Math 530 Lecture Notes. Xi Chen

Similar documents
Hungry, Hungry Homology

LECTURE 3: RELATIVE SINGULAR HOMOLOGY

ALGEBRAICALLY TRIVIAL, BUT TOPOLOGICALLY NON-TRIVIAL MAP. Contents 1. Introduction 1

The Hurewicz Theorem

121B: ALGEBRAIC TOPOLOGY. Contents. 6. Poincaré Duality

Eilenberg-Steenrod properties. (Hatcher, 2.1, 2.3, 3.1; Conlon, 2.6, 8.1, )

58 CHAPTER 2. COMPUTATIONAL METHODS

FILTERED RINGS AND MODULES. GRADINGS AND COMPLETIONS.

6 Axiomatic Homology Theory

Solution: We can cut the 2-simplex in two, perform the identification and then stitch it back up. The best way to see this is with the picture:

Manifolds and Poincaré duality

p,q H (X), H (Y ) ), where the index p has the same meaning as the

Review of Linear Algebra

Algebraic Geometry Spring 2009

HOMOLOGY THEORIES INGRID STARKEY

SELF-EQUIVALENCES OF DIHEDRAL SPHERES

An Outline of Homology Theory

7.3 Singular Homology Groups

Math 341: Convex Geometry. Xi Chen

HOMOLOGY AND COHOMOLOGY. 1. Introduction

Smith theory. Andrew Putman. Abstract

STABLE MODULE THEORY WITH KERNELS

Topological K-theory

Algebraic Geometry Spring 2009

Math 757 Homology theory

CW-complexes. Stephen A. Mitchell. November 1997

PERVERSE SHEAVES ON A TRIANGULATED SPACE

MATH 101B: ALGEBRA II PART A: HOMOLOGICAL ALGEBRA

Algebraic Topology Homework 4 Solutions

A duality on simplicial complexes

Corrections to Introduction to Topological Manifolds (First edition) by John M. Lee December 7, 2015

CAUCHY INTEGRAL THEOREM

Direct Limits. Mathematics 683, Fall 2013

Lecture 1. Toric Varieties: Basics

LECTURE 11: SYMPLECTIC TORIC MANIFOLDS. Contents 1. Symplectic toric manifolds 1 2. Delzant s theorem 4 3. Symplectic cut 8

Part II. Algebraic Topology. Year

Tree-adjoined spaces and the Hawaiian earring

A Note on the Inverse Limits of Linear Algebraic Groups

THE GROUP OF UNITS OF SOME FINITE LOCAL RINGS I

TCC Homological Algebra: Assignment #3 (Solutions)

C n.,..., z i 1., z i+1., w i+1,..., wn. =,..., w i 1. : : w i+1. :... : w j 1 1.,..., w j 1. z 0 0} = {[1 : w] w C} S 1 { },

Course 311: Michaelmas Term 2005 Part III: Topics in Commutative Algebra

CELLULAR HOMOLOGY AND THE CELLULAR BOUNDARY FORMULA. Contents 1. Introduction 1

Theorem 5.3. Let E/F, E = F (u), be a simple field extension. Then u is algebraic if and only if E/F is finite. In this case, [E : F ] = deg f u.

Lecture 6: Etale Fundamental Group

DISCRETIZED CONFIGURATIONS AND PARTIAL PARTITIONS

Topological Data Analysis - Spring 2018

A TALE OF TWO FUNCTORS. Marc Culler. 1. Hom and Tensor

Algebraic Topology exam

A Primer on Homological Algebra

Homotopy and homology groups of the n-dimensional Hawaiian earring

EILENBERG-ZILBER VIA ACYCLIC MODELS, AND PRODUCTS IN HOMOLOGY AND COHOMOLOGY

PERVERSE SHEAVES. Contents

HOMOTOPY THEORY ADAM KAYE

Formal power series rings, inverse limits, and I-adic completions of rings

Honors Algebra 4, MATH 371 Winter 2010 Assignment 4 Due Wednesday, February 17 at 08:35

ALGEBRAIC GEOMETRY COURSE NOTES, LECTURE 2: HILBERT S NULLSTELLENSATZ.

10. Smooth Varieties. 82 Andreas Gathmann

THE POINCARE-HOPF THEOREM

Patrick Iglesias-Zemmour

arxiv: v1 [math.co] 25 Jun 2014

Homological Decision Problems for Finitely Generated Groups with Solvable Word Problem

MATH 215B. SOLUTIONS TO HOMEWORK (6 marks) Construct a path connected space X such that π 1 (X, x 0 ) = D 4, the dihedral group with 8 elements.

FREUDENTHAL SUSPENSION THEOREM

EXT, TOR AND THE UCT

Equivariant cohomology of infinite-dimensional Grassmannian and shifted Schur functions

132 C. L. Schochet with exact rows in some abelian category. 2 asserts that there is a morphism :Ker( )! Cok( ) The Snake Lemma (cf. [Mac, page5]) whi

Tensor, Tor, UCF, and Kunneth

MATH 205B NOTES 2010 COMMUTATIVE ALGEBRA 53

Math 6510 Homework 10

BLOCKS IN THE CATEGORY OF FINITE-DIMENSIONAL REPRESENTATIONS OF gl(m n)

Exercises for Algebraic Topology

PERVERSE SHEAVES: PART I

Scissors Congruence in Mixed Dimensions

Homework 3: Relative homology and excision

INVARIANT IDEALS OF ABELIAN GROUP ALGEBRAS UNDER THE MULTIPLICATIVE ACTION OF A FIELD, II

Pacific Journal of Mathematics

De Rham Cohomology. Smooth singular cochains. (Hatcher, 2.1)

for some n i (possibly infinite).

ON RIGIDITY OF ALGEBRAIC DYNAMICAL SYSTEMS

Matsumura: Commutative Algebra Part 2

Topology Hmwk 6 All problems are from Allen Hatcher Algebraic Topology (online) ch 2

Lie Algebra Cohomology

0.1 Universal Coefficient Theorem for Homology

Cohomology and Base Change

LECTURE 7: STABLE RATIONALITY AND DECOMPOSITION OF THE DIAGONAL

Math 210B. Artin Rees and completions

The rational cohomology of real quasi-toric manifolds

Course notes in algebraic topology

Free Subgroups of the Fundamental Group of the Hawaiian Earring

Determinant lines and determinant line bundles

Groups of Prime Power Order with Derived Subgroup of Prime Order

Topological properties

SECTION 5: EILENBERG ZILBER EQUIVALENCES AND THE KÜNNETH THEOREMS

Representations of algebraic groups and their Lie algebras Jens Carsten Jantzen Lecture III

MATH 221 NOTES BRENT HO. Date: January 3, 2009.

Algebraic Topology I Homework Spring 2014

LECTURE: KOBORDISMENTHEORIE, WINTER TERM 2011/12; SUMMARY AND LITERATURE

and this makes M into an R-module by (1.2). 2

32 Proof of the orientation theorem

Transcription:

Math 530 Lecture Notes Xi Chen 632 Central Academic Building, University of Alberta, Edmonton, Alberta T6G 2G1, CANADA E-mail address: xichen@math.ualberta.ca

1991 Mathematics Subject Classification. Primary 14J28; Secondary 14E05 Key words and phrases. Algebraic Topology, Homology, Cohomology

CHAPTER 1 Homotopy, Fundamental Groups and Covering spaces 1.1. Homotopy Definition 1.1.1. Let f : X Y and g : X Y be two continuous maps between topological spaces X and Y. We say that f and g are homotopic if there exists a continuous map F : X [0, 1] Y such that F (x, 0) = f(x) and F (x, 1) = g(x) for all x X. If we denote the set of continuous maps between X and Y by C(X, Y ), homotopy is clearly an equivalence relation on C(X, Y ). That is, it is trivial that f hom f and f hom h if f hom g and g hom h. The following is also obvious: Proposition 1.1.2. If f 1 hom f 2 and g 1 hom g 2, then g 1 f 1 hom g 2 f 2, where f 1, f 2 C(X, Y ) and g 1, g 2 C(Y, Z). Let X be a topological space. Example 1.1.3. We claim that all continuous maps f : X R n are homotopic to each other: for f, g C(X, R n ), let F : X [0, 1] R n be given by F (x, t) = f(x)(1 t) + g(x)t. Example 1.1.4. The argument for the above example applies to all convex sets in R n : for V a convex set in R n and f, g C(X, V ), let F : X [0, 1] V be given by F (x, t) = f(x)(1 t) + g(x)t. Example 1.1.5. Furthermore, the above argument works for all starshaped sets in R n. We call a set V R n star-shaped if there exists a point p V such that the line segment pq V for every q V. For f C(X, V ), setting F (x, t) = f(x)(1 t)+tp shows that every f C(X, V ) is homotopic to the constant map g : X V sending g(x) = p for all x X. Hence the continuous maps in C(X, V ) are homotopic to each other. Depending on situation, we sometimes impose some further constraints on f and g for them to be homotopic. For example, for two continuous curves f : [0, 1] X and g : [0, 1] X with the same ending points f(0) = g(0) and f(1) = g(1), we say that f and g are homotopic with fixed ending points if there exists a continuous map F : [0, 1] [0, 1] X such that F (t, 0) = f(t), F (t, 1) = g(t), F (0, s) = f(0) = g(0) and F (1, s) = f(1) = g(1). 3

4 1. HOMOTOPY, FUNDAMENTAL GROUPS AND COVERING SPACES In case that X is a C k manifold, we say that two C k -curves f : [0, 1] X and g : [0, 1] X are C k -homotopic if there exists a C k map F : [0, 1] [0, 1] X such that F (t, 0) = f(t) and F (t, 1) = g(t). We say that f C(X, Y ) and g C(Y, X) are homotopic inverses to each other if f g hom 1 Y and g f hom 1 X, where 1 X and 1 Y are the identity maps on X and Y, respectively. Definition 1.1.6. We say two topological spaces X and Y are homotopic if there exist f C(X, Y ) and g C(Y, X) that are homotopic inverses to each other. If X is homotopic to a point, we call X contractible. It follows directly from Example 1.1.5 that Proposition 1.1.7. Every star-shaped set X in R n is contractible. Proof. Let Y = {p} X and let f : X Y and g : Y X be the maps that f(x) = p and g(p) = p. Clearly, f g = 1 Y and g f hom 1 X since all continuous maps in C(X, X) are homotopic to each other. Example 1.1.8. We claim that (R n ) is homotopic to S n 1. Let g : (R n ) (R n ) be the map ( x 1 x g(x 1, x 2,..., x n ) = x 2 1 + x 2 2 +... +, 2 x2 n x 2 1 + x 2 2 +... +, x2 n ) (1.1) x n... x 2 1 + x 2 2 +... + x2 n ( ) x1 = x, x 2 x,..., x n x So g contracts (R n ) to S n 1. It suffices to find a continuous function F : (R n ) [0, 1] (R n ) such that F (x, 0) is the identity map of (R n ) and F (x, 1) = g(x). We simply take (1.2) F (x, t) = (1 t)x + tx x. Example 1.1.9. Similarly, the annulus {r < x < R} R n is also homotopic to S n 1 by almost the same argument. Example 1.1.10. If we remove a linear subspace V from R n, R n \V is homotopic to S n m 1 R m, where m = dim V. Proposition 1.1.11. S 1 is not contractible. Lemma 1.1.12. Let f : R S 1 be the map sending R to its quotient S 1 = R/Z.

CHAPTER 2 Singular Homology 2.1. Definition of Singular Homological Groups An n-simplex is the convex hull of n+1 points p 0, p 1,..., p n in R n. We call it a nondegenerate n-simplex if p 0, p 1,..., p n do not lie on a hyperplane in R n. The standard n-simplex n is the convex hull of the points e 1, e 2,..., e n, 0, where 0 is the origin. That is, (2.1) n = {(t 0, t 1,..., t n 1 ) : 0 t 0, t 1,..., t n 1 1} R n. To definte homological groups, we embed n to R n+1 by sending (2.2) (t 0, t 1,..., t n 1 ) (t 0, t 1,..., t n 1, 1 t 0 t 1... t n 1 ). That is, we let (2.3) n = {(t 0, t 1,..., t n ) : t 0 + t 1 +... + t n = 1, t 0, t 1,..., t n 0} R n+1. Depending on the context, n refers to either the simplex (2.1) or its embedding (2.2) in R n+1. The k-th face of n given in (2.3) is the embedding (2.4) δ k : n 1 n defined by (2.5) δ k (t 0, t 1,..., t n 1 ) = (t 0, t 1,..., t k 1, 0, t k,..., t n 1 ). Let X be a topological space. We define the group of singular n-chains C n (X) of X to be the abelian group freely generated by all continuous maps f : n X. That is, (2.6) C n (X) = {a 1 f 1 + a 2 f 2 +... + a m f m : a 1, a 2,..., a m Z, f 1, f 2,..., f m continuous functions n X} The singular chain complex C(X) of X is the direct sum of C n (X), i.e., (2.7) C(X) = C n (X) n= where we let C n (X) = 0 for n < 0. We have a group homomorphism : C n (X) C n 1 (X), called boundary operator, defined by sending a continuous map f : n X to n (2.8) f = ( 1) k f δ k k=0 5

6 2. SINGULAR HOMOLOGY and then extending by linearity via (2.9) (a 1 f 1 + a 2 f 2 +... + a m f m ) = a 1 f 1 + a 2 f 2 +... + a m f m. We can think of as an operator : C(X) C(X) from C(X) to itself. Theorem 2.1.1. Let be the boundary operator on C(X) defined as above. Then 2 = 0. Proof. We defined the operator δ k : C n (X) C n 1 (X) by δ k f = f δ k and let δ k f = 0 for k > n. Then (2.10) f = ( 1) k δ k f and (2.11) 2 f = for f C n (X). Note that (2.12) Thus, we have k=0 ( 1) k+l δ k δ l f k,l=0 (δ k δ l f)(t 0, t 1,..., t n 2 ) { f(t 0, t 1,..., t l 1, 0, t l,..., t k 1, 0, t k,..., t n 2 ) = f(t 0, t 1,..., t k 1, 0, t k,..., t l 2, 0, t l 1,..., t n 2 ) (2.13) δ k δ l f δ l 1 δ k f = 0 for all k < l and hence 2 = 0. if k l if k < l Clearly, 2 = 0 implies that the image of is contained in the kernel of. More specifically, (2.14) C n+1 (X) ker(c n (X) C n 1 (X)) for all n. The n-th singular homological group of X is defined to be (2.15) H n (X) = ker(c n(x) C n 1 (X)). C n+1 (X) We can also replace the coefficients a i Z in (2.6) by a i R for a fixed commutative ring R. Namely, we let C n (X) be the module over R freely generated by all continuous functions f : n X. As another way to put it, we can replace C n (X) and C(X) by C n (X) Z R and C(X) Z R, respectively. We call the corresponding homological groups (R-modules) (2.16) H n (X, R) = ker(c n(x) Z R C n 1 (X) Z R) (C n+1 (X) Z R) the singular homology of X over R. The homology given in (2.15) is the singular homology of X over Z and we sometimes write H n (X) = H n (X, Z) to emphansize the fact that these groups are defined over Z. It is very easy to figure out H 0 (X). That is,

2.2. SOME BASIC NOTIONS IN HOMOLOGICAL ALGEBRA 7 Proposition 2.1.2. Let X be a topological space and let X = α A X α with X α the connected components of X. Then (2.17) H n (X) = α A H n (X α ) for all n and (2.18) H 0 (X) = α A Z if all X α are path-connected. Proof. Clearly, C n (X) = α A C n (X α ) and hence (2.17) follows. To show (2.18), it suffices to show that H 0 (X) = Z for X path-connected. For every continuous map f : 1 X, f = f(0, 1) f(1, 0). Since X is path-connected, there is a continuous map f : 1 X such that f(0, 1) = p and f(1, 0) = q for every pair of points p, q X. Therefore, C 1 (X) is generated by p q for all p, q X. It follows that H 0 (X) = Z. Proposition 2.1.3. Let X be the topological space consisting of one point. Then { Z if n = 0 (2.19) H n (X) = 0 if n 0 Proof. Obviously, C n (X) = Z is generated by the constant function f n : n X for n 0. Then (2.19) follows easily from the fact that { 0 if n is odd or n = 0 (2.20) f n = f n 1 if n is even and n > 0 2.2. Some Basic Notions in Homological Algebra For simplicity, we work with abelian groups. One can replace them by R-modules. We call a sequence of homomorphisms of abelian groups (2.21) A f B g C exact at B if im(f) = ker(g), a left exact sequence if it is exact at B and f is injective, a right exact sequence if it is exact at B and g is surjective and a short exact or simply an exact sequence if it is both left and right exact. Usually, we write (2.22) 0 A f B g C for a left exact sequence, (2.23) A f B g C 0

8 2. SINGULAR HOMOLOGY for a right exact sequence and (2.24) 0 A f B g C 0 for a short exact sequence. A long exact sequence is simply a sequence (2.25) A 0 A 1 A 2... A n A n+1 of homomorphisms of abelian groups that is exact at all A 1, A 2,..., A n. A chain complex (2.26) K = n Z K n is an abelian group together with a boundary operator sending K n K n 1 and satisfying 2 = 0. Usually, we write it as (K, ). The homological groups of a chain complex (K, ) are defined to be (2.27) H n (K ) = ker(k n K n 1 ) K n+1. Clearly, H n (K ) = 0 if and only if the sequence (2.28)... K n+1 Kn Kn 1... is exact at K n. In other words, H n (K ) measures how (2.28) fails to be exact. Let (K, ) and (L, ) be two chain complexes. We call f : K L a map of chain complexes if it is a group homomorphism sending K n to L n and commuting with the boundary operators, i.e., f = f. More precisely, we have the commutative diagram (2.29)...... K n+1 K n K n 1... f f L n+1 L n L n 1... f The kernel ker(f) and cokernel coker(f) are also chain complexes in a natural way. That is, we can define boundary operators (2.30) : ker(k n f Ln ) ker(k n 1 f Ln 1 ) and (2.31) : coker(k n f Ln ) coker(k n 1 f Ln 1 )

2.2. SOME BASIC NOTIONS IN HOMOLOGICAL ALGEBRA 9 such that the commutative diagram (2.29) can be extended to (2.32) 0 0 0......... J n+1 J n J n 1... K n+1 K n K n 1... f f L n+1 L n L n 1... f... M n+1 M n M n 1... 0 0 0 where the vertical columns are exact. A short exact sequence of chain complexes is a sequence (2.33) 0 K f L g M 0 of maps of chain complexes that is exact as a sequence of abelian groups. A map f : K L of chain complexes induces a natural homomorphism (2.34) f : H n (K ) H n (L ) of homological groups of K and L sending ker(k n Kn 1 ) to ker(l n L n 1 ). A short exact sequence (2.33) of chain complexes leads to a long exact sequence on homology... H n+1 (K ) f H n+1 (L ) g H n+1 (M ) (2.35) H n (K ) f H n (L ) g H n (M ) H n 1 (K ) f H n 1 (L ) g H n 1 (M )... We say two maps f : K L and g : K L of chain complexes are homotopic if there are homomorphisms P n : K n L n+1 such that (2.36) P n + P n 1 = f g on K n for all n. Or equivalently, if we replace P n by ( 1) n P n, (2.36) becomes (2.37) P n P n 1 = ( 1) n (f g). It is obvious that f and g induce the same map on homology if they are homotopic.

10 2. SINGULAR HOMOLOGY 2.3. Homotopy Invariance of Homology 2.3.1. Push-forward. We are going to establish the homotopic invariance of singular homology. First of all, a continuous map f : X Y induces a map of chain complexes (C (X), ) and (C (Y ), ). That is, the diagram (2.38)... C n+1 (X) C n (X) C n 1 (X)... f f f... C n+1 (Y ) C n (Y ) C n 1 (Y )... commutes for f : C n (X) C n (Y ) defined by f (g) = f g for a continuous map g : n X. So it induces a map on homological groups f : H n (X) H n (Y ), which is called the push-forward of f on homology. It is obvious that Proposition 2.3.1. Let f : X Y and g : Y Z be continuous maps between topological spaces X, Y and Z. Then (g f) = g f for f : H n (X) H n (Y ) and g : H n (Y ) H n (Z) the push-forwards of f and g, respectively. 2.3.2. Homotopy invariance of singular homology. We first show that two homotopic maps f : X Y and g : X Y induce the same map on homology. Theorem 2.3.2. Let f : X Y and g : X Y be two continuous maps between topological spaces X and Y and let f : H n (X) H n (Y ) and g : H n (X) H n (Y ) be the induced maps on homology, respectively. If f and g are homotopic, then f g. Combining the above theorem with Proposition 2.3.1, we arrive at our main result of this section Corollary 2.3.3. Let X and Y be two homotopic topological spaces. Then H n (X) = H n (Y ) for all n. We will spend the rest of the section to prove Theorem 2.3.2. Here is a sketch of our proof. It suffices to show that f : C(X) C(Y ) and g : C(X) C(Y ) are two homotopic maps of chain complexes. Namely, we are trying to find P n : C n (X) C n+1 (Y ) such that (2.39) (P n (σ)) P n 1 ( (σ)) = ( 1) n (f σ g σ) for all σ C n (X) and n. A natural way to construct P n is via the continuous map F : X I Y satisfying F (x, 0) = f and F (x, 1) = g, where I = 1 = [0, 1]. For every continuous map σ : n X, we have a continuous map F (σ 1 I ) : n I Y, where 1 I is the identity map on I and σ 1 = σ 1 I : n I X I is the map sending (p, t) to (σ(p), t).

Roughly, we let 2.3. HOMOTOPY INVARIANCE OF HOMOLOGY 11 (2.40) P n (σ) = F (σ 1) and extend it by linearity to all σ C n (X). The problem is that F (σ 1) is a continuous function on n I, n I is not a simplex and hence F (σ 1) is not an element of C n+1 (Y ). The remedy of this situation involves a technique called subdivision of convex polytopes. Roughly, we will find an (n + 1)-chain ξ n C n+1 ( n I) and define P n to be (2.41) P n (σ) = F (σ 1) ξ n such that (2.39) holds, where F : C(X I) C(Y ) is the push-forward of F on the chain complexes. 2.3.3. Oriented simplices. Let X be a convex set in R N. We let L n (X) denote the abelian group freely generated by the maps (2.42) σ p0 p 1...p n (t 0, t 1,..., t n ) = t 0 p 0 + t 1 p 1 +... + t n p n from n to X for some p 0, p 1,..., p n X. We call σ p0 p 1...p n the oriented n-simplex spanned by p 0, p 1,..., p n. Note that the order of p 0, p 1,..., p n is essential. Obviously, L n (X) is a subgroup of C n (X) and L n (X) L n 1 (X) since n (2.43) σ p0 p 1...p n = ( 1) k σ p0 p 1...ˆp k...p n. k=0 Let p 0, p 1,..., p n be n + 1 points in R n that do not lie on a hyperplane. Or equivalently, the determinant p 0 1 p 1 1 (2.44) det.. p n 1 (n+1) (n+1) does not vanish, where p k is the row vector representing the point p k. We let ξ p0 p 1...p n be the n-chain in L n (R n ) given by (2.45) ξ p0 p 1...p n = λ p0 p 1...p n σ p0 p 1...p n where λ p0 p 1...p n = 1 if the determinant (2.44) is negative, λ p0 p 1...p n = 1 if the determinant (2.44) is positive and 0 if it vanishes. For simplicity, we also write ξ σ = ξ p0 p 1...p n and λ σ = λ p0 p 1...p n for σ = σ p0 p 1...p n. 2.3.4. Convex polytope and subdivision. Definition 2.3.4. A convex polytope T R n is the convex hull of a finite set S of points in R n. The dimension dim T of T is the dimension of the linear subspace spanned by S. We call T R n non-degenerate if dim T = n, or equivalently, the interior of T in R n is not empty. We can find {p 1, p 2,..., p m } S such that T is the convex hull of p 1, p 2,..., p m S

12 2. SINGULAR HOMOLOGY and no one point among p 1, p 2,..., p m is a convex combination of the rest. Then the center (barycenter) of T is given by (2.46) c(t ) = 1 m (p 1 + p 2 +... + p m ). An orientation of T is an ordering of {p 1, p 2,..., p m }. Let m = dim T. An (m 1)-dimensional face of T is the intersection T = T H of T and a hyperplane H R n such that dim T = m 1 and H is a supporting hyperplane of T, i.e., T lies on one side of H. Inductively, every l-dimensional face of T is also an l-dimensional face of T. That is, the faces of T consist of all (m 1)-dimensional faces T and all the faces of such T. For convenience, we call T the m-dimensional face of itself. The faces of an n-simplex conv(p 0, p 1,..., p n ) are conv(s) for all nonempty set S {p 0, p 1,..., p n }. An oriented n-simplex σ = σ p0 p 1...p n gives orientations to all its faces. We call σ pi1 p i2...p im an oriented face of σ for 0 i 1 < i 2 <... < i m n. Definition 2.3.5 (Subdivison of a convex polytope). Let T be a convex polytope in R n of dimension dim T = m. A subdivision of T is a way to divide T into a union of m-simplices (2.47) T = a T a such that T a T b is either empty or a face of both T a and T b for all a and b. A subdivision of T is denoted by a collection {T a } of m-simplices with the above properties. Furthermore, an oriented subdivision of T is a collection {σ p0 p 1...p m } of oriented m-simplices with the above properties. For T R n a nondegenerate convex polytope and an oriented subdivision T of T, we associate an n-chain ξ T L n (T ) by (2.48) ξ T = σ T with ξ σ defined in (2.45). Let T = n I R n+1 and let i 0 and i 1 be the maps n n I sending p to (p, 0) and (p, 1), respectively; i 0 and i 1 can be regarded as oriented faces of T. The vertices of i 0 and i 1 are the vertices of T and we orient T in the way {a 0, a 1,..., a n, b 0, b 1,..., b n } such that (2.49) i 0 = σ a0 a 1...a n and i 1 = σ b0 b 1...b n. The other n-dimensional faces of T are δ k ( n 1 ) I for k = 0, 1,..., n, where δ k : n 1 n is the k-th face of n defined by (2.5). We construct the following pseudo barycentric subdivision, denoted by pbcs(t ), inductively as follows: If n = 0, we let pbcs(t ) = {σ}, where σ = 1 I is the identity map on I = 1. ξ σ

2.3. HOMOTOPY INVARIANCE OF HOMOLOGY 13 Let T be an n-dimensional face of T. If T is i 0 or i 1, we let pbcs(t ) = {σ} with σ the identity map on n. If T = δ k ( n 1 ) I, we let pbcs(t ) be the subdivision obtained by induction. For every oriented n-simplex σ p0 p 1...p n pbcs(t ), we replace it by σ p0 p 1...p np with p = c(t ) the center of T. That is, we let (2.50) pbcs(t ) = {σ p0 p 1...p np} T σ p0 p 1...pn pbcs(t ) for T running over all n-dimensional faces of T. Every simplex σ pbcs(t ) is given in the form of σ = σ p0 p 1...p k p k+1...p n+1, where σ p0 p 1...p k is a face of i 0 or i 1 and p j is the center of a face T j = j 1 I of T for j = k + 1,..., n + 1 satisfying σ p0 p 1...p k T k+1...t n+1 = T. As in (2.48), we define ξ n L n+1 (T ) similarly by (2.51) ξ n = σ pbcs(t ) and we claim that ξ n is the (n + 1)-chain such that (2.39) holds if we define P n as in (2.41). Namely, we can prove Lemma 2.3.6. For every continuous map F : X I Y and every σ C n (X), (2.52) (F (σ 1) ξ n ) F ( σ 1) ξ n 1 = ( 1) n (f σ g σ) where F (x, 0) = f(x) and F (x, 1) = g(x). Proof. It suffices to prove it for X = n, Y = n I, F = 1 T : T = n I T the identity map on n I and σ = 1 n : n X the identity map on n. Indeed, since F commutes with, the LHS of (2.52) becomes (2.53) ξ σ (F (σ 1) ξ n ) F ( σ 1) ξ n 1 = F ( ((σ 1) ξ n ) ( σ 1) ξ n 1 ) If (2.52) holds for (X, Y, F, σ) = ( n, n I, 1 T, 1 n ), then (2.54) ((σ 1) ξ n ) ( σ 1) ξ n 1 = ( 1) n (i 0 i 1 ) for σ = 1 n. For all σ C n (X), σ = σ 1 n, σ = σ 1 n (2.55) and hence ((σ 1) ξ n ) ( σ 1) ξ n 1 = (σ 1) ( ((1 n 1) ξ n ) ( 1 n 1) ξ n 1 ) = ( 1) n (σ 1) (i 0 i 1 ). Combining this with (2.53), we see that (2.52) follows. So it suffices to prove it for (X, Y, F, σ) = ( n, n I, 1 T, 1 n ). Indeed, it remains to verify (2.54) for σ = 1 n. That is, (2.56) ξ n ( 1 n 1) ξ n 1 = ( 1) n (i 0 i 1 )

14 2. SINGULAR HOMOLOGY or equivalently, (2.57) ξ n n ( 1) k (δ k 1) ξ n 1 = ( 1) n (i 0 i 1 ) k=0 where δ k 1 : n 1 I n I is the map sending (p, t) to (δ k (p), t). If an (n+1)-simplex σ pbcs(t ) has an n-dimensional face ε containing the center c(t ) of T, then there must exist another (n + 1)-simplex σ pbcs(t ) such that σ σ = ε. More precisely, there are oriented simplices in the form of σ = σ p0 p 1...p k...p n+1 and σ = σ p0 p 1...p k...p in pbcs(t ) such that n+1 ε = σ p0 p 1...ˆp k...p n+1. Note that p n+1 = c(t ). It is easy to check that (2.58) λ p0 p 1...p k...p n+1 = λ p0 p 1...p k...p n+1 and hence (2.59) ξ p0 p 1...p k...p n+1 δ k = ξ p0 p 1...p k...p n+1 δ k. Therefore, (2.60) ξ n = = σ p0 p 1...p n+1 pbcs(t ) σ p0 p 1...p n+1 pbcs(t ) ξ p0 p 1...p n+1 ( 1) n+1 λ p0 p 1...p n+1 σ p0 p 1...p n. Obviously, σ p0 p 1...p n is either one of i 0 and i 1 or a simplex in pbcs(t ) for a face T = δ k ( n 1 ) I of T. And it is easy to check that λ p0 p 1...p n+1 = 1 if σ p0 p 1...p n = i 0 and λ p0 p 1...p n+1 = 1 if σ p0 p 1...p n = i 1. Therefore, (2.61) ξ n ( 1 n 1) ξ n 1 = ( 1) n (i 0 i 1 ) n + ( 1) n+1 λ q0 q 1...q nqσ q0 q 1...q n k=0 n ( 1) k σ p0 p 1...pn pbcs( n 1 I) k=0 σ p0 p 1...pn pbcs( n 1 I) λ p0 p 1...p n σ q0 q 1...q n where q 0 = δ k (p 0 ), q 1 = δ k (p 1 ),..., q n = δ k (p n ) and q = c(t ). Here we use δ k for the map δ k 1 : n 1 I n I. It remains to verify that (2.62) ( 1) n+1 λ q0 q 1...q nq = ( 1) k λ p0 p 1...p n for all p 0, p 1,..., p n n 1 I, q j = δ k (p j ) for j = 0, 1,..., n and q = c(t ), which is quite straightforward.

2.4. BARYCENTRIC SUBDIVISION 15 2.4. Barycentric Subdivision Definition 2.4.1 (Barycentric subdivision). Let T be a convex polytope of dimension dim T = m in R n. The barycentric subdivision of T, denoted by bcs(t ), is the oriented subdivision of T constructed inductively as follows: If T = {p} is a point, bcs(t ) = {σ p }. For every (m 1)-dimensional face T of T and every oriented (m 1)-simplex σ p0 p 1...p m 1 bcs(t ), we replace it by σ p0 p 1...p m 1 p, where p = c(t ) is the center of T. That is, we let (2.63) bcs(t ) = {σ p0 p 1...p m 1 p} T σ p0 p 1...p m 1 bcs(t ) for T running over all (m 1)-dimensional faces of T. By the construction of bcs(t ), it is easy to see that every oriented m- simplex σ bcs(t ) is spanned by a sequence p 0, p 1,..., p m with the property that p j is the center of a j-dimensional face T j of T for j = 0, 1,..., m with the property that T 0 T 1... T m. For a nondegenerate convex polytope T R n, we associate an n-chain α T L n (T ) to T by (2.64) α T = ξ σ. σ bcs(t ) as in (2.48). For every continuous map g : n X, we let (2.65) β g = g α n = λ σ (g σ) σ bcs( n) and we can define β g for all g C n (X) by extension of linearity. This defines an operator β : C(X) C(X) by sending g to β g. It is not hard to see, though a little tedious to check, that β commutes with the boundary operator. That is, Lemma 2.4.2. Let X be a topological space and β : C(X) C(X) be the operator defined as above via barycentric subdivision. Then β = β. Proof. By the same argument as in the proof of Lemma 2.3.6, we see that it suffices show that β(1 n ) = β( 1 n ) for X = n. That is, (2.66) α n = n ( 1) k δ k α n 1. k=0 Again, by the same argument for Lemma 2.3.6, we see that all (n 1)- dimensional faces of σ bcs( n ) inside σ cancel out each other. Thus,

16 2. SINGULAR HOMOLOGY we have (2.67) α n = ξ q0 q 1...q n 1 q = σ q0 q 1...q n 1 q bcs( n) σ q0 q 1...q n 1 q bcs( n) ( 1) n λ q0 q 1...q n 1 qσ q0 q 1...q n 1 where q = c( n ). Obviously, σ q0 q 1...q n 1 is a simplex in bcs(δ k ( n 1 )). To show (2.66), it is enough to verify that (see also (2.62)) (2.68) ( 1) n λ q0 q 1...q n 1 q = ( 1) k λ p0 p 1...p n 1 for all p 0, p 1,..., p n 1 n 1, q j = δ k (p j ) and q = c( n ), which is quite straightforward. So β defines a map of complexes C(X) C(X). The main result of this section is that it is more or less homotopic to the identity map C(X) C(X). Theorem 2.4.3. Let X be a topological space and β : C(X) C(X) be the operator defined above via barycentric subdivision. Then there exists a homomorphism P n : C n (X) C n+1 (X) for each n such that (2.69) P n (σ) P n ( σ) = β(σ) σ for all σ C n (X). As a consequence, we see that (2.70) β m (σ) = β β... β(σ) = σ }{{} m for all σ H n (X) and m 0. This fact is important to us later on. To construct P n, we again find ξ n L n+1 ( n ) and let P n (σ) = σ ξ n, similar to (2.41). We just have to find ξ n such that n (2.71) ξ n ( 1) k δ k ξ n 1 = α n 1 n. k=0 We let e 1, e 2,..., e n, e n+1 be the n + 1 vertices of n, which are the standard basis of R n+1 when we embed n to R n+1 by (2.2). For every permutation τ of {1, 2,..., n + 1} and every 1 k n + 1 satisfying that τ(1) < τ(2) <... < τ(k), we define an oriented (n + 1)- simplex (2.72) γ τ,k = σ eτ(1) e τ(2)...e τ(k) q k...q n+1 where q j is the center of the simplex σ eτ(1) e τ(2)...e τ(j) for k j n + 1. Finally, we let (2.73) ξ n = ( 1) k 1 sgn(τ)γ τ,k τ k

2.5. PRELIMINARIES IN HOMOLOGICAL ALGEBRA 17 where sgn(τ) is the sign of the permutation τ, i.e., it is 1 if τ is an even permutation and 1 if τ is an odd permutation. It is not hard, though quite tedious, to check that (2.71) holds. 2.5. Preliminaries in Homological Algebra 2.5.1. Direct sums and products. Let {A i : i I} be a set of abelian groups indexed by a set I. Then the direct sum of A i is (2.74) A i = {(a i ) i I : a i A i, a i 0 only for finitely many i I}, i I while the direct product of A i is (2.75) A i = {(a i ) i I : a i A i }. i I Clearly, the direct sum of A i is a subgroup of their direct product. They are the same if I is finite but different if I is infinite and A i is nontrivial. In addition, it is easy to see that ) (2.76) Hom Z ( i I A i, Z = i I Hom Z (A i, Z). Let {A i : i I} and {B j : j J} be two sets of abelian groups. We have natural maps ( ) ( ) (2.77) A i Z B j A i Z B j i I i J and (2.78) ( ) ( ) A i Z B j i I i J (i,j) I J (i,j) I J A i Z B j sending (a i, b j ) to (a i b j ). It is easy to see that (2.77) is an isomorphism. However, (2.78) is usually not a surjection. Proposition 2.5.1. If I and J are infinite, each A i is nontrivial and torsion free and B j = Z, then the map (2.78) is not surjective. Proof. Obviously, we just have to show this for I = J = Z +. For each A i, we choose r i 0 A i. Let v (i,j) I J A i Z B j be given by v ij = r i 1 if i = j and 0 otherwise. Let v j = (v 1j, v 2j,...). Then v 1, v 2,... are linearly independent over Z. That is, the map (2.79) Z j=1 i=1 A i

18 2. SINGULAR HOMOLOGY sending (a j ) to a j v j is injective. If v lies in the image of (2.78), then m (2.80) v = x k y k where x k i I A i and y k j J B j. Let y k = (y k1, y k2,...). Then (2.81) v j = for all j. That is, (2.82) v j k=1 m y kj x k k=1 m Zx k k=1 for all j. This contradicts the fact that v 1, v 2,... are linearly independent over Z. 2.5.2. Direct limits. Let I be a partially ordered the set. A directed system {A i : i I} is a collection of abelian groups together with homomorphisms f ij : A i A j for i j satisfying f ik = f jk f ij for i j k. The direct limit (2.83) A = lim A i i I is an abelian group together with homomorphisms g i : A i A satisfying f i = g j f ij for i j and the universal property that if there are another A and homomorphisms g i : A i A with the same property, then there is a homomorphism h : A A with the commutative diagram g i (2.84) A i A g i h A Example 2.5.2. Let A be an abelian group, I be the set of all finite subsets i of A and A i be the subgroup of A generated by the elements in i. In other words, A i are all finitely generated subgroups of A. Obviously, {A i } forms a directed system with the direct limit A. One can think of lim as a functor from the category of directed systems to that of abelian groups. This is a very nice functor: it is exact. That is, for directed systems {A i }, {B i } and {C i }, the direct limit (2.85) 0 lim A i lim B i lim C i 0 i I i I i I of the exact sequences (2.86) 0 A i B i C i 0

2.6. UNIVERSAL COEFFICIENTS THEOREM FOR HOMOLOGY 19 is still exact. Therefore, the direct limit of the kernels is the kernel of the direct limits (2.87) lim ker(a i B i ) = ker(lim A i lim B i ) i I i I i I and the direct limit of the images is the image of the direct limits (2.88) lim im(a i B i ) = im(lim A i lim B i ). i I i I i I Consequently, for a directed system of chain complexes {K,i }, the direct limit of holomology is the homology of the direct limit: (2.89) lim H n (K,i ) = H n (lim K,i ) i I i I for all n. 2.6. Universal coefficients theorem for homology

Bibliography [F] Bill Fulton, Algebraic topology, a first course, GTM 153. [M] William Massey, Singular homology theory, GTM 70. [B-T] Raoul Bott and Loring W. Tu, Differential Forms in Algebraic Topology, GTM 82. 21