You are on page 1of 222

Abel’s Proof

This page intentionally left blank

Abel’s Proof
An Essay on the
Sources and Meaning
of Mathematical
Unsolvability

Peter Pesic

The MIT Press
Cambridge, Massachusetts
London, England

Abel. Robert W. Judy Feldmann. Chry- seis Fox. which appeared in Investigación y Ciencia. July 1990. roots of.3). Oslo Division (figure 10.P47 2003 512. Norway (figure 6. Lucent Technologies. Includes bibliographical references and index. Library of Congress Cataloging-in-Publication Data Pesic. Mary Reilly. I.3). ISBN 0-262-16216-4 (hc : alk. c 2003 Massachusetts Institute of Technology All rights reserved. or information storage and retrieval) without permission in writing from the publisher. Peter. Permission for the use of the figures has been kindly given by the follow- ing: Dover Publications (figures 2. Lisa Reeve. Oslo Division).2). paper) 1.1. 1802–1829. Eckert (Bell Labs/Lucent Technology). Equations. Nils Klitkou (National Library of Norway. This book was set in Palatino by Interactive Composition Corporation and was printed and bound in the United States of America. John Grafton (Dover Publications). 3. Busichio and Edward J. Gray (fig- ure 8. Thomas Hull. Conservation Departmentale de Musées de Vendée (figure 2.2). recording. 2. Abel’s proof: an essay on the sources and meaning of mathematical unsolvability / Peter Pesic. p. Purificación Mayoral (Investi- gación y Ciencia). Niels Henrik. Special thanks to Wan-go Weng for his calligraphy of the Chinese character ssu (meaning “thought”) on the dedication page. and Ssu Weng for their help with the figures. Peter M. University of Oslo). Yngvar Reichelt (Department of Mathematics. Francisco González de Posada (figure 10./Bell Labs (figure 10.9 4—dc21 2002031991 . University of Oslo. QA212 . No part of this book may be reproduced in any form by any electronic or mechanical means (including photocopying. Department of Mathematics. cm.1). Title. George Nichols. I am grateful to Jean Buck (Wolfram Research).1. page 82). Inc. National Library of Norway.1).

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 Abel’s 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: Abel’s 1824 Paper 155 Appendix B: Abel on the General Form of an Algebraic Solution 171 Appendix C: Cauchy’s Theorem on Permutations 175 Notes 181 Acknowledgments 203 Index 205 .

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. that is. Five years later. the highest power. Abel died at age twenty-six. Few read his pamphlet or paid much attention. a n−1 . apart from mathematicians. just before his work started to receive wide acclaim. . Boxes develop arguments and give examples in greater depth. Devel- oped by others. . Abel’s proof concerns the solution of algebraic equations. The notes give references and sug- gestions for further reading. Yet. The appendixes go further still. . If n = 1. . This book tells the story of Abel’s problem and his proof. a 0 are constant “coefficients. including an annotated transla- tion of Abel’s pamphlet. Introduction In 1824. The text contains a minimum of equations. where x is the “variable” whose “solutions” or “roots” we are seeking and a n .” We classify algebraic equations by n. but these can be skipped without guilt. a young Norwegian named Niels Henrik Abel pub- lished a small pamphlet announcing a new mathematical proof for an old problem. . his ideas remain generally unknown. his insights became a cornerstone of mod- ern mathematics. equations of the form a n x n + a n−1 x n−1 + · · · + a 0 = 0.

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

however complex. nth degree are not solvable in radicals either. To find it. . Mathe- matical symbols may indicate hidden truths that have deep human significance. fifth roots. multiplied. . Absorbed in advanced studies. a 5 x 5 + a 4 x 4 + a 3 x 3 + a 2 x 2 + a 1 x + a 0 = 0. I could understand. experts may cease to wonder about the elementary. as we could for all equations up to the quintic. and other algebraic expressions. subtracted. I studied modern texts. They might not notice the kind of basic insight I was seeking. equations of the sixth. if one can use such a word? Such questions have moved me since childhood.Introduction 3 special quintic equations that have solutions we can express in terms of radicals. but the key re- mained elusive. if we take the general equation of the fifth degree. What is worse. retracing the journey recounted in this book. 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. but we cannot express them in any finite formula. The story begins in ancient Greece and has climactic scenes in Norway and France in the 1820s. eighth. Abel’s proof contains a prime secret: how can a search for so- lutions yield the unsolvable? Perhaps if I tried hard enough. That is. What Abel found is indeed surprising and strangely beautiful. involving just roots and powers. and divided in some way. . . what is the signifi- cance of this breakdown. 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. I needed to return to the sources. even as they transcend the human. added. the same holds for equations of any degree higher than five: in general. seventh. there are values of x that will satisfy the given quintic equation.

This page intentionally left blank .

About 2. and the icosahedron (figure 1.” the only solid figures that are regular (having identical equal-sided faces): the tetrahedron. The Elements. culminating in the construction of the five “Platonic solids. a philosopher named Pythagoras and his followers adopted the motto “All is number. Euclid’s masterful compilation. Because of their secrecy. the cube. Those who followed looked back on the Pythagoreans as the source of mathematics.1 The Scandal of the Irrational The story begins with a secret and a scandal. con- sidering themselves guardians of the secrets of mathematics from the profane world. the idea that one could construct irrefutable demon- strations of theoretical propositions that would admit of no .500 years ago. the dodecahedron. The major contribution of the Pythagoreans. and even the degree to which they were indebted to prior discoveries made in Mesopotamia and Egypt remains obscure. includes Pythagorean discoveries along with later work.1). But they also wrapped themselves in mystery. was the concept of mathematical proof. the octahedron. written several hundred years later.” The Pythagorean brotherhood discovered many important math- ematical truths and explored the ways they were manifest in the world. though. many de- tails of their work are lost. in Greece.

d. icosahedron. cube. a. as illustrated after Leonardo da Vinci in Luca Pacioli. On the Divine Proportion (1509). (a) (b) (c) (d) (e) Figure 1. dodecahedron. e. c. octahedron.1 The five regular Platonic solids. b. . tetrahedron.

That is. 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). nor as any whole-number ratio based on it. 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 . There is a third Pythagorean myth that tells of an unfore- seen catastrophe. who. The myths surrounding the Pythagorean brotherhood hide exactly who made their discoveries and how. Despite their motto that “all is number.” implying certain and secure knowledge. For in- stance. consider a square with unit side. Here they went beyond the Babylonians. the fourth (4:3). whereas the ancient Greeks used the ratio 3:2 to denote a relation between unbroken wholes. Its diagonal cannot be expressed as any integral multiple of its side.) 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. a word whose meaning is “the things that are learned. as ratios of the weights of the blacksmith’s hammers. seem not to have been in- terested in proving propositions. The Pythagoreans and their followers created “mathematics” in the sense we still know it. de- spite their many accomplishments. This re- vealed to him that music was number made audible.” the Pythagoreans discovered the existence of magnitudes that are radically different from ordinary numbers. the fifth (3:2). they are incommen- surable. Box 1.The Scandal of the Irrational 7 exception.1 describes the simple argument recounted by Aristotle to prove this. (This is a good point to note an important distinction: the modern fraction 32 denotes a breaking of the unit into parts.

Therefore the diagonal of a square cannot be expressed as a ratio of two whole numbers. Therefore m2 is even (being two times an integer). which is therefore irrational. they have no common factors. Thus the hypothesis must have been wrong: no ratio can represent the relation of di- agonal to side. since the square on the hypotenuse s is double the square on the side. s = m:n. If m is even. our original assumption that s = m:n must be wrong. and n2 = 2 p 2 . Then m2 = 4 p 2 = 2n2 . that is. namely that one and the same num- ber must be both even and odd. But then n must be odd. Now note that s 2 = m2 :n2 = 2:1. by the Pythagorean theorem. We can assume further that m and n are expressed in lowest terms. Since a whole number cannot be both even and odd.8 Chapter 1 Box 1. since otherwise one could divide m and n by a factor of 2 and simplify them further. and so too is n. leads to an absurdity. we can let m = 2 p. . where p is some number.1 The diagonal of a square is incommensurable with its side S 1 1 Let the square have unit side and a diagonal length s. to use a modern term. and so too is m (since the square of an even num- ber is even). Then suppose that s can be expressed as a ratio of two whole numbers. But this means that n2 is even.

arithmos) as opposed to a magni- tude (megethos).” The new magni- tudes are called alogon. a word whose Latin root means unbroken or whole.ta meant the cosmic order. mirrored in proper ritual: certain things must come first. This distinction later became blurred.” Thus. due order is important. recognizable in such English words as rite and rhythm. we do not hear people counting objects as “zero. account. Here. but also irrational quantities. the expression “there are no ap- ples here” means “there aren’t any apples here” more than “there are zero apples here.The Scandal of the Irrational 9 The original Greek term is more pungent. In Vedic India. .ta) stood for untruth and sin. The word arithmos prob- ably goes back to the Indo-European root (a)rı̄. The word for ratio is logos. meaning “inexpressible. r. The word arithmos denotes the counting numbers beginning with two. unsayable. and so on. it is not possible to stick in upstart quantities like “one-half” or (worse still) “the square root of 2” between the integers. which means “word. three. . . but they are inexpressible in terms of ordinary numbers. whose opposite (anr. for “the unit” or “the One” (the Greeks called it the “monad”) was not a number in their judgment. but for now it is crucial to insist on it. . Ancient mathematicians emphasized the distinctions between differ- ent sorts of mathematical quantities. reckoning. . the Greek word for “counting number” goes back to a con- cept of cosmic order. the regular course of days and sea- sons. Thus. one. two.” It was only in the seventeenth century that the word “number” was extended to include not only the counting numbers from two on.” Irrational magnitudes are logical consequences of geometry. and the Greeks were careful to use entirely different words to denote a number (in Greek. others second. The Greeks did not know the Hindu-Arabic zero and surely would not have recognized it as an arithmos. even now.” from a root meaning “picking up or gathering.

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.
Pappus’s 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 Plato’s 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 Meno’s 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 boy’s 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. he then draws the diagonals AB. later voted to execute the philosopher. but realized that in fact that HEFG has four times the area of AEBO. a fact that was not lost on Plato or his hearers. BC. so all four of these together give the true doubled square ABCD. Their outrage points to the power of the new mathematical insights. The process of doubling a square (an eminently rational enterprise) required recourse to the irrational. The slave boy thought that the square on the doubled side HE would have twice the area. DA within the square HEFG. it was cru- cial. Each triangle AOB = BOC = COD = DOA is exactly half the area of the original square. CD. Though Socrates had not referred to the irrationality of the diagonal.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. . At Socrates’ prompting.

5. a danger as fear- ful as drowning. Also present during their conversa- tion was Theaetetus’ teacher Theodorus. In Plato’s 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. to use an unforgettable image from another dialogue. Socrates was deeply impressed with this youth. we learn that he made pro- found discoveries about the irrationals and the five regu- lar solids. along with music and gymnastics. Socrates pre- scribes mathematics. for these young irrationals to tame what is most disorderly and incommensurate in their souls. Certainly the Pythagoreans took this dire view. may be harmo- nized with the rational. It is named after Theaetetus. where for some reason he stopped. What is irrational. the black horse of passion may be yoked to the white horse of reason. clearly the word “irrational” already had acquired the emotional conno- tations it still retains. all the way to 17.. who explains his discovery that there are . 7. dying from battlefield wounds and dysentery. who seemed destined to do great things and who also resembled Socrates physically. a math- ematician who is introduced to us as he is being carried back to Athens. in the soul or in mathematics. an √older √ mathemati- √ cian who had √ proved the irrationality of 3. . down to the snub nose and bulging eyes. . In a flashback to Theaetetus’ youth. .” Appropriately and yet ironically.The Scandal of the Irrational 15 Though a consequence of logical mathematics. Socrates’ characteristic irony is not in evidence as he ques- tions Theaetetus. but Plato’s dialogues open a larger perspective. His joke points to a widely held sense that irrationality in mathematics was a troubling sign of confusion and disorder in the world. and he conversed with Socrates shortly before the philosopher’s trial and death. Plato’s great dialogue on the nature of knowledge rests on this mathematical crux.

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.” √since 2 their squares √ 2 have a common measure (that is. Though he depicts himself as sterile. Theaetetus’ mathematical insight is commensurate with the bravery that will allow him to fight for his city and die with exceptional honor. so much like himself. Anticipating his indictment on the very next day. Socrates helps Theaetetus bring his conception to birth and tests its health. explaining that he is really a midwife who helps people deliver them- selves of their conceptions. Socrates had often made fun of his own ugly features. but he describes Theaetetus as beautiful. since 3 = 3 and 17 = 17 are both in- tegers). barren of wisdom. In Greek. Far from feeling antagonistic.” a mineral that is able to distinguish gold from base metal by the mark each makes on it. Such is the courage of one who could wrestle with the irrational.” by which he means that they should struggle fearlessly to test and refine their opinions together. the word for “torture” can also mean the “touchstone. Socrates encourages his guests to “put themselves to torture. testing whether other knowledge comes through the senses or more mysteriously from within the soul. Theaetetus is ready to enter a searching inquiry that begins with mathematics as a touchstone of true knowl- edge. 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. Socrates justifies himself not to his angry accusers but to this gentle and gifted young man. they are still “com- mensurable in square.16 Chapter 1 degrees of irrationality. During their conversation. This extreme metaphor has overtones of the judicial torture used to coerce . Though such magnitudes as the square roots of 3 or of 17 are irrational.

Be- yond presenting his own results. he proposes that if we want to check whether two ratios are equal. Nevertheless. he introduces a far-reaching definition of equality or inequality that extends to ratios of irrational magnitudes. In Book V. whose study often seems painful to those who do not know the pleasure of insight. This . or less (box 1. who kept numbers and irrational magnitudes strictly apart.” The discoveries of Theaetetus and the test of mathemati- cal proof were enshrined in Euclid’s Elements. Euclid set the discoveries of others in order as a touchstone of mathematical lucidity and logical force.The Scandal of the Irrational 17 truth from slaves. In the case of the irrational. Euclid’s contribution went far beyond this separation of realms. This they have learned from mathematics.4). Euclid drew on a compromise introduced by Eudoxus. whose mixing might lead to incalculable confusion. but Socrates uses it to signify a search for truth that defies even intense pain and humiliation. he would never mix the two dis- tinct types so as to allow a ratio between a number and a magnitude. 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). It is no wonder that Plato placed over the door to his academy the admonition: “Let no one ignorant of geometry enter here. Socrates and Theaetetus see in suffering the path to the superlative pleasure of ultimate truth. which remains even today a living fountainhead of mathematics. equal. invaluable for beginners as well as experienced mathematicians. Following Eudoxus. but yet in propor- tion. Like sol- diers or athletes. This was not simply mathematical apartheid but a decision to consider numbers and magnitudes as entirely distinct genera. we should multiply the terms by various integers to check whether these mul- tiples are respectively greater. For instance.

the following holds: if ma > nb. To show this indefinite divisibility. whose inte- gral multiples constitute all the counting numbers. 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. Definition 5) The ratio a :b is said to be equal to the ratio c:d if. a trial by multiplication. for integers are commensurable because they have a common unit. then mc = nd. if ma = nb.18 Chapter 1 Box 1. it is really a test. when ma is compared with nb and mc is compared with nd.4 Euclid’s definition of equal ratio. meaning a continuously and endlessly divisible magnitude. Thus. then mc < nd. Euclid demonstrates how we can successively subtract from any magnitude half or more of that magnitude. even though the magnitudes themselves may have no common measure. definition of equality still depends on testing multiples of magnitudes. Though implicit in geome- try. and then keep repeating this process until . as opposed to the indivisible One. Euclid’s most daring inquiry into the irrational occurs in Book X. he brings into prominence what later came to be called the continuum. for any whole numbers m and n. then mc > nd. and if ma < nb. a way to nav- igate the irrational sea. as if to examine all possible multiples in order to de- termine whether the multiplied ratios could ever be equal. It also uses any multiples what- ever. Euclid puts it in strong contrast with the way he treats whole numbers in Books VI and VII. which is applicable to any magnitude (Book V.

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. Then Euclid sets out to classify different kinds of irra- tionals. finally we have left a magnitude that is smaller than any given amount (box 1. for if there were. and the same proportion yet again of what remains. If we try to divide an octave (whose ratio . all magnitudes would share that smallest magnitude as their common measure. which might be the side or diagonal of another square. Euclid sets in play a process indefinitely repeated. This means we can formulate a musical version of the mathematical crisis of the irrational. but we can also divide a string to make it sound different intervals. continuing the process as far as necessary so that the remainder can be made less than any given line. no geometrical “atom” or least possible magnitude making up all others.The Scandal of the Irrational 19 Box 1. The diagonal of a square is irrational compared to the side. Proposition 1) Take half (or more) of the given magnitude. Here again. As Theaetetus had shown. there is no smallest magnitude. Thus. but it can be commensurable with another line. nonetheless.5 Euclid’s statement of the indefinite divisibility of any magni- tude (Book X. irrationality is a relative term. What is speakable depends above all on the relation between figures.5). and then the same proportion of what remains. Euclid’s classification of irrationals is intricate. naming them and showing their interrelations. not pic- turable in a single figure but intelligible and logically com- pelling.

which later was called the “devil in music. 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. such harmonic problems are critical.6 The sound of square roots Take two strings. in modern decimals. Then find the geometric ratio (also called the mean proportional) between these strings. If the whole universe is based on number. one can go on to define still other irrational lines that are “infinite in number. this is a portentous statement. is 2:1) exactly at the√point of the geometric mean.414. the “ratio” needed is 2:1 ≈ 1.6).” the tritone. he says something amazing in the final proposition of the book: From the lines already drawn. The discovery of the irrational disclosed an infinitely branching path. each a different species with infinitely many ex- amples. so that their lengths are in the ratio 2:1. and none of them is the same as any of the preceding. . This is close to the dissonant interval called the tritone. Euclid presents his classification of irrationals through a hundred careful propositions. This is very close to the highly dissonant interval later called the “devil in music. cross-multiplying √ this gives x 2 = 2. After this tour de force. This means that 2:x = x:1. The tritone is thus 9:8 × 9:8 × 9:8 = 93 :83 = 729:512 ≈ 1.424. one sounding an octave higher than the other. we get the mongrel “ratio” 2:1 (box 1. the length x at which 2:x is the same proportion as x:1.” namely the interval composed of three equal whole steps each of ratio 9:8. Thus.20 Chapter 1 Box 1.” Although his tone is impassive.

and wonder. By this final proposition. trackless sea of end- lessly different magnitudes. as if to say: Here lies an unfathomable. infinite in number though each is finite in mag- nitude. He might have meant: Here lies an inexhaustible store of treasures. But there is another possible reading of his si- lence. Euclid could have meant to indicate a disturbing glance into the irrational abyss. .The Scandal of the Irrational 21 Euclid’s impassive tone does not disclose what he thought of this situation. from which one should turn away in horror. Behold.

This page intentionally left blank .

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

as in box 2.1). Similar statements could be made about the methods of the Arabic algebraists. Such a notation of place-value was unknown in Europe until the Renaissance. the notation for the modern fraction 0. but they could not go far without the symbols that might have enabled them to speak with more generality. 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).2. though they could solve specific examples. they could not express the general solution of an equation with arbitrary values.24 Chapter 2 force of geometry and tried to illustrate their solutions with geometric examples. the Babylonians made difficult calcu- lations using a notation essentially equivalent to modern place-value. the Babylonians were able to solve many similar problems four thousand years ago. 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 14.5 would be 0. With no symbolic notation for unknowns. When they considered a problem. 30 = 30 60 = 0(60)0 + 30(60)−1 . What is surprising. but based on the number sixty (box 2. and the rich achievements of Babylonian mathematics came fully to light only in the twentieth century. . 30 = 14(60)1 + 30(60)0 . In the case of the quadratic equation. is that some of the greatest ad- vances in solving algebraic problems came long before such powerful symbols were devised. Notably more advanced than the Egyptians. then. the Babylonians wrote it out and solved it in words.

Like the Babylonians. the side of the square.5. . . and a 0 = −870. Compare the general expression of this method in modern symbols shown in 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. which is 0. lawsuits. al- Khwārizmı̄ focuses on specific problems. add this to 870 to get 870.5. . trade.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). and canal building. namely x = −a 1 ±     2 a 12 − 4a 2 a 0 2a 2 gives x = 1 2 + 870 + 12 . Now add 0.” Compare this with the modern solution of the equation a 2 x 2 + a 1 x +a 0 = 0. we will consider the history and signifi- cance of these symbols. each term of which corresponds to a step in the Babylonian solution. His method is that of “completing the square.” Box 2.5. but he adds that “it is necessary that we should demonstrate geometrically the truth of the same problems which we have explained in numbers.5.25.5 by 0. with a 2 = 1. This is the square of 29.5 to 29. a 1 = −1. Writing in the ninth century A.Controversy and Coefficients 25 Box 2. and the result is 30..3.25. Shortly.” though al-Khwārizmı̄ confines it to this one example.D. and multiply 0. in all their dealings with one another”. he sets forth al-jabr as an eminently practical art. West was Muhammed ibn-Musa al-Khwārizmı̄. with the numerical val- ues translated into decimal notation: “Take half of 1. land measurement. “such as men constantly require . he gives as instances the prob- lems that emerge from inheritances. which is 0.

a 1 = 10. in modern notation). labeled e. so the large square JKLM has area 39 + 25 = 64 and hence side 8. We are given that the central square plus the rectangles equals 39. h. But the side of the large square is 2 12 + x + 2 12 = 5 + x = 8.  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 . Each of these is a rectangle of sides x and 2 12 . g. f.26 Chapter 2 Box 2. a 2 = 1. Add to this square four equal areas.3 Al-Khwārizmı̄’s geometric explanation of the solution to quadratic equations x 2¹⁄₂ J K e 2¹⁄₂ 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 comparison. so x = (−10 ± 100 + 156)/2 = (−10 ± 16)/2 = 3 or −13. each of area 2 12 x. Now complete the larger square JKLM by including the little squares at each corner. its side is thus x. so we conclude that x = 3. . Let the central square ABCD have area x 2 . the total area of all four corner squares is 4 × 6 14 = 25. In this case. so that together they have area 10x. whose area is each 2 12 × 2 12 = 6 14 . 2 √ a 0 = −39.

At this time. as the Arabs’ work had also been. a trader has “an argosy bound to Tripolis. this system requires a twofold listing of all debits and credits.ifr gave birth . after long periods of inactivity. a fourth for England—and other ventures he hath. As merchants strove for rapid increase of capital. as did bills of exchange in draft form and modes of international credit. The new bookkeeping was extremely awkward with Greek or Roman numerals but flourished with the advent of Hindu- Arabic numerals. as Shylock describes it. he hath a third at Mexico. These mathematical innovations were often related to problems raised by economic changes. the Arabic word s. Europeans began to work in what they called “the great art. moreover. The basis of all subsequent accounting practice. I understand. double-entry bookkeeping was an in- valuable innovation.” also called the art of the coss or cosa. 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. great strides were made in a matter of years by a few individuals. squand’red abroad. recorded in a single currency and checked by balancing both the debit and the credit sides of the ledger.” In this new situation. European goods ranged widely. meaning the “thing” or unknown. upon the Rialto. another to the Indies. which Leonardo of Pisa (better known as Fibonacci) helped introduce to Europe in his Book of the Abacus (1202). During the centuries following the Crusades. The story of early modern algebra is tumultuous. The ancient methods of bookkeeping also broke down when. they found they needed new ways to deal with dispersed partners and factors operating in diverse local economic situations. Marine insurance emerged.Controversy and Coefficients 27 As the algebraic writings of the Arabs were translated into Latin. traded for spices and silks.

Such practical economic problems lead to more speculative questions: “How many pairs of rabbits will be produced in a year. . 13. . Luca Pacioli’s Summary of Arithmetic.718 . “for without systematic recording. who illustrated his book On the Divine Proportion (1509). 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. . This se- quence could describe the growth of capital as well as an in- creasing population of rabbits. which he argues cannot be solved with integers. 1. a treatment of the “golden ratio” of the Greeks. . . the key to exponential growth. 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. Using Fibonacci as its main source.” essential concepts for reckoning with numbers as a symbolic code. Fibonacci’s book concerns problems in commercial trans- actions. their minds would always be so tired and troubled that it would be impossible for them to conduct business. . Proportions.” 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. square roots. 3.28 Chapter 2 to the English words “zero” and “cipher. emerges during a further consideration of compound interest. and Proportionality (1494) is the first treatise on double-entry bookkeeping. . Fibonacci also considers cubic equations. but for which he presents quite accurate approximations. 8. in which each term is the sum of the two immediately preceding it. Pacioli was a friend and colleague of the artists Piero della Francesca (from whom he plagiarized his mathematics) and Leonardo da Vinci. Geometry. 5. . beginning with a single pair. especially currency exchanges that involve fractions. the quantity now called e = 2. and fractions. 2. Later.

the oldest book in which the familiar “+” and “−” signs appear in print.Controversy and Coefficients 29 It is appropriate that Pacioli’s Summary was one of the first books to be printed on movable type. Here again these symbols at first refer to surplus and deficiency in warehouse inventory. it is not the crucial innovation of the double-entry scheme. including questions about the solvabil- ity of equations. such as the “Treviso Arithmetic” (1478) and Johann Widman’s Mercantile Arithmetic (1489). We gain a similar impression from other early works. . since it appealed to the increasing audience of businessmen who needed systematic and ironclad schemes of accounting for their businesses. for the accounts of Roman slaves or medieval factors were also able to be balanced. 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.” such as “depreciation” or “good-will. Though this is surely useful. Per- haps the best known aspect of the double-entry system is the duality between credit and debit. practical considerations lead sur- prisingly quickly to questions that transcend the symbols’ commercial origins.” Pacioli places his discussion of bookkeeping and commer- cial mathematics next to his exposition of algebra. As with Pacioli. 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. including both “real accounts” that rep- resent concrete transactions with a specific client or credi- tor and what now are called “nominal accounts. only later becoming signs of abstract operations. manifest in the dual record- ing of every transaction on both sides of the ledger. Double entry allows not only bal- ance but completeness of accounting because it records every kind of transaction.

Their gain is equal to the cube of the tenth part of their capital. al-Khwārizmı̄) as well as Fibonacci.1). who had done extensive work classifying cubics and exploring their solvability. and he cites “Mahomet the son of Moses the Arab” (that is. It was his interest in gambling that led Cardano to write the first book on the mathematics of probability. Piero della Francesca. quartic. and his Book of My Life is astonishingly frank. Despite all this. and even quintic equations. If they had made . Cardano was a physician. the subject of contention even among his contemporaries. all of their work had been lost. whose religious opinions were sufficiently heretical that the Inquisition jailed him in his old age. detailing not only his own passions and obsessions but also the misadventures of his children (one son poisoned his wife and was executed. an inveterate gambler. His role in the history of algebra is central but controversial. and an astrologer.30 Chapter 2 Though the Babylonians had made significant strides in solving what we would call cubic equations. the other was a ne’er-do-well who robbed his father). Cardano treats problems that we would call quadratic as al- ready solved. or the Rules of Algebra (1545). whose life spanned most of the sixteenth century (figure 2. 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. of whom the most colorful was Girolamo Cardano. In his book The Great Art. Fibonacci thought such equations could not be solved by magnitudes and square roots. The breakthrough came from a motley bunch of Italians. he held prestigious professorships and eventually re- ceived a pension from the pope. He wrote prolifically. Pacioli thought them quite unsolvable by algebra. a great mathematician as well as painter. an opinion he derived from the Arabic poet-mathematician Omar Khayyām. tried unsuccessfully to solve cubic.

three ducats less.1 Girolamo Cardano. classical statues were regularly being . What was the capital and their profit?” In  x modern 3 terms. if the unknown capital is x. The commercial aspect of such problems may have been less important to Cardano than the possibility of rivaling the ancients. In those days. they would have gained an amount exactly equal to their capital. then  x the 3 gain is 10 and the equation defined by the problem is 10 −3 = x.Controversy and Coefficients 31 Figure 2.

which allowed him to win a public contest with del Ferro’s student. what is clear is that the solution of the cubic became the first example of a sordid modern genre: the scientific priority fight. At any rate. Tartaglia may well have had a hint from this earlier source. The chain of events is not completely clear. and here the moderns might still emulate and even surpass them. Niccolò Fontana. The popular appeal of these mathematical gladiato- rial matches was on a par with bear-baiting and involved something of the same atmosphere. standards of attribution and citation were lax. Were these ancients gods. they were rowdy tournaments of skill. in which each contestant would try to stump the other with a problem to win a handsome prize. who did not publish it but disclosed it to a student before his own death. if del Ferro’s student had learned the solution from his teacher? In those days. astonishing beholders with the immense achievements of antiquity. Tartaglia did make a break- through in solving cubic equations. But how could Tartaglia win. Tartaglia had already published someone else’s translation of Archimedes as if it were his own.32 Chapter 2 unearthed in Italy. These contests show something of the status of algebra at the time. In those days. but in truth he started his work after had heard that it had already been solved by Scipione del Ferro. and the Greek mathemati- cians were likewise just beginning to be read. The works of Plato had only re- cently been translated into Latin. claimed to have solved the problem. not to be touched by the moderns? Yet the an- cients had not solved the cubic. there were many variations of cubic problems that seemed . nicknamed Tartaglia (“Stammerer”). and he also tried to pass as the author of the law of the inclined plane without giving proper credit to its true discoverer.

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. Years passed. attributing it to Tartaglia. 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 says that before seeing this work.” as Cardano calls it (for instance. almost riddling enough to be in code. Cardano visited Bologna and gained access to del Ferro’s papers.Controversy and Coefficients 33 separate. the student might have been able to solve a problem like “the cube and the number equal to the first power. during which Tartaglia did not publish his so- lution. Thus. In any case. but without mention of his oath of secrecy. who pressed him for details of his methods. he had believed the claim of . On a more basic level. since only positive integers were then rec- ognized. in 1543. He disclosed his se- cret to Cardano in 1539. Tartaglia was wary but needed money. Then. Tartaglia communicated the solution in the form of a cumbersome poem. news of Tartaglia’s victory reached Cardano. and del Ferro before him. hinting that he would arrange a meeting with a wealthy patron. By giving credit. x 3 + 21 = 16x). but only after making him swear not to reveal it. Cardano then felt free to publish the method in his Great Art. which showed that del Ferro had in fact solved the cubic before Tartaglia. any case involving negative numbers would have had to be treated separately. he surely sought to avoid the appearance of trying to steal the secret. and to which various techniques were applied. by moving such numbers to the other side of the equation. 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).

which he di- vides into thirteen cases. Tartaglia was outraged and published accusations of plagiarism. or so it seems from subsequent events. reduce it to a sim- pler problem you have already solved. How. But by mentioning del Ferro’s earlier discoveries. he only remarks that. Though they did not present their deduction with a three-dimensional diagram. Tartaglia did not have the complete solution to every variety of cubic that Cardano presents. Ferrari challenged Tartaglia to a mathematical duel that ended with Ferrari’s victory. after learning it. we can solve a quadratic equation by reducing it to the form of a per- fect square and then taking the square root. then. many with subcases.34 Chapter 2 Luca Pacioli that there was no rule that solved equations beyond the quadratic. he went further than Tartaglia ever did. Still. Tartaglia. including a professorship in Bologna. the technique of “completing the square” is the key. while Tartaglia’s account reveals that he left even before the contest was over. in response to his en- treaties. 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. Here. and Cardano showed how any given cu- bic equation can be put in the form of a perfect cube equal to some numerical value. As noted earlier. With his master’s honor at stake. It is natural to try to attack the cubic in the same way by “completing the cube.” Del Ferro. Tartaglia “gave” it to him and also that. for Ferrari obtained many flattering offers. Then we can complete the solution by taking the cube root. it is useful to . At this point a new challenger stepped in. In explanation. he diminishes Tartaglia’s claim to uniqueness. In fact. Cardano had a ser- vant named Ludovico Ferrari who showed such keen interest in mathematics that Cardano took him on as a student.

along with its solution in modern notation. that is. and both have a geometric significance that is lacking in the case of “square-square” equations. which they called “square- square. It turns out only to require complet- ing the square in a somewhat different way. cubic problems seem entirely separate from quadratic ones. completing the cube involves a series of interlock- ing solids that must be completed and reassembled.4 shows an example from Cardano.Controversy and Coefficients 35 consider such a picture (box 2.4). Because cubic equations can be solved by a three- dimensional puzzle. Ferrari addressed the problem of solving what we would call equations of the fourth degree. it only involves nested radicals. At Cardano’s request. That is. which in turn contain expressions involv- ing square roots.3) it shows that this problem is essentially like a jigsaw puzzle. Though the result looks a little daunting. for (like the parallel pic- ture shown for the quadratic equation in box 2. The solution of this three-dimensional puzzle requires breaking it into slices. one might think that quartic equations would somehow require struggling with a puzzle in four di- mensions. Just as the solution of the quadratic involves square roots. Box 2. In Cardano’s book. Later. roots of roots (such as the square root of a cube root). the solution of the quartic does not require a fourth spatial dimension. 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. In fact.” Their use of this term shows that they did not yet have our more general conception of equations of arbitrary degrees. or quartics. the solution of the cubic in- volves cube roots. using x 2 as the .

in modern notation.36 Chapter 2 Box 2. but the overlaps also overlap in the little shaded cube of volume v 3 . In the diagram. let x 3 be the upper shaded cube. To carve the cube with side x from the cube with side u. now embedded in a larger cube with side AC = u. To this expression for x 3 we must add 6x = 6u − 6v in order to satisfy the original problem. Thus x 3 = u3 − {(volume of slices) − [(volume of overlap) − (volume of overlap of overlap)]} = u3 − 3u2 v + 3uv 2 − v 3 . Each overlap has volume uv 2 .” or x 3 + 6x = 20. There is also a lower shaded cube having side BC = v.4 Cardano’s method of completing the cube x v x v x v x x x v A x B v C u=x+v Consider Cardano’s example “Let the cube and six times the side be equal to 20. but they overlap. we need to take three slices. and the trick needed is to set x = u − v. . Each slice individually has volume u2 v. We want to make this problem take the form of a perfect cube.

Substituting this back in.Controversy and Coefficients 37 Box 2. but it is not because it is quadratic in u3 . here grouped to highlight uv. The rest is clear sailing: with this condition applied. Taking  the cube root 3 √ 3 √ of each side gives u = 108 + 10 and v = 108 − 10. v must be subject to one additional condition connecting them. So to complete the cube. That is. Now if uv = 2. v are being substituted for the one variable x. though doubtless guided by Cardano’s example. or u6 = 20u3 + 8 if (like Cardano) we want to keep everything positive. and so u. . Multiplying both sides by u3 gives u6 − 8 = 20u3 . then this equation becomes y2 = 20y + 8. we have u3 − 8/u3 = 20. To make this take the form of a perfect cube. This looks worse. Ferrari succeeded on his own. We know how to solve this quadratic equation:   y = (20 ± 400 − 4(−8))/2 = (20 ± 4(108) )/2 √ √ = 10 ± 108 = 10 + 108. x = 108 + 10 − 108 − 10. we get Cardano’s final 3 √ 3 √ answer. we need to get rid of the “cross terms” −3u2 v + 3uv 2 + 6u − 6v = −3(uv)u − 6v +3(uv)v +6u. because it seems to be of sixth order in u. which we are free to do since the two variables u. This allowed Ferrari and Cardano to treat square-square problems in terms of square problems.4 Continued Now we have x 3 + 6x = u3 − 3u2 v + 3uv 2 − v 3 + 6u − 6v. our cubic looks like u3 − v 3 = 20. where the negative choice of sign is neglected because it leads to a negative √ root. all we need to do is to stipulate that u and v are determined by the relation uv = 2. Finally. which Cardano √ would not have accepted. with v = u2 . Here there was no dispute about priority. let y = u2 . then these terms cancel: −3(uv)u − 6v + 3(uv)v + 6u = −6u − 6v + 6v + 6u = 0. sides of the initial square. So u3 = 10 + 108 and v 3 = 108 − 10. recalling that x = u − v.

in general. In general. We can treat this as a complicated problem of factoring. to use the modern term.5. B. where A.5 Ferrari’s 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 . we could take the square root of both sides and reduce the equation to a quadratic. Ax 2 + Bx + C = (e x + f )2 . it must have seemed that all such problems were now ripe for solution.2) 4 We can make the right-hand side a perfect square by ad- justing y.2).5. it is not a perfect square.38 Chapter 2 Box 2. This is equivalent to requiring that. Box 2. 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. C are the expressions in parentheses on the right-hand side of equation (2.5 shows Ferrari’s solution in modern notation. If .  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 + .1) 4 If the right-hand side were a perfect square. (2.5. Because it followed so quickly on the solution of the cubic.

So (substituting the values for A. y are now determined. How can the human mind under- stand what it cannot see? As Cardano puts it. In going past the cubic. These two equations can readily be solved by the quadratic formula. however. the obvious references to two and three dimensions disappear and the problem becomes more abstract.4) 2 2 which is a perfect square whose square root yields two qua- dratic equations.2).5. (2.5 Continued this is to hold. now written as  2 a3 y x + x+ 2 = (e x + f )2 . B.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. a3 y x2 + x+ = e x + f. “it would be very foolish to go beyond this point. (2.2) into B 2 = 4AC). yielding the four solutions of the quartic equation. Then we can return to (2. 2e To make this consistent requires that B 2 − 4AC = (2e f )2 − 4e 2 f 2 = 0. f.5.5) 2 2 where e. then we can use (2. (2. B = 2e f. Nature does not permit . C from eq.5. Once we have determined y. 2 2 a3 y x2 + x + = −e x − f. (2.5. y must satisfy  2    a3 y a 32 y2 −a 1 + =4 − a2 + y −a 0 + . then we must have A = e 2 .3) to determine e and f . f = C = .5. C = f 2 or √ √ B e = A.Controversy and Coefficients 39 Box 2.5.

10. . “things”) or coefficients. aequari Z solido 2” for the modern x 3 + 3B 2 x = 2Z3 . . 108.6 shows the notation of two Renaissance algebraists. v. it. “A cubus” is x 3 . consonants for coefficients. they were primarily a means of abbrevi- ating ordinary language. 6. He expresses his solution to the specific cubic equation in box 2.” meaning that it is basi- cally abbreviated words. m̃.” and “plano” for “squared”. m̃. The earlier manners of expression operated on entirely dif- ferent principles. “cubus” or “solido” for “cubed. In contrast. and “Z solido 2” is 2Z3 .6 An illustration of the algebraic notation of Cardano and Viète Cardano’s notation is “syncopated. the manner of expression of the algebraists left much to be desired. thus. allowing systematic manipulations that are foreign to natural language. Here p̄. Modern symbolic mathematics steps decisively away from ordinary language. .” Moreover. Although we can translate an equation into words. and m̃. He does not have a general sym- bolism for unknowns (which he calls rebus. 108.40 Chapter 2 Box 2. like shorthand. Note that he uses the modern + and − signs and also the radical √ sign . compared with the modern form. cu. stand for plus and minus. rebus æqualis 20”) as follows: . while . stands for “radix” (square root). stands for “cube root.4 (which he states as “cubus p̄. “B plano 3” is 3B 2 . 10. cu. . Box 2. cu. p̄. He uses vowels for unknowns. v. and . v.” Compare this with its modern form: 3 √ 3 √ x= 108 + 10 − 108 − 10. Viète writes “A cubus + B plano 3 in A.

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, François Viète (figure 2.2), who was

Figure 2.2
François Viète.

42 Chapter 2

able to merge ciphers, legal abstractions, and mathematics
to devise the basic algebraic symbolism we still use. Viète
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 Euclid’s 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 Viète
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, Viète 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 Viète’s 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
Viète 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. Viète 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, Viète 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 user’s 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.
Viète 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 ω, 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 Plato’s, toward a stream of
purely conventional symbols, devoid of intrinsic sense.
The abstractness of mathematics is amplified by Viète’s
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 don’t 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 Viète to

Compare the partic- ular case x 2 − 3x + 2 = 0 with a 2 x 2 + a 1 x + a 0 = 0. Viète’s royal master.” Algebraic symbols avoid the debacle of Babel. bypassing the human confusion of tongues.7). Though he calls himself merely a restorer of what was lost. through his newly systematic codebreaking. Here we have the birth of the coefficient. In 1593. 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. ironically of- fered the king his condolences that there were no French mathematicians up to the challenge. as if they were false alchemists. 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. whereas he has the royal touchstone that will actually separate mathematical gold from dross. He considers Arabic algebra “tedious” and “barbaric. a 1 . he wraps himself in a heroic mantle. Viète was cold to the bitter religious controversies of his age. who sat down with the problem and was able within a few minutes to find the positive roots of the equation. He exults in these analytic powers. he uses to protect France. which enables a huge increase in generality and power. Henry called for Viète. .” rejecting even their word “algebra” for what he calls instead “the analytic art. Viète not only made these important advances but also grasped their immense power. realizing that he has purified an ancient art and given it new life. Viète accepted a challenge that pointed to a new frontier in algebra. which. paying a call on Henry IV. He saves his excitement for mathematics. which in- cludes the specific equation as a single special case among an infinite expanse of possibilities in which the coefficients a 2 . At the height of those powers. a 0 roam over all possible values.” Viète rebukes them. The Dutch ambassador.Controversy and Coefficients 45 allow that quantity to become as indefinite as the unknown itself. Indeed. Stung.

030.7 Adriaan van Roomen’s test problem Solve x 45 − 45x 43 + 945x 41 − 12. Viète emphasized the connection between his “new algebra” and geometry.074x 11 + 7.” .306.512.652x 25 − 378.” Viète’s triumph confirmed his sense that what he had solved was not just one problem but “the proud problem of problems.46 Chapter 2 Box 2.565x 33 − 14.375x 9 − 1.795x 3 + 45x = K .679.100x 27 + 236. though with more tranquil animals.800x 21 − 488.800x 23 + 483.500x 7 + 95.764.040x 31 + 469. Humorously. noting that earlier mathematicians were not able to reconcile them.634x 5 − 3. ironically echo- ing Pythagoras’ mythical gesture. which is: TO LEAVE NO PROBLEM UNSOLVED.658.557. along with his expla- nation of his method.459x 35 + 3. where K is a given number.800x 29 − 117.375x 17 − 232.811.138.280x 15 + 105. farewell.676.150x 37 − 740. he offered to sacrifice a hundred sheep to celebrate the occasion.125x 19 + 384. “But was that not because algebra was up to that time practiced impurely? Friends of learning.942.841.075x 13 − 34.945. Viète published this result in 1595.300x 39 + 111. and look to the just and the good.494. embrace a new algebra.

Was it really possible that Viète had the tools to solve every possible equation. and indeed Viète’s solution in- volved trigonometry. The tenor of Viète’s conviction was clearly that all equations are solvable by the same general procedures used to solve the quadratic. for he recognized that par- ticular equation as closely related to formulas he found in trigonometry. which (as we shall see) goes beyond radicals. Clearly. Had van Roomen changed a single coefficient. Thus. and quartic equations. of whatever degree? In the case of van Roomen’s equation of the forty- fifth degree.3 Impossibilities and Imaginaries Van Roomen was so impressed by Viète’s astonishing feat that he traveled to France expressly to meet him. cubic. Viète’s solution would no longer have been valid. Viète was lucky. the solution of van Roomen’s equation did not after all illuminate the general question of whether all equa- tions are solvable in radicals. Here was a man who claimed to have found the master art that would solve any algebraic problem. 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. Other . and was testing to see if others had discovered it also.

The new foundation would be based on the ratios between the sides of the simplest regular polygons. He was aware.1). holding that “nothing is proved by symbols. his mathematical touchstone. rather than simply the Pythagorean scales (which only used numbers up to 4). in place of the old Pythagorean numerology of simple whole numbers. One of his goals was to create a theory that would encompass contemporary musical scales (which used intervals like 5:4 and 6:5). the equation is cubic and thus solvable by the del Ferro-Cardano-Tartaglia method (box 3. that the Swiss mathematician and clock-builder Jost Bürgi had found algebraic equations whose solutions gave the sides of regular polygons. The question about algebra comes to a head in Kepler’s Harmony of the World (1619). a seven- sided figure that would have introduced harsh dissonances like 7:3 into his scheme. nothing hidden is discovered in natural phi- losophy through geometrical symbols. In the case of the heptagon. This would have caused a crisis for Kepler’s musical theory. and their reservations were an important countertheme to the triumphal march of modern algebra. He had a geometrical argument for this. To do this.” but always thought it fell short of the certainty of geometry. in that the regular heptagon cannot be constructed with straightedge and compass. For all his boldness in astronomy. He considered algebra to be “ingenious and surprising. But the solution involves factors . since accepting this solution would open the door to those terrible dissonances. he needed to use all the regular figures up to and including the hexagon but excluding the heptagon. Johannes Kepler was one who remained criti- cal of the claims of algebra. in which he tried to place music on a new geometrical foundation.48 Chapter 3 mathematicians were skeptical about algebraic solutions. however.” This went beyond Kepler’s vehement declaration that he hated all kabbalists and their esoteric symbolism.

which he took to rest on musical relations. algebra might be irrelevant or inadequate. for on it hung not just musical questions but the whole structure of the universe. Galileo Galilei never once men- tioned algebra. neither can it be known by the human mind. y 3 − 7y2 + 14y − 7 = 0. The issue was important.” he . Others paid even less heed. which seemed at that time bewildering if 3 not absurd. When he famously described the “Book of Nature” as written in “the characters of mathematics. though it would seem likely that he had at least heard of it.1 If we call the side of the polygon x and let y = x 2 . but the solution of the heptagon’s equation using the del Ferro- Cardano-Tartaglia method yields       1 3 7 3 √ √ x=  3 1 − −27 + 1 + −27 . Kepler’s 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. Since even the “Omniscient Mind” of God could not know such roots in a “simple eternal act . These are both solvable.” Because geometry always mirrors the “simple eternal act” of the divine intelligence. If the cosmos were organized geometrically. . Kepler preferred it to the question- able processes required to solve cubic equations. Kepler anticipated correctly that the solution of such complex roots would require an infinite process. . as Kepler believed. . Bürgi’s equation for the pentagon is y2 − 5y + 5 = 0.Impossibilities and Imaginaries 49 Box 3. and for the heptagon. 3 2  √ like 1 − −27.

as an earlier work called it). if x has the dimensions of length and an equa- tion contains an x 3 . Descartes was for some reason reluctant to acknowledge Viète’s pri- ority. René Descartes. before reading the older mathematician’s books.50 Chapter 3 explicitly meant triangles and geometrical figures. Despite the closeness of their work. mentioning him only grudgingly and claiming to have made the same discoveries independently. These technical works vindicated his insistence on correct method (Rules for the Direction of the Mind. La Geometrie and essays on optics and meteorology form an appendix to his Discourse on the Method of Rightly Conducting the Reason. The thread of algebraic mathematics runs through Descartes’s work. reaching toward a larger vision of understanding the world as matter whose motion is mathe- matically intelligible. Nevertheless. which stood aloof from the commercial aspects of early algebra. that is. Viète’s discoveries drew wide attention in France. not alge- braic symbols. Descartes also applied his criteria of method to bring algebra essentially to its modern form. Indeed. Viète had stipulated that all the terms in an algebraic equation should have what we would call the same dimen- sion. Nevertheless. all the other terms must also have the . He also put the new discoveries at the ser- vice of a new philosophy. it must be said that Descartes employed the new algebra with unparalleled clarity and force. who was seven when Viète died in 1603 and whose semi- nal book La Geometrie (1637) would consolidate and extend Viète’s results. He applied this touchstone by systematically doubting the received learning of his time. His optimism about the possibility of solving every problem was shared by his great successor. Symbolic mathematics and rigorous rules exemplify his watchword: method. Perhaps here Galileo was influenced by the academic mathematics of his time.

As he shows how the geometry of Euclid and the conic sections of Apollonius can be described by quadratic equa- tions. Here again the primal notion of number shows its tenacity. some textbooks denied that the prod- uct of two negative numbers is a positive number. if 6 and x each have the dimensions of length. Already the Arabs had realized that some quadratic equations seem to have solutions that are not positive numbers. for the concept of a “negative number” seems to violate our intuitive sense of counting: how can one count negations? Even Viète did not allow his unknowns or coefficients to be negative. where the positive is an advance”—essentially the modern idea of a “number line. that is.” As late as the eighteenth century. the newcomers were not fully welcome. we can simply multiply by a unit length to give 6x the di- mension of volume.” and even Descartes called them “false or less than nothing. Descartes illuminates nagging problems that had shad- owed these equations. .” in which positive and negative represent directions forward and backward along the line. explaining that “the negative in ge- ometry indicates a retrogression.1). They did not recognize negative numbers as numbers. naming them “absurd numbers. Still.Impossibilities and Imaginaries 51 dimension of volume (length cubed). Only in 1629 did Albert Girard recognize negative roots. one sees equations that are almost exactly like ours (figure 3. a rule that Pierre Simon Laplace said “presents some difficulties” even in 1795. although they did seem to have had ways of dealing with debits and credits that approach our concept of signed numbers. Looking at a page of La Geometrie. Other mathematicians of the time refused to admit them. 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.

373). .1 From René Descartes.52 Chapter 3 Figure 3. La Geometrie (1637) (p.

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 . Descartes’s ar- gument is perfectly general and applies to equations of any degree.1.” Box 3.” Box 3.Impossibilities and Imaginaries 53 Though Descartes refers to “false roots. and all those that were true shall become false. He also shows that “it is also easy to transform an equation so that all the roots that were false shall become true roots. He even constructs an ingenious way to tell how many positive and negative roots any equa- tion has.” he does concede that they exist and are roots. the negative sign comes from the odd number of positive roots. Here.2 Descartes’s rule of signs In the page shown in figure 3.” It is amazingly simple: “An equation can have as many true [positive] roots as it contains changes of sign. Here the power of the new notation comes to the fore. Similar considerations lead to Descartes’s general rule of signs.2 shows an example. and as many false [negative] roots as the number of times two + signs or two − signs are found in succession. Note that the equation can be factored into (x − 2)(x − 3)(x − 4)(x + 5) = 0. 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. from + to − or from − to +. . known as “Descartes’s rule of signs. along with Descartes’s argument for why the rule holds. when multiplied out. even if we don’t know their numerical values. because merely by looking at the signs of the successive terms in the equation one can determine the signs of the roots.

” Indeed. and many educated people even today would be hard pressed to account for them. 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. 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. Their alien status is emphasized by calling all non- imaginary numbers “real numbers. in such cases.3).” As difficult as it is to understand what a negative number might mean.” Cardano thought such a quantity was “sophistic” and “as subtle as it is use- less. in the case of cubic equations. if the root is real. . Descartes also opens the door for an even more problem- atic set of solutions that includes square roots of negative numbers. if the root is an imaginary number.54 Chapter 3 by changing the signs of every other term. then the formula does not include any imag- inary numbers. al-Khwārizmı̄ says “there is no equation. we could reject it on the grounds that such a solution is not allowable because it is not real.” The Arabs simply did not allow the square root of a negative number. when one is dealing with quadratic equations. However. In 1560. 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. the del Ferro–Cardano–Tartaglia formula explicitly involves imaginary numbers. and all is well. in accord with his rule. however. 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. even when all the roots are real. This trick works in only a few cases. which we still hold at arm’s length by using the loaded term “imaginary numbers. Conversely. an imagi- nary number is more perplexing still.

This impression is perhaps best dispelled by thinking of complex numbers as points on a plane. he granted the imaginary roots a kind of legitimacy. ( The solution of the heptagon equation in box 3.) Descartes introduced the terms “real” and “imaginary” in their modern sense. but his description perpetuates a certain spooki- ness about quantities that arguably have as much reality as “real” numbers. which we call the imaginary roots. Bombelli showed that 2 − 3 √ −121 =√2 − −1.Impossibilities and Imaginaries 55 Box 3.” By giving both parts a name. with . which is only true if b = 1. by x = 3 2 + −121 + 2 − −121. only spe- cial cases. it cannot be applied to every cubic equation. Bombelli proposed 3 √ √ the substitution 2 + −121 = 2 + b −1. albeit under a name that suggests unreality. where b re- mains to be determined.” Leibniz may have been alluding to the mysterious procession of the Holy Spirit in Christian theology. To reconcile  the two solutions.3 Bombelli’s wild thought Bombelli considered the equation x 3 = 15x + 4. Numbers that have both real and imagi- nary parts he called “complex. This ingenious trick “seemed to rest on sophistry” to Bombelli. This aura persisted even in Gottfried Wilhelm Leibniz’s intriguing description of “almost an amphibian between being and nonbeing. Now by sim- 3 ple substitution we can show that x = 4 also solves this equation. √ 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. whose roots are given.1 that perplexed Kepler is one example that cannot be treated in this way. Sim- √ √ ilarly. in- deed. according  to the del  Ferro-Cardano-Tartaglia so- √ √ lution. The whole solution x is then x = 2 + −1 + 2 − −1 = 4.

In this view. Beyond counting complex roots.2). a complex num- ber is fundamentally two-dimensional. If you wish to increase the value of each root by 3. simply substitute y − 3 into the given equation wherever x occurs. and an nth degree equation n roots. the real part giving the x coordinate and the imaginary part the y coordinate (figure 3. the new equation in powers of y will. a cubic three roots. have roots y = x + 3. Descartes also notes that it is possible to shift the roots of any equation by a certain value. which states that all equations have at least one root.56 Chapter 3 imaginary axis 2i (1. even if those roots are unknown. but for the moment that is only a gleam in Descartes’s eye. substitute sy for x everywhere. ori- ginally formulated in 1629 by Girard. To scale the value of each root by a factor s.2 The two-dimensional representation of complex numbers. by definition. This is a crucial insight. . which expresses the confidence of Viète and Descartes that all algebraic prob- lems have solutions. Descartes states (but does not prove) that each equation has as many roots as its degree: a quadratic equation has two roots. i ) i 1 2 real axis Figure 3. whereas the “real” numbers are one-dimensional. It will later lead to the Fundamental Theorem of Algebra. shifted as required.

which are all affected equally. He does give one example. the four-line locus problem posed by Pappus: to find the curve that lies at a given distance from each of four given lines. in those cases he has to use tools not allowed in an- cient geometry. Not only does Descartes solve this problem with almost ar- rogant ease. with different specific values of the roots in each case. which the ancients did not even attempt. for in the case of a mathematical progression. whenever the first two . However. To be sure. solving an equation of the sixth degree by ingenious manipulations of conic sections. Not only does he have the work of Cardano and Ferrari to rely on. Descartes’s proudest achievement is to have solved another problem that troubled the ancients. ad infinitum. but instead offers a “general rule”: try to simplify the equation to one of lower degree by factoring it.Impossibilities and Imaginaries 57 the new equation in y has roots y = x/s. that is. These triumphs oc- cupy him far more than the problem of solving equations beyond the quartic. but he goes on to solve a five-line locus problem. By such simple transformations. not just straightedge and compass but also conic sections. these transformations do not affect the relative relation of the roots. he restricts its coefficients to allow his geometric method to succeed. Descartes is so sure that equations of degree beyond four can be solved that he doesn’t bother to give specific proce- dures. but he (like Viète 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. With an ironic flourish. the duplication of the cube. We will return to this “relativity” of roots (which Descartes did not discuss) in chapter 9. more and more complex. he concludes that “it is only necessary to follow the same general method to construct all problems. any given equation can generate a family of similar equations. but his equation is not completely general.

it is easy to find the rest. he evidently does not realize how difficult that exercise will prove to be. and Gauss. 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. The next centuries contain many attempts to realize the vision of Viète and Descartes. After Descartes.” But having left the problem of equations beyond the fourth degree as an exalted exercise for the reader. Euler.58 Chapter 3 or three terms are given. .” He asks pardon for what he has “intentionally omitted. a whole huge field of investi- gation lay open. it is perhaps not so surprising that the issue of the quintic equation lay unresolved for two hundred years more. Though there had been extraordi- nary progress in solving cubic and quartic equations during the sixteenth century. so as to leave to others the pleasure of discovery. which will produce the great- est mathematicians yet to have lived: Newton.

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

that is. Newton’s calculus goes beyond anything known to the ancients. without knowing the value of the roots. Though he phrases it in terms of geometry. Newton also makes a number of discoveries that will bear on the problem of the quintic. whom Newton considered his precursor. every root appears in exactly the same way as ev- ery other root. he lectures on the topic (during 1672–1683). constant term is equal to the negative of the product of all the roots. as well as whether they are negative or positive in value. however. as would the products of them taken two at a time. known as “Newton’s identities. then three at a time. the product of all of them would be unchanged. That is. all the other coefficients are equal successively to the sum of all products of the roots taken two at a time.60 Chapter 4 has created: the calculus. He was able to make simple connections between the coefficients of an equation and its roots. Newton’s deepest insight. though his lectures on Universal Arithmetic are not published until 1707. until the final. 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. buried in a passage in his Principia that was . and so on. even Archimedes. to show how large (or small) they could possibly be. for a quintic equa- tion) is equal to the negative sum of all the roots (see box 4. we can look at any equation and determine the range within which its roots lie. though he tends to prefer geometry. remained hidden for many years. Newton is also able to obtain upper and lower bounds for the roots. Likewise. Newton is steeped in algebra. Later on.1).” These generalize “Girard’s identities”: the coeffi- cient of the next-lowest degree term (of x 4 . Thus. The im- portance of these rules is that they let us see direct relations between coefficients and roots. Using these tools. As a young professor. so that if one were to exchange two roots.

” not much noticed until the twentieth century. Multiplying it out. and so on. In lemma 28 of Book I. while the co- efficient of the x term is the symmetric product of all the roots.” which Newton greatly generalized in expressions he derived for the sum of the square of all the roots. Finally. −(x1 + x2 + x3 ). x2 .1 Girard’s and Newton’s identities Consider a cubic equation with the roots x1 . The equa- tion can be written (x − x1 )(x − x2 )(x − x3 ) = 0. the area between those lines cannot be expressed by a finite algebraic equation. if we draw lines across an oval. We can apply the same reasoning to an equation of any degree. x3 . Though he does not define the word explicitly. can in general be found by means of equations finite in the number of their terms and dimensions. cut off by straight lines at will.” in the language of modern mathematicians) and is infinitely smooth (it always has finite curvature. in an investigation of the curved paths that bod- ies can take (as in the orbits of planets). by “oval” Newton seems to mean any closed curve that does not cross itself (it is “simple. The . never is “flat”).Spirals and Seashores 61 Box 4.” That is. the constant term of the equation is the negative product of all three roots: −x1 x2 x3 . he shows that “No oval figure exists whose area. so that the coefficient of the next-to-highest power of the unknown must be the negative sum of all the roots. “Newton’s identities. Note that the coefficient of the x 2 term is the negative sum of all three roots. the next coefficient must be the symmetric sum of all the roots taken two at a time. We will refer to these simple relations as “Girard’s identities. or the sum of their nth powers. taken two at a time: (x1 x2 + x1 x3 + x2 x3 ). we get x 3 − (x1 + x2 + x3 )x 2 + (x1 x2 + x1 x3 + x2 x3 )x − x1 x2 x3 = 0.

Newton indicates that such magnitudes exist (because circles exist. for each hour the hand sweeps through that area. cube roots. Newton’s argument indicates that the area of a circle is not given by any algebraic equation. however high its degree. Inside the oval. and it had long been suspected that the area of a circle is irrational with respect to its radius. To use the term that Euler later introduced. the area of a circle is transcendental. pick any point whatever. for which he does not even bother to draw a picture or write down a line of algebra. At one stroke. fifth roots. Now imagine a point of light moving along that hand. let us call it the pole. His proof is a miracle of simplicity and power. and the moving point keeps track of the area because it is traveling with speed proportional to the area swept out. meaning it cannot be expressed as the root of any equation of finite degree whose coefficients are rational numbers. P.1). and also that there are infinitely many of them. and thus that area (and hence π also) cannot be expressed in terms of any finite number of square roots. Newton has set up a way of measuring the area of the cir- cle. for he knows . Here Newton is implicitly using his new calculus of motion. since his proof is not restricted to circles but holds for any “oval” curve. the name later given (by Euler) to the value of a circle’s perimeter divided by its diameter. It follows from a single brilliant contrivance. Picture a clock hand that makes a complete revolution in one hour. But what Newton surmises goes far beyond the irrationality of π . and beyond the irrational- ity even of π 2 . 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. Now let a straight line come out from that pole and rotate around it at uniform angular speed.62 Chapter 4 simplest such curve is a circle. and so on. and have areas).

Spirals and Seashores 63 A P Figure 4. However. that. we don’t need to know anything about calculus in what follows. the moving point is registering the area that the hand sweeps out. in an infinitesimally short time. A A′ P Figure 4. From any point P inside the oval.2). the point travels a dis- tance from the pole equal to the area the hand has swept out in that time (figure 4.2 Detail of Newton’s lemma 28. which argues that no oval curve has an area expressible by a finite algebraic equation. we can measure the whole . All that matters is that since. draw a straight line that rotates about P at uniform angular speed. the speed of the moving point is proportional to the area swept out between A and A . second by second.1 Newton’s diagram for his lemma 28.

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. Then con- sider a straight line running across the spiral that we will call the x-axis. For the two-dimensional problem of measuring an area. and hence Newton’s conclusion follows: there is no such equation that gives the area of the oval. The hand moves around uniformly. For instance. in pro- portion to the square of the distance from the pole to the oval at any given moment. you would see it move in a spiral. 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. But the spiral in its “infinite number of gyrations” crosses the line an infinite number of times. 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. 0) = 0. starting at the pole and making “an infinite number of gyrations. no matter how large. Thus. Descartes showed that all the conic sections can be described by equations of the second degree. This contradicts our hypothesis that the equation has finite degree. defined by y = 0. f (x. there should be an infinite number of intersection points. Each hour it returns to its initial speed of an hour before. Now Newton relies on the fact that an equa- tion of finite degree can have only a finite number of roots. . Newton has substituted a one- dimensional problem that gives the same answer: find the length traveled outward by the moving point during an hour. but the moving point speeds up and slows down in the course of each hour. If you were to watch the lighted point (or if you were to open the shutter of your camera and make a long exposure). and those curves can be cut by a straight line no more than two times.3).” as Newton puts it (figure 4. y) = 0. corresponding to an infinite number of roots of the equation.

∞) or that is not closed (for example. The argument seemed so simple that it made Newton’s contemporaries sus- picious. Daniel Bernoulli and Leibniz tried to state counterex- amples. but these each involved a curve that intersects itself (for example. The moving point travels an equal distance XX = P X during the next sweep. Newton determines the area of the curve by com- paring the distance from the pole P to the point X after one full revolution. the lemniscate. and so forth. the track of the moving point forms a spiral. such as the circle or the ellipse. a figure-8 on its side. This brilliant argument indicates that all simple closed curves.Spirals and Seashores 65 X′ X P Figure 4. com- posed of the motion of the point along the line and the uniform rotation of the line itself about P. Later mathematicians demanded greater rigor than Newton’s beautifully simple arguments (how can we prove that the spiral must have .3 In Newton’s lemma 28. which sweeps out the full area of the oval. a parabola). have areas that can- not be described by finite algebraic equations.

Leibniz. There are infinitely many magnitudes that are more irrational than any radical. In Newton’s view. Count Ehrenfried Walter von Tschirnhaus. the power and beauty of the algebraic notation ensured that interest in basic questions of algebra remained high. His insights about transcendental magnitudes would be rediscovered centuries later. Despite Newton. succeed- ing generations of mathematicians would take up Newton’s work in the simpler algebraic notation devised by his arch rival. with him as their modern champion. This makes his own preference for geometry (and his geometrically phrased calculus) more persuasive. why worry about the details of quintics? Ironically. but the basic thrust of his insight was sustained. though in his younger days he did spend much energy in classifying cubic equations and contributing to the advance of algebra (for instance. New hope for the solution of equations beyond the quartic was offered by the work of a Saxon noble- man. “Newton’s method” for the approximation of roots). He served in the Dutch army and spent . 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. 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. We will return to them later in light of the unfolding story of the quintic. in the sense that no finite root is commensurable with them.66 Chapter 4 infinitely many intersections with the line?). Tschirnhaus had varied interests. have de- feated the upstart Descartes by subsuming algebra under the larger umbrella of geometry. the ancients. they are transcendental. If the real battle concerns geometric magnitudes not expressible in finite equations. for he thereby places his master theory beyond the limitations of algebra. Implicitly. In this sense.

Tschirnhaus showed that. was doomed to fail. Jerrard was convinced that he had found a solution. the terms pro- portional to the two next-highest powers of x (namely.Spirals and Seashores 67 time in Paris and England. already in the 1680s. The quintic’s intractability was considered a challenge that simply awaited the right assault. in a polynomial of degree n (greater than 2). 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). not less. A century later. In algebra. Leibniz wondered if an extension of Tschirnhaus’s technique could simplify the quintic into the form x 5 equals a constant. Thus. when Tschirnhaus’s friend Leibniz turned his atten- tion to the solution of the quintic. he noticed what later emerged as evidence of a fatal difficulty. Unfortunately. we can always change variables to get rid of the x 4 and x 3 terms. S. Leibniz thus judged that Tschirnhaus’s method. defending it as late as 1858. the equations needed to do this are of higher degree than five. Indeed. Thus. at least. . which would be readily solvable. In 1683. and it was widely believed to be solvable until almost 1800. this approach would not really simplify the problem of the quintic because it leads to equations that are even more complicated. The quintic now looked so simple that it became increasingly perplexing why it would not yield. he sought to generalize Cardano’s approach to the cubic by setting up transformations that would simplify a higher-degree equation.2). though he did not draw the inference that no other method could succeed. x n−1 and x n−2 ) can always be eliminated with one of his transfor- mations (box 4. leaving the general quintic in the much simplified form x 5 + px + q = 0. the Swedish mathematician E. Bring showed that we can also get rid of the x 2 term. His interest in optics led him to set up glassworks in Italy. However. George B. in a quintic.

b 0 through solving these quartic equations. In this period much attention was also paid to the general question of whether all equations do have at least one so- lution. . Jean Le Rond d’Alembert offered a proof. . zn . . where b 4 . George B. it takes megabytes of memory to store the result. b 3 . . In 1843. it proved difficult to establish because of the vast generality . Thus. . . b 3 . these transforma- tions allow any quintic equation to be rewritten in the much simpler form x 5 + px + q = 0. x2 . xn into new roots z1 . a 1 . in general this pro- cedure is extremely complex. . . . Tschirnhaus showed that we can eliminate the two next- higher powers of the unknown. which is the Fundamental Theorem of Algebra that Girard and Descartes had stated without real proof. . Despite the basic simplicity of the theorem. such that z1 = b 4 x14 + b 3 x13 + b 2 x12 + b 1 x1 + b 0 . . However. . a n . .2 Tschirnhaus’s transformation (1683) For any equation of the nth degree. x n−1 and x n−2 by choosing b 4 .68 Chapter 4 Box 4. zn = b 4 xn4 + b 3 xn3 + b 2 xn2 + b 1 xn + b 0 . z2 . z2 = b 4 x24 + b 3 x23 + b 2 x22 + b 1 x2 + b 0 .. . a n x n + a n−1 x n−1 + · · · + a 2 x 2 + a 1 x + a 0 = 0. but it was insuf- ficient. . . . Jerrard showed that the x n−3 term could also be re- moved by a similar transformation. In 1748. . b 0 are functions in radicals of the original coefficients a 0 . . . as were the efforts of Leonhard Euler in the years following. a 2 . . even if we use a computer. . consider a substitution that transforms the roots x1 .

That is.2). and quartic equations.4 The graph of a quintic equation.Spirals and Seashores 69 of the possibilities. cubic. We already know that this happens for the quadratic. y = 0. Ferrari. Note that two of the roots are positive. for those intersections are just the roots that were found by Cardano. But how can we be sure that there isn’t 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.4). solving an equation can be visu- alized (following Descartes) as finding the point (or points) at which a curve y = x n + a n−1 x n−1 + · · · + a 0 crosses a given line (say the x axis. as given by Descartes’s rule of signs (see box 3. . see figure 4. It turns out that these roots cannot be expressed in radicals. whose three crossings of the x axis give the three real roots of the equation y = 0. y = f (x) = x 5 − 15x 2 + 9x + 3. but. and the rest. despite his immense ability and productivity (which f(x) = x3 – 15x2 + 9x + 3 4 3 2 1 x 0 1 2 Figure 4.

The last (and perhaps the most elegant) allows x to be a complex number. this means that solutions of the corresponding equations can in- volve only square roots. a seventeen-sided polygon could be constructed. when r is greater than R. with both a real and an imaginary part.” Carl Friedrich Gauss. Even before he was eighteen. and allows the coefficients also to be com- plex. we have a circle whose radius R is chosen to be so large that the equation’s behavior is dominated by the behavior of its leading term.2. Euclid’s geometry only allows the use of straightedge and compass. 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. Gauss was to give four proofs of the Fundamental Theorem over the course of his long career. as defined in figure 3. It depicts a quintic equation. expressible solely in terms of square roots. Descartes showed that. The figure shows a complex plane. Figure 4.5 shows the heart of his proof. in polar coordinates). meaning that he had found a solution of its special sixteenth- degree equation. x n (or r n . For this reason. on the contrary. he could not give a completely general result. That was left for “the prince of mathematicians. in algebraic terms. The use of complex numbers enables Gauss to ground general statements in a kind of picture. but the same considerations will apply to any equation.70 Chapter 4 continued even after he lost his sight). Gauss represents the behavior of the equation in terms of the polar coordi- nates r and θ. Gauss had made an important discovery about a special class of equa- tions that correspond to inscribing regular polygons inside a circle. This discovery confirmed his decision to devote himself to mathematics rather than philology. The first was his doctoral dissertation of 1799. The young Gauss showed that. Gauss shows . In that plane.

a double root). the imaginary part is always the opposite in sign.5 Gauss’s diagram to establish the Fundamental Theorem of Algebra.Spirals and Seashores 71 Q A P C B D Figure 4. The shaded areas are the “seas. going around that circle. Gauss draws lines showing the regions where the real part of the polynomial is positive . at those points. that one can always find such a circle. Likewise.” in which the real part of the polynomial is less than zero.” in which the real part is greater than zero. He also shows that. the unshaded areas are “land. alternately negative and positive. we find roots at D and C (here. To complete the proof. the real part of the equation takes alternately positive and nega- tive values. no matter what the equation. at A). at some point we must find a root (for instance. Walking along the “shore” from P to Q.

So let’s take a walk along the seashore from P to Q. there are also roots at B and D and a double root at C (that is. along which the real and imaginary parts will both vanish at some point. the lines show the seashore. This picture can be understood if we think of the shaded (negative) region as “sea” and the unshaded (positive) region as “land”. From Gauss’s earlier result. By his choice of examples. and that is the root. where the real part of the polynomial is zero. we know that the imagi- nary part of the equation must go from being negative to being positive as we walk along the shore. meaning that they make use of the overall features of a figure that do not depend on its de- tailed shape. it is clear that Gauss thought he was tracking the solutions of the quintic. separated from the regions where it is negative by lines indicating where the polynomial is zero. Thus. for there will always be shorelines curving in some manner. This means that at some point in the walk the imaginary part will be zero. and one can make a similar argument for any other quintic. keeping the “sea” always on our left. both the imaginary and the real parts of the polynomial must vanish. two points along the big circle. right at the water’s edge. Now if there is a solution. Here Gauss invokes arguments that later came to be called topological. This is point A in the picture. note the intersecting shorelines there). because the real part is zero all along the shore.72 Chapter 4 (unshaded in the figure). . no matter what “geography” the seas and land have. There they are. The problem was simply how to express them. He had just shown that they must exist. this particular quintic has five roots (two of them equal). two roots that happen to be equal.

In doing this. x2 . Lagrange began by asking why the solutions of the cubic and quartic had succeeded in the first place. Assume. Of course. x3 . but only that it was insoluble through this well-tried method. Lagrange noticed that. Then. “Reflections on the Algebraic Theory of Equations. this did not mean that the quintic could not be solved. Joseph-Louis Lagrange brought to light evidence that would prove crucial. and most were not daunted by the centuries-long roadblock at the quintic. 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. that we know the three roots of the cubic equation. he re- vealed an approach that would open the way to a new kind of mathematics. Mathematicians an- ticipated that all equations would have solutions. Those who came afterwards regarded his 1771 paper. As he analyzed these roots. in the late eighteenth century. then.5 Premonitions and Permutations Gauss’s full proof of the Fundamental Theorem was pre- ceded by many inconclusive attempts.” as a foundational document. . a 3 x 3 + a 2 x 2 + a 1 x + a 0 = 0. we can call them x1 . following the vision of Viète and Descartes.

the cube of the sum of the products of each of the three roots times each of the three cube roots of unity. mentioned in the preceding chapter.1). other than α = 1. We can start by factoring out a term (α − 1). Lagrange noted a simple . In the case of the quadratic equation. which are the solutions of the simple equation α 3 = 1 (see box 5. as can be verified by multi- plying it out. At this point. x3 and formed a very simple expression: R = (1x1 + αx2 + α 2 x3 )3 . x2 . then α must satisfy α n−1 + α n−2 + · · · + α 2 + α + 1 = 0. in the solution of cubic equations. a special role is played by the three cube roots of unity. This is not surprising. Lagrange then took the three roots x1 . yielding α 3 − 1 = (α − 1)(α 2 + α + 1) = 0. can be factored in a similar way.74 Chapter 5 Box 5. Gauss’s work. The equation for the nth root of unity. also showed the general importance of the equation α n = 1 in constructing an n-sided regular polygon. which are the other two cube roots of unity besides 1. yielding α n − 1 = (α − 1)(α n−1 + α n−2 + · · · + α 2 + α + 1) = 0. For α = 1. This√ is a simple quadratic equation whose solution is α = −1± −3 2 . we must then have α 2 + α + 1 = 0. If α = 1. the solution of α 2 = 1 gives +1 and −1. Clearly.1 Roots of unity We are looking for roots that satisfy α 3 − 1 = 0. the two fundamental ele- ments from which all other (positive and negative) numbers are built. α n − 1 = 0. These are solutions of the simplest possible cubic equation and thus give the basic elements of the solution of more complicated equations. 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.

Like- wise. there are just 6 = 3 × 2 × 1 different ways to permute the 3 roots. In general. (The point was made independently in 1770 by Alexandre- Théophile Vandermonde. Thus there are n! = n × (n − 1) × (n − 2) × · · · × 2 × 1 permutations. . 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. Lagrange extended this idea to higher-degree equations: There is always a quantity analogous to R that takes a certain Box 5. x2 .3 shows. 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. Our naming the roots x1 . In fact. where n! is read “n factorial” (see box 5. given n quantities. For each of these. x3 or x3 . there are (n − 2) ways to pick the third.2). there are (n − 1) ways to pick the second (51). x1 . There are n ways to pick the first one of them (52 for a deck). x2 . What this means is that we can permute the roots in various ways.2 Permutations Consider n objects. there are n! = n × (n − 1) × (n − 2) × · · · 2 × 1 permutations. .Premonitions and Permutations 75 and striking property. we could just as well switch the names around and call them x2 . . for instance the 52 cards in a deck. although there are six different permutations of the three roots.) Lagrange asked what happens to R when we permute the roots. . . As box 5. x1 without changing anything. x3 is quite arbitrary. all the way to 1 way to pick the final item. hence n(n − 1) ways to pick the first two (52 × 51). they lead to only two possible values for R.4).

3 Lagrange’s permutation of the roots of a cubic equation Consider R1 = (1x1 + αx2 + α 2 x3 )3 and make the substitution (123) x1 x2 x3 −→ x2 x3 x1 . the “identity”). the resolvent R takes on fewer values than there are permutations. (132) (132) Under the substitution x1 x2 x3 −→ x3 x1 x2 . there are only two distinct values for R under these permutations. R1 −→ R2 = (1x2 + αx3 + α 2 x1 )3 .5). As a result. which we call t1 = R 1 = R 2 = R 3 and t2 = R 4 = R 5 = R 6 . in all the soluble equations so far. yielding R5 = (1x2 + αx1 + α 2 x3 )3 . the three roots of a cubic. These roots can be permuted in 4! = 24 dif- ferent ways. There are three odd permutations left: (32). However. or (132). and this number is an important indicator of solvability. yielding R 6 = (1x3 + αx2 + α 2 x1 )3 . which can be written symbolically as (123). This reflects the fact that the equation can be reduced to solving one of lower degree. Thus. Of the six different permutations of the three roots. but under these 24 permutations R takes on only three values. zero for the identity). In the case of the quartic. R1 −→ R3 = (1x3 + αx1 + α 2 x2 )3 . R5 = R 4 .76 Chapter 5 Box 5. meaning that an even number of transpositions have been made (two for R2 and R3 . and (13). The cubic’s resolvent has two values and the quartic’s resolvent . (123) Under this substitution. yielding R 4 = (1x1 + αx3 + α 2 x2 )3 . (12). Lagrange formed a resolvent R from the four roots (box 5. these expressions turn out to be related: R2 = R1 . number of values when the roots are permuted. and R 6 = R 4 . and from there Lagrange could re-derive Ferrari’s solution. which means interchange x2 ↔ x3 . these three are even permutations (counting the possibility of leaving them unpermuted. leaving x1 alone. R3 = R1 . when the roots are shuffled around.

he found that the resolvent takes on six values. as mentioned in the previous chapter. √ 3 t2 = 3 R 4 = 1x1 + α 2 x2 + αx3 provide three linear equations √ √ that determine x1 . each one smaller than the degree of the equation being permuted. x2 . The final re- sult is identical to the del Ferro-Cardano-Tartaglia formula. x3 in terms of 3 t1 . consider again the cubic equation x 3 + a 2 x 2 + a 1 x + a 0 = 0. This was also realized independently by the Italian Gianfrancesco Malfatti (1770) and may have been noted by Leibniz much earlier. if only by steering him away from the false paths of the past.Premonitions and Permutations 77 Box 5. the “resolvent equation” mentioned in the text. Newton’s identitiesallow us to calculate 3 t1 + t2 = −2a 23 + 9a 1 a 2 − 27a 0 and t1 t2 = a 23 − 3a 1 . and he probably hoped that his work would enable him to find it. When Lagrange turned to the quintic.1) −a 2 = x1 + √ x2 + x3√and the definitions 3 t1 = 3 R1 = 1x1 + αx2 + α 2 x3 . What he did argue is that “it is very doubtful that the methods we have just discussed can give a complete solution of equations . has three. and the constant a 2 . 3 t2 . Lagrange drew from this work the moral that a new way of attacking the quintic was needed. This showed that the method that worked to solve equations up to quintic was breaking down.3. To find t1 and t2 . Girard’s identity √ (box 4.4 Lagrange’s resolvent for a cubic equation Following box 5. But he was not yet close to being convinced that the quintic is not solvable. a larger number than the degree of the equation. 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.

” Thus Lagrange still remained hopeful that new methods might succeed where the old ones failed. x2 . various sums and products of the resolvent can be expressed in terms of the coefficients of the original quartic equation. R3 = (x1 + x4 − x2 − x3 )2 . which are exactly as given by Ferrari’s formula. namely R1 . From these expressions. of degree five and. x2 . all the more so. x4 and the function R = (x1 + x2 − x3 − x4 )2 . Lagrange showed that R must satisfy a cubic equation. using Newton’s identities. R3 . R2 . For him. whose three roots are R1 .78 Chapter 5 Box 5. R2 = (x1 + x3 − x2 − x4 )2 . Having solved this cubic equation. . His contemporary. con- sider the four roots x1 . and the condition that the sum of all four roots equals −a 3 . In turn. x4 . we can recover x1 . x3 .” In the process of working out the resolvent for the quintic.” by which he means human calculators. 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.5 Lagrange’s 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. of higher degrees. As in the case of the cubic. As we go through the 24 permutations of the 4 roots. R2 . R3 . he noted that the computations are “so long and complicated that they can discourage the most intrepid calculators. the four roots can be expressed √ as simple √ √ combinations √ of these different values of R. x3 . R takes on only 3 distinct values: R1 = (x1 + x2 − x3 − x4 )2 . in that pencil and paper era.

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. Gauss published his opinion in his masterwork.” Gauss. he did not explain why the quintic is unsolvable. in this context. . would have been the last to give up had he not divined far deeper problems than those his ingenuity and persistence might solve. Disquisitiones Arithmeticae. . enclosed in its last redoubt.Premonitions and Permutations 79 the mathematical historian Jean Étienne Montucla. the problem defends itself desperately. He never seemed to be attracted to permutations and combinations. a little-known mathematician in Italy formulated and sought . Among them. In his doc- toral dissertation of 1799. It was natural that. And there is little doubt that this problem does not so much defy modern meth- ods of analysis as that it proposes the impossible. In 1801. But while Gauss guessed this new level of difficulty. knowing the extent of his own powers. as if the quintic were a town under siege: “the ramparts are raised all around but. that even before Gauss announced his opinion. which involved the question of the solvability of their equations. “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. 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. he first recorded his belief that the quintic equation has no general solution in radicals. then. his thoughts turned to the more general problem of solvabil- ity. expressed this optimistic spirit in a military metaphor. a compendium of many remarkable results. Gauss included his famous work on constructing regular polygons. . Gauss announced this possibility along with his first proof of the Fundamental Theorem of Algebra.

he would jot down his thoughts and ideas on letters he received and in the margins of papers he read. He also maintained his philosophical and religious interests. Characteristically. his mind was preoccupied with mathematics. Ruffini refused to swear a loyalty oath to the French and was barred from his university post until 1799. though the bulk of his time was devoted to medicine. Ruffini has been some- what neglected. he used his experience to write a paper on contagious typhoid. During this enforced absence from academic life. His thoughts turned to the theory of equations as he prepared his first work. medicine. but declined because he was reluctant to leave his many patients. .80 Chapter 5 to prove the assertion that the quintic is unsolvable in rad- icals. despite his deep merits and the singular interest of his dual life as physician and mathematician. and surgery. In 1817 to 1818. graduating in 1788. despite his math- ematical preoccupations. As a child. which included his first proof of the impossibility of solving the quintic. there was an epi- demic of typhoid fever. Perhaps because of this modesty. during which Ruffini tended to the sick with great dedication and courage. His contemporaries noted his self-effacing mod- esty and spirituality. He himself fell ill with the disease. At the end of the day. Though not a revolutionary himself. in later life Ruffini was offered a mathematical post at the University of Padua. but he finally chose to study philosophy. The son of a physician. writing about the definition of life and against Laplace’s mechanistic theory of morality. Ruffini’s devotion to medicine was deep. He served as a professor of mathematics in Modena until the upheavals in the wake of the French Revolution hit Italy. Indeed. he seemed destined for the priesthood. Paolo Ruffini was born in a small Italian town. moving in childhood to Reggio near the city of Modena (figure 5.1).

.1 Paolo Ruffini.Premonitions and Permutations 81 Figure 5.

several of them responding to questions and criti- cisms raised both by his friend Pietro Abbati and by the older mathematician Malfatti. though of a French family) as a fellow countryman who they thought should recognize this Italian achievement. Lagrange was on a committee to examine Ruffini’s memoir. He sent his proof to Lagrange. a bitter disappointment to Ruffini and his friends. for they considered Lagrange (born in Turin. 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 . finding “little in it worthy of attention. with his sublime reflections. the last in 1813).” As it later emerged. hoping for recognition from the man whose work had given him his inspiration.” Indeed. Later. The immortal Lagrange. Unfortunately. Lagrange never replied. Ruffini expressed these proofs in long and obscure forms. who could not accept the notion of unsolvability. Ruffini was among the first to grasp the immense importance of Lagrange’s idea of “permutations” (as Ruffini called them) and was the first to offer a proof of the unsolvability of the quintic. 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). which rendered them hard to understand. to present the proof of it is the main reason for publishing this volume. but he was cool toward it.82 Chapter 5 Though he did not seem to covet the usual rewards of pro- fessional glory. has pro- vided the basis of my proof. 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. however.

in my judgment. Thus Ruffini’s proof fell short. be solvable in radicals. later mathematicians confirmed Lagrange’s con- tention that there was an important missing step in Ruffini’s arguments.” Cauchy also told Ruffini that he had cited this work in his lectures. Indeed. Despite Cauchy’s strong endorsement.Premonitions and Permutations 83 of its assertions. in general. six months before his own death. . One can imagine Ruffini’s gratification when. 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. To be sure. Ruffini’s work still did not achieve general acceptance among mathematicians. Most important. the great French mathematician Augustin Cauchy recognized the importance of Ruffini’s work and generalized some of its results during 1813 to 1815 (as will be discussed later and in appendix C). Other mathematicians were more favorable. This might have been behind Lagrange’s wari- ness. He deserves credit for his vision and his courage. for he described Ruffini’s proof as a “difficult matter. 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. but it remained for others to resolve the issue decisively. he may not have spent much time studying Ruffini’s proof. and for which he thought Ruffini’s reasoning insufficient. Some read- ers at the Royal Society were “quite satisfied” with Ruffini’s proof. at least in the sense that it left others unconvinced.” something that he himself thought improbable. if not incredi- ble. proves com- pletely the unsolvability of the general equation of degree greater than 4.

This page intentionally left blank .

Abel’s father was a pastor.6 Abel’s Proof Ideas. like his father before him. When Abel was 12. Karl XIII. But Norway could not com- pletely separate from her more powerful neighbors. Without knowing Ruffini’s work. though struggling constantly with poverty and misunderstanding. especially mathematical ideas. have a life that goes be- yond their human creators. he discovered whole new territories in mathematics. Abel’s story is one of the most moving in the history of mathematics. Norway separated from Denmark. His native Norway was then in its youth as an independent nation. He was a man of the Enlightenment who read Voltaire and was active in the movement for literacy and vaccination in rural Norway. and Abel’s checkered fortunes to some extent mir- rored Norway’s struggle. Abel’s father was among the delegates sent to offer the crown of Norway to the reigning Swedish king. He wrote several popular catechisms and books of prayer . long the dominant partner in their dual kingdom. the Storting. and set up her own parliament. In his twenty-six years of life. In 1814. and had a notable career as a member of the Storting.1) discovered the same argument only a few years later and also gave the first essentially com- plete demonstration. Niels Henrik Abel (figure 6.

1 Niels Henrik Abel.86 Chapter 6 Figure 6. .

) Holmboe rec- ognized and extolled his student’s “excellent mathematical genius. which had opened in 1819 in the new capital. Abel during this time set himself to solve the most difficult and famous problems. though in fact he had little to learn even from the greatest mathematicians of the time. leaving nine children and a widow who also turned to alcohol for solace. He had been fortunate to find a mentor in a young mathematics teacher. In 1823. though. However. who inspired him and became his lifelong friend. Berndt Michael Holmboe. Somehow. and c that would . his political career ended ignominiously when he pressed unfounded accusations against certain powerful persons. found himself without support and obliged to act as the responsible adult for his younger sib- lings. he read Lagrange’s work and Cauchy’s 1815 pa- per on permutations. Abel soon showed an amazing ability to solve difficult problems. he tried his hand at Fermat’s celebrated last theorem. After his funeral. then a city with only 11. He was interested in natural sci- ence and woke his children to see lunar eclipses. Abel did not read Ruffini. While always modest.” In 1821. perhaps because his works were hard to find and in Italian. she received visiting clergy while in bed with her peasant paramour. b. Abel.Abel’s Proof 87 doubtless read by his son. he continued his education. those teachers proposed that he be granted a special fellowship to visit Paris and Berlin.000 inhabitants. then eighteen. Christiana (now called Oslo). He continued to make rapid strides. Abel entered the Royal Frederick’s Univer- sity. While still a high school student. beginning to write orig- inal papers and going far beyond the skill of his teachers. He died an alcoholic. to show the impossibility of finding integers a . (Though Cauchy’s paper was based on Ruffini’s work. With admirable understanding.

which in many ways is close to Ruffini’s proof. He never explained his reasons. two years after Ruffini’s death.” which showed that if any solutions existed. Instead. At this point. for this problem resisted solution until 1993. Gauss offered no proof for his assertion.88 Chapter 6 satisfy the equation a n = b n + c n . This was a decisive moment. he found himself “at the end of my tether. he nonetheless refused to give up the search for a solution. Even so. Abel soon realized that it was not as general as he had thought. he made an about-face and turned his ef- forts to proving the unsolvability of the quintic. could well consider that the search was not over. the Oslo fjord remained frozen and mail was often delayed. and one wonders what moved him. Not surprisingly. Abel read . twenty-two years after Ruffini had pub- lished the first version of his proof. This was in 1821. which is not surprising consid- ering the isolation of Norway and its lack of mathematical libraries. when his teachers asked him to give some numerical examples of his solution to the quintic. during the winter months. and Abel. Indeed. At first. Abel also turned his mind to the problem of the quintic equation. although it fills in an im- portant gap that Ruffini had not noticed. After all. Abel still did not know Ruffini’s work. like Lagrange and most other mathematicians. In 1826. when it finally succumbed to very elaborate abstract techniques. Abel published his first proof of the unsolvability of the quintic. for he could have stubbornly resumed the quest for a fully general solution to repair the gaps in his work.” as he put it. he found what now are called “Abel’s formulas. they would have to be extremely large numbers. However. on the assumption that a solution had to exist. he thought he had managed to solve it and was very excited. where n is an integer greater than 2. In 1824. Though Abel had probably taken note of Gauss’s opinion that the quintic was unsolvable.

However. I will summarize Abel’s version. Accordingly. If you wish to read Abel’s own words. But his paper is so complicated that it is very difficult to decide the correct- ness of his reasoning. which he amplified in his later accounts of the proof. as well as Lagrange’s hesitation. and if I am not mistaken the only one. In each case. Appendixes B and C fill out details he merely mentions. Gauss above all. Abel had this earliest version of the proof printed at his own expense as a pamphlet. Here I will present the four crucial stages of the proof as Abel presented them in 1826. The theorem of un- solvability may better be called the Abel-Ruffini Theorem. indicate aspects of decisive proof in which Ruffini fell short. Accordingly. which Abel acknowledged in his final (posthumous) paper: “The first. indicating along the way the common elements in their arguments and the crucial gap in Ruffini’s reasoning that Abel filled. . appendix A contains a translation of his 1824 paper.” Because Ruffini’s work overlaps largely with what Abel went on to do. Yet Abel’s remark. followed by my explanation. I do not mean to belittle the recognition Ruffini is due.Abel’s Proof 89 anonymous articles summarizing the work of Ruffini. It seems to me that his reasoning is not always satisfactory. without spelling out all the technicalities to be found in the appendixes. who before me had tried to show the impossibility of the algebraic solution of general equations is the geometer Ruffini. I have added a running commentary to help the reader follow Abel’s argument. I will not discuss it separately. A proof that lacks a decisive step is not yet a proof. I will present a summary statement. Abel compressed his argument to tele- graphic terseness. to save paper and money. hoping to use it as a “calling card” that would gain the attention of the great mathematicians.

. Given a general equation of the mth degree (taking m as a prime number. which is just the general form above with m = 2 and p a simple polynomial. are finite sums of radicals and polynomials and 1 R m is in general an irrational function of the coefficients of the original equation. but the idea can readily be illustrated with a few examples.) This is very close to the general form that Euler had already conjectured some years before. As box 6. That is. which follows the general form. a m ym + a m−1 ym−1 + a m−2 ym−2 + · · · + a 2 y2 + a 1 y + a 0 = 0. the solution of the cubic can be expressed as y = 1 2 p + R 3 + p2 R 3 . (Remember √ that R m is just another way of writing the mth root of R. His first step is to specify the form that a solution must have. (6. the solution of the quadratic 1 equation can be expressed as y = p + R 2 . m = 4 is not prime and the general solution . For the quartic. .1 shows. Abel gives a detailed proof (see appendix B. y can be expressed as a series of terms that contain nested roots involving the coefficients and irrational expressions like 1 1 R m .2) where p. m is a prime number. so that it cannot be factored further). m = 2 and m = 3. In both of these cases. m R. p2 . including some subtleties neglected here). with m = 3. .1) Abel proves a very general statement about any solution. if y is the solution of an algebraic equation of degree m. Likewise. (6. which he calls an “algebraic function”: (I) All algebraic functions y can be expressed in the form 1 2 m−1 y = p + R m + p2 R m + · · · + pm−1 R m .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.

But Abel brilliantly proves that we can express . involves only combinations of the quadratic and cubic forms. (6.1 Abel’s form for the quadratic equation In the quadratic equation. substitute y = 1 1 p + R 2 to yield ( p 2 + R − a 1 p + a 0 ) + (2 p − a 1 )R 2 = 0. So Abel assumes hypothetically that the quintic does have a solution of exactly this form. R 5 . R 5 ) of the coefficients. R 5 . 174 in appendix B. y = a21 ± 12 a 12 − 4a 0 . the fami- liar form of the quadratic solution. or there is no such solution.3). each parenthesis must be separately zero (as Abel discusses in [A8] in appendix A). Ruffini had assumed it without giving a proof. He now goes on to show that this form leads to a contradiction. see p. but Abel remedies this.2) becomes 1 2 3 4 y = p + R 5 + p2 R 5 + p3 R 5 + p4 R 5 . R 5 . The next step is crucial. In the case of the quintic. For the case of the cubic solution.3) Because this follows from his general result about the form of the solution. so p = a21 and a2  then R = −a 0 + 41 . y2 − a 1 y + a 0 = 0. Abel’s general form (6. the general form of y is expressed in terms of polynomials and 1 2 3 4 also various irrational functions (here. as was discussed earlier. In equation (6. To satisfy this in general. (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.Abel’s Proof 91 Box 6. Thus. This requires three further steps. either the solution of the quintic has this form.

this is indeed a simple rational function of the roots. is equal to the difference of the two roots. Here there is an interesting tension: The roots are irrational functions of the coefficients. (y1 − y2 ). whose general solution is y = p + R 2 . Now consider the two roots. y in terms of the roots instead of the original coefficients of the equation. positive when the roots are real. 2 The name discriminant is given to a 12 − 4a 0 . R 5 . y2 are the two roots 1 of the quadratic. which is zero when the roots are equal.1.. which also shows a2 a that R = −a 0 + 41 . . Because it is made up of products of all these rational functions. y1 = 21 +   1 a 12 − 4a 0 . Abel’s step II implies that R 5 is a 2 3 4 rational function of the roots. as well as p. y1 = y2 . To choose a simple example. . box 6.2 shows that 4R = (y1 − y2 )2 . as shown in box 6.2 The relation between roots and coefficients 1 Let y = p + R 2 . p2 . R 5 . What is important here is that all the various irrational functions of the coefficients that appear in y 1 (e. R 5 . In agreement with Abel’s proof. which showed that the coefficients were sums and products of the roots. Then (y1 − y2 ) =  2 2 2  a2  a 12 − 4a 0 and thus (y1 − y2 ) = a 12 −4a 0 = 4 −a 0 + 41 = 4R. . in the quadratic equation 1 above. y2 = a1 − 1 a 12 − 4a 0 . R 5 ) are rational functions of the roots of the equation.g. but those coefficients are always rational sums or products of the roots.1). 1 In the case of the quintic. and negative when they are imaginary. . So 2R 2 .92 Chapter 6 Box 6. y is thus a rational function of the roots. This harks back to Girard’s identities (see box 4. where y1 . as are its powers. . the square root of 4R.

So there can be no more than five values for y. on the one hand. or one value. This gives Abel leverage to show that the solution 1 cannot work. but never three or four values. Roughly speaking. not five roots. it is harder and harder to reconcile this tension. find a solution in radicals. this idea must be amplified much further. as the degree of the equation grows higher. As before. with the quintic equation. two. in general. on the other. (III) If a rational function of five quantities takes fewer than five values when the five quantities are permuted. and we cannot. take only one value. or five values as the roots are permuted. Then. it can take only two different values (equal in magnitude and opposite in sign). let us consider any of the expressions in the 1 solution that is a rational function of the roots. 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. The next step limits the hypothetical solution in a way that will prove to be decisive. note that R 5 cannot. Abel drew this theorem from the work of Cauchy. Finally. First. such as R 5 . 1 Abel investigates what happens if R 5 takes on five values . it is no longer possible. which we have assumed to be unequal. of which it is a special case. according to step II. Cauchy’s proof is presented in appendix C. but never three or four values. 1 Cauchy’s result requires that R 5 can take only one. is a rational function of the roots of the equation. However.Abel’s Proof 93 Let us return to the tension between the irrationality of the roots as functions of the coefficients. and the rationality of the coefficients as products of the roots. in general. It relies on looking at the different ways we can permute the roots of the equation. which. because then it could lead to a single solution.

which is impossible. he 1 shows that R 5 . For instance. This contradiction rests on a special property of the number five. Clearly. shown by the number of values the hypothetical solution can take when subjected to permu- tations of its five roots. . The strategy of Abel’s 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. Therefore Abel concludes that (IV) It is impossible to solve the general equation of the fifth degree in radicals. The argument also applies for degrees higher than five. Though Abel’s argument shows the impossibility of solv- ing the quintic in general. [A15]–[A16]). and it would then be a sixth- degree equation that has one root y = 0 and five unsolvable roots. So Cauchy’s result must require that there be only two (appendix A. having five possible values when the roots are permuted. This leads to a contradiction. The ques- tion remains: why this impossibility? To examine this further means we must seek the heart of Abel’s proof. it still seems opaque. because when we switch the five roots around.94 Chapter 6 (appendix A. He derives an equa- tion whose left-hand side has 120 possible values. would have to be equal to an expression having 120 values. [A27]). while the right-hand side has only 10. and similarly for any higher degree. But this too leads to a contradiction. Abel shows that we get an inconsistent result again. such an equation cannot be solved in general. we can multiply an unsolvable quintic by a factor of y = y − 0. That excludes the possibility that there are five values for y. and so the hypothetical solution leads to absurdities.

Abel did meet August Crelle. For this reason he did not go directly to Paris. only the stipend given him by his university made such a dream voyage possible. fellow Norwegians also pursuing studies abroad. Crelle was hospitable to Abel and soon recognized his new friend’s extraordinary gifts.7 Abel and Galois Abel published his proof shortly before he began his travels abroad. He published seven of Abel’s papers in the first volume of his journal and also began working to find . He hoped to meet Gauss. He knew that this trip was a godsend. 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 Crelle’s Journal by its readers. but headed rather to Berlin. he decided to travel with friends. Because he was shy and prone to loneliness. turned inward on his own concerns. but the great man received very few visitors. He was so poor that he barely scraped by at home. setting it aside as if it were from a crank: “Here is another of those monstrosities!” In Berlin. He received Abel’s proof but did not cut its pages. his only chance to make a mark on the larger world. the mathematical center of the world at that time. hoping that it would open doors for him.

. Despite the requirement that he spend the bulk of his time in Paris. well qualified to appreciate Abel’s achieve- ments and promote his career. without bothering others. he had a premonition that this would be his only chance to see the world. . he could meet Adrien-Marie Legendre and Cauchy. as he spread his wings ever further. even though the trip strained his precarious finances to the limit. this attention buoyed his hopes that he might secure a living as a mathematician that would allow him to marry his fiancée. Everyone works by himself here. who. Christine Kemp. . For Abel. Though he was then in good health and young. an important journal of the day.” Though Abel was disappointed. He did make an impression on some Frenchmen.96 Chapter 7 him a position in Berlin. asked . There. Abel took leave of his friends and steeled himself to face Paris alone. 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 . had no means and remained in Norway working as a governess. They paid no attention to his proof. Everyone wants to teach and no one wants to learn. un- commonly reserved with respect to foreigners. His reception was chilly. like Abel. . . Abel could not resist accompanying his friends on their journeys to Italy and Switzerland. eminent members of the Académie des Sciences. he kept working diligently and continued to send Crelle more and more papers. after all. His formal visits to the great men were received politely but with- out real warmth or recognition. the mathematical editor of Baron de Férrusac’s Bulletin. Jacques Frédéric Saigey. At last. at least. His letters pleaded with his professors in Norway not to disapprove of his desire to see these fabled places.

seeking its larger implications. for example. He played billiards and indulged his pas- sion for the theater. he had not stopped thinking about his proof of unsolvability. Abel also met an important self-taught scientist. Nevertheless. Such are. Despite his loneliness and dwindling funds. he had made steps toward the development of the cell theory of biology and the notion of microbe-borne disease. and I have come to the result that . ironically by his teacher and mentor. He hoped to find a post in Norway. which he described in animated letters. Raspail was quite struck by Abel. In March 1828. At the end of 1826.” which Gauss had solved so brilliantly when younger than Abel. Abel wrote that “Although the alge- braic solution of equations is not possible in general. François-Vincent Raspail. In his scientific work. “The solution of these equations is founded on certain relations among the roots. there are nevertheless particular equations of all degrees that ad- mit such solutions. In a new paper he sent to Crelle. Abel found diversion in Paris. his funds ran out and he had to return home months early. 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. but he must have been deeply perplexed. but the only vacancy in the whole country had been filled during his absence. the equations of the form x n − 1 = 0. and their friendship continued. He could not think of mar- rying in such penury. an ardent par- tisan of the French Republic who was several times exiled or imprisoned. in- cluding his unsolvability proof for the quintic. Holmboe. he came to a crucial realization. Abel was not bitter. he had to eke out his living as a substi- tute teacher in a military academy.Abel and Galois 97 Abel to write brief accounts of articles in other journals. It was at this point (1826) that Abel became aware of Ruffini’s work and noted its unsatisfactory quality.

in these solvable equations. x6 = f 5 (x1 ) = x1 . If this is true. meaning the function f applied two successive times.” as they are now known. f (x1 ). He then keeps applying f to each successive root and gets the series of roots x1 . f n−1 (x1 ). . He calls the first root x1 . where f (x1 ) is a rational function of x1 . then . if n = 5. . then he assumes that the next root x2 = f (x1 ). . . f 3 (x1 ).” Abel gives this interrelation of the roots a simple form. which he calls f 2 (x1 ). which one can arrange in a circle. where n is the degree of the equation. For instance.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. X1 X5 X2 X4 X3 showing the cyclical symmetry of the roots of such solvable “abelian equations. any two roots are related by a rational function. since the equation can have only five different roots. He then as- sumes that the next root is the same function of the previous root. x3 = f (x2 ) = f ( f (x1 )). Abel’s insight is that. f 2 (x1 ).

then the equation is solvable. especially in the con- text of simple equations where we would least expect such . Thus. Indeed. however. it may seem merely formal or devoid of significance. the equation may not be solvable if the order does matter.Abel and Galois 99 all the roots are rational functions of each other and of the coefficients of the given equation. and it is crucial. Contrariwise. Until this point. This is the insight that I consider most helpful. a + b = b + a and a × b = b × a . We need one further con- dition. It is implicit in the case we have considered. he opened the whole issue of commutativity for consideration. Abel’s insight connects solvability with commutativity. But a great surprise is hidden in this seem- ingly flat statement. none of the basic opera- tions of arithmetic and algebra has been “noncommutative. so that the roots can be grouped into different cycles if the equation is solvable. These so-called com- mutative laws express an important quality of numbers. and they hold sway in the operations used in every equation. which are two (possibly different) rational functions of x. I believe that this was the first time that the possibility that operations might not commute emerged in mathematics. of whatever degree.” Thus. if the order in which these two functions are applied to x does not matter. That is. then Abel concludes that the equation is always solvable if g( f (x)) = f (g(x)). at least for abelian equations. but becomes explicit when Abel con- siders an equation whose degree can be factored into several primes. 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. At first. If x is a root of such an equation and two other roots are given by f (x) and g(x).

for it does not matter whether we go around the circle backward or forward. Gauss now spoke of the “depth. At age twenty-three. which he signed “Yours destroyed. and it may be that Abel heard of this through Crelle. who was vying with Abel to extend his results still further. He was able to organize the roots of abelian equations into the cyclic pattern shown above. Abel must have seen that he needed to take it much further. who described himself as being “as poor as a churchmouse” in a note begging an old friend for a loan. In the rest of this book. Indeed. Although he was not given to undue enthusiasm. I hope to follow this story and explore its implications.” . Though Abel did not live to see it. Abel’s brief strand of life ran out. Jacobi was already an associate professor at the University of Königsberg. both Jacobi and Abel were proposed for membership in the Institut de France in 1828. Crelle was working hard to find a place for Abel. for Norway still remained isolated. though neither was elected. Just at this point. It is clear that in such cases the relations between the roots are commutative. Yet this insight is only a beginning. Abel emphasizes this com- mutative symmetry as the crux of this pattern. first of all. He probably did not yet know of the admiration of Legendre or Carl Gustav Jacob Jacobi. for it does not explain how to determine whether the pattern of a given equation is or is not cyclic in this way. and elegance” of Abel’s work. how Abel’s insight into solvable equa- tions squares with this. Notice. His 1828 paper leaves us wondering how he might have carried his question about commutativity even further. delicacy. however. He was ill at the beginning of 1828 and did not know how widely his papers were being read. his suc- cessors gradually extended and elucidated the subtle con- nection between noncommutativity and unsolvability.100 Chapter 7 a strange thing.

this lack of friends is not foremost in my mind because I have so horribly much to do for [Crelle’s] Journal. He began to receive some letters from Legendre that comforted him by their frank admiration and interest in his work. however. almost completely alone. 1828. He was in- creasingly ill. and Abel probably never knew of this attempt to intervene on his behalf.Abel and Galois 101 Crelle was not the only one trying to help Abel. He also wrote: “I am. Nevertheless. Abel could not resist the opportu- nity to spend the holiday with his fiancée in the country. Monsieur Abel. Generously. with only socks to warm his hands. He set off. Four dis- tinguished French mathematicians (including Legendre and Siméon-Denis Poisson) wrote to the king of Sweden implor- ing some help for one of his own subjects.” At Christmas. he took to his bed with . 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). Abel quoted such praise to his friends. not just on the solv- ability of equations but also on elliptic functions and what now are called Abelian integrals. After the Christmas balls and festivities. I assure you that in the most profound sense I am not in asso- ciation with a single human being. Abel continued to work. making fundamental con- tributions.” The king never responded. “a young math- ematician. For his part. Abel wrote back that Legendre’s recognition gave him “one of the happiest moments” of his life. but his friends could not dissuade him from going. “for I would have ac- complished nothing without having been led by your light.” With childlike pride. even though the winter was exceptionally cold. 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.

He died on April 6. During the nights. Abel died at twenty-six. just south of Paris. his anger and despair fright- ened those around him. After Napoleon’s first . On his deathbed.102 Chapter 7 pneumonia. Abel’s new insight of 1828 appeared in print only after his death. 1829. but Galois died at only twenty years old. Abel had sent word to his friend Keilhau urging him to marry Christine after his death. and biographies. Though he had never met her.1) stands beside that of Abel among the most dramatic lives in mathematics. a young Frenchman had already taken the steps that would generalize Abel’s insight into a full account of the solvability of equations. however. but still made light of his condition. Galois’s story is. not knowing of his friend’s death. His family had adopted the ideals of the Revolution of 1789. But the matter was not completely resolved. Crelle wrote him that he was almost sure of having secured him a position in Berlin. in a mysterious duel linked to love and politics. He went on to complete one more paper on tran- scendental functions. which would allow him financial security and the possibility of marriage at last. far better known. Abel kept denying that he was dying of tuberculosis. and without knowing his final work. elaborated and romanticized in novels. films. Two days later. in the presence of his fiancée Christine. Thus. The story of Évariste Galois (figure 7. Galois was born in 1811 in Bourg-la-Reine. Crelle sent final word that Abel’s appoint- ment was approved. Yet even before Abel’s death. We can only speculate where he might have taken these ideas. He cursed his poverty and the in- justice of the neglect he had suffered. he was brave. which ended more than eight hundred years of rule by the Bourbon-Capetian dynasty. Keilhau proposed in a letter and they were soon married. During his last days. in order to record part of the treatise lost in Paris.

1 Évariste Galois. .Abel and Galois 103 Figure 7.

who presented one of Galois’s papers to the Académie des Sciences in 1829. Though shy. Yet the whole story is sadly reminiscent of Abel’s disappointment in . It was a crushing disappointment when that judgment never came. He approached Cauchy. indeed. to study in silence for two hours before a breakfast of bread and wa- ter. to be reviewed by himself and other distinguished members.M. and Louis XVIII took the throne vacated by his executed brother. Often punished. Galois blamed Cauchy for losing the manuscript. they were put in solitary cells for days at a time. a position he held for fifteen years. Nonethe- less.104 Chapter 7 downfall and exile to Elba. but some documents sug- gest instead that Cauchy urged Galois to revise and expand his work. Galois might well have been somewhat paranoid. The mathematician Sophie Germain describes Galois attending the sessions of the Académie and insulting the speakers. Richard went out of his way to seek oppor- tunities for his amazing student. At age eleven. Galois never even got his manuscript back. By now Louis XVIII had returned to power. Like Abel. The romantic myth of misunderstood genius ignores such provocative and probably outrageous behavior. the school was as grim as a prison. Louis XVI. Galois’s father became the mayor of Bourg-la-Reine. and the school’s administration was rigidly royalist. Évariste entered the famous Parisian Collège de Louis-le-Grand as a boarding student. When Napoleon attempted a comeback from Elba. Despite its great name. The students rose at 5:30 A. first Hippolyte Jean Vernier and then Louis-Paul-Émile Richard. Galois made brilliant progress and won prizes in Greek and Latin. even after Napoleon’s defeat at Waterloo in 1815. and stories told after his death do reveal a darker side. Galois was helped by his teachers. France turned back to its royalist past. for unclear reasons. however.

though Galois was not aware of Abel’s work as he formed his own theory. Galois was able to gain admission to the École Préparatoire (now called the École Normale Superieure). That summer.”) Though they were in Paris at the same time. Charles X. Those years were terrible for Galois. In 1830. a moderate like his friend George Washington. During them. was even more desirous of a return to the ancien régime. saying that the answers were completely obvious. Galois’s father felt forced to leave the town and was so humiliated that he com- mitted suicide in 1829. the duc d’Orléans. In July. Galois was ques- tioned on logarithmic series. His obstinacy ended his dream of being a polytechnicien. and the clerical party was also gaining as- cendancy. Through the intervention of Richard. he took the entrance examinations for the presti- gious École Polytechnique. allowed Charles’s dy- nastic rival. In their little town. Galois wanted to join in the fight but was kept locked up in the school. the Jesuit parish priest forged Galois’s father’s name to scurrilous epigrams directed to his own relatives. The grieving Évariste saw the parish priest insulted and hit with stones at his father’s funeral. driven from Paris. unknowing. In the ensuing scandal. he refused.Abel and Galois 105 Paris in 1826 to 1827. rioting began in Paris against the government. in the street. Charles X. though we might imagine them passing each other. When his examiners asked him to prove his statements. (Abel described Cauchy as “a Catholic bigot. it does not seem that Abel and Galois ever met. was un- able to regain power. The old General Lafayette. to become Louis-Philippe . Certainly they were pondering the same questions about the solvability of equations in 1828. when his submission to the Académie also disappeared. where he did not bother to hide his scorn for his teachers and studied only mathematics. Louis XVIII’s suc- cessor.

the youngest prisoners were transferred on parole to a clinic. who later described how Galois. who still looked like a child. Raspail remembered that Galois had to be restrained from a drunken attempt at suicide. Galois escaped a bullet fired by a guard into his crowded cell. Galois fell in love with Stéphanie Poterin-Dumotel. infuriating him even further. along with Raspail. On another.” Though acquitted of the charge of treason. He sent yet another paper outlin- ing his theory of equations to the Académie. which this time rejected it. for so many of their leaders were in prison. During a banquet of the Société. Galois soon was put in prison on other charges. In 1832. Along with Raspail. Two fragmentary letters indicate that. Galois joined the radical Société des Amis du Peuple and became more active politically.” Galois and many others considered this “July monarchy” a grotesque betrayal of their struggle to restore the Republic. Taunted by the other inmates. the daughter of one of the doctors. he wrote: “Happiness and hope are at an end. kept working on his mathematics even there. Galois tried to show he could hold his liquor. he was not heard as he continued: “if he betrays his oaths. the republi- cans were unable to organize further demonstrations. To a friend. His situation then became even more strained.” whom he blamed for his own neglect. When an outbreak of cholera struck Paris. the “peoples’ king.106 Chapter 7 I. and Galois in prison vented his anger about “the very men who already have Abel’s death on their conscience. There. Galois raised his jackknife and sarcastically toasted “To Louis-Philippe!” In the commotion. He tried to eke out a living by giving private lessons in mathematics. On one such occasion. she finally declined to accept his love. the scandal about the Institut’s treatment of Abel came out. That year. though initially encourag- ing. for which he was finally punished by be- ing expelled from school. now surely con- .

though it leaves much un- explained: “I die the victim of an infamous coquette and two of her dupes. After being released from prison. his love rejected. but because of their old friendship they could not bear to look at one another and left the decision to blind fate.” Consider also this account of the duel given by a newspaper in Lyon: “The young Évariste Galois . never! Hatred. It is said that love was the cause of the combat. At point-blank range. . but do not permit a decisive solution. Yet the mystery remains.” indicating that he died for purely personal reasons.” His final letters con- tain his most explicit testimony.” If this newspaper account is accurate. used Russian roulette to decide who would sur- vive. like himself a member of the Société des Amis du Peuple. the full truth will probably never be clear. I go to the .Abel and Galois 107 sumed for the rest of my life. But to the end I told the truth. He died only a few weeks af- ter this letter in a duel whose motives and details remain obscure. Galois wrote: “Pity. Only one of the pistols was loaded. two comrades in the same secret political society. The surviving documents contain tantalizing evidence. that’s all. and who was known to have figured equally in a political trial. . The pistol was the chosen weapon of the adversaries. others politi- cal.” The legend of Galois comes from the succeeding events. Galois begged his comrades “not to reproach me for dying otherwise than for my country. In his final letters. both involved with the same woman. was fighting with one of his old friends. Galois wrote that “only under compulsion and force have I yielded to a provocation which I have tried to avert by every means. a young man like himself.” He reproached himself with hav- ing told “the hateful truth to those who could not listen to it with dispassion. each of them was given a pistol and fired. Some writers ascribe romantic causes.

I need all my courage to die at twenty. the general uprising was postponed a few days until what seemed a more auspicious occasion. including a letter dated on the eve of his duel. though the last letter contains indications of the larger sweep of his ideas. What is new in Galois is a turn toward abstraction in an essentially modern way. Again. for they are of good faith. he “had no time” for that. This testament became the centerpiece of his legend. But his life was the stuff of legend.108 Chapter 7 grave with a conscience free from patriots’ blood. Here the hard evidence ends. Indeed. the hateful truth he spoke? We do not know. many of Galois’s ideas were already present in his earlier published works (he published five papers in his life- time) and unpublished papers.” he wrote in the margin. “I have no time. In 1870. If it was meant for political purposes. the funeral of a prominent general appointed by Napoleon. the truth is less romantic. I would like to have given my life for the public good. He was often too impatient to spell out the tedious details of proofs. Forgive those who kill me.” At his funeral. as if Galois had feverishly scrawled his great theory on the edge of death.” What was the provo- cation to which he yielded. But at the last moment. His last words were to his brother: “Don’t cry. the tragic epitome of doomed genius. three thousand people were ready to attack the police. Yet al- ready in the 1840s mathematicians began to read and study Galois’s posthumous writings. leading to a complete understanding of solv- . Galois’s work agrees entirely with Abel’s results. 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. Galois’s death was in vain.

It remains a great question to what extent this gap can be made intelligible. 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. In so doing.Abel and Galois 109 ability. however. with all its generality and power. As he reformulated and extended Abel’s work. he opened a crucial gap between the new language and common under- standing. which Abel lacked. Galois was the first to speak of these per- mutations as instances of a new kind of mathematical object: the group. . That which empowered modern mathematics left many highly educated persons clueless. much less bridged. At the same time. he opened the way to the characteris- tic language and projects of modern mathematics.

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 Abel’s 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:

1

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.)

consider first exchanging (12) and then exchang- ing (13): 1 (12) 2 (13) 2 3 2 3 1 1 3 .Seeing Symmetries 115 There are two other possibilities: (12) means that 3 is still.) Since each exchange is an odd permutation. (The identity may be considered an even permutation. we can put together two of them to make an even permutation. since it involves zero pairs. while 1 and 3 change places: 1 3 (13) 3 2 1 2 Note that each exchange is an odd permutation. For instance. while 1 and 2 change places: 1 2 (12) 3 2 3 1 and finally (13) means 2 is still. since it in- volves transposing only one pair of dancers.

2 stands where 1 used to stand). Likewise. so that 1 becomes 3. meaning that the dancers have rotated by one-third of a complete circle (120◦ ).116 Chapter 8 The result of these two sequential exchanges is the same as a rotation (123). Note that the rotations involve a single axis standing vertically at the center of the triangle. and 2 becomes 1: 1 3 (132) 3 2 2 1 . so that 1 becomes 2 (that is. we see: 1 2 (123) 3 2 1 3 For simplicity. they can only rotate by 120◦ (one-third of a complete circle) or multiples of it. 3 becomes 2. and 3 becomes 1. We have just described a single shift of 120◦ by (123). like a maypole. we record the angle of rotations in terms of counterclockwise motion. In order to obey the rule that the dancers always return to the original triangular formation. 2 becomes 3. (132) will shift the dancers over by 240◦ . Looking at them from above.

. returning us to the identity: (123) ∗ (123) ∗ (123) = (132) ∗ (123) = (1): 1 (132) 3 (123) 1 3 2 2 1 3 2 Thus. Thus. (13). (1). two successive rotations of 120◦ are the same as one rotation of 240◦ . and the three exchanges (12). . So the dance S3 has six steps: the identity (1). the rotations of the three dancers are cyclic: the basic rotation (123) repeated over and over gives a “3-cycle”: (123). . Notice also that these cyclic rotations by themselves are abelian: whatever the order in which you do them yields the same result. (132). (1). (123).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◦ . . (132). 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◦ . (23). . the two rotations (123) and (132).

2. Notice that this set of rotations is “closed” in the sense that these rotations stay within themselves: two sequential rotations always give another rotation. contain- ing both odd and even permutations.118 Chapter 8 S3 is more complex than S2 because S3 is not commutative. Of course. whereas exchanges are odd. you get (23): 1 3 1 (13) (123) 3 2 1 2 2 3 Thus. if you rotate (123) and then exchange (13). and which are rotations with- out exchanges. For instance. S3 is nonabelian: When you mix rotations and ex- changes. which we will call A3 (for “alternating”). the result is not commutative. never an exchange. So from the whole nonabelian dance S3 . shown shaded in table 8. we can single out just those that are even. you get (12): 1 (123) 2 (13) 2 3 2 1 3 3 1 But if you first exchange (13) and then rotate (123). Note also that ro- tations are always even permutations. we must include the identity (1) among .

If we look within A3 . Here is a symbolic way of writing down this interrelation- ship of the dances: S3  A3  I. which in turn contains only the identity. (132). The shaded entries show the abelian subgroup A3 . The chain goes from the “largest. We will call A3 an “invariant subgroup” of S3 . (123) ∗ (13) = (12). but (13) ∗ (123) = (23). For instance. it involves all other steps of the same kind—here the rotations (123). which simply means: S3 contains the invariant subgroup A3 . I = (1). to its . For instance. shown by the lack of symmetry about the main diagonal.Seeing Symmetries 119 Table 8. and (1). 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. the rotations. which is symmetric about the diagonal.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. (1) and (12) are a subgroup but not in- variant because the other exchanges (13) and (23) are not included.” most inclusive symmetry. we find no other such invariant subgroup except the identity itself. since it corresponds to a rotation by 0◦ (or any multiple of 360◦ ).

then the next largest subsubsymmetry. Solving a cubic equation corresponds to finding an invariant subsymmetry (A3 ) within the next larger symmetry of the equation. and 4 rotate among themselves. including both rotations and exchanges. For instance. S4 is the rotational symmetry of a cube or .120 Chapter 8 next largest invariant subsymmetry. In addition. “Completing the cube” corresponds to going from S3 down to A3 . corresponds to the dance of four. and 4 rotate by 120◦ : 1 3 4 2 Thus. we need a solid figure. ending with the smallest of all. defining an axis around which 2. whose possibilities go far beyond a simple square pattern. 3. To rep- resent such a step. If you include all possible steps. the full symmetry of a tetrahedron. Now we take a breath and get ready for four dancers. S4 . the square becomes so tan- gled that it no longer represents the geometry of the dance. S3 . not at all in a square. and so on. This nested structure mirrors the solution of a cubic equation. A tetrahedron does nicely: (234) corresponds to dancer 1 standing still. 3. while 2. the identity. consider a step like (234): 1 stays still.

1 Johannes Kepler’s figure showing an octahedron constructed within a cube by joining the midpoints of its faces. S4 is a complex dance. showing that they share the same symmetry (Harmonices mundi. 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. . Not surprisingly. Its table has 12 × 12 = 144 entries.1). The subgroup of even permutations. These are so numerous that it is difficult to show the table of S4 . which has 24 × 24 = 576 entries. for its exchanges and rotations in general are not commuta- tive. 1619). too many to list here. A4 . S4 is nonabelian. an octahedron (figure 8. with 4! = 24 possible steps. is nonabelian.Seeing Symmetries 121 Figure 8.

3. something very beautiful has hap- pened.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 . where are the other two. in which all four dancers participate. Though A4 is nonabelian. can be factored and solved as Ferrari did. shown in table 8.122 Chapter 8 Table 8. but in sepa- rate pairs. abelian sub- group. like the do-si-dos of square dancing. (13)(24). One by one. two dancers exchanging places while the other . it contains a subsymmetry V that is abelian. V is abelian (though not cyclic) because the double exchanges do not in- terfere with each other and hence can occur in any order. we fully expect that the dancers’ moves will not be adequately represented by a pentagon or any other plane figure. In this progression. From our expe- rience with S4 . too. and (1). They have been conjured up by mathematical necessities that are purely matters of permutations. V is invariant because it includes all the double exchanges. we expect that the quintic will be illuminated by the dance of five. (14)(23). cube. S5 . the Platonic solids have appeared. we now know that the quartic equation. since these cannot represent possibilities such as (12)(345). Having found within the full dance an invariant. and octahedron. the do- decahedron and icosahedron? Just as with the preceding cases. Having encountered the tetrahedron. without any mention of geometry. we might ask. (12)(34).

2). Their five-fold symmetries corre- spond to the patterns of five dancers (figure 8. Yet it is amazing that neither the polyhedron of symmetry S5 nor the solution of a quintic equation with the same symmetry exists. .Seeing Symmetries 123 three rotate among themselves. we regain contact with a geometric model. S5 ) that they necessarily appear in any simple algebraic or geometric situation. There is no intrinsic relation between the properties of three-dimensional space and the solution of equations of the fifth degree. In seeking these invariant subgroups. It would be tempting to draw a connection between the fact that no such polyhedron exists and the unsolvability of the quintic.3). But no regular polyhedron has the symmetry S5 . S3 . that would correspond to the possibility of solving the quintic equation through taking roots. Just the table for its even steps. let us examine these solids to look for further subsymme- tries within them. numbers 60 × 60 = 3. Here the final two Platonic solids find their place in the scheme. If this is a coincidence. S4 . At this point. A5 . we must consider not only rotations about just one of the axes of the icosahedron. We must look to solid figures.600 entries. as before. There are so few fundamental symmetries (S2 . If we could find even one. seeking one that is invariant within the next higher symmetry. but such a “deduction” would falsely ascribe a cause and effect relation when all we have is an analogy. As before. for A5 describes the rotational symmetry of both the icosahedron and the dodec- ahedron (figure 8. A5 is also non- abelian: all the exchanges and rotations interfere with each other so that the order of steps matters. it is a beautiful one. The dance S5 has 120 × 120 = 14.400 possible combina- tions of two steps and is nonabelian.

when it is extended throughout all the possible axes that are equivalent.1 shows that the symmetries of the icosahe- dron are so interconnected that each leads to all the others.124 Chapter 8 Figure 8. it also includes rotations of the same magni- tude around any other axis that admits such rotations.600 combina- tions of steps.1 shows. other than the identity. Let each pair of dancers make all the possible exchanges. There are different sorts of rotations possible about different axes. But there is no such invariant subgroup.2 Kepler’s diagram showing an icosahedron constructed within a dodecahe- dron by joining the midpoints of its faces. In the process. Box 8. for each type of symmetry. showing that both of these figures share the same symmetry. but rotations about all of them. intricate pattern of A5 is so completely interwoven that it cannot be broken down into separable parts. as box 8. If an invariant subgroup includes a certain rotation. they will eventually go through all 3. No invariant subgroup of steps remains closed within itself. there is no invariant subgroup . The beautiful. Thus.

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 ∗).Seeing Symmetries 125 Figure 8. They show the five-fold symmetry shared by the dodeca- hedron. group its twenty vertices into five sets of four equidistant vertices.” A5 . so that if a and b are elements of the group. What we have called a set of possible dance steps. Connecting each of these five sets yields five intersect- ing tetrahedra. Each element a has an inverse a −1 . be solved.3 Within a dodecahedron. and the “dance of five. There- fore the quintic equation cannot. There is an identity element I (so that a ∗ I = I ∗ a = a ). smaller than the whole noncommutative dance itself. a ∗ b is also an element. In the abstract sense. icosahedron. mod- ern math calls a group. in general.

their group op- eration ∗ is successive permutation. The associative law holds: a ∗ (b ∗ c) = (a ∗ b) ∗ c. It was only in 1854 that Cayley set out the basic definition of the the- ory of groups in this abstract form. Now free of any visual or intuitive content. 12 vertices (points at which 5 lines meet). Gradually. (1).1 The symmetry A5 of an icosahedron has no proper invariant subgroup Consider the different classes of symmetries of an icosahe- dron. Here. The concept of group thus represents a step in generalization that recalls the step from some particular pair of objects to the num- ber 2.126 Chapter 8 (so that a −1 ∗ a = a ∗ a −1 = I ). which might have the value 2 in some case but Box 8. the elements are permutations. and 30 edges (lines connecting two adjacent vertices). which correspond to the 60 elements of A5 . the theory found new power in its abstractness. He also showed that any finite group is identical in structure to some permuta- tion group. which is what Galois had in mind when he first used the word “group” in this sense. Another such step occurred when Viète went from a particular number like 2 to a coefficient a in a general equation. S4 . including not only permutations but any set of ele- ments that would satisfy the basic definition. S5 . the concept of group became more and more general. The earliest examples of groups were permutation symmetries like S3 . Note first that an icosahedron has 20 faces (each of them an equilateral triangle). There are 6 axes connecting the opposite vertices. each of which allows 4 different “pentagonal” rotations (each a multiple of 15 of a . and the identity I is the nonpermutation.

It turns out that each of them can be made up of a sequence of “triangular” rotations. There are 15 axes connecting the midpoints of the 30 edges. consider the “pentagonal” rotations. around each of these we can do an exchange. . These rotations comprise two distinct classes within A5 .1 Continued full rotation. Then there are 10 axes connecting the midpoints of opposite faces. hence 10 × 2 = 20 elements more. So the 60 symmetries have the classes given by 60 = 12 + 12 + 20 + 15 + 1. since the faces are triangles). each having 6 × 2 = 12 elements. around each of which we can make 2 different “triangular” rotations (each 13 of a full rotation. making 15 elements more. corresponding to odd or even multiples of rotations by 72◦ . Finally.Seeing Symmetries 127 Box 8. there is the identity. For instance. since 5 faces surround each vertex in a penta- gonal array).

S4 has 24 elements. each “triangular” rota- tion can be made up of two exchanges: 1 (13) (12) 3 2 (12) (13) = (123) * Thus. These arguments are confirmed by an important general rule. Likewise. any subgroup must have a number of elements that is a divisor of n.1 Continued so that these two classes of steps are interconnected and are not separate subgroups. while its invariant subgroup A3 has 3. called Lagrange’s Theorem: If a symmetry has n ele- ments.128 Chapter 8 Box 8.) S3 has 6 elements. (A proof of this theorem is given in ap- pendix C. and there is no invariant subgroup that does not include the whole. Furthermore. the possible subsymmetries of A5 are made up of all the possible exchanges. while . a divisor of 6.

Now any subgroup of A5 must be a divisor of 60. a divisor of 120. sometimes complex. S3 . Each of these groups contains subgroups. The . So the only invariant subgroups of A5 are “im- proper”: either all of A5 .) But S5 has 120 elements and A5 has 60. I . The symmetries we discussed—S2 . there are many possible groups. S5 — are each groups with different numbers of elements. Now any sub- group would both have to be a divisor of 60 and also be com- posed of either 12 + 1 = 13. because 60 and 1 are the only divisors of 60 that are possible. both divisors of 24. nor 16 (nor any sum of them) is a divisor of 60.1 Continued A4 has 12 and V has 4 elements. 21. or 15 + 1 = 16 elements. in general has many possible values. But neither 13.” as mathematicians put it. as S4 contains the subgroup A4 . Some groups are “trivial.Seeing Symmetries 129 Box 8. each with its own peculiarities. each capable of being realized by many possible sets of ele- ments. Thus. The concept of group allows us to contemplate pure relationships without the dis- traction of considering the things related. (Re- member that the identity is always counted as an element of the subgroup. We have found that the possible invariant subgroups are given by the list 60 = 12 + 12 + 20 + 15 + 1. all 60 elements. like the one whose sole member is the identity. The concepts of group theory parallel every step of our dance analogy. each group has its own character and peculiar beauty. 20 + 1 = 21. A5 has no proper invariant subgroup. As we have seen with S5 . S4 . What we called invariant subgroups are usually called “normal” (see notes to this chapter). expressing the interrelation of the subsymmetries within the next larger symmetry. Groups can be abelian (commu- tative) or nonabelian. Moreover. or just the identity. even bizarre.

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 Galois’s great advance over Abel, specifying ex-
actly which equations are solvable and which not. By 1889,
Camille Jordan and Otto Hölder 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. 176–177, 193–194) 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 Abel’s work, he did not
emphasize the theme of noncommutativity, nor did those
who later systematized his ideas, such as Jordan and Hölder.
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, Galois’s
radical viewpoint won out over the older approach because
it gave a complete understanding of solvability, where Abel’s
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 Viète’s 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.
Peacock’s 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 Abel’s 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 Abel’s 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
Abel’s ideas to be widely accepted. In Hungary during the
same period, the brilliant János 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 Abel’s in-
sights, Hamilton showed his contemporaries how important
and beautiful Abel’s 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.

metaphysics. The carving has long since worn away.” reflecting his devotion to philosophy and poetry (he wrote . he had to make this operation noncommutative: i j = k but ji = −k. For ten years. so that one might say that these quantities anticommute. say of geometry. but the strange beauty of noncommutative numbers lives on. In 1843. To make multiplication work for these “quaternions” (as he called them). which he described as “a curious offspring of a quaternion of parents. Since i j = − ji. where i 2 = j 2 = k 2 = −1. (a . but could be if he took quadruples. Begin- ning in 1833.2). Hamilton sought a new kind of number that would extend the idea of complex numbers. To mark this special moment. as he and his wife were walking along the Royal Canal in Dublin. As noted earlier. but not how they should multiply. such an ordered pair could be represented as a point in a plane (see figure 3. b). His difficulties could not be solved with triples. Hamilton spent the rest of his life developing the new algebra of quaternions.134 Chapter 9 Such generous admiration brought a rich reward. algebra. the answer came to him suddenly. 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. Hamilton tried to make these ordered triples work. He could understand how they should add (term by term. 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. from an ordered pair a + bi to an or- dered triple like a + bi + c j. Hamilton then sought to generalize this idea. where j would generalize i so that i 2 = j 2 = −1. and poetry. grouped according to common factors of 1 or i or j). and likewise jk = i = −k j and ki = j = −ik. reversing the order of multiplication reverses the sign of the product. from the plane to three dimensions. a +bi +c j +dk.

though interested in quaternions.” from “outer” products that were not. Hamilton believed that his quaternions promised a new view of the cosmos. Hamilton noted that there were two kinds of multiplica- tion that emerged. Grassmann published his “Theory of Exten- sion” (Ausdehnungslehre). One he called the “scalar product.” the number of dimensions in the “space.The Order of Things 135 poetry enthusiastically. He gradually developed quater- nions. In 1844. only two types of linear product. j can go from 1 to n. and a number of British mathematicians also took them as the key to mathematics and mathematical physics. 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 . is commutative. The name has stuck. leading to the modern notion of vectors (magni- tudes with direction) and scalars (pure magnitudes with no direction).” which takes two vectors.” That is. As if responding to the same unspoken question.” At that time. finally did not join the true be- lievers. there was no hint of what that could possibly be. not merely three. commutative like Hamilton’s “scalar product. The other (not named) was not commutative and produced a vector as a result. . Grassmann proved that “for products of two factors there are . The difficulty was that. As he studied the possibil- ities. . and was a friend of Samuel Taylor Coleridge). which distinguished “inner” prod- ucts. The great Scottish physicist James Clerk Maxwell. either the factors must . if not well. just as complex numbers imply a two-dimensional plane. in which the indices i. a secondary-school teacher in Germany who became a specialist in Sanskrit. Hermann Grassmann. the four-component quater- nion seemed to require a four-dimensional “space. and produces a scalar as a result. independently developed an algebra based on a “space” of n dimensions.

Grassmann’s “outer” product became what we call the “cross” or “vector” product. for matrix algebra allows great simplification of the process of solution (box 9. matrices are perfect vehicles for expressing the . the problem of solving equations stimulated research in a new direction. in which A × B = − B × A. In America. Gradually. the “invariant twins” Arthur Cayley and James Joseph Sylvester. it is routinely taught without much atten- tion to its striking noncommutativity. Josiah Willard Gibbs.2) and can be used to represent the algebras of Hamilton and Grassmann. Found in every textbook. Since they are not restricted to any specific number of dimensions. Indeed. The idea of ma- trices emerges from looking at the coefficients of simultane- ous equations and arranging them in an array.  as with Hamilton’s quaternion multiplication. Once again. which was developed beginning about 1858 by two inseparable English friends. others began to express similar ideas in their own ways. for matri- ces with real values (but in an ordered array) can realize every algebraic requirement of complex numbers or of quaternions. Grassmann’s work was only slowly assimilated because it was expressed in unconventional notation. have noncommutative multiplication (box 9. Grassmann’s algebras emphasized the general- ity of commuting versus anticommuting. Following Gibbs.1). too. Matrices show that such algebras do not neces- sarily require “complex” or exotic kinds of number. This noncommutative algebra is the basis of the modern understanding of physics.136 Chapter 9 commute (e i e j = +e j e i ) or they must anticommute (e i e j = −e j e i ). It is also central to matrix algebra. Soon it was realized that matrices. reformulated Grassmann’s ideas in the language of vectors that is now used universally. whether Newton’s mechanics or Maxwell’s theory of electricity and magnetism. one of the greatest mathematical physi- cists of the century.

y in terms of x  .1 Two successive linear transformations. y involves solving and re- substituting these two transformations. Expressing x  . Cayley discovered an eight-dimensional generalization of quater- nions. y : b 11 x  + b 12 y = x  b 21 x  + b 22 y = y . now called octonions or Cayley numbers. operations of groups.” giving names (not all of . Consider a second such transformation that expresses x  . 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. 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 . Sylvester called himself “the new Adam.The Order of Things 137 Box 9. y in terms of x. 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. Grassmann’s generalization of the pos- sibilities of algebra opened an ever-widening field.

Benjamin Peirce had enu- merated 162 different algebras.) The historical unfolding of the notion of noncommutativ- ity implies larger questions: What is its significance? Why should we care about it? In 1872.2 Matrix multiplication is not commutative in general Consider. in which he proposed that group theory be considered the heart of mathematics. S. (His remarkable son C. zetaic multipliers. discriminants. the following noncommuting matrices:        0 1 1 0 0 + 0 0 + (−1) 0 −1 = = . considering ever more general groups would be the royal road to understanding ever more complex spaces. 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 = = . He argued that each geometrical figure should be considered more in terms of its fundamental symmetry— specified by the group governing it—than by any figure or equation in itself. In Klein’s view. 0 −1 1 0 0−1 0+0 −1 0 which stuck) to a whole new universe of algebraic beings: invariants. not long before Cayley set out group theory in full abstraction. for instance. among many others.138 Chapter 9 Box 9. By 1860. Peirce carried forward Boole’s logical work as a philosopher. more fundamental in its import than geometry or algebra alone. . allotrious fac- tors. the great mathematician Felix Klein gave a lecture at Erlangen. reaching past three dimensions to manifolds of higher dimensions. besides being a chemist and astronomer.

Klein saw them as incarnations of fundamental groups. Indeed. about 123 km away. (b) going directly north for 1. you end up somewhere else. the rotations and reflections that leave it invariant.000 km. the Platonic solids are incarnations of more complex rotational symmetries: the tetrahedron of A4 . (b) then (a). This was one of many instances that moved Klein to deem group . As part of his pro- gram. In fact. Imagine leaving Santa Fe for two journeys: (a) going directly west for 1. the cube and octahedron of S4 . then (b). as Eu- clidean areas and lengths do when moved in the plane. Where the Greeks considered these solids as emerging from the necessity of fitting together reg- ular polyhedra in three-dimensional space. Klein went back to reconsider the Platonic solids in terms of group theory. But if you do them in the reverse order. If you do these journeys in the order first (a). The order matters since you are traveling on a globe. you will arrive at a certain spot. the dodecahedron and icosa- hedron of A5 . an equilateral triangle should be considered most of all in terms of its sym- metry group S3 . It was an influential rallying cry that still re- sounds in contemporary mathematics. Klein’s presentation lies behind the dance analogy used in the preceding chapter.000 km. noncommutativity is everywhere. In general. not a flat plane. Thus.The Order of Things 139 The “Erlangen Program” was the clarion call of a new abstract vision. most general concepts underneath. He classified geometries by the algebraic properties that remain invariant under a particular group of transformations. rota- tions in three-dimensional space do not commute. understood as a manifestation of its underlying symmetry of rotations. Klein considered this peculiar noncommutativity of spa- tial rotations to be the hallmark of space itself. Similarly. going past the traditional starting points of mathematics to reveal the deepest.

In 1927. Looking back from the perspective of the twenti- eth century. scaled up or down by a multiplicative factor. Perhaps this is simply because the order in which things are done far more natu- rally applies to algebra. Long before.140 Chapter 9 theory the master key that would unlock the secrets of both algebra and geometry by uncovering what underlies both of them. than to geometry. despite such simple cases as the jour- neys around the Earth just described.” The original development of group theory in this discrete. Weyl noted that “it was a lucky chance for the development of mathematics that the relativity problem was first tackled. algebra felt the impact of noncommutativ- ity before geometry. Galois’s fi- nal letter was “perhaps the most substantial piece of writing in the whole history of mankind. not for the continuous point space. the eminent mathematician Hermann Weyl took this as a kind of “relativity” that emerged long before Albert Einstein’s theories. because of its novelty and profundity. in which symbols and operations are read in a certain order. As the comparison of the regular solids and the solvable equations shows. which seems to rely on timeless diagrams. 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). the symmetry between all uniformly moving frames of reference points to an invariant. while leaving invariant their essen- tial constellation: the number of the roots and their relative spacing. Here again.” Weyl even wrote that. algebraic context prepared the way for Einstein and Hermann Minkowski’s understand- ing of relativity in terms of groups of continuous motions that leave the speed of light invariant. Plato considered the eternal . Descartes had noticed that the roots of equa- tions could be increased or decreased.

As I have discussed . It was.6×10 erg-seconds) and −27 where i = −1. It gives dignity to human choice by emphasizing its irrevocable consequences. unless we expend work and also expel heat into the rest of the universe. once done. a shock to realize that the order of things mattered. did not necessarily commute. to pure being beyond existence and becom- ing. a statement that natural processes are in general irreversible. Max Planck emphasized irreversibil- ity as he hesitatingly introduced the quantum into physics. or at least tame it into a more classical form. Dirac’s relation is the direct source of the fa- mous Heisenberg uncertainty principle. then. The heart of thermodynamics is its Second Law. irreversibility means that every action. Time is no longer the perpetually reversible stream of Newtonian or Maxwellian dynamics. The question of order is crucial for modern physics. Struck by its sweep and scope. not vice versa. flowing ever onward. As quantum theory took shape. Formally. On the human scale. irreversible time is the dark river of history. Instead. Since h is extremely small by human standards. it revealed noncommutativity at its very core. he recognized that here too irreversibility held sway. Paul Dirac showed that its crucial realization was that physical quantities.The Order of Things 141 changelessness of mathematics a hallmark of its closeness to absolute reality. Heat flows from hotter to colder bodies. can never be undone. √ h is Planck’s constant (h = 6. Ironically. the child of ir- reversible thermodynamics married to reversible mechanics. Planck himself tried to reverse the flow of events and undo the shock of the quantum. this noncommutativity usually manifests itself only on the atomic scale. In a famous equation. when expressed in algebra. 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/2π . Ruefully.

noncommutativity in quantum theory stood for the irreversibility of measurement in general. Going from these most general principles to more specific cases. This theory rests on what are called nonabelian gauge fields. Here group theory expresses beautifully the sym- metries of space-time. Without even exploring what such fields might be in de- tail. Such nonabelian theories promise to be . Any change in this principle would require overturning this fundamental principle of causality. then all physical quantities of the events must commute. For all of them. whether the quantum theory of fields at space-time points or alternative string theories that use not points but tiny lengths or membranes. physical quantities cannot commute. within the realm of causal influence traveling at or less than the speed of light. Commutativity versus noncom- mutativity lies at the heart of causality itself. Noncommutativ- ity continues to govern the deepest level of physical theory. which has so far been confirmed in myriad ways. noncommutativity is crucial to the development of the “standard theory” that has very successfully drawn the known elementary particles into a largely unified scheme. Initially. Conversely. these are the distinctive hallmarks of the utter indistinguishability of quanta. 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. noncommutativity is basic and promises to be so in the future. expressing their causal relation. expressing their independence.142 Chapter 9 in my book Seeing Double. If events are sufficiently separated in space and time that no light signal can connect them. note that their very name announces that they rely on a special sort of noncommutativity in the symmetries of funda- mental particles.

They also are the starting points of attempts to unite quantum theory with Einstein’s gen- eral theory of relativity. which is a geometrical theory of gravitation as the curvature of space-time. the rotational invari- ance of three-dimensional space leads to the five Platonic solids. if only as a point of departure. The search has only just begun. As noted in the last chapter. Perhaps such considerations of groups may someday explain why we perceive three dimensions of space and only one of time. .” Klein’s conjecture continues to be deeply provocative: nonabelian groups may give deep clues to the nature of space and time. notably the theory of “noncommutative geom- etry.The Order of Things 143 the inescapable vehicles of future physical theory. These directions in physics have in turn stimulated further developments in mathematics.

This page intentionally left blank .

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

new solu- tions of the quintic eclipsed the question of solution in radi- cals. The name “alge- braic numbers” came to signify all the solutions of algebraic equations whose coefficients are integers. which could inspire horror or won- der. the development of algebra led to the term “number” √ being applied to irrationals as well as integers. of whatever degree. To give a name to those alge- braic numbers that are not expressible in radicals. if only to avoid paradox and confusion. It turns out that their solutions can be identified with the so-called elliptic modular functions (defined as sums of powers times trigonometric functions). for they could simply annex the algebraic numbers to those expressible in radicals. they must be a new kind of magnitude. giving them a name seemed tantamount to dispelling the sense of crisis or paradox. the moderns united. Once symbolic algebra was established. Since solutions of all equations are known to exist by the Fundamental Theorem of Algebra.” Now we can complete the odyssey of the quintic. Yet what the ancients separated.146 Chapter 10 lead to an infinite vista. Abel’s proof showed that radicals are not sufficient. . For mathematicians. just as the ancient proof showed that ra- tionals are not enough. could be solved by using these generalized functions. and Fernando Brioschi independently showed their relation to the quintic. These functions are based on the “theta functions” that Abel and Jacobi began to study in 1827. Gradually but inex- orably. we might call them “ultraradical numbers. As with the irra- tionals. Leopold Kronecker. though it was not until 1858 that Charles Hermite. By 1870. the crisis of un- solvability was inevitable. Euclid tried to keep them rigorously separate from the 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. Jordan had proved that any polynomial equation.

At Bell Laboratories in 1937. By the end of the twen- tieth century. the Isograph. scientists developed a larger but less elegant-looking machine. In 1840. which could approximate equations up to the fifteenth degree (figure 10. But what.Solving the Unsolvable 147 During the same period. In 1895.2). far more powerful programs were available to users of personal computers.1). . 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. About 1642. These developments simply sidestepped the problem of solution in radicals. then. Blaise Pascal had the first vision of machines that could compute. people started building machines that could find approximate solutions. Leonardo Torres Quevedo developed an elegant machine that could calculate logarithms needed to solve diffi- cult algebraic problems (figure 10. Leon Lalanne devised a mechanical computer that could approximate equations up to the seventh degree.1 A machine built in 1895 by Leonardo Torres Quevedo to calculate logarithms.

between arithmetic and geometry: the Greeks learned that expressing an irrational requires an infi- nite number of digits. in dif- ferent contexts. Each crisis opened a new insight into the infinite. As .2 The “Isograph. Abel’s proof. they were so busy forging ahead that they had little time for retrospection.” built in 1938 at Bell Laboratories to solve polynomial equa- tions up to the fifteenth degree. Finally. Yet surely there remains valuable insight to be gained by reflecting on what each successive crisis really meant. between geometry and algebra: to express the area of an oval requires an infinite number of algebraic terms. First. aside the older scruples? Perhaps they were right not to look back. within algebra itself: solving a finite equation requires an infinite number of terms. as Newton showed.148 Chapter 10 Figure 10. Next. Even solving a cubic equation with real roots requires that we take the cube root of an arbitrary complex number.

then zn = r n e inθ . but infinite series.1 De Moivre’s formula Express the complex number z in terms of its absolute value r and its argument θ. If we write z = r e . The problem was resolved by Abraham De Moivre in 1707 and given its modern form by Euler in 1748. .718 . Using the Taylor series 2 2 2 for e inθ = 1 + inθ + i n2! θ + · · ·. Most calculators provide sines or cosines by pressing a button. De Moivre’s 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. Yet algebraically sines and cosines are not finite polynomials. z = r (cos θ + i sin θ ). . We must start by expressing each complex number in terms of a magnitude and an angle. is the basis of iθ 2 natural logarithms. where i = −1 and e = 2. De Moivre’s formula follows if we group the real and imaginary terms separately:     n2 θ 2 n3 θ 3 zn = r n e inθ = r n 1− + ··· + i nθ − + ··· . then e iπ = −1 (since sin π = 0 and cos π = −1) or e iπ + 1 = 0. Euler’s beautiful formula. consider the two “Taylor series” sin x = x − x3 5 7 2 4 6 3! + x5! − x7! + · · · and cos x = 1 − x2! + x4! − x6! + · · ·.Solving the Unsolvable 149 complex numbers were accepted.1). 2! 3! If we set nθ = π and r = 1. understood as polar coordi- nates in a plane. De Moivre proved that zn = r n (cos nθ + i sin nθ ). taking their roots became imperative. Geometrically. . these functions are ratios of the sides and hypotenuse of a right triangle. To show this. Newton had already discerned this algebraic infinitude in studying Box 10.

1. Thus. In this.150 Chapter 10 the areas of ovals. calculators sum a few terms to provide a value of required accuracy. finally in 1873 Hermite showed rigorously that e was transcendental. uniting e. To do these later developments justice would require an- other book and would take us away from Abel. The ul- traradical solutions of the quintic equation. they are a kind of amphibian between land (the radicals) and the vast tran- scendental sea. this implied that geometry can express in a few lines an infinite series of algebraic terms. . but e and π were only the beginning. These proofs ended the dreams of the circle-squarers. He also showed that algebraic numbers are countable. Here. For him. these series lead to the formula e iπ + 1 = 0. π . and in 1882 Ferdinand Lindemann did the same for π . Using such series. to whom we now return one final time. Though it was long suspected. In 1874. are neither the one nor the other.1 shows. but they are not composed of radicals. they are algebraic numbers. in general. whose hopes rested on the possibility of an algebraic formulation of π . 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! + · · ·. not the solu- tion of any algebraic equation of finite degree. 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. 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. they are only one among innumerable kinds of irrational numbers. One might have thought that such transcendentals were rare and exotic. and 0 in one pregnant expression. As box 10. one of Euler’s most beautiful thoughts. i. whose infinitude was already implicit in Euclid’s Book X.

The end of the old assumption that all equa- tions have a finite solution revealed a new mathematics of infinite series and noncommutativity. Where the young Gauss had divided a circle into seventeen parts to construct a regular seventeen- sided polygon. He also took Gauss’s seminal insights even further than the master. He went on to do profound work on tran- scendental functions. his proof of the unsolv- ability of the quintic came during his school years. Abel gazed at the dawn of a new mathematics. the “Abelian integrals” and “Abelian Addition Theorem” that deeply influenced those who came after him and that many mathematicians call his greatest work.” Abel recorded the power of the infinite extension of finite algebra. By transcending this limitation. Abel solved the unsolvable. As if taking up Viète’s bold claim “to leave no problem un- solved.” He had shown that solv- ability in radicals is a condition that need not always exist. the lemniscate. . which he began to explore in the few years remaining him. Great as it was. and integrals that gen- eralize trigonometric functions. Nevertheless.” Though not much given to speculating about the significance of what he had done. Instead of looking for a relation that one does not know exists or not. It would be unjust to represent it as his only or greatest accomplishment in mathematics. 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. Abel was able to make a similar division of a more complex curve. Abel called the solvability of equations “my favorite subject. one must ask if such a relation is really possible. infinite series.Solving the Unsolvable 151 Standing on the edge of this ocean. which one can always do with any problem.

The large lemniscate (∞) is at the top of the page.152 Chapter 10 There is another document that records something of Abel’s inner feelings about his work and life. but it is cryptic and private.3 A page from Abel’s Paris notebook (1826). . Written in 1826.3) combine mathematics and reverie in so unguarded Figure 10. Abel’s doodlings in one of his Paris notebooks (fig- ure 10. during his work on the lem- niscate. the passages discussed in the text are to the right and under it.

. . . Abel writes in French and Norwegian. Abel has cross-hatched this curve and repeatedly outlined it.” No page more daringly suggests the mysterious intersec- tion of desire and mathematics. not. Goddamn my ∞. 0 (1 − x 2 )(1 − α 2 x 2 ) Then: “Listen . . the felt meaning of this outburst? Is it humor. . . . .” Then. . in French.” 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 .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. Abel writes “complete solution to the equations in which . a large figure ∞. featuring a striking drawing of a lemniscate.  1 d x Elisa w= √ . . . a remembered or even an imagined woman? Why does he scrawl the name of “Soliman den Anden.” her name oddly entwined in an elliptic integral. . Listen for once. beloved . an ironic prayer: “Our Father who art in Heaven. . interspersed with equations and drawings. “speak to me. Come to me. . Come to me in God’s name . . give me bread and beer. . What remains is equations and ∞. . . . my solutions to algebraic equations . . This page was doubtless meant for no other eyes than his own. wonder? We will never know. one of the tempting Parisians whose beauty he men- tioned in a letter. my friend . come to me . .” In the midst of this. in all your lewdness.” breaking off into equations and intimate thoughts: “My friend . . At the top of this page. Listen .” What is the tone. for once. Who is Elisa? An unknown beloved. . my dear. vexation.

This page intentionally left blank .

b the sum of their products. which aims to fill this gap in the theory of algebraic equations. which does not follow the conventions of our text.] . but (if I am not mis- taken) none has succeeded until now.. according to Girard’s identities (box 4. etc. 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. 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. Therefore I hope that geometers will receive kindly this memoir. and e. c the sum of their triples. [The commentary is italics in brackets []. equation numbers have been added also in brackets. Appendix A Abel’s 1824 Paper Memoir on algebraic equations. d. Note Abel’s notation for the coefficients. He has chosen the signs of [A1] so that the coefficient a is the sum of the roots.1). b. c. that is. one can express y by a function formed by radicals of the quantities a .

etc. 1 Then 1 2 m−1 y = p + R m + p2 R m + · · · + pm−1 R m . . [A4] q . . understood in terms of the successive orders of functions as nested radicals. In what follows. e. . [For Abel’s argument for the generality of [A2]. c. . . c. can always be expressed in the form [A2]. p1 . b. so that eventually one reaches only rational functions inside the innermost radicals. p2 . and by putting pRm in place of R. etc. which he con- sidered Step I of the larger proof. b. p2 . including putting all the terms on a common denominator. p. e. and e. being rational and entire functions [i. [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 . p2 . are each algebraic functions of the coeffi- cients a . .e. explained in appendix B. b.156 Appendix A It is clear that in this case we can express y in the form 1 2 m−1 y = p + p1 R m + p2 R m + · · · + pm−1 R m . p. b. q 1 . etc. p. p1 .. Essentially. d. d. p. . [A2] m being a prime number and R. he shows that a finite sum of radicals. in which R. 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. see appendix B.] 1 We can also assume that it is impossible to express R m by a rational function of the quantities a . d. we obtain a result of this form: 1 2 m−1 P = q + q 1 R m + q 2 R m + · · · + q m−1 R m . however they are nested. and R. Abel assumes 1 that this has been done to remove p1 . and so on until we come to rational functions of the quantities a . polyno- mials] of the quantities a . p1 .] Substituting this value for y in the equation proposed [A1] and reducing. q 2 . functions of the same form as y. it is clear that we can make p1 = 1. p2 . c.

. . q 1 = 0. q 1 . Note that since [A1] sets a polynomial to zero. . p. giving q + q 1 z + · · · + q m−1 zm−1 = 0. c. . .Abel’s 1824 Paper 157 [Now Abel substitutes the presumed solution [A3] back into the main equation. . . q . . ending with [A8]. it is necessary that q = 0. q 2 = 0. q m−1 = 0. Doing so leads to [A4]. p2 . Then substituting [into [A6] z → αµ z]. Also. [A1]. . . Let r + r1 z + r2 z2 + · · · + rk zk = 0 [A6] be that equation. q 2 . are not equal to zero. . . these equations [A5] necessarily have one or more common roots. . [A5] [Abel sets out to prove q = 0. so that each multiplies the 1 appropriate power of R m . It has these roots in common with the equa- tion zm − R = 0. where αµ designates one of the roots of the equation αµm − 1 = 0. b. p1 . depend on all the previous quantities: the coefficients a . so that zm = R or zm − R = 0. where the new functions q . q 2 . q 1 . all the roots of this equation have the form αµ z. we have . . . calling R m = z. as shown in appendix B. . Note also that. . q 1 . for con- venience. Abel has defined q . These two parts of [A5] constrain z. R. He begins by 1 defining z = R m . in [A4] 1 substitute z = R m .] If now the quantities q . his proof requires several steps. Note also that the highest power of R in [A4] will be m−1 m . q 1 = 0. q 2 = 0. we have two equations zm − R = 0 and q + q 1 z + · · · + q m−1 zm−1 = 0. . thus.] In order for this equation to be valid. this rewritten equation also does so. . pp. 173–174. . q m−1 = 0. q 1 . . 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. e and the quantities he has just used. d. If k is the number of these [common] roots. P = 0. 1 In fact. and q m−1 .

the subscript µ is an index that goes from 1 to k (the total number of common roots in [A5]. . αz. of the k roots. which is contrary to the hypothesis. Now αµ stands for the whole series of values 1. . z1 . αk−2 z). Therefore. . zk . but zm = R and so this means that αµm R − R = 0. one can always find the value of z expressed by a rational function of the quantities r. . r1 . αz. p. . e. .) Abel now substi- tutes the successive values αµ z = z. b. α. . rk . r + r1 z + r2 z2 + · · · + rk zk = 0 [A7a] r + αr1 z + α 2r2 z2 + · · · + α k rk zk = 0 [A7b] ······ r + αk−2r1 z + αk−2 2 r2 z2 + · · · + αk−2 k rk zk = 0. z. z2 . rk and α because we have k simultaneous equations to de- termine the k unknowns. r1 . . . . . . . . q m−1 = 0. . it follows that z is also a rational function of these same quantities. . r2 . . . . . [A8] [Now Abel makes the critical observation: from the k equa- tions [A7a–A7k]. . . [A7k] [In z = αµ z. R. αk−2 .. . Dividing by R (which = 0) shows that αµ is a root of the equation αµm − 1 = 0. the first and second are 1 and α. . . . . . α2 . Substituting 1 for αµ gives [A7a]. Substituting z = αµ z back into zm − R = 0 yields αµm zm − R = 0. α1 z. we can always find z as a rational function of r. . and. until substituting αk−2 yields [A7k].158 Appendix A the following equations.. α2 z. . (Note that there are then only k − 2 values of αµ since. . as these quantities are themselves rational functions of a . . it is necessary that q = 0. .] From these k equations. c. d. q 1 = 0. substituting α gives [A7b]. Notice that this is different . p2 . αk−2 z back into [A6]. α2 z. where Abel has noted that 1 is a root of this equation and calls α1 = α. . . . .

except that there is now a factor of α raised to some power (a + 2b + · · ·) multiplied times these previous factors of q and R.] . But we assumed that z = R m is not a rational function of its variables. The same argument also applies for y3 (R m → α 2 R m ) and all the other values of [A9]. q = q 1 = q 2 = · · · = q m−1 = 0. 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 . concluding the proof of [A8]. But since the qs are zero. [A9] α being a root of the equation α m−1 + α m−2 + · · · + α + 1 = 0. α R m . α 2 R m . so we are forced to the only alternative. this becomes (prod- 1 ucts of p.] If now these equations are valid. [A10] 1 2 m [If y1 = p + R + p2 R + · · · + pm−1 R m m m−1 [A3]. 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. . · · · in [A4]. . then each term still vanishes and this y2 also satisfies P = 0 in 1 1 [A4]. 1 then the integral powers of Ri = (R m )mi can be factored out and included with the products of p. we get terms like (products of p. α m−1 R m . . p2 . where i. . Now if we consider y2 . . · · ·) (R m )a +2b+··· . in substi- tuting this form y1 into [A1]. Collecting powers. which was one equation to determine five values of y. a similar argument applies. q 1 . . Abel has just shown that all these qs are zero.Abel’s 1824 Paper 159 from the situation with the original equation [A1]. [A7a–A7k] are k linear equations to determine k unknowns.) (R m )a (R m )b · · · . 1 2 p2 . in 1 1 which R m → α R m . p2 . · · · as q . j are integers ( j ≤ m − 1). Since the exponent a + 2b + · · · can always be written as mi + j. This happens because. α 3 R m . . then q + 1 m−1 q 1 R m + · · · + q m−1 R m = 0.

[A12c] m ······ m−1 1 pm−1 R m = (y1 + αy2 + · · · + α m−1 ym ). then [A11a] = [A11b]. we easily deduce that 1 p= (y1 + y2 + · · · + ym ). we will have 1 2 m−1 y1 = p + R m + p2 R m + · · · + pm−1 R m . [A12a] m 1 1 R m = (y1 + α m−1 y2 + · · · + αym ). [A11m] [Note that. The number m thus cannot exceed 5. and clearly that can only hold if the coefficients are highly restricted. [A11a] 1 2 m−1 y2 = p + α R m + α 2 p2 R m + · · · + α m−1 pm−1 R m . y2 . For instance. y4 . there are special cases of quintics having equal roots. if on the contrary y1 = y2 (for example). y = y0 . and such an equation leads. and y5 the roots of the proposed equation [A1]. for in the contrary case we would have an equation of the same form as the equation P = 0. if all the roots are equal. but they will prove to have simple. which contradicts [A10]. Therefore all the values of y are different. Of course. [A12b] m 2 1 p2 R m = (y1 + α m−2 y2 + · · · + α 2 ym ). [A12m] m .160 Appendix A We see also that all these values of y are different. Abel also reminds us that the roots can be no more than five in number.] From these equations. y3 . rational solutions. Therefore designating by y1 . which would require that α − 1 = 0 = α 2 − 1 = α 3 − 1 = · · ·. as we have seen. then the quintic equation can just be factored into (y − y0 )5 = 0. [A11b] ······ 1 2 m−1 ym = p + α m−1 R m + α m−2 p2 R m + · · · + αpm−1 R m . to a result that cannot be valid.

as these [values] are rational functions of y1 . . y5 and uses [A3] and the result of [A9–A10] to write them out explicitly in [A11a–A11m].. . Let 1 2 n−1 R = S + v n + S2 v n + · · · + Sn−1 v n . [A11m] by α. . as follows. S2 . Then he adds these equations up and gets 1 y1 + y2 + · · · + ym = mp + (1 + α + α 2 + · · · + α m−1 )R m 2 + p2 (1 + α + α 2 + · · · + α m−1 )R m + · · · m−1 + pm−1 (1 + α + α 2 + · · · + α m−1 )R m . y2 . . [A11b] by α m−1 . . . Now Abel will tease out the other terms in [A4] by multiplying each equation in such a way as to 1 isolate each term. . and R m are rational functions of the [roots of the] proposed equation [A1]. p2 . . . we obtain a 1 similar result showing that the quantities v n . for example R. and. multiply [A11a] by 1. [A12. S. S2 . are also. Then add them up. etc. y2 . . [A13] Treating this quantity in the same manner as y.] Let us now consider one of these quantities.1] But from [A10] this leads immediately to [A12a]. . v. [A11c] by α m−2 . R. [Now he calls these five roots y1 . . are rational functions of the different values of the function R. Exactly similar tricks yield the rest of [A12b–m]. [A12. . .2] 1 Since α m = 1. since all the sums (1 + α + α 2 + · · · + α m−1 ) vanish. then mR = y1 + α m−1 y2 + α m−2 y3 + · · · + αym m [A12b].Abel’s 1824 Paper 161 1 We see from this that p. pm−1 . v. To find out what R m is. . the 1 functions v n . . We get: y1 + α m−1 y2 + α m−2 y3 + · · · + αym = (1 + α + α 2 + · · · + α m−1 ) p 1 2 + mα m R m + p2 α m (1 + α + α 2 + · · · + α m−1 )R m + · · · m−1 + α m pm−1 (1 + α + α 2 + · · · + α m−1 )R m . . . . S.

y4 . R being a rational function of a . y3 . To show this. 17. Furthermore. and e.) [See appendix C for Abel’s expansion of this argument. 1 Let R m = r . Cauchy’s theorem . It also means that the equation 1 R m = r can be permuted among the m roots. y4 . (See the memoir by M. and 1 y5 among themselves in the equation R m = r . y2 . . and y5 as independent variables.] This being established. . it is necessary that m = 5 or m = 2. thus 1 the equation R m = r can hold under this supposition. we conclude that all the irrational functions contained in the expression for y are rational func- tions of the roots of the proposed equation. and r will then assume the m different values. y5 . Abel lets R m = r and reminds us that he has just shown it to be a rational function of the roots y1 . . it is clear that one can consider y1 . y3 . . y3 . Let us first consider the irrational functions of 1 the form R m . R is a symmetric function of these roots. and y5 and R is a symmetric function of these quantities. however they are permuted. vol. Cauchy’s 1 theorem states that if m = 5. Cauchy in the Journal de l’école polytechnique. since by this 1 interchange R m necessarily takes m different values since R is a symmetric function. [This concludes Step II of the whole proof. b. y2 . we can interchange the quantities y1 . Con- sequently. c. [where] r is a rational function of [the roots] y1 . 1 [Now the final stage of the reductio begins. our function r = R m can take on only five or two values. it is not hard to complete the de- monstration.] The function r must also take m different values in per- muting in all ways possible the five variables it contains. d. This means that we can permute the roots among themselves without changing R.162 Appendix A Pursuing this reasoning. since m is a prime number. y2 . Now since the case in question is the general so- lution of the fifth-degree equation. y4 . never three or four values. y2 .

Then we can express v in terms of y1 and of the coef- ficients of an equation of which y2 . [A16] But this equation is impossible. the simple argument: Consider expressing one of 1 2 4 the roots y1 as in [A11a]. 1 as in [A12b]. where the case m = 5 has been taken. as well as a much simpler argument for the impossibility of m = 5. Alternatively. y4 . y3 . y5 are the solutions. . First. y3 . let m = 5. y5 . y5 and is symmetric under permutations of four out of five of these. this relation is impossible. Abel gives a much fuller account of these terse assertions. 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 . To . [A14] p. p + p1 y1 + p2 y12 + p3 y13 + p4 y14 = αp + αp1 y2 + αp2 y22 + αp3 y23 + αp4 y24 [A15] where α 4 + α 3 + α 2 + α + 1 = 0.Abel’s 1824 Paper 163 allows r to take only one value. . being symmetric functions of y1 . p2 . .] First. . y4 . . on interchanging y1 and y2 . . y4 . 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). . y1 = p + R 5 + p2 R 5 + · · · + p4 R 5 and 1 then derive the expression for R 5 = 5 (y1 +α 4 y2 +α 3 y3 +α 2 y4 +αy5 ). so m consequently must equal 2. p1 . here is Abel’s longer reasoning for [A14–A15]: Consider a rational function v that depends on y1 . This equation gives. y2 . [In his 1826 paper. y2 . y2 . but that would contradict our initial assumption that all the roots are different. However. y3 . Therefore the case m = 5 is excluded.

y2 . . d. q . pm ). d. c. as was just as- sumed. the function v can be expressed rationally in terms of y1 . Now y1 . y4 . . q = b − a y1 + y12 . y5 . . y5 and now is also in y1 ) or it takes five values (one for each of y1 → y1 . y5 . since this is a quartic equation and hence solvable. . Then it follows that v can be written as a series of terms v = r0 + p1 y1 + p2 y12 + · · · + pm y1m . 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 . Thus (since we exclude the symmetric case). . y3 . then y15 = a y14 − by13 + cy12 − dy1 + e. s = d − cy1 + by12 − a y13 + y14 . y3 . y3 . y2 . c. y4 . . b. y5 must . d. a . y5 in the form v = p0 + p1 y1 + p2 y12 + p3 y13 + p4 y14 . Then we can compare these forms of the same equation and conclude that p. 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). y4 . this argument shows that any function that has five different values y1 . pm are polynomial functions of a . r. y5 are all the roots of a quintic equa- tion y5 − a y 4 + by3 − cy2 + dy − e = 0 [A1]. y3 . y3 . y3 . The roots y2 . b. r = c − by1 + a y12 − y13 . b. y2 . e. where p0 . c. y4 . s can be expressed in terms of the root y1 and the original coefficients a .164 Appendix A see this. y4 . y2 . r. p1 . s. write (y − y2 )(y − y3 )(y − y4 )(y − y5 ) = y 4 − py3 + q y2 − r y + s. Now if in any such function we permute all five of y1 . y4 .) Thus. y3 . y5 can be expressed in terms of the coefficients p. Since y1 is a root of [A1]. Thus. we can write any function that is symmetric under per- mutations of y2 . (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 . q . p1 . 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. . then it follows that it takes only one value and is thus sym- metric (since it is already symmetric in y2 . y4 . so that p = a − y1 .

] Then let 1 R 2 = r. However.Abel’s 1824 Paper 165 take the form v = p0 + p1 y1 + p2 y12 + p3 y13 + p4 y14 [A14]. y4 . S 2 = (y1 − y2 )(y1 − y3 ) · · · (y4 − y5 ) discussed in appendix C. [A17] and then r must have two different values of opposite sign. [Abel writes the two possible values for r as [A17].] . By doing this. as Abel does) p + p1 y1 + p2 y12 + p3 y13 + p4 y14 = αp + αp1 y2 + αp2 y22 + αp3 y23 + αp4 y24 . but this is the essence of his argument. neither of which is allowed since the roots are all different. (Abel discusses various cases in which v takes different numbers of values. Therefore there can be only two values of r . Note that S cannot be zero because no two roots are the same. α 2 R 5 .) 1 We also know that these different roots are related by [A9]. this equation cannot be satisfied unless either α = 1 or y1 = y2 . α 4 R 5 . We then have (see the memoir of M. α R 5 . remembering that the square root is always ambiguous up to a ± sign. 1 1 so as to take R 5 → α R 5 and hence y1 → y2 . which is [A15]. R 5 → 1 1 1 1 1 1 R 5 . where v is a symmetric function (dependent on the coeffi- 1 1 cients) and S 2 is a special function. [A18] v being a symmetric function. y3 . α 3 R 5 . Cauchy’s theorem also implies in this case that r can be written in the form [A18]. Then take [A14] in the form R 5 = r = p + p1 y2 + p2 y22 + p3 y23 + p4 y24 and multiply both sides by α. we get α( 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 . Cauchy) 1 R 2 = r = v(y1 − y2 )(y1 − y3 ) · · · (y2 − y3 ) · · · (y4 − y5 ) 1 = vS 2 . and simi- larly for the other possible four values y2 . by Cauchy’s theorem. y5 .

rather than R 5 . d. we have  1 rr1 = p 2 − p12 S m . . if it were not. y4 . R = v 2 S. R1 = v12 S. m must equal 2. As we have seen. [A23] [Abel reminds us that we now know that we are dealing with 1 1 R 2 . and e and consequently symmetric functions of y1 . where he has dropped out the other terms in [A19] since his object is merely to show us the general form they will take. Using the definitions R = v 2 S. Multiplying out rr1 gives [A23]. etc. The preceding function [A19] can thus be written in the form  1  m1 p + p1 S 2 . R1 = v12 S. [A20] Let  1  m1 r = p + p1 S 2 . etc. we must have ν = µ = etc. b. Likewise. he chooses the minus sign in defining r1 to give this convenient form. What would this . y2 . [A19] p. etc. but in this case r would have four different values. [Abel argues that the product rr1 is a symmetric function. [A21]  1  m1 r 1 = p − p1 S 2 . R. then (by Cauchy’s theorem) m = 2. and y5 . c. . he can express irrational functions such as [A19] in the form r = 1 1 ( p + p1 S 2 ) m [A21]. . = 2.166 Appendix A Let us now consider irrational functions of the form  1 1  m1 p + p1 R ν + p2 R1µ + · · · . y3 . which is impossible. p1 . p2 .] Now if rr1 is not a symmetric function. being rational functions of a . R1 . therefore rr1 must be a symmetric function. [A22] Multiplying. he writes another irrational term r1 in the 1 1 similar general form r1 = ( p − p1 S 2 ) m [A22].

then  1  m1  1 − m1 r + r 1 = p + p1 S 2 + v p + p1 S 2 = z. . c. and e. [A24]  − m1 1 [Remember that r1 = vr = v p + p1 S 2 . only m = 5 or m = 2 is possible. b. then m must equal 5. c. c. So then rr1 is a symmetric function.” as he has done many times before. and e. d. using [A14].] This function having m different values. y2 . p1 . 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 .] . d. to express y as a function of z and the coefficients a . P3 . [Now he turns the expression [A25] “inside out. he writes z in [A25] in terms of the roots of the quintic. q 2 . d. 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 . P2 . [A25] q . . a . and thus rational functions of a . since it would involve a square root of terms including square roots. This is not allowed. y3 . [A23]. can be written in the form 1 p + p1 S 2 (because of the reasoning around [A20]). b. r would have four values. by elimination. b. Because of [A11a].] Combining this equation with the proposed equation. being symmetric functions of y1 . . e. m = 5. and e. c.Abel’s 1824 Paper 167 mean? In [A21]. where all the functions P. R. q 1 . b. [Since he has excluded the possibility that m = 2. and S being rational functions of a . we can express y in terms of a rational function of z. he can write this in the form of [A26]. To regain contact with his quest for the roots of the original equation. etc. p. z = q + q 1 y + q 2 y2 + q 3 y3 + q 4 y 4 . since m is a prime number. R. etc. and P4 are functions of the form p + p1 S 2 . d. [A26] 1 where P.] Let v be this [symmetric] function [v = rr1 ].

] Now the left-hand side [A27] has 120 different values and the right-hand side has only 10. Therefore. that is. . It follows immediately from this theorem that it is also impossible to solve in radicals general equations of degrees higher than the fifth. with m = 5. that he has also established 1 1 1 that R 5 can be equated to ( p + p1 S 2 ) 5 . consequently. But the right-hand side of [A27] has only ten possible values. [The left-hand side of [A27] has 120 different values because y1 can take any of the five values of the five roots. He notes. he turns [A26] inside out. That means the total number of possible values of the left-hand side is 5! = 120. the premise fails that the equation can be solved algebraically. crucially. by y given by [A2]. [A27] where α 4 + α 3 + α 2 + α + 1 = 0. two possibilities for y4 . if the proposed equation is solvable.] Thus we conclude that it is impossible to solve in radicals the general equation of the fifth degree. since the square root has two possibilities. . to express R 5 in terms of the roots y1 . [A28] 1 [For the last time. . and only one for y5 . yielding [A27].168 Appendix A From this expression for y we derive that 1 1 R5 = (y1 + α 4 y2 + α 3 y3 + α 2 y4 + αy5 ) 5 1 1 = ( p + p1 S 2 ) 5 . but we have proved that y must necessarily have this form. y2 . . . There is no way that something with 120 possible values can always equal something with only 10. three possibilities for y3 . multiplied by the five possibilities for the fifth root. this is just [A12b] again. leaving four possibilities for y2 . y cannot have the form we have found.

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. It is worth considering why this doesn’t affect degrees 2. or 1 values.)] . and 4. For them. and we can reapply all of Abel’s arguments that these will all lead to contradic- tions exactly parallel to those for degree 5. with one root y = 0 and the other roots unsolvable. it becomes a sixth-degree equation. 3. but never 3 or 4. the same argument that he has given will continue to apply when n > 5 since Cauchy’s theorem still holds.Abel’s 1824 Paper 169 [Quite simply. Then Cauchy’s theorem says that any function in the solution can take only 7. the number of roots matches the possible values for functions in the solution (respectively 2. 1. 4). con- sider an equation of degree 7. if you multiply an unsolvable quintic by y. 3. In terms of Abel’s proof. For instance. 2. and likewise for higher degrees.

This page intentionally left blank .

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. p. . each x raised to some power. a messy but straightforward process. involving a finite number of radicals. however complex such an expression might be. First. we can always arrange it in the form [B1]. p2 . He did not spell out this argument in his 1824 paper. etc. A rational func- tion is defined as the quotient of two polynomial functions. The basic idea is sim- ple: We assume that a solution has a finite number of terms. so we follow here the account in his 1826 paper. The main idea is to make all the terms (whether inside radicals or coefficients) ratio- nal by always putting them over a common denominator. p1 . which is a sum of terms. What Abel proves is that. f (x). [B1 = A2] where m is a prime number and R. Abel considers a polynomial function. are func- tions of the same form as y. 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 m−1 y = p + p1 R m + p2 R m + · · · + pm−1 R m .

We can now write the algebraic function v as VT .. a “second order” function f ( p1m ) involves first-order functions p1 and f .172 Appendix B similar to the way a rational number is the quotient of two integers. V → V1 = 1 2 m 1 1 v0 + αv1 R n + α 2 v2 R n + · · · + α m vm R n . He now sets up a hier- archy of complexity and defines an algebraic function of the “zeroth order” as f (a . Under these n − 1 substitutions. We can keep going to define an algebraic function of the kth order that has k lev- els of roots-within-roots. d. Vn−1 Abel now shows that the denominator of [B2] is a poly- 1 2 nomial. where f is a rational func- tion of the coefficients a .Vn−1 . where p0 is a function of the zeroth order and f is a rational function of the coefficients 1 in zeroth order.. If we multiply the definition v = VT by 1 = VV11 VV22 . Similarly. where α = 1 but satisfies α n = 1. b. e. . . R n → α n−1 R n . as in [A10]. . c. . when R n → α 2 R n . V2 . . . e). . we call this quite general algebraic 1 function v. [B2] VV1 V2 . . Under the substitution R n → α R n . . To see this. By factoring m and writing R m as a sequence of successive radicals. remember that V = v0 + v1 R n + v2 R n + m 1 1 · · · + vm R n .. . where V is 1 2 m expressed in the form V = v0 + v1 R n + v2 R n + · · · + vm R n . then T V1 V2 . Vn−1 v= . d.Vn−1 . c. a special case of the modern use of the term “algebraic func- tion” for all polynomials f (x. y) = 0. Abel means by “algebraic function” one that can be expressed by rational functions along with taking roots. and likewise for T. . Vn−1 . (Here we have reverted to Abel’s notation in the 1824 paper.. V .) Now Abel considers the series of substitu- 1 1 1 1 1 1 tions that take R n → α R n . V will in general take the n − 1 val- ues V1 . R n → α 2 R n . b. we may assume that m is prime. . Then an algebraic function 1 of the “first order” is f ( p0m ).

would all be functions of order 4. Then m a n+b b Rn = R n = Ra R n . . this same process can be repeated as many times as required until we finally reach polynomials. . we multiply these out 1 and then gather similar terms: VV1 V2 · · · Vn−1 = (v0 + v1 R n + 2 m 1 2 m v2 R n + · · · + vm R n ) × (v0 + αv1 R n + α 2 v2 R n + · · · + α m vm R n ) × 1 2 m 1 (v0 +α 2 v1 R n +α 4 v2 R n +· · ·+α 2m vm R n )×· · ·×(v0 +α n−1 v1 R n + 2 m α 2(n−1) v2 R n +· · ·+α m(n−1) vm R n ) = v0n +αv0n−1 v1 (1+α +α 2 +· · ·+ 2 α n−1 ) + α 2 v0n−2 v2 R n (1 + α + α 2 + · · · + α n−1 ) + · · · = v0n . . [B3] a function of order k. so that it can be written as 1 2 m v = q0 + q1 R n + q2 R n + · · · + qm R n . where a and b < n are integers. where R. R n . we are done. the rational functions in the numerator can be redefined to include the polynomial denominator. . . the denominator of [B2] is a polyno- mial and all the irrational functions are in the numerator. roots of a quartic equation. Thus. and noted that (1 + α + α 2 + · · · + α n−1 ) = 0 [A10]. . In the case of the quintic equation. Furthermore. q 0 . that is. . q 1 . . q 1 . If v0 con- tains further radicals nested inside it. R n . where we 1 2 m have collected the factors of each power of R n .Abel on the General Form of an Algebraic Solution 173 1 2 m becomes V2 = v0 + α 2 v1 R n + α 4 v2 R n + · · · + α 2m vm R n . for we know that there is only a finite number of subradicals. for otherwise we can write m = a n + b. so that only v0n remains. In [B3]. we can always assume that m < n. . . q 0 . This is the crucial realization. R. the 1 numerator is a polynomial in terms of R n and powers of the variables. If v0 is a polynomial. To find the denominator of [B2]. are functions that are of order k − 1 and include the factors absorbing the polynomial denominator. and so forth. [B4] . By the same reasoning.

Because we have factored k with a . Therefore the solution of the cubic equation can be written in Abel’s form. This leaves the fractional b powers of R expressed in terms of lower-order fractions. But R+3 R−3 = 4  2 1 1 a a3 − a31 . The Cardano solution of this equation (see box 2. namely R n .4) is     3 a0 a 02 a 13 3 a0 a 02 a3 y= + + + − + 1. where R− = a20 − 40 + 271 . 3 p2 R + a 1 = 0. Thus p2 = − 3R a1 . Thus y = R+3 + R−3 . as claimed. the highest power 1 n−1 of R n is less than n. write y = p + R 3 + 2 p2 R 3 . and 3Rp22 + a 1 p2 = 0. Hence p = 0. the sum of the 1 roots is 0. but the sum of the roots is also 3 p + R 3 (1 + α + 2 α 2 ) + R 3 (1 + α + α 2 ) = 3 p. [B6] By [A8] in appendix A. 1 2 Now substitute y = R 3 + p2 R 3 into the cubic equation and factor out the various powers of R. Since our cubic equation has no y2 term. and by substitution. R2 − a3 a 0 R − 271 = 0. Then y = R+3 + p2 R+3 = R+3 − a31 R+ 3 . which is exactly the form shown in [B5]. R n . Let us consider some examples.174 Appendix B We can absorb the integral powers of the polynomial Ra into a redefined coefficient q b = q b Ra . .1 shows that the quadratic equation obeys this form. [B5] 2 4 27 2 4 27 1 To show that this really is of Abel’s form. One gets: 1 2 (R + p23 R2 − a 0 ) + R 3 (3 p2 R + a 1 ) + R 3 (3Rp22 + a 1 p2 ) = 0. The case of the cubic equation y3 + a 1 y − a 0 = 0 is only slightly more complicated. One solution of this quadratic is R+ = a20 +  2 1 2 1 a a3 −1 1 1 0 + 271 . This shows that [B4] will take the stated form [B1]. this means that the separate co- efficients are each zero: R + p23 R2 − a 0 = 0. since 1 + α + α 2 = 0. Box 6. b < n.

An! are all the permutations of the original order x1 . If the function is not symmetric. Call the values f (A1 ). . x1 ) = · · ·. the greatest number of such different values of the function cannot exceed the total number of per- mutations. x3 . Let us suppose that . x2 . . If the value of the function is unchanged when the values of the variables are permuted. x2 . f (An! ). where A1 . it does not matter which variable is which: f (x1 . A2 . First we show that p must be a divisor of n!. . which in turn generalizes Ruffini’s results. x5 ) (here for the case n = 5). x2 . Assume that the functions takes p different values. where p < n!. x3 . n!. In general. . . . x4 . x4 . x3 . x3 .. x3 . summarized here and related to the modern language of group theory. if the function is symmetric. . x5 ) = f (x5 . Then some of the n! values must be equal. x4 . x1 . x2 . . . different permu- tations of x1 . . Abel gives a further account. f (A2 ). x2 . f (x1 . . x3 . In his 1826 paper. Cauchy con- siders any function of n variables. x5 will lead to different values of the function. it is said to be a symmetric function. x5 ) = f (x2 . x4 . x4 . and so on through all the 5! = 120 possible permutations of the five variables. Appendix C Cauchy’s Theorem on Permutations Abel’s 1824 paper uses Cauchy’s argument about permuta- tions. That is.

see the works listed in the notes. dividing them into p sets. f (A1 ) = f (A2 ) = · · · = f (Am ). we take the “quotient” of the 24-hour clock divided by 2. pm = n! and so p is a divisor of n!. 13 + 48 = 61 o’clock. namely |H|. . This gives another m equal values (though in general not equal to the first m). This is only a brief clar- ification. A good example of this is the arithmetic of a clock. Since the cosets divide up G into these exhaus- tive and exclusive classes. 191–192. Let H be the sub- group of permutations that fix one of the p possible values of our function. “left” cosets are distinguished from “right. Consider any subgroup H contained in G. A2 → Am+2 . . If we then shift to a 12-hour clock. each numeral is a coset. etc. it is natural to think of them as the “quotient G/H. pp. finally exhausting the n! possibilities. let us make the successive substitutions A1 → Am+1 . representing an equivalence class of each hour: 13 o’clock is equivalent to 13 + 24 = 37 o’clock.”) The argument just given shows that these cosets partition the elements of G into exhaustive.” (If the group G is noncommutative. 193–194). say f (A1 ). each having m elements. Am → A2m . . We can then continue this process of substitution. In modern group theory. and so forth. In this series.” which is a group when H is a normal subgroup (defined on pp. On a 24-hour clock.176 Appendix C the first m values are equal. as are 25 and 37 o’clock. must be a divisor of |G|. . Then the procedure described above corresponds to partitioning the group G into “cosets of H in G. this is called Lagrange’s Theo- rem. for more examples and details. Lagrange’s Theorem states that the number of elements of H. Therefore. exclusive classes of p elements and so p must be a divisor of |G| = n!. . since now 13 and 1 o’clock are also equivalent. 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).

of course.Cauchy’s Theorem on Permutations 177 In general. As the same chemi- cal atoms can form different molecules. multiply this antisymmetric function by a symmetric function. In fact. appendix A). Then the quotient group G/H and H are constituent “sub-molecules” of G. Consider a group G that contains a normal subgroup H that is “proper” (H is neither simply G nor the identity). We can also. until we cannot break them down fur- ther: the collection of indecomposable pieces that we get are the “atoms” of the group G. in which all five variables appear antisymmetrically. Jordan and Hölder’s famous theorem proves that this collection of “atoms” is determined entirely by the group G. the same mathemat- ical “atoms” can form different groups. opposite in sign. Continue breaking these “sub-molecules” down in the same way into ever smaller pieces. . For exam- ple. Abel’s proof means that an equation is solvable in radicals if and only if the “atoms” of its group are all abelian (in particular. Other functions may have two values. cyclic and having orders that are prime numbers. We began by considering a symmetric function of n vari- ables. the function √ changes √ sign but is equal in absolute magnitude: S → − S. which has only one value when we permute the variables. which will not change upon permuta- tion (Abel uses just this function in his [A18]. mean- ing that if we make the permutation x1 → x2 . if the group is finite). consider the function that Cauchy and Abel call √ S = (x1 − x2 )(x1 − x3 )(x1 − x4 )(x1 − x5 )(x2 − x3 ) · · · (x4 − x5 ). From this standpoint. finite groups can be compared to molecules that can be analyzed by breaking them into smaller molecules and ultimately into their constituent atoms. any interchange of two variables will produce two possible values for the function.

which we will write as f (Am ) p = f (Am )0 . f (A m ) = f (A m ) j . any arbitrary permutation is composed of a certain number of transpositions between the variables. as will be true in the case of the quintic. Cauchy concludes. x2 . we can always find two integers a . or f (A m ) = f (A m ) j . even if it is a divisor of n!. This number can be even or odd. If we assume that the function takes on fewer than p values. x4 . Therefore. Likewise. where b is any integer. x5 ). r < p − 1. meaning that when it is applied p times we return to the starting point. From our last result. and we define j = r + p − r . x3 . x3 . For example. we know that this function is unchanged by a permutation of order p > 3. Not only are two of the values equal. b such that b j = pa + 1. Hence f (A m )b j = f (A m ) pa +1 = f (A m ). we must trans- pose variables. Now to make such a permutation. In what follows. Let us imagine a function that takes fewer values than p (assuming now that p > 3. Then apply the permutation A m to both sides of this equation ( p − r ) times: f (A m )r + p−r = f (A m )r + p−r . . to go from f (x1 . but also f (A m ) = f (A m )b j . if the function takes fewer values than p. say f (A m )r = f (A m )r . where 0 < r. it is not changed by a permutation of order p. which we showed is also equal to f (A m ) = f (A m ) pa . for which n = p = 5). so that the permuta- tion A m does not change the value of the function. each multiple of j representing that many times around the circle. f (A m ) p = f (A m ). Since p is a prime number.178 Appendix C It is not always possible to make a function that has any given number of values larger than 2. then two among those values must be equal. Let Am be any permutation of order p. since (r + p − r ) = p. x1 . This means that adjacent values of the function are equal. x4 . we must transpose x1 ↔ x2 . let p be the largest prime less than or equal to n. f (A m ) pa = f (A m ) pa +1 . x5 ) to f (x2 .

The upshot is that. xb → xc → xa . xb → xa . for after p such transpositions. which are just transpositions. After the first three terms. x p → xa → x p or (abcd · · · p) ∗ (ap · · · edbc) = (acb). given any 3-cycle. This too is a substitution of order p. and thus neither can a permutation of order 3. . . as they are called in chapter 8) whose product is the given 3-cycle. xb → xc . It can be written as the transposition . . . by the result in the preceding paragraph. xc → xb . xb → xa . xc → xd . . in general. the variables remain unchanged after both sub- stitutions are performed. consider a second series: xb → xc . we would come around to the starting po- sition. xa → xc . Consider the one we discussed earlier. . x f → xg → x f . . Since neither permutation of order p can change the value of the function (by our earlier result). Clearly. this is (abcd · · · p). for after p such substitutions. xb . xc → xa . xe → xd . they are not. . . Indeed. . . xd → xb . x f → xe . for we could have used any combination of variables in place of xa . xc . xc → xb . . xc → xd → xb . . we re- turn to the starting point. . So these two permutations per- formed in sequence yield xa → xc . neither can both of them. Any permutation of order p must take the form of these examples. replacing each vari- able with the next in the list. xe → x f → xe . We now show that a permutation of order 3 can always be writ- ten as the result of two permutations of order 2. xa → x p or (ap · · · edbc). Similarly. To use the notation intro- duced in chapter 8. Now notice what happens when we perform the first and then the second substitution in that order: xa → xb → xc . So both will leave the function unchanged. . this is a substitution of order p. with all the other variables unchanged. . Though these examples may seem arbitrary. it is a slightly varied version of the first substi- tution (moving xa to the end). let xa → xb .Cauchy’s Theorem on Permutations 179 Now consider two permutations of order p: In the first. we can al- ways find two permutations of order p ( p-cycles. x p → xa .

xb → xa . verifying the assertion: (ab) ∗ (cd) = (abc) ∗ (cad). To check this. since two transpositions (or any even number) do not change the function. giving us the same result: (ab) ∗ (bc) = (acb). Remember that all this followed if p ≤ n. either the function has five values or it has at most two values. to use the term introduced earlier. it is unchanged by an even number of transpositions. Thus. restoring the original sign. It can also be accomplished by xa → xb → xc followed by xc → xa → xd . the conclusion is that the function we have been discussing can have at most two values. Note that. xc → xb .180 Appendix C xa ↔ xb followed by the transposition xb ↔ xc . Similarly. whereas any odd number of trans- positions must change the sign of the function. Cauchy and Ruffini’s insight is crucial to the last stage of Abel’s proof. In either case. if n = 5. then. xa → xb → xc . They are anti- symmetric. therefore. it suffices to show that any two distinct transpositions can be written as two 3-cycles. so any permutation of order 3 is equiv- alent to two permutations of order 2 (transpositions). p = 5 also. To prove this. Two transpositions change the sign twice. so that in the case of the quintic. this is quite general in form. 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. Again. Consider. do them in the order specified. xa ↔ xb followed by xc ↔ xd . But we just learned that the function is unchanged by a permuta- tion of order 3. In this form. each transposition changes the sign of the function. . it is unchanged by the product of two transpositions.

For the Indo-European roots. 1987). Euclid: The standard annotated edition is The Thirteen Books of Euclid’s Elements. Pythagoras—A Life (London: Routledge and Kegan Paul. Heath (New York: Dover. Long (Cambridge: Cambridge University Press. For an engaging modern encounter with this classic.” . Huntley. E. (Cambridge. The Wonder That Was India (New York: Grove Press. 1999). 1961). Basham. Thomas L. see The Pythagorean Source- book and Library. tr. 1970). see Julius Pokorny. L. Pythagoras—A Short Account of His Life and Philosophy (New York: Krikos. A. ed. A. Lore and Science in Ancient Pythagoreanism. Huffman. 1972) and also C. The Divine Proportion: A Study in Mathematical Beauty (New York: Dover. Notes 1 The Scandal of the Irrational Pythagoras of Samos: Scholars have questioned every aspect of the tradi- tional account. which gives the proof cited by Aristotle (Prior Analytics 41a. Indogermanis- ches Etymologisches Wörterbuch (Bern: Francke Verlag. 26–27) at 3:2. A. Edwin L. 236–237. 1954). 66–87. I thank Eva Brann for drawing my attention to the connection between arithmos and “rite. 1979) and Leslie Ralph. MI: Phares Press. see Robin Hartshorne.” in The Cambridge Companion to Early Greek Philosophy. tr. Geometry: Euclid and Beyond (New York: Springer-Verlag. Minar Jr. ta: A. Pythagorean mathematics and the golden ratio: See H. 113. ed. R . 1959). MA: Harvard University Press. See Walter Burkert. Ken Sylvan Guthrie (Grand Rapids. “The Pythagorean Tradition. For a collection of original sources. 1956). 2000). 1:56–60. Other accounts include Peter Gorman.

Greek Mathematical Thought and the Origin of Algebra. NC: University of North Carolina Press. See also Jacob Klein. Greek mathematics: For a helpful survey. Pentagons and irrationality: See Kurt von Fritz.. 242– 264 (1945). 81–94 (1999). The Evolution of the Euclidean Elements (Dordrecht: D. 1930). The “young irrationals” are discussed in Plato’s Republic. Anabasis 2: 21–29. so aporia signifies being stuck. A Commentary on Plato’s Meno (Chapel Hill.” Isis 90:1. note 10. from the context it is clear that he understands it to be noble. Book VII. hereafter cited as MT. Socrates as midwife. 1965). see Xenophon. MA: MIT Press. although the propositions came from Socrates. 1991). 28–34. tr. 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. and Polity: Francis Bacon’s Account of Eros. William Thomson (Cambridge. chapter 2. see my “Wrestling with Proteus: Francis Bacon and the ‘Torture’ of Nature. especially his point that.182 Notes to pp. cited hereafter as HM. MA: Harvard University Press. 103–107. Though Bacon did not use the charged term “torture” in this context. Mathematical Thought from Ancient to Modern Times (New York: Oxford University Press. which includes a plausible account why Theodorus stopped at 17 (181–193) and a critique of earlier explanations (109–130).” Annals of Mathematics 46. 62–108. Eva Brann (New York: Dover. not base or abusive. revised by Uta C. Merzbach (New York: John Wiley. Pappus on the irrational: The Commentary of Pappus on Book X of Euclid’s Elements. Boyer. A poros is a means of passage. . Reidel. 1992). however. for the story of Theaetetus. see the classic book by Carl B. like a bridge or ferry. 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. See also Morris Kline. For Meno’s character. For the later thinkers. This also applies to his daring metaphor of philosophical inquiry as “parricide”. 2000). second ed. Science. see Theaetetus 142a–150b. see Wilbur Richard Knorr. 534d. translation revised thanks to Bruce Perry. 64–65. see my paper “Desire. Plato uses it to describe the ardu- ous ordeal of dialectic and inquiry. A History of Mathematics. 1–17 Plato: The episode of the slave boy is in Meno 84d–85b. see Phaedrus 253d–254e. 1972). 3–113.” Interpretation 26:3. 100–119 (Euclid). For the story of the black and white horses. “The Discovery of Incom- mensurability by Hippasus of Metapontum. 333–352 (1999). 1975). tr. 43–99 (especially 72–74 on incommensurability). For a helpful discussion of Theodorus and Theaetetus. Another essential classic is Jacob Klein. unable to cross over. “the assent and the rejection came from nobody but the boy himself” (105). 150c– 151d.

” Mathematical Chronicle 7. 2 Controversy and Coefficients Babylonian mathematics: HM. For a discussion of an important text. MT. A History of Algebra from al-Khwārizmı̄ . 1:137. 2000). tr. with valuable excerpts and translations. see Sir Thomas Heath. Richard Rees (London: Oxford University Press. 1200– 1800. propositions 1 (on the indef- inite divisibility of any magnitude) and 115 (the infinite number of kinds of irrationals). van der Waerden. D. 23–42. chapter 7. 188.” Studies in Philology 98:4. For an interesting compar- ison of Greek and Babylonian approaches to mathematics. who argues that the “inherent symbolism of the Chinese language” moves in the direction of modern algebraic symbolism. which is understandable given their primal assumption that musical interval is inherently a ratio. 411. 1:90– 91. Science and Civilisation in China (Cambridge: Cambridge University Press.√I have found no passage where an ancient writer noted the near equality of 2 and the tritone. West. 1989). ed.” Studia Leibnitiana 29. J. Al-Khwārizmı̄: See my Labyrinth. For the development of the Greek theory of ratio. A History of Greek Mathematics (New York: Dover. with a comment on the cubic problem on 125–126. MA: Harvard University Press. Hoe. see Midhat Gazalé.Notes to pp. L. 428–456 (2001). L. 189–197 (1998). Seventy Letters. 1969). 154–157. Struik (Cambridge. and the invaluable collection of original texts with commen- tary in Greek Musical Writings. 3:1– 170. Andrew Barker (Cambridge: Cambridge University Press. see J. 112–127. 233–242. “Eudoxos and Dedekind: On the Ancient Greek Theory of Ratios and Its Relation to Modern Mathematics. Euclid on irrationals: See The Elements. His work (called in Arabic Al-jabr wa’l muqabālah) is included in A Source Book in Mathematics. Ancient Greek Music (Oxford: Clarendon Press. 1959). 18–25 183 “Nature on the Rack: Leibniz’ Attitude towards Judicial Torture and the ‘Torture’ of Nature. Book X. 1965). and Howard Stein. 419. Number: From Ahmes to Cantor (Princeton: Princeton Uni- versity Press. See also B. including also Euclid’s musical treatise in 2:190–208. “The Jade Mirror of the Four Unknowns—Some Reflections.” Synthese 84. ed. 3–14. see Joseph Needham. 1981). see Simone Weil. and “Proteus Un- bound: Francis Bacon’s Successors and the Defense of Experiment. For a general survey of the context of numbers and irrationality. Chinese algebra: For a general overview. 163–211 (1990). 55–60. Greek music: See M. 1992). 125–156 (1978).

129–164 (1988). 548. L. 88–117. For the connections with the development of printing. 1980). Augusto Marinoni. 10–16. reprint Maslianico: Dominioni. 1979). Parker (New York: Garland. ed. A.18–21.” in Accounting History from the Renaissance to the Present: A Remembrance of Luca Pacioli. Gies. Johnston. 1965). see R. Shylock on commerce: The Merchant of Venice. at the end of one year the account will be worth e = 2. and Anna Maria Brizio. S. 1980). e: See Eli Maor. “The Art of Algebra from al-Khwārizmı̄ to Viète: A Study in the Natural Selection of Ideas. Emmett Taylor. as does B. 1996). R. Pacioli includes idealized fonts based on geometry at the end of his De divina proportione (Venice: 1507. Vahabzadeh. 1987). Eisenstein. . .” History of Science 26. H. If $1 is invested in a bank account at 100% per annum. continuously. Fibonacci: See HM. and J.184 Notes to pp. Yamey. But if the interest is compounded at every instant. 1982). Leonardo the Scientist (New York: McGraw-Hill. Omar Khayyam the Mathematician (New York: Bibliotheca Persica. Leonardo da Vinci and Pacioli: See Carlo Zammattio. The History of Mathematics: A Reader (London: Macmillan. T. G. 1980). see Karen Hunger Parshall. Venice 1494 (Venice: Abrizzi. e: The Story of a Number (Princeton: Princeton University Press.718 . See also R. 2000). No Royal Road: Luca Pacioli (New York: Arno Press. at the end of one year its value will be $2. 1994). Gies and F. 1994). . There is a valuable collection of essays in Roshdi Rashed. and Emanuel Winternitz. “Pacioli’s Lecacy. Rashed and B. Macve. 1998). 241–243. 1969). and Richard H. The Printing Press as an Agent of Social Change: Communications and Cultural Transformations in Early-Modern Europe (Cambridge: Cambridge University Press. Leonard of Pisa and the New Mathematics of the Middle Ages (New York: Crowell. his assertion remains controversial. 26–28 to Emmy Noether (New York: Springer-Verlag. contains a translation of the portions of the Summa relating to double- entry bookkeeping. Pacioli: R. 3–30.iii. Pacioli on Accounting (New York: McGraw- Hill. Brown and K. 1967). John Fauvel and Jeremy Gray. see E. Bishop. Luca Pacioli’s Exposition of Double- Entry Bookkeeping. Lee. 9–33. S. Khayyām: For a valuable collection of translations with helpful commen- tary. The De- velopment of Arabic Mathematics: Between Arithmetic and Algebra (Dordrecht: Kluwer Academic. Leonardo da Vinci as a Musician (New Haven: Yale University Press. dollars. 254–257. 95–171. For an excellent survey. I. com- pounded annually. Rashed argues that Omar the poet and Omar the mathematician may not be the same person. A.

see Morton J. 255–256. See also Anthony Grafton. see Mark A. 19:3. 1953). For his colorful autobiography. which also includes other documents from the controversy on 253–263. 82–98. Solutions of cubic and quartic equations: For a complete and lucid account. My ac- count of completing the cube draws on Parshall. “The Art of Algebra.” For the connection between economic and mathematical developments.” Mathematical Intelligencer. and Quartic. 410 (1959). 2000).” 144– 147. Tartaglia’s poem: Fauvel and Gray. 1990). integration of real and nominal accounts [in double-entry bookkeeping] far surpasses every other aspect of accounting development. which also clari- fies Pacioli’s plagiarisms from Piero’s works and discusses the relation of algebra and geometry in Piero’s time. see the classic account by J. 282–288. 274–276 (1958). or the Rules of Algebra. see Girolamo Cardano. Ars magna. see also Frank Swetz. History of Mathematics. 133–154. which treats the cubic and quartic equations on 59–107 (the commercial problem cited is on 69–70). 1977). 3–18. Uspensky. Cardano’s Cosmos: The Worlds and Works of a Renaissance Astrologer (Cambridge. IL: Open Court. History of double-entry bookkeeping: See Michael Chatfield. 1948). 33–40 (1997). and MT.Notes to pp. William Dunham gives an engaging presenta- tion in Journey through Genius: The Great Theorems of Mathematics (New York: John Wiley. Richard Witmer (New York: Dover. 28–40 185 Piero della Francesca: For an outstanding account of his great mathemat- ical achievements. Capitalism and Arithmetic: The New Math of the 15th Century (La Salle. The Book of My Life (New York: Dover. Theory of Equations (New York: McGraw-Hill. tr. 263–272. Cardano: See Girolamo Cardano.” American Mathematical Monthly 66. Cardano. See also HM. Peterson. 1987). 9. the Gambling Scholar (Princeton: Princeton University Press. 1993). “Nature does not permit it”: Cardano. A History of Accounting Thought (Huntington. Hellman. who emphasizes that “the significance of the . MA: Harvard University Press. . Cubic. 1962). NY: Krieger Publishing. For a neat exposition of the factorization of these equations. giving very helpful excerpts from the invective of the combatants and also the complete text of Cardano’s Book on Games of Chance (first printed posthumously in 1663). Hellman shows the failure of this technique to the quintic in “The Insolvability of the Quintic Re-Examined. which unfortunately ren- ders Cardano’s mathematical expressions into modern algebraic notation. T. “The Geometry of Piero della Francesca. Ars magna. V. instead of retaining his own rhetorical mode of expression that does not rely on such symbolism. “A Unifying Technique for the Solution of the Quadratic.” American Mathematical Monthly 65. . and also Øystein Ore. .

Impossible Numbers. Kepler’s Geo- metrical Cosmology (Chicago: University of Chicago Press.” Bulletin des sciences mathématiques (2) 54. in his Opera Mathematica (Hildesheim: Georg Olms. giving x = 2 sin θ. 1993). E. 674– 692 (1997). tr. 309–311. 27–39. yielding van Roomen’s equation (see box 2. 1580–1700. Symbols.186 Notes to pp. Field (Philadelphia: American Philosophical Society. which his formula could then express in terms of x = 2 sin θ. 54–59 (2000). discussed in HM. 91–94. “The Relation between Geometry and Algebra: Cardano and Kepler on the Regular Heptagon. 1997). and also by Guido Vetter. and D. 1–29 (1997). 99–105. I cite Viète’s “Introduction to the Analytical Art” from the translation in this volume by J. see my paper “François Viète. n. and Sys- tems: Parallels between Cryptanalysis and Algebra. E.” Cryptologia 21:1. 34–62. 1994). To solve it. Duncan. 277–283 (1954). Walker. such as sin nθ = n cosn−1 θ sin θ − n(n−1)(n−2) 3·2·1 cosn−3 θ sin3 θ + · · ·.7). P. J. ed. 117–123 (Cardano). NY: Dover. A History of Mathematical Notations (Mineola. J. Field. “Sur l’équation du quarante-cinquième degré d’Adriaan van Roomen. which discusses John Wallis’s legal account of “species” at 321–322. he found θ = (arcsin K )/45.” Mathematical Intel- ligencer 22:4. Arzt. For Kepler’s treatment of the heptagon. 219–242. For translations of the original Viète documents. and (in greater detail) in “Secrets. V. Field. 3 Impossibilities and Imaginaries Van Roomen’s problem: See Viète’s response. J. qvod omnibvs mathematicis totivs orbis contruendum proposuit Adrianus Romanus” (1595). A. M. I have dis- cussed the connections between Viète’s work in codebreaking and algebra in my Labyrinth. Aiton. 60–79. 1997). 41–49 Viète and the “new algebra”: The seminal work is Klein. J. which gives the full quotations and sources. See also Helena M. Winfree Smith (315–353). 161–185.” in Girolamo Cardano: Philosoph. “Kepler’s Celestial Music. Father of Modern Cryptanalysis—Two New Manuscripts. Kepler: See my paper “Kepler’s Critique of Algebra. V. Brill. History of symbolic notation: See Florian Cajori. “Ad Problema.” in his Studies in Musical Science in the Late Renaissance (Leiden: E. 1978). 10. 181–187 (Viète). Symbols. Greek Mathemat- ical Thought. 73–83. and J. For further . Viète set K = sin 45θ. and Geometric Entan- glements (Cambridge: Cambridge University Press. see Johannes Kepler. The Harmony of the World. 2001). Pycior. 1988). V. Viète had derived expressions for the sines or cosines of multiple angles.” Isis 88. Kessler (Wiesbaden: Harrassowitz Verlag. Naturforscher. 305–324. and also my Labyrinth. where the subsequent signs alternate and the coefficients are the alternate numbers in the arithmetic triangle (Pascal’s triangle).

For Gauss’s account of negatives and imaginaries. ed. and the Triskaidecagon. see Barry Mazur. Whiteside (Cambridge: Cambridge University Press. Stillman Drake (New York: Doubleday. Latham (New York: Dover. William B. For Galileo’s attitude toward mathematics. the Heptagon. 160–161 (Descartes’s rule of signs). 1972). Ewald (Oxford: Clarendon Press. 287–289 (Bombelli). Ernan McMullin (New York: Basic Books. See also Paul J. see The Mathematical Pa- pers of Isaac Newton. Greek Mathematical Thought. 12–21 (2002). D. 1673–1683. Girard: See his 1629 treatise Invention nouvelle en l’algèbre. 49–60 187 discussion of the heptagon and its relation to angle trisection. 185–194 (1988). 1986). 2002). 1954). Descartes’s “relativity”). 95. 1:310–313.” Mathematical Intelligencer 24:1. Negative and imaginary quantities: HM. 1957). 305–306 (Girard). using conic sections).Notes to pp. See also Pycior. Imagining Numbers (particularly the square root of minus fifteen) (New York: Farrar √Straus Giroux. 1998). 176–192 (methods for solving higher- degree equations). in The Early Theory of Equations (Annapolis. 232–255. see Federica La Nave and Barry Mazur. MD: Golden Hind Press. T. Descartes: See The Geometry of René Descartes. 220–239 (solution of a special sixth-degree equation. 1967). 307 (Leibniz on the amphibian). 197–211.” in Galileo. see the excellent anthology From Kant to Hilbert: A Source Book in the Foundations of Mathematics. 4 Spirals and Seashores Newton: For his Lectures on Algebra. 174–175 (introduction of “imag- inary” roots). Impossible Numbers. Symbols. David Eugene Smith and Marcia L.” Ameri- can Mathematical Monthly. 240–241 (“the general method to construct all problems”. see Andrew Gleason. the pleasure of discovery). “Angle Trisection. Man of Science. 22–37 (the locus problem for three or more lines). tr. Bombelli: For a thoughtful introduction that includes translated selections. 5:130–135. “Reading Bombelli. 219–220 (negative numbers in Hindu mathematics). tr. ed. 565 (two sample passages showing his treatment . Nahin. 1996). 162–175 (methods for in- creasing or scaling the value of roots. Galileo on the Book of Nature: See “The Assayer” in Discoveries and Opin- ions of Galileo. An Imaginary Tale: The Story of −1 (Princeton: Princeton University Press. 276–278 (Chuquet). see Klein. For an engaging account of the developing conception of imaginary numbers. For a discussion of Descartes’s concept of number. 237–238. Boyer. see Carl B. ed. “Galileo’s Place in the History of Mathematics.

which I also follow in my exposition. Bring.” College Mathematics Journal 28:1. 5 Premonitions and Permutations Lagrange and Vandermonde: See the classic sturdy by Hans Wussing. 58 (1999). 100 Great Problems of Elementary Mathematics.” Archive for the History of Exact Sciences. 1999). 485 (geometrical synthesis). Abe Shenitzer (Cambridge. for Jerrard’s mistaken solu- tion. For further discussion of the significance and validity of lemma 28. tr. Their treatment is based on Uspensky. 113–133. especially 182–186 on Gauss’s original proof. The Genesis of the Abstract Group Concept. Bernard Cohen and Anne Whitman (Berkeley: University of California Press. There is a nice one-page proof of this theorem by Uwe F. 511–513 (lemma 28). tr. . David Antin (New York: Dover. 1984). 71–79. History of Algebra. 432–434. 293–297. 81. below. 600– 606. 1994). 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. “The Integrability of Ovals: Newton’s Lemma 28 and Its Counterexamples. The Fundamental The- orem of Algebra (New York: Springer-Verlag. tr. 483–485 (Newton’s solution to locus problem). see my Labyrinth. John Meldrum (New York: Springer- Verlag. For the relation of Newton’s mathematical preferences to his other projects. 1965). 479–499 (2001). 76–83. “A Proof That Polynomials Have Roots. Theory of Equa- tions. see my paper “The Validity of Newton’s Lemma 28. and MT.” Historia Mathematica 28. Tschirnhaus. The Principia: Mathematical Principles of Natural Philosophy. See also Heinrich Dörrie.” though Lagrange never expressed this opinion explicitly. Elements of the History of Mathematics. Fundamental Theorem of Algebra: For a complete treatment of all of Gauss’s proofs. tr. Quotations in the text are from Isaac Newton. 108–112. 55. see Benjamin Fine and Gerhard Rosenberger. His paper “Réflexions sur la résolution algébrique des équations” (1770–1771) is in Œuvres de Lagrange (Paris: Gauthier-Villars. 3:205–422. Citations in the text are taken from van der Waerden. and Jerrard: HM. 60–78 of equations). History of Algebra. 1997).188 Notes to pp. and also Bruce Pourciau. Mayer. 1869). see chapter 9 notes. Leibniz on the quintic equation: See the helpful historical overview in Nicolas Bourbaki (the pseudonym of a group of French mathematicians). van der Waerden. I. 215–219 (2001). 69–80 at 74. MA: MIT Press.

K. A. 178–183 (the arrival of Holmboe). 1981). Gauss: W. 1965). Victoria: Department of Mathematics. History of Al- gebra. for a popular history through Andrew Wiles’s 1993 proof. 100 Great Problems. 1957). Montucla (1725–1799): Cited from the excellent article by Raymond G. 2000). Bühler. 253–277 (1980).” in First Australian Conference on the History of Mathematics. 1962). van der Waerden. 36 (Gauss’s lack of interest in combinatorics). Still useful is the briefer. John N. Another useful article is R. 5–31. Disquisitiones Arithmeticae. W. “Paolo Ruffini’s Contributions to the Quintic. Ruffini: See Wussing. and particularly Ayoub. 2 volumes. 445 (a passage added in proof. Gauss on the quintic: Carl Friedrich Gauss. Niels Henrik Abel: Mathematician Extraordinaire (Minneapolis: University of Minnesota Press. 1902). There is also an excellent account of this in Felix Klein. 6 Abel’s Proof For a comprehensive account of Abel’s life and its Norwegian context. 77). Group Concept. earlier work by Øystein Ore. Niels Henrik Abel and His Times (Berlin: Springer-Verlag. 1915–1953). Ayoub. 239–240 (Abel’s false solution to the quintic). Famous Problems of Elementary Geom- etry. Clarke (New Haven: Yale University Press. There are interesting essays also in the volume marking the centenary of Abel’s birth. “Ruffini and the Quintic Equation. Niels Henrik Abel: Mémorial publié à l’occasion du centénaire de sa naissance (Kristiania: Jacob Dybwad.” including the quote from Ruffini at 263 (“behold a very important theorem”). “Paolo Ruffini’s Contributions to the Quintic. E.” Archive for History of Exact Sciences 23. which points out a number of errors in Ruffini’s work that substantiate the attribution of the theorem to Abel alone.Notes to pp. Fermat’s Theorem: For Gauss’s proof in the case n = 3. For a helpful simplified discussion of the 17-gon and the cyclo- tomic equation. . ed. according to Bühler. 100 Great Problems. at 257. To my knowl- edge. none of Ruffini’s papers has been translated into English. tr. see Arild Stubhaug. tr. see Dörrie. 1981). 297 (Abel and Fermat’s Last Theorem). Bryce. 24–41. 177–184. W. 83–85. see Dörrie. Arthur A. they are available in the original Italian in Opere Matematiche di Paolo Ruffini (Palermo: Tipografia Matematica di Palermo. 407–460 (general discussion of the problem of equal division of a circle. Beman and David Eugene Smith (New York: Chelsea. also known as the cyclotomic equation). Gauss. Gauss: A Biographical Study (New York: Springer-Verlag. 80–84. 96–104. 79–88 189 G. Crossley (Clayton. Monash University.

see his 1826 letter to Holmboe. Abel on commutativity: My translation of the opening of his 1828 paper “Mémoire sur une classe particulière d’équations résolubles algébrique- ment. France: A Modern History (Ann Arbor: University of Michigan Press. Évariste Galois 1811–1832 (Boston: Birkhäuser Verlag. 1997). Rosen. 81–103 (1994) and Michael I. see the classic work by Albert Guérard. L. Lie (New York: Johnson Reprint. Abel’s Theorem: Modern treatments tend to subsume Abel’s own work in Galois theory. see the citations below under Galois Theory and J. though sometimes in an unfamiliar notation. French history during the time of Abel and Galois: For a superb overview. Some older textbooks give an account of Abel’s work. 1965). “Niels Hendrik Abel and Equations of the Fifth Degree. 468 (appeal of the four French mathematicians to the Swedish king).” in Œuvres Complètes. Green. The Theory of Equations (London: Longmans. Sylow and S. For Abel’s work on this problem. Dörrie. “On the Ruffini-Abelian Theorem. 395–420 (Abel in Paris). 1959). 2:217–243 at 218. Quotes from Abel: 471 (monstrous egotists). see Lars Gårding and Christian Skau. 1996). 2:478–507 at 478. 495– 505 (1995).” Bulletin of the American Mathematical Society 2. 88–108 see Simon Singh. Sylow notes (1:293) that Abel first commented on Ruffini in 1826 in an unsigned note in Férussac’s Bulletin. 471 (quite alone). See also William Snow Burn- side and Arthur William Panton. 100 Great Problems.” American Mathematical Monthly 102.” Archive for the History of Exact Sciences 48. 329–331 (Crelle). cited in Œuvres Complètes de Niels Henrik Abel. 288 (Lafayette). 282–305. “Niels Henrik Abel and Solvable Equations. 398–402 (billiards and theater). 278–296 (the period 1814–1848). 475–493 (Abel’s death and his despair). Fermat’s Enigma (New York: Walker.190 Notes to pp.” Œuvres Complètes de Niels Henrik Abel. 200–221 (1896). Abel. 2:254–255. 7 Abel and Galois Abel’s later life: Stubhaug. ed. gives an account of Leopold Kronecker’s clarified version of the theorem. 1928). which includes a helpful . 116–127. 3 (Capetian dynasty). Abel and Ruffini: The quote is my translation from Abel’s 1828 paper “Sur la resolution algebrique des équations. For helpful accounts phrased in the language of modern algebra. 409 (on Cauchy). 424 (poorer than a church mouse). Pierpont. 474 (correspondence with Legendre). Galois’s life: The most reliable recent study is Laura Toti Rigatelli.

John E. for his acquaintance with Abel. found in a brief form in his classic lectures to schoolteachers. 107–118 (1983). 136–149 (1982). R.3. 362–377. 278–285. 1959). Maxfield. and very fully (though with notation that is quite hard to follow) in his Lectures on the Icosahedron (New York: Dover.” Bulletin of the London Mathematical Society 15. See also the helpful articles by René Taton. 84–102 (1982) and “The Short Life of Évariste Galois. 1986). Raspail: For his account of Galois in prison.Notes to pp. Edwards. 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). Hedrick and C.” Scientific American 246:4. Abel. Among earlier brief biographical works. several stand out as especially accessible. see also George Steiner. Saul Stahl. Men of Mathmatics (New York: Simon & Schuster. Cardano. 410–411. is widely known. 98– 100. “Évariste Galois and His Contemporaries. n. Abstract Algebra and Solution by Radicals (New York: Dover. see Stubhaug. E. 416–417 (Galois’s comment on Abel’s death). see Toti Rigatelli.” American Mathematical Monthly 89:2. Galois. 1989). 1992) is particularly suitable for self- study. . though it propagates many of the myths concerning Galois. 1956). Noble (New York: Dover. There is a helpful corrective in Tony Rothman. 8 Seeing Symmetries My treatment of the relation between symmetries and geometric fig- ures is inspired by Felix Klein’s treatment of the symmetries of regular solids as paradigms of the symmetries of equations. 108–130 191 bibliography. Maxfield and Margaret W. 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.” Archive for the History of Exact Science 41. I was greatly helped by Hartshorne’s treatment of the symmetry groups of polyhedra in his Geometry: Euclid and Beyond. 163–169 (1990). A. Grammars of Cre- ation (New Haven: Yale University Press. 469–480. see David Eugene Smith. 1:101– 143. tr. also included in his Science à la Mode: Physical Fashions and Fictions (Princeton: Princeton University Press.).d. There are many treatments. “Genius and Biographers: The Fictionaliza- tion of Évariste Galois. 207–212. For Galois’s final letter. 1997) is an admirably written textbook that includes valu- able excerpts from the writings of al-Khwārizmı̄. Introductory Modern Algebra: A Historical Approach (New York: John Wiley. “A Note on Galois Theory. Elementary Mathematics from an Advanced Standpoint. 3–20. for poetic aspects of Abel and Galois. 2001). Galois. and Harold M. A Source Book in Mathematics (New York: Dover. Abel. Eric Temple Bell.

N. “On the Nonsolvability of the General Polynomial. 1926). Lillian R. 1950). Équations Algébriques et Théorie de Galois (Saint-Amand-Montrond: Librarie Vuibert. is quite accessible. Charles Robert Hadlock. which is thorough and rich in examples. German-speaking readers may find helpful N. I. 1960). Old textbooks sometimes present the theory less abstractly: See Leonard E. M. tr. 1997). 1996). Littlewood gives an interesting overview in The Skele- ton Key of Mathematics: A Simple Account of Complex Algebraic Theories (New York: Harper. long out of print. Galois’ Theory of Algebraic Equations (London: Longman Scientific Technical. 65–76. though somewhat more abstract in its approach than the Maxfields’ book. and R. For some helpful articles. 1962). Freeman. see Raymond G. which is ori- ented to the symmetry concerns of chemists. Burn- side and Panton. 244–305. Dickson. 1934). Fundamen- tals of Galois Theory. Évariste Galois et la Théorie des Équations Algébriques (Paris: Gauthier-Villars. Grundzüge der Galois’schen Theorie. tr. expressed in a kind of verse with drawings. Leo F. gives a lucid outline in sixty pages. Other treatments of Galois theory: A full listing of the vast number of treat- ments is scarcely possible here. and Jerry Shurman. Algebraic Equations: An Introduction to the Theories of Lagrange and Galois (New York: Dover. 1985). but I would like to mention some that I found helpful. 1932). Galois and the Theory of Groups: A Bright Star in Mathesis (Lancaster. 1980). ed. see especially Nathan Jacobson. For more modern treat- ments that are friendly but still rigorous. second edition (New York: W.” American . H. second edition (New York: Wiley. I would also recommend Jean-Pierre Tignol. Boron (Groningen: P. 1973). Verriest. 210–270.192 Notes to pp. which is included in Œuvres Mathématiques d’Évariste Galois. A more recent French treatment is Claude Mutafian. 1960). Schwerdtfeger (Groningen: P. 1980). Field Theory and Its Classical Problems (Mathemati- cal Association of America. The Theory of Equations. Topics in Algebra. despite its austere title. Mathematical Thought. Galois. 135–250. D. 115–138. 237–259. Geometry of the Quintic (New York: Wiley. see M. Many graduate and undergrad- uate textbooks include Galois theory. Herstein. Lieber. Noordhoff. as does Kline. a companion piece to her similarly delightful treatments of non-Euclidean geometry and general relativity. tr. There is also a nice overview in I. There is a nice account of the theory in an old French edition of Galois’s works by G. 1897). E. which concentrates on geometric aspects. 1–21. Galois Theory (London: Chapman and Hall. Basic Algebra I. Sergei Sossinsky (Boston: Birkhäuser. 752–771. Tschebotaröw. M. H. Modern Algebraic Theories (Chicago: Sanborn. Felix Klein and Sophus Lie: Evolution of the Idea of Symmetry in the Nineteenth Century. Émile Picard (Paris: Gauthier-Villars. gives a valuable outline of his work. 111–130 and Cayley (261–287). Toti Rigatelli. 1975). Noordhoff. Postnikov. Ayoub. Beyond the Quartic Equation (Boston: Birkhäuser. and Edgar Dehn. PA: Science Press. Ian Stewart. Bruce King. Yaglom. 1978). 1988).

In particular. John’s Review 35. Melvin Kiernan. 119. it also contains all the conjugates of h. 122. These visualizations go back to Klein. Lectures on the Icosahedron. and John Stillwell. Groups: For a helpful introduction that stresses the larger philosophical significance of group theory. and R. A normal subgroup need not be abelian but is always “self-conjugate”: if it contains an element h. For the grad- ual interpretation of Galois’s ideas. 1964). 397–401 (1982). 27–41 (1984). Burnside. “Galois Theory for Beginners. .” American Mathematical Monthly 101.” Archive for the History of Exact Sciences 8. where h  is not necessarily the same as h. and Lattices. 111–130 193 Mathematical Monthly 89. Wilson notes that “the most general or universal aim of intellectual work is the discovery of invariants” (6). 40–154 (1971–1972). second edition (Cambridge: Cambridge University Press. 1911). Theory of Groups of Finite Order. “Groups.Notes to pp. 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. Group Concept. see Curtis Wilson. Visualization of groups: Two books offer many ways to grasp group sym- metries visually: Israel Grossman and Wilhelm Magnus. Making models of Platonic solids: See David Mitchell. 22–27 (1994).” Historia Scientiarum 27. also treated by W. 752–771. 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. and Yoı̈chi Hirano. Symmetry (Princeton: Princeton University Press. see Wussing. This is the precise meaning of “elements of the same kind” on pp. 1980). History of Algebra. 3–11 (1985).: Mathematical Association of America. Note that if we multiply both sides of this equation on the right by g −1 . Rings. 402–408. All possible g ∗ h ∗ g −1 are called the “conjugates” or the “equivalence class” of h. “The Develop- ment of Galois Theory from Lagrange to Artin. P. 118–141. Mathematical Origami: Geometrical Shapes by Paper Folding (Norfolk: Tarquin Publications. Groups and Their Graphs (Washington. 149–156. B. 123–124.” St. we find then h  ∗ g ∗ g −1 = h  = g ∗ h ∗ g −1 (since by definition g ∗ g −1 = I ). Groups: A Path to Geometry (Cambridge: Cambridge University Press. van der Waerden. D. 1985). there is some h  such that h  ∗ g = g ∗ h.C. MT. “Note sur les diffusions de la théorie de Galois: Première clarification des idées de Galois par Liouville. Burn. both of which assume considerable familiarity with modern algebra. 103–116. the inverse of the element g. see Hermann Weyl. 1999).

Then the whole group G is called solvable if and only if the quotient group of each successive group in the chain. the orbit is undisturbed and the subgroup is “normal. p. That is. Given such a “Galois group. Proof: Let g be an element of G other than the identity (since p > 1). 111–130 Here is a more intuitive way of thinking about normal subgroups. g −1 takes the vertex B back to A. if H is a normal subgroup of G.” namely all the sets {g1 ∗ h 1 . . Thus. .” we can determine in sequence a maximal normal subgroup (a normal subgroup of largest order less than that of the group). the chain is broken. of G and h 1 .” Also.}. then a performs the rotation. for the example of clock arithmetic illustrating cosets and quotient groups. then b = g ∗ a ∗ g −1 means that b is the same kind of rotation around the vertex B = g(A).194 Notes to pp. . G is abelian since it is “cyclic. by definition. g. g 2 . for all actions g. . p. But p is prime. 370–397 (composition series). Jordan and Hölder: See Camille Jordan. g2 ∗ h 1 .}. By Lagrange’s Theorem (see appendix C). then the quotient group G/H is defined to be the set of all “cosets of H in G. which is automatically a (normal) subgroup of any group. n must equal p and H is of the same order as G. made from all the elements g1 . . 1957). . . If the quotient group G i /G i+1 is nonabelian. 474. and Wussing. this can involve complex calculations. and so on until we reach a sub-sub- subgroup that includes only the identity element. where G i  G i+1 means: “G i contains the normal subgroup G i+1 . which is a subgroup of order n. . namely . of H. . . then it is abelian. (See Hartshorne. so that H = G.) Specifically. This chain can be written symbolically as G  G 1  G 2  · · ·  G n−1  G n  I. 286–291 (abelian equations). Geometry: Euclid and Beyond. if a is a rotation around a vertex A. and the equation is not solvable in radicals. . where n > 1. g2 . n divides p. Picture the elements h of the subgroup H as forming an “orbit. . . g. Then the combined operation g ∗ h ∗ g −1 tries to “tip” H out of its orbit by imposing an “external” action g. Group Concept. 135 ff. . and g takes the vertex A back to B. . 176. Then consider H = {I. {g1 ∗ h 2 . Traité des Substitutions (Paris: Gauthier-Villars. . In practice. it has no divisors besides itself and 1. If g ∗h ∗g −1 always remains within its original orbit. g2 ∗ h 2 . a maximal normal sub-subgroup.” though G i+1 may not be normal in the whole group G. G i /G i+1 .” meaning that it is composed of powers of one element. . normal subgroups (which obey this condition) signify all the rotations of the same kind around all allowable vertices.}. See appendix C. . Note also that if a finite group G has a prime number of elements. is abelian. Then if n = 1. . balanced by an equal and opposite “counteraction” g −1 .” consisting of its possible values. h 2 . Solvable chains: Galois theory begins by determining a given equation’s group of permutations.

Respecting the Impossibility of Expressing a Root of Any General Equation above the Fourth Degree. see The Mathematical Papers of Sir William Rowan Hamilton (Cambridge: Cambridge University Press. “On Finite Simple Groups and Their Classification. Note that not every abelian group is cyclic. by any Finite Combination of Radicals and Rational Func- tions” (1839). see Ron Solomon. 1:293–296 (Gauss on commutative law). For a helpful overview of these figures. 71–94. 130–137 195 {I. 321–330 (Duncan Gregory). Monster group: See Stahl. any cyclic group is abelian.Notes to pp. 362 (B. . Since g m ∗ g n = g n ∗ g m . 1967). in the same vol- ume. S. see Ewald. Sir William Rowan Hamilton (Baltimore: Johns Hopkins University Press. 247–248. Mathematical Researches (London: Longmans. 575–582. Jerrard” (1837). and Yaglom. Peirce). commuting and anticommuting (49). 331–361 (De Morgan). 1834). 42:2. see 3:481–516. “On the Argument of Abel.” 41–51. Jerrard. 277–278 (Jerrard’s faulty solution). Elements of History of Mathematics. 248–252 (Hamilton and Peacock). Fearnly-Sander also has a helpful paper on “Hermann Grassmann and the Prehistory of Universal Algebra. From Kant to Hilbert.” Scientific American. Introductory Modern Algebra. g 2 . 117–123. . 1980). Jerrard: See George B. Hamilton and the quintic: Thomas L.” Notices of the Ameri- can Mathmatical Society. see HM. Grassmann: For a translation of part of his Ausdehnungslehre (1862). 362–441 (Hamilton). Klein and Lie. 276–279. 442–509 (Boole). 542–573 (Cayley). Hankins. “Grassmann’s Theory of Dimension. 574– 648 (C. which includes the statement quoted from Grassmann about the two possibilities for products of two fac- tors.” in First Australian Conference on the History of Mathematics. see Desmond Fearnly-Sander. and so too is G. 231–239 (1995). for instance. .3) is abelian but not cyclic. 52–82. For Hamilton’s long account of Abel’s argument. For a popular account. 279:5. for a tech- nical overview of the modern study of simple groups. the four-group V (table 8. 3:517–569. Bourbaki. Wayt Gibbs. see W. for Hamilton’s “Inquiry into the Validity of a Method Recently Proposed by George B. 9 The Order of Things History of noncommutativity: For an excellent selection of original writ- ings with commentary. 510– 522 (Sylvester). 40– 41 (1998). g. 314–321 (Peacock and Cauchy). Peirce and his 162 algebras). “Monstrous Moonshine Is True.}. .

Peter Pesic (Mineola. vii–xiii. 138–143 Bolyai on quintics: See Yaglom. For a popular introduction. 50–54 (com- mutation relations of the Lorentz group). especially the Erlangen pro- gram. The Elegant Universe (New York: W. sacrificing commutativity seemed to him to be the only way to get any results at all” (300). 137–137 (Galois’s letter). which requires greater- than-light-speed travel between them. see Yaglom. Philosophy of Mathematics and Natural Science (Princeton: Princeton University Press. 295–406 (gauge theories and the standard model). W. 2 [series 1]. Nonabelian gauge theories: See Michio Kaku. Duplication of the cube. what is even more likely. 101–120. Causality and noncommutativity: In order to obey the principle that no causal influence can travel faster than the speed of light. and Literature (Cambridge. 111–137. 1950). Galois theory as “relativity”: See Hermann Weyl. 1998). 215–249 (1893). 1999). Felix Klein: For a good overview of his work. See also his 1895 Göttingen 1ecture “The Arithmetizing of Mathematics. and The Theory of Groups and Quantum Mechanics (New York: Dover. 2 [series 1]. or possibly his attempts to find a geometrical representation of triplets had shown him that rotations in three-dimensional space do not commute either.” Bulletin of the American Mathematical Society. relativity theory imposes the constraint that quantum fields must commute when the points in question are separated by a spacelike interval. 1993). see his Eight Lectures on Theoretical Physics. or. Symmetry. Klein and Lie. who notes that “possibly Hamilton had been prepared for the rejection of the commutative law by discussions with Eisenstein. 5–23. Hamilton on quaternions: HM. 74 (Galois theory and relativity) and his Symmetry. NY: Dover. Famous Problems of Elementary Geometry. 241–249 (1896). Quantum theory and noncommutativity: See Hermann Weyl.” Bulletin of the American Mathematical Society. Quantum Field Theory: A Mod- ern Introduction (New York: Oxford University Press. 2002). 1–20. 1949). 133–135. 247 (parents). 94–98. Klein and Lie. Philosophy. For Max Planck’s view of irreversibility. 59. MA: MIT Press. squaring circles: See Klein. His famous 1872 Erlangen lecture is translated as “A Comparative Review of Recent Researches in Geometry. . 582–584 and Hankins. 283–325. Hamilton. see Brian Greene. ed.196 Notes to pp. trisection of angles. Norton. I discuss this question further in my book Seeing Double: Shared Identities in Physics.

O. D. See also Pierre Cartier. Ultraradical numbers: This is the term used by Stewart. “Wittgenstein’s Remarks on the Significance of Gödel’s Theorem. 155–256 at 161–168. 148. Shanker (London: Routledge. as opposed to “simply transcendental numbers. 7–15 (1998). 2000). 10 Solving the Unsolvable De Moivre’s formula: See Smith. “Mathe- matical Problems for the Next Century. See also Eli Maor. tr. Morduhai-Boltovski calls “hypertranscendental numbers” those that are the solutions of ordinary differential equations with constant integral co- efficients.” Bulletin of the American Mathematical Society. G. Non-Commutative Geometry (San Diego: Academic Press.” Mathematical Intelligencer 20:2. Shanker. which in- cludes the text of Hilbert’s 1900 address (240–282). 1969). E. ed. which sets the scale for electromagnetic and quantum interactions. “The Consistency of the Simple Theory of Types. However. 96. and Steve Smale. . ed. 1998). Gel’fond.” which are not the solution of any finite ordinary polynomial equation. which established that arithmetic is consistent if transfinite induction is allowed. 143–147 197 Noncommutative geometry: See the authoritative (and highly technical) de- scription in Alain Connes. 1988). Szabo (Amsterdam: North-Holland. “A Mad Day’s Work: From Grothendieck to Connes and Kontsevich: The Evolution of Concepts of Space and Symme- try. S.Notes to pp. G. Solvability in the problems of David Hilbert and Steve Smale: See Jeremy Gray. Gödel: For a thoughtful commentary on Gödel that raises the possible anal- ogy with Abel and Ruffini.” in The Collected Papers of Gerhard Gentzen. Transcendental and Algebraic Numbers. 389–408 (2001). 1960). The Hilbert Challenge (Oxford: Oxford University Press. Boron (New York: Dover. M. Galois Theory. 214–222. 38. Trigonometric Delights (Princeton: Princeton University Press. Less well known than Gödel’s 1931 paper but of great interest is the 1936 paper by Gerhard Gentzen. α = e 2 /h̄c ≈ 1/137. Source Book in Mathematics. 440–454. 1145–1153 (1993).” Nuovo Cimento 108B. which ends with speculations connecting the “cosmic Galois group” with the group of general relativity and the fine structure constant. the term ultraradical refers only to algebraic numbers not expressible in radicals.” in Gödel’s Theorem in Focus. Leo F. Groups and space-time: I have discussed this matter in my paper “Eu- clidean Hyperspace and Its Physical Significance. See A. includ- ing Euler’s later statements of the formula. see S. 1994).

which is a quotient of general p products of series of powers. z)4 φ(z) = .com/ examples/quintic/. The two theta functions are defined as  ∞ q (n+1/2) cos((2n + 1)z). 166:7. q ) = 2 n=0  ∞ 2 θ3 (z. The notation x means the sum of the terms xn from n = 0 to n = ∞. that is capable of dealing with solu- tions to the quintic. The Computer from Pascal to von Neumann (Princeton: Princeton University Press. The solutions of any polynomial equation can also be expressed in terms of the generalized hypergeometric function. 1972). 106 (Torres Quevedo).” American Mathematical Monthly 101.198 Notes to pp. n=0 n namely x0 + x1 + x2 + · · · . Charles Hermite. 3 θ3 (0. Leopold Kronecker. and Francesco Brioschi independently showed that any quintic could be solved by elliptic modular functions derived from these which theta functions.wolfram. such as MathematicaTM . Spearman and Kenneth S. Williams were able to give a simple criterion in their paper “Characterization of Solvable Quintics x 5 + a x + b. ∞ are infinite sums composed of powers times cosines. Blair K. 2 θ2 (z. Computer solutions of equations: See Herman H. “Leonardo Torres Quevedo. 1985). 986–992 (1994). Goldstine.” Investigación y Ciencia. See the helpful website at http://library. z)4 This function remains of central interest in contemporary mathematics. n=1 The elliptic modular function φ(z) is defined in terms of these two theta functions:  θ2 (0. Here the notation for products is used: n=1 xn . Modular and generalized hypergeometric functions: Abel and Jacobi began to study what they called theta functions in 1827. q ) = 1 + 2 q n cos(2nz). Torres Quevedo: See Francisco González de Posada. Progress con- tinues to be made even without computers. For instance. by 1858. easily available com- puter software. 80–87 (1990) and his book Leonardo Torres Quevedo (Madrid: Fundación Santillana. 146–147 Solution of the quintic: Today students have powerful. which is available also as a wonderful poster with all kinds of information about the history and solution of quintics.

There is a nice brief account in Klein. b q : z) = i=1 q . The generalized hypergeometric function can be expressed in this very compressed notation: ∞ p  (a i )k zk p Fq (a 1 . The Invention of Infinity: Mathematics and Art in the Renaissance (Oxford: Oxford University Press. 1997). Georg Cantor: His Mathematics and Philosophy of the Infinite (Princeton: Princeton University Press. For a general history. . Lennart Berggren.wolfram. . j=1 All the elementary functions (such as the trigonometric functions) can be de- fined in terms of these extremely general functions. see J. second edition (Boulder. 61–77. and Peter Borwein (New York: Springer. Famous Problems of Elementary Geometry. ed. . 1971). namely x1 × x2 × x3 × · · · × x p . for a valuable collection of original papers. For a detailed study of the development of Cantor’s ideas. For the connection with art.html and http://library. “A Simple Proof That π is Irrational. a p : b 1 . . Cantor’s proof: For a translation of his crucial paper.Notes to pp. Field.” in his collection Great Currents of Mathematical Thought (New York: Dover. which includes Lindemann’s original proof (194–229). see Eli Maor. 2:121–158. . see his Essays on the Theory of Numbers (New York: Dover. From Kant to Hilbert. 2:838–940. . see Petr Beckmann. . Concepts of infinity: For Richard Dedekind’s introduction of the infinite as a fundamental notion of set theory. . Transcendentality of e and π : See the superb account of the proofs in Klein. 49–55. 245–283. Beauty in mathematics: For one possible realization of “classic” versus “romantic” mathematical styles. To Infinity and Beyond: A Cultural History of the Infinite (Boston: Birkhäuser. there are brief summaries at http://library. Golem Press. .html. Famous Problems of Elementary Geometry. see Pi: A Source Book.” 509.com/examples/quintic/theta. see Joseph Warren Dauben. CO. Jonathan Borwein.com/examples/quintic/hypergeo. see Ewald. and Ivan Niven. 1963). 63–64. 1987). Dunham gives a helpful and accessible account of Cantor’s work in his Journey through Genius. 1971). 1997). “Beauty in Mathematics. see François Le Lionnais. 150–151 199 means the product of the terms xn from n = 1 to n = p. A History of π . V.wolfram. For a general overview. 1990). 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).

131–149.200 Notes to pp. 2000). along with another from this notebook.” Œuvres. For Abel’s summary comments on the lemniscate. reproduces and translates this page (504). Appendix B Abel on the General Form of an Algebraic Solution My account is drawn from Abel’s “Démonstration de l’impossibilité. The origi- nal text is in his Œuvres. 261–266. 1:28–33. cited in Œuvres Complètes de Niels Henrik Abel. Paul Guyer and Eric Matthews (Cam- bridge: Cambridge University Press. unfortunately has several signif- icant misprints. Critique of the Power of Judgement. in Smith. Rowe and John McCleary (New York: Academic Press. which contains a virtuoso scene (III. Source Book in Mathematics. 389–421. Abel’s work on topics besides solvability of equations: See Michael Rosen. 151–170 Kant: For his account of the “mathematical sublime.” Œuvres Complètes de Niels Henrik Abel.” in The History of Modern Mathematics. cited in Œuvres Complètes de Niels Henrik Abel. see his 1826 letter to Holmboe. 504–505. he may well have seen Beaumarchais’s popular play Le Mariage de Figaro. Abel’s Paris notebooks: Stubhaug. Though apparantly unmoved by music. David E. the editors comment (329–338) that this paper was first published by Holmboe in 1839 and also give interesting alternative readings of Abel’s text. My notes try to explain Abel’s brief indications. “Abel’s Theorem.” American Mathematical Monthly 88. 1989). 1:66–94. 2:260. “Abel’s Theorem on the Lemniscate. Abel’s credo: My translation of part of his introduction to “Sur la Resolution Algébrique des Équations. Abel’s “favorite subject”: See his 1826 letter to Holmboe. The only other published translation. 2:261–262.” 66–72. ed. During his travels. “Démonstration de l’impossibilité de la résolution algébrique des équations générales qui passent le quatrième degré. 2:217–243 at 217.v) in which Figaro plays with the English expression “God-dam” (as Abel also spells it). 387–394 (1981). Abel. tr. . in my translation.” see Immanuel Kant. and (concerning Abel’s contributions on elliptic inte- grals) Roger Cooke. Appendix A Abel’s 1824 Paper This is Abel’s original version of his proof. which he expanded in his 1826 version of the proof. Abel was very fond of the theater.

1905). . My account follows closely Abel’s account in his “Démonstration de l’impossibilité. in Œuvres Complète d’Augustin Cauchy (Paris: Gauthier-Villars. 171–180 201 Appendix C Cauchy’s Theorem on Permutations The original source is Cauchy’s “Mémoire sur le nombre des valeurs qu’une fonction peut acquérir” (1815). series II.Notes to pp.” 75–79. 1:62–90.

This page intentionally left blank .

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

This page intentionally left blank .

15 notebooks of. 96. 200n Anrta. 87–89. 59 155–170. 42. 51. 200n Alternating groups. and Ruffini. 98–100 ibn-Musa. 89–94. 181n noncommutative. Francis. 14. Index Abbati. 195n early life. 97. 132 Abel. 30. 44 Abelian Addition Theorem. 108–109. Niels Henrik. 96 Anticommutation. 183n–184n abelian groups. 181n and Hamilton. 87. 181n illness and death of. 60 130–131. 54 Bacon. 134–136. 98 de. Algebraic functions. 61–66 and Gauss. 95–97 Ayoub. 59 formulas of. Pierre Augustin roots. 151. 200n al-Khwārizmı̄. 1. 88 Aporia. 105. 1. 101–102. 189n. See Groups Abel-Ruffini Theorem. 133 Arithmos. 95. 7. 23–28. 40–45. Pietro. 45. 32. 9 and Cauchy. 1–3. 30. Archimedes. 82 symbolic notation. 192n–193n Académie des Sciences. 182n and Galois. 200n Ausdehnungslehre (Grassmann). 9. 1. 93–94. 152–153. Analytic mathematics. 151 Aristotle. 9 Abelian integrals. A. Raymond. unknown. 85–102. Algebra 24–25. 189n Apollonius of Perga. 92. 7. 97. 1–3. 90–91 151. 145 Area problem.. 190n Athens. 5. 182n–183n coefficient. 190n 135–136 travels in Europe. 88–89. Muhammed abelian equations. 131–143 Beaumarchais. 91–93 Basham. L. 25–26. 44–45. 183n Arabic. 85–89. 43 189n–190n variable. 104–106 Accounting. 200n . See Bookkeeping Babylonian mathematics. 112–113 Alogon. 42.

René. 18 Burkert. 190n. Daniel. 50–59. 185n d’Alembert. 27–31 Bring. See Multiplication Cajori. Edgar. 185n Bombelli.” Brown. 30–40. Girolamo. Abraham. Jost. Miguel de. 150. Jean Le Rond. 200n Burn. See Cipher Bourbaki. 126. 65 Charles X. 96 Connes. 105 Bolyai. 100. Nicolas (pseudonym). Florian. Georg. Samuel Taylor. 27–29. 199n Cauchy’s theorem. 11. Roger. 112. 112. R. 64 . 68. A. 5–6 Calculators. 149. August. 62 Boolean algebra. 96. 182n–195n Commercial arithmetic. 197n 136–138. 1981n Cooke. 197n–198n Bühler. 195n DeMorgan. 62–63 Cyclic groups and symmetries. 183n. 132. 87. 122. 187n Cayley tables. 198n “Completing the cube. 187n Chinese mathematics. 147. 7–8. 137 Descartes. See Oslo 184n–185n Cipher. method. 95–97. 192n 175–180. 184n 25–26. Richard. 121 Calculus. Cantor. 67. 183n Bookkeeping. 27–28. 166.. 122 and conic sections. Dehn. William Snow. Raphael. 27. 48–49 Continuum. 195n del Ferro–Cardano–Tartaglia Causality and noncommutativity. E. 54. S. De Moivre. Walter. 196n Chatfield. 95.. 111–130 Cauchy. 132–133. R. 196n 77.. Alain. P. 174 Cayley. Augustin. 132–133 Code. Eric Temple. R. Arthur. 146.. Scipione. 120 Brizio. 132. 54–55. 16 Boyer. 83. 68 Cartier. 112 Brioschi. 197n Dance. 93.” 36–37. 101 Cross product. 78. Coleridge. Augustus. János. 69. 199n 104–105 Dedekind. 197n Bürgi. 67. 44 Burnside. Fernando. 149–150 symmetries of. 184n “Completing the square. double-entry. 189n Computers. 33. George. 188n Commutativity.206 Index Bell. 195n Commensurable. 119. 32–35. 191n Cervantes. Carl. 163. 42 Bulletin (Baron de Férussac). Christiania. 54–55. 99. 60. K. 23 Bernoulli. Anna Maria. 201n del Ferro. 48–49. 195n Cayley numbers.. 57. 199n 117. W. 142. 113. 195n Circle. Joseph Warren. 186n Cube. 42 Boole. 102 192n–193n Crelle’s Journal. 135 188n. G. 133. 189n Conic sections. Crelle. 113 Bryce. 193n Cosa (coss). Pierre. Dauben. Michael. 195n Cardano. 32–34 and commutativity.

Leonhard. legend of. See Tartaglia Superieure). 30–37. Fermat’s Last Theorem. 1–3 and Cauchy. 59. 125–130. 150. 195n. 57. 28. 199n Fine. 87–88. 1 and Société des Amis du Peuple. 196n . Paul. 191n–193n quintic.. 23. 17–18. 142–143 e. Harold M. 108–109. 195n Disquisitiones Arithmeticae (Gauss).. 192n Dimension (algebraic). 106 150. 190n Einstein. John. 196n Fields (physics). 2. 147. 191n France. 145–146. and Abel. 185n education of. 190n quartic. 23 Fibonacci. 199n Dunham. 197n École Préparatoire (École Normale Fontana. 108 roots. Évariste. 196n and Stéphanie Poterin- Euclid. 3. 90–91. 139 Duplication of cube. E. 49–50. 183n Gårding. 184n. 102–106. nonabelian. 191n 122–129. 50–51 Fauvel. 102–109. William. 2–3. 104–105 120–122 Galois theory. 57. Galileo. 187n. Eudoxus. 187n approximate solutions. 25–26. 199n Dickson. 122 Don Quixote (Cervantes). 68. 113–120. 190n Euclidian geometry. 79. La Geometrie. death of. 140. 199n Fields (mathematics). 141 Fearnly-Sander. 185n 130–131..Index 207 Descartes’s rule of signs. 17–23. 105 Fractions. 69. Desmond. 140 Ewald. 105–106. 184n–185n Dirac. 66. William B. 42. 64. Leonard E. 105 Fine structure constant. L. 90. 79. 34–35. 35. Ludovico (Luigi). 38–39. 188n Elements. See Euclid Equations. 102–108. 185n. 106–107 138–140. symmetries of. Eisenstein. 183m 142–143. 77–78. 28. 189n 189n–190n Dodecahedron. Heinrich. 143 Fundamental Theorem of Algebra. Albert. 146. 62. 2.. V. 90. 5–6 Ferrari. 188n École Polytechnique. Galois. 126–130 37–39. algebraic Galilei. 96–97. 108. Dumotel. Erlangen Program (Felix Klein). 104–105 quadratic. 124–125. See Leonardo of Pisa Dörrie. 50–58 149. 91–99. 148–149. 68–73. 184n 56. and his father. 196n “relativity” of roots. 188n–190n Field. 76. quantum theory of. 76–78. Benjamin.. 198n posthumous writings of. 106–108 111–113. 7 Edwards. 53 Euler. 198n 190n–191n cubic. 5. 139 Gauge fields. 186n. Niccolò. J. 145 general formulation. 45. Lars.

195n identity. 175–180 Gibbs. 186n–187n abelian. 190n Gorman. 187n S4 . Geometry. Hamilton. 196n Great Art (Cardano). 197n Hankins. 48. 195n 133–136. W. Duncan.” 28 solvable chains of. 193n Gies. 125 and unsolvability of quintic. 192n cosets. 119. 135–136. 123–129. 192n Grassmann. Albert. 104 order. 97. 121–122 Henry IV. Midhat.208 Index Gauss. 195n and Abel. 129 Hermite. A. 49. 197n simple groups. 118. 129 Gazalé.. Albert. 129. 79. Gregory. 51. Morton J. Josiah Willard. 128. Wayt. 112. 130. 113–120. O. 195n philosophical aspects. Andrew. 119. 195n 191n. Sophie. 129. 89. 183n Lagrange’s Theorem. 176–177. 146. Carl Friedrich. 196n Gray. 184n quotient. 56. 151 definition of. I. 196n Gentzen. Greek mathematics. 112–113 Girard’s identities. 194n Grossmann. 198n V. 186n Greene. 142–143 50–54. J. 198n Guérard. Robin. Heisenberg uncertainty 193n–195n principle. 176 Hexagon. 140 Herrstein. 150. 109. 185n Hadlock. 196n Hartshorne. 181n Grafton. 30–40. 130. 125–126 and commutativity. 175–176 Gel’fond. 50. 122–124 Gödel. 70–74. 68. 139 Gleason. 130 “Golden ratio. Herman H. 183n Groups. 181n. 185n Harmony of the World (Kepler). 194n–195n Goldstine. 122 González de Posada. 139 God. 88 invariant subgroups. visualization of. Thomas. Thomas L. Kurt. 132. and F. Israel. 138–140. Charles. 193n Heath. Gerhard. Peter. 112–113. Charles Robert. 130. 197n monster group. nonabelian. 136 and permutations. 141 A3 . 193n–194n Girard. 187n–189n 175–180. 111–130. La (Descartes).. 61. cyclical. 48. N. 95. 60. 120–122. 184n–185n. 139 Heptagon.. 100. 45 A5 . 187n S2 . 55 S5 . 122. William Rowan. 195n Geometrie. 117. 198n continuous. 113. 187n normal subgroups.. 176 Gibbs. Jeremy. 5–21 121. 92 S3 . 124. Anthony. 48 . 119. 118–120 Hellman.. 193n Francisco. 185n A4 . 197n Lorentz. Brian. Hermann. 131–132.. 66 193n–195n Germain.

Omar.. 59 Kaku. 199n “July monarchy.. 197n Kemp. Yoı̈chi. 87. Lillian R. 55. 184n Jordan. 100. 106 175–176. 181n Klein. Wilbur Richard.. 146. 196n Logos. 153 Lagrange’s Theorem. 182n. B. King. 193n Holmboe. 192n 177. C. David. Michio. Nathan. 189n. Indian mathematics. 73–83. General. 105 symmetries of. Christine. 28. 147 Invariant subgroups. 194n Invariance. 188n Indistinguishability of quanta. Irreducible case (cubic 100–101 equations). Federica. 27–28. 181n Kline. 182n. H. 74–79 Infinity. See Groups La Nave. Johnston. 128. 100. 51. 48 Locus problem. 187n–188n Isograph. 146. 7–14 Lagrange. 113. 193n 186n–187n Hoe. 183n. 67. Leopold. 30. 19–21. 55 191n. 57. Laplace. 190n. Lieber. D.. 54 Leibniz. Johannes. Pierre Simon. 148. 146. 96. Camille. 187n Irrational magnitudes. 200n Jerrard. 152–153. 200n Louis XVI. E. 187n Huntley. 133. 200n Jacobson. 192n Kabbalists. 130–131. 96. 198n Icosahedron. A. 123–129 La Geometrie (Descartes). 195n Leonardo of Pisa (Fibonacci). 9 Kant. 138–140. Holy Spirit. 143. 147 Le Lionnais. K. 184n 30. 101–102 Hippias of Mesopontum. 9 87. Adrien-Marie. 5–6 Lafayette. 199n Huffman. 183n Legendre. 199n Le Mariage de Figaro Jacobi. 184n 188n. 133 Littlewood. 177. Leonardo da Vinci. Joseph. E. Morris. J. 80 145–146. 192n Lemniscate. 192n 190n. George B. Gottfried Wilhelm. 104–105 . 65. Felix. 194n Kiernan. Knorr. 10 Kepler. 142 Lagrange resolvent. 200n Klein. 133.. 151. Bruce. 182n 198n–199n Kronecker. R.Index 209 Hilbert. 130–131. 141 65–67. 183n Khayyām. Carl Gustav Jacob. 192n Hypergeometric functions. 194n Lindemann. 85 Louis XVIII. 7–14. Joseph-Louis. 22. 146 (Beaumarchais).. Jacob. 196n. Institut de France. Leon. Berndt Michael. 124. François. 97. 139 Lalanne. 184n Hölder. Hirano. 50–58 Incommensurability. 104 Karl XIII.” 105–106 Liouville. Melvin. Immanuel. 6. 150.. S. Ferdinand. Irreversibility. 48–49. 121. Otto.

146. 187n Ore. Joseph. 199n scalar product. 150. Wilhelm. 183n Oval. 99–100. 108 Oslo. See numbers Nahin. 198n line. See Groups Maxwell. 148 Lycée Louis-le-Grand. Paul J. 189n Napoleon. 192n symmetries of. 105–106 lemma 28. Claude. 135 ultraradical. 135 Music. 182n 148–149. 59. 61–66. 131–143. 131–132 rational. 198n 143. 185n. 18–19.. and Margaret W.210 Index Louis-Philippe I. 146 Grassmann algebra. 146 Modular functions. 183n Octahedron. Isaac. Ivan. 61–66 Newton. 149–150. 24 Multiplication quaternions. 136–138 transcendental. Maxfield. 60–61 Newton’s method.. 197n place value. James Clerk. 104. Malfatti. Augusto. 6. 196n commutativity of. 193n Nonabelian gauge fields. 136. Jean Étienne. 141 Numbers Mayer. 87.. 135–136 Norway. 66 Panton.. 197n vector product. 24–25 matrix. 66 Macve. 101 Needham. 184n Noncommutative geometry. 82 142–143. 121 Octonions. 190n. 48. 134 197n. Richard. 193n 23. 134–135. David. MathematicaTM . 188n algebraic. 19–20. 189n octonions (Cayley numbers). quaternion. 196n Maor. See Groups negative. Arthur William. Hermann. 150. 187n Mercantile Arithmetic (Widman). 66. 182n–195n in Greek mathematics. 136–138 Noncommutativity. 28–30. 104 Newton’s identities. 23 Niven. 140 irrational magnitudes. John E. 195n 191n–192n Normal subgroups. 85 Maxwellian dynamics. 7. 62. 137 Morduhai-Boltovsky. Uta C. Eli. 184n–185n and Descartes. 5–6 Mutafian. 184n. 146. 7–8. Uwe F. 187n–188n Pacioli. 77. 197n Mazur. 9 Minkowski. D. 79. 197n Matrix. Mitchell. Meno. 70. 197n. 192n . 141 Magnitudes. Gianfrancesco. 187n Montucla. 199n Nonabelian groups. Øystein. 54–56. See Groups Marinoni. 184n Newtonian dynamics. 51 Monster. Luca. 51–54.. 29 counting. Barry. 9 Merzbach. 135–136 sexagesimal. 7–8. 187n complex and imaginary. 199n Magnus. 7–8. 59–66. 13–14. 150.

97. 101 Scalars. 196n–197n Parabola. Louis-Paul-Émile. Peter. M. 44. J. 106. C. See numbers Société des Amis du Peuple. 135 Postnikov. 185n Christiania (Oslo). Mark. 9. Gerhard. 64. 70. 65 general. 140 Peacock. 106–107 Radicals. 44. G. 197n Rational magnitudes. 138 Ratios. 192n Second Law of Poterin-Dumotel. 2. 182n and Galois theory. 11–17. Steve. 57. Jaques Frédéric. 5–11. 7–8. 10. 11 Shylock. 90 Spearman. 150. 10–11. Helena M. 143. 122. 138. 182n Platonic solids. 200n Pi (π ). Resolvent.. 141–143. Bruce.. 195n Rashed. 28.. Michael. 46. 140–141. 75–77. 147 special. 30. M. Blair K. 191n Pierce. 199n Rosenberger. Siméon–Denis. 140 Pentagon. Roman law. 198n Relativity Species. 191n four-dimensional 135. Roshdi. 188n Pierce. Principia (Newton). 184n Pythagoreans. 197n Pythagoras. 182n Permutations. 27. 192n Pythagorean theorem. Karen Hunger. Simon. 87 Pierpont. 196n Parshall. 139–141. 74. Piero della Francesca. 141. 185n Rosen. 195n Royal Frederick’s University. 196 Smale. 49 Republic (Plato). 190n. 42. 35 Socrates. 15. Ron. 104–105 Pesic. 46 143. 186n–187n Shanker.. 82. 140 Pascal. 190n Quantum theory. 190n Skau. 62. Space 108. 132 . Christian. 96 Poisson. Tony. Sacrifice. Max. 7 three-dimensional. 196n–197n Roots of unity. 195n Rothman. 138. 108–109. 141 Pourciau. 184n–185n of roots. 15 Singh. 140. George. Jerry. 9 n-dimensional. Stéphanie. 181n Shurman.Index 211 Pappus. 132. see Lagrange resolvent 111–130. 189n Pycior. 181n Solomon. 190n Rta.. 195n of space-time. 197n Quaternions. 188n Seventeen-sided polygon. 196n Ruffini. 57. S. Blaise. 106 Thermodynamics. Leslie. logic of. S. 80–83 Plato. 186n. 135–136. 182n Ralph. 5–6. 138.. François-Vincent. 142. 2 Raspail. 59–66. 5–11. 184n Solution in radicals. 13–17. 143 Reductio ad absurdum. 97 Peterson. 182n–183n. 175–180 Richard. Benjamin. 193n Saigey. 43 188n. 181n Planck. 187n 74. Paolo.

Universal Arithmetic (Newton). 141 Weil. N. 192n Squaring the circle.” 29 Stubhaug. Adriaan. 49 Square roots. 190n–192n Stewart. 182n–183n Steiner. 196n 147. 135 Dodecahedron. 186n Thermodynamics. 186n Tetrahedron. B. 187n. 60 Théophile. Laura. 198n West. 186n Tartaglia. 189n–191n. 186n Summary of Arithmetic (Pacioli). 191n. 185n Sylvester. symmetries of. 120–121 Tschirnhaus transformation. R. James Joseph. François. 72 “Standard theory” (physics). 183n–184n. M. 73. 183n Theta functions. 153 Trigonometry. 196n Third Law of Planetary Motion Square. 45–47. 113 Vandermonde. 183n . Hippolyte Jean. 192n. Guido.. François Marie Taton. 196n 28–29 Tschebotaröw. 188n of fundamental particles... 20 Tignol.212 Index Speed of light. 41–47. 183n Torture. irreversibility of. 198n Stein. 132.. John. D. Ian. 188n Symmetric groups. van Roomen. Arild. 192n Swetz. Leonardo. 191n Toti Rigatelli. 191n Arouet de. 32–34. John.. 193n “Treviso arithmetic. 185n Voltaire. René. 196n Time. 15–17. 85 Taylor. 142–143 188n. 104 Octahedron) Verriest. Vernier. 66–68. Simone. P. 195n Topology. 42–43 Viète. Frank. 182n Walker. 20. Alexandre- in algebraic expressions. See Groups Triangle. 192n of three-dimensional space. Jean–Pierre. 188n symmetries of. 184n von Fritz. 60 136–138. 7–14 (Kepler). See Groups Symmetry. 186n Synthetic mathematics. Icosahedron. 17. See Number Stillwell. V. 182n Theodorus. 200n Trial. Vectors. L. George. Kurt. 149–150. 55 Wallis. 142. 142. 113–120 Suleiman II.. 195n Uspensky. 56–58. 68 Theaetetus. 47. Howard. 185n. J. Emmett. 141 Stahl. 146. 140. Christian. 124 Vetter. 186n Theology. Saul. 182n Tetractys. van der Waerden. 10 von Tschirnhaus. 18 Subgroups. Cube. 150. sound of. Count Ehrenfried Tetrahedron. 193n of polyhedra (see Triangle. 75. Torres Quevedo. 197n Transcendental. G. 5–6 Walter. L. Trisection of angle. 151. 16–17.

I. Curtis. 140. Hermann. Johann. 192n. 193n. 184n Zammattio.. 189n Williams.. 193n Xenophon. 196n Widman. 195n–196n Yamey. M. Emmanuel. 193n Winternitz. Andrew. 184n Zero. 184n Wussing. Carlo. Kenneth S. 43 . B. 9. Hans. 29 Wiles.. 198n Wilson. S. 182n Yaglom.Index 213 Weyl. 188n–189n.