
inscribed in the circle with radius .δ centered at the origin O. Then, for any point
.p 0 /= O in the plane .R , we have
2

⎲
3
. ‖x i − p0 ‖2 = 3(δ 2 + ‖p 0 ‖2 ) and
i=1

⎲
3
‖x i − p0 ‖4 = 6δ 2 ‖p0 ‖2 + 3(δ 2 + ‖p 0 ‖2 )2 .
i=1

Proof Let ./ p 0 Ox i = θi (oriented angle) for .i = 1, 2, 3. Then the three points

(cos θi , sin θi ), i = 1, 2, 3, are the vertices of an equilateral triangle. In this case,
.

the three points ∑ .(cos 2θi , sin 2θi ), i = 1, 2, 3, are also the vertices of an equilateral
∑
triangle. Hence . 3i=1 cos θi = 3i=1 cos 2θi = 0.
(i) From .‖x i − p 0 ‖2 = δ 2 + ‖p 0 ‖2 − 2δ‖p 0 ‖ cos θi , we have

⎲
3
. ‖x i − p 0 ‖2 = 3(δ 2 + ‖p 0 ‖2 ).
i=1
2.4 Soddy’s Formula 39

(ii) From
( )2
‖x i − p 0 ‖4 = δ 2 + ‖p 0 ‖2 − 2δ‖p 0 ‖ cos θi
.

= 4δ 2 ‖p 0 ‖2 cos2 θi − 4δ‖p 0 ‖(δ 2 + ‖p 0 ‖2 ) cos θi

+ (δ 2 + ‖p 0 ‖2 )2 ,

cos 2θi
.cos2 θi = 1
2 + 2 , and .cos 2θ1 + cos 2θ2 + cos 2θ3 = 0, it follows that

⎲
3
. ‖x i − p 0 ‖4 = 6δ 2 ‖p 0 ‖2 + 3(δ 2 + |p 0 |2 )2 .
i=1

⨆
⨅

Theorem 2.4 (Soddy’s Formula) If .{σ1 , σ2 , . . . , σ5 } is a normal family of mutu-

ally tangent five spheres in .R3 , then we have

⎛⎲
5 ⎞2 ⎲
5
. b(σi ) =3 b(σi )2 .
i=1 i=1

Remark 2.3 This formula became well-known by the poem “The kiss precise”
published in Nature in 1936 by Frederic Soddy. But also this formula was already
known to Japanese mathematicians in the Edo-period, see [284].
Proof Let .p be the contact point of .σ4 , σ5 , and denote by .fp the inversion of .R3
with respect to the unit sphere centered at .p. Then .fp (σ4 ) and .fp (σ5 ) are parallel
planes, .fp (σi ), i = 1, 2, 3, are mutually tangent spheres each tangent to these two
planes. Since the formula in Theorem 2.4 is independent of the choice of the unit of
length, we may suppose that .fp (σi ), i = 1, 2, 3, are unit spheres. Hence the centers
.x i , i = 1, 2, 3, of these spheres are the vertices of an equilateral triangle with edge-

length 2. We may suppose that these three points .x i , i = 1, 2, 3, lie on the plane
.z = 0 in .R , and .x 1 + x 2 + x 3 = O (the origin). Put .p = (x, y, z), 0 ≤ z < 1,
3
√
and .p0 = (x, y, 0). Since .δ := ‖x 1 ‖ = ‖x 2 ‖ = ‖x 3 ‖ = 2/ 3, it follows from
Lemma 2.4 that, for .i = 1, 2, 3,

b(σi ) = ‖x i − p‖2 − 1 = ‖x i − p0 ‖2 + z2 − 1
.

b(σi )2 = ‖x i − p0 ‖4 + 2‖x i − p 0 ‖2 (z2 − 1) + (z2 − 1)2 .

By Lemma 2.4,

⎲
3 ⎲
3
. b(σi ) = ‖x i − p0 ‖2 + 3(z2 − 1) = 3(δ 2 + ‖p 0 ‖2 ) + 3(z2 − 1),
i=1 i=1
40 2 Bend Formulas

and

⎲
3 3 ⎛
⎲ ⎞
. b(σi )2 = ‖x i − p 0 ‖4 + 2‖x i − p0 ‖2 (z2 − 1) + 3(z2 − 1)2
i=1 i=1

= 6δ 2 ‖p 0 ‖2 + 3(δ 2 + ‖p 0 ‖2 )2 + 6(z2 − 1)(δ 2 + ‖p 0 ‖2 ) + 3(z2 − 1)2

⎛ ⎞2
= 3 (δ 2 + ‖p 0 ‖2 ) + (z2 − 1) + 6δ 2 ‖p 0 ‖2 .

On the other hand, since .b(σ4 ) = 2(1 − z), b(σ5 ) = 2(1 + z),

⎲
5
. b(σi ) = 3(δ 2 + ‖p 0 ‖2 ) + 3(z2 − 1) + 4,
i=1

⎲
5 ⎛ ⎞2
. b(σi )2 = 3 (δ 2 + ‖p 0 ‖2 ) + (z2 − 1) + 6δ 2 ‖p 0 ‖2 + 8(z2 + 1).
i=1

From these two equalities and .δ 2 = 4/3, we can verify that

⎲
5 ⎛⎲
5 ⎞2
3
. b(σi )2 = b(σi ) .
i=1 i=1

⨆
⨅

2.5 Linked 4-Cycle Pairs

Lemma 2.5 Let α1 , α2 , α3 , α4 be a 4-cycle of spheres, and β1 , β2 , . . . , βn be an

n-cycle of spheres, n ≥ 3. If αi and βj are tangent for every i = 1, 2, 3, 4, j =
1, 2, . . . , n, then n = 4.
Proof Since α1 and α3 are disjoint spheres, there is a point p such that fp (α1 ) and
fp (α3 ) are concentric spheres by Theorem 1.9. We may suppose that the center of
these concentric spheres is the origin O of R3 . Put αi∗ = fp (αi ), i = 1, 2, 3, 4,
and βj∗ = fp (βj ), j = 1, 2, . . . , n. The spheres α2∗ , α4∗ and βj∗ , j = 1, 2, . . . , n,
are all tangent to the concentric spheres, and hence they have the same radius, say
r. See Fig. 2.3. Let x i be the center of αi∗ for i = 2, 4, and y j be the center of βj∗
for j = 1, 2, . . . , n. Then ‖y j − x i ‖ = 2r for i = 2, 4, j = 1, 2, . . . , n. Hence
y j , j = 1, 2, . . . , n, lie on the plane H that perpendicularly bisects the line segment
x 2 x 4 . Since ‖x 2 ‖ = ‖x 4 ‖, O also lies on the plane H . Let q be the midpoint of
x 2 x 4 . Then ‖y 1 − q‖ = ‖y 2 − q‖ = · · · = ‖y n − √q‖. Since n ≥ 3, we can deduce
that q = O. Therefore, ‖x i −O‖ = ‖y j −O‖ = 2r for i = 2, 4, j = 1, 2, . . . , n.
2.5 Linked 4-Cycle Pairs 41

Fig. 2.3 α1∗ , α3∗ are

concentric

This implies that / y 1 Oy 2 = · · · = / y n Oy 1 = 90◦ . Therefore, we must have

n = 4. ⨆
⨅
If a pair of 4-cycles α1 , α2 , α3 , α4 and β1 , β2 , β3 , β4 satisfies that each αi is
tangent to all βj , j = 1, 2, 3, 4, then the pair of 4-cycles is called a linked 4-cycle
pair.
Example 2.2 √ and α3 be concentric spheres in R3 centered at the origin O,
Let α1 , √
and with radii 2−1,√ 2+1, respectively.
√ Further on, let α2 , α4 be the unit spheres
with centers (0, 0, 2),√ (0, 0, − 2),
√ √ Let βj , j = √
respectively. 1, 2, 3, 4, be the unit
spheres with centers ( 2, 0, 0), (0, 2, 0), (− 2, 0, 0), (0, − 2, 0), respectively.
Then the 4-cycles α1 , α2 , α3 , α4 and β1 , β2 , β3 , β4 form a linked 4-cycle pair.
Theorem 2.5 (Bend Formula for a Linked 4-Cycle Pair) Suppose that a pair of
4-cycles α1 , α2 , α3 , α4 and β1 , β2 , β3 , β4 forms a linked 4-cycle pair. Then

b(α1 ) + b(α3 ) = b(α2 ) + b(α4 ) = b(β1 ) + b(β3 ) = b(β2 ) + b(β4 ).

.

Proof By Theorem 1.9, there is a point p on the line connecting the centers of
α1 , α3 such that fp (α1 ) and fp (α3 ) are concentric spheres. We may suppose that
their common center is the origin O of R3 . Put αi∗ = fp (αi ), i = 1, 2, 3, 4, and
βj∗ = fp (βj ), j = 1, 2, 3, 4. Note that these spheres have the same radius. We may
suppose that these are all unit spheres. Let x i be the center of αi∗ for i = 2, 4, and
y j be the center of βj∗ for j = 1, 2, 3, 4. Then

‖x 2 − y j ‖ = ‖x 4 − y j ‖ = 2 for j = 1, 2, 3, 4,
.

and

‖y 1 − y 2 ‖ = ‖y 2 − y 3 ‖ = ‖y 3 − y 4 ‖ = ‖y 4 − y 1 ‖ = 2.
.

Hence

.y1y2y3y4, x2y2x4y4, x2y1x4y3

42 2 Bend Formulas

√
are squares with side-length 2 centered at O. Thus, ‖x i ‖ = ‖y j ‖ = 2, and the
√
radii of fp (α1 ), fp (α3 ) are 2 ± 1. By Lemma 2.1, we have

b(fp (β1∗ )) + b(fp (β2∗ )) + b(fp (β3∗ )) + b(fp (β4∗ )) = 4(‖p‖2 + 1)

.

b(fp (α2∗ )) + b(fp (β2∗ )) + b(fp (α4∗ )) + b(fp (β4∗ )) = 4(‖p‖2 + 1)

b(fp (α2∗ )) + b(fp (β1∗ )) + b(fp (α4∗ )) + b(fp (β3∗ )) = 4(‖p‖2 + 1),

from which we get

. b(fp (β1∗ )) + b(fp (β3∗ )) = b(fp (β2∗ )) + b(fp (β4∗ ))

= b(fp (α2 )) + b(fp (α4 )) = 2(‖p‖2 + 1),

that is,

b(β1 ) + b(β3 ) = b(β2 ) + b(β4 ) = b(α2 ) + b(α4 ) = 2(‖p‖2 + 1).

.

On the other hand, by Lemma 2.2,

b(fp (α1∗ )) + b(fp (α3∗ )) = 2(‖p‖2 + 1),

.

that is, b(α1 ) + b(α3 ) = 2(‖p‖2 + 1). Hence the theorem follows. ⨆
⨅
For bend-formulas in other sphere-systems and in higher dimensions see [524].

2.6 Exercises

2.1 Given a Steiner 6-cycle σ1 , . . . , σ6 for (α, β), prove that if the radii of
α, β, σ1 , σ2 are rational numbers, then the radii of σi , i = 3, 4, 5, 6, are all
rational numbers.
2.2 Let A, B, C be mutually externally tangent three circles with radii 1, 1, 2,
respectively. Find the radius of the circle that is internally tangent to all
A, B, C.
2.3 Suppose that four balls lie on a table, and these four balls are mutually tangent.
When three of the balls have unit radius, find the radius of the remaining ball.
2.4∗ Let σ1 , σ2 , σ3 , σ4 , τ1 , τ2 be a system of six spheres in R3 such that, for i =
1, 2, the five spheres σ1 , σ2 , σ3 , σ4 , τi are mutually tangent. Prove that

b(σ1 ) + b(σ2 ) + b(σ3 ) + b(σ4 ) = b(τ1 ) + b(τ2 ).

.
2.7 Notes 43

Fig. 2.4 Circles α and β are

internally tangent
α σ1
β σ2
σ4 σ3

2.5∗ Consulting Fig. 2.4, let σ1 , σ2 , σ3 , σ4 be four circles such that each is tangent
to both α, β, and σi is externally tangent to σi+1 for i = 1, 2, 3. Prove that
b(σ1 ) + 3b(σ3 ) = b(σ4 ) + 3b(σ2 ).

2.7 Notes

The notes completing Chap. 2 refer to the substantial themes of Descartes’ circle
theorem, Soddy’s formula, Apollonian circle packings, and the problem of Apollo-
nius on touching circles.

2.7.1 Notes on Descartes’ Circle Theorem and Soddy’s

Formula

Given three mutually tangent circles, there are two circles touching all of them (and
are also called Soddy circles). Descartes’ circle theorem states that for each four
mutually tangent circles, the radii of the circles satisfy a certain quadratic equation.
It is closely related to the famous problem of Apollonius (on touching circles) and
goes back to Descartes (namely to the year 1643, see the historical contributions
[114] and [70] for more details). Independently, the Japanese mathematician Y.
Nushizumi stated a form of Descartes’ circle theorem in 1751.
Descartes’ theorem occurs in not too many basic geometry books. It is, however,
e.g. discussed in Coxeter’s basic book [182, 1.56–1.59], in Pedoe’s monograph [625,
Section 40.3], and in the book [284, Section 1.7] on the Japanese temple geometry.
The papers [7], [624], [185], and [810] from the Sixties contain new proofs
and nice generalizations of Descartes’ theorem. These generalizations refer, for
example, to the number of participating circles, and the paper [624] contains also a
survey. Further straightforward and recent proofs of Descartes’ theorem are given
by Levrie [503] and Bradford [121].
Coxeter [185] finds a new approach to the n-dimensional version with .n + 2
spheres (using inversions) and studies their spiral-shaped occurrence, where the
44 2 Bend Formulas

contact points of consecutive spheres belong to a “loxodromic curve”. Various

further interesting properties of these sphere arrangements are obtained, and Weiss
[803] continues these investigations in view of Möbius transformations.
In the next two papers, relations between Steiner chains and Descartes’ circle
theorem are investigated. Schwartz and Tabachnikov [719] investigate curvatures
in Steiner chains, getting consequences of Descartes’ circle theorem as subcase. And
Moritsugu [585] studies extensions of Descartes’ theorem regarding the number of
participating circles, and instead of inversions or related methods he uses Gröbner
bases or resultants for the equations of in- and circumscribed circles.
Lagarias, Mallows, and Wilks [487] prove that the “curvature-centers” (cur-
vature times center, where the center is considered as a complex number) satisfy
Descartes’ four-numbers relation. Northshield [598] presents an elegant, short proof
of this result and gives an analogous version for spheres. Analogously, Tupan [784]
gives a short proof of Descartes’ theorem using only basic theorems from plane
geometry and complex numbers.
Finally we cite the nice and elementary exposition [400] of Holshouser et al.
discussing results around the problem of Apollonius and Descartes’ theorem; in
particular, Soddy’s contribution like also Gosset’s approach to the n-dimensional
analogue for .n + 2 spheres are discussed.
As already mentioned in Remark 2.3, Soddy’s formula for five spheres in 3-space
became known due to Soddy’s poem “The kiss precise”, but was already known
in Japan during the Edo period. An interesting historical article to this formula
is [729] by Sharygin and Shtogrin; they emphasize this older source from Japan
(going back to the eighteenth century). In addition, the authors give a careful proof
(based on determinants) of the n-dimensional analogue of Soddy’s result, with .n + 2
mutually touching spheres. Kontorovich [474] studies sphere packings in 3-space
whose bends (reciprocals of radii) are integers, calling them Soddy sphere packings.
The quadratic form he uses is a natural analogue of that coming from Descartes’
circle theorem. The paper has an interestingly rich collection of related references.

2.7.2 Notes on Apollonian Circle Packings

The configuration in Descartes’ circle theorem can be seen as starting point for
constructing Apollonian circle packings, coming from Apollonius’ statement: Given
three mutually tangent circles in the plane, there exist exactly two circles tangent to
all three. (The given circle configuration presents a special case of the arbitrarily
given configuration of three circles for Apollonius’ famous touching problem, see
below.)
An Apollonian circle packing arises by repeatedly filling the interstices between
mutually tangent circles with further tangent circles; see, e.g., [78, II.9] and also
[644, Chapter 4]. Lagarias et al. [487] present a new method for proving Descartes’
theorem and various further results, such as analogues in higher dimensions, the
2.7 Notes 45

complex variant of Descartes’ theorem, extensions to Apollonian circle packings,

and extensions to spherical and hyperbolic geometry.
They continue, with two additional authors, their deep work on Apollonian circle
packings by the trilogy [319], [320], and [321] on group theoretic and geometric
aspects, which follows also the number-theoretic paper [318]. An Apollonian circle
packing is called integral if every circle in the packing has an integer curvature.
In [318] number-theoretic properties of the set of integer curvatures occurring in
integral Apollonian circle packings are studied.
In the mentioned trilogy of articles, the authors observe that there are Apollonian
packings which have strong integral properties; i.e., all circles in such packings have
integer curvatures and rational centers. They start with the Descartes arrangement
(four mutually tangent circles) and generate further circles via inversions in the four
starting circles.
In Part I (i.e., in [319]) it is shown that there is a discrete group, called Apollonian
group and acting on a parameter space of Descartes configurations, such that these
configurations in a packing form an orbit under the action of this group. The authors
show that there exist infinitely many types of integral Apollonian packings, where
the integral structure is related to the integral nature of the Apollonian group, and
they obtain many further related results.
In Part II (i.e., in [320]) the action of a larger discrete group, namely the super-
Apollonian group also having an integral structure, is studied. The orbits describe
the Descartes quadruples of a so-called super-packing. Its circles never cross but
are nested to an arbitrary depth. It turns out that certain Apollonian packings and
super-packings are strongly integral, and it is proved that there are exactly eight
types of strongly integral super-packings, each containing a copy of every integral
Apollonian circle packing (as before, up to scale). Further properties of the super-
Apollonian group are shown.
In the third part [321], the authors extend the results of the other two parts to
n-dimensional analogues of the Apollonian circle packings, having n-dimensional
Desargues configurations consisting of .n + 2 mutually touching spheres as starting
point. They study n-dimensional analogues of the Apollonian group, its dual,
and the super-Apollonian group, and they show that the Apollonian group and
the dual Apollonian group are Coxeter groups. An Apollonian cluster ensemble
is introduced, being any orbit under the Apollonian group; similar notions are
developed for the other two types of groups. Among further results, it is shown
in which dimensions there are rational Apollonian cluster ensembles (i.e., all
curvatures are rational).
The curvatures of four mutually tangent circles with disjoint interiors clearly
form a Descartes quadruple. The four least curvatures in an integral Apollonian
circle packing form a so-called root Descartes quadruple, and it is a primitive root
quadruple if the curvatures are relatively prime. Deriving a closed formula for the
number of primitive root quadruples with minimum curvature, Northshield [597]
confirms a conjecture of Mallows. The paper contains further interesting results in
that direction.
46 2 Bend Formulas

It is known that if the original four circles of an Apollonian circle packing have
integer curvatures, all of the circles in the packing will have this property. Bourgain
and Fuchs [118] present a lower bound for the number .k(P , X) of integers up to X
occurring as curvatures in a bounded integer Apollonian packing P . They confirm
a positive-density conjecture of Graham et al., namely that the ratio .k(P , X)/X
is larger than 0 for X tending to infinity. In other words, there exists a constant
.c > 0 depending on P such that .k(P , X) ≥ cX. There exist various further and

similar interesting results on Apollonian circle packings with integer curvatures,

and Fuchs [282] gives a nice survey on related results and methods. Another
expository paper on recent results on Apollonian circle packings and the classical
Apollonian problem in view of counting problems is [602] by Oh. The author
discusses integral and primitive Apollonian packings, circles with prime curvatures
in such structures, the importance of group-theoretical methods (leading to so-called
thin subgroups) and various further aspects. Here we should also mention the older
review [693] of Sarnak, who treats the construction of integral Apollonian circle
packings emphasizing also a number of Diophantine problems, which come to be in
the context of such packings.

2.7.3 Notes on Apollonius’ Touching Problem

The famous problem of Apollonius on constructing all circles that touch three given
ones in arbitrary positions has a long history and is, in particular, an excellent
“exercise” in using inversion. Mostly in the literature, the given or constructed
circles are allowed to be presented also by the borderline cases lines (radius infinity)
or points (radius zero).

Book Chapters and Surveys

Ways of solution can be found in several book chapters; see, e.g., Johnson [430,
Chapter VI], Schmidt [698, Chapter VII in Part A and Chapter II in Part B],
Dörrie [219, pp. 154–160], Pedoe [625, Section 26.1], Berger’s book [75, 10.11.1],
Ogilvy [601, Section 4.3], Hartshorne’s monograph [382, Section 38], Aumann [48,
Chapter 8], as well as Posamentier and Geretschläger [644, Chapter 7]. Also the
survey [820, Section 12.2] and Section 1.4 in [738] (on conics and the problem of
Apollonius) should be mentioned here. Many of these references contain historical
material on this problem. A paper dedicated as a whole to the history of the problem
is [514] due to Lyuter.
2.7 Notes 47

Some Research Results

Gyarmathi [351] solves the n-dimensional Apollonius problem constructively by

using the cyclographic mapping. Also Iwata [423] solves the n-dimensional version
and extends the problem to elliptic space; he presents also Descartes’ circle theorem.
Coxeter [185] investigates special cases of Apollonius’ problem depending on
the position of the given circles, and he applies inversion as well as hyperbolic
geometry (via stereographic projection). It is clear that the type of the three given
figures (if we allow also points and lines) and their position to each other determine
the number of figures (points, lines or circles) that solve the problem. Pedoe [626]
shows that the numbers .0, 1, 2, 3, 4, 5, 6, 8 are all possible, but that 7 cannot occur.
Fitz-Gerald [270] completes this by showing that the case of one solution cannot
exist if points are excluded from the definition of circles. Based on inversion, Bruen
et al. [130] present a very detailed treatment of all possible cases in the real Möbius
plane. Special care is given to Pedoe’s result from [626].
The three-dimensional analogue (with four given spheres) is studied by Langlet
[491]. He reduces the case that the sphere is externally or internally tangent to all
given spheres to solving a simple quadratic equation for the radius and a system
of three linear equations (for the three coordinates of the centers). This method
can be extended to higher dimensions. Paluszny and Wilker [611] discuss the n-
dimensional case referring to (.n − 1)-spheres touching .n + 1 given spheres. E.g.,
if no two of the given spheres intersect, and no three have the property that one
separates the other two, the expected solution number is .2n+1 . The number of
solutions depends then on the relative size and position of the given spheres. This is
discussed in detail for .n = 3, regarding the six inversely invariant parameters which
represent configurations of four given spheres. If the solution number is finite, it can
be any integer from 0 to 16, and if it is infinite, it can be a one, two or threefold
infinity depending on the number of one-parameter families of solutions.
The two papers [713] and [714] of Schroth are dedicated to the Apollonius
problem in Laguerre geometry, and the problem is completely solved there for flat
Laguerre planes and for 4-dimensional locally compact topological Laguerre planes,
respectively.
Already Schoute contributed to an n-dimensional analogue of Apollonius’
problem with given hyperplanes, and Drechsler and Sterz [226] continue these
studies using modern cohomological theory and enumerative geometry in projective
space.
Knight [457] investigates the Apollonius problem with the help of Lie contact
geometry. He develops a related concept, called signature Lie triads, for the
classification task. The established proof yielding the classification is relatively
long.
With the help of inversion, Morita [584] studies the following configuration
clearly related to Apollonius’ problem: One starts with circles .O1 , O2 , O3 touching
each other externally, and takes S as the circle touching all .O1 , O2 , O3 and having
them in its interior. Now let .I1 , I2 , I3 be the three circles tangent to .S, O2 , O3 , to
.S, O1 , O3 and to .S, O1 , O2 , respectively. Then let .si and .ti be the two common
48 2 Bend Formulas

external tangents of .Oi and .Ii , i = 1, 2, 3. It follows that there is a circle tangent to
all of .s1 , t1 , s2 , t2 , s3 and .t3 . Furthermore, let .A1 A2 A3 be the triangle the edges of
which lie on the outer common external tangents of .O1 , O2 and .O3 , and let .B1 B2 B3
be determined analogously by the outer common external tangents of .I1 , I2 and
.I3 . Let, finally, .l1 , l2 and .l3 be the lines joining opposite vertices of the triangles

.A1 A2 A3 and .B1 B2 B3 . Then the lines .l1 , l2 , l3 have a common point. The author

also derives an analogous three-dimensional configuration.

Berger et al. [74] study a typical location problem based on the Apollonius task,
namely to determine the optimal location of production facilities in the plane. On
the base of a uniform representation of Möbius, Laguerre, and Minkowski geometry,
Taherian and Mohseni [770] obtain a new approach to the Apollonius problem,
where also stereographic projection is used.
Baragar and Kontorovich [57] present some history of the Apollonius problem,
including the names Viète and Gergonne, and lay special emphasize on the
efficiency (i.e., the number of moves) in constructing the circle which is tangent
to three mutually touching circles. It is shown that seven moves are sufficient.
Chapter 3
Graphs and Circle-Systems

Mathematics compares the most diverse phenomena and

discovers the secret analogies that unite them.
— Joseph Fourier

We will explain some fundamentals on graphs, including planar graphs and Euler’s
formula. As a representation of a graph by a circle-system, we consider first the
orthogonal-circle representation (abbreviated as OCR) of a special type of graphs,
called quadrangulations. An OCR of a graph is a circle-system consisting of
circles each corresponding to a vertex of the graph, in which each pair of circles
corresponding to a pair of adjacent vertices is orthogonally crossing, whereas each
pair of circles corresponding to non-adjacent vertices do not cross each other. The
OCR theorem gives a necessary and sufficient condition for a quadrangulation
to have an OCR. From the OCR theorem, the coin graph theorem (also called
Koebe’s theorem, or the Koebe-Andreev-Thurston theorem) and Steinitz’ theorem
characterizing the polyhedral graphs, are derived easily.

3.1 Graphs and Planar Graphs

Suppose that for any two distinct elements of a set, it is somehow defined whether
the two elements are adjacent or not. Then we say the set is equipped with the
adjacency relation. A non-empty finite set V equipped with an adjacency relation is
called a graph, and each element of V is called a vertex of the graph. For example:
• For a set consisting of a finite number of points in .R3 , two points are defined to
be adjacent if and only if this pair of points is exactly unit distance apart. Then
we have a graph. This graph is known as a unit-distance graph in .R3 .
• In the set of vertices of a convex polyhedron .Π in .R3 , two vertices are defined to
be adjacent if they are the end-points of an edge of the polyhedron .Π . Then we
have a graph. This graph is called the graph of the polyhedron .Π .

© The Author(s), under exclusive license to Springer Nature Switzerland AG 2024 49

H. Maehara, H. Martini, Circles, Spheres and Spherical Geometry,
Birkhäuser Advanced Texts Basler Lehrbücher,
https://doi.org/10.1007/978-3-031-62776-7_3
50 3 Graphs and Circle-Systems

• In a set consisting of a finite number of mutually non-overlapping disks in the

plane, two disks are defined to be adjacent if and only if they are tangent. Then
we have a graph whose vertices are disks. This graph is called a coin graph.
In a graph, each pair .(u, v) of mutually adjacent vertices is called an edge of the
graph, and it is simply denoted by uv or vu. Since the adjacency-relation of a graph
can be given by the set .E = {uv, xy, . . . } of all edges, we may regard a graph as a
pair .G = (V , E) of the set V of all vertices and the set E of all edges.
Vertices of a graph .G = (V , E) are (tentatively) denoted by .u, v, w, . . . and
its edges are denoted by .a, b, c, . . . . If .a = uv ∈ E, then a is called the edge
connecting u and v; the vertices .u, v are called the endpoints of the edge a, and u is
said to be adjacent to v, or u is a neighbor of v. For a vertex .v ∈ V , the set of all
vertices that are adjacent to v is denoted by .N(v), and called the neighborhood of
v. Thus

N(v) = {x ∈ V : xv ∈ E}.
.

The number of vertices in .N(v) is called the degree of v and denoted by .deg v. For a
graph .G = (V , E), the minimum value and the maximum value of .deg v for .v ∈ V ,
are called the minimum degree, and the maximum degree of G, respectively. If, in a
graph, the degree of every vertex is known, then the number of edges of the graph is
derived from the following “hand-shaking lemma”, the proof of which is easy and
will be omitted.
Lemma 3.1 (Hand-Shaking Lemma) In a graph .G = (V , E),
⎲
. deg v = 2 × (the number of edges).
v∈V

Suppose that two graphs .G = (V , E) and .G' = (V ' , E ' ) satisfy .V ' ⊂ V and
.E
' ⊂ E. Then .G' is called a subgraph of G. If, further, .G /= G' holds, then .G' is
called a proper subgraph of G. If a graph has n vertices, and every pair of distinct
vertices is adjacent, then the graph is called the complete graph with n vertices, and
( ) by .Kn . Each vertex of .Kn has degree .n − 1, and the number of
it is usually denoted
edges of .Kn is . n2 .
Every graph can be “realized” as a figure in .R3 . For a graph .G = (V , E),
represent each vertex by a point in .R3 , and connect each pair of points corresponding
to a pair of adjacent vertices with a simple curve, so that different curves never
intersect except in their common endpoints. This is clearly possible in .R3 . The
figure obtained in this way is called a realization of the graph G. In a realization of
a graph G, simple curves corresponding to edges of G are also called edges of the
realization, and points corresponding to the vertices of G are also called the vertices
of the realization. Thus the realization of a graph is also regarded as a graph. For
a convex polyhedron .Π in .R3 , its 1-skeleton (that is, the set of vertices and edges
of .Π ) is a realization of the graph of .Π . Hereafter, by the graph of the convex
polyhedron we mean the 1-skeleton of the convex polyhedron.
3.1 Graphs and Planar Graphs 51

Fig. 3.1 The complete graphs .K5 , K6 and the graph of a cube

In many cases, it is convenient to “represent” graphs by figures in the plane. So,

for a graph, we use
a suitable projection of one of its realization in .R3 into a plane
as a representation of the graph. In such a figure, projections of edges of the
realization (simple curves) could cross each other. So, to distinguish vertices of the
graph from such crossing points of curves, vertices are usually denoted by .◦ or .•;
see Fig. 3.1 which shows the complete graphs .K5 , K6 and the graph of a cube.
In a graph .G = (V , E), a sequence of edges for which the corresponding
sequence of edges in its realization constitute a simple curve is called a path of G,
and if .u, v are vertices corresponding to the end-points of the simple curve, the path
is said to connect u and v. A sequence of edges in G for which the corresponding
sequence of edges in its realization constitutes a simple closed curve, is called a
cycle. The number of edges contained in a path or cycle is called the length of the
path or the cycle. A cycle of length k (that is, a cycle consisting of k edges) is called
a k-cycle. A cycle of even length (resp. odd length) is simply called an even-cycle
(resp. odd-cycle). For example, the graph of a cube contains 4-cycles, 6-cycles, 8-
cycles, but contains no odd-cycle. If .n > 3, then the complete graph .Kn contains
k-cycles for .k = 3, 4, . . . , n.
Lemma 3.2 If each vertex of G has degree at least 2, then G contains a cycle.
Proof Let us regard a realization of G as a road network. Starting from a vertex u
we go along an edge to an adjacent vertex v. Since .deg v ≥ 2, without returning the
same edge, we can proceed along another edge to another vertex w. Since .deg w ≥
2, without returning the last edge, we can proceed further along an edge to another
vertex. If we continue in this way, since the number of vertices is finite, sooner or
later we come to a vertex that we already visited. Let z be the first vertex we visit
twice. Then the vertices we visited after the first visit of z till the second visit of
z are all different. Hence the sequence of edges during the first visit of z and the
second visit of z form a cycle. ⨆
⨅
A graph G is called connected if for any two vertices of G there is a path in G
that connects the two vertices. A graph consisting of a single vertex is defined to
be connected. A graph that is not connected is called disconnected. A disconnected
graph can be regarded as the disjoint union of connected graphs. In this sense, each
52 3 Graphs and Circle-Systems

Fig. 3.2 Addition of a bypass (left) and a pointed-edge (right)

Fig. 3.3 A realization of the

graph of a cube in the plane

connected graph is called a connected component (or simply a component) of the

disconnected graph.
In a connected graph .G = (V , E), the operation to add a new edge .e /∈ E that
connects a pair of nonadjacent vertices in G is called the addition of a bypass. The
operation to add a new vertex v and an edge that connects v to a vertex of G is called
the addition of a pointed-edge (Fig. 3.2).
The following theorem is obvious.
Theorem 3.1 Every connected graph with at least two vertices can be obtained
from a single vertex graph by repeating addition of a pointed-edge and addition of
a bypass. ⨆
⨅
If a graph G can be realized as a figure contained in a subset .X ⊂ R3 , then we say
that G is embeddable in X. A graph that is embeddable in a plane is called a planar
graph, whereas a graph realized in a plane is called a plane graph. For example, the
graph of a cube (Fig. 3.1 right) is a planar graph, and Fig. 3.3 shows a plane graph
obtained as a realization of the graph of a cube.
A graph “embedded” on a sphere (that is, realized on a sphere) can be mapped by
a stereographic projection to a plane graph. And every plane graph can be mapped
to a graph embedded on a sphere by the inverse mapping of some stereographic
projection. Thus, we have the following theorem.
Theorem 3.2 A graph G is planar if and only if G is embeddable on a sphere. ⨆
⨅
.R ,
Let G be the graph of a convex polyhedron .Π in 3
that is, the 1-skeleton of
Π . Let P be an interior point of the convex polyhedron .Π , and .S be a sphere with
.

center P . By the central projection .R3 → S from P , G can be embedded in .S.

Corollary 3.1 The graph of a convex polyhedron is a planar graph. ⨆
⨅
A plane graph G divides the plane into several regions. Each region is called a
face of G. Note that the outermost infinite region is also a face. For example, the
plane graph in Fig. 3.3 has six faces. A plane graph consisting of a single vertex
has one face. Faces of a plane graph are denoted by .α, β, . . . . An edge of a plane
graph G is called a boundary edge of a face .α if one side (or both sides) of the edge
3.1 Graphs and Planar Graphs 53

form the face .α. A subgraph of a plane graph is called the boundary of a face .α, if
it consists of all boundary edges of .α and the end-vertices of the boundary edges.
Let us note here that every connected plane graph with at least two vertices is
obtained from a single vertex by repeating addition of a pointed-edge and addition
of a bypass in the plane. For a plane graph G, let .v(G), e(G), f (G) denote the
number of vertices, the number of edges, and the number of faces of G, respectively.

Theorem 3.3 (Euler’s Formula) For every connected plane graph G,

v(G) − e(G) + f (G) = 2

.

holds.
For example, if G is the plane graph in Fig. 3.3, then .v(G) = 8, e(G) =
12, f (G) = 8. Thus, .v(G) − e(G) + f (G) = 8 − 12 + 6 = 2.
Proof For a graph G, denote .v(G) − e(G) + f (G) simply by .ψ(G). If we apply
“addition of a pointed-edge in the plane” to a connected plane graph, then the
number of vertices and the number of edges increase by 1, but the number of faces
does not change. Hence, the value of .ψ is invariant by addition of pointed-edge in
the plane. If we apply “addition of a bypass in the plane” to a connected plane graph,
then the number of vertices does not change, and the number of edges increases by
1. If .u, v are the vertices connected by the bypass, then, since the original graph
is connected, there is a path in the original graph that connects u and v. This path
and the added bypass make together a simple closed curve in the plane, and this
simple closed curve divides the plane into two regions, inside and outside of it.
Since one side of the added bypass becomes the interior of this simple closed curve,
and the other side becomes the exterior of this simple closed curve, the face where
the bypass passes through is divided by the bypass into two regions. Hence, by
addition of a bypass in the plane, the number of faces increases by 1. Therefore, the
value of .ψ does not change by addition of a bypass in the plane. Recall that every
connected plane graph with at least two vertices is obtained from a single vertex
graph by repeating addition of a pointed-edge and addition of a bypass in the plane.
In a single vertex graph, .ψ = 1 − 0 + 1 = 2, and the value of .ψ is invariant under
addition of a pointed-edge and a bypass in the plane, yielding .ψ(G) = 2. ⨆
⨅
Lemma 3.3 If a planar graph G has n vertices, .n ≥ 3, then

e(G) ≤ 3n − 6.
.

Proof We may suppose that G is realized in the plane. If G is not connected, then
by adding edges we can change it to a connected plane graph. So, it is enough to
prove the lemma in the case that G is connected.
Now, imagine that for every edge of the plane graph G, we put one stone .⋄ to
each side of the edge, see Fig. 3.4. The number of put stones is .2e(G). Since G has
at least two edges, each face of G must contain at least three stones. Hence, we
54 3 Graphs and Circle-Systems

Fig. 3.4 One stone .⋄ to each

side of every edge

have .3f (G) ≤ 2e(G). From this and three times Euler’s formula .3v(G) − 3e(G) +
3f (G) = 3 · 2, we have .e(G) ≤ 3n − 6. ⨆
⨅

Remark 3.1 Suppose that for every edge of a plane graph we put one stone to each
side of the edge. If a face .α contains exactly three stones, then the boundary of the
face .α must be a 3-cycle. Such a face .α is called a triangular face, or simply a
triangle.
()
Example 3.1 In the complete graph .K5 , we have .v(K5 ) = 5 and .e(K5 ) = 52 =
10. Since .10 /≤ 9 = 3v(K5 ) − 6, .K5 is not planar.
A connected plane graph whose faces (including the outermost face) are all
triangular faces is called a maximal plane graph. From the proof of Lemma 3.3,
we can see that a maximal plane graph with n vertices has exactly .3n − 6 edges and
.2n − 4 faces.

3.2 Quadrangulations

A connected plane graph with minimum degree .≥ 3 whose faces are all quadrilater-
als (that is, the boundary of every face is a 4-cycle) is called a quadrangulation
(of the plane). For example, the plane graph in Fig. 3.3 is a quadrangulation,
but the plane graph in Fig. 3.5 is not a quadrangulation (though its faces are all
quadrilaterals) since its minimum degree is 2.
Since the outer region of a quadrangulation is also a quadrilateral (i.e., its
boundary is a 4-cycle), and its four vertices have degree at least three, it follows
easily that there must be at least four more vertices. Thus, every quadrangulation
has at least eight vertices.
Example 3.2 If G is a quadrangulation and .v(G) = n, then .e(G) = 2n −
4, f (G) = n − 2. This can be seen as follows: Since .2e(G) = 4f (G), from Euler’s

Fig. 3.5 The minimum

degree is 2
3.2 Quadrangulations 55

formula we have .4 = 2n − 2e(G) + 2f (G) = 2n − e(G). Hence .e(G) = 2n − 4

and .f (G) = n − 2.
If we can divide the vertex set of a graph G into two groups so that the vertices
in the same group become nonadjacent, then the graph is called a bipartite graph.
For example, the graph of a cube (Fig. 3.3) is a bipartite graph. It is obvious that a
bipartite graph contains no odd cycle.
Lemma 3.4 Every quadrangulation is a bipartite graph.
Proof First, we show that no odd cycle exists in a quadrangulation. Suppose that a
quadrangulation .G = (V , E) has a .(2k + 1)-cycle C. Since G is a quadrangulation,
there must be quadrilateral faces inside the simple closed curve C, and outside C.
Let .H = (VH , EH ) be the graph obtained from G by removing the edges and
vertices lying outside C. Let m be the number of quadrilateral faces inside C. Then

2|EH | = 4m + 2k + 1,
.

where .| · | denotes the number of elements of the set. This contradicts the fact that
.|EH | is an integer. Hence G has no odd cycle.
Next, let .P , Q be two paths that connect a vertex x to another vertex y, and
let .EP , EQ denote the set of edges in P and in Q, respectively, see Fig. 3.6. We
show that .|EP | ≡ |EQ | (mod 2). In order to do this, it is sufficient to show that
' '
.|E | ≡ |E | (mod 2), where .E
' '
P Q P = EP \ (EP ∩ EQ ), EQ = EQ \ (EP ∩ EQ ).
Note that .EP' ∪ EQ ' is divided into a several subsets, each of which consists of

edges of a cycle. Then, since every cycle of G is an even cycle, we can deduce that
.|E
' ' ' ' ' '
P ∪ EQ | ≡ 0 (mod 2). Hence, .|EP | + |EQ | = |EP ∪ EQ | ≡ 0 (mod 2), and
' '
.|E | ≡ |E | (mod 2).
P Q
Therefore, if there is a path of even length that connects x and y, then every path
connecting x and y has even length. Now, take a pair .x, y of adjacent vertices of G,
and put .w ∈ V \ {x, y} in the same group as x if x and w are connected by a path of
even length. Otherwise, put w in the same group as y. Then no two vertices in the
same group are adjacent. Hence G is a bipartite graph. ⨆
⨅
A 4-cycle of a quadrangulation G is called a separating 4-cycle if each side of
the simple closed curve formed by the 4-cycle contains a vertex of G, see Fig. 3.7.
A quadrangulation is called irreducible if it has no separating 4-cycle. The graph of
a cube in Fig. 3.3 is an irreducible quadrangulation.

Fig. 3.6 Solid lines for P , P •

.........
•
...................................
......... ...
dotted lines for Q •............
...
...
...
...
...
.... ....
.... .......
•
. ...
.... ...
.....
....
....
... ... .. ....
...................................
x•......
.... . . .....•
.. .... .... .... .... ... •
...
. ...
. • •y
.....................................................
....
.... .... . ...
. .....
.. .. ... .
. .
.... .. . .
... .
.... ... ... ...
.... ... ... ......
.... ..... ... ...
..... Q .....
56 3 Graphs and Circle-Systems

Fig. 3.7 A separating

4-cycle of a quadrangulation

Exercise 3.1 Let H be a plane graph with .n ≥ 4 vertices. Prove that if H is a

bipartite graph and has a face that is not a quadrilateral, then .e(H ) ≤ 2n − 5.
Solution If H is disconnected, then by adding the fewest necessary edges, we can
get a connected plane graph that is still bipartite. Then the outermost region of this
new graph is not a quadrilateral face. Hence, we may consider the case that H is
connected. Since H has a face that is neither a triangle nor a quadrilateral, we can
see, by putting one stone to each side of every edge, that .2 × e(H ) > 4 × f (H ).
From this and Euler’s formula we can deduce that .e(H ) < 2n − 4. ⨆
⨅

3.3 Orthogonal-Circle Representations

As we said, two circles in the plane are said to be mutually crossing if they share
two points. And if their tangent lines at a crossing point intersect orthogonally, then
the two circles are said to be orthogonally crossing.
When two circles .C1 and .C2 cross orthogonally, then the line passing through the
center of .C1 and one of the crossing points is a tangent line of .C2 . Hence the center
of .C1 lies outside .C2 . When .C1 and .C2 cross orthogonally, we write .C1 ⊥ C2 . And
.C1 ⊥ C2 ⊥ C3 ⊥ C4 ⊥ . . . means that .C1 ⊥ C2 , .C2 ⊥ C3 , .C3 ⊥ C4 , etc.

Lemma 3.5 Let .C1 , C2 be two circles with centers .z1 , z2 and .z1 /= z2 , respectively.
If a circle C is orthogonally crossing both .C1 , C2 , then the center of C lies on a
fixed line L that is perpendicular to the line .z1 z2 . If .C1 and .C2 are disjoint, then L
is disjoint from both .C1 , C2 .
The line L is called the radical axis of the circles .C1 , C2 .
Proof Put .z1 = (0, 0), z2 = (a, 0), and let .ri be the radius of .Ci , for .i = 1, 2. Let
(x, y) be the center of C. Then, by the Pythagorean theorem,
.

r 2 = x 2 + y 2 − r12
.

r 2 = (x − a)2 + y 2 − r22 .

From this we have .x = 2a1

(a 2 + r12 − r22 ). Since the value of x is constant, the center
.(x, y) of C lies on a line L that is perpendicular to the line .z1 z2 . If .C1 and .C2 are
3.3 Orthogonal-Circle Representations 57

disjoint, then either .a > r1 + r2 or .a < |r1 − r2 | holds. If .a > r1 + r2 , then

a 2 + r12 − r22 (a − r1 )2 − r22

x − r1 =
. − r1 = > 0, (3.1)
2a 2a
and hence L is disjoint from .C1 . And, similarly, L is disjoint from .C2 .
In the case .a < |r1 − r2 |, by assuming .r1 > r2 , we have .r1 − a > r2 . Hence,
by (3.1), L is disjoint from .C1 . (Since .C2 lies inside .C1 , L is also disjoint from
.C2 .) ⨆
⨅
Lemma 3.6 If four circles .Ci , i = 1, 2, 3, 4, are such that .C1 ⊥ C2 ⊥ C3 ⊥ C4 ⊥
C1 , the following two assertions (i) and (ii) hold.
(i) If .C1 , C3 are disjoint, then .C2 , C4 cross each other.
(ii) If .C1 , C3 do not cross and .C2 , C4 do not cross, then four circles meet at a point.
Proof Let .zi be the center of .Ci , i = 1, 2, 3, 4. By Lemma 3.5, the line .z1 z3 and
the line .z2 z4 are perpendicular.
(i) Let .u, v be the intersection points of .C1 , C2 , and .x, y be the intersection points
of .C1 , C4 . The lines .z2 u, z2 v are tangent lines of .C1 , and .u, v are the contact
points. Similarly, the lines .z4 x, z4 y are tangent lines of .C1 , and .x, y are contact
points, see Fig. 3.8. If .C1 and .C3 are disjoint, then by Lemma 3.5, the line .z2 z4
(
is did joint from both .C1 , C3 . Hence, none of the minor arc .uv of .C1 and the
(

minor arc .xy of .C1 contains the other. (If .uv contains .xy, then .z4 lies in the
angular region ./ uz2 v, the line .z2 z4 must intersect .C1 .) Let p be the point on .C1
(
that is nearest to the line .z2 z4 . Then, since both lines .z2 p, z4 p cut .C1 , .uv and
(

.xy contain p. Hence the circle .C2 (which intercepts .uv from .C1 ) and the circle
(

.C4 (which intercepts .xy from .C1 ) cross each other.

(ii) Suppose that .C1 , C3 do not cross, and .C2 , C4 do not cross. Since .C2 , C4 do not
cross, .C1 , C3 are not disjoint by (i). And since .C1 , C3 do not cross, they must
be tangent. Let w be their contact point. Then the line .z2 z4 is a common tangent
line of .C1 , C3 passing through w. In this case, w must lie on .C2 and .C4 . ⨆
⨅

Fig. 3.8 .C4 ⊥ C1 ⊥ C2 C2

u z2
C1 x

z1 p
v z4
y
L
C4
58 3 Graphs and Circle-Systems

... C
.... 1 C ω ..
....
.... ...........
.................. ............................ ....
....
.... ..
.............. ........ ......
.
. ......
.............................. ....... ...... ....
.......................................... ......................................... .... ............................
ω 3 ...................
............ .........
........
....
... .......
....... .... ...............................
... .......
C 2 ...........
... ...............................
. ... ..
..
. ..
.. ...
................ .....
..
.. C 3
..... .... ........................................................................... ....
..... .... ..
.... ........ ...... ........ ......
. . .. ...
.. .. .. ...
... .... .... .. .. ... ... ....
1 5 .... ....
.. ........ 4 C .... ... ... ... . ...
C
.... ...................7...... C .... ..
.
... ...
.. ............. ............................ ... ...................... .......5 ...... ....
.. ...... ... ......... ........ ... ...... ...... .
... ... .
... ......... .. .
. . .... ..
... .. ... . ... ...
... ... ... .... .. ..
..
... ...
. . .
. .
. .
.
.. .. . ..
....
....
.... .....
.. C.. ..
.. ..
....
6 ..... ..........
....... ... .. ..
... ... ..............
4 7 ........ ...
.................. ... ...
............................................................
.
..
...
....
2 6 .. ..
... ...
.. ...
... ...
... ....
.... ...
.. ...
.. .

Fig. 3.9 .Cω is the special circle

For a quadrangulation .G = (V , E), a family .F = {Cv : v ∈ V } of .|V | circles on

the plane is called an orthogonal-circle representation (abbreviated OCR) of G if it
satisfies the following conditions .1◦ , 2◦ .
.1◦ .Cu , Cv cross each other ⇔ Cu ⊥ Cv ⇔ uv ∈ E.
.2
◦ There is a special circle .C ∈ F that encloses .C ∈ F \ {C } unless .C ⊥ C ,
ω v ω v ω
and no other circle encloses another circle.
In any OCR of a quadrangulation .G = (V , E), the following .3◦ also holds, by
Lemma 3.6(ii).
3◦ For each face (quadrilateral) abcd of G, four circles .Ca , Cb , Cc , Cd meet at a
.

point and .Ca ⊥ Cb ⊥ Cc ⊥ Cd ⊥ Ca . (Hence .Ca and .Cc are tangent, and also
.Cb and .Cd are tangent.)

Figure 3.9 right shows an OCR of the graph of a cube (Fig. 3.9 left).
Theorem 3.4 (OCR Theorem) A quadrangulation .G = (V , E) has an OCR if
and only if G is irreducible.
Proof Suppose that there is an OCR .F = {Cv : v ∈ V } of G, and G is still
not irreducible. Let abcd be a separating 4-cycle. Let .abij, abkl be two faces of
G sharing the edge ab. Let .p, q be the intersection points of .Ca and .Cb . Then, by
Lemma 3.6(ii), .Ca and .Cc are tangent at either p or q. Similarly, .Ca and .Ci , and also
.Ca and .Ck are tangent at one of .p, q. This implies that some three of .Ca , Cc , Ci , Ck

are tangent at the same point. In this case, one of the three circles encloses just one
other circle. This contradicts the condition .2◦ of OCR. Thus, if G has an OCR, then
G must be irreducible.
Next, suppose G is irreducible and let .n = |V |. Then .n > 4. By Example 3.2,
.e(G) = 2n − 4. Hence G has a vertex of degree 3. (If the minimum degree is greater

than 3, then we have .2e(G) ≥ 4n and .e(G) ≥ 2n.) Let .ω be a vertex of degree 3,
and denote other vertices by .1, 2, 3, . . . , n − 1, where .N(ω) = {1, 2, 3}.
3.3 Orthogonal-Circle Representations 59

Fig. 3.10 A pair of v ............

... .
v ..
......
v ..
......
... ... . ... .
congruent right triangles for ....
. .
.... .... .
.... ....
.... .... ... .... ...
.... .... ....
each edge uv
.
...
...
.
=⇒ .
...
...
.
...
..
...
xv .
...
...
.
...
..
...
xv
.... ...
. .. ...
. ..
... ...
. ... ...
. ...
... ..
. ..
.
... . .. .
u ............. u .... . u ....
xu xu

b b jj
ab bc ij jk

a c i k
a c (no label)
ad cd
for quadrilateral wijk
d d
Fig. 3.11 Make a quadrangle and a (composite) triangle

In order to find an OCR for G, take .n − 1 positive reals

x1 , x2 , . . . , xn−1
.

and, corresponding to each edge .uv (u /= ω, v /= ω), prepare a pair of congruent

right triangles (made from cardboard) with arms of lengths .xu , xv , see Fig. 3.10. The
hypotenuses of such right triangles are labeled by uv.
For each quadrilateral face of G with vertices .a, b, c, d ∈ {1, 2, . . . n − 1}, create
a quadrangle abcd by attaching four right triangles with hypotenuse .ab, bc, cd, da,
at the arms of the lengths .xa , xb , xc , xd , see Fig. 3.11 left. For each quadrilateral
face .ωij k with .ω as one of its vertices, create a (composite) triangle ij k by attaching
right triangles with hypotenuses .ij, j k at their arms of length .xj , see Fig. 3.11 right.
Then, by Example 3.2, there arise three (composite) triangles and .n−5 quadrangles.
In the edges of these (composite) triangles and quadrangles, each “label” of edge of
G other than .12, 23, 31 appears exactly twice as labels.
Try to attach these triangles and quadrangles along the edges of the same labels
in the plane. Suppose that we succeed to fit them together without gaps and without
overlappings such that the resulting figure becomes a triangle with vertices .1, 2, 3
in the plane. Then, for each vertex .v /= ω, draw a circle .Cv of radius .xv centered
at the point with label v, and finally draw the inscribed circle of the triangle 123 as
.Cω . Then these circles form an OCR of G. Figure 3.12 shows the case where G is

the graph of a cube, see also Fig. 3.9.

Now we ask for what positive reals .x1 , . . . , xn−1 it is possible to fit these triangles
and quadrangles together into a (large) triangle 123 without gaps and overlappings?
60 3 Graphs and Circle-Systems

Fig. 3.12 Draw a circle with 1

radius .xv and center v

2 3

Let .2θi (x1 , x2 , . . . , xn−1 ) denote the sum of the interior angles at vertex i for all
right triangles with vertex i. Then
⎲ xj
θi (x1 , x2 , . . . , xn−1 ) =
. tan−1 ,
xi
j ∈N (i)

where .N(i) is the neighborhood of the vertex i. If .2θi (x1 , x2 , . . . , xn−1 ) = 2π holds
for every .i ∈ V \ {ω, 1, 2, 3}, then it is possible to fit these (composite) triangles and
quadrangles together into a triangle with three vertices .1, 2, 3. Hence the theorem
will follow from the next lemma. ⨆
⨅
Lemma 3.7 Let .G = (V , E) be an irreducible quadrangulation, and .ω ∈ V be
a vertex of degree 3. Let .1, 2, . . . , n − 1 denote the vertices of V other than .ω,
and suppose that .{1, 2, 3} is the nighborhood of .ω. Then there are positive reals
.x1 , x2 , . . . , xn−1 for which

⎲ xj
. tan−1 = π, i = 4, 5, . . . , n − 1
xi
j ∈N (i)

hold, where .N(i) denotes the neighborhood of the vertex i.

Proof Let us regard an n-tuple of real numbers .(x1 , . . . , xn ) as a point in n-
dimensional space .Rn , and call each .xi its ith coordinate. The set of points in .Rn
whose coordinates are all positive is denoted by .Rn+ .
Since .V = {1, 2, 3, . . . , n − 1, ω}, from now on we denote the vertex .ω of G by
n. For a point .x = (x1 , x2 , . . . , xn ) ∈ Rn+ , we put
⎲ xj
.θi (x) = tan−1 .
xi
j ∈N (i)
3.3 Orthogonal-Circle Representations 61

To prove the lemma, it is enough to show that there is an .x ∈ Rn+ such that

π π π
(θ1 (x), θ2 (x), . . . , θn (x)) = ( , , , π, . . . , π )
. (3.2)
3 3 3
holds. Hereafter, we regard each coordinate .xi of .x as a variable, and denote the
right hand side of (3.2) by .(η1 , η2 , . . . , ηn ). From the identity
xj xi π
. tan−1 + tan−1 = (3.3)
xi xj 2

and .e(G) = 2n − 4, we have

⎲
n ⎲( xj xi ) (2n − 4)π ⎲ n
−1
. θi (x) = tan + tan−1 = = ηi (3.4)
xi xj 2
i=1 ij ∈E i=1

for any .x ∈ Rn+ . Since .tan−1 (xj /xi ) is monotonically decreasing in .xi and
monotonically increasing in .xj , .θi (x) is monotonically decreasing in .xi , and
monotonically non-decreasing in the other variables. Since

. lim tan−1 (xj /xi ) = 0

xi →∞

for a fixed .xj , for any .x ∈ Rn+ and .1 ≤ i ≤ n, there is a smallest .ξi = ξi (x) ≥ 0
such that

θi (x1 , . . . , xi−1 , xi + ξi , xi+1 , . . . , xn ) ≤ ηi .

. (3.5)

(If .θi (x) ≤ ηi , then .ξi = 0, whereas, if .θi (x) > ηi , then .ξi is the unique solution of
the equation .θi (x1 , . . . , xi−1 , xi + ξi , xi+1 , . . . , xn ) = ηi .) Define a map .f : Rn+ →
Rn+ by

.f (x1 , x2 , . . . , xn ) = (x1 + ξ1 , x2 + ξ2 , . . . , xn + ξn ),

where each .ξi ≥ 0 is the minimum number satisfying (3.5). When each coordinate
of .x varies in a very small range, then .ξi varies also in a small range; so, .ξi = ξi (x)
is a continuous function of .x = (x1 , x2 , . . . , xn ). If there is a point .x 0 ∈ Rn+ such
that .f (x 0 ) = x 0 , then .θi (x 0 ) ≤ ηi holds for every i. Hence, by (3.4), we must have
.θi (x 0 ) = ηi for all i. Thus, to prove the lemma, it is sufficient to show that there is

a point .x ∈ Rn+ for which .f (x) = x holds.

Now consider the sequence of points in .Rn+

x (1) = (1, 1, . . . , 1), x (2) = f (x (1) ), x (3) = f (x (2) ), x (4) = f (x (3) ), . . . .

.
62 3 Graphs and Circle-Systems

(1) (2) (3)

By the definition of the map f , the sequence .xi , xi , xi , . . . of the ith coordinate
of this sequence .x (1) , x (2) , x (3) , . . . is a monotonically increasing sequence. Put

(ν)
I = {i : lim xi
. = ∞}.
ν→∞

(1) (2) (3)

If .I = ∅, then the sequence .xi , xi , xi , . . . has a limit value for each i, and
hence the limit point .limν→∞ x (ν) = x (∞) exists. Since .ξi (x) is a continuous
function of .x, we have
( (ν+1) (ν) )
.ξi (x (∞) ) = lim ξi (x (ν) ) = lim xi − xi = 0.
ν→∞ ν→∞

Therefore, .f (x (∞) ) = x (∞) . Hence it is sufficient to show that .I = ∅.

Suppose .I = / ∅. Then for every .i ∈ I , there is a .ν for which .θi (x (ν) ) > ηi .
(Indeed, if .θi (x (ν) ) ≤ ηi for all .ν = 1, 2, 3, . . . , then by the definition of the map
f , we must have .xi(1) = xi(2) = xi(3) = . . . , contradicting .i ∈ I .) Since .θi (x)
is monotonically non-decreasing in the variables other than .xi , .θi (x (ν) ) > ηi for
.ν = ν0 implies that .θi (x
(ν) ) ≥ η for all .ν > ν by the definition of the map f . Thus,
0
there is a .ν1 such that for .ν ≥ ν1 ,

i ∈ I ⇒ θi (x (ν) ) ≥ ηi
.

holds. Therefore, .|I | < n, and otherwise, (3.2) holds for sufficiently
∑ large .ν by (3.4).
Put .EI = {ij ∈ E : i, j ∈ I }. In the computation of . i∈I θi (x (ν) ), .π/2 will be
added from each edge in .EI by (3.3), and .tan−1 (xj(ν) /xi(ν) ) will be added from each
edge .ij, i ∈ I, j /∈ I . However, since

lim tan−1 (xj /xi ) = 0

(ν) (ν)
.
ν→∞

for .i ∈ I, j /∈ I , we need not consider the latter case. Therefore

⎲ ⎲
. ηi ≤ lim θi (x (ν) ) = |EI |π/2.
ν→∞
i∈I i∈I

Since we assumed .I = / ∅, the left hand side is positive, and hence .|EI | > 0, which
implies .|I | ≥ 2. Since G is a quadrangulation, there is no edge connecting .1, 2, 3.
Hence I contains a vertex other than .1, 2, 3.
Let us derive a contradiction in the following, by dividing our argument into two
cases.
(i) The case .{1, 2, 3} /⊂ I : Since .|I | ≥ 2, and I contains a vertex other than .1, 2, 3,
we have
⎲
.π + π/3 ≤ ηi ≤ |EI |π/2.
i∈I
3.4 3-Connected Graphs and Radial Graphs 63

Hence .3 ≤ |EI |, and hence .4 ≤ |I |, which means that I contains at least two
vertices other than .1, 2, 3. Since the graph .(I, EI ) is a bipartite graph, we have
.|EI | ≤ 2|I | − 4 for .|I | ≥ 4. Therefore,

⎲
(|I | − 2)π + 2π/3 ≤
. ηi ≤ |EI |π/2 ≤ (2|I | − 4)π/2,
i∈I

a contradiction.
(ii) The case .{1, 2, 3} ⊂ I : In this case, we must have .n ∈ I . (Otherwise, we have

⎲
3
. lim θn (x (ν) ) = lim tan−1 (xi(ν) /xn(ν) ) = 3π/2 > π = ηn ,
ν→∞ ν→∞
i=1

which contradicts the definition of the sequence .x (ν) .) Hence .|I | ≥ 4. The
plane graph .H := (I, EI ) is clearly not a 4-cycle, and H has a face that is not
a quadrilateral. (If every face of H is a quadrilateral, then the boundary of the
outer face of H is a 4-cycle of G that contains a vertex of G in its interior. Since
H is a proper subgraph of G, this contradicts the fact that G is irreducible.)
Hence, by Exercise 3.1, we have .|EI | ≤ 2|I | − 5. Then,
⎲
π(|I | − 3) + 3(π/3) =
. ηi ≤ |EI |π/2 ≤ (2|I | − 5)π/2,
i∈I

which is a contradiction. ⨆
⨅
Remark 3.2 Theorem 3.4 is a reformulation of a result obtained by Brightwell and
Scheinerman [128] as OCR theorem for a quadrangulation. Their proof is slightly
improved in our OCR theorem.

3.4 3-Connected Graphs and Radial Graphs

For a graph .G = (V , E) and a subset .S ⊊ V , .G−S denotes the graph obtained from
G by removing the vertices in S and removing edges that are incident to vertices in
S. If .G−S is disconnected, then S is called a cut set of G. In a complete graph, there
is no cut set. For a connected graph, a cut set with minimum cardinality is called the
minimal cut set. A connected graph with at least .k + 1 vertices is called k-connected
if the graph has no cut set consisting of .k − 1 vertices. Thus, if G is a 3-connected
graph, then for any two vertices .x, y of G, .G − {x, y} is connected. A complete
graph .Kn is 3-connected for .n ≥ 4. The graph of a cube is also 3-connected as
easily verified. The minimum degree of a 3-connected graph G is at least 3. (If a
vertex v of G has degree at most 2, then its neighborhood .N(v) becomes a cut set
with two vertices, and hence G is not 3-connected.)
64 3 Graphs and Circle-Systems

Lemma 3.8 Let S be a minimal cut set of a connected graph G, and let H be a
connected component of .G − S. Then every .x ∈ S is adjacent to some vertex in H .
Proof Suppose that .x ∈ S is not adjacent to any vertex of H . Then H is still a
connected component of .G − (S \ {x}). This contradicts the fact that S is a minimal
cut set. ⨆
⨅
Theorem 3.5 Let .G = (V , E) be a graph such that .|V | ≥ 4. If for every .x ∈ V
there is a cycle of .G − {x} that contains .N(x), then G is 3-connected.
Proof Suppose that G is not 3-connected, and let S be a minimal cut set, .|S| ≤ 2.
Let x be a vertex in S. Then .N(x) is contained in a cycle C of .G − {x}. Let .C ' be
the graph obtained from C by removing a vertex in S other than x (if it exists), and
edges adjacent to the vertices in S. Since a cycle cannot be made disconnected by
removing a vertex, .C ' is connected. Hence .C ' is contained in a connected component
H of .G − S. Since .N(x) is contained in C, a connected component of .G − S other
than H has no vertex that is adjacent to x. This contradicts Lemma 3.8. ⨆
⨅
Exercise 3.2 Prove that a maximal plane graph with at least four vertices is 3-
connected.
Solution Let G be a maximal plane graph with at least 4 vertices. For a vertex x
of G, let .v1 , v2 , . . . , vk be the vertices (in anti-clockwise order around x) that are
adjacent to x. Since each face of G is a triangle, .v1 v2 , v2 v3 , . . . , vk v1 are all edges
of G. Hence these k vertices and these k edges form a cycle in .G − {x}. Since .N(x)
is contained in this cycle, G is 3-connected by Theorem 3.5. ⨆
⨅
Lemma 3.9 In a 3-connected plane graph G, the following properties .(1) and .(2)
hold. .(1) For every edge .e = xy of G, its both sides are distinct faces of G. .(2) The
boundaries of two different faces of G cannot share a pair of non-adjacent vertices.
Proof
(1) Suppose that both sides of an edge .e = xy are one and the same face .α of G.
Then there is a simple closed curve .Γ that crosses the edge e at a point, and
except that point, the curve lies in the interior of the face .α. Then .Γ meets G
only at the edge e, and never meets .G − {x}. Since .deg x ≥ 3, there is a vertex
.z /= y of G that is adjacent to x. Then y and z do not lie in the same side of .Γ .

Hence, .G − {x} is disconnected, contradicting the fact that G is 3-connected.

(2) Suppose that the boundaries of two different faces .α, β of G share a pair of
non-adjacent vertices .x, y. Then there is a simple curve connecting x and y
and passing through the interior of .α except the endpoints .x, y. Similarly, there
is a simple curve connecting .x, y passing through the interior of the face .β.
Attaching there two simple curves end to end, we obtain a simple closed curve
.Ω, that meets G only at .x, y. Then each side of .Ω contains a vertex of .G −

{x, y}, and hence .G − {x, y} is disconnected. This contradicts the fact that G is
3-connected. ⨆
⨅
3.5 The Coin Graph Theorem 65

Fig. 3.13 A plane graph and its radial graph

For each face of a plane graph G, put one point inside the face. Such a point
is called a face-point. The face-point inside the face .α is denoted by .α ∗ . For each
face of .α, connect the face-point .α ∗ and the vertices on the boundary of .α by edges
(simple curves) so that these edges do not meet in the midway. These edges are
called radial edges. After that, remove the original edges of G. Then we have a
plane graph whose edges are all radial edges. This graph is called the radial graph
of G and denoted by .R(G). The vertices of the radial graph .R(G) are divided into
two groups, the vertices of G, and the face-points. The radial graph .R(G) is clearly
a bipartite plane graph. Figure 3.13 shows a plane graph and its radial graph.
Lemma 3.10 The radial graph of a 3-connected plane graph G is an irreducible
quadrangulation.
Proof Since G is 3-connected, both sides of an edge are distinct faces. Hence,
each face of .R(G) is a quadrilateral whose one diagonal is an edge of G. Since
the degree of each vertex of G is at least 3, each vertex of G is a common point
of the boundaries of at least three faces of G. Hence the degree of each vertex of
.R(G) is at least 3, and hence .R(G) is a quadrangulation. Suppose that .R(G) is not

irreducible. Then .R(G) has a separating 4-cycle .uα ∗ vβ ∗ , where .u, v are vertices of
G and .α ∗ , β ∗ are face-points. Since there are vertices of .R(G) inside this 4-cycle,
and outside this 4-cycle, there are quadrilaterals inside this 4-cycle and outside this
4-cycle. This implies that .G − {u, v} is disconnected, contradicting that G is 3-
connected. Thus, .R(G) is an irreducible quadrangulation. ⨆
⨅

3.5 The Coin Graph Theorem

For a given graph .G = (V , E), a family of mutually nonoverlapping disks .D =

{Dv : v ∈ V } in the plane that satisfies the condition

.uv ∈ E ⇔ Du and Dv are tangent

is called the coin representation of G. A graph for which a coin representation exists
is called a coin graph.
66 3 Graphs and Circle-Systems

Let .G = (V , E) be a coin graph, and .D = {Dv : v ∈ V } be its coin

representation. Then, by representing each vertex v by the center of .Dv , and
representing an edge .uv ∈ E by a line segment connecting the centers of .Du , Dv ,
we have a realization of G in the plane. Thus, we have the following theorem.
Theorem 3.6 Every coin graph is a planar graph. ⨆
⨅
The converse is also true.
Theorem 3.7 (Coin Graph Theorem) Every planar graph is a coin graph.
Proof A graph with at most 3 vertices is obviously a coin graph. Let G be a plane
graph with at least four vertices. If G is not connected, then by taking a face point
in the outer face, and adding some edges that connect the face point to a vertex of
each component of G, we can get a connected plane graph. If this connected plane
graph has a coin representation, then, by removing the disks corresponding to the
face point, we can get a coin representation of G. Thus, we may suppose that G is
connected. If G has a face that is not a triangle, then take a face-point inside the face,
and divide the face into triangular faces by adding radial edges. If this new graph
has a coin representation .D, then by removing from .D the disks corresponding to
the added face-points, we have a coin representation of the original plane graph.
Therefore, to prove the theorem, it is enough to show that every maximal plane
graph G with at least 4 vertices has a coin representation. Since G is 3-connected
by Exercise 3.2, its radial graph .R(G) is an irreducible quadrangulation. Let .ω
be the face-point of the outermost face of G. Then the degree of .ω is 3. By the
OCR theorem, there is an OCR .F of .R(G) in which any circle other than the
circle corresponding to .ω does not enclose any other circle. By removing the circles
corresponding to face-points and then replacing each circle by disks, we have from
such a family .F a coin representation of G. ⨆
⨅
In a coin representation .D = {Dv : v ∈ V } of a planar graph .G = (V , E),
connect the centers of mutually tangent disks with line segments. Then the line
segments never cross. Hence we have the following corollary.
Corollary 3.2 Every planar graph can be realized in the plane such that each edge
of it is a line segment. ⨆
⨅
Remark 3.3 The coin graph theorem was discovered by P. Koebe in 1936, and
rediscovered again by E. Andreev in 1970 and by W. Thurston in 1985 (see [776]).
This theorem is also called the circle-packing theorem of Koebe-Andreev-Thurston.
In the book [610, Ch. 8] of J. Pach and P. K. Agarwal a direct proof of the coin graph
theorem is presented.
3.6 The Theorem of Steinitz 67

3.6 The Theorem of Steinitz

We already noted in Corollary 3.1 that the graph of a convex polyhedron is a planar
graph. Let .Π be a convex polyhedron in .R3 . For a vertex v of .Π , let .α, β, γ , . . . be
the faces (facets) of .Π (in clockwise order) around the vertex v. Then the boundary
of the union of these faces becomes a cycle of the graph of .Π , and the neighbors
of v are vertices contained in this cycle. Hence, by Theorem 3.5, the graph of .Π is
3-connected.
Theorem 3.8 The graph of a convex polyhedron is a 3-connected planar graph.
⨆
⨅
The converse is also true, known as Steinitz’ theorem.
Theorem 3.9 (Steinitz) Every 3-connected planar graph can be realized as the
1-skeleton of a convex polyhedron in .R3 .
Proof Let .G = (V , E) be a 3-connected plane graph. Then its radial graph .R(G)
is an irreducible quadrangulation, and hence there is an OCR .F of .R(G). Roughly
speaking, the proof goes along the following line. By applying the inverse map of
the stereographic projection, this OCR .F is mapped to a system of circles on a
sphere. Then the planes determined by the circles on the sphere corresponding to
face-points of .R(G) enclose a convex polyhedron whose 1-skeleton is a realization
of G.
Now the details. Suppose that an OCR .F of .R(G) is given in the xy-plane in
.R , and the special circle .Cω (which encloses all circles that do not cross .Cω ) is
3

centered at the origin .O = (0, 0, 0) with radius r. Put .P = (0, 0, r) and let .S be
the sphere with diameter OP . The sphere .S is tangent to the xy-plane at O. Let .ϕ
denote the inverse mapping of the stereographic projection .S \ {P } → {xy-plane}
from the point P . Then .ϕ(Cω ) is the great circle (the equator) of .S, and for other
circles .C ∈ F, its image .ϕ(C) is a small circle of .S. Indeed, if C is enclosed by .Cω ,
then .ϕ(C) is a small circle lying in the “southern hemisphere” of .S, and if C crosses
.Cω orthogonally, then the center O lies outside of C, and hence .ϕ(C) is also a small

circle on .S. Thus, .{ϕ(C) : C ∈ F, C /= Cω } is a family of small circles on .S, and

for each .C ∈ F, no circle lies in the interior of the “smaller” cap bounded by C.
For each face-point .α ∗ of .R(G), let .πα denote the plane determined by .ϕ(Cα ),
where .Cα ∈ F is a circle corresponding to .α ∗ . If the boundaries of the faces .α, β
have an edge uv in common, then, in .R(G), .α ∗ uβ ∗ v is a quadrilateral face, and
hence the circles .ϕ(Cα ) and .ϕ(Cβ ) are tangent to each other. The intersection of
the planes .πα and .πβ is a common tangent line of the circles .ϕ(Cα ) and .ϕ(Cβ ),
and also a tangent line of .S. Let .α, β, γ , . . . , δ be the faces of G around the
vertex v in clockwise order. Then the intersection lines of the pairs of planes
.(πα , πβ ), (πβ , πγ ), . . . , (πδ , πα ) are all tangent lines of .S orthogonally crossing

the circle .ϕ(Cv ). In this case, these tangent lines of .S intersect at one and the same
point .v̂, see Fig. 3.14. Therefore, the planes .πα for all faces .α of G enclose the
68 3 Graphs and Circle-Systems

Fig. 3.14 Tangent lines of .S

orthogonal to a small circle

Cv −→ v̂

convex polyhedron with vertices .{v̂ : v ∈ V }, and its 1-skeleton is a realization of

G. ⨆
⨅

Remark 3.4 Steinitz’ theorem was obtained by E. Steinitz in 1922. B. Grünbaum

[340] presented an improved proof of this theorem, but it is long and not easy. A
comparatively easy proof is given by G. M. Ziegler [834], and it is also mentioned
there that Steinitz’ theorem can be proved using the OCR theorem. An elementary
simple proof of the special case of Steinitz’ theorem that states “every maximal
planar graph with more than 3 vertices can be represented by a 1-skeleton of a
convex polyhedron” was obtained by H. Maehara [518]. D. Barnette [61] gave an
intuitive explanation of the following fact: Every 3-connected planar graph with
maximal degree 3 can be realized by the 1-skeleton of a convex polyhedron.

3.7 Exercises

3.1 Prove the hand-shaking lemma.

3.2 A sphere is divided into several regions by n ≥ 3 great circles. Assuming that
no three great circles intersect at the same point, find the number of regions.
3.3 Prove that if a connected graph is not a complete graph then ( it) has a cut set.
3.4 Prove that if a graph with n(≥ 3) vertices has more than n−1 2 edges, then the
graph is connected.
3.5∗ Prove that every quadrangulation has at least eight vertices.
3.6 Find an irreducible quadrangulation that has an odd number of vertices.
3.7∗ Let G be a 3-connected plane graph. For every edge e of G, we can draw a
simple curve e∗ that connects the face points in the both sides of e and crosses
the edge e at a point. This curve e∗ is called a dual edge of e. It is also possible
that no two dual edges cross each other in the midway. The plane graph whose
vertices are the face points of G and whose edges are the dual edges of the
edges of G is called the dual graph of G, and it is denoted by G∗ . Now, a
question. Is it possible to re-draw G and G∗ on the plane so that the following
(i) and (ii) hold?
(i) All edges of G and G∗ are line-segments or half lines.
(ii) Every edge of G and its dual edge cross each other orthogonally.
3.8 Notes 69

(We regard that the face-point w ∗ of the outermost face of G is lying at infinity,
and the edges of G∗ incident to w ∗ are half-lines.)

3.8 Notes

The notes to Chap. 3 here restrict to the following two important themes: results
around the coin graph theorem and the famous theorem of Steinitz on polyhedral
graphs.

3.8.1 Notes on the Coin Graph Theorem

A unit disk graph is the intersection graph of a family of unit disks in the plane. If
the disks form a packing (i.e., they do not overlap), it is called a unit coin graph (or
penny graph, or sometimes also tangency graph). An important observation in this
direction goes back to Koebe [461] and is named today the Koebe-Andreev-Thurston
theorem or coin graph theorem (see Koebe [461] and Thurston [776]). Namely, the
vertices of any planar graph can be represented as midpoints of nonoverlapping
circular disks in the plane such that any two of them are tangent to each other if
the corresponding vertices are joined by an edge in the graph. And for triangulated
planar graphs this representation is unique, up to conformal transformations of the
plane (there are various further cases of uniqueness). Thus, connecting the centers of
the disks via segments in the touching case, a straight-line drawing of the considered
graph is established.

Book Parts and Surveys

Various problems and results on unit coin graphs can be found in several problem
books or collections, see, e.g., Hartsfield and Ringel [381, Problems 8.4.6, 8.4.7
and 8.4.8] or Brass, Moser and Pach [125, Sections 5.7 and 9.2]. In connection
also with Steinitz’s theorem, discussions and applications of the Koebe-Andreev-
Thurston theorem and its extension due to Brightwell and Scheinerman [128] are
given in monographs by Pach and Agarwal [610, Chapter 13], Ziegler [834, Notes
to Lecture 4], Matoušek [560, Section 5.3], Grünbaum [340, Section 13.7], Felsner
[263, Section 2.7], Gruber [336, Sections 15 and 34], and Lovász [512, Chapter 5].
A closely related little survey is due to Sachs, see [683]. He discusses several
known generalizations of the Koebe-Andreev-Thurston theorem and related results
as well as interesting interconnections between them. Pisanski and Randić [637]
survey (in a beautifully illustrated way) interesting relations between graph theory
and geometry. In Sections 1 and 2 there, properties of and relations between coin
graphs and polyhedral graphs (in view of Steinitz’s theorem) are presented, like
70 3 Graphs and Circle-Systems

also Koebe’s result. And in the survey [399], Hlinený and Kratochvíl give an
overview to recognition-complexity results for disk graphs and coin graphs. Koebe’s
theorem and its consequences are one main theme. They also show several results
on coin graphs in higher dimensions. Among other things, it is proved here that the
recognition of unit-ball coin graphs is NP-hard in dimensions 3, 4, 8, and 24. The
paper contains interesting constructions and a helpful bibliography.

Some Research Papers

The Koebe-Andreev-Thurston theorem, unit coin graphs and extensions thereof

have various further properties and applications, e.g. with respect to distance
problems in finite point sets, geometric graphs and polytope theory. We give here
some selected papers presenting such investigations.
Schramm [711] proves that any 3-connected planar graph even represents a 3-
polytope whose edges are all tangent to the unit sphere such that the origin is the
barycenter of the contact points. Sachs [683] presents several known generalizations
of the Koebe-Andreev-Thurston theorem. Also it is clarified that Schramm’s result
remains true if the unit sphere is replaced by the boundary of any smooth, strictly
convex body.
It is known that to find a maximum independent set (i.e., an independent set that
is not a subset of any other independent set) in a unit disk graph is NP-hard. In
[150], Cerioli et al. extend this result to penny graphs. Additionally they prove that
to find a minimum clique partition in a penny graph is NP-hard, too. They also offer
linear-time approximation algorithms for the computation of clique partitions, one
for unit disk graphs and one for penny graphs.
Eppstein [244] studies triangle free penny graphs. Squaregraphs are the graphs
that can be embedded in the plane such that each bounded face is a 4-cycle, and
every vertex not on the unbounded face has degree at least 4. The author first
discusses results about these graphs and their relations to other families of planar
graphs, and he derives then deeper results on them. These refer to degeneracy,
diameter, and upper bounds on edge numbers.
Colin de Verdière introduces in [173] a new spectral invariant of graphs
which is shown to be monotonic regarding edge removal and an operation having
homeomorphic reduction as a subcase. It is essential for planarity of graphs.
Particularly, it turns out that this invariant is at most 3 iff the respective graph is
planar. Alfakih [21] studies this invariant for penny graphs. He proves lower bounds
on the graph invariant introduced in [173] when the complement of the considered
graph is a penny graph. And Mitchell [577] extends results of Alfakih [21] getting
insights that apply to complements of contact graphs of unit spheres in arbitrary
dimensions.
3.8 Notes 71

3.8.2 Notes on the Theorem of Steinitz

Steinitz’s theorem belongs to the interesting little field of polytopal graphs, and
there particularly to the three-dimensional case. For this whole field we refer to the
following general references: the nice book [61] of Barnette and the excellent survey
[339] of Grünbaum. The Steinitz problem, namely to find necessary and sufficient
conditions for a cell-composed sphere to be isomorphic to the boundary complex
of a (higher dimensional) polytope (or to find a complete intrinsic characterization
for boundary complexes of convex polytopes, or to give a characterization of the
polytopal (.n − 1)-spheres among all (.n − 1)-spheres) is in general open. But for
convex three-dimensional polytopes (shortly called 3-polytopes), Steinitz’ theorem
holds. It says that a given graph is the graph of a 3-polytope if and only if it is
simple, planar, and 3-connected. Due to this result, the graphs with these properties
are usually called polyhedral graphs.

Book Chapters

For the history of his famous theorem we refer to Steinitz himself (see [758] and
[759]), Sachs [683], Ziegler [834, Lecture 4 and its notes], and Grünbaum [340,
Chapter 13]; Section 13.7 from [340] should be particularly mentioned since there
different proofs, extensions, and analogues until 2003 are cited.
Extensive representations (with full proofs or explanations of proof methods and
related questions) are given by Grünbaum [340, Chapter 13], Ziegler [834, Lecture
4 and its notes], Richter-Gebert [666], Matoušek [560, Section 5.3], Felsner [263,
Chapters 1 and 2], Gruber [336, Chapter 15], and Lovász [512, Chapters 2 and 3].
Also in the book [251, Chapter III] of Ewald, strongly related to algebraic geometry,
the Steinitz problem is taken care of.
In all these references further topics and notions are introduced or applied, since
they efficiently help in proofs or for extending the field. Because we do not go
into details here, we mention only some of them. All of them are well-studied
mathematical tools or notions (and therefore easy to find). We mean here Schlegel
diagrams, the stereographic projection, Möbius transformations, semialgebraic sets,
realization spaces, the Koebe-Andreev-Thurston circle-packing theorem and the
Koebe-Brightwell-Scheinerman theorem (the latter used, e.g., by Gruber for his
proof of the Steinitz theorem given in [336, Section 15.3]).

Surveys

Besides in books, Steintz’s theorem is also discussed in several surveys. In his

exposition [339] on polytopal graphs (but still mainly treating the three-dimensional
case), Grünbaum excellently presents typical proof methods, types of related results,
and open questions around polyhedral and polytopal graphs. A large variety of
72 3 Graphs and Circle-Systems

problems related to the combinatorial boundary structure of polytopes is surveyed

in the excellent exposition [252] of Ewald et al.; in its second section partial
results regarding the Steinitz problem for .n ≥ 4 (e.g., for simple complexes
or combinatorial spheres with few vertices) are presented. In Section 5.2 of the
handbook article [68], Bayer and Lee present, among other things, Balinski’s
result (that the graph of any n-polytope is still n-connected) and discuss also
other connectivity criteria. And in Sect. 6.3, numbers of combinatorial types of 3-
polytopes are studied, also with the help of Steinitz’ theorem. Bokowski’s handbook
article [99] gives a good survey of the results related to the Steinitz problem from
the viewpoint of oriented matroids. In the first part of his handbook survey [437],
treating 3-polytopes, Kalai discusses Steinitz’s theorem together with the classical
theorems of Kotzig, Eberhard, and Motzkin.

Some Research Results and Directions

One interesting research direction directly coming from Steinitz’s theorem can
be described as follows: Steinitz’s theorem gives a completely combinatorial
characterization of the face lattices of convex 3-polytopes, and already Steinitz
himself observed that his approach implies that for any 3-polytope the set of all
its realizations is a trivial topological set (i.e., realization spaces of 3-polytopes
are contractible and therefore connected). Further results in this direction and for
dimensions .n ≥ 4 showed that realization spaces of n-polytopes with .n + 4 vertices
can have arbitrary homotopy type. Richter-Gebert investigates this and presents
and uses also related basic results of Mnëv [580] in his book [666]. He studies
the structure of realization spaces of polytopes in fixed dimension and proves
a universality theorem for 4-polytopes. This theorem says that for any primary
basic semialgebraic set V there is a 4-polytope whose realization space is stably
equivalent to V . (Recall that a basic semialgebraic set is a set defined by polynomial
equalities and polynomial inequalities, and that a semialgebraic set is a finite union
of basic semialgebraic sets.)
There are many deep and surprising results around such problems. E.g., Barnette
[62] constructs a certain complex of six triangles that, when embedded in a 3-
sphere, prevents the polyhedrality of this sphere. (A triangulated 3-sphere is called
polyhedral if it is isomorphic to the boundary complex of a 4-polytope.) Improving
former bounds, Onn and Sturmfels [604] show that any 3-polytope with m
vertices can be realized in Euclidean 3-space with all vertex coordinates being
3
integers whose absolute value does not exceed .m169m . Ziegler [833] poses research
problems related to Steinitz’s theorem and treating 4-polytopes. He discusses known
results, methods, constructions, and examples - the latter with detailed proofs.
Ewald and Schulz [253] present (.n − 1)-spheres which cannot be embedded in
n-space as starshaped sets. Namely, they verify that for any .n ≥ 4 non-starshaped
simplicial (.n−1)-spheres in n-space exist that have .n+8 vertices. E.g., they also ask
whether this number .n + 8 is best possible and conjecture that every (.n − 1)-sphere
3.8 Notes 73

with .n + 4 vertices has a starshaped or even convex embedding in n-space. They

show the relations of their investigations to algebraic geometry (more precisely, to
toric varieties).
Searching analogues and extensions of the classical Steinitz theorem, Takayama
and Hibi [771] try to find combinatorial characterizations of graphs referring
to Cohen-Macaulay simplicial complexes. They derive a necessary and sufficient
condition for a finite simple graph to be the graph of a two-dimensional simplicial
complex that is Cohen-Macaulay over an arbitrary field. They also present the
example of a graph which cannot be the graph of any Cohen-Macaulay simplicial
complex over a field of characteristic 2, but is the graph of a two-dimensional
Cohen-Macaulay simplicial complex over a field of characteristic ./= 2. Almoham-
mad, Lángi and Naszódi [22] extend Steinitz’s theorem to ball polytopes in 3-space,
i.e., to intersections of finitely many unit balls.
Belotti et al. [71] treat realization spaces of polytopes, in particular realizations
of 3-polytopes whose edges are tangent to the unit sphere (such realization spaces, in
terms of the vertex coordinates or the facets, are semialgebraic sets). They consider
Koebe realizations (with edges tangent to the sphere) and Springborn realizations
(Koebe realizations whose edge barycenter is the origin), see Koebe [461] and
Springborn [753]. Based on the fact that Koebe and Springborn realizations admit
descriptions as semialgebraic sets defined by rational polynomials, the authors study
the minimal degrees of such polynomials.
Chapter 4
Spherical Geometry I

If I were again beginning my studies, I would follow the advice

of Plato and start with mathematics.
— Galileo Galilei

In this chapter, we show that geodesic curves on the sphere are great circular arcs,
and we verify that the cylindrical projection preserves the areas of figures. Girard’s
formula for the area of a spherical triangle is also proved. From the area formula
for spherical polygons obtained by applying Girard’s formula, Legendre’s proof of
Euler’s polyhedral formula is derived. The theorem on inscribed angles is presented,
and the notion of polar set is introduced.

4.1 Geodesic Segments

Let us denote by .S(r), or simply by .S, the sphere of radius r centered at the origin
.O = (0, 0, 0) in the 3-dimensional Euclidean space .R3 . Every circle on the sphere
.S is obtained as the section of .S by a plane in .R . The section of .S by a plane that
3

passes through the origin O is a great circle on .S. A circle on .S that is not a great
circle is called a small circle on .S. Every circle on .S divides .S into two regions,
and each such region is called a spherical cap or simply a cap. If the circle is a
great circle, each region is called a hemisphere instead of a cap. For a point .X ∈ S,
∗
.X denotes its antipodal point, that is, the intersection point of the line XO and
∗ ∗
.S other than X. Clearly, .(X ) = X. For any two distinct points .A, B on .S, there

is a great circle passing through .A, B. (If .A∗ /= B, then there is unique plane that
passes through .A, B, O, and the great circle that passes through .A, B is unique.
On the other hand, if .A∗ = B, then there are infinitely many great circles that pass
through .A, B.) When this great circle is divided by .A, B, the shorter arc (minor arc
or semicircle) is simply denoted by AB. (Since the line segment connecting .A, B is
also denoted by AB, we need to judge from the context which one AB represents.)

© The Author(s), under exclusive license to Springer Nature Switzerland AG 2024 75

H. Maehara, H. Martini, Circles, Spheres and Spherical Geometry,
Birkhäuser Advanced Texts Basler Lehrbücher,
https://doi.org/10.1007/978-3-031-62776-7_4
76 4 Spherical Geometry I

Among the curves on .S that connect A and B, the great circular arc AB is the
shortest curve.
In general, among the curves on a smooth surface M in .R3 that connect two given
points .P , Q ∈ M, the shortest one is called a geodesic segment connecting P and
Q.
Similarly to a point on the surface of the Earth, a point on .S can be represented
by its longitude .τ, −π ≤ τ ≤ π (where the meridian passing through .(r, 0, 0) has
longitude 0; east longitude has .+ sign, west longitude has .− sign) and its latitude
.θ, −π/2 ≤ θ ≤ π/2 (where north latitude has .+ sign, south latitude has .− sign).

Thus, .(x, y, z)-coordinates and .(τ, θ ) of a point on .S are related by

x = r cos τ cos θ, y = r sin τ cos θ, z = r sin θ.

.

Let .A, B ∈ S be two points, .A∗ /= B. Then .A, B determine a great circle
uniquely. Let us show that among the curves on .S that connect A and B, the
great circular arc AB has the minimum length. We may suppose, by rotating .S if
necessary, that .A, B have the same longitude. Let .γ (t), a ≤ t ≤ b, be a curve on .S
that connects A and B:

γ (t) = (x(t), y(t), z(t)) (a ≤ t ≤ b), γ (a) = A, γ (b) = B.

.

The length of the curve .γ (t) is given by

⎰ b/
. ẋ(t)2 + ẏ(t)2 + ż(t)2 dt,
a

where .( ˙ ) denotes the derivative by t. (Regarding t as the time parameter,

ẋ(t), ẏ(t), ż(t) are the velocity
. √ of the point .γ (t) in the direction of .x, y, z-axes,
respectively, and hence . ẋ(t)2 + ẏ(t)2 + ż(t)2 is the magnitude of the velocity
vector of .γ (t). Hence, by integrating it from .t = a to .t = b, we have the distance
that the point .γ (t) has moved, namely, the length of the curve .γ (t).) Let .θ (t), τ (t)
be the latitude and the longitude of .γ (t). Then

ẋ(t) = −r τ̇ (t) sin τ (t) cos θ (t) − r θ̇(t) cos τ (t) sin θ (t)
.

ẏ(t) = r τ̇ (t) cos τ (t) cos θ (t) − r θ̇(t) sin τ (t) sin θ (t)
ż(t) = r θ̇(t) cos θ (t),

from which we get .ẋ(t)2 + ẏ(t)2 + ż(t)2 = r 2 τ̇ (t)2 cos2 θ (t) + r 2 θ̇ (t)2 . Therefore,
⎰ b/ ⎰ b
.the length of γ = ẋ(t) + ẏ(t) + ż(t) dt ≥ r
2 2 2 |θ̇(t)|dt
a a
|⎰ |
| b |
≥ r || θ̇ (t)dt || = r|θ (b) − θ (a)|.
a
4.2 Cylindrical Projections 77

Since we assumed that A and B have the same longitude, .r|θ (b) − θ (a)| is equal
to the length of the great circular arc AB. Thus, the great circular arc AB has the
minimum length among the curves on .S connecting A and B.
Let us state this as a theorem.
Theorem 4.1 For two points .A, B on a sphere .S(r), the great circular arc AB is a
geodesic segment connecting A and B. ⨅
⨆

4.2 Cylindrical Projections

In 3-dimensional space .R3 , a plane or a line that is parallel to the xy-plane is simply
called a horizontal plane or a horizontal line. Let .Γ be the cylinder (the surface of
revolution) obtained by rotating the line segment connecting .(r, 0, r) and .(r, 0, −r)
around the z-axis. The sphere .S(r) is inscribed in this cylinder .Γ and tangent to it
at the equator. Put

Š = S(r) \ {(0, 0, r), (0, 0, −r)}.

.

For .X = (x, y, z) ∈ Š, let .ϕ(X) ∈ Γ denote the intersection of the horizontal ray
−−→
.ZX X and the cylinder .Γ , where .ZX = (0, 0, z). Two points X and .ϕ(X) have the

same z-coordinate, and if X lies on the equator of .S, then .ϕ(X) = X, see Fig. 4.1.
The map

Š ϶ X ⍿→ ϕ(X) ∈ Γ
.

is called the cylindrical projection.

Remark 4.1 A point on .Š is represented by .(τ, θ ) using its longitude .τ ∈ [−π, π ]
and its latitude .θ ∈ [−π/2, π/2], and a point on .Γ can be also represented by .(τ, h)
using its longitude .τ and its height .h ∈ [−r, r]. Then

ϕ(τ, θ ) = (τ, r sin θ ) ∈ Γ.

.

Fig. 4.1 The cylindrical

projection
78 4 Spherical Geometry I

z. B M ................................A
●....................................● ..... ... ... ... ... ... ... ... ... ... .P
...● .●
.... .
. .. .. .. ..
.... ..
.. .. .. . ...
.........
...
..
..
..
.
..
.
..
..
.
..
.
...
..
θ ... .
.. ..
. ...
.
.. .. ... ... .. ..
.. ... ... . ...
....................................................................... ... ..............
..●
........ ....
A ● .......
.
● . ....
.. ..........................................................................................
. ...
...
...
.
... ......
. .. ....
.
........
........
........
.. ... ... ... ... .... . ...
.........
... ... .... . .
....... ...... ●
.... ...............
.
.
.. .. ... ... ... ... . .. . . .
. ..
. ... ... ... ... ... ..... ... ..... ........... ........
.....
..
.... . ... ... .. ... .... ....
................. ...........
.... . .. ........
...●
M .
.●..
...... ●. .
.. ............................................................................................................. ....
.
........ ....
....
. ....
....
..
.........................
.. .....
... ..
...
..
...
.
.. ... ... ... ... . . . ... . ... ... ... ... ..
. .. .... .
....... ........ ....
.
.. .. ... . . ... . . ... ... .... .... ........
......... ........
...
..
..... . .. ... .. .... .......... ........ ....
●......... ●.... .. .... .............. .......... ....
B ............... ......
. ....
.... ................................................................ ....
....................................................................................................................... ...... ......
.. ....... .......
........ .......
.. ........ ........
.........
........... ..... ..
...........
............. ...
..................... .............
......................................................

Fig. 4.2 The development of the surface of revolution

Theorem 4.2 (Archimedes-Lambert) The cylindrical projection preserves the

area, that is, for any region W on .Š, W and .ϕ(W ) have the same area.
Remark 4.2 Archimedes proved that the area of .S is equal to the area of .Γ .
J. H. Lambert generalized this as in Theorem 4.2. See, e.g., [255, 569].
To prove Theorem 4.2, we start with a lemma.
Lemma 4.1 Let AB be a line segment of length d in the xz-plane which does
not intersect the z-axis. By rotating AB around the z-axis, we have a surface of
revolution (frustum of right cone), see Fig. 4.2 left. The area of this surface of
revolution is equal to .d × l, where l is the length of the circle obtained as the locus
of the midpoint M of AB.
Proof If the line segment AB is parallel to the z-axis, then the surface of revolution
is a circular cylinder, and the lemma is obvious in this case. Suppose that the ray
−→
.BA intersects the z-axis at a point P . By cutting the surface of revolution along the

line segment AB and flattening it into the plane, we have the right figure in Fig. 4.2.
Let .θ be the angle at the pivot P of the sector as shown in Fig. 4.2 right, and put
.ρ = |P A|, the length of the line segment P A. Since .|P M| = ρ +
d
2 , we have
.l = (ρ + )θ . Since .P B = ρ + d, the area of this is
d
2
⎛ ⎞
. π(ρ + d)2 − πρ 2 ) · θ
2π = d(ρ + d2 )θ = d × l.

⨆
⨅

Proof of Theorem 4.2 For a small .h > 0 and .θ with .|θ | < π/2, |r sin θ | < r − h/2,
consider the zonal region U of .Š lying between the horizontal planes .z = r sin θ ± h2 .
If h is sufficiently small, then U can be approximated by the lateral face of a frustum
of a right cone. Hence, by Lemma 4.1, the area of U is approximated by the product
of the length of the “center curve” and the width of the zone. The length l of the
center curve is .2π r cos θ . Consulting Fig. 4.3, we can see that the width d of the
zone U satisfies .h/d = cos θ . Hence .d = h/ cos θ , and therefore the area of U is
4.2 Cylindrical Projections 79

Fig. 4.3 A zone U and the

lateral face of the frustum of a
cone
A
U h
B
θ x

approximately .d × l = 2π rh. On the other hand, the area of .ϕ(U ) is clearly .2π rh.
Thus, the areas of U and of .ϕ(U ) are (approximately) equal.
Now choose a large integer n, and divide .Š by the planes .z = ±kr/n (k =
0, 1, 2, . . . , n) into 2n zones. Furthermore, take 2n meridians with longitude
.±π k/n, k = 0, 1, 2, . . . n, and divide each zone into 2n equal pieces. Then .Š is

divided into .2n × 2n pieces. The images of 2n meridians by .ϕ are 2n generating

lines of the cylinder .Γ that divide .Γ into 2n equal regions. Hence, each of the .4n2
pieces of .Š and its image by .ϕ have the same area. For a region W of .Š, its area can
be approximated by the sum of the areas of small pieces of .Š that intersect W . And
by increasing n, we can increase the accuracy of the approximation. Thus, we can
deduce that the area of W is equal to the area of .ϕ(W ). ⨆
⨅
Corollary 4.1 (Archimedes) The area of .S is equal to .4π r 2 . ⨆
⨅
For a point P on .S(r), the set

{X ∈ S : / P OX ≤ θ}
.

is called a cap of angular radius (or spherical radius) .θ with center P . A cap with
angular radius .π/2 is a hemisphere. The center of a hemisphere is also called its
pole.
Example 4.1 The area of a cap with angular radius .θ on .S(r) is equal to .2π r 2 (1 −
cos θ ).
This can be seen as follows: We may suppose that the center of the cap is the
north-pole .(0, 0, r). Then the cap is

{(x, y, z) ∈ S : z ≥ r − r cos θ}.

.

By the cylindrical projection .ϕ, this cap is mapped to the part .z ≥ r − r cos θ of .Γ ,
which has area .2π r 2 (1 − cos θ ).
80 4 Spherical Geometry I

4.3 Spherical Polygons

A simple polygonal curve .A0 A1 A2 . . . An on .S(r) is defined to be a curve consisting

of geodesic segments .Ai Ai+1 , i = 0, 1, 2, . . . , n − 1, each of length less than .π r,
such that
(i) .Ai , i = 1, 2, . . . , n, are all different,
(ii) two distinct segments are disjoint unless they share a common endpoint,
(iii) no two consecutive segments lie on the same great circle.
Each segment .Ai Ai+1 is called an edge, and each .Ai is called a vertex of the
polygonal curve. If .A0 = An , then the simple polygonal curve is called a simple
closed polygonal curve. A spherical polygon is a region on .S(r) bounded by a
simple closed polygonal curve. The vertices and edges of the bounding polygonal
curve are called the vertices and edges of the spherical polygon. Since a simple
closed polygonal curve divides .S(r) into two regions we need, for specifying a
spherical polygon bounded by the polygonal curve, to choose one of the regions
as the spherical polygon. If a simple closed polygonal curve lies in a hemisphere,
then the region in the hemisphere is chosen as the spherical polygon bounded by the
polygonal curve. When a spherical polygon has n edges, then it is called a spherical
n-gon. If .n = 3, then it is called a spherical triangle. Since a simple polygonal
curve with three vertices always lies in a hemisphere by the condition (iii) of a
simple polygonal curve, a spherical triangle always lies in a hemisphere. A spherical
triangle with three vertices .A, B, C is denoted by .Δ̆ABC. The interior angle of a
spherical polygon at a vertex is defined to be the plane angle formed by two tangent
lines of the edges emanating from the vertex. The interior angle of .Δ̆ABC at A is
denoted by ./ CAB or ./ A.
A subset X of .S is called convex if X lies in a hemisphere, and for any two points
in X, the geodesic segment connecting the two points always lies in X. A convex set
contains no pair of mutually antipodal points. By the condition (iii) of the polygonal
curve, the vertices of a spherical triangle all lie on a small circle, and a spherical
triangle is convex. Note that if a spherical polygon is convex, then for every vertex,
the interior angle of the spherical polygon at the vertex is less than .π .
If the intersection of a finite number of hemispheres has positive area (that is,
it has interior points) and contains no pair of mutually antipodal points, then the
intersection is a region bounded by a closed polygonal curve. Hence, it is a spherical
polygon, and it is convex.
Lemma 4.2 Let .Γ be a convex spherical n-gon on .S(r), .n ≥ 3. For an edge .ϵ of .Γ ,
let .G(ϵ) denote the great circle determined by the edge .ϵ. Then .Γ lies in one side of
.G(ϵ).

Proof Let .ϵ = AB, where .A, B are vertices of .Γ . Suppose that .Γ contains two
points .X, Y that lie in the interiors of different sides of .G(ϵ). Then, since .Γ is
convex, .Δ̆ABX and .Δ̆ABY are both contained in .Γ . Hence the edge .ϵ = AB does
not lie on the boundary of .Γ , a contradiction. ⨆
⨅
4.3 Spherical Polygons 81

Thus, every edge .ϵi of .Γ determines a hemisphere .H (ϵi ) that is bounded by

G(ϵi ) and contains .Γ . For an exterior point P of a convex spherical polygon .Γ ,
.

there is a geodesic segment P Q connecting P to an interior point Q of .Γ . Then

P Q crosses some edge .ϵj of .Γ , hence P and Q lie in different sides of the great
circle .G(ϵj ), and hence .P /∈ H (ϵj ). Thus, we have the next corollary.
Corollary 4.2 Let .Γ be a convex spherical n-gon with edges .ϵ1 , ϵ2 , . . . , ϵn , and let
.Hi := H (ϵi ) be the hemisphere bounded by the great circle determined by the edge
.ϵi and containing .Γ . Then .Γ = H1 ∩ H2 ∩ · · · ∩ Hn . ⨆
⨅
The intersection of two hemispheres is called a lune. The two great semicircles in
the boundary of a lune are called the edges of the lune, and their (mutually antipodal)
end-points are called the vertices of the lune.
Remark 4.3 The tangent line of an edge emanating from a vertex P of a spherical
polygon on .S(r) lies on the tangent plane of .S(r) at P . So, if several spherical
polygons with a vertex P in common tessellate together a neighborhood of P
without overlapping, then the sum of the interior angles of these spherical polygons
at P is equal to .2π .
In a lune L, the two interior angles at its vertices are equal. The area of a lune
θ
with interior angle .θ is equal to . 2π ×(the area of the sphere). Thus, the area of a lune
with interior angle .θ is equal to .2θ r 2 .
In the following, we take the radius of the sphere as the unit of distance, and
regard the sphere .S as a unit sphere.
Theorem 4.3 (Girard’s Formula) Let .α, β, γ denote the interior angle of a
spherical triangle .Δ̆ABC on a unit sphere .S. Then the area of .Δ̆ABC is equal
to

α + β + γ − π.
.

Remark 4.4 .α + β + γ − π is called the spherical excess of .Δ̆ABC, and it is

denoted by .2ε:

2ε = α + β + γ − π.
.

Proof of Theorem 4.3 Denote the areas of .Δ̆ABC, Δ̆CBA∗ , Δ̆CA∗ B ∗ , .Δ̆CB ∗ A
by .Δ, x, y, z, respectively (see Fig. 4.4). Since .Δ̆CA∗ B ∗ and .Δ̆C ∗ AB are symmet-
ric with respect to the center of .S, .Δ + y is equal to the area of the lune .CAC ∗ B.
Hence .Δ + y = 2γ . Since .Δ + x = 2α, Δ + z = 2β, we have

.(Δ + x) + (Δ + y) + (Δ + z) = 2(α + β + γ ).

Since the left-hand side is .2Δ + (Δ + x + y + z) = 2Δ + 2π, we have .2Δ =

2(α + β + γ ) − 2π, and the formula of the theorem follows. ⨆
⨅
82 4 Spherical Geometry I

Fig. 4.4 The areas of four A ..●........ .. ... .....

spherical triangles .Δ, x, y, z ...... ... ....
∗
... ..
... ..
.... ....
z ...
..
C
.. ∗ ... ... .........●...
..
B
... ... .. ... ... ....
.... ... .. .. .
..... ....
... ...● .. .. ...
... . ... .... ..
..
... .. .. .. ..... ..
..
... ... . .. ............. ..
.... ... ... ..
.....
.. ..
.... ... .... . ...
..
.. Δ .
.. ...
.. ...
.
..........
. .....
. .
. ..
.
..
...
... .. ..
..
C....
.... .........
...........
..
.
....
.......
y
.
..
..
.
...
.
.
.
... .. ..........●........ ..
... ........... .. ....
● ........................... .....
..... .. .....
....
B ....
....
....... x
.......
........
...........
. ...................●.....
.........

........ .
...........
.............................................. ..
. ..
............
.
A ∗

Since a convex spherical n-gon can be partitioned into .n − 2 spherical triangles

by .n − 3 diagonals, we have the following corollary.
Corollary 4.3 For .n > 3, the area of a convex spherical n-gon on the unit sphere
S is given by
.

.[the sum of the interior angles] − (n − 2)π.

⨆
⨅
Let us present here Legendre’s proof of Euler’s polyhedral formula. A polyhe-
dron in .R3 is called convex if for any two points of the polyhedron, the line segment
connecting the two points belongs to the polyhedron.
Theorem 4.4 (Euler’s Polyhedral Formula) Let .v, e, f denote the numbers of
vertices, edges and faces, respectively, of a convex polyhedron .Π in .R3 . Then .v −
e + f = 2 holds.
Proof Take a point O in the interior of the convex polyhedron .Π , and let .S be the
unit sphere centered at O. Consider the central projection of .R3 \ {O} from O onto
−→
.S (i.e., the map that sends .O /= X ∈ R to the intersection point of the ray .OX
3

and .S). The images of the vertices, edges and faces of .Π by this central projection
constitute together a partition of .S into f convex spherical polygons .Γ1 , Γ2 , . . . , Γf .
Let .a(Γi ) denote the area of .Γi and let .ei denote the number of edges of .Γi . Then,
by Corollary 4.3, we have

.a(Γi ) = [the sum of interior angles of Γi ] − (ei − 2)π.

∑f
By Remark 4.3, we have . i=1 [the sum of interior angles of Γi ] = 2π v. Note also
∑f
that . i=1 (ei − 2)π = (2e − 2f )π. Therefore

⎲
f
4π =
. a(Γi ) = 2π v − (2e − 2f )π,
i=1

from which .v − e + f = 2 follows. ⨆

⨅
 4.4 The Inscribed Angle Theorem 83

Remark 4.5 There are many different proofs of Euler’s polyhedral formula. (We
already presented a proof in Chap. 3.) According to [131, Ch. 7, p.117], Legendre’s
proof seems to be the first one among the published rigorous proofs of Euler’s
polyhedral formula.

4.4 The Inscribed Angle Theorem

For a spherical triangle .Δ̆ABC, three vertices lie on a small circle. The spherical
cap bounded by this small circle and containing .Δ̆ABC is called the circumscribed
cap of .Δ̆ABC, and it is denoted by .cap(ABC).
Theorem 4.5 (The Inscribed Angle Theorem on the Sphere) Let P be the center
of the circumscribed cap of .Δ̆ABC. Then

./ C − (/ A + / B) = ±2/ P BA,
⎛

⎛
where the sign .± takes .− if .ACB is a major arc, and .+ if .ACB is a minor arc. (If
⎛

.ACB is a semicircle, then ./ BAP = 0.) Therefore, for every point .X(/= A, B) on
⎛

.ACB, we have

./ AXB − (/ XAB + / XBA) = constant.

⎛

Proof We consider the case (which looks most complicated) that .ACB is a major
arc, and yet P does not lie in .Δ̆ABC. Consulting Fig. 4.5, we have ./ P AB =
/ P BA, / P BC = / P CB, / P AC = / P CA, by symmetry. Hence

./ C = / P CB − / P CA
= / P BC − / P AC
/ A = / P AB − / P AC
= / P BA − / P AC
/ B = / P BA + / P BC

From this it follows that ./ C − (/ A + / B) = −2/ P BA. ⨆

⨅

Corollary 4.4 In a spherical triangle .Δ̆ABC,

⎛

./ C < / A + / B .⇔ .ACB is a major arc,

⎛

./ C = / A + / B .⇔ .ACB is a semicircle,
⎛

./ C > / A + / B .⇔ .ACB is a minor arc.

⎛

Since ./ A + / B + / C > π , ./ C ≤ π/2 means that .ACB is a major arc. ⨆

⨅
84 4 Spherical Geometry I

Fig. 4.5 The arc ACB is a

major arc and .P /∈ Δ̆ABC
C
P

A B

Lemma 4.3 For a spherical triangle .Δ̆ABC on the unit sphere .S, we have

⎛
the area of Δ̆ABC = π ⇔ A∗ CB ∗ is a semicircle.
.

Proof Let .α, β, γ be the interior angles of .Δ̆ABC at .A, B, C, respectively. Then
the interior angles of .Δ̆A∗ B ∗ C at .A∗ , B ∗ , C are .π − α, π − β, γ , respectively, and

the area of Δ̆ABC = π ⇔ α + β + γ − π = π

.

⇔ γ = (π − α) + (π − β)
⎛

⇔ A∗ CB ∗ is a semicircle.

⨆
⨅
Two spherical triangles on the unit sphere .S are called (directly) congruent if one
is obtained from the other by applying rotations around the center of .S.
Applying Lemma 4.3 we prove the following theorem, see also [129].
Theorem 4.6 If the area of .Δ̆ABC is .π , then the unit sphere .S can be partitioned
into four congruent copies of .Δ̆ABC.
Proof By Lemma 4.3, .A∗ B ∗ is a diameter of the circle .A∗ CB ∗ . Let D be a point on
this circle .A∗ CB ∗ such that .A∗ B ∗ = CD. Then .Δ̆ABC and .Δ̆BAD are congruent.
(Indeed, by the .180◦ rotation around the line through the midpoint M of AB and
the midpoint .M ∗ of .A∗ B ∗ , .Δ̆ABC becomes .Δ̆BAD.) Similarly, since the area of
∗ ∗ ∗ ∗
.Δ̆ACB is .π , .A C is a diameter of the circle .A BC , and BD is also a diameter of
◦
this circle. Hence, by .180 rotation around the line through the midpoint of AC and
the midpoint of .A∗ C ∗ , .Δ̆ABC becomes .Δ̆CDA. Thus four spherical triangles

Δ̆ABC, Δ̆ABD, Δ̆ACD, Δ̆BCD

.

are mutually congruent, and .S is partitioned into these four spherical triangles. ⨆
⨅
4.5 The Polar Set 85

4.5 The Polar Set

For a point .P ∈ S, let .H (P ) denote the hemisphere with pole P , that is,

H (P ) = {Y ∈ S : / Y OP ≤ π/2},
.

where O is the center of .S. For a nonempty subset .X of .S, the set
⋂
X◦ =
. H (X)
X∈X

is called the polar set of .X . For example, .{A, B, C}◦ = H (A) ∩ H (B) ∩ H (C). The
polar set .(X ◦ )◦ of .X ◦ is simply denoted by .X ◦◦ .
Lemma 4.4 For a geodesic segment P Q of length less than .π ,

(P Q)◦ = H (P ) ∩ H (Q).
.

Proof Obviously, .(P Q)◦ ⊂ {P , Q}◦ = H (P ) ∩ H (Q). If .X ∈ H (P ) ∩ H (Q),

then .XP ≤ π/2, XQ ≤ π/2. Among the points on P Q, P or Q is farthest from
X. Hence for every point .Y ∈ P Q we have .XY ≤ π/2, and hence .X ∈ H (Y ).
Thus, .X ∈ H (P ) ∩ H (Q) implies that .X ∈ H (Y ) for every .Y ∈ P Q, and hence
◦ ◦
.X ∈ (P Q) . Therefore, .H (P ) ∩ H (Q) ⊂ (P Q) . ⨆
⨅
Theorem 4.7 The operation to take the polar set .( )◦ satisfies the following four
conditions:
(i) For any family .{Xi : i ∈ J } of subsets of .S,
(⋃ )◦ ⋂
. Xi = Xi◦ .
i∈J i∈J

(ii) .X ⊂ Y ⇒ X ◦ ⊃ Y ◦ .
(iii) .X ⊂ X ◦◦ .
(iv) .X ◦ = X ◦◦◦ .
Proof
(i) is obvious from the definition of the operation .( )◦ .
(ii) Since .Y = X ∪ (Y \ X ), we have .Y ◦ = X ◦ ∩ (Y \ X )◦ by (i). Hence .Y ◦ ⊂ X ◦ .
(iii) .P ∈ X ⇒ X ◦ ⊂ H (P ) ⇒ for every X ∈ X ◦ , XP ≤ π/2. Therefore,
⋂
P ∈
. H (X) = X ◦◦ .
X∈X ◦

(iv) By (iii), we have .X ◦ ⊂ (X ◦ )◦◦ . From .X ⊂ X ◦◦ and (ii), we have .X ◦◦◦ ⊂ X ◦ .

⨆
⨅
86 4 Spherical Geometry I

Lemma 4.5 Let .Λ(P , θ ) = {X ∈ S : / XOP ≤ θ }, the cap of angular radius .θ

and center P . If .0 < θ < π/2 , then .Λ(P , θ )◦ = Λ(P , π/2 − θ ).
Proof If .X ∈ Λ(P , θ ), Y ∈ Λ(P , π/2 − θ ), then

XY ≤ XP + Y P ≤ π/2.
.

Hence, .Λ(P , π/2 − θ ) ⊂ H (X), and hence .Λ(P , π/2 − θ ) ⊂ Λ(P , θ )◦ . On the
other hand, if .Z ∈ S is not contained in .Λ(P , π/2 − θ ), then .ZP > π/2 − θ . Hence
.ZP + θ > π/2. Then there is a point .X ∈ Λ(P , θ ) that lies on the extension of ZP

beyond P and satisfies .ZX > π/2. This implies that .Z /∈ H (X), and hence .Z /∈
Λ(P , θ )◦ . Thus, if .Z ∈ S is not contained in .Λ(P , π/2 − θ ), then .Z /∈ Λ(P , θ )◦ .
Therefore, .Λ(P , θ )◦ = Λ(P , π/2 − θ ). ⨆
⨅
Theorem 4.8 Every convex spherical n-gon .Γ = P1 P2 . . . Pn , .n ≥ 3, on .S has the
following three properties:
(1) .Γ ◦◦ = Γ ,
(2) .Γ ◦ = H (P1 ) ∩ H (P2 ) ∩ · · · ∩ H (Pn ),
(3) .Γ ◦ is a convex spherical n-gon.
Proof
(1) By Corollary 4.2, .Γ is the intersection of n hemispheres. Let .X denote the set of
poles of these n hemispheres. Then .Γ = X ◦ . Hence .Γ ◦◦ = X ◦◦◦ = X ◦ = Γ .
(2) Denote the boundary of .Γ by .∂Γ . Since .Γ is convex, it can be represented as
⋃
Γ =
. (P1 X).
X∈∂Γ

Then, by Theorem 4.7 (i) and Lemma 4.5, we have

⋂ ⋂ ⋂
Γ◦ =
. (P1 X)◦ = (H (P1 ) ∩ H (X)) = H (X) = (∂Γ )◦ .
X∈∂Γ X∈∂Γ X∈∂Γ

Since .∂Γ = (P1 P2 ) ∪ (P2 P3 ) ∪ · · · ∪ (Pn P1 ),

Γ ◦ = (∂Γ )◦ = (P1 P2 )◦ ∩ (P2 P3 )◦ ∩ · · · ∩ (Pn P1 )◦ = H (P1 ) ∩ · · · ∩ H (Pn ).

.

(3) Since .Γ is contained in the interior of some hemisphere, there is a cap of angular
radius .π/2 − ε (with sufficiently small .ε > 0) that contains .Γ . By Lemma 4.5,
the polar set of a cap with angular radius .π/2−ε is a cap of angular radius .ε, and
we can deduce from Theorem 4.7(ii) that .Γ ◦ contains a cap of angular radius
.ε. Hence .Γ
◦ has positive area, and hence .Γ ◦ is a convex spherical polygon.
◦
If .Γ is a convex m-gon, then .m ≤ n by (2). Applying the same argument to
◦ ◦
.Γ = (Γ ) , we have .n ≤ m. Hence .m = n. ⨆
⨅
4.5 The Polar Set 87

By Theorem 4.8 (2), the polar set of a spherical triangle .Δ̆ABC is a spherical
triangle .Δ̆A' B ' C ' , where .A' is a vertex not lying on the boundary of .H (A), .B ' is
a vertex not lying on the boundary of .H (B), etc. This .Δ̆A' B ' C ' is called the polar
triangle of .Δ̆ABC. In other words, if .A' , B ' , C ' ∈ S satisfy
⎧
⎪
⎪ ' BA' = CA' = π/2,
⎨AA < π/2,
.(∗) BB ' < π/2, AB ' = CB ' = π/2,
⎪
⎪
⎩CC ' < π/2, AC ' = BC ' = π/2,

then .Δ̆A' B ' C ' is called the polar triangle of .Δ̆ABC. In this case, .Δ̆ABC is the polar
triangle of .Δ̆A' B ' C ' .
Theorem 4.9 Suppose that .Δ̆ABC and .Δ̆A' B ' C ' are polar triangles of each other.
Let .a ' , b' , c' be the lengths of the edges of .Δ̆A' B ' C ' opposite to the vertices
' ' '
.A , B , C , respectively. Then we have

./ A + a ' = π, / B + b' = π, / C + c' = π.

Proof Consulting Fig. 4.6, we see that ./ B ' AC = π/2 = / BAC ' . Hence

./ B ' AC ' + / BAC = / B ' AC + / BAC ' = π.

Since .B ' A = C ' A = π/2, we have ./ B ' AC ' = B ' C ' = a ' . Therefore, .a ' + / A = π .
Similarly we have .b' + / B = π, c' + / C = π . ⨆
⨅
From Theorem 4.9 and Girard’s formula, we have the following corollary.
Corollary 4.5 The sum of the area of a spherical triangle and the perimeter of its
polar triangle is equal to .2π . ⨆
⨅

Fig. 4.6 .Δ̆ABC and its polar ●

triangle .Δ̆A' B ' C '

A
................. .............
............ .............
.......... ...........
.........
........ ....
............
....... ..........
..........
.●.
..... .........
..... ....
...
....
.... ....
........
A ....
....
....
.... ...... ....
....
.... ....
.... .. ...
.... . . ...
.... .. ...
..... ... ...
....... ..
.........
● B
.................
...
...
...
... ..........
...
.. ............
................ C
....................................................●
...
.........
....................
... ..
..
.. ..
B ..
..
..
..
..
.. C
.. ..
88 4 Spherical Geometry I

4.6 Exercises

4.1 Let A, B be two points on the unit sphere S such that A lies at longitude 0,
latitude 0, and B lies at longitude π/4, latitude π/4. Find the spherical distance
AB.
4.2 The state of Alaska in US is approximated by a rectangular region with
longitude between 140◦ W and 165◦ W , latitude between 60◦ N and 70◦ N. Find
its area. (The radius of the Earth is approximately 6370 km.)
4.3 Let A, B, C, D be points on a small circle on S lying in this cyclic order. Prove
that, in the spherical quadrilateral ABCD, the equality / A + / C = / B + / D
holds.
4.4∗ Prove that for a spherical triangle Δ̆ABC on the unit sphere, the following
two statements (i) and (ii) are equivalent:
(i) The area of a spherical triangle Δ̆ABC is greater than π .
⎛

(ii) A∗ CB ∗ is a minor arc.

4.5∗ For X ⊂ S the intersection of all hemispheres that contain X is called the
convex hull of X, and it is denoted by conv(X). If there is no hemisphere that
contains X, then we put conv(X) = S. Prove that if conv(X) /= S, then
⋂
conv(X) =
. H (P ) = X◦◦ .
P ∈X◦

4.6∗ (Rusty Compass) Using a rusty compass that can draw only circles of a fixed
radius ρ, draw a circle on a big sphere. Then the area of the spherical cap
bounded by the circle and containing the needle-point on the sphere is constant
and independent of the radius of the big sphere. Prove this fact.

4.7 Notes

We start these notes with some thoughts about spherical geometry in general, also
referring to its history. Then we treat, in a somewhat longer part, mathematical
cartography and come, in particular, to the cylindrical projection. Finally, also in
a longer part, we discuss the geometry of spherical polygons and triangles.

4.7.1 Notes on Spherical Geometry

In this subsection we will cite some general references on classical spherical

geometry, meaning the geometry of spherical spaces in the spirit of non-Euclidean
geometries. The reason for having chosen also this theme (spherical geometry) for
4.7 Notes 89

one notes part is the main headline of our Chaps. 4, 5, and 7. As the text and the
reference list will show, the literature on classical spherical geometry is relatively
scattered.
Busemann’s more general investigations of types of geodesic metric spaces
(yielding also characterizations of spherical spaces) will not be discussed in our
book. For such concepts and their importance nowadays we refer to the monographs
[136] and [617] of Busemann and Papadopoulos, respectively, and to the very recent
exposition [620].
We start with the history of the field. Rosenfeld dedicates the whole first chapter
of his monograph [676] to spherical geometry, clarifying the contributions of
Menelaus, Ptolemy, the Arabic period, Regiomontanus, Vièta, Copernicus, Girard,
Euler, and many others.
Also mainly historical in nature are the comprehensive papers [616] and [618]
of Papadopoulos, in which the contributions of Euler and his academic students
to spherical geometry are presented. These refer to spherical trigonometry, various
spherical analogues of and differences to Euclidean results, and mathematical
cartography (for the latter see below).
Regarding applications we should mention here the detailed monograph [743] of
Smart from 1931, treating spherical astronomy and containing all needed tools and
notions from spherical geometry.
In Part III of his classical book [96], Blumenthal discusses the distance geometry
(also) in spherical n-space with the help of determinants which generalize those of
Cayley-Menger. Some spherical Helly-type theorems are studied, like also results
on congruence in hemispheres and small caps.
Also the book [428] of Jennings should be cited here. Its second chapter refers
completely to the geometry of spherical spaces. The author treats geodesics as arcs
of great circles and as distance minimizing curves. Fundamental results of spherical
trigonometry and the laws of sine and cosine for sides and angles are treated. The
relation between area and the angular defect is deduced, and interesting applications
of all this and more material are given. These refer to navigation, map making and
a discussion of conformal mappings.
Also Hartshorne [382, pp. 318 and 458–459] discusses topics from spheri-
cal geometry and, in particular, properties of spherical polygons. Prasolov and
Tikhomirov discuss spherical geometry in their monograph [646, Sections 5.1 and
5.3], also in view of Riemannian geometry.
Similarly, the second chapter of Wilson’s monograph [815] is dedicated to
spherical spaces, giving spherical sine and cosine formulas, some related curve
theory, finite groups of isometries, the Gauss-Bonnet theorem, and an extension to
Möbius geometry.
Borceux treats spherical triangles and spherical trigonometry in Section 4.12 of
his trilogy book [103].
The monograph [809] of Whittlesey is completely dedicated to the geometry
of spherical spaces, introduced in an axiomatic way (several proofs are given)
and showing also different approaches (one of them based on quaternions) and
interesting applications. These applications refer e.g. to the surface of the earth,
90 4 Spherical Geometry I

the study of our planet system, three- and four-dimensional polyhedra, and also
crystallography.
Melzak gives a thorough treatment of spherical trigonometry in Chapter 7 of his
book [569], and Gowers [313] presents parts of spherical geometry and hyperbolic
geometry (using a disc model) to give useful information on the parallel postulate.
Berger [75, Chapter 18] collects a lot of material on spherical geometry.
He presents a comprehensive investigation of spherical triangles and (convex)
spherical polygons, the intrinsic metric of the sphere, its isometry group, spherical
trigonometry, and further topics. Also in his more recent book [78] he studies
spherical geometry (e.g., its metric and spherical trigonometry) in Section III.1.
In Chapter 3 of his book [227], Dunajski presents (among other non-Euclidean
geometries) basics of spherical geometry, e.g. studying mapmaking, Mercator’s
projection, and a derivation of the area formula for spherical triangles.
The book [131] of van Brummelen is the most comprehensive, recent contribu-
tion to spherical trigonometry. Starting with the representation of the contributions
from the fruitful ancient, Arabic, Indian, medieval, and modern periods, the author
gives a fascinating overview to needed or related tools, notions and applications
(e.g., astronomy, regular polyhedra, stereographic projection, crystallography and
the Cèsaro method). Also the book [643] of Popko and Kitrick should be mentioned
here. It is not directly referring to spherical space, but contains a lot of related
material (like trigonometric formulae, geodesics and loxodromes, applied topics like
climate models etc.).
Lassak’s survey [492] gives an overview on results in the spirit of convex
geometry, but holding for spherical spaces and mainly depending on the width
function of convex sets. They are related to quantities like minimal width and
diameter, and therefore also bodies of constant width, diametrically complete sets
and reduced bodies in such spaces are discussed. Also the papers [581], [290],
[345], and [346] refer to typical convexity results in spherical spaces; see also page
242 in the book [558] of Martini, Montejano, and Oliversos.

4.7.2 Notes on Cylindrical Projections and Related Topics

In these notes we reflect results and references from the field of mathematical
cartography, i.e., we treat map projections (which are not always real projections)
to obtain planar images from parts of the earth’s surface. Our interest here, in these
notes, is mainly (but not only) directed to cylindrical projections.
The cylindrical projection introduced in our Sect. 4.2 is, more precisely, the so-
called cylindrical equal-area projection of Archimedes and Lambert. Here it is our
aim to give a short overview to references which contain informations to this and
to related (mainly cylindrical and conical) cartographical maps. The topic of such
“map projections” is classical, and therefore most of the references are of older date,
but have sometimes also recent new editions. There are complete books treating
only this area, or also book chapters and survey-like publications. The case of book
4.7 Notes 91

chapters refers often to parts of books on differential geometry, on non-Euclidean

(spherical) geometry or on classical geometry; see below.

Complete Books

The classical books [398] of Hinks and [568] of Melluish present systematic studies
of the whole field of map projections and have modern editions. In [398] numer-
ous map projections are compared, where also informations on the relationships
between used methods of projection and the qualities and disadvantages of each
method are given. Since it is the author’s intention to present a source for “map
makers” and “map users”, the practical point of view is central. Thus, too extensive
mathematical theory is avoided, and many nice figures are presented. This book
is interesting for readers specialized in geography, the history of cartography and
the relations between different map projections. The book [568] addresses more the
mathematical theories underlying the map constructions. It will be a good source
of knowledge for eveybody interested in the mathematical foundations of map
projections.
König and Weise [466] introduce complex vector coordinates of the spheroid
such that complex function theory can be applied. These coordinates are the complex
longitude, latitude and meridian arc length, and they lead to known projections
(Mercator, Gauss-Krüger etc.). Then the pattern of conformal projections (including
various cylindrical and conical ones) can be obtained by selected complex function
transformations. These projections are presented for the geoid and for the sphere.
Many further interesting methods and results are presented.
Also the comprehensive book [316] of Grafarend and Krumm goes much
deeper into the mathematical basics needed for investigating a broad variety of
map projections. The first four chapters treat (without detailed proofs) the basic
facts from differential geometry. E.g., these are: mappings from Riemann mani-
folds to Riemann manifolds (mainly of dimension 2), the needed coordinates (or
parameters), and surfaces of Gaussian curvature zero which are necessary for car-
tography. Concrete mappings follow, for instance sphere and ellipsoid-of-revolution
to tangential planes, to cylinders, or to cones. For studying geodesic mappings,
the Riemann, Soldner, and Fermi coordinates on the ellipsoid-of-revolution are
presented, and various further related topics are given. The appendices offer, e.g.,
elliptic functions and elliptic integrals, for geodesics the geodesic curvature and
torsion, mixed cylindric map projections, and Gauss surface normal coordinates.
The authors give also interesting historical comments, and the bibliography contains
1387 references!
Clearly, spherical harmonics and related topics are close to the geometry of
the sphere, but would open here several further doors which are not so close to
the aims of this book. However, the monograph [277] of Freeden and Schreiner
(giving a unified, impressive overview to the theory of spherical functions) should
be considered as an exception (i.e., should be mentioned here), due to its closeness to
92 4 Spherical Geometry I

mathematical geosciences and mathematical geography. E.g., the behavior of such

functions to be used in data analysis and geo-applications is taken care of.
Feeman’s monograph [255] is an excellent and modern book on map projections.
Topics presented (after giving basic facts on the earth) are: basic spherical geometry,
the reasons why one cannot make a perfect flat map of the planet, different properties
that maps can have, facts on how to design maps that have such desired properties,
conformal and equal-area maps, and so on. The author quantitatively analyzes
distortions that arise when making world maps, and he indicates how to generalize
methods to produce maps of arbitrary surfaces of revolution. There is also a chapter
that shows how to use Maple add-on software to get maps from geographic data
points. The book would excellently complete and enrich courses in linear algebra,
differential geometry, or calculus. On the other hand, it would excellently function
as fundament of an undergraduate course in mathematics or a geography course,
especially appealing to the teacher who might search exciting visual applications
even for the classroom.
There are further books completely dedicated to map projections (and containing
particular representations of cylindrical map projections). For example, we mention
here Wagner [799], Urmaev [789], Fiala [265], Hoschek [404], and Schröder [712].
E.g., in [789] the author collects special applications of the differential equations of
mathematical cartography together with some approximate and numerical methods
of integration, and Hoschek’s book also collects several “optimal drafts” (like that
of Winkel etc.).

Book Chapters

Books in differential geometry with own chapters on map projections, in particular

treating also cylindrical ones, are due to Gdowski [297], Klotzek [456, Section 3.10,
Appendix], Wünsch [818, Section 7.4], and Trapp [781]. E.g., Gdowski clearly
presents the fundaments of differential geometry needed in the fields of higher
geodesy and cartography, and Trapp studies mathematical cartography based on
conformal and equiareal functions.
Chapter 18.1 of Berger’s monograph [75] presents various needed properties of
the sphere and related basic projections (like the stereographic one), to show then a
certain variety of different cartographical projections. A very detailed discussion of
map projections is presented by Bigalke [87, Chapter 5], including also the already
mentioned draft of Winkel (1913), which is one of the most popular map designs
generated by a “mixture” of different methods.
Agricola and Friedrich wrote the nice book [10] on elementary geometry. Its
last chapter is dedicated to spherical geometry, and for teaching purposes this
book contains a large collection of interesting exercises and problems. Since also
being close to elementary geometry, we mention here the book [462] of Koecher
and Krieg; Chap. 4 presents a (today rarely occurring) collection of theorems on
triangles and their circles.
4.7 Notes 93

Survey-Like and Historical Articles

Proskurowski [648] refers to cylindrical projections and Mercator’s contributions;

the author gives also an interesting overview regarding the Dutch cartographers of
the sixteenth century.
Also very famous mathematicians contributed to mathematical cartography and
related fields. For example, Euler published in 1777 three memoirs on mathematical
cartography, namely on geographical maps in view of three criteria (one being
conformality, and the others prescribing properties of images of meridians), and he
also rediscovered Lambert’s cylindrical equal-area projection (originally introduced
in 1772). All this, including also valuable work of Lagrange, is nicely presented
by Charitos [157]. Also Papadopoulos [619] deals with Euler’s contributions to
cartography, stimulated by the French geographer Delisle who introduced a useful
method of drawing geographical maps. This method motivated Euler to write one
of the mentioned memoirs.
The work of Gauss on geodesy, topography and cartography, distributed over
different periods of his life, is discussed by Kautzleben in [444]. Via land measure-
ment and triangulation, yet in the environment of Braunschweig, Gauss reduced data
successfully via the method of least squares, and his interests in cartography were
clearly influenced by his interests in differential geometry, and certainly vice versa.
Also Chebyshev contributed to mathematical cartography. Inspired by Lagrange,
who developed differential-geometric formulas for angle-preserving maps of the
sphere to the plane, he studied the following problem: for a region C given on the
sphere, find the angle-presering map which minimizes the variation of distance-
dilatation within that region. The solution, that he found himself, is a map for which
the distance-dilatation is constant along the boundary of C. Chebyshev obtained
also interesting approximation results regarding distances in that framework.
Greinke’s paper [325] is remarkable: it contains an elementary approach to
various conformal and area-preserving maps of the sphere onto certain surfaces,
including also the plane. In its elementary style (that also Schröder’s book [712]
has) it is practically a complete teaching module for high schools.
Finally we mention that there are several “picture galleries of map projections”
which can be visited via internet. Here is an excellent example, created by H.
Havlicek (TU Vienna) and his research group:
https://www.geometrie.tuwien.ac.at/karto/
Here is the right place to mention also the recent Proceedings [138] edited by
Caddeo and Papadopoulos on mathematical geography of the nineteenth century.
They contain many articles (partially already mentioned here) with deep insights on
many findings (mainly of Euler, Lagrange and Lambert) from this fruitful century
that determine our world view still today.
94 4 Spherical Geometry I

4.7.3 Notes on Spherical Polygons

It turns out that the theory of spherical polygons (i.e., the knowledge about polygons
defined for spherical spaces or as tiles of interesting spherical tilings) yields a
particularly rich geometric topic. We select here references treating this notion, on
this way also showing the large variety of related inspiring further problems.

Books, Book Parts, and Surveys

Strongly related to his two other books on highly symmetric polyhedra, Wenninger
[808] nicely describes constructions of paper models of spherical tilings (i.e.,
spherical polygons play a natural role) connected with the symmetries of Platonic
and Archimedean solids, with geodesic domes, honeycombs and so on. The author
was strongly inspired by Coxeter.
The chapter on spherical geometry in Fenn’s book [264] contains a detailed
section on geodesic triangles, and another one on the celestial sphere.
Wilson [815] studies spherical geometry in Chap. 2, particularly also sine and
cosine formulae, interesting related properties of curves and the Gauss-Bonnet
theorem (including its Euclidean borderline case). Shortly we mention here also
the book [809] of Wittlesey with an axiomatic approach to spherical geometry, see
the notes above. And also Bigalke’s book [87] contains interesting knowledge about
spherical triangles and polygons.
In the monograph [643] of Popko and Kitrick several topics based on the
geometry of the sphere are collected, among them spherical subdivisions (e.g.,
by great circles), packing problems and interesting point distributions on the
sphere, spherical polygons, and various other concepts. Interesting application fields
are presented, such as architecture (“Bucky’s dome”), celestial catalogs, climate
models, fishpharming, polyhedra with regularity properties, golf ball design and so
on.
In the booklet [166] of Clemens on two-dimensional geometries a problem-
solving approach to planar geometries of constant curvature is used. Part III treats
elementary spherical geometry. E.g., the area of spherical triangles is studied, and
a synthetic treatment of spherical geometry through the angle-excess formula is
given. Part VI refers again to spherical geometry, but now from a more advanced
point of view. Spherical coordinates, shortest paths between any two points, and
also formulas for spherical polygon areas and the sum of their interior angles are
discussed. Various further related topics can be found.
Berger [75, Chapter 18] discusses spherical triangles and (convex) spherical
polygons in a broader way, and his later book [78, Chapter III] contains also
various results on spherical geometry. Furthermore, we refer here to Hartshorne
[382, Section 45 and its exercises]. Van Brummelen devotes the whole book [131]
to spherical trigonometry.
4.7 Notes 95

Regarding survey articles, we mention Rosenfeld [674, §2 and §4] who mainly
studies various properties of spherical triangles, like area, congruence, cosine
theorem for sides and for angles, and also cotangent formulae. And we shortly
mention the survey [492], regarding the type of studied questions mainly dedicated
to convex geometry, but delivering various results also for polygons (see below for
a better description).

Some Research Contributions

Here we present various research results on spherical polygons in some thematic

order.
1. Spherical Tilings by Polygons
Grünbaum and Shephard [341] consider spherical tilings with strong symmetry
properties. Namely, by central projection from the centroid the boundary structures
of regular and uniform 3-polytopes can be projected onto a centered 2-sphere, to
form there a spherical tiling whose transitivity properties come from the symmetry
group of the projected polytopes. The authors classify all tiling types whose
symmetry groups are transitive on the tiles (called isohedral tilings), on the edges
(isotoxal tilings), or on the vertices (isogonal tilings). It should be noticed that these
authors published various papers on the closely related notion of spherical patterns.
L. Fejes Tóth [259] discusses problems on spherical polygons in spherical tilings
created by great circles, making also use of isoperimetric properties of regular
spherical polygons. Florian [272] studies, for .A > 0 and .P > 0 with certain natural
restrictions, the class of all compact, spherically convex sets in the 2-sphere having
area at least A and perimeter at most P . He describes spherically convex sets closest
to circumscribed [resp. inscribed] convex spherical m-gons; the distance considered
here is measured in terms of area and perimeter deviation.
Types of spherical triangles which tile the sphere edge-to-edge are well-known,
but relaxing the edge-to-edge demand, one gets a larger variety. As a first step
in this direction, Dawson [200] characterizes all isosceles triangles that tile the
sphere. He finds one infinite family and three sporadic tiles that tile only edge-
to-edge. The same author together with Doyle published a series of papers on right
spherical triangles which can tile the sphere. In the third and provisionally last part
[201] they study the so-called asymptotically obtuse families of triangles, and they
establish that six triangle types within such families can actually tile the sphere. The
corresponding tilings are classified.
Gao et al. [291] study edge-to-edge tilings of the 2-sphere by twelve geomet-
rically congruent pentagonal tiles. They first look at combinatorially congruent
tiles, to investigate then all cases where the tiles have even the same edge length
combination and arrangement, doing then the same with the cases where all the
tiles have the same angle combination and arrangement. The classification of the
spherical tilings in five classes, which require the edges of the tiles to be great arcs,
is derived.
96 4 Spherical Geometry I

A folding tiling of the sphere is an edge-to-edge covering of it by spherical

geodesic polygons without gaps or overlaps. Avelino and dos Santos [49] study
dihedral folding tilings of the sphere by kites and isosceles triangles. Assuming that
each tiling has at least two non-congruent tiles in some case of adjacency, they give
geometric conditions on the tiles with consequences to local configurations. Finally
presented in a table, this yields a characterization of a discrete family of all possible
tilings. The authors continued these investigations, and Part IV of a series of related
papers is [49].
Sakano and Akama [684] investigate edge-to-edge tilings of the 2-sphere
by mutually congruent polygons. They first characterize all anisohedral spherical
triangles (admitting tilings of the sphere such that the symmetry group of the tiling
does not act transitively on the tiles). Further on, they give a complete classification
of all tilings by kites (convex quadrangles whose cyclic list of edge-lengths is aabb
for .a /= b), darts (non-convex quadrangles whose cyclic list of edge-lengths is also
.aabb, a /= b), and rhombi. These results are based on classification results on tilings

of the 2-sphere by congruent triangles obtained by Ueno and Agaoka [786].

Breda and dos Santos [127] demonstrate how to generate new families of specific
spherical tilings with the help of the free software GeoGebra. For this purpose some
spherical geometry capabilities of GeoGebra had to be extended.
Lee et al. [498] present a new sphere subdivision method which generates a large
number of spherical pentagons and is based on a successive subdivision of a module
of a spherical dodecahedron. Using special design parameters, this method controls
the shapes of generated spherical pentagons. The authors are able to divide a sphere
into spherical pentagons of equal area and to minimize the number of different arc
lengths used in the pentagonal structure. The efficiency of the new method is shown
by various examples, and taking additional constraints on the optimization problem
into consideration may deliver sphere subdivisions of interesting specific types.
Akama et al. [12] continue with a series of results and classify tilings of the
sphere by congruent equilateral pentagons. They show that there are eight edge-to-
edge tilings of the sphere by congruent equilateral pentagons, and that the number
of participating tiles is between 12 and 60.
2. Isoperimetric-Type Results
Wimmer [816] confirms the following theorem also for the spherical plane:
Among all quadrilaterals with given side-lengths there is one of greatest area, and
for this one the sum of two opposite angles is the same as the sum of the two
remaining angles. These quadrilaterals are cyclic and do not exceed the half-sphere
(they consist of Eulerian triangles, i.e., triangles whose angles and edges are smaller
than .π ).
Liu and Chang [511] prove various results of isoperimetric type for spherical
m-gons, .m ≥ 3. E.g., for all spherical convex k-gons of fixed diameter .π/2 with
.3 ≤ k ≤ m and odd m, the regular spherical triangle has maximal perimeter, and

the regular spherical m-gon has maximal area. For the same family, but of fixed
minimal width .π/2, the regular spherical m-gon attains the minimum perimeter,
4.7 Notes 97

and the regular triangle has minimal area. Some further results (e.g., also on reduced
spherical polygons) are obtained.
Also Maehara and Martini [537] derive several inequalities for spherical poly-
gons of isoperimetric type, and they use stereographic projection. E.g., they show
that among spherical polygons with fixed side-lengths and perimeter less than .2π , a
cyclic polygon has the largest area, and that among spherical k-gons inscribed in a
fixed spherical cap, spherical regular k-gons have the largest perimeter. Also some
of the following results on cyclic or regular spherical polygons are of isoperimetric
type.
3. Cyclic and Regular Spherical Polygons
We say that K, a compact set in the n-dimensional sphere (.n ≥ 2), has the
Pompeiu property if only the zero function has zero surface integral over every
rotated copy of K. There exist regular spherical polygons bounded by arcs of great
circles and not having the Pompeiu property. In [40], by Armitage the number of
sides of such polygons is studied for the 2-sphere. It is proved that if .m ≥ 4, then
there are m-sided regular spherical polygons which are bounded by arcs of great
circles and do not have the Pompeiu property.
Maehara [520] shows that among the spherical polygons obtained by deforming
a convex polygon on the unit 2-sphere, without changing the edge lengths, a cyclic
polygon has maximum area. He also reproves that the unit 2-sphere in 3-space
cannot be touched simultaneously by 13 further unit spheres. This famous result
is due to Schütte and van der Waerden [716] and has an interesting history, see
the reports [635] and [145] as well as the very recent contribution [304] due to
Glazyrin.
Vu [798] studies the symplectic volume of the moduli space of closed spherical
polygons and proves that regular m-gons maximize the volume inside the class of
m-gons with fixed perimeter, and that under certain assumptions the regular m-gons
are the unique m-gons maximizing the volume.
Guo and Sönmez [349] show that any formula on side-lengths and diagonal-
lengths and the circumradius of cyclic polygons in Euclidean, hyperbolic and
spherical geometry, which holds in one of these three geometries, is (under certain
additional conditions) also true in the other two ones. The authors give also a useful
little survey on cyclic polygons in these geometries.
Kakulashvili [436] collects calculations for polygons in spherical geometry. E.g.,
spherical cyclic quadrilaterals are investigated, and explicit calculations of metric
quantities of spherical polygons are presented.
Kamiyama [440] studies the configuration space .Mm (a) of regular spherical m-
gons of side-length .a ∈ (0, π ) to determine the corresponding Euler characteristic.
Here he computes this characteristic when m is even by showing that .Mm (a) may
be obtained by Morse surgeries on the configuration space of equilateral m-gons in
the plane. In [441] the author completes these studies.
It is known that for general cyclic m-gons in the Euclidean setting there is no
area formula for .m ≥ 5 in terms of its side-lengths based only on the four arithmetic
98 4 Spherical Geometry I

operations and k-th roots. By Komori et al. [469] it is shown that a similar result
holds in spherical and hyperbolic geometry.
4. Convexity-Related Results
Close to sets of constant width, Goldberg (see [305] and [306]) presents a
kinematic construction for spherical rotors, and in the second paper he constructs
two types of “basic” rotors for polygons in a kinematic way.
From the viewpoint of computational geometry, Ha and Shin [352] present a
linear-time algorithm for computing the intersection of a pair of spherical convex
polygons.
As already announced, we mention once more the excellent survey [492] of
Lassak on spherical convex bodies, polygons, and polytopes. This survey is mainly
related to typical questions and notions of convex geometry. The author collects
results related to width and minimal width of these bodies. Important topics
collected here in connection with spherical convex bodies (the polytope case always
contained) are: lunes and strips containing the studied bodies, minimal width (=
thickness) and bodies of constant width and of constant diameter, diametrically
complete sets and reduced bodies. (Note that reduced bodies are convex bodies with
the property that each proper convex subset of them has smaller minimal width; and
diametrically complete sets have the property that each superset of them has a larger
diameter. In Euclidean space, bodies of constant width share both these properties.)
Some of the quantities and notions under particular consideration in that survey are
width functions (in particular, minimal width and diameter), perimeter, area and
extreme points of spherical convex bodies and polygons.
The perimeter deviation of two convex bodies .K1 and .K2 in the Euclidean plane
is defined as the perimeter of .K1 ∪ K2 minus the perimeter of .K1 ∩ K2 . It is known
that a convex polygon with at most m vertices minimizing the perimeter deviation
from .K1 is always inscribed in .K1 . Fodor [273] shows that the analogous result does
not hold on the sphere.
Chapter 5
Spherical Geometry II

Mathematics consists of proving the most obvious thing in the

least obvious way.
— George Pólya

Using a known method of Cesàro, we prove the spherical cosine law, the spherical
sine law, and Euler’s formula for the spherical excess. The spherical cosine law
is applied to prove a triangle comparison theorem for spheres. Euler’s formula for
spherical excess is applied to prove a theorem for the area of a spherical triangle with
two fixed edges and one variable edge. We also derive an isoperimetric theorem for
spherical quadrilaterals.

5.1 The Cesàro Triangle

Let us take the radius of the sphere .S as unit length, and regard .S as the unit
sphere. For a non-antipodal pair of points .A, B ∈ S, the shortest geodesic segment
connecting A and B and its length are denoted by AB. (The linear distance between
A and B is denoted by .|AB| or .‖A−B‖.) Recall also that .Δ̆ABC denotes a spherical
triangle, whereas .ΔABC denotes a planar triangle. The antipodal point of .X ∈ S is
denoted by .X∗ .
For a spherical triangle .Δ̆ABC on .S, let .Π be the plane that is tangent to .S at A,
and let .ϕ : S \ {A∗ } → Π be the stereographic projection of .S from .A∗ . Denote the
image of .X ∈ S by .X̄:

X̄ = ϕ(X).
.

Then, the planar triangle .ΔĀB̄ C̄ is called the Cesàro triangle of .Δ̆ABC with
respect to A. Planar triangles .ΔĀB̄ ∗ C̄ and .ΔĀB̄ C̄ ∗ are called (Cesáro’s) derived
triangles, see Fig. 5.1.

© The Author(s), under exclusive license to Springer Nature Switzerland AG 2024 99

H. Maehara, H. Martini, Circles, Spheres and Spherical Geometry,
Birkhäuser Advanced Texts Basler Lehrbücher,
https://doi.org/10.1007/978-3-031-62776-7_5
100 5 Spherical Geometry II

Fig. 5.1 Vertices .Ā(= A), B̄, C̄ of a Cesàro triangle, and .B̄ ∗

...................................................
A
................................................
.............................................
.............................................
ε .... B
....
α π−α
.
B .............................
.......
......
. .
........
.......
......
.
..
.. ....
....
....
.....
.....
.....β −ε ..
..
.. ....
... ...
..... .. ....
..... .. ....
..... ..
.
. .......
..... .. ... .
..... .. ....
..... .. ....
.. ....
γ −ε α−ε
.....
..... ..
. .......
..... .. ....
..... .. ....
..... .. ....
..... .. ....
..... ...
.
. ......
..... ... ......
....... .....
.....
C̄

Fig. 5.2 The Cesàro triangle .ΔĀB̄ C̄, and the derived triangle .ΔĀB̄ ∗ C̄

Theorem 5.1 The interior angles of the Cesáro triangle .ΔĀB̄ C̄ of .Δ̆ABC with
respect to A are given by

./ Ā = α, / B̄ = β − ε, / C̄ = γ − ε,

and the interior angles of the derived triangle .ΔĀB̄ ∗ C̄ are given by

./ Ā = π − α, / B̄ ∗ = ε, / C̄ = α − ε

(see Fig. 5.2), where .α, β, γ are the interior angles of .Δ̆ABC at .A, B, C, respec-
tively, and .2ε is the spherical excess of .Δ̆ABC, that is, .2ε = α + β + γ − π.
Proof By the stereographic projection from .A∗ , every circle passing through .A∗
goes to a line on the tangent plane of .S at A, and every circle that does not pass
through .A∗ goes to a circle on the tangent plane. Hence the image of any arc of a
great circle that passes through A goes to a line segment, and since the great circle
BC does not pass through .A∗ , it goes to a circle .Γ on the tangent plane of .S at A.
Hence we can see that the image of .Δ̆ABC under the stereographic projection is the
(

curvilinear triangle enclosed by the line segments .ĀB̄, ĀC̄ and the circular arc .B̄ C̄
5.2 Edge-Lengths of Cesàro Triangles 101

Fig. 5.3 Curvilinear triangle

.ĀB̄ C̄
...
..... Ā ... .... .... .... ....... ............ B
.... .... .... .... .... .... .... .... .
................................................ ε
.... .... .... .... .... .... .... .... .... ...... ...
.... .... .... .... .......... .... ....
B̄ ..................
...
.......... .... .... .... .... .... .... .... .... .... .... .... .... ... ....
.. .....
...
.......
... ..... ... ..
.. ....
...
..
...
.
..

...ε
............ ....
..... ..... ... ..
. ....... .
....

....
.....
.. ...
....
.... ... ..
..
....
....
....
..
...

....
.. ....
.. .....
....
....
. ... ..
.. .. ... ....
...

.....
.. ....
.. .....
.. ....
...
....
..
.. .. .
....
.
....
....
...

........ ....
.... ...
....
.... ..
.. ....
....
....
Γ
....
.......
.....
. .. . . .. ..
..

..........ε ....
....... .... .. ... . .
.....
........ .... . .... .......
......... .... ... .... ........
.......... .........
................. . ..... . ...
.... .... .... .... ............................................. ..................
... ................................................
...
...
....
.............

C̄

of .Γ not containing .B ∗ , see Fig. 5.3. Since the sizes of angles are preserved by the
stereographic projection, the interior angles of this “curvilinear” triangle at .Ā, B̄, C̄
become .α, β, γ , respectively. Since the spherical excess .2ε of .Δ̆ABC is equal to the
sum of three interior angles of this curvilinear triangle minus .π, we can see that .2ε
is equal to the sum of three interior angles of the curvilinear triangle minus the sum
of interior angles of .ΔĀB̄ C̄. Therefore, the angle formed by .Γ and its chord .B̄ C̄
(the part not containing .Ā) is equal to .ε. Thus, ./ ĀB̄ C̄ = β − ε, / ĀC̄ B̄ = γ − ε.
In .ΔĀB̄ ∗ C̄, we have clearly ./ B̄ ∗ ĀC̄ = π − α, and by the alternate segment
theorem, we have ./ ĀB̄ ∗ C̄ = ε. And ./ ĀC̄ B̄ ∗ = α − ε.
⨆
⨅
Remark 5.1 Concerning Cesàro triangles, we refer to J. D. H. Donnay [220], van
Brummelen [131], and H. Maehara and H. Martini [537].
Remark 5.2 In a planar triangle, “the angle opposite to the greater edge is greater”
(see Euclid’s Elements). The same assertion is also true for spherical triangles. This
can be seen as follows. Suppose that .AC > AB in a spherical triangle .Δ̆ABC on
.S. Then B lies in the interior of the spherical cap with center A and radius AC.

By stereographic projection from .A∗ , this spherical cap is mapped to a disk with
center .Ā and radius .|ĀC̄|, and .B̄ lies in the interior of this disk. Hence we have
.|ĀC̄| > |ĀB̄| in the planar triangle .ΔĀB̄ C̄. Hence ./ B̄ > / C̄ by Proposition 18 of

Euclid’s Elements. Since ./ B̄ = β − ε, / C̄ = γ − ε, we have .β > γ .

5.2 Edge-Lengths of Cesàro Triangles

Theorem 5.2 The edge-lengths of the Cesàro triangle ΔĀB̄ C̄ of Δ̆ABC with
respect to A and the edge-lengths of the derived triangle ΔĀB̄ ∗ C̄ are given as
follows:

|ĀB̄| = 2 tan 2c , |B̄ C̄| = 2 sin a2 /(cos b2 cos 2c ), |ĀC̄| = 2 tan b2 ,

.

|B̄ ∗ C̄| = 2 cos a2 /(sin b2 sin 2c ), |ĀB̄ ∗ | = 2 cot 2c ,

102 5 Spherical Geometry II

c 2 cot
2 tan 2 Ā .............................................
2
.............................................
...........................................
... ε B̄ ∗
...................................................... ....
B̄ .......
......
. .. ....
...............................
.....
..... β −ε
.....
......
. .
........
α ..
..
.
.. ........
....
.. ....
..... .. ....
..... . ....
..... ....
..... .. .....
..... ...
b ....
.....
.....
..... 2 tan 2 ....
....
....
b
2 sin a2 /(cos cos ) c γ −ε a
2 cos /(cos 2b sin 2c )
..... .. ... .
..... .. ..
..... .. .... 2
2 2 .. ....
.....
.....
..
.. .......
..... ... ....
..... .. ........
........ .....
....
C̄

Fig. 5.4 Edge-lengths of a Cesàro triangle

A∗
...........................................................
............. ..........
......... .. ... ........
........ ... .. .......
......... .. . ....
..... .
.... .... ....
....
..... ... .
. ....
..... ... ..
. ....
.... .... ... ...
...
... ... ... ...
.... ... .
. ...
... ... ... ...
.. .. .
. ..
... .... ... ..
..
... ... ... ..
...
...
..
.. ....
..
..... ... . ... O
..... .
..
..
..
..
..
... ..
....
.
.......
.....
. c
...... .
.
.
... ...
.
.
.
... ... ...... ... ...
......
... ... ....... ... ..
..
... ... ............ .
. ..
... ... ........ ... ..
.............. ...
.
........
.. ....
.. .......................... ... ....
B ...
.. .... ..........
....... .........
.
.
... .....
.....
... ........ ......... .......
... ......... .......... .... ........
..... ... ..........
...........................................................................................................................................

B̄ A = Ā
Fig. 5.5 Section by the plane A∗ AB

see Fig. 5.4, where a, b, c denote the lengths of the edges of Δ̆ABC opposite to
A, B, C.
Proof
(i) Edge-lengths of ΔĀB̄ C̄: Fig. 5.5 shows the section of S and its tangent plane
at A by the plane ABO (O is the center of S). Since / AOB = c, we have
/ AA∗ B = c/2, and since |AA∗ | = 2, we have |AB| = 2 sin c , |ĀB̄| = 2 tan c .
2 2
Similarly we have |ĀC̄| = 2 tan b2 . From ΔA∗ BA ∼ ΔA∗ AB̄ (similar), we
have |A∗ B| · |A∗ B̄| = |A∗ A|2 . Similarly, from ΔA∗ CA ∼ ΔA∗ AC̄, we
have |A∗ C| · |A∗ C̄| = |A∗ A|2 . Hence |A∗ B| · |A∗ B̄| = |A∗ C| · |A∗ C̄|, and
therefore |A∗ B| : |A∗ C| = |A∗ C̄| : |A∗ B̄|. Since ΔA∗ BC and ΔA∗ C̄ B̄
have the common angle / A∗ , |A∗ B| : |A∗ C| = |A∗ C̄| : |A∗ B̄| implies that
ΔA∗ BC ∼ ΔA∗ C̄ B̄. Therefore,|BC| : |C̄ B̄| = |A∗ B| : |A∗ C̄|, and hence

|BC| · |A∗ C̄|

.|C̄ B̄| = .
|A∗ B|
5.3 Spherical Cosine Law and Sine Law 103

Similarly to |AB| = 2 sin 2c , we can obtain |BC| = 2 sin a2 . From |A∗ B| =

|A∗ A|
|A∗ A| · cos 2c and cos b2 = |A ∗
∗ C̄| = |A∗ C̄| , we have |A C̄| = 2/cos 2 , and hence
2 b

2 sin a2 · (2/ cos b2 ) 2 sin a2

|C̄ B̄| =
.
c = .
2 cos 2 cos 2c cos b2

(ii) Edge-lengths of ΔĀB̄ ∗ C̄: Let a ' , b' , c' be the edge-length of Δ̆AB ∗ C opposite
to A, B ∗ , C, respectively. By (i), the edge-lengths of the Cesàro triangle
'
ΔĀB̄ ∗ C̄ of Δ̆AB ∗ C with respect to A are |ĀB̄ ∗ | = 2 tan c2 , |ĀC̄| =
' ' ' '
2 tan b2 , |B̄ ∗ C̄| = 2 sin a2 /(cos b2 cos c2 ). Since a ' = π −a, b' = b, c' = π −c,
we have

|B̄ ∗ C̄| = 2 cos a2 /(cos b2 sin 2c ), |ĀC̄| = 2 tan b2 , |ĀB̄ ∗ | = 2 cot 2c .

.

⨆
⨅

5.3 Spherical Cosine Law and Sine Law

Theorem 5.3 (Spherical Cosine Law) For Δ̆ABC on S, the following holds:

. cos a = cos b cos c + sin b sin c cos α. (5.1)

Proof Figure 5.6 shows the enlarged Cesàro triangle ΔĀB̄ C̄ of Fig. 5.4 with ratio
1 b c
2 cos 2 cos 2 . Applying the (planar) cosine law to this enlarged triangle, we have

. sin2 a
2 = sin2 b
2 cos2 c
2 + sin2 c
2 cos2 b
2 − 2 sin b2 cos 2c sin 2c cos b2 cos α.

Rewriting this by using the identities

2 sin2
.
x
2 = 1 − cos x, 2 cos2 x
2 = 1 + cos x, 2 sin x2 cos x2 = sin x,

Fig. 5.6 Enlarged triangle

ΔĀB̄ C̄, with ratio
sin 2c cos 2b
..................................
.............................................
α
1 b c
2 cos 2 cos 2 ................................................ ..
..
..... ..
..... ..
..... ..
..... ..
..... ..
..... ..
..
.....
.....
.....
.....
..
..
..
..
sin 2b cos 2c
..... ..
..
a ......... ..
sin 2 .....
.....
..
..
..
..
..... ..
..... ..
..... ..
..... .
..... ....
..... ...
..... ..
..
104 5 Spherical Geometry II

we have

1 − cos a = 12 (1 − cos b)(1 + cos c) + 12 (1 − cos c)(1 + cos b) − sin b sin c cos α,
.

from which (5.1) follows. ⨆

⨅
Applying the spherical cosine law to the polar triangle of Δ̆ABC we have the
following corollary.
Corollary 5.1 In Δ̆ABC, we have

. cos / A = cos / B cos / C − sin / B sin / C cos a.

⨆
⨅
Remark 5.3 By G. A. Jennings [428] and L. Fejes Tóth [257], the spherical cosine
law is proved using the vector product.
Theorem 5.4 (Spherical Sine Law) In Δ̆ABC on S, we have

sin a sin b sin c

. = = . (5.2)
sin α sin β sin γ

Proof This time we use the enlarged triangle of the derived triangle ΔĀB¯∗ C̄ of
Δ̆ABC (Fig. 5.4) with ratio 12 cos b2 sin 2c , see Fig. 5.7. Applying the (planar) sine
law to this enlarged triangle, we have

cos a2 sin b2 sin 2c

. = .
sin(π − α) sin ε

Multiplying 2 sin a2 to both sides, and using 2 sin a2 cos a2 = sin a, we have

sin a 2 sin a2 sin b2 sin 2c

. = .
sin α sin ε

Fig. 5.7 Enlarged triangle

ΔĀB̄ ∗ C̄, with ratio
cos b cos 2
c
2 .........................................................................................
1 b c ..............................................
............................................
....
ε....
....
2 cos 2 sin 2 ..
..
..
..
π−α .......
....
.. ....
.. ....
.. ....
.. ....
.. .......
.. ....
.. ....
.. .... a
.. ....
sin 2b sin 2c ..
..
.. ....
........ cos 2
..
.. .......
.. ....
..
.. .. .....
.. ...
.. ....
.. ....
.. ....
.. ....
.. .......
.. .....
......
..
5.4 The Triangle Comparison Theorem for Spheres 105

Since the right hand side is independent of the choice of an interior angle (and its
opposite edge), (5.2) follows.
⨆
⨅
Theorem 5.5 (Euler’s Formula for the Spherical Excess) For Δ̆ABC,

1 + cos a + cos b + cos c

. cos ε = (5.3)
4 cos a2 cos b2 cos 2c

holds, where a, b, c are lengths of the edges opposite to A, B, C respectively, and

2ε is the spherical excess of Δ̆ABC.
Proof Applying the (planar) cosine law to the enlarged triangle of Fig. 5.7, we have

. sin2 b
2 sin2 c
2 = cos2 a
2 + cos2 b
2 cos2 c
2

− 2 cos a2 cos b2 cos 2c cos ε.

From this we get

cos2 a
+ cos2 b
2 cos
2 c
− sin2 b
sin2 c
. cos ε = 2 2 2 2
.
2 cos a2 cos b
2 cos 2c

The numerator of the right hand side fraction is equal to

( )( )
. cos2 a
2 + cos b2 cos 2c + sin b2 sin 2c cos b2 cos 2c − sin b2 sin 2c .

Rewriting this by using the identities

2 cos2
.
x
2 = 1 + cos x,
2 cos x cos y = cos(x + y) + cos(x − y),

we have (5.3). ⨆
⨅

5.4 The Triangle Comparison Theorem for Spheres

Applying the spherical cosine law, we prove the following result in an elementary
way, without invoking Riemannian geometry.
Theorem 5.6 For a spherical triangle T on the unit sphere, draw a spherical
triangle .T ' on a sphere of radius .R > 1 with the same edge-lengths as T . Then
each interior angle of .T ' is smaller than the corresponding interior angle of T .
106 5 Spherical Geometry II

Since a triangle in the plane is regarded as the limit case .R → ∞ of a spherical

triangle on a sphere of radius R, we have the following corollary.
Corollary 5.2 For a spherical triangle T on the unit sphere, draw a triangle on
the plane with the same edge-lengths as T . Then every interior angle of the planar
triangle is smaller than the corresponding interior angle of T . ⨆
⨅
Theorem 5.6 is a consequence of Toponogov’s triangle comparison theorem, see
[614] and [780, Theorem 3.91, p. 189]. We present here a proof of Theorem 5.6
given in [530].
First, we prepare two lemmas.
Lemma 5.1 For .0 < λ < 1, .sin λx/ sin x is monotonically increasing in .0 < x <
π, and

sin λx
. lim = λ.
x→0 sin x

Proof Put .f (x) = sin λx/ sin x. Then

f ' (x) = (λ cos λx sin x − sin λx cos x)/ sin2 x.

.

Denote the numerator by .g(x), that is,

g(x) = λ cos λx sin x − sin λx cos x.

.

Then .g(0) = 0 and

g ' (x) = −λ2 sin λx sin x + λ cos λx cos x − λ cos λx cos λx + sin λx sin x
.

= (1 − λ2 ) sin λx sin x > 0.

Hence .g(x) > 0 in .0 < x < π , and hence .f ' (x) > 0 in this range. Now, using
L’Hopital’s Rule, we have

sin λx λ cos λx
. lim = lim = λ.
x→0 sin x x→0 cos x

⨆
⨅
Lemma 5.2 For fixed reals .b, c, λ; 0 < b ≤ c < π, 0 < λ < 1, put

cos λx − cos λb cos λc cos x − cos b cos c

ϕ(x) =
. − .
sin λb sin λc sin b sin c

Then, .ϕ(x) > 0 for .c − b < x < min{b + c, 2π − (b + c)}.

5.4 The Triangle Comparison Theorem for Spheres 107

Proof Put .d = min{b + c, 2π − (b + c)}. Since

cos λx − cos λb cos λc cos x − cos b cos c

ϕ(x) =
. −1− +1
sin λb sin λc sin b sin c
cos λx − cos(λb + λc) cos x − cos(b + c)
= − ,
sin λb sin λc sin b sin c

we have .ϕ(c − b) = ϕ(d) = 0. Now

−λ sin λx sin x
ϕ ' (x) =
. +
sin λb sin λc sin b sin c
−λ2 cos λx cos x
ϕ '' (x) = +
sin λb sin λc sin b sin c
λ3 sin λx sin x
ϕ ''' (x) = − .
sin λb sin λc sin b sin c
If .0 < x ≤ c, then

ϕ ' (x) sin λb sin λc λ sin λx sin λb sin λc

. =− + .
sin x sin x sin b sin c
sin λb sin λc sin λx
Since . > λ and . ≥ by Lemma 5.1, we have
sin b sin λc sin x

ϕ ' (x) > 0 in 0 < x ≤ c.

. (5.4)

Hence .ϕ(c − b) = 0 implies

.ϕ(x) > 0 in c − b < x ≤ c. (5.5)

Next, we show that

ϕ '' (x) is monotonically decreasing in 0 < x < π/2, .

. (5.6)
ϕ '' (x) < 0 in π/2 ≤ x < d. (5.7)

sin λx sin(λπ/2)
If .0 < x < π/2, then . < < 1, and
sin x sin(π/2)
⎛ ⎞
λ sin λx
λ sin λx − sin x = sin x
. −1
sin x
⎛ ⎞
λ sin(λπ/2)
< sin x − 1 < (sin x)(λ − 1) < 0.
sin(π/2)
108 5 Spherical Geometry II

Since
sin b sin c 3
(sin b sin c)ϕ ''' (x) =
. λ sin λx − sin x
sin λb sin λc
sin b sin c
and . < 1/λ2 by Lemma 5.1, we have
sin λb sin λc

(sin b sin c)ϕ ''' (x) < λ sin λx − sin x < 0.

.

Therefore, .ϕ ''' (x) < 0 in .0 < x < π/2, and hence (5.6) holds.
sin λb sin λc
If .π/2 ≤ x < π , then .cos x ≤ 0, and . > λ2 by Lemma 5.1. Hence
sin b sin c
sin λb sin λc
(sin λb sin λc)ϕ '' (x) = −λ2 cos λx +
. cos x < λ2 (cos x − cos λx).
sin b sin c
Since .cos x is monotonically decreasing in .0 < x < π , we have .cos x − cos λx < 0
in .0 < x < π . Since .sin λb sin λc > 0, we have .ϕ '' (x) < 0 in .π/2 ≤ x < d.
Thus (5.7) holds.
Now, since

sin λb sin λc
(sin λb sin λc)ϕ '' (0) = −λ2 +
. > 0,
sin b sin c

we have .ϕ '' (0) > 0. Since .ϕ '' (π/2) < 0 by (5.7), and since .ϕ '' (x) is (strictly)
monotonically decreasing in .0 < x < π/2 by (5.6), there must be a unique .x0 ∈
(0, π/2) such that .ϕ '' (x0 ) = 0. Then .ϕ ' (x) is monotonically increasing in .0 <
x < x0 , and monotonically decreasing in .x0 < x < π/2. Since .ϕ ' (x) is also
monotonically decreasing in .π/2 < x < d by (5.7), we can deduce that .ϕ ' (x) is
monotonically decreasing in .x0 < x < d. Therefore,

there is just one x ∈ (0, d) such that ϕ ' (x) = 0.

. (5.8)

Now, if the lemma is not true, then there must be .x1 ∈ (c, d) such that .ϕ(x1 ) = 0.
Then, since .ϕ(c − b) = ϕ(x1 ) = ϕ(d) = 0, .ϕ ' (x) must be 0 at some .x ∈ (c − b, x1 ),
and at some .x ∈ (x1 , d). This contradicts (5.8). Therefore, the lemma holds. ⨆
⨅
Proof of Theorem 5.6 Let .a, b, c denote the edge-lengths of T . Then .a + b + c <
2π. We assume that .b ≤ c. Then

c − b < a < min{b + c, 2π − (b + c)} ≤ π.

.

Let .α be the interior angle of T at the vertex opposite to the edge of length a, and .α '
be the corresponding interior angle of the spherical triangle .T ' drawn on the sphere
of radius .R > 1 with the same edge-lengths as T . To prove the theorem, it is enough
5.5 Triangles with Two Fixed Edge-Lengths 109

to show that .α ' < α. By the spherical cosine law, we have

cos a − cos b cos c cos λa − cos λb cos λc

. cos α = , cos α ' = ,
sin b sin c sin λb sin λc

where .λ = 1/R. Now, by Lemma 5.2, we have .cos α ' − cos α = ϕ(a) > 0. Since
'
.cos x is monotonically decreasing in .0 < x < π , we can deduce that .α < α. ⨆
⨅

5.5 Triangles with Two Fixed Edge-Lengths

For fixed positive numbers .b, c and a variable x satisfying .|b − c| < x < |b + c|,
the area .S(x) of a “planar” triangle
√ with three edge-lengths .b, c, x becomes the
maximum when .x = x0 := b2 + c2 . And .S(x) is monotonically increasing in
.x < x0 , and monotonically decreasing in .x > x0 . Let us show a similar result for

spherical triangles.
Lemma 5.3 Let .a, b, c denote the lengths / of the edges of .Δ̆ABC opposite to
−1
.A, B, C, respectively. Put .a0 = 2 sin sin (b/2) + sin2 (c/2). Then the following
2

statements hold:
⎛

(i) .BAC is a semicircle .⇔ .a = a0 ,

⎛

(ii) .BAC is a major arc .⇔ .a < a0 ,

⎛

(iii) .BAC is a minor arc .⇔ .a > a0 .

Proof Note that if x is the length of the geodesic segment P Q on .S, then the length
⎛

of the line segment P Q is .2 sin(x/2). When .BAC is a semicircle, the interior angle
of the planar triangle .ΔABC at A is the right angle. Hence we have
/
2 sin(a/2) =
. 4 sin2 (b/2) + 4 sin2 (c/2),

from which (i) follows. The other two equivalences follow similarly. ⨆
⨅
Remark 5.4 Note that the length of a semicircle is at most .π . Hence, if .b + c ≥ π,
⎛

then .BAC is always a major arc, and hence .a < a0 .

Denote by .Δ̆(x, b, c) a spherical triangle with edge-lengths .x, b, c, where .b, c
are fixed positive numbers and x is a variable satisfying .|b − c| < x < b + c.
Since .cos t is monotonically decreasing in .0 < t < π , from the spherical cosine law
(Theorem 5.3) we have the following lemma.
Lemma 5.4 Let .y(x) denote the interior angle of .Δ̆(x, b, c) opposite to the edge of
length x. Then .y = y(x) is a monotonically increasing function of x. ⨆
⨅
110 5 Spherical Geometry II

Theorem 5.7 Let .f (x) denote the area of .Δ̆(x, b, c). Then the following two
implications hold.
(1) If .b + c ≥ π, then .f (x) is monotonically increasing in .x > 0.
(2) If .b + c < π , then .f (x) is monotonically increasing in .x < x0 , and
monotonically decreasing in .x > x0 , where
/
x0 = 2 sin−1
. sin2 (b/2) + sin2 (c/2).

Thus the maximum value of .f (x) is .f (x0 ).

Proof By Euler’s formula for the spherical excess, we have

1 + cos x + cos b + cos c 2 cos2 (x/2) + cos b + cos c

. cos ε = = ,
4 cos(x/2) cos(b/2) cos(c/2) 4 cos(x/2) cos(b/2) cos(c/2)

where .f (x) = 2ε. Put .t = x/2 and .g(t) = cos(b/2) cos(c/2) cos ε. Since

cos b + cos c
.g(t) = 1
2 cos t + ,
4 cos t
we have
⎛ ⎞
cos b + cos c sin t
g ' (t) = − 21 sin t +
.
4 cos2 t
sin t
= (cos b + cos c − 2 cos2 t).
4 cos2 t
Hence

g ' (t) ⪌ 0 ⇔ cos b + cos c ⪌ 2 cos2 t

.

⇔ 2 − 2 sin2 (b/2) − 2 sin2 (c/2) ⪌ 2 cos2 t

⇔ sin2 t ⪌ sin2 (b/2) + sin2 (c/2)

/
−1
⇔ t ⪌ sin sin2 (b/2) + sin2 (c/2),

where the triple-symbol (.⪌) is meant in the same order. Since .cos(b/2) cos(c/2) >
0, and .cos t is monotonically decreasing in .0 < t < π , we have
/
f ' (x) ⪌ 0 ⇔ x ⪋ sin−1
. sin2 (b/2) + sin2 (c/2).

Now, from Lemma 5.2 and Remark 5.4 the theorem follows. ⨆
⨅
5.6 The Isoperimetric Theorem for Quadrilaterals 111

For a figure F on .S, its area is denoted by .[F ]. For example, the area of a spherical
triangle ABC is denoted by .[ABC], the area of .Δ̆(a, b, c) by .[Δ̆(a, b, c)], the area
of a spherical quadrilateral ABCD by .[ABCD].
From Theorem 5.7 (1), we have the following corollary.
Corollary 5.3 If the length of the shortest edge of a spherical triangle .Δ̆ABC is at
least .π/2, then .[Δ̆ABC] ≥ [Δ̆(π/2, π/2, π/2)]
Corollary 5.4 In the range .x1 ≤ x ≤ x2 , the area of .Δ̆(x, b, c) becomes minimum
either at .x = x1 or at .x = x2 . ⨆
⨅

5.6 The Isoperimetric Theorem for Quadrilaterals

The next lemma is proved in [520], and it is also listed as a problem in “Curve on a
Sphere” [817, p. 52].
Lemma 5.5 If a simple closed curve on .S has length smaller than .2π , then it is
contained in some hemisphere.
Proof Let .Γ be a simple closed curve on the unit sphere .S whose length is less than
2π . Let .P , Q be two points on .Γ that bisect the length of .Γ . Then, since .P , Q are
.

connected by a curve of length less than .π , the length of the geodesic segment P Q
is less than .π . Let M be the midpoint of P Q. Let .H (M) the hemisphere with pole
M. Since .P Q < π , .P , Q lie in the interior of .H (M).
We show that .Γ is contained in the interior of .H (M). Suppose that .Γ is not
contained in the interior of .H (M). Then .Γ meets the boundary of .H (M) at some
point R. Since M is the midpoint of P Q, by the rotation through .180◦ around M, P
goes to Q, and R goes to .R ∗ (the antipodal point of R that also lies on the boundary
of .H (M)). Hence .P R = QR ∗ . Let s be the length of the partial curve .P → R → Q
along .Γ (hence 2s is the length of .Γ ). Then we have

s ≥ P R + RQ = QR ∗ + RQ ≥ RR ∗ = π,
.

which implies that the length 2s of .Γ is greater than or equal to .2π , a contradiction.
⨆
⨅
A spherical polygon on .S is called cyclic if its vertices are concyclic, that is, all
vertices lie on the same circle.
Theorem 5.8 Let .a, b, c be positive numbers satisfying that .a + b + c < π. Among
the spherical quadrilaterals ABCD on .S such that .AB = a, BC = b, CD = c, a
⎛

cyclic one in which .ABCD is a semicircle has the maximum area.

Proof Since the perimeter of ABCD is less than .2π , ABCD lies in a hemisphere.
Note that we always have .[ABCD] ≤ [ABC] + [ACD]. If the diagonal segment
AC is not contained in the spherical quadrilateral ABCD, then, replacing the part
112 5 Spherical Geometry II

Δ̆ABC by its “mirror image” with respect to the plane ACO, we have one for
.

which .[ABCD] = [ABC] + [ACD] holds. Now, put .t = AC. Then .[ABCD] =
[Δ̆(a, b, t)]+[ACD]. For every .t ≤ a+b, the maximum value of .[ACD] is attained
⎛
when .ACD is a semicircle by Theorem 5.7, and hence the maximum value is given
by a continuous function .g(t) of t. Thus, the maximum value .[Δ̆(a, b, t)] + g(t) of
.[ABCD] for .AC = t is also a continuous function of t. Since the range of t is a

closed interval .a − b ≤ t ≤ a + b, there is a t where the maximum value is attained.

Therefore, there is a spherical quadrilateral ABCD that has the maximum area.
⎛

In this ABCD, .ACD is a semicircle. The quadrilateral ABCD also contains the
diagonal BD. (For otherwise, replacing the part .Δ̆ABD by its “mirror image” with
respect to the plane BDO, we could have a quadrilateral with larger area.) Then, by

⎛
a similar argument as it was used above, we see that .ABD is also a semicircle. Then,

(
since AC is contained in ABCD, B must lie on the subarc .AC of the semicircle
⎛
⎛

ACD. Therefore, ABCD is cyclic and .ABCD is a semicircle.

. ⨆
⨅
Lemma 5.6 Every spherical quadrilateral with perimeter less than .2π can be
continuously deformed, with keeping its four edge-lengths, to a cyclic one.
Proof Let ABCD be a non-cyclic spherical quadrilateral with perimeter less than
2π . We may suppose that D lies in the interior of the circumscribed cap .cap(ABC)
.

of .Δ̆ABC. We also assume that .BC +CD ≥ DA+AB. If we increase the geodesic
distance between D and B (with keeping the four edge-lengths fixed) so that it
becomes nearly .DA + AB (in other words, ./ DAB becomes nearly .180◦ ), then D
goes to the exterior of .cap(ABC). Therefore, there is a moment that D lies on the
boundary of .cap(ABC). At that moment, ABCD becomes cyclic. ⨆
⨅
A spherical polygon is said to be inscribed in a cap if the spherical polygon is
contained in the cap and all vertices of the spherical polygon lie on the boundary
circle of the cap. The cap is also called the circumscribed cap of the spherical
polygon. Thus a spherical polygon inscribed in a cap is a cyclic one.
Theorem 5.9 (Isoperimetric Theorem for Spherical Quadrilaterals) Among
the spherical quadrilaterals on .S with fixed four edge-lengths and perimeter less
than .2π , a cyclic one has the maximum area.
Proof Let ABCD be a non-cyclic quadrilateral on .S with given fixed edge-lengths.
We may suppose that CD is the longest edge, and .AD ≥ BC. Then, since
.AD + DC ≥ AB + BC, D is not contained in .Δ̆ABC. If B lies in the interior

of .Δ̆ACD, then replace B by its “mirror image” with respect to the plane ACO.
By such a replacement .[ABCD] becomes larger, and B goes outside .Δ̆ACD. By
Lemma 5.4, ABCD can be deformed with fixing four edge-lengths, and fixing two
vertices .C, D, to a cyclic quadrilateral .A' B ' CD. We may suppose that .A' and A lie
in the same side of the great circle CD. We show that .[ABCD] < [A' B ' CD]. Let
' ' '
.A P be a (spherical) diameter of the circumscribed spherical cap of .A B CD, see

Fig. 5.8.
5.7 Exercises 113

Fig. 5.8 .A' P is a diameter of D

the spherical cap

A P

B C

Since CD is the longest edge and .A' D ≥ B ' C, it follows that .A' P and CD
intersect. (It is possible that .P = C, but we assume .P /= C in the following.) We
have

[A' B ' CD] = [A' B ' CP ] + [A' DP ] − [CP D].

.

Let us divide our argument into two cases.

(i) The case that AP and CD intersect: Since .B /∈ Δ̆ACD, D /∈ Δ̆ABC, AP
never intersects BC. Hence we have .[ABCD] = [ABCP ]+[ADP ]−[CDP ].
Since .[ADP ] < [A' DP ], [ABCP ] < [A' B ' CP ] by Theorems 5.7 and 5.8,

[ABCD] ≤ [ABCP ] + [ADP ] − [CP D]

.

< [A' B ' CP ] + [A' DP ] − [CP D] = [A' B ' CD].

(ii) The case that AP and CD do not intersect: In this case, the spherical
quadrilateral ABCP contains D in its interior. Hence we have .[ABCD] =
[ABCP ] − [ADP ] − [CDP ]. Since .[ABCP ] < [A' B ' CP ] by Theorem 5.8,

[ABCD] ≤ [ABCP ] − [ADP ] − [CP D]

.

< [A' B ' CP ] + [A' DP ] − [CP D] = [A' B ' CD].

⨆
⨅

5.7 Exercises

5.1∗ Let A, B be two points on the unit sphere with the same latitude π/4 and
the difference of their longitudes π/2. Find the spherical distance between A
and B.
114 5 Spherical Geometry II

5.2∗ In an equilateral spherical triangle on the unit sphere, let d be its edge-length
and θ be its interior angle. Prove that

cos θ cos d
. cos d = and cos θ = .
1 − cos θ 1 + cos d

5.3∗ Prove that in Δ̆ABC the following implications hold.

(i) AC > AB ⇒ / B > / C.
(ii) AC = AB ⇒ / B = / C.
(iii) / B > / C ⇒ AC > AB.
5.4∗ Let Δ̆ABC be a non-obtuse triangle (that is, none of the interior angles
is greater than π/2). Prove that if X, Y ∈ Δ̆ABC, then XY ≤
max{AB, BC, CA}.
5.5 Prove that it is impossible to take four points A, B, C, D on a sphere (of any
radius) such that
√
.AB = BC = CA = 1, AD = BD = CD = 1/ 3.

5.6∗ From Theorem 5.7 (2), derive the following result: Among the convex
quadrilaterals with edge-lengths a, a, b, b on the unit sphere S, one that is
inscribed in a spherical cap has the maximum area.

5.8 Notes

In these notes we will concentrate on the following topics: spherical trigonometry

(particularly, also the corresponding sine and cosine theorems) and triangle compar-
ison theorems (due to Toponogov).

5.8.1 Notes on Spherical Trigonometry and the Respective

Cosine and Sine Laws

Spherical trigonometry, and particularly the spherical sine and cosine theorems (see
our Sect. 5.3), represent an old, well-researched mathematical topic. Therefore the
following part cannot be dominated by recent research papers and is more dedicated
to books with interesting representations of the little field. However, the presented
references contain nevertheless interesting sides and new applications. Most of
them refer to classical spherical geometry, i.e., to the geometry of the 2-sphere.
Extensions to spherical n-dimensional spaces are normally identified as such.
5.8 Notes 115

Some History

Highlights from the history of spherical trigonometry are connected with names
like Menelaus, several Indian and Arabic mathematicians from the Middle Ages,
Regiomontanus, Pitiscus (a Silesian mathematician from the sixteenth century),
Napier, Euler and various further contributors.
In Björnbo’s long paper [90] (JFM 33.0062.01) Menelaus’ essential achieve-
ments in spherical trigonometry are summarized and honored, and further existing
material is ordered.
The contributions of Islamic scientists during the early Middle Ages are nicely
discussed and summarized in [204], [657] and [658], showing also once more the
impact of Menelaus’ findings. Although various efforts had their origins in problems
from astronomy, the Arabic mathematicians also discussed notions from spherical
geometry (e.g., sines and chords) in more abstract terms.
In view of spherical geometry and trigonometry in India (Middle Ages), e.g. also
going back to works of Brahmagupta (seventh century), we refer to Datta and Singh
[198], i.e., mainly to the last part of [198], which is dedicated to spherical geometry.
The role of Regiomontanus (fifteenth century) as well as Clavius and Pitiscus
(sixteenth century) in spherical trigonometry is studied by Olivares [603]. Pitiscus
introduced the name “trigonometry” and presented possibly the first proper proof
of the spherical cosine theorem (in 1595). The paper [603] shows the evolution of
the cosine theorem over all these centuries, and also its proof due to Pitiscus is
presented.
Dietrich and Girstmair [213] study Napier’s achievements in spherical trigonom-
etry (namely Napier’s rules, his aim to use his logarithmic calculus mainly in
trigonometry and related topics). One sees also that the right notation in this little
field was not really optimal (a level that Euler reached a little later). This is also
the right place to cite Havil’s book [385] on the life and mathematical work of
Napier, including of course also spherical geometry. A further essential contribution
to Napier’s work on spherical trigonometry is [732] due to Silverberg.
The deplorable condition that spherical trigonometry disappeared from the cur-
riculum of general elementary mathematics (like also its connections to navigation
and astronomy) led Plofker to write the very nice historical article [640] on this little
field, representing contributions from the respective Hellenistic and Islamic period
to the corresponding Indian and medieval European time. Later the obtained insights
were exceeded by the Europeans to the framework of non-Euclidean geometries,
to which spherical geometry clearly belongs. Especially Euler’s achievements are
recognized in detail. E.g., it is shown in [640] how Euler established the foundations
of spherical trigonometry via the method of maxima and minima, and how he derived
the spherical cosine theorem by an argument using minima. Euler’s contributions
to spherical geometry and trigonometry are also nicely summarized in [618] by
Papadopoulos, and an informative historical survey of the whole field is presented
there. The author gives also a mathematical overview of Euler’s results in spherical
geometry, particularly including spherical projections and cartography. An outline
116 5 Spherical Geometry II

of the scientific connection between Euler and Lagrange is given, too. We mention
here also once more the Proceedings [138] on mathematical geography.
It can be seen by the collection [439] of Kambly that, at the beginning of the
twentieth century, the level of the material on spherical geometry covered in the
mathematical Abitur was high. This example refers to Prussia. A similar example is
the book [188] of Crawley. It is a selection of one thousand exercises (with solutions)
in plane and spherical trigonometry, which appeared 1914 in Pennsylvania.
Two essential overviews from the first half of the twentieth century which contain
much material on the history and on the (in those days) recent state in spherical
geometry and trigonometry are those due to Zacharias [828, Parts 18 and 19] and
Tropfke [782, Part II]. And, e.g., Hessenberg [395], Macaulay [516, Chapter XIII]
as well as Kells et al. [445] were popular related reference books in this period.
We finish this little part on historical aspects by the relatively recent and excellent
book [131] of Van Brummelen. The third chapter discusses the basic ancient
contributions of Hipparchus of Rhodes and Menelaus of Alexandria (the author
of Sphaerica—the spherical version of Menelaus’ theorem plays a central role
here), and in Chap. 4 the achievements of Indian and Islamic mathematicians are
appreciated. The latter had a special motivation to apply spherical geometry: their
praying in the direction to Mecca. Further on, the contribution of Napier, Euler
and others are comprehensively discussed. In the very final part, applications like
modern celestial navigation are presented.

More Recent Representations

In the middle of the twentieth century, the Hilbert doctorands Lietzmann and Dörrie
wrote the two nice and comprehensive monographs [510] and [217], respectively.
The first one clearly and systematically provides all the needed tools from school
geometry and presents numerical and constructive methods needed in spherical
geometry; applications in geodesy and astronomy are discussed, too. The second
book [217], extremely rich in content and dedicated to planar as well as spherical
trigonometry, shows many applications far from each other and therefore presenting
a broad view on the richness of the topic. E.g., extensions of Ceva’s theorem on the
sphere, location of ships, sundials, and Gauss’s method from orbital mechanics (a
subfield of celestial mechanics) are presented.
Not far from this, the three books [315] by Graf, [731] by Sigl and [757]
by Steinert have a strong reference to cartographic, geodesic and astronomic
applications of spherical geometry and trigonometry in common. Here we can
add Hammond’s monograph [377] on spherical trigonometry and applications,
e.g. emphasizing Napier’s rules and navigation, and showing also a balance in its
analytical and geometric interpretations.
A very nice book chapter devoted to spherical trigonometry and respective sine
and cosine theorems is [100, § 6] by Boltyanski and Yaglom (where a detailed
study of spherical trigonometry, based on vector calculus, is given). An analogous
way to the cosine theorem of spherical trigonometry is presented in [45], see also
5.8 Notes 117

the related classroom note [396]). Further on we refer to [675, § 4] by Rosenfeld,

where a complete derivation of these theorems and of relatives is given.
Also the important geometry book [75] of Berger contains a longer discussion
of spherical triangles and trigonometry, namely the whole Section 18.6 (extensions
to spherical polygons follow in Section 18.7). Due to the close relations between
spherical and Riemann geometry, also the exposition [677, § 2] should at least be
mentioned here.
Bigalke’s comprehensive monograph [87] is generally seen as a more modern
continuation of “classical books” (such as [510] and [217]) in the field considered
here. In Chapter 2 of [87] a whole theory of spherical triangles is given, and one
can find as basic theorems also the spherical sine and cosine theorems and various
interesting applications.
Jennings dedicates the whole Chapter 2 of his monograph [428] to spherical
geometry. He derives basic results from spherical trigonometry using cross products,
shows the relation between angular defect and area, and derives the laws of sines
and cosines for sides and angles. Discussed applications come from navigation and
cartography.
Section 5.1 of the geometry book [646] of Prasolov and Tikhomirov contains
only a short part treating spherical geometry, but the laws of sines and cosines are
carefully derived there. From the viewpoint of differential geometry, Wilson [815]
studies spherical geometry in his Chapter 2. He presents properties of curves on
the sphere using the spherical distance metric, and he treats the spherical laws of
cosines and sines.
The fifth chapter of the book [605] of Ostermann and Wanner comprehensively
presents material on (spherical) trigonometry and its applications, and particularly
Sects. 5.6 and 5.7 study also sine and cosine laws with nice applications. And we
mention the monograph [103] of Borceux whose Section 14.2 contains also related
results. Of course, also here the basic book [131] of van Brummelen should be cited
once more, especially Chaps. 2 and 5.
There are various books in the field of celestial mechanics. In their monograph
[596], Neutsch and Scherer introduce modern methods from this field as mathemat-
ical ones, to present also different aspects of current research there. The first two
parts present historical facts and needed mathematical notions and concepts, among
them also spherical trigonometry.
Regarding applications, we mention the book [135] of Burkholder. He introduces
a global spatial data system which is able to use different types of geoinformation
frameworks; it is a collection of mathematical concepts (e.g., certain equations
from geometrical geodesy and solid geometry are used) that can handle spatial data
locally as well as globally. In Chap. 3 mathematical fundaments are given, among
them also basics from spherical trigonometry not commonly covered in elementary
mathematical courses.
Also the book [416] of Hull should be mentioned here. It shows the relations
between Origami and various mathematical concepts; spherical trigonometry is one
of them.
118 5 Spherical Geometry II

A Few Related Journal Articles

Yang and Zhang [824] establish a higher dimensional extension of the Neuberg-
Pedoe inequality yielding pairs of directly similar simplices. They obtain also an
application of this result in spherical trigonometry and conjecture, that it holds also
in n-dimensional elliptic geometry.
With the help of stereographic projection, Jeger [427] elegantly derives sine
and cosine formulas for spherical triangles. McCleary [562] succeeds in proving
simultaneously various basic trigonometric relations for spherical and hyperbolic
geometry, among them also the laws of sines and cosines. The list of references in
this article is very interesting.
Supporting the thesis that there are close relations between the theory of
integrability of (discrete) dynamical systems and fundamental geometric structures,
Petrera and Suris [634] verify that the cosine laws for simplices in spherical space
define in various ways integrable systems. For this purpose, e.g. the derivation of
the spherical cosine and sine laws in terms of Gram matrices is recalled, and then
nice examples of integrable dynamical systems are obtained.
Slutskiy [742] describes an interesting, alternative approach (150 years old)
of the Belgian scientist de Tilly to hyperbolic geometry, and also his studies of
spherical geometry. In his original contributions, de Tilly also deduced the laws of
sines and cosines in spherical space (see particularly Section 4 in [742]).

5.8.2 Notes on Triangle Comparison Theorems

Roughly speaking, the triangle comparison theorem (often called Toponogov’s

theorem) compares the angles of two geodesic triangles with the same side
lengths lying on surfaces with different curvatures; see our Sect. 5.4 above. By
the following overview, one sees the importance of this theorem in differential
geometry, particularly in Riemannian geometry.

Some Historical Remarks

More precisely, the names Aleksandrov Comparison Theorem for the Angles of a
Triangle (see Toponogov [780, Theorem 3.9.1]) for two dimensions, Aleksandrov-
Toponogov theorem (cf. Chavel’s book [158, p. 400] and [15]) and Cartan-
Aleksandrov-Toponogov theorem (see Burago et al. [134, p. 240] in higher dimen-
sions are common, and this theorem has two variants. In the first one, lengths are
compared: given two geodesic triangles with two sides of the correspondingly same
length and the angle between them of the same measure, lying in two surfaces
with Gaussian curvature of the first everywhere larger than or equal to the maximal
Gaussian curvature of the second, then the third triangle side is smaller than or equal
to the third triangle side. The second variant compares angles: if, under the same
5.8 Notes 119

conditions for the two surfaces, the triangles have correspondingly congruent sides,
then the angles are respectively larger or equal to their corresponding counterparts.
The importance of these statements is nicely expressed by Petersen [633, p. 333]:
“Toponogov’s theorem is a very useful refinement of Gauss’ early realization that
curvature and angle excess of triangles are related.”
Pambuccian and Zamfirescu [614] correct statements around the history of the
theorem in the following way: The two-dimensional case is usually presented as
result due to A. D. Aleksandrov (see [16]), and the n-dimensional Riemannian
case as result of Toponogov [779]. But as the authors of [614] underline, the two-
dimensional case was first derived by Pizzetti [639], and they present a biographical
sketch and a short outline of Pizzetti’s proof of the triangle comparison theorem. In
[614] it is also explained how the names of Cartan and also of Darboux came into
this framework, yielding finally the name of the related CAT(k) spaces by the initials
of Cartan, Aleksandrov, and Toponogov.
It turned out that the proofs of Pizzetti, Aleksandrov, and Toponogov cannot be
conveniently compared, since the definitions of the used terms are very different. In
this direction, the interesting role of Rauch [660] was evaluated by Chavel [158, p.
399] (saying “by triangle comparison theorems we mean global forms of the Rauch
comparison theorem”) and Grove and Petersen [335, p. ix]. Further interesting
remarks on the history of triangle comparison theorems can be found in [443], [77,
pp. 52 and 6] and [632, p. 325].
Maehara [530] gave very recently an elementary proof of one classical form
of Toponogov’s theorem (which is essential for our book here), namely: if two
spherical triangles on two different spheres have the same side-lengths, then the
corresponding angles of that triangle which lies on the sphere with larger radius
are smaller; see the beginning of our Sect. 5.4. Replacing the two spheres in this
version by suitable surfaces or manifolds, one gets Toponogov’s theorem and more
general statements, very important in Riemannian geometry and further settings. As
a helpful tool it found entrance into various further mathematical disciplines. In the
following we present a suitable selection of related works.

The Comparison Theorem in Several Books

Toponogov himself published, late in his life, a book on differential geometry (see
[780]). Its third chapter treats topics like geodesics, shortest paths, and the intrinsic
geometry of surfaces (Gauss-Bonnet and Hopf-Rinow theorem). At the end, a proof
of Aleksandrov’s version of the comparison theorem is given.
Going chronologically back, we start with the lecture notes [333] of Gromoll,
Klingenberg, and Meyer. In Chap. 6 the authors collect various comparison
theorems, among them the angle comparison theorem of Toponogov.
Cheeger and Ebin present with [159] a comprehensive, elegantly written book
on global Riemannian geometry; they are particularly interested in geometric and
topological consequences following from assumptions on curvature. They create
techniques for comparing the geometry of a general manifold with that of a simply
120 5 Spherical Geometry II

connected model space of constant curvature. E.g., these are based on comparison
theorems, like also that of Toponogov.
Chavel’s monograph [158] gives a comprehensive and broad approach to Rie-
mannian geometry, having unexpectedly extensive parts of notes and exercises at
the end of each chapter and, following from this, a large bibliography, thus opening
enough space for further ideas and extensions. Discretization of Riemann manifolds,
Hamiltonian and Lagrangian mechanics, and manifolds without conjugate points
(including Hopf’s theorem) are interesting topics. In the last chapter, entitled
“Comparison and finiteness theorems”, the author studies Rauch’s comparison
theorem and applications and also Aleksandrov-Toponogov triangle comparison
theorems.
Also in the book [583] of Morgan and Tian on Ricci flow and the Poincaré
conjecture one can find Toponogov’s result as a helpful tool. In the second chapter
the theory of manifolds of non-negative sectional curvature is discussed, and it
contains the Toponogov theory on the geometry of triangles in this setting.

Research Papers

We continue with a selection of research papers, mainly in chronological order, thus

showing also some of the historical milestones.
Elerath [241] extends Toponogov’s comparison theorem by studying the curva-
ture along geodesics with the curvature along geodesics in a suitable comparison
space S. The usual comparison theorems take S as space of constant curvature, but
the author uses instead a “flattening” surface of revolution in which the curvature
decreases with the distance from the vertex. Thus, first he has to establish certain
crucial properties of geodesics and cut points in such a surface. These ideas with
geodesics are continued by Itokawa et al. in [420] and [421] regarding manifolds
with restricted radial curvature. (Recall that the radial curvature of a Riemannian
manifold S is the curvature of S along and tangent to the geodesics emanating from
a fixed base point.)
Harris [379] derives an analogue of Toponogov’s triangle comparison theorem
for Lorentz manifolds. He obtains first two Rauch comparison theorems using
timelike geodesics, and using a new concept of sectional curvature he finally gives
his triangle comparison theorem for triangles which have all three sides consisting
of timelike geodesic segments.
In [1], Abresch studies the topology of asymptotically non-negatively curve
manifolds. He considers an extension of Toponogov’s triangle comparison theorem
to the situation that the comparison manifold is not simply connected and of constant
curvature, but a surface with only rotational symmetries. In [2] Abresch continues
[1] improving Gromov’s “Betti number theorem” from [334] by extending it to non-
compact asymptotically nonnegatively curved manifolds. Important expositions,
containing survey parts, around Riemannian comparison theorems and related topics
on manifolds of nonnegative sectional curvature are those of Karcher [443] and of
Kobayashi (the pages 140–169 before [443]). A series of essential results in this area
5.8 Notes 121

is selected and presented with detailed proofs. Also interesting technical details are
emphasized, and a comprehensive survey on essential developments in the studied
field is given. In particular, a new proof of Toponogov’s theorem is given, with
applications to critical points of distance functions, the cut locus and the sphere
theorems. Karcher’s exposition is very interesting even for experts in the field,
since in many cases his arguments differ essentially from presentations in standard
textbooks. Marenich [544] proves a characterization theorem of hyperspheres
as special Riemannian manifolds with positive sectional curvature containing an
isometric embedded neighbourhood of an equatorial Euclidean sphere. The proofs
are based on an extremal case of Toponogov’s triangle comparison theorem.
Dai and Wei [194] establish a triangle comparison of Toponogov type for
Riemannian manifolds without sectional curvature bound. They also give an
application to the topological type of complete manifolds with non-negative Ricci
curvature and small volume growth.
Colding [171] investigates n-dimensional closed Riemannian manifolds with
positive Ricci curvature regarding their Gromov-Hausdorff distance to an n-sphere,
and for the proof he develops a suitable version of Toponogov’s triangle comparison
theorem for positive Ricci curvature. The same author presents in [172] a new
integral estimate of distances and angles on manifolds having a lower bound for
their Ricci curvature. He thus obtains an integral version of Toponogov’s theorem
for Ricci curvature.
In [479], Kürzel proves a version of the classical Toponogov triangle comparison
theorem, where the model space is a surface (possibly singular) of revolution having
non-constant curvature.
For a given Kähler manifold S, Adachi [6] studies a nontrivial Kähler magnetic
field (i.e., a closed two-form being a constant multiple of S) and finds a comparison
theorem analogous to Toponogov’s theorem in Riemannian geometry. Toponogov’s
triangle is here replaced by a bow-shape which consists of a trajectory segment and
a kind of geodesic segment on manifolds with sectional curvature bounded from
above.
Kondo and Tanaka [470] prove the finiteness of the topological type of a
complete open Riemannian manifold S with base point .x ∈ S whose radial curvature
at x is bounded below by a non-compact model surface of revolution .S ' with finite
total curvature and has no pair of cut points in a sector based at .x ' ∈ S ' . To prove this,
they use a new comparison theorem of Toponogov type, and they apply the obtained
theorem to get a partial answer to a conjecture of Milnor that complete open
Riemannian manifolds with nonnegative Ricci curvature have finitely generated
fundamental group.
To get similar results, Kondo and Tanaka derive in [471] and [472] a Toponogov
comparison theorem for open triangles standing on the boundary of complete
connected n-dimensional Riemannian manifolds with smooth convex boundary
whose radial curvature is bounded below by certain model surfaces.
Lang and Schroeder present the expository note [488] with a transparent proof
of Toponogov’s theorem for Aleksandrov spaces in the general case (not assuming
local compactness of the underlying metric space). In other words, they show that
122 5 Spherical Geometry II

with certain local assumptions in geodesic metric spaces the comparisons under
consideration hold also in the large.
Kondo et al. [473] study the relationship between the topology and the curvature
of Finsler manifolds. They prove a finiteness theorem of topological type and a
diffeomorphism theorem as applications of a Toponogov-type triangle comparison
theorem of Finsler geometry.
Hebda and Ikeda [388] prove the following result: let S be a complete
Riemannian manifold with a base point x, and .S ' be a complete, simply connected
surface, rotationally symmetric about the vertex .x ' such that the cut locus of every
point .x ' lies in the opposite meridian of .x ' . Now suppose that .(S ' , x ' ) has weaker
radial attraction than .(S, x). Then, for every suitable geodesic triangle in S, there
is a corresponding geodesic triangle in .S ' whose respective sides are equal and that
satisfies the angle comparison and Aleksandrov convexity in Toponogov’s Theorem.
In addition, it is shown that rigidity fails in this system. Hebda and Ikeda [389]
prove an analogue of Toponogov’s triangle theorem for surfaces of revolution as
model spaces. Namely, for a model surface and a Riemannian manifold with a
fixed base point, the authors give necessary and sufficient conditions under which
each geodesic triangle in the manifold, having a vertex at the base point, has a
corresponding Aleksandrov triangle in the model.
Hu et al. [406] present an elegant new proof of Toponogov’s Theorem for
triangles in complete Aleksandrov spaces. This proof does not depend on further
important, basic approaches to that theorem for Aleksandrov spaces in the literature;
see, e.g., Burago, Gromov and Perel’man [133] and Shiohama [730]. The new proof
in [406] yields a new idea useful in the framework of Aleksandrov spaces, namely
that of using variations of geodesics to detect curvature.
Chapter 6
The Problem of Thirteen Balls

Millions saw the apple fall, but Newton asked why.

— Bernard Baruch

How many unit balls can touch one and the same unit ball simultaneously? (We
assume that unit balls cannot overlap each other.) This problem is known as the
problem of thirteen balls, or the Gregory-Newton problem. Applying spherical
geometry and a lemma on planar graphs, we answer this problem.

6.1 The Lemma on Proper Diagonals

We start with a geometric statement.

Lemma 6.1 Let .Δ̆ABC be a spherical triangle with edge-lengths .c = AB, b =
AC, c ≤ b, in which ./ B > π/2. Suppose that .A, C are fixed and B moves on
.S with keeping the length .c = AB fixed. Then the following three statements are

equivalent:
(i) the length of BC decreases,
(ii) ./ BAC decreases,
(iii) ./ BCA decreases.
And any of (i), (ii), and (iii) implies that the area .[ABC] decreases.
Proof The equivalence of (i) and (ii) follows from the spherical cosine law. Let
.Λ be the spherical cap with center A, radius .c = AB, and let .γ the arc of the
boundary circle of .Λ intercepted by the interior angle ./ A of .Δ̆ABC, see Fig. 6.1.
Since .c = AB ≤ b = AC and ./ ABC > π/2, the arc .γ is contained in .Δ̆ABC.
Hence (ii) is equivalent to the property that B becomes an interior point of the arc
.γ , which is equivalent to (iii), and also implies that the area .[ABC] of .Δ̆ABC

decreases. ⨆
⨅

© The Author(s), under exclusive license to Springer Nature Switzerland AG 2024 123
H. Maehara, H. Martini, Circles, Spheres and Spherical Geometry,
Birkhäuser Advanced Texts Basler Lehrbücher,
https://doi.org/10.1007/978-3-031-62776-7_6
124 6 The Problem of Thirteen Balls

Fig. 6.1 Lemma 6.1 A

Λ
c

γ B
b

Remark 6.1 Suppose that .AB = AC in .Δ̆ABC. Then

./ ABC ⪌ π/2 ⇐⇒ AB ⪌ π/2 (the triple symbol ⪌ in the same order)

holds.
Let .Q = ABCD be a spherical quadrilateral. Note that .A∗ /= C. For otherwise,
⎛

.ABC is a geodesic segment, and B cannot be a vertex. Suppose that the diagonal AC
is contained in .Q. Then .Q is divided by AC into two spherical triangles .Δ̆ABC and
.Δ̆ACD. If D is not an interior point of the cap circumscribed to .Δ̆ABC, then AC

is called a proper diagonal of the spherical quadrilateral .Q. In this case, B is not an
interior point of .cap(ACD). If D lies in the interior of .cap(ABC), then the diagonal
BD is also contained in .Q, and .Q is divided by BD into two spherical triangles
.Δ̆ABD, Δ̆BCD. In this case, C is not an interior point of .cap(ABD), and hence

BD is a proper diagonal of .Q. If .Q is a cyclic quadrilateral, then both .AC, BD are

proper diagonals of .Q. Thus, a spherical quadrilateral that can be divided into two
spherical triangles always has a proper diagonal.
Lemma 6.2 (The Lemma on Proper Diagonals) Let .Q be a spherical quadrilat-
eral that has a proper diagonal. If we deform .Q with keeping its four edge-lengths
so that a proper diagonal becomes shorter, then the area of .Q decreases.
Proof Let .Q = ABCD and suppose that AC is a proper diagonal of .Q. If B lies on
the boundary of .cap(ADC), then .Q is cyclic. In this case, if we decrease AC with
keeping the length of the edges of .Q, then .Q becomes not cyclic, and hence its area
decreases. So, we may suppose that B lies outside .cap(ADC).
(1) First, consider the case that one of ./ BAC, / BCA is greater than .π/2, and
one of ./ DAC, / DCA is greater than .π/2. If we shorten AC with keeping
the lengths of .AB, BC, CD, DA, then both .[ABC], [ACD] decrease by
Lemma 6.1, and hence .[Q] decreases.
(2) Next, suppose ./ BAC ≤ π/2 and ./ BCA ≤ π/2. Let .Z1 , Z2 be the centers
⎛

of .cap(ABC), cap(ACD), respectively. Since ./ BAC ≤ π/2, the arc .BAC

is a major arc by Corollary 4.4. Hence .Z1 and A lie in the same side of BC.
6.2 Major Triangles 125

Fig. 6.2 Proper diagonal AC A ..................... .... ....

.. .... .... .....●
...... .. . .. .... ....
.. .. ................. .... ....
.... ..
... .... ..... ... ........ ..................... ...
. .... ..
.. .. ....
..
.. .... ....................
.. .... ...........
... .. .... ... ..........
.. .... .. ....
.... .. ...........
.. .. .. .... ... .........
... .... .. ..... . .........
.........
.. .. .
.. ..... .. ........
. .... ...
.....
. .
.. . . .
. . ..
.... .... .............● B
... ... . .. .
. .. . . . . .... .... .... ..● .... .
.
.
.
... .
...
...
. .. .. ..
.. . .... .... .... .. .. .. .... . ....
..
...
... ● .. .P
.. ...
... . ....
...
..
..
.. ..
... Z 2 ..
.. ... ..
.... . .....
....
.. ... ..
. ..... ... ......
.. ... . .. . . .
.. ... .. ... .......
............ .. ..... ........
●... ...................................... .. ... .....
D ...
... .
... ...
............................ . . .
.............................................
..
..
. ..●
. .... .... ...
.... .... .
... .... ......... .... .... .... .... C

Similarly, since ./ BCA ≤ π/2, .Z1 and C lie in the same side of BA. Therefore,
.Z1 is contained in the lune determined by the angle ./ ABC of .Δ̆ABC. Let
⎛

M be the midpoint of the arc .ABC. Then .Z1 lies between M and .Z2 and the
great circle .MZ2 perpendicularly bisects AC. If .Z1 , Z2 , M lie in the same side
of AC, then .Z2 is an interior point of .Δ̆ABC. If .Z2 /∈ Δ̆ABC, then since
∗
./ Z2 AC < π/2, / Z2 CA < π/2 and ./ B AC ≥ π/2, / B CA ≥ π/2, .Z2
∗
∗
is an interior point of .Δ̆ACB , and hence .Z2 is an interior point of the lune
determined by the angle ./ ABC of .Δ̆ABC. Let P be the intersection of .BZ2
and the boundary circle of .cap(ACD). Since .Z2 is an interior point of the lune
with angle ./ ABC, P is an interior point of .Δ̆ABC, see Fig. 6.2.
Since ./ AP Z2 < π/2, / CP Z2 < π/2, we have ./ AP B > π/2 and ./ CP B >
π/2. Suppose that AC is shortened with keeping the lengths of .AB, BC, CD, DA
and .AP , P C.
(i) Since the cyclic quadrilateral ADCP becomes not cyclic, the area .[ADCP ]
decreases.
(ii) Since ./ ABC becomes smaller, one of ./ ABP , / CBP , say ./ ABP , becomes
smaller. Then, since ./ AP B > π/2, ./ BAP also becomes smaller by
Lemma 6.1, and hence P B becomes shorter, which implies that .[ABP ]
and .[BCP ] decrease. Therefore .[ABCD] = [ADCP ] + [ABP ] + [ACP ]
decreases. ⨆
⨅

6.2 Major Triangles

A spherical triangle is called a major triangle if it contains the center of the

circumscribed cap.
Example 6.1
(i) For every .x, y, z ∈ [π/3, π/2], .Δ̆(x, y, z) is a major triangle.
(ii) For every .x, y ∈ [π/3, 2π/3], .Δ̆(x, y, π/2) is a major triangle.
126 6 The Problem of Thirteen Balls

To see (i), let .Δ̆ABC be a spherical triangle with edge-lengths .x, y, z. Then
the edge-lengths√ of the planar triangle .ΔABC lie between .2 sin(π/6) = 1 and
.2 sin(π/4) = 2. Thus the planar triangle .ΔABC is not an obtuse triangle, and
hence it contains the center of the circumscribed circle. Hence .Δ̆ABC also contains
the center of circumscribed cap, and it is a major triangle.
(ii) is seen as follows. The planar √ triangle formed by the three vertices of
.Δ̆(x, y, π/2) has one edge of length . 2, and the lengths of the other two edges lie
√
between 1 and .2 sin(π/3) = 3. Hence it is not an obtuse triangle, and it contains
the center of the circumscribed circle. Hence .Δ̆(x, y, π/2) also contains the center
of its circumscribed cap.
Lemma 6.3 (Major Triangle Lemma) In a major triangle .Δ̆ABC, if we decrease
the length of AC with keeping the lengths of edges .AB, BC, then the area .[ABC]
decreases.

⎛
Proof For a major triangle .Δ̆ABC, the arc .ABC is a major arc, and the lemma
follows from Lemma 5.1 and Theorem 5.7. ⨆
⨅

6.3 On a Triangulation of a Quadrilateral

To solve the problem of thirteen balls, let us prepare a lemma concerning plane
graphs. By a triangulation of a quadrilateral, we mean a plane graph such that the
boundary of the unbounded outer face is a 4-cycle, and other faces are all triangles.
If the number of triangles in a triangulation G of a quadrilateral is N , then G is
called a triangulation of a quadrilateral into N triangles. A 3-cycle of a plane graph
is called a non-facial 3-cycle if it is not the boundary of a face of G.
For example, Fig. 6.3 shows a triangulation of a quadrilateral into 20 triangles. It
has 13 vertices and has no non-facial 3-cycle. The maximum degree of this graph is
6.
Lemma 6.4 Let G be a triangulation of a quadrilateral into 20 triangles. If (i) G
has 13 vertices and (ii) G has no non-facial 3-cycle, then the maximum degree of G
is greater than 5.

Fig. 6.3 A triangulation of a

quadrilateral into 20 triangles
6.3 On a Triangulation of a Quadrilateral 127

Proof The proof works by contradiction. Suppose that there is a triangulation .G =

(V , E) of a quadrilateral into 20 triangles such that (i) it has 13 vertices, (ii) it has
no non-facial 3-cycle, and (iii) the maximum degree is at most∑ 5.
Since the number of edges is .(4 + 3 × 20)/2 = 32, we have . v∈V deg v = 64 by
the hand-shaking lemma. Since .5 × 13 = 65, G must have one vertex of degree 4,
and 12 vertices of degree 5. Let ABCD be the boundary 4-cycle of the unbounded
face of G. In the following we are going to derive contradictions by dividing our
argument into two cases.
Case (1): The degrees of the vertices .A, B, C, D are all 5:
From each of .A, B, C, D, three edges go into the interior of the “quadrilateral”
ABCD. Since there is no edge that connects A and C, or B and D by
condition (ii), the terminal vertices of these edges emanating from .A, B, C, D
are different from .A, B, C, D. Noting that each of the edges .AB, BC, CD, DA
of the quadrilateral ABCD must be an edge of a triangular face of G, we may
put .v1 , v2 , v3 , . . . , v7 , v8 as the terminal vertices of these edges emanating from
.A, B, C, D, as shown in Fig. 6.4 left. (At present, we do not know whether these

8 vertices .v1 , v2 , . . . , v8 are all different or not. But we know, for example, that
.v2 and .v7 are different by condition (ii).) Among these 12 edges emanating from

.A, B, C, D into the interior of the quadrilateral ABCD, any pair of “neighboring

edges” emanating from the same vertex must be two edges of one and the same
triangular face. Hence .v1 and .v2 must be adjacent. Similarly, .v2 and .v3 , .. . . ,
.v7 and .v8 , .v8 and .v1 must be adjacent, respectively. See Fig. 6.4 right. Then,

since the maximum degree of each vertex of G is at most 5, we can deduce

that the vertices .v1 , v2 , v3 , . . . , v8 are all different. Since the graph G has just
one vertex of degree 4, we may suppose that .v7 or .v8 is the vertex of degree 4.
Then, from each of the vertices .v2 , v4 , v6 , one more edge emanates. Let .x, y, z
be the terminal vertices of these three edges emanating from .v4 , v6 , v8 . (At
present, we do not know whether .x, y, z are different or not.) By condition (ii),
.x /= v4 , v8 . Since no other edge emanates from .v2 , the edges .v2 v1 and .v2 x are

neighboring edges. Hence .v1 and x are adjacent. Similarly .v3 and x are adjacent.

A●....................... ●D
.....
.............
A●....................... .....
.............●D
.. . ...
.. ........ ............... ................... . . ...
.. ........ ............... ...................
... .... ....... .................... . . . . ........... ........... ... ... .... ....... .................... . . . . . ........... ........... ...
.. ... .....
... .... ......
v
...........
........... ....2
..........
...........
.... ...............
...
. .
. . .
.... .. ..
.... .. ..
.. ... .....
... .... ......
........... v ........... ....2
..... ................
...........
............. ..
.... .. ..
.. .
.... ... ...
.. ...
.. ... v ....
.... 3 ..... v 1 ........
.... .... ...
... ...
.. ...
.. ... v .... ● .
.... 3 ................ ...... .................. 1....... v .... .... ..
... ...
..
...
...
...
.........
...... ...............
...
... ..
...
..
...
...
... ● ...............
..
.................
... .... .... ............................
..
.. ●...............
.........
. ..... .... ...
... ..
...
.. ... ..... .. .. ... .. .. ..
..
..
...
... .
... ..
..
..
..
... ....
..... x ..
.. .
... .. ... ..
. .
. ..
..
v...
4 ... ............. v
...............
... 8 ...
.. v● ... ................ .... .... .... ..............
4... .... ●v
................
8 ...
..
..
.. ...
... ...
...
..
..
..
..
.. ...
... ..
... . y z .. .
..
. ..........
. ..
..
. ...
.
. ... . ..
..
..
... . ... ... ... ... ... .. ............ ..... ............. ... ...
... ...
.... ....................
..
.. .. .
.........
....
..........
. . . ..
...
.... .. .
. ..
.. ....
..
● . .... ..
. .
..
. ●
............
... ........... ...
... ...
.
... ...
... .... .......
.
v .......
v
5 ..................................... 7 .......... ..... ....
....
. . .. ..
... .... ....... v . ...... ......... . .........
●
....
.
. . ..
v
. ....
5 .............................................. 7 .......... ..... .....
. .
.. ... .....
.. .. .....
... ......... .....................
.......... v..
...........
.......
6 ................................................ .....
.. .. . .. ... .....
.. .. .....
... ......... .....................
..........
........... v.. .......
6 ................................................ .....
.. .. .

.. .. ...
.. ........... ........ . . .. ..
. .. ........... ........
.. . .. .
● ..
................................................................................................................................................................................. ● ● .
................................................................................................................................................................................. ●
B C B C
Fig. 6.4 No edge of G lies outside the quadrilateral ABCD
128 6 The Problem of Thirteen Balls

A.●...................... ●D
......
............
A.●...................... ●D
......
............
............................... ........... ... ............................... ........... ...
... .... ...... ................ .......... ..... .... ... .... ...... ................ .......... ..... ....
........... ...........
... ... .....
.. ... ......
........... v
........... 2 ................ .
.. .... . .. .
.
.. ..
. ... ... .....
.. ... ......
........... v
........... 2 ................ ....
...... .
.
.. ..
.
.. ...
.. ... v ....
.... 3
.........
...................
......
.
.
... ..
.. ...
.
.. ...
.. ... v .... ●
.... 3 ............................
..................
...................
.
.
... ..
.. ...
.
.. ... ....... .
...
..
...
... .......
...
... ...
..
...
..
...
... ● ..............
.. . .... ...
... ...
..
..
..
...
... ....
.
..
..
..
..
...
... ...
... ... ........ ..
... .... ... ........................
..
x ....
....
.
.......... ...... ... ..... ..
..
... ... ... ... ... .................... ...
v ..
.. 4...
........
..
.. .
........
..... v
... 8 ...
.
v● ..
..
........... .... .... .... ...........
4... .... ...
..
.. .. ..... ●v
.... ...
. ... 8 ...
...
..

...
..
.....
...
...
..
... ...
.. y z
..... .... ... .....................
... .. ........ . ....... .
..
.
. .
. ...
...
.
...
.. . ... .. .. . ........ ....... ... ..
..... .. . . ... ...
..
.. . .
...
.......... ...
............ ...
...
..
.
..
.. ....
. ● .....................
.
.. ............ .. ............. ...... . ● .
....
....
..
..
...
...
..
..
... .....
.. ...
.. .. .......
..v.... .....
5 ..
.... .......
.................... v 7
....
....
.....
... ...
... ..
... ....
.. ...
.. .. .......
v.. . . ..
.. 5
.
● .
.....................
..... .
.................... v 7 ....
....
... ...
... ..
.. ... .....
... .... ..... .....................
...........
........v
.... ....... . ... .. .
6 ................................................ ..... .. ... .....
... .... ..... .....................
........... v
............ ....... .... .. .
6 ................................................ .....
.......................... ........... ........ .......................... ........... ........
. .
● ............................................................................................................................................................................ ● ● ............................................................................................................................................................................ ●
B C B C
Fig. 6.5 .deg D = 4

And similarly, y is adjacent to .v3 and .v5 ; z is adjacent to .v5 and .v7 . Note that
from .v3 two more edges must go out other than the edges .v3 A, v3 v2 , v3 v4 . Hence
x and y must be different by condition (ii). By a similar reasoning, y and z
must be different. Since the degree of every vertex is at most 5, we can now
deduce that .A, B, C, D, v1 , v2 , v3 , v4 , v5 , v6 , v7 , x, y, z are all different. This
contradicts condition (i).
Case (2): The vertex D has degree 4 (see Fig. 6.5).
In this case, .v2 is adjacent to .v3 , v8 . Similarly to the case (1), from each
of .v2 , v4 , v6 , one more edge emanates. Let .x, y, z be the terminal vertices
of theses edges emanating from .v2 , v4 , v6 . Then we can see that 14 ver-
tices .A, B, C, D, v2 , v3 , . . . , v8 , x, y, z are all different, which contradicts the
assumption (i) of the theorem.
⨆
⨅

6.4 The Problem

How many unit balls can touch one and the same unit ball simultaneously? (We
assume unit balls cannot overlap each other.) When two unit balls touch the unit
sphere .S, the spherical distance between the two contact points is clearly greater
than or equal to .π/3, see Fig. 6.6. Conversely, if .P , Q are points on the unit sphere
.S such that the spherical distance between them is at least .π/3, then mutually non-

overlapping two unit balls can touch .S at P and Q. So, the problem is equivalent to
the following one: How many points can be put on .S so that the spherical distance
between any pair is at least .π/3?
Consider a regular icosahedron inscribed in .S. By projecting the edges of the
icosahedron onto .S by the central projection from the center of .S, the sphere .S is
divided into 20 equilateral spherical triangles with area .4π/20 ≈ 0.628. The area of
6.5 Solution 129

Fig. 6.6 ./ P OQ ≥ π/3

an equilateral spherical triangle with edge-length .π/3 is computed as follows: By

Euler’s formula for the spherical excess, we have

1 + 3/2 5
. cos ε = √ = √ ,
4 · 3 3/8 3 3

and .ε = cos−1 5
√ ≈ 0.2756. Hence
3 3

. the area of Δ̆(π/3, π/3, π/3) = 2ε ≈ 0.551.

Since .4π/20 ≈ 0.628 > 0.551, we see that the spherical distance between any two
vertices of the regular icosahedron is greater than .π/3. Since an icosahedron has 12
vertices, we can deduce that at least 12 unit balls can simultaneously touch one and
the same unit ball.

6.5 Solution

Theorem 6.1 The maximum number of unit balls that can simultaneously touch
one and the same unit ball is twelve.
This section is dedicated to the proof of this theorem. Let n be the maximum
number of unit balls that can simultaneously touch one and the same unit ball. As
shown in the previous section, we have n ≥ 12.
Suppose that n unit balls simultaneously touch a unit ball with boundary S. Then
the spherical distance between any two contact points on S is at least

.a := π/3 ≈ 1.047.

(Hereafter, the symbols a, b, c are used to stand for special constants.) Let X denote
the set of n contact points on S. Then X does not lie in a hemisphere. For otherwise,
another unit ball can touch S. Hence, the convex hull conv(X ) of X is a convex
polyhedron containing the center of S in its interior. For any facet F of conv(X ), the
circumscribed spherical cap of F has spherical radius less than π/3. For otherwise,
130 6 The Problem of Thirteen Balls

another unit ball can touch at the center of the spherical cap of F . Triangulate those
facets of conv(X ) that have more than 3 edges by adding suitable diagonals. Then it
follows easily from Euler’s formula that the number of triangles is equal to 2n − 4.
Project the triangulation of the surface of conv(X ) onto S by the central projection
from the center of S. Then S is triangulated into 2n − 4 spherical triangles. In the
following, we are going to show that if n ≥ 13, then
[A] at most one edge of the above triangulation of S has length ≥ cos−1 17 ,
[B] the planar graph obtained from the triangulation of S by eliminating the edges
of length ≥ cos−1 (1/7) has maximum degree ≤ 5, but
[C] no such planar graph exists.
Remark 6.2 The lower bound cos−1 (1/7) in [B] is chosen to guarantee the degree
condition.
Cut S along the edges of the spherical triangles in the triangulation of S into 2n−4
pieces. Let D denote the collection of these 2n − 4 pieces (spherical triangles). Note
that for any triangle in D,
(∗) the length of the shortest edge is at least a = π/3, and the radius of the circumscribed
spherical cap is less than a.

Lemma 6.5 The length of the second longest edge of each triangle in D is less than
√
c := cos−1 ( 1−4 6 ) ≈ 1.942.
.

Proof Among the spherical triangles satisfying (∗), the supremum s of the length
of the second longest edge is attained by an isosceles triangle with edge lengths
s, s, π/3 inscribed in a spherical cap of radius π/3, see Fig. 6.7.
Let ϕ be the interior angle of an equilateral triangle with edge length a = π/3.
By the spherical cosine law, we have

cos a − cos2 a 1
. cos ϕ = = .
sin2 a 3

Fig. 6.7 An isosceles

triangle with base a = π/3

a
s s
a a
a
6.5 Solution 131

Hence

. cos s = cos2 a + sin2 a cos(π − ϕ/2),

√
from which we have s = cos−1 ( 1−4 6 ). ⨆
⨅
Put

b = cos−1 (1/7) ≈ 1.427.

.

By the spherical cosine law, the base angle of the isosceles triangle with edge lengths
a, a, b is computed as π/3.
Lemma 6.6 Let ABCD be a convex quadrilateral each of whose edges has length
at least a = π/3. Suppose that AD = b and b < BD < c. Then AC > b.
Proof If the lengths of three edges AB, BC, CD are not all equal to a = π/3,
then, with keeping the lengths of CD, DA, BD, shorten the edges AB, BC to the
length a. Then, by the spherical cosine law, the angles / ADB, / BDC decrease,
and hence / ADC decreases. Therefore AC becomes shorter. Next, with keeping
BC = a and the lengths of BD, shorten CD to length a. Then / DBC (and hence
/ ABC) decreases, and AC decreases. Hence it is sufficient to prove the lemma
in the case AB = BC = CD = a. In a spherical triangle Δ̆(x, y, z), denote by
f (x, y, z) the cosine of the angle opposite to the edge of length z, that is,

cos z − cos x cos y

f (x, y, z) =
. .
sin x sin y

Then fz (x, y, z) < 0 and

cos y − cos x cos z

fx (x, y, z) =
. .
sin2 x sin y

Put BD = x. Then cos / CBD = f (x, a, a), and cos / ABD = f (x, a, b). Since
fx (x, a, a) > 0, fx (x, a, b) > 0, we have f (x, a, a) < f (c, a, a) ≈ 0.8439 and
f (x, a, b) < f (c, a, b) ≈ 0.4016. Then

./ ABC = / ABD + / CBD > cos−1 f (c, a, b) + cos−1 f (c, a, a) ≈ 1.7238,

and hence
1 3
. cos AC < + cos / ABC ≈ 0.1357.
4 4

Therefore, AC > cos−1 0.1357 ≈ 1.437 > b. ⨆

⨅
132 6 The Problem of Thirteen Balls

Now we return to the collection D of cut-out spherical triangles. If there are

edges longer than b in the edges of spherical triangles in D, then eliminate them by
repeating the following procedure 1, 2, 3.
1. Find the longest edge e and take out from D the pair of spherical triangles that
have the edge e in common.
2. Make a spherical quadrilateral by attaching these two spherical triangles along
the edge e.
3. Deform the quadrilateral with keeping its edge-lengths so that the length of a
proper diagonal becomes b. Then cut the quadrilateral along the diagonal of
length b into two spherical triangles, and paint these triangles red, and return
them to D.
Comments
(i) If neither of a pair taken out from D is red, then e itself is the proper diagonal of
the quadrilateral obtained by attaching the pair of triangles. (This can be seen
as follows: If a triangle is not red, then its three vertices must have been the
vertices of a facet of conv(X ). Since the interior of the circumscribed spherical
cap of a facet of conv(X ) cannot contain vertices of conv(X ), e itself must be
a proper diagonal of the quadrilateral.) Thus, in this case, the proper diagonal
e is shortened to length b.
(ii) In every red triangle, all edges are shorter than c by Lemma 6.5. Hence, if
one of a pair of triangles taken out from D is red, then e is shorter than c,
and the proper diagonal of the quadrilateral (maybe not e) is longer than b by
Lemma 6.6. The proper diagonal is also shortened then.
(iii) Thus, proper diagonals are always shortened by the above procedure, and
the sum of the areas of the triangles in D decreases by the lemma of proper
diagonals.
(iv) The number of edges with length ≥ b is preserved.

If all edge-lengths of the triangles in D become less than or equal to b, then

shorten those edges with length shorter than b to the length a. By the major triangle
lemma, the sum of areas of the triangles in D also decreases in this new change.
Thus, finally every triangle in D falls into one of the four types of triangles with
edge-lengths

(a, a, a), (a, a, b), (a, b, b), (b, b, b),

.

respectively. Let

.δ, α, β, γ

denote the area of these four types of triangles, respectively. Applying Euler’s
formula for spherical excess, we can compute their values as

δ ≈ 0.5513, α ≈ 0.667, β ≈ 0.8923, γ ≈ 1.1948.

.
6.5 Solution 133

Since the sum of these 2n − 4 triangles in D is less than the surface area 4π of S,
we have

2n − 4 < 4π/δ ≈ 22.8.

.

Hence n ≤ 13.
Suppose n = 13. Then there are 22 spherical triangles of the above four
types in D. Let nδ , nα , nβ , nγ denote the number of triangles in the types
(a, a, a), (a, a, b), (a, b, b), (b, b, b), respectively. Then

12.566 ≈ 4π > (22 − nα − nβ − nγ )δ + nα α + nβ β + nγ γ

.

= 22δ + nα (α − δ) + nβ (β − δ) + nγ (γ − δ)
≈ 0.5513 × 22 + 0.1157 nα + 0.341 nβ + 0.6435 nγ ,

namely,

.0.438 > 0.1157 nα + 0.341 nβ + 0.6435 nγ .

Therefore nγ = 0. If there is a triangle Δ̆ of area β in D, then there must be at least

two other triangles each of which shares an edge of length b with Δ̆. Then the value
of the right-hand side in the above inequality becomes at least 0.1157 × 2 + 0.314 =
0.5724, which is impossible. Hence nβ = 0. Since the number of edges with length
b is even, nα is also even. Hence nα = 0 or 2. Therefore, only the following two
possibilities remain: (1) D consists of 22 triangles of area δ, or (2) D consists of 20
triangles of area δ and 2 triangles of area α. This implies that in the triangulation
of S with vertex set X , at most one edge has length greater than of equal to b. This
proves [A].
Let G be the graph obtained from the triangulation of S by removing the edge
of length ≥ b. By the next lemma, for every vertex of G, the angle determined by
two neighboring edges emanating from the vertex is greater than π/3. Hence the
maximum degree of G is at most 5, which proves [B].
Lemma 6.7 If a spherical triangle Δ̆ABC on S satisfies

a ≤ AB ≤ BC < b, a ≤ AC,
.

then / B > π/3.

Proof Let x = AB, y = BC, z = CA and f (x, y, z) = cos / B. By the spherical
cosine law cos z = cos x cos y + sin x sin y cos / B, we have

cos z − cos x cos y

f (x, y, z) = cos / B =
. .
sin x sin y
134 6 The Problem of Thirteen Balls

Since
− sin z
fz (x, y, z) =
. < 0,
sin x sin y

we have f (x, y, z) ≤ f (x, y, a) for a ≤ x ≤ y ≤ b, a ≤ z < π . And since

cos x − cos y cos a

fy (x, y, a) =
. >0
sin2 y sin x

for a < x ≤ y < b, we have f (x, y, a) < f (x, b, a). Finally, from

−2 + 7 cos x
fx (x, b, a) =
. √
8 3 sin2 x
−4 cos x + 7(1 + cos2 x)
fxx (x, b, a) = √ ,
3 2 sin3 x

we have fxx (x, b, a) > 0 in a ≤ x ≤ b, and hence f (x, b, a) takes its maximum
value either at x = a or at x = b. Since f (a, b, a) = 1/2 > 47/96 = f (b, b, a),
we have f (x, y, z) < f (a, b, a) = 1/2. Thus, cos / B < 1/2, that is, / B > π/3.
⨆
⨅
If there is no edge of length ≥ b in the triangulation of S, then G has (3×22)/2 =
33 edges. Hence the average degree of the vertices is 66/13, which is greater than
5, a contradiction. So there must be an edge of length ≥ b in the triangulation of
S. Then G has 20 triangular faces and one quadrilateral face. If XY Z be a 3-cycle
of G, then since the lengths of XY, Y Z, ZX are all less than b, the area [XY Z] is
less than γ . If Δ̆XY Z contains a vertex of X in its interior, then Δ̆XY Z contains at
least three triangles in the triangulation of S, and hence the area [XY Z] is at least
3δ ≈ 1.6539. This is greater than γ ≈ 1.1948, and impossible. Thus, every 3-cycle
of G is a boundary cycle of a triangular face.
Thus, we derived that if n = 13, then there is a graph G on the sphere such that
it has 13 vertices, one quadrilateral face, and 20 triangular face, every 3-cycle is the
boundary of a triangular face, and maximum degree is at most 5.
Let P ∈ S be a point inside the quadrilateral face of G, Then, by the stereographic
projection ϕ from P , G is mapped to a plane graph ϕ(G) on the plane tangent
to S at P ∗ , the antipodal point of P . This graph ϕ(G) is a triangulation of a
quadrilateral into 20 triangles, with 13 vertices and having no non-facial 3-cycle.
And the maximum degree of ϕ(Ǧ) is less than 6, which contradicts Lemma 6.4.
This proves [C].
Remark 6.3 The first rigorous proof of the fact that the maximum number of unit
spheres that can touch one and the same unit sphere is 12 was obtained by K. Schütte
and B. L. van der Waerden [716]. There appeared many proofs of this problem, e.g.,
in [33, 106, 145, 499, 523, 588]. The above proof is based on [520].
6.7 Notes 135

6.6 Exercises

6.1 Prove that if the shortest edge of a spherical triangle T on the unit sphere S
has length at least π/2, then T is a major triangle.
6.2∗ Prove that there is no quadrangulation with nine vertices.
6.3 Prove that the maximum number of balls of radius 1.1 that can simultaneously
touch one and the same unit ball is twelve.
6.4∗ Let T be a spherical triangle on the unit sphere S with edge lengths a, b, c (a ≤
b ≤ c). If the spherical radius of the circumscribed cap of T is less than a,
then the area of T is at least the area of Δ̆(a, a, a). This result is known as
Fejes Tóth’s Lemma. (For a proof of this lemma, see Exercise 7.3∗ .) Using
this lemma prove the following: The maximum number of balls of radius 2
that can simultaneously touch one and the same unit ball is six.
6.5 (i) Check that the distance between any two points in the eight points
√ √ √ √ √
− 7
(
.
5
7
, 0, ± 3 2
5 ), ( 5
7
, ± 3 2
5 , 0), ( 5 , ±5, ±5)
3 3

on the unit sphere is at least 6/5.

(ii) Using Fejes Tóth’s Lemma prove that the maximum number of balls of
radius 3/2 that can simultaneously touch one and the same unit ball is
eight.
6.6 (i) Check that the distance between any two points in the nine points
√ √ √
(0, 0, 1), (±
. √83 , ± 4 , √
4
), (± √83 , 0, − √
8
), (0, ± 23 , − 12 )
7 3 7 7 3 7 3 7 3

on the unit sphere S is greater than 8/7.

(ii) Using Fejes Tóth’s Lemma, prove that the maximum number of balls of
radius 4/3 that can simultaneously touch one and the same unit ball is
nine.

6.7 Notes

We restrict these notes to only one point, namely the problem of thirteen balls
itself. The reason is simple: this problem created sufficiently many activities and
interesting further discussions.

6.7.1 Notes on the Problem of Thirteen Balls

We want to show parts of the network that the 13-spheres problem was able to
create. As already its discussion in the basic monograph [262] shows, this structure
136 6 The Problem of Thirteen Balls

of notions and results clearly belongs to the field of discrete geometry. Due to the
large variety of related papers that exist, we shall lay emphasize mainly (but not
only) on survey-like papers.

The Problem Itself and Its Proofs

In 1694, Newton and Gregory discussed the problem of thirteen balls. The first
complete, rigorous solution to this problem is due to Schütte and van der Waerden
[716]. They confirm Newton’s opinion that no more than twelve unit balls can do
this, and they use an interesting mixture of trigonometric and topological ideas.
Shortly after this success, Leech [499] presented an elegant new proof of that result.
His approach is based on the confirmation of the non-existence of a polyhedron with
13 vertices (one of valence 4, and twelve of valence 5) and 21 facets (20 triangular
ones and 1 quadrangle), and this proof is usually chosen when the result and its
proofs are presented somewhere (meanwhile there are several proofs of the result).
E.g., in the books [11] of Aigner and Ziegler as well as [839] due to Zong one can
find variants of Leech’s proof. And Maehara [520] nicely reproves this result with
the help of the isoperimetric theorem for spherical polygons (which is also elegantly
derived).
Another approach (using spherical Delaunay triangulations) is due to K.
Böröczky [106], see also Sect. 4.4 of the book [107]. And also Hsiang [405, Chapter
2] gives a new proof of Newton’s opinion. Anstreicher [33] presents a proof based
on linear programming bounds and (in another way than above) on properties
of spherical Delaunay triangulations which is not (as in former proofs) directly
depending on the dimension. Following the arguments of Leech, but using also a
tool from spherical geometry, Maehara [523] succeeded in giving an elementary
proof, suitable even for undergraduate students. And a very short proof, using
linear programming methods, is also given by Glazyrin [304]. Chapter 1 of Zong’s
booklet [839] gives a concise and elegantly written overview to the whole subject.

First Generalizations: Related Book Chapters and Surveys

We continue with the concept of optimal packings, where the following steps of
generalizations are already included: to replace Euclidean spheres by convex bodies,
to consider the global situation, and to look also at higher dimensions. Due to the
richness of this framework, we discuss mainly surveys, sometimes also given in
form of book chapters.
Coxeter [184] gives a possible upper bound for the number of congruent non-
overlapping (.n − 1)-spheres in n-space which can touch a given one of the same
radius. He comprehensively discusses also the history of the problem (from the 17th
till to the middle of the twentieth century, with an impressive list of references). The
contributions of famous mathematicians like Kepler, Newton and Gregory, Schläfli
6.7 Notes 137

(who was the first to bring spherical trigonometry up to n dimensions), Poincaré,

Sommerville, Guinand, and Minkowski are comprehensively presented.
In view of generalizations, we define now Newton numbers and related concepts
referring to convex bodies. For each n-dimensional convex body K (i.e., for each
compact, convex set with interior points in n-space), the maximum number of
neighbors of a member of any packing with congruent copies of K is called the
Newton number .N(K) of K. Also the name kissing number is often used. Thus,
the dispute between Newton and Gregory refers to the subcase that K is a three-
dimensional ball in 3-space. If “congruent” is replaced by “translative”, we speak
about the Hadwiger number .H (K) of K (or translative kissing number). The latter
notion can be conveniently used also for normed or even gauge spaces, identifying
K with the respective unit ball (in the non-symmetric case, i.e., for gauge spaces,
inducing even an asymmetric distance function). E.g., from this viewpoint Hadwiger
numbers are discussed by Swanepoel in the survey [765].
Surveys on Newton and Hadwiger numbers of convex bodies are Chapters 13.1
and 13.2 of the monograph [262], Section 5.3 of the handbook article [261], and
Chapter 4 of Zong’s book [837]. Also K. Bezdek surveys in [85] Hadwiger and
Newton numbers of convex bodies, but also variants of them, called one-sided
Hadwiger and kissing numbers.
Sloane [741] gives a nice exposition on the historical development of sphere
packing problems and related methods. He excellently discusses kissing numbers
and considers related lattice and non-lattice packings. Connections to other fields
are shown, too. The article [741] appeared shortly before the first announcement
of Hales regarding the Kepler conjecture, and the reader is referred there to the
forthcoming third edition of the book [176] of Conway and Sloane, containing
(with respect to the second edition) 800 new referencces. This important book
contains much material on problems treating sphere packings. In particular, also
kissing numbers are discussed (see in [175] Section 2.1 of Chapter 1 for the 13-
spheres problem, Section 2.2 of Chapter 1 for other dimensions, and Chap. 13
for bounds and numerical results). Various results where lattice arrangements are
optimal can be found there, too. Interesting connections to related fields (like finite
groups, quadratic forms, error-correcting codes, number theory etc.) are discussed
and successfully applied, and also applications outside of mathematics (chemistry,
superstring theory and so on) are studied.
Zong [838] gives a nice survey on Hadwiger numbers of n-dimensional convex
bodies K, and the subcase where the translates touching K are taken from a lattice
packing (thus yielding lattice kissing numbers) is also taken care of. In some
cases also proofs are presented. The same author surveys results on the relations
between kissing numbers, covering numbers (i.e., the smallest numbers of translates
of the interior of K covering K) and the relatively new related concept of blocking
numbers in [842]. (Note that the blocking number of a convex body K is the smallest
number of non-overlapping translates of K all of which touch K at its boundary such
that their union makes it impossible that any further non-overlapping translate of K
can touch K.) It turns out that blocking numbers form a “bridge” between the other
138 6 The Problem of Thirteen Balls

two notions, and such relations like also further results are excellently presented in
[842].
Coming back to the Newton-Gregory problem, we refer to the surveys of
Casselman [145] as well as Pfender and Ziegler [635]. They present related
historical facts, from the classical beginnings till to extended problems, e.g. also
referring to 4-space. Based on a method initiated by Delsarte in the early seventies
(concerning inequalities for the distance distributions of kissing configurations; see
the fundamental work [211] of Delsarte et al. on spherical codes and designs),
interesting results regarding dimensions eight and twenty-four are discussed. We
note that Delsarte’s method strongly influenced further developments; see again
[635]. So the authors discuss also Musin’s excellent solution of the four-dimensional
case (see [589]), based on a modification of Delsarte’s method, and the paper [169]
of Cohn and Elkies, opening new possibilities to deal with optimal sphere packings
more efficiently. Various further results are offered, also treating optimal lattice
packings of spheres. Basic ideas and tools are nicely explained.
In [337] Gruber gives an interesting overview regarding lattice packing and
covering problems, related mainly to the geometry of numbers and referring to n-
dimensional balls and convex bodies. He offers more recent results on uniqueness of
densest lattice packings and the thinnest lattice coverings of a generic convex body.
Estimates and bounds for the kissing number of densest lattice packings of generic
convex bodies are characterized.
The concept of spherical codes (cf. again Delsarte et al. [211]) is more general
than that of kissing numbers. (Recall that a spherical code with parameters .(n, m, t)
is a set of m points on the unit hypersphere in n dimensions for which the dot product
of unit vectors from the origin to any two points is less than or equal to t. E.g., the
kissing number problem and the Tammes problem can be formulated as subcases
in terms of this concept.) Keeping this more general level of spherical codes,
Boyvalenkov et al. [120] give a comprehensive survey on results and methods
referring to kissing numbers.
Liberti [507] gives a relatively short review on bounds for kissing numbers
of n-dimensional unit balls based on mathematical programming, forming also
a cornerstone for the theory of spherical codes. This might be the right place
to mention also a few important research papers, in which (motivated also by
Delsarte’s approach) methods from semidefinite programming are used to get
upper bounds for kissing numbers in higher dimensions (different to the known
cases .n = 1, 2, 3, 4, 8, 24). Bachoc and Vallentin [51] consider .A(n, α), i.e.,
the maximum number of points on the unit (.n − 1)-sphere with minimal angular
distance .α; the kissing number problem is then equivalent to finding .A(n, π/3).
A “semidefinite program” is created which yields, for some dimensions n, better
upper bounds for .A(n, α) than known ones. Better software, faster computers and
optimized use of computation time enable Mittelmann and Vallentin [579] to get
further improvements in several cases. The results of Kuklin [480] go into the same
direction. And results on lattice kissing numbers (where the packed spheres are
centered at the points of an n-dimensional lattice and have midpoints at minimal
distance from the center of the kissed sphere) are obtained by Skoruppa [740].
6.7 Notes 139

Cohn [167] gives, based on five lectures (from 2014), an excellent overview
to several aspects of sphere packings, e.g. including lattices, linear programming
bounds, kissing numbers and spherical codes. Unfortunately, the later spectacular
results of Viazovska on sphere packings in dimension 8 (see [795]) and of Viazovska
with colleagues (cf. [170]) in dimension 24 are not contained in this very readable
material. (Note that Viazovska got the Fields medal for her results in dimension
8.) However, Cohn’s later survey [168] contains these deep results and is highly
recommendable. This overview is completed by Viazovska’s survey [796], where
the so-called sharp sphere packings are discussed, whose density attains the linear
programming upper bound. In [796] she mainly concentrates on dimensions 8
and 24, where her work is located; in these dimensions no unit ball packing has
larger density than the 8-dimensional lattice packing and the Leech lattice packing,
respectively. A new lower bound on the density of packings of congruent spheres is
given by Campos et al. in [140].
There are also problem books discussing the issue of 13 spheres and related
topics. The first one is [190] by Croft, Falconer and Guy, where problems closely
related to our discussion here (i.e., the Kepler conjecture and kissing numbers) are
summarized in Sections D10 and D12. And the monograph [125] of Brass, Moser
and Pach treats such questions in Sections 1.8, 1.9, and 2.4.

Further Extensions and Variations

We start with a few older (but refreshing) extensions and modifications, and we
continue then with more recent contributions.
L. Fejes Tóth and Heppes consider the following extension of the problem of
thirteen spheres in n-space: Let A be a unit ball in n-space, let B be a family of
unit balls all of them touching A, and C be a family of unit balls, each of them
touching a member of B. Let .Q = B ∪ C, and denote by .qn the maximal cardinality
that such a set Q can have. The authors prove that .q2 = 18, 56 ≤ q3 ≤ 63, and
.168 ≤ q4 ≤ 232.

Inspired by the Gregory-Newton problem, Propp [647] discusses the following

problem: what permutations can be obtained if one marks the 12 touching spheres
and lets them change their positions at will? Using analytic and group theoretic
methods, interesting results (in some sense related to Rubik’s famous cube) are
obtained. E.g., the arrangement of 12 such balls, with midpoints forming the vertex
set of a regular icosahedron, can be deformed in an arrangement forming the vertex
set of an Archimedean cuboctahedron. Also it is conjectured that each general
position of 12 equal balls around a congruent one can be generated in such a
way from any other arrangement. Grünbaum proves in [338] that the Hadwiger
number of a planar convex body is realized by a lattice packing which is not true
for 9-dimensional balls. Zong [836] constructs a counterexample for any dimension
.n ≥ 3. It is some suitable truncation of the cube. G. Kuperberg and Schramm [481]

consider finite packings P of p non-congruent balls in 3-space. For m giving

the number of contact points of the given balls, the authors call .k(P ) = 2m/p
140 6 The Problem of Thirteen Balls

the average kissing number of P . If k denotes the supremum of all numbers .k(P )
over all such finite packings, the authors show that .12.566 ≤ k ≤ 14.928, which
is particularly interesting regarding the lower bound (since for congruent balls this
number is 12). Using semidefinite programming, in [223] by Dostert et al. new upper
bounds on the average kissing number for certain dimensions are derived. And in
[538] kissing numbers of non-congruent balls are investigated, namely regarding
the smallest number of non-overlapping unit balls which can touch a ball of radius
.r > 0 in n-space.

Dostert and Kolpakov [222] find new upper and lower bounds for kissing
numbers of congruent spheres in spherical and hyperbolic n-space.
Clearly, the 13-spheres problem is essential for the local situation in the sphere
packing problem, asking for the densest packing of infinitely many congruent
spheres in 3-space. The Kepler conjecture (or the so-called sphere packing problem)
says that the densest packing of congruent spheres √ in 3-space is reached by the
face-centered cubic lattice packing with density .π/ 18. With the help of Ferguson,
Hales published (in a series of papers) a proof in 2005 and 2006. A revision followed
in 2010, and all related papers, including this revision and further supporting
material, can be found in the book [374] of Hales and Ferguson. Around 2003 Hales
started with a project named FLYSPECK to create an automated (computerized)
formal verification of his proof. In the strongly related book publication [372] Hales
deals with a proof of the Kepler conjecture written in a formal logical system
and being logically independent of the former proof. Presenting, more precisely, a
blueprint version of this formal proof in [372], this book brings the above mentioned
project closer to the reader. Not requiring knowledge on details from parts of the
previous approaches, it is self-contained; see also the review [485] of Lagarias
on it. The officially published completion of the FLYSPECK project is confirmed
in [376] (2017). For comprehensive overviews to the Kepler conjecture and its
history we refer to the book [768] of Szpiro and to Sections 7.1 and 7.2 as well as
subsection 8.7.1 of the monograph [262] of L. and G. Fejes Tóth and W. Kuperberg.
Two further nice expositions on the Kepler conjecture and its history are due to Henk
and Ziegler [392] as well as to Joswig [434].

Related Problems

Böröczky and Szabó [111] study the strongly related Tammes problem of arranging
k points on the unit 2-sphere so that the minimum distance between any two of
them is maximized; see also the classical book [257] of L. Fejes Tóth and its
strong extension [262, Section 8.5] for related references. Denoting this maximum
by .ak , in [111] it is shown that .a13 < 1.02746114. This implies the following: If
.B1 , . . . , B14 are 14 different unit balls in 3-space, forming a packing so that each

of the balls .B2 , . . . , B13 touches .B1 , then the distance between the centres of .B1
and .B14 is at least .2.20527 . . . . Also Musin and Tarasov [592] treat the Tammes
problem (placing 13 points suitably on the unit sphere) and its importance for the
thirteen-spheres problem. Musin and Tarasov solve this problem by showing that
6.7 Notes 141

a certain arrangement of 13 points is best possible, namely at the minimal angular

distance of ca. 57.1367 (an equivalent formulation of the thirteen-spheres problem
asks whether the respective distance is smaller than 60). Depending on this approach
to the Tammes problem, Böröczky and Szabó [112] present a geometric proof of the
related Fejes Tóth conjecture that the balls in any 12-neighbour packing of unit
balls in 3-space are positioned in hexagonal layers, as in a densest lattice packing.
The Tammes problem and similar problems around circle packing types are nicely
discussed in Berger’s refreshing survey [76] which treats also various related things.
The dodecahedral conjecture, going back to L. Fejes Tóth [256], refers to the
Voronoi cells of all ball centers from packings of unit balls in 3-space. It says
that in any such packing every Voronoi cell has volume at least that of a Platonic
dodecahedron having a unit sphere as insphere, where equality holds exactly when
the Voronoi cell is such a Platonic solid. It was positively solved in 1998, and the
paper [375] by Hales and McLaughlin contains a strongly revised version of the
original proof which, however, is essentially the same as the original one. The paper
[375] contains all basic references to the problem. The dodecahedral conjecture is
not directly related to the Kepler conjecture, but the proofs of both share various
ideas and methods.
And in the paper [373] Hales gives outlines of proofs of two further problems
on packings of unit spheres in 3-space: the strong dodecahedral conjecture, and
the conjecture of L. Fejes Tóth on sphere packings with kissing number 12. The
strong dodecahedral conjecture is due to K. Bezdek [84] and says that in such a
sphere packing the surface area of any Voronoi cell is at least as large as the surface
area of the Platonic dodecahedron with inradius 1. The other, already mentioned
conjecture goes back to the paper [258] of L. Fejes Tóth saying that if each sphere
in a packing of unit spheres in 3-space is touched precisely by twelve other spheres,
then this packing consists of hexagonal layers. Hales [373] carefully explains the
basic geometric ideas and methods as well as their combinations with the computer
calculation parts, to give a clear picture of the computer assisted proofs of both
conjectures.
Musin presents in [591] five essays on topics close to L. Fejes Tóth, treating
kissing numbers and related topics. Among them are the Tammes problem and his
bound on circle packings, the maximum kissing number of packings on the sphere,
and one-sided kissing numbers.
Chapter 7
Spherical Geometry III

The art of doing mathematics consists in finding that special

case which contains all the germs of generality.
— David Hilbert

A spherical ellipse is similarly defined as an ellipse in the plane. It is the locus of

a point X on .S for which the sum of the distances from two fixed points to X is
constant. A regular spherical polygon is a cyclic polygon whose edge-lengths are
all equal. Using spherical ellipses in this chapter, we show that among cyclic n-
gons inscribed in a cap of fixed spherical radius, regular n-gons have the maximum
perimeter, and among the n-gons on .S with fixed perimeter, regular n-gons have the
maximum area.

7.1 Spherical Ellipses

Let .S be the unit sphere centered at the origin O of .R3 . Let .A, B ∈ S be a pair of
points such that .AB = 2φ < π. For a given .s, φ < s < π − φ, the set

E(A, B : s) = {X ∈ S : AX + BX = 2s}
.

is called a spherical ellipse with foci .A, B and parameter s.

First, let us observe that a spherical ellipse on .S is a simple closed curve on
.S. Suppose that the midpoint M of AB is the south-pole .(0, 0, −1) of .S. Then,

each meridian (a semicircle connecting the north-pole to the south-pole) intersects

.E(A, B : s) exactly in one point. (Since .AM + BM = 2φ < 2s, and .AM +
∗
∗
BM = 2π − 2φ > 2s, each meridian intersects .E(A, B : s). If two distinct
points .X, X' ∈ E(A, B : s) lie on the same meridian, then one of .Δ̆ABX, Δ̆ABX'
contains the other. Then .AX + BX /= AX' + BX' , a contradiction.) Similarly, for
two distinct points .X, Y ∈ E(A, B : s), none of .Δ̆ABX, Δ̆ABY contains the other.

© The Author(s), under exclusive license to Springer Nature Switzerland AG 2024 143
H. Maehara, H. Martini, Circles, Spheres and Spherical Geometry,
Birkhäuser Advanced Texts Basler Lehrbücher,
https://doi.org/10.1007/978-3-031-62776-7_7
144 7 Spherical Geometry III

Fig. 7.1 None of meridians

..
.....
.Δ̆ABX, Δ̆ABY contains the .. ..
.. ..
other .. ..
... ...
. .
... .. ..
. .... ......... .... .
.. ........ ... ...............
.. .... .. ..... ...
.. .... . ... . ..
.. .......... . ..
...
..
.. X ..● Y
........ ●...
... ... ....... ...
.
..
.
... ... . ...... ..
...... .. .. ...... .
...... .... .. ...
... ............ .. ........
... .. ... ....
..... .. ..
....
....
.... ... .. .. .... ...
.... ...
.... ..
....... .. .. .... ......
....... .. .. .........
● ........ ...●
.
........... ....... ........
A .......................................................
● B
M

This implies that Y lies in the angular region (lune) of an exterior angle of .Δ̆ABX
at the vertex X, see Fig. 7.1. Therefore, for any given .ϵ-neighborhood of X, if the
longitude of Y is sufficiently close to the longitude of X, then Y is contained in the
.ϵ-neighborhood, and the difference of the latitudes of X and Y becomes less than

.ϵ. This implies that the latitude of X is a continuous function of the longitude of X.

Thus, .E(A, B : s) is a simple closed curve on .S.

A spherical ellipse .E(A, B : s) divides .S into two regions

{X ∈ S : AX + BY < 2s} and {X ∈ S : AX + BX > 2s}.

.

The former is called the interior of .E(A, B : s), and the latter is called the exterior
of .E(A, B : s). The spherical ellipse with foci .A, B and passing through C is also
denoted by .E(A, B : C). If .AC + BC = 2s, then .E(A, B : C) = E(A, B : s).
Remark 7.1 (i) The foci .A, B of .E(A, B : s) lie in the interior of the spherical
ellipse, and their antipodals .A∗ , B ∗ lie in the exterior of the spherical ellipse. (ii)
If .X ∈ E(A, B : s), then .A∗ X + B ∗ X = 2π − (AX + BX) = 2π − 2s. Hence
the spherical ellipses .E(A∗ , B ∗ : π − s) and .E(A, B : s) are identical curves, and
hence .E(A∗ , B ∗ : C) = E(A, B : C) as figures. Note that, however, the interior of
∗ ∗
.E(A , B : C) is the exterior of .E(A, B : C).

Lemma 7.1 For a spherical triangle ABC with .AC /= BC, let D be the point on
⎛

ACB such that .AD = BC, BD = AC. Then the spherical ellipse .E(A, B : C) and
.
⎛

the circular arc .ACB intersect only at .C, D, with crossing each other at .C, D, and
⎛
(

the subarc .CD of the circular arc .ACB lies, except for the two endpoints, in the
exterior of .E(A, B : C).
The proof of this lemma will be given in Sect. 7.6, Appendix.
Corollary 7.1 Let .Δ̆ABC be a spherical triangle on .S such that .AC < BC, and
⎛

let .ACB be a point such that .AC < AE, AC < BE. Then .AE + BE > AC + BC.
⨆
⨅
 7.2 Lexell’s Theorem 145

7.2 Lexell’s Theorem

⎛
For a geodesic segment AB and a circular arc .AXY on the unit sphere .S, the angle

⎛
between the tangent lines of AB and .AXY at A is denoted by .AXY .
Theorem 7.1 Let .2ε be the area of .Δ̆ABC on the unit sphere .S. Then
⎛

./ (B ∗ A, B ∗ CA∗ ) = ε, see Fig. 7.2.

Proof Let .ϕ : S → Π be the stereographic projection of .S from .A∗ onto the plane
∗
.Π tangent to .S at A. For every .X ∈ S, X /= A , .ϕ(X) is denoted by .X̄. By this

stereographic projection, any circle that passes through .A∗ is mapped to a line on
−−→
⎛

.Π . So, the arc .B CA is mapped to a half line .B¯∗ C̄. Therefore, the region enclosed
∗ ∗
⎛

by .B ∗ CA∗ and the great circular arcs .AB ∗ , AC is mapped to the planar triangle
∗
.ΔĀB̄ C̄. Since this triangle is a derived triangle .Δ̆ABC, (consulting Fig. 5.2 in

Chap. 5) we have ./ ĀB¯∗ C̄ = ε. Since the stereographic projection preserves the

⎛

magnitude of angles, ./ (B ∗ A, B ∗ CA∗ ) = ε. ⨆

⨅
⎛

For a spherical triangle .Δ̆ABC, the arc .B ∗ CA∗ is called the Lexell arc of .Δ̆ABC
with base AB.
Theorem 7.2 (Lexell) For a spherical triangle .Δ̆ABC on .S, let .H denote the
interior of the hemisphere bounded by the great circle AB and containing C. For a

⎛
point .X ∈ H, .Δ̆ABX has the same area as .Δ̆ABC if and only if X lies on .B ∗ CA∗ .

Fig. 7.2 Theorem 7.1

C

A∗ B
B ∗ ε
A
146 7 Spherical Geometry III

⎛
Proof By Theorem 7.1, the area of .Δ̆ABX is equal to .2 × / (B ∗ A, B ∗ XA∗ ). Now,

⎛
./ (B ∗ A, B ∗ XA∗ ) = / (B ∗ A, B ∗ CA∗ ) ⇔ B ∗ XA∗ = B ∗ CA∗

⎛
⇔ X ∈ B ∗ CA∗ .

⨆
⨅
Remark 7.2 There are many different proofs of this theorem, see, e.g., [257,
Ch.1, §8] (cf. analogously [262]), [616, §5], and [520, 531, 537]. A proof using
stereographic projection was first obtained by A. Simonič [734]. Recently, the
authors presented a collection of seven proofs whose main ideas go back to famous
mathematicians, see [539].
Lemma 7.2 Let .Δ̆ABC be a spherical triangle on .S such that .AC /= BC. Let
⎛

(
D ∈ A∗ CB ∗ be a point that satisfies .AD = BC, BD = AC. Then the subarc .CD
.
⎛

of .A∗ CB ∗ lies inside .E(A, B : C).

Proof By Remark 7.1(ii), .E(A∗ , B ∗ : C) = E(A, B : C) as curves, and the interior

(
of .E(A∗ , B ∗ : C) is the exterior of .E(A, B : C). By Lemma 7.1, the subarc .CD of
⎛

A∗ CB ∗ lies outside .E(A∗ , B ∗ , C). Hence .CD lies inside .E(A, B : C).
. ⨆
⨅

7.3 Equilateral Spherical Triangles

Theorem 7.3 Among the spherical triangles inscribed in a fixed cap Λ on S,

equilateral triangles have the maximum perimeter.
Proof Let Z be the center of the cap Λ. Let us show that the perimeter of the
non-equilateral spherical triangle Δ̆ABC inscribed in Λ is smaller than that of an
equilateral spherical triangle inscribed in Λ. Let AC be the shortest edge, and BC be
the longest edge of Δ̆ABC. We may consider the case that Z lies in Δ̆ABC. (If Z is
⎛

not an interior point of Δ̆ABC, then let M be the midpoint of the arc ACB. Clearly,
AM > AC, BM > AC. By Corollary 7.1, we have that AM + BM > AC + BC,
the perimeter of Δ̆ABM is greater than that of Δ̆ABC, and Z is an interior point of
Δ̆ABM. Hence we may replace Δ̆ABC by Δ̆ABM.)
⎛

Let D be the point on ACB such that BD = AC. By Corollary 7.1, the subarc
⎛
(

CD of ACB lies in the exterior of E(A, B : C) except for its two endpoints. Since
Z is an interior point of Δ̆ABC, we have / AZC < 2π/3 < / CZB. Let E be
⎛

a point on ACB such that / AZE = 2π/3. This point E lies on the subarc CD,
7.4 Spherical Polygons Inscribed in a Cap 147

and is different to C, D. Hence E is an exterior point of E(A, B : C), and we have

AE + BE > AC + BC, which means that the perimeter of Δ̆ABE is greater than
the perimeter of Δ̆ABC. If Δ̆AEB is equilateral, then we are done. If Δ̆AEB is

⎛
not equilateral, then AB /= EB. Let F be the midpoint of the arc EBA. Then the
spherical triangle Δ̆EBA is an equilateral triangle whose perimeter is longer than
that of Δ̆ABC. ⨆
⨅
Theorem 7.4 Let Δ̆(a, b, c) be a spherical triangle with a > b. Then, for every
e, a > e > b, the area of Δ̆(e, a + b − e, c) is greater than the area of Δ̆(a, b, c).
Proof Let Δ̆ABC be a spherical triangle in which the opposite edges of A, B, C
have lengths a, b, c, respectively. Let H be the hemisphere bounded by the great

⎛
circle AB and containing C. Let D be a point on ACB such that AD = BC, BD =
AC. Since AC = b < a = BC = AD and a > e > b, there is a point E on
̂ of the “half ellipse” H ∩ E(A, B : C) such that AE = e. Since
the subarc CD
BE = a + b − e, the perimeter of Δ̆ABE is equal to a + b + c. Since C and D lie
in the different sides of the semicircle AEA∗ (which divides H into two regions),
⎛
(

the subarc CD of A∗ CB ∗ intersects the semicircle AEA∗ at a point, say F . Then,

⎛

since F lies on the Lexcell arc A∗ CB ∗ , we have [ABF ] = [ABC]. Since Δ̆ABF
is contained in Δ̆ABE, we have [ABF ] < [ABE]. Therefore, [ABE] > [ABC].
Thus,

[Δ̆(e, a + b − e, c)] = [ABE] > [ABC] = [Δ̆(a, b, c)].

.

⨆
⨅
Theorem 7.5 Among the spherical triangles with fixed perimeter p, equilateral
triangles have the maximum area.
Proof Let Δ̆(a, b, c) be a spherical triangle with a + b + c = p, and suppose that
a is the maximum and b is the minimum, a > b. Put e = p/3. Then a > e > b.
By Theorem 7.4, [Δ̆(e, a + b − e, c)] > [Δ̆(a, b, c)]. If a + b − e = c, then
Δ̆(e, e, e) = Δ̆(e, a + b − e, c). If a + b − e /= c, then by applying Theorem 7.4, we
have [Δ̆(e, e, e)] > [Δ̆(e, a + b − e, c)]. Therefore, [Δ̆(e, e, e)] > [Δ̆(a, b, c)]. ⨅
⨆

7.4 Spherical Polygons Inscribed in a Cap

In a planar polygon, a vertex x is called an ear of the polygon if the line segment
connecting the two neighboring vertices of x lies in the interior of the polygon
except for its two endpoints. For example, in a pentagon ABCDE, if the diagonal
AD lies inside the polygon, then E is an ear of the polygon, see Fig. 7.3. The two
ears theorem states that if .n ≥ 4, then every n-gon contains two ears that are not
148 7 Spherical Geometry III

Fig. 7.3 The vertex E is an D ...........................................................................

ear .
...
..
..
..
..
........
..
.
.. C
... .. ..
.. .... .... ...
.. .. ... .
..
...
.. A
...
... ...
...
..
..
..
... .... ... .
.. .. ... ..
... .... ... ....
.. .. ... ..
........ ... ...
....
E .
B

adjacent. Similarly to the planar polygon case, an ear of a spherical polygon is a

vertex x such that the geodesic segment between the two neighboring vertices of x
lies in the interior of the spherical polygon, except for its two endpoints.
Lemma 7.3 Every spherical n-gon, .n ≥ 4, lying in the interior of a hemisphere
has at least two ears.
Proof Let .Γ be a spherical n-gon, .n ≥ 4, lying in the interior of the lower
hemisphere .z ≤ 0. Project it from the center O of .S into the plane that is tangent to
.S at .(0, 0, −1). Then we have a planar n-gon. It has at least two ears. By the inverse

projection these ears are mapped to the ears of .Γ . ⨆

⨅
Lemma 7.4 For every spherical polygon .Γ with perimeter less than .2π , there is a
cyclic one with the same edge-length sequence as .Γ .
Proof Let .a1 , a2 , . . . , an be the edge-lengths of .Γ , and suppose that .an is the
maximum. Let K be a circle on .S whose spherical radius is nearly .π/2. Take n
points .P1 , P2 , . . . , Pn on K so that .P1 P2 = a1 , P2 P3 = a2 , . . . , Pn−1 Pn = an−1 . If
the spherical radius of K is sufficiently close to .π/2, then we have .P1 Pn > an .
Squeeze the radius of K with keeping .P1 , P2 , . . . , Pn on it and preserving the
spherical distances .P1 P2 , P2 P3 , . . . , Pn−1 Pn until the length .P1 Pn becomes .an . ⨅ ⨆
Let us denote by .Ψ(a1 , a2 , . . . , an ) a cyclic spherical n-gon with edge-lengths
a1 , a2 , . . . , an in this cyclic order.
.

Theorem 7.6 Among the spherical n-gons with fixed edge-lengths

a1 , a2 , . . . , an , where a1 + a2 + · · · + an < 2π,

.

a cyclic n-gon has the maximum area.

Proof The proof works by induction on n. The case .n = 4 follows from
Theorem 5.8. Suppose that .n ≥ 4 and the theorem is true for n. Let .F be the set
of spherical .(n + 1)-gons .〈X〉 with vertices .X0 , X1 , . . . , Xn and .Xi Xi+1 = ai , i =
0, 1, 2, . . . , n (where the indices are taken modulo .n + 1). By Lemma 7.3, each
.〈X〉 ∈ F has at least two ears. Let

F(i) = {〈X〉 ∈ F : Xi is an ear of 〈X〉}.

.
7.4 Spherical Polygons Inscribed in a Cap 149

Then, for each .〈X〉 ∈ F(i),

the area of 〈X〉 = [Xi−1 Xi Xi+1 ] + [X0 X1 . . . Xi−1 Xi+1 . . . Xn ].

.

If we put .t = Xi−1 Xi+1 , then by the inductive assumption we have

[X0 X1 . . . Xi−1 Xi+1 . . . Xn ] ≤ Ψ(a0 , a1 , . . . , ai−2 , t, ai+1 , . . . , an ).

.

Hence

the area of 〈X〉 ≤ [Δ̆(t, ai−1 , ai+1 )]

.

+ [Ψ(a0 , a1 , . . . , ai−2 , t, ai+1 , . . . , an )], (7.1)

and equality holds only when the vertices other than .Xi lie on the same circle. The
right hand side of (7.1) is a continuous function of t, and the range of t is a closed
interval

. max{δ, |ai−1 − ai |} ≤ t ≤ min{p − ai−1 − ai , ai−1 + ai }

where .δ is the infimum of x such that there exists a cyclic n-gon with edge-lengths
.a0 , a1 , . . . , ai−2 , x, ai+1 , . . . , an , and .p = a0 + a1 + · · · + an . Hence there is a
maximum value .f (i) of the right hand side of (7.1). This maximum value is attained
by an .〈X〉 ∈ F(i) for which the vertices other than .Xi lie on the same circle. Put
.f (k) = max{f (i) : i = 0, 1, . . . , n}, and suppose that .〈Y 〉 ∈ F(k) has the area

.f (k). Then the vertices of .〈Y 〉 other than .Yk lie on the same circle. Since .〈Y 〉 has

another ear .Yj (j /= k), we have also .〈Y 〉 ∈ F(j ), and hence .f (k) ≤ f (j ). Thus
.f (k) = f (j ), and hence the vertices of .〈Y 〉 other than .Yj also lie on the same circle.

Since .n + 1 ≥ 5, we can deduce that all vertices of .〈Y 〉 lie on the same circle. Thus,
.〈Y 〉 ∈ F(i), i = 0, 1, . . . , n, and .f (0) = f (1) = · · · = f (n). Therefore the cyclic

.(n + 1)-gon has the maximum area among the .(n + 1)-gons in .F. ⨆
⨅
Lemma 7.5 For every permutation .σ of .{1, 2, . . . , n}, two spherical n-gons

Ψ(a1 , a2 , . . . , an ) and Ψ(aσ (1) , aσ (2) , . . . , aσ (n) ),

.

have the same area and the same perimeter.

Proof Let Z be the center of the cap circumscribed to .Ψ(a1 , a2 , . . . , an ). Suppose
that Z lies in the interior of .Ψ(a1 , a2 , . . . , an ). Dividing this n-gon by geodesic
segments connecting Z to the vertices of this n-gon into n spherical triangles, and
then rearranging these spherical triangles, we can obtain .Ψ(aσ (1) , aσ (2) , . . . , aσ (n) ).
Hence they have the same area and the same perimeter.
If Z does not lie in the interior of .Ψ(a1 , a2 , . . . , an ) we need to change the proof
slightly (with introducing a “negative” spherical triangle), but it would be essentially
the same. ⨆
⨅
150 7 Spherical Geometry III

7.5 Regular Spherical Polygons

Theorem 7.7 Among the spherical n-gons inscribed in a fixed cap Λ, regular n-
gons have the maximum perimeter.
Proof Let Z be the center of the cap Λ, and let Γ = A1 A2 . . . An be a non-regular
n-gon inscribed in Γ . Let us show that a regular n-gon inscribed in Λ has a perimeter
larger than the perimeter of Γ . By Lemma 7.5, we may suppose that A2 A3 is the
longest edge, and A1 A2 is the shortest edge. Thus A1 A2 < A2 A3 . We may consider
the case that Z lies in the interior of Γ . (If Z is not contained in the interior of Γ ,

⎛
then, denoting by M the midpoint of A1 A2 A3 , we have A1 M > A1 A2 , A3 M >
A1 A2 , and hence, by Corollary 7.1, we have A1 M + MA3 > A1 A2 + A2 A3 . Then
the perimeter of the cyclic n-gon A1 MA3 . . . An is greater than that of Γ . So, we
may replace Γ by A1 MA3 . . . An .)
Now, since Γ contains Z in its interior,

./ A1 ZA2 < 2π/n < / A2 ZA3 .

⎛

Let E2 be a point on A1 A2 A3 satisfying / A1 ZE2 = 2π/n. Then, A1 E2 > A1 A2 .

Since

./ A1 ZA2 + / A2 ZA3 = / A1 ZE2 + / E2 ZA3 = 2π/n + / E2 ZA3 ,

we have / E2 ZA3 = / A1 ZA2 + / A2 ZA3 − 2π/n > / A1 ZA2 , and so we get

E2 A3 > A1 A2 . Hence by Corollary 7.1, A1 E2 +E2 A3 > A1 A2 +A2 A3 . Therefore,
the perimeter of the n-gon Γ1 := A1 E2 A3 . . . An is greater than the perimeter of Γ .
Notice that Γ1 contains Z in its interior. If Γ1 is not a regular n-gon, then, replacing
the order of the edges, we can bring the longest edge and the shortest edge into
consecutive position. For instance, suppose that E2 A3 is the shortest edge and A3 A4
⎛

is the longest edge. Let E3 be a point on E2 A3 A4 satisfying / E2 ZE3 = 2π/n.

Then the n-gon Γ2 := A1 E2 E3 A4 . . . An is inscribed in Λ, and its perimeter is
greater than that of Γ1 . Repeating a similar process, we can get a regular n-gon
whose perimeter is greater than that of Γ . ⨆
⨅
Theorem 7.8 Among the spherical n-gons with fixed perimeter p < 2π , regular
n-gons have the maximum area.
Proof Let Γ be a spherical n-gon with perimeter p.
(i) By Lemma 7.4, there is a cyclic n-gon with the same edge-length as Γ . Then
by Theorem 7.6, the area of this cyclic n-gon is greater than or equal to the
area of Γ .
7.6 Appendix 151

(ii) By Lemma 7.5, we can reorder the edge-lengths of this cyclic n-gon so that
the longest edge (of length a) and the shortest edge (of length b) become
consecutive. (Thus, it becomes a Ψ(. . . , a, b, . . . ).)
(iii) Replace a, b in Ψ(. . . , a, b, . . . ) with e, a + b − e (where e = p/n) and then
change the n-gon to a cyclic one. Then the area of the resulting n-gon is at least
the area of [Γ ] by Theorem 7.4.
By repeating (i), (ii) and (iii) we can show that a regular n-gon with perimeter p
has area like at least that of Γ . ⨆
⨅

7.6 Appendix

7.6.1 Orthogonal Projections of a Spherical Ellipse

Suppose that .A, B lie on the xz-plane, the midpoint of AB is the south-pole
(0, 0, −1), and
.

A = (− sin φ, 0, − cos φ), B = (sin φ, 0, − cos φ),

.

where .2φ < π . For .X = (x, y, z) ∈ E(A, B : s), put .AX = s − t, BX = s + t,

where .−φ < t < φ. Then, since ./ AOX = AX = s − t, / BOX = BX = s + t,
we have
−→ −→
. cos(s − t) = cos(AX) = OA · OX = −x sin φ − z cos φ,
−→ −→
cos(s + t) = cos(BX) = OB · OX = +x sin φ − z cos φ,

and therefore

. cos s cos t + sin s sin t = −x sin φ − z cos φ,

cos s cos t − sin s sin t = +x sin φ − z cos φ,

from which we have

⌠
cos s cos t = −z cos φ
. (7.2)
sin s sin t = −x sin φ.

If .2s = π , then we have .z = 0, and .E(A, B : s) is a great circle of .S. If .2s < π ,
then .E lies in the lower hemisphere, whereas, if .2s > π , then .E lies in the upper
hemisphere. Suppose .2s /= π . From (7.2), we have
⎛ ⎞2 ⎛ ⎞2
cos φ sin φ
. z2 + x 2 = cos2 t + sin2 t = 1.
cos s sin s
152 7 Spherical Geometry III

Fig. 7.4 The foci of the

ellipse are denoted by .•

Put .p = cos φ/cos s, q = sin φ/sin s. Then, since .φ < s < π − φ, we have
p2 > 1 > q 2 and
.

p2 z2 + q 2 x 2 = 1.
. (7.3)

From this and .z2 = 1 − x 2 − x 2 , we have

1 = p2 (1 − x 2 − y 2 ) + q 2 x 2 = p2 − (p2 − q 2 )x 2 − p2 y 2 ,
.

and hence
⎛ ⎞ ⎛ ⎞
p2 − q 2 p2
. x2 + y 2 = 1. (7.4)
p2 − 1 p2 − 1

This represents an ellipse in the xy-plane. Thus, the orthogonal projection of .E into
the xy-plane is an ellipse, which is contained in the unit circle centered at the origin.
Figure 7.4 shows the ellipse (7.4) for the case .φ = π/4, s = π/3, and .Ā, B̄ show the
projections of .A, B, respectively. Note that .Ā, B̄ are not the foci of the ellipse (7.4).
The Eq. (7.3) also represents an ellipse in the xz-plane, but in this case the
projection of .E(A, B : s) is only a part of the ellipse (7.3).
Next, let us consider the figure obtained by the orthogonal projection of .E(A, B :
s) into the yz-plane. From (7.3) and .x 2 = 1 − y 2 − z2 , we have

(p2 − q 2 )y 2 − q 2 z2 = 1 − q 2 .
. (7.5)

In the yz-plane, the Eq. (7.5) represents a hyperbola. See Fig. 7.5, which shows the
hyperbola in the case .φ = π/4, s = π/3.
Thus, the projection of .E into the yz-plane is a part of a hyperbola, see also [425,
Theorem 3.1]. The hyperbola (7.5) divides the yz-plane into three regions, and the
region that contains the y-axis is called the exterior. The other two regions are called
the interior of the hyperbola. Denote by .Ē the orthogonal projection of .E into the
yz-plane, and by .X̄ the orthogonal projection of a point .X ∈ S. Note that if .2s /= π ,
then since .E is connected, .Ē is contained in one of the connected components of
7.6 Appendix 153

Fig. 7.5 The case

.φ= π/4, s = π/3

the hyperbola (7.5). Since .A, B, A∗ , B ∗ lie on the xz-plane, the geodesic segments
∗
.AB , AB are mapped into the z-axis by the orthogonal projection into the yz-plane.

Since the geodesic segment .AB ∗ crosses the ellipse at a point, and .AB ∗ is mapped
onto the line segment .ĀB̄ ∗ bijectively, the line segment .ĀB̄ ∗ must cross .Ē, and since
the hyperbola (7.5) is symmetric with respect to the y-axis, we can deduce that both
∗
.Ā, B̄ lie in the interior of the hyperbola (7.5).

7.6.2 Proof of Lemma 7.1

Applying the orthogonal projection into the yz-plane, we prove Lemma 7.1.
Let .2s = AC + BC. In the case .s = π/2 (that is, when .E(A, B : C) is a great
circle), the lemma will be trivial. So, we assume that .s /= π/2. Then .Ē is contained
in one of the components of the hyperbola (7.5). If .s < π/2, then .Ē is contained in
the .(z < 0)-component of the hyperbola (7.5), whereas .s > π/2 implies that .Ē is
contained in the .(z > 0)-component of the hyperbola (7.5). Recall that .Ā and .B̄ ∗ lie
in the interior of the hyperbola (7.5).
Since the line through A and B is parallel to the x-axis, we have .Ā = B̄, and
⎛

the plane that contains the arc .ACB is perpendicular to the yz-plane. Hence the
⎛

orthogonal projection of .ACB into the yz-plane becomes a line segment connecting
⎛

.Ā and .M̄, where M is the midpoint of the arc .ACB, see Fig. 7.5. Since .Ā is a point
−−→
in the interior of the hyperbola (7.5), the ray .ĀM̄ intersects .Ē only at the point
.C̄(= D̄). Since any line orthogonal to the yz-plane intersects .S in at most two
⎛

points, we have .E(A, B : C) ∩ ACB = {C, D}. Since the line segment .ĀM̄ crosses
(

Ē at .C̄, M must lie in the exterior of .E, and we can deduce that the subarc .CD lies,
.

except its two endpoints, in the exterior of .E. ⨆

⨅
154 7 Spherical Geometry III

Fig. 7.6 Exercise 7.2

(see below)

P C Q

A B

7.7 Exercises

7.1 For a spherical triangle Δ̆ABC on S, let X ∈ S, X /= A∗ , be a point such

⎛
that the shortest geodesic AX intersects the Lexell arc A∗ CB ∗ . Prove that
[ABX] ≥ [ABC].
7.2∗ Let Δ̆ABC be an equilateral spherical triangle with edge length d < π/2.
Let KA , KB , KC be circles on S with spherical radius d and centers A, B, C,
respectively. Let P be the crossing points of KA , KC different to B, and Q be
the crossing point of KB , KC different to A, as shown in Fig. 7.6. Prove that
⎛
⎛

the arc P CQ is a subarc of the Lexell arc A∗ CB ∗ .

7.3∗ (Fejes Tóth’s lemma) Let Δ̆(a, b, c), a ≤ b ≤ c, be a spherical triangle on S.
Prove that if the spherical radius of the circumscribed cap of this triangle is
smaller than a, then [Δ̆(a, b, c)] ≥ [Δ̆(a, a, a)]. (If a ≥ π/2, then Δ̆(a, b, c)
is a major triangle, and the result follows from the major triangle lemma.)
7.4 Represent the edge-length of an equilateral spherical triangle on S using its
area.
7.5 By the orthogonal projection of the spherical ellipse E(A, B : s) with foci
A = (− sin φ, 0, − cos φ) and B = (sin φ, 0, − cos φ) on S to the xy-plane,
we get the ellipse (5/3)x 2 + (3/2)y 2 = 1. Find φ and s.
7.6∗ Among the spherical convex regions on S with a fixed perimeter, there exists
one that has the maximal area. Assuming this is true, prove that one that has
the maximum area is a spherical cap.

7.8 Notes

Since some of the themes and topics discussed in this chapter here occur already in
other parts of the book, we focus in these notes on spherical ellipses and Lexell’s
theorem.
7.8 Notes 155

7.8.1 Notes on Spherical Ellipses

The authors were not able to locate many references on spherical ellipses and related
topics, and only two (partially) corresponding surveys (namely [403] and [425])
could be found. Therefore all identified papers are listed below in chronological
order.
Fabricius-Bjerre [254] presents a nicely unified and simplified derivation of geo-
metric properties of spherical conics. Schilling [696] considers spherical ellipses in
3-space as intersections of a 2-sphere S and a suitable cone whose apex is the center
of S. These curves can be seen as ellipses defined analogously to the usual concept
with foci, and with the help of spherical distances. He shows that many of the
elementary focal properties of usual ellipses can be carried over to these curves in S.
This yields then corresponding theorems on ellipses in elliptic geometry. Pursuing
the same goals, Namikawa [593] studies spherical ellipses and hyperbolas.
Similar to Schilling [696], Dirnböck [214] introduces spherical conics as inter-
section curves of the 2-sphere and certain cones, and he studies their polar
conics and focal properties with the help of spherical evolutes. With analogous
methods, spherical loxodromic curves and the spherical tractrix are studied, and
their kinematic generation is presented.
A somehow special topological result is obtained by Chilakamarri et al. [163].
They complete the construction of all possible types of spherical Venn diagrams on
five curves by a careful case analysis. Recall that a Venn diagram is simple if at
most two curves intersect at a point, and it is exposed if each of its curves has an
arc on the boundary of the unbounded exterior region. Further on, it is convex if it is
isomorphic with a Venn diagram all of whose curves are convex. Completing a series
of earlier papers dedicated to this subject, the authors use graph theory to enumerate
all simple spherical Venn diagrams generated by five participating curves. E.g., it is
shown that there are 20 nonisomorphic simple spherical Venn diagrams of 5 curves,
11 of which are convex. They establish which of these spherical Venn diagrams are
convex, which are exposed, and which can be drawn with congruent ellipses. Kang
and Lin [442] investigate equilateral spherical drawings of planar Cayley graphs,
concentrated on the case that the underlying group is generated by two rotations.
Due to this restriction, the set of such equilateral drawings can be parametrized by
spherical ellipses on the unit sphere.
Schröcker [715] generates so-called sphero-conics on the sphere S as Schilling
[696] (see above) does it with his spherical ellipses (as intersection of S and
quadratic cones). Schröcker characterizes a sphero-conic as locus of points such
that the absolute value of the sum or difference of tangent distances to two fixed
circles is constant. From the two given proofs the first one is based on methods
of descriptive and projective geometry, and the second one on algebraic methods.
The proofs remain valid in case of purely imaginary tangent distances (i.e., when
the sphero-conic is enclosed by both circles). Minor modifications of the algebraic
proof yield also the hyperbolic case.
156 7 Spherical Geometry III

Besides other topics, Weber and Schröcker [800] study, for the elliptic plane,
the analogue of the unique ellipse of minimal area enclosing a compact, convex
set. They show that in this non-Euclidean situation the uniqueness of this minimal
enclosing ellipse is guaranteed if the center or the axis of the conic is prescribed.
With new methods they establish also sufficient conditions for the uniqueness of
such ellipses. Finally, for ellipses of line sets corresponding uniqueness results are
obtained. (This follows the duality principle between points and lines in the elliptic
plane.) Finally, some open questions concerning extremal quadrics are formulated.
For analogous results on minimal area ellipses in the hyperbolic case we refer to
[801].
Horváth [403] presents an informative survey on conics and roulettes (generated
by points that maintain a fixed position relative to a curve rolling on another
one, without slipping) in non-Euclidean geometries, among them also spherical
geometry.
Izmestiev [425] nicely surveys results on conics (also) in spherical geometry.
After defining conic sections and polarities in spherical geometry, the author gives
a corresponding classification and presents several theorems with proofs. E.g., he
shows that a tangent to a spherical conic cuts from the lune formed by the focal lines
a triangle of constant area. Also Ivory’s theorem (the diagonals in a quadrilateral
formed by four suitable confocal conics have equal lengths) is studied for the case
of spherical geometry.
Maehara [529] shows that the following holds for spherical ellipses and triangles
ABC: For ABC with .AC /= BC the circular arc from A to B via C and the spherical
ellipse with foci A and B passing through C intersect in precisely two points, and
these two curves intersect each other at the intersection points. From this, extremal
properties of spherical equilateral triangles are derived.

7.8.2 Notes on Lexell’s Theorem

Lexell’s theorem refers to spherical triangles of equal surface area, and so it has
various applications and extensions in spherical geometry. An additional peculiarity
of that theorem is the fact that many famous mathematicians proved it in different
ways. By that reason, this subsection here is strongly historical in nature.

The Theorem Itself and Different Proofs

Lexell’s theorem says that each spherical triangle on the 2-sphere, with a fixed base
AB and having the same surface area, has its third vertex on a small circle (called
“Lexell’s circle”) which passes through each of the two points being antipodal to A
and B. Thus, the locus of the moving vertex of spherical triangles with fixed base
and fixed spherical area is a small circle. Leonhard Euler and Anders Johan Lexell
gave written proofs of this theorem at the same time, and Euler mentioned that he
7.8 Notes 157

was inspired by Lexell. But it turned out that the publications of the result followed
in different years, namely in 1784 by Lexell [504] and in 1797 by Euler [247].
In [539], Maehara and Martini present seven proofs of Lexell’s theorem, most
of them coming from famous mathematicians (namely Leonhard Euler, Jakob
Steiner, Carl Friedrich Gauss, Josef-Alfred Serret, and László Fejes Tóth) or being
of particular elegance. The sequence of the following selection of very shortly
described proofs of Lexell’s theorem agrees with that in [539]. Lexell’s own proof,
using angle relations, is given and commented in [504] and [616]. Euler (see [247]
and [248]) proved Lexell’s theorem in a way similar to the proof of Proposition 37
in Euclid’s Elements (based on the elementary statement that “triangles which are
on the same base and between the same parallels are equal to one another”); see
again [616]. And also Gauss took a similar way, cf. [46]; he inserted a proof into
a letter to Schumacher in 1841, in response to a related proof sent from Clausen
via Schumacher to him. Jacob Steiner (see [755], [144, p. 93], and [616]) used
the inscribed angle theorem, and Maehara [519] found independently an analogous
approach. According to Casey [144, p. 93], Josef-Alfred Serret proved Lexell’s
theorem by showing that the radius of the Lexell circle of the triangle ABC depends
only on the length of AB and the area. Using polar triangles, also Lászlo Fejes Tóth
gave an own proof of Lexell’s theorem in [257, Section 1.8], see also [262, Section
1.8]. And Aleksander Simonić [734] presented an elegant proof via stereographic
projection.
Besides these proofs reflected in [539] there are further ones due to famous
mathematicians. Jacob Steiner, already mentioned above, continued in 1841 with
a new proof of Lexell’s theorem, see [756]. In addition, he mentions in this paper
that the theorem was also known to Adrien-Marie Legendre who published a proof
based on spherical trigonometry in [500, Note X, Problem III]. Later, in 1855, Henri
Leon Lebesgue became involved, repeating in [496] one of Euler’s approaches and
mentioning that a related proof of Eugene Charles Catalan exists, which is indeed
the case (see [146, Book VII, Problem VII]). Since in some of these publications
Euler’s related work is underestimated or even not mentioned, we refer once more
to the excellent modern reference [616]. Related contributions of Nicholas Fuss
(the secretary of Euler in Saint Petersburg) and other collaborators of Euler are
nicely discussed in [46]. And we mention that further modern proofs and related
discussions can be found in [519, 531], and [537].

Some History

The scientific life of the Swedish astronomer and mathematician Anders Johan
Lexell (1740–1784) is one of the best examples for a strong combination of
mathematics and astronomy. The first complete biography of him was written by
Johan Stén [760] (here we refer also to Calinger’s review [139]). Lexell was born
in Turku (today located in Finland, but in those days belonging to the Swedish
kingdom), and there he also began his mathematical studies on differential and
integral calculus. For “mathematical tourists” it should be mentioned that there is a
158 7 Spherical Geometry III

monument of Lexell at the University of Turku (see [628]). In 1766 (2 years after his
return to the Petersburg Academy) Leonhard Euler invited Lexell to work with him
at the Kunstkammer there (see [248] and Coxeter’s corresponding review [183]),
and a fruitful period of collaboration began. Lexell created a complete theory of
comets, and various famous astronomical findings are due to him. For instance, the
discovery of the first near-Earth object, a Jupiter-family comet, resulted in naming
this object after him (see [825] and [139]). A lunar crater is also named after him.
During a journey throughout Europe, Lexell was meeting several famous math-
ematicians, such as Johann III Bernoulli, Jean d’Alembert, Marie-Jean-Antoine-
Nicolas de Caritat (Marquis de Concordet), and Pierre-Simon de Laplace. His
mathematical work was mainly concentrated on the theory of ordinary differential
equations and spherical trigonometry. In astronomy, he was the first “calculator”.
Namely, he could show that a new heavenly object (which was discovered by
Herschel in 1781, called “King George’s Star”, and thought to be a comet) was
in fact a planet, later named Uranus. Lexell was also involved in the determination
of the solar parallax in connection with the Venus passage 1769, and he obtained a
very good approximation of that parallax.
In 1783, Lexell became the successor of Euler at the Imperial Academy of
Sciences in Saint Petersburg, but he filled that position only briefly, due to his
premature death in 1784. Nevertheless, he became one of the most successful
members of the Russian Academy of Sciences at that time. His mathematical work
was praised by Leonhard Euler and Daniel Bernoulli. Further interesting historical
articles on Lexell and his scientific activities are Atzema [46], Lysenko [513], and
Zhukova [832], the first of which directly treats Lexell’s theorem as a central topic.
This theorem was also included into influential monographs of Lexell’s time, e.g.,
into Legendre’s book [500] from 1808. Later it was discussed in Hadamard’s book
[353, pp. 392–393], and more recent examples of monographs discussing it are
Berger’s book [75, 18.11.10] and Section 1.8 in the already mentioned classical
book [257] of L. Fejes Tóth and its strongly expanded edition [262].

Extensions and Modifications

Regarding research aspects around Lexell’s theorem, especially non-Euclidean

variants are of interest. In these directions, various modifications and analogues
of Lexell’s theorem have been derived. For the hyperbolic case the reader should
consult Papadopoulos and Su [621] as well as Frenkel and Su [281]. And regarding
analogues in absolute geometry we refer to the papers [631] of Persson and [612] of
Pambuccian.
Using so-called Cesàro triangles (recall that these are planar triangles obtained
via stereographic projection from spherical ones), various results on spherical
triangles (which are projected to get the Cesàro ones) are obtained by Maehara and
Martini [537]. These concern the relation of sides of a spherical triangle and their
opposite angles, Lexell’s theorem, and a result on the area of spherical triangles
with one variable side-length. The obtained results are applied to derive several
isoperimetric theorems for spherical polygons.
Chapter 8
Geometric Probability on the Sphere

If we knew what it was we were doing, it would not be called

research, would it?
— Albert Einstein

In this chapter, we study the probabilities concerning several random variables or

events, determined by random points or random great circles on the sphere.

8.1 Random Points on a Sphere

Let .S denote the unit sphere centered at the origin O in .R3 . A randomly chosen
point on .S is called a random point on .S. More precisely, a random point is a point
uniformly distributed on .S, that is, a point .p ∈ S chosen in such a way that, for any
subset .U ⊂ S, the probability that .p ∈ U is equal to .[the area of U ]/(4π ).
Example 8.1 Let .u, v be independent random numbers, both uniformly distributed
in the interval .[0, 1]. Let .p be a point on .S with longitude .τ = π(2u−1) and latitude
−1
.θ = sin (2v − 1). Then .p is a random point on .S.
This can be checked as follows. Put .p = (τ, θ ), and let .ϕ : Š → Γ be the cylindrical
projection. Consulting Remark 4.1 in Chap. 4, for any .U ⊂ S we have

.p ∈ U ⇔ ϕ(τ, θ ) ∈ ϕ(U ) ⇔ (π(2u − 1), 2v − 1) ∈ ϕ(U ).

Note that .(π(2u−1), 2v −1) is uniformly distributed on the quadrilateral .[−π, π ]×

[−1, 1] obtained from the lateral face of the cylinder .Γ by cutting along the line
.τ = 0 and unfolding it. Hence the probability that .(π(2u − 1), 2v − 2) ∈ ϕ(U )

is equal to .[the area of ϕ(U )]/(4π ). Since the cylindrical projection .ϕ preserves the
area of a figure, it follows that .Pr(p ∈ U ) = (the area of U )/(4π ). Thus, .p is a
random point on .S.

© The Author(s), under exclusive license to Springer Nature Switzerland AG 2024 159
H. Maehara, H. Martini, Circles, Spheres and Spherical Geometry,
Birkhäuser Advanced Texts Basler Lehrbücher,
https://doi.org/10.1007/978-3-031-62776-7_8
160 8 Geometric Probability on the Sphere

Example 8.2 If .p is a random point of .S, then its antipodal point .p ∗ is also a
random point.
Indeed, if we denote by .U ∗ ⊂ S the set which is symmetric to .U ⊂ S with respect
to the center O of .S, then “.p ∗ ∈ U ⇔ p ∈ U ∗ ” holds, and since the areas of U and
∗
.U are equal, we have

the area of U ∗ the area of U

. Pr(p∗ ∈ U ) = Pr(p ∈ U ∗ ) = = .
4π 4π
For a random point .p on .S, the pair .{p, p ∗ } is called a random diameter. Thus,
each endpoint of a random diameter is a random point. A great circle divides .S into
two hemispheres; the poles (centers) of the two hemispheres are called the poles
of the great circle. Thus, every great circle has two poles, and they are mutually
antipodal. Any point on .S determines a great circle, being one of its poles. A random
great circle is a great circle determined by taking a random point (or a random
diameter) as its pole (or its poles).
Example 8.3 If .G is a random great circle on .S, then each pole of .G is a random
point on .S.
Example 8.4 The probability that independently chosen three random great circles
pass through the same point is 0.
This can be seen as follows. Let us take three random great circles independently
in the order .G1 , G2 , G3 . Let .x, x ∗ be the intersection points of .G1 , G2 . Then .G3
passes x whenever the pole (which is a random point) of .G3 lies on the great circle
with pole x. Since the area of a great circle is 0, the probability that independently
taken three random great circles pass through the same point is 0.

8.2 Random Spherical Triangles

Let .x, y be independent random points on .S. Then the length of the geodesic
segment .θ = xy is a random variable. Let us find the probability density function
.f (θ ) of .θ . When we consider the distribution of .θ = xy, by the symmetry of .S, we

may suppose that .x is the north pole .(0, 0, 1) of .S. Then .xy ≤ θ if and only if .y
lies in the spherical cap with center .x and angular radius .θ . Since this cap has area
.2π(1 − cos θ ), the distribution function .F (θ ) = Pr(xy ≤ θ ) is given by

2π(1 − cos θ ) 1 − cos θ

F (θ ) =
. = .
4π 2
dF (θ)
Hence the probability density function .f (θ ) = dθ is given by

sin θ
f (θ ) =
. .
2
 8.2 Random Spherical Triangles 161

Therefore, the expected value .E(θ ) is

⎰ π ⎰ π
θ sin θ
E(θ ) =
. θf (θ )dθ = dθ
0 0 2
1⎾ ⏋π π
= sin θ − θ cos θ 0 = .
2 2
Theorem 8.1 For a spherical triangle determined by three random points chosen
independently on .S, the expected value of its perimeter L is .3π/2, and the expected
value of its area A is .π/2.
Proof Let .x, y, z be three random points on .S. Then

E(L) = E(xy + yz + zx) = 3π/2.

.

By Girard’s formula, .A = / x + / y + / z − π . When we consider the size of

./ x = / yxz, we may regard .x, y as fixed points. Then .Pr(/ yxz < t) is equal to

the probability that .z lies in the lune of angle 2t with bisector .xy, which is equal to
.t/π . Therefore, ./ yxz is uniformly distributed in the interval .[0, π ]. Hence .E(/ x) =

π/2. Thus .E(A) = 3π/2 − π = π/2. ⨆

⨅
For two random circles .Gi , i = 1, 2, chosen independently on .S, let .ϕ =
/ (G1 , G2 ) be the non-obtuse angle formed by .G1 , G2 at their intersection. When
we consider the distribution of .ϕ, we may regard .G1 as a fixed great circle. Let .p
be a pole of this .G1 , and let .H (p) be the hemisphere with center .p. Let .x 2 denote
the pole of .G2 contained in .H (p). Then .x 2 is uniformly distributed in .H (p). Note
that .ϕ = px 2 . Hence .Pr(ϕ < t) = 2π(1 − cos t)/(2π ) = 1 − cos t. Thus, we have
the following
Theorem 8.2 For two independent random great circles .G1 , G2 , the probability
density function of the intersection angle .ϕ = / (G1 , G2 ) is given by .f (ϕ) =
sin ϕ, 0 ≤ ϕ ≤ π/2. ⨆
⨅
The expected value .E(ϕ) is computed as follows.
⎰ π/2
E(ϕ) =
. ϕ sin ϕdϕ
0
⎰ π/2
= (cos ϕ − (ϕ cos ϕ)' )dϕ
0
π/2 π/2
= [sin ϕ]0 − [ϕ cos ϕ]0 = 1.
162 8 Geometric Probability on the Sphere

8.3 Random Quartets

For a subset .V ⊂ S, the intersection of all hemispheres that contain V is called the
convex hull of V in .S. (If no hemisphere contains V , then the convex hull of V is
.S.) For example, the convex hull of a pair of points .x, y ∈ S lying in the interior

of a hemisphere is the geodesic segment xy. If three points .x, y, z ∈ S lying in the
interior of a hemisphere do not lie on a great circle, then the convex hull of .x, y, z
is a spherical triangle.
A set of .n (≥ 3) great circles is said to be in general position, if no three circles
of them pass through the same point.
Lemma 8.1
(1) If four great circles on .S are in general position, then these four great circles
divide the sphere .S into 14 regions. Among them, eight are spherical triangles,
and six are spherical quadrilaterals. See Fig. 8.1, with .G1 , G2 , G3 , G4 as four
great circles in general position.
(2) If four great circles on .S are not in general position, then among the regions
divided by these great circles, no region is a quadrilateral.

Proof
(1) (Though it is almost clear from Fig. 8.1, let us present a proof.) Two great circles
meet at(two) points, and since no three great circles share the same point, there
arise .2 42 = 12 points as the intersections of great circles. Since each great
circle is divided into 6 arcs by other three circles, there arise 24 arcs in total.
Hence, we obtain a connected graph realized on .S that has 12 vertices, and 24
edges. Then by Euler’s formula .12 − 24 + f = 2, and hence the number f of
faces of this graph is 14. Since the number of great circles is four, each face is
either a triangle or a (convex) quadrilateral. Let .f3 , f4 denote the numbers of
triangles and quadrilaterals, respectively. Then .f3 + f4 = 14 and .3f3 + 4f4 =
2 × 24, and we have .f3 = 8, f4 = 6.
(2) Among the four great circles, suppose that three of them pass through the same
point x. By these three circles, .S is divided into six lunes. The remaining

Fig. 8.1 Division of .S by

four great circles in general
position G1
G4
G3

G2
8.3 Random Quartets 163

great circle divides each lune into at most two triangles, and there arises no
quadrilateral.
⨆
⨅
Theorem 8.3 When we choose four random points on .S independently, then the
probability that the convex hull of these four points in .S becomes a spherical
quadrilateral on .S is .3/8.
Proof We take four random points on .S in the following two-steps way. First choose
four random great circles .G1 , G2 , G3 , G4 independently. Next, from the pair of
poles of each great circle .Gi , choose either one with equal probability. In this
two-steps way, we can take four independent random points .p1 , p 2 , p 3 , p 4 . Let
.K denote the convex hull of these four points in .S. If .K is a spherical quadrilateral,

then .p1 , p 2 , p 3 , p 4 are the vertices of the quadrilateral. Now, by Theorem 4.7, we
have

K is a quadrilateral
.

⇔ K ◦ = H (p 1 ) ∩ H (p2 ) ∩ H (p3 ) ∩ H (p 4 ) is a quadrilateral.

Four great circles .G1 , G2 , G3 , G4 divide .S into several regions. Since to choose
a pole of .Gi corresponds to choose one side of .Gi , .K ◦ = H (p1 ) ∩ H (p 2 ) ∩
H (p 3 ) ∩ H (p4 ) is nonempty if and only if .K ◦ is one of the regions in .S divided
by .G1 , . . . , G4 . By Lemma 8.1(2), in the case of four great circles that are not in
general position, there arises no quadrilateral. Hence .K ◦ is a spherical quadrilateral
if and only if
(i) .G1 , G2 , G3 , G4 are in general position, and
(ii) for each .Gi , one side of it is chosen so that their intersection becomes a quadrilateral.

By Example 8.4, (i) occurs with probability 1. Under the condition (i), the number
of ways to choose one side of each .Gi for .i = 1, 2, 3, 4 is .24 = 16, and among these
16 ways, just 6 ways determine quadrilaterals. Namely, under the condition (i), the
probability that (ii) occurs is .6/16 = 3/8. Therefore, the probability that the convex
hull of .p 1 , p 2 , p 3 , p 4 becomes a spherical quadrilateral is .3/8. ⨆
⨅
Remark 8.1 When we choose independently four random points on .S, the prob-
ability that the convex hull of these four points becomes a spherical triangle is
.8/16 = 1/2.

Remark 8.2 When we choose independently random four points, each uniformly
distributed in a bounded region W in the plane, what is the probability .P (W ) that
the convex hull of these four points becomes a quadrilateral? This is known as
Sylvester’s four-point-problem. It is known that, for example,

35
P (triangle) = 2/3, P (rectangle) = 25/36, P (circle) = 1 −
. .
12π 2
Theorem 8.3 is regarded as the sphere version of Sylvester’s four-point-problem.
164 8 Geometric Probability on the Sphere

Corollary 8.1 For four random points .x i , i = 1, 2, 3, 4, taken independently on .S,

the probability that the geodesic segments .x 1 x 2 and .x 3 x 4 cross each other is .1/8.
Proof The two segments .x 1 x 2 and .x 3 x 4 cross each other if and only if the convex
hull of .V := {x 1 , x 2 , x 3 , x 4 } is a spherical quadrilateral, and .x 1 x 2 is a diagonal
of the quadrilateral. Under the condition that the convex hull of V is a spherical
quadrilateral, each of the three segments .x 1 x 2 , x 1 x 3 , x 1 x 4 becomes a diagonal of
the quadrilateral with equal probability. Hence .x 1 x 2 and .x 3 x 4 cross each other with
probability . 38 × 13 = 18 . ⨆
⨅
Remark 8.3 For more about geometric probability on the sphere, see H. Maehara
and H. Martini [532]. On higher dimensional spherical versions of Sylvester’s
problem, see [534].

8.4 Wendel’s Theorem

Lemma 8.2 If n great circles on S are in general position, then they divide S into
n2 − n + 2 regions.
( )
Proof Since two great circles intersect at two points, we have 2 n2 = n2 − n points
as the intersection of n great circles. Let us regard the figure obtained by these n
great circles as a connected graph realized on S. It has n2 − n vertices, and each
vertex has degree 4. Hence due to the Hand-shaking lemma, it has 2(n2 − n) edges.
Now, by Euler’s formula, the number of faces is equal to n2 − n + 2. ⨆
⨅
The next theorem is the 2-dimensional version of a theorem due to J. G. Wendel
[807].
Theorem 8.4 (J .G .Wendel) The probability that independently chosen n random
points on S lie on a hemisphere is equal to (n2 − n + 2)/2n .
Proof We take n random points on S in the following two-steps way. First choose
n random great circles G1 , G2 , . . . , Gn independently. Next, from the pair of poles
of each great circle Gi , choose either one with equal probability. In this two-steps
way, we can take n independent random points p1 , p 2 , . . . , p n . Let H (pi ) be the
hemisphere with center p i . If p i , i = 1, 2, . . . , n, lie in a hemisphere with center
q, then p i q ≤ π/2 for every i, which implies that q ∈ H (pi ) for i = 1, 2, . . . , n.
Thus,

p1 , p 2 , . . . , p n lie on a hemisphere ⇔ H (p1 ) ∩ H (p2 ) ∩ · · · ∩ H (pn ) /= ∅.

.

Random great circles Gi , i = 1, 2, . . . , n, are in general position with probability

1. Hence they divide S into n2 − n + 2 regions by Lemma 8.2. H (p1 ) ∩ H (p 2 ) ∩
· · · ∩ H (pn ) =
/ ∅ if and only if H (p1 ) ∩ H (p 2 ) ∩ · · · ∩ H (p n ) is one of n2 − n + 2
regions. Once n random great circles are chosen, then there are 2n ways to choose
one pole for each great circle. Thus, once n random great circles are chosen, then the
8.5 Crofton’s Formula 165

probability that p i , i = 1, 2, . . . , n, lie on a hemisphere is equal to (n2 − n + 2)/2n .

Since this is constant and independent from the choice of n random great circles,
the theorem follows. ⨆
⨅
Let us prove here the circle version of the above theorem in a similar way.
Theorem 8.5 The probability that n random points independently and uniformly
distributed on a circle lie on a semicircle is equal to 2n/2n .
Proof Let us take n random points on a unit circle in the following two-steps way:
First, take n random “diameters” Ii , i = 1, 2, . . . , n, of the circle independently,
and then choose one endpoint x i from each diameter Ii with equal probability. The
n points x i , i = 1, 2, . . . , n, are random independent points. For a point p on the
circle, let us denote by S(p) the semicircle with center (midpoint) p. Then

.x 1 , x 2 , . . . , x n ∈ S(p) ⇔ px ≤ π/2, for i = 1, 2, . . . , n

⇔ p ∈ S(x i ), for i = 1, 2, . . . , n.

Hence, x i , i = 1, 2, . . . , n, lie on a semicircle if and only if

S(x 1 ) ∩ S(x 2 ) ∩ · · · ∩ S(x n ) /= ∅.

. (8.1)

The 2n endpoints of S(x i ), i = 1, 2, . . . , n, divide the circle into 2n arcs, and (8.1)
holds whenever S(x 1 ) ∩ · · · ∩ S(x n ) is one of the 2n arcs. Once n random diameters
are chosen, x i , i = 1, 2, . . . , n, are chosen in 2n ways, and (8.1) holds in 2n ways.
Thus, the probability that x i , i = 1, 2, . . . , n, lie on a semicircle under the condition
that n diameters are given is equal to 2n/2n . Since this is constant, we can deduce
that the probability that x i , i = 1, 2, . . . , n, lie on a semicircle is 2n/2n . ⨆
⨅
Example 8.5 For three random points x, y, z chosen independently on a circle, the
probability that the planar triangle Δxyz contains the center of the circle is equal to
1 − 2 · 3/23 = 1/4.

8.5 Crofton’s Formula

Let .γ be a geodesic segment with length .θ (< π ), and .G be a random great circle
with poles .x, x ∗ on .S. Then .G passes through a point .p on .γ if and only if .xp =
π/2. Hence, if we put

W = {y ∈ S : yp = π/2 for some p ∈ γ },

.
166 8 Geometric Probability on the Sphere

Fig. 8.2 A segment .γ and

two lunes of angle .θ

θ
γ θ

then we have .G ∩ γ /= ∅ ⇔ x ∈ W . Since W is the sum of two lunes with angle .θ

(see Fig. 8.2), it follows that

2 × 2θ θ
. Pr(G ∩ γ /= ∅) = = .
4π π
Let .Γ be a simple polygonal curve consisting of k segments .γi of length .θi < π ,
.i = 1, 2, . . . , k. Let .νi := ν(G ∩ γi ) denote the number of points in .G ∩ γi . Then
.νi is a random variable with possible values .0, 1, ∞, and .Pr(νi = ∞) = 0. Put

ν(G ∩ Γ ) = ν1 + ν2 + · · · + νk .
.

Since the expected value of .νi is given by .E(νi ) = θi /π , the expected value
E(ν(G ∩ Γ ) is given by
.

θ1 θk θ1 + · · · + θk
ν(G ∩ Γ ) =
. + ··· + = .
π π π

By approximating a given curve on .S by a polygonal curve on .S, the next theorem

follows.
Theorem 8.6 Let .Γ be a curve of length L on .S. Then .E(ν(G ∩ Γ )) = πL , where
.G is a random great circle on .S, and .ν(G ∩ Γ ) is the number of intersection points

of .G and .Γ . ⨆
⨅
Let K be a convex region on .S whose boundary .∂K has length L. The probability
that a random great circle .G is tangent to .∂K is 0. If .G cuts K, then .ν(G∩∂K) = 2.
Thus, the value of .ν(G ∩ ∂K) is either 0 or 2. Hence

L
. = E(ν(G ∩ ∂K)) = 2 × Pr(G ∩ K /= ∅).
π
Thus, we have the next theorem.
8.6 Random Points on a Hemisphere 167

Theorem 8.7 (Crofton) Let K be a convex region with perimeter L. Then

L
. Pr(G ∩ K /= ∅) = .
2π
⨆
⨅

8.6 Random Points on a Hemisphere

Let H be a fixed hemisphere of .S. By a random point .x on H we mean a

point uniformly distributed on H , that is, for any .W ⊂ H , we have .Pr(x ∈
W ) = (area W )/(2π ). Since the hemisphere H is not isotropic, computations
of probabilities and expected values for random figures on H are generally more
difficult than computations on .S. Applying Crofton’s formula, let us calculate the
expected value of the length of a geodesic arc connecting two random points chosen
independently on H .
First, let us recall the law of total probability: Let T be a continuous random
variable with probability density function .f (t). Then
⎰ ∞
. Pr(· · · ) = Pr(· · · |T = t)f (t)dt,
−∞

where .Pr(· · · |T = t) is the conditional probability of the event “.· · · ” under the
condition .T = t.
Theorem 8.8 For two random points .x, y chosen independently on the hemisphere
H , the expected value of the spherical distance between .x and .y is equal to .4/π .
Proof Let T be the length of the segment .γ = xy, and let .f (t) be the probability
density function of T . For a random great circle .G on .S, we have
⎰ π
. Pr(G ∩ γ /= ∅) = Pr(G ∩ γ =
/ ∅ | T = t)f (t)dt,
0

which is, applying Crofton’s formula, equal to

⎰ π
t E(T )
. f (t)dt = .
0 π π

Hence we have .E(T ) = π Pr(G ∩ γ = / ∅). The hemisphere H is divided by .G

into two regions, and “.G ∩ γ /= ∅” is equivalent to the property that .x and .y lie
in different regions. Let .p be the center (pole) of H , and .z be a pole of .G. Let .θ
denote the spherical distance between .z and .p. Then .θ is a random variable, and
its probability density function is given by . 12 sin θ as already seen in Sect. 8.2. The
intersection angle ./ (G, ∂H ) of .G and .∂H (the boundary of hemisphere H ) is equal
to .θ , see Fig. 8.3.
 168 8 Geometric Probability on the Sphere

Fig. 8.3 ..............................................................

./(G, ∂H ) = zp = θ ...................
............................ ●..
.................. p ..
.........
.
... .. ........
... ... ... .... .......
....
. .........
........ θ .
.
. .
.. .
.
..
.. ... . ....
....
........ ... ... .. ...
.....●........... .. ... .... ...
...
z
... .......
.......
.....
.
.
..
. .. ... .. ...
...
.. ....... .... ...
. ..
.
... .... .... .... .... .... .... .... ........................... ........... .... .... .... .... .... .... .... ... ....
.................. . ..●. . .
. . .......
.. .............................................................................................................................................
.. . . . ..
..
...
...
. .
.. ..
.. .... ∂H ..
...
.. ..... ....
...
...
....
. .
.. .... G ..
..
.... .. .... ...
.... ... ..... ..
.....
..... .. ... .
....... .. .... .....
......... .......
........... ........
.....................................................

Now, .G divides H into two lunes with angle .θ and angle .π −θ . Note that .G∩γ =
∅ if and only if both .x, y lie in the same lune (of angle .θ or angle .π − θ). Hence

⎛ t ⎞2 Å ã2
π −t
. Pr(G ∩ γ = ∅ | θ = t) = + ,
π π

and hence

. Pr(G ∩ γ /= ∅ | θ = t) = 1 − Pr(G ∩ γ = ∅ | θ = t)
2t (π − t)
= 2t/π − 2t 2 /π 2 = .
π2
Thus,
⎰ π
1
. Pr(G ∩ γ /= ∅) = Pr(G ∩ γ /= ∅ | θ = t) sin tdt
0 2
⎰ π
2t (π − t) 1
= sin tdt.
0 π2 2

Since
⎰
. t sin tdt = t sin t − t cos t,
⎰
t 2 sin tdt = 2t sin t − (t 2 − 2) cos t,

we have
1 1 î óπ
. Pr(G ∩ γ /= ∅) = [t sin t − t cos t]π0 − 2 t sin t − (t 2 − 2) cos t
π π 0

4
= 2.
π

Therefore, .E(T )/π = 4/π 2 , and .E(T ) = 4/π. ⨆

⨅
8.7 Santaló’s Chord Theorem 169

Corollary 8.2 The expected value of the perimeter of a spherical triangle whose
vertices are random points taken independently on a fixed hemisphere is equal to
.12/π . ⨆
⨅

8.7 Santaló’s Chord Theorem

Since .S is symmetric and isotropic, any point on .S chosen randomly not depending
on the already given subset of .S can be regarded as a random point uniformly
distributed on .S. In particular, each of the intersection points of two independent
random great circles on .S is a random point when we ignore the two random great
circles. Let us state this as a lemma.
Lemma 8.3 Each of the intersection points of two independent random great
circles is a random point on .S. ⨆
⨅
Theorem 8.9 (Santaló’s Chord Theorem) Let K be a convex region on .S. For a
random great circle .G on .S, let .ϕ denote the length of the chord .G ∩ K, see Fig. 8.4.
Then

E(ϕ) = (the area of K)/2.

.

Proof Let .Gi , i = 1, 2, be two random great circles chosen independently on .S.
We may regard each of the intersection points of .G1 , G2 as a random point. Hence
.Pr(G1 ∩ G2 ∩ K /= ∅) = (the area of K)/(2π ).

On the other hand, denoting the probability density function of .ϕ by .f (ϕ), and
applying Theorem 8.6, we can compute .Pr(G1 ∩ G2 ∩ K /= ∅) in the following way:
⎰ π
. Pr(G1 ∩ (G2 ∩ K)) =
/ ∅) | the length of G2 ∩ K = ϕ)f (ϕ) dϕ
0
⎰ π
ϕ E(ϕ)
= f (ϕ) dϕ = .
0 π π

Therefore, .E(ϕ) = (the area of K)/2. ⨆

⨅

Fig. 8.4 Theorem 8.9

K
ϕ

G
170 8 Geometric Probability on the Sphere

The sphere-version of Crofton’s formula and Santaló’s chord theorem were

obtained by L. A. Santaló in [687].

8.8 Exercises

8.1 When we choose five random great circles on S, it is divided into 22

regions with probability 1. Find the numbers of triangles, quadrilaterals, and
pentagons. (Study this by drawing figures.)
8.2∗ Prove that the probability that five random points on S, chosen independently,
form the vertices of a convex pentagon is 1/16.
8.3 Prove that for four random points x 1 , x 2 , x 3 , x 4 on S, chosen independently,
the probability that the geodesic segments x 1 x 2 and x 3 x 4 cross each other is
1/8.
8.4 What is the probability that a random great circle on S intersects a fixed
spherical cap of angular radius ρ < π/2?
8.5 What is the probability that the geodesic segment connecting two random
points on S chosen independently, intersects a fixed great circle on S?
8.6∗ What is the probability that a random great circle on S intersects a given
semicircle of a circle with radius ρ < π/2 in two points?

8.9 Notes

In this chapter the notes consist of the following two parts: First we discuss a longer
part on random points and triangles on the sphere, and then a comprehensive report
on Crofton’s formula and related topics is given.

8.9.1 Notes on Random Points and Triangles on the Sphere

These announced results on configurations of random points and triangles (mainly)

on the sphere have clearly also to do with random polygons, polytopes, and with
famous problems (like Sylvester’s four-point problem, see Sect. 8.3) from the field
of geometric probability. In most cases, the studied objects are finite sets of
independent, uniformly distributed and random points and their convex hulls. After
listing survey-like publications, we present a selection of research papers; in both
cases, mainly chronological order is chosen.
8.9 Notes 171

Book Parts and Surveys

Miles presents with [574] the first large survey on problems and results from
geometric probability on the sphere. E.g., Section 3 of this expository paper refers
to spherical caps that contain two, three or m uniform independent particles. Using
integral geometry, such results are also extended to arbitrary domains (instead of
caps). Section 5 treats random spherical triangles, getting the joint distributions of
the lengths of the three sides, and of the sizes of the three angles (the sides being
pairwise independent). In Section 6, tessellations of the sphere determined by n
uniform random great circles are studied. For the polygons in such a tessellation a
long list of first and second order moments of the area, perimeter, and number of
angles is presented. Section 7 offers then related results on Voronoi and Delaunay
tessellations, and finally the duality between points and great circles in this
framework is discussed, like also extensions to higher dimensions.
Deák [203] gives an overview to the known (computerized) techniques for gen-
erating points uniformly inside an n-dimensional hypersphere and also uniformly
on the surface of such a sphere. He presents modifications converting an interior
algorithm into a surface algorithm and vice versa. An inductive method of projecting
points on the surface of an (.n − 1)-dimensional sphere onto the surface of an
n-dimensional sphere is discussed. Further on, Deák investigates the relationship
between the inductive projection method and the polar transformation method for
getting points within a hypersphere.
In Chapter 5 of their book [63], Barrett and Mackay discuss topics from the
geometry on the sphere, including packing problems and arranging points on
spheres at random. Tichy [777] discusses distribution properties of random points
on the sphere and respective applications to numerical integration and to numerical
methods for solving related equation systems.
Croft et al. [190] survey in their Section B5 several questions on random
polygons and polyhedra, e.g. of the following type: Find bounds on .V (K, m), which
denotes the expected volume of the convex hull of m random points in a convex body
K (which can also be a ball/ellipsoid) of unit volume. For instance, Groemer [331]
showed that the expected volume .V (K, n + 1) is maximal when K is an ellipsoid.
Also Sylvester’s four-point problem (see Sect. 8.3) is discussed in Section B5 of
[190].
Together with several coauthors Kendall obtained many results on a theory of
shape in the plane, with statistical applications, e.g. to the asymptotic probability of
given random sets being nearly circular. We refer here to his monograph [449]. And
Kendall surveys in [447] analogous results on the 2-sphere, which mainly refer to
spherical triangles.
In the monograph [706] Schneider and Weil present results from spherical
integral geometry mainly in Sect. 6.5. Subjects like the local spherical Steiner
formula, (generalized) support measures, spherical intrinsic volumes, and the
spherical convex ring are discussed there. See also the dissertation [303] of Glasauer,
where (among other results) local kinematic formulas for spherically convex bodies
are studied.
172 8 Geometric Probability on the Sphere

Reitzner [662] gives a nice survey on random polytopes. In particular he

underlines the distributional sides of functionals of random polytopes, such as
estimates for higher moments, limit theorems, and large deviation inequalities. The
floating body problem and visibility regions flank this part as geometric tools. And
also Hug [409] presents a nice and comprehensive survey on random polytopes.
In the first part he discusses topics from two main directions: results for random
points in smooth convex bodies that have been extended to convex bodies with
weaker smoothness assumptions, and results on expectations which have been
supplemented by distributional results. In the second part, results on intersecting
random halfspaces, hyperplane tessellations, and cells of random mosaics are
studied. The author presents also an impressive collection of used geometric notions
and methods as well as probabilistic tools (such as duality, stability of results,
associated bodies, zonoids and further topics). This rich representation of methods
and tools is one of the reasons for citing these surveys of Reitzner and Hug here,
although their goal is not the study of spherical polytopes. The paper [410] of Hug
and Schneider is dedicated to tessellations of the Euclidean .(n − 1)-spheres by great
.(n − 2)-subspheres. These are clearly equivalent to tessellations of Euclidean n-

space by hyperplanes passing through the origin, also called conical tessellations.
The authors study expectations for general classes of geometric functionals of
random polyhedral cones. E.g., these are: vertex number, total edge length, surface
area and volume of the generated spherical polytopes. The complete covariance
structure of these geometric functionals for an isotropic random Schläfli cone is
determined, yielding the spherical analogue of a result due to Miles. This research
paper contains also a comprehensive list of references showing various aspects of
this little field between stochastic geometry and polytope theory; therefore it is cited
here (also as survey).
The exposition [531] of Maehara and Martini collects many interesting results
from geometric probability on the 2-sphere in Euclidean 3-space. First the authors
present typical tools and notions from classical spherical geometry which enable
them to offer the results in an intuitive, geometric way. Based on this, they study
random spherical polygons, expectations for perimeters, and areas of spherical
triangles, determined by three random points, and probabilities for the shape of
the convex hull of four and five random points. The increasing difficulties for more
points are explained. Then they derive Crofton’s formula and the chord theorem of
Santaló, yielding probability results on spherical caps (e.g., conditions for covering
the whole 2-sphere when being of the same size). Also a higher dimensional outlook
is given. The authors lay emphasize on elementary proofs and pose interesting open
problems.
Hug and Thäle [411] study splitting tessellations of the Euclidean standard (.n −
1)-hypersphere. These are obtained when an existing tessellation cell is splitted into
two cells by a random great (.n − 2)-hypersphere properly meeting that cell. The
authors give a detailed description of such tessellations, e.g. treating the total cell
curvatures, the geometry of a “typical cell” and various further geometric properties.
The whole little field is comprehensively presented, such that we also decided to
bring the paper here, although it is a research paper.
8.9 Notes 173

In Chapter 3 of the Lecture Notes [705], Schneider treats several results and
notions from spherical geometry. These are integral-geometric identities, a spherical
version of Blaschke’s known selection theorem, and facts about the gnomonic pro-
jection. Valuations on spherical polytopes are discussed, yielding several interesting
problems, and inequalities of isoperimetric type are discussed, too. Several formulas
(like the Steiner formula or the kinematic formula) are studied in Chaps. 4, and
5 refers to arrangements of finitely many hyperplanes through the origin, which
naturally generate tessellations of the space into polyhedral convex cones and can
be conveniently presented with the help of spherical tessellations generated by great
(.n − 2)-spheres. Different models of random cones, based on random hyperplane
arrangements with certain distributions subject to mild assumptions, are investigated
in Chap. 5, and this study of random cones is also continued in Chap. 6.

Some Selected Research Papers

Le (see [494] and [495]) investigates random point configurations and triangles in
the 2-sphere, namely in view of evaluations of shape densities on the corresponding
shape space (see the book [449] for general background). The author studies
statistical properties of occurring shapes when the used points are generated by
a random mechanism.
Arbeiter and Zähle [38] deal with random mosaics on the .(n − 1)-sphere whose
cells are piecewise smooth. The authors choose the k-dimensional surface area
measure and .k−1 different curvature measures concentrated on the k-skeleton of the
mosaic, .k = 0, ..., n − 1, and they derive relations between the respective densities.
This yields a parameter system, based on which all mean values can be obtained.
Also great hypersphere mosaics are studied.
Among various further results, the authors of [108] derive a precise asymptotic
formula for the volume of spherical regular simplices as the dimension n tends to
infinity.
Sugimoto and Tanemura [762] study the random sequential coverings of congru-
ent spherical caps under the condition that none of them covers the center of another
one. An algorithm for such type of coverings is given.
Gao et al. [290] prove that among all spherically convex bodies K of given
volume, the total measure of great hyperspheres hitting K is minimal if K is
a spherical cap. This isoperimetric type result gives a spherical analogue of the
usual Urysohn inequality (volume against mean width), and it can be seen as
generalization of the Blasche-Santaló inequality, since in this setting for prescribed
volume spherical caps maximize the volume of the polar body. The authors present
two approaches, each of them interesting for itself.
Maehara studies intersection graphs of random spherical caps in [522]. Let m
be a positive integer, and .r := r(m) be a non-negative real number that is at most
1. Now choose m points independently and uniformly at random from the 2-sphere
in 3-space to be centers .C1 , . . . , Cm of the spherical caps of area .4rπ . We call a
graph .Gr the intersection graph of the family .C1 , . . . , Cm when .C1 , . . . , Cm form
174 8 Geometric Probability on the Sphere

its vertices, and the edges of .Gr come to be whenever there is nonempty intersection
between two caps. The author studies various interesting asymptotic properties of
.Gr with m going to infinity.

Pivovarov [638] derives threshold results for expectations of volume ratios

referring to random Gaussian polytopes (being the convex hull of m independent
identically distributed (i.i.d.) Gaussian random vectors) and random polytopes
whose vertex set is taken from the Euclidean sphere. E.g., for such spherical
polytopes and n tending to infinity, a super-exponential result is obtained, and in
the dual sense corresponding results for spherical polytopes generated by random
facets are obtained, too.
Maeda [517] classifies spherical triangles into seven types regarding sides and
angles. He calculates for random spherical triangles the probabilities of occurrence
of these types. As an application, the expectation of numbers of acute triangulations
of a random spherical triangle in the 2-sphere can be estimated. For the spherical
convex hulls .Pm of m independent, uniformly distributed random points in a
halfsphere, Bárány et al. [59] study the asymptotic behaviour of the expectation of
certain characteristics when m tends to infinity. They succeed with explicit integral
expressions for the expected facet number, surface area, and spherical mean width.
Going on, they use for mean volume and vertex number more sophisticated methods.
Finally, the results are compared with analogues for Euclidean space, and interesting
problems are posed.
Results on f -vectors (which express the numbers of k-faces) of so-called Poisson
zero n-polytopes .Z n of homogeneous and isotropic Poisson hyperplane tessellation
in n-space are derived by Kabluchko [435]. More precisely, he gives an explicit
combinatorial formula in terms of a special array of numbers satisfying a recurrence
relation. A result on intrinsic volumes of .Z n has applications for a type of random
spherical polytopes. Namely, the expected f -vector and also the expected volume
of the spherical convex hull of random points chosen uniformly and independently
(again) from the n-dimensional halfsphere can be computed, and an extension of
Sylvester’s four-point problem is obtained, too.
Also Brauchart et al. [126] study the convex hulls of m random points from the
unit sphere of n-space, regarding surface area measure and asymptotics (as m goes
to infinity) for the expected moments of the radii of spherical caps corresponding to
the facets of the obtained polytopes. The authors pose conjectures for the asymptotic
distribution of the scaled radii of these spherical caps, and also for the expected value
of the largest of these radii (the covering radius). Numerical results, also given in
this paper, support these conjectures.
Sylvester’s four-point problem (asking for the probability that four random points
in the plane form a re-entrant quadrilateral) has a long history, which is nicely
described by Pfiefer [636] (see also our Sect. 8.3). Maehara and Martini [534] prove
a spherical extension of Sylvester’s four-point problem: Let X be a finite subset of
the unit (.n − 1)-sphere in Euclidean n-space. This set is called extremal if for every
.x ∈ X there is a hemisphere that has all points of X as interior points, except for x

being on its boundary. The probability that a random sample of .n + 1 points, chosen
uniformly from the unit (.n − 1)-sphere, forms an extremal set is .1 − (n + 2)/2n . In
8.9 Notes 175

important exposition dealing with further extensions of Sylvester’s problem is due

to Bárány [58].
Inspired by results on the asymptotic behaviour of the volume of Euclidean n-
polytopes determined by i.i.d. points uniformly distributed in a convex body, Besau
and Thäle [81] derive analogous properties of random polytopes also in spherical
spaces, and in addition they look at the hyperbolic case and Hilbert spaces (using
there the Busemann and Holmes-Thompson notions of volumes) and dual intrinsic
volumes in Euclidean case. Among other tools, weighted floating bodies are used
in the proofs. Somehow as counterpart to these results, Besau et al. [82] derive
interesting central limit theorems (with n going to infinity) for random n-polytopes
inscribed to the boundary of a smooth convex body. Besides in spherical geometry,
these results also hold in hyperbolic and projective Finsler geometry. This way, a
large family of well-known theorems is extended and unified.
And again we consider, for m random points distributed independently and
uniformly on the unit (.n − 1)-sphere, the convex hull .Pm . Inspired by high
dimensional statistics, Bonnet and O’Reilly [102] study the n-polytopes .Pm when
both numbers m and n tend simultaneously to infinity. Under these assumptions,
they study the asymptotic behaviour of the facet number, of the facet shape and the
facet distances from the origin.

8.9.2 Notes on Crofton’s Formulae

The contribution of M. W. Crofton (1826–1915) to integral geometry is fundamen-

tal. The famous Crofton formula (also called the Cauchy-Crofton formula, see our
Sect. 8.5) in its classical form is a typical basic result in this field. It relates the length
of a planar curve to the expected number of times that a “random” line intersects
this curve (see [191]).

Some History

We refer also to the paper [497] of Lebesgue, the obituary [192] and the historical
articles [489, 490, 682] and [727] on integral geometry and geometric probability,
treating Crofton’s formulae in various ways and with interesting connections to
further geometric fields.
Later extended Crofton-type formulae became one cornerstone of the fast
development of integral geometry that, roughly speaking, started with Blaschke’s
school (1935–1939), Santaló, Chern, as well as Hadwiger, the latter in the middle
of the twentieth century. The corresponding development before this is documented
in the books [212] of Deltheil, [95] of Blaschke, and [448] of Kendall and Moran.
Examples of early publications in this direction, partially also combined with typical
problems from convex geometry, are again Blaschke’s book [95, §7] and his report
[93], as well as Santaló’s papers [686] (containing a corresponding new proof of
176 8 Geometric Probability on the Sphere

the isoperimetric inequality in the plane), [687] (treating spherical curves), and
[688] (referring to surfaces of constant negative curvature with Crofton’s formula
for chords). This is completed by Santaló’s exposition [689] on n-dimensional
spaces of constant curvature containing needed generalizations of the classical
Crofton formula. Also Chern’s contributions [162, Part II] (see below) and [161,
§3] have to be mentioned here, since the corresponding generalizations of Crofton
formulas needed there are discussed in a very detailed way. In [162, Part II] also
the developments around applications, combined with the names of Helgason and
Gelfand, are represented. And Hadwiger’s contributions can, in great parts, be found
in his books [355, §34] and [356, Chapter 6], followed by Stoka’s monograph [761].

Later Basic Books

After his book [690], Santaló presented with great success the standard monograph
[691] on integral geometry and geometric probability, treating Crofton’s formulae
for Euclidean n-space in Chapter 14 (Section 3), and for non-Euclidean n-spaces in
Chapter 18 (Section 1). A strongly related historical article is [692] of Santaló. For
introducing the concept of vectorial integral geometry, extending common integral
geometry to vector-valued functions, we refer to the paper [357] of Hadwiger and
Schneider. Crofton type integral formulae have their natural analogues in this theory.
Since after the appearance of [691] students in China did not have the benefit
of a similar book in Chinese, Ren collected material from related courses to write
finally an excellent monograph. This follows, in its structure of contents, closely
Santaló’s book and contains also several new results; an English translation of this
book is [663]. In Chapter 5 of his book [749], Solomon applies Crofton’s mean
value theorems for a detailed approach to Sylvester’s four-points problem and its
extension to 3-space. After clarifying the notion of valuation and characterizing, in
Chapters 8 and 9 of the monograph [453], the various intrinsic volumes of convex
bodies, Klain and Rota give in Chapter 10 a general Crofton formula for the intrinsic
valuations in all dimensions.
In their basic monograph [706] on stochastic and integral geometry, Schneider
and Weil present Crofton type formulas in Sections 5.1 and 6.4, and in the notes to
Section 5.1 they also mention differential-geometric versions of Crofton formulas
(see also Sulanke and Wintgen [763] and again Santaló [691]). Closer to convex
geometry, in Section 6.4 of [706], Crofton formulas for mixed measures are derived,
first in a translative and then in its kinematic form. These integral formulas for
mixed measures and their global versions (i.e., mixed functionals) lead to kinematic
formulas for certain mixed volumes, projection functions and support functions of
convex bodies. And this yields, also presented in Section 6.4, Crofton type formulas
for projection functions.
Also in books dedicated directly and mainly to convex geometry one can find
various generalizations and specifications of Crofton-type formulas, forming a
central topic which is connected with projection and intersection formulas. As
8.9 Notes 177

examples, we mention here Section 4.4 in Schneider’s monograph [704] (Crofton’s

intersection formulae) and Chapter 5 of the book [412] of Hug and Weil.
In the second of the two volumes dedicated to selected papers and lectures
of S.-S. Chern on differential and integral geometry (see [162, Part II] one finds
eight lectures reflecting the two essential periods of integral geometry, namely
the classical one with Blaschke and Santaló (see above), and the second one with
Gelfand and Helgason, mainly referring to deep applications. Crofton’s theorem is
carefully introduced at the end of the first lecture, and in the second lecture essential
corollaries of it, holding for closed convex curves, are studied. Inspired by this,
further related results are presented, combined also with notions like width of such
curves and kinematic measures (which are needed in further lectures there).

Surveys

The following surveys reflect results on and applications of Crofton-type formulae

in the spirit of convex geometry. The handbook article [707] on integral geometry,
due to Schneider and Wieacker, deals with measures of sets of geometric objects
being invariant regarding an underlying geometric transformation group. The
authors have chosen the direct measure-theoretic approach on the lines of Federer.
Among other intersection formulae given in Section 2, Crofton’s intersection
formula and also a translative Crofton-type formula are treated there. And in the
handbook article [802] of Weil and Wieacker, dedicated to stochastic geometry,
versions of the principal kinematic formula and the Crofton formula are discussed.
Further on, in Section 8 (treating stereology) Crofton formulae are applied to obtain
an expectation formula for the intrinsic volumes of convex bodies.
Schneider [703] connects these two fields (namely integral geometry and
stochastic geometry) in the following sense: he presents essential techniques
coming from integral geometry and being essential for stochastic geometry. He
deduces integral geometric formulae from Hadwiger’s characterization theorem
for the quermassintegrals of convex bodies (see [356]), namely (among others)
the Crofton formula and the principal kinematic formula. The author calculates
integrals with respect to a motion-invariant or Haar measure of the intrinsic volumes
of sections of convex bodies, or of intersections of a convex body with congruent
copies of a second convex body. In the last chapter of [703], generalizations of
Crofton formulae are studied. E.g., these concern extensions to normed spaces, or
to sections of intermediate dimensions.
A nice lecture (or “quick course”) treating integral geometry, still in the spirit of
problems from convex geometry, is given by Zhang [830]. This work shows nicely
how fundamental ideas of Crofton, Poincaré, Blaschke, Santaló and others were
fruitfully extended in modern times. Kiderlen [452] presents a nice and mathemat-
ically rigorous introduction into integral geometry and stereology, starting with a
nice collection of the most important integral-geometric theorems. He also proves
most of them, and he derives, coming from Hadwiger’s characterization of intrinsic
volumes, the Crofton formula together with other basic results. Applications of such
178 8 Geometric Probability on the Sphere

results in modern stereology are presented in the second part. In Section 4 of the
exposition [532], Crofton’s formula (like also Santaló’s chord theorem) are derived
in a convenient way. Here we also mention once more the two comprehensive
expository papers [489] and [490] of Langevin.

Some Selected, Recent Research Papers

Using sections with random planes or projections in random directions, Goodey et

al. [310] study random sets generated from a deterministic n-dimensional convex
body K. The kth Blaschke section body of K with .2 ≤ k ≤ n − 1 is a convex body
whose surface area measure is derived by integrating the surface area measure of
the sections of K over all k-dimensional flats. Analogously, the authors introduce
the notion of kth mean projection body of K, whose support function is the integral
of the support functions of the projections K over all k-dimensional flats. They
continue with establishing relations between these associated bodies; in particular,
a Crofton-type formula for Blaschke sums is obtained.
Curvature measures of sets with positive reach are suitable extensions of classical
indefinite integrals of elementary symmetric functions of principal curvatures from
smooth hypersurfaces to sets with singularities. In this field, Rother and Zähle [679]
derive a kinematic Crofton-type formula relating the kinematic mean of the absolute
curvature measure of a flat section of a set X of positive reach to another absolute
curvature measure of X. Continuing this, Rataj establishes in [659] a more general
related translative and Crofton-type integral formula.
It is well known that Hilbert’s fourth problem asks for the description of all
geometries in which usual line segments are the shortest connections of their
endpoints. E.g., there are the classical examples of Hilbert geometries, and of the
metrics induced by norms. To generalize this concept, Busemann naturally proposed
to study notions of k-dimensional areas in n-dimensional affine spaces in which k-
flats minimize area. Via an integral geometric approach, Crofton measures came
into that framework, and in Schneider’s exposition [701] this topic is presented in
a comprehensive, insightful way. A continuation is the paper [702], where also
the settings of projective Finsler metrics and k-dimensional Holmes-Thompson
area (important in normed spaces) are considered. It turns out that there is a
unique positive measure on the space of lines, yielding a Crofton-type formula
for the (.n − 1)-dimensional Holmes-Thompson area. Schneider [702] presents an
explicit construction of this measure on the space of lines in a polytopal Hilbert
geometry. This yields an integral geometric Crofton-type representation of the
Holmes-Thompson area of hypersurfaces.
The classical Crofton formula determines an integrated intrinsic volume of the
intersection of a given convex body with an invariantly moving k-flat. One way of
generalizing this concept is given by the replacement of intrinsic volumes either
by local counterparts, such as curvature measures, or by certain tensor-valued
valuations, the Minkowski tensors. Hug and Schneider [415] derive a complete
set of integral geometric formulas of Crofton type for the Minkowski tensors of
8.9 Notes 179

convex bodies, involving integrations over Grassmannians. The principal kinematic

formulae for Minkowski tensors can be obtained from Crofton formulae. Among
other tools, for their proofs they use a translative Crofton formula for curvature
measures. In the paper [413], Hug and Weis study versions of the Crofton formula
with a tensor valuation, i.e., a valuation with values in the space of symmetric
tensors. They present an interesting framework, expressing integrals of tensorial
curvature measures via linear combinations of tensorial curvature measures of the
considered convex body. The paper is a continuation of [415], generalizing and
simplifying the Crofton formulas for tensor valuations given there in various ways.
Further papers in this direction are [413] and [414], also presenting a collection
of extensions of the classical Crofton formula for generalized tensorial curvature
measures of convex bodies. The authors prove that the considered system of Crofton
formulas for such generalized measures is complete and cannot be simplified further.
And it is also shown that the generalized tensorial curvature measures, multiplied
with powers of the metric tensor, are linearly independent.
Goodey et al. [311] derive principal kinematic and Crofton-type formulae for
surface area measures of convex bodies. These formulae are related to a localization
of Hadwiger’s integral geometric theorem, and they are successfully applied to get
an extension of Koldobsky’s orthogonality relation for Fourier operators on the
sphere.
Chapter 9
Intersection Graphs of Spherical Caps

You cannot apply mathematics as long as words becloud reality.

— Hermann Weyl

For a family .C = {C1 , C2 , . . . , CN } of spherical caps on .S, its intersection graph

.Ω(C) is defined as the graph whose vertices are the centers .xi of the caps .Ci ,
.i = 1, 2, . . . , N , and two vertices .xi , xj are adjacent if and only if .Ci ∩ Cj /= ∅. We

present a sufficient condition for the connectedness of .Ω(C). Concerning the caps
of the same radius, we specify the family of complete bipartite graphs that can be
represented as the intersection graphs of caps, and also determine the range of radius
necessary for the representation. In the last section, we consider the asymptotic
probability that .Ω(C) becomes connected for a family of random spherical caps
of the same radius .r = r(N) that is a function of the number N of caps.

9.1 Connectivity

Let .C be a family of spherical caps on the unit sphere .S not necessarily of the same
size. We always assume that their centers are all distinct. Let V denote the set of
the centers of the caps in .C. The cap in .C with center .v ∈ V is denoted by .C(v).
The intersection graph .Ω(C) of .C is a graph with vertex set V in which two vertices
.u, v ∈ V are adjacent (u is said to be a neighbor of v) whenever .C(u) ∩ C(v) /= ∅.

For a pair of adjacent vertices .u, v ∈ V of .Ω(C), we regard the geodesic segment
uv on .S as the edge connecting u and v.
Lemma 9.1 For the intersection graph .Ω(C) of a family .C of spherical caps the
following holds: if two vertices .x, y are connected by a path, then they can be
connected by a path that forms a simple polygonal curve on .S.
Proof Among the paths connecting .x, y in .Ω(C), let .v1 v2 . . . vk (.v1 = x, vk = y)
be the one that contains the minimum number of vertices. Suppose that .vi vi+1

© The Author(s), under exclusive license to Springer Nature Switzerland AG 2024 181
H. Maehara, H. Martini, Circles, Spheres and Spherical Geometry,
Birkhäuser Advanced Texts Basler Lehrbücher,
https://doi.org/10.1007/978-3-031-62776-7_9
182 9 Intersection Graphs of Spherical Caps

Fig. 9.1 An extremal cap

.C(v)

C(v)
v
great circle

(and .vj vj +1 cross) each other, where .j > i + 1. Then .(C(vi ) ∪ C(vi+1 )) ∩
C(vj ) ∩ C(vj +1 ) (/= ∅ since it contains ) the intersection point of .vi vi+1 and
.vj vj +1 . If .C(vi ) ∩ C(vj ) ∪ C(vj +1 ) /= ∅, say .C(vi ) ∩ C(vj ) /= ∅, then we
have the( path .v1 . . . vi vj v)j +1 . . . vk containing less vertices,
( a contradiction.
) If
.C(vi ) ∩ C(vj ) ∪ C(vj +1 ) = ∅, then we have .C(vi+1 ) ∩ C(vj ) ∪ C(vj +1 ) /= ∅,

and similarly we can get a path connecting .x, y and containing less vertices, a
contradiction. Hence the path connecting .x, y that contains the smallest number
of vertices forms a simple polygonal curve on .S. ⨆
⨅
A cap .C(u) ∈ C is called isolated if .C(u) ∩ C(v) = ∅ for all .v ∈ V \ {u}. It is
obvious that if .C contains an isolated cap, then .Ω(C) is disconnected.
A cap .C(v) ∈ C is called an extremal cap if there is a great circle passing through
the center v of .C(v) for which all the neighbors of v lie in the same side of the great
circle, allowing that some of them lie on the great circle, see Fig. 9.1. Note that an
isolated cap is also an extremal cap. Note also that if a family .C of caps contains at
most 3 caps, then all caps in .C are extremal caps.
Remark 9.1 If all centers of caps in .C lie in a hemisphere, then .C has an extremal
cap. This can be seen as follows: Let D be the spherical cap of minimum angular
radius .θ that contains all centers of the caps of .C. Since the centers of the caps in .C
are all contained in a hemisphere, we have .θ ≤ π/2, and at least one center v of a
cap in .C lies on the boundary of the cap D. Then the cap .C(v) is an extremal cap
of .C.
Theorem 9.1 Let .C = {C(v) : v ∈ V } be a family of more than 3 spherical caps
on .S, and let .G = Ω(C) be their intersection graph. If .Ω(C) is disconnected, then
.C has an extremal cap.

Proof Let .F, H be two connected components of G, and let .U, W be their vertex
sets, respectively. Then there is a simple closed curve .Γ on .S that separates F from
H , such that .Γ ∩ C(v) = ∅ for all .v ∈ U ∪ W . (For example, such a curve can
be obtained by modifying a certain component of the boundary curve of the region
⋃
.XU := u∈U C(u).) Let x be a point on .Γ . We may suppose that the antipodal point
∗ ∗ ∗
.x of x lies in the same side of .Γ as H , allowing .x ∈ Γ . Thus, .{x, x } ∩ XU = ∅.
∗
Note that then we can connect x and .x by a curve on .S \ XU .
Now, let us regard x as the north-pole, and .x ∗ as the south-pole of .S. Then,
since .{x, x ∗ } ∩ XU = ∅, for every vertex in U , its longitude is well defined. If
9.2 On the Intersection of a Few Equal Caps 183

.u, u' ∈ U are adjacent, then the difference of their longitude is less than .π . (This
can be seen as follows: Suppose that the difference of their longitudes is equal to
' ∗
.π . Then .u, x, u , x lie on a great circle, and since .C(u) ∩ C(u ) =
' / ∅, we must
' ∗ '
have either .x ∈ C(u) ∪ C(u ) or .x ∈ C(u) ∪ C(u ), which contradicts the equality
∗
.{x, x } ∩ XU = ∅.) Therefore, for every pair of adjacent vertices of U we can

tell which vertex lies in the east (or west) of the other, unless they have the same
longitude. Let us call a vertex of U easternmost if it has no neighbor that lies east
of the vertex. Note that if .v ∈ U is the easternmost vertex, then the cap .C(v) is an
extremal cap.
Now, starting from an arbitrary vertex .u ∈ U , we try to go eastward along
the edges of F . If u has a neighbor .u' lying east of u, then we go to .u' . If .u'
has a neighbor .u'' lying east of .u' , then we go to .u'' . In this way we go east and
east. Eventually, we either return to a vertex that once we visited, or we reach an
eastermost vertex and cannot go further. If the former happens, we have a cycle of
F that forms a simple closed curve separating the north-pole x and the south-pole
∗
.x . This is, however, impossible, because there is a curve on .S \ XU that connects

x and .x ∗ , as already noted. Therefore, we finally reach an easternmost vertex, and

hence, there is an extremal cap. ⨆
⨅
Thus, for a family .C of more than 3 spherical caps the following holds:

∃ an isolated cap ⇒ Ω(C) is disconnected ⇒ ∃ an extremal cap.

.

The idea to use “nonexistence of extremal caps” to show the connectedness of the
intersection graph appeared first in [521].

9.2 On the Intersection of a Few Equal Caps

A spherical cap of radius d with center .v ∈ S is denoted by .Cd (v), and its boundary
circle is denoted by .∂Cd (v). For a set .X on .S, the maximum value of the length uv
for .u, v ∈ X is called the diameter of .X . A point-set .X on .S is called d-independent
if the caps .Cd (x), x ∈ X , are all disjoint. Thus, for any two points .u, v in a d-
independent set, the spherical distance uv is greater than 2d. For any three points
.x, y, z ∈ S, we have .xy + yz + zx ≤ 2π . Hence, if .d ≥ π/3, then there is no

d-independent set of three points.

Let .Δ̆xyz be a spherical triangle such that .yz ≥ d, zx ≥ d and the intersection
.Cd (x) ∩ Cd (y) ∩ Cd (z) has interior points. Let .w, w̄ be the intersection points of

.∂Cd (x) and .∂Cd (y). Then, .w = / w̄. (If .w = w̄, then .∂Cd (x) and .∂Cd (y) are tangent
and .x, y, w lie on a great circle. Since .Cd (x) ∩ Cd (y) has interior points, we can
deduce that .Cd (x) ∪ Cd (y) = S. Since z is neither an interior point of .Cd (x),
nor an interior point of .Cd (y), we must have .z = w. But this contradicts the fact
that .x, y, z are vertices of a spherical triangle.) Consulting Fig. 9.2, we can see that
.Cd (x) ∩ Cd (y) ∩ Cd (z) is disconnected if and only if .w, w̄ ∈ Cd (z). If .Cd (x) ∩
184 9 Intersection Graphs of Spherical Caps

Fig. 9.2
.Cd (x)∩ Cd (y) ∩ Cd (z) is
disconnected z
w y

Cd (y)∩Cd (z) is disconnected, it has two components, consisting of two curvilinear

triangles (possibly one is degenerated to a point), and they are separated by the
boundary circle of the circumscribed cap of .Δ̆xyz.
For an equilateral triangle .Δ̆xyz of edge length d, the connected component
of .Cd (x) ∩ Cd (y) ∩ Cd (z) that is contained in the circumscribed cap of .Δ̆xyz is
called the (spherical) Reuleaux d-triangle with vertices .x, y, z, and it is denoted by
.Td (xyz).

Let .Δ̆xyz be an equilateral triangle of edge length d, and p be the center of the
circumscribed cap of .Δ̆xyz, .p∗ be the antipodal point of p. Then .Cd (x) ∩ Cd (y) ∩
Cd (z) is disconnected if and only if .xp∗ ≤ d. If .xp∗ = d, then .x, y, z, p∗ are
the vertices of a regular tetrahedron inscribed in .S. In this case, .d = xy = xp∗
is computed as .cos−1 (−1/3) and the intersection .Cd (x) ∩ Cd (y) ∩ Cd (z) has two
connected components: .{p∗ } and the Reulaeux d-triangle .Td (xyz). Thus the next
lemma follows.
Lemma 9.2 For an equilateral triangle .Δ̆xyz of edge-length d, the intersection
Cd (x) ∩ Cd (y) ∩ Cd (z) is disconnected if and only if
.

. cos−1 (−1/3) ≤ d < cos−1 (−1/2) = 2π/3.

⨆
⨅
Lemma 9.3 The diameter of the Reuleaux d-triangle is d.
Proof Let .Td (xyz) be a Reuleaux d-triangle. We show that for any two points .u, v ∈
Td (xyz), the length uv is at most d. If one of .u, v coincides with a vertex of .Td (xyz),
then .uv ≤ d is obvious. So, we may consider the case that none of .u, v is a vertex of
.Td (xyz), and u lies on the boundary arc x ı y of .Td (xyz). Then .Cd (u) contains .x, y, z,
and .x, y lie in the interior of .Cd (u). Let .y ' be the mirror image of y with respect
to the plane uxO. Then the boundary circles .∂Cd (u) and .∂Cd (x) also intersect at
'
.y . Hence the boundary arc y ız of .Cd (x) lies in the interior of .Cd (u) except for the
endpoint y. Similarly, the boundary arc xı y is contained in .Cd (u). Therefore, v lies
in .Cd (u), and .uv ≤ d. ⨆
⨅
9.2 On the Intersection of a Few Equal Caps 185

Fig. 9.3 Curvilinear z

. ..............................
..........................● ............
........... .. ........ .
triangles wuv and .w̄ ūv̄
...............
......... .. ..● w
... .
.. ... .... .
.........
........
......
....
.. .. .
... . .
... ....
... . . ....
.
.... ... .. . .... ....
.
..... ..... ... . ..... ...
...
... .. .. . .... ...
.... .... ... ... ...
...
...
..
.. w̄
... ●... .
. .......................
...
...
.. . . .............. ... ..
... .... . .. ... ..
..
..
...
.. .. . ..
..
ū .
...... ... ...●.... ... ... ..●
.
..
. v̄ ..
. ... ... ......... ... .
..
..
.. ... ... ... ... ... ... .. . .
.... .
. ... .... . .. .
..
.
.
..... ... ... ... ... ....
.
..
.... .. ... ...... ...
.......... .... ... .. ... .. ....
... .● ...................... .
...............................................................................●
.. ...............
...............● ..................................... ● ....
x
...
...
...
...
...
. u ...
.. ..
.. v ...
..
y ...
..
.... .. . .. .. ....
.... ............ .
.. ................ .. .... ....
....
.... .... ... .. ... ....
..... . .. .. ... .......
......... . .. ..........
......... .
.......... .........
.............. ..........
.................................................

Corollary 9.1 Let .d < 2π/3 and .Δ̆xyz be a spherical triangle such that .d ≤
zx, zy, xy < 2d. Then the diameter of each component of the intersection .Cd (x) ∩
Cd (y) ∩ Cd (z) is at most d.
Proof Let .w, w̄ denote the intersection points of .∂Cd (x) and .∂Cd (y) such that w
and z lie in the same side of the plane Oxy, see Fig. 9.3. Then .Cd (z)∩Cd (x)∩Cd (y)
is disconnected if and only if .w̄ ∈ Cd (z). Let us consider the case that .Cd (z) ∩
Cd (x) ∩ Cd (y) is disconnected.
Denote the intersection points of .∂Cd (x), ∂Cd (y), ∂Cd (z) by .u, v, ū, v̄, as shown
in Fig. 9.3. Then the connected components of .Cd (x)∩Cd (y)∩Cd (z) are curvilinear
triangles wuv and .w̄ ūv̄. Note that .uw < uz = d. Let p be a point on the arc w ¯w̄ū
such that .up = d. Then .Δ̆puy is an equilateral triangle with edge length d, and the
curvilinear triangle wuv is contained in the Reuleaux d-triangle .Td (puy). Hence
the diameter of the curvilinear triangle is at most d.
Similarly, the curvilinear triangle .w̄ ūv̄ is contained in a Reuleaux d-triangle
¯ such that .ūq = d. Hence its diameter
.Td (y ūq), where q is a point on the arc w̄wu

is at most d. ⨆
⨅
Corollary 9.2 Let .d < π/2 and .x, y ∈ S be two points such that .d ≤ xy < 2d.
Then .Cd (x) ∩ Cd (y) cannot contain .(d/2)-independent three points.
Proof Since .Cd (x) ∩ Cd (y) is covered by the union of two Reuleaux d-triangles,
the corollary follows. ⨆
⨅
The name “Reuleaux triangle” comes from convexity describing the most
elementary “curves of constant width” different to the circle. These are special
bodies of constant width, for which we refer to the survey [153] and to the
monograph [558].
186 9 Intersection Graphs of Spherical Caps

9.3 Intersection Graphs of Equal Caps

Recall that a bipartite graph is a graph in which the vertices can be classified into
two groups so that no pair of vertices in the same group is adjacent. A complete
bipartite graph is a bipartite graph in which every vertex in one group is adjacent to
all vertices in the other group. A complete bipartite graph in which one group has
m vertices, and the other group has n vertices, is denoted by .K(m, n). Figure 9.4
shows the complete bipartite graphs .K(2, 3), K(3, 3), K(3, 4). For .n ≥ 3, .K(1, n)
is also called the n-star. Figure 9.5 shows the graphs called 4-star, 5-star, 6-star.
A graph G is said to be a cap graph if it is the intersection graph of a set of
spherical caps of the same radius. If, furthermore, the radius of the caps is d, then G
is called a d-cap graph. For example, every complete graph .Kn is a cap graph, and
a d-cap graph for every .0 < d < π. The complete bipartite graph .K(2, 3) is also
a cap graph. (It is not difficult to find a set of five caps of the same radius whose
intersection graph is .K(2, 3).)
If the vertex set V of a d-cap graph G contains an independent set of k vertices
(that is, no pair in these k vertices is adjacent), then the centers of the caps
corresponding to these k vertices form a d-independent set on .S. If the n-star .K(1, n)
is a d-cap graph, then since .K(1, n) contains n independent vertices, the cap of
radius 2d whose center corresponds to the vertex of degree n must contain a d-
independent set of n points. Since no spherical cap of spherical radius 2d can contain
a d-independent set of six points, .K(1, n) is not a cap graph for .n ≥ 6. However, for
.n ≤ 5, each n-star .K(1, n) is a cap graph. Indeed, for small .d > 0, we can easily

find the set of six caps of radius d whose intersection graph is .K(1, 5). On the other
hand, if d is large, say, .d ≥ π/3, then it is impossible to obtain .K(1, 5) as a d-cap
graph. In the following, we present, for each .n = 2, 3, 4, 5, the bound of d such that
.K(1, n) is a d-cap graph. We also show that among the complete bipartite graphs

.K(m, n), n ≥ m ≥ 2, only .K(2, 2) and .K(2, 3) are cap graphs.

Theorem 9.2 The n-star .K(1, n), n ≤ 5, is a d-cap graph if and only if .d < ρn ,
where .ρn = π/n for .2 ≤ n ≤ 4, and

1 Ä √ ä
ρ5 =
. cos−1 1/ 5 ≈ 0.553575.
2

Fig. 9.4 .K(2, 3), K(3, 3)

and .K(3, 4)

Fig. 9.5 The stars

.K(1, 4), K(1, 5), K(1, 6)
9.3 Intersection Graphs of Equal Caps 187

Proof
(i) Since any two hemispheres have nonempty intersection, .K(1, 2) is not a d-cap
graph for .d ≥ π/2. If .0 < d < π/2, then a 2d-cap contains d-independent
two points. Hence .ρ2 = π/2.
(ii) For any three points on .S, the minimum distance among the three points is
less than or equal to .2π/3. Hence, .ρ3 ≤ π/3. If .d < π/3, we can take a d-
independent set of three points on the boundary of a cap of radius 2d. Hence
.K(1, 3) is the intersection graph of caps of radius d. Therefore, .ρ3 = π/3.

(iii) For .π/4 ≤ d < π/3 and .z ∈ S, the cap .C2d (z) cannot contain d-
independent four points. To see this, divide .C2d (z) into four equal “sectors”
by four geodesic segments .l1 , l2 , l3 , l4 of length 2d emanating from z. Let .l1 =
zx, l2 = zy, and let .S(z, xy) denote the sector bounded by .l1 = zx, l2 = zy
and a minor circular arc xı y of .∂C2d (z).
First, we show that the diameter of .S(z, xy) is 2d. If .d = π/4, then
.S(z, xy) is a Reuleaux 2d-triangle, and its diameter is 2d. So, we suppose

that .2d > π/2. Let u be the point on xz such that .zu = π/2 and let
G be the great circle with a pole u. Then .l2 is an arc of G. Note that G
and .∂C2d (x) are tangent at z. Since the area of .Δ̆zxy is greater than the
area of .Δ̆(π/2, π/2, π/2), we must have ./ zxy > π/2 = / xzy. Therefore,
∗
.xy < zy = 2d (cf. Exercise .5.2 ). Thus y is an interior point of .C2d (x).

Since G and .∂C2d (x) are tangent at z, it follows that .l2 (= yz) is contained in
.C2d (x). Let v be the point on .∂C2d (z) such that .xv = 2d and x ıv contains y.
Then the Reuleaux 2d-triangle .T2d (zxv) contains .S(z, xy). Thus the diameter
of .S(z, xy) = 2d. Now suppose that .C2d (z) contains d-independent four points
.w1 , w2 , w3 , w4 . We may assume that .w1 lies on .l1 = zx. Then the two sectors

having .l1 (= zx) in common cannot contain .w2 , w3 , w4 . Hence these three
points lie on the union of remaining two sectors, which implies that some two
points of .w2 , w3 , w4 are contained in a sector, contradicting the fact that the
diameter of each sector is 2d. Thus, .C2d (z) cannot contain d-independent four
points. Therefore, .ρ4 ≤ π/4.
For .0 < d < π/4 and a point .z ∈ S, let .u1 , u2 , u3 , u4 be four points (in
this cyclic order) on the boundary circle of .C2d (z) such that .u1 u2 = u2 u3 =
u3 u4 = u4 u1 . Since the isosceles triangle .Δ̆zu1 u2 is a part of an equilateral
triangle of side-length .π/2 (and interior angle .π/2), its area is smaller than
the area .π/2 of an equilateral triangle with interior angle .π/2. Then, since
./ u1 zu2 = π/2, we must have ./ zu1 u2 = / zu2 u1 < π/2 = / u1 zu2 . This

implies that .u1 u2 > zu1 = 2d. Thus, .u1 u2 = u2 u3 = u3 u4 = u4 u1 > 2d.
Therefore, the four caps .Cd (ui ), i = 1, 2, 3, 4, are disjoint, and the intersection
graph of these four caps and .Cd (z) is the 4-star .K(1, 4). Therefore, .ρ4 = π/4.
(iv) Since .ρ4 = π/4, we have .ρ5 ≤ π/4. For a point .z ∈ S and a radius .d, 0 < d <
π/4, let .v1 , v2 , v3 , v4 , v5 be five points in this cyclic order on the boundary
of .C2d (z) that divide the boundary circle into five equal arcs. Then ./ v1 zv2 =
2π/5.
188 9 Intersection Graphs of Spherical Caps

Fig. 9.6 The sector

.T= zv1 v2
T
v1 v2

First, we show that .C2d (z) contains a d-independent set of five points if and
only if .v1 v2 > zv1 = 2d. Let T be the sector obtained as the intersection
of .C2d (z) and the lune with the interior angle ./ v1 zv2 , see Fig. 9.6. Note that
the boundary arc v¯ 1 v2 of the sector T is not a geodesic segment .v1 v2 . Since
.Δ̆zv1 v2 ⊂ T , the interior angle of .Δ̆zv1 v2 at .v1 (and at .v2 ) is less than .π/2.

Note that the diameter of the sector T is .s := max{2d, v1 v2 }. (To see this, let
.p, q be two points on the boundary of T . If both .p, q lie on the boundary curve

v¯1 v2 , then .pq ≤ v1 v2 . If p lies on .zv1 , then .Cs (p) contains T , and hence .pq ≤
s.) Suppose that the cap .C2d (z) contains a d-independent set .X consisting of
five points. We may assume that one point .y0 ∈ X lies on the segment .zv1 .
Then one of the five sectors .zv1 v2 , zv2 v3 , zv3 v4 , zv4 v5 , zv5 v1 must contain
two points of .X . (If neither of the sectors .zv1 v2 , zv5 v1 contains two points of
.X , then the four points in .X \ {y0 } must lie in the union of the three sectors

.zv2 v3 , zv3 v4 , zv4 v5 , which implies that one of the three sectors contains some

two points of .X \ {y0 }.) Hence the diameter of each sector must be greater
than 2d. Since the diameter of T is .max{2d, v1 v2 }, we have .v1 v2 > 2d. Thus,
.v1 v2 = v2 v3 = · · · = v5 v1 > 2d. On the other hand, if .v1 v2 > 2d, then

.{v1 , v2 , v3 , v4 , v5 } is a d-independent set in .C2d (z). Thus, .C2d (z) contains a

d-independent set of five points if and only if .v1 v2 > 2d.

Now, we could ask: For what value .δ of d does .v1 v2 = 2d hold? If .zv1 =
v1 v2 = 2δ, then, by the spherical cosine law, we have

. cos(2δ) = cos2 (2δ) + sin2 (2δ) cos(2π/5),

from which we get

cos(2π/5)
. cos(2δ) = .
1 − cos(2π/5)
√
Since .cos(2π/5) = 14 ( 5 − 1), we have
√
cos(2π/5) 5−1 1
. cos(2δ) = = √ =√ ,
1 − cos(2π/5) 5− 5 5
Ä √ ä
and hence .δ = 12 cos−1 1/ 5 . If .d = δ, then .Δ̆zv1 v2 is an equilateral
triangle with interior angle .2π/5. If .d < δ, then .Δ̆zv1 v2 is an isosceles triangle
9.3 Intersection Graphs of Equal Caps 189

whose interior angle at the vertex z is .2π/5, contained in an equilateral triangle

with interior angle .2π/5. Hence it has smaller area than the equilateral triangle,
and hence ./ zv1 v2 < 2π/5, which implies .v1 v2 > zv1 . In this case, the
intersection graph of the family .{Cd (z)}∪{Cd (vi ), i = 1, 2, 3, 4, 5} is .K(1, 5).
Therefore, .K(1, 5) is a d-cap graph if and only if .d < δ. Thus .ρ5 = δ.
⨆
⨅
Theorem 9.3
(1) .K(2, 2) is a d-cap graph if and only if .0 < d < π/2.
.

(2) .K(2, 3) is a d-cap graph if and only if .π/4 ≤ d < π/3.

.

Proof
(1) The proof is easy and omitted.
(2) If .K(2, 3) is a d-cap graph, then .K(1, 3) is also a d-cap graph. Hence .K(2, 3)
is not a d-cap graph for .d ≥ π/3.
Suppose that .π/4 ≤ d < π/3. Let G be a great circle on .S, and .x, x ∗ be the
poles of G. On the great circle G, we can take three points .u, v, w such that the
three caps .Cd (u), Cd (v), Cd (w) are disjoint. Then these three caps and the caps
∗
.Cd (x), Cd (x ) constitute together a family of r-caps whose intersection graph

is .K(2, 3).
Next, suppose that .d < π/4. Let let .x, y ∈ S be a pair of points such that
.Cd (x) ∩ Cd (y) = ∅. Then .2d < π/2, and hence .C2d (x) ∩ C2d (y) cannot

contain a d-independent set of three points by Cororally 9.2. Hence .K(2, 3) is

not a d-cap graph.
Therefore, .K(2, 3) is a d-cap graph if and only if .π/4 ≤ d < π/3. ⨆
⨅
Theorem 9.4
(1) .K(2, 4) is not a cap graph.
.

(2) The complete bipartite graph .K(3, 3) is not a cap graph.

.

Proof
(1) The proof is easy and omitted.
(2) If .K(3, 3) is a d-cap graph, then .K(2, 3) is also a d-cap graph. So, it is enough
to show that for .π/4 ≤ d < π/3, .K(3, 3) is not a d-cap graph. Let .{x, y, z} ⊂ S
be a d-independent set. Then .C2d (x) ∩ C2d (y) ∩ C2d (z) consists of at most two
components, and each component has diameter at most 2d by Cororally 9.1.
Hence it cannot contain a d-independent set of three points. Therefore, .K(3, 3)
is not a cap graph.
⨆
⨅
Remark 9.2 The results in this section are obtained in [540].
190 9 Intersection Graphs of Spherical Caps

9.4 Random Spherical Caps

If we put N spherical caps of the same angular radius .θ < π/2 at random on .S (that
is, the centers of the N caps are random points on .S chosen independently), what
is the probability that the union of these N caps is connected? Let us estimate this
probability using Theorem 9.1.
Lemma 9.4 Let .C1 , C2 , . . . , CN be N spherical caps of angular radius .θ < π/2
put at random on .S. Under the condition that .CN intersects only the k caps
.C1 , C2 , . . . , Ck , the conditional probability that .CN is an extremal cap of the family

of N caps is equal to .2k/2k .

Proof Let .vi denote the center of .Ci , i = 1, 2, . . . , N . We may suppose that .vN is
a fixed point on .S. Let D denote the cap of angular radius .2θ centered at .vN . Then,
.Ci ∩ CN /= ∅ is equivalent to .vi ∈ D. Hence, under the condition that .CN intersects

.C1 , C2 , . . . , Ck , the centers .v1 , v2 , . . . , vk are independent random points uniformly

distributed in D. Since the probability that some of .vi , i = 1, 2, . . . k, coincide with

.vN is 0, we may assume that .v1 , . . . , vk are distributed uniformly in .D \ {vN }. For
'
.i = 1, 2, . . . , k, let .v denote the intersection point of the boundary circle .∂D of D
i
and the extension of the geodesic segment .vN vi beyond .vi . Then .vi' , i = 1, 2, . . . vk'
are independently and uniformly distributed random points on the circle .∂D. Under
the condition that .CN intersects only the k caps .C1 , C2 , . . . , Ck , the following two
events (1) and (2) are equivalent:
(1) .CN is an extremal cap,
(2) .v1' , v2' , . . . , vk' lie on a semicircle of .∂D.
The probability that the event (2) occurs is already computed as .2k/2k in
Theorem 8.5.
⨆
⨅
Lemma 9.5 Let .{C1 , C2 , . . . , CN } be a family of N spherical caps of the same
angular radius .θ < π/2 put at random on .S. Then the probability that .Ci is an
extremal cap of the family is given by
Ä äN −2
(cos2 θ )N −1 + (N − 1) sin2 θ 1 −
.
1
2 sin2 θ .

Proof Let .vi denote the center of .Ci , and let .Di denote the spherical cap of angular
radius .2θ with center .vi . The area of .Di is

2π(1 − cos 2θ ) = 4π sin2 θ.

.

Hence, for a random point v on .S, .Pr(v ∈ Di ) = sin2 θ . The probability that
a certain k-tuple of points from .v1 , v2 , . . . , vN (other than .vi ) lies in .Di and the
remaining .N − 1 − k points lie outside .Di is

(sin2 θ )k (1 − sin2 θ )N −1−k = (sin2 θ )k (cos2 θ )N −1−k .

.
9.4 Random Spherical Caps 191

Hence, by Lemma 9.4,

. Pr(Ci is an extremal cap)

⎲
N −1 Ç å
N −1 N −1 2k
= (cos θ )
2
+ (sin2 θ )k (cos2 θ )N −1−k k
k 2
k=1
−1 Ç å Ç 2 åk
⎲
N
N −1 sin θ
N −1
= (cos θ )
2
+2 k (cos2 θ )N −1−k .
k 2
k=0

∑ ( )
Since . nk=0 nk kx k y n−k = x · ∂
∂x (x + y)n = nx(x + y)n−1 , we have
Ä äN −2
. Pr(Ci is extremal) = (cos2 θ )N −1 + (N − 1) sin2 θ sin2 θ + cos2 θ
1
2
Ä äN −2
= (cos2 θ )N −1 + (N − 1) sin2 θ 1 − 12 sin2 θ .

⨆
⨅
Theorem 9.5 Let .P (θ, N) denote the probability that the intersection graph of the
family .C of N spherical caps of the same angular radius .θ < π/2 put at random on
.S is connected. We have

P (θ, N ) ≥ 1 − N(cos2 θ )N −1 − N(N − 1)(sin2 θ )(1 −

.
1
2 sin 2θ )N −2 .

Proof By Lemma 9.5, the probability that one of the caps in .C is an extremal cap is
at most
1 2 N −2
N (cos2 θ )N −1 + N(N − 1) sin2 θ (1 −
. sin θ ) .
2
Hence the probability that .C contains no extremal cap is at least

1 2 N −2
1 − N(cos2 θ )N −1 − N(N − 1) sin2 θ (1 −
. sin θ ) .
2
By Theorem 9.1, the property that “there is no extremal cap” implies that “the
intersection graph of the family is connected”. Hence .P (θ, N ) is greater than or
equal to the probability that .C contains no extremal cap. Therefore,

1 2 N −2
P (θ, N ) ≥ 1 − N(cos2 θ )N −1 − N(N − 1) sin2 θ (1 −
. sin θ ) .
2
⨆
⨅
192 9 Intersection Graphs of Spherical Caps

9.5 Chebyshev’s Inequality and a Lemma

From Theorem 9.5 we can derive some non-trivial lower bounds of .P (θ, N) for
some .θ, N. For example,

P ( π3 , 15) > 0.65, P ( π6 , 60) > 0.61, P ( 10

.
π
, 200) > 0.76.

Although these lower bounds look very rough, Theorem 9.5 can be still applied
to derive a reasonably good asymptotic result. Preceding to derive this asymptotic
result, let us recall here Chebyshev’s inequality and prepare a useful lemma.
Let X be a ⎰nonnegative random variable, and .f (x) be its density function. Thus,
a
.Pr(X ≤ a) =
0 f (x)dx. Let .μ = E(X) > 0 be the mean of X. Since
⎰ ∞ ⎰ ∞ ⎰ ∞
μ=
. xf (x)dx ≥ xf (x)dx ≥ tμf (x)dx = tμ Pr(X ≥ tμ),
0 tμ tμ

for every .t > 0, we have

1
. Pr(X ≥ tE(X)) ≤ .
t
This is known as Markov’s inequality. Let Y be a random variable with mean .μY
and variance .σ 2 . Let .X = (Y − μY )2 . Then X is nonnegative, and .E(X) = σ 2 . In
Markov’s inequality, put .t = k 2 /σ 2 . Then .Pr(X ≥ k 2 ) ≤ σ 2 /k 2 . Hence

σ2
. Pr(|Y − μY | ≥ k) ≤ .
k2
This is Chebyshev’s inequality.
Lemma 9.6 Let .f = f (N ), g = g(N ) be two positive functions of N. If .f → 0 as
.N → ∞, then for sufficiently large N we have .(1 − f )g < e−f ·g . If, furthermore,
.f g → 0 as .N → ∞, then
2

(1 − f )g = e−f ·g (1 + o(1)),
.

where .o(1) denotes some function of N that tends to 0 as .N → ∞.

−1 −1
Proof Put .h(x) = log(1 − x). Then .h' (x) = ''
1−x , h (x) = (1−x)2 . Hence, for
0 < t < 1, we have .h(t) = h(0) + h' (0)t +
.
1 '' 2
2 h (ξ t)t for some .0 < ξ < 1, by
Taylor’s theorem. Therefore

−t 2
. log(1 − t) = −t + .
2(1 − ξ t)2
9.6 Asymptotic Probability 193

Hence, for large N (such that .f = f (N) < 1), we have

f2 · g
g · log(1 − f ) = −f · g −
. < −f · g.
2(1 − ξf )2

Thus, .(1 − f )g < e−f ·g . If, furthermore, .f 2 · g → 0 as .N → ∞, then

f2 · g
. = o(1),
2(1 − ξf )2

and therefore we have

(1 − f )g = e−f ·g+o(1) = e−f ·g · eo(1) = e−f ·g (1 + o(1)).

.

⨆
⨅

9.6 Asymptotic Probability

»
Theorem 9.6 Let θ = c 1
N log N, c > 0. Then
⌠
0 (if c < 1 )
. lim P (θ, N) = √
N →∞ 1 (if c > 2).
√
Proof First, suppose that c > 2. To show P (θ, N ) → 1 as N → ∞, it is enough
to verify, by Theorem 9.5, that

N(cos2 θ )N −1 + N(N − 1)(sin2 θ )(1 −

.
1
2 sin 2θ )N −2 → 0
»
as N → ∞. Since θ = c 1
N log N → 0 as N → ∞, we have sin2 θ = θ 2 (1 +
2
o(1)) = (1 + o(1)) cN log N . Hence, applying Lemma 9.6,

N(1 − sin2 θ )N −1 + N(N − 1)(sin2 θ )(1 −

.
1
2 sin 2θ )N −2

< N e−(N −1) sin sin2 θ )N −2

2θ
+ N(N − 1)(sin2 θ )(1 − 1
2

1 c2 log N
∼ N e−c + N(c2 log N)e−(c
2 log N 2 /2) log N
∼ + ,
N c2 −1 Nc
2 /2−1
194 9 Intersection Graphs of Spherical Caps

where the notation f ∼ g implies f = (1 + o(1))g. Since c2 > 2, we have

1 c2 log N
. 2 −1 + 2 /2−1 →0
Nc Nc
as N → ∞.
Now, suppose that c < 1. We have to show that P (θ, N ) → 0 as N → ∞.
Let C = {C1 , C2 , . . . , CN }, and let vi be the center of Ci . Then v1 , v2 , . . . , vN
are independent and uniformly distributed random points on S. Denote by Di the
spherical cap of angular radius 2θ with center vi . The area of Di is 4π sin2 θ . Since
Ci ∩ Cj /= ∅ ⇔ vi ∈ Di , we have

. Pr(Ci ∩ Cj /= ∅) = sin2 θ ∼ θ 2 .

Similarly,

. Pr(Di ∩ Dj /= ∅) = sin2 (2θ ) ∼ 4θ 2 .

Recall that Ci is an isolated cap of C if Ci ∩Cj = ∅ for all j /= i. If C has an isolated

cap, then Ω(C) is disconnected. For each i define the random variable Zi by
⌠
1 if Ci is an isolated cap
Zi =
.
0 otherwise.

Then the sum Z = Z1 + Z2 + · · · + ZN counts the number of isolated caps in C.

Since N (θ 2 )2 = o(1), we have by Lemma 9.6 that the expected value of Zi is

E(Zi ) = Pr(Zi = 1) ∼ (1 − θ 2 )N −1 ∼ e−N θ ∼ e−c = N −c .

2 2 log N 2
.

Hence the expected value of Z is

2
.E(Z) = NE(Z1 ) ∼ N 1−c .

Next, let us find the expected value E(Zi Zj ) for i /= j . Since E(Zi Zj ) =
Pr(Zi Zj = 1), we have

E(Zi Zj ) < Pr(Di ∩ Dj = ∅) Pr(vk /∈ Di ∪ Dj for k /= i, j | Di ∩ Dj = ∅)

.

+ Pr(Di ∩ Dj /= ∅) Pr(vk /∈ Di ∪ Dj for k /= i, j | Di ∩ Dj /= ∅)

< (1 − 4θ 2 )(1 − 2θ 2 )N −2 + 4θ 2 (1 − θ 2 )N −2

∼ (1 − 4θ 2 )e−2(N −2)θ + 4θ 2 e−(N −2)θ

2 2
9.7 Exercises 195

∼ (1 − 4θ 2 )N −2c + (4c2 log N)N −(1+c

2 2)

= N −2c (1 − 4θ 2 + (4c2 log N)N c

2 −1
) ∼ N −2c .
2 2

Hence
⎲ ⎲ ⎲
E(Z 2 ) =
. E(Zi Zj ) = E(Zi2 ) + E(Zi Zj )
i,j i i/=j

= N · E(Z1 ) + N(N − 1) · E(Z1 Z2 )

Å ã
−c2 −2c2 2(1−c2 ) 1
<N ·N +N ·N 2
=N 2 +1
N 1−c
2
∼ N 2(1−c ) ∼ E(Z)2 .

Since E(Z 2 ) ≥ E(Z)2 , we have E(Z 2 ) ∼ E(Z)2 . Note that Z = 0 implies that
|Z − E(Z)| ≥ E(Z) and the variance σZ2 of Z is equal to E(Z 2 ) − E(Z)2 . Now,
applying Chebyshev’s inequality, we have

E(Z 2 ) − E(Z)2
. Pr(Z = 0) ≤ Pr(|Z − E(Z)| ≥ E(Z)) < → 0.
E(Z)2

Thus, Pr(Z ≥ 1) → 1 (N → ∞), that is, the probability that there is an isolated
cap tends to 1 as N → ∞. Therefore, P (θ, N) → 0 as N → ∞. ⨆
⨅
For more results on intersection graphs of spherical caps see [522].

9.7 Exercises

9.1 Prove that for a regular tetrahedron wxyz, the angle / xOy is given by
cos−1 (−1/3), where O is the center of the circumsphere of the regular
tetrahedron.
9.2∗ Prove that K(2, 2) is a d-cap graph if and only if 0 < d < π/2.
9.3∗ Prove that K(2, 4) is not a cap graph.
9.4 Let X be a random variable with mean μ = 2 and variance σ 2 = 3. Using
Chebyshev’s inequality, estimate the lower bound of

. Pr(−2 < X < 6).

9.5 Find the probability that two random caps of angular radius r < π/2 put
independently on the unit sphere S have nonempty intersection.
9.6∗ Find the expected value of the area of the intersection of two random caps of
angular radius θ < π/2 put independently on the unit sphere S.
196 9 Intersection Graphs of Spherical Caps

9.7∗ For n spherical caps of area 4πp put at random on S, let α(n, p) denote the
area of the part of S that is not covered by any of these n spherical caps. Put

α(n, p)
U (n, p) =
. .
4π

Then U (n, p) is a random variable. Prove that the expected value E(U (n, p))
of U (n, p) is given by

E(U (n, p)) = (1 − p)n .

.

9.8 Notes

The most substantial lists of references related thematically to our Chap. 9 refer to
intersection graphs as well as Markov’s and Chebyshev’s inequality. Hence these
themes are chosen as corresponding topics of the two subsections presented here.

9.8.1 Notes on Intersection Graphs

For our purpose here we want to define the intersection graph of a set family F
as the graph whose vertices are presented by the sets from F , and adjacency holds
if and only if the corresponding sets intersect. E.g., a well known result of Tietze
from 1905 says that any finite graph is the intersection graph of convex 3-polytopes;
this is no longer true (in any fixed dimension) if the polytopes have only a bounded
number of faces, or if instead of them simple geometric figures are taken which
can be described in terms of a bounded number of real parameters. It turns out that
intersection graphs of various classes of geometric objects have, even in the plane,
interesting structural and extremal properties. We want to reflect some essential
survey-like publications from this lively subfield of graph theory, and we note that
the graph-theoretic methods discussed here are also interesting in view of other
chapters in our book (see, e.g., Chap. 3). Since there are too many related research
articles, we refrain from quoting such papers.

Books and Book Parts

In his book [670], Roberts describes how useful graph theory (also combined with
other fields) can be to solve complex problems in our daily life, e.g. referring to
urban services, energy, pollution problems and many other branches. In Chap. 3 the
particular importance of intersection graphs within this framework is established.
9.8 Notes 197

A book which is completely dedicated to intersection graphs of set families

is the monograph [564] of McKee and McMorris. The authors treat general
intersection representations as well as chordal, interval and competition graphs,
clearly showing the common underlying theory and offering also the role that
properties of hypergraphs can play within this setting. Chapter 5 is particularly
devoted to threshold graphs which have additional properties. Although presenting
about 500 references, the authors state that this list is far from being complete.
The comprehensive, encyclopedic work [123] of Brandstädt et al. contains also
a part on intersection graphs, namely the whole Chap. 4 treating such graphs with
respect to combinatorial objects (e.g., as line graphs, interval graphs and visibility
graphs of polygons). A forerunner of [123] is the book [308] of Golumbic whose
second edition we cite here. Also this book contains a lot of material on intersection
graphs and subcases thereof.

Surveys

In Sect. 8.1 of his important handbook article [233], Eckhoff discusses intersection
graphs in combinatorial convex geometry. So the studied set families consist mainly
of (special types of) convex sets, and the subcases of interval graphs and circular-
arc graphs or properties like boxicity and sphericity occur there. Kostochka [476]
surveys the state of the art regarding bounds on the chromatic number of the
intersection graphs of types of geometric figures (such as, e.g., boxes, translates
of convex compacta, or circle chords) with given clique number or girth. A few new
results in that direction are also derived.
Circular-arc graphs (i.e., intersection graphs of open arcs on a circle) and
circle graphs (the intersection graphs of chords on a circle) were comprehensively
studied by many authors, and there are various interesting subclasses. Dúran et
al. [231] give a structural overview and pose interesting problems. Pach [609]
gives an overview to results and problems around intersection graphs and the more
general notion of intersection patterns. Many of the discussed problems go back to
famous mathematicians from convexity, graph theory, combinatorics and topology
(e.g., treating Helly-type theorems or Ramsey theory). This survey takes the
rapid development of computational geometry and algorithms in graph theory into
account. In particular, the author presents also applications of results on intersection
graphs and patterns in the theory of topological graphs. Based on a broad spectrum
of mathematical structures and methods, Ellis [242] surveys intersection problems
in extremal combinatorics. These structures include families of vector subspaces,
subsets of finite group actions and also graphs and hypergraphs, combined with
stronger or weaker intersection conditions. And the used methods, which are also
presented, can be purely combinatorial, algebraic, analytic, and probabilistic in
nature. The author treats old and new problems and methods, and he gives elegant
proofs of several central results. This interesting survey is useful for graduates and
researchers.
198 9 Intersection Graphs of Spherical Caps

More algebraic than the definitions of intersection graphs above, Chakrabarty

and Kureethara [155] denote by .L(R) the set of all non-trivial left ideals of a ring R
and define the intersection graph of ideals of a ring R as an undirected simple graph
.G(R) whose vertices are in one-to-one correspondence with .L(R), and whose edges

come to be iff the corresponding left ideals of R have a non-zero intersection. The
comprehensive research on this notion and on properties of .G(R) gives rise to this
survey.

9.8.2 Notes on Chebyshev’s Inequality

Chebyshev’s inequality in probability theory (also called Bienaymé-Chebyshev

inequality) gives an upper bound on the probability of deviation of a random variable
(of finite variance) from its mean. It can be used to prove the weak law of large
numbers, and we use it here in Sect. 9.5. Also the name “Chebyshev’s theorem” is
common, and this can also refer to the closely related Markov inequality, which is
perhaps more used in analysis. Below we collect some basic references related to
Chebyshev’s famous inequality.

Book Parts

Due to its high applicability, Chebyshev’s inequality can be found in various

monographs on probability theory and statistics, but also from other fields. We
present a little collection, mainly in chronological order.
Johnson and Leone [431] present the Chebyshev inequality in their introductory
part of Volume I, treating as groundwork the themes probability and distribution.
Tuckwell discusses in [783] applications of probability. Chapter 6 is dedicated
to sequences of random variables and develops also convergence in distribution
and convergence in probability. As main tools for this, characteristic functions and
Chebyshev’s inequality are treated. In Jiang’s monograph [429] one can find the
Chebyshev inequality in Chapter 5, which collects many numerical and probability
inequalities. In Example 5.6 an exponential probability inequality for sums of
independent random variables is nicely derived from the Chebyshev inequality.
Epps [243, Chapter 3] treats the topics of integration in abstract spaces and
mathematical expectation, leading to classical inequalities like the Chebyshev one.
McKean [563] gives a broad view onto classical limit theorems in probability
and their applications. As one of the cornerstones in this framework, Chebyshev’s
inequality is presented in the opening chapter. And in Chap. 2 the Markov property
and extensions thereof are discussed.
In his basic monograph [50] on measure theory, integration and real analysis,
Axler dedicates the last chapter to probability theory, introducing a random variable
as a measurable function on a probability space, and its expected value as its
Lebesgue integral with respect to the probability measure. It is shown that the
9.8 Notes 199

expected value of a product of independent random variables is the product of their

expected values. And also the Chebyshev inequality is presented, leading to the
weak law of large numbers. The book [438] of Kale and Muralidharan, treating the
classical theory of statistical inference, is an example for the following situation:
the author assumes that the reader is familiar with various basic notions and tools,
among them also Chebyshev’s inequality.

Surveys and Expositions

Here we start with the nice classroom notes [450] of Khatri presenting several
elementary inequalities of multivariate Chebyshev type. They are derived in an
elementary way, in the spirit of Exercise 6 (p. 256) of the classical monograph [187]
of Cramer.
Tong [778] treats the fact that notions like association and dependence of random
variables, rearrangements, and heterogeneity via majorization ordering are (directly
or indirectly) fundamental for developments around stochastic inequalities. It turns
out that these notions are strongly related to Chebyshev’s inequality, the Hardy-
Littlewood-Pólya rearrangement inequality, and Schur functions from analysis. So,
in particular, Chebyshev’s inequality says that the integral of the product of two
increasing functions is larger than or equal to the product of their integrals. This can
be viewed as confirmation of the positive dependence of random variables which are
monotone functions of a common random variable. In this nice expository paper,
Tong wants to emphasize how classical mathematical inequalities of the above
described types play a central role for stochastic inequalities.
And in the very interesting and detailed paper [726] of Seneta, the history of the
law of large numbers is presented. In a successful way, the Chebyshev inequality
is characterized as meeting point of French and Russian research efforts. Especially
interesting is the presentation of the contributions from the Russian side (besides
Chebyshev meaning particularly also Markov and Bernstein).

Selected Research Papers

Huber [407] studies the equality cases of the classical Markov and Chebyshev
inequalities, which sharply bound the tail probabilities of random variables. Using
a simple smoothing via auxiliary randomness, he shows that each of the obtained
bounds can be halved, and that in many cases this halving procedure can be reached
without using such auxiliary randomness.
Ogasawara [600] derives multivariate versions of the sharp Markov and Cheby-
shev inequality, when associated probabilities are extended to segments of the
supports of non-negative random variables. The probabilities take echelon forms,
and when some positive lower bounds of these probabilities are reached, then
improvements without the echelon forms are attained. Saw et al. [695] study the
200 9 Intersection Graphs of Spherical Caps

necessary (and finally simple) modifications of the Chebyshev inequality for the
case that the studied population mean and variance are estimated from a sample.
Based on operations called pseudo-addition and pseudo-multiplication, Agahi
et al. [9] define so-called pseudo-integrals which are more general than Lebesgue
integrals. Generalizing the classical Chebyshev inequality in probability theory, they
establish Chebyshev-type inequalities combined with two classes of such pseudo-
integrals. Interested in the multivariate version of Chebyshev’s in non-additive
cases, Agahi [8] shows that the (multivariate) Chebyshev inequality still holds in
a generalized probability theory based on the notion of Choquet integral.
And in the three short papers [645] of Prakasa Rao, [831] of Zhou and Hu,
and [132] of Budny, Chebyshev’s inequality is suitably extended to Hilbert-space-
valued random and Banach-space-valued random elements.
Chapter 10
Quartets on a Sphere

Doubt is the origin of wisdom.

— Rene Descartes

For any region on a sphere, it is impossible to construct a (plane) map of the region
in which distances are faithfully represented. This fact can be proved by showing
that any given region on a sphere contains a quartet (i.e., a four-point space) that is
not isometrically embeddable in the plane. Actually, any region on a sphere contains
a quartet whose six distances determine the radius of the sphere uniquely. Moreover,
we present an equation to judge if a given quartet can be a quartet on a sphere of a
suitable radius.

10.1 Metric Spaces

A nonempty set .X is called a metric space if for any two elements .P , Q ∈ X a real
number .d(P , Q) is assigned so that the following three conditions hold:
(i) .d(P , Q) = d(Q, P ) ≥ 0,
(ii) .d(P , Q) = 0 ⇔ P = Q,
(iii) .d(P , Q) + d(Q, R) ≥ d(R, P ) (triangle inequality).
The real number .d(P , Q) is called the distance between P and Q. Elements of a
metric space are called points.
For example, the plane .R2 and the 3-dimensional space .R3 are metric spaces by
the usual distances. A sphere is also a metric space by the spherical distance. A
metric space consisting of three points is called a trio, and a metric space consisting
of four points is called a quartet. Every nonempty subset of a metric space .X is a
metric space by the distance defined in .X , and it is called the subspace of .X . Two
metric spaces .X and .Y are called isometric, if there is a one-to-one correspondence
between them such that for any two points in .X , the distance between them is equal

© The Author(s), under exclusive license to Springer Nature Switzerland AG 2024 201
H. Maehara, H. Martini, Circles, Spheres and Spherical Geometry,
Birkhäuser Advanced Texts Basler Lehrbücher,
https://doi.org/10.1007/978-3-031-62776-7_10
202 10 Quartets on a Sphere

to the distance between the corresponding two points in .Y. If a metric space .X is
isometric to a subspace of another metric space .Z, then .X is said to be isometrically
embeddable in .Z. A metric space is called linear (resp. planar), if the metric space
is isometrically embeddable in a line (resp. in a plane). For example, any trio (a
metric space consisting of three points) is planar. Indeed, if the distances .a, b, c
between the three points satisfy the triangle inequality, we can take three points in
the plane so that the distances between them become .a, b, c. If, further on, .a, b, c
satisfy .a + b = c (or .b + c = a, or .c + a = b), then the trio is linear.
If a metric space is linear, then it is also planar, and a planar metric space is also
isometrically embeddable in .R3 .
So far, mostly we take the radius of a sphere as unit of length and regard the
sphere as a unit sphere. In this chapter, however, we deal with spheres of different
radii at the same time. So we regard unit of length defined in another way. Suppose
that a spherical triangle .Δ̆ABC on a sphere with radius r has edge-lengths .a, b, c. If
we take r as unit of length, then the edge-lengths of .Δ̆ABC become .a/r, b/r, c/r.
Since the size of an angle is invariant under a change of unit of length, the cosine
law concerning a spherical triangle .Δ̆ABC with edge-lengths .a, b, c on a sphere of
radius r becomes

. cos(a/r) = cos(b/r) cos(c/r) + sin(b/r) sin(c/r) cos / A.

A metric space that is embeddable in a sphere of some radius is called spherical.

Any trio is obviously spherical. If .A, B, C are three points lying on a sphere .S(r)
of radius r, then .A, B, C lie on a circle on .S(r), and hence .AB + BC + CA ≤
2π r. Thus, we can see that a trio with mutual distances .x, y, z is isometrically
embeddable in .S(r) if and only if .x + y + z ≤ 2π r. A linear metric space consisting
of a finite number of points is also spherical. Indeed, it can be isometrically
embeddable in a great circle of a sphere with large radius.
Exercise 10.1 Prove that the quartet consisting of the four vertices of a rhombus in
the plane shown in Fig. 10.1 is spherical.

Solution On a sphere .S(r) of radius .r ≥ 3d, take four points .A, B, C, D in such a
way that

.AB = AD = BC = BA = BD = d,

Fig. 10.1 The length √ of the

longest diagonal is . 3d d d
d

d d
10.2 The Vertices of a Convex Quadrilateral 203

Fig. 10.2 Four points on

.S(r)
C d
D
d d
y
d
B
A d

Fig. 10.3 A convex ....

...........●
................ ..... ....
A
quadrilateral in the plane ................ ... ..
................
........
.....
.................. ...... ....
................ .. ..
B ●...................
... ....... ...
...
... ..
..
..
..
... ........ .
.... ..
... ...... ... ..
... .
..... ..
. ..
... ........ ... ..
.. .
.. ..
... ...... .. ..
... . .
........ ..
. . . ..
...
.... .......... ..
..
... . .
. ..... ..
... ... ...... ..
... .. .. ...... ..
... .. ...... ..
... . .
... ... .. .. ...... ..
..... . . ..
● .
C D
Ā

as shown in Fig. 10.2, and put .y = AC. Then .y = y(r) is a continuous function of
the radius r. When .r = 3d/(2π ), .A, B, D lie on a great circle of .S(r), and .C, B, D
also lie on a great circle. Hence A and C coincide, and .y(r) = 0. When .r = 4d/(2π )
(that is, .d/r = π/2), the interior angle of .Δ̆ABD is .π/2 by√the spherical cosine
law. Hence .y(r) = AC = AB + BC = 2d. Since .0 √ < 3d < 2d, there is
an .r0 between .3d/(2π ) and .4π/(2π ) such that .y(r0 ) = 3d by the intermediate
value theorem. Therefore, the quartet consisting of the vertices of the rhombus is
isometrically embeddable in .S(r0 ) ⨆
⨅

Remark 10.1 By a computation using Theorem 10.4 in Sect. 10.4, we have .r0 /d ≈
0.602289

10.2 The Vertices of a Convex Quadrilateral

More generally, the following theorem holds.

Theorem 10.1 For any convex quadrilateral in the plane, the metric space consist-
ing of its four vertices is spherical.

Proof Let ABCD be a convex quadrilateral in the plane (Fig. 10.3). Among the four
triangles .ΔABC, ΔBCD, ΔACD, ΔABD, suppose that .ΔABD has the largest
perimeter 2s. Put .d = AC. Let .S(r) denote the sphere of radius r centered at the
origin O of .R3 .
204 10 Quartets on a Sphere

............
............................. C
....
....
.......
......
.......
.....●
.... ..........
C
.... .... .. ....
...
....
. A ...
.. .. ...
.. ... D
.... .... ....●.... .... .... .... ............ ..... .... .... ....●
.........
.. ●
.
.... .....
.. ..................................................................................... −→
... ..
...
...
.... B ....
...
D
....
......
.........
........................................
......
....
B

Fig. 10.4 Increasing the radius r

First, put .r = r0 := s/π , and take four points .A' , B ' , C ' , D ' on .S(r) so that
.◦ .B ' = (r0 , 0, 0), D ' = (x0 , y0 , 0), y0 > 0,
' '
.◦ .A lies on the great circle .z = 0, and .C lies on the hemisphere .z ≥ 0,
' '
.◦ the spherical distances other than .A C are equal to the corresponding distances

in .{A, B, C, D}.
This is clearly possible, see Fig. 10.4 left. If .A' C ' = d for .r = r0 , then
' ' ' ' ' '
.{A , B , C , D } is an isometric embedding of .{A, B, C, D} in .S(r0 ). If .A C /= d,

then we divide the argument into two cases.

Case (i). .A' C ' < d at .r = r0 :
Let us increase the radius r of .S(r) keeping four points .A' , B ' , C ' , D ' on it
satisfying the conditions
.◦ .B ' = (r, 0, 0), D ' = (x, y, 0), y > 0,
' '
.◦ .A lies on the hemisphere .z ≤ 0, and .C lies on the hemisphere .z ≥ 0,
' '
.◦ the spherical distances other than .A C are the same as in the case .r = r0 .

This is also possible. When r increases continuously from .r0 , four points
A' , B ' , C ' , D ' move continuously, and the spherical distance .A' C ' varies con-
.

tinuously. At .r = r0 , .B ' D ' is less than the half length of the great circle of
' ' ' '
.S(r0 ), that is, .B D < π r0 , which implies that .A and .D do not lie in the same
' '
side of the great circle .B C . On the other hand, when r becomes very large,
' ' ' '
.A B C D becomes closer to a convex quadrilateral in a plane. We can see that
' ' ' '
.A and .D come to lie in the same side of the great circle .B C , see Fig. 10.4

right. Therefore, at some .r = r1 > r0 , .A lies on the great circle .B ' C ' . In this
'

case, .A' C ' = A' B ' + B ' C ' . Thus, at .r = r1 , .A' C ' = A' B ' + B ' C ' > d, and
hence, when r increases from .r0 to .r1 , there is a moment when .A' C ' becomes d.
Case (ii). .A' C ' > d at .r = r0 :
10.3 Unispherical Quartets 205

Let us increase the radius r of .S(r) with keeping the four points .A' , B ' , C ' , D '
on it satisfying the conditions
.◦ .B ' = (r, 0, 0), D ' = (x, y, 0), y > 0,
.◦ both .A' and .C ' lie on the hemisphere .z ≥ 0,
' '
.◦ spherical distances other than .A C are the same as in the case .r = r0 .

Of course this is possible, too. In the convex quadrilateral ABCD, let .Ā denote
the mirror image of A with respect to the line BD, and put .d̄ = ĀC, see Fig. 10.3.
Clearly .d̄ < d. When r becomes very large, the distance .A' C ' comes closer
to .ĀC = d̄. Hence for sufficiently large r we have .A' C ' < d. Thus, when r
increases, there is a moment with .A' C ' = d.
⨆
⨅
Example 10.1 [A Nonplanar Quartet on a Sphere] Regarding .S(r) as the surface
of the Earth, let N be the north-pole, and .A, B, C be the points on the equator with
longitudes .10◦ , 20◦ , 30◦ , respectively. Then .AB = BC, AC = 2AB, AN =
BN = CN, and the trio .{A, B, C} is linear. Suppose that there is a quartet
' ' ' '
.{A , B , C , N } in the plane that is isometric to .{A, B, C, N} ⊂ S(r). Then, since
' ' ' ' ' ' ' ' '
.A C = 2A B = A B + B C , .B is the midpoint of the line segment .A C . Since
' '
' ' ' ' ' ' ' ' ' ' ' '
.A N = B N = C N , we have .ΔA B N ≡ ΔC B N . Hence . A B N = 90 , / ' ' ' ◦

and .A N is a hypotenuse of the right triangle .A B N . This means .A N > A B ' ,

' ' ' ' ' ' ' '

contradicting that the two quartets are isometric. Therefore, .{A, B, C, N} is not
planar.
Remark 10.2 P. L. Robinson [671] proved that any region on a sphere contains
a quartet that cannot be isometrically embedded in the plane. Most results in the
following are obtained by H. Maehara in [528].

10.3 Unispherical Quartets

A spherical metric space is called unispherical if the radius of the sphere in which
the space is isometrically embeddable is unique. Any trio is not unispherical. Indeed,
if .a, b, c are the mutual distances in the trio, then for any .r > (a + b + c)/(2π )
the trio is isometrically embeddable in a sphere of radius r. Thus, any trio is not
unisherical.
Similarly to the proof of Theorem 10.1, the following theorem can be proved.
Theorem 10.2 A quartet consisting of the four vertices of a convex quadrilateral
on a sphere is not unispherical. ⨆
⨅
Is there a quartet on a sphere that is unispherical? Yes. Any (small) region on a
sphere contains such a quartet.
Let us call a quartet on a sphere a three-pointed star if it consists of the three
vertices of an equilateral triangle and the center of the circumscribed cap of the
 206 10 Quartets on a Sphere

Fig. 10.5 A three-pointed C ... .... .... .... ..

star .... .... .... ...●... .. ....
.. ....
.... .. .. .... ..
.. ..
....... ... ...
... . ...
..
. . ...
.. . ..
.. . ... ..
.. .. ..
.. .. ..
.. ... .
..
... ...●
D .
..
.......
..
..
.
. .
..
..
. ..
.... . .......
. ....... ...
..
...
. ..... . ........ ..
.. .... .... ........ ..
.. ........ ....... ...
.......... .●
●... ....
A ...
...
... . ....
..
..
B
.... . .
... .... .... .
.... .... ... ... ....
. .... ...... .... .... .... .

triangle, see Fig. 10.5. In a three-pointed star (consulting Fig. 10.5) the angles
./ ADB, ./ BDC, ./ CDA are called the center-angles of the three-pointed star, which
are all equal to .2π/3. It is clear that if a quartet on a sphere of radius R is isometric
to a three-pointed star on a sphere of radius r, then the quartet is also a three-
pointed star, and hence its center angle must be equal to .2π/3. Then the triangle
comparison theorem (Theorem 5.6) implies that .R = r. Hence any three-pointed
star is unispherical.
Theorem 10.3 Since any region on a sphere contains a three-pointed star, any
region on a sphere contains a quartet whose six distances determine the radius of
the sphere uniquely. ⨆
⨅
Corollary 10.1 For any small region on a sphere, it is impossible to make a .(plane.)
map of the region in which distances are faithfully represented.

10.4 An Equation for the Radius

−→ → For a quartet
For .X ∈ R3 , we denote the vector .OX simply by .X.

.X = {P1 , P2 , P3 , P4 },

denote by .Dij the distance .d(Pi , Pj ), and define the matrix .MX (x) in the following
way:
⎛ ⎞
1 cos(D12 /x) cos(D13 /x) cos(D14 /x)
⎜cos(D21 /x) 1 cos(D23 /x) cos(D24 /x)⎟
.MX (x) = ⎜ ⎟
⎝cos(D31 /x) cos(D32 /x) 1 cos(D34 /x)⎠
cos(D41 /x) cos(D42 /x) cos(D43 /x) 1

If the four points of .X lie on the sphere .S(x) of radius x centered at the origin O of
.R3 , then the .(i, j )-entry of the matrix .x 2 MX (x) is the inner product .P→i · P→j of the
vectors .P→i and .P→j .
10.4 An Equation for the Radius 207

In a trio .Δ with mutual distances .a, b, c, the value .s(Δ) = (a + b + c)/2 is called
the semi-perimeter of the trio .Δ. Recall that a trio .Δ is isometrically embeddable in
a sphere of radius r if and only if .s(Δ) ≤ π r. For a quartet .X , we define .s(X ) by

s(X ) = max{s(Δ) : Δ ⊂ X }.
.

Theorem 10.4 For a quartet .X = {P1 , P2 , P3 , P4 } that is not linear, the following
statements are equivalent.
(1) .X is spherical.
(2) The equation .det MX (x) = 0 has a solution .x = r that satisfies .π r ≥ s(X ).
.(Such a solution is the radius of the sphere in which .X is isometrically

embeddable..)
Proof
(1) ⇒ (2): Suppose that a sphere .S(r) of radius r centered at O in .R3 contains
.
' ' ' '
.{P , P , P , P } that is isometric to .X . Then .π r ≥ s(X ). Since ./ P OP
' '
1 2 3 4 i j =
Dij /r, the inner product of two vectors .P→ ' and .P→ ' in .R3 is equal to .r 2 cos(Dij /r).
i j
Let U be the .(4 × 3)-matrix with ith row .P→i' for .i = 1, 2, 3, 4:
⎛ →' ⎞
P1
⎜P→ ' ⎟
.U = ⎜ 2 ⎟ .
⎝P→ ' ⎠
3
P→ ' 4

Then .r 2 MX (r) = U U t , where .U t is the transpose of U . Let .Ũ be the matrix

obtained from U by adding .(0, 0, 0, 0)t as the 4th column:
⎛ →' ⎞
P1 0
⎜P→ ' 0⎟
.Ũ = ⎜ 2 ⎟.
⎝P→ ' 0⎠
3
P→ ' 4 0

Then .Ũ Ũ t = U U t . Since .det(Ũ ) = 0,

. det(r 2 MX (r)) = det(U U t ) = det(Ũ Ũ t ) = 0,

and hence .det MX (r) = 0. That is, .det MX (x) = 0 has a solution .x = r that
satisfies .π r ≥ s(X ).
.(2) ⇒ (1): Suppose that .x = r is a solution of .det MX (x) = 0 that satisfies
.π r ≥ s(X ). Then every trio of .X is isometrically embeddable in the sphere .S(r)

of radius r. Let us divide the argument into two cases.

208 10 Quartets on a Sphere

Case (i). The rank of .MX (r) is equal to 2:

Suppose that the first row and the second row of .MX (r) are linearly independent.
Then we can eliminate the third row and the fourth row by subtracting suitable
linear combinations of the first row and the second row. Since .MX (r) is a
symmetric matrix, we can further eliminate the third and the fourth columns
by subtracting suitable linear combinations of the first and the second columns.
Since the rank of a matrix is invariant under such operations, we have
⎛ ⎞
1 cos(D12 /r)
. det /= 0.
cos(D21 /r) 1

Now, suppose that the fourth row of .MX (r) is represented as

.(the fourth row) = λ1 (the first row) + λ2 (the second row).

Let .{P1' , P2' , P3' } be the isometric embedding of the trio .{P1 , P2 , P3 } of .X into
'
.S(r). Define .P ∈ R by
3
4

P→4' = λ1 P→1' + λ2 P→2' .

.

Since .{P1' , P2' , P3' } ⊂ S(r) is isometric to .{P1 , P2 , P3 }, we have

⎛ ⎞ ⎛ ' ⎞ ⎛ ' ⎞t
1 cos(D12 /r) cos(D13 (r) P→1 P→1
.r ⎝cos(D21 /r) cos(D23 /r)⎠ = ⎝P→2' ⎠ ⎝P→2' ⎠ .
2
1
cos(D31 /r) cos(D32 /r) 1 P→3' P→3'

And for .i = 1, 2 we have

.P→4' · P→i' = (λ1 P→1' + λ2 P→2' ) · P→i' = λ1 P→1' · P→i' + λ2 P→2' · P→i'
( )
= r 2 λ1 cos(D1i /r) + λ2 cos(D2i /r) = r 2 cos(D4i /r).

Therefore,
⎛ ⎞ ⎛ ' ⎞t ⎛ ⎞
P→1' P→1 1 cos(D12 /r) cos(D14 /r)
. ⎝P→ ' ⎠ ⎝P→ ' ⎠ = r 2 ⎝cos(D21 /r) 1 cos(D24 /r)⎠ .
2 2
→
P4 ' →
P4 ' cos(D41 /r) cos(D42 /r) |P→4' |2 /r 2

Since the three vectors .P→1' , P→2' , P→4' are linearly dependent, the determinant of this
matrix is 0. On the other hand, the first row, the second row, and the fourth row
of .MX (r) are linearly dependent,
⎛ ⎞
1 cos(D12 /r) cos(D14 /r)
. det ⎝cos(D21 /r) 1 cos(D24 /r)⎠ = 0.
cos(D41 /r) cos(D42 /r) 1
10.4 An Equation for the Radius 209

Hence we must have .|P→4' |2 /r 2 = 1 (and hence .P4' ∈ S(r)). Thus,

⎛ ⎞t ⎛ ⎞
⎛ ' ' ⎞ P→1' 1 cos(D12 /r) cos(D13 /r) cos(D14 /r)
→ →
P1 P2 ⎜ → ' ⎟
P ⎜ cos(D /r) 1 cos(D23 /r) cos(D24 /r) ⎟
.⎝ P → ' ⎠ ⎜ 2' ⎟ = r 2 ⎜ 21 ⎟.
3 ⎝P→ ⎠ ⎝cos(D31 /r) cos(D32 /r) 1 (P→3' · P→4' )/r 2 ⎠
P→4' 3
P→4' cos(D41 /r) cos(D42 /r) (P→3' · P→4' )/r 2 1

The determinant of the matrix on the left hand side is 0. In the matrix of the right
hand side, by subtracting suitable linear combinations of the first row and the
second row from the third row and the fourth row, we have the matrix
⎛ ⎞
1 cos(D12 /r) cos(D13 /r) cos(D14 /r)
⎜cos(D /r) ⎟
⎜ 21 1 cos(D23 /r) cos(D24 /r) ⎟
.⎜ P→3' ·P→4' ⎟.
⎜ 0 0 0 − cos(D34 /r)⎟
⎝ r2 ⎠
P→3' ·P→4'
0 0 r2
− cos(D43 /r) 0

Since the determinant of this matrix is 0, we have .(P→3' · P→4' )/r 2 = cos(D43 /r).
Therefore, .{P1' , P2' , P3' , P4' } is an isometric embedding of .{P1 , P2 , P3 , P4 } in
.S(r).

Case (ii). The rank of .MX (r) is equal to 3:

Suppose the first row, the second row, and the third row of .MX (r) are linearly
independent. Then we can represent the fourth row by

λ1 (the first row) + λ2 (the second row) + λ3 (the third row).

.

First, we take an isometric embedding .{P1' , P2' , P3' } of .{P1 , P2 , P3 } in .S(r), and
define .P4' ∈ R3 by .P→4' = λ1 P→1' + λ2 P→2' + λ3 P→3' . Then, similarly to the case (i), we
can prove that .P4' ∈ S(r) and .{P1' , P2' , P3' , P4' } ⊂ S(r) is isometric to .X .
⨆
⨅
Example 10.2 Let .X = {P1 , P2 , P3 , P4 } be a metric space in which distances
Dij = d(Pi , Pj ) are given by
.

⎛ ⎞
0 5 5 3
⎜5 0 5 3⎟
.(Dij ) = ⎜ ⎟,
⎝5 5 0 6⎠
3 3 6 0

see also the diagram given in Fig. 10.6. Then .s(X )/π = 7.5/π = 2.38732. By
numerical computation using, say, Mathematica, we have

r1 = 2.4014, r2 = 23.2921
.
210 10 Quartets on a Sphere

Fig. 10.6 Distances between P2 ........

four points ...............
........ .... .......
......
......... ... ......
...
........
5 ............
........
..
...
....
3....
....

.................
..........
.
5 ..
.
.
.
.
.
....
....
.
.......................................................................................................................................................
P3 ... .........
........
6 .
.
. ...
..... P4
........
........
.
.
. .
....
........ .
.
. ....
........ ..
...
5
........
........
........
........
... .....
. ..
3
...
........ ... .....
........ .. ...
.........
......
P1

as solutions of .det MX (x) = 0 that satisfy .π r ≥ s(X ). Therefore, .X is spherical,

but not unispherical.

10.5 A Planar Quartet That Is also Spherical

By using the triangle comparison theorem for spheres (Theorem 5.6 and Corol-
lary 5.2), the following theorem can be proved easily.
Theorem 10.5 For a planar triangle .ΔABC and a point D contained in .ΔABC,
the quartet .{A, B, C, D} is not spherical.
Proof Suppose that there is a quartet .{A' , B ' , C ' , D ' } of a sphere that is isometric
to .{A, B, C, D}.
Case (i). D is an interior point of .ΔABC: By Corollary 5.2, we have
' ' ' > / ADB, / B ' D ' C ' > / BDC, / C ' D ' A' > / CDA. Hence
./ A D B
/. A' D ' B ' + / B ' D ' C ' + / C ' D ' A' > 2π . If .D ' lies in the interior of .Δ̆A' B ' C ' ,
then ./ A' D ' B ' + / B ' D ' C ' + / C ' D ' A' = 2π, and if .D ' does not lie in the interior
of .Δ̆A' B ' C ' , then ./ A' D ' B ' + / B ' D ' C ' + / C ' D ' A' < 2π. Thus we have a
contradiction.
Case (ii). D lies on the line segment AB: In this case, .D ' lies on .A' B ' . Hence,
' ' ' ' ' '
./ A D C + / B D C = π . However, by Corollary 5.2 we have ./ A D C >
' ' '
/ ADC, / B ' D ' C ' > / BDC, and hence .π = / A' D ' C ' + / B ' D ' C ' > / ADC +
/ BDC = π, a contradiction.

⨆
⨅
Corollary 10.2 A quartet of the plane .R2 is spherical if and only if either the four
points are collinear, or the four points are the vertices of a convex quadrilateral in
the plane. ⨆
⨅
10.7 Notes 211

Fig. 10.7 Six connected graphs

10.6 Exercises

10.1∗ Prove that the quartet shown in Fig. 10.6 is not isometrically embeddable in
R3 .
10.2 Let X be a quartet with six distances 1, 1, 1, 1, 1, 2.
(i) Is X a planar quartet?
(ii) Is X a unispherical? If so, find the radius of the sphere that contains a
quartet isometric to X .
10.3∗ Let ABCD be convex quadrilateral on the unit sphere S such that

.AB = AC = BC = π/2, BD = CD = π/3.

Find the spherical distance AD.

10.4∗ Let ABCD be the convex quadrilateral on the unit sphere S given in 10.3∗ .
Find the radius r /= 1 of a sphere into which the quartet {A, B, C, D} can be
isometrically embedded.
10.5 Every connected graph can be regarded as a metric space by the path-
metric, that is, the distance d(x, y) between two vertices x, y is defined
as the minimum number of edges contained in a path connecting x and y.
Classify the six quartets given by the six connected graphs in Fig. 10.7 into
the following three types of quartets:

{not spherical}, {spherical but not unispherical}, {unispherical}.

.

10.7 Notes

The following notes are very short, since the theme of Chap. 10 is (compared
with other chapters) very specific in nature. However, we try to give shortly some
background.
212 10 Quartets on a Sphere

10.7.1 Notes on Four-Point Systems

Based on the notions of metric and semi-metric spaces, Menger and Blumen-
thal presented with [570] and [96] trend-setting publications giving insights into
embeddings and corresponding characterizations of Euclidean, Hilbert, spherical,
hyperbolic, and elliptic spaces. Within this framework, so-called four-point proper-
ties (referring to metrically defined quadruples of points) play an essential role.
These quadruples yield, roughly speaking and for example, characterizations of
generalized Euclidean spaces obtained via their metrical four-point properties.
These say that each complete metric space containing a metric line that connects
every pair of its points is a generalized Euclidean space iff each quadruple (from a
certain class of quadruples of this space) is congruent with a quadruple of points in
a Euclidean space. Clearly, this is closely related to generalizations of the triangle
inequality. The book [96] shows various types of such characterizations, also and in
particular referring to spherical spaces (see, e.g., Chapter VII and VIII).
Later contributions to four-point properties which characterize different types of
spaces are, for example, contained in the following publications: Day [202], Freese
[278], Valentine [792], Valentine [793], Freese et al. [280] as well as Freese and
Andalafte [279]. For instance, many characterizations of real inner product spaces
among metric spaces have been proven based on Euclidean four-point embedding
properties, and many results in this direction interestingly refer also to Banach
spaces.
Chapter 11
Higher Dimensions

Once we accept our limits, we go beyond them.

— Albert Einstein

First, we compute the volume of an n-dimensional ball .B n (r) of radius r in .Rn , and
we prove that for every fixed .r > 0, the volume of the n-dimensional ball tends to
0 as n tends to .∞. How many unit balls in .Rn can touch .B n (r) simultaneously?
The maximum number of such unit balls is called the kissing number of .B n (r). We
already showed in Chap. 8 that the kissing number of .B 3 (1) is twelve. It is generally
difficult to find the kissing number even for .r = 1 for large n. For small radius,
however, it is possible to determine the kissing number n
√ of .B (r). In this chapter, we
determine the kissing number of .B n (r) when .r ≤ 2 − 1 for every .n > 0.

11.1 Figures in High Dimensions

The n-dimensional Euclidean space .Rn is a vector space over .R consisting of all n-
tuple row vectors .x = (x1 , x2 , . . . , xn ) where .xi ∈ R for .i = 1, 2, . . . , n, in which
addition, subtraction, scalar multiple, and inner product are defined as follows: for
.x = (x1 , x2 , . . . , xn ), y = (y1 , y2 , . . . , yn ) and .λ ∈ R:

.x ± y = (x1 ± y1 , x2 ± y2 , . . . , xn ± yn )
λx = (λx1 , λx2 , . . . , λxn )
x · y = x1 y1 + x2 y2 + · · · + xn yn .

Elements of .Rn are called points. For .x ∈ Rn , its norm .‖x‖ is defined as

√ /
‖x‖ =
. x · x = x12 + x22 + · · · + xn2 .

© The Author(s), under exclusive license to Springer Nature Switzerland AG 2024 213
H. Maehara, H. Martini, Circles, Spheres and Spherical Geometry,
Birkhäuser Advanced Texts Basler Lehrbücher,
https://doi.org/10.1007/978-3-031-62776-7_11
214 11 Higher Dimensions

The distance .d(x, y) between two points .x, y ∈ Rn is defined as .d(x, y) = ‖x−y‖.
For .r > 0, the subset

B n (r) = {x ∈ Rn : ‖x − p‖ ≤ r}
.

of .Rn is called the n-dimensional ball of radius r centered at .p. Its “boundary"

Sn (r) = {x ∈ Rn : ‖x − p‖ = r}
.

is a hypersphere (simply, a sphere) in .Rn with radius r and center .p. For .a > 0, the
subset

{(x1 , x2 , . . . , xn ) ∈ Rn : 0 ≤ xi ≤ a, i = 1, 2, . . . , n}
.

of .Rn is called an n-dimensional cube of edge-length a. An n-dimensional cube

with edge-length 1 is called an n-dimensional unit cube.
For .k + 1 points .x i ∈ Rn , i = 0, 1, 2, . . . , k, such that the k vectors
−
x−→ −−→ −−→
.0x1, x0x2, . . . , x0xk

are linearly independent, the set

{λ0 x 0 + λ1 x 1 + · · · + λk x k ∈ Rn : λ0 + λ1 + · · · + λk = 1, λi ≥ 0}
.

is called the k-dimensional simplex (simply, k-simplex) spanned by .x 0 , x 1 , . . . , x k ,

and each .x i is called a vertex of that simplex. Of course, .k ≤ n. A 1-simplex is a line
segment, a 2-simplex is a triangle, and a 3-simplex is a terahedron. In an n-simplex,
the line segment connecting a pair of vertices
( ) is called an edge of the simplex. The
number of edges of an n-simplex is . n+1 2 = n(n + 1)/2. In an n-simplex, any
n vertices span an .(n − 1)-simplex, which is called a facet of the n-simplex. An
n-simplex has .n + 1 facets.
For an n-simplex in .Rn , the (hyper)sphere that passes through all vertices of the
simplex is called the circumscribed sphere (or circumsphere) of the simplex. This
sphere is uniquely determined by the n-simplex. An n-simplex whose edges are all
of the same length is called a regular n-simplex.
√ √
1+ n+1 1− n+1
Example 11.1 For .t = , or .t = , the .n + 1 points
n n
⎧
⎪
⎪ x 0 = (t, t, t, . . . , t, t)
⎪
⎪
⎪
⎪
⎨x 1 = (1, 0, 0, . . . , 0, 0)
⎪
. x2 = (0, 1, 0, . . . , 0, 0)
⎪
⎪
⎪
⎪
⎪. . .
⎪ ...
⎪
⎩x
n = (0, 0, 0, . . . , 0, 1)
√
in .Rn span a regular n-simplex with edge-length . 2.
11.2 Moser’s Paradox 215

Theorem 11.1 The circumscribed sphere of a regular n-simplex in .Rn with unit
edge-length has radius
/
n
. .
2(n + 1)
/
n
Remark 11.1 .
2(n+1) is a monotonically increasing function of n.

Proof Let .x 0 , x 1 , . . . , x n be the vertices of a regular n-simplex in .Rn . We may

suppose that .x 0 is the origin O of .Rn . Since

1 = ‖x 1 − x 2 ‖2 = ‖x 1 ‖2 + ‖x 2 ‖2 − 2x 1 · x 2 = 2 − 2x 1 · x 2 ,
.

we have .x 1 · x 2 = 1/2. Hence we have

⎧
1 (for 1 ≤ i = j ≤ n)
xi · xj =
.
1/2 (for 1 ≤ i /= j ≤ n).

Put .p = 1
n+1 (x 0 + x 1 + · · · + x n ). Since .x 0 = O,

1 ( ⎲ )
‖p − x 0 ‖2 = p · p =
. n+ 2x i · x j
(n + 1)2
1≤i<j ≤n

1 ( ) n
= n + n(n − 1)/2 = .
(n + 1)2 2(n + 1)
/
Hence .‖p − x 0 ‖ = n
2(n+1) . Since

‖p − x 1 ‖2 = p · p + x 1 · x 1 − 2p · x 1
.

n 2 ( n − 1) n
= +1− 1+ = ,
2(n + 1) n+1 2 2(n + 1)
/
we have .|p − x 1 | = 2(n+1) n
. Similarly, for .1 ≤ i ≤ n, we have .‖p − x i ‖ =
/ /
n n
2(n+1) . Therefore, the sphere with center .p and radius . 2(n+1) passes through
the vertices .x 0 , x 1 , . . . , x n . ⨆
⨅

11.2 Moser’s Paradox

In a square with edge-length 4 in .R2 , we can pack four unit disks as shown in
Fig. 11.1. Then, in the room enclosed √ by these four unit disks at the center of the
square, we can pack a disk of radius . 2 − 1. (The centers of the four unit disks form
216 11 Higher Dimensions

Fig. 11.1 A small disk

enclosed by four unit disks

√
a square of edge-length 2, and its diagonal has√length .2 2. Hence the diameter of
the disk packed in the room at the center is .2 2 − 2, and therefore the radius is
√
. 2 − 1.)

In a 3-dimensional cube of edge length 4 in .R3 , we can pack eight unit balls at
the eight corners of √ the cube. Then in the room at the center of the cube, we can
pack a ball of radius . 3 − 1. In the case of the 4-dimensional cube with edge length
4, we can pack 16 unit balls at the 16√ corners of the cube. Then, at the center of the
cube, we can pack a ball of radius . 4 − 1. That is, a unit ball can lie in the room.
Similarly, in the n-dimensional cube of edge-length 4, we can pack .2n unit balls at
the .2n corners of√the cube, and in the room at the center of the cube, we can pack
a ball of radius . n − 1. If, for example,√.n = 9, then the radius of the ball that can
be packed in the room at the center is . 9 − 1 = 2, and the ball is tangent to the
boundary of the cube. If .n > 9, then the central ball protrudes from the cube!
Remark 11.2 M. Gardner introduced this observation as paradox by Leo Moser in
[294, Chapter 3], see also the related notes to this chapter here.

11.3 The Volume of a Ball

Let us compute the volume (n-dimensional volume) of the ball .B n (r) of radius
r. Let .V (n) denote the volume of an n-dimensional ball of unit radius. Then the
volume of the n-dimensional ball of radius r becomes .V (n)r n . Since the volume
of the “sphere-shell" between two concentric hyperspheres of radii .r + h and r is
approximated by .(surface area of B n (r)) × h, and

V (n)(r + h)n − V (n)r n

. lim = nV (n)r n−1 ,
h→0 h

the area of a hypersphere .Sn (r) of radius r is given by .nV (n)r n−1 .
Preceding the computation of the volume .V (n) of the n-dimensional unit ball,
we try an exercise.
11.3 The Volume of a Ball 217

⎰∞ −x 2 dx.
Exercise 11.1 Find .α := −∞ e
Solution We have
⎛⎰ ∞ ⎞⎛ ⎰ ∞ ⎞ ⎰
e−x dx e−y dy = e−(x
2 2 2 +y 2 )
α2 =
. dxdy.
−∞ −∞ R2

By converting cartesian coordinates .(x, y) to polar coordinates .(r, θ ), we get

⎰ ∞ ⎛ ⎰ 2π ⎞ ⎰ ∞ ⎾ −1 ⏋∞
e−r dθ rdr = 2π re−r dr = 2π e−r
2 2 2
α2 =
.
2 0
= π.
0 0 0
√
Therefore, .α = π. ⨅
⨆
To represent .V (n), we need the gamma function .𝚪(s) defined by a convergent
improper integral:
⎰ ∞
𝚪(s) =
. zs−1 e−z dz (s > 0).
0

Concerning the gamma function, the following equalities hold:

• .𝚪(s + 1) = s𝚪(s).
• For an integer .m ≥ 0,

( ) (2m)! √
𝚪(m + 1) = m!,
. 𝚪 m + 21 = 2m π.
2 m!

Remark 11.3 For these equalities and the gamma function see, e.g., the book [41]
of E. Artin.
Theorem 11.2 The volume .V (n) of an n-dimensional unit ball is given by

π n/2
V (n) =
. .
𝚪(n/2 + 1)

Proof To find the volume .V (n), let us consider the integral of the function
e−(x1 +x2 +···+xn ) in the whole .Rn :
2 2 2
.

⎰
e−(x1 +x2 +···+xn ) dx1 dx2 . . . dxn .
2 2 2
.I =
Rn
(√ )n
From Exercise 11.1, we have .I = π = π n/2 . Let us compute I in a different
way: First, integrate the function over the surface of the hypersphere .Sn (r) of radius
r centered at the origin of .Rn , and then integrate the result by r from .r = 0 to
.r = ∞. Since the value of the integrand is constant (.= e
−r 2 ) on the surface of
218 11 Higher Dimensions

the hypersphere .Sn (r), and the area of .Sn (r) is .nV (n)r n−1 , its integral over .Sn (r)
becomes

e−r nV (n)r n−1 .

2
.

Therefore, we have
⎰ ∞ ⎰ ∞
e−r nV (n)r n−1 dr = nV (n) r n−1 e−r dr.
2 2
I=
.
0 0

By changing variables as .z = r 2 , and noting that .dz = 2rdr, this yields

⎰ ∞
nV (n)
I=
. zn/2−1 e−z dz.
2 0

Since the improper integral is equal to .𝚪(n/2),

nV (n)
I=
. 𝚪(n/2) = V (n)(n/2)𝚪(n/2) = V (n)𝚪(n/2 + 1).
2

Then, from .I = π n/2 , we have

π n/2
V (n) =
. .
𝚪(n/2 + 1)

⨆
⨅
π x/2
Figure 11.2 shows the graph of .y = 𝚪(x/2+1) for .0 ≤ x ≤ 20. Using the second
equality of the gamma function, we have

π2
V (1) = 2, V (2) = π, V (3) =
.
4π
3 ≈ 4.19, V (4) = 2 ≈ 4.93,
8π 2 π3 16π 3
V (5) = 15 ≈ 5.26, V (6) = 6 ≈ 5.17, V (7) = 105 ≈ 4.72.

π x/2
Fig. 11.2 .y = 𝚪(x/2+1) 5

0
0 5 10 15 20
11.3 The Volume of a Ball 219

In Fig. 11.2, one sees that the maximum value of .V (n) is .V (5) and that .V (n) → 0
as .n → ∞.
Exercise 11.2 Prove that .V (5) > V (n) for .n > 5.
Solution From the approximated value of .V (n) for .n ≤ 7, we see that .V (5) is the
maximum in .{V (n) : 1 ≤ n ≤ 7}. Since .𝚪(n/2 + 1) = (n/2)𝚪(n/2), and .2π/n < 1
for .n ≥ 7, we have

π n/2 / 𝚪(n/2 + 1) 2π
V (n)/V (n − 2) =
. = < 1.
π n/2−1 / 𝚪(n/2) n

Thus, .V (n) < V (n − 2) for .n ≥ 7, and hence .V (n) < V (5) follows for .n > 7. ⨆
⨅
Theorem 11.3 For any .r > 0, the volume of .B n (r) tends to 0 as n tends to .∞.
Proof From the equality .𝚪(s + 1) = s𝚪(s), we have .𝚪(s)/ 𝚪(s + 1) = 1/s. Hence,
for .n ≥ 3,

r n V (n) n/2 / 𝚪(n/2 + 1) πr2 2π r 2

2π
. = r = = ,
r n−2 V (n − 2) π n/2−1 / 𝚪(n/2) n/2 n

and
⎛ 2π r 2 ⎞
r n V (n) = r n−2 V (n − 2)
. .
n

Let .N > 0 be an integer such that .2π r 2 /N < 1/2. Then, for .n ≥ N , we have
n
.r V (n) < r
n−2 V (n − 2)/2. From this we get

r 2m+N V (2m + N) < r 2m−2+N V (2m − 2 + N)/2

.

< r 2m−4+N V (2m − 4 + N)/22

< ···
< r 2 V (N)/2m .

Therefore,

. lim r 2m+N V (2m + N) = 0,

m→∞

that is, .r n V (n) → 0 (n → ∞). ⨆

⨅
It follows from Theorem 11.3 that for any .ε > 0 and any .r > 0, there is an .n0
such that .r n V (n) < ε holds for .n > n0 . Thus, when .n > n0 , a subset of .Rn with
volume .> ε cannot be covered by a ball of radius r. Hence, we have the following
corollary.
220 11 Higher Dimensions

Corollary 11.1 For any .ε > 0 and for any .r > 0, there is an .n0 > 0 such that
if .n > n0 , then a subset of .Rn with volume .> ε contains a pair of points at least
distance r apart. ⨆
⨅
For example, a subset of .Rn with volume .1cmn can contain, if n is very large, a
pair of points more than 100 kilometers apart.

11.4 Two Lemmas

Lemma 11.1 Let {x 1 , x 2 , . . . , x m } be a set of m points lying on the sphere Sn (r)

of radius r in Rn . Then
⎲
.m2 r 2 ≥ ‖x i − x j ‖2
i<j

holds. Equality holds if and only if the barycenter

p := (x 1 + x 2 + · · · + x m )/m
.

of x 1 , x 2 , . . . , x m coincides with the center of Sn (r).

Proof Let q be the center of the sphere, and compute ‖x i − q‖2 by using inner
products:

‖x i − q‖2 = (x i − q) · (x i − q)
.

= (x i − p + p − q) · (x i − p + p − q)
= ‖x i − p‖2 + ‖p − q‖2 + 2(x i − p) · (p − q).

Since p is the barycenter of x 1 , . . . , x m , we have

⎲
m
. 2(x i − p) · (p − q) = 0.
i=1

Hence

⎲
m ⎲
m
. ‖x i − q‖ = 2
‖x i − p‖2 + m‖p − q‖2 ,
i=1 i=1

and therefore

⎲
m ⎲
m
. ‖x i − p‖2 = ‖x i − q‖2 − m‖p − q‖2 . (11.1)
i=1 i=1
11.4 Two Lemmas 221

On the other hand,

⎲ ⎲
m
2
. ‖x i − x j ‖2 = ‖x i − x j ‖2
i<j i,j =1

⎲
m ⎲
m
= (x i − x j ) · x i − (x i − x j ) · x j
i,j =1 i,j =1

⎲
m ⎲
m
= (x i − x j ) · x i + (x j − x i ) · x j
i,j =1 i,j =1

⎲
m
=2 (x i − x j ) · x i .
i,j =1

⎲
m
Since (x i − x j ) · x i = mx i · x i − mp · x i = m(x i − p) · x i , this equals
j =1

⎲
m
2m
. (x i − p) · x i ,
i=1

⎲
m
and since (x i − p) · p = 0, it becomes
i=1

⎲
m ⎲
m
.2m (x i − p) · (x i − p) = 2m ‖x i − p‖2 .
i=1 i=1

Therefore
⎲ ⎲
m
. ‖x i − x j ‖2 = m ‖x i − p‖2 .
i<j i=1

From this and (11.1), we have

⎲ ⎲
m
. ‖x i − x j ‖2 = m ‖x i − q‖2 − m2 ‖p − q‖2
i<j i=1

⎲
m
≤m ‖x i − q‖2
i=1

≤m r . 2 2

Clearly, equality holds only when p = q. ⨆

⨅
222 11 Higher Dimensions

Lemma 11.2 For m points x i , i = 1, 2, . . . , m, in Rn \ {O}, the following

statements (i) and (ii) hold:
(i) If x i · x j < 0 for every i /= j , then m ≤ n + 1.
(ii) If x i · x j ≤ 0 for every i /= j , then m ≤ 2n.
Proof
(i) The assertion is clearly true for n = 1. Suppose the assertion is true for n − 1
dimensions, and let us consider the n-dimensional case. Let

x i = (x1(i) , x2(i) , . . . , xn(i) ), i = 1, 2, . . . , m,

.

be m points in Rn such that x i · x j < 0 for i /= j . By a suitable transformation

of coordinates of Rn , we may suppose that the first coordinate of x 1 is positive
and all other coordinates are 0, namely,

(1) (1)
x 1 = (x1 , 0, 0, . . . , 0), x1 > 0.
.

(j )
Then, since x 1 · x j < 0, j = 2, 3, . . . , m, the first coordinates x1 of x j (j =
(1)
2, 3, . . . , m) are all negative, that is, xj < 0, j = 2, 3, . . . , m. Put

(j ) (j ) (j )
y j = (x2 , x3 , . . . , xn ), j = 2, 3, . . . , m.
.

(i) (j )
Then, y 2 , y 3 , . . . , y m are m − 1 points in Rn−1 . Since x1 x1 > 0 and
(i) (j )
x1 x1 + y i · y j = x i · x j < 0 for 2 ≤ i < j ≤ m, we must have y i · y j < 0.
(Hence y j /= O for j = 2, 3, . . . , m.) Then by the inductive assumption, we
have m − 1 ≤ (n − 1) + 1 = n, and hence m ≤ n + 1.
(ii) The assertion is clearly true for n = 1. Assuming that the assertion is true for
n − 1 dimensions, let us consider the n-dimensional case. Let
(i) (i)
x i = (x1 , x2 , . . . , xn(i) ), i = 1, 2, . . . , m,
.

be m points in Rn such that x i · x j ≤ 0 for i /= j . We divide the argument into

two cases.

Case (1). The origin O and some two points x i , x j are collinear:
Suppose O and x 1 , x 2 are collinear. Since x 1 · x 2 ≤ 0, O must lie between x 1
and x 2 . So, by a suitable coordinate-transformation of Rn , we may assume

.x 1 = (x1(1) , 0, 0, . . . , 0), x1(1) > 0,

(2) (2)
x 2 = (x1 , 0, 0, . . . , 0), x1 < 0.
11.5 Kissing a Small Ball 223

(j )
Then, since x 1 · x j ≤ 0, x 2 · x j ≤ 0, j = 3, 4, . . . , m, the first coordinate x1
of x j must be 0 for j = 3, 4, . . . , m. Put

(j ) (j ) (j )
y j = (x2 , x3 , . . . , xn ), j = 3, 4, . . . , m.
.

Then, y 3 , y 4 , . . . , y m are m−2 points in Rn−1 \{O}, and y i ·y j = x i ·x j ≤ 0 for

3 ≤ i < j ≤ m. Then, by the inductive assumption, we have m − 2 ≤ 2(n − 1).
Hence, m ≤ 2n.
Case (2). For any two points x i , x j , the three points O, x 1 , x 2 are not collinear:
We may assume that the first coordinate of x 1 is positive, and all other
coordinates of x 1 are 0. That is,

x 1 = (x1(1) , 0, 0, . . . , 0), x1(1) > 0.

.

Since x 1 · x j ≤ 0, j = 2, 3, . . . , m, the first coordinate xj(1) is negative or 0 for

j = 2, 3, . . . , m. Put

(j ) (j ) (j )
y j = (x2 , x3 , . . . , xn ), j = 2, 3, . . . , m.
.

Then, y 2 , y 3 , . . . , y m are m − 1 points in Rn−1 . Since O, x 1 , x j are not collinear

(i) (j ) (i) (j )
for j ≥ 2, y j /= (0, 0, . . . , 0). Moreover, since x1 x1 ≥ 0 and x1 x1 + y i ·
y j = x i · x j ≤ 0 for 2 ≤ i < j ≤ m, we have y i · y j ≤ 0 for i /= j . Then by the
inductive assumption, we have m − 1 ≤ 2(n − 1), and hence m ≤ 2n − 1 ≤ 2n.
⨆
⨅

11.5 Kissing a Small Ball

A point-set .X in .Rn is called dispersed if .‖x − y‖ ≥ 1 holds for every distinct

pair .x, y ∈ X . As we saw in the problem of thirteen balls in Chap. 6, there is a
dispersed set of twelve points on the unit sphere in .R3 . Let us denote by .ν[Sn (r)]
the maximum number of points that a dispersed set on .Sn (r) can contain.
√
Exercise 11.3 Prove that if .r = 1/ 2, then .ν[Sn (r)] = 2n for every .n > 0.
√
Proof (Solution) For .r = 1/ 2, the set of 2n points

(r, 0, 0, . . . , 0)
. (−r, 0, 0, . . . , 0)
(0, r, 0, . . . , 0) (0, −r, 0, . . . , 0)
··· ···
(0, 0, 0 . . . , r) (0, 0, 0 . . . , −r)

in .Rn is a dispersed set on .Sn (r). Hence .ν[Sn (r)] ≥ 2n.

224 11 Higher Dimensions

On the other hand, if .{x 1 , . . . , x m } is a dispersed set on .Sn (r), then, since

1 ≤ ‖x i − x j ‖2 = (x i − x j ) · (x i − x j ) = 1/2 + 1/2 − 2x i · x j
.

for .i /= j , we have .x i · x j ≤ 0. This implies, by Lemma 11.2(ii), .m ≤ 2n, and thus

ν[Sn (r)] ≤ 2n.
. ⨆
⨅
√
From Exercise 11.3, we see that when .r ≥ 1/ 2, √ .ν[Sn (r)] increases as n
increases. To find√ the exact value of .ν[Sn (r)] for .r > 1/ 2 is difficult. However, in

the case .r < 1/ 2, we can derive the exact value of .ν[Sn (r)] for every n.
√
Theorem 11.4 For .0 < r < 1/ 2,
⎧
n + 1 for n < nr − 1
ν[Sn (r)] =
. ,
nr for n ≥ nr − 1

1
where .nr denotes the integer part of . 1−2r 2.

For example, when .r = 0.65, we have . 1−2×(0.65)

1
2 = 6.4516, and hence .nr = 6.
Thus, .ν[Sn (0.65)] = n + 1 for .n < 5, and .ν[Sn (0.65)] = 6 for .n ≥ 5.
Proof
(1) The case .n < nr − 1:
Let .{x 1 , . . . , x m } be a dispersed set of m points on .Sn (r). Since

1 ≤ ‖x i − x j ‖2 = 2r 2 − 2x i · x j ,
.

we have .x i ·x j < 0. Then, by Lemma 11.2(i), .m ≤ n+1. On the other hand, the
/radius of the circumscribed sphere of a regular n-simplex of unit edge-length is
2(n+1) by Theorem 11.1. Since .n < nr − 1, by the remark after Theorem 11.1
n
.

we have
/ /
n nr − 1
. ≤ .
2(n + 1) 2nr
/ /
nr −1
Since .nr ≤ 1−2r 1
2 implies .r ≥ 2nr , we have .r ≥
n
2(n+1) . Hence, .Sn (r)
is a circumscribed sphere of a regular n-simplex with edge length .≥ 1. Thus,
.ν[Sn (r)] ≥ n + 1, and hence .ν[Sn (r)] = n + 1

(2) The case .n ≥ nr − 1:

Let .{x 1 , . . . , x m } be independent m points on .Sn (r). By Lemma 11.1,
⎲ m(m − 1)
m2 r 2 ≥
. ‖x i − x j ‖2 ≥ ,
2
i<j
11.5 Kissing a Small Ball 225

from which we have .m ≤ 1−2r 1

2 . Hence .m ≤ nr . On the other hand, the radius
the circumscribed sphere of a regular .(nr − 1)-simplex with/unit edge length
of /
nr −1 nr −1
is . 2nr by Theorem 11.1. Since .nr ≤ 1
1−2r 2
implies .r ≥ 2nr , a sphere
of radius r in .Rnr −1
is the circumscribed sphere of a regular .(nr − 1)-simplex
with edge length .≥ 1, and hence there is a dispersed set of .nr points. Therefore,
.ν[Sn (r)] = nr .

⨆
⨅
Remark 11.4 Theorem 11.4 was essentially obtained by R. A. Rankin in [655].
For an n-dimensional ball .B n (r) of radius r in .Rn , the maximum number of unit
balls in .Rn that can touch .B n (r) simultaneously (without overlapping) is called the
kissing number of .B n (r).
Lemma 11.3 For .r > 1/2, the kissing number of an n-dimensional ball of radius
2r − 1 is equal to .ν[S n (r)].
.

Proof Let O be the center of .Sn (r) and .x, y ∈ Sn (r) be two points such that .‖x −
−→ −→
y‖ = 1. Let .x ' , y ' be the points on the rays .Ox, Oy, respectively, such that .‖x ' −
O‖ = ‖y ' − O‖ = 2r. Then .‖x ' − y ' ‖ = 2, and the unit balls with centers .x ' , y '
touch the ball .B n (2r −1), see Fig. 11.3. From this we can see that the kissing number
of .B n (2r − 1) is at least .ν[Sn (r)]. Similarly, we can see that the kissing number of
.B (2r − 1) is at most .ν[Sn (r)]. ⨆
⨅
n

√
Theorem 11.5 For .0 < ρ < 2 − 1, the kissing number of .B n (ρ) is
⎧
n+1 for n < n(ρ) − 1,
.
n(ρ) for n ≥ n(ρ) − 1,

2
where .n(ρ) is the integer part of . .
2 − (ρ + 1)2

Fig. 11.3 Two unit balls

touch .Sn (2r − 1)
x
B n (2r − 1)
x
r
O
1
y
Sn (r) y
226 11 Higher Dimensions

Proof Let .ρ = 2r−1. Then .r = (ρ+1)/2, and the kissing number of .B n (ρ) is equal
to .ν[Sn ( ρ+1
2 )] by Lemma 11.3. Since .nr = 1−(ρ+1)2 /2 = 2/(2 − (ρ + 1) ) = n(ρ),
1 2

the theorem follows from Theorem 11.4. ⨆

⨅

11.6 Exercises

11.1∗ Prove that among the volumes V (n) of the n-dimensional unit ball, n =
1, 2, 3, . . . /
, the one that has the maximum value is V (5).
n
11.2 Show that 2(n+1) is monotonically increasing in n.
11.3∗ Prove that 𝚪(s + 1) = s𝚪(s).
11.4 Prove that 𝚪(m + 1) = m! for every integer m > 0.
11.5 Using Lemma 11.1, prove Theorem 11.1.
11.6 In R 3 , find the maximum number of balls of radius 5 that can simultaneously
touch one and the same ball of radius 2.

11.7 Notes

In the above part of Chap. 11, the following geometric figures (in high dimensions)
were studied or used: the (unit) ball, the simplex, and the cube. Therefore in the
following three subsections (of the notes) precisely these three types of figures are
discussed, in n-dimensional version and (due to technical reasons) in reverse order.

11.7.1 Notes on Cubes

In the following we give a selection of nice overviews and results on the geometry
of the cube. We see this (like also the next subsection on simplices) as a good “exer-
cise” in high-dimensional geometry, then needed for high-dimensional geometry
of balls. In particular, here we lay emphasize on sections and projections of n-
dimensional cubes.
More elementary results on and properties of cubes, mainly in dimension 3,
are discussed in the booklets Ranucci and Rollins [656] as well as Ehrenfeucht
[240], treating curiosities on dissections, slices, colorings, Pohlke’s theorem from
descriptive geometry, and geodesics. Similarly, the exposition [804] of Weiss
collects also results from 3-space, e.g. referring to geodesics (applying Snellius’
refraction law) and billiards.
A broader and deeper view is given in Zong’s booklet [841] which is somehow
prepared by the excellent survey [840]. The author treats cross-sections and
projections of n-cubes, inscribed simplices and optimal (regarding the number of
11.7 Notes 227

needed simplices) triangulations of cubes, also in view of hyperbolic geometry, cube

tilings (e.g., Keller’s conjecture on twins in cube tilings) and further topics, mainly
from discrete and convex geometry.

Sections of Cubes

In Ball’s lecture [55, Lecture 6] the extremum property of the cube in the symmetric
case of the reverse isoperimetric problem is nicely discussed. For new contributions
extending Ball’s well known result on maximal hyperplanar cube slices (namely,
that
√ the sharp upper bound for the volume of hyperplane sections of a unit cube is
. 2) we refer to the recent paper [422] of Ivanov and Tsiutsiurupa, containing the

important references. And we mention here the lecture notes [680] in which also
Mahler’s conjecture (the volume product regarding polar bodies is, in the symmetric
case, minimal at cubes) is studied. (Maximal) cube slicing is also related to section
functions of convex bodies or polytopes, and therefore to geometric tomography,
i.e., also to Gardner’s important monograph [295, Chapters 7 and 8] (in particular, to
Note 7.8) and to Martini’s survey [547, Section 2.3]. We repeat here Ball’s essential
result on the largest hyperplanar cube slice in view of a prominent application,
namely the famous Busemann-Petty problem. And Melbourne and Roberto [567]
derive a quantitative form of this result of Ball. Frank and Riede [275] succeed in
deducing an elementary formula for the volume of arbitrary hyperplane slices of the
n-cube.
Chakerian and Logothetti [154] deal with the geometry and combinatorics of
hyperplanar cube-slicing, studying interesting sequences of cube slices and, related
to this, binomial coefficients and properties of Pascal’s triangle. The volumes of
slices and related slabs are studied, and probabilistic questions coming from these
cube slices are also discussed. Also Ball’s result and its “Busemann-Petty applica-
tion" (see above) is discussed. The survey [685] by Saks refers to combinatorial
problems related to sections, partitions and coverings of the n-cube. E.g., the
maximal number of edges that can be intersected by a single hyperplane is precisely
determined, and also bounds for the minimal number of hyperplanes needed to meet
each edge are discussed. Bartha et al. [64] study volumes of central slices of unit n-
cubes orthogonal to their main diagonals for increasing n, showing that they give a
strictly monotonically increasing function of the dimension.
The volume of central slices of unit cubes with slabs of fixed width are studied by
Barthe and Koldobsky [66]. They show that if the width is at most .3/4, then slabs
parallel to facets yield the minimum, whereas for large width, the slabs orthogonal
to the main diagonal become asymptotically minimal as n goes to infinity. Moody
et al. [582] investigate extremal volumes of non-central slices of the unit n-cube,
meaning that the cutting hyperplanes have to be at a certain distance .d > 0 from
the center. This can also be extended to non-central slices of the unit n-cube with
slabs. In [582] the authors solve this completely for one-dimensional √ slices, and
they give a partial solution for extremal hyperplane slices with .d > n − 1/2.
Slices of minimal volume by slabs of width 2d, with the same bound for d, are
228 11 Higher Dimensions

also discussed. Further papers treating slab slices of unit n-cubes are [464] of König
and Koldobsky and [463] of König. And in [465] these authors investigate the
(.n − 2)-dimensional surface area of central hyperplane slices of the unit n-cube. For
instance, they show that the maximum is attained perpendicularly to the direction
vector .(1, 1, 0, ..., 0). This is applied to give a negative answer to an analogue of
the Busemann-Petty problem for surface area, namely for dimensions .n ≥ 14. And
Ivanov and Tsiutsiurupa [422] discuss maximal volumes of k-dimensional sections
of the unit n-cube, for .n ≥ 2 presenting a necessary and sufficient condition for a
k-subspace to be a local maximizer of this section function.

Projections of Cubes

It is known already for a long time that k-dimensional parallel and, in particular,
orthogonal projections of n-cubes are k-dimensional zonotopes, i.e., k-dimensional
finite sums of line segments. E.g., Naumann [594] uses vector stars and properties
of zonotopes to characterize orthogonal projections of n-cubes among arbitrary
parallel projections. This is practically a result from descriptive geometry, more
precisely a generalization of Pohlke’s theorem from the theory of axonometry. Also
Bergold [80] slightly generalizes Pohlke’s theorem: He shows that two arbitrary
parallelepipeds in 3-space can always be turned in such a way that they appear
identical, i.e., that their parallel projections onto a 2-plane perpendicular to the
direction of sight are similar. Maehara [527] proves that for any .k ≥ 2 and .n ≥ k,
there is an n-dimensional unit cube which is projected to a regular 2k-gon using
an orthogonal projection from n-space onto a 2-dimensional subspace. The author
studies these projections for increasing dimension n.
McMullen [565] proves that the orthogonal projections of the unit n-cube onto
totally orthogonal subspaces of dimensions k and .n − k of Euclidean n-space have
the property that their k-volume and (.n − k)-volume are, respectively, equal. He
himself and Schnell [708] extended this result suitably to affine cubes and finite
sums of segments (= zonotopes). Chakerian and Filliman [152] obtain several
inequalities for the k-dimensional Lebesgue measure of the orthogonal projection of
the unit n-cube into a k-subspace. It is shown that the studied problem is symmetric
in k and .n−k, and for .k = 2 and .k = n−2 the authors give a sharp upper bound and
also related results for general k. Filliman [268] adds results on zonotopes, among
them also some about special zonotopes which are even orthogonal k-dimensional
projections of n-cubes with .1 ≤ k ≤ n. Also a necessary condition for k-zonotopes
to have largest k-volume among all k-dimensional orthogonal projections of the
n-cube is given. (For the subcase of maximal orthogonal (.n − 1)-dimensional
projections of n-cubes we refer to the paper [553].) In this spirit, a nice concrete
example is derived: the classical rhombic triacontahedron has largest 3-volume
among all orthogonal projections of a 6-cube into 3-space.
For unit n-cubes, Ostrovskii [607] investigates orthogonal projections having
minimal k-volume in k-dimensional linear subspaces. He proves that a k-zonotope
is linearly equivalent to a minimal-volume projection of a unit n-cube if and only if
11.7 Notes 229

it is linearly equivalent to the zonotope spanned by the multiples of rows of a totally

unimodular .n × r matrix of rank k. And for every subspace L, the set of minimal-
volume projections of unit n-cubes in L contains a parallelepiped, see [606]. On
the other hand, for some subspaces there are minimal-volume projections being far
away from parallelepipeds (meant with respect to the Banach-Mazur distance).
Paouris et al. [615] deal with the volume of random projections of the unit n-
cube. They derive a central limit theorem for the volume of the orthogonal projection
of an n-cube onto a random k-dimensional subspace, distributed according to the
appropriate Haar probability measure as .n → ∞.

Further Topics

We continue now with some further aspects of the geometry of the n-cube,
mentioning for each topic only a few essential and representative publications. For
more details we refer to subsection 2.9.1.2 of the recent book [53].
We start here with vertex embedding of simplices in cubes, particularly interesting
since it is equivalent to the question for the existence of Hadamard matrices, for
Jung constants of the form .n/(n + 1) in .l1n , and for a known problem from coding
theory. From the viewpoint of computational convexity, Gritzmann and Klee study
this topic in [329, subsection 9.6], discussing also its relations to the problem of
largest k-simplices in n-cubes. For .k = n, this is equivalent to the Hadamard
determinant problem subsuming the problem above on Hadamard matrices. Largest
simplices in cubes are also studied in the expositions [330] of Gritzmann, Klee and
Larman as well as [408], due to Hudelson, Klee, and Larman. Interesting conjectures
are stated, and also other problems regarding simplices and cubes are posed. More
recent references on Hadamard matrices are treated in the monograph [722] of
Seberry.
The next theme is triangulations of cubes, i.e., minimal (regarding number)
dissections of the n-cube into n-simplices all whose vertices are vertices of the
given n-cube. Good surveys on this topic are given in the books [210, Section 6.3]
of De Loera, Rambau, and Santos, and [841, Chapters 3 and 4] of Zong; see also
the excellent survey [840] of Zong. For more recent references we refer also to the
book [53], more precisely to its subsection 2.9.1.2.
Our next topic is the famous Keller conjecture from 1930: If the Euclidean n-
space is tiled by congruent parallel unit n-cubes, then there exists some pair of
cubes sharing a complete facet. This conjecture was refuted by Lagarias and Shor
[486] for .n ≥ 10; see also the corresponding Mathematical Review MR1155280.
For further reading regarding the time after the appearance of [486] the references
[840] and [841] are again recommended. After a longer period, finally Brakensiek
et al. [122] succeeded (excluding the existence of certain types of graphs) to solve
Keller’s conjecture for all dimensions. And Horak and Kim [402] could show that
extensions of Keller’s conjecture to tiles having a more complex shape cannot be
verified.
230 11 Higher Dimensions

Further interesting topics and themes treating the geometry of the n-cube that one
might study with the help of MathSciNet, Zentralblatt or problem books like [190]
are: decomposing the cube into cubes, cube colorings, different types of separations
of the vertex sets of cubes, Hamming distances, cubes and graphs, cube coverings,
dissections of cubes, packings of or in cubes, the Banach-Mazur distance, and many
further subjects. Also suitably modified cubes (like, e.g., the Klee-Minty cube, see
Gale’s excellent survey [288]) and their applications present interesting fields.

11.7.2 Notes on Simplices

Simplices of dimension n (shortly called n-simplices) are those convex n-polytopes

which have the smallest possible vertex (or facet) number, namely .n + 1. They
are obtained as convex hulls of .n + 1 affinely independent points, and all their k-
dimensional faces are k-simplices, .0 ≤ k ≤ n − 1. For .k = 0 and .k = n − 1
we call these faces vertices and facets, respectively. They are n-pyramids over
each of their facets (and characterized by this property), and the two- and three-
dimensional subcases are called triangles and tetrahedra, respectively. They have
many applications, e.g. in view of triangulations, simplicial complexes, as simple
and simplicial polytopes, and as extremal polytopes or even extremal convex bodies
with respect to many problems in discrete and convex geometry as well as in
polytope theory.
We survey here selected results on simplices which are geometrically intriguing.
Some of these results are not directly connected with the content goals of this book,
but they are hopefully a welcome completion of the high-dimensional geometry
discussed in this chapter.

Books, Book Chapters and Surveys

A comprehensive survey of results on simplices is given in the recent book [53],

where the following subsections contain the most important parts: 2.9.1.1 on
simplices in general, 2.9.1.2 on themes like vertex embeddings or triangulations of
simplices or by simplices in cubes, and 4.7.3 on volumes (of sections) of simplices.
And there are many further survey-like references where simplices are compre-
hensively studied or applied. E.g., the following books contain parts dedicated to
the geometry of tetrahedra or simplices, or concentrate on problems from fields
like elementary geometry, geometric inequalities, non-Euclidean geometries, or
convexity (results from Banach-space theory we ignore here): Simon [733], Roché
and Comberousse [672], Court [180], Couderc and Ballicioni [178], Thébault
[774], Satyanarayana [694], Baston [67], Benson [72], Grünbaum [340], Boltyanski
and Yaglom [101], Kelly and Weiss [446], Lay [493], Mitrinović et al. [578], Part
I and Part II of Berger’s monograph [75], and Matoušek [560]. In several of these
references one has (due to the variety of properties treated) to deal with the index of
11.7 Notes 231

terms, in others partial chapters on simplices are contained. Clearly, this list is not
complete, there are too many further publications and results.
An excellent reference on results and open problems about simplices as extremal
bodies in view of convex geometry are Schneider’s educational talks [700]. The
author treats results and problems mainly in the light of themes like measures
of symmetry with applications, stability results, and various affine inequalities.
Some discussed applications show relations to stochastic geometry, and the affine
inequalities there refer to difference bodies, polars of projection bodies, the
Blaschke-Santaló inequality, the .Lp -Busemann-Petty centroid inequality and the
Brascamp-Lieb inequality and its reverse. A survey on special types of convex
bodies, in its second section also treating simplices in convex geometry, is the
handbook article [391] of Heil and Martini.
An n-simplex whose .n + 1 altitudes have a point O in common is said to
be orthocentric, O being its orthocenter. The introductory part of Edmonds et
al. [237] contains a comprehensive survey on the rich geometry of orthocentric
simplices (studied by many mathematicians). Based on this overview and also the
(partially new) results obtained in [237], the following observation is made: many
theorems from planar triangle geometry have natural extensions/analogues in higher
dimensions if the considered simplices are orthocentric! For general simplices many
theorems do not have such analogues; see the exposition [363], where this viewpoint
is discussed.
We should emphasize the monograph [266] of Fiedler, which nicely combines
simplex geometry with matrix theory and graph theory. Many scattered results on
n-simplices are usefully collected, and used methods and notions are Gram matrices,
methods for solving large systems of linear equations, eigenvalue problems, notions
from simplex geometry (e.g., circumcenters, dihedral angles etc.), and tools like
signed graphs. Special types of simplices (e.g., orthocentric ones) are presented,
and in addition the author shows that his methods can also be used for studying
other geometric objects like finite point sets, simplicial cones, spherical cones, and
also interesting types of degenerate simplices. The author completes this book by the
later paper [267], where with similar methods various problems on pairs of simplices
are solved.

Selected Properties of General Simplices

Any dihedral angle of an n-simplex is determined by the other .(n2 + n − 2)/2

dihedral angles (for the analogous problem regarding edge-lengths, also in non-
Euclidean geometries, see [207]). Maehara and Martini [533] show that for all
dimensions .n ≥ 2 infinitely many similarity classes of n-simplices exist all whose
dihedral angles are rational multiples of .π , where Hill simplices are essential in
this structure. (Recall that a Hill simplex (see, e.g., [354] or [526]) is defined as
follows: Let .x1 , . . . , xn be vectors with the same norm such that the angle between
any two of them is .γ . An n-dimensional Hill simplex of angle .γ is the convex hull
of the origin together with .x1 , x1 + x2 , . . . , x1 + x2 + · · · + xn . Further on, recall
232 11 Higher Dimensions

that a finite set of real numbers is .π -independent if the numbers and .π are linearly
independent over the rational field.) Based on the latter notion, it is verified in [533]
that for .n ≥ 3 there are uncountably many, mutually non-similar n-simplices whose
dihedral angles and .π are linearly independent in that sense, even forming a dense
set (in the Hausdorff sense) within the family of all n-simplices.
Napoleon’s theorem (see the survey [549]) says that if one erects (to the
exterior) equilateral triangles on the sides of an arbitrary triangle, then their three
circumcenters form the vertex set of an equilateral triangle. A direct extension
to n-space is clearly not possible, but if a triangle abc has its Fermat-Torricelli
point f (here meaning to have minimal distance sum to the vertices) in its interior,
then (an analogue of) Napoleon’s theorem can be formulated without these erected
equilateral triangles and extendable to higher dimensions. Namely, the vertices
' ' '
.a , b , c of the finally obtained regular triangle are also the intersection points of

the lines .f a, f b, f c with the circumcircles of the triangles .f bc, f ac, f ab, or
the points on the three rays from .a, b, c to f for which the distance equalities
' ' '
.|aa | = |bb | = |cc | = |f a| + |f b| + |f c| hold, respectively. The n-dimensional

version (.n ≥ 3) of the first viewpoint is obtained by Martini and Weissbach [554],
that of the second one by Hajja et al. [364].
Let K be an n-dimensional convex body of mean width .M(K), and .M(SK ) be
the minimal mean width of all simplices circumscribed about K. Let S denote a
regular simplex circumscribed about the unit n-ball. Böröczky and Schneider [110]
prove the inequality .2M(SK ) ≤ M(K)M(S), where equality holds iff K is a ball.
The dual problem (to find, among all simplices in a given ball, those of maximal
mean width) remains open.
Assuming constant density, Krantz et al. [477] consider centroids of different
k-skeletons of simplices in Euclidean n-space. They start with the well-known fact
that the centroids of the solid simplex and of its vertex set coincide. Then it is shown
that the centroids of the other values for k, namely .1 ≤ k ≤ n − 1, are generically
different. The authors present types of simplices which, for all values of k, have this
coincidence. E.g., for any .n = pq − 1, where p and q are integers larger than 2,
there exists a non-regular n-simplex for which the centroids of all k-skeletons, with
.1 ≤ k ≤ n, coincide. The existence of a non-regular 4-simplex all whose centroids

coincide is also shown, and various further related results are obtained. Here is also
the right place to cite the (relatively unknown) book [54] of Balk and Boltyanski,
also dedicated to types of centroids.
The intersections of lines from the vertices to the points where the incircle
and excircles of a triangle touch the sides are called Gergonne and Nagel centers,
respectively. There is no direct higher dimensional extension since the correspond-
ing lines for n-simplices, .n ≥ 3, are not necessarily concurrent. However, Hajja
[359] presents equivalent definitions that make such higher dimensional analogues
possible. Corresponding nice results are then derived.
Based on earlier results (e.g., due to Butler and Schneider), Martini [545] derives
some characterizations of simplices as special convex bodies in an elementary way.
Let K be an n-dimensional convex body, and xy be an affine diameter of K (i.e.,
a chord of K having maximal length in its own direction u). Then K is said to be
11.7 Notes 233

an oblique double cone with respect to u if every 2-dimensional half-plane whose

bounding line is the affine hull of x and y cuts out of K a (possibly degenerate)
triangle. The following (and further) properties are equivalent to the fact that K is
an n-simplex: (1) For each direction, K is an oblique double cone. (2) For every
translation vector t creating nonempty intersection .K ∩ (K + t), this intersection
is either a point or a positive homothet of K. (3) Every Steiner symmetral of K has
exactly two extreme points outside the corresponding symmetrization hyperplane.
(4) The projection body and the difference body of K √ are polar to each other
regarding the sphere (centered at the origin) of radius . n · V (K). The second
characteristic property is essential in view of the notion of Choquet simplices (see
the basic survey [751]), and V. Soltan [750] derives characterizations of homothetic
simplices by intersections of the type .K1 ∩(K2 +t), where the convex bodies .K1 , K2
belong to at most countably many homothety equivalence classes.
Let again K be an n-dimensional convex body. A finite set X of exterior points
of K is said to see the whole boundary of K if for any boundary point y there is
some .x ∈ X such that K has empty intersection with the open segment xy. It is
proved in [550] that for any exterior point .x1 of K there are n further exterior points
.x2 , . . . , xn+1 such that the point set .x1 , x2 , . . . , xn+1 sees the whole boundary of K.

Further on, it is shown that a convex body K is an n-simplex iff, for every exterior
point, there is a second exterior point such that this two-points set sees the whole
boundary of K.
An asymptotic estimate on the volume ratio .V (S)/V (K) for K an n-dimensional
convex body and S a simplex containing K is obtained by Galicer et al. [289]. If
both have the same centroid, then .(V (S)/V (K))1/n ≤ cn1/2 ; here .c > 0 is an
absolute constant. The dual situation (regarding simplices contained in K) is also
discussed, and a reason for the coincidence of centroids is also given; it has to do
with the Mahler product of the volumes of S and its polar.
Let S be a minimal area n-simplex circumscribed about an n-dimensional strictly
convex body K, and h represents the homothety with ratio .1/(1 − n) whose center
is the centroid of a facet of S. Extending results for two and three dimensions,
Miernowski et al. [572] confirm that the touching points of K with S are precisely
the tangency points of spheres inscribed in images .h(S) of S.

Selected Results on Special Types of Simplices

Denoting by .SI the convex hull of the incenters of the facets of an n-simplex S, Ma
et al. [515] show that the volume ratio of S and .SI satisfies .nn ≤ V (S)/V (SI ).
Equality holds iff S is a regular simplex. They obtain analogous results for convex
hulls of the touching points of the insphere, and for the convex hull of the
circumcenters of the facets of S.
The following deep result is due to Barthe [65]: each convex body in Euclidean
n-space has an affine image whose mean width is not larger than that of a regular
simplex of the same volume.
234 11 Higher Dimensions

Stability versions of Ball’s reverse isoperimetric inequality and of the corre-

sponding inequality for isotropic measures are obtained by K. J. Böröczky and Hug
[109]. They consider affine invariant concepts of distance between convex bodies
(based on the Banach-Mazur and on the volume difference distance) and obtain
characterizations of regular simplices.
The following is due to Frankl and Maehara [276]: Any finite point set F of
cardinality at least 5 has the property that if all triangles formed by triples from F
have the same positive area, then F is the vertex set of a regular simplex. A result
on k-distance sets constructed from such point sets is also given, namely in terms
of different triangle areas. The characterization of regular simplices is extended
by Martini [548] to spherical and hyperbolic space. McMullen [566] calls an n-
simplex k-equiareal if all its k-faces have the same k-volume (thus, due to [276]
any 2-equiareal n-simplex for .n ≥ 4 is regular). Various properties of k-equiareal
simplices are proved in [566]. E.g., an n-simplex is (.n − 1)-equiareal iff its incenter
and centroid coincide. Also, the author constructs (.n − 2)-equiareal n-simplices
not being regular and gives examples of (.n − 2)-equiareal n-simplices which are
not (.n − 1)-equiareal. One important question for .3 ≤ k ≤ n − 2 and .n ≥ 5 is
whether an (.n − 2)-equiareal n-simplex is necessarily (.n − 1)-equiareal. Although
all constructed examples have this property, this is not proved. There exist such
non-regular simplices for .n = 5 and .n ≥ 7, and one for .n = 11 even having
no symmetry. The author confirms also that if the important question above would
have a positive answer, then every k-equiareal n-simplex would be a regular simplex
for .1 ≤ k ≤ n − 3.
Martini and Wenzel [555] replace “equiareality” by congruence. They show that
an n-simplex is regular if for some k with .1 ≤ k ≤ n−2 all its k-faces are congruent,
and that this does not hold for .k = n − 1 > 2. Thus, the simplices with congruent
facets, called equifacetal simplices, remain interesting in this setting. Edmonds
shows in [234] that the isometry group of every equifacetal simplex acts vertex-
transitively (and facet-transitively). It follows that each isometry of an equifacetal
simplex has only one fixed point: the centroid (i.e., the barycenter of the vertices).
He conjectures that any simplex whose isometry group acts vertex-transitively is
equifacetal. There are interesting further results and problems in [234]. And in [235]
it is shown that the set of centers of a simplex is a subset of the fixed point set of
its group of isometries. It is a single point iff this group acts transitively on the
vertices, implying that an equifacetal simplex has a unique center. This yields the
known center conjecture: If a simplex has a unique center, then it is equifacetal.
For the dimensions at most six Edmonds proves a strong version of this conjecture.
Associated with any equifacetal n-simplex is also a well-defined partition of n that
counts the number of edges of each possible length incident at a given vertex. The
related partition problem asks for a characterization of the partitions coming from
equifacetal simplices. In [236] this problem is solved, and in this setting extremal
equifacetal simplices are characterized.
An n-simplex S is called replicating if one can partition it into at least two
simplices such that all simplex tiles are mutually congruent and similar to S.
The notion of Hill-simplex is defined above and goes (for n dimensions) back to
11.7 Notes 235

Hadwiger [354]. Every Hill n-simplex is replicating, and hence all such simplices
tile n-spce. Hertel (see [393] and [394]) obtained related results and conjectured
that a tetrahedron is replicating iff it is a Hill tetrahedron. Maehara [525] constructs
counterexamples to this conjecture and shows even that for any .n ≥ 3 there are
replicating n-simplices which are not Hill simplices. In [526] he calls a simplex
reflection-symmetric if it is invariant under some orientation reversing isometry.
He characterizes all reflection-symmetric Hill simplices, and for this purpose he
explicitly provides all isometries that preserve any given Hill simplex, finding also
the orientation reversing cases.
It is well-known that a triangle is regular or equilateral if two of its classical
centers (namely the centroid, incenter, circumcenter, and orthocenter) coincide.
Edmonds et al. [238] aim to extend this concept to higher dimensions, asking for
consequences in the facial structure of simplices if two or more centers coincide.
They add also the Fermat-Torricelli point, having minimum distance sum to the
vertices, and replace the orthocenter, in general not existing for .n ≥ 3, by the
Monge point. It turns out that in 3-space one gets characterizations of isosceles
tetrahedra, and that for .n ≥ 4 a more complicated machinery is necessary to get
analogous results. If two of the centers coincide, one can geometrically describe
consequences for their facial structure, but one cannot usually infer much about
other centers. Nevertheless, the authors give a modern and unifying presentation,
including many new and also known results in the plane and in 3-space. They
continue these investigations in [237], getting much more for the subfamily of
orthocentric simplices (i.e., simplices which really have an orthocenter). It is
confirmed that orthocentric simplices, either with congruent faces, or with any pair
from (centroid, circumcentre, incentre, orthocentre) coinciding, have to be regular.
The used tools are mainly Gram matrices and a characterization of the barycentric
coordinates of the orthocenter. Hajja [360] studies degrees of regularity of simplices
by using coincidences of two or three centers. An n-simplex is called biregular if
the set of fixed points of its motion group is a line (or, in degenerate cases, a point,
yielding the subcase or regular simplices). Solving an old open problem at least for
orthocentric simplices of arbitrary dimension, Edmonds et al. [239] prove that the
incenter of an orthocentric simplex lies on its Euler line if and only if this simplex
is biregular.
In [365] and [366] the following essential types of n-simplices are studied:
orthocentric simplices (having concurrent altitudes), circumscriptible simplices (for
which there exists an (.n − 1)-sphere that touches all edges internally), isodynamic
simplices (for which the cevians that join the vertices to the incenters of the opposite
facets are concurrent), and isogonic simplices (a higher-dimensional generalization
of isogonic tetrahedra, for which the cevians that join the four vertices to the points
where the insphere touches the opposite faces are concurrent). In the first paper
[365] it is shown that the intersection between any two such families yields precisely
the class of n-kites (these are simplices with n of their vertices forming a regular
(.n − 1)-simplex whose vertices are equidistant from the remaining vertex), and in
[366] several characterizations of these families are derived or reproved.
236 11 Higher Dimensions

Bédaride and Rao [69] investigate the billiard flow inside a regular n-simplex,
n > 3. It is shown that there are two types of periodic trajectories, the first of which
.

has period .n+1 and hits each facet once. The second one, with period 2n, hits one of
the facets n times, and thus any other facet once. In both cases, for the points where
the trajectory hits the simplex boundary, the exact coordinates are determined. The
authors also study the convex hull of the periodic orbit of length .n + 1, and it is
shown that the orbits obtained for .n = 3 are stable with respect to perturbations of
the respective tetrahedron.
Pompeiu’s theorem says that the distances from the vertices of an equilateral
triangle to any point in its plane yield the side lengths of some triangle. Al-Afifi
et al. [14] generalize this to given regular n-simplices. The authors start with the
classical Möbius-Pompeiu theorem and an extension to regular n-simplices obtained
by Fiedler. Presenting their new generalization, they show that for a regular n-
simplex, .n ≥ 2, and a point x in its affine hull, the distances from x to the vertices
of the given simplex can be taken as the volumes of the facets of some n-simplex
precisely if x is not a vertex of the considered regular simplex. For this purpose,
they prove a higher-dimensional analogue of the triangle inequality.

11.7.3 Notes on the Volume of (Unit) Balls

High Dimensional Phenomena

In Sect. 11.3 above we consider the volume of a unit ball. It is well-known that
if .n → ∞, this quantity goes down to zero and, at the same time, concentrates
more and more on the thin shell which is close to its spherical boundary. This
can be verified in different ways (e.g., by methods from functional analysis,
probability theory, Euclidean geometry, convex geometry, and even combinatorics).
An interesting further and recent way is due to Li [505]. Namely, he uses regularity
theory of elliptic partial differential equations and calculus of variations. And
Hayes [387] (in one of the best mathematical writings of 2012) describes nicely
the known behaviour of an n-cube C of edge-length 2 and an inscribed n-ball B of
radius 1 if the dimension increases (see also the Mathematical Review MR2976645).
The author shows the maximality of dimension 5 with respect to the volume of B
(and of dimension 7 regarding its surface area). Since B remains inscribed, already
for a dimension number like .n = 100, say, the part of C inside B is close to zero,
although this ball further on remains in touching position with each facet of C, and
no larger ball is inside C. Thus, the ball somehow “disappears”!
Another nice representation of such phenomena can be found in Artin’s classical
monograph [41] and Section 13.1 of Matoušek’s book [560], and Rabiei [652]
presents related mass concentration phenomena in balls. There also exact formulas
for the mass content of ball caps and upper bounds for related ratios of volumes are
given, cf. also again Chapter 13 in [560]. In addition we refer here to Gipple [301],
where a recursion formula for the volume of the ball is given, and in [746] different
approaches, for n even and n odd, to the volume of the n-dimensional ball of radius
.r > 0 are shown. Clearly, there are various further such derivations in the literature.
11.7 Notes 237

Inequalities

It might be interesting to add some results on inequalities giving bounds on the

volume of the unit ball in Euclidean space. Based on earlier results, Alzer [30]
derives sharp upper and lower bounds on this volume, and quotients of it using
the values for different dimensions are given, too. His tools are the gamma function
and its logarithmic derivative. He continues in [31], presenting there also a related
interesting double inequality. Further results in this direction are presented by Chen
and Lin [160] and Yin [826], namely sharp upper and lower bounds on ratios of
suitably combined volumes of balls of different dimensions. Anderson et al. [32]
expose analytical inequalities involving special functions from classical analysis
and geometric function theory. Among them the gamma function is discussed, with
extensions and again its role regarding the volume of the unit ball.

On Ball Volumes (Mainly) in Three Dimensions

It is well-known that Archimedes [39] discovered the following result for three
dimensions: The volume of a ball is two thirds of the volume of the smallest right
cylinder containing it, and the surface area of the ball is also two thirds of the total
surface area of this cylinder. Thus, the ratio "volume/surface area" of this cylinder
equals the corresponding ratio of the considered ball. (As third convex body,
sometimes a double cone suitably inscribed to the cylinder and, clearly, with volume
a third of that of the cylinder, is added.) This nice result is discussed in the book
sections [250, Section 1.5, Problem 8] of Eves and [103, Section 4.3] of Borceux.
We refer also to the books of Stein [754, Chapter 10] and Aumann [47], the latter
nicely collecting and presenting almost all famous contributions of Archimedes to
mathematics on a good level for high school students and undergraduates.
A result somehow converse to that of Archimedes is due to Knothe [459], again
holding in 3-space: If the volume of each right cylinder circumscribed about a
convex body K is one and a half times that of K, then K is a ball. Firey [269]
extends this slightly characterizing bodies of constant width (see the book [558]
for this class of convex bodies): Let .C(u) be the cylinder whose generators have
direction u and which circumscribes a convex body K (thus, the generators are
supporting lines of K). Denote by .S(u) the lateral area of the part of .C(u) lying
between the two supporting planes of K orthogonal to u. If .S(u) is constant over
all directions, then K is a body of constant width. This yields an easy proof of
Knothe’s theorem. If K is the ball inscribed to .C(u), and .V (C(u)), V (K), A(C(u))
and .A(K) denote the respective volumes and areas, then the ratio .V (K)/V (C(u))
equals .A(S)/SA(C(u)). This is confirmed by Du Val and Saban [232] and also
suitably generalized to n dimensions. Groemer [332] generalizes Knothe’s result to
n dimensions and derives interesting stability versions. These can be formulated,
e.g. in dimension 3, as follows: if the lateral boundary of any cylinder has area close
to the surface area of K, then K is close to a ball. Furthermore, the author gives
238 11 Higher Dimensions

upper and lower bounds for the average of the second intrinsic volumes of all right
cylinders circumscribed to K.
The authors of [215] study the analogues of the above ratios of Archimedes for all
small geodesic spheres of Riemannian manifolds. Then they continue in proving that
these ratios determine the geometry of any two-point homogeneous space inside the
family of all Riemannian manifolds. Archimedes’ balancing methods led them to the
volumes of balls and cylindrical wedges. Apostol and Mnatsakanian [36] introduce
new balancing methods and derive interesting relations involving volumes and
surface areas of certain types of solid bodies. This helps also to extend Archimedes’
classical results on spheres and cylinders to higher dimensions. Besides all this, new
related results on centroids of hemispheres are obtained in [36]. The same authors
continue in [37]. Namely, inspired by Archimedes’ equality of ratios, they introduce
classes of solids called globes and having the property that the ratio of volume to
surface area is invariant.
The two papers [151] and [699] are slightly related to our presentation here.
Replacing the circumscribed cylinders in our argumentation by suitably circum-
scribed boxes, their mean volume is considered. It is minimal when K is a ball (see
Chakerian [151]), and Schneider [699] obtains a similar result for averaged surface
areas.
Richter and Schicker [668] extend the interesting notion of ball number,
combining the volume and the surface area of a ball of radius .r > 0 with extensions
to normed spaces. They consider the three-dimensional case with the regular
polyhedra as unit balls (including also the non-symmetric tetrahedral gauge). In
the introduction a survey on related results in this direction is presented.

Further Results

Some monotonicity and concavity properties of several functions, strongly related

to the volume of the unit ball, are generalized by Guo and Qi [348], and properties of
corresponding sequences are also concluded. Further related papers are [347] and
[651]. Also Mortici [586] derives new bounds for ratios involving the volume of
the unit ball. In addition, he shows related monotonicity properties and discusses
problems around the surface area of the unit ball, too.
In a detailed manner, Dodd and Coll [216] describe hypersurfaces in n-
dimensional Euclidean space satisfying higher dimensional analogues of the
so-called equal area zones property of the sphere, saying that the surface area of a
zone between two parallel hyperplanes depends only on the distance between these
hyperplanes. These studied objects are called equizonal n-ovaloids, and they are
hypersurfaces of revolution. Interesting further problems are posed and discussed
in [216], e.g. regarding the volume of ovaloids. E.g., the authors prove that the
ratio of this volume for radius 1 to that of the unit ball approaches .π as .n → ∞.
Further on, equizonal n-ovaloids are of class .C ∞ for even dimension, and only of
class .C (n−1)/2 for the odd case. This is improved in [175], where it is verified that,
for even n, equizonal n-ovaloids are analytic, and that for odd dimensions they are
everywhere analytic except at their two poles (being there of class .C n−1 ).
Chapter 12
The Cayley-Menger Determinant

Poetry is the art of giving different names to the same thing;

Mathematics is the art of giving the same name to different
things.
— Henri Poincaré

In this chapter, the Cayley-Menger determinant for a finite semi-metric space

is introduced. Using Cayley-Menger determinants, Soddy’s formula for mutually
tangent spheres and Ptolemy’s classical theorem are generalized to n dimensions.

12.1 The Cayley-Menger Matrix

A semi-metric space .X is a set in which the distance .d(x, y) is defined for every
x, y ∈ X so that
.

d(x, y) = d(y, x) ≥ 0 and d(x, y) = 0 ⇔ x = y

.

hold. If, furthermore, the distance .d(x, y) satisfies the triangle inequality, then .X is
a metric space. Every subset of a semi-metric space is a semi-metric space by the
same distance. Every subset of .Rn is a (semi-)metric space by the distance .d(x, y) =
‖x −y‖. Two semi- metric spaces .X , Y are said to be isometric if there is a bijection
.f : X → Y such that .d(x, y) = d(f (x), f (y)) holds for every .x, y ∈ X . If .X is

isometric to a subset of .Rn , then .X is said to be isometrically embeddable in .Rn .

For a semi-metric space .X = {x1 , x2 , . . . , xk } with k elements, the .k × k-matrix
whose .(i, j )-entry is the squared distance .d(xi , xj )2 for .i, j = 1, 2, . . . k, is called

© The Author(s), under exclusive license to Springer Nature Switzerland AG 2024 239
H. Maehara, H. Martini, Circles, Spheres and Spherical Geometry,
Birkhäuser Advanced Texts Basler Lehrbücher,
https://doi.org/10.1007/978-3-031-62776-7_12
240 12 The Cayley-Menger Determinant

the squared-distance matrix of .X , and denoted by .SD(X ). And the .(k +1)×(k +1)-
matrix
⎛ ⎞
1
⎜ .. ⎟
⎜ .⎟
.𝚪(X ) = ⎜ SD(X ) ⎟
⎝ 1⎠
1 ··· 10

is called the Cayley-Menger matrix of .X . Cayley-Menger matrices are real

symmetric matrices. For example, for the semi-metric space .X = {x1 , x2 , x3 , x4 }
with distances .dij = d(xi , xj ), 1 ≤ i, j ≤ 4, we have
⎛ ⎞
⎛ 2 2 2 ⎞ 0 2
d12 2
d13 2
d14 1
0 d12 d13 d14 ⎜d 2 2 2 1⎟
⎜d 2 0 2 2 ⎟ ⎜ 21 0 d23 d24 ⎟
d23 d24 ⎜ 2 ⎟
.SD(X ) = ⎜ 21 ⎟ 𝚪(X ) = ⎜d31 1⎟ .
2 2
⎝d 2 d 2 0 2 ⎠,
d34 ⎜ 2
d32 0 d34
⎟
31 32
2 d2 d2 0 ⎝d41 2
d42 2
d43 0 1⎠
d41 42 43 1 1 1 1 0

The determinant .det 𝚪(X ) is called the corresponding Cayley-Menger determi-

nant. If we permute the order of the points .x1 , . . . , xk , then the order of the rows
and columns of .𝚪(X ) are permuted, but there arises no essential difference.
Lemma 12.1 Let .X = {x1 , x2 , x3 } be a semi-metric space with distances .a =
d(x1 , x2 ), b = d(x1 , x3 ), c = d(x2 , x3 ). If .det 𝚪(X ) = 0, then .X is isometrically
embeddable in .R1 .
Proof The determinant of .𝚪(X ) is calculated as follows.
| | | |
|0 a 2 b2 1|| || 0 a2 b2 1|| | 2 |
| 2 |a −a 2 c2 − b2 |
|a 0 c2 1|| ||a 2 −a 2 c2 − b2 0|| | |
.| = = − ||b2 c2 − a 2 −b2 ||
c2 0 1|| ||b2 c2 − a 2 −b2 0||
|b 2
| |1 1 |
|1 1 1 0| | 1 1 1 0|
1
| 2 |
|a −2a 2 c2 − a 2 − b2 ||
|
= − ||b2 c2 − a 2 − b2 −2b2 | = −4a 2 b2 + (c2 − a 2 − b2 )2
|
|1 0 0 |

= −(2ab − c2 + a 2 + b2 )(2ab + c2 − a 2 − b2 )
= −((a + b)2 − c2 )(c2 − (a − b)2 )
= −(a + b + c)(a + b − c)(c − a + b)(c + a − b).

Hence, .det 𝚪(X ) = 0 implies that one of .a + b − c, c − a + b, c + a − b is 0. In

the case .a + b − c = 0, the correspondence .x1 → a, x2 → 0, x3 → a + b gives an
isometric embedding of .X into .R1 . Other cases are treated similarly. ⨆
⨅
12.1 The Cayley-Menger Matrix 241

Remark 12.1 Cayley-Menger determinants can be used to make it clear whether a

given finite semi-mertic space can be isometrically embedded in a Euclidean space.
Chapter 4 of L. M. Blumenthal’s book [96] is devoted to the study of relevant related
theory.
Theorem 12.1 For any finite subset .X ⊂ Rn , the rank of the Cayley-Menger
matrix .𝚪(X ) is at most .n + 2. Hence, if .X contains more than .n + 1 points, then
.det 𝚪(X ) = 0.

Proof Let .X = {x1 , x2 , . . . , xk }. Let X be the .(k − 1) × n-matrix whose .(i − 1)th
row is .x−−
→
1 xi for .i = 2, 3, . . . , k:

⎛−−→⎞
x1 x2
⎜x−−→⎟
⎜ 1 x3 ⎟
.X = ⎜ . ⎟ .
⎝ .. ⎠
x−−→x1 k

Since each .x−−

→
1 xi is an n-dimensional vector, it is obvious that the rank of X is at most
n. Therefore, the .(k − 1) × (k − 1)-square matrix .XXt (where .Xt is the transpose
of X) has rank at most n. The .(i, j )-entry of .XXt is .−x1−x−i+1
→·−x−−−→
1 xj +1 (inner product
of two vectors). Since

dij2 = −
. x−→ −−→ −−→ −−→ −−→ −−→
i xj · xi xj = (xi x1 + x1 xj ) · (xi x1 + x1 xj )

= (−x−−
→ −−→ −−→ −−→ 2 2 −−→ −−→
1 xi + x1 xj ) · (−x1 xi + x1 xj ) = d1i + d1j − 2(x1 xi · x1 xj ),

we have .2(x−−
→ −−→
1 xi · x1 xj ) = d1i + d1j − dij . Put
2 2 2

⎛ ⎞
0 0 ... 0 1
⎜0 0⎟
⎜ ⎟
⎜. .. ⎟ .
.M = ⎜ . t
⎜ . 2(XX ) .⎟⎟
⎝0 0⎠
10 ... 00

This matrix M is a square matrix of degree .k + 1, and since the rank of .XXt is
at most n, the rank of M is at most .n + 2. By adding the first row to the ith row,
.i = 2, 3, . . . , k, and then by adding the first column to the j th column for .j =

2, 3, . . . , k, M is changed to
⎛ ⎞
0 0 ... 0 1
⎜0 1⎟
⎜ ⎟
⎜ .. .. ⎟ .
.M1 = ⎜ t ⎟
⎜ . 2(XX ) . ⎟
⎝0 1⎠
11 ... 10
242 12 The Cayley-Menger Determinant

Since the .(i, j )-entry of .2(XXt ) (the .(i + 1, j + 1)-entry of .M1 ) is .di1
2 + d2 − d2 ,
1j ij
by subtracting .di1 2 times the last row from the ith row of .M for .i = 2, . . . , k, and
1
then by subtracting .dj21 times the last column from the j th column for .j = 2, . . . , k,
.M1 becomes

⎛ ⎞
0 −d12
2 −d 2
13 ... −d1k
2 1
⎜−d 2 0 −d232 −d2k
2 1⎟
⎜ 21 ... ⎟
⎜ 2 ⎟
⎜−d31 −d32
2 0 ... −d3k
2 1⎟
.M2 = ⎜ .⎟
⎜ .. ⎟.
⎜ ··· ··· ··· . · · · .. ⎟
⎜ 2 ⎟
⎝−dk1 −dk22 . . . −d 2
k(k−1) 0 1⎠
1 1 ··· 1 1 0

Finally, by mutiplying all rows of .M2 with .−1, and then the last row and the last
column with .−1, .M2 becomes .𝚪(X ). Since the rank is invariant under such changes,
the rank of .𝚪(X ) is equal to (the rank of .XXt ).+2. Hence the rank of .𝚪(X ) is at most
.n + 2. ⨆
⨅
The matrix .XXt used in the above proof is called the Gram matrix of the .k − 1
vectors .x−−
→
1 xi , i = 2, 3, . . . , k. The above proof shows that

.the rank of 𝚪(X ) = (the rank of XXt ) + 2.

12.2 Odd Integral Distances

As an application of Theorem 12.1, let us show here the following theorem.

Theorem 12.2 ([317]) There is no four-point-set in the plane in which the distances
between the points are all odd integers
Proof The proof works by contradiction. Suppose that there is a four-point-set
.X = {x0 , x1 , x2 , x3 } in the plane .R2 in which all distances between points are
odd integers. Its Cayley-Menger matrix .𝚪(X ) is of degree 5, and its rank is at most
.n + 2 = 2 + 2 = 4 by Theorem 12.1. Thus, .det 𝚪(X ) = 0. Hence, the determinant

must be 0 when it is computed modulo 8. However, since .(2k+1)2 = 4k 2 +4k+1 =

4k(k + 1) + 1 ≡ 1 (mod 8), if we compute the determinant modulo 8,
| |
|
|0
| |0 1 1 1 1 0||
1| ||
|
| 0||
1 1 1
|1 1|| ||
1 0 1 1 1
| 0 1 1 |
| | |1 1 0 1 1 0|
. det 𝚪(X ) = |1 1 0 1 1| = | |
| | |1 1 1 0 1 0|
|1 1 1 0 1| | |
| | |1 1 1 1 0 0||
|1 1 1 1 0| ||
0 0 0 0 0 1|
12.3 Orthogonal-Sphere Systems 243

| | | |
|0 1 1 1 1 0| |−1 0 0 0 0 −1||
| | |
|1 0 1 1 1 0| | 0 −1 0 0 0 −1||
| | |
| | | |
|1 1 0 1 1 0| | 0 0 −1 0 0 −1|
=| |=| |
|1 1 1 0 1 0| | 0 0 0 −1 0 −1|
| | | |
|1 1 1 1 0 0| | 0 0 0 0 −1 −1||
| | |
|1 1 1 1 1 1| | 1 1 1 1 1 1|
| |
|−1 0 0 0 0 −1|
| |
| 0 −1 0 0 0 −1|
| |
| |
| 0 0 −1 0 0 −1|
=| | = 4 /= 0,
| 0 0 0 −1 0 −1|
| |
| 0 0 0 0 −1 −1|
| |
| 0 0 0 0 0 −4|

yielding a contradiction. ⨆
⨅
.R
Similarly, it can be proved that there is no five-point-set in 3
with all integral
distances being odd, see the paper [317] of R. L. Graham, B. L. Rothschild, and
E. G. Straus. They proved that there is an .(n + 2)-point-set in .Rn with all integral
distances being odd if and only if .n ≡ 14 (mod 16).

12.3 Orthogonal-Sphere Systems

Two spheres .σ1 , σ2 in .Rn with centers .p1 , p2 and radii .r1 , r2 are called orthogonally
crossing if

‖p1 − p2 ‖2 = r12 + r22 .

. (12.1)

Thus, for any intersection point .x of .σ1 and .σ2 , ./ p1 xp2 = 90◦ holds. A family of
spheres is called an orthogonal-sphere system if every pair of spheres in the family
is orthogonally crossing.
Example 12.1 In .Rn , n unit spheres with centers

(1, 0, 0, . . . , 0), (0, 1, 0, . . . , 0), . . . , (0, 0, . . . , 0, 1)

.

√
and a sphere with center .(−1, −1, . . . , −1) and radius . n + 2 form together an
orthogonal-sphere system. This can be seen by checking (12.1) for every pair of
spheres.
Theorem 12.3 The maximum number of spheres in an orthogonal-sphere system in
Rn is .n + 1.
.
244 12 The Cayley-Menger Determinant

Proof As seen in Example 12.1, there is an orthogonal-sphere system in .Rn

consisting of .n + 1 spheres.
Let .{σ1 , σ2 , . . . , σk } be an orthogonal-sphere system of k spheres in .Rn .
Let .pi , ri be the center and the radius of .σi for .i = 1, 2, . . . , k. Put .X =
{p1 , p2 , . . . , pk } and let .dij be the distance between .pi and .pj . Then, by (12.1),
ij = ri + rj . Hence, for .i, j ≤ k the .(i, j )-entry of the Cayley-Menger matrix
.d
2 2 2

.𝚪(X ) is .r + r . For .1 ≤ i, j ≤ k, by subtracting from the ith row of .𝚪(X ), .r

2 2 2
i j j
2
times the last row, and subtracting from the j th column .rj times the last column,
we have
| |
|−2r 2 0 . . . 0 1|
| |
| 0 1 −2r 2 . . . 0 1|
| |
| 1
|
. det 𝚪(X ) = | ... . . . 1|
| |
| 0 . . . . . . −2rk2 1|
| |
| 1 1 . . . 1 0|

| |
|−2r 2 0 . . . 0 1 ||
| 1
| 0 −2r 2 . . . 0 1 ||
| 1
| 1 || /= 0.
=| ... ...
| 0 . . . . . . −2rk 1 ||
2
| ∑ 1 |
|
| 0 0 ... 0 2|2ri

Since the rank of .𝚪(X ) is at most .n+2 by Theorem 12.1, we must have .k+1 ≤ n+2.
Therefore, .k ≤ n + 1. ⨆
⨅

12.4 Tangent-Sphere Systems

Two spheres in .Rn (two circles for .n = 2) are tangent if they share exactly one
point, and this common point is the contact point of the two spheres. If two spheres
have the radii .r1 , r2 , then the two spheres are tangent if and only if

(the distance between the centers of the sphers) = |r1 ± r2 |,

.

where the signs .+, − correspond to “externally tangent" or “internally tangent",

respectively.
A family of spheres in .Rn is called a tangent-sphere system if the spheres are
mutually tangent, and no three spheres are tangent at the same point. Thus, in a
tangent-sphere system in .Rn , if a sphere .σ encloses another sphere (that is, the ball
bounded by .σ contains another sphere), then .σ encloses all other spheres.
12.4 Tangent-Sphere Systems 245

Example 12.2 In .Rn , .n+1 unit spheres centered at the vertices of an n-dimensional
regular simplex with edge-lengths 2 are mutually tangent. And there is a sphere of
suitable radius, centered at the center (.= centroid) of the regular simplex that is
tangent to all .n + 1 unit spheres. Thus, there are mutually tangent .n + 2 spheres in
.R that form a tangent-sphere system.
n

Let .{σ1 , σ2 , . . . , σk } be a tangent-sphere system in .Rn . Let us define the signed

radius .ρi of each .σi as follows:
⌠
−(the radius of σi ) (if σi encloses another σj ),
ρi =
.
+(the radius of σi ) (otherwise).

By using the signed radius, the distance between the centers of tangent spheres .σi
and .σj can be always represented by .|ρi + ρj |, no matter whether they are internally
tangent or externally tangent.
Descartes’ circle theorem and Soddy’s formula in Chap. 2 are generalized in the
following way.
Theorem 12.4 The maximum number of spheres in a tangent-sphere system in
.Rn , n ≥ 2, is equal to .n + 2. If .{σ1 , σ2 , . . . , σn+2 } is a tangent-sphere-ststem in
.R , then
n

(⎲
n+2
1 )2 ⎲
n+2
( 1 )2
. =n (12.2)
ρi ρi
i=1 i=1

holds, where .ρi is the signed radius of .σi .

Preceding the proof of Theorem 12.4, we present an example to compute a
determinant.
Example 12.3 To compute the determinant
| |
|−1 1 1 . . . 1 a1 ||
|
|1 −1 1 . . . 1 a2 ||
|
| |
| . |
| . . . .. ... |
.D = | |,
| .. |
| ... . ... |
| |
|1 1 1 . . . −1 ak ||
|
| a1 a2 a3 . . . ak 0 |
246 12 The Cayley-Menger Determinant

we use a method to increase the degree of the matrix. Namely,

| | | |
|−1 1 1 . . . 1 a1 1|| ||−2 0 0 . . . 0 a1 1||
|
|1 −1 1 . . . 1 a2 1|| || 0 −2 0 . . . 0 a2 1||
|
| | | |
| .
. . . .. 1|| ||
.
. . . .. 1||
| ... ...
| | | |
.D = | .. |=| .. |
| ... . ... | | ... . ... |
| | | |
|1 1 1 . . . −1 ak 1| | 0 0 0 . . . −2 ak 1|
| | | |
| a1 a2 a3 . . . ak 0 0|| || a1 a2 a3 . . . ak 0 0||
|
|0 0 0 . . . 0 0 1| |−1 −1 −1 . . . −1 0 1|
| |
|−2 0 0 ... 0 a1 1 |
| |
|0 −2 0 . . . 0 |
| a2 1 |
| |
| .
. . . .. |
| ... 1 |
| |
=| .. |
| ... . ... |
| |
|0 0 0 . . . −2 ak 1 |
| ∑ 2 ∑ |
|0 0 0 ... 0 (ai /2) (ai /2)||
| ∑
|0 0 0 ... 0 (−ai /2) 1 − k/2 |

| 1∑ 2 ∑ |
| ai || (−2)k ⎛( ⎲ )2 ⎲ ⎞
1 k k
a
= (−2) || 2 1 ∑ i
k 2
| = ai − (k − 2) a 2
i .
− 2 ai 2 (2 − k)
1
4
i=1 i=1

Remark 12.2 Similarly to the above computation, we can see that the determinant
of the square matrix of degree k whose entries in the main diagonal are all .−1, and
whose remaining entries are all 1, is equal to

(−2)k (1 − k/2),
.

and hence, if .k ≥ 3, then such a matrix is non-singular.

Proof of Theorem 12.4 We already showed that there is a tangent-sphere system of
.n + 2 spheres in .Rn . Now, let .{σ1 , σ2 , . . . , σk } be a tangent-sphere system in .Rn .
Let .ρ1 , ρ2 , . . . , ρk be the signed radii of these spheres, and .X = {p1 , p2 , . . . , pk }
be the set of their centers. Let us compute the Cayley-Menger determinant of .X .
Let .dij denote the distance between .pi and .pj . Then .dij = |ρi + ρj |, and hence
ij = ρi + ρj + 2ρi ρj . For .i, j = 1, 2, . . . , k + 1, by subtracting .ρi times the
.d
2 2 2 2
12.4 Tangent-Sphere Systems 247

last row from the ith row, and by subtracting .ρj2 times the last column from the j th
column, we have
| |
| −2ρ 2 2ρ ρ . . . . . . 2ρ ρ 1|
| 1 1 2 1 k |
|2ρ ρ −2ρ 2 . . . . . . 2ρ ρ 1|
| 2 1 2 k |
| 2
|
| .. |
| ... ... . ... ... |
. det 𝚪(X ) = | |
| .. |
| ... ... ... . ... |
| |
|2ρk ρ1 2ρk ρ2 . . . 2ρk ρk−1 −2ρk2 1|
| |
| 1 1 ... ... 1 0|
| |
| −2 2 . . . . . . 2 1/ρ |
| 1|
| 2 −2 . . . . . . 2 1/ρ |
| 2|
| |
| .. |
| . . . . . . . . . . . . . |
= (ρ1 ρ2 . . . ρk )2 | |
| .. |
| ... ... ... . ... |
| |
| 2 2 . . . 2 −2 1/ρk ||
|
|1/ρ1 1/ρ2 . . . . . . 1/ρk 0 |
| |
| −1 1 . . . . . . 1 1/ρ |
| 1|
| 1 −1 . . . . . . 1 1/ρ |
| 2|
| |
| .. |
2 | ... ... . ... ... |
= 2 (ρ1 ρ2 . . . ρk ) |
k−1
|.
| .. |
| ... ... ... . ... |
| |
| 1 1 . . . 1 −1 1/ρk ||
|
|1/ρ1 1/ρ2 . . . . . . 1/ρk 0 |

In this determinant, the .(k × k)-square matrix obtained from it by removing the last
row and the last column (in which the entries in the main diagonal are all .−1 and
the remaining entries are all 1) is nonsingular for .k ≥ 3 by Remark 12.2. Hence the
rank of .𝚪(X ) is at least k. Since the rank of .𝚪(X ) is at most .n + 2 by Theorem 12.1,
we can deduce that .k ≤ n + 2. Thus, the number of mutually tangent spheres with
distinct contact points is at most .n + 2.
The last determinant can be computed as in Example 12.3:

(−2)k ⎛( ⎲ 1 )2 ⎲ 1⎞
k k
. det 𝚪(X ) = 2k−1 (ρ1 ρ2 . . . ρk )2 − (k − 2) .
4
i=1
ρi ρi2
i=1
248 12 The Cayley-Menger Determinant

Therefore, if .k = n + 2 (that is, .{σ1 , σ2 , . . . , σn+2 } is a family of mutually tangent

spheres with distinct contact points), then .𝚪(X ) is a square matrix of degree .n + 3,
and hence .det 𝚪(X ) = 0 by Theorem 12.1. Therefore,

⎛⎲
n+2
1 ⎞2 ⎲
n+2
1
. −n =0
ρi
i=1
ρi2
i=1

holds. ⨆
⨅

12.5 A Generalization of Ptolemy’s Theorem

The inversion of .Rn with respect to a sphere with center .p and radius r is the
transformation of .Rn \ {p} that sends .x ∈ Rn \ {p} to a point .x ' on the ray .− →
px
such that .|px| · |px ' | = r 2 . Similarly to Theorem 1.8 on inversions of .R3 , it can
be easily proved that an inversion of .Rn maps a sphere to a sphere or a hyperplane.
(Instead of “hypersphere” we use also here the word “sphere”.) More precisely,
• a sphere that does not pass through the center p of the inversion is transformed
to a sphere that does not pass through p.
• a sphere that passes through p is transformed to a hyperplane that does not pass
through p and vice versa, and
• a hyperplane that passes through p is transformed to itself.
Lemma 12.2
| |
|0 a2 b2 c2 ||
| 2
|a 0 d2 e2 ||
.| = (af + be + cd)(af + be − cd)(af − be + cd)(af − be − ce).
|b 2 d2 0 f 2 ||
|
| c2 e2 f2 0|

Proof Expanding the left-hand side determinant along the first column, we get
(af )4 + (be)4 + (cd)4 − 2(af be)2 − 2(af cd)2 − 2(becd)2 . On the other hand,
.

the right-hand side can be expanded in the following way:

(af + be + cd)(af + be − cd)(af − be + cd)(−af + be + cd)

.

= [(af + be)2 − (cd)2 ][(af − be)2 − (cd)2 ]

= [(af )2 + (be)2 − (cd)2 + 2(af be)][(af )2 + (be)2 − (cd)2 − 2(af be)]
= [(af )2 + (be)2 − (cd)2 ]2 − 4(af be)2
= (af )4 + (be)4 + (cd)4 − 2(af be)2 − 2(af cd)2 − 2(becd)2 .

⨆
⨅
12.5 A Generalization of Ptolemy’s Theorem 249

Theorem 12.5 An .(n + 2)-point-set .X in .Rn lies on a sphere or a hyperplane in

.R if and only if .det SD(X ) = 0.
n

Remark 12.3 For a 4-point-set .X = {x1 , x2 , x3 , x4 } in .R2 , the determinant

.det SD(X ) of the .4 × 4-matrix .SD(X ) can be factorized by Lemma 12.2, as

.(d12 d34 + d13 d24 + d23 d14 )(d12 d34 − d13 d24 + d23 d14 )
× (d12 d34 + d13 d24 − d23 d14 )(d12 d34 − d13 d24 − d23 d14 ),

where .dij = d(xi , xj ). Hence, .SD(X ) = 0 if and only if (just) one of

d12 d34 − d13 d24 + d23 d14 ,

.

d12 d34 + d13 d24 − d23 d14 ,

−d12 d34 + d13 d24 + d23 d14

is 0. Thus, the case .n = 2 of Theorem 12.5 implies Ptolemy’s classical theorem.

Proof of Theorem 12.5 Let .X = {x1 , x2 , . . . , xn+1 , xn+2 }, and denote .xn+2 by .z.
Let .ϕ be the inversion of .Rn with respect to the unit sphere centered at .z. Put .yi =
ϕ(xi ), i = 1, 2, . . . , n + 1, and put .Y = {y1 , y2 , . . . , yn+1 }. Then .‖yi − z‖ =
1/‖xi − z‖ for .i = 1, 2, . . . , n + 1. Since .‖yi − z‖‖xi − z‖ = 1 = ‖yj − z‖‖xj − z‖,
we have

‖yj − z‖ : ‖yi − z‖ = ‖xi − z‖ : ‖xj − z‖,

.

and hence .Δzxi xj is similar to .Δzyj yi . Therefore,

‖xj − xi ‖‖yi − z‖ ‖xi − xj ‖

‖yi − yj ‖ =
. = .
‖xj − z‖ ‖xi − z‖‖xj − z‖

Then, by multiplying the ith row of the Cayley-Menger matrix .𝚪(Y), for .i =
1, 2, . . . , n + 1, with .‖xi − z‖2 and the j th column of the resulting matrix, for
.j = 1, 2, . . . , n + 1, with .‖xj − z‖ , we get the matrix .SD(X ). Therefore
2

⎛n+1 ⎞
∏
. ‖xi − xn+2 ‖ 4
· det 𝚪(Y) = det SD(X ). (12.3)
i=1

Now, .x1 , x2 , . . . , xn+2 lie on a sphere or a hyperplane

. ⇔ y1 , y2 , . . . , yn+1 lie on a hyperplane

⇔ det 𝚪(Y) = 0 (by Theorem 12.1)
⇔ det SD(X ) = 0 (by (12.3)).

⨆
⨅
250 12 The Cayley-Menger Determinant

12.6 Exercises

12.1 Confirm that in Rn , n unit spheres with centers

.(1, 0, 0, . . . , 0), (0, 1, 0, . . . , 0), . . . , (0, . . . , 0, 1)

√
and the sphere with center (−1, −1, . . . , −1) and radius n + 2 form
together an orthogonal-sphere system.
12.2∗ Compute the determinant of the square matrix of degree k whose entries in
the main diagonal are all −1, and whose remaining entries are all 1.
12.3 Prove that there is no five-point set in R3 with all integral distances being
odd. √
12.4 Confirm that in Rn , n + 1 spheres of radius 1/ 2 with centers

(1, 0, . . . , 0), (0, 1, 0, . . . , 0), . . . , (0, . . . , 0, 1), (a, a, . . . , a),

.

√
where a = (1 − 1 + n)/n, and the sphere centered at the barycenter of
these n + 1 points having radius
/
n 1
. −√
n+1 2

form a tangent-sphere system.

12.5 In R4 , five unit spheres and a sphere of unknown radius r form a tangent-
sphere system. Find r.
12.6∗ Let x i , i = 1, 2, 3, 4 be four points in the plane R2 and

d(x 1 , x 2 ) = 1, d(x 1 , x 3 ) = 3, d(x 1 , x 4 ) = 2, d(x 2 , x 3 ) = 3,

.

d(x 3 , x 4 ) = 3.

Find the distance d(x 2 , x 4 ).

12.7 Notes

In these notes, we bring together knowledge only about Cayley-Menger deter-

minants and related topics. To the other subjects we could not locate too many
references, except for Ptolemy’s famous theorem. But this is discussed in the notes
to the next chapter, since Casey’s theorem directly generalizes it.
12.7 Notes 251

12.7.1 Notes on Cayley-Menger Determinants

As helpful tools and for their own sake, Cayley-Menger matrices and Cayley-
Menger determinants are introduced in Sect. 12.1.

Assignment and Some History

Volume computation is an important subfield of pure and applied mathematics.

Just referring only to geometric aspects and applications and strongly related
notions (like, e.g., that of mixed volumes in convexity) one gets already a large
and interesting area. A comprehensive variety of approaches to volumes (especially
from the viewpoint of computational convexity, where notions from convexity are
studied in normed vector spaces of finite, but generally non-restricted dimension)
is collected in the important exposition [329] of Gritzmann and Klee. Basic notions
and concepts studied there are representations of polytopes, triangulations, lattice-
point enumeration, sweeping-plane methods, approximation theory, randomized
approaches and also special types of convex bodies (like simplices or zonotopes).
For simplices, Section 3.6 of [329] is particularly interesting. There the authors
discuss volumes of simplices in terms of the Gram matrix of inner products of their
vertices, in terms of the coefficients appearing in the affine functionals defining the
facets, and (last, but not least) in terms of their (squared) edge-lengths, thus using
Cayley-Menger determinants.
We start now with some historical remarks treating Cayley-Menger determinants.
In the 1930s there was a strong tendency to look for axiomatization possibilities in
several mathematical fields, also influenced by Hilbert’s formalism. This led Menger
[570] to present in 1928 a new axiomatic approach to metric spaces based on the
concept of distance and on the congruence relation. Menger understood that an
algebraic machinery, developed by Cayley, is very useful in this framework, and so
the Cayley-Menger determinants came to be. With the help of this notion, Menger
extended also Heron’s theorem to the computation of the volume of n-simplices.
More precisely, the Cayley-Menger determinant, essential mainly in linear
algebra, geometry, and trigonometry, is a determinant yielding the volume of
an n-dimensional simplex S in terms of all squared distances between vertices
of S (edge-lengths). Besides the proper sources [147] due to Cayley and [570,
second investigation] of Menger, recommendable older references treating this
notion are due to Dörrie [218, pp. 285-289] (discussing the three-dimensional case),
Blumenthal [96, Part III] (presenting it as a cornerstone of distance geometry, also
in spherical, elliptic, and hyperbolic spaces as well as in semi-metric ones; see also
Menger’s booklet [571]), Blumenthal and Gillam [97] (treating the geometrical
interpretation of the signs of the cofactors of the elements of a Cayley-Menger
determinant), and Sommerville [752, Chapter VIII] (for volume calculations).
A good expository article on Cayley-Menger determinants from those days was
written by Bottema [116].
252 12 The Cayley-Menger Determinant

For more recent expositions and overviews we refer to the book parts and surveys
discussed below, and to the web page [174].
So Cayley-Menger determinants became mainly a useful tool for several par-
tial fields of and concepts in mathematics, and in applications. Among these
mathematical fields and concepts we might emphasize: volume computation,
distance geometry, non-Euclidean geometries, embeddibility and rigidity questions,
k-distance sets, (semi-)metric spaces, graph theory etc.; and essential application
fields are e.g. robotics, network theory, physical chemistry and molecular structures.

Overviews in Books, and Surveys

Here we bring book- and survey-like publications treating different aspects of

Cayley-Menger determinants in chronological order.
The essential monograph [96] of Blumenthal, which is one of the most fun-
damental references treating distance geometry, uses Cayley-Menger determinants
and generalizations of them as important tools. This can be clearly seen in the Parts
II, III, and IV of the book. Various embedding problems and derived characteriza-
tions (e.g., of Euclidean or Hilbert spaces) are given in terms of these determinants.
And one can also see that some properties of Cayley-Menger determinants, hardly
derivable in a purely algebraic way, get a better approach when they are presented
in the language of the metric relations studied in [96].
In his nice monograph [266], Fiedler summarizes many (mainly own) theorems
from the geometry of simplices, going back to the early fifties, but containing also
modern results (and various parts are even new). All these results are obtained by the
successful use of the interplay between Euclidean geometry, graph theory, matrices
and determinants corresponding with simplices. Also the Cayley-Menger determi-
nant (introduced in Sect. 1.4) is studied, and Fiedler presents many applications
(e.g., in electrical network theory).
Hampton [378] gives a nice survey on the geometry of triangles, tetrahedra and
higher-dimensional simplices. Old and new formulas and identities are presented,
in particular referring to distances, areas, and angles. The author also treats Cayley-
Menger determinants within that framework.
With more than 200 references, Liberti et al. [509] give a comprehensive survey
on the Euclidean geometry based on the concept of distance. The authors emphasize
information for applications mainly in the following direction: the input data are
given in a non-complete system of distances, and the output is a point set in
Euclidean n-space where the given distances are realized. After an introductory
part, the main mathematical results are given in Sects. 12.2, and 12.3 presents
the applications of distance geometry to molecular conformation problems. In the
final fourth section applications to wireless sensor networks, statics, and robotics
are discussed. We continue with a nice exposition of problems around distance
geometry (and in some cases combinable with Cayley-Menger determinants),
namely the paper [508] of Liberti and Lavor. It contains, nicely presented, a beautiful
12.7 Notes 253

discussion of Heron’s formula, Cauchy’s rigidity theorem on polyhedra, Cayley’s

extension of Heron’s formula to higher dimensions, Menger’s characterization of
abstract semi-metric spaces, Gödel’s result on metric spaces on the sphere, and
Schoenberg’s equivalence of distance and positive semidefinite matrices, giving the
fundament of multidimensional scaling.
Ungar encoded in 1988 the Einstein velocity addition law and got an algebraic
structure that he called gyroalgebra. Based on this, Ungar’s book [788] introduces
for the first time the concept of hyperbolic n-simplex as an important tool and subject
of research in this direction of n-dimensional hyperbolic geometry. The basic
notions of gyroalgebra and related geometric notions are excellently introduced
in Chapter 1 of [788], and in Part IV (Chapter 10) Ungar discusses hyperbolic
hyperplanes, hyperspheres and simplices in this context, getting, e.g., notions like
gyrofaces, gyrocenters, and gyoradii for gyrosimplices. Successfully using Cayley-
Menger determinants, this concept enables him to study an interesting portion of
related hyperbolic simplex geometry, e.g. treating in- and circumcenters, in- and
circumradii, altitudes, and Lemoine points.
The book [20] of Alfakih covers knowledge about Euclidean distance matri-
ces (EDMs) and their applications mainly in rigidity theory. After introducing
mathematical notions from matrix theory, graph theory, convexity, and semidefinite
programming, Chaps. 3 and 4 deal with basics from the theory of EDMs themselves
and various special types, such as Manhattan distance matrices on grids, distance
matrices of hypercubes and trees, and in particular also the Cayley-Menger matrices.
Several interesting topics are discussed in later chapters, such as spherical and
nonspherical EDMs, connections among EDMs, adjacency matrices of graphs and
combinatorial designs, and the completion of incomplete EDMs. In the last two
chapters notions like rigidity, local rigidity, infinitesimal rigidity, and universal
rigidity are introduced and studied. In a balanced and detailed manner, the author
nicely presents classical results and various recent developments. It makes sense
to present directly here, close to the book [20], the nicely written exposition [18]
treating also results on EDMs regarding the old question: how to find out whether
a real, symmetric, zero-diagonal quadratic matrix is an EDM? The authors show
the correctness of a related algorithm which is faster than the classical algorithms
and based on sphere intersections. The complexity of this algorithm is linear in the
number of rows and columns of the input matrix, and quadratically in the minimum
embedding dimension.
Havel and Li [384] give a survey on distance geometry, conformal geometric
algebra and its applications to physical chemistry against the background of
metric spaces on which the group of isometries acts transitively. The authors
start with Cayley-Menger determinants, Schoenberg’s quadratic forms and further
related and needed tools, and they describe then conformal molecular chemistry
as an application field of distance geometry. More general Euclidean invariants are
discussed, too. Summarizing the discussed subjects one can say that this survey
is mainly focused on applications of basic concepts and classical results of metric
geometry in physical chemistry.
254 12 The Cayley-Menger Determinant

In the recent exposition [86] a computational technique to determine the volume

of n-dimensional polytopes is presented. Initially, the volume is computed for an
n-simplex which is obtained by signed simplex decomposition of the polytope. A
recursive algorithm makes it possible to compute then the volume of the whole
polytope. In this procedure, Cayley-Menger determinants are used to get the needed
simplex volumes.

Simplex Geometry

Blumenthal and Sen [98] deal with curvature in metric spaces (where the curvature
of three and the torsion of four points can be introduced via Cayley-Menger
determinants). The latter yields two extensions of the sine law of tetrahedra, the
first referring to dihedral angles, and the second to solid angles. Using their tools,
they extend this also to Euclidean n-simplices.
Alexander [19] defines so-called r-acute n-simplices with the help of equal
squared and non-squared interpoint distances and proves a metric simplex result
implying a lower bound for some Cayley-Menger determinants.
Yang and Zhang [824] extend the Neuberg-Pedoe inequality on pairs of triangles
in the plane, whose case of equality characterizes pairs of directly similar triangles,
to Euclidean n-space and give also some related non-Euclidean results. Namely,
the generalized inequalities for pairs of n-simplices contain, besides the volumes
and edge-lengths of the two participating n-simplices, also cofactors of the Cayley-
Menger determinants for one squared volume. Regarding non-Euclidean analogues
it is shown that some analogues do not hold in the hyperbolic case, but at least in the
2-dimensional elliptic case. Further interesting results of these authors in the same
direction are cited.
Gregorac [324] solves the following problem: Given non-negative numbers
.a1 , ..., an+1 and a regular n-simplex S with vertices .vi . Under what circumstances

is there a point p and a scaling factor s such that .pvi = ai /s? If one solution exists,
inversion in the circumsphere of S yields another one. Using the Cayley-Menger
determinant regarding the distances, the author solves the problem and applies the
result to other problems.
Kokkendorff [467] studies finite metric spaces and creates Gram matrices as
tool for studying the geometry of simplices in space forms. He gives a formula
relating the determinant of a normalized Gram matrix (here defined via cosines
of interpoint distances) to the geometry of the considered simplex. A machinery
is constructed that allows to work with Gram matrices in a very general context,
including also a formula for the determinant being a generalization of the Cayley-
Menger determinant formula and an extension of Heron’s formula. Based on
these results, the author gives conditions for a metric space to be realizable as an
isometric embedding into Riemannian and semi-Riemannian space forms, and also
into hyperbolic spaces.
The starting point of D’Andrea and Sombra [196] is the classical Cayley-Menger
determinant regarding an n-simplex. The main result says that the Cayley-Menger
12.7 Notes 255

determinant is absolutely irreducible over C.[dij ] for any .n ≥ 3, with .dij as set of
distances between all vertices. Moreover, the authors prove that a further polynomial
related to the problem of computing the radius of the circumsphere of a simplex is
also absolutely irreducible in all dimensions, but the smallest one (this has to do
with Ptolemy’s theorem). Hajja et al. [368] deal with the equality, holding for
regular n-simplices (.n ≥ 2), which combines the total edge-length and the .n + 1
vertex distances from a point in its affine hull. As a by-product the authors get new
approaches to the irreducibility of Cayley-Menger determinants.
With the help of the Cayley-Menger determinant √ one can write the n-volume of
an n-simplex with integer edge-lengths of the form .q k, where q is a rational num-
ber and k denotes a square-free positive integer which is called the characteristic
of the simplex. Any triangle spanned by three non-collinear points being from this
family is known to have the same characteristic. By Kurz [484] this is extended to
n dimensions, and he also shows that for any n-dimensional integral point set all
non-degenerate n-simplices have the same characteristic. The author continues with
generating finite integral point sets, e.g. of minimal diameter.
Inspired by the fact that the sum of the angles of all Euclidean triangles is equal
for all triangles, Hajja and Martini raised the corresponding question for higher
dimensions in [363], namely whether something similar holds for n-simplices with
.n ≥ 3. Some related property (via corner angles) holds for the orthocentric case,

but for various further important classes of simplices (e.g., circumscriptible ones)
some property like this is not holding, as shown by Hajja and Hayajneh [361].
Here characterizations of special Cayley-Menger determinants corresponding with
special simplex classes form an important tool, like also embeddability properties.
Gaifullin [287] asks whether there are relations between k-volumes of k-faces
of an n-simplex S similar to those between edge-lengths and the volume of S (the
latter ones leading to Cayley-Menger determinants). He has special interest in the
case .k = 2 and shows that, for .n ≥ 4, the square volume of S satisfies a polynomial
relation with coefficients dependent on the squares of the areas of its 2-faces. The
respective minimal degree is computed, and the existence of a monic polynomial
relation between the volume of S with coefficients depending on the areas of the
2-faces of S is confirmed if and only if n is even and at least 6. The leading
coefficients of polynomial relations satisfied by the volume for other dimensions
are also studied.
Schwartz [717] proves the following two non-trivial results: for any tetrahedron
T there exists a tetrahedron .T ' such that every edge of .T ' is one unit longer than the
corresponding edge of T . Furthermore, the ratio of the volume of .T ' to that of T has
the sharp lower bound .(1 + 6/s)3 , where s denotes the sum of edge-lengths of T .
An important tool in proving these results is given by Cayley-Menger determinants.
And so also results for higher dimensions are obtained.
In [367], Hajja et al. prove many nice results on pre-kites, i.e., on n-simplices
having one regular (.n − 1)-face. Their proofs use Cayley-Menger determinants
specified for pre-kites. E.g., they prove the following result: for .n ≥ 3 an arbitrary
non-regular n-dimensional pre-kite cannot have more than two regular facets. The
metric structure of pre-kites with two regular facets is completely clarified. An n-
256 12 The Cayley-Menger Determinant

dimensional pre-kite P is a regular n-simplex if one of the following items holds

true: (i) the circumcenter and the centroid of P coincide; (ii) the circumcenter and
the incenter of P coincide; (iii) the incenter and the centroid of P coincide and
.n ≤ 5. We add that, if .n ≥ 6, then an arbitrary pre-kite with coinciding incenter and

centroid is not necessarily a regular simplex.

A circumscriptible tetrahedron has a sphere touching all its edges. Clearly,
a general tetrahedron is not uniquely determined by its face areas (even not, if
additionally the volume and circumradius are known). Using a characterization of
the tetrahedron by the Cayley-Menger determinant, Hajja and Krasopoulos [362]
prove that a circumscriptible tetrahedron is determined by its face areas.
Aomoto and Machida (see [34]) present a new Schläfli-type formula for the
volume of pseudo-simplices which are spherically faced in Euclidean space. This
formula is described in terms of Cayley-Menger determinants and their differentials
involved with hypersphere arrangements in [35]. This is based on hypergeometric
integrals and goes back to a variational formula also coming from hypergeometric
integrals, but associated with hyperplane arrangements. Also using Cayley-Menger
determinants, this was first studied by them (see [34]).
Taking Heron’s formula (giving the area of a triangle in terms of its side-
lengths) as starting point, Kock [460] presents the corresponding Cayley-Menger
determinant with the possibility to extend this way everything to higher dimensions.
He applies the formula to infinitesimal simplices in Riemannian manifolds, yielding
a geometric description of the volume form of the respective manifold.
It is a natural question whether there can be several non-congruent tetrahedra
with the same volume, circumradius and facet areas. Dehbi et al. proved (in an
older paper) that for these given metric invariants at most eight real tetrahedra
types exist (except that three facet areas are equal). Now, in [205], they reprove
this more elegantly, and they show (using Cayley-Menger determinants) that
the determination of tetrahedra from given metric invariants can be reduced to
solving two polynomial equations having two unknowns. Further simplifications are
discussed, and finally numerical experiments show that the number of real solutions
might be not larger than 6.
At the end of this subsection here we mention once more Fiedler’s monograph
[266] which contains results from various research papers of the author dedicated to
simplex geometry.

Extensions and Modifications

Some extensions and modifications of Cayley-Menger determinants and their

applications were already mentioned above, and we will not repeat them here.
Seidel [724] presents a distance-geometric theory of planar Euclidean geometry
based on properties of Cayley-Menger determinants, and he continues in [725]
giving the fundaments for metrical geometry in spherical and hyperbolic planes
using Cayley-Menger determinants based on cosines or hyperbolic cosines of the
mutual distances in finite sets.
12.7 Notes 257

Havel [383] gives an algebraic approach to prove theorems on Euclidean point

configurations by using the interpoint distances instead of the coordinates of the
given points. Thus Cayley-Menger determinants and Cayley-Menger bideterminants
(see below) are taken to present distances and further measures in an algebraic way.
The author states four geometric theorems (e.g., Simson’s theorem is one of them),
but gives on this way algebraic proofs, and further applications (e.g., to get the
topological structure of a simple linkage mechanism) are also presented.
Basically using Cayley-Menger bideterminants whose entries are 0, 1 and
squared distances, Dalbec [195] derives the first and the second fundamental
theorem for vector invariants of the Euclidean group over the field of characteristic
./= 2. Also a presented related algorithm is based on these determinants.

Rylov [681] presents a new method for investigating a metric space X which is
based on a classification of finite subspaces of X. Due to this one gets informations
on X and can remove constraints imposed usually on the metric. So one can use
X for the description of the space-time geometry and also further geometries, with
indefinite metrics. Interestingly, as a measure for a finite subspace in this framework
(called length or volume), the author introduces a determinant of metric functions
which is similar to the Cayley-Menger determinant.
Li and Leng present in [506] a discrete version of mixed volumes, e.g. of
two finite vector sets, deriving also the corresponding analogue to the Minkowski
inequality for mixed volumes. They continue by defining the concepts of mixed
Cayley-Menger determinants and mixed vertex angles of simplices, thus extending
the concept of classical Cayley-Menger determinants. The relations between all
these notions are studied, and based on this, interesting inequalities and equalities
for simplices containing also the mixed Cayley-Merger determinant and the mixed
vertex angles are derived.
Soland [748] elegantly shows that one can characterize the Euclidean plane by
using the common axioms for metric spaces together with three further simple
axioms involving solely the Cayley-Menger determinants.
Dordovskii [221] shows that the metric spaces which are pretangent to a
Euclidean or unitary space (finite-dimensional) are isometric to this space. Based
on this property, he describes the metric pretangent spaces at the non-singular
points of smooth surfaces. The proof of the main result is strongly based on the
fact that the Cayley-Menger determinant of .n + 2 points in n-space is zero. The
author proves also that there exist spaces pretangent to Hilbert spaces without this
isometry property. Also Bilet and Dovgoshey [88] deal with pretangent spaces,
more precisely with conditions for the isometric embeddability of pretangent metric
spaces in the real Euclidean space. This means that they extend results on isometric
embeddability due to Menger, Schoenberg and Blumenthal to this more general
case.
Thomas and Porta [775] use a known identity of Clifford to extend Euclidean
distance problems regarding point sets to families of hyperspheres (taking the
so-called power between hyperspheres as distance). Choosing the radii suitably,
the Clifford identity comes conveniently down to generalized Cayley-Menger
determinants, simplifying the approaches to the extended distance problems.
258 12 The Cayley-Menger Determinant

Applications

Schoenberg [709] uses Cayley-Menger determinants to describe the mobility of

rigid segment structures (like, e.g., the edge system of the regular octahedron) which
lose their rigidity and become movable linkages with increasing dimension of the
spaces in which they are embedded. His motivation is to study general classes of
such linkages.
Szenthe deals in [767] with the concepts of curvature in the metric theory of
curves and the standard differential geometry of curves in Euclidean spaces. Using
Cayley-Menger determinants, the relations between these concepts are studied.
Graham et al. [317] give an exact criterion that .n + 2 points in Euclidean n-space
have only odd integers as their pairwise distances, and they prove this (and further
similar results) with the help of a factorization of the respective Cayley-Menger
determinant.
Sippl and Scheraga [735] expose results on the problem of calculating Cartesian
coordinates from a matrix of interpoint distances (thus yielding the so-called
embedding problem). They present an efficient algorithm for the transformation of
distances to coordinates and show this way that the embedding problem is related
to the theory of symmetric matrices. Using methods from distance geometry, they
continue in [736] with attacking the embedding problem regarding nuclear magnetic
resonance for macromolecules (determining a proper subset of the complete system
of distances between given points, which then has to be filled with a minimum set
of errors).
Crippen [189] proposes to apply methods from Blumenthal’s distance geometry
(see [96]) for the calculation of conformation of molecules, i.e., to use direct
calculation of the interpoint distances and results on Cayley-Menger determinants
and embeddability from distance geometry. His concept is to use all known distances
at their right places, and to approximate or modify the unknown distances so that
they satisfy geometric constraints. Following the proposals from [189], Klapper
and DeBrota [454] provide an algorithm to compute the occurring distances in a
molecular structure directly, and they present an interesting example.
To make some useful results from distance geometry accessible to applied
scientists, Gower [312] nicely discusses properties of the interpoint distance matrix
of finite point sets in Euclidean n-space. And he studies under what circumstances
a symmetric matrix can be such a distance matrix.
By gradual contraction of the dimensionality, Purisima and Scheraga [650]
succeed to get a low-energy three-dimensional structure from a very-low-energy
high-dimensional conformation. For this dimension contraction they use Cayley-
Menger determinants, and interesting examples (e.g., full-atom representations of
certain terminally blocked amino acids) are given.
Using a Cayley-Menger determinant, Volenec [797] proves that if all distances
occurring between five points in the Euclidean plane take only two different values,
then they form the vertex set of a regular pentagon. (Note that here a “correct"
12.7 Notes 259

labelling of the distances is not assumed; otherwise the result is well known and
true even without assuming 2-dimensionality.)
Serré et al. [728] deal with conceptual design, where designers have to verify
the consistency of engineering specifications; more precisely, here the analysis of
the consistency of a 3D geometric specification is described. Since the usual way
via Cartesian modelling does not give easy analysis of consistency, the authors
present two non-Cartesian modellings: the first by using points and Cayley-Menger
determinants regarding their mutual distances, and the second based on Gram
determinants regarding points and vectors. Finally, the advantages due to these two
ways of modelling are shown. Namely, distance geometry is often used in molecular
geometry, and vectorial modelling is used in fields like kinematics, dynamics and
tolerancing.
For Cayley-Menger determinants used regarding k-distance sets (i.e., point sets
having only k different values of interpoint distances) we refer to Section 5.4 of the
problem book [125] of Brass, Moser and Pach.
In geometric constraint solving Cayley-Menger determinants are preferably
used in robotics and molecular chemistry. Distances are considered as coordinates
yielding systems where the unknowns are interpoint distances. This can reduce the
expense of them to solve, but these systems can also be hard to solve. Mathis and
Schreck [559] describe two algorithms making it possible to get such systems with
a minimal number of equations.
Giving proofs of two recent conjectures concerning the m-body matrix, the paper
[327] mainly refers to the problem of m particles under a potential field that is only
a function of the distances among the pairs of the particles. Whenever the positions
of the considered masses are in a nonsingular configuration (i.e., they do not lie in
an affine subspace of dimension at most .m − 2), this matrix defines a Riemannian
metric on the space coordinatized by their interpoint distances. The corresponding
determinant can be factored into the product of the participating Cayley-Menger
determinant (of order m) and a factor which is mass dependent and of one sign on
all nonsingular mass configurations.
Based on information from protein geometry and distance data, obtained via
nuclear magnetic resonance technique, the authors of the paper [595] treat a branch-
and-prune algorithm to search graph realizations in molecular distance geometry
and, with the help of its dual, to look for completions of corresponding non-
complete distance matrices. The authors use a global optimization method which
makes it possible that also the dual branch-and-prune algorithm can efficiently
attack problems from nuclear magnetic resonance experiments.
Inversive distance circle packings were introduced as a generalization of
Thurston’s circle packings on surfaces, allowing adjacent circles to separate. They
live on triangulated surfaces with given weights on all undirected edges and circle
centers at the vertices, clearly satisfying certain conditions regarding the inversive
distance of neighboring circles. A central question refers to their rigidity, and for
the proof of that a variational principle has to be applied. With new proof steps, Xu
[819] follows that way, and in the first step he uses Cayley-Menger determinants
allowing also to extend at least this proof part to higher dimensions.
Chapter 13
Casey’s Theorem

The more you know, the more you know you don’t know.
— Aristotle

Casey’s theorem is an extension of Ptolemy’s theorem: four points in Ptolemy’s

theorem are replaced by four circles, and distances between points are replaced
by common tangent distances between circles. The proof of Casey’s theorem is
elaborated and requires a few new techniques. Casey’s theorem can be also extended
to a set of circles consisting of more than four circles.

13.1 Bicolored Sets of Circles

To state Casey’s theorem, let us introduce bicolored sets of circles. A set .F of circles
is called a bicolored set of circles if each circle in .F has one of two colores (red and
blue, say) and the following properties (1) and (2) hold.
(1) No circle in .F is contained in the closed disk bounded by another circle in .F.
(2) Circles in .F of different colors are disjoint to each other.
In a bicolored set of circles, one color class may be empty. We also allow that some
colored circles have radius 0 (i.e., are colored points, called null circles).
Let .F = {c1 , c2 , . . . , cn } be a bicolored set of circles in .R2 . We define the
distance .t (ci , cj ) (called the common tangent distance) between two circles .ci , cj
in .F by the distance between the two contact points of a common tangent, where we
take an external tangent if they have the same color. Otherwise, we take an internal
tangent. Let us state this more precisely. For a circle .ci in .F, let .zi be its center, and

© The Author(s), under exclusive license to Springer Nature Switzerland AG 2024 261
H. Maehara, H. Martini, Circles, Spheres and Spherical Geometry,
Birkhäuser Advanced Texts Basler Lehrbücher,
https://doi.org/10.1007/978-3-031-62776-7_13
262 13 Casey’s Theorem

...................................... ci
.. ..
.
....
........ c j ..........
.

... ...
...
...
z j●....... j .....
r zi
........ ..
. .. .
..
.. .
. .
... ........ ....
.......... ...................... .
........... ............ ..........
..
........
........ ●
...... .... ......... .. ...
.. .
. .
...... . .
...... .....
. .
. .
.
......
.
....
.
.. . .
. .. . .
.. ...........................................
. ...
..
ri
..
...
. ... ... ...
... ... . ...
... ...
.. . t(ci , cj )
.. ... ..
... .
.. z i ..
.......
● ........ .. .. . ..
..
.
..
c
..
... i
..........
. ..
t(ci , cj )
..
....
...
... r
........
........ ..
......... rj
... i .●
..
.... ...
....
....
.....
........ ...
. ..... zj
..........
................................................
....... cj
same color case different colors case
Fig. 13.1 Common tangent distances

ri denote its radius. For a pair of circles .ci , cj in .F, their common tangent distance
.

t (ci , cj ) is defined by
.

⎧/
⎨ |zi zj |2 − (ri − rj )2 if ci , cj have the same color,
.t (ci , cj ) = /
⎩ |zi zj |2 − (ri + rj )2 if ci , cj have different colors,

see Fig. 13.1. If .ci , cj are null circles, then .t (ci , cj ) = ‖zi −zj ‖. Due to the common
tangent distances, a bicolored set of circles becomes a semi-metric space.
A circle or a line is called properly tangent to the circles in a bicolored set of
circles, if circles of the same color are tangent to the circle (or line) from the same
side, and the circles of different colors are tangent to the circle (or the line) from the
opposite side.
Now, Casey’s theorem can be stated as follows.
Theorem 13.1 (Casey’s Theorem) For a bicolored set .F = {c1 , c2 , c3 , c4 } of four
circles, the following two statements are equivalent.
(i) There is a circle or a line that is properly tangent to the circles in .F.
(ii) .det SD(F) = 0.
Remark 13.1 By Lemma 12.2, the statement (ii) implies that one of

.t12 t34 − t13 t24 + t23 t14 ,

t12 t34 + t13 t24 − t23 t14 ,
−t12 t34 + t13 t24 + t23 t14

is equal to 0, where .tij = t (ci , cj ) are the common tangent distances between .ci and
cj . Thus, if all circles in .F are null circles, then Casey’s theorem becomes Ptolemy’s
.

classical theorem.
To prove Casey’s theorem, we prepare several lemmas.
13.2 Inversion Invariants 263

13.2 Inversion Invariants

Lemma 13.1 Let c1 , c2 be two circles in a bicolored set of circles, and let x ∈ c1
be a point that lies outside of c2 . Then a circle (or a line) that passes through x and
orthogonally crosses both c1 , c2 is uniquely determined.
Proof Let us consider the case that x does not lie on the line through the centers of
c1 , c2 . Then the center of a circle that orthogonally crosses both c1 , c2 and passes
through x is the intersection point of the tangent line of c1 at x and the radical axis
of c1 , c2 , see Lemma 3.5. Hence such a circle is uniquely determined. ⨆
⨅
Lemma 13.2 Let ϕ be the inversion of the plane with respect to a circle with center
p and radius r. Let us denote ϕ(x) by x ' . For any two points x, y (both different
from p) we have

r 2 |xy|
|x ' y ' | =
. .
|px| · |py|

Proof Since |px| · |px ' | = r 2 = |py| · |py ' |, the triangles Δpxy and Δpy ' x ' are
similar. Hence

|x ' y ' | |px ' | r2

. = = .
|yx| |py| |px| · |py|

From this the lemma follows. ⨆

⨅
For four points w, x, y, z in the plane, let us define [w, x, y, z] by

|wx| · |yz|
[w, x, y, z] =
. ,
|wy| · |xz|

which is called a Möbius invariant of {w, x, y, z}. The next corollary is obtained
by straightforward calculation using the above lemma.
Corollary 13.1 The Möbius invariant [w, x, y, z] is invariant under any inversion
ϕ whose center p is different from w, x, y, z, namely,

.[ϕ(w), ϕ(x), ϕ(y), ϕ(z)] = [w, x, y, z].

⨆
⨅
Lemma 13.3 Let c1 , c2 be two circles in a bicolored set of circles, and let zi , ri be
the center and the radius of ci for i = 1, 2. Let γ be a circle that orthogonally cuts
both c1 , c2 . Let x1 ∈ γ ∩ c1 be a point that lies outside c2 . When we trace γ from
x1 in the direction towards the interior of c1 , let x2 be the point where we meet c1
264 13 Casey’s Theorem

Fig. 13.2 c1 ⊥ γ ' ⊥ c2 ... .....

... .... .●........ .... ....
.. .... . ....... ..... .... .... ....
.
...
.. .. .
.. .
... . .. . ...
...
...
.....
...
...
γ
... ...... ... ...
.. . ...
.. ..... ..
... ..... c 3
...
. .. ..
x
... . ...1
.●
..
..
.. ... ............................................................... .....
.
.
.
. . .
... .
.
................. .
. ............ ... ...
.
.
..
.
.
.. .. . . ............. ..
..
...
.... .
.... ............. ...
.
..... ............... ... .........
.......
..
.
.. .
....... ..
.
. .
y
.......................
.. .. ..
... ..... ....
.... ........... ● . .. ...
.. ... ...... . .........
... ..... ... .. ... .... .. ...................
p
...............●
.
...............
.
...
.
. p ..
.
. q
1............................................................................................2...............................1
.
.....●
.....
.. ..... ........
...... 2
..............................●
...................................●
q
......................
.. ● .. .... ● .
...... ....
... ..
...
z 1 .... .... ....
....
..... .... ...... .... .. ... .
.......... ..
.
.
....
z 2 ....... .......
. ...

... . ...... .... ..

.... .. ................................. .
..... ....
c .
1 .......................................................
....
c 2

again, y1 be the point where we first meet c2 , and y2 be the point where we meet c2
the second time. Then we have

|z1 z2 |2 − (r1 − r2 )2
[x1 , y1 , x2 , y2 ] =
. . (13.1)
(2r1 )(2r2 )

If |z1 z2 | > r1 + r2 , then we also have

|z1 z2 |2 − (r1 + r2 )2
[x1 , y2 , x2 , y1 ] =
. . (13.2)
(2r1 )(2r2 )

Proof Let us consider the case that |z1 z2 | > r1 + r2 . Let p1 , p2 and q1 , q2 be the
points where the line z1 z2 cuts c1 and c2 so that p2 , q1 lie between p1 , q2 . Let c3 be
the circle determined by the three points p1 , q2 , x1 , and let y be the intersection of
c2 and c3 different to q2 , see Fig. 13.2. Let v be the intersection of the lines p1 x1 and
q2 y. Let γ ' be the circle determined by v, x1 , y. Then |vx1 | · |vp1 | = |vy| · |vq2 |. If
we denote by ψ the inversion with center v and power |vx1 | · |vp1 |, then ψ(ci ) = ci ,
for i = 1, 2, 3, where ψ(γ ' ) is the line z1 z2 . Since the inversion preserves the
size of angles, we can deduce that γ ' orthogonally cuts c1 , c2 . Hence γ = γ ' by
Lemma 13.1, and therefore v ∈ γ and y = y2 . Thus,

ψ(p1 ) = x1 , ψ(p2 ) = x2 , ψ(q1 ) = y1 , ψ(q2 ) = y2 .

.

Since the Möbius invariant is invariant under inversions, we have

[p1 , q1 , p2 , q2 ] = [x1 , y1 , x2 , y2 ] and [p1 , q2 , p2 , q1 ] = [x1 , y2 , x2 , y1 ].

.
13.2 Inversion Invariants 265

Now, since |p1 q1 | = |z1 z2 | + r1 − r2 , |p2 q2 | = |z1 z2 | − r1 − r2 , |p1 p2 | = 2r1 , and

|q1 q2 | = 2r2 , we have

[x1 , y1 , x2 , y2 ] = [p1 , q1 , p2 , q2 ]
.

(|z1 z2 | + r1 − r2 )(|z1 z2 | − r1 − r2 )
=
(2r1 )(2r2 )
|z1 z2 |2 − (r1 − r2 )2
= .
(2r1 )(2r2 )

Similarly, (13.2) can be obtained. ⨆

⨅
Let F be a bicolored set of circles, and let p be a point exterior to all circles in F.
Then by the inversion with respect to the unit circle centered at p, circles in F are
transformed to circles. Since p is exterior to all circles in F, no transformed circle is
enclosed in another transformed circle. Hence, by assuming that every transformed
circle bears the same color as its predecessor (and we always assume this), we get
another bicolored set of circles.
For a circle c, let us denote its center by z(c), and its radius by r(c).
Theorem 13.2 For a pair of circles c1 , c2 in a bicolored set of circles,

t (c1 , c2 )2
.
r(c1 )r(c2 )

is invariant under any inversion with center lying outside of both c1 , c2 .

Proof Let us consider the case where c1 , c2 have the same color. Let ϕ be an
inversion with center p lying outside of both c1 , c2 . Let σi = ϕ(ci ), i = 1, 2, and
γ be a circle such that c1 ⊥ γ ⊥ c2 and p /∈ γ . Let x1 ∈ c1 ∩ γ be a point lying
outside c1 . Define x2 , y1 , y2 as in Lemma 13.3. Then,

t (c1 , c2 )2
4[x1 , y1 , x2 , y2 ] =
.
r(c1 )r(c2 )

by Lemma 13.3. Let η = ϕ(γ ). Then σ1 ⊥ η ⊥ σ2 . Put xi' = ϕ(xi ), yi' = ϕ(yi ), i =
1, 2. Then x1' ∈ σ1 ∩ η, which lies outside σ2 , and

t (σ1 , σ2 )2
4[x1' , y1' , x2' , y2' ] =
.
r(σ1 )r(σ2 )

holds. Since the Möbius invariant is invariant under inversions, we have the equality
[x1 , y1 , x2 , y2 ] = [x1' , y1' , x2' , y2' ]. Hence the theorem follows. ⨆
⨅
266 13 Casey’s Theorem

13.3 Isometric Radii Change

Recall that for a pair of circles .ci , cj in a bicolored set .F of circles,

⎧
|zi zj |2 − (ri − rj )2 if ci , cj have the same color
t (ci , cj )2 =
.
|zi zj |2 − (ri + rj )2 otherwise,

where .zi , ri denote the center and the radius of .ci . Hence, if we simultaneously
increase the radii of the circles in .F of the same color by the same amount, say .ε,
and decrease the radii of circles of the other color by .ε, then the values of .t (ci , cj )
are kept unchanged. More precisely, if .εr denote the radius of the smallest red circle
in .F, then for any .0 < ε ≤ εr , it is possible to decrease the radii of all red circles
by .ε, and to increase the radii of all blue circles by .ε without changing the common
tangent distances. Similarly, if .εb denotes the radius of the smallest blue circle in
.F, then for any .0 < ε ≤ εb , it is possible to decrease the radii of all blue circles

by .ε and to increase the radii of all red circles by .ε without changing the common
tangent distances. Thus, by assuming that every modified circle bears the same color
as its predecessor (and we always assume this), the bicolored set of modified circles
is isometric to the original bicolored set of circles as semi-metric space. Let us call
such a simultaneous radii change an isometric radii change. If there is a circle that
is properly tangent to the circles in .F, then by increasing or decreasing the radius
of the circle by .ε, we get a circle that is properly tangent to the modified circles.
If there is a line that is properly tangent to the circles in .F, then, by translating the
line in a direction through distance .ε, we get a line that is properly tangent to the
modified circles. Hence we have the following lemma.
Lemma 13.4 The circles in a bicolored set have a circle or a line that is properly
tangent to them if and only if the circles modified by an isometric radii change have
a circle or a line that is properly tangent to them. ⨆
⨅
Example 13.1 For any bicolored set of three circles .c1 , c2 , c3 , there is a circle (or
a line) that is properly tangent to the three circles. This can be seen as follows. By
Lemma 13.4, we may assume that one of them, say .c3 , is a null circle .{p}. The point
p is exterior to both .c1 , c2 . Let .ϕ be the inversion with respect to the unit circle with
center p. Then .{ϕ(c1 ), ϕ(c2 )} is a bicolored set of circles. Obviously, there is a line
l properly tangent to .ϕ(c1 ), ϕ(c2 ). Then .ϕ(l) is a circle (or a line) that is properly
tangent to .c1 , c2 , c3 .
Lemma 13.5 Let .F = {c1 , c2 , c3 , c4 } be a bicolored set of four circles, where .c4 =
{p}, a null circle. Let .ϕ be the inversion of .R2 with respect to the unit circle with
center p, and let .σi = ϕ(ci ) for .i = 1, 2, 3. Then

t (ci , cj )2
t (σi , σj )2 =
. for 1 ≤ i < j ≤ 3. (13.3)
t (ci , c4 )2 t (cj , c4 )2
13.3 Isometric Radii Change 267

Fig. 13.3
.Δpqi zi∼ Δpϕ(qi )ζi

σi ζi

ci zi
p
qi ϕ(qi )

Proof Let .zi , ri be the center and the radius of .ci , and let .ζi , ρi be the center and
the radius of .σi for .i = 1, 2, 3. First suppose that .ri > 0, ρi > 0. Let .qi be the point
on .ci such that ./ pqi zi = π/2. Then ./ pϕ(qi )ζi = π/2, and .Δpqi zi ∼ Δpϕ(qi )ζi ,
see Fig. 13.3. Hence .|pϕ(qi )|/ρi = |pqi |/ri , and therefore .|pϕ(qi )| = |pqi |ρi /ri .
Since .|pqi ||pϕ(qi )| = 1, we have .|pqi |2 ρi /ri = 1, and hence .t (ci , c4 )2 = |pqi |2 =
ri /ρi . Similarly, .t (cj , c4 )2 = rj /ρj . Now, by Theorem 13.2, we have

t (σi , σj )2 t (ci , cj )2
. = .
ρi ρj ri rj

Hence

t (ci , cj )2 t (ci , cj )2
t (σi , σj )2 =
. = .
(ri /ρi )(rj /ρj ) t (ci , c4 )2 t (cj , c4 )2

Now, for fixed .p, zi , zj ,

t (ci , cj ), t (σi , σj ), t (ci , c4 ), t (σi , σ4 ), ρi , ρj

.

are continuous functions of .ri , rj . For .k ∈ {i, j },

rk
. lim ρk = 0, lim = lim |pqk |2 = |pzk |2 ,
rk →0 ρk →0 ρk rk →0

and when .rk = 0, we have .|pzk |2 = t (c4 , ck )2 . Hence (13.3) holds even in the case
where one of .ci , cj or both are null circles. ⨆
⨅

Lemma 13.6 Let .F = {c1 , c2 , c3 , c4 } be a bicolored set of four circles, where .c4 =
{p}, a null circle. Let .ϕ be the inversion of .R2 with respect to the unit circle with
center p, let .σi = ϕ(ci ), i = 1, 2, 3, and put .E = {σ1 , σ2 , σ3 }. Then

∏
3
. t (ci , c4 )4 det 𝚪(E) = det SD(F).
i=1
268 13 Casey’s Theorem

Proof For .1 ≤ i, j ≤ 3, the .(i, j )-entry of .SD(E) is given by (13.3). Multiplying

the ith row of .𝚪(E), for .i = 1, 2, 3, with .t (ci , c4 )2 and then the j th column, for
.j = 1, 2, 3, with .t (cj , c4 ) , we have .SD(F). Hence the theorem follows. ⨆
⨅
2

Lemma 13.7 For a bicolored set of three circles .E = {σ1 , σ2 , σ3 }, the following
two statements are equivalent.
(i) There is a line that is properly tangent to all circles in .E.
(ii) As a semi-metric space, .E is isometrically embeddable in .R1 .
Proof
(i) .⇒ (ii). Suppose that there is a line L that is properly tangent to all circles in .E.
Then the set of contact points .L ∩ σi , i = 1, 2, 3, is a subset of L that is isometric
to .E. Hence .E is isometrically embeddable in .R1 .
(ii) .⇒ (i). Noting that the common tangent distance is preserved by isometric
radii change, and by Lemma 13.4, we may suppose that one circle in .E is a
null circle. Let .σ3 = {q}, a null circle. Obviously, there is a line L that is
properly tangent to .σ1 , σ2 . Let .qi be the contact points .L ∩ σi , i = 1, 2. Since .E
is isometrically embeddable in a line, there is a point u on L such that .{q1 , q2 , u}
is isometric to .E. If .p = u, then we are done. Suppose .p /= u. Let .zi , ri be the
center and the radius of .σi , for .i = 1, 2. Then .‖zi − p‖2 = ri2 + t (σi , σ3 )2 , for
/
.i = 1, 2. Hence, p lies on the circles with center .zi and radius . r + t (σi , σ3 )2 ,
2
i
.i = 1, 2, see Fig. 13.4.

The point u also lies on these circles. Hence .Δz1 z2 u and .Δz1 z2 p are congruent.
Let .ψ be the reflection of .R2 with respect to the line .z1 z2 . Since .p /= u, we must
have .ψ(u) = p. Then, since .p ∈ ψ(L), .ψ(L) is a line that is properly tangent to
.σ1 , σ2 and passes through p.

⨆
⨅

Fig. 13.4 .σ3 = {p}

σi
zi
ri
●
t(σi , σ.3. ) p
13.5 A Theorem on Five Circles 269

13.4 Proof of Casey’s Theorem

By Lemma 13.4, we may consider the case that one of the circles in .F is a null
circle. Let .c4 = {p}, a null circle. Note that p is exterior to all other circles in
.F. Let .ϕ be the inversion of .R with respect to the unit circle with center p, let
2

.σi = ϕ(ci ), i = 1, 2, 3, and .E = {σ1 , σ2 , σ3 }. By Lemma 13.6,

. det 𝚪(E) = 0 ⇔ det SD(F) = 0. (13.4)

(i) .⇒ (ii): Suppose that there is a circle .γ that is properly tangent to all circles
in .F. Then .ϕ(γ ) is a line that is properly tangent to the circles in .E. Then, by
Lemma 13.7, .E is isometric to a 3-point-subset of .R1 . Hence .det 𝚪(E) = 0 by
Theorem 12.1. Therefore, .det SD(F) = 0 by (13.4).
(ii) .⇒ (i): Suppose that .det SD(F) = 0. Then .det 𝚪(E) = 0, by (12.7). Hence .E
is isometrically embeddable in .R1 , and hence there is a line L that is properly
tangent to the circles in .E, by Lemma 13.7. Then .ϕ(L) is a circle that is properly
tangent to the circles in .F. ⨆
⨅

Remark 13.2 Casey’s theorem can be also generalized to bicolored sets of spheres
in .Rn , see [535].

13.5 A Theorem on Five Circles

With similar methods, we can derive a related theorem referring to a quintuple of

circles.
Theorem 13.3 Let .F = {c1 , c2 , c3 , c4 , c5 } be a bicolored set of five circles in .R2 .
Suppose that, for each .i = 1, 2, 3, 4, 5, there is a circle (or a line) .γi that is properly
tangent to the circles in .F \ {ci }. Then some two of .γi , i = 1, 2, . . . , 5, coincide,
and hence there is a circle or a line that is properly tangent to all five circles in .F.
Proof We may suppose that at least three circles in .F are red circles. Let us diminish
the radius of the smallest blue circle, to zero if .F has a blue circle. Otherwise,
diminish the radius of the smallest red circle to zero by isometric radii change. So,
we may suppose, from the first, .c5 ∈ F is a null circle .{p}. Let .ϕ be the inversion
of .R2 with respect to the unit circle with center p and let .E = {σ1 , σ2 , σ3 , σ4 } be
the bicolored set, where .σi = ϕ(ci ), i = 1, 2, 3, 4. Let .li = ϕ(γi ) for .i = 1, 2, 3, 4,
and .τ = ϕ(γ5 ). Then .li is a line that is properly tangent to the circles in .E \ {σi } for
.i = 1, 2, 3, 4, and .τ is a circle or a line that is properly tangent to the four circles

in .E. Since at least three circles in .E are red circles, we may suppose that .σ1 , σ2 , σ3
are red circles, and .σ4 is either a blue circle, or the smallest red circle. By applying
isometric radii change, we may furthermore suppose that .σ4 is a null circle .{q}. (Of
course, the lines .li and .τ (if .τ is a line) are then translated accordingly, and the
270 13 Casey’s Theorem

Fig. 13.5 The three angular l3.. l . 2

regions are disjoint ...
... ...
... ........................ ...
... ...... ... .....
... ...
....
... σ
....
1 ..........
.. ..

....
............................
... ...
... ...
. .
l1 ↓ ...............................................................................................................................................................................
....... ......
.... ... ... .... ...
... ↓ l1
......
..
... ..... ........
.. .. σ
.... 2 .....
... ..
...
..
...
σ 3
.. ..
.....
.....
......................
...
... . ...
.... .... ...
........ ......
.............................. ...
...
.... ...
...
l2 l3

radius of .τ (if .τ is a circle) is changed accordingly.) Note that three lines .l1 , l2 , l3
all pass through q, and .l4 is properly tangent to .σ1 , σ2 , σ3 .
To prove the theorem, it is sufficient to show that .li = lj for some .1 ≤ i /= j ≤ 4.
If one of .σ1 , σ2 , σ3 is a null circle, say .σ3 = {q ' }, then since both .l1 , l2 pass
through the two points .q, q ' , we have .l1 = l2 . So, we may consider the case that
none of .σ1 , σ2 , σ3 is a null circle.
Suppose that .l1 , l2 , l3 , l4 are all different lines. Since .σ1 , σ2 , σ3 are all red circles
and .σ4 = {q}, the circles in .E \ {σi } are tangent to the line .li from the same side of
.li , .i ≤ 3. For .i = 1, 2, 3, 4, let us denote by .[li ] the side of .li containing the circles

in .E \ {σi }. Then

σ1 ⊂ [l2 ] ∩ [l3 ], σ2 ⊂ [l1 ] ∩ [l3 ], σ3 ⊂ [l1 ] ∩ [l2 ].

.

Since three lines .l1 , l2 , l3 all pass through the point q, either one of the three angular
regions .[l1 ] ∩ [l2 ], [l1 ] ∩ [l3 ], [l2 ] ∩ [l3 ] is contained in another angular region, or
the interiors of the three angular regions are disjoint. If the interiors of three angular
regions are disjoint, then it is impossible to cut simultaneously all three angular
regions by a single line, and hence it is impossible that .σ1 , σ2 , σ3 are tangent to .l4
from the same side (see Fig. 13.5), which is a contradiction. Hence we may suppose
that one of the three angular regions is contained in another angular region, say

.[l1 ] ∩ [l2 ] ⊂ [l1 ] ∩ [l3 ].

In this case, .[l1 ] ∩ [l2 ] is also contained in .[l2 ] ∩ [l3 ], see Fig. 13.6. Let x be the
contact point of .σ1 and .l3 , and y be the contact point of .σ1 and .l2 .
Since .τ passes through q and .τ is tangent to .σ1 , σ2 , σ3 from the same side, .τ
must be a circle externally tangent to them. Then the contact point of .σ1 and .τ must
(

lie on the minor arc xy of .σ1 . Since .σ3 (⊂ [l1 ] ∩ [l2 ]) is also externally tangent to .τ ,
the contact point of .σ3 and .l2 must lie on the line segment qy (otherwise, .τ cannot
be tangent to .σ3 ). Then .σ3 is contained in the union of the disk bounded by .σ1 and
.Δxyq. Since the contact point of .σ2 (⊂ [l1 ] ∩ [l3 ]) and .l3 lies on the extension of

the line segment xq beyond q, no line can be properly tangent to all three circles
13.6 An Extension of Casey’s Theorem 271

Fig. 13.6 ...

.[l1 ]
∩ [l2 ] ⊂ [l1 ] ∩ [l3 ]
...
l
... 3 l
...
... ... 2
... ...
... . ...
.. ..
... ...... .... .... ... .... ...
...... . . ....... ...
.. .. ...
.. .
... σ 1
.....
x
.. .
...
●..... ....
...
..
.. .... ...
. .. q
l ↓
.....................................................................................................●
1 .. .
... . ...
..
..........................................................................
.... ..........
↓ l1
.. ..........
..
... ..
.... .
.. .. .
y . . . ...
.●
..
.
.
....
....
..... ...... .....
.. .... .... .... ....... .... .... ........ .....
... .. . .. ....
... ... . .
.. ....
... .....
.. .. . ...
..
...
..
...
..
σ 2 .....
..
l
.
...
... . . 2
... ....
.
l .... ... . 3
.... .... ........ .... .... ....

Fig. 13.7 .i /= j .⇒ .γi and .cj

are tangent

γi ci

σ1 , σ2 , σ3 , contradicting the property that .l4 is properly tangent to these three circles.
.

Thus, if we assume that .l1 , l2 , l3 , l4 are all different, then we have a contradiction.
⨆
⨅
Remark 13.3 Theorem 13.3 implies the non-existence of double 5-systems of
circles. For details, see [536].
Remark 13.4 It is impossible to replace “properly tangent” simply by “tangent” in
Theorem 13.3, see Fig. 13.7.

13.6 An Extension of Casey’s Theorem

For a semi-metric space .X and .k > 2, we write

. det SD(X : k) ≡ 0

if .det SD(Y) = 0 holds for every subset .Y ⊂ X with k elements. Using this
notation, Ptolemy’s theorem can be extended as follows: For a subset .X ⊂ R2
consisting of n points, .n ≥ 5, .X lies on a circle or a line if and only if .det SD(X :
4) ≡ 0. However, since three points determine a circle uniquely, this is trivial. On
the other hand, a similar extension of Casey’s theorem seems to be not obvious.
Theorem 13.4 For a bicolored set .F of .n (≥ 4) circles in .R2 , the following two
statements are equivalent.
272 13 Casey’s Theorem

(i) There is a circle or a line that is properly tangent to the circles in .F.
(ii) .det SD(F : 4) ≡ 0.
Lemma 13.8 For a bicolored set of three circles .F = {c1 , c2 , c3 }, the number of
circles or lines that are properly tangent to the circles in .F is at most two.
Proof By Lemma 13.4, we may suppose that a circle in .F, say .c3 , is a null circle
{p}. The point p lies outside both .c1 , c2 . Let .ϕ be the inversion of the plane with
.

respect to the unit circle with center p. Let .σ1 = ϕ(c1 ), σ2 = ϕ(c2 ). If .γ is a circle
or a line that is properly tangent to .c1 , c2 , c3 , then .ϕ(γ ) is a line that is properly
tangent to .σ1 , σ2 . The number of lines that are properly tangent to .σ1 , σ2 is at most
two. (If one of .σ1 , σ2 is not a null circle, then there are exactly two lines that are
properly tangent to .σ1 , σ2 .) Hence the number of circles or lines that are properly
tangent to .c1 , c2 , c3 is at most two. ⨆
⨅
Proof of Theorem 13.4 The part (i) .⇒ (ii) follows from Casey’s theorem. So we
show the part (ii) .⇒ (i). Suppose that (ii) holds. If .n = 4, (i) follows from Casey’s
theorem. If .n = 5, then (i) follows from Theorem 13.3. Assume that (i) holds for an
.m ≥ 5, and consider the case .n = m + 1. Let .c1 , c2 , . . . , cm+1 be colored circles of

.F. For each .i = 1, 2, . . . , m + 1, there is a circle or a line .γi that is properly tangent

to the circles in .F \ {ci } by the inductive assumption. By Lemma 13.8, there are at
most two circles (or lines) that are properly tangent to three circles .c1 , c2 , c3 . Since
.γ4 , γ5 , γ6 are all properly tangent to .c1 , c2 , c3 , some two of .γ4 , γ5 , γ6 coincide.

Suppose .γ4 = γ5 . Then .γ4 (= γ5 ) is properly tangent to all circles in .F, and hence
(i) holds. ⨆
⨅

13.7 Exercises

13.1 Give an example of a bicolored set in R2 in which the common-tangent

distance does not satisfy the triangle inequality.
13.2 Using Lemma 13.2, prove Corollary 13.1.
13.3 Let c1 be a red circle of unit radius centered at (−2, 0), and c2 be the blue
circle of unit radius centered at (2, 0). Let ϕ be the inversion of the plane
with respect to the unit circle centered at (3, 3). What is the common tangent
distance t (ϕ(c1 ), ϕ(c2 ))?
13.4∗ Prove the following assertion: If all circles in a bicolored set F of four circles,
each having positive radius, intersect at a common point, then det SD(F) =
0.
13.5∗ Suppose that two circles c1 , c2 of the same color intersect with angle θ , see
Fig. 13.8. Prove that

t (c1 , c2 )2
. = 2(1 + cos θ ),
r1 r2

where r1 , r2 are radii of c1 , c2 , respectively.

13.8 Notes 273

Fig. 13.8 Two circles

intersecting with angle θ

c1 ● ● c2

13.8 Notes

Although Ptolemy’s theorem was already discussed in Sect. 12.5, we present the
literature around it here in Chap. 13. We do this due to its proximity to Casey’s
theorem, which gets also a broad representation in the notes here.

13.8.1 Notes on Ptolemy’s Theorem

Ptolemy’s theorem from planar Euclidean geometry, going back to the first book of
Claudius Ptolemy’s Almagest (see also our Sect. 12.5 here), and its converse can be
formulated together as follows: a quadrilateral is cyclic (i.e., inscribed to a circle)
iff the product of the lengths of its diagonals equals the sum of the products of
the lengths of the opposite pairs of sides. Directly related is Ptolemy’s inequality
which is an equality exactly when the four considered vertices of a quadrilateral
form a cyclic quadrilateral (otherwise the product of the diagonal lengths is smaller).
Ptolemy’s result can be found at many places in the literature. As already said, we
discuss notes on Ptolemy’s theorem in this chapter here since it is a special case of
and a good starting point for Casey’s theorem, which is the main topic of Chap. 13.

Some History

Due to its popularity, Ptolemy’s theorem is mentioned in various books and surveys
on history of mathematics. For a more detailed discussion, one might look at the
chapter on trigonometry in the monograph [309] of González-Velasco, where also
applications and further work of Ptolemy are studied. Another interesting historical
reference is the recent expository paper [418] of Ibragimov and Suceava, discussing
Ptolemy’s theorem and Ptolemy’s inequality. Also applications are given there, and
Ptolemy’s inequality is discussed in the context of several metrics (such as, e.g.,
the taxicab one). Furthermore, the authors discuss in an informative way Ptolemy’s
work Almagest.
274 13 Casey’s Theorem

Due to Tanaka et al. [772], Ptolemy’s theorem can also be found in the Japanese
temple geometry. And Gupta [350] presents extensions of Ptolemy’s theorem (with
simple proofs) due to Indian mathematicians.

Book Chapters and Surveys

We will mention first the following books where Ptolemy’s theorem is nicely
presented (the list is certainly not complete): Altshiller-Court [28, pp. 127–134],
Johnson [430, Chapter IV], Perfect [630, Chapter 3], Coxeter and Greitzer [186,
pp. 42–43], Eves [250, Section 3.8], Berger [75, 10.9.2], Pedoe [625, Section 24.1],
Bottema [117, Section 22], Borceux [103, Section 4.13], Aumann [48, Section 6.1],
as well as Posamentier and Geretschläger [644, Chapter 2].
Further books containing Ptolemy’s theorem and related topics are [358] due
to Hahn, based on complex numbers, and the book [326] of Grigorieva collecting
elementary geometric problems of competition type for high school or college
students. And also in the relatively recent books [25, Chapter 2] and [26, Chapter 1]
by Alsina and Nelsen, showing the rich geometry of quadrilaterals and polygons,
Ptolemy’s theorem is studied.

Extensions of Ptolemy’s Theorem

The motivation for Bowman’s paper [119] goes back to Möbius who studied the
problem of finding, with given side-lengths, all cyclic configurations of polygons. To
get progress in this question, Bowman uses repeated application of Ptolemy’s theo-
rem. Various elliptic analogues of theorems from elementary geometry are proved
in [342], among them Ptolemy’s theorem. Valentine (see [790] and [791]) presents
spherical and hyperbolic variants of Ptolemy’s theorem. Bilinski [89] develops
a general concept containing, e.g., Ptolemy functions and Ptolemy matrices, to
unify approaches to many (at first glance seemingly independent) theorems from
elementary geometry. These are, e.g., the theorems of Monge, Möbius, Kolmogorov,
Ptolemy, and Casey.
Using systems of forces being in equilibrium, Jacob [426] derives elegantly the
theorems of Ptolemy, Möbius, Neuberg and Pompeiu getting cross-connections to
further theorems. His starting point is a theorem due to Mitrinović, Pécarić and
Volenec which expresses the cyclic position of four points in terms of a suitable
accompanying triangle; it is degenerate iff the four points lie on a circle. Smith
[744] extends the triangle inequality to an alternating inequality for each cyclic
polygon with an odd vertex number, where equality holds for an even number. Via
inversion the triangle inequality is transformed into Ptolemy’s inequality, and this
is extended to cyclic polygons, yielding a generalization of Ptolemy’s theorem. The
author studies the analogous results in Minkowski planes, and he continues these
investigations in [745] showing the connections between Euclidean and Minkowski
geometry based on Ptolemy’s theorem with a Minkowski metric.
13.8 Notes 275

Alsina et al. [27] give a characterization of those bijections that preserve the
property of quadrilaterals to be cyclic, and they extend this to real inner product
spaces. Popescu [642] gives several extensions of Ptolemy’s theorem to sets
consisting of more than four points. Gregorac [323] studies a Feuerbach and a
Möbius analogue of Ptolemy’s theorem, and he extends this (using simplex volumes
instead of triangle areas) to n dimensions. Here the proof is based on inversion in
(.n − 1)-spheres and a determinant rule for the n-dimensional version of Heron’s
formula. A determinant is used which disappears iff .n + 2 considered points lie in
an (.n − 1)-sphere or a hyperplane; several interesting special cases are discussed,
too.
Askey [42] uses complex numbers to extend Ptolemy’s theorem from cyclic
quadrilaterals to general ones. The same author discusses in [43] Brahmagupta’s
extension of Ptolemy’s theorem to find (in the cyclic quadrilaterals) the lengths of
the diagonals, of the segments generated by the intersection points of the diagonals,
and the sides. This is extended also to the figures obtained by prolongation of the
opposite sides until they meet, etc.; the author gives nice proofs.

Interesting Proof Methods

The papers [642] and [323] discussed already above elegantly use inversion
arguments to prove their theorems. And Askey [42] successfully uses complex
numbers.
Pech [622] proves several theorems around Ptolemy’s theorem via computations
based on the theory of automated geometry theorem proving, using Gröbner bases.
The novelty is to use this method for proving geometric inequalities.
It is known that cyclic quadrilaterals have many geometric properties which have
the character of necessary conditions. The fact that most of them are also sufficient
is not so much taken care of. Josefsson (see [432] and [433]) presents altogether
34 characterizations with proofs based on very different tools. The paper [274] of
Fraivert et al. is similar in nature.

Applications

Motivated by Ptolemy’s theorem and the Brahmagupta formula, Khimshiashvili

[451] studies cyclic polygons in relation to polygonal linkages, also in view of
mathematical models of mechanical and chemical systems. The investigations in
[451] are based on a new aspect, namely to interpret cyclic polygons as critical
points of certain functions on the moduli space of the respective polygonal linkage.
Percy and Rogers [629] nicely discuss the relations between the Fermat-
Torricelli problem and Ptolemy’s theorem, including also a historical survey.
Schwarz and Smith [720] study cyclic quadrilaterals having correspondingly
the same four side lengths .a, b, c, and d, but different order of the corresponding
sides. Namely, given such a quadrilateral, there are in general two other ones with
the same side lengths in different order. Their circumradius R and area remain, and
276 13 Casey’s Theorem

the lengths of their diagonals can reach altogether only three values: .p, q, and r. It
is proved that for any of the three configurations the following relations hold: (1)
.(p + q + r) − (a + b + c + d) = (abcd)/tR , where .t := a + b + c + d + p + q + r;
2

(2) .a +b +c +d < p +q +r; (3) .(p +q +r 2 )−(a 2 +b2 +c2 +d 2 ) = (abcd)/R 2 ,

2 2

implying .a 2 + b2 + c2 + d 2 < p2 + q 2 + r 2 . E.g., for proving (1) they use three

applications of Ptolemy’s theorem.

13.8.2 Notes on Casey’s Theorem

Ptolemy’s theorem on cyclic quadrilaterals is a natural starting point for considering

Casey’s theorem.

The Theorem Itself and Some History

Ptolemy’s theorem was generalized by John Casey (cf. [142] and [143]) who
replaced the vertices of Q by circles tangent (outside or inside) to C at the vertices
of Q (see also Taylor [773]). More precisely, Casey proved the following statement,
usually called Casey’s theorem: Four circles are tangent to one circle if and only
if the sum of the products of the two pairs of opposite tangential distances of these
four circles equals the product of the diagonal tangential distances. For example,
one can take in all cases (sides and diagonals) outer tangents throughout. Thus, the
property “four points lie on a circle” (Ptolemy) is analogous to “four circles are
tangent to a common circle” (Casey) if we replace points by circles and the usual
distance of points suitably by the notion of tangential distance of circles.
Like, e.g., the Malfatti problem and various further theorems and problems,
Casey’s discovery is also an example from the set of geometric theorems which
were found, independently and earlier than by Europeans, in the Japanese Temple
Geometry of the Edo period. For interesting comments, proofs, applications, and
connections to related theorems we refer to Hayashi [386], Mikami[573], as well
as Fukagawa and Pedoe [284, Section 3.3]. More precisely, the theorem was
known and used already in 1874, as a Sangaku problem from the Gunma prefecture.
Moreover, the Japanese mathematician Chochu Siraishi knew it already in 1820.
Meanwhile Casey’s theorem is well-established and discussed in various mono-
graphs. E.g., one can find it (differently proved and applied) in the following books:
Coolidge [177, pp. 37–38], Johnson [430, p. 124], Eves [249, p. 129], Bottema
[117, Chapter 22], Yaglom [822, pp. 69–71], and Aumann [48, pp. 121–123]. It is
also cited in the encyclopedia [805], under “Casey’s theorem”, and it even found
access to classroom geometry, as the two short and nicely written articles [739] and
[821] show. In the first paper six problems are treated which are based on tangential
distances of circles inscribed into another circle in various configurations, and in the
second article historical background and applications of the theorems of Ptolemy,
Casey, and Feuerbach are given.
13.8 Notes 277

Extensions, Modifications, and Applications

In view of the tangency notion defined for oriented circles (cycles), Zacharias [829]
presents Casey’s theorem in a more rigorous form. Casey’s theorem has also many
applications, e.g. with respect to Feuerbach circles (cf. [664] and also [343], where
many relations between Ptolemy’s, Casey’s and further basic theorems are derived).
Also Duparc [229] lists interesting applications of Casey’s theorem. Finding it
among other nice theorems from elementary geometry, the same author shows in
[230] how this theorem is intertwined with the others.
Bilinski [89] introduces certain incidence configurations and matrices in projec-
tive n-space to study comprehensively their subcases, which include an impressive
list of classical theorems from elementary geometry, among them theorems of Euler,
Monge, Möbius, Ptolemy, and Casey. Higher-dimensional versions of Casey’s
theorem are given by Abrosimov and Aseev [3] and by Maehara and Martini [535].

There exist also numerous extensions to non-Euclidean geometries: Kubota [478]

gives a hyperbolic version of Casey’s theorem, and an elliptic analogue is derived
by van Gruting [342]. Kurnik and Volenec (see [482] and [483]) obtain non-
Euclidean versions extended to an arbitrary number of circles; these can be also
horo- or hypercycles in the hyperbolic case. Hyperbolic analogues are also due
to Abroshimov and Mikaiylova [4] as well as Kostin and Kostina [475], in the
first paper additionally considering spherical geometry, and in the second pseudo-
Euclidean and pseudo-hyperbolic geometry. Abrosimov and Mednykh [5] present
various interesting results referring to area and volume in spherical and hyperbolic
geometries. Among them, they also present spherical and hyperbolic analogues of
Casey’s theorem.
A bipartite set of spheres in n-space is a set of colored (.n − 1)-spheres, where
two colors are used, no sphere is contained in the closed ball bounded by another
sphere in the set, and spheres of different colors are disjoint. For any two such
spheres, their common-tangent distance is the distance between two tangent points
in a common tangent hyperplane, where an outer common tangent hyperplane is
taken if the two spheres have the same color (otherwise, a common inner tangent
hyperplane is taken). Due to this distance type, a bipartite set becomes a semi-
metric space. E.g., Casey’s theorem presents a condition for a bipartite set of four
circles in the plane having a circle which is suitably tangent to all these circles of
a bipartite set. Based on Ptolemy configurations in n-space obtained via Cayley-
Menger determinants and among various further results, Maehara and Martini [535]
derive a Casey-type extension of the n-dimensional Ptolemy configuration, referring
to bipartite families of .n + 2 spheres in n-space. And this is then further extended
to bipartite sphere families of arbitrarily large cardinality.
Chapter 14
Solutions to the Selected Exercises

Obvious is the most dangerous word in mathematics.

— Eric Temple Bell

1.4∗ Let .a, b be the centers and .rA , rB be the radii of the disjoint circles .A, B,
.

respectively. Since B lies in the interior of A, we have .rA > rB . Let C be a

circle that is internally tangent to A and externally tangent to B. Let .x be the
center of C and r be the radius of C. Then

|ax| = rA − r, |bx| = rB + r.
.

Hence .|ax| + |bx| = rA + rB = const. Therefore the trace of .x is an ellipse

with foci .a, b. ⨆
⨅
.1.7
∗ Let .Π be a regular icosahedron inscribed in a sphere .S, and let V be the

vertex set of .Π . Note that each vertex of .Π has degree five. For each .x ∈
V , let .C(x, r) be the spherical cap on .S with center x and angular radius
r. If r is very small, then the spherical caps .C(x, r), x ∈ V , are disjoint.
When we increase r continuously, there is a moment that each spherical cap
.C(x, r) touches its neighboring caps. At this moment, the boundary circles

.∂C(x, r), x ∈ V , form a family of circles on .S such that each circle is tangent

to exactly five other circles in the family. Let .p ∈ S be a point that is not
covered by any spherical cap, and let .ϕ be the stereographic projection of .S
from p to the plane that is tangent to .S at .p∗ . Then, .{ϕ(∂C(x, r)) : x ∈ V } is
a family of circles in which each circle is tangent to exactly five other circles
in the family. ⨆
⨅
.2.4
∗ By Soddy’s formula, .b(τ ) and .b(τ ) are the solutions of the quadratic
1 2
equation in X:
⎛ ⎞2 ⎛ ⎞
⎲
4 ⎲
4
. X+ b(σi ) =3 X +2
b(σi ) 2
.
i=1 i=1

© The Author(s), under exclusive license to Springer Nature Switzerland AG 2024 279
H. Maehara, H. Martini, Circles, Spheres and Spherical Geometry,
Birkhäuser Advanced Texts Basler Lehrbücher,
https://doi.org/10.1007/978-3-031-62776-7_14
280 14 Solutions to the Selected Exercises

Rewriting this equation, we have

⎛ ⎞ ⎛ 4 ⎞2
⎲
4 ⎲
4 ⎲
2X − 2
.
2
b(σi ) X + 3 b(σi ) − b(σi ) = 0.
i=1 i=1 i=1

Then, by Vieta’s formula, we obtain

b(τ1 ) + b(τ2 ) = b(σ1 ) + b(σ2 ) + b(σ3 ) + b(σ4 ).

.

⨆
⨅
2.5∗ Since .σ1 , σ2 , α, β are mutually tangent circles and .σ3 , σ2 , α, β are also
.

mutually tangent circles, we can see, by Descartes’ formula, that .b(σ1 ) and
.b(σ3 ) are the solutions of the following quadratic equation in X:

. (X + b(σ2 ) + b(α) + b(β))2 = 2(X2 + b(σ2 )2 + b(α)2 + b(β)2 ).

Rewriting this equation we have

X2 − 2 (b(σ2 ) + b(α) + b(β)) X

.

+ 2b(σ )2 + 2b(α)2 + 2b(β)2 − (b(σ2 ) + b(α) + b(β))2 = 0.

Then, by Vieta’s formula, we get

b(σ1 ) + b(σ3 ) = 2 (b(α) + b(β) + b(σ2 )) .

.

Similarly, from the mutually tangent four circles .σi , σ3 , α, β for .i = 2, 4, we

have

b(σ2 ) + b(σ4 ) = 2 (b(α) + b(β) + b(σ3 )) .

.

By eliminating .b(α) + b(β) from these two, we have .b(σ1 ) + 3b(σ3 ) =

b(σ4 ) + 3b(σ2 ). ⨆
⨅
.3.5
∗ Let .v, e, f denote the numbers of vertices, edges, and faces, respectively, of

a quadrangulation G. Noting that the minimum degree of G is at least 3, we

have

3v ≤ 2e = 4f.
.

Hence .e = 2f and .3v ≤ 4f . By Euler’s formula, one obtains

2 = v − e + f = v − 2f + f = v − f,
.
14 Solutions to the Selected Exercises 281

and hence .v = f + 2. Thus .3v = 3f + 6 ≤ 4f , from which we have .6 ≤ f .

Therefore, .v = f + 2 ≥ 6 + 2 = 8. ⨆
⨅
.3.7
∗ Since G is a 3-connected plane graph, its radial graph .R(G) is an irreducible

quadrangulation. Let .F be an OCR of .R(G). For four circles of .F that

meet at a point, connect the centers of each pair of tangent circles by a
line-segment. (If one circle is a special circle .Cω , then take, instead of a
line segment, a half line emanating from the center of the circle that is
internally tangent to .Cω in the direction to the contact point.) Then each
pair of intersecting line-segments (or a line segment and a half line) crosses
orthogonally. Such a pair of line-segments corresponds to an edge of G and
its dual edge of the dual graph .G∗ . Thus, we can get a desired drawing of G
and .G∗ . ⨆
⨅
Remark This question on an orthogonal drawing of G and .G∗ was first
asked by W. T. Tutte in 1963 ([785, §13]), and it was answered by
G. R. Brightwell and E. R. Sheinerman [128].
4.4∗ Let .α, β, γ be the interior angles of the spherical triangles .Δ̆ABC at .A, B, C
.

respectively. Then the interior angles of .Δ̆A∗ B ∗ C at .A∗ , B ∗ , C are .π −α, π −

β, γ respectively. Now

[ABC] > π ⇔ α + β + γ − π > π

.

⇔ γ > (π − α) + (π − β)
⎛

⇔ A∗ CB ∗ is a minor arc.

⋂ ⨆
⨅
.4.5∗ If .P ∈ X◦ , then, since .X◦ = Q∈X H (Q), we have .P Q ≤ π/2 for all .Q ∈
X. Hence .X ⊂ H (P ). Then .conv(X) ⊂ H (P ), and hence .conv(X) ⊂ X◦◦ .
If .X ⊂ H (P ), then .X◦ ⊃ H (P )◦ = {P }, and hence .X◦◦ ⊂ H (P ).
Therefore, .X◦◦ ⊂ conv(X). ⨆
⨅
.4.6
∗ Let .r > ρ be the radius of the sphere, O the center of the sphere, Z the point
on the sphere where the needle of the compass is fixed, and P be a point on
the circle. Let .θ = / P OZ. Then, by Example 4.1, the area of the cap is equal
to .2π r 2 (1 − cos θ ). Since .1 − cos θ = 2 sin2 (θ/2), and .ρ = 2r sin(θ/2), the
area of the cap is

2π r 2 (1 − cos θ ) = 2π r 2 2 sin2 (θ/2) = πρ 2 ,

.

which is independent of r. ⨆
⨅
282 14 Solutions to the Selected Exercises

5.1∗ Consider the spherical triangle ABN , where N is the north-pole. Then .AN =
.

BN = π/2 − π/4 = π/4 and ./ ANB = π/2. Applying the spherical cosine
law we have

. cos AB = cos(π/4) cos(π/4) − 0 = 1/2.

Hence .AB = π/3. ⨆

⨅
5.2∗ By the spherical cosine law,
.

. cos d = cos2 d + sin2 d cos θ = cos2 d + (1 − cos2 d) cos θ.

From this we have

(cos d − 1) ((1 − cos θ ) cos d − cos θ) = 0.

.

Since .0 < d, we have .cos d = 1−cos

cos θ
θ , and .cos θ = 1+cos d .
cos d
⨆
⨅
.5.3
∗ Let .β = B, γ = C. Let .ϕ be the stereographic projection of .S from .A∗ to
/ /
the plane .Π that is tangent to .S at A.
(i) Let .Λ be the spherical cap with center A and radius AC. Then B lies in
the interior of .Λ. The image .ϕ(Λ) is a disk in the plane .Π with center
.Ā = A, .B̄ lies in the interior of the disk .ϕ(Λ), and .C̄ lies on the boundary

of this disk. Hence .|ĀC̄| > |ĀB̄| and so ./ B̄ > / C̄ by the Proposition
18 of Euclid’s Elements. Since the angles of the Cesàro triangle .ĀB̄ C̄ at
.B̄, C̄ are .β − ε, γ − ε, respectively, it follows that .β − ε > γ − γ . Hence

.β > γ .

(ii) can be proved similarly, and then (iii) becomes obvious. ⨆

⨅
.5.4∗ It is sufficient to consider the case that .X, Y lie on the boundary of .Δ̆ABC.
Suppose X lies on AB and Y lies on AC, see Fig. 14.1. Note that one of the
angles ./ AXY, / BXY is not less than .π/2. If ./ BXA ≥ π/2, then in .Δ̆AXY ,
/. A ≤ / X implies .XY ≤ AY by Exercise .5.3∗ . And since .AY < AC, we
have .XY < AC. Suppose that ./ BXY ≥ π/2. In the triangle .Δ̆BXY , we
have ./ BXY > / XBY . Hence .XY < BY . By a similar argument, we have
.AY < max{AB, BC}. Hence .XY < max{AB, BC, CA}. ⨆
⨅
.5.6
∗ Let ABCD be a convex quadrilateral on the sphere .S with .AB = AD =

a, BC = DC = b. Then .[ABCD] = [ABC] + [ADC] = 2[ABC]. Hence

Fig. 14.1 Exercise .5.4∗ A ....

..●
.... ......
.... ....
....... ....
... ....
....... ....
....
....
X ....●
..
.
......... ....
.... .... .... ..
.. .... .... ....
...
...
..

.... .... . . ..
.... .
... .
... .. .
. .... .... ....
..
.
.
.
.
..
..
... Y
...................
●..
...
. .... .... .... ... .
●............................................................................................................● ...

B C
14 Solutions to the Selected Exercises 283

.[ABCD] becomes the maximum when .[ABC] becomes the maximum. By

⎛
Theorem 5.7 (2), .[ABC] becomes the maximum when .ABC is a semicircle.
Hence .[ABCD] becomes the maximum when it is inscribed in a spherical
cap. ⨆
⨅
.6.2
∗ Suppose that there is a quadrangulation G with 9 vertices. Let .e, f be the

numbers of edges, faces, respectively. Then .2e = 4f , and hence .e = 2f . By

Euler’s formula, .2 = 9 − e + f = 9 − f , and .f = 7. Therefore, .e = 14.
Since the degree of each vertex is at least 3, we can deduce, by considering
the hand-shaking lemma, that just one vertex has degree 4, and the remainig
8 vertices have degree 3. Let us try to draw such a quadrangulation G on
the plane. Let A be the vertex with degree 4 and ABCD be a quadrangular
face of G. We may assume that ABCD is not the outer face of G. Then,
from A two edges must emanate, and from each of .B, C, D, one edge
must emanate. Let .AA1 , AA2 , BB ' , CC ' , DD ' be such edges in the counter-
clockwise order. It is obvious that none of .A1 , A2 , B ' , C ' , D ' belongs to
' ' '
.{A, B, C, D}. It is also not difficult to see that .A1 , A2 , B , C , D are all
' ' ' ' ' '
different. Since all faces are quadrangles, .A1 B , B C , C D , D A2 must
be edges of G, see Fig. 14.2. Then, since 13 edges appeared so far, there
remains just one edge we can add. Since .A1 , A2 are not adjacent, we need
(to make their degrees 3) two more edges, which is impossible. Therefore, it
is impossible to draw a quadrangulation with nine vertices. ⨆
⨅
.6.4
∗ Suppose that two balls of radius 2 touch one unit ball with center O at .P , Q.

Then, from

42 ≥ 32 + 32 − 2 × 32 cos / P OQ,
.

we have ./ P OQ ≥ d := cos−1 (1/9). Hence the spherical distance between

any two contact points on the boundary sphere .S of the unit ball is at least

d := cos−1 (1/9) ≈ 1.45945.

.

By Exercise .5.2∗ , the interior angle .θ of a .Δ̆(d, d, d) is given by

⎛ ⎞
−1 cos d
θ = cos
. = cos−1 (1/19) ≈ 1.47063.
1 + cos d

Fig. 14.2 Exercise .6.2∗ D C

D C

A2 A B

A1 B
284 14 Solutions to the Selected Exercises

Hence the area of an equilateral spherical triangle with edge-length d is

.3θ − π ≈ 1.270.

Let n be the maximum value of the number of balls of radius 2 that can
simultaneously touch one and the same unit ball. Let .X be the set of
contact points of n balls with the boundary .S of the unit ball. Then .X
is not contained in a hemisphere. (Otherwise, another ball of radius 2 can
touch the unit ball, contradicing that n is the maximum number. Hence the
convex hull .conv(X ) of .X is a convex polyhedron containing the center
of .S. Triangulate each face of .conv(X ) by adding diagonals, if necessary,
and project each triangle, from the center O onto .S. Then .S is divided into
spherical triangles. Let f be the number of spherical triangles and e be the
number of edges in this triangulation of .S. Note that each spherical triangle
is inscribed in a spherical cap with angular radius less than d. (Otherwise,
another ball of radius 2 can touch at the center of the circumscribed spherical
cap, contradicting the maximality of n.) Then, by Fejes Tóth’s Lemma, each
spherical triangle has area .≥ [Δ̆(d, d, d)] ≈ 1.27. Since the sum of the areas
of f spherical triangles is equal to .4π , and each spherical triangle has area
at least .[Δ̆(d, d, d)], we have .f ≤ 4π/1.27 ≈ 9.894. Thus .f ≤ 9. If .f = 9,
then .2e = 3f which contradicts the fact that e is an integer. Hence .f ≤ 8.
If .f = 8, then from Euler’s formula we have .4 = 2n − 2e + 2f = 2n − f ,
and .n = (8 + 4)/2 = 6. Thus, .n ≤ 6. On the other hand, six balls of radius
2 with centers at

(±3, 0, 0), (0, ±3, 0), (0, 0 ± 3)

.

are disjoint balls that touch the unit ball with center .O = (0, 0, 0). Hence
n = 6.
. ⨆
⨅
.7.2
∗ Consider the spherical quadrilaterals .ABCP , ABQC. It is clear that

.[ABP ] =
2 [ABCP ] = [ABC]. Hence .[ABP ] = [ABC]. Similarly,
1
⎛

we have .[ABQ] = [ABC]. Thus, .P , C, Q lie on the Lexell arc .A∗ CB ∗ .

⎛
⎛

Therefore, .P CQ is a subarc of .A∗ CB ∗ . ⨆

⨅
.7.3
∗ Let .Δ̆ABX be a spherical triangle such that .AB = a, BX = b, XA = c.
Let .KA , KB be the circles on .S of spherical radius a with centers A and B,
respectively. Let .C ∈ S be the intersection point of .KA and .KB lying in the
same side of AB as X. Let .KC be the circle on .S with spherical radius a and
center C. Let P be the intersection point of .KA , KC different to B, and Q
be the intersection point of .KC , KB different to A. Since the circumradius of
.Δ̆(a, b, c) is smaller than a, the vertex X lies in the interior of .KC (that is,

the same side of .KC as C), and since .a ≤ b ≤ c, X lies in the exterior of .KA
and in the exterior of .KB . Since .b ≤ c, the edge AX must intersect the minor
14 Solutions to the Selected Exercises 285

⎛
(
arc .P C of .KA . Therefore, AX intersects the arc .P CQ. Then, by Exercise
7.1, we have .[ABX] ≥ [ABC] = [Δ̆(a, a, a)]. ⨆
⨅
.7.6
∗ Among the spherical regions with a given perimeter s, let .Ω be one that has

the maximum area. We may suppose that .Ω is a convex region. (Otherwise,

it would be possible to increase the area by deforming .Ω with preservation
of the perimeter.) Suppose that .Ω is not a spherical cap. Then there are
four points .A, B, C, D (in this cyclic order) on the boundary of .Ω such
that the spherical quadrilateral ABCD is not cyclic. The four segments
.AB, BC, CD, DA divide .Ω into five regions: the quadrilateral ABCD and

four “lune-like” regions .α, β, γ , δ (some of them could be empty). Since

the convex quadrilateral ABCD is not cyclic, it has a proper diagonal.
By increasing the length of this proper diagonal a little bit keeping four
edge-lengths fixed, we can get a quadrilateral with a larger area by the proper-
diagonal lemma. By attaching four regions congruent to .α, β, γ , δ to this new
quadrilateral from the outside of this new quadrilateral, we have a region
whose area is greater than .Ω and whose perimeter is s. This contradicts the
fact that the spherical region .Ω is one that has the maximum area. ⨆
⨅
.8.2
∗ Let .X = {x : i = 1, 2, 3, 4, 5} be the set of five random points on .S chosen
i
independently. By Exercise .4.5∗ , we have .conv(X) = X◦◦ . Then

X is the vertex set of a convex pentagon

.

⇔ X◦ is a spherical pentagon (by Theorem 4.8)

⇔ H (x 1 ) ∩ H (x 2 ) ∩ H (x 3 ) ∩ H (x 4 ) ∩ H (x 5 ) is a pentagon.

Now let us take five random points in the following two-steps way: First
choose five random great circles .G1 , . . . , G5 independently. Next, from the
pair of poles of each great circle .Gi , choose either one with equal probability.
In this two-steps way, we can take five random points .x i , i = 1, 2, . . . , 5.
Five great circles .Gi , i = 1, 2, . . . , 5, divide .S into 22 regions. To choose
a pole of .Gi corresponds to choosing one side of .Gi , and .X◦ = H (x 1 ) ∩
· · · ∩ H (x 5 ) is nonempty if and only if .X◦ is one of the 22 regions divided by
.Gi , i = 1, . . . , 5. Since there are .2 = 32 ways to choose poles, and among
5

22 regions, only two are pentagons, we can deduce that

. Pr(X◦ is a pentagon) = 2/25 = 1/16.

⨆
⨅
8.6∗ Let .Γ be a semicircle of spherical radius .ρ, and .A, B be the endpoints of
.

.Γ . Then .AB = 2ρ and .conv(Γ ) is bounded by .Γ ∪ AB. The length of

.Γ is .π sin ρ, and the perimeter of .conv(Γ ) is .π sin ρ + 2ρ. Note that the

probability that a random great circle .G intersects .conv(Γ ) in only one point
is zero. A random great circle .G cuts .conv(Γ ) if and only if (i) .G intersects
.Γ twice or (ii) .G crosses AB. Since the events (i) and (ii) are exclusive, we
286 14 Solutions to the Selected Exercises

have

. Pr(G ∩ conv(X) /= ∅) = Pr((i) occurs) + Pr((ii) occurs).

Therefore,

. Pr(G intersects Γ in two points)

= Pr(G ∩ conv(Γ ) /= ∅) − Pr(G ∩ AB /= ∅).

By Crofton’s formula we have

π sin ρ + 2ρ 2ρ
. Pr(G ∩ conv(Γ ) /= ∅) = , Pr(G ∩ AB /= ∅) = ,
2π π
and hence
π sin ρ − 2ρ
. Pr(G intersects Γ in two points) = .
2π
⨆
⨅
.9.2∗ If .d ≥ π/2, then any two caps of radius d intersect. Hence, if .d ≥ π/2, then
.K(2, 2) is not a d-cap graph. Suppose .π/4 ≤ d < π/2. Let .w, x, y, z ∈

S be four points lying on a great circle in this cyclic order and satisfying
.wx = xy = yz = zw = π/2. Then, the intersection graph of the four caps

.Cd (w), Cd (x), Cd (y), Cd (z) is .K(2, 2). Next, suppose that .0 < d < π/4.

Let .w, x, y, z be four points lying on a circle K in this cyclic order such that
.wx = xy = yz = zw = 2d. Then the angular radius of K is greater than d.

Hence the intersection graph of the four caps .Cd (w), Cd (x), Cd (y), Cd (z) is
.K(2, 2). Thus, for any .0 < d < π/2, .K(2, 2) is a d-cap graph. ⨆
⨅
.9.3
∗ If .d ≥ π/4, then since .K(1, 4) is not a d-cap graph, .K(2, 4) is not a d-

cap graph. Suppose .d < π/4. Then for any d-independent two points .x, y,
.C2d (x) ∩ C2d (y) cannot contain d-independent three points by Cororally 9.2.

Hence .K(2, 4) is not a d-cap graph. ⨆

⨅
.9.6
∗ Let .x, y be two random points on the unit sphere .S chosen independently,

and .C(x), C(y) be the caps of angular radius .θ and with centers .x, y,
respectively. Then .C(x), C(y) are two random caps. Let W be the area of
.C(x) ∩ C(y). Then W is a random variable. Let .g(w) denote the density
⎰
function of W . Then .E(W ) = wg(w)dw. Let .z be a random point on .S.
By the law of total probability, we have
⎰
. Pr(z ∈ C(x) ∩ C(y)) = Pr(z ∈ C(x) ∩ C(y)|W = w)g(w)dw
⎰
w E(W )
= g(w)dw = .
4π 4π
14 Solutions to the Selected Exercises 287

Let .C(z) be the spherical cap of angular radius .θ with center .z. Then

. Pr(z ∈ C(x) ∩ C(y)) = Pr(x, y ∈ C(z))

⎛ ⎞
2π(1 − cos θ ) 2 (1 − cos θ )2
= = .
4π 4

Therefore, .E(W ) = π(1 − cos θ )2 . ⨆

⨅
.9.7
∗ Let .V (n, p) = S \ (the union of n random caps). (Thus, .α(n, p) is the area
of .V (n, p).) For a random point .x and n random spherical caps of area .4πp
chosen independently on .S, let us compute the probability .Pr(x ∈ V (n, p))
in two different ways.
First, let us regard that n random caps are chosen before we choose a
random point .x. Since

. Pr(x ∈ V (n, p)|U (n, p) = u) = u,

we have
⎰
. Pr(x ∈ V (n, p)) = Pr(x ∈ V (n, p)|U (n, p) = u)g(u)du
⎰
= ug(u)du = E(U (n, p)).

Next, let us regard that n random caps are chosen after a random point .x is
chosen. Then we may regard .x as a fixed point, and .x ∈ V (n, p) is equivalent
to the event that the center of each random cap lies outside of the spherical
cap of area .4πp with center .x. Hence

. Pr(x ∈ V (n, p)) = (1 − p)n .

Therefore, .E(U (n, p)) = (1 − p)n . ⨆

⨅
.10.1
∗ Let .{P1' , P2' , P3' } ⊂ R3 be a trio that is isometric to .{P1 , P2 , P3 }. Then
every point at the same distance 3 from .P1' and .P2' lies on the plane E
that perpendicularly bisects the line segment .P1' P2' . Among the points on
E at distance 3 from .P1' , P2' , the one that is farthest from .P3' is at distance
√ √
. 52 − (2.5)2 + 32 − (2.5)2 ≈ 5.988 from .P3' . Since .5.988 < 6, there is
no point .P4 on E such that .P4' P3' = 6. Hence the quartet .{P1 , P2 , P3 , P4 } is
'

not isometrically embeddable in .R3 . ⨆

⨅
288 14 Solutions to the Selected Exercises

10.3∗ Let .y = cos AD. Since .X := {A, B, C, D} lies on the unit sphere .S, by
.

putting .x = 1 in the equation .MX (x) = 0, we have

| |
|1 0 0 y ||
|
|0 1 0 1/2||
.| = 0,
|0 0 1 1/2||
|
|y 1/2 1/2 1 |
√
from which we get .y 2 = 1/2. Hence .y = ±1/ 2, and .AD = π/4 or
.AD = 3π/4. Since ABCD is a convex quadrilateral, we have .AD = 3π/4.

⨆
⨅
.10.4
∗ The value of .s(X ) is

1
. (π/2 + 3π/4 + π/3) = 19π/24 ≈ 2.48709.
2
Hence .s(X )/π = 19/24 ≈ 0.791667. The equation .MX (x) = 0 is given by
| |
| 1 cos(π/2x) cos(π/2x) cos(3π/4x)||
|
| cos(π/2x) 1 cos(π/2x) cos(π/3x) ||
.| = 0.
| cos(π/2x) cos(π/2x) 1 cos(π/3x) ||
|
|cos(3π/4x) cos(π/3x) cos(π/3x) 1 |

By numerical computation, we have .r = 1 and .r ≈ 0.80608, as solutions of

det MX (x) = 0 that satisfy .r ≥ s(X )/π = 0.791667. Hence .r ≈ 0.80608.
.

⨆
⨅
.11.2
∗ From the approximated value of .V (n) for .n ≤ 7, we see that .V (5) is the

maximum in .{V (n) : 1 ≤ n ≤ 7}. Since .Γ(n/2 + 1) = (n/2)Γ(n/2), and

.2π/n < 1 for .n ≥ 7, we have

π n/2 / Γ(n/2 + 1) 2π
V (n)/V (n − 2) =
. = < 1.
π n/2−1 / Γ(n/2) n

Thus, .V (n) < V (n − 2) for .n ≥ 7, and hence .V (n) < V (5) follows for
n > 7.
. ⨅
⨆
11.3∗
.

⎰ ∞ ⎰ ∞
sΓ(s) = s
. zs−1 e−z dz = (zs )' e−z dz
0 0
⎰ ∞ ⎰ ∞
= [zs e−z ]∞
0 − zs (e−z )' dz = zs e−z dz = Γ(s + 1). ⨆
⨅
0 0
14 Solutions to the Selected Exercises 289

12.2∗
.

| |
| | ||−1 1 1 . . . 1 1||
|−1 1 1 . . . 1 || | |
| | 1 −1 1 . . . 1 1|
|1 −1 1 . . . 1 || | |
| | 1 1 −1 . . . 1 1|
| 1 −1 . . . 1 || = | . . |
.| 1
| . | . . .. |
| .. .. . . .. || | .. .. . . 1|
| . . . | | |
|1 | 1 1 1 . . . −1 1|
1 1 . . . −1| || |
0 0 0 . . . 0 1|
| | | |
|−2 0 0 . . . 0 1| |−2 0 0 . . . 0 |
| | | 1 |
| 0 −2 0 . . . 0 1| | 0 −2 0 . . . 0 |
| | | 1 |
| | | |
| 0 0 −2 . . . 0 1| | 0 0 −2 . . . 0 1 |
= || . . | |
. . .. | = | .. .. . . ..
|
|
.
| . . . . . 1| | . . . . 1 |
| | | |
| 0 0 0 . . . −2 1| | 0 0 0 . . . −2 1 |
| | | |
|−1 −1 −1 . . . −1 1| | 0 0 0 . . . 0 1 − k|
2

= (−2) (1 −
k k
2 ).

⨆
⨅
12.6∗ Put .x = d(x 2 , x 4 ). Since the Cayley-Menger determinant of .{x 1 , x 2 , x 3 , x 4 }
.

is zero, we have
| | | |
|0 1 9 4 1 | |0 1 4 1||
| | | 9
|1 0 9 x 2 1| |0 −1 8 x 2 − 1 1|
| | | |
| | | |
.0 = |9 9 0 9 1| = |0 0 −9 0 1|
| 2 | | 2 |
|4 x 9 0 1| |0 x − 4 5 −4 1|
| | | |
| 1 1 1 1 0 | |1 1 1 1 0|
| | | |
| 1 9 4 1|| || 1 9 4 1||
|
| −1 8 x 2 − 1 1| | −2 7 x 2 − 2 0||
= || |=|
| | |
| 0 −9 0 1| | −1 −10 −1 0|
|x 2 − 4 5 −4 1| |x 2 − 5 4 −5 0|
| |
|1 9 4 ||
|
= − ||−2 7 x 2 − 2|| = x 2 − 85.
|−1 −10 −1 |

√
Therefore, .x = 85. ⨆
⨅
.13.4
∗ Let .c , c , c , c be four circles intersecting at a point p. Then these four
1 2 3 4
circles have the same color, say red. Let us regard p as a colored point,
say a blue null circle. Then, by performing isometric radii change to
.{c1 , c2 , c3 , c4 , {p}}, we can decrease the radii of .ci , i = 1, 2, 3, 4, by some

small .ε, and we get circles .ci' , i = 1, 2, 3, 4. These circles are properly
290 14 Solutions to the Selected Exercises

tangent to a circle of radius .ε with center p. Hence, by Theorem 13.1 (Casey’s

theorem) we have .SD({c1' , c2' , c3' , c4' })) = 0. Since the isometric radii change
does not change the common tangent distance, we have .SD(F) = 0. ⨆
⨅
.13.5
∗ Let .z be the center of .c for .i = 1, 2, and let p be an intersection point
i i
of .c1 , c2 . Let .li be the tangent lines of .ci at p. Then the intersection angle
of .l1 , l2 is .θ . Since the line .zi p and .li cross orthogonally, it follows that
./ z1 pz2 = π − θ . Hence

|z1 z2 |2 = r12 + r22 − 2r1 r2 cos(π − θ ).

.

Therefore,

.t (c1 , c2 )2 = |z1 z2 |2 − (r1 − r2 )2 = −2r1 r2 cos(π − θ ) + 2r1 r2

= 2r1 r2 (1 + cos θ ).

Thus

t (c1 , c2 )2
. = 2(1 + cos θ ).
r1 r2
⨆
⨅
Bibliography

1. Abresch, U.: Lower curvature bounds, Toponogov’s theorem, and bounded topology. Ann.
Sci. Norm. Supér. (4) 18(4), 651–670 (1985)
2. Abresch, U.: Lower curvature bounds: Toponogov’s theorem, and bounded topology. Ann.
Sci. Norm. Supér. (4) 20(3), 475–502 (1987)
3. Abrosimov, N.V., Aseev, V.V.: Generalizations of Casey’s theorem for higher dimensions.
Lobachevskii J. Math. 39, 1–12 (2018)
4. Abrosimov, N.V., Mikaiylova, L.A.: Casey’s theorem in hyperbolic geometry. Sib. Elektron.
Mat. Izv. 12, 354–360 (2015)
5. Abrosimov, N., Mednykh, A.: Area and volume in non-Eucklidean geometry. In: Eighteen
Essays in non-Euclidean Geometry. IRMA Lectures in Mathematics and Theoretical Physics,
vol. 29, pp. 151–189. European Mathematical Society (EMS), Zürich (2019)
6. Adachi, T.: A comparison theorem on crescents for Kähler magnetic fields. Tokyo J. Math.
28(1), 289–298 (2005)
7. Aeppli, A.: Eine Verallgemeinerung einer Formel von Descartes. Elem. Math. 15, 9–13 (1960)
8. Agahi, H.: Multivariate Chebyshev inequality in generalized measure space based on Choquet
integral. Commun. Stat. Theory Methods 46(19), 9625–9628 (2017)
9. Agahi, H., Mesiar, R., Ouyang, Y.: Chebyshev type inequalities for pseudo-integrals. Nonlin-
ear Anal. 72(6), 2737–2743 (2010)
10. Agricola, I., Friedrich, T.: Elementary Geometry. Specialized Knowledge for Study and
Mathematics Education, 2nd edn. Vieweg, Wiesbaden (2009)
11. Aigner, M., Ziegler, G.M.: Proofs from The Book, 6th edn. (first edition: 1998). Including
illustrations by Karl H. Hofmann. Springer, Berlin (2018)
12. Akama, Y., Wang, E., Yan, M.: Tilings of the sphere by congruent pentagons III: edge
combination .a5 . Adv. Math. 394, Paper No. 107881, 41 pp. (2022)
13. Akopyan, A.: Geometry in Figures. A. Akopyan, Moscow (2011)
14. Al-Afifi, G., Hajja, M., Hamdan, A.: Another n-dimensional generalization of Pompeiu’s
theorem. Am. Math. Mon. 125(7), 612–622 (2018)
15. Aleksandrov, A.D.: A theorem on triangles in a metric space and some of its applications
(Russian). Trudy Mat. Inst. Steklov 38, 5–23 (1951)
16. Aleksandrov, A.D.: Instrinsic Geometry of Convex Surfaces (Russian edition). OGIZ,
Moscow-Leningrad, 1948. German edition: Akademie-Verlag, Berlin (1955). English edition
in: S. S. Kutateladze (Ed.): A. D. Alexandrov – Selected Works. Part II. Chapman &
Hall/CRC, Boca Raton (2006)

© The Author(s), under exclusive license to Springer Nature Switzerland AG 2024 291
H. Maehara, H. Martini, Circles, Spheres and Spherical Geometry,
Birkhäuser Advanced Texts Basler Lehrbücher,
https://doi.org/10.1007/978-3-031-62776-7
292 Bibliography

17. Aleksandrov, P.S., Markushevich, A.I., Chintshin, A.J. (Eds.): Enzyklopädie der Elemen-
tarmathematik, Band IV, Geometrie. Deutscher Verlag der Wissenschaften, Berlin (1969).
Russian original: Moscow, 1963
18. Alencar, J., Lavor, C., Liberti, L.: Realizing Euclidean distance matrices by sphere intersec-
tion. Discrete Appl. Math. 256, 5–10 (2019)
19. Alexander, R.: Two notes on metric geometry. Proc. Am. Math. Soc. 64(2), 317–320 (1977)
20. Alfakih, A.: Euclidean Distance Matrices and their Applications in Rigidity Theory. Springer,
Cham (2018)
21. Alfakih, A.Y.: On the Colin de Verdière graph number and penny graphs. Preprint,
arXiv:2006.05197
22. Almohammad, S.M., Lángi, Z., Naszódi, M.: An analogue of a theorem of Steinitz for ball
polyhedra. Aequationes Math. 96(2), 403–415 (2022)
23. Alonso, J., Martini, H., Spirova, M.: Minimal enclosing discs, circumcircles, and circumcen-
ters in normed planes (Part I). Comput. Geom. 45(5–6), 258–274 (2012)
24. Alsina, C., Nelsen, R.: When Less is More: Visualizing Basic Inequalities. Dolciani Mathe-
matical Expositions, vol. 36. Mathematical Association of America, Washington, DC (2009)
25. Alsina, C., Nelsen, R.B.: A Cornucopia of Quadrilaterals. AMS/MAA Dolciani Mathematical
Expositions, vol. 55. MAA Press/American Mathematical Society, Providence (2020)
26. Alsina, C., Nelsen, R.B.: A Panoply of Polygons. AMS/MAA Dolciani Mathematical
Expositions, vol. 58. MAA Press/American Mathematical Society, Providence (2023)
27. Alsina, C., Garcia-Roig, J.-L., Ger, R.: How to keep the quadrilateral inscribed. Abh. Math.
Sem. Univ. Hamburg 65, 123–133 (1995)
28. Altshiller-Court, N.: Modern Pure Solid Geometry, 2nd edn. Chelsea Publishing Co., New
York (1964). First edition: Macmillan Company, 1935
29. Altshiller-Court, N.: College Geometry. Reprint of the second (1980) edition. Dover Publica-
tions, Inc., Mineola (2007). First edition: Barnes & Noble, Inc., New York, 1952
30. Alzer, H.: Inequalities for the volume of the unit ball in .R n . J. Math. Anal. Appl. 252(1),
353–363 (2000)
31. Alzer, H.: Inequalities for the volume of the unit ball in .R n . II. Mediterr. J. Math. 5(4), 395–
413 (2008)
32. Anderson, G.D., Vuorinen, M., Zhang, X.: Topics in special functions III. In: Analytic
Number Theory, Approximation Theory, and Special Functions, pp. 297–345. Springer, New
York (2014)
33. Anstreicher, K.M.: The thirteen spheres: a new proof. Discrete Comput. Geom. 31(4), 613–
625 (2004)
34. Aomoto, K., Machida, Y.: Generalization of Schläfli formula to the volume of a spherically
faced simplex. J. Math. Soc. Jpn. 72(1), 213–249 (2020)
35. Aomoto, K., Machida, Y.: Hypergeometric integrals associated with hypersphere arrange-
ments and Cayley-Menger determinants. Hokkaido Math. J. 49(1), 1–85 (2020)
36. Apostol, T.M., Mnatsakanian, M.: New balancing principles applied to circumsolids of
revolution, and to n-dimensional spheres, cylindroids, and cylindrical wedges. Am. Math.
Mon. 120(4), 298–321 (2013)
37. Apostol, T.M., Mnatsakanian, M.: Volume/surface area ratios for globes, with applications.
Math. Mag. 88(5), 349–359 (2015)
38. Arbeiter, E., M. Zähle: Geometric measures for random mosaics in spherical spaces.
Stochastics Rep. 46(1–2), 63–77 (1994)
39. Archimedes: The two books on the sphere and the cylinder. In: The Works of Archimedes,
vol. 1. Cambridge University Press, Cambridge (2004). Translated into English, together with
Eutocius’ commentaries, with commentary and critical edition of the diagrams by R. Netz
40. Armitage, D.H.: The Pompeiu property for spherical polygons. Proc. R. Irish Acad. Sect. A
96(1), 25–32 (1996)
41. Artin, E.: The Gamma Function. Courier Dover Publications, New York (1964). Translated
by M. Butler, Holt, Rinehart and Winston
Bibliography 293

42. Askey, R.: Some elementary uses of complex numbers. Comput. Methods Funct. Theory
10(2), 579–584 (2010)
43. Askey, R.: Completing Brahmagupta’s extension of Ptolemy’s theorem. In: The Legacy of
Alladi Ramakrishnan in the Mathematical Sciences, pp. 191–197. Springer, New York (2010)
44. Asplund, E., B. Grünbaum: On the geometry of Minkowski planes. Enseign. Math. (2) 6,
299–306 (1960)
45. Athen, H.: Vektorrechnung auf der Kugelfläche. Math.-Phys. Semesterber. 4 , 90–100 (1954)
46. Atzema, E.J.: “A most elegant property”: on the early history of Lexell’s theorem. In:
Research in History and Philosophy of Mathematics. Proceedings of the The Canadian
Society for the History and Philosophy of Mathematics, pp. 117–132. Birkhuser, Basel (2017)
47. Aumann, G.: Archimedes–Mathematik in bewegten Zeiten. Wissenschaftliche Buchge-
sellschaft, Darmstadt (2013)
48. Aumann, G.: Kreisgeometrie. Springer, Spektrum, Berlin und Heidelberg (2015)
49. Avelino, C.P., dos Santos, A.M.F.: Spherical folding tessellations by kites and isosceles
triangles: a case of adjacency. Math. Commun. 19(1), 1–28 (2014)
50. Axler, S.: Measure, Integration & Real Analysis. Graduate Texts in Mathematics, vol. 282.
Springer, Cham (2020)
51. Bachoc, C., Vallentin, F.: New upper bounds for kissing numbers from semidefinite program-
ming. J. Am. Math. Soc. 21(3), 909–924 (2008)
52. Bakelman, I.J.: Spiegelung am Kreis. Teubner, Leipzig (1976). Original version in Russian:
Moscow, 1966
53. Balestro, V., Martini, H., Teixeira, R.: Convexity From the Geometric Point of View.
Birkhäuser, Cham (2024)
54. Balk, M.B., Boltyanskii, V.G.: Geometry of Masses (Russian). Bibl. Kvant, Vyp., vol. 61.
Nauka, Moskva (1987)
55. Ball, K.: An elementary introduction to modern convex geometry. In: Flavors of Geometry.
Mathematical Sciences Research Institute Publications, vol. 31, pp. 1–58. Cambridge Univer-
sity Press, Cambridge (1997)
56. Baralić, D., Jokanović, D.S., Milićević, M.: Variations on Steiner’s porism. Math. Intell.
39(1), 6–11 (2017)
57. Baragar, A., Kontorovich, A.: Efficiently constructing tangent circles. Math. Mag. 93(1), 27–
32 (2020)
58. Bárány, I.: Random points and lattice points in convex bodies. Bull. Am. Math. Soc. (N.S.)
45(3), 339–365 (2008)
59. Bárány, I., Hug, D., Reitzner, M., Schneider, R.: Random points in halfspheres. Random
Struct. Algor. 50(1), 3–22 (2017)
60. Barnes, J.: Gems of Geometry, first edition: Springer, 2009. Springer-Verlag, Heidelberg
(2012)
61. Barnette, D.W.: Map Coloring, Polyhedra, and the Four Color Theorem. Dolciani Mathemat-
ical Expositions, vol. 8. Mathematical Association America, Washington, DC (1983)
62. Barnette, D.W.: An impediment to polyhedrality. J. Combin. Theory Ser. A 48(2), 259–265
(1988)
63. Barrett, A.N., Mackay, A.L.: Spatial Structure and the Microcomputer. Selected Mathematical
Techniques. Macmillan Computer Science Series. Macmillan Education Ltd., Basingstoke
(1987)
64. Bartha, F.A., Fodor, F., González Merino, B.: Central diagonal sections of the n-cube. Int.
Math. Res. Not. IMRN (4), 2861–2881 (2021)
65. Barthe, F.: An extremal property of the mean width of the simplex. Math. Ann. 310(4), 685–
693 (1998)
66. Barthe, F., Koldobsky, A.: Extremal slabs in the cube and the Laplace transform. Adv. Math.
174(1), 89–114 (2003)
67. Baston, V.J.D.: Some Properties of Polyhedra in Euclidean Space. Pergamon Press, Oxford
(1965)
294 Bibliography

68. Bayer, M.M., Lee, C.W.: Combinatorial aspects of convex polytopes. In: Handbook of Convex
Geometry, vol. A, pp. 485–534. North-Holland, Amsterdam (1993)
69. Bédaride, N., Rao, M.: Regular simplices and periodic billiard orbits. Proc. Am. Math. Soc.
142(10), 3511–3519 (2014)
70. Beecroft, P.: Properties of circles in mutual contact. The Lady’s Gentleman’s Diary 139, 91–
96 (1842)
71. Belotti, M., Joswig, M., Panizzut, M.: Algebraic degrees of 3-dimensional polytopes. Vietnam
J. Math. 50(3), 581–597 (2022)
72. Benson, R.V.: Euclidean Geometry and Convexity. McGraw-Hill Book Co., New York-
Toronto-London (1966)
73. Benz, W.: Möbius sphere geometry in inner product spaces. Aequationes Math. 66(3), 284–
320 (2003)
74. Berger, A., Grigoriev, A., Panin, A., Winokurow, A.: Location, pricing and the problem of
Apollonius. Optim. Lett. 11(8), 1797–1805 (2017)
75. Berger, M.: Geometry I and II. Springer, Berlin and Heidelberg (1987)
76. Berger, M.: Circle packings (French, French summary), In: Differential geometry and
topology (Alghero, 1992), pp. 23–64. World Scientific Publishing, River Edge (1993)
77. Berger, M.: Riemannian geometry during the second half of the twentieth century. Jahres-
bericht der Deutschen Mathematiker-Vereinigung 100, 45–208 (1998). Also: University
Lecture Series, 17. American Mathematical Society, Providence, RI, 2000
78. Berger, M.: Geometry Revealed. A Jacob’s Ladder to Modern Higher Geometry. Springer,
Heidelberg (2010)
79. Berggren, J.L.: Episodes in the Mathematics of Medieval Islam, 2nd edn. Springer, New York
(2016). First edition: Springer, 1986
80. Bergold, H.: Eine Verallgemeinerung des Satzes von Pohlke. Elem. Math. 69(2), 57–60 (2014)
81. Besau, F., Thäle, C.: Asymptotic normality for random polytopes in non-Euclidean geome-
tries. Trans. Am. Math. Soc. 373(12), 8911–8941 (2020)
82. Besau, F., Rosen, D.A., Thäle, C.: Random inscribed polytopes in projective geometries.
Math. Ann. 381(3–4), 1345–1372 (2021)
83. Beskin, N.M., Boltyanski, V.G., et al.: Allgemeine Prinzipien geometrischer Konstruktionen.
In: Enzyklopädie der Elementarmathematik, Band IV, pp. 153–201, Deutscher Verlag der
Wissenschaften, Berlin (1969). Russian original: Moscow, 1963
84. Bezdek, K.: On a stronger form of Rogers’ lemma and the minimum surface area of Voronoi
cells in unit ball packings. J. Reine Angew. Math. 518, 131–143 (2000)
85. Bezdek, K.: Sphere packings revisited. Eur. J. Combin. 27(6), 864–883 (2006)
86. Bhattacharya, A., Dubey, K.K., Mondal, B.: Volume of an n-dimensional polyhedron:
revisited. J. Geom. Graph. 27(1), 1–9 (2023)
87. Bigalke, H.-G.: Kugelgeometrie. Otto Salle Verlag, Frankfurt am (1984)
88. Bilet, V., Dovgoshey, O.: Isometric embeddings of pretangent spaces in .E n . Bull. Belg. Math.
Soc. Simon Stevin 20(1), 91–110 (2013)
89. Bilinski, S.: Über Ptolemäische Sätze. Monatsh. Math. 77, 193–205 (1973)
90. Björnbo, A.A.: Studien über Menelaos’ Sphärik. In: Beiträge zur Geschichte der Sphärik und
Trigonometrie der Griechen. Abh. Geschichte Math. 14, VII, 1–154 (1902)
91. Blair, D.E.: Inversion Theory and Conformal Mapping. Student Mathematical Library, vol. 9.
American Mathematical Society, Providence (2000)
92. Blair, D.E., Wilker, J.B.: When does inversion preserve convexity? Kodai Math. J. 6(2), 186–
192 (1983)
93. Blaschke, W.: Über Integralgeometrie. Jahresber. Dtsch. Math.-Ver. 46, 139–152 (1936)
94. Blaschke, W.: Analytische Geometrie. Birkhäuser, Basel-Stuttgart (1954)
95. Blaschke, W.: Integralgeometrie, first edn. 1933. Deutscher Verlag der Wissenschaften, Berlin
(1955)
96. Blumenthal, L.M.: Theory and Applications of Distance Geometry. Clarendon Press, Oxford
(1953)
Bibliography 295

97. Blumenthal, L.M., Gillam, B.E.: Distribution of points in n-space. Am. Math. Mon. 50, 181–
185 (1943)
98. Blumenthal, L.M., Sen, H.: Sine laws for Euclidean simplices. Math. Student 41(3–4), 361–
369 (1973/1974)
99. Bokowski, J.: Oriented matroids. In: Handbook of Convex Geometry, vol. A, pp. 555–602.
North-Holland, Amsterdam (1993)
100. Boltyanski, V., Yaglom, I.M.: Vektoren und ihre Anwendungen in der Geometrie. In: Enzyk-
lopädie der Elementarmathematik IV, pp. 295–390, Deutscher Verlag der Wissenschaften,
Berlin (1969). First edition in Russian: Moscow, 1963
101. Boltyanski, V., Yaglom, I.M.: Konvexe Figuren und Körper. In: Enzyklopädie der Elemen-
tarmathematik V, pp. 171–257. Deutscher Verlag der Wissenschaften, Berlin (1971). First
editionnin Russian: Moscow, 1966
102. Bonnet, G., O’Reilly, E.: Facets of spherical random polytopes. Math. Nachr. 295(10), 1901–
1933 (2022)
103. Borceux, F.: An Axiomatic Approach to Geometry (Geometric Trilogy I). Springer, Cham
(2014)
104. Boreland, B., Kunze, H.: Circle inversion fractals. In: Mathematical and Computational
Approaches in Advancing Modern Science and Engineering, pp. 609–619. Springer, Cham
(2016)
105. Borsuk, K.: Analytic Geometry in n Dimensions. Mathematical Monographs, Tom XII.
Publisher Unknown, Warszawa-Wroclaw (1950)
106. Böröczky, K.: The Newton-Gregory problem revisited. Discrete Geometry. Monographs and
Surveys in Pure and Applied Mathematics, vol. 253, pp. 103–110. Marcel Dekker, New York
(2003)
107. Böröczky, K.: Finite Packing and Covering. Cambridge Tracts in Mathematics, vol. 154.
Cambridge University Press, Cambridge (2004)
108. Böröczky, K.J., Henk, M.: Random projections of regular polytopes. Arch. Math. (Basel)
73(6), 465–473 (1999)
109. Böröczky, K.J., Hug, D.: Isotropic measures and stronger forms of the reverse isoperimetric
inequality. Trans. Am. Math. Soc. 369(10), 6987–7019 (2017)
110. Böröczky, K., Schneider, R.: Circumscribed simplices of minimal mean width. Beitr. Algebra
Geom. 48(1), 217–224 (2007)
111. Böröczky, K., Szabó, L.: Arrangements of 13 points on a sphere. In: Discrete Geometry.
Monographs and Surveys in Pure and Applied Mathematics, vol. 253, pp. 111–184. Dekker,
New York (2003)
112. Böröczky, K., Szabó, L.: 12-neighbour packings of unit balls in 3-space. Acta Math. Hungar.
146(2), 421–448 (2015)
113. Bos, H.J.M.: Lectures in the History of Mathematics. History of Mathematics, vol. 7.
American Mathematical Society/London Mathematical Society, Providence/London (1993)
114. Bos, H.J.M.: Princess Elizabeth of Bohemia and Descartes’ letters (1650–1665). Hist. Math.
37(3), 485–502 (2010)
115. Bos, H.J.M., Kers, C., Oort, F., Raven, D.W.: Poncelet’s closure theorem. Exp. Math. 5(4),
289–364 (1987)
116. Bottema, O.: What inequalities are satisfied by the edges of a tetrahedron? (Dutch) Nieuw
Tijdschr. Wisk. 48, 166–170 (1960/1961)
117. Bottema, O.: Topics in Elementary Geometry, 2nd edn. Springer, New York (2008). Trans-
lated from the first Dutch edition, 1987
118. Bourgain, J., Fuchs, E.: A proof of the positive density conjecture for integer Apollonian
circle packings. J. Am. Math. Soc. 24(4), 945–967 (2011)
119. Bowman, F.: Cyclic pentagons. Math. Gaz. 36, 244–250 (1952)
120. Boyvalenkov, P., Dodunekov, S., Musin, O.: A survey on the kissing numbers. Serdica Math.
J. 38(4), 507–522 (2012)
121. Bradford, A.: An even more straightforward proof of Descartes’ circle theorem. Math. Intell.
45(3), 263–265 (2023)
296 Bibliography

122. Brakensiek, J., Heule, M., Mackey, J., Narváez, D.: The resolution of Keller’s conjecture. J.
Automat. Reason. 66(3), 277–300 (2022)
123. Brandstädt, A., Le, V.B., Spinrad, J.P.: Graph Classes: A Survey. SIAM Monographs on
Discrete Mathematics and Applications. Society for Industrial and Applied Mathematics
(SIAM), Philadelphia (1999)
124. Brannan, D.A., Esplen, M.F., Gray, J.J.: Geometry. Cambridge University Press, Cambridge
(1999)
125. Brass, P., Moser, W., Pach, J.: Research Problems in Discrete Geometry. Springer, New York
(2005)
126. Brauchart, J.S., Reznikov, A.B., Saff, E.B., Sloan, I.H., Wang, Y.G., Womersley, R.S.:
Random point sets on the sphere – hole radii, covering, and separation. Exp. Math. 27(1),
62–81 (2018)
127. Breda, A.M.R., Dos Santos, J.M.: Spherical geometry and spherical tilings with GeoGebra. J.
Geom. Graph. 22(2), 283–299 (2018)
128. Brightwell, G.R., Scheinerman, E.R.: Representations of planar graphs. SIAM J. Discrete
Math. 6, 214–229 (1993)
129. Brooks, J., Strantzen, J.: Spherical triangles of area .π and isosceles tetraheda. Math. Mag. 78,
311–314 (2005)
130. Bruen, A.A., Fisher, J.C., Wilker, J.B.: Apollonius by inversion. Math. Mag. 56(2), 97–103
(1983)
131. van Brummelen, G.: Heavenly Mathematics. The Forgotten Art of Spherical Trigonometry.
Princeton University Press, Princeton (2013)
132. Budny, K.: A generalization of Chebyshev’s inequality for Hilbert-space-valued random
elements. Statist. Probab. Lett. 88, 62–65 (2014)
133. Burago, Y., Gromov, M., Perel’man, G.Y.: A. D. Aleksandrov spaces with curvatures bounded
below. Uspekhi Mat. Nauk 47(2)(284), 3–51, 222 (1992). Translation: Russian Math. Surveys
47 (1992), no. 2, 1–58
134. Burago, D., Burago, Y., Ivanov, S.: A Course in Metric Geometry. American Mathematical
Society, Providence (2001)
135. Burkholder, E.F.: The 3-D Global Spatial Data Model. Foundation of the Spatial Data
Infrastructure. CRC Press, Boca Raton (2008)
136. Busemann, H.: Geometry of Geodesics. Pure and Applied Mathematics, vol. 6. Academic
Press, New York (1955)
137. Byer, O.D., Smeltzer, D.L.: A 3-D analog of Steiner’s Porism. Math. Mag. 87(2), 95–99
(2014)
138. Caddeo, R., Papadopoulos, A. (Eds.): Mathematical Geography in the eighteenth century:
Euler, Lagrange and Lambert. Springer, Cham (2022)
139. Calinger, R.: A comet of the enlightenment: Anders Johan Lexell’s life and discoveries.Math.
Intell. 39(3), 92–93 (2017)
140. Campos, M., Jenssen, M., Michelen, M., Sahasrabudhe, J.: A new lower bound for sphere
packing. In: arXiv:2312.10026v1
141. Carlitz, L.: Some inequalities related to Euler’s theorem .R ≥ 2r. Publ. Inst. Math. (Beograd)
(N.S.) 12(26), 11–17 (1971)
142. Casey, J.: On the equations and properties: (1) of the system of circles touching three circles
in a plane; (2) of the system of spheres touching four spheres in space; (3) of the system of
circles touching three circles on a sphere; (4) of the system of conics inscribed to a conic,
and touching three inscribed conics in a plane. Proc. R. Irish Acad. (1836–1869) 9, 396–423
(1864)
143. Casey, J.: A Sequel to the First Six Books of the Elements of Euclid Containing an Easy
Introduction to the Modern Geometry. With Numerous Examples. Dublin University Press,
Dublin (1881). JFM 13.0436.02
144. Casey, J.: A Treatise on Spherical Trigonometry, and Its Application to Geodesy and
Astronomy with Numerous Examples. Hodges, Figgis, & Co., Dublin (1889)
Bibliography 297

145. Casselman, B.: The difficulties of kissing in three dimensions. Notices Am. Math. Soc. 51(8),
884–885 (2004)
146. Catalan, E.C.: Éléments de Géométrie. Bachilier, Paris (1847)
147. Cayley, A.: On a theorem in the geometry of position. Cambridge Math. J. 2, 267–271 (1841)
148. Cecil, T.E., Ryan, P.J.: Tight and Taut Immersions of Manifolds. Research Notes in Mathe-
matics, vol. 107. Pitman, London (1985)
149. Cecil, T.E., Ryan, P.J.: Gemetry of Hypersurfaces. Springer, New York (2015)
150. Cerioli, M.R., Faria, L., Ferreira, T.O., Protti, F.: A note on maximum independent sets
and minimum clique partitions in unit disk graphs and penny graphs: complexity and
approximation. RAIRO Theor. Inform. Appl. 45(3), 331–346 (2011)
151. Chakerian, G.D.: The mean volume of boxes and cylinders circumscribed about a convex
body. Isr. J. Math. 12, 249–256 (1972)
152. Chakerian, G.D., Filliman, P.: The measures of the projections of a cube. Stud. Sci. Math.
Hungar. 21(1–2), 103–110 (1986)
153. Chakerian, G.D., Groemer, H.: Bodies of constant width. In: Convexity and Its Applications,
pp. 49–96. Birkhäuser, Basel-Boston-Stuttgart (1983)
154. Chakerian, G.D., Logothetti, D.: Cube slices, pictorial triangles, and probability. Math. Mag.
64(4), 219–241 (1991)
155. Chakrabarty, I., Kureethara, J.V.: A survey on the intersection graphs of ideals of rings.
Commun. Comb. Optim. 7(2), 121–167 (2022)
156. Chapple, W.: An essay on the properties of triangles inscribed in and circumscribed about two
given circles. Miscellanea Curiosa Math. 4, 117–124 (1746)
157. Charitos, C.: Area preserving maps from the sphere to the Euclidean plane. In: Eighteen
Essays in non-Euclidean Geometry, pp. 135–150. IRMA Lectures in Mathematics and
Theoretical Physics, vol. 29. European Mathematical Society, Zürich (2019)
158. Chavel, I.: Riemannian Geometry–a Modern Introduction. Cambridge Tracts in Mathematics,
vol. 108. Cambridge University Press, Cambridge (1993)
159. Cheeger, J., Ebin, D.G.: Comparison Theorems in Riemannian Geometry. North-Holland
Mathematical Library, vol. 9. North-Holland Publishing Co./American Elsevier Publishing
Co. Inc., Amsterdam-Oxford/New York (1975)
160. Chen, C.-P., Lin, L.: Inequalities for the volume of the unit ball in .R n . Mediterr. J. Math.
11(2), 299–314 (2014)
161. Chern, S.-S.: On integral geometry in Klein spaces. Ann. Math. (2) 43, 178–189 (1942)
162. Chern, S.-S.: Differential Geometry & Integral Geometry –Selected Papers & Lectures of
Shiing-Shen Chern. In: Cheng, S.-Y., Ji, L. (Eds.) Surveys of Modern Mathematics, vol. 12.
International Press/Higher Education Press, Somerville/Beijing (2017)
163. Chilakamarri, K.B., Hamburger, P., Pippert, R.E.: Analysis of Venn diagrams using cycles in
graphs. Geom. Dedicata 82(1–3), 193–223 (2000)
164. Cieślak, W., Martini, H., Mozgawa, W.: On the rotation index of bar billiards and Poncelet’s
porism. Bull. Belg. Math. Soc. Simon Stevin 20(2), 287–300 (2013)
165. Clancy, C., Frame, M.: Fractal geometry of restricted sets of circle inversions. In: Symposium
in Honor of Benoit Mandelbrot (Curacao, 1995). Fractals 3(4), 689–699 (1995)
166. Clemens, C.H.: Two-Dimensional Geometries. A Problem-Solving Approach. Pure and
Applied Undergraduate Texts, vol. 34. American Mathematical Society, Providence (2019)
167. Cohn, H.: Packing, coding, and ground states. In: Mathematics and Materials. IAS/Park City
Mathematics Series, vol. 23, pp. 45–102. American Mathematical Society, Providence (2017)
168. Cohn, H.: A conceptual breakthrough in sphere packing. Not. Am. Math. Soc. 64(2), 102–115
(2017)
169. Cohn, H., Elkies, N.: New upper bounds on sphere packings. I. Ann. Math. (2) 157(2), 689–
714 (2003)
170. Cohn, H., Kumar, A., Miller, S.D., Radchenko, D., Viazovska, M.: The sphere packing
problem in dimension 24. Ann. Math. (2) 185(3), 1017–1033 (2017)
171. Colding, T.H.: Shape of manifolds with positive Ricci curvature. Invent. Math. 124(1–3),
175–191 (1996)
298 Bibliography

172. Colding, T.H.: Ricci curvature and volume convergence. Ann. Math. (2) 145(3), 477–501
(1997)
173. Colin de Verdière, Y.: On a new graph invariant and a planarity criterion (French). J. Comb.
Theory, Ser. B 50(1), 11–21 (1990)
174. Colins, K.D.: Cayley-Menger Determinant. From MathWorld–A Wolfram Web Resource,
created by Eric W. Weisstein. https://mathworld.wolfram.com/Cayley-MengerDeterminant.
html
175. Coll, V., Dodd, J., Harrison, M.: On the smoothness of the equizonal ovaloids. J. Geom.
103(3), 409–416 (2012)
176. Conway, J.H., Sloane, N.J.H.: Sphere Packings, Lattices and Groups. Grundlehren der
mathematischen Wissenschaften, vol. 290, 3rd edn. Springer-Verlag, New York (1999). First
edition in 1988. With additional contributions by E. Bannai, R. E. Borcherds, J. Leech, S. P.
Norton, A. M. Odlyzko, R. A. Parker, L. Queen and B. B. Venkov
177. Coolidge, J.L.: A Treatise on the Circle and the Sphere. Clarendon Press, Oxford (1916)
178. Couderc, P., Ballicioni, A.: Premier Livre du Tétraedre. Gauthier-Villars, Paris (1953)
179. Courant, R., Robbins, H.: Was ist Mathematik?, 4th edn. Springer-Verlag, Berlin (1992).
Original version in English: Oxford Univ. Press, New York, 1941. Translated from the English
original by Iris Runge. With a foreword by S. Hildebrandt
180. Court, N.A.: Modern Pure Solid Geometry. Macmillan, New York (1932)
181. Coxeter, H.S.M.: Interlocked rings of spheres. Scripta Math. 18, 113–121 (1952)
182. Coxeter, H.S.M.: Introduction to Geometry. John Wiley & Sons, Inc., New York-London
(1961). Reprinted in 1989
183. Coxeter, H.S.M.: Review of [248]. Item MR0083426 in Mathematical Reviews (1957)
184. Coxeter, H.S.M.: An upper bound for the number of equal nonoverlapping spheres that can
touch another of the same size. In: Proceedings of Symposia in Pure Mathematics, vol. VII,
pp. 53–71. American Mathematical Society, Providence (1963)
185. Coxeter, H.S.M.: The problem of Apollonius. Am. Math. Mon. 75, 5–15 (1968)
186. Coxeter, H.S.M., Greitzer, S.L.: Geometry Revisited. New Mathematical Library, vol. 19.
Random House, Inc., New York (1967)
187. Cramer, H.: Mathematical Methods of Statistics. Princeton University Press, Princeton (1946)
188. Crawley, E.: One Thousand Exercises in Plane and Spherical Trigonometry. University of
Pennsylvania, Philadelphia (1914)
189. Crippen, G.M.: A novel approach to calculation of conformation: distance geometry. J.
Comput. Phys. 24(1), 96–107 (1977)
190. Croft, H.T., Falconer, K.J., Guy, R.K.: Unsolved Problems in Geometry. Springer, New York
(1991)
191. Crofton, M.W.: On the theory of local probability. Trans. R. Soc. 158, 181 (1868)
192. Crofton, M.W.: Obituary. Proc. Lond. Math. Soc. 14, xxix–xxxx (1915)
193. Cundy, H.M., Rollett, A.P.: Mathematical Models. Clarendon Press, Oxford (1952)
194. Dai, X., Wei, G.: A comparison-estimate of Toponogov type for Ricci curvature. Math. Ann.
303(2), 297–306 (1995)
195. Dalbec, J.P.: Straightening Euclidean invariants. Ann. Math. Artif. Intell. 13(1–2), 97–108
(1995)
196. D’Andrea, K., Sombra, M.: The Cayley-Menger determinant is irreducible for .n ≥ 3
(Russian. Russian summary). Sibirsk. Mat. Zh. 46(1), 90–97 (2005). Translation in Siberian
Math. J. 46 (2005), no. 1, 71–76
197. Danzer, L., Grünbaum, B., Klee, V.: Helly’s theorem and its relatives. In: Proceedings of
Symposia in Pure Mathematics, vol. VII, pp. 101–180. American Mathematical Society,
Providence (1963)
198. Datta, B.B., Singh, A.N.: Hindu trigonometry. Revised by Kripa Shankar Shukla. Indian J.
Hist. Sci. 18(1), 38–108 (1983)
199. Davis, P.J., Hersh, R.: The Mathematical Experience. Birkhäuser, Boston (1980). Study
edition: 2012
Bibliography 299

200. Dawson, R.J.M.: Tilings of the sphere with isosceles triangles. Discrete Comput. Geom.
30(3), 467–487 (2003)
201. Dawson, R.J.M., Doyle, B.: Tilings of the sphere with right triangles. III. The asymptotically
obtuse families. Electron. J. Combin. 14(1), Research Paper 48, 9 pp. (2007)
202. Day, M.M.: On criteria of Kasahara and Blumenthal for inner-product spaces. Proc. Am.
Math. Soc. 10, 92–100 (1959)
203. Deák, I.: Comparison of methods for generating uniformly distributed random points in and
on a hypersphere (Russian summary). Problems Control Inform. Theory 8(2), 105–113 (1979)
204. Debarnot, M.-T.: Trigonometry. In: Encyclopedia of the History of Arabic Science, vol. 1–3,
pp. 495–538. Routledge, London (1996)
205. Dehbi, L., Zeng, Z., Yang, L.: The number of tetrahedra sharing the same metric invariants
via symbolic and numerical computations. J. Symbolic Comput. 108, 41–54 (2022)
206. Dekster, B.V.: Simplexes with prescribed edge lengths in Minkowski and Banach spaces. Acta
Math. Hungar. 86(4), 343–358 (2000)
207. Dekster, B.V., Wilker, J.B.: Edge lengths guaranteed to form a simplex. Arch. Math. 49, 351–
366 (1987)
208. Del Centina, A.: Poncelet’s porism: a long story of renewed discoveries, I. Arch. Hist. Exact
Sci. 70(1), 1–122 (2016)
209. Del Centina, A.: Poncelet’s porism: a long story of renewed discoveries, II. Arch. Hist. Exact
Sci. 70(2), 123–173 (2016)
210. De Loera, J., Rambau, J., Santos, F.: Triangulations. Structures for Algorithms and Applica-
tions. Algorithms and Computation in Mathematics, vol. 25. Springer-Verlag, Berlin (2010)
211. Delsarte, P., Goethals, J.M., Seidel, J.J.: Spherical codes and designs. Geom. Dedicata 6(3),
363–388 (1977)
212. Deltheil, R.: Probabilités géométriques. Gauthier-Villars, Paris (1926)
213. Dietrich, U., Girstmair, K.: John Napier’s Trigonometrie – ein Blick zurück. Math. Semester-
ber. 56(2), 215–232 (2009)
214. Dirnböck, H.: Absolute polarity on the sphere; conics, loxodrome, tractrix. Math. Commun.
4, 225–240 (1999)
215. Djorić, M., Vanhecke, L.: A theorem of Archimedes about spheres and cylinders and two-
point homogeneous spaces. Bull. Aust. Math. Soc. 43(2), 283–294 (1991)
216. Dodd, J., Coll, V.: Generalizing the equal zones property of the sphere. J. Geom. 90(1–2),
47–55 (2008)
217. Dörrie, H.: Ebene und sphärische Trigonometrie. Oldenbourg-Verlag, München (1950)
218. Dörrie, H.: 100 Great Problems of Elementary Mathematics. Their History and Solution.
Dover, New York (1982). First edition in German: 1940
219. Dörrie, H.: The tangency problem of Apollonius. In: 100 Great Problems of Elementary
Mathematics: Their History and Solutions. Dover, New York (1965)
220. Donnay, J.D.H.: Spherical Trigonometry after the Cesàro Method. Interscience Publication,
New York (1945)
221. Dordovskii, D.: Metric tangent spaces to Euclidean spaces (Russian. English and Russian
summaries). Ukr. Mat. Visn. 8(2), 159–181, 315 (2011). Translation in J. Math. Sci. (N.Y.)
179 (2011), no. 2, 229–244.
222. Dostert, M., Kolpakov, A.: Kissing number in non-Euclidean spaces of constant sectional
curvature. Math. Comput. 90(331), 2507–2525 (2021)
223. Dostert, M., Kolpakov, A., de Oliveira Filho, F.M.: Semidefinite programming bounds for the
average kissing number. Isr. J. Math. 247(2), 635–659 (2022)
224. Dragović, V., Radnović, M.: Hyperelliptic Jacobians as billiard algebra of pencils of quadrics:
beyond Poncelet porisms. Adv. Math. 219(5), 1577–1607 (2008)
225. Dragović, V., Radnović, M.: Poncelet’s porisms and beyond. Integrable billiards, hyperelliptic
Jacobians and Pencils of Quadrics. Birkhäuser, Basel (2011)
226. Drechsler, K., Sterz, U.: Apollonius’ contact problem in n-space in view of enumerative
geometry. Acta Math. Univ. Comenian. (N.S.) 68(1), 37–47 (1999)
300 Bibliography

227. Dunajski, M.: Geometry. A Very Short Introduction. Very Short Introductions, vol. 695.
Oxford University Press, Oxford (2022)
228. Dunham, W. (Ed.): The Genius of Euler: Reflections on his Life and Work. Spectrum Series,
vol. 2. Mathematical Association of America, Washington, DC (2007)
229. Duparc, H.J.A.: Some applications of Casey’s theorem (Dutch). In: Handelingen van het
XXXIe Nederlands Natuur- en Geneeskundig Congres, pp. 85–87. Haarlem (1949)
230. Duparc, H.J.A.: On the s-formula (Dutch). Summer Course 1991: Geometrical Structures
(Dutch). CWI Syllabi, vol. 28, pp. 27–35. Mathematica Centrum. Wisk Information, Amster-
dam (1991)
231. Durán, G., Grippo, L., Safe, M.: Structural results on circular-arc graphs and circle graphs: a
survey and the main open problems. Discrete Appl. Math. 164, part 2, 427–443 (2014)
232. Du Val, P., Saban, G.: Archimedes in n dimensions. Math. Gaz. 58, 202–205 (1974)
233. Eckhoff, J.: Helly, Radon, and Carathéodory type theorems. In: Handbook of Convex
Geometry, vol. A, pp. 389–448. North-Holland, Amsterdam (1993)
234. Edmonds, A.L.: The geometry of an equifacetal simplex. Mathematika 52(1–2), 31–45
(2005/2006)
235. Edmonds, A.L.: The center conjecture for equifacetal simplices. Adv. Geom. 9(4), 563–576
(2009)
236. Edmonds, A.L.: The partition problem for equifacetal simplices. Beitr. Algebra Geom. 50(1),
193–213 (2009)
237. Edmonds, A.L., Hajja, M., Martini, H.: Orthocentric simplices and their centers. Results
Math. 47(3–4), 266–295 (2005)
238. Edmonds, A.L., Hajja, M., Martini, H.: Coincidences of simplex centers and related facial
structures. Beitr. Algebra Geom. 46(2), 491–512 (2005)
239. Edmonds, A.L., Hajja, M., Martini, H.: Orthocentric simplices and biregularity. Results Math.
52(1–2), 41–50 (2008)
240. Ehrenfeucht, A.: The Cube Made Interesting. A Pergamon Press Book, The Macmillan
Company, New York (1964). Translated from the Polish by Waclaw Zawadowski
241. Elerath, D.E.: An improved Toponogov comparison theorem for nonnegatively curved
manifolds. J. Differ. Geom. 15(2), 187–216 (1980/1981)
242. Ellis, D.C.: Intersection problems in extremal combinatorics: theorems, techniques and
questions old and new. In: Surveys in Combinatorics 2022. London Mathematical Society
Lecture Note series, vol. 481, pp. 115–173. Cambridge University Press, Cambridge (2022)
243. Epps, T.W.: Probability and Statistical Theory for Applied Researchers. World Scientific
Publishing Co. Pte. Ltd., Hackensack (2014)
244. Eppstein, D.: Edge bounds and degeneracy of triangle-free penny graphs and squaregraphs. J.
Graph Algorithms Appl. 22(3), 483–499 (2018)
245. Euclid: Euclid’s Elements of Geometry. All Thirteen Books Complete in One Volume. The
Heath Translation. Green Lion Press, Santa Fe (2002)
246. Euler, L.: On the geographical projection of the surface of the sphere (English summary). Acta
Academiae Scientiarum Imperialis Petropolitanae (1), 133–142 (1777/1778). Translation
of De proiectione geographica superficiei sphaericae. Opera Omnia: Series 1, Volume 28,
pp. 276–287
247. Euler, L.: Variae speculationes super area triangulorum sphaericorum. Nova Acta Acad. Sci.
Imperialis Petropolitanae 10, 47–62 (1797)
248. Euler, L.: Opera omnia. Series prima. In: Speiser, A. (Ed.) Opera Mathematica, vol. XXIX.
Commentationes geometricae, vol. quartum. Societas Scientiarum Naturalium Helveticae,
Lausanne (1956)
249. Eves, H.: Fundamentals of Geometry. Allyn and Bacon, Boston (1969)
250. Eves, H.: Fundamentals of Modern Elementary Geometry. Jones and Bartlett Publishers,
Boston (1992)
251. Ewald, G.: Geometry: An Introduction. Wadsworth Publ. Co., Belmont (1971)
Bibliography 301

252. Ewald, G., Kleinschmidt, P., Pachner, U., Schulz, C.: Neuere Entwicklungen in der kombi-
natorischen Konvexgeometrie. In: Contributions to Geometry (Proc. Geom. Sympos., Siegen,
1978), pp. 131–163. Birkhäuser Verlag, Basel/Boston (1979)
253. Ewald, G., Schulz, C.: Non-starshaped spheres. Arch. Math. (Basel) 59(4), 412–416 (1992)
254. Fabricius-Bjerre, F.: The theory of conic sections on the sphere (Danish). Mat. Tidsskr. A
53–71 (1945)
255. Feeman, T.G.: Portraits of the Earth. A Mathematician Looks at Maps. Mathematical World,
vol. 18. American Mathematical Society, Providence (2002)
256. Fejes Tóth, L.: Über die dichteste Kugellagerung. Math. Z. 48, 676–684 (1943)
257. Fejes Tóth, L.: Lagerungen in der Ebene, auf der Kugel und im Raum (second expanded
edition). Die Grundlehren der mathematischen Wissenschaften, Band, vol. 65. Springer-
Verlag, Berlin-New York (1972). First edition: 1953
258. Fejes Tóth, L.: Remarks on a theorem of R. M. Robinson. Stud. Sci. Math. Hungar. 4, 441–
445 (1969)
259. Fejes Tóth, L.: On spherical tilings generated by great circles. Geom. Dedicata 23(1), 67–71
(1987)
260. Fejes Tóth, L., Heppes, A.: A variant of the problem of the thirteen spheres. Can. J. Math. 19,
1092–1100 (1967)
261. Fejes Tóth, G., Kuperberg, W.: Packing and covering with convex sets. In: Handbook of
Convex Geometry, vol. B, pp. 799–860. North-Holland, Amsterdam (1993)
262. Fejes Tóth, L., Fejes Tóth, G., Kuperberg, W.: Lagerungen–Arrangements in the Plane, on
the Sphere, and in Space. Grundlehren der mathematischen Wissenschaften [Fundamental
Principles of Mathematical Sciences], vol. 360. Springer, Cham (2023). Translated from the
German [MR0057566]. With a foreword by Thomas Hales
263. Felsner, S.: Geometric Graphs and Arrangements. Vieweg, Wiesbaden (2004)
264. Fenn, R.: Geometry. Springer, London (2001)
265. Fiala, F.: Mathematische Kartographie. Verlag Technik, Berlin (1957)
266. Fiedler, M.: Matrices and Graphs in Geometry. Encyclopedia of Mathematics and its
Applications, vol. 139. Cambridge University Press, Cambridge (2011)
267. Fiedler, M.: Majorization in Euclidean geometry and beyond. Linear Algebra Appl. 466, 233–
240 (2015)
268. Filliman, P.: Extremum problems for zonotopes. Geom. Dedicata 27(3), 251–262 (1988)
269. Firey, W.J.: A note on a theorem of H. Knothe. Mich. Math. J. 6, 53–54 (1959)
270. Fitz-Gerald, J.M.: A note on a problem of Apollonius. J. Geom. 5, 15–26 (1974)
271. Flatto, L.: Poncelet’s Theorem. Chapter 15 by S. Tabachnikov. American Mathematical
Society, Providence (2009)
272. Florian, A.: Approximation of spherical caps by polygons. Beitr. Algebra Geom. 29, 171–182
(1989)
273. Fodor, F.: Perimeter approximation of convex discs in the hyperbolic plane and on the sphere.
Discrete Comput. Geom. 66(3), 1190–1201 (2021)
274. Fraivert, D., Sigler, A., Stupel, M.: Necessary and sufficient properties for a cyclic quadrilat-
eral. Int. J. Math. Educ. Sci. Technol. 51(6), 913–938 (2020)
275. Frank, R., Riede, H.: Hyperplane sections of the n-dimensional cube. Am. Math. Mon.
119(10), 868–872 (2012)
276. Frankl, P., Maehara, H.: Simplices with given 2-face areas. Eur. J. Combin. 11(3), 241–247
(1990)
277. Freeden, W., Schreiner, M.: Spherical Functions of Mathematical Geosciences – a Scalar,
Vectorial, and Tensorial Setup. Geosystems Mathematics. Birkhäuser/Springer, Berlin (2022)
278. Freese, R.W.: Criteria for inner product spaces. Proc. Am. Math. Soc. 19, 953–958 (1968)
279. Freese, R.W., Andalafte, E.Z.: Four-point characterizations of real inner product spaces. J.
Geom. 75(1–2), 97–105 (2002)
280. Freese, R.W., Andalafte, E.Z., Diminnie, C.: New four point properties which characterize
inner product spaces. Math. Jpn. 27(2), 253–261 (1982)
302 Bibliography

281. Frenkel, E., Su, W.: The area formula for hyperbolic triangles. In: Eighteen Essays in Non-
Euclidean Geometry, pp. 27–46. IRMA Lectures in Mathematics and Theoretical Physics,
vol. 29. European Mathematical Society (EMS), Zürich (2019)
282. Fuchs, E.: Counting problems in Apollonian packings. Bull. Am. Math. Soc. New Ser. 50(2),
229–266 (2013)
283. Fuchs, D., Tabachnikov, S.: Mathematical Omnibus. Thirty Lectures on Classic Mathematics.
American Mathematical Society, Providence (2007)
284. Fukagawa, H., Pedoe, D.: Japanese Temple Geometry Problems. Sangaku. Devotional
Mathematical Tablets. Charles Babbage Research Centre, Winnipeg (1989)
285. Fukagawa, H., Rigby, J.F.: Traditional Japanese Mathematics Problems of the 18th and
19th Centuries. Science Culture Technology Publishing (SCT), Singapore (2002). With an
introductory note by Dan Pedoe
286. Fukagawa, H., Rothman, T.: Sacred Mathematics: Japanese Temple Geometry. Princeton
University Press, Princeton (2008). With a preface by Freeman Dyson
287. Gaifullin, A.A.: Volume of a simplex as a multivalued algebraic function of the areas
of its two-faces. In: Topology, Geometry, Integrable Systems, and Mathematical Physics.
American Mathematical Society Translations: Series 2, vol. 234, pp. 201–221. Advances in
Mathematics: Scientific, vol. 67. American Mathematical Society, Providence (2014)
288. Gale, D.: Linear programming and the simplex method. Notices Am. Math. Soc. 54(3), 364–
369 (2007)
289. Galicer, D., Merzbacher, M., Pinasco, D.: The minimal volume of simplices containing a
convex body. J. Geom. Anal. 29(1), 717–732 (2019)
290. Gao, F., Hug, D., Schneider, R.: Intrinsic volumes and polar sets in spherical space. Math.
Notae 41, 159–176 (2003)
291. Gao, H., Shi, N., Yan, M.: Spherical tiling by 12 congruent pentagons. J. Combin. Theory Ser.
A 120(4), 744–776 (2013)
292. Garcia, R., Reznik, D.: A matryoshka of Brocard porisms. Eur. J. Math. 8, Suppl. 1, S308–
S329 (2022)
293. Garcia, R., Odehnal, B., Reznik, D.: Loci of Poncelet triangles in the general closure case. J.
Geom. 113(1), Paper No. 17, 30 pp. (2022)
294. Gardner, M.: Mathematical Circus. MAA Spectrum, Washington, DC (1992)
295. Gardner, R.J.: Geometric Tomography, 2nd edn. Encyclopedia of Mathematics and its
Applications, vol. 58. Cambridge University Press, New York (2006)
296. Gdawiec, K.: Star-shaped set inversion fractals. Fractals 22(4) 1450009, 7 pp. (2014)
297. Gdowski, B.: Elements of Differential Geometry with Exercises (Polish). Państwowe
Wydawnictwo Naukowe, Warsaw (1982)
298. Geiges, H.: The Geometry of Celestial Mechanics. London Mathematical Society Student
Texts, vol. 83. Cambridge University Press, Cambridge (2016)
299. Giering, O.: Vorlesungen über höhere Geometrie. Vieweg, Braunschweig (1982)
300. Giering, O.: Perlen der Geometrie. Mitt. Dtsch. Math.-Ver. (1), 15–21 (1998)
301. Gipple, J.: The volume of n-balls. Rose-Hulman Undergrad. Math. J. 15, 237–248 (2014)
302. Glaeser, G.: Geometry and its Applications in Arts, Nature and Technology, 2nd edn.
Springer, Cham (2020)
303. Glasauer, S.: Integral geometry of spherically convex bodies. Diss. Summ. Math. 1(1–2),
219–226 (1996)
304. Glazyrin, A.: A short solution of the kissing number problem in dimension three. Discrete
Comput. Geom. 69(3), 931–935 (2023)
305. Goldberg, M.: Rotors in spherical polygons. J. Math. Phys. 30, 235–244 (1952)
306. Goldberg, M.: Basic rotors in spherical polygons. J. Math. Phys. 34, 322–327 (1956)
307. Goldberg, M.: The Peaucellier linkage on the surface of a sphere. Math. Mag. 38, 308–311
(1965)
308. Golumbic, M.C.: Algorithmic Graph Theory and Perfect Graphs, 2nd edn. Annals of Discrete
Mathematics, vol. 57. Elsevier Science B.V., Amsterdam (2004). First edition: 1980. With a
foreword by Claude Berge
Bibliography 303

309. González-Velasco, E.: Journey through Mathematics. Creative Episodes in Its History.
Springer, New York (2011)
310. Goodey, P.R., Kiderlen, M., Weil, W.: Section and projection means of convex bodies.
Monatsh. Math. 126(1), 37–54 (1998)
311. Goodey, P.R., Hug, D., Weil, W.: Kinematic formulas for area measures. Indiana Univ. Math.
J. 66(3), 997–1018 (2017)
312. Gower, J.C.: Euclidean distance geometry. Math. Sci. 7(1), 1–14 (1982)
313. Gowers, T.: Mathematics: A Very Short Introduction. Oxford University Press, Oxford (2002)
314. Grace, J.H.: Tetrahedra in relation to spheres and quadrics. Proc. Lond. Math. Soc. (2) 17,
259–271 (1918)
315. Graf, U.: Trigonometrie der Ebene. Spärische Geometrie und Kartenentwürfe. Hochschulwis-
sen in Einzeldarstellungen. Quelle u. Meyer, Leipzig (1938)
316. Grafarend, E., Krumm, F.: Map Projections. Cartographic Information Systems. Springer,
Berlin (2006)
317. Graham, R.L., Rothschild, B.L., Straus, E.G.: Are there .n + 2 points in .E n with odd integral
distances? Am. Math. Mon. 81, 21–25 (1974)
318. Graham, R.L., Lagarias, J.C., Mallows, C.L., Wilks, A.R., Yan, C.H.: Apollonian circle
packings: number theory. J. Number Theory 100(1), 1–45 (2003)
319. Graham, R.L., Lagarias, J.C., Mallows, C.L., Wilks, A.R., Yan, C.H.: Apollonian circle
packings: geometry and group theory. I. The Apollonian group. Discrete Comput. Geom.
34(4), 547–585 (2005)
320. Graham, R.L., Lagarias, J.C., Mallows, C.L., Wilks, A.R., Yan, C.H.: Apollonian circle
packings: geometry and group theory. II. Super-Apollonian group and integral packings.
Discrete Comput. Geom. 35(1), 1–36 (2006)
321. Graham, R.L., Lagarias, J.C., Mallows, C.L., Wilks, A.R., Yan, C.H.: Apollonian circle
packings: geometry and group theory. III. Higher dimensions. Discrete Comput. Geom. 35(1),
37–72 (2006)
322. Greenberg, M.J.: Euclidean and non-Euclidean Geometries. Development and History. W. H.
Freeman and Company, New York (1993). First edition: 1973
323. Gregorac, R.J.: Feuerbach’s relation and Ptolemy’s theorem in n-space. Geom. Dedicata
60(1), 65–88 (1996)
324. Gregorac, R.J.: A general 3-4-5 puzzle. Eur. J. Combin. 17(6), 533–541 (1996)
325. Greinke, W.: Mathematische Kartographie. Praxis Math. 30(1), 15–41 (1988)
326. Grigorieva, E.: Methods of Solving Complex Geometry Problems. Birkhäuser/ Springer,
Cham (2013)
327. Grinberg, D., Olver, P.J.: The n body matrix and its determinant. SIAM J. Appl. Algebra
Geom. 3(1), 67–86 (2019)
328. Gritzmann, P., Klee, V.L.: On the complexity of some basic problems in computational
convexity. I. Containment problems. Discrete Math. 136(1–3), 129–174 (1994)
329. Gritzmann, P., Klee, V.: On the complexity of some basic problems in computational
convexity. II. Volume and mixed volumes. In: Polytopes: Abstract, Convex and Computational
(Scarborough, ON, 1993). NATO Advanced Science Institutes Series C: Mathematical and
Physical Sciences, vol. 440, pp. 373–466. Kluwer Academic Publishers, Dordrecht (1994)
330. Gritzmann, P., Klee, V.L., Larman, D.: Largest j -simplices in n-polytopes. Discrete Comput.
Geom. 13(3–4), 477–515 (1995)
331. Groemer, H.: On some mean values associated with a randomly selected simplex in a convex
set. Pac. J. Math. 45, 525–533 (1973)
332. Groemer, H.: On circumscribed cylinders of convex sets. Geom. Dedicata 46(3), 331–338
(1993)
333. Gromoll, D., Klingenberg, W., Meyer, W.: Riemannsche Geometrie im Grossen. Lecture
Notes in Mathematics, vol. 55. Springer-Verlag, Berlin-New York (1968)
334. Gromov, M.: Curvature, diameter and Betti numbers. Comment. Math. Helv. 56, 179–195
(1981)
304 Bibliography

335. Grove, K., Petersen, P. (Eds.): Comparison Geometry, Mathematical Sciences Research
Institute Publications, vol. 30. Cambridge University Press, Cambridge (1997)
336. Gruber, P.M.: Convex and Discrete Geometry. Springer, Berlin (2007)
337. Gruber, P.M.: Lattice packing and covering of convex bodies. Tr. Mat. Inst. Steklova
275, 229–238 (2011). Reprinted in Proc. Steklov Inst. Math. 275 (2011), no. 1, 229–238.
Klassicheskaya i Sovremennaya Matematika v Pole Deyatel’ nosti Borisa Nikolaevicha
Delone, pp. 240–249.
338. Grünbaum, B.: On a conjecture of H. Hadwiger. Pac. J. Math. 11, 215–219 (1961)
339. Grünbaum, B.: Polytopal graphs. In: Studies in Graph Theory, Part II. Studies in Mathematics,
vol. 12, pp. 201–224. Mathematical Association of America, Washington, DC (1975)
340. Grünbaum, B.: Convex Polytopes 2nd edn. Graduate Texts in Mathematics, vol. 221.
Springer-Verlag, New York (2003). First edition: Wiley and Sons, 1967. Prepared and with a
preface by Volker Kaibel, Victor Klee and Günter M. Ziegler.
341. Grünbaum, B., Shephard, G.C.: Spherical tilings with transitivity properties. In: The Geomet-
ric Vein – The Coxeter Festschrift, pp. 65–98. Springer, New York/Berlin (1981)
342. van Gruting, C.J.: Some remarks on properties of triangles and circles in elliptic geometry, II
(Dutch). Simon Stevin 28, 13–39 (1951)
343. Gueron, S.: Two applications of the generalized Ptolemy theorem. Am. Math. Mon. 109(4),
362–370 (2002)
344. Guggenheimer, H.W.: Plane Geometry and its Groups. Holden-Day, Inc., San Francisco
(1967)
345. Guo, Q.: Convexity theory on spherical spaces, I (Chinese, English summary). Sci. Sin. Math.
50(12), 1745–1772 (2020)
346. Guo, Q., Peng, Y.: Spherically convex sets and spherically convex functions. J. Convex Anal.
28(1), 103–122 (2021)
347. Guo, B.-N., Qi, F.: Monotonicity of functions connected with the gamma function and the
volume of the unit ball. Integral Transforms Spec. Funct. 23(9), 701–708 (2012)
348. Guo, B.-N., Qi, F.: Monotonicity and logarithmic convexity relating to the volume of the unit
ball. Optim. Lett. 7(6), 1139–1153 (2013)
349. Guo, R., Sönmez, N.: Cyclic polygons in classical geometry. C. R. Acad. Bulgare Sci. 64(2),
185–194 (2011)
350. Gupta, R.C.: An Indian extension of Ptolemy’s theorem. Ganita Bharati 16(1–4), 70–72
(1994)
351. Gyarmathi, L.: Konstruktive Lösung der Apollonius-Aufgabe im n-dimensionalen Raum
durch Benutzung einer Erweiterung der zyklographischen Abbildung auf mehrdimensionale
Räume. Publ. Math. Debrecen 1, 123–128 (1949)
352. Ha, J.-S., Shin, S.-Y.: Edge advancing rules for intersecting spherical convex polygons.
Internat. J. Comput. Geom. Appl. 12(3), 207–216 (2002)
353. Hadamard, J.: Lessons in Elementary Geometry, vol. 2. American Mathematical Soci-
ety/Education Development Center, Inc., Providence (2008). First edition: 1901
354. Hadwiger, H.: Hillsche Hypertetraeder. Gaz. Mat. (Lisbon) 12(50), 47–48 (1951)
355. Hadwiger, H.: Altes und Neues über konvex Körper. Birkhäuser, Basel/Stuttgart (1955)
356. Hadwiger, H.: Vorlesungen über Inhalt, Oberfläche und Isoperimetrie. Springer, Berlin (1957)
357. Hadwiger, H., Schneider, R.: Vektorielle Integralgeometrie. Elem. Math. 26, 49–57 (1971)
358. Hahn, L.-S.: Complex Numbers and Geometry. MAA Spectrum. Mathematical Association
of America, Washington, DC (1994)
359. Hajja, M.: The Gergonne and Nagel centers of an n-dimensional simplex. J. Geom. 83(1–2),
46–56 (2005)
360. Hajja, M.: Coincidences of centers of edge-incentric, or balloon, simplices. Results Math.
49(3–4), 237–263 (2006)
361. Hajja, M., Hayajneh, M.: Impurity of the corner angles in certain special families of simplices.
J. Geom. 105(3), 539–560 (2014)
362. Hajja, M., Krasopoulos, P.: Recovering a circumscriptible tetrahedron from its face areas.
Mediterr. J. Math. 16(6), Paper No. 156, 7 pp. (2019)
Bibliography 305

363. Hajja, M., Martini, H.: Orthocentric simplices as the true generalizations of triangles. Math.
Intelligencer 35(2), 16–28 (2013)
364. Hajja, M., Martini, H., Spirova, M.: New extensions of Napoleon’s theorem to higher
dimensions. Beitr. Algebra Geom. 49(1), 253–264 (2008)
365. Hajja, M., Hammoudeh, I., Hayajneh, M.: Kites as the only doubly special simplices. Beitr.
Algebra Geom. 56(1), 269–277 (2015)
366. Hajja, M., Hayajneh, M., Martini, H.: More characterizations of certain special families of
simplices. Results Math. 69(1–2), 23–47 (2016)
367. Hajja, M., Hammoudeh, I., Hayajneh, M.: Pre-kites: simplices having a regular facet. Beitr.
Algebra Geom. 58(4), 699–721 (2017)
368. Hajja, M., Hayajneh, M., Nguyen, B., Shaqaqha, S.: Irreducibility of the Cayley-Menger
determinant and of a class of related polynomials. Beitr. Algebra Geom. 59(2), 327–342
(2018)
369. Hajós, G.: Einführung in die Geometrie. Teubner, Leipzig (1970). First edition in Hungarian:
1960
370. Halbeisen, L., Hungerbühler, N.: Closed chains of conics carrying Poncelet triangles. Beitr.
Algebra Geom. 58(2), 277–302 (2017)
371. Hales, T.C.: A proof of the Kepler conjecture. Ann. Math. 162, 1065–1185 (2005)
372. Hales, T.C.: Dense Sphere Packings. A Blueprint for Formal Proofs. London Mathematical
Society Lecture Note Series, vol. 400. Cambridge University Press, Cambridge (2012)
373. Hales, T.C.: The strong dodecahedral conjecture and Fejes Tóth’s conjecture on sphere
packings with kissing number twelve. In: Discrete Geometry and Optimization. Fields
Institute Communications, vol. 69, pp. 121–132. Springer, New York (2013)
374. Hales, T.C., Ferguson, S.P.: The Kepler Conjecture. The Hales-Ferguson Proof. Including
papers reprinted from Discrete Comput. Geom. 36(1), 3–26 (2006). Edited by Jeffrey C.
Lagarias. Springer, New York, 2011
375. Hales, T.C., McLaughlin, S.: The dodecahedral conjecture. J. Am. Math. Soc. 23(2), 299–344
(2010)
376. Hales, T.C., Adams, M., Bauer, G., Dang, T.D., Harrison, J., Hoang, L.T., Kalyszyk, C.,
Magron, V., McLaughlin, S., Nguyen, T.T., Nguyen, Q.T., Nipkow, T., Obua, S., Pleso, J.,
Rute, J., Solovyev, A., Ta, T.H.A., Tran, N.T., Trieu, T.D., Urban, J., Vu, K., Zumkeller, R.:
A formal proof of the Kepler conjecture. Forum Math. Pi 5, e2, 29 pp. (2017). MR3659768.
https://doi.org/10.1017/fmp.2017.1239
377. Hammond, J.R.: Concise Spherical Trigonometry with Applications. Houghton Mifflin Co.,
New York (1943)
378. Hampton, M.: Cosines and Cayley, triangles and tetrahedra. Am. Math. Mon. 121(10), 937–
941 (2014)
379. Harris, S.G.: A triangle comparison theorem for Lorentz manifolds. Indiana Univ. Math. J.
31(3), 289–308 (1982)
380. Hartl, J.: Zerfallende Quadrikenschnitte und stereographische Projektion. J. Geom. 22(2),
149–152 (1984). Correction: J. Geom. 24 (1985), p. 101
381. Hartsfield, N., Ringel, G.: Pearls in Graph Theory: A Comprehensive Introduction. Dover
Publications, Inc., Mineola (2003). Reprint of the 1994 revised edition, first edition: 1990
382. Hartshorne, R.: Geometry: Euclid and Beyond. Undergraduate Texts in Mathematics.
Springer-Verlag, New York (2000)
383. Havel, T.F.: Some examples of the use of distances as coordinates for Euclidean geometry.
In: Invariant-theoretic Algorithms in Geometry (Minneapolis, MN, 1987). J. Symb. Comput.
11(5–6), 579–593 (1991)
384. Havel, T., Li, H.: From molecular distance geometry to conformal geometric algebra. In:
Handbook of Geometric Constraint Systems Principles, pp. 107–137. Discrete Mathematics
and Applications. CRC Press, Boca Raton (2019)
385. Havil, J.: John Napier. Life, Logarithms, and Legacy. Princeton University Press, Princeton
(2014)
306 Bibliography

386. Hayashi, T.: Un théorème de Casey en mathématiques japonaises. Tôhoku Math. J. 1, 204–
206 (1912). JFM 43.0582.06
387. Hayes, B.: An adventure in the nth dimension. In: Pitici, M. (Ed.) The Best Writing on
Mathematics 2012. Princeton University Press, Princeton (2013)
388. Hebda, J.J., Ikeda, Y.: Replacing the lower curvature bound in Toponogov’s comparison
theorem by a weaker hypothesis. Tohoku Math. J. (2) 69(2), 305–325 (2017)
389. Hebda, J.J., Ikeda, Y.: Necessary and sufficient conditions for a triangle comparison theorem.
Tohoku Math. J. 74(3), 329–364 (2022)
390. Heidari, F., Honari, B.: Generalization of Steiner’s porism via Poncelet’s theorem. Bull.
Iranian Math. Soc. 49(3), Paper No. 36, 10 pp. (2023)
391. Heil, E., Martini, H.: Special convex bodies. In: Handbook of Convex Geometry, vol. A,
pp. 347–385. North Holland, Amsterdam (1993)
392. Henk, M., Ziegler, G.M.: Kugeln im Computer–die Kepler-Vermutung. In: Alles Mathematik,
pp. 121–143. Vieweg, Braunschweig (2000)
393. Hertel, E.: Self-similar simplices. Beitr. Algebra Geom. 41(2), 589–595 (2000)
394. Hertel, E.: A class of self-similar d-polytopes. Rev. Roumaine Math. Pures Appl. 50(5–6),
647–655 (2005)
395. Hessenberg, G.: Ebene und sphärische Trigonometrie (4. Auflage). Sammlung Göschen, vol.
99. de Gruyter, Berlin (1940). Erste Auflage: 1899
396. Heyda, J.F.: Classroom Notes: vector derivation of the sine and cosine laws in spherical
trigonometry. Am. Math. Mon. 54(9), 544 (1947)
397. Hilbert, D., Cohn-Vossen, S.: Anschauliche Geometrie. Mit einem Anhang: Einfachste
Grundbegriffe der Topologie von P. S. Aleksandrov. Springer, Berlin und Heidelberg (1996).
Erste Ausgabe: Springer, 1932
398. Hinks, A.R.: Map Projections. Cambridg University Press, Cambridge (2016). First ed.: 1921
399. Hlinený, P., Kratochvíl, J.: Representing graphs by disks and balls (a survey of recognition-
complexity results). Discrete Math. 229(1–3), 101–124 (2001). Combinatorics, Graph
Theory, Algorithms and Applications
400. Holshouser, A.L., Molchanov, S., Reiter, H.: Apollonius problems. Pi Mu Epsilon J. 14(4),
261–267 (2016)
401. Honsberger, R.: Mathematical Gems. II. The Dolciani Mathematical Expositions, vol. 2.
Mathematical Association of America, Washington, DC (1976)
402. Horak, P., Kim, D.: Keller’s conjecture revisited. Int. Electron. J. Geom. 15(2), 175–177
(2022)
403. Horváth, À.: Constructive curves in non-Euclidean planes. Stud. Univ. Zilina, Math. Ser.
28(1), 13–42 (2016)
404. Hoschek, J.: Mathematische Grundlagen der Kartographie. B.I.-Wissenschaftsverlag,
Mannheim (1984). Erste Aufl. 1968
405. Hsiang, W.-Y.: Least Action Principle of Crystal Formation of Dense Packing Type and
Kepler’s Conjecture. Nankai Tracts in Mathematics, vol. 3. World Scientific Publishing Co.,
Inc., River Edge (2001). With a foreword by S. S. Chern
406. Hu, S., Su, X., Wang, Y.: A proof of Toponogov’s theorem in Alexandrov geometry. Proc.
Am. Math. Soc. 151(4), 1743–1748 (2023)
407. Huber, M.: Halving the bounds for the Markov, Chebyshev, and Chernoff inequalities using
smoothing. Am. Math. Mon. 126(10), 915–927 (2019)
408. Hudelson, M., Klee, V., Larman, D.: Largest j -simplices in d-cubes: some relatives of the
Hadamard maximum determinant problem. Linear Algebra Appl. 241/243, 519–598 (1996).
Proceedings of the Fourth Conference of the International Linear Algebra Society (Rotterdam,
1994)
409. Hug, D.: Random polytopes. In: Stochastic Geometry, Spatial Statistics and Random Fields.
Lecture Notes in Mathematics, vol. 2068, pp. 205–238. Springer, Heidelberg (2013)
410. Hug, D., Schneider, R.: Random conical tessellations. Discrete Comput. Geom. 56(2), 395–
426 (2016)
Bibliography 307

411. Hug, D., Thäle, C.: Splitting tessellations in spherical spaces. Electron. J. Probab. 24, Paper
No. 24, 60 pp. (2019)
412. Hug, D., Weil, W.: Lectures on Convex Geometry. Springer, Cham (2020)
413. Hug, D., Weis, J.A.: Crofton formulae for tensor-valued curvature measures. In: Tensor
Valuations and Their Applications in Stochastic Geometry and Imaging. Lecture Notes in
Mathematics, vol. 2177, pp. 111–156. Springer, Cham (2017)
414. Hug, D., Weis, J.A.: Kinematic formulae for tensorial curvature measures. Ann. Mat. Pura
Appl. (4) 197(5), 1349–1384 (2018)
415. Hug, D., Schneider, R., Schuster, R.: Integral geometry of tensor valuations. Adv. Appl. Math.
41(4), 482–509 (2008)
416. Hull, T.: Project Origami. Activities for Exploring Mathematics, 2nd edn. CRC Press, Boca
Raton (2013)
417. Hungerbühler, N., Kusejko, K.: Steiner’s porism in finite Miquelian Möbius planes. Adv.
Geom. 18(1), 55–68 (2018)
418. Ibragimov, Z.S., Suceava, B.: Ptolemy through the centuries. Math. Mag. 96(3), 259–271
(2023)
419. Ionascu, E.J.: Apollonius “circle” in spherical geometry. Int. J. Geom. 9(1), 75–84 (2020)
420. Itokawa, Y., Machigashira, Y., Shiohama, K.: Maximal diameter theorems for manifolds
with restricted radial curvature (English summary). In: Proceedings of the Fifth Pacific Rim
Geometry Conference (Sendai, 2000). Tohoku Mathematical Publications, vol. 20, pp. 61–68.
Tohoku University, Sendai (2001)
421. Itokawa, Y., Machigashira, Y., Shiohama, K.: Generalized Toponogov’s theorem for man-
ifolds with radial curvature bounded below. In: Explorations in Complex and Riemannian
Geometry. Contemporary Mathematics, vol. 332, pp. 121–130. American Mathematical
Society, Providence (2003)
422. Ivanov, G., Tsiutsiurupa, I.: On the volume of sections of the cube. Anal. Geom. Metr. Spaces
9(1), 1–18 (2021)
423. Iwata, S.: The problem of Apollonius in the n-dimensional space (Japanese summary). Sci.
Rep. Fac. Ed. Gifu Univ. Natur. Sci. 4, 138–148 (1969)
424. Iwata, S.: Generalizations of Steiner’s contact circle theorem to the n-dimensional space
(Japanese summary). Sci. Rep. Fac. Ed. Gifu Univ. Natur. Sci. 4, 349–354 (1970/1971)
425. Izmestiev, I.: Spherical and hyperbolic conics. In: Eighteen Essays in non-Euclidean Geome-
try. IRMA Lectures in Mathematics and Theoretical Physics, vol. 29, pp. 263–320. European
Mathematical Society, Zurich (2019)
426. Jacob, C.: On a mechanical meaning of Ptolemy, Möbius, Neuberg and Pompeiu theorems.
Rev. Roumaine Sci. Tech. Ser. Méc. Appl. 34(3), 213–222 (1989)
427. Jeger, M.: Die Gewinnung der Grundformen der sphärischen Trigonometrie aus der stere-
ographischen Projektion. Elem. Math. 41(5), 120–125 (1986)
428. Jennings, G.A.: Modern Geometry with Applications. Universitext, Springer, New York
(1994)
429. Jiang, J.: Large Sample Techniques for Statistics. Springer Texts in Statistics. Springer, New
York (2010)
430. Johnson, R.-A.: Advanced Euclidean Geometry. Dover Publication, New York (1960). First
edition: 1929
431. Johnson, N.L., Leone, F.C.: Statistics and Experimental Design in Engineering and the
Physical Sciences, vols. I, II. John Wiley & Sons, Inc., New York-London-Sydney (1964)
432. Josefsson, M.: Characterizations of cyclic quadrilaterals. Int. J. Geom. 8(1), 5–21 (2019)
433. Josefsson, M.: More characterizations of cyclic quadrilaterals. Int. J. Geom. 8(2), 14–32
(2019)
434. Joswig, M.: From Kepler to Hales, and back to Hilbert. Doc. Math., Extra vol. Optimization
Stories, 439–446 (2012)
435. Kabluchko, Z.: Expected f-vector of the Poisson zero polytope and random convex hulls in
the half-sphere. Mathematika 66(4), 1028–1053 (2020)
308 Bibliography

436. Kakulashvili, G.: Euclidean geometry analogs on the sphere. Proc. I. Vekua Inst. Appl. Math.
68, 46–67 (2018)
437. Kalai, G.: Polytope skeletons and paths. In: Handbook of Discrete and Computational
Geometry, pp. 331–344. CRC Press, Boca Raton (1997)
438. Kale, B.K., Muralidharan, K.: Parametric Inference. An Introduction. Alpha Science Interna-
tional Ltd., Oxford (2015)
439. Kambly, R.: Stereometrie und sphärische Trigonometrie (JFM 34.0571.02). Verlag F. Hirt,
Breslau (1903)
440. Kamiyama, Y.: The configuration space of regular spherical even polygons. Int. J. Math. Math.
Sci. Art. ID 7148538, 6 pp. (2019)
441. Kamiyama, Y.: The Euler characteristic of the regular spherical polygon spaces. Homol.
Homotopy Appl. 22(1), 1–10 (2020)
442. Kang, M.-H., Lin, W.-H.: Equilateral spherical drawings of planar Cayley graphs. J. Graph
Algorithms Appl. 25(1), 97–119 (2021)
443. Karcher, H.: Riemannian comparison constructions. In: Global Differential Geometry. MAA
Studies in Mathematics, vol. 27, pp. 170–222. Mathematical Association of America,
Washington, DC (1989)
444. Kautzleben, H.: Carl Friedrich Gauss und die Astronomie, Geodäsie und Geophysik seiner
Zeit. In: Festakt und Tagung aus Anlass des 200. Geburtstages von Carl Friedrich Gauss
(Berlin 1977). Abh. Akad. Wiss. DDR, Abt. Math. Naturwiss. Tech., vol. 3, pp. 123–136.
Akademie-Verlag, Berlin (1978)
445. Kells, L.M., Kern, W.F., Bland, J.R.: Plane and Spherical Trigonometry. McGraw-Hill Book
Company, New York/London (1940)
446. Kelly, P.J., Weiss, M.L.: Geometry and Convexity. A Study in Mathematical Methods. Pure
and Applied Mathematics. A Wiley-Interscience Publication. John Wiley & Sons, New York-
Chichester-Brisbane (1979)
447. Kendall, D.G.: Spherical triangles revisited. In: The Art of Statistical Science, pp. 105–
113. Wiley Series in Probability and Mathematical Statistics Probability and Mathematical
Statistics. Wiley, Chichester (1992)
448. Kendall, M.G., Moran, P.A.P.: Geometrical Probability. Griffin’s Statistical Monographs &
Courses, vol. 10. Hafner Publishing Co., New York (1963)
449. Kendall, D.G., Barden, D., Carne, T.K., Le, H.: Shape and Shape Theory. Wiley Series in
Probability and Statistics. John Wiley & Sons, Ltd., Chichester (1999)
450. Khatri, C.G.: Tchebycheff’s inequality–a classroom note. Gujarat Statist. Rev. 11(1), 40–43
(1984)
451. Khimshiashvili, G.: Cyclic polygons as critical points (English, Georgian summaries). Proc.
I. Vekua Inst. Appl. Math. 58, 74–83, 114 (2008)
452. Kiderlen, M.: Introduction into integral geometry and stereology. In: Stochastic Geometry,
Spatial Statistics and Random Fields. Lecture Notes in Mathematics, vol. 2068, pp. 21–48.
Springer, Heidelberg (2013)
453. Klain, D., Rota, G.-C.: Introduction to Geometric Probability. Cambridge University Press,
Cambridge (1997)
454. Klapper, M.H., DeBrota, D.: Use of Cayley-Menger determinants in the calculation of
molecular structures. J. Comput. Phys. 37(1), 56–69 (1980)
455. Klein, F.: Vorlesungen über nicht-euklidische Geometrie. Springer, Berlin (1968). Erstauflage
1928
456. Klotzek, B.: Einführung in die Differentialgeometrie, II. Deutscher Verlag der Wis-
senschaften, Berlin (1983)
457. Knight, R.D.: The Apollonius contact problem and Lie contact geometry. J. Geom. 83(1–2),
137–152 (2005)
458. Knörrer, H.: Geometrie. Ein Lehrbuch für Mathematik- und Physikstudierende. Vieweg,
Wiesbaden (2006)
459. Knothe, H.: Inversion of two theorems of Archimedes. Mich. Math. J. 4, 53–56 (1957)
Bibliography 309

460. Kock, A.: Heron’s formula, and volume forms (English, French summary). Cah. Topol. Géom.
Différ. Catég. 63(3), 239–258 (2022)
461. Koebe, P.: Kontaktprobleme der konformen Abbildung. Ber. Verh. Sächs. Akad. Wiss.
Leipzig, Math.-Phys. Kl. 88, 141–164 (1936)
462. Koecher, M., Krieg, A.: Ebene Geometrie. Springer, Berlin (2000)
463. König, H.: Non-central sections of the simplex, the cross-polytope and the cube. Adv. Math.
376, Paper No. 107458, 35 pp. (2021)
464. König, H., Koldobsky, A.: Volumes of low-dimensional slabs and sections in the cube. Adv.
Appl. Math. 47(4), 894–907 (2011)
465. König, H., Koldobsky, A.: On the maximal perimeter of sections of the cube. Adv. Math. 346,
773–804 (2019)
466. König, R., Weise, K.H.: Mathematische Grundlagen der höheren Geodäsie und Kartographie.
Erster Band: Das Erdsphäroid und seine konformen Abbildungen. Springer-Verlag, Berlin-
Göttingen-Heidelberg (1951)
467. Kokkendorff, S.L.: Gram matrix analysis of finite distance spaces in constant curvature.
Discrete Comput. Geom. 31(4), 515–543 (2004)
468. Kollár, J.: Algebraic hypersurfaces. Bull. Am. Math. Soc. (N.S.) 56(4), 543–568 (2019)
469. Komori, Y., Umezawa, R., Yasui, T.: On the area formulas of inscribed polygons in classical
geometry. Pure Appl. Math. Q. 16(3), 557–572 (2020)
470. Kondo, K., Tanaka, M.: Total curvatures of model surfaces control topology of complete
open manifolds with radial curvature bounded below. II. Trans. Am. Math. Soc. 362(12),
6293–6324 (2010)
471. Kondo, K., Tanaka, M.: Toponogov comparison theorem for open triangles. Tohoku Math. J.
(2) 63(3), 363–396 (2011)
472. Kondo, K., Tanaka, M.: Applications of Toponogov’s comparison theorems for open triangles.
Osaka J. Math. 50(2), 541–562 (2013)
473. Kondo, K., Ohta, S., Tanaka, M.: Topology of complete Finsler manifolds with radial flag
curvature bounded below. Kyushu J. Math. 68(2), 347–357 (2014)
474. Kontorovich, A.: The local-global principle for integral Soddy sphere packings. J. Mod. Dyn.
15, 209–236 (2019)
475. Kostin, A.V., Kostina, N.N.: An interpretation of Casey’s theorem and of its hyperbolic
analogue (Russian. English summary). Sib. Elektron. Mat. Izv. 13, 242–251 (2016)
476. Kostochka, A.: Coloring intersection graphs of geometric figures with a given clique number.
In: Towards a Theory of Geometric Graphs. Contemporary Mathematics, vol. 342, pp. 127–
138. American Mathematical Society, Providence (2004)
477. Krantz, S.G., McCarthy, J.E., Parks, H.: Geometric characterizations of centroids of sim-
plices. J. Math. Anal. Appl. 316(1), 87–109 (2006)
478. Kubota, T.: On the extended Ptolemy’s theorem in hyperbolic geometry. Sci. Rep. Tôhoku
Univ. 1, 131–156 (1912)
479. Kürzel, M.: Toponogov’s triangle comparison theorem in model spaces of nonconstant
curvature. Differ. Geom. Appl. 7(2), 161–180 (1997)
480. Kuklin, N.A.: Delsarte method in the problem on kissing numbers in high-dimensional spaces.
Proc. Steklov Inst. Math. 284, suppl. 1, S108–S123 (2014)
481. Kuperberg, G., Schramm, O.: Average kissing numbers for non-congruent sphere packings.
Math. Res. Lett. 1(3), 339–344 (1994)
482. Kurnik, Z., Volenec, V.: Die Verallgemeinerungen des Ptolemäischen Satzes und einiger
seiner Analoga in der euklidischen und den nichteuklidischen Geometrien (German. Serbo-
Croatian summary). Glasnik Mat. Ser. III 2(22), 213–243 (1967)
483. Kurnik, Z., Volenec, V.: Neue Verallgemeinerungen der Ptolemäischen Relationen in der
euklidischen und den nichteuklidischen Geometrien. (German. Serbo-Croatian summary).
Glasnik Mat. Ser. III 3(23), 77–86 (1968)
484. Kurz, S.: On the characteristic of integral point sets. Aust. J. Comb. 36, 241–248 (2006)
485. Lagarias, J.C.: Dense sphere packings: a blueprint for formal proofs [book review of
MR3012355]. Bull. Am. Math. Soc. (N.S.) 53(1), 159–166 (2016)
310 Bibliography

486. Lagarias, J.C., Shor, P.W.: Keller’s cube-tiling conjecture is false in high dimensions. Bull.
Am. Math. Soc. (N.S.) 27(2), 279–283 (1992)
487. Lagarias, J.C., Mallows, C.L., Wilks, A.R.: Beyond the Descartes circle theorem. Am. Math.
Mon 109(4), 338–361 (2002)
488. Lang, U., Schroeder, V.: On Toponogov’s comparison theorem for Aleksandrov spaces.
Enseign. Math. 59(3–4), 325–336 (2013)
489. Langevin, R.: La petite musique de la géométrie intégrale (French. English and French
summaries). In: La recherche de la vérité. Écrit. Mathematics, pp. 117–143. ACL-Éd.
Kangourou, Paris (1999)
490. Langevin, R.: Integral geometry from Buffon to geometers of today. Cours Spécialisés, vol.
23. Société Mathématique de France, Paris (2015)
491. Langlet, G.A.: A new fast direct solution to the problem of the sphere tangent to four spheres.
Acta Cryst. Sect. A 35(5), 836–837 (1979)
492. Lassak, M.: Spherical geometry – a survey on width and thickness of convex bodies. In:
Papadopoulos, A. (Ed.) Surveys in Geometry I, pp. 7–47. Springer, Cham (2022)
493. Lay, S.R.: Convex Sets and their Application. Robert E. Krieger Publishing Co., Inc., Malabar
(1992). Revised reprint of the 1982 original
494. Le, H.-L.: Random spherical triangles, I: Geometrical background. Adv. Appl. Probab. 21(3),
570–580 (1989)
495. Le, H.-L.: Random spherical triangles. II. Shape densities. Adv. Appl. Probab. 21(3), 581–594
(1989)
496. Lebesgue, H.L.: Démonstration du théorème de Lexell. Nouv. Ann. Math. 1re série 14, 24–26
(1855)
497. Lebesgue, H.L.: Exposition d’un Mémoire de M. W. Crofton. Nouvelles Ann. Math. (4) 12,
481–502 (1912)
498. Lee, T.-U., Liu, Y., Xie, Y.M.: Dividing a sphere hierarchically into a large number of
spherical pentagons using equal area or equal length optimization. Comput.-Aided Des. 148,
Paper No. 103259, 13 pp. (2022)
499. Leech, J.: The problem of the thirteen spheres. Math. Gaz. 40, 22–23 (1956)
500. Legendre, A.-M.: Éléments de Géométrie, 7th edn. Firmin Didot, Paris (1808)
501. Leonhardt, U., Philbin, T.: Geometry and Light. The Science of Invisibility. Dover Publica-
tions, Inc., Mineola (2010)
502. Levi, M., Tabachnikov, S.: The Poncelet grid and billiards in ellipses. Am. Math. Mon.
114(10), 895–908 (2007)
503. Levrie, P.: A straightforward proof of Descartes’ circle theorem. Math. Intelligencer 41(3),
24–27 (2019)
504. Lexell, A.J.: Solutio problematis geometrici ex doctrina sphaericorum. Acta Acad. Sci.
Imperialis Petropolitinae 5(1), 112–126 (1781)
505. Li, S.: Volume decay and concentration of high-dimensional Euclidean balls – a PDE and
variational perspective. Analysis (Berlin) 41(1), 25–29 (2021)
506. Li, X., Leng, G.: The mixed volume of two finite vector sets. Discrete Comput. Geom. 33(3),
403–421 (2005)
507. Liberti, L.: Mathematical programming bounds for kissing numbers. In: Optimization and
Decision Science: Methodologies and Applications. Springer Proceedings in Mathematics
and Statistics, vol. 217, pp. 213–222 Springer, Cham (2017)
508. Liberti, L., Lavor, C.: Six mathematical gems from the history of distance geometry. Int.
Trans. Oper. Res. 23(5), 897–920 (2016)
509. Liberti, L., Lavor, C., Maculan, N., Mucherino, A.: Euclidean distance geometry and
applications. SIAM Rev. 56(1), 3–69 (2014)
510. Lietzmann, W.: Elementare Kugelgeometrie mit numerischen und konstruktiven Methoden
(Studia Mathematica, Bd. III). Vandenhoeck & Ruprecht, Göttingen (1949)
511. Liu, C., Chang, Y.: Extremal problems for spherical convex polygons. Arch. Math. (Basel)
118(4), 435–450 (2022)
Bibliography 311

512. Lovász, L.: Graphs and Geometry. American Mathematical Society Colloquium Publications,
vol. 65. American Mathematical Society, Providence (2019)
513. Lysenko, V.I.: On the mathematical works of A. I. Lexell (Russian, English summary). In:
History and Methodology of the Natural Sciences, vol. XXV, pp. 104–112. Moskov. Gos.
University, Moscow (1980)
514. Lyuter, I.O.: On the history of the problem of Apollonius of Perga on the construction of a
circle tangent to three given circles (Russian, English summary). Istor.-Mat. Issled. (2) 1(36),
part 2, 82–94, 262 (1996)
515. Ma, T., Zhao, L., Yuan, J.: Inequalities for the incenter simplices. Math. Inequal. Appl. 10,
(3), 703–709 (2007)
516. Macaulay, W.H.: Solid Geometry. Cambridge University Press, Cambridge (1930)
517. Maeda, Y.: Seven types of random spherical triangle in .S n and their probabilities. In:
Computational Geometry and Graph Theory. Lecture Notes in Computer Science, vol. 4535,
pp. 119–126. Springer, Berlin (2008)
518. Maehara, H.: A simple proof for the maximal planar case of Steinitz’ theorem. Bull. Inst.
Math., Acad. Sinica 16, 373–376 (1988)
519. Maehara, H.: Lexell’s theorem via an inscribed angle theorem. Am. Math. Mon. 106, 352–353
(1999)
520. Maehara, H.: Isoperimetric theorem for spherical polygons and the problem of 13 spheres.
Ryukyu Math. J. 14, 45–57 (2001)
521. Maehara, H.: On a condition for the union of spherical caps to be connected. J. Combin.
Theory, Series A 101, 264–270 (2003)
522. Maehara, H.: On the intersection graph of random caps on a sphere. Eur. J. Combin. 25(5),
707–718 (2004)
523. Maehara, H.: The problem of thirteen spheres – a proof for undergraduates. Eur. J. Combin.
28(6), 1770–1778 (2007)
524. Maehara, H.: Bend-formulas in some sphere-systems. Yokohama Math. J. 60, 1–11 (2014)
525. Maehara, H.: Some replicating simplices other than Hill-simplices. Beitr. Algebra Geom.
55(2), 429–432 (2014)
526. Maehara, H.: When is a Hill-simplex reflection-symmetric? Beitr. Algebra Geom. 56(1), 279–
283 (2015)
527. Maehara, H.: Orthogonal projections of cubes and regular simplices into a plane. J. Geom.
107(3), 567–577 (2016)
528. Maehara, H.: On spherical four-point spaces. J. Geom. 107, 1–8 (2016)
529. Maehara, H.: Spherical ellipses and equilateral spherical triangles. Yokohama Math. J. 68,
69–77 (2022)
530. Maehara, H.: The triangle comparison theorem for spheres. Beitr. Algebra Geom. 64(4),
1027–1031 (2023)
531. Maehara, H., Martini, H.: On Lexell’s theorem. Am. Math. Mon. 124(4), 227–244 (2017)
532. Maehara, H., Martini, H.: Geometric probability on the sphere. Jahresber. Dtsch. Math.-Ver.
119(2), 93–132 (2017)
533. Maehara, H., Martini, H.: Simplices whose dihedral angles are all rational multiples of .π , and
related topics. Acta Math. Hungar. 155(1), 25–35 (2018)
534. Maehara, H., Martini, H.: An analogue of Sylvester’s four-point problem on the sphere. Acta
Math. Hungar. 155(2), 479–488 (2018)
535. Maehara, H., Martini, H.: Bipartite sets of spheres and Casey-type theorems. Results Math.
74(1), Paper No. 47, 20 pp. (2019)
536. Maehara, H., Martini, H.: On double k-systems of spheres. Results Math. 75(3), Paper No.
114, 11 pp. (2020)
537. Maehara, H., Martini, H.: On Cesàro triangles and spherical polygons. Aequationes Math. 96,
361–379 (2022)
538. Maehara, H., Martini, H.: Kissing numbers for balls with varying radii. Graphs Combin.
38(6), Paper No. 183, 9 pp. (2022)
312 Bibliography

539. Maehara, H., Martini, H.: Seven proofs of Lexell’s theorem: an excursion into spherical
geometry. Math. Intell. (2023). https://doi.org/10.1007/s00283-023-10281-7
540. Maehara, H., Martini, H.: Complete bipartite graphs as intersection graphs of equal spherical
caps. Preprint
541. Maehara, H., Martini, H.: On the Ţiţeica-Johnson theorem. Bull. Math. Soc. Sci. Math.
Roumanie Tome 67(115), No. 2, 231–237 (2024)
542. Maehara, H., Oshiro, A.: On Soddy’s hexlet and a linked 4-pair. In: Discrete and Computa-
tional Geometry (Tokyo, 1998). Lecture Notes in Computer Science, vol. 1763, pp. 188–193.
Springer, Berlin (2000)
543. Mandelbrot, B.B.: The Fractal Geometry of Nature. W. H. Freeman and Co., San Francisco
(1982)
544. Marenich, V.B.: Application of an extremal case in a comparison theorem for obtaining gap
theorems (Russian). Sibirsk. Mat. Zh. 33(2), 186–189, 223 (1992). Translation in Siberian
Math. J. 33 (1992), no. 2, 342–345
545. Martini, H.: A new view on some characterizations of simplices. Arch. Math. (Basel) 55(4),
389–393 (1990)
546. Martini, H.: Regular simplices in spaces of constant curvature. Am. Math. Mon. 100(2), 169–
171 (1993)
547. Martini, H.: Cross-sectional measures. In: Intuitive Geometry (Szeged, 1991). Colloquia
Mathematica Societatis János Bolyai vol. 63, pp. 269–310. North-Holland, Amsterdam
(1994)
548. Martini, H.: Recent results in elementary geometry, II. In: Proceedings of the 2nd Gauss
Symposium. Conference A: Mathematics and Theoretical Physics (Munich, 1993), pp. 419–
443. Sympos. Gaussiana, de Gruyter, Berlin (1995)
549. Martini, H.: On the theorem of Napoleon and related topics. Math. Semesterber. 43(1), 47–64
(1996)
550. Martini, H., Soltan, V.: A characterization of simplices in terms of visibility. Arch. Math.
(Basel) 72(6), 461–465 (1999)
551. Martini, H., Spirova, M.: The Feuerbach circle and orthocentricity in normed planes. Enseign.
Math. (2) 53(3–4), 237–258 (2007)
552. Martini, H., Spirova, M.: Clifford’s chain of theorems in strictly convex Minkowski planes.
Publ. Math. Debrecen 72(3–4), 371–383 (2008)
553. Martini, H., Weissbach, B.: Zur besten Beleuchtung konvexer Polyeder. Beitr. Algebra Geom.
17, 151–168 (1984)
554. Martini, H., Weissbach, B.: Napoleon’s theorem with weights in n-space. Geom. Dedicata
74(2), 213–223 (1999)
555. Martini, H., Wenzel, W.: Simplices with congruent k-faces. J. Geom. 77(1–2), 136–139
(2003)
556. Martini, H., Wu, S.: On orthocentric systems in strictly convex normed planes. Extracta Math.
24(1), 31–45 (2009)
557. Martini, H., Swanepoel, K.J., Weiss, G.: The geometry of Minkowski spaces — a survey. I.
Exp. Math. 19(2), 97–142 (2001)
558. Martini, H., Montejano, L., Oliveros, D.: Bodies of Constant Width. An Introduction to
Convex Geometry with Applications. Birkhäuser, Cham (2019)
559. Mathis, P., Schreck, P.: Equation systems with free-coordinates determinants. In: Automated
Deduction in Geometry. Lecture Notes in Computer Science, vol. 7993, pp. 59–70. Lecture
Notes in Artificial Intelligence. Springer, Heidelberg (2013)
560. Matoušek, J.: Lectures on Discrete Geometry. Springer, New York (2002)
561. Matoušek, J.: Using the Borsuk-Ulam Theorem. Springer-Verlag, Berlin (2003)
562. McCleary, J.: Trigonometries. Am. Math. Mon. 109(7), 623–638 (2002)
563. McKean, H.: Probability: the Classical Limit Theorems. Cambridge University Press, Cam-
bridge (2014)
Bibliography 313

564. McKee, T.A., McMorris, F.R.: Topics in Intersection Graph Theory. SIAM Monographs on
Discrete Mathematics and Applications. Society for Industrial and Applied Mathematics
(SIAM), Philadelphia (1999)
565. McMullen, P.: Volumes of projections of unit cubes. Bull. Lond. Math. Soc. 16(3), 278–280
(1984)
566. McMullen, P.: Simplices with equiareal faces. In: The Branko Grnbaum Birthday Issue.
Discrete Comput. Geom. 24(2–3), 397–411 (2000)
567. Melbourne, J., Roberto, C.: Quantitative form of Ball’s cube slicing and equality cases in the
min-entropy power inequality. Proc. Am. Math. Soc. 150(8), 3595–3611 (2022)
568. Melluish, R.K.: An Introduction to the Mathematics of Map Projections. Cambridge Univer-
sity Press, Cambridge (2014). First ed.: 1931
569. Melzak, Z.A.: Invitation to Geometry. A Wiley-Interscience Publication. Pure and Applied
Mathematics. John Wiley & Sons, Inc., New York (1983)
570. Menger, K.: Untersuchungen über allgemeine Metrik. Math. Ann. 100, 75–163 (1928)
571. Menger, K.: Géométrie générale. Memorial des Sciences Mathematiques, vol. 124. Gauthier-
Villars, Paris (1954)
572. Miernowski, A., Mozgawa, W., Rzymowski, W.: Minimal area n-simplex circumscribing a
strictly convex body. J. Geom. 87(1–2), 99–105 (2007)
573. Mikami, Y.: Casey’s theorem in Japanese mathematics. Tôhoku Math. J. 15, 289–296 (1919).
JFM 47.0028.05
574. Miles, R.E.: Random points, sets and tessellations on the surface of a sphere. Sankhya, Ser.
A 33, 145–174 (1971)
575. Milman, R.S., Parker, G.D.: Geometry. A Metric Approach with Models. Springer, New York
(1981)
576. Mirman, B.: Explicit solutions to Poncelet’s porism. Linear Algebra Appl. 436(9), 3531–3552
(2012)
577. Mitchell, L.: On Euclidean distances and sphere representations. Graphs Comb. 39(2), Paper
No. 36, 11 p. (2023)
578. Mitrinović, D.S., Pecarić, J.E., Volenec, V.: Recent Advances in Geometric Inequalities.
Kluwer, Dordrecht (1989)
579. Mittelmann, H., Vallentin, F.: High-accuracy semidefinite programming bounds for kissing
numbers. Experiment. Math. 19(2), 175–179 (2010)
580. Mnëv, N.E.: The universality theorems on the classification problem of configuration varieties
and convex polytopes varieties. In: Topology and Geometry – Rohlin Seminar. Lecture Notes
in Mathematics, vol. 1346, pp. 527–543. Springer, Berlin (1988)
581. Molnár, J.: Über eine Übertragung des Hellyschen Satzes in sphärische Räume. Acta Math.
Acad. Sci. Hungary 8, 315–318 (1957)
582. Moody, J., Stone, C., Zach, D., Zvavitch, A.: A remark on the extremal non-central sections
of the unit cube. In: Asymptotic Geometric Analysis. Fields Institute Communications, vol.
68, pp. 211–228. Springer, New York (2013)
583. Morgan, J., Tian, G.: Ricci Flow and the Poincaré Conjecture. Clay Mathematics
Monographs, vol. 3. American Mathematical Society/Clay Mathematics Institute, Provi-
dence/Cambridge (2007)
584. Morita, K.: Some theorems on kissing circles and spheres. J. Geom. Graph. 15(2), 159–168
(2011)
585. Moritsugu, S.: Extending the Descartes circle theorem for Steiner n-cycles. In: Automated
Deduction in Geometry. Lecture Notes in Computer Science, vol. 7993, pp. 48–58. Lecture
Notes in Artificial Intelligence. Springer, Heidelberg (2013)
586. Mortici, C.: Monotonicity properties of the volume of the unit ball in .R n . Optim. Lett. 4(3),
457–464 (2010)
587. Mozgawa, W., Bar billiards and Poncelet’s porism. Rend. Semin. Mat. Univ. Padova 120,
157–166 (2008)
588. Musin, O.R.: The kissing problem in three dimensions. Discrete Comput. Geom. 35, 375–384
(2006)
314 Bibliography

589. Musin, O.R.: The kissing number in four dimensions. Ann. Math. (2) 168(1), 1–32 (2008)
590. Musin, O.R.: Five essays on the geometry of László Fejes Tóth. In: New Trends in
Intuitive Geometry. Bolyai Society Mathematical Studies, vol. 27, pp. 321–333. János Bolyai
Mathematical Society, Budapest (2018)
591. Musin, O.R.: Analogs of Steiner’s porism and Soddy’s hexlet in higher dimensions via
spherical codes. Arch. Math. (Basel) 111(5), 493–501 (2018)
592. Musin, O., Tarasov, A.S.: The strong thirteen spheres problem. Discrete Comput. Geom.
48(1), 128–141 (2012)
593. Namikawa, Y.: On spherical hyperbolas (Japanese). Sugaku 11, 22–24 (1959/1960)
594. Naumann, H.: Über Vektorsterne und Parallelprojektionen regulärer Polytope. Math. Z. 67,
75–82 (1957)
595. de Salles Neto, L.L., Lavor, C., Lodwick, W.: A note on the Cayley-Menger determinant and
the molecular distance geometry problem. Inform. Sci. 559, 1–7 (2021)
596. Neutsch, W., Scherer, K.: Celestial Mechanics. An Introduction to Classical and Contempo-
rary Methods. Bibliographisches Institut, Mannheim (1992)
597. Northshield, S.: On integral Apollonian circle packings. J. Number Theory 119(2), 171–193
(2006)
598. Northshield, S.: Complex Descartes circle theorem. Am. Math. Mon. 121(10), 927–931
(2014)
599. O’Connor, C.P.J., Wieczorek, S., Amann, A.: Loxodromes in open multisection lasers. Phys.
Rev. A 107(5), Paper No. 053520, 12 pp. (2023)
600. Ogasawara, H.: The echelon Markov and Chebyshev inequalities. Commun. Stat. Theory
Methods 49(7), 1578–1591 (2020)
601. Ogilvy, C.S.: Excursions in Geometry. Dover, New York (1990). First edition: 1976
602. Oh, H.: Apollonian circle packings: dynamics and number theory. Jpn. J. Math. 9(1), 69–97
(2014)
603. Olivares, J.G.: Clavius, Pitiscus and the first proof of the cosine theorem for the sides of any
spherical triangle (Spanish. English and Spanish summaries). LLULL 45(91), 15–35 (2022)
604. Onn, S., Sturmfels, B.: A quantitative Steinitz theorem. In: Festschrift on the occasion of the
65th birthday of Otto Krötenheerdt. Beitr. Algebra Geom. 35(1), 125–129 (1994)
605. Ostermann, A., Wanner, G.: Geometry by its History. Springer, Berlin (2012)
606. Ostrovskii, M.I.: Minimal-volume shadows of cubes. J. Funct. Anal. 176(2), 317–330 (2000)
607. Ostrovskii, M.I.: Minimal-volume projections of cubes and totally unimodular matrices.
Linear Algebra Appl. 364, 91–103 (2003)
608. Ozone, J.: A review of Steiner chains in Wasan (Japanese, English summary). In: The Study
of the History of Mathematics 2016, pp. 87–98. Research Institute for Mathematical Science
(RIMS), Kyoto (2018)
609. Pach, J.: Geometric intersection patterns and the theory of topological graphs. In: Proceedings
of the International Congress of Mathematicians–Seoul 2014, vol. IV, pp. 455–474, Kyung
Moon Sa, Seoul (2014)
610. Pach, J., Agarwal, P.: Combinatorial Geometry. Wiley-Interscience Series in Discrete Mathe-
matics and Optimization. John Wiley & Sons, Inc., New York (1995)
611. Paluszny, M., Wilker, J.B.: A case of the 3-dimensional problem of Apollonius. Aequationes
Math. 41(2–3), 172–186 (1991)
612. Pambuccian, V.: A theorem on equiareal triangles with a fixed base. In: Eighteen Essays
in non-Euclidean Geometry, pp. 427–437. IRMA Lectures in Mathematics and Theoretical
Physics, vol. 29. European Mathematical Society (EMS), Zürich (2019)
613. Pambuccian, V., Schacht, C.: Euler’s inequality in Hilbert’s absolute geometry. J. Geom. 109,
Art. 8, 1–11 (2018)
614. Pambuccian, V., Zamfirescu, T.: Paolo Pizzetti: the forgotten originator of triangle comparison
geometry (English, Italian summary). Historia Math. 38(3), 415–422 (2011)
615. Paouris, G., Pivovarov, P., Zinn, J.: A central limit theorem for projections of the cube. Probab.
Theory Related Fields 159(3–4), 701–719 (2014)
Bibliography 315

616. Papadopoulos, A.: On the works of Euler and his followers on spherical geometry. Ganita
Bharati 36(1), 53–108 (2014)
617. Papadopoulos, A.: Metric Spaces, Convexity and Nonpositive Curvature. IRMA Lectures
in Mathematics and Theoretical Physics, vol. 6, 2nd edn. European Mathematical Society,
Zürich (2014)
618. Papadopoulos, A.: Euler, la géométrie sphérique et le calcul des variations: quelques points de
repère. Leonhard Euler, pp. 349–392. Science Musique Ser. Etudes, CNRS Èd., Paris (2015)
619. Papadopoulos, A.: Euler, Delisle and cartography. In: Mathematical Geography in the 18th
Century – Euler, Lagrange, and Lambert, pp. 113–138. Springer, Cham (2022)
620. Papadopoulos, A.: Geometry in the twentieth century: a return to Euclid - The work of Herbert
Busemann. Preprint (2024). arXiv: 2406.01109[math.HO]
621. Papadopoulos, A., Su, W.: On hyperbolic analogues of some classical theorems in spherical
geometry. In: Hyperbolic Geometry and Geometric Group Theory. Proceedings of the 7th
Seasonal Institute of the Mathematical Society of Japan (MSJ-SI). Advanced Studies in Pure
Mathematics, vol. 73, pp. 225–253. Mathematical Society of Japan (MSJ), Tokyo (2017)
622. Pech, P.: On equivalence of conditions for a quadrilateral to be cyclic. In: Computational
Science and its Applications – ICCSA 2011, Part IV. Lecture Notes in Computer Science,
vol. 6785, pp. 399–411. Springer, Heidelberg (2011)
623. Pecker, D.: Poncelet’s theorem and billiard knots. Geom. Dedicata 161, 323–333 (2012)
624. Pedoe, D.: On a theorem in geometry. Am. Math. Mon. 74, 627–640 (1967)
625. Pedoe, D.: A Course of Geometry for Colleges and Universities. Cambridge University Press,
London-New York (1970). Second edition: Geometry - A Comprehensive Course, Dover
Publ., New York, 1988
626. Pedoe, D.: The missing seventh circle. Elem. Math. 25, 14–15 (1970)
627. Pedoe, D.: Circles - A Mathematical View. Mathematical Association of America, Washing-
ton, DC (1995). First edition by Pergamon Press in 1957. Revised reprint of the 1979 edition.
With a biographical appendix on Karl Feuerbach by Laura Guggenbuhl. MAA Spectrum
628. Pekonen, O., Fenyvesi, K., J. Stén: Mathematical monuments in Finland. In: Bridges 2021,
Conference Proceedings, pp. 367–369. Tessellations Publishing, University of Helsinki,
Helsinki (2021)
629. Percy, A., Rogers, D.G.: Alternative road: from van Schooten to Ptolemy. Normat 57(3), 116–
128 (2009)
630. Perfect, H.: Topics in Geometry. Macmillan, New York (1963)
631. Persson, U.: Lexell’s theorem (Swedish, English summary). Normat 60(3), 133–134, 144
(2012)
632. Petersen, P.: Aspects of Global Riemannian Geometry. Bull. Am. Math. Soc. 36, 297–344
(1999)
633. Petersen, P.: Riemannian Geometry. Graduate Texts in Mathematics, vol. 171. Springer, New
York (2006)
634. Petrera, M., Suris, Y.B.: Spherical geometry and integrable systems. Geom. Dedicata 169,
83–98 (2014)
635. Pfender, F., Ziegler, G.M.: Kissing numbers, sphere packings, and some unexpected proofs.
Not. Am. Math. Soc. 51(8), 873–883 (2004)
636. Pfiefer, R.E.: The historical development of J. J. Sylvester’s four point problem. Math. Mag.
62(5), 309–317 (1989)
637. Pisanski, T., Randić, M.: Bridges between geometry and graph theory. In: Geometry at Work.
MAA Notes, vol. 53, pp. 174–194. Mathematical Association of America, Washington, DC
(2000)
638. Pivovarov, P.: Volume thresholds for Gaussian and spherical random polytopes and their
duals. Studia Math. 183(1), 15–34 (2007)
639. Pizzetti, P.: Paragone fra gli angoli di due triangoli geodetici di eguali lati. Atti della Reale
Accademia dei Lincei, Rendiconti (5). Classe di Scienze Fisiche, Matematiche e Naturali
16(1), 149–155 (1907). JFM 38.0638.01
316 Bibliography

640. Plofker, K.: The “ignominious fate” of spherical trigonometry? Ganita Bharati 32(1–2), 127–
143 (2010/2012)
641. Polya, G.: Mathematical Discovery. John Wiley & Sons, New York (1981)
642. Popescu, C.: Notes on the Ptolemy theorem. Nieuw Arch. Wisk. (4) 15(3), 193–197 (1997)
643. Popko, E.S., Kitrick, C.J.: Divided Spheres. Geodesics and the Orderly Subdivision of the
Sphere. CRC Press/A. K. Peters, Boca Raton (2022). First edition: 2012
644. Posamentier, A.S., Geretschläger, R.: The Circle – a Mathematical Exploration Beyond the
Line. With contributions by Christian Spreitzer and Erwin Rauscher. Prometheus Books,
Amherst (2016)
645. Prakasa Rao, B.L.S.: Chebyshev’s inequality for Hilbert-space-valued random elements.
Statist. Probab. Lett. 80(11–12), 1039–1042 (2010)
646. Prasolov, V.V., Tikhomirov, V.M.: Geometry. American Mathematical Society, Providence
(2001). Original version in Russian: Moscow, 1997
647. Propp, J.G.: Kepler’s spheres and Rubik’s cube. Math. Mag. 61(4), 231–239 (1988)
648. Proskurowski, W.: Flattening the earth: mathematical and historical aspects of Mercator
projection. Antiq. Math. 13, 235–255 (2019)
649. Protasov, V.Y.: Generalized closing theorems. Elem. Math. 66(3), 98–117 (2011)
650. Purisima, E.O., Scheraga, H.A.: An approach to the multiple-minima problem by relaxing
dimensionality. Proc. Nat. Acad. Sci. U.S.A. 83(9), 2782–2786 (1986)
651. Qi, F., Wei, C.-F., Guo, B.-N.: Complete monotonicity of a function involving the ratio of
gamma functions and applications. Banach J. Math. Anal. 6(1), 35–44 (2012)
652. Rabiei, N.: On the concentration of mass in generalized unit balls. Appl. Math. E-Notes 18,
259–267 (2018)
653. Rademacher, H., Toeplitz, O.: Von Zahlen und Figuren. Springer, Berlin (1933)
654. Ramírez, J.L., Rubiano, G.N., Zlobec, B.J.: Generating fractal patterns by using p-circle
inversion. Fractals 23(4), 1550047, 13 pp. (2015)
655. Rankin, R.A.: The closest packing of spherical caps in n dimensions. Proc. Glasgow Math.
Assoc. 2, 139–144 (1955)
656. Ranucci, E.R., Rollins, W.E.: Curiosities of the Cube. Crowell Company, New York (1977)
657. Rashed, R., Papadopoulos, A.: On Menelaus’ Spherics III.5 in Arabic mathematics, I. Arabic
Sci. Philos. 24(1), 1–68 (2014)
658. Rashed, R., Papadopoulos, A.: On Menelaus’ Spherics III.5 in Arabic mathematics, II. Arabic
Sci. Philos. 25(1), 1–32 (2015)
659. Rataj, J.: A translative integral formula for absolute curvature measures. Geom. Dedicata
84(1–3), 245–252 (2001)
660. Rauch, H.E.: A contribution to differential geometry in the large. Ann. Math. 54(2), 38–55
(1951)
661. Reed, N.: Peaucellier and the torus. C. R. Math. Rep. Acad. Sci. Canada 4(2), 87–91 (1982)
662. Reitzner, M.: Random polytopes. In: New Perspectives in Stochastic Geometry, pp. 45–76.
Oxford University Press, Oxford (2010)
663. Ren, D.L.: Topics in Integral Geometry. Series in Pure Mathematics, vol. 19. World Scientific
Publishing Co., Inc., River Edge (1994). Translated from the Chinese and revised by the
author. With forewords by Shiing Shen Chern and Chuan-Chih Hsiung
664. Reuschel, A.: Berührungseigenschaften des Feuerbachkreises. Zusammenhang zwischen
der Caseyschen Verallgemeinerung des Satzes von Ptolemäus und Sätzen über vier
Berührungskreise eines Kreises. Praxis Math. 18(1), 3–10 (1976)
665. Reventós Tarrida, A.: Inversive geometry (Spanish. Spanish summary). Gac. R. Soc. Mat.
Esp. 6(1), 39–79 (2003)
666. Richter-Gebert, J.: Realization Spaces of Polytopes. Springer, Berlin (1996)
667. Richter-Gebert, J.: Perspectives on Projective Geometry. A Guided Tour Through Real and
Complex Geometry. Springer, Heidelberg (2011)
668. Richter, W.-D., Schicker, K.: Ball numbers of Platonic bodies. J. Math. Anal. Appl. 416(2),
783–799 (2014)
Bibliography 317

669. Roanes-Lozano, E.: 3D extension of Steiner chains problem. Math. Comput. Modelling 45(1–
2), 137–148 (2007)
670. Roberts, F.S.: Graph Theory and its Applications to Problems of Society. CBMS-NSF
Regional Conference Series in Applied Mathematics, vol. 29. Society for Industrial and
Applied Mathematics (SIAM), Philadelphia (1978)
671. Robinson, P.L.: The sphere is not flat. Am. Math. Mon. 113, 171–172 (2006)
672. Roché, E., de Comberousse, C.: Traité de Géométrie. Gauthier-Villars, Paris (1922)
673. Ronda, J., Valdés, A.: Conic geometry and autocalibration from two images. J. Math. Imaging
Vision 28(2), 135–149 (2007)
674. Rosenfeld, B.A.: Axiome und Grundbegriffe der Geometrie. In: Enzyklopädie der Elementar-
mathematik IV, pp. 3–42, Deutscher Verlag der Wissenschaften, Berlin (1969). First edition
in Russian: Moscow 1963
675. Rosenfeld, B.A.: Die Grundbegriffe der sphärischen Geometrie und Trigonometrie. In:
Enzyklopädie der Elementarmathematik IV, pp. 527–569, Deutscher Verlag der Wis-
senschaften, Berlin (1969). First edition in Russian: Moscow, 1963
676. Rosenfeld, B.A.: A History of non-Euclidean Geometry. Springer, New York (1988). Original
version in Russian: Moscow, 1976
677. Rosenfeld, B.A., Yaglom, I.M.: Nichteuklidische Geometrie. In: Enzyklopädie der Elemen-
tarmathematik V, pp. 385–469, Deutscher Verlag der Wissenschaften, Berlin (1969). First
edition in Russian: Moscow, 1963
678. Roshdi, R., Papadopoulos, A.: Menelaus’ Spherics: Early Translation and al-
Mahani’/alHarawi’s Version. Critical edition of Menelaus’ Spherics from the Arabic
manuscripts, with historical and mathematical commentaries, dual Arabic-English text.
Scientia Graeco-Arabica, vol. 21. De Gruyter, Berlin (2017)
679. Rother, W., M. Zähle: Absolute curvature measures. II. Geom. Dedicata 41(2), 229–240
(1992)
680. Ryabogin, D., Zvavitch, A.: Analytic Methods in Convex Geometry. Analytical and Prob-
abilistic Methods in the Geometry of Convex Bodies. IMPAN Lecture Notes, vol. 2,
pp. 87–183. Polish Academy of Sciences Institute of Mathematics, Warsaw (2014)
681. Rylov, Y.A.: The metric space: classification of finite subspaces instead of constraints on the
metric (Russian. English and Russian summaries). In: Proceedings on Analysis and Geometry
(Novosibirsk Akademgorodok, 1999), pp. 481–504. Izdat. Ross. Akad. Nauk Sib. Otd. Inst.
Mat., Novosibirsk (2000)
682. Saban, G.: Integral geometry (Italian). Archimede 35(3), 110–126 (1983)
683. Sachs, H.: Coin graphs, polyhedra and coformal mappings. Discrete Math. 134, 133–138
(1994)
684. Sakano, Y., Akama, Y.: Anisohedral spherical triangles and classification of spherical tilings
by congruent kites, darts and rhombi. Hiroshima Math. J. 45(3), 309–339 (2015)
685. Saks, M.E.: Slicing the hypercube. In: Surveys in Combinatorics, 1993 (Keele). London
Mathematical Society Lecture Note series, vol. 187, pp. 211–255. Cambridge University
Press, Cambridge (1993)
686. Santaló, L.A.: A demonstration of the isoperimetric property of the circle (Spanish). Publ.
Inst. Mat. Univ. Nac. Litoral 2, 37–46 (1940)
687. Santaló, L.A.: Integral formulas in Crofton’s style on the sphere and some inequalities
referring to spherical curves. Duke Math. J. 9, 707–722 (1942)
688. Santaló, L.A.: Integral geometry on surfaces of constant negative curvature. Duke Math. J.
10, 687–709 (1943)
689. Santaló, L.A.: Integral geometry in spaces of constant curvature (Spanish. English summary).
Repub. Argentina. Publ. Com. Nac. Energia Atomica. Ser. Fis. 1(1), 68 pp. (1952)
690. Santaló, L.A.: Introduction to Integral Geometry. Publications de l’Institut Mathématique de
l’Université de Nancago, II. Current Scientific and Industrial Topics, vol. 1198. Hermann &
Cie, Paris (1953)
318 Bibliography

691. Santaló, L.A.: Integral Geometry and Geometric Probability. With a foreword by Mark Kac.
Encyclopedia of Mathematics and its Applications, vol. 1. Addison-Wesley Publishing Co.,
Reading-London-Amsterdam (1976). Second edition: 2004
692. Santaló, L.A.: Integral geometry: history and perspectives. In: Proceedings of the IV
International Colloquium on Differential Geometry (Univ. Santiago de Compostela, Santiago
de Compostela, 1978). Cursos Congresses University of Santiago de Compostela, vol. 15,
pp. 1–48. University of Santiago de Compostela, A Coruña (1979)
693. Sarnak, P.: Integral Apollonian packings. Am. Math. Mon. 118(4), 291–306 (2011)
694. Satyanarayana, K.: Angles and In- and Exelements of Triangles and Tetrahedra. Bangalore
Press, Bangalore (1962)
695. Saw, J.G., Yang, M.C., Mo, T.C.: Chebyshev inequality with estimated mean and variance.
Am. Stat. 38(2), 130–132 (1984)
696. Schilling, F.: Die Brennpunktseigenschaften der sphärischen Ellipse und ihre Übertragung auf
die ebene nichteuklidische elliptische Geometrie. Math. Ann. 121, 405–414 (1950)
697. Scheffers, G.: Wie findet und zeichnet man Gradnetze von Land- und Sternkarten? Zweite
Auflage, verbessert und erweitert von Karl Strubecker. Mathematisch-physikalische Biblio-
thek, Reihe I, 85/86. Teubner Verlagsgesellschaft, Stuttgart (1956). Erste Ausgabe: 1934
698. Schmidt, H.: Die Inversion und ihre Anwendungen. Verlag von R. Oldenbourg, Munich
(1950)
699. Schneider, R.: The mean surface area of the boxes circumscribed about a convex body.Ann.
Polon. Math. 25, 325–328 (1971/1972)
700. Schneider, R.: Simplices. Educational Talks in the Research Semester on Geometric Methods
in Analysis and Probability. Erwin Schrödinger Institute, Vienna (2005)
701. Schneider, R.: Crofton measures in projective Finsler spaces. In: Integral Geometry and
Convexity, pp. 67–98. World Scientific Publishing, Hackensack (2006)
702. Schneider, R.: Crofton measures in polytopal Hilbert geometries. Beitr. Algebra Geom. 47(2),
479–488 (2006)
703. Schneider, R.: Integral geometric tools for stochastic geometry. In: Stochastic Geometry.
Lecture Notes in Mathematics, vol. 1892, pp. 119–184. Springer, Berlin (2007)
704. Schneider, R.: Convex Bodies: the Brunn-Minkowski Theory. Second expanded edition.
Encyclopedia of Mathematics and its Applications, vol. 151. Cambridge University Press,
Cambridge (2014). First ed. 1993
705. Schneider, R.: Convex Cones – Geometry and Probability. Lecture Notes in Mathematics,
vol. 2319. Springer, Cham (2022)
706. Schneider, R., Weil, W.: Stochastic and Integral Geometry. Springer, Berlin/Heidelberg
(2008)
707. Schneider, R., Wieacker, J.A.: Integral geometry. In: Handbook of Convex Geometry, vol. A,
B, pp. 1349–1390. North-Holland, Amsterdam (1993)
708. Schnell, U.: Volumes of projections of parallelotopes. Bull. Lond. Math. Soc. 26(2), 180–185
(1994)
709. Schoenberg, I.J.: Linkages and distance geometry. Nederl. Akad. Wetensch. Proc. Ser. A 72.
Indag. Math. 31, 43–52 (1969)
710. Schoenberg, I.J.: Mathematical Time Exposures. MAA, Washington, DC (1982)
711. Schramm, O.: How to cage an egg? Invent. Math. 107, 543–560 (1992)
712. Schröder, E.: Kartenentwürfe der Erde. Kartographische Abbildungsverfahren aus mathema-
tischer und historischer Sicht. Mathematische Schülerbücherei, vol. 128. Teubner Verlagsge-
sellschaft, Leipzig (1988)
713. Schroth, A.E.: The Apollonius problem in flat Laguerre planes. J. Geom. 42(1–2), 141–147
(1991)
714. Schroth, A.E.: The Apollonius problem in four-dimensional Laguerre planes. J. Geom. 51(1–
2), 138–149 (1994)
715. Schröcker, H.-P.: Double tangent circles and focal properties of sphero-conics. J. Geom.
Graph. 12(2), 161–169 (2008)
Bibliography 319

716. Schütte, K., van der Waerden, B.L.: Das Problem der dreizehn Kugeln. Math. Ann. 125, 325–
334 (1953)
717. Schwartz, R.E.: Lengthening a tetrahedron. Geom. Dedicata 174, 121–144 (2015)
718. Schwartz, R., Tabachnikov, S.: Centers of mass of Poncelet polygons, 200 years after. Math.
Intelligencer 38(2), 29–34 (2016)
719. Schwartz, R.E., Tabachnikov, S.: Descartes circle theorem, Steiner porism, and spherical
designs. Am. Math. Mon. 127(3), 238–248 (2020)
720. Schwarz, D., Smith, G.C.: On the three diagonals of a cyclic quadrilateral. J. Geom. 105(2),
307–312 (2014)
721. Schwerdtfeger, H.: Geometry of Complex Numbers. Circle Geometry, Möbius Transforma-
tion, non-Euclidean Geometry. A corrected reprinting of the 1962 edition. Dover Books on
Advanced Mathematics. Dover, Inc., New York (1979)
722. Seberry, J.: Orthogonal Designs: Hadamard Matrices, Quadratic Forms and Algebras.
Springer, Cham, 2017. Revised and updated edition of the 1979 original.
723. Sedrakyan, H., Sedrakyan, N.: Geometric Inequalities. Methods for Proving. Problem Books
in Mathematics. Springer, Cham (2017)
724. Seidel, J.J.: Distance-geometric development of two-dimensional Euclidean, hyperbolical and
spherical geometry. I. Simon Stevin 29, 32–50 (1952)
725. Seidel, J.J.: Distance-geometric development of two-dimensional euclidean, hyperbolical and
spherical geometry. II. Simon Stevin 29, 65–76 (1951/1952)
726. Seneta, E.: A tricentenary history of the law of large numbers. Bernoulli 19(4), 1088–1121
(2013)
727. Seneta, E., Parshall, K., Jongmans, F.: Nineteenth-century developments in geometric prob-
ability: J. J. Sylvester, M. W. Crofton, J.-E. Barbier, and J. Bertrand. Arch. Hist. Exact Sci.
55(6), 501–524 (2001)
728. Serré, P., Ortuzar, A., Rivière, A.: Non-Cartesian modelling for analysis of the consistency of
a geometric specification for conceptual design. Internat. J. Comput. Geom. Appl. 16(5–6),
549–565 (2006)
729. Sharygin, I.F., Shtogrin, M.I.: Who discovered Soddy’s formula? (Russian). Mat. v Shkole
(2–3), 31–33 (1992)
730. Shiohama, K.: An Introduction to the Geometry of Aleksandrov Spaces. Lecture Notes
Mathematics, vol. 8. Seoul National University, Research Institute of Mathematics, Global
Analysis Research Center, Seoul (1993)
731. Sigl, R.: Ebene und sphärische Trigonometrie mit Anwendungen aus Kartographie, Geodäsie
und Astronomie. Akademische Verlagsgesellschaft, Frankfurt am (1977)
732. Silverberg, J.S.: Napier, Torporley, Menelaus, and Ptolemy: Delambre and De Morgan’s
observations on seventeenth-century restructuring of spherical trigonometry. In: Research in
History and Philosophy of Mathematics, pp. 149–168. Proceedings of the Canadian Society
for History Philosophy of Mathematics. Birkhäuser/Springer, Cham (2017)
733. Simon, M.: Über die Entwicklung der Elementar-Geometrie im 19. Jahrhundert. B.G.
Teubner, Leipzig (1906)
734. Simonić, A.: Lexell’s theorem via stereographic projection. Beitr. Algebra Geom. 60(3), 459–
463 (2019)
735. Sippl, M.J., Scheraga, H.A.: Solution of the embedding problem and decomposition of
symmetric matrices. Proc. Nat. Acad. Sci. U.S.A. 82(8), 2197–2201 (1985)
736. Sippl, M.J., Scheraga, H.A.: Cayley-Menger coordinates. Proc. Nat. Acad. Sci. U.S.A. 83(8),
2283–2287 (1986)
737. Skau, C.: Three pearls from elementary mathematics (Norwegian. English summary). Normat
48(2), 56–74, 95 (2000)
738. Skopets, Z.A.: Kegelschnitte. In: Enzyklopädie der Elementarmathematik, Band IV, pp. 571–
625. Deutscher Verlag der Wissenschaften, Berlin (1969). Russian original: Moscow, 1963
739. Skopets, Z.A.: Generalization of a theorem of Ptolemy (Russian). Mat. v Shkole (1), 60–61
(1987)
320 Bibliography

740. Skoruppa, N.-P.: Quick asymptotic upper bounds for lattice kissing numbers. Mathematika
49(1–2), 5–57 (2002/2004)
741. Sloane, N.J.A.: The sphere packing problem. In: Proceedings of the International Congress
of Mathematicians. Documenta Mathematica, vol. III, pp. 387–396. Berlin (1998)
742. Slutskiy, D.: De Tilly’s mechanical view on hyperbolic and spherical geometries. In: Eighteen
Essays in non-Euclidean Geometry. IRMA Lectures in Mathematics and Theoretical Physics,
vol. 29, pp. 93–111. European Mathematical Society, Zürich (2019)
743. Smart, W.M.: Textbook on Spherical Astronomy, 4th edn. Cambridge University Press, New
York (1960). First ed.: 1931
744. Smith, J.D.: Generalization of the triangle and Ptolemy inequalities. Geom. Dedicata 50(3),
251–259 (1994)
745. Smith, J.D.: The Ptolemy inequality and Minkowskian geometry. Math. Mag. 68(2), 98–109
(1995)
746. Smith, D.J., Vamanamurthy, M.K.: How small is a unit ball? Math. Mag. 62(2), 101–107
(1989)
747. Soddy, F.: The Hexlet. Nature 138, 958 (1936)
748. Soland, C.: Géométrie plane: une axiomatique centrée sur la distance (French). Elem. Math.
63(4), 173–183 (2008)
749. Solomon, H.: Geometric Probability. Ten lectures given at the University of Nevada, Las
Vegas, Nev., June 9–13, 1975. Conference Board of the Mathematical Sciences Regional
Conference Series in Applied Mathematics, vol. 28. Society for Industrial and Applied
Mathematics, Philadelphia (1978)
750. Soltan, V.: A characterization of homothetic simplices. Discrete Comput. Geom. 22(2), 193–
200 (1999)
751. Soltan, V.: Choquet simplexes in finite dimension – a survey. Expositiones Math. 22(4), 301–
31 (2004)
752. Sommerville, D.M.Y.: An Introduction to the Geometry of n Dimensions. Dover, New York
(1958)
753. Springborn, B.A.: A unique representation of polyhedral types. Centering via Möbius
transformations. Math. Z. 249(3), 513–517 (2005)
754. Stein, S.: Archimedes. What Did He Do Besides Cry Eureka? Mathematical Association of
America, Washington, DC (1999)
755. Steiner, J.: Verwandlung und Teilung sphärischer Figuren durch Construction. J. Reine
Angew. Math. 2, 45–63 (1827)
756. Steiner, J.: Sur le maximum et le minimum des figures dans le plan, sur la sphère et dans
l’espace général. J. Math. Pures Appl.1res série, 6, 105–170 (1841)
757. Steinert, K.-G.: Sphärische Trigonometrie mit einigen Anwendungen aus Geodäsie,
Astronomie und Kartographie. Teubner Verlagsgesellschaft, Leipzig (1977)
758. Steinitz, E.: Polyeder und Raumeinteilungen, Encyclopädie der mathematischen Wis-
senschaften, Band 3 (Geometrie). Teil 3AB12, pp. 1–139 (1922)
759. Steinitz, E., Rademacher, H.: Vorlesungen über die Theorie der Polyeder. Springer, Berlin
(1934). Reprint: Springer, 1976
760. Stén, J.C.-E.: A Comet of the Enlightenment. Anders Johan Lexell’s Life and Discoveries.
Vita Mathematica, vol. 17. Birkhuser, Basel (2014)
761. Stoka, M.: Géométrie intégrale. Gauthier-Villars, Paris (1968)
762. Sugimoto, T., Tanemura, M.: Random sequential covering of a sphere with identical spherical
caps. Forma 16(3), 209–212 (2001)
763. Sulanke, R., Wintgen, P.: Differentialgeometrie und Faserbündel. Birkhäuser, Basel (1972)
764. Svrtan, D., Veljan, D.: Non-Euclidean versions of some classical triangle inequalities. Forum
Geom. 12, 197–209 (2012)
765. Swanepoel, K.J.: Combinatorial distance geometry in normed spaces. In: New Trends in
Intuitive Geometry, pp. 407–458, Bolyai Society Mathematical Studies, vol. 27. János Bolyai
Mathematical Society, Budapest (2018)
Bibliography 321

766. Swart, D., Torrence, B.: Mathematics meets photography. In: Pitici, M. (Ed.) The Best Writing
on Mathematics 2012. Princeton University Press, Princeton (2013)
767. Szenthe, J.: On the metric theory of Euclidean space curves. I. Period. Math. Hungar. 3(3–4),
319–337 (1973)
768. Szpiro, G.G.: Kepler’s Conjecture. How Some of the Greatest Minds in History Helped Solve
one of the Oldest Math Problems in the World. John Wiley & Sons, Inc., Hoboken (2003)
769. Tabachnikov, S.: Kasner meets Poncelet. Math. Intelligencer 41(4), 56–59 (2019)
770. Taherian, S.G., Mohseni Takaloo, S.: A new solution of Apollonius’ problem based on
stereographic projections of Möbius and Laguerre planes. Beitr. Algebra Geom. 60(3), 465–
469 (2019)
771. Takayama, Y., Hibi, T.: Steinitz’s theorem analogue for two-dimensional Cohen-Macaulay
complexes. Adv. Appl. Math. 22(2), 200–218 (1999)
772. Tanaka, M., Michiwaki, Y., Hamada, T., Oyama, M., Kurosaki, K.: Ptolemy’s theorem
appeared in Wasan (Japanese). Sugakushi Kenkyu 63, 9–18 (1974)
773. Taylor, J.H.: A Euclidean proof of Casey’s extension of Ptolemy’s theorem. Quart. J. 26,
228–231 (1893)
774. Thébault, V.: Parmi les Belles Figures de la Géométrié dans l’Espace (Géométrié du
Tétraedre). Librairie Vuibert, Paris (1955)
775. Thomas, F., Porta, J.M.: Clifford’s identity and generalized Cayley-Menger determinants. In:
Advances in Robot Kinematics, 2020. Springer Proceedings in Advanced Robotics, vol. 15,
pp. 285–292. Springer, Cham (2020)
776. Thurston, W.P.: Three-dimensional Geometry and Topology. Princeton University Press,
Princeton (1997)
777. Tichy, R.F.: Random points on the sphere with applications to numerical analysis. Bericht
über die Wissenschaftliche Jahrestagung der GAMM (Karlsruhe, 1989). Z. Angew. Math.
Mech. 70(6), T642–T646 (1990)
778. Tong, Y.L.: Relationship between stochastic inequalities and some classical mathematical
inequalities. J. Inequal. Appl. 1(1), 85–98 (1997)
779. Toponogov, V.A.: Riemann spaces with curvature bounded below. Uspekhi Mat. Nauk
14(1)(85), 87–130 (1959) (in Russian)
780. Toponogov, V.A.: Differential Geometry of Curves and Surfaces. A Concise Guide.
Birkhäuser, Boston (2006). With the editorial assistance of Vladimir Y. Rovenski
781. Trapp, K.: Differential Geometry of Curves and Surfaces. Springer, Cham (2016)
782. Tropfke, J.: Geschichte der Elementarmathematik in systematischer Darstellung, Bd. 5: I.
Ebene Trigonometrie. II: Sphärik und sphärische Trigonometrie (2. Auflage). de Gruyter,
Berlin (1922)
783. Tuckwell, H.C.: Elementary Applications of Probability Theory. Chapman & Hall, London
(1988)
784. Tupan, A.: On the complex Descartes circle theorem. Am. Math. Mon. 129(9), 876–879
(2022)
785. Tutte, W.T.: How to draw a graph. Proc. Lond. Math. Soc. (3) 13, 743–767 (1963)
786. Ueno, Y., Agaoka, Y.: Classification of tilings of the 2-dimensional sphere by congruent
triangles. Hiroshima Math. J. 32(3), 463–540 (2002)
787. Ueno, K., Shiga, K., Morita, S.: A Mathematical Gift. II. The Interplay Between Topology,
Functions, Geometry, and Algebra. Mathematical World, vol. 20. American Mathematical
Society, Providence (2004). Translated from the 1995 Japanese original by Eiko Tyler
788. Ungar, A.A.: Analytic Hyperbolic Geometry in N dimensions. An Introduction. CRC Press,
Boca Taton (2015)
789. Urmaev, N.A.: Investigations on Mathematical Cartography (Russian). Trudy Central. Nauc.-
Issled. Inst. Geodez., Moscow (1953)
790. Valentine, J.E.: An analogue of Ptolemy’s theorem in spherical geometry. Am. Math. Mon.
77, 47–51 (1970)
791. Valentine, J.E.: An analogue of Ptolemy’s theorem and its converse in hyperbolic geometry.
Pac. J. Math. 34, 817–825 (1970)
322 Bibliography

792. Valentine, J.E.: On criteria of Blumenthal for inner-product spaces. Fund. Math. 72(3), 265–
269 (1971)
793. Valentine, J.E.: Banach-Euclidean four-point properties. Fund. Math. 106(3), 227–230 (1980)
794. Vallès, J.: A vector bundle proof of Poncelet’s closure theorem. Exp. Math. 30(4), 399–405
(2012)
795. Viazovska, M.S.: The sphere packing problem in dimension 8. Ann. Math. (2) 185(3), 991–
1015 (2017)
796. Viazovska, M.: Sharp sphere packings. In: Proceedings of the International Congress of
Mathematicians–Rio de Janeiro 2018, vol. II. Invited lectures, pp. 455–466. World Scientific
Publication, Hackensack (2018)
797. Volenec, V.: A characterization of the regular pentagon. Rad Jugoslav. Akad. Znan. Umjet.
428, 79–82 (1987)
798. Vu, K.T.: On the symplectic volume of the moduli space of spherical and Euclidean polygons.
Kodai Math. J. 28(1), 199–208 (2005)
799. Wagner, K.: Kartographische Netzentwürfe. B.I.-Wissenschaftsverlag, Mannheim (1962).
Erste Aufl. 1949
800. Weber, M.J., Schröcker, H.-P.: Minimal area conics in the elliptic plane. Adv. Geom. 12(4),
665–684 (2012)
801. Weber, M.J., Schröcker, H.-P.: Minimal area ellipses in the hyperbolic plane. Beitr. Algebra
Geom. 54(1), 181–200 (2013)
802. Weil, W., Wieacker, J.A.: Stochastic geometry. In: Handbook of Convex Geometry, vol. A, B,
pp. 1391–1438. North-Holland, Amsterdam (1993)
803. Weiss, A.I.: On Coxeter’s loxodromic sequences of tangent spheres. In: The Geometric Vein
- The Coxeter Festschrift, pp. 243–250. Springer, New York (1981)
804. Weiss, G.: The cube: its billiards, geodesics, and quasi-geodesics. J. Geom. Graph. 23(2),
201–210 (2019)
805. Weisstein, E.W.: CRC Concise Encyclopedia of Mathematics. Second edition. Chapman &
Hall, Boca Raton (2003)
806. Wells, D.: The Penguin Dictionary of Curious and Interesting Geometry. Penguin Books,
New York (1991)
807. Wendel, J.G.: A problem in geometric probability. Math. Scand. 11, 109–111 (1962)
808. Wenninger, M.J.: Spherical Models. With a foreword by Arthur L. Loeb. Cambridge
University Press, Cambridge/New York (1979)
809. Whittlesey, M.A.: Spherical Geometry and its Applications. Textbooks in Mathematics. CRC
Press, Boca Raton (2020)
810. Wilker, J.B.: Four proofs of a generalization of the Descartes circle theorem. Am. Math. Mon.
76, 278–282 (1969)
811. Wilker, J.B.: Inversive geometry. In: The Geometric Vein - The Coxeter Festschrift, pp. 379–
442. Springer, New York/Berlin (1981)
812. Wilker, J.B.: Inversive geometry and the hexlet. Geom. Dedicata 10(1–4), 469–473 (1981)
813. Wilker, J.B.: Inversive geometry and the Hopf fibration. Studia Sci. Math. Hungar. 21(1–2),
91–101 (1986)
814. Willmore, T.J.: Surfaces in conformal geometry. Special issue in memory of Alfred Gray
(1939–1998). Ann. Global Anal. Geom. 18(3–4), 255–264 (2000)
815. Wilson, P.M.H.: Curved Spaces. From Classical Geometries to Elementary Differential
Geometry. Cambridge University Press, Cambridge (2008)
816. Wimmer, L.: Cyclic polygons in non-Euclidean geometry. Elem. Math. 66(2), 74–82 (2011)
817. Winkler, P.: Mathematical Mind-Benders. A K Peters, Ltd. Wellesley (2007)
818. Wünsch, V.: Differentialgeometrie – Kurven und Flächen. Wissenschaftsverlag Thüringen,
Langewiesen (2012)
819. Xu, X.: A new proof of the Bowers-Stephenson conjecture. Math. Res. Lett. 28(4), 1283–1306
(2021)
Bibliography 323

820. Yaglom, I.M.: Geometrie der Kreise. In: Enzyklopädie der Elementarmathematik, Band IV,
pp. 459–526. Deutscher Verlag der Wissenschaften, Berlin (1969). Russian original: Moscow,
1963
821. Yaglom, I.M.: On a theorem of Ptolemy, a theorem of Casey and a paper by Z. A. Skopets
(Russian). Mat. v Shkole (1), 61–63 (1987)
822. Yaglom, I.M.: Geometric Transformations. IV: Circular Transformations. Anneli Lax New
Mathematical Library, vol. 44. Mathematical Association of America (MAA), Washington
(2009). Transl. from the Russian by A. Shenitzer
823. Yaglom, I.M., Atanasian, L.S.: Geometrische Transformationen. In: Enzyklopädie der Ele-
mentarmathematik, Band IV, pp. 43–151. Deutscher Verlag der Wissenschaften, Berlin
(1969). Russian original: Moscow, 1963
824. Yang, L., Zhang, J.Z.: A generalisation to several dimensions of the Neuberg-Pedoe inequal-
ity, with applications. Bull. Aust. Math. Soc. 27(2), 203–214 (1983)
825. Ye, Q.-Z., Wiegert, P.A., Hui, M.-T.: Finding long lost Lexell’s comet: the fate of the first
discovered near-Earth object. Astron. J. 155–163 (2018)
826. Yin, L.: Several inequalities for the volume of the unit ball in .R n . Bull. Malays. Math. Sci.
Soc. (2) 37(4), 1177–1183 (2014)
827. Yiu, P.: Rational Steiner porism. Forum Geom. 11, 237–249 (2011)
828. Zacharias, M.: Elementargeometrie und elementare nicht-euklidische Geometrie in synthetis-
cher Behandlung (JFM 45.0738.01). In: Encykl. math. Wiss., Heft 5, 859–962 (III AB 9)
(1914)
829. Zacharias, M.: Der Caseysche Satz. Jber. Deutsch. Math. Verein. 52, 79–89 (1942)
830. Zhang, G.: A lecture on integral geometry. In: Proceedings of the 14th International Workshop
on Differential Geometry and the 3rd KNUGRG-OCAMI Differential Geometry Workshop,
vol. 14, pp. 13–30. National Institute for Mathematical Sciences (NIMS), Taejon (2010)
831. Zhou, L., Hu, Z.-C.: Chebyshev’s inequality for Banach-space-valued random elements.
Statist. Probab. Lett. 82(5), 925–931 (2012)
832. Zhukova, A.: On the contribution of Anders Johan Lexell in spherical geometry. Ganita-
Bharati 41(1–2), 127–149 (2019)
833. Ziegler, G.M.: Three problems about 4-polytopes. In: Polytopes: Abstract, Convex and
Computational (Scarborough, ON, 1993). NATO NATO Advanced Science Institutes Series
C: Mathematical and Physical Sciences, vol. 440, pp. 499–502. Kluwer Academic Publishers,
Dordrecht (1994)
834. Ziegler, G.M.: Lectures on Polytopes. Springer, New York (1995)
835. Ziggelaar, A.: Francois d’Aguilon (1567–1617), Scientist and Architect. Institutum His-
toricum, Roma (1983)
836. Zong, C.: An example concerning the translative kissing number of a convex body. Discrete
Comput. Geom. 12(2), 183–188 (1994)
837. Zong, C.: Strange Phenomena in Convex and Discrete Geometry. Springer, New York (1996)
838. Zong, C.: The kissing numbers of convex bodies–a brief survey. Bull. Lond. Math. Soc. 30(1),
1–10 (1998)
839. Zong, C.: Sphere Packings. Springer-Verlag, Berlin (1999)
840. Zong, C.: What is known about unit cubes? Bull. Am. Math. Soc. (N.S.) 42(2), 181–211
(2005)
841. Zong, C.: The Cube: a Window to Convex and Discrete Geometry. Cambridge Tracts in
Mathematics, vol. 168. Cambridge University Press, Cambridge (2006)
842. Zong, C.: The kissing number, blocking number and covering number of a convex body. In:
Surveys on Discrete and Computational Geometry. Contemporary Mathematics, vol. 453,
pp. 529–548. American Mathematical Society, Providence (2008)
Index

Symbols n-star .K(1, n), 186

.B
n (r) (n-dimensional ball of radius r), 214 .r(σ ) (radius of .σ ), 31
.C(v) (cap of center v), 181 .s(Δ) (semi-perimeter of .Δ), 207
.Cd (v) (cap of radius d and center v), 183 .s(X ), 207
.H (P ) (hemisphere with pole P ), 85 .t (ci , cj ) (common tangent distance), 261
.K(m, n) (complete bipartite graph), 186 .v(G) (number of vertices of G), 53
.K1 ⊥ K2 (orthogonally crossing), 3 .S, S(r) (sphere of radius r), 75
.R (n-dimensional. Euclidean space), 213
.Kn (complete graph with n vertices), 50
n

.MX (x) (a special matrix), 206

◦
.X (polar set of .X ), 85
.N (v) (neighborhood of v), 50
.SD(X ) (squared-distance matrix of .X ),
240 A
n
.Sn (r) (boundary of .B (r)), 214 Abresch, U., 120
.Td (xyz) (Reuleaux d-triangle), 184 Abroshimov, N.V., 277
n
.V (n) (volume of .B (1)), 216 Adachi, T., 121
∗
.X (antipodal point of X), 75 Addition of a bypass, 52
.[w, x, y, z] (Möbius invariant), 263 Addition of a pointed edge, 52
.Δ̆ABC (spherical triangle ABC), 80 Adjacency relation, 49
.Š (a modification of the 2-sphere), 77 Aeppli, A., 43
.det SD(X : k), 271 Agahi, H., 200
.ν[Sn (r)] (maximum cardinality of a dispersed Agaoka, Y., 96
set on .Sn (r)), 223 Agarwal, P.K., 66, 69
.∂Cd (v) (boundary of .Cd (v)), 183 Agricoly, I., 92
.ρi (signed radius), 245 Aigner, M., 136
.𝚪(s) (gamma function), 217 Akama, Y., 96
.𝚪(X ) (Cayley-Menger matrix of .X ), 240 Akopyan, A., 19
.Ω(C ) (intersection graph of .C ), 181 Al-Afifi, C., 236
.|xy| (the length of the line segment connecting Aleksandrov, A.D., 118, 119
x to y), 1 Aleksandrov, P.S., 17
.b(σ ) (bend of .σ ), 31 Aleksandrov-Toponogov theorem, 118
d-cap graph, 186 Alencar, J., 253
d-independent, 183 Alexander, R., 254
.e(G) (number of edges of G), 53 Alfakih, A., 70, 253
.f (G) (number of faces of G), 53 Almohammad, S.M., 73

© The Author(s), under exclusive license to Springer Nature Switzerland AG 2024 325
H. Maehara, H. Martini, Circles, Spheres and Spherical Geometry,
Birkhäuser Advanced Texts Basler Lehrbücher,
https://doi.org/10.1007/978-3-031-62776-7
326 Index

Alonso, J., 7 Benz, W., 25

Alsina, C., 18, 274, 275 Berger, A., 48
Altshiller-Court, N., 16, 18, 22, 274 Berger, M., 7, 16, 23, 27, 44, 46, 90, 92, 94,
Amann, A., 26 117, 119, 141, 158, 230, 274
Andalafte, E.Z., 212 Berggren, J.L., 22
Andreev, E, 66 Bergold, H., 228
Angular radius, 79 Besau, F., 175
Anstreicher, K.M., 134, 136 Bezdek, K., 137, 141
Antipodal point, 75 Bhattacharya, A., 254
Aomoto, K., 256 Bicolored set of circles, 261, 266
Apollonian circle packings, 44 Bigalke, H.-G., 23, 92, 94, 117
Apollonian group, 45 Bilet, V., 257
Apollonius, 16, 21 Bilinski, S., 274, 277
Apollonius circle, 25 Billiard knots, 28
Apollonius problem, 25, 46 Billiards, 27, 28, 236
Arbeiter, E., 173 Bipartite graph, 55, 186
Archimedes, 21, 78 Birational geometry, 25
Armitage, D.H., 97 Björnbo, A.A., 115
Aseev, V.V., 277 Blair, D.E., 16
Askey, R., 275 Bland, J.R., 116
Asplund, E., 6 Blaschke, W., 22, 175
Asymptotic probability, 193 Blocking number, 137
Atanasian, L.S., 17 Blumenthal, L.M., 89, 212, 241, 251, 252, 254,
Athen. H., 116 258
Atzema, E.J., 157 Bodies of constant width, 90, 98, 185, 237
Aumann, G., 17, 18, 23, 46, 274, 276 Bokowski, J., 72
Avelino, C.P., 96 Boltyanski, V., 116, 230
Average kissing number, 140 Boltyanski, V.G., 232
Axler, S., 198 Bonnet, G., 175
Borceux, F., 17, 23, 89, 117, 274
Boreland, B., 17
B Böröczky, K., 134, 136, 140, 232, 234
Bachoc, C., 138 Böröczky, K.J., 173
Bakelman, I.J., 16 Borsuk, K., 22
Balestro, V., 229, 230 Bos, H.J.M., 26, 43
Balk, M.B., 232 Bottema, O., 251, 274, 276
Ball, K., 227 Boundary edge, 52
Ballicioni, A., 230 Boundary graph of a face, 53
Baragar, A., 48 Bourgain, J., 46
Baralić, D., 19 Bowman, F., 274
Bárány, I., 174, 175 Boyvalenkov, P., 138
Barden, D., 171, 173 Brakensiek, J., 229
Barnes, J., 18, 20 Brandstädt, A., 197
Barnette, D., 68, 71, 72 Brass, P., 69, 139, 259
Barrett, A.N., 171 Brauchart, S., 174
Bartha, F.A., 227 Breda, A.M.R., 96
Barthe, F., 227, 233 Brightwell, G.R., 63, 69
Baston, V.J.D., 230 Brocard porism, 29
Bayer, M.M., 72 Bruen, A.A., 47
Beecroft, P., 43 Budny, K., 200
Bédaride, N., 236 Burago, D., 118
Belotti, M., 73 Burago, Yu., 118, 122
Bend, 31 Burkholder, E.F., 117
Benson, R.V., 230 Byer, O.D., 19
Index 327

C Congruent, 84
Caddeo, R., 93 Connected, 51
Calinger, R., 157, 158 Convex, 80, 82
Campos, M., 139 Convex hull, 88, 162
Cap graph, 186 Coolidge, J.L., 16, 18, 276
Carne, T.K., 171, 173 Couderc, P., 230
Cartography, 24, 93 Courant, R., 16
Casey, J., 157, 276 Court, N.A., 230
Casey’s theorem, 262, 276 Covering number, 137
Casselman, B., 97, 134, 138 Coxeter, H.S.M., 1, 16–18, 20, 22, 43, 47, 136,
Catalan, E.C., 157 158, 274
Cayley, A., 251 Cramer, H., 199
Cayley-Menger determinant, 240 Crawley, E., 116
Cayley-Menger matrix, 240, 251 Crippen, G.M., 258
Cecil, T.E., 24 Croft, H.T., 139, 171, 230
Celestial mechanics, 117 Crofton, M.W., 175
Center-angle, 206 Crofton’s formula, 165, 167, 175
Cerioli, M.R., 70 Crossing angle, 3
Cesàro, G., 99 Crystallography, 23
Cesàro triangle, 99, 158 Cube, 214
Chakerian, G.D., 185, 227, 228 Cube slicing, 227
Chakrabarty, I., 198 Cundy, H.M., 17
Chang, Y., 96 Cut set, 63
Chapple, W., 18, 26 Cycle, 51
Charitos, C., 93 Cyclic, 111
Chavel, I., 118–120 Cylindrical projection, 77, 90
Chebyshev’s inequality, 192, 198
Cheeger, J., 119
Chern, S.-S., 176, 177 D
Chilakamarri, K.B., 155 d’Aguilon, 22
Chintshin, A.J., 17 Dai, X., 121
Cieślak, W., 28 Dalbec, J.P., 257
Circumcircle, 6, 7 D’Andrea, K., 254
Circumscribed cap, 83, 112 Datta, B.B., 115
Circumscribed sphere, 214 Davis, D.J., 6
Clancy, C., 17 Dawson, R.J.MaxG., 95
Classical geometry, 25 Day, M.M., 212
Claudius Ptolemy, 273 Deák, I., 171
Clemens, C.H., 94 Debarnot, M.-T., 115
Cohen-Macaulay simplicial complexes, 73 DeBrota, D., 258
Cohn, H., 138, 139 de Comberousse, C., 230
Cohn-Vossen, S., 5, 17, 22 Degree, 50
Coin graph, 50, 65 Dehbi, L., 256
Coin graph theorem, 66, 69 Dekster, B.V., 231
Coin representation, 65 Del Centina, A., 26
Colding, T.H., 121 De Loera, J., 229
Colin de Verdière, Y., 70 Delsarte, P., 138
Colins, K.D., 252 Deltheil, R., 175
Common tangent distance, 261 Density function, 192
Complete bipartite graph, 186 de Oliveira Filho, F.M., 140
Complete graph, 50, 54, 186 Derived triangle, 99
Complete set, 90, 98 de Salles Neto, L.L., 259
Complex numbers, 16, 274, 275 Descartes’ circle theorem, 37, 43
Component, connected component, 52 Diameter, 183
328 Index

Dietrich, U., 115 Ewald, G., 16, 71, 72

Differential geometry, 23, 24, 91–93, 117, Exterior of ellipse, 144
258 Extremal cap, 182
Diminnie, C., 212
Dirnböck, H., 155
Disconnected, 51 F
Dispersed, 223 Fabricius-Bjerre, F., 155
Distance, 201 Face of a plane graph, 52
Distance geometry, 89 Face-point, 65
Dodecahedral conjecture, 141 Facet, 214
Dodunekov, S., 138 Falconer, K.J., 139, 171, 230
Donnay, J.D.H., 22, 101 Faria. L., 70
Dörrie, H., 46, 116, 117, 251 Feeman, T.G., 24, 92
Dordovskii, D., 257 Fejes Tóth, G., ix, 135, 140, 157, 158
Dos Santos, A.M.F, 96 Fejes Tóth, L., ix, 95, 104, 135, 140, 141, 146,
Dos Santos, J.M., 96 157, 158
Dostert, M., 140 Fejes Tóth’s lemma, 135, 154
Dovgoshey, O., 257 Felsner, S., 69, 71
Doyle, B., 95 Fenn, R., 94
Dragović, V., 27, 28 Fenyvesi, K., 158
Drechsler. K., 47 Ferguson, S.P., 140
Dual edge, 68 Ferreira, T.O., 70
Dual graph, 68 Fiala, F., 92
Dubey, K.K., 254 Fiedler, M., 231, 252
Dunajski, M., 90 Filliman, P., 228
Dunham, W., 18 Fisher, J.C., 47
Duparc, H.J.A., 277 Fitz-Gerald, J.M., 47
Dupin cyclide, 20 Flatto, L., 27
Dúran, G., 197 Florian, A., 95
Foci, 143
Fodor, F., 98, 227
E Four-point properties, 126, 130, 162, 203, 205,
Ebin, D.G., 119 210, 211, 275
Eckhoff, J., 197 Fractals, 17
Edge, 80 Fraivert, D., 275
Edge of a graph, 50 Frame, M, 17
Edmonds, A.L., 231, 234, 235 Frank, R., 227
Ehrenfeucht, A., 226 Frankl, P, 234
Elerath, D.E., 120 Freeden, W., 91
Elkies, N., 138 Freese, R.W., 212
Ellis, D.C., 197 Frenkel, E„ 158
Embeddable, 52 Friedrich, T., 92
Epps, T.W., 198 Fuchs, D., 27
Eppstein, D., 70 Fuchs, E., 46
Equiareal simplices, 234 Fukagawa, H., vii, 19, 20, 37, 43, 276
Equilateral spherical triangle, 146
Euler, L., 23, 156, 157
Euler inequality, 17 G
Euler’s formula, 53 Gaifullin, A.A., 255
Euler’s formula for spherical excess, 105 Gale, D„ 230
Euler’s polyhedral formula, 82 Galicer, D., 233
Euler’s triangle theorem, 6, 17 Gamma function, 217
Even-cycle, 51 Gao, F., 90, 95, 173
Eves, H., 17, 18, 20, 274 Garcia, R., 28
Index 329

Garcia-Roig, J.-L., 275 Gruber, P.M., 69, 71, 138

Gardner, M., 216 Gueron, S., 277
Gardner, R.J., 227 Guggenheimer, H.W., 16
Gauss-Bonnet theorem, 23 Guo, Q., 90
Gdawiec, K., 17 Guo, R., 97
Gdowski, B., 92 Gupta, R.C., 274
Geiges, H., 26 Guy, R.K., 139, 171, 230
General position, 162 Gyarmathi, L., 47
Geodesic segment, 76
Geodesy, 24
Geometric probability, 170, 172 H
Ger, R., 275 Ha, J.-S., 98
Geretschläger, R., 17, 18, 27, 44, 46, 274 Hadamard, J., 16, 158
Giering, O., 16, 23, 27 Hadwiger number, 137
Gillam, B.E„ 251 Hadwiger, H., 176, 177, 231, 235
Girard’s formula, 81 Hahn, L.-S., 274
Girstmair, K., 115 Hajós, G., 16
Glaeser, G., 23 Hajja, M., 231, 232, 235, 236, 255
Glasauer, S., 171 Halbeisen, L., 28
Glazyrin, A., 97, 136 Hales, T.C., vii, 140, 141
Goethals, J.M., 138 Halley, 22
Goldberg, M., 17, 98 Hamada, T., 274
Golumbic, M.C., 197 Hamburger, P., 155
González Merino, B., 227 Hamdan, A., 236
González-Velasco, E., 273 Hammond, J.R., 116
Goodey, P.R., 178, 179 Hammoudeh, I., 235, 255
Gower, J.C., 258 Hampton, M., 252
Gowers, T., 90 Hand-shaking lemma, 50
Grünbaum, B., 6, 68, 69, 71, 95, 139, 230 Harriot, 22
Grace, J.H., 7 Harris, S.G., 120
Grace-Danielsson inequality, 7 Hartl, J., 25
Graf, U., 116 Hartsfield, N., 69
Grafarend, E., 91 Hartshorne, R., 16, 46, 89, 94
Graham, R.L., 45, 243, 258 Havel, T., 253, 257
Gram matrix, 242 Havil, J., 115
Graph, 49 Havlicek, H., 93
Graph of convex polyhedron, 50 Hayajneh, M., 235, 255
Graph of polyhedron, 49 Hayashi, T., 276
Great circle, 75 Hebda, J.J., 122
Greenberg, M.J., 16 Heidari, F., 20
Gregorac, R.J., 254, 275 Heil, E., 231
Gregory-Newton problem, 123, 128, 135, 138 Hemisphere, 75
Greinke, W., 93 Henk, M., 140, 173
Greitzer, S.L., 16–18, 22, 274 Hersh, R., 6
Grigoriev, A., 48 Hertel, E., 235
Grigorieva, E., 274 Hessenberg, G., 116
Grinberg, D., 259 Heule, M., 229
Grippo, L., 197 Hexlet, 10
Gritzmann, P., 229, 251 Hexlet of Soddy, 20
Groemer, H., 171, 185 Heyda, J.F., 117
Gromoll, D., 119 Hibi, T, 73
Gromov, M., 120, 122 Hilbert, D., 5, 17, 22
Grove, K., 119 Hill simplex, 231
330 Index

Hinks, A.R., 91 J
Hipparchus, 21 Jacob, C., 274
Hlinený, P., 70 Japanese temple geometry, 19, 20, 43, 274, 276
Holshouser, A.L., 44 Jeger, M., 118
Honari, B., 20 Jennings, G.A., 89, 104, 117
Honsberger, R., 18, 20 Jenssen, M., 139
Horak, P., 229 Jiang, J., 198
Horizontal line, 77 Johnson, N.L., 198
Horizontal plane, 77 Johnson, R.A., 5, 16–18, 22, 46, 274
Horváth, À., 155, 156 Jokanović, D.S., 19
Hoschek, J., 92 Josefsson, M., 275
Hsiang, W.-Y., 136 Joswig, M., 73, 140
Hu, S., 122
Hu, Z-C., 200
Huber, M., 199 K
Hudelson, H., 229 Kabluchko, Z., 174
Hug, D., 90, 172–174, 177, 178, 234 Kakulashvili, G., 97
Hui, M.-T., 158 Kalai, G., 72
Hull, T., 117 Kale, B.K., 199
Hungerbühler, N., 19, 28 Kambly, R., 116
Hypersphere, 214 Kamiyama, Y., 97
Kang, M.-H., 155
Karcher, H., 119
I Kautzleben, H., 93
Ibragimov, Z.S., 273 k-connected, 63
Ikeda, Y., 122 Keller’s conjecture, 229
Incircle, 5, 7 Kells, L.M., 116
Inscribed, 112 Kelly, P.J., 230
Inscribed angle theorem, 83 Kendall, D.G., 171, 173
Integral geometry, 176, 177 Kendall, M.G., 175
Interior of ellipse, 144 Kepler conjecture, 140
Intersection graphs, 181, 186, 196, 197 Kepler problem, 26
Invariance of angles, 12 Kern, W.F., 116
Invariant of graphs, 70 Kers, C., 26
Inversion, 1, 36 Khatri, C.G., 199
Inversion in circles, 15 Khimshiashvili, G., 275
Inversor of Hart, 17 Kiderlen, M., 177, 178
Invisibility, 23 Kim, D., 229
Ionascu, E.J., 25 Kissing number, 137, 225
Irreducible, 55 Kitrick, C.J., 90, 94
Isolated, 182 Klain, D, 176
Isometric, 201, 239 Klapper, M.H., 258
Isometrically embeddable, 202, 239, 257, 268 Klee, V., 229, 251
Isometric radii change, 266 Klein, F., 22
Isoperimetric theorem for spherical Kleinschmidt, P., 72
quadrilaterals, 112 Klingenberg, W., 119
Isoperimetric-type results, 96, 97, 111, 146, Klotzek, B., 92
150, 158 Knörrer, H., 23
Itokawa, Y., 120 Knight, R.D., 47
Ivanov, G., 227, 228 Kock, A., 256
Ivanov, S., 118 Koebe, P., 66, 69, 73
Ivory’s theorem, 156 Koecher, M., 92
Iwata, S., 19, 47 Kokkendorff, S.L., 254
Izmestiev, I., 155 Koldobsky, A., 227, 228
Index 331

Kollár, J., 25 Levi, S, 28

Kolpakov, A., 140 Levrie, P., 43
Komori, Y., 98 Lexell, A.J., 156, 157
Kondo, K., 121, 122 Lexell’s theorem, 25, 145, 156
König, H., 228 Li, H., 253
König, R., 24, 91 Li, X., 257
Kontorovich, A., 44, 48 Liberti, L., 138, 252, 253
Kostin, A.V., 277 Lie contact geometry, 47
Kostina, N.N., 277 Lietzmann, W., 116, 117
Kostochka, A., 197 Lin, W.-H., 155
Krantz, S.G., 232 Linear, 202
Krasopoulos, P., 256 Linked 4-cycle pair, 41
Kratochvíl, J., 70 Liu, C., 96
Krieg, A., 92 Liu, Y., 96
Krumm, F., 91 Location problem, 48
Kubota, T., 277 Lodwick, W., 259
Kuklin, N.A., 138 Logothetti, D., 227
Kumar, A., 139 Lovász, L., 69, 71
Kunze, H., 17 Lune, 81
Kuperberg, G., 139 Lysenko, V.I., 158
Kuperberg, W., ix, 135, 140, 157, 158 Lyuter, I.O., 46
Kureethara, J.V., 198
Kurnik, Z., 277
Kurosaki, K., 274 M
Kürzel, M., 121 Ma, T., 233
Kusejko, K., 19 Macaulay, W.H., 116
Machida, Y., 256
Machigashira, Y., 120
L Mackay, A.L., 171
Lagarias, J.C., 44, 45, 140, 229 Mackey, J., 229
Lambert, J.H., 78 Maculan, N., 252
Lángi, Z.L., 73 Maeda, Y., 174
Lang, U., 121 Maehara, H., 5, 19, 21, 68, 97, 101, 106, 111,
Langevin, R., 175, 178 119, 134, 136, 140, 146, 156, 157,
Langlet, G.A., 47 164, 172–174, 178, 228, 231, 234,
Larcher, H., 120 235, 271, 277
Larman, D., 229 Magnus, L.J., 1
Lassak, M., 90, 95, 98 Major triangle, 125
Lavor, C., 252, 253, 259 Mallows, C.L., 44
Law of total probability, 167 Mandelbrot, B.B., 17
Lay, S.R., 230 Marenich, V.B., 121
Le, H., 171, 173 Markov’s inequality, 192, 198
Le, H.-L., 173 Markushevich, A.I., 17
Le, V.B., 197 Martini, H., 5–7, 26, 28, 97, 101, 140, 146, 157,
Lebesgue, H.L., 157, 175 164, 172, 174, 178, 185, 227–232,
Lee, C.W., 72 234, 235, 255, 271, 277
Lee, T.-U., 96 Mathematical cartography, 90
Leech, J., 134, 136 Mathis, P., 259
Legendre, A.-M., 82, 157, 158 Matoušek, J., ix, 69, 71, 230
Leng, G., 257 Maximal plane graph, 54
Length of a cycle, 51 Maximum degree, 50
Length of a path, 51 Maximum independent set, 70
Leone, F.C., 198 McCarthy, J.E., 232
Leonhardt, U., 24 McCleary, J., 118
332 Index

McKean, H., 198 Muralidharan, K., 199

McKee, T.A., 197 Musin, O.R., 19, 21, 138, 140, 141
McLaughlin, S., 141
McMorris, F.R., 197
McMullen, P., 228, 234 N
Mean width, 233 Namikawa, Y., 155
Mednykh, A., 277 Napoleon’s theorem, 232
Melbourne, J., 227 Narváez, D., 229
Melluish, R.K., 91 Naszódi, M., 73
Melzak, Z.A., 90 Naumann, H., 228
Menger, K., 212, 251 Neighbor, 50, 181
Mercator, 22 Neighborhood, 50
Merzbacher, M., 233 Nelsen, R., 18, 274
Mesiar, R., 200 Neuberg-Pedoe inequality, 118
Metric space, 201 Neutsch, W., 117
Meyer, W., 119 Newton number, 137
Michelen, M., 139 Nguyen, B., 255
Michiwaki, Y., 274 Non-Euclidean geometries, 16, 25
Miernowski, A., 233 Non-facial 3-cycle, 126
Mikaiylova, L.A., 277 Norm, 214
Mikami, Y., 276 Normal family, 31
Miles, R.E., 171 Northshield, S., 44, 45
Milićević, M., 19 Null circle, 261
Miller, S.D., 139 Nushizumi, Y., 43
Milman, R.S., 22
Minimal area simplices, 233
Minimal cut set, 63 O
Minimal enclosing ellipse, 156 O’Connor, C.P.J., 26
Minimum degree, 50 OCR theorem, 58
Mirman, B., 28 Odd-cycle, 51
Mitchell, L., 70 Odehnal, B., 28
Mitrinović, D.S., 230 Ogasawara, H„ 199
Mittelmann, H., 138 Ogilvy, C.S., 18, 20, 46
Mnëv, N.E., 72 Oh, H., 46
Mo, T.C., 199 Ohta, S., 122
Möbius geometry, 16, 19, 22, 23, 25, 275 Olivares, J.G., 115
Möbius invariant, 263 Oliveros, D., 185
Mohseni Takaloo, S., 25, 48 Olver, P.J., 259
Molchanov, S., 44 Onn, S., 72
Molnár, J., 90 Oort, F., 26
Mondal, B., 254 O’Reilly, E., 175
Montejano, L., 185 Oriented matroid, 72
Moody, J., 227 Origami, 117
Moran, P.A.P., 175 Orthocentric simplices, 231, 235
Morgan, J., 120 Orthogonal-circle-representation, 58
Morita, K., 27, 47 Orthogonally crossing, 3, 243
Moritsugu, S., 44 Orthogonal-sphere system, 243
Moser’s paradox, 215 Ortuzar, A., 259
Moser, L., 216 Oshiro, A, 21
Moser, W., 69, 139, 259 Ostermann, A., 117
Mozgawa, W., 28, 233 Ostrovskii, M.I., 228
Mucherino, A., 252 Ouyang, Y., 200
Multisection lasers, 26 Oyama, M., 274
Index 333

P Popko, E.S., 25, 90, 94

Pach, J., 66, 69, 139, 197, 259 Porta, J.M., 257
Pachner, U., 72 Posamentier, A.S., 17, 18, 27, 44, 46, 274
Paluszny, M., 47 Power, 1
Pambuccian, V., 18, 106, 119, 158 Prakasa Rao, B.S.L., 200
Panin, A., 48 Prasolov, V.V., 22, 27, 89
Panizzut, M., 73 Problem of Apollonius, 43, 46
Paouris, G., 229 Projections of cubes, 228
Papadopoulos, A., vii, 89, 93, 115, 146, 157, Proper diagonal, 124
158 Properly tangent, 262
Parameter, 143 Proper subgraph, 50
Parker, G.D., 22 Propp, J.G., 139
Parks, H., 232 Proskurowski, W., 93
Path, 51 Protasov, V.Yu., 19
Peaucellier linkage, 5 Protti, F., 70
Peaucellier’s inversor, 17 Ptolemy, 21
Pecarić, J.E., 230 Ptolemy’s theorem, 249, 273, 274
Pech, P., 275 Purisima, E.O., 258
Pecker, D., 28
Pedoe, D., vii, 16, 18–20, 23, 37, 43, 46, 47,
274, 276 Q
Pekonen, O., 158 Quadrangulation, 54
Peng, Y., 90 Quadrics, 23
Penny graph, 69 Quartet, 201
Percy, A., 275
Perel’man, G.Y., 122
Perfect, H., 16, 274 R
Perimeter deviation, 98 Rademacher, H., 5, 71
Persson, U., 158 Radial edges, 65
Petersen, P., 119 Radial graph, 65
Petrera, M., 118 Radical axis, 56
Pfender, F., 97, 138 Radnović, M., 27, 28
Pfiefer, R.E., 174 Rambau, J., 229
Philbin, T., 24 Ramiŕez, G.N., 17
Pinasco, D., 233 Randić, M., 69
Pippert, R.E., 155 Random diameter, 160
Pisanski, T., 69 Random great circle, 160
Pivovarov, P., 174, 229 Random point, 159, 171
Pizzetti, P., 119 Random polygons, 170
Planar, 202 Random polyhedral cones, 172
Planar graph, 52 Random polytopes, 170, 172, 174
Plane graph, 52 Random spherical cap, 173, 190
Plofker, K., 115 Random spherical triangle, 160
Polar set, 85 Ranucci, E.R., 226
Polar triangle, 87 Rao, M., 236
Pole, 79 Rashed, R., 115
Polya, G., 6 Rataj, J., 178
Polyhedral graphs, 71 Rauch, H.E., 119
Pompeiu property, 97 Raven, D.W., 26
Pompeiu’s theorem, 236 Realization, 50
Poncelet polygon, 26 Realization spaces for polytopes, 72
Poncelet’s porism, 7, 12, 26, 28 Reduced bodies, 90, 98
Poncelet triangle, 28 Reed, N., 17
Popescu, C., 275 Regular n-simplex, 214
334 Index

Regular spherical polygon, 97, 98, 150 Scheffers, G., 24

Reiter, H., 44 Scheinerman, E.R., 63, 69
Reitzner, M., 172 Scheraga, H.A., 258
Ren, D.L., 176 Scherer, K., 117
Replicating simplices, 234 Schilling, F., 155
Representation of a graph, 51 Schmidt, H., 17, 22, 46
Reuleaux d-triangle, 184 Schneider, R., 90, 171, 173, 174, 176–178, 231,
Reuschel, A., 277 232
Reventós Tarrida, A., 17, 19 Schnell, U., 228
Reznik, D., 28 Schoenberg, I.J., 7, 27, 258
Reznikov, A.B., 174 Schramm, O., 70, 139
Richter-Gebert, J., 25, 71, 72 Schreck, P., 259
Riede, H., 227 Schreiner, M., 91
Riemann’s number ball, 23 Schröcker, H.-P., 155
Rigidity, 253, 259 Schröder, E., 22, 24, 92, 93
Ringel, G., 69 Schroeder, V., 121
Rivière, A., 259 Schroth, A.E., 47
Roanes-Lozano, E., 19, 21 Schulz, C., 72
Robbins, H., 16 Schuster, R., 178
Roberto, C., 227 Schütte, K., 97, 134, 136
Roberts, F.S., 196 Schwartz, R.E., 20, 26, 44, 255
Roché, E., 230 Schwarz, D., 275
Rogers, D.G., 275 Schwerdtfeger, H., 16, 23
Rollett, A.P., 17 Seberry, J., 229
Rollins, W.E., 226 Seidel, J.J., 138, 256
Ronda, J., 28 Semialgebraic sets, 72
Rosen, D.A., 175 Semi-metric space, 239
Rosenfeld, B.A., 15, 22, 89, 95, 117 Semi-perimeter, 207
Roshdi, R., vii Sen, H., 254
Rota, G.-C., 176 Seneta, E., 199
Rother, W., 178 Separating 4-cycle, 55
Rothman, T., vii Serré, P., 259
Rothschild, B.L., 243, 258 Shaqaqha, S., 255
Ryabogin, D., 227 Sharygin, I.F., 44
Ryan, P.J., 24, 70 Shephard, G.C., 95
Rylov, Yu.A., 257 Shi, N., 95
Rzymowski,W., 233 Shiga, K., 27
Shin, S.-Y., 98
Shiohama, K., 120, 122
S Shor, P/ W., 229
Saban, G., 175 Shtogrin, M.I., 44
Sachs, H., 69–71 Sigl, R., 116
Safe, M., 197 Sigler, A., 275
Saff, E.B., 174 Signed radius, 245
Sahasrabudhe, J., 139 Silverberg, J.S., 115
Sakano, Y., 96 Simon, M., 230
Saks, M.E., 227 Simonić, A., 25, 146, 157
Santaló, L.A., 170, 175, 176 Simple polygonal curve, 80
Santaló’s chord theorem, 169 Simplex, 214, 231, 235, 253, 254
Santos, F., 229 Simplex centers, 232, 235
Sarnak, P., 46 Singh, A.N., 115
Satyanarayana, K., 230 Sippl, M.J., 258
Saw, J.G., 199 Skau, C., 27
Schacht , C., 18 Skopets, Z.A., 46, 276
Index 335

Skoruppa, N.-P., 138 Su, W., 158

Sloan, I.H., 174 Su, X., 122
Sloane, N.J.A., 137 Subgraph, 50
Slutskiy, D., 118 Subspace, 201
Small circle, 75 Suceava, B., 273
Smart, W.M., 89 Sugimoto, T., 173
Smeltzer, D.L., 19 Sulanke, R., 176
Smith, G.C., 275 Super-Apollonian group, 45
Smith, J.D., 274 Surface area, 178, 237, 238
Soddy, Frederick, 20, 37 Suris, Y.B., 118
Soddy’s formula, 39, 43 Svrtan, D., 18
Soddy’s hexlet, 10, 21, 35 Swanepoel, K.J., 7, 137
Soland, C., 257 Swart, D., 25
Solomon, H., 176 Sylvester’s four-point-problem, 163, 174
Soltan, V., 233 Szabó, L., 140
Sombra, M., 254 Szenthe, J., 258
Sönmez, N., 97 Szpiro, G.D., 140
Sphere packing problems, 137
Spherical, 202
Spherical n-gon, 80 T
Spherical astronomy, 89 Tabachnikov, S., 20, 26–28, 44
Spherical cap, 75 Taherian, S.Gh., 25, 48
Spherical codes, 19, 138 Takayama, Y, 73
Spherical cosine law, 89, 103, 114 Tammes problem, 140
Spherical Delaunay triangulations, 136 Tanaka, M., 121, 122, 274
Spherical ellipse, 143, 151, 155 Tanemura, M., 173
Spherical excess, 81 Tangent, 3
Spherical polygon, 80, 94 Tangent-sphere system, 244
Spherical rotors, 98 Tarasov, A.S., 140
Spherical sine law, 89, 104, 114 Taylor, J.H., 276
Spherical tilings, 94–96 Teixeira, R., 229, 230
Spherical triangle, 80 Thäle, C., 172, 175
Spherical trigonometry, 90, 114 Thébault, V., 230
Sphero-conics, 155 Thomas, F., 257
Spinrad, J.P., 197 Three-circles theorem, 5
Spirova, M., 6, 7 Three-pointed star, 205
Springborn, B.A„ 73 Thurston, W.P., 66, 69
Squared-distance matrix, 240 Tian, G., 120
Stén, J., 157, 158 Tikhomirov, V.M., 22, 27, 89
Steiner chain of circles, 18, 20, 21 Titeica, G. (Ţiţeica, G.), 5
Steiner cycle, 8, 18, 31 Titeica-Johnson theorem, 5
Steiner’s porism, 8, 18 Toeplitz, O., 5
Steiner, J., 157 Tong, Y.L., 199
Steinert, K.-G., 116 Toponogov’s theorem, 118
Steinitz’ theorem, 67, 71 Toponogov’s triangle comparison theorem, 106
Steinitz, E., 67, 71 Toponogov, V.A., 106, 118, 119
Stereographic projection, 11, 21, 99, 118, 145 Torrence, B., 25
Sterz, U., 47 Trapp, K., 92
Stoka, M., 176 Triangulations of cubes, 229
Stone, C., 227 Triangle comparison theorem for spheres, 105,
Straus, E.G., 243, 258 118
Strong dodecahedral conjecture, 141 Triangle free penny graphs, 70
Stupel, M., 275 Triangular face, triangle, 54
Sturmfels, B., 72 Triangulation of a quadrilateral, 126
336 Index

Triangulations of cubes, 229 Weisstein, E.W., 276

Trio, 201 Wells, D., 5, 18, 20
Tropfke, J., 116 Wendel’s theorem, 164
Tsiutsiurupa, I., 227, 228 Wendel, J.G., 164
Tuckwell, H.C., 198 Wenninger, M.J., 94
Tupan, A., 44 Wenzel, W., 234
Two ears theorem, 147 Whittlesey, M.A., vii, 89, 94
Wieacker, J.A., 177
Wieczorek, S., 26
U Wiegert, P.A., 158
Ueno, K., 27 Wilker, J.B., 17, 21, 25, 43, 47, 231
Ueno, Y., 96 Wilks, A.R., 44, 45
Umezawa, R., 98 Wilson, P.M.H., 23, 89, 94, 117
Ungar, A.A„ 253 Wimmer, L., 96
Unispherical, 205 Winkler, P., 111
Unit-distance graph, 49 Wintgen, P., 176
Urmaev, N.A., 92 Womersley, R.S., 174
Wu, S., 6
Wünsch, V., 92
V
Valdés, A., 28
Valentine, J.E., 212, 274 X
Vallès, J., 28 Xie, Y.M., 96
Vallentin, F., 138 Xu, X., 259
Van Brummelen, G., vii, 23, 90, 94, 101, 116,
117
Van der Waerden, B.L., 97, 134, 136 Y
Van Gruting, C.J., 274, 277 Yaglom, I.M., 16–18, 22, 46, 116, 117, 230,
Veljan, D., 18 276
Vertex, 49, 80 Yan, C.H., 45
Vertex embeddings of simplices in cubes, 229 Yan, M., 95, 96
Viazovska, M.S., 139 Yang, L., 118, 254, 256
Volenec, V., 230, 258, 277 Yang, M.C., 199
Volume, 213, 216, 236, 237 Yasui, T., 98
Volume of a ball, 216, 237 Ye. Q.-Z., 158
Vu, T.Khoi, 97 Yiu, P., 19
Yuan, J., 233

W
Wagner, K., 92 Z
Wang, E., 96 Zach, D., 227
Wang, Y., 122 Zacharias, M., 116, 277
Wang, Y.G., 174 Zähle, M., 173, 178
Wanner, G., 117 Zamfirescu, T., 106, 119
Wasan, 19, 44, 276 Zeng, Z., 256
Weber, M.J., 156 Zhang, G., 177
Wei, G., 121 Zhang, J.Z., 118, 254
Weil, W., 171, 176–178 Zhao, L., 233
Weis, J.A., 179 Zhou, L., 200
Weise, K.H., 24, 91 Zhukova, A., 158
Weiss, A.I., 44 Ziegler, G.M., 68, 69, 71, 72, 97, 136, 138, 140
Weiss, G., 7, 226 Zinn, J., 229
Weiss, M.L., 230 Zong, C., 136–139, 226, 229
Weissbach, B., 228, 232 Zvavitch, A., 227