(Kristopher Tapp) Symmetry A Mathematical Explora

Symmetry
Kristopher Tapp
Symmetry
A Mathematical Exploration
Kristopher Tapp
Department of Mathematics
Saint Joseph’s University
5600 City Ave.
Philadelphia, PA 19131
USA
[email protected]
ISBN 978-1-4614-0298-5 e-ISBN 978-1-4614-0299-2

DOI 10.1007/ 978-1-4614-0299-2
Springer New York Dordrecht Heidelberg London
Library of Congress Control Number: 2011940690
Mathematics Subject Classification (2010): 00-01, 00A06, 00A09, 00A66, 00A99, 20-01, 51-01, 97G40,
97U20
© Springer Science+Business Media, LLC 2012

All rights reserved. This work may not be translated or copied in whole or in part without the written
permission of the publisher (Springer Science+Business Media, LLC, 233 Spring Street, New York, NY
10013, USA), except for brief excerpts in connection with reviews or scholarly analysis. Use in connection
with any form of information storage and retrieval, electronic adaptation, computer software, or by similar
or dissimilar methodology now known or hereafter developed is forbidden.
The use in this publication of trade names, trademarks, service marks, and similar terms, even if they are not
identified as such, is not to be taken as an expression of opinion as to whether or not they are subject to
proprietary rights.
Printed on acid-free paper
Springer is part of Springer Science+Business Media (www.springer.com)

Symmetry is a ubiquitous
concept in mathematics and
science. Certain shapes and
images seem more symmetric than
others, yet it is not immediately
obvious how to best measure and
understand an object’s symmetry.
In fact, the quest to more precisely
understand symmetry has been a
driving force in science and mathe-
Vitruvian Man by Leonardo Da Vinci
matics, and will form the central
theme of this book. You will learn the ways in which mathe-
maticians study the topic of symmetry.
If you are curious about the

mathematical patterns underlying the
symmetry that you observe in the phy-
sical world, then this book is for you.
Why are honeycombs hexagonal?
Why are bubbles spherical? Why did
the HIV virus evolve its icosahedral
shape? What is the shape of the
An icosahedral HIV virus
universe, and how might this shape be
related to the shape of a virus? How
can one understand the symmetry of molecules or crystal
formations? How might the symmetry in a painting enhance its
artistic appeal? Parts of these answers are found in other
disciplines – biology, chemistry, physics, and art – but the
common thread is mathematics. Mathematics provides the tools
to understand and classify the possible types of symmetry that
v
vi Preface
objects may possess, which is a crucial prerequisite for

addressing questions like those above.
No background beyond high school level algebra is

required to read this book. The mathematical topics are drawn
from diverse fields including graph theory, abstract algebra, linear
algebra and topology, all of which are essential to rigorously
study symmetry. Although some of these topics are advanced,
the presentation in this book is intended to be precise and
rigorous, yet accessible to a general audience. The only real pre-
requisite is that you discard any preconceived notions of what
math is and is not, and begin this mathematical journey with an
open mind and a willingness to begin actively doing what
mathematicians do: dis-
covering patterns, invent-
ing precise language for
discussing the mathe-
matical principles under-
lying those patterns,
forming conjectures, and
eventually proving beauti-
ful theorems.
Honeycomb photo by Ken Tapp
Intended Audience
This book is primarily intended as a textbook for a one-

semester math course for math or non-math majors, including
humanities majors, with the goal of encouraging effective ana-
lytical thinking and exposing students to elegant mathematical
ideas. It includes some of the topics which are commonly found
in sampler textbooks, such as Platonic solids, Euler’s formula,
irrational numbers, countable sets, permutations and a proof of
the Pythagorean Theorem. All of these topics serve a single
compelling goal: to understand the mathematical patterns under-
lying the symmetry that we observe in the physical world. I hope
Preface vii
that students from all majors will enjoy the many beautiful
mathematical topics herein, and will come to better appreciate
the powerful cumulative nature of mathematics as these topics
are woven together into a single story about symmetry.
Instructor resources, including PowerPoint lectures and

access to all images in the book, can be found at http://www.sju.edu/
~ktapp/Symmetry
Acknowledgments
I am delighted to thank Paul Klingsberg and Susan Ramee

for suggesting many improvements to the exposition. I am also
grateful to the following artists who generously permitted me
to include their images in my book: Vladimir Bulatov, Robert
Fathauer, Brian Sanderson, Paul Söderholm and Ken Tapp.
Ta
ablle of No
ota
atio
onn
Dn = the
e n
nth dih hed
drall grouup = tthe
e syymme up of a rreg
etryy grrou gula
ar
n-g
gon
n. Itt ha
as 2
2n tottal m
meemb
berrs (n
n ro ations and n flips)).
ota
Cn = the
e n
nth cycclicc grou
up = tthe
e syym
mme etryy grou up of an n orrien
nte
ed
reg
gula
ar n-g
gon n. It ha
as n tottal mem mbeer (alll ro ota
ations)). A
sho
orth nd nottation forr th
han hese n ro
otattion
ns iis {0, 1, 2
2, …
…, n––1}.
Pn = the
e nnth peerm
muta atio
on grooupp = tthe co olle
ection off all p
perrmu
u-
tation
ns o
of n orrde
eredd th
hing
gs. It h otal me
hass n! to em
mbe
ers.
An = the
e n
nth alterna
atin
ng grooupp = the e ccollecttionn o of alll e
eve
en
permuta
atio
ons of n o
ordere
ed tthin
ngss. Itt ha
as h
half th
he ssize
eoof Pn.
Z = {…
… –3 –2, –1, 0, 1, 2,, 3, …}
3, – he sset of all inttege
… = th erss.
Q = the
e se
et o
of a
all rratio
ona
al n
num
mbe
ers (frracttion
ns llike
e 8//5 a
and
d –3
3/7
7).
R = the
e se
et o
of a
all rreal nu
umbers ((rattion
nal an
nd irration
nal nu
umb
berrs).
Rn = n
n-dime enssion
nal Euuclid
deaan space e. F mple, R2 is ca
Forr exxam alle
ed
3
the
e pllane aand
d R iss caalle
ed sspa
ace.
ix
Table of Contents
PREFACE v
INTENDED AUDIENCE ........................................................................................ VI

ACKNOWLEDGMENTS ...................................................................................... VII
TABLE OF NOTATION ........................................................................................ IX
1. INTRODUCTION TO SYMMETRY 1
A PRECISE DEFINITION OF “SYMMETRY”................................................................ 2

TYPES OF SYMMETRIES AND TYPES OF OBJECTS ....................................................... 6
THE CLASSIFICATION OF PLANE RIGID MOTIONS .................................................... 12
EXERCISES.................................................................................................... 14
2. THE ALGEBRA OF SYMMETRY 17
CAYLEY TABLES ............................................................................................. 18

SYMMETRY GROUPS....................................................................................... 21
THE POWER OF INVERSES ................................................................................ 25
AN IMPROVED CLASSIFICATION OF PLANE RIGID MOTIONS ...................................... 28
EXERCISES.................................................................................................... 30
3. ISOMORPHISM 35
WHAT IS AN ISOMORPHISM? ........................................................................... 36

ISOMORPHISM EXAMPLES ............................................................................... 39
RIGID EQUIVALENCE....................................................................................... 43
A BETTER NOTATION FOR THE CYCLIC GROUPS ...................................................... 46
EXERCISES.................................................................................................... 46
xi
xii Table of Contents
4. THE CLASSIFICATION THEOREMS 51
BOUNDED OBJECTS ........................................................................................ 51

BORDER PATTERNS ........................................................................................ 54
WALLPAPER PATTERNS ................................................................................... 56
SUMMARY ................................................................................................... 59
EXERCISES .................................................................................................... 60
5. SUBGROUPS AND PRODUCT GROUPS 63
SUBGROUPS ................................................................................................. 63
GENERATING SUBGROUPS ............................................................................... 66
PRODUCT GROUPS ......................................................................................... 69
EXERCISES .................................................................................................... 73
6. PERMUTATIONS 75
PERMUTATION GROUPS ................................................................................. 76

EVEN AND ODD PERMUTATIONS ....................................................................... 80
EXERCISES .................................................................................................... 84
7. SYMMETRIES OF SOLID OBJECTS 87
RIGID MOTIONS OF SPACE ............................................................................... 87

THE SYMMETRY GROUP A TETRAHEDRON ............................................................ 93
THE PROPER SYMMETRY GROUP A CUBE ............................................................. 96
THE PROPER SYMMETRY GROUP A DODECAHEDRON .............................................. 97
SOLID OBJECTS WHICH ARE “ESSENTIALLY TWO-DIMENSIONAL” ............................... 99
THE CLASSIFICATION THEOREM FOR BOUNDED OBJECTS ........................................ 101
CHIRALITY .................................................................................................. 104
PROPER VERSUS FULL SYMMETRY GROUPS ........................................................ 107
EXERCISES .................................................................................................. 111
Table of Contents xiii
8. THE FIVE PLATONIC SOLIDS 115
COUNTING THEIR PARTS ................................................................................ 120

DUALITY .................................................................................................... 121
EULER’S FORMULA ....................................................................................... 124
THE EULER CHARACTERISTIC........................................................................... 129
AN ALGEBRAIC PROOF THAT THERE ARE ONLY FIVE PLATONIC SOLIDS ....................... 131
THE PLATONIC SOLIDS THROUGH THE AGES ....................................................... 134
EXERCISES.................................................................................................. 135
9. SYMMETRY AND OPTIMIZATION 139
MINIMAL SURFACES ..................................................................................... 139

THE CIRCLE WINS ......................................................................................... 142
EXERCISES.................................................................................................. 147
10. WHAT IS A NUMBER? 149
NATURAL NUMBERS ..................................................................................... 149

RATIONAL NUMBERS .................................................................................... 151
REAL NUMBERS ........................................................................................... 153
WHICH REAL NUMBERS ARE RATIONAL? ........................................................... 156
HOW MANY PRIMES ARE THERE? .................................................................... 158
EXERCISES.................................................................................................. 162
11. CANTOR’S INFINITY 167
THE MODERN MEANING OF “SAME SIZE” .......................................................... 168

ARE THE RATIONAL NUMBERS COUNTABLE? ...................................................... 172
CANTOR’S THEOREM .................................................................................... 174
EXERCISES.................................................................................................. 177
xiv Table of Contents
12. EUCLIDEAN SPACE 179
THE PYTHAGORAN THEOREM AND DISTANCE FORMULA ....................................... 180

NAMING THE POINTS ON THE UNIT CIRCLE ......................................................... 183
THE DOT PRODUCT AND PERPENDICULARITY ...................................................... 185
USING THE DOT PRODUCT TO FIND A LOVER OR A SONG........................................ 187
WHAT IS A RIGID MOTION? ........................................................................... 190
TWO EXOTIC EXAMPLES ................................................................................ 193
EXERCISES .................................................................................................. 195
13. SYMMETRY AND MATRICES 199
MATRIX COMPUTATIONS .............................................................................. 199

REPRESENTING RIGID MOTIONS AS MATRICES .................................................... 202
ORTHOGONAL MATRICES .............................................................................. 206
YOU FINISHED THE BOOK. NOW WHAT?........................................................... 208
EXERCISES ............................................................................................... 209
INDEX 213
ͳǤ
Our journey starts with the question: what does “symmetry”
mean? Look at the following four objects, and rank them from the
most symmetric to the least symmetric:
How do you interpret this question in a manner which is precise

enough to lead you to a justifiable ranking of the four objects?
And how symmetric is the following painting by Robert Fathauer?
Seahorses and Eels by Robert Fathauer

http://members.cox.net/fathauerart/
To answer any of these questions, we must first make the

questions more precise.
K. Tapp, Symmetry: A Mathematical Exploration, DOI 10.1007/978-1-4614-0299-2_1, 1

2 1. Introduction to Symmetry
A Precise Definition of “Symmetry”
Most people would agree that Seahorses and Eels looks

symmetric, perhaps because it contains repeated images. But
symmetry involves more than just repeated images. A haphazard
arrangement of green eels and purple seahorses would not look
nearly as symmetric. Fathauer arranged his seahorses and eels
together like jigsaw puzzle pieces, so that the pattern of
neighbors surrounding one eel is the same as the pattern
surrounding any other.
To phrase this idea more precisely, let us imagine that the

pattern is painted onto an infinite glass wall that extends
indefinitely up, down, right and left. Imagine that the painted
pattern extends indefinitely so as to cover the whole infinite wall,
which requires infinitely many seahorses and infinitely many eels.
This infinite painting looks exactly the same from many different
positions. If the viewer is positioned in front of the eye of one
right-facing eel, then what she sees is exactly the same as if she
were positioned in front of the eye of another right-facing eel.
Here is an equivalent way

to say the same thing, rephrased
in terms of moving the glass wall
rather than moving the viewer:
there are many ways in which
the glass wall could be moved/
repositioned so that the painted
image looks exactly the same
before and after the re-position-
ing. For example the wall,
together with the pattern painted
onto it, could be translated A translation symmetry
(which means slid) so that each right-facing eel moves one

position to the right. This translation is called a symmetry of the
A Precise Definition of “Symmetry” 3
infinite painting because a viewer who closes her eyes while the
wall is moved, could not, after she opens her eyes, detect that
any change has occurred. This translation is encoded by the
length (about an inch) and direction (right) of the red arrow
pictured above. Several copies of the red arrow are included to
demonstrate that each composition element (the eye of a right-
facing eel, the tail of a down-facing eel, the center of the purple tail
spiral, etc.) moves exactly onto an identical element. That is why
a viewer would not detect the change.
This way of thinking about symmetry applies equally well

to other objects, including the orange and the purple star pictured
here. Imagine this star image is painted onto our
infinite glass wall. Again, there are several ways
in which the wall could be repositioned/moved
so that the image looks exactly the same
before and after the repositioning. For example,
it could be rotated 90° about the star’s center, or
180 or 270°. Each of these motions is called a
symmetry of the star image.
Here is our first attempt at formulating a precise definition of the

word “symmetry”:
DEFINITION: A symmetry of an object in the plane is a rigid

motion of the plane that leaves the object apparently unchanged.
In the above discussions, the “plane” was represented as

an infinite glass wall, the “object” was represented as a painting
on the wall (of a star or an infinite pattern), and a “rigid motion”
meant a moving or repositioning of the glass wall, like a rotation
or a translation. What does it mean for a rigid motion to “leave the
object apparently unchanged”? It means that, if a viewer was to
close her eyes during the repositioning, she would not detect a
difference; the object would look exactly the same when she
opened her eyes.
Precise language is crucially important in mathematics!

Our above definition of symmetry cuts straight to the underlying
reason that objects intuitively appear symmetric (they look the
same from many positions and/or contain repeated images), but
in a manner which is precise enough to form a foundation for a
rigorous mathematical investigation of symmetry. To really pull
this off, we will eventually require a more precise definition of the
term “rigid motion”. But for now, it will be enough to think of a
rigid motion as a moving/repositioning of the glass wall, like a
rotation or a translation. A rigid motion may NOT break, bend,
stretch, compress or otherwise distort distances on the glass wall
(you cannot use a glass cutter or a blow torch).
A rigid motion is always a motion of the whole plane (the

whole glass wall); one may then ask whether it is a symmetry of
any object in the plane. For example, in the illustration below, the
72° rotation about the red point is a rigid motion of the plane that
is a symmetry of the purple pentagon, but is not a symmetry of
the other shapes.
A Precise Definition of “Symmetry” 5
We emphasize that a rigid motion is completely

determined by its effect on each point of the plane; that is, two
motions that do the same thing to every point are considered the
same motion. For example, a rotation by 90° is the same as a
rotation by 90+360 = 450° about the same point. When
enumerating the symmetries of an object, we would NOT list both
a 90 and a 450° rotation, or any other such redundancies. What
matters is the effect of a motion, not the motion itself.
One type of rigid motion that we have not yet considered is

a flip over a line. Visualize a flip over a line as achieved by
flipping the glass wall over to expose its back surface (its
underside). Points along the line remain in their original position,
while points on one side of the line flip over to the opposite side
of the line. Imagine that the plane is a completely transparent
glass wall, so that any image in the plane shows through the
back surface of the glass, and looks reversed after the flip is
performed. For example, in the illustration below, the flip over the
red line is a rigid motion of the plane that is a symmetry of the
purple pentagon but is not a symmetry of the other shapes.
The words flip and reflection are synonymous; we will

henceforth use these terms interchangeably.
To better understand why the previously illustrated star

image appears symmetric, we will list all of its symmetries. There
are three obvious symmetries; namely
rotations about its center point by 90, 180,
and 270° (rotation angles are always
counterclockwise in this book). A fourth
valid symmetry is called the identity. This
is the motion that does nothing; it leaves
every point of the plane in its original
position. The identity can be considered a
The star's 4 reflection lines
rotation (by 0°) or a translation (by zero
distance). The identity is the only rigid motion that is a symmetry
of every object. Thus, the star has four rotation symmetries. The
star can also be flipped over any of the four colored lines
illustrated here. In summary, the star has four rotation
symmetries and four reflection symmetries, giving exactly eight
symmetries in total. The number of symmetries which an object
has provides a measurement of how symmetric that object is.
Types of Symmetries and Types of Objects
To describe to me your favorite rotation or translation or

flip, what information must you give me? A rotation is specified by
its center point and its (counterclockwise) angle. A translation is
specified by the length and direction of a single arrow. A flip is
specified by its reflection line.
Rotations and translations are called “proper” rigid motions.

Flips are called “improper” rigid motions. The intuitive difference
is that improper motions leave the plane’s underside facing
Types of Symmetries and Types of Objects 7
the viewer. A more precise

explanation of this difference
is obtained by comparing
how proper and improper
motions affect a right hand
or a clock. In the illustration
on the right, flipping the top
image over the red line
transforms the right hand
into a left hand and the clock
into a “counterclock” (a clock
that turns counterclockwise). A flip reverses orientation
DEFINITION: A rigid motion is called proper if it preserves

orientation, which means that after the motion is applied, an
image of a right hand still looks like a right hand and a clock still
looks like a clock. It is called improper if it reverses orientation,
which means that it turns a right hand into a left hand and a clock
into a counterclock.
An even more precise definition of proper/improper will be

discussed later when we learn about matrices. For now, let us
turn our attention to another intuitive concept which needs to be
described more precisely. We previously imagined that the
pattern in the Seahorses and Eels painting was extended
infinitely up, down, right and left, so the resulting object is
“unbounded”. On the other hand, the star image did not extend
infinitely in any direction; we could fit the entire star image into
a frame, so it is called “bounded”. We make this distinction
precise by focusing, not on the imprecise “extended infinitely”
verbiage, but instead on more precise issue of whether the object

can be framed (say by a square frame):
DEFINITION: An object in the plane is called bounded if it is fully

contained in some square in the plane. Otherwise it is called
unbounded.
The meaning would remain unaltered if the word “square” were

replaced by “circle” or “pentagon” or many other possibilities. If
an image can be framed by one of these frame shapes, then it
can be framed by all of them.
In the study of symmetry, the most important bounded

objects in the plane are the “regular polygons.”
DEFINITION: The regular n-sided polygon (also called the

regular n-gon) is the shape in the plane enclosed by n equal
length straight sides, assembled so that all n of its angles are
equal.
Thus, a regular 3-gon means an equilateral triangle, a regular

4-gon means a square, a regular 5-gon means a pentagon,
a regular 6-gon means a hexagon, and so on.
Regular polygons
The two sides of the 2-gon lie on top of each other (because they
meet at angles of 0°), so the 2-gon looks like a line segment.
Notice that each of these regular polygons has both

rotation and reflection symmetries. Can you think of a way to
orient each of these polygons, which means to alter it in such a

way that its rotation symmetries are preserved but it no longer
has any reflection symmetries? For example, here are a few
artistic ways to orient the pentagon.
Oriented pentagons
There are many other possibilities; whichever you choose, the

result is called an oriented pentagon. Each oriented pentagon
pictured above has five rotation symmetries but NO reflection
symmetries. Do you see why? A reflection would reverse the
issue of whether it appears to spin clockwise or counterclock-
wise, and would therefore not be a symmetry. Since the first
method is the simplest, we will use it to orient the other regular
polygons:
Oriented regular polygons
Here is the general definition:
DEFINITION: An object in the plane is called oriented if it has NO

improper symmetries.
You can always detect that an oriented

object has been flipped. A clock is oriented
because flipping it would make it look like a
counterclock. Similarly, each oriented
regular polygon above appears to spin
counterclockwise, but would appear to spin
clockwise after being flipped. The knotted An oriented object
blue object pictured here is oriented; hold it up to a mirror, and
notice how its over/under crossing pattern differs from that of its
mirror image. A flip would make it look like its mirror image, so a
flip could not be a symmetry.
There are two types of unbounded objects that are

classically important within the study of symmetry. First, a
wallpaper pattern intuitively means an unbounded pattern that
extends infinitely in all directions (left, right, up, and down)
according to some organized scheme. The Seahorse and Eels
painting is a wallpaper pattern (after being indefinitely extended).
Second, a border pattern (also called a Frieze pattern) means an
unbounded pattern that only extends infinitely along one line
(usually the x-axis). For example, if the following pattern is
extended infinitely to the right and left, then the result is a border
pattern:
A border pattern
Border patterns are usually drawn horizontally as above, so they

extend infinitely to the right and left, but not up or down. When
positioned like this, all of the pattern’s translation symmetries are
in directions parallel to a horizontal line. This observation helps
us to formulate a more precise definition.
DEFINITION: An unbounded pattern in the plane that has at least

one translation symmetry (besides the identity) is called
a border pattern if all its translations are parallel to a single line.
a wallpaper pattern otherwise.
The previously illustrated border pattern has many types of

symmetries. You can translate it any number of positions to the
right or left. You can reflect it over the horizontal center line.
There are also vertical lines over which you can reflect it. If you
perform any pair of the above-mentioned symmetries, one after
the other, the result will also be a symmetry. For example, if you
translate it any number of positions to the right or left and then
reflect it over the horizontal center line, then the result is aptly
called a glide reflection.
DEFINITION: A glide reflection means the result of performing a

translation (other than the identity) followed by a reflection over a
line that is parallel to the direction of the translation.
It does not matter which you do first: translate or reflect. In the

illustration below, either order has the same effect of moving the
bottom-left gnome to the top-right position.
A glide reflection translates and flips

Gnome image (used here and elsewhere) created by Paul Söderholm, www.gnurf.net.
Can you invent a border pattern with a glide reflection

symmetry that has the peculiar property that the reflection and
translation out of which it is built are not themselves symmetries
of the border pattern? An answer is found in an exercise at the
end of this chapter.
The Classification of Plane Rigid Motions
We began with vague intuitive notions of the word

“symmetry.” A symmetric object often contains repeated images,
and often looks the same from many positions. Based on these,
we formulated a mathematically precise definition: a symmetry of
an object is a rigid motion of the plane that leaves the object
apparently unchanged. We then formulated precise definitions of
other terms: “bounded”, “proper”, “oriented”, “border pattern”, and
“wallpaper pattern”. This provides us with a vocabulary for more
precisely discussing symmetry. In the remainder of the book, this
precision will serve us well. It will allow us to ask and answer
many precise questions, and eventually to prove beautiful
theorems about the possible types of symmetries that objects
may have. In this book, definitions are placed in green boxes and
theorems are placed in blue boxes.
What is still missing? Well, we defined a “symmetry” using

the term “rigid motion” but we have not yet precisely defined the
term “rigid motion”. Rather, we have relied on an intuitive feeling
for this concept. When it becomes necessary, we will eventually
give a more precise definition of “rigid motion.” To help us get by
for now, we mention the following (which our eventual precise
definition will allow us to prove):
Th
he C
Classificcatio
on off Pla
ane Rig
gid M
Motio
ons 1
13
CL
LAS
SSIIFIC
CA
ATIO
ON OF
FPPLA
ANE
ERRIG
GID MO
OTION
NS (V
VER
RSIO
ON
N 1)):
Evveryy propper riggid mo
otio
on oof tthe
e plane
e iss a tra
ansslattion
n, a ro
otattion
n,
or a rrota
atio
on ffollo
owed byy a translatio
on..
Evveryy immprrop
per rigid mo
otion o
of th
he pla
ane
e iss a flip
p orr a flip
p fo
ollow
weed
byy a ttran
nslatio
on.
In oth
herr word
w ds, there e a are noo rig
gid moti
m onss o othe
er tha an thee tyype
es
tha we have alrrea
at w ady co onsideered
d (a
andd co
ombin nations th herreo
of). Yo
ou
maay takke thiss cclasssiffica
ation a
as you ur deffinition
n oof rrigid
d m
mottionn fo
or
noow, if yyou
u likke.
T
Thee sttoryy iss evenn ssimpleer ffor rig
gid mo otio
ons th
hat are e ssymm-
meetries off a b boun ndeed ob bjecct. Thhe syymm es of a bo
mettrie oun
nde ed
ob
bjecct in
nclu
ude
e onlyy flipps andd ro
otaations (no o trranslaations)). In
n fa
act::
THHE CE ENTE ER PO OIN

NT TH HEO ORREM M: An ny b
bou
unddedd o
objeect in thhe
pla
anee ha
as a “cen nte
er poinnt” ssucch ttha
at:
(1)) E
Eve
ery proopeer ssym
mme etryy iss a rottatio
on ab
bout th
his cen
nteer p
poin
nt.
(2)) E
Eve
ery immproope er ssym
mmmetrry is a fflip ovver a lin ne tthro
ouggh thiis
centter po
oint.
Yoou m
migght thiink of an n obbjecct’ss ce
entter point as a ballancinng ppoinnt; if
you ccut thee obje ect outt off ca
ardboa ard
d annd wissh tto b
balancce it o
on yyouur
finger tip
p, tthiss is the
e ccorrrectt pllace
e to
opposiition yyou
ur finge
er.
Exercises
Challenge problems are designated ٍ.
(1) Which are proper and which are improper:
1. A proper symmetry followed by an proper symmetry

2. A proper symmetry followed by an improper symmetry
3. An improper symmetry followed by an improper symmetry
(2) How many symmetries does each capital letter in the English
language have:
A B C D E F G H I J K L M N O P Q R S T U V W X Y Z.
(3) How many symmetries does the n-gon have for each of the
values n = 2, 3, 4, 5, 6, 7 (these polygons are illustrated in the
chapter)? Guess a general formula for the number of symmetries
of an n-gon. What about an oriented n-gon?
(4) Draw a wallpaper pattern with NO improper symmetries.
(5) How many symmetries does each object have? Which object
has the most? The least?
(6) Any capital letter in the English language can be used to

create a border pattern like this:
…A A A A A A A A….
…B B B B B B B B …
…C C C C C C C C …
Exercises 15
For each of the 26 letters, decide whether the resulting border

pattern (a) contains reflections across any horizontal lines,
(b) contains reflections across any vertical lines, and (c) contains
rotations (other than by 0°).
(7) Consider the following border pattern to be extended infinitely

to the right and left.
Characterize all of its translation, rotation, reflection, and glide

reflection symmetries. What if the pattern was built from As rather
than Rs?
(8) Draw two different objects that have exactly the same
collection of symmetries.
(9) How many bounded objects can you think of that have
infinitely many symmetries?
(10) What do you think is the most symmetric object in the plane?
(11) Make sketches of several bounded objects that have

interesting collections of symmetries. Try to sketch a bounded
object whose collection of symmetries is significantly different
from that of an oriented or non-oriented polygon.
(12) “If an object has any translation symmetries (other than the
identity), then it must have infinitely many translation
symmetries.” Explain why this statement is true.
(ٍ13) “If an object has any translation symmetries (other than

the identity), then it must be an unbounded object.” Prove this
statement from scratch (without using The Center Point
Theorem). HINT: Visualize the object as painted on the (glass)
plane. Where will a single drop of paint be moved as one

translation symmetry is repeatedly performed?
(14) “A bounded object could never have any translation

symmetries other than the identity.” Explain how this statement is
related to the statement in the previous problem?
(15) Do you think that a circle can be oriented? In other words, do

you think there is an oriented object that has the same proper
symmetries as the circle? Guessing is fine – you do not need to
prove your answer.
(16) Count the proper and improper symmetries of each star

pictured below. Which are oriented?
(17) Identify some rotation symmetries of the previously pictured

Seahorses and Eels wallpaper pattern.
(18) Count the symmetries of the Seahorses and Eels painting

considered at face value as a bounded image (not extended
indefinitely into a wallpaper pattern).
(19) Perform a web image search for “M.C. Escher symmetry”.

Many of Escher’s paintings are wallpaper patterns (if you imagine
them infinitely extended). How many different angles of rotational
symmetry can you find among his wallpaper pattern paintings?
Identify some patterns that are oriented and some that are not.
ʹǤ
You learned as a child how to count, and thus became
familiar with the numbers, but that was not nearly enough. To
really understand numbers, you also needed to learn how to
perform algebraic operations on numbers, such as addition and
multiplication. Addition and multiplication are algebraic ways of
combining two numbers to get a number back as the answer.
These algebraic operations greatly enriched your ability to
understand, appreciate, and work effectively with numbers.
It is the same with symmetries. In the previous chapter,

you listed all symmetries of an object (such as a triangle or a
square). But just listing them is not nearly enough. To really
understand, appreciate, and work effectively with the collection of
symmetries that you listed, you need to learn a crucial algebraic
operation called “composition.” It allows you to combine two
symmetries of your object to get back a symmetry of that object
as the answer.
DEFINITION: If A and B are rigid motions of the plane, then A*B

denotes the rigid motion obtained by first performing B and then
performing A. It is called the composition of A with B.
If A and B are both symmetries of an object (say a triangle or

rectangle or wallpaper pattern), then A*B is also a symmetry of
that object. Thus, composition is an algebraic operation on the
collection of symmetries of that object. It combines two
symmetries of the object to get back a symmetry of the object as

18 2. The Algebra of Symmetry
the answer. Think of “A*B” as meaning “A following B” or “A

performed after B.” It is important to keep the order straight – it is
the opposite of what you might have expected.
Cayley Tables
As a child, you became familiar with the algebraic

operations of addition and multiplication by memorizing tables.
Similarly, you will now study the algebraic operation of
composition by building a table that exhibits the result of
composing any pair of symmetries.
Let us start with a square. Its eight symmetries are:
{I, R90, R180, R270, H, V, D, Dcc}
where H means horizontal flip, V means vertical flip, D and Dcc

mean the two diagonal flips, and R means a counterclockwise
rotation by the subscripted angle. The illustrations below show
the effect of these eight symmetries on a square whose corners
are labeled A, B, C, D and whose center is decorated with a
picture of a gnome. The front of the square is green and the back
is yellow.
These illustrations are followed by a table which exhibits

the composition of any pair of these symmetries. This table is
called a Cayley table for the square (or a Cayley table for the
symmetries of the square). You find a composition, A*B, in a
Cayley table like this one, by locating A along the left edge and B
along the top edge.
Cayley Tables 19
* I R90 R180 R270 H V D Dcc

I I R90 R180 R270 H V D Dcc
R90 R90 R180 R270 I Dcc D H V
R180 R180 R270 I R90 V H Dcc D
R270 R270 I R90 R180 D Dcc V H
H H D V Dcc I R180 R90 R270
V V Dcc H D R 180 I R270 R90
D D V Dcc H R270 R90 I R180
Dcc Dcc H D V R90 R270 R180 I
A Cayley table for the symmetries of a square
For example, the yellow cell in the table says that H*R90 = D.
That is, first performing R90 and then performing H results in the
same final position as D, as illustrated below:
H*R90 = D because “performing H after performing R90” yields the same ending position as
performing D.
Notice that the H reflection line does not rotate along with R90
because H is a fixed rigid motion of the plane.
This table provides a wealth of information about the symmetries

of the square. What patterns do you see? What can you learn
from this table? Here is one important observation: H*R90 = D,
while R90*H = Dcc. Order matters! “Performing R90 and then H” IS
DIFFERENT THAN “performing H and then R90.”
Next, let us try an equilateral triangle, which has the

following six symmetries:
I = the rotation by zero (the identity)

R120 = rotation by 120°
R240 = rotation by 240°
F1 = flip over line 1
F2 = flip over line 2
F3 = flip over line 3.
Fill in the table below showing the

The triangle's reflection lines
composition of any two of these
symmetries. For this task, you will need a cardboard triangle with
vertices labeled A, B, and C. To get you started, the yellow entry
means that R120*F1 = F2. When you are done, you will have a
Cayley table for the triangle. What patterns do you see? How is it
similar to the square’s Cayley table? How is it different?
Symmetry Groups 21
* I R120 R240 F1 F2 F3
I
R120 F2
R240
F1
F2
F3
A Cayley table for the symmetries of a triangle
Symmetry Groups
Composing symmetries has a lot in common with adding and

multiplying numbers. To investigate this similarity, let us first
review the familiar algebraic properties of multiplication and
addition of numbers.
Î The order in which a pair of numbers are added (or

multiplied) does not affect the result. This is called the
commutative property. In symbols:
A+B = B+A and A × B = B×A
Î The order in which a pair of additions (or a pair of

multiplications) is performed does not affect the result. This
is called the associative property. In symbols:
(A+B)+C = A+(B+C) and (A × B) × C = A×(B×C)
Î Adding 0 to a number has no effect. Multiplying a number

by 1 has no effect. In symbols:
A + 0 = 0 + A = A and 1 × A = A × 1 = A
We call 0 and 1 the identities; 0 is the additive identity and

1 is the multiplicative identity.
Î For any number A, the sum of A and –A equals 0, the

additive identity. For any nonzero number A, the product of
A and 1/A equals 1, the multiplicative identity. We call –A

the additive inverse of A, and 1/A the multiplicative inverse
of A.
Composition of symmetries shares all of these features except for

the commutative property. Do you see the similarities? Imagine
three friends:
(1) Adam likes to add numbers. Adam has a special number,

0, which has no effect when he adds it to other numbers.
That is his identity. He can find any number’s “inverse”
(the thing that adds to it to give his identity). For example,
the inverse of 35 is –35.
(2) Michelle likes to multiply numbers. Michelle has a special
number, 1, which has no effect when she multiplies it by
other numbers. That is her identity. She can find any
number’s “inverse” (the thing that multiplies by it to give
her identity). For example, the inverse of 35 is 1/35.
(3) Chris likes to compose symmetries of the square. Chris
has a special symmetry, I, which has no effect when he
composes it with other symmetries. That is his identity. He
can find any symmetry’s “inverse” (the thing that
composes with it to give his identity). For example, the
inverse of R90 is R270.
What do these three stories have in common, aside from corny

alliteration? Each of these three friends is working with a system
that mathematicians refer to as a “group.” In common English,
the word “group” has a very general meaning, but in
mathematics, it has a very specific technical meaning. Its
meaning is meant to capture the commonalities between the
systems studied by Adam, Michelle, and Chris in the above
stories.
Symmetry Groups 23
DEFINITION: A group is a set (denoted G) with an algebraic

operation (denoted •) that satisfies the following properties:
(1) G has an “identity” (denoted I) which has no effect on other

members; that is, A•I = A and I•A = A for all members, A, of G.
(2) Each member, A, of G has an inverse in G (usually denoted

A–1), which combines with it in either order to give the identity:
A•A–1 = I and A–1•A = I.
(3) The associative property holds: (A•B)•C = A•(B•C) for all

triples A,B,C of members of G.
If the commutative property also holds (which means A•B = B•A

for all pairs A,B of members of G), then we call G a commutative
group; otherwise we call G a non-commutative group.
Think of “•” as a generic symbol, which Adam replaces with

“+,”Michelle replaces with “×” and Chris replaces with “*.” In our
previous discussion of these three friends, we were too vague
about what the word “number” means. We now clarify this point.
EXAMPLE: Adam likes the infinite set of all integers, Z = {… –3,

–2, –1, 0, 1, 2, 3, …}, which is a commutative group under the
operation of addition. The identity is 0, and the inverse of A is –A.
EXAMPLE: Michelle likes the set of all non-zero real numbers,

which is a commutative group under the operation of multiplication.
The identity is 1, and the inverse of A is 1/A.
NON-EXAMPLE: The integers, Z = {… –3, –2, –1, 0, 1, 2, 3, …},

is NOT a group under the operation of division. Most integer
divisions, like 5/7, do not result in integers. A valid algebraic
operation for a group G must be a method of combining each pair
of members of G to give an answer that is a member of G.
NON-EXAMPLE: The integers, Z = {… –3, –2, –1, 0, 1, 2, 3, …},

is NOT a group under the operation of multiplication. Although 1
is the identity, the number 3 does not have an inverse in Z
(because 1/3 is not an integer).
Chris likes the eight symmetries of a square, which form a

group. There is nothing special about a square; the collection of
symmetries of ANY object forms a group!
THEOREM: The collection of symmetries of any object in the

plane forms a group under the operation of composition “*.”
This group is called the symmetry group of that object. Its identity
is the rigid motion that we previously called “the identity.” The
inverse of a symmetry, A, is denoted A–1; it is the rigid motion that
undoes what A does.
One step of proving this theorem is checking that the

associative property is valid within symmetry groups. If A, B, and
C are rigid motions, it is difficult to visualize why (A*B)*C is the
same rigid motion as A*(B*C). The trick is to imagine what
happens to any single point of the plane, visualized as a single
drop of paint on the infinite glass wall. Both (A*B)*C and A*(B*C)
effect a single point by first performing C then B and then A. They
have the same effect on each single point, so they must be the
same rigid motion.
Another step is verifying that every symmetry of an object

has an inverse. More generally, every rigid motion of the plane
has an inverse – a rigid motion that undoes it.
THEOREM: Every rigid motion of the plane has an inverse.
Although we will not give a formal proof of this fact, it should

seem believable. No matter what rigid motion I apply to the plane,
you can always move it back into its starting position. If I rotate
The Power of Inverses 25
27° clockwise, then you can rotate 27° counterclockwise. If I flip

and then translate, then you can translate back and then flip
back. If the original motion is video recorded, then its inverse can
be visualized by imagining running the video backwards.
The symmetry groups of the oriented and non-oriented

regular polygons are so important, we have special names and
symbols for them:
DEFINITION: Suppose that n 2.

The symmetry group of a regular n-gon is denoted as Dn and is
called the nth dihedral group.
The symmetry group of an oriented regular n-gon is denoted as

Cn and is called the nth cyclic group.
Recall (from Exercise #3 in Chap. 1) that Dn has n rotations

and n flips, while Cn has n rotations and no flips (the identity
counts as a rotation). This pattern can be extended to n = 1 by
defining D1 and C1 to mean the symmetry groups of the smiley
face and gnome, respectively, as shown in the illustration below.
Representative objects whose symmetry groups are Dn and Cn
The Power of Inverses
The existence of inverses is surprisingly useful. The

remainder of this chapter is devoted to some of its powerful
consequences. For example, did you notice that the square’s
Cayley table looked like a Sudoku board – all eight symmetries

appeared in each row and column. There is nothing special about
a square; we will prove that any Cayley table has this property.
THE SUDOKU THEOREM: If an object has a finite symmetry

group, then every symmetry appears exactly once in each row
and each column of its Cayley table.
PROOF: To show that an arbitrary symmetry, called A, must

appear in the column of the symmetry B, we must locate a
symmetry which, when performed after B, yields A. The correct
choice is (A*B–1). This works because:
(A*B–1)*B = A*(B–1*B) = A*I = A.
There are as many symmetries as positions in the column; since

each symmetry appears there, each must appear exactly once.
The claim about rows is proven similarly…...……..…..….….…….Ƒ
The idea of this proof actually shows that the Cayley table of any
finite group (not just a symmetry group) has the Sudoku property.
Here is another beautiful consequence of the existence of
inverses. It is not a coincidence that a regular polygon has equal
numbers of rotations and flips. If an object has any improper
symmetries at all, then it has equal numbers of proper and
improper symmetries. In other words:
THE ALL-OR-HALF THEOREM: If an object has a finite

symmetry group, then either all or half of its symmetries are
proper.
PROOF: Suppose that NOT all of the object’s symmetries are

proper. Choose one of the improper symmetries and call it F (for
Flip). On the left column of a parchment of ancient scroll, list all of
the proper symmetries: R1, R2, R3, etc. On the right column, show
the result of composing these proper symmetries with F (as
The Power of Inverses 27
illustrated). Your left column is now a list of all of

the proper symmetries. Your right column is a list
of all of the improper symmetries. The left and
right columns have the same sizes; thus, there
are equal numbers of proper and improper
symmetries. Why are the symmetries in the right
column all improper? Because an improper
symmetry composed with a proper symmetry is
always improper. Why does every improper symmetry appear
somewhere in the right column, with no repetitions? Because
every symmetry appears exactly once in F’s row of the Cayley
table……….…………………………………………………..………Ƒ
Our final application of the existence of inverses has to do

with objects that lack symmetry.
DEFINITION: An object is called asymmetric if it has no

symmetries other than the identity.
A haphazard doodle will almost certainly be asymmetric. The

next theorem says that asymmetric objects are very useful as
rigid motion detectors:
RIGID MOTION DETECTOR THEOREM: If an object is

asymmetric, then any rigid motion of the plane is uniquely
determined by knowing the object’s appearance after that motion
is applied.
To understand this theorem, let us first think about why it is NOT

true for a symmetric object, like a square. Suppose you close
your eyes and then reopen them to discover that the square has
moved 3 in. to the right. From this, you can NOT tell what motion
I performed while your eyes were closed. I might have translated
3 in. to the right or I might have rotated 90q and then translated
3 in. to the right. You have no way of knowing. The rotation part
28 2
2. Th
he A
Algeb
bra of S
Symmettry
is unnde
etecctab ble beeca
ausse it is a ssym mmmetrry of thee ssquuaree. The
follow
winggp proo
of ssho
owss th
hatt, iff the ssqu
uare
e in
n th
his sto
oryy is re
epla
aceed
byy an
naasym mm bjecct, the
mettric ob en the ere co
ouldd not be an ny und dettecct-
abble parrt.
PR
ROO OFF: SSup
ppose tha at ttwo
o differrennt riigid m onss, ccallled A and B
d moti B,
ave exxacctly the ssam
ha me eff ffecct o
on an assym mmetrric objjecct, sso tha at
wh
henn yo ou op
pen yo ourr eyyess yo ou caannot telll w whethe er A o or B w
werre
pe ormed. B
erfo But if A a andd B w werre g gennuin nely d diffe
ereent,, th n A–1*B
hen
wo
ouldd be a syymmmetry of the e ob bject tthat iss diffferrent fro
om
m the idden
ntityy,
wh
hich
h iss im ossible
mpo e beecaause th he objject is assym mme etric……… …….……Ƒ Ƒ
T
Thee prrevviou
us gnoom
me iima
agee iss a
asymmmmetric;; th
hat is wh
hy it
vissua
ally disstin
nguish
hed the
e ssym
mme
etriies of the
e sq
qua
are
e.
An
n Im
mppro
ove
ed Cla
asssifiica
atio
on of Plane R
Rig
gid
d Mot
M tion
ns
S
Sincce the e bab by ph hotto show wn
here iss asyymme etricc, th he prreviou us
the
eorremm im mpliess tthatt th herre is ON NLYY
ONNE riggid mo otio m es it ffrom
on thatt mov m itts
boottom-righ ht po osittionn tto itss ttop p-le
eft
p
possitio
on. It ssee
ems to oh havve b
bee en rrota
ate
ed
counttercclockwwisse b
by 277° ((yoou ccouuld meassurre thiss a ang gle witth a
pro actor) an
otra nd then tra d up a
ansslatted andd leeft.. In
n faactt, yyou
u coould
eve th
acchie hat same e re
epoosittion
ningg o
of thhe ph hotoo bby o onlyy d ng on
doin ne
thing: ro
otatingg bby 2
27°° a
about the e co
orrrecttly cho oseen po oint. HHoww do
you findd this p
point?
?
H
Here e’ss howw. FFirsst, ch
hosse a disstinguished poin nt oon the
ph
hoto
o, llike
e th he noose, aand w a rred lin
d draw ne conne ectting he tw
g th wo
no Next, dra
osess. N aw
w a gre eennpperp
pen ar bissector. T
ndiccula The po ointt we
w
seek liees on n tthiss g
gree ne, and it iss cchosen
en lin n sso th hat thhe
An Improve
ed C
Classificcatio
on of Pla
ane Rig
gid M
Motio
ons 2
29
illu
ustrrate
ed b
bluee
hinngee m
mea asuuress
exxacttly 27 7°. Ro otat-
ingg tthe e pplan ne byy
27 7° a aboout tthiss
speciial poin nt willl
acchieeve the illlus-
traatedd reepo
osittionningg
of the e bbab
by pho otoo,
sinncee it move es the e
no ose to the p oper
pro
plaacee a andd iit tiltss
the e p
pho oto the e ppro-
pe er a
amo ounnt! If yyouu
do o no ot owwn a p proo-
traactoor, the
ere are a alte
ernaativve m
meethoodss yoou cou uld usse tto ssolve thiis
pro oblemm. F Forr e
examp ple, yo ou coouldd sseleect se eve
eral d distiinguissheed
po ointss (n
nosse, eaarloobe, annd eyela ash), a andd lo
ocatte tthee intterssecctio
on o
of
all off the co orre
espponndinng grreen n llinees. But witth ou ur p pro
otraacto
or
me ethod, w we can eassily exxpla
n e ain why the e mmetthood w woorkss. AAs yo ou
slidde do ownn thhe gre een
n lin
ne,, th
heree iss cclea
arlyy a unniquue point wh herre
the ebbluee hing ge aanggle will cchange e frrom oo larrge to too ssma
m to all. We
W
no ow havve the e kkey idea for p provvingg thhe firsst p of the follow
parrt o win
ng
importan nt tthe
eoreem:
CL
LAS
SSIFIC
CAATIO
ONN OF
O PPLA
ANNE R
RIG
GID
D MOT
M TIO
ONS
S (V
VERS
SIO
ON 2
2):
Evveryy pproperr rrigid

d m
motion
n o
of the
e p
plan
ne is a tra
ansslattion
n o
or a
rottation.
Evveryy immprrop
per rig
gid mo
otio
on o
of tthe pla
ane
e iss a reflecctio
on o
or a g
glide
refflecctio
on.
30 2
2. Th
he A
Algeb
bra of S
Symmettry
PRROO OFF OF
O FIR RST T CCLA AIMM : Asssu uming ve ersiion 1 of the e cclasssiffi-
cationn thheoore
em (in Ch hap p. 11), all weemmusst pprovve is thiss: A rrota atio
on
followwedd byy a traansslattionn iss th
he ssam
me ass a sin nglee ro
otattion
n. WWe e will
use the ba abyy ph hotoo (or anyy o othe
er aasyymmmetricc immag ge) ass ou ur rrigiid
mo otio
on ddettecttor.. A rottation folllow
wed d by a traansslattion
n ha
as the e saamme
efffectt o e babyy p
on tthe phooto ass a siingle rottatio on wh hosse ce enteer iis
fouundd usinng tthee prrevviouuslyy exxplainnedd prrotrracctorr methhod Sincce it
d. S
haas the sa ame e efffecct o
on the e baabyy p must be the
photto, it m e ssamme rrigiid
mo otio
on…………… ………… ………… …… ………… ………… ………… ………… …… ………… ……....…Ƒ Ƒ
W
We will n not provee th
he seecond cla aim
m (a
abo
out im
mproopeer rrigiid
mo
otio
onss), b e enco
but we ourrage yyou
u to thinkk ab
bou
ut h
how
w to
o prrove
e itt.
Exxerrcissess
ou ha
(1)) Iff yo ave no ot yyet do
one
e so, fill in the C
Cayyley ttable for th
he
tria
anggle in thiss chap
pte
er.
(2)) Desccrib
be anyy p
patternns yyou
u seee in the
e Caylley tab
ble
es fo
or D3 an
nd
D4, whic
w ch w were cconnstruccted
d in
n th
he ccha
apte
er.
(3)) Consstru
uct a C
Cayle
ey ta
ablle ffor D2. Iss D2 a comm
muttative gro
oup
p?
(4)) C
Connstrrucct a CCayyleyy ttablle forr Cn forr ea achh n = 2, 3,, 44, 5
5.
Deesccribe a
anyy pa ernss and sim
atte milaarittiess yo
ou ssee
e. H
How
w iss th
he C
Cayleey
ble forr Cn re
tab elated o the C
d to Cayyleyy taable e fo
or Dn?
(ٍ5)) C Connstrrucct a Ca ayle

ey tab ble fo or D5. HIN NTT: C
Connsttrucct a
cardb boa
ard peenta agoon witth llabe ele es. Usse your pen
ed vverrtice nta
ago on tto
fill in so
omee of thhe Ca ayle ey ttab
ble, an nd ttheen ssavve tim
me b
by usiingg th
he
Su udooku prropeertyy to
o filll in
n th
he rrestt.
Exercises 31
(6) Find the inverse of each of the symmetries in the Cayley

tables you constructed in Exercises (3), (4), and (5).
(7) Does the G border pattern (an infinite strip of Gs) have a
commutative symmetry group?
…. G G G G G G G G G G G G …..
What about the C border pattern? What about Z? What about Y?
(8) Which symmetry of the square must be performed after H to

yield R270? Which must be performed before D to yield V?
(9) In D3 = the symmetry group of a triangle, solve the following

equation for X:
R120*X = F1
Solve this in two ways. First, scan down R120’s row of the Cayley
table until you find F1 – the answer is the column in which you
find it. Second, left compose each side of the equation with the
inverse of R120. In any group, do you think you can always solve
an equation of the form A*X = B for X? How?
(10) In D3, solve the following equation for X:
X*R120 = F1
Solve this in two ways. First, scan down R120’s column of the
Cayley table until you find F1 – the answer is the row in which you
find it. Second, right compose each side of the equation with the
inverse of R120. In any group, do you think you can always solve
an equation of the form X*A = B for X? How?
(11) Decide whether the following statements are true or false,

and discuss:
(1) Each cyclic group is commutative

(2) Each dihedral group is noncommutative
(ٍ12) Prove that every rotation symmetry of a border pattern is

by 0 or 180°.
HINT: Suppose that some rotation (called R) by a different angle

were a symmetry. If T is a translation symmetry, explain why
R*T*R–1 equals a translation symmetry that is NOT parallel to T?
What can you conclude?
(13) Do the four improper symmetries of a square alone form a

group under composition?
(14) Is the set of even integers a group under addition? What

about the set of odd integers?
(ٍ15) If an object has infinitely many symmetries, explain why it

must have either zero or infinitely many improper symmetries.
(16) What does the scroll in the proof of the All-or-Half Theorem
look like if the object is a square and the improper symmetry you
choose is H?
(ٍ17) What type of symmetry results if you first perform a

reflection over one line, and next perform a reflection over a
second line? (HINT: either the two lines are parallel or they inter-
sect in a point. Consider each of these possibilities separately.)
(ٍ18) Can a bounded object have only reflection symmetries (no

rotations other than the identity)? If so, then how many different
reflection symmetries could it have? Explain. (HINT: if it had more
than one reflection symmetry, what could you conclude using
your solution to the previous exercise?)
Exercises 33
(19) Physicists still do not fully understand why galaxies have

spiral shapes. What is the symmetry group of the spiral
approximated by each spiral galaxy pictured below?
NGC 1300 photo by European Southern

Observatory Whirlpool galaxy photo by NASA and ESA
͵Ǥ
This chapter is about the things that matter and the things
that do not matter. Consider this. In the last chapter, you learned
that D4 means the symmetry group of a square. Did you respond
by asking: “Which square? Where is it centered? How big is it? Is
it upright or tilted? Is it green or purple?” Later, when you studied
the symmetry group of an infinite strip of Gs, did you ask: “How
tall are the Gs? How far apart are they spaced? What color are
they?” You probably did NOT ask these questions because you
intuitively sensed that their answers do not matter. In exactly
what sense do these things not matter? To focus on what does
matter, we will need a more precise way of understanding exactly
what does not matter. That is the purpose of this chapter. That is
the purpose of an isomorphism.
First recall that it is possible for two

different objects to have exactly the same
symmetry groups. In the illustration on the
right, every symmetry of the larger red square
is also a symmetry of the smaller green square,
and vice-versa, so the symmetry groups of the
red square and of the green square are literally Two squares with the
same symmetries
identical.
But what happens if the two squares have different centers

and are tilted at different angles, as pictured below? If a rigid
motion of the plane is performed that is a symmetry of the red
square, it will move the green square to another location; thus,
the symmetry groups of the red and green squares are not

36 3. Isomorphism
literally identical. Neverthe-

less, we intuitively sense that
the two symmetry groups are
still essentially the same –
they have the same essen-
tial algebraic structure within
their Cayley tables. We
Two squares with different symmetries
will call a pair of groups
“isomorphic” if they are essentially the same. The goal of this
chapter is to formulate a precise way to express this concept.
What Is an Isomorphism?
Let us think carefully about the relationship between the

symmetry groups of the red and green squares which are
illustrated above.
Rene is studying the symmetry group of the red square. It

has the following eight
symmetries: {I, R90, R180,
R270, H, V, D, Dc}. Her
Cayley table looks exactly
like the Cayley table in
Chap. 2 (typeset in red).
Meanwhile, Gretchen is
studying the symmetry
group of the green square.
Her square has the following eight symmetries: {I, R90, R180, R270,
F1, F2, F3, F4}. Since her reflection lines are not quite horizontal,
vertical or diagonal, she chose different symbols for naming
them. Even her rotations differ from Rene’s rotations. For example,
Rene’s R90 is a rotation about the center of the red square, while
Gretchen’s R90 is a rotation about the center of the green square –
these are different rigid motions of the plane.
What Is an Isomorphism? 37
Nevertheless, Gretchen’s Cayley table is essentially the

same as Rene’s in the following precise sense. Starting with
Rene’s red Cayley table, if I ask my word processor to translate
red symbols into green symbols using the following symbol-
replacement dictionary:
I R90 R180 R270 H Dc V D

Ľ Ľ Ľ Ľ Ľ Ľ Ľ Ľ
I R90 R180 R270 F1 F2 F3 F4
I will end up with a valid Cayley table for Gretchen’s green group.
Do you see
* I R90 R180 R270 H V D D’
why the resultant I I R90 R180 R270 H V D D’
green table is a R90 R90 R180 R270 I D’ D H V
R180 R180 R270 I R90 V H D’ D
valid Cayley table R270 R270 I R90 R180 D D’ V H
for Gretchen’s green H H D V D’ I R180 R90 R270
V V D’ H D R180 I R270 R90
group (which means
D D V D’ H R270 R90 I R180
that each of its 64 D’ D’ H D V R90 R270 R180 I
cells is filled in
correctly)? If you * I R90 R180 R270 F1 F3 F4 F2
start checking cells, I I R90 R180 R270 F1 F3 F4 F2
you will quickly see R 90 R 90 R 180 R 270 I F 2 F 4 F 1 F3
R180 R180 R270 I R90 F3 F1 F2 F4
why things are R270 R270 I R90 R180 F4 F2 F3 F1
working out. Our F1 F1 F4 F3 F2 I R180 R90 R270
F F F F F R I R270 R90
symbol-replacement F34 F34 F23 F12 F41 R180 270 R90 I R180
dictionary was very F2 F2 F1 F4 F3 R90 R270 R180 I
carefully constructed. The reflection lines listed counterclockwise
around the red square were matched with the reflection lines
listed counterclockwise around the green square. Said differently,
this dictionary is induced by the rigid motion that slides and tilts
the red square on top of the green square, matching up the two
sets of reflection lines.
38 3. Isomorphism
In summary, each cell in the green table is filled in correctly,

which really means this: Our dictionary translates every true red
equation into a true green equation! For example the yellow
highlighted cell of the red table represents “H*R90 = D” which is a
true red equation. This equation is symbol-by-symbol translated
into “F1*R90 = F4” which is a true green equation, represented by
the yellow highlighted cell of the green table.
The symmetry groups of the two squares are “isomorphic.”

The dictionary between their members is an “isomorphism.” Here
is the precise definition:
DEFINITION: An isomorphism between two groups means a

one-to-one matching (dictionary) between their members that
translates each true equation in one group into a true equation in
the other. This is the same as saying that it translates an entire
Cayley table for one group into a valid Cayley table for the other.
We say two groups are isomorphic if there exists an isomorphism

between them.
When two groups are isomorphic, we think of them as

essentially the same; they have the same algebraic structure
within their Cayley tables. Imagine an observer who does not
know or care what the names of the group members represent
(symmetries of a red square or of a green square or whatever).
This observer would study the two Cayley tables and discover
exactly the same patterns. From this observer’s perspective, the
two groups would look like a single group represented in two
different notational systems.
An isomorphism is analogous to a language dictionary.

For example, the square’s Cayley table in Chap. 2 would look
different if this textbook was translated into Swahili. Cosmetic
Isomorphism Examples 39
differences would arise because the Swahili word for “Rotation”

does not begin with “R,” nor does the word for “Horizontal” begin
with “H.” But these irrelevant notational differences do not matter
– Swahili readers will learn the same things as English readers. It
is a single group represented in two different notational systems.
The dictionary between the English and Swahili notational
systems is like an isomorphism. It translates true equations into
true equations. For example, when the Swahili symbol for H is
composed with the Swahili symbol for R90 it had better equal the
Swahili symbol for D; otherwise the translator needs to be fired.
Isomorphism Examples
Example: The Star and The Moth
Moth photo by Ken Tapp
The swirly star above has exactly two symmetries: the

identity and a rotation by 180° about its center (like C2). The moth
also has two symmetries: the identity and a vertical flip (like D1).
Here are their Cayley tables:
star I R180 moth I V

I I R180 I I V
R180 R180 I V V I
The symmetry group of the star is isomorphic to the symmetry

group of the moth. The isomorphism is the following one-to-one
matching between their members: I ļ I R180 ļ V.
40 3. Isomorphism
This isomorphism translates each star symmetry into a

moth symmetry. It deserves to be called an isomorphism
because it translates the star’s entire Cayley table into the moth’s
entire Cayley table. It does not matter that a flip is geometrically
different from a rotation. All that matters is the underlying
algebraic patterns in the Cayley tables. An observer who did not
know the meaning of the symbols R180 or V would look at the two
Cayley tables and describe them in the same way: they both
have exactly one non-identity member which is its own inverse.
Thus, C2 is isomorphic to D1.
Example: George and Peter
George is studying the symmetry group of a border pattern

constructed from an infinite strip of green Gs:
…. G G G G G G G G G G G G G G ….
The subsequent Gs are 1 cm apart. He calls his group G. Its

members are: G = {…,T–3, T–2, T–1, T0, T1, T2, T3,…}, where Tn
denotes the translation by n centimeters to the right (if n is
positive) or to the left (if n is negative).
Meanwhile, Peter is studying the symmetry group of a

border pattern constructed from an infinite strip of purple Ps:
…P P P P P P P P P…
The subsequent Ps are 2 cm apart. He calls his group P. Its

members are P = {…,T–6, T–4, T–2, T0, T2, T4, T6,…}. George and
Peter chat over coffee and come to suspect that their groups
might be isomorphic. The alleged isomorphism they construct is
the following one-to-one matching:
… T–4 T–3 T–2 T–1 T0 T1 T2 T3 T4 …

Ľ Ľ Ľ Ľ Ľ Ľ Ľ Ľ Ľ
… T–8 T–6 T–4 T–2 T0 T2 T4 T6 T8 …
Isomorphism Examples 41
The pattern doubles each green subscript, matching T3 with T6

for example. Since T3 and T6 each translates its border pattern
three letters to the right, George and Peter believe that this is an
isomorphism. They feel that counting letters matters more than
counting centimeters, and that color does not matter a bit.
To check that their matching is really an isomorphism, they

must verify that each true green equation becomes a true purple
equation. For example, “T5*T8 = T13” is a true green equation. If
we double all subscripts, we get “T10*T16 = T26”, which is in fact a
true purple equation. There is nothing special about 5 and 8; for
any integers m and n, their matching gives the following:
Tm * Tn = Tm+n
Ļ Ļ Ļ
T2m * T2n = T2m+2n.
The translated purple equation is true. This verifies that each true
green equation is translated into a true purple equation, and
therefore that their matching is really an isomorphism!
Example: George and The Integers
When our friend George (from the previous example)

composes his symmetries, he is really just adding centimeters.
For example, T5*T8 = T13 because 5+8 = 13. He is performing
integer addition in disguise. He is suddenly struck by a revelation:
his group is isomorphic to the additive group of integers,
Z = {…,–3, –2, –1, 0, 1, 2, 3, …}.

The isomorphism that he constructs between G and Z
simply matches Tn with n. For example, it matches the symmetry
T7 with the integer 7. As evidence that this is really an iso-
morphism, George observes that the true equation “T5*T8 = T13”
in G is translated into “5 + 8 = 13,” which is a true equation in Z.
Notice that the translated equation has “+” instead of “*” because
+ is the algebraic operation in Z. As before, there is nothing
42 3. Isomorphism
special about 5 and 8; every true equation in G is translated into

a true equation in Z:
Tm * Tn = Tm+n
Ļ Ļ Ļ
m + n = m+n.
Since G is a symmetry group of a border pattern and Z is
not a symmetry group at all (it is just a bunch of integers), George
worries that he is comparing apples and oranges when he claims
that G and Z are isomorphic. In fact, he is allowed to make this
comparison because G and Z have in common that they are both
groups.
NON-EXAMPLE: The dihedral group D4 is NOT isomorphic to the

cyclic group C10 because they have different sizes. D4 has eight
members, while C10 has ten members. Isomorphic groups always
have the same size, since an isomorphism is a one-to-one
matching between their members.
NON-EXAMPLE: The dihedral group D4 is NOT isomorphic to the

cyclic group C8. Even though they both have eight members, C8
is commutative while D4 is non-commutative. A commutative
group could never be isomorphic to a non-commutative group.
Think about why. Thus, although each object pictured below has
eight symmetries, there is an essential difference between their
symmetry groups (which are D4 and C8, respectively). These two
objects are symmetric in different ways.
Each star has 8 symmetries, but their symmetry groups are not isomorphic.
Rigid Equivalence 43
Rigid Equivalence
In the previous chapter, we defined Dn to mean the

symmetry group of a regular n-gon. We did not bother to specify
irrelevant details like where the n-gon was centered or how it was
tilted. These details do not matter because moving or rotating an
object (or transforming it by any rigid motion) does not really
change its symmetry group. Here is the precise way to express
this:
THEOREM: Performing a rigid motion does not essentially

change an object’s symmetry group. More precisely, its
symmetry groups before and after the object is transformed by
the rigid motion are isomorphic to each other.
PROOF: Let us call the rigid motion M, the original object X, and
the transformed object Y, as illustrated below. It is visually
apparent that M matches every symmetry of X with a symmetry
of Y. For example, M moves each reflection line of X to a
reflection line of Y, and M moves the rotation center of X to the
rotation center of Y.
Here is a clever way to

more precisely specify the
manner in which M induces a
matching between the sym-
metries of X and Y. Every
symmetry, A, of X gets
matched with M*A*M–1. Do you
see why M*A*M–1 is a symmetry of Y? First M–1 is performed
(which moves Y to X), then A is performed (which moves X to X
because A is a symmetry of X), and then M is performed (which
moves X back to Y). The net result is a symmetry of Y.
44 3. Isomorphism
To verify that this matching is an isomorphism, we must

check that it translates true equations into true equations.
Suppose A*B = C is a true equation in the symmetry group of X.
This equation translates into (M*A*M–1)*(M*B*M–1) = M*C*M–1,
which is a true equation because its left side simplifies to its right
side as follows:
LEFT = M*A*M–1*M*B*M–1 = M*A*I*B*M–1 = M*A*B*M–1

= M*C*M–1 = RIGHT.
Think on your own about why this matching is one-to-one……....Ƒ
Since the symmetry group of an object is unaffected by rigid

motions, it is useful to define the following:
DEFINITION: Two objects in the plane are called rigidly

equivalent if there exists a rigid motion of the plane which, when
applied to the first object, repositions it so that afterwards the two
objects have exactly the same symmetries.
The two stars pictured

are rigidly equivalent because
the purple one can be tilted and
slid on top of the silver one so
that, after this repositioning, the
two objects have symmetry
groups which are not just
isomorphic but are literally
identical. Every symmetry of the These two stars are rigidly equivalent
silver star is a symmetry of the
repositioned purple star, and every symmetry of the repositioned
purple star is a symmetry of the silver star.
THEOREM: If two objects are rigidly equivalent, then their

symmetry groups are isomorphic.
Rigid Equivalence 45
PROOF: The symmetry group of the first object is isomorphic to

the symmetry group of the repositioned first object, which is identi-
cal to the symmetry group of the second object………...….…….…Ƒ
Let us reconsider the examples from the previous section

in terms of this new notion of rigid equivalence:
[The red & green squares] These two squares are rigidly
equivalent, which is the real reason that their symmetry groups
turned out to be isomorphic.
[The star & the moth] The symmetry group of the star is
isomorphic to the symmetry group of the moth, even though the
star is NOT rigidly equivalent to the moth. A rigid motion cannot
change the fact that the star has a rotation, while the moth has a
flip.
[George & Peter] George’s G border pattern is NOT rigidly

equivalent to Peter’s P border pattern. A rigid motion cannot
change the fact that their smallest translation lengths (1 and 2 cm
respectively) are different. If the G border pattern was enlarged
by a scaling factor of 2, then the two border patterns would
become rigidly equivalent.
These examples help us to understand the interrelated

concepts of rigid equivalence and isomorphism. It is all about
what matters and what does not matter. In the next chapter, we
will answer this difficult question: What are all of the possible
ways in which (1) a bounded objects, (2) a border pattern, and
(3) a wallpaper patterns can be symmetric? But before answering
this question, we must decide what it means. When do we
consider two objects to be symmetric in the same way? Should
this mean that the objects are rigidly equivalent? Or should it
mean that the objects have isomorphic symmetry groups? Or
maybe it should mean that one object is rigidly equivalent to a
rescaling (enlarging or shrinking) of the other object. All of these
46 3. Isomorphism
possibilities are reasonable, and in some cases, these different

possibilities lead to different answers to the question.
That is the reason this chapter is important. Our main

challenge in this book is to understand the different ways in which
objects can be symmetric. The concepts of isomorphism and rigid
equivalence provide us with a language for more precisely saying
what exactly this challenge entails.
A Better Notation for the Cyclic Groups
We end this chapter by describing a

convenient and simple new notation system for the
members of a cyclic group. The 5th cyclic group
C5 = {I, R72, R144, R216, R288} contains the
five rotations of the oriented star pictured on
the left. Keeping track of all of these angles is
cumbersome. Instead, it is common conven-
tion to name the members of C5 like this:
C5 = {0, 1, 2, 3, 4}. Think of “3” as representing “R216” which is a
rotation of the star by 3 counterclockwise “turns.” We are using
the word “turn” here to mean the smallest possible rotation angle,
72°, which moves each star point to its counterclockwise neighbor.
Thus, we are counting turns instead of angles. The Cayley table
for C5 looks particularly simple when translated into this concise
new notational system:
* I R72 R144 R216 R288 + 0 1 2 3 4

I I R72 R144 R216 R288 0 0 1 2 3 4
R72 R72 R144 R216 R288 I 1 1 2 3 4 0
R144 R144 R216 R288 I R72 2 2 3 4 0 1
R216 R216 R288 I R72 R144 3 3 4 0 1 2
R288 R288 I R72 R144 R216 4 4 0 1 2 3
A Better Notation for the Cyclic Groups 47
It is also common convention to use “+” to denote the

algebraic operation in a cyclic group. The yellow highlighted cells
above represent the equation that we previously wrote as “R216*
R288 = R144” and which we now write as “3 + 4 = 2 (in C5)”.
This notational change encourages you to think of the

algebraic operation in C5 as encoding a kind of “addition with
wrap-around.” As you count 4 past 3 on your fingers, remember
that when you reach 5, you must wrap back around to zero, so
your count goes: 4, 0, 1, 2.
In our future study of Cn, we will always use this

abbreviated notation system:
Cn = {0,1,2,…,n–1},
and we will always write “+” instead of “*.” For example, we will
write:
5 + 7 = 2 (in C10)
to indicate that, when you rotate a 10-gon by 7 turns and then by

5 more turns, you have done 2 turns more than going all of the
way around, so the result is the same as rotating just two turns
(here the word “turn” means a rotation by 360/10 = 36°).
Alternatively, just add on your fingers, wrapping back to zero
when you reach ten, so your count of 7 past 5 goes “6, 7, 8, 9, 0,
1, 2.” Alternatively, just add 5 + 7 = 12, and then subtract 10 to
get 2. It is all the same thing.
Here is another illuminating example:
9 + 7 = 4 (in C12)
In C12, addition-with-wrap-around works like a clock: 9 h after 7

o’clock, it will be 4 o’clock. However, this analogy is imperfect
because, when computing in C12, only the numbers between 0
and 11 are used. The analogy between C12 and clock-arithmetic
48 3
3. Iso
omo
orphism
on
nly b
beccom
mess a
airtig
ghtt if wee all ag
gre
ee tto hhen
nce eforrth sayy “0
0” insttea
ad
of “12
2” w
wheen rep
porrtingg thhe tim
me oof d
dayy, sso ttha
at m
midn night and dn
noo on
are ed “ze
e calle ero o’cclocck.”” We
W hop pe tha not mind..
at yyou wiill n
T
Thiss n
new w n nottatio
onaal sysste
em fo or Cn iss n not exxacctlyy aan
iso
omo orp
phissm be etw
ween ttwo ups be
o grrou ecaausse tthe ere wa as onlly e eve
er
onne g oup in the sstorry. Ne
gro eveerth
hele well in
esss, itt fitts w nto the
e ccha apte
er
abbout issom morrph
hismm. Wh henn a pair off grrou ups is isom morpphic, it iis
oftten be est to
o th
hinkk o
of th hem
m a as a sin
ngle e g grou up de ed in tw
escrribe wo
diffferrentt no
ota
ation
n ssysttem
ms.
Exxerrcissess
(1)) Fiill in
neeach
hbblan
nk:
5+7=_
___
_ (iin C9), 6 + _
___
_ = 2 ((in C10), 80 + 35 = _
___
_ (in C1000).
(2)) Consstru
uct a ccom
mplete
eCCayyleyy ta
able
e fo
or C7.
(3)) Draw
w sseve era
al ddiffe
eren
nt o
objectts w
who e syymme
ose etryy grrou
ups are
eaall
iso
omo orp
phicc to
o D5 .
(4)) Draw
w sseve era
al ddiffe
eren
nt o
objectts w
who
ose
e syymme
etryy grrou
ups are
eaall
iso
omo orp
phicc to
o C6 .
(5)) Coulld e
enla
arg
ging
g or sh
hrin
nkin
ng an ob
bjecct e
eve
er e
esse
enttiallly ccha
ang
ge
its syymm
mettry gro
oupp?
(6)) S
Som me copy ma achhine es lett yo
ou se eparatelyy coonttrol the vverrticaal
nd horizo
an onttal sccalinng facto ors when n yyou
u e enla ge or sh
arg nk a
hrin
piccturre. Co
ouldd a ve nly shrrinkking
erticcal--on geeverr essse
entiiallyy cchangeeaan
ob
bjecct’s syymmmettry gro oupp?
Exercises 49
(7) Take the 26 capital English letters, and sort them into piles
according to whether their symmetry groups are isomorphic.
Within each pile, sub-sort them according to whether the letters
are rigidly equivalent.
(8) Explain why the symmetry group of the Z border pattern is

NOT isomorphic to the symmetry group of the G border pattern.
(9) Verify that the following one-to-one matching is NOT an

isomorphism between the red and green squares in this chapter:
I R90 R180 R270 H Dc V D

Ľ Ľ Ľ Ľ Ľ Ľ Ľ Ľ
I F1 R180 R270 R90 F2 F3 F4
Is a random haphazard matching of the red and green symbols

likely to be an isomorphism?
(ٍ10) Make Cayley tables for D2 and C4 and explain why these
groups are NOT isomorphic.
(11) In the following list of groups, identify a pair which is

isomorphic, and explain why there are no other such pairs:
C1, D1, C2, D2, C3, D3, C4, D4, C5, D5, C6, D6,…
(12) In the following list of objects, identify a pair which has

isomorphic symmetry groups and explain why there are no other
such pairs. Is any pair rigidly equivalent?
50 3. Isomorphism
(ٍ13) Draw two border patterns which have isomorphic

symmetry groups but yet can NOT be enlarged or shrunk to
make them rigidly equivalent.
(14) Must any two asymmetric objects be rigidly equivalent?

ͶǤ
Historically, the concept of
symmetry evolved slowly from a
vague idea to a precise notion as
scientists and mathematicians sought
to study the symmetry of their world
using ever more precise language and
methods. The classification theorems
in this chapter represent some of the
pinnacles of this historic journey.
These theorems provide complete
classifications of the possible ways in
which each kind of planar object we
have studied (bounded objects, border
patterns, and wallpaper patterns) can
Leonardo Da Vinci’s self-portrait
be symmetric!
Bounded Objects
The description of all possible symmetry groups of

bounded objects is usually attributed to Leonardo Da Vinci
(1452–1519).
DA VINCI’S THEOREM: The symmetry group of any bounded

object in the plane is either infinite or is isomorphic to a dihedral
or cyclic group.
Thus, if the object’s symmetry group is finite, then Da Vinci’s

Theorem guarantees that it is isomorphic to the symmetry group
of one of these objects pictured below:

52 4. The Classification Theorems
Representatives of all the ways in which a bounded object can be symmetric
Actually, a stronger “rigid” version of this theorem is true:

Any bounded object in the plane with a finite symmetry group is
rigidly equivalent to one of the objects pictured above! Of course,
you must imagine that the list of pictures above does not stop at
n = 7, but goes on indefinitely. This rigid version is the one that
Leonardo Da Vinci understood, since he predated the develop-
ment of group theory.
Notice that Da Vinci’s Theorem tells you nothing about a

bounded object whose symmetry group is infinite. You might
guess that such an object must be rigidly equivalent to a circle,
but this is not quite true. There are some nuances involved which
we will not be equipped to discuss until a later chapter.
PROOF OF RIGID VERSION OF DA VINCI’S THEOREM:

Imagine that you have a bounded object with a finite symmetry
group. The Center Point Theorem tells us that all of its rotations
have the same center. In fact, we claim that they must all be
multiples of the smallest-angle rotation. Let us use an example to
think about why. If the smallest rotation symmetry was R10, then
R20, R30, R40,…,R350 would also be symmetries. Something else,
like R37, could not also be a symmetry, because that would make
R37*(R30)–1 = R7 be a rotation symmetry smaller than R10. Do you
see how this example leads you to the general idea needed to
justify the claim?
Bounded Objects 53
Say your object has n rotation symmetries. Since these

rotations are all multiples of the smallest, they have the same
angles as the rotations of a regular n-gon. If your object has no
flips, then it is rigidly equivalent to an oriented regular n-gon via
any rigid motion that matches up their center points.
If your object has any flips, then choose one flip and call it
F. Composing F with each of the n rotations yields a list of n
different flips. By the All-Or-Half Theorem, your object has no
flips other than these n.
Next draw a regular n-gon, choose one of its flips, and call
it Fcc. All of the n-gon’s flips are obtained by composing its
n rotations with Fcc. Now move your object on top of the regular
n-gon which you drew,
using a rigid motion,
M, that matches their
center points and
aligns the F-flip line
with the Fcc-flip line, as
illustrated on the right.
After this repositioning,
your object will have
exactly the same symmetries as the regular n-gon, namely, the
same rotations about their now-common center point, and the
compositions of these rotations with their now-common chosen
flip. Thus, your object is rigidly equivalent to the regular n-gon. In
particular, your object’s symmetry group is isomorphic to that of
the regular n-gon ...……………………………………………………Ƒ
COROLLARY: If a bounded object has exactly n rotations and

zero flips, then its symmetry group is isomorphic to Cn . If a
bounded object has exactly n rotations and n flips, then its
symmetry group is isomorphic to Dn .
It is interesting that there are no other possibilities. Before

reading Da Vinci’s Theorem, you might have imagined that it was
possible to sketch a picture whose n rotations and n flips fit
together into a Cayley table that is essentially different from Dn.
Border Patterns
Photo by Horia Varlan from Flickr.com

Images by Becky F from Flickr.com
The border patterns which you encounter in art and

architecture exhibit a seemingly infinite variety of artistic motifs.
However, from a mathematical point of view, there are only seven
different ways in which a border pattern can be symmetric!
THE CLASSIFICATION OF BORDER PATTERNS: Any border

pattern is rigidly equivalent to a rescaling of one of the seven
model border patterns illustrated below (provided it has a
smallest non-identity translation).
The requirement that the border pattern has a smallest

non-identity translation excludes patterns like an infinite
horizontal line, which differs from the seven model patterns in
that it can be translated any distance right or left, no matter how
small. Some texts include the smallest translation requirement
into their definition of a border pattern.
Border Patterns 55
A “rescaling” (enlarging or
shrinking) is just needed to ensure
that the border pattern has the
same smallest translation length
as the model pattern to which it is
being compared.
Rescaling a border pattern

does not essentially change its
symmetry group. Thus, the group
version of the classification theorem
says this: The symmetry group of
any border pattern is isomorphic to TheAndrewKepert
7 model border patterns: Image by
on Wikipedia.org
the symmetry group of one of the
seven model border patterns (provided it has a smallest
translation). However, among the seven model patterns, some of
the symmetry groups are isomorphic to each other. After
removing the redundant ones, the list of seven shrinks to a list of
four. Not all isomorphisms are explained by rigid motions.
If you have a border pattern, how do you know which of

the seven model patterns yours is rigidly equivalent to? All you
must do is fill out the following “identification card”:
Q1 – Does it have any horizontal reflection symmetry?

Q2 – Does it have any vertical reflection symmetry?
Q3 – Does it have any 180° rotation symmetry?
Q4 – Does it have any glide reflection symmetry?
Border pattern identification card
Your pattern will be rigidly equivalent to the model pattern that

has the same yes/no responses to all four questions on this
identification card. The term “horizontal” here really means the
direction of your border pattern’s translations, and the term
“vertical” really means the direction perpendicular to your border

pattern’s translations.
In fact, the steps involved in proving the classification
theorem are: (1) Prove that a border pattern cannot have any types
of symmetries other than translations and those types mentioned
on the identification card, (2) prove that two border patterns with
the same identification cards must be rigidly equivalent to rescaling
of each other, and (3) decide which combinations of yes/no
responses are identification cards of actual border patterns.
Wallpaper Patterns
Like border patterns, wallpaper patterns exhibit endless

artistic variety, but yet we will soon learn that there are only 17
ways in which a wallpaper pattern can be symmetric.
Qubbah Ba'adiyim in Marrakesh WoodCut QBert Block Texture

photo by amerune, Flickr.com by Patrick Hoesly, Flickr.com
Just like border patterns, wallpaper patterns are classified

according to their answers to the questions on an identification
card. The first question is this: What is the maximal order of a
rotation symmetry of the pattern? This means the number of
times that the rotation must be repeatedly performed to return to
the starting position. For example a 90° rotation has order
360/90 = 4, while a 60° rotation has order 360/60 = 6. Do you see
Wallpaper Patterns 57
the pattern? It is quite surprising that the only possible orders for
the rotation symmetries of wallpaper patterns are 1, 2, 3, 4 and 6,
as we will soon see.
Another identification card question asks whether there are
any glide reflection symmetries that are indecomposable. This
means that the translation and reflection out of which the glide
reflection is built are not themselves individually symmetries of
the pattern. Below are the identification card and the flow chart by
which the 17 wallpaper patterns are classified:
O – What is the maximum Order of a rotation symmetry?

R – Does it have any Reflection symmetries?
G – Does it have an indecomposable Glide reflection symmetries?
ON – Does it have any rotations centered ON reflection lines?
OFF – Does it have any rotations centered OFF reflection lines?
The 17 model wallpaper patterns: diagram by Brian Sanderson,http://www.warwick.ac.uk/~maaac/

If you have a wallpaper pattern, then your pattern will be

“symmetric in the same way” as one of the 17 model patterns;
namely, the one whose identification card places it in the same
flow chart position as your pattern. To state a precise theorem,
we must decide what “symmetric in the same way” means. One
possibility is to focus on symmetry groups:
THE CLASSIFICATION OF WALLPAPER PATTERNS: The

symmetry group of any wallpaper pattern is isomorphic to the
symmetry group of one of the 17 model patterns (provided it has
a smallest non-identity translation).
However, this theorem is not quite optimal because,

among the 17 model patterns, certain pairs have isomorphic
symmetry groups, so a list of fewer than 17 model patterns would
have sufficed. There is a rigid version of this theorem, which
genuinely requires all 17 model patterns, but it is a bit too
technical to fully describe in this book. Here is the rough idea. To
make an arbitrary wallpaper pattern become rigidly equivalent to
one of the 17 model patterns, you must alter it by something
called a linear transformation, which is slightly more general than
a rescaling because it includes things like vertical-only-
shrinkings/enlargings and shears (which are illustrated in
Exercise 9). We are not quite equipped to give a precise
definition of a linear transformation. That is why we only stated
the group version of the classification theorem. This theorem is
difficult to prove; a complete proof did not appear until 1891.
The symmetry groups of the 17 model patterns are often

called “wallpaper groups.” Chemists call them “plane crystallo-
graphic groups” because they represent the possible con-
figurations into which two dimensional crystal structures can
form. M.C. Escher incorporated many of these patterns into his
paintings. They also occur throughout nature, for example in
honeycombs.
Summ
mary 5
59
Su
ummmaryy
In
n thiss ch hap pter, wwe diid som meething reemaarkaablle. We d dess-
cribedd a
all possible w s in w
e way which (1) a b bouunded ob bjecct, (2) a
bo
ordeer pattterrn, and ((3) a wa allpapeer patterrn can be syymme etricc.
Foor e
each oof th
hesse tthre ee categ w ffound a list of mo
goriies, we odeels. Fo
or
ounded
bo do objeectss, oourr modeels we ere
e th
he o
orie
enteed annd n
non n-oriennte
ed
reg
gular po olyg
gonns, plu us a ggnoome a andd a happy facce. Fo or bo orde
er
atternss, o
pa our lisst cconttainned
deexacctlyy se en mo
eve odeels. Fo
or wwallpa ape
er
pa
atternss, o
our listt co
onta ained exxactly 17 mo ode
els.
A
Anyy obbjecct in tthe caateg gorry iss “ssym
mmmetrric in tthee saame w wayy”
as on ne of tthee mod
m delss. W Wha at e
exaactlly ddoees iit m
meaan for tw wo o objectts
to bee “ssymmme etric in the sa ame ewwayy?” It d dep
pen ndss onn th
he catteg goryy.
Foor bbouundded ob bjeectss, the be est mean ning iis ttha he ob
at th bjeccts arre
rig
gidlyy e
equuivaalennt. Fo or b
bord derr paatte
ernns, thee b best mmea aninng is tha at
onne ppattterrn is rrigid
dlyy eqquivvalentt too a re esccalinng of the e o
othe er. Foor
wa paper pattterrnss, th
allp he best me ean ning hat on
g iss th ne oobjectt is riggidlly
eqquivvale
ent too thhe rressult off pe erfo
ormming a linea ar tra
ansfformma ation tto
thee o
otheer, buut ssincce llineearr tra
anssforrma atio ult tto deffine
onss arre diffficu e,
wee ssetttled
d fo or the e nnexxt-b
best m mea aning: th hatt th
he pa atte
ernss hhavve
iso
omo orp
phicc syymme etryy grroups..
T
Thiss ccha apteer rep pre
eseentss yyet an notthe er e exhhibiition of th he
importan nceeo of p
preccisee la
anggua age e inn math
m hem matticss. B Befforee yoou caan
an
nswwer a quesstio on, yo ou mus
m st u und derrsta
and d e exa
actlyy w what th he
qu
uesttion
nm mea anss. S
Sincce tthe
e firrst ccha
apte er oof tthiss bo
ookk, wwe have ha ad
a g
gro
owing de esirred to unndeersttan nd tthe e ways
w s in whicch ob
n w bjeccts ca
an
be
e syymm meetricc. O Oneeoof the mo ostt difficcult chhalleng gess tu
urnned ou ut tto
be
e the cclarrificcatioon of pre
ecisselyy wwha at thhis me ean ns.
Exercises
(1) For each of the seven model border patterns, answer the four
identification card questions. If Q1 and Q2 both answer yes for a
pattern, why must Q3 and Q4 also answer yes?
(2) For each capital English letter, decide which of the 7 model
border patterns is formed by repeating it indefinitely along a line.
(ٍ3) Of the 7 model border patterns, find two which have

isomorphic symmetry groups.
(ٍ4) Which of the 7 model border patterns have commutative

symmetry groups?
(5) How many of the 7 model border patterns are oriented?
(6) How many of the 17 model wallpaper patterns are oriented?
(7) Of the 17 model wallpaper patterns, identify at least one

which has a commutative symmetry group and at least one which
does not. Explain why.
(ٍ8) Prove that a border pattern could not have any reflection
symmetries other than horizontal and vertical.
(9) Brick patterns 2 and 3 below were obtained from 1 by

applying linear transformations; namely, a vertical-only-shrinking
for 2 and a “shear” for 3. Imagine that all three patterns are
extended indefinitely to form wallpaper patterns. Decide which of
the three are “symmetric in the same way” as each other,
according to several meanings for this term: (A) they have
isomorphic symmetry groups, (B) they are rigidly equivalent,
(C) they are rigidly equivalent to rescaling of each other, (D) they
have identical identification cards.
Exercises 61
(10) Classify each of these borders patterns as type 1-7.
Image by User:Cyp from wikipedia.org
(11) Perform a web image search for “M.C. Escher symmetry.”

Many of Escher’s paintings are wallpaper patterns (if you imagine
them infinitely extended). How many of the 17 model wallpaper
patterns can you find represented in his paintings?
(12) Identify the model wallpaper pattern that matches with the
paintings Seahorses and Eels and Three Fishes by Robert
Fathauer. How would the answer change if the fish all had the
same color?
(13) Identify the model wallpaper patterns that match with

Qubbah Ba'adiyim and WoodCut QBert Block Texture, illustrated
in the chapter.
(14) Identify the model border patterns that match with the border
patterns by Beck F and Horia Varlan illustrated in the chapter.
ͷǤ

In this chapter, we learn how to find small groups inside of
large groups, and then how to build large groups out of small
groups. The point is to help you better understand symmetry
groups. If you can recognize an object’s symmetry group as
having been built out of smaller groups, then this realization
might help you to much more clearly understand its underlying
algebraic structure.
Subgroups
First, we learn how to find small groups inside of large

groups. For example, inside D4 = {I, R90, R180, R270, H, V, D, Dc},
let us separately consider the rotations R = {I, R90, R180, R270} and
the flips F = {H, V, D, Dc}. Let us build red-only and green-only
tables, like this:
* I R90 R180 R270 * H V D D’

I I R90 R180 R270 H I R180 R90 R270
R90 R90 R180 R270 I V R180 I R270 R90
R180 R180 R270 I R90 D R270 R90 I R180
R270 R270 I R90 R180 D’ R90 R270 R180 I
The red table is the Cayley table for R = {I, R90, R180, R270}, which
is a self-contained group that happens to be isomorphic to C4.
The green table does not look like a Cayley table at all, due to the
fact that F = {H, V, D, Dc} is not a self-contained group. After
studying the green table, you can identify several reasons why F
is not a group. First, there is no identity. Second, the cells of the
green table are filled with symbols that are not in F. If F were

64 5. Subgroups and Product Groups
really a group, then each cell of its Cayley table would be filled
with something in F – there could be no foreign symbols.
To summarize: the rotations, R, form a special type of

collection, called a “subgroup,” because it is a self-contained
group in its own right. The special properties of R, by virtue of
which we know it is a self-contained group are: (1) the identity is
a rotation, (2) the inverse of every rotation is a rotation, and
(3) the composition of every pair of rotations is a rotation (so that
the red Cayley table has no foreign symbols). Look back at the
definition of “group” and think about why these three properties
are exactly what is needed to verify that R is a group. The flips, F,
do NOT form a self-contained group because (1) the identity is
not a flip and (2) the composition of a pair of flips is not a flip.
DEFINITION: Suppose that G is a group. A collection, H, of G’s

members is called a subgroup if it forms a self-contained group
on its own, which means that it satisfies all three of these
requirements:
(Identity) H must include the identity of G.
(Products) If A and B are in H, then A • B must be in H.
(Inverses) The inverse of anything in H must be in H.
We have seen that the rotations form a subgroup of D4, but the
flips do not. Here are some more examples.
EXAMPLE: The collection K = {I, R180, H, V} is a subgroup of D4.

Why? First, it contains the identity, I. Second, each thing in K is
its own inverse. Third, the product
* I H V R180
(composition) of any two things in
I I H V R180
K always lies in K, because every
H H I R180 V
cell of the table at the right is
filled in with one of the four things V V R180 I H
in K (no foreign symbols). In fact, R180 R180 V H I
Subgroups 65
K looks like D2 = the symmetry group of the

2-gon. Think of K here as the symmetries of the
square which are also symmetries of the 2-gon
drawn as a horizontal middle stripe.
EXAMPLE: The symmetry group of the W border pattern,
…WWWWWWWWWWWWWWWWW…
includes translations and also vertical flips. The translations alone

form a subgroup. We will denote these translations like this: T =
{…,T–3,T–2,T–1,T0,T1,T2,T3,…}. Why is it a subgroup? First, it
includes the identity, T0. Second, the inverse of every translation
is a translation. For example, the inverse of T7 is T–7. Third, the
composition of any pair of translations is a translation. For
example, the composition of T7 and T8 is T15.
NON-EXAMPLE: In the symmetry group of the W border pattern,

the flips alone do NOT form a subgroup. The problems are:
(1) the identity is not a flip, and (2) the composition of a pair of
flips is not a flip – it is a translation.
EXAMPLE: In the additive group of integers, Z = {…,–3, –2, –1,

0, 1, 2, 3, …}, the evens, E = {…, –6, –4, –2, 0, 2, 4, 6, …} form a
subgroup. Why is E a subgroup of Z? First, the identity, 0, is
even. Second, the inverse of every even number is even. For
example, the inverse of 6 is –6. Third, the sum of any two even
numbers is even (remember that adding is the algebraic
operation in this group). Think about why the odd numbers do
NOT form a subgroup of Z.
EXAMPLE: In the cyclic group C10 = {0, 1, 2, 3, 4, 5, 6, 7, 8, 9},

the even numbers, E = {0, 2, 4, 6, 8} form a subgroup. Why is E a
subgroup of C10?
First, the identity, 0, is even. Second, the + 0 2 4 6 8

inverse of every even number is even: the 0 0 2 4 6 8
inverse of 2 is 8, the inverse of 4 is 6, and 2 2 4 6 8 0
the inverse of 0 is 0. Third, the “sum with 4 4 6 8 0 2
wrap-around” of any two even numbers is 6 6 8 0 2 4
even, because the Cayley table for E (shown 8 8 0 2 4 6
here) is entirely populated by even numbers.
Can you explain why E is isomorphic to C5?
NON-EXAMPLE: In the cyclic group C9 = {0, 1, 2, 3, 4, 5, 6, 7, 8},

the even numbers, E = { 0, 2, 4, 6, 8 } do NOT form a subgroup.
The first problem is that the inverse of an even number is NOT
always even in this group. For example, the inverse of 2 is 7 and
the inverse of 4 is 5. Second, the sum-with-wrap-around of two
even numbers is not always even, like 6 + 8 = 5 (in C9).
Generating Subgroups
Suppose you have a group, G, and you wish to find a subgroup

of it. You could list all of G’s members, and then circle the ones
you are putting into your subgroup. But how do you do the
circling so that you end up with a valid subgroup? Here is one
way. First, circle the identity, I, because it is in every subgroup.
Next choose some other member, A, to circle. Now you are
forced to circle the inverse of A. You are also forced to circle the
product of A with itself any number of times, and the product of
the inverse of A with itself any number of times. After you have
done all of this, you will have a valid subgroup. It contains A plus
everything else that is forced to be there by the three
requirements of being a subgroup. We call this subgroup <A>
because it is “generated” by A.
Generating Subgroups 67
DEFINITION: If G is a group, and A is a member of G, then,

<A> = {…, A–1•A–1•A–1, A–1•A–1, A–1, I, A, A•A, A•A•A,…}
is called the subgroup of G generated by A. It contains A and A–1
combined with themselves any number of times.
EXAMPLE: In Z = {…, –3, –2, –1, 0, 1, 2, 3, …}, let us take A = 3.

The inverse of 3 is –3. The symbol “•” must be replaced with “+,”
which is the algebraic operation in Z. For example, the expres-
sion “A•A•A” becomes “3+3+3 = 9”. Similarly, “A–1•A–1•A–1•A–1”
becomes “–3+–3+–3+–3 = –12” and so on. We conclude that
<3> = { …, –12, –9, –6, –3, 0, 3, 6, 9, 12, …} = the multiples of 3.
EXAMPLE: In D4, let us take A = R90. The inverse of A is R270.

The symbol “•” must be replaced with “*” because composition is
the algebraic operation in D4. We have:
A = R90, A•A = R180, A•A•A = R270, A•A•A•A = I, …
(we can stop here because the pattern repeats).
A–1 = R270, A–1•A–1 = R180, A–1•A–1•A–1 = R90, A–1•A–1•A–1•A–1 = I,...
(we can stop here because the pattern repeats). In summary,
<A> = {I, R90, R180, R270}.
This makes sense – if you perform a 90° rotation forward or

backwards as many times as you like, you generate all of the
rotations of the square.
In the last example, when we composed A with itself

repeatedly, we eventually arrived at the identity, after which the
pattern began repeating. Furthermore, the things obtained by
composing A–1 with itself repeatedly were redundant to the things
obtained by composing A with itself repeatedly. In fact, this is

exactly what always happens in any finite group.
THEOREM: If G is a finite group, and A is a member of G, then

<A> = { I, A, A•A, A•A•A, A•A•A•A, …}
(this list starts repeating as soon as one of these expressions
equals I, and not before).
This theorem tells us how to quickly find the members of <A>.

Just write the identity, then write A, then A•A, and so on until you
arrive at the identity again, at which point you are done. For
example, in C10 = {0, 1, 2, 3, 4, 5, 6, 7, 8, 9},
<2> = {0, 2, 4, 6, 8} (the evens)
<3> = {0, 3, 6, 9, 2, 5, 8, 1, 4, 7} (all of C10)
<4> = {0, 4, 8, 2, 6} (the evens – the same as <2>)
<5> = {0, 5}.
There is a nice way to visualize these subgroups of C10.

Imagine 10 friends (numbered 0–9) standing in a circle. They
pass a ball amongst themselves. Friend number 0 starts with the
ball and passes it to her right. The subgroup <2> is the collection
of friends who touch the ball if everyone passes 2 to their right.
The subgroup <3> is the collection of friends who touch the ball if
everyone passes 3 to their right and so on.
Product Groups 69
Try drawing your own such circle diagram to illustrate <4> and
<5> in C10. Perhaps these diagrams help you see why the Cs are
called cyclic groups.
We previously defined the “order” of a rotation symmetry,
so that R360/n has order n. More generally, we now define “order”
of any member of any group:
DEFINITION: If A is a member of a finite group, then the order of

A is the size of the subgroup <A>.
The “size” of a group or subgroup just means its number of

members. Thus, the order of A is the smallest number of copies
of A which must be combined to get the identity. For example, the
order of the member 2 of the group C10 equals 5 because
2+2+2+2+2 = 0, and no smaller list of 2s sum to the identity. The
order of the member R90 of the group D4 equals 4 because a 90°
rotation must be done 4 times in succession to return a square to
its starting position.
Product Groups
Next, we consider an important way of putting together a

pair of groups, G1 and G2, to build a single new group, which is
denoted G1 × G2 and is called the product of G1 and G2. The
members of this new group are all of the possible ways of pairing
together a member of the first group with a member of the
second group (wrapped in parentheses and separated by a
comma). For example, if G1 = {A, B, C} and G2 = {1, 2, 3, 4},
then:
G1 × G2 = {(A,1), (A,2), (A,3), (A,4),

(B,1), (B,2), (B,3), (B,4),
(C,1), (C,2), (C,3), (C,4)}.
Thus, G1 × G2 catalogs all ways to pair a red G1 member with a

green G2 member. The coloring is optional, but is helpful for
understanding the idea here. If the red letters are men dancers
and the green numbers are women dancers, then G1 × G2
catalogs all ways in which a mixed-gender dance couple could be
formed. In the last example, G1 had 3 members, G2 had 4
members, and G1 × G2 had 3 × 4=12 members. More generally,
the size of the product of two finite groups equals the product of
their sizes.
We will denote the algebraic operation in G1 × G2 as “•.”

How do you think it works? You simply do the red G1 part and the
green G2 part separately. For example, to find (5,H)•(7,V) in the
group Z × D4, the answer is (5+7,H*V) = (12,R180).
For more practice, let us construct the Cayley table for the
product of the groups
C3 = {0, 1, 2} and C2 = {0, 1}.
The product of these two groups has the following six members:
C3 × C2 = {(0,0), (0,1), (1,0), (1,1), (2,0), (2,1)}.
The Cayley table looks like this:
• (0,0) (0,1) (1,0) (1,1) (2,0) (2,1)

(0,0) (0,0) (0,1) (1,0) (1,1) (2,0) (2,1)
(0,1) (0,1) (0,0) (1,1) (1,0) (2,1) (2,0)
(1,0) (1,0) (1,1) (2,0) (2,1) (0,0) (0,1)
(1,1) (1,1) (1,0) (2,1) (2,0) (0,1) (0,0)
(2,0) (2,0) (2,1) (0,0) (0,1) (1,0) (1,1)
(2,1) (2,1) (2,0) (0,1) (0,0) (1,1) (1,0)
The yellow cell in the above table represents (1,1) • (1,0) = (2,1).
Just think of this equation as a red equation (1 + 1 = 2 in C3)
Product Groups 71
packaged together with a green equation (1 + 0 = 1 in C2).

Product groups may be large, but they are simple. Since their
algebraic operation involves doing two unrelated things, product
groups are no more complicated than the smaller groups out of
which they are built.
Why are product groups important? For starters, when we

study the symmetry group of a new shape, we usually ask
whether it is isomorphic to any familiar group. Now that the
product of two familiar groups is familiar to you, you have a larger
library of familiar groups against which to compare a new
symmetry group. Also, if you discover that a symmetry group is
isomorphic to a product group, then this discovery often helps
you better understand its underlying algebraic structure.
For example, we will explain why D2 = {I, H, V, R180} is

isomorphic to C2 × C2 = {(0,0), (1,0), (0,1), (1,1)}. Their Cayley
tables look like this:
I H V R180 (0,0) (1,0) (0,1) (1,1)

I I H V R180 (0,0) (0,0) (1,0) (0,1) (1,1)
H H I R180 V (1,0) (1,0) (0,0) (1,1) (0,1)
V V R180 I H (0,1) (0,1) (1,1) (0,0) (1,0)
R180 R180 V H I (1,1) (1,1) (0,1) (1,0) (0,0)
The isomorphism is the following one-to-one correspondence:
I ļ (0,0), H ļ (1,0), V ļ (0,1), R180 ļ (1,1).
It is simple to check that this dictionary translates each of

the above Cayley tables into the other and is therefore an
isomorphism.
Even better, there is a nice visual

way to see this isomorphism. Think of
D2 as the symmetry group of this red and
green cross. Notice that H exchanges the ends of the red
rectangle, V exchanges the ends of the green rectangle, and R180

exchanges both. In C2 × C2, think of the red number as recording
whether the ends of the red rectangle were exchanged (0 = no,
1 = yes), and think of the green number as recording whether the
ends of the green rectangle were exchanged. Thus, R180 ļ (1,1)
because the 180° rotation exchanges the ends of both
rectangles. The group D2 appears simpler once you notice that
computing in this group really just involves simultaneously
keeping track of the answers to two yes/no questions. Doing two
unrelated things simultaneously is what product groups are all
about.
Here is a second visual illustration of the utility of product

groups. The symmetry group of the B border pattern includes
translations and the reflection over the horizontal center line.
…BBBBBBBBBBBBBBBBB…
My favorite symmetry of this border pattern is (T–5,YES), which is

my shorthand for “translate 5 letters to the left and flip over the
horizontal center line.” My second favorite symmetry is (T8,NO),
which is my shorthand for “translate 8 letters to the right and do
not flip over the horizontal center line.” Composing symmetries
is easy. For example, (T–5,YES) • (T8,NO) = (T3,YES). Similarly,
(T6,YES) • (T4,YES) = (T10,NO). Two unrelated things are happen-
ing simultaneously here: the T-subscripts NO YES
are added and the YES/NOs are combined NO NO YES
according to the table at the right. YES YES NO
Since these two things are unrelated, we suspect that the

symmetry group of the B border pattern is isomorphic to a
product group. In fact, since the above table looks just like the
Cayley table for C2 = {0,1} (with 0 = NO and 1 = YES), the
symmetry group of the B-frieze is isomorphic to Z × C2. The iso-
morphism matches the symmetry (Tn,NO) with (n,0) and matches
Exe
ercisses 7
73
(Tn,Y
YESS) w
with
h (n
n,1)). D
Do you ssee
e wh
hy this iss an issom
morph
hismm? Fo or
exxam
mple wattch this iisomo
e, w orphhism ttran
nslate
e a tru
ue eequ
uattion nto a
n in
tru
ue e uatiion:
equ
(T
T8,Y
YES) * (T
T7,Y
YEES) = ((T15,N
NO))
Ļ Ļ Ļ
8,1) • (7,,1) = ((15
(8 5,0)
Doo you seee w whhy it w

will tra
anslate
e eeveery tru
ue eequuattion nto a true
n in
quation
eq n? Whatt abou ut an eq qua
atio
on w with
h tw
woo NOss? O Or with one
YEES annd one e N
NOO? In sum mm
marry, thee ssymmme group off th
etryy g he B
bo
ordeer pattterrn is rea
allyy a pprodducct g
grooup inn disg guisse, annd thiis
dissco helps sim
overy h mpllify yoour underrstaand
d th
he ssym
mm metrry ggroup..
Exxerrcissess
m y ssubgro
(1)) Fiind ass man oup
ps o
of D4 a
as you
u ccan.
(2)) Reca
R all tha
at a sqquaare caan bbe orieentted so
o th
hat only tthee rootationns
in D4 arre symmmmetriess oof tthe re
esulting ima agee (n not the fflipss). Fo
or
eaach suubg
grou of
up f D4 w ch
whic yo fou
ou und d, ccan
n yo fin
ou nd a w way tto
ornnamment/d
deccorrate
e th
he sq quare to make e tthatt ssubg
gro p beco
oup om
me
the
e syym
mmeetryy grrou of the ressultting
up o g im
magge??
(3)) In
n th
he gro
oupp C3 × C2, ffindd a su
ubggrouup wh hich
h iss issom
morphiic tto
C3 an nd a ssubg
gro
oup
p whicch iss issom
morrphic tto C2.
(4)) Findd th
he sub
bgrrou
up ggen ed by ea
nerrate ach
h mem
m mbe er o
of tthe
e cyycliic
grooupps C4, C5, C6, …,
… C11. Make
M e a co onjectture
eaabout the
e po osssiblle
ordderrs o
of m mberss off Cn.
mem
(ٍ5)) In
n a finite gro
oupp, tthe LA
AST
T th
hin
ng in the list <A>> = {A,, A • A
A,
–1
1
A•A•A A, …} iss a ways A . T
alw The
e nnexxt-to
o-la
astt th
hing
g is aalw
wayys
A– 1• A–1, and so on. Verify this is true in of all your lists in

problem (4). How does this help justify the claim in the chapter
that “the things obtained by composing A–1 with itself repeatedly
are redundant to the things obtained by composing A with itself
repeatedly”?
(6) Make a complete Cayley table for C3×C3.
(ٍ7) Explain why C12 is NOT isomorphic to C2 × C6, even though

these two groups both have order 12.
(8) In the group C7 × C10, list the members of <(0,2)> and

<(1,5)>.
(9) Explain why the proper symmetries of any object in the plane
form a subgroup of its symmetry group. This subgroup is called
the proper symmetry group of the object.
͸
͸Ǥ

E
Eacch syymm mettry of a squarre pe erm
mutees th he sq quaare’’s
verticces, aand
d a o perrmu
also es its edge
ute es. YYouu ccann teell whicch
symm mettry waas p
perfforme by obsserrvin
ed b ng hhow he verticces orr ed
w th dge
es
we
ere pe erm
muted. FFor this rea asoon and d ffor manyy mmorre to co omee,
pe
ermuta ons are a crrucial keyy to
atio o unde
ersstanndin
ng sym
mm mettry.
A “p perrmu
utation
n” meeanns a rea arra
ang
gemmennt. Fo
or exam
mple
e,
the
ere
e arre ssix perrmu ations of the
uta e le
etters A
A, B, and
dCC; n
nam
mely:
A
ABBC, AC
CB,, BA
ACC, B
BCA
A, C
CAB
B, C
CBBA
hinkk off A, B, and C as lettterrs o
Th on a m mag gne et b
boa nd think of a
ard,, an
pe
ermuta on as a word
atio d (nott n necesssarrily a re eal En nglish word
w d)
wh
hichh yoou can sspe
ell u
usin
ng a
all oof tthe
e lettterrs.
DEEFINITTIO
ON:: W
Wheen n2, thee ccolle
ecttion
n off all p
permuutattion
ns of n
ord
derred th
hing
gs is den
notted Pn aandd iss ca
alle
ed the e nth permuta atio
on
gro
oupp.
We
ewwill soo
on see th
hatt Pn iss a gro
oup
p, a
as itts n
nam
me sugge
estts.
U
Usuuallyy the “o ordere
ed th hings”” are e letters o orde
ereed
alp
pha
abeetica
allyy. F mple, P3 = {AB
Forr exxam BC, A ACBB, BBAAC, BCCA, CCAB B,
CBBA}} ha
as sixx mmemmbe ers. H
Howwmman ny p
perrmu ns does {A
utattion A, B
B, CC,
D, E, F}} ha
avee? In oth her woords, w e sizze of P6? T
what is tthe To speell a

76 6. Permutations
six-letter word on the magnet board, you have six choices for the
first letter, then five choices left for the second letter, then four
choices left for the next letter, and so on. The number of ways in
which you can make these choices in succession equals:
6 × 5 × 4 × 3 × 2 × 1 = 720. Therefore, the size of P6 equals 720.
The general rule is:
THEOREM: the size of Pn equals n!
The symbol “n!” means the product of all of the integers between
1 and n. It is pronounced “n factorial”. Here are the first few:
2! = 1 × 2 = 2
3! = 1 × 2 × 3 = 6
4! = 1 × 2 × 3 × 4 = 24
5! = 1 × 2 × 3 × 4 × 5 = 120
6! = 1 × 2 × 3 × 4 × 5 × 6 = 720
Factorials grow large very quickly.
Permutation Groups
Listing and counting permutations is not nearly enough. To

uncover the full power of permutations, we must learn the
algebraic operation which turns Pn into a group. This operation is
called “composition” and is denoted “*” (exactly as in symmetry
groups).
To compose permutations, the key is to regard each

permutation, not as a word, but as the action (exchanging/
moving/cycling of letters) that occurred to build that word on the
magnet board from the (ordered) starting position. For example,
in P3, the starting position of the magnet board is alphabetical,
ABC, so this word is the identity. Each other word was obtained
from this starting position by moving/exchanging/cycling some
letters. For example, CBA was obtained by exchanging the 1st
Permutation Groups 77
and 3rd letters, encoded by writing (13). ACB was obtained by

exchanging the 2nd and 3rd letters, encoded by writing (23).
Their composition, ACB * CBA, simply means the result of
performing these exchanges in succession: first perform CBA
(exchange the 1st and 3rd letters) and then perform ACB
(exchange the 2nd and 3rd letters), like this:
We learn that: ACB * CBA = CAB.
This equation might seem like a dreamlike intermixing of math

class with spelling class. The algebra of permutations forces us
to constantly alternate between two points of view: regarding a
permutation as a word and as an action. To summarize:
DEFINITION: If W1 and W2 are permutations (words), then their

composition, W1 * W2, is the permutation (word) obtained from
the starting position by first performing the action for W2 and then
performing the action for W1.
The shorthand symbols (13) and (23) were useful above,

so let us similarly encode the remaining words in P3. For
example, the word BCA is obtained by moving the 1st letter (A) to
the 3rd position, bumping the original 3rd letter (C) to the 2nd
position, and finally bumping the original 2nd letter (B) back to the
1st position. We’ll encode this as (132). The red circle
diagram on the right is a better code for this, but it uses
too much typesetting space. The other words in P3 are
coded as follows:
ABC ļ I BCA ļ (132) CAB ļ (123)

BAC ļ (12) ACB ļ (23) CBA ļ (13)
78 6. Permutations
THEOREM: Pn is a group.
This is not hard to prove. For example, every permutation

has an inverse because, no matter how I scramble the letters on
a magnet board, it is always possible for you to unscramble them
back to the starting position. Let us practice by constructing a
Cayley table for P3.
* ABC BCA CAB BAC ACB CBA

ABC
BCA
CAB
BAC
ACB CAB
CBA
A Cayley table for P3
The yellow cell represents the equation ACB * CBA = CAB, which
in our green “action notation” looks like (23) * (13) = (123). Fill in
the rest of the table. Along the way you will have to think about
both the words and the actions that the green shorthand symbols
represent. When you are done, your table should look sus-
piciously like the Cayley table for D3 from Chap. 2. In fact:
THEOREM: P3 is isomorphic to D3.
PROOF: The isomorphism from P3 to D3 is:
ABC ļ I BCA ļ R120 CAB ļ R240

BAC ļ F1 ACB ļ F2 CBA ļ F3.
This dictionary translates the Cayley table for P3 above into the
Cayley table for D3 in Chap. 2. There is an important visual way
to understand the isomorphism. Instead of writing three-letter
words left-to-right, let us write them clockwise around the vertices
of a triangular magnet board, starting at the top. For example, the
Permutation Groups 79
identity word, ABC, is pictured here. Our isomorphism simply

matches a symmetry of the triangle with the word that we see
after that symmetry is performed (always
reading words clockwise starting at the
top). That is, we match each symmetry of
the triangle with the permutation of its
vertices that the symmetry induces. The
permutation induced by a composition of
two symmetries equals the composition
of the individual permutations which
these two symmetries induce. That is why it is an iso-
morphism………………………………………………..…………….Ƒ
For n > 3, Pn is NOT isomorphic to Dn because Pn has far

more members than Dn. The issue here is that not every
permutation of the vertices of an n-gon can be achieved by a
symmetry. For example, no symmetry of a square could exchange
its top two vertices without also exchanging its bottom two
vertices.
The green action symbols for P3 are called cycle notation.

This is a notation system (not just in P3 but also in Pn) for
describing the action required to build each word from the starting
position on a magnet board. For example, in P6, let us translate
the word BADEFC into cycle notation. On a magnet board, this
word is obtained from the following action:
Its cycle notation is (12)(3654), which means: exchange the 1st

and the 2nd letters, and then cycle the remaining letters by
sending 3rd ĺ 6th ĺ 5th ĺ 4th ĺ 3rd. Notice how the closing “)”
80 6
6. Pe
ermu
utationss
rep
pre
esentss ann in
nsttrucctio
on tto lloo
op b
bacck tto the
e sttartting
g “(”. Byy the
waay, he
ere is sommee se elf-e
exp argon:: (1
planattoryy ja 12)((36
654 4) ccontain
ns a
cycle
e off le
ength 2, na ame ely (12 nd a ccyccle of len
2), an ngth 4 4, n
nam
melly
(36
6544).
Y
You u ca
an translate e an ny wo ord intto ccycle nnottatio
on byy follow win
ng
theese e step ps: First wrritee “((” ffollo
owe ed byy th ositiion off th
he po he firsst
hat mo
lettterr th ovees. Foor e exaamp ple
e, iff A movm es,, thhen n yo ou will w writte
“(1
1”. Ne ext wrrite thee p
possitio on tha at ffirstt le
etter m
movvess too, then n wwritte
thee posiition thhatt the le
ette er in thatt poosittionn mov
m ves to,, an nd so on n. A
As
soon ass a lettterr move
m es ba ack to the ep possitio
on o e firrst lettter yo
of tthe ou
consiidered d, w
writte “)”.. Iff anny rema aining lettterrs mo ove e, thhenn w writte
annoth her “(”” fo
ollo
oweed by the ssma est moviing po
alle osittionn wwhicch yo ou
haave no ot yeet cconnsid
dere ed,, fo
ollo
owe ed byy th he poosittion
n thatt le ette
er
mo ove es tto a
andd so
o on.
F
For ad
ddittion
nal praacttice
e with
w co omp possing
g sym
mme etries,, leet u
us
find tthe co
omp
possitio
on BA ADEEFC C * EF
FAB BCCD in P6. Foor tthiss, w
we usse
EFFABBCD Daas the
e sttartting
gppositio
on oof tthe magnett booarrd, an nd tthe
en
peerfo
orm theea
aboove--picctured on ffor BA
d acctio ADE EFC 12)(36
C = (1 654 4):
Wee le
earrn thatt in P6, B DEF
BAD FC * E
EFA
ABCD
D = FE
EBC
CDA. Tryy
pro
onooun ng tha
ncin at!
Evven
naand
dOOdd
dPPerrmuta
atio
onss
Otther th
han
n the id
den
ntityy, tthe
e sim
mpplesst p muttatiions a
perm are the
e “sswa
apss”
DE
EFINIT
TIO
ON:: A sw
wap
p means an exxchange of two
o le
ette
ers..
Even and Odd Permutations 81
For example, ABEDCF is a swap in P6. Its cycle notation is (35)

because it was obtained from the alphabetical starting position by
swapping (exchanging) the 3rd and 5th letters. The cycle notation
for any swap is a single cycle of length 2.
THEOREM: The swaps generate Pn. In other words, every

permutation in Pn can be expressed as a composition of swaps.
PROOF: To warm up, let us think about how to write the word
EADCFB in P6 as a composition of swaps. With our magnet
board in its starting position, ABCDEF, we must perform a
sequence of swaps (letter exchanges) to arrive at the ending
position EADCFB. We will work left-to-right, moving the E to the
first position, then the A to the second, and so on:
ABCDEF ĺ EBCDAF
EBCDAF ĺ EACDBF
EACDBF ĺ EADCBF
EADCBF ĺ EADCFB.
The colored letters are being swapped. We started with positions
1 & 5 colored, then 2 & 5, then 3 & 4, and finally 5 & 6. Thus:
EADCFB = (56) * (34) * (25) * (15).
We have successfully expressed EADBFC as a composition of

four swaps in P6. The same strategy will work to write any permu-
tation in Pn as a composition of swaps……...…………………….…Ƒ
In the EADCFB example above, what if we instead work

right-to-left? How many swaps are needed? What if we only use
adjacent swaps (which means exchanges of adjacent letters)?
What if we correctly position the vowels first and then the
consonants? Try several different strategies, and count the
number of swaps needed for each strategy. Some strategies
require four swaps. Others might require 6 or 8 or 10 swaps.
82 6. Permutations
A grossly inefficient strategy might require 76 swaps. Here is the

important point. Sit down first because it is the most important
thing you will ever learn. No matter what strategy you employ,
you will require an even number of swaps. This is because
EADCFB is an “even permutation.”
DEFINITION: A permutation that can be obtained by composing

an even number of swaps is called an even permutation. A
permutation that can be obtained by composing an odd number
of swaps is called an odd permutation.
The importance stems from the following theorem:
THEOREM: A permutation cannot be both even and odd.
In other words, if you obtain your favorite word in Pn using

an even number of swaps, then this word is even – anyone else
would also require an even number of swaps to obtain it. We will
not discuss the proof of this theorem, which is a bit difficult, but
please do spend some time investigating it. For example, verify
using several different strategies that CADEFB requires an odd
numbers of swaps and is, therefore, an odd permutation.
THEOREM: Exactly half of the permutations in Pn are even.

Furthermore, the even permutations form a subgroup of Pn.
PROOF: To show that the even permutations form a subgroup,

we must verify three things. First, the identity is an even
permutation because we can obtain it using 0 swaps, and 0 is an
even number. Second, the composition of two even permutations
is an even permutation. For example, if the first permutation uses
8 swaps and the second uses 12 swaps, then the composition
uses 8 + 12 = 20 swaps (the sum of two even numbers is even).
Finally, the inverse of an even permutation is even. For example,
the inverse of (56) * (34) * (25) * (15) is (15) * (25) * (34) * (56) – the
Even and Odd Permutations 83
same swaps done in the reverse order. If your friend performs a

sequence of swaps, you can return to the starting position by
undoing your friend’s swaps in reverse order. This verifies that
the even permutations form a subgroup of Pn. In the exercises,
you will be asked to prove that exactly half of them are even.…..…Ƒ
DEFINITION: The subgroup of all even permutations in Pn is

denoted An and is called the nth alternating group.
The size of An equals half the size of Pn. The sizes of the
first few permutation and alternating groups are:
Size of P2 = 2 Size of A2 = 1
Size of P5 = 120 Size of A5 = 60.
The groups P4, A4, and A5 will play starring roles in the next
chapter, so remember their sizes, and keep an eye out for them.
We end this chapter with a final useful observation:
THEOREM: A cycle of length m can be obtained by composing

m – 1 swaps. Thus, the cycle is even if m is odd, and viceversa.
For example, the word BCDEFA in P6 is expressed in cycle

notation as (165432). This is a cycle of length 6. Using a magnet
board, check that BCDEFA = (56) * (45) * (34) * (23) * (12), so it
is obtained from five swaps. These five swaps “bubble” the first
letter A to the end of the word.
This theorem is useful for quickly deciding whether a

permutation is even or odd. Using cycle notation is often faster
than using a magnet board. For example, we previously decided
that BADEFC = (12)(3654). The first cycle has length 2, so it is
84 6
6. Pe
ermu
utationss
oddd. Thhe ssecconnd cyccle haas len

ngth
h 4
4, sso it iis o
odd
d. B
BAADE
EFC
C iss,
the
ere
eforre, thee ccom
mpoosittion
n o
of a
an odd ccyccle wiith ann odd cyycle
e,
whhich
h iss evven
n.
Exxerrcissess
(1)) If you h
havve n
not yet do e so, ffill in tthe
one e Ca ey tab
ayle ble forr P3 in
n the
chaptter.
(2)) E
Exp
presss each off thhesse w
wordss in
n P6 in cyccle no
ota
ation, an
nd
de
ecidde wh eacch is even
hettherr e n or od dd: CCBAAFD DE, BBCCDAAFE
E,
FAABCCDE, ED DAF FCBB.
(3)) Fiind CB
BAFD
DE * BCD
DAF
FE and F
FAB
BCDE
E * ED
DAF
FCB
B in
n P 6.
(4)) F
Find he fivve-lette
d th er woord co
orre esp
ponndinng witth each off thhesse
cyclee no
ota
ation e pressions in P5, an
exp nd de de wh
ecid hettherr e
eacch iis
evven or od
dd: (133)(2
2455), (13
352
24),, (13)((45
5).
(5)) Wha
W at iss th
he o der of a ccyclle o
ord of le gth m?
eng
(6)) Liist the

e siix m
membberss o
of P3, andd ccircle tthe
e evven
n oness. V
Via
a th
he
isoomoorp phissm be etw
wee
en P3 a andd D3, whicw ch syymm mettrie
es of thhe
tria
anggle are emmatcheed witth e
eve
en p
perrmu utattion
ns??
(7)) List all 24

4 p perm
muutattion
ns o
of thee le
ette
ers {A
A,B,C,D
D}. Write
W e e eacch
ermuta
pe atio
on in ccycle n on. Ciircle th
notatio he eve
en onnes.
(8)) La
abeel the veerticcess off a squ
uarre aas ssho own n. IIn yyou
ur
listt o
of tthe
e 224 permuta ations off {A A,BB,CC,D}}, deccide
wh hich
h p
perm mutationn is inducced d b
by e eacch of the e eeighht
symm mettries o
of the sq quaare. Thhatt is,, affterr pe
erfo ming
orm
thee ssymmm metrry, wh hich w woord doo yyou u ssee e (a alw
wayys
Exercises 85
reading words clockwise around the square starting at the top-left

vertex)? Verify that the symmetries which induce even
permutations are exactly {I, H, V, R180}.
(9) In the previous exercise, instead of reading

words clockwise around the square, what if we read
in the “page of a book” order, so that the identity
position is the one illustrated here. Verify that the
symmetries which induce even permutations are {I, H, V, R180},
as before.
The punch line: It makes sense to ask whether a symmetry

induces an even or odd permutation of the vertices; the answer
does not depend on how the vertices were ordered.
(ٍ10) Which of the symmetries of a regular pentagon induce

even permutations of its vertices? What about a regular
hexagon? Conjecture a pattern: which symmetries of a regular
n-gon induce even permutations of its vertices? HINT: Consider
separately the cases where n is even and where n is odd.
(11) Explain why the adjacent swaps generate Pn. In other words,
every permutation in Pn can be expressed as a composition of
adjacent swaps.
(12) Express the identity permutation in P7 in several different

ways as a composition of swaps. Check that you always have an
even number of swaps.
(ٍ13) Prove that exactly half of the permutations in Pn are even.

HINT: use the “ancient scroll” idea in the proof of the All-Or-Half
Theorem from Chap. 2.
(ٍ14) On the final exam for their symmetry course, students are
asked to match five objects with their five symmetry groups. What
do you think is the fairest method for grading such a matching
problem? For example, try regarding each student’s answer as a
86 6. Permutations
permutation in P5, and scoring based on how “close” it is to the

identity permutation (the correct answer). How should “close” be
measured?
(ٍ15) If H is a subgroup of Pn, prove that either all or half of the

members of H are even.
(16) Is P2 isomorphic to C2?

͹Ǥ
So far, we have only studied the symmetries of flat two-
dimensional objects in the plane. In this chapter, we discuss the
symmetries of solid three-dimensional objects in space. Look at
the following solid objects, and rank them from the most
symmetric to the least symmetric.
Images from Robert Webb’s Great Stella software, http://www.software3d.com/Stella.html
Each of these objects is built from red and yellow faces, silver
edges, and gold vertices. To rank them, we must decide how our
previous methods for studying symmetry apply to solid objects.
Rigid Motions of Space

A symmetry of a solid object is defined in the same way
that a symmetry of a flat object was defined:
DEFINITION: A symmetry of a solid object in space means a

rigid motion of space which leaves the object apparently
unchanged.
But what is a “rigid motion of space”? Intuitively, it is a moving/

repositioning of all of space that does not compress, expand, or
otherwise distort distances. Imagine that space is completely
filled with transparent ice, with an object entombed somewhere

88 7. Symmetries of Solid Objects
within. A rigid motion moves the entire infinite expanse of ice,

and it is called a symmetry of the object if it leaves the object
apparently unchanged. How can an infinite expanse of ice be
moved, and where does the mover stand? We do not have a
good answer; the ice image is imperfect, but it is the best we can
offer until we provide a more precise definition of a rigid motion in
a later chapter. For now, here are some examples.
TRANSLATIONS: A translation moves every point of space the

direction and distance specified by a single
arrow. Imagine that the ball pattern
illustrated here is continued infinitely in
all directions, so that the balls fill up all
of space. Each of the three arrows
represents a translation symmetry of this
infinite ball pattern. For example, the red
arrow translates each ball 5 ball positions in
the right-back direction. Since each ball moves
exactly on top of another ball, an observer would not notice that
the translation occurred – that is what makes it a symmetry.
ROTATIONS: In space, we rotate about an axis (which means a

line), not about a point. Imagine children rotating about a
maypole. The axes of some rotation symmetries of a cube are
shown below. Try to do each of these rotations to a cardboard
cube to convince yourself that each one really is a symmetry.
Entombing your cardboard cube in an infinite expanse of ice is
optional; it is fine here to think of each rotation as something
done just to the cube rather than to all of the surrounding space.
Rigid Motions of Space 89
3 axes of rotation symmetries of the cube
The order of an axis means the order of the smallest non-identity

rotation symmetry about that axis. For example, the axes pictured
above have orders 4, 3, and 2, respectively.
REFLECTIONS: A reflection
across a plane is visualized by
thinking of the plane as a
mirror, and moving each point
of space (each speck of ice) to
the position of its mirror image
on the opposite side of the
mirror. For example, the
reflection across the green
plane exchanges the ice above
and below this plane; it is NOT
Image by YassineMrabet, Wikipedia.org
a symmetry of the human figure
because, after this reflection occurs, he would appear upside
down. Neither is the reflection across the blue plane, which would
leave him facing backwards. Only the reflection across the
orange plane is a symmetry of the human figure.
It is easy to visualize a translation or a rotation being

physically done to a solid object (and to the surrounding ice, if
you wish). However, reflections are different. A reflection cannot
be physically done to a solid object. There is nothing you can

physically do to the human figure that will exchange his right and
left halves. The reflected image of his right hand looks like his left
hand, but there is nothing you can physically do to a solid right
hand to turn it into a solid left hand. Why should we care about
reflections if we cannot do them? Even though reflections cannot
be physically performed, they still help explain why solid objects
appear to be symmetric. The man has bilateral symmetry, which
means that his only symmetry (other than the identity) is a single
reflection. That is why his left half looks the same as (or at least
like the mirror image of) his right half. What other living things
have approximately bilateral symmetry?
A rigid motion is called proper, if it transforms a solid right

hand into a solid right hand and thus can be physically done to
solid objects. Rotations and translations are proper. A rigid
motion is called improper, if it transforms a solid right hand into a
solid left hand, and thus cannot be physically done to solid
objects. A reflection is improper, and so is a reflection followed by
a rotation or translation.
We will eventually define “rigid motion of space” more

precisely, but to get by for now, we offer this:
CLASSIFICATION OF RIGID MOTIONS OF SPACE: Any rigid

motion of space can be obtained by composing rotations,
reflections, and translations (no more than one of each kind is
needed).
In other words, there are no rigid motions of space other than the
types we just considered (and compositions thereof).
An object in space is called bounded if it is entirely

contained in some cube. Otherwise, it is called unbounded. The
meaning would remain unaltered if the word “cube” was replaced
by “sphere” or “pyramid” or many other possibilities; if an object
Rigid Motions of Space 91
can be fit inside one of these shapes, then it can be fit into all
of them. The previously mentioned infinite ball pattern was
unbounded; we used it to illustrate translation symmetries.
Bounded object never have translation symmetries. In fact:
THE CENTER POINT THEOREM: Any bounded solid object in

space has a “center point” such that:
(1) Every proper symmetry is a rotation about an axis through this

center point and
(2) Every improper symmetry is a reflection across a plane

through this center point, possibly composed with a rotation
about an axis through this center point.
Part (2) is necessarily lengthier than its analog for bounded flat
objects, because for bounded solid objects, a reflection
composed with a rotation is NOT necessarily equal to a
reflection.
The symmetries of any solid object form a group (under

the operation of composition). As before, this group is called the
symmetry group (or the full symmetry group) of the object. The
proper symmetries form a subgroup, which is called the proper
symmetry group of the object.
Our main goal in this chapter is to understand all of the

possible ways in which bounded solid objects can be symmetric.
But what exactly does this mean? When should we consider two
bounded solid objects to be “symmetric in the same way”?
Should this mean that they have isomorphic symmetry groups?
Or should it mean that the have isomorphic proper symmetry
groups? Or should it mean that they are rigidly equivalent or
perhaps properly rigidly equivalent, defined as follows:
DEFINITION: Two solid objects are called rigidly equivalent if

there exists a rigid motion of space which, when applied to the
first object, repositions it so that afterwards, the two objects have
exactly the same symmetries.
Two solid objects are called properly rigidly equivalent if there

exists a rigid motion of space which, when applied to the first
object, repositions it so that afterwards, the two objects have
exactly the same proper symmetries.
These are all reasonable ways to capture the idea that a

pair of objects is “symmetric in the same way.” Some imply
others. For example, if two objects are rigidly equivalent, then
their full symmetry groups are isomorphic (just as before).
Similarly, if two objects are properly rigidly equivalent, then their
proper symmetry groups are isomorphic.
Why are proper symmetries getting so much attention?

Because understanding an object’s proper symmetries is often
easier than understanding all of its symmetries. You will see. The
Center Point Theorem tells us that the proper symmetries of a
solid bounded object are all rotations.
Our next goal is to understand the full symmetry group (or

at least the proper symmetry group) of a tetrahedron, cube, and
dodecahedron pictured below.
Tetrahedron Cube Dodecahedron

The Symmetry Group a Tetrahedron 93
Take time now to build these three shapes. Cardboard

works well. To keep track of the required building materials, we
will denote by “F” the number of faces and by “S” the number of
sides that each face has. A tetrahedron has F = 4 faces that are
identical equilateral triangles (S = 3). A cube has F = 6 faces that
are identical squares (S = 4). A dodecahedron has F = 12 faces
that are identical pentagons (S = 5). Cut out the faces and glue or
tape them together. Another approach is to find cut-out templates
for each shape by doing a web image search for something like
“dodecahedron template.”
Set one of these three cardboard shapes on a table. To

perform a proper symmetry, you might pick it up, rotate it, and
then set it back down on the table in its original footprint. In how
many ways can you do this? You really only have two choices to
make: which face goes down and how this bottom face is rotated.
The number of ways to make these two choices in succession
equals F × S. We, therefore, expect that the proper symmetry
groups of these three shapes have the following sizes:
F S Number of proper symmetries

Tetrahedron 4 3 12
Cube 6 4 24
Dodecahedron 12 5 60
Counting symmetries is a good start, but it is not enough. We will

now look in more depth at the symmetry group (or at least the
proper symmetry group) of each of these three shapes.
The Symmetry Group a Tetrahedron

Illustrated below are three types of symmetries of a tetrahedron:
Pictured first is an axis of order 3, which connects a vertex

to the midpoint of its opposite face. There are four such axes
(one for each vertex), with rotations by 120 and 240° about each,
yielding a total of 8 non-identity rotations about such axes.
Pictured second is an axis of order 2, which connects the

midpoints of a pair of opposite edges. There are three such axes
(one for each pair of edges), with a 180° rotation about each,
yielding a total of 3 non-identity rotations about such axes. We
thus have 8 + 3 + 1 = 12 total rotation symmetries (counting the
identity), as expected. These are all of the tetrahedron’s proper
symmetries.
Pictured third is a reflection plane. There are six such

planes, one for each edge. These are all of the tetrahedron’s
reflection planes; however, they might NOT be all of the
tetrahedron’s improper symmetries. The Center Point Theorem
does NOT say that every improper symmetry is a reflection.
We wish to fully understand the tetra-

hedron’s symmetry group. To build a Cayley
table, we must first choose a name for each
symmetry. Here is an illuminating naming
system. Let us label the four vertices A, B,
C and D as illustrated – think of this
illustration as the identity position, which we
The Symmetry Group a Tetrahedron 95
call ABCD. Name each other symmetry with the four-letter word
that you see after that symmetry is performed. Always read
words in the order indicated by the illustration (top, bottom-right,
bottom-front, and bottom-left). Thus, we name symmetries
according to the permutations of the tetrahedron’s vertices which
they induce. This system will inevitably lead you to stumble upon
the following theorem:
THEOREM: A tetrahedron’s symmetry group is isomorphic to P4,

and its proper symmetry group is isomorphic to A4.
PROOF: The previously described naming strategy associates

each symmetry with a permutation in P4 (a four letter word). No
two different symmetries could induce the same permutation; that
is, if you know how the vertices were permuted, then you know
what symmetry was performed.
Further, for any permutation of the vertices, we can find a

symmetry that achieves that permutation. How? First, the top
vertex (A) can easily be rotated into the correct position. In both
illustrations below, it has just been rotated to the bottom-right
position. Next, the rotation and reflection pictured below can be
combined to obtain all six possible permutations of the remaining
three vertices (B, C, and D around the green face).
In summary, there is a one-to-one correspondence

between the symmetries of the tetrahedron and the permutations
of its four vertices. This one-to-one correspondence provides the

desired isomorphism. The permutation induced by the composi-
tion of two symmetries equals the composition of the permutations
which these two symmetries induce. That is why it is an
isomorphism. In the exercises, you will verify that the proper
symmetries correspond to the even permutations, which is why the
proper symmetry group is isomorphic to A4………….……………Ƒ
The Proper Symmetry Group a Cube

Next, let us study the proper symmetry
group of a cube (we will worry about the full
symmetry group later). The cube has 24 proper
symmetries. The only permutation group or
alternating group with this size is P4, which
might lead you to guess the following:
THEOREM: The proper symmetry group of a cube is iso-

morphic to P4.
PROOF: A cube has 6 faces, 8 vertices, and 12 edges. To mimic

the idea of the previous proof, we must
find something of which the cube has 4.
The solution is pictured here; it has 4
diagonals (colored green, blue, red, and
yellow). The four diagonals are permuted
by the proper symmetries of the cube.
Check that any permutation of these four
diagonals can be achieved by a proper
symmetry. Also check that no two different
proper symmetries ever induce the same permutation of the
colors. You will need a physical cube to convince yourself of
these claims. If your cube is solid, you cannot run colored
diagonals through it, but you could instead just color its vertices.
In summary, there is a one-to-one correspondence between
The Proper Symmetry Group a Dodecahedron 97
the proper symmetries of the cube and the permutations

of its four diagonals. This correspondence is the desired
isomorphism………………………………………………..…………Ƒ
Here is an alternative phrasing of the key idea from the

previous proof. The largest possible stick (line segment) that is
able to fit inside of a cube can actually fit in exactly four different
ways, and these four ways are permuted by the symmetries of
the cube.
The Proper Symmetry Group a Dodecahedron

Finally, let us turn our attention to the dodeca-
hedron. The dodecahedron has 60 proper
symmetries. The only permutation group or
alternating group with this size is A5, which might
lead you to conjecture that:
THEOREM: The proper symmetry group of the dodecahedron is

isomorphic to A5.
IDEA OF PROOF: We need to find

something of which the dodecahedron
has five, which get permuted by its
proper symmetries. Here is the
answer: it is possible to color all of
its edges using five colors (blue,
green, pink, purple, and yellow) in
such a way that each proper
symmetry permutes these colors. To
verify this, color the edges of your
cardboard model as illustrated. If you color correctly, then you
can verify the following: (1) Each proper symmetry permutes the
colors; for example, if a proper symmetry sends one pink edge
to a yellow edge, then it sends all pink edges to yellow edges.
(2) Each proper symmetry permutes the colors by an even
permutation. (3) No two different proper symmetries ever per-

mute the colors in the same way. (4) Every even permutation of
the colors can be achieved by a proper symmetry. Use your
cardboard model to make sure you understand and believe each of
these claims…………………..……………………………..…………Ƒ
Here are explicit instructions for coloring your

dodecahedron model to make the previous proof work. You need
a total of six pink edges. Color some first edge pink, and the
remaining five pink edges should be the edges that are either
parallel or perpendicular to the first. Repeat with the other colors.
Since the six edges of a single color are chosen to be

mutually parallel or perpendicular, each color determines a
“frame.” Each such color frame instructs you how to position a
cube that you might wish to inscribe inside of the dodecahedron
(so that every edge of the cube is parallel
to an edge of the dodecahedron with
that color). The illustration on the left
shows a cube positioned by the pink
frame. In fact, a largest possible
cube that is able to fit inside of a
dodecahedron can actually fit in five
different ways (determined by the five
color frames), and these five ways are
permuted by the symmetries of the dodecahedron.
This draws a nice analogy between the cube’s story and

the dodecahedron’s story. Fitting largest possible sticks into a
cube is analogous to fitting largest possible cubes into a
dodecahedron.
Solid Objects Which Are “Essentially Two-Dimensional” 99
Solid Objects Which Are “Essentially Two-Dimensional”

It was hard work figuring out the proper symmetry groups
of the tetrahedron, cube, and dodecahedron,
largely because these shapes had so many
rotation axes. What about a shape with
only one rotation axis, like the star-
based pyramid illustrated here? That
is much easier. Its proper sym-
metry group is isomorphic to
the proper symmetry group
of the star-shaped
shadow that it would
cast on the ground,
namely C5. We will call
this pyramid “essentially
two-dimensional” be- Images from Robert Webb's Great Stella software
cause its proper sym- http://www.software3d.com/Stella.html
metries are no more
complicated than the symmetries of its two-dimensional shadow
(cast when light is aimed down its only rotation axis).
The thick hexagon pictured on the right

is also “essentially two-dimensional.” At first
glance, it might appear that the central blue
axis (which has order 6) is its only rotation
axis, but look again! In fact, it has six
other axes, each of order 2. We are talking
about the side axes, like the one drawn below
in black.
Like all rotations, the 180° side

axis rotation is a proper symmetry, which
might seem confusing because it reminds
you of an improper symmetry of the flat
hexagon. Thus, the proper symmetry group
of the thick hexagon is isomorphic to the full
symmetry group of its flat hexagon shadow,
namely the dihedral group D6. In fact, the word “dihedral” comes
from the Greek “dihedron” which means a solid with two faces,
like the thick hexagon. Whenever a flat two-dimensional object is
thickened a bit, its proper symmetry group becomes isomorphic
to the flat object’s full symmetry group. This is because the third
dimension of space can be used to rotate the object upside
down.
DEFINITION: A solid three-dimensional object is called

essentially two-dimensional if it has no more than one rotation
axis of order >2.
Thus, an essentially two-dimensional solid object only has

one rotation axis (if you do not count order 2 axes). The solid
object’s proper symmetries are no more interesting than the
symmetries of the two-dimensional shadow cast when light is
shown down this main symmetry axis.
More specifically, the star-based pyramid and the thick

hexagon above exemplify the only two ways that an essentially
two-dimensional solid object can be related its shadow: (1) There
are no side axes, and the proper symmetry group of the solid is
isomorphic to the proper symmetry group of its shadow or
(2) There are side axes, and the proper symmetry group of the
solid is isomorphic to the full symmetry group of the shadow. In
either case, Da Vinci’s Theorem implies:
The Classification Theorem for Bounded Objects 101
THEOREM: The proper symmetry group of any essentially two-

dimensional solid object is either infinite or isomorphic to a
dihedral or cyclic group.
The hockey puck and the cone shown here are examples
of essentially two-dimensional solid objects
with infinitely many symmetries. Each has
a circular shadow. The proper symmetry
group of the hockey puck is isomorphic
to the full symmetry group of this circular
shadow. The proper symmetry group of the cone is isomorphic to
the proper symmetry group of this circular shadow.
The Classification Theorem for Bounded Objects

In the last several sections, each new example was more
complicated than the previous. The tetrahedron had 12 proper
symmetries, the cube had 24, and the dodecahedron had 60.
Can we continue building more and more complicated solid objects
with more and more symmetries? We did so with flat objects
when we considered 2-gons, 3-gons, 4-gons, 5-gons…this list
goes on indefinitely, with each shape having more symmetries
than the previous one. But quite surprisingly, we cannot build
solid objects which are any more complicated than the objects
that we have already considered! The three-dimensional analog
of Da Vinci’s theorem is much more restrictive that you might
have expected. Here it is:
THE CLASSIFICATION THEOREM: Every bounded solid object

is either essentially two-dimensional or is properly rigidly
equivalent to a tetrahedron, cube, dodecahedron, or sphere.
If the object is essentially two-dimensional, then its proper

symmetry group must be infinite or isomorphic to a cyclic or
dihedral group. If the object is genuinely three-dimensional, it is

amazing that there are only four more possibilities for its proper
symmetry group! Thus, the group version of the classification
theorem says this: The proper symmetry group of any bounded
solid object is infinite or isomorphic to a dihedral or cyclic group or
A4, P4, or A5.
It is remarkable that there are so few possibilities

for the proper symmetry groups of solid bounded
objects! To appreciate what this means, let us think
about people who might wish that there were more
possibilities. For example, imagine a medieval blacksmith
named Robin who is commissioned by a knight named Sir
Dim to build a spike ball weapon with 12 sharp spikes. Sir
Dim insists that the 12 spikes be “spread out with no bald
spots,” to best smite his enemies. That is, he wants the
weapon to look the same to an enemy facing one spike
as it does to an enemy facing any other spike. He is really
requesting here that the spike arrangement be symmetric –
ignoring the chain, there should be a proper symmetry moving
any spike to any other spike. Since a dodecahedron
has 12 faces, Robin is in luck – he can arrange the
12 spikes at the midpoints of the faces of a
dodecahedron.
The next year, Sir Dim demands a new weapon built with
20 spikes. Robin is again in luck, because the dodecahedron has
20 vertices, so he can arrange the spikes at the vertices of a
dodecahedron. Robin could also manage 30 spikes, arranged at
the midpoints of the 30 edges of the dodecahedron, or even 60
spikes (think about how). But the next year, Sir Dim requests a
weapon with 100 spikes – the largest number he knows, and
Robin is out of luck. It is impossible to build a perfectly symmetric
The Classification Theorem for Bounded Objects 103
spike ball with more than 60 spikes because such a weapon

would have more proper symmetries than a dodecahedron,
contradicting the classification theorem. The only way it could be
done is with an essentially two-dimensional arrangement of
spikes around the equator of the ball, but this arrangement would
have large bald spots, and is not what Sir Dim had in mind.
This is not an engineering restriction or an issue of

manufacturing imperfections – it is a purely mathematical
limitation. There is a limit to the number of points which can be
symmetrically arranged around a sphere! Making the sphere
larger or the points smaller does not help. It just cannot be done.
There are many modern versions of this tale. For example,

a golf ball engineer is unable to distribute more
than 60 dimples symmetrically around a golf
ball, and must, therefore, settle for an
approximately symmetric arrangement of the
250–450 dimples found on most balls today.
Unfortunately, this means that the aero-
dynamics could in principle change depending
on how the ball is set on the tee. On the other hand, a lack of
dimple symmetry is not all bad. Certain intentionally asymmetric
dimple patterns, like those used on Polara brand balls, can cause
a ball to fly straighter. The USGA banned the use in tournament
play of asymmetric Polara balls, which lead Polara to file a
lawsuit against the USGA. The USGA was in an interesting legal
and mathematical situation here; it was trying to regulate that
balls must be as symmetric as possible, even while the
classification theorem says that no ball could ever be perfectly
symmetric.
Architects are also limited by the classification theorem.

For example, the arrangement of triangles around a geodesic
dome structure like the Epcot center is approximately symmetric
at best. Similarly, the arrangement of more than 60 seeds around

a goats beard or dandelion could not be perfectly symmetric. In
biology, proteins have been discovered representing all possible
types of proper symmetry groups allowed by the classification
theorem: cyclic, dihedral, A4, P4, and A5. It may not be obvious
how to account for the symmetry of a particular protein, but the
serious lack of options is certainly part of the answer.
Goats beard photo by Ken Tapp

Epcot Center photo by Ylebru, Wikipedia.org
Chirality
A solid object is called chiral if all of its symmetries are
proper. For example, the decorated cube
pictured on the left is chiral; after a reflection, it
would look different because the red cycle on
each face would turn clockwise instead of
counterclockwise, so a reflection could not be
a symmetry. Hold your book up to a mirror to see
what we mean. Imagine that the cube and its reflection
are both solid objects (not flat images on the page and on the
mirror surface). Do you see how they differ?
You could create a chiral tetrahedron or dodecahedron in

the analogous way. A more artistic variation is shown on the right.
Chirality 105
Bulatov’s metal sculpture is

approximately a chiral dodeca-
hedron; it has the same proper
symmetry group as a dodeca-
hedron, but it has no improper
symmetries. Hold it up to a mirror
to check this.
“Looking-glass” is an old
English term for a mirror. In the
opening scene of Lewis Carroll’s Dodecahedron XIV metal sculpture by
novel, Through the Looking Glass Vladimir Bulatov (bulatov.org)
and What Alice Found There, Alice imagines that the reflected
image of her room in the Looking glass is part of a real other
world. She asks her cat, “How would you like to live in Looking-
glass House, Kitty? I wonder if they’d give you milk in there?
Perhaps Looking-glass milk isn’t good to drink.” This comment is
far more interesting that it might at first appear.
Alice goes through the looking glass

All chiral objects act and look differently in Looking-glass

world. For example, the writing on books is backwards. Also,
Looking-glass screws are “counter screws,” so Looking-glass
carpenters learn the rule “righty loosey, lefty tighty,” which does
not even rhyme. This makes it difficult to graduate from Looking-
glass carpenter school.
More serious differences lurk at

the microscopic level. Many molecules
are chiral, which means that their
reflections in Looking-glass world will
look and act differently. In the
illustration here, the reflected image of
the left-handed amino acid is the right-handed amino acid. In fact,
the word “chiral” derives from the Greek word for “hand.”
Naturally occurring organic chiral molecules are almost always
found only in the left- or right-hand version, not both. The right-
handed version of a molecule may interact differently with
mechanisms in living cells that evolved to interact with the left-
hand version. For example, the molecule carvone is responsible
for the smell of caraway seeds, but the mirror-image version of
carvone smells like spearmint!
Thalidomide is a deadly example of the difference

between the left- and right-hand versions of molecules. It was
prescribed in the late 1950s to control morning sickness, but it
turned out to cause birth defects. Before being recalled in 1961,
thalidomide was responsible for birth defects in tens of thousands
of infants worldwide. This tragedy may have been caused by the
manner in which thalidomide was produced. The synthetic
manufacturing of chiral molecules yields equal amounts of the
left- and right-hand versions of the molecule. It is thought that
one version of thalidomide controls morning sickness, while the
other causes birth defects and that the two were not properly
Proper Versus Full Symmetry Groups 107
separated in the manufacturing process. Evidently, it is not safe

for Kitty to drink Looking-glass milk.
Proper Versus Full Symmetry Groups

We have so far only classified the proper symmetry groups
of bounded solid objects. What about the full symmetry groups?
The All-Or-Half Theorem and its proof are valid for solid objects
(check this). That is, either all or half of a solid object’s
symmetries are proper. In certain situations, it is easy to explain
how the proper and improper symmetries fit together into a single
Cayley table.
THEOREM: If an object (two-dimensional or three-dimensional)

has a single improper symmetry of order 2 that commutes with
every one of its proper symmetries, then its full symmetry group
is isomorphic to the product of its proper symmetry group and C2.
A symmetry, A, is said to commute with another symmetry, B, if

the order in which the pair are performed does not matter; that is,
A * B = B * A.
PROOF IDEA: Suppose there is an improper symmetry of order

2 that commutes with all proper symmetries and call it F. Name
the proper symmetries as R1, R2,
R3, and so on. Recall from the proof
of the All-Or-Half Theorem that the
green scroll pictured here is a listing
of all of the object’s symmetries
(proper on the left and improper on
the right). The red scroll is a listing
of all things in the product of the
object’s proper symmetry group and C2. The two scrolls are
aligned so as to suggest a one-to-one correspondence. We
match the left of the green scroll with the left of the red scroll, so
that Rn ļ (Rn,0). We match the right of the green scroll with the
right of the red scroll, so that F * Rn ļ (Rn,1). Think about why

this pairing is an isomorphism. That is, why does it translate
every true green equation into a true red equation? To answer
this, separately consider green equations that combine a pair of
green symmetries from the left, the right, or one of each…………Ƒ
NON-EXAMPLE: In D4, there is no flip that commutes with all of

the rotations, or even just with R90 (check this using the Cayley
table in Chap. 2). Thus, the square does not satisfy the
hypothesis of the theorem, nor does it satisfy the conclusion,
since D4 is NOT isomorphic to C4 × C2.
EXAMPLE: At the end of Chap. 5, we discovered that the

symmetry group of the B border pattern is isomorphic to Z × C2.
Since the horizontal flip commutes with each translation, the
above theorem leads us to this same conclusion. Compare our
reasoning in Chap. 5 to the logic of the above proof.
EXAMPLE: We previously discovered that the proper symmetry

group of a thick hexagon is isomorphic to D6.
We now claim that its full symmetry group is
isomorphic to D6 × C2. To justify this claim,
we must find an improper symmetry which
commutes with all 12 proper symmetries.
The answer is the reflection across the plane
that is parallel to the top and bottom hexa-
gons, and half way between them. With the 12 vertices labeled
as shown, this reflection exchanges lower case with upper case
letters, which might help you visualize why it commutes with the
12 rotations.
As for the cube and dodecahedron, each one of these has

a special improper symmetry called central inversion that
commutes with all of its proper symmetries. Therefore, the full
Proper Versus Full Symmetry Groups 109
symmetry group of the cube is isomorphic to P4 × C2, and the full

symmetry group of the dodecahedron is isomorphic to A5 × C2.
Central inversion is similar to a reflection, but instead of

reflecting across a plane, it reflects across a point, namely the
center point of the object. What does it mean to “reflect across a
point”? It means to do two things in succession: First rotate 180°
about an axis through that point and then reflect across the plane
through that point that is perpendicular to this axis. In the diagram
below, central inversion across the green point moves the each
position (like the yellow one) to its “antipode” across this green
point.
Central inversion moves the yellow start point to its antipode across the green center-point
An object is called centrally symmetric if central inversion

(across its center point) is a symmetry of the object. When we
later learn about matrices, we will verify that central inversion
always commutes with every other symmetry of a centrally
symmetric object. For now, can you visualize why this might be
true? Thus, the full symmetry group of any centrally symmetric
object is isomorphic to the product of its proper symmetry group
with C2. Can you visualize why cubes and dodecahedrons are
centrally symmetric, but tetrahedrons are not?
110
0 7. S
Symmettriess of S
Solid O
Objeccts
Y
You u migh ht cconjjeccturre thatt an ny bound ded
d soolid
d objeect (that iis
noot e
esse enttiallly twoo-dime enssion nal)) is rig
gidlly e
equuiva
alennt tto o
onee th
hesse
seven n o objectts: tetrahed dro
on, cu ube e, d doddeccahheddronn, chiiral teetra
a-
heedroon, ch al cube
hira e, cchiral doode ecah hed on, spherre. Th
dro his is a g goo
od
guuesss, b butt to ma akee it tru
ue, an 8th o objeect mu ust
bee addde ed to the e lisst; naame ely,, a vo olleyyba
all.
Thhe vvollleyballl illusstrated d on n thhe right hass
thee ssamme pro ope er ssym mmmetrry g oup ass a
gro
tettrah
hed dron, butt it diffferrs ffrom m the tettra--
heedroon in tha at iit iss ccentrally symm mettric,
so itss fuull ssym
mm metrry g grooup is iso omorp phicc
to A4 × C2. In su umm maary, anny b bouund ded d so
olidd
obbjecct (tha at iss nnot esssen ntia
allyy tw
wo-dim mennsio ona
al)
is rigiidlyy eq
quiivallent to o a spherre o or tto o
onee off th
hese oobjeectss:
P
Properr syym. grrou Full sym. gro
up. F oup
p.
T
Tetrrahhed dron n P4 (2
24)
C
Chirral tettrah
heddron n A4 (12
2) A4 (12)
Voolle
eyba all A4 × C2 (224)
Cuube P4 × C2 (4
48)
P4 (24
4)
C
Chirall cuubee P4 (2
24)
Do
ode ecaahe edro on A5 × C2 (1220))
A5 (60
0)
Ch
hira
al dode eca
ahe edron A5 (6
60)
N
Notiice thhat we e hhavve not yyet de escribe
ed the p posssib
ble fu
ull
symm mettry grroup ps off essse entiallyy ttwo
o-diime
enssion
nal so olid
d o
obje
ectss.
Th
his top pic willl b
be ddisccussse es. It willl ccom
ed in tthe exxerccise mpleete ou
ur
piccturre o
of a
all p
posssib
ble wa ayss in wh
hich hbbounded solid obj bjeccts cannbbe
symm mettric.
Exercises 111
Exercises
(1) Identify the proper and full symmetry group of several of
Vladimir Bulatov’s sculptures displayed at http://bulatov.org.
Which are chiral?
(2) Verify that all 12 proper symmetries of the tetrahedron

permute the vertices by even permutations; thus, the proper
symmetry group of a tetrahedron is isomorphic to A4.
(3) Is every improper symmetry of the tetrahedron equal to a

reflection?
(4) Describe the full symmetry groups of the star-based pyramid,

the hockey puck and the cone (pictured in this chapter). Which of
these objects are centrally symmetric?
(5) Fill in the following blanks:
The tetrahedron has ___ axes of order 3 and ___ axes of order 2.
The cube has ___ axes of order 4, ___ axes of order 3, and ___
axes of order 2.
The dodecahedron has __ axes of order 5, ___ axes of order 3,

and ___ axes of order 2.
Using nothing but this information, how could you count the
proper symmetries of each object?
(6) We saw that the symmetry group of the tetrahedron is P4,

thought of as the group of all permutations of its four vertices.
Could you alternatively think of this as the group of all
permutations of its four faces? In other words, can we replace
vertices with faces in the proof?
(7) Explain what is wrong with the following reasoning: “The six
faces of the cube get permuted by each symmetry of the cube;
therefore, the symmetry group of the cube is isomorphic to P6.”
(ٍ8) Explain what is wrong with the following reasoning: “The

four colored diagonals of the cube get permuted by each
symmetry of the cube; therefore, the full symmetry group of the
cube is isomorphic to P4.”
(9) Explain what is wrong with the following

reasoning: “The four colors (painted on the
vertices in the illustration on the right) get
permuted by each proper symmetry of the cube;
therefore, the proper symmetry group of the cube
is isomorphic to P4.” How could you recolor the
vertices to make this reasoning correct?
(10) Determine the proper and full symmetry group of a box

whose length, width, and height are all different. What if exactly
two of these measurements are the same?
(11) If a solid object has an odd number of total symmetries, what is

the strongest conclusion you can make about its symmetry group?
(12) Is Looking-glass milk safe for Looking-glass Kitty to drink?
(13) The faces of a soccer ball are pentagons and

hexagons. Obtain a soccer ball and determine its
full and proper symmetry group. Hint: What kinds
of rotation axes does it have? Compare its
answer to that of the tetrahedron, cube, and
dodecahedron.
Exercises 113
(14) Is a thick hexagon centrally symmetric? What about a thick

pentagon? What about a thick n-gon?
(15) identify the proper and full symmetry groups of the single,
double, and triple inner tubes pictured below.
(ٍ16) The seven bracelets below are obtained by wrapping the

seven border patterns around cylindrical bands. Each bracelet is
essentially two-dimensional.
Image by AndrewKepert, Wikipedia.org
(a) Which of these bracelets are chiral?

(b) Which of these bracelets have rotation axes of order 2?
(c) The proper symmetry group of each bracelet is
isomorphic to either a cyclic group or a dihedral group.
Which are cyclic and which are dihedral?
(d) Identify an improper symmetry of the 6th bracelet that
commutes with all of its proper symmetries.
(e) The first bracelet has exactly three proper symmetries

and three improper symmetries. Describe all of them.
Notice that none of the improper symmetries are
reflections. What familiar group is this bracelet’s
symmetry group isomorphic to?
(f) Find a pair of bracelets whose full symmetry groups
are isomorphic.
COMMENT: Each bracelet was obtained from a border patterns

by wrapping some number (usually 6) of iterations of the border
pattern’s basic design element around a band. By using different
numbers of iterations, you obtain not just seven bracelets, but
seven families of bracelets. These families of bracelets represent
all possible ways in which essentially two-dimensional bounded
objects can be symmetric. More precisely, any essentially two-
dimensional bounded object with a finite symmetry group is
rigidly equivalent to one of the bracelets in one of the families.
ͺǤ
The regular polygons played the staring rolls in our study
of symmetry among two-dimensional bounded objects. Though
not as symmetric as a circle, they did pretty well considering they
were built out of straight lines. Each corner looked the same as
each other corner, and each edge looked the same as each other
edge. The three-dimensional analog of a regular polygon is called
a “regular polyhedron” or a “Platonic solid”:
DEFINITION: A regular polyhedron (also called a Platonic Solid)

means a bounded three-dimensional object whose faces are all
identical regular polygons assembled such that each vertex has
the same number of faces meeting at it.
We have already encountered three Platonic solids: the

tetrahedron, cube and dodecahedron. How many others are
there? Since there is an infinite list of regular polygons (each next
one looking more like a circle), you might expect that there is also
an infinite list of Platonic solids (each next one looking more like
a sphere). Or perhaps you expect the opposite – that the
tetrahedron, cube and dodecahedron are the only Platonic solids,
since these were the only objects referenced in the classification
theorem of the previous chapter. Or perhaps you suspect that
there are exactly five Platonic solids because the chapter title
ruined the surprise.
Whatever your guess, we will now attempt to system-

atically find and classify the Platonic solids. You will need some
supplies. Obtain scissors and cardboard, and cut out a small pile
of identical copies of each of these regular polygons: 3-gons
(triangles), 4-gons (squares) and 5-gons (pentagons). Start by
choosing a pile, which we will call “S” (for Sides). For example,
S = 3 means you are building your Platonic solid entirely out of
116 8. The Five Platonic Solids
triangles, S = 4 means squares, S = 5 means pentagons, and so

on. Next decide how many copies of this shape will meet at each
vertex. We will call this choice “C” (for Copies). Assemble C
copies of an S-gon together to form your first vertex, and tape up
the sides. Then do the same for each new vertex, and see
whether you end up with a valid Platonic solid. For example, the
choice S = 3 and C = 3 yields a tetrahedron:
S=3 and C=3 (triangles meeting 3 to a vertex) yields a tetrahedron.
The tetrahedron is not the only Platonic solid that you can
build with triangles. The choices C = 4 and C = 5 yield new
Platonic solids called the octahedron and the icosahedron.
S=3 and C=4 (triangles meeting 4 to a vertex) yields an octahedron.

The Five Platonic Solids 117
S=3 and C=5 (triangles meeting 5 to a vertex) yields an icosahedron.
What else can be built with triangles? The choice S = 3

and C = 6 does not seem to work because six triangles meeting
at a vertex leaves no extra room. There are no
gaps between the six triangles, so your
construction will never bend upwards to
become a bounded solid; rather, no matter
how many triangles you add, you will end up
with a flat triangular tiling of a portion of the
plane. What about S = 3 and C > 6? For
example, try taping together seven triangles meeting at a vertex.
This is actually possible to do, as long as you allow some of the
tape lines to extend up above the plane and others to extend
down below the plane. Try to convince yourself that this
construction could never be completed to form a valid (bounded)
Platonic solid. We will use the term overcrowding to refer to the
phenomenon where more regular polygons meet at a vertex than
can fit together in the plane. We conjecture that overcrowding
could never lead to a valid Platonic solid, so we will not waste
time trying any more overcrowded possibilities for S and C.
What can be built out of squares? The cube is one familiar

possibility, illustrated below:
S=4 and C=3 (squares meeting 3 to a vertex) yields a cube.
S = 4 and C = 4 does not work because four squares meeting at

a vertex leaves no extra room, and would, therefore, yield a
square-tiling of a portion of the plane. S = 4 and C > 4 does not
work due to overcrowding.
What can be built from pentagons (S = 5)? The only

possibility is C = 3, since C > 3 would be overcrowded. Thus,
only the dodecahedron can be built from pentagons:
S=5 and C=3 (pentagons meeting 3 to a vertex) yields a dodecahedron.

The Five Platonic Solids 119
Are there any possibilities with S > 5? Exactly three

hexagons fit together in the plane with no gaps, so the
choice S = 6 and C = 3 would lead to a hexagonal
tiling of the plane, not a bounded Platonic solids.
Any attempt to build a Platonic solid with S > 6
would fail because of overcrowding. We have
arrived at an important theorem, usually attributed to
Plato:
PLATO’S THEOREM: There are exactly five Platonic solids: the

tetrahedron, cube, octahedron, dodecahedron, and icosahedron.
Tetrahedron Cube Octahedron Dodecahedron Icosahedron

Images from Robert Webb's Great Stella software, http://www.software3d.com/Stella.html
The above tape-and-cardboard discussion provides very

strong evidence that this theorem is true, but we must
acknowledge that more work is required to achieve a completely
airtight proof of this theorem. In particular, it is still necessary to
prove more rigorously that overcrowding never yields a valid
Platonic solid. Also, it is necessary to prove that the tetrahedron,
cube, octahedron, dodecahedron, and icosahedron are all perfect
(not just approximate) Platonic solids. For example, what if it
turned out that a dodecahedron could only be assembled using
slightly irregular pentagons, whose angles vary a degree or two
from being all the same? Our cardboard and tape was too sloppy
to detect such a minor issue. We will not discuss these remaining
details. Mathematicians have taken care of these issues, so we
can move forward with assurance that the theorem is true – there
are exactly five Platonic solids!
Counting Their Parts
If you have not already done so, take time now to build the
remaining Platonic solids out of cardboard. You need to build
them to count their parts. Use your cardboard models to verify
that the number of vertices (V), edges (E) and faces (F) for each
Platonic solid is as follows:
V E F S C
Tetrahedron 4 6 4 3 3
Cube 8 12 6 4 3
Octahedron 6 12 8 3 4
Dodecahedron 20 30 12 5 3
Icosahedron 12 30 20 3 5
As before, “S” means the number of sides that each face has,
and “C” means the number of faces meeting at each vertex.
Examine your Platonic solid models, and notice that “C” also
equals the number of edges emanating from each vertex. That is
another way to say the same thing.
In case we miss-counted the edges, let us check our work by

deriving E from the other numbers in the above table. There are
actually two useful methods for deriving E:
(1) What is wrong with this logic: “Since the cube has six
faces, each with four edges, the cube must have 6 × 4=24
total edges”? Why did this computation mistakenly double
the correct answer that the cube has 12 total edges? Each
edge was mistakenly double-counted because each edge
belongs to two faces. The correct formula is this: the number
of faces times the number of edges per face equals twice
the number of edges. In symbols: F × S = 2 × E.
Duality 121
(2) What is wrong with this logic: “Since the cube has eight
vertices, each with three edges emanating from it, the
cube must have 8 × 3=24 total edges”? As before, each
edge was mistakenly double-counted, this time because
each edge emanates from two vertices. The correct
formula is: the number of vertices times the number of
edges emanating from each vertex equals twice the
number of edges. In symbols: V × C = 2 × E.
Check that the two blue-boxed formulas above are true in each
row of the Platonic solid table.
Duality
What other patterns do you notice in the Platonic solid

table? The color coding is a hint. The cube and octahedron rows
are colored purple. How are these rows similar to each other?
The dodecahedron and icosahedron rows are colored green.
How are these rows similar to each other? What patterns can you
identify? Do these numerical patterns indicate a coincidence or a
fundamental geometric relationship between the color-coded
pairs of Platonic solids? Hoping for the latter might lead you to
discover the concept of “duality.”
DEFINITION: Duality is a procedure for starting with one Platonic

solid and constructing another Platonic solid. The new solid is
built from the old solid via these steps:
1. The vertices of the new solid are the centers of the faces of
the old solid.
2. Two vertices of the new solid are connected via a new

edge exactly when the faces of the old solid (on which these
vertices are centered) share an old edge.
122
2 8. T
The Five
e Pla
aton
nic S
Solid
ds
If the
e old ssolid is a cu ubee, leet u
us ima agine wh hat the en neww soolidd will
loook like To hellp yyou
e. T u viisualizze the e prroccess, ffindd a rooomm whic
w ch iis
shape ed rou ughhly like a cu ubee. PPaint a an “X” on the ce ente er o
of itts rroo
of,
its flo oorr, aandd e eacch of itss fo ourr w walls. Th hesse X-m markss a are th he
verticces off the ennew w so olid
d. U
Use e kiite string
g fo
or thhe edgess o he new
of th w
solid.. Run a pie ece of kitte sstring fro om the e cceiliing-X to ea ach wa all-X
X
(beeca ausse tthe ceeilinng ssha aress a
an e edgge wwith each w
he walll), a
andd siimilarlly
om the floo
fro or-XX to ea achh w walll-X. RRun n a p piecce off kkite sttrin ng
beetwe eenn thhe X-m maarkss on e eacch p pair oof
waalls th hat sh harre an n eedge (fro ont-
rig
ght, frront-le eft, ba ackk-rightt, and d
baack--lefft). Do o NNOT n a piiece
T rrun e
of kitte strring g bbetw wee en X--ma arks
onn p pairrs of wa allss thatt d do no ot
share e ann edge e (cceilling
g-flo
oorr, rig
ghtt-
lefft, ffron
nt-bbacck).. AAs d dep d on
pictted
thee rrigh ht, the e new w sollid is an
occtah hed dronn!
Thhis iillusstra on help

atio ps uss Image bby User:44C, W Wikiiped
dia.o
org
acccouuntt foor a of tthe relatiionshiips that we
all o e prrevviouuslyy diisco oveere
ed
beetweeen n tthee ccub be an nd occtah hed dron row ws off tthe ta able e. Fo or
exxammple e, the e nnum mbeer of ve erticcess oof tthe ne ew so olid
d (the e ooctaa-
heedroon) eq quaals the e n
num mbe er of faccess off th he oldd so olid
d (tthe e cuubee)
beecause e thhe new w ssolid’ss ve
erticcess are the e ceentterss off th
he oold so olid’’s
faccess. F
Furttheer, tthee neew
w an nd old d so olid
dhhavee the sa ame en nummbe er oof
eddges, becau usee th he duuality pro oceedu ure requiired d uus to consttrucct
onne (kitte striing g) e ge of the
edg e n neww ssoliid forr eaachh (ceiiling g–wwalll,
floor––wa all, orr wall– all bo
–wa oundarry) ed dge
e off th
he old d so olid
d. F Finallyy,
thee nnummbe er of facess o of tthee nnew w ssolidd eequualss tthe num mbe er oof
verticces off th he old d soolid
d. IIf e
eigh ht sspiderrs buiilt tthe
eir wwebs at th he
eigght veerticcess off the o
old solid (the ccorn nerrs o
of the roo omm), tthen e eacch
spideer wwou uld ha ave a difffere entt fa
ace of the enneww so olid
d so o sspend it iis
daay sstarring
g att.
Duality 123
The duality procedure undoes itself. If the old solid is the

octahedron, then the new solid is the cube! We say that the cube
and octahedron form a dual pair or that each is dual to the other.
Imagine yourself in an octahedron-shaped room, painting
X-marks and running kite string. Can you see the cube taking
shape? It is illustrated below.
Cube-octahedron duality. Images by Peter Steinberg, commons.wikimedia.org
If you start with a dodecahedron, the duality procedure yields

an icosahedron, and vise versa; these two Platonic solids form
another dual pair.
Dodecahedron-Icosahedron duality. Images by Peter Steinberg, commons.wikimedia.org

If you start with a tetrahedron, then the duality procedure

yields a tetrahedron. The tetrahedron is therefore called self-dual.
Try to draw a picture of this.
In each of the above duality illustrations, it is easy to see

that the old solid and the new solid have the same proper and
improper symmetries. Every symmetry of the old solid is also a
symmetry of the new solid inscribed inside it, and vice versa. In
particular, the octahedron has the same (proper and full) sym-
metry group as the cube, while the icosahedron has the same
(proper and full) symmetry group as the dodecahedron. With no
extra work required, we have identified the proper and full
symmetry groups of our two new Platonic solids!
Euler’s Formula
Here is a reprinting of our Platonic solid table, with two

new columns added at the end to celebrate our new under-
standing of the proper and full symmetry groups of dual pairs,
and with a mysterious new white column added in the middle.
V E F V+F–E S C Prop Full

Tetrahedron 4 6 4 2 3 3 A4 P4
Cube 8 6 2 4 3
12 P4 P4 × C2
30 A5 A5 × C2
As the white column shows, each of the five Platonic solids

has the mysterious property that its vertex count plus its face
count minus its edge count equals 2. In symbols: V + F – E = 2.
Why always 2? Is this a coincidence, or an indication of some
Euler’s Formula 125
underlying geometric principle waiting to be discovered? Hoping

for the latter, let us hunt for the most general setting in which this
formula is valid.
First, this formula appears in the study of “connected

planar graphs.” What are those? A graph means a finite
collection of vertices (dots in the plane) together with a finite
collection of edges (straight or curved paths beginning and
ending at the vertices). Think of the vertices as towns and the
edges as roads between the towns. A graph is called connected
if it is possible to travel between any pair of vertices (towns)
along the edges (roads). A graph is called planar if the edges
only meet each other at vertices (the roads may meet each other
at towns but do not otherwise cross each other). In the illustration
below, the left graph is a connected and planar graph. The right
illustration show how to change it into a non-planar graph (by
adding a new edge that crosses another edge at a non-vertex
location) and how to change it into a non-connected graph (by
adding a new cluster of edges and vertices with no bridge to the
original cluster).
This graph is connected and planar This graph is NOT

Take out a pen and paper, and draw your own connected
planar graph. Make it as complicated as
you like. When you finish, count
the number of vertices (V), edges
(E), and faces (F) that your
graph has. For example, our
count for the pictured graph
came out V = 9, E = 14, F = 7.
A face means one of the grassy
pastures into which the network of
towns and roads divides the land-
scape. If a farmer placed one cow on
each face, then the cows could not share each other’s grass,
because the type of cow we are talking about here is afraid to
cross roads. One of the faces is always unbounded (number 5 in
our graph), so one lucky cow will have an unlimited supply of
grass to munch. If you count carefully, your numbers will satisfy
Euler’s magic formula V + F – E = 2. In words, the number of
combined faces and vertices equals two more than the number of
edges. This formula is true for our graph (9 + 7 – 14 = 2), and it is
true for your graph also, no matter how complicated you made it.
This is guaranteed by:
EULER’S FORMULA FOR THE PLANE: For any connected

planar graph, V + F – E = 2.
PROOF: The theorem seems magical, but its proof is simple.

Imagine putting a piece of tracing paper over the connected
planar graph that you drew. The idea is to retrace your graph,
one edge at a time, in such a way that V + F – E starts equal to 2
and remains equal to 2 at each step. Start by tracing any one
vertex. At this point, V = 1, F = 1 and E = 0, so V + F – E = 2 for
this starting graph on your tracing paper. Next trace any edge
that starts at that vertex, and also trace the vertex at which this
edge ends. Continue tracing one step at a time. At each step, you
Euler’s Formula 127
must trace a new edge that begins at a previously traced vertex,

and also trace the vertex at which this edge ends (unless the
ending vertex was previously traced). Since your graph was
connected, you will be able to completely trace it in this “one
edge at a time” order. The expression V + F – E started equal to
2, and we claim that it will continue to equal 2 after each new
edge is traced. Why? Each time you trace a new edge, there are
two possibilities:
(1) If the new edge terminates at a previously untraced vertex,

then you trace it, so at this step you just added one new
edge and one new vertex and no new faces.
(2) If the new edge terminates in a previously traced vertex,
then this new edge will divide one large face into two
smaller faces, so at this step you just added one new edge
and one new face and no new vertices.
So either you increased E and V by one each, or you increased E

and F by one each. Neither of these changes effects the
expression V + F – E. If this expression equaled 2 before you
made the change, then it equals 2 afterwards as well. As you
trace the graph, the expression V + F – E starts at 2 and never
changes, so after the last edge is traced, it still equals 2………….Ƒ
Are there other settings in which Euler’s formula is true?

Let us try balloons. Blow up a balloon and draw a graph on it with
a permanent marker. As before, your graph must be connected,
and its edges may only meet each other at its vertices. Now
count V, F, and E. Does the number of combined faces and
vertices equal 2 more than the number of edges? If you counted
carefully, then it must!
EULER’S FORMULA FOR THE SPHERE: For any connected

graph embedded on a sphere, V + F – E = 2.
“Embedded on a sphere” means that the graph is drawn on the

surface of a sphere with its edges only meeting each other at its
vertices. You might visualize it as a network of towns and roads
on the surface of a planet. We do not need a new proof here; just
check that the previous proof of Euler’s Theorem works equally
well on a sphere.
How is Euler’s formula for Platonic solids related to Euler’s

formula for the plane and the sphere? Try this. Instead of
cardboard triangles, squares, and pentagons, use rubber ones to
build your Platonic solids. For example, build an icosahedron out
of 20 rubber triangles, carefully glued together so that the seams
are airtight. Blowing air into your rubber icosahedron will balloon
it into a spherical shape, and its seam lines will form a connected
graph embedded on the sphere, whose vertices, edges, and
faces correspond to those of the original flat-faced icosahedron.
Thus, Euler’s theorem is valid for the five Platonic solids because
it is valid the balloon graphs that they determine on the sphere.
Alternately, if you hold the hollow frame of a Platonic solid

over a piece of paper, shine a light from above, and trace the
resulting edge shadows on the paper, your tracing will be a
connected graph. Choose a good angle to shine the light from so
that your resulting graph is planar. After you straighten their
edges and pretty them up a bit, your tracings might look like this:
Edge-shadow graphs of the 5 Platonic solids
Can you tell which graph corresponds to which Platonic solid?

The Euler Characteristic 129
The Euler Characteristic
We have seen that Euler’s formula works in at least three

settings: the plane, the sphere, and the Platonic solids. We have
described natural connections between these three settings.
Are there any settings in which Euler’s formula does NOT

work? Try this. Instead of drawing your connected graph on a
plane or a balloon, draw it on an inner tube – the kind used to
float down a lazy river. Or draw it on a “double inner tube” for
couples who like to float together or on a “triple inner tube.”
Use a permanent marker, but draw your graph when the water
park attendant is not looking. Should you expect the expression
V + F – E to equal 2 as before? That depends on how you draw
the graph. If you use only a small part
of the rubber surface, like the
illustration on the left, then you will get
the same V, E, and F counts that you
got when you drew that same graph on
a piece of paper, so V + F – E = 2 as
before. But the largest face of this graph is very peculiar. Unlike
the other faces, its shape could not be formed by deforming
(bending or stretching) a rubber polygon.
Let us start over and draw a new

graph whose faces are all shaped
like deformed polygons. For
example, the faces of the graph
on the right are all deformed
squares. We carefully counted
F = 576, V = 576 and E = 1,152.
Therefore, the expression V + F – E
equals zero (not 2).
In fact, the expression V + F – E equals zero for ANY

connected graph embedded on an inner tube, no matter how
complicated, as long as its faces are all deformed polygons. This
is not much of a restriction. It is OK, for example, if some of its
faces are triangles, while others are squares. The ones that are
squares need not be perfect squares, but may be deformed
(stretched and curved) rubber squares, like those on the inner
tube graph pictured above.
It gets better. Any such graph on the “double inner tube”

will satisfy V + F – E = –2. Any such graph on the “triple inner
tube” will satisfy V + F – E = –4. These observations are the
beginning of an entire field of mathematics called topology.
Different surfaces can be distinguished by their value of V + F – E.
This value is called the Euler characteristic of the surface. Thus,
the Euler characteristic of the sphere equals 2, of the inner tube
equals 0, of the double inner tube equals –2, and of the triple
inner tube equals –4. Do you see the pattern? The Euler char-
acteristic of a quadruple inner tube equals –6, and so on.
The Euler characteristic is unaffected by bending and

stretching. You could stretch your spherical balloon into an egg
shape, and this would not change its Euler characteristic because
it would not change the V, E, or F counts for a graph drawn on
the balloon. Thus, the Euler characteristic of a surface measures
An Algebraic Proof that There Are Only Five Platonic Solids 131
some essential quality of its shape that is unaffected by bending

and stretching. Surfaces with different Euler characteristics
cannot be bent or stretched into each other. If you meet a friend
while floating down a lazy river, you cannot make room for her by
stretching your single inner tube into a double, because the
single and double tubes have different Euler characteristics.
An Algebraic Proof that There Are Only Five Platonic

Solids
Let us summarize all of the relationships that we have

discovered among the numbers in our Platonic solid table, which
is reprinted here:
V E F S C
Tetrahedron 4 6 4 3 3
Cube 8 12 6 4 3
First, we compared a Platonic solid to its dual: they have

the same number of edges, their vertex and face counts are
exchanged, and their “S” and “C” values are exchanged. These
relationships are all explained by the geometric process of
duality.
In addition, we discovered that the following four

relationships are valid for each row of the table:
(1) F × S = 2 × E (2) V × C = 2 × E (3) V + F – E = 2

(4) S 3 and C 3.
We will prove now that the five rows of our Platonic solid
table are the ONLY solutions to this system of four equations.
ALGEBRA LEMMA: There are only five ways to choose positive

integers {V, E, F, S, C} so that equations (1)–(4) are satisfied;
namely, the five rows of the Platonic solid table.
In other words, if you gave the four equations to a friend who had
never heard of Platonic solids and who did not know what the
variable stood for, your friend could, using only algebra, conclude
that there are only five solutions to the system of equations. Here
is how your friend would do it.
PROOF: Solving (1) for F gives F = (2 × E)/S. Solving (2) for V

gives V = (2 × E)/C. Substituting these expressions for F and V
into (3) gives: (2 × E)/C + (2 × E)/S – E = 2, which is the same
thing as E × (2/C + 2/S – 1) = 2. In particular, we learn that:
(2/C + 2/S – 1) is positive.
Both S and C are at least 3. If either S or C were larger than 5,

then (2/C + 2/S – 1) would NOT be positive. Therefore S = 3, 4,
or 5 and C = 3, 4, or 5. If C and S were both 5, or both 4, or if one
of them were 5 and the other were 4, then (2/C + 2/S – 1) would
not be positive. Here are the only remaining possibilities:
S = 3 and C = 3 (the tetrahedron), S = 3 and C = 4 (the

octahedron), S = 3 and C = 5 (the icosahedron), S = 4 and C = 3
(the cube), S = 5 and C = 3 (the dodecahedron).
Once you know the values of S and C, you can use the equation
“E×(2/C + 2/S – 1) = 2” to find the value of E, then use (2) to find
the value of V and then use (1) to find the value of F…………….Ƒ
It is fascinating that we can learn so much using only

algebra. But how does The Algebra Lemma advance our study of
the Platonic solids? A “lemma” means a mathematical fact whose
primary purpose is to help prove a more important theorem. What
important theorem will the Algebra Lemma help us prove?
An Algebraic Proof that There Are Only Five Platonic Solids 133
For one thing, the Algebra Lemma provides an alternative

way to prove that there are only five Platonic solids. Here is the
idea. Suppose that your cousin Karl claims he discovered a 6th
Platonic solid. We know that Karl’s new solid must satisfy
equations (1)–(4). The Algebra Lemma tells us that Karl’s solid
must have the same values for {V, E, F, S, C} as one of the five
familiar Platonic solids. But this means that Karl’s solid is the
same as one of these five familiar Platonic solids – it is not really
new.
Most people regard this new proof as better than our

previous cardboard-and-scissors proof. You can’t argue with
algebra. Nevertheless, it is important to confess that the new
proof has its own shortcomings. For example, how do we know
that Karl’s solid satisfies Euler’s formula? What if it has an Euler
characteristic different from 2? The construction pictured here
has an Euler character-
istic zero. It is not a
Platonic solid because
different vertices have
different numbers of tri-
angular faces meeting at
them. In fact, seven tri-
angles meet at each
vertex along the inner
white rim, which is a
nice illustration of over- Not a Platonic Solid.
crowding. But maybe Photo by Karl Horton, Flickr.com
Karl can figure out how to construct an inner tube shape like this
that actually is a Platonic solid. A proof is required to rule out this
possibility. We are not trying to cast doubt on the theorem – it is
true that there are exactly five Platonic solids – we are just
acknowledging that the proofs which we have discussed so far
are not quite complete.
The Platonic Solids Through the Ages
Human fascination with the Platonic solids extends back

thousands of years. These solids probably first appeared in the
artwork of the Neolithic people of Scotland (~1400 B.C.). They
were studied extensively by Pythagoras
(570-495 B.C.) and the ancient Greeks.
They are named after Plato (428-
328 B.C.), who associated them
with earth (cube), air (octa-
hedron), water (icosahedron),
fire (tetrahedron) and the
universe (dodecahedron). Euclid’s
Elements (~300 B.C.) includes mathe-
matical constructions for all five Platonic
solids, and a proof that there are no
others. There are only five of these
perfect shapes, which elevated their
Kepler's model of the solar system importance in scientific and theo-
logical writings. These five shapes
found their way into all manner of theories. They were claimed to
represent the fundamental pieces out of which all matter is
formed. In 1659, Kepler explained the motions of the known
planets using a model of the solar system that was based on the
five Platonic solids inscribed inside each other (illustrated above).
Today, scientists do not view the Platonic solids as directly

relevant to the motions of the planets or the fundamental building
blocks of matter. Nevertheless, these five solids maintain an
important status within math and science. As we saw in Chap. 7,
every bounded solid object (which is not essentially two-
dimensional) has the same proper symmetry group as one of the
five Platonic solids (or a sphere). Thus, the Platonic solids are
models for the possible ways in which the solid objects around
E
Exerrcise
es 13
35
us, likke mooleccule

es in che emmistrry a and d cell strructture es in bioologgy, ca an
be
e sym mme etric. Wh hy did tthe e H HIVV viruss e evoolvee aan ico osaahe edraal
ssha
apee, a as illustrate ed on the lleftt? A b biologgist’’s
a
ansswe er tto tthiss quuesstioon is in ormed byy th
nfo he
m
matthe ema aticcal facct thhatt theree are so few w
posssibbilittiess. D
Doe es mo odeern sccien nce still
rrega ard
d the P Plato oniic solidss aas repre e-
se entting
g eeartth, airr, w watter, fire, and th he
unive ersee? No ot reallly, buut hheree iss a
an intrri-
5 gu
uing g up pdaate. Scie entissts annd mmattheema aticcian
ns
are
e currrentlyy u using data a ccolllectedd b by the
Huubbble te elessco
ope to o aatteempt to d disccovver th he shape of th he
un
niveerse e. On ne off th he toop co onteend derss is an n a act three
absstra e-
mensiona
dim al shhap pe wh hich h is ca alleed “Poin ncare do odeecaahe edraal
space e” b
beccau
usee itss geeommettry is intima atelly rrela
atedd to
o th
he ssymmmmetrry
gro
oup p o of a do odecah hed on. Pe
dro erh
hap ps the e d doddeccahedrron n d doees
rep
preesent tthee un
nive erse!
Exxerrcissess
(1)) Fo
or e
eacch duaal p
pair off Platoonic ssolid
ds in o
ourr ta
able
e, the va
aluees o
of
“S” aand “CC” aare
e exxchhan nged. Ho ow does the eggeo omeetric p
procced
durre
of dualitty a
acccount ffor this re ationsh
ela hip??
(2)) Th
he du
ualitty p
pro
oceddurre ccan
n be a
appplied
d too ccerttain
n ob
bjeectss whicch
areennot Pla
atonicc so
olid
ds. Wh hat so
olid resullts if yyou
u ap he dualitty
pplyy th
prooce ure to a ssqu
edu uare based pyyram
e-b mid
d?
(3)) C
Couunt the vverrtice
es, edge es, and fa acess oof tthe
e p nar grrap
plan ph
be
eloww, a
and
d veeriffy E
Eule
er’ss fo
orm
mula
a fo
or this grraph
h.
(4) You are asked to draw a connected planar graph with exactly
10 edges. Use Euler’s formula to decide the largest and smallest
possible number of faces that your graph could have. Then draw
two graphs, one with the largest and one with the smallest
possible number of faces.
(5) If a connected planar graph has 12 vertices and 14 faces,

how many edges does it have?
(6) We claimed that we carefully counted F = 576, V = 576 and

E = 1,152 for the inner tube graph pictured in this chapter. That
was a lie. Really, we just guessed the number of faces. We then
knew that the graph had as many vertices as faces, and twice as
many edges as faces. How did we know this?
COMMENT: with these relationships, V + F – E = F + F – 2 × F = 0,

so the Euler characteristic comes out right even if our face count
guess was wrong.
(7) A soccer ball is made from 12 pentagon faces

and 20 hexagon faces. Use this information to
figure out how many vertices and edges it has,
and then use this to verify that the Euler
characteristic equals 2.
Exercises 137
(ٍ8) The planar graph shown here is not connected; rather, it is

built from two “connected compo-
nents,” with no bridge between them.
Compute V, E, and F for this graph,
and verify that V + F – E = 3.
Prove that ANY planar graph with
exactly two connected components
will satisfy V + F – E = 3. Conjecture a
formula for V + F – E in any planar graph with exactly n
connected components. HINT: What is the simplest such graph
you can draw? From the simplest one, you can construct any
other “one edge at a time.”
(ٍ9) For each Platonic solid, the number of proper symmetries

equals F × S and also equals V × C. Explain why.
ͻ
ͻǤ

S
Sym mmetryy iss beau utifful, buut th
hat is no ot w
whyy there e iss so
ommucch
of it in tthe wo orld
d. VViruusees e evoolveed the eir ico
osahhed dra
al shap pe, no
ot tto
bee prrettty, but b beca ausse this ssha apee op ptim
mizzes pe erfo
ormman nce. BBeees
evvolvved th he be eha avioor of bu uild
dingg h hexxaggonal ho oneeyco om
mbs fo or
funnctionnal, no esthettic rea
ot e asoonss. TThe e soolution
n to o a
an opttimmizaatio
on
prooblemm iss o ofte
en hig ghlyy ssymmme etriic. In otther w worrds, ssymmme etriic
shape es arre oftten th he be est sh happess, and d tthiss p
prinncip
ple help ps
acccou untt for th
heirr prrevvale
encce in nn natuure.
Minim
maal S
Surfa
ace
es
W
Whe en yoou bblowwab bub bble
e, itt migh
m ht b be cig gar--shhapped att th
he
insstannt it leeavvess th
he wa and, b but the en it iimmmedia aps into a
atelyy ssna
spherica al shaapee. Wh hy?? WWh hat op ptim
mizatio on prrob blemm is na aturre
almmosst insstan ntaneo oussly soolviing he ere
e? So oap p film is likke ela astiic
strretcche ed tauughht. It wa antss to g gett sma aller; tha at iis, it waantss tto
deecre easse iits surrfacce areea. Buut it iss fo ed to encclo
orce ose yo our bre
eatth
of airr beca ausse surfacce ten nsioon prreve entts iit frrom
m ppop ppin
ng. Give en
thaat it m must en ncloosee a fixxedd vvolu ume e oof airr, itt findss tthe le
east-
surface--are ea wa ay too do so.. T Thee sphe ere
e iss n not ju ust the
e b besst
solutiion th hat thee bbubbblee ccann fin
nd; math
m hem matticianss h havve pro
ove en
thaat itt is the
ebbest amo ong alll co
oncceivvab ble sha ape es:
TH
HE BU UBBL LE T THEO ORE EM
M: TThee sp
phe
ere
e is the le
easst-ssurffacce-a
are
ea
wa
ay tto e
encclosse a g
give
en vvolu
umme.

140 9. Symmetry and Optimization
The sphere is also the most symmetric bounded three-

dimensional shape, so at least in this case, the most symmetric
shape is the optimum shape! As you read this chapter, keep in
mind two principles: (1) The most symmetric shapes are often the
best and (2) the solution to an optimization problem often has the
same symmetries as the problem itself.
Soap solution can answer other optimization problems as

well. If you dip a plastic frame into soap solution, nature finds the
least-area surface stretching over that frame.
Least-area surfaces stretching over the saddle and tetrahedron frames
At the instant that the tetrahedron frame was removed

from the bucket of soap solution, the soap film probably coated
each of its four faces separately, but then it almost instantaneously
snapped into the improved configuration pictured above. The
surface could reduce its surface area further by letting go of the
plastic frame, but chemical bonds prevent this. Given that it is
compelled to cling to the plastic frame, it finds the least-area
surface that does so. Notice that the soap film surface inside
each frame pictured above has the same symmetries as the
frame it clings to. This illustrates our second principle.
Minimal Surfaces 141
A least-area surface stretching over a cube frame is

illustrated below, viewed from two different angles.
A least-area surface stretching over a cube frame, viewed from 2 angles
You might have expected the small square film in the center to
collapse to a point, but a small square turns out to be better than
a point. A good calculus student could verify this, but not nearly
as quickly as the soap film figured it out.
The solution to an optimization problem often, but not

always, has the same symmetries as the problem itself. Unlike
the saddle and tetrahedron frames, the soap film in the cube
frame does not have the same symmetries as the frame itself.
This is possible because there are actually three soap film con-
figurations that tie as least-surface-area winners, corresponding
to whether the small square film in the center is parallel to the
front, right, or top face of the cube frame. These three con-
figurations are permuted by the symmetries of the cube (or you
can actually shake the frame to make it switch between these
three configurations). This is reminiscent of our observation in
Chap. 7 that a largest possible cube that is able to fit inside a
dodecahedron can fit in five different ways, and these five ways
are permuted by the symmetries of the dodecahedron.
The study of least-area surfaces is currently an extremely

active area of mathematics research. If you perform a web image
search for “minimal surface,” you will find a gallery of beautiful
soap-film-like images, including bounded surfaces stretching over
frames like those above and also unbounded surfaces extending
indefinitely in all directions.
The Circle Wins
There is a natural two-dimensional analog of The Bubble

Theorem. Suppose that a farmer wants to enclose exactly 25
acres of grass land for his cows to graze on. How can he do so
with the smallest possible length of fence? If you expect the most
symmetric shape to be the best shape, then you will correctly
guess the following theorem:
THE CIRCLE THEOREM: The circle is the least-perimeter way to

enclose a given area in the plane.
Thus the circle, which is the most symmetric bounded object in

the plane, is also the solution to the farmer’s optimization
problem. He should build a circular fence. The proof of this
theorem is all about symmetry.
SKETCH OF PROOF. All of the farmers in the land competed in

a contest to design the least-peri-
meter fence enclosing a given area
(say 25 acres, although this number
does not matter). Farmer Don won!
His fence not only beat the other
farmers’ fences but also it beat all
possible other fences. In other words,
Farmer Don found the least-perimeter
possible way to enclose the given Farmer Don's winning fence
area. We wish to prove that Farmer might look like this.
The Circle Wins 143
Don’s winning fence is a circle, but for all we know now, it could
have a shape as crazy as the one illustrated above.
Consider the horizontal line that exactly divides the area of

Farmer Don’s fence in half. We claim that this line must also
divide the perimeter of his fence in half. Why? Because if, say,
the top had more perimeter than the bottom, then his fence would
not have really been a winner – replacing the long top with the
mirror-reflection of the short bottom would produce a fence that
beats Farmer Don’s original fence, contradicting our assumption
that his original fence is the winner. In summary, because his
fence is a winner, we know that the horizontal line that divides its
area in half must also divide its perimeter in half. There are equal
amounts of grass on each side and equal lengths of fence on
each side.
Now consider the new fence obtained by replacing the top

half of Farmer Don’s fence with the mirror reflection of the bottom
half (over this horizontal halving-line), like this:
Replacing the top half with the mirror reflection of the bottom yields a tied-winner!
This new fence has the same area and same perimeter as the
original, so it is a tied-winner! It ties with the original, but it is
guaranteed to have at least two symmetries, whereas the original
might have had only one (the identity).
Next consider the vertical line that divides this new fence
in half. As before, it must also divide the perimeter in half, so if
we replace the right side with the mirror reflection of the left side,
as pictured, then we obtain yet another tied winner.
Replacing the right half with the mirror reflection of the left yields another tied-winner!
This new tied winner is guaranteed to have at least four sym-

metries, namely I, H, V and R180. Notice that R180 is a symmetry
because it is the composition of H and V, each of which is a
symmetries. The white dot at which the horizontal and vertical
lines cross is the center point of this new winner.
Observe that EVERY line through this white center point

divides the perimeter and area of the new winner in half, simply
because R180 exchanges the two sides of such a line. But this
implies that every such line must meet the fence at right angles.
Why is a non-right angle, like the angle at which the blue line
pictured below meets the fence, impossible? Because the tied
winner obtained by replacing one side of the blue line with the
mirror reflection of the other side has an “innie-point” (two fence
segments meeting at an angle pointing into the pasture side). But
this is impossible; winners never have innie-points. The fence
near any innie-point can be rounded off (as illustrated in purple)
to yield a fence that is better on all counts: the purple modification
encloses more area and has a smaller perimeter than the
original.
The Circle Wins 145
If the blue line met the fence at a non-right angle, then replacing one side with the mirror
image of the other produces something impossible: a winner with an innie-point.
In summary, starting with Farmer Don’s winning fence, we

built a tied winner with four symmetries, which is guaranteed to
meet every line through its center point at a right angle.
Now let us brainstorm. Think of a fence shape that meets

every radial line (meaning every line through its center point) at a
right angle. How many different such
shapes can you come up with? In
fact, the only solution is a circle!
To convince yourself of this,
image tracing along the fence
edge with a compass that’s
anchored at the center point,
expanding and contracting the
compass width as necessary to
stay on the fence path. Notice that
expanding the compass creates obtuse
angles with the radial lines, while contracting creates acute
angles. The only way to avoid acute and obtuse angles is to
never expand or contract your compass, which only happens

when you are tracing a perfect circle.
Thus, the tied winner with at least four symmetries must be

a perfect circle. This winner looks like four copies of the bottom-
left quadrant of Farmer Don’s original winner, so the original
winner’s bottom-left quadrant must be a quarter-circle. There is
nothing special about the bottom-left. A slight modification to
the above proof establishes that the bottom-right, top-left, and
top-right quadrants are also quarter circles. Thus, Farmer Don’s
original winning fence must have been a circle…….………………Ƒ
That was a long proof! Hopefully it convinced you that the

theorem is true. It is. Nevertheless, we must confess that more
work is needed to make this proof precise and rigorous enough to
satisfy mathematicians today. For example, the compass-tracing
part is really a calculus problem which involves verifying that the
polar coordinate function, ‫ݎ‬ሺߠሻ, has zero derivative (if you do not
know calculus, then please ignore that sentence). A more subtle
issue is this: our proof showed only that a winning fence must be
a circle. This conclusion is vacuous unless we can verify that a
winning fence exists; in other words, that there exists a fence
enclosing the given area that has smaller perimeter than any
other fence (except for possible ties). This is true, and probably
seems obvious, but it does require a separate proof. After all,
there does not exist a largest-perimeter fence enclosing a given
area.
It is not a bad thing that our proof was insufficiently

rigorous. Beautiful new ideas are often discovered when we
strive to put visual proofs and vague heuristic arguments on more
solid and rigorous footing. In this case, proving that a winner
exists lead to important ideas in the field of analysis. In any case,
the details have been filled in, and we now know beyond doubt
that the theorem is true: the circle wins!
E
Exerrcise
es 14
47
Exxerrcissess
(1)) F
Findd tthe
e pperime
eterr o
of tthe
e ccirclle, sq
quaare, a
and
d e
equilattera
al
tria
ang encllosing
gle e g a
are
ea 25 5. Orderr tthe ese thhre
ee sh
hap
pes b by
inccrea
asing peerim
meter.
(2)) Fa
armmerr Ann wa antss to
o fe
encce o
off 25 accress o
of grasss lland a
alon
ng a
ght rivver. Sh
strraig he do
oes no ot n
nee
ed tto ffen
nce aloong he river’ss e
g th edg
ge
beecause eh her cowws arre aafra
aid of waater. G
Gueesss th
he llea
ast-perrim
mete
er
fenncee sh
hap pe.
(3)) W
Whe en tw
wo bub bbles co ollid
de, thhey fo ormm a
“do
oub ble buubble” co gurration wh
onfig hichheenclosses
and ssep para
atess tw
wo (poossiblyy differren
nt) vvolu
ummes
of airr. D
Do aacttual sooap
pb bubble es ffind he lleast-
d th
surfacce--are
ea wa ay to do o sso, o or d do math he-
ma ns kn
aticcian now w o of a be etteer sollution th hat
hysiical so
ph oapp bbub
bblees aaree unab ble o atttain?
e to
Usse online e rresource es as ne eeddedd to o llea
arn the a
ansswe
er to th
he
ouble b
do bub
bble ep blem.
prob
(ٍ4)) F Four ccitie

es lie att th he co orne
erss of a ssquare e. WWee wwishh tto
conne ectt them
m byy a ne etwworkk o
of ro
oadds sso thaat it iss po
osssible to
oddrivve
be
etwe eenn any pa air o
of ccitie
es. Whicch oof the thrree
e ro oad co onfiiguration
ns
piccturred
d be elo
ow use es the e ssma est tottal len
alle ngtth o of roaadwwayy? Ca an
you invventt a ro oadd ccon nfig
gura
atioon thaat bea atss all tthreee off thhesse
config gurrationss?
(ٍ5) The circle is the most-perimeter way to enclose a given

area in the plane. This sentence is completely false, but my friend
PK tried to prove it anyways. To prove it, he copied our proof of
The Circle Theorem, replacing “least” with “most”, and making
other such modifications as needed. For example, he easily
changed the purple path near the innie-point to make it enclose
less area and have larger perimeter than the original. What goes
wrong with his attempt?
(ٍ6) The Circle Theorem is more commonly called The Isoperi-

metric Theorem, from the Greek word for “same perimeter.” This
name is more appropriate for the following alternative version of
the theorem: Among all curves with the same given perimeter,
the circle encloses the most area. In other words, if a farmer has
a fixed length of fence to work with, and he wishes to enclose the
most possible area, he should build a circular fence. Explain why
the two versions of this theorem are equivalent. HINT: Explain
why a counterexample to one version could be rescaled to
become a counterexample to the other.
ͳͲǤǫ
By now you understand the importance of precise
language. The history of mathematics is, among other things, a
story about the invention of ever-more-precise language and
techniques to explore the abstract ideas required to model the
physical world. New ideas force us to look back and more
precisely redefine our old vocabulary. This might sound like
tedious backtracking, but it has been a driving force that has
sparked some of the greatest breakthroughs in mathematics.
It is time now for us to backtrack. We began our story by

defining a symmetry of an object as a rigid motion that leaves the
object apparently unchanged. Now it is finally time to more
precisely define the term “rigid motion.” Historically, this was
necessary to prove many of the theorems found in this book.
Furthermore, since backtracking involves finding new ways to
think about old ideas, it will lead us to unexpected discoveries
and truths.
How far must we backtrack? Before deciding what a rigid

motion of the plane or of space means, we must decide what “the
plane” and “space” mean. Since the plane is made from pairs of
numbers, ሺ‫ݔ‬ǡ ‫ݕ‬ሻ, and space from triples of numbers ሺ‫ݔ‬ǡ ‫ݕ‬ǡ ‫ݖ‬ሻ, we
must first decide what a number is. So let’s start with numbers.
Natural Numbers
In school, you first studied the natural numbers:
N = {1, 2, 3, 4, 5, 6, …} “the natural numbers”
You learned that the most important natural numbers are the
prime numbers: 2, 3, 5, 7, 11, 13, 17, 19, 23, and so on.

150 10. What Is a Number?
DEFINITION: A prime number is a natural number greater than 1

that cannot be expressed as a product of two smaller natural
numbers.
You probably learned that primes are the building blocks of all
natural numbers. More precisely:
THEOREM: Every natural number greater than 1 is either prime

or can be expressed in a unique way as a product of primes.
For example, the prime factorization of 300 is:
300 = 2 u 2 u 3 u 5 u 5.
We figured this out by breaking 300 down step-by-step until the

pieces could not be further broken down…
…like this: 300 = 3 u 100 = 3 u 4u 25 = 3 u 2 u 2 u 5 u 5.
…or like this: 300 = 10 u 30 = 2 u 5 u 3 u 10 = 2 u 5 u 3 u 2 u 5.
The two answers above become the same if the primes are
relisted in increasing order. That is what “in a unique way”
means. Each natural number has only one prime factorization!
By the way, once you know the prime factorization of a

number, it is easy to find the prime factorization of its square. For
example, the square of 300 is 3002 = 300 × 300 = 90,000, which
has the following prime factorization:
90,000 = 300 u 300 = 2 u 2 u 3 u 5 u 5 u 2 u 2 u 3 u 5 u 5
(reorder) = 2 u 2 u 2 u 2 u 3 u 3 u 5 u 5 u 5 u 5.
Notice that 3002 has twice as many of each prime as 300 has in
its prime factorization. In particular, the square of any natural
number greater than 1 has an even number of occurrences of
Rational Numbers 151
each prime in its prime factorization. This observation will be

important in the next section. You’ll see.
Rational Numbers
After becoming acquainted with the natural numbers, you

learned about the integers (which include zero and negatives):
Z = {…, –3, –2, –1, 0, 1, 2, 3, …} “the integers”
(The German word for “integer” starts with Z). Next you
studied rational numbers, which means fractions such as 3/5,
17/8, and –22/3:
Q = {all quotients “a/b” of integers with b0} “the rationals”
The terms “rational number”, “ratio”, “fraction,” and “quotient”

are roughly synonymous. Of course, some fractions are the same
as others. For example, 2/3 = 4/6. If you were finding all solutions
to an equation, like 3x=2, you would not separately list 2/3 and
4/6, because these are the same number. Here is the general
rule: a/b = c/d whenever ad = bc. Regard this comment as an
addendum to our definition of Q. Thus, Q means the set of all
quotients “a/b” of integers with b0, with the understanding that
certain quotients are really the same as others.
If you studied mathematics 25 centuries ago in ancient

Greece, you would have learned that all numbers are rational.
What else could a number be? All physical matter was thought to
be finitely dividable (that is, every object was thought to be made
up of finitely many undividable building blocks), so rational
numbers could describe the exact size of any portion of any
object. Furthermore, the Greeks regarded rational numbers as
divine gifts from their gods. When clues first appeared that other
types of numbers existed, it was deemed blasphemous to pursue
such thoughts.
These early Greek mathematicians developed number

theory and geometry in parallel. They required numbers to
represent not just portions of physical objects but also lengths in
geometric constructions. For example, they
knew that the diagonal of a 1-by-1 square
has a length that squares to 2 (when
multiplied times itself, the answer is 2).
Today, we call this length “the square root of
two,” denoted ξʹ. They searched in vain for
a fraction that squares to two, but their
search led instead to a proof that no such fraction could ever
be found.
THEOREM: There is no rational number that squares to 2.
Today, we say “ξʹ is a number which is not rational – an

irrational number.” This sentence would have puzzled the early
Greek mathematicians because, to them, “number” meant
“rational number” (and also because they spoke Greek).
PROOF: Suppose your uncle Pete claims to have found two

positive integers, p and q, such that the fraction p/q squares to
exactly 2. That is, (p/q)2 = 2. How do we know that Pete is
mistaken? Let us explore the consequences of his claim and
demonstrate that it leads to a contradiction. Re-write Pete’s claim
with some simple algebra like this:
(p/q)2 = 2 ļ p2/q2 = 2 ļ p2 = 2 u q2
Now ask the question: how many 2s are in the prime

factorizations of the left and the right sides of the equation
p2 = 2 u q2? The left side (p2) has an even number of 2s because
it has twice as many as p has. The right side (2uq2) has an odd
number of 2s because it has one more than twice as many as q
has. But the prime factorization is unique. If the left and right
Real Numbers 153
sides really equaled each other, there would not be a difference

between their prime factorizations. There is only one possible
conclusion: Pete was mistaken. There is no rational number
which squares to 2……………………………………………………Ƒ
When the early Greek mathematicians discovered this

theorem, it was an abrupt challenge to their mathematical and
religious belief systems, so they closely guarded this
uncomfortable truth. Only a select few mathematicians were privy
to the secret, and those who were caught sharing it with the
uninitiated were executed!
Real Numbers
Inevitably the secret got out, and it led to the invention
of a number system including more than just fractions. What do
rational numbers like 2/3 and lengths like ξʹ have in common, so
that both can be incorporated into a more general concept of
number? Today, we put them on equal footing by writing them as
decimal expressions:
2/3 = 0.66666666… and ξʹ = 1.41421356237…
In the decimal expression for 2/3, the symbol “…” indicates

a continuation of the pattern (an unending string of 6s). In the
decimal expression for ξʹ, the symbol “…” just indicates some
unending string of digits, whose pattern you might not
understand. We are lead to a precise definition of a more general
type of number:
DEFINITION: A real number means a “decimal expression”; that

is, an expression formed from an integer followed by a decimal
point followed by infinitely many digits. The set of all real
numbers is denoted R.
A real number like 4/5 = 0.8 should be appended with an

unending string of zeros (4/5 = 0.800000…) so that is has
infinitely many digits after its decimal point, as required by the
definition. The purpose of this requirement is to put all real
numbers on equal footing, so that pairs of them can be more
easily added or multiplied.
Speaking of which, you might be embarrassed to learn

that you do not really know how to add or multiply real numbers.
For example, what is 2/3 + ξʹ? Familiar rules of addition require
you to begin at the right-most digit and work left, but the above
decimal expressions for 2/3 and ξʹ go on indefinitely to the right.
There is no right-most digit at which to begin!
Precisely defining real number addition, multiplication,

subtraction, and division involves technical intricacies that are
beyond the scope of this book. We will settle for discussing only
one important subtraction problem:
12.75000000000… – 12.74999999999… = ???
Can you guess the answer? The difference between the red and
green number is clearly less than 12.75 – 12.74 = 0.01 and less
than 12.750 – 12.749 = 0.001, and less than 12.7500 –
12.7499 = 0.0001, and so on. Since this difference is less than
arbitrarily small numbers, the only reasonable guess is that the
difference equals zero. In fact, zero is the correct answer! But if
the difference between two numbers equals zero, then those
numbers must equal each other:
12.74999999999… = 12.75000000000…
In case you have trouble believing this, a second purely

algebraic way to see that these two numbers are the same is
found in the exercises at the end of this section. The red and
green decimal expressions are different ways of writing the same
Real Numbers 155
real number, just like 2/3 and 4/6 are different ways of writing the
same rational number. Here is the general rule:
REAL REDUNDANCY RULE: A digit (other than 9) followed by

an unending string of 9s can be replaced by the next larger digit
followed by an unending string of 0s. There are no other
redundancies among real numbers.
The strings of 0s and 9s are even allowed to straddle the

decimal point. So if a car dealer quotes you $19,999.99…, then
you will be making out your check for exactly 20,000.
The Real Redundancy rule is green-boxed because it is a

definition, not a theorem. More precisely, the rule is an
addendum to our definition of a real number. Thus, R means the
set of all decimal expressions, with the understanding that certain
decimal expressions are really the same as others.
Our “decimal expression” definition of a real number fits

nicely with our visual intuition that the real numbers should
represent all possible (positive and negative) lengths; that is, a
real number should be a point on an idealized (infinitely thin) line,
which we call the “real number line,” pictured here:
The digits of a real number tell us how to locate it on this

number line. For example, to locate ξʹ= 1.41421356237…, the
first digit “1” tells us to look between 1 and 2. If we divide this
interval into 10 equal bins (numbered 0–9), the next digit “4” tells
us which bin to look in. If we subdivide that bin into 10 equal sub-
bins, the next digit “1” tells us which sub-bin to look in, and so on.
Here is the picture:
Each next digit provides a ten fold increase in the

accuracy with which we know the number’s location on the real
number line. In fact, if you were asked to find ξʹ without a
calculator, you would probably identify it digit-by-digit like in the
above picture. First, ξʹ lies between 1 and 2 because 1 squares
to less than 2, while 2 squares to more than 2. Next, ξʹ lies
between 1.4 and 1.5 because 1.4 squares to less than 2, while
1.5 squares to more than 2, and so on.
Which Real Numbers Are Rational?

Every rational number is a real number because long
division can be used to convert any fraction into a decimal
expression. For example:
3/7 = 0.428571428571428571… (6-digit string repeats)

W
Whicch R
Real Num
mbe
ers A
Are R
Ratiiona
al? 15
57
T
Thee long divvisiion woorkk off finndinng
this a anssweer is picctured on the
d o e rrigh ht.
Nooticce tthat th
he dig
gits staarte
ed repea atin
ng
as so oon
n aas tthee oranngee num mbe ers (th he
remmaaindderss) staarte
ed re epeatin ng. S Sincce
the
eree are on even posssib
nly se bilittiess fo
or
this ooranngee re
emain ndeer (0
0–66), you kkne ew
in ad dvaancee tthat itt w
wouuld rep pea at afte er
noot m
more th hann seveen steepss.
C
Cou uld we e work
w k th his lon
ng ddivisioon
prooblemm ba ackw waardss? T Thaat is, if
N=0 0.42
285571428 857 714
4288571… haad
beeen given n, cou uld we h havve figu e out
ure
at N = 3
tha 3/7? ? HHerre is a cle eveer ttricck.
Sinncee N ha as a 6-d digit re
epe g sttring,
eating
wee w
will mu ply it by 1,000
ultip 0,0
000 (wwhicch ha as ssix ze
eross), an
nd ttheen
subtrractt N fro
om 1,0 000
0,0000 u N too re
emove e th
he reppea
ating strin
ng, likke
this:
1,,000,0 0uN=4
000 428,57 428
71.4 857 428
714 8571…
N= 0
0.4
428 285
857142 1… Å
571 ubtrractt
Åsu
99
999 9uN=4
999 428
8,57
71.0 000
000 000
000 00…
000
We earrn thatt N = 4
e le 8,571//99
428 99,9
999
9 (w
which red
duccess to
o N=
=33/7)).
THHEOORREM M: A rreall nu
ummber iss raatio
ona
al preccise
ely wh
hen
n itss d
decima
al
exxpre
esssion
n iss evven
ntua
allyy re
epeatin
ng.
Thhe ttermm ““evventtua ally repeaatin

ng”” means tha at, po
ossiiblyy affterr so
omme
inittial digitss, tthe tail o
of the de
ecim
mall exxprresssion iss fo orm
med d frromma an
definitte rrepetittion
ind n off a sin ng of dig
ngle finitte sstrin gitss. F
For exam mplee,
122.34 45992149 9214 4922144922144… iss evven ntuallyy re
epeeatting
g, aand d sso iis
the
e rratiional nu 4/5 = 0
umberr 4 0.8
8 b
beccause it getss rrew
writtten a
as
0.8
800
0000…
PROOF: The previous examples show how to convert the

fraction 3/7 into an eventually repeating decimal expression, and
how to convert the eventually repeating decimal expression
0.428571428571428571… into a fraction. Look back at these two
examples and convince yourself that the same techniques would
work to convert any fraction into an eventually repeating decimal
expression, and any eventually repeating decimal expression into
a fraction. If you are not yet convinced, try some practice
problems from the exercises……...……..……………….…………Ƒ
Because of this theorem, it is easy to find irrational

numbers (real numbers that are not rational). Just design a
decimal expression whose digits follow a pattern that’s more
complicated than “eventually repeating.” For example, the
following number is irrational:
N = 0.01001000100001000001… (pattern continues).
Try to create your own examples of irrational numbers. Although

it is easy to make up irrational numbers, it is often difficult to
prove that particular numbers are irrational. For example, the
famous numbers ߨ and ݁ are both irrational, but proving this is far
beyond the scope of this book. It is still unknown whether ߨ ξଶ is
rational or irrational, although most everyone expects it to be
irrational.
How Many Primes Are There?
The prime numbers played a very minimal role in our story

about symmetry. In case you missed their relevance, here is a
quick recap of our backtracking progress. To precisely define
rigid motions of the plane and space, we must precisely define
the plane and space, which requires us to precisely define real
numbers, which requires an understanding that not all numbers
How Many Primes Are There? 159
are rational. The simplest provably irrational number is ξʹ, and

this proof requires an understanding of prime factorization.
In this final section, we embark on a short excursion for

readers who desire to learn a few more fundamental and beauti-
ful facts about prime numbers before returning to symmetry.
More specifically, we will explore the question: how many

prime numbers are there? We take it as self-evident that there
are infinitely many natural numbers. That is what the “…” meant
when we wrote N = {1, 2, 3, 4, 5, 6, …}. But are there infinitely
many primes? This is less obvious. After all, there are only finitely
many Lego block shapes, but yet there are infinitely many
different Lego constructions that could be built from unlimited
supplies of them. It is similarly conceivable that there are only
finitely many prime numbers out of which all of the infinitely many
natural numbers can be built. This matter was settled by Euclid
around 300 BC.
EUCLID’S THEOREM: There are infinitely many prime numbers.
PROOF: Imagine that your sister’s boyfriend Andy insists that

there are only finitely many prime numbers, and to prove it, he list
all of them together on a piece of paper: p1, p2, p3, …, pn. How do
we know that Andy is wrong? No matter how lengthy his list, we
will describe a strategy for identifying a prime number that is not
on his list. Thus, no finite list of prime numbers could ever be
complete.
Here is how we will identify a prime number that is missing

from Andy’s list (or from any finite list of prime numbers). First
compute the result of multiplying all of Andy’s primes together
and adding 1. Call this number “L” because it is so large:
L = p1 u p2 u p3 u … u pn + 1.
Notice that L is not on Andy’s list because it is much larger than

anything on his list. If L happens to be prime, then it is exactly
what we seek: a prime number that is not on Andy’s list. If L
happens not to be prime, then any single prime number that
appears in L’s prime factorization is exactly what we seek: a
prime number that is not on Andy’s list. This is because none of
Andy’s primes divide evenly into L. In fact, L was custom built so
that each of Andy’s primes p1, p2, p3, …, pn leaves a remainder 1
when you divide it into L. Thus, no finite list of prime numbers
could ever be complete, which means there must be infinitely
many prime numbers …………………………..…..………………..Ƒ
Here is a quick example to help you understand the logic
of the above proof. Suppose Andy’s list is {2, 3, 5, 7, 11, 13}. We
consider L = 2 u 3 u 5 × 7 × 11 × 13 + 1 = 30,031. It turns out that
L is not prime; its prime factorization is 30,031 = 59 × 509. The
numbers 59 and 509 are primes missing from Andy’s list.
So now we know that there are infinitely many primes, but
we still might wonder how frequently occurring the prime numbers
are among the natural numbers. Are primes in abundance, or are
they a rare breed? Do most US citizens have a prime social
security number, or very few? To warm up to questions like
these, let us make a list of all natural numbers between 1 and 10,
and highlight the primes, like this:
1, 2, 3, 4, 5, 6, 7, 8, 9, 10.
Notice that 40% of the numbers on this list are prime, which we
will write as 0.40. Next we will go all the way to 100, like this:
1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20,
21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31, 32, 33, 34, 35, 36, 37,
38, 39, 40, 41, 42, 43, 44, 45, 46, 47, 48, 49, 50, 51, 52, 53, 54,
55, 56, 57, 58, 59, 60, 61, 62, 63, 64, 65, 66, 67, 68, 69, 70, 71,
72, 73, 74, 75, 76, 77, 78, 79, 80, 81, 82, 83, 84, 85, 86, 87, 88,
89, 90, 91, 92, 93, 94, 95, 96, 97, 98, 99, 100.
How Many Primes Are There? 161
Notice that 25% of the numbers on this list are prime, which we
will write as 0.25. As you travel further and further down the list of
natural numbers, prime numbers occur in these proportions:
N Fraction of numbers up to N that are prime

10 0.40
100 0.25
1,000 0.168
10,000 0.1229
100,000 0.09592
1,000,000 0.078498
The question is: what is the pattern in this table, and how
does the pattern continue for larger and larger choices of N?
Depending on how accurately you intend to answer this question,
it is either difficult or extremely difficult or worth a million dollars.
The easiest answer is this: as N increases, the fraction

approaches zero. The more difficult issue is specifying how
quickly the fraction approaches zero. Here is a famous theorem,
which was conjectured in the early 1800s and not successfully
proven until 1896:
THE PRIME NUMBER THEOREM: The fraction of numbers up to

ଵ
N that are prime is approximately .
twice the number of digits in N
This theorem is difficult to prove, but it is easy to use. For

example, what fraction of numbers between 1 and 1,000,000 are
prime? Since there are seven digits here, the answer is
ଵ
approximately ൎ ͲǤͲ͹ͳͶ, which is close to the exact value from
ଵସ
the table. Thus, about 7% of the numbers less than a million are
prime.
162
2 10. Wh
hat Iss a N
Num
mber?
O
Ourr veersion n off th
he Priimee NNummbe er Theorrem m aaboove is no ot
the
e mmosst a
acccuraate
e ve ersionn. F
For reeaderss faamiiliarr w
with ogarith
h lo hmss,
weemmenntio
on tthe more eaaccu ura
ate annd mmuch mo ore
e faamo ouss verssion
n,
whhich
h sa his: Th
ayss th he fractioon of num mbberss up
p to
o N th hat areepprim
me iis
ͳ
apppro
oxim
maatelyy .OOur ccrudde ve ersion ab bovve followws froom thiis

moore
e acccuratte vverssion bbeccausse::
twice the number digits of N ൎ ʹ ൈ ൎ ͳͲ ൈ = Ǥ
Ouur ccrude ve ersion saacrifice

es som
mee acccu
uraccy (pa
artlyy b
beccausse 2 iis
noot vveryy cclosse to ͳͲ ൎ ʹǤ͵Ͳʹ ǥ)), bbut it ha
as tthe
e ad dvaanttage
e o
of
beeingg uund andablle to re
derssta ead who are
derss w e nnott faam
milia
ar witth
log
garrithmms.
E
Eveen tthe
e lo ogarith hm ve ersioon of theeP Primme nu umb berr th
heo
oremm
can b be fin ned an
ne ttun nd impro ove ed on.. In
n faact, thhe questt to o m
morre
preecissely un ndersta and d exa acttly hoow th he prim me nu umberrs arre
disstrib
butted ammong the e nnatu ura
al nnum
mbeers is intterttwin nedd wwithh so
omme
of the e m
mosst ddiffficult ttheore emss inn mmatthema aticcs a
andd also o soome o of
thee mo ost innfam mous unso olved p probblemss. These e incclud de thhe
Rieemmann H Hyp
potthe esiss, w which todayy sstan ndss a as tthe
e m
mosst ffam mouus
unnsolved m matth pro oble em, w with h a onne mmilllion
n dollaar pprizze proomiseed
to thee peersson wh ho firsst ssolvves it.
Exxerrcissess
C verrt the folllow
(1)) Con wing ffracctio
ons intto d
deccim
mal expre
esssion
ns: 5/7
7,
23
3/21 1, 1
14/1
13.
(2)) Convverrt th
he ffollow
wing
g de
ecima
al e
expresssio
onss intto ffracctio
onss:
10
0.79
929
929 92…, 5
929 5.2
200
030030
003 03…, 0.3
300 344 4…
444
Exercises 163
(3) Prove that ξ͵ is an irrational number; in other words, there is

no rational number that squares to 3. Hint: Copy the proof in the
chapter that ξʹ is an irrational number.
(4) Prove that (5+ξʹ)/7 is an irrational number. HINT: If you could

write it as a fraction, how could you use this to write ξʹ as a
fraction?
(ٍ5) Prove that ξʹ+ξ͵ is an irrational number. HINT: Assume it

is rational, which means ξʹ+ξ͵= p/q, then square both sides of
this equation and find a contradiction.
(6) Find a rational number between M = 15.235950… and

N = 15.237146… (to answer this, you do not need to know any
more of the digits of M and N than are shown). Can you find
infinitely many different rational numbers between M and N?
What about between M = 15.236950… and N = 15.237146…?
(7) Find an irrational number between M = 15.235950… and

N = 15.237146… (to answer this, you do not need to know any
more of the digits of M and N than are shown). Can you find
infinitely many different irrational numbers between M and N?
What about between M = 15.236950… and N = 15.237146…?
(ٍ8) Prove that there are infinitely many different rational

numbers and infinitely many different irrational numbers between
any pair of distinct real numbers.
(ٍ9) In your solution to the previous problem, why does your

proof not work when the two real numbers are equal, like
M = 12.7499999… and N = 12.750000…?
(10) What is the smallest rational number larger than 0? If you do

not think there is one, then explain why. What about the smallest
irrational number larger than 0?
(ٍ11) What can you say about the decimal expression for a
fraction whose denominator has two digits? What is the longest
the repeating string could possibly be? What is the longest
possible string of initial digits before the repeating string begins?
(12) In this chapter, we saw that N = 12.74999999999… equals

12.75. Reprove this claim in a purely algebraic manner by
converting N to a fraction, and then showing that this fraction
equals 12.75.
(13) To intelligently understand political news, you often need a

frame of reference for comprehending big numbers. Filling in the
following blanks (using internet resources as needed) may help
you develop such a frame of reference:
A billion equals _______ million. The US population

equals about ______ million. If the US government spends a
billion dollars, this amount averages to about $______ per
citizen. A trillion equals _____ billion. If the government spends a
trillion dollars, this amount averages to about $______ per
citizen. The US national debt is about $______ trillion, which
averages to about $________ per citizen. The most recent
national deficit equals about $______ billion, which averages to
about $_______ per citizen. The difference between the
meanings of the words “debt” and “deficit”
is:_________________________________________________.
(14) About what fraction of natural numbers between 1 and one

trillion are prime? About how many natural numbers between 1
and one trillion are prime?
(15) About what fraction of US citizens have a prime social

security number? About how many US citizens have a prime
social security number?
Exercises 165
(16) Suppose that X and Y are real numbers. For X, the third digit
after its decimal point equals 5. For Y, the third digit after its
decimal point equals 7. You do not know anything else about X or
Y. In other words, X and Y look like this:
X = *.**5**** … and Y = *.**7****…
From this information, can you conclude that X and Y are

DIFFERENT real numbers? What if X and Y instead look like this:
X = *.**9**** … and Y = *.**0****…
State a general rule for comparing two numbers which differ at a

single decimal position.
(17) If X and Y are real numbers which differ at a single decimal

position, and Y does NOT end in an infinite string of 9s or an
infinite string of 0s, explain why X and Y must be different real
numbers.
ͳͳǤǯ
Thus far, we have introduced the following important sets
of numbers:
N = {1, 2, 3, 4, 5, 6, …} “the natural numbers”
Z = {…, –3, –2, –1, 0, 1, 2, 3, …} “the integers”
Q = {all quotients “a/b” of integers with b 0} “the rationals”
R = {all real numbers} “the real numbers”
Which of these sets is the largest? You might respond that R is

the largest because it contains the others. Or you might respond
that they all have the same size, namely infinity. Until a little more
than a century ago, mathematicians were content with the
decision that every infinite set has the same size as every other
infinite set. They were not right or wrong – this is simply what
they meant by the phrase “same size.”
OLD-FASHIONED DEFINITION OF “SAME SIZE”: A pair of sets

is said to have the same size if either (1) they are both finite and
have the same number of members or (2) they are both infinite.
This definition probably seems reasonable, but you are

about to learn a beautiful truth about infinity to which this
definition blinds you. Mathematicians who used this definition did
not understand their blind spot any more than the ancient Greek
mathematicians understood the truths to which they were blinded
when they defined “number” to mean “rational number.” In the
history of mathematical thought, this “infinity” blind spot was just
as significant as the “number” blind spot, and its removal
unleashed a rich world of fundamentally new ideas.

168 11. Cantor’s Infinity
The Modern Meaning of “Same Size”

What else could the phrase “same size” possibly mean?
To answer this question, let us think more carefully about how we
compare the sizes of sets. When my niece was a toddler, I gave
her ten candles and ten candle holders, and I asked her whether
there were as many candles as candle holders. An adult would
have separately counted the candles and holders and compared
the answers, but my niece did not yet know how to count to ten.
So instead, she simply placed one candle into each holder. Since
the candles and holders matched up perfectly, she knew there
were equal numbers of each.
If you are given two infinite sets and asked whether they
have the same size, then your situation is very analogous to my
niece’s. You do not have the ability to separately count each set
because you do not know how to “count to infinity.” Your most
reasonable solution is the one my niece used – you should try to
find a one-to-one correspondence (a matching) between the
members of the two sets. This idea is not child’s play – it is so
important, it will become our new meaning of “same size.”
MODERN DEFINITION OF “SAME SIZE”: A pair of sets is said

to have the same size if their members can be matched with a
one-to-one correspondence.
It is time to forget the old-fashioned definition, and from

now on, use only the modern definition. To decide whether two
sets have the same size, your only job is to determine whether
their members can be matched with a one-to-one correspondence.
For example, to decide whether you have the same number of
fingers as the stranger who you just met at the aquarium, you
may not count and compare; rather, you must attempt a finger-to-
finger matching as in the illustration below. It is often very natural
to compare sizes by matching rather than counting. For example,
The Modern Meaning of “Same Size” 169
in a truck load of new pairs

of shoes headed to Payless,
you know that the number
of right shoes equals the
number of left shoes with-
out knowing how many of
either are in the truck. Still,
it is difficult to change old
habits into new habits, so
We have the same number of fingers.
let’s practice.
EXAMPLE (A pair of finite sets): How do we verify that the sets

S1 = {a, b, c, d, e} and S2 = {A, B, C, D, E} have the same size? If
you answered “they both have five members,” then you have not
yet let go of the old-fashioned definition. From now on, the only
way to confirm that two sets have the same size is to exhibit a
one-to-one correspondence between their members, like this:
a b c d e
Ľ Ľ Ľ Ľ Ľ
A B C D E
EXAMPLE (N and E): How do we decide if N = {1, 2, 3, 4, …}

(the set of all natural numbers) and E = {2, 4, 6, 8, …} (the set of
all even natural numbers) have the same size? Here are some
WRONG ANSWERS:
“They have the same size because they are both infinite.”
“N is larger than E because N has all of E’s members plus more.”
These answers are WRONG because they do NOT refer to the

modern definition of “same size.” If you are tempted by these
wrong responses, then you need to let go of your previous
associations with the phrase “same size” and let the modern
definition above become your ONLY meaning for this phrase. To
answer this question, your ONLY job is to decide whether

the member of N and E can be put into a one-to-one
correspondence. After some trial and error, you will find that they
can, like this:
1 2 3 4 5 6 7 8 9 …
2 4 6 8 10 12 14 16 18 …
A one-to-one correspondence between the natural numbers and the even numbers
In case the pattern is not clear, we could describe it with a

formula: n ļ 2n. Do you see why this pattern is a one-to-one
correspondence? For any even number you ask me about, I can
find the natural number that matches with it. For example, if you
ask me about 100, I report back that 50 matches with it. I will only
ever have one choice for the number I report back, because this
pattern never allows multiple natural numbers to match with the
same even number. Thus, it is a one-to-one correspondence! We
learn that N and E have the same size! How strange that an
infinite set can have the same size as a subset of itself!
EXAMPLE (N and Z): Does N = {1, 2, 3, 4, …} have the same

size as Z = {…, –3, –2, –1, 0, 1, 2, 3, …}? To answer this
question, your ONLY job is to decide whether the member of N
and Z can be put into a one-to-one correspondence. This seems
difficult at first because the members of Z extend indefinitely in
both the right (positive) and the left (negative) direction. But the
following clever matching overcomes this difficulty:
1 2 3 4 5 6 7 8 9 …
0 1 –1 2 –2 3 –3 4 –4 …
A one-to-one correspondence between the natural numbers and the integers
Do you see the pattern? The even natural numbers get matched
with the positive integers, while the odd natural numbers get
The Modern Meaning of “Same Size” 171
matched with the negative integers. In case the pattern is not

clear, we could clarify it using a formula:
(even n) ļ n/2, (odd n) ļ –(n–1)/2
Thus, N and Z have the same size!
Finding a one-to-one correspondence can require clever-

ness and persistence. In the previous example, you might have
first tried the matching n ļ n, but then realized that this matching
misses all of the negative members of Z. But just because one
attempted matching fails, you can NOT conclude that the sets
have different sizes – a cleverer attempt might still succeed.
DEFINITION: An infinite set is called countable if it has the same

size as N (the set of natural numbers).
In the previous two examples, we concluded that E and Z

are both countable. In general, to prove that an infinite set is
countable, you must match its member with the natural numbers.
That is, you must decide which 1st member of your set matches
with 1, which 2nd member matches with 2, which 3rd member
matches with 3, and so on. For E and Z, our matching looked like
this:
1st 2nd 3rd 4th 5th 6th 7th …

E 2 4 6 8 10 12 14 …
Z 0 1 –1 2 –2 3 –3 …
In summary, to prove that an infinite set is countable, we must

find an infinite listing of its members, {1st, 2nd, 3rd, … }, which is
organized so as to eventually include each member.
Are the Rational Numbers Countable?

Does Q (the set of all rational numbers) have the same
size as N? In other words, is Q a countable set? If you believe
that Q is countable, then to prove it you must find an infinite
listing {1st rational, 2nd rational, and 3rd rational,…} organized so
that your list eventually includes each rational number. Here is a
first attempted pattern:
1/1, 1/2, 1/3, 1/4, 1/5, …
This attempt fails because it only includes positive fractions

whose numerators equal 1. Let us improve this attempt by
squeezing in more numerators:
1/1,1/2, 1/3, 2/3, 1/4, 2/4, 3/4, 1/5, 2/5, 3/5, 4/5, …
n removed because 2/4 = 1/2
This attempt is better – it eventually includes all positive fractions

whose numerators are smaller than their denominators. Next, let
us squeeze in their reciprocals:
1/1, 1/2, 2/1, 1/3, 3/1, 2/3, 3/2, 1/4, 4/1, 3/4, 4/3,
1/5, 5/1, 2/5, 5/2, 3/5, 5/3, 4/5, 5/4, …
The pattern here is: 1st fraction from the previous list, then its
reciprocal, then the 2nd, then its reciprocal, and so on. This new
pattern is better still – it eventually includes all positive fractions.
All that remains is to insert 0 at the front, and intersperse the
negatives:
0, 1/1, –1/1, 1/2, –1/2, 2/1, –2/1, 1/3, –1/3, 3/1, –3/1,
2/3, –2/3, 3/2, –3/2, …
The pattern here is: zero, then the 1st fraction from the previous
list, then its negative, then the 2nd, then its negative, and so on.
Behold the power of trial and error! We just proved:
Are the Rational Numbers Countable? 173
THEOREM: The set of rational numbers, Q, is countable.
The previous proof is perfectly valid, but the following alternative

proof is also important to understand.
ALTERNATIVE PROOF: Here is an alternative method for listing

all of the positive rational numbers. This means organizing them
into an infinite list, but we first settle for organizing them into an
infinite grid which (like a computer spreadsheet grid) has a top
edge and left edge, but extends indefinitely down and right. The
most natural arrangement is like this, with the column determining
the numerator and the row determining the denominator:
1 2 3 4 5 …
1 1/1 2/1 3/1 4/1 5/1 …
2 1/2 2/2 3/2 4/2 5/2 …
3 1/3 2/3 3/3 4/3 5/3 …
4 1/4 2/4 3/4 4/4 5/4 …
5 1/5 2/5 3/5 4/5 5/5 …
: : : : : :
Now we organize the cells of this infinite grid into an

infinite list by meandering through the grid like this:
If we record the fractions we visit along this meandering purple

path, and remove the redundant ones as we go, our list will begin
like this:
1/1, 2/1, 2/2, 1/2, 1/3, 2/3, 3/3, 3/2, 3/1, 4/1, 4/2, 4/3, 4/4, 3/4,…
Now that we have successfully listed all of the positive rational

numbers, we can insert zero at the front and intersperse the
negatives as before…………………………………..……………….Ƒ
Cantor’s Theorem
Our next goal is to decide whether R (the set of all real
numbers) is countable. To appreciate the question, try to
construct an infinite listing {1st real, 2nd real, 3rd real, …}. You
might start with a listing of the rational numbers and then insert
some famous irrational numbers like ߨ and ξʹ at the front of your
list. But what about the less famous irrationals, like the ones you
made up yourself in the last chapter? The more you add to your
list, the more you discover is missing. Are there too many real
numbers to squeeze into a single infinite list? The answer to this
difficult question was discovered by Georg Cantor around 1872.
CANTOR’S THEOREM: The set of real numbers, R, is NOT

countable (so we call it uncountable).
I know lots of ways to construct an infinite listing of real

numbers that fails to include them all. But this does not prove
Cantor’s theorem, since someone cleverer than me might some-
day succeed in including them all. To prove his theorem, Cantor
had to show that NO listing, no matter how cleverly constructed,
could ever succeed in including all real numbers. In other words,
he had to prove that every attempted listing is doomed in
advance. Here is how he did it:
PROOF: We will prove that any listing of real numbers is

incomplete. No matter how scrupulously the list was organized,
some real numbers were definitely left off. More precisely, we will
describe a concrete procedure for identifying a real number that
is missing from any given listing of real numbers.
Cantor’s Theorem 175
Imagine a listing of real numbers. Maybe it was created by

your Aunt Clair, who tried her best to include all of the real
numbers on her list. Maybe it begins like this:
1st ļ 3.1415926635… (ߨ)

2nd ļ 0.3333333333… (1/3)
3rd
ļ 1.41421356237… (ξʹ)
4th ļ 256655643.00000000000… (Aunt Clair’s SSN)
5th ļ 509.73737373737… (Her favorite number)
6th ļ 5.04749726737… (ߨ ξଶ )
Here is a concrete procedure for identifying a real number

that is missing from the list. We will call this missing number M. It
will lie between 0 and 1, so it will have the form
M = 0.d1d2d3d4d5d6d7d8…
where each dn is a digit (0-9). How should we choose these digits

to insure that M is NOT on the list? The answer is ingenious, and
is hinted at by the red digits in Aunt Clair’s list. Here it is: Choose
M’s first digit, d1, to be anything other than the first digit (after the
decimal point) of the first number on the list. This insures that M
is different from the 1st number on the list, since it has a different
first digit. Choose M’s second digit, d2, to be anything other than
the second digit of the second number on the list. This insures
that M is different from the 2nd number on the list, since it has a
different second digit. Do you see the idea? Chose M’s nth digit,
dn, to be anything other than the nth digit of the nth number on
the list, which insures that M is different from the nth number on
the list, since it has a different nth digit.
In the Aunt Clair example, the red diagonal includes the

numbers {1, 3, 4, 0, 7, 7, …}, so we must choose
M = 0.(not 1)(not 3)(not 4)(not 0)(not 7)(not 7) ….

This leaves us a lot of freedom. M = 0.258163… works fine, as

would many other choices. With each digit, there are ten choices
(0–9), and only one choice is disallowed, which still leaves us
nine options. To be on the safe side, we will also avoid 0s and 9s,
which still leaves at least seven options for each digit. See
Exercise #17 from Chap. 10 to understand the reason for
avoiding 0s and 9s.
In summary, we can use this diagonal procedure to build a

real number, M, which is missing from any given listing of real
numbers. Therefore, no listing of real numbers could possibly be
complete. Thus, the real numbers could never all be arranged
into a single list – they are uncountable........................................Ƒ
Cantor’s Theorem says that, in a very precise sense, the

infinite sets N and R do NOT have the same size. Thus, the
modern definition of “same size” leads to this truth: not all infinite
sets have the same size – some are genuinely larger than others!
This is a surprising and remarkable phenomenon. In popular
writing, it is described with phrases like “different sizes of infinity”
or “more infinite than infinity.”
During Cantor’s life, his work was

criticized by theologians who considered it
a challenge to the notion of God as the
one and only infinite and also by mathe-
maticians who were uncomfortable with his
counterintuitive conclusions. But in the end,
you can’t argue with a solid proof. Cantor’s
conclusions were eventually accepted,
causing a paradigm shift in the way
mathematicians thought about fundamental
concepts such as numbers and sets. The Georg Cantor
E
Exerrcise
es 17
77
fam
mous math hem
matticiaan Da avid
d HHilb
berrt p
preddicted
d th
he lon
ng--lasstin
ng
importan
nceeo
of C
Can ntorr’s wo
ork whhenn wrote e: ““No
ooone sh
hall exxpe
el uus
fro
om the
e Para
adisse w
whhichh Canttor ha
as ccrea
ateed.””
Exxerrcissess
(1)) You
Y r brotherr in aw has a lisst o
n la 4-digit nu
of ffour 4 umbberrs, butt hiis
andwriiting
ha g iss sso p
pooor, youu ccan on nly ma e out o
ake oneeddigitt off ea
achh:
9*, ****0. Witho
3****, *7***, **9 out kn nowwing gaany mo ore
e in
nforrmaatio
on, ca an
you ffind
d a 4-d digit nnummbeer tthat iss no
ot oon your bro othe
er in llaw
w’s listt?
Hoow is tthiss prrob blem
m relaated
d to C tor’s p
o Cant proof?
?
(2)) Dec
D ide e whettheer th
hesse setts hhavve thee sa
am e: S1 = {all odd
me ssize
na
aturral n
nummb
berss} a
andd S2 = {alll natu
urall nu
umberrs ggre
eateer than
n99}.
(3)) Iss itt posssible to list a

all o
of the
e p nal nu
possitivve ration umb
berrs iin
inccreaasing ord derr?
(ٍ4)) If a sspre
eaddsh
hee et g
grid exxten ndeed ind
definite
ely lefft, rrigh
ht, u
up annd
do
ownn, w
wou uld its cells be co ounntabble?
?HHintt: C
Can
n yoou ffind d a paath tha
at
meean
nde ers thrrou
ugh alll off itss ce
ells?
(ٍ5)) P Provve tha at the e set of irrratio onaal num mbberss iss u
unccounta
ablee.
HINT T: IIf iit w
werre cou unttabble, th hen
n in nterrsp
perssing
g a llistiing off a
all
irra
atio
ona
al n nummbeerss with
w a lisstin
ng of all ra atio
onall n
num mbeers wooulld
prooduuce
e a listting
g off alll re
eal nu
umb berrs.
(ٍ6)) If you rreplacce tthe wo ord

d “re
eall” wwith
h “rratio
onaal” thrrougghooutt th
he
proooff off Ca
anttor’ss T
The
eore
em, yoou geet a faultyy proo of th
hatt the rratio
onaal
nuumbbers are e u unccouunta
able
e. Why
W iss itt fa
aullty?
? WWh hichh sstep p iis
inccorrecct?
(7) Is the set of all real numbers between 0 and 1 countable or

uncountable? Why?
(8) Is the set of all real numbers whose decimal expressions

contain no 5s countable or uncountable? Why?
(9) Is the set of all pairs of natural numbers countable or

uncountable? Why?
(10) You have infinitely many piles, one for each natural number.
The first pile has one marble, the second has 2 marbles, the third
has 3 marbles, and so on. Is the total number of marbles a
countable set?
(ٍ11) You have infinitely many piles – one for each natural
number. Each pile has infinitely many marbles – one for each
natural number. Is the total number of marbles a countable set?
(ٍ12) If you remove one member from an infinite set, will the
new set always have the same size as the original set? Hint: first
consider the case when the original set is countable.
(ٍ13) Which of the following objects have countable symmetry

groups: a circle in the plane, a sphere in space, the O border
pattern. Explain your answers.
ͳʹǤ
ʹ

T
Thiss bboook is aabo out tw
wo-d dim
men nsioonaal o
obje
ectts in tthe
e pllane
annd tthre
ee--dimmennsionaal oobjectts in sspa acee, a
and he wa
d th ays in whicch
such ob bjeccts ca
an be syymm meetricc. B
Butt what
w t do tthe e te
erm
ms ““plaane
e”
annd ““space e” meean n? T
To so olidiify and tthe found dattion
no on wwhiich ou
ur
the
eorry oof ssym
mmetryy reestts, wee will nnowwp preccise
ely de ne tthe wo
efin ord
ds
“pllane”, “sppacce,”” and “rig
gid mootioonss” th
herreo Let’s begin.
of. L
DEEFINITTIOON:: n n-ddimenssionall E
Eucclid
dea
an sp pacce, deno oteed Rn,
meeanns tthe
e se
et o
of a
all o
orde
ere
ed n
n-tu
uple
es of rea
al n
num
mbe ers.. R2 iss ca
alle
ed
3
the
e plan
ne, and R is ccalle ed spacee.
T
The e, R2, is thu
e plane us tthe
e se
et o
of a
all o
ordere
ed 2
2-tu
uples (pa
airss)
eal num
of re mbeers, like ሺͳǡ Ȃ ͹ሻ
andሺξʹǡ ߨሻ. The illustration on
e rrigh
the ht ssho
owss sseve eraal p
poinnts
2 2
of R . A g gennera al p
point of R is
deenotedd ሺ‫ݔ‬ǡ ‫ݕ‬ሻ . No oticce ho oww
the
ese wo real num
e tw mbberss re ecoord
the
e loca on of a po
atio ointt inn tthe
pla
anee. T
Thee firrst numb ber,, ‫ ݔ‬, de
es--
cribess its ea astt–wwest p possitio
on,
whhile
e th
he se eco ond
d nnum mbe er, ‫ ݕ‬,
deescribees itss no
orth h–ssouuth
po
osition..
S
Spa e, R3, m
ace meeans th he sett off all orrde
ered
d 3-tuple
es ((trip
pless) o
of
real numbers, like ሺͳǡ Ȃ ͹ǡͷሻ and ሺξʹǡ ߨǡ Ȃ ͳͺሻ. A general point of
R3 iss deno ed ሺ‫ݔ‬ǡ ‫ݕ‬ǡ ‫ݖ‬ሻ. N
ote Notice hoow th
hese tthre
ee re eal nu
umbberrs
de
escrribe
e th
he loccatio
on of a poin
nt in space.. Th
he firsst num
mbeer, ‫ ݔ‬,

180
0 12. Eucclide
ean Spa
ace
deescrribees its ea astt–

we est posiition, th he
secon nd nu umb berr, ‫ ݕ‬,
deescrribees its noorthh–
soouthh ppositioon, an nd
thee th
hirdd num mbeer, ‫ ݖ‬,
deescribe es itss up p–
doown n pposiition. Th he
illu
ustrratiion oon th he
3
lefft sh
how
ws the
e lo
oca
ation
noof th
he poiint ሺʹǡʹǡ͵ሻ in
nR .
R4 mmeeans th he se et o
of a
all o
ordere
ed 44-tuplles ሺ‫ݔ‬ǡ ‫ݕ‬ǡ ‫ݖ‬ǡ ‫ݓ‬ሻ of rea
al
numbers, for example ሺξʹǡ ߨǡ Ȃ ͳͺǡͳͻሻ. Although high dimensional
Eu
uclid
deaan space es like e R4 ccan
nnoot b
be vvisu
ualize
ed aas w we as R2 an
ell a nd
3
R , th
heyy a
are sttill of pra actica
al im
mportancce. For exxam mple e, a ffoo
od
ma ufaccturerr who
anu w reecordss his ssales of se eve
en d ent prroductts
difffere
7
ea
ach weeekk is reallyy re
ecoording a ppoint o
of R .
The
T urpose
e pu eo exp
his chaptter is to e
of th plorre tthe
e su
ubtlleties and
2
acies off Euclide
inttrica eann spaccess, p ularly R a
parrticu and R3, and d tto
deefine mot
e “rrigid m ns”
tion of Euuclid an
dea spaacees.
Th
he Pyythag
gorran
n The
eore
emma
and
d Dist
D tannce
eFForrmu
ula
a
T
The
e most
m t im mpo orta
ant facct a
abo out the pplan ne, R2,
waas disscoverredd byy tthe Pyythagore eanns – tthe sa ame
gro p of a
oup anciennt Grreek ma ath
hem maticia
anss w who
overed
dissco d th nal.. It is a rellatio
hat ξʹ iss irrrattion onsship
be
etwe een
n th
he llen
ngth
hs o dess off a right ttria
of tthe sid angle:
THHE PY YTHHAG GOORE

EAN T THEO ORE EM: TThe len
ngtths, a,, b and c off th
he
sid
dess of a rightt trrian
ngle
e ((listted
d in
n in
ncre
eassing
g o
ord
der)) saatissfy th
he
2 2 2
rellatio
onsship: a + b = c .
The Pythagoran Theorem and Distance Formula 181
For example, if a = 3 and b = 4, then c2 = 32 + 42 = 9 + 16 = 25,

so c = ξʹͷ = 5.
PROOF: For the proof, you

will need some supplies.
Cut out four copies of the
same right triangle. We will
call its side lengths a, b,
and c (in increasing order).
Next, cut out a square
whose side length is a + b.
These supplies are pictured
in the illustration on the right.
The key observation is that there are two very different

patterns with which to arrange the four triangles onto the square.
These two patterns are
shown in the left illus-
tration. For the first pattern,
the uncovered orange
area equals a 2 + b2
(because it is made of
two squares with side
lengths a and b). For the second pattern, the uncovered orange
area equals c2 (because it is made of a single square with side
length c). Since rearranging the pattern could not change the
uncovered area, we conclude that a2 + b2 must equal c2. If we
had started with a fatter or narrower right triangle, think about
why the two arrangements still work out, with corners meeting
perfectly as in the illustration above, and with right angles at the
corners of the orange c-by-c square…..……………………………Ƒ
Why is the Pythagorean Theorem important? Because it

allows us to measure distances between points in the plane:
182 12. Euclidean Space
DISTANCE FORMULA FOR THE PLANE: The distance between

ሺ‫ݔ‬ଵ ǡ ‫ݕ‬ଵ ሻ and ሺ‫ݔ‬ଶ ǡ ‫ݕ‬ଶ ሻ in R2 equals ඥሺ‫ݔ‬ଶ െ ‫ݔ‬ଵ ሻଶ ൅ ሺ‫ݕ‬ଶ െ ‫ݕ‬ଵ ሻଶ .
PROOF: This distance formula is nothing more than the

Pythagorean Theorem in disguise.
The distance between ሺ‫ݔ‬ଵ ǡ ‫ݕ‬ଵ ሻ and
ሺ‫ݔ‬ଶ ǡ ‫ݕ‬ଶ ሻ equals the length, ܿ, of
the hypotenuse (the longest side)
of the right triangle pictured
in the illustration on the left. The
Pythagorean Theorem says
ܿ ൌ ξܽଶ ൅ ܾ ଶ , which is exactly the
distance formula………………….Ƒ
The distance formula in space is analogous:
DISTANCE FORMULA FOR SPACE: The distance

betweenሺ‫ݔ‬ଵ ǡ ‫ݕ‬ଵ ǡ ‫ݖ‬ଵ ሻ and ሺ‫ݔ‬ଶ ǡ ‫ݕ‬ଶ ǡ ‫ݖ‬ଶ ሻ in R3 equals
ඥሺ‫ݔ‬ଶ െ ‫ݔ‬ଵ ሻଶ ൅ ሺ‫ݕ‬ଶ െ ‫ݕ‬ଵ ሻଶ ൅ ሺ‫ݖ‬ଶ െ ‫ݖ‬ଵ ሻଶ .
PROOF: The proof involves a

clever double application of the
Pythagorean Theorem. In the
illustration below, applying the
Pythagorean Theorem to the red
triangle gives that the distance
between the two orange points in
R3 equals:
Distance = ܿ = ξܽଶ ൅ ܾ ଶ .
Now simplify this expression for ܿ
after substituting the following:
ܾ ൌ ‫ݖ‬ଶ െ ‫ݖ‬ଵ , ܽ ൌ ඥሺ‫ݔ‬ଶ െ ‫ݔ‬ଵ ሻଶ ൅ ሺ‫ݕ‬ଶ െ ‫ݕ‬ଵ ሻଶ .

Naming the Points on the Unit Circle 183
This expression for ܽ comes from applying the Pythagorean

Theorem to the yellow triangle……………......……………….……Ƒ
Our distance formulas generalize naturally to Rn in the

most obvious manner. For example, the distance between the
points ሺ‫ݔ‬ଵ ǡ ‫ݕ‬ଵ ǡ ‫ݖ‬ଵ ǡ ‫ݓ‬ଵ ሻ and ሺ‫ݔ‬ଶ ǡ ‫ݕ‬ଶ ǡ ‫ݖ‬ଶ ǡ ‫ݓ‬ଶ ሻ in R4 equals:
ඥሺ‫ݔ‬ଶ െ ‫ݔ‬ଵ ሻଶ ൅ ሺ‫ݕ‬ଶ െ ‫ݕ‬ଵ ሻଶ ൅ ሺ‫ݖ‬ଶ െ ‫ݖ‬ଵ ሻଶ ൅ ሺ‫ݓ‬ଶ െ ‫ݓ‬ଵ ሻଶ
and so on. We do not need to prove that this formula agrees with
any previous meaning of “distance” in Rn, because when n > 3,
there is no previous meaning. We can, therefore, take this
formula as our definition of the word “distance.” For example, a
food manufacturer who records his sales of seven different
products each week might use the distance formula in R7 to
compare this week’s sales to last week’s sales.
Naming the Points on the Unit Circle

We will now take a closer look at the
circle – more specifically, the unit
circle in R2, which means the circle
with radius 1 centered at the
origin, ሺͲǡͲሻ. How many points on
the unit circle can you identify?
The easy ones to identify are its
intersections with the ‫ ݔ‬and ‫ ݕ‬axes:
ሺͳǡͲሻ, ሺȂ ͳǡͲሻ, ሺͲǡͳሻ, and ሺͲǡ Ȃ ͳሻ. Do you
know the ‫ ݔ‬and ‫ ݕ‬coordinates of any other points on the unit
circle? You could probably use the distance formula to find more,
but to really understand the unit circle, it is not enough to just
haphazardly name some points. What we really need to do is
parameterize all of the unit circle’s points in a systematic way.
Here is the idea. Beginning at the point

ሺͳǡͲሻ, you walk counterclockwise around
the unit circle. In your left hand, you
hold one end of a purple rope. The
other end of the rope is tethered to
ሺͲǡͲሻ. Consider the angle that the
positive ‫ݔ‬-axis makes with the purple
rope. As you walk one full trip around the
circle, this angle steadily increases from 0 to
360. We will call this angle your “angle-position” (because it
determines where you are) and we will denote this angle-position
with the letter “‫ݐ‬.”
The ‫ ݔ‬and ‫ ݕ‬coordinates of your position (when your

angle-position equals some value of ‫ ݐ‬between 0 and 360) have
special names:
DEFINITION: If you are at angle-position ‫ ݐ‬on the unit circle,
ܿ‫ݏ݋‬ሺ‫ݐ‬ሻ means your ‫ݔ‬-coordinate (pronounced “the cosine of ‫)”ݐ‬.
‫݊݅ݏ‬ሺ‫ݐ‬ሻ means your ‫ݕ‬-coordinate (pronounced “the sine of ‫)”ݐ‬.
Most calculators can compute the cosine and sine of any

angle-position with dedicated buttons labeled “cos” and “sin.”
Here is a table showing the cosine and sine of all multiples of 15
between 0 and 180 (the values are rounded to two decimal
places):
‫ݐ‬ 0 15 30 45 60 75 90 105 120 135 150 165 180
ܿ‫ݏ݋‬ሺ‫ݐ‬ሻ 1 .97 .87 .71 .50 .26 0 -.26 -.50 -.71 -.87 -.97 -1
‫݊݅ݏ‬ሺ‫ݐ‬ሻ 0 .26 .50 .71 .87 .97 1 .96 .87 .71 .50 .26 0
Cosine and Sine table
The Dot Product and Perpendicularity 185
What this table really tells you is the ‫ ݔ‬and ‫ ݕ‬coordinates

of the 13 points on the top half of the unit
circle whose angle-positions are
labeled in the diagram on the
right. Could you have pre-
dicted which numbers in
the table would be posi-
tive and which would be
negative? Could you Angle-positions around the unit circle
have predicted which would be close to zero and which would be
close to 1?
The Dot Product and Perpendicularity

In this section, you will learn a simple method for testing
perpendicularity. To get started, we must introduce the terms
“norm” and “dot product.”
DEFINITION: If ‫ ݌‬ൌ ሺܽǡ ܾǡ ܿሻ and ‫ ݍ‬ൌ ሺ‫ݔ‬ǡ ‫ݕ‬ǡ ‫ݖ‬ሻ are points of R3,
(1) The norm of ‫( ݌‬denoted ȁ‫݌‬ȁ) means the distance from ‫ ݌‬to the
origin ሺͲǡͲǡͲሻ, which is computed as: ȁ‫݌‬ȁ ൌ ξܽଶ ൅ ܾ ଶ ൅ ܿ ଶ .
(2) The dot product of ‫ ݌‬and ‫(ݍ‬denoted ‫ ݌‬Ȉ ‫ )ݍ‬is defined as:
‫ ݌‬Ȉ ‫ ݍ‬ൌ ܽ‫ ݔ‬൅ ܾ‫ ݕ‬൅ ܿ‫ݖ‬.
Norms and dot products of points in Rn are defined analogously,

as these examples demonstrate.
Example in R2: If ‫ ݌‬ൌ ሺ͵ǡͶሻ and ‫ ݍ‬ൌ ሺʹǡ͹ሻ in R2, then the norm
of ‫ ݌‬isȁ‫݌‬ȁ ൌ ξ͵ଶ ൅ Ͷଶ ൌ ξʹͷ ൌ ͷ, and the dot product of ‫ ݌‬and ‫ݍ‬
equals: ‫ ݌‬Ȉ ‫ ݍ‬ൌ ͵ ൈ ʹ ൅ Ͷ ൈ ͹ ൌ ͸ ൅ ʹͺ ൌ ͵Ͷ.
Example in R3: If ‫ ݌‬ൌ ሺʹǡ Ȃ ͵ǡͷሻ and ‫ ݍ‬ൌ ሺͷǡ ͹ǡ Ȃ ͳሻ, then the
norm of ‫ ݌‬is ȁ‫݌‬ȁ ൌ ඥʹଶ ൅ ሺെ͵ሻଶ ൅ ͷଶ ൌ ξ͵ͺ|͸Ǥͳ͸, and their dot
is: ‫ ݌‬Ȉ ‫ ݍ‬ൌ ʹ ൈ ͷ൅Ȃ ͵ ൈ ͹ ൅ ͷ ൈȂ ͳ ൌ ͳͲȂ ʹͳȂ ͷ ൌȂ ͳ͸.
Example in R4: If ‫ ݌‬ൌ ሺʹǡ Ȃ ͵ǡͶǡͳሻ and ‫ ݍ‬ൌ ሺͷǡ ͳͲǡ Ȃ ͳǡͲሻ, then the
norm of ‫ ݌‬is ȁ‫݌‬ȁ ൌ ඥʹଶ ൅ ሺെ͵ሻଶ ൅ Ͷଶ ൅ ͳଶ ൌ ξ͵Ͳ|ͷǤͶͺ, and
their dot product is:
‫ ݌‬Ȉ ‫ ݍ‬ൌ ʹ ൈ ͷ൅Ȃ ͵ ൈ ͳͲ ൅ Ͷ ൈȂ ͳ ൅ ͳ ൈ Ͳ ൌȂ ʹͶ.
Norms and dot products are easy to

calculate, but what are their geometric
meanings? To answer this, draw a purple
arrow from the origin to ‫ ݌‬and draw an
orange arrow from the origin to ‫ݍ‬, as
shown in the illustration on the
right. In this picture, ȁ‫݌‬ȁ is
simply the length of the purple
arrow. Their dot product, ‫ ݌‬Ȉ ‫ݍ‬, has a geometric meaning that
depends on the angle (which we’ll call D) between the purple and
orange arrows.
THE MEANING OF THE DOT PRODUCT: If ‫ ݌‬and ‫ ݍ‬are points

in Rn, and D is the angle between the arrows pointing to them
from the origin, then ‫ ݌‬Ȉ ‫ ݍ‬ൌ ȁ‫݌‬ȁȁ‫ݍ‬ȁܿ‫ݏ݋‬ሺDሻ.
That is, the dot product can be found by multiplying these

three numbers together: the norm of ‫݌‬, the norm of ‫ݍ‬, and the
cosine of the angle between them. We will not prove this formula,
but we will tell you why you should care about it.
Our main purpose for this dot product formula is the

following quick and easy method for testing whether the purple
and orange arrows are perpendicular:
PERPENDICULARITY TEST:
If ‫ ݌‬Ȉ ‫ ݍ‬൐ Ͳ, then D is acute (less than 90°).
If ‫ ݌‬Ȉ ‫ ݍ‬ൌ Ͳ, then D is right (equal to 90°).
If ‫ ݌‬Ȉ ‫ ݍ‬൏ Ͳ, thenD is obtuse (greater than 90°).
Using the Dot Product to Find a Lover or a Song 187
In particular, the orange and purple arrows are perpendicular

exactly when the dot product equals zero. This perpendicularity
test works because the cosine of an acute angle is positive, the
cosine of a right angle is zero, and the cosine of an obtuse angle
is negative. Think about why.
EXAMPLE: If ‫ ݌‬ൌ ሺʹǡ Ȃ ͵ǡͷሻ and ‫ ݍ‬ൌ ሺͷǡ ͹ǡ Ȃ ͳሻ in R3, then their
dot product is ‫ ݌‬Ȉ ‫ ݍ‬ൌȂ ͳ͸, so the angle is obtuse.
EXAMPLE: If ‫ ݌‬ൌ ሺ͵ǡͶሻ and ‫ ݍ‬ൌ ሺʹǡ͹ሻin R2, then ‫ ݌‬Ȉ ‫ ݍ‬ൌ ͵Ͷ, so
the angle is acute.
EXAMPLE: If ‫ ݌‬ൌ ሺͶǡʹሻand ‫ ݍ‬ൌ ሺȂ ͵ǡ͸ሻ in R2, then their dot

product is ‫ ݌‬Ȉ ‫ ݍ‬ൌ Ͷ ൈȂ ͵ ൅ ʹ ൈ ͸ ൌȂ ͳʹ ൅ ͳʹ ൌ Ͳ, so the
angle is right, which means that the arrows are perpendicular.
See how easy it is to determine whether two arrows are

perpendicular? But what if you wish to know, not just whether the
angle is acute or obtuse, but exactly what the angle equals? For
this, just solve the “Meaning of the Dot Product” formula for
ܿ‫ݏ݋‬ሺDሻ, which yields:
௣Ȉ௤
ܿ‫ݏ݋‬ሺDሻ ൌ .
ȁ௣ȁȁ௤ȁ
This tells you what the cosine of D equals. Then use the
–1
“cos ” button on your calculator, which un-does the cosine
function, to find what D equals. In the plane R2 and space R3, this
answer is exactly the angle that you would measure with a
protractor. In higher dimensional Euclidean spaces, protractors
don’t make sense, so we simply take this answer as our definition
of the word “angle” in higher dimensional Euclidean spaces.
Using the Dot Product to Find a Lover or a Song

Here is an activity that might give you a better feeling for
distances, dot products and angles. This activity will also help
you understand the need to compute these quantities in high

dimensional Euclidean spaces. But let’s start with R2.
Write down your own personal point ሺ‫ݔ‬ǡ ‫ݕ‬ሻ in R2 which

represents ‫ = ݔ‬how much you love cats and ‫ = ݕ‬how much you
love sushi, each on a scale from –5 (hate) to +5 (love). For
example, my personal point is K = ሺെ͵ǡ ͷሻ because I am allergic
to cats and I love raw salmon. I named my point “K” because my
first name is Kris. Next, ask a friend to write down his or her
personal point. Draw your point and your friend’s point in R2, and
draw arrows to them from the origin. Next do three calculations:
(1) Calculate the dot product of your point and your friend’s point.
To what extent do you think that this dot product measures the
compatibility of your interests? If the dot product is largely
positive, does this mean that your interests are closely aligned
with your friend’s? If the dot product is largely negative, does this
mean that the two of you have opposite interests? What does it
mean if the dot product equals zero?
(2) Calculate the angle between the arrows. To what extent do

you think that this angle measures the compatibility of your
interests? Does an acute angle mean that your interests are
closely aligned? Does an obtuse angle indicate opposite
interests? What does a right angle indicate?
(3) Calculate the distance between your point and your friend’s
point. To what extent do you think that this distance measures
the compatibility of your interests? If the distance is small, does
this mean that your interests are closely aligned? If the distance
is large, does this indicate that you have opposite interests?
Which quantity does the best job of measuring the

compatibility of your interests: (1) the dot product, (2) the angle,
or (3) the distance? There is no right answer – each strategy has
Using the Dot Product to Find a Lover or a Song 189
advantages and disadvantages. But notice that ALL three

quantities can be computed using dot products. Here is why:
NORMS, DISTANCES, AND ANGLES CAN ALL BE

COMPUTED USING DOT PRODUCTS:
(1) ȁ‫݌‬ȁ ൌ ඥ‫ ݌‬Ȉ ‫( ݌‬the norm of a point is the square root of its dot
product with itself).
(2) The distance between ‫ ݌‬ൌ ሺܽǡ ܾǡ ܿሻ and ‫ ݍ‬ൌ ሺ‫ݔ‬ǡ ‫ݕ‬ǡ ‫ݖ‬ሻ equals
ȁ‫ ݌‬െ ‫ݍ‬ȁ ൌ ȁሺܽ െ ‫ݔ‬ǡ ܾ െ ‫ݕ‬ǡ ܿ െ ‫ݖ‬ሻȁ
ൌ ඥሺܽ െ ‫ݔ‬ǡ ܾ െ ‫ݕ‬ǡ ܿ െ ‫ݖ‬ሻ Ȉ ሺܽ െ ‫ݔ‬ǡ ܾ െ ‫ݕ‬ǡ ܿ െ ‫ݖ‬ሻ.
௣Ȉ௤
(3) ܿ‫ݏ݋‬ሺDሻ ൌ .
ȁ௣ȁȁ௤ȁ
In summary, knowing how to compute dot products allows

you to also compute norms, angles, and distances. All three
strategies for measuring the compatibility of your interests are
really just based on dot product calculations.
The previous activity might remind you of those dating

websites that help single people find other single people with
compatible interests. How does such a website match you up
with your perfect future spouse? First, the site asks you about
your interest level in more than just cats and sushi. Let us
suppose the site asks you 50 questions, so that your personal
point lies in R50. As discussed above, there are several
reasonable strategies for measuring how “close” your personal
point is to the personal point of someone else, like say Matt
Damon or Penelope Cruz. The dot product is at the heart of all
such strategies.
The administrators of dating websites are not the only

folks who make money by calculating dot products in high
dimensional Euclidean spaces. The creators of the Music
Genome Project and Pandora Radio have also made a bundle.
Their idea was to encode the musical essence of a song as a

point in a high dimensional Euclidean space – as high as R400 for
some music genres. The 400 numbers represent 400 musical
attributes of the song: how twangy are the guitar solos, how
angry are the lyrics, what gender is the lead vocalist, what is the
song’s time signature, and so on. Identifying the point in R400 that
encodes one song takes a trained musician about a half an hour.
They analyzed a large collection of songs and founded a music
recommendation website based on their song-point database. A
user of their website enters a song that he or she enjoys, and the
website creates a personal radio station of similar songs. The
user provides further information about his or her preferences by
indicating approval or disapproval of each song played, and the
station takes this added preference information into account in
selecting future songs to play.
Pandora Radio keeps their algorithms secret, but it is fun

to guess how the songs are selected. To find new songs similar
to the songs that you like, their algorithm at core must measure
the similarity of songs based on the “closeness” of the
corresponding points in R400. As mentioned previously, the most
natural measurements of closeness are all based on the dot
product – in this case the dot product in R400. You might think of
this as the “dot” in Pandora.com.
What is a Rigid Motion?

We have been studying rigid motions since Chap. 1. At
long last, we are finally equipped to explain precisely what this
term means. The key idea is to regard a rigid motion of Rn as a
function from Rn to Rn. That is, a rigid motion is a rule or formula
which allows one to determine the output point in Rn associated
to each input point in Rn. Visually, the output point represents
where the motion “moves” the input point to.
What is a Rigid Motion? 191
For example, one of the first rigid motions of R2 that we

studied was R90 – the 90° counterclockwise rotation of the plane
about the origin. For practice, let us regard R90 as a function from
R2 to R2. It inputs a point of the plane, called
‫݌‬, and it outputs a point of the plane, called
R90ሺ‫݌‬ሻ. The illustration on the right
shows how several colored points are
moved by this rotation:
R90ሺͲǡͲሻ ൌ ሺͲǡͲሻ
R90ሺͳǡͲሻ ൌ ሺͲǡͳሻ
R90ሺ͵ǡͳሻ ൌ ሺȂ ͳǡ͵ሻ
R90ሺͲǡʹሻ ൌ ሺȂ ʹǡͲሻ
R90ሺͲǡ Ȃ ʹሻ ൌ ሺʹǡͲሻ
Do you see that pattern? You might guess that the pattern
is this: R90ሺ‫ݔ‬ǡ ‫ݕ‬ሻ ൌ ሺȂ ‫ݕ‬ǡ ‫ݔ‬ሻ. In the next chapter, we will prove that
this formula is correct. Assuming this for now, this example helps
us get used to thinking of a rigid motion as a function. Do you see
how R90 is a definite rule that associates an output point to each
input point? The rule is described by the formula R90ሺ‫ݔ‬ǡ ‫ݕ‬ሻ ൌ
ሺȂ ‫ݕ‬ǡ ‫ݔ‬ሻ. This formula empowers you to quickly determine R90’s
effect on any input point you like. For example, R90ሺ͵͹ǡͷͷሻ ൌ
ሺȂ ͷͷǡ͵͹ሻ. This is a whole new way to think about R90.
Translations are even easier than rotations to describe as

functions. For example, let T3 denote the translation of the plane
a distance 3 to the right. Considered as a function, T3 simply
increases the x-coordinate of each point by 3. A general formula
is T3ሺ‫ݔ‬ǡ ‫ݕ‬ሻ ൌ ሺ‫ ݔ‬൅ ͵ǡ ‫ݕ‬ሻ. For example, T3ሺ͵ǡͷሻ ൌ ሺ͸ǡͷሻ and
similarly T3ሺͳͲǡ Ȃ ͵ሻ ൌ ሺͳ͵ǡ Ȃ ͵ሻ.
When regarded as a function,

the defining property of a rigid motion,
F, is this: it preserves distances. This
means that the distance between any
pair of points is the same before and
after the rigid motion is applied. That
is what makes it rigid! In other words,
the distance from ‫ ݌‬to ‫ ݍ‬must equal
the distance from Fሺ‫݌‬ሻ to Fሺ‫)ݍ‬, for
any pair of points ‫ ݌‬and ‫ݍ‬. For R90 preserves distances
example, in the illustration on the right, the two dashed lines have
the same length because R90 preserves the distance between the
pink and green points.
At last, we can precisely define the vocabulary words upon

which our entire study of symmetry has been built:
DEFINITIONS:
A rigid motion of Rn means a function, F, from Rn to Rn that

preserves distances.
An object in Rn means a set of points in Rn.
A symmetry of an object in Rn means a rigid motion, F, of Rn

which moves the object onto itself; that is, Fሺ‫݌‬ሻ is a point of the
object whenever ‫ ݌‬is a point of the object.
In truth, this definition of “object” only models single-

colored or single-material objects. For example, we previously
visualized a two-dimensional object as pained on the glass plane.
If only white paint was used, then you may now more precisely
regard this object as the set of points of R2 that have paint on
them. But the above definitions must be modified to allow for
multicolored objects. Think about how. Similarly, we previously
visualized a three-dimensional object as sculpted, perhaps out of
Two Exotic Examples 193
bronze, and surrounded by an infinite expanse of ice. You may

now more precisely regard this object as the set of points of R3
where there is bronze rather than ice. But if multiple building
materials were used, such as bronze, wood and paint, then you
will need to think on your own about how to generalize the above
definitions to incorporate this added generality.
It is difficult to picture an object in higher dimensional

Euclidean spaces, but a picture is not always required. For
example, the collection of songs written by Leonard Cohen is an
object in R400. This object is unlikely to have any symmetries
other than the identity.
Two Exotic Examples

How well do the precise definitions of “object”, “rigid
motion,” and “symmetry” agree with the intuitive feelings you
have by now developed for these terms? It is important for the
intuition and the rigor to be closely aligned, or at least to
understand the ways in which they differ. Perhaps you previously
thought of an object in R2 as something that you could paint or
draw in the plane, and an object in R3 as something you could
build out of plastic or wood or metal. However, our precise
definition of the word “object” allows some things that can only be
made with mathematical formulas not with paint or wood. Here
are two examples.
EXAMPLE: AN EXOTIC BORDER PATTERN: When we

classified the seven border patterns, we only considered border
patterns that have a “smallest translation”. What was the purpose
of that restriction? For one thing, it ruled out the ‫ݔ‬-axis, which is a
border pattern that can be translated any amount right or left. But
it also ruled out much more complicated border patterns. For
example, consider the set of all rational numbers on the ‫ݔ‬-axis. In
other words, the set of all point in R2 of the form ሺ‫ݔ‬ǡ Ͳሻ, where ‫ ݔ‬is
a rational number. This is a border pattern, but you would be hard

pressed to draw an accurate picture of it. A translation right or left
by any rational length is a symmetry of this border pattern
because the sum of two rational numbers is rational, so the
pattern is moved onto itself. A translation right or left by any
irrational length is NOT a symmetry of this border pattern
because a rational plus an irrational is irrational, so the pattern is
not moved onto itself. This border pattern’s symmetry group is
very different from the symmetry groups of the seven border
patterns in the classification theorem.
EXAMPLE: AN EXOTIC BOUNDED OBJECT: Look back at Da

Vinci’s classification of the possible symmetry groups of bounded
objects in the plane. Notice that his theorem said nothing about
bounded objects with infinite symmetry groups. Surprisingly, the
circle is NOT the only such object. For example, consider the set
of all points on the unit circle whose angle-positions are rational;
that is, points that have the form ሺܿ‫ݏ݋‬ሺ‫ݐ‬ሻǡ ‫݊݅ݏ‬ሺ‫ݐ‬ሻሻ for some rational
number ‫ݐ‬. This is a bounded object in R2, although you would be
hard pressed to draw an accurate picture of it. A rotation
(centered at the origin) by any rational angle is a symmetry of this
object, but a rotation by any irrational angle is not a symmetry of
this object. To see why this is true, just notice that the rotation by
angle D moves each point of this object according to this formula:
RDሺܿ‫ݏ݋‬ሺ‫ݐ‬ሻǡ ‫݊݅ݏ‬ሺ‫ݐ‬ሻሻ ൌ ሺܿ‫ݏ݋‬ሺ‫ ݐ‬൅ Dሻǡ ‫݊݅ݏ‬ሺ‫ ݐ‬൅ Dሻሻ. So if D is rational,
then RD moves this object onto itself, but if D is irrational, then it
does not.
We mention the above examples in part to demonstrate

the necessity of the fine print in the classification theorems from
Chap. 4. We may not have anticipated examples like these, but
we have no choice but to accept them. Surprising and
counterintuitive examples play a crucial role in mathematics. The
help delineate the true from the false, and studying them carefully
can lead us to rich new understandings.
E
Exerrcise
es 19
95
Exxerrcissess
(1)) In
n th
he prooff of th
he Pyytha agooreean Th heoore
em, exxpla
ain why
w the
oraangge squ
uarre w
with
h siide ngtth ܿ iss re
e len eallyy a sq
qua
are.
(2)) If ‫ ݌‬ൌ ሺʹǡ͹ሻ and ‫ ݍ‬ൌ ሺ͵ǡ Ȃ ͷሻ in R2, fin he follow

nd th win
ng
qu es: ȁ‫݌‬ȁǡ ȁ‫ݍ‬ȁǡ ‫ ݌‬Ȉ ‫ݍ‬, the
uantitie disttance fro
e d om ‫ ݌‬to ‫ݍ‬, an
nd tthe
e an
nglle
be
etweeenn th
he a
arrowws frromm thhe orig n
gin to ‫݌‬ and ‫ݍ‬.
(3)) Iff ‫ ݌‬ൌ ሺͳǡͶǡ͵ሻ and ‫ ݍ‬ൌ ሺȂ ͳǡʹǡ͹ሻ iin R3, find
d th
he follow
win
ng
uantitie
qu es: ȁ‫݌‬ȁǡ ȁ‫ݍ‬ȁǡ ‫ ݌‬Ȉ ‫ݍ‬, the distance from ‫ ݌‬to ‫ݍ‬, annd tthe
e an
nglle
be
etwe een
n th
he aarrowws frrom
m thhe orig n to ‫ ݌‬and ‫ݍ‬.
gin
n R5, do
(4)) In o th om the o
ws fro
he arrrow orig o ሺͳǡ Ȃ ʹǡ͵ǡ Ȃ Ͷǡͷሻ an
gin to nd
ሺͳǡͲǡ Ȃ ͳǡ͵ǡʹሻ form an acute, obtuse or right angle?
(5)) Let ‫ ݌‬ൌ ሺ͵ǡ͹ሻ. Show that ‫ݍ‬ଵ ൌ ሺെ͹ǡ͵ሻ is perp pen ndicculaar to p
an
nd h hass th
he sa ame e nnorm
m a as ‫ ݌‬. Sh howw tthatt ‫ݍ‬ଶ ൌ ሺ͹ǡ െ͵ሻ is also
pe
erpe end
dicuularr to
o p a and ha as the e ssamme no ormm aas p. Draw w thhesse
po
ointss. Wh hich
h o onee of ‫ݍ‬ଵ ,‫ݍ‬ଶ is oobtaained byy rrota atinng ‫ ݌‬b by 900°
clo ? Whicch iss obtain
ockkwisse? nedd byy rottating ‫ ݌‬byy 90 0°
counttercclockw G ss a forrmu
wisse? Gue or tthe 90
ula fo 0° clo ockkwisse annd
counttercclockwwisse rrota
atio
on o
of aan aarb ary point ‫ ݌‬ൌ ሺ‫ݔ‬ǡ ‫ݕ‬ሻ.
bitra
ula for R90 is: R90ሺ‫ݔ‬ǡ ‫ݕ‬ሻ ൌ ሺȂ ‫ݕ‬ǡ ‫ݔ‬ሻ. G

(6)) A forrmu Gue
esss an
naana
alog
gou
us
forrmuula for R180ሺ‫ݔ‬ǡ ‫ݕ‬ሻ and for R270ሺ‫ݔ‬ǡ ‫ݕ‬ሻ.
(7)) Gue
G ess a fformu ula forr Hሺ‫ݔ‬ǡ ‫ݕ‬ሻ and Vሺ‫ݔ‬ǡ ‫ݕ‬ሻ, where H m mea anss thhe
hoorizo
onttal flip he ‫ ݔ‬-a
p ovverr th axiss and V me
eanns the
e ve
ertiical fliip o
ove
er
2
e ‫ ݕ‬-axxis in R .
the
(8) If R is the rotation by 90° about the ‫ݖ‬-axis in R3, guess a

formula for Rሺ‫ݔ‬ǡ ‫ݕ‬ǡ ‫ݖ‬ሻ.
(9) If F is the reflection across the ‫ݕݔ‬-plane in R3, guess a

formula for Fሺ‫ݔ‬ǡ ‫ݕ‬ǡ ‫ݖ‬ሻ. Do the same for reflections across the ‫ݖݔ‬-
and the ‫ݖݕ‬-planes.
(ٍ10) If D is any angle, find a formula in terms of D for RDሺͳǡͲሻ

and RDሺͲǡͳሻ.
(ٍ11) In the sine and cosine table in the chapter, the values are
rounded to 2 decimals, but these values can be determined
exactly. Determine the exact values of ܿ‫ݏ݋‬ሺͶͷሻ ൌ ‫݊݅ݏ‬ሺͶͷሻ|Ǥ͹ͳ
and of ‫݊݅ݏ‬ሺ͸Ͳሻ|Ǥͺ͹. Hint: ܿ‫ݏ݋‬ሺ͸Ͳሻ equals exactly ½.
(12) Fill in the following table with rounded values for the sine and
cosine of angles between 180 and 360. HINT: Use the same
numbers that are in the table in the chapter; namely, plus and
minus 0, .26, .50, .71, .87, .97 and 1.
‫ݐ‬ 180 195 210 225 240 255 270 285 300 315 330 345 360
ܿ‫ݏ݋‬ሺ‫ݐ‬ሻ
‫݊݅ݏ‬ሺ‫ݐ‬ሻ
(ٍ13) Which of the following types of symmetries does the

“exotic border pattern” described in this chapter have: vertical
flips, horizontal flips, 180° rotations, glide-reflections.
(ٍ14) Does the exotic bounded object described in this chapter

have any improper symmetries? In other words, is this object
oriented?
(ٍ15) Consider the border pattern consisting of all rational

numbers on the ‫ݔ‬-axis that can be expressed as a fraction whose
denominators is a power of 2. Describe all of its translation sym-
metries. Does this border pattern have a “smallest translation”?
Exercises 197
(ٍ16) The exotic bounded object described in this chapter is a

countable infinite collection of points of the unit circle with the
property that any point can be moved to any other point by a
symmetry of the object. Is there an analogous three-dimensional
object? That is, can you find a countable infinite collection of
points on the sphere such that any point of the collection can be
moved to any other point by a symmetry of the collection? HINT:
Because of the classification theorem, any such example must be
essentially two-dimensional.
(17) If a cube with side-length 2 is centered at the origin (0,0,0),

then the locations of its eight vertices are: ሺͳǡͳǡͳሻ, ሺͳǡͳǡ െͳሻ,
ሺͳǡ െͳǡͳሻ, ሺͳǡ െͳǡ െͳሻ, ሺെͳǡͳǡͳሻ, ሺെͳǡͳǡ െͳሻ, ሺെͳǡ െͳǡͳሻ,
ሺെͳǡ െͳǡ െͳሻ. Draw a picture of this cube. Prove that it is a
Platonic solid by verifying that all six of its faces are identical
regular 4-gons. In particular, you must verify that the edges meet
at right angles.
ͳ͵Ǥ
In the early chapters of this book, we described rigid
motions with phrases like “the 90° rotation about a point.” Then
we learned to describe rigid motions with formulas like F(x,y) =
(–y,x). But for many purposes, the very best way to describe a
rigid motion is with a matrix. That is what we will do in this final
chapter.
Matrix Computations
A matrix simply means a grid of real numbers.
DEFINITION: An n-by-n matrix means n2 real numbers arranged

into a square grid.
For example, here are a few 2-by-2 matrices:
ʹ െ͵ Ͳ ൰, ‫ ܥ‬ൌ ቀെʹͲ െ͵
‫ܣ‬ൌቀ ቁ, ‫ ܤ‬ൌ ൬ ξʹ ቁ
ͳ ͹ െͷͳ ͵Τͷ ߨ Ͳ
and here are a few 3-by-3 matrices:
െʹ ͷ ͻ Ͳ ͳ െ͵ ͳ Ͳ Ͳ
‫ܦ‬ൌ൭ Ͳ Ͷ ͸ ൱, ‫ ܧ‬ൌ ቌͷ െʹȀ͹ ʹ͹ ቍ, ‫ ܨ‬ൌ ൭Ͳ ͵ Ͳ൱
െʹȀ͵ ͺ െͳ Ͳ ξ͹ Ͳ Ͳ Ͳ ͻ
We will think of each row and each column of an n-by-n

matrix as a point in Rn. For example, the rows of the above
matrix ‫ ܧ‬are: ሺͲǡ ͳǡ െ͵ሻǡ ሺͷǡ െʹȀ͹ǡʹ͹ሻ and ሺͲǡ ξ͹ǡ Ͳሻ, while the
columns of ‫ܨ‬are: ሺͳǡ Ͳǡ Ͳሻǡ ሺͲǡ ͵ǡͲሻ, and ሺͲǡ Ͳǡ ͻሻ. The entries of a
matrix are the numbers out of which it is built. For example, the
(2,3)-entry of ‫ ܦ‬is ͸, while the (3,2)-entry of ‫ ܦ‬is ͺ. Notice the
convention of indexing the entry’s row first and then its column.

200 13. Symmetry and Matrices
There are three important types of matrix computations,

which we will now describe. First, we will describe how to multiply
a pair of matrices. The answer is another matrix.
HOW TO MULTIPLY A PAIR OF MATRICES: If ‫ ܣ‬and ‫ ܤ‬are n-

by-n matrices, then ‫ ܤ כ ܣ‬is the n-by-n matrix whose (i,j)-entry
equals the dot product of the ith row of ‫ ܣ‬with the jth column of‫ܤ‬.
2-BY-2 EXAMPLE:
ʹ െ͵ ͵ Ͷ ሺʹǡ െ͵ሻ Ȉ ሺ͵ǡͲሻ ሺʹǡ െ͵ሻ Ȉ ሺͶǡͳͲሻ
ቀ ቁ‫כ‬ቀ ቁൌ൬ ൰
ͳ ͹ Ͳ ͳͲ ሺͳǡ͹ሻ Ȉ ሺ͵ǡͲሻ ሺͳǡ͹ሻ Ȉ ሺͶǡͳͲሻ
͸ െʹʹ
ൌቀ ቁ.
͵ ͹Ͷ
Each of the four entries of the answer equals the dot product of
the same-colored row of the first matrix with same-highlighted
column of the second matrix.
3-BY-3 EXAMPLE:
ʹ െ͵ Ͷ ͳ ͳͲ ͵ ͺ െͳʹ ͵͵

൭Ͳ ͳ ͷ൱ ‫ כ‬൭െʹ Ͷ െͳ൱ ൌ ൭െʹ െʹͳ ʹͻ൱.
െͶ Ͳ ͵ Ͳ െͷ ͸ െͶ െͷͷ ͸
As before, each of the nine entries of the answer equals the dot
product of the same-colored row of the first matrix with same-
highlighted column of the second matrix.
Second, we will describe how to compute the determinant

of a matrix.
THE DETERMINANT OF A 2-BY-2 MATRIX: The determinant of a

2-by-2 matrix ‫ܣ‬, denoted as ሺ‫ܣ‬ሻ, is the real number defined as:
ܽ ܾ
݀݁‫ ݐ‬ቀ ቁ ൌ ܽ݀ െ ܾܿ.
ܿ ݀
Matrix Computations 201
This is simple enough. For example,
ʹ ͷ
ቀ ቁ ൌ ʹ ൈ ͵ െ ͷ ൈ ͹ ൌ ͸ െ ͵ͷ ൌ െʹͻǤ
͹ ͵
The determinant of a matrix is a number, which might be positive
or negative or zero.
THE DETERMINANT OF A 3-BY-3 MATRIX: The determinant of

a 3-by-3 matrix ‫ܣ‬, denoted as ሺ‫ܣ‬ሻ, is the real number defined
as:
ܽ ܾ ܿ
݀݁‫ ݐ‬൭݀ ݁ ݂ ൱ ൌ ܽ݁݅ ൅ ܾ݂݃ ൅ ݄ܿ݀ െ ܿ݁݃ െ ݂݄ܽ െ ܾ݀݅.
݃ ݄ ݅
You do not need to memorize this formula because there is a

simple way to visualize it. Just append copies of the first two
columns of A to the right edge of A and draw red and green
circles like this:
Notice that the three green circles correspond to the three

positive terms in the determinant formula, while the three red
circles correspond to the negative terms.
There is an analogous (but messier) formula for the

determinant of a 4-by-4 matrix and a 5-by-5 matrix, and so on. In
this book, you will only need to compute determinants of 2-by-2
and 3-by-3 matrices, but be aware that the word “determinant”
does make sense for larger matrices.
Third, we will describe how to multiply a matrix times a

point. The answer is a point:
HOW TO MULTIPLY A MATRIX TIMES A POINT: If ‫ ܯ‬is an n-

by-n matrix and ‫ ݌‬is a point in Rn, then ‫ ݌ כ ܯ‬means the point in
Rn whose ith coordinate equals the dot products of the ith row of
‫ ܯ‬with ‫݌‬.
ʹ െ͵
EXAMPLE: If ‫ ܯ‬ൌ ቀ ቁ and ‫ ݌‬ൌ ሺͳͲǡ Ͷሻ, then:
ͳ ͷ
‫ ܲ כ ܯ‬ൌ ൫ሺʹǡ െ͵ሻ Ȉ ሺͳͲǡͶሻǡ ሺͳǡͷሻ Ȉ ሺͳͲǡͶሻ൯ ൌ ሺͺǡ͵Ͳሻ
ͳ ʹ െ͵
EXAMPLE: If ‫ ܯ‬ൌ ൭ Ͳ Ͷ ͷ ൱ and ܲ ൌ ሺʹǡ Ͷǡͳሻ, then:
െʹ Ͳ ͳ
‫ ܲ כ ܯ‬ൌ ൫ሺͳǡʹǡ െ͵ሻ Ȉ ሺʹǡ Ͷǡͳሻǡ ሺͲǡͶǡͷሻ Ȉ ሺʹǡ Ͷǡͳሻǡ ሺെʹǡͲǡͳሻ Ȉ ሺʹǡ Ͷǡͳሻ൯
ൌ ሺ͹ǡʹͳǡ െ͵ሻ
In summary, you just learned three algebraic skills: (1) multiplying

a matrix times a matrix, (2) computing the determinant of a matrix,
and (3) multiplying a matrix times a point. All three computations
involve only adding and multiplying. In the next section, we
discuss the geometric meanings of these computations, and the
relationship between matrices and rigid motions.
Representing Rigid Motions as Matrices

Matrices are great, but they are really only useful for
studying the symmetries of a bounded object. Here is why. The
familiar Center Point Theorem for two- and three-dimensional
objects generalizes to objects in Rn. The generalization says this:
GENERAL CENTER POINT THEOREM: Any bounded object in

Rn has a “center point” that is fixed by each of its symmetries.
Representing Rigid Motions as Matrices 203
In other words, each of its symmetries leaves its center point

unmoved. Any bounded object can be translated so that its
center point becomes the origin ሺͲǡͲǡ ǥ ǡͲሻ of Rn. After this re-
positioning, all of its symmetries will be rigid motions that fix the
origin (they will leave the origin unmoved). This is exactly the
type of rigid motions that matrices can help us study. In this
chapter, we only consider bounded objects centered at the origin.
Each symmetry of such an object fixes the origin.
You previously learned to regard a rigid motion as a

function and represent it using a formula. For example, R90 has
the formula: R90(x,y) = (–y,x). Now, we will redescribe a rigid
motion as a matrix. Here is how.
THE MATRIX THAT REPRESENTS A RIGID MOTION: The

matrix that represents a rigid motion, F, of Rn is the n-by-n matrix
whose columns are FሺͳǡͲǡ ǥ ǡͲሻ, FሺͲǡͳǡͲǡ ǥ ǡͲሻ,…, FሺͲǡ ǥ ǡͲǡͳሻ, in
this order.
So the matrix representing a rigid motion, F, of R2 is the 2-

by-2 matrix whose two columns are FሺͳǡͲሻ and FሺͲǡͳሻ. The matrix
representing a rigid motion, F, of R3 is the 3-by-3 matrix whose
three columns are FሺͳǡͲǡͲሻ, FሺͲǡͳǡͲሻ and FሺͲǡͲǡͳሻ.
EXAMPLE: The rigid motion R90 is represented by the 2-by-2

matrix whose first column equals R90ሺͳǡͲሻ ൌ ሺͲǡͳሻ and whose
second column equals R90ሺͲǡͳሻ ൌ ሺȂ ͳǡͲሻ. Thus:
Ͳ െͳ
R90 = ቀ ቁ
ͳ Ͳ
In what way exactly does this matrix “represent” R90? At first
glance, this matrix seems to only tell us which outputs R90 asso-
ciates to the inputs ሺͳǡͲሻ and ሺͲǡͳሻ. Magically, this matrix also
tells us the outputs R90 associate to every possible input point of

R2. Remember how to multiply a matrix times a point? Watch:
Ͳ െͳ ሺ
ቀ ቁ ‫ݔ כ‬ǡ ‫ݕ‬ሻ ൌ ሺെ‫ݕ‬ǡ ‫ݔ‬ሻ
ͳ Ͳ
See what happened here? When the matrix that represents R90
was multiplied times a point of R2, the answer was exactly the
output that R90 associates with that input point. For example:
Ͳ െͳ ሺ ሻ
R90(3,1) = (–1,3) ՞ ቀ ቁ ‫͵ כ‬ǡͳ ൌ ሺെͳǡ͵ሻ
ͳ Ͳ
This is exactly what always happens!
THEOREM: If F is a rigid motion of Rn that fixes the origin, and ‫ܯ‬

is the matrix that represents F, then for any point ‫ ݌‬of Rn, we
have: ‫ = ݌ כ ܯ‬F(‫)݌‬.
In words: “To learn the output point that F associates to the input
point ‫݌‬, you multiply ‫ ܯ‬by ‫݌‬.” This is a powerful theorem. For
example, when n = 3, the columns of ‫ ܯ‬are defined to record
where F moves the three points ሺͳǡͲǡͲሻ, ሺͲǡͳǡͲሻ and ሺͲǡͲǡͳሻ. It is
surprising that this information is enough to determine where F
moves ALL points of R3, and that this determination is achieved
via something as simple as matrix-point multiplication.
Let us revisit D4 = the symmetry group of the square. Here

are the matrices representing all eight of its members (assuming
the square is centered at the origin of R2):
ͳ Ͳ Ͳ െͳ െͳ Ͳ Ͳ ͳ
Iൌ ቀ ቁǡR90ൌ ቀ ቁǡR180ൌ ቀ ቁǡ R270ൌ ቀ ቁǡ
Ͳ ͳ ͳ Ͳ Ͳ െͳ െͳ Ͳ
ͳ Ͳ െͳ Ͳ Ͳ െͳ Ͳ ͳ
Hൌ ቀ ቁǡVൌ ቀ ቁǡDൌ ቀ ቁǡD'ൌ ቀ ቁǤ
Ͳ െͳ Ͳ ͳ െͳ Ͳ ͳ Ͳ
Representing Rigid Motions as Matrices 205
Look back at the pictures of these eight

symmetries in Chap. 2 and verify that
these eight matrices are all correct. For
example, the matrix for D is correct
because its columns are D(1,0) = (0,–1)
and D(0,1) = (–1,0) as shown in the
illustration on the right.
What happens when you multiply two of these matrices?

For example, what is the matrix that represents R90 times the
matrix that represents D? Try it. You will discover the answer is
the matrix that represents H, agreeing with the fact that
R90*D = H. It seems that matrix multiplication achieves com-
position of symmetries! Here is the general rule:
MATRIX MULTIPLICATION ACHIEVES COMPOSITION: If F1

and F2 are rigid motions of Rn that fix the origin, and ‫ܯ‬ଵ and ‫ܯ‬ଶ
are the matrices that represent them, then ‫ܯ‬ଵ ‫ܯ כ‬ଶ is the matrix
that represents their composition F1*F2.
We now have a very effective way to translate visual

questions about symmetries into algebraic questions about
matrices. Do you see how it works? If your cousin Henry had
never heard of D4, you could describe this group to him simply by
showing him the eight matrices and nothing else. He could use
matrix multiplication to build the Cayley table, without ever cutting
out a cardboard square or visualizing a rotation or flip.
The rotations in D4 were all multiples of 90°, which made it

easy to find the matrices representing them. To do the same for
other cyclic and dihedral groups, the rotations are more
complicated, so you will need this:
THEOREM: The matrix that represents a rotation about the origin

ܿ‫ݏ݋‬ሺ‫ݐ‬ሻ െ‫݊݅ݏ‬ሺ‫ݐ‬ሻ
of R2 by ‫ ݐ‬degrees is: ൬ ൰.
‫݊݅ݏ‬ሺ‫ݐ‬ሻ ܿ‫ݏ݋‬ሺ‫ݐ‬ሻ
This matrix’s first column is correct because cosine and

sine are exactly defined so that the rotation sends ሺͳǡͲሻ to
ሺܿ‫ݏ݋‬ሺ‫ݐ‬ሻǡ ‫݊݅ݏ‬ሺ‫ݐ‬ሻሻ. Think about why the second column is also
correct.
Orthogonal Matrices
Not just any matrix represents a rigid motion. The matrices
that do have a special property called “orthogonal.”
DEFINITION: A matrix is called orthogonal if the norm of each

column equals 1 and the dot product of each pair of different
columns equals 0.
Suppose that F is a rigid motion of R3 that fixes the origin,

and ‫ ܯ‬is the matrix that represents F. Let us think about why we
might expect ‫ ܯ‬to be orthogonal. Imagine
watching F move the orange, green, red,
and purple points in the illustration on
the right. Since F is rigid, the distance
between pairs of these points is
the same before and after the
motion. The orange point
stays put because it is the
origin (try saying “orange
origin” 10 times fast). Imagine the colored arrows moving along
with to aim at the new locations of the green, red and purple
points. These new locations are the three columns of ‫ܯ‬. After the
motion, the arrows will still have length 1, which is to say that the
columns of ‫ ܯ‬have norm 1. This is simply because the distance
Orthogonal Matrices 207
between the orange and each other point is unchanged by the

motion. After the motion, these arrows will still be mutually
perpendicular, which is to say that the dot product of each pair of
different columns of ‫ ܯ‬equals 0. Why? If the angle between two
arrows became acute, the points they aim at would have grown
closer together. If the angle became obtuse, the points would
have grown further apart. Can you picture this? This visual dis-
cussion is not a proof, but it should help you believe the following:
THEOREM: If ‫ ܯ‬is a matrix that represents a rigid motion that

fixes the origin, then ‫ ܯ‬is an orthogonal matrix. Conversely,
every orthogonal matrix represents a rigid motion that fixes the
origin.
It turns out that orthogonal matrices have very limited

possibilities for their determinants. Calculate the determinant of
each of the eight matrices for D4. You will discover that the
determinant of each rotation is 1, while the determinant of each
flip is –1. The general rule is:
THEOREM: The determinant of any orthogonal matrix equals

either 1 (if it represents a proper rigid motion) or –1 (if improper).
This theorem should perhaps be green-boxed instead of blue-

boxed because it is in part a definition. It is a more precise
definition of the terms “proper” and “improper” for rigid motions of
the plane and space. For rigid motions of higher dimensional
Euclidean spaces, it is our only definition of these terms.
What about rigid motions which do not fix the origin? Can
we use determinants to distinguish whether they are proper or
improper? This turns out to be easy because:
THEOREM: Every rigid motion of Rn equals a rigid motion that

fixes the origin followed by a translation.
So a rigid motion is called proper or improper depending on

whether its origin-fixing part has determinant equal to 1 or –1.
This is a good definition. Defining “proper” using determinants is
much more precise and unambiguous than our previous verbiage
about right hands and clocks.
You Finished the Book. Now What?

In this chapter, we intended only to describe how matrices
are related to symmetry. We did not include any proofs because
the proofs belong to a linear algebra book. We also did not carry
out the important work of using matrices to prove the previously
unproven theorems scattered throughout this book.
Thus, this book ends with a beginning – a more rigorous

starting point from which you can revisit the topic of symmetry
with more precise definitions and more complete proofs. We
hope we have piqued your interest in someday reading the books
and taking the classes in which this idea is fully developed. Here
is a brief glimpse of some what’s left to learn about symmetries
and matrices.
Matrices provide a precise definition of “rigid motion” from

which one can quickly prove the classifications of plane and
space rigid motions found in this book. Some of the most difficult
theorems in this book can then be rephrased and proved using
matrices. For example, the classification of symmetry groups of
solid bounded objects boils down to understanding the possible
finite subgroups of the group of all orthogonal 3-by-3 matrices.
Things get really interesting when you move into higher

dimensional Euclidean spaces. Consider this question:
Classify the possible symmetry groups of bounded objects in Rn.

Exe
ercisses 20
09
Y
Youu alrea adyy kknow the e answ wer fo or n = 2 annd n = 3. W Wha at
abbout gene eraal n
n? T
Thiss qque estio
on turrned o out to bee veery diffficult an nd
very im mpo orta
ant.. It mo otivateed so ome e of th he mostt ssign nificcan
nt
maathemmatics of the e pa astt ce
entuuryy. F
Firstt, th uestion
his qu n iss in
nterrtwiine
ed
e prob
witth the blemm of cla assifying all posssible finiite gro oup ps, whicch
ad to on
lea ne off thhe mo ost ce elebraatedd a ach
hievvem
men nts off m mod derrn
maathemmatics,, a aptlyy ccallled
d “TThee EEno orm mouus Th heorem m.” Seco ondd,
this que esttion
n iss rrela
ated to the cla assificatio on off ccom mpa act Liie
grooup
ps, whicch iis a b beaautiiful piece e oof m maathe
ema aticcs up pon whicch
muuchh off modern n phhyssicss deepeendds.
AAlte
erna
ately, there e iss a vveryy m
mod derrn me etho od for p provvin
ng
cla on the
asssificcatio eorrem
ms witthout ma atriccess. T
Thiss m
mod derrn a
app proaachh,
cham mpio oneed by John Co onwwayy and oth herrs, usees orb bifo
old geo ommetrry
to claassify theew which bo
wayys in w ordeer ppattterrns,, wallppapper pa atte
ernss,
annd bbou undded soolid ob
bjeccts caan b
be ssym
mm metrric. A generral--audie encce
ducction o
inttrod his prrogram
of th m iss fo
ounnd iin a beau utifful boook title
ed Thhe
Syymm metry of Th gs, by Co
hing onwwayy, B
Burrgieel and Go ooddma an--Strrau
uss.
Exxerrcissess
ʹ െ͵ ͵ Ͳ െͳͲ െ͵
(1)) Consider: ‫ ܣ‬ൌ ቀ ቁ, ‫ ܤ‬ൌ ൬ ൰, ‫ ܥ‬ൌ ቀ ቁ.
ͳ ͹ െͷ ͳȀʹ Ͷ Ͳ
(a
a) C
Comp
pute
e th
he d
detterm
min
nan
nt o
of each
hmmattrix..
b) Compute ‫ ܤ כ ܣ‬and ‫ܣ כ ܤ‬. Are they equal?

(b
(cc) Verify that ݀݁‫ݐ‬ሺ‫ܥ כ ܣ‬ሻ ൌ ݀݁‫ݐ‬ሺ‫ܣ‬ሻ ‫ݐ݁݀ כ‬ሺ‫ܥ‬ሻ.
d) Compute ‫ כ ܣ‬ሺʹǡ͵ሻ and ‫ כ ܣ‬ሺെʹǡ͹ሻ.

(d
(e Verrifyy that ‫ ܣ‬iss not o

e) V orth
hog
gon
nal..
(f) The function from R 2 to R2 determined by ‫( ܣ‬sending ܲ

to ‫ )ܲ כ ܣ‬is not a rigid motion because ‫ ܣ‬is not orthogonal.
Show this directly by finding a pair of points whose
distance is not preserved.
െʹ ͷ ͻ Ͳ ͳ െ͵
(2) Consider: ‫ ܣ‬ൌ ൭ Ͳ Ͷ ͸ ൱, ‫ ܤ‬ൌ ൭ͷ െʹ ͳͲ ൱.
െͳ ʹ െͳ Ͳ ͳ Ͳ
(a) Compute the determinants of ‫ ܣ‬and ‫ܤ‬.
(b) Compute ‫ܤ כ ܣ‬and ‫ܣ כ ܤ‬. Are they equal?
(c) Compute ‫ כ ܣ‬ሺͳǡʹǡ͵ሻand ‫ כ ܣ‬ሺͲǡͶǡ െʹሻ.
(d) The function from R3 to R3 determined by ‫( ܣ‬sending ܲ

to ‫ )ܲ כ ܣ‬is not a rigid motion because ‫ ܣ‬is not orthogonal.
Show this directly by finding a pair of points whose
distance is not preserved.
(ٍ3) Write the matrix that represents each of the six symmetries
in D3.
(4) Determine the matrix that represents each of the following

rigid motion of R3:
(a) The 90° rotation about the ‫ݖ‬-axis or the ‫ݔ‬-axis.
(b) A reflection across the ‫ݖݕ‬-plane or the ‫ݖݔ‬-plane.
(ٍ5) In this problem, we will revisit the central inversion rigid

motion defined at the end of Chap. 7.
(a) Find the 3-by-3 matrix, ‫ܯ‬ଵ , that represents F1 = the

reflection across the ‫ݕݔ‬-plane.
(b) Find the 3-by-3 matrix, ‫ܯ‬ଶ , that represents F2 = the

180° rotation about the ‫ݖ‬-axis.
Exercises 211
(c) Verify that F1 and F2 commute by checking that

‫ܯ‬ଵ ‫ܯ כ‬ଶ ൌ ‫ܯ‬ଶ ‫ܯ כ‬ଵ . Can you picture this using the
illustration found in Chap. 7?
(d) The composition F = F1*F2 is called central inversion.

Find the matrix ‫ ܯ‬ൌ ‫ܯ‬ଵ ‫ܯ כ‬ଶ that represents F. Verify
that F has the formula Fሺ‫ݔ‬ǡ ‫ݕ‬ǡ ‫ݖ‬ሻ ൌ ሺെ‫ݔ‬ǡ െ‫ݕ‬ǡ െ‫ݖ‬ሻ.
(e) Verify that F is improper by showing ݀݁‫ݐ‬ሺ‫ܯ‬ሻ ൌ െͳ.
(f) Verify that F commutes with every rigid motion of R3

that fixes the origin. Do this by checking that ‫ ܯ כ ܣ‬ൌ
‫ ܣ כ ܯ‬for every 3-by-3 matrix ‫ܣ‬.
(g) Explain why F is a symmetry of the cube and the

dodecahedron but not of the tetrahedron.
(ٍ6) Describe in words the rigid motion of R3 represented by:
ሺ‫ݐ‬ሻ Ͳ െሺ‫ݐ‬ሻ
൭ Ͳ ͳ Ͳ ൱
ሺ‫ݐ‬ሻ Ͳ ሺ‫ݐ‬ሻ
Index
A Countable, 171–174, 178, 197
All-or-Half theorem, 26, 32, 53, 85, Cube, 88–90, 92, 93, 96–99, 101,
107 104, 108, 109, 111, 112, 115,
Alternating group, 83, 96, 97 118–121, 123, 124, 131, 132,
Associative property, 21, 23, 24 134, 141, 197, 211
Asymmetric, 27, 103 Cycle notation, 79, 81, 83
Axis, 10, 88, 89, 91, 94, 99, 100, Cyclic group, 25, 32, 42, 46, 47, 51,
109, 184, 193, 195, 196, 210 65, 66, 101, 102, 113
B D
Bilateral symmetry, 90 Da Vinci’s theorem, 51–54, 100, 101
Border pattern, classification, Determinant, 200–202, 207, 208
12–13, 54–56 Dihedral group, 25, 32, 42, 100,
Bounded, 7, 8, 12, 15, 16, 32, 45, 102, 113
51–54, 90–92, 101–104, 107, Distance formula, 180–183
114, 115, 117, 119, 134, 140, Dodecahedron, 92, 93, 97–99,
142, 194, 196, 197, 202, 203, 101–105, 108, 109, 111, 112,
208 115, 118–121, 123, 124, 131,
Bubble theorem, 142 132, 134, 141, 211
Dot product, 185–190, 200, 206,
C 207
Cantor’s theorem, 174–177 Duality, 121–124, 131
Cayley table, 18–21, 26, 27, 31,
36–38, 40, 46, 54, 63, 64, 66, E
70–72, 74, 78, 94, 107, 108, 205 Edge, 18, 94, 97, 98, 115, 120, 121,
Center Point theorem, 15, 52, 91, 124–128, 137, 145, 173, 201
92, 94, 202 Essentially two-dimensional,
Central inversion, 108, 109, 210, 99–101, 103, 113, 114, 197
211 Euclidean space, 179–197
Centrally symmetric, 109, 111, 113 Euclid’s theorem, 159
Chiral, 104–106, 111, 113 Euler characteristic, 129–131, 133,
Circle theorem, 142, 148 136
Commutative group, 23, 42 Euler’s formula
Commutative property, 21–23 For plane, 126–127
Commute, 107, 211 For sphere, 127–128
Composition Even permutation, 82, 83, 85, 96, 98,
of a permutation, 76–77 111
of symmetries, 21–22
Connected, 121, 125–130, 136, F
137 Face, 16, 25, 93–95, 104, 120, 124,
Cosine, 184, 186, 187, 196, 206 126, 127, 129, 131, 136, 141
K. Tapp, Symmetry: A Mathematical Exploration, DOI 10.1007/978-1-4614-0299-2, 213

214 Index
Flip. See Reflection O

Frieze pattern. See Border pattern Object, 3–10, 12, 14–17, 24, 26, 27,
32, 42–45, 51–53, 74, 87–92,
G 100–102, 104, 107, 109, 111,
Glide reflection, 11, 12, 15, 55, 57 112, 114, 115, 134, 142, 149,
Graph, 125–130, 136, 137 151, 192–194, 196, 197, 202,
Group, 22–26, 31–33, 35–47, 49, 203
51–53, 55, 58, 60, 63–74, 76, Octahedron, 116, 119–121, 123,
78, 83, 91–102, 105, 107–109, 124, 131, 132, 134
111–114, 124, 134, 194, 204, Odd permutation, 82, 85
205, 208 Order
of an axis, 89, 111
I of a group member, 69
Icosahedron, 116, 117, 119–121, of a rotation, 56, 69, 113
123, 124, 128, 131, 134 Oriented, 9, 10, 12, 14–16, 25, 46,
Identity 53, 60, 196
additive/multiplicative, 21 Orthogonal matrix, 207
of a group, 23, 24 Overcrowding, 117–119, 133
symmetry, 6, 21, 24
Improper, 6, 7, 9, 14, 16, 26, 27, 32, P
90, 91, 94, 100, 105, 107, 108, Permutation group, 76–80
111, 113, 114, 124, 196, 207, Perpendicularity test, 186, 187
208, 211 Planar, 51, 125, 126, 128, 136, 137
Integer, 23, 24, 41, 151, 153 Platonic solid, 115–137, 197
Inverse Plato’s theorem, 119
additive/multiplicative, 22 Prime factorization, 150–152, 159,
of a group member, 23–24, 77–78 160
of permutations, 78, 82 Prime number, 150, 159–161
of a symmetry, 22 Prime number theorem, 161
Isomorphic, 36, 38–45, 49–51, 53, Product group, 71, 72
55, 58, 60, 63, 66, 71, 72, 74, Proper, 6, 7, 12, 14, 16, 26, 27, 74,
78, 79, 86, 91, 92, 95–97, 90–105, 107–114, 124, 134,
99–102, 107–109, 111–114 137, 207, 208
Isomorphism, 35–50, 71, 72, 78, 79, Proper rigid equivalence, 43–46
96, 97, 108 Proper symmetry group, 74, 91–93,
95–102, 104, 105, 107–109,
M 111–113, 134
Matrix, 199–211 Pythagorean theorem, 181–183
Matrix multiplication, 205
Minimal surface, 142 R
Rational number, 151–153, 155,
N 156, 163, 167, 172, 194
Natural number, 150, 170, 178 Real number, 153–156, 174–176,
Norm, 185, 186, 189, 195, 206 200, 201
Index 215
Real redundancy rule, 155 Symmetry group, 24–26, 31, 33, 35,
Reflection, 5, 6, 8, 9, 11, 12, 15, 20, 36, 39, 40, 42–45, 49, 51–53,
32, 36, 37, 43, 55, 57, 60, 72, 55, 58, 60, 63, 65, 71, 72, 74,
89–95, 104, 108, 109, 111, 143, 92–102, 105, 107–109,
144, 196, 210 111–114, 124, 134, 194, 204
Regular polygon, 10, 26, 115
Regular polyhedron, 115 T
Riemann hypothesis, 162 Tetrahedron, 92–96, 99, 101, 104,
Rigid equivalence, 43–46 111, 112, 115, 116, 119, 120,
Rigid motion 124, 131, 132, 134, 140, 141,
definition, 3 211
plane classification, 12–13 Topology, 130
space classification, 90–93 Translation, 2–4, 6, 10–12, 15, 16,
Rigid motion detector theorem, 27 32, 40, 45, 54, 55, 57, 58, 65,
Rotation, 3–6, 8, 9, 15, 16, 18, 20, 88–91, 108, 191, 193, 194, 196,
25–27, 32, 36, 39, 40, 43, 207
45–47, 52–57, 63, 64, 67, 69,
72, 88–92, 94, 95, 99, 100, 108, U
112, 113, 191, 194, 196, 199, Unbounded, 7, 8, 10, 11, 15, 90, 91,
205–207, 210 126, 142
Unit circle, 183–185, 194, 197
S
Same size, 27, 42, 167–172, 176, V
178 Vertex, 85, 94, 95, 115–118, 120,
Sine, 184, 196, 206 121, 124, 126, 127, 131, 133
Subgroup, generated, 66–69 Volleyball, 110
Sudoku theorem, 26
Swap, 81 W
Symmetry, 1–33, 35, 36, 38–45, Wallpaper pattern, classification, 58
49–53, 55–58, 60, 61, 63, 65,
69, 71, 72, 74, 76, 79, 85,
87–115, 124, 134, 139–149,
158, 159, 178, 192–194,
197–211

(Kristopher Tapp) Symmetry A Mathematical Explora

Uploaded by

Copyright:

Available Formats

(Kristopher Tapp) Symmetry A Mathematical Explora

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

(Kristopher Tapp) Symmetry A Mathematical Explora

Uploaded by

Copyright:

Available Formats

Symmetry

ISBN 978-1-4614-0298-5 e-ISBN 978-1-4614-0299-2

© Springer Science+Business Media, LLC 2012

Printed on acid-free paper

Springer is part of Springer Science+Business Media (www.springer.com)

If you are curious about the

objects may possess, which is a crucial prerequisite for

No background beyond high school level algebra is

This book is primarily intended as a textbook for a one-

Instructor resources, including PowerPoint lectures and

I am delighted to thank Paul Klingsberg and Susan Ramee

INTENDED AUDIENCE ........................................................................................ VI

A PRECISE DEFINITION OF “SYMMETRY”................................................................ 2

2. THE ALGEBRA OF SYMMETRY 17

CAYLEY TABLES ............................................................................................. 18

WHAT IS AN ISOMORPHISM? ........................................................................... 36

4. THE CLASSIFICATION THEOREMS 51

BOUNDED OBJECTS ........................................................................................ 51

5. SUBGROUPS AND PRODUCT GROUPS 63

PERMUTATION GROUPS ................................................................................. 76

7. SYMMETRIES OF SOLID OBJECTS 87

RIGID MOTIONS OF SPACE ............................................................................... 87

8. THE FIVE PLATONIC SOLIDS 115

COUNTING THEIR PARTS ................................................................................ 120

9. SYMMETRY AND OPTIMIZATION 139

MINIMAL SURFACES ..................................................................................... 139

10. WHAT IS A NUMBER? 149

NATURAL NUMBERS ..................................................................................... 149

11. CANTOR’S INFINITY 167

THE MODERN MEANING OF “SAME SIZE” .......................................................... 168

12. EUCLIDEAN SPACE 179

THE PYTHAGORAN THEOREM AND DISTANCE FORMULA ....................................... 180

13. SYMMETRY AND MATRICES 199

MATRIX COMPUTATIONS .............................................................................. 199

How do you interpret this question in a manner which is precise

Seahorses and Eels by Robert Fathauer

To answer any of these questions, we must first make the

K. Tapp, Symmetry: A Mathematical Exploration, DOI 10.1007/978-1-4614-0299-2_1, 1

A Precise Definition of “Symmetry”

Most people would agree that Seahorses and Eels looks

To phrase this idea more precisely, let us imagine that the

Here is an equivalent way

(which means slid) so that each right-facing eel moves one

This way of thinking about symmetry applies equally well

Here is our first attempt at formulating a precise definition of the

DEFINITION: A symmetry of an object in the plane is a rigid

In the above discussions, the “plane” was represented as

Precise language is crucially important in mathematics!

A rigid motion is always a motion of the whole plane (the

We emphasize that a rigid motion is completely

One type of rigid motion that we have not yet considered is

The words flip and reflection are synonymous; we will

To better understand why the previously illustrated star

Types of Symmetries and Types of Objects

To describe to me your favorite rotation or translation or

Rotations and translations are called “proper” rigid motions.

the viewer. A more precise

DEFINITION: A rigid motion is called proper if it preserves

An even more precise definition of proper/improper will be