P. 1
Algebraic topology An intuitive approach (Sato)

Algebraic topology An intuitive approach (Sato)

|Views: 518|Likes:
Published by Spirdym

More info:

Published by: Spirdym on Apr 17, 2011
Copyright:Attribution Non-commercial


Read on Scribd mobile: iPhone, iPad and Android.
download as PDF, TXT or read online from Scribd
See more
See less






Volume 183

Algebraic Topology: An Intuitive Approach
Hajime Sato
Translated by Kiki Hu son



Shoshichi Kobayashi (Chair) Masamichi Takesaki

by Copyright Originally by Iwanami

Sato Hajime Sato 1996

Hajime 1996 by published



Japanese Tokyo,


Publishers, Japanese by



Kiki Hudson Primary 55-01;

1991 Mathematics

Subject Classification. Secondary 57-01.

A_BgffiACT.This book develops an introduction to algebraic topology mainly through simple examples built on cell complexes. The topics covered include homeomorphisms, homotopy equivalences, the torus, the Mobius strip, closed surfaces, the Klein bottle, cell complexes, fundamental groups, homotopy groups, homology groups, cohomology groups, fiber bundles, vector bundles, spectral sequences, and characteristic classes.






SaW, Hajime, 19444 [[so kika. English] Algebraic topology : an intuitive approach Hajime Bato ,translated by Kiki Hudson. p. cm. (Translations of mathematical monographs, ISSN 0065-9282 : v. 183) (Iwanami series in modem mathematics) Included bibliographical references and index. ISBN 0-8218-1046-4 (softcover : alk. paper) 1. Algebraic topology. I. Title. IT. Series. Ill. Series: Iwanami series In modem mathematics. QA612.S3713 1999 514'.2-dc21 98-53247




1999 by the American Mathematical Society. All rights reserved. The American Mathematical Society retains all rights except those granted to the United States Government. Printed in the United States of America

The paper used in this book is acid-free and falls within the guidelines established to ensure permanence and durability. Information on copying and reprinting can be found in the back of this volume. www ams . 10987654321 04 03 02 01 00 99


Preface Preface to the English Translation Objectives Chapter 1. Homeomorphisms and Homotopy Equivalences 1.1. Homeomorphisms 1.2. Homotopy equivalences 1.3. Topological pairs Summary Exercises Chapter 2. Topological Spaces and Cell Cornplexes 2.1. Basic spaces 2.2. Product spaces and quotient spaces 2.3. Topological sums and attaching spaces 2.4. Cell complexes Summary Exercises Chapter 3. Homotopy 3.1. Homotopy sets 3.2. Fundamental groups 3.3. Higher homotopy groups 3.4. Homotopy invariance Summary Exercises Chapter 4. Homology 4.1. Homology groups 4.2. Homology axioms 4.3. Immediate consequences of the axioms
vii Xl

'" xU1

2 3

5 6

9 9


13 16 16

17 17
22 24 26 26

27 27


Exact couples and spectral sequences 9..4.Vll! CONTENTS Summary Exercises Chapter 5.1.3. _['ruer 'U iu res all U v ec w r T""'I 11 1 . Fiber bundles 8. Cohomology of classifying spaces Summary Exercises 73 75 79 80 81 81 84 00 90 91 92 93 97 97 .. Cup products 7. Calculation of homology groups of cell complexes 5.1. Cohomology 0. The universal coefficient theorem Summary Exercises \. Pro ducts of abelian groups 7. Homology of simplicial complexes 5.4. Collapsing cohomology sequences 9.. Cohomology spectral sequences 9. The Universal Coefficient Theorem 7. T T""'I "')1 Ir _J_ res 11 8.2.2. vonomolOgy axioms 6. Homology calculation of cell complexes Summary Exercises Chapter 6.1. Vector bundles 8..1. The Kiinneth formula Homology Groups of Cell Complexes 5.3. Applications of spectral sequences 9.J O.. Cohomology groups of pk(C) 9.6. Spectral sequences of fiber bundles 9. Spectral Sequences 9.7.1. ~r rn '"' 61 61 64 64 66 (j(j 67 69 T:'1. Cohomology of simplicial complexes Summary Exercises 37 37 39 40 44 51 52 53 55 uu 56 59 59 Chapter 7.2.. Grassmann manifolds Summary Exercises Chapter 9.2.3. LC::t.

CONTENTS IX A View from Current Mathematics Appendix Sets Topological spaces and continuous maps Groups Answers to Exercises Recommended Reading 99 105 105 107 109 111 115 117 Index .

Classification in topology is a crude tool. To somebody who has mastered the subject this essential common sense should be as familiar as the air around him. Thus the terms one employs in topology are increasingly becoming important and useful in other branches of mathematics as well as in various fields in the natural sciences. such as any form of change in a geometrical object that stretching or shrinking might cause. By touching it one can feel its physical quality and then keep this as one's own. X1 . -Iist as the state of connectedness characterizes the essence of many phenomena we encounter in our daily lives. it is often necessary to describe to what extent a certain object is connected or separated. I believe that the most efficient way to fulfill this purpose is to investigate simple but meaningful examples in some concrete terms. There are numerous algebraic topology books and many of them are excellent. This book is a simple manual that the reader can follow. The single most difficult thing one faces when one begins to learn a new branch of mathematics is to get a feel for the mathematical sense of this subject. yet we have dared to add another book on this subject. and in fact the reader who follows ourinstructions step by step will end up with areal working model of algebraic topology. The purpose of this book is to help an aspiring first-time reader acquire this topological atmosphere in a short period of time. but one that never fails to determine if a geometrical object is connected or not.Preface Topology forms a branch of geometry emphasizing connectedness as the most fundamental aspect of a geometrical object. In topology. therefore. If a geometrical object is connected then we investigate to what degree it is connected. one ignores virtually all geoIIletricai traits other than connectedness. It takes a long time for a beginner to get to this point. It is important that the reader grasp a mathematical object with his or her own hands.

but also suggested various ways to improve the final product. We did not try to expand the theory to its fullest extent to make our book an encyclopedic reference. Hajime Sato ill. My thanks also go to Tadayoshi Mizutani. Yoshinori Machida. finding many mistakes.In order to pursue this objective we have therefore sacrificed gen- erality and limited the objects of our discussion to the simplest but most essential cases. instead.ly 1996 'See Preface to the Engtish Translation . Tsuchiya for permitting me to use his senrinar notes as well as for giving me much useful advice throughout every stage of the writing. we use the easiest possible examples to help the reader see the backbone of our discussion. and Shigeo Ichiraku. Last but not least. We will be greatly pleased if the reader enjoys reading our book while acquiring several essential methods or approaches to discuss "Practical Topology for Physicists" given by Akihiro Tsuchiya and compiled by Yasuhiko Yamada at the University of Nagoya in 1986. I would like to thank the editors at Iwanami Shoten. Tetsuya Ozawa. who not only read the entire manuscript carefully. I am deeply indebted to Mr.

. So I have provided the reader who has no knowledge of sets. since it simply shows that good vibes alone cannot conquer everything. Yoshinori Machida for spotting numerous typos. several of my friends (including the translator) have complained that the gap between my claim that lW previous klWwledge of mathematics is required .. and consequently she had to do more writing than translating. but they will not do much good unless both the reader and the author have "good vibes" about the subject. This is especially so with the Appendix. I would also like to thank Martin Guest for valuable suggestions. Kiki Hudson. with a basic mllllmal list of definitions and results that may prove useful. This does not really change my original view that the book is readable for anybody who wishes to find out about algebraic topology. filling in missing links and so on. etc. topology. I am grateful to the translator. however. for conveying my writing style and philosophy as faithfully as possible in her translation.Preface to the English Translation It is a great pleasure to me that the American Mathematical Society chose to publish my book" Algebraic Topology: An Intuitive Approach" in their translation series. Since the publication of the original version of this book in 1996. groups. and the AMS editors for presenting the book in splendid style. I have modified some of those troublesome spots. and the actual contents of the book is too big. I think that technical terms help both the reader and the author organize their thoughts. We discussed all the changes verbally. to gether with readable references. This is in the Appendix at the end of the book. Hajirne Sato September 1998 XIII . I have also used the book for my topology seminar (for seniors) and came to see that the reading got a little rough toward the end of the book. This is all right too.

and "cohomology groups". If the concept and reality are far apart you will have . Then after finishing the book the reader should compare its contents with this original concept. translating geometrical concepts into algebraic terms. We actually go beyond counting the number of holes and develop "characteristic classes" to describe how a geometrical 0 bject bends globally. whether a given geometrical object is connected or not connected. In the problem described above. It has managed to express these problems cleanly and algebraically in group-theoretical terms (involving almost only the additive group of integers or cyclic groups of integers modulo prime numbers). One can measure the length of a geometrical object in meters and the weight in kilograms. One focuses on the connectivity. We classify objects according to the nature of their connectedness. "homology groups". ignoring changes caused by stretching or shrinking.Objectives As I stated in the Preface. which may appear to be too slippery to grasp. Intuitively the "i-th homotopy group" describes the "i-dimensional round holes" and "i-th homology group" reveals the number of "i-dimensional rooms" in a geometrical object. I want the reader to spend a few minutes before beginning the book imagining the problem of classifying geometrical objects only with a yardstick that measures their connectedness. But then what is a hole and how do we count the number of holes? In this book. in topology we investigate one aspect of geometrical objects almost exclusively of the others: that is. you will find a mathematical interpretation of these concepts. How do we measure the extent to which a geometrical object is connected? Can we develop a system with suitable units and numbered scales? For example. we can use the number of holes in a geometrical object. termed "homotopy groups". These are some of the major concerns in algebraic topology. it would be nice IT the reader would come to understand and appreciate how contemporary mathematics has constructed the theory of algebraic topology.

ITyou have had nothing to do with them so far. however. I emphasize again that the purpose of this book is to familiarize the reader with the way to think about algebraic topology. but essentially all you need in order to read this book is to understand the o concep : 1 2 Topology by James Dugundji. fundamental groups. Needless to say. homology groups. cell complexes. fiber bundles. If you have already studied general topology (especially its geometrical aspects). Mobius strip. spectral sequences. you might rightly guess from the table of contents that the following are homeomorphisms. I hope that the strange sound they make intrigues you enough to start the book. and if they are rather close your mathematical intuition will have proved to be excellent (and you will ITyou already have any familiarity with algebraic topology. etc. as I feel that this order might work better to sharpen the reader's intuitive understanding. Brown. homotopy equivalences. The reader might feel a need for the theory of groups. 1989 . I use the and will later construct concrete examples such as simplicial homology groups. algebraic topology evolved from general topology (the theory of topological spaces). William C. cohomology groups. Topology has developed (perhaps unintentionally) on the strength we 1 V 1 Ill. then you will find upon finishing the book that they are not difficult at all but that they form some of the basic concepts in contemporary mathematics. characteristic classes. torus. homotopy groups. for instance if you have read Chapters from I to XI in Topology of Junes Dugundji ' you will be ideally prepared. If you have seen some (or all) of these words somewhere before and they have vaguely interested you.xvi OBJECTIVES opened a doorto a brave new world. I have tried to keep my explanation basically intuitive so that even readers with no previous knowledge of general topology will be able to follow the book. Klein bottle. closed surfaces. vector bundies.

The addition and subtraction of integers carry over to those operations modp (we say that modp is a cyclic group of order p).. (X.D-dimensional sphere closed unit interval [0. The only talent this book demands of the reader is a flexible and (1) LIST OF SYMBOLS homotopic homotopy 3 set ball 4 Dn sn-l I X and Y have the same homotopy type n-dimensional 4 9 9 t= (JR:) e 1 e (n ... 7L) hP(X) S simplex (Jj belongs to the boundary a is a ace 0 an at is i erent q-th chain group of S over Z q-th homology group of S over Z n-dimensional simplicial complex projective of o" 48 52 53 space 56 p-th cohomology group of X complex 59 49 . (2) In certain situations. 1] n-dimensional real projective plane (open) i-cell closed i-cell 10 11 13 13 25 31 pt Hp(X. we regard two integers which differ by a fixed prime number p to be equal (we say that we consider integers modp). singleton set ("-. G) G) CA Zp(C) Bp(C) oj h*(X) c for ho(X) G direct sum I:~o of n. G) cone overA reduced homology of group X hp(X) 32 32 32 36 38 45 45 45 -< chain complex group of p-cycles group of p-boundaries (jn Cq(S.7L) pn(c) Hq(S..J H*(X.xvii The addition or subtraction of two integers gives another integer (we say that the set Z of the integers is an additive group). We write 7Lp for the set of the integers modp.

G) Bq(S. rd BO(n) BU(n) Lk(o. n) GL(m. S) . x 6 U (E.ODJECTIVES Symbol h*(X) (jP. HlJ\?5.lUUUL. 8 Gq(S.L Hom(G11 G2) G'2) abelian group of homomorphisms from G 1 to G2 torsion 66 66 67 68 69 69 74 74 74 74 74 74 80 80 83 83 106 Tor(G11 G2) Ext(G1. 7r. F) F----+E~B cross product diagonal map cup product fiber bundle fiber bundle total space base space fiber projection real Grassmannian manifold complex Grassmannian manifold classifying space of real n-vector bundles classifying space of complex n-vector bundles link: complex of a in S G1 abelian group of the extensions of G2 by E B F 7r CIR(m. GJ Lq \61 U2 r'I r'I Meaning direct sum hP(X) of hP(X) coboundary homomorphism q-th cohomology chain of S over G cochain complex of S over G q-th cohomology group of S over G group of q-cochains of S over G group of q-coboundaries of S over G E~o Page 59 60 60 61 61 61 61 1 P. G*(S. B. G) G) G) zq(S.

M. we consider topological spaces and tw-o types of continuous maps betw-een them. B. The letters in anyone of these classes are homeomorphic but no tw-o belonging to distinct classes are. . . {C. N. R. D. X}. {B. strictly speaking. I. L. T. On the other hand. V.CHAPTER 1 Homeomorphisms Homotopy and Equivalences Throughout this book a map means a continuous map. homotopy classification breaks the alphabet into three distinct classes according to their "homotopy types": {A. U. G. The classification of the capital letters A. K. because geometrical properties of homotopy equivalences translate themselves most successfully into modern algebra.' In topology we essentially discuss the connectedness of geometrical objects called topological spaces. J. {H. 5. which are called "homeomorphisms" and "homotopy equivalences" respectively. F. {D. {X}. T. H. C. and here we use the sans-serif style. but it is the one that plays the more important role in algebraic topology. . Z. W. 'See the Appendix for the definition .Q}. O}. We might classify topological spaces up to homeomorphism. Y. K}: {P}. {E. or we might do so up to homotopy equivalence. however. F. V}. Our choice depends on how strong we want our classification to be. The classification according to homotopy equivalences is weaker (there are many spaces not "homeomorphic" to each other that are of the same "homotopy type"). {Q}. 0. . P}. for example we write I and not 1). Z by homeomorphisms results in the following nine classes (this also depends on the choice of font.

.. g 0 f = id and fog = id. PROBLEM 1. We say that topological spaces X and Y are homeomorphic if there exist continuous maps f : X --t Y and g : Y --t X such that the composites g 0 f and fog are the identity maps of X and Y respectively.2. Think of f M FIGURE N 1. where id denotes the identity map.1 Let f be a map which sends the left half A of the M onto the left vertical line plus the center diagonal A of the N without changing anything. . DEFINITION The fact that go f is the identity map implies that f is an injection and g is a surjection. that both SAMPLE f and g are continuous bijective (1-1 onto) maps. fog = id.2 1. . while straightening the right half A of M and sending it onto the right vertical line I ofN (see Figure 1.. . homotopy equivalences? 1. . HOMEOMORPHISMS AND HOMOTOPY EQUIVALENCES Two letters have the same homotopy type if and only if they belong to the same class. Then we get g 0 f = id and SOLUTION.1.We want to transfer the left vertical line and the center di onalline of N onto the left half A of M. Homeomorphisms 1. Consider the letters M and N. in short.1). In this case f is a ho meo mo rphism from X to Y and g is a homeomorphism of Y to X. Similarly the fact that fog is the identity map .1. and to bend the right vertical line of N and map it onto the right half A of M.

1]) is a homotopy between them. For any point Xo in X.4. This fact is intuitively obvious (we can never change the identity map of 0 to a constant map through continuous maps: we cannot shrink the letter 0 to a point without breaking it). 1. We consider two maps fo and f1 from the letter X to the letter Y: fa sends every point of X to the crossing point of Y. the crossing point of X as Xo. Then X-xo consists offour disjoint line segments (each being half open. SUppose there eXIsted a homeomorphIsm f : X ----1 1. in particular. 1]. SOLUTION. These two spaces are not homeomorphic. it. EXAMPLE . Let fo be the map of 0 into itself which sends every point to the apex of the 0 and let !I be the identity map. to be the map sending each point x of X to the point obtained by shrinking II(x) by t from the center crossing ofY. Two maps from atopoiogical space Xto atopo- logical space Y. fi : X ----1 Y (i = 0.1. and I.2.!(xo)· Take. DEFINITION 1. 1. by from fo to h· 'life indicate this situation it (t E [0. having one open end and one closed end).6.Xo onto 1. Show that the topological spaces X and I are not homeomorphic. and I .5. A precise proof. Homotopy equivalences In order to define homotopy equivalences we must first say when two maps are homotopic. the definition of a homeomorphism insures that the map II(X-xo) which is the restriction of f to the space X minus the point xa is a homeomorphism of X . Then fa and /I are not homotopic. 1. HOMOTOPY EQUIVALENCES 3 SAMPLE PROBLEM 1. maps the upper vee v of X onto the upper vee v of Y and the lower wedge A of X onto the lower vertical I of Y by closing A like a tweezer.f( Xo )consists of two disjoint line segments (each of which is half open). 1) I are lWmotoplc if there exists a family of continuous maps It : X varying continuously fo ~ h and say that EXAMPLE ---t Y (t E [0.2. 1]). The basic stance of topology is to regard all spaces lwmeorrwrphic to each other as identical. Take the letter O.3. Then 10 and it are homotopic because we can define t E [0.

If . and we will see it in Example 4.f is homotopic to g : X --+ Y then g is homotopic to f. In general a homotopy equivalence is neither injective nor surjective. This is an algebraic simplification. 1. which we call the homotopy set of X to Y. We are using the same symbol for homotopic maps. A f " X ~ Y is a homotopy equivalence of X and Y if for some g : Y ~ X. Y and O. map map Y are exists 1.Y] the set of the homotopy classes of maps from X to Y. we regard all homotopic maps from X to Y as identical and place them in the same homotopy class. depends on homology theory. We say that X and Y have the same homotopy type if there a homotopy equivalence between them. Show that the map fo : X -+ Y from the letter X to the letter Y in Example 1.9.0] tv Z (the set of the integers).5 is a homotopy equivalence (Hint: for a suitable g : Y ---1 X construct a homotopy between g 0 fo and the identity map as well as a homotopy between fo 0 g and the identity map). we need to look at only one of them. H01'v1EOMORPIIISMS AND HOMOTOPY EQUIVALENCES however. We will discuss the following result in Chapter Three: EXAMPLE [X. If f is homotopic to g and g is homotopic to h : X --+ Y then f is homotopic to h. but this should not cause any confusion here SInce both sides are topologtcal spaces. DEFINITION [0. 2. Consider the letters X. even if a homotopy class has a large number of maps. Suppose we look at the set S of the maps from a topological space X to a topological space Y. In other words. Let X and Y be topological spaces. . A map f : X --+ Y is homotopic to itself. Therefore. Therefore the relation of being homotopic is an equivalence relation on S that breaks S into equivalence classes called homotopy classes. We write X ~ Y when X and Y have the same homotopy type. the composites g 0 f: X ----1 X and fog :Y ~ homotopic to the identity map of X and Y respectively. 3. Y] "-' one point. PROBLEM. We denote by [X. 1.4 1.8.7. The following properties are easy to check.

10.3.1. Given two pairs (X. B) we mean a map f : X ---t Y such that f(A) c B. By a topological]X1ir (X. In later chapters we will study homology groups and cohomology groups (of topological spaces). and has the same homotopy type as the letter O.. Make a loop B in IR:3by tying a knot in the string before splicing its ends (see Fig. A) we mean a topological space X and a subspace A of X. X and fog : y ---+ Y are the Identity maps of X and Y respectively. We will introduce other tools such as characteristic classes to determine if the given spaces are homeomorphic. however. including EXAMPLE of course three-dimensional 1. A) ---+ (Y. two pairs (X. A doughnut is homeomorphic to a coffee cup with a handle. namely. 1. the definitions of homeomorphisms spaces.2). TOPOLOGICAL PAIRS From the definition we see that two topological spaces that are homeomorphic have the same homotopy type. therefore. We have so far used only letters of the alphabet. A) and (Y.A) and (Y.3. A) rather than a single space X. 1. The concept of homeomorphisms for topological pairs parallels the case for single spaces. by a map of pairs f : (X. Passing from single spaces to pairs of spaces as 0 bjects of study was a great breakthrough in algebraic topology in the past. and homotopy equivalences carry over to geometrical objects of dimen- sions two or higher. homotopy equivalences are a looser (less strict) way of classifying topological spaces. The restrictions f 104 : A -----1 B and gl B : B ---+ A are both homeomorphisms. In the (x. B).9. EXAMPLE 1 . Y Z )-space lR:3. 1.splice the ends of a string to make a simple loop A. Topological p arrs In topology we frequently consider a pair of topological spaces (X. B) are homeomorphic if we can find maps f : X ---+ Y and g : Y _ X such that the composItes go f : X _. They each offer the identical information for spaces of the same homotopy type. These are one-dImensIOnal geometncal objects (tOPOlOgICalspaces) consIsting of lines and curves.

0) and (Y. 1. B) into homotopy classes. We partition the continuous maps from a pair (X. and (JR3. B). A) (Y. B)] is the homotopy set of maps from (X.2.ce if it has an "inverse map" in the homotopy . Two topological spaces are homeomorphic if there exists a homeomorphism between them. t]. B). are homotopic if there exists a family of continuous maps of it : (X. Knots We can show that such pairs (JRY. A) _. B).G 1. i = 0. 0). we look at the set denoted by [(X. Three. A). HOMEOMORPHISMS AND HOMOTOPY EQUIVAlENCES A FIGURE B 1.2 A map from one topological space to another topological space is a homotopy equivale:r. A) to another pair (Y. B). B)] I each element is a homotopy class consisting of all homotopic maps from (X. A) to (Y. B) are not homeoA) morphie to each other (later we will compute the "fundamental groups of the com lements of A and the 0 ". varying continuously from fa to h. (Y.. Then we have in which Xy= X Y as the right-hand side of the equality is the homotopy set in which an element is a set of homotopic maps from X to Y. A) to rv. A) ~ (Y. In particular.1 A map from one topological space to another is a homeomorphism if it has an inverse map. that is. Summary 1. We say that [(X. We say that two continuous maps of pairs fz : (X. t E [0. if A = B = 0 we write X and Y in place of (X. 1. (Y.

Show that the pairs (A.3 The upper portion t6 of the letter A is a subspace of A and the upper portion D of R is a subspace of R. 1. b.sense. 1. D) are homeomorphic. . Two topological spaces have the same homotopy type if there exists a homotopy equivalence between them.3 The same ideas carry over to homeomorphisms.) and (R. homotopy equivalences and homotopy types for maps of topological pairs.

1. . We must state precisely which geometrical objects are subjects of our investigation in this book. In this chapter we explain how to build various topological spaces and cell complexes. among which the most basic are (solid) balls. E JR'I ~X. also referred to as disks or cells. In other words. we only consider those objects on which we can impose the concept of continuity. In the ensuing chapters we will deal with cell complexes only. We must be able to determine if a geometrical object is connected or separated. and the (n . We make the convention that the O-disk DO is a one-point space. two and three. < 1 }. ranging from basic ones such as line segments and disks to fuzzy ones whose boundaries are blurry. The dimensions of cells we study do not stop after one.X2. unless otherwise stated. We construct a topological space called a cell complex by splicing to gether finitely many cells of suitable dimensions. There is a wide variety of topological spaces..D-dimensional sphere (or (n I)-sphere) by The O-sphere SO consists of two points {± I}. and we will call them topological spaces. The boundary surface of a ball is a sphere. 2.Xn) tr: = { (Xl. Basic spaces For a natural number n n-ball) D" by > 1.CHAPTER 2 Topological Spaces and Cell Complexes There is a large selection of geometrical objects around us.. we define the n-dimensional ball (or . The boundary fJDn of the . but run up to n in general.







n-disk Dn is the (n -I)-sphere


that is,

aDn = 8n-1.
The interior of a ball D71 - 8Dn is called a ball without boundary or an open ball. Denote by 1 the closed interval [0, 1] between 0 and 1. Then I and Dl are homeomorphic. 2.2. Product spaces and quotient spaces

For topological spaces X and Y the set of all ordered pairs (x, y) of points x E X and y E Y, denoted by X x Y, becomes a new topological space, which we call the product space (or the product) of 2.1. A unit square
I x I = { (Xl y)





< I, 0 < y < 1 }

is homeomorphic to D2. We write 12 for I x 1. 2.2. The product space 81 x 81 is homeomorphic to the surface of a doughnut. We call this space the torus and denote it by T2 (cf. Example 2. 15).

Consider a topological space X and an equivalence relation-e on X. We partition X into mutually disjoint subsets according to this relation; namely, elements x and y of X belong to the same subset if and only if x y. These subsets are called equivalence classes. Denote by X the family of the equivalence classes of X under the equivalence relation N; then X is a new topol~gical space in which each point is an equivalenceclass. We say that X is the quotient space ofX~(formed) under the equivalence relation N, Often we write X/rv


1 of the interval I. This circle is homeomorphic to both 51 and the letter O. Ifwe collapse the boundary 8Dn = 8n-1 of D" to one point, relation on I)" is x Ny {:} x, Y E Bl)" (we regard the boundary points as a single point), and = Dn / rv is homeomorphic to 8'.


2.4. We identify the upper and lower edges of the square /2 = {(x,y)jO < x <1, a <y< I} by (O,y) (l,y), o <y < 1, and the left and right edges by (x, 0) (x, 1),0 < x < 1.








In other words, we stitch the upper and lower edges together without twisting the square (we now have a cylinder) and then stitch the right and left edges (two circles) fugether without twisting the cylinder. Then the quotient space 12 = [2/ rv is homeomorphic to the torus T2 = Sl x S". and left edges of the square [2 = {(x,y)IO < x <1,0 <y:; 1 } by identifying the points on these edges symmetrically with respect to the point (IJ 2,1/ 2); that is, we regard (0, y) and (1,1 y), <s < 1, as identical ((0, Y) (1, 1 - y), <y < 1). The resulting quotient space j is the Mobius strip, which is well-known for its one-sidedness (see Figure 2.1).

2.5. We splice the right









2.1. The Mobius


2.6. The quotient space of the n-dimensional sphere S" with the identification of each point x of S" with its antipodal point -x is the n-dimensional real projective space (here x = (Xl,X2) ... ,xn+d, -X= (-Xl,-X21" -xn+dEsn). We denote this space by pn(IR). When n = 2, in particular, p2(IR) is the


real projective plane.

The quotient space of D2 under the equivalence relation identifying each pair of antipodal points on the boundary EJD2 = 31 is homeomorphic to the real projective plane p2(lR).
2.3. Topological sums and attaching spaces

Let X and Y be topological spaces with the intersection A = X r Y. Their union XU Y is a topological space called the topological sum of X and Y, which we might also denote by XUAY. IfXnY = 0, X U Y is a disjoint union of X and Y (Figure 2.2), A onto B by h and pretend that X n Y = A = B to construct the topological sum of X and Y. We indicate this sum by X Uh Y.

Suppose now that X n Y - 0 but that there is a homeomorphism h : B ---1 A between some subsets A c X and BeY. We paste











2.2. Topological

sum XUy


2.3. Attaching


More generally, if we have a continuous map h : A ---+ B where A c X and B c Y, we can still make a quotient space of the topological sum X U Y by identifying each b E B with every a E A such that h(u) = b. This is the attaching space X Uh Y of X and Y by the attaching map h : B ~ A (Figure 2.3).

2.7. Let X be the one-point X=A =p t,

space pt. Consider B=8D2=Sl,


Then attaching map h : B ----1 A "collapses" B to the point pt, and the resulting attaching space X Uh Y is homeomorphic to the twodimensional

2 .8.


Let X = A = S1, Y = I x S 1, B = { O} X 81, Then both A and B are homeomorphic to a circle. Let h : B ----1 A be the map that sends B twice around A, i.e., h(z) = Z2, where we think of our circles as the set of complex numbers of modulus 1. The attaching space X UhY is homeomorphic to the Mobius strip (cf. Exercise 2.2 for the proof). 2.9. X = A = SI, Y = D2, we have the same X, A and B as above. Only If we use the above h : B ----1 A, the attaching morphic to the space that we get by attaching

B = 8D2 = S1. Here, Y is the new addition. space X Uh Y is homeoa two-dimensional ball

Closed surfaces 2. N 2. to form a topological space. T{OU can pretend that the n-ple torus is an inflated n-person life buoy (Figure 2.4 .f by removing the interior of an embedded disk D2 from the torus 81 x 51. We build an attaching space called a (finite) cell complex inductively acconling to the following recipe. every (orientable) closed surface is homeomorphic to an n-ple torus l'\.4.2 .oDl = Di . ae~. We join two T.? 's along their boundaries to make a double torus M2• We say that M2 is a closed surface (more accurately. which is in tum homeomorphic to the two-dimensional projective space p2(lR).10. EXAMPLE Mn F IGORE 2.rention that ifJ = eO is a singleton space and that oeD 8-1 is the empty set. an orientable closed surface) of germs two.4). Cell complexes III this section we put together a finite number of spaces each homeomorphic to some open i-ball D2 .re that ~ h or} e~. Ifwe remove the interior of D2 from M2 and attach To along the boundary aD2 = 81. DEFINITION . 0 < i < n.4. we obtain a triple torus (a torus with three holes) M3J and we continue this process to obtain an n-ple torus (a torus with n holes). 2. The boundary of T(} is SI . It is customary to call each component (homeomorphic to an open i-ball) an i-cell and denote it by ei (a cell in this definition has no boundary).of et is homeomorphic to s='. The boundary. Hence the closed O-cell is also an open O-cell (O-cell without boundary). Thus YyTew. CELL COMPLEXES 13 to the Mobius strip along its boundary. we say that n is the genus of jv1n. We adopt the cow. In fact. We make a robot's glove T. A closed cell e is homeomorphic to the i-dimensional ball.11.sr>.[n for some n.

Let X be a cell complex. we have the natJ ural inclusion map i : ej --+ x(q) the natural identification map 1r : X(q) --+ X" = Xq-l UJ. ko closed O-cells " ~1 ...1 •• I -0 -0 e1.~. 1 ~~11~ -1 -0 eko -} ' ki closed i-cells k n clo sed n -cells Our construction begins with X" = Ueg U·· .14 2. and attach XCI) to XO by h. and attach X(2) to Xl by h2 to obtain the attaching X2 == Xl Uh2 We continue this process till we reach Set We have now exhausted our ingredients.. x(n). For each q-cell eq. -} •• .. .. U Bet.. U (a disjoint sum) and ex(l) = U U . Ue~o' which is a disjoint sum..x(q) and th e inclusion map L : Xq ---1 X The composite of these maps 1 1 ¢j = '-' 0 1r a i :ej ~X is the characteristic map of the cell eJ. X" is the q-skeleton of the cell complex X. h2: ax(2) ~ x'. The restncnon or the cnaractensnc map ¢j to the boundary aej agrees with the restriction of the attaching map hq : ex'» ~ Xq-1 to the boundary Be. .. oet ee~ ei u e1 ~ et ey hI : ax(l) x". and obtain the attaching space Set X(2) == u e~U . 0 < q < n.. . TOPOLOGICAL SPACES AND CELL COMPLEXES INGREDIENTS.... e2. U oet. Specify an attaching map CONSTRUCTION. Specify an attaching map ei Xl _ XO Uk1 X(I).. space X(2). U eL and ax(2) = Bey U Be§ U . Set x(1) == U· .. The final product X = X" is an n-dimensional cell complex. For each q.

..13... This structure suggests that if one opens up the torus along a suitable pair of circles intersecting at a single point. l (eO Uhl e1) "" S1 is the map which sends SI 2.. -n n If we attach a closed n-cell en to sn-l .. one will get a square (Figure 2. = eO Uh sn-l n _ 1 en-1 by the 111(:1-1-' ii.. REMARK. THEOREM 2... and one closed 2-cell e-: SI x 81 = eO. The torus T2 Sl x 51 consists of one O-cell two closed l-cells.. The theorem below is an obvious consequence of the above definition. EXAMPLE ---1 = (eO Uh e1) Uh2 e2.vq 1 2 (a disjoint sum) must always be attached to some subcomplex X" by an attaching map: r <q.... ....... The n-dimensional sphere S" is a cell complex consisting of one O-cell eO and one closed n-cell en with the attaching map EXAMPLE that is.15. . : ox'» Dn EXAMPLE = s::' Uh n-1 -7 Xn-1 = n the resulting cell complex is an n-dimensional ball = (-0 e -n-l) e U h IS • -n 2. -.. one closed l-cell e1 and one closed 2-cell e2: p2(~) where h2 : ae2 '" S' around 51twice... that is. . lue11LlLY .. 2. but the boundary fJX(q) of X(q) = eq u e5 ... u e1. (eO Uhl (ei u e~)) Uh2 e2. A cell complex X is a union of cells (without v P....q q: boundary).. ... "" ..5).... The real projective plane p2(IR) consists of one closed O-cell eO. S" = eo Uh e .12..14. ei and e-:z. It may be that cells of some dimensions are missing in our construction ..

2 Show that the attaching space X Uh Y of Example 2.1 In Example 2. What about the n-ple torus (torus with n holes)? . 2. In the following chapters our topological space (a topological pair) will always be a cell complex (a pair of cell complexes) unless otherwise stated. 2. we defined the torus as a quotient space of [2 = { (x. For the cells e~ of the cell complex X and the cells e ce s eJ x ez .l y) E I yields the two-dimensional sphere S".3 Represent the double torus (torus with two holes) as a cell complex. SAMPLE PROBLEM Y of cell SOLUTION. we say that (X.1 The basic building blocks of manifolds we use in this book are balls B" and spheres sn-l which are the boundaries of B": 2.3 We glue two spaces using an attaching map to make an attaching space.16. Show that the further identification of (x. Summary 2.2 Equivalence relations on topological spaces define quotient spaces. 0 < y < 1 }.8 is homeomorphic to the Mobius strip. A) is a pair of cell complexes. Torus -0 e 2. Show that the product X x complexes X and Y is again a cell complex. Exercises 2.5. y) I 0 < x < 1. y) E I with (1 x. In this case A is also a cell complex: whose attaclring maps are the restrictions of the attaching maps ofX.4. When X is a cell complex and A c X is a subspace consisting of cells of X. 2.IG 2 TOPOLOGICAL SPACES AND eEl COMPLEXES e2 FIGURE 2.

For a natural number n. sn-l). 17 . xo). Then the pair (X. On the other hand the second homotopy group of a geometrical object measures the degree to which one can shrink a large piece of cloth. that there is a pool of water which your lasso is enclosing (We assume that the interior of the pool does not belong to our geometrical object). however. B) (tw 0 maps are ill a same homotopy class IT they are pairwise homotopic). If there is no pool of water on this grassland the lasso will smoothly converge at your feet. to that point while keeping the cloth always in the object. A) = (In. Homotopy sets Recall that in Section 1. We habitually drop the curly braces around Xo and write (X. spread out with its border being gathered at a single point. Imagine.3 we denoted by [(X. 3. In this case you cannot bring the lasso to a single point without getting it wet. A)~ (Y.A) to another pair (Y. you might skip this chapter and proceed to the next one.CHAPTER 3 Homotopy Suppose that you are standing on a geometrical object (say on the Pampas). aIn) is homeomorphic to iD". By a topological space X with a base point Xo (we might simply say a pointed topological space) we mean a pair w hose second component is a singleton subspace {xo}. You toss a lasso and try to shrink it to a single point at your feet (the lasso must stay on dry ground). If the above explanation gives you some idea of what is going on. The fundamental group or the first homotopy group of the Pampas measures the degree of the possIbIhty: In shrInking the lasso to a pOInt. B)] the set of homotopy classes of the continuous maps from a pair (X. set X = In and A = BIn.1.

3. Xo)] N {one point}.1. HOMOTOPY A pointed topological space (X. The family of maps . xo)]. (X. To see this. an arbitrary map topic to the constant map fo : (In. Let X be the topological space representing the letter X and let Xo be the crossing point of X. seemingly more complicated at the first glance.xO) ---+ (8. 8In). defined by it (x) = t h (x ) .3) constitute one of major characteristics of tile space X. XO). (X. (X. ---1 I". we get the following one-to-one correspondence THFDREM There exists a natural You will see that the set [(f'. BIn). xo) is XE homo- io(x) (In. n = 1.fI. In other words.ttl t E [0. BIn). (X. (X.. xo)] has as a representative some map f : (Sl. 3.. consider the maps II: (In. 8In) ---+ (X. §§ 3. xo) with its rotation number (a negative . 1]. 3. BIn). cf. (SI. gives a homotopy between fo and . 3. aln) ---1 (X. is easier to handle than [(8'. Pick a base point Xo on the n-dimensional sphere sn so that we have a pointed space (sn. xo). it: (X. BIn) = XO. when we give a group structure to these sets. Xo) . Denote by X the letter 0 = 51 and by Xo its apex. all). (each x E I" It multiplies the distance from Xo to /1(x) by t). aln) __. These homotopy sets (each of which turns out to be a group. . (X. We also get a pair homeomorphic to (sn) xo) by sqlleezing the bOllndary 8P~ of the n-ball I" to a point Since we can represent an element of [(r'. Then for each natural number n we have EXAMPLE [(In. Xo) 1 by some map f : (In. XO) 1 in the following section and beyond. where f(8In) = Xo.2. xo) and natural numbers n determine the homotopy sets [(In.2. Then we have a one-to-one correspondence since every element of [(11. xo).2.. EXAMPLE 3. Xo)]. xo).18 3.

xo) is the n-th homotopy group of the space (X. . [0. Yand [X. (X. BIn).xd]. we have the following theorem. and we obtain the following results. however.3. (X.4. Xo)] '" [(In. where (X. Suppose that X is a connected topological space (note. xc)].4.xo)] N [(In. and such maps with the same rotation number are homotopic. When X is a cell complex.T/ithistinct base points d are isomorphic. xo) is a finitely generated abelian group isomorphic to a direct sum of some copies of the infinite cyclic group Z and some copies of finite cyclic groups Z/(Pi). whose simple proofwill appearin§3. we have N {one point}.xo)· For a connected space X. The first homotopy group 7rl(X. 7fl (X. we have a one-to-one correspondence THEOREM [(In. 3. FUNDAMENTAL GROUPS 19 integer IT f changes the direction of the rotation) indicating how many times the map f makes Sl wrap around 81. which we saw in Example 17. in introducing a certain group structure to homotopy sets in the next section and beyond we will see that base points are very important indeed.(Sl.2. BIn). BIn). that for a finite cell complex connectedness and arcwise connectedness are equivalent). YJ rv 0. (Sl. in particular. if you are an advanced reader.8In).2.5. xo) and say that 7fn(X. Xl)]' The reader might perhaps conclude from this theorem that base points are not necessary. xo) is especially important. More generally. PROPOSITION 3. the respective n th homotopy groups ofitB pointed spaces "J. xo) is finitely generated but not necessarily abelian. Then for any pair of points Xo and Xl of X and each natural number n. Fundamental groups We give a group structure to the homotopy set [(f'. 3. if 7fl (X.3). xo) = {I}. For the letters X. 0] Z. Evidently taking a different base point spondence Xl in Sl gives the corre- [(In. (X. then for n > 2. We denote this group by 7fn(X. Xo) is a pointed topological space ($3. The same argument carries over to homotopy sets of topological spaces without base points. and so we often write 7rn(X). BIn). 7rv (X.

the reason being that for any map f : (1. 1 a 1) ~ (X. xo)] with the above "product" satisfies the group axioms. aI). In this section we study the case n = 1. 1/2] and [1/2.81) ~ (X. and hUh = h on [1/2.81). xo). xo). We say that a topological space X is simply connected if X is connected and 1fl (X.1]). a2 E [(1. imagine the interval I as the time interval (so that tEl reads" at time t"). xo) ~ {l} (the unit element). 1]into the subintervals [0. Any change of h or h through a homotopy results only in a change of huh by a homotopy. Divide the interval 1= [0. 11'1 (X. (X. 2. we can define a continuous map huh : (1. ~r'~"fln'n i= . represents the unit element Of7rl (X. xo) into a group. fo u f and f U fa are both homotopic to f.1/2]). fundamental groups. What sort of maps f : (1. Since both It and 12 map 81 to xo. For two elements C¥l and a2 in [(1. 81).I in of Ii this selection can be anything as long as it is homotopic to Ii). We make the homotopy set [(1. 81) ~ (X. Then the map fo u f stays still at TO for the first .and we have traditionally come to call it the fundamental group of (X. xo) represent the unit element of the group SOLUTION. When X is connected we often write 1fl (X).xo) by h u/2 = 11 on [0. 1/2] (identifying J with [0. To see this. xo)] we define their "product" «i . (X. xo)]) determined by a pointed topological space (X.81)' (X. xo)] i = 1. TO) the resulting group. Xo The constant map fo defined by f( 5) = xo. (X. It : (I. xo). SAMPLE PROBLEM 3. hence The homotopy set [(1. We denote by 1fl (X. a1) --1 (X. 1] (identifying I with [1/2. TO).1] both of which are homeomorphic to I = [0. which we call the fundamental rou or the rsi homoto of (X. xo).6. You may perhaps be familiar with the fact that Cauchy's integration theorem holds for holomorphic functions defined on a simply connected domain in the complex plane. S E I. 1].

30 seconds, say, and then hurries up to cover in the next 30 seconds what f would cover in a minute. If we decrease the interval of time during which It U f remains still from 0 to (1/2 1/2t) minute, then we have 11 = i, and the family of maps it is a desired homotopy.



3.7. Suppose that I : (1, (1) --t (X, xo) repE tt 1 (X, xu). Find a map that represents the inverse a -1.


Denote by a the inversion of the interval I about its
, a




f 0 a : (1, aI) maps f U (f 0

(X, Xo) represents a- 1 for the following reason: the a) and (f 0 a) U f are both homotopic to the constant

closer and closer to O.

3.8. We saw


Example 1.7 that there


a bijection

[(I] ,811),




The rotation number of the product of two elements in 7r) (SI) equals the sum of their respective rotation numbers, and hence we have a group isomorphism


To show the first isomorphism, let

I : (I", aIn)


iD", *) repre,

by it (X) = t i (X), t E [0, 1]. Then the family of maps {it}, t E [0, 1], is a homotopy between f and the constant map taking the entire In to the center point 0 of Dn, For the second isomorphism, notice that if n > 2, then for an arbitrary map from (In, oIn) to (sn, TO) there is a point Xl (# XO) in such that, by perturbing the image of the map by a small homotopy if necessary, we can make it avoid Xl . Since S" - {x l} is homeomorphic to the interior of Dn, the proof reduces to the case for D".


3.10. We consider the fi ure ei ht: that is two circles joined at a point, say Xo (the choice of the intersection point is immaterial). Let a E Hi (co) be the homotopy class of a map going once around the left circle only, and let (3 E 7f1(00) be the homotopy class of a map going once around the right circle only (cf. Figure 3.1). In this situation we have



free group generated by









i. e., every element of

7f1 (00)

has some expression

Note that a{3 # {3a, and hence


(co) is not abelian

3.11. We have the isomorphism

(31 x Sl)



The homotopy class 0: of a map of 31 into 81 x 81 going around its first component Sl once, generates the first component Z, and the homotopy class (3 of a map of 81, going around the second 81, generates the second Z. The following simple argument shows that a{3 = {3a. The product Sl x S1 is homeomorphic to the torus (Example 2.15) that we constructed by gluing together the upper and lower edges of [2 by (x, 1) (x, 0) as well as the right and left edges by (0, y) (1, y). Then a map of 81 into itself going around the boundary of 12 once represents aj3a-1 {3-1, which is homotopic to the constant map that sends the entire S 1 to the center of 12; hence 0;(30:-1/3-1 is the unit element of the group.

3.3. Higher



We adopt the same trick that we used for n = 1 to define a product structure on the homotopy set [(f', tun), (X, xo)] for n > 2; that is, we decompose the n-cube In as T' = ([0,1/2] u [1/2,1]) x 1-1

(I u



Suppose two maps BIn) ~ (X, xo), i = 1,2, represent 01 and a2 E [(In, 8In), (X, xo)] respectively. Then we define the product
0:'1 02

Ji ; (In,

to be the homotopy class of

hUh : (In, 8In)

(X, xo).

We denote by 1fn(X, xo) or simply 7rn(X) the homotopy set with this product, and say that 7r(X, TO) is the n-th homotopy group of the , topological space X. We show in the following theorem that the n-th homotopy group of X is abelian for n > 2.

3.12. For every natural
a2 = 02.


group; i.e., al.

number n > 2, 1fn (X) is an Q1 for aI, G:2 E [(In, aln), (X,xo)]. for n = 2. The proof is exactly

We prove the theorem the same for the case n > 2. We can write


= =

([0, 1/2] x [1/2,1]) x I ([0,1/2] U [1/2,1]) x ([0, 1/2] x [1/2, 1]).

In matrix form, we have




I2l =

[0, 1/2] x [0, 1/2]' [1/2,1] x [0, 1/2],

112 =

122 =

[0, 1/2] x [1/2, 1] [1/2,1] x [1/2, 1],



on 111 U hI; on

t., U h2.

Through a suitable homotopy we can change it and fz so that they map 121 and 112 respectively onto the point TO. Denoting by * the constant map taking hI and h2 to xo, we can write



h) ~

(II 12 . *)

We think of 12 as D2 (since they are homeomorphic) and rotate D2 through the angle ni: Then the composite

changes from ( ~l to ( ~2 continuously as t goes from 0 to 1. The last matrix represents a map homotopic to h U !I and hence we conclude that !I U 12:::: 12 U 11; that is,




Here are several examples. Srppose we have a map f : (X. . BIn). cation of points which differ componentwise by integers. IS a group homomorphism. ye).15. n > 2. EXAMPLE 3. and the conclusion follows.Z) = 0.(Y. EXAMPLE 3. xo) ex. : 1fn by g : (In. . 1::. -t (Y. Yo).xo)]-t [(In. we regard SI as the quotient space of IR under the identification of points that differ by integers. BIn) from the definition that [. for instance. The Hopf map f: 53 -t 8 representing the generator 1 E Z will come aboard as the projection of some fiber bundle in Chapter Eight. . EXAMPLE ° n > 2. .. and so it is isomorphic to a direct product of finitely many copies of the infinite cyclic group Z with finitely many copies of some finite cyclic groupsiZ/(Pi). B. xo) ----. n < k. HOMOTOPY Ifa cell complex Xis simply connected. Then we can show that 7rn(Sl X 51) '" 7fn(1R). Homotopy mvarrance . 71n(Sk) n >2' . Then we can show that7frJ81 )'" 1fn(IR). and hence our claim. 3. (X. = 0. YO)] I that sends the homotopy class generated to the homotopy class of fog: (F'. XO) -t 7fn(Y. In order to show that 7fn(Sl) = (n > 2).4. It is easy to see (X. Yo) an induced map [«: [(In.24 3. then for n ~ 2. BlnL (Y. 71n(Sn) N k> 0'. 7fn(X. xo) is a finitely generated abelian group. n n (letters A.13. 8In) ___. 7f3(S2) 2 ~ Z.

t E [0.3). The proof of homotopy type is not straightforward. If X is a connected topological space. 3. Consider a continuous curve Xl EX.4. $1. Xl) for every ti.. where tE [0. Xl) for each natural number n. xu) I'V l'fn (X. the pointed spaces (X.xd. then for any two points Xl and Xl we have a group isomorphism THEOREM 7rn (X. Xl. Yo). We define a family of maps from T' to X taking the boundary of I" along this curve. and hence we have an isomorphism 71n(X. yo) have the same homotopy type.17. and so we will directly establish isomorphisms of homotopy groups in the following N 3. xo) 7l"n(X. 1]. This implies that there exists a homotopy equivalence between (X. X is a manifold). xd for all n. HOMOTOPY INVARIANCE 25 The following theorem will become evident as we recall the definition of homotopy equivalences (cr. Similarly we get homomorphisms so that we have o . xo) and (Y. 1fn (X. xI). xo) -1 (X.16 (homotopy invariance). there is a homeomorphism li : (X. and so we have group isomorphisms THEOREM r-. xo) 7l"n(Y. which naturally defines a family of homomorphisms PROOF. for a connected cell complex X. xo) and (X. When a connected topological space X enjoys a very nice geometrical property called homogeneity (for instance. xI) have the same homotopy type for any pair of points Xo. not necessarily homogeneous. Xl) for any pair of points Xo and Xl ofX.xo) ---)0 71n (X. xu) and (X.3. 1]. xo) ::::7rn(X. Actually. however. ht : 7rn(X. If two pairs (X. then for each n we have a group isomorphism 7rn(X. which connects Xo and Xl.

bOlll1dmy to a pointed space has a (natural) group structure.1 The set of homotopy classes of the maps from the pair consisting of the unit interval and itt.3 Compute the fundamental group 1l'1(M2) of the two-holed torus M2· . Exercises 3.3 The set of homotopy classes of the maps from the topological pair of I" and its boundary to a pointed topological space becomes an abelian group for n > 2.1 Show that 7rn(Sk) = 0. 3. 3. n < k. 3.2 We say that a topological space is simply connected if its fundamental group is trivial.26 3. HOMOTOPY Summary 3. and we call it the n-th homotopy group of the pointed space.2 Determine the fundamental group 7r1 (P2 (IR)) of the real projective plane p2(IR). 3. and it is called the fundamental group of tills pointed space.

1. 4. therefore. we denote by Hp(X.MAT00030257462 CHAPTER 4 Homology We are now ready to discuss homology.(X) = 2:::. while actually computing the homology groups of simplicial complexes in the next chapter. . however. any abelian group G can appear as the O-th homology group ho(pl) of a onepoint space pt. P = 0. and so we conclude that they are not homeomorphic. and hence their direct sum h. although we will not give a direct proof. you might come to appreciate these axioms.0 hp(X) assigns to a topological space X a family of abelian groups hp(X). G) 27 . you may simply skim through tills chapter. (X) is a homology of X.. In this sense we might say that homology groups are basic topological invariants. rest assured that a complete homology theory that satisfies homology axioms does exist. Our approach in this chapter will be neat and tidy for the reader who is inclined towards abstraction.0 hp(X) with the property that if X and X' have the same homotopy type (this is of course the case if they are homeomorphic) then hp(X) rv hp(X') (isomorphic as groups). In fact. for each p. The direct sum h. Homology theories are not unique. . If you prefer a more concrete approach to the subject. We start with the axiomatic treatment of homology. the history of homology reveals how great mathematicians in the present and past struggled with computations before finally formulating suitable axioms to ease their work (or sharpen their results or whatever).. Homology groups A homology theory h. In particular.(X) = 2:. and each summand hp(Xp) is the pth homology group ofX. Thus if for some p the p-th homology groupshp(X) and hp(X') are not isomorphic. When ho(pt) = G. then X and X' are not of the same homotopy type.2. G) the p-th homology group hp(X) of X and by H*(Xi G) = 2::::0 Hp(X. 1.

H* (X.). and so on.: (a) Two homology theories with the identical ho(pt) agree on every h. H* (X. lR). among which are the real numbers JR. Homology axioms Let X be a topological space. We say that H* (X. G) is the homology of X over the coefficient group G (or simply over G). We Inention in passing that the homology group Hp(X. If G is abelian. A) we will certainly have one for (X. Z) (the universal coefficient theorem).(X) for any cell complex X. and if Gis Z then Hp(X. G) algebraically from H*(X.2. In this book we use only (finite) cell complexes. however. JR. Conversely for an arbitrarY abelIan group G there masts a homology theory h. A).t a finitely generated a belian group is a direct sum of finitely many copies of the infinite cyclic group Z and finite cyclic groups 'lLq of period q~ where q is a prime number. Thus H*(X.) with an infinitely generated coefficient group 1R turns out to be the direct sum of a certain number of copies of R. the pth homology group Hp(X. 0) = X .(X) £.:0 hp(X) in many different ways. Accordingly.(X) = 2:::=0 hp(X). Recall that we defined a pair to be an ordered pair (X. we can conceive homologies such as H* (X. 1. One can indeed derme a homology h. for the following two reasons we will postpone giving a certain simple definition till later and start here by taking the basic properties of homology theories as axioms which hold no matter how we defme ti. 1. but then in many cases H* (X. and besides we can understand how to define . In a later chapter we will show that for any abelian group G. the complex numbers C. (b) Computing homologies from the definition is difficult even if we defme them first. 0). The numbers of copies of the respective distinct summands dictate the topology of the space X. the rational numbers Q. etc..28 4. When we have determined a homology theory for every pair (X.. There is a wide variety of abelian groups not finitely generated.) is the most basic homology here. where A is a subspace of X (cf 1 3) If A is the empty set.) is a direct sum of a suitable number of copies of Z and Zq's.:0 hp with ho (pt) ~ G. 4. 1. Zq). and this means that if Gis finitely generated so is Hp(X. we can calculate H*(X. What sort of abelian groups do we know? We have a famous theorem (the fundamental theorem of abelian groups) which says tha. we simply write X in place of (X. G) is also abelian. lR) is easier to compute. = 2::. HOMOLOGY the homology h. G).

where ker jp is the subset of Gp mapped to by jp and imjp+l is the image of Gp+ 1 by !p+ 1.4. A). (1) To an arbitrary continuous map f : (X. . In order to state the properties of homology groups inclusive of all conceivable coefficients groups. HOMOLOGY AXIOMS 29 a homology of X more clearly when we include pairs. then --1 !* = t. where p is 0. 2. however. A) '8. A) hp(X'. 1. A) --1 (X. assgns to each topological parr (X. A) determines homomorphism id.1. .:0 hp is a homology theory jf h. (XI..A') .A). : hp(X. which satisfy the following properties.2 . .BnA) -+ (AUB. (b) If we have a second map g : (X'. (X. Given two topological spacesA and B. = 2:. there corresponds for eachp a homomorphism of abelian groups AXIOMS ° [. By setting hp(X) = 0 for p < 0. p = 0. A) -+ hp(X". Once again recall that our pairs are always pairs of cell complexes. we denote by hp the p-th homology group. we may even assume that p E'z. A) --1 (X". A") then the composition (g 0 f)* deteTIrrined by g 0 f : (X.2. A). A) into (X'. are homotopic (f ~ t' : (X. since i(B n A) cA. --+ ---+ is exact if we have ker !p = imfp+l . : h (X. The inclusion map i :B ----1 (A u B) determines the map of pairs i: (B. : hp(X. B n A) and (A u B. A") satisfies (g (c) 0 1)* = g* 0 [.. A') H>MOTOPY AXIOM. Aj. A) If two lImps -+ f and f' of (X. A).. in our consideration. 4. A) abelian groups hp(X.1 (HOMOLOGY AXIOMS). A sequence of abelian groups and homomorphisms -+ Ip G p+l !Jp+l Gp ---+ G p-l !p-l Gp-2 --+ . A'J. A) --1 (X'. A')). We say that h. A') ~ (X" . A) --1 hp(X'. A) --1 hp(X. : hp(X. we form pairs (B. the identity satisfying the following three properties: (a) The identity map i: (X. . A ').

1. A') and for each p we have a 0 i. satisfying the homology axioms. h*(X.30 4. The computations of homologies of various spaces that you will read ill the following' sections will shoo some light upon this fact .= (fIA)* 0 o. To a pair (X. Suppose that G is an abelian group. is the coe We sa thatG = theory h.A) induces an isomorphism For a topological pair (X.':0 hp(X. We can prove that for cell complex pairs (X.A) = 2:. A) with the natural inclusion maps i : A ---t X and j : X = (X. HOMOLOGY (2) BOUNDARY AxIOM. (3) EXCISION AxIOM. the following diagram commutes: f. for eachp: (we often drop the subscript p and simply write 8). In short.A). A) there is an exact sequence (4) EXACTNESS AxIOM. A) there is a unique homology group over the coefficient group G. B nA) ---t (A U B. For each p. the inclusion map i : (B. 0) ---t (X. called the boundary homomorphism homomorphism or differential). A) there corresponds (connecting a homomorphism. This completes AXIOM 4.. where pt is a one-point space. such that for any continuous map f: (X. A) --t (X'.

A) --t (X'. The homology axioms lead to the following theorem. A. A) = = Hp(X. The inclusion map i :X --t X is the same as the identity map id : X ---t X. however. The exactness of the above sequence implies that i. IMMEDIATE CONSEQUENCES OF THE AXIOMS 31 When G is the coefficient group of h we write hp(X. A) -I. Then from Axiom 4. hp(pt) f= 0 for some p > 1. A.O. Aj is an homotopy equivaeach natural number p. For any X. Hence we have hp(X. known as the homotopy invariance of homology groups.A) --t Hp(X/~ Aj is an homotopy equivalence. lence. . A'. = id : hp(X) -+ hp(X). Immediate consequences of the axioms (a) HOMOTOPY INvARIANCE. G) --+ Hp(X'. . 4. for instance.3. or hp(X.2 . G) --+ H(X. A) such that 9 0 f id and fog ~ id. If f : (X. X) is a zero map and 8p+1 : hp+l (X. such as K-theory (that comprise one of the main topics in the Iwanami series: Developments in Contemporary Mathematics). for the following reason. G). there are some important ones. A) h. f* 09* =id : Hp(X/~A'. and so i. THEOREM 4. G). EXAMPLE 4. We do not discuss this type of homology in this book.4 . is a generalized homology theory. A. G) = p=l L Hp(X. X) . X) ---t hp(X) is also a zero map. we have that hp(X.A. .3. PROOF.. then for is an isomorphism. A. G). X) = O~ p = 0. f* : Hp(X. H* (X. Aj ---t (X. there exists a map 9 : (X'. SUppose tt.2.. G). X) ~ hp(X) ~ hp(X) --t • of the pair (X.1. In this case we say that h. 0 [.3 . satisfies the axioms (1) through (4) but not (5). X) is exact. = id : Hp(X. By Axiom (4) the long exact sequence • I I --+ hp+1 (X) ~ hp+1 (X) ~ hp+l (X.3 (1) we get the equalities N Since f g1. : hp+1(X) -+ hp+l (X. (X.0 for some p < O.

X n CA) conclusion follows from homotopy invariance. the exact hp-1 (xo) -+ . we have the isomorphisms (b) HOMOLOGY OF QuOTIENT f* and g* of abelian groups. we have a homotopy equivalence (X UA CA. . Since hp(xo) = 0 for p > 0 and ho(xo) hp(xo) ~ hp(X) ~ hp(X. and we regard A as A c CA. For each nat- ural number p. but we have X n CA = A. p :» O. and so (X. Let there (X. CA). exists an isomorphism hp(X. The excision axiom implies that hp (X~ X n CA) f'.. Let f :A x {O} ---+ * be the CA = * U f (A x I) the cone over A.J hp (X UA CA. --t = G. 0 Let X be a topological space and A a subspace of X. xo) (xo E X). A) be a cell complex pair. HOMOLOGY therefore.(X) and h*(X. hp(Xj A. = (X. are different THEOREM 4. respectively. A) r-. When X is a cell complex the excision axiom (Axiom 4. xo) ofa topological space X and a pointed space (X.1 (3)) gives the following result SPACES. The D like the homotopy groups) h.pt).32 4.. We define the quotient space X/A to be the space where the set A is regarded as a single point denoted by pt. A). CA) ~ (X/A. XI)) ~ sequence . hence.4. PROOF. THEOREM 4.5. pt) constant map of the product PROOF. We can extend the homotopy which collapses CA to a point to a homotopy of the sum X UA CA still collapsing CA to the point (this is possible because X is a topological pair of cell complexes).

f : X.3. =0 group oc We say that h. xo). hence. IMMEDIATE I"V CONSEQUENCES OF '!HE AXIOMS 33 yields hp(X) hp(X. xo). hl(X) '" h1(X. xo). f o i = id: Xo ---7 xo. the middle of the above short exact ho(X) :: ho(X.tXo induces the homomorphism f* : ho(X) ge the constant map ---t ho(xo). hp(xo) = 0 for every natural number p.(X) is the reduced homology theorem these groups are related by ofX. A) simply by For an arbitrary map f : (X. A) ---t (X'. = id : ho(xo) ---1 ho(xo). ~ We define the reduced homology group h*(X~A) ofa pair (X. The reduced homology group of a one-point space 18 always zero: h*(xo) = 0. p :£0) EEl ho(xo). xo) -+ O. > 1.. and we f* as 0 i.4. i. It follows from hp(xo) = 0. By the above FXAMPLE 4. We may also define the reduced homology group hp(X) to be the kernel of the homomorphism f* : hp(X) --1 hp(xo) induced by the constant map f : X ---1 Xo. The expression ecomes h*(X) = L hp(X). ho(X. is injective and we have the short ho(X) exact sequence o sequence splits: ---1 ho(xo) ~ ». that oc D Set hp(X) . Aj we may define the reduced homomorphism . p > 2.-.hp(X. Furthermore. that is..6. Because [. is the left inverse of i..

. p > 0. . 4. u. hp(Xl' 0) -----+ hp(xo) N hp(SO.) to be the restriction of the boundary homomorphism 8 of the homology groups.xo) hp-1 (xu) p > 0.n. A) hp+ 3- ~ (d) HOMOLOGY hp(A) ~ hp(X) ---1 • Since the n-dimensional ball I)" has the homotopy type of a singleton set.pt) rv hp(Dn. in other words. p =I. lip \ .8Dn) N kp(Dn. ~ N The long exact sequence for the pair (SO. .xd. hp(So) = 0.8.. .ur-. xo) ~ together with hp(xo) = ho(So) N ° N hp(SO) A hp(SO. since the quotient space D" to S". "" . hp(sn) PROOF.tVA 1_ /. PROPOSITION 4. 8Dn). I groups: hp+ dA) ~ - ---1 hp+ 1( X) ___:__. The n-sphere S" is also easy to calculate. If one examines the zero-dimensional situation carefully .. xo).f1. one can immeuiareiy estaonsn tne rouowing.. 1 (X. we have p = n.7.to be the restriction of 1*. We have a long exact sequence of reduced ho1 - mology I . We now note that.. or equivalently that ho(SO) G..1\ +r ILp -1 \.r1J ~ '.i\. . /8D" is homeomorphic hp(sn) = hp(sn. PROPOSITION G) = 0. and the boundary homomorphism v . =t (p > 0) and ho(xo) G implies that G ttl G. rv 0. 0. P 2:0. We first compute the homology groups of So From the excision axiom (AxIOM 4. For any n > {G. we have for every natural number p GRoups OF SPHERES.1 (3» we get = {xo. .

9.(b )*..1 (1) ((go !)* = 9* 0 i-). AxIOM G) --+ HI (OJ G)..4. G) is contained in the image of the map i. .1 (1).l (em-I) 7' and proceeding inductively with p = 1. sn-l).1 (1) that fo and h are not homotopic.. HI (0. by Axiom 4. D 4. G) (\. Then as fo = i 0 fa. aDn) = ~ hp~l(sn 1) ~ hp_1(Dn) (Dn. xo) and the excision axiom for SO to determine its homology groups. G) --+ Hp(O. 4. let i : Xo --+ 0 be the natural embedding.l i up l\. : Hp(xo. Our starting point was the homology group of a singleton space. and we used the exact sequence of the pair (80. we have that (fI)* = id : Hp(O. Since 0 and Sl are homeomorphic. and let h be the identity map of O. G).J G. Hence we have that (em) j ~ aDn) --t • Up\. The homotopy invariance enabled us to compute the homology groups of the n-ball D". we computed the homology groups of S" = D" / sn-l. Then by AxIOM 4. HI (xo. Thus we have that (!o)* and we conclude by OBSERVATION. EXAMPLE . Let fa be the map of the letter 0 into itself which send every point of 0 to the apex Xo of 0. which says id. using the exact sequence of the pair iD".3. sn-l). we get what we wanted. the image of the map (10)* : Hp(O. IMMEDIATE CONSEQUENCES OF THE AXIOMS 35 We also have the exact sequence " • --t hp(Dn) ~ kp(Dn. and Xl. This example suggests that exact sequences of suitable pall'S would help us calculate the homology groups of a variety of new spaces. The zero-dimensional sphere SO consists of two points z. t. we also have HI(O. Then. . On the other hand. = id. Let us review how we used the homology axioms 00 compute the homology groups of a general n-dimensional sphere.. i for the pair tD". G). By the dimension axiom we get G) ----+ Hp(O.. G) = O.2. G) ~ Hp(O.

1. and so forth as homology theories which satisfy the homology axioms.. REMARK. B) 8p+1 = j~' 0 8p+1 : hp+l THEOREl\![ --t 1$ (X. B) ~ hn(A B) ~ hv(X. A. Cech homology. B) -----t ••• exact. HOMOLOGY (e) HOMOLOGY EXACT SEQUENCES OF TRIPLES. PROOF. B) ---. o If you are one of those who find diagram-chasing too tedious to go through. B) the group (X.(A) s: hp(A. By the homology axioms. i: (A. Make the use otthe diagram shown In Figure 4. though we need some lengthy but fun diagram-chasing. A) ~ h. . B) J: (X. The long sequence hv+l(X.36 4. more general spaces one can count singular homolFor ogy. B).. We define the boundary homomorphism 8p+1 : hp+l by A) ---t hp(A. A and B. • Consider three topological spaces X. The natural determine inclusion -t maps and and (X. We then get the following exact sequence of of the triple (X. Exact sequences of triples will be quite useful in various ways in the later chapters. B) -t h*(X.h*(X. A) ~ hv+1(X. A) homomorphism B) l*: h*(A. J*: h*(X. you may instead adopt this theorem as an axiom. A). 4. we have the following exact sequences: .(A) ~ hp(X) --1 •• . Enjoy the fun game of a diagram chase applying the homology axioms repeatedly. Wecanobtain homology exact sequences of topological triples usmg only the homol- ogy axioms.hp+dX) ~ hp+1(X.10. B). B cAe X. ---. A) ~ h. B) --1 (X.

-. X2 and A be topological spaces.:s I h.~~ .-f' ~~ ".."f'Yo .. ~r~ .1 There are five homology axioms.2 For a fixed homology theory. Set ~ = (hhl h2*) and 'IjJ = mh m2*. the homology group of a pair IS equal to the reduced homology of the corresponding quotient space.. Exercises 4.B) FIGURE 4.(B) hp(A.'.. where Xi C X (i = 1.".......... ~r~~~~~~ ".. T b~~ rl-rr>11-"'".2 We identify a point on the p-dimensional sphere SP with a point on the q-dimensional sphere and denote by SP V the resulting quotient space...... A Q n"h'" ~.2).4 The homology exact sequence of a triple will be useful later..-l~~~b~~~ -f'-rr>"f'Yo ~~~~~~ the homology axioms... Show that there exists a long exact sequence (called the Mayer.... • --1 exact sequence): ~ ~ h...1..rl1.1 Summary 4... .1 MORE CHALLENGING.£>" . sq sq sq ..".. '~~bJ 1. "f'Yo"h11-1-'" -1-1...~ ~~~~ .. ~~~~~r~~~ ~~~ 1.(A) hq(XI) EB hq(X2) ~ ~ hq(X) --+ hq~l (A) --+ I II 4. X = Xl U X2 and A = Xl n X2· Denote by hI : A --+ Xi and tru .f-v. 4.. Let X."h . ... Xi --+ X the natural embeddings of A and Xi in Xl and X respectively... Xl. Calculate the homology groups of SP V using the Mayer-Vietoris exact sequence.Vietoris .. 4.

The calculation of cell complexes requires that you know the behavior of the boundary operator. but becomes quite simple to handle once you get to know it.CHAPTER 5 Homology Groups of Cell Complexes We are going to compute the homology groups (for homologies satisfying the axioms) of some concrete cell complexes. We will favor a more theoretical approach in this portion 0 our ISCUSSlOn oug t IS may contra ICt our po ICY0 introducing you to the theory with concrete examples). one can surmise the situation by saying that the calculation of the homology groups of simplicial complexes is clear in its direction but a bit complicated in the actual manipulation. First we outorganize your strategy if you are half-way familiar with these words. The boundary operator will become quite easy to see once you get acclimatized to your environment. You decide which is more palatable to you. If you have never heard of them. 39 . just skip this introductory gibberish. We will define everything in the coming sections. and we first explain how to compute the homology mology axioms. We will compute the i-th homology group as the quotient group of a subgroup of i-cycles (i-th chains which get sent to the zero element by the boundary operator) by a subgroup of i-boundaries (the image of the i + l-chains by the boundary operator). Once you finish this chapter you will be able to calculate the homology groups of all sorts of topological spaces which you treat as cell complexes. groups of simplicial complexes In short. which is not immediate from the homology axioms.

1. Then q" PROOF. . Choose a point Xo on the n-sphere S": Take qn copies n copies of the point Xo are regarded as a single point. we obtain the result o . -----+ • .1 (Xn-1 . • . an : Cn(X) --1 Cn-2(X) > O. 5.40 5. we define abelian groups Cn(X) by = hn(Xn. 8n-1 satisfies an-lOan PROOF. xn-l.. .2.For n> k = dimX we have and so Cn(X) = O..We regard as the map an : Cn(X) 0 an xn = Xn-1.1. HOMOLOGY GROUPS OF CELL COMPLEXES 5. Then Cn(X) = 0 for n < O. We call this quotient space a bouquet of 8". U8(e~)) 'V hn(sn v··· V s-. Cn(X) Xn-1) ----t an h n. (Uer. --1 Xn-2): hn(Xn. It follows from the excision axiom h hn(xn. Xn-2) »: hn(Xn. 0 0 = 0 for every integer n an-lOan Recall that we defined =i an as j* an. ~ Cn-1(X).xn-l) r-v n. The next proposition computes PROPOSITION C. Forn = 0. Xn-2) Xn-1). Let X be a cell complex and let qn be the number of n-cells in X. t. 1 so we get D oj*) o8n = O. By induction starting with qn = 1. o8n-1 oj* oan =].2. Calculation of homology groups of cell complexes Let X be a k-dimensional cell complex. and (8n- j. We denoted by Xq the qof the triple (X". Let ho(pt) '" G. (X).

we have the following: PROPOSITION 5. : OBSERVATION. We define the pth homology group hp (C) of C by DEFINITION hp( C) = ker jp/ Im /p+l' We can say that an exact sequence is a chain complex whose homology groups all vanish. Xn-1) = 0.3 and the exact sequence of (X" Xn-1) to obtain For a cell complex X and the natural sion map i : X" -+ X. hq(xn) --+ hq(X) is an isomorphism. whereas that of a chain complex refers to the existence of boundary operators which satisfy ip 0 jp+l = O. CALCULATION 41 Similarly we can show PROPOSITION 5.6. the following result For an arbitrary cell complex X. = 0. Note that this condition is a bit weaker than the exactness condition. . We shall now move to a general theory: homology groups of chain complexes. 1 We use Proposition 5. and homomorphisms jp DEFINITION 5. P > q + 1. Let C be a chain complex. Here we do not build axioms but compute aIgebraicaIIy.(C) (c Gp) of p-cycles to be the kernel of jPl and a subgroup Bp(C) (c Gp) of pboundaries to be the image of Gp+1 under jp+l' With this notation . ker jp = im jp+l' The term complex cf a cell complex refers to a combined object. such that jp 0 jp+l = 0 (i.5. We use only the homology axioms to prove the last proposition. q < n. The homomorphisms ip are boundary operators or boundary homomorphisms. ker /p c im jp+l ) for every p.e. A chain complex C consists of a long sequence of abelian groups G. for the q-th homology group we need only to consider the p-dimensional skeleton XP.1. For a chain complex C.4.. q > n. The q-th homology group vanishes on any skeleton X" if n ~ q . 5. but it nicely exhibits one basic property of the homology groups of a cell complex. that is. . we define a subgroup Z. hp(xn.5 . ---t G p+ 1 -----1 11'+1 GP -------1 JP G P _ 1 -------1 1p-1 GP - 2. SIDETRACK. p =I=- n. ..1. inclu- hq(xn) i.3.

while the identity imply a +1 = j. Xp-l) horizontally and that of PROOF. (X) o. C-'p-l (X) 8p-1 ----+ =r» ' . 1 Cp(X) = hp(XP.* = hp.(XP . We make a commutative diagram by arranging the exact sequence of a triple (XP+\XP..J Xp-2) .we can express the p-th homology group of the chain complex Cas hp(C) The boundary operators complex X."-. For each p. . The right&hand side is the homology of C (X) computed algebraically. Xp-1) 11+ 11 o lap From 1_ rv' l\. '- = Zp(C)j Bp(C). p 0 8p+1 and the fact that j* is injective .er a-p = im J.7. We give the fundamental theorem on the homology groups of a 5. . define the homology groups hp(C(X)) of C(X). long sequence C(X) of a cell C-p-2(X) a of the ---+ C-p+l (X) p+l p 8p+1 C-'p _____. there exists a natural isomorphism THEOREM hp(X) ~ hp(C(X))1 where the left-hand side is the axiomatic honwlogy of the cell complex X which gives rise to the chain complex c(X).

l 8e>. CALCULATION 43 Hence it follow s that hp(C(X)) _ ker8pj im8p+1 N hp(XP.) the generator 1 E Z of Cp(X) corresponding to a p-cell ex .. How should we interpret the boundary operator [) : Cp(X) ---+ Cp-1(X)? For simplicity we discuss the case G = Z. You might ask if you can actually calculate the homology groups of a cell complex from the right-hand side of the equality in the thethat you need to calculate some homology groups (so far unknown to you) in order to build the chain complex on the right-hand side... But from Proposition 5. 'l-'f-L* • p. .1.l e~l. Denote by (e>.. Recall."'-.. D< Z) by Z) induces eJ. Z) to define a homomorphism .i Hp(e).il ¢11 : e injective omomorp ism ¢11* : Hp-1 (eM) de.J Z ---t Hp_1(8e\. Let qp-l be the number of p-I-cells ofX. and eJ1. Xp-2) N hp(X).- p- l(e J.' When we have the cell complex. 7l) Z ---+ Hp_dxp-l.* of Hp(e).l e~J the incidence number of e). -+ Xp-1 of eA and the boundary isomorphism B : Hp(e>. : 8e>..5.. 7l) rv Zqp-l.4 one gets hp(XP+l.. Be). _ h>. We also have a well-defined homomorphism )-l'H 1(Xp-1 Xp-2·Z)NZqp-1-tH (A. 0 a: h). Xp-2)/ imB* ~ hp(Xp+1. Z) tv Z h>. So let us investigate the incidence number [e). Xp-2). ---+ Xp-1 Furthermore. Then we can write eIJeEX(p-lj We call [e). We use the attaching map h>.l Be>..i' de'Z)rvZ " - .. the characteristic map of eJ. Z) N Z ---+ Hp-1 (Xp-1.. that each p--chain Cp(X) is isomorphic to the direct sum of the-number-of-the-cells many copies of G. however... Xp-2..

the vectors iJOiJ1. 0) --t (Xp-1.etLl = 0 or ±1. Denote by da" the collection of the faces of a" of dimension less than n. . . Xp-2). . a 5. 5. Hp-1 (Xp~ll Xp-2.'l). In the coming sections. the homology groups of a simplicial complex are determined uniquely from homology axioms. pi.Pn are the vertices and n is the dimension of (In. THEOREM logy g ro up 5.44 Consider 5.J E Z of the homowith Z coefficients of a cell complex X is well-defined by oj* 0 r--- [eA. and that Po. By a regular cell complex we mean a cell complex whose characteristic maps are homeomorphisms. Then we can calculate the incidence number tnd hence the boundary operator given by (5. We say that da" is the boundary DEFINITION l.POPn are linearly independent). Z). . .: Hp-1 (XP-l. and we will look into them in the following sections. then we can determine directly from the homology axioms that [eA. The incidence number [eA) e.etLJ = (4)tL. ELI.. We first define a simplex as a higherdimensional triangle. •• . 1EZ Il.. BeA. Pn be n + 1 points of R N in general position (i.1) of an arbitrary cell complex. Pn) the smallest convex set spanned by these points. COMPLEXES the induced homomorphism Z) -1 i. Let Po. The following is an easy consequence of definitionchasing.Pn} is called a face of o". .9 (SDlPLEX). Hence. and denote by an = (Po.. .. HOMOLOGY GROUPS OF ( .e.. we will give a direct definition of a homology of simplicial complexes which agrees with the homology determined by the homology axioms. Z) of the inclusion (Xp-1.. a" is its own face. Calculating the homology groups of a two-dimensional sphere is easy if we treat it as the surface of a regular cube. Homology of simplicial complexes (a) DEFINITION OF SIMPLICIAL COMPLEX. . We say that an is the n-simplex.. Simplicial complexes are regular.B. Suppose N is a sufficiently large natural number. We will investigate the possibility of regarding more general spaces (including spaces of higher dimensions) as collections of "triangles".Pi] } of {Po. In particular. The j-simplex spanned by a subset {Pi .)-l hA*)1) EZ I"V Hp-1(efl.2. If X is a regular cell complex. .. (eA.8etL. .

0) span a two-simplex. 2. not all on the same line) defines a two-simplex.0. A simplicial complex is a regular cell complex. P3). . 5.Pl. Let S = {(Tn} be a finite family of simplexes of various dimensions in 1RN. P2).5 . A zero-simplex in S is called a vertex of S. 0). 0). then (Jl n 0'2 is a face of each of 0'1 and 02' We say that S is a simplicial complex. where N is large.P2). P3 four two-simplexes: (Tf (T5 SIX = = (PO. 0. P2 (1. 0. (T~ = (Pl. that is.11 (SIMPLICIAL COMPLEX). z)~space IR3 the pOInts Po .P2). ag = (P2).P3).P3).e. (T~ = (PI). satisfying the following conditions: (i) If (In E S. any set of four points not on the sallie plane determines a three-simplex.) f'( 5.P3).10. a~ = (Pl. We indicate that rr" by writing (JJ --< o".12.P3). . SirIlilarly. DEFINITION In short a simplicial complex S is a collection of simplexes spliced together along some faces such that any face of a slinplex ill S is again a simplex in S. P1 (1. 5. Consider four points in JR3: EXAMPLE PO (0.2 .P2. The highest dimension of the simplexes in S is the dimension of S. (J~ = O"~ = (Pl. EXAMPLE (jJ is a simplex belonging to the boundary of PI = (1. (P2. its characteristic maps are homeomorphisms . (Po. and four zero-simplexes: a? = (PO). HOMOLOGY OF SIMPLICIAL COMPLEXES 45 of o".(0.P3). Any set of three points in general position (i. (T~ = a~ = (PO. then every face of o" is also in S. 0) and P2 = (1. (Po.0).. IIi the (x. (ii) If (J11{~ E S. pd. (PO.P2.Pl. and so it is a two-dimensional simplicial complex which is the boundary of the three-dimensional simplex spanned by the Pi'S (Figure 5. a~ = (P3)' The set S consisting of these fourteen simplexes satisfy the conditions (i) and (ii).1). y.0. one-simplexes: (Ti (T~ = = (Po. Here is an example of a simplex c" whose boundary is a simplicial complex.2. 0).

.0.2.0) and p~ = (-2. If we reorder the vertices in a sim. we still have the same simplex. .0. y. z )-space ]R3. A specific . which is the quotient space of two copies of S2 attached at a single point. The s aces 8 and 3' are both homeomo hie to the two-dimensional sphere S2. DEFINITION 5.13. Pl == (1.0). P2 = (1. -3. -4) conssting of fourteen simplexes is a1so a simplicial complex.12. 5.14.15. plex a" = (Po. In the (x.46 5. Simplicial complex 5.1. Let S be a simplicial complex and denote by lSI the subset of ]RN consisting of the points belonging to simplexes of S.. ORIENTATIONS SIMPLEXES. The union s u 8' is another simplicial complex of dimension two which consists of twenty-one simplexes (the number of simplexes in S n S being seven). Since we can regard each simplex as a cell. while IS U S'I is homeomorphic to the space which one obtains from the sphere 52 by putting the divider EXAMPLE 2 in it.13. I (b) HOMOLOGY OF OF SIMPLICIAL COMPLEXES. and SUS' of Example 5. is a cell complex. S'. Look at S.12 and Exam le 5. -4) to the points in Example 5. Then lSI has the relative topology as a subset of IRN.0).Pn). -3. HCMOLOGY GROUPS z OF CELL COMPLEXES Pa x y FIGURE EXAMPLE 5. Then the boundary S of the simplex spanned by four points Po = (0. The space IS U Sf has the homotopy type of S2 V S2. add p~ = (-2.

(Po.2. EBZ is the q-chain group of S over the coefficient group Z. we set Cq(3. PI 5. When we onent an by Po. PO) = .Pnl' If we change the order through an odd permutation we add the minus sign. and those differing by an odd number of permutations are regarded as unequal.2. .5. P2 . = \PO. .2. Let S be a simplicial complex. We now construct a chain complex by defining a boundary ator 8q: Cq(S.. INTEGRAL DEFINITION HOMOLOGY OF SIMPLICIAL COMPLEXES..Z) -t Cq~l(S.17. Po. We orient the two-simplex and P2 by (5. which we show by an arrow as in Figure 5.16. HOMOLOGY OF SIMPLICIAL COMPLEXES 47 order of the vertices defines an orientation of the simplex an. Pb Po) .Pn In tlus order we wnte ((In) (pa.(P2 . P2 . Z) = 71}q = Z EBZ EB. Denote by kq the number of q~simp]exes in S.Pl) Po FIGURE 5. and we say that we have oriented an.. . Z) . for example. so that there are essentially tw 0 distinct orientations. < 0 or oper- .PO). Oriented l-simplex and 2-simplex spanned by vertices EXAMPLE PO. when q q > dim S. P2) = .Z)..O.2) (PO.P2) = \P 1. PI) = (P2 . Then we say that the abelian group k copies Cq(S.18. We can orient the one-simplex with vertices Po and PI by: -\Pl. Those orientations which can reach each other through an even number of permutations are regarded as equal.Pl. EXAMPLE 5. 5... which we can indicate by an arrow as in Figure 5.2. P 1) - (PI. PO . .

.Pq). HOMOLOGY GROUPS OF CELL COMPLEXES 5. Z).. Z) by a aq(al((Ji)+a2(a~)+. Prove Proposition 5.. For every q.Pj+l..20. .Pq)) . and so we can define a boundary homomorphism q: Aq(S... Z) --t Aq 1 (S. . I . +akq(ert») = aq (al ((Ji)) + 8q (a2((J~») + .20. .. . .+(akq is the free + '" + bk'l ((Jt) = (al+bd(eri)+ (technically we say that bk'l)(erk4) Aq(S.(er%)). . Z) abelian group generated by (eri). b y a (ai(a{)) = L( -l)jai(PO. Hence aq(ai~(Jf)) is an element of Aq-1 (8. . Z) = 71}q = Z ffi Z ffi . the composition PROBLEM..pj.. (J{. j=O ." where PJ means PJ is being deleted.Pl..pq) we define 8q(ai(aJ) kq = erq(ai(po.PJ-l.48 DEFINITION 5. (Jk4' Then form a formal set: We make this set into an abelian group isomorphic to the q-chain group Cq (8. .. ffi Z by al (eri) + "+ akq (er%) + b: \ai)+ ". er~. Label the q-simplexes by 1 through kq.19. + aq(akq ((Jk))' Thus we have defined a boundary homomorphism (w hich depends on the choice of the orientation on each simplex and the simplexes) We also set the order in which we arranged 8q - 0 for q > dim S or q < O. . If (eri) = (po. A simple calculation leads to PROPOSITION 5. Orient each simplex in 8 in any way you wish.

)jBq(S. '1.o.£4 .12.Z)) keI8q/im8q+11 is the q-th homology group of the simplicial complex S over the coefficients Z.5. Z) '::'_ -4' C1 (S. Z)). depends only on the simplicial complex 8 and does not depend either on the choice of the orientation of simplexes or on the way the simplexes are indexed.21. '1. Z) = {al (lT~) + a2(a~) + a3(0'~) + a4(er2) 12:: a. Z) B2(S. IIq(C*(S. We use the orientation of each simplex and the indexing of the simplexes as in Example 5. Z) ~ ».a((J'~) + a(0'5) a(O'~) a E Z} N Z. Also. Z) ~ . e. C (3. we get a chain complex C. Z) = Co (S.. Here we compute the homology groups over ::2 of the simplicial complex in Example 5. Z) . Z) = B 1 ( S.£4 correspond to the following 6 x 4 and 4 x 6 matrices respectively: EXAMPLE t'V N By a simple calculation one can show thatHq(3. {a\ar) . Z) . im8q+1 (so Hq(S. Z) ~ . . HOMOLOG'! OF SIMPLICIAL COMPLEXES 49 Thus if dim 8 = n.Z) 1 -1 0 1 0 0 0 0 0 1 -1 1 0 1 -1 0 0 1 1 0 -1 0 1 0 -1 1 -1 -1 0 0 0 0 0 1 0 0 0 -1 1 0 -1 0 0 0 1 1 -1 1 Hence we have (i) Z2(S. 1 o~ Cn(S. ::2): Cn-1 (S. Then the boundary operators 82 : C2(3. Z) the group of boundaries.2. We say that the q-th homology group of the chain complex C * (8. '1. 5. we denote by Zq(S. We will write Hq(S.(al + a2)(a§) + (al . ker 8q? and by Bq(S. Z) Z6 -1 Co(3..22. 5. = O} rv = {al (ert) + a2(lT~) . (iii) z.12. Z) = the following: (ii) Z 1 (S. Z) ::24. Z) = Zq(S.) instead of Hq (C* (S.) the group of cycles.£6 and tJt : C1 (8.Z) ~ O. (S. Z)? i. +(a3(0"~) + (al + a2 a3)(ffH I al E Z} rv Za.::2)).2) DEFINITION ~ Co(S.a3)(0'1) rv 3 . Bo(S. (8.

and carry out the computation. Z) "-l Z. REAL PROJECTIVE PLANE P2(1R.(u~).Z2) ('. Let us compute H* (p2(IR). For an abelian group G.3 one sees its subdivision with curved simplexes. which we derive from the homology axioms. . (ii) HI (8. define the chain as C (S. Homology groups for simplicial complexes are called simplicial homology groups ExAMPLE 5. The generator of H2(S.Z2) Z) ('. HOMOLOGY GROUPS OF CEIL COMPlEXES Thus we get the result: (i) H2 (S. We can develop a homology theory over a more general coefficient group G as follows. IZ). sphere 82) and H* (8.50 5. it is a 2 of its boundary 81 identified. Z) = 0. H2(p2(IR). In order to identify a topological space as a simplicial complex we can choose to be homeomorphic to a simplicial complex in some real space IRN of high enough dimension.(O"~) + (O"~) . Z) JS exactly the same as the homology of 82. the process of finding a simplicial complex 8 homeomorphic to X is called a triangulation of X (or a simplicial decomposition). HO(p2(IR).)Z~ HdP2(IR).Z2) Z) = 0. The reader should examine the figure (Figure 5. Z) and extend the definitions. Z) ~ Z (corresponding to ±1 E Z) is generated by (ern . H1(p2(IR). in other words. The homology theory over Z is called the integral homology. Z) @ G = C (S. The integral homology and &-homology of p2 (Itt) are as follows: HO(p2(Itt). Z) ~' Z.) £::2. Z) rv /£2. H2(Itt.23. In Figure 5. Given a topological space X.).1) to confirm that this is a two-cycle which is not a geometrical boundary. rv ~ 22. The projective plane p2(Itt) is a quotient space of 82 with antipodal points identified. 22. (iii) Ho(S.

We do not give its (easy) proof here. In Example real projective plane: ElrAMPIB 2. Homology calculation of cell complexes Now that we know the homology groups of simplicial complexes.24.3.a FIGURE Recall that a triangulation of a topological space meant identifying the space with a simplicial complex. You will have fun proving the exactness axiom (Axiom 4. Any two homologies of a triangulable s are e ual over the same coe cient and so t with its simplicial homology too). The simplicial homology of a triangulable space satisfies the homology axioms.25. 2 . THEOREM 5. you can approximate any map between them by a simplicial map (sending a simplex to a simplex). If you triangulate triangulable spaces finely enough. The following theorem will shed some light on the validity of our axiomatic homology theory. 5. 5. we gave a cell division of the p2(IR) = (e" Uhl e1) Uh e2. which induces homomorphisms between their simplicial homology groups.14. as it is somewhat lengthy. we can proceed to compute the incidence numbers of a cell complex and then its boundary operators We discuss this procedure by examples. It is also easy to prove the excision axiom.1 (4) ) for simplicial homology.

2) = o. . HI (p2(JR). 5. .2 The n-th chain group (group of n-chains) of a cell-complex is a tree sum 0 nk copies 0 I S coe icien group e number of n-cells. -2 The integral incidence numbers . 5. The n-dimensional pn(c) = U(n complex projective x V(l) space + l)jU(n) has a cell division: Hence s.. Z) N 221 H2(p2(JR). 1 Hence we get the following groups: Ho(P2(JR). G\-l (X) 2) = 0 for each i. and the boundary operators of a simplicial complex. = 1 2 and so we et are [eJ' eO] = 0 and [e2.. HOMOLOGY GROUPS OF CELL COMPLEXES The incidence numbers in the integral coefficients are [el ~eO] = 0 and [e2.52 5. 8(e2) = 2(e ).2n. hence we have ell a(e1) = 0. EXAMPLE Z) r>-J Z. C. 1z i = 0.. 5.2.. The torus in Example 2. Ifi(pn(c).15 has a cell division SI x SI = (" e U hI e] u e2 -1 ~1)) Uh2 e . = 2. .4 We compute the homology of a cell complex by investigating its boundary operators. (X) ~ r: (C) = tJ Uh2 e2 U .3 We define the chain complex. therefore.27. and thereby compute its homology.26. Uh2n e2n. eJ] = 0 for EXAMPLE 5.. 5. Summary e omo ogy 0 e same as its chain complex.

1 Using cell divisions of S" and D" compute their respective integ-ral homolog-y g-roups. 5.3 Calculate the integral homology of the torus with n holes using a cell division of Jv12• l .EXERCISES 53 Exercises 5. 5.2 Derive the integral homology groups of the double toruslvl2 from the result on the torus T2 = Sl x Sl together with the Mayer-Vietoris exact sequence.

. To an arbitrary map f: (X. pair (X. of abelian groups such that where id : (X.) and each natural number p. We have a unique cohomology theory once we specify the coefficient group. A. 1. We remind the reader that our topological space P= IS a cell complex. as cohomology and homology are on parallel tracks. A'') is another continuous map. which we study in the next chapter. A)a direct sum 2:.A'..) ---+ (X". A) is the identity map. A') ---+ hP(X. we will just state the cohomology axioms and give a direct definition of simplicial cohomology and its calculation. however. which satisfy the following properties. If g(X'.CHAPTER 6 Cohomology We obtain the cohomology axioms by reversing the direction of arrows for induced homomorphisms and boundary homomorphisms of maps used in developing homology theory. is always equal to the identity homomorphism. . In any case.A).0 hP(X. We can also compute cohomology groups directly from the corresponding homology groups with the aid of the universal coefficienttheorem. . A) (1) HOMOTOPY AxIOM. there corresponds a homomorphism t' : hP(X'. 2. A) ---+ (X. A).A) of abelian groups hP(X..A.) 55 ---+hP(X. A) ---+ (X' . P = 0. then (g 0 it = f* 0 g*: hP(X". in many instances cohomology is much easier to handle.

A) and each natural a homomorphism. A) i. Z. A).2. Z) = Hom(Cq(S. B The inclusion (A u B .' Cq(S. j : X = (X. for the q-th integral chain group (the group of q-chains over Z) and define the q-cochain group over G. we have Cq(S. hP(X) ~ hP(A) -+ . -+ hP-1 (X) .ut. A') (fIA)* op. Z). 0) -----+ (X. by Cq(S. A) -----+ (XI. map f : (X. (5) DIMENSION AxIOM.56 G. G) -+ Cp+1(S. then 0. Z)). B n A) for every p. We take . . G). we label the q-simplexeswith numbers 1 through kq: ai. G). A').A) induces an isomorphism (3) EXCISION AXIOM. A) f* 0 (X'. a E Cp+1(S. A') (2) COBOUNDARY AxIoM. > = 0. Z). called the co boundary -4 bP : hP(A) 8P 0 hP+1(X. G) by =X = bq(x)(a) (8(a)): XE CfJ(S. a~. map i: (B. For a pair (X. Cq(S.A) -----+ hP(B. n A) -----+ i*: hP(A u B. hP (pt) where pt IS a singleton space. -7 such that for any continuous We often write I) for I)P. We orient each simplex in Sin any old way. A) and the natural inclusion maps i : A -+ X. number p there corresponds homomorphism. -----+ hP(X. COHOMOLOGY If f ~ /' : (X. To a pair (X.s hP- 1 (A) ~ If p hP(X. = A). A) we have a long exact sequence (4) EXACTNESS AXIOM. then f* = J'*: hP(X'. G). Fbr each q. We define operators 8P.G) In the case G coboundary = Hom(Cq(S.··· where kq is the number of q-simplexes in S. G). Let Sbe a simplicial complex. . Cohomology of simplicial complexes 6.

Then. i= i i.12.)* ((crJ)) = = 8. G). 6.{pi)* (o(pj. We say that C*(S. G» = ker8q/im8q-1 of the cochain complex C* (3. G) ° f- 0. .2. which we shall denote by Hq(8j G). First we notice that 8° ((pil*) ((PJ..22 we calculated the integral homology of the simplicial complex of Example 5.4. G) denotes the group of q-boundaries of C* (8. Z) ---t The calculation of the coboundary homomorphism C1 (3. which we denote by C*(S. . 6. G) over G of 8.(PO. By the q-th cohomology group of a simplicial complex 8 over coefficient group G we mean the q-th homology group DEFINITION Hq(C*(S. where 8j is the Kronecker delta. Here we calculate its integral cohomology.Pk)) . C1(3. 6. i: Then Gq(S. EXAMPLE G) = ZQ(3. In Example 5. C2(3. Z) N Z4.3. Similarly. Z) by ((J. G). Z) N Z6. .pd) - (Pi)* ((Pk) (Pj)) - ok - 8. COHOMOLOGY OF SIMPLICIAL COMPLEXES 57 Then we get a chain complex . The elements of B" ( 8. i 8 {1' J 0. kq.. For each oriented simplex (an. we get 8° ( (po) *) for example as 8° ((po)*) = -(p01Pl)* . +- 6 1 C 1 (S.. Z) is isomorphic to {E7!1 al (a{)* at E Z}. G)/Bq(3. We denote by Z" (3.. G) are q-cocycles DEFINITION of the simplicial complex 8 over the group G.. G) the group of q-cycles in the G-cochain complex C*(S. G) is the cochain complex of 8 over G (G-cochain complex for short).P3)*. Z) goes as follows. 6° : CO(3. G). a free abelian group generated by (an* ~i = 1. G) ~6° C (8. Z) ~ Z4. and we get the following: CO(3.2 .. Bq(8.j.(p01P2)* . we define ((Jf)* E Cq(3. We can now write Hq(3. G). Elements of zq(S. G) are called the q-coboundaries of 8.6.

E Z} Z3. - a. Z) = = Bl(S.58 6.) . Z). since most likely it makes no sense to you yet).(o-~) is the fundamental cycle which generates H2(S.(u~) + (U5) . Z) {al(ui)* + (-al (iii) + a2(O-~)* + a3(ul)* + (-al + a2)\o-l)* + a3)\ug)* + (-a2 + a3)(u~)* a. . Z) to the element I'V Z by sending an element adar)* + a2(u~)* + a3(o-a)* + a4(O-~)* {O". Z) rv Z. Z)""" Z by "integrating" the two-cochain at this fundamental cycle (do not fret about this. (i i) Hl(3. I a 1 a2 + a3 - a4 = 1=1 o} I'V Z3. Z4 to H2 (8. Z) rv r'V 0. r"v f'V Z2 (S. We can say that we realize the correspondence to the cohomology group H2 (8. Hence we get the following result: (i) HO(S. (iii) H2(3. Z). Z) = C2 (3. Z) -+ C2 (8. Thus we have the following 6 x 4 and 4 x 6 matrices representing 8° and 61 respectively: '-1 -1 -1 0 0 0 1 0 1 0 0 -1 -1 0 0 0 1 -1 0 1 0 -1 ° 1 0 0 1 010 0 0 1 -1 1 -1 0 0 0 -1 0 1 0 1 1 0 These matrices are the transposes of the matrices for the boundary operators in the homology calculation. Z) B2 (S. Hence we get (ii) Zl(S. COHOMOLOGY Similarly we can compute the coboundary operator 81 : Cl (S. Z) Z. Z) = {t Z4. (ITt) .

')9 Summary 6. 6.2 Triangulate the real projective plane p2 (JR) to compute its inte- .1 One obtains the cohomology axioms from the homology axioms by reversing the arrow in every homomorphism appearing in the homology axioms.2 We can define and compute any simplicial cohomology directly.EXERCISES . 6. Exercises integral cohomology groups.

which are generalized version of product spaces. by using spectral sequences. which are enough for the calculation of (co)homology groups. We hope that you consult the recommended reading at the end of the book if you are interested in knowing their precise definitions. Moreover.1. G. for every abelian group G we have that @ G G®Z 61 rv . The tensor product also satisfies G. we do no more than state some properties of each of these products.J @ G~. however.) G2 (29 G). For two abelian groups Gland G2. (a) TENSOR PRODUCTS. Products of abelian groups We will discuss four types of products of abelian groups. This will be handy for computing (co)homology groups over various coefficient groups of concrete geometrical objects. In Chapter Nine we will deal with the calculation of the (co)homology groups of fiber spaces. 7. Z (29 G2 rv f"'. G1 = then there is an isomorphism bi Gi and G'. G1 @ G2~ such that if G1 and G2 are direct sums. G1 @ G2 ~ L Gl Z.CHAPTER 7 The Universal Coefficient Theorem In this chapter we state without proofhow to calculate homology and cohomology groups (we will write (co)homology groups) of product spaces as well as (co)homology groups with a change of coeffic1ent groups. there is a well-defined abelian group called their tensor product (over Z)..l = by G~. We also define the cup product of cohomology groups.

bn ~ Note that these relations completely determine of two finitely generated abelian groups.1. If Gland G2 are direct sums. TORSION PRODUCTS.I( m. G) Finally. G) ~ Tor(G. THE UNIVERSAL COEFFICIENT THEOREM and for every pair of natural numbers m.. Z) . t'.. Hom(Zm' Zn) ~ Z(m. Z) ~ Z. w. (c) Two abelian groups G1 and G2 determine an abelian group. whieh depends only on the torsion parts of Gland G 2 (their respective subgroups consisting of the elements whose integral multiples become for some integers).. n and their greatest divisor (m. Tor(G1.10. In particular....l'V. j 1J" 2 ' ~ 1-1. Tor(Z.n)' Note agam that these properties completely detemrine the abelian group Hom(G11 G2) of any two finitely generated abelian groups G1 and G2. the tensor product (b) Hom. the following relations hold: l/-Jm '.G2). n) . we have Hom(G1.. the following relations hold: Hom(Zm. and tor any abelian group U we = 0. rv G. G2) We also have tor any abelian groups have that r-. n). For any two abelian groups G1 and G2 we have the abelian group Ham(G1..l yy 'V "t ..j An abelian group G satisfies Hom(Z. G2) of the homomorphisms from G1 to G2. Z) = 0.l.] G~).. L i. G 1 = Gl and ° 2:~ " ~ "'" nJ . if G1 = L:i Gl and G2 = L:J G~ are direct sums. Hom(Z..(Y common iL. 2: Tar(GL i. called their torsion product (over Z). G2) r-v Hom(G~~ G~).. Ch and G2.. Y 'V Tor(G1.62 7.

denoted by Ext( G1. Zn) rv Ext(Zm' Z) rv Zm. Zm) rv Tor(Zml Z) = 0.n). IR) = 0. Ext(Zm' IR) = 0. Zn) rv Z(m. IR) Tor(IR.1.Horn(IR. we have the following: Tor(Z. G2).) = 0.7. G2 = I:j G~. . We list the similar relations using llt Hom(IR. Z(rn. G2) for two finitely generated abelian groups G1 and G2. G2). G) We also have the following: Exl(Z. Note that all these groups are regarded as Z-modules. Horn(Zml JR. PRODUCTS OF ABELIAN GROUPS f53 In addition. in this case we set Toro(G11 G2) G1 ® G2. G2). Ext(Zm.G2). Z) rv = 0. Z) rv Tor(Z. (d) Ext. Sometimes we write Ext 1 (G1. and in this case we set ExtO(G1. Note that these properties completely determine Ext( G1. Z) N r-J IR. then there For any abelian group G. G2) . G1 = is an isomorphism I:i Gl. Z) = 0. For two abelian groups G1 and G2. Zm) = 0. Sometimes we write Tori (GI. G2) in place ofExt( G1. Note that 'we can determine the torsion product of two finitely generated abelian groups from these properties. Ext(Z. Zm) = 0. an abehan Equivalence classes of extensions of G2 by G1 determine group. we have that Ext(Z. Zm) Tor(Zm. IR) = 0. Hom(Z. Ext(Z.n). Tor(IR. ff G1 and G2 are direct sums. Tor(Zm. an extension of G2 by G1 is a group G together with an exact sequence ° ---7 Gz ---1 G ~ G1 --+ 0. I'V Tor(Z.IR) = 0.Hom(G1. G2) in place ofTor(GI.

6. Hence the composition of the cross product and the induced map . Z) ~ @ L HP(X. We have a very strong result in the following theorems. Z)) p+q=n+l 7. THEOREM 7.7.Z) ~ L Hp(X.6. We calculated the homology groups of a cell complex from its chain complex (Theorem 5. Cup products We define: in this section. Hn(X x Y. which imply that the map induced by the cross product is injective.". x) E X x X. . a cup product which gives a product structure to cohomolo to 010 'cal s ace X the diagonal rna p . Z).2. The Kiinneth formula We will now state a formuia to compute the (co)homology groups of the product spaces X x Y of topological spaces X and Y given their respective (co)homology groups. Hq(y. We can regard the tensor product of the respective chains of X and Y naturally as a chain on X x Y. is continuous. which induces a homomo hism Similarly we get the induced homomorphism We say that these maps are induced by the cross product. Hq(y. sending x E X to (x.7).Z) p+q=n Hn(x x Y.3.1 (the homology Kiinneth formula). Z) EB Z) p+q=n L Tor (HP(X. : X -----+ X X X.

G). otherwise. Z) x H2(p2(C). the cup product u : H2(p2(C).4 with its hint).:z Z. G) thus equipped with a product structure has become a ring. Exercise 7. G) and b E Hq(X. 0. EXAMPLE Hj (S2 V 84: Z) . although the complex projective plane p2(C) has the same integral cohomology . au G). Z) is the zero map (cf. Z) --+ H4(52 V 54. The definition implies that the structure induced on a cohomology ucts satisfy the following properties (which follow naturally from the properties of tensor products together with the trick of regarding the For a E HP(X.7 •3 . Z) X H2(S2 V S4. For a map j : X Y. . G) 4 Hp+q(X. CUP PRODUCTS 65 defines a homomorphism u : HP(X. c E H'{ X: G). G) = Lp HP(X. U j*(b).. j = 0. satisfies U (1. Z) ---t H4(p2(C. Hence one can show that the cup product homomorphism u : H2(S2 v 54. J = 0. we define their cup product aUb by b = 6 * (a x b) E Hp+q (X. otherwise. (a U G). 4. b) c =a ----+ U (b U c). (-. a U b = (-1) pq (b U a). . b E Hq(X. . G) x Hq(X. j*(a U b) = j*(a) The cohomology group H*(X. and j* is a productpreserving homomorphism (ring homomorphism). G). On the other hand.J Z. U G). For aEHP(X. ---1 Z) 1) = 1 if we look at U as U : . 2.:z x .

G) EB Ext(Hn+l(x. and we obtain cohomology groups over a general group from the integral homology or integral cohomology groups. Tor(Gl~ G2) and Ext(G1. . 7. Z). of the respective (co)homology groups of two topological spaces with the (co)homology groups of the product space. 7.4 With the structure of a cup product. Z). coefficient uct: THEOREM 7. Z) ® G EBTor(Hn-1(X. logy groups of a product space fall out from the Kiinneth formula. We can also compute cohomology over a general group G from the integral cohomology and the torsion prodG) r'v Hn(x. Z) ® G EB Tor(Hn+l(x. We list four formulae known as the universal coefficient theorem. n-ix. 7. THEOREM G) ~ Hom(Hn(X. cohomology groups become nn s. Cup products of manifolds are closely related to intersection theory. 2l G). 7.The above example shows that the cup product observes topo10 ical s aces in finer detail than cohomolo does' nevertheless the geometric implication of the cup product is in general hard to grasp. G) EB Ext(Hn-1(X.5. G). one defines G1 ®G2. G2). THEOREM G) cv Hn(X.6. Hom(G1. We can calculate cohomology over a general coefficient group G using the corresponding integral homology and the extension product: n-ix. We can calculate homology over a general coefficient group G by using the corresponding integral homology and the torsion product: THEOREM Hn(X.4. 7.1 For abelian groups G1 and G2.7. ZL G). eorem We compute homology groups over a general group from the corresponding integral homology groups. G2). 2). Z). We can calculate homology over a general coefficient group G from the integral cohomology and the extension product: Hn(X. G). ummary G) cv Hom(Hn(X.

7.2 Determine the integral cohomology groups of the real projective pe usmg e on e correspon mg omo ogy groups.5 Compute the homology groups of the real projective plane P2(IR) over Z2 using its integral homology groups. Z) N £:.Z) ""-'Z~ HI (P2(IR). Z) is trivial (Hint: consider an onto map F : S2 V S4 ~ 82 which collapses 54 to a point). HO(p2(IR).7. p2(IR) to compute the homology group H*(P (IR) xP (IR). Z) of the product space P"(R) x p2 (JR). . 7.1 Use the integral homology of the real projective plane. Z) X H2(S2 V 84.2) H2(P2(IR). Z) --+ H4(S2 V S4. 7.3 Compute the integral cohomology groups of the product space x P"(R). 7.5 One computes (co)homology groups over a general coefficient group using the integral (co)homology groups together with the universal coefficient theorem. Exercises 7.4 Show that the map u : H2(S2 v 84. Z) = 0.

We first consider the product space E = 51 x D I.1.CHAPTER 8 Fiber Bundles and Vector Bundles In order to investigate a curve on the plane we take its derivative at each point. too. We can do the same for geometrical objects in higherdimensional spaces.1). and to their firm establishment as powerful tools for the exploration of manifolds. we get some global information about the surface. and then we define Grassmann manifolds. general vector bundles and fiber bundles. FIber bundles Let SI be the unit circle and Dl = [-1. Evidently there exists a natural projection rr : E ~ 51 such that rr-l(U) = U x Dl holds for any open interval U in S1 (Figure 8.1 to SI. This leads us to tangent bundles of manifolds. if we look at the tangent plane at each point of a surface and observe how these planes change as we move the points. In this chapter we discuss fiber bundles and vector bundles. while building field theories for elementary particle physics. 8.1. which have absolute control over the equivalence classes of vector bundles. therefore. in investigating a higher-dimensional smooth geometrical object (differentiable manifold. 69 . In this example. Similarly. We get a rough idea of the shape of the curve from the distribution of the signs of the derivatives. Meanwhile physicists too established gauge theories. this turns out to be essentially equivalent to the concept of fiber bundles. 1] the unit interval. We see that -rr-1 (D) = U x Dl for any open interval U of 51 (Figure 8. the center of the band is a circle S1 and there is a projection tt from 1\. Next we look at the Mobius band 1\. and investigate their roles. to be precise) we must consider the linear space tangent to it at each of its points.2).

F) is a fiber bundle with the total space E. B. an . We also denote this fiber bundle by We say that the homeomorphism h between 1f-l(U) and U x F is a local trivialization. We say that n "! b is the ber over b.1.an a map 1 (b) is homeomorphic to 1f : E --+ B such that for each b E B. /Pl Then we say that the quadruple (E.70 8. 1fF and there is a neighborhood U of b with a homeomorphism h : 1l'-1 (U) --+ U x F making the following diagram commute. IT. ppose we ave opo ogre spaces. 81 x Dl U FIGURE 82 The Mohills strip Keep the above examples in mind for the definition of a fiber un e. FIBER BUNDLES AND VECTOR BUNDLES u FIGURE 8. . the base space B. where PI : U x F ---t U is the projection of U onto the first component: ]f-l(U) 10 ~ ~ Ux F U. the fiber F and the projection t: .

F) . With the natural D(l) __.H and K respectively. it is called a trivial bundle (over B). 7r. Suppose that we have two fiber bundles (with the same fiber). 8. The product space E = B x F is the simplest fiber bundle. F). FIBER BUNDLES 71 If you have read other books on the subject. consider G = SO 3 H = 80(2) and K = {e} (the trivial group consisting of the unit element). More generally.1.3.1) = SO(n + SO(n) 1) . 8O(n) and SO(n -1) for G. Then a bundle map DEFINITION f : (E. S2 is the so- Z (cf. B'. F) consists of a pair of maps ---t (E: 7r'. F) and (E'. (E. B. This is a fiber bundle known as the (unit) tangent sphere bundle. Example N ow that we have defined fiber bundles. Let G = SU(2). H = U(l) and K = {el. REMARK. n . EXAMPLE EXAMPLE H /K ---t G/K ~ G/H. = N The projection map 7r: 8U(2) = S3 --+ 8U(2)/U(1) called Hop] map and induces a generator of 7r3(S2) 3. B. we obtain a fiber bundle over the base space S" with the fiber S": 1 : s::' = SO(n) --+ SO(n + 1) ~ S" SO(n . you might be concerned with the structural group.5).8.1) SO(n . Our structural group in this book is always the group of self-homeomorphisms Homeo(F). with the fiber Sl and the base space 52. we can define bundle maps between any two of them.1. n'. Then we have a fiber bundle Sl = SO(2) -7 SO(3) --+ S2 = SO(3) / SO(2). 8.SU(2) = S3 ---t SU(2)/U(1) = S2. ifwe take the groups SO(n+ 1). and settle the question of bundle equivalences. B'.

For any map g : A ~ B we have the following natural construction of a fiber bundle D F over A and a bundle ma g = (g.j (h (b) L a homeomorphism. such that it . B. 1['. Put {ta. B. In this case for any fiber bundle ~ = (E. Let g : A --. Let (E. B. We say that (D: p. F) by the map g : A -4 B. F) and e maps f =(11112) : (E. 12 0 rr = n' 0 It). g) is a bundle map. g2): (E'. B'. B. F) D= -t (E. the induced bundle g*(J) is the trivial bundle (over A). A. 7f. and we denote it by g*(E. F) is the induced bundle or the pullback of (E. F) over B. B: F). F) is a fiber bundle and that (9. = a. 8. DEFINITION 8. 9 =(91. EXAMPLE the definition of bundle maps. e) E A x E g(a) = rte) }. F) e F) --t (E'. e) g(a. B.IT. --1 (E. p.B. e) = e. A. 11". g) : (D. F). CONSTRUCTION. B. p. A.F) (org"(~). F). and that for every b E B the restriction of 11 to the fiber 11"-1 (b) over b.4. You will see immediately that (D.rr.B. We say that two fiber bundles (E. 12 B' II : tt -1 (b) IS -4 n' . IT.5. 7f. 7f. where ~= (E. FIBER BUNDLES AND VECTOR BUNDLES subject to the condition that the diagram E 1 rr' B commutes (that is to say. lr.4 B be a constant map. IT'. . p(a. F) be a fiber bundle and let A be an arbitrary topological space.F)).72 8.

7f. Tangent bundles form a special class among vector bundles. the addition and the scalar product are 1 defined on rr. and we will use the notation for either IRn or en. which are easier to handle. Physicists use a simple expression. F) there are infinitely many bundle equivalences of B x F inducing the identity map B -+ B (since we can rotate the fibers). B. (E'. B the inverse image rr-1 (b) of b has the structure of the n-dimensional vector space (that is. Vector bundles are linear approximations of manifolds and are essential tools for the investigation of manifolds. characterize a manifold as a topological space each point of which has a neighborhood homeomorphic to some fixed n-dimensional Euclidean space ~n. we can find a neighborhood U of b in B and a homeomorphism DEFINITION vn . 8. Then 12 ((). B'. Vector bundles Perhaps you are not yet familiar with manifolds. "choosing a gauge" . n'. B'. The underlying vector space is either real or complex. and B and B' --1 be a bundle map of ~ = (E. f2): (E. In order to study manifolds. B.8. which in tum form a special class among fiber bundles. So we first define vector bundles.2. having the following property of locally triviality: For each b E B.2. VECTOR BUNDLES 73 Notice that in the above definition. 7f. vn 8. F). we look at their tangent bundles.8. and you will find an abundance of literature concerning them (see the References) everywhere you look. B. ]f. one might. F) ~ is 8.when they pick one fiber bundle which is bundle-equivalent to the trivial bundle. 1f'. and there are homeomorphisms respectively. we have g2 012= J idB. but they form a family of topological spaces which share extremely beautiful properties. By an n-dimensional vector bundle we mean a map 7f : E ---') of topological spaces E and B such that for any b E B. This example may be a bit hard to understand. In the most pithy way.(b) and satisfy the vector space axioms).6. EXAMPLE between E and E'. F) e = (E'. For a trivial bundle ~ = (E. Let (II.7. F) to bundle equivalent to the ind uced bundle EXAMPLE 8.

so you n. By a (vector) bundle map f between two ndimensional vector bundles (E. 8. !lInIS DEFINITION 1 (b) : tt -1 (b) -t 7[. h) : (E. 7rt. B. B').10. h) with the property that the restriction of !I to each fiber IT-I (b). b E B. EXAMPLE We can also define bundle maps and bundle isomorphisms vector bundles in a natural way. The Mobius strip is a one-dimensional vector bundle whose fiber is the real one-dimensional vector space. 11 = idE) iI 091 = idE'. see that an n-vector bundle is a fiber bundle whose fiber is the n-dimensional vector space. 910 B') (E. 92) : (E'. EXAMPLE You will some day come to see that the following is an example of ultimate importance. t -t (E'. B) 1f . n-dimensionalvector bundles with a given base space (of course. in a natural way. x) is an isomorphism of vector spaces 1f-l(bl) and V". Vector bundles (E. here we count isomorphic vector bundles as one) are there? many . 8. -----1 1fI. Then the space TM consisting of the tangent vectors of M becomes. FIBER BUNDLES AND VECTOR BUNDLES such that for every b' E U the assignment of x E 7r-1 (b') to h( b'.11.9.-1 ( 12 (b)) . E = B x R is a one-dimensional trivial real vector bundle.74 8. 7r. since its fiber is V The direct product E = B x is the simplest vector bundle. B) for an n-vector bundle. the total space of a real n-vector bundle whose base space isM. Let M be an n-dimensional differentiable manifold. and locally trivial maps preserve the vector structure.12. 1f. an isomorphism of vector spaces. which we call a trivial vector bundle. 1f. vn 8. 1f. B) 9 = (91. We often write (E. B. DEFINITION for 8.) we mean a bundle map f = (11. B) and (E: 7r'.) are bundle isomorphic (isomorphic as vector bundles) when we can find (vector) bundle maps f such that = (iI. r . B) and (E'.





8.13. There are exactly two distinct

real vector bundles Mobius band (cf. 8.21).

one-dimensional over 8. These are the trivial bundle and the

A BIT OF HISTORY. There exist infinitely many four-dimensional real vector bundles over the four-dimensional sphere. In 1957, J Milnor showed that among these bundles there are some which are not vector-bundle isomorphic despite the fact that their total spaces are homeomorphic; and using this, he came up with several manifolds that are homeomorphic but not diffeomorphic to the sevendimensional sphere S7. His discovery made a big splash in the mathematical community. The fundamental problem above boils down to a question of classifYjng homotopy classes of maps from the base space to a Grassmann manifold, where we utilize the remarkable result that characteristic classes which are elements in some suitable cohomology groups can detect differences among vector bundles. We will try to give a lucid explanation of this result in the rest of this chapter (and beyond). 8.3. Grassmann manifolds

(a) DEFINITION OF GRASSMANN MANIFOLD. For natural numbers m and nwithm > n, we define the Grassmann manifold GlP!.(m, n) to be the space of all n-dimensional linear subspaces of the real vector space IRm of dimension m. Similarly we define the complex Grassmann manifold GC (m, n). Can you visualize the space of the n-dimensional subspaces (each point is an n-dimensional linear subspace)? Let m = 2 and n = 1. A one-dimensional subspace of the plane JR2 is a line through the origin with slope -cc < e < +00 The lines with slopes -00 and +00 respectively are identical(the y-axis). Hence Gn:t(2, 1) is homeomorphic to the circle which is the straight line (inti) with -00 and +00 identified; that is,



GlR(2, 1) _ 31.
8.15. Let m = 3 and n = 1. Each line through the origin in IR3 intersects the unit sphere 52 in two antipodal points, so

that the correspondence of a one-dimensional linear subspace to the corresponding pair of antipodal points on 52 is one-to-one. Hence, the Grassmann manifold CIR(3,l) is the quotient space of S2 where the







antipodal points are identified; in other words, it is the real projective plane p2 (IR): GIR(3, 1) = p2(IR). It is evident, further, that GlR(3, 2) is homeomorphic to GlR(3,l) (consider the line intersecting orthogonally with a two-dimensional linear subspace), and in general, CIR(m, n) = CIR(m, m - n). We should mention further that CIR (m, n) is an n x (m - n)dimensional manifold (i.e., ClR(m, n) is a compact topological space such that each of its points has some neighborhood homeomorphic to IRnx(m-n)). If you are familiar with the definition of the k-th orthogonal group O(k), you will probably guess that GR(m, n) = O(m)jO(m n) x O(n)

So you see that Grassmann manifolds have many fascinating facets,
but in fact, beyond just being amusing, they dominate the universe of the equivalence classes of all n-dinlensional vector bundles over all possible base spaces, with the limit lim ClR(m, n) as their tsar. We pursue this topic next. m-ex::






look at the subset E = {( X, x) E

eIlt (m, n)

x IRm I x EX}

of the product space GlR (rn, n) x IRm. Tills is a topological space where every point X (an n-linear subspace of Rm) in elR(m, n) appears with a train (homeomorphic to R'') of attendants (points of H'" belonging to X). ITwe define 7r: E ---t CR(m, n) by 7r(X, x) = X it is obvious that is an n-dimensional vector bundle over the base space CR(m, n) with the total space E. We say that I'n(clR(m,n)) the canonical vector bundle over the Grassmann manifold CR(m, n). 8.16. The Grassmann manifold morphic to the Mobius strip.

(cIR:(2, 1» is homeo-

8.17. It is a bit hard to visualize might look at GIR(m, n) in the following way:

,,1 (G.rt.(3,

1». You

e (3,

1).:: p2(IR)

= (Mobius strip) U D2.





First you think of the one-dimensional vector bundle over the circle S1 whose total space is the Mobius strip (this can be a different copy from the Mobius strip in the above decomposition). Jfyou think of the base space Sl as the center circle of the Mobius band and extend this bundle over the entire Mobius band, you have the trivial one-dimensional vector bundle along its boundmy edge (which is also the boundary of D2). Now extend this bundle over all of D2. The resulting total space is E.

we may think of n-dimensional subspaces of have the following inclusion sequence: CID:(n


Since as those of lRm+1) we

+ 1,n)

c CIPI.(n + 2,n) c···

C ClR(n

+ N,n)


We also have the inclusion sequence of the corresponding vector bundles:

"t (CIR(n + 1, n))


~t (CIR(n

+ 2,2)) c

c ... We have a very important outline.

yn (CIR(n+N,n))

theorem, whose proof we will only

8.18. Let X be a cell complex and let ; = (E,n,X) be an arbitrary n-dimensional vector bundle. Then for a sufficiently large number N (in fact, N > dim X + 2), there exists a vector bundle map frum to the canonical vcetor bundle l'n (G~ (n + 11, n»; thm; is an induced bundle of




To show that there exists a vector bundle map from j to the canonical bundle "n(cIR(n+ N, n)), we need a map from E to Rn+N such that the image of its restriction to each fiber is always an n-dimensional linear subspace ofJRn+N. We can show that such a map fromE to rn:.n+N exists ifN is large enough. For instance, for a trivial vector bundle ~ = (E, 7r, B) all you need is N = O. D

Recall (Example 8.6) that if a vector bundle map of ~ to '"'f exists then its pullback is equivalent to ~. Denote by 9 = (fj g) a bundle map of the above theorem. Then g : X ~ CR(n + N, n) and



('''fn (CIPI.(n + N, n)))


So the gist of the theorem is that the bundle "'t = ,n(GlR(n + N, n)) is more complicated than any other n-dimensional vector bundle ~

n . with no exceptions. where N is at least dimX + 3. and write We denote by limN---+oo BO(n) = N-+oc lim GR(n + N.20. The proof of the following theorem is analogous to that of Theorem 8. n) with a large N. It is evident that the set of homotonv classes of mans from 81 to pN (R) also consists of two points (the homotopy class of a constant map and the homotopy class of a homeomorphism of 51 to the oneedimensional real projective line pI (JR) = Sl c pN (JR)).21. n) any Grassmann manifold ClR(n + N. . n). Since an arbitrary map f : X ~ GfR(n N. as a corollary to Theorem 8. Take X = Sl and consider the set of all (isomorphism classes of) one-dimensional vector bundles over Sl . n). 1) is homeomorphic to the N-dimensional real projective space pN OR). Now we must refer you to some results concerning . FIBER BUNDLES AND and hence that ~is induced from (actually reduced from Tn). their induced maps of the base spaces are homotopic (as ordinary maps). THEOREM 8. n) induces a pullback over X.18. Under this cireumstanee. Then for a large enough N (for instance. y\I be an n-dimensional vector bundle. Example 8. m. . There is a one-to-one correspondence between the set of all homotopy classes of maps from X to ClR(n + N. GlR(1 + N. EXAMPLE ClR(n + N. cf. f* ree ui u.13). '7T" . We say that two vector bundle maps are bundle homotopic if they change continuously from one to the other through vector bundle maps. we say that BO(n) is the classifying space for the n-dimensional real vector bundles.78 8. N > dim +3)) any two vector bundle maps from ~ to the canonical vector bundle ~n (G~ (n + N n)) are bundleehomotovic. which consists of two points (the trivial vector bundle and the Mobius band. On the other hand. and the set of all isomorphism classes of n-dimensional vector bundles over X (two bundles belong to the same isomorphism class if they are vector-bundleeisomorphic). Since there is a one-to-one correspondence between the set of all vector bundles over X and the set of all homotopy classes of X to BO( n).n VECTOR BUNDLES R 1Q Lof Y no rr '-'011 orvrn ril av: -r rrn r] lof t {H' \ .18 we get the following + 8.

Have we simplified the problem? Now cohomology plays a useful role.Lie groups.n) over X is guaranteed to be non-trivial. but not globally a product in generaL . G). complex Grassmann manifold GIC(n + N. JR) of the classifying space of n- Summary 8. then the corresponding pullback J*(/. . G) is a characteristic class of the vector bundle J* (/.1 A fiber bundle is locally a product. Recall that homotopic maps induce the identical homomorphism. n) has a rep- + N. For a non-zero c E H*(BO(n). n) = O(n + n)jO(n) J. A map f from X to BO(n) induces the cohomology homomor- phism for any coefficient group G. the Grassmann manifold GIR (n its Lie subgroup) GR(n resentation as a homogeneous space (quotient space of a Lie group by can write BO(n) = + N. G)? In the next chapter we will investigate the cohomology H*(BU(n). If a map is constant then its induced a constant map corresponds to the trivial vector bundle. we say that f*(c) E H* (X. that is. H --+c U(n + N)/U(n) x U(N) = U(n + oo)jU(n) x U(oo). The above argument has reduced the fundamental problem of classifying the n-dimensional real (complex) vector bundles to that of investigating the set of homotopy classes of maps into the classifying space BO(n) (BU(n)). and so if f*(c) E H*(X. G) does not vanish for some non-zero element c E H*(BO(n).J-+oc x O(N). G). Hence we Fm O(n + N)jO(n) x O(N) = O(n + oo)/O(n) x 0(00).n ) over X We can thus rephrase our problem: how many non-zero elements are there in H*(BO(n). n) and the classifying space BU(n) = JIll.

8. Show that the Klein bottle is the total space of a certain fiber bundle. 8. identifying the point {O}x X on {O} x 81 with the point {I} x r(x) on {l} x Sl for every x E 81. FIBER BUNDLES AND VECTOR BUNDLES FIGURE 8. . over which we have the canonical vector bundle.3.3) as the quotient space of the product space I x 81. w x morphic to the two-dimensional sphere 82. where I is the interval [0.4 There is a one-to-one correspondence between the set of all isomorphism classes of vector bundles (of a fixed dimension. say n) an which large. We define the Klein bottle (Figure 8. The Klein bottle 8. 1].2 We sa that a fiber bundle is a vector bundle if each of its fibers has the structure of a vector space.80 8.3 A Grassmann manifold consists of all subspaces of a certain dimension in a vector space.1 Defme a homeomorphism r : Sl ~ 81 of the circle to be the reflection along the plumb line dividing symmetrically right and left. s: . Exercises 8.

ketj = imi. 9. however. Given an exact couple as above. R.1. set Then d satisfies d2 dd (jk) (jk) j(kJ')k O. d .CHAPTER 9 Spectral Sequences We can compute homology groups of product spaces using the Kiinneth formula that we discussed in an earlier chapter. ker k = imj. and homomorphisms i : D ---+ D. which relate homology groups of the total spaces of fiber bundles to tho se of the base spaces and fibers. 81 . keri .jk : R ---+ at E. at D. at D. looks somewhat complicated. Exact couples and spectral sequences By an exact couple we mean two abelian groups D and F. that is. We now introduce the theory of spectral sequences. once one masters its usages one will find it to be quite an attractive and practical computational tooL Spectral sequences enable us to show the existence of the Chern classes of vector bundles which occupy an important space in current mathematics.imk. J' : D ~ E and k : E ---+ D between them with the exact triangle: D~D c-. at a glance. This theory. but we can do better. /j E.

that is...El2 :9..k(y).) W'e have thus eonstructsdthe derioed couple as shown in the above diagram .. S1?ECT. the :fact keri :.E' ---!o D' by s f' ..~. a given exact couple we derive a new exact couple as follows.iiD' : D' ~ D~ ji-1:. Hence. and define i' : D' .E~.R1JL SEQUENCES From. that is. then from y' .: 0. Take [y'] ~ [y].y ~ d(z) we get k(y') . l :D' .. where we set j'(i(x))· b'(x)] (the definition of/ is good. Set D' E' = H(E. and so we get [j(x)] == [j(x')J:). d) -= i(D) (= ker j) c D. d) _.. that is.. (~ kerd/ im"d). E'..1. for [y] E E: y E kerd. SOW'iI'lION. k' ~ gle. for i{x) E D' we have d(j(x)):. Take your time and endoy.: im i' .z) ~ (kj)k(Z') :. We define k' by kf[y] . SAMPLE P~OBLEM 9. k(y) ED'.: mi. (This k' is well-defined.~.D1 = i(D) restriction of i to IY. and so we ean determine i k' [y] as an element of Y..:imk implies x' x:. the fact jk(y) ~ d(y) ~ 0 implies that k(y) E keIj :.j(x) == jk(y) == d(y). k f : . and so j(x) determines the homology class [j (x) 1 E H{ E. ker kl' ::":: imj'. tr.key) ~ kd(.. ----7 ker(dj) ~.: (jk)j(x) == 0. since if we take Xl with l i(x ) == i(x) E D'. . ker jf :. ker d] im d .E'. The derived couple gives an exact triimker i' == im..: k(y)~ and j(x') .

Then d" : En -+ En is an endomorphism of En (dndn = 0). k) be an exact couple and consider the sequence of the derived exact couples iD".2. We say that an abelian group A is bigraded if it has a direct sum representation DEFINITION A= EEAp..:.9. hence 1 Ek :::. The spectral sequence converges if for some integer k > 0. #. d'')). Set d" = i" k".3. suppose we attach A.. in 1 jn 1 kn).. we will arrive at the n-th derived couple ill a natural way. and suppose that A is first quadrant bigraded. E.q pEZ qEZ -+ A = LLAp. . k) (En+ 1 = H (En. E We denote by Eoo these mutually identical abelian groups. For the same A and A". ti. n = 1. 9. EXACT COUPLES AND SPECTRAL SEQUENCES 83 Ifwe continue deriving a new exact couple out of the old. i. we have En = 0 for every n > k. qE'l CX) CX) The abelian group A is first quadrant bigraded if it has the expression A = LLAp. dn). .1.q pEIl qEIl has bidegree (a . q E Z we have f(Ap.q) C Ap+a. q) in the plane always stays in the first quadrant. DEFINITION 9.k+1 = Ek+2 = .2. i. In this case ker dn = En and im d" = 0. We say that the sequence (En. . j. as above. q).q. is the spectral sequence of (D E.0 to the lattice point (p. i.q+b. Then the dot to represent (p. Let (D. b) if for every p. We need some new terminology. pE'i!. En.q' p=o q=O We say that a homomorphism f: A = LLAp.

SPECTRAL SEQUENCES o FIGURE 9. p. To be more explicit. . we can prove the following Theorems 9. In fact one looks at the limit E= of a certain spectral sequence to obtain some information aboutH*(E. denote by B" the q-skeleton of B. 1 i+ -----+ Ai ------t d. Then we have the chain complex -"---"1 Ad. sequences of fiber bundles Spectral In Chapter Seven we computed homology groups of a product space B x F via the Kiinneth formula.Iii ~ = L ~. .q pEZ qE'l di = dl +1 p+q=i Ai : A. For q = 0.q.9.4.2. a differential when it satisfies Suppose a differential d : A has the bidegree (-n.. A i-I.-. sequence. we will end up with a bigraded spectral 9. Let B be a simply connected cell complex. Serre developed a method to calculate the homology groups of the total space E of a fiber bundle (E.1). which consists of all cells of dimension .-------t 1 A i.1 2: 2:A pEl qEZ ~P. 'Jr. ITwe start out with a bigraded exact couple..2 a.i ---1 Ai-I. Put .6 and 9.q ---lo A = L L Ap.1.- ---1 • . B. G).1.1.9. We say that a homomorphism dd = O. •• See Figure 9. In the latter case it is more complicated than passing to the tensor products and so on.2. F) from those of B and F. Differential d :A ~ A IS DEFINITION 9. n .

an abelian group..5. dn(E.:q '" Fp. PROPOSITION 9. Consider the chain complexes C* (E) .-n. .q+l This is the Serre homology spectral sequence of the fiber bundle (E.85 less than or equal to q. G).. we get We choose an abelian group G and set We can reduce the calculation of the n-th homology group of E to its calculation over the n-cells of the base space B..n c·· · c Fn-1. Jr. F) be a fiber bundle. 1. Then we can define in a natural wayan exact n) with the couple which has the convergent spectral sequence (En. C Bb = B. We obtain the initial exact couple for the theorem as follow s. b = dim B. C* (EP) and C* (EP-l ) of the cell . n = 0. we have o= THEOREM F-l..qjFp-1.. F) over G. d following properties. Let (E.2.q+n-l' (3) E.n+1 C FO.7. where B a simply connected cell complex. (1) En = differential L J!.1 c Fn. . B.q) c p=Oq=O of bidegree t-n. B.o = Hn(E. n E.7r. Thus we can prove the next proposition by an argument similar to the one we used for Theorem 5. For any n.q oo oo (first quadrani bigraded). and dn is a 1). Thus we have that o= Putting B-1 C BO C Bl C .

we see that the pointsE5.1 c Fn. G) of subgroups of Hn(E).. uy "Ilt:: UIllVt::H:iW Z) 0 Hq(F..A. Z). 1... v~ .q Do you understand the implication of the theorem? What the theorem says is the following. G) til Tor (Hp~l (B.q+l of the adjacent groups in the sequence = F~l. 9. E.1. In the above spectral sequence..0 0.7.J OIU1. Then the sequence converges to E:J:q. EP and EP-l respectively.L0 1. c Fn-1. and hence when E2 = E3 = . Hq(F. dn = for n > 2.n+l C Fo. From this we keep taking homologies (compute inductively E~.o = Hn(E. we have p+q=n-l which is nothing but the Kiinneth DEFINITION formula 9.q = Hp(B.q.vIU L.. G)) LeJ L" w'1t::UI't::llI. We say that a spectral sequence (En. I .1. dn) collapses when dn = 0.6 and beyond we will encounter some non-trivial collapsing spectral sequences which will work a miracle in the computation of cohomologies of classifying spaces.n c .8.1.1.. In §9. .0 v in the second quadrant and hence is zero)..1.. = Eoc. the equality EXAMPLE ° = B. Hence a natural map ET o. which is the quotient Fp. \01. n > 2..1.86 9. '. For a trivial bundle (E. we have E". 1.q/ Fp-1.UO v1.« . and so E2 = E3 = . p. G)) is isomorphic to ° Hp(B. SPECTRAL SEQUENCES complexes E. .q+T-l. Moreover. ).1..IO. (\ I V • JT U . F) = B x F.1. E. Hq(F. IT.q ----+ ET+l O.v1.1 C'. • \IV 1.q on the •• ~ -a.g..q from the homologies of B and F. First we compute the E..q L'T DT IAJ LJ -T. They induce the homology exact sequence that we can use to define the initial exact couple: El OBSERVATION = ~Hp+q(C*(EP)/C*(EP-l)).IU 1. For instance..

GJ. G).2). Hn(B. in Serre's spectral sequence for ( E . G) ::: E6. 7r* = 7T : Hn(E. that i* = i: Hn(F. It follows from Theorem 9.o onto the homology implies that E(fq '" 9. B. we have a natural map tv 7T : Hp(E: GJ ---t E.q+l = 0 Fo. Hence we classes exists.q o E~. The Serre spectral sequence is also useful in THEOREM 9. GJ .q ---t EG:q ---t Hq(E. the boundary of a point ET on the x-axis comes from the fourth quadrant and so it is zero (Figure 9. that is. Let (E.. By Theorem EJ. 'IT. and there is a natural one-to-one map We also have a projection Hp(E) = Fp. have a natural map rv i: Hq(F. Hence the cycles and homology classes agree.. GJ.q c Hq(E. for all n > 0. G).9. from Serre's spectral sequence. G). G).q/ F_1. G». F) we have.4) that E. GJ ---t ---t Hn(E. Hq(F.6 (2) and the universal coefficient theorem (Theorem 7. F) be a /'iher bundle whose base space some fixed fiber over a point of B. Similarly.q C::!.q+l = FO. B.6 (2) we have that coefficient theorem Fq(F.o ---+ E.1. but then the universal and the fact Ho(B. Furthermore. Therefore. Z) Z imply that E6.o '" Hp(B.:o Fp.q rv Ho(B. GJ.o rv Hp(B.:'oJ because Er.Eo. the fact F_1. Then we can calculate the induced maps 1T * and i.:o ---t E.o/ Fp-1. 'IT.

Z) (j> 0) using the Serre spectral sequence of the fiber bundle (S2k=1.3. E.o:::: E~. = 0.t1 = H(E. q<0 0 r q> I . Figure 9. Z). 81) over Z (cf..q Since = 0.6 (2) says that E. Hq(F. 2k 0 + 1. We wish to compute the homology groups Hj(pk(C. Theorem 9. so that E. we have.l :: Hp(B. E..q ~ Hp(B.3.. for all n > 2.3). pk(C).9.SPECTRAL SEQUENCES FIGURE 9. 9. Applications of spectral sequences j < 0 and i> 2k j ::::: . IT. q <0 En or q > 1. otherwise. Z)).(E) using non-collapsing spectral sequences? In the next section we apply this theorem to compute some homology groups.'q: dn).. Homology of pk(() Can we actually compute H.

2. we hence it follows that E. p.Hence we get .+2.) Fp.O 0 =~ ---7 E2 ) p-2._1 implies that im(d2 : Ep2+2-1 -..) = 0. ..q/Fp-1.9. and so E. .q p.q ~ Ep-n.l). Enp. p i.:q I"'-' (3).. 7l._n If n > 2. p f= 2k From these two equalities we conclude that if P i.o = H. (S2k+\ 7l.l ker(d 2 = im(d : 2 : E. Furthermore.2k. p + q i=. We can compute the cohomology groups of pk(C) using the universal coefficient theorem. + 1. p. the factE~+2. E2 p.2) = O. then E. then dn = 0 .1) (c E.. d2 : E.-2.+2. Therefore.O ~ Hp+2(B. and so from the sequence FO.O E..O rv p. for all p and q we have rv rv E3 rv E4 = .q: We have d2(E.q Recalling Theorem 9.0. d 0.j C we get I •• C Fj-l. E p... = Eoo p.) Z) r-: is an isomorphism.+2.q+n-l' we have Ep..2k. .q = 0 or q+n-l = 0. Fp.E20) = 0 C E2 o.. and B = to obtain cs: (Bi 7l.2. we have E = S2k+l. 1 N Hp(B.. We now recall that Hi.q+l. Z j = 0. 2k + 1.6 conclude that = 0.0 --7 --7 E.2k.o ~ E3 = ker (d2.. Ho(B.1' On the other hand.3.l C Fj. j odd. which says that E.. .q = p. APPLICATIONS OF SPECTRAL SEQUENCES 89 Look at thnehomomorphism dn : E.q ---f En p-n q+n-l' n > 2. Z) ----t E.

which will turn out to commute with the convergence of cohomology spectral sequences. Then we can define a natural exact couple which has a convergent spectral sequence satisfying the following properties. . Cohomology spectral sequences ". In working with cohomology we habitually use subindices for derived couples (En) Dn.ker (Hp+q(E. G) ---1 Hp+q(EP~l.q E~q . where En and D. SPECTRAL SEQUENCES 9. .q+l I'. of abelian groups. (r-v HP+pPrime(B.q.q . B. whose coefficient product We derive the product for En+l from for we derive the product the product for En. n. Hq+q'(F. (7) of B.q) c Eh+n. 9.q / FP+l. (1) (2) E~.q ~ HP (B. is a differential dn (Eh.q ® EK EP+P' n (5) dn(ab) = dn(a)b+ (6) The product (-l)p+qadn(b).. in~ jn.q+q.q-n+l of bidegree (n. In a like manner we derive cohomology spectral sequences as follows. Hq(F. Setting FP. :J Fn+1.. -n -f. ( 3) G)).-1 =0 En = l:~o 2:~o EK.q (first quadrant bigraded).10 (Serre's cohomology spectral sequence). . b E Eh·q .1). . Recall that cohomology groups have cup products. a E E~. where d. -4 E~+pl.90 9..4. are bigraded. G)) ..i FP. THEOREM G) :::::: FO. kn). we get a sequence Hn(E. (4) For every n ---1 > 1 there is a well-dejined product E~. G))) equals (-1 )qp times the cup product is the cup plVduct of F.n~l:J .q-l. F) a fiber bundle. B a simply connected cell complex and (E. In this sense cohomology spectral sequences are even more useful than homology sequences. Let G be an abelian group. .Ie obtained Serre' s homology spectral sequences from certain exact couples of fiber bundles. and ECXJ from the cup product for E.n :J Fl.

Cohomology groups of pk(C) In §9. suppose that al E E~.9. Hn(E. which is different from the homology bidegree (-n. B.n + 1).l and ao E E~'o correspond under isoIIlorphisIIIS E~. Z) (it may be a bit tedious. Z). E E~. p* : Hn(B.3 we calculated the homology groups ofP"(C) from the homology spectral sequence over Z of a certain fiber bundle. by TheoreIIl 9.1 ~ HO(B. For this ao E E~'o HP(B.10 is the Serre cohomology spectral sequence of a fiber bundle (E.2k the map is an isomorphism. is a generator of the group Z. 223) and E~'o _ HP( B.5. G). COHOMOLOGY GROUPS OF pk(C) 91 The spectral sequence in Theorem 9. 1 E E~. On the other hand. n . Z) ~ 21 as well as its multiple by -1. We can also compute the maps of cohomology groups. i* : Hn(E.l ~ HP(B.1). is (n.. .l N HP(B. G) -+ Hn(F.. G) -----+ G). and noted that we can obtain its cohomology groups from the universal coefficient theorem. 9.5. 7r. F) over G.10 (5) we have d2(aI) = d2(ao = u 1) = d2(aO) u u1 +(-l)Pao u d2(1) (-l)P ao d2 ( 1). Z). Note that the bidegree of d. Z) Z satisfies N t'. and for p i. Thus we have shown the following result. Hence. we want to take advantage of Serre's cohomology spectral sequence to compute the cohomology groups of the projective space pk(C) and investigate their cup-product structure. .I an u 1 = a. where d2(1) E H2(B. Our situation here is just as it is with homology. induced by the embedding i: F -1 E of the fiber F and the projection 7r : E -1 B of the total space E in the same way as we did in the homology case. In this section. but write down three copies of the spectral sequence and consider the cup product).

of odd-dimensional PROPOSITION G) = L H + (X. ~) = 0 and Z ® JR lR IR @lft lRimply that N E~.. hold: . HDdd(X. JR) = O. oo 2i 1 G) i=O cohomology groups of X over G. IR) = Hodd(F.q ~ Ef+r. Hq(F. JR). for every r PROOF. . Suppose that a fiber bundle (E. (pk (C).. f'. SPECTRAL SEQUENCES THEOREM 9. /l) . > 2. q .6. Z) ® Hq(F. But at least one ofp. sequence.2k. Moreover we get d. H2j 75) ~ {z. ase space at.. q.z are j = 0.q-r+l. The cohomology IP (pk(C).I The universal coefficient theorem (Theorem 7. if either p or q is odd then E~. . groups of Pk(C) over .q = 0 for r 2:: 2.q = O. D We have one more proposition restricted to cohomology groups. 9. F) over HOdd(B.q '" HP (B. U u generates cohomology sequences 9. JR) ®lR Hq(F. Z).12.r + 1 is always odd. By G) we mean the direct sum u=tx. B. Collapsing We see in the next proposition an example of non-trivial fiber bundles with collapsing spectral sequences. If u generates H2(pk(C). = 0 : E~. p + r.6) and the facts that Tor(Z. j odd. Therefore. then uj = u U u U . We only investigate cohomology which will be useful for us later. and so dn = 0 holds whenever r > 2.2. R) HP(B. 0. rv JR)) ::: HP(B.92 9.11. Thus in this case Ef.

H*(E. E~. R). hence (as we are working over R). G) ::::FO.HP(B.7. i.n-ll'. G).'o -t E~O '" Fn.Q E~O is an isomorphism implies that it is By the universal coefficient theorem we see that E~..n Furthermore. the map i* is equal to the E~n -* composition Hn(E. G) is a surjection. ----t E.7.G) (2) p* : (3) ----t ----t H" (F. . G) f"V p* is equal to the composition 0)) tv H" (B. U(n + N)/U(n) x UrN). is an isomorphism implies that it is E~. G). IR) 0~ Hq(F.n _____.o C Hn(E. G) is an injection. HO(F..n surjective.JR) ~ H*(B.R) 0IR Hq(F.. / F1. Hq(F. but the fact that E. p+q=n o 9.e.JR.R) :: L E~q rv 2:= HP(B. c Urn).n (Y Hn(F.R). it follows that Hn(E. G) Hn(E. and the fact that E~n _____. -Iist as in the homology case.l FO. Cohomology of classifying spaces Let us calculate the cohomology groups of the classifying space BU(n) - lim N--+OCl U(n + lV)/U(n) x U(1V) - U(n + oo)jU(n) x U(oo) of n-dimensional complex vector bundles over the real coefficients R. COHOMOLOGY OF CLASSIFYING SPACES 93 (1) i*: Hn(E. Hn(B. Hn(B. PROOF. IR)) ". There is a natural identification of T" which is the product of n copies ofU(1) '" Sl as a subgroup of Urn).q ~ HP (B.9.IR).) ®[R H*(F. We have the fiber bundle U(n)/Tn -* U(n + N)/Tn x UrN) --.

You're Reading a Free Preview

/*********** DO NOT ALTER ANYTHING BELOW THIS LINE ! ************/ var s_code=s.t();if(s_code)document.write(s_code)//-->