Discovering Poncelet Invariants

in the Plane

Ronaldo A. Garcia
Dan S. Reznik

a
impa
 Ronaldo A. Garcia
Dan S. Reznik

Discovering Poncelet Invariants

in the Plane
Discovering Poncelet invariants in the plane
Primeira impressão, julho de 2021
Copyright © 2021 Ronaldo A. Garcia e Dan S. Reznik.
Publicado no Brasil / Published in Brazil.

ISBN 978-65-89124-43-6
MSC (2020) Primary: 37M05, Secondary: 14H70, 37C83, 51M15, 51N20, 14Q05

Coordenação Geral Carolina Araujo

Produção Books in Bytes Capa Izabella Freitas & Jack Salvador

Realização da Editora do IMPA

IMPA [Link]
Estrada Dona Castorina, 110 editora@[Link]
Jardim Botânico
22460-320 Rio de Janeiro RJ
 Preface

Since the discovery of Poncelet’s porism in the 1810s, a steady stream of proofs has
been put forth, drawing upon the ever-evolving language and abstraction of math-
ematics. These started in the 19th century with Poncelet’s own synthetic/analytic
proof, passing through Jacobi’s treatment with elliptic functions, all the way to our
era where the phenomenon is understood on an abstract torus. See Del Centina
(2016a,b) for the historical background.
Indeed, for the past 200 years, the focus has been on refining proofs and un-
derstanding ramifications of the porism with respect to other areas of Mathematics.
One consequence has been that the ambient, dynamic planar geometry of Poncelet
polygons has been mostly unexplored.
In this book we take this less-traveled road, i.e., utilizing tools of interactive
simulation, we set off to discover curious phenomena manifested by Poncelet poly-
gons in the Euclidean plane. These include invariant metric quantities, the shape
of loci of certain points, etc. Luckily, we have stumbled upon many interesting
phenomena. Whenever possible, we illustrate the results with pictures and/or an-
imations. To further engage the reader, we propose many exercises and research
questions.
This research started in 2011 following lively conversations with Jair Koiller
about the path of light rays in an ellipse. This resulted in several Mathematica
simulations and a few videos uploaded to YouTube. After an 8-year hiatus, we
resumed the work in early 2019 following a few very auspicious events: (i) one of
the authors learned other mathematicians had watched our videos and published
proofs of phenomena therein, (ii) Sergei Tabachnikov’s invitation for us to pub-
lish an article (jointly with Jair Koiller) in the Mathematical Intelligencer, and (ii)
our expository talk at IMPA’s 32nd colloquium of Brazilian mathematics (see this
video). Following this process our research sped up and we ended up producing
dozens of papers and hundreds of experimental videos, which form the basis of
this book.
We are indebted to several mathematicians and friends who have answered
hundreds of our emails, and shared with us much needed insights: Arseniy Akopyan,
Michael Bialy, Ana Chávez-Caliz, Mário Jorge Carneiro, Manish Chakrabarti,
Ethan Cotterill, Marcos Craizer, Iverton Darlan, Carlos Esperança, Robert Ferréol,
Corentin Fierobe, Sergey Galkin, Liliana Gheorghe, Bernard Gibert, João Gondim,
Darij Grinberg, Mark Helman, Daniel Jaud, Clark Kimberling, Jair Koiller, Do-
minique Laurain, Nicholas McDonald, Peter Moses, Oliver Nash, Boris Odehnal,
Matt Perlmutter, Pedro Roitman, Olga Romaskevich, Richard Schwartz, Hellmuth
Stachel, Sergei Tabachnikov, Israel Vainsencher, Daniel Weller, Jorge Zubelli, and
others.
We also thank IMPA for the opportunity to publish this book supporting our
course in the 33rd Colloquium of Brazilian Mathematics (2021), and Paulo Ney
de Souza for his encouragement, and editorial support.

Ronaldo Garcia & Dan Reznik

Goiânia & Rio de Janeiro, Brazil

July, 2021
 Contents

1 Introduction 1
1.1 Poncelet preliminaries . . . . . . . . . . . . . . . . . . . . . . . . 1
1.2 The elliptic billiard . . . . . . . . . . . . . . . . . . . . . . . . . 3
1.3 Focusing on 3-periodics . . . . . . . . . . . . . . . . . . . . . . . 3
1.4 Asking simple questions . . . . . . . . . . . . . . . . . . . . . . 5
1.5 On to more confocal results . . . . . . . . . . . . . . . . . . . . . 6
1.6 Branching out to non-confocal families . . . . . . . . . . . . . . . 7
1.7 Analysis methods . . . . . . . . . . . . . . . . . . . . . . . . . . 7
1.8 Related work . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
1.9 Book organization . . . . . . . . . . . . . . . . . . . . . . . . . . 9

2 Confocal Pair 11
2.1 Preliminaries . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
2.2 Caustic semiaxes . . . . . . . . . . . . . . . . . . . . . . . . . . 12
2.3 Incenter and excenter loci . . . . . . . . . . . . . . . . . . . . . . 13
2.4 A stationary point . . . . . . . . . . . . . . . . . . . . . . . . . . 15
2.5 Conserved quantities . . . . . . . . . . . . . . . . . . . . . . . . 17
2.6 An interpretation for Darboux’s constant . . . . . . . . . . . . . . 20
2.7 Confocal vertex parametrization . . . . . . . . . . . . . . . . . . 20
2.7.1 Standard . . . . . . . . . . . . . . . . . . . . . . . . . . 20
2.7.2 Jacobi’s universal measure . . . . . . . . . . . . . . . . . 21
2.8 Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
2.9 Research questions . . . . . . . . . . . . . . . . . . . . . . . . . 27
3 Concentric, Axis-Parallel (CAP) 28
3.1 Excentral family . . . . . . . . . . . . . . . . . . . . . . . . . . . 28
3.2 Incircle family . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32
3.2.1 Confocal affine image . . . . . . . . . . . . . . . . . . . 33
3.3 Circumcircle family . . . . . . . . . . . . . . . . . . . . . . . . . 34
3.3.1 Confocal affine image . . . . . . . . . . . . . . . . . . . 36
3.4 Homothetic family . . . . . . . . . . . . . . . . . . . . . . . . . 37
3.5 Dual family . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39
3.6 Vertex parametrization for a generic CAP pair . . . . . . . . . . . 41
3.7 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42
3.8 Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44
3.9 Research questions . . . . . . . . . . . . . . . . . . . . . . . . . 46

4 Non-concentric, Axis-Parallel (NCAP) 48

4.1 Poristic family (Bicentric triangles) . . . . . . . . . . . . . . . . . 48
4.2 Poristic excentrals . . . . . . . . . . . . . . . . . . . . . . . . . . 55
4.3 The Brocard porism . . . . . . . . . . . . . . . . . . . . . . . . . 59
4.3.1 A digression: equilateral isodynamic pedals . . . . . . . . 67
4.4 Vertex parametrization . . . . . . . . . . . . . . . . . . . . . . . 69
4.4.1 Poristic family . . . . . . . . . . . . . . . . . . . . . . . 69
4.4.2 Poristic excentrals . . . . . . . . . . . . . . . . . . . . . 70
4.4.3 Brocard porism . . . . . . . . . . . . . . . . . . . . . . . 70
4.5 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 71
4.6 Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 72
4.7 Research questions . . . . . . . . . . . . . . . . . . . . . . . . . 73

5 Confocal Loci 75
5.1 Kimberling centers with elliptic loci . . . . . . . . . . . . . . . . 76
5.2 When billiard 3-periodics are obtuse . . . . . . . . . . . . . . . . 78
5.3 Quartic locus of the symmedian point X6 . . . . . . . . . . . . . 79
5.4 Feuerbach point and its anticomplement . . . . . . . . . . . . . . 82
5.5 A locus with singularities . . . . . . . . . . . . . . . . . . . . . . 83
5.6 A self-intersecting locus . . . . . . . . . . . . . . . . . . . . . . . 85
5.7 A non-compact locus . . . . . . . . . . . . . . . . . . . . . . . . 85
5.8 A golden locus . . . . . . . . . . . . . . . . . . . . . . . . . . . . 88
5.9 When the billiard is swept non-monotonically . . . . . . . . . . . 88
5.10 The dance of the swans . . . . . . . . . . . . . . . . . . . . . . . 90
5.11 Locus of vertices of derived triangles . . . . . . . . . . . . . . . . 93
 5.12 Locus triple winding . . . . . . . . . . . . . . . . . . . . . . . . 93
5.13 Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 98
5.14 Research questions . . . . . . . . . . . . . . . . . . . . . . . . . 101

6 Loci in CAP Pairs 102

6.1 Incircle family . . . . . . . . . . . . . . . . . . . . . . . . . . . . 103
6.2 Circumcircle family . . . . . . . . . . . . . . . . . . . . . . . . . 104
6.3 Homothetic family . . . . . . . . . . . . . . . . . . . . . . . . . 106
6.3.1 Four circular loci . . . . . . . . . . . . . . . . . . . . . . 107
6.3.2 Loci of the Brocard points . . . . . . . . . . . . . . . . . 109
6.3.3 First Brocard triangle: vertex locus . . . . . . . . . . . . 110
6.3.4 Loci of Fermat and isodynamic equilaterals . . . . . . . . 111
6.4 Dual family . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 112
6.5 Excentral family . . . . . . . . . . . . . . . . . . . . . . . . . . . 112
6.6 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 112
6.6.1 Loci types, CAP families . . . . . . . . . . . . . . . . . . 116
6.6.2 Loci types, NCAP families . . . . . . . . . . . . . . . . . 117
6.7 Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 118
6.8 Research questions . . . . . . . . . . . . . . . . . . . . . . . . . 120

7 Analyzing Loci 121

7.1 When are loci algebraic? . . . . . . . . . . . . . . . . . . . . . . 122
7.2 Review: Blaschke products . . . . . . . . . . . . . . . . . . . . . 124
7.3 Locus of the incenter in a generic pair . . . . . . . . . . . . . . . 127
7.4 Loci in generic nested ellipses . . . . . . . . . . . . . . . . . . . 131
7.5 Circular loci in the circumcircle family . . . . . . . . . . . . . . . 136
7.6 Elliptic loci in the confocal pair . . . . . . . . . . . . . . . . . . . 140
7.7 Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 145
7.8 Research questions . . . . . . . . . . . . . . . . . . . . . . . . . 146

8 The Focus-Inversive Family 147

8.1 Non-Ponceletian . . . . . . . . . . . . . . . . . . . . . . . . . . . 147
8.2 A stationary point . . . . . . . . . . . . . . . . . . . . . . . . . . 148
8.3 Billiard-like invariants . . . . . . . . . . . . . . . . . . . . . . . 150
8.4 The rotating billiard table . . . . . . . . . . . . . . . . . . . . . . 150
8.5 Invariant area product . . . . . . . . . . . . . . . . . . . . . . . . 151
8.6 Circular loci galore! . . . . . . . . . . . . . . . . . . . . . . . . . 154
 8.7 A rule for circular loci? . . . . . . . . . . . . . . . . . . . . . . . 155
8.7.1 Centroidal loci: a tale of three circles . . . . . . . . . . . 156
8.8 A focus-inversive Doppelgänger . . . . . . . . . . . . . . . . . . 158
8.9 Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 161
8.10 Research questions . . . . . . . . . . . . . . . . . . . . . . . . . 162

9 A Locus Visualization App 163

9.1 Main ellipse and animation controls . . . . . . . . . . . . . . . . 164
9.1.1 Convenience animation controls . . . . . . . . . . . . . . 165
9.2 Channel controls . . . . . . . . . . . . . . . . . . . . . . . . . . . 166
9.3 Choosing a triangle family . . . . . . . . . . . . . . . . . . . . . 166
9.3.1 Poncelet families . . . . . . . . . . . . . . . . . . . . . . 166
9.3.2 Ellipse “mounted” . . . . . . . . . . . . . . . . . . . . . 168
9.4 Triangle type . . . . . . . . . . . . . . . . . . . . . . . . . . . . 170
9.4.1 Standard triangles . . . . . . . . . . . . . . . . . . . . . . 170
9.4.2 Exotic triangles . . . . . . . . . . . . . . . . . . . . . . . 170
9.4.3 Inversive triangles . . . . . . . . . . . . . . . . . . . . . 171
9.5 Locus type . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 172
9.5.1 Centers and vertices . . . . . . . . . . . . . . . . . . . . 172
9.5.2 Envelopes . . . . . . . . . . . . . . . . . . . . . . . . . . 173
9.5.3 Bicentric pairs . . . . . . . . . . . . . . . . . . . . . . . 173
9.6 Triangle center . . . . . . . . . . . . . . . . . . . . . . . . . . . . 173
9.7 Cevians, pedals, & Co. . . . . . . . . . . . . . . . . . . . . . . . 174
9.7.1 Traditional . . . . . . . . . . . . . . . . . . . . . . . . . 174
9.7.2 Inversive . . . . . . . . . . . . . . . . . . . . . . . . . . 174
9.7.3 Reflexive . . . . . . . . . . . . . . . . . . . . . . . . . . 176
9.7.4 Triangulated . . . . . . . . . . . . . . . . . . . . . . . . 176
9.8 Notable circles . . . . . . . . . . . . . . . . . . . . . . . . . . . . 176
9.8.1 Ellipse-affixed circles . . . . . . . . . . . . . . . . . . . . 177
9.8.2 Central circles . . . . . . . . . . . . . . . . . . . . . . . 177
9.9 Inversive transformations with respect to a circle . . . . . . . . . . 179
9.10 Conic and invariant detection . . . . . . . . . . . . . . . . . . . . 180
9.10.1 Curve type . . . . . . . . . . . . . . . . . . . . . . . . . 180
9.10.2 Detection of metric invariants . . . . . . . . . . . . . . . 180
9.11 The tandem bar . . . . . . . . . . . . . . . . . . . . . . . . . . . 182
9.12 Odds & ends . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 184
9.12.1 Ellipse, locus tange, and animation background . . . . . . 184
9.12.2 Resetting the UI and centering the animation . . . . . . . 184
 9.12.3 Setting the locus color . . . . . . . . . . . . . . . . . . . 187
9.12.4 Collapsing the locus control area . . . . . . . . . . . . . . 187
9.13 Artsy loci . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 187
9.14 Sharing and exporting . . . . . . . . . . . . . . . . . . . . . . . . 188
9.15 Jukebox playback . . . . . . . . . . . . . . . . . . . . . . . . . . 191

A Notes in Triangle Geometry 193

A.1 Trilinear coordinates . . . . . . . . . . . . . . . . . . . . . . . . 193
A.2 More calculations with distances . . . . . . . . . . . . . . . . . . 194
A.3 Barycentric coordinates . . . . . . . . . . . . . . . . . . . . . . . 196
A.4 Conversion to and from cartesians . . . . . . . . . . . . . . . . . 196
A.5 Triangle centers . . . . . . . . . . . . . . . . . . . . . . . . . . . 197
A.6 Selected triangle centers . . . . . . . . . . . . . . . . . . . . . . . 197
A.7 Some derived triangles . . . . . . . . . . . . . . . . . . . . . . . 200
A.8 The (first) Brocard triangle . . . . . . . . . . . . . . . . . . . . . 204
A.9 Pedal and antipedal triangles . . . . . . . . . . . . . . . . . . . . 205
A.10 Cevian triangle . . . . . . . . . . . . . . . . . . . . . . . . . . . 208
A.11 Perspective triangles . . . . . . . . . . . . . . . . . . . . . . . . . 208
A.12 Polar triangle . . . . . . . . . . . . . . . . . . . . . . . . . . . . 208
A.13 Circumconic . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 209
A.14 Inconic . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 210
A.15 Brocard inellipse . . . . . . . . . . . . . . . . . . . . . . . . . . 210
A.16 Ceva conjugate . . . . . . . . . . . . . . . . . . . . . . . . . . . 211
A.17 Isogonal conjugation . . . . . . . . . . . . . . . . . . . . . . . . 211
A.18 Isotomic conjugation . . . . . . . . . . . . . . . . . . . . . . . . 213
A.19 The Euler line . . . . . . . . . . . . . . . . . . . . . . . . . . . . 215
A.20 Circumconic and inconic . . . . . . . . . . . . . . . . . . . . . . 216
A.21 Billiard notes . . . . . . . . . . . . . . . . . . . . . . . . . . . . 217
A.22 Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 218

B Jacobi Elliptic Functions 224

B.1 Jacobi elliptic integral and inverse . . . . . . . . . . . . . . . . . 224
B.2 Jacobi elliptic functions . . . . . . . . . . . . . . . . . . . . . . . 224
B.3 Basic identities . . . . . . . . . . . . . . . . . . . . . . . . . . . 225
B.4 Connection with differential equations . . . . . . . . . . . . . . . 226
B.5 Inverse Jacobi elliptic functions . . . . . . . . . . . . . . . . . . . 226
B.6 Complex plane extension . . . . . . . . . . . . . . . . . . . . . . 227
C Ellipse-Mounted Brocard loci 228
C.1 Circular sweep, one vertex at center . . . . . . . . . . . . . . . . 228
C.2 Circular sweep, two vertices at 90-degrees . . . . . . . . . . . . . 229
C.3 Circular sweep, antipodal vertices . . . . . . . . . . . . . . . . . 229
C.4 Ellipse sweep, two vertices at major endpoints . . . . . . . . . . . 230
C.5 Elliptic sweep, vertices on major axis . . . . . . . . . . . . . . . . 230

Bibliography 235

Index 240

Glossary 244
1 Introduction

1.1 Poncelet preliminaries

Poncelet’s closure theorem is illustrated in Figure 1.1. It is based on a simple
geometric iteration. Given two nested ellipses1 E and Ec , pick a point P1 on the
boundary of E. Let P2 be where a ray shot from P1 along one of the tangents to Ec
meets E again. Repeat this from P2 , yielding P3 , etc. This produces a piecewise-
linear Poncelet trajectory.
For most choices of .E; Ec /, the trajectory will never close, i.e., it will never
meet P1 again. In fact, it will fill a region between the two conics. However, for
certain choices 2 , the trajectory will indeed close. Let N , and integer greater than
2, be the number of steps required for P1 to be met again. We call such polygonal
trajectories “N -periodic”.
Still referring to Figure 1.1, Poncelet’s theorem states that if a trajectory de-
parting from some point P1 on E closes after N steps, then a porism is triggered
which prescribes a 1d family of N -gons: a trajectory departing from any other
point on the boundary of E will also close in N steps. We say such a pair “admits”
a 1d family of N -periodic trajectories.
1 The theorem is projective, i.e., it works for any pair of conics, nested or not.
2 Those which satisfy Cayley’s conditions, see Dragović and Radnović (2011).
2 1. Introduction

Figure 1.1: Top left: 3 Poncelet iterations within a pair of ellipses in general posi-
tion; their centers are labeled O and Oc , respectively. Top right: 5 more iterations
executed (starting at P4 ), showing the trajectory is not likely to close. Bottom: a
new ellipse pair for which an iteration departing from P1 closes after 7 steps (blue
polygon). Poncelet’s porism guarantees that if the iteration were to start anywhere
else on the outer ellipse, e.g., P10 , it will also yield a closed, 7-gon (dashed red).
Video, Live
1.2. The elliptic billiard 3

Poncelet’s theorem has been widely studied for over 2 centuries. It is regarded
as a fundamental result in algebraic geometry, see the surveys in Bos, Kers, and
Raven (1987), Del Centina (2016b), and Dragović and Radnović (2014).

1.2 The elliptic billiard

A special case with remarkable properties is when E and Ec are constrained to be
confocal, i.e., to have coinciding foci, see Figure 1.2. As a corollary to Graves’ the-
orem described in Tabachnikov (2005), consecutive segments of a trajectory (open
or closed) are bisected by the normal to E. Thus, the iteration can be regarded as
the path of a small billiard ball undergoing elastic collisions against the boundary
of E. For this reason, the confocal pair is also termed the elliptic billiard. An
elliptic billiard trajectory can also be interpreted as the path of a light ray reflect-
ing in a mirrored elliptic cavity. Since all rays are tangent to an internal, virtual
ellipse, one borrows a term from optics and calls the latter the caustic3 . From a
dynamical systems’ point of view, the path of the billiard ball is constrained by
two integrals of motion, namely, linear and angular momentum. This renders the
system integrable: the trajectory can be fully computed from initial conditions.
For a more formal definition, see Tabachnikov (ibid.). Indeed, the elliptic billiard
is conjectured as the only integrable planar billiard, see Kaloshin and Sorrentino
(2018).
Assume a pair of confocal ellipses has been chosen which admits an N -periodic
family. A first remarkable property is that in the confocal case, the family con-
serves perimeter, i.e., perimeter is invariant. This is all the more impressive4 given
the non-linearities constraining the dynamic geometry of billiard N -periodics.

1.3 Focusing on 3-periodics

Though invariant perimeter is a prime example of a clearly inspectable metric phe-
nomenon, little attention has been paid as to whether confocal (or other Poncelet
families) may manifest any additional, interesting, euclidean phenomena. Here,
partly owing to our limitations, we did get a head start exploring that less-traveled
road, i.e., using simulation to probe the dynamic geometry of Poncelet families for
any salient Euclidean properties.
3 Inoptics, this term refers to the envelope of rays reflected from a curved mirror.
4 Infact, this result is rooted on the fact that billiard trajectories are extrema of the perimeter
function, see Tabachnikov (2005).
4 1. Introduction

Figure 1.2: Top left: A confocal pair of ellipses; shown are the first four Pon-
celet iterations departing from P1 . Graves’ theorem guarantees that each two con-
secutive segments are bisected by the ellipse normal nO i . Top right: A closing
3-periodic trajectory. Bottom left: The first 50 segments of a non-periodic trajec-
tory starting at P1 and directed toward P2 , notice P1 P2 does not pass between the
two foci. Bottom right: a confocal pair comprising an ellipse and a hyperbola.
All trajectory segments pass between the foci and are tangent to the hyperbola.
Early Video 1, Video 2, Video 3
1.4. Asking simple questions 5

P''
P1 P'
P
P3 O X1''
X1
X1'
X1

P2

Figure 1.3: Left: An N D 3 orbit. Its incenter X1 is where angular bisectors

(black arrows) concur. Right: Three billiard 3-periodics tangent to a confocal
caustic (brown). Over positions P; P 0 ; P 00 of a first vertex. Also shown are the
corresponding incenters X1 ; X10 ; X100 . Video 1, Video2

At a first moment, we further restricted our search to N D 3 families only, con-

focal or not. Indeed, this book is a tale of the offspring – properties and invariants
– of the unlikely marriage of triangle geometry and Poncelet constraints.
The literature covering triangle geometry is extensive, though have often re-
ferred to Gallatly (1914) and Johnson (1960). For informal use we have relied on
Weisstein (2009, 2019).

1.4 Asking simple questions

An early observation is illustrated in Figure 1.3. Consider a confocal ellipse pair
admitting a 3-periodic family. These are triangles whose internal angles are bi-
sected by ellipse normals (see above). Known to the Greeks was the fact that the
bisectors of a triangle concur (i.e., they meet) at a point known as the incenter.
Immediately one is compelled to ask: “what could be the path, or locus, of the
incenter, over the family of triangles in the confocal pair?” As it turns out, and
despite all non-linearities, it is pure ellipse5 .
A second observation is illustrated in Figure 1.4. Directly associated with both
a triangle and its incenter is the incircle, the unique circle tangent to each side of a
triangle. The points of tangency are known as the intouchpoints. Again a natural
question is “what is the locus of the intouchpoints over billiard 3-periodics?” As
5 Welater realized this is a very rare occurrence. The space of ellipse pair choices is 5d. That of
confocal pairs 1d. There is growing evidence the locus of the incenter can only be an ellipse if the
pair is confocal.
6 1. Introduction

Figure 1.4: An N D 3 orbit (blue), its Incircle (transparent green), Incenter (green
dot) and Intouch Points (brown dots). Over the N D 3 family, the Incenter locus is
a perfect ellipse (green), while the Intouchpoints produce a self-intersecting sextic
(dashed brown). Video, Live

shown in the figure, it turns out this locus is not as simple: it is a two-lobed, self-
intersecting curve.
As one plays with other locus phenomena, one is quickly led to ask “what
determines locus shape?” In this book we will explore these and many other related
questions.

1.5 On to more confocal results

Upon a more systematic probing of confocal 3-periodics, there emerges a list of
surprising facts:

• There is a special center of the triangle – the mittenpunkt – which remains

stationary over the family at the common center.

• The ratio of the radii of two classic circles associated with a triangle – the
circumcircle and incircle – is invariant over the family.

• In turn, the previous observation implies (via a well-known theorem) that

the sum of internal angle cosines is also invariant. Interestingly, this fact
1.6. Branching out to non-confocal families 7

remains true6 for N > 3 families!

1.6 Branching out to non-confocal families

Beyond confocal 3-periodics, we also investigate a few other “famous” concen-
tric ellipse pairs, shown in Figure 1.5. These include a pair (i) with an incircle, (ii)
with circumcircle, (iii) of homothetic ellipses, (iv) of “dual” ellipses, and (v) of
the excentral triangles to the confocal pair itself. As it turns out, each aforemen-
tioned family maintains a specific triangle center7 stationary at the common center.
Experiments quickly identify special euclidean quantities each family conserves.
For example, both the circumcircle and homothetic families conserves the sum
of their squared sidelengths. While the former conserves the product of internal
angle cosines, the latter conserves the sum of internal angle cotangents! Compare
these with the confocal family, whose perimeter (i.e., the sum of sidelengths), and
sum of its internal angle cosines are conserved.
Indeed, as properties, invariants, and loci types for each family were unearthed,
we began to organize families in groups with shared “behaviors”.

1.7 Analysis methods

We have used a number of methods to analyze and prove some of the facts de-
tected experimentally. These include: (i) analytic geometry often assisted by a
computer algebra system (CAS), (ii) synthetic and/or inversive geometry of con-
ics, often relying on Akopyan and Zaslavsky (2007), Coxeter and Greitzer (1967),
and Glaeser, Stachel, and Odehnal (2016), (iii) the theory of resultants to clas-
sify loci as algebraic, (iv) family parametrizations of various kinds, such as stan-
dard, elliptic-function-based, and one based on Blaschke products, as described in
Daepp et al. (2019), etc.
In fact, the latter has also helped us approach two central questions: (i) what de-
termines whether the locus of a triangle center is an ellipse or not, and (ii) whether
the locus of the incenter (and/or excenters) can be an ellipse in a pair other than
the confocal one. Experiments suggest it cannot, and a comprehensive proof is
still lacking.
6 Indeed, to us this opened a Pandora’s box, since many properties of N D 3 systems continue
to hold for N > 3.
7 These are special points on a triangle, thousands of which are catalogued in Kimberling (2019).
8 1. Introduction

Figure 1.5: The confocal family is shown at the top left. Also shown are 5 other
“famous” concentric families. Video
1.8. Related work 9

1.8 Related work

Loci of triangle centers over the N D 3 “poristic” family (interscribed between
two circles) were studied by Odehnal (2011). In Schwartz and Tabachnikov (2016a)
the loci of vertex, perimeter, and area centroids are studied over a generic Pon-
celet family indicating that the first and last are always ellipses while in general
the perimeter one is not a conic. The locus of the “circumcenter-of-mass” (a gener-
alization of the circumcenter for N-gons), studied in Tabachnikov and Tsukerman
(2014), is shown to be a conic over Poncelet N-periodics in Chavez-Caliz (2020).

Over confocal 3-periodics, the elliptic locus of (i) the incenter was proved in
Fierobe (2021), Garcia (2019), and Romaskevich (2014); (ii) of the barycenter in
Garcia (2019) and Schwartz and Tabachnikov (2016a); and (iii) of the circumcen-
ter in Fierobe (2021) and Garcia (2019). The elliptic locus of the Spieker center
(which is the perimeter centroid of a triangle) was proved in Garcia (2019). Some
properties and invariants of confocal N-periodics are described in Reznik, Gar-
cia, and Koiller (2020a); N D 3 subcases are proved in Garcia, Reznik, and
Koiller (2020b). Some invariants have been proved for all N ⩾3 in Akopyan,
Schwartz, and Tabachnikov (2020), Bialy and Tabachnikov (2020), and Chavez-
Caliz (2020).

1.9 Book organization

The next 3 chapters describe the basic geometry, several phenomena, and invari-
ants of Poncelet 3-periodic families, as follows:

• Confocal family, Chapter 2: these are billiard 3-periodics, i.e., interscribed

in a confocal pair of ellipses.

• Concentric, axis-parallel (CAP), Chapter 3: these are non-confocal triangle

families interscribed in a pair of concentric, axis-parallel ellipses.

• Non-concentric, axis-parallel (NCAP), Chapter 4: these are triangle families

interscribed in a pair of non-concentric, axis-parallel ellipses (and/or circles).

Following the above, we redirect our attention to the geometry and properties
of loci of triangle centers and vertices over some of the aforementioned families.
To be sure:
10 1. Introduction

• Loci of the confocal family, Chapter 5.

• Loci in non-confocal CAP families Chapter 6.

• A framework for analyzing and explaining locus phenomena, based on Blashke

products, Chapter 7.

In Chapter 8 we introduce a property-rich, non-Ponceletian family called the

“focus inversive” family. These are images of billiard 3-periodics under an inver-
sion with respect to a focus-centered circle.
Most figures herein will contain links to either YouTube videos or live demon-
strations. The latter are rendered by a specially-built locus visualization applica-
tion. Its functionality is described in Chapter 9.
At the end of every chapter we include a set of exercises and whenever possible,
research questions. We very much encourage the reader to give them a try.
The following appendices are included:

• Appendix A contains some notes about triangle geometry. A more complete

reference is of course Weisstein (2009).

• Appendix B contains a review of Jacobi elliptic functions used for one of

our family parametrizations.

• In Appendix C we describe the loci of the Brocard points over a certain

non-Ponceletian family of triangles defined with respect to an ellipse or a
circle.
2 Confocal Pair

In this chapter we describe the geometry and properties of billiard 3-periodics, i.e.,
the 1d family of Poncelet triangles “interscribed”1 in a pair of confocal ellipses.
We begin by (i) reviewing some elliptic billiard preliminaries; then (ii) deriving
conditions for the geometry of the inner ellipse (also known as the caustic); fol-
lowing that we review and prove some early results in regards to (iii) the elliptic
loci of the incenter and excenter of the family, and (iv) the stationarity of a special
triangle center known as the Mittenpunkt. Then (v) the key metric conservation of
the ratio of inradius-to-circumradius is discussed. The chapter ends by proposing
two alternative parameterizations for the vertices of billiard 3-periodics which we
call standard and Jacobi-based.

2.1 Preliminaries
Let the confocal pair of ellipses E and Ec be given by:

x2 y2 x2 y2
EW C 1 D 0; Ec W C 1D0
a2 b2 ac2 bc2
1 This means inscribed in a first ellipse while circumscribing a second one.
12 2. Confocal Pair

where c 2 D a2 b 2 D ac2 bc2 .

Indeed, the family of billiard N -periodics classically conserve perimeter L
and Joachimsthal’s constant J . The latter one is equivalent to stating all trajectory
segments are tangent to a confocal caustic, see Tabachnikov (2005, Thm 4.4) and
Arnold and Tabachnikov (2020).
When N D 3, we can derive these explicitly using the vertex parametrization
given in Equation (2.5).
Proposition 2.1. For billiard 3-periodics, the perimeter and Joachimsthal’s con-
stant are given by:
p
2ı a2 b 2
J D ; L D 2.ı C a2 C b 2 /J
c2
p
where ı D a4 a2 b 2 C b 4 , called here Darboux’s constant.
Proof. We compute the values considering an isosceles 3-periodic with P1 D
Œa; 0, and
"
p # "
p #
a b 2 ı b 2 2ı a2 b 2 a b2 ı b 2 2ı a2 b2
P2 D ; ; P3 D ;
a2 b 2 a2 b 2 a2 b 2 a2 b2
(2.1)
We have that
  
P1 P3 1
L D jP2 P3 j C 2jP1 P2 j; J D ; ;0
jP1 P3 j a
Straightforward calculations using the vertex parametrization in Equation (2.5),
leads to the stated result.
Henceforth the oft-occurring quantity ı will be referred to as the Darboux con-
stant. An interesting geometric interpretation for it appears in Proposition 2.5.

2.2 Caustic semiaxes

The Cayley condition for a concentric, axis parallel (CAP) pair of ellipses to admit
a 3-periodic family is given by:
ac bc
C D1 (2.2)
a b
In turn, this constrains the semiaxes of the confocal caustic.
2.3. Incenter and excenter loci 13

Proposition 2.2. The semiaxes ac ; bc of the confocal caustic are given by:



a ı b2 b a2 ı
ac D ; bc D

c2 c2

When a D b, we have that ac D bc D a=2.

Proposition 2.3. The semiaxes a and b of the ellipse in terms of the semiaxes ac
and bc of the confocal ellipse are given by:
s
1p 1 2ac3 4c 2 ac ac p
aD w1 C w2 p C ; b D a2 c 2
2 2 w1 2
2 2
w1 D ac2 .4cac bc / 3 ; w2 D 2ac2 C .4cac bc / 3
The implicit equation that defines a above is the quartic given by
c 2 .ac2 2ac a/ C a2 .2ac a a2 / D 0

2.3 Incenter and excenter loci

An intriguing phenomenon is that over billiard 3-periodics, the locus of both incen-
ter and the excenters are ellipses, as was initially detected experimentally (see an
early Video). This was proved in Romaskevich (2014) and Garcia (2019). Indeed,
we haven’t yet found another Poncelet pair where this is the case2 , see Conjecture 2
and Conjecture 3.
Referring to Figure 2.1:
Theorem 2.1. Over billiard 3-periodics, the locus of the incenter X1 and excenter
are ellipses E1 and Ee concentric and axis-parallel with the confocal pair whose
axes .a1 ; b1 / and .ae ; be / are given by:
ı b2 a2 ı
a1 D ; b1 D
a b
b2 Cı 2
a Cı
ae D ; be D
a b
Furthermore, E1 and Ee have reciprocal aspect ratios, i.e., a1 =b1 D be =ae .
2 One exception is the poristic family, for which the locus of the excenters is a circle and that of
the incenter is a point.
14 2. Confocal Pair

Figure 2.1: Top Left: An elliptic billiard 3-periodic (solid blue) is shown inscribed
in an outer ellipse (black) and a confocal caustic (brown). Graves’ theorem im-
plies its internal angles will be bisected by ellipse normals (black arrows). Also
shown is the incenter X1 defined as the intersection of said bisectors. Bottom Left:
Poncelet’s porism implies a 1d family of such triangles exists. Some samples are
shown (dashed blue). A classic invariant is perimeter. The Mittenpunkt X9 re-
mains stationary at the center. The incenter X1 sweeps an ellipse (dashed green).
Right: The excentral triangle (solid green) has sides perpendicular to the bisec-
tors. Over billiard 3-periodics, the excentral is of variable perimeter. Its vertices
(known as the “excenters”) also sweep an ellipse (dashed green) whose aspect ra-
tio is the reciprocal of that of the incenter locus. The symmedian point X60 of the
excentral triangle coincides with X9 of the reference and is therefore stationary.
Live
2.4. A stationary point 15

Proof. It follows from the vertex parametrization in Equation (2.5) and the defini-
tion of incenter and excenters. We have that
s1 P1 C s2 P2 C s3 P3 1
X1 D D .s1 P1 C s2 P2 C s3 P3 /
s1 C s2 C s3 L
where s1 D jP2 P3 j, s2 D jP1 P3 j and s3 D jP1 P2 j. A careful symbolic
analysis shows that E1 .X1 / D 0. A similar analysis considering the excenters
shows that the locus of the three points is the ellipse Ee stated.
A more general treatment to the above is given in Chapter 7.
Corollary 2.1. The pair fE; Ee g is Ponceletian.
Proof. Direct from Cayley condition

a b a2 b2
C D 2 C 2 D1
ae be b Cı a Cı

2.4 A stationary point

The Mittenpunkt X9 is a triangle center where lines from each excenter thru the
side midpoint meet. Referring to Figure 2.2:
Theorem 2.2. Over the family of 3-periodics in the elliptic billiard, X9 is station-
ary at the common center.
An elegant synthetic proof was kindly contributed by Romaskevich (2019):
Proof. Let E be the outer ellipse in the confocal pair, O. By definition, the Mit-
tenpunkt X9 is where lines from the excenters Ei through the side midpoints Mi
concur. Notice each side is an ellipse chord between tangents to E seen from the
Ei (this is because in the confocal pair the excentral triangle is tangent to E). Con-
sider the image of lines Ei Mi under an affine transform which sends E to a circle
C 0 , let O 0 be its center. The transformed lines will pass through the midpoints of
chords of C 0 between tangents seen from Ei0 (the affine image of Ei ). By circular
symmetry, such lines must also pass through O 0 , and therefore remain stationary.
But O 0 is the affine image of O, so the result follows.
16 2. Confocal Pair

M2 M2
X9 X9
M1 M3 M1 M3

Figure 2.2: Left: 3-periodic billiard triangle (blue), its excentral triangle (green).
The Mittenpunkt X9 is the point of concurrence of lines drawn from the excenters
through sides’ midpoints Mi . Right: the affine image which sends the billiard to
a circle. Lines from imaged excenters through sides’ midpoints must pass through
the origin. Since the latter is stationary, so must be its pre-image X9 , which is
stationary at the billiard center. Video
2.5. Conserved quantities 17

2.5 Conserved quantities

Given a triangle, let r and R denote the radius of its incircle and circumcircle,
known as the inradius and circumradius, respectively. Over billiard 3-periodics,
note these two radii are variable. Referring to Figure 2.3:

Theorem 2.3. r=R is invariant over billiard 3-periodics and given by:

r 2.ı b 2 /.a2 ı/
D :
R c4

Proof. The following relation, found in Johnson (1960), holds for any triangle:

s1 s2 s3
rR D ;
2L

where L D s1 C s2 C s3 is the perimeter, constant over billiard 3-periodics. There-

fore:

r 1 s1 s2 s3
D
(2.3)
R 2L R2

Next, let P1 D .a; 0/ be a vertex of an isosceles 3-periodic. Obtain a candidate

expression for r=R. This yields (2.3) exactly. Using the vertex parametrization in
Equation (2.5), derive an expression for the square of the right-hand side of (2.3)
as a function of x1 and subtract from it the square of (2.3). In Garcia, Reznik, and

2
Koiller (2020b) it is shown s1 s2 s3 =R2 is rational on x1 . For simplification, use
R D s1 s2 s3 =.4A/, where A is the triangle area. With a CAS, show said difference
is identically zero for all x1 2 . a; a/.

Let i , r, R, and A denote the ith internal angle, inradius, circumradius, and
area of a reference triangle. Primed quantities refer to the excentral triangle. The
relations below, appearing in Johnson (1960), hold for any triangle:
18 2. Confocal Pair

Figure 2.3: The incircle (green), circumcircle (purple), and 9-point (Euler’s) circle
(pink) of a billiard triangle (blue). These are centered on X1 , X3 , and X5 , respec-
tively. Their radii are the inradius r, circumradius R, and 9-point circle radius
r9 D 2R. Over the family, the ratio r=R is invariant. In turn this implies an in-
variant sum of cosines. Live
2.5. Conserved quantities 19

3
X r
cos i D 1 C (2.4)
R
i D1
Y 3
r
cos i0 D
4R
i D1
A r
0
D
A 2R

Corollary 2.2. Over billiard 3-periodics, also invariant are the sum of 3-periodic
cosines, the product of excentral cosines, and the ratio of excentral-to-3-periodic
areas.
Direct calculations yields an expression for the invariant sum of cosines in
terms of elliptic billiard constants J and L.
P3
Corollary 2.3. i D1 cos i D JL 3
Indeed in Akopyan, Schwartz, and Tabachnikov (2020) it is shown that for all
N the sum of cosines is invariant and equal to JL N .
Let Pi be a billiard 3-periodic vertex and dj;i D jPi fj j its distance to
billiard focus fj .
Proposition 2.4. Over billiard 3-periodics, the following sum is invariant:
X 1 X 1 a2 C b 2 C ı
D D
d1;i d2;i ab 2

Proof. Direct computation with CAS using vertex parametrization given in Sec-
tion 2.7.
Let P D .x; y/ be a point on an ellipse with semiaxes a; b. In Weisstein (2019,
Ellipse), the curvature at P is expressed both in terms of its coordinates and the
distances d1 ; d2 to the foci as follows:
 2  3=2
1 x y2 ab
D 2 2 4
C 4 D
a b a b .d1 d2 /3=2
Let i denote the billiard ellipse curvature at vertex Pi of a Poncelet 3-periodic.
From the above and Proposition 2.4 obtain:
20 2. Confocal Pair

Corollary 2.4. Over billiard 3-periodics, the following quantity is conserved:

3
X 2 a2 C b 2 C ı
i3 D 4
i D1 .ab/ 3

2.6 An interpretation for Darboux’s constant

Darboux’s constant ı appearing above has a curious geometric interpretation. Re-
call the power of a point Q with respect to a circle C D .C0 ; R0 / is given by
jQ C0 j2 R02 , see Weisstein (2019, Circle Power). Let C denote the (moving)
circumcircle of billiard 3-periodic, and O D X9 the billiard center.

Proposition 2.5. The power of O with respect to C is constant and equal to ı.

Proof. Consider an isosceles billiard 3-periodic given by Equation (2.1). Its cir-
2 2 Cı
cumcircle will be centered at C0 D Œ b 2b ı ; 0 with circumradius R0 D b 2b :
Therefore, the power of the center of the ellipse with respect to the circumcircle is
given by
 2 2  2 2
2 2 b ı b Cı
jOC0 j R0 D D ı:
2b 2b
The stated invariance is confirmed with a CAS using the vertex parametrization
in Equation (2.5).

2.7 Confocal vertex parametrization

We describe two parametrizations for billiard 3-periodic vertices: (i) standard and
(ii) Jacobi.

2.7.1 Standard
We call “standard” parametrization that where a first vertex P1 .t / of the billiard
3-periodic is parametrized as P1 .t/ D Œx1 ; y1  D Œa cos t; b sin t .
As derived in Garcia (2019), P2 D .x2 ; y2 /=q2 and P3 D .x3 ; y3 /=q3 where:
2.7. Confocal vertex parametrization 21

x2 D b4 a2 C b 2 cos2 ˛ a2 x13 2a6 cos ˛ sin ˛ y13


C a4 .a2 3 b 2 / cos2 ˛ C b 2 x1 y12 2 a4 b 2 cos ˛ sin ˛x12 y1 ;



y2 D2b 6 cos ˛ sin ˛ x13 a4 a2 C b 2 cos2 ˛ b 2 y13

(2.5)
C 2 a2 b 4 cos ˛ sin ˛ x1 y12 C b 4 .b 2 3 a2 / cos2 ˛ C a2 x12 y1



q2 Db 4 a2 .a2 b 2 / cos2 ˛ x12 C a4 b 2 C .a2 b 2 / cos2 ˛ y12


2 a2 b 2 a2 b 2 cos ˛ sin ˛ x1 y1 :

x3 D b 4 a 2 b 2 C a2 cos2 ˛ x13 C 2 a6 cos ˛ sin ˛ y13



C a4 cos2 ˛ a2 3 b 2 C b 2 x1 y12 C 2 a4 b 2 cos ˛ sin ˛ x12 y1



y3 D 2 b 6 cos ˛ sin ˛ x13 C a4 b 2 b 2 C a2 cos2 ˛ y13


2 a2 b 4 cos ˛ sin ˛ x1 y12 C b 4 a2 C b 2 3 a2 .cos ˛/2 x1 2 y1 ;





q3 D b 4 a2 a2 b 2 cos2 ˛ x12 C a4 b 2 C a2 b 2 cos2 ˛ y12


C 2 a2 b 2 a2 b 2 cos ˛ sin ˛ x1 y1 :

where:
p p p
a2 b a2 b 2 C 2 a4 b 2 c 2 a2 b 2 2ı a2 b 2
cos ˛ D q D q
c 2 a4 c 2 x12 c 2 a4 y12 C b 4 x12
q
2
2
b4 a2 ı x12 C a4 b 2 ı y12
sin ˛ D q
c 2 a4 y12 C b 4 x12

Note that in Section 3.6 we generalize the above to any concentric, axis-parallel
pair.

2.7.2 Jacobi’s universal measure

Under the standard parametrization, we can obtain the “position” t of P D Œx; y
on an ellipse:
1 ay
t D tan :
bx
22 2. Confocal Pair

Figure 2.4: The cosines [Link] / of billiard 3-periodic internal angles for the
standard (top) and Jacobi parametrizations (bottom). While in the former case the
three curves are distinct, in the latter case all cosines follow the same curve at
different phases.

As shown in Figure 2.5(top), when a first vertex P1 .t / in a billiard 3-periodic is

parametrized in the standard way, though its position is linear on the t parameter,
it will drive motions of the other two vertices P2 .t / and P3 .t / which are both
distinct and non-linear.
Fortunately, a uniform parametrization exists, which goes back to Jacobi, for
all Poncelet families, based on the so-called “universal measure”, which linearizes
the Poncelet map, see Koiller, Reznik, and Garcia (2021). Specifically, vertices
are obtained at fixed multiples of a constant
u in the argument of certain Jacobi
elliptic functions. This parametrization, adapted to the elliptic billiard case, ap-
pears in Stachel (2021a,b) and is reproduced below. First let’s recall a few useful
definitions.
The notation adopted below takes after Armitage and Eberlein (2006).
Definition 2.1. The incomplete elliptic integral of the first kind K.'; k/ is given
2.7. Confocal vertex parametrization 23

by: Z '
d
K.'; k/ D p (2.6)
0 1 k 2 sin2 
The complete elliptic integral of the first kind K.k/ is simply K.=2; k/.

Definition 2.2. The elliptic sine sn, cosine cn, and delta-amplitude dn are given
by:
sn.u; k/ D sin '
cn.u; k/ D cos '
q
dn.u; k/ D 1 k 2 sin2 '

where ' D am.u; k/ is known as the amplitude, i.e., the upper-limit in the integral
in Equation (2.6) such that K.'; k/ D u.
A review of these functions appears in Appendix B.
Remark 2.1. Note to the reader: Mathematica (resp. Maple) expects m D k 2
(resp. k) as the second parameter to elliptic functions.
Theorem 2.4. A billiard 3-periodic Pi .i D 1; : : : ; N / of period N with turning
number , where gcd.N;  / D 1, is parametrized on u with period 4K where:

Pi D Πa sn .u C i
u; m/ ; b cn .u C i
u; m/

where,
ac2 bc2 4K
m D k2 D ;
u D
ac2 N
q
bc
aD b 2 C ac2 bc2 ; b D
cn.
u
2 ; m/
Proof. See Stachel (2021b).
Since in this chapter we are considering billiard 3-periodics, so above N D 3,
and  D 1. As shown in Figure 2.5, under the Jacobi parametrization each of
the 3 vertices of billiard 3-periodics follows the exact same curve, albeit with a
120-degree phase.
Recall the sum of cosines is constant for billiard 3-periodics. Figure 2.4 shows
how individual cosines follow either (i) 3-distinct curves, or (ii) the same exact
curve (at different phases) if the parametrization is standard or Jacobi, respectively.
24 2. Confocal Pair

Figure 2.5: The “position” i (vertical axis) of a point on an ellipse with semiaxes
a; b vs the billiard 3-periodic parameter (horizontal axis). Top: vertex under “stan-
dard parametrization”, i.e., P1 .t/ D Œa cos t; b sin t. Notice while P1 ’s position
evolves linearly, those of P2 and P3 are different curves. Bottom: Said positions
under Jacobi’s parametrization. Notice the three positions are 120-degree delayed
copies of one another.
.
2.8. Exercises 25

Figure 2.6: Two random triangles shown with their circumbilliards. Video

2.8 Exercises
Exercise 2.1. Referring to Figure 2.6, show that every triangle has a circumbil-
liard, i.e., an ellipse to which it is inscribed and to which it is a billiard 3-periodic.
Compute the axes of said circumbilliard with respect to triangle vertices.

Exercise 2.2. A pair of circles uniquely defines a pencil of coaxial circles, see
Weisstein (2019, Limiting Points). The pencil contains exactly two circles which
degenerate to a point, known as limiting points. Derive the location of such points
for the poristic pair obtained from the image of two confocal ellipses centered at
Œ0; 0 and with axes a; b and a0 ; b 0 .

Exercise 2.3. Let `1 ; `2 be the limiting points of the two circles which are polar
images of a confocal pair E; E 0 with respect to a circle centered on f1 . At what
aspect ratio a=b of E will `2 coincide with f2 ?
Exercise 2.4. A well-known result is that the inversion of a circle pair C; C 0 with
respect to a circle C1 centered on `1 (resp. C2 centered on `2 ) is a pair of concentric
circles C10 and C100 (resp. C10 and C100 ). Prove the following lesser known result: the
ratio of radii between C10 and C100 is the same as the ratio between C20 and C200 .

Exercise 2.5. Referring to Figure 8.9, let C; C 0 be the pair of circles which are the
polar image of a confocal pair of ellipses E; E 0 . Let C10 ; C100 be the inversive images
of C; C 0 wrt to a circle centered on a focus of the ellipse pair. Prove that: (i) C10 and
C100 are concentric with the ellipse pair and (ii) C10 (resp. C100 ) is externally tangent
to E (resp. E 0 ) at its left and right major vertices.
26 2. Confocal Pair

Exercise 2.6. Prove the inversive image of billiard 3-periodics with respect to a
focus-centered circle is a non-Ponceletian family inscribed in Pascal’s Limaçon
whose Gergonne point X7 is stationary; see it Live. Indeed, this family has con-
stant perimeter (to be shown later).

Exercise 2.7. Consider the ellipse x 2 =a2 C y 2 =b 2 D 1. For a 3-periodic billiard

orbit with vertices Pi D Œxi ; yi  (i=1,2,3) show that:

.x2 y3 x3 y2 / x1 y1 C .x3 y1 x1 y3 / x2 y2 C .x1 y2 x2 y1 / x3 y3 D 0

Exercise 2.8. For a 3-periodic billiard orbit with vertices Pi D Œxi .t/; yi .t /
(i=1,2,3) Let Ci .t / D Œ1=xi .t/; 1=yi .t/.
Show that the polygon fC1 .t/; C2 .t /; C3 .t/g is a segment that can be bounded
or unbounded.

Exercise 2.9. Which simple or self-intersected N - gon (closed polygon with N

vertices and N sides) can be an orbit on an elliptic billiard?
For N D 4 only the parallelogram can be a non self-intersected orbit on an
elliptic billiard, see Connes and Zagier (2007). See Garcia and Reznik (2020) for
the analysis of self-intersected 4 gons.

Exercise 2.10. Consider a 3-periodic billiard orbit and its antipodal orbit. Show
that the six points of intersections of the two triangles are contained in a stationary
confocal ellipse Eh : x 2 =ah2 C y 2 =bh2 D 1 where:

p
ı b2 a2 C b 2 C 2 ı 2ı a2 b2
ah D
2
3 a2 b2


p
a2 ı a2 C b 2 C 2 ı 2ı a2 b2
bh D
2
3 a 2 b 2

Conclude that the pair of ellipses fEh ; E1 g is a billiard pair having all orbits of
period 6. Also show that the pair fE; Eh g defines a zig-zag billiard and that the
orbits have period 12 and that the perimeter is constant. See Live
2.9. Research questions 27

2.9 Research questions

Question 2.1. Weisstein (2019, Extouch triangle) defines the extouch triangle
as having vertices at the points of contact of the excircles with a triangle’s side-
lines. In Chapter 5 we show that the vertices of the extouch triangles of billiard
3-periodics coincide with the caustic touchpoints, see it Live. Show that said ex-
touch family is also Ponceletian and concentric with the elliptic billiard; derive
expressions for the semiaxes of its elliptic caustic. Is its center a triangle center?
3
Concentric,
Axis-Parallel
(CAP)

Below we introduce, five additional notable 3-periodic Poncelet families inter-

scribed between a pair of concentric, axis-parallel (CAP) ellipses. They were
shown in Figure 1.5 and we name them (i) incircle, (ii) circumcircle, (iii) homo-
thetic, (iv) dual, and (v) excentral (their geometry is defined below). As before,
the Cayley condition Equation (2.2) will be used to constrain the ellipse pair. For
each family we derive geometric properties and invariants. The chapter concludes
with a parameterization which is general for any CAP pair and is used to support
several proofs. In Section 3.7 we summarize properties, fixed points, and other
traits of the families treated herein.

3.1 Excentral family

This is the Ponceletian family of excentral triangles to billiard 3-periodics. If the
axes of its caustic are a; b, this family is inscribed in an ellipse with ae ; be given
in Theorem 2.1; see Figure 3.1(left). Indeed, by inverting the relations in the latter,
we can express a; b 1 in terms of ae ; be :
1 There is a slight abuse of notation in that these should have been labeled a ; b . However we
c c
maintained these as a; b since the caustic to the excentral family is the elliptic billiard.
3.1. Excentral family 29

Figure 3.1: Left: billiard 3-periodic (blue) and its excentral triangle (green). The
former conserves the sum of its cosines. The latter is inscribed in an ellipse (dashed
green) and conserves the product of its cosines. Middle: Affine image of confocal
family which sends caustic (brown) to a circle. This family also conserves the sum
of cosines, equal to that conserved by its confocal pre-image. Right: Affine image
of confocal family which sends billiard ellipse (black) to a circle. This family also
conserves the product of cosines, equal to that conserved by the excentral family
of its pre-image. Video
30 3. Concentric, Axis-Parallel (CAP)

Proposition 3.1. Given the semiaxes ae , be of Ee the semi axis of the caustic E
are given by:

.ıe ae2 3be2 /ae .3ae2 C be2 ıe /be

aD ; b D
2ce2 2ce2
p
where ce2 D ae2 be2 and ıe D ae4 C 14ae2 be2 C be4 .
The symmedian point X6 of the excentral triangle coincides with the mitten-
punkt of its reference, see Kimberling (2019, X(6)). Therefore:
Corollary 3.1. Over excentral 3-periodics, the symmedian point X6 of the excen-
tral family is stationary.
Recall that in Corollary 2.2 two other notable invariants are mentioned: prod-
uct of internal angle cosines, and ratio of its area by billiard 3-periodics.
Corollary 3.2. The invariant product of cosines of excentral 3-periodics is a quar-
ter of the quantity in Corollary 2.2. Furthermore the area ratio of billiard 3-
periodics by excentrals is half the quantity in Corollary 2.2.
Let si0 denote the variable sidelengths of the excentral family, i D 1; 2; 3. Here
is an additional curious invariant:
Proposition 3.2. Over the excentral family, the sum squared sidelines divided by
the product of sidelines is constant. Furthermore it is equal to Joachimsthal’s
constant J of its parent 3-periodic billiard family. Explicitly:
P 0 2 p
.si / 2ı a2 b 2
Q 0 D DJ
si c2
Proof. Derive explicit expressions for excentral sidelengths and arrive at claim
via CAS simplification.
Referring to Figure 3.2, Weisstein (2019, Cosine Circle) defines the cosine cir-
cle as centered on the symmedian point X6 , and passing through the 6 intersections
of lines through X6 parallel to the sides of the orthic triangle. Its radius r  is given
by the product of sidelengths divided by the sum of their squares.
Corollary 3.3. The cosine circle of the excentral family is stationary with radius
r  D 1=J .
3.1. Excentral family 31

Figure 3.2: The cosine circle (red) of the excentral family (green) is stationary.
It contains the 6 intersections of lines (dashed blue) through the common center
(family’s X6 and billiard periodics’ X9 ) which are parallel to their orthic, i.e., side-
lines of billiard 3 periodics. Video
32 3. Concentric, Axis-Parallel (CAP)

Figure 3.3: A 3-periodic (blue) in the incircle family, and its fixed-radius circum-
circle (purple). The locus of the circumcenter X3 is a concentric circle (dashed
purple). Live

3.2 Incircle family

The incircle family, shown in Figure 3.1(middle), is the Poncelet family in a CAP
pair for which the caustic is a circle (let r denote its radius). It follows immediately
that the family’s incenter X1 is stationary. Let a; b be the axes of the ellipse the
family is inscribed in. Cayley yields:

Proposition 3.3. The inradius r of the incircle family is given by:

ab
r D ac D bc D
aCb

As shown in Figure 3.3:

Proposition 3.4. The incircle family has invariant circumradius given by R D

.a C b/=2. Furthermore, the locus of its circumcenter X3 is a circle of radius
d D R b D a R centered on the common center O D X1 .
3.2. Incircle family 33

Proof. Let P1 D .x1 ; y1 / be a first vertex of the incircle family. Using an explicit
parametrization for P2 and P3 , obtain via CAS the following coordinates for the
moving circumcenter X3 :

2    3
a b4 x1 x12 .a C b/2 C a2 b .2 a C b/ y1 x12 .a C b/2 a2 b 2
X3 D

;

5
2 a a2 b 2 x12 C a2 b 2 b a2 x12 C b 2 a2 x12

And circumradius R D jP1 X3 j D .a C b/=2. Also obtain that the locus of

X3 is a circle concentric with the incircle and of radius .a b/=2.

Referring to Equation (2.4):

Corollary 3.4. The incircle family conserves its sum of cosines given by:
X r a2 C 4ab C b 2
cos i D 1 C D
R .a C b/2

3.2.1 Confocal affine image

As Figure 3.1(middle) depicts, the incircle family can also be obtained from an
affine image of billiard 3-periodics which sends the confocal caustic to a circle.
Let ˛; ˇ be the semiaxes of its billiard ellipse pre-image.

Lemma 3.1. The confocal family is sent to the incircle family by scaling it along
the major axis by an amount s given by:
q
ˇ.ıN ˛2 C ˇ2/ N
sD ; ıD ˛4 .˛ˇ/2 C ˇ 4
˛3

a D s˛, and
Proof. The scaled family will be inscribed in an ellipse with semiaxes
b D ˇ. Its caustic will be the circle r D bc , where bc D ˇ ˛ 2 ıN =.˛ 2 ˇ 2 / is
the confocal caustic minor axis given in Proposition 2.2. The Cayley condition for
the incircle family imposes that r D bc D .ab/=.a C b/, i.e., the result follows
from solving bc D .s˛ˇ/=.s˛ C ˇ/ for s.

Surprisingly:
34 3. Concentric, Axis-Parallel (CAP)

Proposition 3.5. The sum of cosines conserved by the incircle family is identical
to that conserved by billiard 3-periodics which are its affine pre-image.
Proof. Let s be the scaling along the major axis in Lemma 3.1. Plug a D s˛ and
b D ˇ into Corollary 3.4, subtract one (to obtain r=R for the incircle family) and
verify it yields the expression in Theorem 2.3.

3.3 Circumcircle family

The circumcircle family, shown in Figure 3.1(right), is the Poncelet family in a
CAP pair for which the outer ellipse is a circle (let R denote its radius). It follows
immediately that the family’s circumcenter X3 is stationary. Let ac ; bc be the axes
of its inellipse and si the sidelengths. Cayley imposes ac C bc D R.
Lemma 3.2. Poncelet triangles in the circumcircle family are always acute.
Proof. Since the stationary circumcenter X3 is interior to the caustic caustic, it
will be interior to circumcircle family triangles, and the result follows.
Proposition 3.6. The circumcircle family conserves the sum of squared sidelengths.
This is given by:
3
X
si2 D [Link] C 2bc /.2ac C bc /
i D1
Proof. CAS-assisted simplification from the vertex parametrization in Proposi-
tion 3.17.
Proposition 3.7. The circumcircle family conserves the product of its internal an-
gle cosines. This is given by:
3
Y ac bc ac bc
cos i D 2
D
[Link] C bc / 2R2
i D1

Proof. CAS-assisted simplification from vertex parametrization.

Recall the orthic triangle has vertices at the feet a triangle’s altitudes. Let Rh
denote its circumradius. The well-known identity Rh D R=2 appears in Weisstein
(2019, Orthic Triangle, Eqn 7). Therefore Rh is invariant over the circumcircle
family. Let rh denote the orthic’s inradius. Referring to Figure 3.4:
3.3. Circumcircle family 35

Figure 3.4: A circumcircle family 3-periodic (blue) and its orthic triangle (orange).
Over the family the orthic’s circumcircle (dashed orange) and incircle (dashed
green) have invariant radii. Also shown are their centers X3;h and X1;h which, for
any reference triangle, correspond the nine-point center X5 and orthocenter X1 .
Video, Live

Proposition 3.8. Over the circumcircle family rh is invariant and given by rh D

ac bc =.ac C bc /.

Q In Weisstein (ibid., Orthic Triangle, Eqn. 5) one finds the identity rh D

Proof.
2R 3iD1 cos i . Recalling R D ac C bc , substitution into Proposition 3.7 yields
the claim.
Referring to Figure 3.4, it can be shown (see Exercise 3.3):
Lemma 3.3. Over the circumcircle family, the locus of the orthic circumcenter
(i.e., the 9-point center X5 of the family) is a circle concentric with the pair.
Let i0 denote the curvature of the inner ellipse at the points of contact of cir-
cumcircle 3-periodics (the proof is left to the reader in Question 3.10):
36 3. Concentric, Axis-Parallel (CAP)

Proposition 3.9. Over circumcircle 3-periodics, the following quantity is con-

served:
X 3
2 a2 C ac bc C bc2
.i0 / 3 D c 4
i D1 .ac bc / 3

3.3.1 Confocal affine image

As Figure 3.1(right) depicts, the circumcircle family can also be obtained from an
affine image of billiard 3-periodics which sends the billiard ellipse with semiaxes
˛; ˇ to a circle with radius R D ˇ. Therefore billiard 3-periodics are sent to the
circumcircle family by scaling it along the major axis by an amount s 0 D ˇ=˛.
Therefore Proposition 2.2 implies:

Lemma 3.4. The caustic semiaxes ac ; bc of the circumcircle family which is the
s 0 -affine image of the confocal family are given by:

ˇ ˇ.ıN ˇ2/ N
ˇ.˛ 2 ı/
ac D ˛c D 2 ; bc D ˇc D
˛ ˛ ˇ2 ˛ 2 ˇ 2

where ˛c ; ˇc are the caustic semiaxes of the confocal pre-image, and ˛; ˇ; ıN are
as previously defined.

Note that the s 0 -affine image of billiard excentrals becomes a Poncelet family with
fixed incircle; see Figure 3.1(right, dashed green triangles). We have seen above
such a family conserves its sum of cosines. Surprisingly, the following invariant
“role reversal” takes place:

Proposition 3.10. The sum of cosines conserved by billiard 3-periodics is the same
as the one conserved by the s 0 -affine image of billiard excentrals. Furthermore
product of cosines conserved by billiard excentrals is the same as the one con-
served by the s 0 -affine image of billiard 3-periodics (circumcircle family).

Proof. For the first statement it suffices to show that the s 0 -affine image of billiard
excentrals has sides parallel to those of the s-image of billiard 3-periodics, i.e.,
the incircle family and use Proposition 3.5. The second statement can be proved
algebraically from vertex parametrization.
3.4. Homothetic family 37

Figure 3.5: A 3-periodic (blue) interscribed between two homothetic ellipses

(black, brown). Since this family is an affine image of one of equilaterals inter-
scribed between two concentric circles, (i) the barycenter X2 is stationary at the
common center, and (ii) the area is conserved. Also conserved is (iii) the sum of
squared sidelengths. (ii) and (iii) imply the Brocard angle ! is invariant. Also
shown are the two (moving) the Brocard points ˝1 and ˝2 . Video, Live

3.4 Homothetic family

The homothetic family, shown in Figure 3.5, is the Poncelet family in a CAP pair
for which the outer and inner ellipse are homothetic to each other, i.e., a D kac and
b D kbc , where a; b and ac ; bc are outer and inner ellipse semiaxes, respectively.
Cayley implies:
Proposition 3.11. The semiaxes of a CAP pair of homothetic ellipses which admits
a 3-periodic are given by:
a b
ac D ; bc D
2 2

Proposition 3.12. The barycenter X2 is stationary at the common center and area
A is invariant and given by: p
3 3
AD ab
4
38 3. Concentric, Axis-Parallel (CAP)

Proof. Consider an affine transformation that sends both outer and inner ellipse to
a unit circle, e.g., by scaling the system along the major (resp. minor) axis by 1=a
(resp. 1=b). Uniquely amongst all triangle centers, the barycenter X2 is invariant
under affine transformations. By symmetry of the equilateral centroid, it will be
identified with the center of the homothetic pair. Affine transformations preserve
area ratios, so A will be the the area of an equilateral triangle inscribed in a unit
circle scaled by the inverse Jacobian ab. This completes the proof.
Curiously, the homothetic family shares the following invariant with the cir-
cumcircle family:
Proposition 3.13. Over the homothetic family, the sum of squared sidelengths si2
is invariant and given by:
3
X 9 2

si2 D a C b2
2
i D1

The proof below was kindly contributed by Tabachnikov (2020).

Proof. Invariant sum of squared sidelengths follows from the fact that the aver-
age of the harmonics of degree 1 and 2 over the group of rotations of order 3 is
zero. Namely, consider a unit vector v.'/ D .cos '; sin '/ and a matrix A taking
concentric circles to homothetic ellipses. Then jAv.'/j2 is a trigonometric polyno-
mial of degree 2. Average it over Z3 by adding 2=3 and 4=3 to '. The result is
independent of ', as needed. The actual value is obtained via CAS simplification
from vertex parametrization.
Referring to Figure 3.5, recall the definition of a triangle’s Brocard angle !,
given in Weisstein (2019, Brocard Angle): sidelines Pi Pi C1 rotated about Pi by
some angle  will only concur (at the first Brocard point ˝1 ) if  D !. A second,
distinct Brocard point ˝2 exists if sidelines Pi Pi 1 are rotated about Pi by !.
P relation appearing in Weisstein (ibid., Brocard Angle, Eqn. 2) is
A known
cot ! D . 3iD1 si2 /=.4A/. Therefore:
Corollary 3.5. Over the homothetic family, the Brocard angle ! is invariant. Its
cotangent is given by: p 2
3 a C b2
cot ! D
2 ab
Proof. Direct calculations using the explicit parametrization of homothetic ver-
tices.
3.5. Dual family 39

P
Another known relation valid for any triangle is cot ! D cot i :
Corollary 3.6. The homothetic family conserves the sum of its internal angle cotan-
gents.
As in Corollary 2.4, let i denote the curvature of the outer ellipse at vertex
Pi .
Proposition 3.14. Over homothetic 3-periodics, the following quantities are con-
served:
3
X 2
3
3 .a2 C b 2 /
i D
2 .ab/ 32
i D1
X 3 4
3
3 .3a4 C a2 b 2 C 3b 4 /
i D 4
8 .ab/ 3
i D1

Proof. In the homothetic pair, a 3-periodic orbit is given by:

2i  2i 
Pi D Œa cos.u C /; b sin.u C /
3 3
2
3
and ki is the following
2
3
b 4 xi2 C a4 yi2
ki D 8
.ab/ 3
The result follows by direct computation using CAS.

3.5 Dual family

The dual family, shown in Figure 3.6, is the Poncelet family in a CAP pair such
that the outer and inner ellipses are “duals” curves of each other, i.e., tangents to
one are sent to points on the other and vice versa. For ellipses, this simply implies
their aspect ratios a=b and ac =bc will be reciprocals of one another. Cayley yields:
Proposition 3.15. The caustic semiaxes of the dual family are given by:
ab
ac D  b; bc D  a; D
a2 C b2
40 3. Concentric, Axis-Parallel (CAP)

Figure 3.6: A Dual family 3-periodic (blue) interscribed in a pair of “dual” ellipse
(black, brown). Their aspect ratios are reciprocals of each other. No invariants
have yet been detected for this family other than the fact that the orthocenter X4 is
stationary at the common center. Also shown is the orthic triangle (dashed orange)
whose vertices lie at the feet of the altitudes (dashed blue). Live

Remarkably:
Proposition 3.16. The orthocenter X4 of the dual family is stationary.
Proof. Follows directly from the vertex parametrization in Proposition 3.17.
In terms of the vertices of a triangle A D Œxa ; ya , B D Œxb ; yb , C D Œxc ; yc 
the orthocenter X4 D Œx4n =A4 ; y4n =A4  is given by the following rational func-
tions

x4n D .xc xb / xa ya C .xb yb xc yc / xa C .yc yb / ya2 C yb2 yc2 ya
C xc xb . yc yb // C yb yc .yc yb /

y4n D .xb xc / xa2 C .yb yc / xa ya C xc2 xb2 xa C . xb yb C xc yc / ya
C xb xc .xb xc / C yb yc .xb xc /
A4 D .yb yc / xa C . xb C xc / ya C xb yc xc yb
The results follows from CAS-assisted simplification from the vertex parame-
trization in Proposition 3.17.
3.6. Vertex parametrization for a generic CAP pair 41

Figure 3.7: Left: Two CAP ellipses (black and brown), and a point P1 on the outer
one. The lines thru P1 tangent to the inner ellipse intersect the outer one at P2 and
P3 . Notice that P2 P3 cut thru the inner ellipse, i.e., the pair of ellipses does not
satisfy Cayley’s conditions. Right: the minor axis of the inner ellipse has been
scaled such that P1 P2 P3 is now a Poncelet triangle.

Despite much searching, no invariant quantities have yet been found for this
family.

3.6 Vertex parametrization for a generic CAP pair

Consider a general CAP pair of ellipses denoted E and Ec . We will derive a generic
parametrization for the vertices of 3-periodics in such a pair. A first calculation
will be helpful. Referring to Figure 3.7(left):

Proposition 3.17. The intersections P2 and P3 on E of the two tangents to Ec seen

from a point P1 D Œx1 ; y1  also on E are given by:

 
1 p1 x1 C p2 y1 w1 x1 C w2 y1
P2 D Œx2 ; y2  D ;
k2 b a
 
1 p1 x1 p2 y1 w1 x1 C w2 y1
P3 D Œx3 ; y3  D ;
k2 b a
42 3. Concentric, Axis-Parallel (CAP)

CAP family ac
bc
note

confocal a ı b 2 =c 2 b a2 ı =c 2 Proposition 2.2
incircle .ab/=.a C b/ .ab/=.a C b/ Proposition 3.3
circumcircle choose ac < R R ac RDaDb
homothetic a=2 b=2 Proposition 3.11
dual .ab 2 /=.a2 C b 2 / .a2 b/=.a2 C b 2 / Proposition 3.15
conf. .ıe ae2 3be2 /ae .3ae2 Cbe2 ıe /be
Proposition 3.1
excentrals 2ce2 2ce2

Table 3.1: Values for the caustic semiaxes ac ; bc to be used in the generic vertex
parametrization for a CAP pair in Proposition 3.17

p1 D b a4 bc4 .a2 ac2 /2 b 4


p2 D 2a .a2 C ac2 /b 2 a2 bc2 k1
q
k1 D b 2 bc2 .a2 ac2 /x12 C ac2 a2 .b 2 bc2 /y12
 2 2 2  2 2 2
a .b C bc2 / ac2 b 2 a .b bc2 / C ac2 b 2
k2 D x1 C y1
a b


w1 D 2b .b 2 C bc2 /a2 ac2 b 2 k1


w2 D a ac4 b 4 a4 .b 2 bc2 /2

Parametrizations for specific CAP families can be obtained from Proposition 3.17
by setting the caustic semiaxes ac , bc as in Table 3.1.

3.7 Summary
Fixed points and (known) conserved quantities for the concentric, axis-parallel
(CAP) families in this chapter appear in Table 3.2.
Also of interest is data about caustics, regarded as a family’s fixed inconic,
shown in Table 3.3.
3.7. Summary 43

Family Fixed Conserves Notes

L, J , r=R,
Confocal X9 P 2=3 i.e., billiard 3-periodics
cos i ; i
P sum of cosines same as
Incircle X1 R, cos i
confocal affine pre-image
P Q product of cosines same as
si2 , cos i ,
Circumcircle X3 excentrals’ in confocal affine
rh ,Rh ,.i0 /2=3
Q pre-image
0 =A, 0
Confocal AP Q 0i ,
cos primed quantities refer to those
X6 0 2
Excentrals .sP i/ = si of the excentral family
2
A, si , !,
Homothetic X2 P 2=3 4=3 affine image of concentric circles
cot i , i , i
Dual X4 n/a

Table 3.2: Summary of fixed points and (known) conserved quantities for the con-
centric, axis-parallel (CAP) families mentioned in this chapter.

Poncelet caustic Brianchon caustic contact tri

center
family (inconic) point contact tri [Link]/*
incircle X1 incircle X7 intouch 3tYYu3h
homothetic X2 Steiner X2 medial 3474753
circumcircle X3 ? X69 X69 -cev. 2T3qu9f
dual X4 ? X253 X253 -cev. 2SUfomB
excentral X6 orthic X4 orthic 3uXXI7H
confocal X9 Mandart X8 extouch 3wiBeyv

Table 3.3: Information about the caustic to various CAP families, regarded as a
fixed inconic. The Brianchon point is the perspector to the triangle whose vertices
are at the touchpoints with the inconic, see Weisstein (2019, Brianchon point).
When named, this triangle appears in the “caustic contact tri” column. An link to
an animation for each case is provided.
44 3. Concentric, Axis-Parallel (CAP)

Figure 3.8: Normals to the ellipse at vertices of 3-periodics in the homothetic

family concur at the orthocenter X4 .

3.8 Exercises
Exercise 3.1. Prove that the power of the circumcircle with with respect to the
common center in each of the following 3-periodic families is constant and given
by the listed expressions. (i) incircle: ae be ; (ii) homothetic: .ae2 C be2 /=2, and
(iii) excentral: a2 b 2 2ı.
Exercise 3.2. Prove the radius r  of the stationary cosine circle of the excentral
family is larger than the major axis a of its caustic.
Exercise 3.3. Prove Lemma 3.3.
Exercise 3.4. Derive the proof details to the 2nd part of Proposition 3.10.

Exercise 3.5. (contributed by L. Gheorghe) Prove that ellipse normals at vertices

of 3-periodics in the homothetic family are concurrent at the orthocenter X4 , see
Figure 3.8. Derive the semiaxes of the elliptic locus of X4 as a function of a; b of
the outer ellipse, see it Live.

Exercise 3.6. The Thomson cubic is the locus of centers of circumconics such
that normals at the vertices concur (on the Darboux cubic), see Gibert (2021a,
Darboux and Thomson cubics). Prove that vertex normals off of the X1 - and X6 -
centered circumconics concur on X84 and X64 , respectively. This readily implies
normals to the outer ellipse at the incircle and excentral family vertices will concur
on said points; see Table 3.4.
3.8. Exercises 45

Poncelet norms. concur

center
family concur [Link]/*
incircle X1 X84 3eVuCQY
homothetic X2 X4 3eXSRhC
circumcircle X3 X3 2RqMqul
dual X4 X3346 n/a
excentral X6 X64 3hwCTfN
confocal X9 X1 3uTvqLI

Table 3.4: CAP families studied herein. Coincidentally, their centers lie on the
Thomson cubic which is the loci of circumconic centers such that normals at
vertices concur, see Gibert (2021a, Thomson Cubic). The third column lists the
experimentally-found concurrence points. These lie on the Darboux cubic de-
scribed in Gibert (2021a, Darboux cubic).

Exercise 3.7. Recall the dual family has stationary orthocenter X4 . Prove that the
inversive image of the dual family wrt to a circle concentric with the ellipse pair is
a non-Ponceletian family with incenter X10 stationary at the common center. This
inversive family is inscribed in Booth’s curve and its caustic can contain multiple
spikes; see it Live.

Exercise 3.8. Given a reference triangle T , its tangential triangle T 0 has sides
tangent to the circumcircle at the vertices of T . A known fact is that the sides of
T 0 are parallel to those of the orthic Th of T , see Weisstein (2019, Tangential Tri-
angle). For any acute triangle T , the Gergonne point X70 of the tangential triangle
coincides with the symmedian X6 of T , see Weisstein (ibid., Contact Triangle).
Let T denote the Poncelet family of excentral triangles. We’ve seen above that
(i) this family is all acute, and that (ii) its symmedian point X6 is stationary. Let T 0
denote their tangential triangles. This family will be non-Ponceletian: its vertices
do not sweep a conic nor do its sides envelop one.
Since the T are all acute, they can be thought of as the contact triangles of
the T 0 . Therefore the Gergonne point X70 of the tangentials to the excentrals will
coincide with X6 of the excentrals and be stationary, see Weisstein (ibid., Contact
Triangle).
Prove that the ratio of homothety between the orthics (billiard 3-periodics) and
the tangentials is invariant. Corollaries: (i) the T 0 conserve perimeter; (ii) they
conserve the same r=R as the excentral orthics, i.e., the corresponding billiard
3-periodics. See it Live.
46 3. Concentric, Axis-Parallel (CAP)

Also prove that the locus of X90 of the T 0 is an ellipse.

Derive equations for the curves swept by the vertices of T 0 as well as their
caustic.

3.9 Research questions

Question 3.1. Referring to Figure 3.6, are there any conserved quantities for the
dual family besides stationarity of X4 at the common center?
Question 3.2. Referring to the dashed green triangle in Figure 3.1(middle), are
there any conserved quantities and/or fixed triangle centers for the family which
is an s-affine image of billiard excentrals?
Question 3.3. Consider the homothetic family and its polar image with respect
to a focus of the outer ellipse E. Prove that (i) the caustic is a circle, derive its
location and radius. (ii) the family is inscribed in a conic, namely, below (resp.
above) a certain aspect ratio a=b of E, the conic is an an ellipse (resp. hyperbola).
(iii) the Gergonne point X7 of the family is stationary. Live: family inscribed in
ellipse, hyperbola.
Question 3.4. Prove that a necessary condition for a triangle center to be sta-
tionary over Poncelet 3-periodics is that it lies on the Thomson Cubic, defined in
Gibert (2021b).
Question 3.5. Xk , k D 1; 2; 3; 4; 6; 9 lie on the Thomson cubic K002 and as
seen above, are stationary over Poncelet families centered on them. Referring to
Figure 3.9, prove that X1249 , also on K002 , is stationary over Poncelet 3-periodics
centered on it. Show that in the X1249 -centered system, X4 (resp. X20 ) is the
perspector of the circumconic (resp. inconic).
Question 3.6. The perspector of a circumconic centered on X is the X2 -Ceva
conjugate of X. Likewise, the perspector (Brianchon point) of an inconic centered
on X is the isotomic conjugate of the anticomplement of X . Given a triangle, let
C and I be the X -centered circumconic and inconic. Show that for a triangle,
the inconic perspector (i.e., the Brianchon point) is the anticomplement of the
circumconic one.
Question 3.7. In Gibert (ibid.), it is stated that if Xk lies on the Thomson cubic
K002 , circumconic normals at the vertices (and inconic normals at the contact
points) concur. Consider the (axis-parallel, concentric) circumconic and inconic
3.9. Research questions 47

Figure 3.9: A Poncelet 3-periodic (blue) interscribed between the X1249 -centered
circumconic C (black) and inconic I (brown); over the family, said center remains
stationary. Since the center lies on the Thomson cubic, (i) C and I are axis-parallel,
(ii) the normals to C at the vertices concur (at X20 in this case), and (iii) the normals
to I at the contact points also concur (at X1498 ). It turns out X20 doubles up as
the Brianchon point of I, and X4 is the perspector of C, i.e., the perspector of its
polar triangle (magenta) with respect to C. Video

pair centered on X1249 . Prove that the circumconic (resp. inconic) normals at the
vertices (resp. contact points) meet at X20 (resp. X1498 ).
Question 3.8. Gibert (ibid.) lists the following triangle centers as lying on the
Thomson cubic: Xk , k D57, 223, 282, 1073, 3341, 3342, 3343, 3344, 3349,
3350, 3351, 3352, 3356, 14481. Experimentally they are not stationary over Pon-
celet 3-periodics centered on them. Why is that? Conversely, why is it that Xk ,
k D1,2,3,4,6,9,1249 , also on the Thomson, can be stationary?

Question 3.9. Given a triangle, compute its X7 -centered inconic and circumconic.
Prove that 3-periodics interscribed in said conics will not maintain X7 stationary.
Prove the same by taking the conic pair’s center to be X8 and X10 , i.e., in neither
of these cases will the original center remain stationary. Report the Brianchon
point for all said inconics.
Question 3.10. Prove Proposition 3.9.
4
Non-concentric,
Axis-Parallel
(NCAP)

Here we introduce a few Poncelet triangle families interscribed in non-concentric,

axis-parallel (NCAP) ellipse pairs. These are known as (i) the poristic family (in-
terscribed between two non-concentric circles), (ii) the poristic “excentral” fam-
ily, comprising the excentral triangles to the poristic family, and (iii) the Brocard
porism, a special family whose Brocard points remain stationary (defined and ex-
plained below). In each case we describe their geometry, present properties and
invariants, and propose a vertex parameterization. In Section 4.5, properties and
fixed points for the families treated herein are summarized. In Figure 4.11 we or-
ganize all families studied in this chapter and in Chapters 2 and 3, grouping them
by “similarity” of properties and invariants.

4.1 Poristic family (Bicentric triangles)

Poristic triangles, shown in Figure 4.1, are the simplest case of Poncelet’s porism:
a 1d family of triangles with fixed incircle and circumcircle. They are also known
as the (N D 3) bicentric family.
First described by Chapple (1746), the family was later studied by both Euler
4.1. Poristic family (Bicentric triangles) 49

Figure 4.1: Poristic Triangle family (blue): fixed incircle (green) and circumcircle
(purple). Left: a few poristic triangles (blue and dashed blue) in a pair of circles
such that d < r, i.e., all poristic triangles are acute. Right: the same but with
d > r; since X3 can be either interior or exterior to the family, both acute and
obtuse triangles will be present. Live
50 4. Non-concentric, Axis-Parallel (NCAP)

Figure 4.2: Over the poristic family, the antiorthic axis (solid blue) is stationary
and perpendicular to X1 X3 . Video.

and Poncelet. The so-called Euler’s triangle formula1 , constrains the distance d
between incenter X1 and circumcenter X3 as follows:

d 2 D R.R 2r/ (4.1)

where r; R are the radii of outer and inner circle. Referring to Figure 4.1:

Proposition 4.1. The Poristic family will contain obtuse triangles iff d > r.

Proof. This stems from the fact that when d < r, X3 is always interior to the
incircle, i.e., the caustic of the Poncelet family.

In consonance with both billiard 3-periodics and the incircle family:

Proposition 4.2. The poristic family conserves the sum of its internal angle cosines.

Proof. Direct application of Equation (2.4), noting by definition r=R is constant.

In Weisstein (2019, Antiorthic axis), the antiorthic axis is defined as containing

the three intersections of a triangle’s sidelines with those of the excentral triangle.
As illustrated in Figure 4.2, the following was proved by Weaver (1927):
1 Chapple had stated it in 1746, Euler in 1765, and Poncelet’s porism was published in 1822, see
Del Centina (2016a).
4.1. Poristic family (Bicentric triangles) 51

Proposition 4.3. The antiorthic family is stationary over the poristic family and
perpendicular to X1 X3 .
Let a first vertex P1 of the poristic family be parametrized by P1 .t / D RŒcos t; sin t.
Proposition 4.4. The perimeter L.t/ of poristic triangles is given by:

p
3 R2 4 dR cos t C d 2 3 R2 C 2 dR cos t d 2
L.t/ D p
R R2 2 dR cos t C d 2
Proof. Follows directly computing the 3-vertices explicitly and using L.t/ D
jP1 P2 j C jP2 P3 j C jP3 P1 j and simplifying it with a CAS.
It turns out that poristic triangles can be regarded as the image of billiard 3-
periodics (and vice versa) under (i) a variable similarity transform, and (ii) a polar
transformation wrt to a focus-centered circle. We now proceed to prove these
results, but first we will need a couple of lemmas. In Odehnal (2011, page 17),
one finds the following result, illustrated in Figure 4.3:
Lemma 4.1. Over the poristic family, the locus of the Mittenpunkt X9 is a circle
with radius is Rd 2 R=.9R2 d 2 / centered on X1 C.X1 X3 /.2R r/=.4RCr/ D
d.3R2 C d 2 /=.9R2 d 2 /.
In fact we can derive X9 .t/ explicitly:
Lemma 4.2.
2   3


2
d 4 d c .Rct d / r .3 d ct C R/ r 2 4Rd 2 st R2 .2 Rct d /2
X9 .t/ D 4 ; 5
.4 R C r/ .d ct R C r/ .R2 C d 2 2 dRct / .9 R2 d 2/

where ct and st are shorthand for cos.t/ and sin.t / respectively.

Let Pi D Œxi ; yi  denote the vertices of billiard 3-periodics and Pi0 D Œxi0 ; yi0 
those of a poristic family, i D 1; 2; 3.
Theorem 4.1. The Pi0 are an image of the Pi under a variable similarity transform
comprising of (i) a rigid rotation by .t/, (ii) a rigid translation by X9 .t/, and (iii)
uniform scaling by L.t/. These are given by:

xi0 DL.t/.cos .t/xi C sin .t/yi C x9 .t //

yi0 DL.t/. si n.t /xi C cos .t /yi C y9 .t//
.1 cos t/.R C d 2R cos t /
tan .t / D
.2R cos t C R d / sin t
52 4. Non-concentric, Axis-Parallel (NCAP)

Figure 4.3: A poristic triangles (blue) is shown along as its circumbilliard (dashed
magenta) whose aspect ratio is invariant. The locus of the Mittenpunkt X9 is a
circle (red). Video, Live
4.1. Poristic family (Bicentric triangles) 53

Proof. CAS-assisted simplification.

Reznik and Garcia (2021a) defines the “circumbilliard” E9 of a triangle as the

circumellipse centered on X9 . Let a9 ; b9 its semiaxes. CAS manipulation yields:

Corollary 4.1. Over the poristic family, a9 .t/ and b9 .t/ are given by:
p
R 3 R2 C 2 dR d 2
a9 DL.t/
R2 d 2
9p
R R d
b9 DL.t/ p
3 R C d .3 R d /
q p
2 2 2R dR
c9 D a9 b9 D L.t/ 2 :
9R d2

Corollary 4.2. The ratios a9 .t/=L.t/, b9 .t /=L.t /, and c9 .t /=L.t / are invariant
over the Poristic family.

Corollary 4.3. Over the poristic family, the aspect ratio of the (varying) circum-
billiard is invariant and given by:
s
a9 .t/ .R C d / .3R d /
D
b9 .t/ .R d / .3 R C d /

Recall the definition of the polar of a point P with respect to a circle C, men-
tioned in Weisstein (2019, Polar): it is defined as the line perpendicular to OP
which contains the inversion of P wrt to C. Dually, the pole of a line L with re-
spect to C is the inversion of the foot of the perpendicular dropped from O onto P
wrt to C. So given a smooth curve, we can speak of its polar image with respect
to a circle as the set of poles of the curve’s tangents with respect to C.
The fact that the polar image of an ellipse with respect to a focus is a circle
is a well-known result, mentioned in Akopyan and Zaslavsky (2007) and Glaeser,
Stachel, and Odehnal (2016).
Let E and E 0 be a confocal ellipse pair centered at Œ0; 0, with major axes along
x. Let a; b and a0 ; b 0 denote their major and minor semiaxes, respectively. The
foci f1 and f2 are at Œ˙c; 0, where c 2 D a2 b 2 . A known classical result which
we reproduce below is:
54 4. Non-concentric, Axis-Parallel (NCAP)

Figure 4.4: The poristic family (orange) is the polar image of billiard 3-periodics
(blue) with respect to a circle (dashed gray) centered on one of the foci of the
confocal pair (f1 in the picture). Live
4.2. Poristic excentrals 55

Lemma 4.3. The polar image of the E; E 0 pair with respect to a circle of radius 
centered on f1 is a pair of nested circles Ci nt ; Cext with centers given by:
c 2 c
Oi nt D Œ c 2 ; 0; O ext D Œ c  ; 0
b2 b 02
Their radii r; R and distance d between their centers are given by:

a 2 a
0 c .a2 a0 2 /
r D 2 ; R D  ; d D 2
b2 b 02 b2 b02
Referring to Figure 4.4:
Corollary 4.4. The poristic family is the polar image of billiard 3-periodics with
respect to a circle centered on a focus.
Corollary 4.5. The sum of cosines of the polar image of billiard 3-periodics with
respect to a focus-centered circle is given by:
X r ab 02
cos  0 D 1 C D1C 0 2 (4.2)
R ab
Given a triangle, an inconic is is fully defined by its center and is tangent to
the three sidelines, see Weisstein (2019, Inconic).
Referring to Figure 4.5, let E1 be the X1 -centered circumconcic to the poristic
family, i.e., it contains the vertices. Let 1 , and 1 denote its semiaxes. Interest-
ingly:
Proposition 4.5. 1 D R C d and 1 D R d are invariant over the poristic
family, i.e., E1 rigidly rotates about X1 .
A proof appears in Garcia and Reznik (2021, Appendix C).

4.2 Poristic excentrals

The family of excentral triangles to the poristic family, shown in Figure 4.6 is also
Ponceletian: in Odehnal (2011), it is shown to be inscribed in a circle of radius
2R where R is the circumradius of its reference poristic family, centered on X40 ,
the Bevan point, or X30 of the family in questions (to avoid confusion, we will be
priming quantities associated with this family).
Proposition 4.6. The barycenter X20 of poristic excentrals is stationary.
56 4. Non-concentric, Axis-Parallel (NCAP)

Figure 4.5: A poristic triangle (blue) is shown, as well as I1 , the X1 -centered

inconic (light blue). Also shown is the excentral triangle (green), the circle (or-
ange) the excentral family is inscribed in and their MacBeath caustic (dashed
green). Also shown is I30 (dark red), the X30 -centered excentral inconic (red). Note
X30 D X40 . Over the poristic family, both I1 and I30 rotated rigidly at 90-degrees
from each other. Video
4.2. Poristic excentrals 57

Figure 4.6: The poristic family (blue) is interscribed between two fixed circles, i.e.,
their circumcenter X3 and incenter X1 are stationary. The family of its excentral
triangles (solid green) are inscribed in a circle (orange) centered on the Bevan point
X40 and of radius twice the original circumradius. This family circumscribes the
MacBeath inellipse (dashed orange), centered on X3 with foci on X1 and X40 . A
second for both poristics and excentrals configuration is also shown (dashed blue
and dashed green). Video, Live
58 4. Non-concentric, Axis-Parallel (NCAP)

excentral excentral reference

MacBeath center center
center X50 X3
focus X40 X1
focus X30 X40

Table 4.1: Center and foci of the MacBeath inconic of an excentral triangle and
the corresponding triangle center of the reference.

Proof. Recall a triangle’s barycenter X2 is a third of the way from the circumcenter
to the orthocenter, see Weisstein (2019, Euler Line, Eqn. 6). The result follows
from the fact that both X30 D X40 and X40 D X1 are stationary.
The MacBeath inconic, defined in Weisstein (ibid., MacBeath Inconic), is an
ellipse centered on a triangle’s 9-point center X5 , with foci at the circumcenter X3
and orthocenter X5 . Poristic excentrals are Ponceletian since:
Proposition 4.7. The MacBeath inconic to the excentral poristics is stationary and
is therefore the caustic. Let 05 and 50 denote its major and minor semiaxes. These
are given by: p
05 D R; 50 D R2 d 2
Proof. It is straightforward to verify the sidelines of poristic excentrals are dynam-
ically tangent to the ellipse:
.x d /2 y2
C D1
R2 R2 d2
with center X3 D .d; 0/ and foci X40 D .0; 0/ and X1 D .2d; 0/.
Correspondences between the centers and foci of the excentral MacBeath in-
conic and those of the reference
p triangle appear in Table 4.1.
0 0 2
Since 5 =5 D R= R d 2 , use Equation (4.1) to obtain:
Corollary 4.6. The aspect ratio of the caustic to the excentral poristics is given
by: r
05 R
0 D
5 2r
As shown in Figure 4.5, let I30 be the X30 -centered inconic to poristic excentrals.
Let 03 and 30 denote its major and minor semiaxes, respectively.
4.3. The Brocard porism 59

Proposition 4.8. Over poristic excentrals, 03 D R C d and 30 D R d are

invariant, i.e., I30 rigidly rotates about X30 .

We omit the long proof kindly contributed by B. Odehnal and appearing in

Garcia and Reznik (2021, Appendix C).
Interestingly:

Theorem 4.2. Excentral poristics are the image of the circumcircle family under
a variable rigid rotation. The rigidly-rotating I30 is identified with the caustic of
the circumcircle family.

Proof. Recall Proposition 3.8: the orthic triangles of the circumcircle family has
invariant inradius and circumradius. Also recall Lemma 3.3: the locus of the orthic
circumcenter is a circle concentric with the common center. Also notice in the
circumcircle family, the caustic is the stationary inconic centered on X3 .

4.3 The Brocard porism

A property-rich family of Poncelet triangles is the so-called “Brocard porism”, in-
troduced in Bradley (2011), Bradley and Smith (2007), and Shail (1996). It is
inscribed in a circle and circumscribes the so-called Brocard inellipse. Remark-
ably, its foci coincide with the stationary Brocard points ˝1 and ˝2 of the family;
see Figure 4.7.
Let R denote the radius of the outer circle and a0 ; b 0 the caustic semiaxes, with
.c / D .a0 /2 .b 0 /2 . Let also, d D jX3 X39 j the distance between the centers
0 2

of the circle and of the caustic.

Proposition 4.9. A pair of circle (outer) .x x0 /2 Cy 2 D R2 and ellipse (caustic)

.x x1 /2 =.a0 /2 C y 2 =.b 0 /2 D 1 admit a 1d family of Poncelet triangles if and
only if
.a0 /2 2Ra0 .b 0 /2 C R2 d 2 D 0; d D jx1 x0 j

Proof. Follows from Cayley condition for 3-periodics in an NCAP pair.

Proposition 4.10. For any triangle T , the circumcircle and Brocard inellipse are
Ponceletian, they admit a 1d family of Poncelet triangles. Furthermore, their Bro-
card inellipse is stationary.

Proof. Follows from vertex parametrization for the Brocard family and/or from
stationarity of X6 , see Proposition 4.13.
60 4. Non-concentric, Axis-Parallel (NCAP)

Proposition 4.11. The stationary circumcenter X3 and circumradius R are given

by:  
c 0 ı1 2.a0 /2
X3 D 0; ; R D
b0 b0
p
where ı1 D 4.a0 /2 .b 0 /2 .
The following is a known requirement for the Brocard porism to be possible,
appearing in Shail (1996, Eqs. 15–17):
Corollary 4.7. R⩾2c 0
Remarkably, and echoing a property seen above for the homothetic family,
leaving the proof as an exercise:
Proposition 4.12. Over the Brocard porism, the Brocard angle ! is invariant and
given by:
ı1 p
cot ! D 0 ⩾ 3
b

P3 Henceforth, we shall use symbol ш for cot !. Recall for any triangle ш D
i D1 cot i , i.e.:

Corollary 4.8. The Brocard porism conserves the sum of its internal angle cotan-
gents.
Shail (ibid.) derives the distance between Brocard points (the foci of the Bro-
card inellipse), in terms of invariant R and !:

j˝1 ˝2 j2 D 4R2 sin2 !.1 4 sin2 !/ D .c 0 /2 (4.3)

Corollary 4.9. p
c 0 D R sin ! 1 4 sin2 !
Referring to Figure 4.7, all of ˝1 , ˝2 , X3 , and X6 are concyclic on the so-
called Brocard circle, see Weisstein (2019, Brocard Circle), whose center is X182 .
The Brocard axis is defined in Weisstein (ibid., Brocard Axis) as the line containing
the circumcenter X3 and symmedian point X6 of a triangle.
Proposition 4.13. Over the Brocard porism, the following 3 objects are stationary:
(i) the Brocard circle, (ii) the Brocard axis, and (iii) the symmedian point X6 are
stationary.
4.3. The Brocard porism 61

Figure 4.7: A triangle (blue) in the Brocard porism is shown inscribed in an outer
circle (black) and having the Brocard inellipse (brown) as its caustic, with foci at
the stationary Brocard points ˝1 and ˝2 of the family, and centered on the Brocard
midpoint X39 . The Brocard points as well as the stationary circumcenter X3 and
symmedian point X6 are concyclic on the Brocard circle (dashed magenta), whose
center is X182 . Live
62 4. Non-concentric, Axis-Parallel (NCAP)

Proof. The Brocard circle is stationary since it passes through 3 stationary points:
˝1 ; ˝2 ; X3 are stationary. The Brocard axis is stationary since it contains station-
ary X3 and stationary Brocard midpoint X39 . For any triangle, X6 is antipodal to
X3 on the Brocard circle.
Recall the two isodynamic points X15 and X16 of a triangle as the two unique
intersections of the 3 Apollonius circles2 . X15 (resp. X16 ) is interior (resp. exte-
rior) to the circumcircle. In fact they are inverse images of each other with respect
to the latter, see Weisstein (2019, Isodynamic Points).
Proposition 4.14. Over the Brocard porism, the two Isodynamic points X15 and
X16 are stationary and given by:
" p # " p #
R. 3 ш/ R. 3 C ш/
X15 D 0; p ; X16 D 0; p
ш2 3 ш2 3

Proof. Let P and U be finite points on a triangle’s plane with normalized barycen-
tric coordinates .p; q; r/ and .u; v; w/, respectively. Let f and g be homogeneous
functions of the sidelengths. The .f; g/ barycentric combo of P and U , also de-
noted f  P C g  U , is the point with barycentric coordinates .f p C g u; f q C
g v; f r C g w/. In Kimberling (2019, X(15), X(16)), the following combos (see
below), derived by Peter Moses, are provided:

p
X15 D 3  X3 C ш  X6
p
X16 D 3  X3 ш  X6
With all involved quantities invariant, the result follows.
Proposition 4.15. The semiaxes a0 and b 0 and center X39 of the Brocard inellipse
are given by:
 
0 0
 2
 1 2
Œa ; b  DR sin !; 2 sin ! D R p ;
1 C ш2 1 C ш2
" p #
Rш ш2 3
X39 D 0;
ш2 C 1

2 These are circles which contain a vertex and the intersection of the corresponding internal and
external bisectors with the opposite side.
4.3. The Brocard porism 63

Proof. Consider a triangle T with sidelengths s1 ; s2 ; s3 , area A, and circumradius

R. The following identities appear in Bradley and Smith (2007) and Shail (1996):
s1 s2 s3 2A
RD ; sin ! D p (4.4)
4A
where D .s1 s2 /2 C .s2 s3 /2 C .s3 s1 /2 . Bringing in Equation (4.3), the result
follows from combining the above into expressions for the Brocard inellipse semi-
axes, given in Weisstein (2019, Brocard Inellipse):
s1 s2 s3 2s1 s2 s3 A
a0 D p ; b0 D (4.5)
2

With the results above, we can derive the following quantities and centers ex-
plicitly:

" p #
R ш2 3
X6 D 0;
ш
p
R ш2 3
jX3 X6 j D
pш
R ш2 3
˝1;2 .R; ш/ D Œ˙1; ш
ш2 C 1
Let a0 be the major axis of a generic triangle’s Brocard inellipse. Interestingly,
we have:

Lemma 4.4.
X 3
1 1
2
D
s 4.a0 /2
i D1 i

Proof. As
1 1 1 s22 s32 C s12 s32 C s12 s22
C C D
s12 s22 s32 s12 s22 s32
the result follows from Equation (4.5).

Therefore, we have:
64 4. Non-concentric, Axis-Parallel (NCAP)

Figure 4.8: Over the homothetic family, the Brocard inellipse has semiaxes of
variable lengths but invariant aspect ratio. Video

Corollary 4.10. Over the Brocard porism, the sum of inverse squared sidelengths
is invariant
Similarity and Polar image: As it will be seen below, the Brocard porism is
the image of the Homothetic family under two different transformations: variable
similarity and polar.
Referring to Figure 4.8, we will first prove a handy lemma, introduced in
Reznik and Garcia (2021b):
Lemma 4.5. Over the homothetic family, though the semiaxes of the Brocard in-
ellipse have variable lengths, their ratio ˇ is invariant and given by:
p
3a4 C 10a2 b 2 C 3b 4
ˇD >1
4ab

Proof. The result follows from combining Equation (4.4) with Equation (4.5), us-
ing the sidelengths si of the homothetic family using the parametrization in Sec-
tion 2.7.1.
The result below was introduced in Reznik and Garcia (ibid., Thm 4.1):
4.3. The Brocard porism 65

Proposition 4.16. The Brocard family is the image of the homothetic one under a
variable similarity transform.
Proof. Consider the following similarity transform which sends points X in the
plane of the homothetic family to new ones X 0 :

X 0 D Scale.1=b 00 /:Rot . /:T ransl. X39 /:X

where b 00 is the variable minor semiaxis length of the the (moving) Brocard in-
ellipse E 00 in the homothetic pair,  the angle between said minor axis and the
horizontal, and X39 the moving center of E 00 . Clearly, E 00 will be sent to an origin-
centered ellipse which is axis-parallel to the homothetic ones. By Lemma 4.5, the
aspect ratio ˇ of E 00 is invariant over the homothetic family, implying the trans-
formed inellipse will have fixed axes .ˇ; 1/. Notice its circumcenter and circum-
radius are prescribed by the semiaxes of the caustic (see Proposition 4.11). This
completes the proof.
Referring to Figure 4.9, let a; b be the semiaxes of the outer ellipse in the
homothetic pair.
Proposition 4.17. The Brocard porism is the polar image of the homothetic family
with respect to a circle centered on a caustic focus f 0 . The symmedian point X6
of the image coincides with f 0 . Its outer circle and ellipse are given by:

C W .x x0 /2 C y 2 D R2
.x x1 /2 y2
EW C D1
.a0 /2 .b 0 /2
c.b 2 C 42 / c.4a2 c 2 C 42 /
x0 D ; x 1 D
2b 2 2.4a2 c 2 /
4a2 2a2 2a2
.a0 / D 2 ; .b 0
/ D p ; R D
4a c2 b 4a2 c 2 b2
Here b 0 > a0 .
Proof. Proof is left as an exercise.
Remark 4.1. From the relations obtained in Proposition 4.17 it follows that
p
0 2 .a0 /2 / 0 2 .a0 /2
2 .4.b / 2 4.b /
aD ; b D  p
3a0 .b 0 /2 3.b 0 /2
66 4. Non-concentric, Axis-Parallel (NCAP)

Figure 4.9: The Brocard family (magenta) is the polar image of the homothetic
family (solid blue) with respect to a circle (dashed gray) centered on a focus of the
homothetic caustic (light brown) which is sent to the Brocard circumcircle (dashed
magenta). The outer ellipse (black) is sent to the Brocard inellipse (green). Live
4.3. The Brocard porism 67

Since two polar transformations with respect to the same circle is the identity:

Corollary 4.11. The homothetic family is the polar image of the Brocard family
with respect to its stationary symmedian point X6 .

Corollary 4.12. In terms of the homothetic pre-image semiaxes a; b, the invariant

sum of inverse squared sidelengths and Brocard angle are given by:

X 3
1 1 b 2 .3a2 C b 2 /
D D
s2
i D1 i
4.b 0 /2 164 a2
p a
cot ! D ш D 3
b
0 0
p By Proposition 4.15 and Proposition 4.17 it follows that sin ! D .a /=.2b / D
Proof.
b= 4a2 c 2 : Using that csc2 ! cot2 ! D 1 the result follows.

4.3.1 A digression: equilateral isodynamic pedals

Referring to Figure 4.10, the pedal (resp. antipedal) triangles of the isodynamic
points X15 and X16 (resp. isogonic points X13 and X14 ) are equilateral triangles
centered on X396 and X395 (resp. X5463 and X5464 ). These facts appear in Kim-
berling (2019).
Let A denote the area of a triangle and Ak , k D 13; 14; 15; 16 denote the area
of said equilaterals. Moses (2020) has kindly contributed the following expres-
sions:

Proposition 4.18.
2ш 2ш
A13 =A D2 C p ; A14 =A D 2 p
3 3
p p
3 C 3ш 3 3ш
A15 =A D ; A 16 =A D
2.csc2 ! 4/ 2.csc2 ! 4/

Corollary 4.13. For any triangle:

p
A13 A16 шC 3
D D p
A14 A15 ш 3
68 4. Non-concentric, Axis-Parallel (NCAP)

Figure 4.10: The pedal (resp. antipedal) triangles (orange, resp. purple) of the iso-
dynamic points X15 and X16 (resp. isogonic points X13 and X14 ) are equilaterals
centered on X396 and X395 , collinear with X6 (resp. X5463 and X5464 , collinear
with X3 ), see Kimberling (2019). Over the porism, the area ratios A16 =A15 and
A13 =A14 are invariant and identical. The loci of X396 and X395 are circles (dashed
orange) as are those of X5463 and X5464 (not shown). Video
4.4. Vertex parametrization 69

Corollary 4.14. The Brocard porism conserves A13 =A14 and A16 =A15 .
The centroid of the pedal triangle of X15 (resp. X16 ) is X396 (resp. X395 ).
Proposition 4.19. The locus of X15 and X16 pedal centroids X396 and X395 are
the following circles:

2  p  p 3
2
R 3.ш C 1/ 2 3 ш ш C 3 p
R. 3ш C 3/
C395 D 40; p
5 ; r395 D
3 ш2 3 ш2 C 1 3.ш2 C 1/
2  p  p 3
R 3 .ш2 C 1/ C 2 3 ш ш 3 p
C396 D 40; p
5 ; r396 D R. 3ш 3/
3 ш2 3 ш2 C 1 3.ш2 C 1/

Proof. Obtained via CAS.

Remark 4.2. Notice the ratio r395 =r396 is equal to A396 =A395 .
Still referring to Figure 4.10:
Proposition 4.20. The locus of X13 and X14 antipedal centroids X5463 and X5464
are the following circles:

4.4 Vertex parametrization

4.4.1 Poristic family
Consider a pair of circles x 2 Cy 2 D R2 , .x d /2 Cy 2 D r 2 , with d 2 D R.R 2r/.
Then a 3-periodic orbit is parametrized by:

P1 D Œx1 ; y1 
1  
P2 D 2 4Rr 2 qy1 w1 w2 ; 2rRw1 y1 C 2rqw2
w
1  
P3 D 2 4Rr 2 qy1 w1 w2 ; 2rRw1 y1 2rqw2
w
p
q D R2 r 2 2dx1 C d 2 ; w D R2 2dx1 C d 2
w1 D R2 2r 2 2dx1 C d 2 ; w2 D .R2 C d 2 /x1 2R2 d
70 4. Non-concentric, Axis-Parallel (NCAP)

4.4.2 Poristic excentrals

Its vertices sweep a circle centered on X40 of the original poristic family, so we
omit the parametrization.

4.4.3 Brocard porism

Consider an isosceles Poncelet triangle T D ABC in the Brocard porism, where
AB is tangent to E at one of its minor vertices. Let jABj D 2d and the height be
h. Let  D d 2 C h2 . Let the origin .0; 0/ be at its circumcenter X3 . Its vertices
will be given by:
     
d2 h2 d2 h2 
AD d; ; B D d; ; 0;
2h 2h 2h

Proposition 4.21. The Brocard porism containing T as a Poncelet triangle is de-

fined by the following circumcircle K0 and Brocard inellipse E:


K0 W x 2 C y 2 R2 D 0; R D
2h
EW 64d 2 h4 x 2 4h2 .9d 2 C h2 /y 2 C 4h.3d 2 C h2 /.3d 2 h2 /y
.d 2 h2 /.9d 2 h2 / 2 D 0

Proof. The proof follows from T , and isosceles Poncelet triangle. Recall that the
Brocard inellipse is centered at X39 . Its perspector is X6 , i.e., it will be tangent to
T where cevians through X6 intersect it, see Weisstein (2019, Brocard inellipse).

Consider the pair: circle x 2 C y 2 D R2 D .d 2 C h2 /2 =.4h2 / and ellipse

x 2 =a2C .y y0 /2 =b 2 D 1, with semiaxes
p !
d d 2 C h2 4d 2
.a; b/ D ;
9d 2 C h2 9d 2 C h2

and center .0; y0 /, y0 D .9d 4 h4 /=.2h.9d 2 C h3 //.

Vertices Pi D Œxi ; yi , i D 1; 2; 3 of Brocard porism triangles are given by:
4.5. Summary 71

Family Fixed Conserves Notes

Poristic P polar image of Confocal
X1 ,X3 ,X40 ,: : : cos i ; a9 =b9
(bicentric) P 2 Q family wrt to a focus
Poristic X2 ,X3 ,X4 ,X5 si , cos i Inscribed in circle;
Excentrals caustic is MacBeath inconic
P 2 polar image of Homothetic
X3 ,X6 ,X15 ,X16 ,
Brocard X39 ,X182 ; : : :, P si , !, family wrt caustic focus;
cot i inscribed in circle;
˝1 , ˝2
caustic is Brocard inellipse

Table 4.2: Summary of fixed points and (known) conserved quantities for the non-
concentric, axis-parallel (NCAP) families in this chapter.

x1 D cos t=q1
y1 D sin t=q1
x2 D d.d 2 C h2 /..3d 2 C h2 / sin t C 2dh cos t 3d 2 C h2 /=q2
y2 D .d 2 C h2 /..9d 4 2d 2 h2 C h4 / sin t 2dh.3d 2 C h2 / cos t 9d 4 C h4 /=.2bq2 /
x3 D d.d 2 C h2 /.2dh cos t .3d 2 C h2 / sin t C 3d 2 h2 /=q3
y3 D .d 2 C h2 /.2dh.3d 2 C h2 / cos t C .9d 4 2d 2 h2 C h4 / sin t 9d 4 C h4 /=.2bq3 /
q1 D .2h/=.d 2 C h2 /
q2 D 2dh.3d 2 h2 / cos t .9d 4 h4 / sin t C 9d 4 C 2d 2 h2 C h4
q3 D 2dh.3d 2 h2 / cos t C .9d 4 h4 / sin t 9d 4 2d 2 h2 h4

4.5 Summary

Fixed points and (known) conserved quantities for the non-concentric (NCAP)
families in this chapter appear in Table 4.2 (compare with Table 3.2).
A diagram depicting how certain pairs of families are interrelated by either
similarity or polar transformations appears in in Figure 4.11.
72 4. Non-concentric, Axis-Parallel (NCAP)

Incircle Rigid Rot Poristic Similarity Confocal (Billiard)

X1, ∑ 𝐜𝐨𝐬 X1,X3, ∑ 𝐜𝐨𝐬 f1-polar X9, 𝐋, ∑ 𝐜𝐨𝐬

Affine I

Circumcircle Rigid Rot

Orthics Orthic/
X40, ∑ 𝐜𝐨𝐬 Excentral

Orthic/ Affine II
Excentral

Rigid Rot Poristic Excentrals Similarity Confocal

Circumcircle
X2,X3,X4,X5 Excentrals
X3, ∏ 𝐜𝐨𝐬, ∑ 𝒔𝒊 𝟐 f1c-polar
∏ 𝐜𝐨𝐬 X6, ∏ 𝐜𝐨𝐬

Similarity Brocard Porism

Homothetic
X3,X6,X39,X182
X2, A, ∑ 𝒔𝒊 𝟐 , 𝝎 f1c-polar Ω , Ω , 𝝎, ∑ 𝒔𝒊 𝟐

Figure 4.11: Families mentioned in this chapter (blue ones are concentric, tan ones
are non-concentric), as well as the transformations under which certain families
are interrelated.

4.6 Exercises
Exercise 4.1. Show that over the poristic family, the locus of the foci of the X9 -
centered circumconic (the circumbilliard) is a circle.
Exercise 4.2. Prove Proposition 4.3. Furthermore, prove the intersection point of
X1 X3 with the antiorthic axis is the Schröder point X1155 .
Exercise 4.3. Prove that over the poristic family the inconic centered on X1 is
axis-parallel with the circumconic centered on X9 (i.e., the circumbilliard), see
this Video.
Exercise 4.4. Recall the cosine circle C (also known as the second Lemoine circle)
is centered on a triangle’s symmedian point X6 . Let E 0 be the Brocard ellipse of
some triangle T . Let ˇ be the aspect ratio of E 0 , i.e., a0 =b 0 . Show that for any T ,
above (resp. below) a certain ˇ, C is tangent to E 0 at two distinct points (resp. it
is exterior to E 0 ). See it Live.
4.7. Research questions 73

Exercise 4.5. Show that the poristic excentral family is also the polar image of
billiard excentrals wrt to a circle centered on a billiard (i.e., the caustic) focus.
See it Live.

Exercise 4.6. Show that over the Brocard porism the radius r  of the cosine circle
is invariant.

Exercise 4.7. Show that the first Lemoine circle (centered on X182 is stationary
over the Brocard porism. Above a certain a0 =b 0 , this circle is tangent to one of the
minor vertices of the caustic. See it Live.

Exercise 4.8. Ehrmann’s “third” Lemoine circle is studied in Grinberg (2012). It

is centered on X576 , is defined as follows: for each vertex, consider the 3 circles
containing pairs of vertices and the symmedian point X6 . The third Lemoine circle
contains the 6 intersections of said circles (2 each) with the sidelines. Prove this
circle is also stationary over the Brocard porism, i.e., all three Lemoine circles
are; see it Live.

Exercise 4.9. Prove the expression and inequality for cot ! in Proposition 4.12.

Exercise 4.10. That the Brocard axis X3 X6 is stationary over the Brocard porism
is established. Prove that the Lemoine axis, which intersects the Brocard axis at
the Schoutte point X187 , is also stationary; see it Live.

Exercise 4.11. The so-called “second” Brocard triangle, defined in Weisstein (2019,
Second Brocard Triangle), has vertices at the intersections of symmedians (cevians
through X6 ) with the Brocard circle. Show that over the Brocard porism, the fam-
ily of second Brocard triangles is a new, smaller Brocard porism which shares
the isodynamic points X15 and X16 with the original family. Prove that if this is
iterated, the shrinking porisms converge to X15 . See it Live.

4.7 Research questions

Question 4.1. Show that (i) the family of tangential triangles to the Brocard porism
is also Ponceletian (caustic is the Brocard circumcircle).(ii) Derive the axes for
the ellipse it is inscribed in. and that (iii) its Gergonne point X7 is stationary and
coincides with the symmedian point X6 of the Brocard porism; (iv) the locus of
X20 of the tangentials is a segment along the Brocard axis of the original family.
Live
74 4. Non-concentric, Axis-Parallel (NCAP)

Question 4.2. The 3 Apollonius’ circles of a triangle pass through a vertex and
its two isodynamic points X15 and X16 , see Weisstein (2019, Isodynamic points).
Prove that over the Brocard porism, the sum of the inverse squared radii of the
three Apollonius circles is invariant, see them Live.

Question 4.3. Prove that the polar image of the Brocard porism with respect to a
circle centered on a caustic focus is another (tilted, smaller) Brocard porism whose
Brocard inellipse shares a focus with the original one. Where does the sequence
of Porisms converge? See it Live.

Question 4.4. Prove that over the poristic family, the barycenter X2 of the intouch
triangles is stationary. Derive its coordinates. See it Live.
 Locus

5 Phenomena in
the Confocal
Family

When we consider Poncelet 3-periodic families, a natural (and indeed early) ques-
tion was “what are the loci of certain triangle centers”. Recall one of our early
experimental finds: that over billiard 3-periodics, the locus of the incenter X1 is
an ellipse (as is that of the excenters), see Section 2.3. Also an early find was that
the “locus” of the Mittenpunkt X9 is a point, see Section 2.4.
In this chapter we expand this exploration by touring a gallery of interesting
locus-related phenomena. Our hope is to give the reader an appreciation for the
beauty and variety of loci obtainable. These include:

• The loci of some notable centers of a triangle, showing they are ellipses;

• Billiard 3-periodics which can be both acute and obtuse;

• A triangle center with a non-elliptic (quartic) locus nearly identical to an

ellipse;

• Two special triangle centers railed to either the billiard or the confocal caus-
tic;
76 5. Confocal Loci

• A non-smooth locus with four singularities;

• A self-intersecting locus;
• A non-compact, non-elliptic locus;
• An elliptic locus whose aspect ratio is the golden ratio ';
• A triangle center railed to the elliptic billiard whose motion with respect to
3-periodic vertices is “non-monotonic”;
• The non-elliptic loci of the vertices of certain derived triangles.
• The “triple-winding” of triangle center loci over themselves.

5.1 Kimberling centers with elliptic loci

The semiaxes a1 ; b1 for the elliptic locus of the incenter X1 were given in The-
orem 2.1. As shown in Figure 5.1, it turns the loci of the next four centers on
Kimberling (2019) are also ellipses. These are the barycenter X2 , the circumcen-
ter X3 , the orthocenter X4 , and the center of the 9-point circle (also known as
Euler’s circle) X5 . Their semiaxes are given by:

2ı a2 b 2
.a2 ; b2 / Dk2 .a; b/ ; with k2 D
3c 2
 2 2 
a ı ı b
.a3 ; b3 / D ;
2a 2b
 
k4 k4 .a2 C b 2 /ı 2 a2 b 2
.a4 ; b4 / D ; ; with k4 D
a b c2
 
0 00
w5 .a; b/ C w5 .a; b/ı w5 .b; a/ w500 .b; a/ı
0
.a5 ; b5 / D ;
w5 .a; b/ w5 .b; a/

where w50 .u; v/ D u2 .u2 C 3v 2 /, w500 .u; v/ D 3u2 C v 2 , and w5 .u; v/ D 4u.u2
v 2 /. Note that (i) a2 =b2 D a=b and (ii) b4 =a4 D a=b.
As it turns out, the locus of 49 out of the first 200 centers on Kimberling (ibid.)
are ellipses. These are: Xk , k D1, 2, 3, 4, 5, 7, 8, 10, 11, 12, 20, 21, 35, 36, 40, 46,
55, 56, 57, 63, 65, 72, 78, 79, 80, 84, 88, 90, 100, 104, 119, 140, 142, 144, 145,
149, 153, 162, 165, 190, 191, 200. Links to live animations as well as expressions
for their semiaxes are provided in Garcia, Reznik, and Koiller (2021).
5.1. Kimberling centers with elliptic loci 77

X3

X2
X1
X5

X4

Figure 5.1: Over billiard 3-periodics, the loci of incenter X1 , barycenter X2 , cir-
cumcenter X3 , orthocenter X4 , and 9-point center X5 are all ellipses. The Euler
line (dashed black) is shown passing through all but the first center. Video, Live
78 5. Confocal Loci

Figure 5.2: Locus of the orthocenter (orange) over elliptic billiards with different
aspect ratios. If a=b is (i) less than (resp. (ii) equal, (iii) greater than) ˛4 '1:352,
the locus of the orthocenter X4 (orange) is (i) interior (resp. (ii) internally tangent,
(iii) intersecting) with the elliptic billiard. In (i) and (ii) all 3-periodics are acute,
whereas in (iii) some will be obtuse.

5.2 When billiard 3-periodics are obtuse

It turns out the locus of X4 can be used to determine if the billiard 3-period family
will contain obtuse triangles. Referring to Figure 5.2:
Proposition 5.1. The locus of X4 is internally tangent to the elliptic billiard at its
top and bottom vertices when a=b D ˛4 given by:
q
p
˛4 D 2 2 1 ' [Link]

Proof. The equation b4 D b is equivalent 4 2 2 7b 4 D 0: Therefore,

p p to a C 2a b
as a > b > 0, it follows that a=b D 2 2 1:
5.3. Quartic locus of the symmedian point X6 79

Let ˛4 be the positive root of x 6 C x 4 4 x3 x2 1 D 0, i.e., ˛4 D ' 1:51.

Proposition 5.2. When a=b D ˛4 , then a4 D b and b4 D a, i.e., the locus of X4
is identical to a rotated copy of Billiard.
Proof. The condition a4 D b, or equivalently b4 D a, is defined by a6 C a4 b 2
4a3 b 3 a2 b 4 b 6 D 0. Graphic analysis shows that x 6 C x 4 4 x 3 x 2 1 D 0
has only one positive real root which we call ˛4 .
Theorem 5.1. If a=b < ˛4 (resp. a=b > ˛4 ) the 3-periodic family will not (resp.
will) contain obtuse triangles.
Proof. If the 3-periodic is acute, X4 is in its interior, therefore also internal to the
EB. If the 3-periodic is a right triangle, X4 lies on the right-angle vertex and is
therefore on the EB. If the 3-periodic is obtuse, X4 lies on exterior wedge between
sides incident on the obtuse vertex (feet of altitudes are exterior). Since the latter
is on the EB, X4 is exterior to the EB.
Another way to think of this is depicted in Figure 5.3: a=b > ˛4 , opens up
two “zones” along the top and bottom halves of the elliptic billiard. A 3-periodic
will be obtuse if and only if one of its vertices is on either zone. These zones are
precisely portions of the elliptic billiard which are interior to the locus of X4 ; see
Figure 5.2(right). When a=b D ˛4 said zones collapse to the top and bottom
vertices of the elliptic billiard; see Figure 5.2(bottom left).

5.3 Quartic locus of the symmedian point X6

The symmedian point X6 is replete with properties. Honsberger (1995, Ch. 7) calls
it “one of the crown jewels of triangle geometry”. Its construction is deceptively
simple: the point where a triangle’s symmedians concur; these are reflections of
medians on the bisectors. Its trilinear coordinates could not be simpler: Œa W b W c.
However, it is the first Kimberling center whose locus over billiard 3-periodics is
not an ellipse.
In fact, when 1 < a=b < 2, its locus is visually indistinguishable from a true
ellipse; see Figure 5.4. Fortunately, its fit error is easily detectable with numerical
methods. Indeed:
Proposition 5.3. The locus of X6 is a convex quartic given by:

X6 .x; y/ D c1 x 4 C c2 y 4 C c3 x 2 y 2 C c4 x 2 C c5 y 2 D 0
80 5. Confocal Loci

Figure 5.3: Both acute (blue) and obtuse (dashed blue) billiard 3-periodics are
shown. In this case a=b D 1:618 > ˛4 . If a 3-periodic vertex is located in the red
arcs along the top and bottom halves of the elliptic billiard, the 3-periodic will be
obtuse.

where:

c1 D b 4 .5ı 2 4.a2 b 2 /ı a2 b 2 / c2 D a4 .5ı 2 C 4.a2 b 2 /ı a2 b 2 /

2 2 2 2
c3 D 2a b .a b C 3ı / 2 c4 D a 2 b 4 .3b 4 C 2.2a2 b 2 /ı 5ı 2 /
p
c5 D a4 b 2 .3a4 C 2.2b 2 a2 /ı 5ı 2 / ı D a4 a2 b 2 C b 4

Proof. Using a CAS, obtain symbolic expressions for the coefficients of a quar-
tic symmetric about both axes (no odd-degree terms), passing through 5 known-
points. Still using a CAS, verify the symbolic parametric for the locus satisfies the
quartic.

Note the above is also satisfied by a degenerate level curve .x; y/ D .0; 0/, which
we ignore.

Remark 5.1. We term the “best-fit” ellipse E6 the one internally-tangent to X6 .x; y/ D
0 at its four vertices. Its semiaxes are given by:

   2 
.3 a2 b 2 /ı .a2 C b 2 /b 2 a .a 3 b 2 /ı C .a2 C b 2 /a2 b
a6 D ; b6 D
a2 b 2 C 3ı 2 a2 b 2 C 3ı 2

Table 5.1 shows the above coefficients numerically for a few values of a=b.
5.3. Quartic locus of the symmedian point X6 81

a/b a6 b6 c1 =c3 c2 =c3 c4 =c3 c5 =c3 A.E6 /=A.X6 /

1:25 0:433 0:282 0:211 1:185 0:040 0:095 0:9999
1:50 0:874 0:427 0:114 2:184 0:087 0:399 0:9998
2:00 1:612 0:549 0:052 4:850 0:134 1:461 0:9983
3:00 2:791 0:620 0:020 12:423 0:157 4:769 0:9949

Table 5.1: Coefficients ci =c3 , i D 1; 2; 4; 5 for the quartic locus of X6 as well as

the axes a6 ; b6 for the best-fit ellipse, for various values of a=b. The last-column
reports the area ratio of the internal ellipse E6 (with axes a6 ; b6 ) to that of the
quartic locus X6 , showing an almost exact match.

Figure 5.4: Over billiard 3-periodics (blue), the locus of the symmedian point X6
is a quartic (green). At the billiard aspect ratio shown, it is visually identical to an
ellipse. Also shown is a copy of the quartic (red) such that the distance to a best-fit
ellipse (green) is scaled 1000 fold. Live
82 5. Confocal Loci

P1(t)

e2
X11
X9
e3
e1 X2

X100

Figure 5.5: A billiard 3-periodic (blue). Also shown are the incircle (green) and
9-point circle (pink) which touch at the Feuerbach point X11 . Also shown is the
latter’s anticomplement X100 , and the three extouchpoints e1 ; e2 ; e3 . Over the bil-
liard family, X100 sweep the billiard while both X11 and the extouchpoints sweep
the caustic (though in opposite directions). Video, Live

5.4 The locus of the Feuerbach point and its anticomple-

ment
Referring to Figure 5.5, the Feuerbach point X11 is the single point of contact
between the incircle and the 9-point circle, see Weisstein (2019, X(11)). X11 is
known to lie on the X9 -centered inconic, called the Mandart inellipse, see Weis-
stein (ibid., Mandart inellipse). Since the latter is unique:
Observation 5.1. The confocal caustic is the stationary Mandart inellipse of bil-
liard 3-periodics.
Therefore:
Proposition 5.4. Over billiard 3-periodics, X11 sweepts the confocal caustic.
The anticomplement of a point P is its double-length reflection about the
barycenter X2 , i.e., A.P / D X2 C 2X2 P . Still referring, Figure 5.17, X100
is the anticomplement of X11 . This point is known to lie on (i) the circumcircle,
(ii) the Steiner circumellipse (centered on X2 ), and most relevantly here, (iii) on
5.5. A locus with singularities 83

the X9 -centered circumellipse, see Kimberling (2019, X(9)). Since the latter is
unique:

Observation 5.2. The elliptic billiard is the stationary X9 -centered circumconic

of billiard 3-periodics.

Therefore (proof is left as Exercise 5.8):

Proposition 5.5. Over billiard 3-periodics, the locus of X100 is the elliptic billiard.
It sweeps it in the direction opposite to that of the 3-periodic vertices along the
billiard.

The vertices of the so-called extouch triangle are the points of contact of the
excircles with a triangle’s sidelines, see Weisstein (2019, Extouch triangle). These
are also known as extouchpoints. A known fact is that the Mandart inellipse (i.e.,
the caustic) touches a triangle’s sidelines at the extouchpoints, see Weisstein (ibid.,
Mandart inellipse). Therefore:

Proposition 5.6. Over billiard 3-periodics, the locus of the extouchpoints is the
confocal caustic.

This is also illustrated in Figure 5.17. A curious dynamic phenomenon is that

while the extouchpoints follow the direction of motion of billiard 3-periodics along
the outer ellipse (e.g., counter- or clockwise), X11 rotates in the opposite direction;
see this Live.

5.5 A locus with singularities

Loci considered thus far have been smooth, regular curves. Here we give an ex-
ample of one with four corners. Recall that given a triangle T , the orthic triangle
has vertices at the feet of T ’s altitudes.
Referring to Figure 5.6, it is easy to see that if a triangle T is acute (resp.
obtuse), all three vertices (only one vertex) of the orthic will lie on a sideline. In
the obtuse case, the other two will lie on extensions of two sidelines, i.e., they will
be exterior to T .
An interesting result is the “switching” behavior of the incenter of the orthic
triangle, mentioned in Coxeter and Greitzer (1967, Chapter 1):

Lemma 5.1. If a triangle is acute (resp. obtuse), the incenter of the orthic will
coincide with the orthocenter (resp. the obtuse vertex of T ).
84 5. Confocal Loci

Figure 5.6: Left: the orthic triangle (orange) is shown of an acute reference trian-
gle T (blue), for with an interior orthocenter X4 . In this case, the orthic incenter
X10 coincides with X4 . Right: When T (blue) is obtuse, X4 is exterior. Further-
more, two orthic vertices are outside of T and X10 coincides with the obtuse vertex,
B in the picture. Video

Figure 5.7: From left to right: the orthic triangle (purple) of billiard 3-periodics
(blue) is shown at 3 different positions. The locus of X4 (orange ellipse) intersects
the billiard, i.e., a=b > ˛4 . When a 3-periodic is acute (left), the orthic incenter
coincides with X4 . When it is a right triangle (middle), X4 is on the elliptic billiard
and the orthic is a degenerate segment. When it is obtuse (right), the orthic incenter
remains “pinned” to the obtuse vertex. The end result is that the locus of the orthic
incenter is a quadrilateral with four elliptic arcs (thick purple, right) with four
corners. Video, Live
5.6. A self-intersecting locus 85

Recall that for billiard 3-periodics to include obtuse triangles, a=b > ˛4 ; see
Proposition 5.1. Referring to Figure 5.7:

Corollary 5.1. If a=b > ˛4 , the locus of the incenter of the orthic triangle of
billiard 3-periodics is an elliptic arc “quadrilateral” with four corners.

5.6 A self-intersecting locus

Consider the curious case of a triangle center which is the isogonal conjugate of
the Feuerbach point, listed on Kimberling (2019) as X59 . We revisit its intriguing
locus.
As shown in Figure 5.8, this is a continuous curve with four self-intersections,
internally tangent to the elliptic billiard on its four vertices independently of a=b.
Since it intersects a line parallel to and infinitesimally away from either axis at six
points, its degree must be at least 6.
We propose leave it as a research question (below) the derivation of this locus
(as an implicit and/or parametric equation) and of its critical points.

5.7 A non-compact locus

Given a triangle T , Weisstein (2019, Tangential triangle) defines the tangential
triangle T 0 as having sides tangent to the circumcircle at the vertices. Notice T 0 is
unbounded for a right triangle since the hypotenuse is a diameter of the circumcir-
cle. Consider a smooth deformation of an acute triangle to an obtuse one: one of
the vertices of the tangential triangle will undergo a discontinuous jump. Recall
that the family of billiard 3-periodics with a=b > ˛4 (resp. a=b < ˛4 ) contains
both acute and obtuse (resp. only acute) triangles. Therefore the family of T 0 will
(will not) undergo discontinuous jumps, and the locus of triangle centers thereof
will be non-compact (resp. compact).
As an example, consider the locus of the circumcenter of the tangential triangle,
listed as X26 on Kimberling (2019). It can be shown it is non-elliptic. As shown
in Figure 5.9, it is non-compact (resp. compact) when a=b > ˛4 (resp. a=b < ˛4 ).
In the former case, the locus is compactified by an inversion with respect to the
center.
86 5. Confocal Loci

Figure 5.8: Over billiard 3-periodics, the locus of X59 is a continuous curve with
four self-intersections, tangent to the billiard at its four vertices. Top Left: if a=b
is slightly above 1, the locus of X59 is nearly four-fold symmetric. Not shown: if
a=b D 1, X59 will be on the line at infinity. Bottom Left: An acute 3 periodic
a=b < ˛4 , and an acute 3-periodic. Right: a right-angle 3-periodic in an a=b > ˛4
elliptic billiard. Video, Live
5.7. A non-compact locus 87

Figure 5.9: Left: The tangential triangle (dashed green) is shown for a 3-periodic
in an a=b < ˛4 elliptic billiard. The center of the tangential circumcircle (green)
is X26 . In this case all 3-periodics are acute, and the locus of X26 is compact (and
non-elliptic). Right inset: the image of the (non-compact) locus of X26 under an
inversion with respect to a circle concentric with the billiard, for various values
88 5. Confocal Loci

5.8 A golden locus

The circumcenter of the excentral Triangle is known as the Bevan point X40 , see
the corresponding entry on Kimberling (2019). The following was shown in Gar-
cia, Reznik, and Koiller (2020a):

Proposition 5.7. Over billiard 3-periodics, the locus of X40 is an ellipse similar
to a rotated copy of the elliptic billiard. Its semiaxes are given by

a40 D c 2 =a; b40 D c 2 =b:

p
Corollary 5.2. At a=b D 2, the top and bottom vertices of the locus of X40
touches the top and bottom vertices of the elliptic billiard. .

Referring to Figure 5.10, the following is a harmonious fact associated with

the locus of X40 :
p
Corollary 5.3. At a=b D .1 C 5/=2 D ', the Golden Ratio, the locus of X40 is
identical to a 90ı -rotated copy of the elliptic billiard.

5.9 When the billiard is swept non-monotonically

In Proposition 5.5, we saw that X100 sweeps the elliptic billiard in the direction
opposite to the motion of billiard 3-periodic vertices.
The next triangle center on Kimberling (2019) which is on the X9 -centered cir-
cumconic is X88 , known to be collinear with X1 and X100 . Assume a monotonic
traversal of billiard 3-periodic vertices along the billiard. It turns out at a certain
aspect ratio, the “motion” of X88 can be made to stop.

Proposition 5.8. At a=b D ˛88 , the y velocity of X88 vanishes when the 3-periodic
is a sideways isosceles, where
q
p
˛88 D . 6 C 2 2 /=2 ' 1:485

Proof. Parametrize a 3-periodic vertex P1 .t / D Œa cos t; b sin t. At t D 0, P1 D

0
.a; 0/ it can be easily checked that X88 D . a; 0/. Solve y88 .t /jt D0 D 0 for a=b.
After some algebraic manipulation, this equivalent to solving 4x 4 12x 2 C7 D 0,
p p
whose positive roots are . 6 ˙ 2 2 /=2. ˛88 is the largest of the two.
5.9. When the billiard is swept non-monotonically 89

Figure 5.10: A 3-periodic (blue) is shown within an a=b D ' elliptic billiard
(gold) as well as its excentral triangle (green). At this aspect ratio, the locus of the
Bevan point X40 (purple) is a 90ı -rotated copy of the billiard. Recall this point is
the circumcenter of the excentral triangle. Video, Live
90 5. Confocal Loci

Indeed, there are three types of X88 motion with respect to P1 .t/: (i) a=b <
˛88 : monotonic and opposite to P1 .t/; (ii) a=b D ˛88 : monotonic and opposite,
but with full stop at the billiard major vertices; (iii) a=b < ˛88 : non-monotonic,
containing two retrograde phases.
An equivalent statement, illustrated in Figure 5.11, is that the line family X1 X100
is instantaneously tangent to its envelope at X88 . Referring to Figure 5.11:
Proposition 5.9. Over billiard 3-periodics, the envelope of X1 X100 is (i) entirely
inside, (ii) touches at vertices of, or (iii) intersects the billiard, for a=b (i) less
than, (ii) equal to, or (iii) greater than ˛88 , respectively.
Interestingly:
Proposition 5.10. The motion of X88 is instantaneously (i) opposite to P1 , (ii)
stationary, or (iii) in the direction of P1 , if the tangency E of X1 X100 with the
envelope lies inside, on, or outside the billiard.

5.10 The dance of the swans

Several triangle centers were identified by Peter Moses which lie on the X9 -centered
circumconic of any triangle. These are listed on Kimberling (2019, X(9)) as fol-
lows: Xk , k D88, 100, 162, 190, 651, 653, 655, 658, 660, 662, 673, 771, 799, 823,
897, 1156, 1492, 1821, 2349, 2580, 2581, 3257, 4598, 4599, 4604, 4606, 4607,
8052, 20332, 23707, 24624, 27834, 32680.
In Reznik, Garcia, and Koiller (2020b) we called such centers swans, since
over billiard 3-periodics they will elegantly glide along the margins of an elliptic
“pond”. The first four centers on said list are Xk , k D88, 100, 162, and 190, and
are shown in Figure 5.12.
Above we saw that the motion of X100 is “monotonic” whereas that with
a=b > ˛88 that of X88 isn’t. The next two swans on Kimberling (2019) are X162
and X190 .
Proposition 5.11. The motion of X162 with respect to P1 .t/ is non-monotonic if
a=b > ˛162 where ˛162 '1:1639 is the only positive root of:
5x 8 C 3x 6 32x 4 C 52x 2 36
Proof. The trilinear coordinates of X162 are given by
1 1 1


W

W



s22 s32 s22 C s32 s12 s32 s12 s32 C s12 s22 s12 s22 s12 C s32 s32
5.10. The dance of the swans 91

Figure 5.11: Collinear points X1 ; X100 ; X88 shown in an elliptic billiard with
a=b (i) less than (top-left), (ii) equal to (top-right), or (iii) greater than (bottom),
˛88 '1:486. The motion of X88 relative to 3-periodic vertices will be: (i) mono-
tonic and opposite to the vertices, (ii) monotonic and opposite but will full stops
at the vertices, and (iii) non-monotonic. The envelope (purple) of line X1 X100 in-
tersects the billiard if a=b > ˛88 (bottom). The motion of X88 is instantaneously
(i) opposite to P1 , (ii) stationary, or (iii) in the direction of P1 , if the tangency E
of X1 X100 with the envelope lies inside, on, or outside the billiard. Video, Live
92 5. Confocal Loci

Figure 5.12: A billiard 3-periodic (blue) and the swans Xk , k D88, 100, 162, and
190. Live. This Video shows 29 swans from Moses’ list on Kimberling (2019,
X(9)).

We use the standard parametrization for vertices of the confocal family found
in Section 2.7.1. Using the trilinear coordinates above, we have

X162 .t/ D .x126 .t/; y126 .t //

At t D 2 , P1 D .0; b/ and X162 . 2 / D .0; b/.

0
Solve x162 .t/jtD 2 D 0 for a=b. After some long algebraic symbolic manipu-
lation, this is equivalent to solving 5x 8 C 3x 6 32x 4 C 52x 2 36 D 0, whose
positive roots is ˛162 ' 1:16369:

Since ˛88 > ˛162 , setting a=b > ˛88 implies both centers will move non-
monotonically. Curiously:

Proposition 5.12. With a=b > 1, X88 and X162 never coincide. Therefore over
the billiard 3-periodic family, they never cross each other.

Proof. Consider an elementary triangle P1 D . 1; 0/, P2 D .1; 0/ and P3 D

.u; v/. Obtain cartesian coordinates for X88 and X162 using their trilinears. The
equation X88 D X162 is given by two algebraic p equations F .u; v; s1 ; s2 / D
; s2 / D 0 of degree 17 with s1 D .u 1/2 C v 2 Dj P3 P2 j
G.u; v; s1p
and s2 D .u C 1/2 C v 2 Dj P2 P1 j. Particular solutions of these equations
5.11. Locus of vertices of derived triangles 93

p
are equilateral triangles with P3 D .0; ˙ 3/ in which case X88 and X162 go to
infinity, i.e., these centers can never meet with a=b > 1.
In Figure 5.13, X88 and X162 are imagined as “swans” executing an elegant,
choreographing a never-crossing dance along the margins of an elliptic “pond”.
The joint motion of P1 .t/, X88 , and X162 can also be visualized on the surface
of a torus where the meridians (circles around the smaller radius) correspond to
a given t and the parallels represent a fixed location on the billiard boundary. As
shown in Figure 5.14, the curves for X88 and X162 are thrice-winding, though
never intersecting.
Referring to Figure 5.15, we summarize the monotonicity in the motion of
the first four swans on Kimberling (2019) with respect to a vertex of billiard 3-
periodics as follows:
Proposition 5.13. Over the family of billiard 3-periodics, for any a=b > 1, the
motion of X100 and X190 is monotonic and opposite with respect to that of a vertex
in the family.
Proposition 5.14. Over the family of billiard 3-periodics, if a=b is below (resp.
above) a certain ˛162 > 1 (resp. ˛88 > ˛162 ), the motion of X162 (resp. X88 ) is
monotonic and opposite (resp. non-monotonic) with respect to that of a vertex in
the family.

5.11 Locus of vertices of derived triangles

Some triangles derived from billiard 3-periodics are shown in Figure 5.16. For
their constructions see Appendix A and Weisstein (2019).
Mentioned in Chapter 1 was an early experiment which showed that over bil-
liard 3-periodics, the locus of the vertices of the intouch triangle (i.e., the intouch-
points) is a 2-lobed, self-intersect curve; see Figure 1.4.
As shown in Figure 5.17, the loci of vertices of some other triangles derived
from billiard 3-periodics aren’t ellipses. A noteworthy exception is the extouch
triangle, mentioned above.

5.12 Locus triple winding

Consider one turn of vertex P1 .t/ of billiard 3-periodics around the billiard. Given
3-periodic triple periodicity, over said motion a triangle center will sweep its locus
94 5. Confocal Loci

Figure 5.13: The dance of swans X88 and X162 along the margins of an elliptic
pond. (i) while P1 moves CCW, X88 and X162 approach each other; (ii) at their
closest, they almost kiss. (iii) Suddenly, X162 reverses course, (iv) and a short-
lived same-direction pursuit ensues. (v) An unswooned X88 also changes course,
(vi) with now both swimming away from each other. The duo meets again on
2nd, 3rd and 4th quadrants, where the dance steps are played back in alternating
forward and backward order. A black mittenswan guards his clutch at the center
of the lake. Video, two Live swans, four Live swans.
5.12. Locus triple winding 95

Figure 5.14: The coordinated motion of P1 .t / (blue), X88 (red) and X162 (green)
on the surface of a translucent torus, whose (i) meridians represent position along
the elliptic billiard, and (ii) parallels the family parameter t . Notice the green and
red curves are non-monotonic around the torus but never cross each other. A solid
black meridian is wound at t D 0 and a dashed one appears at one of the 12 instants
of closest distance between X88 and X162 , see Question 5.5.
96 5. Confocal Loci

Figure 5.15: Signed angular velocities of swans Xk , k D88,100,162,190 vs the

parameter t of P1 .t / D Œa cos.t/; b sin.t/ of a billiard 3-periodic vertex, for vari-
ous values of a=b. Cubic roots of the velocities are shown for better visualization
near zero. Top left: a=b D 1:15 is sufficiently small such that all centers move
with variable, negative velocity (monotonic). Top right: At a=b D ˛162 ⩾1:164,
the motion of X162 comes to a stop at discrete values of t. Bottom left: At
a=b D ˛88 ⩾1:486, it is X88 ’s turn to touch zero velocity at discrete moments.
Bottom right: at a=b D 1:5 > ˛88 > ˛162 , both X162 and X88 are engaged in
non-monotonic motion. Notice X100 and X190 remain monotonic (negative ve-
locity).
5.12. Locus triple winding 97

Figure 5.16: Triangles derived from an isosceles billiard 3-periodic (blue). These
contain one vertex on the axis of symmetry. Video, Live

Figure 5.17: Non-elliptic loci of the vertices of triangles derived from billiard
3-periodics: the (i) intouch (green), (ii) Feuerbach (not to be confused with the
Feuerbach point) (blue), (iii) medial (red), triangles. A noteworthy excpetion is the
extouch triangle (light brown), whose vertices sweep the confocal caustic. Video,
Live
98 5. Confocal Loci

Figure 5.18: Depicted is the convex combination Y1 ./ of the incenter X1 and an
intouchpoint I1 of a billiard 3-periodic. app, Video 1, Video 2.

thrice (excluding X9 which doesn’t move).

Referring to Figure 5.18, consider the convex combination Y1 .t / of incenter
X1 and an intouch point I1 .t/, namely:

Y1 .t/ D .1 /X1 .t/ C I1 .t /

where  is a real number.

Loci obtained for Y1 at different values of  are shown in Figure 5.19. At
 D 1 (top-left), Y1 .t / is the recognizable two-lobe locus of the intouchpoints. As
 decreases, the two lobes approach each other. At some critical  they will touch
each other at single point. Decreasing  further causes the lobes to self-intersect
and contain the center of the confocal ellipse pair,. which entails that the turning
number about the origin of the locus suddenly jumps from 1 to 3. As  approaches
zero, the lobes further interpenetrate, and when  D 0, they collapse to the elliptic
locus of the incenter which by continuity, will thrice wind over itself.

5.13 Exercises
Exercise 5.1. Calculate the elliptic billiard aspect ratio a=b such that top and
bottom vertices of the elliptic locus of X3 coincide each with the billiard top and
bottom vertices. Repeat for the locus of X5 .
5.13. Exercises 99

Figure 5.19: The locus (pink) of convex combination Y1 of the incenter and an
intouchpoint at different values of . The elliptic locus of the incenter appears
in all four frames (green). When  D 1 (top left), one obtains the original two-
lobed locus of the intouchpoints (pink). As  decreases (top right, bottom left), the
two lobes approach each other and at some point touch. Decreasing  further still
causes loves to self-intersect and contain the ellipse pair center. As  approaches
zero (bottom right), the lobes further interpenetrate and when  D 0 (not shown),
they collapse to the elliptic locus of the incenter (green). Video 1, Video 2
100 5. Confocal Loci

Exercise 5.2. Calculate the elliptic billiard aspect ration a=b such that the locus
of X4 is identical to a 90ı rotated copy of billiard.
Exercise 5.3. Over billiard 3-periodics, the envelope of the Euler line is an astroidal-
like curve with four cusps, see it Live. Derive its equation. Also, find the elliptic
billiard aspect ratio a=b such that the top and bottom cusps of said curve coincide
each with top and bottom vertices of the elliptic billiard.
Exercise 5.4. Referring to Figure 5.6(right), let Th denote the orthic triangle of
an obtuse triangle T . Is there another (acute) triangle T 0 whose orthic is also Th ?
Exercise 5.5. Express in terms of a; b of the elliptic billiard, the coordinates of
the endpoints of the obtuse “zones” labeled Pi? , i D 1; 2; 3; 4 in Figure 5.3.
Exercise 5.6. Prove Corollaries 5.2 and 5.3.
Exercise 5.7. Let A be area of the four-corner region common to an ellipse and its
90ı -rotated
hp copy and Aiel l D ab be the area of the ellipse. Show that A=Ael l D
4 csc 1 1 C .a=b/2 =. What is this ara ratio for a=b D '?
Exercise 5.8. Prove that the motion of X100 along the elliptic billiard is opposite
to that of the vertices of billiard 3-periodics.
Exercise 5.9. Assume a=b > ˛88 . Find t in P1 .t / D Œa cos t; b sin t where the
motion of X88 changes direction.
Exercise 5.10. Prove that X88 coincides with a 3-periodic vertex if and only if
s2 D .s1 C s3 /=2. In this case, X1 is the midpoint between X100 and X88
Exercise 5.11. Prove Proposition 5.9.
Exercise 5.12. Find the unique aspect ratio a=b > ˛4 of an elliptic billiard which
contains right-triangle 3-periodics with sides as [Link]. Find aspect ratios for
billiards with the next up Pythagorean 3-periodics: [Link], [Link], [Link],
[Link].
Exercise 5.13. Let T be a billiard 3-periodic, and T 0 its anticomplementary trian-
[Link] its sides contain each of the vertices of T and are parallel to the latter’s
opposite sides, see Weisstein (2019). Let T 00 be the intouch triangle of T 0 . Show
that the vertices of T 00 (the intouchpoints) are always on the elliptic billiard. See
it Live. Bonus: prove that the motion of the intouchpoints of T 00 is non-monotonic
assuming P1 .t/ D Œa cos t; b sin t is monotonic along the billiard.
Exercise 5.14. Referring to Figure 5.19, compute the  such that the lobes of Y.t /
touch.
5.14. Research questions 101

5.14 Research questions

Question 5.1. Concerning the locus of X59 over billiard 3-periodics (Figure 5.8,
determine:

• An implicit and/or parametric equation;

• The locations of its four self-intersections;

• The a=b such that if X59 is on a self-intersections on the elliptic billiard

minor axis, the 3-periodic is a right triangle? (it is close to 1:58, see Fig-
ure 5.8(right).

Question 5.2. Prove that over billiard 3-periodics traversed continuously, the ver-
tices of the extouch triangle, i.e., the 3 extouchpoints, will move in the same direc-
tion as 3-periodic vertices, whereas the Feuerbach point will move in the opposite
direction.

Question 5.3. Derive an expression (implicit and/or parametric) for the locus of
X26 in either the compact or non-compact case.

Question 5.4. Derive an expression of the non-elliptic locus of the vertices of the
anticomplementary triangle over billiard 3-periodics. Show it is always external
to the elliptic billiard. Derive its inflection points. See it Live.

Question 5.5. Derive an expression for t where X88 and X162 are closest (there
are 12 solutions). In Figure 5.14, the dashed meridian represents one such mini-
mum which for a=b D 2 occurs at t'41ı . Notice it does not coincide with any
critical points of motion.

Question 5.6. Show that the locus of the inversion of X1 with respect to the moving
circumcircle of billiard 3-periodics is also an ellipse. See it Live.

Question 5.7. Show that the locus of the inversion of X3 with respect to the moving
incircle of billiard 3-periodics is also an ellipse. See it Live.

Question 5.8. Prove Proposition 5.13.

Question 5.9. Prove Proposition 5.14, and derive ˛162 and ˛88 . Numerically,
these are approximately 1:164 and 1:486, respectively, see Figure 5.15 (top right
and bottom left).
 Locus

6 Phenomena in
other CAP
Families

In the previous chapter we toured loci phenomena over billiard 3-periodics. Here
we continue this exploration over the five concentric, axis-parallel (CAP) families
depicted in Figure 1.5. In Section 6.6, we review and discuss locus phenomena
for each such family, organizing them according to similarity in locus curve types
(for various triangle centers). Interestingly, the following 3 groups emerge:

• Confocal, incircle, and poristics

• Excentrals, circumcircle, and poristic excentrals

• Homothetic pair and Brocard porism.

For reference, and as we march toward a theory for locus ellipticity (next chap-
ter), at the chapter’s end we provide a list of triangle centers which over each of
the families studied so far are either stationary, trace out a circle, or an ellipse.
In the discussion below, recall Cayley’s condition for a CAP pair to admit a
Poncelet 3-periodic family: ac =a C bc =b D 1, where a > b > 0, ac > 0, and
bc > 0.
6.1. Incircle family 103

Figure 6.1: Loci of Xk , k D2,4,5,6 over incircle 3-periodics. These are ellipses
for all but X5 whose locus is a circle. Live 1, Live 2

6.1 Incircle family

A 3-periodic interscribed in a CAP pair with incircle is shown in Figure 1.5(top
middle).
ab
Recall that for this pair, Cayley implies the inradius r D aCb . Also recall that
in Proposition 3.4, we show that incircle 3-periodics have invariant circumradius
R D .a C b/=2 and that the locus of the circumcenter X3 is a circle of radius
d D .a b/=2 centered on X1 , see Figure 3.3.
The next 4 propositions are illustrated in Figure 6.1.
Proposition 6.1. Over incircle 3-periodics, the locus of the barycenter X2 is an
ellipse with axes a2 D a.a b/=.3a C 3b/ and b2 D b.a b/=.3a C 3b/ centered
on O D X1 .
Proof. Follows from Section 3.6.
Proposition 6.2. Over incircle 3-periodics, the locus of the orthocenter X4 is an
ellipse of axes a4 D .a b/b=.a C b/ and b4 D .a b/a=.a C b/ centered on
O D X1 .
104 6. Loci in CAP Pairs

Proposition 6.3. Over incircle 3-periodics, the locus of the center X5 of the nine-
.a b/2
point circle is a circle of radius d D [Link]/ centered on O D X1 .

Proof. Direct, analogous to Garcia, Reznik, and Koiller (2020b, Thm.3).

Proposition 6.4. Over incircle 3-periodics, the locus of X6 is a quartic given by

the following implicit equation:

2
b .b C 2 a/ a2 C 2 ab C 3 b 2 x 2 C a .a C 2 b/ 3 a2 C 2 ab C b 2 y 2
 
a2 b 2 .a b/2 b 2 .b C 2 a/2 x 2 C a2 .a C 2 b/2 y 2 D 0

6.2 Circumcircle family

This family is inscribed in a circle of radius R centered on O D X3 and circum-
scribes a concentric ellipse with semiaxes a; b; see Figure 1.5(top right). Recall
that the Cayley condition implies R D a C b.

Proposition 6.5. Over circumcircle 3-periodics, the locus of the barycenter X2 is

a concentric circle with radius r2 given by:

1
r2 D .a b/
3
Referring to Figure 6.2:

Proposition 6.6. Over circumcircle 3-periodics, the loci of both the orthocenter
X4 and the center X5 of the 9-point circle are concentric circles centered on X3 ,
with radii 2d 0 and d 0 respectively, where d 0 D .a b/=2 .

Proof. Based on 3-periodic vertex parametrization and CAS-assisted algebraic

simplification.

In Section 2.3 (resp. Section 5.3) we showed that over the confocal family, the
locus of X1 (resp. X6 ) is an ellipse (resp. quartic). Interestingly:

Proposition 6.7. Over circumcircle 3-periodics, the locus of the symmedian point
X6 (resp. the incenter X1 ) is an ellipse (resp. the convex component of a quartic
6.2. Circumcircle family 105

Figure 6.2: A circumcircle 3-periodic: The loci of both orthocenter X4 (pink) and
nine-point center X5 (olive green) are concentric with the external circle (black).
Their radii are 2d 0 and d 0 , respectively where d 0 D jX4 X5 j. In contradistinction
to the elliptic billiard, the locus of the incenter X1 (dashed brown) is non-elliptic
while that of the symmedian point X6 (dashed blue) is an ellipse. Video, Live
106 6. Loci in CAP Pairs

– note the other component corresponds to the locus of the 3 excenters which can
be concave). These are given by:

x2 y2 a2 b 2 a2 b 2
locus of X6 W C D 1; a 6 D ; b 6 D ;
a62 b62 a C 2b 2a C b

2
locus of X1 W x 2 C y 2 2 .a C 3 b/ .a C b/ x 2 2 .a C b/ .3 a C b/ y 2

2
C a2 b 2 D 0

Proof. CAS-assisted simplification.

6.3 Homothetic family

The family of 3-periodics interscribed in a pair of homothetic ellipses is depicted in
Figure 1.5(bottom left). Let a; b be the semiaxes of the outer ellipse. The Cayley
condition for this pair implies that ac D a=2 and bc D b=2, see Proposition 3.11.
p the barycenter X2 is stationary at the common center and the area A D
Recall
.3ab 3/=4 is invariant.
Recall that over the confocal family, the locus of the incenter X1 (resp. sym-
median point X6 ) was an ellipse (resp. a quartic). Referring to Figure 6.3, this is
reversed in the homothetic family:

Proposition 6.8. Over homothetic 3-periodics, the locus of the incenter X1 (resp.
symmedian point X6 / is a quartic (resp. an ellipse). These are given by:

locus of X1 W 16 a2 y 2 C b 2 x 2 a2 x 2 C b 2 y 2 8 b 2 a4 C 5 a2 b 2 C 2 b 4 x 2


2
8 a2 2 a4 C 5 a2 b 2 C b 4 y 2 C a2 b 2 a2 b 2 D 0;
x2 y2 a.a2 b 2 / b.a2 b 2 /
locus of X6 W C D 1; a 6 D ; b 6 D

a62 b62 2.a2 C b 2 / 2.a2 C b 2 /

Proof. CAS-assisted simplification.

6.3. Homothetic family 107

Figure 6.3: A homothetic 3-periodic and the quartic (resp. elliptic) locus of the
incenter X1 (resp. symmedian point X6 ). Live

6.3.1 Four circular loci

The two Fermat points X13 and X14 as well as the two isodynamic points X15 and
X16 have trilinear coordinates which are irrational on the sidelengths of a triangle,
see Kimberling (2019). Indeed, over billiard 3-periodics, their loci are non-elliptic.

Referring to Figure 6.4:

Proposition 6.9. Over homothetic 3-periodics, the loci of Xk , k D13,14,15,16

are four distinct circles. Their radii are .a b/=2, .a C b/=2, .a b/2 =z, and
.a C b/2 =z, respectively, where z D 2.a C b/.
108 6. Loci in CAP Pairs

Figure 6.4: Circular loci of the first and second Fermat points X13 and X14 (red and
green) as well as the first and second isodynamic points X15 and X16 (purple and
orange) for two aspect ratios of the homothetic pair: a=b D 3 (left) and a=b D 5
(right). The radius of the X16 locus is minimal at the first case. Video, Live
6.3. Homothetic family 109

6.3.2 Loci of the Brocard points

Figure 6.5: Over homothetic 3-periodics, the loci of the two Brocard points ˝1
and ˝2 are tilted ellipses (red and green) of aspect ratio equal to those in the pair,
see Video. Also shown (dashed orange) is the locus of the vertices of the first
Brocard triangle (orange): this is an axis-aligned ellipse also homothetic to the
[Link], Live

Referring to Figure 6.5:

Proposition 6.10. Over homothetic 3-periodics, the loci of the Brocard points ˝1
and ˝2 are ellipses E1 and E2 which modulo rotation are homothetic to the ellipses
in the pair. The loci are reflected images of each other about either the x or y axis.
Proof. The loci are given by


p

7 a4 C 6 a2 b 2 C 3 b 4 x 2 3 a4 C 6 a2 b 2 C 7 b 4 y 2 4 3 a2 C b 2 xy
E1 D C 1
a2 .a2 b 2 /2 b 2 .a2 b 2 /2 ab .a2 b 2 /


p

7 a4 C 6 a2 b 2 C 3 b 4 x 2 3 a4 C 6 a2 b 2 C 7 b 4 y 2 4 3 a2 C b 2 xy
E2 D C C 1
a2 .a2 b 2 /2 b 2 .a2 b 2 /2 ab .a2 b 2 /
110 6. Loci in CAP Pairs

The angle  between the axes of ellipses E1 and E2 is given by

p
4 3.a2 C b 2 /ab
tan  D 4 :
3a C 2a2 b 2 C 3b 4

In no other CAP family so far studied, is the locus of either Brocard point an
ellipse. This informs:

Conjecture 1. Over 3-periodics in a CAP family, the locus of the Brocard points
is an ellipse if and only if the ellipses are homothetic.

In Appendix C we describe the loci of the Brocard points over a certain non-
Ponceletian family of triangles with two vertices affixed to the boundary of an
ellipse (or circle) and the other one which sweeps it.

6.3.3 First Brocard triangle: vertex locus

Consider a triangle T D P1 P2 P3 with Brocard points ˝1 and ˝2 . Referring to
Figure 6.6:

Definition 6.1 (First Brocard Triangle). The vertices P10 , P20 , P30 of the First Bro-
card Triangle T1 are defined as follows: P10 (resp. P20 , P30 ) is the intersection of
P2 ˝1 (resp. P3 ˝1 , P1 ˝1 ) with P3 ˝2 (resp. P1 ˝2 , P2 ˝2 ).

Know properties of the T1 include that (i) it is inversely similar to T , (ii) its
barycenter X2 coincides with that of the reference triangle, and (iii) its vertices are
concyclic with ˝1 , ˝2 , X3 , and X6 on the Brocard circle, defined in Weisstein
(2019, Brocard Circle), whose center is X182 . Referring to Figure 6.5:

Proposition 6.11. Over 3-periodics in the homothetic pair, the locus of the vertices
of T1 is an axis-aligned, concentric ellipse, homothetic to the ones in the pair and
interior to the caustic. Its axes are given by:

a.a2 b 2 / b.a2 b 2 /
a0 D ; b 0
D
2.a2 C b 2 / 2.a2 C b 2 /

Proof. The locus must be an ellipse since T1 is inversely similar to the 3-periodics
whose vertices are inscribed in an ellipse and their barycenters coincide. A vertex
of the Brocard triangle is parametrized by
6.3. Homothetic family 111

Figure 6.6: Construction for the First Brocard Triangle (orange) taken from Weis-
stein (2019, First Brocard Triangle). It is inversely similar to the reference one
(blue), and their barycenters X2 are common. Its vertices B1 ; B2 ; B3 are con-
cyclic with the Brocard points ˝1 and ˝2 on the Brocard circle (orange).

x2 y2
C D1
a02 b 02

Since homothetic 3-periodics conserve area (they are the affine image of regu-
lar polygons interscribed in two concentric circles, see Reznik and Garcia (2021b)),
so must T1 (inversely similar). Its area can be computed explicitly:
Proposition 6.12. Over 3-periodics in the homothetic pair, the area of T1 is in-
variant and given by
p
2
3 3ab a2 b 2

2
16 a2 C b 2

6.3.4 Loci of Fermat and isodynamic equilaterals

As seen in Section 4.3.1, the pedal triangles from either the Fermat X13 ; X14 and
Isodynamic X15 ; X16 points are equilateral. Since the homothetic family con-
serves ш D cot !, Proposition 4.18 implies:
112 6. Loci in CAP Pairs

Corollary 6.1. Over homothetic 3-periodics, the areas Ak , k D 13; 14; 15; 16 of
the equilateral pedals from the Fermat and Isodynamic points are invariant.

Over the Brocard porism, the loci of said equilaterals, were shown to be circles,
see Figure 4.10. Interestingly, and referring to Figure 6.7:

6.4 Dual family

The dual family (bottom middle in Figure 1.5) is interscribed between two CAP
ellipses with reciprocal aspect ratios. Its orthocenter X4 is stationary. Referring
to Figure 6.8:

Proposition 6.13. Over dual 3-periodics, the loci of X2 , X3 , and X5 are ellipses.

6.5 Excentral family

Recall in Theorem 2.1, we derived the semiaxes of the locus of the excenters, i.e.,
the ellipse in which the excentral family is inscribed.
Also recall Section 5.8 where it was noted that over the excentral family, the
locus its circumcenter (X40 in terms of billiard 3-periodics) was identical to a
rotated copy of caustic (i.e., the elliptic billiard) when the latter’s aspect ratio is '
the golden ratio.
Referring to Figure 6.9:

Proposition 6.14. Over excentral 3-periodics the locus of X2 , X3 and X4 are

ellipses.

6.6 Summary
Table 6.1 summarizes the types of loci (point, circle, ellipse, etc.) for some tri-
angle centers for families analyzed in this and previous chapters (including non-
concentric such as poristic triangles and Brocard’s porism). Families are grouped
according to similar patterns in their loci types.
The first row reveals that out of the 8 families considered only in the confocal
case is the locus of the incenter X1 an ellipse, suggesting this is a rare phenomenon.
6.6. Summary 113

Figure 6.7: The homothetic Poncelet family (stationary X2 ) is equibrocardal (con-

serves !) and its triangles (blue) have invariant area. The loci of Xk ; k D
13; 14; 15; 16 are circles concentric with the ellipses Video. Since areas A15 , A16 ,
A13 , A14 only depend on cot !, they are individually invariant. The loci of X396
and X395 are ellipses whereas those of X5463 and X5464 are circles. Video
114 6. Loci in CAP Pairs

Figure 6.8: Over dual 3-periodics (stationary X4 ), the loci of X2 , X3 , and X5 are
ellipses. Live

Figure 6.9: Elliptic loci of Xk , k D2,3,4 over excentral 3-periodics (the symme-
dian X6 is stationary at the center). Live
6.6. Summary 115

Group A Group B Group C

Por.
Conf. Inc. Por. Exc. Circ. Hom. Broc.
Exc.
X1 E P P X X X 4 X
X2 E E C E C P P C
X3 E C P E P P E P
X4 E E C E C P E C
X5 E C C E C P E C
X6 4 4 E P E C E P
X7 E E C X X X X X
X8 E E C X X X X X
X9 P E C X X X X X
X10 E E C X X X X X
X11 E00 C00 C00 X X C5 X X
X12 E C C X X X X X
X13 X X X X X X C C
X14 X X X X X X C C
X15 X X X X X X C P
X16 X X X X X X C P
X99 X X C0 X C0 C0 E0 C0
X100 E0 E0 C0 X C0 C0 X C0
X110 X X C0 E0 C0 C0 X C0

Table 6.1: Loci types (P, C, E, X indicate point, circle, ellipse, and non-elliptic (de-
gree not yet derived) loci, respectively) of some triangle centers over 3-periodic
families. These are clustered in in 3 groups A,B,C sharing many metric phenom-
ena: (i) confocal, incircle, poristic; (ii) excentral, circumcircle, poristic-excentral;
(iii) homothetic and Brocard porism. A numeric entry indicates the degree of the
non-elliptic implicit, e.g., ’4’ for quartic. A singly (resp. doubly) primed letter
indicates a perfect match with the outer (resp. inner) conic in the pair. The sym-
bol C5 refers to the nine-point circle. The boldface entries indicate a discrepancy
in the cluster. Note: Xn for the confocal and poristic excentral triangles refer to
triangle centers of the family itself (not of their reference triangles).
116 6. Loci in CAP Pairs

The plethora of circles in the poristic family had already been shown in Odehnal
(2011). A significant occurrence of ellipses in the confocal pair was signalled in
Garcia, Reznik, and Koiller (2020a). As mentioned above, irrational centers Xk ,
k 2 Œ13; 16 sweep out circles for the homothetic pair. X15 and X16 are known to
be stationary over the Brocard family studied in Bradley and Smith (2007). How-
ever, the locus of X13 and X14 are circles. Also noticeable is the fact that (i)
though in the confocal pair the locus of X1 (resp. X6 ) is an ellipse (resp. quartic),
locus types are swapped for both circumcircle and homothetic families.
It is well-known that there is a projective transformation that takes any Pon-
celet family to the the confocal pair, see Dragović and Radnović (2011). In this
case only projective properties are preserved.
As mentioned above, the confocal family is the affine image of either the incir-
cle or circumcircle family. In the first (resp. second) case the caustic (resp. outer
ellipse) is sent to a circle. Though the affine group is non-conformal, we showed
above that both families conserve the sum of cosines. One way to see this is that
there is an alternate, conformal path which takes incircle 3-periodics to confocal
ones, namely through a rigid rotation (yielding poristic triangles), followed by a
variable similarity (yielding the confocal family).
A similar argument is valid for circumcircle triangles: there is an affine path
(non-conformal) to the confocal family though both conserve the product of cosines.
Notice there is also an alternate conformal composition of rotation (yielding poris-
tic excentral triangles) and a variable similarity (yielding confocal excentral trian-
gles). All in this path conserve the product of cosines.
Finally, homothetic and Brocard porism 3-periodics form an isolated clique.
As mentioned in Reznik and Garcia (2021b), though these are variable similarity
images of one another, they are not affinely-related.

6.6.1 Loci types, CAP families

Below we list triangle centers such that their loci types are either points or con-
ics. These are obtained via numerical simulation amongst the first 200 centers on
Kimberling (2019).

• Confocal pair (stationary X9 )

– Ellipses: 1, 2, 3, 4, 5, 7, 8, 10, 11, 12, 20, 21, 35, 36, 40, 46, 55, 56, 57,
63, 65, 72, 78, 79, 80, 84, 88, 90, 100, 104, 119, 140, 142, 144, 145,
149, 153, 162, 165, 190, 191, 200.
– Circles: n/a
6.6. Summary 117

• Incircle: (stationary X1 )

– Ellipses: 2, 4, 7, 8, 9, 10, 20, 21, 63, 72, 78, 79, 84, 90, 100, 104, 140,
142, 144, 145, 149, 153, 191, 200.
– Circles: 3, 5, 11, 12, 35, 36, 40, 46, 55, 56, 57, 65, 80, 119, 165.

• Circumcircle: (stationary X3 )

– Ellipses: 6, 49, 51, 52, 54, 64, 66, 67, 68, 69, 70, 113, 125, 141, 143,
146, 154, 155, 159, 161, 182, 184, 185, 193, 195.
– Circles: 2, 4, 5, 20, 22, 23, 24, 25, 26, 74, 98, 99, 100, 101, 102, 103,
104, 105, 106, 107, 108, 109, 110, 111, 112, 140, 156, 186.

• Homothetic: (stationary X2 )

– Ellipses: 3, 4, 5, 6, 17, 20, 32, 39, 62, 69, 76, 83, 98, 99, 114, 115, 140,
141, 147, 148, 182, 183, 187, 190, 193, 194.
– Circles: 13, 14, 15, 16.

• Dual: (stationary: X4 )

– Ellipses: 2, 3, 5, 20, 64, 107, 122, 133, 140, 154.

– Circles n/a

• Excentral: (stationary: X6 )

– Ellipses: 2, 3, 4, 5, 20, 22, 23, 24, 25, 26, 49, 51, 52, 54, 64, 66, 67,
68, 69, 70, 74, 110, 113, 125, 140, 141, 143, 146, 154, 155, 156, 159,
161, 182, 184, 185, 186, 193, 195.
– Circles n/a

Expressions for the semiaxes of the elliptic loci for many triangle centers are
available in Garcia, Reznik, and Koiller (2021).

6.6.2 Loci types, NCAP families

For completeness, included below are point and/or conic loci for both Poristic and
Brocard triangles. These include many stationary centers as well as segment and
hyperbolic loci.
118 6. Loci in CAP Pairs

• Poristic, see Odehnal (2011)

– Points (11): 1, 3, 35, 36, 40, 46, 55, 56, 57, 65, 165.
– Segments (2): 44, 171.
– Circles (46): 2, 4, 5, 7, 8, 9, 10, 11, 12, 20, 21, 23, 63, 72, 74, 78, 79,
80, 84, 90, 98, 99, 100, 101, 102, 103, 104, 105, 106, 107, 108, 109,
110, 111, 112, 119, 140, 142, 143, 144, 145, 149, 153, 186, 191, 200.
– Ellipses (39): 6, 19, 22, 24, 25, 28, 31, 33, 34, 37, 38, 41, 42, 43, 45,
47, 48, 51, 52, 54, 58, 59, 60, 71, 73, 77, 81, 88, 89, 169, 170, 181,
182, 184, 185, 195, 197, 198, 199.
– Hyperbolas (7): 26, 49, 64, 154, 155, 156, 196.

• Brocard porism

– Points (10): 3, 6, 15, 16, 32, 39, 61, 62, 182, 187.
– Segments (3): 50, 52, 58.
– Circles (38): 2, 4, 5, 13, 14, 17, 18, 20, 23, 69, 74, 76, 83, 98, 99, 100,
101, 102, 103, 104, 105, 106, 107, 108, 109, 110, 111, 112, 114, 115,
140, 141, 147, 148, 183, 186, 193, 194.
– Ellipses (6): 24, 25, 51, 143, 157, 18.
– Hyperbolas (5): 26, 49, 64, 154, 159.

6.7 Exercises
Exercise 6.1. Prove that over incircle 3-periodics, the power of the center with
respect to the (fixed radius) circumcircle is invariant and equal to ab.

Exercise 6.2. Compute a=b of the external ellipse in the incircle CAP family such
that (i) the circular locus of X3 coincides with the incircle, (ii) the elliptic locus
of X4 touches the outer ellipse at its top and bottom vertices, and (iii) the circular
locus of X5 coincides with the incircle. See it Live1, Live2.

Exercise 6.3. Derive the radius of the circumcircle in the same-named family such
that the quartic locus of X1 and the circular locus of X4 intersect at four points
on the inner ellipse, see it Live.

Exercise 6.4. Prove Proposition 6.9.

6.7. Exercises 119

Exercise 6.5. Prove that over homothetic 3-periodics, the radius of the circular
locus of X16 is minimum when a=b D 3.
p
Exercise 6.6. Prove that at a=b D 5, the elliptic loci of the Brocard points over
homothetic 3-periodics are internally tangent to the inner ellipse. See it Live.

Exercise 6.7. Derive the a=b such that the elliptic loci of the Brocard points over
the homothetic family intersect the y axis at b=2, i.e., at the top vertex of the caustic.
See it Live.

Exercise 6.8. Prove that over homothetic 3-periodics, the locus of the Brocard
midpoint X39 is an ellipse, derive its axis.

Exercise 6.9. Show that over homothetic 3-periodics, the elliptic locus of the ver-
tices of the first Brocard triangle is interior to the inner ellipse.

Exercise 6.10. Compute the invariant similarity ratio of homothetic 3-periodics

to the first Brocard triangles.

Exercise 6.11. Derive expressions for the areas in Corollary 6.1.

Exercise 6.12. Synthesize a triangle center such that over billiard 3-periodics its
locus is a circle? Hint: it will be an affine combination of X2 and X3 .

Exercise 6.13. Derive the semiaxes for the dual family elliptic loci of X2 , X3 , and
X5 in Proposition 6.13.

Exercise 6.14. As shown in Section 6.6.2, over poristic triangles, the locus of X44 ,
and X171 are segments. Derive their data. Do the same for the segment-loci of
X50 , X52 , X58 over Brocard porism 3-periodics.

Exercise 6.15. Given an ellipse E with semiaxes a, b, consider a non-Ponceletian

family of triangles with two vertices fixed on the foci of E and a third one which
sweeps the boundary. Show the locus of the incenter of this family is an ellipse.
See it Live.

Exercise 6.16. Prove that over the Brocard porism, the locus of X114 is a circle
concentric with, and exterior to, the Brocard inellipse. Derive its radius. Live

Exercise 6.17. Prove that over the Brocard porism, the locus of X115 is a circle
concentric with the Brocard inellipse of radius equal to the latter’s minor semiaxis.
Live
120 6. Loci in CAP Pairs

Exercise 6.18. Over the Brocard porism, the locus of X185 is an ellipse which
intersects the major axis of the Brocard inellipse E 0 in two points A and B, see it
Live. In the 1 < a=b < 2 range, A; B appear to lie between the foci of E 0 , however
for larger a=b, e.g., a=b D 3, the locus seems to pass through the foci, see it Live.
Prove or disprove this statement. Derive the center and semiaxes of the locus.

6.8 Research questions

Question 6.1. Prove (or disprove) Conjecture 1.

Question 6.2. Prove that over homothetic 3-periodics, the locus of center X5463
(resp. X5464 ) of the first (resp. second) isogonic equilateral antipedal coincides
with the circular locus of X13 (resp. X14 ).

Question 6.3. Prove that over homothetic 3-periodics, the locus of the centers
X396 (resp. X395 ) of the isodynamic equilateral pedals are two ellipses, and derive
their semiaxes.

Question 6.4. Can a triangle center be found such that over excentral (or dual)
3-periodics its locus is a circle?

Question 6.5. Show that over the poristic family (see Section 4.1), the locus of
X59 is an ellipse whose major vertices are internally tangent to the outer circle
poristics are inscribed in. See it Live.

Question 6.6. Prove Proposition 6.14 and derive the semiaxes of the elliptic loci
of the named centers.
7 Analyzing Loci

In Chapters 5 and 6, we took an informal, observational approach in describing in-

teresting locus-related phenomena. In this chapter we take a first step at analyzing
such phenomena. Namely, we attempt to answer the following questions:
• When are loci algebraic?
• What type of curve is swept by the incenter (and excenters) in a generic
concentric, axis-parallel (CAP) pair?
• Given 3-periodics in a generic pair of ellipses, when is the locus of a triangle
center an ellipse?
• We consider the special case of 3-periodics in the circumcircle family, show-
ing that many such loci are circles.
Half through the chapter, we review Blaschke products, described in Daepp
et al. (2019), since they will be used to answer some of the above questions.
Section 7.6 closes the chapter by presenting a theory for locus ellipticity in
the confocal pair, as initially described in Helman, Laurain, Reznik, et al. (2021).
This theory can be easily generalizable to any other pair. We also derive conditions
under which a given locus in the confocal pair is a perfect circle or degenerates
into a segment.
122 7. Analyzing Loci

7.1 When are loci algebraic?

Consider a Triangle Center X whose Trilinears p W q W r are rational on the
sidelengths s1 ; s2 ; s3 , i.e., the Triangle Center Function h is rational, see Equa-
tion (A.1)

Theorem 7.1. The locus of a rational triangle center over 3-periodics in a CAP
pair is an algebraic curve.

Our proof is based on the following 3-steps which yield an algebraic curve
L.x; y/ D 0 which contains the locus. We refer to Lemma 7.1 and Lemma 7.2
appearing below.

Proof.
Step 1. Introduce the symbolic variables u; u1 ; u2

u2 C u21 D 1; u22 D r1 u2 C r2 :
Let sidelengths s1 D jP3 P2 j; s2 D jP1 P3 j; s3 D jP2 P1 j. Define
g1 D s12 jP3 P2 j2 , g2 D s22 jP3 P1 j2 and g3 D s32 jP2 P1 j2 .
Therefore, gi (i=1,2,3) are polynomial expressions on si and .u; u1 ; u2 /:

g1 D h1 s12 C h0
g2 D h1 s22 h2 u1 u2 C h3
g3 D h1 s32 C h2 u1 u2 C h3

Here hi are polynomials in the variable u. The long expressions will be omitted,
but can be evaluated from the vertex parametrization given in Proposition 3.17.
The vertices will be given by rational functions of u; u1 ; u2

P1 D .a u; b u1 /; P2 D .p2x ; p2y /=p3 ; P3 D .p3x ; p3y /=p3

Equations gi D 0 (i D 1; 2; 3), are polynomial in .si ; u; u1 ; u2 /.

Step 2. Express the locus X as a rational function on u; u1 ; u2 ; s1 ; s2 ; s3 .
Convert p W q W r to Cartesians X D .x; y/ via Equation (A.1). From
Lemma 7.1, it follows that .x; y/ is rational on u; u1 ; u2 ; s1 ; s2 ; s3 .

x D Q=R; y D S=T
7.1. When are loci algebraic? 123

To obtain the polynomials Q; R; S; T on said variables u; u1 ; u2 ; s1 ; s2 ; s3 , one

substitutes the p; q; r by the corresponding rational functions of s1 ; s2 ; s3 that de-
fine a specific Triangle Center X . Other than that, the method proceeds identically.
Step 3. Computing resultants. Our problem is now cast in terms of the polynomial
equations:

E0 D Q x R D 0; F0 D S yT D0
Firstly, compute the resultants, in chain fashion:

E1 DRes.g1 ; E0 ; s1 / D 0; F1 D Res.g1 ; F0 ; s1 / D 0
E2 DRes.g2 ; E1 ; s2 / D 0; F2 D Res.g2 ; F1 ; s2 / D 0
E3 DRes.g3 ; E2 ; s3 / D 0; F3 D Res.g3 ; F2 ; s3 / D 0
It follows that E3 .x; u; u1 ; u2 / D 0 and F3 .y; u; u1 ; u2 / D 0 are polynomial
equations. In other words, s1 ; s2 ; s3 have been eliminated.
Now eliminate the variables u1 and u2 by taking the following resultants:

E4 .x; u; u2 / DRes.E3 ; u21 C u2 1; u1 / D 0

F4 .y; u; u2 / DRes.F3 ; u21 C u2 1; u1 / D 0
E5 .x; u/ DRes.E4 ; u22 C 1 u2 1; u2 / D 0
F5 .y; u/ DRes.F4 ; u22 C 1 u2 1; u2 / D 0
This yields two polynomial equations E5 .x; u/ D 0 and F5 .y; u/ D 0.
Finally compute the resultant
L D Res.E5 ; F5 ; u/ D 0
that eliminates u and gives the implicit algebraic equation for the locus X .
Remark 7.1. In practice, after obtaining a resultant, a human assists the CAS by
factoring out spurious branches (when recognized), in order to get the final answer
in more reduced form.
When not rational in the sidelengths, except a few cases1 , Triangle Centers in
Kimberling’s list have explicit Trilinears involving fractional powers and/or terms
1 For instance Hofstadter points X.359/; X.360/.
124 7. Analyzing Loci

containing the triangle area. Those can be made implicit, i.e, given by zero sets
of polynomials involving p; q; r; s1 ; s2 ; s3 . The chain of resultants to be computed
will be increased by three, in order to eliminate the variables p; q; r before (or
after) s1 ; s2 ; s3 .

Theorem 7.2. In the family of 3-periodic orbits in a generic Poncelet pair of conics
the locus of a rational triangle center is an algebraic curve.

Proof. The analysis follow the same steps as in the case of a Poncelet pair of el-
lipses. See proof of Theorem 7.1.

Supporting lemmas
p
Lemma 7.1. Let P1 D .au; b 1 u2 /: The coordinates of P2 and P3 of the 3-
periodic
p p are rational functions in the variables u; u1 ; u2 , where u1 D
billiard orbit
1 u , u2 D r1 C r2 u2 and r1 D ac2 .b 2 bc2 /a2 b 2 ; r2 D a2 b 2 .a2 bc2
2

ac b 2 /.
2

Proof. Follows directly from the vertex parametrization in Proposition 3.17.

p
Lemma 7.2. Let P1 D .au; b 1 u2 /: Let s1 , s2 and s3 the sides of the triangu-
lar orbit P1 P2 P3 . Then g1 .u; s1 / D 0, g2 .s2 ; u2 ; u/ D 0 and g3 .s3 ; u2 ; u/ D 0
for polynomial functions gi .

Proof. Using the parametrization of the 3-periodic Poncelet orbit it follows that
s12 jP2 P3 j2 D 0 is a rational equation in the variables u; s1 . Simplifying,
leads to g1 .s1 ; u/ D 0:
2 2
Analogously for s2 and s3 . In this case, the pequations s2 jP1 Pp3 j D 0 and
s32 jP1 P2 j2 D 0 have square roots u2 D r1 C r2 u2 and u1 D 1 u2 and
are rational in the variables s2 ; u2 ; u1 ; u and s3 ; u2 ; u1 ; u respectively. It follows
that the degrees of g1 , g2 , and g3 are 10. Simplifying, leads to g2 .s2 ; u2 ; u1 ; u/ D
0 and g3 .s3 ; u2 ; u1 ; u/ D 0.

7.2 Review: Blaschke products

As a tool for further results in this chapter, we will use a special parametrization
of Poncelet 3-periodics based on em Blaschke Products, which we originally used
in Helman, Laurain, Garcia, et al. (2021).
7.2. Review: Blaschke products 125

Figure 7.1: Blaschke complex parametrization of Poncelet 3-periodics (blue). Ver-

tices are z1 ; z2 ; z3 . The foci of the caustic are f; g.

Here we consider 3-periodics inscribed in a unit circle and circumscribing a

non-concentric ellipse. We will work in the complex plane. Under the Blaschke
parametrization, Poncelet 3-periodic vertices become symmetric with respect to
the information of the circle-ellipse pair.
Let T D fz 2 C W jzj D 1g the unit circle and D D fz 2 C W jzj < 1g the
open unit disk bounded by T :
Consider a Moebius map Mw0 D .w0 z/=.1 w0 z/ and the Blaschke product
of degree 3 given by

B D Mw1 Mw2 Mw3

Referring to Figure 7.1, let z1 ; z2 ; z3 2 C denote the vertices of Poncelet
3-periodics in a generic N D 3 family with fixed (unit) circumcircle denoted
T D fz 2 C W jzj D 1g. Let f; g be the foci of the caustic. Using Viète’s formula,
we obtain the following parametrization of the elementary symmetric polynomials
126 7. Analyzing Loci

on z1 ; z2 ; z3 :

Definition 7.1 (Blaschke’s Parametrization).

1 WDz1 C z2 C z3 D f C g C f g
2 WDz1 z2 C z2 z3 C z3 z1 D fg C .f C g/
3 WDz1 z2 z3 D 

where  2 T is the varying parameter.

Note that the concentric case occurs when g D f .

For each  2 T , the three solutions of B.z/ D  are the vertices of a 3-periodic
orbit of the Poncelet family of triangles in the complex plane, see Daepp et al.
(2019, Chapter 4). Furthermore, as  varies in T , the whole family of triangles is
covered. Clearing the denominator in this equation and passing everything to the
left-hand side, we get

z3 .f C g C f g/z 2 C .fg C .f C g//z D0

Lemma 7.3. If u; v; w 2 C and  is a parameter that varies over the unit circle
T  C, then the curve parametrized by
v
F ./ D u C Cw

ˇ ˇ
is an ellipse centered at w, with semiaxis juj C jvj and ˇjuj jvjˇ, rotated with
respect to the horizontal axis of C by an angle of .arg u C arg v/=2.

Proof. If either u D 0 or v D 0, the curve h.T / is clearly the translation of a

multiple of the unit circle T , and the result follows. Thus, we may assume u ¤ 0
and v ¤ 0.
Choose k 2 C such that k 2 D u=v. Write k in polar form, as k D r, where
r > 0 (r 2 R) and jj D 1. We define the following complex-valued functions:

R.z/ WD z; S.z/ WD rz C .1=r/z; H.z/ WD kvz; T .z/ WD z C w

It is straightforward to check that F D T ı H ı S ı R.

Since jj D 1, R is a rotation of the plane, thus R sends the unit circle T to
itself. Since r 2 R, r > 0, if we identify C with R2 , S can be seen as a linear
transformation that sends .x; y/ 7! ..r C 1=r/ x; .r 1=r/ y/. Thus, S sends T
7.3. Locus of the incenter in a generic pair 127

to an axis-aligned, origin-centered ellipse E1 with semiaxis r C 1=r and jr 1=rj.

H is the composition of a rotation and a homothety. H sends the ellipse E1 to an
origin-centered ellipse E2 rotated by an angle of [Link]/ D arg.k/ C arg.v/ D
.arg.u/ arg.v//=2 C arg.v/ D .arg.u/ C arg.v//=2. The semiaxis of E2 have
length
jkvj.r C 1=r/ D rjvj.r C 1=r/ D jr 2 vj C jvj D jk 2 vj C jvj D juj C jvj, and
ˇ ˇ ˇ ˇ ˇ ˇ
jkvjjr 1=rj D rjvjjr 1=rj D ˇjr 2 vj jvjˇ D ˇjk 2 vj jvjˇ D ˇjuj jvjˇ
Finally, T is a translation, thus T sends E2 to an ellipse E3 centered at w,
rotated by ˇ .arg.u/ C arg.v//=2 from the axis, with semiaxis lengths juj C
ˇ an angle
jvj and ˇjuj jvjˇ, as desired.
Theorem 7.3. Let B be a Blaschke product of degree 3 with zeros 0; f; g: For
 2 T , let z1 ; z2 ; z3 denote the three distinct solutions to B.z/ D . Then the
lines joining zj and zk , .j ¤ k/ are tangent to the ellipse given by
jw f j C jw gj D j1 f gj:
Proof. See Daepp et al. (ibid., Theorem 2.9, page 37).
Theorem 7.4. Given two points f; g 2 D. Then there exists a unique conic E
with the foci f; g which is 3-Poncelet caustic with respect to T . Moreover, E is an
ellipse. That ellipse is the Blaschke ellipse with the major axis of length j1 f gj:
Proof. See Daepp et al. (ibid., Corollary 4.4, page 44) and Dragović and Radnović
(2021).

7.3 Locus of the incenter in a generic pair

Recall the locus of the incenter and excenters are ellipses if the pair is confocal, see
Theorem 2.1. Here we expand the analysis, starting with the circumcircle family.
The techniques developed here will help us expand the result to any generic pair.
Proposition 7.1. Over Poncelet 3-periodics in the pair with an outer circle and
an ellipse in generic position, the locus X1 is given by:
X1 W z 4 2..fN C g/
N C fg/z 2 C 8z
C .fN g/N 2 2 C [Link] j2 g C f jgj2 2f 2g/ C f 2 g 2 D 0
W z4 2ˇz 2 C 8z C .ˇ 2 4˛/ D 0
128 7. Analyzing Loci

Proof. The incenter of a triangle with vertices fz1 ; z2 ; z3 g is given by:

p p p
s1 z1 C s2 z2 C s3 z3
X1 D p p p
s1 C s2 C s3
s1 D jz2 z3 j2 ; s2 D jz1 z3 j2 ; s3 D jz2 z1 j2

Using that zi 2 T it follows that

     
z3 z2 z1 z3 z1 z2
s1 D 2 C ; s2 D 2 C ; s3 D 2 C
z2 z2 z3 z1 z2 z1
Eliminating the square roots in the equation X1 z D 0 and using the relations
i (i=1,2,3) given in Blaschke’s parametrization the result follows.
p p p
Note that for zi 2 T 1 we have that X1 W z1 z2 z1 z3 z2 z3 , see
Exercise 7.5.
Using techniques similar to those in the last proof, we derive the following
expression for the locus of the incenter and the excenters over 3-periodics in any
ellipse pair:
Proposition 7.2. Over Poncelet 3-periodics in a generic nested ellipse pair, the
locus of X1 and the excenters are the roots of the following quartic polynomial in
z:

.p 2 q 2 /2 2 z 4 4  pq..2 C ˇ/p 2 .2 ˛  C 2/pq C .2 C ˇ/q 2 /z 3

C . 2 ˇ 2 p 4 C 2  .2 ˛ 2 C ˛ ˇ C 9 /p 3 q C . 4 ˛ 2 2 8 ˇ 2 20 ˛  C 4 ˇ 2 /p 2 q 2
C 2  .2 ˛ 2 C ˛ ˇ C 9 /pq 3 2 ˇ 2 q 4 /z 2 C .8 3 p 4 4  .ˇ 2 C 6 ˛  ˇ 2 /p 3 q
C .4 ˛ ˇ 2 C 16 ˛ 2  4 ˇ 2 ˛ C 20 3 4 ˇ /p 2 q 2 4  .ˇ 2 C 6 ˛  ˇ 2 /pq 3
C 8 3 q 4 /X 2 .4 ˛  ˇ 2 /p 4 C 2  .4 ˛ 2  ˇ 2 ˛ C 2 3 ˇ /p 3 q
C . 4 ˛3 C ˛2ˇ2 12 ˛ 3 C 2 ˇ 2 2 C 2 ˛ ˇ  C 2 /p 2 q 2
C 2  .4 ˛ 2  ˇ 2 ˛ C 2 3 ˇ /pq 3 2 .4 ˛  ˇ 2 /q 4 D 0

Proof. Let p; q 2 R. Consider the affine transformation T .z/ D pz C qz and set

wi D T .zi /. The proof is similar to that given in Proposition 7.1. Here, in order
to simplify the vertices zi it is necessary to evaluate the sums pk D z1k C z2k C z3k
.k D 1; : : : ; 5/, expressing the result in terms of ˛, ˇ and . See Figure 7.3
7.3. Locus of the incenter in a generic pair 129

Figure 7.2: Consider a Poncelet 3-periodic family (blue) interscribed between two
non-concentric, unaligned ellipses (centers O and Oc ). The locus of the incenter
X1 (solid green) is non-elliptic and skewed. The locus of the excenters (dashed),
i.e., the vertices of the excentral triangle (solid green), is a non-convex curve.
Video

Figure 7.3: Left: Poncelet 3-periodic in a non-concentric pair with fixed circum-
circle. Right: an affine image thereof.
130 7. Analyzing Loci

Referring to Figure 7.2, note that in general, the locus of the incenter is not
even four-fold symmetric, and that of the excenters may be non-convex.
The above entails yet another alternative proof for the ellipticity of the X1 locus
over billiard 3-periodics:

Corollary 7.1. Over billiard 3-periodics, the locus of X1 is given by:

a4 C b 4 C c 2 ı  .a2 C b 2 /ı a4 b4
X1 D z D0
2 a2 b 2 2a2 b 2 

Proof. Let a and b be the semiaxes of the billiard. In the confocal pair we have
that
1p 1p
f D 2 ı a2 b 2 ; g D 2 ı a2 b 2
c c
This is obtained by taking an affine map T .z/ D .a C b/z=2 C .a b/z=2 N sending
the pair with an unitary outer circle to the confocal pair, see Section 2.2.
The result follows by factorization of the quartic polynomial that defines X1
in Proposition 7.2.
Using CAS we obtain that X1 is factorizable as E1 E2 , where


a4 C b 4 C c 2 ı  .a2 C b 2 /ı a4 b 4
E1 D z D0
 2 a2 b 2 2a2 b 2 

E2 D 2c 2  2 a2 b 2  z 3 C c 2 2 a2 b 2 ıC



C c 2 a2 C b 2 2 a4 C 4 a2 b 2 b 4 z 2



4  c 2 a 2 C b 2 C ı z 4 2 a 4 C b 4 C a 2 ı C b 2 ı

Corollary 7.2. The locus X1 is the ellipse with semiaxes given by a1 D .a2 ı/=b
and b1 D .ı b 2 /=a:

Proof. Follows directly from Lemma 7.3 and Proposition 7.2

Remark 7.2. The space of possible choices of two conics which admits a 3-periodic
family is five dimensional.
7.4. Loci in generic nested ellipses 131

This stems from the fact that a conic has five degrees of freedom, so two conics
have 10; the euclidean transformation group is 4-dimensional, and Cayley deducts
one degree of freedom. Therefore: 10 4 1 D 5.
Note that over said 5d space, the possible confocal configurations are 1-di-
mensional. Interestingly, experimental evidence suggests that our very first result
(elliptic locus of the incenter and excenters) is actually very rare. Referring to
Figure 7.4:

Conjecture 2. Given a pair of ellipses admitting Poncelet 3-periodics, the locus

of the incenter is a non-degenerate ellipse (i.e., not a point) iff the pair is confocal.

Recall a result by Odehnal (2011), illustrated in Figure 4.6: the locus of the
excenters over the poristic family is a circle twice the radius of the circumcircle.
Referring to Figures 2.1 and 7.2:

Conjecture 3. Given a pair of ellipses admitting Poncelet 3-periodics, the locus

of the excenters is an ellipse iff the pair is either confocal or poristic, in which
case the locus is a circle.

7.4 Loci in generic nested ellipses

In this Section we prove the locus of a given fixed affine combination of X2 and
X3 is an ellipse. We will use Blaschke products since, as shown in Figure 7.5, a
generic non-concentric pair is always the affine image of a pair with circumcircle.
Consider the generic pair of nested ellipses E D .O; a; b/ and Ec D .Oc ; ac ;
bc ;  / in Figure 7.6, where  is the counterclockwise tilt2 of Oc with respect to
O. Let s, c denote the sine and cosine of  , respectively. Define cc2 D ac2 bc2 .
The Cayley condition for the pair to admit a 3-periodic family is given by:

b 4 xc4 C 2 a2 b 2 xc2 yc2 C 2cc2 b 2 .a2 C b 2 / c 2
2 b 2 bc2 b 2 a2 2 b 4 bc2 xc2




8 a2 b 2 xc yc cc2 sc C a4 yc4 C 2cc2 a2 a2 C b 2 c 2 2 bc2 C b 2 a4 C 2 a2 b 2 bc2 yc2


C cc4 c 4 c 4 2 cc2 c 2 a2 ac2 b 2 a2 C bc2 b 2 c 2
C .aac C ab bbc / .aac ab bbc / .aac C ab C bbc / .aac ab C bbc / D 0

2 Not to be confused with i , used before to denote an N -periodic internal angle.

132 7. Analyzing Loci

Figure 7.4: Locus of the incenter over Poncelet 3-periodics interscribed in 4 differ-
ent pairs of nested ellipses. (i) confocal pair (top left): the locus of X1 is an ellipse
(this could be unique to the confocal family); (ii) homothetic (top right): the lo-
cus is non-elliptic (we know this via numerics); (iii) circumcircle family (bottom
left): again, the locus is non-elliptic; (iv) 3-periodics in a non-concentric, non-axis-
parallel ellipse pair (bottom right): the locus of the incenter is not even four-fold
symmetric.
7.4. Loci in generic nested ellipses 133

Figure 7.5: Affine transformation that sends a generic ellipse pair and its 3-periodic
family (left) to a new pair with circumcircle (right). We parametrize the 3-periodic
orbit with vertices zi in the circumcircle pair using the foci of the latter’s caustic
f and g, and then apply the inverse affine transformation to get a parametrization
of the vertices Pi of the original Poncelet pair. Video

Figure 7.6: A pair of ellipses in general position which admits a Poncelet 3-

periodic family (blue). Let the outer one be centered at the origin O. Their major
axes are tilted by  , and their centers displaced by Oc D .xc ; yc /. Video
134 7. Analyzing Loci

Referring to Figure 7.7:

Theorem 7.5. Over the family of 3-periodics interscribed in an ellipse pair in

general position (non-concentric, non-axis-aligned), if X˛;ˇ is a fixed linear com-
bination of X2 and X3 , i.e., X˛;ˇ D ˛X2 C ˇX3 for some fixed ˛; ˇ 2 C, then its
locus is an ellipse.

Proof. Consider a general N D 3 Poncelet pair of ellipses that forms a 1-parameter

family of triangles. Without loss of generality, by translation and rotation, we may
assume the outer ellipse is centered at the origin and axis-aligned with the plane
R2 , which we will also identify with the complex plane C. Let a; b be the semi-
axis of the outer ellipse, and ac ; bc the semi-axis of the inner ellipse, as usual.
Referring to Figure 7.5, consider the linear transformation that takes .x; y/ 7!
.x=a; y=b/. This transformation takes the outer ellipse to the unit circle T and
the inner ellipse to another ellipse. Thus, it transforms the general Poncelet N D
3 system into a pair where the outer ellipse is the circumcircle, which we can
parametrize using Blaschke products. In fact, to get back to the original system,
we must apply the inverse transformation that takes .x; y/ 7! .ax; by/. As a
linear transformation from C to C, we can write it as L.z/ WD pz C qz, where
p WD .a C b/=2; q WD .a b/=2.
Let z1 ; z2 ; z3 2 T  C be the three vertices of the circumcircle family,
parametrized as in Definition 7.1, and let v1 WD L.z1 /; v2 WD L.z2 /; v3 WD L.z3 /
be the three vertices of the original general family. The barycenter X2 of the origi-
nal family is given by .v1 C v2 C v3 /=3, and the circumcenter X3 is given by Tak
(n.d.):
ˇ ˇ ˇ ˇ
ˇ v1 jv1 j2 1 ˇ , ˇ v1 v1 1 ˇ
ˇ ˇ ˇ ˇ
X3 D ˇˇ v2 jv2 j2 1 ˇˇ ˇ v2 v2 1 ˇ
ˇ ˇ
ˇ v3 jv3 j2 1 ˇ ˇ v3 v3 1 ˇ

Since z1 D 1=z1 ; z2 D 1=z2 ; z3 D 1=z3 , we can write v1 ; v2 ; v3 as rational

functions of z1 ; z2 ; z3 , respectively. Thus, both X2 and X3 are symmetric rational
functions on z1 ; z2 ; z3 . Defining X˛;ˇ D ˛X2 C ˇX3 , we have consequently that
X˛;ˇ is also a symmetric rational function on z1 ; z2 ; z3 . Hence, we can reduce its
numerator and denominator to functions on the elementary symmetric polynomials
on z1 ; z2 ; z3 . This is exactly what we need in order to use the parametrization by
Blaschke products.
7.4. Loci in generic nested ellipses 135

In fact, we explicitly compute:

p 2 q 2 .˛ C 3ˇ/ C 3ˇ32 C ˛p 3 1 3 pq 2 .3ˇ C 1 3 .˛ C 3ˇ// ˛q 3 2
X˛;ˇ D
33 .p q/.p C q/

where 1 ; 2 ; 3 are the elementary symmetric polynomials on z1 ; z2 ; z3 .

Let f; g 2 C be the foci of the inner ellipse in the circumcircle system. Using
Definition 7.1, with the parameter  varying on the unit circle T , we get:

1
X˛;ˇ D u C v Cw

where:



p f g ˛p 2 q 2 .˛ C 3ˇ/ C 3ˇpq
u WD
3.p q/.p C q/
ˇpq.q fgp/ 1
v WD C ˛fgq
.q p/.p C q/ 3
 


q f C g p 2 .˛ C 3ˇ/ ˛q 2 C p.f C g/ ˛p 2 q 2 .˛ C 3ˇ/
w WD
3.p q/.p C q/

By Lemma 7.3, this is the parametrization of an ellipse centered at w, as de-

sired. As in Lemma 7.3, it is also possible to explicitly calculate its axis and rota-
tion angle, but these expressions become very long.

In Theorem 7.5 a linear combination of X2 and X3 was considered in terms of

complex parameters ˛; ˇ. Below this result is specialized to the case of an affine
combination of said centers in terms of a real parameter .

Corollary 7.3. Over the family of 3-periodics interscribed in an ellipse pair in

general position (non-concentric, non-axis-aligned), if X is a real affine combi-
nation of X2 and X3 , i.e., X D .1 /X2 C X3 for some fixed 2 R, then its
locus is an ellipse. Moreover, as we vary , the centers of the loci of the X are
collinear.

Proof. Apply Theorem 7.5 with ˛ D 1 ; ˇ D to get the elliptical loci. As in

the end of the proof of Theorem 7.5, the center of the locus of X can be computed
136 7. Analyzing Loci

explicitly as
w Dw0 C w1 , where
1    
w0 D q f C g C p.f C g/
3

 

q 2p 2 C q 2 f C g p.f C g/ p 2 C 2q 2
w1 D
3.p q/.p C q/
As 2 R varies, it is clear the center w sweeps a line.
We proved that all of the following triangle centers have elliptic loci in the
general N=3 Poncelet system, including the barycenter, circumcenter, orthocenter,
nine-point center, and de Longchamps point (reflection of the orthocenter about
the circumcenter of a triangle):
Observation 7.1. Amongst the 40k+ centers listed on Kimberling (2019), Kimber-
ling (2020b) identifies about 4.9k which lie on the Euler line. Out of these, only
226 are fixed affine combinations of X2 and X3 . For k < 1000, these amount to
Xk ; k D2, 3, 4, 5, 20, 140, 376, 381, 382, 546, 547, 548, 549, 550, 631, 632.

Observation 7.2. The elliptic loci of X2 and X4 are axis-aligned with the outer
ellipse.
Experimental evidence suggests the converse of Theorem 7.5 is also true:
Conjecture 4. Over 3-periodics interscribed between two ellipses in general po-
sition, the locus of a triangle center Xk is an ellipse if and only if Xk is a fixed
linear combination of X2 and X3 .

7.5 Circular loci in the circumcircle family

Referring to Figure 7.8:
Proposition 7.3. If a triangle center X˛;ˇ D ˛X2 C ˇX3 is a fixed linear com-
bination of X2 and X3 for some ˛; ˇ 2 C, its locus over 3-periodics in the non-
concentric pair with a circumcircle is a circle centered on O˛ and of radius R˛
given by:

˛.f C g/ j˛fgj
O˛ D ; R˛ D
3 3
7.5. Circular loci in the circumcircle family 137

Figure 7.7: A 3-periodic is shown interscribed between two non-concentric, non-

aligned ellipses (black). The loci of Xk , k D 2; 3; 4; 5; 20 (and many others)
remain ellipses. Those of X2 and X4 remain axis-aligned with the outer one. Fur-
thermore the centers of all said elliptic loci are collinear (magenta line). Video
138 7. Analyzing Loci

Furthermore, the center and radius of the locus do not depend on ˇ since the
circumcenter X3 is stationary at the origin of this system.

Proof. Since, z1 ; z2 ; z3 are the 3 vertices of the Poncelet triangle inscribed in the
unit circle, its barycenter and circumcenter are given by X2 D .z1 Cz2 Cz3 /=3 and
X3 D 0, respectively. We define X˛;ˇ WD ˛X2 CˇX3 D ˛.z1 Cz2 Cz3 /=3. Using
Definition 7.1, we get X˛;ˇ D ˛.f C g C f g/=3 D ˛.f C g/=3 C .˛f g/=3,
where the parameter  varies on the unit circle T . Thus, the locus of X over the
Poncelet family of triangles is a circle with center O˛ WD ˛.f C g/=3 and radius
R˛ WD j˛f gj=3 D j˛fgj=3.

Using ˛ D 1 ;ˇ D for a fixed 2 R in Proposition 7.3, we get:

Corollary 7.4. If a triangle center X D .1 /X2 C X3 is a real affine combina-

tion of X2 and X3 for some 2 R, its locus over 3-periodics in the non-concentric
pair with a circumcircle is a circle. Moreover, as we vary , the centers of these
loci are collinear with the fixed circumcenter.

Many triangle centers on Kimberling (2019) are affine combinations of the

barycenter X2 and circumcenter X3 . See Observation 7.1 for a partial list.

Observation 7.3. For a generic triangle, only X98 , and X99 are simultaneously
on the Euler line and on the circumcircle. However these are not linear combina-
tions of X2 and X3 . Still, if a triangle center is always on the circumcircle of a
generic triangle (there are many of these, see Weisstein (2019, Circumcircle)), its
locus over 3-periodics in the non-concentric pair with circumcircle is trivially a
circle.

Corollary 7.5. Over the family of 3-periodics inscribed in a circle and circum-
scribing a non-concentric inellipse centered at Oc , the locus of Xk , k in 2,4,5,20
are circles whose centers are collinear. The locus of X5 is centered on Oc . The
centers and radii of these circular loci are given by:

f Cg f Cg
O2 D ; O4 D f C g; O5 D ; O20 D .f C g/
3 2
jfgj jfgj
r2 D ; r4 D jfgj; r5 D ; r20 D jfgj
3 2
7.5. Circular loci in the circumcircle family 139

Figure 7.8: Left: 3-periodic family (blue) in the pair with circumcircle where the
caustic contains X3 , i.e., all 3-periodics are acute. The loci of X4 and X20 are inte-
rior to the circumcircle. Right: X3 is exterior to the caustic, and 3-periodics can
be either acute or obtuse. Equivalently, the locus of X4 intersects the circumcircle.
In both cases (left and right), the loci of Xk , k in 2,4,5,20 are circles with collinear
centers (magenta line). The locus of X5 is centered on Oc . The center of the X2
locus is at 2=3 along OOc . Video

Proof. As in Corollary 7.4, we can use Proposition 7.3 with D 0; 2; 1=2; 4

to get the center and radius for X2 ; X4 ; X5 ; X20 , respectively. All of these centers
are real multiples of f C g, so they are all collinear. Moreover, the center O5 of
the circular loci of X5 is .f Cg/=2, that is, the midpoint of the foci of the inellipse,
or in other words, the center Oc of the inellipse.

Referring to Figure 7.8:

Observation 7.4. The family of 3-periodics in the pair with circumcircle includes
obtuse triangles if and only if X3 is exterior to the caustic.

This is due to the fact that when X3 is interior to the caustic, said triangle center
can never be exterior to the 3-periodic. Conversely, if X3 is exterior, it must also
be external to some 3-periodic, rendering the latter obtuse.
140 7. Analyzing Loci

Author Year Technique Reference

D. Reznik 2011 Experimental Video Reznik (2011)
O. Romaskevich 2014 Complex Analytic Geometry Romaskevich (2014)
R. Garcia 2016 Real Analytic Geometry Garcia (2019)
(in this book) 2021 Specialize X1 locus Corollary 7.1
to confocal pair
M. Helman et al. 2021 3-Center Linear Combination see Ž

Table 7.1: Various proof methods for the ellipticity of X1 over billiard 3-periodics.
Ž Helman, Laurain, Reznik, et al. (2021)

7.6 Epilogue: a theory for elliptic loci in the confocal pair

Proposition 7.4. If a triangle center Xk is stationary over a Poncelet 3-periodic
family, then the locus of any triangle center X which is a fixed linear combination
of X2 ; X3 ; Xk will be an ellipse.

Proof. The triangle center X D ˛X2 C ˇX3 C Xk is the linear combination

X˛;ˇ WD ˛X2 C ˇX3 under a fixed translation by Xk , because both Xk and Xk
are fixed over the family.

This entails the most compact rendition of the following result (appearing orig-
inally in Helman, Laurain, Reznik, et al. (2021)):

Corollary 7.6. Over billiard 3-periodics, the locus of X1 is an ellipse.

Proof. For any triangle, X1 can be expressed as the linear combination X1 D

˛X2 C ˇX3 C X9 of X2 , X3 and X9 with:
6 2  4
˛D ; ˇD ; D
C2 C2 C2
where  D r=R, is the ratio of inradius to circumradius. In Reznik, Garcia, and
Koiller (2020a) we showed that over the confocal family, X9 is stationary and 
is invariant, so the claim follows.

Table Table 7.1 shows a history of proof techniques of the ellipticity of X1

over billiard 3-periodics:
We can expand the above result to other triangle centers in the confocal pair,
as many of these are fixed linear combinations of X2 , X3 , and X9 .
7.6. Elliptic loci in the confocal pair 141

Proposition 7.5. In the confocal pair, from X1 to X200 , the loci of Xk are ellipses,
k D1, 2, 3, 4, 5, 7, 8, 10, 11, 12, 20, 21, 35, 36, 40, 46, 55, 56, 57, 63, 65, 72, 78,
79, 80, 88Ž , 84, 90, 100, 104, 119, 140, 142, 144, 145, 149, 153, 162Ž , 165, 190Ž ,
191, 200.

Proof. As in the previous corollary, one can write X1 as a fixed linear combination
of X2 , X3 , and X9 , given that the ratio  D r=R is constant in the confocal pair.
In Helman, Laurain, Reznik, et al. (2021, Table 2 ), a table of fixed coefficients
˛; ˇ; is provided expressing each of the triangle centers in the claim as fixed lin-
ear combinations of X1 , X2 and X3 . Table 7.2 reproduces those results. Therefore
all triangle centers in the claim (except for X88 , X162 , and X190 ) are fixed linear
combinations of X1 , X2 , and X3 , and therefore they are fixed linear combinations
of X2 , X3 , and X9 as well. By Proposition 7.4, given that X9 is stationary over the
confocal family, this implies the loci of all these triangle centers are ellipses.
Ž Note: the loci of X88 , X162 , and X190 (called “swans” before) are also el-
lipses because by definition they lie on the circumconic centered on X9 , see Kim-
berling (2019, X(9)).
Referring to Figure 7.9:

Proposition 7.6. In the confocal pair, the locus of X D ˛X2 C ˇX3 for ˛; ˇ 2 R
is a circle when:
 

˛ ı 3ab ˙ 2 a2 C b 2
D
ˇ ˙ 2ab
ˇ ˇ
Proof. By Lemma 7.3, this will happen when juj C jvj D ˇjuj jvjˇ with u; v
from Theorem 7.5. In the confocal pair, when ˛; ˇ 2 R, both u and v are real
numbers as well. Thus, this condition holds if and only if either u D 0 or v D 0.
The ratios ˛=ˇ that yield circular loci can then be computed directly.

Observation 7.5. It follows that .˛=ˇ/C C .˛=ˇ/ D 3.

Definition 7.2 (Degenerate Locus). When the elliptic locus of a triangle center is
a segment, i.e., one of its axes has shrunk to zero, we will call it “degenerate”.

Proposition 7.7. Let X be a fixed linear combination of X2 , X3 , and Xk , where

Xk is some stationary center over the family of 3-periodics. As the vertices of the
3-periodics sweep the outer ellipse monotonically, the path of X in its elliptical
locus is monotonic as well, except for when this locus is degenerate.
142 7. Analyzing Loci

Xk ˛ ˇ
Xk ˛ ˇ
X1 1 0 0
X2 0 1 0 X65 C1 0 
X3 0 0 1 X72  2 3 
C2 3
X4 0 3 2 X78  1  1 0
3 1 6 6
X5 0 X79 1 2C3 2C3
2 2
2C4 3 4 2C1 6 2
X7 C4 C4 C4
X80 1 2 1 2 1 2
X8 2 3 0  2 6 2 4
X84   
 2 6 2
X9 .C1/2 6 2. 1/
C4 C4 C4 X90 2 C2 1 2 C2 1 2 C2 1
1 3
X10 2 2 0 2 3 2
1 3  X100 2 1 2 1 2 1
X11 1 2 1 2 1 2 2 3 2 2
1 3  X104 2 1 2 1 2 1
X12 1C2 1C2 1C2 1 3 3 C1
X119 2 1 2 1 2 1
X20 0 3 4 3 1
3 2 X140 0 4 4
X21 0 2C3 2C3 C2 3C6 2
1 2 X142 2C8 2C8 2C8
X35 2C1 0 2C1 4 8 12 3 8
1 2 X144 C4 C4 C4
X36 1 2 0 1 2 4 3
X145 7 7 0
X40 1 0 2 4 6C9 12 8
1C 2 X149 6 3 6 3 6 3
X46 1  0 1  4 6 3 12 4
1  X153 6 3 6 3 6 3
X55 1C 0 1C 1 4
1  X165 3 0 3
X56 1  0 1  6 4
2C 2 X191 1 2C3 2C3
X57 2  0 2  C4 6
 2 3 2 X200  2  2 0
X63 C1 C1 C1

Table 7.2: Triples ˛; ˇ; used to express a given triangle center Xk as the linear
combinations ˛X1 C ˇX2 C X3 . Note:  D r=R. Note also that though the
loci of X88 , X162 , and X190 are ellipses over the confocal family (in fact, they
sweep the elliptic billiard), they are not included since they are not fixed linear
combinations.
7.6. Elliptic loci in the confocal pair 143

Proof. By Theorem 7.5, the locus of X can be parametrized by u C v 1 C w for

some u; v; w 2 C, where  sweeps the unit circle in C in the same direction as
the 3-periodic vertices sweep the outer ellipse of the Poncelet pair. We can thus
parametrize X as X .t/ D ue i t C ve i t C w. If either u D 0 or v D 0, it is clear
from this parametrization that X sweeps its locus monotonically. Thus, we can
now assume that u ¤ 0 and v ¤ 0.
Denoting u D u0 C iu1 and v D v0 C iv1 with u0 ; u1 ; v0 ; v1 2 R, we can
directly compute
ˇ ˇ2
ˇd ˇ
ˇ X .t /ˇ D juj2 C jvj2 C 2 sin.2t/.u2 v1 u1 v2 / 2 cos.2t /.u1 v1 C u2 v2 /
ˇ dt ˇ

Since .u2 v1 u1 v2 /2 C .u1 v1 C u2 v2 /2 D .u21 C u22 /.v12 C v22 / D juj2 jvj2 ,

there is some angle  2 Œ0; 2/ (the angle between the vectors .u1 ; u2 / and
.v1 ; v2 /) such that u1 v1 C u2 v2 D jujjvj cos  and u2 v1 u1 v2 D jujjvj sin .
Substituting this back in the previous equation, we derive
ˇ ˇ2  
ˇd ˇ
ˇ X .t/ˇ D jujjvj juj C jvj C 2 sin.2t / sin./ 2 cos.2t / cos./
ˇ dt ˇ jvj juj
   
juj jvj juj jvj
D jujjvj C 2 cos.2t C / ⩾ jujjvj C 2
jvj juj jvj juj
By AM-GM inequality, this last quantity is always strictly greater than 0 unless
ˇ ˇ2
ˇd ˇ
juj D jvj. If juj ¤ jvj, we will have ˇ dt X .t/ˇ > 0, and hence the velocity vector
never vanishes, meaning that the X sweeps its smooth locus monotonically. By
Lemma 7.3, this means that X sweeps its locus monotonically except when this
locus is degenerate.

In Section 5.12 we provided a continuity argument for the three turns executed
by a triangle center over a traversal of the billiard 3-periodic family.
Remark 7.3. In Daepp et al. (2019, Lemma 3.4, p. 28) it is shown that (i) the
complex argument of the Blaschke product is monotonic on the unit circle, and that
(ii) for each  there are 3 solutions for the equation B.z/ D . This means that as
 sweeps the unit circle monotonically, the 3-periodics sweep the outer Poncelet
ellipse monotonically and in the same direction as . Moreover for every 3 full
cycles of  over the complex the unit circle, each vertex of the 3-periodics sweep
the outer ellipse exactly once.
144 7. Analyzing Loci

Figure 7.9: A 3-periodic (blue) in a pair of confocal ellipses (black) with a=b D
1:5. Also shown are two degenerate (segment-like) loci (purple) obtained with
'f:27; :73g and two circular loci (orange), obtained with 'f:43; :3g. Video

Proposition 7.8. Let X be a fixed linear combination of X2 , X3 , and Xk , where

Xk is some stationary center over the family of 3-periodics. Over a full cycle of
3-periodics, the winding number of X over its elliptical locus is ˙3, except for
when this locus is degenerate.

Proof. By Theorem 7.5, the locus of X can be parametrized by u C v 1 C w for

some u; v; w 2 C. From Remark 7.3, one can see that the winding number of
 associated to 3-periodics is C3 for each full cycle of 3-periodics over the outer
Poncelet ellipse. Thus, it is sufficient to prove that the winding number of X over
its elliptical locus is ˙1 as  goes around the complex unit circle just once.
Since w is the center of the elliptic locus of X (see Lemma 7.3), we compute
the winding number of X around w. Parametrizing X as X .t / D ue it Cve i t Cw
where  D e it , one can directly compute the winding number as in Ahlfors (1979,
Lemma 1, p. 114):
I Z
1 d 1 2
X 0 .t /
D dt D sign.juj2 jvj2 /
2 i X  w 2 i 0 X .t/ w
By Lemma 7.3, the only way we can have juj D jvj is if the locus of X is
7.7. Exercises 145

degenerate. Thus, whenever this locus is not degenerate, the winding number of
X around its locus as  sweeps the unit circle once is equal to 1 if juj > jvj and
1 when juj < jvj, as desired.

7.7 Exercises
Exercise 7.1. Consider a cubic polynomial p.z/ D .z ˛1 /.z ˛2 /.z ˛3 / with
simple roots ˛i (i=1,2,3). Let ˇ1 and ˇ2 the roots of p 0 .z/. Consider the family
of confocal ellipses having foci ˇ1 and ˇ2 .
Show that there exists a unique ellipse E in this family passing through the
midpoints .˛i C ˛j /=2, and that it is tangent to the sides of the triangle T D
f˛1 ; ˛2 ; ˛3 g. This ellipse is known as Steiner innelipse of T .
Conclude that the center of E is the triangular center X2 of T and that T is a
3-periodic orbit of a homothetic Poncelet pair.
Exercise 7.2. Consider an ellipse E and the set of tangent lines. Show that the
set of points of intersection between any two perpendicular tangents to E lie on a
circle. Find the radius and the center of this circle.
Exercise 7.3. Consider a circle C and a point P0 . Consider the family of circles
passing through P0 and internally tangent to C. Show that the set of centers of this
family of circles is an ellipse. Find the semiaxes and the foci of the ellipse.
Exercise 7.4. In the proof of Proposition 7.2, let z1 ./, z2 ./ and z3 ./ the roots
of E2 .z; / D 0,  2 T . Show that the trace of these three curves is an ellipse,
i.e., they parametrize the excentral locus.
Exercise 7.5. Consider a triangle inscribed in T 1 with vertices w12 , w22 and w32 .
Show that:
• The incenter X1 is w1 w2 w1 w3 w2 w3 .
• The excenters are w1 w2 w1 w3 w2 w3 , w1 w2 C w1 w3 w2 w3 and
w1 w2 w1 w3 C w2 w3 .
• The barycenter X2 is .w12 C w22 C w32 /=3.
• The orthocenter X4 is w12 C w22 C w32 .
• The nine-point center X5 is .w12 C w22 C w32 /=2.

Exercise 7.6. Derive the conditions under which a locus of a triangle center be-
comes degenerate (segment-like) over billiard 3-periodics.
146 7. Analyzing Loci

7.8 Research questions

Question 7.1. In Question 2.1 one is asked to prove that the family of billiard 3-
periodic extouch triangles is Ponceletian. Prove that over this family the loci of
Xk , k D 2; 3; 4; 5; 20 are ellipses, derive their semiaxes. See it Live.

Question 7.2. Prove Conjecture 2 and/or Conjecture 3.

Question 7.3. Prove (or disprove) Conjecture 4.

Question 7.4. Recall the Brocard porism, described in Section 4.4.3. The Brian-
chon point of the Brocard inellipse is X6 (stationary over the porism), i.e., the
sidelines of the Brocard porism family touch the inellipse at the vertices of the X6 -
cevian. Weisstein (2019, Symmmedial triangle) calls this the symmedial triangle.
Show that over Brocard porism 3-periodics, (i) the symmedial triangles are Pon-
celetian, (ii) compute the center and semiaxes of its inellipse, (iii) show that the
locus of Xk , k D 13; 14; 15; 16 are circles. See it Live.
8
The Focus-
Inversive
Family

This chapter describes a multi-talented triangle family directly derived from bil-
liard 3-periodics. We call it the focus-inversive family of triangles. These are
inversive images of billiard N-periodics with respect to a circle centered on (say
the left) focus. The N D 3 case is shown in Figure 8.1. Amongst its curious prop-
erties, we show that: (i) it has a stationary triangle center (the Gergonne point),
(ii) its perimeter and sum of cosines is invariant (mirroring the behavior of billiard
3-periodics), (iii) it is also a billiard 3-periodic family but of a gyrating elliptic
billiard, (iv) the product of its area with that of focus-inversives with respect to
the (right) focus is invariant, (v) that any triangle center whose locus is an ellipse
in the elliptic billiard traces out a circle over focus-inversives, and finally, that (vi)
the loci of its three centroids (vertex, perimeter, and area) are all circles! What’s
more, most of these properties generalize to focus-inversives for N > 3, though
we leave these to part II of this book.

8.1 Non-Ponceletian
A known result is that the inversive image of an ellipse with respect to one of its
foci is a loopless Pascal’s Limaçon, see Weisstein (ibid.). Therefore, the focus-
inversive will be inscribed in such a curve and is therefore non-Ponceletian. In-
148 8. The Focus-Inversive Family

Figure 8.1: The N D 3 focus-inversive family (pink), i.e., the inversive image of
billiard 3-periodics (blue) with respect to a focused-centered circle C (dashed gray).
Focus-inversives are inscribed in a loopless Pascal’s Limaçon (olive green). Both
Ž
perimeter and sum of cosines are invariant. The Gergonne point X7 is stationary.
 Ž
Also shown is X7 , the inversive image of X7 with respect to C, inquired about in
Exercise 8.1. Live

deed, the caustic is also non-elliptic, as shown in Figure 8.2: a continuously in-
creasing billiard aspect ratio will transition the caustic from (i) a regular curve, to
(ii) one with a self-intersection and two cusps, to (iii) a non-compact curve with
two infinite branches.

8.2 A stationary point

Recall in the confocal family the Mittenpunkt X9 is stationary. Henceforth we
shall append a Ž to all quantities referring to the focus-inversive family. Let a; b
denote the semiaxes of the pre-inversion billiard which we assume to be centered
on Œ0; 0 and be axis-parallel to x, and y respectively. Let  denote the radius of
f1 D Œ c; 0, the (left) focus-centered inversion circle, c 2 D a2 b 2 . Interest-
ingly:
8.2. A stationary point 149

Figure 8.2: Non-conic caustic (pink) to the focus-inversive family (pink). A bil-
liard 3-periodic (dashed blue) and the corresponding focus-inversive triangle are
shown at the top-left picture only. The billiard caustic is shown on every frame
(brown). From left-to-right, top-to-bottom, a=b is increased in small steps. Over
this range, the caustic transitions from (i) a regular curve, to (ii) a curve with one
self-intersection and two cusps, to (iii) a non-compact curve. Live
150 8. The Focus-Inversive Family

Ž
Proposition 8.1. The Gergonne point X7 of focus-inversives is stationary on the
major axis of the pre-image confocal pair. Its coordinates are given by:
   
Ž 2
X7 D c 1 ;0
ı C c2

where as before: ı 2 D a4 .ab/2 C b 4 .

8.3 Billiard-like invariants

The following two surprising invariants – constant perimeter and sum of cosines –
are analogues to those displayed by billiard 3-periodics. Interestingly they are not
consequences of elementary principles or transformations.

Proposition 8.2. The perimeter LŽ of focus-inversives is invariant and given by:

q

8 a4 C 4 a2 b 2 C 2 b 4 ı C 8 a6 C 3 a2 b 4 C 2 b 6
LŽ D  2
a2 b 2

Ž
Let i denote angles internal to focus-inversives.

Proposition 8.3. The sum of internal angle cosines of focus-inversives is invariant

and given by:
X Ž ı.a2 C c 2 ı/
cos 1;i D
a2 c 2

8.4 The rotating billiard table

Recall that in Figure 2.6 we introduced the concept of the circumbilliard: given a
triangle T , this is the X9 -centered circumellipse of which T is a billiard 3-periodic
(circumellipse normals are angular bisectors). Let C Ž denote the (moving) circum-
Ž
billiard (X9 -centered circumellipse) of focus-inversives. Indeed, and referring to
Figure 8.3, focus-inversives are billiard 3-periodics of a rigidly-moving virtual
elliptic billiard (see Exercise 8.4):
8.5. Invariant area product 151

Proposition 8.4. Over focus-inversives, the semiaxes aŽ ; b Ž of C Ž are invariant

and given by:

p
aŽ D  k1 k2 .ı C a c/
p
b Ž D  k1 k2 .ı a c/
where:
p q
c 2

k1 D 8 a4 C 4 a2 b 2 C 2 b 4 ı C 8 a6 C 3 a2 b 4 C 2 b 6
k3
k2 D 2a2 b 2 ı



k3 D 2ab 2 2 a2 b 2 ı C 2 a4 2 a2 b 2 b 4

Ž
Proposition 8.5. Over the 3-periodic family, the Mittenpunkt X9 of focus-inversive
triangles moves along a circle with center and radius given by:
   
Ž 2 1
C9 D c 1C ;0
2b 2
Ž 2a2 b 2 ı
R9 D2
2ab 2

8.5 Invariant area product

Ž Ž
Let A1 (resp. A2 ) denote the area of the f1 - (resp. f2 ) inversive triangle family.
Referring to Figure 8.4:

Ž Ž
Proposition 8.6. For N D 3, the area product A1 A2 of the two focus-inversive
triangles is given by:

 
Ž Ž 8 4 2 2 4

3a4 b 2
A1 A2 D 8 2 a C 2a b C 4b ıC C a6 C 4 b 6
8a b 2
152 8. The Focus-Inversive Family

Figure 8.3: The moving circumbilliard (orange) to focus-inversives (pink) rigidly

Ž
translate and rotate (invariant semiaxes). Their center X9 sweeps a circle. The
Ž
location of the stationary Gergonne point X7 is also shown. Video 1, Video 2
8.5. Invariant area product 153

Figure 8.4: The area product of f1 - and f2 -inversive triangles (pink) is invariant.
Video, Live

Figure 8.5: A focus-inversive 3-periodic (pink) is shown inscribed in Pascal’s

Ž
Limaçon (dashed pink). Also shown are the circular loci of Xk , k D 1; 2; 3; 4; 5
whose centers Oi all lie on the billiard’s major axis. Video, Live
154 8. The Focus-Inversive Family

8.6 Circular loci galore!

Ž
We saw above that the locus of X9 is a circle. A remarkable property of the focus-
inversive family is its ability to produce circular loci of many triangle centers.
Referring to Figure 8.5, through CAS-assisted simplification we obtain:
Ž
Proposition 8.7. The locus of X1 is the circle given by:
  2 2  
Ž 2 2a C b C 2ı
C1 D c 1C ;0
2b 4
Ž 2ı 2 C b 4 C .2a2 b 2 /ı
R1 D2
2ab 4
Ž
Proposition 8.8. The locus of X2 is the circle given by:
  2  
Ž 2 2a b2 ı
C2 D c 1C ;0
3a2 b 2
Ž 2a2 b2 ı
R2 D 2
3ab 2
Ž
Proposition 8.9. The locus of X3 is the circle given by:
  2 2 
Ž 2a C b
C3 D c 1 C  ;0
2b 4
Ž a. b 2 C ı/
R3 D2
2b 4
Ž
Proposition 8.10. The locus of X4 is the circle given by:
  2  
Ž 2 .b C ı/ı
C4 D c 1C ;0
a2 b 4
Ž c 2 .b 2 C ı/
R4 D2
ab 4
Ž
Proposition 8.11. The locus of X5 is the circle given by:
  4  
Ž 2a 3a2 b 2 C 2b 4 C 2b 2 ı
C5 D c 1C ;0
4a2 b 4
Ž .3a2 2b 2 /b 2 C .a2 2b 2 /ı
R5 D2
4ab 2
8.7. A rule for circular loci? 155

Ž
Proposition 8.12. The locus of X100 is the circle given by:

   
Ž 2 1
C100 D c 1 C  2 ;0
b
Ž a
R100 D 2 2
b
Ž
Note: The locus of X11 is also a circle, see Exercise 8.5.

8.7 A rule for circular loci?

Recall Section 6.6.1 where 42 triangle centers are identified (from within the first
200 on Kimberling (2019)), whose loci over billiard 3-periodics are ellipses.

Observation 8.1. Amongst the first 200 triangle centers listed on Kimberling
Ž
(ibid.), the following triangle centers Xk sweep conics over the focus-inversive
family:

• Circles (40): 1, 2, 3, 4, 5, 8, 9, 10, 11, 12, 20, 21, 35, 36, 40, 46, 55, 56, 57,
63, 65, 72, 78, 79, 80, 84, 90, 100, 104, 119, 140, 142, 144, 145, 149, 1501 ,
153, 165, 191, 200.

• Ellipses (4): 69, 75, 85, 86.

Comparing these with the list in Section 6.6.1 for the confocal family, one
realizes that the only ones missing are X7 (stationary over the inversive family) and
the “swans” Xk , k D88, 162, and 190. i.e., triangle centers which by construction
lie on the billiard, see Figure 8.6.
Experimentally, in the range k⩽1000, if the locus of Xk is an ellipse over
billiard 3-periodics (excluding the cases where the locus is the billiard itself), then
Ž
the locus of Xk over the focus-inversive family is a circle. Therefore:

Conjecture 5. If the the locus of some triangle center X is an ellipse over billiard
3-periodics, then the locus of X Ž over the inversive family is a circle.

The following cases do not invalidate the conjecture but are noteworthy:
1 See Question 8.1.
156 8. The Focus-Inversive Family

Figure 8.6: Over billiard 3-periodics (dashed red) the loci of both X88 and X162
coincide with the billiard (blue). However, when taken as centers of the the focus-
inversive triangles (not shown), their loci are clearly non-elliptic (green and pur-
ple). Live app: X88+X162, X88+X100

• Though X100 is a swan, its focus-inversive locus is circular.

• Though X658 is swan, its focus-inversive locus is an ellipse.
• Though the locus of X150 is non-elliptic over billiard 3-periodics, that of
Ž
X150 is a circle, see Figure 8.7.
• Though the billiard locus of X934 (blue) is a curve with two self-intersections,
its focus-inversive locus is a circle, see it Live.

8.7.1 Centroidal loci: a tale of three circles

Let C0 , C1 , C2 denote the vertex, perimeter, and area centroids of polygon, respec-
tively. In Schwartz and Tabachnikov (2016b, Thm 1) it was shown that the loci of
C0 ; C2 over Poncelet families are ellipses, though this not hold in general for C1 .
For triangles, C0 D C2 D X2 and C1 D X10 , see Weisstein (2019, Spieker
Center). Per above we already know that the loci of X2 and X10 over the focus-
inversive family are circles. Therefore, and referring to Figure 8.8:
Ž
Corollary 8.1. The loci of the Ci , i D 1; 2; 3 of the focus-inversive family are
circles.
8.7. A rule for circular loci? 157

Figure 8.7: Though over billiard 3-periodics the locus of X150 is non-elliptic, its
locus over the focus-inversive family is a circle. Live.

Figure 8.8: Circular locus of the focus-inversive X2 and the perimeter centroid
Ž Ž
C1 D X10 . Note that for triangles, the former coincides with both the vertex and
area centroids. App
158 8. The Focus-Inversive Family

8.8 A focus-inversive Doppelgänger

Corollary 4.4 in Section 4.1 states that the poristic family is the polar image of
billiard 3-periodics with respect to a circle of radius  centered on a focus. Let R
and d denote the radius of the poristic circumcircle and d D jX3 X1 j. If a; b
are the semiaxes of the outer ellipse in the confocal pre-image, then these can be
expressed as:
Proposition 8.13.

R D.2a4 2a2 b 2 C b 4 C .2a2 b 2 /ı/a2 =b 6

d D.2a2 b 2 C 2ı/c2 a2 =b 6

Recall every pair of circles is associated with two so-called limiting points
`1 and `2 about which the inversion of the pair yields a new pair of concentric
circles, see Weisstein (2019, Limiting points). Let C and C 0 denote the incircle
and circumcircle of the poristic family. Referring to Figure 8.9, it can be shown:
Observation 8.2. One of the limiting points – call it `1 – of the bicentric circle pair
coincides with a focus – call it f1 – of its confocal polar pre-image. Furthermore,
`1 is internal to both circles.
Classic inversive geometry yields:
Proposition 8.14. Triangles of the focus-inversive family are identical to the pedal
triangles of the poristic family with respect to f1 D `1 .
Let LŽŽ denote the perimeter of the pedal triangle with respect to the non-focal
limiting point `2 . Referring to Figure 8.10:
Proposition 8.15. Over the poristic family, LŽŽ is invariant and given by:


p 2q
ŽŽ 9 R2 d 2 R2 d 2 2 3 p
L D 4
.R2 d 2/ 2 9 R2 d 2 C 3R4 C 6R2 d 2 d4
16 R d
Equation (4.2) in Section 4.1 provides an expression for the invariant sum of
cosines of the poristic family in terms of the semiaxes a0 ; b 0 of its billiard polar
pre-image. Interestingly:
Proposition 8.16. The sum of cosines of pedal triangles to the bicentric family
with respect to either limiting point is invariant and identical to that of the poristic
themselves.
8.8. A focus-inversive Doppelgänger 159

Figure 8.9: Billiard 3-periodic (blue) and its polar image (orange) with respect
to a circle (dashed gray) centered on the left focus. Said polar family is poristic
and interscribed between two circles C and C 0 (dashed orange) whose centers are
labeled X3 and X1 . Also shown are the two limiting points `1 and `2 of this pair
of circles. Notice that `1 (resp. `2 ) is internal (resp. external) to C and C 0 and
coincides with the billiard left focus (resp. lies to the right of the billiard center).
Also shown are the two pairs of concentric circles (light blue) which are inversive
images of C and C 0 about `1 and `2 , respectively. Notice the circles in the first
pair are tangent to the billiard (black) and confocal caustic (brown) at their major
vertices, respectively.
160 8. The Focus-Inversive Family

X7,1 X7,2
X3 X1 f1 X9 f2 l2

Figure 8.10: A billiard 3-periodic (Blue), and its polar image with respect to the
left focus, i.e., the poristic family (solid orange). The focus-inversive family (pink)
has invariant perimeter and can also be regarded as the pedal triangle of the poris-
tic family with respect to said focus. The latter coincides with the interior limit
point of the poristic circle pair. A second triangle (purple) is shown which is the
pedal with respect to the (exterior) limiting point `2 of the poristic circle pair. Its
perimeter is also invariant. The Gergonne points X7;1 and X7;2 of either pedal
family are stationary. live
8.9. Exercises 161

ŽŽ
Let X7 denote the Gergonne point of the pedal of bicentrics with respect to
their (external) limiting point:
ŽŽ
Proposition 8.17. X7 is stationary over the poristic family and given by:

p
ŽŽ .R2 d 2 /..R2 d 2 /3=2 9R2 d 2 3R4 6R2 d 2 C d 4 /
X D
16dR4

8.9 Exercises
 Ž
Exercise 8.1. Referring to Figure 8.1, let X7 denote the inversion of X7 with
respect to the inversion circle used to produce the focus-inversive family. Show it
is given by:  
 ı
X7 D ;0
c

Exercise 8.2. Derive the expression in Proposition 8.2.

Exercise 8.3. Derive the expression in Proposition 8.3.
Ž
Exercise 8.4. Show that the locus of the Mittenpunkt X9 of focus-inversives is a
Ž Ž
circle with center O9 and radius R9 given by:
   
Ž 2 1
O9 D c 1 C  ;0
2b 2
Ž 2a2 b 2 ı
R9 D2
2ab 2

Ž
Exercise 8.5. Show that over N D 3 focus-inversives, the locus of X11 is the
circle given by:
  2 2  
Ž 2 a Cb Cı
C11 D c 1C ;0
2a2 b 2
Ž a2 C b 2 C ı
R11 D 2
2ab 2
162 8. The Focus-Inversive Family

Ž
Exercise 8.6. Show that over N D 3 focus-inversives, the locus of X100 is the
circle given by:
   
Ž 2 1
C100 D c 1 C  2 ;0
b
Ž a
R100 D 2 2
b
Exercise 8.7. Consider the bicentric (poristic) family which is the polar image of
billiard 3-periodics wrt to a focus, see Section 4.1. Show the focus-inversive family
are the pedal triangles of said bicentric family wrt to said focus. Bonus: show this
focus coincides with one of the limiting points of the bicentric circle pair.

Exercise 8.8. Show that poristic pedals with respect to their exterior limiting
points are also a constant-perimeter family whose Gergonne point is stationary.

Exercise 8.9. Prove Observation 8.2.

Exercise 8.10. Prove Proposition 8.14.

8.10 Research questions

Ž
Question 8.1. Prove that the locus of X150 is a circle. Derive center and radius.
Ž
Question 8.2. Prove that the locus of X934 is a circle, derive center and radius.
Ž
Question 8.3. Prove that the locus of X658 is an ellipse. Derive its center and
semiaxes.
Ž
Question 8.4. Prove the loci of Xk , k D69, 75, 85, and 86 are ellipses, derive
centers and semiaxes.

Question 8.5. Consider the family of inversive images of excentral 3-periodics

with respect to a circle centered at a point M in the plane. Show the symmedian
point X6 of such a family will be stationary regardless of M . Compute the location
of X6 . See it in this video Video.
9
A Locus
Visualization
App

Many insights described in previous chapters were obtained from experimentation

and observation of pictures, videos, and interaction with the dynamic geometry of
Poncelet configurations. Dozens of notebooks and 100s of small interactive apps
were written with Wolfram (2019). Most of our videos, and other digital artifacts
are compiled in Reznik (2021b).
To further facilitate exploratory discovery of invariants and locus properties of
3-periodic families, we developed a Javascript-based locus visualization app, orig-
inally described in Darlan and Reznik (2021). It borrows many interactivity ideas
from Wolfram (2019) Manipulate, and the sharing and usability model from Ho-
henwarter et al. (2013).
A typical screenshot of the application is depicted in Figure 9.1. A large area
called here the animation window is where the dynamic geometry of a particular
triangular family and its associated loci are drawn To its left is a strip of channel
controls, comprising four identical groups, which define which objects are to be
used as a basis to compute and draw loci from.
The most common usage pattern is depicted in Figure 9.2, namely: the user
selects (i) a triangle family (Poncelet or ellipse-mounted, see below); (ii) the tri-
angle on which computations will be made (the default is “reference” but dozens
of derived triangles can be chosen); (iii) the locus type, i.e., whether one wishes
164 9. A Locus Visualization App

Figure 9.1: Locus Visualization app to explore 3-periodic families. Shown are the
loci of Xk , k D1,2,3,4, over billiard 3-periodics. The “(E)” suffix indicated they
are numerically ellipses. Live. Also see our tutorial playlist.

Figure 9.2: Caption

to trace out a triangle center, a vertex, an envelope, etc., and (iv) which triangle
center should the locus be drawn for. The first one thousand triangle centers listed
on Kimberling (2019) are currently supported.
In the sections below we describe the main functions of the user interface. A
video-based tutorial is available in Reznik (2021c).

9.1 Main ellipse and animation controls

Before a particular triangle family can be setup and its loci visualized, one must set
certain basic animation controls, using the various areas highlighted in Figure 9.3.
9.1. Main ellipse and animation controls 165

Figure 9.3: Basic animation controls include (i) the setting of the base ellipse
aspect ratio a/b either via typing into the textbox (showing 1:618 in the picture)
or via the scrollbar next to it; (ii) above the animation area, pausing or running the
animation and choosing a speed – slow, medium, or fast. Note: a small “anim”
dropbox located below the a/b scrollbar, when not in the “off” position, triggers
a smooth oscillation of the aspect ratio over the range specified in the “min” and
“max” input boxes to its right.

These include (i) the setting of the base ellipse aspect ratio a/b either via typing
into the textbox (showing 1:618 in the picture) or via the scrollbar next to it; (ii)
above the animation area, pausing or running the animation and choosing a speed
– slow, medium, or fast. Note: a small “anim” dropbox located below the a/b
scrollbar, when not in the “off” position, triggers a smooth oscillation of the aspect
ratio over the range specified in the “min” and “max” input boxes to its right.

9.1.1 Convenience animation controls

If the animation is paused, hitting the up (or right) and down (or left) arrows on
the keyboard allows one to carefully step forward or backward over the triangle
family.
The mouse wheel allows for the simulation image to be zoomed or unzoomed.
By clicking and dragging into the main animation area one can pan and repo-
sition the image.
166 9. A Locus Visualization App

Figure 9.4: Four identical groups of “channel” controls positioned to theleft of the
main animation window.

9.2 Channel controls

As shown in Figure 9.4, four identical groups of “channel” controls are positioned
to the left of the main animation window. Figure 9.5 zooms in one of them, whose
individual settings are explained next.

9.3 Choosing a triangle family

The first step in Figure 9.2 is the choice of a triangle family. A specific one is se-
lected via the mnt drop-down, see Figure 9.6. Two types of families are supported:
(i) Poncelet, and (ii) ellipse “mounted” (see below), which originated the name of
the control.

9.3.1 Poncelet families

Currently we support the following 8 types of 3-periodic Poncelet families inter-
scribed between axis-parallel ellipses, whose names are familiar from previous
sections: (i) Confocal (i.e., elliptic billiard), (ii) Homothetic, (iii) with Incircle,
(iv) with Circumcircle, (v) Dual, (vi) Excentral (to confocals), (vii) Poristic, and
(viii) the Brocard Porism. Note (i)-(vi), while the last two are non-concentric.
9.3. Choosing a triangle family 167

Figure 9.5: Various settings in a single channel control.

Figure 9.6: The mnt drop-down selects a triangle family.

168 9. A Locus Visualization App

9.3.2 Ellipse “mounted”

Also selectable are triangle families T .t / D V1 V2 P .t /, where V1 ; V2 are pinned to
two points on or near an ellipse, and P .t / D Œa cos t; b sin t  sweeps the boundary.
Let The following fixed locations for V1 and V2 are currently supported:

1. major: left and right ellipse vertices (EVs)

2. minor: top and bottom EVs

3. mixed: left and top EVs

4. ctrMajor: center and left EV

5. ctrMinor: center and top EV

6. fs: the 2 foci f1 and f2

7. fsCtr: center and right focus (f2 )

8. fsLeft: left EV and f2

9. fsRight: right EV and f2

10. fsTop: top EV and f2

11. tl-bl: top left corner of ellipse bounding box (TL) and bottom left of the
same (BL)

12. tl-tr: TL and top right corner (TR) of ellipse bounding box

13. tl-l: TL and left EV

14. tl-t: TL and top EV

15. tl-b: TL and bottom EV

16. tl-o: TL and center of ellipse

17. tl-br: TL and center of ellipse

9.4. Triangle type 169

Figure 9.7: The triangle menu selects whether a *reference* or some derived
triangle should be used to compute loci. The tri checkbox immediate to the left
selects whether the triangle should be drawn or not.
170 9. A Locus Visualization App

9.4 Triangle type

The second step in Figure 9.2 is the choice of the type of triangle with respect
to which centers and loci will be computed. This is done the tri checkbox and
drop-down, as shown in Figure 9.7.
While the checkbox controls whether selected triangle is drawn or not, the
drop-down contains some four-dozen derived triangles. Below the default setting
*reference* (this indicates a plain triangle in the family should be used), the
choices are organized in three groups:

1. Standard “named” triangles (undecorated abbreviations), such as anticompl

for anticomplementary, bci for BCI triangle, etc., whose construction can
be looked up on Weisstein (2019).

2. Exotic triangles (prefixed by a “.”): .andromeda, .antlia, etc., obtained

from Lozada (2016).

3. Inversive triangles, e.g., *inv-f1*, *inv-f1c*, etc. (decorated with aster-

isks).

Below we document triangles both in the “standard” and “exotic” groups:

9.4.1 Standard triangles

These include: Reference, Anticomplementary, BCI, 1st Brocard, 2nd Brocard,
3rd Brocard, 4th Brocard, 5th Brocard, 6th Brocard, 7th Brocard, Circum-Medial,
Circum-Mid-Arc, Circum-Orthic, Excentral, Extouch, Extangents, Feuerbach,
Fuhrmann, Half-Altitude, Hexyl, Incentral, Inner Vecten, Intangents, Intouch, John-
son, Lemoine, Lucas Central, Lucas Inner, Lucas Tangents, MacBeath, Medial,
Mixtilinear, 1st Morley Adj, 2nd Morley Adj, 3rd Morley Adj, 1st Neuberg, 2nd
Neuberg, Orthic, Outer Vecten, Reflection, Steiner, Symmedial, Tangential, Tan-
gential Mid-Arc, Yff Central, Yff Contact.

9.4.2 Exotic triangles

These include: Andromeda, Antlia, Apollonius, Apus, Atik, Ayme, Bevan-Antipodal,
1st Circumperp, 2nd Circumperp, Excenters–Incenter, Reflections, Excenters–Mid-
points, Honsberger, Inverse–in–Excircles, Inverse–in–Incircle, Kosnita, Mandart
Excircles, Mandart Incircles, Ursa Major, Ursa Minor.
9.4. Triangle type 171

9.4.3 Inversive triangles

The options below are images of the reference triangle in a given family under an
inversive-like transformation with respect to unit circle centered on a stationary
notable point of the family’s underlying ellipse (or caustic), e.g., center, focus,
etc.

• *inv-ctr*,*inv-f1*,*inv-f1c*,*inv-f2*: inversion of vertices with

respect to a unit circle centered on the outer ellipse center, outer ellipse left
focus, inner ellipse left focus, or outer ellipse right focus, respectively.

• *pol-ctr*,*pol-f1*,*pol-f1c*: a new, “polar” triangle is computed

bounded by the polars of the vertices with respect to ellipse center, outer
ellipse left focus, or inner ellipse left focus, respectively.

• *ped-lim2*: this is specific to the confocal family. Computes the pedal

triangle with respect to the non-focal limiting point of the bicentric family
which is the polar image of the confocal family.

• *x3map-ctr*,*x3map-f1*,*x3map-f1c*: consider a triangulation of the

original triangle in 3 subtriangles, each of which contains two vertices of the
original triangle and either (i) the center of the outer ellipse, (ii) its left fo-
cus, or (iii) the inner ellipse left focus, respectively. These transformations
compute a new triangle with vertices at the circumcenter of each subtriangle.

• *x3inv-ctr*,*x3inv-f1*,*x3inv-f1c*: these compute the inverses of

the previous transform with respect to the same points.

• *crem-ctr*,*crem-f1*,*crem-f2*: sends the reference vertices to their

images under a quadratic Cremona transformation, which sends .x; y/ !
.1=x; 1=y/. The origin will be the center of the outer ellipse, its left focus,
or its right focus, respectively.

Note: four additional settings *inf-x*,*inf-y*,*inf-x2*,*inf-y2* are

provided and are experimental and non-inversive. They dynamically set the x or
y coordinate of each vertex so they slide along infinity-like Lissajous curves.
172 9. A Locus Visualization App

Figure 9.8: The Li menu selects the locus type to (triangle center, vertex, envelope,
etc.).

9.5 Locus type

The third step in Figure 9.2 is the choice of type of locus to be drawn, or more
precisely, the feature selected from the family/triangle combination previously se-
lected. This is done with the Li menu at the top of the control group, i D 1; 2; 3; 4,
shown in Figure 9.8.
There are three conceptual groups of locus types: (i) triangle centers and ver-
tices, (ii) segment envelopes, and (iii) bicentric pairs. These are explained next.

9.5.1 Centers and vertices

1. off: it indicates the trace (locus) of this channel should not be drawn. It is
the default setting for channels 2; 3; 4 upon startup.
2. xn: draw the locus of the selected triangle center, as in Section 9.6;
3. v1, v2, v3: show the trace of one of thee vertices of the triangle family. In
Poncelet families, these will sweep out the same curve, but this is not the
9.6. Triangle center 173

case for ellipse-mounted families.

4. ort: the orthopole of line Xm Xn , see Weisstein (2019, Orthopole), where

m and n are selected triangle and Cevian centers, see Section 9.6 and Sec-
tion 9.7.

9.5.2 Envelopes
1. env: the envelope of segment Xm Xn , m¤n, where m (resp. n) is the se-
lected triangle (resp. Cevian) center.

2. e12, e23, e31: the envelope of side Vi Vj of the triangle family. Note these
are one and the same (resp. distinct) for Poncelet (ellipse-mounted) families.

3. e1x, e2x, e3x: the envelope of Vi Xn , i.e., the line from a given vertex to
a selected triangle center. In a concentric Poncelet family, the envelope of
Vi X1 will be the outer ellipse’s evolute, see it Live.

9.5.3 Bicentric pairs

Only a few have so far been implemented, from the copious list in Kimberling
(2020a).

1. ˝1 ; ˝2 : the Brocard points

2. ˇ1 ; ˇ2 : the Beltrami points: inversions of the Brocard points with respect

to the circumcircle

3. 1 ; 2 : also known as “Moses” points: inversion of the Brocard points with

respect to the incircle.

4. 1 ; 2 : the two foci of the Steiner circumellipse (aka. the Bickart points)

9.6 Triangle center

The fourth and final step in Figure 9.2 is the choice of triangle center Xk in the
region highlighted in Figure 9.9. There are three ways to choose k 2 Œ1; 1000: (i)
by typing/editing the text field showing k, (ii) incrementing or decrementing k by
174 9. A Locus Visualization App

Figure 9.9: Controls used for the selection of a particular triangle center Xk .

clicking on the “-” and “+” symbols around the text field; (iii) using the scrollbar
to the right of the “+” control, to quickly scroll through all 1000 values of k. In
fact after any of these is performed, this set of controls becomes “focused” in such
a way that (iv) left (resp. right) arrow keystrokes will decrement (resp. increment)
the value, allowing mouse-free traversal of triangle centers.

9.7 Cevians, pedals, & Co.

An additional “Cevian-like” transformation with respect to an additional triangle
center Xm can be applied to the triangle type selected in Section 9.4. Let us call
the latter the “parent” triangle. The specific transformation is selected via the drop-
down menu in Figure 9.10 (the default setting is pn off, meaning this additional
transformation is inactive), and Xm via the numeric input box to the right of the
menu.
The Xm -transformations possible are grouped into (i) traditional, (ii) inversive,
(iv) reflexive, and (iv) triangulated. Below, let Tm denote the transformed triangle,
and Pi , i D 1; 2; 3, the vertices of the parent triangle.

9.7.1 Traditional
Available in this groups are the standard constructions for (i) Cevian, (ii) Antice-
vian, (iii) Circumcevian, (iv) Pedal, (v) Antipedal, and (vi) Trilinear Polar trian-
gles described in Weisstein (2019). Recall that the latter produces a degenerate
(segment-like) triangle, see Weisstein (ibid., Trilinear Polar).

9.7.2 Inversive
• invert: Tm will have vertices at inversions of the parent one with respect
to a unit circle centered on Xm .
9.7. Cevians, pedals, & Co. 175

Figure 9.10: Cevian-like triangles and number box to select a triangle center play-
ing the role of Q (see text).
176 9. A Locus Visualization App

• polar: Tm will be bounded by the polars (infinite lines) of the parent’s

vertices with respect to a unit circle centered on Xm , see Weisstein (2019,
Polar).

• inv-excircs: Tm will have vertices at inversions of Xm with respect to its

excircles, see Weisstein (ibid., Excircle).

• polar-exc: Tm will be bounded by the polars (infinite lines) of Xm with

respect to each of the parent’s excircles.

9.7.3 Reflexive
• vtx-refl: Tm has vertices at the reflections of Xm on the parent vertices.

• side-refl: Tm has vertices at the reflections of Xm on the sidelines of the

parent triangle.

9.7.4 Triangulated
Triangulate the parent with respect to Xm , i.e., consider the following subtriangles:
T23 D Xm P2 P3 , T31 D Xm P3 P1 , and T12 D Xm P1 P2 .

• 3-circums: Tm has vertices at the circumcenters of T23 , T31 , and T12 .

• 3-inv: The inverse of 3-circums. Tm is such that the circumcenters of its

three subtriangles are the vertices of the parent. The vertices of Tm are the
non-Xm intersections of a circle through Xm and Pi with a circle through
Xm and Pi C1 , cyclically.

• Xk -map, k 2 Œ1; 11: Tm has vertices at the Xk of T23 , T31 , and T12 . Note:
X3 -map is the same as the 3-circums setting.

9.8 Notable circles

Dozens of circles can be visualized with respect to the triangle family selected in
Section 9.4. These are selected via the (left) drop-down menu highlighted in Fig-
ure 9.11. The circs off setting is the default. The possible choices are organized
in two groups: (i) ellipse-affixed, and (ii) central circles.
9.8. Notable circles 177

Figure 9.11: The left drop-down selects an ellipse-based circle or a “central” circle
for both visualization and/or as references for inversive transformations.

9.8.1 Ellipse-affixed circles

These are asterisk-decorated to indicate that they refer to a unit circle centered on
a notable point of the ellipse (or caustic) used to generate a given triangle family,
to be sure:

• *f1*: the left focus of the outer ellipse. Note: in the poristic (resp. ex-
central) family this becomes the center of the outer circle (resp. caustic =
elliptic billiard).
• *f1c*: the left focus of the inner ellipse.
• *f2*: the right focus of the outer ellipse. Note: in the poristic (resp. excen-
tral) family this becomes the incenter (resp. a focus of the outer ellipse.
• *ctr*: the center of the system.

9.8.2 Central circles

Most of these are defined in Weisstein (ibid., Central Circles):

• adams: the Adams circle

178 9. A Locus Visualization App

• appollon1,appollon2,appollon3: 1st, 2nd, and 3rd Apollonius’ cir-

cles (which intersect on the isodynamic points)
• bevan: the Bevan circle, circumcircle of the excentral triangle
• brocard,brocard2: the Brocard circle and the so-called “2nd” Brocard
circle.
• circum: the circumcircle
• conway: Conway’s circle
• cosine: the cosine (or 2nd Lemoine) circle
• [Link]: the cosine circle of the excentral triangle
• ehrmann: Ehrmann’s 3rd Lemoine circle, see Grinberg (2012).
• excircle1,excircle2,excircle3: the three excircles
• euler: Euler’s circle
• furhmann: Furhmann’s circle
• gallatly: Gallatly’s circle
• gheorghe: Gheorghe’s circle, see Kimberling (2019, X(649))
• incircle: Incircle
• lemoine: 1st Lemoine circle
• lester: Lester’s circle
• mandart: Mandart’s circle
• moses,moses rad: Moses’s circle and Moses’ radical circle
• parry: Parry’s circle
• reflection: the “reflection” circle (circumcircle of the reflection triangle)
• schoutte: Schoutte’s circle
• spieker: Spieker’s circle
9.9. Inversive transformations with respect to a circle 179

Figure 9.12: The incircle, circumcircle, and Bevan circle are viewer simultane-
ously, by choosing them on the circle menu in 3 separate channels. Notice that to
make the circle appear, one must check the tri checkbox in the lower left of that
channel control area. Live

• taylor: Taylor’s circle

As shown in Figure 9.12, several circles can be shown simultaneously. To do

this select one for each channel (maintaining the same triangle family and type),
and make sure to check the

9.9 Inversive transformations with respect to a circle

Provided a circle C is selected (see above section), one can add an inversive-type
transformation with respect to it. This is done via the (right) drop-down menu
highlighted in Figure 9.11. The possible transformations are as follows:

• inv off: No transformation is performed.

• inv xn: invert the selected triangle center (see Section 9.6) with respect to
C.

• inv tri: invert the vertices of triangles in the family with respect to C.
180 9. A Locus Visualization App

• pol tri: compute a new triangle bounded by the polars of the original
vertices with respect to C.

• cre xn: send the selected triangle center to its image under a quadratic
Cremona transformation (QCT) .x; y/ ! .1=x; 1=y/, where .x; y/ are the
coordinated of the center of C.

• cre tri: compute a new triangle whose vertices are images of the QCT
with respect to the center of C.

9.10 Conic and invariant detection

9.10.1 Curve type
As shown in Figure 9.13, when one or more loci are displayed, the app indicates
in the lower right-hand side of the corresponding control group, the curve type of
the locus (detected via least-squares curve fitting). The following codes are used:

• X: non-conic

• E: ellipse

• H: hyperbola

• P: parabola (very rare)

• L: line or segment

• *: a stationary point.

The same code is also appended (in parenthesis) to the (moving) triangle center
being displayed, for example, Figure 9.13, X2(E),X3(E),X4(E), indicate the loci
of barycenter, circumcenter, and orthocenter are ellipses over billiard 3-periodics.

9.10.2 Detection of metric invariants

The app also reports when certain basic, metric quantities are invariant, currently
over triangles in the first channel only. These appear as a single line at the bottom
of the animation area of a given experiment, see Figure 9.13. In the example, the
following line of text is reported:
9.10. Conic and invariant detection 181

Figure 9.13: An indication as to curve type of each locus appears in a small box
in the lower right-hand side of each control group. In the picture, X means the first
locus (incenter over the excentral family) is non-conic. An E in the remainder 3
channels indicates their loci are ellipses. Notice the same indicator is appended to
the instantaneous location of the triangle centers being tracked, e.g., X3(E) indi-
cates the locus of the circumcenter is an ellipse. Live
182 9. A Locus Visualization App

Y
L D [Link]; r=R D [Link]; cos0 D 0:0811; A0 =A D [Link]
In turn, this means that perimeter L and ratio r=R of inradius-to-circumradius
are numerically
Q invariant over the reference family selected in channel 1, and that
the product cos of cosines, and ratio A0 =A of derived-by-reference areas is
0

constant (these are observations first introduced in Reznik, Garcia, and Koiller
(2020a)).
Reported invariants appear unprimed to refer to the reference triangle in a
given family. Primed quantities will appear when a derived triangle has been se-
lected (e.g., “excentral”), allowing for mutual comparison.
The following quantities are currently reported, when numerically invariant:

• L; A: perimeter and area

• r; R: inradius and circumradius
• r=R: ratio
P of inradius-to-circumradius, tantamount to invariant sum of cosines
since cos D 1 C r=R.
• cot.!/: the cotangent of the Brocard angle
P 2 P P
• s ; 1=s; s 2 : sum of squared, reciprocal, or reciprocal-squared side-
lengths, respectively.
Q
• cos: the product of internal
Q
• s: the product of sidelengths
• Rc : if a circle is selected via the circle menu (Section 9.8), whether its radius
is constant.

Also reported whenever a derived triangle is selected, are one of L0 =L, A0 =A,
A0 :A, Rc0 =Rc if these are invariant.

9.11 The tandem bar

A common exploratory pattern is to observe the behavior of loci across all chan-
nels simultaneously while a single setting is varied, e.g., triangle family, triangle
type, etc. This could be done with tedious mouse-based changes (of the varying
parameter) across all controls.
9.11. The tandem bar 183

Figure 9.14: The tandem bar. Checking one or more checkboxes ensures locks all
corresponding drop-downs in the channel controls to take the same value.

As an example, consider observing the loci of Xk , k D 1; 2; 3; 4 for the billiard

family and then for the homothetic family. This would require one to reset each of
the four mnt drop-downs from billiard to homothetic. If the user now wished
to examine said loci over the incircle family, all mnt drop-downs wouldhave to be
reset to incircle, etc.
Referring to Figure 9.14, the tandem bar, makes this rather common usage
pattern very [Link], the user can set one or more tandem checkboxes
causing a given setting to be “short circuited” across all channels. Specifically:

• L: the locus type

• mnt: the Poncelet or ellipse-mounted family

• xn: the triangle center number

• tri: the reference or derived triangle

• pn: the triangle center with respect to which Cevian-like triangles are calcu-
lated

As an example, consider the sequence shown in Figure 9.15. Tandem check-

boxes L and mnt are checked, indicating both locus type and triangle family are
in unison across all channels. This automatically sets xn across all channels, i.e.,
184 9. A Locus Visualization App

loci will be drawn (as opposed to, e.g., the envelope). The use then needs to man-
ually set the triangle center values of 1,2,3,4 for each channel. To now observe
these across all families, since mnt is set, the user simply needs to flip through the
triangle families using the triangle family drop-down on any one of the channels
in the strip. In fact this can be done with the up and down arrows on the keyboard
once that control comes into focus (e.g., by expanding the drop-down), allowing
for very quick perusal of this phenomenon across all triangle families.

9.12 Odds & ends

9.12.1 Ellipse, locus tange, and animation background
As shown in Figure 9.16, the area immediately below the four sets channel controls
the following parameters:

• ell checkbox: whether the main ellipse underlying a triangle family of

choice should be drawn or not.
• Rotation menu: the default rot off setting leaves the animation window
as is. Settings 90ı ,180ı ,270ı apply a global rotation to the picture drawn.
• rmax menu: the (half-side) of the square bounding box where points in all
loci are evaluated, respective to the minor semiaxis of the ellipse, assume
to be of unit length. Ideally, this should be set to as small a value as able to
contain all loci.
• bg: used to set the background color of the main animation window, dark
blue by default. By clicking on the colored square an RGB picker window
pops-up permitting fine control of the color.
• invert button: single-click inversion (in RGB space) of colors of back-
ground and loci currently in being drawn.

9.12.2 Resetting the UI and centering the animation

Figure 9.17 highlights reset UI and center UI push-buttons are located at the
top-left corner of the app. These are used to (i) restore all controls in the app to their
default values, and (ii) recenter the geometry drawn to the center of the animation,
respectively.
9.12. Odds & ends 185

Figure 9.15: Usage of the tandem feature. The user has previously selected trian-
gle centers k D 1; 2; 3; 4 for each of the channels. Top: the user is about to flip, in
tandem, triangle family from “billiard” to “homothetic” for the first channel; Bot-
tom: since the tandem mnt is checked, all channels flip in unisn to “homothetic”,
with the visualization being updated in one shot. To quickly flip through all other
families, the user can hit the up and down keys on the keyboard.
186 9. A Locus Visualization App

Figure 9.16: In the highlighted area controls are available to (i) ell: show or hide
the main ellipse, (ii) rot xxx: apply a global rotation to the animation widow, (iii)
rmax: set the bounding box of the area in which loci are computed, (iv) bg: set
the animation window’s background color, and (v) invert: invert (RGB negative)
colors of background and all loci drawn.
9.13. Artsy loci 187

Figure 9.17: Reset and center push-buttons are located at the top-left corner of
the app. which (i) reset UI: all controls in the app are restored to their default
values; (ii) center UI: the center of the simulation is panned back to the center
of the animation area. This is useful after having previously panned the picture
via a mouse drag. Also shown is (iii) a color selector square located to the right
of every triangle center scrollbar, through which a new color can be selected for
displaying the corresponding locus. Finally, (iv) a bbox push button is provided to
repositing and scale the geometric scene so as to best fit it in the available space.

9.12.3 Setting the locus color

Also shown in Figure 9.17 are color and rescaling controls to the right of the trian-
gle center scrollbar. These are permit (i) selection of a color specific to a particular
locus being drawn, and (ii) a resizing/recentering of the particular locus so as to
best fit the animation window.

9.12.4 Collapsing the locus control area

The “hamburger” control shown in Figure 9.18 can be used to hide/expand the set
of controls on the left marging of the app, sometimes useful for demonstration
purposes.

9.13 Artsy loci

A set of controls, highlighted in Figure 9.19, can be used to color fill connected
regions of loci. A first clicking on the palette icon in the middle-right section of
188 9. A Locus Visualization App

Figure 9.18: The hamburger control (three horizontal bars) located to the right of
the the main controls can be clicked to hides/expand the main controls.

a channel’s control group selects a random set of pastel colors. Subsequent clicks
(or hitting the right arrow key) generate a new random color set. Hitting the left
arrow goes back to color sets previously generated. Right-clicking on the palette
icon and or changing any other setting in the user interface causes the color fills to
disappear.
Also highlighted in Figure 9.19 is a scrollbar and color chooser located below
the bottom-most channel control group. These are used to set (i) the transparency
of colors fills, and (ii) the color of the border of connected regions (default is
white).
A collage of four colored-filled curvaceous loci is shown in Figure 9.20. Some
two hundred such “artsy” loci are showcased in Reznik (2021a).

9.14 Sharing and exporting

As shown in Figure 9.21, four buttons on the global control strip (top left of the
interface) can be used to copy and export a link, an image, or a vector graphics
representation the loci currently rendered. The buttons are as follows:
9.14. Sharing and exporting 189

Figure 9.19: Clicking on the highlighted palette icons in the mid-right section of
every channel control area triggers color fills in any drawn loci. Clicking it several
times (resp. right clicking on it) randomizes colors (resp. removes the color fills).
At the bottom of the channel control strip a scrollbar can be used to control the
transparency of the fills. A color chooser at its right side can be clicked to select
the color of region borders (default is white).
190 9. A Locus Visualization App

Figure 9.20: Four examples of the kinds of color-filled loci which can be produced
with the app. Gallery and Video

Figure 9.21: Buttons on the upper control strip for copying, sharing, and exporting
experiment configuration and images.
9.15. Jukebox playback 191

• copy config: A URL containing all information pertaining to the current

geometric scene (as defined by the channel controls and other pieces of UI)
is copied to the clipboard. This URL can be shared with another user and/or
shortened prior to sharing, e.g., with [Link].

• copy image: the image currently on the animation window is copied to the
clipboard. It can then be pasted anywhere else as an image.

• export .png: what is currently on the animation window is downloaded

to the local file system.

• export .json: a vector graphics representation of all loci drawn is ex-

ported in human-readable JSON format.

9.15 Jukebox playback

Many experiments constructed with the tool are stored in a database accessible
by the app. These are organized in different thematic groups which can be played
back in continuously in “jukebox” mode (each experiment being displayed 5-10
seconds). This is initiated by selecting a thematic group from the drop-down at the
top right hand side of the app window, highlighted in Figure 9.22. To quickly zip
forward or backward thru items in the series, click (or right-click) on the jukebox
icon to the left of the drop-down. Jukebox mode can be stopped at anytime by
returning the drop-down to the Juke off setting.
192 9. A Locus Visualization App

Figure 9.22: To play sequentially through one of many groups of experiments, se-
lect an item from the (highlighted) drop-down “jukebox” drop-down menu, at the
top right hand corner of the app. To stop the jukebox playback, select Juke off.
Click (or right-click) on the jukebox icon to quickly move forwards or backwards
in a given sequence.
A
Notes in
Triangle
Geometry

A.1 Trilinear coordinates

Let a triangle T be labeled “Euler style”: vertices (and/or angles) A, B, C , and
sidelengths a, b, c, where a D jBC j; b D jAC j; c D jABj.
Note: since in this book triangles are studied within the context of Poncelet 3-
periodics, we often refer to vertices (resp. sidelengths) as Pi (resp. si ), i D 1; 2; 3.
If X is a point in the plane T , then its position is completely determined by
the ratios of directed distances (with signal) from X to the sidelines. Such ratios
can therefore serve as coordinates for X. Any ordered triple Œp; q; r of numbers
respectively proportional to the directed distances (with signal) from X to the side-
lines BC , CA, AB are called homogeneous trilinear coordinates, or, trilinears,
for short.
Consider a point X whose trilinears are Œp; q; r. Then kp; kq and kr are the
directed, signed distances to the sidelines of T , where:
2

kD

ap C bq C cq
The above distances are known as exact trilinear coordinates. Often though, it is
sufficient to use trilinears in their unnormalized homogeneous form. The trilinears
194 A. Notes in Triangle Geometry

Figure A.1: Trilinear coordinates in the plane.

of A; B; C are given by:

     
2

2

AD ; 0; 0 ; B D 0; ; 0 ; C D 0; 0; ;0 :
a b c
As an example, consider the incenter X1 . As the center of the inscribed circle
(incircle), the distances from it to the sidelines are one and the same (the inradius),
therefore its trilinears are Œ1; 1; 1. Recall the inradius is given by r D .a C b C
c/=.2
/, where
is the area of T , see Weisstein (2019, Inradius).

A.2 More calculations with distances

Proposition A.1. Let the trilinear coordinates of two points P and Q be Œp1 ; q1 ; r1 
and Œp2 ; q2 ; r2 . Suppose that api Cbqi Ccri D 2
, i.e., the trilinear coordinates
are the actual distances to the sidelines. Then the Euclidean distance  D jP Qj
is given by:

2 sin2 B D .p1 p2 /2 C .r1 r2 /2 2j.p1 p2 /.r1 r2 /j cos B:

A.2. More calculations with distances 195

Figure A.2: Distance between points P and Q.

Also we have

2 sin2 A D .q1 q2 /2 C .r1 r2 /2 2j.r1 r2 /.q1 q2 /j cos A:

2 sin2 C D .p1 p2 /2 C .q1 q2 /2 2j.p1 p2 /.q1 q2 /j cos C:

Referring to Figure A.2:

Proof. Consider the circle having PQ as diameter and center O. Draw the seg-
ments PA1 and P C1 parallels to the sidelines AB and BC . Let C1 C 0 be also a
diameter.
By the law of cosines we have that

jA1 C1 j2 D jP A1 j2 C jP C1 j2 2jP A1 j jP C1 j cos ˛

Since C1 C 0 is a diameter it follows that ∠A1 P C1 D ∠A1 C 0 C1 . Therefore,

jA1 C1 j D jC C 0 j sin ˛ D jP Qj sin ˛ D  sin B:

Now, by the construction of A1 and C1 , we have that

jP C1 j D jp1 p2 j; jP A1 j D jr1 r2 j:

This ends the proof.

196 A. Notes in Triangle Geometry

Let ˇ ˇ ˇ ˇ ˇ ˇ
ˇq r ˇ ˇr p1 ˇ ˇ ˇ
L D ˇˇ 1 1 ˇˇ ; M D ˇˇ 1 ˇ ; N D ˇp1 q1 ˇ
q2 r2 r2 p2 ˇ ˇ p2 q2 ˇ
Denote

fL; M; N g D L2 C M 2 C N 2 2MN cos A 2LN cos B 2LM cos C:

Proposition A.2.
R2 fL; M; N g
2 D :

2
Proof. The result follows from algebraic manipulations of the three formulas ob-
tained in Proposition A.1. The details are left to the reader.

Proposition A.3. Let lx C my C nz D 0 be a straight line. Then the distances of

the vertices of a reference triangle ABC to this line are:

2
l 2
m 2
n
pD ; qD ; rD :
a fl; m; ng b fl; m; ng c fl; m; ng

A.3 Barycentric coordinates

If the trilinears of a point are Œp; q; r then its barycentric coordinates are Œap; bq; cq.
Observe that 2ap (resp. 2bq, 2cr) is the oriented area of triangle AXB (resp.
CXA, BXC ).

A.4 Conversion to and from cartesians

Trilinears Œp; q; r of a point X D .x; y/ 2 R2 can be converted to cartesians
coordinates using Kimberling (2019):

paA C qbB C rcC

XD (A.1)
pa C qb C rc
where A D .xa ; ya /, B D .xb ; yb /, C D .xc ; yc / are the vertices of the triangle
ABC expressed in cartesian coordinates.
A.5. Triangle centers 197

To convert cartesians to trilinears, consider a triangle T D ABC with vertices

A D .xa ; ya /; B D .xb ; yb / and C D .xc ; yc /. The trilinears of P D .x0 ; y0 /
are given by:
 
1 1 1
..C B/ ^ P C B ^ C / W ..A C / ^ P C C ^ A/ W ..B A/ ^ P C A ^ B/
a b c

Where u ^ v denotes the area of the oriented parallelogram generated by u and v.

A.5 Triangle centers

Let T be the set of all real triples .a; b; c/ which are sidelengths of a triangle ABC .
That is,

T D f.a; b; c/ W 0 < a < b C c; 0 < b < c C a; 0 < c < a C bg:

On any subset U of T , define a triangle center function as a nonzero function

f .a; b; c/ such that:

• f is homogeneous in a, b, c (i.e., f .t a; t b; t c/ D t n f .a; b; c/ for some

non negative integer n, t > 0, and all .a; b; c/ in U .

• f is symmetric in b and c (i.e., f .a; c; b/ D f .a; b; c/ for all .a; b; c/ in U.

A center on U is an equivalence class Œp; q; r of ordered triples .p; q; r/ given

by
p D f .a; b; c/; q D f .b; c; a/; r D f .c; a; b/

for some center function f defined on U .

Note: for compactness, sometimes the center function is expressed in terms of
sines and cosines of the angles of the triangle.
Constructions for a few triangle centers and related objects are illustrated in
Figure A.3.

A.6 Trilinear coordinates for selected triangle centers

Proposition A.4. The trilinear coordinates for X2 are given by Œ1=a; 1=b; 1=c:
198 A. Notes in Triangle Geometry

Figure A.3: Constructions for Triangle Centers Xi , i D 1; 2; 3; 4; 5; 9; 11, bor-

rowed from Reznik, Garcia, and Koiller (2020a). The incenter X1 is the inter-
section of angular bisectors, and center of the incircle (green), whose inradius is
denoted r. The barycenter X2 is where lines drawn from the vertices to oppo-
site sides’ midpoints meet. Side midpoints define the medial triangle (red). The
circumcenter X3 is the intersection of perpendicular bisectors, the center of the
circumcircle (purple) whose circumradius is denoted R. The orthocenter X4 is
where altitudes concur. Their feet define the orthic (orange). X5 is the center of
the 9-point circle (pink). The Feuerbach point X11 is the single point of contact
between the Incircle and the 9-Point circle. The excenters, i.e., the vertices of the
excentral triangle, are pairwise intersections of external bisectors. The excircles
(dashed green) are centered on the excenters and touch each side at an extouch
point. Lines drawn from each excenter through sides’ midpoints (dashed red) con-
cur at the mittenpunkt X9 . Also shown (brown) is the triangle’s Mandart inellipse,
centered on X9 and internally tangent to each side at an extouchpoint.
A.6. Selected triangle centers 199

Proof. Consider a triangle of reference T D ABC . The midpoint of the segment

BC has trilinear coordinates Œ0; 1=b; 1=c: In fact,
h a a i
0; sin C; sin B  Œ0; sin C; sin B D Œ0; sin C; sin B
2 2    
R R 1 1
 0; ;  0; ; :
b c b c
Analogously, Œ1=a; 0; 1=c and Œ1=a; 1=b; 0 are the trilinear coordinates of the
other two midpoints of ABC: Therefore, the medial lines are given by by cz D 0,
ax cz D 0 and ax by D 0. The intersection of these lines is the point
Œ1=a; 1=b; 1=c:
Proposition A.5. The trilinear coordinates of X3 are given by Œcos A; cos B; cos C :
Proof. Let O be the center of the circumcirle of ABC . Draw a perpendicular line
from P to the sideline BC .
As PO is a perpendicular bisector line, it follows that
1
∠BOP D ∠BOC D ∠BAC D A:
2
Therefore,
jBP j a
D cot A
OP 2
As a= sin A D 2R it follows that
a
cot A D R sin A cot A D R cos A:
2
Performing the same analysis with the other two vertices the result follows.
Proposition A.6. The trilinear coordinates of X4 are given by Œsec A; sec B; sec C :
Proof. The trilinear coordinates of the altitude feet relative to the side BC is given
by Œ0; 2
2

a cos C; a cos B  Œ0; cos C; cos B. This follows directly by elemen-
tary analysis of the geometry of the triangle. Analogously, the other two are given
by Œcos C; 0; cos A and Œcos B; cos A; 0. Therefore, computing the intersection of
the straight lines cos By cos C z D 0 and cos Ax cos C z D 0 it follows that
X4 D ŒsecA; secB; secC :
Proposition A.7. The trilinear coordinates of X9 are given by Œb C c a; a C c
b; a C b c  Œcot A B C
2 ; cot 2 ; cot 2 :
200 A. Notes in Triangle Geometry

Proof. Consider the excentral triangle T 0 D A0 BC 0 of T We have that X9 is the

point of concurrence of lines drawn from each excentral point to the midpoint of
the corresponding side of ABC . The excentral points have trilinear coordinates
A0 D Œ 1; 1; 1; B 0 D Œ1; 1; 1 and C 0 D Œ1; 1; 1. The lines passing through
the excentral points and the correspondent midpoints Œ0; 1=b; 1=c, Œ1=a; 0; 1=c
and Œ1=a; 1=b; 0 of the sides of the triangle T are given by

.b c/x C by cz D 0; ax C .a c/y cz D 0; ax by C .a b/z D 0:

Solving the linear system above it follows that Œx; y; z D Œb C c a; a C c

b; a C b c: Also, using the laws of cosine and sine it follows that

A cos A C 1 .a C b C c/.b C c a/R

cot D D D k.b C c a/
2 sin A abc
R.a C b C c/
kD
abc

Analogously, cot B2 D k.a C c b/ and cot C2 D k.a C b c/.

Table A.1 lists trilinears for some centers mentioned above as well as a few
others.

A.7 Some derived triangles

A derived triangle T 0 is constructed from the vertices of a reference triangle T .
It can be represented by a 3  3 vertex matrix, whose rows the trilinears of the
vertices of T 0 with respect to T . A few examples include:

• The excentral triangle is bounded by the external bisectors of a triangle. Its

vertices are known as the excenters.

• The medial triangle has vertices at the midpoints of the reference’s sides.

• The intouch triangle has vertices at the points of tangency of the incircle
with the sidelines of a reference triangle.

• The extouch triangle has vertices at the tangency points of the excircles with
the sidelines, see Figure A.4. These are also the midpoints of the perimeter
of T . For example, jA C j C jA Ce j D jB C j C jB Ce j.
A.7. Some derived triangles 201

Xk triangle center f .a; b; c/

X1 incenter 1
X2 barycenter 1=a
X3 circumcenter cos A
X4 orthocenter sec A
X5 center of Euler’s circle cos.B C /
X6 symmedian point a
X7 Gergonne point .bc/=.b C c a/
X8 Nagel point .b C c a/=a
X9 mittenpunkt bCc a
X10 Spieker center bc.b C c/
X11 Feuerbach point 1 cos.B C /
X15 1st isodynamic point sin.A C 3 /
X16 2nd isodynamic point sin.A 3 /

Table A.1: Some triangle centers and their first trilinear coordinate expressed as a
symmetric function f .a; b; c/ on the sidelengths. The complete trilinear vector is
given cyclically by Œf .a; b; c/; f .b; c; a/; f .c; a; b/. Note that sometimes these
are more concisely expressed as trig functions on the angles A; B; C which can be
converted back to f .a; b; c/ via the law of sines and/or cosines.
202 A. Notes in Triangle Geometry

Figure A.4: Intouch triangle (blue) and extouch triangle (brown).

• The anticomplementary triangle is such that its medial triangle is the original
reference triangle, see Figure A.5.
• The Feuerbach triangle has vertices at the points where the 9-point circle
touches each of the excircles, Figure A.6.

The vertex matrices for the first three are given by:

2 3 2 3 2 ac ab
3
1 1 1 0 b 1 c 1 0 a bCc aCb c
41 6 bc ab 7
1 1 5 ; 4a 1 0 c 1 5 ; 4 bCc a
0 aCb c 5
1 1 1 a 1 b 1 0 bc ac
0
bCc a a bCc

And that for the extouch triangle is:

0 1
a bCc aCb c
0 b c
B aCbCc aCb c C
@ a 0 c A
aCbCc a bCc
a b
0
A.7. Some derived triangles 203

Figure A.5: The anticomplementary triangle A0 B 0 C 0 of ABC has sides which pass
through each vertex of a reference triangle and are parallel to the opposite side. Its
circumcenter X30 coincides with the reference’s orthocenter X4 .

The trilinear vertex matrix of the anticomplementary triangle T 0 D A0 B 0 C 0 is

given by Weisstein (2019):

0 1 1 1 1
a b c
@ 1 1 1 A
a b c
1 1 1
a b c

In Weisstein (ibid., Feuerbach triangle), the trilinear vertex matrix for the Feuer-
bach triangle A1 B1 C1 is defined as:

0 1
sin2 . B 2 C / cos2 . C 2 A / cos2 . A 2 B /
@ cos2 . B C / sin2 . C 2 A / cos2 . A 2 B / A
2
cos2 . B 2 C / cos2 . C 2 A / sin2 . A 2 B /

The trilinear coordinates of X11 is given by:

Œ1 cos.B C/ W 1 cos.C A/ W 1 cos.A B/

204 A. Notes in Triangle Geometry

Figure A.6: The Feuerbach triangle A1 B1 C1 of ABC has vertices where the 9-
point circle touches the excircles. Point F D X11 is the Feuerbach point.

A.8 The (first) Brocard triangle

Referring to Figure A.7, the first Brocard point ˝ (resp. second Brocard point ˝ 0 )
of a triangle T D ABC (labeled in counterclockwise order) is the unique point
interior to T such that angles ∠˝AB, ∠˝BC and ∠˝ CA (resp. angles ∠˝ 0 BA,
∠˝ 0 CB and ∠˝ 0 AC ) are equal.
The trilinear coordinates of ˝ (resp ˝ 0 ) are Œc=b W a=c W b=a (resp. Œb=c W
c=a W a=b:). Since neither is symmetric in the last two coordinates, they are not
proper triangle centers. In fact, they are known as a bicentric pair, see Kimberling
(2020a).
Consider the six straight lines passing through A; B; C and the Brocard points
˝ and ˝.
Referring to Figure A.8, the triangle with vertices B1 D A˝ \ B˝ 0 , B2 D
C ˝ \ A˝ 0 and B3 D B˝ \ C ˝ 0 is called the first Brocard triangle.
The Brocard circle is the circumcircle of the first Brocard triangle, Figure A.8.
It contains the two Brocard points and the circumcenter X3 and the symmedian
point X6 . Its center is X182 , at the midpoint of X6 and X3 . Lines X3 X6 and ˝˝ 0
are orthogonal.
In Johnson (1929), one finds many identities concerning the Brocard points, a
few of which are reproduced below:

cot A C cot B C cot C D cot !:

A.9. Pedal and antipedal triangles 205

Figure A.7: Brocard points ˝ and ˝ 0 of a triangle ABC . X39 sits at the midpoint
of ˝˝ 0 .

p
j˝ X3 j D j˝ 0 X3 j D R 1 4 sin2 !:

p
R 1 4 sin2 !
RB D :
2 cos !
The trilinear vertex matrix of the Brocard triangle B1 B2 B3 is given by:

0 1
abc c 3 b3
@ c 3 abc a3 A
b3 a3 abc

A.9 Pedal and antipedal triangles

Referring to Figure A.9(left), the pedal triangle with respect to a point P has ver-
tices at the feet of perpendiculars dropped from P onto the sidelines of a reference
triangle.
206 A. Notes in Triangle Geometry

Figure A.8: The first Brocard triangle B1 B2 B3 and its circumcircle, also known
as the Brocard circle.
A.9. Pedal and antipedal triangles 207

Figure A.9: Left: A0 B 0 C 0 is the pedal triangle of ABC with respect to P . Right:
A00 B 00 C 00 is the antipedal of ABC wrt to P , i.e., thepedal of A00 B 00 C 00 is ABC .

Referring to Figure A.9(right), the antipedal triangle is the triangle whose

pedal triangle is the reference. One construction for it is as follows: connect P
to the vertices of reference T . For each vertex draw a line perpendicular to the
corresponding line. The antipedal triangle will be bounded by said lines.
Let Œp; q; r denote the trilinears of a point P . The vertex matrix for the pedal
(resp. antipedal) triangle with respect to P will be given by:

0 1
0 q C p cos C r C p cos B
@p C q cos C 0 r C q cos A A
p C r cos B q C r cos A 0

Respectively:

0 qCp cos C rCp cos B qCp cos C 1

rCp cos B pCq cos C pCr cos B
B .rCq cos A/ rCq cos A qCp cos C C
@ qCp cos C pCq cos C qCr cos A A
qCr cos A pCr cos B pCr cos B
rCp cos B rCq cos A qCr cos A

The pedal triangle with respect to the incenter, circumcenter, and orthocenter
are the intouch, medial, and orthic triangles.
208 A. Notes in Triangle Geometry

Figure A.10: Ceva triangle A1 B1 C1 of ABC .

A.10 Cevian triangle

Referring to Figure A.10, the Cevian (or Ceva) triangle with respect to a point
P has vertices at the intersections of lines from the vertices through P with the
opposite sides. These lines are also known as Cevians. Let P D Œp; q; r. Then
the vertex matrix for said Cevian triangle will be given by:
0 1
0 q r
@p 0 r A
p q 0

A.11 Perspective triangles

Two triangles T D ABC and T1 D A1 B1 C1 are in perspective when the three
lines AA1 , BB1 and C C1 are concurrent. This point is called the perspector of the
pair fT ; T1 g. The perspective axis of a pair of triangles is the line through the three
points of intersection of the corresponding sidelines AB \ A1 B1 , AC \ A1 C1 and
BC \ B1 C1 . See Figure A.11.

A.12 Polar triangle

Referring to Figure A.12, given a reference triangle T and a conic E, the polar tri-
angle is bounded by the polars of the vertices of T with respect to E. The reference
A.13. Circumconic 209

Figure A.11: The perspector X of triangles ABC and A1 B1 C1 and the perspective
axis.

and its polar are always perspective at a some point X , known as the perspector
of E.

A.13 Circumconic
A circumconic of a triangle contains its three vertices. A 2d family of such conics
exists. If center (or perspector) is specified, then the circumconic is unique. Points
Œx; y; z on a circumconic satisfy:
p q r
C C D0
x y z
If the perspector is supplied, the center of the circumconic is given by:

Œp. ap C bq C cr/; [Link] bq C cr/; [Link] C bq cr/

The circumcircle is the circumconic centered on X3 . Its points satisfy:

ayz C bxz C cxy D 0:

210 A. Notes in Triangle Geometry

Figure A.12: The polar triangle A1 B1 C1 of ABC with respect to a conic E are in
perspective at X .

A.14 Inconic
The inconic E 0 is tangent to the sidelines of a triangle. A 2d family of inconics
exists for any triangle. If either center or perspector is specified, then the inconic
is unique. For inconics, the perspector is also called the Brianchon point. Let its
trilinears be Œ p1 ; q1 ; 1r . Then the center is given by Œcq C br W ar C cp W bp C aq
and the inconic will satisfy:

p2x2 C q2y 2 C r 2z2 2qryz 2prxz 2pqxy D 0

A.15 Brocard inellipse

The Brocard inellipse is the inconic with parameters Œp; q; r D Œ1=a; 1=b; 1=c.
Points Œx; y; z on it satisify:

a2 b 2 z 2 2a2 bcyz C a2 c 2 y 2 2ab 2 cxz 2abc 2 xy C b 2 c 2 x 2 D 0

A.16. Ceva conjugate 211

Figure A.13: Point X is the P -Ceva conjugate of Q.

Its center is X39 D Œa.b 2 C c 2 /; b.a2 C c 2 /; c.a2 C b 2 / and the perspector

is X6 D Œa; b; c.

A.16 Ceva conjugate

Referring to Figure A.13, Let P and Q be points, neither of which lie on a sideline
of the reference triangle T D ABC . The P -Ceva conjugate X of Q is the perspec-
tor of the Cevian triangle of P and the anticevian triangle of Q, see Kimberling
(1998).

A.17 Isogonal conjugation

In the investigation of triangle geometry the isogonal conjugation is an important
tool. Referring to Figure A.14, consider a reference triangle T D ABC . Two rays
CP and CP 0 are isogonal relative to C when ∠ACP D ∠BCP 0 . Equivalently,
the ray CX1 is the common bisector of P 0 CP and ACB.
If rays AP and AP 0 are isogonal (∠BAP D ∠BAP 0 ), we say points P and
P are isogonal conjugates. It can be shown that ∠PBC D ∠P 0 BC .
0
212 A. Notes in Triangle Geometry

Figure A.14: Isogonal conjugation.

Proposition A.8. Let P and P 0 be isogonal conjugates. If P D Œp; q; r then

P 0 D Œ1=p; 1=q; 1=r D Œqr; pr; pq. The map '.P / D P 0 is an involution, i.e.,
' 2 D id .
Proof. Direct from the definition of the map '.
The isogonal conjugation is a special quadratic Cremona transformation.
Another useful construction of the isogonal conjugate is as follows. Consider
triangle T D ABC and a point P . Denote by A1 , B1 and C1 the contact points
of the sidelines of T with the incircle. Consider the reflection of line AP on AA1 ,
repeating this cyclically, i.e., reflect line BP (resp. P C ) on the bisector BB1 (resp.
C C1 ). The intersection of said reflected lines is the isogonal conjugate of P .
Proposition A.9. Let Œp; q; r be the trilinears of a point P . Its isogonal conjugate
is given by Œ1=p; 1=q; 1=r.
Proof. Consider trilinears P D Œp; q; r, A D Œ1; 0; 0, B D Œ0; 1; 0 and C D
Œ0; 0; 1. Then A1 D Œ0; 1; 1, B1 D Œ1; 0; 1 and C1 D Œ1; 0; 0. The line AP
is given by qz ry D 0 and its reflection relative to the bisector z D y is line
qy rz D 0. Cyclically, the other two reflected lines are given by xp rz D 0
and px qy D 0. The intersection of these three lines is
   
rz rz rz 1 1 1
; ; D ; ;
p q r p q r
A.18. Isotomic conjugation 213

Under isogonal conjugation, it follows that

∠PAB D ∠CAIg .P /; ∠PBC D ∠ABIg .P /; ∠P CA D ∠BCIg .P /

The isogonal conjugation is the map

Ig .Œp; q; r/ D Œ1=p; 1=q; 1=r D Œqr; pr; pq

which is an involution .Ig ı Ig D id /.

Referring to Figure A.15:

Proposition A.10. The circumcenter X3 and orthocenter X4 are isogonal conju-

gates.

Proof. Let T D ABC . The isosceles triangle BOC has angle 2A (or 2 2A/
at vertex O. Therefore the angle between ray CO and segment BC is equal to
=2 A (or A =2 ). The angle between ray CX4 and the segment CA is
=2 A (or A =2 ). Therefore, X4 and X3 are isogonal relative to C . The
same conclusion follows for the other vertices.

A.18 Isotomic conjugation

Referring to Figure A.16, consider a triangle T D ABC and a point P , consider
the intersection A1 of the line AP with the sideline BC . Reflect A1 with respect
to the midpoint Am of side BC , obtaining the point A01 . Repeat cyclically and
obtain B10 and C10 . The intersection of the three lines AA01 , BB10 and C C10 is the
isotomic conjugate of P .

Proposition A.11. Consider P D Œp; q; r specified in trilinear coordinates. Then

the isotomic conjugate of P is Œ1=.a2 p/; 1=.b 2 q/; 1=.c 2 r/. If P D Œp; q; r is
specified in barycentric coordinates, then the isotomic conjugate of P is simply
Œ1=p; 1=q; 1=r.

Proof. Let P D Œp; q; r, A D Œ1; 0; 0, B D Œ0; 1; 0 and C D Œ0; 0; 1 be the
trilinears for said points. The midpoint of the side BC is Am D Œ0; c; b. Also,
Bm D Œc; 0; a and Cm D Œb; a; 0. The line AP is given by qz ry D 0. Therefore
A1 D Œ0; q; r and A01 D Œ0; c 2 r; b 2 q.
214 A. Notes in Triangle Geometry

Figure A.15: X3 and X4 are isogonal conjugates.

Figure A.16: Isotomic conjugation.

A.19. The Euler line 215

Analogously, the line BP is given by rx pz D 0. Therefore B1 D Œp; 0; r

and B10 D Œr 2 c; 0; a2 p. Therefore the intersection of lines AA01 and BB10 is the
point  
1 1 1
; ;
a2 p b 2 q c 2 r

The isotomic conjugation is the map Im .Œp; q; r/ D Œ1=.a2 p/; 1=.b 2 q/; 1=.c 2 r/
D Œqr=a2 ; pr=b 2 ; pq=c 2  which is also an involution.

A.19 The Euler line

The trilinears Œx; y; z of a line passing through points (given in trilinears) Œu; v; w
and Œp; q; r satisfy:
ˇ ˇ
ˇx y z ˇ
ˇ ˇ
ˇ u v w ˇ D 0;
ˇ ˇ
ˇp q r ˇ

Since the Euler line contains X3 and X4 , it is given by:

ˇ ˇ
ˇ x y z ˇˇ
ˇ
ˇcos A cos B cos C ˇ D 0;
ˇ 1 ˇ
ˇ 1 1 ˇ
cos A cos B cos C

Equivalently, points Œx; y; z on the Euler line satisfy:

cos A.cos2 B cos2 C /x C cos B cos2 C cos2 A y C cos C cos2 A cos2 B z D 0:
The Euler line can also be written in terms of the sidelengths a, b, c:

a.c 2 b 2 /. a2 Cb 2 Cc 2 /xCb.a2 c 2 /.a2 b 2 Cc 2 /yCc.b 2 a2 /.a2 Cb 2 c 2 /z D 0

The point at infinity on the Euler line is its intersection with the line ax C by C
cz D 0; equivalently sin A x C sin B y C sin C z D 0.
A notable point in the Euler line is the barycenter X2 , whose trilinear are
Œ1=a; 1=b; 1=c  Œsin A; sin B; sin C .
216 A. Notes in Triangle Geometry

Proposition A.12. Consider a triangle with vertices at the following cartesian

coordinates: A D Œ˛; ˇ, B D Œ 1; 0; C D Œ1; 0. The Euler line is given by:
.3 3˛ 2 ˇ 2 /x 2˛ˇy C ˛.˛ 2 C ˇ 2 1/ D 0
Proof. In cartesians:
   
˛2 C ˇ2 1 1 ˛2
X3 D 0; ; X4 D ˛;
2ˇ ˇ
The claim follows from direct calculations.

Proposition A.13. The image of the Euler line by isogonal conjugation Ig is the
circumconic (hyperbola) given by

IE .x; y; z/ D c a2 b2 a2 C b 2 c 2 xy C b a2 c2 a2 b 2 C c 2 xz



C a b2 c 2
a2 b2 c 2 yz D 0
More general, the image of any line by Ig is a circumconic.
Proof. Write the Euler line in the parametric form E.u/ D .1 u/X3 C uX4
and compute Ig .E.u//. Now, writing in the implicit form, it is straightforward to
obtain the result stated.
Proposition A.14. The image of the Euler line under the isotomic conjugation Im
is circumconic (hyperbola) by:

IH .x; y; z/ D ab.a2 b 2 /.a2 C b 2 c 2 /xy ac.a2 c 2 /.a2 b 2 C c 2 /xz

bc.b 2 c 2 /.a2 b2 c 2 /yz D 0

A.20 Circumconic and inconic with conjugation of

barycentrics
Let the center Œu; v; w be the barycentrics of a circumconic’s center U . Then
its perspector is given by G.U /, the X2 -Ceva conjugate of U . Algebraically:
G.Œu; v; w/ D Œu.v C w u/; v.w C u v/; w.u C v w/. Let U D Œu; v; w
be the barycentrics of an inconic’s center. Then its perspector (Brianchon) will be
given by I.A.U //, where I.Œu; v; w/ D Œ1=u; 1=v; 1=w is the isotomic conju-
gate of a point and A.Œu; v; w/ D Œv C w u; w C u v; u C v w is the
anticomplement.
A.21. Billiard notes 217

Figure A.17: The elliptic billiard has X1 as its perspector; the caustic is the Man-
dart inellipse, whose Brianchon point is the Nagel point X8 .

A.21 Billiard notes

Referring to Figure A.17, the elliptic billiard can be regarded as the (fixed) cir-
cumconic centered on X9 , called elsewhere the “circumbilliard”. Its perspector is
X1 .
The caustic of the elliptic billiard is the Mandart inellipse, whose Brianchon
(perspector) is the Nagel point X8 .

Proposition A.15. The isogonal conjugate of billiard 3-periodics is the antiorthic

axis of T .

Proof. Recall that the antiorthic axis is the perspective axis of T and its excentral
triangle, and that the elliptic billiard is centered at X9 (mittenpunkt) and its per-
spector is X1 (incenter). Therefore, it follows that the antiorthic axis is given by
x C y C z D 0. The elliptic billiard is xy C xz C yz D 0.
218 A. Notes in Triangle Geometry

A.22 Exercises
Exercise A.1. Let P D Œp; q; r be an interior point of an equilateral triangle
T D ABC . Show that:
• the distances from P to the sidelines are given by Œpk; qk; rk, where k D
.2A/=.pa C qb C rc/ and A is the area of the triangle.
• k.p C q C r/ is equal to the length of the triangle’s altitude, i.e., the sum is
independent of the position of the point.
• The result is true for P exterior to the triangle, though here we need to
consider the signed distance, as in Figure A.1.
• If the sum of distances is independent of the point, then the triangle is equi-
lateral.
• Generalize the above to regular polygons.
Exercise A.2. Show that mittenpunkt X9 is the symmedian point of the excentral
triangle.
Exercise A.3. Let P and Q be isogonal conjugates in a triangle ABC . Then the
circumcenters of BP C and BQC are inverses with respect to the circumcircle of
the triangle ABC .
Exercise A.4. Let P and Q be isogonal conjugates in the of triangle ABC . Then
the pedal triangles with respect to P and Q share a circumcircle. Moreover, the
center of this circle is the midpoint of PQ.
Exercise A.5. Let P and Q be isogonal conjugates in a triangle T D ABC .
Consider triangle T 0 whose vertices lie at the reflections of P with respect to sides
AB, BC and AC . Show the circumcenter of T 0 is Q.
Exercise A.6. Let E be an ellipse inscribed in a triangle ABC (i.e., an inconic).
Then foci f1 and f2 of E are isogonal conjugates.
Exercise A.7. Consider two lines x and y passing through a point P0 . Let u
and v be conjugate lines with respect to x and y. Let P 2 u and Px 2 x and
Py 2 y be the pedal points of P . Referring to Figure A.18, show that (i) the
points fP0 ; P; Px ; Py g are concyclic; (ii) line h D Px Py is orthogonal to the line
v.
A.22. Exercises 219

Figure A.18: Isogonal line v is orthogonal to h D Px Py .

220 A. Notes in Triangle Geometry

Figure A.19: Geometric construction of Brocard points: let ABC be a triangle.

Draw a line through A and parallel to BC . Consider the circle through C and
tangent to AB at A. Let D be the intersection of said line and circle. Line BD
intersects the circle at Brocard point ˝1 . Repeat this construction for the other
vertices and obtain ˝2 .

Exercise A.8. Show the Brocard points can be obtained through the construction
in Figure A.19.

Exercise A.9. Let the incircle of triangle ABC touch side BC at A1 , and let
A1 A01 be a diameter of the incircle. Denote by A2 the intersection of lines AA01
and BC . Show that BA2 D CA1 . Consider a similar construction with respect to
the two other sides. Show that the three lines AA2 , BB2 and C C2 are concurrent.
Show that this point is X8 and that it lies on line X1 X6 .
Exercise A.10. A similarity about a point O is a composition of a rotation and
a dilation, both centered at O. Consider a quadrilateral ABCD which is not a
parallelogram. Show that there is a unique similarity sending AC to BD.
Exercise A.11. Let A; B; C; D be four distinct points in the plane, such that AC is
not parallel to BD. Let X be the intersection of AC and BD. Let the circumcircles
of ABX and CDX meet again at O. Show that O is the center of the unique
similarity that sends A to C and B to D.
Exercise A.12. If O is the center of the similarity that sends A to C and B to D,
then O is also the center of the similarity that sends A to B and C to D. See Zhao
(2021) and Chavez-Caliz (2020).
A.22. Exercises 221

Figure A.20: Polar circles intersecting orthogonally.

Exercise A.13. Let ABC be a triangle and C its circumcircle. Let the tangent
lines to C at B and C meet at D. Show that AD is a symmedian of ABC . Use
this fact to construct X6 .
Exercise A.14. Consider a triangle ABC . On sideline BC construct two points
A1 , A2 such that A2 C D c and A1 B D b. The segment A1 A2 has length
a C b C c. Repeat the construction for the other two sidelines. Show that the six
points obtained lie on a circle known as Conway’s circle.
Exercise A.15. A polar circle of an obtuse triangle T D ABC is the circle cen-
tered at the orthocenter X4 with radius r given by r 2 D 4R2 .a2 C b 2 C c 2 /=2,
where R is the radius of the circumcircle of T . Define analogously the polar cir-
cles of triangles ABH , BCH and ACH . Referring to Figure A.20, show that all
polar circles intersect orthogonally. Determine the nine-point-circles of the four
triangles.

Exercise A.16. An orthocentric system is a set of four points Pi D .xi ; yi / .i D

1; : : : ; 4/ in the plane, such that each point Pi is the orthocenter of the triangle
defined by the other three points. Show that the set of orthocentric systems is an
algebraic set of co-dimension 2 in R8 .
Exercise A.17. Show that the circumconic that is the isogonal image of Euler line
is never an ellipse.
222 A. Notes in Triangle Geometry

Exercise A.18. Consider the triangle with cartesian

p vertices A Dp Œ˛; ˇ, B D
Œ 1; 0 and C D Œ1; 0 and sides a D 2; b D .˛ 1/2 C ˇ 2 ; c D .˛ C 1/2 C ˇ 2 .
Let P1 D Œp1 ; q1 ; r1  and P2 D Œp2 ; q2 ; r2  be the trilinears of three points. Let
P .u/ D uP1 C .1 u/P2 . Consider the cartesian coordinates for said points
obtained with Equation (A.1). Let X.P1 / D P1 and X.P2 / D P2 . Show that
P  .u/ D X.P .u// is a line defined by wP1 C .1 w/P2 . Find the relation
between parameters u and w, obtaining that w D k1 u=..k1 k2 /u C k2 /.

Exercise A.19. Show that the antiorthic axis of a triangle ABC is orthogonal to
the line X1 X3 .

Exercise A.20. Consider the triangle T with vertices A D Œ 1; 0, B D Œ1; 0

and C D Œu; v. A point P is called equilateral if the Ceva triangle with respect
to P is equilateral. Find an equilateral Ceva triangle with P in the interiof of T .
Find the trilinear coordinates of the equilateral point and confirm that it is X370 .
Analyze the case where P is in the exterior of T . See Kimberling (2019).

Exercise A.21. Let I D X1 denote the incenter of a triangle ABC . Consider the
Euler lines of the four triangles BCI , CAI , ABI , and ABC . Show that these
four lines are concurrent. This point of concurrence is called Schiffler point, and
it is X21 on Kimberling (ibid.). Determine the trilinear coordinates of this point.

Exercise A.22. Consider a triangle center given in barycentric coordinates as

P D Œf .a; b; c/; f .b; c; a/; f .c; a; b/ and define the point E.P / D Œf . a; b; c/;
f .b; c; a/; f .c; a; b/. (i) Show that E is an involution and that X2 is a fixed
point of E. (ii) Determine the points E.X1 /, E.X3 / and E.X4 /. See Kimberling
(ibid.).

Exercise A.23. Consider a triangle T D ABC and triangular centers X1 and

X9 . Construct the circumconic E having center X9 and perspector X1 . Construct
the inconic Ec with the same perspector X1 . Show that the center of the inconic is
X37 and that the three points X1 , X9 , and X37 are collinear.

Exercise A.24. Consider the space of 3-gons (triangles) in R2 , up to translations

and positive homotheties. Denote this space by P.3; 2/. Show that P.3; 2/ is
diffeomorphic to the unitary sphere S3 . Let SO.2/ be the set of positive rotations
in the plane and S3 the set of permutations. Let ' W S 0.2/  P.3; 2/ ! P.3; 2/
be an action. Analyze the quocient space P.3; 2/=S 0.2/.
A.22. Exercises 223

Exercise A.25. Consider a triangle T D ABC , with sidelengths a, b and c.

Suppose that a C b C c is normalized to 2. Let sa D .b C c a/=2, sb D
.a C c b/=2 and sc D .a C b c/=2. Therefore sa ⩾ 0, sb ⩾ 0 and sc ⩾ 0.
Finally, define x 2 D 1 a, y 2 D 1 b and z 2 D 1 c. Show that the unit
sphere S2 (x 2 C y 2 C z 2 D 1) is an eight-fold covering of the space of ordered
triangles of the plane up to translations, homotheties and rotations. See Bowden
et al. (2019).

Exercise A.26. In the family of billiard 3-periodics, analyze properties of the as-
sociated spherical curve given in Exercise A.25.
B Jacobi Elliptic
Functions

A well-known reference on this topic is Armitage and Eberlein (2006).

B.1 Jacobi elliptic integral and inverse

The incomplete Jacobi elliptic integral of the first kind is defined as:
Z '
dx
u D F .'; k/ D p
0 1 k 2 sin2 x
where 0 < k < 1 is known as the elliptic modulus. The complete Jacobi integral
of the first kind is obtained by setting ' D =2.
The Jacobi amplitude function am.u; k/ computes the inverse ' of F , i.e.,
given a u, what is ' such that F .'; k/ D u.

B.2 Jacobi elliptic functions

Referring to Figure B.1, the following are the three Jacobi elliptic functions:
B.3. Basic identities 225

cn.u; k/ D [Link].u; k//

sn.u; k/ D [Link].u; k//
q
dn.u; k/ D 1 k 2 sn2 .u; k/

Note: assuming a given k, we can occasionally omit it when writing the above,
e.g., cn.u/ is shorthand for cn.u; k/.

B.3 Basic identities

cn.0/ D 1; sn.0/ D 0; dn.0/ D 1;

p
cn.K/ D 0; sn.K/ D 1; dn.K/ D 1 k 2 D k1 ;
cn.2K/ D 1; sn.2K/ D 0; dn.2K/ D 1:

Also,

sn2 .u/ C cn2 .u/ D 1

dn2 .u/ C k 2 sn2 .u/ D 1
sn0 .u/ D cn.u/dn.u/
cn0 .u/ D sn.u/dn.u/
dn0 .u/ D k 2 sn.u/cn.u/
am0 .u/ D dn.u/

cn.u/cn.v/ sn.u/sn.v/dn.u/dn.v/
cn.u C v/ D

.u; v/
sn.u/cn.v/dn.v/ C sn.v/cn.u/dn.u/
sn.u C v/ D

.u; v/
dn.u/dn.v/ k 2 sn.u/sn.v/cn.u/cn.v/
dn.u C v/ D

.u; v/

.u; v/ D 1 k sn .u/ sn2 .v/
2 2
226 B. Jacobi Elliptic Functions

1.0

0.5
sn[u,2]
dn[u,2]
2 4 6 8 10
cn[u,2]
-0.5

-1.0

Figure B.1: The three Jacobi elliptic functions sn, cn, and dn.

B.4 Connection with differential equations

It turns out sn.u; k/ is the solution to the implicit differential equation

 2
dy
D .1 y 2 /.1 k2y 2/
du

Likewise, cn.u; k/ is the solution to:

 2
dy
D .1 y 2 /.1 k2 C k2y 2/
du

Finally, dn.u; k/ is the solution to:

 2
dy
D .y 2 1/.1 k2 y 2/
du

B.5 Inverse Jacobi elliptic functions

The inverse Jacobi elliptic functions are defined as:

B.6. Complex plane extension 227

Z u
dy
arcsn.u; k/ D p
0 .1 y 2 /.1 k2y 2/
Z 1
dy
arccn.u; k/ D p
u .1 y 2 /.1 k2 C k2y 2/
Z 1
dy
arcdn.u; k/ D p
u .1 y 2 /.k 2 1 C y 2/

B.6 Complex plane extension

Jacobi’s elliptic functions can be extended to the complex plane:
p sn.z; k/ D
sin .am.z; k//, cn.z; k/ D [Link].z; k// and d n.z; k/ D 1 k 2 sn2 .z; k/,
where z 2 C, and 0 < k < 1.
These functions have two independent periods and have simple poles at the
same points. In fact:

sn.u C 4K/ D sn.u C 2iK 0 / D sn.u/

cn.u C 4K/ D cn.u C 2K C 2iK 0 / D cn.u/
dn.u C 2K/ D dn.u C 4iK 0 / D dn.u/
p
K 0 D K.k 0 /; k 0 D 1 k 2

The poles of these three functions, which are simple, occur at points

2mK C i.2n C 1/K 0 ; m; n 2 Z

They display a certain symmetry around the poles. Namely, if zp is a pole of

sn.z/, cn.z/ and dn.z/, then, for every w 2 C, we have Armitage and Eberlein
(2006, Chapter 2):

[Link] C w/ D [Link] w/
[Link] C w/ D [Link] w/
[Link] C w/ D [Link] w/
C
Ellipse-
Mounted
Brocard loci

Let a family of triangles be defined with two vertices V1 ; V2 stationary with re-
spect to an ellipse with semiaxes a; b, and a third vertex V3 D P .t / which sweeps
the boundary, P .t / D Œa cos t; b sin t. Notice this family is non-Ponceletian. We
show that over certain combinations of V1 and, V2 , the Brocard points sweep beau-
tiful, teardrop-shaped curves.

C.1 Circular sweep, one vertex at center

Referring to Figure C.1 (left):
Proposition C.1. Let b D a. The locus of the Brocard points ˝1 (resp. ˝2 ) with
V1 D .0; 0/ and V2 D .0; a/ is a circle of radius a3 (resp. a teardrop curve) of
a2 2a2
area 9 (resp. 9 ).
Proof. In this case we have that:
 
cos t 2 a sin t
˝1 .t/ D a ;
5 4 sin t 5 4 sin t
 
2 cos t sin 2t 2 sin t C cos 2t
˝2 .t / D a ;
5 4 sin t 5 4 sin t
C.2. Circular sweep, two vertices at 90-degrees 229

p
Remark C.1. The above loci intersect at aŒ˙ 3=6; 1=2; along with V2p D .0; a/
they define an equilateral. This stems from the fact that when P .t / D aŒ˙ 3=2; 1=2,
V1 V2 P .t / is equilateral and the two Brocard points coincide at the Barycenter X2 .

C.2 Circular sweep, two vertices at 90-degrees

Referring to Figure C.1 (right):

Proposition C.2. The locus of ˝1 and ˝2 with V1 D Œ0; a and V2 D Œa; 0 are a
pair of identical skewed teardrop shapes given by:

" #
sin2 t C cos t sin t 1 cos t
˝1 .t / D a ;
.sin t 2/ cos t 2 sin t C 3 .sin t 2/ cos t 2 sin t C 3

"
#
1 sin t cos2 t cos t C sin t
˝2 .t/ D a ;
.cos t 2/ sin t 2 cos t C 3 .sin t 2/ cos t 2 sin t C 3

Let D.Œx; y/ D Œy; x, the reflection about the diagonal. It follows that
˝2 .t/ D .D ı ˝1 /.t 2 /.
In Figure C.2 we show the shape of the locus varies in a complicated way when
V2 D Œ0; a and V1 D Œx; 0, with 0⩽x⩽a.

C.3 Circular sweep, antipodal vertices

Referring to Figure C.3 (left), Ferréol (2020) has contributed the following state-
ment:

Proposition C.3. With V1 D Œ a; 0 and V2 D Œa; 0, the loci of the Brocard p
points are a pair of inversely-identical teardrop shapes whose areas are a2 = 5.
The one with a cusp on V1 is given by the following quartic:

x4 2x 3 C 2y 2 x 2 C 2x 2y 2 x 1 C y 4 C 4y 2 D 0
230 C. Ellipse-Mounted Brocard loci

Proof.
 
a cos 2t 8 a cos t C a 2 a sin 2t 4 a sin t
˝1 .t / D ;
cos 2t 9 cos 2t 9
 
8 a cos t C a cos 2t a 2 a sin 2t 4 a sin t
˝2 .t / D ; :
cos .2 u/ 9 cos 2t 9
Let R.Œx; y/ D Œ x; y. Then ˝2 .t/ D .R ı ˝1 /.t/. The implicit form of ˝2 is
given by

B2 .Œx; y/ D a2 .a2 2ax 4y 2 / C 2ax.x 2 C y 2 / .x 2 C y 2 /2 D 0:

Analogously, B1 .Œx; y/ D B2 .Œ x; y/ D 0 is the implicit

R form of ˝1 .
The area of the region bounded by ˝i is given by 12 ˝i xdy ydx. It follows
p
5a2
that A.˝i / D 5 .

C.4 Ellipse sweep, two vertices at major endpoints

Figure C.3 (right) depicts the loci of ˝1 and ˝2 with P .t / on an ellipse with
semi-axes .a; b/ and with V1 D Œ a; 0 and V2 D Œa; 0. Their loci are a pair of
symmetric teardrop curves whose complicated parametric equations we omit.

C.5 Elliptic sweep, vertices on major axis

Proposition C.4. The locus of ˝1 and ˝2 with V1 D Œx1 ; 0, jx1 j ⩽ a, V2 D
Œ a; 0 and V3 D Œa cos t; a sin t are a pair of singular teardrop curves with the
following areas:

4 .x1 C a/2 a5 
A1 D
q
2 2
2
3 a C x1 4 a2 C x12


2 a2 ax1 C x12 .x1 C a/3 a2 
A2 D
2 q
3 a2 C x12 4 a2 C x12

2
p x1 D a, the ratio of A1 and A2 by the area of the circle a  both reduce
When
to 1= 5'0:4472.
C.5. Elliptic sweep, vertices on major axis 231

Figure C.1: Left: V1 and V2 are affixed to the center and top vertex of the unit
circle and a third one P .t/ revolves around the circumference. The locus of the
Brocard points ˝1 ; ˝2 are a circle (red) and a teardrop (green) whose areas are
1/9 and 2/9 that of the generating circle. The sample triangle (blue) shown is equi-
lateral, so the two Brocard points coincide. Notice the curves’ two intersections
along with the top vertex form an equilateral (orange). Right: V1 ; V2 are now
placed at the left and top vertices of the unit circle. The Brocard points of the fam-
ily describe to inversely-similar teardrop shapes. Video, Live
232 C. Ellipse-Mounted Brocard loci

Figure C.2: Loci of Brocard Points ˝1 (red) and ˝2 (green) with V2 fixed at .0; 1/
and as V1 slides from the origin along the x axis toward Œ1; 0. P .t / performs a
complete revolution on a unit circle (black). Top left: V1 D Œ0; 0, the locus of
˝1 (resp. ˝2 / is a circle (resp. a teardrop curve) of 1=9 (resp. 2=9) the area of
the external. Bottom right: when V D Œa; 0 the two loci are inversely-similar
C.5. Elliptic sweep, vertices on major axis 233

Figure C.3: Left: With antipodal V1 and V2 and P .t / revolving on the circumfer-
p
ence, the loci of the Brocards are symmetric teardrops whose area are 1= 5 that
of the circle. Right:. With V1 ; V2 at the major vertices of an ellipse of axes .a; b/,
and P .t / revolving on its boundary, the the Brocard loci (red and green) are still
symmetric (though stretched) teardrop shapes. In this case a=b D 1:5. Live
234 C. Ellipse-Mounted Brocard loci

Proof. The above is obtained with direct integration and simplification with a com-
puter algebra system (CAS).
 Bibliography

L. V. Ahlfors (1979). Complex Analysis: an Introduction to Theory of Analytic

Functions of One Complex Variable. McGraw Hill. Zbl: 0395.30001 (cit. on
p. 144).
A. Akopyan, R. Schwartz, and S. Tabachnikov (Sept. 2020). “Billiards in Ellipses
Revisited.” Eur. J. Math. (cit. on pp. 9, 19).
A. Akopyan and A. Zaslavsky (2007). Geometry of Conics. Providence, RI: Amer.
Math. Soc. (cit. on pp. 7, 53).
J. V. Armitage and W. F. Eberlein (2006). Elliptic Functions. London: Cambridge
University Press. Zbl: 1105.14001 (cit. on pp. 22, 224, 227).
M. Arnold and S. Tabachnikov (Sept. 2020). “Remarks on Joachimsthal integral
and Poritsky property.” arXiv: 2009.04988 (cit. on p. 12).
M. Bialy and S. Tabachnikov (Sept. 2020). “Dan Reznik’s Identities and More.”
Eur. J. Math. (cit. on p. 9).
H. J. M. Bos, C. Kers, and D. W. Raven (1987). “Poncelet’s Closure Theorem.”
Expo. Math. 5, pp. 289–364. Zbl: 0633.51014 (cit. on p. 3).
L. Bowden, A. Haynes, C. Shonkwiler, and A. Shukert (2019). “Spherical geom-
etry and the least symmetric triangle.” Geom. Dedicata 198, pp. 19–34. Zbl:
1418.53003 (cit. on p. 223).
C. Bradley (2011). “The Geometry of the Brocard Axis and Associated Conics.”
CJB/2011/170 (cit. on p. 59).
C. Bradley and G. Smith (2007). “On a Construction of Hagge.” Forum Geomet-
ricorum 7, pp. 231–247. Zbl: 1162.51010 (cit. on pp. 59, 63, 116).
236 Bibliography

W. Chapple (1746). “An essay on the properties of triangles inscribed in, and cir-
cumscribed about two given circles.” Miscellanea Curiosa Mathematica, The
Gentleman’s Magazine 4, pp. 117–124 (cit. on p. 48).
A. Chavez-Caliz (Aug. 2020). “More About Areas and Centers of Poncelet Poly-
gons.” Arnold Math J. (cit. on pp. 9, 220).
A. Connes and D. Zagier (2007). “A Property of Parallelograms Inscribed in El-
lipses.” The American Mathematical Monthly 114.10, pp. 909–914. Zbl: 1140.
51010 (cit. on p. 26).
H. S. M. Coxeter and S. L. Greitzer (1967). Geometry Revisited. Vol. 19. New
Mathematical Library. New York: Random House, Inc., pp. xiv+193. MR:
3155265. Zbl: 0166.16402 (cit. on pp. 7, 83).
U. Daepp, P. Gorkin, A. Shaffer, and K. Voss (2019). Finding Ellipses: What
Blaschke Products, Poncelet’s Theorem, and the Numerical Range Know about
Each Other. MAA Press/AMS. Zbl: 1419.51001 (cit. on pp. 7, 121, 126, 127,
143).
I. Darlan and D. Reznik (June 2021). “An App for Visual Exploration, Discovery,
and Sharing of Poncelet 3-Periodic Phenomena.” arXiv: 2106.04521 (cit. on
p. 163).
A. Del Centina (2016a). “Poncelet’s porism: a long story of renewed discoveries,
I.” Arch. Hist. Exact Sci. 70.1, pp. 1–122. Zbl: 1333.01015 (cit. on pp. i, 50).
— (2016b). “Poncelet’s porism: a long story of renewed discoveries, II.” Arch.
Hist. Exact Sci. 70.2, pp. 123–173. Zbl: 1345.01007 (cit. on pp. i, 3).
V. Dragović and M. Radnović (2021). “Poncelet polygons and monotonicity of
rotation numbers: iso-periodic confocal pencils of conics, hidden traps, and
marvels.” arXiv: 2103.01215 (cit. on p. 127).
V. Dragović and M. Radnović (2011). Poncelet Porisms and Beyond: Integrable
Billiards, Hyperelliptic Jacobians and Pencils of Quadrics. Frontiers in Math-
ematics. Basel: Springer (cit. on pp. 1, 116).
— (2014). “Bicentennial of the great Poncelet theorem (1813–2013): current ad-
vances.” Bull. Amer. Math. Soc. (N.S.) 51.3, pp. 373–445. Zbl: 1417.37034
(cit. on p. 3).
R. Ferréol (July 2020). Équation de la larme de Brocard 1. Private Communication
(cit. on p. 229).
C. Fierobe (2021). “On the circumcenters of triangular orbits in elliptic billiard.”
Journal of Dynamical and Control Systems. Accepted. arXiv: 1807 . 11903
(cit. on p. 9).
W. Gallatly (1914). The modern geometry of the triangle. Francis Hodgson. Zbl:
44.0595.17 (cit. on p. 5).
Bibliography 237

R. Garcia (2019). “Elliptic Billiards and Ellipses Associated to the 3-Periodic

Orbits.” American Mathematical Monthly 126.06, pp. 491–504. Zbl: 1444 .
37023 (cit. on pp. 9, 13, 20, 140).
R. Garcia and D. Reznik (Oct. 2020). “Invariants of Self-Intersected and Inversive
N-Periodics in the Elliptic Billiard.” arXiv: 2011.06640 (cit. on p. 26).
— (June 2021). “Related By Similarity I: Poristic Triangles and 3-Periodics in
the Elliptic Billiard.” Intl. J. of Geom. 10.3, pp. 52–70 (cit. on pp. 55, 59).
R. Garcia, D. Reznik, and J. Koiller (2020a). “Loci of 3-periodics in an Elliptic
Billiard: why so many ellipses?” arXiv: 2001.08041 (cit. on pp. 88, 116).
— (2020b). “New Properties of Triangular Orbits in Elliptic Billiards.” Amer.
Math. Monthly to appear. arXiv: 2001 . 08054. Zbl: 1453 . 37024 (cit. on
pp. 9, 17, 104).
— (Jan. 2021). Companion Website to “Loci of 3-periodics in an Elliptic Billiard:
Why so many ellipses?” GitHub Pages (cit. on pp. 76, 117).
B. Gibert (2021a). Catalogue of triangle cubics. URL: https : / / bernard -
[Link]/[Link] (cit. on pp. 44, 45).
— (2021b). Thomson cubic. URL: [Link] [Link]-
[Link]/Exemples/[Link] (cit. on pp. 46, 47).
G. Glaeser, H. Stachel, and B. Odehnal (2016). The Universe of Conics: From the
ancient Greeks to 21st century developments. Springer. Zbl: 1354.51001 (cit.
on pp. 7, 53).
D. Grinberg (2012). “Ehrmann’s third Lemoine Circle.” J. of Classical Geometry
1, pp. 40–52 (cit. on pp. 73, 178).
M. Helman, D. Laurain, R. Garcia, and D. Reznik (Feb. 2021). Center power and
loci of Poncelet triangles. arXiv: 2102.09438 (cit. on p. 124).
M. Helman, D. Laurain, D. Reznik, and R. Garcia (June 2021). A theory for locus
ellipticity of Poncelet 3-periodic centers. arXiv: 2106.00715 (cit. on pp. 121,
140, 141).
M. Hohenwarter et al. (2013). GeoGebra 4.4 (cit. on p. 163).
R. Honsberger (1995). Episodes in Nineteenth and Twentieth Century Euclidean
Geometry. Washington, DC: Mathematical Association of America. Zbl: 0829.
01001 (cit. on p. 79).
R. A. Johnson (1929). Modern Geometry: An Elementary Treatise on the Geometry
of the Triangle and the Circle. Boston, MA: Houghton Mifflin (cit. on p. 204).
— (1960). Advanced Euclidean Geometry. 2nd. editor John W. Young. New York,
NY: Dover. Zbl: 0090.37301 (cit. on pp. 5, 17).
V. Kaloshin and A. Sorrentino (2018). “On the Integrability of Birkhoff Billiards.”
Phil. Trans. R. Soc. A.376. Zbl: 1407.37085 (cit. on p. 3).
238 Bibliography

C. Kimberling (1998). Triangle Centers and Central Triangles. Vol. 129. Congr.
Numer. Utilitas Mathematica Publishing, Inc. Zbl: 0912 . 51009 (cit. on
p. 211).
— (2019). Encyclopedia of Triangle Centers. Zbl: 1423.05023. URL: http://
[Link]/ck6/encyclopedia/[Link] (cit. on pp. 7,
30, 62, 67, 68, 76, 83, 85, 88, 90, 92, 93, 107, 116, 136, 138, 141, 155, 164,
178, 196, 222).
— (2020a). Bicentric Pairs. URL: [Link] (cit. on pp. 173,
204).
— (2020b). Central Lines of Triangle Centers. URL: [Link]
(cit. on p. 136).
J. Koiller, D. Reznik, and R. Garcia (2021). “Average Elliptic Billiard Invariants
with Spatial Integrals.” arXiv: 2102.10899 (cit. on p. 22).
C. E. Lozada (Dec. 2016). Index of Triangles Referenced in ETC. last update: May
2021. URL: [Link]
[Link] (cit. on p. 170).
P. Moses (Sept. 2020). Expressions for the Area of Isodynamic Pedal and Isogonic
Antipedal Equilaterals. Private Communication (cit. on p. 67).
B. Odehnal (2011). “Poristic loci of triangle centers.” J. Geom. Graph. 15.1,
pp. 45–67. MR: 2858157. Zbl: 1266.51024 (cit. on pp. 9, 51, 55, 116, 118,
131).
D. Reznik (2011). Path of Incenter for Family of Triangular Orbits in Elliptical
Billiard. YouTube (cit. on p. 140).
— (2021a). Arful Loci of Triangles in Ellipses (cit. on p. 188).
— (May 2021b). Experimental Poncelet Media Assets. Observable (cit. on
p. 163).
— (2021c). Locus App Tutorial: a Video Walkthrough. YouTube (cit. on p. 164).
D. Reznik and R. Garcia (2021a). “Circuminvariants of 3-periodics in the elliptic
billiard.” Intl. J. Geometry 10.1, pp. 31–57 (cit. on p. 53).
— (Mar. 2021b). “Related By Similarity II: Poncelet 3-Periodics in the Homoth-
etic Pair and the Brocard Porism.” Intl. J. of Geom. 10.4, pp. 18–31 (cit. on
pp. 64, 111, 116).
D. Reznik, R. Garcia, and J. Koiller (2020a). “Can the Elliptic Billiard still surprise
us?” Math Intelligencer 42, pp. 6–17. Zbl: 1439.37031 (cit. on pp. 9, 140, 182,
198).
— (2020b). “The Ballet of Triangle Centers on the Elliptic Billiard.” Journal for
Geometry and Graphics 24.1, pp. 079–101. Zbl: 1453.37024 (cit. on p. 90).
Bibliography 239

O. Romaskevich (2014). “On the incenters of triangular orbits on elliptic billiards.”

Enseign. Math. 60, pp. 247–255. arXiv: 1304 . 7588. MR: 3342645. Zbl:
1371.37073 (cit. on pp. 9, 13, 140).
— (May 2019). Proof the Mittenpunkt is Stationary. Private Communication.
Rennes, France (cit. on p. 15).
R. Schwartz and S. Tabachnikov (2016a). “Centers of mass of Poncelet polygons,
200 years after.” Math. Intelligencer 38.2, pp. 29–34. MR: 3507116. Zbl:
1351.01013 (cit. on p. 9).
— (2016b). “Centers of mass of Poncelet polygons, 200 years after.” Math. Intel-
ligencer 38.2, pp. 29–34. MR: 3507116. Zbl: 1351.01013 (cit. on p. 156).
R. Shail (1996). “Some properties of Brocard points.” The Mathematical Gazette
80.489, pp. 485–491. Zbl: 0882.51009 (cit. on pp. 59, 60, 63).
H. Stachel (May 2021a). “On the Motion of Billiards in Ellipses.” arXiv: 2105.
03624v1. Zbl: 07336456 (cit. on p. 22).
— (May 2021b). “The Geometry of Billiards in Ellipses and their Poncelet Grids.”
arXiv: 2105.03362. Zbl: 07336456 (cit. on pp. 22, 23).
S. Tabachnikov (2005). Geometry and Billiards. Vol. 30. Student Mathematical
Library. Mathematics Advanced Study Semesters, University Park, PA. Prov-
idence, RI: American Mathematical Society, pp. xii+176. MR: 2168892. Zbl:
1119.37001 (cit. on pp. 3, 12).
— (July 2020). Invariant sum of squared sidelengths for N-periodics in the Ho-
mothetic Poncelet Pair. Private Communication (cit. on p. 38).
S. Tabachnikov and E. Tsukerman (2014). “Circumcenter of Mass and Generalized
Euler Line.” Discrete Comput. Geom. 51, pp. 815–836. Zbl: 1301.51023 (cit.
on p. 9).
N. C. Tak (n.d.). General Formula for Equidistant Locus of Three Points. Version
accessed on: 2016-01-27). URL: [Link]
1628779 (cit. on p. 134).
J. H. Weaver (1927). “Invariants of a poristic system of triangles.” Bull. Amer.
Math. Soc. 33.2, pp. 235–240. MR: 1561353. Zbl: 53.0604.08 (cit. on p. 50).
E. Weisstein (2009). The CRC Encyclopedia of Mathematics. 3rd. London: Chap-
man and Hall/CRC. Zbl: 1179.00007 (cit. on pp. 5, 10).
— (2019). “Mathworld.” MathWorld–A Wolfram Web Resource (cit. on pp. 5, 19,
20, 25, 27, 30, 34, 35, 38, 43, 45, 50, 53, 55, 58, 60, 62, 63, 70, 73, 74, 82, 83,
85, 93, 100, 110, 111, 138, 146, 147, 156, 158, 170, 173, 174, 176, 177, 194,
203).
S. Wolfram (2019). Mathematica, Version 10.0. Champaign, IL (cit. on p. 163).
Y. Zhao (2021). “Three Lemmas in Geometry” (cit. on p. 220).
 Index

A notable, 176
affine combination, 131, 138 circumbilliard, 53, 150
algebraic, 121 circumcircle, 17
axis circumradius, 17
antiorthic, 50, 217 closure, 1
perspective, 208 combo (barycentric), 62
computer algebra system, 7
B
concentric, 121
bicentric pair, 173
confocal, 3
billiard, 3, 19, 80, 217
conic, 9
bisected, 3
conjugate
bisection, 3
Ceva, 211
Brocard points, 109, 204, 229
isogonal, 211
C isotomic, 213
CAP, 102 conservation, 3
caustic, 3, 82 constant
Cevian, 174 Darboux’s, 20
circle Joachimsthal’s, 12
Apollonius, 62 coordinates
Brocard, 60, 110, 204 barycentric, 196
cosine, 30 trilinear, 193
Euler, 76 cubic
Lemoine, 73 Darboux, 44
240
Index 241

Thomson, 44 G
geometry
E dynamic, 3
ellipse-mounted, 168 inversive, 10
ellipses Graves’ theorem, 3
circumcircle family, 34
confocal pair, 13, 25, 131 I
dual pair, 39 incircle, 17
excentral family, 28 inellipse
homothetic pair, 38 Brocard, 59, 64, 210
incircle family, 32 MacBeath, 58
nested, 131 Mandart, 82, 217
elliptic inradius, 17
invariant, 3
functions, 224
area, 38, 111
integral, 23, 224
Brocard angle, 38
inverse functions, 226
detection, 180
modulus, 224
perimeter, 13, 150, 158
envelope, 3, 90, 173
product of areas, 151
product of excentral cosines, 19
F
ratio r=R, 17
family
semiaxes, 58, 150
antiorthic, 50
square sidelengths, 38
bicentric, 48
sum of cosines, 17, 36, 150, 158
Brocard, 64
inversion, 10
Brocard porism, 102
inversive, 147
circumcircle, 7, 34, 102, 104
iteration, 1
concentric axis-parallel, 28
confocal, 9, 11, 102 L
dual, 7, 39 limaçon, 147
excentral, 28, 102, 112 limiting points, 25, 158
focus-inversive, 147 line
homothetic, 7, 36, 64, 106 Cevian, 207
incircle, 7, 30, 103 Euler, 136, 215
non-confocal, 7 linear combination, 140
Poncelet, 166 loci, 7
poristic, 13, 48, 53, 69, 102 locus, 6
poristic excentral, 55 algebraic, 121
triangle, 166 barycenter, 104
242 Index

centroid, 156 perspector, 208

circumcenter, 76 Spieker, 200
excenters, 13 stationary, 15, 62
Feuerbach, 80 symmedian, 65, 200
incenter, 13, 127 polar image, 25, 53
Mittenpunkt, 51, 151 porism
nine-point circle, 104 bicentric, 48
non-compact, 85 Brocard, 59, 62, 70
orthocenter, 78, 104 excentral, 59
phenomena, 102 Poncelet, 1, 13
self-intersecting, 85 power of a point, 20
symmedian, 80, 104 product of excentral cosines, 19
type, 172
R
P reflection, 3
parametrization resultants (theory of), 7
Blaschke, 126
S
Jacobi, 22
semiaxes
standard, 20 caustic, 12
pencil (of circles), 25, 55 ellipse, 13
perimeter, 3 simulation, 3
point stationary, 7
1st isodynamic, 62, 200 sum
2st isodynamic, 62, 200 of cosines, 19, 150
barycenter, 200 of cotangents, 60
Bevan, 88 swans, 90
circumcenter, 200
excenter, 13 T
extouch, 83 trajectory, 1
Fermat, 107 transformation
Feuerbach, 80, 93, 200 polar, 51, 53
Gergonne, 147, 200 similarity, 51
incenter, 13, 200 triangle
limit, 158 anticomplementary, 202
Mittenpunkt, 15, 200 antipedal, 205
Nagel, 200 Brocard, 110, 204
nine-point, 200 center, 7, 121, 197
orthocenter, 200 Cevian, 173, 208
Index 243

circumcircle, 7 orthic, 83
circumradius, 17 orthopole, 173
cosine circle, 30 pedal, 69, 205
excenters, 7 perspective, 208
excentral, 202 polar, 208
exotic, 170 sidelengths, 7
extouch, 83, 202 standard, 170
extouchpoints, 83 symmedial, 146
family, 166 symmedian, 79
Feuerbach, 202 tangential, 85
incenter, 5 type, 170
incircle, 6 triangle center, 122, 173
inradius, 17
intouch, 74, 202 V
inversive, 171 vertex parametrization, 20, 41, 69
medial, 202 videos, 10
Mittenpunkt, 7 visualization, 10
 Glossary

A Poncelet polygon signed area, page 17

A0 Poncelet outer polygon signed area, page 17

J Joachimsthal’s constant of an elliptic billiard, page 12

L perimeter of a billiard N -periodic, page 12

O center of outer ellipse, page 1

Oc center of inner ellipse, page 1

Pi Poncelet polygon vertex, page 1

R triangle circumradius, page 17

Xi Kimberling center X.i /, page 13

E outer ellipse, page 1

Ec inner ellipse, page 1

Ee locus of the excenters over billiard 3-periodics, page 13

Ei locus of Xi over billiard 3-periodics, page 13

˝1 ; ˝2 a triangle Brocard points, page 60

Glossary 245

ш cot !, page 60

ı Darboux’s constant of an elliptic billiard , page 12

; i ellipse curvature at point Œx; y or at vertex Pi , respectively, page 19

! Brocard angle, page 62

sn, cn, dn elliptic Jacobi functions, page 23

 counterclockwise angle from the major axis of E to that of Ec , page 131

i Poncelet polygon internal angle, page 17

i0 Poncelet outer polygon internal angle, page 17

a; b major and minor semiaxes of E, page 12

ac ; bc major and minor semiaxes of Ec , page 12

ae ; be semiaxes of Ee , page 13

ai ; b i semiaxes of Ei , when elliptic, page 13

c half focal length of E, page 12

cc half focal lenght of Ec , page 131

dj;i distance from vertex Pi to focus fj , page 19

fj a focus of E, page 19

r triangle inradius, page 17

si Poncelet polygon sidelength, page 15

xc ; yc coordinates of Oc , page 131

altitude A segment from a vertex P on a triangle to the foot of a perpendicular

dropped from P to the opposite side., page 197

anticomplementary triangle A triangle whose sides contain the reference’s ver-

tices and are parallel to the opposite sides, page 100
246 Glossary

antiorthic axis The perspective axis of a triangle and its excentral triangle,
page 217

antipedal triangle Given a triangle T and a point P , the triangle T 0 such that T is
the pedal triangle of T 0 with respect to P , page 207

barycenter (X2 ) The center of mass of a triangle, obtained by intersecting the me-
dians, page 197

barycentric coordinates Given a triangle ABC and a point P , a triple of numbers

proportional to the oriented areas of APB, BP C and CPA, page 196

Brianchon point Point of concurrence of the three lines through the vertices of a
triangle and the points of contact of an inconic with the triangle, page 210

Brocard inellipse The inellipse centered on X39 with foci on the Brocard points,
page 210

Brocard points Unique points ˝1 and ˝2 interior to a triangle ABC such that
∠˝1 AB D ∠˝1 BC D ∠˝1 CA D ! and ∠˝2 BA D ∠˝1 CB D
∠˝2 AC D !. They are a bicentric pair of points, page 109

CAP A concentric, axis-parallel pair of ellipses admitting Poncelet 3-periodics,

page 9

CAS Computer algebra system, page 7

cevian A line from a vertex to a point on opposite sideline of a triangle, page 208

cevian triangle The triangle with vertices at the intersection of cevians through a
point P with the opposite sidelines, page 208

circumbilliard The mittenpunkt- (i.e., X9 -) centered circumellipse, page 72

circumcenter (X3 ) The center of the circumcircle, obtained by intersecting the per-
pendicular bisectors, page 197

circumcircle A circle passing through the vertices of a triangle. Its center is X3 ,

page 7

circumconic A conic passing through each of the vertices of a triangle, page 44

Glossary 247

circumconic perspector The perspector of the polar triangle with respect to the
circumconic and the reference triangle, page 222

circumellipse A circumconic which is an ellipse, page 53

circumhyperbola An circumconic which is a hyperbola, page 216

circumradius Radius of the circumcircle, page 17

confocal caustic The confocal conic to which all segments of a billiard trajectory
are tangent, as implied by Joachimsthal’s integral, page 5

contact triangle See intouch triangle, page 200

duality An incidence-preserving transformation with respect to a conic in which

points are sent to lines (their polars) or conversely, lines are sent to points
(their poles), page 112

ETC Kimberling’s Encyclopedia of Triangle Centers, page 76

ETC Kimberling’s Encyclopedia of Triangle Centers, page 196

Euler circle See nine-point circle, page 17

excenters The three intersections of the external bisectors to a triangle, page 200

excentral triangle The triangle with vertices at the excenters, page 200

excircles The three circles centered on each excenter and tangent to all sidelines,
page 26

extouch triangle The triangle whose vertices are the extouchpoints, page 200

extouchpoints The points of tangency of the excircles with the sidelines, page 200

Feuerbach point (X11 ) Where the incircle touches the nine-point circle, page 197

Feuerbach triangle The triangle whose vertices are where the 9-point circle
touches each of the excircles, page 203

focus-inversive family The inversive image of billiard N-periodics with respect to

a circle centered on a focus, page 147
248 Glossary

Gergonne point (X7 ) The perspector of a triangle and its intouch triangle,
page 197
incenter (X1 ) The center of the incircle, obtained by intersecting the three internal
angle bisectors of a reference triangle, page 197
incircle The circle touching each side of a triangle. Its center is X1 , page 7
inconic A conic tangent to the each of the sidelines of a triangle, page 42
inconic perspector See Brianchon point, page 210
inellipse An inconic which is an ellipse, page 34
inradius Radius of the incircle, page 17
intouch triangle A triangle whose vertices are the intouchpoints, page 200
intouchpoints The points of tangency of the incircle with the sidelines, page 200
invariant a quantity that is conserved in a 1d-family of periodic trajectories,
page 17
isogonal conjugate Given a point P , reflect the three P -cevians about the angular
bisectors. These meet at the isogonal conjugate of P , page 211
isotomic conjugate Given a triangle Pi , i D 1; 2; 3, and a point X, consider the
intersections Qi of cevians through X with the opposite side. Let Qi0 be the
reflection of Qi about the midpoint of the corresponding side side. Lines
Pi Qi0 meet at X 0 , the isotomic conjugate of X, page 213
Kimberling center A triangle center catalogued as Xk on Kimberling’s Encyclo-
pedia of Triangle Centers (ETC), page 76
MacBeath inconic The inellipse centered on the center X5 of the nine-point circle.
Its foci are X3 and X4 , page 58
Mandart inellipse The inconic centered on X9 , whose perspector is X8 , page 217
medial triangle A triangle with vertices at the midpoints of the sides of a reference
triangle, page 202
mittenpunkt (X9 ) Where lines from each excenter through sides’ midpoints con-
cur, page 197
Glossary 249

Nagel point (X8 ) The perspector of a triangle and its extouch triangle, page 197

NCAP A non-concentric, axis-parallel pair of ellipses admitting Poncelet 3-

periodics, page 9

nine-point center (X5 ) The center of the nine-point circle, page 197

nine-point circle A circle passing through sides’ midpoints. It also contains the
feet of altitudes and the midpoints between vertices and the orthocenter.
Its center is X5 , page 104

orthic triangle A triangle whose vertices are the feet of the three altitudes, page 39

orthocenter (X4 ) Where altitudes concur, page 197

Pascal’s limaçon Given a point P and a circle C, the limaçon (small snail) is the
envelope of all circles with centers on C which pass through P. The inver-
sive image of an ellipse with respect to a focus-centered circle is a loopless
limaçon, page 148

pedal triangle Given a point P , the triangle with vertices at the feet of perpendic-
ulars from P dropped onto the sidelines of a reference triangle, page 207

perpendicular bisector A line through a triangle’s side midpoint and perpendicular

to said side, page 197

perspective axis For two perspective triangles, the line through the (collinear) in-
tersections of corresponding sidelines, page 208

perspector For two perspective triangles, point of concurrence of lines connecting

corresponding vertices, page 208

polar The line which is the dual of a point with respect to conic, page 208

polar triangle The triangle bounded by the polars of the vertices of a triangle with
respect to a conic, page 208

pole The point which is the dual of a line with respect to a conic, page 208

poncelet iteration Sends a chord AB of a first conic E tangent to a second conic

E 0 , to a new chord BC of E, such that BC is also tangent to E 0 , page 1
250 Glossary

Poncelet trajectory The piecewise linear trajectory resulting from sequential Pon-
celet iterations, page 1

Poncelet’s porism A pair of conics which admits a closed Poncelet trajectory after
N iterations (N-periodic) is associated with a 1d family of such N-periodic
trajectories, page 13

Spieker center (X10 ) The incenter of the medial triangle, page 197

stationary point A triangle center which remains stationary over the 1d family of
3-periodic trajectories in some conic pair, page 6

Steiner circumellipse The circumellipse centered on the barycenter X2 , page 82

symmedian A cevian through X6 , page 221

symmedian point (X6 ) The intersection of the symmedians, page 197

triangle center A point on the plane of a triangle whose trilinears are

Œf .a; b; c/; f .b; c; a/; f .c; a; b/ such f is a triangle center function,
page 197

triangle center function Given a triangle with sidelenghts a; b; c, a function

f .a; b; c/ which is both homogeneous and symmetric, i.e., f .t a; t b; t c/ D
t k f .a; b; c/ and f .a; b; c/ D f .a; c; b/, page 197

trilinear coordinates Given a triangle and a point P , a triple of numbers propor-

tional to the signed distances from P to the each sideline, page 193
 Títulos Publicados — 33º Colóquio Brasileiro de Matemática

Geometria Lipschitz das singularidades – Lev Birbrair e Edvalter Sena

Combinatória – Fábio Botler, Maurício Collares, Taísa Martins, Walner Mendonça, Rob Morris e
Guilherme Mota
Códigos geométricos, uma introdução via corpos de funções algébricas – Gilberto Brito de Al-
meida Filho e Saeed Tafazolian
Topologia e geometria de 3-variedades, uma agradável introdução – André Salles de Carvalho
e Rafał Marian Siejakowski
Ciência de dados: algoritmos e aplicações – Luerbio Faria, Fabiano de Souza Oliveira, Paulo
Eustáquio Duarte Pinto e Jayme Luiz Szwarcfiter
Discovering Poncelet invariants in the plane – Ronaldo A. Garcia e Dan S. Reznik
Introdução à geometria e topologia dos sistemas dinâmicos em superfícies e além – Víctor León
e Bruno Scárdua
Equações diferenciais e modelos epidemiológicos – Marlon M. López-Flores, Dan Marchesin,
Vítor Matos e Stephen Schecter
Differential Equation Models in Epidemiology – Marlon M. López-Flores, Dan Marchesin, Vítor
Matos e Stephen Schecter
A friendly invitation to Fourier analysis on polytopes – Sinai Robins
PI-álgebras: uma introdução à PI-teoria – Rafael Bezerra dos Santos e Ana Cristina Vieira
First steps into Model Order Reduction – Alessandro Alla
The Einstein Constraint Equations – Rodrigo Avalos e Jorge H. Lira
Dynamics of Circle Mappings – Edson de Faria e Pablo Guarino
Statistical model selection for stochastic systems with applications to Bioinformatics, Linguis-
tics and Neurobiology – Antonio Galves, Florencia Leonardi e Guilherme Ost
Transfer operators in Hyperbolic Dynamics - an introduction – Mark F. Demers, Niloofar Kia-
mari e Carlangelo Liverani
A course in Hodge Theory: Periods of Algebraic Cycles – Hossein Movasati e Roberto Villaflor
Loyola
A dynamical system approach for Lane-Emden type problems – Liliane Maia, Gabrielle Norn-
berg e Filomena Pacella
Visualizing Thurston’s geometries – Tiago Novello, Vinícius da Silva e Luiz Velho
Scaling problems, algorithms and applications to Computer Science and Statistics – Rafael
Oliveira e Akshay Ramachandran
An introduction to Characteristic Classes – Jean-Paul Brasselet
a
ISBN 978-65-89124-43-6

impa
Instituto de
Matemática
Pura e Aplicada
9 786589 124436