0% found this document useful (0 votes)
224 views158 pagesMichio Kuga - Susan Addington - Motohico Mulase - Galois' Dream - Group Theory and Differential Equations - Group Theory and Differential Equations-Birkhauser (1993)
Uploaded by
Lance Randall GamierCopyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF or read online on Scribd
0% found this document useful (0 votes)
224 views158 pagesMichio Kuga - Susan Addington - Motohico Mulase - Galois' Dream - Group Theory and Differential Equations - Group Theory and Differential Equations-Birkhauser (1993)
Uploaded by
Lance Randall GamierCopyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF or read online on Scribd
/ 158
Michio Kuga
Galois' Dream:
Group Theory and Differential Equations
Susan Addington
Motohico Mulase
Translators
B
Birkhauser
Boston - Basel - Berlin Translator: Translator:
Susan Addington Motohico Mulase
Department of Mathematics Department of Mathematics
California State University University of California
San Bernardino, CA 92407 Davis, CA 95616
USA USA
Library of Congress Cataloging-in-Publication Data
Kuga, Michio, 1928-1990.
[Garoa no yume. English]
Galois’ dream: group theory and differential equations / Michio
Kuga ; translated by Susan Addington and Motohico Mulase.
pcm.
Includes bibliographical references and index.
ISBN 0-8176-3688-9 (alk. paper). -- ISBN 3-7643-3688-9 (alk.
paper )
1. Galois theory. 2. Differential equations. 3. Monodromy
groups. I. Title.
QA174.2.K8413 1993 92-41486
$12'.3--de20 cIP
Printed on acid-free paper.
© Birkhiiuser Boston 1993.
Originally published in Japanese in 1968 by Nippon Hyoron Sha Co., Tokyo, Japan.
Alll 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, photo-
copying, recording, or otherwise, without prior permission of the copyright owner.
Permission to photocopy for internal or personal use of specific clients is granted by
Birkhiuser Boston for libraries and other users registered with the Copyright Clearance
Center (CCC), provided that the base fee of $6.00 per copy, plus $0.20 per page is paid
directly to CCC, 21 Congress Street, Salem, MA 01970, U.S.A. Special requests should
be addressed directly to Birkhiuser Boston, 675 Massachusetts Avenue, Cambridge, MA
02139, U.S.A.
ISBN 0-8176-3688-9
ISBN 3-7643-3688-9
‘Typeset in TeX by by Susan Addington.
Printed and bound by Quinn-Woodbine, Woodbine, NJ.
Printed in the U.S.A.
987654321 Contents
Preface... ee Lo iv
Pre-Mathematies
Oth Week No prerequisites. 2... 2.2... -. 8
Ist Week Sets and Maps ........... a)
2nd Week Equivalence Classes... 2... re)
3rd Week The Story of Free Groups... 22... 28
Heave Ho! (Pull it Tight)
4th Week Fundamental Groups of Surfaces... .. . 33
5th Week Fundamental Groups... 2... . . 38
6th Week Examples of Fundamental Groups. . . 45
7th Week Examples of Fundamental Groups, continued . 49
Men Who Don’t Realize
That Their Wives Have Been Interchanged
8th Week Coverings . . L883
9th Week Covering Surfaces and Fundamental Groups. 59
10th Week Covering Surfaces and Fundamental Groups,
continied 2.0... fee 61
1th Week The Group of Covering Transformations . |. 70
Everyone Has a Tail
12th Week The Universal Covering Space... . 7
13th Week ‘The Correspondence Between Coverings of @ 0)
and Subgroups of ™(D;0) . fees 88
Seeing Galois Theory
14th Week Continuous Functions on Covering Surfaces . 89
15th Week Function Theory on Covering Surfaces... . 93
Solvable or Not?
16th Week Differential Equations . . . cee 105
17th Week Elementary Methods of Solving _
Differential Equations... ...... 109
18th Week Regular Singularities... . Lee. Hd
19th Week Differential Equations of Fuchsian Type... 129
References... ee ee .. ML
Notation. 2... ee L143
Index ..... So -. 147 vi
Contents
Contents
Preface... ee iv
Pre-Mathematics
Oth Week No prerequisites. 5... 2... Le. 8
Ist Week Sets andMaps .. 2... 0. ee 5
2nd Week Equivalence Classes... 1... Lene 19
3rd Week The Story of Free Groups... . 2... 23
Heave Ho! (Pull it Tight)
4th Week Fundamental Groups of Surfaces... 2... 33
5th Week Fundamental Groups... 2.2.2... 38
6th Week Examples of Fundamental Groups... . . 45
7th Week Examples of Fundamental Groups, continued . 49
Men Who Don’t Realize
That Their Wives Have Been Interchanged
8th Week Coverings 2... L288
9th Week Covering Surfaces and Fundamental Groups. 59
10th Week Covering Surfaces and Fundamental Groups
continied .. 1.2.0... . 61
11th Week The Group of Covering Transformations . . . 70
Everyone Has a Tail
12th Week The Universal Covering Space. 77
13th Week ‘The Correspondence Between Coverings of (D;O)
and Subgroups of ™(D;O).. . . « .. 83
Seeing Galois Theory
Lath Week Continuous Functions on Covering Surfaces. 89
15th Week Function Theory on Covering Surfaces... . 93
Solvable or Not?
16th Week Differential Equations... . cee 105
17th Week Elementary Methods of Solving
Differential Equations ......... 109
18th Week Regular Singularities... . fee. 1d
19th Week Differential Equations of Fuchsian Type . . . 129
References... 2... 02 ee . ul
Notation. 2... 0. ee 143
Index 2... ee 147 Preface
They came back to say that they all fell into the crevasse.
These are lecture notes for a course I gave at the University of Tokyo a few
years ago. The forty students who attended were first year undergraduates. Among
these, about twenty five completed the course. Rather than giving a simplified
presentation of the lectures here, I will reproduce the complete set of lectures,
preserving the original style faithfully. The title of the course was “Group Theory
and Differential Equations.”
What should one expect from the title “Group Theory and Differential Equa-
tions”?
It is well known that group theory, at its inception, was deeply connected to al-
gebraic equations. The success of the theory of Galois groups in algebraic equations
led to the hope that similar group-theoretic methods could be a powerful arsenal
to attack the problems of differential equations. In fact, Sophus Lie devoted his
life to this problem, inventing the theory of Lie groups as a weapon. Although this
brought him lasting fame, he was unable to make substantial progress in his original
aim of a group-theoretic study of differential equations. Picard and Vessiot inher-
ited the problem, and later it was studied by Ritt and Kolchin from the algebraic
point of view. This work has been very successful for some special types of differ-
ential equations, including linear ordinary differential equations, but it seemed far
from satisfactory compared with the theory Lie envisioned. In order to follow Lie’s
original program, it would be necessary to use the ideas and techniques of modern
analysis, such as the infinite dimensional Lie groups of E. Cartan and Kuranishi and
the cohomology theory of D. C. Spencer. [More recently (1990), Kimura, Hattori,
Sato, Kashiwara, and Yoshida have also contributed to the subject.]
By the way, these lectures are not an introduction to these important theories
(although the subjects are not completely unrelated).
‘My lectures are on Fuchsian differential equations and their monodromy groups.
Riemann also tried a group-theoretic treatment of linear ordinary differential equa-
tions. He used discontinuous groups rather than the continuous groups that Lie
used later. These discontinuous groups are defined purely topologically, and are
used to understand the way the domains of the differential equations are connected.
‘These groups are called fundamental groups or monodromy groups. They also rep-
resent the multivalued nature of the solutions of the differential equation. In fact,
the fundamental group I’ is contained in the continuous group G considered by
Lie, Picard, and Vessiot. (We call it the Picard-Vessiot group later on.) However, Preface
since T # G, T is not as powerful as the Picard-Vessiot group in solving differential
equations. But if we restrict ourselves to only Fuchsian type differential equations,
then the solutions are completely characterized by the monodromy group I. This
topic will be the main subject of this book.
The main importance of the monodromy group, however, is not in the theory
of solvability of differential equations. Rather, P shows its true power when the
differential equation cannot be solved by elementary methods. The monodromy
group connects the theory of Fuchsian differential equations of special types with
the theory of automorphic functions and enables us to investigate the deep ana-
lytic and algebraic structure of the solutions. Riemann treated the case with three
singularities—that is, the case in which the solutions are essentially described by
hypergeometric functions. Riemann’s P-function theory is nothing but this. This
approach was continued later by Fuchs and Poincaré. Poincaré studied the subject
with relation to automorphic functions.
Some might criticize my choice of such a dusty old-fashioned topic. It seems
puny compared with the healthy, gushing stream of modern analysis.
Nevertheless, I have chosen this topic for pedagogical reasons. Namely,
(i) It can serve as an introduction to algebraic systems (groups), topology, and
analysis (function theory).
(ii) Only elementary knowledge of these three subjects is needed, but all three
are necessary. This theory is the most primitive example that the intersection of
many different disciplines is the most interesting mathematics. It can help con-
vince students that mathematics is a unity; it also gives a wider perspective of
mathematics.
(iii) It gives a geometric understanding of Galois theory.
(iv) There are many challenging problems which are not too difficult for student
exercises.
(v) Above all, I cannot forget the old dogma that the most interesting aspect
of analysis is its algebraic structure.
It may be true that there are no more good research problems left in the
theory of Fuchsian differential equations, but the situation is more difficult in several
variables. There seem to be many new directions to explore here. For example, it
doesn’t seem possible to reduce Appell’s theory of hypergeometric functions in two
variables to the theory of symmetric spaces. I wonder if it is possible to reduce
the case of discrete monodromy groups to the moduli theory of algebraic varieties.
I think that Kahler studied such problems in the 1930’s, but they seem to be
untouched since then. Is it possible to understand Leray’s recent work on the
Cauchy problem in partial differential equations as preliminary to a group-theoretic
study of differential equations?
[Recent work on hypergeometric functions in two variables includes a long paper
by Deligne and a book by Manin.] Preface
Actually, I believe that the lecture series was quite successful. Some of the
students became math majors, and went on to become good mathematicians. And
they came back to say that they all fell into the crevasse.
September, 1967
Translator’s note: When this book was first published, it was a best seller; stu-
dents would carry it around to be hip, whether or not they could read it. However,
it got Kuga in trouble with the mathematical establishment, as it was considered
undignified for a professor at Tokyo University to publish a book with cartoons and
funny examples.
More recently Kuga was asked to write another book in the same style, since
Galois’ Dream was so successful in attracting students to mathematics. He declined,
saying, “I was only the Pied Piper.”
Feminist readers are asked not to take umbrage at some of Kuga’s examples;
keep in mind that the book was written in another era and in another culture. Kuga
had a deep respect (and admiration) for women.
This translation was a joint effort; it could not have been done with only one
of us. Mulase translated from the Japanese; then I refined the text into idiomatic
English. We also consulted a preliminary translation that Tadatoshi Akiba did in
1975.
I would like to thank the many people and institutions who provided assistance
for this project: the Paul and Gabriella Rosenbaum Foundation, for providing some
financial support; California State University at San Bernardino, for help with
computer software and hardware, especially the staff of the Computer Center and
Audio-Visual Departments; Lillian Kinkade, who TeXed the entire first draft of the
book, learning TeX especially for the occasion; the many people whose laser printers
I borrowed, especially Dave and Shari Stockwell; Madge Goldman, for serving as
a liaison between translators, editors, and others; and Han Sah, for mathematical
and editorial advice.
Finally, I wish to acknowledge a profound debt to Michio Kuga, my mathe-
matical parent, who died February 13, 1990.
Susan Addington
February, 1992.
ix This page intentionally left blank Pre-Mathematics This page intentionally left blank Week 0 3
The Oth Week:
No Prerequisites
The title of this series of lectures is “Group Theory and Differential Equations”.
‘The contents are briefly explained in the preface. The Oth week is a summary of
the preliminary meeting of the course. (In the lecture, I outlined everything in the
course, and left the students in a fog.)
Let me describe the prerequisites you need to follow these lectures.
(0) There is no prerequisite for the first two weeks. Later you will need some group
theory, introductory topology, theory of functions, and differential equations. Do
not be intimidated at this point. As you will see, you won’t need more than the
introductory parts of these subjects.
(1) By the third week (you still have two full weeks to study!) you will need some
group theory. Learn the definition of a group and some fundamental results. It is
enough to understand the concepts of group, subgroup, normal subgroup, homomor-
phism, and isomorphism. Nothing beyond this will be used. While this material
can be found in any introductory book on abstract algebra (also called modern
algebra), concise treatments are given in Pontryagin [11] and van der Waerden [15].
(2) If you do not know the language of topology, find out what the terms “neigh-
borhood”, “open set”, “continuous map”, and “homeomorphism” mean by the 12th
week. You can find these definitions in any textbook on point-set topology (also
known as general topology). Singer and Thorpe [12] covers these topics in the first
few pages, and [11] also contains a brief but complete summary.
(3) Starting in the 15th week, you will need some knowledge of the theory of func-
tions of a complex variable. You need to know about holomorphic functions, the
Cauchy-Riemann equations, contour integrals, Cauchy’s Theorem (i.e. fi, f(z)dz =
0), Morera’s Theorem, Cauchy’s integration formula, power series, meromorphic
functions, poles, and Laurent series. (There is a rapid review of these topics in
Week 15.) This material can be found in any textbook on complex analysis; Ablfors
[1] is a standard one.
(4) Some facts about linear differential equations in a complex variable will be
needed starting in the 16th week (especially the theorem on existence of solutions).
By the 17th week, you will need to know about differential equations of Fuchs type.
A good reference is Birkhoff and Rota [4]; an older book is Ince {8}. 4
Galois’ Dream
(5) Fron the 16th week on, some knowledge of the simplest notions of linear algebra
will be needed: linear transformations and their matrix expressions. (Only two-
dimensional vector spaces are used.) See, for example, Curtis [5], or any book on
linear algebra. Also look up the definition of a linear representation of a group in a
graduate-level algebra book. It can be found in [15], for example.
The Schaum Outline Series generally presents subjects in a straightforward
way; the series has books on topology, algebra, linear algebra, and complex analysis.
See the references for titles and authors. Week 1
The First Week:
Sets and Maps
This week I will explain the concepts of “set” and “map”, which are funda-
mental in mathematics. I am sure that many of you know these terms already, so
my explanation will be just an appetizer before the main course.
Sets.
We call any collection of objects that is clearly determined a set. Each object
in this collection (= set) is called an element. We use capital letters M, N, etc.,
to denote various sets, and small letters x, y, etc., to denote elements of these sets.
When =r is an element of (say) M, we sometimes use such expressions as “x belongs
to M” and “z is contained in M”. In mathematical notation,
Ms3nz or «EM.
If y is not an element of M, we use the notations
Mty or yeM.
Example 1. Consider all the natural numbers. Of course, you know what they
are: 1,2,3,4,5,.... These numbers are obtained by starting from 1 and adding 1
some number of times. Although there are infinitely many natural numbers, we will
consider them all at once. This collection of all natural numbers is a set. We use
the letter N to represent this set. That is,
N= {1,2,3,4,5, ...}
All the natural numbers belong to N, and nothing but natural numbers belongs to
N. The number 1966 is a natural number; so 1966 € N. But a negative number
7031 is not a natural number; so —7031 ¢ N.
Example 2. Now let’s consider all the integers. An integer is any of
1) Natural numbers: 1, 2,3,4,5,...
2) Natural numbers numbers with a negative sign: —1,—2,-3,—4,—5,... and
3) 0.
In mathematics, the letter Z is used to denote the set of integers.
—7031 € Z, because — 7031 is an integer.
5 6 = Galois’ Dream
355 355 | .
Tag f Z decause [T= is not an integer.
Example 3. Next, consider all the rational numbers. A rational number is a
number that can be written as the quotient of two integers m and n. (n # 0, of
course!) The letter Q denotes the set of all rational numbers.
355 355 F
“3 €Q, because — 3 84 rational number.
v2 ¢Q, because V2 is not a rational number.
Example 4. The totality of real numbers forms a set, which we denote by R.
V2ER,1+2V-1¢R.
Example 5. The set of all complex numbers is denoted by C. A complex number
is of the form z+ ¥—ly (a and y are real numbers.)
Example 6. The set of all human beings alive at this precise moment (= October
31, 1960, 1:32 + 17.8322... seconds.)
Example 7. The set of all students in this room right now.
Example 8. The set of all prime numbers smaller than 100. This is the set
{2,3,5, 7, 11, 13, 17, 19, 23, 29, 31, 37, 41, 43, 47, 53, 59, 61, 67, 71, 73, 79, 83, 89, 97}
consisting of 25 elements.
Example 9. Consider a plane. For example, we can imagine stretching the black-
board in this room infinitely in all possible directions. Let’s call this infinite black-
board m. There are infinitely many points on this plane 7. (So saying, the instructor
marks many points on the board.) All of these are points of x. Therefore, the plane
m can be considered as a set (that is, the set consisting of the points of m). If the
point P is on 7, then we write P € 7. If the point R is not on 7, then we write
R¢n.
Example 10. Similarly, consider a circle C on 7. (The instructor proudly draws
an almost perfect circle on the board.) This circle C’ is also a collection of points.
Therefore, it is also a set. In Figure 1.1,
PEC, and Q¢C.
A set consisting of (geometric) points is called a point set. The sets in Examples
9, 10, and 11 are, therefore, point sets.
Example 11. Let’s consider an arbitrary shape on 7. (The instructor tries to draw
O-ba Q.*) Let’s call the shaded part D. Since D consists of the points that are part
of D, D is another set.
* a well-known Japanese cartoon ghost Week 1
Figure 1.1.
Warning: It is not necessary to consider shapes which are a single piece. For
example, (the instructor draws a figure next to D, and calls it P-ko.**) You can
consider these two figures at once and call them a set F.
Figure 1.2.
A point set that consists of more than one piece, such as F, is called discon-
nected. We will discuss this later. (The 4th week.) For example, the set of points
on the land on this globe is a disconnected point set.
Figure 1.3.
The sets in Examples 6, 7, and 8 have only finite numbers of elements. The
set in Example 8 consists of 25 elements, the set of Example 6 has approximately 3
** O-ba Q's sister
7 8 Galois’ Dream
billion members, and the set in Example 7 (er . . . let me count: one, two, three, .
- .) consists of _____ members.
Sets such as these which have finite numbers of elements are called finite sets.
On the other hand, sets with infinite numbers of members or elements are called
infinite sets, as in Examples 1, 2, 3, 4, 5, 9, 10, and 11. If a finite set M has m
elements, we express this by
[M| =m.
‘That is, |M| stands for the number of elements the finite set M contains. If M is
an infinite set, we write
|M| = ov.
Example 12. If M is the set of all prime numbers less than 10, then M =
{2,3,5,7}. Therefore |M| = 4.
Example 13. If M is the set of prime numbers between 6 and 9, then M = {7}.
Therefore |M| = 1.
We defined a set to be a collection of objects. In English, a collection usually
means a set containing more than one thing. However, we will find it easier in
describing various mathematical situations to interpret a “collection of objects” as
a collection of one object, as a special case. By this convention we allow M in
Example 13 to be considered as a set, even though |M| = 1.
But even more bizarre, we will consider a set M that satisfies |M| = 0. As
long as we stick to the original description of a set being a collection of objects, it
is invalid to say that this definition makes a collection of no objects a set. But in
mathematics (as in many other places) we turn a blind eye to the law. So we insist
that the collection without any elements is a set, and call it the empty set. We use
the symbol @ to denote it. (In other words, we introduce the new concept of the
empty set.)
Mathematese
If you learn a few words of mathematese now, you will be rewarded later.
The first word is the symbol V. This is a mathematese term meaning “all”. You
could also interpret this to mean “any”. Actually, these two words, all and any,
have logically equivalent meanings. For example, the statement “All women are
stupid” (Excuse me!) and the statement “Any woman is stupid” have the same
meaning, don’t they? (Of course, whether or not the statements are true is a
different question.)
The symbol V was invented by upending the letter A to emphasize that it
stands for the words all and any.
‘The symbol J is another word in mathematese meaning “there exist(s)”. This
is a reversed E, which stands for the word exist.
The symbol = represents negation (i.e., No... or Not. . ..)
The symbol | is used often instead of “such that” or “satisfying”. Week 1
The brackets {...} represent a set whose elements are listed or described between.
the brackets. For example,
{ 0-ba Q, O-ba P, Dodompa* }
denotes the set that has O-ba Q, O-ba P, and Dodompa as its elements.
Pocket Mathematese-English Dictionary
Mathematese English
a exist
= negation,
I such that
{...} the set with ... as its elements
Examples of Usage
(1) {ec € R | 2? < 1} is the set of all real numbers z such that £, then MD L.
According to our definition of a subset, the set M is its own subset; i.e., we
have M C M. Now suppose that NV is a subset of M, and furthermore suppose
* standard Japanese nonsense word
9 10
Galois’ Dream
that NV is not M itself; ie., MC M and NV # M. We call NV a proper subset of M.
When WN is a proper subset of M, and we want to emphasize this fact, we write
M2ZN or NGM
Convention: We always consider the empty set § to be a subset of any set. In other
words, for any set M, M > 0.
Exercise. Let M = {a,b,c} be a finite set consisting of three elements a, b, and c.
Find the total number of subsets of M.
(Solution):
(0) M is a subset of M.
(1) The sets consisting of two elements
{a, b}, {b,c}, {c,0}
are subsets of M.
(2) The sets consisting of one element
{a}, {b}, {c}
are subsets of M.
(3) The empty set 0 is a subset of M by our convention.
Therefore M has the eight subsets listed above.
Figure 1.5.
Exercise. Let M be a finite set with m elements. Find the number of subsets of
M.
(Answer): 2”, Prove this yourself. Week 1 11
Set Operations
‘We choose one set M and investigate the relations among its subsets A, B, C,
When A and B are two subsets of M, we want to consider the collection of
elements of M that belong to at least one of A or B. In mathematical notation,
consider
{x €M | 2 (at least one of A or B)}.
This is certainly a subset, and we denote it by AU B. The subset AU B is called
the union of A and B.
Figure 1.6.
In other words, AUB = {x € M | x € (at least one of A or B)}. AUB can
be represented by the shaded shape in Figure 1.6. If there are three subsets, A, B,
and C’ of M, we can still define the union AU BUC by
{x€M | «€ (at least one of A, B, or C)}.
More generally, if you have subsets A1,A2,Az,... (a finite or infinite number of sets),
you can still define the union
Ua
={zeEM | xé (at least one of Ai, Aa, As, -.)}
={reM|%| ce Aj}.
A, U AQU A3U..
Figure 1.7. 12 Galois’ Dream
Figure 1.8.
Next consider the elements of M that belong to both A and B. This set is
called the intersection of A and B, and is denoted by AN B.
ANB={ceM | céAand re B)}.
See the first part of Figure 1.8 for a pictorial explanation.
If A and B are separated, as in the second part of Figure 1.8, then ANB = 0.
Also,
ANBNC={reM | cé€A,c€B, and z€C)}.
Figure 1.9.
The intersection of a finite or infinite number of subsets A1,A2,A3,... can be
defined similarly (and is illustrated in Figure 1.10):
ALN Ag NASON. = (Ai
= {xeEM | x (all of Aj, Ag, As,-.-)}
={xEM | Vi,x € Aj}. Week 1
rw
ie
,... satisfying C2. When a partition of M is given, each of the
subsets Na,Ng,Ny in the sum M = U, Ng, is called a class by mathematicians.
For nonmathematicians, a “compartment”, a “country”, or a “territory” might be
easier to understand. We will use any of these words in this lecture. 20
Galois’ Dream
Example 1. Suppose M is the set of all human beings on earth at this moment
(= 1:36:27 p.m. on December 19, 1960). Let Nj be the set of males and N2 be the
set of females. Then M = Ny UN2 and Ni M.N2 = 0. Therefore, M = Ni UNo
gives a partition of M. The classes in this partition are Ny and N2.
Example 2. The set M is as in Example 1. Let No be the set of all babies of age 0,
Nj be the set of children of age 1, Nz the set of children of age 2, ..., and, in general,
Nj is the set of all people of age k. Clearly, M = Ugg Ne, and M,N; = 0 if
Nu # Nj. Therefore, M = U9 Nx gives a partition.
It is convenient to use equivalence relations in order to find a partition of a
set. Now let me explain what an equivalence relation is. First, a binary relation is
a relation between two objects (or elements of a set). For example, the relation of
being brothers, as in “Mr. X and Mr. Y are brothers”, is such a relation.
Example 3. Some binary relations in the set of human beings.
(i) wand yare brothers. Let us write By to abbreviate this expression. For exam-
ple, John Kennedy B Ted Kennedy, Groucho Marx B Harpo Marx, Tweedledee
B Tweedledum, etc. When Mr. U and Mr. V are not brothers, we write UBV.
For example, unfortunately Kuga B Rockefeller.
(ii) We write 2Fy to mean z is the father of a boy y. The following theorem holds.
Theorem. If xFy and Fz and y # z, then yBz.
(iii) 2 is in love with y. Notation: «Ly. One of the misfortunes of life is that 2Ly
does not imply that ya. When there is a relation among three elements such as
aLy and zLy, such a relation is called a ternary relation. Non-mathematicians
would call this a love triangle.
Example 4. Let £ be the set of all straight lines in three-dimensional Euclidean
space E°. Here are some binary relations between two elements of L:
(i) Land U are parallel. Notation: Illl/; negation is 1 U.
(ii) Land are perpendicular. Notation: [LI’; negation is | LU’.
(iii) | and I’ intersect.
Here we define ! and 1’ to be parallel if 1 = 1’, or 1 and I’ are contained in the
same plane without intersecting. The following theorem is well known.
Theorem 2.1.
(1) aiid
(2) Uji! implies U0
(3) Ut’ and 10” imply Uj”.
(1) and (2) are obvious from the definition. (3) is not so direct, but the proof is
easy.
If a binary relation ~ on a set M satisfies the following three conditions E1, Week 2 21
E2, and E3, we call it an equivalence relation.
El: For any element ¢ € M, 2 ~ a.
E2: c~y implies y ~ x
E3: ¢~y and y~z imply 2 ~ z.
Example 5. Some equivalence relations in the set of all humans.
(i) Have the same age.
(ii) Live in the same country.
(iii) Are of the same sex.
Example 6. In the previous example, if we stretch the meaning of “brothers” so
that a person is his own brother, the relation B becomes an equivalence relation in
the set of all males. El: 2Bz (now that B has a new meaning), E2: «By implies
yBz, and E3: By and yBz imply 2Bz.
Example 7. In the set of straight lines C, the relation || is an equivalence relation.
Now suppose ~ is an equivalence relation on a set M. We will denote by V(x)
the subset of M consisting of all those elements y that satisfy y ~ x. Namely,
() N(@) = {ye M | y~ 2}
Of course,
(2) cE N(a),
because x ~ x (El).
Thus when you choose 2, the subset M(x) is determined. As you change « within
M, various N(x) emerge. But the union of all these subsets is, of course, M. That
is,
Cl: M =Usem (2):
This is because any element y of M is contained in N(y).
C2: N(2) # N(y) implies N(x) NN (y) = 0.
Indeed, if N(x) N.N(y) # 9, then there is an element z € N(x) NA (y). By E2 and
E3, 2 ~ y. Therefore, by E2 and E3 again, if u ~ x, then u ~ y and vice versa.
This means N(x) = N(y). This proves C2.
What we have proved now is this: M = U,¢.4,N(2) gives a partition of M.
‘This shows how an equivalence relation ~ on M gives rise to a partition of M.
‘As we change x in M, we obtain an N(z) corresponding to each x. However, the
sets M(x) are not necessarily distinct. In other words, even if « # y it is possible
that M(x) = N(y), as we have seen. For simplicity, assume M is finite. Although
it seems that the number of sets \V(x) is the same as the number of elements of
M, denoted |M|, there may be much repetition of the sets N(x). So, the number
of distinct M(x)’ is usually considerably smaller than |M|. In the expression
M = UzemN(a) there is much waste, though the statement is certainly correct.
See the following example. 22
Galois’ Dream
Example 8. Let M be the set of all humans, and x ~ y be the equivalence relation
“x and y are of the same sex”. Then it follows
Kuga ~ Albert Einstein
Richard Nixon ~ Elvis Presley
Yoko Ono ~ Shirley MacLaine
Ann Landers ~ Abigail van Buren % Kuga.
If we make different N(x) by changing x, we have
N (Kuga) = \’(Binstein) = M’(Nixon) = N'(Mao Tse Tung) = N(Blvis) = ....
We also have
N (Kyoko Kuga) = (Shirley MacLaine) = M(Brigitte Bardot) =...
Each of these is identical to either {men} = Ny or {women} = No, and there are
only two distinct classes. In order to simplify the expression of M as the union
of A(z), simply choose one representative of all men and one representative of all
women; for example, M = (Michio Kuga) U (Shirley MacLaine). In general,
let us consider the set of all different compartments in a partition M = U, N(z)
which is generated by an equivalence relation ~ . This set is denoted by M/~ and
is called the quotient space (or quotient set) of M by ~ . (An element of M/~ is
itself a subset N(x). The set M/ ~ is a set of sets.)
Example 9. Let M be the set of all humans, and ~ the equivalence relation of
being of the same sex. Then M/ ~ consists of only two elements Nj and No, i.e.,
M/~ = {Ni,No}-
Example 10. Let £ be the set of straight lines as before. We want to find a
partition of £L by the equivalence relation || (parallelness). Each equivalence class
consists of parallel lines. We call each class a “direction”. (Actually, an equivalence
class is a pair consisting of a direction and its opposite, since a line that points
northwest also points southeast. So, £/|| is the set of all directions.
The map which assigns the class (2) to each element x of M is a map from
M to M/ ~. This is obviously a surjection. We call this map the natural map. If
we denote this map by v,
Moda => u(z) =N(c).
(Soliloquy: This is not thorough enough, but maybe there will be a second chance. I
wonder if the students understood the subtlety involved in putting the statement of
C2 as Na # Ng implies NaNNg = 0 instead of using a # implies Na MNg = 07!)
Exercise. Let f be a map from M to N. If we define ~ by x ~ y if and only if
f(x) = f(y), then ~ is an equivalence relation on M. Show that every equivalence
relation coincides with some equivalence relation ~ defined in this way using some
function f.
Answer: N= .M/~, f =v (natural map). Week 3 23
The Third Week:
The Story of Free Groups
We shall start with a set of 2n + 1 “letters”,
Ey A, Aa, Ag, Ady oy Ans
Ay’, A", 4g, Ag?) Ant
An arbitrary sequence of these letters is called a word. For example,
A5A3Az1 A100 A29Z Ax EAgi Aoi (a word of length 10),
AgAz'AzApArAz* Ay BEB Ayo (length 11),
Ay Aj? (length 2),
Aj? (length 1),
E (length 1),
are all words. The number of letters used in a word is the length of the word. If A;
occurs m times, we count it m times. We denote the set of all words by W. This is
an infinite set.
Warning: Needless to say, two words containing exactly the same set of letters can
be different if the orders of these letters are different. For example,
A\As1A5EA2AgEAs Ao
and
AsAsA2AgEA31BAi Aa
are two different words. We use W, Wa, ... to denote elements (i.e., words) of W. A
word is itself a sequence of letters, but we consider this sequence as a single object
and represent it by a single letter such as W. This is very common in mathematics.
For example, a complex number x + yi is determined only when you specify two
real numbers « and y; we often express such a number by one letter z. Also, we use
x or & to represent a vector (21,22, ...Tn)-
Given two words such as A,A5'AsHA4A>! and A3A,A7*EA2, we can join
them to make a new and longer word,
A,A31AsEAqAz'A3A1A71E Ap. 24
Galois’ Dream
We call this longer word the product (or concatenation) of A,A;+AsEAqA3‘ and
A3A1A7'EAg. In general, the product of two words W; and W2 is the word ob-
tained by joining W; and W2. We use the symbol W; - W2 to denote the product
of Wi and Wo.
Caution: W; - We and We -W, are not necessarily the same word. For example, if
W, = A, A31A5EAqAz! and Wz = A3A1A71E Ag,
Wy» We = AAs AsEBAgAz'AgA1 Az *E Ap,
but
WoW, = A3A,A71BA2A,A3 1 AsEAgAz!.
As you see, W; -W2 # W2- Wi.
It is obvious that
Formula 3.1. (W, - W2)- Ws = Wi - (We - Ws)
holds. From now on we denote the product of three words by W - Wz - Ws without
parentheses. Of course, this is nothing but the three words W1, We, and Ws joined.
Similarly, the product W; - W-Wg- +» - Wm is the expression obtained by joining
these words.
Next we want to define some operations on elements of W.
(I) If the expression A;A;? occurs within a word, erase these two letters and
replace them with E. This is an operation of type I.
#0 AAD ee
ee tA;
Il 4 Il (operation of type I)
—
E Geen?
(II) Similarly if Ay1A; occurs in a word, replace these by E.
0 RADIAL Ree
ei
Il 4 Il (operation of type II)
Fee EB tae
(III) The reverse of an operation of type I. That is, replace the letter E by A;A;?.
The index i can be anything between one and n.
. :
ll 4 ll (operation of type III) Week 3 25
(IV) Similarly, the reverse of an operation of type II is an operation of type IV. That
is, replace E by Ay 1Aj.
eee BO ees
Il 4 I (operation of type IV)
TR RATIA THER
(V) Ifa word has length > 2, and if it contains E, erase E.
oe ae
tes
| 4 od (operation of type V)
—
eR THR
Caution: Of course, after E is erased, there will be a space of length 1, so
we must compress the word to eliminate this space.
(VI) The reverse of the preceding operation. That is, insert the letter E between
any two letters of a word (or at the beginning or the end of the word). For
example,
EAgA\ As‘ AsEAs1= AgA,As'AsEAs) => AgAiA31A5EAs1E
4
AgAi As EAsE As"
4
AgA1 As AsSEEAs*
‘We call a combination (composition) of a finite number of these six operations
a fundamental transformation. When the word W, can be changed to W2 by a
fundamental transformation, we say that W; and W, are equivalent, and denote
this by W; ~ We. For example, the following fundamental transformations
AyoAs'EA3Az! = AioAz1A3Az1=>AioE Az !=>A10 Az!
v
AyoAg*Az'A1 A343"
4
A1oEA3*A,*A1A3A3*
4
Ayo AsA51A3*Ay'A1 A343"
BA,oA5A51A51Az1A1A3A3+
show that.
AwA3*EA3A3* ~ AioAz*, 26
Galois’ Dream
and
AoA3!EA3Az1 ~ EAyoA3A51A51Az*A,A3A5*.
Obviously we have
Proposition 3.1. The relation ~ above is an equivalence relation. That is,
(a) W~m
(b) Wi ~ We > Wa ~ Wi
(c) Wi ~ We and W2 ~ W3 > W, ~ Ws
We also have
Proposition 3.2.
Wy ~ We
(a) Wim oi ea
You can prove these very easily.
Example. Let
Wi = AioA31EA3Az1, We = AroAz*
Ws = A2AsEA5*AG1, Wa = A2dg?.
Then W, ~ W2 and Ws ~ W,.
Now
Wi: Wa = AioAs*BA3Aq*AzAsEA5*AG) ~ AioAz"A2AG*
=W.-Wa.
Next is a procedure for producing a new word from a word W. We replace
each letter A; in W by Ay, Ay? by Aj, and keep E untouched. Then we reverse
the order of the letters in the word. The new word we obtain is denoted by W~?.
For example, if
W = A,A51A3A71A2EA3A31E,
then
W-) = BA3A51BAz1AgA5 1 A541.
Obviously,
Proposition 3.3. Week 3 27
Proposition 3.4.
Wi ~W2=> W,'~w,t
Proposition 3.5.
WW ~W Wek.
Here, E is the word consisting of the letter E alone.
For example, when
W = A,A51A3A71 A EA5A3E,
then
W-} = EA3A51BAz1AqA31A54;2.
Therefore,
W-W7) = A\Az1A34z1 A,B As Az }BEA3A5 1EAz1AgAs*A5 Ay?
~ AA51A3Az1A2EAsA31A3A51BAz1AgAg As Ay?
~ A, A51A3A7 Ag BASE A; 1EAz1AgA51A5A;*
~ A, A51A3A;'A2EA5A5\EAz'AgAg* ASA"
~ A, A51A3A7!AgEEEA31AqA31A5Az*
~ A, A51A3A71A2A31Ag AS AS Ay?
~ A, A51A3A71AgA31A5 Az?
~ Ay A51A3A31A5 Az
~ A, A51A5Az*
~ AAT)
~E.
Similarly, W-! - W = E. The general situation is the same. QED.
Let us denote the quotient set W/~ of W by this equivalence relation ~ by
the symbol F; ie., F = W/ ~ . We denote elements of F by wi, wa,-..,Wp)-+-
Each of these is an equivalence class by ~.
Suppose w; and ws are two elements of F. Choose any word W; from w; and
Wp from w2. Make the product W; - Ws, and denote the equivalence class to which
W, - We belongs by w. The procedure is summarized as follows:
wy 3 Wy (arbitrary choice)
— WW,
tw > We (arbitrary se} product Wi - We
—+ equivalence class of Wi - W2 = w. 28
Galois’ Dream
Since both w; and w2 contain an infinite number of elements (words), let’s change
W, to another W{ still belonging to w,, and replace Wz by W4 from we.
w > Wi
we 3 Wy
} — W{- Ws — eqnivalence class of W}- Wi =u".
Actually, the equivalence class w’ is nothing but w. This is because W, and Wi
belong to the same equivalence class and W2 and W4 belong to the same equivalence
class, so W; ~ W/ and Wz ~ W3. Proposition 3.2 states that Wy - Ws ~ Wi- W3.
Therefore, both products belong to the same equivalence class, namely w = w'.
QED.
This means that no matter which representatives W; and W2 you choose from
‘wy and w respectively, the equivalence class to which W, - W2 belongs is w. That
is, the class w is determined by w; and we, and does not depend on the choice of
W, and W. We can now call w the product of w; and we, and adopt the notation
wy we.
Now we have defined a product
Faw
—w-wer
F> ai wre
in the set F. The class which contains the word consisting of the letter F alone is
denoted by e. Since
Ai Ay! ~ E,
then
E€e, A;A;1 ee, AT1A; €e, EE €e, etc.
Similarly, we choose a representative W from a class w, and make the word
W-}, We denote by w~? the class to which W-} belongs. As before, w~? does not
depend on the choice of W; it depends on w alone (by Proposition 3.4).
Theorem 3.1. If wi, w2, ws, w € F, we have
(1) (wy - we) - wg = wy - (we - ws)
(2)
(3)
The proof is left to the reader.
To those who know the definition of groups, it is better to say Week 3 29
Theorem 3.2. The set F is a group under product. The unit element is e. The
inverse of w is w~1.
In the Oth week, I recommended that you teach yourself the concept of groups
by this week. Have you done it? We will now use some terms from group theory.
As I said, the set F becomes a group. This group is called the free group
generated by A1, Az,..., and An. The letters Aj, Ag,..., An are called the gen-
erators of the group. When you want to emphasize the fact that F is generated by
Aj, Aa)... Any F is written as F(A1, Az,..., An). The element of F containing
the one-letter word A; is denoted by the same symbol Aj.
Finally, an interesting theorem. (Prove it yourself.)
Theorem 3.3. Let G be an arbitrary group, and let a1, a2,..., aq be any
elements of G. Then there is a unique homomorphism ¢ from F(A, Ao,---, An)
to G satisfying
(Ax) = a1, (Az) = @2,..-,9(An) = an
Hint: Let y(A3451A4g£A7!) = aga5‘agaz}, for example. This page intentionally left blank Heave Ho! (Pull it tight) This page intentionally left blank Week 4 33
The Fourth Week:
The fundamental group of a surface
The scene of today’s lecture is set in a region in a plane. We define a region to
be a part of a plane surrounded by some closed curves. For example, the portion D
in Figure 4.1 surrounded by the closed curves Cy, Cz, and C3 is a region (i.e., the
unshaded part of the figure).
Figure 4.1.
In the lectures we often call a region “the land” and the outside of a region
“the sea”, or “lakes”, or “ponds”. The main character in today’s story is a curve.
The curves in the lectures will be limited to those that have two end points (one
initial point and one terminal point), and have a direction (or orientation) from the
initial point to the terminal point. For example, the “curve” C in Figure 4.2, which
approaches a circle infinitely closely, is excluded from the curves under consideration
because C' does not have a terminal point.
c
—<——
Figure 4.2.
However, a curve could be zigzag, straight, or even have intersection with itself. 34 = Galois’ Dream
Q_ terminal
initial point
point P.
initial
point P* .
ge terminal
2 point
initial
point P,
ee, Q terminal
ani point
initial
int P terminal
nn WAY Gas
Figure 4.3.
(See Figure 4.3.)
It is not necessary that the initial point and the terminal point be different.
When the two endpoints coincide, we call the curve a closed curve.
initial point =
terminal point
=P=Q
Figure 4.4.
Since we are considering oriented curves only, we must consider the curve ob-
tained from a curve C' by reversing the orientation as distinct from C itself. We
express the curve with reversed orientation by C7}.
terminal point ofC
= initial point ofC ~!
initial point of _,
= terminal point of C YS
Figure 4.5.
Obviously,
the initial point of C = the terminal point of C~*
the terminal point of C = the initial point of C7! Week 4 35
When the terminal point of C; coincides with the initial point of C2, we can
obtain a third curve by joining C and C2. Of course, this is the curve obtained by
tracing C; first, then tracing C2. We call this new curve the product (or concatena-
tion) of Cy and Cp, and denote it by Cy «C2.
Q =terminal point of C
P
\ GR
Figure 4.6.
nitial point of C ,
It is clear that we have the following formula when the terminal point of C, is
the initial point of C, and the terminal point of C> is the initial point of C3.
Formula 4.1.
(Ci C2) Cz = Ci « (C2 Cs)
Both sides of Formula 4.1 represent a curve which traces Ci, C2, and Cy in
this order. From now on we write Cj - Cp - Cy to denote this curve.
If you can always draw a curve in D from P to @ for an arbitrary pair of points
P and Q, we say that D is connected. In other words, it is not always possible to
travel from P to Q without swimming, jumping, or flying. We will consider only
connected regions in the rest of the lecture.
Figure 4.7.
Now we consider the set of curves in a connected region D. We denote the set
of curves in D by W(D). Then we define an equivalence relation ~ as follows:
Hi:
H2:
Two curves C; and Cy (both elements of W(D)) are equivalent (i.e., Cy ~ Ce)
ift
The initial point of C, is the initial point of C2;
The terminal point of C; is the terminal point of C2; 36
Galois’ Dream
H3: Cj can be deformed continuously to Cz without moving the end points.
In other words, we consider the curve C; to be made of a rubber string. Then
move C; around D, while keeping the end points fixed, by stretching or contracting
the rubber string C;. Be careful to keep C; strictly inside D while you do this.
Never let the string O; get wet in the lakes.
Figure 4.8. Men at work. The railroad is deforming continuously.
Figure 4.9. You can’t deform continuously if there is a pond.
It is obvious that the relation ~ is an equivalence relation. Therefore, we
have an equivalence relation ~ on W(D). Mathematicians say that C; and C2 are
homotopic when C; ~ Cp. In Figure 4.10, C1, C2, and Cy are all homotopic to each
other, but C; and Cy are not. We will define homotopy groups next week based on
this idea. Week 4 37
Figure 4.10.
Note: This discussion on curves is qualitative and pictorial. In order to work
with curves, it will be necessary to view them as equivalence classes of continuous
functions defined on an interval. Two equivalent curves look exactly the same, but
may be parametrized by different functions. 38 Galois’ Dream
The Fifth Week:
Fundamental Groups
Last week we were considering curves in a region. A region was a part of the
plane surrounded by closed curves (for example, the unshaded part of figure 5.1).
We decided to call such a region land and the outside region a sea or a lake (such
as the shaded part of Figure 5.1) .
Figure 5.1.
When we considered a curve in the region D, we always meant a curve with an
initial point, a terminal point, and orientation. In these lectures, a region means a
connected region; namely, any two points in the region can be connected by a curve
in the region.
D
Figure 5.2. Disconnected region, or country consisting of at least two islands.
When C; and C2 are two curves in D, we call them homotopic, and use the
notation C ~ C2, if the following conditions are satisfied: Week 5 39
Hi: The initial point of C; is the initial point of C2.
H2: The terminal point of C; is the terminal point of C2.
H3: C, can be deformed continuously into C2 without moving the end points.
Figure 5.5. C1 ~ Cp ~ Cy # B.
So far this is a review of the last week.
It is clear that we have the following proposition: 40 Galois’ Dream
Proposition 5.1. If Ci ~ C2, C3 ~ C4, and the terminal point of C, the initial
point of C3, then Cy -C3 ~ C2-C4.
Recall that C; - Cs is the product curve obtained by tracing C; first, then C3.
‘Try to prove this proposition by looking at Figure 5.6.
Figure 5.6. & ~ a => OCs ~ C2 + Cs.
The following proposition is clear, too.
Proposition 5.2. If Cy ~ C2, then Cy! ~ Cy}.
The symbol C-? stands for the curve C’ with the orientation reversed.
Now we consider a subset of W(D). Choose a point O in D. Consider the set
W(D;) of closed curves in D with O as both the initial and terminal point. (See
Figure 5.7)
Figure 5.7. Closed curves which belong to W(D; 0) Week 5
‘The product of the elements Ci and C2 of W(D;Q) is a curve which starts
at O, returns to O once by way of C1, and returns again to O via C2. Therefore,
C\ + C2 is a closed curve starting (and ending) at O, so C; - C2 is again an element
of W(D;0).
41
Lemma 5.1. If C; and Cz € W(D;0), then Cy - Cz € W(D;0).
It is also clear that we have
Formula 5.1.
(Ci + C2) - C3 = C1 - (C2 -Cs)
See Formula 4.1, which is similar.
Now let us consider the quotient space, W(D;O)/ ~, of the set W(D;O)
by the equivalence relation ~ of homotopy. It is denoted by 71(D;O); namely,
™(D;O) = W(D;0)/~. Each element of 7,(D; 0) is a set of equivalence classes
of closed curves from O to O in D. Such an equivalence class is called a homotopy
class. The homotopy class which contains the closed curve C' is denoted [C]. Let
a and b be arbitrary elements of ™(D;Q). By definition, each of a and 6 is a
set of closed curves, all of which are homotopic to each other. Now choose closed
curves A from a, and B from b. (Of course, we have a = [A] and = [B].) Then,
since the product A- B is again a closed curve starting and ending at O, (i.e.,
A-B €W(D;0)), the curve A- B determines a homotopy class [A- B] € ™(D;0).
Let c = [A- B]. We can verify that this class c is completely determined as soon as
the classes a and b are chosen, and it does not depend on which curves A and B
we have chosen from a and 6 respectively. Indeed, choose A’ from a and BY from b.
‘Then the product A’ - B’ ~ A- B by Proposition 5.1, because A~ A’ and B ~ B’.
That is, both A- B and A’ - B’ belong to the same homotopy class.
[A-B]=[4’ BI =C
Therefore, we will use the symbol a-b to denote c, because c is determined by a
and b alone, and call it the product of homotopy class a and b. In other words,
[A-B]=a-b.
‘This defines an operation which assigns to an arbitrary pair of homotopy classes a
and ba third homotopy class a-b. Let me use a diagram to summarize the situation:
m(D;0) 3.ar+ choose A(E a)
Eos Se ED | 4B =o ben(Di0) 42 Galois’ Dream
The following is clear for a, b, and c in ;(D; 0):
Proposition 5.3.
(a-b)-c=a:(b-c)
When a closed curve having O as its end points can be deformed continuously
to the point O, we call the curve null-homotopic.
Figure 5.8. A null-homotopic curve can be contracted to O.
Figure 5.9. A loop railroad which is not null-homotopic.
We denote the homotopy class consisting of all null-homotopic curves by 1.
Obviously, the following is true for any a € m(D;0):
Proposition 5.4.
a-l=l-a= Week 5
The inverse class of a homotopy class a is defined as [A~] where A is a curve
in a. We denote this by a~! : a~! = [A~]. The class a~? depends only on the
homotopy class a and is independent of the choice of A. The proof is obvious from
Proposition 5.2.
For any a € 7;(D;O),
43
Proposition 5.5. For any a € 7(D;0O),
Proof: Take a € A. Since a~! = [A~4], a-a~1 = [A- A“4]. A- AW? is a closed
curve starting from O, stopping at O on the way, and finally coming back to O,
following A in reverse. (See Figure 5.10.) Let’s deform this continuously. Since
all we need to fix are two end points, we can move the points in between without
any restriction. The curve B, which follows A just until the point O comes into
view, then returns, is equivalent to A. A~?, as in Figure 5.11. Similarly the curves
B', B",... of Figures 5.12, 5.13, and 5.14 are all homotopic, and
A-ATA BA BN Bs...
Finally, the curve A- A? will be deformed continuously to the point O. Therefore,
A. A-} = 1 (= the constant curve) and a-a~! = 1. The argument is similar for
ata. QED.
° a.
Figures 5.10, 5.11, 5.12, 5.13, 5.14.
Using the language of groups, 44 Galois’ Dream
‘Theorem 5.1. The set of homotopy classes ™(D;O) is a group under product
abra-b.
‘The unit is 1. The group 7(D; 0) is called the fundamental group of D relative to
the base point O. (Sometimes it is called the Poincaré group or the first homotopy
group.) When we change the point O to another point O’, what happens? There
is a theorem.
Theorem 5.2. Let D be a connected region, and O, O' two points of D. Then
m(D;O) and ™(D; 0") are isomorphic.
The proof is left to the reader. The idea is that for a curve A connecting O and O’,
the assignment of [A~!-C’- A] € m1(D;0’) to [C] in 3(D; 0) gives an isomorphism.
(See Figure 5.15.)
When only the abstract group structure is considered, ™,(D; 0) is denoted by
m(D). When ™(D) = {1} = the trivial group, we say that D is simply connected.
This means that any closed curve in D can be contracted to a point. Week 6 45
The Sixth Week:
Examples of fundamental groups
Example 1. The region D is the plane minus one point. There is no sea, but
there is an infinitely small lake Py. (See Figure 6.1.) According to our rule, we are
not allowed to move the curve across Py. No matter how small, a lake is a lake, so
we may not move our rubber railroad across Pp.
~~
‘Om base point
Figure 6.1.
If this is difficult to visualize, think of Pp) instead as an infinitely tall pole
reaching to heaven. (See Figure 6.2).
Let C be the curve at O and going around the pole once counterclockwise
before coming back to O. Let C2 be a similar closed curve that goes around the
pole twice. In general, C,, is a closed curve that goes around the pole n times
(n=1,2,...). Also, we define C_,, as a closed curve going around the pole n times
clockwise. Finally, let Co be a closed curve that is null-homotopic. Intuitively
it is clear that C, is not homotopic to Co, (Ci # Co), C2 is not homotopic to
Ci, (C2 # Ci), nor is Cz homotopic to Cy, (C2 % Cp). If you are not convinced,
tie a rope around a pole twice, and heave ho! Pull it tight! If the rope comes loose,
then Cz ~ Co. But this won’t happen.
In general, if n # m, then Cy, £ Cm. So if we define cn = [Cn],
60 = 1, C14 C25 0065 Cny 2225-1, C-25 C235 ++ 46
Galois’ Dream
Figure 6.2.
are distinct elements of 1(D;O). It can be proved that there are no other elements
in ™(D;O) than cy (n = 0,1, +2,...):
m1(D;O) = {co, €1,€25+++5C-1, C-25 ++}
Intuitively this is obvious, but a rigorous proof is difficult (and will not be given
here). It is also clear that Cm-Cn = Cm4n. Therefore, 71(D; O) is isomorphic to Z
(the additive group of integers): 1(D;O) ~ Z. In particular, it is an abelian group.
In order to write down the correspondence that gives this isomorphism, we
do the following: Let C be the complex plane, and let a be the complex number
corresponding to Py. Since our closed curve C’ does not pass through the point
Pp, the complex line integral [ 4. makes sense. (We are assuming that C is
lz
rectifiable, i.e., C has length). If we let
=i f[%
© Oni Joz
then n(C) is an integer, and C ~ C’ implies that n(C) = n(C’). Therefore, the
correspondence
n(C)
a
W(D;0) 3 Cr n(C) = = f *
eZ
gives an isomorphism. (Look this up in any book on the theory of complex func-
tions.) Week 6
‘This example shows the usefulness of fundamental groups in analysis.
Example 2. D = the plane — {Po,Qo}, the plane with two points removed. Is
the fundamental group of D commutative?
Answer: No, it is not. In order to see this, let P be a closed curve going around
a pole Pp once counterclockwise, and let Q be a closed curve around a pole Qo
counterclockwise. Now it is sufficient to show that P-Q ¢ Q-P. This is equivalent
to showing P-Q-P7!-Q-1. 41.
To see this, wind a rope around the two poles so that the rope represents
P-Q-P-1.Q-1. Without letting go of the ends of the rope, heave ho! Pull it tight.
If the entire rope comes loose, then P-Q-P~!-Q-1 ~ 1. When you try it, you will
see that this doesn’t happen. Therefore, P-Q-P~1-Q-! % 1. This means that
P-Q¢#Q-: P,which shows that (Plane — {two points}; O) is not commutative.
Actually, the fundamental group is known to be the free group with two generators.
(See Figure 6.3.)
peer
Figure 6.3.
Example n. Similarly, (Plane — {n points}; O) ~ the free group with n gener-
ators. If n > 1, this group is not abelian.
Example 1’. The fundamental group of an annulus. As in Example 1, the
fundamental group is isomorphic to Z. Actually, this case can be reduced to Example
1. Let D’ be the annulus with outer radius a and inner radius b. When we use
polar coordinates (r,@) to represent points in D’, the correspondence y given by
=>
DI > (7,0) (5,0) € the plane ~ {the origin}
47 48
Galois’ Dream
is a one-to-one and onto map from D’ to (Plane — {origin}) = D. Since both y
and y7? are continuous, D’ and D are homeomorphic. (In general, two spaces are
homeomorphic if there is a bijection such that both 7 and 7! are continuous.)
Figure 6.4.
‘We use D’ = D to denote this. In general, we have the following theorem.
Theorem 6.1.
If D and D’ are homeomorphic, m(D;O) and m(D’;0") are isomorphic.
Here O and O’ are points in D and D’ respectively. Since the proof is easy, it is left
as an exercise. (Just assign C’ to 4(C).)
Example 2. The region D of Figure 5.1 is homeomorphic to the plane minus
two points. Therefore, this is reduced to Example 2.
Example n’/. D = an island with n lakes on it. Then D is homeomorphic to the
plane minus n points. Week 7 49
The Seventh Week:
Examples of fundamental groups, continued
So far we have discussed fundamental groups of plane regions. Needless to say,
we can define fundamental groups for surfaces in space.
Example 1. The surface in Figure 7.1 is called a torus. The surface of a doughnut
and an inner tube (without considering the air inside) are tori.
Figure 7.1.
Consider the fundamental group (TO) of a torus T. Is it abelian?
Let us consider whether the equivalence classes a = [A] and b = [B] commute,
where A and B are the curves in Figure 7.1. In order to see if a-b = b-a, we check
whether B-1A~!BA ~ 1. If you tie a rope around a torus to represent B~! A~1 BA,
you can see that the rope will come loose without leaving the surface of the torus.
This shows B~1A~1BA ~ 1. Hence a-b = b-a. (See Figure 7.2.)
Figure 7.2.
It is known that every element of ™(7;O) can be written in the form a” -
b™ (n,m = 0, +1, +2,...); therefore, m(T;O) = Z@ Z. In particular, (T;O) is 50 Galois’ Dream
abelian.
Example 2. The fundamental group of an inner tube for two, three, or four
becomes more complicated. These groups are no longer abelian. (See Figure 7.3.)
Figure 7.3.
We can define fundamental groups not only for two-dimensional surfaces, but
also for three-dimensional regions in space. For example, a solid torus (that is, the
essential part of a doughnut) or the air inside an inner tube. The fundamental
group of a solid torus is isomorphic to Z.
Fundamental groups are defined for higher dimensional connected manifolds,
as well as connected regions of higher dimensional Euclidean spaces.
In general, as long as you can define closed curves and continuous deformations
of curves in a set, you can define a fundamental group.
Exercise. Determine the fundamental group of the space obtained from R® by
deleting a circle S and a line ! through it, i.e., D = R3 — (SU).
cD.
Figure 7.4. Men Who Don’t Realize
That Their Wives Have Been Interchanged This page intentionally left blank Week 8 53
The Eighth Week:
Coverings
Let us start with an example. Take two distinct points O’ and O in the plane.
Let a and b be real numbers such that 0 < a < b. Let D’ be the set of points P
whose distance from O! is between a and b. Similarly, let D be the region bounded
by two circles of radii a and b and center O. See Figure 8.1.
We will use a polar cooordinate system having O’ as the origin to express points
in D’, and another having O as the origin to express points in D.
We define a map f from D’ to D by sending a point (r,@) of D’ to the point
(r,20) of D. Then f is a continuous map from D’ to D. The point f(P) = (r, 20)
moves in D as P = (r,8) moves in D’. As P traces out the figure F in D’, as in
Figure 8.2, {(P) traces out the figure in the right side of Figure 8.2. It is denoted
by f(F). Asa point P moves in D’ along a closed curve around O! once, f(P)
moves around O in D twice.
Let Q be a point in D, and consider f~1(Q). That is, find a point P in D’ such
that f(P) = Q. For example, if Q = (c,40°), then the point P; = (c, 20°) satisfies
(Pr) = Q, but this is not the only solution of the equation f(P) = Q; P2 = (c, 200°)
also satisfies f(P) = Q. Indeed, f(c, 200°) = (c, 400° — 360°) = (c, 40°) = Q. There
are no other solutions to the equation f(P) = (c,40°) besides P, = (c,20°) and
Pp = (c,200°). That is, f-1(Q) = {Pi, Po}. In general, for every Q = (r,0) of D, 54 Galois’ Dream
Figure 8.2.
£-1(Q) consists of two points P, = (r, 36) and P2 = (r, 30+ 180°). Thus f is a
two-to-one map. See Figure 8.3.
Figure 8.3.
Let U be a small neighborhood of Q = (r,@) in D and consider its inverse image
{71(U). Since f-1(U) consists of points P in D’ such that f(P) is in U, it is a union
of two small neighborhoods U; of P = (r,30) and Uz around P, = (r,30 + 180°).
In other words, f-1(U) = U; U Ug. (When U is a small disc around Q, U; and Up
are egg-shaped sets around P; and P2.) Of course, U and Uz have no intersection:
U,NU2=0
As P sweeps out all of U;, {(P) sweeps out all of U. If P and P2 are two points of
Uy, f(P,) and f (Ps) are two points of U: Py, P, € Ui, P, # Po => f(Pr) # f(P2)-
That is, f is a two-to-one map of D’ to D, but its restriction to Uj is a one-to-one
map of U; onto U. Moreover,
flv, :U; — U is a homeomorphism onto U
flu, : Uz — U is a homeomorphism onto U. Week 8
Explanation: A continuous map of a domain or region U onto U is a homeomor-
phism if it is one-to-one and its inverse map is continuous. Our f, as a map of D/
to D, is globally two-to-one, so it is not a homeomorphism. However its restriction
to U; is one-to-one, so we can define the inverse map (f|y,)~1 : U — Uj, which is
obviously continuous. Thus f|y, is a homeomorphism.
Generalizing this example, we define the notion of a covering space.
Definition: Let D’ and D be two connected regions in the plane. When a map
{ : D! D satisfies the following conditions C1 and C2, we call f a covering map
from D’ to D. We also say that D’ covers D by f, and that D’ is a covering space
(or surface) of D.
C1: f is a continuous surjection.
C2: If Q € D, then f-1(Q) is a finite or countable set {P,, P2,...,Pay+-.}. HEU is
a small neighborhood of Q in D, then
(i) f-1(U) is the union of small neighborhoods U; of P;, Uz of Pp, .... That
is, f-\(U) =U, UU2 UU» UU Us
(ii) These small neighborhoods U; do not intersect each other: U; MU; = 0 if
i#j.
(iii) f|v, is a homeomorphism onto U.
Let me paraphrase the latter half of C2: by taking a small enough neighborhood
U of Q, we can always write f~1(U) as the union of U;,U2,... that satisfy the
conditions above.
A continuous surjective map f is called a covering map if we can find such a U
for each point Q of D.
Example 1. Let D be the surface of a right circular cylinder with radius 1 em and
height 10 cm. Let D’ be an infinite horizontal strip of width 10 cm in the ey-plane:
D' ={(2,y) | 0 f(A) ~ f(B)
Fix a point O! in D’ and let O = f(0"). Consider m1(D’;0") and ™(D;0).
For every element a of 7(D';0’), take a closed curve A such that a = [A]. Then
f(A) is a closed curve from O to O. Thus it defines a homotopy class a € m(D;O)
which contains f(A).
Since a depends only on the homotopy class a € ™(D';O") and does not
depend on the representative A (by Formula 9,2), let us denote a by f.(a). We
denote by f. the map assigning f,(a) in 1(D;O) to the element a of ™(D/;0’).
Formula 9.1 shows that f, is a homomorphism from 71(D’; 0’) into ™(D;0). Thus
the image f.(m1(D’;0’)) of f, is a subgroup of ™(D;0).
Theorem 9.1. The covering f : D’ —+ D defines a subgroup f.(m(D';0’)) of
m(D;O). Conversely, for every subgroup +y of 14(D;0), there exists a covering
f: D! — D such that 7 = f.(m(D50’)). 60 Galois’ Dream
Figure 9.1.
The proof of this theorem will be postponed until next week. Then we will
also see that f. is injective. Therefore, the subgroup f,(m(D’; O’)) is isomorphic
to ™(D’; 0’). Week 10 61
The Tenth Week:
Covering surfaces and fundamental groups, continued
Let’s recall what a covering map is:
Cl: A continuous surjective map f from D’ to D.
C2: For every point Q of D, f-(Q) consists of finitely or countably many points
{P1, Po, Ps...} of D’. Also, there exists a small neighborhood U around each
point Q such that there is a small neighborhood V; around each P; satisfying
@) f'U) =VUWUVU--, REV.
(ii) Vin Vj = 0 for i #5.
iii) The restriction of f to V; is a homeomorphism from V; to U. That is, fly,
is injective and surjective and (f|v,)~! is continuous. See Figure 10.1.
Figure 10.1.
Such a neighborhood U of D is called a copiable neighborhood. Each of the Vi, Va,..-
is called a copy of U around P;, P2,....
Let f : D! —+ D be a covering. For a curve C’ in D’, f(C’) is again a curve
in D. C = f(C’) is called the projection of C’. We also say that C’ covers C. See
Figure 10.2. 62
Galois’ Dream
Figure 10.2.
Conversely, we can construct a curve C’ in D! from a curve C in D by the following
procedure. For the initial point Q of C, choose a point from f-1(Q) = {P,, Ps,.--},
say P3.
@)
i)
‘Take a copiable closed neighborhood U around Q and consider a copy V3 around
P3. Since the restriction of f to V3, flv, : Vz — U, is a homeomorphism, its
inverse map (f|v,)~! maps U onto V3 continuously. Therefore the segment
CNU = Co of the curve C in U goes to a curve C49 in Vs via (fly,)~!. The
initial point of Cf. is, of course, Ps. Let Ps,:be the endpoint of C4, and let
(Ps) = Q1. Then Q, is the endpoint of CNU = Co. See Figure 10.3-1. Of
course, Co is the projection of Co.
Figure 10.3-1.
Choose a copiable closed neighborhood U; around Q; and take a copy V3: of
U; around P31. Map the segment CMU; = C; of the curve C contained in
Uy into Vs,1 by (flvs1)7?. It intersects C4, and extends it a little. See Figure
10.3-2.
We call the extension C4... Let P32 be the endpoint of C4, and let f(Ps,2) = Qo- Week 10 63
Figure 10.3-2.
Then Co U C; is nothing but the projection f(C4 9 UC5,1) of Ch, UC51-
(iii) Choose a copiable closed neighborhood Uz around Q2 and take a copy V3,2 of
U2 around P3,2. Map the segment CM U2 = C2 of the curve C contained in
Up into Vso by (flvs,2)~}. It intersects C4, and extends it a little. See Figure
10.3-3.
Figure 10.3-3.
We call the extension C4,. Let Ps,3 be the endpoint of C4, and let f(Ps3) = Qs-
Then Co UC; U C2 is the projection f (C39 U C3, UC3,2) of C39 UC3, UC32-
(iv) Choose a copiable closed neighborhood U3 around Qs and take a copy Vs,3 of
Us around P3,3. Map the segment C U3 = Cs of the curve C contained in Us
into Vs,3 by (flvs,3)~1. It intersects C32 and extends it a little. We call the
extension C,3. Let Ps,4 be the endpoint of C3, and let f(P3,4) = Qa. It is the
endpoint of the curve C3. Needless to say,
F(C39 UC31 UC3.2 C33) 64
Galois’ Dream
Figure 10.3-4.
fills up the curve C between Q and Qy. See Figure 10.3-4.
Did you really bother to read all of this? Did you follow this explanation to
the end?
By a successive application of these procedures, the arc C3, starting at P3 in
the covering space D’ will be extended bit by bit (such as C49 U C51; C49 UCS,1 U
03,23 C39 UC3,1 UC3,2 UC5,35 C39 UC3,1 UC5,2 UC3,3 UC3,43--.) and eventually
UGs
0
will cover the whole curve C. That is, f(Uj O%;) = C- (The reason for this is
that, since C contains both endpoints, it is compact. See any topology book.)
So if we let Cz = UR 9 C3,» then C4 is a curve in D’ starting at P3; whose
projection is C. Since the curve C4, is determined by the curve C’ and the initial
point P3; in D’, we call it the lift of C with initial point P;. Because f is a homeo-
morphism on each copy of a copiable neighborhood, the lift is unique on each small
neighborhood. Since the lift is unique at each step, and the small segments overlap,
the lift of the entire curve is unique. In other words, the lift of a curve C depends
only on C and the initial point chosen for the lift.
In the procedure, we chose P in f~1(Q) = {Pi, Ps, Ps,--.} in order to con-
struct the curve C4, but of course we can do the same procedure for any other
points P; or P2 or P3 .... Let Ci be the lift of C starting at Pj, C} be the lift of C
starting at P».... The number of lifts of the curve C’ is equal to the cardinality of
{-1(Q), where Q is the initial point of C. Let M, be the endpoint of Ci, Mz be
the endpoint of C3, .... It is obvious that the points My, M2,... cover the endpoint
N of C. That is, f-!(N) = {Mi, Mp,...}.
Take a point R on C. Every point S1,52,... of the inverse image f-1(R) =
{51,52,...} is ona curve C1, C},.... See Figure 10.4.
We can renumber the indices i so that S; is on Cj. As R moves along the curve C,
each of the points S1, S2,... moves along C’,C%,.... We call R the original and 5; Week 10 65
Figure 10.4.
the shadow of R. As we noted, when the original R moves along C, its shadows
54, 52,... move along the curves C/,C4,... in D’. But can they crash into each
other in D’? It doesn’t happen. See Figure 10.5.
Figure 10.5.
Proof: Suppose $; crashes into Sz when R comes to a point Ro in C. Let Sp be
the site of the accident. That is, S; = Sp = Sp when R = Ro. Let U be a copiable
neighborhood of Ro and let Vo be a copy of U around Sp. Then it is obvious from
Figure 10.6 that fly, is no longer one-to-one. This contradicts the definition of a
covering. Therefore, 5; will never crash into 5}. QED.
Problem: For a covering f : D’ —> D, let n (1
