Infinity

In the l ast century, the theory of i nfni te sets, created by the German mathe
matician Georg Cantor, has occupied one of the key positions i n mathematics.
Its concepts refect the most general properties of mathematical objects.
However, i n view of the di scovery of a series of paradoxes i n this theory, many
emi nent schol ars doubt the soundness of its foundations.
Thi s book i s a popul ar account of the roads fol lowed by human thought i n
attempts to understand the i dea of the i nfnite in physics and i n mathematics.
It tel l s the reader about fundamental concepts of set theory, about its evolution,
and about the relevant contributions of Russian scienti sts.
The book i s meant for a broad range of readers who want to know how the
notion of the i nfnite has changed in ti me, what one studies in set theory and
what is the present state of that theory.
Editor-i n-chief
A. S. Sol odovni kov,
Doctor of Physical and
Mathematical Sciences.
In Search of Infinity
N. Ya. Vilenkin
Translated by Abe Shenitzer
with the editorial assistance of
Hardy Grant and Stefan Mykytiuk
Birkauser
Boston Basel Berlin
rranslator
Ab Shenitzer
Department of Mathematics & Statistics
York University
North York, Ontario M3J IP3
Canada
Library of Congress Cataloging-in-Publication Data
Vilenkin, N. Va. (Naum Yakovlevich)
[ V poiskakh bskonechnosti. English]
In searh of infinity I N. Va. Vilenkin : translated by Abe
Shenitzer, with the editorial assistance of Hardy Grant and Stefan
Mykytiuk.
p. c.m
Includes bibliographical references.
ISBN 0-8176-3819-9 (H : alk. paper). -- ISBN 3-7643-3819-9 (H
alk. paper)
1. Infinite. 2. Set theory. I. Title.
QA9.V5513 1995 95-34319
511.3'22--dc20 CIP
Printed on acid-free paper
Birkhluser Boston 1995
Birkhiuser $
Copyright is not claimed for works of U.S. Goverment employees.
All rights reserved. No part of this publication may be reproduced, stored in a retrieval
system, or transmitted, in any form or by any means, electronic, mechanical, photoopy
ing, recording, or otherwise, without prior permission of the copyright owner.
Permission to photocopy for internal or personal use of specifc clients is granted by
Birkhluser Boston for libraries and other users registered with the Copyright Clearance
Center (CCC), provided that the base fee of $6.0 per copy, plus $0.20 per page is paid
directly to CCC, 222 Rosewood Drive, Danvers, MA 01923, U.S.A. Special rquests
should be addressed directly to Birkhluser Boston, 675 Massachusetts Avenue,
Cambridge, MA 02139, U.S.A.
ISBN 0-8\76-3819-9
ISBN 3-7643-3819-9
Reformatted by Texniques, Inc., Brighton, MA from translator's diskettes
Printed and bound by Quinn-Woodbine, Woodbine, NJ
Printed in the United States of America
9 8 7 6 5 4 3 2 I
1ablc ol Lontcnts
Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . vi i
Chapter 1 . Infni ty and the universe . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
Chapter 2. The mysteries of i nfni te sets . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33
Chapter 3. Remarkable functi ons and curves, or a strol l
through a mathematical hal l of wonders . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7 1
Chapter 4. In search of the absol ute . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1 1 7
Concl usion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1 34
Notes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1 37
ntroducton
The concept of i nfnity is one of the most important, and at the same ti me,
one of the most mysteri ous concepts of science. Al ready in antiquity many
phi l osophers and mathematicians pondered over its contradictory nature. In
mathematics, the contradictions connected with i nfnity i ntensi fed afer the
creation, at the end of the 1 9th century, of the theory of i nfnite sets and the
subsequent discovery, soon afer, of paradoxes i n thi s theory. At the time,
many scienti sts i gnored the paradoxes and used set theory extensively i n their
work, whi l e others subjected set-theoretic methods i n mathematics to harsh
critici sm. The debate i ntensi fed when a group of French mathematicians,
who wrote under the pseudonym of Nicol as Bourbaki, tried to erect the whole
edifce of mathematics on the si ngle notion of a set. Some mathematicians
greeted this attempt enthusi astical l y whil e others regarded it as an unnecessary
formalization, an attempt to tear mathematics away from l i fe-giving practical
appl ications that sustain it. These differences notwithstanding, Bourbaki has
had a si gni fcant i nfuence on the evol ution of mathematics i n the twentieth
century.
In this book we try to tel l the reader how the idea of the i nfni te arose and
devel oped in physics and i n mathematics, how the theory of i nfnite sets was
constructed, what paradoxes it has led to, what signi fcant efforts have been
made to el i mi nate the resul ting contradictions, and what routes scientists are
tryi ng to fnd that woul d provide a way out of the many difculties.
Chapter 1
nhnly and lhc unVcrsc
From mythology to science. The study of stone fgures and of drawi ngs,
made on cave wal l s wi th ochre and charcoal , shows convincingly that tens of
thousands of years ago humans not only collected i nformation about ani mals
and bi rds, about properties of grasses and fruits, but also tried to establ i sh
regul arities i n the world around them and to understand its origi ns. Thi s i s
how the ol dest myths arose: myths about the "Great Origi nal Mother," the
common ancestor of humans and ani mal s, and about totems - protectors of
tribes. Ancient humans attributed menacing natural phenomena - thunder,
l i ghtning, earthquakes and hurricanes - to the infuence of otherworl dl y forces.
The devel opment of agriculture and cattle breedi ng i ncreased the i mpor
tance of celestial observations. The conjunction of heavenl y bodies was a
signal for sowi ng and harvesting. From the heavens came rai n that irrigated
felds and pastures. The locust, one of the most formidable enemies of the
ancient farmer, fel l from the heavens on crops and orchards. Small wonder that
ancient mythology embraced both Heaven and Earth.
Smal l wonder also that humans dei fed the forces of nature and thought
that the heavens were the abode of the gods. Thi s was the source of the frst
myths about the ori gi n of the worl d and about the connections between natural
phenomena. In al most al l these myths the i ni ti al state of the world was described
as either chaos or an ocean. Then came the separation of heaven and earth,
and, on the earth, of water and land. Thus Akkadi an myths tell of the god
Marduk, who vanqui shed the goddess Tiamat, symbol of the pri meval ocean,
and fashioned heaven and earth out of her gigantic body which he spl i t i n two.
The symbolic element i s more pronounced i n the myths of ancient India.
These myths are due to Aryan tribes that i nvaded Hindustan between 1 500 and
1 200 Be. These myths tel l of the battle of the Danavs and the Adityas, of the
2 In Search of Innit
constrained and the unbounded, or the limited and the unlimited. The triumph
of the Adityas was secured by Indra, apparently the son of Heaven and Earth.
When, by a mi racle, he grew to frightful size, the terri fed Heaven and Earth
scattered in opposite di rections.
Si mi lar myths exi sted i n ancient Iran. In the distant past the culture of the
Irani ans was the same as that of the Aryans. According to Irani an stories, the
god Ormazd created out of sparkl i ng metal the l i ght and bright heaven in the
form of an egg. Its top reached the Infni te World. Al l creatures were created
in the i nteri or of heaven. The Irani ans thought that the earth was round and
suspended in the center of heaven l i ke the yoke in the center of an egg.
According to ancient Greek mythology, the world was divided into three
regi ons - the underground ki ngdom, the earth, and the sky, surrounded on al l
si des by an ocean (thi s explai ned why water spurts upward from spri ngs and
fal l s to the earth as rai n).
Myths enabled people to rel ate unconnected concepts and to make some
kind of sense of the world around them. The element of action i n myths was
outside space and ti me and was not subject to the restrictions of l ogi c. Because
i t refected the experience of earl ier generations, a myth could not be refuted
by the argument that it contradicted the experience of an i ndividual .
The devel opment of manufacturing and the increased complexity of so
ci al relations were accompanied by the slow evolution of a new conceptual
framework that made i t possible to bypass mythology and to use logic to verify
the consi stency of deductions. Myths began to contradict reason and logical
criteria, fewer of them were created, and the begi nni ngs of science appeared.
Tradition l i nks the begi nni ng of the scientifc mindset to the Greek phil oso
pher Thales
l
who predicted the solar ecl i pse of 585 Be. Thi s prediction could
not have been made at a ti me when the sun and moon were thought of as the
eyes of Heaven or as the gods Hel ios and Selene.
Basi ng hi mself on the Indo-Irani an i dea of the contest between the bounded
and the i nfnite, Anaxi mander
2
devel oped the doctrine that the uni verse derives
from the i nfnite - the ape iron - which is inexhaustible and eternal . He taught
that there are an i nfnite number of worlds that stand apart, that these worlds
come into bei ng, and when they peri sh they are reabsorbed into the i nfni te.
Much i n the doctri nes of Thales, Anaxi mander and thei r students strikes us
now as naive (for example, Thal es' notion than humans derive from fsh). But
these doctrines al so included bri l l i ant guesses, such as the spherical shape of
the earth, the evol ution of l i vi ng creatures, and many others.
From the viewpoint of thi s book, what is most i mportant in the doctrines
Chapter 1. Infnit and the universe
3
of these phi l osophers is that they i ni ti ated the discussion of l i mi tlessness and
of i nfnity, and that they clarifed the question of the fnite or i nfni te nature
of the universe. To cl arify the idea of unboundedness, Anaxi mander said that
"wherever the warrior stands he can extend hi s spear farther." Thi s i dea was l ater
repeated by Archytas, an eminent mathematician of the Pythagorean school who
l i ved i n the fourth century BC. Democritus, the famous atomist of antiquity,
thought that the universe is not only i nfnite but al so centerless. In the frst
century BC, thi s thought was expounded i n verse form by the Roman poet and
philosopher Titus Lucreti us Carus:
It has no bounds, no end, no l i mit,
And i t matters not what part of the uni verse you are i n;
Wherever you are, from the spot you take up,
It stretches to i nfnity i n all di rections.
A few centuries l ater, Nizami, the great poet of Central Asi a, asked:
Is there di rection i n the i nfnite world?
Can endless di stances be measured?
Thus the idea of the i nfnite arose i n the East, reached the Greek philosophers
and schol ars who studied it, and returned to the East. The emi nent German
mathematician Hermann Weyl 3 wrote that the i ntuition of the i nfni te, and its
cal m and unquestioned acknowl edgment, were present i n the Eastern world,
but there thi s i ntuition remai ned an abstract awareness. Accordi ng to Weyl ,
the great achievement of Greek science was the transformation of the pol ar
opposition between the fnite and the i nfnite into a prol i fc tool for acqui ri ng
knowledge of real ity.
We note that the i dea of the eternity of the uni verse also ori gi nated i n the
East, as indicated by the fol lowi ng ancient Eastern parable: "There i s a di amond
mountain a thousand cubits tal l . Once i n a hundred years a bi rd comes fyi ng
and gnaws the mountain wi th i ts beak. When i t has gnawed away the whole
*
mountai n, the frst i nstant of eternity wi l l have passed."
The ancient Greek philosopher Heraclitus very clearl y expressed the idea
that the worl d i s eternal and that i t changes i n time. He taught that " al l thi ngs
fow, and i t i s not possible to step i nto the same river twice, for he who does
wi l l be washed by other waters." He also said that "thi s cosmos was made
*
In other vaiants of this paable the mountain is ground down by the fairy Lilavati, who
comes dancing once in a hundred years and touches the mountain with the end of her train.
4 In Search of Infnit
by neither god nor man, but always was, and i s, and wi l l be: ever-l i vi ng fre,
ki ndl i ng i n measures and quenchi ng i n measures."
Heracl i tus' cosmol ogy is very far removed from the myths of antiqui ty.
Indeed, a study of ancient myths shows that space and ti me were once thought
to be fnite, di scontinuous, and made up of qual i tativel y different pieces. Then
the idea of evol ution of the worl d was foreign to the human mind, and ti me
seemed to fow in a ci rcul ar manner, with constant repetition of the seasons
and of the positions of the pl anets and stars. (It i s no accident that i n ancient
Greek mythology Chronos, the god of ti me, i s the son of Uranus, the god of
Heaven, and phi l ologists fnd that the words time and rotation have a common
origi n. ) And even when men began to think of ti me as fowing in one di rection,
they assumed that the world had a begi nni ng (creation of the world) and woul d
end
(
the last judgment of the Gospel s, the twilight of the gods i n Scandinavi an
mythol ogy, and so on). Space seemed bounded by the heavenl y frmament
that was the abode of the gods. It took a trul y revolutionary change of human
consciousness for the notion of a boundless world to emerge, a world without
begi nni ng or end.
(
As we shal l see bel ow this notion struck many of the best
minds in the ancient world as rather bol d. )
Like grains of sand at the seashore. In antiquity, phi l osophy, science, and
mathematics were so cl osely rel ated that, as a rule, all three were pursued
by the same people. For example, Thales not only phi l osophized about the
l i mitl ess and predicted ecl ipses but also proved, for the frst time, geomet
ric theorems. Again, tradition credits Pythagoras, founder of an infuential
phi l osophical school i n the si xth century BC, with the di scovery of the famous
theorem about a right tri angle.
Pythagoras cl ai med that "al l i s number," and some numbers
( 1 ,4, 7, 1 0)
pl ayed a speci al rol e in hi s world view. Incidental l y, the number 7 pl ayed
a special role for the ancient Sumerians, long before mathematics began to
develop in Greece. In thei r myths there are seven storm spirits, seven nasty
di seases, seven regions of the underworld l ocked with seven doors, and so on.
Agai n, seven celestial bodies move across the frmament. An unknown Greek
phi l osopher, a contemporary of Thales and Anaximander, clai med that the ex
ternal form of things as wel l as their essence must refect the number seven; in
particul ar, he l i sted seven spheres of the uni verse (the sphere of pure zephyr, the
stars, the sun, the moon, of air, water, and earth). Some schol ars thi nk that the
mystical role of the number seven goes back to the ti me when people counted to
six and seven stood for many
(
even today we use the expressions "seven don' t
wait for one," "measure seven times, cut once," and so on).
Chapter I. Infnit and the universe
5
Nei ther the Greek nor the Egyptian schol ars coul d express very l arge num
bers. That was because they were unaware of the i dea of positional notati on.
The Babyl oni ans, who knew positional notati on, had too util itarian an approach
to mathematics to consider such i mpractical questions as, say, the number of
grai ns of sand at the seashore. It is i nteresting that according to the ancient
Greek "Texts of the pyramids" the ferryman of the underworld tests the soul of
a dead king to see i f he can count to ten. The dead soul ' s answer takes the form
of a poem in which the ten fngers are l i sted in thei r order. Thi s is undoubtedl y
an echo of a ti me when fnger counting bordered on magi c.
And even when people learned to express by means of numbers the quantity
of warriors i n an army or the quantity of l oaves of bread needed to feed the
sl aves who bui l t colossal temples and pyramids, they had no knowledge of
numbers now called bi l l ion, tri l l ion, quadri l l ion, and so on. To describe l arge
numbers they resorted to compari sons: "as much as there is sand and dust," "as
much as there i s sand at the seashore," "l i ke the weight of a mountain balanced
by weights," "l i ke the number of leaves on trees." At that ti me, even the best
scientists couldn' t tel l which was l arger: the number of grai ns of sand at the
seashore or of leaves i n the forest.
In the third century Be, some 250 years after the Greek phi l osophers be
gan to discuss the concept of boundlessness, Archi medes created a system for
designating very l arge numbers. In hi s "Sand-reckoner" he created a calcul us
for designating numbers up to 1 0
8 1 0
16
How large thi s number i s can be sur
mi sed by noting that 1 01
5
0 neutrons sufce to fl l al l of our metagal axy
(
that i s,
the universe). Agai n, if one used decimal notation and wrote 400 di gi ts on a
meter-long strip of paper, then the strip of paper needed to record Archi medes'
number woul d be so long that i t would take a ray of l ight about eight days to
traverse i t.
Large though i t i s, Archi medes' number i s fnite. One can' t wri te down the
i nfnite sequence of natural numbers on a strip of paper so long that it can be
wound into a bal l the size of our gal axy, or into 1 0
8 1 0
16
such bal l s.
The ancient Hi ndus were also fascinated by l arge numbers. Some Hindu
stories tel l of battles that i nvol ved 1 0
2
3 monkeys, and of tests i n the course
of which Buddha named enormous numbers. No matter that the whole so
lar system coul d not accommodate 1 0
2
3 monkeys. Such number games were
i ncarnations of the notion of i nfnity and they bewitched Greeks and Hi ndus
alike. Archi medes and the Hindus were del i ghted with the thought of creati ng
a symbol ic cal cul us that made it possible to express such huge numbers on a
small wax tablet or a strip of papyrus, thus going beyond visual contempl ati on.
A mysterious paradox. The i dea of the i nfnite in science came from various
sources. The questions: Is the uni verse bounded? Did it have a begi nni ng?
Wi l l it have an an end? were one such source. One of the most vital questions
that ancient Greek philosophers contended with was the structure of the world
i n the smal l . Dai l y experience showed that a l oaf of bread could be shared by
two, three, or at most ten people, and it could be broken i nto some ten thousand
crumbs. Could the l oaf of bread be further subdivided? And i s there a l i mi t to
di vi si bi l ity of materi al objects? Experience alone could not supply an answer
to this question, and so the question of the l i mits of di vi si bi l ity of objects shifted
from the real m of experience to that of specul ation.
Some phi l osophers maintained that matter is i nfnitely divisible. Thus
Anaxagoras cl ai med that "there i s no least among smal l thi ngs; there i s al
ways something smal l er. For that which exists cannot cease to exi st as a resul t
of di vi si on regardless of how far the l atter conti nues." He thought that the con
tinuous cannot consist of discrete elements "separated from one another, as if
spl it from one another by the bl ows of an axe. " Anaxagoras and hi s fol l owers
seemed to thi nk of a set of discrete points i n terms of a heap of dust and of a
conti nuum in terms of a bronze or i ron sword.
The other school derived from the Pythagoreans. Its fol l owers assumed
that there exist smal l est particles of matter, atoms, whose hardness makes them
i ndi vi si bl e (i n Greek "atom" means indivisibl e). These ideas were devel oped by
Leucippus
4
and Democritus. Atomists also introduced the notion of indivisible
and di mensionless parts of space (amers). Some schol ars claim that these ideas
go back to Democritus whil e others ascribe them to Epicurus.
5
It i s difcult to
decide who i s right, because all that has come down to us from Democritus'
many works are meagre fragments.
The confict between the two phil osophical school s was exacerbated when,
in the mi ddle of the 5th century BC, the Greek phi losopher Zeno of Elea
6
showed the paradoxical consequences of careless use of the assumption of the
unl i mited di vi si bi l ity of space and time. The best known of Zeno' s paradoxes
are the Dichotomy and the Achilles. In the Dichotomy Zeno shows that a moving
object i s at rest. After al l , argued Zeno, before it reaches its aim the object must
cover half the di stance i nvol ved, and before that a quarter, and before that an
eighth, and so on. Si nce space i s i nfnitely di vi si ble, the process of halving wi l l
never end. In the Achilles, Zeno shows i n a si mi lar manner that the feetfooted
Achi l l es, who covers ten stadi a in a minute, wi l l never catch up with the sl ow
tortoi se who covers one stadi um in a minute.
Chapter I. Infnit and the universe 7
Of course, Zeno' s arguments about the impossibil ity of motion in the real
world are contradicted by everyday experience. When the famous phil osopher
cynic Di ogenes
7
learned about Zeno' s arguments he si mpl y got up and
*
walked. But Zeno' s arguments showed that the views on i nfnity in contem-
porary mathematics were extremely naive. In particul ar, Zeno was the frst
to show that a segment can be decomposed i nto i nfni tel y many parts each of
which has nonzero length. If we repl aced hi s geometric segment by a fni te
ti me interval then we could prove the paradoxical concl usi on that we coul d
name al l of the i nfnitely many natural numbers i n an hour' s t i me. To do thi s
we woul d name the frst natural number i n the frst half hour, the second i n the
next quarter hour, the third in the next eighth of an hour, and so on. We coul d
thus ft the i nfni te i n the fni te - the i nfnite sequence of natural numbers i n a
fnite ti me interval .
Thi s attempt cannot be real ized; the number 100, say, woul d have to be
uttered in 2
-
1 00 hours - a ti me i nterval far shorter than the fastest process
known to modern science. Another barrier to such speed i s that no i nformation
can be transmitted with speed greater than the speed of l i ght. Al so, when
one attempts to subdivide space into smaller and smaller parts one comes up
agai nst i ts quantum properties. But these are considerations that pertai n to
the real worl d, whereas (to use modern parlance) Zeno investigated one of i ts
mathematical model s that admits unl i mited hal vi ng of space and ti me i nterval s.
For this model objections of physici sts are i rrelevant.
Over the centuries, judgments concerni ng Zeno' s paradoxes have changed
many ti mes. Just when it seemed that they had been completel y refuted, a care
ful anal ysi s woul d show that, whatever the level of knowl edge, there remained
something unexpl ai ned - a germ of new contradicti ons, of new knowledge. A.
Fraenkel ,
8
the emi nent special ist in the area of i nfnite sets and the foundations
of mathematics, put i t thus:
The bridgi ng of the chasm between the domai ns of the discrete
and the continuous, or between arithmetic and geometry, is one
of the most i mportant - nay, the most i mportant - problem of
the foundations of mathemati cs. . . . Of course, the character of
reasoning has changed, but, as always, the difcul ties are due to
the chasm between the discrete and the continuous -that permanent
stumbl i ng block which also plays an extremel y i mportant role in
mathematics, philosophy, and even physi cs.
Of course, Diogenes' reactIOn had everything to do wi th Zeno's concl usi on and nothi ng
to do wIth hi s arguments. (Transl ator).
Later we wi ll lea of other paradoxes associated with the notion of i nfnity
that may make Zeno' s paradoxes look quite naive. But the emnent Russi an
phi l osopher G. I. Naan
9
observed that mankind wi l l never be able to completely
refute the Eleatic philosopher because the i nfnite i s i nexhaustible, and Zeno
managed to capture i n nai ve but bril l i ant form three eteral problems that are
very close to one another and to the problem of i nfni ty, namely the problem
of nothi ng, the problem of conti nui ty, and the problem of exi stence. It was
no acci dent that Ari stotle called Zeno "the founder of di alectic," and Hegel
regarded him as the father of di alectic in the modem sense of the word.
As could be expected, there were attempts to use Zeno' s paradoxes to bol
ster i deal i sm. As one famous German ideal ist philosopher put it, in Zeno' s
paradoxes i nfnity pl ays the role of a di ssolvent of reality.
We won' t go into further detai l s of the role of Zeno' s paradoxes in physi cs
and phi l osophy. Thei r rol e i n mathematics was that they uncovered the confict
between the di screte and the conti nuous and showed the danger of thoughtless
use of the i nfnite. Afer Zeno one coul d not, li ke the sophist Antiphon, 1 0 regard
a ci rcle as a polygon wi th i nfnitely many si des and in this way compute its
area. Thi s was the begi nni ng of the period of el i mi nation of the i nfnite from
mathematics.
A desperate attempt to do wi thout i nfnite processes i n geometry was due
to Democritus . He tried to base geometry on his atomic doctri ne. Had he
succeeded, the form of geometry would have been very diferent from what
it i s. But l ong before Democritus, the di scovery was made in the school of
Pythagoras that the side and di agonal of a square are i ncommensurable. And
thi s coul d not be so i f the si de of a square and its di agonal consi sted of a fni te
number of i ndi vi si bl e parts. Al so, while Democritus managed to compute the
vol ume of a pyrami d usi ng hi s method, he could not decide whether or not
neighboring cross sections of a pyramid are equal ; i ndeed, i f they are, then
the pyramid could not contract to a point, and i f they are not then the pyrami d
woul d not be smooth. As a result of these diffculties hi s textbook of geometry
l ost out to Eucl i d' s famous Elements, based on the idea of unl i mited di vi si bi li ty
of space. And philosophers i ntent on refuting Zeno' s paradoxes had to l ook for
approaches other than atomi sm.
On shaky ground. Aristotle, one of the greatest phi l osophers of antiquity,
gave a great deal of thought to the notion of the i nfni te and its properties.
When he discusses thi s subject i n hi s works he warns that the topic entai l s
wal king on very shaky ground. Indeed, afer Zeno and Democritus the concept
9
of the i nfni te was mired in contradictions. Ari stotle admitted that "much that
is i mpossible fol l ows from negating the exi stence of the i nfni te as wel l as from
accepti ng i t." He gave fve reasons for bel ieving in the existence of the i nfnite.
Four are: the i nfnity of ti me, the i nfnite di vi si bi l ity of magni tudes employed i n
mathematics, the fact that the i nfnite prevents the cessation of the phenomena
of coming i nto exi stence and of anni hi l ation, and the fact that the fnite always
abuts on somethi ng, and thus there i s no l i mi t to the fnite. The ffth reason -
the reason Ari stotle regarded as the weightiest - was that there are no bounds
to thought. Speci fcal l y, there are no bounds on numbers, or on mathematical
magnitudes, or on what i s beyond heaven. And i f what i s beyond heaven is
i nfnite, then there are many worlds.
And yet, al l these weighty arguments notwithstanding, "the Phil osopher"
(Ari stotle' s sobriquet in antiquity and (especi al l y) l ater) refused to accept the
idea of the exi stence of an i nfnite worl d, saying that in matters of the i nfni te
one cannot trust thought.
We won' t go i nto the subtleties of Ari stotl e' s reasoning i n which he analyzes
vari ous consequences of the assumption of the exi stence of the i nfnite.
What i s most i mportant for us i s the di sti nction Ari stotle drew between
the actual infnite and the potential infnite, that is between an existing i nfnity
and an i nfni ty i n the state of becoming. To obtain an exampl e of each type
of i nfnity, consider Zeno' s process of unl i mi ted halvi ng of a segment. An
example of an actual i nfni ty i s the total ity of parts of a segment resul ti ng
from its completed subdi vi si on by repeated hal vi ng. An example of a potenti al
infnity i s the process, evol vi ng i n ti me, of its continued subdi vi si on by hal vi ng.
Ari stotle al so i ntroduced two other notions of i nfnity now known as extensive
and intensive. The frst arises as a result of successive and unl i mi ted addition
of new objects, and the second i s the resul t of del vi ng i ndefnitely deeper into
the structure of an object under i nvestigati on.
Ari stotle acknowledged the exi stence of onl y the potential i nfnite. He
said that "The i nfnite does not actual l y exist as an i nfni te sol id or magnitude
apprehended by the senses. . . . The i nfni te exi sts potential l y, the i nfnite i s
motion . . . "
Ari stotle rejected Zeno' s arguments by sayi ng that what moves does so
unconsciousl y. He also rejected the notion of an i nfni te uni verse; he thought
that i t was bounded by the ul ti mate sphere beyond which there was neither
matter nor space. Unl ike Pl ato, who thought that the worl d was the work of a
Demi urge (creator), Aristotle cl ai med that it was not created and was eternal .
(Aristotle' s view that the worl d was not created earned hi m l ater the enmity of
1 0
In Search of Infnit
many theologi ans. )
After Ari stotle it was acknowledged that "science i s true t o the extent to
which i t is based on the assumption that the di scontinuous does not consist of
the i ndi vi si ble."
Mathematicians also stopped usi ng the concept of i nfnity. For example, i n
hi s Elements Eucl i d doesn' t use the concept of the i nfnite even where it woul d
be qui te natural to do so. Euclid does not say that there are i nfni tel y many
primes but that the number of primes exceeds any preassigned natural number.
He moves triangles and other fnite fgures but never moves an i nfnite plane.
He tries to use motions of even fni te fgures as l ittle as possible - afer Zeno
the concept of motion became logical ly suspect.
To avoid the use of the i nfnite, Eucl i d' s predecessor, the Greek mathemati
ci an Eudoxus,
I I
formul ated an axiom which, in effect, denied the exi stence
of i nfnitely smal l and i nfnitel y l arge magnitudes. It was to the effect that a
suitable i ntegral multiple of the smaller of two given magnitudes wi l l exceed
the l arger one. Thi s axiom was the basis of the method of exhaustion used by
Eucl i d and Archi medes to prove theorems about areas and vol umes.
Rebirth of the infnite. The l ast centuries of antique ci vi l i zation were marked
by the decl i ne of learni ng and by the spread of superstitions and of bel iefs
in mi racles and omens. There was a gradual decl i ne of the oral tradition that
enabled people to understand the complex reasoni ng of the great schol ars of the
fouri shi ng period of that ci vi l izati on. Faith in magicians, miracle workers and
prophets fouri shed. The triumph of Chri stian dogma completed the process of
sl ow death of antique phi l osophy and learni ng. In 41 5 AD a mob of Chri sti an
fanatics, incited by the Alexandrian bi shop Cyri l , l ynched Hypati a, 1
2
one of the
last representatives of ancient cul ture, and burned the Alexandrian l ibrary that
preserved the treasures of that cul ture.
For many centuries after that, phil osophy was reduced to the rol e of a servant
of theology. Old superstitions that the fat Earth is supported by three whales
got a new lease on l i fe. The words of the theologian Tertul l ian - "It i s certain
because i t i s i mpossibl e" - became the sl ogan of that ti me. The i nfnite also
ended up i n the theological sphere - it became an attribute of God.
But sl owly the i deas of Pl ato and Ari stotle again acqui red currency and
the difcul t job of reconci l i ng Church dogma and the teachings of the ancient
phi l osophers began. While the most zealous theologi ans rejected the teachings
of the pagan phil osophers, an ever l arger number of schol ars adopted the view
point of Albert the Great: 1
3
"I am not concerned with divine miracl es when I
1 1
reason about natural things usi ng natural logic." Al ready i n the 1 4th century
some schol astics acknowledged the eternity of the worl d and the mortal ity of
the human soul .
The controversies between theologians and philosophers gave rise t o new
ideas that undermined rel igious dogmas as wel l as the worl d view of the follow
ers of Ari stotle. In 1 277 the Pari s bi shop Etienne Tempier attacked the teachi ng
of the Phi l osopher about the eternity and the noncreatedness of the uni verse but,
at the same ti me, admitted the i dea of the possi bl e pl ural ity of worl ds. He even
advanced the idea that the heavenl y spheres can have not onl y ci rcul ar but al so
recti l i near moti on. But that meant that there was space beyond the ul ti mate
sphere ! Thi s was a frst step to admitti ng the possi bi l ity of an i nfnite uni verse,
although Tempier' s teaching contai ns no explicit statement to thi s efect.
In the 1 4th century Thomas Bradwardine
l
4
arrived at the idea of the exis
tence of a vacuum. Then there were scholastics who rejected Ari stotle' s thesi s
about the nonexi stence of the actual i nfnite. The so-cal led i nfni ti sts cl ai med
that the notion of the actual i nfni te contai ned no contradiction whatever and
such an i nfnite coul d therefore exist. At the begi nni ng of the 1 4th century
John Baconthorpe
l
5
asserted that every magnitude - number, time, a col l ection
of sol ids - could be actually i nfnite, and that a sol id could be subdivided into
an i nfnite number of parts. Thi s undermi ned the foundation of Ari stotel i an
cosmology.
The fnal demol ition of ancient cosmology came i n the 1 5th and 1 6th cen
turies and i s l i nked to the names of Nicol as Cusa
l
6
and Gi ordano Bruno. Cusa
devel oped the doctrine of the maxi mum, that i s something that cannot be ex
ceeded. This was one of the greatest achievements of Renai ssance di alecti cs
and prepared a revol ution not onl y i n cosmology but al so i n the mathematical
way of thi nki ng. Thus Cusa maintai ned that a l i ne i s a ci rcl e of i nfnite radi us
and considered not individual fgures but the l i miting states of fgures resul ti ng
from vari ous modi fcations of thei r form.
The notion of the i nfnite al so attracted the attention of astronomers. In hi s
great work Nichol as Copernicus maintained that the di stance between the Earth
and the Sun is imperceptibly small in comparison with the heavenl y frmament.
The sphere of stars he regarded as very si mi l ar to the i nfnite. He sometimes
said that Heaven i s i mmeasurabl y greater than the Earth and represents an
i nfnitely l arge magni tude. But he lef i t up to other schol ars to deci de whether
the Universe i s i nfnite or merely i mmeasurabl y l arge.
The fnal step i n the demol ition of old dogmas was taken by Gi ordano Bruno
who pai d with hi s l i fe for hi s scientifc exploit. Bruno wrote:
1 2 In Search of Infnit
The universe is one, i nfni te, i mmovable . . . It cannot be grasped
and is therefore i ncalcul able and l i mi tless, and thus i nfnite and
boundless, and, therefore, i mmovable. It does not move i n space,
for there i s nothi ng outside i t where i t coul d transfer, because i t i s
al l . It i s not born . . . for i t i s al l exi stence. It i s not annihi l ated, for
there is no other thing i nto which it coul d change. It can nei ther
decrease nor increase, for it is i nfnite.
This i s how the human spi ri t was freed from the l i mi tations that fettered it.
Bruno expressed the spirit of the new era i n the fol l owing verse:
I rise high and boldly break
The i magi nary barrier of the crystal sphere.
I rush to i nfnity, to di fferent di stances.
Some are desti ned for grief and some for joy, -
The Mi l ky Way I leave bel ow for you.
Of course, many found the new world now revealed to the eye of the mi nd
most uncomfortabl e. The heavenl y spheres contai ni ng the orderly universe
of antiquity and the Middle Ages were demoli shed. The worl d appeared be
fore man located in nothi ngness, surrounded by nothi ngness and permeated by
nothingness through and through. At the begi nni ng of the 1 7th century the
Paris parl iament i ssued a decree that anyone coming out with a polemic against
ancient and uni versal l y recognized authors woul d be subject to capital pun
ishment. At the same ti me, the Hol y Inqui sition conducted two tri al s against
Gal i leo Gali l ei , the greatest physicist and mathematician of the time, and, by
threateni ng to burn him at the stake, forced him to publ icly reject the ideas of
Copernicus and Bruno, and sentenced him to house arrest for the rest of his l ife.
But ti mes were changi ng. To sol ve practical problems scientists found
it necessary to appl y methods forbidden by Aristotel i an science and to use
i ndi vi si bl e and i nfnitely smal l magnitudes. There was renewed interest in the
ideas of Democritus. Usi ng "i l l egitimate" methods, Kepler obtained formul as
for the volumes of various solids that bafed adherents of ancient rigor. He also
used these methods i n the i mmortal works i n which he establ i shed the l aws of
motion of the pl anets around the Sun. Using the ideas of his teacher Gali leo, the
Ital i an mathematici an Bonaventura Caval ieri 1
7
wrote a book titled Geometr
of indivisibles in which he stated principles that made it possible to compute
areas and volumes of various fgures by means of a general method. Toward
the end of the 1 7th century Newton and Leibniz i ndependently systematized
Chapter 1. Infnit and the universe 1 3
the methods of solution of a tremendous variety of problems, methods based
on the use of i nfni tely small and i nfni tely l arge magnitudes. This was the
genesis of mathematical anal ysi s (diferential and i ntegral calcul us), one of the
most remarkable creations of the human mind. Knowi ng the forces acti ng on
a body one could, by the methods of mathematical anal ysi s, determine that
body' s trajectory. In particul ar, one could in thi s way deterine the orbits of
planets and comets.
The i deas of Copernicus, Bruno, Gali leo, and Newton i nterested not only
scienti sts. At the end of the 1 7th century, the French writer Fontenellel
S
wrote
On the mUltiplicit o/worlds, a work transl ated i nto many l anguages. And the
1 8th-century German poet Albert von Haller wrote:
I amass countless numbers,
I pile mi l l ions into a mountain,
I pour ti me i n a heap,
Ranges of innumerabl e worlds.
When I look at you
From a reckless height,
Then I see that you are
Far above al l numbers and measures:
They are just part of you.
The new ideas were propagated i n Russi a by Lomonosov, the poet and great
founder of Russi an science. He expressed his i dea of the worl d' s infnity in
these inspired l i nes:
The chasm opened, ful l of stars,
The stars numberless, the chasm bottomless.
And the poet Sumarokov embedded these i deas i n hi s transl ation of the
bibl ical psal ms. His rival Tred' yakovski i denounced hi m to the Holy Synod i n
these words: "While reading the September 1 755 issue of ' Monthly works' ,
I, the undersigned, found i n i t holy odes by Colonel Aleksandr Petrov, son
of Sumarokov. Among them I found an ode based on Psalm 1 06, and saw
that, from the ei ghth stanza to the eleventh inclusive, it speaks - in its own
words and not those of the psal mist - of the i nfnity of the universe and of
the actual multitude of worl ds, and not about what i s possible owi ng to di vi ne
omni potence."
Afer receivi ng thi s denunciation, the Holy Fathers demanded that Czari na
El i sabeth should suppress the journal which, they said, "contai ns much that i s
contrary to the holy faith, especi al l y certain translations and works that assert
the exi stence of many, nay, innumerable worl ds, which is utterly repul sive to
the Holy Writ and to Christian faith and gives weak soul s reasons for natural i sm
and athei sm."
But the eighteenth century was coming, and the czari na "took no action"
on the humble request of the Holiest Synod.
Newton's world. In astronomy, physics and mathematics, the end of the 1 7th
century wi tnessed the triumph of ideas connected, in one way or another, with
the use of the i nfnite. There came into bei ng a picture of the world ruled by
Eucl i dean geometry and Newton' s laws of moti on. Scientists assumed that
knowi ng the positions of all material bodies at a given moment they coul d,
by sol vi ng the appropri ate di fferential equations, predict their posi tions at an
arbi trary future moment.
The two foundations of the whole edifce had nothing to do with one another.
Infni te space was completely unrelated to the matter that fl led i t - it was just
a stage that accommodated the drama of the worl d. By its very nature thi s
space was unrel ated t o everything external and remai ned forever the same and
i mmobi le. It woul d not change even i f al l matter suddenl y vani shed. In thi s
connection Einstein wrote:
Newton found that observed geometric magnitudes (di stances be
tween material points) and thei r changes i n time do not, i n a physi
cal sense, ful l y characterize motion . . . Thus, i n addition to masses
and to di stances between points that change i n ti me there exists
somethi ng else that determi nes the occurri ng events; this "some
thi ng" he took to be the relation to absolute space.
As a resul t of the successes of Newtonian mechanics and astronomy, the
world picture proposed by Newton gained universal acceptance. Doubts about
it came to be regarded as anti scienti fc.
The famous German phi l osopher Kant described the picture of the uni verse
accepted in the 1 8th century in these words:
In the i nfnite distance there are many such systems of stars, and
its parts are mutually rel ated. . . . We see the frst members of an
unbroken series of worlds and systems, and the frst part of thi s
i nfnite progression gives us an idea of the whole. There i s no
end here, there i s a trul y i mmeasurable chasm . . . . The universe i s
fl led wi th worl ds without number and end . . .
1 5
It shoul d be noted that the recognition of the i nfnity of the universe "co
existed peaceful ly" with faith i n God in the mind of Kant and i n the mi nds
of most of hi s contemporaries. In fact, some theol ogians argued that i t took
a more powerful God to create an i nfnite rather than a fnite world and saw
in the i nfnity of the universe a "proof of God' s omnipotence." It took hal f a
century of the activities of Vol taire and the Encycl opedi sts, and the horrors of
the French revolution, to make possible an i ntellectual cl i mate in which Lapl ace
could say to Napoleon who asked hi m why hi s works on celestial mechani cs
make no mention of God: "Your Highness, I have no need of this hypothesi s."
New complications. Not in vai n did Ari stotle warn about the shaky and vague
nature of the concept of i nfnity and of the complicati ons it coul d lead to. The
frst compl ications i n Newtonian physics and mathematical anal ysi s arose soon
after thei r creati on.
The students and fol l owers of Newton and Leibniz used the vague concepts
of the i nfnitely smal l and i nfnitely l arge, ful l of mystery, to sol ve the most
complex problems of astronomy, physics and mechani cs. They proceeded
recklessly. They unceremoniously added i nfni tely many terms without pausi ng
to ask whether or not the rules of operation applicable to fni te sums carried
over to i nfnite sums. But while the fundamental concepts of the new cal culus
struck mathematicians brought up on ancient rigor as nebulous, its practical
triumphs al l ayed for a ti me al l doubts. D' Al embert, 1 9 one of the great 1 8th
century French mathematicians, told his students "Go ahead, and faith wi l l
come."
But at the end of the 1 8th century came the frst si gns of trouble. Cases began
to accumul ate where i ncorrect appl ication of i nfni tel y small and i nfnitely l arge
magnitudes led to paradoxes. As a result, in the begi nni ng of the 1 9th century
these magni tudes were bani shed from mathematics and repl aced by the idea
of l i mi t. In this the works of Abel ,
2
o
Cauchy,
2 1
and Gauss,
22
the "prince of
mathematicians," pl ayed a collective role. The following excerpt from Gauss'
letter t o Schumacher,
2
3 written i n 1 83 1 , i s typical of hi s view of the i nfni te:
I object t o the use of an i nfni te magnitude as somethi ng completed;
this is never admi ssible in mathematics. One must not i nterpret
i nfnity l i teral l y when, strictl y speaking, one has in mi nd a l i mi t ap
proached with arbi trary cl oseness by ratios as other things i ncrease
without bounds.
Another area where complications arose was cosmology. The natural as
sumptions about the uni form di stribution of stars i n i nfnite space led unexpect-
edly to a paradox. It turned out that their collective brightness would be the
same as if a Sun glittered at every point i n the sky. It was an i mage an Indian
poet had i n mind many centuries earl ier when he exclaimed:
The sky above would shine
With boundless and awesome force
If a thousand Suns at once fashed in it.
Thi s was the so-called photometric, or Ol bers' paradox. There was another,
so-called gravitational paradox. It tured out that if the i nfni te Universe con
tained a fni te amount of matter, then all of i t would collect i n one place, i n a
si ngl e l ump. And if the total mass were i nfnite and uni formly distributed, then
thi s woul d lead to a mutual equi l i bration of gravitational forces.
Both paradoxes can be el i mi nated by assuming that matter i s di stributed in
the Universe nonuni forml y. But this leads to the hypothesis that the Universe
has a center. Thi s is no less surpri si ng than the notion of fniteness of mass i n
the Universe. Here we note that contemporary observations show that matter i s
more or less uni forml y distributed throughout the Metagalaxy ( = total physical
Universe) .
The photometric and gravi tational paradoxes coul d be removed only afer
the i ntroduction, in the 20th century, of a new theory of the structure of the
Universe based on Einstei n' s general theory of rel ativity. Before tel l i ng you
about these new notions we' l l make another excursion i nto mathematics.
Curved space. The fol l owing passage i s taken from the novel The Astronauts,
by the science-fction writer Stani sl aw Lem:
The picture of the stary sky changed very rapidly. Yesterday' s
photographs didn' t match today' s. I t seemed that a mysterious
force was pushing them apart. It was as i f they were specks on the
surface of a constantly expanding bal loon. The gravitologist of the
astral expedition was at the computer for days. The expedition was
approachi ng a heavy star that was strongl y curving the surrounding
space. To determi ne the course of the ship thi s curvature had
to be constantly computed. This was a cruci al check. Humans
were encounteri ng such strong gravitational felds, such l arge space
curvatures, for the frst time. Einstei n' s equations were now bei ng
put to a test. The success of the expedition, and even the l ives of
its members, depended on them.
Chapter I. Infnit and the universe 1 7
What i s most bafi ng i n thi s passage to a reader unfami l i ar with modern
mathematics i s the reference to the curvature of space. It is much more di fcul t
to i magine curved space than a curved l i ne or surface. Smal l wonder that the
creation of the concept of curved space i n the second half of the 1 9th century
i nspired the (German) epigram:
Die Menschen fassen kaum es
Das Krimmungsmass des Raumes
(people seem unable to grasp what i s [meant by] the measure of curvature of
space).
The common objections to the noti on of curved space are these: A curved
l i ne cannot be brought into coincidence with a straight l i ne; i t can be di sposed
i n a pl ane or i n space. Si mi l arl y, a curved surface can' t be disposed i n a plane
- this requi res at least three-di mensi onal space. Thus curved space must l ie
in some ambient space of four, or perhaps fve, dimensi ons. And si nce no
one has observed four-dimensional space, the space we l i ve i n can' t possibly
be curved. Incidental l y, the notion of four-di mensi onal space i s a favorite of
science-fction writers. In many of H. G. Wel l s' stories voyages take place in
the fourth dimensi on.
It turns out that curvature of space has nothi ng to do with the fourth di men
sion and i s, so to say, one of its i ntri nsi c aspects. In fact, the curvature of space
can be determi ned wi thout leavi ng it, by just caryi ng out measurements i n i t.
To cl arify how thi s i s done we frst explain how to fnd the curvature of a
surface without l eaving it, just by measuring di stances between its points .
Geometry on planet Eks. For millennia people thought that the Earth was fat.
But observations of the shadow of the Earth during l unar eclipses suggested to
the ancient Greek scientists that the Earth was spherical ; i n fact, Eratosthenes
managed to compute its radi us with considerable accuracy. For a ti me, the idea
of the fatness of the Earth reasserted itself, and i t took Magel l an' s ci rcum
navigation of the worl d to establ ish the spherical shape of the Earth.
And now i magine pl anet Eks whose sky i s permanentl y shrouded i n cl ouds
and whose oceans are not navigable. It i s i nhabited by creatures capabl e of
rational thought. Can they decide whether they l ive on a fat piece of l and
surrounded by water or on a spherical pl anet? In earthl y terms, the question i s:
Coul d the spherical shape of the Earth have been establ i shed without Magel l an' s
voyage? Here we are not questioning the geographi cal or hi storical si gni fcance
of his voyage. Ours i s a purel y mathematical concern, namely the possi bi li ty
of establ i shi ng the curvature of the Earth wi thout ci rcumnavi gating it.
1 8
In Search of Infnit
To understand how the idea of the curvature of a surface ari ses, we trace
the evolution of geometry on the pl anet Eks.
Geometry begi ns as an experi mental endeavor. It becomes a ful l -fedged
science as a resul t of a theoreti cal general ization of centuries of observation
pertai ni ng to properties of space. Si nce the sky of pl anet Eks i s permanently
obscured by clouds, its geometers were restricted to making measurements on
its surface. It turned out that among al l the l i nes joi ni ng two points there is
al ways a shortest one ( see Figure 1 ) . These shortest l i nes were cal l ed "straight
line segments."
Figure 1 .
Of course, an "outside" observer would say that these l i nes are curved rather
than straight, that they are arcs of sections of a sphere by means of di ametrical
planes. But Eksl i ngs couldn' t look at thei r pl anet from the outside, and cal l ed
the shortest curves on the surface straight l i nes. (It is worth poi nting out that
we often say that "the road is as straight as an arrow" when, in fact, i t is an arc
of a di ametrical section. )
Next the Eksl i ngs establ i shed vari ous properties of these "straight l i nes."
At frst they could perform onl y local measurements, with poor preci si on. As
a resul t, they concluded that the properties of "straight l i nes" on the surface of
thei r pl anet were the same as the properties of straight l i nes i n the pl ane. In
parti cul ar, they thought that two points determine a uni que "straight l i ne" and
that two "straight l i nes" i ntersect i n just one point (the di ametrical l y opposite
second point of i ntersection was i naccessible to the observers). Final ly, the
Eksl i ng geometers concl uded that thei r "straight l ines" are i nfnitel y l ong. The
very idea that if they moved along a "straight l i ne" in the same di rection they
woul d eventually return to the starting point struck them as absurd, inconceiv
abl e, and contrary to common sense. They also concl uded that the surface of
thei r pl anet is i nfnite, and accused opponents of thi s idea of being gui l ty of
mortal si n.
Further study of the properties of "straight l i nes" showed that, withi n the
l i mits of avai l able preci si on of measurement, the sum of the angles of a triangle
Chapter 1. Infnit and the universe 1 9
was equal to 1 800, that the square of the hypotenuse of a right triangle was equal
to the sum of the squares of its shorter sides, and so on. In other words, the
Eksl i ngs constructed Eucl i dean geometry and were certain that i t was appl i cable
to the surface of thei r planet.
In ti me, the Eksl ings devel oped thei r technology and coul d survey ever
larger pieces of thei r pl anet' s l and (we recal l that their oceans were not navi ga
ble) . These more gl obal measurements contradicted Euclidean geometry. To
understand what was at i ssue, take three points A, B, C on a sphere as i n Figure
2a. Connect them by Eksl i ng "straight l i nes," or, in our parlance, by arcs of
meri di ans AB and A C and the equatorial arc B C. The angles of the triangle
AB C are all right angles, and so thei r sum i s 2700 and not, as i n Eucl i dean
geometry, 1 800 It i s "too l arge" by 900
Cal l the difference (a
+ f +
y) - 1 800 between the sum of the angles a,
f
, y
of a spherical tri angl e and 1 800 its angular excess. (The notion of angular excess
retai ns its signi fcance on an arbitrary surface. ) Then the angul ar excess of our
triangle i s 900 Other triangles have different angul ar excesses. Thus i n the
triangle A B D in Figure 2b the angles B and Dare 900 each and the angle BA D
is 1 800 so that the angul ar excess of triangle A B D is 1 800
B
D
Figure 2a. Figure 2b.
Parallel transport. The nonzero angul ar excess of tri angles was not the onl y
thing that persuaded Eksl ings that they li ved on a curved surface. Pythagoras '
theorem also fai led. The triangle in Figure 2a is at the same ti me a right tri angle
and an equi l ateral triangl e!
The study of paral lel straight l i nes on the surface of the pl anet (which they
defned as straight l i nes perendicular to one and the same straight l i ne) also
led Eksli ngs to unexpected resul ts. If a segment i s transported along a cl osed
plane curve paral lel to itself, then i t wi l l return to its i ni ti al position without
change of di rection (see Figure 3a) . The Ekslings veri fed this in the smal l .
But measurements i n the l arge on the surface of Eks led t o al together diferent
results.
For example, consider the triangle ABC i n Figure 3b. Thi s triangle has three
right angles. We draw at A a "segment" perpendicular to A B and transport it,
paral lel to itself, along the contour of tri angle ABC. When we reach B, the
di rection of our segment is the same as that of the equator. Si nce the equator
is also a "straight l i ne," parallel transport of our segment along BC yields a
segment at C that is again di rected along the equator. When this segment
i s paral lel-transported along the meridian C A, then we obtai n at A a segment
rotated rel ative to the i ni ti al segment at A through 900 Again, paral lel transport
of a segment along the contour of A B D i n Figure 2b rotates it through 1 800
Note that in both cases the transported segment is rotated through an angle
equal to the angul ar excess of the relevant triangl e. Thi s turns out to be true for
all triangles !
a
Figure 3a. Figure 3b.
An i nteresting case of paral lel transport i s that of parallel transport of a
segment around the equator (see Figure 4) . At frst one is i ncl i ned to think that
the segment returns to its i ni ti al position without any rotation. But thi s is not
so. Actually, the segment is rotated through 3600 To see this supplement the
equator wi th an arc A B of a meri dian traversed i n both di rections. In this way
we obtain a "triangle" A B A two of whose angles are 900 each and the third is
3600 The angul ar excess of thi s triangle i s 3600 (To understand thi s case of
paral lel transport, it may be hel pful to think of "tri angle" ABA as the union of
two tri angles l ike the one i n Figure 2b. (Transl .
Measurement of curvature. We repeat: By measuring the angle sum of
triangles, by noting the rotation of a segment under paral lel transport along a
cl osed contour, and by tryi ng to verify Pythagoras' theorem, the i nhabitants of
2 1
pl anet Eks concluded that they l ived not on a plane but on a curved surface. As
a measure of curvature of a piece of this surface, they took the angle of rotation
of a segment under paral lel translation along its boundary. Thi s curvature coul d
also be computed by subdividing the piece of surface i n question i nto tri angles
and adding thei r angular excesses. (Thi s approach i s justifed by noting that
the angular excess of a triangle that is the uni on of two triangles is the sum of
their angular excesses. )
Figure 4. Figure 5.
It turned out that the l arger the area of the piece of surface the greater its
curvature. Specifcally, the angul ar excess of a tri angle with area S and angles
with radi an measures a, f, and y is
a
+
f
+
y - r = kS. ( 1 )
This showed that the curvature per uni t area was constant. The Eksl i ngs took
k as a measure of curvature of the surface of thei r planet.
Among all surfaces there i s just one for which the angular excess of a triangle
per unit area i s constant. Thi s surface i s a sphere. And so the Eksl i ng geometers
reali zed that they l i ved on a sphere and not on a plane. They computed its radius
with rel ative ease. Note that the number k does not depend on the choice of
triangle. Take, say, the triangle ABC i n Figure 2a. The radian measure of its
angul ar excess i s r /2. Its area i s one eighth of the area of a sphere of radi us
R, that i s r R
2
/2. Hence ( 1 ) reduces i n thi s case to the equality r /2 = k
.
7 [' ,
so that k = 1 / R
2
. It fol l ows that for an arbitrary spherical triangle with angl es
a, f, y and area S we have
where R is the radi us of the sphere. Thi s formul a al l ows us to compute the
radius of the sphere by measuring the angles and area of a triangle on that
22
In Search of Infnit
sphere. Of course, this is not the most convenient way of computing the radi us
of a sphere, for i t cal l s for accurate measurements of the angles and area of a
triangle on that sphere. To compute the radius of the Earth, one computed the
length of an arc of a meri di an by making certain observations of the stars.
Gaussian curvature. We saw that the curvature k of a sphere per uni t area
is 1 / R
2
. Thus the l arger its radius the smaller the curvature of a piece of its
surface of unit area; the surface of a ball i s far more curved than the surface of
the Earth.
Gauss proposed a si mi l ar measure of the curvature of an arbitrary surface.
On any surface we can construct a geometry i n a manner si mi l ar to the con
struction of spherical geometry. The role of recti l i near segments i s pl ayed by
the shortest curves, or geodesics, that is curves whose length i s shorter than the
lengths of al l other curves connecti ng two given points. The frst to encounter
these curves were surveyors. Incidental ly, Gauss' interest i n the geometry of
surfaces was the resul t of his i nvol vement with a geodesic survey of the ki ngdom
of Hannover that l asted several years.
One can construct geodesic triangles, quadrangles, and so on. Just as on a
sphere, the angle sum of a (geodesic) triangle on an arbitrary curved surface
i s, in general , di ferent from r. As on a sphere, we defne the curvature of a
triangle per uni t area as (a
+
f
+
y - r) / S. On non-spherical surfaces thi s
number can b different for different tri angl es; i n fact, it can be positive for
some triangles and negative for others.
To compute the curvature of a surface at a point, we compute the number
(a
+
f
+
y - r) / S for a sequence of ever smaller triangles around this poi nt
and seek its l i mi t. Thi s i s Gauss' defnition of the curvature of a surface at a
poi nt; hence the term Gaussian curvature. It is positive i f the triangles i nvol ved
have angle sums greater than r and negative i f these sums are less than r .
The Gaussi an curvature of a convex surface is everywhere positive. The
Gaussi an curvature of a torus (see Figure 5) is positive at some poi nts and
negative at others.
A remarkable property of Gaussi an curvature i s that it i s unchanged by
bendi ng the surface, that i s by subjecting i t to transformations that don' t change
distances between poi nts. (Such transformations are called isometries. ) It fol
l ows that the Gaussi an curvature of a cyl i nder i s zero at al l points; afer al l , a
cyl i nder is the resul t of bending a piece of a pl ane, and the Gaussian curvature
of the l atter is zero everywhere. Si mi l arly, the Gaussi an curvature of a cone i s
zero at al l of i ts points other than the vertex.
23
The pseudosphere and hyperbolic geometry. The Gaussi an curvature of a
sphere is constant and positive at al l of its points. It is interesting that there
is a surface of constant negative Gaussi an curvature. Thi s surface is called a
pseudosphere and is obtained in the following manner.
Consider Fi gure 6. Imagine a person at A with a dog on a taut leash at O.
The dog runs at constant speed along the straight l ine 0 x and its owner runs
afer i t so that her velocity i s always di rected along the (taut) leash. (Thi s means
that the owner i s initially runni ng i n the di rection A 0. ) As the dog continues
to run along the straight l i ne Ox, the angle between that l i ne and the directi on
of the owner' s velocity continues to decrease, whil e the di stance between dog
and owner stays constant. The curve traversed by the owner i s called a tractri
and has the following property: Draw a tangent to the tractrix at a point M.
I f the tangent at M i ntersects Ox at N, then the length of the segment M N i s
constant. (Thi s i s clear if we bear i n mi nd that M N represents the taut leash. )
If we refect the tractri x i n Figure 6 i n the straight l i ne 0 A and rotate the
resulting curve about the Ox-axi s, then we obtai n the surface of revolution rep
resented i n Figure 7. Thi s surface i s the promised pseudosphere. As noted, its
Gaussi an curvature i s constant and negative at all points (other than the points
of the ci rcle generated by the point A i n Figure 6) . What i s remarkable about
thi s surface i s that its geometry i s the same as that of a part of the hyperboli c
pl ane. Thus the di scovery of the pseudosphere was an i mportant event i n the
evolution of non-Euclidean geometry.
Gauss and Riemann. Gauss gave a preci se defnition of the notion of curvature
of a surface. Next on the agenda was the problem of defni ng the curvature of
space. Thi s problem was sol ved by Bernhard Riemann,
24
one of the most
remarkable mathematicians of the 1 9th century.
Riemann dealt with the problem of curvature of space i n his i naugural lecture
of 1 854. At that ti me, the (l audable) custom was that a begi nni ng i nstructor was
expected to present a lecture before the members of the faculty who would then
be i n a position to judge hi s teaching abi l ities. Riemann ofered a few topics
for such a lecture and Gauss selected the one that i nterested hi m most - "On
the hypotheses which lie at the foundations of geometry." It i s safe to assume
that the l isteners were not greatly i mpressed by Riemann' s pedagogical talent.
The only l i stener who completely understood Riemann' s lecture was Gauss.
Usi ng vi rtually no formul as and computations, Riemann devel oped general
ideas on multidimensional mani folds, on measuring length on such mani fol ds,
24
In Search of Infnit
on thei r curvature, and so on. In thi s lecture Riemann posed questions that con
tinue to i nterest theoretical physi ci sts to thi s day, namely, whether our space
is conti nuous or di screte, and whether Euclidean geometry is appl icable to
i nfnitesimal regions of space. The profundity of Riemann' s ideas fascinated
Gauss, and, accordi ng to some members of the audience, he went home sunk
i n deep thought.
Figure 6. Figure 7.
Curvature of a space. What was Riemann' s notion of curvature of a space and
how was it to be measured? Riemann used the same idea that Gauss had used
for measuring the curvature of a surface, that i s, he computed the angle sum of
a geodesic tri angle and the amount by which i t difered from 1. But here the
di ffculty i s that there are many planes passi ng through a point i n a space, and
therefore many relevant tri angles. Hence Riemann spoke not about curvature
at a point but about curvature at a point in the di rection of a given plane.
If for al l triangles i n a space the angle sum i s 1, then i n this space Eucl idean
geometry is val i d. Such a space has zero curvature and i s said to befat. If there
are tri angles wi th angle sum greater than 1, then the curvature of the space at
the relevant points is positive. Fi nally, if there are tri angles with angle sum less
than 1, then the curvature of the space at the relevant points i s negative. Thi s
shows that there i s nothi ng mysteri ous about the notion of curvature of a space.
What counts i s the possi bi l ity that the angle sum of a tri angle can devi ate from
the prescribed Eucl i dean val ue.
Spaces that have the same curvature at al l points and i n al l planes are of
great i nterest (they are known as constant-curvature spaces) . In such spaces
bodies can move from pl ace to pl ace wi thout changi ng their di mensi ons. If the
curvature of a space varies, then the di mensions of a moving body are subject
to change - the body i s di storted.
Riemann posed the question of the curvature of the space we l ive i n. He
wrote:
Either the actual things forming the groundwork of a space must
consti tute a di screte mani fold, or el se the basi s of metric rel ations
must be sought outsi de that actual ity, i n binding forces that operate
upon i t.
A deci sion on these questions can be found only by starti ng from
the structure of phenomena that has been approved in experience
hitherto, for which Newton laid the foundation, and by modifyi ng
thi s structure gradual l y under the compul sion of facts whi ch it
cannot explain . . .
Thi s path leads into the domain of another science, i nto the real m
of physics, into which the nature of thi s present occasion forbids
us to penetrate.
25
The pursuit of curvature. Riemann' s prediction began to be real ized only at
the begi nni ng of the 20th century. It was then that the question of the curvature of
space moved from the domain of abstract mathematical considerati ons to that
of concrete physical theories. The great physicist Albert Einstei n pondered
the reason for the equality of inertial and gravitational masses and eventuall y
created the general theory of rel ativity that radical l y changed our notions of
the relation between matter and space. Whereas, as noted earlier, i n Newton' s
physics space was completely i ndependent of the matter i n i t , the new theory
ruled out the very exi stence of empty, that is fel d-free, space. Al so, i t turned out
that space and ti me cannot exist independently, but only i n a state of i nseparable
connection with one another, and only as a structural property of a fel d.
Einstei n' s theory was so complex and unusual that many scientists reacted to
it with di sbel ief (and many fai led to understand it). An additional di ffcul ty was
that Einstei n had no experi mental proofs of hi s theory other than the equal ity
of the two forms of mass, and physi ci sts acknowledge a new theory onl y i f, in
addition to expl ai ni ng al l relevant known phenomena, i t predicts new, as yet
unobserved ones.
Among the many unexpected, and at frst si ght paradoxical , consequences
of general rel ativity, there i s one that lends itself rel ativel y easi l y to experi mental
veri fcation. The theory i mplies that attracting masses wi l l curve space. Thus
one had to give an experi mental proof of the curvature of space.
A long ti me ago it was establ i shed by means of physical experiments that
l ight fol lows a path that mi ni mizes ti me, and that the speed of l i ght in empty
space i s always the same. Therefore (the paths of) l ight rays (in empty space)
are taken as geodesics i n space.
To determi ne the curvature of space i t was necessary to measure the angle
26
In Search of Infnit
sum of a triangle made up of l ight rays. But even in the case of tri angles with a
star as one vertex, the angle sum i s, within experi mental error, equal to 1 , so that
the curvature of space, i f any, is very small . Incidental l y, unsuccessful attempts
to prove the non-Eucl i dean nature of ordi nary space by di rect measurement of
the angle sum of cosmic tri angles go back to Lobachevski .
25
Ei nstein suggested that rather than determine the angle sum of cosmic tri
angles one should observe changes i n paths of l ight rays. His theory i mplied
that if one photographed a star twice - once when a beam of its l ight passes
far from the Sun and a second ti me when i t passes cl ose to the Sun, then one
woul d notice a shift i n the star' s position due to the bending of the l ight beam.
Cl assical physi cs al so predicts such an effect, but Einstei n' s theory predicted
an effect that was double that predicted by the cl assical theory. Even thi s "l arge"
effect is very small ; in fact, it is less than two seconds of arc (the angle subtended
by a two-kopeck coin at a di stance of 1 200 meters) . Nevertheless, the effect i s
measurabl e. And whi l e sunl ight "drowns out" stars i n the Sun' s vicini ty, the
necessary observations can be carried out during a total ecl ipse of the Sun.
Accordingly, i n the spri ng of 1 9 1 9 two scienti fc expeditions were dis
patched to measure the curvature of space, one to the southern shore of Africa,
and the other to northern Brazi l . Observations carried out by the expeditions
on 29 May 1 9 1 9 fully confrmed Einstei n' s prediction: the actual star shif was
equal to that predicted by Einstei n' s theory.
Another proof of the curvature of space was derived from observations of
Mercury. Thi s pl anet i s cl osest to the Sun and so i s most affected by the cur
vature of the space around the Sun. As a resul t of thi s curvature, the orbit of
Mercury rotates sl i ghtly wi th each rotation of the pl anet around the Sun. The
orbit al so rotates because of planetary attraction. But the observed rotation of
the orbit of Mercury exceeded by 41 seconds of arc per 1 00 years the rotation
due to pl anetary attraction alone. Based on Einstei n' s formulas, the computa
tion of the rotation of the orbi t of Mercury due to the curvature of space yielded
the fgure of 41 seconds of arc per 1 00 years. Thi s resul t expl ained a riddle that
had been troubl i ng astronomers for many years and suppl ied a new confrma
tion of the theory of rel ativity.
The expanding Universe. Relativity theory confronted astronomers with the
fundamental question of the structure of actual space. If i t is curved and its
curvature i s positive, then i t coul d very wel l be l i ke a three-di mensional sphere
- that i s, unbounded but fni te in size. Some phi l osophers rejected the very
possi bi l ity of the fnite size of actual space. But their arguments carried as
Chapter 1. Infnit and the universe 27
much conviction as the arguments of Eksl i ng scienti sts who thought that they
l i ved on an i nfnite pl ane and not on a sphere of fnite si ze. The answer to the
question of the structure of actual space called for astronomical observations
and not for mere specul ation. The frst attempt to construct a model of the
Universe on the basi s of the new theory of gravi tation was by Einstein hi msel f,
but it was not very successful . Einstein tried to construct a model of a stationary
Universe that did not change i n time. It seemed to echo the words of Aristotle,
who wrote that "throughout time, according to the chronicles passed on from
generation to generati on, we fnd no trace of changes ei ther in the di stant sky
as a whole or i n any of its accessible parts."
To construct a stationary model of the Universe, Einstei n had to assume the
exi stence of forces that repel galaxies from one another and are proportional to
the di stance between them. At that time, there were no known experi ments that
confrmed the exi stence of such forces. Later Einstei n described the assumption
of the existence of such forces of repul si on as the "bi ggest mi stake of my l i fe."
An unexpected solution was proposed in 1 922 by the young Leni ngrad
mathematician Aleksandr Aleksandrovich Friedmann.
2
6
In a paper "On the
curvature of space" Friedmann showed that the equations of general rel ativity
have non stationary solutions - that i s, solutions that i mpl y that the Uni verse
expands or contracts. In the frst case the curvature and density of matter
decrease, and in the second case both i ncrease. It is i nteresting that Friedmann' s
major works deal with dynamical meteorol ogy and not wi th theoretical physi cs.
Friedmann' s solution was so diferent from establ i shed views of the struc
ture of the Universe that the frst reaction of the scientifc community was to
assume that he had made a mistake. This was al so the tenor of a short note by
Einstein that appeared in 1 922 i n the same journal that had publ i shed Fried
mann' s paper. Soon Einstei n received from Friedmann a detailed l etter that
di spelled al l hi s doubts. And while Einstein was at the ti me universally recog
nized as the leading physicist and Friedmann was just a begi nni ng researcher,
the venerable scientist di d not hesitate for a moment how to respond. On 1 3
May 1 923 the editor of the Joural of Physics (Zeitschrif fUr Physik) obtained
a letter from Einstein which he subsequently publ i shed under the title "A note
on A. Friedmann' s paper on the curvature of space." The letter read:
In a previous note I criticized the paper menti oned in the title of
thi s note. But Friedmann' s l etter convi nced me that my critici sm
was based on computational errors.
I regard Friedmann' s results as correct and as shedding new l ight.
It turns out that the feld equations admit not onl y static but also
dynamic (that is varyi ng with ti me) central l y-symmetric solutions
for the structure of space.
Later i t turned out that there are no static solutions. Einstei n' s model was
unstable, and so a homogeneous and i sotropic (the same i n al l di recti ons) model
of the Universe must be dynamic. Following Friedmann, many physi ci sts and
astronomers - Lemaitre,
27
de Si tter,
2
8
and others - constructed dynamic models
of the Universe.
The question of an experi mental confrmation of the val idity of one or an
other model of the Universe arose once more; i n particul ar, the question to
be decided was whether the Universe expands or contracts. Then astronomers
recalled that al ready at the begi nni ng of the 20th century the American as
trophysi ci st V. M. Sl iphe
r
9
had measured the radial velocities of galaxies. It
turned out that 36 of the 41 gal axies i nvestigated by Sl i pher recede from us,
and some of them do so at speeds of 2000krsec (l ater i t turned out that the
remai ni ng 5 gal axies approach us because of the actual motion of the Sun i n
our Galaxy) . Sl ipher' s i nvestigations were continued by E. P. Hubble,
3
D
who
determi ned not only the velocities of galaxies but also thei r di stances from the
Sun. His i nvestigations establ i shed the fol lowing remarkable l aw: the speed
with which a gal axy recedes from us is proportional to its di stance. In other
words, the most di stant galaxies recede from us fastest.
Thi s was a conclusi ve proof of Friedmann' s guess. The Universe i s expand
i ng. Thi s is not to say that the Earth i s the center from which galaxies recede
in al l di recti ons. What it does mean is that every observer wi l l see the same
picture - galaxies receding i n all di rections from the spot she occupies.
The picture of the expanding Universe made possible an expl anation of the
gravitational and photometric paradoxes. It turned out that l ight from di stant
stars does not reach the Earth, that we can only see galaxies that are cl oser than
the so-cal led visible horizon. As for the possi bi l ity of a contracting Universe,
we note that during such periods, i f any, the photometric paradox woul d set in
with a vengeance, and the food of energy engul fng the Earth would make l ife
impossibl e. In other words, l i fe is possible during periods of expansion of the
Uni verse but i mpossible during peri ods of its contraction.
The past and future of the Universe. Afer the appearance of dynamic models
of the Uni verse there arose a number of questions now pondered by theoretical
physici sts: When di d the scatteri ng of the galaxies begi n and what preceded it?
Wi l l the scatteri ng persi st forever or wi l l there be a period of contraction? and
so on. Today most scienti sts agree that many bi l l ions of years ago the Universe
Chapter 1. Infnit and the universe 29
was in a superdense state. It appears that the density of this clot was greater
than the density of matter i n atomic nuclei - that is, greater than 1 0
1 4
g/cm
3
.
Some computations give densities of the order of 1 0
93
g/cm
3
. One thi ng is clear,
namely that matter was in a state completely unknown to today' s physi ci sts,
and that neither gravitational l aws nor quantum theory were appl icable to it.
On the basi s of observations of the Universe i n its present state, theoretical
physici sts try to arrive at its state shortly after the "big bang" wi th which, they
think, its hi story began. They compute when the elementary particles began to
form, when these particles began to coalesce into atoms, how the temperature of
the Universe began to change. In particular, these i nvestigations have shown that
if the Universe was origi nally "hot," then we must still be recei vi ng today a form
of electromagnetic radi ation called cosmic background radi ati on. Di ferent
vari ants of the theory have assigned to thi s radi ation temperatures as low as a
fraction of one degree on the Kel vi n scale and as high a 20-30 degrees on that
scale.
In 1 964 two Soviet theoretici ans, A.G. Doroshkevich
3
1
and I. D. Novi kov,
3
2
computed by how much cosmic background radi ation exceeds the i ntensi ty
of other sources of radiation i n the centi meter range of the spectrum. But
their paper, poi nti ng out the possi bi li ty of a deci si ve experi ment, was somehow
overlooked. Cosmic background radi ation was accidentally di scovered (by
Americans) i n 1 965 i n the course of work with an antenna for tracking satel lites.
Its temperature turned out to be 3 degrees Kel vi n. This discovery greatl y
i ncreases the probabi l ity of the model of a hot i ni ti al state of the Universe.
We won' t try to give a detai led account, based on modern views, of the
creation and evolution of stars and galaxies. In thi s area there are many con
tentious poi nts. Of al l these theories (i ncl uding the "big bang") the eminent
physicist L. N. Bri l louin
33
said that "ai l thi s i s too beautiful to be true and too
unl ikely to be believed." Nevertheless, modern science has no other theory that
expl ains the mul titude of puzzl i ng phenomena i n the Universe.
Another question of i nterest to scientists i s, of course, the future of the
Universe. Here there are two possible scenari os: either the Universe wi l l
continue to expand i ndefni tel y and at some very di stant ti me the galaxies wi l l
have drifed so far apart that the inhabitants of one won' t be able t o see any of
the others, or at some moment the scattering of the galaxies wi l l be arrested and
they wi l l begi n to come closer t o one another. Computations show that which
of these tendencies will prevail depends on the average density of matter i n the
Universe. If thi s density i s less than 1 0
-
2
9
g/c m
3
, then expansi on will continue
forever. Otherwi se, at some moment, the Universe wil l begi n to contract. Some
scienti sts believe that the contraction wi ll continue until the Universe wi ll agai n
achieve the state of a superdense clot and the same chain of events wi l l agai n
unfol d. One cannot help but remember the words of Heracl itus about the fre
that regularly blazes up and is regul arl y extingui shed.
And what, i n fact, i s the density of matter i n the Universe? What makes i t
di ffcul t to answer thi s question i s the possi bi l ity that a great part of matter i s i n
a state that makes observations i mpossible. (Here we mention bl ack holes that
emit no radi ation, i ntergalactic gas, and possi bl y neutrinos that can penetrate
enormous masses of matter wi thout i nteracting with i t. ) According to current
estimates, thi s density is one-thirtieth less than the critical value, so that it seems
that we need not fear that the Universe wi l l begi n to contract. But each year
bri ngs so much new knowledge that it i s premature to try to give a defnitive
answer to thi s questi on.
Is the Universe infnite? One of the most difcult questions of cosmol ogy i s
the question i n the title of thi s section. If the average density of matter i n the
Metagalaxy is more than the critical value of 1 O
-
2
9
g/cm
3
, then we obtain an
open model of the Metagal axy. If thi s density is less than 1 0
-
2
9
glcm
3
, then we
obtain a cl osed model . Whereas the open model accords with the assumption
of a Uni verse of i nfni te size, the cl osed model has positive curvature, and so
its si ze i s bounded from above by some number. At the moment there are no
conceivable methods of di sproving the claim that our Universe i s cl osed and
that its si ze does not exceed 1 0
25
km.
The question of the fni teness or i nfniteness of the Universe i s further
compl i cated by the fact that, from the viewpoint of rel ativity, ti me and space
don' t exi st separatel y but only as a singl e entity. Using this fact, the Russi an
cosmologist A. L. Zel ' manovich
3
4
constructed absol utely bafing model s. In
one of these model s space has an i nfnite vol ume i n one coordinate system
and a fni te vol ume i n another. In another model , space of i nfnite vol ume
in one coordi nate system is part of space of fnite vol ume in another system.
There are si mi l ar paradoxes i nvol ving ti me and connected wi th observations of
gravitational col l apse (the shri nking of a burned-out star to a point under the
action of gravi ty) . In one coordinate system this process fts i nto a fni te ti me
i nterval and in another coordi nate system i t l asts i nfnitely long.
We see that the usual contraposition of the fnite and i nfnite i n space and
the fnite and i nfnite i n ti me as mutual l y excl usive possi bi l ities, and, quite
general l y, our customary formul ation of questions i nvol ving the fni te and the
i nfni te in space and ti me cannot be regarded as correct in all cases.
3
1
Another complicating factor is the possi bi l ity of an inhomogeneous and
anisotropic (having diferent properties i n di ferent di rections) Universe. These
possi bi l ities can favor model s of the Universe very di ferent from a sphere in
three-dimensional space. (Note that even surfaces of zero curvature i ncl ude not
only parts of the pl ane but al so cyl i nders and cones. ) Thus there are i nfni tel y
many admissibl e model s of the Universe. But whil e the supply of model s i s
i mmeasurably richer than it was before the advent of rel ativi stic cosmol ogy,
deal ing wi th them i s i mmeasurabl y more complex than before.
Al l these complexi ties show that i t is hardl y possibl e to give a si mple answer
to the question posed at the begi nni ng of thi s secti on.
The question of the i nfni te di vi si bi l ity of matter has also become i mmea
surabl y more compl i cated than before. At the begi nni ng of the 20th century i t
was thought that matter consi sted of indivisibl e atoms. Then Rutherford found
that atoms consist of nuclei and electrons. Next came the di scovery of protons,
and, a few years l ater, of neutrons and positrons. At the end of the thirties,
the neutron-positron model of the atomic nucleus came i nto bei ng. Thi s model
was so apt that it made possible the sol ution of the problem of the release of
atomic energy. But even before the activation of the frst atomic reactor came
the di scovery of new elementary particles - mesons of vari ous ki nds. After that,
the number of elementary particles grew al most exponentiaIly. There were an
tiprotons and anti neutrons, various hyperi ons and antihyperions, resonances,
and so on. Then i t turned out that neutrons can change i nto protons and protons
into neutrons, so that for these particles the term "consists of" l ost al l meani ng.
To bri ng order i nto the worl d of elementary particles, quarks, particles of
a new ki nd, were devised. Whi l e they can be used to explain everything i n a
satisfactory manner, no one has observed them. Quarks differ by color (thi s term
is strictly a convention) . Thi s led to a new branch of physi cs - chromodynamics.
According to very recent views we wi J I never see "J i ve" quarks. The reason for
this is that i nsi de elementary particles the forces that bind quarks increase with
the di stance between them (much as stretching a spri ng increases its tension) .
Al l this shows that it i s di ffcul t t o say how one shoul d, at thi s point, think of
the i nfni te divisibil ity of matter.
Chapter 2
1hc myslcrcs 0l nhnlc scls
A concept without defnition. Pythagoras, Zeno, Pl ato, and Ari stotle al l di s
cussed the one and the many. One Pythagorean defned a (natural) number as a
collection of units, and in Book VII of the Elements Eucl i d writes that "Number
i s a col lection consisting of units" (i n ancient Greek mathematics one was not
regarded as a number) .
But the "set-theoretic boom," that i s, the wi de use of set theory in various
areas of science and technol ogy, came onl y in the 20th century. Why was it pos
sible to get along without thi s concept before? There i s a si mple answer to thi s
question: those who di d not know about sets resembled Mol iere' s hero who di d
not know that he spoke prose. They dealt with sets al l the time but di d not refer
to them as such. For example, an economist planning the connections between
departments i n a factory did not think about each individual machine tool but
about, say, al l l athes or mi l l i ng machi nes and about thei r output. Si mi l arl y, an
ofcer of a certain rank prepari ng a mi l i tary operation thought of the actions
of hi s whol e company, batal l i on, regi ment or di vi si on and not of the acti ons of
individual soldiers. Al l these people worked wi th sets of objects, studied them
as si ngl e collectivities. A mathematician woul d say that they dealt with sets of
elements and not with individual elements.
Unfortunatel y, he coul d not have said what a set is. When defni ng a
new concept mathematicians reduce it to known concepts. But one must start
somewhere, and there is no concept in mathematics more basic than a set.
Thi s i s not surpri sing i f we bear i n mind that al most al l sciences begi n
with cl assi fcati on, with grouping of si mi l ar objects or concepts i nto a si ngl e
whol e and wi th separating of di ssi mi l ar objects. Before there can be biol ogy,
people must learn to tel l a wolf from a jackal and a rabbit from a hare. Before
there can be mi neralogy people must for centuries col l ect stones and tel l grani te
from fi nt and mal achite from j asper. From the viewpoint of mathematics al l
cl assi fcation reduces to forming sets on the basi s of certain cri teri a. That' s
why it i s i mpossible to reduce the concepts of a set to si mpler concepts.
We can speak of the set of chairs i n a room, of the set of protons on Jupiter,
of the set of words in the works of Pushki n, of the set of cel l s in a human body,
of the set of fsh in an ocean, of the set of natural numbers, of the set of points
i n the pl ane, of the set of spheres i n space, and so on.
The objects that make up a set are cal led its elements. The examples just
mentioned show that elements can be real objects (chairs, protons, fsh, and so
on) as wel l as abstract objects (numbers, points, geometric fgures, and so on).
Elements can also be such fgments of the human i magi nation as mythological
heroes, ghosts and the gods of vari ous rel i gi ons.
A set of real objects i s usual l y fnite -that i s, contai ns fnitely many elements.
Finite sets are usual l y given by l i sting thei r elements. For example, the set of
days in the week is given by the l i st { Monday, Tuesday, Wednesday, Thursday,
Fri day, Saturday, Sunday} . Of course, it is hardly possible to l i st the elements
of l arge sets such as fsh i n an ocean or grai ns of sand at the seashore.
Sets of abstract concepts or mythological heroes and gods can also be fni te.
Examples of such sets are the set of even primes, the set of Ol ympic gods, and
so on. The set of natural numbers i s i nfni te, and so is the set of points i n the
pl ane.
Braces that contain a l i st of elements of a set symbol i ze the uniting of these
elements i nto a singl e whole. If an element a belongs to a set A then we write
a E A. If an element a does not belong to a set A then we write a f A. For
example, if N is the set of natural numbers, then 6 E N, 3/4 f N and crocodi l e
f N. If A i s the set of months i n the year then May E A and Wednesday f A.
Thus when speaking of a set we uni te certain elements or concepts i n a
si ngl e whole and then operate with this whole. The German mathematician
Georg Cantor,
1
the creator of set theory, put it thus: "A set is a Many which
al l ows itself to be thought of by us as a One."
The academician N. N. Luzi n,
2
one of the founders of the Russian set
theoretic school , suggested the fol lowi ng intui tive representation of a set. Think
of a transparent, i mpenetrable shel l , something l i ke a transparent and tightly
cl osed bag. Suppose that the shel l contai ns al l the elements of some set A and
no others. Thi s shel l and the objects i n i t can serve as a representation of the
set A consisting of these elements. The transparent shel l that contai ns al l the
elements of the set and no others represents the act of uniting the elements that
gi ves rise to the set A.
Chapter 2. The mysteries of infnite sets
35
Sets and properties of objects. One cannot l i st the elements of an i nfnite set.
Nor i s i t practical to list the elements of a very l arge fnite set. To determi ne
a set of ei ther kind we speci fy a property shared by al l of its el ements and not
bel onging to any element not i n the set. Such a property of the el ements of a
set is called its characteristic property.
For example, a characteristic property of the set of prime (natural) numbers
i s that all its elements have exactly two divisors. Usi ng thi s property we can
i mmediately tel l that nei ther I , nor 1 8, nor 2/3 are primes. Indeed, I has just
one divi sor, 1 8 has si x ( l , 2, 3, 6, 9, 1 8), and 213 is not a natural number. On the
other hand, 7 i s a prime because it has exactly two divi sors, I and 7.
The philosophers of antiquity tried hard t o fnd characteri stic properties of
vari ous sets. For example, Pl ato is said to have given the fol l owi ng defni tion:
"A human is a l i ve featherless biped." The story has it that hi s contemporary
Diogenes pl ucked a rooster and said: "Here is Pl ato' s human." Thi s forced
Pl ato to add "with wi de toe nail s" to hi s defni ti on. This too was an obvi ously
specious characteri stic property of the concept of a human.
We write P (x) for "the element x has the property P," and {x I P (x) } for
the set of elements each of which has the property P. For example, A =
{x I x
2
- 3x
+
2 = O} consists of al l roots of the equation x
2
- 3x
+
2 = 0, and
B = {x i x E N and 0 < x < 3} consi sts of al l natural numbers that satisfy the
i nequal i ty 0 < x < 3. These sets both consist of the numbers I and 2, that is
they have the same elements. We cal l such sets equal and write A = B = { I , 2} .
Thi s example shows that while the concepts of set and characteri stic property
are closely rel ated, they are not the same - the same set can be given by means
of different characteri stic properties. Characteristic properties that describe the
same set are usual l y called equivalent.
Many mathematical theorems concern the equality of two sets, say, of the
set of natural numbers divisible by 3 and the set of natural numbers (written
i n deci mal notation) such that the sum of thei r digits i s divisibl e by 3, or, for
another example, of the set of equi l ateral triangl es and the set of equi angul ar
triangl es. Some problems of thi s kind are sti l l open. For example, we don' t
know whether the set of natural numbers n for which the equation x
n
+
y
n
= z
n
has solutions in N (= the set of natural numbers) is equal to the set { I , 2}
(Ferat 's Last Theorem), or, for another example, whether the set of primes p
for which 2P - 1 is divisible by p
2
is equal to the set { 1 093, 35 1 1 } .
Sets and the real world. Set-theoretic methods are now used in all areas of
learni ng. For example, li ngui sts study the set of letters or the set of (gram
matical) cases of a l anguage, ethnographers study the set of forms of fami l y
rel ati ons of members of a tribe, and physici sts study the set of molecules i n a
given volume of a gas.
Al l these sets are fnite. In many cases they can be given by l i sting thei r
el ements. For example, a teacher l i sts the students i n hi s cl ass i n the cl ass roster,
a l i brari an catalogues the set of books in her l ibrary, and a geographer l i sts the
set of states on the earth.
The progress of physics has demonstrated the growing difculty of describ
ing particles i n terms of set theory. Thi s i s due to the fact that many particles
transmute i nto one another. For example, a proton can transmute i nto a neutron
and conversely, so that here the term "consi sts of" loses its intuitive sense.
When maki ng up sets of objects of the real world i t i s frequently necessary
to identify various objects or concepts. For example, the maker of a dictionary
ignores the fact that a word is pronounced di fferently in different parts of a
country. To her, vari ations of pronunciation are i rrelevant. What counts is the
underl yi ng si ngle element of the set of words of the language i nvol ved. (The
student of di alects l ooks very diferently at the same set. What interests her
most is preci sel y the different pronunciations of a si ngle word. )
Thus i n talking about elements of a set (of real objects or of abstract noti ons)
we empl oy a kind of identi fcation operation and feel i ntuitively that i t won' t
lead t o contradictions. In other words, sets arise from vaguer concepts by
identifcation of various elements.
Other compl i cations i nvol ved in the use of set-theoretic concepts in the study
of the real world are due to the vagueness of many concepts, the i ndefniteness
of many properties of objects, and the difculty of spl itting reality i nto di stinct
objects. Some of these difculties are di scussed bel ow.
Of course, al l the indicated compl i cations and difculties are no reason for
not usi ng set-theoretic methods in describing real ity or in constructing scien
tifc theories. They do show, however, that the set-theoretic treatment of an
area of science i mposes seri ous l i mi tations on our approach to the i nvestigated
phenomena, and in many cases "crudifes" them.
On the positive si de, as noted by Yu. I. Mani n,3 the concepts of set theory
are very useful for constructi ng mathematical model s of phenomena in the real
world, because
they have laid a universal foundation for defni ng all mathematical
constructions usi ng . . . "general ized geometric i mages." These
mental constructs represent both a receptacle for the meani ng of
mathematical formal i sms and a means of extracting meaningful
assertions from the vast sea of derivable mathematical formul as.
37
They are therefore natural i ntermediaries between mathematics and physi cs. In
Mani n' s view, another advantage of set-theoretic l anguage i s that it forces one
to avoid all that i s superfuous.
Sets and language. We noted earl ier that prescribing sets of real objects by
means of thei r characteristic properties i nvol ves vari ous difculties. These
difculties stem from the large number of i ntermedi ate forms and from the
i mpreci sion of everyday language.
For example, consider the term the set of planets in the Solar system. We
cl ai m that this term i s not ful l y defned. For one thi ng, we do not know whether
or not there are pl anets beyond Pluto. For another, i n addition to the maj or
pl anets - Mercury, Venus, the Earth, Mars, Jupiter, Satur, Uranus, Neptune,
and Pl uto - some 1 600 small pl anets, the so-called asteroids, revol ve around
the Sun. The diameters of some of them, Ceres, Pal l as and Juno, say, measure
hundreds of kilometers. On the other hand, there are asteroids whose di ameters
do not exceed I km. Wi th i mproved methods of observation astronomers are
certai n to di scover celestial objects of ever smaller dimensions, and the question
*
wi l l be asked where pl anets end and meteorites and cosmic dust begi n.
Fuzzy sets. The American scientist L. Zadeh
4
found an ori gi nal way out of the
difculties just described. Zadeh i ntroduced the concept of afuz set and the
cl osel y rel ated concept of a linguistic variable. Just as with preci se properties
(being a pri me, bei ng a triangle) there are associ ated ordi nary or precise sets,
so too with fuzzy properties (being a young person, being a long street) there
are associ ated fuzzy sets (of young persons, of long streets) . For exampl e, a
distinguished academician wi l l l i kel y refer to a forty-year-old colleague as a
At this point, in the last paragraph of thi s section, the author notes that the di ference
between planets and meteorites i s of i nterest only to astronomers and i s not of great i mportance.
On the other hand, he says, there are compel l i ng reasons for precision in, say, the cl assifcation
of crimes. I agree, but left out this paragraph because it strikes me a irelevant.
I also left out the frst two paragraphs of this section because they deal with the term the set of
Russian words and its imprecision. It i s clear that i t makes l ittle sense to transl ate thi s material .
On the other hand, I am sure that English readers wi l l have no difculty in spotting some
elements of imprecision i n the term the set of English words. (Translator)
38
In Search of Infnit
young person whereas a frst-year student wi ll li kely regard his forty-year-ol d
professor as elderly. With every person (or, more precisely, with every "expert")
there is associ ated a preci se set of people he or she regards as young. But then
with every person x there is associ ated a number min, where n is the number of
all experts and m is the number of experts who regard x as young. We denote
min by p(x) and say that x belongs to the fuzzy set of young peopl e with
membership coefcient p (x) . Of course, the l atter takes on val ues from 0 to I .
Preci se sets di fer from fuzzy sets i n that for them p(x) can take on just
two values, namely 0 if x A and I if x E A. The exi stence of experts makes
i t possi ble to form a fuzzy set from the set of preci se sets associ ated with each
expert. Of course, notwithstandi ng the fuzziness of the set of young people, we
can say wi th certainty that for some x, p (x) = I (for example, no one wi l l deny
that a newborn is young) and for certain other x, p(x) = 0 (for example, no
one wi l l call an 80-year-old young) . No one? Wel l , there is the story that when
the 65-year-ol d chessmaster Tartakover defeated the 70-year-old chessmaster
Bernstei n he exclai med "youth wi ns ! "
Of course, i t makes l ittle sense t o cal l a commission of experts every ti me to
determi ne "membership coeffcients." For the most part such coefcients are
i ntroduced differently, on the basi s, say, of statistical data. But once they have
been selected, they can be used to obtai n membership coefcients for other sets.
The noti on of a fuzzy set led to the i ntroduction of fuzzy rel ations and fuzzy
algorithms. But fuzzy algorithms had been in use long before they were defned
by Zadeh. Every cookbook contai ns al gorithms with advice such as thi s: "Fi rst
whip the cream until it is very thick. Then add the four and whip the mixture
wel l . . . " True, authors of cookbooks don' t defne precisely when the whi pped
cream is very thick and how much whipping is appropriate. But there are good
reasons for thi nking that di shes prepared accordi ng to such reci pes usual l y turn
out al l right. It is i nteresti ng that fuzzy algorithms are now begi nni ng to be used
i n, for example, mathematical computing. Sti l l , onl y the future wi l l determi ne
the suitabi l ity and useful ness of Zadeh' s way of introducing fuzzy sets.
Infnite sets. What we said earl ier about sets referred for the most part to
fni te sets. For mi l lenni a the study of i nfnite sets was kept out of science on
Aristotl e' s authority. But in the 1 3th century the School man Robert Grosseteste,
a professor at Oxford (and the teacher of the famous Roger Bacon) decl ared that
the actual l y i nfnite is a defnite number, and while it is beyond our cognition
i t actual ly exists. Grosseteste thought that i nfni ties can be compared with one
another; that there are more moments i n a l arger ti me interval than i n a smaller
Chapter 2. The mysteries of infnite sets 39
one; that there are more points i n a l arger magnitude than i n a smal l er one; and
that the number "of points i n a segment one el l l ong" i s its true measure. Thus
Aristotle' s potential i nfni te was again opposed by the actual i nfni te of uni ts.
The gradual use of the actual i nfni te began i n the 1 8th century (i nfni te
series were actual l y viewed as sums of i nfnitely many terms), and in the 1 9th
century Gauss, supposedl y so very much against the use of the actual i nfnite
i n mathematics, used it i n his number-theoretic work. More expl icit uses of
the actual i nfni te are found i n the works of some of Gauss' successors such as
Di richlet
5
and Dedekind.
6
However, there was al most no systematic study of i nfnite sets. Paradoxes
of the infnite, by the Czech phi l osopher and mathematician Bernard Bolzano,
7
appeared posthumousl y in 1 85 1 . In thi s book, Bol zano made the frst attempt to
study the properties of the actual l y i nfnite. The book anticipated many concepts
of the theory of i nfnite sets, but these anticipations l acked the preci si on and
cl arity that Georg Cantor endowed them with i n his works written over a period
of twenty years.
In connection wi th hi s study of i nfnite series of trigonometric functions
Cantor faced the need to cl assi fy i n a certain sense sets made up of points of
a straight l i ne (now briefy referred to as point sets) . In particul ar, he wi shed
to know i f al l such sets can be l i sted. In the course of these i nvestigations
he real i zed that fni te and i nfni te sets have radical l y di ferent properties. In
particul ar, he noticed that operations that are impossible i n the case of fni te
sets can be easi l y real ized for i nfnite sets. For example, try to put up additional
guests i n a hotel al l of whose rooms al ready have a singl e guest, without any
doubl i ng up or evicti ons. Impossible? Only i f the hotel i s fni te ! And what
if it has i nfni tel y many rooms? . . Such hotel s are found in the stories of the
i ntergal actic wanderer Ion the Quiet, and it is only right that we should gi ve
hi m the foor.
The extraordinary hotel, or the thousand and frst journey of Ion the Quiet.
I got home rather l ate - the get-together at the club Andromeda Nebul a dragged
on long after midnight. I was tormented by ni ghtmares the whole night. I dreamt
that I had swal l owed an enormous Kurdl ; then I dreamt that I was again on the
pl anet Durditov and didn' t know how to escape one of those terrible machines
they have there that turn peopl e into hexagons; then . . . People general l y advi se
agai nst mi xi ng ol d age with seasoned mead. An unexpected telephone cal l
brought me back to real ity. It was my oId friend and companion i n i nterstel l ar
travel s Professor Tarantog.
"A pressi ng problem, my dear Ion," I heard. "Astronomers have di scovered
a strange object in the cosmos - a mysterious bl ack l i ne stretching from one
galaxy to another. No one knows what i s going on. Even the best telescopes
and radi o-telescopes placed on rockets cannot help in unravel ing the mystery.
You are our l ast hope. Fly right away in the direction of nebul a ACD- 1 587."
The next day I got my oId photon rocket back from the repair shop and
i nstal led i n i t my ti me accelerator and my electronic robot who knows al l
the languages of the cosmos and al l the stories about interstel l ar travel (it i s
guaranteed t o keep me entertai ned for at least a fve-year journey). Then I took
of to attend to the matter at hand.
Just as the robot exhausted hi s entire supply of stories and had begun to
repeat hi msel f (nothing i s worse than l istening to an electronic robot repeati ng
an old story for the tenth ti me), the goal of my journey appeared i n the di stance.
The galaxies which covered up the mysterious l ine l ay behind me, and in front
of me was . . . the hotel Cosmos. Some ti me ago I constructed a smal l pl anet
for wanderi ng i nterstel l ar exil es, but they tore it apart and again were without
a refuge. After that, they deci ded to give up wanderi ng i nto foreign galaxies
and to put up a grandiose bui l di ng - a hotel for al l travelers i n the cosmos. Thi s
hotel extended across almost al l the galaxies. I say "al most al l " because the
exi l es di smantled a few uni nhabited gal axies and made off with a few poorl y
situated constel l ations from each of the remai ni ng ones.
But they did a marvelous job of building the hotel . In each room there were
faucets from which hot and cold pl asma fowed. If you wi shed, you coul d be
spl i t i nto atoms for the ni ght, and in the morni ng the porter woul d put your
atoms back together agai n.
But, most i mportant of al l , there was an infnite number of rooms i n the
hotel . The exi les hoped that from now on no one woul d have to hear that
irksome phrase that had pl agued them during their ti me of wanderi ng: "no
room avai l abl e."
In spite of thi s I had no luck. The frst thing that caught my eye when I
entered the vestibul e of the hotel was a si gn: Delegates to the cosmic zoologists'
congress are to regi ster on the 1 27th foor.
Si nce cosmic zoologi sts came from all the galaxies and there are an i nfnite
number of these, i t turned out that al l the rooms were occupied by participants
i n the conference. There was no pl ace for me. True, the manager tried to get
some of the delegates to agree to double up so that I coul d share a room with one
of them. But when I found out that one proposed roommate breathed fuorine
and another considered i t normal to have the temperature of his envi ronment at
41
about 8600 , I pol itel y turned down such "pleasant" neighbors.
Lucki l y the di rector of the hotel had been an exile and well remembered the
good turn I had done him and his fel l ows. He would try to fnd me a pl ace at the
hotel . Afer al l , you could catch pneumoni a spending the night in i nterstel l ar
space. Afer some meditation, he turned to the manager and said:
"Put hi m i n number I ."
"Where am I going to put the guest in number I ?"
"Put hi m in number 2. Shif the guest in number 2 to number 3, number 3
to number 4, and so on."
It was onl y at this point that I began to appreci ate the unusual qual ities of
the hotel . If there had been only a fni te number of rooms, the guest in the l ast
room woul d have had to move out i nto i nterstel l ar space. But because the hotel
had i nfnitely many rooms, there was space for all, and I was able to move in
without depri vi ng any of the cosmic zool ogists of his room.
The fol l owi ng morning, I was not astonished to fnd that I was asked to
move i nto number 1 ,000,00. It was si mpl y that some cosmic zoologists had
arrived bel atedl y from gal axy VSK-3472, and they had to fnd room for another
999,999 guests. But while I was going to the manager to pay for my room
on the third day of my stay at the hotel, I was di smayed to see that from the
manager' s wi ndow there extended a l ine whose end di sappeared somewhere
near the cl ouds of Magel l an. Just then I heard a voice:
"I wi l l exchange two stamps from the Andromeda nebul a for a stamp from
Si ri us."
"Who has the Whale stamp from the 57th year of the cosmic era?"
I turned i n bewi l derment to the manager and asked:
"Who are these people?"
"Thi s i s the i nterstel l ar congress of phi l atel ists."
"Are there many of them?"
"An i nfnite set - one representative from each galaxy."
"But how wi l l you fnd room for them? Afer all , the cosmic zool ogists
don' t l eave ti l l tomorrow."
"I don' t know; I am on my way now to speak to the di rector about it for a
few minutes."
However, thi s ti me the problem turned out to be much more difcult and
the few mi nutes extended i nto an hour. Finally, the manager left the ofce of
the di rector and proceeded to make his arrangements. First he asked the guest
i n number I to move to number 2. This seemed strange to me, si nce I knew
from my own experience that such a shif woul d only free one room, whereas
he had to fnd pl aces for nothi ng less than an i nfnite set of phi l atel i sts. But the
manager continued to give orders:
"Put the guest from number 2 into number 4, the one from number 3 into
number 6; i n general , put the guest from number n i nto number 2n."
Now hi s pl an became clear: by thi s scheme he woul d free the i nfni te set of
odd-numbered rooms and woul d be able to settl e the phi l atel i sts in them. So
i n the end the even numbers turned out to be occupied by cosmic zoologists
and the odd numbers by philatel ists. (I didn' t say anything about myself - after
three days of acquaintance I became so friendly wi th the cosmic zoologists
that I had been chosen an honorary representative to thei r congress; so I had
to abandon my own room along with all the cosmic zoologi sts and move from
number 1 ,00,000 to number 2, 000,0() . And a phi l atel ist friend of mine who
was 574th i n l i ne got room number 1 1 47. In general , the phi l atel ist who was
nth i n line got room number 2n - 1 .
The fol l owi ng day the room si tuation eased up - the cosmic zool ogists'
congress ended and they took off for home. I moved in with the di rector, i n
whose apartment there was a vacant room. But what i s good for the guests does
not always please the management. Afer a few days my generous host became
sad.
"What ' s the trouble?" I asked hi m.
"Hal f the rooms are empty. We won' t ful fl l the fnancial plan."
Actual l y, I was not quite sure what fnancial plan he was talking about; after
al l , he was getting the fee for an i nfnite number of rooms, but I nevertheless
gave him some advice:
"Wel l , why don' t you move the guests cl oser together; move them around
so as to fl l al l the rooms."
Chapter 2. The mysteries of innite sets
43
Thi s turned out to be easy to do. The philatel i sts occupied onl y the odd
rooms: 1 , 3, 5, 7, 9, etc. They left the guest i n number I alone. They moved
number 3 into number 2, number 5 into number 3, number 7 into number 4,
etc. At the end all the rooms were once agai n flled and not even one new guest
had arrived.
But thi s did not end the director' s unhappiness. It was explained that the
exiles did not content themselves with the erection of the hotel Cosmos. The
i ndefatigable bui l ders then went on to construct an i nfnite set of hotel s, each of
which had i nfni tel y many rooms. To do this they di smantled so many galaxi es
that the i ntergal actic equi l ibri um was upset and thi s could entai l seri ous con
sequences. They were therefore asked to cl ose al l the hotels except ours and
put the material used back i nto pl ace. But i t was di ffcult to carry out thi s order
when all the hotels (ours i ncluded) were flled. He was asked to move al l the
guests from i nfnitel y many hotels, each of which had i nfni tel y many guests,
into one hotel , and thi s one was al ready fl led!
"I' ve had enough! " the di rector shouted. "First I put up one guest in an
al ready ful l hotel , then another 999,999, then even an i nfni te set of guests; and
now they want me to fnd room i n it for an additional i nfni te set of i nfni te sets
of guests. No, the hotel i sn' t made of rubber; let them put them where they
want."
But an order was an order, and they had fve days to get ready for the
arrival of the new guests. Nobody worked in the hotel during these fve days -
everybody was pondering how to solve the problem. A contest was announced
- the prize woul d be a tour of one of the galaxies. But all the solutions proposed
were turned down as unsuccessful . Then a cook i n trai ni ng made the foll owi ng
proposal : leave the guest i n number 1 i n hi s present quarters, move number 2
into number 1 00 I , number 3 i nto number 200 1 , etc. After thi s, put the guest
from the second hotel into numbers 2, 1 002, 2002, etc. of our hotel, the guests
from the thi rd hotel into numbers 3, 1 003, 2003, etc. The project was turned
down, for i t was not clear where the guests of the 1 001 st hotel were to be pl aced;
after al l , the guests from the frst 1 000 hotels would occupy all the rooms. We
recalled on this occasion that when the servi l e Roman senate ofered to rename
the month of September "Tiberi us" to honor the emperor (the preceding months
had al ready been given the names of Juli us and Augustus), Tiberius asked them
caustical l y "and what wi l l you offer the thirteenth Caesar?"
The hotel ' s bookeeper proposed a pretty good vari ant. He advi sed us to
44
In Search of Infnit
make use of the properties of the geometric progression and resettl e the guests as
follows: the guests from the frst hotel are to be put i n rooms 2, 4, 8, 1 6, 32, etc.
(these numbers form a geometric progression with multiplier 2). The guests
from the second hotel are to be put i n rooms 3, 9, 27, 8 1 , etc. (these are the terms
of the geometric progressi on wi th mul tiplier 3). He proposed that we resettle
the guests from the other hotel s i n a si mi l ar manner. But the director asked hi m:
"And we are to use the progression wi th multipl ier 4 for the third hotel ?"
"Of course," the bookeeper repl ied.
"Then nothi ng i s accompl i shed; afer al l , we al ready have someone from
the frst hotel in room 4, so where are we going to put the people from the third
hotel ?"
My turn to speak came; i t was not for nothi ng that they made you study
mathematics for fve years at the Stel l ar Academy.
"Use pri me numbers. Put the guests from the frst hotel into numbers 2, 4,
8, 1 6, . . . , from the second hotel i nto numbers 3, 9, 27, 8 1 , . . . , from the third
into numbers 5, 25, 1 25, 625, . . . , the fourth i nto numbers 7, 49, 343, . . . "
"And it won' t happen again that some room wi l l have two guests?" the
di rector asked.
"No. Afer al l , if you take two prime numbers, none of their positive integer
powers can equal one another. If p and q are prime numbers, p = q, and m
and n are natural numbers, then pm = q
n
."
The di rector agreed with me and i mmedi atel y found an i mprovement on
the method I had proposed, in which onl y the pri mes 2 and 3 were needed.
Namely, he proposed to put the guest from the mth room of the nth hotel into
room number 2m3
n
This works because i f m = p or n = q, 2m3
n
= 2P 3Q SO
no room woul d have two occupants.
This proposal del ighted everyone. It was a sol ution of the problem that
everyone had supposed i nsoluble. But neither the di rector nor I got the prize; too
many rooms woul d be lef unoccupied if our solutions were adopted (according
to mi ne - such rooms as 6, 1 0, 1 2, and, more general ly, all rooms whose numbers
were not powers of primes, and according to the director' s - all rooms whose
numbers coul d not be written in the form 2
n
3m) . The best solution was proposed
by one of the phi l atel i sts, the president of the Academy of Mathematics of the
gal axy Swan.
He proposed that we construct a tabul ation, i n whose rows the number of
the hotel would appear, and i n whose columns the room numbers woul d appear.
For example, at the intersection of the 4th row and the 6th col umn there woul d
appear the 6th room of the 4th hotel . Here i s the tabulation
(
actually, , onl y i ts
upper l ef part, for to write down the entire tabul ation we would have to employ
infnitely many rows and col umns):
( 1 , 1 ) ( 1 , 2) ( 1 , 3) ( 1 , 4) ( 1 , 5)
( I , n)
(2, 1 ) (2, 2) (2, 3) (2,4) (2, 5) (2, n)
(3, I ) (3, 2) (3, 3) (3,4) (3, 5) (3, n)
(4, 1 ) (4, 2) (4, 3) (4,4) (4, 5) (4, n)
(5, I ) (5, 2) (5, 3) (5,4) (5, 5) (5, n)
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
(m, 1 ) (m, 2) (m, 3) (m,4) (m, 5) . . . (m, n) . . .
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
"And now settle the guests according to squares," the mathematician-phi l atel ist
said.
"How?" The di rector did not understand.
"By squares. In number 1 put the guest from ( I , I ), i . e. , from the frst room
of the frst hotel ; i n number 2 put the guest from
(
1 , 2), i . e. , from the second
room of the frst hotel ; in number 3 put the guest from (2, 2), the second room
of the second hotel , and i n number 4 - the guest from (2, 1 ), the frst room of the
second hotel . We will thus have settled the guests from the upper lef square of
side 2. After thi s, put the guest from
( 1 , 3) i n number 5, from (2, 3) i n number 6,
from
(
3, 3) i n number 7, from (3, 2) i n number 8, from (3, 1 ) i n number 9.
(
These
rooms fl l out the square of si de 3. ) And we carry on in thi s way:
46
In Search of Infnit
( 1 , 1 )
( 1 , 2)
( 1 , 3)
( 1 ,4)
( 1 , 5)
( 1 , n)
, , , , ,
(2, 1 ) +
(2, 2) (2, 3)
(2,4)
(2, 5)
(2, n)
, , , ,
( 3, 1 ) +
(3, 2) +
(3, 3) (3,4)
(3, 5)
(3, n)
, , ,
(4, 1 ) +
(4, 2) +
(4, 3) +
(4,4) (4, 5)
(4, n)
, ,
(5, 1 ) +
(5, 2) +
(5, 3) +
(5,4) +
(5, 5)
(5, n)
,
(
n, 1 ) +
(
n, 2) +
(
n, 3) +
(n,4) + (n, 5) + . . . (n, n) . . .
"Wi l l there real l y be enough room for al l ?" The director was doubtful .
"Of course. After al l , according t o thi s scheme we settle the guests from
the frst n rooms of the frst n hotel s in the frst n
2
rooms. So sooner or l ater
every guest wi l l get a room. For example, i f we are talking about the guest from
number 1 36 in hotel number 2 1 7, he wi l l get a room at the 2 1 7th stage. We can
even easi l y fgure out which room. It wi l l have the number 2 1 7
2
- 1 36
+
1 .
More generally, i f the guest occupies room n i n the mth hotel , then if n :: m he
wi l l occupy number (n - 1 )
2
+
m, and i f n < m, number m
2
- n
+
I ."
The proposed project was recognized to be the best - al l the guests from al l
hotel s woul d fnd a pl ace in our hotel , and not even one room woul d be empty.
The mathemati ci an-phi l atel i st received the prize - a tour of galaxy LCR-287.
In honor of thi s so successful solution, the di rector organized a reception
to which he invited al l the guests. The reception, too, had its problems. The
occupants of the even-numbered rooms arrived a hal f hour l ate, and when they
appeared, it turned out that al l the chairs were occupied, even though our kind
host had arranged to have a chair for each guest. They had to wait while
everyone shifted to new pl aces so as to free the necessary quantity of seats
(of course, not one new chair was brought i nto the hal l ) . Later on, when they
began to serve the guests ice cream, i t was di scovered that each guest had two
porti ons, although, as a matter of fact, the cook had only prepared one portion
per guest. I hope that by now the reader can fgure out by himself how this
happened.
47
At the end of the reception I got into my photon rocket and took off for
Earth. I had to i nform the cosmonauts of Earth about the new haven existing i n
the cosmos. Besides, I wanted to consult some promi nent mathematicians and
my friend Professor Tarantog about the properties of i nfni te sets.
From the author. With this we temporarily take l eave of our hero. Many of
hi s stories give rise to doubt - after al l , according to the laws of the theory
of rel ativity it is impossible to transmit signal s at speeds greater than 1 86,000
mi les/sec. Thus, even the very frst order of the di rector woul d requi re an
i nfni tel y large interval of ti me to carry out. But let us not ask too much of Ion
the Quiet - he has had even more i mprobable adventures during his travel s.
The rest of the book i s devoted to the story of the theory of sets. And
although the events wi l l no l onger take pl ace i n i nterstel l ar space but on the
interval [0, 1 ] or the square of si de I , many of them wi l l seem no less unusual.
How to compare sets in terms of size. At the begi nni ng of this chapter we
were concerned wi th properties of sets whi ch hol d general l y for both fni te
and i nfnite sets. Here we shal l be i nterested in properties characteri stic of
i nfnite sets alone. We have already seen in the story of Ion the Quiet that these
properties are quite different from those of fnite sets - things i mpossible for
fni te sets turn out to be possible for i nfnite sets.
The frst question whi ch we shal l now discuss i s the problem of deci di ng
when two i nfnite sets are of equal si ze. For fni te sets of the most varied types
we can always say which of them contai ns the l arger number of elements. This
problem i s much more complicated for i nfni te sets. For example, which i s the
l arger set, that of the natural numbers or that of the rational numbers? That of
the rational numbers or that of the real numbers? Are there more points on the
entire l i ne than there are on a segment? More points i n a square than on a l i ne?
It appears quite si mple at frst gl ance to answer these questi ons. After al l ,
the set of natural numbers i s only a part of the set of rational numbers and the
segment is onl y a part of the l i ne. Isn' t it obvious, therefore, that there are fewer
natural numbers than there are rational numbers and that there are fewer points
on a segment than on the l i ne? It turns out not to be so obvi ous. It does not
fol low at all that when we go from fni te to i nfnite sets, the l aws derived from
the study of the former, for example, a l aw such as "the part is less than the
whole," remain val i d.
Above al l , an attempt to compare i nfnite sets usi ng the cri terion that one
i s a part of the other i s doomed to fai l ure i n advance. For example, where are
there more poi nts, in a square or on the whole of an i nfnite li ne? After al l , the
square cannot b contai ned in a l i ne and, without breaki ng it, it is i mpossible
to put a line i nside a square. Of course, it i s possible to break the line up into
segments of length equal to the si de of the square, and after that place each
segment i nsi de the square i n such a way that no two i ntersect. But how do
we know that we can fnd a way to brea up the square so that the parts can
be strung out along the li ne without overlapping? And how many i nfnite sets
there are which are not parts of one another! The set of squares in the plane
and the set of ci rcles i n the same plane do not have even a si ngle element in
common. How can we compare them? How can we fnd out i f there are more
atoms of ni trogen or of oxygen in the universe?
We have now posed the problem. First we i nvestigate under what conditions
it can be said that one set contai ns j ust as many elements as another. In other
words, we study the conditions under which two i nfni te sets have "the same
Peasure" of elements.
On the dance foor. The problem of compari son is easily sol ved for fni te
sets. In order to fnd out if the number of elements is the same for two sets,
we have only to count them. If we get the same numbers, this means that both
sets have the same size. But such a procedure is not suitable for i nfni te sets;
for, having begun to count the elements of an i nfnite set, we run the risk of
devoting our entire l ives to this job and sti lI not completing the enterprise we
have undertaken.
And the method of counting i s not always convenient even for fni te sets.
For i nstance, let us go to a dance hal l . How can we tell if the numbers of boys
and girls here are the same? Of course, we could ask the boys to go off to one
si de and the girls to the other, and undertake to count both groups. But, i n the
frst place, thi s would give us superfuous i nformation; we are not interested
in how many boys and girls are here, but only in whether the numbers are the
same. And then, the young people on the dance foor did not get together to
stand around and wait for the end of the count, but to dance.
Wel l , what then? Let us sati sfy thei r wish and ask the orchestra to play
a dance that everybody knows how to do. Then the boys wi l l ask the gi rls to
dance and . . . our problem wi l l b solved. After all , if it turs out that all the
boys and girls are danci ng, i . e. , if all the young people are paired off, then it i s
obvi ous that there are just as many boys as gi rl s on the dance foor.
We coul d fnd out by an identical procedure whether the number of specta
tors in a theater i s equal to the number of seats. If during the performance al l
49
the pl aces are taken, no spectator is standing in the aisles and one spectator i s
sitting i n each seat, then we can be sure that there are just as many spectators
as seats.
For every fow there is an ebb. We have seen how it i s possible to determine
that two fnite sets have equal l y many elements without havi ng recourse to
counting. We can also appl y thi s method to i nfnite sets. But there we can
no longer get an orchestra to do the job; we ourselves have to di stribute the
elements of the two sets to be compared into "couples."
Suppose we are given two sets A and B. We shall say that we have estab
l i shed a one-to-one correspondence between them, i f the elements of these sets
have been joi ned i n pairs (a , b) such that:
(
1 ) element a belongs to set A, and element b belongs to set B;
(
2) every element of the two sets occurs i n one and onl y one pair.
For i nstance, if set A consi sts of the boys on the dance foor and set B
consi sts of the gi rl s found there, then pair (a , b) is composed of the boy and gi rl
dancing together. If set A consi sts of the spectators and set B consists of the
seats i n the theater, then pair (a, b) i s composed of the spectator and the seat i n
which he sits. The reader can easily think of other examples of correspondences
between sets of the same size.
Natural ly, not every correspondence between sets i s one-to-one. If set A
consi sts of all the trees in the world and set B consi sts of al l the fruit growi ng
on these trees, then we can set up the fol lowi ng correspondence between these
sets: to each fruit we make correspond the tree on which it grows. But thi s
is not a one-to-one correspondence: on some trees many pieces of fruit grow,
while other trees do not even bear frui t. Thus, some elements a (trees) wi l l
appear i n many pairs, whi l e other elements a wi l l not appear i n any.
For two fnite sets, it means the same thi ng to say that there is a one-to-one
correspondence between them, or to say that they have equal l y many el ements.
The fundamental turni ng point in the theory of sets came when Cantor deci ded
to compare i nfnite sets i n the same manner.
In other words, Cantor said that two (possibly i nfni te) sets A and B have
equal l y many elements, if it is possible to set up a one-to-one correspondence
between them.
Mathematicians do not usual l y say: "sets A and B have equal l y many
elements"; they say: "A and B have the same cardinalit" or they say: "sets A
and B are equivalent."
Because of this, the word cardinalit means the same thing for i nfnite sets
as the words "number of elements" do for fnite sets.
The Czech savant B. Bol zano arrived at the notion of one-to-one corre
spondence i ndependently of Cantor; but he gave up further pursui t of the idea
because of the diffculties into which it l ed hi m. As we shall soon see, we shal l
have to set aside many cheri shed habits of thought once we accept the principle
of compari ng i nfni te sets with the aid of the one-to-one correspondence.
Can a part be equal to the whole? One dogma that we have to brush aside i s
the statement, establi shed at the begi nni ng of the devel opment of mathematics:
a part i s less than the whole. Thi s statement i s indi sputably true for fni te sets,
but it loses its force when we try to appl y i t to i nfnite sets. Let us recall how
the di rector of the extraordinary hotel shifted the cosmic zoologists to even
numbered rooms. He moved the i nhabitant of room n to room 2n. In other
words, he moved them according to the fol lowing scheme:
| 2 3 n
' ' ' '
2 4 6 2n
But this scheme sets up a one-to-one correspondence between the set of
natural numbers
1 , 2, 3, . . . , n, . . .
and a part of this set: the set of even numbers
2, 4, 6, . . . , 2n, . . .
But we agreed to assume that two sets contai n equal l y many elements i f it
i s possible to set up a one-to-one correspondence between them. Thi s means
that the set of natural numbers contai ns as many and only as many elements as
one of its subsets, the set of even numbers.
In exactly the same way we could set up a one-to-one correspondence
between the set of natural numbers and the set of numbers of the form
1 0, 1 00, 1 000, 1 000, . . .
To do thi s we need only associ ate the natural number n with the number
1 0" :
Chapter 2. The mysteries of infnite sets 5 1
This establ ishes the desi red one-to-one correspondence. In the same way we
can set up a one-to-one correspondence between the set of al l natural numbers
and the set of all squares of natural numbers:
the set of al l cubes of natural numbers:
and so on.
General l y speaking, we can set up a one-to-one correspondence between
the set of all natural numbers and any of its i nfni te subsets. To do thi s we need
only write down the numbers of this subset i n a sequence.
Inci dental l y, there are good reasons why they say that there i s nothi ng new
under the sun and the new i s the quite forgotten old. Al ready at the begi nni ng
of the 1 7th century Gal i leo pondered the paradoxes of the i nfnite and noticed
that one can set up a one-to-one correspondence between the natural numbers
and thei r squares. His book Discourses and mathematical proofs, concering
two new sciences pertaining to mechanics and local motions
(
1 638) contai ns a
dial ogue in which Sal vi ati , expressi ng the ideas of Gal i l eo, says the fol l owi ng:
What we sai d pertai ns to a number of difculties that ari se because,
when we use our l i mited powers of reasoning to discuss the i nfni te,
we ascribe to it properties we know from things fni te and bounded.
But thi s is wrong, for properties such as a greater and a lesser
magnitude and equal ity are not appl icable to the i nfnite, of which
we cannot say that one i nfnity i s greater or smaller than another,
or that it is equal to it.
To prove hi s i dea, Sal vi ati notes that, on the one hand,
there are as many squares as there are roots, for every square has
its root and every root its square; no square can have more than
*
one root and no root more than one square. . . Al so, there are
as many roots as there are numbers, for there is no number that
cannot be a root of some square; this bei ng so, we must admi t that
there are as many squares as there are numbers . . .
We are tal king of just the natural numbers.

But on the other hand,
also,
the quantity of all numbers - squares and nonsquares - i s greater
than that of squares alone,
as we go to l arge numbers, the number of squares decreases con
tinual ly and very quickly.
Accordi ng to Salviati, there i s just one way of avoiding this contradiction:
I see no other solution than to admit that there are i nfnitel y many
numbers, i nfnitely many squares, and i nfnitel y many roots. One
cannot say that there are fewer squares than numbers or more num
bers than squares: i n the fnal analysi s, the properties of equality,
and of bei ng a l arger or a smaller magnitude, are appl icable onl y
to fni te quantities and not when deal i ng wi th the i nfnite.
Thi s shows that, i n essence, Gal ileo was aware of the notion of a one-to
one correspondence and real i zed that such a correspondence can be establ i shed
between all the natural numbers and the squares, and that these sets can therefore
be said to have the same number of elements. He also real ized that in the case of
an i nfnite set a part can be equal to the whole. But then he jumped to the fal se
concl usi on that al 1 i nfnities are the same. Thi s is true of the i nfnite subsets of
the set of natural numbers, the elements of each of which can be enumerated.
GaJileo could not i magi ne that the set of points of an interval cannot be
enumerated (we wi l l prove thi s below). Like the ancient atomists, he assumed
that an i nterval consists of i nfnitel y many atoms that can be enumerated.
Countble sets. We cal 1 sets wi th as many elements as the set of natural
numbers countable sets. In other words, a set i s caIJed countable if i t i s i nfni te
and its el ements can be counted with the aid of the natural numbers. For
example, the set of even numbers, the set of odd numbers, the set of primes,
and, i n general , any i nfnite subset of the natural numbers are countable sets.
We someti mes have to empl oy considerable i ngenui ty i n order to show that
thi s or that set is countable. Let us take as our example the set of al1 i ntegers
(both positive and negative) :
. . . , -n, . . . ,
-
3,
-
2,
-
1 , O, 1 , 2, 3, . . . , n, . . .
I f we try t o number them begi nni ng at some given place, we fnd that the
numberi ng i s i ncomplete; for all the numbers occuring before the given place
have not been counted. In order not to leave out any numbers we have to write
the set in two l i nes:
0, 1 , 2, 3, 4, 5, 6, . . .
- 1 , -2, -3, -4, -5, -6, -7, . . .
and number by columns. Here 0 is assigned the number 1 , - 1 is assi gned the
number 2, 1 the number 3, -2 the number 4, etc. In other words, zero and
all the positive integers are numbered with odd numbers, while all the negative
integers are numbered with even numbers. This resembles the way the hotel
director placed the phi l atel i sts i n a hotel already flled with cosmic zoologists.
But if it i s easy to show that the set of i ntegers i s countable, i t i s more
difcult to show that the same is true of the rational numbers. After all, the
rationals are densely distributed: between any two rational numbers we can
still fnd i nfnitely many rational numbers. So it is quite unclear how we should
go about numbering them; it would seem that between any two numbers we
would still have to number an i nfnite set, so that the process would never end.
And it really i s i mpossible to write down the rational s i n a sequence i n which
each number i s greater than its predecessor.
But if we do not concern oursel ves about the magnitude of the numbers in our
sequence, we can succeed i n numbering them. Let us frst write down all positive
fractions with denominator 1 , then all positive fractions wi th denomi nator 2,
then wi th denominator 3, and so on. We get a tabulation li ke the followi ng:
1 2 3 4 5
T' T
'
T
'
T' T' . . .
1 2 3 4 5
'
'
'
'
'
'
'' '
'
. . .
1 2 3 4 5
j
'
j
'
j
'
j
'
j
'
. . .
1 2 3 4 5
4>
4
' 4' 4' 4
' . . .
1 2 3 4 5
)
'
)
'
)
'
)
'
)
'
. . .
Cl earl y, every positive rational number wi l l appear i n this table, and more
than once. For example, the number 3 occurs in the form of the fractions 3/ 1 ,
6/2, 9/3, and so on.
Now we commence the numberi ng. For thi s we recal l the l ast exploit of
the di rector of the extraordi nary hotel , the one i n whi ch he found pl aces for the
guests of i nfnitely many such hotel s. In doing this he numbered by squares.
We shal l proceed i n the same manner, but with this compl icati on: we shall
l eave out some of the fractions (for example, si nce I I I i s assigned number I ,
we drop the fractions 2/2, 3/3, etc. for they express the same number). We get
the fol lowi ng enumeration of the positive rational s: I , 2, 1 12, 3, 3/2, 2/3, 1 /3,
4, 4/3, 3/4, 1 14, . . . .
Thus we can number al l the positive rati onal s. It i s now easy to explain
how all the rational numbers (both positive and negative) can be numbered. We
separate them i nto two tables, usi ng even numbers to number one table and odd
numbers for the other (rememberi ng to reserve a number for zero).
In general , if we take the union of a countable set of countable sets, we
again get a countable set. We could prove this by usi ng this same techni que of
numberi ng by squares.
Algebraic numbers. We managed to enumerate al l rational numbers. The
rational numbers are obtai ned from the integers by usi ng the operation of divi
sion. Now we consider the set of numbers obtained from the i ntegers by usi ng
the al gebraic operations and the extraction of roots. This set incl udes numbers
such as
;
+ 1 , - , and "monsters" such as
11 47 +
-
- 6 +

V 289 - i 4 +

+ 1
We ask: Can one enumerate the numbers in thi s set? Thi s seems more difcult
than enumerating the rational s. But it turns out that thi s set of numbers is
countable, that i s, its elements can be enumerated.
To prove this we note that every number of our set i s a root of an equation
of the form
( 1 )
where aQ , . . . , a
n
are i ntegers. For example, 317 i s a root of 7x - 3 = 0, , i s
a root of x
3
- 5 = 0, and }2 + W is a root of x
6
- 6x
4
+ 1 2x
2
- 1 1 = 0. It i s
55
sometimes difcult, but always possible, t o write down an equation of which a
particul ar one of the numbers we consider is a root.
We note that not al l roots of equati ons of the form ( 1 ) wi th a
o
, . . . , a
n
integers can be obtained from the integers by means of the algebraic operations
and extraction of roots. For example, the roots of
x
5
- 3x + 3 = 0
are not so obtainable, or, as we say, this equation is not sol vable in radical s. The
numbers that are roots of equati ons of the form ( 1 ) wi th integer coefcients are
cal l ed algebraic numbers. Thus the set of algebraic numbers contai ns the set
of numbers obtai ned from the i ntegers by means of al gebraic operations and
extraction of roots. If the former set can be shown to be countable, then so can
the l atter.
Note that if we manage to enumerate al l al gebraic equations of the form
(
1 ), then our job is al most done. Recal l that every algebraic equation of degree
n has at most n roots. If we manage to enumerate al l equati ons wi th integer
coefcients, then we can set up a table whose frst row consi sts of the di stinct
roots of the frst equation, the second row of the di sti nct roots of the second
equation not i n the frst row, the third row of the di stinct roots of the thi rd
equation not in the frst two rows, and so on. The resul t is a table of the form
It is obvious how to enumerate the elements of this table.
Now we set about enumerating the elements of the set of al gebraic equations
with integer coefcients. Here we i mitate the scheme used by the di rector of the
hotel in sol vi ng his most difcult problem. The one diference is that whereas
the di rector used numbers of the form 2m3
n
, that i s powers of two primes, we
wi l l use powers of all pri mes. The reader shoul d bear i n mind that every natural
number is uniquely representable as the product of pri mes, or, equivalently, as
the product of di sti nct prime powers.
We proceed as fol lows. First we enumerate the integers (see p. 53) . Let a be
the number of the integer a. Then with the equation a
o
x
n
+a
l
x
n
-
I
+
. .
+a
n
= 0
(where, we recal l , a
o
, . . . , a
n
are i ntegers) we associ ate the number
56
2
1
3
a.- 1 pan
. . .
n+
1
In Search of Infnit
(
P
n+
1
is the (n + l )th pri me) . For example, wi th the equation 3x
2
- 2 = 0
we associ ate the number 2
4
3
1
57 = 37500; thi s i s so because when enumer
ati ng the i ntegers we assi gned to -2,0, 3 the numbers 4, 1 , 7 respecti vel y. Now
each equation has a number and different equations have different numbers.
(Every number N is a uni que product of pri me powers. The uni que expo
nents a
n
, an -
I
, . . . , ao of these prime powers determine uni quel y the i ntegers
a
n
, a
n
-
I , . . . , a
o
, and thus the equation a
o
x
n
+
. . .
+ a
n
= 0. )
Sets of unequal size. We have al ready explai ned what we mean when we say:
"two sets have equal l y many elements." Now we are going t o explain what we
mean when we say: "one set has more elements than another." For fni te sets
thi s too can be found out wi thout resorting to counting. Recal l our example
i nvol vi ng the dance foor. If, afer the orchestra starts playing and the boys have
invited the gi rl s to dance, there are some boys leaning agai nst the wal l , then it i s
clear (i f al l the girls are dancing) that there are more boys. On the other hand,
if we see some girls sadly watching thei r friends dancing, it i s clear (i f al l the
boys are danci ng) that there are more gi rl s.
In these examples we proceeded as fol lows: we tried t o establ ish a one
to-one correspondence between one set, the frst, and part of another set, the
second. If thi s worked out, then the second set had more elements than the frst.
By employi ng thi s method we could prove, for example, that there are fewer
fsh in the ocean than atoms on the Earth ( although both these sets are fni te, it
i s hardl y possible to count them). We can do this by si mpl y letting each fsh
correspond to one of the atoms constituting its body. This sets up a one-to-one
correspondence between the set of all fsh and part of the set of all atoms on
Earth.
Unfortunatel y, this si mple procedure fai l s to hold good for i nfnite sets.
Indeed, we recently saw that a set can have as many elements as one of its parts.
So we are i n no position to concl ude from the sol e fact that A has as many
el ements as a part of set B, that set A has fewer elements than B.
We shal l be more modest i n our demands and say that i f we can set up a
one-to-one correspondence between set A and part of set B, then set B has no
fewer elements than set A. We could prove that this rel ation possesses al l the
fundamental properties of inequal ities:
( I ) Each set A has no fewer elements than itself.
57
(2) I f set A has no fewer elements than set B, and B has no fewer elements
than set C, then A has no fewer elements than C.
(3) I f A has no fewer elements than B, and B has no fewer elements than A,
then they have equal l y many elements (that i s, we can set up a one-to-one
correspondence between the elements of these sets) .
The frst property fol l ows from the fact that by associ ating with an element
of A that very element we obtain i s a one-to-one mapping of A onto itself. The
second property i s also clear: i f A can be mapped i n a one-to-one manner onto
a part of B, and B can be mapped i n a one-to-one manner onto a part of C, then
A can be mapped i n a one-to-one manner onto a part of C.
I n spite of its apparent si mpl icity, the thi rd assertion i s fai rl y di fcul t to
prove. Cantor thought that it was true but for many years was unable to fnd
a proof. In 1 897 he lectured at the uni versity of Hal le on set theory and men
ti oned this di ffcul ty. A few days l ater, one of the students, 1 9-year-old Fel i x
Bernstei n,
8
brought Cantor a proof based on the very same idea the di rector of
the cosmic hotel used to accommodate new guests. Hence thi s assertion i s now
cal led the Cantor-Bernstei n theorem. Many years l ater a proof of thi s theorem
dating back to 1 887 was found i n the Nachlass of the German mathematician
Dedekind.
We will now expl ain what we mean when we say that the cardi nal i ty of a
set B is greater than the cardinal ity of a set A.
It can happen that set B has n o fewer elements than set A, but these sets are
not equivalent. In other words, there could exist a one-to-one correspondence
between set A and part B
I
of set B without there existing a one-to-one corre
spondence between A and al l of set B. Thi s i s the case i n which we shal l say
that B has more elements than A.
The countable set - the smallest of the infnite sets. We al ready said that any
i nfni te subset of the set of natural numbers is countabl e. Let us now prove that
any i nfnite set contains a countable subset. We can concl ude from thi s that the
cardinal ity of a countable set is not greater than the cardi nal ity of any i nfni te
set, i . e. , that this cardi nal i ty is the smal lest i nfnite cardinal ity.
We can select a countable subset from an i nfnite set A i n the fol lowi ng
way: Take any element X
I
- we can do thi s because the set A i s i nfnite, and so
i s certai nl y not empty. Clearl y, we have not exhausted the el ements of A wi th
the selection of element X
I
. so that we can proceed to select a second el ement
X
2
. After that, we choose a third element X
3
, etc. We have thus extracted from
set A a countable subset X of indexed elements:
x = {Xl , X
2
, , Xn , } .
By maki ng a sl i ght change i n the argument we can arrange matters so that
an i nfni te set wi l l be left even after the extraction of the countable subset. All
we have to do i s put back i nto A all those elements from X that have even
indices. Afer doing thi s, we have extracted a countable subset
and the remai ni ng part of the set sti l l contains an i nfnite subset of el ements:
{X
2
, X
4
, X6 , , X
2
n , } (and possibly other elements) .
It is not difcult to prove the following theorems.
The cardinalit of an infnite set is not changed when we adjoin a countable
set to it.
The cardinalit of an uncountable set is not changed when we extract a
countable subset from it.
These theorems again assert that countable sets are the smallest i nfnite sets.
Uncountable sets. Al l the sets we have constructed so far have been countable.
Thi s natural l y leads us to ask whether all i nfnite sets are countable. If so, the
mathematician woul d have an easy l i fe: all i nfnite sets would have equally
many elements and no further anal ysi s of i nfnity woul d be necessary. But the
si tuation turns out to be more compl icated than that; uncountable sets exist, and
of more than one cardi nal ity. We are already acquai nted with one uncountable
set - the set of all points on a straight l i ne. But rather than speak of thi s set, we
are going to di scuss a set closely rel ated to it, the set A of ways in which the
rooms of the extraordinary hotel can b occupied.
Note that it i s usual l y not easy to prove that a set i s uncountable. Afer al l ,
t o prove that a set i s countable means si mpl y t o i nvent a method of enumerating
its elements. But to prove that a set i s uncountable we have to prove that no
such method exists. In other words, no matter what method we appl ied, some
element of the set would fai l to be counted. Cantor conceived of a very clever
method for provi ng the uncountability of sets which is called the di agonal pro
cess. Cantor' s method of proof i s made clear by the fol lowing story about Ion
the Quiet.
The census that never took place. Up to now I have talked about the successes
of the di rector of the extraordinary hotel : about how he managed to fnd places
for an i nfni te set of new guests i n hi s already full hotel, and how he l ater was
able to fnd pl aces even for the guests from i nfnitely many such unusual hotel s.
But there was a ti me when even thi s wizard met fai lure.
An order came down from the commi ssi oner of cosmic hotel s to compile
a l i st as quickly as possible of aU the possible ways i n which the rooms of the
hotel could be occupied. The l i st was to be presented in the form of a tabl e,
each l ine of whi ch was to refect one of the various ways of occupying the hotel .
The fl l ed rooms were to be i ndicated by ones and the empty rooms by zeros.
For example, the sequence
1 01 01 01 01 01 0 . . .
meant that al l the odd rooms were fl led and all the even rooms were empty.
The sequence
1 1 1 1 1 1 1 1 1 1 1 . . .
meant that the entire hotel was fl led, whil e the sequence
00. . .
indicated a fnancial catastrophe - al l the rooms were empty.
The di rector was overloaded with work and therefore conceived of a si mple
way out of the si tuati on. He charged the man on duty on each foor to compose
a l ist of the ways i n which just the rooms i n his charge could be occupied.
No two ways on the l i st were to be the same. Afer a few days the l i sts were
presented to the di rector and he combi ned them al l i nto one l i st.
"Are you sure that this l i st i s complete?" I asked the di rector. "Isn' t there
some other way of occupying the rooms?"
"I don' t know" he repl ied. "There are i nfnitel y many ways l i sted and I
don' t know how to test the l i st for completeness."
At this point an idea fashed into my head (by the way, I may be overes
ti mati ng my talents, because not all traces of my di scussions wi th Professor
Tarantog on i nfnite sets had vani shed from my mi nd) .
"I can guarantee that the l i st is i ncomplete. I can select a way that is sure
to be l acking."
"I agree that the l i st i s probably i ncomplete. But you won' t succeed in se
lecti ng a way that i sn' t l i sted; after al l , there are already i nfnitely many l i sted."
We made a bet. I proposed to wi n it by nai l i ng each sequence on the door
of the room to which it corresponded (the reader wi l l recall that there were just
as many ways l i sted as rooms in the hotel ). I then proceeded i n a very si mple
fashi on. Going up to the door of the frst room, I saw that the corresponding
sequence started with the digit O. The digit I quickly appeared on my writing
pad; this was the frst digit of the sequence I wanted to construct.
When I went up to the door of the second room, I wasn' t interested i n the
frst digit of the sequence; after al l , I al ready had the frst digit of my sequence.
So I di rected my attention to the second digit. Seeing that this was I , I wrote the
di gi t 0 on my pad. Si mi l arl y, when I noticed that the third digit of the sequence
nailed to the third room was also 1 , I agai n wrote the digit 0 on my pad. In
general , when I found that the nth digit of the nth sequence was 0, then I wrote
the digit I in the nth pl ace on my pad, but if the nth digit of the nth sequence
was 1 , then I wrote O.
After I had gone past al l the rooms of the hote
(
a sequence of zeros and
ones had been written on my pad.
Going to the di rector' s ofce, I said:
"Here, feast your eyes on the mi ssi ng sequence."
"And how do you know that i t' s l acki ng?"
"It can' t be the frst because it has a di ferent frst digit. It can' t be the
second because it has a di fferent second digit; in general , it can' t be the nth
because it has a diferent nth digit."
The bet was won, and I gained the privi l ege of stayi ng at the hotel whenever
I wanted at no charge.
But it at once became clear that no matter what countable set of sequences
you took, there woul d always be a sequence that didn' t appear i n the l i st (you
woul d always be able to hang them on the doors of the rooms). This means that
the set of al l ways of occupying the hotel is uncountable, and the task given the
di rector was not one that could be carried out.
We deci ded to send a telegram descri bing the situation. I shoul d point out
that the telegraph i n use at the extraordi nary hotel was itself unusual : it could
send telegrams composed of an i nfnite set (more preci sel y, a countable set) of
' Hm. how much time did he have to spend?
Chapter 2. The mysteries o/infnite sets
6 1
dots and dashes. For example, the telegram might have the form
-. - -. - - - . etc.
I quickly grasped the fact that the set of al l such telegrams was al so uncount
abl e; after al l , you could just as wel l put zeros and ones i n place of the dots and
dashes, and then there woul d be no di ference at all between the telegrams with
a countable set of signs and the set of all ways of occupyi ng the hotel .
Afer sending the telegram, I took leave of the di rector of the hotel and took
of for gal axy RGC-8067, where I was to carry out an astrographical survey.
The uncountability of the continuum. Now it wi l l not be di fcul t to prove
that the set of all points on a l i ne is uncountable. In pl ace of i t we can di scuss
the set of al l real numbers, si nce to each poi nt on the l i ne there corresponds a
real number, and conversely.
Any real number can be given an i nfnite decimal expansion of the form
Some even have two expansions, for example: 0. 500 . . . and 0. 49999 . . .
represent the same number. To si mpl i fy matters we shal l employ the expansi on
with the zeros.
Suppose that by some scheme we had managed to enumerate al l the real
numbers. In order to show that this can' t happen we need only show that some
number has not been enumerated. Fol lowi ng i n Ion the Quiet' s footsteps, we
proceed i n the fol lowing manner.
We frst write zero fol l owed by a deci mal point. We then take the frst
number i n our enumeration and examine its frst pl ace afer the decimal point
(i . e. , its tenths pl ace) . If it differs from 1 , then we write a 1 after the deci mal
point in the number we are constructi ng; but if it is I , we put a 2 afer the
deci mal point. Afer that we choose the second number in our enumeration
and examine its second pl ace afer the deci mal point. Again, i f thi s number is
different from 1 , we put the number 1 i n the hundredths pl ace i n our number;
and i f it i s 1 , then we use 2. We carry on i n thi s way, each ti me looki ng at the
nth place i n the nth number of our enumeration. As a result of these operations
we get some number, for example:
N = 0. 1 1 2 1 2 1 1 . . .
It is clear that this number is not one of those enumerated: it di fers from
the frst number in the frst deci mal pl ace; it differs from the second number
i n the second deci mal place; it differs from the nth number i n the nth deci mal
pl ace, etc.
In order to make it clearer to the reader how we determine our number
different from all those enumerated, suppose that in the given enumeration the
frst fve numbers have the fol lowi ng form:
4. 27364 . . .
-1 . 3 1 225 . . .
7. 9547 1 . . .
0. 6241 9 . . .
8. 56280 . . .
Then the number not in the enumeration wi l l begi n with the fol l owi ng
deci mal s:
0. 1 2 1 2 1 . . .
Natural l y, thi s i s not the onl y number that i s not on the l i st (we could have
repl aced all the deci mals except 2 by 2 and replaced 2 by 7, or chosen some
other rule) . But we only needed to establ i sh the exi stence of a si ngle number
which does not appear in the enumeration in order to demonstrate that the sup
posed enumeration of al l the real numbers coul d not exist.
The existence of transcendental numbers. Numbers are called transcendental
numbers if they are not the roots of any equation with integer coefcients of
the form
As we saw, the numbers which are roots of such equations are cal led algebraic
numbers.
During a long period i n its hi story, mathematics only dealt with algebraic
numbers, such as 7/ 1 5, : . + :, etc. It was only at the cost of a great
effort that, in 1 844, the French mathematician Liouvi l l e
9
was able to fnd a few
transcendental numbers. But the proof that the number r i s transcendental ,
carried out by Li ndemann 1 0 in 1 882, was a great mathematical event; i ndeed,
it fol l owed that it is i mpossible to square the ci rcl e. And suddenly it became
cl ear that the al gebraic numbers met with at every step in mathematics are
real l y extremely rare, whil e the transcendental numbers, so hard to construct,
are real l y the common ones. Afer al l , we have already seen that the al gebraic
63
numbers onl y form a countable set; while the set of all real numbers, as we just
demonstrated, is uncountable. Thi s means that the real numbers which are not
al gebraic, i . e. , the transcendental numbers, must also be uncountable.
This proof of the exi stence of transcendental numbers, obtai ned by Can
tor i n 1 873, differed from Liouvi l l e' s proof i n that i t used only the notions of
countabi l ity and uncountabi l i ty of sets and not speci al properties of al gebraic
numbers. Liouvi l l e' s theorem i mpl ies that the number 0. 1 0 1 ( 1 00000 1 . . . , i n
whose deci mal expansion the n-th 1 i s followed by n ! zeros, i s transcendental .
To obtai n an example of a transcendental number through Cantor' s proof one
must fol low a far longer road: frst enumerate the algebraic numbers, then write
them in deci mal form, then use the di agonal procedure to obtai n a transcenden
tal number. It woul d take a "practical i nfni ty" to fnd, say, the 1 Ol-th di gi t i n
the deci mal expansion of thi s number. On the other hand, Liouvi l l e' s method
enables us to form transcendental numbers for which one can, admittedly wi th
difcul ty, answer such questions. Thus the downsi de of the general method of
proof is its rel ative uselessness in deal ing with concrete questions.
Long and short line segments have equally many points. I f the reader had
been unacquainted with the remarkable properties of i nfni te sets, the question
"Are there more points on a line segment I foot long or on a line segment 1 mile
long?" woul d hardl y have raised a shadow of doubt i n hi s mind. The answer
woul d be clear; there are many more points on the segment of length 1 mile,
for i sn' t i t 5280 ti mes longer? But by now the reader has probably learned to
beware of maki ng categorical statements - the properties of i nfnite sets are too
di ssi mi lar to what he has been taught to expect by dai l y l i fe.
D
Figure 8.
And the long and short segments do i n fact have equal l y many points ! In
other words, i t i s possible to set up a one-to-one correspondence between the
points of these segments. Figure 8 shows the easiest way of doing thi s.
It i s hard t o reconci l e oneself t o the thought that a path a million l i ght years
long has only as many points as the radi us of an atomic nucleus !
But even more unexpected i s the result that there are not even more points
on the entire i nfnite l i ne than on a segment, i . e. , a one-to-one correspondence
64
In Search of Infnit
can b set up between the set of points on the l i ne and the set of points on the
segment.
We do not even need the whole segment, but can di scard its endpoi nts (i . e. ,
we use the open i nterval). It i s clear from Figure 9 how to set up a one-to-one
correspondence between the i nterval AB and the li ne. Clearly, each point on
the i nterval corresponds to exactly one point on the l i ne and every point on the
line has a partner on the i nterval .
A
8
Figure 9
However, this correspondence can b set up in another way with the help of
a curve - the tangent curve, the graph of the function y = tan x (Figure 1 0).
Y
!= t on X
Figure 1 0
..
2
Segment and square. Mathematicians reluctantly reconciled themsel ves to
the fact that there are as many points on a segment as on an i nfnite l i ne. But the
fol l owi ng resul t of Cantor turned out to be even more unexpected. Searching
for a set which woul d have more points than a segment, he turned to the set
of points of a square. He had no doubt of the resul t - afer al l, the segment
occupied onl y one si de of the square, whereas the set of all segments which
composed the square had the cardi nality of the conti nuum.
Cantor searched for three years (from 1 87 1 to 1 874) for a proof that it was
i mpossible to set up a one-to-one correspondence between the poi nts of the
segment and the points of the square.
The years went by, but the desi red resul t coul d not be obtai ned. And then
the completel y unexpected happened. He succeeded i n setting up the corre
spondence he bel ieved i mpossible ! He wrote to the mathematici an Dedeki nd:
"I see it, but I don' t bel ieve it."
But we have to resign oursel ves to the fact that our i ntuition lets us down
again here - i t turns out that there are exactly as many points i n the square as on
the segment. A rigorous proof of the statement i s made somewhat complicated
by the l ack of uniqueness of the deci mal expansion of numbers. We shal l
therefore present only a sketch of Cantor' s proof.
Let us take the segment [0, 1 ] and the square of si de I . We may suppose
that the square i s situated as i n Figure I I . We have to set up a one-to-one
correspondence between the points of the segment and the points of the square.
Projection of the points of the square onto the segment A B wi l l not help here;
i ndeed, under projection an i nfnite set of points of the square are sent into one
poi nt of the segment (for example, all the points of segment DA go i nto point
A) .
We can solve the problem as fol l ows: We can speci fy any point T of the
square ABC D by means of two numbers, its coordi nates x and y (or more
simply its di stances along the sides AB and AD) . These numbers can be
written as i nfnite deci mal s. Si nce x and y are not
J
D t-&
A
t
T {..gJ
y
B
Figure I I
more than I , these deci mals have the form
y = 0
f
d
3
2
f
n
. . .
(2. 2)
(2. 3)
(for the sake of si mplicity we do not take points l yi ng on the sides of the
square, but onl y take i nterior poi nts) . For example, if x = 0. 63205 . . . and y =
66
In Search of Infnit
0. 2 1 357 . . . , then al = 6, a
2
= 3, a
3
= 2, etc. , and fl = 2, f
2
= 1 , f
3
= 3,
etc.
Now we have to pick the point Q of the segment AB which is to correspond
to T. It is enough to say what the length of the segment A Q is. We choose thi s
length t o be equal t o the number z
,
whose deci mal expansion i s obtained by
i nterlaci ng the deci mal expansions of x and y. In other words, we form a thi rd
expansi on from the two expansions (2) and (3) by combi ni ng thei r deci mals:
For i nstance, i f
x = 0. 5 1 5623 . . .
and
y = 0. 734856 . . .
then we obtain
z = 0. 57 1 354682536 . . .
The point z lies on the segment [0, 1 ] , and i t i s clear that diferent points of
the square correspond to di fferent points of the segment. Indeed, i f points T
and T
'
are not the same, then the deci mal expansion of x and x' or y and y
'
must di ffer at least in one place. But thi s wi ll lead to a diference in the deci mal
expansi ons of the numbers z and z
'
. A somewhat more detailed anal ysi s shows
that, conversel y, the points corresponding to two di fferent deci mal expansions
also do not coincide.
Thus we have set up a one-to-one correspondence between the points of the
square and the points of a part of the segment [0, 1 ] . Thi s shows that the set of
points of the square has a cardi nality no l arger than that of the set of points of
the segment. But its cardi nali ty i s certainly no smaller, so that the cardi nal ities
must coincide.
Not j ust the square, but the cube as well has only as many points as the
segment. In general , any geometric fgure contai ni ng at least one l i ne segment
wi l l have just as many points as the segment. Such sets are called sets with the
cardi nali ty of the continuum (from the Lati n continuum - unbroken).
Is there a set of largest cardinality? Ti ll now the l argest cardinality we have
become acquainted with is that of the set of points on the l ine, i . e. , the cardinality
of the conti nuum. Neither the set of points of the square nor the set of points
of the cube has a l arger cardi nal ity. Perhaps the cardinality of the conti nuum is
the l argest possible? Thi s turns out not to be the case. Indeed, there i s no set
of l argest cardi nality. Given any set A, there i s a set B of cardi nality greater
than the cardinal ity of A. We can construct i t by frst associ ati ng with each
poi nt a of a set A the function assuming the val ue 1 at this point and the value 0
at the remaining poi nts. Clearl y, di stinct points give rise to di sti nct functions .
For example, i f set A consists of the points 1 , 2, 3, then point 1 corresponds to
a function which assumes the val ue 1 at thi s point, while point 2 corresponds
to a function assuming the val ue 0 at point 1 . These functions are distinct. We
take our set B to be the set of all functions on A with values 0 and 1 .
Thus, the cardinal ity of set B i s not less than the cardi nal i ty of set A.
Let us now show that these cardinal ities are not equal , i . e. , no one-to-one
correspondence can be found between the elements of sets A and B. Indeed,
suppose such a correspondence exi sted.
We designate the function corresponding to element a of A by fa (x) . Re
member that al l the functions fa (x) assume onl y the two values 0 and 1 .
Let us defne a new function ((x) by means of the equation:
(
(X
)
= 1
-
f
x
(x)
,
( 2. 3)
Thus, i n order t o determine the val ue of the function ((x) at some point a of A
we frst must fnd the function fa (x) corresponding to this point and subtract its
val ue at x = a from 1 . It i s now clear that the function ((x) i s defned on the
set A and assumes only the val ues 0 and 1 . Consequently, ((x) i s an element
of set B. But then by our assumption ((x) corresponds to some point b of A;
thi s means that
(
(X
)
= f
b
(x) . (2. 4)
It fol l ows from (3) and (4) that for al l x i n A
Let us set x = b in this equation. Then we get
so that
But thi s contradicts the requi rement that the values of the function f
b
(x)
be 0 and I . The contradiction we have obtained shows that there can be no
one-to-one corespondence between sets A and B
.
Thus, given any set A, we can construct a set B of l arger cardinal ity. There
fore, no set of l argest cardi nality can exist. Beginning wi th the smallest i nfnite
cardi nality - that of the set of natural numbers, we obtain frst the cardinality
of the conti nuum, then the cardinal i ty of the set of functions on the set of real
numbers, and continue to ascend the dizzying staircase of ever i ncreasing i nf
nite cardi nalities.
The arithmetic of the infnite. The arithmetic of natural numbers i s not just
counti ng "one, two, three." We can add and multiply natural numbers and
raise a natural number to a power. There i s a close connection between these
operations and operations on fni te sets. When adding natural numbers m and n
we are computing the number of elements in the uni on of two disjoint sets with
m and n elements respectively. When multipl yi ng m by n we are computing
the number of pairs (a , b) , where the frst element belongs to a set A with m
elements, and the second to a set B wi th n elements. In mathematics thi s set
of pairs is called the Cartesian prduct of the sets A and B and is denoted by
A x B.
Denote the union of di sj oint sets A and B by A
+
B and the cardi nality of
A by
I
A
I
. Then what we just said can be written as fol lows:
I
A
+
B
I
=
I
A
I + I
B
I
I
A x B
I
=
I
A
I I
B
I
.
The left-hand sides of these equalities make sense for i nfni te sets. Thi s
allows us to defne addition and mUl tiplication of i nfnite cardi nal ities. Denote
by
N
the set of natural numbers and by
/
the set of points of the i nterval [0, I J .
Then our earl ier resul ts on cardinal ities can be wri tten as fol lows:
n
+ I NI
-
I NI , I NI + I NI
=
I NI , I NI I NI
=
I NI , I NI + 1 /1
=
1 /1 , I NI I /I
=
1 /1 ,
1 /1 1 /1
=
1 /1 ,
and so on. For example, the equal i ty
I NI I NI
=
I NI
states that the union of a
countable set of countable sets is countable, and the equality
1 /1 1 /1
=
I / I
states that a square has as many poi nts as a segment.
It is not difcul t to prove that for fnite sets A and B the number of al l
mappi ngs with domain A and range in B is
I
B
I
I AI . It is therefore reasonable to
69
defne I BI I A I , for sets A and B with arbitrar cardi nal ities I AI and I BI , as the
cardi nal ity of the set of mappings with domain A and range in B. For example,
21 NI = I I states that the cardi nality of the set of sequences of zeros and ones
i s the same as the cardi nality of the conti nuum.
We can now understand what Cantor had i n mi nd when he said that the laws
of arithmetic of the i nfnite differ radical l y from those of the real m of the fni te.
Tansfnite numbers. Natural numbers are used not onl y to answer the question
of "how many?" but also, i n the case of an ordered set, the question "i n which
pl ace i s it?" Thus the natural numbers are used as quantitative numbers as well
as ordinal numbers. Cardi nalities can be used onl y as quantitative numbers.
To describe order we must use other concepts. After al l , the si mplest i nfni te
set, namely the set
N
of natural numbers, can be ordered in i nfnitely many
ways. In addition to the usual orderi ng 1 , 2, 3,4, 5, 6, . . . we can start with the odd
numbers (i n their usual order) and then set down the even numbers (also i n their
usual order) : 1 , 3, 5, . . . , 2, 4, 6, . . . . If we try to enumerate the natural numbers
in this ordering then we get into trouble - all numbers wi l l have been used to
enumerate the odd numbers and i t wi l l be necessary to i ntroduce addi ti onal
symbol s. To deal with thi s case, Cantor suggested the label w for 2, w + I for
4, and so on.
Even more symbols are needed i f we take frst all numbers di vi sible by 3,
then those which leave the remainder I when divided by 3, and then those that
leave the remai nder 2 upon division by 3: 3, 6, 9, . . . , I ,4, 7, . . . , 2, 5, 8, . . . . Here 2
is l abel led w
.
2, 5 is l abel led w
.
2 + 1 , and so on. And if we order the natural
numbers by writing frst the pri mes, then products of two primes, then products
of three pri mes, and so on, and at the very end the number 1 (which is neither
prime nor composite), i t will have to be l abelled with the new symbol w
tJ
.
Cantor thought up many other orderi ngs of the natural numbers. They
al l share (with the orderings just introduced) the property that every subset
of the natural numbers has i n each of these orderings a least element. Can
tor cal led (possibly uncountable) ordered sets with the correspondi ng property
well-ordered, and the symbol s he introduced for thei r i ndexi ng transfnite num
bers or transfnite ordinals (trans i s Latin for beyond). In the study of transfni te
numbers Cantor encountered the problem of the cardinal ity of the set of count
able transfnite numbers. It i s easy to show that this cardi nal ity i s uncountable
and not greater than the cardinal ity of the conti nuum. But neither Cantor, nor
hi s many students and fol l owers coul d say whether i t i s equal to or less than the
cardi nal ity of the conti nuum. We will di scuss the present state of thi s co-called
continuum prblem i n Chapter 4.
At the begi nni ng of the 20th century the theory of i nfnite sets became a
fashionable area of mathematics. Some experts attached very great importance
to i nvestigations in this area. For example, A. Fraenkel wrote:
The conquest of the actual i nfnite by means of set theory can be
regarded as an extension of our scienti fc horizon whose i mpor
tance is comparable to the i mportance of the Copernican system
i n astronomy and the theory of rel ativity and quantum mechanics
i n physi cs.
But i n the end, ti me, the harshest of al l judges, puts everything i n i ts proper
perspective. The number of papers in which the authors used transfnite numbers
or i nvesti gated cardinal ities beyond the countable and beyond the cardinal ity of
the conti nuum decreased steadi l y. Sets with such cardi nal i ties can be obtained
by considering all subsets of the pl ane or al l functions on [0, 1 J. But the fact i s
that vi rtual ly al l theoretical i nvestigations and the solution of practical problems
i nvol ve not arbitrary subsets of the plane and arbitrary functions but those
obtai ned by certai n defnite processes from very simple ones. And sets of such
"good" subsets and functions have the cardinal ity of the conti nuum.
In spi te of the vi ew of P. S. Al eksandrov
l l
and A. N. Kol mogorov
l
2
that "the
tremendous i nfuence of set theory on the evolution of mathematics in the l ast
half-century i s at present a universal l y conceded fact," this i nfuence moves at
present i n very di fferent channel s. In the next chapter we wi l l describe the
changes i n certain areas of mathematics due to set-theoretic concepti ons.
Chapter 3
Hcmarkablc luncl0ns and curVcs, 0r a slr0ll
lhr0ugh a malhcmalcal hall 0l w0ndcrs
General foundation. The whole history of the evolution of mathematics i s
marked by di alectical opposition and unity of i t s two parts devoted t o the study
of numbers and fgures, respectively. The natural numbers di fer from one
another by thei r properties: some are even and some odd, some are prime and
some composite, some can be written as sums of two squares and some can
not. Thi s i nfnite variety of properties that change so stri ki ngl y upon addition
of just one uni t to a number endows number theory with a speci al charm. Of
course, geometric fgures are just as varied - tri angles and squares, ci rcles and
parabolas, astroids and cardioids. But each curve by itself, a straight l i ne or a
ci rcle, is composed of points with identical properties.
The idea of the i nfni te also takes di fferent forms in these two branches of
mathematics. In arithmetic it is embodied in the sequence of natural numbers,
while in geometry it appears in the i nfnity of space and, at the same time,
i n the i nfnite di vi si bi l ity of fgures. And yet, notwithstanding this seemi ngl y
unbri dgeable chasm - due perhaps to some profound characteri stics of human
reason - throughout the history of mathematics there have been endl ess attempts
to link arithmetic and geometry and to derive all of mathematics from a si ngl e
foundation.
When mathematics was not so much a science as a trade pl ied by Egypti an
and Babylonian scribes, the unity of arithmetic and geometry manifested itself
i n naive form- the many problems dealt wi th included computations of areas
of fgures and of vol umes of sol ids. The frst attempt to effect a theoretical
uni fcation of arithmetic and geometry was undertaken i n the si xth century Be
in the school of the ancient Greek mathematician and phi l osopher Pythagoras.
One of Pythagoras' pronouncements that has come down to us i s "Al l is number."
He not onl y tried to "veri fy harmony by means of al gebra" by creating one of
the frst mathematical theories of the musical scale, but also attempted to base
the science of measurement of geometric magnitudes on the natural numbers.
That is why the di scovery of the i ncommensurabi l ity of the side and di agonal of
a square by a Pythagorean spel led a catastrophe for the Pythagorean world view.
(For a long ti me thi s di scovery was withheld from the uni ni ti ated. )
When it became cl ear that geometry could not be based on the concept
of natural number, the ancient Greek mathematicians adopted the opposite
approach and began to express rel ations between arbitrary magnitudes in geo
metric terms. Whereas the di screte lent itself more readi l y to l ogical anal ysi s,
the conti nuous could be more readi l y grasped by i ntuition. Rel yi ng on geomet
ric l anguage, Greek schol ars stated al gebraic l aws (this was when terms such as
square and cube of a number, geometric mean, and geometric progression frst
entered mathematics), i nvestigated quadratic i rrationalities, and sol ved cubic
equati ons. Geometry itself they based on the idea of the i nfni te di vi si bi l ity of
l i nes, (pl ane) fgures, and sol ids. They created the concepts of a di mensionless
point, of a l i ne that had only length, and of a geometric surface without thick
ness. And in spite of the fact that these were just bold abstractions from real
points, l i nes and surfaces, they served thei r inventors wel l in thei r i nvestigations
and enabled them to obtain correct formul as for areas and vol umes.
Afer the fal l of antique ci vi l i zation, the center of mathematical investi
gations shifted to Arabic-speaki ng countries. Arabic schol ars were fami l iar
not only with the ancient Greek heritage but also with the tradition, derived
from Babyl oni an scribes, that incl uded general methods of solution of arith
metical probl ems. They were also i nfuenced by the di scoveries of Indian
mathematicians, who constructed the deci mal number system and, unl ike the
ancient Greek scholars, made free use of negative numbers. All thi s prepared
the ground for the creation of al gebra, which frst came into being i n the 9th
century as the art of sol ving equati ons. At that ti me, mathematicians, many of
whom l i ved in Central Asia, paid l i ttle attention to subtleties connected with
incommensurable segments and freely used numbers i n the study of geometric
problems.
In Western Europe. the study of al gebra began a few centuries l ater. Ital y
was the frst center of such studies. Western European mathematicians obtai ned
formul as for the sol ution of cubic and quartic equati ons, began the study of
complex numbers, and devel oped l i teral symbol i sm. At frst, in accordance
with ancient Greek traditions, l i teral calcul us was geometric in form. and thi s
Chapter 3. Remarkable functions and curves, or a strll through
a mathematical hall of wonders
blocked the use of expressions invol vi ng di fferent powers.
73
Mathematics was faced with the need to construct al gebra without rel i ance
on geometric concepts and to free it from an essential l y extri nsic geometric
termi nology. A deci sive step i n this direction was taken by Descartes.
I
At frst
si ght there seemed to be nothing special i n hi s approach. He suggested that a
segment e be fxed and called a unit segment. But thi s meant that a product
of two segments a and b could be regarded not as the area of a rectangle
with si des a and b but as the length of a segment c such that a : e = c : b
(we note that the theory of proportions had been worked out in great detail
by the ancient Greek schol ars). The square root of a was i nterpreted as the
geometric mean of the segments a and e. After Descartes, mathematicians
could freely use arbi trary algebraic expressions without worryi ng about thei r
geometric sense. Descartes' works i ni ti ated a period of "arithmetizati on" of
mathematics that l asted more than two hundred years. Thi s process repl aced
its geometric foundation with an arithmetic one.
A high point of thi s devel opment was the nongeometric construction of the
real numbers. Thi s was done i n the I 870s by Cantor, Weierstrass,
2
Dedeki nd
and Mcray. 3 The starting poi nt of thei r defnitions was the concept of a natural
number. After Descartes, mathematicians coul d describe geometric objects
by means of real numbers (for example, a point by its coordi nates). Thi s
made possible the "arithmetizati on" of geometry. The resul t was a new unity
of mathematics based on an arithmetical foundati on. At that time, schol ars
bel ieved that they had reduced the continuous to the di screte.
It was no accident that one of the creators of the ari thmetical construction
of the real numbers was Cantor, the creator of set theory, another hi s teacher
and i nspirer Weierstrass, and the third Dedekind, who in hi s works came very
cl ose to the i deas of set theory. Whi l e thei r constructions were di fferent, each
of them defned a real number in terms of an i nfnite set of rational numbers.
For example, in Weierstrass' theory, a real number was defned as an i nfni te
deci mal . To gi ve such an i nfnite deci mal one had to speci fy ten subsets of
the natural numbers, each of which speci fed the l ocati ons of a particular di gi t.
In Dedekind' s theory, a real number e was defned by si mply subdi vi di ng the
rationals into two subsets, one of which contai ned the numbers not exceedi ng
e, and the other the numbers greater than e.
At the same ti me Gottlob Frege
4
tried to base the arithmetic of natural
numbers on the set concept . Thi s made the theory of i nfnite sets the common
foundation of arithmetic and geometry, of the discrete and the conti nuous. It
seemed that frst one must master i nfni te sets, then set aside i n thei r theory
74
In Search of Infnit
a tiny corner where fnite sets woul d humbly huddle together, and onl y then
obtai n the natural numbers. In Hi l bert ' s
5
words,
owi ng to the gigantic si multaneous efforts of Frege, Dedekind and
Cantor, the i nfni te was set on a throne and revel led i n its total
triumph. In its dari ng fi ght the i nfnite reached dizzyi ng heights
of success.
But not al l schol ars accepted without demur the new view of mathematics.
At frst, Cantor's di scoveries were received with di sbel ief and even outright
antagoni sm by many mathematicians and with indifference by an overwhel mi ng
majority of phi l osophers. The l eader of the opposition agai nst the new view
was Kronecker,
6
one of the most eminent mathematicians of that ti me. Hi s key
tenet was that mathematics can deal only with what is bui l t up out of natural
numbers i n fnitely many steps. His famous motto was that "God created the
natural numbers and al l the rest is the work of man." He therefore rejected not
onl y the theory of i nfnite sets, general l y regarded as extravagant, but also the
newfangled theory of real numbers.
The unenthusi astic reception accorded Cantor' s work by many mathemati
cians can in part be expl ained by the fact that the very idea of regarding the
i nfni te as somethi ng completed contradicted the establ i shed views. Conse
quently, many schol ars viewed research in the area of i nfni te sets as somethi ng
far from the vital tasks of science. In part, this unhappy state of affairs was due
to Cantor' s style - a medley of mathematical investigations with phil osophical
and theol ogical digressi ons. Transfnite numbers met with a guarded recep
tion because they could not be used to eval uate a complicated integral , sum
a difcult series, or solve a di fferential equati on. And Cantor' s successes i n
the real m of concrete mathematics (for example, hi s proof of the exi stence of
transcendental numbers) seemed rather uni mpressive.
But there was one area of mathematics in whi ch set theory was somethi ng
l i ke dai l y bread. This was the then ul tramodern theory of functions of a real
vari able. To understand why Cantor' s ideas were particul arly useful in thi s
area of mathematics, we must recall the course of evol ution of the concept of a
functi on.
How the notion of a function developed. The majority of mathematical
concepts underwent a long period of devel opment. They frst arose as gener
al izati ons of intuitive ideas derived from everyday experience. With gradual
el i mi nation of detai l s and accidental aspects, these intuitive ideas sl owl y crys
tal l i zed into exact mathemati cal defnitions. But it often happened that these
Chapter 3. Remarkable functions and curves, or a stroll through
a mathematical hall of wonders 75
defnitions appl ied not only to those objects whose study led to thei r formul a
tion, but also to other objects that had not been thought of earl ier. The study
of these new objects was begun and the process of abstraction was carried to
ever higher level s; next came the extension of the ori gi nal defni ti ons on the
basi s provided by the studies. Ever broader meani ng came to be attributed
to mathematical concepts; they embraced wi der and wi der cl asses of objects,
occurri ng i n more varied felds of mathematics.
The noti on of a function also fol l owed a tortuous path. The idea of the
interdependence of two quantities apparently arose in cl assical Greek science,
where i t was used in geometry. At the begi nni ng of the 1 7th century Gali leo,
Kepler and other scientists began to develop ki nematics, the study of the motion
of bodies. Under the i nfuence of thei r works, Descartes introduced i nto math
ematics the general notion of a vari able magnitude. Here i s Engel s' eval uation
of thi s development:
Descartes' notion of a variable magnitude was a watershed i n math
emati cs. Through it motion, and thus dialectic, entered mathemat
ics . . . and made dif erential and integral calculus necessar. In
fact, the calcul us came i nto bei ng at that ti me, and, by and l arge,
it was completed rather than invented by Newton and Lei bni z.
Seventeenth-century scienti sts used variable magnitudes to describe the
most varied moti ons. As a resul t, the geometric l anguage of mathematics
was gradual l y repl aced by the l anguage of mechanics. For example, when he
i ntroduced the logarithmic function i nto science, Napier
7
rel ied on recti l inear
motions of points, and Newton felt that mathematical anal ysi s must study the
ti me dependence of vari ous magnitudes - such as displacements, vel ocities and
accelerations - that characterize moti ons. Incidental ly, for some ti me geometric
l anguage was retained. For example, instead of trigonometric functions of a
numeri cal argument, one studied the lengths of certain segments in a disk as
functi ons of an angle. In fact, Leibniz frst i ntroduced the very notion of a
function as a connection between certain segments characterizi ng points on a
curve (absci ssas and ordinates, absci ssas and subtangents, and so on).
But i n 1 7 1 8, J. Bernoul l i
8
gave the fol lowi ng defnition of a function free
of geometric i magery:
A function of a vari able quantity is a magnitude formed in some
manner from this vari able quantity and constants
and Euler
9
defned a function as fol l ows:
76 In Searh of Infnit
Quantities dependent on others such that as the second change, so
do the frst, are said to be functi ons.
Al so, Euler managed to free trigonometry of geometric l anguage and to
introduce trigonometric functions as functions of numerical vari abl es.
In Eul er' s ti me it was thought that a function must be expressed by means
of a si ngl e formula, so a rel ation not so expressible was regarded as "sewn
together" from functions. For example, the rel ation
{
X
,
y =
2
X
,
if x < 0
if x :: 0,
was thought to consist of to dif erent functions.
It soon became cl ear that the matter was si gni fcantly more complex. When
he sol ved the problem of the vi brating stri ng, D. Bernoul l i
1
0 obtai ned an answer
i n the form of an i nfnite series whose terms were products of two trigonometric
functi ons, one of the moment t i n ti me and the other of the coordinate of the point
of the stri ng. Accordi ng to the then accepted view, this meant that the defection
of the stri ng was a function of two variables given by a single expression.
Thi s same problem of the vi brating stri ng was solved by D' Alembert.
D' Al embert ' s solution had a form quite different from Bernoul l i ' s, and, what
is most i mportant, could be given by different formul as for di fferent values of
the argument.
What l ooked to be an i nsoluble contradiction now l oomed up before 1 8th
century mathematics: two answers had been obtained for the same problem,
one expressed by a si ngle formul a for al l val ues of the argument and another
by several formul as. D. Bernoul l i ' s solution was questi oned because it was
thought that he had not found all solutions to the problem, just the solutions
expressi bl e i n a single formul a. A bitter controversy arose in which al l the
promi nent mathematicians of the 1 8th century - Eul er, d' Alembert, and others
- took part.
The controversy, i n essence, was over the concept of a function, the con
nection between functional dependence and the possi bi l ity of expressing thi s
dependence by means of a formul a. A defni tive solution to the question was
obtained at the begi nni ng of the 1 9th century, when the French mathemati ci an
J. Fourier
l l
showed that the sum of an i nfni te series of trigonometric functions
can be expressed by different formul as over different interval s. He then gave a
new defnition of function, stressing that the main thi ng was the assignment of
values for the function; whether this assignment was carried out by means of a
si ngl e formul a or not was uni mportant.
Chapter 3. Remarkable functions and curves, or a stroll through a
mathematical hall of wonders 77
Fourier's resul t was refned by the German mathemati ci an Di richlet, who
showed that any gi ven curve can be the graph of the sum of a trigonometric
seri es. It is requi red only that the number of maxi ma and mi ni ma on the curve
be fni te and the curve be bounded in ampl itude.
Afer a long debate that i nvol ved many emi nent schol ars (i ncl udi ng Loba-
chevski ) the fol l owi ng defni ti on of function was uni versal l y accepted:
A variable quantity y i s sai d to be a function of a variable quantity
x, i f to each value of the quantity x there corresponds a uni quely
determi ned val ue of the quantity y.
In thi s defnition not a word i s said about the function havi ng to be given
by a si ngle formul a throughout its i nterval of defni ti on. From the modern
standpoint, the only possi ble faw of thi s defnition is that i t menti ons vari able
magni tudes. From the viewpoint of "pure mathematics" this term i s not pre
ci sely defned. At the begi nni ng of the 1 9th century schol ars just gave examples
of variable magnitudes encountered i n physics (temperature of a cool i ng body,
displacement or velocity of nonuni form moti on, and so on). They thought
that one can associ ate wi th each of these variable magni tudes a mathematical
variable whose vari ation described the manner of vari ation of the physi cal mag
nitude. But then one of the most fundamental concepts of mathematics seemed
to depend on the physical notion of ti me.
The creation of the theory of real numbers and of set theory i n the second
half of the 1 9th century made possi ble, among other thi ngs, the rigorization
of the vague concept of a vari able magni tude. It turned out that one shoul d
thi nk of a vari able as a letter for which one can substitute numbers bel ongi ng to
some number set X. Of course, thi s approach to the concept of a variable was
more static than the concept accepted by schol ars at the begi nni ng of the 1 9th
century. It l acked the sense of motion and vari abi l ity. But i t made possi bl e a
defnition of function free from extramathematical concepts:
A function f defned on a number set X i s a correspondence rul e
that associ ates wi th every x i n X a number f(x) .
I n thi s very general formul ation, the notion of a function merges with those
of correspondence. mappi ng, transformation. operator, and so on.
For exampl e, from thi s point of vi ew the area of a triangle i s a function
defned on the set of all triangles and assumi ng its val ues i n the set of posi ti ve
numbers. And the ci rcle i nscri bed in a tri angle is a function defned on the
set of all triangles wi th values in the set of ci rcl es. In view of the fact, that i n
78 In Search o
f
Infnit
a coordi nati zed pl ane, triangles and ci rcles are determi ned by certain sets of
numbers, these functi ons can be reduced to numerical functions. In general ,
numerical functions are one of the most i mportant kinds of functi ons. That i s
why, i n the sequel, we wi l l restrict oursel ves t o functions defned on sets of
numbers and assumi ng numerical val ues.
Under a microscope. The rigorization of mathematical concepts cuts both
ways. On the positive side, it el i mi nates much that is uncl ear, i ncreases the
preci si on of mathematical di scourse, and results in more concl usi ve proofs.
But then there are losses. What science gains in terms of rigor it often pays
for wi th l oss of i ntui tive appeal . Al so, there is always the question whether the
rigorous defni ti ons correspond to the crude, intuitive i mages they are supposed
to model mathematically. Thus the stumbl i ng blocks removed from the feld of
mathemati cs often do not di sappear but are merely transferred to the boundary
between mathematics and its appl icati ons.
But for mathematics preci se defni tions are a vi tal necessi ty. By studyi ng
the properties of the concepts determi ned by these defni ti ons, schol ars learn
the properti es of the mathematical model s by means of which they are tryi ng to
describe the real world. If the properties of a parti cul ar model do not quite l ive
up to expectati ons, then al l thi s means is that in the process of its construction
one left out i mportant aspects of the objects the model was supposed to descri be.
Therefore, as soon as they rigorized the notion of a function, mathematicians
embarked on its thorough study. Then i t turned out that many of the objects
covered by the rigorous defni tion were most unl i kely to have been studied by
mathemati ci ans of past centuries. For example, Di richlet observed that the
correspondence
f
(X) =
{
:
i s a functi on.
i f x i s i rrational,
i f x i s rational
No 1 8th-century mathematician woul d have studied such a correspondence.
They studied only functi ons that described dependencies between physical or
geometric magni tudes. But all measurements of concrete magni tudes i nvol ve
errors, and for such magni tudes it makes no sense to ask whether thei r values
are rational or i rrational . Of course, one can object that the value of the function
I
| , i f x < 0,
sgn x = 0 , if x = 0,
| , if x > 0,
is not well determi ned near the point x = 0 - a small error of measurement can
change the function val ue sharply from negati ve to positive. But 1 8th-century
a mathematical hall ofwonders 79
mathematici ans knew that a function l i ke sgn x i s a mathematical i deal i zation
of a conti nuous function that grows abruptl y i n the region around x = O. As
for the Di richlet function, i t was useless for the most i deal i zed description of
any real process.
Even the addition of the requi rement of conti nuity hel ped l i ttle. Usi ng
thei r new freedom, mathemati ci ans began to construct compl i cated examples
of continuous functions that contradicted al l the fami l i ar expectati ons of thei r
predecessors. Henri Poi ncare
l
2
characterized the changed views of schol ars
at the end of the 1 9th century concerni ng the notion of a function in these
words: "There was a ti me when the search for new functions was motivated
by some practical ai m. Nowadays one i nvents functions to demonstrate the
faws i n the arguments of our predecessors; other than thi s, no concl usi on can
be obtai ned from them." The subsequent evolution of mathemati cs showed that
Poi ncare' s view was onesi ded. In modem physi cs one deal s with functions and
curves that have very strange properti es. But at the end of the 1 9th century
such appl ications were sti l I in the distant future. Mathemati ci ans del i ghted
i n i nvestigating properties of the weirdest functi ons, which thei r predecessors
woul d have certai nl y put in a hall of wonders. Not for nothi ng did some
mathematicians of the cl assical persuasion refer to the new theory of functions
as the "teratology of functions."
Mathemati ci ans at the end of the 1 9th century seemed to have put functi ons
under a microscope, whereas thei r predecessors vi ewed them wi th the naked eye
and could not have di scovered the fne features of thei r "microscopic structure."
In theory, 1 8th-century mathematicians real i zed that, l i ke all curves, the graphs
of functi ons had no thickness. But when they thought of functi ons, they i mag
i ned thei r graphs as drawn on a sheet of paper with a penci l , that is as havi ng
thickness. Such curves are piecewi se monotonic, that i s they consi st of a fni te
number of pieces that go up or down. Except for a few points, these graphs
have tangents everywhere, any two curves in a bounded part of the pl ane have
only fni tel y many points of i ntersection, and so on. Al l thi s seemed obvi ous.
These mathematicians di d not suspect that there exi st functi ons and curves
whose properties do not at all resemble the properties of such "respectable"
functions as pol ynomi al s, trigonometric and exponenti al functions, and so on.
But the mathematical apparatus they had devel oped contai ned the charge of
dynamite which subsequently expl oded the deceptive state of wel lbei ng. Thi s
dynamite was the theory of i nfni te seri es. At frst, such series made i t easier
to compute functional val ues. But then they became the means of obtai ni ng
new functi ons. And then, in the context of addition of "nice" pol ynomi al s,
8c
In Search of Infnit
i t became clear that addi ng more and more terms i ntroduced fner and fner
"futters" of the i nfni te sum-to-be, and i n the end one obtained a function
with properti es very di fferent from those of the functi ons studied in cl assical
anal ysi s. The behavi or of such functions made cl assical mathematicians think
of a madhouse. Thus here too the i nfni te, the i dea of addi ng i nfni tel y many
terms, had a revolutionary effect on the evolution of science.
And now let me i nvi te you on a strol l through a mathematical hall of wonders
whose exhi bi ts di ffer from school displays as much as an ichthyosaur or a
di nosaur di ffers from today' s l i vi ng creatures.
The genie escapes from the bottle. Di richlet' s own functi on, of whi ch we
spoke earl ier, was al ready unusual . After al l , there are i nfnitely many rati onal
and i rrational numbers on even the smal lest i nterval of the x axi s. But Di richlet' s
function i s | for rational numbers and c for i rrational numbers. Thus, as we
move along the x-axi s, the val ue of the function constantly jumps back and forth
between c and | It is i mpossi bl e to graph thi s function, si nce it is di scontinuous
at every poi nt.
And even among the continuous functi ons are some with unexpected prop
erti es. For example, can a conti nuous function have i nfni tel y many maxi ma
and mi ni ma on a fni te i nterval ? Thi s seems i mpossi ble at frst gl ance. After
al l , the curve has to take up space in fal l i ng from a maxi mum to a mi ni mum,
then agai n ri si ng to a maxi mum, etc. How can it do all thi s in a fnite interval ?
Neverthel ess, such odd functions do exi st and it is quite si mple to construct
one.
We shal l construct such a function on the segment c, | We frst cut the
segment in two and construct an equi l ateral triangle on the lef half. Now we
di vide the right half i nto two equal parts and construct a second equi l ateral
triangle on the segment | , We carry out the descri bed operation i nfnitely
many ti mes. As a result we fnd a mountain range wi th i nfnitely many peaks
gradual l y droppi ng down to the point | (Figure 1 2). We take the curve obtai ned
as the graph of the function f(x) . Thus, the function is defned at each poi nt
of the segment c. | wi th the exception of the right endpoint | Here we put
f( l ) = O.
Si nce the height of the peaks approaches c as x approaches | , we obtai n
a function conti nuous at al l points of the segment c, | But the number of
maxi ma and mi ni ma on thi s segment is i nfni te!
I n order t o construct such a strange function a mathematici an of the 1 8th
century woul d have to spend a lot of ti me tryi ng out combi nations of functions
Figure 1 2
before he would conjecture that the function
f(x)
=
{ X cos ,
0 ,
i f x # 0
if x
=
0
has i nfni tely many maxi ma and mi ni ma on the segment [0, I ] .
8 1
But functions wi th i nfni tel y many maxi ma and mi ni ma were onl y the frst
of the unpleasant surprises i n store for mathemati ci ans. The genie had only
begun to escape from the bottle.
Wet points. The function we constructed i n the precedi ng section has onl y one
point near which there are i nfni tely many maxi ma and mi ni ma; thi s i s the point
I . Now we shal l construct another function with many more such poi nts.
Imagi ne that rai n i s fal l i ng on the segment [0, I ] of the x-axi s. We go about
provi di ng shelter from the rai n as fol l ows. We di vi de the segment [0, I ] i nto
three equal parts and erect a tent i n the form of an equi l ateral tri angle i n the
central part. It protects all the points of the central part from the rai n (except
the endpoi nts, i . e. , the points [ and ;
Now we divide each of the two pieces left over i nto three parts and protect
the central part with a tent of the same form (but only one thi rd as wi de). We
now have the curve sketched i n Figure 1 3. In the thi rd step of thi s procedure
we erect four more tents, then ei ght more, etc.
Now we come to the question of whether al l the points of the segment have
been shel tered by the saw-toothed curve, or whether there remai n points wetted
by the rai n? It is easy to poi nt out some of the "wet" points - these are the
endpoints of the shel tered segments ( i . e. , such points as [ , , , , ,
etc. ) Al l these points were left unprotected when the correspondi ng tents were
erected, and they remain unprotected by the tents erected subsequently. It i s
82
o
1
:
z
"
1
J
Figure 1 3
2 7
Y 9
In Search of Infnit
8
9
I
easy to see that there are i nfnitely many such endpoints, but that they sti l l form
onl y a countable set.
But i t turns out that there i s an uncountable set of "wet" points i n addition
to these. It i s convenient to use the ternary representation i n order to describe
them. As we know, the ternary representation i s formed in the same way as the
deci mal representati on, except that the numbers are grouped in threes i nstead
of tens. Thus, in the ternary representation we onl y employ the three digits
0, 1 , 2 for writing numbers i n pl ace of the ten ordi nari l y used.
It i s easy to learn how to change the representation of a number whose
ternary representation i s
0. 02020202 . . .
It i s represented i n the deci mal system by the i nfni te geometric series
2 2 2
3
2
+
3
4
+
3
6
+
. . .
The sum of thi s series i s Thus,
1 /4 = 0. 020202 . . .
Now we can say exactl y which points remain wet after al l the protective
tents have been set up. The frst tent shelters the poi nts l yi ng between and
But these are just the points whose ternary representations have the form
0. 1 . . .
where the dots stand for any combination of di gi ts 0, 1 , 2 (i n the same way that
al l the points whose deci mal representations begi n wi th the di gi t 1 , i . e. , have
the form 0. 1 . . . , l ie between the points 1 / 1 0 and 2/ 1 0).
The poi nts sti l l wet after the frst step are those whose ternary representations
have the form
0. 0 . . .
or the form
0. 2 . . .
83
We can prove in the same way that after the two tents of the second step
have been set up, the points remai ni ng wet are only those whose ternary repre
sentations begi n wi th one of the fol l owi ng four combi nati ons:
0. 00 . . .
0. 02 . . .
0. 20 . . .
0. 22 . . .
Thus, any point in whose ternary representation 1 occurs wi l l at some stage
be protected from the rai n. In the end only those points remai n wet whose
ternary representations can be wri tten without using I . For example, the points
1 /4 = 0. 020202 . . .
and
3/4 = 0. 20202 . . .
remain wet.
But now i t must be clear why the set of "wet" poi nts has the cardi nal ity of
the conti nuum. Afer al l , thi s set can be put i nto one-to-one correspondence
wi th the set of i nfnite telegrams (see p. 6 1 ). We can do this by putti ng a point
such as
0. 20220200 . . .
i n correspondence with an i nfnite telegram by repl aci ng 0 by a dot and 2 by a
dash. Di fferent numbers correspond to di fferent telegrams when thi s procedure
is fol l owed. We al ready know that the set of i nfnite telegrams has the cardi nal i ty
of the conti nuum; thus, the set of wet points wi l l also have thi s cardi nal i ty.
The set of points we cal l ed wet was frst constructed by Cantor, and is now
cal led Cantor 's set. It i s clear from the construction of the tents that there are
i nfni tel y many maxi ma and mi ni ma of the saw-toothed curve near each point
of Cantor' s set.
The Devil 's Staircase. There i s sti l l another i nteresting functi on rel ated to
Cantor' s set. It is defned as fol lows. We frst di vide the segment [0, I ] i nto
three equal parts and stipul ate that our function equal at each point of the
mi ddle thi rd. Then we di vi de the left and right thirds i nto three equal parts and
84
In Search of Infnit
y
1
Figure 1 4
sti pul ate that the function equal from to , and equal from to . We
now have four segments on which the function i s not yet defned:
[ 0, ] [ . ] [ , ] [ , I ]
We di vide each of these i nto three equal parts and set the function equal to ,
, , , respectivel y, on the four mi ddle pieces.
Conti nui ng thi s process, we obtai n a function whi ch i s defned on all the
"dry" poi nts, i . e. , on al l the points not bel onging to Cantor's set. It is easy
to defne it on the points of thi s set too, and in such a way that i t becomes
conti nuous and nondecreasing on the segment [0, I ] . An approxi mation to the
graph of the function obtained is shown in Figure 1 4. It has the form of a
stai rcase with an i nfnite number of steps (of course, not all the steps are shown
on the graph).
Afer learni ng about curves wi t h i nfni tely many maxi ma and mi ni ma, we
are not l ikel y to be surprised at a staircase with an i nfni te number of steps. But
here i s somethi ng surpri si ng. Let us compute the total length of our staircase.
The frst step we constructed has length [, the next two have length apiece,
the next four have length _ apiece, etc. Thus, the sum of the lengths of al l the
steps is expressed by the i nfnite geometric seri es:
The sum of thi s series i s
I 2 4
+
+
+ 27 + . . .
I
" - I
-2
-
.
1 -
-
3
a mathematical hall ofwonders
(b)
Figure 1 5
85
Hence, the total length of the stai rcase i s I . But the function does not
increase at al l along these steps ; al l its rising i s concentrated at the points of
Cantor' s set. But very "few" points fal l to the share of this set - even though
its cardinal ity i s that of the conti nuum, its length i s zero! (The length of the
segment [0, I ] is I and the total length of the steps is I . ) Thus, our function
manages somehow to rise from 0 to I , even though i t onl y increases on a set of
zero length and never makes any jumps ! Isn' t thi s real l y surpri si ng?
A prickly curve. For many centuries mathemati ci ans dealt onl y wi th curves at
each of whose points a tangent could be constructed. If there were excepti ons
these occurred at only a few poi nts. The curve seemed to break at these poi nts,
and they were therefore called points of frcture. The curve drawn i n Fi gure
I 5a has two points of fracture, while the curve drawn in Figure I 5b has ten
points offracture.
But the curves that we just now constructed al ready have i nfnitel y many
points offracture: the curve associ ated with Figure 1 3 has a whole conti nuum
of them. It breaks at each poi nt of the Cantor set and, i n addi ti on, at the peaks of
al l the tri angles. However, even the curve associ ated wi th Figure 1 4 has breaks
on a comparatively "small" set of points: its length i s zero.
For a long ti me no mathemati ci an bel ieved that there coul d exi st a conti n
uous curve whol l y composed of "sawteeth," "breaks," and "pri ckl es." Math
ematicians were greatly amazed, therefore, when someone succeeded i n con
structi ng such a curve, and what i s more, a function with a graph l ike a picket
fence. The frst to do this was the Czech mathemati ci an Bolzano. But hi s work
remai ned unpubl i shed for a long ti me, and the frst publ i shed example was
that of the German mathemati ci an Weierstrass. However, i t is di fcul t for us
to present Weierstrass' example, for i t i s based on the theory of trigonometric
series.
86 In Search 01 Infnit
o
(a)
4
- .
o
< - - - .
(c)
o
f
(b)
Figure 1 6
We shal l now di scuss Bolzano' s example, maki ng a few sl i ght changes. We
frst di vi de the segment [0, 1 ] i nto four equal parts and construct an i sosceles
right triangle over the two central parts (Figure 1 6a). The resul ti ng curve i s the
graph of some functi on, which we shall denote by y = I
I
(x) . We next di vi de
each of the four pieces agai n i nto four equal parts and correspondingly construct
four more i sosceles right tri angles (Figure 1 6b). Thi s gi ves us the graph of a
second function y = h (x) . If we add these two functi ons, the graph of the sum
y = I
I
(x) + h (x) has the form sketched i n Figure 1 6c. It i s clear that thi s curve
al ready has more breaks and that these breaks are more densely di stributed. In
the next stage we again di vide each piece i nto four parts, now constructing 1 6
i sosceles right triangles and then addi ng the corresponding function y = h (x)
to the function y = I
I
(x) + hex) .
As we continue thi s process, we obtain a curve wi th a larger number of
breaks. In the l i mi t we obtai n a curve wi th a break at each poi nt and possessi ng
a tangent at no poi nt.
A si mi l ar exampl e of a curve possessi ng a tangent at no poi nt was con
structed by the Swedish mathemati ci an Helge von Koch. 1 3
He took an equi l at
eral triangle, di vi ded each of its sides i nto three equal parts, and then constructed
new equi l ateral triangles with peaks pointing out over the three central secti ons.
Thi s gave hi m a fgure somethi ng l i ke a si xpoi nted star (Figure 1 7a). He then
went on to divide each of the twel ve sides of thi s star i nto three equal parts,
agai n constructi ng equi l ateral tri angles. Thi s gave the even more pri ckl y curve
drawn in Fi gure 1 7b. After i nfnitely many di vi sions and constructions of right
triangles he obtai ned a curve at each point of which there was a break or a
prickl e.
Mathemati ci ans constructed many continuous functions whose
graphs possess a tangent at no point, and began to study thei r properties. These
(a)
(b)
Figure 1 7
87
properties have no si mi l arity to those of the "wel l -behaved" smooth functi ons
with whi ch they had dealt up to that time. It i s no wonder, then, that mathe
maticians trained i n the cl assical tradition regarded these new functions wi th
astoni shment. Going even beyond thi s, the promi nent exponent of cl assi cal
anal ysi s Charles Hermi te
l
4
wrote as fol l ows to hi s friend, the Dutch math
ematician Stieltjes
l
5
: "I turn away i n horror from thi s regrettabl e pl ague of
conti nuous functions that do not have a derivati ve at even one poi nt" (Le. , as
we have named them, everywhere-pri ckl y curves).
In physi cs we encounter curves hi ghl y remi ni scent of the everywhere prickl y
curves of von Koch and others. These curves are the trajectories of parti cl es
undergoing Browni an motion caused by col l i si ons wi th molecul es. The French
scientist J. Perrin
l
6
made a sketch of the motion of these parti cl es. He observed
thei r posi ti ons every 30 seconds and connected the points thus obtained wi th
straight l i ne segments. Hi s resul t was a tangle of broken l i nes somethi ng l i ke that
sketched in Figure 1 8. But it should not be thought that the parti cles observed
actual l y moved i n straight l i nes between the separate observati ons. If Perri n had
observed them every half second instead of every half mi nute, he would have
had to repl ace each straight l i ne segment by a much more compl icated broken
l i ne l i ke that in Fi gure 1 8. And the shorter the i nterval between observati ons, the
more compl icated and "prickl y" the broken l i ne woul d become. The American
mathemati ci an N. Wiener
l
7
showed that i f the particles i n Browni an motion are
sufciently small that thei r i nerti a can be neglected, they move along curves
which have no tangent at any poi nt.
A closed curve of infnite length. We have ofen encountered curves of i nfni te
length: the straight l i ne, the parabola, etc. , al l have i nfni te length. But al l these
88
In Search of Innit
Figure 1 8
curves go off to i nfni ty, so it i s not surpri si ng that they have i nfnite length.
However, it is not di fcul t to construct a curve enti rel y contai ned in a fnite
regi on of the pl ane and sti l l having i nfni te length. For thi s we can take a ci rcle
and wind a spi ral wi th i nfni tel y many turns around it. Si nce the number of turns
i s i nfnite and the l ength of each turn i s greater that that of the ci rcumference
of the ci rcl e, the length of the spi ral must be i nfni te.
But can we construct a closed curve of i nfnite length? The ordi nary cl osed
curves - the ci rcle, the el l i pse, the cardi oi d - all have fni te length. However,
the length of von Koch' s prickl y curve is i nfni te. Indeed, the peri meter of the
ori gi nal triangle is 3. As is easi l y computed, the star obtai ned in the frst stage
has length 4. And in the fol l owi ng stage we obtai n a curve composed of 48
segments each of length 1 19. Thus its peri meter i s 48/9. Next we obtai n a
curve of length 1 92/27, etc. In general , at the nth stage we obtai n a curve wi th
peri meter 3

, ,

But thi s expression approaches i nfnity as n i ncreases, so
that the length of von Koch' s curve is i nfnite.
A mathematical carpet. It i s told that Catheri ne the Second asked one of
her general s what the di fference is between a mortar and a howitzer. The
embarrassed general repl i ed: "You see, Queen Mother, a mortar is one thing
and a howitzer is somethi ng el se." We woul d probably receive an i nformative

. . W

.
.

.
.

.
.
.
. . .

Figure 1 9
89
answer l i ke thi s one i f we were to ask a person knowi ng l ittle about mathemati cs
what the diference i s between a curve, a surface and a sol i d. Moreover, he
woul d be surprised that we asked about such obvi ous thi ngs. After al l , i t i s
qui te cl ear that a curve, a surface, and a sol i d are qui te di fferent thi ngs, and no
one woul d cal l a ci rcle a surface or a sphere a curve.
But a wi tty chess master once sai d that the di fference between a master and
a begi nni ng chess pl ayer i s that the begi nner has everything clearl y fxed i n hi s
mi nd, whi l e to the master everything i s a mystery. That i s al so how matters
stand wi th our questi on. Of course, when we are speaki ng of such geometric
fgures as a ci rcle or the perimeter of a square, no one has any doubts that these
are curves and not surfaces. But i n the course of mathematical devel opment,
si nce Cantor' s di scoveries, there have appeared many strange geometric fgures,
and even an experienced, knowledgeable professor, not to speak of a student,
wi l l not be able to deci de right away whether they are curves, surfaces, or sol i ds.
We shal l present some of these fgures.
We take a square of side 1 and divi de i t i nto 9 equal parts; then we di scard
the central part (leavi ng the si des of the di scarded square). Afer thi s we di vide
each of the remai ni ng squares i nto 9 equal squares, and agai n di scard the central
squares. Afer one more such operation we arrive at the fgure drawn in Fi gure
1 9 (the squares to be discarded are hatched). It is clear that the fgure in Fi gure
1 9 is sti l l a surface. But we wi l l not stop at the thi rd step; the squares wi l l be
di vi ded i nto ni ne equal parts i nfni tel y many ti mes, and each ti me the central part
wi l l be di scarded. In the end we obtain a geometric fgure cal l ed Sierpinski 's
carpet after the Pol i sh mathematician who devi sed i t.
The fgure l ooks l i ke cl oth woven by some mad weaver. The threads of
90
In Search of Infnit
Figure 20
warp and woof run down and across and i nterl ace to form a very symmetric
and beauti ful desi gn. But the resul ti ng carpet is ful l of holes - there is not an
uncut piece in it; even the smallest square had to have its center cut out.
And i t is not at all clear whether thi s carpet is a curve or a surface. After
al l , on the one hand, it does not contai n a si ngl e whole piece, and so can hardl y
be called a surface; but, on the other hand, the threads formi ng i t were woven
i nto such a complex pattern that probabl y no one woul d unhesitati ngly cal l
Sierpi nski ' s carpet a curve. In any case, it woul d be very hard to draw thi s
"curve."
But Sierpi nski 's carpet is not the most compl i cated geometric fgure. Instead
of a square we could have taken a cube, di vi ded it i nto 27 equal small cubes
and di scarded the central smal l cube along wi th its 6 nei ghbori ng cubes. Then
we woul d have di vi ded each smal l remai ni ng cube i nto 27 equal parts and
agai n woul d have carried out the operation of di scardi ng certain parts (the sol id
remai ni ng afer two such operations i s shown i n Figure 20) . Suppose that the
operati on had been carried out i nfnitel y many ti mes. What ki nd of fgure woul d
we get afer al l the pieces had been di scarded - a curve, a surface or a sol i d?
Euclid does not rely on Euclid. When a complicated geometric problem
was pl aced before mathemati ci ans of an earl ier ti me, they frst proceeded to
exami ne what Eucl i d had written about it. After al l , for al most two thousand
years Eucl id was the standard of mathematical rigor and an encycl opedia of
geometric knowl edge. It i s si gni fcant that even phi l osophers, stri vi ng to secure
themsel ves from reproach regardi ng the rigor of thei r arguments, had recourse
9 1
to Eucl i d' s l anguage and formul ated thei r statements as axi oms, lemmas, and
theorems.
But as far as our question i s concerned, everything Eucl i d wrote was enti rel y
too vague. The frst l i nes of Eucl i d' s book Elements read as fol l ows:
1 . A point is that which has no parts.
2. A curve i s length without width.
3. The extremity of a curve is a point.
4. A surface is that which has only length and width.
5. The extremity of a surface i s a curve.
6. A boundary is that which is the extremi ty of somethi ng.
7. A fgure i s that which i s contai ned wi thi n somethi ng or wi thi n some
boundari es.
Now, l i ke these or not, they are not rigorous mathematical defni ti ons. A
person not knowi ng what points, curves, or l i nes are wi l l hardl y get much
useful i nformation from these "defni ti ons," so remi ni scent of the answer of the
confused general ("a curve is one thing, and a surface is somethi ng el se"). And,
i n any case, we shall not succeed i n fndi ng out from these defni ti ons whether
Si erpi nski ' s carpet i s a curve or a surface, whether i t has j ust length wi thout
width or both length and wi dth.
However, such compl icated fgures as Sierpi nski ' s carpet were unknown i n
Eucl i d' s ti me, and defnitions were not real l y necessary for si mple fgures -
everyone coul d pick out which were the curves and which were the surfaces i n
a fgure. I t seems though, that Eucl i d hi mself felt that not al l was right with hi s
defni ti ons of the fundamental concepts. I n any case, havi ng presented these
defni ti ons at the begi nni ng of the book, he went on to completel y forget about
them and did not employ them even once i n the remai nder of his work.
Ar rigorous defnitions needed? Eucl i d' s authority stood unquesti oned for
two thousand years. To doubt hi s statements in any way was to deci si vel y and
i rrevocabl y undermi ne your own mathematical reputati on. One of the greatest
mathematicians of the 1 9th century, Carl Friedrich Gauss, arri ved at the i dea
of a non-Eucl i dean geometry even before Lobachevski , but di d not publ i sh hi s
*
investigations, feari ng, as he wrote one friend, the screams of the Boeoti ans.
92
In Search of Infnit
It was fnal l y the mathematical exploit of the great Russian geometer Nikolai
Ivanovich Lobachevski , who di d publ i sh hi s di scoveries i n spite of the derision
of the uncomprehending savants, that gave the worl d non-Eucl i dean geometry.
It became clear after the appearance of Lobachevski ' s work that there exi st
two geometri es, both i rreproachable logical l y, but arri vi ng at entirel y di fferent
theorems. But if thi s is so, then every appeal to "geometric obviousness"
completely lost its value. Each geometric assertion now had to be based on
rigorous defni ti ons and i rreproachable logical arguments. And now it was
especi al l y i mportant that the fundamental geometric concepts of curve, fgure,
and sol i d be gi ven exact defni ti ons, i n no way l i ke those of the type "thi s i s one
thi ng, and that is somethi ng el se."
Thi s attempt at r
i
gorous defni tion characterized not only the geometry, but
also the anal ysi s of the 1 9th century.
Science had succeeded in sol vi ng the most varied problems, from calcu
l ati ng the trajectory of an arti 1 1ery shel l to predicting the moti ons of pl anets
and comets, wi th the aid of the di fferential and integral calcul us based on the
work of Newton, Lei bni z, Euler, Lagrange,
I
S
and other great mathemati ci ans
of the 1 7th and 1 8th centuri es. But the fundamental concepts wi th whose aid
these remarkabl e results were achieved were defned in a hi ghly unri gorous
manner. The mathematical anal ysi s of that ti me was based on the concept of
an i nfnitesimal quanti ty, something bal anci ng on the border of exi stence and
nonexi stence; somethi ng l i ke zero, but not real l y zero. And mathematici ans
of the 1 8th century were forced to encourage thei r dubious students with the
words: "Work, and bel i ef wi 1 1 come to you."
But, mathematics is not rel i gi on; it cannot be founded on fai th. And what
was most i mportant, the methods yi eldi ng such remarkable results in the hands
of the great masters began to lead to errors and paradoxes when employed by
thei r less talented students. The masters were kept from error by thei r perfect
mathematical i ntui ti on, that subconsci ous feel i ng that often leads to the right
answer more qui ckl y than lengthy logical reasoni ng. But the students di d
not possess thi s i ntui ti on, and the end of the 1 8th century was marked by an
unprecedented scandal in mathematics - an i nfux of formul as worth less than
the paper they were pri nted on and questionable theorems whose domain of
appl i cabi l i ty was enti rel y unclear.
So, l i ke chi ldren who break a beauti ful toy in order to see what makes
i t work, 1 9th-century mathemati ci ans subjected all the concepts employed up
to that ti me to a severe critique and then began to rebui l d mathematics on a
A proverbI al l y dul l Greek trIbe.
Chapter 3. Remarkable functions and curves, or a stroll thrugh
foundation of ri gorous defni ti ons. Appeals to i ntui ti on were rejected; i n pl ace
*
of thi s they demanded the most ri gorous logi c. Found wanti ng in logic were
the si mple statements met wi th in a course in anal ysi s, such as: "Consi der the
domain G bounded by the cl osed curve r." What is a cl osed curve? Why is i t
the boundary of a domai n? Into how many parts does a cl osed curve di vide the
pl ane, and which of the these parts is bei ng studied?
The mathemati ci ans of the 1 8th century did not repl y to these questions.
They just drew an oval and thought that thi s was al l that needed to be done. But
no one bel i eved i n pictures i n the 1 9th century. The question "what i s a curve?"
was only one of the vital questions facing anal ysts.
However, a l ong ti me went by before they succeeded i n gi vi ng a compre
hensi ve answer to thi s questi on.
A curve is the path of a moving point. In order to arrive at a rigorous defni ti on
of curve i t was necessary to move away from the concrete objects on whi ch
the formation of the mathematical concept was based: long, thi n threads; l ight
rays; long, narrows roads, etc. In all these cases the length i s so much greater
than the wi dth that the l atter can be neglected. After mathematical i deal i zati on
we arri ve at the notion of length without wi dth.
The frst to try to give a rigorous defni tion of curve was the French math
emati ci an Cami l l e Jordan. 1 9 He proceeded from the fact that the trajectory of
the motion of a very smal l body may be represented by a long, narrow tube.
As we di mi ni sh the si ze of the body, the tube becomes more and more narrow
and in the l i mit becomes the trajectory of a moving point - a curve possessi ng
no wi dth. Jordan appl ied thi s i mage i n hi s defni tion of curve. Namely, he
called the trajectory of a movi ng point a curve. Here the point i s to move in a
continuous manner, not maki ng any jumps.
Jordan' s defni ti on can be more exactly stated as fol l ows: In order to de
termi ne the posi ti on of a movi ng point its coordi nates must be given for each
moment during the moti on. Si nce the motion takes pl ace over a fnite ti me
interval , we can assume wi thout loss of general ity that thi s i nterval is [0, 1 ] . In
other words, the point begi ns to move at some moment of ti me taken as the
start of the observation and completes its motion after a certain uni t of ti me
has el apsed (a second, a mi nute, a year, etc. ) . The coordi nates of the movi ng
point are gi ven for each moment of ti me t during the passi ng of thi s i nterval .
Thus, the coordi nates of the point depend on the moment of ti me t , and so are
"
True, they frequently tended to throw out the baby with the bathwater; i n the 20th century
much of what was thrown out became once more part of science.
94
In Search oj Infnit
r
Figure 2 1
functi ons of t . We shal 1 denote these functions by J(t ) and ge t ) :
x = J(t ) y = g(t)
The requi rement that the point move conti nuously amounts to the requi re
ment that the functi ons J(t ) and get ) be conti nuous at each point of the segment
[0, I J. Roughly speaki ng, a smal 1 change i n t shoul d produce only a smal1 change
in the functi ons J(t ) and ge t ) .
Jordan' s defni ti on turned out t o be a rather successful one. Al 1 the curves
with which mathemati ci ans had dealt up to thi s ti me turned out to be curves
in Jordan' s sense, or, as we say, Jordan curves. Curves made up of arcs of
di fferent curves are also Jordan curves.
The theorem is obvious, but the proof is not. Employi ng hi s concept of
curve, Jordan was successful i n gi vi ng a preci se meani ng to the sentence from
the anal ysi s textbook that we spoke of earl ier: "Let the cl osed curve r bound
the domai n G." A cl osed Jordan curve i s a curve which at t = I passes through
the point that was passed through at t = O. The curve does not i ntersect itsel f
as long as no two values of ti me t, and t
2
between 0 and | correspond to the
same point on the curve.
Jordan proved the fol 1 owi ng theorem.
A closed Jordan curve r which does not intersect itsel divides the plane into
two parts. Two points contained in the same part can be connected by a brken
line that does not intersect the curve r, but two points contained in dif erent
parts cannot be connected by such a brken line; any brken line connecting
them must intersect the curve r (Figure 2 1 ).
Chapter 3. Remarkable/unctions and curves, or a stroll through
a mathematical hall o/wonders
95
Thi s theorem seems completely obvi ous. Its proof, however, requi red very
subtle arguments. Even when the curve r is the boundary of a polygon, the
proof remai ns qui te compl icated.
The two parts i nto which a closed Jordan curve di vi des the plane are cal l ed
the exteri or and i nterior domai ns bounded by thi s curve. The concept of a
domain bounded by a cl osed curve thus acqui red an exact meani ng.
A curve passing through al l the points of a square. It appeared at frst when
Jordan gave hi s defnition of a curve that the goal had been achi eved; a ri gorous
defnition of the concept of a curve was now avai l able that di d not depend on
intuition. But it was quickly found out that thi s was not the case - Jordan' s
defni ti on embraced not onl y what mathemati ci ans usual l y called curves, but
also geometric fgures that no one would cal l curves. Mathemati ci ans coul d
somehow reconci l e themsel ves to everywhere-prickl y curves, but no one had
the heart to cal l a square a curve. But i t did turn out that the square, the triangle
(not the peri meter of the fgure, but i n each case the fgure itsel f with al l i ts
i nterior poi nts) and the di sk were curves i n Jordan' s sense. Thi s was proved by
the Ital ian mathematician Peano.
2
0
We al ready menti oned that Cantor set up a one-to-one correspondence be
tween the points of a segment and those of a square, i . e. , he showed that there
are just as many poi nts on the segment as are in the square (see p. 65). But hi s
correspondence was not conti nuous. As the point moved along the segment,
the corresponding point on the square did not crawl around l ike a beetle, but
jumped around l i ke a fea. Indeed, let us take the points
0. 5000000 . . . and 0. 49999999000000 . . .
on the segment. These points are qui te cl ose together. But the correspondi ng
points on the square are far apart. For the poi nt correspondi ng to the frst of
these i s (0. 5000 . . . , 0. 000 . . . ) situated on the bottom of the square, whi l e the
poi nt correspondi ng to the second is (0. 4999000 . . . , 0. 9999000 . . . ) si tuated
very cl ose to the top of the square. And if we i ncrease the number of ni nes i n
the second point, thus bri ngi ng i t cl oser t o the frst, the correspondi ng points
on the square do not begi n to approach one another.
Thus, Cantor' s mappi ng of the segment onto the square, although one
to-one, was not conti nuous, and so did not give rise to a Jordan curve. Peano
succeeded i n setti ng up another mapping of the set of points of the segment onto
the set of points of the square which sent nei ghbori ng points on the segment into
neighbori ng points on the square. In other words, Peano was able to construct
a curve (in Jordan' s sense) whicQ passed through al l the points of the square.
96
In Search of Infnit
8
G
.
A E A E A E
(a)
(b) (c)
Figure 22
But, after all, the square is uni form, so we wi l l not be able to see where the
curve begi ns, where it ends and how i t moves about the square. Therefore, we
shal l fol l ow the example of the physi ci st Perri n, rather than that of an abstract
painter, and sketch the position of the movi ng point usi ng l i ne segments. The
shorter the i nterval s of ti me taken between separate "observations," the more
accuratel y wi l l the broken l i ne thus obtained represent Peano' s curve.
We shal l frst observe the position of the movi ng poi nt every second. In
other words, we observe its position at the begi nni ng of the motion, at second
after the begi nni ng of the moti on, at second afer the begi nni ng of the moti on,
at second, and at the end of the moti on. This gi ves us 5 points. Connecting
them, we obtai n the l i ne ABCDE drawn i n Figure 22a.
Natural l y, thi s l i ne does not pass through al l points of the square. Now we
reduce the i nterval of ti me between i ndi vi dual observations and observe the
posi ti on of the point every _ second. Now the curve twi sts more, the number
of breaks i ncreases, and i t takes the form sketched i n Figure 22b. If we observe
the position of the movi ng point sti l l more often, we obtain the curve sketched
in Fi gure 22c. We see that the curve fl l s the square more and more densely,
that i t approaches more and more cl osely to each of its points. In the l i mit, in
which we woul d be constantly observi ng the movi ng point, we woul d obtai n a
curve passi ng through al l points of the square wi thout excepti on.
It shoul d be noted that, while Peano has an advantage over Cantor in that
hi s curve is conti nuous, he fal l s short in another respect. Hi s curve no l onger
gives rise to a one-to-one mappi ng of the segment onto the the square; i t passes
through some points of the square several ti mes. It was l ater proved that i t i s
i mpossi bl e t o obtain a correspondence that i s both one-to-one and conti nuous:
there does not exi st a Jordan curve passi ng through al l the poi nts of the square
exactly once !
Chapter 3. Remarkable functions and curves, or a strll thrugh
Everything had come unstrung. It i s di ffcul t to put i nto words the effect that
Peano' s resul t had on the mathematical world. It seemed that everythi ng was
i n rui ns, that al 1 the basic mathematical concepts had l ost thei r meani ng; the
di fference between curve and surface, between surface and sol i d was no longer
clear. (The resul t showing the i mpossi bi l i ty of a one-to-one conti nuous corre
spondence between the segment and the square was sti l 1 unknown. ) Poincare
bitterl y exclai med: "How is it possible that i ntuition could so deceive us?"
It soon became clear that Jordan' s defni ti on had its faults. On the one hand
i t was too broad: Peano' s curve fts i t. But on the other hand i t was too narrow.
For example, a circle with a spiral wrapped around it is not a Jordan curve.
So the question was agai n rai sed: what is a curve and how does it di ffer
from a surface? The answer is rel ated to Cantor' s general studies of geometric
fgures.
How to make a statue. Havi ng founded the theory of sets, Cantor now turned
his attention to the questi on: what is a geometric fgure ? The most general
answer to thi s question woul d read: a geometric fgure is any set of points i n
space. I f thi s set l ies i n the pl ane, then we obtai n a pl ane geometric fgure. But
thi s answer would be too general - a "fgure" i n thi s sense woul d have no real 1 y
i nteresting properties. The geometry of such fgures would be al most devoi d
of theorems.
So it was frst of al 1 necessary to l i mi t the cl ass of sets to be studied,
separati ng out those which had properties cl ose to those of ordi nary geometric
fgures.
In order to separate out thi s cl ass of fgures we have to deci de what i t i s that
ordi nary fgures such as the square, di sk, l i ne segment, etc. , have in common. It
turns out that we can construct al1 these fgures by means of a si ngl e procedure.
When Rodi n was asked how he managed to make his remarkable statues,
he replied: "I choose a bl ock of marble and chop off whatever I don' t need."
We can obtai n any "reasonable" bounded pl ane geometric fgure by thi s
same method: we take a square which contai ns i t and chop off whatever we do
not need. Of course, we do not chop everything off at once, but proceed step
by step, at each step removi ng di sks. Here we remove the i nterior of the di sk
whi l e its boundary, a ci rcle, i s l ef i n the fgure. Briefy, we are removi ng an
open disk.
At frst we mi ght think that thi s procedure woul d only yi el d fgures l i ke those
i n Figure 23. But the secret l ies in the fact that we remove not just one or two
disks, but a countable set of disks. In thi s way we can obtai n any "reasonable"
fgure we l i ke. To see thi s, note that the set of di sks wi th rati onal radi i and
98
In Search of Infnit
Figure 23
centers is countable. (Thi s is easi l y shown by the methods of Chapter 2. ) To
obtai n a requi red fgure we need only remove from a square that contai ns i t (the
"block of marble") all open rational di sks that don' t contai n a si ngl e one of its
poi nts. And i f we remove such di sks from the pl ane rather than from a square,
then we can obtai n unbounded fgures as wel l .
Usi ng the procedure just described we can obtain disks and squares, el l i pses
and astroi ds, regul ar polygons and stars, but not a square mi nus a si ngle vertex.
Thi s i s because removi ng the vertex woul d entai l removing a neighborhood of
that vertex. Figures obtai ned from the pl ane by removing from i t a countable
number of open di sks are said to be closed.
Continua. In addition to the ordi nary geometric fgures, i t turns out that by
removi ng a countable set of open di sks (or squares, etc. ) we can also obtain
other sets qui te unl i ke the ordi nary fgures but sti l l possessi ng many i nteresti ng
properti es. For i nstance, Si erpi nski ' s carpet, of which we have already spoken
at length, can be obtai ned in the fol l owi ng manner: from the square of si de |
di scard smal l squares one by one, leavi ng thei r si des behi nd.
However, thi s di scardi ng process can yield "fgures" not composed of a
*
si ngl e whole piece. For example, if we remove "crosses," as in Figure 24,
i n the end we obtai n a set not contai ni ng a si ngle whole piece. (Such a set
i s said to be completely disconnected) . Hence, we make the requi rement that
afer each di scardi ng operation there must remain a set consi sti ng of a si ngle
piece. Then after all the removal s there wi l l remai n a set composed of a si ngle
piece (or, as mathemati ci ans say, a connected set) . The set obtai ned wi ll al so
I ncl udmg termmal segments such as, for example, the segments A B . CD . E F . G H.
A B
E F
Figure 24
Figure 25
be bounded, i . e. , i t i s entirel y contai ned i n some square.
A set F sati sfyi ng the fol l owing conditions:
99
1 . it i s obtai ned from a square by di scardi ng a countable set of open di sks
(or squares, etc. ),
2. it i s composed of a si ngle piece ( i t i s connected),
was said by Cantor to be a continuum (recall that the Lati n word continuum
means unbroken). A conti nuum turns out to be the most general set sti l l pos
sessi ng properties qui te si mi lar to those of ordi nary geometric fgures.
Cantor curves. Now we are i n a position to answer the question: what i s a
pl ane curve? Si nce pl ane curves must be geometric fgures, it is cl ear that we
must search for them among the conti nua. But a square and di sk are conti nua,
and we certai nl y do not want to cal l these fgures curves. Thus, we have to add
on some other requi rement which woul d el i mi nate such fgures.
Note that both the ci rcle and the square contai n whole pieces of the plane.
But a curve woul d not contai n whole pieces of the plane; no matter how small
1 0
In Search of Infnit
1
Q
Figure 26
a square we took, there woul d always be points on it not bel onging to the curve
(Fi gure 25).
So here i s the supplementary condition we need:
A plane curve in Cantor 's sense is a continuum contained in the plane which
does notfll a whole piece of the plane ( i . e. , in every square there are points not
bel ongi ng to thi s curve).
For example, a segment, the peri meter of a triangle, a ci rcle, and a four-l eaf
rose are curves. So is the Sierpi nski carpet; i ndeed, we put a hole in ever
square obtai ned in the di vi sion process i nvol ved in its construction, so that no
whole piece of the pl ane i s contained i n i t. Other Cantor curves i ncl ude a ci rcle
with a spi ral wound around i t and the saw-toothed curve of Figure 26 together
with the segment [0, 1 ] of the y-axi s. More general l y, all those fgures that seem
to our i ntui ti on to be curves are also curves i n Cantor' s sense, whi l e any fgure
contai ni ng even one whole piece of the pl ane does not belong to the cl ass of
Cantor curves.
But even among Cantor curves are some whose properties are qui te unl i ke
those of ordi nary curves. We shal l now discuss some of these.
Can a curve have area? Of course, now that the reader has made the ac
quai ntance of curves passi ng through all the points of a square, he wi l l not be
surprised by anythi ng. But even so, can a curve have area? After al l , Eucl i d
di d say that a curve i s length wi thout wi dth. And how can we get area from
somethi ng wi thout wi dth? Cantor' s defnition of a curve, too, says that the
curve cannot contain a whole pi ece of the plane. Where wi l l we fnd area i n
thi s case? But we have seen many ti mes that i nfni te processes lead to enti rely
unexpected resul ts.
Before we study the questi on, we must come to an understanding about
the exact meani ng of the words used. What is meant by the words "a curve
a mathematical hall of wonders 1 0 1
Figure 27
has zero area" or "a curve has nonzer area"? Let us take the most ordi nary
curve- a strai ght l i ne segment. Si nce its wi dth is zero, we can pl ace it i nsi de
a rectangle of arbi trari l y small area; we onl y have to choose a rectangle of
sufciently small wi dth. In exactly the same way we can put a ci rcl e i nsi de a
polygonal domai n of arbitrari l y smal l area. This can be done by i nscri bi ng i n
i t a regul ar pol ygon wi th a very l arge number of sides and then ci rcumscri bi ng
a si mi l ar pol ygon. The region incl uded between the two pol ygons wi l l have
smal l area (the more si des our polygons have, the smaller the area), and the
ci rcle is entirel y contai ned in thi s region (Figure 27) .
Now it is clear what is meant by the words a curve has zero area. They
mean that no matter how smal l a positive number e we take, we can fnd a
pol ygonal domain which contai ns the curve and has an area less than e . And
i f we cannot fnd such a domai n, the area of the curve i s not equal to zero.
In order to make the defnition clearer we shall appl y it to a more compl icated
curve than a si mple segment or ci rcl e. Sierpi nski ' s carpet represents, of course,
a very compl icated curve. Let us fnd its area. Recal l frst that the area of the
whole square was | In the frst step we di scarded the central square of area 1 19.
We thus got a pol ygonal domai n of area 8/9. In the second step we discarded
8 squares each of which has area 1 /8 1 . Thi s left a polygonal domai n of area
8 8
_
64
_
8
2
s s
'
?

It is now clear that after the thi rd step there wi l l remai n a polygonal domai n
of area (8/9)3 , afer the fourth a domain wi th area (8/9)4 , etc. But if you
1 02
In Search of Infnit
Figure 28
take any proper fraction and raise i t to higher and hi gher powers, the resul t tends
to zero. Thi s tel l s us that for every e > 0 , after sufciently many steps, we get
a polygonal domai n of area less than e . And thi s domain covers Sierpi nski ' s
carpet. Hence the area of Si erpi nski ' s carpet i s zero.
Thi s would seem to mark the complete tri umph of EucI i d' s defni ti on. Even
such a complicated curve as Si erpi nski ' s carpet has area zero. But it woul d be
premature to celebrate the tri umph now. After al l , no one forced us to di scard
such l arge pieces. Figure 28 i l l ustrates a more economical way of removing
squares. If we choose squares of the right size, then the total area of the removed
squares won' t exceed | . But then after each step of the removal process we
are left wi th a polygonal domai n of area not less than | . The fgure left over
after completion of our process cannot be covered by a polygonal domai n of
area less than | Therefore its area cannot be said to be zero. Now, thi s
remai nder, just as i n the case of Si erpi nski ' s carpet, i s a curve (i n Cantor' s
sense) - in constructing i t we made a hol e i n every square and rectangle and
not a si ngl e whole rectangle or square was left behi nd.
As a resul t, therefore, a curve i n Cantor' s sense can have nonzero area!
Domains without ara. The example we just anal yzed is not too convi nci ng:
the curve we obtai ned consi sts of poi nts of self-i ntersection and does not bound
any domai n. So the question ari ses: can a "good" curve that does not i ntersect
i tself have nonzero area? It happens that it can !
We can construct such a curve by changing a l i ttle the construction carried
out before. We frst construct a set i n which you not only cannot fnd a whole
piece of square, but not even a whole piece of curve, and the area of thi s set
1 03
Figure 29
wi l l not be zero. To do thi s we must di scard whole crosses rather than central
squares, as shown in Figure 29. Here we select the di mensions of the discarded
fgures so that thei r combi ned area is less than hal f the area of the whole square.
But then the area of the lefover set is at least half the area of the whole square.
Note that in constructing the set we discarded whole crosses, ruthl essl y teari ng
up the square. No two points of the set remai ni ng can be connected by a curve,
not even a curve in Cantor' s sense; every connection between its points has
been broken. As mathematicians would say, the remai nder i s a completel y
di sconnected set. And sti l l the area of thi s set, not contai ni ng a si ngl e whole
piece of the pl ane nor an arc of a curve, i s di fferent from zero; you cannot cover
this set with a polygonal domain of area less than 1 /2.
Now it i s easy to construct an example of a cl osed curve that does not
intersect i tsel f and has nonzero area. To do thi s we need onl y connect the
poi nts we al ready have, just as we drew a curve through all the points of the
square. And because we di scarded whole crosses at each step, our curve wi l l
not intersect itself (i n this it differs from Peano' s curve). But si nce i t passes
through al l points of the set, whose area must be at least 1 12, the area of the
curve obtained must be at least 1 12.
It i s al so no trouble now to construct a domain wi thout area. For thi s we
need onl y connect two points A and B of our curve with some ki nd of curve,
perhaps a semicircle. Then we obtain a curve which bounds some domain G .
And what is its area? The answer depends on whether or not we i ncl ude the
boundary with the domain - after al l , the boundary itself has an area of at least
1 /2. Clearl y, our domain has no area in the ordi nary sense of the word. In
mathematics domai ns with no area in the usual sense are called nonsquareable.
1 0
In Search of Infnit
And yet, they can be measured. Mathematicians had thought about what
is meant by the area of a fgure before the di scovery of nonsquarable fgures.
Before that, schol ars used the notions of length, area, and vol ume without
subjecting them to rigorous critical anal ysi s. The story i s told of a French "circle
squarer" who brought hi s "solution" to the Paris Academy of Science. Asked
what he meant by the area of a ci rcl e the man exclaimed: "You don' t defne
areas, you compute them." And this very point of view was also widespread
among mathematicians. They thought that area is a number associ ated with a
geometric fgure and having obvious properties (the area of a fgure is the sum
of the areas of its parts, congruent fgures have the same area, and so on). Not
for a moment did they doubt that every pl ane fgure has area (possibly equal to
zero or i nfni te).
A characteristic of mathematics i s that i t pursues new methods of solution of
practical probl ems, studies and sharpens its tool s, and l ooks for the wi dest and
most natural area of appl icabil ity for each new concept and for the most general
conditions of val idity of each proved theorem. All thi s i s a necessity and not
the fri volous pastime of mathematical snobs. Only by establ ishing concepts
and theorems in greatest general ity and by freeing them from unnecessary
restrictions rel ated to the concrete problems from which they arose can one
see the connections between di stant areas of science and learn to use methods
created in one context i n situations which, at frst si ght, have nothing i n common
with that context.
That is why such seemingl y obvious concepts as length, area and vol ume
( l ater to be subsumed under the si ngle concept of measure) were subjected to
detai l ed anal ysi s. One of the frst mathematicians who tried to make preci se
the concept of measure was Jordan. For many decades he presented in Pari s a
course of mathematical anal ysi s based on precise defnitions, fawless proofs,
and rigorous logic. Of course, he couldn' t use in his lectures the vague concept
of area. Hi s defnition of area can be stated as fol l ows: the area of a fgure
is a number between the areas of polygons contai ned in that fgure and the
areas of polygons contai ni ng that fgure. It turned out that the onl y fgures that
have Jordan area are fgures whose boundary has zero area. Unfortunately,
very many fgures have no Jordan area. In particul ar, the nonsquarable fgures
described earl ier have no Jordan area.
Inspi red by Jordan' s lectures, young schol ars tackled the problem. At the
end of the 1 9th century, Emile Borel
2
1 proposed one of the frst defnitions
appl i cable to a very wide cl ass of fgures. He observed that all fgures used
a mathematical hall ofwonders 1 05
in sci ence, regardless of whether they are on a l i ne, i n the plane, or in space,
can be obtained from very si mple fgures - segme,!s, squares, and cubes - by
means of the two operations of formi ng countable l :i ons of sets and taki ng
complements of sets ( i n parti cul ar, as we saw earl ier, al l cl osed sets can be
obtai ned in thi s way). By transfni te al ternation of these operati ons one can
obtai n even more complex sets, cal led B-sets i n Borel ' s honor. (Using Zeno' s
idea, we can obtai n al l such sets i n a fnite ti me i nterval by constantly doubl i ng
the speed of the operations invol ved. )
It turned out that one can assign t o each Borel set a number as i t s measure
by rel yi ng on the fol l owi ng two principles:
J. I a set A is representable as a countable union of disjoint measurable
subsets then its measure is the sum of the measures of these subsets;
2. The measure of the complement of a measurable subset is the dif erence
between the measure of the whole and the measure of that subset.
In parti cul ar, Borel ' s pri nciples i mpl y that the measure of a countable set i s
zero (such a set consi sts of countabl y many poi nts each of which has measure
zero).
Unfortunately, it l ater turned out that Borel ' s process has a seri ous short
comi ng. It i s obvious that the measure of a set must not depend on the way
in which i t i s bui l t up from si mple sets. And Borel was unable to produce the
requi red proof of this i ndependence.
Henri Lebesgue,
22
who began his scientifc work at that ti me, proceeded
di fferently. Before we discuss hi s approach to the measure problem we say
something about hi s earl ier achievements.
Lebesgue' s frst papers angered cl assical mathemati cians. The very title of
one of them, "On unruled devel opable surfaces," seemed to them as outlandi sh
as, say "On gaseous ice" to a physi ci st or "On fshl i ke elephants" to a biologist.
Even bad students knew that a devel opable surface ( a cyl i nder, a cone, and
so on) i s made up of straight l i nes, that i t can be obtained by the motion of a
recti l i near generator. But the source of the trouble was that the young author
thought of developable surfaces i n a way di fferent from that of the cl assi cal
geometers. He regarded as developable not only surfaces obtai ned by careful l y
bending a sheet of paper but al so surfaces obtained by crumpl i ng i t ( i n expl ai ni ng
hi s paper to a fri end Lebesgue sai d: "Imagine a crumpled handkerchief"). He
showed that one can so "crumple" a piece of paper that it doesn' t contain a
single straight-l i ne segment. Of course, the resul ti ng surface is al l fol ds and
breaks, whi ch is why i t was mi ssed by geometers who cl assi fed devel opable
surfaces. They studied only the smooth case.
Next Lebesgue tackled the problem of the area of a non-smooth surface, a
surface that admits no tangent pl anes anywhere. For a crumpled devel opable
surface the problem has a si mple sol uti on: one fattens i t and computes the area
of the resul ti ng pi ece of the plane. But thi s answer can not be obtai ned from the
formul as of cl assical mathemati cs. The l atter appl y only to smooth surfaces.
An attempt to measure the area of a surface by approxi mati ng it by the areas
of i nscri bed polyhedra is bound to fai l . In fact, the German mathemati ci an
Hermann Schwarz
2
3 showed that thi s approach fai l s even in the case of an
ordi nary cyl i nder. Speci fcal l y, Schwarz showed that one can i nscribe in a
cyl i nder a polyhedron wi th so many folds that i ts area i s far greater than that
of the cyl i nder. Lebesgue managed to come up wi th a defni ti on of the area
of a surface that di spensed wi th the need for tangent pl anes and avoided the
di fcul ti es connected wi th the "Schwarz accordi on." Work on this speci al
problem brought Lebesgue to general i deas of the measure of sets and enabled
him to fnd a way of measuri ng the lengths, areas, and vol umes of the strangest
fgures.
Lebesgue took from Borel the i dea of summi ng series and modi fed Jordan' s
defni ti on of measure by usi ng, i n addi ti on to pol ygons, fgures obtai ned by
formi ng countable uni ons of polygons. Speci fcal l y, a fgure is Lebesgue t
coverble i f i t can be covered by a countable uni on of pol ygons such that the
sum of the series of thei r areas is less than t. Now call a set X Lebesgue
measurble i f for every t > 0 it can be represented as a polygon Ae to which
one adds an t-coverabl e set and from which one removes another t-coverable
set. If I A I denotes the measure of a polygon A , then, clearl y, the measure of
the set X must l ie between the numbers I Ae l - t and I Ae l +t . If turns out
that for Lebesgue-measurable sets there always exi sts just one number wi th thi s
property regardless of the choice of t > 0 and of the approxi mati ng polygon
Ae . Thi s number i s cal led the Lebesgue measure of the set X .
The concept of Lebesgue measure turned out to be an unqual i fed success.
It was found t o appl y t o al l sets encountered i n science up t o that ti me. Sub
sequently di scovered i nstances of nonmeasurable sets i nvol ved the use of the
axiom of choice (to be di scussed below) and were not constructive. We can
therefore say that Lebesgue sol ved the problem of measure for all sets encoun
tered in mathematical practice.
By means of his concept of measure Lebesgue could obtain the i ntegral s
of al l di sconti nuous functi ons that could be constructed by the methods of
1 07
that time. Predictably, the i ntegral based on Lebesgue measure i s called the
Lebesgue integral.
The triumph of Lebesgues' ideas brought Gaston Darboux
24
to a change of
heart. Darboux had been one of the leaders of the cl assical mathematici ans.
But at the i nternational mathematical congress that took pl ace i n Rome i n 1 909
Darboux spoke of the fervent and i nqui si ti ve spirit of 20th century mathemati cs,
about science that pursued its i nvestigations in a completel y new area wi th
unexplored perspectives. He stressed that 20th century science does not hesi tate
to probe the very foundations of the constructions that had seemed unshakeable
for so long.
Later, the i deas that resul ted i n the creation of Lebesgue measure and the
Lebesgue i ntegral enabled Kol mogorov to construct an axi omatic theory of
probabi l i ty and Norbert Wiener to defne the concepts of measure and i ntegral
for spaces of functi ons. All who worked with the i deas of measure rel ied on
constructions and theorems that ori gi nated wi th Lebesgue. After the publ icati on
of the works of Kolmogorov just referred to, these i deas began to be widel y used
i n the theory of probabi l i ty and its appl icati ons, and, i n parti cul ar, i n statistical
physi cs. They were also appl i ed i n the study of dynamical systems. As for
theoretical mathematics, the i mportance of the Lebesgue i ntegral cannot be
overesti mated. Sufce it to say that many of its most i mportant resul ts coul d
not even be formul ated wi thout the Lebesgue i ntegral . It was not for nothi ng
that the hymn of "Luzitania," a group of young Soviet mathemati ci ans who i n
the 20s studied the newest results i n the theory of functi ons, procl ai med that
"Our god is Lebesgue, Our idol the i ntegral ." The group "Luzitani a," a name
that honors N. N. Luzi n, founder of the Russian school of functi ons of a real
vari able and teacher of the members of that group, meri ts more than a mere
reference. Indeed, it is difcult to overesti mate its i mportance in the study of
mathematical problems i nvol vi ng i nfnite sets.
Luzitania. The devel opments just described resul ted i n a new theory of
functi ons. Interest i n thi s theory grew i n an extraordi nary manner. In Gotti ngen,
Hi l bert and hi s school used the concept of the Lebesgue i ntegral to study the
ci rcle of problems i nvol vi ng so-called i ntegral equati ons, and Ital i an mathe
maticians proved a number of extremel y i nteresting resul ts.
In the 1 9th century, S1. Petersburg was the acknowl edged center of Russi an
science. Withi n the wal l s of the S1. Petersburg Academy of Science, the
traditions of Euler and Daniel Bernoul l i were pi ousl y preserved. Here l abored
great Russian mathemati ci ans - Ostrogradski ,
25
Chebyshev,
2
6
A. A. Markov,
27
Lyapunov,
2
8
Stekl ov,
2
9
and Korki n. 30 They al l had in common an i ntense i nterest
i n i nvestigati ons deal i ng with mathematical anal ysi s and in the solution of
di fcul t, concrete mathematical problems. The newfangled investigati ons in
set theory and i n the theory of di scontinuous functions did not appeal to them.
These i nvestigations struck them as far removed from the problems they were
studyi ng (although it turned out l ater that results obtained by fol l owing these
newfangled roads were of great use in many traditional areas of mathematics).
Thi ngs were di fferent i n the ol d Russian capital . Here, at the famous
Moscow universi ty, courses in set theory were offered as earl y as the beginning
of the 20th century and i n 1 907 1. 1. Zhegal ki n
3
1 defended a master' s thesis on
transfnite numbers. One of the peopl e who got interested i n this real m of ideas
was the di fferential geometer D. F. Egorov,
3
2
one of the best Moscow mathe
maticians of that time. He was able to prove a theorem about the convergence
of series that became one of the most important tool s in all studies in the theory
of functions. What was most i mportant, however, was that Egorov drew the
attention of hi s young students to this area of research. One of these students,
who was just embarking on a scientifc career, was Luzi n.
At that ti me many schol ars tried to understand the rel ation between the
"wi l d" functions of Di richlet, Riemann, Borel and Lebesgue and the functions
studied by the previous generations of schol ars. Luzin found that by "adjusting"
a di sconti nuous function on a set of arbitrarily smal l measure one can turn it
into a conti nuous function. And a continuous function can be approximated
with arbitrary cl oseness by a polynomial . Thus the most tangled functi ons can,
i n a sense, be reduced to the best known functions, namely polynomi al s.
Luzin al so studied problems connected with trigonometric (Fourier) series,
an issue of abiding interest to speciali sts in the theory of functions of a real
vari able si nce the ti me of Cantor' s earl iest papers. In this area too Luzin proved
a number of extremel y i nteresting results that discl osed some of the subtle
mechani sms that control the convergence of these series. These, and many
other resul ts he had proved, were the basi s of hi s master' s dissertation titled
"Integral s and trigonometric series." The scienti fc merit of hi s thesis was
so great that, notwithstanding the opposition of some mathemati cians of the
cl assical persuasion, he was awarded for it a doctorate i n pure mathematics - a
very rare event in the practice of Russian universities.
Luzi n' s scientifc enthusi asm, the novelty of hi s ideas, and hi s outstanding
teaching abi l i ties attracted to hi m the most talented young mathematicians,
most of whom had been associ ated with hi m si nce their school days. Many
of them obtained outstanding scientifc results even before the completion of
thei r university studies. For many years mathematicians had tried to show that
a mathematical hall of wonders 1 09
if a Fourier series converges to zero al most everywhere (that i s, everywhere
except for a set of measure zero) then its coeffcients must al l be equal to
zero. To everybody' s amazement the student D. E. Men' shov
33 showed that thi s
i s not the case. Hi s counterexample, l i ke many bri l l i ant ones constructed by
Luzin himself, was very intricate. Men' shov' s paper was the frst i n a series
of investigations devoted to the class of problems he di scovered. N. K. Bari,
34
who l ater wrote a splendid book on Fourier series, obtained strong resul ts in thi s
area. In hi s student days Kol mogorov al so worked on a number of problems
in the theory of Fourier series. He constructed a remarkable example of a
Lebesgue-integrable function whose Fourier series diverges everywhere.
In another di rection, Luzi n' s students worked on the structure of Borel sets.
To show that a Borel set is ei ther countable or contai ns a subset of the cardi nality
of the conti nuum, Pavel Sergeevich Aleksandrov, while sti l l a student, i nvented
an extremel y subtle construction (called i n his honor the A-operation) that can
be used to obtain any such set. Some ti me l ater, M. Va. Susl i n, 3
5
another young
student of Luzi n' s, showed that the A-operation can be used to obtai n certai n
sets that are not Borel sets. Thi s gave rise to the problem of describing thi s cl ass
of sets now known as Susl i n sets. Unhappi l y, Susl i n' s premature death from
typhus in 1 9 1 9 put an end to hi s research. It was conti nued by Luzi n, who was
joi ned by P. S. Novikov
36
and Liudmi l a Kel dysh. 3
7
Thei r resul ts gave rise to a
new branch of mathematics known as descriptive set theor. Further work i n
thi s area touched on the very core of the foundati ons of set theory and showed
the bounds of set-theoretic thought. Many problems that are now sol ved were
posed by Luzin and the results obtained confrm hi s remarkable i ntui ti on.
The work of Luzi n and hi s students made Moscow the universal l y acknowl
edged center of research in the area of functions of a real vari abl e. Thi s state
of affai rs was not affected by either World War I, or the ci vi l war, or the i n
tervention, or the blockade. In Pol and, Luzi n' s i deas were devel oped by W.
Sierpi nski ,
38
who l i ved in Moscow during W.W.I and associ ated with Luzi n.
We have menti oned onl y some of Luzi n' s many students. Many of them
have achieved i nternational fame.
Due to thei r i ntense preoccupation with problems of set theory and the
theory of functions of a real variable, members of the Luzitani a group tended at
times to underestimate the i mportance of cl assical di rections in mathemati cs.
But the scientifc interests of many of them shifted l ater to areas much cl oser
to practical problems. For example, as menti oned earl ier, Kol mogorov used
the ideas of Lebesgue measure in the theory of probabi l ity and then studied its
practical applications. Si mi l arl y, the outstandi ng appl ied mathemati ci an M. A.
1 1 0
In Search of Infnit
/ II
Figure 30
Lavrent ' ev3
9
studied the subtlest i ssues of set theory i n hi s youth.
The great irrigation project. It woul d take a book on the hi story of Soviet
mathemati cs to describe the areas of mathematics pursued by former members
of the Luzi tani a group. Thi s bei ng so, we here describe only some of the uses
of ideas from the theory of i nfni te sets in topology, the branch that i nvestigates
the properti es of fgures unaffected by very general transformations. All that i s
requi red i s that these transformations be one-to-one and that they not i nvol ve
teari ng or gl ui ng together.
Ini ti al l y, topol ogi sts studied only fgures composed of a fnite number of
very si mple fgures cal l ed si mplexes (points, segments, tri angles, tetrahedrons).
But l ater they appl ied thei r ideas to more complex sets. And here i t became
clear that geometric i ntui ti on, so faithful a servant in the past, gave i ncorrect
answers to many questi ons.
The Dutch mathemati ci an Brouwer
4
o
constructed a number of amazi ng
exampl es of curves and pl ane domai ns. We wi l l now present one of the most
surpri si ng of these exampl es.
Let us draw the map of some country and the countries contiguous to i t.
Al most every point of the boundary of thi s country bel ongs to two and only two
countries: the given one and one of its neighbors. At each of these points there
are two border guards. On the map there are some poi nts where three countries
come together (Fi gure 30). Three border guards stand at such points. But there
is onl y a fni te number of such pl aces on the map. And it seems quite obvious
that such poi nts could not occupy the whole boundary of a country, so that
there could not be three domai ns (three countries) shari ng the same boundary.
In other words, it seems obvious that three border guards from three di fferent
countries wi l l not be standi ng at every point of the boundary.
I I I
Figure 3 1
But Brouwer constructed three such domai ns. In order to understand hi s
example, i magine an i sl and i n the ocean on which there are two l akes wi th
fresh water. One l ake i s cold and the other i s warm. Now we shal 1 carry out the
fol 1 owi ng i rrigation project. During the frst day we construct canals leadi ng
from the ocean and from both l akes i n such a way that each canal i s "bl i nd"
(i . e. , is only a creek of the corresponding reservoir), and that the canal s nowhere
touch one another, so that when we have fni shed, each point of dry l and is at
a di stance of less than I ki l ometer from sea water and from the water of both
l akes (Figure 3 1 ) .
During the fol 1 owi ng half day we extend these canals in such a way that
they remai n "bli nd" as before and do not touch one another, and so that now the
di stance from any point of dry l and to any of the three canal s is less than |
ki l ometer. In doing thi s, of course, the canals have to be made narrower than
they were before. In the fol lowing quarter day we carry on, arrangi ng matters
so that each point of dry l and is less than ki l ometer from any canal , etc. As
we continue the process, the canal s become ever more wi ndi ng and narrower.
After two days' work the enti re i sl and wi l 1 be permeated by these three canal s
and converted i nto a Cantor curve. No matter what point of the curve we stand
on, we can scoop up, accordi ng to our whim, salt water or warm or cold fresh
water. And things are so arranged that the waters do not mi x wi th one another.
If we repl aced the ocean and l akes by three countries, we would obtai n the
unusual map we spoke of at the begi nni ng - three border guards, one from each
country, woul d be pl aced at each poi nt of the boundary.
1 1 2
In Search of Infnit
A "nondissertable" subject. Cantor' s defni tion had one fauI t- i t was not
at al l suitable for curves i n space. And what i s a surface in space? No one
knew. Thi s probl em - to determine what curves and surfaces i n space are -
was put in the summer of 1 92 1 to hi s 23-year-old student P. S. Urysohn4 1 by the
venerable Professor D. E Yegorov of Moscow Uni versity. (It is evident that he
gave more thought to the mathematical si gni fcance of the problem than to its
"dissertabi l i ty" - thi s problem was one of the hardest ! ) .
Urysohn qui ckl y comprehended that Yegorov' s problem i s only a speci al
case of a much more general problem: what is the di mension of a geometric
fgure, i . e. , what are the characteristics of the fgure which cause us to say
that a segment or a ci rcle has di mension I , a square has di mensi on 2, and a
cube or a bal l has di mension 3? Thi s period in the l i fe of P. S. Urysohn was
recalled by his cl osest fri end, a young doctoral candi date in those days and now
an academi ci an, the honorary presi dent of the Moscow Mathematical Society,
PS. Al eksandrov, who wrote: " . . . the whole summer of 1 92 1 was spent in
tryi ng to fnd an ' up-to-date' defnition (of di mension) ; P. S. shifted hi s i nterest
from one vari ant to another, constantl y setti ng up examples showi ng why thi s
or that vari ant had to be el i mi nated. He spent two months total l y absorbed
in hi s meditati ons. At l ast, one morni ng near the end of August, PS. awoke
with hi s now wel l -known i nductive defnition of di mension in its fnal form
. . . . That very morni ng, whi l e we were bathi ng in the Kl ayz' ma, P. S. Urysohn
told me about his defnition of di mension and there, during a conversation that
extended over several hours, outl i ned a pl an for a complete theory of di mension
composed of a series of theorems, which were then hypotheses that he did not
yet know how to go about proving and which were l ater proved one after another
in the months that fol l owed. I never again ei ther partici pated in or wi tnessed a
mathematical conversation composed of such a dense fow of new ideas as the
conversati on of that August morni ng. The whole program outl i ned then was
real i zed duri ng the wi nter of 1 92 1 /22; by the spri ng of 1 922 the whole theory
of di mension was ready . . . . ".
The basic idea of Urysohn' s defni ti on of di mension consi sted of the fol
l owi ng. Two or perhaps several poi nts usual l y suffce for separati ng a portion
of a curve from the remai nder (the part of the four-leafed rose of Fi gure 32
contai ni ng the center can be separated from the rest of the curve by using eight
poi nts) . But i t is al ready i mpossi ble to separate a part of a surace from the
remai nder by removing several points - for thi s you have to take a complete
curve - no matter how many poi nts you take on the surface, it is always possible
to go around them. In the same way a surface i s needed to separate a part of
1 1 3
Figure 32
three-di mensi onal space from the rest of the space.
All thi s sti l l had to be made more preci se: for some curves an i nfnite set of
points has to be removed i n order to separate some part, but the totality of these
points does not form a curve. U rysohn succeeded i n gi vi ng a preci se formul ati on
of al l the defnitions requi red. In a way hi s defni ti ons are remi ni scent of those
of Eucl id (the ends of a curve are poi nts, the ends of a surface are curves. ) But
thi s resembl ance i s somethi ng l i ke the one between the hol l owed-out tree trunk
of pri mitive man and a modern ocean l i ner.
We wi l l make these defni ti ons preci se.
Let A be a subset of a set X . A point p of X i s called a boundar
point of A if there are points of A and points of X not in A that are
arbi trari ly close to p . The boundar of A is the set of al l i ts boundary
points. For example, the boundary of a square in the pl ane i s its usual boundary
and the boundary of that square i n space is the square itself.
A set A is open in X if it contai ns none of its boundary points in X .
An example of an open set is the i nteri or of a di sk in the pl ane.
A set X has dimension 0 i f each of its points can be contained i n an
arbitrari l y smal l set whose boundary in X is empty. Examples of such sets
are any fnite set, the points with rati onal coordinates on a straight l i ne, the
Cantor set, and so on.
A set X has dimension 1 i f i t is not O-di mensi onal and each of its points
can be encl osed i n an arbi trari l y small set open i n X whose boundary in X
is O-di mensi onal .
It turned out that not only al l the ordi nary curves (ci rcle, l i ne segment,
1 1 4 In Search of Infnit
Figure 33
el l ipse, etc. ) but al so al l Cantor curves have di mension 1 i n Urysohn' s sense.
Thus, i t now became possible to defne the notion of a curve i n space as wel l
as in the pl ane, namely, a curve is a continuum of dimension 1.
And i t was also clear how t o defne a surface, a 3-dimensional solid, and,
i n general , a set of any di mension. Si nce the defnition proceeds by numerical
order, frst defni ng a set of di mension 0, then a set of dimension 1 , then of
di mensi on 2, etc. , Urysohn' s defnition i s called inductive.
The article should be printed, not reviewed! Urysohn proved many very
i nteresting theorems rel ating to the notion of dimension that he introduced.
But he was unable to fnd a way to prove one very important theorem; he coul d
not prove that an ordi nary cube has di mension 3. After prolonged effort he
found a remarkable way out of the difculty, conceiving a new defnition of
dimension i n the process. We shal l not discuss thi s defnition i n detai l , but shal l
si mpl y i l l ustrate it with very si mple fgures.
If we take a segment or a ci rcle, we can divide i t into arbitrari l y small pieces
in such a way that each point belongs to at most two pieces (Figure 33). Here
we take the pieces together with thei r boundaries ( i . e. , thei r endpoints). But a
square cannot be divided this way. It seems at frst gl ance that if we divide a
square i nto pieces, there wi l l always be points belonging to four pieces (Figure
34a). But if we pl ace the pieces the way bricks are laid in construction, then
each point belongs to at most three di fferent pieces (Figure 34b). In the same
way, we can di vi de a cube into small paral lelopipeds in such a way that each
Chapter 3. Remarkable functions and cures, or a stroll through
1 1 5
(a) (b)
Figure 34
point belongs to at most four parallel opipeds.
Thi s i s the property that Urysohn took for hi s new defni ti on of di mensi on.
A fgure X i s said to have dimension n , i f n i s the smal lest nonnegati ve
i nteger for which the fol l owi ng hol ds: X can be di vi ded i nto arbitrari l y smal l
cl osed pieces i n such a way that each of its poi nts belongs to at most n + 1
*
different pieces.
Employi ng this defni ti on of di mensi on, Urysohn proved that the di mensi on
of a square i s 2, the di mension of a cube i s 3 , etc. And he then proved that
thi s defnition is equivalent to hi s earl ier one.
Urysohn' s theory of di mension made a great i mpression on the mathemati
cal worl d. This i s vi vi dl y shown by the fol l owi ng epi sode. Duri ng a tri p abroad
Urysohn gave a report on his results i n Gotti ngen. Before the rise of the Fascists
to power the University of Gotti ngen was one of the leadi ng mathematical cen
ters. After the report, the head of the Gotti ngen school of mathemati cs, Davi d
Hi l bert, sai d that the results shoul d be publ i shed in Mathematische Annalen -
one of the most respected mathematical journal s of the ti me. A few months
l ater, Urysohn again gave a report at Gotti ngen, and Hi l bert asked the editor of
Mathematische Annalen, Richard Courant, whether Urysohn' s arti cl e had been
printed yet. The l atter repl ied that the article was bei ng reviewed. "But I cl earl y
stated that it was to be printed, not revi ewed! " Hi l bert exclai med. Afer such
"
This defnItion of di mension goes back to Lebesgue and Brouwer. (Trans. )
an unequivocal decl aration the article was soon pri nted.
For the next three years Urysohn carried on mathematical research un
equaled i n depth and i ntensi ty (during this ti me he publ i shed several dozen
articles) . A tragic accident abruptly ended his l i fe - he drowned August 1 7,
1 924 whi l e swi mmi ng in the Bay of Bi scay during a storm. He completed hi s
l ast mathematical article the day before hi s death.
After Urysohn' s death there sti l I remai ned numerous rough drafts and out
l i nes of unpubli shed resul ts. His cl osest friend (and co-author of many of hi s
articles) P. S. Aleksandrov interrupted hi s own studies for a time and prepared
these articles for publ ication, thus making these additional results of Urysohn
avai l able to all mathematicians. The theory of di mension at present constitutes
an i mportant chapter of mathematics.
Chapter 4
n scarch 0l lhc abs0lulc
New complications. The success attai ned i n the study of functions and curves
with the help of set theory made i t a full and equal member of the fami l y of math
ematical sciences. Thi s recognition was acknowledged at the frst i nternati onal
congress of mathematici ans in Zurich in 1 897. Hurwitzl and Hadamard,
2
the greatest experts in mathematical analysi s, demonstrated in thei r lectures
extremely varied appl ications of sets and di sclosed their connection wi th the
theory of so-cal led analytic functions. Three years l ater, at the next i nterna
tional mathematical congress, David Hi lbert ' s l i st of 23 of the most i mportant
unsolved mathematical problems i ncluded problems i n set theory. In his lecture
at the congress Henri Poincare gave a high rating to Cantor' s works. Speak
ing of the rol e of intuition and logic i n mathematics, he said that mathematics
fnds i n set theory an absolutely permanent and rel i able foundati on, and now
al l that remai ns are the natural numbers and fnite or i nfni te systems of such
numbers. In hi s view, mathematics had become completel y ari thmetized and,
fnally, absol utel y rigorous.
I n view of this high appraisal of set theory by the leading mathematician of
the ti me i t i s not surpri si ng that academic honors rai ned upon its creator, Georg
Cantor. He was made an honorary member of the London Royal Society, a
corresponding member of the Institute of Science, Literature and the Arts i n
Venice, honorary Doctor of Mathematics of the University of Chri sti ani a (now
Oslo), and so on.
But every fami l y has its skeleton in the cl oset. As for mathematics, duri ng
the many mi l l enni a of i ts evolution, the skeleton that woul d pop out of the
cl oset at the most i nopportune moments was the i nconsi stency i mpl i ci t in the
very concept of the i nfnite. From the moment when Zeno recognized thi s i n
consi stency, repeated attempts had been made to restore everythi ng to normal
and to make the cl oset more secure. After the frst of these attempts by Eu
doxus and Eucl i d, two mi l lenni a passed before Weierstrass and Cantor made
the second attempt. We saw that the best mathematicians of that ti me thought
that the attempt had been compl etely successful . But this time the "skeleton"
turned out to be extremel y disturbed and popped out of the cl oset afer sl ightly
more than two decades. David Hi l bert put the matter in these words:
What took pl ace was si mi l ar to what happened in the evolution of
the i nfnitesi mal calcul us. Overjoyed by new and rich resul ts, one
openl y adopted an i nsufciently critical atti tude concerni ng the
l awful ness of i nferences. As a result, contradictions arose, at frst
rare then progressivel y more serious, already in connection with the
basic formation of concepts and the use of inferences that gradual l y
became routine . . . Cantor' s doctrine was violently attacked from
many di recti ons. The countermovement was so i mpetuous that the
most commonly used and most fruitful concepts of mathemati cs,
its si mplest and most i mportant i nferences were jeopardized and
their use was forbidden.
The frst sign of trouble in the very foundations of set theory was the paradox,
frst di scovered by Cantor hi mself in 1 895 and publ i shed two years l ater by
the Ital ian mathematician Bural i -Forti . 3 At i ssue was the set of all transfnite
numbers. From its defni tion, this was no worse than any other set, for it seemed
to many to be conceivable as a si ngle object. But this set turned out to have
an essential faw. Because it was-wel l ordered it was describable by some
transfni te number Q. But then Q must be greater than al l transfnite numbers,
and thus greater than itself, which i s obviousl y i mpossi bl e.
Later it became cl ear that the set of all sets also involves a contradicti on.
After al l , thi s set must contai n al l of i ts subsets. But thi s i s i mpossible because
the cardinal ity of the set of subsets of any set i s greater than the cardinal ity of
that set (see p. 67) .
Another remarkable example of a speci fc inconsi stent set was publi shed
i n 1 903 by Bertrand Russel 1 . 4 In general , sets do not contain themsel ves as
el ements (for example, the set of natural numbers is not a natural number, the
set of all triangles is not a triangle, etc. ) . But there are sets that do contain
themsel ves as elements. For example, the set of all abstract notions is an
abstract notion (don' t you agree?). Because such sets are rare we' l l cal l them
extraordinar, and all other sets ordinar.
Now l et A be the set of all ordinary sets. At frst si ght there is nothing wrong
with this defni tion; it is not clear why the phrase "the set of all ordinary sets"
Chapter 4. In search of the absolute 1 1 9
is worse than the phrase "the set of al l triangles." But, in fact, we are faced here
with a serious contradi ction. Let' s try to determi ne whether A is ordi nary or
extraordi nary. If it is ordi nary then it contai ns itself as an element (remember
that A i s the set of al l ordinary sets) . But then, by defni ti on, i t i s extraordi nary.
A contradiction. If i t i s extraordi nary, then i t must contain itself as an element.
But the elements of A are i nvari abl y ordi nary sets. Again a contradi cti on. Thus
the set A i s neither ordi nary nor extraordi nary.
Such logical contradictions arise in much si mpler cases. For example, a
soldier was ordered to shave those soldi ers and onl y those soldi ers of hi s pl atoon
who did not shave themsel ves. The question arose whether he shoul d shave
hi msel f. If he shaved hi msel f, then he woul d be among the group of soldiers
who shaved themsel ves, but he doesn' t have the right to shave those soldi ers.
If he doesn' t shave hi mself, then he belongs to the cl ass of soldiers who do not
shave themsel ves, but then accordi ng to the order he has been gi ven he must
shave hi mself.
There are other wel l -known examples of sets which at frst gl ance appear
to be wel l defned, but turn out on cl oser i nspection to be very poorl y defned,
and we would be better off sayi ng that these sets are not defned at al l . For
example, let A be the set of rational numbers which can be defned wi th the
aid of at most two hundred Engl i sh words (here we i ncl ude the words "zero,"
"one," "two," etc. ) . Si nce the set of all Engl ish words is fni te (for si mpl i ci ty
we may assume that we onl y choose words found in Webster' s di cti onary and
their grammatical derivati ves), the set of all such numbers i s fnite. Suppose
that the numbers in question are rl , r
2
, . . . , rN . Consi der the rati onal number
where n, i s I i f the i th deci mal di gi t of r, i s different from I and 2 otherwi se.
The number r di ffers from rl i n the frst decimal di gi t, from r
2
i n the second
deci mal di gi t, and so on. Therefore r does not belong to the set A. But r
must belong to the set A, because we used less than two hundred words i n its
defni tion.
Thi s paradox i s cl osely rel ated to the fol l owi ng one:
What is the smallest integer that cannot be defned by means of a sentence
having less than two hundred English words ?
Such a number exi sts since the number of words in the Engl i sh l anguage i s
fni te; so there must be a number that cannot be defned by means of a sentence
1 20
In Search of Infnit
havi ng less than two hundred words. And, of course, among these numbers
there woul d have to be a smallest.
On the other hand, thi s number cannot exi st, si nce its defnition i nvol ves a
contradiction. Indeed, thi s number is defned by the sentence written above i n
ital i cs, which we see contains l ess than two hundred words; whi l e according to
its defnition thi s number cannot be determi ned by such a sentence.
Here is a more complicated example of a fnite set of which we cannot
deci de whether or not it contains a particular element.
We separate al l Engl i sh adjectives into two cl asses. In the frst cl ass we put
every adjective which has the property it describes. In the second class we put
all the remai ni ng adjectives. For example, "Engl i sh" belongs i n the frst cl ass
and "German" i n the second. Agai n, "hissing" bel ongs i n the frst cl ass and
"blue" i n the second, as does "monosyl l abic."
To distinguish one cl ass from the other we i ntroduce two more adjectives.
The adjectives i n the frst cl ass can al l be sai d to be autologous whereas the
adjectives in the second cl ass are all heterologous (auto = self, hetero = other,
logos = sense, law). Clearly, the word autologous belongs i n the frst cl ass.
But what of heterlogous? We leave i t to the reader to see that thi s adjective
belongs to nei ther of the two cl asses.
These and si mi l ar paradoxes can be traced back to the ancient "l i ar" paradox,
ascribed to the Greek phil osopher Eubul ides of Mi letus. It i s to the effect that
a person who says "I am l yi ng" is neither tel l i ng the truth nor l yi ng. (Indeed,
i f he is tel l i ng the truth then he is l yi ng. But i f he is l yi ng then he is tel l i ng the
truth. )
Of course, it woul d be si mplest t o say that i n set theory one does not con
si der such whi msical sets. But not giving an exact defnition of which sets can
be considered and which must be rejected i s i nviting trouble.
Somehow one problem does not work out. The di scovery of paradoxes i n set
theory made a profound i mpression on mathemati ci ans. While, for example,
Poincare had only good things to say about Cantor' s theory before the publ i
cation of the set-theoretic paradoxes, he aferwards spoke of it deri sively and
maintai ned that the actual l y i nfni te does not exist. After receiving Russel l ' s
letter with a description of hi s paradox, Frege, whose book was just coming
off the press, was forced to incl ude i n it a remark that, i n essence, amounted to
crossi ng out all of its content. The discovery of the paradoxes had an especi al l y
painful effect on Cantor. He i mmersed hi msel f i n thoughts of how to el i mi nate
them. He fai led, became seriousl y i l l , and gave up scientifc work many years
Chapter 4. In searh of the absolute
before his death. Hi l bert excl aimed:
One must admit that the state we are i n now vi s-a-vi s the paradoxes
i s i n the long run unendurable. Just think of it: i n mathematics, thi s
standard of trustworthi ness and truth - the forming of concepts and
of i nferences, as learned, taught, and used by al l of us, can l ead to
nonsense. Where i s one to fnd rel iabi l ity and truth i f mathematical
thought itself can fai l ?
1 2 1
But i n view of the triumphant successes of mathematical anal ysi s based
on set-theoretic conceptions, most mathematicians were i ni ti al l y completel y
untroubled by the di scovery of paradoxes i n Cantor' s theory. If there was a
certain unease, then it pertai ned to the "most distant" areas of mathemati cs.
What was far more troublesome was the emergence of probl ems i n set
theory that defed solution for a very long time. Many of these problems had to
do with the chasm between the di screte and the continuous, between arithmetic
and geometry - a chasm that seemed to have been surmounted for al l time. Of
the two principal cardinal ities i nvestigated by Cantor - the countable and the
conti nual - one derived from arithmetic, from the notion of natural number,
and the other from conti nui ty, from the notion of a continuum. The question
arose how these two cardinal ities are rel ated.
It was natural to assume that the cardinal ity of the conti nuum is "next"
to the countable cardi nal ity - that i s, that there is no uncountable set whose
cardinal ity i s less than that of the conti nuum. If thi s hypothesis were true, then
the cardinal ity of the conti nuum woul d be the frst transfni te number after all
transfnite numbers of the countable type. The cardinal ity of the set of countable
transfnite numbers was cal l ed aleph one, and so the question was whether or
not the cardinal ity of the conti nuum i s aleph one. That it is aleph one i s the
so-cal led continuum hypothesis, and Cantor pondered over it for many years.
Many ti mes he thought that he had establ i shed it but in the end the "solution"
turned out to have been an i l l usi on.
Nor were other schol ars more successful i n thei r attempts to prove or di s
prove the conti nuum hypothesi s. It was number one on Hi l bert' s l i st of prob
lems. Luzi n pondered it for many years, but i n his case too the solution behaved
l i ke a mirage i n the desert.
One day a ffeen-year-ol d boy, Lev Shni relman
5
was brought i n to Luzi n.
He was said t o possess exceptional mathematical abi l ity - and indeed, he later
became one of the most promi nent Soviet mathematici ans and a correspondi ng
member of the Academy of Sciences of the USSR. In order to test the capabi l i
ties of thi s young mathematician Luzin proposed that he solve thi rty extremely
di fcul t probl ems. He was able to do 29 of the problems. The si ngle unsol ved
probl em was the conti nuum problem. At the end of a month the young math
ematician came back to Luzin and sadl y told hi m: "Somehow one problem
doesn' t work out."
Havi ng despai red of ever resol ving the conti nuum hypothesi s, Luzin used to
tel 1 hi s cl osest students that he didn' t know why the cardinality of the conti nuum
must be aleph one. "Who knows" - he qui pped sadly - "maybe it ' l 1 turn out to
be equal to aleph seventeen."
Then came other problems which also resi sted solution in the framework of
the usual set theory. They included general i zations and vari ants of the contin
uum hypothesi s as wel 1 as other propositions that could nei ther be proved nor
di sproved.
A bafing axiom. The fai l ure to prove the conti nuum hypothesi s gave rise to
the question of whether the set of points of an i nterval can be wel 1 -ordered. Put
differentl y, the question was whether or not the scale of transfnite numbers can
accommodate the conti nuum. Cantor was sure that it can. He was convi nced
that every set can be wel 1 -ordered. But i n this case too he fai l ed to fnd a
promi si ng approach to the solution of the problem.
An unexpectedl y si mple and short solution was publ i shed i n 1 90 by Zer
mel 0
6
- he was able to show that any set can be wel 1 -ordered. However,
Zermel o' s proof did not satisfy a11 mathematicians. The problem was that the
proof depended on one assumption which appeared to be far from obvious to
both its author and others. This statement came to be cal 1ed the axiom of choice
or Zerelo 's axiom and can be i J I ustrated as fol 1ows:
Suppose that i n front of you l i e several piles of appl es. It i s obvious that you
can select an apple from each pile and put them i n a new pile. It woul d also seem
to be true that the same can be done i f there are i nfni tel y many apples i n each
pi l e as wel 1 as i nfnitely many pi l es. Thi s is what consti tutes the axiom of choice:
If an infnite set of infnite sets is given, then it is possible to choose one
element from each set without giving the rule of choice in advance.
Indeed, al1 the trouble arises from these l ast words - the axiom of choice
leads to completely nonconstructive proofs: with it you can prove, for example,
that every set can be wel 1 -ordered, but it does not give any i nformation about
how to go about it.
Mathematicians employed the axiom of choice for many years, consideri ng
Chapte r 4. In search of the absolute
1 23
it to be completel y obvi ous. But when they began to refect on it more deeply,
it came to appear more and more probl ematic. Many of the theorems proved
with the hel p of the axi om of choice completel y contradict our mathematical
i ntui ti on. Thi s led Bertrand Russell to speak of thi s axi om as fol l ows:
"At frst i t seems obvi ous, but the more you think about i t, the stranger the
deductions from thi s axiom seem to become; in the end you cease to understand
what i s meant by i t."
Luzi n admitted: "I ponder Zermel o' s axi om day and night. If onl y someone
knew what ki nd of thing i t i s."
At thi s poi nt one reluctantly remembers the words of Mephi sto in Goethe' s
"Faust":
To try to understand it is a wasted effort,
Wise man and fool al i ke
Are confused
By the sorry mass of contradictions.
True, thi s axiom di d not lead to contradictions but one could obtain from i t
as many i ncomprehensible consequences as one wi shed.
To apples from one. Let us talk about one of the most surpri si ng conse
quences of the axiom of choice. Probably everyone has seen a cl ever magi ci an
at work on the stage. Fi rst he shows the spectators an empty sack, then he
drops a ball into the sack, only to draw out . . . two; droppi ng in the two bal l s,
be pul l s out four; dropping i n the four, he pul l s out eight. Of course, everyone
knows that i t is no mi racle, but is si mply "slei ght of hand." Such mi racles can,
however, happen i n the theory of sets.
We take an ordi nary apple and divide i t in any way into four pieces. It seems
clear that i f we take only two of the pieces, it wi l l not be possi ble to form an
enti re appl e from them (i n the same way, if you have eaten half an orange, you
cannot form an enti re orange from the remai ni ng sl ices).
However, mathematicians can di vi de a bal l i nto fve parts i n such a way that
an entire bal l of the same radi us can be formed from two of the parts, wi thout
supplementing them i n any way, si mply by transl ati ng them as rigid bodi es.
A second, identical bal l can be formed from the other three parts. Thus, we
can obtain two di sti nct bal l s from the one. It i s a pi ty that thi s problem i s onl y
capable of bei ng sol ved in theory, otherwise we could make two apples from
one, then four, then eight, etc. Of course, the problem cannot be sol ved in the
real world - it woul d contradict the l aw of conservation of matter.
But mathematics does not i nvestigate the i mmedi ate materi al world. Rather,
it i nvesti gates mathematical model s of thi s world. Therefore, if one obtains
resul ts that contradict physical intuition, then this i s not the faul t of mathematics
but the consequence of the wrong choice of model .
Thus our strange resul t, just l i ke some other paradoxical consequences of
the axi om of choice, shows only that thi s axiom must be treated with care.
Some mathematicians try to separate sharpl y assertions derived with the help
of thi s axiom from the others. Dangerous or not, some theorems of mathemat
i cal analysi s cannot be proved wi thout rel i ance on thi s fateful axiom.
A quintet of demons. One of the reasons why the greatest mathematicians
refused to bel ieve i n the possi bi l ity of wel l -ordering the conti nuum was the
absence of even a hi nt of a transparent construction for such a well -orderi ng.
In thi s connection there ensued a spi rited di scussion of the meani ng of the term
"there exi sts" i n mathematics. Does thi s expression mean that an appropri ate
mathematical object admits a defni te construction, or can one also consider sets
that exist only on the strength of the axiom of choice? How meaningful is the
concept of the set of all subsets of the conti nuum i f we cannot constructively
describe most of these subsets? Afer a certain "period of i ncubation" the
"sickness" came to l i ght, and i n 1 905 the most famous French mathematicians
(Hadamard, Borel , Bai re,
7
Lebesgue) publi shed thei r correspondence on the
meani ng of i nfnity and on the question of which i nfnite sets shoul d be regarded
as havi ng exi stence. These questions also preoccupied Hi l bert and the young
Dutch mathemati ci an Brouwer.
The participants in the debate disagreed profoundly with one another. In
Hil bert' s vi ew, the debate was sharp because, begi nni ng with the di stant past,
no problem troubl ed the human mind as deepl y as that of the nature of the
i nfni te. Al most no other idea i nfuenced human reason i n so sti mul ating and
fruitful a manner as the idea of the i nfnite, and, i n Hi l bert' s vi ew, no other idea
requi red cl ari fcation so urgentl y.
To give the reader an i dea of the viewpoints of the participants in thi s heated
debate (smal l wonder the debate was heated - afer all these were mathematical
Ol ympi ans) we quote a bri l l i ant account taken from Luzi n' s book The present
state of the theor offunctions of a real variable. In it he makes use of vari ants
*
of "Maxwell ' s demon." Luzin wrote:
*
Maxwel l ' s demon made heat fow from a cold to a warm region by opening and cl osing a
Chapter 4. In search of the absolute
If we analyze the views of the creators of the modern theory of
functi ons, then we can easily notice that each of them adopts i n the
course of his work a defni te conception of what is possible and
admi ssi bl e. Its bounds are the bounds of mathematics and beyond
them begi ns a domain that lies - to use Borel ' s term - outside
mathematics . . . If, fol l owi ng Maxwel l ' s example, we assign each
author' s domain of the possible and real izable to the appropriate
imaginary creature, then we get the fol lowi ng scheme.
1 . Brouwer' s "demon." Hi s domain i s the fnite with a speci fed
fni te upper bound. Al l beyond this domain is "outside mathemat
i cs."
2. Bai re' s "demon." His domain is si mpl y the domain of the
fnite without the prescription of a fni te upper bound. The i nfnite
i s merely afaon de parler and i s "outside mathematics."
3. Borel ' s "demon." Hi s domain is the domain of the countabl y
i nfnite. Every uncountable set i s "outside mathematics."
4. Lebesgue' s "demon." Hi s domain is the domain of the cardi
nal ity of the conti nuum. Every operation that requi res a conti nuum
of si mple steps i s avai l able to thi s "demon." . . .
5. Zermel o' s "demon." Hi s feld of operations extends to al l
cardi nal ities. In particul ar, Zermel o' s demon can well -order every
set.
1 25
The fol lowing example i l l ustrates the di fference between the respective
strengths of the Borel demon and of the Lebesgue demon. Suppose that we
wish to fnd out if the inequali ty x :: a holds for all elements of a set X. If
X i s countable, then Borel ' s demon can sol ve the problem because he must
verify just countably many i nequal ities Xk :: a. If X has the cardi nal ity of the
conti nuum, then Borel ' s demon can' t solve the problem but Lebesgue' s demon
can.
Luzi n' s own view of these matters was ambiguous. For the most part he
shared Borel ' s position, noting that "the concept of an uncountable i nfnity i s
purel y negative and has no objective reality whatsoever; i t i s a concept due onl y
to man' s abi l ity to devise proofs ' by contradiction' and does not correspond to
any attai nable real ity . . . " At other times he tended more to Bai re' s viewpoint
and maintained that we l ack a sufciently clear conception of the actual i nfni te
even though this concept can be defned in terms of abstract logic. To the Pol i sh
hole i n a di aphragm that divided a gas-fl l ed vessel i nto t wo halves A and B. The demon would
open the hole to let a fast molecule pass from A to B and cl ose i t otherwise. (Tanslator)
mathemati ci an Kuratowski
8
he wrote: "No matter what, I cannot consider the
set of positive i ntegers as gi ven, for the very concept of the actual i nfni te
strikes me as i nsufciently natural to consi der it by itself." And further: "The
fundamental problem i s to expl ain whether or not the sequence of positive
i ntegers is completel y objective. It seems that it is al most objective and that
there are [ i n it] traces of unquesti onable subjecti vi ty, so that i t is i mpossible
to speak of the sequence of positive i ntegers at al l ti mes, in all cases, in one
and the same sense." He thought that i t was too early to formul ate the burni ng
problem of the uni queness of the sequence of positive i ntegers and to speak of
fni te numbers that are i naccessible i f we start wi th I .
Bourbaki ' S9 viewpoint of the essence of mathematical problems i s the very
opposite of Luzi n ' so He cl ai ms that whatever is free of contradictions exi sts, and
so shares Zermel o' s posi ti on. Thi s means that he admits arbi trary cardi nali ti es,
and accepts wi thout reservation the axi om of choi ce and al l its consequences,
i ncl udi ng the paradoxical subdivi sion of the sphere and the cl ai m of the wel l
orderedness of the conti nuum. He seems to take no i nterest whatever in the
question of the possi bl e appl icabi l i ty of such a mathematics to the problem of
knowi ng the real world.
P. S. Al eksandrov al so approved of worki ng with sets of arbitrary cardinal ity.
For example, he extended the concept of dimension to a very large set of spaces
that sati sfy no countabi l i ty conditions, devel oped geometry i n such spaces, and
so on. Thus Zermel o' s demon enables one to obtai n extremel y beautiful results
but, on the other hand, leads to asserti ons whose meani ng cannot be grasped
by i ntui ti on.
What compl icates the choice of one of the fve demons l i sted earl ier is that
the "unpleasantnesses" that arise i n connection wi th sets of arbitrary cardi nal i ty
can be modeled in countable sets. It fol l ows that the bl ame for the compl ica
ti ons that arise in mathematics must be laid at the door not of sets of too high
cardi nal ity but of the i dea of the actual i nfni te.
Banishing the infnite. David Hi l bert embarked on a bold and profound attempt
to resolve the di fcul ties of the theory of i nfni te sets. He would not give up the
achievements of thi s theory (he put it in a colorful phrase: "No one wi l l banish
us from the paradi se created for us by Cantor") . In hi s work On the infnite
Hi l bert noted that whi l e the i nfni tel y small and i nfnitely l arge had once been
removed from mathemati cal anal ysi s, the i nfni te managed to el bow its way
back i n, frst through the i nfnite sequences used to defne real numbers and
then in the real numbers thought of as a compl eted set.
1 27
Weierstrass had reduced the concepts of the i nfnitely smal l and the i nfnitel y
large to inequalities invol vi ng fni te magnitudes. Si mi l arly, Hi l bert wanted to
el i mi nate i nfni te sets from mathematics. In hi s view, the use of such sets i n
mathematical arguments shoul d be regarded as a fafon de parler that makes
it possible to briefy describe complex properties of fnite sets. The i nfnite is
i nadmissible as the basis of rational thought because it i s not part of nature.
That was a refection of Hi l bert' s notion of the harmony between exi stence and
thought. In hi s vi ew, the road to rel i able operation with the i nfnite led through
the fnite.
Thi s is the fnitar viewpoint. To realize it rigorousl y Hi l bert gave a re
stricted l i st of admissibl e symbol s. To guard mathematics against the intrusi on
of notions of the i nfni te connected with vi sual appeal and the use of i ntuition
he el aborated a theory of formal proofs. In it, symbol s that express logical
assertions are transformed by means of rigorousl y formul ated rules i n a manner
si mi l ar to that used in al gebra.
The frst stated aim of the new calculus was a formal proof of the consi stency
of the arithmetic of positive integers. For more than two decades Hi l bert and
hi s students l ooked for a road to the solution of this problem. They achieved
many "intermedi ate" successes but the fnal tri umph el uded them.
In 1 93 1 there appeared a paper of Kurt GOdel 1 O that was l i ke a bol t from
the blue. By sophi sticated refnement and formalization of arguments that went
back i n essence to the "l i ar" paradox, GOdel proved the astounding resul t that
in any formal system that incl udes the arithmetic of the natural numbers it i s
possible t o formul ate an undecidable proposition - that i s, a statement that can
neither be proved nor disproved. At the same ti me, if one accepts the exi stence
of the compl ete set of natural numbers, then this statement i s true or false. Thus
Borel ' s demon, capable of carryi ng out countably many checks, could fnd out
which of the two possibil ities holds.
GDdel ' s resul t was one of the greatest achievements of logic i n the more than
two thousand years of its exi stence. (Once, to be sure, he happened to hear, at
a conference on l ogic, a lecture that cl ai med that nothi ng had been achieved i n
logic si nce Aristotl e' s ti me! ) Hi s di scovery brought to l i ght the chasm between
the true and the provable. We shall not go into the ci rcl e of probl ems connected
with GDdel ' s di scovery but wi l l refer the reader to the splendid [Russi an] book
The provable and the unprovable by Yu. I. Manin, publ i shed in 1 979 by Soviet
Radi o.
While GDdel ' s proof showed that Hi l bert' s program can not be real i zed, hi s
efforts were not wasted. Hi s i nvestigati ons gave ri se to metamathematics, a new
branch of mathematics devoted to proof theory. This led to a great deepening
of i deas and to the evolution of the methods of mathematical logic. Later, al l
thi s turned out to be of great use for the el aboration of algorithmic languages
for computers.
Axiomatization of the infnite. Another path toward overcoming the difcul
ti es i n the theory of i nfnite sets was chosen by mathematicians who tried to
base i t on a system of axi oms. One such system was proposed in 1 908 by Zer
melo and l ater refned by A. Fraenkel . The Zermelo-Fraenkel system describes
the properties of the relation of bel onging, x E y, which i s used to defne the
relation of inclusion, x C y, for sets and the notion of equality of sets. Axioms
are formulated that assert that two sets contai ni ng the same elements are equal ,
and that equal sets are contai ned i n the same sets. Then come axi oms that
codify the rul es for set bui ldi ng - the forming of a pair of sets and of the union
of an arbitrary set of sets. Then there i s an axiom that asserts the exi stence of
the set of subsets of a set. Another axiom i n thi s group i s the rul e that si ngles
out a subset of a set on the basi s of certain properties of its elements. Thi s
axi om el i mi nates the paradoxical sets of Cantor, Bural i-Forti and Russel l . Al l
these sets were given by properties of thei r elements but were not subsets of
some "l awful" set.
These axioms enable one to prove the exi stence of the empty set and to
obtai n from any set x a new set {x} whose only element is x. Of course,
the Zermelo-Fraenkel system includes the axiom of choice. It also includes
an axiom which says that the i mage of a set under some mapping i s al so a
set. Fi nal ly, the system contai ns "the axiom of the i nfni te" which, in essence,
asserts the exi stence of the i nfni te set of natural numbers (although thi s concept
does not appear in the statement of the axiom).
The two critical questions faci ng any system of axioms are: i s it possible
to deduce from i t two assertions that contradict one another, and can one prove
or di sprove al l asserti ons formul ated by means of its terms. The adherents
of the Zermel o-Fraenkel system feel that i t must be consistent (that i s free of
contradictions) because no contradictions have turned up thus far (they may, of
course, turn up in the future) . To test its power, attempts were made to use i t to
prove or disprove the conti nuum hypothesi s. But here some remarkabl e results
were found.
It al l began wi th a di scovery GOdel made i n 1 939. By adjoi ni ng the contin
uum hypothesi s to the axioms of set theory GOdel obtai ned a consi stent system
(of course, thi s consi stency is rel ative; i t i s predicated on the assumption that the
1 29
axioms of set theory are consistent) . Luzi n had earl ier anticipated the paradox
ical possi bi l ity that the axioms of set theory contradict neither the conti nuum
hypothesi s nor its negation. In 1 963 Paul Cohen
I I
showed that thi s i s i ndeed
the case. He proved that one can not deduce the conti nuum hypothesi s from
the Zermelo-Fraenkel system of axioms. Al so, i t turned out that the axiom of
choice i s independent of the remai ning axioms of the Zermel o-Fraenkel sys
tem, much as the paral lel axi om cannot be proved or disproved on the basi s
of the remai ning axioms of geometry. It al so became clear that it i s possible,
without contradiction, to adjoi n to the system of axioms obtained from the
Zermelo-Fraenkel system by repl acing the axiom of choice by its negation the
assertion of the i mpossibl ity of wel l -orderi ng the continuum. At vi rtual l y the
same ti me as Cohen, the Czech mathematici an P. Vopenka obtained si mi l ar
(and even stronger) resul ts.
The situation i n mathematics that resul ted from the publ ication of the pa
pers by GOdel , Cohen and Vopenka was si mi l ar to that i n geometry after the
publ ication of the works of Lobachevski and J. Bol yai .
1 2
But Eucl i dean and
non-Eucl i dean geometry were regarded as different mathematical model s of the
real world, and the choice between the two affected physics and not mathemat
ics. The discovery of hyperbol ic geometry did not i mpinge on the foundations
of mathematics. But now i t i s precisely the foundations of mathematics that
are i nvol ved. Now a mathematici an can choose the set theory he prefers. He
can choose the set theory in which the axiom of choice and the conti nuum
hypothesi s hol d, or the set theory i n whi ch one rejects the axiom of choice and
the possi bi l ity of wel l -ordering the conti nuum. He can also choose other possi
bil ities, for example, to accept the axi om of choice but to reject the conti nuum
hypothesi s. In the l atter case he must assume that the conti nuum has a pl ace on
the scale of transfni te numbers but i ts pl ace on that scale i s unknown.
If one bears i n mi nd that set theory cl ai ms to be the basi s of mathematics,
then there are many mathematics, and it i s up to the i ndividual mathemati ci an
to choose among them. Of course, each mathematics yiel ds its model of the
real worl d, but the differences among them are deep and touch on the most
fundamental questions of the theory of cognition. It i s safe to assume that had
Hil bert lived to read Cohen' s papers, he would have taken back hi s proud words:
Mathematics i s a science without hypotheses. To j ustify i t , I don' t
need, l i ke Kronecker, the Lord God, or, l i ke Poi ncare, the assump
tion of the special abi l ity of reason based on the pri nciple of math
ematical i nduction, or, l i ke Brouwer, pri mary intuition, or, fnal l y,
l i ke Russell and Whi tehead,
1
3 the axioms of i nfnity, reduction
1 30
In Search of Infnit
and completeness, which are genuine hypotheses of a substantial
nature, and, i n addition, completel y improbable.
A modern mathematici an i s much more l ikely to approve of the viewpoint
of the noted American mathematician and logician Wi l l ard Quine: 1
4
Begi nni ng in 1 90 I , there have appeared many set theories none
of which is indi sputably superior to the others. Even the question
of whether they are free of individual contradictions i s debatable
i n the setting of such an examination, because we can no l onger
rel y on common sense when establ ishing the l i kel ihood of vari
ous propositions. Set theory is di scredited by paradoxes and as
a foundation for mathematics i t is far less rel i able than its super
structure. It is therefore clear that we cannot regard set theory as
the foundation of mathematics and hope that it wi l l del iver us from
the fear of unsoundness of cl assical mathematics. In devel oping
al l possible systems we try to fnd a scheme that woul d reproduce
i n the corresponding superstructure the accepted l aws of cl assical
mathematics. At this stage we regard set theory as a convenient
short dictionary of mathematical terms used for the formul ation of
the common system of axioms of cl assical mathematics.
We conclude this section wi th Kol mogorov' s relevant opi nion:
The cl ari fcation of the question to what extent, and under what
condi tions, it is admi ssi ble, in the study of i nfni te sets, to i gnore
thei r process of formation, cannot as yet be regarded as complete.
A lost bet. It remai ns to tel l of an attempt to take set theory, and so al l of
mathematics, out of its state of prolonged cri si s. Thi s was undertaken i n 1 907
by Brouwer, who rel ied to si gni fcant extent on the oft-repeated opinions of
Kronecker and Poi ncare. According to Brouwer and his fol l owers, the speci al
character of the concept of the i nfnite was completel y ignored in anal ysi s and
i n geometry from the 1 7th century onwards. Thus, the presumably rigorous
methods of the theory of real numbers and of mathematical anal ysi s, introduced
into mathematics by schol ars in the 1 9th century, not onl y failed to achieve thei r
ai m but led to the creation of an el aborate system based on the completel y fal se
tendency to deal with the i nfni te usi ng means developed for fnite sets. Thi s
view spel led total rejection of the conception of mathematics that went back to
Cauchy, Weierstrass and Cantor.
1 3 1
Brouwer and hi s school bel ieved that this conception of real number and
of function masks the dangers that l urk i n the concept of the i nfnite, abounds
in vicious ci rcles, and lays cl ai m to an extraordi nary general ity that inevitably
leads to contradictions. Thi s view was tantamount to complete rejection of the
1 9th century' s progress i n strengthening the foundations of cl assical mathemat
ics. In particul ar, Brouwer and hi s school viewed Cantor' s set theory as "an
i nteresting pathological speci al case" in the hi story of mathematics that woul d
l i kely cause grief t o future generations. What was especi al l y interesting i n al l
thi s i s that Brouwer had t o hi s credit signi fcant achievements i n set-theoretic
mathematics.
To set mathematics on what they regarded as the right path i t was necessary
to rely on intuition, hence the name intuitionism for this school of thought.
The i ntuitionists rejected the idea of the conti nuum as a point set because they
regarded the idea of the conti nuum as prior to the idea of a poi nt. They said
that the conti nuum i s the medi um for free formati on of poi nts and not a point
set.
The intuitioni sts subjected to witheri ng critici sm the l ogic used by math
ematicians in the 1 9th century and earl ier. In particul ar, they categorical l y
rejected the l aw of the excl uded middle, one of the fundamental l aws of Ari s
totel i an logic, which asserts that every proposition i s either true or fal se. They
argued that this l aw was infered from observations on fnite sets of objects and
applies onl y to such sets. For example, to establ ish the truth of the proposition:
"Among the people l i vi ng on the Earth on | January 1 983 none is 200 years
old" it sufces to check the age of everyone al i ve on that day. But this method
of veri fcation cannot be used to determine the properties of elements of i nfni te
sets. The elements of such a set cannot be l i ned up for a check of documents.
Thi s meant that the "arsenal" of the i ntuitioni sts l acked so powerful a tool as
proof by contradiction. They rejected "pure existence proofs" and requi red i n
each case a presentation of a concrete example of an object with gi ven properti es.
Thus, they woul d accept as an exi stence proof of an object only a descri ption
of a construction of that object. Hermann Weyl , who joined the camp of the
intuitionists, compared concrete asserti ons to treasures and exi stence proofs to
documents contai ning instructions for fndi ng them.
In other words, the i ntuitioni sts required a transition from assertions of the
form "there exist even numbers" to assertions of the form "2 is an even number."
In a lecture on intuitioni sm Brouwer gave the fol lowing example of an as
sertion that can nei ther be proved nor di sproved: "In the deci mal representation
of the number 7 there are ten nines i n a row." At that ti me the number of known
1 32
In Search of Infnit
digits in the expansion of 7 was 707 (many of which turned out to be i ncorrect) .
Computers have enabled us to fnd a great many more digits of the deci mal ex
pansion of 7, so that among them there are, possibl y, ten ni nes i n a row. But
i f the number ten i s repl aced by 1 01 000, then all attempts at veri fcation of our
hypothesi s are forever doomed, the speed of future computers notwithstanding.
Si nce the problem admits of no theoretical solution, it fol lows that the assertion
about the exi stence of 1 01 000 ni nes i n the deci mal expansion of 7 i s necessar
i l y unveri fabl e. Incidental l y, one of the mathematicians present at Brouwer' s
lecture remarked that whi l e we don' t know whether the assertion i s true or not
the Lord God knows. "I have no direct contact with God" was Brouwer's dry
retort.
The i ntuitionists changed mathematics in radical ways. Thei r anal ysi s con
tai ns no di sconti nuous functi ons. In their arithmetic, the fact that a product
is zero does not di rectl y i mpl y that one of its factors i s zero. Al most every
assertion of cl assical mathematics must be repl aced by a strange-sounding i n
tuitioni stic analogue, and much had to be given up. "I don' t regard any of the
theorems in the usual textbooks as i nviol able" decl ared the i ntuitionist Skol em.
The cal l to so radical a transformation won the approval of only a smal l
- but very i nfuential - group of schol ars. Hi l bert was a ferce opponent of
Brouwer' s reform. He said: "What Weyl and Brouwer are doing is none other
than a revi val of Kronecker' s idea. They try to save mathematics by tossing
overboard al l that provokes concern . . . They crumble and chop science. If we
accepted the reform they propose, then we woul d be runni ng the risk of losing
the greatest part of our precious treasures."
Hi lbert angri l y asserted that taking away from mathemati ci ans the l aw of
the excl uded middle was the same as depri vi ng astronomers of telescopes or
forbidding boxers to use thei r fsts. He wrote that the prohibition of exi stence
proofs and of the l aw of the excl uded middle is al most synonymous with a
repudi ation of mathematical science, and the sorry remnants, the few, i ncom
plete and unrel ated resul ts worked out by intuitioni sts, cannot compare with the
power of modern mathematics. Hi l bert complai ned bi tterly that the hypnotic
power of a si ngl e temperamental and clever man could have so unbel ievable
and eccentric an i nfuence on the mathematical mi l ieu.
The intuitioni sts di d not l eave Hi l bert ' s strictures unanswered; they cl ai med
that hi s program for savi ng mathematics was bound to banish from it al l mean
i ng. But the majority of mathematicians sided with Hil bert. They felt that the
very exi stence of mathematics and the vast scope of its appl ications prove that
it is neither absurd nor empty, and to cure a toe one need not amputate the leg.
Chapte r 4. In search of the absolute 1 33
In spite of the fact that most mathematicians rejected the i deas of the i n
tuitionists, the l atter were sure of thei r future victory. In 1 9 1 8 Hermann Weyl
bet hi s friend George P6l yai S that i n 20 years the i deas of the i ntuitionists
woul d tri umph. As a cri terion, he indicated two theorems of cl assical anal ysi s
that can be found i n every textbook of hi gher mathematics but are meani ngl ess
in intuitioni stic mathematics. He thought that within 20 years these theorems
would vanish from general l y accepted mathematics. But after 20 years Weyl
admitted that he' d l ost the bet.
We note that in the l ast few decades interest i n intuitionism has again been
on the i ncrease and many emi nent logicians have expl icitly or i mplicitly joi ned
thi s trend i n mathematical thought.
1 34 In Search of Innit
Loncluson
We have fni shed our journey along the meandering road taken by the human
mind i n its attempt to master that most contradictory of concepts, i nfnity, to
tame i t and use it to apprehend real ity. First there were the myths. Then came the
frst scientifc quests of Pythagoras, Zeno and Ari stotle. By now manki nd has
attained such i mpressive achievements as the modern cosmological theori es,
the i ntricate constructions of mathematical anal ysi s, and the theory of i nfni te
dimensional spaces.
In some of these creations of the human mind one i s al most unaware of the
complexity of the underl yi ng conceptions. While carryi ng out in the course of
technical computations the usual operations of di fferentiation and integrati on
of functions, the engi neer does not stop to think that there was a ti me when
these operations were viewed as vi rtual l y unattai nabl e. As Engel s put it:
When variable magni tudes entered mathematics and when their
vari abi l ity was extended to the i nfnitely smal l and i nfnitely l arge,
then mathematics, usual l y so very moral , perpetrated the Fal l : it
ate the apple of knowledge and thi s opened for it the road to gigan
tic achievements but also to del usi ons. The vi rgi n state of absol ute
meaningful ness, of i rrefutable provabil ity of all things mathemat
ical . . . belonged to the past. An era of di scord had arrived.
Now these words seem al most l ike prophetic antici pations of the modern state
of the i nvestigations of the deep problems of the theory of i nfnite sets.
The reader of thi s book must have more than once asked hi msel f the ques
tion of the practical val ue of the quests it describes, of the possible practical
appl icabi l ity of cardinal ities of i nfni te sets, transfnite numbers, unusual func
tions and curves from the mathematical hal l of wonders, and so on. This i s the
most di fcul t question in the majority of scientifc i nvestigations. It was for a
good reason that Michael Faraday said to some l ord who wanted to know the
practical val ue of el ectromagnetic researches: "And what can one say of the
future of a newborn baby?" At one time or another geneti cs, relativity theory
and quantum physics were accused of being out of touch with the real worl d.
Many regarded Tsiol kovski ' s i dea of interpl anetary voyages as just a dream.
When tryi ng to answer the question of the practical value of set-theoretic in
vestigations one must keep i n mind that set theory has three essenti al l y different
aspects: operations on sets, cardi nal ities of i nfni te sets, and the set-theoretic
approach to mathematics as a whole. The roots of the frst of these aspects reach
Conclusion 1 35
back to the di stant past; indeed, in essence, Ari stotle' s syl l ogi sms express def
ni te rel ations between sets and operations on them. Set-theoretic operati ons are
i ndi spensable i f, say, computers are to do thei r job of col lecti ng and processi ng
data, and such operations are therefore provided for i n algorithmi c l anguages.
We need hardl y mention the wi despread use of symbol s for such operati ons i n
modern mathematical l i terature - it woul d be hopeless to try to rewrite many
cl assical works of 20th-century mathematics wi thout usi ng the symbol s for
i ncl usi on, and uni on and i ntersection of sets.
The question of the practical si gni fcance of the theory of i nfni te sets i s
more di fcul t. Its role i s rather i ndi rect, i n the sense that set-theoretic concepts
underl ie many frequentl y used mathematical theorems. Relevant examples
are fxed-point theorems that are the basi s for many computational schemes,
the theorem of the compactness of the product of compact spaces, and so on.
One of the most i mportant concepts of modern mathemati cs, that of i nfni te
di mensi onal Hi l bert space, arose out of Hi l bert ' s i nvesti gati ons i n the theory
of systems of equations with i nfnitely many unknowns. The most i mportant
model of such a space consists of di sconti nuous functi ons that can be i ntegrated
only in the sense of the Lebesgue i ntegral . It is also in Hi l bert space that we
fnd a most remarkable construct - the Wiener spi ral , whose di recti ons at any
two of its points are orthogonal . Its strangeness notwithstandi ng, thi s spi ral
is the geometric i mage of stochastic processes with i ndependent i ncrements, a
concept of i mportance for appl ications of probabi l ity theory.
We must not forget that the conceptions of mathematical logic that pl ay such
an i mportant role i n the theory of algorithmic languages were to a l arge extent
created i n connection wi th attempts to resol ve the paradoxes that arose i n the
theory of i nfni te sets. In recent years, other aspects of this theory have begun to
acqui re practical val ue. Thus the wel l -known American appl i ed mathemati ci an
Richtmayer thi nks that concepts deri ved in part from set theory are l i kel y to
become part of everyday practice in physi cs. He is also of the opi ni on that
every area of mathematics is of potential i nterest for physi cs.
The most contentious aspect of set theory i s the attempt to construct al l math
ematics on a set-theoretic base (the so-cal led "bourbaki sm"). Many schol ars
hold radical l y di fferent views on this i ssue. Some feel that thi s attempt makes
possi ble a uni fed treatment of di fferent mathematical problems as wel l as the
appl ication of methods developed for the solution of problems of one ki nd to the
study of seemi ngly unrel ated problems. Other mathemati ci ans, (usual l y cl osely
connected wi th the appl ied si de of the subject) accuse thi s approach of bei ng
riddled wi th unnecessary formal i sm and regard the tool s used as i ncompati ble
with the problems under study. These disagreements are due to the divergence
of the worl dviews of the opposi ng si des and of thei r different estimates of the
i mportance of vari ous problems and achievements of mathematics. The truth
may turn out to be some di alectic synthesi s of the now di ametrically opposite
views.
The outcome of these disagreements notwithstanding, the rise and evol ution
of set theory was one of the most i mportant stages i n the hi story of mathematics
and i n manki nd' s attempt to master the concept of the i nfnite. Thi s process
wi l l never end, for the i dea of the i nfnite i s inexhaustible and the human mi nd
wi l l forever fnd i n it new aspects.
Notes 1 37
Notcs
Chapter 1
I . Thales of Mi letus (c. 625-547 BC). Ancient Greek mathematici an and
astronomer. The frst to i ntroduce the notion of proof i nto mathematics.
Proved some very si mple geometric theorems.
2. Anaximander of Mi l etus (c. 585-525 BC). Anci ent Greek phi l osopher.
Student of Thales. First to make the conjecture about the i nfni ty of
worlds and of the universe.
3. Weyl , Hermann ( 1 885- 1 955). German mathematician. Author of i mpor
tant works on mathematical anal ysi s, differential geometry, al gebra and
number theory. Interested in the phi l osophy of mathematics. One of the
founders of i ntuitioni sm.
4. Leuci ppus (c. 500-440 BC). Ancient Greek phil osopher, atomi st.
5. Epicurus (c. 341 -27 1 BC). Ancient Greek phi l osopher, atomist.
6. Zeno of El ea (c. 490-430BC). Ancient Greek phil osopher and author
of paradoxes that constituted a critique of notions such as mul ti pli ci ty,
i nfni ty, motion, and (a nai ve view of) the conti nuum.
7. Diogenes of Samos (4 1 3-323 BC). Ancient Greek phil osopher.
8. Fraenkel , Abraham ( 1 89 1 - 1 965) . German mathematici an. Author of
works on mathematical logic.
9. Naan, Gustav Ivanovich (b. 1 9 1 9). Soviet phil osopher. Interested in the
phil osophy of science.
1 0. Antiphon (second half of the ffth century BC). Ancient Greek phi l osopher
sophi st.
1 1 . Eudoxus of Cnidos (c. 408- c. 355 BC). Anci ent Greek mathematician
and astronomer, author of a theory of i ncommensurable magni tudes and
of the "method of exhaustion" for proving theorems on areas and vol umes.
1 2. Hypati a of Alexandria (370-41 5) . Phi l osopher, mathematician and as
tronomer. Lynched by a fanatical Chri stian mob incited by bi shop Cyril
of Alexandri a.
1 3. Al bert the Great (c. 1 1 93- 1 280) . German phi l osopher, natural ist, theolo
gian and logician. Attempted to adapt Ari stotle' s teachings to theology.
1 4. Bradwardi ne, Thomas ( 1 300- 1 349) . Engl i sh mathemati cian, author of
Trctatus de continuo. First to use the word "i rrational" i n a mathemati cal
sense.
1 5 . Baconthorpe, John. Engl i sh scholastic of the 1 4th century.
1 6. Cusa, Nicholas ( 1 401 - 1 464). German phi l osopher and theologian. Hi s
works prepared the ground for Renaissance pantheism and created premi ses
for the justi fcation of the notion of i nfniteness of the Universe.
1 7. Caval ieri , Bonaventura ( 1 598- 1 647). Ital ian mathematician, student of
Gal ileo, and author of a method of indivisibles for fnding areas and
vol umes.
1 8. Fontenel le, Bernard ( 1 657- 1 757). French author and popul arizer of sci
ence. One of the pi oneers of the phi losophy of the Enl ightenment.
1 9. D' Al embert, Jean ( 1 7 1 7- 1 783). French mathematician, mechanician and
phi losopher. One of the creators of mathematical physics.
20. Abel , Niel s ( 1 802- 1 829). Norwegian mathemati ci an. Proved the un
sol vabi l ity i n radical s of the general quintic. One of the creators of the
modern cri teri a of rigor in mathematical anal ysi s.
2 1 . Cauchy, Augustin-Loui s ( 1 789 - 1 857). French mathemati ci an. Author
of the theory of functions of a complex vari able. Developed mathematical
anal ysi s based on the l i mi t concept.
22. Gauss, Carl ( 1 777- 1 855). The greatest German mathematician of the
1 9th century. Obtained many fundamental resul ts in al gebra, geometry,
number theory, and mathematical analysi s.
23. Schumacher, Hei nrich ( 1 780- 1 850) . German astronomer.
Notes 1 39
24. Riemann, Bernhard ( 1 826- 1 866). German mathemati ci an. Obtai ned
many outstanding results i n the theory of functions of a complex vari able,
in geometry, and i n other areas of mathematics. One of the creators of
the concept of a multidi mensi onal space (Riemanni an geometry) .
25. Lobachevski , Ni kol ai Ivanovich ( 1 792- 1 856) . Russi an mathemati ci an.
Creator of non-Eucl i dean geometry.
26. Fridman (Friedmann), Aleksandr Aleksandrovich ( 1 888- 1 925). Sovi et
physi ci st and cosmol ogist.
27. Lemaitre, Georges ( 1 894- 1 966) . Bel gi an astronomer and astrophysi ci st.
Author of the theory of an expanding Uni verse.
28. Si tter, Wi Ilem de ( 1 872- 1 934). Dutch astronomer. Pioneered the appl i
cation of rel ati vi ty theory to cosmology.
29. Sl i pher, Vesto ( 1 875- 1 969). American astronomer. Fi rst to measure
radi al velocities of galaxies.
30. Hubble, Edwi n ( 1 889- 1 953). American astronomer. Showed that galax
ies are systems of stars.
3 1 . Doroshkevich, Andrei Georgievich (b. 1 937). Soviet physi ci st and cos
mologi st.
32. Novikov, Igor Dmitrievich (b. 1 935). Soviet astrophysi ci st and cosmol
ogist.
33. Bri I loui n, Leon ( 1 889- 1 969). French physi ci st. Author of works on
quantum mechanics, information theory, and phi l osophy of science.
34. Zel manovich, Abram Leoni dovich (b. 1 9 1 3). Soviet cosmologi st.
Chapter 2
I . Cantor, Georg ( 1 845- 1 9 1 8). German mathemati ci an. Creator of set
theory.
2. Luzi n, Nikolai Nikol aievich ( 1 883- 1 950) . Soviet mathemati ci an. One
of the founders of the Moscow mathematical school . Author of a number
of distingui shed works on set theory and the theory of functi ons of a real
variable.
1 40
In Search of Infnit
3. Mani n, Yuri Ivanovich (b. 1 937). Soviet mathematician. Author of
works on al gebra and the appl icati ons of mathematics to modern physi cs.
4. Zadeh, Lotf (b. 1 9 1 8). American mathematician. Author of the theory
of fuzzy sets.
5. Di richlet, Lejeune ( 1 805- 1 859). German mathematician. Author of
works on number theory, mathematical physics and the theory of series.
6. Dedekind, Richard ( 1 83 1 - 1 9 1 6). German mathematician. One of the
creators of modern algebra. In hi s works Dedekind came close to the
ideas of set theory.
7. Bol zano, Bernard ( 1 78 1 - 1 848) . Czech mathematician, phil osopher and
l ogician. Author of a number of works on the foundations of mathemat
ical anal ysi s and the theory of i nfnite sets.
8. Bernstei n, Fel i x ( 1 878- 1 956). German mathematician.
9. Liouvi l le, Joseph ( 1 809- 1 882). French mathematician. Author of works
on mathematical anal ysi s and number theory.
1 0. Lindeman, Carl ( 1 852- 1 939). German mathematician. Proved that 7 i s
transcendental .
I I . Aleksandrov, Pavel Sergeevich ( 1 896- 1 982). Soviet mathematician. Founder
of the Soviet school of topology.
1 2. Kol mogorov, Andrei Nikolaevich ( 1 903- 1 987). Soviet mathematici an.
Author of many outstanding works on functions of a real vari able, the
theory of probabi l ity and topology.
Chapter 3
1 . Descartes, Rene ( 1 596- 1 650). French mathematician and philosopher.
Created analytic geometry, i ntroduced the notion of a variable and studied
rel ations between variables.
2. Weierstrass, Carl ( 1 8 1 5- 1 897). German mathematician. One of the
creators of the theory of functi ons of a complex variable. Created an
arithmetical theory of real numbers and based on it a restructured account
of mathematical anal ysi s.
Notes
1 41
3. Meray, Charl es ( 1 835- 1 9 1 1 ) . French mathematician. Created an arith
metical theory of real numbers at the same time as Weierstrass.
4. Frege, Gottlob ( 1 848- 1 925) . German mathemati ci an and logician.
5. Hilbert, Davi d ( 1 862- 1 943). The greatest German mathemati ci an of
the 20th century. Author of outstanding works i n the fol l owing areas:
the theory of i nvari ants, the theory of algebraic numbers, the cal culus
of variations, the foundations of mathematics, and functi onal anal ysi s.
Introduced the concept of an i nfnite-di mensional space.
6. Kronecker, Leopol d ( 1 823- 1 89 1 ) . German mathematici an who worked
i n the areas of al gebra and number theory.
7. Napier, John ( 1 550- 1 6 1 7) . Scotti sh mathematician. Created the theory
of logarithms.
8. Bernoul l i , Johann ( 1 667- 1 748). Swi ss mathematician. One of the founders
of mathematical anal ysi s. Contributed substanti al l y to the evolution of
the concept of functi on.
9. Euler, Leonhard ( 1 707- 1 783). Mathematician, physi ci st, mechani ci an
and astronomer. Born in Swi tzerl and. Spent most of hi s l i fe in Russi a.
Author of outstanding works i n anal ysi s i n whi ch he devel oped general
methods of i ntegration and of sol ution of differential equations. Studied
functi ons of a complex variable.
1 0. Bernoul l i , Daniel ( 1 700- 1 782). Physicist and mathematician. Son of
Johann Bernoul l i . Author of a number of works on mathematical physi cs.
I I . Fouri er, Joseph ( 1 768- 1 830). French mathematician. Studied equations
of mathematical physics and made extensive use of trigonometric series
for thei r solution.
1 2. Poincare, Henri ( 1 854- 1 9 1 2) . French mathematician, physi ci st, astronomer
and phil osopher. One of the greatest mathematicians of the 20th century.
Author of outstanding works i n al most al l areas of mathemati cs.
1 3. Von Koch, Hel ge ( 1 870- 1 924). Swedi sh mathematician. Hi s work on
i nfnite systems of l i near equations prepared the ground for functional
anal ysi s.
1 4. Hermite, Charles ( 1 822- 1 901 ). French mathematician. Author of nu
merous i nvestigations in mathematical anal ysi s, algebra, and number
theory.
1 5 . StieItjes, Thomas ( 1 856- 1 894). Dutch mathematician. Author of a num
ber of works i n mathematical anal ysi s.
1 6. Perri n, Jean ( 1 870- 1 942). French physicist and physical chemist. Made
experimental studies of Brownian moti on.
1 7. Wiener, Norbert ( 1 894- 1 964). American schol ar, "father of cyberneti cs."
A number of hi s works are devoted to functional anal ysi s and the study
of i nfni te-dimensi onal spaces.
1 8. Lagrange, Joseph ( 1 736- 1 8 1 3). French mathematician and mechanician.
One of the creators of the calcul us of variations, author of a number of
works on mathematical anal ysi s and its appl ications.
1 9. Jordan, Cami l l e ( 1 838- 1 922). French mathematician. One of the founders
of group theory.
20. Peano, Gi useppe ( 1 858- 1 932). Ital ian mathematician. Studied the theory
of di fferential equations and formal-logical foundations of mathematics.
2 1 . Borel , Emil e ( 1 87 1 - 1 956). French mathemati cian. Author of works
in the area of functions of a real variable and in many other areas of
mathematical anal ysi s.
22. Lebesgue, Henri ( 1 875- 1 94 1 ) . French mathemati ci an. Introduced a
general ization of the notion of integral that is widely used in modern
mathematics.
23. Schwartz, Hermann ( 1 843- 1 92 1 ) . German mathemati cian. Author of
works in the area of mathematical anal ysi s.
24. Darboux, Gaston ( 1 842- 1 9 1 7). French mathematician. Author of works
on di fferential geometry, the theory of differential equati ons, and other
areas of mathematics.
25. Ostrogradski , Mi khai l Vasi l ' evich ( 1 80 1 - 1 862). Russi an mathemat i ci an.
One of the founders of the St. Petersburg school of mathemati cs. Author
of outstanding works on the equati ons of mathematical physics, mathe
matical anal ysi s, and theoretical mechanics.
26. Chebyshev, Pafnuti L' vovich ( 1 82 1 - 1 894) . Russi an mathematician and
mechani ci an. Founder of the St. Petersburg school of mathemati cs,
author of outstanding works on mathematical anal ysi s, number theory,
the theory of probabi l i ty, and the theory of mechani sms.
27. Markov, Andrei Andreevich ( 1 856- 1 922). Russi an mathematici an. Au
thor of outstanding works on number theory, mathemati cal anal ysi s, and
probabi l i ty theory.
28. Lyapunov, Aleksandr Mi khai l ovich ( 1 857- 1 9 1 8) . Russi an mathemati
ci an and mechani ci an. Author of outstandi ng works deal i ng wi th poten
tial theory, stabi l i ty of moti on, the theory of equi l i bi um of rotati ng fuids,
and probabi l i ty.
29. Stekl ov, Vl adi mi r Andreevich ( 1 864- 1 926). Russi an and Soviet mathe
matici an. Author of outstandi ng works on mathematical physi cs.
30. Korki n, Aleksandr Ni kol aevich ( 1 837- 1 908). Russi an mathemati ci an.
Author of works on number theory and on partial di fferential equati ons.
3 1 . Zhegal ki n, Ivan Ivanovich ( 1 869- 1 947). Soviet mathematici an. Author
of works on mathematical logic, the foundations of mathemati cs, and set
theory.
32. Egorov, Dmitri Fedorovich ( 1 869- 1 93 1 ) . Soviet mathemati ci an. Author
of works on di fferential geometry, the theory of i ntegral equati ons, and
the theory of functi ons of a real vari abl e.
33. Men' shov, Dmitri Eugenevich (b. 1 892). Soviet mathemati ci an. Author
of outstandi ng works on trigonometric and orthogonal seri es.
34. Bari , Ni na Karl ovna ( 1 90 1 - 1 96 1 ). Soviet mathematici an. Author of
works on tri gonometric series.
35. Susl i n, Mi khai l Yakovlevich ( 1 894- 1 9 1 9) . Russi an mathemati ci an. One
of the creators of descriptive set theory.
36. Novi kov, Petr Sergeevich ( 1 90 1 - 1 975). Soviet mathemati ci an. Author
of outstanding works on descriptive set theory and mathematical l ogi c.
37. Keldysh, Liudmi l a Vsevol odovna ( 1 904- 1 976). Sovi et mathemati ci an.
Author of works on descri ptive set theory.
38. Sierpi nski , Wacl aw ( 1 882- 1 969) . Pol ish mathematician. Founder of the
Poli sh school of mathematics. Author of works on the theory of functions
of a real variable, topol ogy, and number theory.
39. Lavrent' ev, Mikhail Al ekseevich ( 1 900- 1 980). Soviet mathematician
and mechanici an. One of the founders of the Si berian branch of the
Academy of Science of the USSR. Author of outstanding works on the
theory of functions, quasiconformal mappings, aero and hydrodynamics.
40. Brouwer, Luitzen
(
1 88 1 - 1 966) . Dutch mathematician. Author of i mpor
tant resul ts in topology, functional anal ysi s, and mathematical physics.
4 1 . Urysohn, Paul Samui l ovich ( 1 898- 1 924) . Soviet mathematician. Author
of i mportant works i n the area of general topology and one of the creators
of general di mension theory.
Chapter 4
1 . Hurwitz, Adolf ( 1 859- 1 9 1 9) . German mathematician. Author of works
i n mathematical anal ysi s, the theory of functions, al gebra, and number
theory.
2. Hadamard, Jacques ( 1 865- 1 963). French mathematician. Author of
outstanding works i n the area of mathematical physics, the theory of
functi ons, and number theory.
3. Bural i -Forti , Cesaro ( 1 86 1 - 1 93 1 ). Italian mathematician.
4. Russel l , Bertrand ( 1 872- 1 970) . Engl i sh logician, philosopher, mathe
matici an and soci al activist. Founder of l ogici sm.
5. Shni rel man, Lev Henrikhovich ( 1 905- 1 938). Soviet mathematician. Au
thor of outstandi ng works i n number theory, topol ogy, and topological
and quantitative methods of mathematical anal ysi s.
6. Zermelo, Ernst ( 1 87 1 - 1 953). German mathematician. Author of works
on set theory, the calcul us of variations, and probabil ity theory.
7. Bai re, Rene ( 1 874- 1 932). French mathematician. Worked in the area of
the theory of functions of a real vari able.
Notes 1 45
8. Kuratowski, Kazi mierz ( 1 896- 1 980) . Pol i sh mathematici an. Author of
works in the area of topology, the theory of fUt 'tions of a real vari abl e,
and mathematical l ogic.
9. Bourbaki , Nicolas. Pseudonym of a group of modern French mathemati
ci ans who have publ i shed a treatise of many vol umes titled The elements
of mathematics.
1 0. Godel , Kurt ( 1 906- 1 978). Abstract mathematician and l ogi ci an.
1 1 . Cohen, Paul (b. 1 934). American mathematician. Proved the i ndepen
dence of the conti nuum hypothesi s.
1 2. Bol yai , Janos ( 1 802- 1 860) . Hungarian mathematician. Created hyper
bol ic geometry shortly after Lobachevski and i ndependently of hi m.
1 3 . Whitehead, Al fred ( 1 86 1 - 1 947) . Engl ish mathemati ci an and logici an.
Coauthor (with B. Russel l ) of Principia mathematica.
1 4. Quine, Wi l l ard (b. 1 908) . American mathematician. One of the greatest
experts i n mathematical l ogic and the foundations of mathemati cs.
1 5 . P6l ya, George ( 1 887- 1 985). Hungarian mathematician. Author of works
on functional anal ysi s, mathematical stati stics, and combinatori cs.

Infinity

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Infinity

Uploaded by

Copyright:

Available Formats

In the l ast century, the theory of i nfni te sets, created by the German mathe

We are tal king of just the natural numbers.

You might also like