Abels Proof

This page intentionally left blank

Abels Proof
An Essay on the
Sources and Meaning
of Mathematical
Unsolvability

Peter Pesic

The MIT Press

Cambridge, Massachusetts
London, England


c 2003 Massachusetts Institute of Technology

All rights reserved. No part of this book may be reproduced in any form by
any electronic or mechanical means (including photocopying, recording, or
information storage and retrieval) without permission in writing from the
publisher.

This book was set in Palatino by Interactive Composition Corporation and

was printed and bound in the United States of America.

Permission for the use of the figures has been kindly given by the follow-
ing: Dover Publications (figures 2.1, 3.1); Conservation Departmentale de
Musees de Vendee (figure 2.2); Department of Mathematics, University of
Oslo, Norway (figure 6.1); Francisco Gonzalez de Posada (figure 10.1, which
appeared in Investigacion y Ciencia, July 1990, page 82); Robert W. Gray (fig-
ure 8.3); Lucent Technologies, Inc./Bell Labs (figure 10.2); National Library
of Norway, Oslo Division (figure 10.3).
I am grateful to Jean Buck (Wolfram Research), Peter M. Busichio and
Edward J. Eckert (Bell Labs/Lucent Technology), Judy Feldmann, Chry-
seis Fox, John Grafton (Dover Publications), Thomas Hull, Nils Klitkou
(National Library of Norway, Oslo Division), Purificacion Mayoral (Investi-
gacion y Ciencia), George Nichols, Lisa Reeve, Yngvar Reichelt (Department
of Mathematics, University of Oslo), Mary Reilly, and Ssu Weng for their
help with the figures. Special thanks to Wan-go Weng for his calligraphy of
the Chinese character ssu (meaning thought) on the dedication page.

Library of Congress Cataloging-in-Publication Data

Pesic, Peter.
Abels proof: an essay on the sources and meaning of mathematical
unsolvability / Peter Pesic.
p. cm.
Includes bibliographical references and index.
ISBN 0-262-16216-4 (hc : alk. paper)
1. Equations, roots of. 2. Abel, Niels Henrik, 18021829. I. Title.

QA212 .P47 2003

512.9 4dc21 2002031991
for Ssu,
thoughtful and beloved
This page intentionally left blank
 Contents

Introduction 1

1 The Scandal of the Irrational 5

2 Controversy and Coefficients 23

3 Impossibilities and Imaginaries 47

4 Spirals and Seashores 59

5 Premonitions and Permutations 73

6 Abels Proof 85

7 Abel and Galois 95

8 Seeing Symmetries 111

9 The Order of Things 131

10 Solving the Unsolvable 145

viii Contents

Appendix A: Abels 1824 Paper 155

Appendix B: Abel on the General Form of an Algebraic
Solution 171
Appendix C: Cauchys Theorem on Permutations 175
Notes 181
Acknowledgments 203
Index 205
 Introduction

In 1824, a young Norwegian named Niels Henrik Abel pub-

lished a small pamphlet announcing a new mathematical
proof for an old problem. Few read his pamphlet or paid
much attention. Five years later, Abel died at age twenty-six,
just before his work started to receive wide acclaim. Devel-
oped by others, his insights became a cornerstone of mod-
ern mathematics. Yet, apart from mathematicians, his ideas
remain generally unknown.
This book tells the story of Abels problem and his proof.
The text contains a minimum of equations, so that the main
lines of the argument should be accessible to people who are
intrigued by ideas but feel uncomfortable with mathemat-
ical detail. Boxes develop arguments and give examples in
greater depth, but these can be skipped without guilt. The
appendixes go further still, including an annotated transla-
tion of Abels pamphlet. The notes give references and sug-
gestions for further reading.
Abels proof concerns the solution of algebraic equations,
that is, equations of the form a n x n + a n1 x n1 + + a 0 = 0,
where x is the variable whose solutions or roots we
are seeking and a n , a n1 , . . . , a 0 are constant coefficients. We
classify algebraic equations by n, the highest power. If n = 1,
2 Introduction

the equation is linear and x has one root. If n = 2, the equation

is quadratic, and we are taught in high school that the gen-
eral solutions for any quadratic 2 equation a 2 x 2 + a 1 x + a 0 = 0
a 1 a 4a 2 a 0
are given by x = 1
2a 2
, no matter what integers or
fractions a 2 , a 1 , a 0 might be. The quadratic equation is thus
solvable in radicals, as mathematicians say, meaning that
its solutions can be expressed in terms of radicals (in this case,
the square root, but in general meaning any root, such as a
cube root, a fifth root, etc.), combined with the four ordinary
algebraic operations (addition, subtraction, multiplication,
division). This is as far as one goes in high school, except per-
haps to learn that there are equations of still higher degree:
if n = 3, we get equations like x 3 + 2x 2 + x 1 = 0, called
cubic or of the third degree. Likewise, there are fourth-degree
equations (also called quartic) that involve x 4 , fifth-degree
equations (called quintic) that involve x 5 , and so on, so that an
nth-degree equation has powers of x up to and including x n .
For many people, the quadratic formula is a dim memory
at best, though most of us recall that a quadratic equation
has solutions that can be found from a certain formula. What
about cubic equations or quartic equations? It turns out that
they are also solvable in radicals. Their solutions are more
complex than the quadratic formula, but they are available
in books and used to be studied routinely in school texts
of the nineteenth century. In the case of the cubic equation,
the solutions involve cube roots of square roots. It all seems
very routine, even tedious; we expect that all equations have
solutions, no matter what their degree, but that the solutions
get more and more complex the higher the degree. So far, this
all appears very tame.
But there is a great surprise. In general, a quintic equation
is not solvable in radicals. That is, although there are certain
Introduction 3

special quintic equations that have solutions we can express

in terms of radicals, if we take the general equation of the
fifth degree, a 5 x 5 + a 4 x 4 + a 3 x 3 + a 2 x 2 + a 1 x + a 0 = 0, there are
infinitely many values of the coefficients for which there is no
way of expressing what values of x would solve this equation
in terms of some formula involving a finite number of square
roots, cube roots, fifth roots, and other algebraic expressions.
That is, there are values of x that will satisfy the given quintic
equation, but we cannot express them in any finite formula,
however complex, involving just roots and powers, added,
subtracted, multiplied, and divided in some way, as we could
for all equations up to the quintic. What is worse, the same
holds for equations of any degree higher than five: in general,
equations of the sixth, seventh, eighth, . . . , nth degree are not
solvable in radicals either.
Why? What happened to the pattern of algebraic equa-
tions each having solutions? What is it about the fifth degree
that causes the problem? Why does it then go on to affect all
higher degrees of equations? Most of all, what is the signifi-
cance of this breakdown, if one can use such a word?
Such questions have moved me since childhood. Mathe-
matical symbols may indicate hidden truths that have deep
human significance, even as they transcend the human.
Abels proof contains a prime secret: how can a search for so-
lutions yield the unsolvable? Perhaps if I tried hard enough,
I could understand. I studied modern texts, but the key re-
mained elusive. Absorbed in advanced studies, experts may
cease to wonder about the elementary. They might not notice
the kind of basic insight I was seeking. To find it, I needed to
return to the sources, retracing the journey recounted in this
book. The story begins in ancient Greece and has climactic
scenes in Norway and France in the 1820s. What Abel found
is indeed surprising and strangely beautiful.
This page intentionally left blank
1 The Scandal of the
Irrational

The story begins with a secret and a scandal. About 2,500

years ago, in Greece, a philosopher named Pythagoras and
his followers adopted the motto All is number. The
Pythagorean brotherhood discovered many important math-
ematical truths and explored the ways they were manifest in
the world. But they also wrapped themselves in mystery, con-
sidering themselves guardians of the secrets of mathematics
from the profane world. Because of their secrecy, many de-
tails of their work are lost, and even the degree to which they
were indebted to prior discoveries made in Mesopotamia and
Egypt remains obscure.
Those who followed looked back on the Pythagoreans as
the source of mathematics. Euclids masterful compilation,
The Elements, written several hundred years later, includes
Pythagorean discoveries along with later work, culminating
in the construction of the five Platonic solids, the only solid
figures that are regular (having identical equal-sided faces):
the tetrahedron, the cube, the octahedron, the dodecahedron,
and the icosahedron (figure 1.1). The major contribution of
the Pythagoreans, though, was the concept of mathematical
proof, the idea that one could construct irrefutable demon-
strations of theoretical propositions that would admit of no
 (a) (b)

(c) (d)

(e)

Figure 1.1
The five regular Platonic solids, as illustrated after Leonardo da Vinci
in Luca Pacioli, On the Divine Proportion (1509). a. tetrahedron, b. cube,
c. octahedron, d. dodecahedron, e. icosahedron.
The Scandal of the Irrational 7

exception. Here they went beyond the Babylonians, who, de-

spite their many accomplishments, seem not to have been in-
terested in proving propositions. The Pythagoreans and their
followers created mathematics in the sense we still know
it, a word whose meaning is the things that are learned,
implying certain and secure knowledge.
The myths surrounding the Pythagorean brotherhood hide
exactly who made their discoveries and how. Pythagoras
himself was said to have recognized the proportions of sim-
ple whole-number ratios in the musical intervals he heard
resounding from the anvils of a blacksmith shop: the octave
(corresponding to the ratio 2:1), the fifth (3:2), the fourth (4:3),
as ratios of the weights of the blacksmiths hammers. This re-
vealed to him that music was number made audible. (This
is a good point to note an important distinction: the modern
fraction 32 denotes a breaking of the unit into parts, whereas
the ancient Greeks used the ratio 3:2 to denote a relation
between unbroken wholes.) Another story tells how he sac-
rificed a hundred oxen after discovering what we now call
the Pythagorean Theorem. These stories describe events that
were felt to be of such primal importance that they demanded
mythic retelling.
There is a third Pythagorean myth that tells of an unfore-
seen catastrophe. Despite their motto that all is number,
the Pythagoreans discovered the existence of magnitudes
that are radically different from ordinary numbers. For in-
stance, consider a square with unit side. Its diagonal cannot
be expressed as any integral multiple of its side, nor as any
whole-number ratio based on it. That is, they are incommen-
surable. Box 1.1 describes the simple argument recounted
by Aristotle to prove this. This argument is an example of
a reductio ad absurdum: We begin by assuming hypothetically
that such a ratio exists and then show that this assumption
8 Chapter 1

Box 1.1
The diagonal of a square is incommensurable with its side

S 1

Let the square have unit side and a diagonal length s. Then
suppose that s can be expressed as a ratio of two whole
numbers, s = m:n. We can assume further that m and n are
expressed in lowest terms, that is, they have no common
factors. Now note that s 2 = m2 :n2 = 2:1, since the square on
the hypotenuse s is double the square on the side, by the
Pythagorean theorem. Therefore m2 is even (being two times
an integer), and so too is m (since the square of an even num-
ber is even). But then n must be odd, since otherwise one
could divide m and n by a factor of 2 and simplify them
further. If m is even, we can let m = 2 p, where p is some
number. Then m2 = 4 p 2 = 2n2 , and n2 = 2 p 2 . But this means
that n2 is even, and so too is n. Since a whole number cannot
be both even and odd, our original assumption that s = m:n
must be wrong. Therefore the diagonal of a square cannot be
expressed as a ratio of two whole numbers.

leads to an absurdity, namely that one and the same num-

ber must be both even and odd. Thus the hypothesis must
have been wrong: no ratio can represent the relation of di-
agonal to side, which is therefore irrational, to use a modern
term.
The Scandal of the Irrational 9

The original Greek term is more pungent. The word for

ratio is logos, which means word, reckoning, account, from
a root meaning picking up or gathering. The new magni-
tudes are called alogon, meaning inexpressible, unsayable.
Irrational magnitudes are logical consequences of geometry,
but they are inexpressible in terms of ordinary numbers, and
the Greeks were careful to use entirely different words to
denote a number (in Greek, arithmos) as opposed to a magni-
tude (megethos). This distinction later became blurred, but for
now it is crucial to insist on it. The word arithmos denotes the
counting numbers beginning with two, for the unit or the
One (the Greeks called it the monad) was not a number in
their judgment. The Greeks did not know the Hindu-Arabic
zero and surely would not have recognized it as an arithmos;
even now, we do not hear people counting objects as zero,
one, two, three, . . . . Thus, the expression there are no ap-
ples here means there arent any apples here more than
there are zero apples here.
It was only in the seventeenth century that the word
number was extended to include not only the counting
numbers from two on, but also irrational quantities. Ancient
mathematicians emphasized the distinctions between differ-
ent sorts of mathematical quantities. The word arithmos prob-
ably goes back to the Indo-European root (a)r, recognizable
in such English words as rite and rhythm. In Vedic India, r.ta
meant the cosmic order, the regular course of days and sea-
sons, whose opposite (anr.ta) stood for untruth and sin. Thus,
the Greek word for counting number goes back to a con-
cept of cosmic order, mirrored in proper ritual: certain things
must come first, others second, and so on. Here, due order is
important; it is not possible to stick in upstart quantities like
one-half or (worse still) the square root of 2 between the
integers, a word whose Latin root means unbroken or whole.
10 Chapter 1

The integers are paragons of integrity; they should not be

confused with magnitudes, which are divisible.
At first, the Pythagoreans supposed that all things were
made of counting numbers. In the beginning, the primal
One overflowed into the Two, then the Three, then the Four.
The Pythagoreans considered these numbers holy, for
1 + 2 + 3 + 4 = 10, a complete decade. They also observed
that the musical consonances have ratios involving only the
numbers up to four, which they called the holy Tetractys.
Out of such simple ratios, they conjectured, all the world
was made. The discovery of magnitudes that cannot be
expressed as whole-number ratios was therefore deeply dis-
turbing, for it threatened the entire project of explaining na-
ture in terms of number alone. This discovery was the darkest
secret of the Pythagoreans, its disclosure their greatest scan-
dal. The identity of the discoverer is lost, as is that of the one
who disclosed it to the profane world. Some suspect them
to have been one and the same person, perhaps Hippias of
Mesopontum, somewhere around the end of the fifth century
B.C., but probably not Pythagoras himself or his early fol-
lowers. Where Pythagoras had called for animal sacrifice to
celebrate his theorem, legend has it that the irrational called
for a human sacrifice: the betrayer of the secret drowned at
sea. Centuries later, the Alexandrian mathematician Pappus
speculated that

they intended by this, by way of a parable, first that everything

in the world that is surd, or irrational, or inconceivable be veiled.
Second, any soul who by error or heedlessness discovers or reveals
anything of this nature in it or in this world wanders on the sea of
nonidentity, immersed in the flow of becoming, in which there is no
standard of regularity.

Those who immerse themselves in the irrational drown

not by divine vengeance or by the hand of an outraged
The Scandal of the Irrational 11

brotherhood but in the dark ocean of nameless magnitudes.

Ironically, this is a consequence of geometry and the
Pythagorean Theorem itself. When Pythagoras realized that
the square on the hypotenuse was equal to the sum of the
squares on the other two sides, he was very close to the
further realization that, though the squares might be com-
mensurable, the sides are not. Indeed, the argument given in
box 1.1 depends crucially on the Pythagorean Theorem. That
argument suggests that, had Pythagoras tried to express the
ratio of the diagonal of a square to its side, he would have
realized its impossibility immediately. He probably did not
take this step, but his successors did.
The discovery of the irrational had profound implications.
From it, Pappus drew a distinction between such contin-
uous quantities and integers, which progressing by de-
grees, advance by addition from that which is a minimum,
and proceed indefinitely, whereas the continuous quantities
begin with a definite whole and are divisible indefinitely.
That is, if we start with an irreducible ratio such as 2:3,
we can build a series of similar ratios in a straightforward
manner: 2:3 = 4:6 = 6:9 = . But if there is no smallest
ratio in a series, there can be no ratio expressing the whole.
Pappuss words suggest that it was this argument that may
have opened the eyes of the Pythagoreans. Consider again
the diagonal and side of a square. The attempt to express
both of them as multiples of a common unit requires an in-
finite regress (box 1.2). However small we take the unit, the
argument requires it to be smaller still. Again we see that no
such unit can exist.
The challenge of Greek mathematics was to cope with two
incommensurable mathematical worlds, arithmetic and ge-
ometry, each a perfect realm of intelligible order within itself,
but with a certain tension between them. In Platos dialogues,
12 Chapter 1

Box 1.2
A geometric proof of the incommensurability of the diagonal
of a square to its side, using an infinite regress:

A E G B

H
F

D C

In the square ABCD, use a compass to lay off DF = DA along

the diagonal BD. At F, erect the perpendicular EF. Then
the ratio of BE to BF (hypotenuse to side) will be the same
as the ratio of DB to DA, since the triangles BAD and EFB are
similar. Suppose that AB and BD were commensurable. Then
there would be a segment I such that both AB and BD were
integral multiples of I. Since DF = DA, then BF = BD DF is
also a multiple of I. Note also that BF = EF, because the sides
of triangle EFB correspond to the equal sides of triangle BAD.
Further, EF = AE because (connecting D and E) triangles EAD
and EFD are congruent. Thus, AE = BF is a multiple of I. Then
BE = BA AE is also a multiple of I. Therefore, both the side
(BF) and hypotenuse (BE) are multiples of I, which therefore
is a common measure for the diagonal and side of the square
of side BF. The process can now be repeated: on EB lay off
EG = EF and construct GH perpendicular to BG. The ratio of
hypotenuse to side will still be the same as it was before and
hence the side of the square on BG and its diagonal also share
I as a common measure. Because we can keep repeating this
process, we will eventually reach a square whose side is less
than I, contradicting our initial assumption. Therefore, there
is no such common measure I.
The Scandal of the Irrational 13

this challenge elicits deep responses, reaching out from math-

ematics to touch emotional and political life. A pivotal mo-
ment occurs in the dialogue between Socrates and Meno, a
visiting Thessalian magnate who was a friend and ally of
the Persian king. Meno was notoriously amoral, a greedy
and cynical opportunist. Strangely, on his last day in Athens
he asks Socrates over and over again whether virtue can be
taught or comes naturally. Their conversation turns on the
difference between true knowledge and opinion.
At the heart of their discussion, Socrates calls for a slave
boy, with whom he converses about how to double the area
of a square of a given size. Unlike Meno, the boy is inno-
cent and frank; he confidently expresses his opinion that if
you double the side of a square, you double its area. Their
conversation is a perfect example of Socrates practice of phi-
losophy through dialogue. As they talk, the boy realizes that
a square of double side has four times the area, which leaves
him surprised and perplexed. The Greek word for his situa-
tion is aporia, which means an impasse, an internal contradic-
tion. Just before this conversation, Socrates questioning had
revealed contradictions in Menos confident opinions about
virtue, and Meno had lashed out angrily. Socrates, he said,
was like an ugly stingray that harms his victims and renders
them helpless. Socrates answer is to show how well the slave
boy could take being stung. The boy is amazed and curious,
not angry. He readily follows Socrates lead in drawing a new
picture (box 1.3). In a few strokes, the real doubled square
emerges by drawing the diagonals within the boys fourfold
square. Responding freely to Socrates suggestions, the boy
grasps this himself. Meno is forced to admit that the sting of
realizing his ignorance did not harm the boy, who replaced
his false opinion with a true one. The dialogue ends with
Meno smoldering, foreshadowing the angry Athenians who
14 Chapter 1

Box 1.3
Socrates construction of the doubled square in the Meno

A
H E

O
D B

G F
C

Let the original square be AEBO. The slave boy thought that
the square on the doubled side HE would have twice the
area, but realized that in fact that HEFG has four times
the area of AEBO. At Socrates prompting, he then draws the
diagonals AB, BC, CD, DA within the square HEFG. Each
triangle AOB = BOC = COD = DOA is exactly half the area
of the original square, so all four of these together give the
true doubled square ABCD.

later voted to execute the philosopher. Their outrage points to

the power of the new mathematical insights. Though Socrates
had not referred to the irrationality of the diagonal, it was cru-
cial. The process of doubling a square (an eminently rational
enterprise) required recourse to the irrational, a fact that was
not lost on Plato or his hearers.
The Scandal of the Irrational 15

Though a consequence of logical mathematics, clearly the

word irrational already had acquired the emotional conno-
tations it still retains. In Platos Republic, Socrates jokes that
young people are as irrational as lines and hence not yet
suited to rule in the city and be the sovereigns of the great-
est things. Appropriately and yet ironically, Socrates pre-
scribes mathematics, along with music and gymnastics, for
these young irrationals to tame what is most disorderly and
incommensurate in their souls. His joke points to a widely
held sense that irrationality in mathematics was a troubling
sign of confusion and disorder in the world, a danger as fear-
ful as drowning. Certainly the Pythagoreans took this dire
view, but Platos dialogues open a larger perspective. What
is irrational, in the soul or in mathematics, may be harmo-
nized with the rational; to use an unforgettable image from
another dialogue, the black horse of passion may be yoked
to the white horse of reason.
Platos great dialogue on the nature of knowledge rests on
this mathematical crux. It is named after Theaetetus, a math-
ematician who is introduced to us as he is being carried back
to Athens, dying from battlefield wounds and dysentery. In
a flashback to Theaetetus youth, we learn that he made pro-
found discoveries about the irrationals and the five regu-
lar solids, and he conversed with Socrates shortly before the
philosophers trial and death. Socrates was deeply impressed
with this youth, who seemed destined to do great things and
who also resembled Socrates physically, down to the snub
nose and bulging eyes. Also present during their conversa-
tion was Theaetetus teacher Theodorus, an older
mathemati-

cian who had proved the irrationality of 3, 5, 7, . . ., all
the way to 17, where for some reason he stopped.
Socrates characteristic irony is not in evidence as he ques-
tions Theaetetus, who explains his discovery that there are
16 Chapter 1

degrees of irrationality. Though such magnitudes as the

square roots of 3 or of 17 are irrational, they are still com-
mensurable in square, since
2 their squares
 2 have a common
measure (that is, since 3 = 3 and 17 = 17 are both in-
tegers). Socrates is struck with the truth and beauty of these
insights and uses them as examples that lead to a broader
conversation about the nature of knowledge. He reminds
Theaetetus and Theodorus that he has a reputation as one
who stings by inducing perplexity and jokingly asks
Theaetetus not to denounce him as an evil wizard, explaining
that he is really a midwife who helps people deliver them-
selves of their conceptions.
Anticipating his indictment on the very next day, Socrates
justifies himself not to his angry accusers but to this gentle
and gifted young man, so much like himself. Far from feeling
antagonistic, Theaetetus is ready to enter a searching inquiry
that begins with mathematics as a touchstone of true knowl-
edge, testing whether other knowledge comes through the
senses or more mysteriously from within the soul. Though
he depicts himself as sterile, barren of wisdom, Socrates helps
Theaetetus bring his conception to birth and tests its health.
Socrates had often made fun of his own ugly features, but he
describes Theaetetus as beautiful. Theaetetus mathematical
insight is commensurate with the bravery that will allow him
to fight for his city and die with exceptional honor. Such is
the courage of one who could wrestle with the irrational.
During their conversation, Socrates encourages his guests
to put themselves to torture, by which he means that they
should struggle fearlessly to test and refine their opinions
together. In Greek, the word for torture can also mean
the touchstone, a mineral that is able to distinguish gold
from base metal by the mark each makes on it. This extreme
metaphor has overtones of the judicial torture used to coerce
The Scandal of the Irrational 17

truth from slaves, but Socrates uses it to signify a search for

truth that defies even intense pain and humiliation. Like sol-
diers or athletes, Socrates and Theaetetus see in suffering the
path to the superlative pleasure of ultimate truth. This they
have learned from mathematics, whose study often seems
painful to those who do not know the pleasure of insight. It is
no wonder that Plato placed over the door to his academy the
admonition: Let no one ignorant of geometry enter here.
The discoveries of Theaetetus and the test of mathemati-
cal proof were enshrined in Euclids Elements, which remains
even today a living fountainhead of mathematics, invaluable
for beginners as well as experienced mathematicians. Be-
yond presenting his own results, Euclid set the discoveries of
others in order as a touchstone of mathematical lucidity and
logical force. In the case of the irrational, Euclid drew on
a compromise introduced by Eudoxus, who kept numbers
and irrational magnitudes strictly apart, but yet in propor-
tion. For instance, Euclid considers two numbers in a certain
ratio (say 2:3) and shows how this proportion could be equal
to that between two irrational magnitudes (as 2 2:3 2 is
equal to 2:3). Nevertheless, he would never mix the two dis-
tinct types so as to allow a ratio between a number and a
magnitude. This was not simply mathematical apartheid but
a decision to consider numbers and magnitudes as entirely
distinct genera, whose mixing might lead to incalculable
confusion.
Euclids contribution went far beyond this separation of
realms. In Book V, he introduces a far-reaching definition
of equality or inequality that extends to ratios of irrational
magnitudes. Following Eudoxus, he proposes that if we want
to check whether two ratios are equal, we should multiply
the terms by various integers to check whether these mul-
tiples are respectively greater, equal, or less (box 1.4). This
18 Chapter 1

Box 1.4
Euclids definition of equal ratio, which is applicable to any
magnitude (Book V, Definition 5)

The ratio a :b is said to be equal to the ratio c:d if, for any
whole numbers m and n, when ma is compared with nb and
mc is compared with nd, the following holds: if ma > nb,
then mc > nd; if ma = nb, then mc = nd; and if ma < nb, then
mc < nd.

definition of equality still depends on testing multiples of

magnitudes, even though the magnitudes themselves may
have no common measure. It also uses any multiples what-
ever, as if to examine all possible multiples in order to de-
termine whether the multiplied ratios could ever be equal.
Thus, it is really a test, a trial by multiplication, a way to nav-
igate the irrational sea. Euclid puts it in strong contrast with
the way he treats whole numbers in Books VI and VII, for
integers are commensurable because they have a common
unit.
Euclids most daring inquiry into the irrational occurs in
Book X, which asks: Do the irrational magnitudes have some
intelligible order? Can one classify them into clear categories
by genus and species? He begins by showing that any magni-
tude can be indefinitely divided. Though implicit in geome-
try, he brings into prominence what later came to be called the
continuum, meaning a continuously and endlessly divisible
magnitude, as opposed to the indivisible One, whose inte-
gral multiples constitute all the counting numbers. To show
this indefinite divisibility, Euclid demonstrates how we can
successively subtract from any magnitude half or more of
that magnitude, and then keep repeating this process until
The Scandal of the Irrational 19

Box 1.5
Euclids statement of the indefinite divisibility of any magni-
tude (Book X, Proposition 1)

Take half (or more) of the given magnitude, and then the
same proportion of what remains, and the same proportion
yet again of what remains, continuing the process as far as
necessary so that the remainder can be made less than any
given line.

finally we have left a magnitude that is smaller than any given

amount (box 1.5). Thus, there is no smallest magnitude, no
geometrical atom or least possible magnitude making up
all others, for if there were, all magnitudes would share that
smallest magnitude as their common measure. Here again,
Euclid sets in play a process indefinitely repeated, not pic-
turable in a single figure but intelligible and logically com-
pelling, nonetheless.
Then Euclid sets out to classify different kinds of irra-
tionals, naming them and showing their interrelations. As
Theaetetus had shown, irrationality is a relative term. The
diagonal of a square is irrational compared to the side, but
it can be commensurable with another line, which might be
the side or diagonal of another square. What is speakable
depends above all on the relation between figures. Euclids
classification of irrationals is intricate, though it does not go
beyond what we would call the square root of the sum or
difference of two square roots. He identifies such quantities
in the division of a geometrical line, but we can also divide
a string to make it sound different intervals. This means we
can formulate a musical version of the mathematical crisis
of the irrational. If we try to divide an octave (whose ratio
20 Chapter 1

Box 1.6
The sound of square roots

Take two strings, one sounding an octave higher than the

other, so that their lengths are in the ratio 2:1. Then find the
geometric ratio (also called the mean proportional) between
these strings, the length x at which 2:x is the same proportion
as x:1. This means that 2:x = x:1; cross-multiplying
this gives
x 2 = 2. Thus, the ratio needed is 2:1 1.414, in modern
decimals. This is close to the dissonant interval called the
tritone, which later was called the devil in music, namely
the interval composed of three equal whole steps each of ratio
9:8. The tritone is thus 9:8 9:8 9:8 = 93 :83 = 729:512
1.424.

is 2:1) exactly at thepoint of the geometric mean, we get

the mongrel ratio 2:1 (box 1.6). This is very close to the
highly dissonant interval later called the devil in music,
the tritone. If the whole universe is based on number, such
harmonic problems are critical.
Euclid presents his classification of irrationals through a
hundred careful propositions. After this tour de force, he
says something amazing in the final proposition of the book:
From the lines already drawn, one can go on to define still
other irrational lines that are infinite in number, and none
of them is the same as any of the preceding. Although his
tone is impassive, this is a portentous statement. The realm
of the irrational is infinite not just because there are an un-
limited number of irrational magnitudes of each type but
even more because there is an infinite variety of kinds of such
magnitudes, each a different species with infinitely many ex-
amples. The discovery of the irrational disclosed an infinitely
branching path.
The Scandal of the Irrational 21

Euclids impassive tone does not disclose what he thought

of this situation. By this final proposition, Euclid could have
meant to indicate a disturbing glance into the irrational abyss,
as if to say: Here lies an unfathomable, trackless sea of end-
lessly different magnitudes, from which one should turn
away in horror. But there is another possible reading of his si-
lence. He might have meant: Here lies an inexhaustible store
of treasures, infinite in number though each is finite in mag-
nitude. Behold, and wonder.
This page intentionally left blank
2 Controversy and
Coefficients

In his attempt to survey the irrational magnitudes, Euclid

was laying out possible solutions of what we now call the
quadratic equation. But that statement distorts his own way
of understanding these matters and thus is seriously mis-
leading. The ancient Greek mathematicians did not know the
word algebra, or the symbolic system we mean by it, even
though many of their propositions can be translated into al-
gebraic language. Euclid kept numbers rigorously separate
from lines and magnitudes. He might have been horrified by
the way algebra mixes rational and irrational, numbers and
roots. So, as we prepare to consider the meaning of equations,
we need to pause and assess the significance of the mathemat-
ical revolution reflected in the rise of algebra as we know it.
The word algebra is Arabic, not Greek, and refers to
the joining together of what is broken. Thus, in Don Quixote,
Cervantes calls someone who sets broken bones an algebrista.
The original Arabic algebraists created recipes for solving
different kinds of problems, usually involving whole num-
bers. They worked by presenting examples with solutions,
but without the methodical symbolism we now associate
with algebra and without the systematic processes of solution
that are equally important. They did recognize the logical
24 Chapter 2

force of geometry and tried to illustrate their solutions with

geometric examples, but they could not go far without the
symbols that might have enabled them to speak with more
generality.
What is surprising, then, is that some of the greatest ad-
vances in solving algebraic problems came long before such
powerful symbols were devised. In the case of the quadratic
equation, the Babylonians were able to solve many similar
problems four thousand years ago. Notably more advanced
than the Egyptians, the Babylonians made difficult calcu-
lations using a notation essentially equivalent to modern
place-value, but based on the number sixty (box 2.1). We
still use sexagesimal notation in our measurements of time
(60 minutes in the hour) and of angle (60 minutes of arc
in a degree). Such a notation of place-value was unknown
in Europe until the Renaissance, and the rich achievements
of Babylonian mathematics came fully to light only in the
twentieth century. When they considered a problem, the
Babylonians wrote it out and solved it in words. With no
symbolic notation for unknowns, they could not express the
general solution of an equation with arbitrary values, though
they could solve specific examples, as in box 2.2.
Similar statements could be made about the methods of
the Arabic algebraists, whose principal representative to the

Box 2.1
Babylonian sexagesimal (base 60) notation

The Babylonian notation for our decimal number 870 can

be written as 14, 30 = 14(60)1 + 30(60)0 ; the notation for the
modern fraction 0.5 would be 0; 30 = 30
60
= 0(60)0 + 30(60)1 .
Controversy and Coefficients 25

Box 2.2
Babylonian solution of equations

Here is an example of a Babylonian solution of what we

would call the equation x 2 x = 870, with the numerical val-
ues translated into decimal notation: Take half of 1, which
is 0.5, and multiply 0.5 by 0.5, which is 0.25; add this to 870
to get 870.25. This is the square of 29.5. Now add 0.5 to 29.5,
and the result is 30, the side of the square. Compare this
with the modern solution of the equation a 2 x 2 + a 1 x +a 0 = 0,
with a 2 = 1, a 1 = 1, and a 0 = 870, namely x = a 1
   
2
a 12 4a 2 a 0 2a 2 gives x = 1
2
+ 870 + 12 , each term of
which corresponds to a step in the Babylonian solution.

West was Muhammed ibn-Musa al-Khwarizm. Writing in

the ninth century A.D., he sets forth al-jabr as an eminently
practical art, such as men constantly require . . . in all their
dealings with one another; he gives as instances the prob-
lems that emerge from inheritances, lawsuits, trade, land
measurement, and canal building. Like the Babylonians, al-
Khwarizm focuses on specific problems, but he adds that
it is necessary that we should demonstrate geometrically
the truth of the same problems which we have explained
in numbers. Box 2.3 shows his geometrical explanation of
one problem: to solve for the unknown root if a square
and 10 roots equal 39 units (x 2 + 10x = 39, as we would
say). His method is that of completing the square, though
al-Khwarizm confines it to this one example. Compare the
general expression of this method in modern symbols shown
in box 2.3. Shortly, we will consider the history and signifi-
cance of these symbols.
26 Chapter 2

Box 2.3
Al-Khwarizms geometric explanation of the solution to
quadratic equations

x 2
J K

e 2
A B

h f

D C
g

M L

Consider the proposition that a square and 10 roots equal

39 units (x 2 + 10x = 39, in modern notation). Let the central
square ABCD have area x 2 ; its side is thus x. Add to this
square four equal areas, labeled e, f, g, h, each of area 2 12 x, so
that together they have area 10x. Each of these is a rectangle
of sides x and 2 12 . Now complete the larger square JKLM
by including the little squares at each corner, whose area is
each 2 12 2 12 = 6 14 ; the total area of all four corner squares
is 4 6 14 = 25. We are given that the central square plus
the rectangles equals 39, so the large square JKLM has area
39 + 25 = 64 and hence side 8. But the side of the large square
is 2 12 + x + 2 12 = 5 + x = 8, so we conclude that x = 3.
In comparison,  the algebraic solution of a 2 x 2 + a 1 x + a 0 = 0
 
is x = a 1 a 1 4a 2 a 0 2a 2 . In this case, a 2 = 1, a 1 = 10,
2

a 0 = 39, so x = (10 100 + 156)/2 = (10 16)/2 = 3
or 13.
Controversy and Coefficients 27

As the algebraic writings of the Arabs were translated into

Latin, Europeans began to work in what they called the
great art, also called the art of the coss or cosa, meaning the
thing or unknown. The story of early modern algebra is
tumultuous; after long periods of inactivity, great strides
were made in a matter of years by a few individuals. These
mathematical innovations were often related to problems
raised by economic changes, as the Arabs work had also
been. During the centuries following the Crusades, new con-
ditions of commerce involving extensive traffic in commodi-
ties required widespread use of currency as a medium of
exchange and a new way of analyzing the flow of assets.
European goods ranged widely, traded for spices and silks.
As merchants strove for rapid increase of capital, they found
they needed new ways to deal with dispersed partners and
factors operating in diverse local economic situations. Marine
insurance emerged, as did bills of exchange in draft form
and modes of international credit. The ancient methods of
bookkeeping also broke down when, as Shylock describes
it, a trader has an argosy bound to Tripolis, another to the
Indies; I understand, moreover, upon the Rialto, he hath a
third at Mexico, a fourth for Englandand other ventures
he hath, squandred abroad.
In this new situation, double-entry bookkeeping was an in-
valuable innovation. The basis of all subsequent accounting
practice, this system requires a twofold listing of all debits
and credits, recorded in a single currency and checked by
balancing both the debit and the credit sides of the ledger.
The new bookkeeping was extremely awkward with Greek
or Roman numerals but flourished with the advent of Hindu-
Arabic numerals, which Leonardo of Pisa (better known as
Fibonacci) helped introduce to Europe in his Book of the
Abacus (1202). At this time, the Arabic word s.ifr gave birth
28 Chapter 2

to the English words zero and cipher, essential concepts

for reckoning with numbers as a symbolic code.
Fibonaccis book concerns problems in commercial trans-
actions, especially currency exchanges that involve fractions.
Such practical economic problems lead to more speculative
questions: How many pairs of rabbits will be produced in a
year, beginning with a single pair, if in every month each pair
bears a new pair that becomes productive from the second
month on? Considering such problems leads Fibonacci to
his celebrated sequence: 1, 1, 2, 3, 5, 8, 13, . . . , in which each
term is the sum of the two immediately preceding it. This se-
quence could describe the growth of capital as well as an in-
creasing population of rabbits. Later, the quantity now called
e = 2.718 . . . , the key to exponential growth, emerges during
a further consideration of compound interest. Fibonacci also
considers cubic equations, which he argues cannot be solved
with integers, square roots, and fractions, but for which he
presents quite accurate approximations.
Using Fibonacci as its main source, Luca Paciolis Summary
of Arithmetic, Geometry, Proportions, and Proportionality (1494)
is the first treatise on double-entry bookkeeping. Pacioli was
a friend and colleague of the artists Piero della Francesca
(from whom he plagiarized his mathematics) and Leonardo
da Vinci, who illustrated his book On the Divine Proportion
(1509), a treatment of the golden ratio of the Greeks.
Pacioli considers that it is necessary in business to be a
good accountant and a ready mathematician and that busi-
ness affairs must be arranged systematically, for without
systematic recording, their minds would always be so tired
and troubled that it would be impossible for them to conduct
business. The system he expounds has the great advantage
that it envisions an exhaustive accounting in which all the
enterprises of a company can be drawn together.
Controversy and Coefficients 29

It is appropriate that Paciolis Summary was one of the first

books to be printed on movable type, since it appealed to the
increasing audience of businessmen who needed systematic
and ironclad schemes of accounting for their businesses. Per-
haps the best known aspect of the double-entry system is the
duality between credit and debit, manifest in the dual record-
ing of every transaction on both sides of the ledger; this leads
to the possibility of double-checking the accuracy of accounts
through the comparison of trial balances struck indepen-
dently on each side of the ledger. Though this is surely useful,
it is not the crucial innovation of the double-entry scheme,
for the accounts of Roman slaves or medieval factors were
also able to be balanced. Double entry allows not only bal-
ance but completeness of accounting because it records every
kind of transaction, including both real accounts that rep-
resent concrete transactions with a specific client or credi-
tor and what now are called nominal accounts, such as
depreciation or good-will.
Pacioli places his discussion of bookkeeping and commer-
cial mathematics next to his exposition of algebra. Here he
was not an innovator but an influential compiler of tech-
niques. His exposition indicates the close association
between commercial and what we would consider pure
mathematics. We gain a similar impression from other early
works, such as the Treviso Arithmetic (1478) and Johann
Widmans Mercantile Arithmetic (1489), the oldest book
in which the familiar + and signs appear in print. Here
again these symbols at first refer to surplus and deficiency in
warehouse inventory, only later becoming signs of abstract
operations. As with Pacioli, practical considerations lead sur-
prisingly quickly to questions that transcend the symbols
commercial origins, including questions about the solvabil-
ity of equations.
30 Chapter 2

Though the Babylonians had made significant strides in

solving what we would call cubic equations, all of their work
had been lost. Fibonacci thought such equations could not
be solved by magnitudes and square roots; Pacioli thought
them quite unsolvable by algebra, an opinion he derived from
the Arabic poet-mathematician Omar Khayyam, who had
done extensive work classifying cubics and exploring their
solvability. Piero della Francesca, a great mathematician as
well as painter, tried unsuccessfully to solve cubic, quartic,
and even quintic equations. The breakthrough came from
a motley bunch of Italians, of whom the most colorful was
Girolamo Cardano, whose life spanned most of the sixteenth
century (figure 2.1). Cardano was a physician, an inveterate
gambler, and an astrologer, whose religious opinions were
sufficiently heretical that the Inquisition jailed him in his old
age. It was his interest in gambling that led Cardano to write
the first book on the mathematics of probability. Despite all
this, he held prestigious professorships and eventually re-
ceived a pension from the pope. He wrote prolifically, and
his Book of My Life is astonishingly frank, detailing not only
his own passions and obsessions but also the misadventures
of his children (one son poisoned his wife and was executed,
the other was a neer-do-well who robbed his father). His
role in the history of algebra is central but controversial, the
subject of contention even among his contemporaries.
In his book The Great Art, or the Rules of Algebra (1545),
Cardano treats problems that we would call quadratic as al-
ready solved, and he cites Mahomet the son of Moses the
Arab (that is, al-Khwarizm) as well as Fibonacci. Cardano
gives the problem of the cubic equation a rhetorical form
and a commercial context: Two men go into business to-
gether and have an unknown capital. Their gain is equal to
the cube of the tenth part of their capital. If they had made
Controversy and Coefficients 31

Figure 2.1
Girolamo Cardano.

three ducats less, they would have gained an amount exactly

equal to their capital. What was the capital and their profit?
In
 x modern
3 terms, if the unknown capital is x, then
 x the
3 gain is
10
and the equation defined by the problem is 10
3 = x.
The commercial aspect of such problems may have been
less important to Cardano than the possibility of rivaling the
ancients. In those days, classical statues were regularly being
32 Chapter 2

unearthed in Italy, astonishing beholders with the immense

achievements of antiquity. The works of Plato had only re-
cently been translated into Latin, and the Greek mathemati-
cians were likewise just beginning to be read. Were these
ancients gods, not to be touched by the moderns? Yet the an-
cients had not solved the cubic, and here the moderns might
still emulate and even surpass them.
The chain of events is not completely clear; what is clear
is that the solution of the cubic became the first example
of a sordid modern genre: the scientific priority fight.
Niccolo Fontana, nicknamed Tartaglia (Stammerer),
claimed to have solved the problem, but in truth he started
his work after had heard that it had already been solved by
Scipione del Ferro, who did not publish it but disclosed it
to a student before his own death. Tartaglia may well have
had a hint from this earlier source. In those days, standards
of attribution and citation were lax; Tartaglia had already
published someone elses translation of Archimedes as if it
were his own, and he also tried to pass as the author of the
law of the inclined plane without giving proper credit to
its true discoverer. At any rate, Tartaglia did make a break-
through in solving cubic equations, which allowed him to
win a public contest with del Ferros student. These contests
show something of the status of algebra at the time; they were
rowdy tournaments of skill, in which each contestant would
try to stump the other with a problem to win a handsome
prize. The popular appeal of these mathematical gladiato-
rial matches was on a par with bear-baiting and involved
something of the same atmosphere.
But how could Tartaglia win, if del Ferros student had
learned the solution from his teacher? In those days, there
were many variations of cubic problems that seemed
Controversy and Coefficients 33

separate, and to which various techniques were applied,

because mathematicians lacked the modern symbolism that
can include all possible varieties of cubic equations under
the general form a 3 x 3 + a 2 x 2 + a 1 x + a 0 = 0. Thus, the student
might have been able to solve a problem like the cube and
the number equal to the first power, as Cardano calls it (for
instance, x 3 + 21 = 16x), but still be stymied by the cube
and first power equal to square and number (as he would
have written an equation like x 3 + 20x = 6x 2 + 33). On a
more basic level, since only positive integers were then rec-
ognized, any case involving negative numbers would have
had to be treated separately, by moving such numbers to the
other side of the equation. In any case, news of Tartaglias
victory reached Cardano, who pressed him for details of his
methods, hinting that he would arrange a meeting with a
wealthy patron.
Tartaglia was wary but needed money, perhaps hindered in
earning his living because of a speech impediment caused by
saber wounds near the mouth that he received as a child dur-
ing the French invasion of Italy in 1512. He disclosed his se-
cret to Cardano in 1539, but only after making him swear not
to reveal it. Tartaglia communicated the solution in the form
of a cumbersome poem, almost riddling enough to be in code.
Years passed, during which Tartaglia did not publish his so-
lution. Then, in 1543, Cardano visited Bologna and gained
access to del Ferros papers, which showed that del Ferro
had in fact solved the cubic before Tartaglia. Cardano then
felt free to publish the method in his Great Art, attributing it
to Tartaglia, and del Ferro before him, but without mention
of his oath of secrecy. By giving credit, he surely sought to
avoid the appearance of trying to steal the secret. He says
that before seeing this work, he had believed the claim of
34 Chapter 2

Luca Pacioli that there was no rule that solved equations

beyond the quadratic. But by mentioning del Ferros earlier
discoveries, he diminishes Tartaglias claim to uniqueness.
In explanation, he only remarks that, in response to his en-
treaties, Tartaglia gave it to him and also that, after learning
it, he went further than Tartaglia ever did.
In fact, Tartaglia did not have the complete solution to
every variety of cubic that Cardano presents, which he di-
vides into thirteen cases, many with subcases. Still, Tartaglia
was outraged and published accusations of plagiarism. At
this point a new challenger stepped in. Cardano had a ser-
vant named Ludovico Ferrari who showed such keen interest
in mathematics that Cardano took him on as a student. With
his masters honor at stake, Ferrari challenged Tartaglia to
a mathematical duel that ended with Ferraris victory, or so
it seems from subsequent events, for Ferrari obtained many
flattering offers, including a professorship in Bologna, while
Tartaglias account reveals that he left even before the contest
was over.
How, then, was the cubic equation solved? The method
turns out to be an early instance of a powerful mathematical
strategy: to solve a more difficult problem, reduce it to a sim-
pler problem you have already solved. Here, the technique of
completing the square is the key. As noted earlier, we can
solve a quadratic equation by reducing it to the form of a per-
fect square and then taking the square root. It is natural to try
to attack the cubic in the same way by completing the cube.
Del Ferro, Tartaglia, and Cardano showed how any given cu-
bic equation can be put in the form of a perfect cube equal
to some numerical value. Then we can complete the solution
by taking the cube root. Though they did not present their
deduction with a three-dimensional diagram, it is useful to
Controversy and Coefficients 35

consider such a picture (box 2.4), for (like the parallel pic-
ture shown for the quadratic equation in box 2.3) it shows
that this problem is essentially like a jigsaw puzzle. The
solution of this three-dimensional puzzle requires breaking
it into slices.
That is, completing the cube involves a series of interlock-
ing solids that must be completed and reassembled. Box 2.4
shows an example from Cardano, along with its solution in
modern notation. Though the result looks a little daunting,
it only involves nested radicals, that is, roots of roots (such
as the square root of a cube root). Just as the solution of the
quadratic involves square roots, the solution of the cubic in-
volves cube roots, which in turn contain expressions involv-
ing square roots. Later, we will see that this characteristic
structure of nested radicals for the solution of the cubic gives
an important clue in the search for the solution of higher-
order equations beyond the cubic.
Ferrari not only defended his master but also made an ad-
vance of his own. At Cardanos request, Ferrari addressed
the problem of solving what we would call equations of
the fourth degree, or quartics, which they called square-
square. Their use of this term shows that they did not yet
have our more general conception of equations of arbitrary
degrees. In Cardanos book, cubic problems seem entirely
separate from quadratic ones, and both have a geometric
significance that is lacking in the case of square-square
equations. Because cubic equations can be solved by a three-
dimensional puzzle, one might think that quartic equations
would somehow require struggling with a puzzle in four di-
mensions. In fact, the solution of the quartic does not require a
fourth spatial dimension. It turns out only to require complet-
ing the square in a somewhat different way, using x 2 as the
36 Chapter 2

Box 2.4
Cardanos method of completing the cube
x v
x

v
x v x

A x B v C
u=x+v

Consider Cardanos example Let the cube and six times the
side be equal to 20, or x 3 + 6x = 20, in modern notation. We
want to make this problem take the form of a perfect cube,
and the trick needed is to set x = u v. In the diagram, let x 3
be the upper shaded cube, now embedded in a larger cube
with side AC = u. There is also a lower shaded cube having
side BC = v. To carve the cube with side x from the cube with
side u, we need to take three slices. Each slice individually
has volume u2 v, but they overlap. Each overlap has volume
uv 2 , but the overlaps also overlap in the little shaded cube of
volume v 3 . Thus x 3 = u3 {(volume of slices) [(volume of
overlap) (volume of overlap of overlap)]} = u3 3u2 v +
3uv 2 v 3 . To this expression for x 3 we must add 6x = 6u 6v
in order to satisfy the original problem.
Controversy and Coefficients 37

Box 2.4 Continued

Now we have x 3 + 6x = u3 3u2 v + 3uv 2 v 3 + 6u 6v.

To make this take the form of a perfect cube, we need to get
rid of the cross terms 3u2 v + 3uv 2 + 6u 6v = 3(uv)u
6v +3(uv)v +6u, here grouped to highlight uv. Now if uv = 2,
then these terms cancel: 3(uv)u 6v + 3(uv)v + 6u = 6u
6v + 6v + 6u = 0. So to complete the cube, all we need to
do is to stipulate that u and v are determined by the relation
uv = 2, which we are free to do since the two variables u, v
are being substituted for the one variable x, and so u, v must
be subject to one additional condition connecting them.
The rest is clear sailing: with this condition applied, our
cubic looks like u3 v 3 = 20, with v = u2 . Substituting this
back in, we have u3 8/u3 = 20. Multiplying both sides by
u3 gives u6 8 = 20u3 , or u6 = 20u3 + 8 if (like Cardano) we
want to keep everything positive. This looks worse, because
it seems to be of sixth order in u, but it is not because it is
quadratic in u3 . That is, let y = u2 ; then this equation becomes
y2 = 20y + 8. We know how to solve this quadratic equation:
 
y = (20 400 4(8))/2 = (20 4(108) )/2

= 10 108 = 10 + 108,
where the negative choice of sign is neglected because it leads
to a negative
root, which Cardano
would not have accepted.
So u3 = 10 + 108 and v 3
= 108 10. Taking  the cube root
3 3
of each side gives u = 108 + 10 and v = 108 10.
Finally, recalling that x = u v, we get Cardanos final
3
3
answer, x = 108 + 10 108 10.

sides of the initial square. This allowed Ferrari and Cardano

to treat square-square problems in terms of square problems.
Here there was no dispute about priority; Ferrari succeeded
on his own, though doubtless guided by Cardanos example.
38 Chapter 2

Box 2.5 shows Ferraris solution in modern notation. We can

treat this as a complicated problem of factoring, to use the
modern term. Because it followed so quickly on the solution
of the cubic, it must have seemed that all such problems were
now ripe for solution.

Box 2.5
Ferraris solution to the quartic equation

Take the equation x 4 + a 3 x 3 + a 2 x 2 + a 1 x + a 0 = 0 and shift

the terms so that it becomes x 4 + a 3 x 3 = a 2 x 2 a 1 x a 0 .
a2
Then add 43 x 2 to both sides to make the left-hand side a
perfect square:
 2
a 32 2 a3 x
x4 + a 3 x3 + x = x2 +
4 2
 
a 32
= a 2 x2 a 1 x a 0 (2.5.1)
4
If the right-hand side were a perfect square, we could take
the square root of both sides and reduce the equation to a
quadratic. In general, it is not a perfect square,
 but we  can2
make it so by adding to both sides the term y x 2 + a 32x + y4 :
 2    
a3 y a 32 a3 y
x2 + x+ = a 2 + y x2 + a 1 + x
2 2 4 2
 
y2
+ a 0 + . (2.5.2)
4
We can make the right-hand side a perfect square by ad-
justing y. This is equivalent to requiring that, in general,
Ax 2 + Bx + C = (e x + f )2 , where A, B, C are the expressions
in parentheses on the right-hand side of equation (2.5.2). If
Controversy and Coefficients 39

Box 2.5 Continued

this is to hold, then we must have A = e 2 , B = 2e f, C = f 2 or

B
e = A, f = C = .
2e
To make this consistent requires that B 2 4AC = (2e f )2
4e 2 f 2 = 0. So (substituting the values for A, B, C from
eq. (2.5.2) into B 2 = 4AC), y must satisfy
 2   
a3 y a 32 y2
a 1 + =4 a2 + y a 0 + , (2.5.3)
2 4 4
which is a cubic equation in y (called the resolvent of
the quartic equation) and hence solvable by the del Ferro-
Cardano-Tartaglia method. Once we have determined y, then
we can use (2.5.3) to determine e and f . Then we can return
to (2.5.2), now written as
 2
a3 y
x + x+
2
= (e x + f )2 , (2.5.4)
2 2
which is a perfect square whose square root yields two qua-
dratic equations,
a3 y
x2 + x+ = e x + f,
2 2
a3 y
x2 + x + = e x f, (2.5.5)
2 2
where e, f, y are now determined. These two equations can
readily be solved by the quadratic formula, yielding the four
solutions of the quartic equation.

In going past the cubic, however, the obvious references

to two and three dimensions disappear and the problem
becomes more abstract. How can the human mind under-
stand what it cannot see? As Cardano puts it, it would be
very foolish to go beyond this point. Nature does not permit
40 Chapter 2

Box 2.6
An illustration of the algebraic notation of Cardano and Viete

Cardanos notation is syncopated, meaning that it is basi-

cally abbreviated words. He does not have a general sym-
bolism for unknowns (which he calls rebus, things) or
coefficients.
He expresses his solution to the specific cubic equation in
box 2.4 (which he states as cubus p. 6. rebus qualis 20) as
follows:
. v. cu. . 108. p. 10. m. . v. cu. . 108. m. 10.
Here p. and m. stand for plus and minus, and . stands for
radix (square root), while . v. cu. stands for cube root.
Compare this with its modern form:
3
3
x= 108 + 10 108 10.
In contrast, Viete writes A cubus + B plano 3 in A,
aequari Z solido 2 for the modern x 3 + 3B 2 x = 2Z3 . He uses
vowels for unknowns, consonants for coefficients, cubus
or solido for cubed, and plano for squared; thus, A
cubus is x 3 , B plano 3 is 3B 2 , and Z solido 2 is 2Z3 . Note
that he uses the modern + and signs and also the radical

sign .

it. Moreover, the manner of expression of the algebraists

left much to be desired. Box 2.6 shows the notation of two
Renaissance algebraists, compared with the modern form.
The earlier manners of expression operated on entirely dif-
ferent principles; they were primarily a means of abbrevi-
ating ordinary language, like shorthand. Modern symbolic
mathematics steps decisively away from ordinary language,
allowing systematic manipulations that are foreign to natural
language. Although we can translate an equation into words,
Controversy and Coefficients 41

we cannot manipulate those words as we can the equation.

However impressive the achievements of the Italian mathe-
maticians, they lacked the insights that would be achieved
by symbolic mathematics.
Modern mathematical notation depends on the innova-
tions of a seventeenth-century French lawyer, royal coun-
selor, and codebreaker, Francois Viete (figure 2.2), who was

Figure 2.2
Francois Viete.
42 Chapter 2

able to merge ciphers, legal abstractions, and mathematics

to devise the basic algebraic symbolism we still use. Viete
considered himself a rediscoverer of a lost and ancient art,
for he could not bring himself to believe that the ancient
mathematicians could have achieved what they did with the
tools that they claimed to be using. This is not unreasonable;
anyone who has read one of Euclids magnificent geometric
proofs cannot help wonder what gave him the idea to draw
the crucial lines on which a proof depends. Certainly any
student of geometry who has tried to create an original proof
understands that there is no royal road (as Euclid put it)
to finding them. The student may wander down many blind
paths and suffer much frustration before hitting on a trick
that works.
This is what Pappus had called synthetic mathematics.
By this he meant that Euclid begins with certain axioms
and shows how they can be synthesized to yield the theo-
rem to be proved. But Pappus also alluded to a process of
analytic mathematics, in which the mathematician would
work backward from the desired result, finding on the way
what was necessary to arrive at the result in question. There
are some tantalizing examples of this working-backward in
Apollonius seminal work on the conic sections, and Viete
studied them carefully. He concluded that this analytic art
was probably the way the ancients had devised their mirac-
ulous proofs: to reach a certain theorem, work backward until
you find out what steps are necessary to reach that theorem,
then turn around and begin again from that starting point,
setting out the steps that you have found to be necessary.
Thus, analytic problem solving sets up a kind of scaffold-
ing from which the perfected architecture of the proof can
be constructed. At the end, Viete inferred, the ancient mas-
ters would remove the scaffolding so that the beholder could
Controversy and Coefficients 43

admire the beauty of the proof, unobstructed by the tools that

built it.
In Vietes view, the ancients obscured traces of their work-
ing methods both to excite more admiration and to accentu-
ate the beauty of the finished product. He thought he was
only rediscovering their methods, not innovating. However,
there is no evidence that the ancients really had the art that
Viete reconstructed; and if they did, it was most likely a
heuristic working-backward that lacked the symbolism he
created. Not knowing the place-system or the Hindu-Arabic
zero, the ancients had not taken the decisive steps toward true
symbolism. Viete himself was guided by his work in code-
breaking, so that the process of solving an algebraic problem
through symbolic manipulation became, for him, closely tied
to the methods used to solve ciphers, another practice un-
known to the ancients. He also may have been influenced
by Roman law, which would use a letter to symbolize an
unknown defendant (a John Doe, in modern terms). If one
could use such a symbol to represent a person, why not use
a symbol to stand for the cosa, the thing sought for in a
problem?
Further, Viete realized that such a symbol might itself be
manipulated as if it were a number, rather than an alphabetic
letter. Here we have the crucial moment in which algebraic
symbolism steps beyond the bounds of ordinary language
and semantics, if only through conciseness: compare the
brevity of 4x 2 with four times the unknown squared. But
the advantage of algebraic symbolism goes further: we can
add two equations (or otherwise manipulate them) in ways
that are impossible for two sentences. What language accom-
plishes with syntax, algebra can do by juxtaposing a few sym-
bols. Moreover, the ambiguities of language are eliminated
in algebra, whose symbols are unequivocal. Where language
44 Chapter 2

calls for judgment and logic, algebra requires only symbolic

manipulation according to fixed rules. We might even say
that the symbols do the thinking, because they embody log-
ical and mathematical power without the users needing to
reinvent it each time, or even understand it beyond following
the rules. As our story unfolds, it will become increasingly
clear that this power has complex consequences, for it enables
amazing discoveries while hiding their meaning behind the
very symbolism that discloses them.
Viete called his new discovery a logic of species, as op-
posed to a logic of numbers. He meant by species the
way the unknown could stand not just for a single num-
ber but for a whole class of possible values. That is, the
unknown is not just a stand-in for a single value but repre-
sents the whole spectrum of possibilities: it is a variable, as we
now say. Here again language falls behind, for Plato had sug-
gested that each word aspires to a natural connection with
what it represents, at least for nouns as naming words. Al-
gebra is more like a discourse composed solely of pronouns
and verbs and is self-consciously artificial: there is no natural
relation between whatever we label the unknown, whether x
or , and the variable it represents. Unlike language, which is
meaning-laden and intentional, algebra removes all trace of
ordinary intentionality and meaning. It points toward a very
different view of discourse than Platos, toward a stream of
purely conventional symbols, devoid of intrinsic sense.
The abstractness of mathematics is amplified by Vietes
even more remarkable discovery: the coefficient. It may not
have been so startling to symbolize the unknown by a single
symbol, for the Italians were already symbolizing the com-
plex notion of a number I am looking for but dont know
by the word cosa or even by a single character (sometimes
even as simple as a dot, ). But that unknown was always
multiplied by a definite quantity, and it was bold of Viete to
Controversy and Coefficients 45

allow that quantity to become as indefinite as the unknown

itself. Here we have the birth of the coefficient, which enables
a huge increase in generality and power. Compare the partic-
ular case x 2 3x + 2 = 0 with a 2 x 2 + a 1 x + a 0 = 0, which in-
cludes the specific equation as a single special case among
an infinite expanse of possibilities in which the coefficients
a 2 , a 1 , a 0 roam over all possible values. Indeed, an expression
like a 1 x is awkward to render in ordinary language: Multi-
ply an arbitrary number called a 1 by another arbitrary num-
ber called x. Algebraic symbols avoid the debacle of Babel,
bypassing the human confusion of tongues.
Viete not only made these important advances but also
grasped their immense power. Though he calls himself
merely a restorer of what was lost, he wraps himself in a
heroic mantle. He considers Arabic algebra tedious and
barbaric, rejecting even their word algebra for what he
calls instead the analytic art. Viete rebukes them, as if they
were false alchemists, whereas he has the royal touchstone
that will actually separate mathematical gold from dross.
Viete was cold to the bitter religious controversies of his age.
He saves his excitement for mathematics, which, through his
newly systematic codebreaking, he uses to protect France. He
exults in these analytic powers, realizing that he has purified
an ancient art and given it new life.
At the height of those powers, Viete accepted a challenge
that pointed to a new frontier in algebra. In 1593, the Belgian
mathematician Adriaan van Roomen challenged all the
mathematicians of the whole world to solve a monster equa-
tion of the forty-fifth degree (box 2.7). The Dutch ambassador,
paying a call on Henry IV, Vietes royal master, ironically of-
fered the king his condolences that there were no French
mathematicians up to the challenge. Stung, Henry called for
Viete, who sat down with the problem and was able within
a few minutes to find the positive roots of the equation.
46 Chapter 2

Box 2.7
Adriaan van Roomens test problem

Solve
x 45 45x 43 + 945x 41 12,300x 39 + 111,150x 37 740,459x 35
+ 3,764,565x 33 14,945,040x 31 + 469,557,800x 29
117,679,100x 27 + 236,030,652x 25 378,658,800x 23
+ 483,841,800x 21 488,494,125x 19 + 384,942,375x 17
232,676,280x 15 + 105,306,075x 13 34,512,074x 11
+ 7,811,375x 9 1,138,500x 7 + 95,634x 5 3,795x 3
+ 45x = K ,
where K is a given number.

Viete published this result in 1595, along with his expla-

nation of his method. Humorously, he offered to sacrifice
a hundred sheep to celebrate the occasion, ironically echo-
ing Pythagoras mythical gesture, though with more tranquil
animals. Viete emphasized the connection between his new
algebra and geometry, noting that earlier mathematicians
were not able to reconcile them. But was that not because
algebra was up to that time practiced impurely? Friends of
learning, embrace a new algebra; farewell, and look to the
just and the good. Vietes triumph confirmed his sense that
what he had solved was not just one problem but the proud
problem of problems, which is: TO LEAVE NO PROBLEM
UNSOLVED.
3 Impossibilities and
Imaginaries

Van Roomen was so impressed by Vietes astonishing feat

that he traveled to France expressly to meet him. Here was
a man who claimed to have found the master art that would
solve any algebraic problem. Was it really possible that Viete
had the tools to solve every possible equation, of whatever
degree? In the case of van Roomens equation of the forty-
fifth degree, Viete was lucky, for he recognized that par-
ticular equation as closely related to formulas he found in
trigonometry. Had van Roomen changed a single coefficient,
Vietes solution would no longer have been valid. Clearly,
van Roomen had not been envisaging the general problem
of solving equations of higher degrees but had come upon
that specific equation in the course of his trigonometric stud-
ies, and was testing to see if others had discovered it also.
Thus, the solution of van Roomens equation did not
after all illuminate the general question of whether all equa-
tions are solvable in radicals, and indeed Vietes solution in-
volved trigonometry, which (as we shall see) goes beyond
radicals. The tenor of Vietes conviction was clearly that all
equations are solvable by the same general procedures used
to solve the quadratic, cubic, and quartic equations. Other
48 Chapter 3

mathematicians were skeptical about algebraic solutions,

and their reservations were an important countertheme to
the triumphal march of modern algebra. For all his boldness
in astronomy, Johannes Kepler was one who remained criti-
cal of the claims of algebra, holding that nothing is proved
by symbols, nothing hidden is discovered in natural phi-
losophy through geometrical symbols. This went beyond
Keplers vehement declaration that he hated all kabbalists
and their esoteric symbolism. He considered algebra to be
ingenious and surprising, but always thought it fell short
of the certainty of geometry, his mathematical touchstone.
The question about algebra comes to a head in Keplers
Harmony of the World (1619), in which he tried to place
music on a new geometrical foundation, in place of the old
Pythagorean numerology of simple whole numbers. The new
foundation would be based on the ratios between the sides of
the simplest regular polygons. One of his goals was to create
a theory that would encompass contemporary musical scales
(which used intervals like 5:4 and 6:5), rather than simply the
Pythagorean scales (which only used numbers up to 4). To
do this, he needed to use all the regular figures up to and
including the hexagon but excluding the heptagon, a seven-
sided figure that would have introduced harsh dissonances
like 7:3 into his scheme. He had a geometrical argument for
this, in that the regular heptagon cannot be constructed with
straightedge and compass. He was aware, however, that the
Swiss mathematician and clock-builder Jost Burgi had found
algebraic equations whose solutions gave the sides of regular
polygons. In the case of the heptagon, the equation is cubic
and thus solvable by the del Ferro-Cardano-Tartaglia method
(box 3.1). This would have caused a crisis for Keplers musical
theory, since accepting this solution would open the door to
those terrible dissonances. But the solution involves factors
Impossibilities and Imaginaries 49

Box 3.1

If we call the side of the polygon x and let y = x 2 , Burgis

equation for the pentagon is y2 5y + 5 = 0, and for the
heptagon, y 3 7y2 + 14y 7 = 0. These are both solvable, but
the solution of the heptagons equation using the del Ferro-
Cardano-Tartaglia method yields
 

   
1 3 7 3
x=  3
1 27 + 1 + 27 .
3 2


like 1 27, which seemed at that time bewildering if
3

not absurd. Kepler anticipated correctly that the solution of

such complex roots would require an infinite process. Since
even the Omniscient Mind of God could not know such
roots in a simple eternal act . . . , neither can it be known by
the human mind.
Because geometry always mirrors the simple eternal act
of the divine intelligence, Kepler preferred it to the question-
able processes required to solve cubic equations. The issue
was important, for on it hung not just musical questions but
the whole structure of the universe, which he took to rest
on musical relations. Keplers search for cosmic harmony
led to his Third Law of planetary motion: The square of the
period of a planet is proportional to the cube of its mean or-
bital radius. If the cosmos were organized geometrically, as
Kepler believed, algebra might be irrelevant or inadequate.
Others paid even less heed. Galileo Galilei never once men-
tioned algebra, though it would seem likely that he had at
least heard of it. When he famously described the Book of
Nature as written in the characters of mathematics, he
50 Chapter 3

explicitly meant triangles and geometrical figures, not alge-

braic symbols. Perhaps here Galileo was influenced by the
academic mathematics of his time, which stood aloof from
the commercial aspects of early algebra.
Nevertheless, Vietes discoveries drew wide attention in
France. His optimism about the possibility of solving every
problem was shared by his great successor, Rene Descartes,
who was seven when Viete died in 1603 and whose semi-
nal book La Geometrie (1637) would consolidate and extend
Vietes results. Despite the closeness of their work, Descartes
was for some reason reluctant to acknowledge Vietes pri-
ority, mentioning him only grudgingly and claiming to have
made the same discoveries independently, before reading the
older mathematicians books. Nevertheless, it must be said
that Descartes employed the new algebra with unparalleled
clarity and force. He also put the new discoveries at the ser-
vice of a new philosophy, reaching toward a larger vision of
understanding the world as matter whose motion is mathe-
matically intelligible.
The thread of algebraic mathematics runs through
Descartess work. Symbolic mathematics and rigorous rules
exemplify his watchword: method. Indeed, La Geometrie and
essays on optics and meteorology form an appendix to his
Discourse on the Method of Rightly Conducting the Reason. These
technical works vindicated his insistence on correct method
(Rules for the Direction of the Mind, as an earlier work called
it). He applied this touchstone by systematically doubting
the received learning of his time. Descartes also applied his
criteria of method to bring algebra essentially to its modern
form. Viete had stipulated that all the terms in an algebraic
equation should have what we would call the same dimen-
sion; that is, if x has the dimensions of length and an equa-
tion contains an x 3 , all the other terms must also have the
Impossibilities and Imaginaries 51

dimension of volume (length cubed). Descartes showed that

this was unnecessary: we can make all terms have the same
dimension by multiplying by a unit length as many times as
needed, that is, if 6 and x each have the dimensions of length,
we can simply multiply by a unit length to give 6x the di-
mension of volume. Looking at a page of La Geometrie, one
sees equations that are almost exactly like ours (figure 3.1).
As he shows how the geometry of Euclid and the conic
sections of Apollonius can be described by quadratic equa-
tions, Descartes illuminates nagging problems that had shad-
owed these equations. Already the Arabs had realized that
some quadratic equations seem to have solutions that are not
positive numbers. They did not recognize negative numbers
as numbers, although they did seem to have had ways of
dealing with debits and credits that approach our concept
of signed numbers. Here again the primal notion of number
shows its tenacity, for the concept of a negative number
seems to violate our intuitive sense of counting: how can one
count negations? Even Viete did not allow his unknowns or
coefficients to be negative. Only in 1629 did Albert Girard
recognize negative roots, explaining that the negative in ge-
ometry indicates a retrogression, where the positive is an
advanceessentially the modern idea of a number line,
in which positive and negative represent directions forward
and backward along the line. Still, the newcomers were not
fully welcome. Other mathematicians of the time refused
to admit them, naming them absurd numbers, and even
Descartes called them false or less than nothing. As late as
the eighteenth century, some textbooks denied that the prod-
uct of two negative numbers is a positive number, a rule that
Pierre Simon Laplace said presents some difficulties even
in 1795.
52 Chapter 3

Figure 3.1
From Rene Descartes, La Geometrie (1637) (p. 373).
Impossibilities and Imaginaries 53

Though Descartes refers to false roots, he does concede

that they exist and are roots. He even constructs an ingenious
way to tell how many positive and negative roots any equa-
tion has, known as Descartess rule of signs. It is amazingly
simple: An equation can have as many true [positive] roots
as it contains changes of sign, from + to or from to +; and
as many false [negative] roots as the number of times two +
signs or two signs are found in succession. Box 3.2 shows
an example, along with Descartess argument for why the
rule holds. Here the power of the new notation comes to the
fore, because merely by looking at the signs of the successive
terms in the equation one can determine the signs of the roots,
even if we dont know their numerical values. Descartess ar-
gument is perfectly general and applies to equations of any
degree. He also shows that it is also easy to transform an
equation so that all the roots that were false shall become
true roots, and all those that were true shall become false,

Box 3.2
Descartess rule of signs

In the page shown in figure 3.1, Descartes explains that

the equation x 4 4x 3 19x 2 + 106x 120 = 0 has three pos-
itive (true) roots and one negative (false) root because
there are three changes of sign between the terms. Note that
the equation can be factored into (x 2)(x 3)(x 4)(x + 5) =
0; when multiplied out, note that the x 4 term is positive, but
the x 3 term then must have a negative coefficient since it is
formed by the terms (2 3 4 + 5)x 3 = 4x 3 . Here, the
negative sign comes from the odd number of positive roots.
Similar considerations lead to Descartess general rule of
signs.
54 Chapter 3

by changing the signs of every other term, in accord with his

rule. Surely his awareness that true roots could be changed
into false and vice versa must have helped him accept both
as different but comparable kinds of solutions.
Descartes also opens the door for an even more problem-
atic set of solutions that includes square roots of negative
numbers, which we still hold at arms length by using the
loaded term imaginary numbers. As difficult as it is to
understand what a negative number might mean, an imagi-
nary number is more perplexing still, and many educated
people even today would be hard pressed to account for
them. Their alien status is emphasized by calling all non-
imaginary numbers real numbers. The Arabs simply did
not allow the square root of a negative number; in such cases,
al-Khwarizm says there is no equation. Cardano thought
such a quantity was sophistic and as subtle as it is use-
less. Indeed, when one is dealing with quadratic equations,
if the root is real, then the formula does not include any imag-
inary numbers, and all is well. Conversely, if the root is an
imaginary number, we could reject it on the grounds that such
a solution is not allowable because it is not real. However, in
the case of cubic equations, even when all the roots are real,
the del FerroCardanoTartaglia formula explicitly involves
imaginary numbers. This irreducible case (as it came to be
known) means that we must come to terms with imaginary
numbers if we want to use the formula for the cubic equa-
tion. In 1560, Raphael Bombelli had what he called a wild
thought about how he might resolve this puzzle by showing
that the imaginary parts might cancel each other (box 3.3).
This trick works in only a few cases, however, leaving open
the general question of how to understand imaginary num-
bers and (even worse) how to take their cube roots or manip-
ulate them when they appear alongside real numbers.
Impossibilities and Imaginaries 55

Box 3.3
Bombellis wild thought

Bombelli considered the equation x 3 = 15x + 4, whose roots

are given, according
 to the del 
Ferro-Cardano-Tartaglia so-

lution, by x =
3
2 + 121 + 2 121. Now by sim-
3

ple substitution we can show that x = 4 also solves this

equation. To reconcile
 the two solutions, Bombelli proposed
3
the substitution 2 + 121 = 2 + b 1, where b re-
mains to be determined.
Cubing both
sides of this expres-
sion yields
2 + 121 = 2 + 11 1 = (2 + b 1)3 =
8 + 12b 1 6b 2 b 3 1, which is only true if b = 1. Sim-

ilarly, Bombelli showed that 2
3
121 =2 1. The
whole solution x is then x = 2 + 1 + 2 1 = 4. This
ingenious trick seemed to rest on sophistry to Bombelli; in-
deed, it cannot be applied to every cubic equation, only spe-
cial cases. ( The solution of the heptagon equation in box 3.1
that perplexed Kepler is one example that cannot be treated
in this way.)

Descartes introduced the terms real and imaginary in

their modern sense. Numbers that have both real and imagi-
nary parts he called complex. By giving both parts a name,
he granted the imaginary roots a kind of legitimacy, albeit
under a name that suggests unreality. This aura persisted
even in Gottfried Wilhelm Leibnizs intriguing description of
almost an amphibian between being and nonbeing, which
we call the imaginary roots. Leibniz may have been alluding
to the mysterious procession of the Holy Spirit in Christian
theology, but his description perpetuates a certain spooki-
ness about quantities that arguably have as much reality as
real numbers. This impression is perhaps best dispelled
by thinking of complex numbers as points on a plane, with
56 Chapter 3

imaginary
axis
2i

(1, i )
i

1 2 real axis

Figure 3.2
The two-dimensional representation of complex numbers.

the real part giving the x coordinate and the imaginary part
the y coordinate (figure 3.2). In this view, a complex num-
ber is fundamentally two-dimensional, whereas the real
numbers are one-dimensional.
Beyond counting complex roots, Descartes states (but does
not prove) that each equation has as many roots as its degree:
a quadratic equation has two roots, a cubic three roots, and
an nth degree equation n roots. This is a crucial insight, ori-
ginally formulated in 1629 by Girard, which expresses the
confidence of Viete and Descartes that all algebraic prob-
lems have solutions. It will later lead to the Fundamental
Theorem of Algebra, which states that all equations have at
least one root, but for the moment that is only a gleam in
Descartess eye.
Descartes also notes that it is possible to shift the roots
of any equation by a certain value, even if those roots are
unknown. If you wish to increase the value of each root by 3,
simply substitute y 3 into the given equation wherever x
occurs; the new equation in powers of y will, by definition,
have roots y = x + 3, shifted as required. To scale the value
of each root by a factor s, substitute sy for x everywhere;
Impossibilities and Imaginaries 57

the new equation in y has roots y = x/s. By such simple

transformations, any given equation can generate a family of
similar equations, with different specific values of the roots in
each case. However, these transformations do not affect the
relative relation of the roots, which are all affected equally. We
will return to this relativity of roots (which Descartes did
not discuss) in chapter 9.
Descartes is so sure that equations of degree beyond four
can be solved that he doesnt bother to give specific proce-
dures, but instead offers a general rule: try to simplify the
equation to one of lower degree by factoring it. Not only does
he have the work of Cardano and Ferrari to rely on, but he
(like Viete before him) has been able to use his analytic ap-
proach to geometry to solve problems that had baffled the an-
cients: the trisection of an angle, the duplication of the cube.
To be sure, in those cases he has to use tools not allowed in an-
cient geometry, that is, not just straightedge and compass but
also conic sections. Descartess proudest achievement is to
have solved another problem that troubled the ancients, the
four-line locus problem posed by Pappus: to find the curve
that lies at a given distance from each of four given lines.
Not only does Descartes solve this problem with almost ar-
rogant ease, but he goes on to solve a five-line locus problem,
which the ancients did not even attempt. These triumphs oc-
cupy him far more than the problem of solving equations
beyond the quartic. He does give one example, solving an
equation of the sixth degree by ingenious manipulations of
conic sections, but his equation is not completely general;
he restricts its coefficients to allow his geometric method to
succeed. With an ironic flourish, he concludes that it is only
necessary to follow the same general method to construct all
problems, more and more complex, ad infinitum; for in the
case of a mathematical progression, whenever the first two
58 Chapter 3

or three terms are given, it is easy to find the rest. He asks

pardon for what he has intentionally omitted, so as to leave
to others the pleasure of discovery.
But having left the problem of equations beyond the fourth
degree as an exalted exercise for the reader, he evidently does
not realize how difficult that exercise will prove to be. The
next centuries contain many attempts to realize the vision
of Viete and Descartes. Though there had been extraordi-
nary progress in solving cubic and quartic equations during
the sixteenth century, it is perhaps not so surprising that the
issue of the quintic equation lay unresolved for two hundred
years more. After Descartes, a whole huge field of investi-
gation lay open, in which the question of solving higher-
degree equations was only one problem among many. Yet this
strand is richly interwoven in the mathematical history of the
two centuries after Descartes, which will produce the great-
est mathematicians yet to have lived: Newton, Euler, and
Gauss.
4 Spirals and Seashores

Sir Isaac Newtons relation to algebra was peculiarly ambiva-

lent. As a young man, he studied Descartess works carefully.
In maturity, he so loathed Descartes that he sometimes seems
to have been unwilling to utter or write Descartess name, as
if it would defile him. The reasons for this are not totally
clear. Newton despised Descartess philosophy as cloaked
atheism, but there were also mathematical reasons for his
displeasure. Newton, like Viete, styled himself a devotee
and restorer of antiquity, whereas Descartes, he felt, had be-
trayed the profundity of Greek geometry for the questionable
advantages of analytic geometry.
To demonstrate his claim, in his Principia Newton offers
a purely geometric resolution of the four-line locus problem
Descartes so prided himself on, noting caustically that this
gives not an [analytical] computation but a geometrical syn-
thesis, such as the ancients required, of the classical problem
of four lines, which was begun by Euclid and carried on by
Apollonius. In the rest of the Principia, Newton also tends to
avoid analytical algebra, preferring to state his propositions
in the manner of Euclid. However, this appearance is decep-
tive, for in fact Newton does use algebraic expressions as
well as a new tool of analytical mathematics that he himself
60 Chapter 4

has created: the calculus. Though he phrases it in terms of

geometry, Newtons calculus goes beyond anything known
to the ancients, even Archimedes, whom Newton considered
his precursor.
Thus, though he tends to prefer geometry, Newton is
steeped in algebra. As a young professor, he lectures on the
topic (during 16721683), though his lectures on Universal
Arithmetic are not published until 1707. Newton also makes
a number of discoveries that will bear on the problem of the
quintic. He was able to make simple connections between the
coefficients of an equation and its roots, known as Newtons
identities. These generalize Girards identities: the coeffi-
cient of the next-lowest degree term (of x 4 , for a quintic equa-
tion) is equal to the negative sum of all the roots (see box 4.1).
Likewise, all the other coefficients are equal successively to
the sum of all products of the roots taken two at a time, then
three at a time, until the final, constant term is equal to the
negative of the product of all the roots. Later on, it will be-
come extremely important to note that all of the roots appear
in a symmetric manner in each of these products and sums.
That is, every root appears in exactly the same way as ev-
ery other root, so that if one were to exchange two roots, the
product of all of them would be unchanged, as would the
products of them taken two at a time, and so on. The im-
portance of these rules is that they let us see direct relations
between coefficients and roots, without knowing the value
of the roots. Newton is also able to obtain upper and lower
bounds for the roots, that is, to show how large (or small)
they could possibly be. Using these tools, we can look at any
equation and determine the range within which its roots lie,
as well as whether they are negative or positive in value.
Newtons deepest insight, however, remained hidden for
many years, buried in a passage in his Principia that was
Spirals and Seashores 61

Box 4.1
Girards and Newtons identities

Consider a cubic equation with the roots x1 , x2 , x3 . The equa-

tion can be written (x x1 )(x x2 )(x x3 ) = 0. Multiplying
it out, we get x 3 (x1 + x2 + x3 )x 2 + (x1 x2 + x1 x3 + x2 x3 )x
x1 x2 x3 = 0. Note that the coefficient of the x 2 term is the
negative sum of all three roots, (x1 + x2 + x3 ), while the co-
efficient of the x term is the symmetric product of all the roots,
taken two at a time: (x1 x2 + x1 x3 + x2 x3 ). Finally, the constant
term of the equation is the negative product of all three roots:
x1 x2 x3 . We can apply the same reasoning to an equation of
any degree, so that the coefficient of the next-to-highest
power of the unknown must be the negative sum of all the
roots, the next coefficient must be the symmetric sum of all
the roots taken two at a time, and so on. We will refer to
these simple relations as Girards identities, which Newton
greatly generalized in expressions he derived for the sum of
the square of all the roots, or the sum of their nth powers,
Newtons identities.

not much noticed until the twentieth century. In lemma 28

of Book I, in an investigation of the curved paths that bod-
ies can take (as in the orbits of planets), he shows that No
oval figure exists whose area, cut off by straight lines at will,
can in general be found by means of equations finite in the
number of their terms and dimensions. That is, if we draw
lines across an oval, the area between those lines cannot be
expressed by a finite algebraic equation. Though he does not
define the word explicitly, by oval Newton seems to mean
any closed curve that does not cross itself (it is simple,
in the language of modern mathematicians) and is infinitely
smooth (it always has finite curvature, never is flat). The
62 Chapter 4

simplest such curve is a circle, and it had long been suspected

that the area of a circle is irrational with respect to its radius.
But what Newton surmises goes far beyond the irrationality
of , the name later given (by Euler) to the value of a circles
perimeter divided by its diameter, and beyond the irrational-
ity even of 2 . Newtons argument indicates that the area of
a circle is not given by any algebraic equation, however high
its degree, and thus that area (and hence also) cannot be
expressed in terms of any finite number of square roots, cube
roots, fifth roots, and so on.
To use the term that Euler later introduced, the area of a
circle is transcendental, meaning it cannot be expressed as the
root of any equation of finite degree whose coefficients are
rational numbers. At one stroke, Newton indicates that such
magnitudes exist (because circles exist, and have areas), and
also that there are infinitely many of them, since his proof
is not restricted to circles but holds for any oval curve. His
proof is a miracle of simplicity and power, for which he does
not even bother to draw a picture or write down a line of
algebra. It follows from a single brilliant contrivance. Inside
the oval, pick any point whatever; let us call it the pole, P.
Now let a straight line come out from that pole and rotate
around it at uniform angular speed. Picture a clock hand
that makes a complete revolution in one hour. Now imagine
a point of light moving along that hand, starting from the
pole and moving outward along the hand with speed given
by the square of the distance from the pole to the point A
where the hand intersects the oval (figure 4.1).
Newton has set up a way of measuring the area of the cir-
cle, for each hour the hand sweeps through that area, and the
moving point keeps track of the area because it is traveling
with speed proportional to the area swept out. Here Newton
is implicitly using his new calculus of motion, for he knows
Spirals and Seashores 63

Figure 4.1
Newtons diagram for his lemma 28, which argues that no oval curve has
an area expressible by a finite algebraic equation. From any point P inside
the oval, draw a straight line that rotates about P at uniform angular speed.

A
P

Figure 4.2
Detail of Newtons lemma 28; the speed of the moving point is proportional
to the area swept out between A and A .

that, in an infinitesimally short time, the point travels a dis-

tance from the pole equal to the area the hand has swept out
in that time (figure 4.2). However, we dont need to know
anything about calculus in what follows. All that matters is
that since, second by second, the moving point is registering
the area that the hand sweeps out, we can measure the whole
64 Chapter 4

area of the oval merely by waiting until an hour has elapsed

and measuring the distance the moving point has traveled
radially outward from the pole. For the two-dimensional
problem of measuring an area, Newton has substituted a one-
dimensional problem that gives the same answer: find the
length traveled outward by the moving point during an hour.
The hand moves around uniformly, but the moving point
speeds up and slows down in the course of each hour, in pro-
portion to the square of the distance from the pole to the oval
at any given moment. Each hour it returns to its initial speed
of an hour before. If you were to watch the lighted point (or
if you were to open the shutter of your camera and make a
long exposure), you would see it move in a spiral, starting
at the pole and making an infinite number of gyrations, as
Newton puts it (figure 4.3). Now Newton applies a reductio
ad absurdum: Suppose that it is possible to describe this spi-
ral (and hence also the area of the oval) by some polynomial
equation with a finite number of terms, f (x, y) = 0. Then con-
sider a straight line running across the spiral that we will call
the x-axis, defined by y = 0. What can be said about the inter-
sections of this line and the spiral? Each of them is a root of the
equation f (x, 0) = 0. For instance, Descartes showed that all
the conic sections can be described by equations of the second
degree, and those curves can be cut by a straight line no more
than two times. Now Newton relies on the fact that an equa-
tion of finite degree can have only a finite number of roots,
no matter how large. But the spiral in its infinite number of
gyrations crosses the line an infinite number of times. Thus,
there should be an infinite number of intersection points,
corresponding to an infinite number of roots of the equation.
This contradicts our hypothesis that the equation has finite
degree, and hence Newtons conclusion follows: there is no
such equation that gives the area of the oval.
Spirals and Seashores 65

Figure 4.3
In Newtons lemma 28, the track of the moving point forms a spiral, com-
posed of the motion of the point along the line and the uniform rotation of
the line itself about P. Newton determines the area of the curve by com-
paring the distance from the pole P to the point X after one full revolution,
which sweeps out the full area of the oval. The moving point travels an equal
distance XX = P X during the next sweep, and so forth.

This brilliant argument indicates that all simple closed

curves, such as the circle or the ellipse, have areas that can-
not be described by finite algebraic equations. The argument
seemed so simple that it made Newtons contemporaries sus-
picious. Daniel Bernoulli and Leibniz tried to state counterex-
amples, but these each involved a curve that intersects itself
(for example, the lemniscate, a figure-8 on its side, ) or that
is not closed (for example, a parabola). Later mathematicians
demanded greater rigor than Newtons beautifully simple
arguments (how can we prove that the spiral must have
66 Chapter 4

infinitely many intersections with the line?), but the basic

thrust of his insight was sustained. There are infinitely many
magnitudes that are more irrational than any radical, in the
sense that no finite root is commensurable with them. In this
sense, they are transcendental. Implicitly, Newton considers
that his proof argues the priority of geometry over algebra
by showing that a simple geometric figure includes quanti-
ties that defeat any finite amount of algebra. This makes his
own preference for geometry (and his geometrically phrased
calculus) more persuasive, for he thereby places his master
theory beyond the limitations of algebra. In Newtons view,
the ancients, with him as their modern champion, have de-
feated the upstart Descartes by subsuming algebra under the
larger umbrella of geometry.
With this in mind, it is understandable that Newton may
not have considered the solution of the quintic to be germane
to his larger project, though in his younger days he did spend
much energy in classifying cubic equations and contributing
to the advance of algebra (for instance, Newtons method
for the approximation of roots). If the real battle concerns
geometric magnitudes not expressible in finite equations,
why worry about the details of quintics? Ironically, succeed-
ing generations of mathematicians would take up Newtons
work in the simpler algebraic notation devised by his arch
rival, Leibniz. His insights about transcendental magnitudes
would be rediscovered centuries later. We will return to them
later in light of the unfolding story of the quintic.
Despite Newton, the power and beauty of the algebraic
notation ensured that interest in basic questions of algebra
remained high. New hope for the solution of equations
beyond the quartic was offered by the work of a Saxon noble-
man, Count Ehrenfried Walter von Tschirnhaus. Tschirnhaus
had varied interests. He served in the Dutch army and spent
Spirals and Seashores 67

time in Paris and England. His interest in optics led him to

set up glassworks in Italy. He was called the discoverer of
porcelain because he helped set up the pottery works at
Dresden (though the art had been long known in China). In
algebra, he sought to generalize Cardanos approach to the
cubic by setting up transformations that would simplify a
higher-degree equation. In 1683, Tschirnhaus showed that,
in a polynomial of degree n (greater than 2), the terms pro-
portional to the two next-highest powers of x (namely, x n1
and x n2 ) can always be eliminated with one of his transfor-
mations (box 4.2). Thus, in a quintic, we can always change
variables to get rid of the x 4 and x 3 terms. A century later, the
Swedish mathematician E. S. Bring showed that we can also
get rid of the x 2 term, leaving the general quintic in the much
simplified form x 5 + px + q = 0. The quintic now looked so
simple that it became increasingly perplexing why it would
not yield. Indeed, George B. Jerrard was convinced that he
had found a solution, defending it as late as 1858.
The quintics intractability was considered a challenge that
simply awaited the right assault, and it was widely believed
to be solvable until almost 1800. However, already in the
1680s, when Tschirnhauss friend Leibniz turned his atten-
tion to the solution of the quintic, he noticed what later
emerged as evidence of a fatal difficulty. Leibniz wondered
if an extension of Tschirnhauss technique could simplify the
quintic into the form x 5 equals a constant, which would be
readily solvable. Unfortunately, the equations needed to do
this are of higher degree than five. Thus, this approach would
not really simplify the problem of the quintic because it leads
to equations that are even more complicated, not less. Leibniz
thus judged that Tschirnhauss method, at least, was doomed
to fail, though he did not draw the inference that no other
method could succeed.
68 Chapter 4

Box 4.2
Tschirnhauss transformation (1683)

For any equation of the nth degree,

a n x n + a n1 x n1 + + a 2 x 2 + a 1 x + a 0 = 0,
consider a substitution that transforms the roots x1 , x2 , . . . , xn
into new roots z1 , z2 , . . . , zn , such that
z1 = b 4 x14 + b 3 x13 + b 2 x12 + b 1 x1 + b 0 ,
z2 = b 4 x24 + b 3 x23 + b 2 x22 + b 1 x2 + b 0 ,
..
.
zn = b 4 xn4 + b 3 xn3 + b 2 xn2 + b 1 xn + b 0 ,
where b 4 , b 3 , . . . , b 0 are functions in radicals of the original
coefficients a 0 , a 1 , a 2 , . . . , a n .
Tschirnhaus showed that we can eliminate the two next-
higher powers of the unknown, x n1 and x n2 by choosing
b 4 , b 3 , . . . , b 0 through solving these quartic equations. In 1843,
George B. Jerrard showed that the x n3 term could also be re-
moved by a similar transformation. Thus, these transforma-
tions allow any quintic equation to be rewritten in the much
simpler form x 5 + px + q = 0. However, in general this pro-
cedure is extremely complex; even if we use a computer, it
takes megabytes of memory to store the result.

In this period much attention was also paid to the general

question of whether all equations do have at least one so-
lution, which is the Fundamental Theorem of Algebra that
Girard and Descartes had stated without real proof. In 1748,
Jean Le Rond dAlembert offered a proof, but it was insuf-
ficient, as were the efforts of Leonhard Euler in the years
following. Despite the basic simplicity of the theorem, it
proved difficult to establish because of the vast generality
Spirals and Seashores 69

of the possibilities. That is, solving an equation can be visu-

alized (following Descartes) as finding the point (or points) at
which a curve y = x n + a n1 x n1 + + a 0 crosses a given line
(say the x axis, y = 0; see figure 4.4). We already know that
this happens for the quadratic, cubic, and quartic equations,
for those intersections are just the roots that were found by
Cardano, Ferrari, and the rest. But how can we be sure that
there isnt even one weird equation that does not cross the
line? Euler was able to give many ways to factor equations
of higher degree into products of lower-degree equations,
but, despite his immense ability and productivity (which

f(x) = x3 15x2 + 9x + 3

4
3
2
1
x
0 1 2

Figure 4.4
The graph of a quintic equation, y = f (x) = x 5 15x 2 + 9x + 3, whose three
crossings of the x axis give the three real roots of the equation y = 0. Note
that two of the roots are positive, as given by Descartess rule of signs (see
box 3.2). It turns out that these roots cannot be expressed in radicals.
70 Chapter 4

continued even after he lost his sight), he could not give a

completely general result.
That was left for the prince of mathematicians, Carl
Friedrich Gauss. Even before he was eighteen, Gauss had
made an important discovery about a special class of equa-
tions that correspond to inscribing regular polygons inside a
circle. Euclids geometry only allows the use of straightedge
and compass. Descartes showed that, in algebraic terms, this
means that solutions of the corresponding equations can in-
volve only square roots. For this reason, it seemed impossi-
ble to construct a regular polygon having a prime number
of sides greater than five without recourse to higher roots
than square roots. The young Gauss showed that, on the
contrary, a seventeen-sided polygon could be constructed,
meaning that he had found a solution of its special sixteenth-
degree equation, expressible solely in terms of square roots.
This discovery confirmed his decision to devote himself to
mathematics rather than philology.
Gauss was to give four proofs of the Fundamental Theorem
over the course of his long career. The first was his doctoral
dissertation of 1799. The last (and perhaps the most elegant)
allows x to be a complex number, with both a real and an
imaginary part, and allows the coefficients also to be com-
plex. The use of complex numbers enables Gauss to ground
general statements in a kind of picture. Figure 4.5 shows the
heart of his proof. It depicts a quintic equation, but the same
considerations will apply to any equation. Gauss represents
the behavior of the equation in terms of the polar coordi-
nates r and . The figure shows a complex plane, as defined
in figure 3.2. In that plane, we have a circle whose radius
R is chosen to be so large that the equations behavior is
dominated by the behavior of its leading term, x n (or r n , in
polar coordinates), when r is greater than R. Gauss shows
Spirals and Seashores 71

A
P
C B

Figure 4.5
Gausss diagram to establish the Fundamental Theorem of Algebra. The
shaded areas are the seas, in which the real part of the polynomial is less
than zero; the unshaded areas are land, in which the real part is greater
than zero. Walking along the shore from P to Q, at some point we must
find a root (for instance, at A). Likewise, we find roots at D and C (here, a
double root).

that one can always find such a circle, no matter what the
equation. He also shows that, going around that circle, the
real part of the equation takes alternately positive and nega-
tive values; at those points, the imaginary part is always the
opposite in sign, alternately negative and positive.
To complete the proof, Gauss draws lines showing the
regions where the real part of the polynomial is positive
72 Chapter 4

(unshaded in the figure), separated from the regions where it

is negative by lines indicating where the polynomial is zero.
This picture can be understood if we think of the shaded
(negative) region as sea and the unshaded (positive) region
as land; the lines show the seashore, where the real part of
the polynomial is zero. Now if there is a solution, both the
imaginary and the real parts of the polynomial must vanish.
So lets take a walk along the seashore from P to Q, two
points along the big circle, keeping the sea always on our
left. From Gausss earlier result, we know that the imagi-
nary part of the equation must go from being negative to
being positive as we walk along the shore. This means that
at some point in the walk the imaginary part will be zero,
and that is the root, because the real part is zero all along the
shore. This is point A in the picture; there are also roots at B
and D and a double root at C (that is, two roots that happen
to be equal; note the intersecting shorelines there).
Thus, this particular quintic has five roots (two of them
equal), and one can make a similar argument for any other
quintic, no matter what geography the seas and land have,
for there will always be shorelines curving in some manner,
along which the real and imaginary parts will both vanish at
some point. Here Gauss invokes arguments that later came
to be called topological, meaning that they make use of the
overall features of a figure that do not depend on its de-
tailed shape. By his choice of examples, it is clear that Gauss
thought he was tracking the solutions of the quintic. He had
just shown that they must exist. There they are, right at the
waters edge. The problem was simply how to express them.
5 Premonitions and
Permutations

Gausss full proof of the Fundamental Theorem was pre-

ceded by many inconclusive attempts. Mathematicians an-
ticipated that all equations would have solutions, following
the vision of Viete and Descartes, and most were not daunted
by the centuries-long roadblock at the quintic. Then, in the
late eighteenth century, Joseph-Louis Lagrange brought to
light evidence that would prove crucial. In doing this, he re-
vealed an approach that would open the way to a new kind of
mathematics. Those who came afterwards regarded his 1771
paper, Reflections on the Algebraic Theory of Equations,
as a foundational document.
Lagrange began by asking why the solutions of the cubic
and quartic had succeeded in the first place. He reviewed
the process of factoring (completing the square or the cube)
and concluded that this sort of procedure could not work
for the quintic. Of course, this did not mean that the quintic
could not be solved, but only that it was insoluble through
this well-tried method.
Assume, then, that we know the three roots of the cubic
equation, a 3 x 3 + a 2 x 2 + a 1 x + a 0 = 0; we can call them
x1 , x2 , x3 . As he analyzed these roots, Lagrange noticed that,
74 Chapter 5

Box 5.1
Roots of unity

We are looking for roots that satisfy 3 1 = 0, other than

= 1. We can start by factoring out a term ( 1), yielding
3 1 = ( 1)( 2 + + 1) = 0, as can be verified by multi-
plying it out. For = 1, we must then have 2 + + 1 = 0.
This is a simple quadratic equation whose solution is =
1 3
2
, which are the other two cube roots of unity besides 1.
The equation for the nth root of unity, n 1 = 0, can be
factored in a similar way, yielding n 1 = ( 1)( n1 +
n2 + + 2 + + 1) = 0. If = 1, then must satisfy
n1 + n2 + + 2 + + 1 = 0.

in the solution of cubic equations, a special role is played by

the three cube roots of unity, which are the solutions of the
simple equation 3 = 1 (see box 5.1). These are solutions of
the simplest possible cubic equation and thus give the basic
elements of the solution of more complicated equations. This
is not surprising. In the case of the quadratic equation, the
solution of 2 = 1 gives +1 and 1, the two fundamental ele-
ments from which all other (positive and negative) numbers
are built. Gausss work, mentioned in the preceding chapter,
also showed the general importance of the equation n = 1
in constructing an n-sided regular polygon. Clearly, the so-
lutions of this equation would also be important in finding
the solution of an nth degree algebraic equation.
Returning to the general cubic equation, Lagrange then
took the three roots x1 , x2 , x3 and formed a very simple
expression: R = (1x1 + x2 + 2 x3 )3 , the cube of the sum of
the products of each of the three roots times each of the three
cube roots of unity. At this point, Lagrange noted a simple
Premonitions and Permutations 75

and striking property. Our naming the roots x1 , x2 , x3 is quite

arbitrary; we could just as well switch the names around and
call them x2 , x1 , x3 or x3 , x2 , x1 without changing anything.
What this means is that we can permute the roots in various
ways. In fact, there are just 6 = 3 2 1 different ways to
permute the 3 roots. In general, given n quantities, there are
n! = n (n 1) (n 2) 2 1 permutations, where n!
is read n factorial (see box 5.2).
By examining the permutations of the roots, Lagrange
found a crucial clue to understanding the solution of
the equation that will be all-important in what follows. (The
point was made independently in 1770 by Alexandre-
Theophile Vandermonde.) Lagrange asked what happens to
R when we permute the roots. As box 5.3 shows, although
there are six different permutations of the three roots, they
lead to only two possible values for R. Lagrange thought
this was very remarkable and used it to show how a
cubic equation can be reduced to the solution of a quadratic
(box 5.4).
Lagrange extended this idea to higher-degree equations:
There is always a quantity analogous to R that takes a certain

Box 5.2
Permutations

Consider n objects, for instance the 52 cards in a deck. There

are n ways to pick the first one of them (52 for a deck). For
each of these, there are (n 1) ways to pick the second (51),
hence n(n 1) ways to pick the first two (52 51). Like-
wise, there are (n 2) ways to pick the third, . . . , all the
way to 1 way to pick the final item. Thus there are n! = n
(n 1) (n 2) 2 1 permutations.
76 Chapter 5

Box 5.3
Lagranges permutation of the roots of a cubic equation

Consider R1 = (1x1 + x2 + 2 x3 )3 and make the substitution

(123)
x1 x2 x3 x2 x3 x1 , which can be written symbolically as (123).
(123)
Under this substitution, R1 R2 = (1x2 + x3 + 2 x1 )3 .
(132) (132)
Under the substitution x1 x2 x3 x3 x1 x2 , or (132), R1
R3 = (1x3 + x1 + 2 x2 )3 . Of the six different permutations of
the three roots, these three are even permutations (counting
the possibility of leaving them unpermuted, the identity),
meaning that an even number of transpositions have been
made (two for R2 and R3 , zero for the identity). There are
three odd permutations left: (32), which means interchange
x2 x3 , leaving x1 alone, yielding R 4 = (1x1 + x3 + 2 x2 )3 ;
(12), yielding R5 = (1x2 + x1 + 2 x3 )3 ; and (13), yielding
R 6 = (1x3 + x2 + 2 x1 )3 . However, these expressions turn
out to be related: R2 = R1 , R3 = R1 , R5 = R 4 , and R 6 = R 4 .
As a result, there are only two distinct values for R under
these permutations, which we call t1 = R 1 = R 2 = R 3 and
t2 = R 4 = R 5 = R 6 .

number of values when the roots are permuted, and this

number is an important indicator of solvability. In the case
of the quartic, Lagrange formed a resolvent R from the four
roots (box 5.5). These roots can be permuted in 4! = 24 dif-
ferent ways, but under these 24 permutations R takes on
only three values, the three roots of a cubic, and from there
Lagrange could re-derive Ferraris solution.
Thus, in all the soluble equations so far, when the roots
are shuffled around, the resolvent R takes on fewer values
than there are permutations. This reflects the fact that the
equation can be reduced to solving one of lower degree. The
cubics resolvent has two values and the quartics resolvent
Premonitions and Permutations 77

Box 5.4
Lagranges resolvent for a cubic equation

Following box 5.3, consider again the cubic equation x 3 +

a 2 x 2 + a 1 x + a 0 = 0. Girards identity
(box 4.1) a 2 = x1 +

x2 + x3and the definitions 3 t1 = 3 R1 = 1x1 + x2 + 2 x3 ,

3
t2 = 3 R 4 = 1x1 + 2 x2 + x3 provide three linear equations

that determine x1 , x2 , x3 in terms of 3 t1 , 3 t2 , and the constant
a 2 . To find t1 and t2 , Newtons identitiesallow us to calculate
3
t1 + t2 = 2a 23 + 9a 1 a 2 27a 0 and t1 t2 = a 23 3a 1 . Then we
can express t1 and t2 as the two roots of the quadratic equation
(t t1 )(t t2 ) = 0 = t 2 t(t1 + t2 ) + t1 t2 or
   3
t 2 + t 2a 23 9a 1 a 2 + 27a 0 + a 23 3a 1 = 0,
the resolvent equation mentioned in the text. The final re-
sult is identical to the del Ferro-Cardano-Tartaglia formula.

has three, each one smaller than the degree of the equation
being permuted. When Lagrange turned to the quintic, he
found that the resolvent takes on six values, a larger number
than the degree of the equation. This showed that the method
that worked to solve equations up to quintic was breaking
down. This was also realized independently by the Italian
Gianfrancesco Malfatti (1770) and may have been noted
by Leibniz much earlier, as mentioned in the previous
chapter.
Lagrange drew from this work the moral that a new way
of attacking the quintic was needed, and he probably hoped
that his work would enable him to find it, if only by steering
him away from the false paths of the past. But he was not yet
close to being convinced that the quintic is not solvable. What
he did argue is that it is very doubtful that the methods we
have just discussed can give a complete solution of equations
78 Chapter 5

Box 5.5
Lagranges resolvent for the quartic equation

For the quartic equation x 4 + a 3 x 3 + a 2 x 2 + a 1 x + a 0 = 0, con-

sider the four roots x1 , x2 , x3 , x4 and the function R = (x1 +
x2 x3 x4 )2 . As we go through the 24 permutations of the
4 roots, R takes on only 3 distinct values:
R1 = (x1 + x2 x3 x4 )2 ,
R2 = (x1 + x3 x2 x4 )2 ,
R3 = (x1 + x4 x2 x3 )2 .
As in the case of the cubic, the four roots can be expressed
as
simple
combinations
of these different values of R, namely
R1 , R2 , R3 , and the condition that the sum of all four
roots equals a 3 . In turn, various sums and products of the
resolvent can be expressed in terms of the coefficients of
the original quartic equation, using Newtons identities.
From these expressions, Lagrange showed that R must satisfy
a cubic equation. Having solved this cubic equation, whose
three roots are R1 , R2 , R3 , we can recover x1 , x2 , x3 , x4 , which
are exactly as given by Ferraris formula.

of degree five and, all the more so, of higher degrees. In the
process of working out the resolvent for the quintic, he noted
that the computations are so long and complicated that they
can discourage the most intrepid calculators, by which he
means human calculators, in that pencil and paper era. For
him, all this reflected an uncertainty [that] will discourage
in advance all those who might be tempted to use [the older
methods] to solve one of the most celebrated and important
problems of algebra.
Thus Lagrange still remained hopeful that new methods
might succeed where the old ones failed. His contemporary,
Premonitions and Permutations 79

the mathematical historian Jean Etienne Montucla, expressed

this optimistic spirit in a military metaphor, as if the quintic
were a town under siege: the ramparts are raised all around
but, enclosed in its last redoubt, the problem defends itself
desperately. Who will be the fortunate genius who will lead
the assault upon it or force it to capitulate?
Here Gauss came to a different conclusion. In his doc-
toral dissertation of 1799, he first recorded his belief that the
quintic equation has no general solution in radicals. Gauss
announced this possibility along with his first proof of the
Fundamental Theorem of Algebra. In 1801, Gauss published
his opinion in his masterwork, Disquisitiones Arithmeticae,
a compendium of many remarkable results. Among them,
Gauss included his famous work on constructing regular
polygons, which involved the question of the solvability of
their equations. It was natural that, in this context, his
thoughts turned to the more general problem of solvabil-
ity. Everyone knows that the most eminent geometers have
been ineffectual in the search for the general solution of equa-
tions higher than the fourth degree. . . . And there is little
doubt that this problem does not so much defy modern meth-
ods of analysis as that it proposes the impossible. Gauss,
knowing the extent of his own powers, would have been the
last to give up had he not divined far deeper problems than
those his ingenuity and persistence might solve.
But while Gauss guessed this new level of difficulty, he did
not explain why the quintic is unsolvable. He never seemed
to be attracted to permutations and combinations, so it is
not clear that his opinion was founded on the kind of per-
mutational arguments that Lagrange had made. All the more
striking, then, that even before Gauss announced his opinion,
a little-known mathematician in Italy formulated and sought
80 Chapter 5

to prove the assertion that the quintic is unsolvable in rad-

icals. The son of a physician, Paolo Ruffini was born in a
small Italian town, moving in childhood to Reggio near the
city of Modena (figure 5.1). As a child, he seemed destined
for the priesthood, but he finally chose to study philosophy,
medicine, and surgery, graduating in 1788. He served as a
professor of mathematics in Modena until the upheavals in
the wake of the French Revolution hit Italy. Though not a
revolutionary himself, Ruffini refused to swear a loyalty oath
to the French and was barred from his university post until
1799. During this enforced absence from academic life, his
mind was preoccupied with mathematics, though the bulk
of his time was devoted to medicine. At the end of the day, he
would jot down his thoughts and ideas on letters he received
and in the margins of papers he read. His thoughts turned to
the theory of equations as he prepared his first work, which
included his first proof of the impossibility of solving the
quintic.
Ruffinis devotion to medicine was deep, despite his math-
ematical preoccupations. In 1817 to 1818, there was an epi-
demic of typhoid fever, during which Ruffini tended to the
sick with great dedication and courage. He himself fell ill
with the disease. Characteristically, he used his experience
to write a paper on contagious typhoid. He also maintained
his philosophical and religious interests, writing about the
definition of life and against Laplaces mechanistic theory
of morality. His contemporaries noted his self-effacing mod-
esty and spirituality. Indeed, in later life Ruffini was offered
a mathematical post at the University of Padua, but declined
because he was reluctant to leave his many patients.
Perhaps because of this modesty, Ruffini has been some-
what neglected, despite his deep merits and the singular
interest of his dual life as physician and mathematician.
Premonitions and Permutations 81

Figure 5.1
Paolo Ruffini.
82 Chapter 5

Though he did not seem to covet the usual rewards of pro-

fessional glory, however, Ruffini was tenacious in his quest
to prove the unsolvability of the quintic and to bring his
work before the mathematical world. He published no fewer
than six versions of his proof (the first in 1799, the last in
1813), several of them responding to questions and criti-
cisms raised both by his friend Pietro Abbati and by the older
mathematician Malfatti, who could not accept the notion of
unsolvability.
He began his first proof by announcing that the algebraic
solution of general equations of degree greater than 4 is al-
ways impossible. Behold a very important theorem which
I believe I am able to assert (if I do not err); to present the
proof of it is the main reason for publishing this volume. The
immortal Lagrange, with his sublime reflections, has pro-
vided the basis of my proof. Indeed, Ruffini was among the
first to grasp the immense importance of Lagranges idea of
permutations (as Ruffini called them) and was the first to
offer a proof of the unsolvability of the quintic.
Unfortunately, Ruffini expressed these proofs in long and
obscure forms, which rendered them hard to understand.
He sent his proof to Lagrange, hoping for recognition from
the man whose work had given him his inspiration. Lagrange
never replied, a bitter disappointment to Ruffini and his
friends, for they considered Lagrange (born in Turin, though
of a French family) as a fellow countryman who they thought
should recognize this Italian achievement. Later, Lagrange
was on a committee to examine Ruffinis memoir, but he was
cool toward it, finding little in it worthy of attention. As it
later emerged, Lagrange (then elderly) may have had trouble
wading through the paper and eventually told a third party
that it was good but did not give sufficient proof of some
Premonitions and Permutations 83

of its assertions. Indeed, he may not have spent much time

studying Ruffinis proof.
Other mathematicians were more favorable. Some read-
ers at the Royal Society were quite satisfied with Ruffinis
proof. Most important, the great French mathematician
Augustin Cauchy recognized the importance of Ruffinis
work and generalized some of its results during 1813 to
1815 (as will be discussed later and in appendix C). One can
imagine Ruffinis gratification when, six months before his
own death, he received a letter from Cauchy, who wrote that
your memoir on the general resolution of equations is a
work that has always seemed to me worthy of the attention
of mathematicians and that, in my judgment, proves com-
pletely the unsolvability of the general equation of degree
greater than 4. Cauchy also told Ruffini that he had cited
this work in his lectures.
Despite Cauchys strong endorsement, Ruffinis work still
did not achieve general acceptance among mathematicians.
The reasons are unclear but seem to center on the novelty of
arguments from permutations and even more on the radical
notion that some equations might not, in general, be solvable
in radicals. This might have been behind Lagranges wari-
ness, for he described Ruffinis proof as a difficult matter,
something that he himself thought improbable, if not incredi-
ble, and for which he thought Ruffinis reasoning insufficient.
To be sure, later mathematicians confirmed Lagranges con-
tention that there was an important missing step in Ruffinis
arguments. Thus Ruffinis proof fell short, at least in the sense
that it left others unconvinced. He deserves credit for his
vision and his courage, but it remained for others to resolve
the issue decisively.
This page intentionally left blank
6 Abels Proof

Ideas, especially mathematical ideas, have a life that goes be-

yond their human creators. Without knowing Ruffinis work,
Niels Henrik Abel (figure 6.1) discovered the same argument
only a few years later and also gave the first essentially com-
plete demonstration. Abels story is one of the most moving
in the history of mathematics. In his twenty-six years of life,
he discovered whole new territories in mathematics, though
struggling constantly with poverty and misunderstanding.
His native Norway was then in its youth as an independent
nation, and Abels checkered fortunes to some extent mir-
rored Norways struggle.
When Abel was 12, Norway separated from Denmark, long
the dominant partner in their dual kingdom, and set up her
own parliament, the Storting. But Norway could not com-
pletely separate from her more powerful neighbors. In 1814,
Abels father was among the delegates sent to offer the crown
of Norway to the reigning Swedish king, Karl XIII. Abels
father was a pastor, like his father before him, and had a
notable career as a member of the Storting. He was a man
of the Enlightenment who read Voltaire and was active in
the movement for literacy and vaccination in rural Norway.
He wrote several popular catechisms and books of prayer
86 Chapter 6

Figure 6.1
Niels Henrik Abel.
Abels Proof 87

doubtless read by his son. He was interested in natural sci-

ence and woke his children to see lunar eclipses. However,
his political career ended ignominiously when he pressed
unfounded accusations against certain powerful persons.
He died an alcoholic, leaving nine children and a widow
who also turned to alcohol for solace. After his funeral,
she received visiting clergy while in bed with her peasant
paramour.
Abel, then eighteen, found himself without support and
obliged to act as the responsible adult for his younger sib-
lings. Somehow, though, he continued his education. He had
been fortunate to find a mentor in a young mathematics
teacher, Berndt Michael Holmboe, who inspired him and
became his lifelong friend. Abel soon showed an amazing
ability to solve difficult problems. While still a high school
student, he read Lagranges work and Cauchys 1815 pa-
per on permutations. (Though Cauchys paper was based on
Ruffinis work, Abel did not read Ruffini, perhaps because
his works were hard to find and in Italian.) Holmboe rec-
ognized and extolled his students excellent mathematical
genius. In 1821, Abel entered the Royal Fredericks Univer-
sity, which had opened in 1819 in the new capital, Christiana
(now called Oslo), then a city with only 11,000 inhabitants.
He continued to make rapid strides, beginning to write orig-
inal papers and going far beyond the skill of his teachers.
With admirable understanding, those teachers proposed that
he be granted a special fellowship to visit Paris and Berlin,
though in fact he had little to learn even from the greatest
mathematicians of the time.
While always modest, Abel during this time set himself
to solve the most difficult and famous problems. In 1823,
he tried his hand at Fermats celebrated last theorem, to show
the impossibility of finding integers a , b, and c that would
88 Chapter 6

satisfy the equation a n = b n + c n , where n is an integer greater

than 2. Not surprisingly, he found himself at the end of my
tether, as he put it, for this problem resisted solution until
1993, when it finally succumbed to very elaborate abstract
techniques. Even so, he found what now are called Abels
formulas, which showed that if any solutions existed, they
would have to be extremely large numbers. Abel also turned
his mind to the problem of the quintic equation. At first, he
thought he had managed to solve it and was very excited.
This was in 1821, twenty-two years after Ruffini had pub-
lished the first version of his proof. At this point, Abel still
did not know Ruffinis work, which is not surprising consid-
ering the isolation of Norway and its lack of mathematical
libraries. Indeed, during the winter months, the Oslo fjord
remained frozen and mail was often delayed.
Though Abel had probably taken note of Gausss opinion
that the quintic was unsolvable, he nonetheless refused to
give up the search for a solution. After all, Gauss offered no
proof for his assertion, and Abel, like Lagrange and most
other mathematicians, could well consider that the search
was not over. However, when his teachers asked him to give
some numerical examples of his solution to the quintic, Abel
soon realized that it was not as general as he had thought.
This was a decisive moment, for he could have stubbornly
resumed the quest for a fully general solution to repair the
gaps in his work, on the assumption that a solution had
to exist. Instead, he made an about-face and turned his ef-
forts to proving the unsolvability of the quintic. He never
explained his reasons, and one wonders what moved him.
In 1824, two years after Ruffinis death, Abel published his
first proof of the unsolvability of the quintic, which in many
ways is close to Ruffinis proof, although it fills in an im-
portant gap that Ruffini had not noticed. In 1826, Abel read
Abels Proof 89

anonymous articles summarizing the work of Ruffini, which

Abel acknowledged in his final (posthumous) paper: The
first, and if I am not mistaken the only one, who before me
had tried to show the impossibility of the algebraic solution
of general equations is the geometer Ruffini. But his paper is
so complicated that it is very difficult to decide the correct-
ness of his reasoning. It seems to me that his reasoning is not
always satisfactory.
Because Ruffinis work overlaps largely with what Abel
went on to do, I will not discuss it separately. I do not mean
to belittle the recognition Ruffini is due. The theorem of un-
solvability may better be called the Abel-Ruffini Theorem.
Yet Abels remark, as well as Lagranges hesitation, indicate
aspects of decisive proof in which Ruffini fell short. A proof
that lacks a decisive step is not yet a proof. Accordingly, I
will summarize Abels version, indicating along the way the
common elements in their arguments and the crucial gap in
Ruffinis reasoning that Abel filled.
If you wish to read Abels own words, appendix A contains
a translation of his 1824 paper. Abel had this earliest version
of the proof printed at his own expense as a pamphlet, hoping
to use it as a calling card that would gain the attention of
the great mathematicians, Gauss above all. However, to save
paper and money, Abel compressed his argument to tele-
graphic terseness, which he amplified in his later accounts of
the proof. Accordingly, I have added a running commentary
to help the reader follow Abels argument. Appendixes B and
C fill out details he merely mentions. Here I will present the
four crucial stages of the proof as Abel presented them in
1826, without spelling out all the technicalities to be found
in the appendixes. In each case, I will present a summary
statement, followed by my explanation.
90 Chapter 6

Abel uses the time-honored method of reductio ad absur-

dum: he begins by assuming that the quintic is solvable and
shows that this leads to a contradiction. His first step is to
specify the form that a solution must have. Given a general
equation of the mth degree (taking m as a prime number, so
that it cannot be factored further),

a m ym + a m1 ym1 + a m2 ym2 + + a 2 y2 + a 1 y + a 0 = 0, (6.1)

Abel proves a very general statement about any solution,

which he calls an algebraic function:

(I) All algebraic functions y can be expressed in the form

1 2 m1
y = p + R m + p2 R m + + pm1 R m , (6.2)

where p, p2 , . . . are finite sums of radicals and polynomials and

1
R m is in general an irrational function of the coefficients of the
original equation.
That is, if y is the solution of an algebraic equation of degree
m, y can be expressed as a series of terms that contain nested
roots involving the coefficients and irrational expressions like
1 1
R m . (Remember that R m is just another way of writing the
mth root of R, m R.) This is very close to the general form that
Euler had already conjectured some years before. Abel gives
a detailed proof (see appendix B, including some subtleties
neglected here), but the idea can readily be illustrated with a
few examples. As box 6.1 shows, the solution of the quadratic
1
equation can be expressed as y = p + R 2 , which is just the
general form above with m = 2 and p a simple polynomial.
Likewise, the solution of the cubic can be expressed as y =
1 2
p + R 3 + p2 R 3 , which follows the general form, with m = 3.
In both of these cases, m = 2 and m = 3, m is a prime number.
For the quartic, m = 4 is not prime and the general solution
Abels Proof 91

Box 6.1
Abels form for the quadratic equation

In the quadratic equation, y2 a 1 y + a 0 = 0, substitute y =

1 1
p + R 2 to yield ( p 2 + R a 1 p + a 0 ) + (2 p a 1 )R 2 = 0. To
satisfy this in general, each parenthesis must be separately
zero (as Abel discusses in [A8] in appendix A), so p = a21 and
a2 
then R = a 0 + 41 . Thus, y = a21 12 a 12 4a 0 , the fami-
liar form of the quadratic solution. For the case of the cubic
solution, see p. 174 in appendix B.

involves only combinations of the quadratic and cubic forms,

as was discussed earlier.
In the case of the quintic, Abels general form (6.2) becomes
1 2 3 4
y = p + R 5 + p2 R 5 + p3 R 5 + p4 R 5 . (6.3)

Because this follows from his general result about the form of
the solution, either the solution of the quintic has this form,
or there is no such solution. So Abel assumes hypothetically
that the quintic does have a solution of exactly this form. He
now goes on to show that this form leads to a contradiction.
This requires three further steps.
The next step is crucial; Ruffini had assumed it without
giving a proof, but Abel remedies this.
(II) All algebraic functions y can be expressed in terms of
rational functions of the roots of an equation.

The idea here is simple but telling. In equation (6.3), the

general form of y is expressed in terms of polynomials and
1 2 3 4
also various irrational functions (here, R 5 , R 5 , R 5 , R 5 ) of the
coefficients. But Abel brilliantly proves that we can express
92 Chapter 6

Box 6.2
The relation between roots and coefficients
1
Let y = p + R 2 , as shown in box 6.1, which also shows
a2 a
that R = a 0 + 41 . Now consider the two roots, y1 = 21 +
 
1
a 12 4a 0 , y2 = a1
1
a 12 4a 0 . Then (y1 y2 ) =

2 2 2
 a2 
a 12 4a 0 and thus (y1 y2 ) = a 12 4a 0 = 4 a 0 + 41 = 4R.
2

The name discriminant is given to a 12 4a 0 , which is zero

when the roots are equal, y1 = y2 , positive when the roots
are real, and negative when they are imaginary.

y in terms of the roots instead of the original coefficients of

the equation. What is important here is that all the various
irrational functions of the coefficients that appear in y
1
(e.g., R 5 ) are rational functions of the roots of the equation.
To choose a simple example, in the quadratic equation
1
above, whose general solution is y = p + R 2 , box 6.2
shows that 4R = (y1 y2 )2 , where y1 , y2 are the two roots
1
of the quadratic. So 2R 2 , the square root of 4R, is equal
to the difference of the two roots, (y1 y2 ). In agreement
with Abels proof, this is indeed a simple rational function of
the roots.
1
In the case of the quintic, Abels step II implies that R 5 is a
2 3 4
rational function of the roots, as are its powers, R 5 , R 5 , R 5 ,
as well as p, p2 , . . . . Because it is made up of products of
all these rational functions, y is thus a rational function of
the roots. Here there is an interesting tension: The roots are
irrational functions of the coefficients, but those coefficients
are always rational sums or products of the roots. This harks
back to Girards identities (see box 4.1), which showed that
the coefficients were sums and products of the roots.
Abels Proof 93

Let us return to the tension between the irrationality of the

roots as functions of the coefficients, on the one hand, and
the rationality of the coefficients as products of the roots, on
the other. Roughly speaking, as the degree of the equation
grows higher, it is harder and harder to reconcile this tension.
Finally, with the quintic equation, it is no longer possible, and
we cannot, in general, find a solution in radicals. However,
this idea must be amplified much further.
The next step limits the hypothetical solution in a way that
will prove to be decisive.

(III) If a rational function of five quantities takes fewer than five

values when the five quantities are permuted, it can take only two
different values (equal in magnitude and opposite in sign), or one
value, but never three or four values.

Abel drew this theorem from the work of Cauchy, of which

it is a special case. Cauchys proof is presented in appendix C.
It relies on looking at the different ways we can permute the
roots of the equation. Abel also relies on the Fundamental
Theorem of Algebra: A quintic equation must have at least
one root and no more than five different roots. So there can
be no more than five values for y, which, according to step II,
is a rational function of the roots of the equation.
As before, let us consider any of the expressions in the
1
solution that is a rational function of the roots, such as R 5 .
1
Cauchys result requires that R 5 can take only one, two, or
five values as the roots are permuted, but never three or four
values. This gives Abel leverage to show that the solution
1
cannot work. First, note that R 5 cannot, in general, take only
one value, because then it could lead to a single solution,
not five roots, which we have assumed to be unequal. Then,
1
Abel investigates what happens if R 5 takes on five values
94 Chapter 6

(appendix A, [A15][A16]). This leads to a contradiction; he

1
shows that R 5 , having five possible values when the roots are
permuted, would have to be equal to an expression having
120 values, which is impossible.
That excludes the possibility that there are five values
for y. So Cauchys result must require that there be only two
(appendix A, [A27]). But this too leads to a contradiction,
because when we switch the five roots around, Abel shows
that we get an inconsistent result again. He derives an equa-
tion whose left-hand side has 120 possible values, while the
right-hand side has only 10. Clearly, such an equation cannot
be solved in general, and so the hypothetical solution leads
to absurdities. Therefore Abel concludes that

(IV) It is impossible to solve the general equation of the fifth

degree in radicals.

The strategy of Abels argument is straightforward: he

takes the only possible form a solution could have and shows
that it leads to contradictory results when we permute the
roots of the equation. This contradiction rests on a special
property of the number five, shown by the number of values
the hypothetical solution can take when subjected to permu-
tations of its five roots. The argument also applies for degrees
higher than five. For instance, we can multiply an unsolvable
quintic by a factor of y = y 0, and it would then be a sixth-
degree equation that has one root y = 0 and five unsolvable
roots, and similarly for any higher degree.
Though Abels argument shows the impossibility of solv-
ing the quintic in general, it still seems opaque. The ques-
tion remains: why this impossibility? To examine this further
means we must seek the heart of Abels proof.
7 Abel and Galois

Abel published his proof shortly before he began his travels

abroad, hoping that it would open doors for him. He knew
that this trip was a godsend, his only chance to make a mark
on the larger world. He was so poor that he barely scraped
by at home; only the stipend given him by his university
made such a dream voyage possible. Because he was shy
and prone to loneliness, he decided to travel with friends,
fellow Norwegians also pursuing studies abroad. For this
reason he did not go directly to Paris, the mathematical
center of the world at that time, but headed rather to
Berlin. He hoped to meet Gauss, but the great man received
very few visitors, turned inward on his own concerns. He
received Abels proof but did not cut its pages, setting it
aside as if it were from a crank: Here is another of those
monstrosities!
In Berlin, Abel did meet August Crelle, an amateur math-
ematician of some distinction who had founded a journal
of mathematics that would become so well known that it
would be called simply Crelles Journal by its readers. Crelle
was hospitable to Abel and soon recognized his new friends
extraordinary gifts. He published seven of Abels papers in
the first volume of his journal and also began working to find
96 Chapter 7

him a position in Berlin. For Abel, this attention buoyed his

hopes that he might secure a living as a mathematician that
would allow him to marry his fiancee, Christine Kemp, who,
like Abel, had no means and remained in Norway working
as a governess.
Despite the requirement that he spend the bulk of his time
in Paris, Abel could not resist accompanying his friends on
their journeys to Italy and Switzerland. His letters pleaded
with his professors in Norway not to disapprove of his desire
to see these fabled places, even though the trip strained his
precarious finances to the limit. Though he was then in good
health and young, he had a premonition that this would be
his only chance to see the world.
At last, Abel took leave of his friends and steeled himself to
face Paris alone. There, after all, he could meet Adrien-Marie
Legendre and Cauchy, eminent members of the Academie
des Sciences, well qualified to appreciate Abels achieve-
ments and promote his career. His reception was chilly. His
formal visits to the great men were received politely but with-
out real warmth or recognition. They paid no attention to his
proof. They were more absorbed in their own concerns and
saw him only as one more bright young man among so many
that had come seeking their favor over the years. Abel wrote
to a friend that the French were monstrous egotists . . . un-
commonly reserved with respect to foreigners. . . . Everyone
works by himself here, without bothering others. Everyone
wants to teach and no one wants to learn. Though Abel was
disappointed, he kept working diligently and continued to
send Crelle more and more papers, as he spread his wings
ever further.
He did make an impression on some Frenchmen, at least.
Jacques Frederic Saigey, the mathematical editor of Baron de
Ferrusacs Bulletin, an important journal of the day, asked
Abel and Galois 97

Abel to write brief accounts of articles in other journals, in-

cluding his unsolvability proof for the quintic. It was at this
point (1826) that Abel became aware of Ruffinis work and
noted its unsatisfactory quality. Abel also met an important
self-taught scientist, Francois-Vincent Raspail, an ardent par-
tisan of the French Republic who was several times exiled or
imprisoned. In his scientific work, he had made steps toward
the development of the cell theory of biology and the notion
of microbe-borne disease. Raspail was quite struck by Abel.
Despite his loneliness and dwindling funds, Abel found
diversion in Paris. He played billiards and indulged his pas-
sion for the theater, which he described in animated letters.
At the end of 1826, his funds ran out and he had to return
home months early. He hoped to find a post in Norway, but
the only vacancy in the whole country had been filled during
his absence, ironically by his teacher and mentor, Holmboe.
Abel was not bitter, and their friendship continued, but he
must have been deeply perplexed. He could not think of mar-
rying in such penury; he had to eke out his living as a substi-
tute teacher in a military academy. Nevertheless, he had not
stopped thinking about his proof of unsolvability, seeking its
larger implications.
In March 1828, he came to a crucial realization. In a new
paper he sent to Crelle, Abel wrote that Although the alge-
braic solution of equations is not possible in general, there
are nevertheless particular equations of all degrees that ad-
mit such solutions. Such are, for example, the equations of
the form x n 1 = 0, which Gauss had solved so brilliantly
when younger than Abel. The solution of these equations is
founded on certain relations among the roots. I have sought
to generalize this method by supposing that two roots of a
given equation are so connected that one can express ratio-
nally the one by the other, and I have come to the result that
98 Chapter 7

such an equation can always be solved with the help of a cer-

tain number of equations of lower degree. There are even cases
where one can solve algebraically the given equation itself.
Abel gives this interrelation of the roots a simple form.
He calls the first root x1 ; then he assumes that the next root
x2 = f (x1 ), where f (x1 ) is a rational function of x1 . He then as-
sumes that the next root is the same function of the previous
root, x3 = f (x2 ) = f ( f (x1 )), which he calls f 2 (x1 ), meaning
the function f applied two successive times. He then keeps
applying f to each successive root and gets the series of roots
x1 , f (x1 ), f 2 (x1 ), f 3 (x1 ), . . . , f n1 (x1 ), where n is the degree
of the equation. For instance, if n = 5, x6 = f 5 (x1 ) = x1 ,
since the equation can have only five different roots, which
one can arrange in a circle,

X1

X5 X2

X4 X3

showing the cyclical symmetry of the roots of such solvable

abelian equations, as they are now known.
Abels insight is that, in these solvable equations, any two
roots are related by a rational function. If this is true, then
Abel and Galois 99

all the roots are rational functions of each other and of the
coefficients of the given equation. We need one further con-
dition, however, and it is crucial. It is implicit in the case
we have considered, but becomes explicit when Abel con-
siders an equation whose degree can be factored into several
primes, so that the roots can be grouped into different cycles
if the equation is solvable. If x is a root of such an equation
and two other roots are given by f (x) and g(x), which are
two (possibly different) rational functions of x, then Abel
concludes that the equation is always solvable if

g( f (x)) = f (g(x)).
That is, if the order in which these two functions are applied
to x does not matter, then the equation is solvable.
This is the insight that I consider most helpful. At first, it
may seem merely formal or devoid of significance. Abelian
equations are solvable if their roots are related by functions
such that it does not matter in what order we apply the func-
tions. Contrariwise, the equation may not be solvable if the
order does matter. But a great surprise is hidden in this seem-
ingly flat statement. Until this point, none of the basic opera-
tions of arithmetic and algebra has been noncommutative.
Thus, a + b = b + a and a b = b a . These so-called com-
mutative laws express an important quality of numbers, and
they hold sway in the operations used in every equation, of
whatever degree.
Abels insight connects solvability with commutativity, at
least for abelian equations. Thus, he opened the whole issue
of commutativity for consideration. Indeed, I believe that this
was the first time that the possibility that operations might
not commute emerged in mathematics, especially in the con-
text of simple equations where we would least expect such
100 Chapter 7

a strange thing. Though Abel did not live to see it, his suc-
cessors gradually extended and elucidated the subtle con-
nection between noncommutativity and unsolvability. In the
rest of this book, I hope to follow this story and explore its
implications.
Notice, first of all, how Abels insight into solvable equa-
tions squares with this. He was able to organize the roots of
abelian equations into the cyclic pattern shown above. It is
clear that in such cases the relations between the roots are
commutative, for it does not matter whether we go around
the circle backward or forward. Abel emphasizes this com-
mutative symmetry as the crux of this pattern. Yet this insight
is only a beginning, for it does not explain how to determine
whether the pattern of a given equation is or is not cyclic in
this way. Abel must have seen that he needed to take it much
further. His 1828 paper leaves us wondering how he might
have carried his question about commutativity even further.
Just at this point, however, Abels brief strand of life ran
out. He was ill at the beginning of 1828 and did not know how
widely his papers were being read, for Norway still remained
isolated. Although he was not given to undue enthusiasm,
Gauss now spoke of the depth, delicacy, and elegance of
Abels work, and it may be that Abel heard of this through
Crelle. He probably did not yet know of the admiration of
Legendre or Carl Gustav Jacob Jacobi, who was vying with
Abel to extend his results still further. At age twenty-three,
Jacobi was already an associate professor at the University of
Konigsberg. Indeed, both Jacobi and Abel were proposed for
membership in the Institut de France in 1828, though neither
was elected. Crelle was working hard to find a place for Abel,
who described himself as being as poor as a churchmouse
in a note begging an old friend for a loan, which he signed
Yours destroyed.
Abel and Galois 101

Crelle was not the only one trying to help Abel. Four dis-
tinguished French mathematicians (including Legendre and
Simeon-Denis Poisson) wrote to the king of Sweden implor-
ing some help for one of his own subjects, a young math-
ematician, Monsieur Abel, whose works show that he has
mental powers of the highest rank and who nonetheless
grows ill there in Christiania in a position of too little value
for one of his so rare and early-developed talent. The king
never responded, and Abel probably never knew of this
attempt to intervene on his behalf.
For his part, Abel continued to work, not just on the solv-
ability of equations but also on elliptic functions and what
now are called Abelian integrals, making fundamental con-
tributions. He began to receive some letters from Legendre
that comforted him by their frank admiration and interest
in his work, even though Legendre seemed to have lost an
important paper that Abel had sent to the Institut years be-
fore (which Abel did not mention in his replies). Generously,
Abel wrote back that Legendres recognition gave him one
of the happiest moments of his life, for I would have ac-
complished nothing without having been led by your light.
With childlike pride, Abel quoted such praise to his friends.
He also wrote: I am, however, almost completely alone. I
assure you that in the most profound sense I am not in asso-
ciation with a single human being. Nevertheless, this lack of
friends is not foremost in my mind because I have so horribly
much to do for [Crelles] Journal.
At Christmas, 1828, Abel could not resist the opportu-
nity to spend the holiday with his fiancee in the country,
even though the winter was exceptionally cold. He was in-
creasingly ill, but his friends could not dissuade him from
going. He set off, with only socks to warm his hands. After
the Christmas balls and festivities, he took to his bed with
102 Chapter 7

pneumonia. He went on to complete one more paper on tran-

scendental functions, in order to record part of the treatise lost
in Paris. Crelle wrote him that he was almost sure of having
secured him a position in Berlin, which would allow him
financial security and the possibility of marriage at last. But
the matter was not completely resolved. Abel kept denying
that he was dying of tuberculosis. During his last days, he
was brave. During the nights, his anger and despair fright-
ened those around him. He cursed his poverty and the in-
justice of the neglect he had suffered, but still made light
of his condition. He died on April 6, 1829, in the presence
of his fiancee Christine. Two days later, not knowing of his
friends death, Crelle sent final word that Abels appoint-
ment was approved. On his deathbed, Abel had sent word
to his friend Keilhau urging him to marry Christine after his
death. Though he had never met her, Keilhau proposed in a
letter and they were soon married.
Thus, Abels new insight of 1828 appeared in print only
after his death. We can only speculate where he might have
taken these ideas. Yet even before Abels death, and without
knowing his final work, a young Frenchman had already
taken the steps that would generalize Abels insight into a full
account of the solvability of equations. The story of Evariste
Galois (figure 7.1) stands beside that of Abel among the most
dramatic lives in mathematics. Galoiss story is, however, far
better known, elaborated and romanticized in novels, films,
and biographies. Abel died at twenty-six, but Galois died at
only twenty years old, in a mysterious duel linked to love
and politics.
Galois was born in 1811 in Bourg-la-Reine, just south of
Paris. His family had adopted the ideals of the Revolution
of 1789, which ended more than eight hundred years of rule
by the Bourbon-Capetian dynasty. After Napoleons first
Abel and Galois 103

Figure 7.1
Evariste Galois.
104 Chapter 7

downfall and exile to Elba, however, France turned back to

its royalist past, and Louis XVIII took the throne vacated by
his executed brother, Louis XVI. When Napoleon attempted
a comeback from Elba, Galoiss father became the mayor of
Bourg-la-Reine, a position he held for fifteen years, even after
Napoleons defeat at Waterloo in 1815. At age eleven, Evariste
entered the famous Parisian College de Louis-le-Grand as a
boarding student. Despite its great name, the school was as
grim as a prison. The students rose at 5:30 A.M. to study in
silence for two hours before a breakfast of bread and wa-
ter. Often punished, they were put in solitary cells for days
at a time. By now Louis XVIII had returned to power, and
the schools administration was rigidly royalist. Nonethe-
less, Galois made brilliant progress and won prizes in Greek
and Latin.
Like Abel, Galois was helped by his teachers, first
Hippolyte Jean Vernier and then Louis-Paul-Emile Richard.
Though shy, Richard went out of his way to seek oppor-
tunities for his amazing student. He approached Cauchy,
who presented one of Galoiss papers to the Academie des
Sciences in 1829, to be reviewed by himself and other
distinguished members. It was a crushing disappointment
when that judgment never came; indeed, Galois never even
got his manuscript back, for unclear reasons. Galois blamed
Cauchy for losing the manuscript, but some documents sug-
gest instead that Cauchy urged Galois to revise and expand
his work. Galois might well have been somewhat paranoid,
and stories told after his death do reveal a darker side. The
mathematician Sophie Germain describes Galois attending
the sessions of the Academie and insulting the speakers.
The romantic myth of misunderstood genius ignores such
provocative and probably outrageous behavior. Yet the
whole story is sadly reminiscent of Abels disappointment in
Abel and Galois 105

Paris in 1826 to 1827, when his submission to the Academie

also disappeared. (Abel described Cauchy as a Catholic
bigot.) Though they were in Paris at the same time, it does
not seem that Abel and Galois ever met, though we might
imagine them passing each other, unknowing, in the street.
Certainly they were pondering the same questions about the
solvability of equations in 1828, though Galois was not aware
of Abels work as he formed his own theory.
Those years were terrible for Galois. Louis XVIIIs suc-
cessor, Charles X, was even more desirous of a return to
the ancien regime, and the clerical party was also gaining as-
cendancy. In their little town, the Jesuit parish priest forged
Galoiss fathers name to scurrilous epigrams directed to his
own relatives. In the ensuing scandal, Galoiss father felt
forced to leave the town and was so humiliated that he com-
mitted suicide in 1829. The grieving Evariste saw the parish
priest insulted and hit with stones at his fathers funeral. That
summer, he took the entrance examinations for the presti-
gious Ecole Polytechnique. During them, Galois was ques-
tioned on logarithmic series. When his examiners asked him
to prove his statements, he refused, saying that the answers
were completely obvious. His obstinacy ended his dream of
being a polytechnicien. Through the intervention of Richard,
Galois was able to gain admission to the Ecole Preparatoire
(now called the Ecole Normale Superieure), where he did
not bother to hide his scorn for his teachers and studied only
mathematics.
In 1830, rioting began in Paris against the government;
Galois wanted to join in the fight but was kept locked up
in the school. In July, Charles X, driven from Paris, was un-
able to regain power. The old General Lafayette, a moderate
like his friend George Washington, allowed Charless dy-
nastic rival, the duc dOrleans, to become Louis-Philippe
106 Chapter 7

I, the peoples king. Galois and many others considered

this July monarchy a grotesque betrayal of their struggle
to restore the Republic. Along with Raspail, Galois joined
the radical Societe des Amis du Peuple and became more
active politically, for which he was finally punished by be-
ing expelled from school. His situation then became even
more strained. He tried to eke out a living by giving private
lessons in mathematics. He sent yet another paper outlin-
ing his theory of equations to the Academie, which this time
rejected it, infuriating him even further. During a banquet
of the Societe, Galois raised his jackknife and sarcastically
toasted To Louis-Philippe! In the commotion, he was not
heard as he continued: if he betrays his oaths.
Though acquitted of the charge of treason, Galois soon
was put in prison on other charges, along with Raspail, who
later described how Galois, who still looked like a child, kept
working on his mathematics even there. Taunted by the other
inmates, Galois tried to show he could hold his liquor. On
one such occasion, Raspail remembered that Galois had to
be restrained from a drunken attempt at suicide. On another,
Galois escaped a bullet fired by a guard into his crowded cell.
In 1832, the scandal about the Instituts treatment of Abel
came out, and Galois in prison vented his anger about the
very men who already have Abels death on their conscience,
whom he blamed for his own neglect. That year, the republi-
cans were unable to organize further demonstrations, for so
many of their leaders were in prison. When an outbreak of
cholera struck Paris, the youngest prisoners were transferred
on parole to a clinic. There, Galois fell in love with Stephanie
Poterin-Dumotel, the daughter of one of the doctors. Two
fragmentary letters indicate that, though initially encourag-
ing, she finally declined to accept his love. To a friend, he
wrote: Happiness and hope are at an end, now surely con-
Abel and Galois 107

sumed for the rest of my life. The legend of Galois comes

from the succeeding events. He died only a few weeks af-
ter this letter in a duel whose motives and details remain
obscure. Some writers ascribe romantic causes, others politi-
cal; the full truth will probably never be clear. The surviving
documents contain tantalizing evidence, but do not permit a
decisive solution.
After being released from prison, his love rejected, Galois
wrote: Pity, never! Hatred, thats all. His final letters con-
tain his most explicit testimony, though it leaves much un-
explained: I die the victim of an infamous coquette and two
of her dupes. Consider also this account of the duel given
by a newspaper in Lyon: The young Evariste Galois . . . was
fighting with one of his old friends, a young man like himself,
like himself a member of the Societe des Amis du Peuple, and
who was known to have figured equally in a political trial. It
is said that love was the cause of the combat. The pistol was
the chosen weapon of the adversaries, but because of their
old friendship they could not bear to look at one another and
left the decision to blind fate. At point-blank range, each of
them was given a pistol and fired. Only one of the pistols was
loaded.
If this newspaper account is accurate, two comrades in the
same secret political society, both involved with the same
woman, used Russian roulette to decide who would sur-
vive. In his final letters, Galois begged his comrades not
to reproach me for dying otherwise than for my country,
indicating that he died for purely personal reasons. Yet the
mystery remains. Galois wrote that only under compulsion
and force have I yielded to a provocation which I have tried
to avert by every means. He reproached himself with hav-
ing told the hateful truth to those who could not listen to
it with dispassion. But to the end I told the truth. I go to the
108 Chapter 7

grave with a conscience free from patriots blood. I would

like to have given my life for the public good. Forgive those
who kill me, for they are of good faith. What was the provo-
cation to which he yielded, the hateful truth he spoke? We
do not know.
Here the hard evidence ends. His last words were to his
brother: Dont cry. I need all my courage to die at twenty.
At his funeral, three thousand people were ready to attack
the police. But at the last moment, the general uprising was
postponed a few days until what seemed a more auspicious
occasion, the funeral of a prominent general appointed by
Napoleon. If it was meant for political purposes, Galoiss
death was in vain.
But his life was the stuff of legend. In 1870, Raspail used
the examples of Abel and Galois in a powerful oration in the
French Chamber of Deputies that excoriated the established
order for its cruelty and indifference to young genius. Yet al-
ready in the 1840s mathematicians began to read and study
Galoiss posthumous writings, including a letter dated on
the eve of his duel. This testament became the centerpiece
of his legend, as if Galois had feverishly scrawled his great
theory on the edge of death. I have no time, he wrote in
the margin, the tragic epitome of doomed genius. Again, the
truth is less romantic. He was often too impatient to spell
out the tedious details of proofs; he had no time for that.
Indeed, many of Galoiss ideas were already present in his
earlier published works (he published five papers in his life-
time) and unpublished papers, though the last letter contains
indications of the larger sweep of his ideas.
Galoiss work agrees entirely with Abels results. What is
new in Galois is a turn toward abstraction in an essentially
modern way, leading to a complete understanding of solv-
Abel and Galois 109

ability, which Abel lacked. As he reformulated and extended

Abels work, Galois found it expedient to discuss permuta-
tions of the roots of equations not case by case but through
a master abstraction that would encompass many permu-
tations at once. Galois was the first to speak of these per-
mutations as instances of a new kind of mathematical object:
the group. In so doing, he opened the way to the characteris-
tic language and projects of modern mathematics, with all its
generality and power. At the same time, however, he opened
a crucial gap between the new language and common under-
standing. That which empowered modern mathematics left
many highly educated persons clueless. It remains a great
question to what extent this gap can be made intelligible,
much less bridged.
This page intentionally left blank
8 Seeing Symmetries

Once, a mathematician was giving a talk and had just stated a

complex theorem bristling with abstract concepts and sym-
bols. Before he could begin to prove it, suddenly someone
in the audience blurted out: Wait! Is that really true? The
speaker paused and drew a small equilateral triangle on the
board. He labeled its vertices A, B, and C, as a schoolchild
might. He stared at his triangle for a while, then erased it.
Yes, it is true. Let G be a group . . . , going back to his
theorem. Even someone comfortable with abstraction felt the
need to think about a simple example before moving to the
abstract statement.
Likewise, let us seek an intuitive view of the process of
solving equations before returning to the story of abstrac-
tions. Here, symmetry and permutation are central issues.
To show their importance, I will present them as movements
in a dance.
Consider first a quadratic equation. In general, it has a pair
of roots. To visualize their symmetry, think of two dancers
who always return to the same line from which they began
112 Chapter 8

and to the original distance between them. They can either

stay still or exchange places:

(12)

1 2

Changing places pivots them halfway around (180 ) and cor-

responds to exchanging the two roots. Changing places twice
means a full rotation (360 ) that returns them to their original
positions.
Let us call this dance S2 , because it is symmetric and has
two dancers. It consists of two elements: (1), meaning do
nothing, which we will also call the identity, and (12),
meaning exchange dancers 1 and 2. To describe the dance
more fully, we also list the ways the basic steps can be com-
bined. The possible combinations of these steps in sequence
can be recorded in, a Cayley table, introduced by Arthur
Cayley in 1854 (table 8.1). We read these steps left to right. For
instance, (12) (12) = (1) means that two exchanges return
us to the starting point, the identity. Since the do-nothing
step changes nothing, (12) (1) = (1) (12) = (12). Here,
the order in which you do the steps does not matter, so that
S2 is commutative or abelian, to use the modern term that

Table 8.1

S2 (1) (12)

(1) (1) (12)

(12) (12) (1)
Seeing Symmetries 113

honors Abels discovery of commutation as a characteristic

of his solvable equations.
At the same time, S2 is cyclic, meaning that if you keep
repeating the step (12) over and over, the dancers keep re-
peating the cyclical pattern (1), (12), (1), (12), (1), . . . . Call this
a 2-cycle, since they return to the original pose (1) after two
repetitions of (12).
Now S2 could represent the permutations of any two ob-
jects, but we are specifically interested in the two roots of a
quadratic equation. Since the roots of a quadratic equation
are paired by the in its solution, we must seek them as
a pair. By completing the square, we find the square root
whose two values can switch back and forth in a 2-cycle,
just like the two dancers. We have just seen that this cyclical
dance is also abelian. Here emerges our basic insight: Solving
an equation corresponds to a certain commutative symmetry. We
will gradually test and refine this insight, which will require
subtle qualification.
The analogy between S2 and the process of solving a
quadratic equation rests on a parallelism between the steps
of a dance and the steps of solving an equation. In each case,
there is something invariant, like the central axis around
which the two dancers turn. Here we anticipate an impor-
tant general insight: Where there is a symmetry, there must be an
invariant. We will now follow this insight in the case of more
complicated dances and equations.
As quadratic equations involve two dancers, cubic equa-
tions involve three. Think of them arranged at the corners of
an equilateral triangle and moving so that they always re-
turn to this same triangular formation, though perhaps with
different dancers at different vertices of the triangle. As we
shall see, there are just six steps they can take: three different
exchanges of two dancers, two different rotations involving
114 Chapter 8

all three dancers, and the identity. As before, we seek the

invariant axis around which the dancers move in each case.
Consider the different ways the dancers can exchange
places in pairs. We label the dancers 1, 2, and 3 (perhaps with
different costumes). Here is how they are arranged to start,
equidistant and viewed from above:

3 2

Dancers 2 and 3 can exchange, while dancer 1 stays still,

which we call (23):

1 1
(23)

3 2 2 3

Notice the symmetry axis, which goes from the still dancer
to the midpoint of the two exchanging dancers. (If a dancer
does not move, we leave its number out of the symbol.)
Seeing Symmetries 115

There are two other possibilities: (12) means that 3 is still,

while 1 and 2 change places:

1 2
(12)

3 2 3 1

and finally (13) means 2 is still, while 1 and 3 change places:

1 3
(13)

3 2 1 2

Note that each exchange is an odd permutation, since it in-

volves transposing only one pair of dancers. (The identity
may be considered an even permutation, since it involves
zero pairs.)
Since each exchange is an odd permutation, we can put
together two of them to make an even permutation. For
instance, consider first exchanging (12) and then exchang-
ing (13):

1 (12) 2 (13) 2

3 2 3 1 1 3
116 Chapter 8

The result of these two sequential exchanges is the same as

a rotation (123), meaning that the dancers have rotated by
one-third of a complete circle (120 ), so that 1 becomes 2
(that is, 2 stands where 1 used to stand), 2 becomes 3, and 3
becomes 1. Looking at them from above, we see:

1 2
(123)

3 2 1 3

For simplicity, we record the angle of rotations in terms of

counterclockwise motion. Note that the rotations involve a
single axis standing vertically at the center of the triangle,
like a maypole.
In order to obey the rule that the dancers always return
to the original triangular formation, they can only rotate by
120 (one-third of a complete circle) or multiples of it. We
have just described a single shift of 120 by (123). Likewise,
(132) will shift the dancers over by 240 , so that 1 becomes 3,
3 becomes 2, and 2 becomes 1:

1 3
(132)

3 2 2 1
Seeing Symmetries 117

We can build up all the possible rotations that respect the

original triangle out of repetitions of the basic rotation of
120 . Thus, two successive rotations of 120 are the same as
one rotation of 240 , which we can write as (123) (123) =
(132):

1 (123) 2 (123) 3

3 2 1 3 2 1

Three successive rotations by 120 add up to a full rotation

of 360 , returning us to the identity: (123) (123) (123) =
(132) (123) = (1):

1 (132) 3 (123) 1

3 2 2 1 3 2

Thus, the rotations of the three dancers are cyclic: the basic
rotation (123) repeated over and over gives a 3-cycle:
(123), (132), (1), (123), (132), (1), . . . . Notice also that these
cyclic rotations by themselves are abelian: whatever the order
in which you do them yields the same result. So the dance
S3 has six steps: the identity (1), the two rotations (123) and
(132), and the three exchanges (12), (13), (23).
118 Chapter 8

S3 is more complex than S2 because S3 is not commutative.

For instance, if you rotate (123) and then exchange (13), you
get (12):

1 (123) 2 (13) 2

3 2 1 3 3 1

But if you first exchange (13) and then rotate (123), you
get (23):

1 3 1
(13) (123)

3 2 1 2 2 3

Thus, S3 is nonabelian: When you mix rotations and ex-

changes, the result is not commutative. Note also that ro-
tations are always even permutations, whereas exchanges
are odd. So from the whole nonabelian dance S3 , contain-
ing both odd and even permutations, we can single out just
those that are even, which we will call A3 (for alternating),
shown shaded in table 8.2, and which are rotations with-
out exchanges. Notice that this set of rotations is closed
in the sense that these rotations stay within themselves: two
sequential rotations always give another rotation, never an
exchange. Of course, we must include the identity (1) among
Seeing Symmetries 119

Table 8.2

S3 (1) (123) (132) (12) (13) (23)

(1) (1) (123) (132) (12) (13) (23)

(123) (123) (132) (1) (23) (12) (13)
(132) (132) (1) (123) (13) (32) (12)
(12) (12) (13) (23) (1) (123) (132)
(13) (13) (23) (12) (132) (1) (123)
(23) (23) (12) (13) (123) (132) (1)

S3 is nonabelian, shown by the lack of symmetry about the main diagonal.

For instance, (123) (13) = (12), but (13) (123) = (23). The shaded entries
show the abelian subgroup A3 , which is symmetric about the diagonal.

the rotations, since it corresponds to a rotation by 0 (or any

multiple of 360 ).
We will call A3 an invariant subgroup of S3 , meaning
that A3 involves some (but not all) of the steps of S3 and
is closed and invariant. This term (defined technically in the
notes) means that if an invariant subgroup involves one kind
of step, it involves all other steps of the same kindhere
the rotations (123), (132), and (1). If we look within A3 , we
find no other such invariant subgroup except the identity
itself. For instance, (1) and (12) are a subgroup but not in-
variant because the other exchanges (13) and (23) are not
included.
Here is a symbolic way of writing down this interrelation-
ship of the dances:

S3  A3  I,

which simply means: S3 contains the invariant subgroup A3 ,

which in turn contains only the identity, I = (1). The chain
goes from the largest, most inclusive symmetry, to its
120 Chapter 8

next largest invariant subsymmetry, then the next largest

subsubsymmetry, and so on, ending with the smallest of all,
the identity. This nested structure mirrors the solution of a
cubic equation. Completing the cube corresponds to going
from S3 down to A3 . Solving a cubic equation corresponds to
finding an invariant subsymmetry (A3 ) within the next larger
symmetry of the equation, S3 .
Now we take a breath and get ready for four dancers,
whose possibilities go far beyond a simple square pattern.
If you include all possible steps, the square becomes so tan-
gled that it no longer represents the geometry of the dance.
For instance, consider a step like (234): 1 stays still, while 2, 3,
and 4 rotate among themselves, not at all in a square. To rep-
resent such a step, we need a solid figure. A tetrahedron does
nicely: (234) corresponds to dancer 1 standing still, defining
an axis around which 2, 3, and 4 rotate by 120 :

3
4

Thus, the full symmetry of a tetrahedron, S4 , including

both rotations and exchanges, corresponds to the dance of
four. In addition, S4 is the rotational symmetry of a cube or
Seeing Symmetries 121

Figure 8.1
Johannes Keplers figure showing an octahedron constructed within a cube
by joining the midpoints of its faces, showing that they share the same
symmetry (Harmonices mundi, 1619).

an octahedron (figure 8.1). These solids have three axes of

symmetry:

The four diagonals of a cube also correspond to the four

dancers or to the four roots of a quartic.
S4 is a complex dance, with 4! = 24 possible steps. These are
so numerous that it is difficult to show the table of S4 , which
has 24 24 = 576 entries. Not surprisingly, S4 is nonabelian,
for its exchanges and rotations in general are not commuta-
tive. The subgroup of even permutations, A4 , is nonabelian.
Its table has 12 12 = 144 entries, too many to list here.
122 Chapter 8

Table 8.3

V (1) (12) (34) (13) (24) (14) (23)

(1) (1) (12) (34) (13) (24) (14) (23)

(12) (34) (12) (34) (1) (14) (23) (13) (24)
(13) (24) (13) (24) (14) (23) (1) (12) (34)
(14) (23) (14) (23) (13) (24) (12) (34) (1)

V is an abelian subgroup of A4 .

Though A4 is nonabelian, it contains a subsymmetry V

that is abelian, shown in table 8.3. V is invariant because it
includes all the double exchanges, (12)(34), (13)(24), (14)(23),
and (1), in which all four dancers participate, but in sepa-
rate pairs, like the do-si-dos of square dancing. V is abelian
(though not cyclic) because the double exchanges do not in-
terfere with each other and hence can occur in any order.
Having found within the full dance an invariant, abelian sub-
group, we now know that the quartic equation, too, can be
factored and solved as Ferrari did.
In this progression, something very beautiful has hap-
pened. One by one, the Platonic solids have appeared. They
have been conjured up by mathematical necessities that are
purely matters of permutations, without any mention of
geometry. Having encountered the tetrahedron, cube, and
octahedron, we might ask, where are the other two, the do-
decahedron and icosahedron?
Just as with the preceding cases, we expect that the quintic
will be illuminated by the dance of five, S5 . From our expe-
rience with S4 , we fully expect that the dancers moves will
not be adequately represented by a pentagon or any other
plane figure, since these cannot represent possibilities such
as (12)(345), two dancers exchanging places while the other
Seeing Symmetries 123

three rotate among themselves. We must look to solid figures,

as before.
But no regular polyhedron has the symmetry S5 . It would be
tempting to draw a connection between the fact that no such
polyhedron exists and the unsolvability of the quintic, but
such a deduction would falsely ascribe a cause and effect
relation when all we have is an analogy. There is no intrinsic
relation between the properties of three-dimensional space
and the solution of equations of the fifth degree. Yet it is
amazing that neither the polyhedron of symmetry S5 nor the
solution of a quintic equation with the same symmetry exists.
If this is a coincidence, it is a beautiful one. There are so few
fundamental symmetries (S2 , S3 , S4 , S5 ) that they necessarily
appear in any simple algebraic or geometric situation.
The dance S5 has 120 120 = 14,400 possible combina-
tions of two steps and is nonabelian. Just the table for its even
steps, A5 , numbers 60 60 = 3,600 entries. A5 is also non-
abelian: all the exchanges and rotations interfere with each
other so that the order of steps matters. At this point, we
regain contact with a geometric model, for A5 describes the
rotational symmetry of both the icosahedron and the dodec-
ahedron (figure 8.2). Here the final two Platonic solids find
their place in the scheme. Their five-fold symmetries corre-
spond to the patterns of five dancers (figure 8.3). As before,
let us examine these solids to look for further subsymme-
tries within them, seeking one that is invariant within the
next higher symmetry. If we could find even one, that would
correspond to the possibility of solving the quintic equation
through taking roots.
In seeking these invariant subgroups, we must consider not
only rotations about just one of the axes of the icosahedron,
124 Chapter 8

Figure 8.2
Keplers diagram showing an icosahedron constructed within a dodecahe-
dron by joining the midpoints of its faces, showing that both of these figures
share the same symmetry.

for each type of symmetry, but rotations about all of them.

There are different sorts of rotations possible about different
axes, as box 8.1 shows. If an invariant subgroup includes a
certain rotation, it also includes rotations of the same magni-
tude around any other axis that admits such rotations.
But there is no such invariant subgroup, other than the
identity. Box 8.1 shows that the symmetries of the icosahe-
dron are so interconnected that each leads to all the others. Let
each pair of dancers make all the possible exchanges. In the
process, they will eventually go through all 3,600 combina-
tions of steps. No invariant subgroup of steps remains closed
within itself, when it is extended throughout all the possible
axes that are equivalent. The beautiful, intricate pattern of A5
is so completely interwoven that it cannot be broken down
into separable parts. Thus, there is no invariant subgroup
Seeing Symmetries 125

Figure 8.3
Within a dodecahedron, group its twenty vertices into five sets of four
equidistant vertices. Connecting each of these five sets yields five intersect-
ing tetrahedra. They show the five-fold symmetry shared by the dodeca-
hedron, icosahedron, and the dance of five, A5 .

smaller than the whole noncommutative dance itself. There-

fore the quintic equation cannot, in general, be solved.
What we have called a set of possible dance steps, mod-
ern math calls a group. In the abstract sense, a group is a
set of elements having a particular structure, defined by four
central requirements: The group is closed under a certain op-
eration (call it ), so that if a and b are elements of the group,
a b is also an element. There is an identity element I (so
that a I = I a = a ). Each element a has an inverse a 1
126 Chapter 8

(so that a 1 a = a a 1 = I ). The associative law holds:

a (b c) = (a b) c. The earliest examples of groups were
permutation symmetries like S3 , S4 , S5 , which is what Galois
had in mind when he first used the word group in this
sense. Here, the elements are permutations, their group op-
eration is successive permutation, and the identity I is the
nonpermutation, (1).
Gradually, the concept of group became more and more
general, including not only permutations but any set of ele-
ments that would satisfy the basic definition. It was only
in 1854 that Cayley set out the basic definition of the the-
ory of groups in this abstract form. He also showed that
any finite group is identical in structure to some permuta-
tion group. Now free of any visual or intuitive content, the
theory found new power in its abstractness. The concept of
group thus represents a step in generalization that recalls
the step from some particular pair of objects to the num-
ber 2. Another such step occurred when Viete went from
a particular number like 2 to a coefficient a in a general
equation, which might have the value 2 in some case but

Box 8.1
The symmetry A5 of an icosahedron has no proper invariant
subgroup

Consider the different classes of symmetries of an icosahe-

dron, which correspond to the 60 elements of A5 . Note first
that an icosahedron has 20 faces (each of them an equilateral
triangle), 12 vertices (points at which 5 lines meet), and 30
edges (lines connecting two adjacent vertices). There are 6
axes connecting the opposite vertices, each of which allows
4 different pentagonal rotations (each a multiple of 15 of a
Seeing Symmetries 127

Box 8.1 Continued

full rotation, since 5 faces surround each vertex in a penta-

gonal array). These rotations comprise two distinct classes
within A5 , each having 6 2 = 12 elements, corresponding
to odd or even multiples of rotations by 72 . Then there are
10 axes connecting the midpoints of opposite faces, around
each of which we can make 2 different triangular rotations
(each 13 of a full rotation, since the faces are triangles), hence
10 2 = 20 elements more. There are 15 axes connecting
the midpoints of the 30 edges; around each of these we can
do an exchange, making 15 elements more. Finally, there is
the identity. So the 60 symmetries have the classes given by
60 = 12 + 12 + 20 + 15 + 1.
For instance, consider the pentagonal rotations. It turns
out that each of them can be made up of a sequence of
triangular rotations,
128 Chapter 8

Box 8.1 Continued

so that these two classes of steps are interconnected and are

not separate subgroups. Furthermore, each triangular rota-
tion can be made up of two exchanges:

(13)

(12)

3 2
(12) (13) = (123)
*

Thus, the possible subsymmetries of A5 are made up of all

the possible exchanges, and there is no invariant subgroup
that does not include the whole.
These arguments are confirmed by an important general
rule, called Lagranges Theorem: If a symmetry has n ele-
ments, any subgroup must have a number of elements that
is a divisor of n. (A proof of this theorem is given in ap-
pendix C.) S3 has 6 elements, while its invariant subgroup
A3 has 3, a divisor of 6. Likewise, S4 has 24 elements, while
Seeing Symmetries 129

Box 8.1 Continued

A4 has 12 and V has 4 elements, both divisors of 24. (Re-

member that the identity is always counted as an element
of the subgroup.) But S5 has 120 elements and A5 has 60, a
divisor of 120. Now any subgroup of A5 must be a divisor of
60. We have found that the possible invariant subgroups are
given by the list 60 = 12 + 12 + 20 + 15 + 1. Now any sub-
group would both have to be a divisor of 60 and also be com-
posed of either 12 + 1 = 13, 20 + 1 = 21, or 15 + 1 = 16
elements. But neither 13, 21, nor 16 (nor any sum of them) is a
divisor of 60. So the only invariant subgroups of A5 are im-
proper: either all of A5 , all 60 elements, or just the identity,
because 60 and 1 are the only divisors of 60 that are possible.
Thus, A5 has no proper invariant subgroup.

in general has many possible values. The concept of group

allows us to contemplate pure relationships without the dis-
traction of considering the things related. Moreover, there
are many possible groups, each with its own peculiarities,
each capable of being realized by many possible sets of ele-
ments. Some groups are trivial, as mathematicians put it,
like the one whose sole member is the identity, I . As we have
seen with S5 , each group has its own character and peculiar
beauty, sometimes complex, even bizarre.
The concepts of group theory parallel every step of our
dance analogy. The symmetries we discussedS2 , S3 , S4 , S5
are each groups with different numbers of elements. Each of
these groups contains subgroups, as S4 contains the subgroup
A4 , expressing the interrelation of the subsymmetries within
the next larger symmetry. Groups can be abelian (commu-
tative) or nonabelian. What we called invariant subgroups
are usually called normal (see notes to this chapter). The
130 Chapter 8

nesting of symmetry and subsymmetry is formalized as a

solvable chain, written (in the case of the cubic equation) as

S3  A3  I,

where S3  A3 means: S3 contains A3 as a normal subgroup.

This was Galoiss great advance over Abel, specifying ex-
actly which equations are solvable and which not. By 1889,
Camille Jordan and Otto Holder expressed this in a form
that highlights commutativity: An equation is solvable if the
quotient group of each successive pair of links in the chain
is abelian. That is, the chain shows how the group is made
up of building blocks, namely invariant subgroups nested
and linked, in which each successive link is as large as pos-
sible. To be solvable, each successive link must fit into the
next in an abelian way. Here, the criterion of commutativity
is subtle, for it is not necessary that the links be abelian, only
their linkage. The concept of quotient group (defined
and illustrated on pp. 176177, 193194) gives precise form
to this.
A group whose subgroups do not form a solvable chain is
called simple, though (like the icosahedron and its symme-
try group, A5 ) intuitively it may be formidably complex. Even
so, A5 turns out to be the smallest nonabelian simple group,
in the sense of having the smallest number of elements. A
great tour de force of twentieth-century mathematics was
the exhaustive enumeration of finite simple groups, starting
with A5 and including the so-called monster, which corre-
sponds to a certain symmetry of a solid in a space of 196,883
dimensions. Alas for intuition, with nothing to visualize or
grasp. But the abstract theory stands ready to help when in-
tuition is overwhelmed and clueless. Here the question of
commutativity can be the guiding thread.
9 The Order of Things

Abel first noticed the commutativity in his special equa-

tions, so it is only fitting that this property is named for him.
Though Galois immensely extended Abels work, he did not
emphasize the theme of noncommutativity, nor did those
who later systematized his ideas, such as Jordan and Holder.
Doubtless their attention was on precise mathematical for-
mulation rather than philosophical implications. Neverthe-
less, their insight entered deep into the mainstream of
mathematics, largely through the influence of the theory of
equations they shaped. So intrinsic is the thread of noncom-
mutativity that modern treatments take it for granted. That
being the case, it is important to reconsider its import and
seek ways to express it.
For the first generations that followed Abel and Galois, this
new theory was still rich and strange. In retrospect, Galoiss
radical viewpoint won out over the older approach because
it gave a complete understanding of solvability, where Abels
results were only partial. Yet it is worth pausing to consider
the confrontation between them and how it played out, for
it reflects deep currents in the development of mathematics.
At the beginning of the nineteenth century, algebra was
still regarded as an offspring of arithmetic, closely tied to its
132 Chapter 9

fundamental operations. Gauss considered a b = b a the

first principal truth about multiplication. In 1830, George
Peacock was the first to try to give algebra the character of
a demonstrative science. He distinguished between arith-
metical algebra, whose elements are numbers and whose
operations are those of arithmetic, and symbolical algebra,
a science, which regards the combinations of signs and sym-
bols only according to determinate laws, which are altogether
independent of the specific values of the symbols them-
selves. This distinction is not far from Vietes contrast
between the logistic of numbers and of species. It em-
phasizes the abstract quality of the symbols, opening the
door to new possibilities, though Peacock thought of the
symbols fundamentally as generalizations of ordinary
numbers.
Peacock also set forth what he called a principle of the
permanence of equivalent forms, which essentially means
that, though the symbols may be general, the laws of alge-
bra should always be the same as those of arithmetic. For
instance, Cauchy wrote that if a and b be whole numbers, it
may be proved that a b is identical with ba ; therefore, a b is iden-
tical with ba , whatever a and b may denote, and whatever
may be the interpretation of the operation which connects
them. There is no hint that these symbols could ever escape
from the laws of ordinary numbers.
Peacocks contemporaries gradually threw off these re-
strictions. In 1830, Augustus De Morgan wrote that no word
or sign of arithmetic or algebra has one atom of meaning,
so that their interpretations should be open and arbitrary,
not restricted to ordinary numbers and magnitudes. By 1840,
Duncan Gregory had added that these symbols could repre-
sent operations, not just numbers. Building on this, in 1847
George Boole used algebraic symbols to represent members
of a set, where the whole universe of discourse is called 1
The Order of Things 133

and the empty set 0. In modern Boolean algebra, addition

means uniting two sets and multiplication means their inter-
section. Let 1 mean the totality of humans, x all Americans,
and y all women. Then (1 x) means all non-Americans,
(x+y) all Americans and all women, xy all American women.
Not only did Boole show the close relation of logic to math-
ematics, he also emphasized that the form, rather than the
content, of the symbols is crucial. For this reason, Bertrand
Russell credits Boole with having discovered pure mathe-
matics, the greatest discovery of the nineteenth century.
The way now stood open for algebras that would break de-
cisively with the rules of ordinary arithmetic. The first was
found by an Irishman, William Rowan Hamilton. He was
an early admirer of Abels proof and wrote an extensive ac-
count of it that helped other mathematicians at a time when
they found it puzzling in the extreme. As mentioned earlier,
Jerrard thought he had found a solution to the quintic in
1835, which stimulated Hamilton to study Abels proof care-
fully. Hamilton tried tactfully to persuade Jerrard that his
proof was mistaken, but Jerrard died unconvinced, still de-
fending his proof in 1858, which shows how long it took for
Abels ideas to be widely accepted. In Hungary during the
same period, the brilliant Janos Bolyai, co-discoverer of the
first non-Euclidean geometry, also tried to solve the quintic,
evidently unaware of Abel and Ruffini.
By emphasizing the power and generality of Abels in-
sights, Hamilton showed his contemporaries how important
and beautiful Abels theory really was. In so doing, he helped
it to live and to flourish in other minds, so that it would not
be neglected because of its novelty and difficulty. In France,
Joseph Liouville and Camille Jordan performed an even more
important service for Galois, whose fragmentary and abbre-
viated writings needed extensive amplification before their
import could emerge.
134 Chapter 9

Such generous admiration brought a rich reward. Begin-

ning in 1833, Hamilton sought a new kind of number that
would extend the idea of complex numbers. He realized that
a complex number like a + bi (where i = 1) could be
thought of as an ordered pair composed of its real and its
imaginary parts, (a , b). As noted earlier, such an ordered pair
could be represented as a point in a plane (see figure 3.2).
Hamilton then sought to generalize this idea, from the plane
to three dimensions, from an ordered pair a + bi to an or-
dered triple like a + bi + c j, where j would generalize i so
that i 2 = j 2 = 1.
For ten years, Hamilton tried to make these ordered triples
work. He could understand how they should add (term by
term, grouped according to common factors of 1 or i or j),
but not how they should multiply. In 1843, the answer came
to him suddenly, as he and his wife were walking along the
Royal Canal in Dublin. His difficulties could not be solved
with triples, but could be if he took quadruples, a +bi +c j +dk,
where i 2 = j 2 = k 2 = 1. To make multiplication work for
these quaternions (as he called them), he had to make this
operation noncommutative: i j = k but ji = k, and likewise
jk = i = k j and ki = j = ik. Since i j = ji, reversing the
order of multiplication reverses the sign of the product, so
that one might say that these quantities anticommute. To mark
this special moment, Hamilton took out his pocket knife and
carved the equation i 2 = j 2 = k 2 = i jk (since i jk = kk = 1)
into a stone of the Brougham Bridge.
The carving has long since worn away, but the strange
beauty of noncommutative numbers lives on. Hamilton spent
the rest of his life developing the new algebra of quaternions,
which he described as a curious offspring of a quaternion of
parents, say of geometry, algebra, metaphysics, and poetry,
reflecting his devotion to philosophy and poetry (he wrote
The Order of Things 135

poetry enthusiastically, if not well, and was a friend of

Samuel Taylor Coleridge). He gradually developed quater-
nions, leading to the modern notion of vectors (magni-
tudes with direction) and scalars (pure magnitudes with no
direction).
Hamilton noted that there were two kinds of multiplica-
tion that emerged. One he called the scalar product, which
takes two vectors, is commutative, and produces a scalar as
a result. The name has stuck. The other (not named) was not
commutative and produced a vector as a result. Hamilton
believed that his quaternions promised a new view of the
cosmos, and a number of British mathematicians also took
them as the key to mathematics and mathematical physics.
The great Scottish physicist James Clerk Maxwell, though
interested in quaternions, finally did not join the true be-
lievers. The difficulty was that, just as complex numbers
imply a two-dimensional plane, the four-component quater-
nion seemed to require a four-dimensional space. At that
time, there was no hint of what that could possibly be.
As if responding to the same unspoken question, Hermann
Grassmann, a secondary-school teacher in Germany who
became a specialist in Sanskrit, independently developed
an algebra based on a space of n dimensions, not merely
three. In 1844, Grassmann published his Theory of Exten-
sion (Ausdehnungslehre), which distinguished inner prod-
ucts, commutative like Hamiltons scalar product, from
outer products that were not. As he studied the possibil-
ities, Grassmann proved that for products of two factors
there are . . . only two types of linear product, namely the
one whose system of determining equations has the form
e i e j + e j e i = 0 and the one for which it has the form e i e j = e j e i ,
in which the indices i, j can go from 1 to n, the number of
dimensions in the space. That is, either the factors must
136 Chapter 9

commute (e i e j = +e j e i ) or they must anticommute (e i e j =

e j e i ). Since they are not restricted to any specific number of
dimensions, Grassmanns algebras emphasized the general-
ity of commuting versus anticommuting.
Grassmanns work was only slowly assimilated because it
was expressed in unconventional notation. Gradually, others
began to express similar ideas in their own ways. In America,
Josiah Willard Gibbs, one of the greatest mathematical physi-
cists of the century, reformulated Grassmanns ideas in the
language of vectors that is now used universally. Following
Gibbs, Grassmanns outer product became what we call
the cross or vector product, in which A B = B A, 
as with Hamiltons quaternion multiplication.
This noncommutative algebra is the basis of the modern
understanding of physics, whether Newtons mechanics or
Maxwells theory of electricity and magnetism. Found in
every textbook, it is routinely taught without much atten-
tion to its striking noncommutativity. It is also central to
matrix algebra, which was developed beginning about 1858
by two inseparable English friends, the invariant twins
Arthur Cayley and James Joseph Sylvester. The idea of ma-
trices emerges from looking at the coefficients of simultane-
ous equations and arranging them in an array. Once again,
the problem of solving equations stimulated research in a
new direction, for matrix algebra allows great simplification
of the process of solution (box 9.1). Soon it was realized that
matrices, too, have noncommutative multiplication (box 9.2)
and can be used to represent the algebras of Hamilton and
Grassmann. Matrices show that such algebras do not neces-
sarily require complex or exotic kinds of number, for matri-
ces with real values (but in an ordered array) can realize every
algebraic requirement of complex numbers or of quaternions.
Indeed, matrices are perfect vehicles for expressing the
The Order of Things 137

Box 9.1
Two successive linear transformations, compared with their
matrix form

Consider a transformation (such as a rotation of coordinates)

that takes the variables x and y and defines new variables x 
and y :
a 11 x + a 12 y = x 
a 21 x + a 22 y = y .
Consider a second such transformation that expresses x  , y
in terms of x  , y :
b 11 x  + b 12 y = x 
b 21 x  + b 22 y = y .
Expressing x  , y in terms of x, y involves solving and re-
substituting these two transformations, but can be expressed
much more simply if we write them in matrix form:
     
b 11 b 12 a 11 a 12 x x 
=
b 21 b 22 a 21 a 22 y y
The two matrices can be multiplied using the rule that each
row of the first matrix be multiplied term-by-term by the first
column of the second matrix, and the sums taken of those
products:
    
b 11 b 12 a 11 a 12 a 11 b 11 + a 21 b 12 a 12 b 11 + a 22 b 12
=
b 21 b 22 a 21 a 22 a 11 b 21 + a 21 b 22 a 12 b 21 + a 22 b 22
This notation greatly simplifies dealing with systems of
equations.

operations of groups. Grassmanns generalization of the pos-

sibilities of algebra opened an ever-widening field. Cayley
discovered an eight-dimensional generalization of quater-
nions, now called octonions or Cayley numbers. Sylvester
called himself the new Adam, giving names (not all of
138 Chapter 9

Box 9.2
Matrix multiplication is not commutative in general

Consider, for instance, the following noncommuting

matrices:
      
0 1 1 0 0 + 0 0 + (1) 0 1
= = ,
1 0 0 1 1+0 0+0 1 0
which is not equal to the product taken in the opposite order:
      
1 0 0 1 0+0 1+0 0 1
= = .
0 1 1 0 01 0+0 1 0

which stuck) to a whole new universe of algebraic beings:

invariants, discriminants, zetaic multipliers, allotrious fac-
tors, among many others. By 1860, Benjamin Peirce had enu-
merated 162 different algebras. (His remarkable son C. S.
Peirce carried forward Booles logical work as a philosopher,
besides being a chemist and astronomer.)
The historical unfolding of the notion of noncommutativ-
ity implies larger questions: What is its significance? Why
should we care about it? In 1872, not long before Cayley set
out group theory in full abstraction, the great mathematician
Felix Klein gave a lecture at Erlangen, in which he proposed
that group theory be considered the heart of mathematics,
more fundamental in its import than geometry or algebra
alone. He argued that each geometrical figure should be
considered more in terms of its fundamental symmetry
specified by the group governing itthan by any figure or
equation in itself. In Kleins view, considering ever more
general groups would be the royal road to understanding
ever more complex spaces, reaching past three dimensions
to manifolds of higher dimensions.
The Order of Things 139

The Erlangen Program was the clarion call of a new

abstract vision, going past the traditional starting points of
mathematics to reveal the deepest, most general concepts
underneath. It was an influential rallying cry that still re-
sounds in contemporary mathematics. As part of his pro-
gram, Klein went back to reconsider the Platonic solids in
terms of group theory. Where the Greeks considered these
solids as emerging from the necessity of fitting together reg-
ular polyhedra in three-dimensional space, Klein saw them
as incarnations of fundamental groups. Thus, an equilateral
triangle should be considered most of all in terms of its sym-
metry group S3 , the rotations and reflections that leave it
invariant. Similarly, the Platonic solids are incarnations of
more complex rotational symmetries: the tetrahedron of A4 ,
the cube and octahedron of S4 , the dodecahedron and icosa-
hedron of A5 . Indeed, Kleins presentation lies behind the
dance analogy used in the preceding chapter.
In fact, noncommutativity is everywhere. In general, rota-
tions in three-dimensional space do not commute. Imagine
leaving Santa Fe for two journeys: (a) going directly west for
1,000 km; (b) going directly north for 1,000 km. If you do
these journeys in the order first (a), then (b), you will arrive
at a certain spot. But if you do them in the reverse order, (b)
then (a), you end up somewhere else, about 123 km away.
The order matters since you are traveling on a globe, not a
flat plane.
Klein considered this peculiar noncommutativity of spa-
tial rotations to be the hallmark of space itself, understood as
a manifestation of its underlying symmetry of rotations. He
classified geometries by the algebraic properties that remain
invariant under a particular group of transformations, as Eu-
clidean areas and lengths do when moved in the plane. This
was one of many instances that moved Klein to deem group
140 Chapter 9

theory the master key that would unlock the secrets of both
algebra and geometry by uncovering what underlies both of
them.
Long before, Descartes had noticed that the roots of equa-
tions could be increased or decreased, scaled up or down by
a multiplicative factor, while leaving invariant their essen-
tial constellation: the number of the roots and their relative
spacing. Looking back from the perspective of the twenti-
eth century, the eminent mathematician Hermann Weyl took
this as a kind of relativity that emerged long before Albert
Einsteins theories. In 1927, Weyl noted that it was a lucky
chance for the development of mathematics that the relativity
problem was first tackled, not for the continuous point space,
but for a system consisting of a finite number of distinct ob-
jects, namely the system of the roots of an algebraic equa-
tion with rational coefficients (Galois theory). Weyl even
wrote that, because of its novelty and profundity, Galoiss fi-
nal letter was perhaps the most substantial piece of writing
in the whole history of mankind. The original development
of group theory in this discrete, algebraic context prepared
the way for Einstein and Hermann Minkowskis understand-
ing of relativity in terms of groups of continuous motions that
leave the speed of light invariant. Here again, the symmetry
between all uniformly moving frames of reference points to
an invariant.
As the comparison of the regular solids and the solvable
equations shows, algebra felt the impact of noncommutativ-
ity before geometry, despite such simple cases as the jour-
neys around the Earth just described. Perhaps this is simply
because the order in which things are done far more natu-
rally applies to algebra, in which symbols and operations
are read in a certain order, than to geometry, which seems
to rely on timeless diagrams. Plato considered the eternal
The Order of Things 141

changelessness of mathematics a hallmark of its closeness to

absolute reality, to pure being beyond existence and becom-
ing. It was, then, a shock to realize that the order of things
mattered.
The question of order is crucial for modern physics. The
heart of thermodynamics is its Second Law, a statement that
natural processes are in general irreversible. Heat flows from
hotter to colder bodies, not vice versa, unless we expend
work and also expel heat into the rest of the universe. Struck
by its sweep and scope, Max Planck emphasized irreversibil-
ity as he hesitatingly introduced the quantum into physics.
On the human scale, irreversibility means that every action,
once done, can never be undone. It gives dignity to human
choice by emphasizing its irrevocable consequences. Time is
no longer the perpetually reversible stream of Newtonian or
Maxwellian dynamics. Instead, irreversible time is the dark
river of history, flowing ever onward.
Ironically, Planck himself tried to reverse the flow of events
and undo the shock of the quantum, or at least tame it into
a more classical form. Ruefully, he recognized that here too
irreversibility held sway. As quantum theory took shape, it
revealed noncommutativity at its very core, the child of ir-
reversible thermodynamics married to reversible mechanics.
Formally, Paul Dirac showed that its crucial realization was
that physical quantities, when expressed in algebra, did not
necessarily commute. In a famous equation, he wrote that
the position of a particle q and its momentum p as oper-
ators obey the noncommuting relation pq q p = i h/2 ,
h is Plancks constant (h = 6.610 erg-seconds) and
27
where
i = 1. Since h is extremely small by human standards,
this noncommutativity usually manifests itself only on the
atomic scale. Diracs relation is the direct source of the fa-
mous Heisenberg uncertainty principle. As I have discussed
142 Chapter 9

in my book Seeing Double, these are the distinctive hallmarks

of the utter indistinguishability of quanta. Noncommutativ-
ity continues to govern the deepest level of physical theory,
whether the quantum theory of fields at space-time points or
alternative string theories that use not points but tiny lengths
or membranes. For all of them, noncommutativity is basic
and promises to be so in the future.
Initially, noncommutativity in quantum theory stood for
the irreversibility of measurement in general. The theory of
relativity also relies on noncommutativity to express another
fundamental principle: Effects do not follow instantaneously
from causes but are limited in their speed by the velocity
of light. Here group theory expresses beautifully the sym-
metries of space-time. If events are sufficiently separated in
space and time that no light signal can connect them, then all
physical quantities of the events must commute, expressing
their independence. Conversely, within the realm of causal
influence traveling at or less than the speed of light, physical
quantities cannot commute, expressing their causal relation.
Any change in this principle would require overturning this
fundamental principle of causality, which has so far been
confirmed in myriad ways. Commutativity versus noncom-
mutativity lies at the heart of causality itself.
Going from these most general principles to more specific
cases, noncommutativity is crucial to the development of
the standard theory that has very successfully drawn the
known elementary particles into a largely unified scheme.
This theory rests on what are called nonabelian gauge fields.
Without even exploring what such fields might be in de-
tail, note that their very name announces that they rely on a
special sort of noncommutativity in the symmetries of funda-
mental particles. Such nonabelian theories promise to be
The Order of Things 143

the inescapable vehicles of future physical theory, if only

as a point of departure. They also are the starting points
of attempts to unite quantum theory with Einsteins gen-
eral theory of relativity, which is a geometrical theory of
gravitation as the curvature of space-time. These directions
in physics have in turn stimulated further developments in
mathematics, notably the theory of noncommutative geom-
etry. Kleins conjecture continues to be deeply provocative:
nonabelian groups may give deep clues to the nature of space
and time. As noted in the last chapter, the rotational invari-
ance of three-dimensional space leads to the five Platonic
solids. Perhaps such considerations of groups may someday
explain why we perceive three dimensions of space and only
one of time. The search has only just begun.
This page intentionally left blank
10 Solving the
Unsolvable

Abels work opened the way to Galoiss explanation of the

failure to solve all equations in radicals. Indeed, radicals had
posed dilemmas since the beginning. The discovery of irra-
tional quantities cast into doubt the nature of mathematics
and, with it, the possibility of definitive human knowledge.
In Greek, krisis means a trial, a decisive act of distinction,
separation, judgment. The impossibility of making the irra-
tional commensurate with the rational tested and formed
Greek mathematics and philosophy.
Much later, Abels impossibility proof was a watershed
for modern mathematics. It had the dimensions of a crisis,
though its significance has not been fully acknowledged or
assessed. To do so requires a special effort of imagination
that would allow us to return to a moment that has been
forgotten, the moment in which Abel confronted what he had
proved.
As discussed earlier, it is not clear whether the ancients
regarded the discovery of the irrational as a disaster or a
miracle. It was ominous; one might drown in such an un-
nameable sea, beyond the security of the counting numbers.
As Euclid showed at the end of his Book X, the irrationals
146 Chapter 10

lead to an infinite vista, which could inspire horror or won-

der. Euclid tried to keep them rigorously separate from the
integers, if only to avoid paradox and confusion. Yet what the
ancients separated, the moderns united. Gradually but inex-
orably, the development of algebra led to the term number
being applied to irrationals as well as integers, for x = 2
seems no less valid a solution of the equation x 2 = 2 than
y = 2 is a solution of y2 = 4.
Once symbolic algebra was established, the crisis of un-
solvability was inevitable. Abels proof showed that radicals
are not sufficient, just as the ancient proof showed that ra-
tionals are not enough. Since solutions of all equations are
known to exist by the Fundamental Theorem of Algebra,
they must be a new kind of magnitude. The name alge-
braic numbers came to signify all the solutions of algebraic
equations whose coefficients are integers. As with the irra-
tionals, giving them a name seemed tantamount to dispelling
the sense of crisis or paradox. For mathematicians, new solu-
tions of the quintic eclipsed the question of solution in radi-
cals, for they could simply annex the algebraic numbers to
those expressible in radicals. To give a name to those alge-
braic numbers that are not expressible in radicals, we might
call them ultraradical numbers.
Now we can complete the odyssey of the quintic. It turns
out that their solutions can be identified with the so-called
elliptic modular functions (defined as sums of powers times
trigonometric functions). These functions are based on the
theta functions that Abel and Jacobi began to study in
1827, though it was not until 1858 that Charles Hermite,
Leopold Kronecker, and Fernando Brioschi independently
showed their relation to the quintic. By 1870, Jordan had
proved that any polynomial equation, of whatever degree,
could be solved by using these generalized functions.
Solving the Unsolvable 147

During the same period, people started building machines

that could find approximate solutions. About 1642, Blaise
Pascal had the first vision of machines that could compute.
In 1840, Leon Lalanne devised a mechanical computer that
could approximate equations up to the seventh degree. In
1895, Leonardo Torres Quevedo developed an elegant
machine that could calculate logarithms needed to solve diffi-
cult algebraic problems (figure 10.1). At Bell Laboratories in
1937, scientists developed a larger but less elegant-looking
machine, the Isograph, which could approximate equations
up to the fifteenth degree (figure 10.2). By the end of the twen-
tieth century, far more powerful programs were available to
users of personal computers.
These developments simply sidestepped the problem of
solution in radicals. But what, then, happened to the crisis?
Was it ever resolved? Or is it a case of history being written by
the victors, the moderns who successfully bypassed and set

Figure 10.1
A machine built in 1895 by Leonardo Torres Quevedo to calculate logarithms.
148 Chapter 10

Figure 10.2
The Isograph, built in 1938 at Bell Laboratories to solve polynomial equa-
tions up to the fifteenth degree.

aside the older scruples? Perhaps they were right not to look
back; they were so busy forging ahead that they had little time
for retrospection. Yet surely there remains valuable insight to
be gained by reflecting on what each successive crisis really
meant.
Each crisis opened a new insight into the infinite, in dif-
ferent contexts. First, between arithmetic and geometry: the
Greeks learned that expressing an irrational requires an infi-
nite number of digits. Next, between geometry and algebra:
to express the area of an oval requires an infinite number
of algebraic terms, as Newton showed. Finally, Abels proof,
within algebra itself: solving a finite equation requires an
infinite number of terms.
Even solving a cubic equation with real roots requires that
we take the cube root of an arbitrary complex number. As
Solving the Unsolvable 149

complex numbers were accepted, taking their roots became

imperative. The problem was resolved by Abraham De
Moivre in 1707 and given its modern form by Euler in 1748.
We must start by expressing each complex number in terms
of a magnitude and an angle, understood as polar coordi-
nates in a plane. De Moivres formula then gives a beautiful
expression for the nth power of the complex number (where
n could be any integer or fraction) in terms of sine and cosine
functions (box 10.1).
Most calculators provide sines or cosines by pressing a
button. Geometrically, these functions are ratios of the sides
and hypotenuse of a right triangle. Yet algebraically sines and
cosines are not finite polynomials, but infinite series. Newton
had already discerned this algebraic infinitude in studying

Box 10.1
De Moivres formula

Express the complex number z in terms of its absolute value

r and its argument , z = r (cos + i sin ). De Moivre proved
that
zn = r n (cos n + i sin n ).
To show this, consider the two Taylor series sin x = x
x3 5 7 2 4 6
3!
+ x5! x7! + and cos x = 1 x2! + x4! x6! + . If we
write z = r e , where i = 1 and e = 2.718 . . . is the basis of
i 2

natural logarithms, then zn = r n e in . Using the Taylor series

2 2 2
for e in = 1 + in + i n2! + , De Moivres formula follows
if we group the real and imaginary terms separately:
   
n2 2 n3 3
zn = r n e in = r n 1 + + i n + .
2! 3!
If we set n = and r = 1, then e i = 1 (since sin = 0
and cos = 1) or e i + 1 = 0, Eulers beautiful formula.
150 Chapter 10

the areas of ovals. For him, this implied that geometry can
express in a few lines an infinite series of algebraic terms.
Thus, Newton was able to write the sine and cosine as sin x =
3 5 7 2 4 6
x x3! + x5! x7! + and cos x = 1 x2! + x4! x6! + . Using
such series, calculators sum a few terms to provide a value of
required accuracy. As box 10.1 shows, these series lead to the
formula e i + 1 = 0, one of Eulers most beautiful thoughts,
uniting e, i, , 1, and 0 in one pregnant expression.
Though it was long suspected, finally in 1873 Hermite
showed rigorously that e was transcendental, not the solu-
tion of any algebraic equation of finite degree, and in 1882
Ferdinand Lindemann did the same for . These proofs
ended the dreams of the circle-squarers, whose hopes rested
on the possibility of an algebraic formulation of . One might
have thought that such transcendentals were rare and exotic,
but e and were only the beginning. In 1874, Georg Cantor
showed that there are uncountably many transcendentals,
which are far more dense in the real line than the count-
able infinitude of integers or of rational numbers. He also
showed that algebraic numbers are countable.
To do these later developments justice would require an-
other book and would take us away from Abel, to whom
we now return one final time. His discovery showed that
there is an intermediate class between algebraic irrationali-
ties (such as the radicals earlier expected to solve the quintic)
and the full generality of transcendental numbers. The ul-
traradical solutions of the quintic equation, in general, are
neither the one nor the other; they are algebraic numbers,
but they are not composed of radicals. In this, they are a kind
of amphibian between land (the radicals) and the vast tran-
scendental sea. Here, they are only one among innumerable
kinds of irrational numbers, whose infinitude was already
implicit in Euclids Book X.
Solving the Unsolvable 151

Standing on the edge of this ocean, Abel gazed at the dawn

of a new mathematics, which he began to explore in the few
years remaining him. Great as it was, his proof of the unsolv-
ability of the quintic came during his school years. It would be
unjust to represent it as his only or greatest accomplishment
in mathematics. He went on to do profound work on tran-
scendental functions, infinite series, and integrals that gen-
eralize trigonometric functions, the Abelian integrals and
Abelian Addition Theorem that deeply influenced those
who came after him and that many mathematicians call his
greatest work. He also took Gausss seminal insights even
further than the master. Where the young Gauss had divided
a circle into seventeen parts to construct a regular seventeen-
sided polygon, Abel was able to make a similar division of a
more complex curve, the lemniscate.
Nevertheless, Abel called the solvability of equations my
favorite subject. Though not much given to speculating
about the significance of what he had done, in the introduc-
tion of his final paper on the solution of equations (written
in 1828 but published only posthumously in 1839) Abel ar-
ticulated a kind of credo: One must give to a problem a
form such that it is always possible to solve it, which one
can always do with any problem. Instead of looking for a
relation that one does not know exists or not, one must ask
if such a relation is really possible. He had shown that solv-
ability in radicals is a condition that need not always exist.
By transcending this limitation, Abel solved the unsolvable.
As if taking up Vietes bold claim to leave no problem un-
solved, Abel recorded the power of the infinite extension of
finite algebra. The end of the old assumption that all equa-
tions have a finite solution revealed a new mathematics of
infinite series and noncommutativity.
152 Chapter 10

There is another document that records something of

Abels inner feelings about his work and life, but it is cryptic
and private. Written in 1826, during his work on the lem-
niscate, Abels doodlings in one of his Paris notebooks (fig-
ure 10.3) combine mathematics and reverie in so unguarded

Figure 10.3
A page from Abels Paris notebook (1826). The large lemniscate () is at
the top of the page; the passages discussed in the text are to the right and
under it.
Solving the Unsolvable 153

a way that he probably would have been amused if not mor-

tified to learn that it would become a treasure of the National
Library in Oslo. This page was doubtless meant for no other
eyes than his own.
Abel writes in French and Norwegian, interspersed with
equations and drawings, featuring a striking drawing of a
lemniscate, a large figure . Abel has cross-hatched this
curve and repeatedly outlined it. At the top of this page, Abel
writes complete solution to the equations in which . . . ,
breaking off into equations and intimate thoughts: My
friend . . . beloved . . . Come to me in Gods name . . . . In
the midst of this, an ironic prayer: Our Father who art in
Heaven, give me bread and beer. Listen for once. Then, in
French, speak to me, my dear, her name oddly entwined
in an elliptic integral,
 1
d x Elisa
w= .
0 (1 x 2 )(1 2 x 2 )

Then: Listen . . . Listen . . . Come to me, my friend . . . not, for

once, my solutions to algebraic equations . . . come to me . . . in
all your lewdness.
No page more daringly suggests the mysterious intersec-
tion of desire and mathematics. Who is Elisa? An unknown
beloved, one of the tempting Parisians whose beauty he men-
tioned in a letter, a remembered or even an imagined woman?
Why does he scrawl the name of Soliman den Anden, the
Ottoman emperor Sulieman II who also appears on another
page? What of the neat equations interwoven among these
dreamy fragments? And what of the words near the top of the
page: Goddamn . . . Goddamn my . What is the tone, the
felt meaning of this outburst? Is it humor, vexation, wonder?
We will never know. What remains is equations and .
This page intentionally left blank
 Appendix A
Abels 1824 Paper

Memoir on algebraic equations, in which is demonstrated the im-

possibility of solving the general equation of the fifth degree (1824)
Geometers have been much concerned with the general
solution of equations of the fifth degree and many of them
have sought to prove their impossibility, but (if I am not mis-
taken) none has succeeded until now. Therefore I hope that
geometers will receive kindly this memoir, which aims to fill
this gap in the theory of algebraic equations.
Let

y5 a y 4 + by3 cy2 + dy e = 0 [A1]

be the general equation of the fifth degree and let us suppose

that it is solvable algebraically, that is, one can express y by a
function formed by radicals of the quantities a , b, c, d, and e.
[The commentary is italics in brackets []; equation numbers have
been added also in brackets. Note Abels notation for the coefficients,
which does not follow the conventions of our text. He has chosen
the signs of [A1] so that the coefficient a is the sum of the roots, b the
sum of their products, c the sum of their triples, etc., according to
Girards identities (box 4.1).]
156 Appendix A

It is clear that in this case we can express y in the form

1 2 m1
y = p + p1 R m + p2 R m + + pm1 R m , [A2]

m being a prime number and R, p, p1 , p2 , etc. functions of the

same form as y, and so on until we come to rational functions
of the quantities a , b, c, d, and e.
[For Abels argument for the generality of [A2], which he con-
sidered Step I of the larger proof, see appendix B. Essentially, he
shows that a finite sum of radicals, however they are nested, can
always be expressed in the form [A2], including putting all the
terms on a common denominator, so that eventually one reaches
only rational functions inside the innermost radicals. Abel then
uses [A2] to frame a reductio ad absurdum: he begins by assum-
ing that y can be written as a finite series of algebraic terms, in
which R, p, p1 , p2 , . . . are each algebraic functions of the coeffi-
cients a , b, c, d, e, understood in terms of the successive orders of
functions as nested radicals, explained in appendix B.]
1
We can also assume that it is impossible to express R m by
a rational function of the quantities a , b, etc. p, p1 , p2 , and by
putting pRm in place of R, it is clear that we can make p1 = 1.
1
Then
1 2 m1
y = p + R m + p2 R m + + pm1 R m . [A3]

[Abel simplifies [A2] by getting rid of p1 : he redefines R pRm

  m1 1
1 1
so that p1 R m p1 pRm = R m . In what follows, Abel assumes
1
that this has been done to remove p1 .]
Substituting this value for y in the equation proposed [A1]
and reducing, we obtain a result of this form:
1 2 m1
P = q + q 1 R m + q 2 R m + + q m1 R m , [A4]

q , q 1 , q 2 , etc. being rational and entire functions [i.e., polyno-

mials] of the quantities a , b, c, d, e, p, p2 , . . . and R.
Abels 1824 Paper 157

[Now Abel substitutes the presumed solution [A3] back into

the main equation, [A1]. Doing so leads to [A4], where the new
functions q , q 1 , q 2 , . . . depend on all the previous quantities: the
coefficients a , b, c, d, e and the quantities he has just used, p, p1 ,
p2 , . . . , R. Note that since [A1] sets a polynomial to zero, this
rewritten equation also does so, P = 0. Note also that, for con-
venience, Abel has defined q , q 1 , q 2 , . . . so that each multiplies the
1
appropriate power of R m . Note also that the highest power of R in
[A4] will be m1 m
, as shown in appendix B, pp. 173174.]
In order for this equation to be valid, it is necessary that
q = 0, q 1 = 0, q 2 = 0, . . . , q m1 = 0.
1
In fact, calling R m = z, we have two equations

zm R = 0 and q + q 1 z + + q m1 zm1 = 0. [A5]

[Abel sets out to prove q = 0, q 1 = 0, q 2 = 0, . . . , q m1 = 0;

his proof requires several steps, ending with [A8]. He begins by
1
defining z = R m , so that zm = R or zm R = 0. Also, in [A4]
1
substitute z = R m , giving q + q 1 z + + q m1 zm1 = 0. These
two parts of [A5] constrain z.]
If now the quantities q , q 1 , . . . are not equal to zero, these
equations [A5] necessarily have one or more common roots.
If k is the number of these [common] roots, we know that
we can find an equation of degree k that has as roots the k
roots mentioned and in which all the coefficients are rational
functions of R, q , q 1 , and q m1 . Let

r + r1 z + r2 z2 + + rk zk = 0 [A6]

be that equation. It has these roots in common with the equa-

tion zm R = 0; thus, all the roots of this equation have the
form z, where designates one of the roots of the equation
m 1 = 0. Then substituting [into [A6] z z], we have
158 Appendix A

the following equations,

r + r1 z + r2 z2 + + rk zk = 0 [A7a]

r + r1 z + 2r2 z2 + + k rk zk = 0 [A7b]

r + k2r1 z + k2
2
r2 z2 + + k2
k
rk zk = 0. [A7k]

[In z = z, the subscript is an index that goes from 1 to k (the

total number of common roots in [A5], z, z, 1 z, 2 z, . . . , k2 z).
Substituting z = z back into zm R = 0 yields m zm R = 0,
but zm = R and so this means that m R R = 0. Dividing by R
(which = 0) shows that is a root of the equation m 1 = 0.
Now stands for the whole series of values 1, , 2 , . . . , k2 ,
where Abel has noted that 1 is a root of this equation and calls
1 = . (Note that there are then only k 2 values of since,
of the k roots, the first and second are 1 and .) Abel now substi-
tutes the successive values z = z, z, 2 z, . . . , k2 z back into
[A6]. Substituting 1 for gives [A7a]; substituting gives [A7b],
. . . until substituting k2 yields [A7k].]
From these k equations, one can always find the value
of z expressed by a rational function of the quantities r, r1 ,
r2 , . . . , rk , and, as these quantities are themselves rational
functions of a , b, c, d, e, R,. . . , p, p2 , . . ., it follows that z is
also a rational function of these same quantities, which is
contrary to the hypothesis. Therefore, it is necessary that

q = 0, q 1 = 0, . . . , q m1 = 0. [A8]

[Now Abel makes the critical observation: from the k equa-

tions [A7aA7k], we can always find z as a rational function of
r, r1 , . . . , rk and because we have k simultaneous equations to de-
termine the k unknowns, z1 , z2 , . . . , zk . Notice that this is different
Abels 1824 Paper 159

from the situation with the original equation [A1], which was one
equation to determine five values of y; [A7aA7k] are k linear
equations to determine k unknowns, and can be solved by elim-
ination: Treat each power of z as a separate unknown and solve
the k equations as if they were simultaneous linear equations for
1
those k unknowns. But we assumed that z = R m is not a rational
function of its variables, so we are forced to the only alternative,
q = q 1 = q 2 = = q m1 = 0, concluding the proof of [A8].]
If now these equations are valid, it is clear that the proposed
equation [A1] is satisfied by all the values that one obtains
1
for y by giving to R m all the values
1 1 1 1 1
R m , R m , 2 R m , 3 R m , . . . , m1 R m , [A9]

being a root of the equation

m1 + m2 + + + 1 = 0. [A10]
1 2 m
[If y1 = p + R + p2 R + + pm1 R
m m m1 [A3], then q +
1 m1
q 1 R m + + q m1 R m = 0. This happens because, in substi-
tuting this form y1 into [A1], we get terms like (products of p,
1 2
p2 , . . .) (R m )a (R m )b . Collecting powers, this becomes (prod-
1
ucts of p, p2 , ) (R m )a +2b+ . Since the exponent a + 2b + can
always be written as mi + j, where i, j are integers ( j m 1),
1
then the integral powers of Ri = (R m )mi can be factored out and
included with the products of p, p2 , as q , q 1 , in [A4]. Abel
has just shown that all these qs are zero. Now if we consider y2 , in
1 1
which R m R m , a similar argument applies, except that there
is now a factor of raised to some power (a + 2b + ) multiplied
times these previous factors of q and R. But since the qs are zero,
then each term still vanishes and this y2 also satisfies P = 0 in
1 1
[A4]. The same argument also applies for y3 (R m 2 R m ) and
all the other values of [A9].]
160 Appendix A

We see also that all these values of y are different, for in

the contrary case we would have an equation of the same
form as the equation P = 0, and such an equation leads, as
we have seen, to a result that cannot be valid. The number m
thus cannot exceed 5. Therefore designating by y1 , y2 , y3 , y4 ,
and y5 the roots of the proposed equation [A1], we will have
1 2 m1
y1 = p + R m + p2 R m + + pm1 R m , [A11a]
1 2 m1
y2 = p + R m + 2 p2 R m + + m1 pm1 R m , [A11b]

1 2 m1
ym = p + m1 R m + m2 p2 R m + + pm1 R m . [A11m]

[Note that, if on the contrary y1 = y2 (for example), then [A11a]

= [A11b], which would require that 1 = 0 = 2 1 =
3 1 = , which contradicts [A10]. Therefore all the values of
y are different. Of course, there are special cases of quintics having
equal roots, but they will prove to have simple, rational solutions.
For instance, if all the roots are equal, y = y0 , then the quintic
equation can just be factored into (y y0 )5 = 0, and clearly that
can only hold if the coefficients are highly restricted. Abel also
reminds us that the roots can be no more than five in number.]
From these equations, we easily deduce that

1
p= (y1 + y2 + + ym ), [A12a]
m
1 1
R m = (y1 + m1 y2 + + ym ), [A12b]
m
2 1
p2 R m = (y1 + m2 y2 + + 2 ym ), [A12c]
m

m1 1
pm1 R m = (y1 + y2 + + m1 ym ). [A12m]
m
Abels 1824 Paper 161

1
We see from this that p, p2 , . . . , pm1 , R, and R m are rational
functions of the [roots of the] proposed equation [A1].
[Now he calls these five roots y1 , y2 , . . . , y5 and uses [A3] and the
result of [A9A10] to write them out explicitly in [A11aA11m].
Then he adds these equations up and gets
1
y1 + y2 + + ym = mp + (1 + + 2 + + m1 )R m
2
+ p2 (1 + + 2 + + m1 )R m +
m1
+ pm1 (1 + + 2 + + m1 )R m . [A12.1]

But from [A10] this leads immediately to [A12a], since all the sums
(1 + + 2 + + m1 ) vanish. Now Abel will tease out the other
terms in [A4] by multiplying each equation in such a way as to
1
isolate each term, as follows. To find out what R m is, multiply
[A11a] by 1, [A11b] by m1 , [A11c] by m2 , . . . , [A11m] by .
Then add them up. We get:
y1 + m1 y2 + m2 y3 + + ym = (1 + + 2 + + m1 ) p
1 2
+ m m R m + p2 m (1 + + 2 + + m1 )R m +
m1
+ m pm1 (1 + + 2 + + m1 )R m . [A12.2]
1
Since m = 1, then mR = y1 + m1 y2 + m2 y3 + + ym
m

[A12b]. Exactly similar tricks yield the rest of [A12bm].]

Let us now consider one of these quantities, for example
R. Let
1 2 n1
R = S + v n + S2 v n + + Sn1 v n . [A13]

Treating this quantity in the same manner as y, we obtain a

1
similar result showing that the quantities v n , v, S, S2 , . . . are
rational functions of the different values of the function R,
and, as these [values] are rational functions of y1 , y2 , etc., the
1
functions v n , v, S, S2 , . . . are also.
162 Appendix A

Pursuing this reasoning, we conclude that all the irrational

functions contained in the expression for y are rational func-
tions of the roots of the proposed equation.
[This concludes Step II of the whole proof.]
This being established, it is not hard to complete the de-
monstration. Let us first consider the irrational functions of
1
the form R m , R being a rational function of a , b, c, d, and e.
1
Let R m = r , [where] r is a rational function of [the roots]
y1 , y2 , y3 , y4 , and y5 and R is a symmetric function of these
quantities. Now since the case in question is the general so-
lution of the fifth-degree equation, it is clear that one can
consider y1 , y2 , y3 , y4 , and y5 as independent variables; thus
1
the equation R m = r can hold under this supposition. Con-
sequently, we can interchange the quantities y1 , y2 , y3 , y4 , and
1
y5 among themselves in the equation R m = r , since by this
1
interchange R m necessarily takes m different values since R
is a symmetric function.
1
[Now the final stage of the reductio begins. Abel lets R m = r
and reminds us that he has just shown it to be a rational function
of the roots y1 , y2 , . . . , y5 . Furthermore, R is a symmetric function
of these roots. This means that we can permute the roots among
themselves without changing R. It also means that the equation
1
R m = r can be permuted among the m roots, and r will then assume
the m different values, however they are permuted.]
The function r must also take m different values in per-
muting in all ways possible the five variables it contains. To
show this, it is necessary that m = 5 or m = 2, since m is a
prime number. (See the memoir by M. Cauchy in the Journal
de lecole polytechnique, vol. 17.)
[See appendix C for Abels expansion of this argument. Cauchys
1
theorem states that if m = 5, our function r = R m can take on
only five or two values, never three or four values. Cauchys theorem
Abels 1824 Paper 163

allows r to take only one value, but that would contradict our initial
assumption that all the roots are different.]
First, let m = 5. The function r therefore has five different
values and can consequently be put in the form
1
R 5 = r = p + p1 y1 + p2 y12 + p3 y13 + p4 y14 , [A14]

p, p1 , p2 , . . . being symmetric functions of y1 , y2 , . . . . This

equation gives, on interchanging y1 and y2 ,

p + p1 y1 + p2 y12 + p3 y13 + p4 y14

= p + p1 y2 + p2 y22 + p3 y23 + p4 y24 [A15]

where

4 + 3 + 2 + + 1 = 0. [A16]

But this equation is impossible, so m consequently must

equal 2.
[In his 1826 paper, Abel gives a much fuller account of these terse
assertions, as well as a much simpler argument for the impossibility
of m = 5. First, the simple argument: Consider expressing one of
1 2 4
the roots y1 as in [A11a], y1 = p + R 5 + p2 R 5 + + p4 R 5 and
1
then derive the expression for R 5 = 5 (y1 + 4 y2 + 3 y3 + 2 y4 +y5 ),
1

as in [A12b], where the case m = 5 has been taken. However, this

relation is impossible, since the left-hand side has five values (the
possible values of the fifth root) while the right-hand side has 120
(the permutations of the five roots). Therefore the case m = 5 is
excluded.
Alternatively, here is Abels longer reasoning for [A14A15]:
Consider a rational function v that depends on y1 , y2 , y3 , y4 , y5
and is symmetric under permutations of four out of five of these,
y2 , y3 , y4 , y5 . Then we can express v in terms of y1 and of the coef-
ficients of an equation of which y2 , y3 , y4 , y5 are the solutions. To
164 Appendix A

see this, write (y y2 )(y y3 )(y y4 )(y y5 ) = y 4 py3 +

q y2 r y + s. The roots y2 , y3 , y4 , y5 can be expressed in terms of
the coefficients p, q , r, s, since this is a quartic equation and hence
solvable. Now y1 , y2 , y3 , y4 , y5 are all the roots of a quintic equa-
tion y5 a y 4 + by3 cy2 + dy e = 0 [A1], which means that
(y y1 )(y y2 )(y y3 )(y y4 )(y y5 ) = y5 a y 4 + by3
cy2 + dy e = (y y1 )(y 4 py3 + q y2 r y + s), as was just as-
sumed. Then we can compare these forms of the same equation and
conclude that p, q , r, s can be expressed in terms of the root y1 and
the original coefficients a , b, c, d. (In detail, writing out this last
expression and factoring gives (y y1 )(y 4 py3 + q y2 r y + s) =
y5 ( p + y1 )y 4 + (q + py1 )y3 (r + q y1 )y2 + (s + r y1 )y sy1 ,
so that p = a y1 , q = b a y1 + y12 , r = c by1 + a y12 y13 ,
s = d cy1 + by12 a y13 + y14 .)
Thus, the function v can be expressed rationally in terms of
y1 , a , b, c, d. Then it follows that v can be written as a series of
terms v = r0 + p1 y1 + p2 y12 + + pm y1m , where p0 , p1 , . . . , pm are
polynomial functions of a , b, c, d, e, the dependence on y1 having
been factored out (here Abel uses an argument very similar to that
in appendix B to argue that the denominator can always be incor-
porated into the rational terms p0 , p1 , . . . , pm ). Since y1 is a root
of [A1], then y15 = a y14 by13 + cy12 dy1 + e, which can be used
to re-express all the terms in v = p0 + p1 y1 + p2 y12 + + pm y1m
in which m is 5 or greater.
Thus, we can write any function that is symmetric under per-
mutations of y2 , y3 , y4 , y5 in the form v = p0 + p1 y1 + p2 y12 +
p3 y13 + p4 y14 .
Now if in any such function we permute all five of y1 , y2 , y3 ,
y4 , y5 , then it follows that it takes only one value and is thus sym-
metric (since it is already symmetric in y2 , y3 , y4 , y5 and now is also
in y1 ) or it takes five values (one for each of y1 y1 , y2 , y3 , y4 , y5 .
Thus (since we exclude the symmetric case), this argument shows
that any function that has five different values y1 , y2 , y3 , y4 , y5 must
Abels 1824 Paper 165

take the form v = p0 + p1 y1 + p2 y12 + p3 y13 + p4 y14 [A14], and simi-

larly for the other possible four values y2 , y3 , y4 , y5 . (Abel discusses
various cases in which v takes different numbers of values, but this
is the essence of his argument.)
1
We also know that these different roots are related by [A9], R 5
1 1 1 1 1 1
R 5 , R 5 , 2 R 5 , 3 R 5 , 4 R 5 . Then take [A14] in the form R 5 =
r = p + p1 y2 + p2 y22 + p3 y23 + p4 y24 and multiply both sides by ,
1 1
so as to take R 5 R 5 and hence y1 y2 . By doing this, we get
( p + p1 y1 + p2 y12 + p3 y13 + p4 y14 ) = p + p1 y2 + p2 y22 + p3 y23 + p4 y24
or (reversing the arbitrary names of y1 and y2 , as Abel does) p +
p1 y1 + p2 y12 + p3 y13 + p4 y14 = p + p1 y2 + p2 y22 + p3 y23 + p4 y24 ,
which is [A15]. However, this equation cannot be satisfied unless
either = 1 or y1 = y2 , neither of which is allowed since the roots
are all different. Therefore there can be only two values of r , by
Cauchys theorem.]
Then let
1
R 2 = r, [A17]

and then r must have two different values of opposite sign.

We then have (see the memoir of M. Cauchy)
1
R 2 = r = v(y1 y2 )(y1 y3 ) (y2 y3 ) (y4 y5 )
1
= vS 2 , [A18]

v being a symmetric function.

[Abel writes the two possible values for r as [A17], remembering
that the square root is always ambiguous up to a sign. Cauchys
theorem also implies in this case that r can be written in the form
[A18], where v is a symmetric function (dependent on the coeffi-
1 1
cients) and S 2 is a special function, S 2 = (y1 y2 )(y1 y3 )
(y4 y5 ) discussed in appendix C. Note that S cannot be zero
because no two roots are the same.]
166 Appendix A

Let us now consider irrational functions of the form

 1 1  m1
p + p1 R + p2 R1 + , [A19]

p, p1 , p2 , etc. R, R1 , etc. being rational functions of a , b, c, d,

and e and consequently symmetric functions of y1 , y2 , y3 , y4 ,
and y5 . As we have seen, we must have = = etc. = 2, R =
v 2 S, R1 = v12 S, etc. The preceding function [A19] can thus be
written in the form
 1
 m1
p + p1 S 2 . [A20]

Let
 1
 m1
r = p + p1 S 2 , [A21]
 1
 m1
r 1 = p p1 S 2 . [A22]

Multiplying, we have
 1
rr1 = p 2 p12 S m . [A23]

[Abel reminds us that we now know that we are dealing with

1 1
R 2 , rather than R 5 . Using the definitions R = v 2 S, R1 = v12 S, . . .
he can express irrational functions such as [A19] in the form r =
1 1
( p + p1 S 2 ) m [A21], where he has dropped out the other terms
in [A19] since his object is merely to show us the general form
they will take. Likewise, he writes another irrational term r1 in the
1 1
similar general form r1 = ( p p1 S 2 ) m [A22]. Multiplying out rr1
gives [A23]; he chooses the minus sign in defining r1 to give this
convenient form.]
Now if rr1 is not a symmetric function, m must equal 2,
but in this case r would have four different values, which is
impossible; therefore rr1 must be a symmetric function.
[Abel argues that the product rr1 is a symmetric function; if it
were not, then (by Cauchys theorem) m = 2. What would this
Abels 1824 Paper 167

mean? In [A21], r would have four values, since it would involve

a square root of terms including square roots. This is not allowed;
only m = 5 or m = 2 is possible. So then rr1 is a symmetric
function, [A23].]
Let v be this [symmetric] function [v = rr1 ], then
 1
 m1  1
 m1
r + r 1 = p + p1 S 2 + v p + p1 S 2 = z. [A24]
  m1
1
[Remember that r1 = vr = v p + p1 S 2 .]
This function having m different values, then m must equal 5,
since m is a prime number. Thus we have

z = q + q 1 y + q 2 y2 + q 3 y3 + q 4 y 4
 1
 15  1
 15
= p + p1 S 2 + v p + p1 S 2 , [A25]

q , q 1 , q 2 , etc. being symmetric functions of y1 , y2 , y3 , etc. and

thus rational functions of a , b, c, d, and e.
[Since he has excluded the possibility that m = 2, by elimination,
m = 5. To regain contact with his quest for the roots of the original
equation, he writes z in [A25] in terms of the roots of the quintic,
z = q + q 1 y + q 2 y2 + q 3 y3 + q 4 y 4 , using [A14].]
Combining this equation with the proposed equation, we
can express y in terms of a rational function of z, a , b, c, d,
and e. Now such a function is always reducible to the form
1 2 3 4
y = P + R 5 + P2 R 5 + P3 R 5 + P4 R 5 , [A26]
1
where P, R, P2 , P3 , and P4 are functions of the form p + p1 S 2 ,
p, p1 , and S being rational functions of a , b, c, d, and e.
[Now he turns the expression [A25] inside out, as he has done
many times before, to express y as a function of z and the coefficients
a , b, c, d, e. Because of [A11a], he can write this in the form of
[A26], where all the functions P, R, . . . can be written in the form
1
p + p1 S 2 (because of the reasoning around [A20]).]
168 Appendix A

From this expression for y we derive that

1 1
R5 = (y1 + 4 y2 + 3 y3 + 2 y4 + y5 )
5
1 1
= ( p + p1 S 2 ) 5 , [A27]

where

4 + 3 + 2 + + 1 = 0. [A28]
1
[For the last time, he turns [A26] inside out, to express R 5 in
terms of the roots y1 , y2 , . . . , yielding [A27]; this is just [A12b]
again, with m = 5. He notes, crucially, that he has also established
1 1 1
that R 5 can be equated to ( p + p1 S 2 ) 5 .]
Now the left-hand side [A27] has 120 different values and
the right-hand side has only 10; consequently, y cannot have
the form we have found, but we have proved that y must
necessarily have this form, if the proposed equation is
solvable.
[The left-hand side of [A27] has 120 different values because
y1 can take any of the five values of the five roots, leaving four
possibilities for y2 , three possibilities for y3 , two possibilities for
y4 , and only one for y5 . That means the total number of possible
values of the left-hand side is 5! = 120. But the right-hand side
of [A27] has only ten possible values, since the square root has
two possibilities, multiplied by the five possibilities for the fifth
root. There is no way that something with 120 possible values can
always equal something with only 10. Therefore, the premise fails
that the equation can be solved algebraically, that is, by y given by
[A2].]
Thus we conclude that it is impossible to solve in radicals the
general equation of the fifth degree.
It follows immediately from this theorem that it is also
impossible to solve in radicals general equations of degrees
higher than the fifth.
Abels 1824 Paper 169

[Quite simply, if you multiply an unsolvable quintic by y, it

becomes a sixth-degree equation, with one root y = 0 and the other
roots unsolvable, and likewise for higher degrees. In terms of Abels
proof, the same argument that he has given will continue to apply
when n > 5 since Cauchys theorem still holds. For instance, con-
sider an equation of degree 7. Then Cauchys theorem says that any
function in the solution can take only 7, 2, or 1 values, and we can
reapply all of Abels arguments that these will all lead to contradic-
tions exactly parallel to those for degree 5. It is worth considering
why this doesnt affect degrees 2, 3, and 4. For them, the number
of roots matches the possible values for functions in the solution
(respectively 2, 3, 4). Only for degree 5 or higher does the gap open
between the number of roots and the permitted number of values
for the functions in the solution (5, 2, 1, but never 3 or 4.)]
This page intentionally left blank
 Appendix B
Abel on the General
Form of an Algebraic
Solution

Abel needs to prove that the general solution y of an equation

has the form
1 2 m1
y = p + p1 R m + p2 R m + + pm1 R m , [B1 = A2]

where m is a prime number and R, p, p1 , p2 , etc. are func-

tions of the same form as y, perhaps containing further nested
radicals inside them but ultimately (after a finite number of
radicals within radicals) containing only rational functions of
the coefficients of the original equation. The basic idea is sim-
ple: We assume that a solution has a finite number of terms,
involving a finite number of radicals. What Abel proves is
that, however complex such an expression might be, we can
always arrange it in the form [B1]. The main idea is to make
all the terms (whether inside radicals or coefficients) ratio-
nal by always putting them over a common denominator, a
messy but straightforward process. He did not spell out this
argument in his 1824 paper, so we follow here the account in
his 1826 paper.
First, Abel considers a polynomial function, f (x), which is
a sum of terms, each x raised to some power. A rational func-
tion is defined as the quotient of two polynomial functions,
172 Appendix B

similar to the way a rational number is the quotient of two

integers. Abel means by algebraic function one that can
be expressed by rational functions along with taking roots, a
special case of the modern use of the term algebraic func-
tion for all polynomials f (x, y) = 0. He now sets up a hier-
archy of complexity and defines an algebraic function of the
zeroth order as f (a , b, c, d, e), where f is a rational func-
tion of the coefficients a , b, c, d, e. Then an algebraic function
1
of the first order is f ( p0m ), where p0 is a function of the
zeroth order and f is a rational function of the coefficients
1
in zeroth order. Similarly, a second order function f ( p1m )
involves first-order functions p1 and f . We can keep going to
define an algebraic function of the kth order that has k lev-
els of roots-within-roots; we call this quite general algebraic
1
function v. By factoring m and writing R m as a sequence of
successive radicals, we may assume that m is prime.
We can now write the algebraic function v as VT , where V is
1 2 m
expressed in the form V = v0 + v1 R n + v2 R n + + vm R n , and
likewise for T. (Here we have reverted to Abels notation in
the 1824 paper.) Now Abel considers the series of substitu-
1 1 1 1 1 1
tions that take R n R n , R n 2 R n , . . . , R n n1 R n ,
where = 1 but satisfies n = 1, as in [A10]. Under these
n 1 substitutions, V will in general take the n 1 val-
ues V1 , V2 , . . . , Vn1 . If we multiply the definition v = VT by
1 = VV11 VV22 ...Vn1
...Vn1
, then
T V1 V2 . . . Vn1
v= . [B2]
VV1 V2 . . . Vn1

Abel now shows that the denominator of [B2] is a poly-

1 2
nomial. To see this, remember that V = v0 + v1 R n + v2 R n +
m 1 1
+ vm R n . Under the substitution R n R n , V V1 =
1 2 m 1 1
v0 + v1 R n + 2 v2 R n + + m vm R n ; when R n 2 R n , V
Abel on the General Form of an Algebraic Solution 173

1 2 m
becomes V2 = v0 + 2 v1 R n + 4 v2 R n + + 2m vm R n ; and so
forth. To find the denominator of [B2], we multiply these out
1
and then gather similar terms: VV1 V2 Vn1 = (v0 + v1 R n +
2 m 1 2 m
v2 R n + + vm R n ) (v0 + v1 R n + 2 v2 R n + + m vm R n )
1 2 m 1
(v0 + 2 v1 R n + 4 v2 R n + + 2m vm R n ) (v0 + n1 v1 R n +
2 m
2(n1) v2 R n + + m(n1) vm R n ) = v0n +v0n1 v1 (1+ + 2 + +
2
n1 ) + 2 v0n2 v2 R n (1 + + 2 + + n1 ) + = v0n , where we
1 2 m
have collected the factors of each power of R n , R n , . . . , R n ,
and noted that (1 + + 2 + + n1 ) = 0 [A10], so that
only v0n remains. If v0 is a polynomial, we are done. If v0 con-
tains further radicals nested inside it, this same process can
be repeated as many times as required until we finally reach
polynomials, for we know that there is only a finite number
of subradicals. Thus, the denominator of [B2] is a polyno-
mial and all the irrational functions are in the numerator.
Furthermore, the rational functions in the numerator can be
redefined to include the polynomial denominator.
This is the crucial realization. By the same reasoning, the
1
numerator is a polynomial in terms of R n and powers of the
variables, so that it can be written as
1 2 m
v = q0 + q1 R n + q2 R n + + qm R n , [B3]

a function of order k, where R, q 0 , q 1 , . . . are functions that

are of order k 1 and include the factors absorbing the
polynomial denominator. In the case of the quintic equation,
R, q 0 , q 1 , . . . would all be functions of order 4, that is, roots of
a quartic equation. In [B3], we can always assume that m < n,
for otherwise we can write m = a n + b, where a and b < n
are integers. Then
m a n+b b
Rn = R n = Ra R n . [B4]
174 Appendix B

We can absorb the integral powers of the polynomial Ra into

a redefined coefficient q b = q b Ra . This leaves the fractional
b
powers of R expressed in terms of lower-order fractions, R n .
Because we have factored k with a , b < n, the highest power
1 n1
of R n is less than n, namely R n . This shows that [B4] will
take the stated form [B1], as claimed.
Let us consider some examples. Box 6.1 shows that the
quadratic equation obeys this form. The case of the cubic
equation y3 + a 1 y a 0 = 0 is only slightly more complicated.
The Cardano solution of this equation (see box 2.4) is
   
3 a0 a 02 a 13 3 a0 a 02 a3
y= + + + + 1. [B5]
2 4 27 2 4 27
1
To show that this really is of Abels form, write y = p + R 3 +
2
p2 R 3 . Since our cubic equation has no y2 term, the sum of the
1
roots is 0; but the sum of the roots is also 3 p + R 3 (1 + +
2
2 ) + R 3 (1 + + 2 ) = 3 p, since 1 + + 2 = 0. Hence p = 0.
1 2
Now substitute y = R 3 + p2 R 3 into the cubic equation and
factor out the various powers of R. One gets:
1 2
(R + p23 R2 a 0 ) + R 3 (3 p2 R + a 1 ) + R 3 (3Rp22 + a 1 p2 ) = 0.
[B6]
By [A8] in appendix A, this means that the separate co-
efficients are each zero: R + p23 R2 a 0 = 0, 3 p2 R + a 1 = 0,
and 3Rp22 + a 1 p2 = 0. Thus p2 = 3R
a1
, and by substitution, R2
a3
a 0 R 271 = 0. One solution of this quadratic is R+ = a20 +
 2 1 2 1
a a3 1 1 1
0
+ 271 . Then y = R+3 + p2 R+3 = R+3 a31 R+ 3 . But R+3 R3 =
4  2 1 1
a a3
a31 , where R = a20 40 + 271 . Thus y = R+3 + R3 , which is
exactly the form shown in [B5]. Therefore the solution of the
cubic equation can be written in Abels form.
 Appendix C
Cauchys Theorem on
Permutations

Abels 1824 paper uses Cauchys argument about permuta-

tions, which in turn generalizes Ruffinis results. In his 1826
paper, Abel gives a further account, summarized here and
related to the modern language of group theory. Cauchy con-
siders any function of n variables, f (x1 , x2 , x3 , x4 , x5 ) (here for
the case n = 5). If the value of the function is unchanged
when the values of the variables are permuted, it is said to be
a symmetric function. That is, if the function is symmetric, it
does not matter which variable is which: f (x1 , x2 , x3 , x4 , x5 ) =
f (x2 , x1 , x3 , x4 , x5 ) = f (x5 , x2 , x3 , x4 , x1 ) = , and so on
through all the 5! = 120 possible permutations of the five
variables. If the function is not symmetric, different permu-
tations of x1 , x2 , x3 , x4 , x5 will lead to different values of the
function. In general, the greatest number of such different
values of the function cannot exceed the total number of per-
mutations, n!. Call the values f (A1 ), f (A2 ), . . . , f (An! ), where
A1 , A2 , . . . , An! are all the permutations of the original order
x1 , x2 , x3 , . . ..
Assume that the functions takes p different values, where
p < n!. First we show that p must be a divisor of n!. Then
some of the n! values must be equal. Let us suppose that
176 Appendix C

the first m values are equal, f (A1 ) = f (A2 ) = = f (Am ).

In this series, let us make the successive substitutions A1
Am+1 , A2 Am+2 , . . . , Am A2m . This gives another m equal
values (though in general not equal to the first m). We can
then continue this process of substitution, finally exhausting
the n! possibilities, dividing them into p sets, each having m
elements. Therefore, pm = n! and so p is a divisor of n!.
In modern group theory, this is called Lagranges Theo-
rem. Consider a group G and call the number of its elements
|G| (where |G| = n! if G is the group of permutations of n
elements). Consider any subgroup H contained in G.
Lagranges Theorem states that the number of elements of
H, namely |H|, must be a divisor of |G|. Let H be the sub-
group of permutations that fix one of the p possible values of
our function, say f (A1 ). Then the procedure described above
corresponds to partitioning the group G into cosets of H
in G. (If the group G is noncommutative, left cosets are
distinguished from right.) The argument just given shows
that these cosets partition the elements of G into exhaustive,
exclusive classes of p elements and so p must be a divisor
of |G| = n!. Since the cosets divide up G into these exhaus-
tive and exclusive classes, it is natural to think of them as
the quotient G/H, which is a group when H is a normal
subgroup (defined on pp. 193194).
A good example of this is the arithmetic of a clock. On
a 24-hour clock, each numeral is a coset, representing an
equivalence class of each hour: 13 oclock is equivalent to
13 + 24 = 37 oclock, 13 + 48 = 61 oclock, etc. If we then shift
to a 12-hour clock, we take the quotient of the 24-hour clock
divided by 2, since now 13 and 1 oclock are also equivalent,
as are 25 and 37 oclock, and so forth. This is only a brief clar-
ification; for more examples and details, see the works listed
in the notes, pp. 191192.
Cauchys Theorem on Permutations 177

In general, finite groups can be compared to molecules that

can be analyzed by breaking them into smaller molecules and
ultimately into their constituent atoms. As the same chemi-
cal atoms can form different molecules, the same mathemat-
ical atoms can form different groups. Consider a group G
that contains a normal subgroup H that is proper (H is
neither simply G nor the identity). Then the quotient group
G/H and H are constituent sub-molecules of G. Continue
breaking these sub-molecules down in the same way into
ever smaller pieces, until we cannot break them down fur-
ther: the collection of indecomposable pieces that we get are
the atoms of the group G. Jordan and Holders famous
theorem proves that this collection of atoms is determined
entirely by the group G. From this standpoint, Abels proof
means that an equation is solvable in radicals if and only if
the atoms of its group are all abelian (in particular, cyclic
and having orders that are prime numbers, if the group is
finite).
We began by considering a symmetric function of n vari-
ables, which has only one value when we permute the
variables. Other functions may have two values. For exam-
ple, consider the function that Cauchy and Abel call

S = (x1 x2 )(x1 x3 )(x1 x4 )(x1 x5 )(x2 x3 ) (x4 x5 ),

in which all five variables appear antisymmetrically, mean-

ing that if we make the permutation x1 x2 , the function

changes
sign but is equal in absolute magnitude: S
S. In fact, any interchange of two variables will produce
two possible values for the function, opposite in sign. We
can also, of course, multiply this antisymmetric function by
a symmetric function, which will not change upon permuta-
tion (Abel uses just this function in his [A18], appendix A).
178 Appendix C

It is not always possible to make a function that has any

given number of values larger than 2, even if it is a divisor
of n!. In what follows, let p be the largest prime less than or
equal to n. Let Am be any permutation of order p, meaning
that when it is applied p times we return to the starting point,
which we will write as f (Am ) p = f (Am )0 . If we assume that
the function takes on fewer than p values, then two among

those values must be equal, say f (A m )r = f (A m )r , where
0 < r, r < p 1. Then apply the permutation A m to both

sides of this equation ( p r ) times: f (A m )r + pr = f (A m )r + pr ,
or f (A m ) = f (A m ) j , since (r + p r ) = p, f (A m ) p = f (A m ),
and we define j = r + p r . Not only are two of the values
equal, f (A m ) = f (A m ) j , but also f (A m ) = f (A m )b j , where b
is any integer, each multiple of j representing that many
times around the circle. Since p is a prime number, we can
always find two integers a , b such that b j = pa + 1. Hence
f (A m )b j = f (A m ) pa +1 = f (A m ), which we showed is also equal
to f (A m ) = f (A m ) pa . This means that adjacent values of the
function are equal, f (A m ) pa = f (A m ) pa +1 , so that the permuta-
tion A m does not change the value of the function. Therefore,
Cauchy concludes, if the function takes fewer values than p,
it is not changed by a permutation of order p.
Let us imagine a function that takes fewer values than p
(assuming now that p > 3, as will be true in the case of the
quintic, for which n = p = 5). From our last result, we know
that this function is unchanged by a permutation of order
p > 3. Now to make such a permutation, we must trans-
pose variables. For example, to go from f (x1 , x2 , x3 , x4 , x5 )
to f (x2 , x1 , x3 , x4 , x5 ), we must transpose x1 x2 . Likewise,
any arbitrary permutation is composed of a certain number
of transpositions between the variables. This number can be
even or odd.
Cauchys Theorem on Permutations 179

Now consider two permutations of order p: In the first,

let xa xb , xb xc , xc xd , . . . , x p xa , replacing each vari-
able with the next in the list. To use the notation intro-
duced in chapter 8, this is (abcd p). Clearly, this is a
substitution of order p, for after p such transpositions, we re-
turn to the starting point. Similarly, consider a second
series: xb xc , xc xa , xd xb , xe xd , x f xe , . . . , xa x p
or (ap edbc). This too is a substitution of order p, for after p
such substitutions, we would come around to the starting po-
sition. Indeed, it is a slightly varied version of the first substi-
tution (moving xa to the end). So both will leave the function
unchanged, by the result in the preceding paragraph. Now
notice what happens when we perform the first and then the
second substitution in that order: xa xb xc , xb xc
xa , xc xd xb , xe x f xe , x f xg x f , . . . , x p
xa x p or (abcd p) (ap edbc) = (acb). After the first
three terms, the variables remain unchanged after both sub-
stitutions are performed. So these two permutations per-
formed in sequence yield xa xc , xb xa , xc xb , with
all the other variables unchanged. Though these examples
may seem arbitrary, they are not. Any permutation of order
p must take the form of these examples, for we could have
used any combination of variables in place of xa , xb , xc , . . . .
The upshot is that, in general, given any 3-cycle, we can al-
ways find two permutations of order p ( p-cycles, as they are
called in chapter 8) whose product is the given 3-cycle.
Since neither permutation of order p can change the value
of the function (by our earlier result), neither can both of
them, and thus neither can a permutation of order 3. We
now show that a permutation of order 3 can always be writ-
ten as the result of two permutations of order 2, which are
just transpositions. Consider the one we discussed earlier,
xa xc , xb xa , xc xb . It can be written as the transposition
180 Appendix C

xa xb followed by the transposition xb xc . To check this,

do them in the order specified, xa xb xc , xb xa , xc xb ,
giving us the same result: (ab) (bc) = (acb). Again, this is
quite general in form, so any permutation of order 3 is equiv-
alent to two permutations of order 2 (transpositions). But we
just learned that the function is unchanged by a permuta-
tion of order 3; therefore, it is unchanged by the product
of two transpositions. Similarly, it is unchanged by an even
number of transpositions. To prove this, it suffices to show
that any two distinct transpositions can be written as two
3-cycles. Consider, then, xa xb followed by xc xd . It
can also be accomplished by xa xb xc followed by
xc xa xd , verifying the assertion: (ab) (cd) = (abc)
(cad). Thus, since two transpositions (or any even number) do
not change the function, each transposition changes the sign
of the function. Two transpositions change the sign twice,
restoring the original sign, whereas any odd number of trans-
positions must change the sign of the function. They are anti-
symmetric, to use the term introduced earlier.
In either case, the conclusion is that the function we have
been discussing can have at most two values. Remember that
all this followed if p n. So finally we reach the main theorem
of Cauchy that is so important for Abel: The number of different
values a function of n values can take cannot be lower than the
largest prime number p n without becoming equal to 2. Note
that, if n = 5, p = 5 also, so that in the case of the quintic,
either the function has five values or it has at most two values.
In this form, Cauchy and Ruffinis insight is crucial to the last
stage of Abels proof.
 Notes

1 The Scandal of the Irrational

Pythagoras of Samos: Scholars have questioned every aspect of the tradi-

tional account. See Walter Burkert, Lore and Science in Ancient Pythagoreanism,
tr. Edwin L. Minar Jr. (Cambridge, MA: Harvard University Press, 1972)
and also C. A. Huffman, The Pythagorean Tradition, in The Cambridge
Companion to Early Greek Philosophy, ed. A. A. Long (Cambridge: Cambridge
University Press, 1999), 6687. Other accounts include Peter Gorman,
PythagorasA Life (London: Routledge and Kegan Paul, 1979) and Leslie
Ralph, PythagorasA Short Account of His Life and Philosophy (New York:
Krikos, 1961). For a collection of original sources, see The Pythagorean Source-
book and Library, ed. Ken Sylvan Guthrie (Grand Rapids, MI: Phares Press,
1987).
Euclid: The standard annotated edition is The Thirteen Books of Euclids
Elements, tr. Thomas L. Heath (New York: Dover, 1956), which gives the
proof cited by Aristotle (Prior Analytics 41a, 2627) at 3:2. For an engaging
modern encounter with this classic, see Robin Hartshorne, Geometry: Euclid
and Beyond (New York: Springer-Verlag, 2000).
Pythagorean mathematics and the golden ratio: See H. E. Huntley, The
Divine Proportion: A Study in Mathematical Beauty (New York: Dover, 1970).
R
. ta: A. L. Basham, The Wonder That Was India (New York: Grove Press, 1954),
113, 236237. For the Indo-European roots, see Julius Pokorny, Indogermanis-
ches Etymologisches Worterbuch (Bern: Francke Verlag, 1959), 1:5660. I thank
Eva Brann for drawing my attention to the connection between arithmos and
rite.
182 Notes to pp. 117

Plato: The episode of the slave boy is in Meno 84d85b. See also Jacob Klein,
A Commentary on Platos Meno (Chapel Hill, NC: University of North Carolina
Press, 1965), 103107, especially his point that, although the propositions
came from Socrates, the assent and the rejection came from nobody but the
boy himself (105). For Menos character, see Xenophon, Anabasis 2: 2129.
The young irrationals are discussed in Platos Republic, Book VII, 534d.
For the story of the black and white horses, see Phaedrus 253d254e; for the
story of Theaetetus, see Theaetetus 142a150b; Socrates as midwife, 150c
151d. A poros is a means of passage, like a bridge or ferry, so aporia signifies
being stuck, unable to cross over. For a helpful discussion of Theodorus
and Theaetetus, see Wilbur Richard Knorr, The Evolution of the Euclidean
Elements (Dordrecht: D. Reidel, 1975), 62108, which includes a plausible
account why Theodorus stopped at 17 (181193) and a critique of earlier
explanations (109130).
Greek mathematics: For a helpful survey, see the classic book by Carl B.
Boyer, A History of Mathematics, second ed., revised by Uta C. Merzbach (New
York: John Wiley, 1991), 4399 (especially 7274 on incommensurability),
100119 (Euclid), cited hereafter as HM. Another essential classic is Jacob
Klein, Greek Mathematical Thought and the Origin of Algebra, tr. Eva Brann
(New York: Dover, 1992), 3113. See also Morris Kline, Mathematical Thought
from Ancient to Modern Times (New York: Oxford University Press, 1972),
2834, hereafter cited as MT.
Pappus on the irrational: The Commentary of Pappus on Book X of Euclids
Elements, tr. William Thomson (Cambridge, MA: Harvard University Press,
1930), 6465, translation revised thanks to Bruce Perry.
Pentagons and irrationality: See Kurt von Fritz, The Discovery of Incom-
mensurability by Hippasus of Metapontum, Annals of Mathematics 46, 242
264 (1945).
Touchstone and torture: I have discussed the issue of experiment as the
torture of nature in my book Labyrinth: A Search for the Hidden Meaning
of Science (Cambridge, MA: MIT Press, 2000), chapter 2, in which I argue
that Francis Bacon did not say or mean that nature was to be abused by
the experimental trials of the new science. Though Bacon did not use the
charged term torture in this context, Plato uses it to describe the ardu-
ous ordeal of dialectic and inquiry; however, from the context it is clear
that he understands it to be noble, not base or abusive. This also applies to
his daring metaphor of philosophical inquiry as parricide; see my paper
Desire, Science, and Polity: Francis Bacons Account of Eros, Interpretation
26:3, 333352 (1999), note 10. For the later thinkers, see my Wrestling with
Proteus: Francis Bacon and the Torture of Nature, Isis 90:1, 8194 (1999),
Notes to pp. 1825 183

Nature on the Rack: Leibniz Attitude towards Judicial Torture and the
Torture of Nature, Studia Leibnitiana 29, 189197 (1998), and Proteus Un-
bound: Francis Bacons Successors and the Defense of Experiment, Studies
in Philology 98:4, 428456 (2001).
Euclid on irrationals: See The Elements, Book X, propositions 1 (on the indef-
inite divisibility of any magnitude) and 115 (the infinite number of kinds of
irrationals). For a general survey of the context of numbers and irrationality,
see Midhat Gazale, Number: From Ahmes to Cantor (Princeton: Princeton Uni-
versity Press, 2000). For the development of the Greek theory of ratio, see Sir
Thomas Heath, A History of Greek Mathematics (New York: Dover, 1981), 1:90
91, 154157, and Howard Stein, Eudoxos and Dedekind: On the Ancient
Greek Theory of Ratios and Its Relation to Modern Mathematics, Synthese
84, 163211 (1990).
Greek music: See M. L. West, Ancient Greek Music (Oxford: Clarendon Press,
1992), 233242, and the invaluable collection of original texts with commen-
tary in Greek Musical Writings, ed. Andrew Barker (Cambridge: Cambridge
University Press, 1989), 1:137, 188, 411, 419, including also Euclids musical
treatise in 2:190208.I have found no passage where an ancient writer noted
the near equality of 2 and the tritone, which is understandable given their
primal assumption that musical interval is inherently a ratio.

2 Controversy and Coefficients

Babylonian mathematics: HM, 2342; MT, 314. For an interesting compar-

ison of Greek and Babylonian approaches to mathematics, see Simone Weil,
Seventy Letters, tr. Richard Rees (London: Oxford University Press, 1965),
112127.
Chinese algebra: For a general overview, see Joseph Needham, Science and
Civilisation in China (Cambridge: Cambridge University Press, 1959), 3:1
170, with a comment on the cubic problem on 125126. For a discussion
of an important text, with valuable excerpts and translations, see J. Hoe,
The Jade Mirror of the Four UnknownsSome Reflections, Mathematical
Chronicle 7, 125156 (1978), who argues that the inherent symbolism of the
Chinese language moves in the direction of modern algebraic symbolism.
Al-Khwarizm: See my Labyrinth, chapter 7. His work (called in Arabic
Al-jabr wal muqabalah) is included in A Source Book in Mathematics, 1200
1800, ed. D. J. Struik (Cambridge, MA: Harvard University Press, 1969),
5560. See also B. L. van der Waerden, A History of Algebra from al-Khwarizm
184 Notes to pp. 2628

to Emmy Noether (New York: Springer-Verlag, 1980). For an excellent survey,

see Karen Hunger Parshall, The Art of Algebra from al-Khwarizm to Viete:
A Study in the Natural Selection of Ideas, History of Science 26, 129164
(1988). There is a valuable collection of essays in Roshdi Rashed, The De-
velopment of Arabic Mathematics: Between Arithmetic and Algebra (Dordrecht:
Kluwer Academic, 1994).

Khayyam: For a valuable collection of translations with helpful commen-

tary, see R. Rashed and B. Vahabzadeh, Omar Khayyam the Mathematician
(New York: Bibliotheca Persica, 2000). Rashed argues that Omar the poet
and Omar the mathematician may not be the same person; his assertion
remains controversial.

Shylock on commerce: The Merchant of Venice, I.iii.1821.

Fibonacci: See HM, 254257, John Fauvel and Jeremy Gray, The History of
Mathematics: A Reader (London: Macmillan, 1987), 241243, and J. Gies and
F. Gies, Leonard of Pisa and the New Mathematics of the Middle Ages (New York:
Crowell, 1969).
e: See Eli Maor, e: The Story of a Number (Princeton: Princeton University
Press, 1998). If $1 is invested in a bank account at 100% per annum, com-
pounded annually, at the end of one year its value will be $2. But if the
interest is compounded at every instant, continuously, at the end of one
year the account will be worth e = 2.718 . . . dollars.
Pacioli: R. Emmett Taylor, No Royal Road: Luca Pacioli (New York: Arno Press,
1980), contains a translation of the portions of the Summa relating to double-
entry bookkeeping, as does B. S. Yamey, Luca Paciolis Exposition of Double-
Entry Bookkeeping; Venice 1494 (Venice: Abrizzi, 1994), 933, 95171. See also
R. G. Brown and K. S. Johnston, Pacioli on Accounting (New York: McGraw-
Hill, 1965), and Richard H. Macve, Paciolis Lecacy, in Accounting History
from the Renaissance to the Present: A Remembrance of Luca Pacioli, ed. T. A. Lee,
A. Bishop, R. H. Parker (New York: Garland, 1996), 330. For the connections
with the development of printing, see E. L. Eisenstein, The Printing Press as
an Agent of Social Change: Communications and Cultural Transformations in
Early-Modern Europe (Cambridge: Cambridge University Press, 1979), 548.
Pacioli includes idealized fonts based on geometry at the end of his De divina
proportione (Venice: 1507; reprint Maslianico: Dominioni, 1967).
Leonardo da Vinci and Pacioli: See Carlo Zammattio, Augusto Marinoni,
and Anna Maria Brizio, Leonardo the Scientist (New York: McGraw-Hill,
1980), 88117, and Emanuel Winternitz, Leonardo da Vinci as a Musician
(New Haven: Yale University Press, 1982), 1016.
Notes to pp. 2840 185

Piero della Francesca: For an outstanding account of his great mathemat-

ical achievements, see Mark A. Peterson, The Geometry of Piero della
Francesca, Mathematical Intelligencer, 19:3, 3340 (1997), which also clari-
fies Paciolis plagiarisms from Pieros works and discusses the relation of
algebra and geometry in Pieros time.
History of double-entry bookkeeping: See Michael Chatfield, A History of
Accounting Thought (Huntington, NY: Krieger Publishing, 1977), 318, who
emphasizes that the significance of the . . . integration of real and nominal
accounts [in double-entry bookkeeping] far surpasses every other aspect
of accounting development. For the connection between economic and
mathematical developments, see also Frank Swetz, Capitalism and Arithmetic:
The New Math of the 15th Century (La Salle, IL: Open Court, 1987).
Cardano: See Girolamo Cardano, Ars magna, or the Rules of Algebra,
tr. T. Richard Witmer (New York: Dover, 1993), which unfortunately ren-
ders Cardanos mathematical expressions into modern algebraic notation,
instead of retaining his own rhetorical mode of expression that does not rely
on such symbolism. For his colorful autobiography, see Girolamo Cardano,
The Book of My Life (New York: Dover, 1962), and also ystein Ore, Cardano,
the Gambling Scholar (Princeton: Princeton University Press, 1953), which
treats the cubic and quartic equations on 59107 (the commercial problem
cited is on 6970), giving very helpful excerpts from the invective of the
combatants and also the complete text of Cardanos Book on Games of Chance
(first printed posthumously in 1663). See also Anthony Grafton, Cardanos
Cosmos: The Worlds and Works of a Renaissance Astrologer (Cambridge, MA:
Harvard University Press, 2000).
Tartaglias poem: Fauvel and Gray, History of Mathematics, 255256, which
also includes other documents from the controversy on 253263.
Solutions of cubic and quartic equations: For a complete and lucid account,
see the classic account by J. V. Uspensky, Theory of Equations (New York:
McGraw-Hill, 1948), 8298. William Dunham gives an engaging presenta-
tion in Journey through Genius: The Great Theorems of Mathematics (New York:
John Wiley, 1990), 133154. See also HM, 282288, and MT, 263272. My ac-
count of completing the cube draws on Parshall, The Art of Algebra, 144
147. For a neat exposition of the factorization of these equations, see Morton
J. Hellman, A Unifying Technique for the Solution of the Quadratic, Cubic,
and Quartic, American Mathematical Monthly 65, 274276 (1958); Hellman
shows the failure of this technique to the quintic in The Insolvability of the
Quintic Re-Examined, American Mathematical Monthly 66, 410 (1959).
Nature does not permit it: Cardano, Ars magna, 9.
186 Notes to pp. 4149

Viete and the new algebra: The seminal work is Klein, Greek Mathemat-
ical Thought, 161185; I cite Vietes Introduction to the Analytical Art
from the translation in this volume by J. Winfree Smith (315353), which
discusses John Walliss legal account of species at 321322, n. 10. See
also Helena M. Pycior, Symbols, Impossible Numbers, and Geometric Entan-
glements (Cambridge: Cambridge University Press, 1997), 2739. I have dis-
cussed the connections between Vietes work in codebreaking and algebra
in my Labyrinth, 7383, and (in greater detail) in Secrets, Symbols, and Sys-
tems: Parallels between Cryptanalysis and Algebra, 15801700, Isis 88, 674
692 (1997). For translations of the original Viete documents, see my paper
Francois Viete, Father of Modern CryptanalysisTwo New Manuscripts,
Cryptologia 21:1, 129 (1997).

History of symbolic notation: See Florian Cajori, A History of Mathematical

Notations (Mineola, NY: Dover, 1993), 117123 (Cardano), 181187 (Viete).

3 Impossibilities and Imaginaries

Van Roomens problem: See Vietes response, Ad Problema, qvod omnibvs

mathematicis totivs orbis contruendum proposuit Adrianus Romanus
(1595), in his Opera Mathematica (Hildesheim: Georg Olms, 2001), 305324,
discussed in HM, 309311, and also by Guido Vetter, Sur lequation du
quarante-cinquieme degre dAdriaan van Roomen, Bulletin des sciences
mathematiques (2) 54, 277283 (1954). Viete had derived expressions for the
sines or cosines of multiple angles, such as sin n = n cosn1 sin
n(n1)(n2)
321 cosn3 sin3 + , where the subsequent signs alternate and
the coefficients are the alternate numbers in the arithmetic triangle (Pascals
triangle). Viete set K = sin 45, which his formula could then express in
terms of x = 2 sin , yielding van Roomens equation (see box 2.7). To solve
it, he found = (arcsin K )/45, giving x = 2 sin .
Kepler: See my paper Keplers Critique of Algebra, Mathematical Intel-
ligencer 22:4, 5459 (2000), which gives the full quotations and sources,
and also my Labyrinth, 9194. For Keplers treatment of the heptagon, see
Johannes Kepler, The Harmony of the World, tr. E. J. Aiton, A. M. Duncan, and
J. V. Field (Philadelphia: American Philosophical Society, 1997), 6079, and
D. P. Walker, Keplers Celestial Music, in his Studies in Musical Science in
the Late Renaissance (Leiden: E. J. Brill, 1978), 3462; J. V. Field, Keplers Geo-
metrical Cosmology (Chicago: University of Chicago Press, 1988), 99105; J. V.
Field, The Relation between Geometry and Algebra: Cardano and Kepler
on the Regular Heptagon, in Girolamo Cardano: Philosoph, Naturforscher, Arzt,
ed. E. Kessler (Wiesbaden: Harrassowitz Verlag, 1994), 219242. For further
Notes to pp. 4960 187

discussion of the heptagon and its relation to angle trisection, see Andrew
Gleason, Angle Trisection, the Heptagon, and the Triskaidecagon, Ameri-
can Mathematical Monthly, 95, 185194 (1988).

Galileo on the Book of Nature: See The Assayer in Discoveries and Opin-
ions of Galileo, tr. Stillman Drake (New York: Doubleday, 1957), 237238. For
Galileos attitude toward mathematics, see Carl B. Boyer, Galileos Place in
the History of Mathematics, in Galileo, Man of Science, ed. Ernan McMullin
(New York: Basic Books, 1967), 232255.
Negative and imaginary quantities: HM, 219220 (negative numbers in
Hindu mathematics), 276278 (Chuquet), 287289 (Bombelli), 305306
(Girard). See also Pycior, Symbols, Impossible Numbers. For Gausss account of
negatives and imaginaries, see the excellent anthology From Kant to Hilbert:
A Source Book in the Foundations of Mathematics, ed. William B. Ewald (Oxford:
Clarendon Press, 1996), 1:310313, 307 (Leibniz on the amphibian). For an
engaging account of the developing conception of imaginary numbers, see
Barry Mazur, Imagining Numbers (particularly the square root of minus fifteen)
(New York: Farrar Straus Giroux, 2002). See also Paul J. Nahin, An Imaginary
Tale: The Story of 1 (Princeton: Princeton University Press, 1998).

Girard: See his 1629 treatise Invention nouvelle en lalgebre, in The Early Theory
of Equations (Annapolis, MD: Golden Hind Press, 1986).
Bombelli: For a thoughtful introduction that includes translated selections,
see Federica La Nave and Barry Mazur, Reading Bombelli, Mathematical
Intelligencer 24:1, 1221 (2002).
Descartes: See The Geometry of Rene Descartes, tr. David Eugene Smith and
Marcia L. Latham (New York: Dover, 1954), 174175 (introduction of imag-
inary roots), 160161 (Descartess rule of signs), 162175 (methods for in-
creasing or scaling the value of roots; Descartess relativity), 2237 (the
locus problem for three or more lines), 176192 (methods for solving higher-
degree equations), 220239 (solution of a special sixth-degree equation, using
conic sections), 240241 (the general method to construct all problems; the
pleasure of discovery). For a discussion of Descartess concept of number,
see Klein, Greek Mathematical Thought, 197211.

4 Spirals and Seashores

Newton: For his Lectures on Algebra, 16731683, see The Mathematical Pa-
pers of Isaac Newton, ed. D. T. Whiteside (Cambridge: Cambridge University
Press, 1972), 5:130135, 565 (two sample passages showing his treatment
188 Notes to pp. 6078

of equations). Quotations in the text are from Isaac Newton, The Principia:
Mathematical Principles of Natural Philosophy, tr. I. Bernard Cohen and Anne
Whitman (Berkeley: University of California Press, 1999), 485 (geometrical
synthesis), 483485 (Newtons solution to locus problem), 511513 (lemma
28). For further discussion of the significance and validity of lemma 28,
see my paper The Validity of Newtons Lemma 28, Historia Mathematica
28, 215219 (2001), and also Bruce Pourciau, The Integrability of Ovals:
Newtons Lemma 28 and Its Counterexamples, Archive for the History of
Exact Sciences, 55, 479499 (2001). For the relation of Newtons mathematical
preferences to his other projects, see my Labyrinth, 113133.
Tschirnhaus, Bring, and Jerrard: HM, 432434; for Jerrards mistaken solu-
tion, see chapter 9 notes, below.
Leibniz on the quintic equation: See the helpful historical overview in
Nicolas Bourbaki (the pseudonym of a group of French mathematicians),
Elements of the History of Mathematics, tr. John Meldrum (New York: Springer-
Verlag, 1994), 6980 at 74.
Fundamental Theorem of Algebra: For a complete treatment of all of Gausss
proofs, see Benjamin Fine and Gerhard Rosenberger, The Fundamental The-
orem of Algebra (New York: Springer-Verlag, 1997), especially 182186 on
Gausss original proof. Their treatment is based on Uspensky, Theory of Equa-
tions, 293297, which I also follow in my exposition. See also Heinrich Dorrie,
100 Great Problems of Elementary Mathematics, tr. David Antin (New York:
Dover, 1965), 108112. There is a nice one-page proof of this theorem by
Uwe F. Mayer, A Proof That Polynomials Have Roots, College Mathematics
Journal 28:1, 58 (1999).

5 Premonitions and Permutations

Lagrange and Vandermonde: See the classic sturdy by Hans Wussing, The
Genesis of the Abstract Group Concept, tr. Abe Shenitzer (Cambridge, MA: MIT
Press, 1984), 7179, van der Waerden, History of Algebra, 7683, and MT, 600
606, which suggests at 605 that Lagrange was drawn to the conclusion that
the solution of the general higher-degree equation (for n > 4) by algebraic
equations was likely to be impossible, though Lagrange never expressed
this opinion explicitly. His paper Reflexions sur la resolution algebrique
des equations (17701771) is in uvres de Lagrange (Paris: Gauthier-Villars,
1869), 3:205422. Citations in the text are taken from van der Waerden,
History of Algebra, 81.
Notes to pp. 7988 189

G. E. Montucla (17251799): Cited from the excellent article by Raymond G.

Ayoub, Paolo Ruffinis Contributions to the Quintic, Archive for History of
Exact Sciences 23, 253277 (1980), at 257.

Gauss: W. K. Buhler, Gauss: A Biographical Study (New York: Springer-Verlag,

1981), 36 (Gausss lack of interest in combinatorics).

Gauss on the quintic: Carl Friedrich Gauss, Disquisitiones Arithmeticae,

tr. Arthur A. Clarke (New Haven: Yale University Press, 1965), 445 (a passage
added in proof, according to Buhler, Gauss, 77), 407460 (general discussion
of the problem of equal division of a circle, also known as the cyclotomic
equation). For a helpful simplified discussion of the 17-gon and the cyclo-
tomic equation, see Dorrie, 100 Great Problems, 177184. There is also an
excellent account of this in Felix Klein, Famous Problems of Elementary Geom-
etry, tr. W. W. Beman and David Eugene Smith (New York: Chelsea, 1962),
2441.

Ruffini: See Wussing, Group Concept, 8084, van der Waerden, History of Al-
gebra, 8385, and particularly Ayoub, Paolo Ruffinis Contributions to the
Quintic, including the quote from Ruffini at 263 (behold a very important
theorem). Another useful article is R. A. Bryce, Ruffini and the Quintic
Equation, in First Australian Conference on the History of Mathematics, ed.
John N. Crossley (Clayton, Victoria: Department of Mathematics, Monash
University, 1981), 531, which points out a number of errors in Ruffinis work
that substantiate the attribution of the theorem to Abel alone. To my knowl-
edge, none of Ruffinis papers has been translated into English; they are
available in the original Italian in Opere Matematiche di Paolo Ruffini (Palermo:
Tipografia Matematica di Palermo, 19151953), 2 volumes.

6 Abels Proof

For a comprehensive account of Abels life and its Norwegian context, see
Arild Stubhaug, Niels Henrik Abel and His Times (Berlin: Springer-Verlag,
2000), 178183 (the arrival of Holmboe), 239240 (Abels false solution to the
quintic), 297 (Abel and Fermats Last Theorem). Still useful is the briefer,
earlier work by ystein Ore, Niels Henrik Abel: Mathematician Extraordinaire
(Minneapolis: University of Minnesota Press, 1957). There are interesting
essays also in the volume marking the centenary of Abels birth, Niels Henrik
Abel: Memorial publie a loccasion du centenaire de sa naissance (Kristiania: Jacob
Dybwad, 1902).

Fermats Theorem: For Gausss proof in the case n = 3, see Dorrie, 100 Great
Problems, 96104; for a popular history through Andrew Wiless 1993 proof,
190 Notes to pp. 88108

see Simon Singh, Fermats Enigma (New York: Walker, 1997). For Abels work
on this problem, see his 1826 letter to Holmboe, cited in uvres Completes
de Niels Henrik Abel, 2:254255.

Abels Theorem: Modern treatments tend to subsume Abels own work

in Galois theory. Some older textbooks give an account of Abels work,
though sometimes in an unfamiliar notation; see the citations below under
Galois Theory and J. Pierpont, On the Ruffini-Abelian Theorem, Bulletin
of the American Mathematical Society 2, 200221 (1896). For helpful accounts
phrased in the language of modern algebra, see Lars Garding and Christian
Skau, Niels Henrik Abel and Solvable Equations, Archive for the History of
Exact Sciences 48, 81103 (1994) and Michael I. Rosen, Niels Hendrik Abel
and Equations of the Fifth Degree, American Mathematical Monthly 102, 495
505 (1995). Dorrie, 100 Great Problems, 116127, gives an account of Leopold
Kroneckers clarified version of the theorem.
Abel and Ruffini: The quote is my translation from Abels 1828 paper Sur
la resolution algebrique des equations, uvres Completes de Niels Henrik
Abel, ed. L. Sylow and S. Lie (New York: Johnson Reprint, 1965), 2:217243
at 218. Sylow notes (1:293) that Abel first commented on Ruffini in 1826 in
an unsigned note in Ferussacs Bulletin.

7 Abel and Galois

Abels later life: Stubhaug, Abel, 329331 (Crelle), 395420 (Abel in Paris),
468 (appeal of the four French mathematicians to the Swedish king). Quotes
from Abel: 471 (monstrous egotists), 398402 (billiards and theater), 424
(poorer than a church mouse), 474 (correspondence with Legendre), 471
(quite alone), 475493 (Abels death and his despair), 409 (on Cauchy).
Abel on commutativity: My translation of the opening of his 1828 paper
Memoire sur une classe particuliere dequations resolubles algebrique-
ment, in uvres Completes, 2:478507 at 478. See also William Snow Burn-
side and Arthur William Panton, The Theory of Equations (London: Longmans,
Green, 1928), 282305.

French history during the time of Abel and Galois: For a superb overview,
see the classic work by Albert Guerard, France: A Modern History (Ann
Arbor: University of Michigan Press, 1959), 3 (Capetian dynasty), 278296
(the period 18141848), 288 (Lafayette).

Galoiss life: The most reliable recent study is Laura Toti Rigatelli, Evariste
Galois 18111832 (Boston: Birkhauser Verlag, 1996), which includes a helpful
Notes to pp. 108130 191

bibliography. Among earlier brief biographical works, Eric Temple Bell,

Men of Mathmatics (New York: Simon & Schuster, 1986), 362377, is widely
known, though it propagates many of the myths concerning Galois; for
poetic aspects of Abel and Galois, see also George Steiner, Grammars of Cre-
ation (New Haven: Yale University Press, 2001), 207212. There is a helpful
corrective in Tony Rothman, Genius and Biographers: The Fictionaliza-
tion of Evariste Galois, American Mathematical Monthly 89:2, 84102 (1982)
and The Short Life of Evariste Galois, Scientific American 246:4, 136149
(1982), also included in his Science a la Mode: Physical Fashions and Fictions
(Princeton: Princeton University Press, 1989). See also the helpful articles by
Rene Taton, Evariste Galois and His Contemporaries, Bulletin of the London
Mathematical Society 15, 107118 (1983), and Harold M. Edwards, A Note on
Galois Theory, Archive for the History of Exact Science 41, 163169 (1990). For
Galoiss final letter, see David Eugene Smith, A Source Book in Mathematics
(New York: Dover, 1959), 278285.

Raspail: For his account of Galois in prison, see Toti Rigatelli, Galois, 98
100; for his acquaintance with Abel, see Stubhaug, Abel, 410411, 416417
(Galoiss comment on Abels death).

8 Seeing Symmetries

My treatment of the relation between symmetries and geometric fig-

ures is inspired by Felix Kleins treatment of the symmetries of regular solids
as paradigms of the symmetries of equations, found in a brief form in his
classic lectures to schoolteachers, Elementary Mathematics from an Advanced
Standpoint, tr. E. R. Hedrick and C. A. Noble (New York: Dover, n.d.), 1:101
143, and very fully (though with notation that is quite hard to follow) in his
Lectures on the Icosahedron (New York: Dover, 1956), 320. I am not aware
of other treatments that use the analogy with dance or the visualizations
I present in the text and in box 8.3. I was greatly helped by Hartshornes
treatment of the symmetry groups of polyhedra in his Geometry: Euclid and
Beyond, 469480.
Basic treatments of Galois theory: My presentation attempts only to give
an intuitive version (though see appendix C and the notes below for a few
more details). There are many treatments; several stand out as especially
accessible. John E. Maxfield and Margaret W. Maxfield, Abstract Algebra and
Solution by Radicals (New York: Dover, 1992) is particularly suitable for self-
study. Saul Stahl, Introductory Modern Algebra: A Historical Approach (New
York: John Wiley, 1997) is an admirably written textbook that includes valu-
able excerpts from the writings of al-Khwarizm, Cardano, Abel, Galois,
192 Notes to pp. 111130

and Cayley (261287); I would also recommend Jean-Pierre Tignol, Galois

Theory of Algebraic Equations (London: Longman Scientific Technical, 1980).
Charles Robert Hadlock, Field Theory and Its Classical Problems (Mathemati-
cal Association of America, 1978), despite its austere title, is quite accessible,
though somewhat more abstract in its approach than the Maxfields book.
There is also a nice overview in I. M. Yaglom, Felix Klein and Sophus Lie:
Evolution of the Idea of Symmetry in the Nineteenth Century, tr. Sergei Sossinsky
(Boston: Birkhauser, 1988), 121. Lillian R. Lieber, Galois and the Theory of
Groups: A Bright Star in Mathesis (Lancaster, PA: Science Press, 1932), long
out of print, gives a lucid outline in sixty pages, expressed in a kind of verse
with drawings, a companion piece to her similarly delightful treatments of
non-Euclidean geometry and general relativity.

Other treatments of Galois theory: A full listing of the vast number of treat-
ments is scarcely possible here, but I would like to mention some that I
found helpful. D. E. Littlewood gives an interesting overview in The Skele-
ton Key of Mathematics: A Simple Account of Complex Algebraic Theories (New
York: Harper, 1960), 6576, as does Kline, Mathematical Thought, 752771.
Old textbooks sometimes present the theory less abstractly: See Leonard E.
Dickson, Modern Algebraic Theories (Chicago: Sanborn, 1926), 135250, Burn-
side and Panton, The Theory of Equations, 244305, and Edgar Dehn, Algebraic
Equations: An Introduction to the Theories of Lagrange and Galois (New York:
Dover, 1960). There is a nice account of the theory in an old French edition of
Galoiss works by G. Verriest, Evariste Galois et la Theorie des Equations
Algebriques (Paris: Gauthier-Villars, 1934), which is included in uvres
Mathematiques dEvariste Galois, ed. Emile Picard (Paris: Gauthier-Villars,
1897). A more recent French treatment is Claude Mutafian, Equations
Algebriques et Theorie de Galois (Saint-Amand-Montrond: Librarie Vuibert,
1980). German-speaking readers may find helpful N. Tschebotarow,
Grundzuge der Galoisschen Theorie, tr. H. Schwerdtfeger (Groningen:
P. Noordhoff, 1950), which is thorough and rich in examples. Toti Rigatelli,
Galois, 115138, gives a valuable outline of his work. For more modern treat-
ments that are friendly but still rigorous, see M. M. Postnikov, Fundamen-
tals of Galois Theory, tr. Leo F. Boron (Groningen: P. Noordhoff, 1962), Ian
Stewart, Galois Theory (London: Chapman and Hall, 1973), and R. Bruce
King, Beyond the Quartic Equation (Boston: Birkhauser, 1996), which is ori-
ented to the symmetry concerns of chemists. Many graduate and undergrad-
uate textbooks include Galois theory; see especially Nathan Jacobson, Basic
Algebra I, second edition (New York: W. H. Freeman, 1985), 210270, I. N.
Herstein, Topics in Algebra, second edition (New York: Wiley, 1975), 237259,
and Jerry Shurman, Geometry of the Quintic (New York: Wiley, 1997), which
concentrates on geometric aspects. For some helpful articles, see Raymond
G. Ayoub, On the Nonsolvability of the General Polynomial, American
Notes to pp. 111130 193

Mathematical Monthly 89, 397401 (1982), and John Stillwell, Galois Theory
for Beginners, American Mathematical Monthly 101, 2227 (1994), both of
which assume considerable familiarity with modern algebra. For the grad-
ual interpretation of Galoiss ideas, see Wussing, Group Concept, 118141, van
der Waerden, History of Algebra, 103116, B. Melvin Kiernan, The Develop-
ment of Galois Theory from Lagrange to Artin, Archive for the History of
Exact Sciences 8, 40154 (19711972), MT, 752771, and Yochi Hirano, Note
sur les diffusions de la theorie de Galois: Premiere clarification des idees de
Galois par Liouville, Historia Scientiarum 27, 2741 (1984).
Making models of Platonic solids: See David Mitchell, Mathematical
Origami: Geometrical Shapes by Paper Folding (Norfolk: Tarquin Publications,
1999).
Groups: For a helpful introduction that stresses the larger philosophical
significance of group theory, see Curtis Wilson, Groups, Rings, and
Lattices, St. Johns Review 35, 311 (1985). In particular, Wilson notes that
the most general or universal aim of intellectual work is the discovery of
invariants (6).
Regular solids and their groups: For a classic work that includes a sketch
of the symmetry argument for the uniqueness of the five Platonic solids,
see Hermann Weyl, Symmetry (Princeton: Princeton University Press, 1980),
149156.

Visualization of groups: Two books offer many ways to grasp group sym-
metries visually: Israel Grossman and Wilhelm Magnus, Groups and Their
Graphs (Washington, D.C.: Mathematical Association of America, 1964), and
R. P. Burn, Groups: A Path to Geometry (Cambridge: Cambridge University
Press, 1985). These visualizations go back to Klein, Lectures on the Icosahedron,
also treated by W. Burnside, Theory of Groups of Finite Order, second edition
(Cambridge: Cambridge University Press, 1911), 402408.
Normal subgroups: A subgroup H of a group G is called a normal or invariant
subgroup of G if for every element g in G and h in H, there is some h  such
that h  g = g h, where h  is not necessarily the same as h. Note that
if we multiply both sides of this equation on the right by g 1 , the inverse
of the element g, we find then h  g g 1 = h  = g h g 1 (since by
definition g g 1 = I ). All possible g h g 1 are called the conjugates
or the equivalence class of h. A normal subgroup need not be abelian but
is always self-conjugate: if it contains an element h, it also contains all the
conjugates of h. This is the precise meaning of elements of the same kind
on pp. 119, 122, 123124.
194 Notes to pp. 111130

Here is a more intuitive way of thinking about normal subgroups. Picture

the elements h of the subgroup H as forming an orbit, consisting of its
possible values. Then the combined operation g h g 1 tries to tip H out
of its orbit by imposing an external action g, balanced by an equal and
opposite counteraction g 1 . If g h g 1 always remains within its original
orbit, for all actions g, the orbit is undisturbed and the subgroup is normal.
Also, if a is a rotation around a vertex A, then b = g a g 1 means that b
is the same kind of rotation around the vertex B = g(A). (See Hartshorne,
Geometry: Euclid and Beyond, 474.) Specifically, g 1 takes the vertex B back
to A, then a performs the rotation, and g takes the vertex A back to B. Thus,
normal subgroups (which obey this condition) signify all the rotations of the
same kind around all allowable vertices.
Jordan and Holder: See Camille Jordan, Traite des Substitutions (Paris:
Gauthier-Villars, 1957), 286291 (abelian equations), 370397 (composition
series); and Wussing, Group Concept, 135 ff.
Solvable chains: Galois theory begins by determining a given equations
group of permutations. In practice, this can involve complex calculations.
Given such a Galois group, we can determine in sequence a maximal
normal subgroup (a normal subgroup of largest order less than that of the
group), a maximal normal sub-subgroup, and so on until we reach a sub-sub-
subgroup that includes only the identity element, which is automatically a
(normal) subgroup of any group, by definition. This chain can be written
symbolically as G  G 1  G 2   G n1  G n  I, where G i  G i+1 means:
G i contains the normal subgroup G i+1 , though G i+1 may not be normal
in the whole group G.

Then the whole group G is called solvable if and only if the quotient group
of each successive group in the chain, G i /G i+1 , is abelian. That is, if H is
a normal subgroup of G, then the quotient group G/H is defined to be the
set of all cosets of H in G, namely all the sets {g1 h 1 , g2 h 1 , . . .}, {g1
h 2 , g2 h 2 , . . .}, . . . , made from all the elements g1 , g2 , . . . of G and h 1 , h 2 , . . .
of H. See appendix C, p. 176, for the example of clock arithmetic illustrating
cosets and quotient groups. If the quotient group G i /G i+1 is nonabelian, the
chain is broken, and the equation is not solvable in radicals.
Note also that if a finite group G has a prime number of elements, p, then
it is abelian. Proof: Let g be an element of G other than the identity (since
p > 1). Then consider H = {I, g, g 2 , . . .}, which is a subgroup of order n,
where n > 1. By Lagranges Theorem (see appendix C), n divides p. But p
is prime; it has no divisors besides itself and 1. Then if n
= 1, n must equal
p and H is of the same order as G, so that H = G. G is abelian since it is
cyclic, meaning that it is composed of powers of one element, g, namely
Notes to pp. 130137 195

{I, g, g 2 , . . .}. Since g m g n = g n g m , any cyclic group is abelian, and so too

is G. Note that not every abelian group is cyclic; for instance, the four-group
V (table 8.3) is abelian but not cyclic.

Monster group: See Stahl, Introductory Modern Algebra, 247248; for a tech-
nical overview of the modern study of simple groups, see Ron Solomon,
On Finite Simple Groups and Their Classification, Notices of the Ameri-
can Mathmatical Society, 42:2, 231239 (1995). For a popular account, see W.
Wayt Gibbs, Monstrous Moonshine Is True, Scientific American, 279:5, 40
41 (1998).

9 The Order of Things

History of noncommutativity: For an excellent selection of original writ-

ings with commentary, see Ewald, From Kant to Hilbert, 1:293296 (Gauss
on commutative law), 314321 (Peacock and Cauchy), 321330 (Duncan
Gregory), 331361 (De Morgan), 362441 (Hamilton), 442509 (Boole), 510
522 (Sylvester), 542573 (Cayley), 362 (B. Peirce and his 162 algebras), 574
648 (C. S. Peirce). For a helpful overview of these figures, see HM, 575582,
Bourbaki, Elements of History of Mathematics, 117123, and Yaglom, Klein and
Lie, 7194.
Grassmann: For a translation of part of his Ausdehnungslehre (1862), see
Desmond Fearnly-Sander, Grassmanns Theory of Dimension, in First
Australian Conference on the History of Mathematics, 5282; in the same vol-
ume, Fearnly-Sander also has a helpful paper on Hermann Grassmann and
the Prehistory of Universal Algebra, 4151, which includes the statement
quoted from Grassmann about the two possibilities for products of two fac-
tors, commuting and anticommuting (49).

Hamilton and the quintic: Thomas L. Hankins, Sir William Rowan Hamilton
(Baltimore: Johns Hopkins University Press, 1980), 276279, 248252
(Hamilton and Peacock), 277278 (Jerrards faulty solution). For Hamiltons
long account of Abels argument, On the Argument of Abel, Respecting the
Impossibility of Expressing a Root of Any General Equation above the
Fourth Degree, by any Finite Combination of Radicals and Rational Func-
tions (1839), see The Mathematical Papers of Sir William Rowan Hamilton
(Cambridge: Cambridge University Press, 1967), 3:517569; for Hamiltons
Inquiry into the Validity of a Method Recently Proposed by George
B. Jerrard (1837), see 3:481516.
Jerrard: See George B. Jerrard, Mathematical Researches (London: Longmans,
1834).
196 Notes to pp. 138143

Bolyai on quintics: See Yaglom, Klein and Lie, 59.

Hamilton on quaternions: HM, 582584 and Hankins, Hamilton, 283325,

247 (parents), who notes that possibly Hamilton had been prepared for the
rejection of the commutative law by discussions with Eisenstein, or possibly
his attempts to find a geometrical representation of triplets had shown him
that rotations in three-dimensional space do not commute either, or, what
is even more likely, sacrificing commutativity seemed to him to be the only
way to get any results at all (300).
Felix Klein: For a good overview of his work, especially the Erlangen pro-
gram, see Yaglom, Klein and Lie, 111137. His famous 1872 Erlangen lecture
is translated as A Comparative Review of Recent Researches in Geometry,
Bulletin of the American Mathematical Society, 2 [series 1], 215249 (1893).
See also his 1895 Gottingen 1ecture The Arithmetizing of Mathematics,
Bulletin of the American Mathematical Society, 2 [series 1], 241249 (1896).

Duplication of the cube, trisection of angles, squaring circles: See Klein,

Famous Problems of Elementary Geometry, 523.
Galois theory as relativity: See Hermann Weyl, Philosophy of Mathematics
and Natural Science (Princeton: Princeton University Press, 1949), 74 (Galois
theory and relativity) and his Symmetry, 137137 (Galoiss letter).
Quantum theory and noncommutativity: See Hermann Weyl, Symmetry,
133135, and The Theory of Groups and Quantum Mechanics (New York: Dover,
1950), 9498. For Max Plancks view of irreversibility, see his Eight Lectures on
Theoretical Physics, ed. Peter Pesic (Mineola, NY: Dover, 1998), viixiii, 120.
I discuss this question further in my book Seeing Double: Shared Identities
in Physics, Philosophy, and Literature (Cambridge, MA: MIT Press, 2002),
101120.
Causality and noncommutativity: In order to obey the principle that no
causal influence can travel faster than the speed of light, relativity theory
imposes the constraint that quantum fields must commute when the points
in question are separated by a spacelike interval, which requires greater-
than-light-speed travel between them.
Nonabelian gauge theories: See Michio Kaku, Quantum Field Theory: A Mod-
ern Introduction (New York: Oxford University Press, 1993), 5054 (com-
mutation relations of the Lorentz group), 295406 (gauge theories and the
standard model). For a popular introduction, see Brian Greene, The Elegant
Universe (New York: W. W. Norton, 1999).
Notes to pp. 143147 197

Noncommutative geometry: See the authoritative (and highly technical) de-

scription in Alain Connes, Non-Commutative Geometry (San Diego: Academic
Press, 1994). See also Pierre Cartier, A Mad Days Work: From Grothendieck
to Connes and Kontsevich: The Evolution of Concepts of Space and Symme-
try, Bulletin of the American Mathematical Society, 38, 389408 (2001), which
ends with speculations connecting the cosmic Galois group with the group
of general relativity and the fine structure constant, = e 2 /hc 1/137,
which sets the scale for electromagnetic and quantum interactions.

Groups and space-time: I have discussed this matter in my paper Eu-

clidean Hyperspace and Its Physical Significance, Nuovo Cimento 108B,
11451153 (1993).

10 Solving the Unsolvable

De Moivres formula: See Smith, Source Book in Mathematics, 440454, includ-

ing Eulers later statements of the formula. See also Eli Maor, Trigonometric
Delights (Princeton: Princeton University Press, 1998).
Solvability in the problems of David Hilbert and Steve Smale: See Jeremy
Gray, The Hilbert Challenge (Oxford: Oxford University Press, 2000), which in-
cludes the text of Hilberts 1900 address (240282), and Steve Smale, Mathe-
matical Problems for the Next Century, Mathematical Intelligencer 20:2, 715
(1998).
Godel: For a thoughtful commentary on Godel that raises the possible anal-
ogy with Abel and Ruffini, see S. G. Shanker, Wittgensteins Remarks on
the Significance of Godels Theorem, in Godels Theorem in Focus, ed. S. G.
Shanker (London: Routledge, 1988), 155256 at 161168. Less well known
than Godels 1931 paper but of great interest is the 1936 paper by Gerhard
Gentzen, The Consistency of the Simple Theory of Types, in The Collected
Papers of Gerhard Gentzen, ed. M. E. Szabo (Amsterdam: North-Holland,
1969), 214222, which established that arithmetic is consistent if transfinite
induction is allowed.
Ultraradical numbers: This is the term used by Stewart, Galois Theory, 148.
D. Morduhai-Boltovski calls hypertranscendental numbers those that are
the solutions of ordinary differential equations with constant integral co-
efficients, as opposed to simply transcendental numbers, which are not
the solution of any finite ordinary polynomial equation. See A. O. Gelfond,
Transcendental and Algebraic Numbers, tr. Leo F. Boron (New York: Dover,
1960), 96. However, the term ultraradical refers only to algebraic numbers not
expressible in radicals.
198 Notes to pp. 146147

Solution of the quintic: Today students have powerful, easily available com-
puter software, such as MathematicaTM , that is capable of dealing with solu-
tions to the quintic. See the helpful website at http://library.wolfram.com/
examples/quintic/, which is available also as a wonderful poster with all
kinds of information about the history and solution of quintics. Progress con-
tinues to be made even without computers. For instance, Blair K. Spearman
and Kenneth S. Williams were able to give a simple criterion in their paper
Characterization of Solvable Quintics x 5 + a x + b, American Mathematical
Monthly 101, 986992 (1994).

Computer solutions of equations: See Herman H. Goldstine, The Computer

from Pascal to von Neumann (Princeton: Princeton University Press, 1972), 106
(Torres Quevedo).

Torres Quevedo: See Francisco Gonzalez de Posada, Leonardo Torres

Quevedo, Investigacion y Ciencia, 166:7, 8087 (1990) and his book Leonardo
Torres Quevedo (Madrid: Fundacion Santillana, 1985).
Modular and generalized hypergeometric functions: Abel and Jacobi began
to study what they called theta functions in 1827; by 1858, Charles Hermite,
Leopold Kronecker, and Francesco Brioschi independently showed that any
quintic could be solved by elliptic modular functions derived from these
which
theta functions,

are infinite sums composed of powers times cosines.
The notation x means the sum of the terms xn from n = 0 to n = ,
n=0 n
namely x0 + x1 + x2 + . The two theta functions are defined as



q (n+1/2) cos((2n + 1)z),
2
2 (z, q ) = 2
n=0



2
3 (z, q ) = 1 + 2 q n cos(2nz).
n=1

The elliptic modular function (z) is defined in terms of these two theta
functions:

2 (0, z)4
(z) = .
3

3 (0, z)4
This function remains of central interest in contemporary mathematics. The
solutions of any polynomial equation can also be expressed in terms of
the generalized hypergeometric function, which is a quotient of general
p
products of series of powers. Here the notation for products is used: n=1 xn
Notes to pp. 150151 199

means the product of the terms xn from n = 1 to n = p, namely x1 x2

x3 x p . The generalized hypergeometric function can be expressed in
this very compressed notation:
p
 (a i )k zk
p Fq (a 1 , . . . , a p : b 1 , . . . , b q : z) = i=1
q ,
i=1
(b i )k k!
k=0

where the Pochhammer symbol (a i )k is defined by


k
(a i )k = (a i + j 1).
j=1

All the elementary functions (such as the trigonometric functions) can be de-
fined in terms of these extremely general functions; there are brief summaries
at http://library.wolfram.com/examples/quintic/hypergeo.html and
http://library.wolfram.com/examples/quintic/theta.html.
Transcendentality of e and : See the superb account of the proofs in Klein,
Famous Problems of Elementary Geometry, 6177. For a general history, see Petr
Beckmann, A History of , second edition (Boulder, CO; Golem Press, 1971);
for a valuable collection of original papers, see Pi: A Source Book, ed. Lennart
Berggren, Jonathan Borwein, and Peter Borwein (New York: Springer, 1997),
which includes Lindemanns original proof (194229), and Ivan Niven, A
Simple Proof That is Irrational, 509.
Cantors proof: For a translation of his crucial paper, see Ewald, From Kant
to Hilbert, 2:838940. There is a nice brief account in Klein, Famous Problems of
Elementary Geometry, 4955. Dunham gives a helpful and accessible account
of Cantors work in his Journey through Genius, 245283. For a detailed study
of the development of Cantors ideas, see Joseph Warren Dauben, Georg
Cantor: His Mathematics and Philosophy of the Infinite (Princeton: Princeton
University Press, 1990).
Concepts of infinity: For Richard Dedekinds introduction of the infinite as a
fundamental notion of set theory, see his Essays on the Theory of Numbers (New
York: Dover, 1963), 6364. For a general overview, see Eli Maor, To Infinity
and Beyond: A Cultural History of the Infinite (Boston: Birkhauser, 1987). For
the connection with art, see J. V. Field, The Invention of Infinity: Mathematics
and Art in the Renaissance (Oxford: Oxford University Press, 1997).
Beauty in mathematics: For one possible realization of classic versus
romantic mathematical styles, see Francois Le Lionnais, Beauty in
Mathematics, in his collection Great Currents of Mathematical Thought (New
York: Dover, 1971), 2:121158.
200 Notes to pp. 151170

Kant: For his account of the mathematical sublime, see Immanuel Kant,
Critique of the Power of Judgement, tr. Paul Guyer and Eric Matthews (Cam-
bridge: Cambridge University Press, 2000), 131149.

Abels work on topics besides solvability of equations: See Michael

Rosen, Abels Theorem on the Lemniscate, American Mathematical Monthly
88, 387394 (1981), and (concerning Abels contributions on elliptic inte-
grals) Roger Cooke, Abels Theorem, in The History of Modern Mathematics,
ed. David E. Rowe and John McCleary (New York: Academic Press, 1989),
389421. For Abels summary comments on the lemniscate, see his 1826 letter
to Holmboe, cited in uvres Completes de Niels Henrik Abel, 2:261262.

Abels favorite subject: See his 1826 letter to Holmboe, cited in uvres
Completes de Niels Henrik Abel, 2:260.

Abels credo: My translation of part of his introduction to Sur la Resolution

Algebrique des Equations, uvres Completes de Niels Henrik Abel, 2:217243
at 217; the editors comment (329338) that this paper was first published by
Holmboe in 1839 and also give interesting alternative readings of Abels
text.

Abels Paris notebooks: Stubhaug, Abel, 504505, reproduces and translates

this page (504), along with another from this notebook. Though apparantly
unmoved by music, Abel was very fond of the theater. During his travels,
he may well have seen Beaumarchaiss popular play Le Mariage de Figaro,
which contains a virtuoso scene (III.v) in which Figaro plays with the English
expression God-dam (as Abel also spells it).

Appendix A Abels 1824 Paper

This is Abels original version of his proof, in my translation. The origi-

nal text is in his uvres, 1:2833. The only other published translation, in
Smith, Source Book in Mathematics, 261266, unfortunately has several signif-
icant misprints. My notes try to explain Abels brief indications, which he
expanded in his 1826 version of the proof, Demonstration de limpossibilite
de la resolution algebrique des equations generales qui passent le quatrieme
degre, uvres, 1:6694.

Appendix B Abel on the General Form of an

Algebraic Solution

My account is drawn from Abels Demonstration de limpossibilite, 6672.

Notes to pp. 171180 201

Appendix C Cauchys Theorem on Permutations

The original source is Cauchys Memoire sur le nombre des valeurs quune
fonction peut acquerir (1815), in uvres Complete dAugustin Cauchy (Paris:
Gauthier-Villars, 1905), series II, 1:6290. My account follows closely Abels
account in his Demonstration de limpossibilite, 7579.
This page intentionally left blank
 Acknowledgments

To Larry Cohen and his associates at the MIT Press for their
support and collaboration in bringing this book to life.
To St. Johns College for released time under the Louise
Trigg tutorship, and to my fellow students who encouraged
my struggle to understand Abels proof.
To Raymond Ayoub, David Cox, David Derbes, William
Dunham, Robin Hartshorne, Barry Mazur, Mark Peterson,
Michael Rosen, Tony Rothman, Jerry Shurman, and Curtis
Wilson, whose comments and criticisms saved me from many
mistakes. The errors that remain are my own.
And to Ssu, Andrei, and Alexei, who solved the unsolvable
for me.
This page intentionally left blank
 Index

Abbati, Pietro, 82 symbolic notation, 4045, 132

Abel, Niels Henrik, 13, 85102, unknown, 43
189n190n variable, 1, 44
Abelian Addition Theorem, Algebraic functions, 9091
151, 200n al-Khwarizm, Muhammed
abelian equations, 98100 ibn-Musa, 2526, 30, 183n184n
abelian groups, 112113 Alogon, 9
Abelian integrals, 151, 200n Alternating groups. See Groups
Abel-Ruffini Theorem, 13, 8994, Analytic mathematics, 42, 59
155170, 200n Anrta, 9
and Cauchy, 87, 9394, 96 Anticommutation, 134136, 195n
early life, 8589, 189n Apollonius of Perga, 42, 51, 59
formulas of, 88 Aporia, 14, 182n
and Galois, 105, 108109, Archimedes, 32, 60
130131, 145 Area problem, 6166
and Gauss, 8889, 95, 151 Aristotle, 7, 181n
and Hamilton, 133 Arithmos, 9, 181n
illness and death of, 101102, 190n Athens, 15
notebooks of, 97, 152153, 200n Ausdehnungslehre (Grassmann),
and Ruffini, 8789, 97, 190n 135136
travels in Europe, 9597 Ayoub, Raymond, 189n, 192n193n
Academie des Sciences, 96, 104106
Accounting. See Bookkeeping Babylonian mathematics, 5, 7,
Algebra 2425, 30, 183n
Arabic, 2328, 45, 54 Bacon, Francis, 182n183n
coefficient, 1, 4445, 9193 Basham, A. L., 181n
noncommutative, 131143 Beaumarchais, Pierre Augustin
roots, 1, 92, 98 de, 200n
206 Index

Bell, Eric Temple, 191n Cervantes, Miguel de, 23

Bernoulli, Daniel, 65 Charles X, 105
Bolyai, Janos, 133, 196n Chatfield, Michael, 185n
Bombelli, Raphael, 5455, 187n Chinese mathematics, 183n
Bookkeeping, double-entry, 2729, Christiania. See Oslo
184n185n Cipher, 2728, 33, 42
Boole, George, 132133, 195n Circle, 62
Boolean algebra, 132133 Code. See Cipher
Bourbaki, Nicolas (pseudonym), Coleridge, Samuel Taylor, 135
188n, 195n Commensurable, 78, 16
Boyer, Carl, 182n195n Commercial arithmetic, 2731
Bring, E. S., 67, 188n Commutativity, 99, 112
Brioschi, Fernando, 146, 198n Completing the cube, 3637, 120
Brizio, Anna Maria, 184n Completing the square,
Brown, R. G., 184n 2526, 113
Bryce, R. A., 189n Computers, 147, 197n198n
Bhler, W. K., 189n Conic sections, 42
Bulletin (Baron de Ferussac), 96 Connes, Alain, 197n
Brgi, Jost, 4849 Continuum, 11, 18
Burkert, Walter, 1981n Cooke, Roger, 200n
Burn, R. P., 193n Cosa (coss), 27, 44
Burnside, William Snow, 190n, Crelle, August, 9597, 100, 102
192n193n Crelles Journal, 95, 101
Cross product. See Multiplication
Cajori, Florian, 186n Cube, 56
Calculators, 78, 149150 symmetries of, 121
Calculus, 60, 6263 Cyclic groups and symmetries, 113,
Cantor, Georg, 150, 199n 117, 122, 195n
Cardano, Girolamo, 3040, 54, 57,
67, 69, 185n dAlembert, Jean Le Rond, 68
Cartier, Pierre, 197n Dance, 111130
Cauchy, Augustin, 83, 87, 96, Dauben, Joseph Warren, 199n
104105 Dedekind, Richard, 183n, 199n
Cauchys theorem, 93, 163, 166, Dehn, Edgar, 192n
175180, 201n del Ferro, Scipione, 3234
and commutativity, 132, 195n del FerroCardanoTartaglia
Causality and noncommutativity, method, 3235, 4849, 5455,
142, 196n 77, 174
Cayley, Arthur, 112, 126, De Moivre, Abraham, 149, 197n
136138, 195n DeMorgan, Augustus, 132, 195n
Cayley numbers, 137 Descartes, Rene, 5059, 68, 187n
Cayley tables, 112, 119, 122 and conic sections, 64
Index 207

Descartess rule of signs, 53 Euler, Leonhard, 62, 68, 90,

La Geometrie, 5058 149, 196n
relativity of roots, 57, 140 Ewald, William B., 187n, 195n, 199n
Dickson, Leonard E., 192n
Dimension (algebraic), 5051 Fauvel, John, 184n185n
Dirac, Paul, 141 Fearnly-Sander, Desmond, 195n
Disquisitiones Arithmeticae (Gauss), Fermats Last Theorem, 8788,
79, 189n 189n190n
Dodecahedron, 56 Ferrari, Ludovico (Luigi), 3435,
symmetries of, 124125, 126130 3739, 57, 69, 76, 122
Don Quixote (Cervantes), 23 Fibonacci. See Leonardo of Pisa
Dorrie, Heinrich, 188n190n Field, J. V., 186n, 199n
Dunham, William, 185n, 199n Fields (mathematics), 139
Duplication of cube, 196n Fields (physics), quantum theory
of, 142143
e, 28, 150, 184n, 199n Fine, Benjamin, 188n
Ecole Polytechnique, 105 Fine structure constant, 197n
Ecole Preparatoire (Ecole Normale Fontana, Niccolo. See Tartaglia
Superieure), 105 Fractions, 7
Edwards, Harold M., 191n France, 45, 9697, 102108, 190n
Einstein, Albert, 140, 143 Fundamental Theorem of Algebra,
Eisenstein, E. L., 184n 56, 6873, 79, 146, 188n
Elements. See Euclid
Equations, algebraic Galilei, Galileo, 4950, 187n
approximate solutions, 66, Galois, Evariste, 102109,
147, 198n 190n191n
cubic, 3, 28, 3037, 90, 113120, and Abel, 105106, 108109,
148149, 185n 130131, 145
general formulation, 13 and Cauchy, 104105
quadratic, 2, 23, 2526, 64, 9091, death of, 106108
111113, 185n education of, 102106, 190n
quartic, 2, 35, 3839, 7678, and his father, 104105
120122 Galois theory, 125130, 191n193n
quintic, 23, 7778, 9199, legend of, 108, 191n
122129, 198n posthumous writings of, 108
roots, 1 and Societe des Amis du Peuple,
Erlangen Program (Felix Klein), 106107
138140, 196n and Stephanie Poterin-
Euclid, 5, 1723, 42, 59, 145146, Dumotel, 106
150, 183n Garding, Lars, 190n
Euclidian geometry, 139 Gauge fields, nonabelian,
Eudoxus, 1718, 183m 142143, 196n
208 Index

Gauss, Carl Friedrich, 7074, 97, cyclical, 113, 117, 122,

187n189n 175180, 195n
and Abel, 89, 95, 100, 151 definition of, 125126
and commutativity, 131132, 195n identity, 112, 119, 125
and unsolvability of quintic, 79, 88 invariant subgroups, 119, 129
Gazale, Midhat, 183n Lagranges Theorem, 128, 175176
Gelfond, A. O., 197n Lorentz, 196n
Gentzen, Gerhard, 197n monster group, 130, 195n
Geometrie, La (Descartes), nonabelian, 118, 129, 142143
5054, 187n normal subgroups, 119, 129,
Geometry, 50, 60, 66 193n195n
Germain, Sophie, 104 order, 176
Gibbs, Josiah Willard, 136 and permutations, 175180
Gibbs, W. Wayt, 195n philosophical aspects, 193n
Gies, J. and F., 184n quotient, 130, 176177, 193n194n
Girard, Albert, 51, 56, 68, 187n S2 , 112113
Girards identities, 61, 92 S3 , 113120, 139
Gleason, Andrew, 187n S4 , 120122, 139
God, 49, 55 S5 , 122124
Godel, Kurt, 197n simple groups, 130
Golden ratio, 28 solvable chains of, 130, 194n195n
Goldstine, Herman H., 198n V, 122
Gonzalez de Posada, visualization of, 193n
Francisco, 198n Guerard, Albert, 190n
Gorman, Peter, 181n
Grafton, Anthony, 185n Hadlock, Charles Robert, 192n
Grassmann, Hermann, Hamilton, William Rowan,
135136, 195n 133136, 196n
Gray, Jeremy, 184n185n, 197n Hankins, Thomas L., 196n
Great Art (Cardano), 3040, 185n Harmony of the World (Kepler), 48,
Greek mathematics, 521 121, 124, 186n
Greene, Brian, 196n Hartshorne, Robin, 181n,
Gregory, Duncan, 132, 195n 191n, 194n
Grossmann, Israel, 193n Heath, Thomas, 183n
Groups, 109, 111130, 138140, Heisenberg uncertainty
193n195n principle, 141
A3 , 118120 Hellman, Morton J., 185n
A4 , 121122 Henry IV, 45
A5 , 123129, 139 Heptagon, 48, 186n187n
abelian, 112113, 129 Hermite, Charles, 146, 150, 198n
continuous, 140 Herrstein, I. N., 192n
cosets, 176 Hexagon, 48
Index 209

Hilbert, David, 197n Kemp, Christine, 96, 101102

Hippias of Mesopontum, 10 Kepler, Johannes, 4849, 121, 124,
Hirano, Yochi, 193n 186n187n
Hoe, J., 183n Khayyam, Omar, 30, 184n
Holder, Otto, 130131, 177, 194n Kiernan, B. Melvin, 193n
Holmboe, Berndt Michael, 87, 97, King, R. Bruce, 192n
190n, 200n Klein, Felix, 138140, 143, 189n,
Holy Spirit, 55 191n, 196n, 199n
Huffman, C. A., 181n Klein, Jacob, 182n, 187n
Huntley, H. E., 181n Kline, Morris, 182n, 192n
Hypergeometric functions, Knorr, Wilbur Richard, 182n
198n199n Kronecker, Leopold, 146, 190n, 198n

Icosahedron, 56 Lafayette, General, 105

symmetries of, 123129 La Geometrie (Descartes), 5058
Incommensurability, 714 Lagrange, Joseph-Louis, 7383,
Indian mathematics, 9 87, 188n
Indistinguishability of quanta, 142 Lagrange resolvent, 7479
Infinity, 22, 146, 148, 151, 153 Lagranges Theorem, 128,
Institut de France, 100, 106 175176, 194n
Invariance, 113, 139 Lalanne, Leon, 147
Invariant subgroups. See Groups La Nave, Federica, 187n
Irrational magnitudes, 714, 1921, Laplace, Pierre Simon, 51, 80
145146, 183n Legendre, Adrien-Marie, 96,
Irreducible case (cubic 100101
equations), 54 Leibniz, Gottfried Wilhelm, 55,
Irreversibility, 141 6567, 183n, 187n188n
Isograph, 147 Le Lionnais, Francois, 199n
Le Mariage de Figaro
Jacobi, Carl Gustav Jacob, 100, 146 (Beaumarchais), 200n
Jacobson, Nathan, 192n Lemniscate, 65, 152153, 200n
Jerrard, George B., 67, 133, Leonardo da Vinci, 6, 28, 184n
188n, 195n Leonardo of Pisa (Fibonacci), 2728,
Johnston, K. S., 184n 30, 184n
Jordan, Camille, 130131, 133, 146, Lieber, Lillian R., 192n
177, 194n Lindemann, Ferdinand, 150, 199n
July monarchy, 105106 Liouville, Joseph, 133
Littlewood, D. E., 192n
Kabbalists, 48 Locus problem, 57, 59
Kaku, Michio, 196n Logos, 9
Kant, Immanuel, 200n Louis XVI, 104
Karl XIII, 85 Louis XVIII, 104105
210 Index

Louis-Philippe I, 105106 lemma 28, 6166, 148

Lycee Louis-le-Grand, 104 Newtons identities, 6061
Newtons method, 66
Macve, Richard, 184n Newtonian dynamics, 136, 141
Magnitudes, 78, 23 Niven, Ivan, 199n
Magnus, Wilhelm, 193n Nonabelian gauge fields,
Malfatti, Gianfrancesco, 77, 82 142143, 196n
Maor, Eli, 184n, 197n, 199n Nonabelian groups. See Groups
Marinoni, Augusto, 184n Noncommutative geometry,
MathematicaTM , 198n 143, 197n
Matrix, 136138 Noncommutativity, 99100,
Maxfield, John E. and Margaret W., 131143, 195n
191n192n Normal subgroups. See Groups
Maxwell, James Clerk, 135136 Norway, 85
Maxwellian dynamics, 141 Numbers
Mayer, Uwe F., 188n algebraic, 146, 150, 197n
Mazur, Barry, 187n complex and imaginary, 5456, 70,
Meno, 1314, 182n 148149, 187n
Mercantile Arithmetic (Widman), 29 counting, 9
Merzbach, Uta C., 182n195n in Greek mathematics, 9
Minkowski, Hermann, 140 irrational magnitudes, 78, 1819,
Mitchell, David, 193n 23, 146
Modular functions, 198n line, 51
Monster. See Groups negative, 5154, 187n
Montucla, Jean Etienne, 79, 189n octonions (Cayley numbers), 137
Morduhai-Boltovsky, D., 197n place value, 24
Multiplication quaternions, 134135, 196n
commutativity of, 131132 rational, 78, 146
Grassmann algebra, 135136 sexagesimal, 2425
matrix, 136138 transcendental, 62, 66, 150,
quaternion, 134 197n, 199n
scalar product, 135 ultraradical, 146, 150, 197n
vector product, 135
Music, 7, 1920, 48, 183n Octahedron, 56
Mutafian, Claude, 192n symmetries of, 121
Octonions. See numbers
Nahin, Paul J., 187n Ore, ystein, 185n, 189n
Napoleon, 104, 108 Oslo, 87, 101
Needham, Joseph, 183n Oval, 6166
Newton, Isaac, 5966, 149150,
187n188n Pacioli, Luca, 6, 2830, 184n185n
and Descartes, 59, 66 Panton, Arthur William, 190n, 192n
Index 211

Pappus, 1011, 42, 57, 182n and Galois theory, 140, 196n197n
Parabola, 65 general, 143, 196n
Parshall, Karen Hunger, 184n185n of roots, 57, 140
Pascal, Blaise, 147 special, 140
Peacock, George, 132, 195n of space-time, 140
Pentagon, 49 Republic (Plato), 15, 182n
Permutations, 7577, 82, 108109, Resolvent, see Lagrange resolvent
111130, 175180 Richard, Louis-Paul-Emile, 104105
Pesic, Peter, 142, 182n183n, 186n, Roman law, 43
188n, 196n197n Roots of unity, 74, 97
Peterson, Mark, 185n Rosen, Michael, 190n, 200n
Pi ( ), 62, 150, 199n Rosenberger, Gerhard, 188n
Pierce, Benjamin, 138, 195n Rothman, Tony, 191n
Pierce, C. S., 138, 195n Royal Fredericks University,
Piero della Francesca, 28, 30, 185n Christiania (Oslo), 87
Pierpont, J., 190n Rta, 9, 181n
Planck, Max, 141, 196n Ruffini, Paolo, 8083
Plato, 1117, 44, 140141, 182n
Platonic solids, 56, 122, 138, Sacrifice, 10, 46
143, 193n Saigey, Jaques Frederic, 96
Poisson, SimeonDenis, 101 Scalars, 135
Postnikov, M. M., 192n Second Law of
Poterin-Dumotel, Stephanie, 106 Thermodynamics, 141
Pourciau, Bruce, 188n Seventeen-sided polygon, 70,
Principia (Newton), 5966, 187n 74, 189n
Pycior, Helena M., 186n187n Shanker, S. G., 197n
Pythagoras, 511, 46, 181n Shurman, Jerry, 192n
Pythagorean theorem, 11 Shylock, 27, 184n
Pythagoreans, 511, 15 Singh, Simon, 190n
Skau, Christian, 190n
Quantum theory, 141143, 196 Smale, Steve, 197n
Quaternions. See numbers Societe des Amis du Peuple,
106107
Radicals, 2, 35 Socrates, 1317, 182n
Ralph, Leslie, 181n Solomon, Ron, 195n
Rashed, Roshdi, 184n Solution in radicals, 2
Raspail, Francois-Vincent, 97, 106, Space
108, 191n four-dimensional 135, 197n
Rational magnitudes, 9 n-dimensional, 135136, 138
Ratios, 7 three-dimensional, 139141, 143
Reductio ad absurdum, 78, 64, 90 Spearman, Blair K., 198n
Relativity Species, logic of, 44, 132
212 Index

Speed of light, 140, 142, 196n Third Law of Planetary Motion

Square, 714 (Kepler), 49
Square roots, sound of, 20 Tignol, JeanPierre, 192n
Squaring the circle, 150, 196n Time, irreversibility of, 141
Stahl, Saul, 191n, 195n Topology, 72
Standard theory (physics), Torres Quevedo, Leonardo,
142, 196n 147, 198n
Stein, Howard, 183n Torture, 1617, 182n183n
Steiner, George, 191n Toti Rigatelli, Laura, 190n192n
Stewart, Ian, 192n, 197n Transcendental. See Number
Stillwell, John, 193n Treviso arithmetic, 29
Stubhaug, Arild, 189n191n, 200n Trial, 18
Subgroups. See Groups Triangle, symmetries of, 113120
Suleiman II, 153 Trigonometry, 47, 149150, 186n
Summary of Arithmetic (Pacioli), Trisection of angle, 187n, 196n
2829 Tschebotarow, N., 192n
Swetz, Frank, 185n
Sylvester, James Joseph, Universal Arithmetic (Newton), 60
136138, 195n Uspensky, J. V., 185n, 188n
Symmetric groups. See Groups
Symmetry, 113 Vandermonde, Alexandre-
in algebraic expressions, 60 Theophile, 75, 188n
of fundamental particles, van der Waerden, B. L., 183n184n,
142143 188n, 193n
of polyhedra (see Triangle; van Roomen, Adriaan, 4547, 186n
Tetrahedron; Cube; Vectors, 135
Dodecahedron; Icosahedron; Vernier, Hippolyte Jean, 104
Octahedron) Verriest, G., 192n
of three-dimensional space, 124 Vetter, Guido, 186n
Synthetic mathematics, 4243 Vite, Francois, 4147, 5658, 73,
132, 151, 186n
Tartaglia, 3234, 185n Voltaire, Francois Marie
Taton, Rene, 191n Arouet de, 85
Taylor, R. Emmett, 184n von Fritz, Kurt, 182n
Tetractys, 10 von Tschirnhaus, Count Ehrenfried
Tetrahedron, 56 Walter, 6668, 188n
symmetries of, 120121 Tschirnhaus transformation, 68
Theaetetus, 1517, 20, 182n
Theodorus, 17, 182n Walker, D. P., 186n
Theology, Christian, 55 Wallis, John, 186n
Thermodynamics, 141 Weil, Simone, 183n
Theta functions, 146, 198n West, M. L., 183n
Index 213

Weyl, Hermann, 140, 193n, 196n

Widman, Johann, 29
Wiles, Andrew, 189n
Williams, Kenneth S., 198n
Wilson, Curtis, 193n
Winternitz, Emmanuel, 184n
Wussing, Hans, 188n189n, 193n

Xenophon, 182n

Yaglom, I. M., 192n, 195n196n

Yamey, B. S., 184n

Zammattio, Carlo, 184n

Zero, 9, 43