II

[]

2010

An Introduction to Algebraic Topology

Simplicial Homology
A Short Course
Katsuro Sakai
Institute of Mathematics University of Tsukuba

Preface
In Algebraic Topology, topological properties of topological spaces and continuous maps are studied by using algebraic methods (e.g., groups, homomorphisms, etc.). Homology Theory and Homotopy Theory are typical and classical. There are several Homology Theories. Among them, Simplicial Homology is most elementary and geometrical. As an introduction to Algebraic Topology, this course is provided. Our purpose is knowing how to associate Topology (spaces) with Algebra (groups) by learning the process of dening the homology, and how to use Algebra in Topology by applying the homology. The reader is required to familiar with basic language of sets and maps, elementary point-set topology of Euclidean spaces, linear algebra and elementary group theory. To prepare this notes, the author was inuenced from the book Topology (Japanese) by Ichiro Tamura (Iwanami Shoten Publ., 1972). In particular, the author owes the use of Mayer-Vietoris sequence for the proof of the topological invariance of homology to Tamuras book. The rst draft was written in 2001 based on the authors lecture notes for Topology I, which he had lectured several times the class of Mathematics Majors in the University of Tsukuba. The author is indebted to his colleague Kazuhiro Kawamura and his student Masato Yaguchi for their helpful suggestions on the rst draft of the manuscript. This has been used as a text in the class of Topology I for Mathematics Majors in the University of Tsukuba, 2001, 2005, 2009. This has been revised and improved after the lectures. Tsukuba, March, 2010 Katsuro Sakai

c 2010, K. Sakai

Contents
1 2 3 4 5 6 7 8 9 Convex Sets and Simplexes . . . . . . . . Simplicial Complexes and Polyhedra . . Simplicial Maps and Approximations . . Homology of an Oriented Complex . . . Homomorphisms and Simplicial Maps . . Mayer-Vietoris Exact Sequences and the Subdivision Operator . . . . . . . . . Topological Invariance of Homology . . . Homotopical Invariance of Homology and Homologies of Surfaces . . . . . . . . . . Relative Homology for Polyhedral Pairs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4 12 28 37 50

. . . . . . . . . . . . 57 . . . . . . . . . . . . 65 . . . . . . . . . . . . 72 . . . . . . . . . . . . 79

ii

Simplicial Homology
A Short Course

In this course, we study Simplicial Homology, which is a sequence of additive groups associated with a topological space, so called a polyhedron. A polyhedron is constructed by small pieces, so called simplexes. The totality of assembled simplexes constructing a polyhedron is called a simplicial complex, which reveals the geometrical structure of the space. These materials are rst studied, and then we see how to dene the homology group for a simplicial complex. As the most important fact, it is shown that homeomorphic polyhedra have isomorphic homologies. Some applications of homologies are given. The homologies of several surfaces are also calculated.

Notations
We use the following notations: N = 1, 2, . . . the set of natural numbers; the set of integers;

Z = 0, 1, 2, . . .

Z+ = N {0} = 0, 1, 2, . . . the set of non-negative integers; R = (, ) the real line; R+ = [0, ) the non-negative half line; I = [0, 1] the unit closed interval;

()

 X n = X X ; x X n , x = (x(1), . . . , x(n)),
n times

x(i) X the i-th coordinate of x, pri : X n X (pri (x) = x(i)) the projection; Rk the k-dimensional Euclidean space, x Rk , x = x, y Rk , t R, x + y = (x(1) + y(1), . . . , x(k) + y(k)) sum, tx = (tx(1), . . . , tx(k)) scalar product, ei Rk (i = 1, . . . , k) dened by ei (j) = B k = x Rk
def def k i=1

x(i)2 (Euclidean) norm,

1 if i = j, 0 if i = j;
k () k 1

x 1 the unit k-ball; x = 1 the unit (k 1)-sphere; x(i) = 1, x(i) 0 (1 i k)

Sk1 = Bk = x Rk k1 = x Rk
def k i=1

the standard (k 1)-simplex;

k i=1

k 1

k1 = x Rk
def

x(i) = 1, x(i) > 0 (1 i k)

k 1

the standard open (k 1)-simplex.

R2 R3 e3 2 e2 e1 e1

e2

In this course, for beginners, spaces are considered as subspaces of Euclidean space Rk . X Y X is homeomorphic to Y ; idX = id the identity map of X; intX A = int A the interior of A in X; bdX A = bd A the boundary of A in X; clX A = cl A the closure of A in X; diam A = sup{ x y | x, y A}
def

the diameter of A;
def

dist(A, B) = inf{ x y | x A, y B}. the distance between A and B

A pair of spaces means a pair (X, A) of a space X and its subspace A. A map (or a homeomorphism) f : (X, A) (Y, B) from a pair (X, A) of spaces to another pair (Y, B) means a map (or a homeomorphism) f : X Y with f (A) B (or f (A) = B), whence f |A : A B is also a map (or a homeomorphism). When there is a homeomorphism f : (X, A) (Y, B), we say that (X, A) is homeomorphic to (Y, B) and denote (X, A) (Y, B).

Convex Sets and Simplexes

1.1 Denition We call C Rk a convex set if x, y C, t [0, 1], tx + (1 t)y C. This condition is equivalent to the following:
n

x0 , x1 , . . . , xn C, z n ,
i=0

z(i + 1)xi C.

1.1 Exercise Show the above equivalence. 1.2 Example Every linear subspace of Rk , Bk and k1 are convex. For any A Sk1 = Bk , Bk \ A is convex. 1.2 Exercise Show that if C Rk is convex then cl C and int C are also convex. 1.3 Denition The convex hull A of A Rk is dened by A =
def

C C : convex in Rk , A C ,

which is the smallest convex set containing A. For a nite set, we simply denote {x0 , x1 , . . . , xn } = x0 , x1 , . . . , xn . Note that n1 = e1 , . . . , en Rn . 1.3 Exercise For A Rk , prove the following: A =
n i=0

z(i + 1)xi n N, xi A, z n .

1.4 Exercise For A Rk , prove the following: cl A = C C: closed convex in Rk , A C .

1.4 Denition We say that x0 , x1 , . . . , xn Rk are anely independent (or geometrically independent) if
n n

ti xi = 0,
i=0 i=0

ti = 0 t0 = t1 = = tn = 0,
()

that is, x1 x0 , . . . , xn x0 are linearly independent, hence n k. It is said that A Rk is in general position if any k + 1 points of A are anely independent. 4

Remark. Let A = {x0 , x1 , . . . , xn } Rk and n k. Then, A is in general position if and only if x0 , x1 , . . . , xn are anely independent. 1.5 Exercise Prove that if v0 , v1 , . . . , vn Rk are anely independent and n < k, there exist k n many vn+1 , . . . , vk Rk such that v0 , v1 , . . . , vk are anely independent. 1.5 Denition We call Rk an (n-dimensional) simplex or an nsimplex if there exist some anely independent v0 , v1 , . . . , vn Rk such that n n , = v0 , v1 , . . . , vn = i=0 z(i + 1)vi z where v0 , v1 , . . . , vn are called vertices of and n = dim is the dimension of . We denote (0) = {v0 , v1 , . . . , vn }. 1.6 Exercise In the above, prove that : n dened by (z) = n z(i + 1)vi is a homeomorphism. In particular, the i=0 representation n z(i + 1)vi of a point in is unique. i=0 1.6 Denition When a point in an n-simplex = v0 , v1 , . . . , vn is represented as x = n z(i + 1)vi , we call z = (z(1), . . . , z(n + 1)) n the i=0 barycentric coordinate of x with respect to v0 , v1 , . . . , vn . 1.7 Proposition If v0 , v1 , . . . , vn Rk and u0 , u1 , . . . , um Rk are anely independent, then v0 , v1 , . . . , vn = u0 , u1 , . . . , um {v0 , v1 , . . . , vn } = {u0 , u1 , . . . , um }. Hence, for each simplex Rk , (0) and dim are uniquely determined. Proof. Assume that {v0 , v1 , . . . , vn } = {u0, u1 , . . . , um }. Then, without loss of generality, we can assume that v0 = uj for any j = 0, 1, . . . , m. When we represent
m

(n ) n

()

v0 =
j=0

z(j + 1)uj , z m ,

there are distinct j1 = j2 such that z(j1 + 1), z(j2 + 1) (0, 1) because z(j + 1) = 1 (i.e., v0 = uj ) for all j = 0, 1, . . . , m. Choose > 0 so that 0 < z(j1 + 1) < 1, 0 < z(j2 + 1) < 1,

and let x =
j=j1 ,j2

z(j + 1)uj + (z(j1 + 1) )uj1 + (z(j2 + 1) + )uj2 , z(j + 1)uj + (z(j1 + 1) + )uj1 + (z(j2 + 1) )uj2 .
j=j1 ,j2

x =

Then, x = x (cf. Exercise 1.6). Since x , x v0 , v1 , . . . , vn , we can write

n n

x =
i=0

z (i + 1)vi , x =
i=0

z (i + 1)vi , z , z n .

It follows that 1 1 v0 = x + x = 2 2
n i=0

1 1 z (i + 1) + z (i + 1) vi , 2 2

which implies 1 z (1) + 1 z (1) = 1 (cf. Exercise 1.6). Since z (1), z (1) I, it 2 2 follows that z (1) = z (1) = 1, which means x = x = v0 a contradiction! Thus we have the result. 1.8 Proposition Let be an n-simplex in Rk . If = v0 , v1 , . . . , vn (where n = dim ), then v0 , v1 , . . . , vn are anely independent. Consequently, (0) = {v0 , v1 , . . . , vn }. Proof. To see that v1 v0 , . . . , vn v0 are linearly independent, let L=
n i=1 ti (vi

v0 ) ti R ,

which is the linear subspace of Rk spanned by v1 v0 , . . . , vn v0 . Note that x L for each x . It suces to show that dim L = n. By the denition of an n-simplex, we have anely independent u0 , u1 , . . . , un Rk such that = u0, u1 , . . . , un , whence u1 u0 , . . . , un u0 u0 L and u1 u0 , . . . , un u0 are linearly independent. For each j = 0, . . . , n, vj = n zj (i + 1)ui for some zj n . Note that zj (1) = 1 n zj (i + 1). i=0 i=1

For every j = 0,
n n

vj v0 =
i=0 n

zj (i + 1)ui
i=0

z0 (i + 1)ui

=
i=1 n

(zj (i + 1) z0 (i + 1))ui + (zj (1) z0 (1))u0

n

=
i=1 n

(zj (i + 1) z0 (i + 1))ui
i=1

(zj (i + 1) z0 (i + 1))u0

=
i=1

(zj (i + 1) z0 (i + 1))(ui u0 ).

Thus, L is spanned by u1 u0 , . . . , un u0 , hence {u1 u0 , . . . , un u0 } is a base for L, which implies dim L = n. 1.9 Denition For an n-simplex = v0 , v1 , . . . , vn Rk , we dene =
def n i=0 n i=0

z(i + 1)vi z n , i = 1, . . . , n + 1, z(i) > 0 z(i + 1)vi z n the interior of ; z(i + 1)vi z n , i = 1, . . . , n + 1, z(i) = 0
() ()

= =
def

= \ the boundary of ;
n

n i=0

b() =

def

i=0

1 vi the barycenter of n+1

(where b() ).

Remark. (1) When dim = 0, we have = and = . We identify a 0-simplex v0 = {v0 } with the point v0 . (2) The barycentric coordinate of x is uniquely determined with respect to the order v0 , v1 , . . . , vn (cf. Exercise 1.6). 1.7 Exercise For each k-simplex in Rk , show that intk = and bdk = . Note that if dim = k then intk = , hence = intk . For every n-simplex , it can be shown that (, ) (Bn , Sn1 ) (Theorem 1.16). 1.10 Proposition For an n-simplex = v0 , v1 , . . . , vn , diam = max vi vj 7 i, j = 0, 1, . . . , n .

Proof. For each x, y , it suces to nd i1 , i2 such that xy First, choose i1 so that x vi1 = maxi x vi . Let
n

vi1 vi2 .

y=
i=0

z(i + 1)vi , z n .

Then, it follows that

n n

xy = x
i=0 n

z(i + 1)vi
i=0

z(i + 1) x vi

z(i + 1) x vi1 = x vi1 .

i=0

Next, choose i2 so that vi1 vi2 = maxi vi1 vi , whence vi1 x vi1 vi2 by the same argument. Consequently, x y vi1 vi2 . 1.11 Denition A simplex is called a face of a simplex if (0) (0) , whence we denote . An n-dimensional face is simply called an n-face (0-face = vertex). When and = , we call a proper face of , and denote < . 1.8 Exercise For each k-simplex in Rk , show =
<
,

Supplement
1.12 Denition For a compact (= bounded closed) convex set A Rk with 0 int A, we dene pA : Rk R+ as follow: pA (x) = inf s > 0 s1 x A = inf s > 0 x sA , which is called the Minkowski functional. x A 0 s1 x sA tx

1.13 Proposition For a compact (= bounded closed) convex set A Rk with 0 int A, the Minkowski functional pA : Rk R+ is well-dened and continuous, and it has the following properties:

(1) pA (x) = 0 x = 0; (2) pA (x)1 x A for every x Rk \ {0}; (3) pA (tx) = tpA (x) for each x Rk and t (4) A = {x Rk | pA (x) 1}; 0;

(5) int A = {x Rk | pA (x) < 1}, hence bd A = {x Rk | pA (x) = 1}. Proof. (Well-denedness) Since 0 int A, we have 0 > 0 such that x < 0 implies x A. For each x Rk , t1 x A if t > x /0 . Thus, pA is well-dened. (1) It is obvious that pA (0) = 0. Assume x = 0. Since A is compact (bounded), tx A for suciently large t > 0, i.e., s1 x A for suciently small s > 0. Then pA (x) = 0. (2) We have t1 > t2 > > 0 such that t1 x A and ti pA (x) (i ). i Note that pA (x) = 0 by (1). Since A is closed in Rk , it follows that pA (x)1 x A, (3) The case t = 0 or x = 0 follows from (1). When t > 0 and x = 0, tpA (x) by the since (tpA (x))1 tx = pA (x)1 x A by (2), we have pA (tx) 1 tx A, it also follows from the denition that denition. Moreover, since pA (tx) pA (x) pA (tx)/t, hence tpA (x) pA (tx). 1 for every x A. Since 0 A, it suces to (4) By the denition, pA (x) 1 implies x A. Since pA (x)1 x A by (2), it follows show that 0 < pA (x) from the convexity of A that x = (1 pA (x))0 + pA (x)(pA (x)1 x) A. (5) If pA (x) = 1 then n+1 x= n n n+1
1

x A and x = lim

n+1 x, n

which means x int A. Hence, pA (x) < 1 for every x int A. Conversely, assume pA (x) < 1. Since 0 int A, we have 0 > 0 such that z < 0 implies z A. If x y < (1 pA (x))0 , then y = (1 pA (x))(1 pA (x))1 (y x) + pA (x)pA (x)1 x A because (1 pA (x))1 (y x) A and pA (x)1 x A. Therefore, x int A. (Continuity) If pA (x) > a > 0 then a1 x Rk \ A by the denition. Since k \ A is open in Rk , we have > 0 such that x y < implies a1 y Rk \ A, R 1 1 hence a1 pA (y) = pA (a1 y) > 1 by (3) and (4), so pA (y) > a. Thus, pA is lower semi-continuous. If pA (x) < b then pA (b1 x) = b1 pA (x) < 1 by (3), which means b1 x int A by (5). Choose 2 > 0 so that x y < 2 implies b1 y int A, hence b1 pA (y) = pA (b1 y) < 1 by (3) and (5), so pA (y) < b. Thus, pA is upper semi-continuous.

1.9 Exercise Show that the Minkowski functional pA in the above has also the following properties: (1) pA (x + y) pA (x) + pA (y) for every x, y Rk ;

(2) pA is a norm on Rk A = A. 1.14 Denition A convex set A Rk with intk A = is called a convex body.

1.15 Theorem Every compact convex body A Rk is homeomorphic to Bk . In fact, there exists a homeomorphism : Rk Rk such that (A) = Bk , hence (A, bdk A) (Bk , Sk1 ). Proof. Without loss of generality, we may assume that 0 int A. By using the Minkowski functional pA for A, we dene maps , : Rk Rk as follows: (x) = (x) = 0 pA (x)1 x x 0 x
1 p

if x = 0, if x = 0;

if x = 0, A (x)x if x = 0.

0 A 0 Bk

Since pA is continuous, these are continuous at x Rk \ {0}. To see the continuity at 0, assume xi Rk \ {0} and xi 0 (i ). Since pA (xi )1 xi A, it follows that pA (xi )1 xi < diam A, whence (xi ) = pA (xi )1 xi
2

xi diam A 0 (i ).

Therefore, is continuous at 0. On the other hand, by the continuity of pA , (xi ) = pA (xi ) pA (0) = 0 (i ), hence is also continuous at 0. For each x Rk \ {0}, observe that (x) = pA (x)1 x 2 , pA ((x)) = pA (x)1 x pA (x) = x , (x) = pA (x) and pA ((x)) = x
1

pA (x)2 ,

10

hence it follows that (x) = (x)

1

pA ((x))(x) = pA (x) x

(x) = x and

((x)) = pA ((x))1 (x) (x) = x pA (x)1 (x) = x Then, = id and = id, that is, is a homeomorphism with 1 = . Observe (Bk ) = {(x) | x (S
k1

1} = {y Rk | (y) 1} = A and

1}

= {y Rk | pA (y)

) = {(x) | x = 1} = {y Rk | (y) = 1} = {y Rk | pA (y) = 1} = bdk A.

Thus we have the result.

Note. In connection with the above, if Bk A Rk then int A = . This follows from the following: Brouwers Theorem on Invariance of Domain: For each X, Y Rk , X Y implies int X int Y . In this theorem, one should remark that X Y is not given by any homeomorphism of Rk onto itself.

1.16 Theorem For every n-simplex , (, ) (Bn , Sn1 ). Proof. Let n = x R n 0 = 0, e1 , . . . , en R . Then, n is a compact convex set in Rn and 0 intn n = n = x Rn 0 0
n i=1 x(i) n i=1 x(i) n

1, x(i)

0 (i = 1, . . . , n)

< 1, x(i) > 0 (i = 1, . . . , n) = .

Therefore, (n , n ) = (n , bdn n ) (Bn , Sn1 ) by Theorem 1.15. 0 0 0 0 For each n-simplex = v0 , v1 , . . . , vn Rk , v1 v0 , . . . , vn v0 are linearly independent, hence we can dene a linear injection : Rn Rk by (ei ) = vi v0 for each i = 1, . . . , n. Let h : Rn Rk be the embedding dened by h(x) = (x) + v0 . Observe that h(n ) = and h(n ) = . Then, it follows that 0 0 (, ) (n , n ) (Bn , Sn1 ). 0 0

11

Simplicial Complexes and Polyhedra

2.1 Denition A nite collection K of simplexes in Rk is called a simplicial complex if the following are satised: (K1) K, K, , .

(K2) , K, =

For a simplicial complex K, we dene dim K = max{dim | K} the dimension of K (K is n-dimensional if dim K = n); |K| =
def def ()

K=
K

=
K

Rk the polyhedron of K; n} the n-skeleton of K

K (n) = { K | dim (K (0) =

def K

(0) is the set of vertices of K);

()

mesh K = max{diam | K} the mesh of K.

def

Remark. In the above, K = K (n) if and only if dim K = n. The polyhedron |K| is a compact set in Rk because of the nite union of simplexes. The name mesh comes from the following fact (cf. Proposition 1.10): mesh K = mesh K (1) = max v0 v1 v0 , v1 K .

Note. In this lecture, we treat only nite simplicial complexes. However, innite simplicial complexes are useful and have many applications. For an innite simplicial complex K, the local niteness is usually assumed because the topology of |K| is unique, where K is said to be locally nite if each v K (0) is the vertex of only nitely many simplexes of K. Replacing Rk by innite-dimensional linear topological space such as Hilbert space 2 , we can also consider innite-dimensional simplicial complexes. In this case, the local niteness is no longer assumed. If K is locally nite, then |K| is locally compact and locally nite-dimensional (i.e., each point has a nite-dimensional neighborhood). There are innite-dimensional locally nite simplicial complexes. When K is not locally nite, the polyhedron |K| has two typical topologies, the metric topology and the weak (Whitehead) topology. Non-locally nite simplicial complexes are also very useful.

2.2 Denition For an n-simplex , we dene K() =

def

the simplicial complex of ,

where |K()| = and K()(0) = (0) ; 12

 K() =
def

< = K()(n1) the simplicial complex of ,

whence |K()| = . 2.3 Denition Let K be a simplicial complex. A simplicial complex L is called a subcomplex of K if L K. In other words, L K is a subcomplex of K if L satises (K1) (i.e., L, K() L). A pair (K, L) of simplicial complexes means a pair of a simplicial complex K and its subcomplex L. Remark. Let L be a subcomplex of K. Then |L| |K| and dim L Every subcomplex of L is also a subcomplex of K. 2.4 Example Let K be a simplicial complex. (1) Each n-skeleton K (n) of K is a subcomplex of K. (2) For each K, K() is a subcomplex of K. 2.5 Denition For a simplicial complex K and K, the star St(, K) and the link Lk(, K) of in K are dened as follows: St(, K) = K K such that
def

dim K.

Lk(, K) = St(, K) = .
def

K Lk(, K)

St(, K)

For each point x |K|, we dene the star St(x, K) and the link Lk(x, K) at x in K as follows: St(x, K) = K K such that
def

, x ,

Lk(x, K) = St(x, K) x .
def

The carrier of x in K is the simplex K with x .

13

2.1 Exercise (1) For each x |K|, show that | St(x, K)| is a neighborhood of x in |K|. (2) Give an example such that | Lk(x, K)| = bd|K| | St(x, K)|. (3) Show | St(x, K)| = {(1 t)x + ty | y | Lk(x, K)|, t I} and | St(, K)| = {(1 t)x + ty | x , y | Lk(x, K)|, t I}. (4) Prove that if x then St(x, K) = St(, K). In particular, St(, K) = St(b(), K) for every K. Moreover, show that Lk(x, K) = Lk(, K) if = {x}. 2.6 Proposition Let K be a simplicial complex and , K. (1) =

(2) = = . Proof. In both (1) and (2), the implication is trivial. The converse implication can be seen as follows: . If has a vertex other (1) Assume that = . Recall than vertices of , then any point of cannot belong to , which is a contradiction. Thus, = . (2) If = , then and by (1), which implies = . 2.7 Denition For a simplicial complex K and v K (0) , we dene OK (v) = | St(v, K)| \ | Lk(v, K)| = |K| \
def

{ K | v (0) }

{ | K, v (0) } the open star at v.

Remark. The open star OK (v) is an open neighborhood of v in |K|, and OK (v) v K (0) is an open cover of |K|. 2.8 Lemma Let v0 , . . . , vn K (0) . Then, v0 , . . . , vn K if and only if OK (v0 ) OK (vn ) = . Proof. If = v0 , . . . , vn K, then OK (vi ) for all vi , hence OK (v0 ) OK (vn ) = . Conversely, assume that OK (v0 ) OK (vn ) contains some x. Let K be the carrier of x, i.e., x . Then, each vi is a vertex of , which implies . Hence, v0 , . . . , vn K. v0 , . . . , vn 14

2.9 Denition A topological space X is a (topological) polyhedron if X |K| for some simplicial complex K, that is, there exists a homeomorphism h : |K| X, whence (K, h) is called a triangulation of X. A pair (X, Y ) of spaces is polyhedral if (X, Y ) (|K|, |L|) for some pair (K, L) of simplicial complexes, that is, there exists a homeomorphism h : |K| X such that h(|L|) = Y , whence (K, L, h) is called a relative triangulation of a pair (X, Y ). Remark. A polyhedron is compact and metrizable.
Note. Recall that only nite simplicial complexes are treated in this lecture. In case locally nite simplicial complexes are considered, polyhedra are locally compact.

()

2.10 Example For an n-simplex , there exists a homeomorphism h : (, ) (Bn , Sn1 ) by Theorem 1.16. Then, (K(), K(), h) is a relative triangulation of (Bn , Sn1 ), so (K(), h) and (K(), h|) are triangulations of Bn and Sn1 , respectively. 2.11 Example In the following, we illustrate triangulations of typical surfaces by gures, where surfaces are spaces which are locally homeomorphic to the Eucliean plane R2 , that is, each point has an open neighborhood homeomorphic to an open set in R2 . A surface is also called a 2-dimensional manifold. An n-dimensional manifold is dened as a space which is locally homeomorphic to the n-dimensional Euclidean space Rn . Surface
R2

15

(1) Sphere S2
v v cut and unfold v A triangulation of S2 v

cut and unfold

A triangulation of S2 w u v v w u w u w cut and unfold v u

w A triangulation of S2

Not a simplicial complex

16

(2) Torus T2

paste

cut and unfold

paste

A triangulation of T2

(3) Klein bottle K2

paste

cut and unfold

paste

A triangulation of K2

17

(4) Projective plane P2 = (R3 \ {0})/ = {[x] | x R3 \ {0}}, where x y t R \ {0} such that tx = y ; [x] = (R \ {0})x
def def

P2 = S2 / , where x y x = y
def

x3 [x] x 1 0 x1

R3

x3 [x] x

R3

x3 = 1

1 1 v = [e1 ] x1 0 1

x3 = 1

x2

x2

x, x S1 are identied v v paste v fold v v

u w v v v

u A triangulation P2

w v Not a triangulation P2 v

18

(5) Mbius band M2 o

cut and unfold

A triangulation of M2

Mbius band is contained in the projective plane. o

u
1

v B2 M2
2

M2 cut
2

B2 +

B2 +

v
1

u B2 paste
1

B2 +
2 1

v B2

B2 A triangulation of P2

19

Remark. (1) The projective plane P2 can be obtained by attaching Mbius o 2 2 band M with the disk B along their boundary circles. P2 = M2 S1 B2

cut and unfold paste S1

S1

S1

paste B2

The cross cap paste M2 S1 S1

B2 B2

Mbius band is also called the cross cap. o

20

Remark. (2) Klein bottle K2 can be obtained by attaching two copies of Mbius band M2 along their boundary circles. In other wards, K2 can be o obtained by attaching two cross caps arong their boundary circles. K2 = M2 S1 M2

M+ M2

M2 cut and paste M+ M M2

2 1

1 2

The under half of Klein bottle

Mbius band o

2.12 Denition Let K and K be simplicial complexes. It is said that K is a subdivision of K or K subdivides K if |K | = |K| and each K is contained in some K, whence we denote K K.
Note. In the above, if innite simplicial complexes are considered, it is required that each K contains only nitely many simplexes of K .

Remark. If K

K and K

K , then K

K.

2.2 Exercise Let L be a subcomplex of K and K a subdivision of K. Show that L = { K | |L|} is a simplicial complex which subdivides L. 2.13 Lemma Let 0 < 1 < < n be a sequence of simplexes and ui i (i = 0, 1, . . . , n). Then, u0 , u1, . . . , un are anely independent, whence u0 , u1, . . . , un is an n-simplex. In particular, b(0 ), b(1 ), . . . , b(n ) is an n-simplex 21

Proof. By inserting faces of decient dimensions, it can be assumed that dim i = i. Then, we can write i = v0 , . . . , vi and
i

ui =
j=0

zi (j + 1)vj , zi i .
n i=0 ti ui

To see that u0, u1 , . . . , un are anely independent, let n i=0 ti = 0. Observe that
n n i n i

= 0 and

ti ui =
i=0 i=0

ti
j=0

zi (j + 1)vj =
n i=0 j=0 n

ti zi (j + 1)vj ti zi (j + 1)vj .

=
j i

ti zi (j + 1)vj =
j=0 i=j

Since v0 , . . . , vn are anely independent and

n n n i n

ti zi (j + 1) =
j=0 i=j i=0

ti
j=0

zi (j + 1) =
i=0

ti = 0,

it follows that n ti zi (j + 1) = 0 for each j = 0, . . . , n. Since zi (j + 1) = 0 i=j for each i and j, it can be seen that tn = tn1 = = t0 = 0 by downward induction. 2.14 Denition For a simplicial complex K, we dene Sd K =
def

b(0 ), . . . , b(n )

0 < 1 < < n K

the barycentric subdivision of K.

The n-th barycentric subdivision is inductively dened as follows: Sdn K = Sd(Sdn1 K), where Sd0 K = K. Remark. In the above, by replacing the barycenters b() with any points v , we can dene a derived subdivision of K. The n-th derived subdivision is also dened by induction. 2.3 Exercise (1) Show that Sd K is a simplicial complex and Sd K K. (2) Show that if L is a subcomplex of K then Sd L is a subcomplex of Sd K. 22

1 b(1 )

b(2 )

0 = b((0 ) b(0 ), b(1 ), b(2 )

2.15 Proposition For an m-dimensional simplicial complex K, mesh Sd K Hence, limn mesh Sdn K = 0. Proof. It suces to show that, for each < K, b() b( ) m mesh K m+1 m mesh K. m+1

(see the remark for Denition 2.1). We write = v0 , v1 , . . . , vn . Choose j so that b( ) vj = maxi b( ) vi . Since b() , it follows from the argument in the proof of 1.13 that b() b( ) b( ) vj . Then,
n

b( ) vj =
i=0 n

1 vi vj n+1

1 1 vi vj = vi vj n+1 n+1 i=0 i=j m n diam mesh K. n+1 m+1 Thus, we have the result.
Note. We can prove the following theorem, but more preparations are necessary and it wont be used in this lecture, so the proof is omitted. Theorem (J.H.C. Whitehead): Let K1 and K2 be simplicial complexes with |K1 | = |K2 |. Then K has a subdivision K3 subdividing both K1 and K2 . Hence, if K1 K and K2 K, then there exists some K3 K subdividing both K1 and K2 . This theorem is valid for any innite simplicial complexes.

23

2.16 Denition A simplicial complex K with an order on K (0) is called an ordered (simplicial) complex if (0) is a totally ordered subset of K (0) for each K. 2.17 Example (1) A simplicial complex K with a total order on K (0) is an ordered complex. (2) For each simplicial complex K, the barycentric subdivision Sd K is an ordered complex with the order dened as follows: b() b( )
def

()

(i.e., b() < b( ) < ).

2.18 Lemma Let v0 , v1 , . . . , vn Rk be anely independent. Then, for each j = 0, . . . , n, (v0 , 0), . . . , (vj , 0), (vj , 1), . . . , (vn , 1) Rk R = Rk+1 are also anely independent. Proof. Assume
j n n+1

ti (vi , 0) +
i=0 i=j

ti+1 (vi , 1) = 0 and

i=0

ti = 0.

By observing each factor of the rst equation, we have

j n n

ti vi +
i=0 i=j j i=0 ti

ti+1 vi = 0 and
i=j

ti+1 = 0,

hence we have also follow that

= 0. Since v0 , . . . , vn are anely independent, it

t0 = = tj1 = tj + tj+1 = tj+2 = = tn+1 = 0. Then, tj (vj , 0) + tj+1 (vj , 1) = 0. Since (vj , 0), (vj , 1) Rk+1 are anely independent, we have also tj = tj+1 = 0. 2.19 Denition For an n-simplex = v0 , v1 , . . . , vn Rk with an order v0 < v1 < < vn (i.e., K() is an ordered complex), we dene K() I =
def

j = 0, 1, . . . , n such that (v0 , 0), . . . , (vj , 0), (vj , 1), . . . , (vn , 1) 0 i0 < < ij 24 ij+1 < < im n .

(vi0 , 0), . . . , (vij , 0), (vij+1 , 1), . . . , (vim , 1)

1 I v2 0 K() I v0 v1

2.20 Lemma In the above, |K() I| = I. Proof. By the denition, it is easy to see that |K() I| I. Let (x, t) I. If t = 1 then (x, t) (v0 , 1), . . . , (vn , 1) |K() I|. When t < 1, let x =
n i=0 n

z(i + 1)vi , z n . Then,

n

z(i + 1)
i=j+1

t<
i=j

z(i + 1) s < 1 so that

for some j = 0, 1, . . . , n, whence we can choose 0

n

t = sz(j + 1) +
i=j+1

z(i + 1).

Then, it follows that

n

(x, t) =
i=0 j1

z(i + 1)vi , t z(i + 1)(vi , 0) + (1 s)z(j + 1)(vj , 0)

i=0 n

+ sz(j + 1)(vj , 1) +
i=j+1

z(i + 1)(vi , 1)

(v0 , 0), . . . , (vj , 0), (vj+1, 1), . . . , (vn , 1) |K() I|. Thus, we have |K() I| = I. 25

2.21 Denition For an ordered complex K, let K I = =

def

K() I
K

(v0 , 0), . . . , (vj , 0), (vj+1, 1), . . . , (vn , 1) v0 , . . . , vn K, v0 < < vj vj+1 < < vn

and K {i} = {i} K , i = 0, 1.

def

2.22 Proposition For an ordered complex K, K I is a simplicial complex and |K I| = |K| I. Moreover, K I is an ordered complex with the lexicographical order, that is, (v, i) < (u, j) v < u or v = u, i < j. Proof. By the denition, it is obvious that K I satises (K1). To see (K2), let , K I such that = . We write = (v0 , 0), . . . , (vj , 0), (vj+1, 1), . . . , (vm , 1) and = (v0 , 0), . . . , (vj , 0), (vj +1 , 1), . . . , (vm , 1) , where v0 , . . . , vm K, v0 < < vj vj+1 < < vm and v0 , . . . , vm vj +1 < < vm . Let K, v0 < < vj {v0 , . . . , vj } = {v0 , . . . , vj } {v0 , . . . , vj } and {vj
+1 , . . . , vm

} = {vj+1 , . . . , vm } {vj +1 , . . . , vm }.
+1

Then, v0 < < vj

vj

< < vm and , 1)

(v0 , 0), . . . , (vj , 0), (vj

+1 , 1), . . . , (vm

= (v0 , 0), . . . , (vj , 0), (vj+1, 1), . . . , (vm , 1) (v0 , 0), . . . , (vj , 0), (vj +1 , 1), . . . , (vm , 1) . The convex hull of the above set is (v0 , 0), . . . , (vj , 0), (vj
+1 , 1), . . . , (vm

, 1) = .

Therefore, is a face of and . By the denition and Lemma 2.20, we have |K I| =

K

K() I =
K

|K() I| =
K

I = |K| I.

The proof is complete. 26

Remark. In the above, K {i}, i = 0, 1, are subcomplexes of K I and |K {i}| = |K| {i}. 2.4 Exercise Let I = {I, 0, 1} be the triangulation of I, with the order 0 < 1. Then, I I and (I I) I are triangulations of I2 and I3 , respectively. Represent them in gures.

27

Simplicial Maps and Approximations

3.1 Denition Let K and L be simplicial complexes. A simplicial map is a function : K (0) L(0) such that = v0 , . . . , vn K, (v0 ), . . . , (vn ) L, hence it induces a function : K L. Then, a simplicial map is denoted by : K L rather than : K (0) L(0) by idenstifying = , that is, :K = v0 , . . . , vn () = (v0 ), . . . , (vn ) L.

Remark. (1) In the above, it is not necessary that all (v0 ), . . . , (vn ) are distinct. Then, dim () dim but dim () = dim in general. When : K (0) L(0) is injective, dim () = dim and = K implies () = ( ), that is, : K L is injective. But, even if : K (0) L(0) is surjective, : K L is not surjective in general. (a) dim () < dim
v2 v0 v1 u0 = (v0 ) u1 = (v1 ) = (v2 )

(b)

(K (0) ) = L(0) but (K) = L

v2 v0 v1 u0 = (v0 ) u1 = (v1 ) u2 = (v2 )

(2) If L is a subcomplex of K, then the inclusion i : L K (i.e., i : L(0) K (0) ) is a simplicial map. In particular, idK : K K (i.e., idK (0) : K (0) K (0) ) is a simplicial map. 3.1 Exercise For a simplicial map : K L, show the following: (1) K implies () ( ). 28

(2) (K) = () K is a subcomplex of L. 3.2 Denition For a simplicial map : K L, the function || : |K| |L| is dened as follows:
n

||(x) =
i=0

z(i + 1)(vi ),

where v0 , . . . , vn K and
n

x=
i=0

z(i + 1)vi , z n .

This || is also called a simplicial map. 3.2 Exercise Show that || is well-dened, that is, ||(x) is uniquely determined independently from choice of K with x and an order of vertices of . Remark. (1) One should observe that | idK | = id|K|.

(2) For a simplicial map : K L and K, ||(b()) = b(()) in general.

v2 u1 = (v2 ) () = b( )

v0 b()

v1

u0 = (v0 ) = (v1 ) ||(b()) = 2 u0 + 1 u1 3 3

3.3 Lemma Let : K L be a simplicial map. (1) || : |K| |L| is continuous. (2) If : K (0) L(0) is injective then || is an embedding.

29

Proof. (1) For each n-simplex = v0 , . . . , vn K, we dene the homeomorphism : n by (z) = n z(i + 1)vi . Then, for each z n , i=0 (|| )(z) = n z(i + 1)(vi ). It follows that || is continuous, whence i=0 the restriction ||| = (|| ) 1 is also continuous. Since |K| is the union of nitely many simplexes, it follows that || is continuous. (2) For each n-simplex = v0 , . . . , vn K, () = (v0 ), . . . , (vn ) L is also an n-simplex, hence we have the homeomorphism : n () dened by (z) = n z(i + 1)(vi ). Let be the same as above. Since i=0 (z) = (|| )(z) for z n , ||| = 1 : () is a homeomorphism. Since () ( ) = ( ) for each , K, it follows that || is an embedding. Remark. In the above, || is not a homeomorphism even if : K (0) L(0) is bijective. See Remark (1) for Denition 3.1. 3.4 Denition A map f : |K| |L| is simplicial (with respect to K and L) if f = || for some simplicial map : K L. A simplicial homeomorphism (or a simplicial embedding) is a homeomorphism (or an embedding) which is simplicial. Remark. (1) In the above denition, f |K (0) = . Hence, a map f : |K| |L| is simplicial if and only if, for each = v0 , . . . , vn K, f (v0 ), . . . , f (vn ) L and n f
n i=0

z(i + 1)vi =
i=0

z(i + 1)f (vi), z n .

(2) In case |K| = |K | and |L| = |L |, a simplicial map f : |K| |L| with respect to K and L is not simplicial with respect to K and L in general (even if K K and L L, e.g., K = Sd K and L = Sd L). See Remark (2) after Denition 3.2
Note. The following can be proved: Theorem: Let f : |K| |L| be a simplicial map. Then, (1) for any L L, there exists some K with respect to K and L ; K such that f is simplicial L such

K and L L, there are K K and L (2) for any K that f is simplicial with respect to K and L . On the contrary to (1), for some K K, there exist no L is simplicial with respect to K and L .

L such that f

30

3.3 Exercise A simplicial embedding f : |K| |L| with respect to K and L is also simplicial with respect to the barycentric subdivisions Sd K and Sd L. 3.5 Lemma Let K1 , K2 and K3 be simplicial complexes. If : K1 K2 and : K2 K3 are simplicial maps, then the composition : K1 K3 is also simplicial and | | = || || : |K1 | |K3 |. 3.4 Exercise Prove the lemma above. 3.6 Denition A map f : |K| |L| be called a PL map (or a piece-wise linear map) if f is simplicial with respect to some subdivisions K and L of K and L, respectively. A PL homeomorphism (or a PL embedding) is a homeomorphism (or an embedding) which is PL.
Note. By using the theorem in the previous notes and the theorem due to J.H.C. Whitehead in the note after Proposition 2.15, it is easy to prove the following: Lemma: Let f : |K| |L| be a PL map. Then, (1) there exists some L L such that for any L with respect to some K K and L ; L , f is simplicial

PL PL PL

(2) if |K| = |K | and |L| = |L |, then f : |K | |L | is also PL. Then, the following is easily follows: Proposition: Let K1 , K2 and K3 be simplicial complexes. If f : |K1 | |K2 | and g : |K2 | |K3 | are PL maps, then the composition g f : |K1 | |K3 | is also PL.

3.7 Denition Let K and L be simplicial complexes. It is said that K is (simplicially) isomorphic to L (denoted by K L) if there exist simplicial maps : K L and : L K such that = idK and = idL , where : K L is called a (simplicial) isomorphism and : L K is the inverse of . (Recall that || : |K| |L| is called a simplicial homeomorphism.) It is said that K is combinatorially equivalent to L (denoted by K L) if K and L have subdivisions which are isomorphic to = each other. Remark. (1) K L K L but K L K L. = = (2) The relations and are equivalence relations among simplicial = complexes.

()

()

31

Note. It follows from the following theorem that the relation is transitive. = Theorem: Let K and K be subdivisions of a simplicial complex K. Then, there is a subdivision K of K such that K K and K K . More generally, the following also hold: Theorem: Let K1 and K2 be simplicial complexes with |K1 | = |K2 |. Then, there is a simplicial complex K such that K K1 and K K2 . In other words, |K1 | = |K2 | implies K1 K2 . =

3.5 Exercise Prove that a bijection : K (0) L(0) is a simplicial isomorphism if and only if v0 , . . . , vn K (v0 ), . . . , (vn ) L. 3.8 Proposition Let h : |K| |L| be a homeomorphism. Then, (1) h : |K| |L| is simplicial if and only if there is a simplicial isomorphism : K L such that h = ||; (2) h : |K| |L| is PL if and only if there are subdivisions K K and L and a simplicial isomorphism : K L such that h = ||. L Proof. (1) Since the if part is trivial, it suces to show the only if part. If h : |K| |L| is a simplicial homeomorphism, then h|K (0) : K (0) L(0) is clearly a simplicial injection (cf. Remark (1) after Denition 3.4). For each u L(0) , let be the carrier of h1 (u). Then, h(v) = u for any vertex v of , which implies that is itself a vertex, i.e., h1 (u) K (0) . Therefore, h|K (0) : K (0) L(0) is a bijection. It remains to show that for each v0 , . . . , vn K (0) , h(v0 ), . . . , h(vn ) L v0 , . . . , vn K. Let = h(v0 ), . . . , h(vn ) L and K be the carrier of h1 (b( )). Then, h() = , whence h(), which means that h(v0 ), . . . , h(vn ) are vertices of h(), i.e., v0 , . . . , vn are vertices of . Therefore, v0 , . . . , vn K. (2) The if part is trivial. If h : |K| |L| be a PL homeomorphism, then h is simplicial with respect to subdivisions K K and L L. Hence, the only if part follows from (1). 3.9 Corollary For simplicial complexes K and L, K L if and only if there = exists a PL homeomorphism h : |K| |L|. Consequently, K L implies = |K| |L|.

32

 Note. It had been conjectured that |K| |L| implies K = L. This conjecture was called Hauptvermutung. All attempts in proving this conjecture were in vain. Finally, this has been known to be false.

()

3.10 Denition Let f : |K| |L| be a continuous map. A simplicial map : K L (or || : |K| |L|) is a simplicial approximation of f if ||(x) is contained in the carrier of f (x) in L for each x |K|.
() (v1 ) v0 x v1 f || (v0 ) ||(x) f (x) f ()

3.11 Proposition Let K be a simplicial complex and : (Sd K)(0) K (0) . Then, is a simplicial map which is a simplicial approximation of id|K| if and only if (b()) is a vertex of for each K. Hence, id|K| has a simplicial approximation : Sd K K. Proof. First, note that is the carrier of b(). Then, the only if part is obvious. To see the if part, assume that satises the condition. To show that (0) (0) is a simplicial map, let 0 < 1 < < n K. Since (b(i )) i n , (b(0 )), . . . , (b(n )) is a face of n , hence (b(0 )), . . . , (b(n )) K. Therefore, is a simplicial map. For each x | Sd K| = |K|, there are 0 < 1 < < n K such that b(0 ), . . . , b(n ) is the carrier of x in Sd K. Then, n is the carrier of x in K. As saw in the above, (b(0 )), . . . , (b(n )) is a face of n , hence ||(x) n . 3.12 Lemma Let f : |K| |L| be a continuous map and : K (0) L(0) . Then, is a simplicial map which is a simplicial approximation of f if and only if f (OK (v)) OL((v)) for each v K (0) . Proof. Assume that is a simplicial approximation of f . Let v K (0) . For each x OK (v), let K be the carrier of x and L the carrier of f (x). Then v is a vertex of . Since ||(x) and () is the carrier of ||(x),

33

() = , which means () , hence (v) is a vertex of . Thus, f (x) OL ((v)). Hence, we have f (OK (v)) OL ((v)). Conversely, assume that satises the condition. For each v0 , . . . , vn K. Then n OK (vi ) = by Lemma 2.8. By the condition, i=0
n n

OL((vi ))
i=0 i=0

f (OK (vi )) f

n i=0

OK (vi ) = ,

which means that (v0 ), . . . , (vn ) L by Lemma 2.8. Then is a simplicial map. For each x |K|, let v0 , . . . , vn K be the carrier of x. Since x n OK (vi ), it follows from the condition that f (x) n OL ((vi )). i=0 i=0 Hence, every (vi ) is a vertex of the carrier of f (x), whence ||(x) (v0 ), . . . , (vn ) . Therefore, is a simplicial approximation of f . 3.13 Proposition Let K1 , K2 and K3 be simplicial complexes. Suppose that continuous maps f : |K1 | |K2 | and g : |K2 | |K3 | have simplicial approximations : K1 K2 and : K2 K3 , respectively. Then, the composition : K1 K3 is a simplicial approximation of the composition gf with respect to K1 and K3 . 3.6 Exercise By using Lemma 3.12, prove the proposition above. 3.14 Theorem Every continuous map f : |K| |L| has a simplicial approximation : Sdn K L for some n N. Proof. Since |K| is compact, there is a Lebesgue number > 0 for the open cover f 1 (OL(u)) u L(0) of |K| (i.e., A |K| with diam A < is contained in some f 1 (OL(u))). By Proposition 2.15, mesh Sdn K < 1 for 2 some n N. Then, for each v (Sdn K)(0) , diam | St(v, Sdn K)| < , whence St(v, Sdn K) is contained in some f 1 (OL (u)). Let : (Sdn K)(0) L(0) be a map such that f (OSdn K (v)) OL ((v)) for each v (Sdn K)(0) . Then, is a simplicial map which is a simplicial approximation of f by Lemma 3.12. Remark. In general, without subdividing K, we cannot obtain a simplicial approximation of a continuous map f : |K| |L|. For instance, let K = L be the triangulation of [0, 2] with K (0) = {0, 1, 2}. Consider the continuous map f : |K| |L| dened by f (x) = 0 for x 1/ 2 and f (x) = 2x 1 for 34

 x 1/ 2. Then, f has no simplicial approximations with respect to K and L because f (OK (1)) = [0, 2 2 1) is not contained in any OL(v), v L(0) . By the way, it should be noticed that f 2 = f f has a simplicial approximation : K L dened by (0) = (1) = 0 and (2) = 1. 3.15 Proposition Let f, g : |K| |L| be continuous maps. Suppose that, for each x |K|, there is a vertex u L(0) such that f (x), g(x) OL(u). Then, there exist n N and a simplicial map : Sdn K L which is a simplicial approximation of both f and g. Proof. By the assumption, f 1 (OL(u)) g 1 (OL (u)) u L(0) is an open cover of |K|. Similarly to Theorem 3.14, let > 0 be a Lebesgue number for this cover and choose n N so that mesh Sdn K < 1 . Then, for 2 each v (Sdn K)(0) , there is some u L(0) such that OSdn K (v) St(v, Sdn K) f 1 (OL(u)) g 1 (OL(u)), hence we have : (Sdn K)(0) L(0) such that f (OSdn K (v)) g(OSdn K (v)) OL((v)) for each v (Sdn K)(0) . Then, is a simplicial map which is a simplicial approximation of both f and g by Lemma 3.12. 3.16 Denition Let f, g : |K| |L| be simplicial maps. It is said that f is simplicially homotopic to g (or f and g are simplicially homotopic) if there exists a simplicial map h : |K I| |L| such that h(x, 0) = f (x) and h(x, 1) = g(x), where K (0) is given some order so that K is an ordered complex. We call h a simplicial homotopy (from f to g) and denote f = h0 and g = h1 . When f = ||, g = | | for simplicial maps , : K L, it is also said that is simplicially homotopic to . Moreover, when h = || for a simplicial map : K I L, is also called a simplicial homotopy (from to ). 3.7 Exercise For an ordered complex K and i = 0, 1, we dene the injection ini : |K| |K I| by ini (x) = (x, i). Prove that ini : |K| |K {i}| is a simplicial homeomorphism and hi = h ini . 3.17 Theorem Let , : K L be simplicial approximations of a continuous map f : |K| |L|. Then, and are simplicially homotopic. 35

Proof. Give an order on K (0) so that K is an ordered complex. We dene : (K I)(0) L(0) by (v, 0) = (v) and (v, 1) = (v) for each v K (0) . Since and are simplicial approximations of f , f (OK (v)) OL ((v)) OL ( (v)) for each v K (0) . For each n-simplex v0 , . . . , vn K with v0 < v1 < < vn and j = 0, . . . , n,
j n j n

OL((vi , 0))
i=0 i=j

OL ((vi , 1)) =
i=0 n

OL((vi ))
i=j

OL ( (vi ))
n i=0

i=0

f (OK (vi )) f

OK (vi ) = ,

which implies that (v0 , 0), . . . , (vj , 0), (vj , 1), . . . , (vn , 1) L. Then, is a simplicial map, namely a simplicial homotopy from to .

36

Homology of an Oriented Complex

4.1 Denition Let = v0 , . . . , vn be an n-simplex. For two ordered (n + 1)-tuples (vi0 , . . . , vin ) and (vj0 , . . . , vjn ) of all vertices v0 , . . . , vn of , we dene i , . . . , in (vi0 , . . . , vin ) (vj0 , . . . , vjn ) 0 is even, j0 , . . . , jn def i0 , . . . , in is a permutation of (0, 1, . . . , n). Then, the relation is j0 , . . . , jn an equivalence relation on the set of (n + 1)-tuples of all vertices v0 , . . . , vn , which has two equivalence classes if n > 0 (only one when n = 0). The equivalence classes are called orientations of and one is called the inverse of the other. When one of them is decided, it is said that is given an orientation. An oriented simplex is a simplex given an orientation. The equivalence class of an (n + 1)-tuple (vi0 , . . . , vin ) is denoted by [vi0 , . . . , vin ] and its inverse is by [vi0 , . . . , vin ]. where Remark. One should observe the following: [vi0 , . . . , vin ] = [v0 , . . . , vn ] [vi0 , . . . , vin ] = [v0 , . . . , vn ]
[v0 , v1 ] v0 v2 [v0 , v1 , v2 ] [v2 , v1 , v0 ] v1 v0 v2

( )

0, . . . , n i0 , . . . , in 0, . . . , n i0 , . . . , in

is even; is odd.

[v1 , v0 ] v1

v0

v1

v0

v1

Note. In the above, an (n + 1)-tuple (vi0 , . . . , vin ) of all vertices is called an ordered (n-)simplex. In other words, an ordered simplex is a simplex given a total order on the vertices. For each n-simplex, (n+1)! many ordered n-simplexes are dened.

37

4.2 Denition It is said that a simplicial complex K is given an orientation when each simplex of K is given an orientation. A simplicial complex K given an orientation is called an oriented (simplicial) complex.

( ) ()

An oriented complex
Note. Let K be a simplicial complex such that K \ K (0) has m many simplexes. Then, K can be given 2m many orientations.

Remark. By giving an order on K (0) so that K is an ordered complex (i.e., (0) is totally ordered for each K), we can give an orientation on K. For instance, the barycentric subdivision Sd K is an oriented complex by the natural order.

The orientation of Sd K by the natural order 4.3 Denition For a nite set = {1, . . . , r }, the free additive group over (or generated by ) is the set of all formal sums r ai i , a1 , . . . , ar i=1 Z, with the addition dened by
n n n

ai i +
i=1 i=1

bi i =
i=1

(ai + bi )i ,

where we write 1i = i and 0i = 0. We call the set of generators of

38

this group. This group is isomorphic to the direct sum Z Z.

r times

We call r the rank of the free additive group over . In other words, the rank is the number of elements of generators.
Note. When is innite, the free additive group over is dened as the set n of all formal nite sums i=1 ai i , n N, a1 , . . . , an Z, 1 , . . . , n , with the addition dened as above. To avoid ambiguity, we can dene the free additive group over as Z f the additive group consisting of functions : Z such that () = 0 except for nitely many , where the addition is dened point-wise (i.e., ( + )() = () + ()). In case = {1 , . . . , r }, Z is the group Z f consisting of all functions from to Z, whence Z Zr Z Z . = =
r times

For each , dene Z by () = 1 and ( ) = 0 for all = . f n Then, every Z can be written as = i=1 ai i , where {1 , . . . , n } = f 1 \ (0) and ai = (i ) Z. By identifying = , we ca regard as the n formal sum i=1 ai i .

4.4 Denition Let K be an oriented complex. For each n Z+ , let K[n] = K (n) \ K (n1) . The (oriented) n-dimensional chain group Cn (K) of K is dened as the free additive group over K[n], that is, all oriented n-simplexes of K are generators of Cn (K). Each c = r ki i Cn (K) is called an i=1 n-chain. For each K, 1 is identied with , the inverse of . The K boundary operator n = n : Cn (K) Cn1 (K) is the homomorphism dened by
n K n ([v0 , . . . , vn ]) = i=0

n ()

(1)i [v0 , . . . , vi , . . . , vn ],

where we mean [v0 , . . . , vi , . . . , vn ] = [v0 , . . . , vi1 , vi+1 , . . . , vn ]. Let 0 = 0. Thus, we have the following sequence:
n C(K) : Cn (K) Cn1 (K) 2 1 0 C1 (K) C0 (K) 0.

n+1

n1

39

K The system C(K) = (Cn (K), n )n 0 or the sequence above is called the (oriented) chain complex of K. For each n Z+ , let

()

Zn (K) = Ker n = c Cn (K) n (c) = 0 ,

def def

Bn (K) = Im n+1 = c Cn (K) c Cn+1 (K) such that n+1 (c ) = c . Each c Zn (K) is called an n-cycle and each c Bn (K) an n-boundary or (a bounding n-cycle). Remark. In the denition of the chain groups Cn (K), it is not necessary to give K an orientation. In fact, we can obtain Cn (K) as the free additive group over the set K[n] of all n-simplexes of K without orientations. However, the K orientation of K is essential to dene the boundary operator n . Each nsimplex has n + 1 many (n 1)-faces but we do not know how to make the alternating sum of (n 1)-faces without orientation. 4.5 Fact For every oriented complex K with dim K = m, (1) Z0 (K) = C0 (K) and (2) Cn (K) = 0 for n > m, hence Bn (K) = 0 for n m.
n n

4.6 Denition For an oriented n-simplex = [v0 , . . . , vn ] K, an (n 1)face is said to be inherited the orientation from if there is some i = 0, . . . , n such that = (1)i [v0 , . . . , vi , . . . , vn ]. For two oriented simplexes , K, the incidence number [; ] is dened as follows: 1 if , dim = dim 1 and is inherited the orientation from , [; ] = 1 if , dim = dim 1 and is not inherited the orientation from , 0 otherwise (i.e., or dim = dim 1). Since K[n] = K (n) \ K (n1) is nite, an n-chain is represented by c=
K[n]

a . 40

By using the incidence number, we can write n (c) =

K[n1] K[n]

a [; ] .

4.1 Exercise Prove the above equation. 4.7 Lemma For each n 0, n n+1 = 0, i.e., Bn (K) Zn (K).

Proof. The case n = 0 is obvious. For n > 0 and [v0 , . . . , vn+1 ] K, n n+1 ([v0 , . . . , vn+1 ]) = n
n+1 n+1 i i=0 (1) [v0 , . . . , vi , . . . , vn+1 ]

=
i=0 n+1

(1)i n ([v0 , . . . , vi , . . . , vn+1 ])

i1

=
i=0

(1) +

i j=0

(1)j [v0 , . . . , vj , . . . , vi , . . . , vn+1 ] (1)j1 [v0 , . . . , vi , . . . , vj , . . . , vn+1 ]

n+1

j=i+1

=
j<i

(1)i+j [v0 , . . . , vj , . . . , vi , . . . , vn+1 ] +

i<j

(1)i+j1 [v0 , . . . , vi , . . . , vj , . . . , vn+1 ]

=
j<i

(1)i+j [v0 , . . . , vj , . . . , vi , . . . , vn+1 ]

i<j

(1)i+j [v0 , . . . , vj , . . . , vi , . . . , vn+1 ] = 0.

Then, it follows that n n+1 = 0. Recall the following basic theorem in Group Theory: Fundamental Homomorphism Theorem: For every homomorphism f : G H between groups, Im f G/ Ker f. = The following is the direct consequence of this theorem. 4.8 Proposition For each n > 0, Bn1 (K) Cn (K)/Zn (K). = 41

4.9 Denition Let K be an oriented complex and n Z+ . The n-dimensional homology group Hn (K) of K dened as the following factor group: Hn (K) = Zn (K)/Bn (K). For each z Zn (K), we denote z K = z + Bn (K) Hn (K), which is called the homology class of z (in K). We denote H (K) = (Hn (K))n 0 . 4.10 Fact For every oriented complex K with dim K = m, (1) H0 (K) = C0 (K)/B0 (K) and Hm (K) = Zm (K); (2) Hn (K) = 0 for n > m. Remark. When L is a subcomplex of K with the orientation inherited from K, we can naturally regard Cn (L) Cn (K), Zn (L) Zn (K) and Bn (L) Bn (K). Then, we have to make clear that the homology class of z Zn (L) means which class z + Bn (L) or z + Bn (K), so we write z L = z + Bn (L) or z K = z + Bn (K).
Note. In general, consider a system of Abelian groups and homomorphisms:
n C : Gn+1 Gn Gn1 .

n+1

We call C = (Gn , n )nZ a chain complex if n1 n = 0. The n-dimensional homology group of C is dened as Hn (C) = Ker n / Im n+1 . Then, Hn (K) = Hn (C(K)), where Cn (K) = 0 for all n < 0.

()

4.11 Example (1) The simplicial complex {v} of only one vertex is the simplest one. For this complex, H0 ({v}) = Z0 ({v}) = C0 ({v}) Z and = Hn ({v}) = 0 for all n > 0. (2) Let = v0 , v1 , v2 be a 2-simplex and give the orientation of K() by the order v0 < v1 < v2 . Then, Hn (K()) = 0 for all n > 2. Observe C2 (K()) = Z[v0 , v1 , v2 ] ( Z); = C1 (K()) = Z[v0 , v1 ] + Z[v0 , v2 ] + Z[v1 , v2 ] ( Z Z Z); = C0 (K()) = Zv0 + Zv1 + Zv2 ( Z Z Z). =

42

For c = a[v0 , v1 , v2 ] C2 (K()), 2 (c) = a[v1 , v2 ] a[v0 , v2 ] + a[v0 , v1 ]. Then, 2 (c) = 0 if and only if a = 0. Therefore, H2 (K()) = Z2 (K()) = 0. Moreover, we have B1 (K()) = Z([v1 , v2 ] [v0 , v2 ] + [v0 , v1 ]). Now, consider a 1-chain c = a1 [v0 , v1 ] + a2 [v0 , v2 ] + a3 [v1 , v2 ] C1 (K()). Then, observe 1 (c) = a1 (v1 v0 ) + a2 (v2 v0 ) + a3 (v2 v1 ) = (a1 + a2 )v0 + (a1 a3 )v1 + (a2 + a3 )v2 . Hence, 1 (c) = 0 if and only if a1 = a2 = a3 . Thus, Z1 (K()) = Z([v0 , v1 ] [v0 , v2 ] + [v1 , v2 ]) = B1 (K()), which means H1 (K()) = 0. Moreover, in the above, 1 (c) = (a1 a3 )(v1 v0 ) + (a2 + a3 )(v2 v0 ) because v2 v1 = (v2 v0 ) (v1 v0 ). Then, it follows that B0 (K()) = Z(v2 v0 ) + Z(v1 v0 ). For c = a1 v0 + a2 v1 + a3 v2 C0 (K()), we can write c = a2 (v1 v0 ) + a3 (v2 v0 ) + (a1 + a2 + a3 )v0 , which implies that Z0 (K()) = C0 (K()) = Z(v2 v0 ) + Z(v1 v0 ) + Zv0 B0 (K()) Z. = Then, we have H0 (K()) Z. = (3) Consider the subcomplex K() of the complex K() in (2) above. Then, Hn (K()) = 0 for all n > 1. Moreover, we have C1 (K()) = C1 (K()) C0 (K()) = C0 (K()) ( Z Z Z); = Z Z Z). (=

By the same argument as (2), H0 (K()) = H0 (K()) Z. Finally, it = Z. follows that H1 (K()) = Z1 (K()) = Z1 (K()) = 43

Remark. Let H be a subgroup of G = Z Z (= Z2 ) with H Z. But it = Z in general. For example, let should be remarked that G/H = H = (2n, 2n) G n Z = Z(2, 2) = 2Z(1, 1), (i.e., H is generated by (2, 2) G). Since each (a, b) G can be written (a, b) = a(1, 1) + (b a)(0, 1), we have G = Z(1, 1) Z(0, 1), hence G/H = Z2 Z, where Z2 = Z/2Z is the cyclic group with order 2 4.2 Exercise For a 3-simplex = v0 , v1 , v2 , v3 , calculate the homology groups (1) Hn (K()), (2) Hn (K()) and (3) Hn (K()(1) ), where the orientation is given by the order v0 < v1 < v2 < v3 . 4.3 Exercise Let K = I I be the triangulation of I2 dened in Exercise 2.4 whose orientation induced by the lexicographic order (cf. Proposition 2.22). Calculate the homology groups (1) Hn (K) and (2) Hn (K (1) ).
Note. For a simplicial complex K without orientation, we can dene the homology of K as follows: (1) We dene Cn (K) as the free additive group over the set of all ordered n-simplexes of K, whence rank Cn (K) = (n + 1)!n (where n is the number of n-simplexes of K). Let n : Cn (K) Cn1 (K) (n > 0) be the homomorphism dened as follows:
n

n (v0 , . . . , vn ) =
i=0

(1)i (v0 , . . . , vi , . . . , vn ).

Then, C (K) = (Cn (K), n )n 0 is a chain complex. Although Cn (K) = Cn (K) for an oriented complex K with dim K > 0, it can be proved that Hn (K) Hn (C (K)) for all n 0. = (2) We can also dene On (K) as the free additive group over the set of all (n + 1)-tuples of vertices of simplexes of K, that is, (v0 , . . . , vn ) v0 , . . . , vn K

is the base for On (K), where it is not required that all vi are distinct. Then, rank On (K) is larger than rank Cn (K). Moreover, On (K) = 0 even if n > dim K. We call On (K) the n-dimensional ordered chain group of K. O Similarly to the above, we dene n : On (K) On1 (K) as follows:
O n (v0 , . . . , vn ) = n i=0

(1)i (v0 , . . . , vi , . . . , vn ).

44

O Thus, another chain complex O(K) = (On (K), n )n 0 is obtained, which is called the ordered chain complex of K. It can be also proved that Hn (K) Hn (O(K)) for all n 0. =

In the above, avoiding orientation causes the increase in the numbers of generators of the chain groups. As a consequence, it is harder to compute the homology groups even in the simple cases as in Example 4.11 or Exercises 4.2 and 4.3. In general case, even if we use orientation, it is not easy to compute directly the homology.

Recall the following very important theorem in Group Theory: Fundamental Theorem on Abelian Groups: Every nitely generated Abelian group G is isomorphic to the direct sum of some cyclic groups, that is, Z Z Z1 Z2 Zs ,
r times

where Zi is the cyclic group with order i > 1, each i is a divisor of i+1 and r = rank G is the rank of G. The following is also important: Theorem on Free Additive Groups: Every subgroup H of a free additive group G is also free and rank H rank G. Moreover, rank G/H = rank G rank H. By these theorems, we have Hn (K) Z Z Z1 Z2 Zs . =
r times

where r = rank Hn (K) = rank Zn (K) rank Bn (K). Combining Proposition 4.8 with the theorem above, we have rank Bn1 (K) = rank Cn (K) rank Zn (K). 4.12 Denition We call rank Hn (K) the n-dimensional Betti number of K. The alternating sum
m n

(K) =
i=0

(1)i rank Hi (K) (m = dim K)

is called the Euler-Poincar characteristic of K. e 4.4 Exercise For each oriented complex in Example 4.11, calculate Euler-Poincar characteristic. e 45

 4.13 Theorem (Euler-Poincare Formula) For an oriented complex K and n Z+ , let n be the number of n-simplexes of K and m = dim K. Then,
m

(K) =
i=0

(1)i i .

Proof. Note that Bm (K) = 0 and Z0 (K) = C0 (K). As saw in the above, we have rank Zi (K) + rank Bi1 (K) = rank Ci (K). Since rank Ci (K) = i , it follows that
m

(K) =
i=0 m

(1)i rank Hi (K) (1)i (rank Zi (K) rank Bi (K))

i=0 m

= rank Z0 (K) +
i=1 m

(1)i (rank Zi (K) + rank Bi1 (K))

m

=
i=0

(1) rank Ci(K) =

i=0

(1)i i .

4.5 Exercise For each ordered complex in Example 4.11, conrm Euler-Poincar formula by computation. e Remark. Later, we shall show that |K| |L| implies H (K) H (L) for = every oriented complexes K and L. This fact is called the topological invariance of homology. Then, it follows that Euler-Poincar characteristic is e a topological invariant, that is, |K| |L| implies (K) = (L) for every simplicial complexes K and L. By this fact, we can see that |K| |L| by computing Euler-Poincar characteristics (K) and (L). e 4.6 Exercise On each surfaces in Example 2.11, apply EulerPoincar formula to calculate Euler-Poincar characteristic. By e e using Euler-Poincar characteristic, which surfaces are not homee omorphic to each others? Recall that a space X is connected if X cannot be the union of disjoint non-empty open sets, and that X is path-connected if each points x, y X can be connected by a path in X, where a path is a continuous map f : I X with f (0) = x and f (1) = y. 46

4.14 Theorem For an oritented complex K, the following conditions are equivalent: (1) H0 (K) Z; = (2) |K| is path-connected; (3) |K| is connected. Proof. (1) (2): By (1), we have H0 (K) = Zv K for some v K (0) . For each x |K|, it suces to construct a path f : I |K| with f (0) = v and f (1) = x. Let K be the carrier of x and u (0) . Then, uK = svK for some s Z, that is, u sv = 1 (c) for some a 1-chain c = m ti [vi , ui ] C1 (K), i=1 where vi = ui for every i. Thus, we have
m

sv = u 1 (c) = u
i=1

(ti ui ti vi ).

We can rearrange [vi , ui ] so that vi = v for 0 < i k; vi = v for k < i m; ; ui = v for < i m (ui = v for 0 < i k). Then, it ui = v for k < i follows that
k

ti
i=1 k i=k+1

ti = s and
m m

1
i=1

ti +
i=k+1

ti
i= +1

ti +
i= +1

ti = 0,

hence we have s = 1. Therefore,

m

u v = 1 (c) =
i=1

(ti ui ti vi ).

By replacing vi and ui when ti < 0, we may assume that ti > 0 for all i. Moreover, since ti [vi , ui] is the sum of ti many [vi , ui], we may assume that ti = 1 for all i, that is, u v = m (ui vi ). Observe that if ui = u then i=1 ui = vj for some j. We can choose n1 so that v = vn1 . Assume distinct n1 , . . . , nk can be chosen so that uni = vni+1 . In case unk = vj for some j = n1 , . . . , nk , let nk+1 be such a j. Otherwise, it follows that
m

u=v+
i=1

(ui vi ) = unk +
j=n1 ,...,n k

(uj vj ),

47

whence u = vnk because unk = vj for any j = n1 , . . . , nk . Thus, we can choose distinct n1 , . . . , nk so that v = vn1 , un1 = vn2 , . . . , unk = u. We can dene a path f : I |K| as follows: i1 i (i (k + 1)t)v + ((k + 1)t i + 1)u if t , ni ni k+1 k+1 f (t) = k (k + 1 (k + 1)t)u + ((k + 1)t k)x t 1. if k+1 Then, f (0) = v and f (1) = x. (2) (1) For each distict v, u K (0) , we have a continuous map f : I |K| such that f (0) = v and f (1) = u. Let I be the natural triangulation of I. By Theorem 3.14, f has a simplicial approximation : Sdn I K for some n N. We can choose v0 , v1 , . . . , vk from {(2n i) | i = 0, . . . , 2n } so that v0 = v, vk = u, vi1 = vi and vi1 , vi K for every i = 1, . . . , k Then, we have c = i = 1k [vi1 , vi ] C1 (K). Since 1 (c) = vk v0 = u v, it follows that v K = uK . Consequently, H0 (K) is generated by only one element v K . It remains to show that sv K = 0 implies s = 0. If sv K = 0 H0 (K), then we have some 1-chain c = m ti [vi , ui] C1 (K) such that kv = 1 (c) = i=1 m (ti ui ti vi ). Note that vi = ui for each i = 1, . . . , m. We can rearrange i=1 k; vi = v for k < i m; ui = v for [vi , ui] so that vi = v for 0 < i k<i ; ui = v for < i m, where note that ui = v for 0 < i k. Then,
k

i=1 k

ti +
i=k+1 m

ti = s and
m

ti
i=1 i=k+1

ti +
i= +1

ti
i= +1

ti = 0,

whence we have s = 0. Therefore, H0 (K) = Zv K Z. = (2) (3) If |K| is disconnected, then |K| has disjoint non-empty open sets U, V such that |K| = U V . Let x U and y V . Since |K| is path-connected, we have a path f : I |K| such that f (0) = x and f (1) = y. Then, f 1 (U) and f 1 (V ) are non-empty open sets in I and I = f 1 (U) f 1 (V ). This contradicts to the connectedness of I. (3) (2) Fix any point v K (0) and dene U = x |K| x can be connected to v by a path in |K| . For each u K (0) , we show that OK (u) U = implies OK (u) U. Let x OK (u) U. Then, we have a path f : I |K| with f (0) = v and 48

f (1) = x. For each y OK (u), we dene a path g : I |K| as follows: f (3t) if 0 t 1 , 3 g(t) = (2 3t)x + (3t 1)u if 1 t 2 , 3 3 (3 3t)u + (3t 2)y if 2 t 1. 3 Then, g(0) = x and g(1) = y. Therefore, OK (u) U. Since |K| = uK (0) OK (u), it follows that U and |K| \ U are open. Since |K| is connected, we have |K| \ U = , that is, |K| = U, which means that every x |K| can be connected to v by a path in |K|. Thus, |K| is path-connected. 4.15 Denition A component of a simplicial complex K is a maximal subcomplex K0 of K such that |K0 | is connected, where |K0 | is a component of a space |K| (i.e., a maximal connected subset of |K|). 4.16 Theorem Let K = r Ki be an oriented complex, where Ki s are i=1 dierent components of K. Then Hn (K) Hn (K1 ) Hn (K2 ) Hn (Kr ) for each n Z+ . = In particular, H0 (K) Z Z, =
r times

hence rankH0 (K) is the number of components of K. Proof. Observe that, for each n 0,

Cn (K) = Cn (K1 ) Cn (K2 ) Cn (Kr ).

K K Since n i = n |Cn (Ki ) for each i = 1, . . . , r, it follows that

Zn (K) = Zn (K1 ) Zn (K2 ) Zn (Kr ), Bn (K) = Bn (K1 ) Bn (K2 ) Bn (Kr ). Therefore, we have the result.

49

Homomorphisms and Simplicial Maps

5.1 Denition Let K and L be oriented complexes. The homologies H (K) and H (L) are dened by using the chain complexes
K C(K) = (Cn (K), n )n 0 L and C(L) = (Cn (L), n )n 0 .

A sequence h = (hn )n 0 of homomorphisms hn : Cn (K) Cn (L) is called a chain homomorphism from C(K) to C(L) (denoted by h : C(K) L K C(L)) if n hn = hn1 n for each n > 0, that is, the following diagram is commutative: h Cn (K) n Cn (L) L K
n n

()

Cn1 (K) Cn1 (L)

hn1

Then, hn (Zn (K)) Zn (L) and hn (Bn (K)) Bn (L) for each n 0. Hence, h induces the homology homomorphism h : H (K) H (L) dened by L h (z K ) = hn (z) (i.e., h (z + Bn (K)) = hn (z) + Bn (L)) for each z Zn (K), which is called the homology homomorphism induced by h. When all hn are isomorphisms, h is called a chain isomorphism, we denote h1 = (h1 ), which is called the inverse of h. In case K = L (with n the same orientation), idC(K) = (idCn (K) )n 0 is the chain isomorphism. Remark. Obviously (idC(K) ) = idH (K) . 5.2 Proposition Let K1 , K2 and K3 be oriented complexes. For chain homomorphisms f = (fn )n 0 : C(K1 ) C(K2 ) and g = (gn )n 0 : C(K2 ) C(K3 ), the composition gf = (gn fn )n 0 : C(K1 ) C(K3 ) is a chain homomorphism. Moreover, (gf ) = g f : H (K1 ) H (K3 ). Proof. See the following diagram: n Cn (K2 ) n Cn (K3 ) Cn (K1 ) K2 K3 K1
n n hn1 hn1 h h

() () ()

Cn1 (K1 ) Cn1 (K2 ) Cn1 (K3 ) The proof is easy. 5.1 Exercise Complete the proof above. 50

5.3 Corollary Let K and L be oriented complexes. A chain isomorphism f = (fn )n 0 : C(K) C(L) induces a homology isomorphism f : H (K) 1 H (L) and f = (f 1 ) . Proof. In Proposition 5.2, let K1 = K3 = K (with the same orientation), K2 = L and g = f 1 . Then, we have (f 1 ) f = (idC(K) ) = idH (K) . Similarly, it follows that f (f 1 ) = idH (L) . Therefore, f is an isomorphism 1 and f = (f 1 ) . 5.4 Denition Let K and L be oriented complexes and : K L a simplicial map. For each n 0, we dene the homomorphism n : Cn (K) Cn (L) by n ([v0 , . . . , vn ]) = [(v0 ), . . . , (vn )] if all (vi ) are distinct, 0 otherwise

for each oriented n-simplex [v0 , . . . , vn ] of K. Remark. In the above, n ([v0 , . . . , vn ]) = 0 dim (v0 ), . . . , (vn ) < n. 5.5 Lemma In the above, = (n )n C(K) to C(L).
0

is a chain homomorphism from

Proof. Let [v0 , . . . , vn ] be an oriented n-simplex of K. Dividing three cases, we show that
L K n n ([v0 , . . . , vn ]) = n1 n ([v0 , . . . , vn ]).

(1) In case all (vi ) are distinct, that is, dim (v0 ), . . . , (vn ) = n,
L L n n ([v0 , . . . , vn ]) = n ([(v0 ), . . . , (vn )]) n

=
i=0 n

(1)i [(v0 ), . . . , (vi ), . . . , (vn )] (1)i n1 ([v0 , . . . , vi , . . . , vn ])

i=0 n i i=0 (1) [v0 , . . . , vi , . . . , vn ]

= n1

K = n1 n ([v0 , . . . , vn ]).

51

(2) In case dim (v0 ), . . . , (vn ) < n 1, for all i = 0, . . . , n, dim (v0 ), . . . , (vi ), . . . , (vn ) < n 1, hence n1 ([v0 , . . . , vi , . . . , vn ]) = 0. Then, it follows that
K L n1 n ([v0 , . . . , vn ]) = 0 = n n ([v0 , . . . , vn ]).

j <k (3) When dim (v0 ), . . . , (vn ) = n 1, we have 0 that (vj ) = (vk ) and the other (vi ) are distinct, whence dim (v0 ), . . . , (vi ), . . . , (vn ) < n 1 for all i = j, k. Observe that [(v0 ), . . . , (vj ), . . . , (vn )] = (1)k1 [(vk ), (v0 ), . . . , (vj ), . . . , (vk ), . . . , (vn )] = (1)k1 [(vj ), (v0 ), . . . , (vj ), . . . , (vk ), . . . , (vn )] = (1)k+j1[(v0 ), . . . , (vk ), . . . , (vn )]. Then, it follows that
K n1 n ([v0 , . . . , vn ]) = n1 n n i i=0 (1) [v0 , . . . , vi , . . . , vn ]

n such

=
i=0

(1)i n1 ([v0 , . . . , vi , . . . , vn ])

= (1)j [(v0 ), . . . , (vj ), . . . , (vn )] + (1)k [(v0 ), . . . , (vk ), . . . , (vn )] = (1)j [(v0 ), . . . , (vj ), . . . , (vn )] + (1)j1 [(v0 ), . . . , (vj ), . . . , (vn )]
L = 0 = n n ([v0 , . . . , vn ]). L K Thus, n n = n1 n , that is, is a chain homomorphism.

5.6 Denition Let K and L be oriented complexes and : K L a simplicial map. The chain homomorphism : C(K) C(L) is called the chain homomorphism induced by . The homomorphism ( ) is called the homology homomorphism induced by and simply denoted by . In case f = || : |K| |L|, is also denoted by f , which is also called the homology homomorphism induced by f . 52

() ()

Remark. (1) In case K = L (with the same orientation), (id|K| ) = (idK ) = idH (K) . (2) In case L is a subcomplex of K (with the orientation inherited from K), each chain group Cn (L) is a subgroup of Cn (K) and the boundary opL K erator n is the restriction of n , but Hn (L) is not a subgroup of Hn (K) in general as seen in Example 4.11 (K() K()). 5.7 Proposition Let K1 , K2 and K3 be oriented complexes. For simplicial maps : K1 K2 and : K2 K3 , () = , hence () = . In case f = || and g = ||, (gf ) = g f . Proof. We may only prove that () = . To this end, it suces to show that ()n ([v0 , . . . , vn ]) = n n ([v0 , . . . , vn ]) for each [v0 , . . . , vn ] K1 . When all (vi ) are distinct, ()n ([v0 , . . . , vn ]) = [(v0 ), . . . , (vn )]. In this case, since all (vi ) are also distinct, n ([v0 , . . . , vn ]) = [(v0 ), . . . , (vn )]. Note that [(v0 ), . . . , (vn )] K2 or [(v0 ), . . . , (vn )] K2 . Then, it follows that n (n ([v0 , . . . , vn ])) = [(v0 ), . . . , (vn )] = ()n ([v0 , . . . , vn ]). Otherwise, whether (vi ) are distinct or not, we have ()n ([v0 , . . . , vn ]) = 0 = n n ([v0 , . . . , vn ]). Thus, we have () = . 5.8 Corollary Let K and L be oriented complexes. A simplicial isomorphism : K L induces a homology isomorphism : H (K) H (L) and 1 = (1 ) . Hence, a simplicial homeomorphism f : |K| |L| induces 1 a homology isomorphism f and f = (f 1 ) . Proof. In Proposition 5.7, let K1 = K3 = K (with the same orientation), K2 = L and = 1 . Then, we have (1 ) = (idK ) = idH (K) . Similarly, it follows that (1 ) = idH (L) . Therefore, is an isomorphism and 1 = (1 ) . 53

Remark. In case L is the same simplicial complex as K with a dierent orientation from K, we apply the above corollary to idK : K L to show that H (K) is determined independent from an orientation of K. Then, the homology H (K) is dened for a simplicial complex K. By this remark, hereafter we do not take care any orientation of a simplicial complex. In case of necessity, we may consider its orientation. Sometimes, we use the term a simplicial complex instead of an oriented complex. 5.9 Proposition Let : K L be a simplicial map. If K (resp. L) is connected, then : H0 (K) H0 (L) is a monomorphism (resp. an epimorphism).1 Hence, in case both K and L are connected, : H0 (K) H0 (L) is an isomorphism. Proof. Let v K (0) . If K is connected, then H0 (K) = Zv K and (vK ) = L (v) = 0, hence : H0 (K) H0 (L) is a monomorphism. On the other L hand, if L is connected, then H0 (L) = Z(v) , hence : H0 (K) H0 (L) is an epimorphism. 5.10 Proposition Let : K L be a constant simplicial map. Then, (Hn (K)) = 0 for each n > 0 and (H0 (K)) Z. = Proof. For each n > 0, n (Cn (K)) = 0, hence (Hn (K)) = 0. Let L0 be the component of L containing u0 , where (K (0) ) = {v0 }. By Proposition 5.9, (H0 (K)) = H0 (L0 ) Z. = In Denition 3.16, it is dened that two simplicial maps are simplicially homotopic. We shall show that they induce the same homology homomorphism, that is, 5.11 Theorem Let f, g : K L be simplicial maps. If f and g are simplicially homotopic then f = g : H (K) H (L). Proof. Let h : K I L be a simplicial homotopy with h0 = f and h1 = g, where K (0) is given some order so that K is an ordered complex. We dene 0, as follows: for each homomorphisms Dn : Cn (K) Cn+1 (K I), n [v0 , . . . , vn ] K,
n

Dn ([v0 , . . . , vn ]) =
i=0
1

(1)i [(v0 , 0), . . . , (vi , 0), (vi, 1), . . . , (vn , 1)].

A homomorphism is called a monomorphism or an epimorphism according to being injective or surjective.

54

KI K For simplicity, we denote n+1 = n+1 and n = n . Then, observe n

n+1 Dn ([v0 , . . . , vn ]) =
i=0

(1)i
j<i

(1)j [(v0 , 0), . . . , (vj , 0), . . . , (vi , 0), (vi, 1), . . . , (vn , 1)]

+ (1)i [(v0 , 0), . . . , (vi1 , 0), (vi, 1), . . . , (vn , 1)] + (1)i+1 [(v0 , 0), . . . , (vi , 0), (vi+1, 1), . . . , (vn , 1)]
n

+
j>i

(1)j+1 [(v0 , 0), . . . , (vi , 0), (vi, 1), . . . , (vj , 1), . . . , (vn , 1)] ,

and
n

Dn1 n ([v0 , . . . , vn ]) = Dn1

j=0 n

(1)j [v0 , . . . , vj , . . . , vn ]

=
j=0 n

(1)j Dn1 ([v0 , . . . , vj , . . . , vn ]) (1)j

j=0 i<j

(1)i [(v0 , 0), . . . , (vi , 0), (vi , 1), . . . , (vj , 1), . . . , (vn , 1)]

+
i>j

(1)i1 [(v0 , 0), . . . , (vj , 0), . . . , (vi , 0), (vi , 1), . . . , (vn , 1)] .

By summing these equations, we have n+1 Dn + Dn1 n ([v0 , . . . , vn ])

n

=
i=0

[(v0 , 0), . . . , (vi1 , 0), (vi, 1), . . . , (vn , 1)] [(v0 , 0), . . . , (vi , 0), (vi+1 , 1), . . . , (vn , 1)]

= [(v0 , 1), . . . , (vn , 1)] [(v0 , 0), . . . , (vn , 0)]. Therefore, hn n+1 Dn + Dn1 n ([v0 , . . . , vn ]) = hn ([(v0 , 1), . . . , (vn , 1)] [(v0 , 0), . . . , (vn , 0)]) = gn ([v0 , . . . , vn ]) fn ([v0 , . . . , vn ]). 55

For any z Zn (K), since Dn1 n (z) = Dn1 (0) = 0, we have gn (z) fn (z) = hn n+1 Dn (z) hn (Bn (K)) Bn (L), which implies that g (z K ) = gn (z) = fn (z) = f (z K ). 5.12 Corollary For any simplicial complex K and any vertex v0 K (0) , H (St(v0 , K)) is isomorphic to H ({v0 }). In particular, for every simplex , Hn (K()) = 0 for n > 0 and H0 (K()) Z. = Proof. We show that H (St(v0 , K)) H ({v0 }). Let L = St(v, K) with an = order on L(0) such that L is an ordered simplicial complex. We dene : (L I)(0) = L(0) {0, 1} L(0) by (v, 0) = v and (v, 1) = v0 for all v L(0) . Then is clearly a simplicial map. Thus, we have a simplicial homotopy h = || such that h0 = id|L| and f = h1 : |L| |L| is the constant map with h1 (|L|) = {v0 }. Since f = (id|L| ) = idH (L) , we have the result by Proposition 5.10.
L L

56

Mayer-Vietoris Exact Sequences and the Subdivision Operator

G1 G2 Gr+1
f1 f2 fr

6.1 Denition A sequence of groups and homomorphisms

is said to be exact if Im fi = Ker fi+1 for each i = 1, . . . , r. Remark. A homomorphism f : G H is injective (resp. surjective) if and only if f f 0 G H (resp. G H 0) is exact. Recall that a monomorphism is an injective homomorphism and an epimorphism is a surjective homomorphism. If H is a subgroup of G then p 0 H G G/H 0 is exact, where p is the quotient homomorphism. 6.2 Theorem (Mayer-Vietoris) Let K be an oriented complex and K0 , K1 and K2 be subcomplexes of K such that K = K1 K2 and K0 = K1 K2 , let i1 : K0 K1 , i2 : K0 K2 , j1 : K1 K and j2 : K2 K be the inclusions and let n : Hn (K0 ) Hn (K1 ) Hn (K2 ) and n : Hn (K1 ) Hn (K2 ) Hn (K) be homomorphisms dened by n (z K0 ) = ((i1 ) (z K0 ), (i2 ) (z K0 )) = (z K1 , z K2 ), n (z1
K1

()

(1) (2)

, z2

K2

) = (j1 ) (z1

K1

) + (j2 ) (z2

K2

) = z1 + z2 .

Then, there are homomorphisms n : Hn (K) Hn1 (K0 ), n the following sequence is exact: Hn (K0 ) Hn (K1 ) Hn (K2 ) Hn (K)
n 1 Hn1 (K0 ) H1 (K) H0 (K0 )

0 such that

n+1

n1

0 H0 (K1 ) H0 (K2 ) H0 (K) 0.

This sequence is called the Mayer-Vietoris exact sequence for (K; K1 , K2 ). Proof. First, we show Im n = Ker n . For each z Zn (K0 ), n n (z K0 ) = n (z K1 , z K2 ) = z z 57
K

= 0 Hn (K).

Hence, n n = 0, i.e., Im n Ker n . To see Ker n Im n , let (z1 K1 , z2 K2 ) Ker n Hn (K1 ) Hn (K2 ), where z1 Zn (K1 ) and z2 Zn (K2 ). Since z1 + z2
K

= n (z1 K1 , z2 K2 ) = 0 Hn (K),

there is some c Cn+1 (K) such that n+1 (c) = z1 + z2 . We can write c = c1 + c2 , c1 Cn+1 (K1 ), c2 Cn+1 (K2 ). Let z0 = z1 n+1 (c1 ) = n+1 (c2 ) z2 Zn (K1 ) Zn (K2 ) = Zn (K0 ). Then, z0 K0 Hn (K0 ), z0 K1 = z1 K1 Hn (K1 ) and z0 K2 = z2 K2 Hn (K2 ), hence n (z0 K0 ) = (z0 K1 , z0 K2 ) = (z1 K1 , z2 K2 ). Therefore, Ker n = Im n . Next, we will dene the homomorphisms n : Hn (K) Hn1 (K0 ), n 0.

Each z Zn (K) can be written as z = c1 + c2 , ci Cn (Ki ) (i = 1, 2). Since n (z) = 0, it follows that n (c1 ) = n (c2 ) Zn1 (K1 ) Zn1 (K2 ) = Zn1 (K0 ). (Note that Bn1 (K1 ) Bn1 (K2 ) = Bn1 (K0 ) in general.) Assume z K = K z Hn (K) for other z Zn (K) and z = c1 + c2 , ci Cn (Ki ) (i = 1, 2). Then, z z = n+1 (c) for some c Cn+1 (K), where we write c = c1 + c2 , ci Cn+1 (Ki ) (i = 1, 2). Then, it follows that c1 c1 n+1 (c1 ) = c2 c2 + n+1 (c2 ) Cn (K1 ) Cn (K2 ) = Cn (K0 ). Therefore, we have n (c1 ) n (c1 ) = n (c1 c1 ) Bn1 (K0 ), which means that n (c1 ) = n (c1 ) n : Hn (K) Hn1 (K0 ) by
K0 K0

Hn1 (K0 ). Thus, we can dene = n (c2 ) ,

K0

n (z K ) = n (c1 )

K0

(3)

where z = c1 + c2 , ci Cn (Ki ) (i = 1, 2). K To verify that n is a homomorphism, let z K , z Hn (K), where z = c1 + c2 , z = c1 + c2 Zn (K), ci , ci Cn (Ki ) (i = 1, 2). Then, n (z K ) = 58

n (c1 ) and n (z ) = n (c1 ) . Since z z = (c1 c1 ) + (c2 c2 ) and ci ci Cn (Ki ) (i = 1, 2), we have n (z K z ) = n (z z ) = n (c1 c1 ) = n (c1 )
K0 K K K0 K K0

K0

K0

n (c1 )

= n (z K ) n (z ).

We show Im n+1 = Ker n . Let z K Hn+1 (K), where z = c1 + c2 K0 Zn+1 (K), ci Cn+1 (Ki ) (i = 1, 2). Then, n+1 (z K ) = n+1 (c1 ) Hn (K0 ). K1 Since n+1 (c1 ) Bn (K1 ), n+1 (c1 ) = 0 Hn (K1 ). On the other hand, K2 n+1 (c1 ) = n+1 (c2 ) Bn (K2 ), hence n+1 (c1 ) = 0 Hn (K2 ). Thus, we have n n+1 (z K ) = n (n+1 (c1 ) ) = (n+1 (c1 ) , n+1 (c1 ) ) = 0 Hn (K1 ) Hn (K2 ). Therefore, n n+1 = 0, i.e., Im n+1 Ker n . To see Ker n Im n+1 , let z K0 Ker n Hn (K0 ), where z Zn (K0 ). Since (z K1 , z K2 ) = n (z K0 ) = 0 Hn (K1 ) Hn (K2 ), we have c1 Cn+1 (K1 ) and c2 Cn+1 (K2 ) such that z = n+1 (c1 ) = n+1 (c2 ). Let z = c1 c2 Cn+1 (K). Then, n+1 (z ) = n+1 (c1 ) n+1 (c2 ) = 0, which means z Zn+1 (K). Observe that z K0 = n+1 (c1 )
K0 K0 K1 K2

= n+1 (z ) Im n+1 .

Thus, we have Im n+1 = Ker n . It remains to show Im n = Ker n . Let (z1 K1 , z2 K2 ) Hn (K1 ) Hn (K2 ), where z1 Zn (K1 ) and z2 Zn (K2 ). Then, z1 + z2 Zn (K) and n n (z1 K1 , z2 K2 ) = n (z1 + z2 ) = n (z1 )
K K0

= 0 Hn1 (K0 ).

Hence, n n = 0, i.e., Im n Ker n . To see Ker n Im n , let z K Ker n Hn (K), where z = c1 + c2 K0 Zn (K), ci C(Ki ) (i = 1, 2). Since n (c1 ) = n (z K ) = 0 Hn (K0 ), there is some c0 Cn (K0 ) such that n (c1 ) = n (c0 ). Then, z1 = c1 c0 Zn (K1 ) and z2 = c2 + c0 Zn (K2 ). Thus, we have z1 K1 Hn (K1 ) and z2 K2 Hn (K2 ). Since z = z1 + z2 , it follows that K n (z1 K1 , z2 K2 ) = z1 + z2 = z K . Therefore, Im n = Ker n . 59

Remark. In case K0 is connected, 0 is a monomorphism by Proposition 5.9, hence 1 = 0. Then, we have the following exact sequence: Hn (K0 ) Hn (K1 ) Hn (K2 ) Hn (K)
n 2 Hn1 (K0 ) H2 (K) H1 (K0 ) 1 H1 (K1 ) H1 (K2 ) H1 (K) 0.

n+1

n1

6.3 Denition For an oriented n-simplex = [v0 , . . . , vn ], let O() be the set of all ordered (n + 1)-tuples (vi0 , . . . , vin ) of all vertices v0 , . . . , vn of . For each = (vi0 , . . . , vin ) O(), let sgn be dened by 0, . . . , n 1 if is even, i0 , . . . , in sgn = 1 if 0, . . . , n is odd, i0 , . . . , in and let b be the barycenter of the j-face vi0 , . . . , vij of . Then, we have j the oriented n-simplex [b , . . . , b ] of Sd K(). One should note that b is the 0 n n barycenter of for any O(). Let K be an oriented complex. The barycentric subdivision Sd K is considered as an oriented complex with the orientation given by the natural order in Example 2.17 (2). We dene the homomorphisms Sdn : Cn (K) Cn (Sd K) as follows: Sdn ([v0 , . . . , vn ]) =
O()

sgn [b , . . . , b ]. 0 n

We call Sd = (Sdn )n

the subdivision operator.

6.4 Lemma The subdivision operator Sd : C(K) C(Sd K) is a chain Sd homomorphism, that is, n Sdn = Sdn1 n for each n 0, where n = n K K and n = n . Proof. Let = [v0 , . . . , vn ] be an oriented n-simplex of K. Then, n Sdn ([v0 , . . . , vn ]) =
O() n

sgn n ([b , . . . , b ]) 0 n (1)i sgn [b , . . . , b , . . . , b ]. 0 n i

O() i=0

60

For each = (vk0 , . . . , vkn ) O() and i < n, there exists a unique = (vk0 , . . . , vkn ) O() such that = but [b , . . . , b , . . . , b ] = [b , . . . , b , . . . , b ]. 0 n 0 n i i Indeed, vki = vki+1 , vki+1 = vki and vk = vk for sgn . Hence, it follows that n Sdn ([v0 , . . . , vn ]) =
O() i i i For each i = 0, . . . , n, we dene i = [v0 , . . . , vn1 ] by vj = vj for j < i i and vj = vj+1 for j i, that is, i i (v0 , . . . , vn1 ) = (v0 , . . . , vi , . . . , vn ). i i For each = (vk0 , . . . , vkn1 ) O(i ), we denote i i = (vk0 , . . . , vkn1 , vi ) O(). Then, sgn = (1)ni sgn and [b , . . . , b ] = [b , . . . , b ]. Let 0 n1 0 n1

= i, i + 1. Then, sgn =

(1)n sgn [b , . . . , b ]. 0 n1

O(i ) = O(i ) O(). Observe that O() = follows that

n i=0

O(i ) and O(i ) O(j ) = if i = j. Then, it

n i i (1)i Sdn1 ([v0 , . . . , vn1 ]) i=0 n

Sdn1 n ([v0 , . . . , vn ]) = =

(1)i sgn [b , . . . , b ] 0 n1
i=0 O(i ) n

= =
O()

(1)n sgn [b , . . . , b ] 0 n1
i=0 O(i )

(1)n sgn [b , . . . , b ] 0 n1

= n Sdn ([v0 , . . . , vn ]). The proof is completed. Applying Mayer-Vietoris exact sequences, we shall prove the following theorem: 61

6.5 Theorem For every oriented complex K, the subdivision operator Sd : C(K) C(Sd K) induces the homology isomorphism Sd : H (K) H (Sd K), hence H (K) H (Sd K). = To prove this theorem, we need the following lemma: 6.6 Lemma (Five Lemma) Consider the following commutative diagram of groups and homomorphisms: A1 A2 A3 A4 A5 h h h h h
1 2 3 4

f1

f2

f3

f4

B1 B2 B3 B4 B5 .
g1 g2 g3 g4

Suppose that the top and bottom sequences are exact, h2 and h4 are isomorphism, h1 is an epimorphism and h5 is a monomorphism. Then, h3 is an isomorphism. Proof. To see that h3 is surjective, let y B3 . Since h4 is an epimorphism, g3 (y) = h4 (x ) for some x A4 . Then, h5 f4 (x ) = g4 h4 (x ) = g4 g3 (y) = 0. Since h5 is a monomorphism, f4 (x ) = 0, i.e., x Ker f4 = Im f3 , hence x = f3 (x) for some x A3 . Then, g3 (y h3 (x)) = g3 (y) g3 h3 (x) = h4 (x ) h4 f3 (x) = 0, i.e., y h3 (x) Ker g3 , hence y h3 (x) = g2 (y ) for some y B2 . Since h2 is an epimorphism, we have x A2 such that h2 (x ) = y , whence h3 f2 (x ) = g2 h2 (x ) = g2 (y ) = y h3 (x). Therefore h3 (x + f2 (x )) = h3 (x) + h3 f2 (x ) = y. To show that h3 is injective, let x Ker h3 A3 . Since h4 f3 (x) = g3 h3 (x) = 0 and h4 is a monomorphism, f3 (x) = 0, i.e., x Ker f3 = Im f2 , hence x = f2 (x ) for some x A2 . Then, g2 h2 (x ) = h3 f2 (x ) = h3 (x) = 0, i.e., h2 (x ) Ker g2 = Im g1 , hence h2 (x ) = g1 (y ) for some y B1 . Since h1 is an epimorphism, we have x A1 such that h1 (x ) = y , whence h2 f1 (x ) = g1 h1 (x ) = h2 (x ). Since h2 is a monomorphism, f1 (x ) = x , so x = f2 (x ) = f2 f1 (x ) = 0. Proof of Theorem 6.5. The theorem is trivial if dim K = 0 because Sd K = K. Furthermore, the theorem holds when K = K() for some simplex . Indeed, since Sd K() = St(b(), Sd K()), 62

Hn (K()) = Hn (Sd K()) = 0 for all n > 0 by Corollary 5.12, hence Sd : Hn (K()) Hn (Sd K()) is an isomorphism for each n > 0. By Theorem 4.14, H0 (K()) = Zv K() and H0 (Sd K()) = Zv Sd K() , where v is a vertex of . Since Sd0 (v) = v by the denition, Sd : H0 (K()) H0 (Sd K()) is also an isomorphism. Now, we prove the theorem by induction on the number of simplexes of K. In the case K has the only one simplex (hence dim K = 0), it has proved in the above. Assume that the theorem holds for simplicial complexes with m 1 many simplexes and that K has m many simplexes. By the above remark, we may assume that dim K > 0. Let be a maximal dimensional simplex of K, that is, dim = dim K > 0. Then, K1 = K \ {} is a subcomplex of K, K = K1 K() and K() = K1 K(). Moreover, Sd K1 and Sd K() are subcomplexes of Sd K, Sd K = Sd K1 Sd K() and Sd K() = Sd K1 Sd K(). Consider the following diagram:
n+1

Hn (K()) Sd

Hn (K1 ) Hn (K()) Sd Sd

Hn (Sd K()) Hn (Sd K1 ) Hn (Sd K())

n+1 n

Hn (K) Sd

Hn1 (K()) Sd

n1

Hn (Sd K) Hn1 (Sd K()) ,

n n n1

where the top and bottom sequences are Mayer-Vietoris exact sequences, Sd : H (K()) H (Sd K()) and Sd : H (K1 ) H (Sd K1 ) are isomorphisms by the inductive assumption, and as seen in the above, Sd : H (K()) H (Sd K()) is also an isomorphism. It is easy to see that (Sd Sd )n = n Sd and Sd n = n (Sd Sd ). To see Sd n = Sd n , let z K Hn (K), where z Zn (K). Choose c1 Cn (K1 ) so that zc1 Cn (K()), whence Sdn (c1 ) Cn (Sd K1 ) and Sdn (z) 63

Sdn (c1 ) Cn (Sd K()). Then, n (z K ) = n (c1 ) Sd K0 n Sdn (c1 ) . It follows from Lemma 6.4 that Sd n (z K ) = Sd (n (c1 ) ) = Sdn1 n (c1 ) = n Sdn (c1 )
Sd K0 K0

K0

and n (Sdn (z)

Sd K

) =

Sd K0

= n (Sdn (z)

Sd K

) = n Sd (z K ).

Thus, the above diagram commutative. It follows from the Five Lemma that Sd : H (K) H (Sd K) is an isomorphism. In Corollary 5.12, we have shown that for every simplex , Hn (K()) = 0 if n > 0 and H0 (K()) Z. Now, applying Mayer-Vietoris exact sequences, = we can calculate the homology groups of K(). 6.7 Proposition Let be an (k + 1)-simplex (k > 0). Then, Hn (K()) = Z if n = 0, k, 0 otherwise.

Proof. For n > k = dim , Cn (K()) = 0, hence Hn (K()) = 0. If n K() K() k, then Cn (K()) = Cn (K()) and n = n . Hence Zn (K()) = k and Bn (K()) = Bn (K()) for n < k. Therefore Zn (K()) for n Hn (K()) = Hn (K()) for n < k. It remains to show that Hk (K()) = Z in case k = 1 (Example 4.11). Z. We have shown that Hk (K()) = In case k > 1, let be an k-face of . By induction, it suces to show that Hk () Hk1 ( ). Let v be the vertex of with v . Since = K() = K( ) St(v, K()) and K( ) = K( ) St(v, K()), we have the following exact sequence: Hk (K( )) Hk (St(v, K()) Hk (K()) Hk1 (K( )) Hk1 (K( )) Hk1 (St(v, K()). In the above sequence, the both side groups are trivial by Corollary 5.12. Therefore, we have Hk () Hk1( ). = Remark. In the above, we have excluded the case k = 0. In this case, consists of just two points, hence H0 (K()) Z Z and Hn (K()) = 0 = for all n > 0.

64

Topological Invariance of Homology

7.1 Lemma Let K be an oriented complex. If : Sd K K a simplicial approximation of id|K|, then : H (Sd K) H (K) is an isomorphism 1 with = Sd . Proof. Since the subdivision operator Sd : C(K) C(Sd K) induces the homology isomorphism Sd : H (K) H (Sd K) by Theorem 6.5, it suces to show Sd = idH (K) . In fact, we can show Sd = idC(K) . Let n 0 and = [v0 , . . . , vn ] each oriented n-simplex of K. Recall Sdn ([v0 , . . . , vn ]) =
O()

sgn [b , . . . , b ], 0 n

where O() denotes the set of all ordered (n + 1)-tuples = (vi0 , . . . , vin ) of all vertices v0 , . . . , vn , b is the barycenter of the j-face vi0 , . . . , vij of and j 0, . . . , n if 1 is even, i0 , . . . , in sgn = 1 if 0, . . . , n is odd. i ,...,i
0 n

We shall observe that there is a unique 0 = (vi0 , . . . , vin ) O() such that all (b0 ) are distinct. Indeed, if such a 0 exists, then each (b0 ) is a j j vertex of vi0 , . . . , vij because is a simplicial approximation of id|K| . Since all (b0 ) are distinct, we have (b0 ) = vij . In particular, (b()) = vin . Let j j n1 be the (n 1)-face of which does not contain vin . Then, (b(n1 )) = vin1 . Let n2 be the (n 2)-face of n1 which does not contain vin1 . Then (b(n2 )) = vin2 . Thus, vin , vin1 , . . . , vi0 are inductively determined, whence 0 O() is a unique element such that all (b0 ) are distinct. j Now, it should be noticed that (b0 ) = vij for each j = 0, . . . , n and j = 0 implies n ([b , . . . , b ]) = 0. Then, it follows that 0 n n Sdn () = n Sdn ([v0 , . . . , vn ]) =
O()

sgn n ([b , . . . , b ]) 0 n

= sgn 0 [vi0 , . . . , vin ] = [v0 , . . . , vn ] = . Therefore, n Sdn = idCn (K) for every n 0. 7.2 Lemma Let f : |K| |L| be a continuous map and : | Sdn K| |L| and : | Sdm K| |L| simplicial approximations of f . Then, Sdn = Sdm : H (K) H (L). 65

Proof. We may assume that m > n. For each i > 0, let i : | Sdi K| | Sdi1 K| be a simplicial approximation of id|K| . Then, by Proposition 3.13,
m n = n+1 m1 m : | Sdm K| | Sdn K| m and n : | Sdm K| |L| are also simplicial approximations of id|K| and f , m respectively. It follows from Theorem 3.17 that n is simplicially homotopic to . By Theorem 5.11, we have m m (n ) = (n ) = : H (Sdm K) H (L). m Since (n ) Sdm = Sdn by Lemma 7.1, it follows that m Sdn = (n ) Sdm = Sdm : H (K) H (L).

H (K) R

m6 mmm mmm mmm mmm

Sdn

H (Sdn K)
S

RRR RRR RRR Sdm RRR(

Sdmn

m (n )

QQQ QQQ QQQ QQQ Q( m6 mmm mmm mmm mmm

H (L)

H (Sdm K)

7.3 Denition Let f : |K| |L| be a continuous map. We dene the homology homomorphism f : H (K) H (L) induced by f as g Sdn : H (K) H (L) (i.e., f = g Sdn ), where g : | Sdn K| |L| is a simplicial approximation of f . Remark. By Lemma 7.2, the homomorphism f is determined independently from a simplicial approximation of f . Recall that (id|K| ) = idH (K) . 7.4 Theorem Let K1 , K2 and K3 be oriented complexes, f : |K1 | |K2 | and g : |K2 | |K3 | continuous maps. Then, (gf ) = g f : H (K1 ) H (K3 ). Proof. Let : Sdn K2 K3 be a simplicial approximation of g. Then, g = Sdn . Let : Sdn K2 K2 and : Sdm K1 Sdn K2 simplicial approxi mations of id|K2 | and f , respectively. Since : Sdm K2 K2 is a simplicial approximation of f by Proposition 3.13, we have f = () Sdm = Sdm . On the other hand, : Sdm K1 K3 is a simplicial approximation gf by Proposition 3.13. Then, we have (gf ) = () Sdm = Sdm = Sdn Sdm = g f . 66

()

Using this theorem, we can show that H (K) is a topological invariant of |K|, that is, 7.5 Theorem (Topological Invariance of Homology) Let K and L be oriented complexes. Then, every homeomorphism f : |K| |L| induces a homology isomorphism f : H (K) H (L) with (f )1 = (f 1 ) . Hence, |K| |L| implies H (K) H (L). = Proof. By the above theorem, (f 1 ) f = (id|K| ) = idH (K) and f (f 1 ) = (id|L| ) = idH (L) , hence (f )1 = (f 1 ) . 7.6 Denition For a (topological) polyhedron X, given a triangulation (KX , hX ), we dene the homology H (X) = H (KX ), where H (KX ) is determind independent from an orientation of KX as observed in Remark for Corollary 5.8. Let Y be another polyhedron with a triangulation (KY , hY ). For a continuous map f : X Y , we dene f = (h1 f hX ) : H (X) = H (KX ) H (KY ) = H (Y ) Y which is called the homology homomorphism induced by f . Remark. In the above, for another triangulation (KX , hX ) of X, identifying H (KX ) = H (KX ) by the isomorphism (h1 hX ) , we can dene the X homology H (X) independently from a triangulation of X. Moreover, for another triangulation (KY , hY ) of Y , H (KY ) is identifying with H (KY ) by the isomorphism (h1 hY ) . Then, by identifying (hY 1 f hX ) : H (K ) Y H (L ) with (h1 f hX ) : H (K) H (L), the homology homomorphism Y f : H (X) H (Y ) can be dened independently from triangulations of X and Y . The following follows from Theorem 4.14. 7.7 Theorem A (topological) polyhedron X is connected if and only if H0 (X) Z. = For each n > 0, it follows from Corollary 5.12 and Proposition 6.7 that Hi (Bn ) = Z if i = 0, 0 if i > 0, and Hi (Sn ) = Z if i = 0, n, 0 otherwise.
()

()

By comparison between the homology groups, n = m implies Sn Sm . By using this observation, we can show the topological invariance of dimension of Euclidean spaces, that is, 67

7.8 Theorem (Topological Invariance of Dimension) If n = m N then Rn Rm , equivalently Rn Rm implies n = m. Proof. If Rn Rm , then Sn Rn {} Rm {} Sm , where Rn {} and Rm {} are the one-point compactications of Rn and Rm , respectively. Therefore, n = m by the above observation. Applying the homology, we can also show the following: 7.9 Theorem (No Retraction Theorem) For each n N, there exist no continuous map r : Bn Sn1 such that r|Sn1 = idSn1 . Such a map r is called a retraction. Proof. When n = 1, B1 is connected but S0 is disconnected, hence S0 is not the image of any continuous map of B1 . Then, we have the result. In case n > 1 (i.e., n 1 > 0), assume that there exists a retraction r : Bn Sn1 . Let i : Sn1 Bn be the inclusion. Since ri = idSn1 , it follows that r i = (ri) = idH (Sn1 ) . Hence, r : Hn1 (Bn ) Hn1 (Sn1 ) is surjective. However, Hn1 (Bn ) = 0 and Hn1 (Sn1 ) Z. This is a = contradiction. Applying this No Retraction Theorem, we can show the following xed point theorem: 7.10 Theorem (Brouwers Fixed Point Theorem) For each n N, every continuous map f : Bn Bn has a xed point x0 Bn , that is, f (x0 ) = x0 . Proof. Suppose that there exists a continuous map f : Bn Bn such that f (x) = x for any x Bn . Then, we can dene a continuous map r : Bn Sn1 as follows: r(x) = (1 s(x))f (x) + s(x)x = f (x) + s(x)(x f (x)), where s : Bn [0, ) is a map satisfying r(x) = 1, that is, s(x) is the non-negative root of the following equation: x f (x) 2 t2 + 2 f (x), x f (x) t + f (x)
2

1 = 0,

68

where x, y is the inner product of x, y Rn . Actually, s(x) = 1 x f (x)

2

f (x), x f (x) + f (x), x f (x)

2

x f (x) 2 ( f (x)

1) .

As is easily observed, r(x) = x for each x Sn1 . This contradicts to No Retraction Theorem 7.9.
r(x) x f (x) Bn Sn1

We can apply No Retraction Theorem to prove the following theorem mentioned in 1: 7.11 Theorem (Invariance of Domain) For each X, Y Rk , X Y implies int X int Y . We use also the following extension theorem: Tietze Extension Theorem: Let A be a closed subset of a space X ( Rk ). Every continuous map f : A I has a continuous extension f : X I.
Note. The above is valid for every normal space X.

7.1 Exercise Let A be a closed subset of a space X ( Rk ). Applying Tietze Extension Theorem coordinate-wise, show that every continuous map f : A In extends to a continuous map f : X In . For each x Rk and > 0, we denote B(x, ) = {y Rk | x y < }, B(x, ) = {y Rk | x y Then, (B(x, ), bd B(x, )) (Bk , Sk1 ). 69 }.

7.12 Lemma Let X be a closed subset of Rk and x X. Then, x bd X if and only if each neighborhood U of x in X contains a neighborhood V of x such that every continuous map f : X \ V Sk1 extends to a continuous map f : X Sk1 . Proof. To see the only if part, for each neighborhood U of x in X, choose > 0 so that B(x, ) X U. Let V = B(x, ) X. Since Bk Ik , every continuous map f : X \ V Sk1 can be extended to a continuous map g : X Bk by Tietze Extension Theorem (cf. Exercise 7.1). Then, we can nd 0 < < so that g(X \ B(x, )) Bk \ {0}. Let r : Bk \ {0} Sk1 be the canonical radial retraction (i.e., r(y) = y 1y) Since x bd X, we have z B(x, ) \ X. Let = + x z > 0. Observe that B(x, ) B(z, ) B(x, ). Let h : X X \ B(z, ) X \ B(x, ) be the continuous map dened by h|X \ B(z, ) = id and h(y) = z + (y z) for y X B(z, ). yz

Then, rgh : X Sk1 is a continuous extension of f . To see the if part, assume that x int X. Then B(x, ) X for some > 0. By the condition, B(x, ) contains a neighborhood V of x such that every continuous map f : X \ V Sk1 extends to a continuous map f : X Sk1 . It is easy to construct a retraction r : Rk \ {x} bd B(x, ) Sk1 . Then, r|X \ V extend to a retraction r : X bd B(x, ). Since B(x, ) X, bd B(x, ) is a retract of B(x, ), which contradicts to No Retraction Theorem 7.9 because (B(x, ), bd B(x, )) (Bk , Sk1 ). Therefore, x bd X. Proof of Theorem 7.11. Let h : X Y be a homeomorphism. For each x bd X and each neighborhood U of h(x) in Y , h1 (U) is a neighborhood of x in X, which contains a neighborhood V of x such that every continuous map f : X \ V Sk1 extends to a continuous map f : X Sk1 . Then, h(V ) is a neighborhood of h(x) in Y such that h(V ) U, and every continuous map g : Y \ h(V ) Sk1 extends to a continuous map g : Y Sk1 . Indeed, k1 : X Sk1 , whence gh : X \ V S extends to a continuous map f f h1 : Y Sk1 is a continuous extension of g. The following is the topological version of Theorem 6.2. 70

7.13 Theorem (Mayer-Vietoris) Let X be a (topological) polyhedron and X0 , X1 and X2 be subpolyhedra of X such that X = X1 X2 and X0 = X1 X2 , let i1 : X0 X1 , i2 : X0 X2 , j1 : X1 X and j2 : X2 X be the inclusions and let n : Hn (X0 ) Hn (X1 ) Hn (X2 ) and n : Hn (X1 ) Hn (X2 ) Hn (X) be homomorphisms dened by n () = ((i1 ) (), (i2 ) ()), n (1 , 2 ) = (j1 ) (1 ) + (j2 ) (2 ). Then, there are homomorphisms n : Hn (X) Hn1 (X0 ), n the following sequence is exact: Hn (X0 ) Hn (X1 ) Hn (X2 ) Hn (X)
n 1 Hn1 (X0 ) H1 (X) H0 (X0 ) 0 H0 (X1 ) H0 (X2 ) H0 (X) 0.

0 such that

n+1

n1 0

71

Homotopical Invariance of Homology and Homologies of Surfaces

8.1 Denition Let f, g : X Y be continuous maps. It is said that f is homotopic to g (denoted by f g) if there exists a continuous map h : X I Y such that h0 = f and h1 = g, where ht : X Y (t I) are dened by ht (x) = h(x, t) for all x X. Here, we call h a homotopy from f to g and denote h : f g. A map f : X Y is said to be null-homotopic (denoted by f 0) if f is homotopic to a constant map. 8.2 Proposition Let C(X, Y ) be the set of all continuous maps from X to Y. (1) The relation is a equivalence relation on C(X, Y ). f and g g imply

(2) For each f, f C(X, Y ) and g, g C(Y, Z), f gf g f .

Proof. (1) Let f, f , f C(X, Y ). Then, f f by the homotopy dened f by the homotopy dened by h (x, t) = by ht = f . If h : f f , then f h(x, 1 t). If h : f f and h : f f , then f f by the homotopy dened by h (x, 2t) if 0 t 1 , 2 h(x, t) = h (x, 2t 1) if 1 t 1. 2 (2) Let h : f f and h : g dened by ht = ht ht . g . Then, gf g f by the homotopy

8.3 Denition It is said that X is homotopically equivalent to Y , or X has the homotopy type of Y , or X and Y have the same homotopy type (denoted by X Y ) if there exist continuous maps f : X Y and g : Y X such that gf idX and f g idY , whence f is called a homotopy equivalence and g a homotopy inverse of f . A space X is contractible if X has the homotopy type of a one point space, that is, the identity idX is null-homotopic. Remark. A homeomorphism f : X Y is a homotopy equivalence with f a homotopy inverse. Hence, X Y implies X Y , namely the homotopy type of X is a topological invariant of X. Every convex set is contractible, so every simplex is contractible. 8.1 Exercise The relation spaces. is an equivalence relation among
1

()

72

8.2 Exercise Let f : X Y be a continuous map. If X or Y is contractible then f is null-homotopic. 8.4 Denition Let L be a simplicial complex and f, g : X |L| continuous maps. It is said that f is contiguous to g (with respect to L) if each f (x) and g(x) are contained in the same simplex of L. Remark. By Denition 3.10, every simplicial approximation of a continuous map f : |K| |L| is contiguous to f , and any two simplicial approximations of f are contiguous to each others. 8.5 Proposition Let L be a simplicial complex and f, g : X |L| continuous maps. If f is contiguous to g, then f is homotopic to g. Proof. For each x X, f (x) and g(x) are contained in the same simplex of L. Then, we can dene the homotopy h : f g by dened by h(x, t) = (1 t)f (x) + tg(x). 8.6 Corollary Let K and L be simplicial complexes. For each continuous map f : |K| |L|, every simplicial approximation of f is homotopic to f , so any two simplicial approximations of f is homotopic to each others. 8.7 Theorem Let K and L be simplicial complexes. For continuous maps f, g : |K| |L|, if f g then f = g : H (K) H (L). Proof. Let h : |K| I |L| be a homotopy with h0 = f and h1 = g. Since |K| I is compact, there is a Lebesgue number > 0 for the open cover h1 (OL (u)) u L(0) of |K| I. Choose n N so that 1/n < . Then, each {x} [(i 1)/n, i/n] is contained in some h1 (OL(u)), i.e., h({x} [(i 1)/n, i/n]) OL (u). Since h(i1)/n and hi/n have a common simplicial approximation by Proposition 3.15, they induce the same homology homomorphism, hence f = (h0 ) = (h1/n ) = = (h1 ) = g . Using Theorems 7.4 and 8.7, we can strengthen Theorem 7.5 as follows: 8.8 Theorem (Homotopical Invariance of Homology) Let K and L be simplicial complexes. If f : |K| |L| is a homotopy equivalence with a homotopy inverse g : |L| |K|, then f : H (K) H (L) is an isomorphism with (f )1 = g . Hence, |K| |L| implies H (K) H (L). = 73

Proof. Since g f = (gf ) = (id|K| ) = idH (K) and f g = (f g) = (id|L| ) = idH (L) , we have the result. By Euler-Poincar Formula 4.13, we have the following: e
K L 8.9 Corollary For simplicial complexes K and L, let n , n be the numbers of n-simplexes of K and L, respectively. Then, |K| |L| implies dim K i=0 dim L

(1)

K i

=
i=0

L (1)i i .

The following is the topological version of Theorem 8.8. 8.10 Theorem (Homotopical Invariance of Homology) Let X and Y be (topological) polyhedra. If f : X Y is a homotopy equivalence with a homotopy inverse g : Y X, then f : H (X) H (Y ) is an isomorphism with (f )1 = g . Hence, X Y implies H (X) H (Y ). = In case Y is a singleton, we have the following: 8.11 Corollary For every contractible polyhedron X, Hn (X) = Z if n = 0, 0 otherwise.

Homologies of Surfaces
In the rest of this section, we calculate the homologies of surfaces. Mbius band M2 : Let i : S1 M2 be the natural embedding of the o circle S1 to the middle circle of Mbius band M2 . There is a continuous map o id 2 . Then, i is a homotopy r : M2 S1 such that ri = idS1 and ir 1 2 equivalence, hence i : H (S ) H (M ) is an isomorphism. Consequently, Hn (M2 ) Hn (S1 ) = = Z if n = 0, 1, 0 otherwise.

Let j : S1 M2 be the natural embedding of the circle S1 to the boundary circle M2 of M2 . Then, rj : S1 S1 is the map winding S1 round itself 2 times, hence (rj) () = 2 for each H1 (S1 ) Z. =

74

Using the homology of Mbius band M2 and the Mayer-Vietoris exact o sequences, we can calculate the homologies of the projective plane P2 and Klein bottle K2 as follows: Projective plane P2 : First, note Hn (P2 ) = 0 for all n > 2 because dim P2 = 2. Since P2 is connected, H0 (P2 ) = Z. We can consider P2 as the space jointed Mbius band M2 and the disk B2 along the boundary circles o 1 M2 and B2 , that is, P2 = M2 B2 (cf. Remark (1) for Example 2.11). Let S1 = M2 = B2 . Then, we have the following Mayer-Vietoris exact sequence: 0 = H2 (M2 ) H2 (B2 ) H2 (P2 ) 2
epi 0 2

H1 (S1 ) H1 (M2 ) H1 (B2 ) H1 (P2 )

mono epi

(2,0)

Z0=Z H0 (S1 ) H0 (M2 ) H0 (B2 ) 1

0 mono 0

(1,1)

Z Z.

In the above, 0 is a monomorphism by the denition of 0 and Proposition 5.9. From the denition of 1 and the observation in calculation of the homology of Mbius band, it follows that 1 is a monomorphism. Thus, we o 2 have H2 (P ) = 0, and H1 (P2 ) Z/2Z = Z2 by Fundamental Homomorphism = Theorem of Group Theory. Consequently, Z if n = 0, 2 Hn (P ) = Z2 if n = 1, 0 otherwise.

75

Klein bottle K2 : Since dim K2 , Hn (K2 ) = 0 for all n > 2. By connectedness of K2 , H0 (K2 ) = Z. We can consider K2 as the space jointed two Mbius band M2 and M2 along the boundary circles M2 and M2 , that is, o 1 2 1 2 2 K = M2 M2 (cf. Remark (2) for Example 2.11). Let S1 = M2 = M2 . 1 2 1 2 Then, we have the following Mayer-Vietoris exact sequence: 2 0 = H2 (M2 ) H2 (M2 ) H2 (K2 ) 1 1
epi 0 2

H1 (S1 ) H1 (M2 ) H1 (M2 ) H1 (K2 ) 1 2

mono epi

(2,2)

ZZ H0 (S1 ) H0 (M2 ) H0 (M2 ) 1 1 2

0 mono 0

(1,1)

Z Z.

In the above, 0 is a monomorphism by the denition of 0 and Proposition 5.9. From the denition of 1 and the observation in calculation of the homology of Mbius band, it follows that 1 is a monomorphism. Thus, o 2 we have H2 (K ) = 0 and H1 (K2 ) Z2 Z by Homomorphism Theorem.) = Consequently, Z if n = 0, 2 Hn (K ) = Z2 Z if n = 1, 0 otherwise.
)

The group G = Z Z is generated by (1, 0) and (0, 1). Let H be the subgroup of G generated by (2, 2). We dene an isomorphism h : G G by h(1, 0) = (1, 1) and h(0, 1) = (0, 1). Since h(1, 1) = (1, 0), h(H) = 2Z 0 G = Z Z, whence G/H G/h(H) = Z/2Z Z = Z2 Z. =

76

Punctured torus T2 : Let Sl and Sm be a latitude circle and a meridian p circle of T2 . Then, Sl and Sm meet each other at a single point v. Take a disk D T2 missing Sl Sm and let C = D be its boundary circle. We call Tp = T2 \ (D \ C) a punctured torus. Then, T2 = Tp D and Tp D = C.

Sm v S

C D S r C

v Sm D v

v Sm

Let i : Sl Sm T2 be the inclusion. We have a map r : Tp Sl Sm such that ri = r|Sl Sm = id and ir idp . Then, i is a homotopy equivalence, hence i : H (Sl Sm ) H (Tp ) is an isomorphism by the Homotopical Invariance of Homology. By using the Mayer-Vietoris exact sequence, it is easy to see that Hn (Sl Sm ) Hn (Sl ) Hn (Sm ). Consequently, = Z if n = 0, Hn (Sl Sm ) Z Z if n = 1, Hn (Tp ) = = 0 otherwise. Observe that r|C : C Sl Sm is a map winding C round Sl and Sm and then conversely Sl and Sm , hence the homomorphism (r|C) : H1 (C) H1 (Sl Sm ) H1 (Sl ) H1 (Sm ) is the null homomorphism. =

77

Using the homology of Tp and the Mayer-Vietoris exact sequences, we can calculate the homologies of the torus T2 as follows: Torus T2 : First, note Hn (T2 ) = 0 for all n > 2 because dim T2 . Consider the following Mayer-Vietoris exact sequences: 2 0 = H2 (Tp ) H2 (D) H2 (T2 )
0 iso 2

H1 (C) H1 (Tp ) H1 (D) H1 (T2 )

0 iso

(0,0)

(Z Z) 0 H0 (C) H0 (Tp ) H0 (D) 1

0 mono 0

(1,1)

Z Z,

where 0 is a monomorphism by the denition of 0 and Proposition 5.9. Then, H2 (T2 ) H1 (C) Z and H1 (T2 ) H1 (Tp ) Z Z. Consequently, = = = = Z Z Z 2 Hn (T ) = Z 0 if n = 0, if n = 1, if n = 2, otherwise.

78

Relative Homology for Polyhedral Pairs

9.1 Denition A pair (K, L) of oriented complexes means a pair of an oriented complex K and its subcomplex L with the orientation inherited from K. For such a pair (K, L), the n-dimensional (oriented) relative chain group Cn (K, L) is dened as the factor group Cn (K)/Cn (L). The relative (K,L) boundary operator n : Cn (K, L) Cn1 (K, L) is the homomorphism K induced by the boundary operator n : Cn (K) Cn1 (K). The (oriented) (K,L) relative chain complex is the system C(K, L) = (Cn (K, L), n )n 0 :
n C(K, L) : Cn (K, L) Cn1 (K, L)

n+1

(K,L)

(K,L)

n1

(K,L)

C1 (K, L) C0 (K, L) 0

(K,L)

(K,L)

(K,L)

The sequence q = (qn )n 0 of the quotient homomorphisms qn : Cn (K) Cn (K)/Cn (L) = Cn (K, L) is a chain homomorphism from C(K) to C(K, L), that is, the following diagram is commutative: Cn (K) Cn (K, L) (K,L) K
n n qn

Cn1 (K) Cn1 (K, L)

qn1

Moreover, we have the following short exact sequence: 0 Cn (L) Cn (K) Cn (K, L) 0, n where i : L K is the inclusion simplicial map. When L = , we can regard C(K, ) = C(K). Remark. (1) Let Cn (K, L) be the subgroup of the n-chain group Cn (K) generated by K[n] \ L, all oriented n-simplexes in K \ L. Then, as is easily observed, the relative n-chain group Cn (K, L) is naturally isomorphic to (K,L) K Cn (K, L). But it should be noticed that n = n |Cn (K, L). (2) For two simplexes , K, the relative incidence number relative with respect to L is dened as follows: 1 if , dim = dim 1, L and is inherited the orientation from , [; ]L = 1 if , dim = dim 1, L and is not inherited the orientation from , 0 otherwise. 79
i qn

We dene n : Cn (K, L) Cn1 (K, L) as follows: n (c) =

K[n1]\L K[n]\L

a [; ]L for each c =
K[n]\L (K,L)

a .

Then, we have n

= n by identifying Cn (K, L) = Cn (K, L) as in (1).

9.1 Exercise Prove the remark (1) and (2) above. 9.2 Denition For a pair (K, L) of oriented complexes, let
(K,L) and Bn (K, L) = Im n+1 . Zn (K, L) = Ker n (K,L)

The following factor group is called the n-dimensional relative homology group: Hn (K, L) = Zn (K, L)/Bn (K, L). We denote H (K, L) = (Hn (K, L))n 0 . For z Zn (K, L), we denote z (K,L) = z + Bn (K, L) Hn (K, L), which is the relative homology class of z. When L = , we can regard H (K, ) = H (K). By the remark for Denition 9.1, we have the following: 9.3 Theorem (Excision Theorem) Let K be an oriented complex with a subcomplex L. If K0 is a subcomplex of K such that K = L K0 (i.e., K \ L K0 ), then C(K0 , L K0 ) C(K, L) by the natural isomorphism, = hence H (K0 , L K0 ) H (K, L). = The following shows the relation between H (K), H (L) and H (K, L). 9.4 Theorem For a pair (K, L) of oriented complexes, there exist homomorphisms : Hn (K, L) Hn1 (L), n > 0, such that the following sequence is exact:
Hn (L) Hn (K) Hn (K, L) Hn1 (L) H0 (K) H0 (K, L) 0,

where i is the homomorphism induced by the inclusion i : L K and q is the homomorphism induced by the quotient chain homomorphism q = (qn )n 0 : C(K) C(K, L). The above sequence is called the homology (long) exact sequence for (K, L). 80

Proof. See the following commutative diagram: 0 Cn+1 (L) Cn+1 (K) Cn+1 (K, L) 0 (K,L) L K
n+1 n+1

in+1

qn+1

n+1

n Cn (K) Cn (K, L) 0 0 Cn (L) (K,L) L K

n n n

qn

0 Cn1 (L) Cn1 (K) Cn1 (K, L) 0 (K,L) L K

n1 n1

in1

qn1

n1

0 Cn2 (L) Cn2 (K) Cn2 (K, L) 0 For each z Zn (K, L), there is c Cn (K) such that qn (c) = z because (K,L) K Zn (K, L) Cn (K, L) = Cn (K)/Cn (L). Since qn1 n (c) = n qn (c) = (K,L) K n (z) = 0, it follows that n (c) Ker qn1 = Im in1 . Thus, we have K d Cn1 (L) such that n (c) = in1 (d), whence
L K K K in2 n1 (d) = n1 in1 (d) = n1 n (c) = 0. L Since in2 is a monomorphism, it follows that n1 (d) = 0, which implies that d Zn1 (L). L Now, we show that the homology class d of d is uniquely determined for (K,L) the homology class z (K,L) of z. Let z Zn (K, L) such that z (K,L) = z , i.e., zz Bn (K, L). As same as above, choose c Cn (K) and d Zn1 (L) K so that qn (c ) = z and n (c ) = in1 (d ). On the other hand, we have (K,L) y Cn+1 (K, L) such that n+1 (y) = z z , whence there is x Cn+1 (K) such that qn+1 (x) = y. Observe that K qn (c c n+1 (x)) = qn (c c ) n+1 (y) = 0. (K,L) K Since Ker qn = Im in , we have b Cn (L) such that in (b) = c c n+1 (x). Then, it follows that L K K K K in1 n (b) = n (c c n+1 (x)) = n (c) n (c ) = in1 (d) in1 (d ) = in1 (d d ) L Since in2 is a monomorphism, it follows that d d = n (b) Bn1 (L). L L Therefore, d = d . L Consequently, we can dene : Hn (K, L) Hn1 (L) by (z (K,L) ) = d , where z Zn (K, L) and d Zn1 (L) such that z = qn (c) and in1 (d) =

in2

qn2

81

K n (c) for some c Cn (K). To see that is a homomorphism, let z, z Zn (K, L) and d, d Zn1 (L) such that z = qn (c), z = qn (c ), in1 (d) = K K n (c) and in1 (d ) = n (c ) for some c, c Cn (K). Observe that z z = K qn (c c ) and in1 (d d ) = n (c c ). Then,

(z (K,L) z

(K,L)

) = (z z

(K,L)

)=dd

=d d ,

which means that is a homomorphism. The proof of the exactness is left to the reader. 9.2 Exercise Complete the proof above by proving the exactness of the homology sequence. 9.5 Denition Let (K, L) and (M, N) be pairs of oriented complexes, and h = (hn )n 0 : C(K) C(M) a chain homomorphism such that hn (Cn (L)) Cn (N ) for every n 0. Each homomorphism hn : Cn (K) Cn (M) induces the homomorphism hn : Cn (K, L) Cn (M, N): n 0 Cn (L) Cn (K) Cn (K, L) 0 h h |C (L)
n n n

qn

hn

0 Cn (N) Cn (M) Cn (M, N) 0,

in qn

whence n

(M,N )

hn = hn1 n

(K,L)

:
h

Cn (K, L) n (K,L)
hn1

Cn (M, N) (M,N )
n

Cn1 (K, L) Cn1 (M, N), that is, h induces the relative chain homomorphism h = (hn )n
0
()

: C(K, L) C(M, N),

and the relative homology homomorphism h : H (K, L) H (M, N). Then, the following diagram is commutative: Hn (K) Hn (K, L) Hn1 (L) h (h|L) h

()

Hn (M) Hn (M, N) Hn1 (N)

q

82

9.3 Exercise Show that the above diagram is commutative. Remark. (1) By Proposition 5.7, the following diagram is also commutative: Hn (L) Hn (K) h (h|L)

Hn (N) Hn (M)
i

(2) The isomorphism H (K0 , LK0 ) H (K, L) in the Excision Theorem = 9.3 is induced by the inclusion i : (K0 , L K0 ) (K, L). 9.4 Exercise Prove the remark (2) above. 9.6 Denition The relative subdivision operator Sd = (Sdn )n
0

: C(K, L) C(Sd K, Sd L),

is the chain homomorphism induced by the subdivision operator Sd = (Sdn )n

0

: C(K) C(Sd K).

This relative subdivision operator induces the relative homology isomorphism Sd : H (K, L) H (Sd K, Sd L). 9.5 Exercise By using Five Lemma 6.6, show that the relative homology homomorphism induced by the relative subdivision operator is an isomorphism. 9.7 Denition When a simplicial map : K M satises f (L) N (i.e., () N for every L), we write : (K, L) (M, N). In this case, induces the relative chain homomorphism = (n )n 0 : C(K, L) C(M, N), which induces homomorphism : H (K, L) H (M, N). Remark. Let (K, L) be a pair of oriented complexes. Then, (1) (id(K,L) ) = idH (K,L) , and (2) for the inclusion j : (K, ) (K, L), j = q in Denition 9.1, hence j = g : H (K) H (K, L). 9.8 Proposition Let (Ki , Li ), i = 1, 2, 3, be pairs of oriented complexes. Then, for simplicial maps : (K1 , L1 ) (K2 , L2 ) and : (K2 , L2 ) (K3 , L3 ), = () : H (K1 , L1 ) H (K3 , L3 ). 83

9.6 Exercise Prove the proposition above. 9.9 Lemma Let f : (|K|, |L|) (|M |, |N |) be a continuous map. Then, f : |K| |M | has a simplicial approximation g : Sdn K M such that g(Sdn L) N (i.e., g : (Sdn K, Sdn L) (M, N)), whence g| Sdn L is a simplicial approximation of f ||L| : |L| |N |. 9.7 Exercise Prove the lemma above. 9.10 Denition A simplicial map g : (Sdn K, Sdn L) (M, N) as in the above lemma a relative simplicial approximation of f : (|K|, |L|) (|M |, |N |). 9.11 Lemma Let f : (|K|, |L|) (|M |, |N |) be a continuous map. If g : (Sdn K, Sdn L) (M, N) and g : (Sdm K, Sdm L) (M, N) are relative simplicial approximations of f , then g Sdn = g Sdm : H (K, L) H (M, N). 9.8 Exercise Prove the lemma above. 9.12 Denition For a continuous map f : (|K|, |L|) (|M |, |N |), by using its relative simplicial approximation g : (Sdn K, Sdn L) (M, N), we dene the relative homology homomorphism f = g Sdn : H (K, L) H (M, N), which is called the relative homology homomorphism induced by f . Remark. For each pair (K, L) of oriented complexes, (id(K,L) ) = idH (K,L) . 9.13 Theorem Let (Ki , Li ), i = 1, 2, 3, be pairs of oriented complexes. For any continuous maps f : (|K1 |, |L1 |) (|K2 |, |L2 |) and g : (|K2 |, |L2 |) (|K3 |, |L3 |), g f = (f g) : H (K1 , L1 ) H (K3 , L3 ). 9.9 Exercise Prove the theorem above. 9.14 Corollary A homeomorphism f : (|K|, |L|) (|M |, |N |) induces to a relative homology isomorphism f : H (K, L) H (M, N). Hence, (|K|, |L|) (|M |, |N |) implies H (K, L) H (M, N). =

84

9.15 Denition For a polyhedral pair (X, A) with a relative triangulation (K, L, h), we dene the relative homology H (X, A) = H (K, L). Let (Y, B) be another polyhedral pair with a relative triangulation (M, N, g). For a continuous map f : (X, A) (Y, B), we dene f = (g 1f h) : H (X, A) = H (K, L) H (M, N) = H (Y, B), which is called the relative homology homomorphism induced by f . Remark. In the above, for other relative triangulation (K , L , h ) of (X, A), identifying H (K , L ) = H (K, L) by the isomorphism (h1 h ) , we can dene the relative homology H (X, A) independently from a relative triangulation of (X, A). Moreover, for another relative triangulation (M , N , g ) of (Y, B), H (M , N ) is identifying with H (M, N) by the isomorphism (g 1 g ) . Then, by identifying (g 1 f h ) : H (K , L ) H (M , N ) with (g 1 f h) : H (K, L) H (M, N), the relative homology homomorphism f : H (X, A) H (Y, B) can be dened independently from relative triangulations of (X, A) and (Y, B). The following is the topological version of Theorem 9.3: 9.16 Theorem (Excision Theorem) Let (X, A) be a polyhedral pair. If X0 is a subpolyhedron of X such that X = A X0 , then the inclusion i : (X0 , A X0 ) (X, A) induces to the isomorphism i : H (X0 , A X0 ) H (X, A). By the remark after Denition 9.7, we have the following topological version of Theorem 9.4: 9.17 Theorem For a polyhedral pair (X, A), there exist homomorphisms : Hn (X, A) Hn1 (A), n > 0, such that the following sequence is exact:
Hn (A) Hn (X) Hn (X, A) Hn1 (A) H0 (X) H0 (X, A) 0,

()

where i is the homomorphism induced by the inclusion i : A X and j is the homomorphism induced by the inclusion j : (X, ) (X, A). The above sequence is called the homology (long) exact sequence for (X, A).

85

Index
anely independent, 4 barycenter of a simplex, 7 barycentric coordinate, 5 barycentric subdivision, 22 Betti number, n-dimensional, 45 boundary of a simplex, 7 boundary operator, 39 boundary, bounding cycle, 40 Brouwers Fixed Point Theorem, 68 Brouwers Theorem on Invariance of Domain, 11, 69 carrier, 13 chain, 39 chain complex, 42 chain complex, ordered, 45 oriented, 40 chain group, ordered, 44 oriented, 39 chain homomorphism, 50 chain homomorphism induced by, a simplicial map, 52 chain isomorphism, 50 combinatorially equivalent, 31 component, 49 contiguous, 73 contractible, 72 convex body, 10 convex hull, 4 convex set, 4 cross cap, 20 cycle, 40 derived subdivision, 22 dimension, 5, 12 epimorphism, 57 Euler-Poincar characteristic, 45 e Euler-Poincar formula, 46 e exact sequence, 57 Excision Theorem, 80, 85 face, 8 Five Lemma, 62 xed point, 68 free additive group, 38 Fundamental Homomorphism Theorem, 41 Fundamental Theorem on Abelian Groups, 45 general position, 4 generator of the free additive group, 38 geometrically independent, 4 Hauptvermutung, 33 homeomorphism between pairs, 3 homology (long) exact sequence, 80, 85 homology class, 42 homology group, 42 homology homomorphism, 50 homology homomorphism induced by a chain homomorphism, 50 homology homomorphism induced by, a continuous map, 66, 67 a simplicial map, 52 homology isomorphism, 50 homology of a polyhedron, 67 Homotopical Invariance of Homology, 73, 74 homotopically equivalent, 72 homotopy, 72 homotopy equivalence, 72 homotopy inverse, 72 homotopy type, 72 incidence number, 40 interior of a simplex, 7

86

J.H.C. Whitehead Theorem on subdivision, 23 Klein bottle, 17, 21, 76 linearly independent, 4 link, in a simplicial complex, 13 locally nite simplicial complex, 12 locally nite-dimensional, 12 Mbius band, 19, 74 o manifold, 15 map between pairs, 3 Mayer-Vietoris exact sequence, 57 Mayer-Vietoris Theorem, 57, 71 mesh, 12 Minkowski functional, 8 monomorphism, 57 No Retraction Theorem, 68 null-homotopic, 72 open star, 14 ordered complex, 24 ordered simplex, 37 orientation, of a simplex, 37 of a simplicial complex, 38 oriented complex, 38 oriented simplex, 37 pair of oriented complexes, 79 pair of simplicial complexes, 13 pair of spaces, 3 PL (piece-wise linear) map, 31 PL embedding, 31 PL homeomorphism, 31 polyhedral pair, 15 polyhedron, 12 polyhedron, topological, 15 projective plane, 18, 19, 20, 75 proper face, 8 punctured torus, 77 rank, 39, 45 relative, boundary operator, 79 chain complex, 79 chain group, 79 chain homomorphism, 82, 83

homology, 85 homology class, 80 homology group, 80 homology homomorphism, 82, 84, 85 incidence number, 79 simplicial approximation, 84 subdivision operator, 83 triangulation, 15 retraction, 68 simplex, 5 simplicial approximation, 33 simplicial complex, 12 simplicial embedding, 30 simplicial homeomorphism, 30 simplicial homotopy, 35 simplicial isomorphism, 31 simplicial map, 28, 29 simplicially homotopic, 35 simplicially isomorphic, 31 skeleton, 12 sphere, 16 star, in a simplicial complex, 13 subcomplex, 13 subdivision operator, 60 subdivision, subdivide, 21 surface, 15 Theorem on free additive groups, 45 Tietze Extension Theorem, 69 Topological Invariance of Dimension, 68 Topological Invariance of Homology, 67 torus, 17, 78 triangulation, 15 vertex, 5

87