Introdu

tion to Algebrai topology

Lothar Gotts he
 Introdu tion

The aim of the short ourse is to give a rst introdu tion to algebrai topology,
in parti ular to singular homology of topologi al spa es.
The idea of algebrai topology is to assign algebrai data (e.g. groups) to topo-
logi al spa es. Then problems in topology translate into problems in algebra, whi h
sometimes are mu h easier. More pre isely one wants to asso iate algebrai data to
topologi al spa es in a fun torial way. This means the following: To ea h topolog-
i al spa e X we assign a group F (X ) and to ea h ontinuous map f : X ! Y a
homomorphism f : F (X ) ! F (Y ) su h that
(1) (IdX ) = idF (X ) ,
(2) For X f!Y ! g
Z ontinuous maps
(g Æ f ) = g Æ f:
This will translate problems of topology into problems of algebra.
Example 0.1 A ne essary ondition for X and Y to be homeomorphi is that
.

F (X ) and F (Y ) are isomorphi groups. (Let f : X ! Y be a homeopmorphism

with inverse f 1. Then f 1 Æ f = idX , f Æ f 1 = idY , thus idF (X ) = f 1 Æ f and
idF (Y ) = f Æ f 1 . Thus f is an isomorphism.) This is usually the easiest way to
prove that two spa es with similar topologi al properties are not homeomorphi .
We will look at two su h fun tors:
(1) The fundamental group, whi h is a group made of equivalen e lasses of loops
in X .
(2) Homology groups whi h are made out of maps from simpli es into X .
 CHAPTER 1
The fundamental group and elements of homotopy theory
1. Homotopy of paths and the fundamental group

Let X be a topologi al spa e. Denote by I the losed interval [0; 1℄.

Definition 1.1 A path in X from x0 to x1 is a ontinuous map : I ! X with
.

(0) = x0 , (1) = x1 . X is alled path onne ted if for any x0 , x1 2 X there is a path
from x0 to x1 . A path from x0 to x0 is alled a loop at x0 . Let ;  : I ! X be
two paths from x0 to x1 . We say that  and  are homotopi with end points xed;
written
 '  relf0; 1g;
if there is a map F : I  I ! X s.th.
(1) F (; 0) = , F (; 1) = 
(2) F (0; t) = x0 , F (1; t) = x1 8t .

We will often drop the relf0; 1g. F is alled a homotopy from  to  . Note that in
this ase all Ft := F (; t) are paths from x0 to x1 with F0 = , F1 =  so that  gets
ontinuously deformed into  .

We write Ft :  '  .
3
 Exer ise 1.2 ' is an equivalen e relation, i.e.
.

(1)  ' ,
(2)  '  =)  ' ,
(3)  '  and  ' =)  '
Therefore we an onsider the equivalen e lasses alled the homotopy lasses [℄
of paths.
If  is a path from x0 to x1 and  is a path from x1 to x2, we an dene the
omposition, a path  from x0 to x2 by
8

 (t) =
< (2t) 0  t  12 :
: (2t 1) 1  1  1
2

Remark 1.3 If Ft :  ' 0 and Gt :  '  0 Then FtGt :  ' 0 0 .

.

Therefore we get a well-dened produ t of homotopy lasses

[℄[ ℄ := [ ℄:
Theorem 1.4 Let 1(X; x0 ) be the set of homotopy lasses of loops in X at x0 ,
.

with the produ t [ ℄[ ℄ := [ ℄: Then 1 (X; x0 ) is a group. The neutral element is the
onstant loop x0 and the inverse of [ ℄ is the lass of  1 dened by  1 (t) =  (1 t).

1 (X; x0 ) is alled the rst homotopy group with respe t to x0 .

4
 Proof. We show  1 ' x0 . The homotopy is given by the diagram

i.e. we put 8
>
>
>
<
 (2s) 0st
F (s; t) :=  (t) t  2s  2 t
>
>
>
: 1 (2s t) 2 t  2s  2:

The fun tions are ontinuous on ea h triangle and agree on the interse tions, thus F
is ontinuous. The asso iativity is proven by

(note that this also proves the asso iativity for produ ts of paths and not just for
loops) and that x0 is the neutral element is proven by


Exer ise 1.5 Write this down in formulas.
.

By denition 1 (X; x) depends on the hoi e of the base point x. Is there a relation
between 1 (X; x0 ) and 1 (X; x1)? Obviously this an only be the ase if x0 and x1
lie in the same path- omponent.
5
 Proposition 1.6 Let  be a path from x0 to x1 . Then the map
.

 : 1 (X; x0 ) ! 1 (X; x1 ); [ ℄ ! [ 1 ℄
is an isomorphism.

Proof.  is learly a group homomorphism, and  1 is the inverse, by the

diagram


Corollary 1.7 . If X is path- onne ted, then 1 (X; x) is independent (up to
isomorphism) of x.
In this ase we will often write 1 (X ) for 1 (X; x0 ) and all it the fundamental
group of X .
Definition 1.8 A topologi al spa e is alled simply onne ted if is path- onne ted
.

and 1(X ) = f1g.

The simplest example of a simply onne ted topologi al spa e is one point fg.
We want to see that the fundamental group is fun torial.
Definition 1.9 Let X , Y be topologi al spa es and let x0 2 X , y0 2 Y . Let
.

f : (X; x0 ) ! (Y; y0 ) (this just means f (x0 ) = y0 )be ontinuous. Then we dene

f : 1 (X; x0 ) ! 1 (Y; y0 ); [ ℄ ! [f Æ  ℄:

It is easy to see that this is well-dened and a homomorphism.

6
 Remark 1.10 The following are obvious:
.

(1) (idX ) = id1(X;x0 ) ,

(2) for (X; x0 ) f!(Y; y0) !g
(Z; z0) ontinuous, we have
(g Æ f ) = g Æ f:
Thus the fundamental group is fun torial.
2. Homotopy of maps

Definition 2.1 Let X , Y be topologi al spa es and f; g : X ! Y ontinuous

.

maps. A homotopy between f and g is a ontinuous map F : X  I ! Y with

F (; 0) = f and F (; 1) = g . (Writing again Ft := F (; t) we get a ontinuous family
of maps with F0 = f , F1 = g).
In this ase f and g are alled homotopi and we write f ' g of Ft : f ' g.
If A  X is a subset and gjA = f jA, F is alled a a homotopy rel A if F (a; t) =
f (a) = g (a) for all t and all a 2 A. Note that two paths  ,  are homotopi with
xed endpoints if they are homotopi rel f0; 1g.
Definition 2.2 Two topologi al spa es X , Y are alled homotopy equivalent if
.

there are ontinuous maps f : X ! Y , g : Y ! X with

g Æ f ' idX ; f Æ g ' idY :
In this ase f (or g) is alled a homotopy equivalen e.
X is alled ontra tible if it is homotopy equivalent to a point.

Homeomorphi topologi al spa es are in parti ular homotopy equivalent.

Exer ise 2.3 Show that R n is ontra tible.
.

Now we want to see that homotopy equivalen es indu e isomorphisms of funda-

mental groups. There is a small problem, be ause if f and g are as above, then we
usually have g(f (x0)) 6= x0.
Lemma 2.4 Let F : I  I ! X be ontinuous.
. Put  = F (0; ),  = F (1; ),
= F (; 0), Æ = F (; 1).

7
 Then Æ '  1  relf0; 1g.
Proof. Let x0 = Æ(0), x1 = Æ(1). Then the proof is by putting together the
squares

That is 8
<x
0 st
E (s; t) =
(1 + t s)
: s  t;
8

E (s; t) =
(t + s) 1
< st
:x
1 1 s  t:
Then
Et Ft Gt : Æ '  1  relf0; 1g:

This implies that homotopi maps indu e almost the same map of homotopy
groups.
Corollary 2.5 Let f; g : X ! Y be homotopi via F : X  I ! Y . Let x0 2 X .
.

Let  = F (x0 ; ) : I ! Y (a path from f (x0 ) to g (x0 )). Then we have

g =  Æ f : 1 (X; x0 ) ! 1 (X; g (x0 )):
(Note that  : 1 (X; f (x0 )) ! 1 (X; g (x0 )) is an isomorphism, thus f is an isomor-
phism if and only g is).

8
 Proof. For any loop  at x0 we have a diagram


Theorem 2.6 Let f : X ! Y be a homotopy equivalen e, x0 2 X . Then
.

f : 1 (X; x0 ) ! 1 (Y; f (x0 )) is an isomorphism.

Proof. We have g Æ f ' idX , f Æ g ' idY . Thus, by the orollary, gf and f g
are isomorphisms. Thus f is an isomorphims. 
Corollary 2.7 If X is ontra tible, then X is simply onne ted.
.

Example 2.8 In parti ular Rn is simply onne ted.

.

Note that however simply onne ted manifolds are not ne essarily ontra table,
e.g. the n-sphere S n for n > 1 is simply onne ted, but not ontra table.
3. Fundamental group of the ir le

We will ompute the fundamental group of the ir le S 1, using its over R . It

turns out that the homotopy lass of a loop is determined by how many times it
winds around the ir le. The methods used here generalize: all "ni e" topologi al
spa es X have a so alled universal overing  : Y ! X , su h that  is a lo al
homeomorphism and the bres of  are dis rete and Y is simply onne ted. Then
1 (X ) an be identied with the group of homeomorphisms f : Y ! Y with  =  Æ f .
The arguments are very similar to the ones we use below for the spe ial ase of S 1.
We will not pursue

this

here.
Let S = z 2 C jzj = 1 . The exponential map e : R ! S 1 ; t ! e2it is a lo al
1 
homeomorphism. In parti ular
ej( 1=2;1=2) : ( 1=2; 1=2) ! S 1 n f 1g
is a homeomorphism, let log be its inverse.
9
 We want to show that paths in S 1 and homotopies between them an be lifted
uniquely to R .
Lemma 3.1. (1) (lifting paths) Let  : I ! S 1 be a path with (0) = 1. Then
there is a unique path  0 : I ! R with  (0) = 0 and e Æ  0 =  .
(2) (lifting homotopies) Let  : I ! S 1 be another path and let Ft :  '
 relf0; 1g. Then there is a unique F 0 : I  I ! R , with Ft0 :  0 '  0 relf0; 1g
and e Æ F 0 = F .

Proof. (1) and (2) are very similar and have the same proof. Let either J = I
or J = I  I . Let f : J ! S 1 be either  or F .
(Existen e) Sin e J is ompa t, F is uniformly ontinuous, i.e. there exists " > 0
s.th. jt t0j < " =) jf (t) f (t0 )j < 1. In parti ular f (t0 ) 6= f (t) so that
log (f (t)=f (t0)) is dened. Choose n so that jtj < n" for all t 2 J . Dene
X n  
f 0 (t) := log f (ti=n)=f (t(i 1)=n) :
i=1

Then f 0 : J ! R is ontinuous, f 0(0) = 0, e Æ f 0 = f .

(Uniqueness) Let f 00 be another lift. Then g := f 0 f 00 is ontinuous and has
image in Z. As J is onne ted, g is onstant and thus f 0 = f 00 . 
Theorem 3.2 . 1 (S 1 ) = Z.

Proof. By the Lemma above the end point of the lift 0 depends only on the
homotopy lass [℄. Dene a map
w : 1 (S; 1) ! Z; [ ℄ 7!  0 (1)

w is a homomorphism: Let [ ℄; [ ℄ 2 1 (S 1 ; 1). Let m =  0 (1), n =  0 (1). Let

 00 : I ! R be the path t 7!  0 (s) + m. Then  0  00 is a lifting of  with endpoint
n + m.
w is surje tive: Given n 2 Z dene  0 (s) = ns. Then w([e Æ  0 ℄) = n.
w is inje tive: Assume w([ ℄) = 0, i.e.  0 is a loop in R . As R is ontra table, we
get Ft : 0 ' 0 relf0; 1g. Thus e Æ Ft :  ' 0 relf0; 1g. 
So we have omputed the fundament group in at least one nontrivial ase.
10
 Exer ise 3.3 1(X  Y; (x; y)) = 1 (X; x)  1 (Y; y).
.

Sket h: The map

1 (X  Y; (x; y )) ! 1 (X; x)  1 (Y; y ); [(;  )℄ ! ([ ℄; [ ℄)
is an isomorphism.
So we see that a torus S 1  S 1 has fundamental group Z  Z.
As an appli ation of the Theorem we want to prove the ase n = 2 of the Brouwer
xed point theorem.
Definition 3.4 Let X be a topologi al spa e and A  X . A is alled a retra t
.

of X if there is a ontinuous map f : X ! A with f jA = idA.


Let E n = z 2 R n  jjzjj  1 be the losed unit n dis . Note that E n is
ontra table.
Proposition 3.5 S 1 is not a retra t of E 2 .
.

Proof. Supposed f : E 2 ! S 1 is a retra tion. Let i : S 1 ! E 2 be the in lusion.

Then f Æ i = idS1 . Thus f Æ i = id, whi h is impossible be ause 1(E 2 ) = 0 and
1 (S 1 ) = Z. 
Corollary 3.6 (Case n = 2 of Brouwer xed point theorem). Any ontinuous
.

map f : E ! E has a xed point.

Proof. Let f : E 2 ! E 2 have no xed point. Dene r : E 2 ! S 1 by letting r(x)
be the interse tion point (on the side of x) of the line through x and f (x) with S 1 .

(Exer ise: write the formula for this.) Then r is a retra tion, a ontradi tion. 
Exer ise 3.7 Assume X is the union of two simply onne ted open sets U , V
.

with U \ V nonempty and path- onne ted. Then X is simply onne ted.
Exer ise 3.8 Show that S n is simply onne ted for n  2.
.

11
 1 (X ) an be very ompli ated and does not have to be ommutative. For instan e
let X be a bouquet of k ir les:

Then 1(X ) is the free group on k generators (i.e. a group with k generators and
no relations).
4. Higher Homotopy groups

Let S n be the n-sphere, and let s 2 S n be a point. It is easy to see from the
denition that 1 (X; x0 ) an be interpreted as the set of homotopy lasses of maps
from (S 1 ; s) to (X; x0 ). Higher homotopy groups n(X; x0 ) an be dened as the
set of homotopy lasses of maps (S n; s) to (X; x0). One an show that n (X; x0 ) is
a ommutative group for n > 1. Higher homotopy groups are in general diÆ ult
to understand. A lot of work has for instan e gone (and still is going) into the
omputation of the higher homotopy groups of spheres. It is not diÆ ult to show
that k (S n) = 0 for k < n and that n(S n) = Z. For k > n the answer is very
ompli ated and not ompletely known.

12
 CHAPTER 2
Homology
The main part of the ourse is about homology. The denition of homology is a
bit more ompli ated and less intuitive than that of the fundamental group and of
the higher homotopy groups. Homology is however in many ways more powerful and
easier to handle. There are many ways of dening homology groups, whi h at least
for ni e spa es are all equivalent. The most elementary and most intuitive (whi h
also was the rst histori ally) is simpli ial homology. For this a topologi al spa e is
triangulated, i.e. divided into pie es homeomorphi to simpli es, and these simpli es
are then used to dene the homology groups.

This has the advantage of being intuitive. There are however a number of te hni al
disadvantages. In parti ular it takes a lot of eort to show that the homology groups
are independent of the triangulation.
We will use singular homology, whi h is dened in terms of maps from simpli es
into X . It has the advantage that it obviously is a topologi al invariant of X . In fa t
it is not diÆ ult to show that the homology groups depend only on the homotopy
type of X and furthermore that they are fun torial.

1. Singular homology

The singular homology groups H (X; Z) of a topologi al spa e X are made out of
ontinuous maps from simpli es (points, lines, triangles, tetraeders) into X . So rst
we need to talk about simpli es.
13
 Definition 1.1 Let p0; : : : ; pk 2 R n . The onvex hull of p0 ; : : : ; pk is
.

k
nX  X o
ti pi  0  ti ; ti = 1 :
i=0

In R n let
e0 = (0; 0; : : : ; 0)
e1 = (1; 0; : : : ; 0)
e2 = (1; 0; : : : ; 0)
:::
en = (0; : : : ; 0; 1)
The standard n-simplex
n  R n is the onvex hull of e0 ; : : : ; en.

Definition 1.2 Let X be a topologi al spa e. A singular n-simplex in X is a

.

ontinuous map  :
n ! X .
Example .1.3 Given p0 ; : : : ; pn in R k , they dene a singular n-simplex
X X
(p0; : : : ; pn) :
n ! R k ; ti ei 7! ti pi:
(p0; : : : ; pn) is alled the singular simplex generated by p0; : : : ; pn.

14
 Definition 1.4 The fa e Fi :
n 1 !
n of the standard simplex is
.

(e0 ; : : : ; ebi; : : : ; en) :
n 1 !
n :

For a singular simplex  :
n ! X the omposition (i) :=  Æ Fi :
n 1 ! X is
alled the i-th fa e of .
Remark 1.5 Note that the image Fi(
n) is indeed an (n 1)-dimensional fa e
.

of
n; in fa t it is the fa e oposite to ei.

Now we want to dene singular hains as formal linear ombinations of singular

simpli es.
Definition 1.6 Let Sk (X ) be the free abelian group generated by the singular
.

q -simpli es, i.e.

n X  o
Sk (X ) = formal sums a    singular k-simplex ; a 2 Z :

That is su h a formal sum is 0 if all oeÆ ients are. Elements of Sk (X ) are alled
singular k-simpli es.
The set of singular k- hains is very big, so we annot use it dire tly.
Definition 1.7 The boundary  of a singular k-simplex  :
k ! X is
.

k
X
 := ( 1)i(i) 2 Sk 1(X ):
i=0
Thus

15
 We extend  by linearity to a group homomorphism
X X
 : Sk (X ) ! Sk 1(X ); a  7! a :
The boundary of a 0 hain is dened to be 0.
The starting point for the denition of homology is the fundamental identity
 Æ  = 0.
Lemma 1.8  Æ  = 0.
.

Proof. It is enough to show  = 0 for all singular k-simpli es  :
k ! X .

k
X
 =  ( 1)i Æ (e0 ; : : : ; ebi; : : : ; ek )
i=0
Xk i 1
X
= ( 1)i ( 1)j  Æ (e0 ; : : : ; ebj ; : : : ; ebi; : : : ; ek )
i=0 j =1
k
!
X
( 1)j  Æ (e0 ; : : : ; ebi; : : : ; ebj ; : : : ; ek )
j =i+1
X
= ( 1) i+j  Æ (e0; : : : ; ebj ; : : : ; ebi; : : : ; ek )
j<i
X
( 1)i+j  Æ (e0 ; : : : ; ebi; : : : ; ebj ; : : : ; ek ) = 0
i<j

Definition 1.9 A singular k- hain z with z = 0 is alled a k- y le. If z = 
.

for a (k + 1)- hain , then z is alled a boundary. We denote Zk (X )  Sk (X ) the

subgroup of k- y les and Bk (X ) the subgroup of k-boundaries. By  Æ  = 0 we have
Bk (X )  Zk (X ). The quotient group Hk (X; Z) = Zk (X )=Bk (X ) is alled the k-th
singular homology group of X .
Remark 1.10 One an instead dene homology Hk (X; R) with oeÆ ients in
.

an arbitrary ommutative ring R with 1. The most important ones are R = Z and
R = Q . One takes
nX  o
Sk (X ) := a    singular k simplex a 2 R ;

and pro eeds as before. Almost everything we do works also in this more general
ontext.
16
 We will however sti k to R = Z and most of the time drop the Z from the notation.
Dire tly from the denition it is very hard to ompute the homology groups,
ex ept in the most simple ase, whi h is however important.
Proposition 1.11 . Let X = fpg be a point. Then Hi (X ) = 0 for i > 0 and
H0 (X ) = Z.
Proof. For ea h n  0 there is a unique singular n-simplex, the onstant map
n :
n ! fpg. We have
8
n 1 n even > 0
Xn <
 n = ( 1) n 1 =
i
:0 n odd or 0
i=0

Thus B n(X ) = Zn(X ) for n > 0. However B0 (X ) = 0 and H0(X ) = Z0(X ) = n 0 
n2Z : 
The following proposition relates homology to path omponents. It is the key to
all results in algebrai topology about onne tedness.
Proposition 1.12
L
. (1) Let (Xk )k be the set of path- omponents of X . Then
Hn (X ) = k Hn (Xk ) for all n  0.
(2) If X is path- onne ted, then H0(X ) = Z.
Proof. (1) As
n is path- onne ted, L
 (
n ) lies in a path omponent of X for
any singular simplex . Thus Sn(X ) = Lk Sn(Xk ). Furthermore L
 is just the dire t
sum
L
of the  on the Xk . Thus Zn(X ) = k Zn(Xk ), Bn(X ) = k Bn(Xk ), Hn(X ) =
k Hn (Xk ).
(2) For p 2 X write also p for the singular simplex
0 ! fpg. Choose x0 2 X .
For any x 2 X , let x :
1 = I ! X be a path from x0 to x. Thus x = x x0 .

Given a 0- hain = P axx, we laim 2 B0(X ) () P

ax = 0.

17
 If P ax = 0, then
X X X 
= ax x ax x0 =  ax x :
The onverse is lear.
By denition every 0- hain is a y le. Thus the map sending to the sum of the
oeÆ ients is a surje tive homomorphism S0 (X ) = Z0(X ) ! Z with kernel B0(X ).
Thus H0(X ) = Z. 
Finally we will show that homology is fun torial, i.e. if f : X ! Y is ontinuous,
then there exists homomorphism f : Hn(X ) ! Hn(Y ) with (idX ) = idH (X ) , (g Æ
n

f ) = g Æf . We will use this o asion to introdu e some basi on epts of homologi al
algebra.
Definition 1.13 A sequen e
.

S = S0  S1  S2  : : :
of abelian groups and group homomorphism is alled a hain omplex if  Æ  = 0.
Given a hain omplex S, its i-th homology group is
Hi (S ) := Ker( jSi )=Im( jS 1 ):
i

A morphism F : S ! T of hain omplexes is a sequen e F : Si ! Ti of

homomorphism su h that  Æ F = F Æ  .

S?i 1 S?i
? ?
yF yF

Ti 1 Ti
Remark 1.14 The homology of hain omplexes is fun torial: If F : S ! T is
.

a morphism of hain omplexes, then we dene Hi(F ) : Hi(S) ! Hi(T) as follows:

Let [a℄ 2 Hi(S ) be represented by a 2 Ker( jS ). Then F (a) 2 Ker( jT ) and we
i i

put Hi(F )([a℄) = [F (a)℄. The relation  Æ F = F Æ  implies that this is well-dened.
We also see Hi(idS ) = idH (S) and Hi(G Æ F ) = Hi(G) Æ Hi(F ).
i

Remark 1.15 We all S(X ) = S0 (X )  S1(X )  : : : the singular hain om-

.

plex. We see that it is a hain omplex and that its homology groups are the Hi(X ).
We see that the map X 7! S(X ) is fun torial: If f : X ! Y is ontinuous, we put
X X
f : Si (X ) ! Si (Y ); a  7! a f Æ :

18
We see that f Æ  =  Æ f . Thus f is a morphism of omplexes and (idX ) = idS(X ) ,
(g Æ f ) = g Æ f.
Definition 1.16 For f : X ! Y ontinuous, we put
.

X  X 
f : Hi (X ) ! Hi (Y ); a  7! a f Æ  :
Thus (f : Hi(X ) ! Hi(Y )) = Hi(f : S(X ) ! S(Y )). So it is fun torial.
Remark 1.17 As the Hi(X ) are fun torial, they are in parti ular topologi al
.

invariants, i.e. if f : X ! Y is a homeomorphism, then Hi(f ) : Hi(X ) ! Hi(Y ) is

an isomorphism.
Remark 1.18 Let Æn := id
A singular simplex  :
n ! X gives
. n

 : Sn (
n ) ! Sn (X );  7!  Æ :
Note that by denition  (Æn) = . This observation is often used to use fun toriality
to redu e the proof of properties of singular simpli es to the spe ial ase of Æn. We
will see an instan e of this in the next se tion.
2. The homotopy invarian e theorem

We want to show that the homology groups are homotopy invariants, not just
topologi al invariants.
Theorem 2.1 Let f; g : X ! Y
. be homotopi maps, then f = g : Hi(X ) !
Hi (Y ) for all i  0.

Corollary 2.2 . If X and Y are homotopy equivalent, then Hi (X ) = Hi (Y ) for

all i  0.
8

Corollary 2.3 . If X is ontra tible, then Hi (X ) =

< 0 i>0
:Z i=0
In parti ular this is true for X = R n .
We shall show that the homotopy indu es a hain homotopy between f ; g :
S (X ) ! S (Y ).

19
 Definition 2.4 Let F; G : S ! T be morphisms of omplexes. A hain homo-
.

topy from F to G is a sequen e of maps s : Si ! Ti+1 , su h that

s s = G F:
i.e. 
?S?
i
s
S?i+1
?
??
yyFG . ??
FG
yy
Si 
Si+1
If a hain homotopy from F to G exists, then F and G are alled homotopi .
Proposition 2.5 . If F and G are homotopi , then Hi (F ) = Hi (G) for all i.
Proof. If a 2 Ker( jS ), then s(a) = 0, thus G(a) F (a) = s(z), so G(a) and
i

F (a) have the same homology lass. 

Proof. (of the Theorem). First we redu e to the spe ial ase of maps t : X !
X  I ; x 7! (x; t). Then we have a homotopy i : 0 ' 1 .
Assume the theorem for 0 and 1. Let F : X  I ! Y be a homotopy from f to
g . Then F Æ 0 = f , F Æ 1 = g and thus
f = F Æ (0 ) = F Æ (1 ) = g :
Thus we only need to prove (0 ) = (1 ).
For this we will onstru t a hain homotopy P : Sn(X ) ! Sn+1(X  I ) from (0 )
to (1 ) i.e. we want
(P P  )() = 1 Æ  0 Æ ;
for all singular n-simpli es . Now we want to use fun toriality to redu e to the ase
X =
n ,  = Æn .
We want to onstru t P in a fun torial way, i.e. we want that for any f : X ! Y
Sn?(X ) P! Sn+1 (? X  I)
?f ?(f 1 )
y y I 

Sn (Y ) ! Sn+1(Y  I )
P

ommutes. This requirement gives us as only hoi e P () = (  1I )(P (Æn)). On

the other hand this denition is obviously fun torial. Write i for i in ase X =
n.
Assume that we have dene P (Æn) in su h a way that
(P P  )(Æn ) = 1 Æ Æn 0 Æ Æn:
20
Let  : Æn ! X be a singular n-simplex. Then, using that also f = f  by denition,
(P P  )( ) = (P P  )( Æn ) = (  1I ) (P P  )(Æn ))
= (  1I )(1 0 ) (Æn ) = (1 0 ) ( ):

Now we have to dene P on Æn. The polyhedron
n  I  R n+1 has 2(n + 1)

verti es A0; : : : ; An, B0 ; : : : ; Bn given by Ai := (ei ; 0), Bi := (ei ; 1). We dene
n
X
P (Æn ) := ( 1)i(A0 ; : : : ; Ai; Bi; : : : ; Bn):
i=0

For instan e
P (Æ0 ) = (A0 ; B0 );

P (Æ1 ) = (A0 ; B0 ; B1 ) (A0; A1 ; B1):

In general the ee t of P is to make a (n + 1)- hain made of prisms whose union

is
n  I .
Claim: ( Æ P + P Æ  )(Æn ) = 1 Æ Æn  0 Æ Æn .
The right hand side gives
1 Æ Æn 0 Æ Æn = (B0 ; : : : ; Bn ) (A0 ; : : : ; An):
21
Now we ompute the left hand side.
n
X
P (Æn ) = ( 1)i  (A0 ; : : : ; Ai; Bi; : : : ; Bn)
i=0
X
= ( 1)i+j (A0; : : : ; A j ; : : : Ai; Bi; : : : ; Bn)
j i
X
( 1)i+j (A0; : : : ; Ai; Bi; : : : B j ; : : : ; Bn)
ij
All terms with i = j an el with the ex eption of (B0 ; : : : ; Bn) (A0; : : : ; An). Thus
P (Æn ) = (B0 ; : : : ; Bn ) (A0 ; : : : ; An )
X
+ ( 1)i+j (A0 ; : : : ; A j ; : : : Ai; Bi; : : : ; Bn)
j<i
X
( 1)i+j (A0; : : : ; Ai; Bi; : : : B j ; : : : ; Bn)
i<j
On the other hand k
X
Æn = ( 1)j (e0; : : : ; ebj ; : : : ; en):
j =0
Sin e
X
P ((e0 ; : : : ; ebj ; : : : ; en )) = ( 1)i(A0 ; : : : ; Ai; Bi; : : : B j ; : : : ; Bn)
i<j
X
( 1)i (A0; : : : ; A j ; : : : Ai; Bi; : : : ; Bn);
i>j
the result follows. 

22
 3. Relation between fundamental group and homology

We now want to show that the fundamental group and the rst homology group
are losely related. We know that the fundamental group is in general not ommu-
tative, while the homology groups are all ommutative. We will show that the rst
homology group is just the fundamental group made ommutative (i.e. divided by its
ommutator subgroup).
If : I ! X is a loop in X , we may regard it as a singular 1-simplex (be ause by
denition
1 = I ). Fix a point x0 2 X and let L(x0 ) be the set of loops at x0 . Then
we have a map  : L(x0 ) ! S1(X ), sending a loop  to itself viewed as a singular
simplex. The following result implies that this indu es a group homomorphism  :
1 (X; x0 ) ! H1 (X ).
Proposition 3.1 If two loops ;  are homotopi , then theys are homologuous
.

as singular 1-simli es. Moreover, given two loops 1 ; 2 at x0 , we have

( 1 2 ) = ( 1 ) + ( 2 ):
Proof. Choose a homotopy F : I  I ! X with
F (t; 0) = ;
F (t; 1) = ;
F (1; s) = F (0; s) = x0 ; 8s :
Consider the points e0 = (0; 0), e1 = (1; 0), e2 = (0; 1), Q = (1; 1) 2 R 2 .

Note that
= F Æ (e0 ; e1 )
 = F Æ (e2; Q)
and
F Æ (e1 ; Q) = F Æ (e0 ; e2 ) = x0 :

23
 Dene the singular 2- hain
 := F Æ (e0 ; e1 ; Q) F Æ (e0 ; e2 ; Q):

Then
 = F Æ (e1 ; Q) F Æ (e0 ; Q) + F Æ (e0 ; e1 )
F Æ (e2 ; Q) + F Æ (e0 ; Q) + f Æ (e0 ; e2 )
= x0 +  x0 = 

This proves that and  are homologuous.

To prove the se ond laim, we need to dene a singular 2-simplex  with
 = +   :

Consider the triangle

where T = (1=2; 0). If Q 2
2 lies on or below the line  joining e1 with T

onsider the line joining e0 with Q parametrized linearly, su h that t = 0 at e0 and
t = 1 at the interse tion point with , and set  (Q) := 1 (t).

24
 Analoguously, if Q lies on or above , onsider the line joining e2 with Q and
parametrize it in su h a way that t = 1 at e2 and t = 0 at the interse tion with 
and set (Q) =  (t).

Then
 ( ) =  Æ (e1 ; e2 )  Æ (e0 ; e2 ) +  Æ (e0 ; e1 )
=   + :
Thus ([ ℄[ ℄) = ([ ℄) + ([ ℄). 
Now we want to see that  is surje tive and the kernel of  is the ommutator
subgroup of 1 (X ).
Definition 3.2 The ommutator subgroup C (G) of a group G is the subgroup
.

generated by all ghg 1h 1 with g; h 2 G. One easily shows that C (G) is a normal
subgroup of G and G=C (G) is abelian, alled the abelianization of G. By denition
it is lear that, if  : G ! H is a homomorphism to an abelian group, then ker() 
C (G).

It turns out that H1(X ) is the abelianization of 1(X ).

Theorem 3.3.Let X be path- onne ted Then  : 1 (X; x0 ) ! H1 (X ) is surje tive
with kernel the ommutator subgroup of 1 (X; x0 ).

25
 Proof. Let z = P nii be a 1- y le, i.e.
X
0 = z = ni (i(1) i(0)):
In this sum, some of the points an oin ide, the sum of the oeÆ ients orresponding
to every point must vanish.
Choose paths i from x0 to i(0) and i from x0 to i(1), where the paths only
depend of the points (i.e. if some points on ide, then also the paths are the same).

Then X
0 = ni(i i );
be ause the paths orrespond pre isely to the points. Now put
 i := i + i i :

Then X X
z= ni i = ni  i :
Let i be the loop ii i 1 Then
Y 
X
 ni i = ni  i

= [z℄:
Thus  is surje tive.
Finally we have to determine the kernel of . First, sin e H1 (X ) is abelian, ker()
ontains the ommutator subgroup. We need to show that every element of ker()
lies in the ommutator subgroup. The proof is not diÆ ult, but a little bit onfusing.
26
 Let be a loop, whi h is a 1-boundary, i.e.
X 
= ni i

Write i = ( 0;i 1;i + 2;i) for paths 0;i; 1;i; 2;i: It means that one of the i;j , say
of the i0;j0 is equal to and o urs in the sum with oeÆ ient 1. After olle ting
terms, all other paths o ur with oeÆ ient 0.

Choose paths
0;i from x0 to 2;i (0) = 1;i (0)
1;i from x0 to 0;i (0) = 2;i (1)
2;i from x0 to 1;i (1) = 0;i (1)

again the paths should only depend on the points, i.e. if some points oin ide we
take the same paths, and if the point is x0 we hoose the onstant path x0 . Consider
the loops
0;j := 1;j 0;j 2;j1
1;j := 0;j 1;j 2;j1
2;j := 0;j 2;j 1;j1

27
Note that i0 ;j0 ' i0 ;j0 = . Note that the loops
j := 0;j 1;j1 2;j = 1;j 0;j 1;j1 i;2 1;j1
are homotopi to the onstant loop x0 sin e the image of a 1-simplex is ontra table.

Hen e Qi [i℄n = 1 in 1 (X; x0 ). Let 10 := 1 (X; x0 )=C (1(P

i X; x0 )), and let ei be
the image of [i℄ in 10 . We use additive notation for 10 . Then i ni ei = 1. Sin e 1
is ommutative, we an olle t terms and get
X X
ni ein =
i
ni (e0;i e1;i + e2;i )
i i
= e
Note that the fa t that the i;j ex ept i0 ;j0 = an el out in the rst sum is
equivalent to the i;j ex ept i0 ;j0 ' i0 ;j0 = an eling out. Thus 2 C (1 (X; x0 )).

Corollary 3.4 If X is simply onne ted, then H1(X ) = 0.
.

Corollary 3.5 . 8
<Z n = 1;
H1 (S ) =
n
:0 n > 1

Corollary 3.6 H1(X  Y ) = H1(X )  H1(Y ).

.

4. Relative Homology

We have to extend the notion of homology to pairs A  X of topologi al spa es.

Definition 4.1 A pair of topologi al spa es is a topologi al spa e together with
.

a subspa e A  X ( subspa e means that A  X and that A arries the indu ed

topology). A morphism f : (X; A) ! (Y; B ) of pairs is a ontinuous map f : X ! Y
with f (A)  B .
28
 Now we want to dene relativ homology groups Hi(X; A) for pairs (X; A).
Definition 4.2 The relative i- hains are
.

Sn (X; A) := Sn (X )=Sn(A);
and we dene a boundary map
 : Sn (X; A) ! Sn 1 (X; A); [a℄ 7! [a℄:
Thus the diagram
Sn?(X ) ! Sn(X;

? A)
? ?
y y
Sn 1 (X ) ! Sn 1(X; A)


where the horizontal maps are the quotient maps, ommutes. As  (Sn (A))  Sn 1(A),
 is well-dened. As  Æ  = 0 it follows that  Æ  = 0. Thus
 
S (X; A) := (S0 (X; A)  S1 (X; A)  : : : )
is a hain omplex. The relative homology groups
Hn(X; A) := Ker(jS (X;A) )=Im(jS 1 (X;A) )
n n

are the homology groups of this omplex.

We an des ribe the relative homology also dire tly in terms of Sn(X ): The group
of relative n- y les modulo A is
 
Zn (X; A) := 2 Sn (X )   2 Sn 1 (A) :
The group of relative n-boundaries modulo A is
Bn (X; A) :=  (Sn+1 (X )) + Sn (A)  Sn (X ):
Then by the isomorphism theorem
(M=P )=(N=P ) = M=N
we get
Hn (X; A) = Zn (X; A)=Bn (X; A):

29
 Example 4.3 A path  : I ! X is a relative 1- y le if modulo A if its end points
.

lie in A.

Remark 4.4 If A is empty, then Sn(A) = 0 for all n, thus Hn(X; ;) = Hn(X ).
.

Thus the usual homology is a spe ial ase of the relative homology.
Remark 4.5 Relative homology is fun torial in the pair (X; A), i.e. if f :
.

(X; A) ! (Y; B ) is a morphism of pairs, then we have a homomorphism

f : Hn (X; A) ! Hn (Y; B )
su h that
id(X;A) = idH (X;A) ;
n (g Æ f ) = g Æ f :
The proof is the same as before. For P aii 2 Zn(X; A) we put
X 
X 
f ai i := ai f Æ i :
Example 4.6 In parti ular we have the in lusion map j : (X; ;) ! (X; A) whi h
.

is the identity on X . giving

j : Hn (X ) ! Hn (X; A):
Proposition 4.7 Let (Xk )k be the family of path- omponents of X , and put
.

Ak = Xk \ A. Then for all n  0,

X
Hn (X; A) ' Hn(Xk ; Ak ):
k
Proof. The proof is the same as for Hn(X ). 
Proposition 4.8 If A is nonempty and X is path onne ted, then H0(X; A) = ;
.

Proof.
P
Choose x0 2 A. For ea h x 2 X let x be a path from x0 to x. For
= x x 2 S0 (X ), we get
X  X
 x x = x x0 :
Thus 2 B0(X; A). 
30
 It follows that in general H0 (X; A) is the free abelian group on the onne ted
omponents of X whi h do not meet A.
Remark 4.9 Two maps f; g : (X; A) ! (Y; B ) are alled homotopi if they are
.

homotopi as maps X ! Y by a homotopy F with F (A  I )  B . It is easy to

modify the above proof of homotopy invarian e of homology to show that if f; g :
(X; A) ! (Y; B ) are homotopi , then
f = g : Hn (X; A) ! Hn (Y; B ):

5. Long exa t homology sequen e

We are not really interested very mu h in the relative homology groups in their
own right, we want to use them as a tool to ompute ordinary homology groups. Thus
the most important property of the relative homology groups Hn(X; A) is the long
exa t homology sequen e, whi h relates them to the usual homology groups Hn(X )
and Hn(A).
First we make a short interlude in homologi al algebra.
Definition 5.1 A sequen e A0 f!0 A1 f!1 : : : f !1 An of homomorphisms of abelian
.
n

groups is alled exa t at Ai if Ker(fi) = Im(fi 1 ). It is alled exa t if it is exa t at

A1 ; : : : ; An 1 .
In parti ular
0 !A f!B ! g
C !0
is exa t ( alled a short exa t sequen e) if f is inje tive g is surje tive and Ker(g) =
Im(f ).
For instan e the sequen e
0 ! Z
! n
Z !Z=(n) !0

is exa t.
Note that by denition a hain omplex (with S 1 = 0) (S ;  ) is exa t if and only
if all homology groups vanish. Thus the homology measures the failure of exa tness.
Definition 5.2 Let S F!T G!U be morphisms of hain omplexes. The se-
.

quen e
0 !S F!T G!U !0
31
is alled a short exa t sequen e of hain omplexes if for all i
0 !Si F!Ti G!Ui !0
is a short exa t sequen e of abelian groups.
We have the following fundamental fa t whi h lies on the basis of many results
about homology and ohomology:
Theorem 5.3 Let
.

0 !S F!T G!U !0

be a short exa t sequen e of hain omplexes. Then there exists a long exa t homology
sequen e
::: !
Æ
Hn (S ) F!
 H (T ) G!H (U ) !
n  n 
Æ
Hn 1 (S ) F!
 H (T ) G!H (U ) !
n 1  n 1 
Æ

::: !
Æ
H0 (S ) F!

H0 (T ) G!H0 (U ) !0
Proof. This is proven by a te hnique alled diagram hasing, i.e. one takes
elements in one of the groups and maps them either forward by one of the maps one
has, or takes a preimage under one of the maps, until they do what you want. At
any given step one only has a small number of hoi es, thus given enough patien e
one an usually nd the proof for a given true statement.
First we want to onstru t Æ : Hn(U ) ! Hn 1(S). I write Zn(U ) for Ker( jU ),
n

Bn 1 (U ) := Im( jU ) and similarly for S , T .

n

(0) Let [a℄ 2 Hn(U ) for a 2 Zn(U ). We need to onstru t Æ([a℄) := [ ℄ for
2 Zn 1 (S ). We do this a ording to the diagram, where the horizontal maps are
F; G and the verti al maps are  and the numbers refer to the order in whi h the
elements are onstru ted.
S T U

n b?(1) ! a?(0)
G
? ?
y y
n 1 ?(3) ! b(2) G! 0
F
?
y
0
(1) There exists b 2 Tn with G(b) = a (be ause G is surje tive).
32
 (2) We have b 2 Tn 1. Note that
G(b) = G(b) = a = 0:
(3) Thus by the exa tness of the sequen e, there exists a unique 2 Sn 1 with
F ( ) = b: Furthermore
F ( ) = F ( ) = b = 0:
As F is inje tive we get  = 0, thus 2 Zn 1(S ), and we put Æ([a℄) = [ ℄ 2 Hn 1(S).
Exer ise 5.4 Che k that this is independent of the hoi e of b.
.

We now need to he k exa tness of the sequen e. We he k it at Un and leave the

rest as an exer ise. First we he k Æ Æ G = 0. Let [b℄ 2 Hn(T ) with b 2 Zn(T ). Then
for [a℄ = G ([b℄) we an take b as the b in the above denition of . Thus b = 0 and
therefore = 0, i.e. G ([b℄) = 0.
Now let [a℄ 2 Hn(U ) with a 2 Zn(U ) and Æ([a℄) = 0. We need to nd b0 2 Zn(T )
with G ([b0℄) = [a℄. That Æ([a℄) = 0 means that in the above diagram
d
?
b G! a
?
? ?
y y
! b
we have 2 Bn 1 (S ), i.e. there exists d 2 Sn with = d. Put b0 = b F (d). Then
G(b0 ) = G(b) = a (be ause G Æ F = 0) and
b0 = b F (d) = b b = 0
Therefore b0 2 Zn(T ) and [a℄ = G ([b0 ℄). 
Corollary 5.5 (long exa t ohomology sequen e) Let A
. X be a subspa e.
There is a long exa t homology sequen e

::: !
Æ
Hn (A) i!

Hn (X ) j!

Hn (X; A) !
Æ
Hn 1 (A) i!

Hn 1 (X ) j!

Hn 1 (X; A) !
Æ

::: !
Æ
H0 (A) i!

H0 (X ) j!

H0 (X; A) !0
where i is the pushforward by the in lusion A ! X and j is the pushforward by the
in lusion X ! (X; A).

33
 Proof. By denition we have that the maps i : S (A) ! S (X ) and j :
S (X ) ! S (X; A) ommute with  , i.e. they are maps of omplexes. By deni-
tion i is inje tive and Sn(X; A) = Sn(X )=Sn(A), i.e
0 ! S (A) i!

S (X ) j!

S (X; A) ! 0

is an exa t sequen e of omplexes. 

Exer ise 5.6 Show that the exa t homology sequen e is fun torial in (X; A),
.

i.e. if f : (X; A) ! (Y; B ) is a map of pairs, then the diagram

: : : ! Hn?(A) ! Hn?(X ) ! Hn (X; ? A) ! Hn ?1(A) ! : : :
? ? ? ?
y y y y
::: ! Hn(B ) ! Hn(Y ) ! Hn(Y; B ) ! Hn 1(B ) ! : : :
is ommutative.
Example 5.7 Let
.
 
E n := x 2 R n  jxj  1
be the n-ball and
 
S n 1 := x 2 R n  jxj = 1
the (n 1)-sphere. E n is ontra table, therefore Hk (E n) = 0 for k > 0 and by the
exa t homology sequen e
 : Hk (E n ; S n 1 ) ! Hk (S n 1 )

is an isomorphism for all k > 1. We have an exa t sequen e

0 ! H1(E n; S n 1) ! H0(S n 1) ! H0(E n) ! 0
If n > 1 the last two groups are Z and thus H1 (E n ; S n 1 ) = 0. One also sees
H1 (E 1 ; S 0 ) = Z.

The exa t homology sequen e will be useful for the omputation of homology
groups, on e we have a method for omputing the relative homology. This we will
introdu e in the next se tion.
34
 6. The Ex ision Theorem

Let X be a topologi al spa e and A  X a subspa e. The relative homology

Hn (X; A) has mostly been introdu ed as a tool to ompute the usual homology
Hn (X ). This obviously only an work if we have a method for omputing relative
homology. The most important result for doing this is the ex ision theorem. It says
that ertain subspa es U  A an be ut out (ex ised) without ee ting the relative
homology.
Definition 6.1 The in lusion map j : (X n U; A n U ) ! (X; A) is alled an
.

ex ision if it indu es an isomorphism j : Hk (X n U; A n U ) ! Hk (X; A) for all k  0.

In this ase we say that U an be ex ised.
Theorem 6.2
(Ex ision) Assume that the losure U of U in X is a subset of the
.

interior Ao of A. Then U an be ex ised.

Definition 6.3 Let (Y; B )  (X; A). (Y; B ) is alled a deformation retra t
.

of (X; A), if there is a retra tion r : (X; A) ! (Y; B ), su h that for the in lusion
i : (Y; B ) ! (X; A), i Æ r is homotopi to the identity on (X; A).

Theorem 6.4 . Let V  U  A and assume

(1) V an be ex ised,
(2) (X n U; A n U ) is a deformation retra t of (X n V; A n V ).
Then U an be ex ised.

Proof. Let i : (X n U; A n U ) ! (X n V; A n V ) be the in lusion. Then i :

Hk (X n U; A n U ) ! Hk (X n V; A n V ) is an isomorphism for all k. Sin e homology is
a fun tor and V an be ex ised, also U an. 
Before proving the ex ision theorem, we give some appli ations.
Proposition 6.5
Let En+ , En be the losed northern and southern hemisphere
.

of S n, n  1 (thus En+ \ En is the equator S n 1 . Then

(En+; S n 1) ! (Sn; En )
is an ex ision.

35
 Proof. Let  
V := x 2 S n  xn+1 < 1=2 :
Then by the ex ision theorem V an be ex ised. We see that (En+; S n 1) is a de-
formation

retra t

of (S n n V; En n V ) (move up along great ir les). Therefore
U := x 2 S n  xn+1 < 0 an be ex ised.


Now we an ompute the homology of S n.
Corollary 6.6 For all n  1 we have
.
8
<Z q = n or q = 0
Hk (S n) =
:0 otherwise.
8
<Z q=n
Hk (En ; S n 1) =
0
: otherwise.
Proof. Proje ting on the rst n oordinates gives a homeomorphism (En+; S n 1) !
(E n; S n 1). We have seen that the onne ting homomorphism Hk (E n; S n 1) !
Hk 1 (S n 1 ) is an isomorphism for k > 1. As En is ontra tible, the long exa t
homology sequen e gives that Hk (S n) ! Hk (S n; En ) is an isomorphism. Combining
this with the isomorphism Hk (E n; S n 1) ! Hk (S n; E n) of the proposition we get
36
Hk (S n ) = Hk (S n; En ) = Hk (E n; S n 1) = Hk 1(S n 1) for all k > 1, n > 0. So we
have for k > 1,n > 0 that Hk (S n) = Hk l(S n l ) and Hk (En; S n 1) = Hk 1(S n 1).
As Hk (S 0) = 0 for k > 0, it follows that Hk (S n) = 0 for k > n. If 0 < k  n, then
Hk (S n ) = Hk (E n ; S n 1) = H1 (S n k+1). Therefore it suÆ es to ompute H1 (En ; S n 1 )
and H1(S n).
We have shown 8
<0 n > 1
H1 (E n ; S n 1 ) =
:Z n = 1

We have an exa t sequen e

0 ! H 1 (S n ) ! a
H1 (S n ; En ) !b
H0 (En ) !

H0 (S n ) ! 0:
As is an isomorphism, b must be 0 and therefore a is an isomorphism. So
8
<0 n > 1;
H1 (S ) = H1 (S ; En ) = H1 (E ; S ) =
n n n n 1
:Z: n = 1


Now we an prove the general version of Brouwers xed point theorem. The proof
is the same as in the n = 2 ase.
Corollary 6.7 S n 1 is not a retra t of S n.
.

Proof. Suppose f : E n ! S n 1 is a ontinuous map whose restri tion to S n 1

is the identity, then we get
Hn 1 (S n 1 ) ! Hn 1 (E n ) ! Hn 1 (S n 1 );
su h that the omposition is the identity. Thus for n > 1 this is
Z ! 0 ! Z;

whi h is impossible. (Exer ise: do the ase n = 1.) 

Theorem 6.8 (Brouwers Fixed point theorem) Any ontinuous map E n ! E n
.

has a xed point.

Proof. The proof is the same as before. 
In order to prove the theorem we will have to show that we an ompute H(X; A)
with simpli es whi h lie either entirely in X n U or in Ao.
37
 Definition 6.9 Let U := (Ui)i be an open over of a topologi al spa e X . For
.

ea h k  0 let
X 
SkU (X ) := a  2 Sk (X )  8k 9i :  (
k )  Ui :
These are the linear ombinations of simpli es whose image is ontained in one of the
open sets of the over. Then
SU (X ) = (S0U (X )  S1U (X )  : : : )
is a sub- hain omplex of S (X ). We dene SkU (X; A) = SkU (X )=SkU (A). The in lu-
sions SkU (X ) ! Sk (X ) and SkU (A) ! Sk (A) indu e maps i : SkU (X; A) ! Sk (X; A).
Thus we get the hain omplex SU (X; A) with a hain map i : SU (X; A) ! S (X; A).
Theorem 6.10(of over) SU (X ) ! S (X ), SU (A)
. ! S (A), SU (X; A) !
S (X; A) indu e isomorphisms on homology for all k  0.

Proof. (of Ex ision Theorem from Theorem of over) Consider the over U :=
fX n U; Aog of X . Then SU (X; A) = SU (X )=SU (A). We have SkU (X ) = Sk (X n U ) +
Sk (Ao ). We have
S (X n U ) + Sk (Ao )
SkU (X )=SkU (A) = k = Sk (X n U )=Sk (A n U ):
Sk (A n U ) + Sk (Ao )
Thus we obtain by the theorem of over
Hk (X; A) = Hk (SU (X; A)) = Hk (S (X n U; A n U )) = Hk (X n U; A n U ):

In order to prove the theorem of the over we need to subdivide simpli es into
smaller simpli es. We do this by bari entri subdivision.
Definition 6.11 Let p0; : : : ; pk 2 R n . Let  := (p0; : : : ; pk ) be the orresponding
.

singular k-simplex. Let B 2 R n . The join B   is dened as

B   := (B; p0 ; : : : ; pk ) :
k+1 ! R n :
We extend this by linearity to a map
X  X
B  : Sk (R ) ! Sk+1 (R ); B 
n n a  = a B  :
 

38
 Note that for = P aii 2 Sk (R n ) we have
8
< B  ( ) k > 0;
 (B  ) =
: P a B k = 0:
i i

We want to dene subdivision operators

Sd : Sn (X ) ! Sn (X )
whi h are fun torial, i.e. if f : X ! Y is a ontinuous map, then the diagram
(X ) Sd! Sn?(X )
Sn?
? ?
yf yf
Sn (Y ) ! Sn(Y )
Sd

ommutes. Thus it suÆ es to dene Sd(Æn), be ause for  :
n ! X a singular

simplex, we an put
Sd( ) := Sn ( (Æn )) =  (Sd(Æn )):
We indu tively dene
Sd(Æ0 ) = Æ0
Sd(Æn) = Bn  Sd(Æn )
Here Bn is the bari enter of
n:
Bn :=
X n
1 e:
i=0
n+1 i

Lemma 6.12 . Sd = Sd .

39
 Proof. It is enough to show Sd(Æn) = Sd (Æn). We prove this by indu tion on
n. The ase n = 0 is obvious. Now
Sd(Æn ) =  (Bn  Sd(Æn )) = Sd(Æn ) Bn   (Sd(Æn )):
By indu tion
Bn  SdÆn = Bn  Sd( 2 Æn ) = 0:

Proposition 6.13 . Sd and idS (X ) are hain homotopi .
Proof. Dene T : Sn(X ) ! Sn+1(X ) again to be fun torial, i.e. so that for all
f : X ! Y the diagram
(X ) T! Sn+1?(X )
Sn?
? ?
yf yf
Sn (Y ) ! Sn+1 (Y )
T

ommutes. Thus it is again enough to dene T (Æn). We put T (Æ0 ) = 0 and indu tively
T (Æn ) := Bn  (Æn SdÆn T Æn ):
Claim:

T + T  = Id Sd:
Again it is enough to show this for Æn. We do this by indu tion on n. By indu tion
T Æn = (Id Sd T  )(Æn ) = Æn SdÆn :
Thus we get
T Æn = B (Æn SdÆn T Æn )
= Æ SdÆ T Æ B (Æn SdÆn T Æn )
= Æn SdÆn T Æn :

Remark 6.14. (1) Note that the operators Sd, T send SU (X ) into itself.
(2) By restri tion we also have the orresponding operators Sd; T : S(A) !
S (A). Thus by passing to the quotient we get Sd; T : S (X; A) ! S (X; A)
whi h again send SU (X; A) to itself.
40
 (3) It also follows that on S(X; A) we have
Sd = Sd
T + T  = Id Sd:
Let  = (p0 ; : : : ; pk ) be a singular simplex in R n . Then (Æk ) is a ompa t set.
Let d() be its diameter.
(Exer ise: d() is the maximum of the length of the edges of .)
Lemma 6.15 Ea h singular simplex appearing in the k- hain Sd() has diameter
.
kd()
at most k+1 .

Proof. Exer ise. 

Proposition 6.16 Let  be a singular n-simplex in X , let U
. := (Ui)i be an open
over of X . Then there is an r > 0 su h that Sdr  2 SnU (X ):
Proof. Sin e
n is ompa t, there is an  > 0 su h that  maps an -neighbourhood
of any point in
n into one of the Ui (exer ise). Sin e
nr
lim
r!1 (n + 1)r
= 0;
there is an r > 0 su h that Sdr Æn is a linear ombination
X
Sdr Æn = ai i
i
of simpli es i of diameter smaller then . But then
X
Sdr  = ai  Æ i 2 SnU (X ):
i

Proof. (of the theorem of over). Let Y = X or Y = A or Y = (X; A). We have
to show that the anoni al map j : Hn(SnU (Y )) ! Hn(Y ) is an isomorphism.
Surje tivity: Let [ ℄ 2 Hn (Y ) represented by 2 Si (Y ) with  = 0. For r
suÆ iently
Pr 1 i
large we have Sdr ( ) 2 SnU (Y ) and Sdr ( ) = Sdr ( ) = 0. Putting
b := i=0 Sd ; we get b = 0 and
Sdr ( ) = (Id Sd)b = (T + T  )b = T b:
Thus [ ℄ = [Sdr ( )℄ and [Sdr ( )℄ is in the image of j .
41
 Inje tivity: Let a 2 SnU (Y ), with j ([a℄) = 0. This means a = b with b 2 Sn+1 (Y ).
Then Sdr (a) = Sdr (b) and for r suÆ iently large Sdr (b) 2 SnU+1(Y ). Finally in
SnU (Y ) we have that a Sdr (a) is a boundary (as in the proof of the surje tivity).
Thus a is a boundary in SnU (Y ). Thus j is inje tive. 
7. Mayer-Vietoris sequen e

The most useful tool for omputing the homology of a spa e is the Mayer-Vietoris
sequen e. Under suitable assumptions it allows to ompute the homology of X [ Y
in terms of that of X , that of Y and that of X \ Y .
Definition 7.1 A triple (X; X1 ; X2) of a topologi al spa e and subspa es X1 ; X2 
.

X with X = X1 [ X2 is alled an exa t triad if the in lusion map (X1 ; X1 \ X2 ) !

(X; X2) is an ex ision (i.e. indu es an isomorphism on homology).
Main example: If X = X1 [ X2 and X1 and X2 are open in X , then (X; X1 ; X2 )

is an exa t triad. Putting A = X1, U = X1 n X1 \ X2 , in the statement of the ex ision

theorem, we see X n U = X2 , thus U is a losed subset of X ontained in the open
A. So the ex ision theorem applies.
Theorem 7.2 (Mayer-Vietoris sequen e): Let (X; X1; X2) be an exa t triad.
.

Denote ik : Xk ! X and jk : X1 \ X2 ! Xk the in lusions. Then there is a long

exa t sequen e
!Hn(X1 \ X2)j1!j2Hn(X1)  Hn(X2 )( i1!
 ;i2 )
Hn (X ) !Hn 1 (X1 \ X2 ) : : :
!H0 (X1 \ X2)j1!j2H0(X1)  H0(X2 )( i1!  ;i2 )
H0 (X ) !0
Furthermore the sequen e is fun torial in exa t triads.
Proof. (in the ase X1; X2  X are open). Let U be the open over fX1; X2g of
X . We have an exa t sequen e of hain omplexes
0 !S (X1 \ X2)j1!j2S(X1 )  S (X2)( i1!  ;i2 ) U
S (X ) !0:
That j1  j2 is inje tive is obvious, also that ( i1 ; i2 ) is surje tive is obvious.
Furthermore ( i1 ; i2) Æ (j1  j2 ) = j + j = 0, where j : X1 \ X2 ! X is the
in lusion. Finally if i1 (a1 ) + i2 (a2 ) = 0 for a1 2 Sn(X1 ), a2 2 Sn(X2), it follows
that a1 = a2 2 Sn(X1 \ X2).
As the homology of SU (X ) is equal to Hn(X ) the result follows from the long
homology sequen e asso iated to an exa t sequen e of hain omplexes. 
42
 For the general ase we need a lemma in homologi al algebra.
Lemma 7.3 (Barrat-Whitehead Lemma). Given a ommutative diagram of abelian
.

groups and homomorphisms with exa t rows

::: !?Cn+1 h !
+1
A?n
n
!
fn
B?n ! C?n h! : : :
gn n

? n+1 ? ? ? n
y y n y n y
h f g
! C n+1 ! An !
n+1 n
: : : Bn ! n
C n h! n
:::
in whi h the i are isomorphisms. There is a long exa t sequen e
( f n ;n) 1
: : : ! An n !fn An  Bn ! B n hn n!gn An 1 : : :
whi h is fun torial in su h diagrams.
Proof. The proof is by diagram hasing. We show exa tness at B n.
hn n 1 g n ( f n ; n )(a; b) = hn n 1 g n ( f n (a); n (b)) = hn n 1 gn (n (b))
as gnf n = 0. But n 1gnn = gn and hngn = 0. Thus hn n 1gn( f n; n) = 0.
Now let b 2 B n with hn n 1gn(b) = 0. There is a b1 2 Bn with gn(b1 ) = n 1gn(b).
Then gn(b) = ngn(b1 ) = gnn(b1 ). Thus gn(b n (b1)) = 0 and there is an a 2 An
su h that f n(a) = b n(b1 ). Then ( f n; n)( a; b1 ) = b n(b1 ) + n(b1 ) = b: 
Proof. (of the Mayer Vietoris sequen e). Apply the Barrat Whitehead lemma
to the diagram
: : : ! Hn(X? 1 \ X2 ) ! Hn (?X1 ) ! Hn (X1 ; ?
j1
X1 \ X2 ) ! : : :
?j ?i ?
y 2 y 1 y
! Hn (X2 ) ! Hn(X )
i2
! Hn (X; X2 ) ! :::

Redu ed homology . Sometimes in order to avoid ase distin tions it is useful
to have a slightly modied denition of the 0th homology group.
Definition 7.4 On S0(X ) repla e the boundary map  by
.
X  X
# x x = x :
x
Leave the other boundary operators the same as before. It is easy to he k that
 #  = 0. The homology of this hain omplex is alled the redu ed homology of
X , and denoted by Hn# (X ). By denition Hn# (X ) = Hn (X ) if n 6= 0, and if X is
onne ted, then H0#(X ) = 0. For A 6= ; we put Hn#(X; A) := Hn(X; A).
43
 Exer ise 7.5 If A 6= ;, the long exa t homology sequen e of (X; A) is also exa t
.

with redu ed homology.

Remark 7.6 . (1) If U 6= A and (X n U; A n U ) ! (X; A) is an ex ision, then

Hn# (X n U; A n U ) = Hn# (X; A) for all n.
(2) If X1 \ X2 is nonempty this implies that we also have the Mayer-Vietoris
sequen e for redu ed homology.

Example 7.7 A graph is a spa e whi h is a union of nitely many losed ar s

.

( alled edges) (i.e. spa es homemorphi to losed intervals), whi h have at most the
endpoints in ommon.

We want to ompute the homology of a graph G. G is homotopi ally equivalent

to an r-leaved rose Gr

(just ontra t su essively all edges whi h are not loops). So we want to ompute
the homology of Gr . G1 = S 1, and for r > 1 Gr = S 1 [ Gr 1 with S 1 \ Gr 1 = fpg.
(Gr ; Gr 1; S 1) is an exa t triad, be ause (Gr 1; p) ! (Gr ; S 1) is an ex ision. This is
by ex ising S 1 n p. If we ex ise a slightly smaller open the ex ision theorem applies,
but then (Gr 1; p) is a deformation retra t.
44
 Thus for n > 0 we get an exa t sequen e
0 = Hn(p) ! Hn(S 1)  Hn(Gr 1) ! Hn(Gr ) ! Hn# 1(p) = 0:
Thus Hn(S 1)  Hn(Gr 1) = Hn(Gr ). So by indu tion we get
8

Hn (Gr ) =
< 0 n > 1;
:Zr n = 1:

Example 7.8 The torus T = S 1  S 1 an be obtained as two annuli both alled

.

A identied along their boundary whi h is the union of two ir les C1 ; C2 (i.e. ea h
Ci is identied with Ci in the other opy of A).

The in lusions of the ir les into A are homotopy equivalen es. Thus the ex ision
theorem implies that (A; C1 t C2) ! (T; A) is an ex ision. So we get the exa t
sequen e
0 ! H2(T ) ! H1(C1 t C2)(j1! ;j2 )
H1 (A)  H1 (A) ! H1 (T ) ! H0#(C1 t C2 ) ! 0:

We see that H1(C1 t C2 ) = Z2 = H1 (A)  H1(A) and the matrix of (j1 ; j2 ) is 11 11 .
 

Thus the kernel has dimension 1 and we get H0(T ) = H2(T ) = Z, H1(T ) = Z2.

45
 8. Spheri al omplexes

We introdu e a te hnique for onstru ting topologi al spa es and develop a te h-

nique for omputing the homology groups of spa es onstru ted this way. Essentially
all reasonable spa es an be onstru ted in this way.
Definition 8.1 Let A  X be a subspa e` of a topologi al spa e. Let f : A ! Y
.

be a ontinuous map. We put X [f Y := (X Y )=  where the equivalen e relation

is generated by x being equivalent to f (x) for all x 2 A. We denote by g : X ` Y !
X [f Y the quotient map, and by f the restri tion of g to X . This spa e is given the
quotient topology, i.e. U  X [f Y is open if and only if g 1(U ) is open.
g maps Y homeopmorphi ally to a subspa e of X [f Y , whi h we identify with
Y . Then with our identi ation f jA = f .
We want to study this under some additional assumptions.
Definition 8.2 A pair (X; A) is alled a ollared pair if
.

(1) X is Hausdor,
(2) A  X is losed,
(3) points in X n A an be separated from A, i.e. for any x 2 X n A, there are
disjoint opens U; V su h that x 2 U and A  V ,
(4) A has a ollaring in X , i.e. there is an open neighbourhood B of A in X ,
su h that A is a strong deformation retra t of B (i.e. there exists a homotopy
F : B  I ! B with F (; 0) = idB , F jAI = idAI , F (; 1) : B ! A.)
Example 8.3 (E n; S n 1) is a ollared pair.
.

If f : S n 1 ! Y is a ontinuous map, we say E n [f Y is obtained from Y by

adjoining an n- ell via f .
Proposition 8.4 Let Y be a Hausdor spa e and let Z := En [f Y be obtained
.

by adjoining an n- ell. Then (Z; Y ) is a ollared pair. In fa t if B is a ollaring for

46
S n 1 , then Y [ f (B ) is a ollaring for Y . Moreover f jE n nS n 1 : En n Sn 1 ! Z n Y
is a homeomorphism.

Proof. g 1 (Z n Y ) = E n n S n 1 , whi h is open in Z . Thus Y is losed and

f : E n n S n 1 ! Z n Y is a homeomorphism.
Let B be a ollaring of S n 1 in E n. Sin e B t Y is open in E n t Y and saturated
under the equivalen e relation, we get that Y [ f (B ) is open in Z . Let D : B  I ! B
be a ontinuous map with
D(a; t) = a for all a 2 A;
D(b; 0) = b for all b 2 B;
D(b; 1) 2 A for all b 2 B:
Let D : (Y [ f (B ))  I ! Y [ f (B ) be given by
8
<z z 2 Y;
D(z; t) :=
:f (D (b; t)) z = f (b); for b 2 B:

Then D is ontinuous be ause it omes from a ontinuous map on (B t Y )  I , and

it makes Y into a strong deformation retra t of Y [ f (B ).
Let x 2 E n n S n 1. Let U; V be disjoint opens in E n su h x 2 U and S n 1  V .
Then the disjoint opens Y [ f (V ) and f (U ) separate Y from f (x).
Let x; y 2 Z . We have to see they an be separated by open sets. Then there are
three possibilities.
(1) x; y 2 Z n Y . As Z n Y is open and homeomorphi to E n n S n 1, they an be
separated by open sets.
(2) x 2 Y y 62 Y . Then this follows as Y and be separated from y.
(3) x; y 2 Y . Let Y1; Y2 be disjoint opens in Y separating x; y. Let r : B ! S n 1
be a retra tion and Bi := r 1(f 1(Yi)). Then Bi is open in E n and Y1 [ f (B1 ),
Y2 [ f (B2 ) are disjoint opens separating x and y . 
Proposition 8.5Let Z be Hausdor, let Y  Z be a losed subspa e and let
.

f: En ! Z be a map sending S n 1 to Y . Assume that f : E n n S n 1 ! Z n Y is a

homeomorphism. Let f := f jS n 1 .
Then Z = E n [f Y .

47
 Proof. f : E n ! Z and the in lusion Y ! Z dene a ontinuous map g :
En t Y ! Z. The indu ed map g : E n [f Y ! Z is bije tive. Furthermore gjE is n

losed be ause E n is ompa t, and gjY is losed. Thus g is losed. Therefore also g
is losed and so it is a homeomorphism. 
Definition 8.6 A spheri al omplex is a topologi al spa e whi h is obtained
.

from a nite set of points by su essively atta hing nitely many ells (of varying
dimensions).
Example 8.7 (Examples of spheri al omplexes)
.

(1) Let Y be a point p, then E n [f p ' S n.

(2) Let Y = S n 1, f = idS 1 , then E n [id S n 1 = E n.
n

(3) Let Y = S 1 , f : S 0 ! S 1 the onstant map to a point. Then Z is a two-leaved

rose G2 .

Repeating this pro ess we get an r-leaved rose Gr .

(4) Let Y = G2 denote the y les ,  .

dene f : S 1 ! G2 a ording to the diagram

Then E 2 [f G2 ' T = S 1  S 1.
48
 Now we want to see that real and omplex proje tive spa es are also spheri al
omplexes.
PnC and PnR are the spa es of omplex (respe tively real) one-dimensional subspa es
in C n+1 (respe tively R n+1 . Points have homogeneous oordinates (a0 : : : : : an), with
ai 2 R (respe tively C ).
Proposition 8.8
. (1) PnC is obtained from PCn 1 by atta hing a 2n- ell.
(2) PnR is obtained from PRn 1 by atta hing an n- ell.
Proof. In both ases Pn 1 an be identied with the set of (a0 : : : : : an 1 : 0)
in Pn , and the omplement is homeomorphi to C n (respe tively R n ). Dene a map
f : E 2n ! PnC (respe tively f : E n ! PnR by
(a0; : : : ; an 1) 7! (a0; : : : ; an 1; 1 k(a0; : : : ; an 1)k)
Here the ai are omplex (respe tively real) numbers. Then f : E n n S n 1 ! Pn n Pn 1
is a homeomorphism and the result follows by the proposition. 
Now we want to ompute the homology of spheri al omplexes. We will use a
very easy but often useful result from homologi al algebra.
Lemma 8.9 (ve lemma). Let
.

A?1 ! A?2 ! A?3 ! A?4 ! A?5

? ? ? ? ?
y y y y y
B1 ! B2 ! B3 ! B4 ! B5
be a ommutative diagram of abelian groups and homomorphisms in whi h the rows are
exa t and the four outer verti al maps are isomorphism. Then  is an isomorphism.
Proof. The proof is by easy diagram hasing. 
49
 Theorem 8.10 Let Z := E n [f Y be obtained by atta hing an n- ell. Then
.

f :( E n; S n 1 ) ! (Z; Y ) indu es an isomorphism f : Hk (E n ; S n 1) ! Hk (Z; Y ) for


all k.

Proof. Let B be a ollaring of S n 1 in E n, and let B be the orresponding

ollaring of Y in Z . We have a ommutative diagramm
Hk (E n?; S n 1 ) !
i
Hk (E?n ; B )
?f ?f
y1 y2
Hk (Z; Y ) j
! Hk (Z; B )
The horizontal maps are indu ed by in lusions and the verti al ones by f . We will
show that i; j; f2 are isomorphisms. This then shows that f1 is an isomorphism. The
ommutative diagram
Hk (E n n S n?1 ; B n S n 1 ) ! Hk (E?n ; B )
? ?
yf3 yf2
Hk (Z n Y; B n Y ) ! Hk (Z; B)
in whi h the horizontal maps are ex isions and f3 is indu ed by a homeomorphism,
shows that f2 is an isomorphism. To show that j (resp. i) is an isomorphism, we use
that Y (respe tively S n 1) is a deformation retra t of B (respe tively B ). Thus we
get a ommutative diagramm with exa t rows
Hk?(Y ) ! Hk?(Z ) ! Hk (Z; ? Y) ! Hk ?1(Y ) ! Hk ?1(Z )
? ? ?j ? ?
y y y y y
Hk (B ) ! Hk (Z ) ! Hk (Z; B) ! Hk 1(B ) ! Hk 1(Z )
in whi h the outer arrows are isomorphism. Thus j is an isomorphism by the ve
lemma. 
Finally we an ompute how the homology hanges under atta hing an n- ell.
Theorem 8.11 Let Z = E n [f Y be obtained from Y by atta hing an n- ell.
.

(1) Hk (Z ) = Hk (Y ) for k 6= n; n 1,
(2) Hn# 1(Z ) = Hn# 1(Y )=f(Hn 1(S n 1)),
(3) Hn#(Z ) = Hn#(Y )  ker(f : Hn 1(S n 1) ! Hn 1(Y )):

50
 Proof. We know that  : Hk (E n; S n 1) ! Hk# 1(S n 1) is an isomorphism. Using
the ommutative diagram
Hk (E n?; S n 1 ) ! Hk# 1?(S n 1 )
? ?
y yf
Hk (Z; Y ) ! Hk# 1(Y )
we an repla e Hk (Z; Y ) by Hk# 1 (S n 1 ) in the long exa t ohomology sequen e for
(Z; Y ). So we get the exa t sequen e
Hk (S n) f!

Hk (Y ) ! Hk (Z ) ! Hk# 1(S n 1 ) f! #
Hk 1(Y ) ! Hk# 1 (Z ):
(1) follows be ause Hk# 1(S n 1) = 0 unless k = n. The sequen e
 #
Hn# 1 (S n 1 ) f! Hn 1 (Y ) ! Hk# 1(Z ) ! 0
implies (2). Finally we have a sequen e
0 ! Hn#(Y ) ! Hn#(Z ) ! ker(f ) ! 0:
We know that ker(f ) is isomorphi to 0 or to Z. Therefore the sequen e splits and
Hn#(Z ) = Hn# (Y )  ker(f ). 
Corollary 8.12 If Z is a spheri al omplex, then Hk (Z ) is nitely generated
.

for all k, and if n is the highest dimension of a ell, then Hk (Z ) = 0 for k > n.
Now we an ompute the homology groups of proje tive spa es.
Corollary 8.13
.
8
<0 k > 2n or k odd,
Hk (PnC ) =
:Z k even, 0  k  2n

Proof. The proof is by indu tion on n, P0C is a point. PnC is obtained from PCn 1
by atta hing a 2n- ell. Therefore Hk (PnC ) = Hk (PCn 1 ) for k 6= n 1; n. Note that
H2n 1 (PCn 1 ) = 0. Therefore f is the zero map. Thus H2n 1 (PnC ) = H2n 1 (PCn 1 ) = 0
and H2n(PnC ) = H2n(PCn 1 )  Z = Z: 
The argument shows that if a speri al omplex Z has only ells of even dimensions,
then H2k 1(Z ) = 0 for all k and H2k (Z ) = Zr where r is the number of 2k- ells in
Z . There is a number of ases from algebrai geometry, like Grassmanians and Flag
varieties, where this happens.
51
 The ase of the real proje tive spa e is more ompli ated, be ause now we have
ells of all dimensions, therefore the atta hing maps will not automati ally indu e
the zero map in homology. So one has to study these indu ed maps, whi h is a bit
ompli ated. The result is
Theorem 8.14 8
.

>
>
>
<
0 k > n;
Hk (PnR ) = Z k = n or k even and 0 6= k  n 1;
>
>
>
:Z=2Z k odd and k  n 1:

52