OUP CORRECTED PROOF – FINAL, 30/07/20, SPi

The Wonder Book of Geometry

OUP CORRECTED PROOF – FINAL, 30/07/20, SPi
 l s t o ry
a tica
m a them
a


OUP CORRECTED PROOF – FINAL, 30/07/20, SPi

1
Great Clarendon Street, Oxford, OX2 6DP,
United Kingdom
Oxford University Press is a department of the University of Oxford.
It furthers the University’s objective of excellence in research, scholarship,
and education by publishing worldwide. Oxford is a registered trade mark of
Oxford University Press in the UK and in certain other countries
© David Acheson 2020
The moral rights of the author have been asserted
First Edition published in 2020
Impression: 1
All rights reserved. No part of this publication may be reproduced, stored in
a retrieval system, or transmitted, in any form or by any means, without the
prior permission in writing of Oxford University Press, or as expressly permitted
by law, by licence or under terms agreed with the appropriate reprographics
rights organization. Enquiries concerning reproduction outside the scope of the
above should be sent to the Rights Department, Oxford University Press, at the
address above
You must not circulate this work in any other form
and you must impose this same condition on any acquirer
Published in the United States of America by Oxford University Press
198 Madison Avenue, New York, NY 10016, United States of America
British Library Cataloguing in Publication Data
Data available
Library of Congress Control Number: 2020932235
ISBN 978–0–19–884638–3
Printed and bound in Great Britain by
Clays Ltd, Elcograf S.p.A.
Links to third party websites are provided by Oxford in good faith and
for information only. Oxford disclaims any responsibility for the materials
contained in any third party website referenced in this work.
 OUP CORRECTED PROOF – FINAL, 30/07/20, SPi

Contents

1. Introduction 1
2. Getting Started 4
3. Euclid’s Elements 9
Euclid, 1732 12
4. Thales’ Theorem 14
The Mathematical World of Ancient Greece 18
5. Geometry in Action 20
6. Pythagoras’ Theorem 26
7. ‘In Love with Geometry’? 36
371 Proofs of Pythagoras 42
8. ‘Imagine my Exultation, Watson . . .’ 44
9. Congruence and Similarity 50
The Golden Ratio 58
10. Conversely . . . 60
11. Circle Theorems 68
12. Off at a Tangent 73
13. From Tangents to Supersonic Flow 79
Galileo and Thales’ Theorem 84
14. What is π, Exactly? 86
15. The Story of the Ellipse 94
 OUP CORRECTED PROOF – FINAL, 30/07/20, SPi

vi CON T E N T S

16. Geometry by Coordinates 101

Inspector Euclid Investigates... 106
17. Geometry and Calculus 108
18. A Royal Road to Geometry? 114
19. Unexpected Meetings 122
20. Ceva’s Theorem 129
Some Further Slices of Pi 136
21. A Kind of Symmetry 138
22. ‘Pyracy’ in Woolwich? 145
23. Fermat’s Problem 154
24. A Soap Solution 164
25. Geometry in The Ladies’ Diary 171
Euclid, 1847 178
26. What Euclid Did 180
27. Euclid on Parallel Lines 189
Proof by Picture? 196
28. ‘A New Theory of Parallels’? 198
29. Anti-Euclid? 205
30. When Geometry Goes Wrong . . . 213
31. New Angles on Geometry 223
32. And Finally . . . 231
 OUP CORRECTED PROOF – FINAL, 30/07/20, SPi

CON T E N T S vii

Notes 241
Further Reading 265
Acknowledgements 269
Publisher’s Acknowledgements 270
Picture Credits 271
Index 273
OUP CORRECTED PROOF – FINAL, 30/07/20, SPi
 OUP CORRECTED PROOF – FINAL, 30/07/20, SPi

Introduction

It all started at school, one cold winter morning in 1956,

when I was ten.
Mr. Harding had been doing some maths at the blackboard,
with chalk dust raining down everywhere, when he suddenly
whirled round and told us all to draw a semicircle, with
diameter AB.
Then we had to choose some point P on the semicircle, join
it to A and B by straight lines, and measure the angle at P (Fig. 1).

Fig. 1 Thales’ theorem.

 OUP CORRECTED PROOF – FINAL, 30/07/20, SPi

2 I N T RODUC T ION

I duly got on with all this, casually assuming that the angle
at P would depend on where P is, exactly, on the semicircle.
But it doesn’t.
It’s always 90°.
* * *
At the time, I had no idea that mathematics is full of surprises
like this.
I had no idea, either, that this is one of the first great theorems
of geometry, due to a mathematician called Thales, in ancient
Greece. And according to Thales – so it is said – the key question
is always not ‘What do we know?’ but rather ‘How do we know it?’
Why is it, then, that the angle in a semicircle is always 90°?
The short answer is that we can prove it, by a sequence of
simple logical steps, from a few apparently obvious starting
assumptions.

Fig. 2 The importance of proof.

 OUP CORRECTED PROOF – FINAL, 30/07/20, SPi

I N T RODUC T ION 3

And by doing just that, in the next few pages, I hope to not
only lay some foundations for geometry, but do something
far more ambitious.
For, with geometry, it is possible to see something of the
whole nature and spirit of mathematics at its best, at almost
any age, within just half an hour of starting.
And in case you don’t quite believe me…
 OUP CORRECTED PROOF – FIRST PROOF, 30/07/20, SPi

Getting Started

The first really major idea is that of parallel lines.

These are lines, in the same plane, which never meet, no
matter how far they are extended.
And I will make two assumptions about them.

Parallel lines
Imagine, if you will, two lines crossed by a third line, produ-
cing the so-called corresponding angles of Fig. 3.

Fig. 3 Corresponding angles.

 OUP CORRECTED PROOF – FIRST PROOF, 30/07/20, SPi

GE T T I NG S TA RT E D 5

Then, throughout most of this book, I will assume that

(1) If two lines are parallel, the corresponding angles are
equal.
(2) If corresponding angles are equal, the two lines are
parallel.
These assumptions are rooted in the intuitive notion that
parallel lines must be, so to speak, ‘in the same direction’, but
however obvious (1) and (2) may seem, they are assumptions.
And, even at this early stage, it is worth noting that they
amount to two very different statements.
In effect, (1) helps us use parallel lines, while (2) helps us
show that we have some.

Angles
We will measure angles in degrees, denoted by °, and the two
parts of a straight line through some point P form an angle of
180° (Fig. 4).

180˚

Fig. 4 A straight line.

A right angle is half this, i.e. an angle of 90°, and the two
 OUP CORRECTED PROOF – FIRST PROOF, 30/07/20, SPi

6 GE T T I NG S TA RT E D

Fig. 5 Right angles.

lines forming it are then said to be perpendicular (Fig. 5).

Opposite angles
When two straight lines intersect, the so-called opposite angles
are equal (Fig. 6).

Fig. 6 Opposite angles.

Alternate angles
If two lines are parallel, and crossed by a third line, then the
so-called alternate angles are equal (Fig. 7).

Fig. 7 Alternate angles.

 OUP CORRECTED PROOF – FIRST PROOF, 30/07/20, SPi

GE T T I NG S TA RT E D 7

b
c

Fig. 8 Proof that alternate angles are equal.

This is because, in Fig. 8, a = b (corresponding angles) and

b = c (opposite angles). So a = c.
The argument works ‘in reverse’, too, so that if alternate
angles are equal, the two lines must be parallel.
And with these ideas in place, we are now ready to prove
the first theorem which, in my view, is not obvious at all . . .

The angle-sum of a triangle

The three angles in any triangle add up to 180° (Fig. 9).

Fig. 9 Angles in a triangle.

 OUP CORRECTED PROOF – FIRST PROOF, 30/07/20, SPi

8 GE T T I NG S TA RT E D

To prove this, draw a straight line through one corner,

parallel to the opposite side (Fig. 10).

a b
c

a b

Fig. 10 Proof of the angle-sum of a triangle.

The angles a are then equal (alternate angles).

The angles b are also equal, for the same reason.
Finally, the new line is straight, so a + b + c = 180°, which
completes the proof.
 OUP CORRECTED PROOF – FINAL, 30/07/20, SPi

Euclid’s Elements

The most famous example of geometry being presented in this

concise, deductive, and carefully ordered way is the Elements,
written by Euclid of Alexandria (Fig. 11), in about 300 BC.

Fig. 11 Euclid.
 OUP CORRECTED PROOF – FINAL, 30/07/20, SPi

10 E UC L I D’S E L E M E N T S

It is best to be clear from the outset, I think, that the precise

theorems and proofs of Euclid’s Elements (Fig. 12) are essen-
tially about imaginary objects.

Fig. 12 The oldest surviving copy of Euclid’s Elements, MS

D’Orville 301, copied by Stephen the Clerk for Arethas of Patras,
in Constantinople in AD 888.

A Euclidean straight line, for instance, isn’t just ‘perfectly’

straight—it has zero thickness. So even if I could draw one
properly, you wouldn’t be able to see it.
And a point isn’t a blob of small dimension—it has no
dimension at all. Or, as Euclid put it:

A point is that which has no part.

It should be said, too, that Euclid makes no use of what we

would call ‘measurement units’ for length. And there are no
 OUP CORRECTED PROOF – FINAL, 30/07/20, SPi

E UC L I D’S E L E M E N T S 11

degrees in Euclid; the nearest he comes to having a unit for

angle is the concept of right angle, which he uses a great deal
(Fig. 13).

Fig. 13 Proof that opposite angles are equal, from a 1732 edition
of Euclid’s Elements.

In spite of this, and the austere style of exposition, the

Elements has had more influence, and more editions, than
almost any other book in human history.
In the end, however, there can be no single ‘best’ way of
doing geometry, and we all have to find our own path into the
subject.
And if, in this book, I unashamedly assume more than
Euclid does, it is because I want to proceed more quickly to
interesting and surprising results . . .
Euclid, 1732

One of the most popular early editions

of Euclid was by Isaac Barrow. It was
first published in 1660 , but my own copy
dates from 1732 .
 OUP CORRECTED PROOF – FINAL, 30/07/20, SPi

Thales’ Theorem

Thales’ theorem says that the angle in a semicircle is always 90º.

And, to prove it, we need just one or two more key ideas.

Congruent triangles
Congruent triangles are ones which have exactly the same size
and shape.
And the most obvious way of fixing the exact size and
shape of a triangle is, perhaps, to specify the lengths of two
sides and the angle between them.
This leads to a very simple test for congruence, known
informally as ‘side-angle-side’, or SAS (Fig. 14).

Fig. 14 Congruence by SAS.

Isosceles triangles
An isosceles triangle is one in which two sides are equal.
 OUP CORRECTED PROOF – FINAL, 30/07/20, SPi

T H A L E S’ T H EOR E M 15

Triangles of this kind play a major part in geometry, largely

because the ‘base’ angles of an isosceles triangle are equal (Fig. 15).

A B Fig. 15 An isosceles triangle.

Many people, I think, find this particular result rather obvi-

ous. After all, if we ‘nip round the back’ of an isosceles tri-
angle it will look exactly the same.
A more formal way of proving the result is to introduce the
line CD bisecting the angle at C (Fig. 16).

Fig. 16 Proof that the base angles of an

A D B isosceles triangle are equal.

The triangles ACD and BCD are then congruent by SAS,

and  one is, in fact, a ‘mirror image’ or ‘overturned’ version
 OUP CORRECTED PROOF – FINAL, 30/07/20, SPi

16 T H A L E S’ T H EOR E M

of the other. In particular, then, the angles at A and B must

be equal.
(If all three sides of a triangle happen to be equal it is said to
be equilateral. The triangle is then isosceles in three different
ways, so all three of its angles are equal.)

Circles
The defining property of a circle is that all its points are the
same distance from one particular point, called the centre, O.
Some other common terminology is introduced in Fig. 17.

circumference
us
di
ra

O diameter
chord

Fig. 17 The circle.

And this gives us all we need to prove Thales’ theorem.

Thales’ theorem
We want to prove that if P is any point on the semicircle in
Fig.  18, then ÐAPB = 90°, where ÐAPB denotes the angle
between AP and PB.
Now, the simplest way of using the fact that P lies on the
semicircle, surely, is to draw in the line OP and observe that
 OUP CORRECTED PROOF – FINAL, 30/07/20, SPi

T H A L E S’ T H EOR E M 17

Fig. 18 Proof of Thales’ theorem.

OP = OA = OB, because all points on a circle are the same

distance from its centre.
Suddenly, then, we have two isosceles triangles, AOP and BOP.
The two ‘base angles’ a are therefore equal, and so are the
two base angles b.
Finally, the three angles of the large triangle APB must add
up to 180°, so
a + ( a + b ) + b = 180°
and therefore a + b = 90°. In consequence, ÐAPB = 90°,
which proves the theorem.
And in all the years since I first saw this proof, on a cold
winter morning in 1956, I have never forgotten it.
After all, the result is, at first sight, rather difficult to believe,
yet just a few minutes later we find ourselves saying, argu-
ably: ‘Oh, it’s sort of obvious, really, isn’t it—when you look at it
the right way.’
And in my experience, at least, this is often one of the hall-
marks of mathematics at its best.
 a t i c a l w orld of
the m athem E EC E
N T G R
ANCIE

Rome

the greek empire

Athens Miletus
Crotona Samos

Syracuse

Mediterranean Sea
Alexandria

Thales lived in Miletus.

Pythagoras came from the island of Samos,
but later moved to Crotona.

Thales
Pythagoras
Euclid
Archimedes
B.C. −500 −400 −300 −200 −100 0
Plato’s Academy in Athens had this famous
inscription over its entrance:

ΑΓΕΩΜΕΤΡΗΤΟΣ
ΜΗΔΕΙΣ ΕΙΣΙΤΩ
“Let no one ignorant of geometry enter here”

The Pharos Lightouse, Alexandria

Euclid wrote The Elements in Alexandria.

Archimedes lived and worked in Syracuse.
 OUP CORRECTED PROOF – FINAL, 31/07/20, SPi

Geometry in Action

Throughout history there have been practical applications

of geometry, and one of the earliest was Thales’ attempt to
calculate the height of the Great Pyramid in Egypt.

Thales and similar triangles

Thales measured the shadow of the Great Pyramid cast by the
Sun, and by adding half the pyramid’s base determined the
distance L in Fig. 19.

Fig. 19 Similar triangles.

 OUP CORRECTED PROOF – FINAL, 31/07/20, SPi

geom e t ry i n ac t ion 21

He then measured the shadow ℓ cast by a vertical pole of

height h.
Assuming the Sun’s rays to be parallel, he reasoned that
the two triangles in Fig.  19, though of very different size,
would have exactly the same shape, and that corresponding
sides would therefore be in the same proportion.
In particular, then, he reasoned that
h "
= ,
H L

and so, having measured the other three lengths, he was able
to determine the Great Pyramid’s height H.

Fig. 20 Thales, on a Greek postage stamp of 1994.

Today we use the term similar to describe triangles which

have exactly the same shape, and, as we will see later, they
play a major part in some of the most striking theorems of
geometry.
 OUP CORRECTED PROOF – FINAL, 31/07/20, SPi

22 geom e t ry i n ac t ion

Measuring the Earth

According to its Greek roots, the word ‘geometry’ means,
quite literally, ‘Earth measurement’. So it seems appropriate
to look next at a famous attempt to measure the circumfer-
ence of the Earth, by Eratosthenes of Alexandria, in about
240 bc.
And, as it happens, he too used the Sun’s rays, but in a
rather different way.
Eratosthenes knew that, at noon on the longest day of the
year, the Sun was directly overhead at his birthplace Syene
(modern-day Aswan), because it illuminated the bottom of a
deep well there.
He also knew that, at the same time, the Sun made an angle
of 7.2° with the vertical at Alexandria, which he took to be
5000 stades due north of Syene.

Alexandria
Sun
7.2°
O
Syene

Fig. 21 Measuring the Earth.

Eratosthenes assumed that the Sun was so far from the

Earth that its rays arrived parallel. The two shaded angles in
 OUP CORRECTED PROOF – FINAL, 31/07/20, SPi

geom e t ry i n ac t ion 23

Fig. 21 are then corresponding angles, so the angle at O, the

centre of the Earth, must also be 7.2°.
Now, 7.2° is one-fiftieth of 360°, so he reasoned that the
circumference of the Earth must be 50 times the distance
between Alexandria and Syene, i.e. 250,000 stades.
In truth, this is probably an oversimplification of what
Eratosthenes actually did. Moreover, what a stade was, as a
unit of distance, is also lost; estimates by subsequent scholars
put it between 0.15 and 0.2 km, leading to a result for
the Earth’s circumference somewhere between 37,500 and
50,000 km. (The actual value is about 40,000 km.)

‘Practical work’, 1929

There is, of course, another, quite different aspect to practical
geometry, namely the actual construction of geometrical fig-
ures using ruler, compasses, and other tools of the trade.
When doing this kind of work, however, we have to be
continually mindful of what physicists would call ‘experi-
mental error’; otherwise, things can get a bit ridiculous.
On my shelves at home, for instance, there is an old geom-
etry exercise book, dating from 1929, that once belonged to a
pupil at a primary school in the north of England.
The book itself has considerable charm, and consists
mostly of simplified Euclid, carried out neatly and well. On
the very last page, however, and without any warning at
all, we suddenly meet something called ‘Practical Work’,
involving—apparently—some actual measurement (Fig. 22).
 OUP CORRECTED PROOF – FINAL, 31/07/20, SPi

24 geom e t ry i n ac t ion

Fig. 22 From a 1929 school exercise book.

Yet, despite the tick of approval from the teacher, there

is something faintly absurd about this particular piece of
work.
So far as I can determine, the angle A is closer to 45° than
50°, and it looks to me as if the various numbers have simply
been cooked up, quite unashamedly, so that the angle-sum
comes out ‘right’.

Area
Perhaps the oldest geometrical idea of real practical import-
ance is that of area, driven largely by problems concerning
land.
 OUP CORRECTED PROOF – FINAL, 31/07/20, SPi

geom e t ry i n ac t ion 25

We begin with a square of side 1 unit, and it quickly becomes

evident how to calculate the area of a rectangle with sides
which are whole numbers (Fig. 23).

4
1
1 3
A=1

A = 4×3
a

b
b

A = ab a
1
A =  ab
2
Fig. 23 Area.

This leads us to define the area of a rectangle, more gener-

ally, as
A = ab,

where the side lengths a and b may now be fractional or even

irrational.
Introducing a diagonal then bisects the rectangle itself, giv-
ing 1 ab as the area of a right-angled triangle.
2
And, improbable as it may seem, these elementary ideas of
area are enough to let us take a first look at one of the most
famous—and far-reaching—theorems of all . . .
 OUP CORRECTED PROOF – FINAL, 30/07/20, SPi

Pythagoras’ Theorem

There is a surprisingly simple relationship between the

lengths of the sides of any right-angled triangle (Fig. 24).
And, like so much that is best in mathematics, it is this
generality that gives the theorem its power.

Fig. 24 Pythagoras’ theorem.

In Fig. 24, c denotes the length of the hypotenuse—meaning

the side opposite the right angle—while a and b denote the
lengths of the other two sides.
 OUP CORRECTED PROOF – FINAL, 30/07/20, SPi

PY T H AG OR A S ’ T H EOR E M 27

A special case
If the two shorter sides happen to be of length 3 and 4, then
a 2 + b2 = 9 + 16 = 25, so c must be 5.
And the idea of a right-angled triangle with sides of length
3, 4, and 5 was known to Babylonian mathematicians (in
what is now Iraq) over a thousand years before Pythagoras.
One ancient clay tablet, for instance, has a geometrical prob-
lem on it equivalent to the following:

A ladder of length 5 units stands upright, flat against a wall.

The upper end slips down a distance 1 unit. How far does the
lower end slide out?

Fig. 25 The 3-4-5 special case of Pythagoras’s theorem, from

John Babington’s Treatise of Geometrie (1635).
 OUP CORRECTED PROOF – FINAL, 30/07/20, SPi

28 PY T H AG OR A S ’ T H EOR E M

The 3-4-5 right-angled triangle is so well known, in fact,

that it is sometimes confused with Pythagoras’ theorem
itself. Yet, as I have emphasized, it is only one very special
case (Fig. 25).
And, in many ways, it isn’t the most important special case
at all . . .

Unexpectedly irrational
If the two shorter sides of a right-angled triangle are equal,
then, according to Pythagoras’ theorem, the three sides are in
proportion 1 : 1 : 2 (Fig. 26a).
This was, again, known long before Pythagoras, and a
famous Babylonian clay tablet, known as YBC 7289, dating
from roughly 1700 BC, shows a square with two diagonals,
and various numbers in the mathematical notation of the
time (Fig. 26b).

45˚

√2
1

45˚
1
(a) (b)
Fig. 26 Another special case of Pythagoras’ theorem.
 OUP CORRECTED PROOF – FINAL, 30/07/20, SPi

PY T H AG OR A S ’ T H EOR E M 29

And one of those numbers, representing the ratio of diag-

onal to side, is, in modern decimal notation:

1× 4142128,

which is within 1 part in a million of 2.

The importance of this particular example stems, in part,
from the Pythagorean discovery that 2 is irrational, so that
the ratio of diagonal to side for a square cannot be written
exactly as the ratio of two whole numbers.

Fig. 27 One of the few things known with some certainty

about Pythagoras is that he investigated the connection
between mathematics and music. This woodcut is from
Franchino Gaffurio’s Theorica Musicae (1492).
 OUP CORRECTED PROOF – FINAL, 30/07/20, SPi

30 PY T H AG OR A S ’ T H EOR E M

To put it another way, it is impossible to find a unit of

length, however small, such that the side and diagonal of a
square are both whole numbers.
This was a monumental shock for the Pythagorean world
view, and even had a profound effect, centuries later, on the
structure of Euclid’s Elements.

Three proofs of Pythagoras

So far, we have presented Pythagoras’ theorem as an unex-
pectedly simple relationship between three lengths.
Evidently, however, a 2 + b2 = c 2 also represents an extraor-
dinary relationship between the areas of the three squares
that one could, if one wished, construct on the sides of the
triangle (Fig. 28).

2
c

b2

a2

Fig. 28 Another view of Pythagoras’ theorem.

 OUP CORRECTED PROOF – FINAL, 30/07/20, SPi

PY T H AG OR A S ’ T H EOR E M 31

And this way of looking at things forms the basis, in fact,

of many of the proofs.

A ‘proof by picture’

Arrange four copies of the original right-angled triangle as in

Fig. 29a. This produces a square with area c2 in the middle.
Now think of the triangles as white tiles on a dark floor,
and move three of them so that they occupy the new posi-
tions indicated in Fig. 29b.

b2

c2
a2

(a) (b)
Fig. 29 The simplest proof of Pythagoras’
theorem?

The floor area not occupied by triangles is now a 2 + b2 , yet

must evidently be the same as before.
So a 2 + b2 = c 2 .
 OUP CORRECTED PROOF – FINAL, 30/07/20, SPi

32 PY T H AG OR A S ’ T H EOR E M

‘Plain and Easie’

This proof starts with the same diagram, but uses a little
algebra instead.

a b

b c
c a

c2
a c
c b

b a
Fig. 30 An algebraic proof.

The area of the large square in Fig. 30 is

( a + b )2 = a 2 + 2ab + b2 .

But the large square is made up of the small square (area c2)
1
and four triangles of area ab each. So the area of the large
2
square is also c 2 + 2ab , and therefore

a 2 + b2 = c 2 .
 OUP CORRECTED PROOF – FINAL, 30/07/20, SPi

PY T H AG OR A S ’ T H EOR E M 33

The earliest clear exposition I have found of this particular

proof is in John Ward’s ‘plain and easie’ Young Mathematician’s
Guide of 1707, which was one of the most popular and best-
selling mathematics books of its time (Fig. 31).

Fig. 31 From the Young Mathematician’s Guide of 1707.

 OUP CORRECTED PROOF – FINAL, 30/07/20, SPi

34 PY T H AG OR A S ’ T H EOR E M

Pythagoras in China?

A similar—but nonetheless different—argument makes use

of Fig. 32.
Again there are four copies of the original triangle, each of
1
area ab, but they are placed inside a square of area c2 so as to
2
leave a small square of area (a − b)2 in the middle.

c
b
a
c

(a–b)2

Fig. 32 Another algebraic proof.

So ( a - b ) + 2ab = c , and as ( a - b ) = a - 2ab + b this

2 2 2 2 2

reduces, yet again, to

a 2 + b2 = c 2 .

And the figure used in this proof can, in fact, be clearly

seen in one of the oldest and most famous Chinese mathem-
atical texts, the Zhou bi suan jing (Fig. 33).
 OUP CORRECTED PROOF – FINAL, 30/07/20, SPi

PY T H AG OR A S ’ T H EOR E M 35

Fig. 33 From a Ming dynasty copy of the Zhou

bi, printed in 1603.

In consequence, this has sometimes been advanced as one

of the earliest ‘proofs’ of Pythagoras’ Theorem. As I under-
stand it, however, the accompanying text is extremely diffi-
cult to translate and interpret, and expert scholars on the
history of Chinese mathematics are still not agreed on what it
really means (see Notes, p. 241).
In any event, as it happens, the proof in Euclid’s Elements is
quite different from anything we have seen so far.
And, some would say, a bit scary . . .
 OUP CORRECTED PROOF – FINAL, 30/07/20, SPi

‘In Love with Geometry’?

In John Aubrey’s Brief Lives there is a famous passage concern-

ing the seventeenth-century philosopher Thomas Hobbes:

He was 40 yeares old before he looked on Geometry; which

happened accidentally. Being in a Gentleman’s Library, Euclid’s
Elements lay open, and ’twas the 47 El. Libri I. He read the
Proposition. By G–, sayd he (he would now and then sweare
an emphaticall Oath by way of emphasis) this is impossible!

Fig. 34 Euclid’s Proof of Pythagoras’ theorem.

 OUP CORRECTED PROOF – FINAL, 30/07/20, SPi

‘I N LOV E W I T H GEOM E T RY’? 37

Now, Elements I.47 is none other than Euclid’s statement

and proof of Pythagoras’ theorem (Fig. 34), and many people
find Euclid’s diagram rather forbidding. (It certainly terrified
me, aged 10, in 1956.)
But Hobbes reacted rather differently, and though he found
the theorem itself almost unbelievable, he persisted with the
proof, step by step, until at last he was convinced.
And, according to Aubrey, this made him

. . . in love with Geometry.

In any event, Euclid’s proof is certainly well worth pursu-

ing, and if we are to do this we need, first, to take a closer look
at the whole idea of area.

The area of a triangle

The general formula for the area of a triangle is shown in Fig. 35.

Fig. 35 The area of a triangle.

And we can prove it quite easily by letting the height divide

the original triangle into two right–angled ones. As we saw in
 OUP CORRECTED PROOF – FINAL, 30/07/20, SPi

38 ‘I N LOV E W I T H GEOM E T RY’?

Chapter 5, each one has area equal to half the height times its
base, so by adding the two the result follows.
And if the original arrangement happens to look like
Fig. 36b, with an obtuse-angled triangle, the result still follows;

C C

h h

A B A B
(a) (b)
Fig. 36 Same area!

we just end up subtracting the areas of two right-angled

triangles, instead of adding them.
As a matter of fact, because the two triangles in Fig.  36
have the same height h and the same base AB they will have
the same area, and this will remain true no matter how far we
move C to the right, provided only that we move it always paral-
lel to AB, so that h stays the same.
And this was apparently one of the very first results in
geometry that impressed the young Isaac Newton, while still
an undergraduate at Cambridge.
For, according to a contemporary, he started reading
Euclid’s Elements in 1663, but found the early propositions so
obvious that

he wondered how anybody would amuse themselves to

write any demonstrations of them.
 OUP CORRECTED PROOF – FINAL, 30/07/20, SPi

‘I N LOV E W I T H GEOM E T RY’? 39

Yet, like most of us, perhaps, he didn’t find the result in

Fig. 36 intuitively obvious at all.
And, interestingly, this particular result was eventually to
play a major part in his researches on planetary motion, as
we will see later in the book.
For the time being, however, it is just what we need to
tackle Euclid’s very distinctive proof of Pythagoras’ theorem.

Euclid’s proof of Pythagoras’ theorem

In Fig. 37, we begin by drawing the three squares on the sides
of the right-angled triangle, and then draw the line CDE per-
pendicular to the hypotenuse AB.

D
A B

Fig. 37 The idea behind Euclid’s proof.

 OUP CORRECTED PROOF – FINAL, 30/07/20, SPi

40 ‘I N LOV E W I T H GEOM E T RY’?

The basic idea is to prove that the areas of the light grey
square and light grey rectangle are equal.
By the same argument, the same will be true of the dark
grey square/rectangle, too.
In this way, then, the sum of the areas of the two small
squares will be equal to the area of the large square on the
hypotenuse, and the theorem will be proved.

Start with half

A B the light grey square.

Slide the top of the triangle

from C to A, keeping the base
A (shown) the same.

Rotate the triangle

through 90˚

With the new base (shown),

D slide the top of the triangle
from C to D.

Fig. 38 The heart of Euclid’s proof. (See also pp. 12, 36, 179.)
 OUP CORRECTED PROOF – FINAL, 30/07/20, SPi

‘I N LOV E W I T H GEOM E T RY’? 41

The key, therefore, is to prove the claim about the light

grey square and rectangle.
And to do this, we focus essentially on half the light grey
square (Fig. 38).
The beauty of this argument is that each of the three steps
preserves the light grey area, so that the area of the light grey
triangle at the end is equal to the area of the light grey triangle
at the start. Doubling both then gives the result.
And the only slightly informal step—rotating the triangle
through 90° and claiming that it ends up in the position
indicated—can be made more formal, if desired, by observing
that the two triangles in question are congruent by SAS, the
angle between the sides in each case being 90°+ ÐABC .
 F PY T H AG O R AS
37 1 P R O O F S O

In 1927, Elisha Scott Loomis, of Cleveland,

Ohio, published a book containing 230
proofs of Pythagoras’ theorem . . .
By the 2nd edition (1940)
the number had risen to
371, including one by a
16-year old schoolgirl
from Indiana . . .

FFrom the Indianapolis Star,

28 Oct 1938

Ann Condit, at Wellesley

(see Notes, p. 241)
College in 1944
 OUP CORRECTED PROOF – FINAL, 30/07/20, SPi

‘Imagine my Exultation,
Watson . . .’

Sherlock Holmes is forever pursuing some master criminal,

but not often, I think, with the aid of geometry.

Tree
Rod

Fig. 39 Sherlock Holmes and similar triangles.

Yet, at a crucial point in ‘The Adventure of the Musgrave

Ritual’, he needs to find the length of the shadow that would
have been cast, at a particular time of day, by a long-vanished
tree of height 64 feet.
So, having waited until the Sun is at the right position in
the sky, he takes a 6-foot fishing rod, measures the length of
its shadow, and scales up the answer by a factor of 64/6.
 OUP CORRECTED PROOF – FINAL, 30/07/20, SPi

‘I M AGI N E M Y E X U LTAT ION , WATSON . . . ’ 45

And when he examines the actual spot where the shadow

of the tree would have ended, he gets rather excited:

‘You can imagine my exultation, Watson, when . . . I saw

a conical depression in the ground. I knew that it was the
mark made by Brunton in his measurements, and that I was
still upon his trail.’

Yet this is, of course, just Thales and the Great Pyramid, all
over again.
It is just similar triangles.

A problem with ladders

Similar triangles also provide the key to a very old problem
indeed, going back to at least AD 850, when it appeared in a
textbook by the Indian mathematician Mahavira.
Two ladders are propped against the walls of an alley, as in
Fig. 40. The heights a and b are given, and we have to deter-
mine the height h at which the ladders cross.
And—surprisingly, perhaps—the answer turns out to be
independent of the width of the alleyway.

1 1 1
— =— +—
b h a b
a
h
c1 c2

Fig. 40 A problem with ladders.

 OUP CORRECTED PROOF – FINAL, 30/07/20, SPi

46 ‘I M AGI N E M Y E X U LTAT ION , WATSON . . . ’

Yet to show this, all we have to do is introduce the distances

c1 and c2, as shown, so that c = c1 + c2 is the width of the alleyway.
Then, by similar triangles,
h c1
=
b c
and h c2
= .
a c

Adding, and dividing by h, then gives the result in Fig. 40.

Pythagoras by similar triangles

Rather more significantly, perhaps, similar triangles provide
some of the most elegant proofs of Pythagoras’ theorem.
In fact, Albert Einstein once recalled inventing such a
proof ‘after much effort’ as a young boy. Sadly, however, we
cannot be sure of the exact details, because there are two
major ways in which such a proof can go.

b
a
I II

c1 c2
c

Fig. 41 Pythagoras’ theorem by similar triangles.

 OUP CORRECTED PROOF – FINAL, 30/07/20, SPi

‘I M AGI N E M Y E X U LTAT ION , WATSON . . . ’ 47

They both start by dropping a perpendicular from the

right angle onto the longest side, or hypotenuse (Fig.  41).
Somewhat remarkably, all three triangles are then similar,
because they all contain the same three angles.

A neat proof

We observe, next, that hypotenuses of these similar triangles

will be in the same proportion as the smaller sides.
From Triangle I and the original, large, triangle, we there-
fore have
a c1
= , so a 2 = cc1 .
c a
And, in the same way, from triangle II and the large triangle:

b c2
= , so b2 = cc 2 .
c b
Adding, and recalling that c1 + c2 = c, then gives the result.

An even neater one?

Areas of similar triangles are in proportion to the squares of

their corresponding sides.
If we focus, then, entirely on hypotenuses in Fig. 41, the area
of triangle I must be a fraction a2/c2 of the area of the large
triangle.
The same argument applied to triangle II produces a frac-
tion b2/c2.
The two fractions must evidently add up to 1, so a2 + b2 = c2.
 OUP CORRECTED PROOF – FINAL, 30/07/20, SPi

48 ‘I M AGI N E M Y E X U LTAT ION , WATSON . . . ’

Similarity and area

So far in this book, we have treated the ideas of similarity and
area quite separately, but they are in fact related.
To see this, draw a rectangle ABCD, and a diagonal AC.
Then pick a point E on the diagonal and draw lines through it
parallel to the sides (Fig. 42).

D C
c a b
—=—
E c d
d
a
A b B

Fig. 42 Area and similarity.

The two right-angled (bold) triangles are similar, because

they share the same three angles. So, according to Thales,
their sides will be in the same proportion.
And however obvious this may seem, it is possible to
deduce it, if we wish, using area.
Note, first, that the two shaded rectangles are bisected by
the diagonal AEC, so that the two light grey triangles have
equal area, and the same is true of the two dark grey triangles.
But the diagonal AEC also bisects the large rectangle, so the
triangles ACD and ACB have equal area.
 OUP CORRECTED PROOF – FINAL, 30/07/20, SPi

‘I M AGI N E M Y E X U LTAT ION , WATSON . . . ’ 49

It then follows at once, by subtraction, that the two (unshaded)

rectangles left over must have equal area. So ad = bc, and dividing
by cd gives the result:
a b
= .
c d

The bigger picture

Ideas of similarity are very powerful, but we have applied
them, so far, only to right-angled triangles.
And our work so far on congruence has a serious draw-
back, too: we have, at present, only one way (SAS) of identify-
ing it.
It is time, then, to take a closer, and more general, look at
congruence and similarity as a whole.
 OUP CORRECTED PROOF – FINAL, 30/07/20, SPi

Congruence and Similarity

The ‘Dam Buster’ raids of 1943 have gone down in history as

one of the most extraordinary episodes of World War II.
But in order for the ‘bouncing bombs’ to skip, repeatedly,
over the water, and reach their target, they had to be dropped
from an aircraft flying at a precise height of just 60 feet.
To achieve that, the ground crew mounted one angled
spotlight in the aeroplane’s nose and another nearer its tail,
so defining a unique triangle with the required height (Fig. 43).

Fig. 43 The dam busters.

 OUP CORRECTED PROOF – FINAL, 30/07/20, SPi

CONGRU E NC E A N D SI M I L A R I T Y 51

And when the two spots of light on the water’s surface

coincided, the bomb was released.
So SAS is not the only way of specifying the exact shape
and size of a triangle, and the method we have just seen is
called—not surprisingly—ASA.
There is in fact a third major method, called SSS, in which
we simply specify the lengths of all three sides. This is, per-
haps, the least ‘obvious’ of the three, but imagine, if you will,
taking a line of given length and drawing a circle of some
given radius about each end (Fig. 44).

Fig. 44 Triangle construction

A B by SSS.

It is evident, I think, that if the two circles meet at all, they

will meet, on any particular side of the original line, at a unique
point P.
We now have three ways, then, of defining the precise shape
and size of a triangle, and these lead directly to three ways of
identifying, and exploiting, congruence.
 OUP CORRECTED PROOF – FINAL, 30/07/20, SPi

52 CONGRU E NC E A N D SI M I L A R I T Y

Congruence
The three major tests for congruent triangles are shown in
Fig. 45.

Fig. 45 Tests for congruence.

This is, perhaps, the place to stress that the angle in ‘SAS’
must be between the two sides, just as the side in ‘ASA’ must be
‘between’ the two angles.
But, having made this point, I would like to move swiftly
on to two examples of the congruence tests in action.

The reflection of light

When light is reflected at a plane mirror, experiment shows

that the incident and reflected rays make equal angles with
the mirror.
 OUP CORRECTED PROOF – FINAL, 30/07/20, SPi

CONGRU E NC E A N D SI M I L A R I T Y 53

B B

A A

Q P O Q P O

B′
B′

(a) (b)
Fig. 46 Finding the shortest path.

And in about AD 100, Heron of Alexandria realized that, in

doing this, light takes—somewhat mysteriously—the shortest
possible path.
To see this, imagine, first, the light travelling from a spe-
cific point A to a specific point B via a general point P on the
mirror (Fig. 46a)
Now draw BO perpendicular to the mirror, and extend it
to B′, where OB′= OB. The triangles POB and POB′ will then
be congruent by SAS (the angle in question being 90°).
So, wherever P is on the mirror,

PB = PB¢ and ÐBPO = ÐB¢PO

Suddenly, then, the problem of picking P to minimize

AP + PB is the same as picking P to minimize AP + PB′—and
we do that, surely, by making APB′ a straight line (Fig. 46b).
 OUP CORRECTED PROOF – FINAL, 30/07/20, SPi

54 CONGRU E NC E A N D SI M I L A R I T Y

And in that particular case, ÐAPQ = ÐB¢PO, because they

are opposite angles. As ÐB¢PO = ÐBPO anyway, it follows
that, in the shortest-path case, the incident and reflected rays
make equal angles with the mirror.

Parallelograms

A parallelogram is defined simply as a four-sided plane figure

in which opposite sides are parallel.

D C

A B

Fig. 47 A parallelogram.

But opposite sides are also equal, and we can prove this
using ASA.
The key step is to draw in a diagonal such as AC in Fig. 47.
The angles • are equal (alternate angles), and the angles ⚬ are
also equal, for the same reason.
The triangles ABC and CDA are then congruent by ASA.
So, in particular, AB = CD and BC = DA, which is what we
were trying to show.

Describing congruence and similarity

Note, incidentally, that in referring to triangles ABC and CDA

just now, I was trying to communicate, through the order of the
 OUP CORRECTED PROOF – FINAL, 30/07/20, SPi

CONGRU E NC E A N D SI M I L A R I T Y 55

lettering, not only the congruence but also which points cor res-
pond to which, and I will try to follow this practice with both
congruence and similarity throughout the book.

Similarity
Similar triangles have exactly the same shape, and therefore the
same three angles.
And, in consequence, their sides are all in the same pro-
portion k, called the scale factor.
Congruence is a special case of similarity, with k = 1, so it is
no surprise, perhaps, that the three main tests for similarity
(Fig. 48) mirror those for congruence very closely.

Fig. 48 Tests for similarity.

 OUP CORRECTED PROOF – FINAL, 30/07/20, SPi

56 CONGRU E NC E A N D SI M I L A R I T Y

And, as with congruence, I would like to offer, straight

away, two examples of the tests in action.

The mid-point theorem

Take any triangle and join the mid-points of two of its sides.
The resulting line is then parallel to the third side (Fig. 49).

P Q

A B

Fig. 49 The mid-point theorem.

To see this, we compare the two triangles PCQ and ACB.

They have the angle C in common, and PC = 1 AC while
2
QC = 1 BC .
2
1
They are therefore similar by SAS, with scale factor k = .
2
So all their angles must correspond, and therefore, in
particular, ÐCPQ = ÐCAB .
So PQ and AB are parallel.

Varignon’s theorem

Take any quadrilateral—i.e. four-sided figure—and join the

mid-points of the sides by straight lines, in order (Fig. 50).
 OUP CORRECTED PROOF – FINAL, 30/07/20, SPi

CONGRU E NC E A N D SI M I L A R I T Y 57

Fig. 50 Varignon’s theorem.

Then the result is always a parallelogram!

And yet, as soon as we draw in one of the diagonals of the
original figure, we see why, because two applications of the
mid-point theorem show that GH and FE are both parallel
to DB, and must therefore be parallel to each other (Fig. 51).
And an equivalent argument, using the other diagonal,
shows that EH is parallel to FG.

C
F
D E

G
B
H Fig. 51 Proof of
A Varignon’s theorem.

Rarely in mathematics, I think, does such an apparently

mysterious result become so ‘obvious’, so quickly, when you
look at it the right way.
 N R AT IO
THE G OLDE
1 + √5
Φ= 2

This is the ratio diagonal/side for

a REGULAR PENTAGON
PROOF Each diagonal is parallel
to one of the sides, so ABDC is
a parallelogram. So DC = 1 and
DE = Φ −1

1 1
Φ
B C
1 1
D
1 1
Φ-1 Φ-1

E 1 F

The triangles ABC and DEF are

similar , so BC/EF = AB/DE, which
leads to

Φ2 = Φ + 1
The positive root of this quadratic is

1 + √5
2
 OUP CORRECTED PROOF – FINAL, 30/07/20, SPi

10

Conversely . . .

In ancient Egypt, there were surveyors of land known as

‘rope-stretchers’, and it is sometimes claimed that they used
the 3–4–5 special case of Pythagoras’ theorem to create right
angles (Fig. 52).
Well, they didn’t.

Fig. 52 Ancient Egypt?

This is because Pythagoras’ theorem says that if a triangle

has a right angle then there is a certain relationship between
the lengths of the three sides.
 OUP CORRECTED PROOF – FINAL, 30/07/20, SPi

CON V E R SE LY . . . 61

But what the rope-stretchers would have needed—if they

did anything like this at all—is the exact opposite.
In short, they would have needed not Pythagoras’ theorem,
but its converse.
And, more generally in mathematics, it is always import-
ant to distinguish carefully between any statement

P implies Q

and its converse

Q implies P,

which, like the original statement itself, may or may not be true.

The converse of Pythagoras’ theorem

So, is the converse of Pythagoras’ theorem true? In particular,
is it true that in a 3–4–5 triangle the angle opposite the long-
est side must necessarily be 90°?
Well, actually it is, and the proof is very simple—almost
laughably so. All we have to do is draw a second triangle with
sides of length 3 and 4, and an angle of 90° between them
(Fig. 53b).

5 3
3

4 4

(a) (b)
Fig. 53 Proving the converse of Pythagoras’ theorem.
 OUP CORRECTED PROOF – FINAL, 30/07/20, SPi

62 CON V E R SE LY . . .

By Pythagoras’ theorem itself, the length of the dashed line

in Fig. 53b will be 5. The two triangles are then congruent by
SSS, so the angle opposite the longest side in Fig. 53a must,
indeed, be 90°.

Proof by contradiction
Proving the converse of some theorem is not usually so
straightforward, and one very helpful device is the idea of
proof by contradiction, or reductio ad absurdum.
The idea is to assume, at the very outset, that what you are
trying to prove is false, and then show that this would lead to
some contradiction or absurdity.

?
Fig. 54 Converse of the isosceles triangles theorem.

Isosceles triangles, for instance, provide a good example.

We saw earlier that if a triangle is isosceles then its base angles
are equal. But what about the converse (Fig. 54)?
It is in fact true, and Euclid proves it in the Elements by
contradiction.
He starts, then, with ÐABC = ÐACB in Fig. 55, but assumes
that AB and AC are not equal. One of them must therefore
be  larger than the other, and he lets AB denote the larger.
Then he introduces the point D such that DB = AC, and draws
in the line CD.
 OUP CORRECTED PROOF – FINAL, 30/07/20, SPi

CON V E R SE LY . . . 63

Fig. 55 A proof by contradiction in Barrow’s edition of Euclid’s

Elements (1732).

The triangles DBC and ACB are then congruent by SAS.

But this is absurd, because triangle DBC is only part of
triangle ACB!
The original assumption that AB and AC are unequal must
therefore be wrong. So they are equal, which proves the con-
verse theorem.

Thales revisited
A few years ago, I was enthusing about Thales’ theorem
(Fig.  56) at some conference, and two members of the

Fig. 56 Thales’ theorem.

 OUP CORRECTED PROOF – FINAL, 30/07/20, SPi

64 CON V E R SE LY . . .

audience came up to me afterwards and claimed that I was

making a bit of a fuss, and that the result was really rather
‘obvious’.
After all, they said, it’s fairly obvious from symmetry that
we can take any rectangle and place it in a circle of suitable
size so that all four corners lie on the circle itself (Fig. 57a).
And it’s also fairly obvious—again from symmetry—that
the diagonals of the rectangle meet at the centre of the circle
(Fig. 57b).
So if we now rub out half the diagram we get Fig. 57c, with
the dotted line as a diameter.
And that’s Thales’ theorem, isn’t it?

(a) (b) (c)

Fig. 57 A proof of . . . what, exactly?

Well . . . no . . . I don’t think it is.

Loosely speaking, with Thales’ theorem, we start with a
diameter and end up with a right angle, not the other way
round.
More precisely, let A and B be two different points on a cir-
cle, and let P be a general point in the plane, not coincident
with either A or B.  And consider, if you will, the following
three statements:
 OUP CORRECTED PROOF – FINAL, 30/07/20, SPi

CON V E R SE LY . . . 65

(a) P is on the circle

(b) AB is a diameter
(c) ÐAPB is a right angle.
Thales’ theorem says that (a) + (b) implies (c).
And, put like this, two possible converses come to mind.
One is that (a) + (c) implies (b), and it seems to me that the
foregoing argument is an informal proof of this.
The other possible converse, and the one I will refer to in
future as the converse, because of its practical value, is (b) + (c)
implies (a).
We will now prove this by a powerful device, namely coup-
ling proof by contradiction with the original theorem itself.

The converse of Thales’ theorem

Draw a circle with AB as diameter, and let P be some point

not on AB.
We want to show that if ÐAPB = 90° , then P lies on the circle.
Begin, then, by supposing that ÐAPB = 90°, but that P
doesn’t lie on the circle. There are then two cases to consider.
Suppose, first, that P lies inside the circle (Fig. 58).
P′

A O B
Fig. 58 Proving the converse of Thales’ theorem.
 OUP CORRECTED PROOF – FINAL, 30/07/20, SPi

66 CON V E R SE LY . . .

We then extend the line AP so that it meets the circle at

some point P′, say, and we join P′B.
Now, ÐAPB = 90° by assumption, and ÐAP¢B = 90° , by
Thales’ theorem itself.
So PB and P′B are both perpendicular to AP′, and must
therefore be parallel.
But they meet, of course, at B, giving a contradiction.
So the assumption that P is inside the circle must be wrong.
The same argument, with a slightly different diagram,
shows that P can’t be outside the circle either.
So P must lie on the circle itself.

An alternative approach

The method we have just introduced is a powerful and general

one, but, as it happens, there is on this occasion a simpler
alternative that avoids ‘contradiction’ altogether.

A O B

Fig. 59 An alternative method.

In Fig. 59, with ÐAPB = 90° , draw OD parallel to BP.

 OUP CORRECTED PROOF – FINAL, 30/07/20, SPi

CON V E R SE LY . . . 67

Triangles AOD and ABP are then similar (by AA), with a
scale factor of 2 (because AB = 2 × AO), so D is the mid-point
of AP.
Draw OP. Triangles ODA and ODP are then congruent by
SAS, so OP = OA, and P therefore lies on the circle.
(Yet another proof of the converse of Thales’ theorem can
be found in Notes, p. 243.)
And, just in case you should wonder, I make no apology
for returning, from time to time in this book, to the very first
theorem in geometry that made a real impression on me,
aged 10.
That’s because, for all its elegance, and element of surprise,
Thales’ theorem is just one special case of something even
more general and far-reaching . . .
 OUP CORRECTED PROOF – FINAL, 30/07/20, SPi

11

Circle Theorems

The main circle theorem, from which so much else follows,

can be loosely stated as:

The angle at the centre is twice the angle at the circumference,

where the angles in question stand on the same arc of the

circle (Fig. 60a).
And it has an immediate and quite extraordinary conse-
quence: all the angles like ÐAPB and ÐAP’B in Fig. 60b are
equal, because they are all half the angle at the centre!

Fig. 60 Circle theorems.

 OUP CORRECTED PROOF – FINAL, 30/07/20, SPi

C I RC L E T H EOR E M S 69

Why, then, is the remarkable theorem in Fig. 60a true?

The main theorem

In Fig. 61, triangle AOP is isosceles, because OA = OP (radii).
The two ‘base angles’ a are therefore equal. Moreover,
ÐAOP = 180° - 2a , so if we extend the straight line PO to Q,
then ÐAOQ = 2a .

ab

O
a 2a 2b b
A Q B
Fig. 61 Proof of the main theorem.

In a similar way, the two angles b are equal, and ÐBOQ = 2b.
Adding then gives the result

ÐAOB = 2ÐAPB.

In truth, there is in fact another case to consider, in which,

for example, P lies so far to the right that AP crosses OB.
Happily, however, the argument goes through almost as
before; we just end up subtracting two angles instead of add-
ing them.
 OUP CORRECTED PROOF – FINAL, 30/07/20, SPi

70 C I RC L E T H EOR E M S

Fig. 62 The main circle theorem, on a Guinea–Bissau postage

stamp.

The theorem in Fig. 60b then follows at once, and this, in

turn, leads to two further remarkable theorems . . .

Four points on a circle

If four points happen to lie on a circle, then opposite angles of
the associated quadrilateral add up to 180° (Fig. 63).
One way of seeing why is to note first that the angles s are
equal, for they stand on the same arc AB. The angles t are also
equal, for a similar reason.
But from triangle ABC we have s + t + ÐABC = 180° . So

ÐADC + ÐABC = 180°.

 OUP CORRECTED PROOF – FINAL, 30/07/20, SPi

C I RC L E T H EOR E M S 71

C
D t s
s

t
A
B

Fig. 63 Four points on a

circle.

Arguably, however, the converse of this result turns out to

be even more important:

If the opposite angles of a quadrilateral add up to 180°, then the four

points concerned lie on a circle.

The standard proof of this is by contradiction, and pro-

ceeds in exactly the same spirit as that for the converse of
Thales’ theorem.

Intersecting chords
If two chords AB, CD meet at a point P, then
AP. BP = CP. DP
To see why, note that in Fig. 64, the angles s are equal, and
the angles t are also equal.
 OUP CORRECTED PROOF – FINAL, 30/07/20, SPi

72 C I RC L E T H EOR E M S

D
B
s s

P
t
t
A
C

Fig. 64 Intersecting chords.

Triangles APD and CPB are therefore similar, by ‘AA’, so

AP/CP = DP/BP, whence the result. And the same argument
works equally well if the two chords happen to meet outside
the circle (Fig. 65).

PA. PB = PC. PD
B

P
C
D

Fig. 65 An external intersection.

 OUP CORRECTED PROOF – FINAL, 30/07/20, SPi

12

Off at a Tangent

‘Am I going to die?’ said Tangent, his mouth full of cake.

Evelyn Waugh, Decline and Fall (1928)

I don’t know why it is, exactly, that non-mathematicians

often find the idea of a tangent somehow amusing.
In Evelyn Waugh’s classic novel for instance, Tangent is the
hapless and dim-witted son of Lord and Lady Circumference,
and gets shot in the foot with a starting pistol on School
Sports Day.
Yet there’s nothing funny, really, about a tangent.
And all we have to do to get one is to pick some point P on
a circle and draw the straight line through P that is at right
angles to the radius OP. All the other points on that line will

Fig. 66 The tangent to a circle.

 OUP CORRECTED PROOF – FINAL, 30/07/20, SPi

74 OF F AT A TA NGE N T

then be further from O than P is, and will therefore lie outside
the circle.
In other words, this tangent line will just touch the circle at
P, whence its name.

The secant–tangent theorem

Figure 67, with a tangent PT, shows a classical theorem dating
back to Euclid.

PA. PB = PT2
A

P
T

Fig. 67 The secant–tangent theorem.

The simplest, but slightly informal, way of seeing why it is

true is, I think, to imagine the ‘secant’ line PAB (so-called
because it cuts the circle) rotating gradually clockwise about
P. The product PA.PB will remain constant in the process, by
Fig. 65, and as we can eventually make both A and B as close
to T as we like, that constant must be PT2.
(A more formal proof can be found in Notes, p. 244.)
And while this theorem may seem a little obscure, it does
in fact have a very practical application . . .
 OUP CORRECTED PROOF – FINAL, 30/07/20, SPi

OF F AT A TA NGE N T 75

Measuring the Earth (again!)

The idea here is to deduce the radius of the Earth, R, by climbing
a mountain of known height h and estimating the distance D
to the horizon.

h D

T
R
R

Fig. 68 Measuring the Earth.

If, in Fig. 68, we draw in the tangent line PT, the secant–tangent

theorem tells us at once that

h( 2 R + h ) = D2 .

In practice, h will be so small in comparison with 2R that

this is sensibly approximated by 2Rh = D2, so that

D2
R» .
2h

And this is essentially what the Persian astronomer and

mathematician Al-Biruni did, in about 1019, obtaining a
value for R of 3939 miles.
 OUP CORRECTED PROOF – FINAL, 30/07/20, SPi

76 OF F AT A TA NGE N T

This was astonishingly accurate for the time, and within

1% of the currently accepted value.

Looking at Euclid
Another nice example of tangents in action is a curious problem
first posed in 1471—in a slightly different way—by the German
astronomer Johann Müller, better known as Regiomontanus.
Imagine, if you will, that you are looking up, with due
reverence, at an enormous statue of Euclid (Fig. 69).

Fig. 69 Looking at Euclid.

Clearly, if you stand too far away, your viewing angle a will
be very small.
But it will also be small if you stand too close, because you
will then be viewing Euclid very obliquely.
So, where should you stand to make the viewing angle a as
large as possible?
 OUP CORRECTED PROOF – FINAL, 30/07/20, SPi

OF F AT A TA NGE N T 77

To find the answer, draw a horizontal line at eye level. Then

draw the (unique) circle through the top and bottom of Euclid
himself that just touches that line (Fig. 70a).

Q Q
O O

P P
eye-level
T line A B
(a) (b)
Fig. 70 What’s the best view?

Viewing from the point of tangency T then gives the largest

possible value of the viewing angle a.
This is because any other point on the eye-level line, such as A
or B in Fig. 70b, will inevitably lie on a bigger circle through P and
Q. And the viewing angles there will consequently be smaller,
because they will both be 1 ÐPOQ, which will, itself, have
2
become smaller as a result of O moving further from P and Q.
So, if Euclid’s head and feet are at heights h1 and h2 above
your eye level, the secant–tangent theorem gives the optimum
viewing distance as h1h2 .

A good try?
As it happens, the same problem arises—in principle—in
rugby, whenever a player has to select a point on the try-line
(Fig. 71), for the viewing angle of the gap between the posts
can be maximized in the same way.
 OUP CORRECTED PROOF – FINAL, 30/07/20, SPi

78 OF F AT A TA NGE N T

touch
-down

try-line

Fig. 71 Another maximization problem?

But there must, I think, be other practical considerations,

because I once had the opportunity to present this strategy to
a former England international rugby player, and he was
deeply unimpressed.
 OUP CORRECTED PROOF – FINAL, 30/07/20, SPi

13

From Tangents to
Supersonic Flow

When an object moves faster than the speed of sound, the

disturbance to the air is confined to the region behind a
v-shaped ‘shock wave’ (Fig. 72).

Fig. 72 Supersonic flow.

And, somewhat surprisingly, it is possible to infer from a

still photograph such as this how fast the object is moving.
 OUP CORRECTED PROOF – FINAL, 30/07/20, SPi

80 F ROM TA NGE N T S T O SU PE R SON IC F LOW

More precisely, we can determine the Mach number

v
M= ,
c
where v is the speed of the object, and c the speed of sound.
To do this, however, we need some tangents.
And a bit of trigonometry . . .

Trigonometry
This branch of the subject begins with the definition of the
sine and cosine of an angle θ, which we do by constructing a
right-angled triangle (Fig. 73).

Fig. 73 Definition of sin θ and cos θ.

It is worth noting at once that these definitions are only
made possible by the whole idea of similar triangles, because
the ratios a/c and b/c depend only on the exact shape of the
right-angled triangle and not on its size.
The quantities sin θ and cos θ are not independent; Pytha-
goras’ theorem tells us that (sin q)2 +(cos q)2 = 1, and a uni-
versal shorthand for this is
sin2 q +cos2 q = 1.
Two values of θ are particularly important, for reasons of
symmetry.
 OUP CORRECTED PROOF – FINAL, 30/07/20, SPi

F ROM TA NGE N T S T O SU PE R SON IC F LOW 81

Fig. 74 A special case.

The first is 45°, which leads to an isosceles right-angled triangle

that we have encountered already (Fig. 74).
The other is 60°, in which case we can determine sin θ and
cos θ by bisecting an equilateral triangle, i.e. one with all three
sides equal (Fig. 75).

Fig. 75 Another special case.

Such a triangle is isosceles in three different ways, so all its

angles are equal, and each one must therefore be 60°. And if
we choose a side length of 2, for convenience, the outcome
is as shown, with the 3 emerging from an application of
Pythagoras’ theorem.
 OUP CORRECTED PROOF – FINAL, 30/07/20, SPi

82 F ROM TA NGE N T S T O SU PE R SON IC F LOW

1 1

sin θ cos θ

0 θ 90° 0 θ 90°
Fig. 76 Graphs of sin θ and cos θ.

Rather more ingenuity (and sheer hard work) is needed to

determine sin θ and cos θ for other values of θ, but Fig.  76
shows the overall results in graphical form.
And yet . . . what has all this got to do with supersonic flow?

Supersonic flow
Recall, first, the idea of Mach number
v
M= ,
c
where v is the speed of the object and c the speed of sound.
Now, as the object passes any particular fixed point P in
space, it generates a sound wave which travels outward at
speed c, so that at time t later that particular acoustic disturbance
is confined to a circle around P of radius ct (Fig. 77).
By that time, the object itself will have travelled a distance
vt from P.
So, if v > c, the object will now be outside the circle in question,
and on drawing tangents from it we find that the angle θ is
such that sin θ = c/v.
 OUP CORRECTED PROOF – FINAL, 30/07/20, SPi

F ROM TA NGE N T S T O SU PE R SON IC F LOW 83

ct

θ vt
object P

Fig. 77 Tangents in action.

In this way, then, the total acoustic disturbance—due to all

such waves—will be confined to a v-shaped wedge of half-
angle θ, where
1
sinq = ,
M
and there will be no disturbance at all outside this region.
So, in the case of Fig. 72, where θ is roughly 45°, we deduce
that the object must be moving with a Mach number M of
roughly 2, or 1·4.
 GALIL EO AND
THAL ES’ THEO REM

In around 1600, Galileo

conducted his famous
experiments on motion
down an inclined plane.

Galileo’s chord theorem: motions down

all inclined planes such as AB, AC etc.
take the same time:
The modern explanation is that the
component of gravity down the plane
is proportional to cos A, so very small if
the plane is almost horizontal.

But, by Thales’ theorem, the length

of the plane is also proportional to cos A,
and the two effects cancel out.
 OUP CORRECTED PROOF – FINAL, 30/07/20, SPi

14

What is π, Exactly?

The circumference of any circle is proportional to its diameter,

and this allows us to define the special number

circumference
p= .
diameter

Fig. 78 Circles and π.

Moreover, the diameter is twice the radius r, so the formula

for the circumference in Fig.  78 follows directly from our
definition of π.
But what about the formula for the area? Why is that true?

The area of a circle

We will use an idea due, essentially, to Archimedes, and inscribe
within the circle a polygon having N equal sides (Fig. 79).
 OUP CORRECTED PROOF – FINAL, 30/07/20, SPi

W H AT IS π, E X AC T LY? 87

h r

A B Fig. 79 An inscribed polygon.

This will consist of N equal triangles, such as OAB, where

O is the centre of the circle.
1
Now, the area of each such triangle will be 2 its height h
times its base AB. The total area of the polygon will be N
times this, i.e. 1 ´ h ´ (AB) ´ N .
2
But (AB) × N is the length of the perimeter of the polygon, so

area of polygon = 1 ´ h ´ (perimeter of polygon).

2

Finally, consider what happens as N gets larger and

larger, so that the polygon has shorter and shorter sides, and
approximates the circle ever more closely (Fig. 80).

Fig. 80 Closer and closer . . .

Plainly, h will get ever closer to the radius of the circle, r,

and the perimeter of the polygon will get ever closer to the
circumference of the circle 2πr.
 OUP CORRECTED PROOF – FINAL, 30/07/20, SPi

88 W H AT IS π, E X AC T LY?

In short, then, by taking N large enough we can make the

1
area of the polygon as close as we like to ´ r ´ 2p r .
2
And that is why the area of the circle itself is πr2.

The volume of a sphere

Using broadly similar ideas, Archimedes proved the result in
Fig. 81 for the volume of a sphere.

r
4
Volume = — πr3
3

Fig. 81 The volume of a sphere.

To use any of these formulae in practice, however, we plainly

need to know something about the actual numerical value of
the constant π.

Approximating π
Archimedes started by taking a circle and then calculating the
perimeters of inscribed and circumscribed hexagons (Fig.  82).

Fig. 82 Approximating π.
 OUP CORRECTED PROOF – FINAL, 30/07/20, SPi

W H AT IS π, E X AC T LY? 89

The circumference of the circle must be greater than the first

of these, but less than the second, and with the help of some
ideas from Chapter 13 this leads to
3 < p < 2 3.
But he then doubled the number of sides, repeatedly, until
with 96-gons he obtained the tighter bounds:
10 1
3 < p <3 ,
71 7
and the upper bound, 22/7, was still being used as a ‘practical’
approximation to π in my early schooldays in the 1950s.

Packing problems
One application of all these ideas is to packing problems, which
can be even more fiendish in mathematics than in real life.

Circle packing

Suppose, for instance, that we want to pack a lot of circles as

tightly as possible on a flat surface.

(a) (b)
Fig. 83 Square packing.
 OUP CORRECTED PROOF – FINAL, 30/07/20, SPi

90 W H AT IS π, E X AC T LY?

One arrangement that comes to mind, surely, is the one in

Fig. 83. If the circles have radius 1, and therefore area π, the
plane is completely covered by square units of area 2 × 2 = 4,
as in Fig. 83b, so
p
square ‘packing fraction’ = ,
4
which is about 78%.
But it is fairly obvious, I think, that we can do better with
the hexagonal arrangement of Fig. 84a, in which each circle is
in direct contact with six other circles, rather than just four.

2 2
√3

1 1
(a) (b)
Fig. 84 Hexagonal packing.

Each horizontal row is exactly as it was before, but the dis-

tance between them has decreased. The gap between the lines
of centres used to be 2, but is now the height of an equilateral
triangle of side 2, which is 3 (Fig. 84b).
So the new packing fraction is the square one multiplied
by 2 / 3 , giving
p
hexagonal packing fraction = ,
2 3
which is about 90%.
 OUP CORRECTED PROOF – FINAL, 30/07/20, SPi

W H AT IS π, E X AC T LY? 91

And while we might find it intuitively obvious that the hex-

agonal packing arrangement is the best of all, this was in fact only
proved in 1890 by the Norwegian mathematician Axel Thue.

Sphere packing

The corresponding problem with spheres is even more difficult.

If we adopt a similar approach to that with circles, we arrive
at the packing arrangement of Fig. 85, with each layer arranged
hexagonally, and nestling in the troughs of the layer immedi-
ately below, as with a stack of oranges at a greengrocer.
And in this case
p
volume packing fraction = ,
3 2
which is about 74%.

Fig. 85 Johannes Kepler and the packing of spheres.

And in 1611, Johannes Kepler—who was apparently think-

ing of cannonballs rather than oranges—famously conjec-
tured that no tighter packing was possible.
Yet the first proof of this conjecture to survive serious critical
scrutiny did not emerge until 1998!
 OUP CORRECTED PROOF – FINAL, 30/07/20, SPi

92 W H AT IS π, E X AC T LY?

The first exact formula for π

We have seen Archimedes’ bounds on π, which imply the
well-known approximation π ≈ 3.14, but in 1593 Francois
Viète obtained the first ever exact expression for the numerical
value of π, in the form of a product involving an infinite number
of terms (Fig. 86).

Fig. 86 The first ever exact formula for π.

To get it, Viète started with an inscribed square (Fig. 87a),

and repeatedly doubled the number of sides—without ever
stopping.
To see the effect of this doubling, suppose that at a certain
stage in the process one side of the polygon, PQ, makes an
angle 2A at the centre (Fig. 87b).

1 O
A

P A/2 Q

R
(a) (b)
Fig. 87 Doubling the number of sides.
 OUP CORRECTED PROOF – FINAL, 30/07/20, SPi

W H AT IS π, E X AC T LY? 93

Then, at the next stage, PQ is replaced by PR + RQ, and the

main circle theorem of Chap. 11 then implies that ÐPQR = 1 A .
2
This means that each time we double the numbers of sides,
the current perimeter of the inscribed polygon gets multi-
plied by a factor
1
.
cos 1 A
2
And A itself will, of course, be reduced by a factor of 2 when
we apply the next stage of the process.
So, if the circle has radius 1, the starting square will have
perimeter 4 2 and the starting value of A will be 45°. The
circumference of the circle itself will be 2π, so by repeatedly
doubling the number of sides we get
1 1 1
2p = 4 2 × × × ¼
45° 45° 45°
cos cos cos
2 4 8
To finish things off we appeal to a useful little result from
trigonometry for the cosine of half a given angle:

q 2 + 2 cos q
cos =
2 2

(see Notes, p. 246). As cos 45° = 1/ 2 , repeated application of

this, and a little rearranging, leads directly to Viète’s extraor-
dinary result.
 OUP CORRECTED PROOF – FINAL, 31/07/20, SPi

15

The Story of the Ellipse

On 24 October 1811, a mathematics student at Oxford called

George Chinnery wrote in his diary

That odious and abominably tedious book called the Conic

Sections . . . . is forced upon us merely because the author of
the treatise is alive and resident in the University.

And I find this rather sad, because a conic section is what

we get by taking a plane slice through a circular cone (Fig. 88).

Ellipse Parabola Hyperbola

Fig. 88 Conic sections.

In particular, if we slice through the cone at a fairly shallow

angle, the result is a closed curve called an ellipse.
 OUP CORRECTED PROOF – FINAL, 31/07/20, SPi

T H E S T ORY OF T H E E L L I PSE 95

And even by 1811, whether George Chinnery knew it or

not, the ellipse had already proved to be one of the most
important curves in the history of science.

The ellipse
There is in fact another, completely different way of making
an ellipse.
Mark out two fixed points, H and I in Fig.  89, and run a
loop of string round them. Then move the point E, while
keeping the string taut; this will trace out an ellipse.

Fig. 89 An ellipse, from van Schooten’s

Exercitationum Mathematicarum (1657).

If the loop of string is long, the ellipse will be almost circu-

lar, but if it is short enough, so that it fits rather tightly about
the two fixed points, the ellipse will be long and thin.
And the two fixed points are called focal points, for good
reason . . .
 OUP CORRECTED PROOF – FINAL, 31/07/20, SPi

96 T H E S T ORY OF T H E E L L I PSE

The reflection property

If we make the perimeter of the ellipse a mirror, and place a
source of light at one focal point, the light will be reflected by
the mirror to the other focal point.
This is because, in Fig. 90, the lines PF and PF′ make equal
angles with the tangent.

Fig. 90 The reflection property of an ellipse. F and

F′ are the focal points.

And to see why this property holds, imagine that we are

moving around the ellipse, towards the fixed point P, keeping
the string taut.
Now, if we suddenly decide to ‘fly off at a tangent’ at P, we
can only do so if we can somehow lengthen the loop of string.
And this means that the shortest path from F to F′ via the
tangent at P must be via the point P itself.
 OUP CORRECTED PROOF – FINAL, 31/07/20, SPi

T H E S T ORY OF T H E E L L I PSE 97

Finally, then, in something of a masterstroke, we can appeal

to Heron’s theorem on the reflection of light (Chapter 9) to tell us
that PF and PF′ must make equal angles with the tangent at P.
Yet, while all this was known in ancient times, no one then
can surely have foreseen the extraordinary way in which the
ellipse would reappear, over 1500 years later . . .

Kepler and planetary motion

In 1609, after a painstaking analysis of the astronomical obser-
vations, Johannes Kepler proposed the following:
1. The orbit of each planet is an ellipse, with the Sun at
one focus.
2. A line drawn from the Sun to a planet sweeps out
equal areas in equal times.
The first law, then, is about the shape of the orbit, and the
second about the variation in speed as a planet goes round its
orbit, moving faster when close to the Sun and slower when
further out.

C
D
B

A
Fig. 91 Kepler’s
equal-area law.
 OUP CORRECTED PROOF – FINAL, 31/07/20, SPi

98 T H E S T ORY OF T H E E L L I PSE

In Fig.  91, for instance, the two shaded regions have the
same area, and—according to Kepler—the planet takes as
long to get from C to D as it does to get from A to B.
These were revolutionary ideas, and, it has to be said, not
readily accepted at the time.
Did the planets really move in this particular way?
And, if so . . . why?

Newton, dynamics, and geometry

In around 1679, Isaac Newton showed that Kepler’s second, area-
sweeping, law could be explained at once if the planet is acted on
by a gravitational force that is directed always towards the Sun.
His whole approach is highly geometrical, and, most inter-
estingly, Newton makes extensive use of the very first theorem in
Euclid that really surprised him, about triangles of quite different
shape having equal area (see p. 38).
To see how he does this, we need to note first that Newton does
not treat a continuously acting force directly, but replaces it by a
succession of short sharp impulses towards S, such as F in Fig. 92.

C
C′

F
B

S A

Fig. 92 Geometry in Newton’s dynamics.

 OUP CORRECTED PROOF – FINAL, 31/07/20, SPi

T H E S T ORY OF T H E E L L I PSE 99

Consider first the motion when there is no such force. The

planet will then—like any other object—move at constant
velocity in a straight line, first from A to B, and then, after an
equal time interval, from B to C, with AB = BC. The triangles SAB
and SBC will therefore have equal area, because they have equal
‘bases’ AB and BC, and the same ‘height’ (the perpendicular dis-
tance from S to AC).
Now consider the effect of the impulse F at B. It produces an
instantaneous change in the velocity, in such a way that the
planet ends up at C′ instead of C, where CC′ is proportional to
the impulse and in the same direction, i.e. parallel to SB.
It follows, then, that the triangles SBC and SBC′ also have
equal area, because they have the same base SB and are
between the same two parallel lines, so have the same height.
So the shaded areas SAB and SBC′ must themselves be equal,
and in this way Kepler’s second law can be explained by assum-
ing a force on each planet directed at all times towards the Sun.

Elliptical motion

It was somewhat later, in 1684, that Newton showed that

Kepler’s first law—elliptical orbit, with the Sun at one focus—
implied that the force must be proportional to 1/ r 2 , where r
denotes distance from the Sun.
This was one of the greatest moments in the history of sci-
ence, and was decisive in Newton’s path towards the theory
of universal gravitation.
His whole approach was, again, highly geometrical, involv-
ing much sophisticated use of theorems from the world of
ancient Greece on conic sections.
 OUP CORRECTED PROOF – FINAL, 31/07/20, SPi

100 T H E S T ORY OF T H E E L L I PSE

Fig. 93 A sketch of orbital motion from Newton’s unpublished

manuscript De Motu corporum in gyrum (1684). S denotes the Sun.

But what looks, at first sight, like pure Euclidean geometry

in Fig. 93, isn’t. At one moment the planet is at P, and a short
time later it is at Q.
And, crucially, right at the end of his analysis, Newton con-
siders what happens as that time interval gets shorter and
shorter, so that Q gets closer and closer to P.
This whole kind of thinking propels our geometry story
along still further, to very different territory indeed.
And in order to see something of this we must first back-
track a little, by considering a much more algebraic approach
to geometry that began to develop in the early seventeenth
century.
 OUP CORRECTED PROOF – FINAL, 31/07/20, SPi

16

Geometry by Coordinates

In the early seventeenth century, mathematics took a completely

new turn, largely at the hands of Viète, Fermat, and Descartes.
The fundamental idea was to use algebra to help solve
problems of geometry, and vice versa.
While the original approaches were rather different, we do
this today by first drawing two perpendicular axes, and then
giving each point a pair of coordinates (x, y).
The result of all this is that a curve can be represented by an
equation, and vice versa.
y
6

4
y = x2
2

x
–3 –2 –1 0 1 2 3
Fig. 94 A parabola.

The equation y = x2, for example, corresponds to a parabola

(Fig.  94), which happens to have some remarkable geomet-
rical properties.
 OUP CORRECTED PROOF – FINAL, 31/07/20, SPi

102 GEOM E T RY BY COOR DI NAT E S

In order to see these, however, we will need to grapple with

the whole idea of the steepness, or slope, of a curve.
And this turns out to be a rather subtle matter, so it makes
sense to tackle first the slope of the simplest ‘curve’ of all: the
straight line.

The slope of a line

To get the slope of a line, all we have to do is take two points
P and Q on the line, calculate the increases in x and y as we
move from P to Q, and divide one by the other (Fig. 95).

y
Q
increase
in y increase in y
P slope =
increase in x
increase
in x

x
Fig. 95 The slope of a straight line.

Notably, it doesn’t matter which two points of the line we

choose—this ratio is always the same, essentially because of
similar triangles, and the fact that the line is straight.
And, fairly evidently, I think, the greater the slope, the
steeper the line.

Perpendicular lines
The whole idea of slope is especially useful if two lines are
perpendicular, because the product of their slopes is then –1.
 OUP CORRECTED PROOF – FINAL, 31/07/20, SPi

GEOM E T RY BY COOR DI NAT E S 103

To see why, imagine turning the line in Fig. 96a anti-clockwise

through 90°, as in Fig. 96b.

b
Q Q′

b a
P
a
P′

(a) (b)
Fig. 96 Perpendicular lines.

The original slope is b/a, but as we move from Q′ to P′ in

Fig. 96b x increases by b but y decreases by a, giving a slope of
–a/b. So the product of the two slopes is –1.
And the converse is also true: if the product of the slopes is
–1, then the two lines are perpendicular.

The distance between two points

If two points have coordinates (x1, y1) and (x2, y2), the distance
between them is

D = ( x2 - x1 )2 + ( y2 - y1 )2 ,
simply because of the right-angled triangle in Fig. 97.
And determining distances by using coordinates in this way is
one of the most common ‘practical’ applications of Pythagoras’
theorem.
 OUP CORRECTED PROOF – FINAL, 31/07/20, SPi

104 GEOM E T RY BY COOR DI NAT E S

(x2, y2)

D y2 – y1

(x1, y1) x2 – x1
Fig. 97 Using Pythagoras’ theorem to find the
distance D between two points.

It is also just what we need to see how Thales’ theorem

emerges from this new, and very different, way of doing
geometry . . .

Thales’ theorem (again!)

Our first task, evidently, is to find the equation of a circle, and
it is simplest, perhaps, to consider a circle of radius 1 centred
on the origin of coordinates (0, 0).
As the distance of each point (x, y) on the circle from (0, 0)
must be 1, it follows at once, from what we have just seen,
that the equation of the circle must be x 2 + y2 = 1.

x2 + y2 = 1 P
(x, y)

A O B
(–1, 0) (1, 0)
Fig. 98 Proving Thales’ theorem.
 OUP CORRECTED PROOF – FINAL, 31/07/20, SPi

GEOM E T RY BY COOR DI NAT E S 105

Moreover, the ends of the diameter AB in Fig. 98 will have

coordinates (–1, 0) and (1, 0). And we want to show that AP
and BP are perpendicular, in accord with Thales’ theorem.
All we have to do, then, is calculate

y-0 y
slope of AP = =
x - ( -1) x + 1

0- y y
slope of PB = =
1- x x -1

y2
and form the product, which is .
( x 2 - 1)

But, as we have seen, x2 + y2 = 1, because P lies on the circle.

So the product of the slopes is –1, and therefore ÐAPB = 90° .
 e c t o r E u clid
Insp e s . .. . . . a break-in at the
inve s t i g a t Gallery of Modern Art

The masterpiece “8 × 8 grid

with 3 lines on it” has been
‘rearranged’ by vandals . . .
Yet nothing has been stolen . . .

In fact, it’s actually BIGGER

than it was before . . .
13 × 5 = 65 !

How could this happen?

(Answer on p.24)
 OUP CORRECTED PROOF – FINAL, 31/07/20, SPi

17

Geometry and Calculus

His name is Mr. Newton; a fellow of our College, & very
young . . . but of an extraordinary genius and proficiency in
these things.
Isaac Barrow, of Trinity College Cambridge,
in a letter of 1669

Calculus is the key to much of modern science and engineer-

ing, largely because it is all about the rates at which things
change.
But this idea is, itself, closely related to a purely geometric
one: the slope of a curve.
Given a curve, then, how can we determine its slope, or
steepness, at some particular point P?
Well, we can presumably get a good approximation to it by
choosing some nearby point Q on the curve, and then calcu-
lating the slope of the straight line PQ (Fig. 99).
But if we really want the slope of the curve at P, it turns out
that we have to do something rather more subtle.
In short, we have to explore what happens to the slope of
PQ as Q gets closer and closer to P.
 OUP CORRECTED PROOF – FINAL, 31/07/20, SPi

GEOM E T RY A N D C A LC U LUS 109

P
y

x
Fig. 99 Finding the slope of a
curve.

The slope of a curve

If we actually carry out this procedure with the curve y = x2,
for example, it turns out that the slope at any point is 2x (see
Fig. 100 and Notes, p. 248).

y
6
y = x2
4
slope = 2x

x
–3 –2 –1 0 1 2 3

Fig. 100 Slope and tangent.

In this particular case, then, the slope increases with x,

which surely makes sense, because the curve y = x2 evidently
‘bends upward’, and so gets steeper as x increases from 0.
 OUP CORRECTED PROOF – FINAL, 31/07/20, SPi

110 GEOM E T RY A N D C A LC U LUS

And one reason why the slope of a curve is important is

that it is also the slope of the tangent to the curve at the point
in question.
The curve y = x2, for instance, is actually a parabola, and
the slope of the tangent is key to proving one extraordinary
property: with a parabolic mirror, parallel rays of light from a
distant source all get reflected to a single focal point F (see
Fig. 101 and Notes, p. 249).

Fig. 101 A parabolic mirror.

Calculus
The whole procedure of obtaining the slope of a curve from
its equation is called differentiation, and this is the first key idea
in calculus.
The subject came fully to life in the second half of the
seventeenth century, largely through the work of Isaac
Newton, in England, and Gottfried Leibniz, in Germany
(Fig. 102).
One major application is to problems where we are trying
to find the maximum or minimum value of some quantity of
 OUP CORRECTED PROOF – FINAL, 31/07/20, SPi

GEOM E T RY A N D C A LC U LUS 111

(a) (b)
Fig. 102 (a) Isaac Newton (1642–1727). (b) Gottfried Leibniz
(1646–1716).

interest. In Fig. 103, for instance, the slope is positive for small

values of x, but negative for larger ones, and one way of deter-
mining the maximum value of y, evidently, would be to calcu-
late the value of x for which the slope is actually zero.

slope = 0
y

+ –

x
Fig. 103 Using slope to find a maximum value.
 OUP CORRECTED PROOF – FINAL, 31/07/20, SPi

112 GEOM E T RY A N D C A LC U LUS

But one of the greatest achievements of both Newton and

Leibniz was to link the whole process of differentiation
with two other, apparently quite different, geometrical
ideas . . .

Area and volume

Finding the area or volume of a region enclosed by curved
boundaries has always been difficult.
Yet, quite independently, Newton and Leibniz discovered
that this can be done, in principle, by ‘undoing’ or ‘reversing’
differentiation—a procedure called integration.

y A

slope = y

x x
(a) (b)
Fig. 104 The fundamental theorem of calculus.

To see something of this, suppose we have a curve, with y

depending on x in some way, as in Fig. 104a. Then the area A
underneath the curve will also depend on x, and if we plot
that against x in Fig. 104b we find something that is, at first
sight, extraordinary: the slope of this new curve, at any par-
ticular value of x, is given simply by the corresponding value
of y from Fig. 104a (see also Notes, p. 250).
 OUP CORRECTED PROOF – FINAL, 31/07/20, SPi

GEOM E T RY A N D C A LC U LUS 113

So, if we know how y depends on x, we can, in principle, find

how the area A depends on x by reversing the differentiation
process.
And similar ideas can be used to find volumes. In 1643, for
instance, Evangelista Torricelli caused something of a stir by
discovering (by other methods) a three-dimensional object
that had infinite extent but finite volume (Fig. 105).

Fig. 105 Torricelli’s trumpet, obtained by rotating the curve

y = 1x , for x ³ 1 , about the x-axis.

When Thomas Hobbes got to hear of this, some 30 years

later, he wrote
to understand this for sense, it is not required that a man
should be a geometrician or a logician, but that he should
be mad.

But Torricelli was right, and calculus can be used to con-

firm it.
More generally, calculus brought about a revolution in
mathematics that would eventually take geometry in wholly
new directions.
Yet, even as all this was happening, the classical Euclidean
geometry of straight lines and circles was still proceeding apace.
And, already, some authors were even trying to ‘popular-
ize’ it . . .
 OUP CORRECTED PROOF – FINAL, 31/07/20, SPi

18

A Royal Road to Geometry?

‘There is no royal road to geometry.’
Euclid’s reply to King Ptolemy I, when asked if there were
some easier way of learning the subject.

Actually, there is a Royal Road to Geometry; it was a book by

Thomas Malton, published in 1774 (Fig. 106).
Malton was a self-taught mathematician who originally
kept an upholsterer’s shop in London. He also gave private
tuition at his house in Poland Street, where he claimed he
could

make any Gentleman (having a Talent for it) a Proficient in

Geometry, in less than half the Time usually spent in it;
making it, at the same time, instructive and entertaining.

In his book, Malton is quite forthright in his criticism of

Euclid’s Elements, which he describes as being

much encumbered with useless Demonstrations.

By this, he appears to mean that there is far too much prov-

ing the obvious, which, in Malton’s view, serves only to
 OUP CORRECTED PROOF – FINAL, 31/07/20, SPi

A ROYA L ROA D T O GEOM E T RY? 115

Fig. 106 A royal road to geometry?

 OUP CORRECTED PROOF – FINAL, 31/07/20, SPi

116 A ROYA L ROA D T O GEOM E T RY?

perplex and embarrass the minds of Youths.

And he adds

I have always found more difficulty in demonstrating, to

another Person, self-evident Propositions, than the most
intricate of others . . .

I will return to Malton’s book later in the chapter, but let us

take a brief look, first, at two even earlier attempts to stream-
line geometry for the wider public.

The pathway to knowledge?

The first geometry book to be published in English was The
Pathway to Knowledge, by Robert Recorde, in 1551.
Recorde is best known, perhaps, as inventor of the ‘equals’
sign =, but he was also, by all accounts, one of the greatest
mathematics teachers of all time.
And I have always found one particular observation in the
Pathway rather haunting:

. . . it is not easie for a man that shall travaile in a straunge

arte, to understand at the beginninge bothe the thing that is
taught and also the juste reason whie it is so . . .

Accordingly, there are plenty of assertions in the Pathway,

but hardly anything by way of proof or deductive argument.
In the case of Pythagoras’ theorem, for instance, Recorde
simply draws a diagram for the 3-4-5 case (Fig. 107) and writes
 OUP CORRECTED PROOF – FINAL, 31/07/20, SPi

A ROYA L ROA D T O GEOM E T RY? 117

Fig. 107 Robert Recorde, and a figure from his Pathway to

Knowledge of 1551.

. . . by the number of the divisions in eche of these squares,

may you perceave . . . that the theoreme is true . . .

But this diagram—sometimes presented as a ‘proof’ even

today—proves nothing, even in the 3-4-5 case, unless we can
show somehow that the 50 little squares in it are all exactly
the same size.
In any event, Recorde clearly took the view that, for the
complete beginner, it was best to focus on the propositions
themselves, and not the proofs.

‘Plain and easie’

A very different point of view comes over strongly in John
Ward’s Young Mathematician’s Guide of 1707 (Fig. 108).
And the author seems to have possessed a certain amount
of modesty, for he describes the Guide as ‘Plain and Easie’ and
 OUP CORRECTED PROOF – FINAL, 31/07/20, SPi

118 A ROYA L ROA D T O GEOM E T RY?

Fig. 108 John Ward’s Young Mathematician’s Guide of 1707.

 OUP CORRECTED PROOF – FINAL, 31/07/20, SPi

A ROYA L ROA D T O GEOM E T RY? 119

wholly intended to Instruct, and not to amuse or Puzzle the

young Learner with hard Words; nor is it my Ambitious
Desire of being thought more Learned or Knowing than
really I am . . .

It is, of course, his treatment of geometry that concerns us

here, and while everything is clearly influenced by Euclid,
Ward frequently goes his own way.
In particular, his proof of the angle-sum of a triangle is dif-
ferent from both ours and Euclid’s, and appeals to two cor-
responding angles and one opposite one (Fig. 109).

Fig. 109 Proving the angle-sum of a triangle.

But the really distinctive aspect of Ward’s treatment of

geometry is that as soon as he has the preliminaries out of the
way, he goes off like a rocket.
For, in the space of just seven pages, he gets through isos-
celes triangles, ideas of area, Pythagoras’ theorem, similar
 OUP CORRECTED PROOF – FINAL, 31/07/20, SPi

120 A ROYA L ROA D T O GEOM E T RY?

triangles (and their connection with area), circle theorems,

and more besides, all accompanied by proof.
Quite what Robert Recorde would have made of all this we
don’t know.
But we do know that The Young Mathematician’s Guide was
one of the most popular and best-selling mathematics books
of its time.

Not quite the ‘pizza theorem’

I should like to end this chapter by returning to Thomas
Malton’s Royal Road to Geometry.
For a most distinctive feature of that book which sets it
apart from several similar ones is that it contains a number of
interesting results and theorems that are not in Euclid.
These include, for instance, what we now know as
Varignon’s theorem, and a little-known theorem which
Malton describes as an ‘extraordinary property of the Circle’.

d R
a2 + b2 + c2 + d2 = 4R2
a b
c

Fig. 110 An ‘extraordinary property of the Circle’.

This theorem concerns any two chords which intersect at

right angles (Fig.  110). And, somewhat remarkably, the sum
a2 + b2 + c2 + d2 is then a constant, independent of where the
two chords intersect.
 OUP CORRECTED PROOF – FINAL, 31/07/20, SPi

A ROYA L ROA D T O GEOM E T RY? 121

More curiously still, some 200 years later, in the 1960s, this
result was coupled with some elementary calculus to produce
the so-called pizza theorem (Fig. 111).

Fig. 111 The pizza theorem.

This gives an exotic way of sharing pizza equally, because,

for any internal point P, the total grey area and the total white
area are the same, provided the angles between the four cuts
are all 45°!
I imagine that Thomas Malton cannot possibly have known
this, but sense from his 1774 book that he did know that
there was still a great deal of interesting and surprising geom-
etry yet to be discovered, even in the thoroughly classical
world of circles and straight lines.
 OUP CORRECTED PROOF – FINAL, 31/07/20, SPi

19

Unexpected Meetings

Some of the biggest surprises in geometry come about when

lots of straight lines meet unexpectedly at a single point.
This happens, for instance, if we take any old triangle, of
no particular shape, and draw a line through each corner
perpendicular to the opposite side (Fig. 112).

Fig. 112 The altitudes of a triangle.

The way these altitudes all meet at a single point has always
struck me as a bit peculiar. (It even happens if the triangle
ABC has one angle greater than 90°—though the meeting
point is then outside the triangle.)
And the history of this result is a little odd, too. It appears
to be classical, but is not in Euclid’s Elements, and the most
 OUP CORRECTED PROOF – FINAL, 31/07/20, SPi

U N E X PEC T E D M E E T I NGS 123

well-known proof, due to Gauss, dates only from the nine-

teenth century.
In any event, in order to understand Gauss’ proof we need
to begin with a related result which I have always found
rather less surprising, and which is in Euclid . . .

The perpendicular bisectors

The perpendicular bisectors of the sides of any triangle ABC
meet at a single point, O (Fig. 113).

A B
Fig. 113 The perpendicular bisectors.

To see why, imagine, if you will, all the circles that pass
through A and B.  Only one of them will pass through the
third corner C, giving the triangle ABC a unique circumcircle,
with centre O.
The sides of the triangle are chords of this circle, and it is
intuitively clear, I think, by symmetry, that their perpendicu-
lar bisectors will all pass through the centre O, and therefore
meet at a single point.
A more formal proof is possible, using congruent triangles
(see Notes, p. 252). And the meeting point is, once again, out-
side the triangle if one of its angles is greater than 90°.
 OUP CORRECTED PROOF – FINAL, 31/07/20, SPi

124 U N E X PEC T E D M E E T I NGS

The altitudes
To prove that the altitudes of a triangle ABC meet at a single
point, Gauss introduces what appears, at first sight, to be a
rather mind-boggling construction. He draws a line through
each corner parallel to the opposite side, so as to embed the
original triangle ABC in a larger one, DEF (Fig. 114).

C D
E

A B

Fig. 114 Gauss’ proof that the

altitudes meet at a single point.

This creates pairs of parallelograms such as ABCE and

ABDC. And as opposite sides of parallelograms are equal,
this implies that EC = CD (because both are equal to AB), so
that C is the mid-point of ED.
In the same way, A and B are also mid-points, and sud-
denly we see that the altitudes of triangle ABC are the perpen-
dicular bisectors of triangle DEF, and therefore meet at a single
point, called the orthocentre of the original triangle ABC, and
denoted by H.
 OUP CORRECTED PROOF – FINAL, 31/07/20, SPi

U N E X PEC T E D M E E T I NGS 125

The angle-bisectors
The three angle-bisectors of a triangle meet at a single point I,
called the incentre (Fig. 115).

I
Fig. 115 The incentre of a
B C triangle.

This is evident, I think, if we regard it as intuitively obvious

that
(a) there is a unique incircle within the triangle touching
all three sides and
(b) the angle-bisectors will, by symmetry, all pass through
its centre I.
Once again, a more formal proof is possible (see Notes,
p. 252).

The medians
Take any old triangle, and draw a line from each corner to the
mid-point of the opposite side.
Then these ‘medians’ also meet at a single point G, called
the centroid (Fig. 116).
 OUP CORRECTED PROOF – FINAL, 31/07/20, SPi

126 U N E X PEC T E D M E E T I NGS

G
A B Fig. 116 The medians of a triangle.

To prove it, consider first just the medians through A and B

(Fig. 117). Triangles ECD and ACB are similar by SAS, with a
‘scale factor’ of 2. So ED is parallel to AB, and AB = 2 × ED.

E D
G Fig. 117 Proof that the medians
A B meet at a point.

Then, after identifying some alternate angles, we see that

triangles EGD and BGA are similar by AA, again with a scale
factor of 2.
So

AG = 2 ´ DG and BG = 2 ´ EG.

In other words, these two medians cut each other at a point

which is, in each case, two-thirds of the way from the triangle
corner to the opposite mid-point.
Applying the same argument to a different pair of medians,
we find that all three medians cut each other in this particular
way, and therefore meet at a single point.
 OUP CORRECTED PROOF – FINAL, 31/07/20, SPi

U N E X PEC T E D M E E T I NGS 127

And there’s a lot more . . .

In a general triangle, of no particular shape
1. The circumcentre O, centroid G, and orthocentre H all
lie on a straight line, called the Euler line (Fig. 118a).
2. The altitudes are the angle-bisectors of the so-called
pedal triangle in Fig. 118b.

H
(a) Euler Line

G
O

(b) Pedal Triangle

H
(c) Nine-Point Circle

Fig. 118 Surprising properties of a triangle. (Further details can

be found in Notes, p. 253.)
 OUP CORRECTED PROOF – FINAL, 31/07/20, SPi

128 U N E X PEC T E D M E E T I NGS

3. The circle through the feet of the altitudes passes

through the mid-points of the sides and the mid-
points of the lines joining H to the corners of the tri-
angle (Fig. 118c).
And the story doesn’t end there.
We have looked at just four special points in a triangle, and
some of the associated properties. Yet there are many other
special points in a triangle, and a famous website currently
lists over 16,000 of them (see p. 266).
We will exercise rather more restraint, and introduce just
one more in the next chapter, which is largely about a new,
and very different, perspective on the ones we have met
already.
 OUP CORRECTED PROOF – FINAL, 31/07/20, SPi

20

Ceva’s Theorem

This theorem has a very classical feel to it, but was in fact only
discovered in 1678.
It concerns what happens when we join each corner of a
triangle to some point on the opposite side (Fig. 119).

Fig. 119 Ceva’s theorem.

Each such point divides its side in a certain ratio, and the
theorem itself says that if the three lines meet at a single point
P, then the product of those three ratios must be 1.
And to prove it we need just one very simple idea.
The triangles A and B in Fig. 120 have the same height, and
as the area of a triangle is ‘ 1 base ´ height ’, their areas must
2
therefore be in proportion to their bases.
 OUP CORRECTED PROOF – FINAL, 31/07/20, SPi

130 C E VA’S T H EOR E M

Area A a
=
A B Area B b

a b

Fig. 120 An area-ratio theorem.

So, to prove Ceva’s theorem we begin by observing in

Fig. 121 that
Area ABX Area ACX
=
BX XC
and
Area PBX Area PCX
= .
BX XC
Subtracting, we get the following:
Area ABP Area ACP
= ,
BX XC

Y
Z
P

Fig. 121 Proof of Ceva’s

B X C theorem.
 OUP CORRECTED PROOF – FINAL, 31/07/20, SPi

C E VA’S T H EOR E M 131

so
BX Area ABP
= .
XC Area ACP

In the same way, then,

CY Area BCP
=
YA Area ABP
and
AZ Area ACP
= .
ZB Area BCP

And when we multiply these three ratios together, every-

thing cancels (!), and the product is therefore 1.
Arguably, however, the converse of Ceva’s theorem is even
more important . . .

The converse of Ceva’s theorem

This says that if, in Fig. 122,
BX CY AZ
× × = 1,
XC YA ZB
then the three lines AX, BY, CZ meet at a single point.

Z Y
P

Fig. 122 Proving the

B X′ X C converse.
 OUP CORRECTED PROOF – FINAL, 31/07/20, SPi

132 C E VA’S T H EOR E M

To prove it, let P be the point at which BY and CZ meet, and

let AP meet BC at some point X′.
Then by Ceva’s theorem itself:
BX¢ CY AZ
× × = 1,
X¢C YA ZB
and on combining this with the given condition on X we
have
BX¢ BX
= ,
X¢C XC
which means that X′ and X divide BC in the same ratio. So X′
must, in fact, be the point X, which proves the result.

The medians revisited

If X, Y, Z are the mid-points of the sides of the triangle, then
BX CY AZ
× × = 1,
XC YA ZB
because each individual ratio is equal to 1 (!).
This provides a new ‘angle’, then, on why the medians
meet at a point.

The altitudes revisited

The converse of Ceva’s theorem provides a new perspective,
too, on why the altitudes meet at a single point.
For, if we draw in one of them (Fig. 123), we see that
BX c cos B
= ,
XC b cos C
and if we continue round the triangle in this way then Ceva’s
product is
 OUP CORRECTED PROOF – FINAL, 31/07/20, SPi

C E VA’S T H EOR E M 133

b
c

B X C Fig. 123 An altitude.

c cos B a cos C b cos A

× × = 1.
b cos C c cos A a cos B

The Gergonne point

A more exotic application arises if we draw the incircle to our
triangle and let X, Y, Z be the points of tangency (Fig. 124).

A
a a
Y
Z

c
b I

B b X c C

Fig. 124 The Gergonne point of a triangle.

As the two tangents to a circle from an external point are

equal, we find that
BX CY AZ b c a
= =1
XC YA ZB c a b
so that AX, BY, and CZ meet at a point.
 OUP CORRECTED PROOF – FINAL, 31/07/20, SPi

134 C E VA’S T H EOR E M

The point in question is named after J.  D.  Gergonne

(1771–1859), and is not typically the same as the centre of
the incircle, I.

How Ceva did it

Ceva was a hydraulic engineer, and discovered his theorem,
apparently, using mechanics.
In particular, Archimedes’ law of the lever concerns the
balance, under gravity, of two masses m and M at different
distances from a pivot (Fig. 125).

d D
md = MD

m M

Fig. 125 Archimedes’ law of the lever. (The rod joining the two
masses is assumed to be weightless.)

Now consider the triangle in Fig. 126, and imagine that we

make it out of weightless material and place masses mA, mB,
and mC at the corners.
Suppose, too, that we decide to choose those masses so that
the point P is the centre of gravity, in which case the triangle,
when arranged horizontally, will balance on a pinhead at P.
It will then certainly balance on any of the three lines
through P in Fig. 126, and three applications of the law of the
lever give
mB a1 = mC a 2 , mCb1 = mA b2 , m A c1 = mB c 2 .
 OUP CORRECTED PROOF – FINAL, 31/07/20, SPi

C E VA’S T H EOR E M 135

mA

c1 b2

P
c2 b1
Fig. 126 ‘Mechanical’ proof
mB a1 a2 mC of Ceva’s theorem.

On eliminating the three masses we obtain

a1b1c1 = a 2 b2 c 2 ,

which is Ceva’s theorem.

And, as it happens, this whole idea of using mechanics to
do geometry was not as original as one might suppose.
For we now know that Archimedes himself used methods
of this general kind to discover some of his most famous
results, including the formula for the volume of a sphere.
123 RU N N I NG H E A D

m e f u t h e r
So
s l i ce s o f
1674 Leibniz uses calculus to show that

1706 The symbol π appears in print for the first

time with its modern meaning, in William Jones’
A New Introduction to the Mathematics.
 R U N N I NG H E A D 123

1748 A discovery by Leonhard Euler links π with e = 2.718 . . .

and the imaginary number i = √−1

1767 Johann Lambert proves that π is irrational:

p
π≠ where p and q are
q whole numbers
 OUP CORRECTED PROOF – FINAL, 31/07/20, SPi

21

A Kind of Symmetry

There is a striking formula, usually credited to Heron of

Alexandria, for the area of a triangle in terms of the lengths of
its sides (Fig. 127).

Fig. 127 Heron’s formula for the area of a triangle.

It is striking, surely, because of how the three lengths a, b,

and c all enter the formula in exactly the same way. Yet a
moment’s thought reveals this to be rather reassuring.
Suppose, for instance, we have a triangle with sides 5, 6,
and 7, and want to know its area. We could set a = 5, b = 6,
c = 7 and work it out.
But we would surely be alarmed if setting, say, a = 7, b = 6,
c = 5 were to lead to a different answer, because (by SSS) these
define essentially the same triangle, but ‘viewed’ in a different
way.
 OUP CORRECTED PROOF – FINAL, 31/07/20, SPi

A K I N D OF S Y M M E T RY 139

In other words, as the a, b, and c in the formula are just ‘the

sides of the triangle’, with nothing to distinguish which is
which, we need to be able to interchange them freely and still
get the same answer.
This kind of symmetry often provides a valuable check on
an answer, and can also lend a certain elegance to mathemat-
ical results.
In particular, the radii of the incircle and circumcircle of a
triangle can be written in symmetric ways in terms of the tri-
angle’s sides and its area Δ (see Fig. 128 and Notes, p. 258).

INCIRCLE

2Δ
r r=
a+b+c

CIRCUMCIRCLE

R
abc
R=
4Δ

Fig. 128 Incircle and circumcircle radii.

 OUP CORRECTED PROOF – FINAL, 31/07/20, SPi

140 A K I N D OF S Y M M E T RY

Newton and the altitudes

Symmetry ideas of this kind can also be used to good effect in
proofs.
A fine example, in my opinion, is a little-known proof by
Isaac Newton that the altitudes of a triangle meet at a single
point, which appears in an unpublished manuscript of 1680
(Fig. 129).

Fig. 129 The diagram for Newton’s proof, from his Geometria
curvilinea and Fluxions, Ms Add. 3963, p54r.

His plan is simply to show, by direct calculation, that the

altitudes from A and B meet the one from C at the same height.
In Fig. 130, then, we begin by drawing the altitude CD, and
denote the three lengths shown by a, b, and L.
And as soon as we draw the altitude through A we find
that the marked angles are equal, because they are both
90° - ÐABC .
 OUP CORRECTED PROOF – FINAL, 31/07/20, SPi

A K I N D OF S Y M M E T RY 141

Fig. 130 Two altitudes of a triangle.

So the triangles ADP and CDB are similar, and therefore

h a
= ,
b L
i.e. ab
h= .
L
But this is—unexpectedly, I think—symmetric in a and b, so
that when we repeat the whole calculation for the altitude
from B, all that will happen is that a and b will change places
in the formula, L will stay the same, and we will end up with
the same value for h as before.

The eyeball theorem

Much the same thing happens in a curious theorem dis-
covered in the 1960s by the Peruvian geometer Antonio
Gutierrez.
 OUP CORRECTED PROOF – FINAL, 31/07/20, SPi

142 A K I N D OF S Y M M E T RY

Take two circles, and draw tangents to each from the centre
of the other (Fig.  131). Then the two straight lines AB and
A′ B′ are always equal in length, regardless of the relative size of
the two circles.

B B′

O O′
A A′

Fig. 131 The eyeball theorem.

Yet the reason becomes clear as soon as we calculate, say,

AB in terms of the radii r, r′, and the distance D between the
centres, for this turns out to be
2 rr ¢
AB = ,
D
which is unexpectedly symmetric in r and r′ (see Notes,
p. 259).

The medians, by coordinate geometry

Yet another neat use of symmetry is a coordinate–geometry
proof that the medians of a triangle meet at a point.
We need, first, one key result that I have not yet mentioned.
If, in Fig. 132, the points P1 and P2 have coordinates (x1, y1) and
(x2, y2), then the coordinates of any point P on the line between
them can be written
 OUP CORRECTED PROOF – FINAL, 31/07/20, SPi

A K I N D OF S Y M M E T RY 143

x = (1 - l )x1 + lx 2
y = (1 - l ) y1 + ly 2 ,

where λ denotes the fraction P1P/P1P2.

P2
P

P1 Fig. 132 A point on a line.

Thus, λ = 0 corresponds to P1, λ = 1 to P2, and the mid-point,

with l = 1 , has coordinates
2
1 1
( x1 + x2 ), ( y1 + y2 ).
2 2

(x3, y3)

G (x2, y2)

1 (x1 + x2), 1 (y1 + y2)

2 2
(x1, y1)

Fig. 133 The medians; a coordinate–geometry approach.

And if we apply this same idea to the median in Fig. 133, we

find that the coordinates of a point two-thirds of the way from
the corner to the mid-point of the opposite side are
1 1
( x1 + x2 + x3 ), ( y1 + y2 + y3 ).
3 3
But these coordinates are completely symmetric with
respect to the three corners of the triangle, so, by the same
 OUP CORRECTED PROOF – FINAL, 31/07/20, SPi

144 A K I N D OF S Y M M E T RY

argument, this particular point must lie two-thirds of the

way down both the other two medians as well.
In short, and with the help of a little symmetry, we have
proved that the three medians of a triangle meet at a single
point by considering in detail just one of them!
 OUP CORRECTED PROOF – FINAL, 31/07/20, SPi

22

‘Pyracy’ in Woolwich?

In 1747, Thomas Simpson, a mathematician at the Royal

Military Academy in Woolwich (Fig. 134), published a book
called Elements of Plane Geometry.
And, most unfortunately, a colleague at the academy
immediately accused him of plagiarism.

Fig. 134 The Royal Military Academy.

This colleague, the Professor of Fortification and Artillery,

claimed that Simpson’s book was merely an ‘incorrect copy’
 OUP CORRECTED PROOF – FINAL, 31/07/20, SPi

146 ‘PY R AC Y’ I N WO OLW IC H?

of his own work, adding

The Editor imagined, I suppose, that changing some

propositions, and mangling the demonstrations of others,
was a sufficient disguise to make it pass for his own
performance; but how far this will justify such a piece of
pyracy, must be left to the judgment of the publick.

For Simpson (Fig.  135) this was, perhaps, just one more
upset in a somewhat turbulent life. For while he eventually
became a Fellow of the Royal Society, he had little formal
education, and taught himself basic mathematics from a
book that he acquired from a fortune-teller. He also married
a widow almost old enough to be his grandmother, who (so it
is said) treated him so badly in later life that he was driven to
‘guzzle porter and gin in low company’.

Fig. 135 Thomas Simpson (1710–1761).

In any event, Simpson fought back against the charge of

plagiarism, in a much-expanded second edition of his book,
 OUP CORRECTED PROOF – FINAL, 31/07/20, SPi

‘PY R AC Y’ I N WO OLW IC H? 147

Fig. 136 From the second edition of Thomas Simpson’s book,

titled Elements of Geometry (1760).
 OUP CORRECTED PROOF – FINAL, 31/07/20, SPi

148 ‘PY R AC Y’ I N WO OLW IC H?

published in 1760. And time does not seem to have soothed

his anger, for he claimed there to have difficulty giving his
side of the story

without trangressing the bounds of decency.

While Simpson won the argument, the whole episode was

something of a shame, because his book really does stand out
from many other geometry books of the time.
It starts boldly, with a veiled criticism of Euclid:

My design, in this little Book, is to lay, before the young

Beginner, an easy Method for acquiring a competent
Knowledge in the Subject of Geometry, without . . . being
obliged to go through a Number of useless and tedious
Propositions.

And he then proceeds quite rapidly, in his own way, with

a square-within-a-square proof of Pythagoras’ theorem,
and, more strikingly still, Varignon’s theorem (which was
quite new at the time) as early as page 29.
But what really makes Simpson’s book distinctive is a later,
and most unusual, section on The Maxima and Minima of
Geometrical Quantities.

What’s the smallest area?

Suppose, for instance, that we have two fixed lines AB and
AC, and a fixed point E somewhere between them (Fig. 137a).
 OUP CORRECTED PROOF – FINAL, 31/07/20, SPi

‘PY R AC Y’ I N WO OLW IC H? 149

B
C
(a) (b)
Fig. 137 A minimization problem, from Thomas Simpson’s
Elements of Plane Geometry (1747).

It is evident, I think, that by choosing the line through E

almost parallel to AB or AC we can make the shaded area as
big as we like.
But how should we choose the line through E to make the
shaded area as small as possible?
The answer, in short, is that we need to choose the line so
that E is at its mid-point.
Simpson’s proof is illustrated in Fig.137b, where DE = EF,
and GEH is some other line through E. By introducing FI, par-
allel to GD, he creates two congruent triangles EDG and EFI.
As a result, the area of EDG is greater than the area EFH, so by
changing from the line DEF to the line GEH we gain more
‘shaded’ area than we lose. The line DEF must therefore be the
one enclosing least area.
And while this solution is surely an elegant one, I would
argue that the next problem from Simpson’s collection has a
more elegant solution still, not least because it features, at a
 OUP CORRECTED PROOF – FINAL, 31/07/20, SPi

150 ‘PY R AC Y’ I N WO OLW IC H?

crucial moment, an ingenious appeal to the converse of

Thales’ theorem…

Queen Dido’s problem

According to legend, Queen Dido arrived at the North
African coast in about 800 BC, and was granted as much land
there as she could encompass with an ox-hide.
She promptly cut the ox-hide into very thin strips, and
stitched them all into one very long strip, but this still left her
with a problem.

Fig. 138 Dido’s problem (with a straight coastline).

With reference to Fig. 138, the problem is to maximize the

shaded area by varying the points A, B, and the shape of the
curve C, while keeping the length of the curve C fixed.
And Queen Dido—so it is said—knew the answer: C
should be a semi-circle, with AB as diameter.
 OUP CORRECTED PROOF – FINAL, 31/07/20, SPi

‘PY R AC Y’ I N WO OLW IC H? 151

Now, if we pass over the question of whether a maximum

actually exists at all (which can be a subtle matter in problems
of this general kind), then Simpson presents an elegant proof
that the answer must, indeed, be a semicircle.
Observe, first, that the area of a triangle with two given
sides will be a maximum when those sides are perpendicu-
lar (Fig.  139), as a direct consequence of the formula
1 base ´ height.
2

Fig. 139 Which has the largest area?

Next, imagine that we have found the arrangement giving

the maximum area in Queen Dido’s problem (Fig. 140a), and
that there is some point D on the enclosing curve such that
ÐADB is not 90°.
We can then quickly obtain a contradiction by changing
the length of AB until ÐADB is 90°, while keeping both the
size and shape of the segments ACD and DEB unchanged
(Fig. 140b).

D
D
E
E
C C 90°

A B A B

Fig.140 The heart of Simpson’s argument.

 OUP CORRECTED PROOF – FINAL, 31/07/20, SPi

152 ‘PY R AC Y’ I N WO OLW IC H?

Fig. 141 Simpson’s argument, as it appears in his Elements of

Geometry of 1760. The result is presented as a corollary to his
Theorem 14. The reference to ‘13.3’ is to Thales’ theorem.
 OUP CORRECTED PROOF – FINAL, 31/07/20, SPi

‘PY R AC Y’ I N WO OLW IC H? 153

The length of the curve ACDEB will be unchanged, but we

will have increased the enclosed area, contrary to hypothesis,
because the new triangle ADB will have greater area than the
old one, while the other two areas will be the same as before.
It follows, then, that ÐADB must be 90° for all points D on
the maximizing curve.
As a final flourish, we appeal to the converse of Thales’ the-
orem, and conclude that the maximizing curve must be a
semicircle with AB as diameter.
Interestingly, Simpson appears to slip up at this very last
step by appealing to Thales’ theorem itself, rather than its
converse.
More curiously still, the whole ingenious argument seems
to be universally credited, in all the books and papers that
I have seen, to the renowned Swiss geometer Jakob Steiner, in
the 1840s.
So far as I can determine, however, it was really invented
by Simpson, almost 100 years earlier.
 OUP CORRECTED PROOF – FINAL, 31/07/20, SPi

23

Fermat’s Problem

One of the most famous minimization problems in geom-

etry is this:

Given a triangle ABC, find the point P which

makes PA + PB + PC as small as possible.

It dates from 1644, when Fermat posed it as a challenge to

Torricelli, who was working in Florence.

B C

Fig. 142 The Fermat problem.

Torricelli solved it, and with one proviso, the solution is

really rather neat: choose P so that all three angles at P are 120°
(Fig. 142).
 OUP CORRECTED PROOF – FINAL, 31/07/20, SPi

F E R M AT’S PROBL E M 155

He did observe, however, that this only works if all the

angles of the original triangle ABC are less than 120°.

Torricelli’s approach
His main argument can be loosely paraphrased as follows.
Let P be the desired minimizing point, or Fermat point, in
Fig. 143, and draw the ellipse with foci B and C which passes
through P.
Imagine, now, moving round the ellipse. As the sum of the
distances from B and C will be constant (by the property
shown in Fig. 89), and P minimizes PA + PB + PC, it follows
that P must be the point on the ellipse that is nearest to A.

B C

Fig. 143 Torricelli’s main argument, as inferred from his Opere,

vol. 1, first published in 1919.

This means, then, that PA is perpendicular to the tangent at

P—as in Fig.  143. Moreover, PB and PC make equal angles
with that tangent (by the ‘reflection property’ of Fig. 90),
which implies that ÐAPB = ÐAPC.
Repeating the argument with, say, A and B as the foci then
gives the result—all three angles at P must be equal.
 OUP CORRECTED PROOF – FINAL, 31/07/20, SPi

156 F E R M AT’S PROBL E M

Fig. 144 Cavalieri seeking the Fermat point of triangle ACF, in his
Exercitationes Geometricae Sex of 1647.

Three years later, in 1647, Bonaventura Cavalieri, pub-

lished a very similar solution, which he had apparently
obtained independently (Fig.  144). Cavalieri also observed
that if one corner of the triangle happens to contain an angle
greater than 120° we need to do something rather more mun-
dane, and choose P as that wide-angled corner.
Then, a little later still, one of Torricelli’s students, Vincenzo
Viviani, found a different approach altogether . . .

Viviani’s approach
This elegant solution to the Fermat problem uses an elemen-
tary theorem due to Viviani himself.
 OUP CORRECTED PROOF – FINAL, 31/07/20, SPi

F E R M AT’S PROBL E M 157

Viviani’s theorem

This states that for any internal point P of an equilateral

triangle the sum of the perpendicular distances from
the  sides is a constant, independent of the position of P
(Fig. 145).

h2
h3
P
h1

Fig. 145 Viviani’s theorem: in an equilateral triangle,

h1 + h2 + h3 = constant.

And the proof is very simple: just join P to the three cor-
ners, dividing the triangle into three smaller ones. Then if
the original triangle has side L and height H we have, from
areas,
1 1
(h1 + h2 + h3 )L = HL ,
2 2
which proves the theorem and shows that the constant
value of h1 + h2 + h3 is, in fact, the height H of the equilateral
triangle.
This gives us what we need, then, for an entirely different
attack on the Fermat problem.
 OUP CORRECTED PROOF – FINAL, 31/07/20, SPi

158 F E R M AT’S PROBL E M

The Fermat problem (again)!

In Fig. 146, let PA, PB, and PC make equal angles of 120° with
each other. We want to show that this minimizes the sum
PA + PB + PC .
To do this, draw in the dashed lines perpendicular to PA, PB,
PC. As the angles of any quadrilateral add up to 360°, the
angles of the triangle DEF must all be 60°, so that triangle
must therefore be equilateral.

F
A
E
P

B C

Fig. 146 An alternative approach to the Fermat problem.

Now consider any other internal point Q of the triangle

ABC.
Plainly, from Fig. 146, the sum QA + QB + QC will be greater
 OUP CORRECTED PROOF – FINAL, 31/07/20, SPi

F E R M AT’S PROBL E M 159

than the sum of Q’s perpendicular distances from the sides of

triangle DEF, which will, itself, be equal to PA + PB + PC , by
Viviani’s theorem.
So the ‘three-equal-angles’ point P does indeed minimize
the sum of the distances to the corners of triangle ABC; any
other internal point Q leads to a greater value.
While it is often claimed that this whole approach is due to
Torricelli, and simply published by Viviani in his De Maximis et
Minimis of 1659, I have found no evidence for this in the sources
that I have seen, and suspect it is due to Viviani himself.
In any event, the treatment above is only a very loose para-
phrase of what Viviani actually does, not least because he
announces and proves his theorem for any regular polygon
(i.e. one with all its sides and angles equal), not just for an
equilateral triangle (Fig. 147).

Fig. 147 Viviani’s theorem, as it appears in his De Maximis et

Minimis (1659).

Thomas Simpson’s approach

Not surprisingly, Simpson includes this problem in his col-
lection of 1747, and tackles it in his own way.
 OUP CORRECTED PROOF – FINAL, 31/07/20, SPi

160 F E R M AT’S PROBL E M

He begins by recalling Heron’s result on the reflection of

light at a plane mirror, which we discussed in Chapter 9.
Next, he shows that a similar result holds for two points
outside a circle with centre O: the shortest path between them,
via a point P on the circle, is such that the two parts of the
path make equal angles with the radius OP.

Fig. 148 From Simpson’s Elements of Geometry (1760).

He then applies this to the Fermat problem, establishing

that all three angles must be 120° in a proof by contradiction
(Fig. 148).
 OUP CORRECTED PROOF – FINAL, 31/07/20, SPi

F E R M AT’S PROBL E M 161

In a separate publication, The Doctrine and Application of

Fluxions (1750), he notes—like Cavalieri before him—the
quite different solution if one angle of the triangle happens to
be greater than 120°.

Constructing the Fermat point

As part of his original solution to the problem, Torricelli gave
a practical way of actually constructing the Fermat point P.
On each side of the original triangle ABC, construct an
equilateral triangle and its circumcircle (Fig. 149a). The three
circumcircles will then meet at the Fermat point P.

A E E
F F A

P P
B
C B
C

D D
(a) (b)
Fig. 149 Constructing the Fermat point.

To see why this happens, let P denote the Fermat point, so

that ÐBPC = 120° . As ÐBDC = 60° , the four points P, C, D, B
all lie on a circle, by the converse of the circle theorem on
p. 70. In other words, the circumcircle of triangle BCD passes
 OUP CORRECTED PROOF – FINAL, 31/07/20, SPi

162 F E R M AT’S PROBL E M

through P, and the same argument applies, of course, to the

other two circles as well.
As it happens, however, when we come to the actual con-
struction process, we don’t really need to bother with the cir-
cles at all.
This is because our three equilateral triangles, together
with the second circle theorem on p. 68, imply that all six
angles around P in Fig.149b are 60°!
This in turn means that APD, BPE, and CPF are all straight
lines. So all we have to do to locate P is construct, say, the top
two equilateral triangles in Fig. 149b, draw in the lines BE and
CF, and see where they meet.

A ‘proof by rotation’
I would like to end this chapter with a particularly elegant
solution to the Fermat problem.
We begin by choosing any point P inside the triangle ABC,
and rotate the triangle PAB about B through 60° (Fig. 150).

A′
P′
P

B C
Fig. 150 Solving Fermat’s problem ‘by rotation’.
 OUP CORRECTED PROOF – FINAL, 31/07/20, SPi

F E R M AT’S PROBL E M 163

This creates an equilateral triangle BPP′, and, in conse-

quence, the original minimization problem turns out to be
equivalent to minimizing the length of the path CPP′A′.
Now, C and A′ are fixed points, but P is ‘movable’, and P′
will also move in response to how we choose P.
And if we choose P so that PA, PB, and PC all make angles of
120° with each other, the path CPP′A′ becomes straight (because
ÐBPP¢ = ÐBP¢P = 60° , while ÐBPC = ÐBP¢A¢ = 120° ), and
therefore the shortest possible.
While this proof is usually credited to J.  E.  Hofmann, in
1929, many of the ideas underlying it are really much older—
especially when we notice that ABA′, too, is an equilateral
triangle, sitting on one side of the original, and that the min-
imizing straight line CPP′A′ from Fig. 150 is none other than
the straight line CPF in Fig. 149b.
 OUP CORRECTED PROOF – FINAL, 31/07/20, SPi

24

A Soap Solution

Sometimes, with a really difficult minimization problem,

physics can lend a helping hand.

Fig. 151 A soap film approach to geometry.

Suppose, for instance, we have two Perspex plates separated

by four pins at the corners of a square, and we dip all this into
a bowl of water with a bit of washing-up liquid.
When we take it out again, there will be some kind of soap
film linking the pins.
 OUP CORRECTED PROOF – FINAL, 31/07/20, SPi

A SOA P SOLU T ION 165

Moreover, if we keep on doing this, then, sooner or later, a

particularly distinctive configuration will emerge, with five
straight portions and 2 three-way intersections (Fig. 151).
And this is, in fact, the solution to a tricky geometrical
problem.

What’s the shortest network?

Imagine, if you will, that our job is to link four towns by a
road network which is as short as possible.
And suppose, too, that the towns lie—somewhat conveni-
ently—at the corners of a square of side 1.
Then two possibilities which come quickly to mind, I
think, are shown in Figs. 152a and 152b. They certainly allow
someone at any particular town to travel to any other town,
which is our basic requirement.

(a) 3 (b) 3

(c) 2√2 = 2.828.. (d) 1+√3 = 2.732..

Fig. 152 In search of the shortest network.
 OUP CORRECTED PROOF – FINAL, 31/07/20, SPi

166 A SOA P SOLU T ION

With a little thought, however, we soon find that we can do

a bit better by using the diagonals (Fig. 152c).
Yet we do better still with the network shown in Fig. 152d,
with two three-way junctions where the roads make 120°
angles with each other.
To see this, observe that we now have some 30°, 60°, 90°
triangles (Fig. 153).
So QBsin60° = 1 , and as sin 60° = 3 / 2 we find that
2
1
QB = .
3
Also, QS = QB cos 60° , and as cos 60° = 1/ 2 we have
1
QS = .
2 3

A B

30°

1
—
2

60°
R S
P Q

Fig. 153 Finding

the length of a
D C network.

Finally, as RS = 1, the middle segment PQ will be 1 - 2 ´QS ,

i.e. 1 - 1/ 3 . So, if we add this to the four parts like QB we
end up with

1 + 3 = 2.732¼
 OUP CORRECTED PROOF – FINAL, 31/07/20, SPi

A SOA P SOLU T ION 167

as our grand total for the whole network, which is certainly

the shortest so far.
Now, as it happens, this is the shortest of all, and therefore the
solution to our problem, but proving this is not easy.
And, unfortunately, our soap film experiment doesn’t
‘prove’ it, either. A soap film will always try to minimize its

(a)

(b)

Fig. 154 Two different soap film ‘solutions’ to the same problem.
 OUP CORRECTED PROOF – FINAL, 31/07/20, SPi

168 A SOA P SOLU T ION

surface energy, which is proportional to its surface area, and

hence—in our particular case—to the length of the network.
But the laws of physics guarantee only that when the soap film
finally settles it will have less surface energy than it would
have in any slightly different state. For all we know, there is
some very substantially different state with less energy still.
The difficulty can be seen more clearly, perhaps, if our sys-
tem has slightly less symmetry, and in Fig. 154 the arrange-
ment of the pins is not quite square.
When I first withdrew the apparatus the film settled into
state (a), but by gently blowing on the film through a drinking
straw I decreased the length of the short middle section until
the film suddenly jumped into a completely different state (b),
which, as it turns out, is slightly shorter than (a).
It’s all a bit like throwing a ball carelessly over some bumpy
ground, with lots of hills and valleys (Fig. 155). With a bit of

Fig. 155 A simple system with multiple equilibrium states.

 OUP CORRECTED PROOF – FINAL, 31/07/20, SPi

A SOA P SOLU T ION 169

friction present, we know that, sooner or later, the ball will

end up at the bottom of one of the valleys, but we have no
reason to suppose that it will necessarily end up at the bottom
of the deepest one.
Nonetheless, the soap film experiments not only help point
the way towards a solution, but offer a fresh perspective on
why the three-way junctions here—and indeed in the Fermat
problem of Chapter 23—always involve angles of 120°.

T T Fig. 156 Equal forces in equilibrium.

Viewed in terms of forces, rather than energy, this is sim-

ply because the different parts of the soap film pull on each
other equally, and the only way three equal forces can be in
balance at a single point is if the angles involved are all 120°
(Fig. 156).
 OUP CORRECTED PROOF – FINAL, 31/07/20, SPi

Fig. 157 The Ladies Diary (1740).

 OUP CORRECTED PROOF – FINAL, 31/07/20, SPi

25

Geometry in The Ladies’ Diary

The Ladies’ Diary was a popular journal, published annually in

London, from 1704 till 1841.
It was, in part, an almanac, but it also contained

Delightful and Entertaining Particulars

which included some mathematical problems sent in by its

readers.
And, as it turned out, it was these problems that gave The
Ladies’ Diary its lasting fame.

Keeping one’s head

Many of the contributors were, in fact, men. In the 1715 edi-
tion, for instance, we find the following problem, posed by
‘Mr. Tho. Shepheard’:

If you walk all the way round the Earth’s equator,

your head travels further than your feet.
How much further, if your height is h?

Actually, this is a very loose paraphrase of the original

problem, because (a) the person is 5 foot 7 inches tall, (b) they
 OUP CORRECTED PROOF – FINAL, 31/07/20, SPi

172 GEOM E T RY I N T HE L A DIE S’ DI A RY

don’t go all the way round, and (c) the whole problem is cast
in verse so appalling that it is, I believe, best forgotten.

Fig. 158 A walk round the Earth.

In any event, the answer to the version of the problem in

Fig. 158 is
2p h
and—somewhat surprisingly, perhaps—independent of the
radius r of the Earth!
Yet this is simply because the feet travel a distance 2πr, the
head a distance 2π(r + h), and on subtraction the terms involv-
ing r cancel.

Pythagoras in the garden

The questions usually required rather more persistence and
ingenuity.
In the 1754 edition, for instance, a problem from ‘Mr
Timothy Doodle’ asks for the radius of a circular fountain,
given the distances of its rim from the four corners of a
 OUP CORRECTED PROOF – FINAL, 31/07/20, SPi

GEOM E T RY I N T HE L A DIE S’ DI A RY 173

rectangular garden. And the published solution (Fig.  159)

involves four applications of Pythagoras’ theorem, together
with a bit of algebra.

Fig. 159 Lots of Pythagoras, in a problem from The Ladies Diary

(1754).

Look through any window . . .

A slightly offbeat similar triangles problem appeared in the
1790 issue, and was solved by Miss Betty Claxton of Benwell,
near Newcastle upon Tyne. Her solution, published the
following year, is shown in Fig. 160.
The problem itself is to determine the height of a window
CD, given three obscure pieces of information. First, from the
point L we can see 3½ yards of the wall of the house opposite.
Second, if we move 5 yards closer to the window, to the point
S, we can see 6 yards of the wall. Finally, the house opposite is
12 yards beyond the window.
 OUP CORRECTED PROOF – FINAL, 31/07/20, SPi

174 GEOM E T RY I N T HE L A DIE S’ DI A RY

Fig. 160 Betty Claxton’s solution. Here a:b::c:d means a/b = c/d.

It is all, essentially, a novel variation on the AD 850 ‘ladder

problem’ of Chapter 8.

A penny-farthing problem
This problem, from the 1800 issue, is about proving the for-
mula for the distance D (Fig. 161).

D = 2√ab

a
b

Fig. 161 A penny-farthing problem.

 OUP CORRECTED PROOF – FINAL, 31/07/20, SPi

GEOM E T RY I N T HE L A DIE S’ DI A RY 175

The symmetric way in which the radii a and b appear in the

answer is reassuring, for if we ‘nip round the back’ we have
essentially the same problem, with plainly the same answer,
except that a and b will have swapped places.
Oddly enough, the whole problem resurfaced 150 years
later, in a most peculiar way.
For, in May 1952, Johanna Mankiewicz, 15-year-old niece
of the famous film producer, was faced with just this problem
as part of her geometry homework in Los Angeles.
And, unable to solve it, she wrote to Albert Einstein.

Fig. 162 A homework hint from Albert Einstein.

 OUP CORRECTED PROOF – FINAL, 31/07/20, SPi

176 GEOM E T RY I N T HE L A DIE S’ DI A RY

Just as remarkably, Einstein replied, possibly spurred on a

little by

I realize that you are a very busy man, but you are the only
person we know of who could supply us with the answer . . .

a remark which did not go down at all well with the principal
of Johanna’s school, especially when the whole story was
splashed all over The New York Times and countless other
newspapers around the world.
Curiously, Einstein’s reply (Fig. 162) doesn’t have the two
circles touching, and gives, in effect, only a general ‘hint’ on
how to solve the problem.

A tricky problem
In truth, many of the questions in The Ladies’ Diary were more
difficult than the ones I have chosen so far.

r
Fig. 163 A tricky
problem from The Ladies
Diary (1730).
 OUP CORRECTED PROOF – FINAL, 31/07/20, SPi

GEOM E T RY I N T HE L A DIE S’ DI A RY 177

One quite early example, from 1730, is shown in Fig. 163,

where the problem is to determine the radius r of the incircle.
Yet we are told nothing directly about the triangle itself, only
the radii a, b, c of the small circles.
This would seem to be a much more difficult problem, and
I, at least, have some trouble viewing it as one of the ‘Delightful
and Entertaining Particulars’ trumpeted on the journal’s
cover (see Notes, p. 260).
Yet the answer itself is something of a delight, because it is
not only reassuringly symmetric in a, b, and c, but unexpect-
edly simple:

r = ab + bc + ca .

* * *

The Ladies’ Diary ceased publication in 1841, and in order to

take our story still further we need to recognize that, by this
time, mathematicians were beginning to seriously question
the very foundations of geometry.
And if we wish to see something of this, the first thing we
need to do—not surprisingly—is take a much closer look at
what Euclid actually did . . .
123 RU N N I NG H E A D

U C L I D 1 8 4 7
E

Oliver Byrne's extraordinary

edition of 1847 claimed that,
with COLOUR , Euclid's
Elements could be mastered
"in less than one third the
time usually employed".
R U N N I NG H E A D 123
 OUP CORRECTED PROOF – FINAL, 31/07/20, SPi

26

What Euclid Did

Euclid’s Elements consists of 13 ‘Books’, each containing a

logical sequence of propositions. And each proposition is
immediately followed by its proof.
Book I deals with many of the fundamentals, including
congruent triangles, parallel lines, and area. It ends on a high
note, with Pythagoras’ theorem.

I Triangles, parallels, and area

II Geometric algebra
III Circles
IV Constructions
V Theory of proportions
VI Similar figures
VII–IX Number theory
X Irrationals
XI–XIII Solid geometry

Fig. 164 The structure of Euclid’s Elements.

Book II takes a geometric approach to several ideas that we

would probably view algebraically, such as that shown in
Fig. 165. But algebra—at least as we know it today—is essen-
tially a sixteenth-century invention, so Euclid had to find
another way.
 OUP CORRECTED PROOF – FINAL, 31/07/20, SPi

W H AT E UC L I D DI D 181

ab b2
(a + b)2 = a2 + 2ab + b2
a2 ab
Fig. 165 Geometric
a b algebra.

Book III is about circle geometry, and the early parts of our
Chapter 11 follow Euclid’s treatment fairly closely.
Euclid deals with similar triangles very late on, in Book VI,
the reason being that the whole subject is full of ratios of
lengths, and these can all too easily be irrational (Fig. 166).

√2
1

1 Fig. 166 An irrational number in geometry.

This was a major problem for the ancient Greeks, who

simply did not view irrational numbers like 2 as numbers at
all. And to deal with this, Euclid precedes his treatment of
similarity with a long, and highly sophisticated, ‘theory of
proportions’, due to Eudoxus, in Book V.
Books VII–X are, for the most part, not directly about
geometry at all. In Book VIII, Proposition 20, for instance, we
find Euclid’s famous proof that there are infinitely many
prime numbers.
He returns to geometry in Books XI–XIII, but mostly, now,
in three dimensions, culminating in a treatment of the five
regular polyhedra, or Platonic solids (Fig. 167).
 OUP CORRECTED PROOF – FINAL, 31/07/20, SPi

182 W H AT E UC L I D DI D

tetrahedron cube octahedron

dodecahedron icosahedron
Fig. 167 The Platonic solids. In each one, all the faces are identical
regular polygons.

While some editions (such as Barrow’s) include a further

two Books, these are not now thought to be by Euclid.

Euclid’s postulates
Euclid tries to state all his assumptions at the very outset.
He also tries to use as few as possible, and just five assump-
tions—variously called ‘postulates’ or ‘axioms’—relate spe-
cifically to geometry.
The first three are about constructions (Fig. 168), which
Euclid often uses to establish the existence of certain geo-
metrical concepts that we might, more casually, take for
granted.
 OUP CORRECTED PROOF – FINAL, 31/07/20, SPi

W H AT E UC L I D DI D 183

Fig. 168 Euclid’s first three postulates, from an edition of the

Elements by Robert Simson (1804).

The fourth postulate is rather different:

All right angles are equal to one another

and, in my experience, totally mystifying to a newcomer to

geometry, who is liable to think: ‘Of course they are! They’re
all 90 degrees!’
It is now usually interpreted as a statement about the nature
of space itself, namely that it is both homogeneous (the same in
all places) and isotropic (the same in all directions).
The idea is used, for instance, in the proof that opposite
angles are equal (Fig.  169), when we assert that, despite
their different directions, the lines AB and CD both contain at
E an angle of 180°—or, as Euclid would say, two right
angles.
 OUP CORRECTED PROOF – FINAL, 31/07/20, SPi

184 W H AT E UC L I D DI D

Fig. 169 Proof that opposite angles are equal, in Thomas

Simpson’s Elements of Plane Geometry (1747).

The fifth and final explicit starting assumption is Euclid’s

famous Parallel Postulate:

If a straight line falling on two straight lines make the

interior angles on the same side less than two right
angles, the two straight lines, if produced indefinitely,
meet on that side on which are the angles less than the
two right angles.

Needless to say, it is quite unlike the others, and in terms of

sheer subtlety it is in a league of its own. For this reason, we
will discuss it separately, in Chapter 27.
Interestingly, Euclid tries to advance as far as he can without
using the parallel postulate in the earlier parts of Book I
(Fig. 170), and that is where we turn next.
 OUP CORRECTED PROOF – FINAL, 31/07/20, SPi

W H AT E UC L I D DI D 185

Fig. 170 Some early propositions from Euclid, Book I.

The earlier parts of Book I

Congruence

Euclid starts with SAS, and ‘proves’ it by arguing that if one

triangle is placed on top of the other it will fit exactly (Fig. 171).
 OUP CORRECTED PROOF – FINAL, 31/07/20, SPi

186 W H AT E UC L I D DI D

Fig. 171 Euclid’s ‘proof’

of SAS congruence.

It is not at all obvious, however, that his fundamental assump-

tions allow us to do this sort of thing. And, to put it another way,
if they do, then do we really need the fourth postulate?
It could be argued, then, that SAS is more or less—as in the
present book—a starting assumption.
But while we also assumed ASA and SSS congruence as
fairly ‘obvious’, Euclid emphatically does not; he proves them
(see Notes, p. 261).
To indicate how this kind of thing can be done I will pre-
sent here a later, and rather elegant, proof of SSS due to Philo
of Byzantium (c.280–220 BC).

A proof of SSS congruence

Let two triangles satisfy the SSS congruence conditions. We

want to prove that they really are congruent.
To do this, take one triangle and place it underneath the
other, in the specific way shown in Fig. 172, and join CR.

C
st

A B

Fig. 172 Philo’s proof of SSS

st
R congruence.
 OUP CORRECTED PROOF – FINAL, 31/07/20, SPi

W H AT E UC L I D DI D 187

Now, triangle RAC is isosceles, so the angles s are equal.

Triangle RBC is also isosceles, so the angles t are equal.
So  ÐACB = ÐARB , and the triangles ACB and ARB are
therefore congruent by SAS.
This proof assumes, of course, the principal result about
isosceles triangles, and we turn next to how Euclid goes
about proving that.

Euclid on isosceles triangles

Euclid’s proof that the base angles of an isosceles triangle are
equal comes very early in the Elements, when all he has at his
disposal is congruence by SAS.
And his argument is a sophisticated one that has flum-
moxed beginners ever since (Fig. 173).

Fig. 173 Euclid’s proof that the base angles of an isosceles triangle
are equal, from Barrow’s edition.
 OUP CORRECTED PROOF – FINAL, 31/07/20, SPi

188 W H AT E UC L I D DI D

Why doesn’t he present a simpler proof?

There is one, for instance, that involves no constructions at
all, and is usually credited to Pappus (end of AD 3rd century).
We observe, simply, that, in Fig. 173, triangles BAC and CAB
are congruent, by SAS, so ÐABC = ÐACB .
It seems to me, at least, that this is logically valid, even
though it is really only a formal version of ‘If we nip round the
back of an isosceles triangle it’ll all look the same, won’t it?’
We can be more clear, I think, why Euclid doesn’t use the
proof I presented in Chapter  4 involving the angle-bisector
AD and SAS (Fig. 174).

Fig. 174 The angle-bisector approach to

B D C isosceles triangles.

The reason, almost certainly, is that Euclid hasn’t at this

stage proved—from his list of stated assumptions—that such
a thing as an angle-bisector exists, and this is, again, just one
more indication of how demanding the Elements can be in
terms of logical rigour.
 OUP CORRECTED PROOF – FINAL, 31/07/20, SPi

27

Euclid on Parallel Lines

Euclid defines parallel lines as follows:

Parallel straight lines are straight lines which, being in

the same plane and being produced indefinitely in both
directions, do not meet one another in either direction.

The parallel postulate

A slightly informal paraphrase of Euclid’s parallel postulate is
given in Fig. 175.

Fig. 175 The parallel postulate (somewhat informally).

Now, there are 180° in a straight line. So if, instead, a + b is

greater than 180° in Fig. 175, the equivalent angle-sum on the
 OUP CORRECTED PROOF – FINAL, 31/07/20, SPi

190 E UC L I D ON PA R A L L E L L I N E S

left will be less than 180°, and by the postulate itself the two
lines will then meet somewhere on the left.
So, according to Euclid’s parallel postulate, the only way
the two lines can be parallel is if a + b = 180°.
Or, to put it slightly differently, if L and L′ are parallel, then
a + b = 180°. And if we appeal again to 180° in a straight line,
it follows that
If two lines are parallel, alternate angles are equal (Fig. 176).

Fig. 176 If two lines are parallel . . .

Moreover, as opposite angles are equal, this is equivalent to

our first major assumption in Chapter 2: if two lines are par-
allel, the corresponding angles are equal.
But while we immediately went on to assume the converse,
too, Euclid emphatically does not; he presents—in effect—a
proof of it, as we see next.

Book I, Proposition 27
This proposition, indicated schematically in Fig.  177, is as
follows:
If alternate angles are equal, the two lines are parallel.
 OUP CORRECTED PROOF – FINAL, 31/07/20, SPi

E UC L I D ON PA R A L L E L L I N E S 191

Fig. 177 If alternate angles are equal . . .

Euclid’s proof of this appeals to a somewhat obscure (and,

I think, relatively little-known) Proposition  I.16, where he
presents a proof that the ‘exterior’ angle of any triangle, c, is
greater than either of the interior angles a and b (Fig. 178a).

B F

b b
D

a c
A C E
(a) (b)

(c)
Fig. 178 Diagrams for Euclid I.16 and I.27.
 OUP CORRECTED PROOF – FINAL, 31/07/20, SPi

192 E UC L I D ON PA R A L L E L L I N E S

To prove that b < c, he joins A to D, the mid-point of BC

(Fig.178b), and extends the line by its own length to
F. Triangles ADB and FDC are then congruent by SAS, so that
b = ÐDCF, which is less than ÐDCE = c .
He then observes, in I.27, that if alternate angles are equal,
and the lines in question were not parallel, they would have to
meet (Fig.  178c)—giving an immediate contradiction with
the result b < c in I.16.

Fig. 179 Some later propositions in Euclid Book I.

 OUP CORRECTED PROOF – FINAL, 31/07/20, SPi

E UC L I D ON PA R A L L E L L I N E S 193

The angle-sum of a triangle

This comes quite late in Book I, as Proposition 32, because it
depends critically on ideas about parallel lines.
Euclid’s proof is slightly different from the one in Chapter 2,
and draws on one pair of alternate angles and one pair of
corresponding angles.

Fig. 180 Euclid I.32, as it appears in Thomas Simpson’s Elements of

Plane Geometry (1747).

With reference to Fig. 180, he now proves, essentially, that

the exterior angle CBD is not simply greater than either angle
A or angle C (Proposition I.16); it is actually equal to their sum.
And from this it quickly follows that the three internal
angles of the triangle must add up to 180°.

Similar triangles
We saw in Chapter  8 how ideas of similarity are related to
those of area, but they are bound up with ideas of parallelism
 OUP CORRECTED PROOF – FINAL, 31/07/20, SPi

194 E UC L I D ON PA R A L L E L L I N E S

as well, and all these things come together in Euclid’s treat-

ment of the subject, in Book VI.
The cornerstone of this treatment is Proposition 2, where
he takes a triangle ABC and draws in DE, parallel to BC
(Fig. 181).

Fig. 181 The figure for

Euclid VI.2, from Barrow’s
edition.

The two triangles ABC and ADE are then similar, and
Euclid establishes that DB/AD = EC/AE as follows.
First, by the area-ratio theorem of Chapter 20:
DB AreaDBE
=
AD AreaADE
and
EC AreaECD
= .
AE AreaADE
Finally,
AreaDBE = AreaECD,
 OUP CORRECTED PROOF – FINAL, 31/07/20, SPi

E UC L I D ON PA R A L L E L L I N E S 195

by viewing these two triangles in a different way, with the same

‘base’ DE and standing between the same two parallels, DE
and BC.
Euclid’s result then follows, and on adding 1 to both sides
we find that

AB AC
= ,
AD AE
so that similar triangles have sides which are in the same proportion.
 123 RU N N I NG H E A D

C T U R E ?
O F b y PI
PRO
One proof of Pythagoras’ theorem has
become quite popular as an animation . . .

a b

c d
 R U N N I NG H E A D 123

This whole idea seems to date from Antonio

Lecchi's Elementa Geometriæ of 1753 :

If we extend HK and ON
to meet at L, triangles AKL
and BAC are congruent
(by SAS) and at right angles
to one another.
So LADF is a straight line,
LA=BC=DF, and everything
in the animation does fit!
 OUP CORRECTED PROOF – FINAL, 31/07/20, SPi

28

‘A New Theory of Parallels’?

The year 1888 saw the appearance of a very strange book by

the Oxford mathematician C. L. Dodgson—better known as
Lewis Carroll, the author of Alice in Wonderland.
Its title was A New Theory of Parallels, and in the book
Dodgson showed that we can replace Euclid’s parallel postu-
late, if we wish, by an obscure assumption involving a hexa-
gon inscribed in a circle (Fig. 182).

(a) (b)
Fig. 182 (a) C. L. Dodgson (self-portrait) and (b) his alternative to
Euclid’s parallel postulate.
 OUP CORRECTED PROOF – FINAL, 31/07/20, SPi

‘a n e w t h eory of pa r a l l e l s ’ ? 199

The assumption is that the area of the hexagon is greater

than the area of any one of the circular segments outside it,
and Dodgson himself wrote of this alternative:

that it is somewhat bizarre I am willing to admit.

But then the parallel postulate itself is a bit odd, not least
because it seems at first sight to be all about two lines not
being parallel.
Little wonder, then, that ever since Euclid’s time mathem-
aticians have looked around for alternatives.
And it eventually became clear that, given his other
assumptions, Euclid’s geometry can be developed perfectly
satisfactorily using various alternatives to the parallel postu-
late, including the assumption that
(i) the angle-sum of any triangle is 180°
or that
(ii) triangles can be similar without being congruent.
But the main alternative, still in frequent use today, is not
nearly so revolutionary.
In fact, it is only subtly different from the parallel postulate
itself . . .

Playfair’s postulate
In its usual modern form, this is:

Given a straight line L and a point P not on the line, there is one
and only one line through P parallel to L (Fig. 183).
 OUP CORRECTED PROOF – FINAL, 31/07/20, SPi

200 ‘a n e w t h eory of pa r a l l e l s ’ ?

Fig. 183 Playfair’s postulate.

While this is commonly attributed to the Scottish mathem-

atician John Playfair, it actually appears rather differently in
his Elements of Geometry of 1795:

Two straight lines which intersect one another cannot be

both parallel to the same straight line.

In addition, Playfair attributes the whole idea to others,

including a book called The Rudiments of Mathematics by
Ludlam, some ten years earlier (Fig. 184).

Fig. 184 From Ludlam’s Rudiments of Mathematics (1785).

In any event, Playfair’s book is significant in other ways too.

In particular, it contains over 50 pages of ‘Notes’ on Euclid’s
Elements, including an extensive discussion of parallel lines.
And he observes that one of the great difficulties occurs at
the very start, when we define parallel lines as lines in the
plane which never meet.
 OUP CORRECTED PROOF – FINAL, 31/07/20, SPi

‘a n e w t h eory of pa r a l l e l s ’ ? 201

It is only natural to ask, then, whether we can avoid some

of the difficulties by actually defining parallel lines in an
entirely different way . . .

‘Parallel’ = ‘constant distance apart’?

This is the definition of parallel lines adopted by Thomas
Simpson in his Elements of Plane Geometry of 1747 (Fig. 185).

Fig. 185 Thomas Simpson’s original (1747) approach to parallel

lines.

But, as Playfair points out, there’s a difficulty.

The original line is AB, and we are invited to draw perpen-
diculars, on the same side, all of equal length, from any point
on the line AB.
But it only takes two such perpendiculars to define a
straight line ‘parallel’ to the original, so how do we know that
the ends of all the other perpendiculars also lie on that line?
In fact, Simpson himself appears to have realized the diffi-
culty, because in the second edition of his book, in 1760, he
adopts a considerably more sophisticated approach.
 OUP CORRECTED PROOF – FINAL, 31/07/20, SPi

202 ‘a n e w t h eory of pa r a l l e l s ’ ?

‘Parallel’ = ‘in same direction’?

It is even more tempting, perhaps, to simply define parallel
lines as being ‘in the same direction’.
And, if asked what we mean by that, exactly, we could
say that they make equal angles with any third line that intersects
them both.
This amounts, really, to taking ‘corresponding angles are
equal’ as our definition of parallel lines (Fig. 186).

Fig. 186 From John Ward’s Compendium of Speculative Geometry,

published (posthumously) in 1765.

But Playfair again points out a difficulty.

We are entitled, presumably, to construct two lines that
make equal angles with one particular third line of our
choice, and call them parallel, but how do we then know that
these two lines also make equal angles with any other third
line that crosses them both?
Can we, perhaps, prove it?
 OUP CORRECTED PROOF – FINAL, 31/07/20, SPi

‘a n e w t h eory of pa r a l l e l s ’ ? 203

A false proof . . .

Here’s how not to do it.

In Fig. 187, let the angles a be given as equal. Then from
one of the triangles

b = 180° - a - d

a b

a c

Fig. 187 A false proof . . .

and from the other

c = 180° - a - d

and therefore b = c.
The trouble with this ‘proof’, of course, is that we have
used the fact that the angle-sum of a triangle is always 180°,
and all the proofs of this that we have seen so far depend on
the very thing that we are trying to prove.

. . . and another . . .

Perhaps we can prove that the angle-sum of a triangle is 180°

without using ideas of parallelism?
 OUP CORRECTED PROOF – FINAL, 31/07/20, SPi

204 ‘a n e w t h eory of pa r a l l e l s ’ ?

Suppose we just let the angle-sum of a triangle be S, say.

c2 c3

Fig. 188 . . . and

c1 a b c4 another one . . .

Then, in Fig. 188,
c1 + c 2 = S - a
c3 + c 4 = S - b.
If we now add, we find on the left-hand side the angle-sum of
the large triangle, so S = 2S – a – b. But a + b = 180°, and therefore

S = 180°.

The mistake this time is perhaps a little more subtle, but in

the very first line. While we didn’t make the classic error of
assuming what we were trying to prove—that the angle-sum
of a triangle is always 180°—we did assume it is a constant, S,
across three different triangles.
At best, then, the argument above shows that if the
angle-sum of a triangle is a constant, then that constant
has to be 180°.
And the only reason I have included both these false proofs
at this stage is to point out the kind of pitfalls that lie in wait
for the unwary as soon as we try to explore the very founda-
tions of geometry in any depth.
 OUP CORRECTED PROOF – FINAL, 31/07/20, SPi

29

Anti-Euclid?

No one really knows, I think, whether Euclid’s Elements was

ever intended for genuine beginners. But by the second half
of the nineteenth century it was certainly being used that way
in UK schools, even with quite young pupils.

Fig. 189 J. M. Wilson (1836–1931), an early

reformer of geometry teaching in schools.
 OUP CORRECTED PROOF – FINAL, 31/07/20, SPi

206 a n t i - e uc l i d?

Schoolteachers themselves were apparently divided on

whether this was a good idea. One mathematics master at
Eton, for instance, claimed that

the Euclid hour is always a bright one,

while J. M. Wilson (Fig. 189), during his time at Rugby School,

wrote of

the extreme repulsiveness of Euclid to almost every boy.

In May 1870, Rawdon Levett, a young schoolmaster in

Birmingham, wrote a letter to Nature which included the sentence

If the leaders of the agitation for the reform of our geometrical

teaching would organise an Anti-Euclid Association,
I feel sure they would meet with considerable and
daily-increasing support.

And, partly in response to this, the year 1871 saw the foun-
dation of the Association for the Improvement of Geometrical
Teaching (Fig.  190), which later became the Mathematical
Association.
The A.I.G.T. was never exactly an ‘anti-Euclid’ body, but it
certainly challenged Euclid as the standard introduction to
geometry in schools.
This was greatly welcomed by some, but bitterly opposed
by others, including Charles Dodgson at Oxford. In Euclid and
his Modern Rivals (1879) he vigorously defended the status
quo, and lampooned the A.I.G.T. mercilessly as

‘The Association for the Improvement of Things in General’.

Yet the case against Euclid for beginners is substantial.

 OUP CORRECTED PROOF – FINAL, 31/07/20, SPi

a n t i - e uc l i d? 207

Fig. 190 An early meeting of the A.I.G.T.

 OUP CORRECTED PROOF – FINAL, 31/07/20, SPi

208 a n t i - e uc l i d?

First, making real sense of Euclid’s logical structure is

extremely demanding, partly because the mathematical
backdrop at the time was so different to that of today.
The Elements can also be criticized as austere—perhaps
unnecessarily so. Here’s J. M. Wilson again:

Euclid places all his theorems and problems on a level,

without giving prominence to the master-theorems.

Another difficulty is the whole idea of proof by contradic-

tion, or reductio ad absurdum. While Euclid uses it a great deal—
and seems, at times, to almost relish it—the idea can be
conceptually difficult for beginners.
In 1867 this led, in fact, to one of the most bizarre editions
of Euclid ever, by Lawrence S. Benson of New York, who had
something of an obsession about removing all ‘proof by con-
tradiction’ and replacing it with something else (Fig.  191).
Unfortunately, he also claimed to prove that the area of a cir-
cle of radius r is exactly 3r2, and Benson’s Geometry eventually
became an object of ridicule, with the Southern Review of 1869
reporting that

No book could, in our humble opinion, be better adapted

to unhinge all the reasoning powers of the juvenile mind.

In any event, the main trouble with Euclid for beginners,

surely, is that there is just so much ‘proving the obvious’, sim-
ply because he tries to assume so little at the outset.
 OUP CORRECTED PROOF – FINAL, 31/07/20, SPi

a n t i - e uc l i d? 209

Fig. 191 From the title page of Benson’s Geometry (1867).

As the mathematician A. N. Whitehead (1861–1947) once

observed:

It requires a very unusual mind to undertake the analysis of

the obvious.

And in W.  D.  Cooley’s Elements of Geometry, Simplified and

Explained, for instance, published in 1860, we find this
passage:

Curiosity is awakened by the discovery of remarkable and

unsuspected properties. . . . These rewards of labour are, in
the ordinary system, too long delayed.

To this end, Cooley races along, not quite as fast as John

Ward in his Young Mathematician’s Guide, but in a similar spirit.
 OUP CORRECTED PROOF – FINAL, 31/07/20, SPi

210 a n t i - e uc l i d?

And there is another feature, too, that makes his book dis-
tinctive. Just occasionally, in what are essentially short
‘asides’, he mentions entirely informal approaches that help
bring the subject to life.

Fig. 192 From W. D. Cooley’s Elements of Geometry,

Simplified and Explained (1860).

In connection with the angle-sum of a triangle for instance,

he advocates tearing the corners off a paper triangle and re-
assembling them (Fig. 192).
This proves nothing, of course, except that the angle-sum
is approximately 180° in one particular case. But it was most
unusual for a book at this time to include such an ‘experi-
mental’ approach alongside the formal development.

The twentieth century

By the first half of the twentieth century, the reforms insti-
gated by the A.I.G.T. (and others) were well under way, and
the leading geometry books for schools included elements of
‘practical geometry’, often before the deductive treatment
(Fig. 193).
 OUP CORRECTED PROOF – FINAL, 31/07/20, SPi

a n t i - e uc l i d? 211

Fig. 193 Geometry, as I first met it. Books by Godfrey and

Siddons (1919), Durell (1934) and, bottom, P. Abbott’s Teach
Yourself Geometry (1959).
 OUP CORRECTED PROOF – FINAL, 31/07/20, SPi

212 a n t i - e uc l i d?

In 1946, the President of the Mathematical Association

observed in his annual lecture that Euclid’s Elements was,
indeed, a peak of Greek culture, but then added:

How quite it came to be inferred from this that the book

was pre-eminently suited for the education of small boys in
Victorian England . . . forms a curious chapter in the history
of education in this country.

In any event, whatever we may think of Euclid as an intro-

duction to geometry, especially for young people, we surely
acknowledge it as the most extraordinary and influential
book on mathematics ever written.
And even today, perhaps, some will have a sneaking sym-
pathy with one headmaster of Eton in the mid-nineteenth
century, of whom it is said

He divided the books of the world into three classes:

Class I: The Bible
Class II: Euclid
Class III: All the rest.
 OUP CORRECTED PROOF – FINAL, 31/07/20, SPi

30

When Geometry Goes Wrong . . .

One of the strangest books on my shelves at home, in Oxford,

claims that Pythagoras’ theorem is actually wrong.
In Fig.  194, for instance, we have (supposedly) a right-
angled triangle ABC, with the two shorter sides both of
length 1. Yet according to the book in question, the ‘true’
length of the hypotenuse is not 2 , but 3 .
2

Fig. 194 From Heisel’s Mathematical and Geometrical

Demonstrations (1934).
 OUP CORRECTED PROOF – FINAL, 31/07/20, SPi

214 w h e n geom e t ry g oe s w rong . . .

The author of the book, Carl Theodore Heisel, published it

himself in 1934, and it is now something of a collector’s item,
though sadly for all the wrong reasons.
So far as I can determine, Heisel simply rejects outright the
idea that irrational numbers such as 2 can have any place
in geometry.
Which brings me to his real obsession—and I do mean
obsession—the numerical value of π . . .

Pi in the sky
Tucked into my copy of Heisel’s book is a typewritten covering
letter, signed by Heisel himself, and dated 6 April 1934 (Fig. 195).
In it, he claims to prove that the exact value of π is
256
= 3.160¼,
81

Fig. 195 Covering letter from Heisel to Mr. E. K. Bennett, of

Wadsworth, Ohio (1934).
 OUP CORRECTED PROOF – FINAL, 31/07/20, SPi

w h e n geom e t ry g oe s w rong . . . 215

and, casting any modesty aside, describes this as ‘the most

wonderful mathematical discovery in the history of the
world’.
Yet we saw in Chapter 14 how Archimedes proved that π
must lie between 3 10
71
and 3 71 , which corresponds to

3 ×1408¼< p < 3 ×1428¼.

So, not content with taking on Pythagoras, Heisel takes on

Archimedes as well.
In the interests of balance we could, I suppose, note than
an earlier π-enthusiast, James Smith of Liverpool, claimed to
prove in 1869 that π is actually
25
= 3 ×125,
8
which, in contrast, is too small to be consistent with
Archimedes.
Smith inflicted his ideas on various mathematicians of the
day, including J. M. Wilson at Rugby School, who made the
mistake of replying, and then had great trouble ending the
correspondence.
Eventually, in desperation, Wilson tried sarcasm, writing

Sir, I admire your indomitable perseverance, your hand-

writing, and your liberal expenditure of postage stamps . . .

though the real ‘killer’ was

your conclusions are as certain as their premises, with

which they are in fact identical.
 OUP CORRECTED PROOF – FINAL, 31/07/20, SPi

216 w h e n geom e t ry g oe s w rong . . .

Sadly, this was all completely lost on James Smith, who

went on to publish his ideas about π (Fig. 196), only to suffer
very public ridicule.

Fig. 196 James Smith’s Geometry of the Circle (1869).

The real trouble, for both Smith and Heisel, was that the
Swiss mathematician Johann Lambert had proved as early as
1761 that π is irrational, so cannot be written exactly as the
ratio of any two whole numbers, let alone 25/8 or 256/81.

Squaring the circle?

The irrationality of π is linked to the problem of ‘squaring the
circle’, though the two are not quite the same thing.
That problem, which dates from ancient times, is to con-
struct, in a finite number of steps, a square equal in area to that
of a given circle, using only an unmarked ruler and compass.
And Ferdinand Lindemann proved in 1882 that this is
impossible.
 OUP CORRECTED PROOF – FINAL, 31/07/20, SPi

w h e n geom e t ry g oe s w rong . . . 217

Yet 15 years later, Edwin J. Goodwin, a country doctor in

Indiana, claimed to have done it, and persuaded his state rep-
resentative, Taylor Record, to introduce a bill establishing his
‘new mathematical truth’ as part of Indiana state law.

(a) (b)
Fig. 197 (a) From The Indianapolis News, Sat. 13 Feb, 1897.
(b) Professor Clarence Waldo, whose intervention stopped the
bill from becoming law.

The bill itself was extremely confused, with different parts

of it seeming to imply no fewer than six different values for π,
all of them rational.
And it was only when the bill reached the Indiana Senate
that it came to grief, sparing further embarrassment, thanks
to an intervention by a professor of mathematics at Purdue
University, who happened to be visiting at the time.
 OUP CORRECTED PROOF – FINAL, 31/07/20, SPi

218 w h e n geom e t ry g oe s w rong . . .

Is every triangle isosceles?

Now and again, geometry is deliberately done wrongly, in order
to make some cautionary point.
In Lewis Carroll’s Picture Book of 1899, for instance, there is
one of his favourite puzzles: a ‘proof’ that all triangles are
isosceles (Fig. 198).

The ‘proof’

Let the bisector of angle BAC meet the perpendicular bisector

of BC at F.
Draw FH perpendicular to AB and FG perpendicular to AC.

Fig. 198 Where’s the mistake?

Triangles AFH and AFG are congruent (by angle-sum +

ASA), so
(i) AH = AG
(ii) FH = FG
Triangles BDF and CDF are congruent (by SAS), so
(iii) FB = FC.
Combining (ii), (iii), and Pythagoras’ theorem gives
(iv) HB = GC.
Adding (i) and (iv) gives AB = AC !
 OUP CORRECTED PROOF – FINAL, 31/07/20, SPi

w h e n geom e t ry g oe s w rong . . . 219

The mistake

If the angle-bisector of A meets the perpendicular bisector of

BC at a single point F at all, that point always lies outside the
triangle not inside (Fig. 199).

B C
D

F Fig. 199 The mistake.

It lies, in fact, on the circumcircle of ABC, midway between

B and C, so that the arcs BF and FC, being equal, form the
same angle at A, consistently with AF being the bisector of
angle BAC.
In consequence of all this, when we drop perpendiculars
from F onto AB and AC, one of the two lands on an extension
of the side (AC in Fig. 199), rather than the side itself, and this
invalidates the very last step in the ‘proof’.
So, while some witty mathematician once said,

Geometry is the art of correct reasoning on

incorrect diagrams,

we are well advised, in general, to ensure that our diagrams

are not too incorrect.
 OUP CORRECTED PROOF – FINAL, 31/07/20, SPi

220 w h e n geom e t ry g oe s w rong . . .

Malfatti’s problem
In 1803, as part of a wider investigation, the Italian mathem-
atician Gianfrancesco Malfatti considered the question

Given a triangle of fixed size and shape, how do you

construct 3 non-overlapping circles inside it so that their
total area is as large as possible?

And he thought he knew the answer: choose the circles so

that each one touches two sides of the triangle and both the
other two circles (Fig. 200).

Fig. 200 Malfatti’s problem.

Bizarrely, it was not until 1930—more than a hundred

years after Malfatti’s conjecture—that two mathematicians
noticed something wrong in the simplest possible case—the
equilateral triangle (Fig. 201).
For, in his configuration, the circles occupy a fraction
0.729 of the triangle’s area, but we do very slightly better by
first sticking in the biggest circle we can, and then two smaller
ones.
 OUP CORRECTED PROOF – FINAL, 31/07/20, SPi

w h e n geom e t ry g oe s w rong . . . 221

π√3
= 0.729
(1+√3)2

11 π
= 0.739
27√3

Fig. 201 Something wrong with Malfatti’s

‘solution’ . . .

Even more oddly—and 35 years later still—someone else

noticed that we don’t really have to do any calculations at all
here to see that something is wrong.
For if the triangle in question is very long and thin,
Malfatti’s solution is rather obviously not correct, and we
surely do better with the second arrangement in Fig. 202.

Fig. 202 . . . something very wrong.

 OUP CORRECTED PROOF – FINAL, 31/07/20, SPi

222 w h e n geom e t ry g oe s w rong . . .

Finally, in 1967, someone showed that Malfatti’s solution

is  never correct, no matter what the shape of the original
triangle.
In geometry, then, a wrong conjecture can remain a wrong
conjecture for a very long time.
 OUP CORRECTED PROOF – FINAL, 30/07/20, SPi

31

New Angles on Geometry

For the past 200 years, classical geometry has continued to

flourish, both through new results and through new ways of
looking at old ones.

Dandelin spheres
In 1822, for instance, the Belgian engineer G. P. Dandelin dis-
covered a novel way of relating the two views of an ellipse,
namely as (a) a cross-section of a cone or (b) a curve for which
PA + PB is constant (Fig. 203).

A B

Fig. 203 Two views of an ellipse.

His idea was to introduce two spheres into the problem, one
above the plane that slices through the cone, and one below
(Fig. 204).
 OUP CORRECTED PROOF – FINAL, 30/07/20, SPi

224 n e w a ngl e s on geom e t ry

The sphere above the plane touches the cone along a hori-
zontal circle (shown) and is just big enough to touch the
plane at A. Similarly, the sphere below touches the cone along
another horizontal circle (shown) and touches the plane at B.
Note that once we have made our particular slice, and got
our particular closed curve, the two Dandelin spheres are
fixed in position and size.
Our aim now is to prove that as a point P moves around
the curve, PA + PB remains constant.
To do this, join P to the top of the cone by a straight line,
and let that meet the two horizontal circles at A′ and B′.
Now comes the final, beautiful step . . .
As PA and PA′ are both tangent to the upper sphere, from
the same external point P, they are equal. Similarly, PB and
PB´ are equal, because both are tangent to the lower sphere.
So PA + PB = PA′ + PB′.

A′

B P

B′

Fig. 204 Dandelin spheres.

 OUP CORRECTED PROOF – FINAL, 30/07/20, SPi

n e w a ngl e s on geom e t ry 225

But PA′ + PB′ is constant, because it is the fixed distance,

measured straight down the side of the cone, between the
two (fixed) horizontal circles.
So PA + PB is constant, too, and that completes the proof.

More unexpected meetings

Miquel’s theorem

Take any old triangle ABC. Let A′ be any point on the side
opposite A, and choose points B′ and C′ similarly.
Then the circles AB′C′, BC′A′, and CA′ B′ meet at a common
point (Fig. 205).

C
A′

B′

A C′ B Fig. 205 Miquel’s theorem.

This remarkable theorem, dating from 1838, is one of

several discovered by the French school teacher Auguste
Miquel.
And it has a most elegant proof (in Notes, p. 263) involving
the four-points-on-a circle theorem of Chapter  11 and its
converse!
 OUP CORRECTED PROOF – FINAL, 30/07/20, SPi

226 n e w a ngl e s on geom e t ry

Common chords

If three circles all intersect one another, the three common

chords meet at a single point (Fig. 206).

Fig. 206 Common

chords.

This, too, has an elegant proof (Notes, p. 264) using

three  applications of the intersecting-chords theorem of
Chapter 11.
I know little of the history of this particular result, but it
appears as something of a throwaway remark on p. 337 of
Hall and Stevens’ School Geometry of 1904, so I imagine it is
much older.

The seven-circles theorem

In 1974 an unusual book of just 68 pages appeared, with a

preface which begins as follows:

This collection of theorems in plane geometry, some of

which we believe to be new, illustrates that there may still
be plenty of interesting theorems awaiting discovery.
 OUP CORRECTED PROOF – FINAL, 30/07/20, SPi

n e w a ngl e s on geom e t ry 227

The main result involves a closed chain of six circles, each

touching the next. And the whole chain touches the inside of
a seventh circle (Fig. 207).

Fig. 207 The seven-circles theorem (1974).

Finally, if we join the points of tangency of ‘opposite’ cir-

cles in the chain, we find that the three lines meet at a single
point.
In this kind of way, then, unexpected meetings in geom-
etry continue to fascinate, even in modern times.

Non-periodic tiling
The year 1974 also saw the Golden Ratio

1+ 5
f=
2
make a remarkable reappearance in geometry through Roger
Penrose’s famous non-periodic tiling of the plane.
 OUP CORRECTED PROOF – FINAL, 30/07/20, SPi

228 n e w a ngl e s on geom e t ry

ϕ ϕ ϕ ϕ
1
ϕ

1
1
1 1

‘KITE’ ‘DART’

Fig. 208 Penrose tiling.

In his most well-known example, just two types of tile are

involved, the ‘kites’ and ‘darts’ of Fig. 208, both made from
the triangles that occur in the regular pentagon on p. 59 and
therefore involving the Golden Ratio.
 OUP CORRECTED PROOF – FINAL, 30/07/20, SPi

n e w a ngl e s on geom e t ry 229

And by adding a couple of simple rules about how the tiles

can be fitted together he produced an example where the
plane can only be tiled in a non-periodic (non-repeating) way.

Up-and-over geometry
Even Thales’ theorem has had something of a makeover in
recent years, and in a most unlikely context: the mechanism
for operating a typical ‘up and over’ garage door (Fig. 209).

P B P B

O
A
O
garage garage

Fig. 209 Up-and-over geometry.

The door itself is divided into three equal parts by the

points A and O.
The point A is constrained to move in a vertical shaft
which runs up to the (fixed) point P at the top of the front of
the garage. The point O, on the other hand, is attached to P by
a rod with length one-third of the door height.
And a direct consequence of this arrangement, which has
practical advantages, is that, when the door is opened, the
top of the door B moves exactly horizontally.
 OUP CORRECTED PROOF – FINAL, 30/07/20, SPi

230 n e w a ngl e s on geom e t ry

And the reason for this is, essentially, Thales’ theorem.

The only difference, really, is that one usually tends to
think of Thales’ theorem with AB as fixed and P as moving,
whereas here the point P and the direction PA are fixed, and it
is the diameter AB which moves!
 OUP CORRECTED PROOF – FINAL, 30/07/20, SPi

32

And Finally . . .

Many years ago, someone tried to impress me with a triangle

that had an angle-sum of 270°.
They took the surface of a sphere, and drew three lines
along ‘great circles’, to produce the triangle in Fig. 210.
And it does, indeed, contain three right angles!
This is an example of spherical geometry, a branch of the sub-
ject with a long history and—not surprisingly—of great
importance in navigation.
And it is in fact a special case of a beautiful theorem by

Fig. 210 Spherical geometry.

 OUP CORRECTED PROOF – FINAL, 30/07/20, SPi

232 a n d f i na l ly . . .

Girard (1629) which says that

æ 4A ö
Angle - sum = 180° ç 1 + ÷,
è S ø

where A is the area of the spherical triangle and S the surface

area of the sphere.
So if the triangle is very small and almost flat, so that A is
tiny compared to S, the angle-sum is almost 180°. For the tri-
angle in Fig.  210, however, A = 1 S , and the angle-sum is,
8
indeed, 270°!
Yet I’m sorry to say that, at the time, I was unimpressed.
‘But that isn’t really a triangle,’ I said. ‘Anyone can see that
the sides aren’t straight!’
I fear I lacked the imagination to see that they would cer-
tainly seem straight to anyone living—so to speak—in the
purely two-dimensional world of the surface itself, unaware
of anything ‘outside’.
More seriously still, there was no one around at the time to
tap me gently on the shoulder and say:

‘Now, young man . . . you seem terribly certain that you

know, in any conceivable context, what a “straight line” is.
Just how would you define one, exactly?’

And, just possibly, if there had been someone around to

say this, I would have had less trouble when I first encoun-
tered geometries in which some of our most familiar and
deeply entrenched ideas just don’t hold . . .
 OUP CORRECTED PROOF – FINAL, 30/07/20, SPi

a n d f i na l ly . . . 233

Non-Euclidean geometry

Throughout the history of geometry there have been numer-

ous attempts to prove Euclid’s parallel postulate from his
other assumptions.
But all failed, and in the 1830s it finally became clear why:
simply dropping Euclid’s parallel postulate altogether, while
keeping the others, can lead to quite different, but entirely
self-consistent, geometries (Fig. 211).

Fig. 211 Euclidean and non-Euclidean geometry.

This discovery is due essentially to the Hungarian mathem-

atician Janos Bolyai and the Russian mathematician Nikolai
 OUP CORRECTED PROOF – FINAL, 30/07/20, SPi

234 a n d f i na l ly . . .

Lobatchevsky, though Bolyai had to overcome the strongest

possible discouragement from his father:

You must not attempt this approach to parallels. I know

this way to the very end. I have traversed this bottomless
night, which extinguished all light and joy of my life.
I entreat you, leave the science of parallels alone.
Letter to Janos Bolyai from his father, 4 April 1820

But Bolyai persisted, and we now know that there are self-
consistent geometries with infinitely many lines through a
given point parallel to a given line, and triangles with angle-
sums which are always less than 180°.
And some years later, Riemann showed that there is a
geometry, too, with no lines through a given point parallel to
a given line, and triangles with angle-sums which are always
greater than 180°.
A further weird feature of both geometries is that there are
no similar triangles that are not congruent. This is because—as in
the case of spherical geometry—the angle-sum depends on
the size of the triangle.

Projective geometry

By the middle of the nineteenth century, the whole idea of

geometry had opened up in other ways, too.
Projective geometry, for instance, came to prominence
through problems of perspective in art, but something of its
 OUP CORRECTED PROOF – FINAL, 30/07/20, SPi

a n d f i na l ly . . . 235

B′ C′
A′

P
Q
R

A
B C Fig. 212 Pappus’ theorem.

spirit can be seen in a remarkable theorem due to Pappus of

Alexandria, in about AD 320.
Let A, B, C be successive points on a line, and A', B', C' suc-
cessive points on another (Fig.  212). Then let AB' and BA'
meet at P, and define the points Q and R, similarly.
Then P, Q, and R lie on a straight line.
And the most distinctive aspect of all this is that there is no
mention whatsoever of length, angle, or area.

Topology

This part of the subject is largely concerned with whether

one geometrical object can (or cannot) be continuously
deformed into another.
In Fig.  213, for instance, and despite appearances, the
objects in (a) and (e) are topologically the same, because one
can be gradually, and continuously, deformed into the other,
through the sequence shown.
 OUP CORRECTED PROOF – FINAL, 30/07/20, SPi

236 a n d f i na l ly . . .

(a) (b) (c)

(d) (e)
Fig. 213 Topology.

The Möbius band in Fig. 214, however, is topologically dif-

ferent from an ordinary (untwisted) band, because it is
impossible to deform one into the other without cutting or
tearing it. And the one-sidedness of the Möbius band leads to
some remarkable properties.

(a) (b)
Fig. 214 (a) A Möbius band. (b) A normal, untwisted band.
 OUP CORRECTED PROOF – FINAL, 30/07/20, SPi

a n d f i na l ly . . . 237

These are, I think, well known, but if you haven’t come

across them, try cutting a Möbius band—right along its
centre-line—into two bands, and see what happens!

Fractal geometry

If topology can be viewed as a sort of ‘broad brush’ approach

to geometry, then fractal geometry is at the opposite extreme,
and challenges our intuition in quite different ways.
An early example is the ‘Koch snowflake’ of 1904.
To obtain this, take an equilateral triangle, replace the mid-
dle third of each side by two lines of equivalent length, and
then continue doing this, with every straight portion of the
boundary, for ever (Fig. 215).

Fig. 215 Making a Koch snowflake.

The result is a geometrical curve which encloses a finite

area—just 8/5 times the area of the original triangle—yet the
curve itself is infinitely long, and has no clearly defined tan-
gent anywhere, because no matter how closely we look at some
little portion of it we continue to see further detail; things
never ‘smooth out’.
And while fractal curves like this were once viewed as odd-
ities, they now arise frequently in the subject of dynamics,
where they are often linked with the idea of chaos.
 OUP CORRECTED PROOF – FINAL, 30/07/20, SPi

238 a n d f i na l ly . . .

Would you believe it?

In 1924, two Polish mathematicians, Stefan Banach and

Alfred Tarski, proved that it is possible to dissect a solid
sphere into a finite number of pieces, and then reassemble
those pieces to form two solid spheres which are each the same
size as the original (Fig. 216).

DISSECT
AND
REASSEMBLE

Fig. 216 The Banach–Tarski paradox.

It is true that the pieces in the ‘in between’ stage are so

mathematically subtle that they have no clearly defined volume
at all, but the result still stunned the mathematical world.

* * *
In short, there are whole new worlds of geometry out
there, each with its own fresh surprises.
And in my experience, at least, the best surprises in geom-
etry tend to stay with you, throughout life—a little bit like old
friends.
 OUP CORRECTED PROOF – FINAL, 30/07/20, SPi

a n d f i na l ly . . . 239

In fact, to be quite honest, there are times when I wonder if

this particular book has been gently brewing ever since my
very first surprise in geometry, all those years ago, at school,
one cold winter morning in 1956, when I was ten.
OUP CORRECTED PROOF – FINAL, 30/07/20, SPi
 OUP CORRECTED PROOF – FINAL, 30/07/20, SPi

Notes

chapter 6: pythagoras’ theorem

The traditional source quoted in connection with the Chinese text

Zhou bi suan jing (Fig. 33, p. 35) is Joseph Needham’s Science and
Civilisation in China (Cambridge, 1959), vol. 3, pp. 22–3.
But this is, I believe, acknowledged to be a very free translation,
and a discussion of some of the difficulties can be found in
Christopher Cullen’s article ‘Learning from Liu Hui? A Different
Way to Do Mathematics’ in Notices of the American Mathematical
Society, vol. 49, pp. 783–90 (2002).

371 proofs of pythagoras

Ann Condit’s proof

In Fig. 217, ABC is our right-angled triangle, and P is the mid-point
of the hypotenuse AB.
We begin by drawing (dotted) lines through P perpendicular to
the sides, dividing all three squares in half.
The plan is to prove Pythagoras’ theorem by proving that Area
DQC + Area FRC = Area APG (and then multiplying by 4!).
By ‘shearing’ the first two triangles in a Euclid-like way this is
equivalent to proving that

Area DPC + AreaFPC = Area APG,

which is the heart of Condit’s proof.

 OUP CORRECTED PROOF – FINAL, 30/07/20, SPi

242 not e s

D
E
F

Q R

A P B

G
Fig. 217 Ann Condit’s proof.

To do this, she joins DF and extends PC to meet DF at E.

Now, triangles ACB and DCF are congruent by SAS, and after a
little angle-chasing, this leads to the important conclusion that PC
is perpendicular to DF.
So, by viewing PC as the ‘base’ of the triangles DPC and FPC,

AreaDPC+AreaFPC= 21 PC.(DE+EF)

= 21 PC.DF

But PC = AP, and because of the congruence just noted, DF = AB =

AG, and the result then follows.
 OUP CORRECTED PROOF – FINAL, 30/07/20, SPi

not e s 243

chapter 10: conversely . . .

Isosceles triangles
An alternative proof of the converse of the isosceles triangle theorem,
avoiding contradiction, goes as follows.
In Fig. 218, draw the angle-bisector AD.

B D C Fig. 218 Is it isosceles?

Then, as the angles of a triangle add up to 180°, ÐADB = ÐADC.

Triangles ADB and ADC are then congruent by ASA (with AD as
the side in question), so AB = AC.
Euclid couldn’t do this in Elements I.6, because he hadn’t at that stage
established either ASA congruence or the angle–sum of a triangle.

The converse of Thales’ theorem

Yet another way of proving this is by just using Pythagoras’
theorem—albeit somewhat relentlessly.
In Fig.  219, O is the mid-point of AB, with OA = OB = r, and
ÐAPB = 90°. And we want to prove that R = r, so that P lies on the
circle with AB as diameter.
 OUP CORRECTED PROOF – FINAL, 30/07/20, SPi

244 not e s

b a
R
h

A r–s s O r B Fig. 219 Is R equal to r?

One application of Pythagoras gives a 2 + b2 = 4 r 2.

Two more turn this into

h 2 + ( r + s )2 + h 2 + ( r - s )2 = 4 r 2 ,

which simplifies to h 2 + s 2 = r 2 . One final application of Pythagoras

gives h 2 + s 2 = R 2 , so R = r.

chapter 12: off at a tangent

The slightly informal proof of the secant–tangent theorem (Fig. 67,

p. 74) may be avoided, if desired, by appealing instead to the alternate
segment theorem, which says that ÐPTQ = ÐPRT in Fig. 220.

Fig. 220 Proving the secant–

P T tangent theorem.

Triangles PTQ and PRT are then similar by AA, so

PT PQ
= ,
PR PT
whence the result.
 OUP CORRECTED PROOF – FINAL, 30/07/20, SPi

not e s 245

The alternate segment theorem

To prove this, introduce the perpendicular to PT at the point of tan-
gency T, which will be a diameter of the circle (Fig. 221).

Fig. 221 Proving the alternate segment

P T theorem.

Then ÐPTQ = ÐQST because of Thales’ theorem, and

ÐQST = ÐQRT because they are ‘angles at the circumference’
standing on the same chord QT.

Measuring the Earth

It is worth noting, perhaps, that we don’t have to use the secant–tangent
theorem in Fig. 68, p. 75. We could, instead, just use Pythagoras’
theorem to give

( R + h )2 = D2 + R 2 ,

which leads directly to the same result.

chapter 13: from tangents to supersonic flow

Two additional results of great value in trigonometry are

sin( A + B ) = sin A cos B + cos A sin B
cos( A + B ) = cos A cos B - sin A sin B,
and I offer a picture-proof in Fig. 222.
 OUP CORRECTED PROOF – FINAL, 30/07/20, SPi

246 not e s

sinA sinB

sinA cosB

sin
A
B
1

sA
co cosA sinB

A
B
cosA cosB
Fig. 222 Finding sin( A + B ) and cos( A + B ) .

chapter 14: what is π, exactly?

Two important special cases of the results in the Note for Chapter 13
are the double-angle formulae:
sin 2 A = 2 sin A cos A
cos 2 A = cos2 A - sin2 A .
In view of Pythagoras’ theorem (cos2 A + sin2 A = 1), the second
of these is equivalent to
cos 2 A = 2 cos2 A - 1.

Setting A = q / 2 then leads to the ‘useful little result from trigo-

nometry’ which I used, at the very end of Chapter 14, when deriving
Viète’s remarkable formula for π.
 OUP CORRECTED PROOF – FINAL, 30/07/20, SPi

not e s 247

Inspector Euclid Investigates . . .

Mercifully, slopes come to the rescue, because the thick lines on the
diagram in Fig. 223 disguise the fact that ABC is not a straight line.

C
5 B
5

3
A 8 5
8
8 × 8 = 64 13 × 5 = 65
Fig. 223 A paradox . . .

Plainly, the slope of AB is 3/8, while the slope of BC is 2/5, which

is very slightly greater, and when we draw the second figure more
accurately (Fig.  224) we find that the re-assembled pieces leave a
very long, thin, parallelogram in the middle. This gap accounts
exactly for the apparent discrepancy.

5 8

5 5
3

8 5
Fig. 224 . . . and its resolution.
 OUP CORRECTED PROOF – FINAL, 30/07/20, SPi

248 not e s

This puzzle is often attributed to the great Sam Loyd, who

is  said to have presented it at the American Chess Congress
in 1858. But puzzles of this general kind have a very long history,
as can be seen from David Singmaster’s article ‘Vanishing Area
Puzzles’, in Recreational Mathematics Magazine, no. 1 pp. 10–21
(2014).

chapter 17: geometry and calculus

The slope of the curve y = x2

To find the slope at the point P, in Fig. 225, choose a nearby point Q,
also on the curve.
If the x-coordinates of P and Q are x and x + h, say, then their
corresponding y coordinates will be x2 and (x + h)2.
In consequence, as we move from P to Q,
increase in y 2 xh + h 2
= ,
increase in x h
and on cancelling the factors of h we have

increase in y
= 2 x + h.
increase in x

Q
y P y = x2

Fig. 225 Finding the slope of the curve

x h y = x2.

Finally, we notice that the closer we take Q to P, i.e. the smaller

we  take h, the closer this ratio comes to 2x, so that the slope of
the curve itself at P must be 2x.
 OUP CORRECTED PROOF – FINAL, 30/07/20, SPi

not e s 249

The parabola: focus and directrix

Associated with the parabola y = x2 there is a special point called the
focus F, with coordinates ( 0, 1 ) , and a special (horizontal) line called
4
the directrix with equation y = - 1 .
4
A key property is that any point P on the parabola, with coordinates
(x, y), is the same distance from F as it is from the directrix, so that

PF = PD

in Fig. 226.
To prove this, the distance formula in Fig. 97, p. 104, gives
PF2 = x 2 + ( y - 41 )2 and PD = ( y + 41 ) , and these turn out to be equal,
2 2

because y = x2.

T R
y

P (x,y)

F
¼ Q
x
¼

directrix D
Fig. 226 Geometry of the parabola.

The parabola: reflection property

Now let the line QPR bisect ÐFPD in Fig. 226. As PF = PD, FPD is an
isosceles triangle, so this angle-bisector will be perpendicular to the
triangle’s ‘base’ FD.
As F has coordinates ( 0, 41 ) and D has coordinates ( x , - 41 ) , the
slope of FD is - 1 .
2x
 OUP CORRECTED PROOF – FINAL, 30/07/20, SPi

250 not e s

The slope of QPR must therefore be 2x, which, as we have seen, is

the slope of the parabola itself at P.
So QPR must be the tangent to the parabola at P.
Moreover, as ÐQPD and ÐTPR are opposite angles, it follows
that

ÐFPQ = ÐTPR,

which is the reflection property.

The area under a curve

To understand the claim made in Fig. 104, p. 112, imagine x increas-
ing very slightly, as in Fig. 227. The area A will also increase slightly,
and the increase will almost correspond to a tall thin rectangle of
height y.

A y

x
Fig. 227 A small increase in area.

In other words, the small increase in area will almost be obtained

by multiplying y by the small increase in x.
Or to put it another way, if we divide the small increase in A by
the small increase in x we will get, very nearly, y.
 OUP CORRECTED PROOF – FINAL, 30/07/20, SPi

not e s 251

And that is, essentially, why the slope of the curve in Fig. 104b,
p. 112, is y.

chapter 18: a royal road to geometry?

To establish Malton’s ‘extraordinary property of the Circle’ in

Fig. 110, p. 120, draw diameters AC and BD in Fig, 228 to create a
rectangle ABCD.

E C
D

R
d

A a b B
c

Fig. 228 Proving Malton’s ‘extraordinary property’.

Then appeal to symmetry to claim that EF = c, and therefore

BC = d – c.
Finally, use Pythagoras’ theorem on triangle ABC to obtain

( a + b )2 + ( d - c )2 = ( 2 R )2 .

And on multiplying out, the result emerges, because any two

intersecting chords have ab = cd, whether they are at right angles or
not (Fig. 64, p. 72).
 OUP CORRECTED PROOF – FINAL, 30/07/20, SPi

252 not e s

chapter 19: unexpected meetings

The perpendicular bisectors

In Fig. 229, let D, E, and F be the mid-points of the sides, and let the
perpendiculars from D and E meet at O.

E D
O

Fig. 229 Proving that the perpen-

A F B dicular bisectors are concurrent.

Triangles OEA and OEC are congruent by SAS, so OA = OC.

Similarly, OC = OB.
Therefore OA = OB.
Now join OF.
Triangles OFA and OFB are then congruent by SSS.
So ÐOFA = ÐOFB, and OF is therefore the perpendicular bisector
of AB.

The angle-bisectors
In Fig. 230, let the angle-bisectors from B and C meet at I. Draw ID,
IE, IF perpendicular to the sides.

F
E
I

Fig. 230 Proving that the angle-

B D C bisectors are concurrent.
 OUP CORRECTED PROOF – FINAL, 30/07/20, SPi

not e s 253

Then triangles IDB and IFB are congruent (by angle-sum + ASA).
So IF = ID.
Similarly, IE = ID.
So IF = IE, and triangles IFA and IEA are therefore congruent
(by Pythagoras + SSS).
So ÐIAF = ÐIAE, and IA is therefore the bisector of angle BAC.

The Euler line

In Fig. 231, consider one perpendicular bisector OC′, one median
CC′, and extend OG by twice its length to a point that we will rather
presumptuously call H, so that GH = 2 × OG.

H
O G

A C′ B
Fig. 231 The Euler line.

As GC = 2 × GC′, the triangles OGC′ and HGC are then similar by

SAS, so ÐHCG = ÐOC¢G . This means that CH is parallel to OC′,
and therefore perpendicular to AB.
In other words, the point we have tentatively called H lies on the
altitude through C.
By repeating the argument, starting with a different corner and
opposite side, H lies on all three altitudes, and must therefore, indeed,
be the orthocentre.
 OUP CORRECTED PROOF – FINAL, 30/07/20, SPi

254 not e s

The pedal triangle

In Fig. 232, the four points A, F, H, and E all lie on a circle (with AH
as diameter, in fact), by the converse of Thales’ theorem (twice). So
the angles a are equal, as they both stand on the same chord EH
(Theorem (b) in Fig. 60, p. 68).
Similarly, the angles b are equal.

D
H

ab b
a

A B
F
Fig. 232 The pedal triangle.

Finally the angles ÐCAD and ÐCBE are equal, because they are
both 90° - ÐACB.
So a = b, and the altitude CF therefore bisects ÐEFD. The same
argument then applies to the other two altitudes.

The nine-point circle

In Fig. 233, we use the suffix 1 to denote the feet of the altitudes, 2
for the mid-points of the sides, and 3 for the mid-points of the lines
joining H to the corners of the triangle.
By the mid-point theorem, C2B2 is parallel to BC.
Again by the mid-point theorem, B2C3 is parallel to AH, and
therefore perpendicular to BC.
 OUP CORRECTED PROOF – FINAL, 30/07/20, SPi

not e s 255

B1 C3

A1
B2
A2
H

A3
B3

A C2 C1 B
Fig. 233 The nine-point circle.

So B2C3 must be perpendicular to C2B2, i.e. ÐC2B2C3 = 90°.

Similarly, ÐC2 A 2C3 = 90°, and we already know that
ÐC2C1C3 = 90°.
So, by the converse of Thales’ theorem (three times!) we learn that
the points B2, A2, and C1, all lie on a circle with C2C3 as diameter.
To put it another way, the (unique) circle which passes through
the mid-points of the sides, A2, B2, and C2, also passes through the
points C1 and C3.
The same argument (with different starting-points) then shows
that it must also pass through B1 and B3 and A1 and A3, establishing
the nine-point circle.

chapter 20: ceva’s theorem

Ceva’s theorem is in the same spirit as a much earlier one, due to

Menelaus of Alexandria, in about ad 100.

Menelaus’ theorem
This concerns a triangle ABC which has its sides cut by a straight
line, as in Fig. 234.
 OUP CORRECTED PROOF – FINAL, 30/07/20, SPi

256 not e s

C
b1 a2 a1 b1 c1
— . — . — =1
a2 b2 c2
b2
a1
A B c2
c1

Fig. 234 Menelaus’ theorem.

The simplest proof I know involves drawing the dashed line

though B parallel to AC. Then, by one pair of similar triangles,
its  length is b1a1 , and by another, b2c 2 , from which the result
a2 c1
follows.

Ceva’s theorem; another proof

Ceva’s theorem can, in fact, be proved from Menelaus’ theorem.

c1 b2
e

c2 b1
d
a1 a2 Fig. 235 Proving Ceva’s theorem.

With reference to Fig.  235, two applications of Menelaus’

theorem give
d c1 ( a1 + a 2 )
× × =1
e c2 a2

and
d b2 ( a1 + a 2 )
× × = 1,
e b1 a1

from which the result follows.

 OUP CORRECTED PROOF – FINAL, 30/07/20, SPi

not e s 257

And, as our original proof of Ceva’s theorem was by means of

area (Chapter  20), while we have just proved Menelaus’ theorem
using similar triangles, the above proof provides yet another
example of the link between the two ideas.

chapter 21: a kind of symmetry

Two results from trigonometry

Trigonometry can be especially useful when dealing with a general
triangle that has no particular symmetry or simplifying features.
In particular, two major results are

Area = 1 ab sin C
2

and the cosine rule

c 2 = a 2 + b2 - 2ab cos C

(see Fig. 236a).

b c b c
b sinC

C C
a a–b cosC
(a) (b)
Fig. 236 Triangle trigonometry.

The first of these is really just ‘ 1 base ´ height ’, while the second
2
comes from applying Pythagoras’ theorem to the right-angled
triangle in Fig. 236b with hypotenuse c.
 OUP CORRECTED PROOF – FINAL, 30/07/20, SPi

258 not e s

Heron’s theorem for the area

This can be established—after considerable algebra—by eliminating
the angle C between the results above, using Pythagoras’ theorem,
i.e. sin2 C + cos2 C = 1.

Incircle and circumcircle

Figure 237 gives a ‘picture proof’ for the radius of the incircle, by
simply viewing r as the height of the three constituent triangles, and
adding the areas.

1 1 1
a b Δ = — ar + — br + — cr
2 2 2
r
c
Fig. 237 Finding the radius of the incircle.

Figure 238 gives a picture proof for the radius of the circumcircle.

It uses the main circle theorem—‘the angle at the centre is twice
that at the circumference’.

θ
b 1
a Δ = — ab sinθ
2
R θθ Rsinθ = —
c
2 Fig. 238 Finding the radius of
c
the circumcircle.

Newton and the altitudes

This delightfully down-to-Earth proof that the altitudes meet at a
point can be found in The Mathematical Papers of Isaac Newton, vol. 4,
 OUP CORRECTED PROOF – FINAL, 30/07/20, SPi

not e s 259

p. 454 (Cambridge University Press, 2008), along with a substantial

historical footnote by the editor, D. T. Whiteside.

The eyeball theorem

In Fig.  239, triangles OQB and OTO′ are similar, so QB/TO′ =
OB/OO′.
If we let D = OO′ this gives

rr ¢
QB = ,
D

T
B
r′
r
O O′
Q

A
S

Fig. 239 The eyeball theorem.

The medians, by coordinate geometry

The formulae for the coordinates of a general point P on a line
(Fig.  132, p. 143) follow from similar triangles (Fig.  240), for then
l = ( x - x1 )/( x2 - x1 ) , with a similar result for the y-coordinates.

P2
P1P P1D
P λ= =
P1P2 P1E
P1
D E
Fig. 240 A point on a line.
 OUP CORRECTED PROOF – FINAL, 30/07/20, SPi

260 not e s

chapter 24: a soap solution

Much more about the geometry of soap films can be found in The
Science of Soap Films and Soap Bubbles by C.  Isenberg (Dover
Publications, 1992).

chapter 25: geometry in the ladies’ diary

A penny-farthing problem
Pythagoras applied to a certain right-angled triangle gives
( a + b )2 = ( a - b )2 + D2 ,

whence the result (Fig. 241).

a
a–b b

Fig. 241 Penny-farthing

D problem.

A tricky problem
The first solution to this problem appeared in 1731 and can be
found on p. 336 of vol. 1 of Hutton’s The Diarian Miscellany (1775). It
makes an appeal to a ‘curious theorem’ by Ozanam, and entirely
misses the neat, symmetric form of the final answer.
A second solution appears on p. 187 of vol. 1 of Thomas
Leybourn’s The Mathematical Questions Proposed in the Ladies’ Diary
(1817). The neat, symmetric form of the answer is given, together
with a generalization of the original problem, but it appeals to
some sophisticated theorems from advanced trigonometry.
 OUP CORRECTED PROOF – FINAL, 30/07/20, SPi

not e s 261

I have managed to construct a solution involving only elemen-

tary trigonometry. It begins by using some Pythagoras and similar
triangles to establish the results in Fig. 242.

R–r
sin θ =
R+r
2√Rr
cos θ = r R
R+r θ

Fig. 242 Nested circles.

The three θ-like angles in the actual problem—say α, β, γ—have

the property that α + β + γ = 90° (because the angles of a triangle add
up to 180°), and I use this to eliminate them. The most effective way of
doing this that I could see using elementary trigonometry is through

cos (a + b ) = cosa cos b - sina sin b = sin g ,

but even that leads to some fairly tedious algebra before a quadratic
equation for R emerges, and the neat, symmetric form of the
answer:

R = ab + bc + ca .

While the whole approach is strictly elementary, it is rather

scrappy, and I cannot help thinking that there must be a better way.

chapter 26: what euclid did

SSS congruence
Euclid’s proof of SSS congruence, Prop. I.8, is based on the idea that
SSS is enough to define a triangle uniquely, proved in Prop. I.7, and the
proof of that is based on Fig 243, where he assumes (for contradiction)
that the ‘third’ point could be in two different places, C and C′.
 OUP CORRECTED PROOF – FINAL, 30/07/20, SPi

262 not e s

C
C′

Fig. 243 Euclid’s proof of SSS

A B congruence.

There are then two isosceles triangles present. So, to paraphrase

Euclid slightly:

ÐACC¢ = ÐAC¢C < ÐBC¢C = ÐBCC¢,

which implies that ÐACC¢ < ÐBCC¢ , which is absurd.

ASA congruence
His proof of ASA congruence is again by contradiction, and in
much the same spirit as his proof of the converse of the isosceles
triangle theorem in Fig. 55, p. 63.

C
C′
D

A B A′ B′
Fig. 244 Euclid’s proof of ASA congruence.

The essence is as follows. Suppose the ASA conditions allow two

different triangles, so that, for instance, AC > A′C′ in Fig. 244. Then
choose the point D so that AD = A′C′, and draw BD.
The triangles DAB and C′A′B′ will then be congruent by SAS, so
ÐABD = ÐA¢B¢C¢. So ÐABD = ÐABC , which is absurd.
 OUP CORRECTED PROOF – FINAL, 30/07/20, SPi

not e s 263

chapter 30: when geometry goes wrong . . .

For more about the 1897 episode in Indiana, see David Singmaster’s
article ‘The Legal Values of Pi’ in Pi: A Source Book, edited by
L. Berggren, J. Borwein, and P. Borwein, pp. 236–9 (Springer, 1997).
That book includes, not surprisingly, much more about π, includ-
ing (i) an extensive chronology (pp. 288–305), (ii) Lambert’s 1761
paper proving irrationality (pp. 129–40), and (iii) Lindemann’s
paper proving that π is transcendental (pp. 194–206).

chapter 31: new angles on geometry

Miquel’s theorem
In Fig. 245, let two of the circles meet at P.
By the circle theorem shown in Fig. 63 (p. 71) we have
ÐA¢PB¢ = 180° - c and ÐB¢PC¢ = 180° - a , from which it follows that
ÐA¢PC¢ = a + c .
But from triangle ABC we have a + c + b = 180°, so that
ÐA¢PC¢ + b = 180°. It then follows, from the converse of the circle the-
orem just quoted, that the four points A′PC′B lie on a circle.

A
a C′

P
b B

B′ c
A′
C
Fig. 245 Proving Miquel’s theorem.
 OUP CORRECTED PROOF – FINAL, 30/07/20, SPi

264 not e s

Common chords
In Fig. 246, let two of the common chords, AB and CD, meet at P.
Extend EP to meet circle 1 at F′ and circle 3 at F′′.
Now apply the intersecting chords theorem (Fig. 64, p. 72) three
times, to circles 1, 2, and 3 in turn:

EP.PF¢ = CP.PD = AP.PB = EP.PF¢¢.

So PF¢ = PF¢¢, and therefore both F′ and F′′ must, in fact, be the
point F.
So P lies on the chord EF.

A 3
2
E
D
P

C
F′
B F′′
F
1
Fig. 246 Common chords.
 OUP CORRECTED PROOF – FINAL, 31/07/20, SPi

Further Reading

general

The Penguin Dictionary of Curious and Interesting Geometry, by David

Wells (Penguin, 1991). (A true miscellany, and great for dipping
into on a cold winter’s evening.)
A Square Peg in a Round Hole; Adventures in the Mathematics of Area, by
Chris Pritchard (Mathematical Association, 2016).
Icons of Mathematics, by C.  Alsina and R.  B.  Nelsen (Mathematical
Association of America, 2011).
The Changing Shape of Geometry, ed. Chris Pritchard (Cambridge
University Press, 2003). (Remarkable collection of essays.)
The Parsimonious Universe, by S.  Hildebrandt and A.  Tromba
(Copernicus, 1996). (Beautifully illustrated account of maximum
and minimum problems.)
The Pythagorean Theorem, by Eli Maor (Princeton University Press,
2007).

the history of geometry

Geometry by its History, by A. Ostermann and G. Wanner (Springer, 2012).

Mathematical Thought from Ancient to Modern Times, by Morris Kline
(Oxford University Press, 1972).

euclid’s elements

The classic edition, originally dating from 1903, is:

The Thirteen Books of Euclid’s Elements, trans. Sir Thomas Heath (Dover,
1956).
 OUP CORRECTED PROOF – FINAL, 31/07/20, SPi

266 f u rt h e r r e a di ng

It includes a great deal of scholarly commentary by Heath.

A much more recent, and relatively compact edition, also based
on Heath’s translation, was published by Green Lion Press in 2002.
A striking, full-colour facsimile of Oliver Byrne’s extraordinary
1847 colour edition of the first six books of Euclid’s Elements was
published by Taschen in 2010.
In addition, there is a much admired website version by David
Joyce, at https://mathcs.clarku.edu/~djoyce/elements/elements.html
I have also found the following commentary on the Elements very
helpful:
Euclid—The Creation of Mathematics, by B. Artmann (Springer, 1999).

textbooks of geometry

When I was first learning the subject, many years ago, my principal
books were Teach Yourself Geometry, by. P. Abbott (English Universities
Press, 1948) and Elementary Geometry, by C. V. Durell (Bell, 1925).
The second part of Abbott’s book is especially concise, and an
excellent modern textbook in a similar spirit, but with many valu-
able exercises, is Crossing the Bridge, by Gerry Leversha (UK Mathematics
Trust, 2008). A more expansive treatment can be found in Geometry,
by Harold Jacobs (Freeman, 2003).
At a more advanced level—and especially if you enjoyed the
‘unexpected meetings’ of Chapter 19—I recommend:
The Geometry of the Triangle, by Gerry Leversha (UK Mathematics
Trust, 2013) as well as Clark Kimberling’s famous website
Encyclopedia of Triangle Centers at http://Faculty.evansville.edu/ck6/
encyclopedia/ETC.html.
For looking ahead further still, towards university-level differen-
tial geometry, I recommend:
Elementary Geometry, by John Roe (Oxford University Press, 1993).
 OUP CORRECTED PROOF – FINAL, 31/07/20, SPi

f u rt h e r r e a di ng 267

the teaching of geometry

Teaching and Learning Geometry, by Doug French (Continuum, 2004).

Mathematics for the Multitude? A History of the Mathematical Association,
by Michael Price (The Mathematical Association, 1994).
The History of the Geometry Curriculum in the United States, by Nathalie
Sinclair (Information Age Publishing, 2008).

and finally . . .

Five very different ways of looking far beyond the geometry in most
of this book:
Things to Make and Do in the Fourth Dimension, by Matt Parker
(Particular Books, 2014).
Gems of Geometry, by John Barnes (Springer, 2009).
Mathematics: The Science of Patterns, by Keith Devlin (Scientific
American Library, 1994).
Topology: A Very Short Introduction, by Richard Earl (Oxford
University Press, 2019).
Ideas of Space, by Jeremy Gray (Clarendon Press, Oxford, 1989).
OUP CORRECTED PROOF – FINAL, 31/07/20, SPi
 OUP CORRECTED PROOF – FINAL, 31/07/20, SPi

Acknowledgements

I wish to thank Richard Earl, Gerry Leversha, and Vicky Neale

for some very helpful comments on the manuscript. I am
also extremely grateful to Latha Menon, Jenny Nugee, and
Emma Slaughter of Oxford University Press for taking such
great care at all stages of production.
 OUP CORRECTED PROOF – FINAL, 31/07/20, SPi

Publisher’s Acknowledgements

Epigraph for Chapter 12:

Twelve (12) words from Decline and Fall by Evelyn Waugh
(Penguin Books 1937, 1962) Text Copyright © Evelyn Waugh,
1928
From Decline and Fall by Evelyn Waugh, copyright © 2012.
Reprinted by permission of Little, Brown and Company, an
imprint of Hachette Book Group, Inc.
 OUP CORRECTED PROOF – FINAL, 31/07/20, SPi

Picture Credits

2. ScienceCartoonsPlus.com
11. AF Fotografie / Alamy Stock Photo
12. The Bodleian Libraries, The University of Oxford,
MS.D’Orville 301, Fols.21v-22r
20. INTERFOTO / Alamy Stock Photo
25. E.F. Smith Collection, Kislak Center for Special
Collections, Rare Books and Manuscripts, University
of Pennsylvania
26b. Yale Babylonian Collection, YBC 7289
27. Granger Historical Picture Archive / Alamy Stock
Photo
33. akg-images / Pictures From History
34. Felix Bennett
43. Northcliffe Collection/ANL/Shutterstock
72. Z J Levensteins, Naval Ordnance Laboratory
85 (left). GL Archive / Alamy Stock Photo
89. From van Schooten, Exercitationum Mathematicorum
(1657)
93. Cambridge University Library
102a. GL Archive / Alamy Stock Photo
107 (right). The Picture Art Collection / Alamy Stock Photo
129. Sir Isaac Newton, Geometria curvilinea and Fluxions,
MS Add. 3963, p 54r Reproduced by kind permission
of the Syndics of Cambridge University Library
134. Image by George Rex
142 (right). Pictorial Press Ltd / Alamy Stock Photo
143 (right). Pictorial Press Ltd / Alamy Stock Photo
 OUP CORRECTED PROOF – FINAL, 14/08/20, SPi

272 Pic t u r e C r e di ts

144. Public Domain

145 (left). History and Art Collection / Alamy Stock
Photo
147. Public Domain
181. From the author’s private collection
182a. Lebrecht Music & Arts / Alamy Stock Photo
182b. Photo: Blackwell’s Rare Books
189. From James M. Wilson, An Autobiography,
Sidgwick and Jackson, 1932
190. The Mathematical Association
191. Public Domain
195. From the author’s private collection
196. Public Domain
197b. Public Domain
200 (right). The History Collection / Alamy Stock Photo

p. 12 (top) 19th era / Alamy Stock Photo

p. 43 (top, right) Copyrighted 1938. Associated Press.
2164996:0820PF
p. 43 (bottom, left) Wellesley College Archives, Library &
Technology Services
p. 136 From William Jones, ‘A New Introduction to
the Mathematics’, London, 1706
p. 137 (top) FALKENSTEINFOTO / Alamy Stock Photo
p. 178 (bottom) History and Art Collection / Alamy Stock
Photo

The publisher and author apologize for any errors or omis-

sions in the above list. If contacted they will be pleased to rec-
tify these at the earliest opportunity.
 OUP CORRECTED PROOF – FINAL, 31/07/20, SPi

Index

A of triangle 25, 37, 87, 99, 192, 257

AA similarity 55, 59, 67, 72, 244, preserved 38, 41, 98, 195, 241
256, 259 using calculus 112, 250
Al-Biruni 75 area-ratio theorem 130
alternate angles 6, 90 ASA congruence 51, 54, 218, 223,
alternate segment theorem 245 253, 262
altitudes 122, 132, 140 Association for the Improvement of
ancient Greece 18 Geometric Teaching 206
angles 5 Aubrey, J. 36
alternate 6
B
at the centre 68
at the circumference 68 Babington, J. 27
bisected 125, 127, 252, 254 Banach-Tarski paradox 238
corresponding 4 Barrow, Isaac 12, 108
in straight line 5 Benson, L. S. 208
measurement 5, 24 bisector
notation for 16 of angle 125, 252
obtuse 38 perpendicular 123, 252
on same arc 68 Bolyai, J. 233
opposite 6 Byrne, O. 178
right 6
C
sum of triangle 7, 119, 193, 203,
204, 210 calculus 108, 248
Archimedes 18, 86, 134, 215 cannonballs 91
area 24 Carroll, Lewis 198, 218
of circle 86 Cavalieri, B. 156
of rectangle 25 centre of gravity 134
of similar triangles 47 centroid 125, 253
 OUP CORRECTED PROOF – FINAL, 31/07/20, SPi

274 i n de x

Ceva’s theorem 129 conic section 94

converse 131 constructions 161
from Menelaus’ theorem 256 contradiction, see proof: by
Chinnery, G. 94 contradiction
chord 16 converse 61
circle 16 of Ceva’s theorem 131
angles on same arc 68, 178, 219, of circle theorem 161, 225
245, 254 of isosceles triangle theorem
area of 86 62, 243
four points on a 71, 225, 263 of Pythagoras’s theorem 61
intersecting chords 72, of Thales’ theorem 65, 153, 243
251, 264 Cooley, W. D. 209
main theorem 68, 77, 92, 258 coordinates 101
nine-point 127, 254 corresponding angles 4, 23, 190
packing 89 cosine rule 257
tangent to 73 cos θ 80
terminology 16 cyclic quadrilateral 71
theorems 68
circumcentre 123. 127, 253 D
circumcircle 123, 139, 258 Dandelin spheres 223
circumference 16, 86, 172 Descartes, R. 101
common chords theorem 226 diagonal of rectangle 48, 64
concurrence: Dido’s problem 150
altitudes 122, 124 differentiation 110
angle bisectors 125, 252 directrix 249
circles 225, 263 distance by coordinates 103
chords 226 Dodgson, C. 198, 207
Gergonne point 133 double-angle formulae 246
medians 125, 142
perpendicular bisectors 123, 252 E
Condit, A. 43, 241 Earth, measurement of 22, 75
cone 94 ellipse 94, 223
congruence 14, 52, 54 in Fermat problem 155
describing 54 and planets 97
see also ASA, SAS, SSS reflection property 96
 OUP CORRECTED PROOF – FINAL, 31/07/20, SPi

i n de x 275

equilateral triangle 16, 81, 90, 157, Goodwin, E. J. 217

161, 162, 220, 237 gradient, see slope
equilibria, multiple 168 gravitation 85, 99
Einstein, Albert 46, 175 Gutierrez, A 141
Eratosthenes 22
Euclid 9, 76, 114 H
Euclid’s Elements 9, 12, 36, 39, 180 half-angle formula 93
ASA congruence 185, 262 Heisel, C. T. 213
Barrow’s edition 12 Heron of Alexandria 52
Book I 85, 192 area formula 138, 258
criticism 208 reflection of light 52
Oliver Byrne’s edition 178 hexagon 88, 90, 198
postulates 182 Hobbes, T. 36, 113
SAS congruence 185, 186 Hofmann, J. E. 163
similarity and area 194 hyperbola 94
SSS congruence 185, 261
structure 180 I
teaching of 206 icosahedron 182
Eudoxus 181 incentre 125
Euler, L. 127, 137, 253 incircle 125, 139, 258
eyeball theorem 141, 259 inclined plane 84
infinite product 92
F infinite series 136
Fermat, P. 101, 154 Inspector Euclid investigates…
focal points of ellipse 95 106, 247
focus of parabola 110, 249 integration 112
fractal 237 intersecting chords theorem 71
irrational number 28, 137, 181,
G 214, 216
Galileo 84 isosceles triangle 14, 17, 62, 69,
garage door geometry 229 187, 218, 243
Gauss, C. F. 123
Gergonne point 133 K
Girard 232 Kepler, J. 91, 97
golden ratio 58, 227 Koch snowflake 237
 OUP CORRECTED PROOF – FINAL, 31/07/20, SPi

276 i n de x

L mirror image 15, 53

Möbius band 236
ladder problems 27, 45
Ladies’ Diary, The 171
N
keeping one’s head 171
penny-farthing problem 174, 260 Newton, Sir Isaac 38, 98, 110,
tricky problem 176, 260 140, 258
window problem 173 nine-point circle 127, 254
Lambert, J. 137, 216, 263 non-Euclidean geometry 233
Lecchi, A. 197
Leibniz, G. L. 110, 136 O
Levett, R. 206 octahedron 182
Lindemann, F. 216, 263 opposite angles 6, 11, 184
line: of cyclic quadrilateral 71
parallel 4 orthocentre 124, 253
straight 5, 10, 232
Lobatchevsky, N. 233 P
Loomis, E. S. 42 packing problems 89
Ludlam, W. 200 Pappus 188, 235
parabola 94, 101, 110, 249
M parallel lines 4, 189, 201, 202
Mach number 80, 82 and alternate angles 6, 190
Mahavira 45 and corresponding angles 4,
Malfatti’s problem 220 190
Malton, T. 114, 120 distance apart 201
Mankiewicz, Johanna 175 ‘same direction’ 202
Mathematical Association, parallel postulate 184, 189, 233
The 206, 212 alternatives 199
maximisation 111, 147, 150, 220 parallelogram 54
mechanics in geometry 134, 164 pedal triangle 127, 254
medians 125, 132, 142, 259 penny-farthing problem 174, 260
Menelaus’ theorem 255 Penrose, R. 227
midpoint theorem 56 pentagon 58, 182
minimisation 110, 147, 149, 154, 165 perpendicular 5
Miquel’s theorem 225, 263 bisectors 123, 252
 OUP CORRECTED PROOF – FINAL, 31/07/20, SPi

i n de x 277

perspective 234 in converse of Thales’

Philo of Byzantium 186 theorem 243
pi: in coordinate geometry 103
definition 86 in The Ladies’ Diary 172, 260
exact formula for 92 in trigonometry 80, 257
infinite product 92 measuring the Earth 225
infinite series 136 proof 31, 32, 34, 36, 46, 196, 241
irrational 137, 216, 263
link with imaginary numbers 137 Q
symbol for 136 quadrilateral 57, 70
value of 89, 137, 215
pizza theorem 121 R
planetary motion 97 radius 16
Plato 19 Recorde, R. 116
Platonic solids 182 rectangle 25, 40, 48, 64, 107, 173,
Playfair’s postulate 199 196, 247, 250, 251
point 10 reductio ad absurdum, see proof: by
Fermat 155 contradiction
Gergonne 133 reflection of light 52, 96, 110, 147,
polygon, regular 87, 159 160, 249
polyhedron 182 Regiomontanus 76
‘practical work’ 23, 210 regular polygon 87, 159
projective geometry 234 Riemann, B. 233
proof: right angle 6, 183
by contradiction 62, 208 road networks 164
by rotation 162 rotation 162
false 203, 204, 218 Royal Road to Geometry 114, 120, 231
importance of 2 rugby 77
Pyramid, Great 20 ruler and compasses 216
Pythagoras’ theorem 26
3-4-5 special case 27, 60, 117 S
Ann Condit’s proof 43, 241 SAS:
converse 61 congruence 14, 41, 51, 53, 63, 185,
Euclid’s proof 39 187, 188, 192, 218, 242, 252, 262
in China 34, 241 similarity 55, 56, 253
 OUP CORRECTED PROOF – FINAL, 31/07/20, SPi

278 i n de x

scale factor 55 superposition 185

secant-tangent theorem 74, 244 supersonic flow 79, 82
semicircle 1, 16, 65, 104, 150 symmetry 15, 64, 123, 125, 138, 143,
seven circles theorem 226 175, 177, 231, 261
Sherlock Holmes 44
shortest network 165 T
similar triangles 20, 45, 55 tangent 73, 79, 110
and golden ratio 59 secant-tangent theorem 74
describing 54 tetrahedron 182
link with area 48 Thales’ theorem 1, 14, 16, 63, 230
right-angled 20, 44, 46, 48 and Galileo 84
Thales and 20 by coordinate geometry 104
see also AA, SAS, SSS converse 65, 153, 243, 254, 255
Simpson, T. 145, 152, 160, 184, theorem:
193, 201 alternate segment 245
Simson, R. 183 angle bisector 125, 252
sin θ 80 Ceva’s 129
slope: circle 68, 71, 72
of a curve 109, 248, 250 common chords 226, 264
of straight line 102, 247 eyeball 141, 259
perpendicular lines 102 Girard’s 232
Smith, J. 215 intersecting chords 71
snowflake 237 isosceles triangle 14
soap films 164, 167, 260 Menelaus’ 255
sphere 88 Miquel’s 225, 263
packing 91 midpoint 56
spherical geometry 231 Pappus’s 235
square 25, 30 Pythagoras’, see Pythagoras’s
SSS: theorem
congruence 51, 62, 186, 252, secant-tangent 74, 244
253, 261 seven circles 227
similarity 55 Thales’, see Thales’ theorem
Steiner, J. 153 Varignon’s 57
squaring the circle 216 Viviani’s 157, 159
 OUP CORRECTED PROOF – FINAL, 31/07/20, SPi

i n de x 279

Thue, A. 91 cosine rule 257

tiling 227 double-angle formulae 246
topology 235
Torricelli, E. U
Fermat problem 155 up-and-over geometry 229
trumpet 113
triangle: V
altitudes 122, 140 Varignon’s theorem 57, 148
angle-bisectors 125 Viète, F. 92, 101
angle sum of 7, 119, 193, 203, viewing angle 76
204, 210 Viviani, V. 156
area of 25 volume:
equilateral 16 sphere 88
incentre 125 trumpet 113
isosceles 14
medians 125, 142 W
obtuse 38 Waldo, C. 217
perpendicular bisectors 123 Ward, J. 33, 117, 202, 209
right-angled 25, 26 Whitehead, A. N. 209
similar 20, 44 Wilson, J. M. 205, 208, 215
trigonometry 80
addition formulae 245, 261 Y
and altitudes 132 Young Mathematician’s Guide
and pi 93 33, 117
and soap films 166
and supersonic flow 83 Z
area of triangle 257 Zhou bi 35
 OUP UNCORRECTED PROOF – FIRST PROOF,

1089 AND ALL THAT

A Journey into Mathematics

-
David Acheson

‘Every so often an author presents scientific

ideas in a new way . . . Not a page passes with-
out at least one intriguing insight . . . Anyone
who is baffled by mathematics should buy it.
My enthusiasm for it knows no bounds.’ Ian
Stewart, New Scientist
‘An instant classic . . . an inspiring little mas-
terpiece.’ Mathematical Association of America
‘Truly inspiring, and a great read.’ Mathematics
Teaching
978-0-19-959002-5 | Paperback | £8.99
This extraordinary little book makes math-
ematics accessible to everyone. From very simple beginnings Acheson
takes us on a journey to some deep mathematical ideas. On the way, via
Kepler and Newton, he explains what calculus really means, gives a brief
history of pi, and introduces us to chaos theory and imaginary numbers.
Every short chapter is packed with puzzles and illustrated by world
famous cartoonists, making this is one of the most readable and imagina-
tive books on mathematics ever written.

0004902476.INDD 278 Dictionary: NOSD 7/25/2020 8:57:58 PM