Beautiful Geometry

Frontispiece: Infinity
Beautiful Geometry
Eli Maor and Eugen Jost

Princeton University Press

Princeton and Oxford
 Copyright © 2014 by Princeton University Press

Published by Princeton University Press, 41 William Street, Princeton, New Jersey 08540

In the United Kingdom: Princeton University Press, 6 Oxford Street, Woodstock, Oxfordshire OX20 1TW

press.princeton.edu

Jacket Illustration: Sierpinski’s Arrowhead by Eugen Jost

All Rights Reserved

Library of Congress Cataloging-in-Publication Data

Maor, Eli.

Beautiful geometry / Eli Maor and Eugen Jost.

pages cm

Summary: “If you’ve ever thought that mathematics and art don’t mix, this stunning visual history of geometry will
change your mind. As much a work of art as a book about mathematics, Beautiful Geometry presents more than sixty
exquisite color plates illustrating a wide range of geometric patterns and theorems, accompanied by brief accounts of the
fascinating history and people behind each. With artwork by Swiss artist Eugen Jost and text by acclaimed math historian
Eli Maor, this unique celebration of geometry covers numerous subjects, from straightedge-and-compass constructions to
intriguing configurations involving infinity. The result is a delightful and informative illustrated tour through the 2,500-year-
old history of one of the most important and beautiful branches of mathematics”— Provided by publisher.

Includes index.

ISBN-13: 978-0-691-15099-4 (cloth : acid-free paper)

ISBN-10: 0-691-15099-0 (cloth : acid-free paper) 1. Geometry—History. 2. Geometry—History—Pictorial

works. 3. Geometry in art. I. Jost, Eugen, 1950– II. Title.

QA447.M37 2014

516—dc23

2013033506

British Library Cataloging-in-Publication Data is available

This book has been composed in Baskerville 10 Pro

Printed on acid-free paper. ∞

Printed in Canada

1 3 5 7 9 10 8 6 4 2
 To Dalia, my dear wife of fifty years
May you enjoy many more years of
good health, happiness, and Naches from your family.
—Eli

To my dear Kathrin and to my whole family

Two are better than one; because they have a good reward for their labor.
For if they fall, the one will lift up his fellow (Ecclesiastes 4:9-10).
—Eugen
 Contents

Prefacesix 13. One Theorem, Three Proofs 39

1. Thales of Miletus 1 14. The Prime Numbers 42

2. Triangles of Equal Area 3 15. Two Prime Mysteries 45

3. Quadrilaterals 6 16. 0.999. . . = ? 49

4.  erfect Numbers and

P 17. Eleven 53
Triangular Numbers 9
18. Euclidean Constructions 56
5. The Pythagorean Theorem I 13
19. Hexagons 59
6. The Pythagorean Theorem II 16
20. Fibonacci Numbers 62
7. Pythagorean Triples 20
21. The Golden Ratio 66
8. The Square Root of 2 23
22. The Pentagon 70
9. A Repertoire of Means 26
23. The 17-Sided
10. More about Means 29 Regular Polygon 73

11. Two Theorems from Euclid 32 24. Fifty 77

12. Different, yet the Same 36 25. Doubling the Cube 81

v i i i    C o n t e n t s

26. Squaring the Circle 84 45. Symmetry II 149

27. Archimedes Measures 46. The Reuleaux Triangle 154

the Circle 88
47. Pick’s Theorem 157
28. The Digit Hunters 91
48. Morley’s Theorem 160
29. Conics 94
3 4
49. The Snowflake Curve 164
30. =
3 4 99
50. Sierpinski’s Triangle 167
31. The Harmonic Series 102
51. Beyond Infinity 170
32. Ceva’s Theorem 105

33. e 108 Appendix: Proofs of Selected

Theorems Mentioned in This Book 175
34. Spira Mirabilis 112
Quadrilaterals 175

35. The Cycloid 116 Pythagorean Triples 176

A Proof That √2 Is Irrational 176
36. Epicycloids and Hypocycloids 119
Euclid’s Proof of the Infinitude
of the Primes 176
37. The Euler Line 123
The Sum of a Geometric
38. Inversion 126 Progression 177
The Sum of the First n Fibonacci
39. Steiner’s Porism 130 Numbers 177
Construction of a
40. Line Designs 134 Regular Pentagon 177
Ceva’s Theorem 179
41. The French Connection 138
Some Properties of Inversion 180
42. The Audible Made Visible 141

Bibliography 183
43. Lissajous Figures 143

44. Symmetry I 146 Index 185

 Prefaces
A r t t h r o ugh Mathemati cal Ey e s

El i Maor

N o doubt many people would agree that art and

mathematics don’t mix. How could they? Art,
after all, is supposed to express feelings, emotions,
repeating itself to infinity, becomes etched in his or
her mind as a pattern. The search for pattern is indeed
the common thread that ties mathematics to art.
and impressions—a subjective image of the world as
the artist sees it. Mathematics is the exact oppo-
•
site—cold, rational, and emotionless. Yet this per- The present book has its origin in May 2009, when
ception can be wrong. In the Renaissance, mathe- my good friend Reny Montandon arranged for me
matics and art not only were practiced together, they to give a talk to the upper mathematics class of the
were regarded as complementary aspects of the Alte Kantonsschule (Old Cantonal High School) of
human mind. Indeed, the great masters of the Re- Aarau, Switzerland. This school has a historic claim
naissance, among them Leonardo da Vinci, Michel- to fame: it was here that a 16-year-old Albert Ein-
angelo, and Albrecht Dürer, considered themselves stein spent two of his happiest years, enrolling there
as architects, engineers, and mathematicians as much at his own initiative to escape the authoritarian edu-
as artists. cational system he so much loathed at home. The
If I had to name just one trait shared by mathe- school still occupies the same building that Einstein
matics and art, I would choose their common search knew, although a modern wing has been added next
for pattern, for recurrence and order. A mathemati- to it. My wife and I were received with great honors,
cian sees the expression a2 + b2 and immediately and at lunchtime I was fortunate to meet Eugen Jost.
thinks of the Pythagorean theorem, with its image of I had already been acquainted with Eugen’s ex-
a right triangle surrounded by squares built on the quisite mathematical artwork through our mutual
three sides. Yet this expression is not confined to ge- friend Reny, but to meet him in person gave me spe-
ometry alone; it appears in nearly every branch of cial pleasure, and we instantly bonded. Our encoun-
mathematics, from number theory and algebra to ter was the spark that led us to collaborate on the
calculus and analysis; it becomes a pattern, a para- present book. To our deep regret, Reny Montandon
digm. Similarly, when an artist looks at a wallpaper passed away shortly before the completion of our
design, the recurrence of a basic motif, seemingly book; just one day before his death, Eugen spoke to
x p r e fac e s

him over the phone and told him about the progress Our aim is to reach a broad audience of high
we were making, which greatly pleased him. Sadly school and college students, mathematics and sci-
he will not be able to see it come to fruition. ence teachers, university instructors, and laypersons
Our book is meant to be enjoyed, pure and sim- who are not afraid of an occasional formula or equa-
ple. Each topic—a theorem, a sequence of numbers, tion. With this in mind, we limited the level of math-
or an intriguing geometric pattern—is explained in ematics to elementary algebra and geometry (“ele-
words and accompanied by one or more color plates mentary” in the sense that no calculus is used). We
of Eugen’s artwork. Most topics are taken from ge- hope that our book will inspire the reader to appre-
ometry; a few deal with numbers and numerical pro- ciate the beauty and aesthetic appeal of mathemat-
gressions. The chapters are largely independent of ics and of geometry in particular.
one another, so the reader can choose what he or she Many people helped us in making this book a re-
likes without affecting the continuity of reading. As ality, but special thanks go to Vickie Kearn, my
a rule we followed a chronological order, but occa- trusted editor at Princeton University Press, whose
sionally we grouped together subjects that are re- continuous enthusiasm and support has encouraged
lated to one another mathematically. I tried to keep us throughout the project; to the editorial and tech-
the technical details to a minimum, deferring some nical staff at Princeton University Press for their ef-
proofs to the appendix and referring others to exter- forts to ensure that the book meets the highest aes-
nal sources (when referring to books already listed thetic and artistic standards; to my son Dror for his
in the bibliography, only the author’s name and the technical help in typing the script of plate 26 in He-
book’s title are given). Thus the book can serve as an brew; and, last but not least, to my dear wife Dalia
informal—and most certainly not complete—survey for her steady encouragement, constructive critique,
of the history of geometry. and meticulous proofreading of the manuscript.
 Pl ay i ng wit h Patterns, Numbe rs, and F o rms

Eu gen Jost

M y artistic life revolves around patterns, numbers,

and forms. I love to play with them, interpret
them, and metamorphose them in endless variations.
should I say discovered?—by the Greeks more than
two thousand years ago.
For me, playing with numbers and patterns al-
My motto is the Pythagorean motto: Alles ist Zahl (“All ways has top priority. That’s why I like to call my
is Number”); it was the title of an earlier project I pictures playgrounds, following a statement by the
worked on with my friends Peter Baptist and Carsten Swiss Artist Max Bill: “perhaps the goal of concrete
Miller in 2008. Beautiful Geometry draws on some of the art is to develop objects for mental use, just like peo-
ideas expressed in that earlier work, but its conception ple created objects for material use.” Some illustra-
is somewhat different. We attempt here to depict a tions in our book can be looked at in this sense. The
wide se­lection of geometric theorems in an artistic onlooker is invited to play: to find out which rules a
way while remaining faithful to their mathematical picture is built on and how the many metamorpho-
message. ses work, to invent his or her own pictures. In some
While working on the present book, my mind was chapters the relation between text and picture is
often with Euclid: A point is that which has no part; loose; in others, however, artistic claim stood be-
a line is a breadthless length. Notwithstanding that hind the goal to enlighten Eli’s text. Most illustra-
claim, Archimedes drew his broad-lined circles with tions were created on the keypad of my computer,
his finger in the sand of Syracuse. Nowadays it is but others are acrylics on canvas. Working with Eli
much easier to meet Euclid’s demands: with a few was a lot of fun. He is one of those mathematicians
clicks of the mouse you can reduce the width of a that teach you: Mathematics did not fall from
line to nothing—in the end there remains only a heaven; it was invented and found by humans; it is
nonexisting path. It was somewhat awe inspiring to full of stories; it is philosophy, history and culture. I
go through the constructions that were invented—or hope the reader will agree.
Beautiful Geometry
Plate 1. Sunrise over Miletus
 1
Thales of Miletus

T hales (ca. 624–546 BCE) was the first of the long

line of mathematicians of ancient Greece that
would continue for nearly a thousand years. As with
they recognized him as one of the Seven Wise Men
of Greece.
Mathematics was already quite advanced during
most of the early Greek sages, we know very little Thales’s time, but it was entirely a practical science,
about his life; what we do know was written several aimed at devising formulas for solving a host of fi-
centuries after he died, making it difficult to distin- nancial, commercial, and engineering problems.
guish fact from fiction. He was born in the town of Thales was the first to ask not only how a particular
Miletus, on the west coast of Asia Minor (modern problem can be solved, but why. Not willing to ac-
Turkey). At a young age he toured the countries of cept facts at face value, he attempted to prove them
the Eastern Mediterranean, spending several years from fundamental principles. For example, he is
in Egypt and absorbing all that their priests could credited with demonstrating that the two base an-
teach him. gles of an isosceles triangle are equal, as are the two
While in Egypt, Thales calculated the height of vertical angles formed by a pair of intersecting lines.
the Great Cheops pyramid, a feat that must have left He also showed that the diameter of a circle cuts it
a deep impression on the locals. He did this by into two equal parts, perhaps by folding over the
planting a staff into the ground and comparing the two halves and observing that they exactly over-
length of its shadow to that cast by the pyramid. lapped. His proofs may not stand up to modern
Thales knew that the pyramid, the staff, and their standards, but they were a first step toward the kind
shadows form two similar right triangles. Let us de- of deductive mathematics in which the Greeks
note by H and h the heights of the pyramid and the would excel.
staff, respectively, and by S and s the lengths of their Thales’s most famous discovery, still named after
shadows (see figure 1.1). We then have the simple him, says that from any point on the circumference
equation H/S = h/s, allowing Thales to find the value of a circle, the diameter always subtends a right
of H from the known values of S, s, and h. This feat angle. This was perhaps the first known invariance
so impressed Thales’s fellow citizens back home that theorem—the fact that in a geometric configuration,
2   B e a u t i f u l Geometry

s S Figure 1.1

some quantities remain the same even as others are N ot e :

changing. Many more invariance theorems would
1. The converse of Thales’s theorem is also true: the
be discovered in the centuries after Thales; we will
locus of all points from which a given line segment sub-
meet some of them in the following chapters.
tends a constant angle is an arc of a circle having the line
Thales’s theorem can actually be generalized to
segment as chord. In particular, if the angle is 90°, the
any chord, not just the diameter. Such a chord di-
locus is a full circle with the chord as diameter.
vides the circle into two unequal arcs. Any point
lying on the larger of these arcs subtends the chord
at a constant angle α < 90°; any point on the smaller
arc subtends it at an angle β = 180° – α > 90°. 1 Plate 1,
Sunrise over Miletus, shows this in vivid color.
 2
Triangles of Equal Area

A round 300 BCE, Euclid of Alexandria wrote his

Elements, a compilation of the state of mathe-
matics as it was known at his time. Written in 13
ces that lie on a line parallel to the base have the
same area.
In proving this theorem, it would be tempting to
parts (“books”) and arranged in strict logical order use the familiar formula for the area of a triangle,
of definitions, postulates (today we call them axi- A = bh/2, and argue that since two triangles with the
oms), and propositions (theorems), it established same base b and top vertices that lie on a line paral-
mathematics as a deductive discipline, in which lel to the base also have the same height h, they must
every theorem must be proved based on previously have the same area. This “algebraic” proof, however,
established theorems, until we fall back on a small would not satisfy Euclid; he insisted that a proof
set of postulates whose validity we assume to be should be based strictly on geometric consider-
true from the outset. Euclid’s 23 opening defini- ations. Here is how he proved it:
tions, 10 axioms, and 465 theorems cover all of
classical (“Euclidean”) geometry—the geometry we In figure 2.1, let triangles ABC and DEF have
learn in school—as well as elementary number the- equal bases BC and EF. Their vertices A and D
ory. The Elements is considered the most influential are on a line parallel to BC and EF. We extend
book in the history of mathematics. It has had an AD to points G and H, where GB is parallel to
enormous influence on generations of mathemati- AC and HF is parallel to DE. Then the figures
cians, scientists, and philosophers, and its terse style GBCA and DEFH are parallelograms with the
and rigid structure of definitions, postulates, propo- same area, for they have equal bases BC and EF
sitions, and demonstrations (proofs) became the and lie between the same parallels BF and GH.
paradigm of how mathematics should be done. Now the area of triangle ABC is half the area of
Proposition 38 of Book I of the Elements says, Tri- parallelogram GBCA, and the area of triangle
angles which are on equal bases and in the same parallels DEF is half the area of parallelogram DEFH.
are equal to one another. In modern language this Therefore the two triangles have the same area—
reads: all triangles with the same base and top verti- QED.1
4

Plate 2. Triangles of Equal Area

 2 . T r i a n g l e s o f E q u a l A r e a    5

G A D H

B C E F Figure 2.1

Our illustration (plate 2) shows three identical This theorem may seem rather unassuming, but
red triangles, each of whose sides can be regarded as Euclid makes good use of it in proving other, more
a base. A series of blue triangles are built on each advanced theorems; most famous among them is the
base, with their vertices moving along a line parallel Pythagorean theorem, to which we will turn in chap-
to that base. They get narrower as the vertices move ters 5 and 6.
farther out, yet they all have exactly the same area,
providing another example of a quantity that re- N ot e :
mains unchanged even as other quantities in the 1. QED stands for quod erat demonstrandum, Latin for
configuration vary. “that which was to be demonstrated.”
 3
Quadrilaterals

H ere is a little-known jewel of a theorem that never

fails to amaze me: take any quadrilateral (four-
sided polygon), connect the midpoints of adjacent
Now, of course, you can repeat the process and
connect the midpoints of PQRS to get another,
smaller parallelogram, as shown in the middle panel
sides, and—surprise—you’ll get a parallelogram! of plate 3. In fact, you can do this again and again,
The surprise lies in the word any. No matter how getting ever smaller parallelograms whose areas are
skewed your quadrilateral is, the outcome will al- 1⁄2, 1⁄4, 1⁄8, . . . of the original quadrilateral, until they

ways be a parallelogram. The theorem holds true seem to converge to a point.

even for the dart-shaped quadrilateral shown in the While we are on the subject of quadrilaterals, here
top-left corner of plate 3. And that’s not all: the area is another little-known fact: the area of any quad­
of the parallelogram will always be one half the area rilateral is completely determined by the lengths of
of the quadrilateral from which it was generated. its two diagonals and the angle between them. In
The proof is rather short and is based on the fol- fact, the area is given by the simple formula A =
lowing theorem: in any triangle, the line joining the 1⁄2d ⋅ d ⋅ sin α (figure 3.3). It doesn’t matter how you
1 2
midpoints of two sides is parallel to the third side
and is half as long (see figure 3.1). Now let’s apply
this to quadrilateral ABCD. Denote the midpoints of
sides AB, BC, CD and DA by P, Q, R, and S, respec- 1 A
tively (figure 3.2). Line PQ is parallel to diagonal AC, MN = BC
2
which in turn is parallel to RS. Thus, PQ and RS are
parallel. By the same argument, lines PS and RQ are
M N
also parallel, so PQRS is a parallelogram—QED.
(The proof that it has half the area of the generating
quadrilateral is just a tad longer and is given in the
appendix.) B C Figure 3.1
 7

Plate 3. Quadrilaterals
8   B e a u t i f u l Geometry

P
S

B
D
d1

α d2
Q
R

C Figure 3.2 Figure 3.3

measure the angle— α or (180° − α)—for we know

from trigonometry that sin α = sin (180° − α). It is a
pity that these little treasures seldom, if ever, find
their way into our geometry textbooks.
 4
Perfect Numbers and Triangular Numbers

T he Pythagoreans—the school founded by Pythago-

ras in the fifth century BCE—had a special relation-
ship with numbers (the term here meaning positive in-
decahedron (12 regular pentagons), and the icosahe-
dron (20 equilateral triangles); they are shown in fig-
ure 4.1. No wonder, then, that the number 5 acquired
tegers). In their mind, numbers were not just a something of a sacred status with the Greeks.
measure of quantity but symbols possessing mythical Even more revered than five was the number 6,
significance. The number 1 was not considered a num- the first perfect number, being the sum of its proper
ber at all, but rather the generator of all numbers, divisors, 1 + 2 + 3.1 The next perfect number is 28
since every number can be obtained from it by re- (= 1 + 2 + 4 + 7 + 14), followed by 496 and 8,128. These
peated addition. Two symbolized the female charac- were the only perfect numbers known in antiquity.
ter, 3 the male character, and 5 their union. Five was As of this writing, 48 perfect numbers are known; the
also the number of Platonic solids—convex polyhedra largest, discovered in 2013, is 257,885,160 ⋅ (257,885,161 − 1),
whose faces are all identical regular polygons (al- an enormous number of nearly 35 million digits.
though the proof that there are exactly 5 of them came The question of how many perfect numbers exist—
only later). These 5 solids are the tetrahedron, having 4 or even whether their number is finite or infinite—is
equilateral triangles as faces, the cube (6 square faces), still unanswered.
the octahedron (2 square pyramids joined at their Six is also a triangular number, so called because
bases and comprising 8 equilateral triangles), the do- these numbers form a triangular pattern when ar-

Figure 4.1
10

Plate 4. Figurative Numbers

 4 . P e r f e c t & T r i a n g u l a r N u m b e r s    11

ranged in rows of 1, 2, 3, . . . dots. The first four trian-

gular number are 1, 1 + 2 = 3, 1 + 2 + 3 = 6, and
1 + 2 + 3 + 4 = 10, followed by 15, 21, and so on. The
Pythagoreans discovered several relations between
these sequences of integers. For example, the nth tri-
angular number is always equal to n(n + 1)/2 (you Figure 4.2
can check this for a few cases: 1 + 2 + 3 = 6 = (3 × 4)/2,
1 + 2 + 3 + 4 = 10 = (4 × 5)/2, etc.). So this gives us a
convenient way—a formula—for finding the sum
of the first n integers without actually adding them
up:
n(n + 1)
1+ 2 + 3 ++ n = .
2
When we add instead the first n odd integers, a
surprise is awaiting us: the result is always a perfect
square: 1 = 12, 1 + 3 = 4 = 22, 1 + 3 + 5 = 9 = 32, and, in Figure 4.3
general,
1 + 3 + 5 + 7 + ⋅⋅⋅ + (2n − 1) = n2.

Still another relation comes from adding two con- perfect numbers exist is unknown and remains one
secutive triangular numbers; again you get a perfect of the great mysteries of mathematics. Should such
square: 1 + 3 = 4 = 22, 6 + 10 = 16 = 42, 10 + 15 = 25 = 52, a number be found, it would be an oddity indeed!
and so on. This is true because The Pythagoreans established these relations, and
many others, by representing numbers as dots ar-
n(n + 1) (n + 1)(n + 2) ranged in various geometric patterns. For example,
+ = (n + 1)2 .
2 2 figure 4.2 shows two triangular arrays, each repre-
senting the sum 1 + 2 + 3 + 4. Taken together, they
Perhaps most surprising of all is the fact that every form a rectangle of 4 × 5 = 20 dots. Therefore, the re-
perfect number is also a triangular number. Thus 6, 28, quired sum is half of that, or 10. Repeating this for
496, and 8,128 are the 3rd, 7th, 31st, and 127th trian- other numbers of dots, it would have been easy for
gular numbers, respectively, and 257,885,160 ⋅ (257,885,161 the Pythagoreans to arrive at the formula 1 + 2 + 3
− 1) is the (257,885,161 − 1)th triangular number.2 Eu- + ⋅⋅⋅ + n = n(n + 1)/2. Similarly, figure 4.3 illustrates
clid, in his Elements (see page 3), proved that if 2n − 1 how they would have established the formula 1 + 3
is prime, then 2n−1 ⋅ (2n − 1) is perfect.3 More than two + 5 + 7 + ⋅⋅⋅ + (2n − 1) = n2, while figure 4.4 demon-
thousand years later, Leonhard Euler proved the strates that the sum of two consecutive triangular
converse: every even perfect number is of the form numbers is always a perfect square. The Pythagore-
2n−1 ⋅ (2n − 1) for some prime value of n. All 48 perfect ans viewed these relations as purely geometric;
numbers known today are even; whether any odd today, of course, we prefer to prove them algebra-
1 2   B e a u t i f u l Geometry

while others are artistic expressions of what a keen

eye can discover in an assembly of dots. Note, in
particular, the second panel in the top row: it illus-
trates the fact that the sum of eight identical triangu-
lar numbers, plus 1, is always a perfect square.4
Figure 4.4
N ot e s :

1. The proper divisors of a number are all positive inte-

ically. Yet, in discovering them, the Greeks sowed gers that divide it evenly, including 1 but excluding the
the seeds that many centuries later would evolve number itself.
into modern number theory, the branch of mathe- 2. To see this, write 257,885,160 ⋅ (257,885,161 − 1) as [257,885,161
matics concerned with the positive integers. ⋅ (257,885,161 − 1)]/2 and let n = 257,885,161 − 1. Then the expres-
Plate 4, Figurative Numbers, is a playful meditation sion has the form (n + 1)n/2, a triangular number.
on ways of arranging 49 dots in different patterns of 3. For more on primes of the form 2n − 1, see chapter 14.
color and shape. Some of these arrangements hint at 4. This is because 8 ⋅ (n + 1)n/2 + 1 = 4n2 + 4n + 1 = (2n +
the number relations we mentioned previously, 2
1) .
 5
The Pythagorean Theorem I

B y any standard, the Pythagorean theorem is the

most well-known theorem in all of mathematics.
It shows up, directly or in disguise, in almost every
lost to us. The Pythagoreans did not leave any writ-
ten records of their discoveries, so we can only spec-
ulate what demonstration he gave. There is, how-
branch of it, pure or applied. It is also a record ever, an old tradition that ascribes to him what
breaker in terms of the number of proofs it has gen- became known as the Chinese proof, so called because
erated since Pythagoras allegedly proved it around it appeared in an ancient Chinese text dating from
500 BCE. And it is the one theorem that almost ev- the Han dynasty (206 BCE – 220 CE; see figure 5.1).
eryone can remember from his or her high school It is perhaps the simplest of the more than 400 proofs
geometry class. that have been given over the centuries.
Most of us remember the Pythagorean theorem The Chinese proof is by dissection. Inside square
by its famous equation, a2 + b2 = c2. The Greeks, how- ABCD (figure 5.2) inscribe a smaller, tilted square
ever, thought of it in purely geometric terms, as a KLMN, as shown in (a). This leaves four congruent
relationship between areas. This is how Euclid stated right triangles (shaded in the figure). By reassem-
it: in all right-angled triangles the square on the side sub- bling these triangles as in (b), we see that the re-
tending the right angle is equal to the squares on the sides maining (unshaded) area is the sum of the areas of
containing the right angle. That is, the area of the squares 1 and 2, that is, the squares built on the sides
square built on the hypotenuse (“the side subtend- of each of the right triangles.
ing the right angle”) is equal to the combined area Elisha Scott Loomis (1852–1940), a high school
of the squares built on the other two sides. principal and mathematics teacher from Ohio, col-
Pythagoras of Samos (ca. 580–ca. 500 BCE) may lected all the proofs known to him in a classic
have been the first to prove the theorem that made book, The Pythagorean Proposition (first published in
his name immortal, but he was not the first to dis- 1927, with a second edition in 1940, the year of his
cover it: the Babylonians, and possibly the Chinese, death). In it you can find a proof attributed to
knew it at least twelve hundred years before him, as Leonardo da Vinci, another by James A. Garfield,
is clear from several clay tablets discovered in Meso- who would become the twentieth president of the
potamia. Furthermore, if indeed he had a proof, it is United States, and yet another by Ann Condit, a
14

Plate 5. 25 + 25 = 49
 5 . T h e P y t h a g o r e a n T h e o r e m I   15

16-year-old high school student from South Bend,

Indiana. And of course, there is the most famous
proof of them all: Euclid’s proof. We will look at it
in the next chapter.
Our illustration (plate 5) shows a 45-45-90-degree
triangle with squares—or what looks like squares—
built on its sides and on the hypotenuse. But wait!
Something strange seems to be going on: 52 + 52 = 72,
or 25 + 25 = 49! Did anything go wrong? Do we see
here an optical illusion? Not really: the illustration
is, after all, an artistic rendition of the Pythagorean
theorem, not the theorem itself; as such it is not
bound by the laws of mathematics. To quote the
American artist Josef Albers (1888–1976): “In sci-
ence, one plus one is always two; in art it can also be
three or more.” Figure 5.1. Joseph Needham, Science and Civilisation in
China, courtesy of Cambridge University Press, Cam-
bridge, UK.

A L B A L B

K K

M M

D N C D N C

(a) (b) Figure 5.2

 6
The Pythagorean Theorem II

T he Pythagorean theorem is listed as Proposition

47 in the first book of Euclid’s Elements. But you
will not find Pythagoras’s name heading it: true to
his terse, matter-of-fact style, Euclid avoided any
reference to persons in his work, instead letting the
geometry speak for itself. So the most famous theo-
rem in mathematics simply became known as Eu-
clid I 47.
Euclid’s proof of I 47 is anything but simple, and
it has tested the patience of generations of students.
In the words of philosopher Arthur Schopenhauer,
“lines are drawn, we know not why, and it afterwards
appears they were traps which close unexpectedly
and take prisoner the assent of the astonished
Figure 6.1. The Pythagorean Theorem in an Arab Text,
reader.” Yet of the 400 or so demonstrations of the
from the Eighth Century. Richard Mankiewicz, The Story
theorem, Euclid’s proof stands out in its sheer aus-
of Mathematics, courtesy of Princeton University Press,
terity, relying on just a bare minimum of previously
Princeton, NJ.
established theorems. Its classic configuration, with
its many auxiliary lines, has become an icon in
nearly every geometry book for the past one thou- triangle has the same area as the rectangle formed by
sand years (see figure 6.1). the hypotenuse and the projection of that side on
At the heart of Euclid’s proof is a double applica- the hypotenuse. Figure 6.2 shows a right triangle
tion of theorem I 38 about triangles of equal area ACB with its right angle at C. Consider the square
(see page 3). But first he proves a lemma (a prelimi- ACHG built on side AC. Project this side on the hy-
nary result): the square built on one side of a right potenuse AB, giving you segment AD. Now construct
 17

Plate 6. Pythagorean Metamorphosis

1 8  

A D B

F E K Figure 6.2
 6 . T h e P y t h a g o r e a n T h e o r e m II   19

AF perpendicular to AB and equal to it in length. sisting of a right angle and the common angle
Euclid’s lemma says that area ACHG = area AFED. ∠BAC). And as congruent triangles, they have the
To show this, divide AFED into two halves by the same area.
diagonal FD. By I 38, area FAD = area FAC, the two Now, what is true for one side of the right triangle
triangles having a common base AF and vertices D is also true of the other side (again, see figure 6.2):
and C that lie on a line parallel to AF. Likewise, di- area BMNC = area BDEK. Thus, area ACHG + area
vide ACHG into two halves by diagonal GC. Again BMNC = area AFED + area BDEK = area AFKB: the Py-
by I 38, area AGB = area AGC, AG serving as a com- thagorean theorem.
mon base and vertices B and C lying on a line paral- Plate 6, Pythagorean Metamorphosis, shows a series
lel to it. But area FAD = 1⁄2 area AFED, and area of right triangles (in white) whose proportions
AGC = 1⁄2 area ACHG. Thus, if we could only show change from one frame to the next, starting with the
that area FAC = area BAG, we would be done. extreme case where one side has zero length and
It is here that Euclid produces his trump card: then going through several phases until the other
triangles FAC and BAG are congruent because they side diminishes to zero. In accordance with Euclid’s
have two pairs of equal sides (AF = AB and AG lemma, the two blue regions in each phase have
= AC) and equal angles ∠FAC and ∠BAG (each con- equal areas, as do the orange regions.
 7
Pythagorean Triples

A triple of positive integers (a, b, c) such that

a2 + b2 = c2 is called a Pythagorean triple; it repre-
sents a right triangle with sides a and b and hypote- b-r

nuse c, all of integer lengths. Some examples are (3,

4, 5), (5, 12, 13), and (8, 15, 17); one can find such b-r
triples even among large numbers: (4,601, 4,800, c
6,649). These four examples are of primitive triples— b r
triples whose members have no common factor
a-r
other than 1. Of course, from any given triple we can r
generate infinitely many others by multiplying it by
an arbitrary integer; for example, the triple (6, 8, 10) r r
is just the triple (3, 4, 5) magnified by a factor of 2.
Such nonprimitive triples represent similar triangles r a-r
and are essentially equivalent. a
Figure 7.1
A Babylonian clay tablet known as Plimpton 322
and dating to about 1800 BCE (now at Columbia
University) lists the hypotenuse and the short side
of 15 Pythagorean triangles, demonstrating that the simple configuration holds within it some surprises.
Babylonians were familiar with the Pythagorean the- Take any right triangle and inscribe in it its incircle.
orem some twelve hundred years before Pythagoras This circle is tangent to all three sides, and its radius
is said to have proved it.1 In Book X of the Elements, is given by the formula
Euclid gives an algorithm for generating every prim-
a +b −c
itive Pythagorean triple (there are infinitely many of r= .
2
them); we give it in the appendix.
It is hard to imagine a simpler geometric structure This follows from figure 7.1: because the two tan-
than a right triangle with integer sides, yet even this gent lines from a point to a circle are of equal length,
 21

Plate 7. The (3, 4, 5) Triangle and Its Four Circles

2 2   B e a u t i f u l Geometry

we have c = (a − r) + (b − r), from which the preceding c +a −b c +b −a c +b +a

ra = , rb = , rc = .
formula follows. 2 2 2
Now, this formula works for any right triangle,
whether Pythagorean or not; but if the triangle is There exist some interesting relations among the
Pythagorean, the radius will always be an integer as four radii; for example, r + ra + rb + rc = a + b + c, that is,
well. This is because in any Pythagorean triple, ei- their sum is equal to the perimeter of the triangle.
ther a, b, and c are even, as in the triple (6, 8, 10), or Another peculiar relation is ra ⋅ rb = r ⋅ rc = (a ⋅ b)/2, the
one of a or b is even, the other odd, and c is odd, as area of the triangle. The proofs of these relations are
in (9, 12, 15) [in a primitive triple, the latter case al- quite simple, and we leave them to the reader.
ways holds, as in (3, 4, 5)]. In either case, a + b − c will Plate 7 shows the (3, 4, 5) triangle (in red) with its
always be even and, therefore, divisible by 2, result- incircle and three excircles (in blue), for which r =
ing in an integer value of r. (3 + 4 − 5)/2 = 1, ra = (5 + 3 − 4)/2 = 2, rb = (5 + 4 − 3)/2 = 3,
But that’s not all. Consider the three excircles of a and rc = (5 + 4 + 3)/2 = 6.
Pythagorean triangle, each being externally tangent
to one side and to the other two sides extended. Sur- N ot e :
prisingly, the radii of all three circles are also inte- 1. See Maor, The Pythagorean Theorem: A 4,000-Year His-
gers. They are given by the formulas tory, chapter 1.
 8
The Square Root of 2

O ne of the most momentous events in the history

of mathematics was the discovery of a new kind
of number that had never been known before—an
But one day a member of the Pythagorean school
by the name Hippasus made a startling discovery:
the square root of 2—the number that when multi-
irrational number. plied by itself results in 2—cannot be expressed as a
To the Pythagoreans, “number” meant either a ratio of two integers, no matter how much one tries.1
positive integer or a ratio of two positive integers, a You can approximate it as closely as you please by
rational number. Examples of such numbers are 2⁄1 rational numbers, but you can never write it exactly
(or simply 2), 3⁄2, and 5⁄3. The Pythagoreans believed as a ratio. For example, 14/10, 141/100, 1,414/1,000,
that any quantity, whether an abstract number or a and 14,142/10,000—or as decimals, 1.4, 1.41, 1.414,
physical entity, is represented by a rational number. and 1.4142—are four rational approximations of 2,
This belief, in all likelihood, came from music, a dis- increasing progressively in accuracy. But to get the
cipline that in ancient Greece ranked equal in im- exact value of 2 would require us to write down an
portance to arithmetic, geometry, and spherics (as- infinite, nonrepeating string of digits, and this can-
tronomy)—the four components of the quadrivium not be expressed as a ratio of integers. Thus, 2 is an
that an educated person was expected to master. irrational number (a proof of the irrationality of 2 is
Pythagoras is said to have discovered that the com- given in the appendix).
mon musical intervals produced by a vibrating The discovery that 2 is irrational shattered the
string correspond to simple ratios of string lengths. Pythagorean belief in the rule of rational numbers,
The octave, for example, corresponds to a ratio of and it brought about a serious intellectual crisis.
2:1; the fifth, to 3:2, the fourth, to 4:3, and so on (the What to do with this new kind of number? Could it
names octave, fifth, and fourth refer to the position be represented geometrically? Take a square of unit
of these intervals on the musical staff). Pythagoras side and draw its diagonal. By the Pythagorean the-
took this as a sign that the entire universe—from the orem, this diagonal has a length equal to 2; but not
laws of musical harmony to the motion of the heav- being able to express it as a rational number, the Py-
enly bodies—is governed by rational numbers. Num- thagoreans were forced to regard it as a purely geo-
ber Rules the Universe became the Pythagorean motto. metric entity—in effect a line segment with an unde-
24

Plate 8. This Is Not the Square Root of Two

 8 . T h e S q u a r e R o o t o f 2    25

fined length! Their confusion can be seen from the are the first few digits of the decimal expansion of
double meaning of the phrase irrational number: a 2, plate 8 provides the answer in lively colors. We
number that is not a ratio of two integers, but also named it This is Not the Square Root of 2, paraphrasing
an erratic number that defies rational behavior. René Magritte’s (1898–1967) famous painting of a
The crisis precipitated by this discovery had far- pipe, which he titled Ceci n’est pas une pipe (“this is
reaching consequences: it opened up a rift between not a pipe”). And, indeed, the long string of deci-
the two major branches of mathematics, geometry mals in our illustration is not the square root of 2,
and arithmetic (and, by extension, algebra), a rift just a close approximation of it.
that would impede the progress of mathematics for
the next two thousand years. It was not until the in-
N ot e :
vention of analytic geometry by Descartes and Fer-
mat in the seventeenth century that the two branches 1. More precisely, the Pythagoreans discovered that the
were reunited. numbers 2 and 1 are incommensurable—they do not have
You have probably seen posters of π or e (the base a common measure. That is, one cannot find two line seg-
of natural logarithms) listing the first few thousand ments of integer lengths m and n such that n ⋅ 2 = m ⋅ 1.
digits of their decimal expansion in row after row of Had such line segments existed, it would mean that
monotonous figures. If you were wondering what 2 = m/n, a rational number.
 9
A Repertoire of Means

A nother subject of great interest to the Pythago-

reans was how to find the average, or mean, of
two positive numbers. At first thought this seems to
and b. For example, the geometric mean of 3 and 5
is 3 ⋅ 5 = 15 ≈ 3.87, a shade less than their arith-
metic mean.
be a trivial question. Let the numbers be a and b. Plate 9 depicts the geometric mean the way the
Add them and divide by 2, getting (a + b)/2: you are Greeks looked at it—as a relation of areas. We think
done. Today, of course, we would compute this of a and b as the lengths of two line segments, and
mean numerically; for example, the mean of 3 and 5 we draw them end to end along a straight line. We
is (3 + 5)/2 = 8⁄2 = 4. The Greeks, however, thought of then draw a semicircle with this line as diameter and
it in geometric terms: they regarded a and b as the erect a perpendicular to it at the point where a and b
lengths of two line segments, drawing them end to are joined. The height of this perpendicular is
end along one line and finding the midpoint of the G = ab , the geometric mean of a and b (we will
combined segment—all doable with straightedge prove this in the next chapter). Therefore, G2 = ab, so
and compass (see chapter 18). that the rectangle with sides a and b has the same
This kind of mean, called the arithmetic mean, is area as a square with side G. The illustration shows a
just one of several possible means. Another often- series of identical semicircles, all with the same di-
used mean is the geometric mean, defined as ab . ameter a + b but with varying proportions a : b. Note
Imagine you own a rectangular plot of land and that G reaches its maximum value when a = b, as
you want to build on it a square-shaped house. shown in the central panel.
Your plot, however, is rather long and narrow, so it And there is a third mean, whose definition may
doesn’t provide enough space for your dream house. at first seem a bit strange. A common situation en-
You therefore propose to your realtor to trade it countered by pilots goes like this: an aircraft can fly
off for a square-shaped plot of equal area. What at 500 mph in still air. But due to the presence of
should be the side of this square? If the length and wind, this so-called airspeed will be different from
width of the rectangle are a and b, its area is ab. Set- the aircraft’s ground speed. On a stretch between two
ting this equal to the area of a square of side x, cities, a pilot encounters a headwind of 50 mph, re-
we get ab = x2, or x = ab —the geometric mean of a ducing the aircraft’s ground speed to 450 mph. On
 27

Plate 9. Geometric Mean

2 8   B e a u t i f u l Geometry

the return flight, the same wind now becomes a tail- This kind of mean, obtained by taking the recipro-
wind, increasing the ground speed to 550 mph. At cals of the two numbers, finding their arithmetic
what speed would the aircraft have to fly in still air to mean, and then taking the reciprocal of the result, is
complete the round trip in the same stretch of time called the harmonic mean. Putting this into the lan-
as when the wind was blowing? guage of algebra, if H denotes the harmonic mean of
Your first impulse might be to give the answer as a and b, we have
500 mph, the aircraft’s own airspeed and the arith-
1 1  1 1 a + b
metic mean of 450 and 550. But first impulses can be =  + = .
H 2a b 2ab
wrong. Let the distance between the cities be d and
the required speed, v. The time it takes the aircraft Taking the reciprocal of this, we get H = 2ab/(a + b).
to complete the outbound stretch (flying against The formulas A = (a + b)/2, G = ab , and H = 2ab/
the headwind) is d/450, whereas the time of the (a + b) comprise the three classical means of the
return stretch is d/550. The total time is, therefore, Pythagoreans.
d/450 + d/550. Setting this equal to the time it would The names arithmetic, geometric, and harmonic for
take to complete the round trip at a constant ground the three means derive from the three classical
speed v, we have progressions with the same names: the arithmetic
d d 2d progression, in which there is a constant difference
+ = . between successive terms; the geometric progres-
450 550 v
sion, which keeps a constant ratio, and the har-
The distance d cancels out, leaving us with the monic progression, whose terms are the reciprocals
equation of those of an arithmetic progression. Here are three
1 1 2 examples:
+ = ,
450 550 v Arithmetic progression: 1, 2, 3, 4, 5, . . .
from which we get Geometric progression: 1, 2, 4, 8, 16, . . .

1 1 1 1  1 450 + 550 1, 000 1 Harmonic progression: 1, 1⁄2, 1⁄3, 1⁄4, 1⁄5, . . .

=  + = ⋅ = =
v 2  450 550  2 450 ⋅ 550 495, 000 495 Each term in these progressions is, respectively,
the arithmetic, geometric, and harmonic mean of the
and, finally, v = 495 mph—slightly less than 500 mph. terms immediately preceding and following it. It may
So the required speed is not the arithmetic mean of have been this connection that caught the Greeks’
the two ground speeds, but a tad less. interest in these means.
 10
More about Means

T urning again to the aircraft making its round trip T

between two cities, we found that the harmonic
mean of the two ground speeds, 495 mph, was less— S
though just barely—than their arithmetic mean, 500
mph. This is not a coincidence. A well-known theo- A
rem says that of the three means, the harmonic mean
is always the smallest, the arithmetic mean the larg- G
est, and the geometric mean somewhere in between.
In the case of the aircraft, the geometric mean of the O Q
P R
two speeds is 450 ⋅ 550 = 497.49 mph, rounded to
a b
two places, so we have H < G < A. This double inequal-
ity can actually be made more general: the less-than
signs become equal signs if and only if the two num-
bers whose mean we are seeking are equal. That is,

H ≤ G ≤ A,

with equality if and only if a = b. This statement is

known as the arithmetic-geometric-harmonic mean in- Figure 10.1
equality.
To see why this is so, let us use a geometric con-
struction, in the spirit of the Greeks. Suppose we are
given two line segments of lengths a and b. Place Now draw a semicircle with center at O and diam-
them end to end along a line, with a = PQ and eter PR (using the same notation as before; again see
b = QR (figure 10.1). Bisect PR and call its midpoint figure 10.1). At Q erect a perpendicular to PR , meet-
O. We have OP = (a + b )/2 = A, the arithmetic mean ing the circle at S. By Thales’s theorem (see chapter
of a and b. 1), ∠PSR = 90° and, therefore, triangles PQS and SQR
30

Plate 10. Mean Constructions

 1 0 . M o r e a b o u t M e a n s    31

S’
are similar, having a common right angle at Q and
equal angles QPS and QSR. Thus, PQ / QS = SQ / QR ,
2
or QS = PQ ⋅ QR = a ⋅ b (we consider all line seg-
ments to be nondirectional, so QS = SQ ). This gives G
us QS = ab = G, the geometric mean of a and b.
To show that the geometric mean of a and b can-
not be larger than their arithmetic mean, we refer Q’
P’ R’
again to figure 10.1. Draw a line segment perpen- A H
dicular to PR at O, meeting the circle at T. Because Figure 10.2
OP is the radius, we have OT = OP = A. A It is then
clear from our figure that QS ≤ OT , that is, G ≤ A.
Furthermore, the two means are equal if and only if P'S' meets the extension of P'Q' at R', with Q'R' = H.
points S and T coincide, in which case PQ = QR , that And since we already know that G ≤ A, it follows
is, a = b. from the similarity of the triangles P'Q'S' and S'Q'R'
What about the harmonic mean? Can it be con- that H ≤ G. Combining the two inequalities, we get
structed as easily as its arithmetic and geometric H ≤ G ≤ A, the arithmetic-geometric-harmonic mean
counterparts? The answer is yes, but in order to do inequality.
so we must first prove a rather surprising result: if It is truly remarkable that the circle— perhaps the
we multiply together the expressions A = (a + b)/2 simplest of all geometric constructs—holds within it
and H = 2ab/(a + b), we get AH = [(a + b)/2] ⋅ [2ab/ so many hidden features waiting to be discovered by
(a + b)] = ab = G2, or G = AH : the geometric mean of a a keen observer. No wonder the Greeks held the cir-
and b is also the geometric mean of A and H. This result cle in such high esteem.
is the key to the construction of H. For, if we rewrite Plate 10, Mean Constructions (no pun intended!),
the equation G2 = AH as a proportion, A/G = G/H, we is a color-coded guide showing how to construct all
see that A and H play the same role vis-à-vis G as did three means from two line segments of given lengths
a and b in our construction of the geometric mean. (shown in red and blue). The arithmetic, geometric,
Thus, having already found A and G from figure and harmonic means are colored in green, yellow,
10.1, construct right triangle P'S'Q' with P'Q' = A and and purple, respectively, while all auxiliary elements
Q'S' = G (figure 10.2). The perpendicular from S' to are in white.
 11
Two Theorems from Euclid

T heorem 35 of the third book of the Elements says,

If in a circle two straight lines cut one another, the rect-
angle contained by the segments of the one is equal to the A
rectangle contained by the segments of the other.
To make sense of this enigmatic statement, we
must understand that the Greeks always used geo-
metric language to describe operations that nowa-
days would be stated in algebraic terms. Thus, “the
rectangle contained by” is code for “the product of”
[the sides of the rectangle]—in other words, the area D
P
of the rectangle. Translated into modern language,
the theorem says: if through a point P inside a circle C
we draw a line that cuts the circle at points A and B,
the product PA × PB is constant; it has the same B Figure 11.1
value for all lines through P (see figure 11.1, where
PA × PB = PC × PD ).
This is followed by theorem 36: If a point be taken
outside a circle and from it there fall on the circle two circle we draw a line that cuts the circle at points A
straight lines, and if one of them cut the circle and the and B, the product PA × PB is constant for all possi-
other touch it, the rectangle contained by the whole of the ble lines through P and is equal to the square of the
straight line which cuts the circle and the straight line in- length of the tangent line from P to the circle (figure
2
tercepted on it outside between the point and the convex 11.2, where PA × PB = PC × PD = PT ). Note that the
circumference will be equal to the square on the tangent. phrase the square on the tangent actually means the
Again, behind this seemingly convoluted lan- area of a square whose side equals the length of the
guage is the statement: if from a point P outside a tangent line.
 33

Plate 11. Circles and Rectangles

3 4   B e a u t i f u l Geometry

C D P

Figure 11.2

gles ∠ACD and ∠DBA subtend the same arc, AD  , on

the circumference (Euclid III, 21). Thus, triangles
PAC and PDB are similar, having two pairs (and,
therefore, three) of equal angles. Consequently,
PA / PC = PD / PB , or PA × PB = PC × PD , which
C
B means that the product PA × PB is constant for all
P possible lines through P.
The proof of theorem 36 follows similar lines,
A and we leave it to the reader to work out the de-
D
tails. The additional statement, that the product
Figure 11.3 PA × PB is equal to the square of the tangent line
from P to the circle, can be thought of as a limiting
case when point A approaches point B, ultimately
to coincide with the point of tangency T. We then
2
The proof of theorem 35 is quite simple. In figure have PA × PB = PT × PT = PT . We should note,
11.3, P is a point inside the circle, and AB and CD are however, that the Greeks would not have accepted
two chords passing through P. We have ∠APC = ∠BPD such a limiting argument into their reasoning be-
and ∠ACD = ∠DBA, the latter equality because an- cause it subtly involves the notions of motion and
 1 1 . T w o T h e o r e m s f r o m E u c l i d    35

continuity, two concepts that Euclid regarded as illustrate theorem 36 and the two lower circles (P in-
physical in nature and thus foreign to pure mathe- side), theorem 35. In each case, PR × PS is repre-
matical thinking. sented by a rectangle of sides PR and PS (shown in
Our illustration (plate 11) shows four identical orange), and since this product is constant for all
circles, each with a pair of chords intercepting at P. possible chords through P, the two rectangles on ei-
The two upper circles (with P outside each circle) ther side of each circle are equal in area.
 12
Different, yet the Same

T he two theorems we just met—numbers 35 and 36

in book III of Euclid—sound strikingly similar:
both are about a circle, a point P, and a line through
Let us denote the two circles by (O, p) and (O, r),
where O is their common center and p = OP and
r = OR are their radii (see figure 12.1, where (O, p) is
P that cuts the circle at points R and S. The two theo- marked in red and (O, r) in blue). We notice that the
rems state that the product PR × PS remains con- two circles can be interchanged by simply switching
stant for all possible lines through P. Yet there is a the roles of P and R and of S and Q. In figure 12.1(a),
2
difference: in theorem 35 P is inside the circle, while P is outside (O, r) and we have PR × PS = PT ; in (b),
2
in theorem 36 it is outside. And theorem 36 has the P is inside (O, r) and we have RP × RQ = RT . But
2
additional result that PR × PS = PT , where PT is note that RQ = PS , because P and Q are symmetri-
the length of the tangent line from P to the circle. So cally positioned with respect to O (as are R and S);
despite their similarity, they are two distinct state- consequently, we can rewrite the last equation as
2
ments, and Euclid was careful to list them as sepa- PR × PS = PT . But this is exactly the same equation
rate theorems, one following the other. as in case (a)! We can, therefore, combine theorems
And yet a closer inspection will show that the two 35 and 36 into a single statement:
theorems are one and the same after all. For when If through a point P a line is drawn that cuts a circle
we say that a point P is given with respect to a circle, (O, r) at points R and S, the product PR × PS is constant
we imply only that the distance OP from P to the for all possible lines through P. If P is outside (O, r), this
center of the circle O is given, and nothing else. The product is the square of the tangent from P to (O, r); if
actual position of P relative to the circle (that is, its P is inside (O, r), it is the square of the tangent from R
direction from O) is immaterial. And since all points to (O, p).
P with the same distance OP from O describe a sec- Plate 12 illustrates the complete symmetry be-
ond circle concentric to the given circle and having tween the two theorems. The top two figures illus-
the radius OP , we conclude that PR × PS has the trate theorem 36, the bottom two, theorem 35. The
same value for all points on this second, “implied” “given” and “implied” circles are marked in blue and
circle. red, respectively; the yellow rectangles and squares
 37

Plate 12. Euclid I, 35 and 36

3 8   B e a u t i f u l Geometry

T T

r p

O p O r
P R
R P

S Q
Q S

(a) (b)
Figure 12.1

represent the product PR × PS and the quantities ties)? Surely he must have noticed the similarity be-
2 2
PT (in the top figure) and RT (bottom figure). The tween the two, but the idea of generalizing several
two configurations are exactly the same, except that particular cases into a broader, sweeping statement
the given and implied circles reverse their roles. was foreign to the Greeks. To Euclid, each case rep-
The French mathematician Lazare Carnot (1753– resented a separate theorem, standing firmly on its
1823) achieved this unification by regarding all line own. Generalizations had to wait for future genera-
segments as directed quantities that can assume posi- tions of mathematicians.1
tive or negative values (for example, PR = − RP ).
This, however, would not have sat well with Euclid,
N ot e :
because the Greeks did not recognize negative quan-
tities. Why, then, didn’t Euclid think of combining 1. This chapter is based on an article by Maor pub-
the two theorems the way we did here (that is, by lished in The Mathematics Teacher (May 1979, pp. 363–
regarding all line segments as nondirected quanti- 367).
 13
One Theorem, Three Proofs

T heorem 13 of Book VI of Euclid tells us how to

find the geometric mean (the mean proportion, as
the Greeks called it) of two line segments. In es-
the hypotenuse, dividing it into segments AD = m
and DB = n . The first proof—the one found in most
geometry textbooks—relies on the similarity of tri-
sence, it says that in a right triangle, the altitude h angles ADC and CDB, since they share a right angle
divides the hypotenuse into two segments m and n at D and equal angles ∠DAC and ∠DCB. Thus,
such that h/m = n/h. From this it follows that h2 = mn, AD / DC = CD / DB , or m/h = h/n, or h2 = mn (as be-
so that h is the geometric mean of m and n. Plate 13 fore, all line segments are nondirectional, so that
illustrates this for m = 9, n = 4, and h = 6. CD = DC ).
We offer here three quite different proofs, with The second proof is a direct consequence of
the question in mind, which of them is the simplest? Thales’s theorem (or rather its converse; see chapter
Let the triangle be ABC, with the right angle at C 1, note 1) and Euclid III 35 (chapter 11). Inscribe
(figure 13.1). From C drop the altitude CD = h to triangle ABC in a circle with diameter AB, as in figure

C C

h h

A B A B
m D n m n

Figure 13.1 h

C’ Figure 13.2
40

Plate 13. 6 2 = 9 x 4
 1 3 . O n e T h e o r e m , T h r e e P r o o f s    41

13.2. This circle passes through C and through its C

mirror image C' when reflected in the diameter.
90°
Hence m ⋅ n = h ⋅ h = h2—QED.
h
The third proof relies on a comparison of areas.
We turn triangle ADC counterclockwise through 90°
A B
about D to get the skewed butterfly-like figure m D n
BCDA'C'D (figure 13.3). Note that angles ∠C'A'D and
∠BCD are equal, and therefore lines BC and A'C' are
C
parallel, with CA' acting as a transversal (Euclid I
27). We now construct square CD'C'D with area h2
and rectangle A'B'BD with area mn, and divide each
h
into two halves by diagonals C'C and A'B. Triangles
C'A'B and C'A'C have the same area, since their verti- C’ B
ces B and C lie on a line parallel to the base A'C' (Eu- h D n
m
clid I 38). From each of these triangles subtract tri-
angle A'DC', resulting in triangles C'DC and A'DB A’
having the same area. But triangle C'DC has half the
area of square CD'C'D, and triangle A'DB has half the
area of rectangle A'B'BD. Thus, h2 = mn.
We return now to the question we posed at the C
D’
beginning of this chapter: which of the three proofs
is the simplest? Judging by their length, the second
proof is the obvious winner, requiring just four lines h
of explanatory text. But length is only one criterion D
C’ B
of what constitutes a simple proof. Another crite- h n
rion—and arguably a more important one—is how m

many previously established propositions the theo- B’

A’ Figure 13.3
rem in question is directly based upon. And this
puts the third proof up front: it rests on just two
earlier propositions—Euclid I 27 and I 38. More-
over, it is in line with the Greek interpretation of a
product as the area of a rectangle, and in this sense proof relies is itself resting on several earlier theo-
it echoes Euclid’s famous proof of the Pythagorean rems. The complete ancestry of a theorem may be
theorem (Euclid I 47). long and arduous, making any absolute judgment of
Of course, even this criterion is not entirely fool- which proof is the simplest almost impossible. Ulti-
proof, as each of the two theorems on which the last mately, simplicity is in the eyes of the beholder.
 14
The Prime Numbers

T he prime numbers have always enjoyed a special

status among the integers. A prime number, or
prime, for short, is an integer greater than 1 that can
therefore, considered the building blocks of all inte-
gers, playing a role somewhat analogous to that of
the chemical elements in the periodic table. In fact,
be divided only by itself and 1. The first 10 primes so fundamental are the primes to mathematics that
are 2, 3, 5, 7, 11, 13, 17, 19, 23, and 29. The smallest— it has been proposed to use them as a means of com-
and the only even prime—is 2. The largest, as of this munication with extraterrestrial beings, if and when
writing, is 257,885,161 − 1, a gargantuan 17,425,170-digit we find them.
number that would fill some 2,500 pages if printed.1 Plate 14 shows a row of dots representing the
An integer greater than 1 that is not prime is called number 17. As much as you may try, you cannot ar-
composite. The number 1 is considered neither prime range these dots in a rectangular array: there will al-
nor composite. ways be at least one left over. That, of course, is be-
The importance of the primes in number theory— cause 17 is prime—it is not the product of any smaller
and by extension, in much of mathematics—comes integers except itself and 1.
from the fact that every positive integer can be factored Primes of the form 2n − 1, like the prime 257,885,161 − 1
into primes in one and only one way. For example, 12 just mentioned, are known as Mersenne primes, after
can be factored into 3 and 4, so 12 = 3 × 4. But 4 itself the Minimite friar and freelance mathematician
can be factored into 2 and 2, so we end up with Marin Mersenne (1588–1648). For 2n − 1 to be prime,
12 = 3 × 2 × 2. We could have started by factoring 12 n itself must be prime, but the converse is false:
into 2 and 6; but 6 = 2 × 3, so we get 12 = 2 × 2 × 3. Ex- 211 − 1 = 2,047 = 23 × 89. The first five Mersenne primes
cept for their order, we end up with the same set of are 22 − 1 = 3, 23 − 1 = 7, 25 − 1 = 31, 27 − 1 = 127, and
primes. This fact, known as the fundamental theorem 213 − 1 = 8,191. To date, only 48 Mersenne primes are
of arithmetic, is true for every positive integer greater known; the largest, discovered in 2013, is the num-
than 1 (although for large numbers the factorization ber quoted earlier. It is not known how many Mer-
process may take a very long time and may even be senne primes exist—or even if their number is finite
impossible to achieve in practice). The primes are, or infinite.
 43

17 = 1 x 17
17 = 2 x 8 + 1

17 = 2 x 7 + 3

17 = 2 x 6 + 5

17 = 3 x 5 + 2

17 = 4 x 4 + 1

17 = 5 x 3 + 2

17 = 6 x 2 + 5

17 = 7 x 2 + 3

17 = 8 x 2 + 1

Plate 14. 17 Is Prime

4 4   B e a u t i f u l Geometry

In chapter 4 we mentioned that if 2n − 1 is prime,

then 2n−1 ⋅ (2n − 1) is perfect. Consequently, each newly
discovered Mersenne prime automatically yields a
new perfect number. Since the largest prime to date,
257,885,161 − 1, is the 48th known Mersenne prime, it
follows that 257,885,160 ⋅ (257,885,161 − 1) is the 48th known
perfect number. Figure 14.1
Many questions about the primes are still unan-
swered, adding to the aura of mystery that has al-
ways surrounded these numbers. Two questions im- come back to this number in chapter 33). Moreover,
mediately come to mind: how many primes are the approximation gets better with increasing n and
there, and what is their law of distribution among approaches n/ln n as n → ∞. It was only in 1896, more
the integers? The first question was answered by Eu- than a hundred years after Gauss’s announcement,
clid around 300 BCE. In a classic proof that became that Jacques Solomon Hadamard of France and
a model of simplicity, Euclid showed that there is no Charles de la Valleé-Poussin of Belgium indepen-
end to the primes: their number is infinite. Although dently proved his conjecture. It is known as the
the primes, on the average, thin out as we go to prime number theorem.
higher numbers, there is no last prime beyond which In years past, the discovery of a new prime often
all integers are composite. We give Euclid’s proof in made the news. In 1963, the University of Illinois
the appendix. issued a special cancellation stamp to commemorate
The second question—the distribution of primes their discovery of the largest prime then known:
among the integers—has intrigued mathematicians 211,213 − 1 (figure 14.1). This 3,376-digit number—the
for centuries. In contrast to the elements of the peri- twenty-third Mersenne prime—was huge at the time
odic table, the primes do not seem to follow any rec- of its discovery but is dwarfed by the primes that
ognizable pattern; in fact, all attempts to find a for- have been discovered since. But rest assured, still
mula that would generate all primes have so far larger primes are lurking just around the corner,
failed. We do know, however, something about their waiting to be discovered by today’s powerful super-
average distribution. In 1792, when he was just 15 computers—or even home computers.
years old, Carl Friedrich Gauss (1777–1855) exam-
ined a table of primes and announced that the num-
N ot e :
ber of primes below a given integer n increases ap-
proximately as n/ln n, where ln stands for natural 1. Source: The Largest Known Primes—A Summary, on the
logarithm (logarithm to the base e = 2.71828 . . .; we’ll Web at http://primes.utm.edu/largest.html, 2013.
 15
Two Prime Mysteries

A mong the many unsettled questions about the

primes, two stand out for their deceptive sim-
plicity. Even a cursory glance at a table of primes
8 = 5 + 3, 10 = 5 + 5 = 7 + 3, and so on. (Note that for
odd numbers this is false: 11 cannot be written as the
sum of two primes). Goldbach conjectured that this
will reveal the abundance of pairs of primes of the is true for all even numbers. He tried to prove it but
form p and p + 2: 3 and 5, 5 and 7, 11 and 13, . . . , failed, so he posed the question to Euler. Euler,
101 and 103, and so on. One can find these twin whose mind was occupied with more pressing math-
primes even among very large numbers: 29,879 and ematical problems, shelved Goldbach’s letter; it was
29,881, 140,737,488,353,699 and 140,737,488,353,701. discovered only after Euler’s death in 1783 among
At the time of writing, the largest known twin pair is his enormous volume of correspondence. Despite
3,756,801,695,685 ⋅ 2666,669 ± 1, each having 200,700 numerous attempts to prove the conjecture or find
digits.1 How many twin primes are there? Mathema- a counterexample, Goldbach’s conjecture remains
ticians are nearly unanimous in their belief that there unsettled.
exist infinitely many of them, just like the primes Until a few decades ago, the primes were consid-
themselves. But belief counts for very little in math- ered the ultimate object of pure mathematics, exist-
ematics; what counts is proof, and thus far no one ing in the ethereal universe of number theory and
has been able to prove this twin prime conjecture. devoid of any practical applications. But that has
The second unanswered question was first raised recently changed: since about 1980 the primes have
by Christian Goldbach (1690–1764), a German played a central role in encoding financial transac-
mathematics teacher and, later, a diplomat in the tions and confidential communications to ensure
service of the Russian czar. In 1742 he wrote a letter the security of the transmitted information over the
to Leonhard Euler, the foremost mathematician of Internet. Thus, it is always possible that a once ob-
the time, in which he reported on a startling obser- scure and abstract subject may suddenly find a prac-
vation: every even number greater than 2 he had ex- tical application in the real world.
amined was the sum of two primes (sometimes in Plate 15.1, Prime and Prime Again, shows a curious
more than one way). For example, 4 = 2 + 2, 6 = 3 + 3, number sequence: start with the top eight-digit
46

73939133
7393913
739391
73939
7393
739
73
7
Plate 15.1. Prime and Prime Again
 47

Plate 15.2. Homage to Martin Gardner

4 8   B e a u t i f u l Geometry

number and keep peeling off the last digits one by in order. If aliens from another planet would look at
one, until only 7 is left. For no apparent reason, each this picture, they would most likely not be able to
number in this sequence is a prime. Plate 15.2 is a read its textual message, but they just might dis-
quote from the eminent popularizer of mathematics cover the hidden primes around it—and perhaps re-
Martin Gardner (1914–2010), whose many articles spond in kind.
on the primes made them a household name. Notice
the dots surrounding the text, starting with the
N ot e :
arrow at the top: if you look carefully, you’ll dis-
cover that the second, third, fifth, seventh, elev- 1. Source: The Largest Known Primes—A Summary, on the
enth, . . . dots are marked in red—the prime numbers Web at http://primes.utm.edu/largest.html, 2013.
 16
0.999 . . . ═ ?

W hen I ask beginning mathematics students, “is

0.999 . . . exactly equal to 1, or only approxi-
mately so?” their responses are usually split evenly,
what’s still left (a quarter of the total distance), then
half of that (an eighth), and so on ad infinitum (see
figure 16.1). “Impossible!” said Zeno. No human
but occasionally the majority will vote for the sec- can exhaust an infinite number of steps, so there will
ond option. Well, let’s see: always be a tiny distance still left to be covered. The
runner will be unable to reach the finish line; in fact,
x = 0.999 …
using the same argument, any kind of motion is im-
10x = 9.999 … .
possible. Yet we do it all the time, giving no thought
Subtract the first equation from the second: at all to the process.
At the heart of the runner’s paradox is the infinite
9x = 9
series
x =1.
1 1 1 1
Surprising? . . . If this simple question can cause + + + + .
2 4 8 16
such disagreement today, how much more so in an-
cient times, when the idea of anything going on to Such a series, in which each term has a fixed ratio
infinity was so confusing that it was avoided entirely, to its predecessor, is called a geometric series (for no
shunned by the Greeks as horror infiniti, the horror particular connection to geometry). In the general
of the infinite. case, a geometric series can be written as
To make the point, the fourth-century BCE phi-
a + ar + ar 2 + ⋅⋅⋅,
losopher Zeno of Elea came up with four para-
doxes—he called them “arguments”—meant to show where a is the initial term and r, the common ratio.
that mathematicians were unable to deal with infin- If r is less than 1 in absolute value (that is, −1 < r < 1),
ity. In one of these paradoxes Zeno purports to show the series will converge—it will reach a definite limit
that motion is impossible. Imagine an athlete about as the number of terms keeps growing. That is to
to run a 1-mile stretch. To do so, the runner would say, by adding more and more terms, the sum will
first have to cross the halfway point, then half of get ever closer to its limiting value, making the re-
50

Plate 16. 0.999... = 1

 1 6 . 0 . 9 9 9 … = ?    51

0 1

0 1/2 1

0 1/2 1/4 1

0 1/2 1/4 1/8 1 Figure 16.1

maining difference as small as we please. In the case and so on. The sum seems to be getting ever closer
of the runner, we have to 1, the limiting value of the series.
Of course, adding the first few terms of our series
1
= 0.5, does not prove that it converges to 1—or even that a
2 limiting sum exists in the first place. Returning to
1 1 3
+ = = 0.75, the general case, it is not difficult to show that as
2 4 4 long as −1 < r < 1, the limiting sum is a/(1 − r) (we
1 1 1 7 give the proof in the appendix). For the runner’s
+ + = = 0.875,
2 4 8 8 paradox we have a = 1⁄2 and r = 1⁄2, so the sum is (1⁄2)/
1 1 1 1 15 (1 − 1⁄2) = 1, and the runner will reach the finish line
+ + + = = 0.9375,
2 4 8 16 16 just fine.
5 2   B e a u t i f u l Geometry

Going back to the question we posed at the be- convincing refutation of the paradoxes. And no
ginning of this chapter—is 0.999 . . . exactly equal to wonder: to explain them, one must first accept the
1, or only approximately so?—we note that the re- existence of infinity as a mathematical reality, a
peating decimal 0.999 . . . is actually a geometric se- mental leap that even nineteenth-century mathe-
ries, 9⁄10 + 9⁄100 + 9⁄1,000 + ⋅⋅⋅ , whose initial term is a = 9⁄10 maticians were not quite ready to take. It took the
and whose common ratio is r = 1⁄10. Since this com- insight of a relatively unknown genius by the name
mon ratio satisfies the condition −1 < r < 1, the series Georg Cantor to take this crucial step, and in
will converge to the limit (9⁄10)/(1 − 1⁄10) = (9⁄10)/ doing so he revolutionized our understanding of
(9⁄10) = 1, settling the issue once and for all. infinity. We will take a closer look at Cantor’s ideas
Zeno’s paradoxes caused a stir in the mathemati- in chapter 51. However, if you are still wondering
cal community that lasted well over two thousand how a string of nines can make up a 1, plate 16,
years. Despite volumes of arguments, mostly philo- 0.999 ... = 1 will provide an answer—albeit a whim-
sophical or religious, no one was able to offer a sical one.
 17
Eleven

T here is a parody about a mathematician who

tries to prove that all numbers (here meaning
positive integers) are interesting. Assume not. The
seems to shun 11: flowers with 2, 3, 4, 5 and 6 petals
are very common, but not with 11. Nor does it play
a role in the inorganic world of crystals and miner-
number 1 is certainly interesting, being the genera- als. In Peter Stevens’s exhaustive Handbook of Regu-
tor of all numbers. So is 2, the first even integer lar Patterns (see the bibliography), with its hundreds
and the only even prime. Three, being the sum of of designs taken from all aspects of art and nature,
1 and 2, makes it interesting as well. What about not a single pattern is based on 11.
4? We have 4 = 2 + 2 = 2 × 2 = 22: no doubt about it, 4 Still, the excluded number has its advocates: as
is definitely interesting. And so it goes, until we ar- told in Genesis 37:9, 11 stars appear in Joseph’s
rive at the first uninteresting number. But this, of dream, together with the Sun and Moon—perhaps a
course, makes it interesting! Thus, all numbers are reference to the 11 constellations of the zodiac visi-
interesting—QED. ble on any given night, the twelfth being hidden by
If we had to choose an uninteresting number, 11 the Sun. The Susan B. Anthony dollar coin, still in
would certainly be a candidate. Tucked unceremoni- circulation but rarely used, is framed by an 11-sided
ously between its two more famous neighbors, 10 regular polygon (figure 17.1). And of course, 11 is
and 12, it seems to lack any defining character­istics: the number of players in a football team—American
it is not a member of any immediately re­cognizable football as well as soccer.
number sequence,1 nor is it a perfect square or a sum Eleven does have some mathematical claim to
of two squares. In mythology, too, 11 seems to have fame: it is the second repunit—a number, all of whose
acquired a negative reputation: in ancient Rome, an digits are 1. These numbers are denoted by Rn, where
assembly of 11 men was charged with apprehending n is the number of 1s. So R1 = 1 = (10 − 1)/9, R2 = 11 =
criminals and bringing them to justice—a precursor (100 − 1)/9, and, in general, Rn = (10n − 1)/9. These re-
of our modern jury system. The sixteenth-century punits have some remarkable properties: not only
numerologist Petrus Bungus deemed 11 as having are all repunits palindromes (numbers that are the
“no connection to divine things, no ladder reaching same whether read forward or backward), but so are
up to things above, nor any merit.”2 Even nature their squares, up to R9:
5 4  

Plate 17. Celtic Motif 1

 1 7 . E l e v e n    55

1,111,111,111,111,111,111, of which the first and

third factors are themselves repunits, R2 and R19,
while the second factor, except for the trailing 1, has
a similar digit structure as that of 1 / 11 = 0.09 .
As with Mersenne numbers (see page 42), Rn can
be prime only if n is prime, but the converse is false:
R3 = 111 = 3 × 37. As of this writing, only five prime
repunits are known: R2, R19, R23, R317, and R1031, the
last discovered in 1986 by Hugh C. Williams and
Harvey Dubner. In addition, there are several “prob-
Figure 17.1
able primes” whose primality has still to be con-
firmed. And like the Mersenne primes, it is unknown
how many repunit primes exist—or even if their
number is finite or infinite.4
Our illustration (plate 17) shows an intriguing
lace pattern winding its way around 11 dots ar-
R12 = 1 ranged in three rows; it is based on an old Celtic
R =
2
2 121 motif. We hope this excursion into an “unpopular”
number will encourage the reader to search for it in
R =
2
3 12321
other places and be rewarded with discovering the
R =
2
1234321
4 unexpected.5

R = 12345678987654321
2
9
N ot e s :
Eleven is one of a small number of primes that 1. It is, however, the fifth member of the Lucas num-
have simple divisibility rules. Take the number 1,529 bers—a Fibonacci-like sequence (see chapter 20) that starts
and alternately add and subtract its digits, going with 1 and 3: 1, 3, 4, 7, 11, 18, 29, . . . .
from left to right: 1 − 5 + 2 − 9 = −11. Since the result is 2. Annemarie Schimmel. The Mystery of Numbers (New
divisible by 11, so is the number itself (indeed, York: Oxford University Press, 1993), p. 189.
1,529 = 11 × 139). You can check this for as many num- 3. Sometimes the result may be 0, as, for example, with
bers as you wish: it always works.3 Repunits in gen- 187. Since 0 is divisible by 11 (0 = 11 × 0), so is 187; indeed,
eral follow some simple divisibility rules: No repunit 187 = 11 × 17.
is divisible by 2 or 5; it is divisible by 3 if and only if n 4. Source: Wolfram MathWorld, on the Web at http://
is a multiple of 3; by 7 and by 13 if and only if n is a mathworld.wolfram.com/Repunit.html, 2013.
multiple of 6; and by 11 if and only if n is even. Among 5. This chapter is based on an article by Maor in the
composite repunits, R38 has a particularly interesting journal Mathematics Teaching in the Middle School (January
prime factorization: 11 × 909,090,909,090,909,091 × 2002).
 18
Euclidean Constructions

A ccording to tradition, it was Plato (ca. 427–347

BCE) who decreed that all geometric construc-
tions should be done with a straightedge (an un-
C

marked ruler) and compass alone. Of course, there

is nothing intrinsically special about these tools, ex-
cept perhaps their simplicity (you can still get them D A B E
for a dollar or two at any drugstore), but Plato made
their use into an art. Hundreds of constructions can
be done with them, from very basic drawings to
highly complex designs. Indeed, straightedge and
compass constructions became so fundamental to
geometry that Euclid incorporated them in his Ele-
Figure 18.1
ments from the very beginning. Proposition 1 of
Book I—the very first of the 465 theorems in his Ele-
ments—shows us how to construct an equilateral tri-
angle when its side is given. It is the same construc-
tion that generations of students have learned in We should note that Euclid’s compass was dif­
their geometry class: ferent from the familiar modern compass: it “col-
lapsed” when lifted from the paper and, therefore,
Let the given side be line segment AB (figure could not be used to transfer line segments from
18.1). With A as center, draw circle BCD. With B one place to another. We do not know whether Eu-
as center, draw circle ACE. The two circles meet clid actually used it as a physical tool or whether he
at C. Now join C with A and with B. We have intended it merely as an abstract device with which
AB = AC and BA = BC ; therefore, AB = BC = one could do the construction in principle. Which-
CA, proving that the required triangle is indeed ever the case, constructions with the straightedge
equilateral. and compass—jointly known as the Euclidean tools—
 57

Plate 18. Seven Circles a Flower Maketh

5 8   B e a u t i f u l Geometry

became the subject of endless fascination, “a geomet- The result is now known as the Mohr-Mascheroni
ric solitaire which over the ages has attracted hosts of theorem.
players, and though now well over two thousand By itself, a geometric construction is a stark,
years old, has lost none of its singular charm and black-and-white array of lines and circles. But add
appeal.”1 color to it, and it can become an exquisite work
In fact, you don’t even need a straightedge. As the or art, as plate 18, Seven Circles a Flower Maketh,
Danish geometer Jørgen Mohr (1640–1697) proved shows. In the coming chapters we will look at some
in 1672, every construction that can be done with a par­
ticular constructions, among them the regular
straightedge and compass can be done with a com- pentagon.
pass alone—provided you think of a line as given by
the two intersection points of a pair of circles. Be-
N ot e :
cause Mohr published his result in Danish rather
than Latin—the language of scientific discourse at 1. The quote is from Howard Eves, A Survey of Geometry,
the time—it received little attention until it was re- p. 154. It might be supposed that the modern compass,
discovered in 1797 by the Italian Lorenzo Masche- being capable of transferring distances, can do more than
roni (1750–1800). It was only by a curious incident its collapsible predecessor, but this is not the case: it can be
that Mohr’s original theorem came to light, when a shown that the two are completely equivalent in the sense
young mathematics student found a copy of his that each can do everything the other can, although per-
work in a secondhand bookstore in Copenhagen. haps requiring more steps. For a proof, see Eves, p. 155.
 19
Hexagons

A regular polygon is a convex polygon whose sides

all have the same length and meet each other at
the same angle. Next to the equilateral triangle, the
D C

simplest regular polygon to construct—using only

the Euclidean tools—is the six-sided hexagon. Let
the side AB be given (figure 19.1). Draw a circle with
center at A and radius AB , place the point of your
compass at B, and without changing the compass’s A B
opening, swing an arc, cutting the circle at C. Now E
place your compass at C and, with the opening still
the same, swing a second arc, cutting the circle at D.
Repeat the process three more times, resulting in
points E, F, and G (if you do it one more time, the
last point should coincide with B—provided, of
course, that your drawing was exact). With a straight-
G
edge, connect pairs of adjacent points and you get a F
perfect hexagon, with its high degree of symmetry Figure 19.1
(six 60° rotations and six reflections).
The hexagon is one of only three regular poly-
gons that can tile, or tessellate the plane—fill it com-
pletely without gaps or overlaps. The other two are a good example is the paving of the subway stations
the square and the equilateral triangle; but since a of Washington, DC.
hexagon can be dissected into six equilateral trian- Suppose you want to place rows of identical coins
gles, the hexagonal and triangular tessellations are on a table so that each coin touches its immediate
not really different. Hexagonal tiling, while not as neighbors. This can be done in two ways: either the
common as square tiling, can be seen at many places; coins in successive rows are placed above one an-
60

Plate 19. Parquet

 1 9 . H e x a g o n s    61

spheres, resulting in a density of π / 18 spheres per

unit volume, or slightly more than 74 percent. Amaz-
ingly, his conjecture—well known to any fruit ven-
dor who packs a pile of oranges in a box—remained
unproved until 1998, when the American mathema-
tician Thomas Hales used a computer program to
Figure 19.2 exhaust the large number of possible cases.
The hexagon was known to humans for thousands
of years, as evidenced by the six-spiked wheels of Bab­
ylonian and Egyptian chariots. Nature, too, takes ad-
vantage of the hexagon’s high degree of symmetry.
Snowflakes, with their infinite variety of fine struc-
ture, invariably crystallize into perfect hexagons, a
fact that never fails to fascinate nature lovers. And
then there is the honeycomb, whose occupants, the
Figure 19.3 hardworking honeybees, diligently shape their wax-
made habitat into hexagonal prisms, making this
shape a fitting logo on honey products.
Plate 19, Parquet, seems at first to show a stack of
other, so that each coin is surrounded by eight identical cubes, arranged so that each layer is offset
neighbors whose centers form a square (figure 19.2); with respect to the one below it, forming the illusion
or else each row is offset with respect to the row of an infinite, three-dimensional staircase structure.
below it, the coins in the second row partially filling But if you look carefully at the cubes, you will notice
the space between those of the first row (figure 19.3). that each corner is the center of a regular hexagon. I
In this second arrangement, every coin is surrounded still remember my amazement when, as a physics
by six neighbors, whose centers form a hexagon. student, I first noticed this sixfold symmetry in an
The resulting array is not only more stable but also object that at first thought should only have two-
more efficient—it packs more coins per unit area. and fourfold symmetries. It takes a while for the eye
The analogous configuration in three dimensions, to recognize this, as the entire array seems to jump
with spheres replacing the coins, has puzzled math- in and out of space, at one moment appearing as if
ematicians for more than three hundred years. The the cubes point straight at you, only to reverse their
great German astronomer Johannes Kepler (1571– orientation at the next. An optical illusion? A trick
1630) conjectured in 1611 that this arrangement that the brain plays with our eyes? I let the reader
represents the most efficient packing of identical decide.
 20
Fibonacci Numbers

A lmost anyone with the slightest interest in math-

ematics will be familiar with the name Fibo-
nacci. Leonardo of Pisa—he later adopted the name
rabbits can produce, assuming that a pair becomes
productive from the second month on and gives
birth to a new pair every subsequent month. This
Fibonacci (son of Bonacci)—was born in Pisa around leads to the sequence of numbers 1, 1, 2, 3, 5, 8, 13,
1170, the son of a wealthy merchant. Pisa at that 21, 34, . . ., in which each number from the third
time was an important commercial center, serving term on is the sum of its two predecessors.1 The Fi-
both Christian Europe and Moslem Middle East bonacci sequence, as it became known, grows very
and North Africa. Fibonacci was thus acquainted fast: the tenth member is 55, the twentieth is 6,765,
with the newly invented Hindu-Arabic numeration and the thirtieth is 832,040. In his famous problem
system, with the numerals (or “ciphers”) 0 through 9 Fibonacci asked how many rabbits will there be
as its centerpiece. Convinced that this system was after one year. The answer is 144, the twelfth Fibo-
superior to the cumbersome Roman numerals, he nacci number.
wrote a book entitled Liber Abaci (“Book of the Fibonacci hardly could have anticipated the stir
Calculation,” sometimes translated as “Book of the his little puzzle would create. The sequence enjoys
Abacus”), in which he advocated the new system numerous properties—so many, in fact, that a schol-
and explained its operation. Published in 1202, it arly journal, the Fibonacci Quarterly, is entirely de-
became an instant bestseller and was in no small voted to it. Fibonacci numbers seem to appear where
measure responsible for the acceptance of the new you least expect them. For example, the seeds of a
system by European merchants and, eventually, by sunflower are arranged in two systems of spirals, one
most of the learned world. winding clockwise, the other, counterclockwise. The
So it is ironic that Fibonacci’s name is remem- number of spirals in each system is always a Fibo-
bered today not for the main thrust of his influential nacci number, typically 34 one way and 55 the other
book—promoting the Hindu-Arabic numeration (see figure 20.1), with occasional higher numbers.
system—but for a little problem he posed in it, per- Smaller Fibonacci numbers also show up in the
haps as a recreational exercise. The problem deals scales arrangement of pinecones and the leaf pat-
with the number of offsprings a hypothetical pair of terns of many plants.
   6 3

Plate 20. Girasole

6 4   B e a u t i f u l Geometry

55
34

Figure 20.1

Among the purely mathematical properties of the nouncing the answer, 143, while they are still doing
Fibonacci numbers, we mention here just one: the their sums. It always works! (See the appendix for a
sum of the first n members of the sequence is always proof.)
equal to the next-to-next member, minus 1; that is, Perhaps most surprising of all is a discovery made
in 1611 by Johannes Kepler: divide any member of
F1 + F2 + F3 + ⋅⋅⋅ + Fn = Fn+2 − 1.
the sequence by its immediate predecessor. As you
For example, the sum of the first 8 Fibonacci num- do this with ever-increasing numbers, the ratios
bers is the tenth number minus 1: 1 + 1 + 2 + 3 + 5 + 8 seem to converge to a fixed number, a limit:
+ 13 + 21 = 54 = 55 − 1. You can use this fact to surprise
your friends by asking them to find the sum of, say, 2 3 5 8
= 2, = 1.5, = 1.666 …, = 1.6,
the first 10 Fibonacci numbers. Most likely they will 1 2 3 5
start by adding the terms one by one, a process that
will take some time. But knowing that the twelfth 13 21
= 1.625, = 1.615…, …
Fibonacci number is 144, you can outdo them by an- 8 13
 2 0 . F i b o n a c c i N u m b e r s    65

This limit, about 1.618, turns out to be one of rectangles whose dimensions are exactly the Fibo-
the most famous numbers in mathematics, nearly nacci numbers. The word Girasole (“turning to the
the equal in status to π and e. Its exact value is (1 + sun” in Italian) refers to the presence of these num-
5 )/2. It came to be known as the golden ratio (sectio bers in the spiral arrangement of the seeds of a sun-
aura in Latin), and it holds the secret for construct- flower—a truly remarkable example of mathematics
ing the regular pentagon, as we will see in chapter 22. at work in nature.
Plate 20, Girasole, shows a series of squares, each
of which, when adjoined to its predecessor, forms a
N ot e :
rectangle. Starting with a black square of unit
length, adjoin to it its white twin, and you get a 2 × 1 1. The sequence is sometimes counted with 0 as the first
rectangle. Adjoin to it the green square, and you get member. It can also be extended to negative numbers: . . .
a 3 × 2 rectangle. Continuing in this manner, you get 5, −3, 2, −1, 1, 0, 1, 1, 2, 3, 5, . . ..
 21
The Golden Ratio

S uppose you are being asked to divide a line seg-

ment into two parts such that the whole segment
is to the longer part as the longer part is to the shorter.
which, after rearranging, yields the quadratic equa-
tion x2 + x − 1 = 0. This equation has two solutions,
one positive and one negative; but since x stands for
The Greeks were greatly intrigued by this seemingly length, it cannot be negative. Using the familiar
simple problem, but exactly why is not quite clear: quadratic formula and taking only the positive solu-
perhaps it was posed by an anonymous scholar as an tion, we get
exercise to his students, or it may have arisen from −1 + 5
the challenge of constructing a regular pentagon x= ,
2
with straightedge and compass (see the next chap-
ter). Whatever its origins, this particular division of a or about 0.618. The golden ratio ϕ, by definition, is
line segment into two parts became known as the 1/x. A little arithmetic manipulation will show that
golden section (sectio aura in Latin). The ratio between 1/x = (1 + 5)/2, which, you will notice, differs from
the lengths of the two parts is called the golden ratio x by exactly 1. Thus, the decimal value of ϕ is about
and is usually denoted by the Greek letter ϕ (phi), 1.618.
although some authors denote it by τ (tau). The number ϕ enjoys many interesting properties,
Let the line segment be of unit length (figure some quite surprising. We have already noticed that
21.1). Denoting the length of the longer part by x, ϕ = 1 + 1/ϕ. Multiplying both sides of this equation
the problem leads to the equation by ϕ results in ϕ2 = ϕ + 1. Multiplying again by ϕ gives
us ϕ 3 = ϕ2 + ϕ = (ϕ + 1) + ϕ = 2ϕ + 1. This process can be
1 x repeated; replacing ϕ2 by ϕ + 1 at each step and col-
= ,
x 1−x lecting like terms, we get the sequence:

x 1- x

1 Figure 21.1
 67

1.618033988749894848204
1, 1, 2, 3, 5, 8, 13, 21, 34,
5868343656381177203091
55, 89, 144, 233, 377,
7980576286213544862270
610, 987, 1597, 2584,
5260462818902449707207
4181, 6765, 10948,
2041893911374847540880
17713, 28661, ...
7538689175212663386222
3536931793180060766726
28661 : 17713 = 1.618... 3544333890865959395829
17713 : 28661 = 0.618... 0563832266131992829026
788067520876689250171...

φ11 = 1φ
φ22 = 1φ + 1 a b

1 φ33 = 2φ + 1 a : b = (a + b) : a
φ44 = 3φ + 2 φ= 1+ 1+ 1+ 1+ ... >
φ55 = 5φ + 3 a = 61.8 % b = 38.2 %
2 φ66 = 8φ + 5
...

1+ 5
φ=
2

Plate 21. The Golden Ratio

6 8   B e a u t i f u l Geometry

ϕ 2 = ϕ + 1, ϕ3 = 2ϕ + 1, ϕ4 = 3ϕ + 2, ϕ5 = 5ϕ + 3, you accept approximations into the picture, you are

ϕ6 = 8ϕ + 5, . . . . opening the door to endless speculation, which is
indeed what happened with the golden ratio. Re-
The coefficients in these expressions turn out to gardless of whether its connections to art and archi-
be none other than the Fibonacci numbers! In the tecture are purely accidental or are anchored in some
previous chapter we saw that the ratio of two con- hidden principle, they have greatly added to the
secutive Fibonacci numbers approaches the golden aura of mystique that surrounds this number. No
ratio as we move higher up in the sequence; now we wonder medieval scholars dubbed it the divine
have a second example of how seemingly unrelated proportion.
mathematical objects may in fact be intimately con- Plate 21 showcases a sample of the many occur-
nected. Who would have expected that the golden rences of the golden ratio in art and nature. Some of
section—a purely geometric entity—would have any- the panels describe scenes mentioned earlier in this
thing to do with the Fibonacci numbers, whose ori- chapter. For the remaining panels, we’ll use a “coor-
gin is in number theory? dinate” notation (x, y) to identify each panel, x
The golden section has also found its way into art standing for the row number (counting from top to
and architecture. It has been claimed that the bottom) and y for the column (from left to right).
Greeks, in their obsessive quest for aesthetic perfec- In panel (1, 1) we see a regular pentagon and its
tion, have examined rectangles of various propor- associated pentagram, with a smaller pentagon
tions and found that the rectangle whose length-to- nested inside; as we will see in the next chapter, the
width ratio is equal to the golden ratio appeared to golden ratio is the key to constructing a regular pen-
be the most aesthetically pleasing. They may have tagon with straightedge and compass.
used this proportion in the construction of their Panel (2, 1) shows one way of dividing a line seg-
temples. The famous Parthenon in Athens has the ment (here of length 2) in the ratio ϕ:1; the dot on
approximate proportions of the golden ratio, but the base line marks the point of division. Panel (4, 2)
whether it was built specifically with this ratio in shows another way; the horizontal line is parallel to
mind is the subject of an ongoing debate.1 Leonardo the triangle’s base and cuts the two lateral sides at
da Vinci noticed that if a man stretches his hands to their midpoints.
their full extent eagle-manner, the ratio of his height Panel (2, 3) is rather intriguing. We see an infinite
to the wingspan of his hands is approximately 1.6— succession of nested square roots, with 1 added to
very close to ϕ. And in our own time, someone with each radical before taking the next, like a Russian
a keen eye has noticed that the dimensions of a stan- Matryoshka doll inside which resides a smaller doll, a
dard credit card are 8.5 × 5.3 cm, resulting in a still smaller one inside that, and so on. Surprisingly,
length-to-width ratio of 1.60377—within less than 1 this infinite expression converges to ϕ.2
percent of ϕ. Panels (3, 1), (3, 2), and (3, 3) bring the golden
But how close is close? Is 3 a close approximation ratio into three-dimensional space. Take three iden-
to π? If so, then we can find a connection to π when- tical cards, each of length-to-width ratio equal to ϕ.
ever 3 shows up, which is to say, everywhere. Once Cut a slit along the centerline in two of the cards,
 2 1 . T h e G o l d e n R a t i o    69

equal in length to the card’s width. Insert the cards

N ot e s :
into each other as shown, resulting in a three-dimen-
sional structure with 12 corners. Connect pairs of 1. On this subject see Livio, The Golden Ratio.
adjacent corners, and you get an icosahedron, a 2. To see this, let us denote the entire expression by x
polyhedron with 20 equilateral triangles as faces— (assuming, of course, that the nested radicals indeed con-
one of the five regular solids we met in chapter 4. verge to a limit). Since the same x also appears inside the
Interestingly, the two structures appear on the logo radical, we have the equation 1 + x = x . Squaring both
of two organizations devoted to mathematical re- sides and rearranging terms, we get the quadratic equa-
search and teaching—the stacked cards as the icon tion x2 − x − 1 = 0, whose positive solution is (1 + 5 )/2, the
of the German DFG Research Center Matheon and golden ratio (we take only the positive solution, because
the icosahedron on the logo of the Mathematical As- by definition the symbol a is the positive solution of the
sociation of America. equation x2 = a).
 22
The Pentagon

T he five-sided pentagon has fascinated mathemati- C

cians for generations, and it still fascinates them
today. In contrast to the three-, four-, and six-sided 36°
regular polygons, it is not at all obvious how to con-
struct a regular pentagon, a fact that presented a
challenge to the Pythagoreans. The secret to its con- O
struction is the 72−72−36-degree triangle ABC formed
by any of the pentagon’s five sides and the vertex 72°
opposite to it (figure 22.1).1 This triangle is known
as the golden triangle because its side-to-base ratio is
72° 72°
exactly the golden ratio ϕ = (1 + 5)/2. Since this A B
ratio can be constructed using a straightedge and Figure 22.1
compass, the pentagon can be constructed too (we
give the proof and construction in the appendix).
The pentagon is related to the golden ratio in
other ways as well. For example, if you connect its In medieval times fortresses were often built in a
five vertices by straight lines, you get a pentagram, a pentagonal shape, supposedly because it afforded
five-cornered star polygon, inside which resides a the best firing field from the watchtowers at its cor-
smaller pentagon, an exact replica of the original ners. But the most famous pentagonal fortress is of
but with side ϕ 2 times smaller. This process can be modern vintage: the Pentagon in Washington, DC.
repeated, generating ever smaller pentagons and Theories abound as to why this unusual shape was
pentagrams whose sides decrease as 1/ϕ 2, 1/ϕ4, chosen for the world’s largest defense department:
1/ϕ 6, . . . (see plate 22). No wonder the Pythagoreans economy in building materials, efficient use of space,
were enchanted by the pentagram, with its many or a tribute to the ancient Pythagorean emblem. The
hidden secrets—so much so that they chose it as truth is more prosaic: when construction began in
their emblem. 1941, the only suitable lot within reasonable dis-
 71

Plate 22. Pentagons and Pentagrams

7 2   B e a u t i f u l Geometry

tance from the capital was tucked between the Po- a cluster of small pentagonal-shaped metal pen-
tomac River, one of its tributaries, and three access nants with the emblem of the Soviet Union, making
roads, an area roughly forming a pentagonal shape. the pentagon the first human-made object to reach
So a pentagon it would be, prompting Hermann another world.
Weyl, in his classic book Symmetry, to comment: “By
its size and distinctive shape, it provides an attrac-
N ot e :
tive landmark for bombers.” Fifty years after he
wrote those words, on September 11, 2001, his pre- 1. To see why this is a 72−72−36-degree triangle, in-
monition tragically came true. scribe the pentagon in a circle with center O. We have
The pentagon also made history of a different ∠AOB = 360°/5 = 72°, so ∠ACB = 36° because its vertex is
kind. When the Soviet spacecraft Luna 2 crashed on  , as
on the circumference and subtends the same arc, AB
the Moon on September 14, 1959, it scattered around ∠AOB.
 23
The 17-Sided Regular Polygon

O nce we have constructed a regular polygon of n

sides—an n-gon, for short—it is easy to con-
struct a regular polygon with twice as many sides—a
A
P

2n-gon: inscribe the n-gon in a circle with center at O

(figure 23.1 shows this for the hexagon). Let A and B
be two adjacent vertices of this n-gon. Bisect ∠AOB
and extend the bisector until it meets the circle at P.
Connect P to either A or B, and you have one side of
the 2n-gon. So from the 3-, 4-, and 5-sided gons we
can get the 6-, 8-, 10-, and 12-sided gons. Add to that
the 15-sided gon (for reasons to be explained O
shortly), and you have the complete list of regular
n-gons the Greeks were able to construct with Eu-
clidean tools.
Imagine the surprise when 19-year-old Carl Fried-
rich Gauss, just beginning his academic career, an-
nounced in 1796 that he could construct a regular
gon of 17 sides! Gauss (1777–1855) would soon be
recognized as one of the greatest mathematicians of
all time, but when he made his sensational discovery Figure 23.1
he was still unknown. In fact, he had previously
planned to become a linguist, but his surprising dis-
covery convinced him that mathematics was his true
calling, and so it was.
Before you attempt to construct a 17-sided poly- of the daunting task awaiting you: the length of each
gon inscribed in a unit circle, you should be warned side of your 17-gon is given by the formula
74

4+
17 = 2 1
Plate 23. Homage to Carl Friedrich Gauss
 2 3 . T h e 1 7 - S i d e d R e g u l a r P o l y g o n   7 5

1 nonnegative k. Indeed, for k = 0, 1, 2, 3, and 4 we get

34 − 17 − 34 − 2 17 − 2 17 + 3 17 1
4 Fk = 3, 5, 17, 257, and 65,537—all primes. But in 1732
Leonhard Euler proved Fermat wrong: for k = 5 we
7 + 170 − 26 17 − 4 34 + 2 17 , 5
get22 + 1 = 4, 294, 967, 297 = 641 × 6, 700, 417, a com-
an expression that is sure to give the chills to anyone posite number. To this day it is not known if any
thinking of pursuing a career in mathematics!1 other Fermat primes exist, so it is possible that there
Gauss actually did more: he showed that it is pos- are other, as yet undiscovered, regular gons that can
sible, in principle, to construct a regular n-gon if n is be constructed with the Euclidean tools. Needless to
a product of a nonnegative power of 2 and distinct say, if such polygons do exist, they must have a huge
k
primes of the form Fk = 22 + 1, where k is a nonnega- number of sides, making any practical construction
tive integer. Primes of this form are known as Fermat totally out of the question. Gauss, however, was not
primes, named after the great French number theo- so much concerned about the actual construction;
rist Pierre de Fermat (1601–1665). Fermat conjec- what mattered to him was the laws that govern such
tured that this expression yields a prime for every a construction.

Figure 23.2
7 6   B e a u t i f u l Geometry

Gauss’s achievement is immortalized in his Ger- It is extremely rare to find examples of a 17-sided
man hometown of Brunswick, where a large statue regular gon in nature, let alone in art or architec-
of him is decorated with an ornamental 17-pointed ture. But if you search around long enough, you will
star (plate 23 is an artistic rendition of the actual star eventually find what you were looking for. One of us
on the pedestal, which has deteriorated over the (Jost) recently discovered in a shopping mall in the
years); reportedly the mason in charge of the job town of Leipzig a 17-sided glass dome, with a 17-
thought that a 17-sided polygon would look too sided pattern decorating the floor under it (figure
much like a circle, so he opted for the star instead.2 23.2).
In 1837, Pierre Laurent Wantzel (1814–1848) proved
that the polygons Gauss identified are in fact the
N ot e s :
only ones constructible with Euclidean tools, mak-
ing Gauss’s discovery a necessary and sufficient con- 1. For further discussion of the 17-gon, see Hartshorne,
dition for the construction. Thus, a regular polygon Geometry, chapter 29.
of 15 sides is constructible, because 15 = 3 × 5, and 2. I wish to thank Professor Manfred Stern of Halle,
0
both 3 and 5 are Fermat primes (3 = 22 + 1 and Germany, for updating me about the Gauss statue in
1
5 = 22 + 1);3 but a regular 14-sided gon is not, be- Brunswick.
cause 14 = 2 × 7, and 7 is not a Fermat prime. Neither 3. Proposition 16 of Book IV of the Elements gives the
is a 50-sided gon, because 50 = 2 × 5 × 5, and the dou- construction of a regular 15-sided polygon. For a nice ani-
ble appearance of 5 in the factorization disqualifies mation of the construction, see http://en.wikipedia.org/
the 50-gon from being constructible. wiki/Pentadecagon#Regular_pentadecagon.
 24
Fifty

N umber aficionados will tell you that every num-

ber has its own personality, its special features
and its unique meaning—although people might
Fifty is also the smallest number that can be writ-
ten as the sum of two squares in two different ways:
50 = 52 + 52 = 72 + 12. In his book Arithmetica, Diophan-
differ as to what exactly that meaning is. But from tus of Alexandria (probably the third century CE)
the perspective of a number theorist, what matters showed that the product of two numbers, each of
most is a number’s prime factors. This prime factor- which is a sum of two squares, is again a sum of two
ization is the key to most, if not all, the mathemati- squares:
cal properties of a number.
Let us take a look at 50—a nice, round number, (a 2 + b 2 )(c 2 + d 2 ) = a 2c 2 + a 2d 2 + b 2c 2 + b 2d 2
the halfway point of a century, the age at which = (a 2c 2 + 2abcd + b 2d 2 )
many of us start to think about our mortality. As + (a 2d 2 − 2abcd + b 2c 2 )
decreed in the Bible (Leviticus 25:10), the year fol-
= (ac + bd )2 + (ad − bc )2 .
lowing a 49-year-long cycle, the year of the Jubilee
(from the Hebrew Yovel) was designated as a time of But the last expression can also be written as
renewal, of emancipation of slaves and restoration (ac − bd)2 + (ad + bc)2, showing that the said product
of leased lands to their former owners. To this day a can be expressed as the sum of two squares in two
fiftieth anniversary is known as a jubilee. different ways.1 Since 50 is the product of 5 and 10,
The prime factorization of 50 is 2 × 5 × 5, and as we each of which is a sum of two squares (5 = 12 + 22 and
saw in the previous chapter, the double presence of 10 = 12 + 32), we have
the Fermat prime 5 in the factorization rules out the
possibility of constructing a regular polygon of 50 50 = (1 × 1 + 2 × 3)2 + (1 × 3 − 2 × 1)2 = 72 + 12,
sides with Euclidean tools. Interestingly, a 51-gon—
practically indistinguishable from its 50-sided neigh- and, again,
bor—is constructible with Euclidean tools; that’s
50 = (1 × 1 − 2 × 3)2 + (1 × 3 + 2 × 1)2 = 52 + 52.
because 51 = 3 × 17, and both 3 and 17 are Fermat
0 2
primes (3 = 22 + 1 and 17 = 22 + 1). The next such number is 65 = 5 × 13 = 82 + 12 = 72 + 42.
7 8  

Plate 24.1. Stars and Stripes

   7 9

Plate 24.2. Celtic Motif 2

8 0   B e a u t i f u l Geometry

Coincidentally, 50 stars appear on the American without offending any other state! We display the
national flag, the Stars and Stripes, each star repre- two flags, the actual and the hypothetical, in plate
senting one state of the Union. This opens up an 24.1.
interesting possibility. Most people, looking at the Plate 24.2 shows a laced pattern of 50 dots, based
pattern of stars on the flag, will see an arrangement on an ancient Celtic motif. Note that the entire array
of nine horizontal rows alternating between six and can be crisscrossed with a single interlacing thread;
five stars per row (5 × 6 + 4 × 5 = 50). But if you look at compare this with the similar pattern of 11 dots (see
the star arrangement diagonally, an entirely new chapter 17), where two separate threads were neces-
pattern emerges: five rows with 1, 3, 5, 7, and 9 stars, sary to cover the entire array. As we said before,
followed by the same pattern in reverse. In chapter 4 every number has its own personality.
we saw that the sum of the first n consecutive odd
integers is n2. In our case, 1 + 3 + 5 + 7 + 9 = 52 = 25,
making the total number of stars 50.
N ot e :
But 50 is also equal to 72 + 12, which can be ar-
ranged in a square of 7 × 7 = 49 stars and a single ad- 1. Except when a = b or c = d, in which case the product
ditional star anywhere outside. This single star could is the sum of two squares in just one way; for example,
10 = 2 × 5 = (12 + 12)(12 + 22) = (1 × 1 ± 1 × 2)2 + (1 × 2 1 × 1)2 =
±
then stand for any one state of the Union, allowing
each state to claim that it has a privileged status 32 + 12.
 25
Doubling the Cube

A ccording to legend, at one time the Greek town

of Delos was afflicted by a devastating plague
that nearly decimated its population. In despera-
The Delian problem, as it has been known since,
was not the only problem that couldn’t be solved
with Euclidean tools. Two other constructions
tion, the city elders consulted the oracles, who deter- caused mathematicians even greater headache: tri-
mined that the god Apollo was unhappy with the secting an arbitrary angle, and squaring the circle—
small size of the pedestal on which his statue was that is, constructing a square equal in area to that of
standing. To appease him, they recommended to a given circle. All three problems, collectively known
double the volume of the cubical pedestal. The task as the classical problems of antiquity, were settled only
was given to the town’s mathematicians, who soon in the nineteenth century—the first two by Pierre
realized that doubling the side of the cube would not Laurent Wantzel in 1837, and the last—squaring the
do it—it would increase the volume eightfold and circle—by Ferdinand Lindemann in 1882 (see chap-
would make the pedestal unreasonably large. What ter 26). Using algebraic methods that were not avail-
to do? able to the Greeks, these mathematicians set the
It took mathematicians some two thousand years conditions under which a geometric construction,
to realize that the problem could not be solved with suitably translated into a set of equations, could be
Euclidean tools, no matter how much one tried. achieved with Euclidean tools. Among the construc-
That is, given a line segment of length a, no con- tions that cannot be done is that of a line segment
struction, using only straightedge and compass, whose length is equal to the cube root of another,
could produce a line segment of length x such that given line segment (except, of course, when the lat-
x3 = 2a3. To be sure, a variety of other tools have been ter is a perfect cube); and since the Delian problem
devised to accomplish the task, but they were re- leads to the equation x = 3 2 a , the construction can-
garded as “mechanical” and thus not befitting a not be achieved with Euclidean tools.
“true” solution in accordance with Plato’s decree. The impossibility of solving the three classical
(Of course, the straightedge and compass are me- problems was not the first time that a mathematical
chanical tools too, but for tradition’s sake they are task has been proved impossible to achieve under
the only tools permitted in the construction.) given restrictive conditions. We recall the intellec-
82

1. 2 5 9 9 2 1 0 4 9 8 9

4
8 7
. .

3 1
.

6 4
.
5

7
7 0 2

6 7
5

2 1
0

8 3 5 2 7 8 2 7 2 0 0 6

Plate 25. The Oracles of Delos

 2 5 . D o u b l i ng th e C u b e   8 3

tual crisis that struck the Pythagoreans when they Plate 25, The Oracles of Delos, shows a fanciful “du-
proved that the square root of 2 could not be ex- plication” of two unfolded dice, which, when added
pressed as a ratio of two integers. it showed that together, produce a third dice of twice the volume of
even mathematics, for all its reputation as the disci- the original dice. Of course, the plus and equal signs
pline of absolute, infallible truth, has its inherent should not be taken literally, reminding us again
limitations. But this hasn’t stopped countless ama- that the artist is not bound by the same constraints
teurs and cranks from submitting their “solutions” that limit a mathematician. And yet not everything
to professional journals, with visions of lasting is fanciful here: the side of the large cube is about
fame and perhaps even a monetary award. Appar- 1.259 . . . as large as that of the small cubes. This
ently some people will never take no for an an- number is the decimal value of 3 2 , the very stum-
swer—even if that no derives its authority from bling block that stood in the way of solving the De-
mathematics! lian problem.
 26
Squaring the Circle

A t first glance, the circle may seem to be the sim-

plest of all geometric shapes and the easiest to
draw: take a string, hold down one end on a sheet of
the circle’s diameter. Taking the diameter to be 1
and equating the circle’s area to that of the square,
we get π(1⁄2)2 = (8⁄9)2, from which we derive a value
paper, tie a pencil to the other end, and swing it of π equal to 256⁄81 ≈ 3.16049—within 0.6 percent of
around—a simplified version of the compass. But the true value. However, as remarkable as this
first impressions can be misleading: the circle has achievement is, it was based on “eyeballing,” not
proved to be one of the most intriguing shapes in all on an exact geometric construction.
of geometry, if not the most intriguing of them all. In the Bible (I Kings 7:23) we find the following
How do you find the area of a circle, when its ra- verse: “And he made a molten sea, ten cubits from
dius is given? You instantly think of the formula one brim to the other; it was round all about . . . and
A = π r 2. But what exactly is that mysterious symbol a line of thirty cubits did compass it round about.”
π? We learn in school that it is approximately 3.14, In this case, “he” refers to King Solomon, and the
but its exact value calls for an endless string of digits “molten sea” was a pond that adorned the entrance
that never repeat in the same order. So it is impos- to the Holy Temple in Jerusalem. Taken literally,
sible to find the exact area of a circle numerically. this would imply that π = 3, and the quadrature of
But perhaps we can do the next best thing—con- the circle would have become a simple task! A great
struct, using only straightedge and compass, a deal of commentary has been written on this one
square equal in area to that of a circle? verse (it also appears, with a slight change, in II
This problem became known as squaring the cir- Chronicles 4:2), but that would take us outside the
cle—or simply the quadrature problem—and its so- realm of geometry. Plate 26.1, π = 3, quotes this fa-
lution eluded mathematicians for well over two mous verse in its original Hebrew; to read it, start at
thousand years. The ancient Egyptians came pretty the red dot and proceed counterclockwise all the
close: In the Rhind Papyrus, a collection of 84 way around.
mathematical problems dating back to around Numerous attempts have been made over the cen-
1800 BCE, there is a statement that the area of a turies to solve the quadrature problem. Many ca-
circle is equal to the area of a square of side 8⁄9 of reers were spent on this task—all in vain. The defini-
 85

π = 3

Plate 26.1. π = 3
86

Plate 26.2. Metamorphosis of a Circle

 2 6 . Sq u a r i ng th e C i r c l e   8 7

tive solution—a negative one—came only in 1882, a circular disk (in red) centered on it. As the squares
when Carl Louis Ferdinand von Lindemann (1852– decrease in size, the circles expand, yet the sum of
1939) proved that the task cannot be done—it is im- their areas remains constant. In the central frame, the
possible to square a circle with Euclidean tools. Ac- square and circle have the same area, thus offering a
tually, Lindemann proved something different: that computer-generated “solution” to the quadrature
the number π, the constant at the heart of the quadra- problem. In the panel on the lower right, the squares
ture problem, is transcendental. A transcendental and circles reverse their roles, but the sum of their
number is a number that is not the solution of a poly- areas is still constant. The entire sequence is thus a
nomial equation with integer coefficients. A num- metamorphosis from square to circle and back.
ber that is not transcendental is called algebraic. All Of course, Euclid would not have approved of
rational numbers are algebraic; for example, 3⁄5 is the such a solution to the quadrature problem, because
solution of the equation 5x − 3 = 0. So are all square it does not employ the Euclidean tools—a straight-
roots, cubic roots, and so on; for example, 2 is the edge and compass. It does, instead, employ a tool of
positive solution of x2 − 2 = 0, and 3 2 − 5 is one so- far greater power—the computer. But this power
lution of x6 − 4x3 − 1 = 0. The name transcendental has comes at a price: the circles, being generated pixel
nothing mysterious about it; it simply implies that by pixel like a pointillist painting, are in reality not
such numbers transcend the realm of algebraic true circles, only simulations of circles.1 As the old
(polynomial) equations. saying goes, “there’s no free lunch”—not even in
Now it had already been known that if π turned geometry.
out to be transcendental, this would at once estab-
lish that the quadrature problem cannot be solved.
N ot e :
Lindemann’s proof of the transcendence of π there-
fore settled the issue once and for all. But settling 1. The very first of the 23 definitions that open Euclid’s
the issue is not the same as putting it to rest; being Elements defines a point as “that which has no part.” And
the most famous of the three classical problems, we since all objects of classical geometry—lines, circles, and
can rest assured that the “circle squarers” will pur- so on—are made of points, they rest on the subtle assump-
sue their pipe dream with unabated zeal, ensuring tion that Euclidean space is continuous. This, of course, is
that the subject will be kept alive forever. not the case with computer space, where Euclid’s dimen-
Plate 26.2, Metamorphosis of a Circle, shows four sionless point is replaced by a pixel—small, yet of finite
large panels. The panel on the upper left contains size—and space between adjacent pixels is empty, con-
nine smaller frames, each with a square (in blue) and taining no points.
 27
Archimedes Measures the Circle

B y any measure, Archimedes (ca. 285–212 BCE) is

considered the greatest scientist of antiquity, the
equal of Newton and Einstein. As with most of the
that the crown was made of base metals rather than
pure gold, Heron asked Archimedes to investigate.
Immersing the crown in Syracuse’s public bath and
ancient Greek sages, much of what we know about comparing its weight to that of the spilled water, he
him was written by later historians, who often con- concluded that the crown was indeed a forgery. Be-
fused legend with fact; thus many of the stories side himself with excitement, Archimedes leaped
about him must be taken with a grain of salt. He was out of the bath and ran down the streets of Syracuse,
born in the town of Syracuse, on the southeast coast shouting eureka—I found it!
of the island of Sicily, where he spent all his life. In In a rare case of exact dating of an ancient event,
popular accounts Archimedes is best remembered Archimedes’s year of death is known because it hap-
for his spectacular engineering feats, but he consid- pened when the Romans finally breached Syracuse’s
ered himself first and foremost a pure mathemati- walls and took the city by surprise. Their com-
cian, interested in mathematical theorems for their mander, General Marcellus, ordered his troops to
own sake, with little regard for their practical appli- capture the renowned scientist alive and treat him
cations. Yet when the Roman navy lay siege to his with dignity. A soldier came across an old man
town, King Heron called upon him to design war crouching over some geometric figures drawn in the
machines with which the city could be defended. sand. Being ordered to stand up, the man ignored
Among his devices was an enormous crane that the soldier, who then drew his sword and killed him:
could pluck a ship out of the sea, hoist it high in the it was Archimedes. The year was 212 BCE.
air, and then let it fall to its doom; he also devised Archimedes wrote on a wide range of subjects,
huge concave mirrors that aimed the sun’s rays at but only a dozen or so of his works survived. Among
the Roman fleet and set its ships ablaze. Whether them is a small tract with the title Measurement of a
any of these machines was actually built is not Circle, in which he devised a method for approximat-
known, but their stories made his name legendary. ing the value of π to any desired accuracy. His idea
By far the most famous story about Archimedes is was to inscribe regular polygons of 6, 12, 24, 48 and
his investigation of the king’s crown. Suspicious 96 sides inside a circle, find the perimeter of each
 89

Plate 27. Homage to Archimedes

9 0   B e a u t i f u l Geometry

polygon, and divide it by the circle’s diameter (by would be another two thousand years before math-
definition, π is the circumference-to-diameter ratio). ematicians would prove this fact conclusively.
With each step, the polygons will grip the circle Plate 27, Homage to Archimedes, shows a black cir-
more tightly from within, resulting in approxima- cle and a series of inscribed and circumscribing reg-
tions of π progressively increasing in accuracy. These ular polygons (in blue and red, respectively) of 3, 6,
approximations, however, are all undervalues of the 12, 24, and 48 sides. We see how the circle is squeezed
exact value of π. Archimedes therefore repeated the between each pair of polygons, the fit getting tighter
process with circumscribing polygons, gripping the as the number of sides increases. The final (central)
circle from the outside and giving a series of overval- circle is practically indistinguishable from the 48-
ues of increasing accuracy. From the 96-sided in- sided polygons that hold it tight. For practical rea-
scribed and circumscribing polygons, he concluded sons Archimedes started with a hexagon rather than
that the actual value of π lies between 310⁄71 and 310⁄70 a triangle, because its perimeter is easy to find; and
(in decimal notation, between 3.14085 and 3.14286). he doubled the number of sides so that he could use
This last value is equal to 22⁄7, an approximation a formula he himself had devised for computing the
which in precomputer times was often used as a perimeter of a regular 2n-gon from that of a regular
rough estimate. n-gon.1
Besides devising the first workable algorithm for
approximating π, Archimedes’s method also gave us
N ot e :
a glimpse into the theoretical nature of this famous
number. It hinted to the fact that the exact value of 1. For a full account of Archimedes’s method, see The
π can never be found, because it involves a process Works of Archimedes, edited by T. L. Heath (New York:
that must be repeated infinitely many times. It Dover. 1953), pp. 91–98.
 28
The Digit Hunters

In the second century BCE, during Archimedes’s

lifetime, the Hindu-Arabic numeration system was
still more than a thousand years in the future. So
portedly had the number inscribed on his tomb-
stone in Leiden, but all traces of it have been lost.
Here it is:
Archimedes had to do all his calculations in a strange
3.14159265358979323846264338327950288.
hybrid of the Babylonian sexagesimal (base 60) sys-
tem and the Greek system, in which each letter of Apparently nothing would stop the digit hunters
the alphabet had a numerical value (alpha = 1, from attempting to find π to ever more decimal
beta = 2, and so on). Today, of course, we associate places. In 2009 a Japanese team set the record at 2.6
the value of π with its decimal expansion—a nonre- trillion digits, but it is only a question of time before
peating, seemingly random string of digits that goes this record, too, will be broken.1 Though of no ap-
on forever. Terminate this expansion after any num- parent practical use, such enormous strings of digits
ber of digits, and you’ll get only an approximation may shed light on some as yet unanswered questions
of π. about π, among them whether it is a normal number
So, where should we stop? For many daily prac- (a number whose digits follow a uniform distribu-
tical tasks, the simple fraction 22 ⁄ 7 ≈ 3.1428571 will tion; that is, all blocks of digits of given length ap-
suffice, differing from π by just 0.04 percent. The pear with equal probability, regardless of the base in
Chinese mathematician Zu Chongzhi (Tsu Ch’ung- which the number is expressed).
Chih, 429–501) around 480 CE discovered the Not far behind the digit hunters are the digit re-
nice approximation 355 ⁄ 113 ≈ 3.1415929, accurate to citers. In 2006 the Japanese Akira Haraguchi, then
six places—that is, to the nearest millionth. The 60 years old, set a record by reciting the first 100,000
Dutch-German mathematician Ludolph van Ceu- digits of π from memory, taking him more than 16
len (1540–1610) computed π to 20 decimal places, hours to read them out. Some feat!
using Archimedes’s method with polygons of 60 × How do the digit hunters achieve their task?
229 sides and spending much of his professional life There are numerous infinite series and products that
on the task (he later improved his calculations to allow us to approximate π to any desired accuracy,
35 places). After Van Ceulen’s death, his widow re- some doing the job better than others. The first to
92

Plate 28. Almost π

 2 8 . T h e D i g i t H u nt e r s   9 3

devise such a formula was the Frenchman François We listed these formulas here mainly because of
Viète (1540–1603), who in 1596 discovered the infi- their historical significance, although their slow rate
nite product of convergence makes them of little use in practice
(with the last series, it takes 600 terms to get π to just
2 2 2+ 2 2+ 2+ 2
= ⋅ ⋅ ⋅. two decimal places, 3.14). And yet they are remark-
π 2 2 2 able because they tie the number π, the circumfer-
In 1671, the Scotsman James Gregory (1638– ence-to-diameter ratio of a circle, to the integers;
1675) discovered the infinite series they show once again the universality of mathemat-
ics—its ability to link together different concepts
π 1 1 1
= 1 − + − + −. that at first sight seem totally unrelated.
4 3 5 7
Plate 28, Almost π, gives the first 12,827 digits of
Not to be outdone, Leonhard Euler (1707–1783) this famous number, enough to give any potential
solved one of the great mysteries of his time: to find digit reciter a full load of numbers to memorize.
the sum of the series 1 + 1⁄22 + 1⁄32 + 1⁄42 + ⋅⋅⋅. In 1734 he
2
announced that the series converges to π ⁄ 6:
N ot e :
π 2
1 1 1 1. Source: BBC News, on the Web at http://news.bbc
= 1+ 2 + 2 + 2 + .
6 2 3 4 .co.uk/2/hi/technology/8442255.stm.
 29
Conics

Imagine slicing a cone—for illustration’s sake think

of it as an ice cream cone—with a swift stroke of a
knife. If you slice it in a plane parallel to the cone’s
cut is along the cone’s axis—comprise the five conic
sections (figure 29.1).
The Greek mathematician Apollonius of Perga
base, you get a circular cross section. Tilt the angle (ca. 262–190 BCE) wrote an extensive treatise on
slightly, and you get an ellipse. Tilt the angle even the conic sections. He gave them the names ellipse,
more, and the ellipse becomes narrower, until it no parabola, and hyperbola, according to whether the
longer closes on itself: it becomes a parabola. This cut is at an angle smaller than, equal to, or greater
happens when the cut is parallel to the side of the than the angle between the cone’s base and its side.
cone. Increase the angle yet again, and you get two The conics are endowed with numerous proper-
disconnected curves, the two branches of a hyper- ties, some shared by the entire family, while others
bola (provided you regard the cone as a double cone are unique to each member. Let us begin with the
joined at its apex). Taken together, these curves—to parabola. On a sheet of paper draw a line d and
which we may add a pair of straight lines when the choose a point F not on d. The parabola is the set

Figure 29.1
 95

Plate 29.1. Reflecting Parabola

9 6   B e a u t i f u l Geometry

d Figure 29.2

(“locus”) of all points equidistant from d and F (fig- screen. It also works in reverse: if a source of light or
ure 29.2). d is called the directrix, and F, the focus of a radio transmitter is placed at the focus, the emitted
the parabola. Like all conics, the parabola is a sym- beam will be reflected in a direction parallel to the
metric curve: its axis of symmetry passes through F parabola’s axis, a feature brought to good use in the
and runs at a right angle to d. As Archimedes had reflective surface of a car’s headlights.
discovered, rays of sunlight arriving at the parabola
in a direction parallel to its axis are reflected toward •
a single point, the focus, as shown in plate 29.1 (in-
deed, “focus” in Latin means fireplace). This prop- The ellipse is the locus of points, the sum of whose
erty finds its modern use in the ubiquitous satellite distances from two fixed points, the ellipse’s foci, is
antenna dish—a concave surface with a parabolic constant. A ray of light emanating from one focus is
cross section that, when aimed at a satellite in geo- reflected by the ellipse toward the other focus. A
stationary orbit, collects its signals at the focal point, striking example of this is the famous Whispering
where they are amplified and fed into your TV Gallery in the United States Capitol in Washington,
 97

Plate 29.2. Ellipses and Hyperbolas

9 8   B e a u t i f u l Geometry

DC, where visitors gathered at one focal point of the When you throw two stones into a pond, each will
elliptically shaped hall can hear their guide, stand- create a disturbance that propagates outward from
ing some distance away at the other focus, whisper the point of impact in concentric circles. The two
some words seemingly to himself or herself. It never systems of circular waves eventually cross each other
fails to impress the audience. and form a pattern of ripples, alternating between

• crests and troughs. Because this interference pattern

depends on the phase difference between the two
oncoming waves, the ripples invariably form a sys-
Like the ellipse, the hyperbola has two foci, but this tem of confocal ellipses and hyperbolas, all sharing
time the difference of the distances from any point on the same two foci. In this system, no two ellipses
the hyperbola to the foci is constant. The hyperbola ever cross one another, nor do two hyperbolas, but
consists of two disconnected branches, separated by every ellipse crosses every hyperbola at right angles.
two lines of demarcation that cross each other mid- The two families form an orthogonal system of curves,
way between the foci. These lines are the hyperbo- as we see in plate 29.2.
la’s asymptotes: they approach the hyperbola ever so But the conic sections also play a role on a much
close but never touch it—which is precisely what grander scale. The German astronomer Johannes
the word asymptote means: “don’t touch me.” The Kepler, whom we met already twice before, discov-
asymptotes are like signposts pointing the way to ered that all planets move around the Sun in ellipti-
infinity. cal orbits. This discovery finally put to rest the old
A ray of light originating at one focus is reflected Greek belief that the planets move around the Earth
by the hyperbola in a direction away from the other in perfect circles or combinations of circles. A cen-
focus. So the word focus in this case is really a misno- tury later Isaac Newton would show that every celes-
mer: rather than concentrate at the focus, beams of tial body—whether a planet, a comet, or a moon—
light are scattered in every direction as they bounce moves around its parent body in an ellipse, a
off the hyperbola; “antifocus” would perhaps be a parabola, or a hyperbola. The conic sections thus
better name. became the cosmic sections.
 30
 
=
 

W e encountered the geometric progression in

chapter 16 in connection with the runner’s
paradox. Many interesting results can be obtained
12
In the first stage there is one shaded square of side
⁄ , so its area is (1⁄2)2 = 1⁄4. In the second stage there
are three shaded squares, each of side 1⁄4 and area
using a geometric progression, some quite unex- (1⁄4)2 = 1⁄16. In the third stage there are nine shaded
pected. Consider a square of unit side, divide it into squares, each of area (1⁄8)2 = 1⁄64. Continuing in this
four equal smaller squares, and shade the upper- way, the total shaded area will be
right square, as in figure 30.1. Now divide each of
1 
2 3
1 3 9 27 3 3  3
the remaining, unshaded squares into four equal + + + +  = 1 + +   +   +  .
4 16 64 256 4 4  4   4  
parts, and shade the upper-right quarter of each
(figure 30.2). Repeating the process again and The expression inside the brackets is a geometric
again, will the shaded area approach a limit? If so, series with an initial term 1 and common ratio 3⁄4.
what is it? Since this common ratio is less than 1, the series con-

Figure 30.1 Figure 30.2

100

3=4
Plate 30. 3 4
 3 0 . 3 / 3 = 4 / 4   1 0 1

verges to the limit 1/(1 − 3⁄4) = 1/(1⁄4) = 4 (see the ap- rage in the early and middle seventeenth century—
pendix, page 177). Therefore, the total shaded area just before the invention of the calculus—when the
will be 1⁄4 × 4 = 1—the area of the original square. In limit concept was not yet fully understood. All kinds
other words, after sufficiently many steps, the shaded of “strange” results followed from them, which left
part will nearly cover the entire original square, even their discoverers with a sense of fascination. In 1668
though at each step we shaded only one-fourth of the Danish-born mathematician Nicolaus Mercator
each square. Plate 30, 3/3 =4/4, carries this process (ca. 1620–1687) —one of many “minor mathemati-
to its sixth stage; the shaded and unshaded squares cians” who paved the way to the invention of the
are shown here in yellow and blue, respectively. calculus—showed that the “infinite polynomial”
2 3 4
The infinite geometric series was already known (today we call it a power series) x − x ⁄ 2 + x ⁄ 3 − x ⁄4 + ⋅⋅⋅
to the Greeks; in fact, it is at the heart of Zeno’s par- converges to ln(1 + x) whenever −1 < x ≤ 1. Put x = 1 in
adoxes (see again chapter 16). The Greeks knew that the two expressions, and you get
under certain conditions—specifically, when the 1 1 1
common ratio is less than 1 in absolute value—the ln 2 = 1 − + − + − ,
2 3 4
series seems to get closer and closer to a specific
number. Today we call this number the limit of the a series as remarkable as the Gregory series (page
series. The Greeks, however, could not perceive that 93). It is indeed puzzling why such a simple series—
the sum of the entire series is actually equal to the the sum of the reciprocals of the integers, taken with
limit. A Greek mathematician would say only that alternating signs—should have anything to do with
with each additional shaded square, the combined ln 2 and by implication with e, a number intimately
area will get closer and closer to 1. tied with the calculus. Unfortunately, there is no ob-
The geometric series is only the simplest of nu- vious way to illustrate such series geometrically; the
merous series that have fascinated mathematicians geometric series is one of the few exceptions, so the
over the ages. Infinite series and products were the blue and yellow squares of our plate will have to do.
 31
The Harmonic Series

In the last chapter we saw that the series 1 − 1⁄2 +

1⁄3 − 1⁄4 + ⋅⋅⋅ converges to ln 2.

It is tempting to ask what will happen if we take

first million terms is 14.357; the first billion terms,
about 21; the first trillion terms, about 28. But to
make the sum exceed, say, 100, we would have to
the terms of this series in absolute value, that is, all add up a staggering 1043 terms (that’s 1 followed by
positive. We then get the harmonic series, the sum of 43 zeros). To get an idea of just how large this num-
the reciprocals of the positive integers: ber is, suppose we were to write down the series,
term by term, on a long paper ribbon until its sum
1 1 1 1 surpasses 100, allocating 1 cm for each term (this is
1+ + + + +
2 3 4 5 actually an underestimate, since the terms will re-
quire more and more digits as we go along). The rib-
The name “harmonic” comes from the fact that a bon will then be 1043 cm long, which is about 1025
vibrating string produces not only one note but infi- light-years. But the size of the observable universe is
nitely many higher notes, whose frequencies are 1, 2, at present estimated at only 1011 light-years, so our
3, 4, 5, . . . times the fundamental, or lowest, fre- ribbon would soon be running out of space to do
quency. It is one more example of the influence that the job! Yet if we could sum up the entire series—all
music, and musical terminology, has had on its infinitely many terms—the sum would grow to
mathematics. infinity.
Since the Middle Ages it was known that the har- The divergence of the harmonic series was first
monic series diverges—its sum grows without bound proved by Nicole Oresme (ca. 1320–1382), a French
as we add more and more terms, despite the fact that theologian, economist, and mathematician. His
the terms themselves get smaller and smaller. But proof is based on comparing the terms of the series
you would never guess this from watching the sum with a second series in which the third and fourth
grow, because the rate of divergence is agonizingly terms are 1⁄4 + 1⁄4 instead of 1⁄3 + 1⁄4, the fifth, sixth, sev-
slow. Some numbers will make this clear: the sum of enth, and eighth terms are 1⁄8 + 1⁄8 + 1⁄8 + 1⁄8 instead of
the first thousand terms of the series is 7.485, 1⁄5 + 1⁄6 + 1⁄7 + 1⁄8, and so on. Since 1⁄3 + 1⁄4 > 1⁄4 + 1⁄4, 1⁄5 +

rounded to the nearest thousandth; the sum of the 1⁄6 + 1⁄7 + 1⁄8 > 1⁄8 + 1⁄8 + 1⁄8 + 1⁄8, etc., we have
 103

Plate 31. A Dangerous Overhang

1 0 4   B e a u t i f u l Geometry

1
2
1
4
1
6
1
8
1
10 Figure 31.1

1 1 1 1 1 1 1 physics. Imagine stacking n identical domino tiles

1+ + + + + + + +
2 3 4 5 6 7 8 one on top of the other, but with each tile offset with
respect to the one below it according to sequence 1⁄2,
1 1 1 1 1 1 1
> 1+ + + + + + + + 1⁄4, 1⁄6, 1⁄8, . . ., 1⁄2n (taking the length of each tile to be
2 
4 4 
8 
8 
8 
8
1; see figure 31.1). The stack of dominoes will gradu-
1 1 1 ally curve and create an increasing overhang, and we
=1+ + + +
2 2 2 would expect that its center of gravity will eventu-
ally extend beyond the bottom tile and cause the
Each group of terms after the first in this second stack to collapse. Surprisingly, this does not hap-
series is equal to 1⁄2, so the series grows to infinity. pen: against all odds, the stack will survive intact,
And since the original series is greater than 1 + 1⁄2 + although just barely so (see plate 31). The overhang
1⁄2 + 1⁄2 + 1⁄2 + ⋅⋅⋅, it too grows to infinity—it diverges. from stacking n dominos in this manner turns out to
The harmonic series gives rise to many surprises. be 1⁄2(1 + 1⁄2 + 1⁄3 + ⋅⋅⋅ + 1⁄n). And since this finite har-
For example, if we remove from it all the terms with monic series diverges as n → ∞, we could add on
composite denominators—leaving only prime-num- more and more tiles (starting at the bottom) and
ber denominators—the series will still diverge! This make the overhang as large as we please while the
is rather remarkable, because the primes thin out as stack will maintain its equilibrium—though I
we go to higher numbers—they become ever more wouldn’t recommend standing under it on a rainy
rare. And yet the sum of their reciprocals still di- day!1
verges. On the other hand, the sum of the recipro-
cals of all twin primes (see page 45) is known to con-
N ot e :
verge—although it is still an open question how
many twin primes are there. So, until the issue is 1. See John Bryant and Chris Sangwin, How Round
settled, we cannot say with unfailing certainty that is Your Circle: Where Engineering and Mathematics Meet
this is an infinite series. (Princeton, NJ: Princeton University Press, 2008), pp.
But perhaps the most amazing aspect of the har- 255–59.
monic series comes not from mathematics but from
 32
Ceva’s Theorem

G iovanni Ceva (1647–1734) was born in Milan

and got his schooling in a Jesuit institute there.
After completing his education in Pisa he was ap-
C

pointed professor of mathematics at the university of

Mantua, where he stayed for the rest of his life. His
work was in geometry and in hydraulics—not an un-
A‘
usual combination in seventeenth-century Europe,
B‘
before academic specialization became the rule: G
Leonardo da Vinci, Galileo, and many other scien-
tists were working on engineering problems as much B
as on art, mathematics, and physics.
The prevailing view at the time was that after Eu- C‘
clid, geometry became largely a closed subject, with
nothing more of value to be discovered. Ceva dis-
proved this view: in 1678 he discovered a new theo-
rem that is remarkable for its simplicity.
Let ABC be any triangle, and let A', B', and C' be
any points on the sides opposite of A, B, and C, re- A Figure 32.1
spectively (figure 32.1). Ceva’s theorem says that if
the lines AA', BB', and CC' pass through one point,
then
verse of Ceva’s theorem is also true: if the triple
BA' CB' AC'
⋅ ⋅ = 1. product equals 1, the three lines are concurrent—
A'C B'A C'B
they pass through one point.
Note that this expression is a product of three ra- In proving his theorem Ceva relied on mechanical
tios; we will simply call it the triple product. The con- principles, an approach that would not sit well with
106

Plate 32. Isocevas

 3 2 . C e va ’ s T h e o r e m   1 0 7

Euclid. Nevertheless, one cannot fail to admire the to great advantage when he invented his version of
simplicity of the idea behind it—to think of A, B, the differential calculus.
and C as if they were physical objects with appro- If we rewrite the triple product as
priate weights and then find their center of gravity
G. Surprisingly, the actual values of these weights BA' ⋅ CB' ⋅ AC' = A'C ⋅ B'A ⋅ C'B ,
disappear from the final equation, leaving us with a we find that interchanging the primed letters with
purely geometric relation between the line seg- the unprimed ones does not affect the value of the
ments in question. You will find Ceva’s proof in the product on each side—it remains invariant. This is
appendix. shown in our illustration (plate 32): if you multiply
But there is more to Ceva’s theorem: if you look together the lengths of the red-colored line segments
again at the triple product, you may be tempted to and then do the same with the green-colored line
regard the letters as variables that can be multiplied segments, you end up with the same product.
and divided as if they were actual numbers. Then the The central triangle in the illustration shows sev-
six letters mutually cancel out, leaving the product eral contour lines, each representing one value of
to be equal to 1! To be sure, this is no more than a the product and thus one position of G. In honor of
formal manipulation of symbols, but such formal- Giovanni Ceva we will call them isocevas, in analogy
ism often plays a role in mathematics and may even with isobars and isotherms—lines connecting points
hint at some deeper connections. Gottfried Wilhelm of equal pressure or equal temperature on a weather
von Leibniz (1646–1716) used a similar symbolism map.
 33
e

T o the trio of special numbers we have met so far—

2, φ, and π—we now add a fourth number, e.
This number, the base of natural logarithms, is of a
Financial institutions use various methods of
compounding money: annually, semiannually, quar-
terly, monthly, weekly, or even daily, in each case
more modern vintage than its ancient companions, using an adjusted nominal interest rate: 0.05,
tracing its origin to the seventeenth century. And 0.05/2 = 0.025, 0.05/4 = 0.0125, and so on. In each
unlike the others, it has its roots not in geometry, case the effective interest rate becomes smaller, but
but in the world of business. the money is compounded that much more often.
Financial matters have been of concern to people Clearly we have here two opposing effects—a smaller
since the dawn of recorded history. A Babylonian interest rate but a higher number of compounding
clay tablet, dating back to about 1700 BCE and now per year. Which will be the winner? Surprisingly, the
in the Louvre, asks how long it would take for a sum two effects almost cancel out, with a very slight gain
of money, invested at 20 percent interest rate com- in favor of more frequent compounding, as shown
pounded annually, to double. This leads to the ex- in the following table:
ponential equation 1.2x = 2, whose solution the tab-
$100 invested for 1 year at 5% interest rate
let gives as about 3.787 years.1
compounded n times a year
The seventeenth century saw a renewed curiosity
Conversion period n 0.05/n Sum
in the law of compound interest. Suppose $100 is
invested at an interest rate of 5 percent compounded Annually 1 0.05 $105
once a year. At the end of one year, this sum will Semiannually 2 0.025 $105.06
grow to 100 × 1.05 = $105. But suppose the bank Quarterly 4 0.00125 $105.09
compounds the interest twice a year, each time at half Monthly 12 0.004166 $105.12
the nominal interest rate, that is, 2.5 percent. Since Weekly 52 0.0009615 $105.12
in the same year there are now two conversion peri- Daily 365 0.0001370 $105.13
ods, the accrued amount will grow to 100 × 1.0252 =
$105.06, that is, 6 cents more than when interest is Nothing in principle prevents us from compound-
compounded annually. ing the sum even more often—every hour, every min-
 109

Plate 33. Euler’s e

1 1 0   B e a u t i f u l Geometry

ute, every second . . . every instant, resulting in a con- y

tinuous compounding of the money. Someone in
the sixteenth or early seventeenth century—we don’t
know who made the discovery or exactly when—no-
ticed that as the interest is compounded more and
more frequently, the accrued sum does not grow to
riches, as one might expect, but seems to approach
some limiting value. This limiting value comes from
the formula S = P(1 + r/n)nt for the amount of money
accrued when P dollars are invested for t years at an
annual interest rate r compounded n times a year. If 1
for convenience we replace r/n by 1/m in this for-
mula, it becomes S = P(1 + 1/m)rmt = P[(1 + 1/m)m]rt.
As n → ∞, so does m, but the expression (1 + 1/m)m x
will approach a limit, as shown in the following O
table:
Figure 33.1
m 10 100 1,000 10,000 100,000
(1 + 1/m)m 2.59374 2.70481 2.71692 2.71815 2.71827

This limit is about 2.71828 and is denoted by the let-

ter e, a notation first used by Leonhard Euler in Closely related to the exponential function are
1727.2 Like 2, φ, and π, it is an irrational number. the expressions (ex + e−x)/2 and (ex − e−x)/2, collec-
And like π, it is a transcendental number, meaning tively known as the hyperbolic functions and denoted
that it is not the solution of any polynomial equa- by cosh x and sinh x, respectively. They bear some
tion with integer coefficients (see page 87). striking similarities to the more familiar trigonomet-
It didn’t take long before e showed up in other ric functions cos x and sin x; for example, in analogy
situations with no direct connection to finance. with the trigonometric identity cos2 x + sin2 x = 1, we
Foremost among them is the exponential function have the hyperbolic identity cosh2 x − sinh2x = 1
y = ex, whose rate of change—its derivative, in the lan- (note, though, the presence of the minus sign). These
guage of calculus—at any point x is always equal to analogies, however, are merely formal; in their
ex [that is, (d/dx)ex = ex]. It is this property that makes graphs the two sets of functions are radically differ-
the exponential function so important in calculus— ent, the hyperbolic functions lacking the periodicity
so much so that e is universally used as the natural of their trigonometric counterparts.
base for all growth or decay processes, from the One of the outstanding problems that occupied
growth of a bacteria population to the disintegra- the mathematical world in the years immediately fol-
tion of a radioactive substance. Figure 33.1 shows its lowing the invention of the calculus was to find the
graph. exact curve taken up by a chain of uniform thickness
 e   1 1 1

hanging freely under the force of gravity. To the eye

a hanging chain looks very much like a parabola
(see figure 33.2), but the Dutch physicist Christiaan
Huygens (1629–1695) disproved this when he was
just 17 years old. The mystery was solved in 1691 by
three of the leading mathematicians of the time, who
worked on it independently—Huygens himeslf
(now 62), Leibniz, and Johann Bernoulii. To every-
one’s surprise, the shape turned out to be the graph
of the hyperbolic cosine, (ex + e−x)/2, which hence-
forth became known as the catenary (from the Latin
catena, a chain). In our own time it has been immor- Figure 33.2. Illustration
talized in one of the world’s most imposing archi- by Eyal Maor.
tectural monuments, the Gateway Arch in St. Louis,
Missouri, whose shape is exactly that of an inverted
catenary.
Comparing the numerical values of 2, φ, e, and
π—perhaps the four most famous numbers in math- nigs­berg (now Kaliningrad in Russia), whose seven
ematics—one cannot fail to notice how close they are bridges inspired Euler to solve a famous problem
on the number line, occupying less than three units that marked the birth of graph theory; and an as-
of its infinite length. Why this is so no one knows; it sortment of formulas associated with e.3
remains one of the enduring mysteries of science.
Plate 33, Euler’s e, gives the first 203 decimal
N ot e s :
places of this famous number—accurate enough for
most practical applications, but still short of the 1. The tablet gives this answer in the peculiar base-60
exact value, which would require an infinite string Babylonian numeration system as (in modern notation) 3;
of nonrepeating digits. In the margins there are sev- 47,13,20, which stands for 3 + 47/60 + 13/602 + 20/603, or
eral allusions to events that played a role in the his- about 3.7870.
tory of e and the person most associated with it, Le- 2. Despite often-heard claims, it is unlikely that Euler
onhard Euler: an owl (“Eule” in German); the chose the letter e because it is the initial of his own name.
Episcopal crosier on the flag of Euler’s birthplace, 3. For a more complete history of e, see Maor, e: The
the city of Basel; the latitude and longitude of Kö­ Story of a Number, of which this chapter is an excerpt.
 34
Spira Mirabilis

O f the numerous curves we encounter in art, ge-

ometry, and nature, perhaps none can match
the exquisite elegance of the logarithmic spiral,
the pole O to any point on the spiral intercepts it at
a constant angle α (figure 34.1).1 It is for this rea-
son that the curve is also known as an equiangular
shown in plate 34.1. This famous curve appears, with spiral. As a consequence, any sector with given an-
remarkable precision, in the shape of a nautilus gular width ∆Θ is similar to any other sector with
shell, in the horns of an antelope, and in the seed the same angular width, regardless of how large or
arrangements of a sunflower (see page 64, figure small it is. This is manifested beautifully in the
20.1). It is also the ornamental motif of countless nautilus shell (plate 34.2): the snail residing inside
artistic designs, from antiquity to modern times. It the shell gradually moves from one chamber to the
was a favorite curve of the Dutch artist M. C. Escher next, slightly larger chamber, yet all chambers are
(1898–1972), who used it in some of his most beauti- exactly similar to one another: a single blueprint
ful works. serves them all.
The logarithmic spiral is best described by its The logarithmic spiral has been known since an-
polar equation r = e aΘ, where r is the distance from cient times, but it was the Swiss mathematician
the spiral’s center O (the “pole”) to any point P on Jakob Bernoulli who discovered most of its proper-
the curve, Θ is the angle between line OP and the x- ties. Bernoulli (1654–1705) was the senior member
axis, a is a constant that determines the spiral’s rate of an eminent dynasty of mathematicians, all hailing
of growth, and e is the base of natural logarithms. It from the town of Basel. He was so enamored with
follows from this equation that if Θ is the sum of two the logarithmic spiral that he dubbed it spira mirabi-
angles Θ1 and Θ2, the radius will be the product of the lis and ordered it to be engraved on his tombstone
corresponding radii: r = e a (θ1 +θ2 ) = e aθ1 ⋅ e aθ2 = r1 ⋅ r2 . Put after his death. His wish was fulfilled—but not quite
differently, if we increase Θ arithmetically (that is, in as he had intended: for some reason, the mason en-
equal amounts), r will increase geometrically (in a graved a linear spiral instead of a logarithmic one (in
constant ratio). a linear spiral the distance from the center increases
The many properties of the spiral all derive from arithmetically—that is, in equal amounts—as in the
this single feature. For example, a straight line from grooves of a vinyl record). The “wrong” spiral on
 113

Plate 34.1. Spira Mirabilis

1 1 4   B e a u t i f u l Geometry

α
α
α

x
O

Figure 34.1

Bernoulli’s headstone can still be seen at the clois-

N ot e :
ters of the Basel Münster, perched high on a steep
hill overlooking the Rhine River (plate 34.3). 1. This angle is determined by the constant a; in fact,
But if a wrong spiral was engraved on Bernoulli’s α = cot−1a. In the special case when a = 0, we have α = 90°
tombstone, at least the inscription around it holds and the spiral becomes the unit circle r = e 0 = 1. For nega-
true: Eadem mutata resurgo—“though changed, I tive values of a, the spiral changes its orientation from
shall arise the same.” The verse summarizes the counterclockwise to clockwise as Θ increases.
many features of this unique curve: stretch it, rotate For more on the logarithmic spiral, see Maor, e: The
it, or invert it, it always stays the same. Story of a Number, chapter 11.
 3 4 . Sp i r a M i r a b i l i s   1 1 5

Plate 34.2. Nautilus Shell Plate 34.3. Jakob Bernoulli’s Tombstone

 35
The Cycloid

R ivaling the logarithmic spiral in elegance is the

cycloid—the curve traced by a point on the rim of
a circle that rolls along a straight line without slip-
sound, the performance of Huygens’s clock fell
short of his expectations.
Shortly thereafter the cycloid made history again.
ping (figure 35.1). The cycloid is characterized by its In 1696 Johann Bernoulli (1667–1748), the younger
arcs and cusps, each cusp marking the instant when brother of Jakob (of logarithmic spiral fame), posed
the point on the wheel’s rim reaches its lowest posi- this problem: to find the curve along which a parti-
tion and stays momentarily ar rest. cle, again subject only to the force of gravity, will
The cycloid has a rich history. In 1673 Christiaan slide down in the least amount of time. You might
Huygens, whom we’ve just met in connection with think this should be the straight line connecting the
the catenary, solved one of the outstanding prob- initial and final positions of the particle, but this is
lems that had intrigued seventeenth-century scien- not so: depending on the path’s curvature, the par-
tists: to find the curve down which a particle, mov- ticle may accelerate faster at one point and slower at
ing only under the force of gravity, will take the another, showing that the path of shortest distance
same amount of time to reach a given final point, between two points is not necessarily the path of
regardless of the initial position of the particle. This shortest time.
problem is known as the tautochrone (from the Greek Known as the brachistochrone (“shortest time”), this
words meaning “the same time”). To his surprise, problem was attempted by some of the greatest minds
Huygens found that the curve is an arc of an in- of the seventeenth century. Among them was Galileo,
verted cycloid. He tried to capitalize on his discov- who incorrectly thought the required path is an arc of
ery by constructing a clock whose pendulum was a circle. In the end, five correct solutions were sub-
constrained to swing between two adjacent arcs of a mitted in response to Johann Bernoulli’s challenge—
cycloid, so that the period of oscillations would be by Newton, Leibniz, L’Hospital (famous for a rule in
independent of the amplitude (in an ordinary pen- calculus named after him), and the two Bernoulli
dulum this condition holds only approximately). brothers, who worked on the problem independently
Unfortunately, although the theory behind it was and used different methods. To their surprise, the
 117

Plate 35. Reflections on a Rolling Wheel

1 1 8   B e a u t i f u l Geometry

x
O

Figure 35.1

curve turned out to be an inverted cycloid—the same the early seventeenth century they presented a chal-
curve that solved the tautochrone problem. But in- lenging task.
stead of rejoicing in their success, the discovery em- Plate 35, Reflections on a Rolling Wheel, shows the
broiled the two brothers in a bitter priority dispute, path of a luminous point attached to a rolling wheel
resulting in a permanent rift between them. at three different distances from the center: at top,
The cycloid had some more surprises in store. the point is outside the wheel’s rim (as on the flank
Evangelista Torricelli (1608–1647), famous for his of a railroad car wheel); at the middle, it is exactly
invention of the mercury barometer, is credited with on the rim; and at the bottom, inside of it. The top
finding the area under one arc of the cycloid: the and bottom curves are called prolate and curtate cy-
area turned out to be 3πa2, where a is the radius of cloids, respectively, while the middle curve is the or-
the generating circle. A few decades later Christo- dinary cycloid. You can see the curtate variant at
pher Wren (1632–1723), London’s venerable archi- night as the path traced by the reflector on a bicycle
tect who rebuilt the city after the Great Fire of 1666, wheel as the cyclist moves forward.1
found that the length of each arc is 8a; surprisingly,
the constant π is not involved. This was one of the
N ot e :
first successful rectifications of a curve—finding the
arc length between two points on the curve. With 1. For a full history of the cycloid, see the article “The
the invention of calculus in the decade 1666–1676, Helen of Geometry” by John Martin, The College Mathe-
problems like these could be solved routinely, but in matics Journal (September 2009, pp. 17–27).
 36
Epicycloids and Hypocycloids

W hereas the cycloid is generated by a point on

the rim of a wheel rolling along a straight line,
we may also consider a wheel rolling on the outside
the ratio 2:1 can be used to draw a straight-line seg-
ment! In the nineteenth century this provided a po-
tential solution to a problem that had vexed engi-
of a second, fixed wheel; the resulting curve is an neers for many years: how to convert the to-and-fro
epicycloid (from the Greek epi, meaning “over” or motion of the piston of a steam engine into a rota-
“above”). Or, we can let the wheel roll along the in- tional motion of the wheels. It was one of many so-
side of a fixed wheel, generating a hypocycloid lutions proposed, but in the end it turned out to be
(hypo = “under”). The epicycloid and hypocycloid impractical.
come in a great variety of shapes, depending on the When R/r = 4, the hypocycloid becomes the star-
ratio of the radii of the two wheels. Let the radii of shaped astroid (from the Greek astron, a star; see fig-
the fixed and moving wheels be R and r, respec- ure 36.4). This curve has some interesting properties
tively. If R/r is a fraction in lowest terms, say m/n, of its own. Its perimeter is 6R (just as with the cy-
the curve will have m cusps (corners), and it will be cloid, this is independent of π), and the area en-
completely traced after n full rotations around the closed by it is 3πR2/8, that is, three-eighths the area
fixed wheel. If, however, R/r is not a fraction—if it is of the fixed circle.1
irrational—the curve will never close completely, al- Imagine a line segment of fixed length with its
though it will nearly close after many rotations. Fig- endpoints resting on the x- and y-axes, like a ladder
ures 36.1 and 36.2 show the formation of a hypocy- leaning against a wall. When the ladder is allowed to
cloid with a ratio R/r = 5 and an epicycloid with assume all possible positions, it describes a region
R/r = 3, respectively. bound by one-quarter of an astroid. This shows that
For some special values of R/r the ensuing curves a curve can be formed not only by a set of points
can be something of a surprise. For example, when lying on it, but also by a set of lines tangent to it. We
R/r = 2, the hypocycloid becomes a straight-line seg- will come back to this subject in chapter 40.
ment: each point on the rim of the rolling wheel will Turning now to the epicycloid, the case where the
move back and forth along the diameter of the fixed fixed and the moving wheels have the same radius
wheel (figure 36.3). Thus, two circles with radii in (R/r = 1) is of particular interest: it results in a cardi-
1 2 0  

Figure 36.1

Figure 36.2

Figure 36.3

Figure 36.4

Figure 36.5
 121

Plate 36. Epicycloids

1 2 2   B e a u t i f u l Geometry

oid, so called because of its heart-shaped form (fig- planetary epicycles. In fact, this latter curve closely
ure 36.5). Its perimeter is 16R and its area is 6πR2.2,3 resembles the path of Venus against the backdrop of
The Greek astronomer Claudius Ptolemaeus the fixed stars, as seen from Earth. This is due to an
(commonly known as Ptolemy, ca. 85–165 CE), in 8-year cycle during which Earth, Venus, and the Sun
an attempt to explain the occasional retrograde mo- will be aligned almost perfectly five times. Surpris-
tion (east to west, instead of the usual west to east) ingly, 8 Earth years also coincide with 13 Venusian
of the planets, ascribed to them a complex path in years, locking the two planets in an 8:13 celestial
which each planet moved along a small circle whose resonance and giving Fibonacci aficionados one
center moved around Earth in a much larger circle. more reason to celebrate!
The resulting orbit, an epicycle, has the shape of a
coil wound around a circle. When even this model
failed to account for the positions of the planets at
N ot e s :
any given time, more epicycles were added on top of
the existing ones, making the system increasingly 1. We might mention in passing that the astroid has the
cumbersome. It was only when Johannes Kepler dis- unusual rectangular equation x2/3 + y2/3 = R2/3.
covered that the planets move around the Sun in el- 2. For nice simulations of how these curves are gener-
lipses that the unwieldy epicycles became unneces- ated, go to http://mathworld.wolfram.com/Hypocycloid
sary and were laid to rest. .html and http://mathworld.wolfram.com/Epicycloid.html.
Plate 36 shows a five-looped epicycloid (in blue) 3. For more on the properties of epicycloids and hypo-
and a prolate epicycloid (in red) similar to Ptolemy’s cycloids, see Maor, Trigonometric Delights, chapter 7.
 37
The Euler Line

Leonhard Euler (1707–1783) was arguably the most

prolific mathematician of all time. Born in Basel,
Switzerland, to a Calvinist minister, he studied with
yet fully published, is estimated to take up 70
volumes.
There are more formulas and theorems named
Johann Bernoulli and received his master’s degree after Euler than any other scientist in history. Among
from the University of Basel at the age of 16. When his most celebrated results are the formula V − E +
two of Bernoulli’s sons, Nicolas and Daniel (the lat- F = 2 for the number of vertices V, the number of
ter famous for a law named after him in hydrody- edges E, and the number of faces F of any simple
namics), moved to St. Petersburg, Euler followed polyhedron (a solid with planar faces and no holes),
them and spent the next 14 years there. He then and the equation eπi + 1 = 0 that relates the funda-
moved to Berlin to head the Prussian Academy of mental constants of arithmetic (0 and 1), of analysis
Sciences but returned to Russia in 1766 and never (e), of geometry (π), and of complex numbers
left again. His last years were beset by tragedies: the (i = −1). This equation has often been hailed by
death of his wife, the loss of vision in both of his eyes, popular writers as having no less than divine power,
and a fire that destroyed his home and library (fortu- but mathematicians generally refrain from such
nately, most of his manuscripts were saved). None of mystical attributes.
these setbacks, however, slowed down his inexhaust- We present here a lesser-known jewel of Euler,
ible creativity: at the age of 70 he remarried, and al- discovered by him in 1765: in any triangle, the cir-
though completely blind, he kept working to the cumcenter (the center of the circumcircle), the cen-
very end, dictating his results to his assistants. troid (the intersection of the three medians, so called
Euler’s work covered nearly every area of mathe- because it is the center of gravity of the triangle),
matical research known in his time. He could move and the orthocenter (the intersection of the three alti-
as easily from number theory to analysis as from ge- tudes), all lie on one line, the Euler line. Moreover, if
ometry to physics and astronomy. He was also a we denote the three points by O, G, and H, respec-
great popularizer of science, writing numerous tively, then GH = 2OG . This result looks so Euclid-
memoirs, letters, and books. His entire output, not ean, yet it had to wait for Euler to be discovered.
124

Plate 37. Nine Points and Ten Lines

 3 7 . T h e E u l e r L i n e   1 2 5

Said the Canadian geometer H.S.M. Coxeter: in green, red, and orange, respectively, and the
“Some of his simplest discoveries are of such a na- Euler line, in yellow. We call this a construction
ture that one can well imagine the ghost of Euclid without words, where the points and lines speak for
saying, ‘Why on earth didn’t I think of that?’ ”1 themselves.
Our illustration Nine Points and Ten Lines (plate
37) shows the point-by-point construction of Eul-
N ot e :
er’s line, beginning with the three points defining
the triangle (marked in blue). The circumcenter O, 1. Coxeter, Introduction to Geometry, p. 17. For a proof
the centroid G, and the orthocenter H are marked of Euler’s line theorem, see pp. 17–18.
 38
Inversion

T he circle can be a source of never-ending fascina- Q

tion. Consider a circle with center at O and radius
1. Two points P and Q on the same ray through O are
called inverses of each other with respect to the circle
if OP ⋅ OQ = 1, or, equivalently, OQ = 1/OP (figure
38.1). Either point can be thought of as the “image” 1 P
of the other. If P is inside the circle (that is, OP < 1),
its image Q will lie outside (OQ > 1), and vice versa.
O
Therefore, the entire interior of the circle is “mapped”
onto the exterior in a one-to-one correspondence.
And, conversely, the interior is a miniature map of
the world outside the circle. If OP = 1, then OQ = 1,
Figure 38.1
so that points on the circle are mapped onto
themselves.
But what about the center O? Where is its image
point? If OP = 0, then OQ = 1 /OP = 1 / 0 . But wait:
didn’t we learn in school that division by 0 is unde- Since inversion maps points that are close to the
fined? To go around this difficulty, we define the center onto points far away from it, we expect that
image of O under inversion to be the point at infinity. figures should be greatly distorted when subjected
But where exactly is this point, and in what direc- to inversion. Yet in spite of this, some figures do not
tion? The answer, metaphorically speaking, is: ev- change at all. Clearly, every ray through O is mapped
erywhere and nowhere! This “point” is not an ordi- on itself, and a circle with center at O and radius r
nary point in the usual sense of the word; we becomes a circle with the same center and radius
introduce it only so as not to exclude the center from 1/r. More surprising is the fact that a circle through O
our discussion. is mapped onto a straight line not through O, and vice
 127

Plate 38. Inverted Circles

1 2 8   B e a u t i f u l Geometry

versa. Note that the two “endpoints” of the line are

A
one and the same point, the point at infinity.
These relations are illustrated in plate 38. The cir-
cle of inversion, in light blue, is the same in all pan-
els, while a red circle of varying size passes through
the center of inversion, marked by a black dot. De-
pending on whether the red circle lies entirely in-
side the blue circle, touches it internally, or inter-
sects it, its image line (in green) will pass outside O
B E D
the inversion circle, be tangent to it, or cut it at the
two points where the circles intersect. The central
panel illustrates the limiting case when the red cir-
cle is so large that it becomes a straight line through
the center, in which case it coincides with its own
image.
Inversion has many applications, of which we
mention one here. In 1864 Charles-Nicolas Peaucel- C
lier (1832–1913), a French army officer, and Yom
Tov Lipman Lipkin of Lithuania, the son of a fa-
mous rabbi, independently invented a mechanical Figure 38.2
linkage that could transform rectilinear motion into
rotary motion (figure 38.2). It had previously been
believed that this could not be done mechanically,
but with the advent of the steam engine it became
paramount to find a way to convert the to-and-fro But OA and AB are given quantities, so the last ex-
motion of the piston into a rotary motion of the en- pression is constant. Call this constant k2 (it is posi-
gine’s wheels. tive because OA > AB ). We then have OB ⋅ OD = k 2 , or
The Peaucellier inverter, as it became known, was
OB OD
one of many attempts at solving this problem. It ⋅ = 1,
consists of six rigid rods OA, OC, AB, BC, CD, and DA k k
connected by linkages (points that allow free rota- showing that points B and D are inverses of each
tion) at O, A, B, C, and D, and such that OA = OC and other in a circle (not shown in the figure) of radius k
AB = BC = CD = DA . We have and center O. Consequently, when D moves along a
2 2
straight line, its image B describes an arc of a circle
OB ⋅ OD = (OE − BE ) ⋅ (OE + BE ) = OE − BE through O, seemingly offering a solution to the rec-
2 2 2 2 2 2
= (OA − AE ) − (AB − AE ) = OA − AB . tilinear-to-circular conversion problem.1
 3 8 . In v e r s i o n   1 2 9

Alas, it would never be possible to draw a com- N ot e s :

plete circle with this device, as this would require us 1. A nice demonstration of the Peaucellier inverter can be
to trace the full length of the line, from one “end- found at http://mechanical-design-handbook.blogspot.
point” (the point at infinity) to the other. So the de- com/2011/02/peaucellierlipkin-and-sarrus-straight.html.
vice was mainly a theoretical curiosity and became a 2. A proof of some of the properties of inversion is
favorite with nineteenth-century popularizers of found in the appendix. For a full discussion, see Coxeter,
science.2 Introduction to Geometry, pp. 77–91.
 39
Steiner’s Porism

T he first half of the nineteenth century saw a re-

vival of interest in classical Euclidean geometry,
in which figures are constructed with straightedge
educator Heinrich Pestalozzi, he became completely
dedicated to mathematics. Among his many beauti-
ful theorems we bring here one that became known
and compass and theorems are proved from a given as Steiner’s porism. Given two nonconcentric circles,
set of axioms. This “synthetic,” or “pure,” geometry one lying entirely inside the other, construct a series
had by and large been thrown by the wayside with of secondary circles, each touching the circle preced-
the invention of analytic geometry by Fermat and ing it in the sequence as well as the two original cir-
Descartes in the first half of the seventeenth century. cles (see Figure 39.1). Will this Steiner chain close
Analytic geometry is based on the idea that every upon itself, so that the last circle in the chain coin-
geometric problem could, at least in principle, be cides with the first? Steiner, in 1826, proved that if
translated into the language of algebra as a set of this happens for any particular choice of the initial
equations, whose solution (or solutions) could then circle of the chain, it will happen for every choice.
be translated back into geometry. This unification of In view of the seeming absence of symmetry in
algebra and geometry reached its high point with the configuration, this result is rather surprising.
the invention of the differential and integral calcu- Steiner devised a clever way of inverting the two
lus by Newton and Leibniz in the decade 1666–1676, original circles into a pair of concentric circles. As a
and it has remained one of the chief tools of mathe- result, the chain of secondary circles (now inverted)
maticians ever since. The renewed interest in syn- will occupy the space between the (inverted) given
thetic geometry came, therefore, as a fresh breath of circles evenly, like the metal balls between the inner
air to a subject that had by that time been consid- and outer rings of a ball bearing wheel; obviously
ered out of fashion. these can be moved around in a cyclic manner with-
One of the chief protagonists in this revival was out affecting the chain.
the Swiss geometer Jacob Steiner (1796–1863). But that’s not all: it turns out that the centers of
Steiner did not learn how to read and write until he the circles of the chain always lie on an ellipse
was 14, but after studying under the famous Swiss (marked in red in figure 39.1), and the points of
 131

Plate 39. Steiner’s Porism

1 3 2   B e a u t i f u l Geometry

Figure 39.1

contact of adjacent circles lie on yet another circle almost half a century before Steiner. An old Japa-
(marked in green).1 nese tradition, going back to the seventeenth cen-
Plate 39 illustrates several Steiner chains, each tury, was to write a geometric problem on a wooden
comprising five circles that touch an outer circle (al- tablet, called sangaku, and hang it in a Buddhist
ternately colored in blue and orange) and an inner temple or Shinto shrine for visitors to see. A fine ex-
black circle. The central panel shows this chain in its ample of Steiner’s—or Chokuyen’s—chain appeared
inverted, symmetric “ball-bearing” configuration. on a sangaku at the Ushijima Chomeiji temple in
As happens occasionally, a theorem that has been Tokyo. The tablet no longer exists, but an image of
known in the West for many years turned out to have it appeared in a book published about the same time
already been discovered earlier in the East. Steiner’s as Steiner’s discovery (see figure 39.2).2
porism is a case in point: a Japanese mathematician, It is somewhat of a mystery why this theorem be-
Ajima Chokuyen (1732–1798), discovered it in 1784, came known as Steiner’s porism. You will not find
 3 9 . St e i n e r ’ s P o r i s m   1 3 3

Figure 39.2. Courtesy of University of Aichi

Education Library.

the word porism in your usual college dictionary,

N ot e s :
but the online Oxford English Dictionary defines it
as follows: In Euclidean geometry: a proposition aris- 1. Steiner chains enjoy many additional properties. See
ing during the investigation of some other propositions http://en.wikipedia.org/wiki/Steiner_chain. For a proof
by immediate deduction from it. Be that as it may, the of Steiner’s porism, see Coxeter, Introduction to Geometry,
theorem again reminds us that even the good old p. 87.
Euclidean geometry can still hold some surprises 2. See Hidetoshi and Rothman, Sacred Mathematics:
within it. Japanese Temple Geometry, p. 292.
 40
Line Designs

J ulius Plücker (1801–1868) is not a household

name among present-day mathematicians, but in
the nineteenth century he carved himself a niche in
lines can be used as building blocks of geometric
figures just as much as points. Plate 40.1 shows a
parabola generated entirely from its tangent lines;
geometry where few others had ventured before. He not a single point was used in its construction (the
realized that a curve need not be regarded as a set of plate also shows the parabola’s reflecting property,
points; it can just as well be described as a set of discussed earlier in chapter 29). Plate 40.2 goes even
tangent lines. The idea was not entirely new. It had further, showing a Star of David–like design made
been known for more than a century that certain for- of 21 line parabolas.
mal statements about points and lines remain valid Plücker’s career took him through strange twists.
when the words point and line are everywhere inter- His major work in geometry was published in two
changed. For example, just as two points determine volumes in 1828 and 1831, and it was in the second
a unique line, so do two lines determine a unique volume that he gave the analytic formulation of the
point—their point of intersection.1 This principle of principle of duality. Yet his work was not favorably
duality became the centerpiece of a new kind of ge- received by the two leading geometers of the time,
ometry, projective geometry, in which dual relations Jacob Steiner and Jean Victor Poncelet, whose syn-
such as “two points determine a line” or “two lines thetic geometry was more in line with the classical
determine a point” became the main focus, rather geometry of Euclid. It didn’t help that Plücker’s aca-
than metric properties such as the length of a line demic position at the University of Bonn was not in
segment or the area of a polygon. mathematics but in physics. This in itself was not
But while the principle of duality was well known unusual (Gauss held the position of director of the
in Plücker’s time, he gave it a new formulation that astronomical observatory at Göttingen), but it was
placed the subject squarely in the realm of analytic used by Plücker’s adversaries to claim that he was
geometry: he showed that a curve can be generated not a true physicist. To prove them wrong, he aban-
from a set of lines obeying a line equation, in much doned mathematics and for the next 18 years de-
the same way as the traditional view of a curve as a voted himself to physics, making contributions in
set of points obeying a point equation. In other words, optics, magnetism, and spectroscopy. It was only
 135

Plate 40.1. Line Parabola

136

Plate 40.2. Line Design

 4 0 . L i n e D e s i gn s   1 3 7

toward the end of his life that Plücker returned N ot e s :

to his first love, geometry, where he made several
more discoveries. After his death his work was com- 1. If the lines are parallel, the point of intersection re-
pleted by the influential German mathematician cedes to infinity and is known as a “vanishing point.”
Felix Klein (1849–1925), who, like Plücker, was as 2. For a fuller discussion of line equations, see Maor,
much versed in algebra and geometry as he was in The Pythagorean Theorem: A 4,000-Year History, chapter 10.
physics.2
 41
The French Connection

T here is a time-honored French tradition, going

back at least to the sixteenth century, that calls
for scientists to pursue a career in the military or
R

civil service in parallel to their academic careers:

François Viète, René Descartes, Jean Baptiste Joseph Q
Fourier, and Victor Poncelet—to name but a few—all
served as army officers, military engineers, or public D E
C
administrators at various levels of government. This B
P
tradition is embodied by the lives of two early nine-
teenth-century French geometers, Charles Julien
Brianchon and Jean Victor Poncelet. A
Not much is known about the early life of Brian- F
chon (1783?–1864); even his year of birth is in dis- Figure 41.1
pute, being given as either 1783 or 1785. At the age
of 18 he enrolled at the prestigious École Polytech-
nique in Paris, where he studied under the geometer an ellipse, the three diagonals joining pairs of op-
Gaspard Monge. While there he discovered a theo- posite vertices are concurrent—they pass through
rem that, in disguise, had already been found by an- one point. The two theorems are illustrated in fig-
other Frenchman, Blaise Pascal (1623–1662) more ures. 41.1 and 41.2.1
than one hundred fifty years earlier. Pascal’s theo- Later in life Brianchon joined the French artillery
rem, which he discovered when just 16 years old, corps under the command of Napoleon, seeing ac-
says that if we inscribe a hexagon in an ellipse, the tion in Spain and Portugal. When Napoleon’s cam-
three points of intersection of pairs of opposite sides paign in the Iberian Peninsula ended in defeat, Bri-
are collinear—they lie on one line (if two opposite anchon got a teaching position at the Royal Artillery
sides are parallel, their point of intersection is “the school in Vincennes. In his later years his scientific
point at infinity”). This is the exact dual of Brian- output declined, and he devoted himself almost en-
chon’s theorem: if we circumscribe a hexagon about tirely to teaching.
 139

Plate 41. French Connections

1 4 0   B e a u t i f u l Geometry

A dark blue, respectively. The nine points and the cir-

cle passing through them are shown in orange, with
B a green dot marking its center. The triangle’s cir-
F cumcenter, centroid, and orthocenter are in yellow,
and the Euler line passing through them (see page
123) in red. Quite a lot of action is packed into this
seemingly simple configuration!
E C The discovery of Brianchon’s and Poncelet’s theo-
rem has other contenders, though: Karl Wilhelm
D Figure 41.2 Feuerbach (1800–1834) of Germany, at the age of
22, proved it for six of the nine points (the three me-
dian points and the three altitude feet); he discov-
Poncelet (1788–1867) entered the École Polytech- ered several additional properties associated with it,
nique when he was 15. Like Brianchon, he then pur- and the circle is often named after him. Olry Ter-
sued a military career in the Corps of Engineers. quem (1782–1862), another French mathematician
When, in 1812, Napoleon launched his military cam- with a teaching career in the artillery corps, seems to
paign against Russia, Poncelet joined the ranks in have been the first to prove the theorem for all nine
the march eastward. In October of that year the tide points, and he gave it the name nine-point circle. The
turned against the French, and in the battle of Kras- occasional attribution of the theorem to Euler seems
noi he was injured and captured by the Russians. To doubtful.3
keep his sanity during his long months in captivity
and having no books to study from, Poncelet recon-
N ot e s :
structed from memory much of the mathematics he
had learned at the Polytechnique and then used it to 1. The two theorems can be generalized to include any
completely rewrite a branch of mathematics known member of the conic sections—not just an ellipse—but
as projective geometry, with the principle of duality this requires the introduction of points and lines at infin-
as its cornerstone (see chapter 40). ity, which are not part of Euclidean geometry.
In 1821 Brianchon and Poncelet together pub- For a proof of Brianchon’s theorem, see Eves, A Survey
lished the theorem for which they are best known, of Geometry, pp. 143–44.
the nine-point circle: in any triangle, the feet of the 2. For a proof of the nine-point circle theorem, see Cox-
altitudes, the midpoints of the sides, and the mid- eter, Introduction to Geometry, pp. 18–19.
points of the line segments joining the vertices to The nine-point circle has many other interesting fea-
the orthocenter (the intersection point of the three tures; see http://en.wikipedia.org/wiki/Nine-point_cir
altitudes), all lie on one circle. Furthermore, the cen- cle and http://mathworld.wolfram.com/Nine-PointCir
ter of this circle is halfway between the circumcenter cle.html.
and orthocenter, and its radius is one-half that of the 3. See the article “History of the Nine-Point Circle” by
circumcircle.2 This is shown in plate 41, where the J. S. Mackay, Proceedings of the Edinburgh Mathematical So-
triangle and its circumcircle are colored in light and ciety [1892 (11), pp. 19–61], accessible on the Internet.
 42
The Audible Made Visible

E rnst Florens Friedrich Chladni (1756–1827) is not

among the giant names in the history of science,
but his discoveries had a visual impact—quite liter-
Chladni evidently was a man of wide interests.
Besides music and physics, he showed an interest in
meteorology and speculated that meteorites—once
ally. Born in Wittenberg, Germany, in the same year thought to be the debris from volcanic eruptions—
as Mozart, but outliving him by nearly forty years, actually originated in outer space, a view that has
Chladni was a rare combination of musician and since become the accepted fact. And in 1791—the
physicist. In his book, Discoveries in the Theory of year of Mozart’s death—he invented a musical in-
Sound (1787) he showed how the vibrations of strument, the euphon, consisting of metal bars and
sound-generating objects could be seen visually. He glass rods that were rubbed with a wet finger, some-
experimented with thin metal plates, over which a what similar to the glass harmonica for which Mo-
fine layer of sand was strewn. When the plate was zart wrote his Adagio and Rondo in C minor, K.
excited by rubbing a violin bow against its edge, the 617. But it is his work on vibrating plates that earned
sand arranged itself in beautiful geometric patterns Chladni his name in the history of science. He liter-
of a high degree of symmetry. Chladni realized that ally made sound visible.
these patterns—ridges on which the sand tends to Plate 42 shows nine Chladni figures generated in
accumulate—are nodal lines, lines along which the the physics lab of the Alte Kantonsschule in Aarau,
plate does not vibrate. This is analogous to the nodal Switzerland, under the supervision of our colleague
points along a vibrating string, those points where Dr. Markus Meier. They were formed on triangular,
the string divides itself into individual segments, square, and circular-shaped brass plates, over which
each vibrating independently of the others. a fine semolina powder was strewn. The vibrations
Chladni went public with his discovery, demon- were initiated by a wave generator that produced a
strating it before large audiences across Europe. nearly pure sine wave with frequencies ranging
Among his listeners was Napoleon—himself an from 0.5 to 1.5 kHz (500–1500 cps). The electric sig-
enthusiastic supporter of scientific research—who nals were then converted to mechanical vibrations
rewarded Chladni with an honorarium of 6,000 by a piezo element, and the ensuing patterns were
francs. photographed.
142

Plate 42. Chladni Patterns

 43
Lissajous Figures

A nother scientist who transformed sound into vi-

sual patterns was Jules Antoine Lissajous (1822–
1880). Lissajous was professor of mathematics at the
dians per second), and φ is the phase difference be-
tween them. As time progresses, a point P whose
coordinates are (x, y) will describe a curve whose
Lyceé Saint-Louis in Paris, where he studied all equation can be obtained by eliminating t between
kinds of vibrations and waves. In 1855 he invented a the two preceding equations. For example, if ω1 = ω2
simple optical device for analyzing compound vi- and φ = 0 (the two vibrations being in tune and in
brations. He attached small mirrors to the prongs of phase), we get y = (b/a)x, the equation of a straight
two tuning forks vibrating at right angles to each line through the origin; if, however, φ = π/2 (a 90°
other. When a beam of light was aimed at one of the phase difference), we get x2/a2 + y2/b2 = 1, an ellipse.1
mirrors, it bounced off to the other mirror and from If the phase difference slowly grows with time, this
there to a screen, where it formed a two-dimensional ellipse will continuously change its orientation and
pattern, the result of superimposing the two vibra- shape, passing (in the case a = b) from the circle
tions. This simple device—a forerunner of the mod- x2 + y2 = 1 to the pair of lines y = ± x. When the fre-
ern oscilloscope—was a novelty in his time; up until quencies are unequal, the resulting figure becomes
then the study of sound depended entirely on the much more complex; if the ratio ω1/ω2 is an irratio-
process of hearing, that is, on the human ear. Lissa- nal number, the figure will never close—it will be
jous literally made it possible to “see sound.” aperiodic.
Just as with snowflakes, Lissajous figures, as they Due to their great variety, Lissajous figures be-
came to be known, come in an infinite variety. Any came a favorite of nineteenth-century popular-sci-
change in the parameters of the two vibrations ence demonstrations. Their greatest expositor was
will drastically affect the ensuing figure. To see this, none other than Lissajous himself, who—just like his
let predecessor, Chladni—made the lecture circuit all
across Europe. And like Chladni, among Lissajous’s
x = a sin ω1t, y = b sin (ω2t + φ),
audience was a distinguished guest—Emperor Na-
where a and b are the amplitudes of the two vibra- poleon, this time Napoleon III. Lissajous’s experi-
tions, ω1 and ω2 are their angular frequencies (in ra- ments were shown at the Paris Universal Exposition
144

Plate 43. Lissajous Figures

 4 3 . L i s s a j o u s F i g u r e s   1 4 5

in 1867, in recognition of which he was awarded the cally with a harmonograph—a device consisting of
prestigious Lacaze Prize in 1873. two coupled pendulums oscillating at right angles
Our illustration (plate 43) shows four Lissajous to each other; the frequencies could be varied by ad-
figures with (going clockwise) frequency ratios of justing the length of each pendulum. You can still
1:2, 2:3, 3:4, and 4:5 and equal amplitudes. With see a harmonograph at work at a few science muse-
graphing software you can create a virtually infinite ums, but the digital age has made it obsolete.
variety of these figures, with arbitrary amplitudes,
frequency ratios, and phase differences. It is fasci-
N ot e :
nating to watch how even a small change in any of
these parameters—and especially the frequency 1. To see this, write x = a sin ω1t, y = b sin (ω1t + π/2) =
ratio—can drastically alter the ensuing figure. In b cos ω1t. Dividing the first equation by a and the second
years past, these figures were generated mechani- by b, squaring, and adding, we get x2/a2 + y2/b2 = 1.
 44
Symmetry I

F ew subjects have bridged the divide between the

humanities and science more successfully than
the concept of symmetry. Symmetry is as central to
stay the same under a reflection in each of the three
altitudes, as well as under 120°, 240°, and 360° rota-
tions about its center (more about this in the next
mathematics and physics as it is to the visual arts, chapter).
architecture, music, and aesthetics. To the Greeks, In the nineteenth century these ideas led to the
symmetry meant a balance between the different creation of a new branch of algebra known as group
parts of an object. Most Greek temples have a per- theory. Two names are associated with this develop-
fect bilateral, or reflection, symmetry: if you were ment: the Norwegian Niels Henrik Abel (1802–
looking at a mirror image of the Parthenon, you 1829) and the Frenchman Évariste Galois (1811–
would hardly be able to distinguish it from the ac- 1832). The concept of a group came to them while
tual shrine. To the aesthetically minded Greeks, attempting to find a formula for solving the general
symmetry was synonymous with beauty and perfec- fifth-degree equation—the quintic—using only the
tion; the human body was the ultimate example of elementary operations of addition, subtraction, mul-
such a perfection, and the Greeks realized it in their tiplication, division, and root extraction, as with the
numerous statues and sculptures. familiar quadratic formulas for solving second-de-
Mirror reflection is but one of several kinds of gree equations. As it happened, this turned out to be
symmetry, indeed, the simplest one. In its broadest a pipe dream: Abel and Galois independently proved
sense, symmetry is defined as the set of all transfor- that no such formula exists for the general quintic,
mations that, when acting on an object, leave that nor for the general equation of any higher degree
object invariant, or unchanged. These transforma- (although particular cases of such equations may be
tions may be reflections, rotations, or translations, solved in the manner described).
and they may act on a physical object, a geometric Tragically, both Abel and Galois died all too
figure, or a group of abstract symbols. An isosceles young: Abel of tuberculosis at the age of 27, Galois
triangle, for example, will remain unchanged from in a gun duel with a rival (ostensibly over a girl-
its original orientation if we reflect it in the altitude friend, but more likely because of his antiroyalist
through the top vertex; an equilateral triangle will political views in the years following the French
 147

Plate 44. Gothic Rose

1 4 8   B e a u t i f u l Geometry

Revolution); he was just 20 years old. In their short tion symmetry, while five of the remaining circles
lives they changed the course of mathematics, mak- exhibit threefold rotation symmetries with or with-
ing it more formal, more abstract, and more general out reflection (if you disregard the inner details in
than ever before. some of them). The circle in the 10-o’clock position
Plate 44, Gothic Rose, shows a rosette, a common has the twofold rotation symmetry of the yin-yang
motif on stained glass windows like those one can icon. In the next chapter we will see how these sym-
find at numerous places of worship. The circle at metry patterns can be given a precise mathematical
the center illustrates a fourfold rotation and reflec- formulation.
 45
Symmetry II

A s we just saw, an equilateral triangle is endowed

with six symmetry elements—three 120° rota-
tions and three mirror reflections. Let the triangle be
A C

r1
ABC, with vertex A at the top, followed clockwise by
vertices B and C. Let us denote the six symmetry op-
erations by letters: r1, r2, r3 for the 120°, 240°, and
360° clockwise rotations and m1, m2, m3 for the reflec- C B B A
tions in the altitudes through the top, lower-right, Figure 45.1
and lower-left vertices (henceforth we’ll refer to
these vertices as nos. 1, 2, and 3 rather than A, B, and
C, because they keep changing their position as we
rotate and reflect the triangle; thus the position-1
vertex is always at the top, 2 is at lower right, and 3
is at lower left, regardless of the positions of A, B, same as “doing nothing”—it leaves everything un-
and C). changed. To summarize,
Now a 120° rotation changes triangle ABC to
r1 : ABC → CAB; r2 : ABC → BCA; r3 : ABC → ABC.
CAB, which is just a cyclic permutation of the let-
ters A, B and C, so that C now occupies the 1-posi- Turning now to the reflections, we can summarize
tion, A occupies the 2-position, and B the 3-posi- them as follows:
tion (see figure 45.1). In a similar way, a 240°
m1 : ABC → ACB; m2 : ABC → CBA; m3 : ABC → BAC.
rotation transforms ABC into BCA, and a 360° rota-
tion changes ABC into . . . ABC—it brings the trian- Note that in each reflection one letter stays put,
gle back to its starting orientation. This, of course, while the other two switch positions—exactly what a
should come as no surprise: a 360° rotation is the mirror reflection does (figure 45.2).
1 5 0   B e a u t i f u l Geometry

A A

m1

C B B C Figure 45.2

A C C

r1 m1

C B B A A B Figure 45.3

We are now ready to produce our trump card: we We have, in effect, created a kind of algebra of
can follow any of the six symmetry operations with symmetry operations, similar in some ways to the or-
a second symmetry operation; for example, a 120° dinary algebra of numbers and variables we learn in
rotation followed by a reflection in the 1-position school, but with one crucial difference: unlike num-
amounts to bers, the product of two symmetry operations is
generally noncommutative—the order in which we
ABC → CAB → CBA.
perform the operations does matter. To see this, let
But CBA can be obtained directly from ABC by a re- us again find the product of r1 and m1, but this time
flection in the 2-position (in our case, vertex B; see in reverse order:
figure 45.3). We call this combined application of
ABC → ACB → BAC.
two symmetry operations a product and denote it
with a dot. And just as a product of two numbers That is to say, m1 ⋅ r1 = m3, whereas r1 ⋅ m1 = m2. Thus,
gives us a third number, a product of two symmetry r1 ⋅ m1 ≠ m1 ⋅ r1.
operations results in a third symmetry operation. In If you are willing to spend a few more minutes on
the example just given, this amounts to writing this exercise (it may remind you of your daily Su-
r1 ⋅ m1 = m2. doku), you can create a complete “multiplication
 151

Plate 45. Symmetry

1 5 2   B e a u t i f u l Geometry

table” of the six symmetry operations of the equilat- multiply, their product will again be one of the
eral triangle. Here it is: six symmetry operations: you can never go out-
⋅ r1 r2 r3 m1 m2 m3
side this set.
2. Among the members of our set there is one ele-
r1 r2 r3 r1 m2 m3 m1 ment that has the effect of “doing nothing.” This,
r2 r3 r1 r2 m3 m1 m2 of course, reminds us of the number 1 in ordinary
r3 r1 r2 r3 m1 m2 m3 multiplication: 1 ⋅ a = a for any number a. And in-
m1 m3 m2 m1 r3 r2 r1 deed, this particular element, called the unit or
m2 m1 m3 m2 r1 r3 r2 identity element, is r3, the 360° rotation, since it
m3 m2 m1 m3 r2 r1 r3 leaves the triangle in its original position.
3. Just as with numbers, any multiplication in our
Plate 45 shows all 36 entries of this table in a rain-
set can be undone; that is, every element has an
bow display of color. Note that the upper-left and
inverse, whose product with the original element
lower-right quadrants (on gray background) consist
results in the identity element. For example, the
of rotations alone, while the remaining two quad-
inverse of r1 is r2, because r1 ⋅ r2 = r3. The inverse of
rants are pure reflections. We can summarize this in
any reflection is the same reflection again, since
a miniature table:
the mirror image of a mirror image is the original
⋅ r m image: a reflection is its own inverse.
r r m 4. If you multiply together three elements, the order
m m r of grouping does not matter; symbolically,
a ⋅ (b ⋅ c) = (a ⋅ b) ⋅ c. This is the associative law, with
Here, r and m stand for rotations and reflections of which we are familiar from ordinary arithmetic.
any kind. As this table shows, a succession of two You can convince yourself of the validity of this
rotations or two reflections always results in a rota- law by trying a few examples (or, if you have
tion, whereas a rotation and a reflection result in a the patience, all 432 = 2 ⋅ 63 possible combina-
reflection. You can convince yourself of this when tions); for example, r1 ⋅ (m3 ⋅ r2) = r1 ⋅ m1 = m2, while
looking in the mirror: what you see is not yourself (r1 ⋅ m3) ⋅ r2 = m1 ⋅ r2 = m2: you get the same answer.
but a mirror image of yourself; if you stretch out
Any collection of objects—whatever their nature—
your right hand, your image will respond with its
that fulfills these four requirements is called a group.
left hand! To see a true image of yourself, you have
It was this concept that Abel and Galois indepen-
to look at the intersection of two mirrors at right
dently introduced in their quest for a formula for
angles to each other; you will see yourself as some-
solving the general quintic (Galois suggested the
one else sees you.
name). For more than a hundred years the group
Returning to the triangle’s multiplication table,
concept was regarded as a purely abstract creation,
four features about this table are worth noting:
devoid of any practical applications. But with the
1. The table is “closed” in the sense that no matter rise of modern physics in the twentieth century,
which two symmetry operations you choose to group theory suddenly assumed central stage in
 4 5 . S y mm e t r y II   1 5 3

nearly every branch of science, from crystallography which all other groups can be created, in much the
and quantum mechanics to relativity and particle same way as any integer can be created by multiply-
physics. ing together its prime factors. This achievement, the
One of the great accomplishments of modern result of a collaboration of more than 100 mathema-
mathematics was achieved between 1995 and 2004 ticians, has been compared in its importance to the
with the successful classification of all finite simple discovery half a century earlier of the DNA double
groups—groups with a finite number of elements from helix and the genetic code.
 46
The Reuleaux Triangle

F ranz Reuleaux (1829–1905), notwithstanding his

French-sounding name, was a German scientist
and engineer who is regarded as the founder of mod-
ern kinematics and machine design. He was born to
a family of mechanical engineers and machine build-
ers, an environment that nourished his future inter-
ests. He got his formal education at the Karlsruhe
Polytechnic School and held his first academic ap-
pointment with the Swiss Federal Technical Insti-
tute (ETH) in Zurich. In 1864 he became professor
at the Royal Industrial Academy in Berlin (later the
Royal Technical College), being active as an educa-
tor, industrial scientist, and consultant. His views on
various technical matters had considerable influence
on the subsequent growth of German industry. But
during the last years of his life his reputation de-
clined: he was criticized for basing his ideas entirely
on kinematic principles—that is, the purely geomet-
ric aspects of machinery—while ignoring the even Figure 46.1

more important dynamic effects of forces and

torques.
Reuleaux’s name is mainly remembered today for centers of the other two (figure 46.1). The overlap-
an idea that would lead to some unexpected practi- ping central area is called the Reuleaux triangle. This
cal uses. It all started with a simple geometric con- simple-looking figure has some remarkable proper-
struction: with a compass, draw three identical cir- ties. For example, the width of the Reuleaux triangle
cles, each of radius r and each passing through the is constant; that is, two parallel tangent lines to the
 155

Plate 46. Reuleaux Triangle

1 5 6   B e a u t i f u l Geometry

triangle are always separated by a constant distance ject as their means of transportation, since the circle
r, regardless of where we place them (note that one is regarded by them as a religious icon that should
of these tangent lines will always pass through a not be used for mundane purposes.
vertex). The Reuleaux triangle is but one of a large family
This feature, of course, reminds us of the circle, of constant-breadth curves. It is known that all such
and indeed the two shapes have several properties in curves with the same breadth r also have the same
common. The circumference of the Reuleaux trian- perimeter. Of all the curves in this set, the Reuleaux
gle is πr, the same as the circumference of a circle of triangle has the smallest area, whereas the circle has
diameter r. We can, therefore, regard the radius of the the largest. This remarkable fact was proved in 1916
generating circles as the diameter of the Reuleaux by the Austrian mathematician Wilhelm Blaschke
triangle.1 A Reuleaux triangle will always fit into a (1885–1962). Thus, the circle and the Reuleaux tri-
square of side r, a fact that inspired Harry James angle occupy the opposite ends of the family of con-
Watt (a descendant of the English inventor James stant-breadth curves.3
Watt of steam engine fame) to issue a patent for a
drill that could cut square holes (well, almost square:
N ot e s :
the corners would still be a bit rounded.)2
Because of its constant width, a Reuleaux triangle 1. This analogy, however, does not extend to the trian-
could be used as a wheel, at least in theory; but a gle’s area, which is (r 2/2)(π − 3 ), compared to that of a
ride on a vehicle using these “wheels” would be any- circle of diameter r, πr 2/4.
thing but smooth, as the axle would wobble up and 2. An animation of this drill can be seen at http://math-
down three times during each revolution (see plate world.wolfram.com/ReuleauxTriangle.html.
46). In Poul Anderson’s science fiction story, The 3. More on curves of constant breadth can be found in
Three-Cornered Wheel (1963), inhabitants of an alien Rademacher and Toeplitz, The Enjoyment of Mathematics,
planet are using a noncircular, constant-width ob- chapter 25.
 47
Pick’s Theorem

G eorg Alexander Pick (1859–1942) was born to a

Jewish family in Vienna but spent most of his
academic life in Prague. Pick began his career as as-
two intersection points, but it may contain more (for
example, if the side is along a coordinate line). Pick
discovered an unexpected relation between the area
sistant to some of the great names in mathematics A of this polygon, the number of points M inside the
and physics at the late nineteenth century, among polygon, and the total number of grid points on the
them Ernst Mach (after whom the Mach number in sides N. Pick’s theorem says that A = M + 1⁄2N − 1.
aerodynamics is named) and Felix Klein, whose re- Plate 47 shows a lattice polygon with 28 grid
form program in mathematics education would points (in red) and 185 interior points (in yellow).
greatly influence subsequent generations of mathe- Pick’s formula gives us the area of this polygon as
maticians. In 1892 Pick became professor at the Ger- A = 185 + 28⁄2 − 1 = 198 square units.
man University of Prague, where 20 years later he Pick’s theorem received little attention during his
would play a key role in offering a professorship to a lifetime, and even today few mathematics books
young physicist by the name Albert Einstein; it mention it. It was mainly through a popular work,
would be Einstein’s first full-time academic appoint- Mathematical Snapshots (1950) by the Polish mathe-
ment. The two forged a close friendship, driven by matician Hugo Dyonizy Steinhaus, that the formula
their common passion for science and music. became known to a wider circle.1
Pick’s work covered a wide range of subjects, in- If we rewrite Pick’s formula as 2M + N − 2A = 2, it
cluding algebra, geometry, and analysis, but he is bears a certain similarity to Euler’s famous polyhe-
remembered today mainly for the theorem that dron formula V + F − E = 2 (see page 123). This similar-
bears his name, which he published in 1899. Imag- ity, however, is superficial: Pick’s formula relates two
ine an infinite plane on which a uniform rectangular pure numbers with a geometric quantity, area, while
grid is superimposed. The points of intersection of Euler’s formula is a relation between three pure num-
this grid all have integer coordinates: (0, 0), (0, 1), bers; it does not tell us anything about the polyhe-
(1, 2), and so on. On this grid draw a polygon whose dron’s surface area or volume. Nevertheless, there is a
vertices all lie on the intersection points. Note that subtle connection between them, which came to light
every side of this lattice polygon must connect at least shortly after Steinhaus made the theorem known.2
158

Plate 47. Pick’s Theorem

 4 7 . P i ck ’ s T h e o r e m   1 5 9

Pick’s later years were clouded by the political at-

N ot e s :
mosphere of the 1920s and 1930s. Now retired, he
hoped to live out his remaining years in Vienna, the 1. For a proof of Pick’s theorem, see Coxeter, Introduc-
place of his birth. But it was not to be: following the tion to Geometry, pp. 208–209.
Anschluss—the Nazi occupation of Austria in 1938— 2. See the following articles in The American Mathemati-
he managed to go back to Prague, only to be cal Monthly: “From Euler’s Formula to Pick’s Formula
rounded up there a year later when the Nazis in- Using an Edge Theorem” by W. W. Funkenbusch [1974
vaded Czechoslovakia. Together with some one (81), pp. 647–48]; “Pick’s Theorem Revisited” by Dale E.
hundred fifty thousand fellow Jews, he was deported Varberg [1985 (92), pp. 584–87]; and “Pick’s Theorem” by
to the Theresienstadt concentration camp, where he Branko Grünbaum and G. C. Shephard [1993 (100)), pp.
died in 1942. 150–61].
 48
Morley’s Theorem

F rank Morley (1860–1937) was an English-born

mathematician who moved to the United States
in 1887. He was professor at Johns Hopkins Univer-
C

sity from 1900 to 1928 and editor of the American

Journal of Mathematics from 1900 to 1921. He pub-
Q
lished several books, including two on the theory of
R
functions of complex variables (1893 and 1898), in
B
which he introduced that subject—one of the high-
lights of nineteenth-century European mathemat- P

ics—to American readers. He was known as an out-

standing teacher who had no fewer than 45 doctoral
students, an incredible number for any professor.
He was also a gifted chess player and once beat the
world champion Emmanuel Lasker.
But if Morley’s name is remembered today, it is A Figure 48.1
mainly for a theorem he discovered in 1899 but did
not publish until 1929. Take any triangle ABC and
draw its three pairs of angle trisectors, each pair origi-
nating at one vertex (see figure 48.1). Now consider could have been included in Euclid’s Elements if only
the two trisectors adjacent to side AB; they meet at Euclid had known about it—but he didn’t, nor did
point P. Similarly, the trisectors adjacent to side BC anyone else for the next 2,200 years. It shows that
meet at Q, and those adjacent to side CA meet at R. even in classical geometry, surprises may still be
Morley’s theorem says that points P, Q, and R form awaiting us around the corner—or perhaps around
an equilateral triangle, appropriately called the Mor- the vertex!
ley triangle and shown in yellow in the figure. This The trisectors we have considered trisected the in-
unexpected result came totally out of the blue; it terior angles at A, B, and C, but Morley’s theorem
 161

Plate 48. Morley’s Theorem

1 6 2  

Figure 48.2
 4 8 . M o r l e y ’ s T h e o r e m   1 6 3

holds true for the exterior angles as well. Let the pairs As an additional bonus, the side of Morley’s inter-
of exterior trisectors adjacent to sides AB, BC, and nal triangle is equal to 8r sin(A/3) sin(B/3) sin(C/3),
CA meet at K, L, and M, respectively (figure 48.2); where r is the circumradius of the original triangle
points K, L, and M again form an equilateral trian- and A, B, C are its angles.
gle, shown in purple in the figure. And that’s not all: Plate 48 is a whimsical play on Morley’s theorem,
if we take the intersections of pairs of the remaining in which three key events of his life are alluded to. It
trisectors with the extended sides of this second Mor- shouldn’t be too difficult to decode the meaning of
ley triangle, we get three more equilateral triangles the various symbols shown, so we leave this task to
(shown in red)! Considering that all five equilateral the reader.
triangles were generated from a single, generic tri-
angle, these results are truly remarkable.
N ot e s :
One may wonder why this jewel of a theorem
hadn’t been discovered earlier. A possible explana- 1. For a proof of Morley’s theorem, see Coxeter, Intro-
tion is that angle trisection is one of the three classi- duction to Geometry, pp. 23–25. Many alternative proofs
cal problems of antiquity that have no solution and related results can be found at the Internet sites listed
using only the Euclidean tools (see chapter 25). in the bibliography.
Consequently, problems involving angle trisection 2. Figures 48.1 and 48.2 are color renditions based on
may have been regarded as of little interest and were, the drawings in Wells, The Penguin Dictionary of Curious
therefore, neglected. and Interesting Geometry, p. 155.
 49
The Snowflake Curve

N iels Fabian Helge von Koch (1870–1924) was a

Swedish mathematician who is remembered
today chiefly, if not solely, for a curious curve he dis-
of the original triangle. Now repeat the process with
each of the 12 new sides, resulting in a 48-sided
shape whose perimeter is 4⁄3 that of the previous pe-
covered in 1904. Take an equilateral triangle of unit rimeter, and thus (4⁄3)2 = 16⁄9 of the original perimeter.
side, divide each side into three equal parts, each of Continuing in this way, the perimeter will increase
length 1⁄3, and delete the middle part (figure 49.1). by a factor of 4⁄3 with each step. Since this factor is
Over the deleted part, construct the two sides of an greater than 1, the perimeter keeps growing without
equilateral triangle of side 1⁄3. This gives you a 12- bound and becomes infinite (more formally, (4⁄3)n →
sided Star of David shape, whose perimeter is 4⁄3 that ∞ as n → ∞).

Figure 49.1
 165

Plate 49. Variations on a Snowflake Curve

1 6 6   B e a u t i f u l Geometry

As you repeat the process again and again, the 3 12 48 1 4 42

curve retains its overall Star of David shape but be- 1+ + + + = 1+ + 3 + 5 +
9 81 729 3 3 3
comes increasingly crinkly. At the same time, the 1
distance between any two points, no matter how = 1 + [1 + 4 / 9 + (4 / 9)2 + ].
3
close they are, becomes infinite, and, consequently,
the curve has no tangent line anywhere. This, in- The expression inside the brackets is a geometric
deed, was the motivation behind its discovery: Koch series with the common ratio 4⁄9, so its sum is
attempted to show that there exist continuous curves 1/(1 − 4⁄9) = 9⁄5 (see page 51). The total area is there-
with no tangent lines (curves that are nowhere dif- fore equal to 1 + 1⁄3 ⋅ 9⁄5 = 8⁄5, or 1.6 times the area of
ferentiable, in the language of calculus). Such curves the original triangle. So here we have a closed shape
were a novelty in Koch’s time, so much so that they with a finite area but an infinite perimeter: you could
were dubbed “pathological curves”—a hint to their never fence it off, as any fence would have infinitely
strange behavior. many corners at every section of it. No wonder the
But what was strange yesterday has become the Koch curve has become the darling of popularizers
commonplace of today in the form of fractals—curves of mathematics.
with the property that any small part of them looks Plate 49 is an artistic interpretation of Koch’s
exactly like the entire curve (mathematicians call curve, starting at the center with an equilateral tri-
this self-similarity). In fact, Koch’s curve—also known angle and a hexagram (Star of David) design but
as the snowflake curve due to its shape—was one of approaching the actual curve as we move toward the
the first known fractals and certainly the first to periphery.
make it into popular culture.1
One more feature of the Koch curve is worth not-
N ot e :
ing: it encompasses a finite area. To see this, start
again with the initial triangle of side 1. In the first 1. The name fractal is credited to the French-born
step we add three small triangles, each of side 1⁄3 and American mathematician Benoit Mandelbrot (1924–
thus of area 1⁄32 = 1⁄9 of the area of the initial triangle. 2010), who coined it in 1975.
In the next step we add 12 new triangles, each of
side 1⁄9 and area 1⁄92 = 1⁄81 of the initial triangle. Con-
tinuing in this manner, we get the series
 50
Sierpinski’s Triangle

W aclaw Franciszek Sierpinski (1882–1969) was

born in Warsaw. He entered the University of
Warsaw in 1899, graduating in 1904; four years later
and gave the first-ever university course devoted en-
tirely to set theory. The topic would occupy him for
the rest of his life.1
he was appointed to the University of Lvov. He be- As an offshoot of his research, Sierpinski in 1915
came interested in set theory after reading about a came up with a seemingly impossible geometric
theorem that allows for a seemingly impossible situ- configuration: a triangle-like figure whose area is
ation. We learn in analytic geometry that any point zero. Start with an equilateral triangle (figure 50.1),
in the plane can be uniquely specified by two num- and remove from it the small central triangle formed
bers, its x- and y-coordinates. This has been the rock by connecting the midpoints of the sides (shown in
foundation of analysis ever since René Descartes in- white in the figure). Now repeat the process with the
vented his analytic geometry in 1637. Not so, said three remaining black triangles, then repeat it again
the theorem that Sierpinski came across: one num- and again . . . forever. In the limit, what will be the
ber suffices. Intrigued, he wrote to the article’s au- area of all the black triangles?
thor, Tadeusz Banachiewicz, asking for an explana- Taking the area of the initial triangle to be 1, each
tion. He got a reply of a single word: “Cantor,” of the three smaller black triangles has an area 1⁄4, so
referring to the creator of modern set theory, Georg their combined area is 3 × 1⁄4 = 3⁄4. In the next step we
Cantor, the subject of our last chapter. Becoming have nine black triangles, each with area 1⁄16, making
hooked on the subject, Sierpinski started studying it the total area 9 × 1⁄16 = (3⁄4)2. Continuing in this man-

Figure 50.1
168

Plate 50. Sierpinski’s Triangle

 5 0 . S i e r p i n s k i ’ s T r i a ng l e   1 6 9

ner, the black areas follow the progression 1, 3⁄4, While we cannot go to infinity in the actual world,
(3⁄4)2, (3⁄4)3, . . ., a geometric progression with the we can get pretty close with a computer, stopping
common ratio 3⁄4. Since this ratio is less than 1, the only when the resolution reaches its limit at the pixel
terms of the progression tend to 0 as n → ∞. So, level. Plate 50 shows the seventh stage of Sierpins-
eventually the original black triangle will become ki’s triangle.
empty, despite the fact that at each stage we removed
only 1⁄4 of each black area.
N ot e s :
On the other hand, the perimeters of the black
triangles follow the sequence 3, 9⁄2, 27⁄4, 81⁄8, . . ., a geo- 1. This biographical sketch is based on the article “Wa-
metric progression with the common ratio 3⁄2. Since claw Sierpinski” by John J. O’Connor and Edmund F.
this ratio is greater than 1, the terms grow without Robertson in The MacTutor History of Mathematics Archive
bound as we keep removing more and more trian- (on the Web), listed in the bibliography.
gles, making the limiting perimeter infinite. That is 2. The formation of Sierpinski’s triangle can be viewed
to say, the limiting shape—known as Sierpinski’s tri- as an animation at http://en.wikipedia.org/wiki/Sierpin
angle—has zero area but infinite length! It shows ski_triangle.
again that when infinity comes into play, strange
things always lurk around the corner.2
 51
Beyond Infinity

C an anything be larger than infinity? No, says

common sense. But who is to say that common
sense is always right—especially when it comes to
O

the infinite, a world beyond our physical reach? “In-

finity is a place where things happen that don’t,” an
anonymous school pupil once said. Railroad tracks,
though perfectly parallel, seem to meet far away on
the horizon—at infinity; indeed, in projective geom-
etry they do meet at infinity (see chapter 40). Yet
when we try to reach that elusive point, it recedes
from us just as fast as we approach it. C D
Again, when comparing two infinite sets, who is
to tell which is the larger? There seem to be twice as
many counting numbers as there are even numbers, A B
and yet we can match every counting number with
Figure 51.1
its double, showing that the two sets are just as large.
And certainly there should be more points on a long
line segment than on a short one, yet they both con-
tain the same number of points—the same, in fact, as his productive years. Beginning in 1874, Cantor
the entire number line (see figure 51.1). Strange, in- published a series of articles that at once put the
deed, is the world of infinity. concept of infinity on its head. To begin with, he
Enter Georg Cantor (1845–1918). Born in St. Pe- insisted that an infinite set should be regarded in its
tersburg, Russia, his family moved to Germany entirety, as one whole. This ran smack against the
when he was 11. In 1869 he settled in the city of accepted notion­—going back to the Greeks—that in-
Halle, at whose university he would spend most of finity can be thought of only as a process, never as a
 171

Plate 51. The Rationals Are Countable!

1 7 2   B e a u t i f u l Geometry

finished product. For example, the act of counting 2. The real numbers, on the other hand, are uncount-
the natural numbers 1, 2, 3, . . . can go on forever, yet able—there are infinitely many more of them than
at each stage only a finite number of them have been there are counting numbers or rational numbers.
counted; the complete process can never be finished. They cannot be put in a 1:1 correspondence with
Cantor thought otherwise: the natural numbers, he any denumerable set; their infinity is of a higher
said, should be regarded as a single, complete ob- rank than that of denumerable sets. Cantor as-
ject, symbolized by enclosing them in braces: {1, 2, signed this kind of infinity the Gothic letter C, the
3, . . .}. In effect, Cantor declared that infinity, far infinity of the continuum. In the hierarchy of in-
from being the elusive goal of a never-ending pro- finities, C > ℵ0.
cess, is a mathematical reality. 3. There exist sets still more numerous than the reals
But this was only the harbinger of things to come. and, therefore, of a higher degree of infinity than
Consider the set of rational numbers. We know that even C. Consider the set of all subsets of a given
between any two fractions, no matter how close, you set—its power set. For example, starting with the
can always squeeze a third fraction. For example, two-element set {a, b}, we can create a new set that
between 1/1,001 and 1/1,000 you can fit the fraction has all these elements as subsets: {{a}, {b}, {a, b},
2/2,001 and, indeed, infinitely many more fractions. { }} (note that we included the empty set, { },
The rational numbers, then, are spread densely among the subsets). This new set has 4 = 22 ele-
along the number line, and it would seem only natu- ments. We can now repeat the process with this
ral to assume that there are many more of them than new set, getting a power set of 16 = 24 elements.
counting numbers—infinitely many more. But “nat- The number of subsets in this process grows very
ural” is a very poor guide when it comes to infinity. fast: the next power set will have 216 = 65,536 ele-
In a series of groundbreaking papers published in ments, and the one after that, 265,536 elements, ap-
the decade 1874–1884, Cantor established three re- proximately 1 followed by 19,728 zeros. The pro-
sults that seem to defy common sense: cess can go on forever, generating ever-larger sets
of enormous, yet still finite, size.
1. There are just as many positive rational numbers
as counting numbers; that is to say, we can enlist Cantor now imagined that we can do the same
all positive rationals in a row and count them one with infinite sets, generating a never-ending chain of
by one, without leaving a single one out, as de- power sets, each larger than its predecessor. Each of
picted in plate 51. Cantor called any such set a these sets stands one rung higher on the ladder of
countable, or denumerable, set. The even numbers, infinities. Cantor pointed out that these sets are
the odd numbers, the squares, the primes, and the purely a creation of the mind; they cannot be con-
rational numbers are all denumerable sets: their structed in any “real” sense; they reside in the ethe-
members can be put in a one-to-one correspon- real sphere of abstract mathematics. And yet their
dence with the counting numbers. Cantor de- existence is as real as that of any other mathematical
noted their infinity with the symbol ℵ0, the infin- object, concrete or abstract.1
ity of countable sets (ℵ, pronounced “aleph,” is Cantor’s last years were not happy ones. Stung by
the first letter of the Hebrew alphabet). the relentless opposition to his radical ideas and suf-
 5 1 . B e y o n d I n f i n i t y   1 7 3

fering from repeated spells of depression, he spent Hold infinity in the palm of your hand,
his final years in a mental institution, where he died And eternity in an hour
in 1918. Yet his ideas slowly gained acceptance. In a
way, he accomplished the vision of William Blake’s
N ot e :
famous verse in Auguries of Innocence:
1. For a more complete account of Cantor’s theory, see
To see the world in a grain of sand, Maor, To Infinity and Beyond, chapters 9 and 10 (Princeton,
And heaven in a wild flower. NJ: Princeton University Press, 1991).
 Appendix
P r o o fs o f S el ec ted T h eo rems Menti o ned i n Th is B ook

Q ua d r i lat e rals ( pag e 6 ) A

We refer to figure A.1. Let □ABCD denote the area of P

S
quadrilateral ABCD, ∆ABD, the area of triangle ABD,
etc. Because triangle APS has half the base length B
and half the height of triangle ABD, we have D
∆APS = 1⁄4∆ABD. Similarly, ∆CQR = 1⁄4∆CBD. Thus,
1 1 1
∆APS + ∆CQR = ∆ABD + ∆CBD =  ABCD . Q
R
4 4 4

By the same argument,

C
Figure A.1
1 1 1
∆BPQ + ∆DSR = ∆BAC + ∆DAC = ABCD . A
4 4 4

Therefore,

PQRS = ABCD − [∆APS + ∆CQR + ∆BPQ + ∆DSR ] S

1 1
= ABCD − ABCD = ABCD . P
2 2 C

The proof still holds when the quadrilateral has R

the shape of a dart (see figure A.2), except that some Q

D

of the internal areas must be subtracted rather than

added. We leave the proof of this case to the reader.
B Figure A.2
1 7 6    A p p e n d i x

P yt hago r ea n T r i ple s ( pag e 2 0 ) and squaring both sides, we get m2 = 2n2, which
means that m2—and, therefore, m—is an even integer
Let u and v be any positive integers, with u > v. Set
(since the square of an odd integer is always odd).
a = u2 − v2, b = 2uv, and c = u2 + v2. Then
So m = 2r for some integer r. Putting this back into
the equation m2 = 2n2, we get 4r 2 = 2n2, or n2 = 2r 2. But
a2 + b2 = (u2 −v2) 2 + (2uv) 2
this means that n2—and, therefore, n—is also even,
= u4 − 2u2v2 + v4 + 4u2v2
so n = 2s for some integer s. Thus, m and n have a
= u4 + 2u2v2 + v4 = (u2 + v2) 2 = c2,
common factor 2, in contradiction to our assump-
showing that (a, b, c) is a Pythagorean triple. For tion that the fraction m/n is in lowest terms. We con-
example, if u = 5 and v = 3, we get a = 52 − 32 = 25 − 9 = 16, clude that 2 is not rational—QED.
b = 2 × 5 × 3 = 30, and c = 52 + 32 = 25 + 9 = 34, giving us
the triple (16, 30, 34).
Eu c l i d ’ s P r o o f o f t he Infinitude of
In this example we got a nonprimitive triple,
t he P r i m e s (page 44)
being merely a magnification of the triple (8,15,17)
by a factor of 2. To get primitive triples, two restric- We again follow an indirect proof (a proof by con-
tions must be imposed on u and v: they must be rela- tradiction). Assume there is only a finite number of
tively prime (having no common factors other than primes, p1, p2, p3, . . ., pn Now consider the number N
1), and they must be of opposite parities— one even = p1 ⋅ p2 ⋅ p3 ⋅ . . . ⋅ pn + 1. This number is obviously
and the other odd. For example, choosing u = 5 and greater than any of the pi, yet it is not divisible by
v = 4 gives us the primitive triple (9, 40, 41). any of them, because any such division will leave a
Since the choice of u and v is otherwise arbitrary, remainder 1. Therefore, N must either be a new
we see that there are infinitely many primitive Py- prime, not included in the original list, or else it is
thagorean triples. The converse is also true: every composite, in which case at least one of its prime fac-
primitive Pythagorean triple corresponds to a pair tors must be a new prime not included in the origi-
of values (u, v) subject to the restrictions mentioned nal list. In either case we have a contradiction, for
previously. For the details, see Maor, The Pythagorean we assumed that the set p1, p2, p3, . . ., pn included all
Theorem: A 4,000-Year History, pp. 221–22. the primes—QED.
To illustrate, suppose there are only three primes,
2, 3, and 5. Then N = 2 ⋅ 3 ⋅ 5 + 1 = 31, a new prime. On
the other hand, if we started with 3, 5, and 7, we
A Proof That  Is Irrational (page 23)
would get N = 3 ⋅ 5 ⋅ 7 + 1 = 106 = 2 ⋅ 53, producing two
The proof is by the indirect method: we assume that new primes that were not in the original set. We can
2 is rational and then show that this leads to a then add these primes to our list and start the pro-
contradiction. cess again, generating ever more primes (although
Suppose 2 is rational, that is, 2 = m/n, where m not all of them, and not in sequential order), like a
and n are two positive integers with no common fac- runaway nuclear chain reaction, but fortunately
tor other than 1. Rewriting this equation as m = 2n with less-dire consequences.
 A p p e n d i x    1 7 7

The Sum o f a Ge o m e tr i c P r o g r e ssi o n terms except −F2 and Fn+2 cancel out. Remembering
(pag e 5 1 ) that F2 = 1, we thus get
Consider the finite geometric progression a, ar, F1 + F2 + F3 + ⋅⋅⋅ + Fn = Fn+2 − 1.
ar 2, . . ., arn−1 of n terms. Let its sum be S:
A slightly different formula holds for the sum of the
S = a + ar + ar 2 + ⋅⋅⋅ + ar n−1. squares of the first n Fibonacci numbers:
F12 + F22 + F32 +  + Fn2 = Fn Fn +1 ,
Multiply this equation by the common ratio r:
as can be proved by mathematical induction.
Sr = ar + ar + ar + ⋅⋅⋅ + ar .
2 3 n Many more Fibonacci-related formulas can be
found in Alfred S. Posamentier and Ingmar L.
Subtract the second equation from the first: all terms Lehmann, The (Fabulous) Fibonacci Numbers (Am-
cancel out except the first and last: herst, NY: Prometheus Books, 2007).

S − Sr = a − ar n
C o n st r u c ti o n o f a R egular
P e ntag o n (pag e 70)
or S(1 − r) = a(1 − r n), from which we get
We first show that the 72-72-36-degree triangle
a(1 − r n ) formed by any side of the pentagon and the vertex
S= ,
1−r opposite it has a side-to-base ratio equal to the
golden ratio φ = (1 + 5)/2.
provided, of course that r ≠ 1. Now, if r is less than 1 Let the triangle be ABC, with the 36° angle at C
in absolute value (that is, −1 < r < 1), the term r n gets (figure A.3). Bisect angle A and extend the bisector
smaller and smaller as n increases; that is, r n → 0 as until it meets side BC at D. This produces two isosce-
n → ∞. Thus, the sum of the infinite progression is
C
a
S∞ = . 36°
1−r

1
Th e S u m o f t h e Fi r st n F ibo n ac c i
x x
N u m b e rs (pag e 64 )
D
Let F1 = F2 = 1, Fi = Fi−2 + Fi−1 for i = 3, 4, . . . . Note that 72°
1
x -1
F1 = F3 − F2, F2 = F4 − F3, F3 = F5 − F4, . . ., Fn = Fn+2 − Fn+1.
36°
Adding up these expressions, we get, on the left 72° 36° 72°
A B
side, F1 + F2 + F3 + ⋅⋅⋅ + Fn, while on the right side all 1 Figure A.3
1 7 8    A p p e n d i x

les triangles—the 72-72-36-degree triangle BDA and 1 x −1

= ,
the 36-36-108-degree triangle ACD. Of these, the for- x 1
mer is similar to triangle ABC, so we have which leads to the quadratic equation x2 − x − 1 = 0.
Solving it and taking only the positive solution (be-
AB BD
= cause x stands for length, which cannot be nega-
AC BA
tive), we get
(all line segments are nondirectional, so that AB 1+ 5
x= ,
= BA ). Setting AB = 1 and AC = x , we have AB = AD 2
= CD = 1, AC = BC = x , and BD = x − 1. We thus get which is exactly the golden ratio φ.

√ 5/2
1

A 1 B 1/2 P √ 5/2 Q

φ φ

E C

D Figure A.4
 A p p e n d i x    1 7 9

To construct a pentagon with unit side, we first gravity of weights WA, WB, and WC attached to the
need to construct a segment equal in length to the respective vertices. Let us find the center of gravity
golden ratio. Referring to figure A.4, draw a unit of any two of these weights, say WB and WC; call this
line segment AB = 1 , extend it to the right by seg­ point A'. By Archimedes’s law of the lever (the “see­
ment BP = 1/2, and at B erect a perpendicular saw rule”), we have
2 2 2
BR = 1. We have PR = BP + BR = (1/2)2 + 12 = 5/4,
so PR = 5/2. Place your compass at P and swing an W B ⋅ BA' = WC ⋅ A'C ,
arc of radius PR , meeting the extended segment AB
at Q. We have from which we get

1 5 (1 + 5) BA' WC
BQ = BP + PQ = + = =ϕ. = .
2 2 2 A'C W B

We are now ready to construct our pentagon. In a similar manner we get for the other two pairs of
Place your compass at B and swing an arc of radius points,
BQ ; then do the same with your compass at A (only CB' W A AC' W B
the first of these arcs is shown in full in figure A.4). = and = .
B'A WC C'B W A
The two arcs intersect at D. Next, from each of the
points A and D swing an arc of radius AB ; the two
arcs intersect at E (see following note). Do the same
from points B and D, producing point C. Connect C
points A, B, C, D and E, and your pentagon is
complete.
Note: for the arcs to intersect, each radius must be
greater than half the segment. Indeed, AD = BQ =
(1 + 5)/2 ≈ 1.618, so AE = 1 > AD/2 ≈ 0.809, ensur­
A‘
ing that the arcs meet.
B‘
For further discussion and alternative ways to G
construct a pentagon, see Hartshorne, Geometry: Eu-
clid and Beyond, chapter. 4, and Heilbron, Geometry B
Civilized, pp. 221–228.
C‘

C e va’s T h e o r e m (pag e 1 0 5 )

Let the triangle be ABC (figure A.5), and let A', B',
and C' be any points on the sides opposite to A, B,
and C such that AA', BB', and CC' pass through one
point G. Ceva thought of this point as the center of A Figure A.5
 1 8 0    A p p e n d i x

Multiplying the three ratios, we get This means that triangles OPQ and OQ'P' are similar.
Since P is the point on l closest to O, line OP is per­
BA' CB' AC' WC W A W B
⋅ ⋅ = ⋅ ⋅ = 1. pendicular to l, so ∆OPQ is a right triangle with its
A'C B'A C'B W B WC W A
right angle at P. Therefore, triangle OQ'P' is also a
Note that in forming this triple product, we went in right triangle with its right angle at Q', and this is
a counterclockwise direction around the triangle, true regardless of the position of Q on l. Thus, by the
from B to A' to C, then to B', and so on. Since the converse of Thales’s theorem (see chapter 1, note 1),
center of gravity G of the entire triangle must lie on as Q moves along l, its image Q' describes an arc of a
each of the lines AA', BB', and CC', it must lie on their circle k with diameter OP'.
intersection, so the three lines are concurrent. Inter­ The converse of this property is also true: inver­
estingly, the actual weights disappeared from the sion carries a circle through O onto a line not through
final result—they just served as a tool and canceled O. This can be seen by simply reading figure A.6
out at the end. “backward,” from circle k to line l. This is a result of
This, of course, is a physically motivated proof
with which a pure mathematician might disagree.
But let us remember that Archimedes, the quint­
essential pure mathematician, often used subtle
physical reasoning in his proofs, yet he always sup­
plemented them with a rigorous mathematical argu­
Q
ment. For a strictly geometric proof, see Eves, A Sur-
vey of Geometry, pp. 247–48.
Q΄

S o m e Pr o p e rt i e s of I n v e r s i o n
P΄ P
(page 1 2 8 ) O

Let the circle of inversion be c, with center at

O and radius 1 (figure A.6). We first show that in­
version transforms a straight line not through O into k
a circle through O. Let the line be l. Choose two
c
points on l, the point P closest to O and any other
point Q, and let their images under inversion be l
P' and Q', respectively. By the definition of inversion
we have OP' = 1/OP and OQ' = 1/OQ , so OP ⋅ OP' = '
' OQ ⋅ OQ' = 1, from which we get

OP OQ' Figure A.6

= .
OQ OP'
 App e n d i x    1 8 1

the fact that inversion is completely symmetric: if P' It can also be shown that inversion carries a circle
is the image of P, then P is the image of P', as can be not through O to another circle not through O. For
seen from the equivalence of the defining equations the details, see Courant and Robins, What is Mathe-
OP' = 1/OP and OP = 1/OP'. matics, pp. 142–144.
 Bibliography

Baptist, Peter, Albrecht Beutelspacher, and Eugen Jost. Hidetoshi, Fukagawa, and Tony Rothman. Sacred Mathe-
Alles ist Zahl (All is Number). Cologne: Cologne matics: Japanese Temple Geometry. Princeton, NJ:
University Press, 2009. Princeton University Press, 2008.
Courant, Richard, and Herbert Robbins. What is Mathe- Kemp, Martin. The Science of Art: Optical Themes in Western
matics? Revised ed. by Ian Stewart. New York and Art from Brunelleschi to Seurat. New Haven, CT.
Oxford: Oxford University Press, 1996. and London: Yale University Press, 1990.
Coxeter, H.S.M. Introduction to Geometry. 2nd ed. New Livio, Mario. The Golden Ratio: The Story of Phi, the World’s
York: John Wiley, 1969. Most Astonishing Number. New York: Broadway
———. The Beauty of Geometry. New York: Dover, 1999. Books, 2002.
Darling, David. The Universal Book of Mathematics: From Mankiewicz, Richard. The Story of Mathematics. Princeton,
Abracadabra to Zeno’s Paradoxes. Hoboken, N.J.: NJ: Princeton University Press (ND).
John Wiley, 2004. Maor, Eli. The Pythagorean Theorem: A 4,000-Year History.
Euclid: The Elements, translated with introduction and Princeton, NJ: Princeton University Press, 2007.
commentary by Sir Thomas Heath. 3 vols. New ———. Trigonometric Delights. Princeton, NJ: Princeton
York: Dover, 1956. University Press, 2002.
Eves, Howard. A Survey of Geometry. Revised ed. Boston: ———. e: the Story of a Number. Princeton, NJ: Princeton
Allyn and Bacon, 1972. University Press, 2009.
———. An Introduction to the History of Mathematics, 6th ed. Pickover, Clifford A. The Math Book: From Pythagoras to the
Fort Worth: Saunders College Publishing, 1990. 57th Dimension, 250 Milestones in the History of Math-
Gombrich, E. H. The Sense of Order: A Study in the Psychol- ematics. New York and London: Sterling, 2009.
ogy of Decorative Art. Ithaca, NY: Cornell Univer- Posamentier, Alfred S., and Ingmar Lehmann. The (Fabu-
sity Press, 1979. lous) Fibonacci Numbers. New York: Prometheus
Hartshorne, Robin. Geometry: Euclid and Beyond. New Books, 2007.
York: Springer Verlag, 2000. Rademacher, Hans, and Otto Toeplitz. The Enjoyment of
Heilbron, J. L. Geometry Civilized: History, Culture, and Mathematics. Princeton, NJ: Princeton University
Technique. Oxford, UK: Clarendon Press, 1998. Press, 1957.
1 8 4   B i b l i o g r a ph y

Seymour, Dale, and Reuben Schadler. Creative Construc-

Web Sites
tions. Oak Lawn, IL: Ideal School Supply Com-
pany, 1994. Cut The Knot: http://www.cut-the-knot.org/front.shtml.
Steinhaus, H. Mathematical Snapshots. New York: Oxford Eugen Jost on the web: http://www.everything-is-number
University Press, 1969. .net.
Stevens, Peter S. Handbook of Regular Patterns: An Introduc- The MacTutor History of Mathematics Archive, created by
tion to Symmetry in Two Dimensions. Cambridge, John J. O’Connor and Edmund F. Robertson,
MA. and London: MIT Press, 1981. The School of Mathematics and Statistics, Uni-
Wells, David. The Penguin Dictionary of Curious and Interest- versity of St. Andrews, Scotland: http://www
ing Geometry. London: Penguin Books, 1991. -history.mcs.st-andrews.ac.uk/history/.
Weyl, Hermann, Symmetry. Princeton, NJ: Princeton Uni- Wolfram MathWorld: http://mathworld.wolfram.com/.
versity Press, 1952.
 Index

Abel, Niels Henrik, 146, 148 countable set(s), 172

Albers, Josef, 15 Coxeter, Harold Scott MacDonald (H.S.M. Coxeter),
algebraic number, 87 125
Alte Kantonsschule Aarau (Old Cantonal High School cycloid, 116–­118
Aarau), ix, 141
Anderson, Poul, The Three-­Cornered Wheel, 156 da Vinci, Leonardo, 68
Apollonius of Perga, 94 Delian problem (doubling the cube), 81–­83
Archimedes of Syracuse, 88, 96, 180; Measurement of a Cir- Descartes, René, 167
cle, 88–­90 Diophantus of Alexandria, 77
arithmetic-­geometric-­harmonic mean inequality, 29–­31 duality, principle of, 134
astroid, 119
e, 101, 108–­111
Banachiewicz, Tadeusz, 167 Einstein, Albert, ix, 157
Bernoulli, Jakob, 112, 116 Elements, the (Euclid), 3, 87n1; propositions: I 1, 56; I 27,
Bernoulli, Johann, 111, 116 41; I 38, 3, 16, 19, 41; I 47, 16–­19, 41; III 35, 36, 32–­35,
Bill, Max, xi 36–­38, 39; IV 16, 76n3; VI 13, 39–­41
Blake, William, Auguries of Innocence, 173 eleven, 53–­55
Blaschke, Wilhelm, 156 ellipse, 96–­98
brachistochrone, 116, 118 epicycle(s), 122
Brianchon, Charles Julien, 138, 140 epicycloids, 119–­122
Bungus, Petrus, 53 Escher, Maurits Cornelis (M. C. Escher), 112
Euclid, xi, 3, 87; and perfect numbers, 11
Cantor, Georg, 52, 167, 170–­173 Euclidean constructions, 56–­58, 87
Carnot, Lazare, 38 Euler, Leonhard, 45, 75, 123; and e, 110–­111; formula V +
catenary (hanging chain), 110–­111 F − E = 2, 123, 157; line, 123–­125, 140; and perfect
Ceulen, Ludolph van, 91 numbers, 9, 11, 44; series for pi, 93
Ceva, Giovanni, 105; theorem of, 105–­107, 179 exponential function, 110
Chladni, Ernst Florens Friedrich, 141; figures, 141–­142
Chokuyen, Ajima, 132 Fermat, Pierre de, 75; primes, 75–­76
classical problems of antiquity, 81–­83, 163 Feuerbach, Karl Wilhelm, 140
conic sections, 94–­98 Fibonacci (Leonardo of Pisa), 62; Liber Abaci, 62; num-
constant-­breadth curves, 156 bers, 62–­65, 68
continuum, infinity of, 172 fifty, 77–­80
1 8 6   i n d e x

fractal(s), 166 Mercator, Nicolaus, infinite series of, 101

fundamental theorem of arithmetic, 42 Mersenne, Marin, 42; primes, 42, 44
Mohr, Jørgen, 58
Galileo, Galilei, 116 Morley, Frank, 160; theorem, 160–­163; triangle, 160
Galois, Évariste, 146, 148 musical intervals, 23
Gardner, Martin, 47–­48
Gauss, Carl Friedrich, 44, 73–­76 Newton, Isaac, 98, 116
geometric series, 49–­52, 99–­101, 177 nine-­point circle, 140
Goldbach, Christian, 45; conjecture, 45
golden ratio, 65, 66–­69, 70, 177–­179 Oresme, Nicole, 102–­103
Gregory, James, infinite series of, 93
group(s), 146, 152–­153; finite simple, 153 parabola, 94–­96; focus of, 96
Parthenon, 68
Hadamard, Jacques Solomon, 44 Pascal, Blaise, 138
Hales, Thomas, 61 Peaucellier, Charles-­Nicolas, 128
Haraguchi, Akira, 91 pentagon, 70–­72; construction of, 177–­179
harmonic series, 102–­104 pentagram, 70–­71
harmonograph, 145 perfect numbers, 9, 11
hexagons, 59–­61 pi (π): Archimedes’s method, 88–­90; biblical value, 84–­
Huygens, Christiaan, 111, 116 85; Egyptian value, 84; record number of digits, 91; in-
hyperbola, 98 finite series and products for, 93; van Ceulen’s value,
hyperbolic functions, 110–­111 91; Zu Chongzhi’s value, 91
hypocycloids, 119–­122 Pick, Georg Alexander, 157, 159; theorem, 157–­159
Plato, 56
infinity, 170–­173 Platonic solids, 9, 69
inversion in a circle, 126–­129, 180–­181 Plimpton 322 (Babylonian clay tablet), 20
irrational number, 23–­25 Plücker, Julius, 134, 137
isocevas, 106–­107 point at infinity, 126
Poncelet, Jean Victor, 138, 140
Kepler, Johannes, 61, 64, 98, 122 power set(s), 172
Klein, Felix, 137, 157 prime numbers, 11, 42–­44, 45–­48, 104, 176
Koch, Niels Fabian Helge von, 164; curve, 164–­166 projective geometry, 134, 140, 170
Ptolemy (Claudius Ptolemaeus), 122
lattice polygon, 157 Pythagoras of Samos, 13
Leibniz, Gottfried Wilhelm von, 107, 111, 116 Pythagoreans, the, 11–­12, 23, 25
L’Hospital, Marquis de, 116 Pythagorean theorem, 13–­15, 16–­19
Lindemann, Carl Louis Ferdinand von, 81, 87 Pythagorean triples, 20–­22, 176
line equations, 134
Lipkin, Yom Tov Lipman, 128 quadrature problem. See squaring the circle
Lissajous, Jules Antoine, 143; figures, 143–­145 quadrilaterals, 6, 175
logarithmic spiral, 112–­115
Loomis, Elisha Scott, The Pythagorean Proposition, 13 rectification, 118
repunits, 53, 55
Magritte, René, Ceci n’est pas une pipe (“this is not a Reuleaux, Franz, 154; triangle, 154–­156
pipe”), 25 Rhind Papyrus, 84
Mandelbrot, Benoit, 166n1
Mascheroni, Lorenzo, 58 sangaku (Japanese wooden tablet), 132–­133
means, 26–­28, 29–­31 Schopenhauer, Arthur, 16
 I n d e x    1 8 7

seventeen-­sided regular polygon, 73–­76 tessellations, 59

Sierpinski, Waclaw Franciszek, 167; triangle, 167–­169 Thales of Miletus, 1; theorem of, 1–­2
snowflake curve. See Koch, Niels Fabian Helge von: Torricelli, Evangelista, 118
curve transcendental number, 87, 110
square root of two, 23–­25, 176 triangular numbers, 9, 11
squaring the circle, 84–­87 trisection of an angle, 81, 163
Stars and Stripes (American national flag), 78, 80 twin primes, 45, 104
Steiner, Jacob, 130; chain, 130–­133; porism, 130–­133
Steinhaus, Hugo Dyonizi, Mathematical Snapshots, 157 Valleé-­Poussin, Charles de la, 44
straightedge and compass constructions. See Euclidean Viète, François, infinite product of, 93
constructions
sum of two squares, 77, 80n1 Wantzel, Pierre Laurent, 76, 81
symmetry, 146–­148, 149–­153 Watt, Harry James, 156
Wren, Christopher, 118
tautochrone, 116, 118
Terquem, Olry, 140 Zeno of Elea, 49; paradoxes of, 49, 52, 101