MA3403 Lecture Notes
MA3403 Lecture Notes
MA3403 Lecture Notes
Lecture Notes
Fall 2018
Gereon Quick
Please send suggestions for corrections to gereon.quick@ntnu.no.
Contents
Lecture 1. Introduction 5
Lecture 2. Cell complexes and homotopy 17
Lecture 3. Singular chains and homology 39
Lecture 4. Singular homology, functoriality and H0 49
Lecture 5. Relative homology and long exact sequences 59
Lecture 6. The Eilenberg-Steenrod Axioms 71
Lecture 7. Generators for Hn (S n ) and first applications 79
Lecture 8. Calculating degrees 89
Lecture 9. Local vs global degrees 95
Lecture 10. Homotopies of chain complexes 99
Lecture 11. Homotopy invariance of singular homology 105
Lecture 12. Locality and the Mayer-Vietoris sequence 115
Lecture 13. Cell complexes 123
Lecture 14. Homology of cell complexes 133
Lecture 15. Computations of cell homologies and Euler characteristic 141
Lecture 16. Designing homology groups and homology with coefficients 151
Lecture 17. Tensor products, Tor and the Universal Coefficient Theorem 161
Lecture 18. Singular cohomology 173
Lecture 19. Ext and the Universal Coefficient Theorem for cohomology 185
Lecture 20. Cup products in cohomology 199
Lecture 21. Applications of cup products in cohomology 211
3
4 CHAPTER 0. CONTENTS
Lecture 22. Poincaré duality and intersection form 221
Lecture 23. Classification of surfaces 229
Lecture 24. More on Poincaré duality 239
Bibliography 249
Appendix A. Exercises 251
1. Exercises after Lecture 2 251
2. Exercises after Lecture 6 252
3. Exercises after Lecture 7 254
4. Exercises after Lecture 9 256
5. Exercises on the Hurewicz map 259
6. Exercises after Lecture 12 264
7. Exercises after Lecture 13 267
8. Exercises after Lecture 15 270
9. Exercises after Lecture 17 272
10. Exercises after Lecture 21 275
LECTURE 1
Introduction
Organization:
Lectures: Tuesdays and Thursdays, both days at 12.15-14.00 in S21 in Sen-
tralbygg 2.
Exercises: We will have some suggested exercises, hopefully on a weekly
basis. You should try to solve as many exercises as possible, not just the ones
I suggest, but also all that you find in other textbooks. We will not have an
exercise class though. But you can discuss exercises with me at any time!
Course webpage: wiki.math.ntnu.no/ma3403/2018h/start
where all news about the class will be announced. You will also find lecture notes
a few hours after class on the webpage.
Office hours: Upon request.
Just send me an email: gereon.quick@ntnu.no
Textbooks: We will not follow just one book... but there are many good
texts out there. For example, you can look at
[H] A. Hatcher, Algebraic Topology. It’s available online for free. It contains
much more than we have time for during one semester.
[Mu] J.R. Munkres, Elements of Algebraic Topology.
[V] J.W. Vick, Homology Theory - An Introduction to Algebraic Topology.
Two books that you can use as an outlook to future topics:
[Ma] J.P. May, A Concise Course in Algebraic Topology. It’s also online
somewhere.
[MS] J.W. Milnor, J. Stasheff, Characteristic Classes.
5
6 CHAPTER 1. INTRODUCTION
There are many other good books and lecture notes out there. Ask me if you
need more.
What is required?
I will assume that you are starting your third year at NTNU (or more). You
should have taken the equivalent of Calculus 1-3 or MA1101-1103, MA 1201-
1202. So you should be familiar with Euclidean space Rn , multivariable calculus
and linear algebra. Ideally, you have taken TMA4190 Introduction to Topology
and/or General Topology.
You should also know a bit about algebra, like what is a group, an abelian
group, a field, ideally also what is a ring and module over a ring.
Finally, it would be good if you knew what a topoogical space is and you
would know what the words open, closed, compact, etc mean. But, in fact,
you could also just have some few examples of topologial spaces in mind, like
n-spheres, torus etc. without knowing too many abstract stuff. For, the class is
much more about the ideas and methods we develop than anything else. And
these methods are useful almost everywhere.
Nevertheless, if you want to refresh your knowledge on Topology, you may
want to have a look at the book
[J] K. Jänich, Topology.
Remark: We will later see that all the invariants we construct are preserved
under homotopy equivalences, a weaker notion than homeomorphisms. This
will finally lead to the idea of the stable homotopy category being the motive of
topological spaces. We will not discuss this in class, but feel free to ask me about
it. :)
For example, the first important tool that we are going to define soon is
singular homology. It will allow us to use a simple algebraic argument to show
≈
that there cannot be a homeomorphism Rm − → Rn .
Just to make you taste a little more of what algebraic topology can do:
Multiplicative Structures on Rn
Let Rn × Rn → Rn be a bilinear map with two-sided identity element e ̸= 0
and no zero-divisors. Then n = 1, 2, 4, or 8.
What we are looking for is a ”multiplication map”. You know the cases
n = 1 and n = 2 very well. It’s just R and C ∼ = R2 . These are actually
fields.
For n = 4, there are the Hamiltonians, or Quaternions, H ∼ = R4 with a
multiplication which as almost as good as the one in C and R, but it is
not commutative. (You add elements i, j, k to R with certain multiplication
rules.)
For n = 8, there are the Octonions O ∼ = R8 . The multiplication is not
associative and not commutative.
And that’s it!
• Before we move on, let us play a game and see an invariant in action.
The rules: Take a piece of paper and draw two crosses, i.e. spots with four
free ends.
Each move involves joining two free ends with a curve which does not cross
any existing line, and then putting a short stroke across the line to create two
new free ends. The players play alternating moves.
If there are no legal moves left, the player who made the last legal move
wins.
Let us assume that we know that the game ends after a finite number of
moves, say m moves. At the end we will have created a connected, closed planar
graph, in particular, a figure which has vertices, edges and faces.
10 CHAPTER 1. INTRODUCTION
We claim that no matter how you play, what strategy you use etc, there are
always 8 moves before the game stops, it is always the second player who wins
and there is a fixed number of vertices, edges and faces!
Why? Well, the number of moves and everything about the figure we create
is determined by Euler’s formula v − e + f = 2. The number 2 is an example
of an algebraic invariant.
To understand how this works, we need to determine how the number v of
vertices, the number e of edges, and the number f of faces depends on the number
m of moves.
For the vertices, when we start the game we have two vertices. In each move,
we create one new vertex. Thus we get
v = 2 + m.
For the edges, when we start the game we have no edges. In each move, we
create one line, but we split it into two edges by adding a vertex in the middle.
Hence in each move, we create 2 edges. Thus we get
e = 2m.
11
For the faces, we have to think backwards. At the end, there is exactly one
free (or loose) end pointing into each face that we created. For, if there was a
face with two free ends pointing into one face, then we could connect these two
ends within that face and the game would not have stopped. Note that there is
also exactly one loose end pointing out of the figure. (Again, of there were two
we could connect them by going around the figure.)
Now we need to check how many free ends we produce. We start with 4 free
ends per cross, that is 8 free ends. In each move, we connect two free ends, but we
also create two new ones. Thus the number of free ends does not change during
the whole game. Hence we get
f = 8.
In total we get
2=v−e+f
2 = 2 + m − 2m + 8
0 = −m + 8
m = 8.
Hence no matter how we play, the game ends after 8 moves. Since this number
is even, the second player always wins. Moreover, we always get v = 10, e = 16,
and f = 8.
Alternative: Changing the starting setup changes the outcome of the game.
For if we start with n crosses (or nodes), then we get with the same reasoning as
above
v =n+m
e = 2m
f = 4n.
Euler’s formula then yields
2=v−e+f
2 = n + m − 2m + 4n
m = 5n − 2.
Thus the game ends after m = 5n − 2 moves. For example, if n = 3, this is an
odd number and the first player always wins.
12 CHAPTER 1. INTRODUCTION
Here is the idea why Euler’s formula holds: We can draw any connected
planar graph (a graph we can draw in the plane such that its edges only intersect
in the vertices and we can walk along the edges between any two vertices) as
follows:
1) We start with a graph consisting of just one vertex and no edges, so
v = 1 and e = 0. And we have one face, the outer face or the plane around
the vertex, so f = 1. So in total the formula holds v−e+f = 1−0+1 = 2.
2) Now we can extend the graph by either
a) adding one vertex and connect it via an edge to the first one; that
is we change v → v + 1 and e → e + 1 or
b) draw an edge from the existing vertex to itself; this way we create
a new face as well, hence we change e → e + 1 and f → f + 1.
Thus after both operations the formula v − e + f = 2 still holds. Now
we continue this process until we have created the planar graph we had
in mind.
f = P + H.
5P + 6H
e= .
2
13
v−e+f =2
5P + 6H 5P + 6H
− +P +H =2 (multiply by 6)
3 2
10P + 12H − 15P − 18H + 6P + 6H = 2 (simplify)
P = 12.
To get H, we count how many hexagons there are per pentagon: Each penta-
gon is surrounded by 5 hexagons which would yield H = 5P . But each hexagon
is attached to 3 pentagons at the same time. Hence we have counted three times
as many hexagons as there really are. This yields
5P 5 · 12 60
H= = = = 20.
3 3 3
• •
3 • 2 P = # pentagons = ?
•
• H = # hexagons = ?
1 • • •
4 2
• • 1 • • f-e+v=2
• 5 • • •
f= P+H
• • 3
• • e = ( 5P + 6H )/2
•
v = ( 5P + 6H )/3
P must be 12
H = ( 5P )/3
H must be 20
14 CHAPTER 1. INTRODUCTION
The Euler characteristic:
The number v − e + f = 2 is an example of the Euler characteristic of a
surface, i.e., a two-dimensional manifold. In fact, 2 is the Euler characteristic
of the sphere. The Euler characteristic can be defined for any topological space
X. It is denoted by χ(X) and it is always an integer number. More generally,
χ(X) is a topological invariant which means it does not change if we transform
X continuously.
Here are some examples for surfaces:
• For a sphere X = S 2 , it is 2: χ(S 2 ) = 2.
• For a torus X = T 2 , it is 0: χ(T 2 ) = 0.
• For a surface with two holes, it is −2.
• In general, for a surface with g holes, it is 2 − 2g.
Now assume we have two spaces X and Y , defined in some complicated way
which makes it difficult to understand how they look. But let us assume we can
calculate their Euler characteristics by some method. Then, if χ(X) ̸= χ(Y ), we
know that we cannot transform X continuously into Y .
And there are also more positive examples. It often happens that an invariant
defined one way turns out to encode a lot of other information as well.
You will learn more about these things soon...
v-e+f=2 v-e+f=0
sphere torus
Open sets in Rn
• Let x be a point in Rn and r > 0 a real number. The ball
Br (x) = {y ∈ Rn : |x − y| < r}
with radius ϵ around x is an open set in Rn .
• The open balls Br (x) are the prototypes of open sets in Rn .
• A subset U ⊆ Rn is called open if for every point x ∈ U there
exists a real number ϵ > 0 such that Bϵ (x) is contained in U .
• A subset Z ⊆ Rn is called closed if its complement Rn \ Z is open
in Rn .
17
18 CHAPTER 2. CELL COMPLEXES AND HOMOTOPY
• Familiar examples of open sets in R are open intervals, e.g. (0,1) etc.
• The cartesian product of n open intervals (an open rectangle) is open in
Rn .
• Similarly, closed intervals are examples of closed sets in R.
• The cartesian product of n closed intervals (a closed rectangle) is closed
in Rn .
• The empty set ∅ and Rn itself are by both open and closed sets.
• Not every subset of Rn is open or closed. There are a lot of subsets
which are neither open nor closed. For example, the interval (0,1] in R;
the product of an open and a closed interval in R2 .
TRn = {U ⊆ Rn open}
Here are some examples of topological spaces which also demonstrate that
some topologies are more interesting than others:
• Rn with TRn as described above.
• An arbitrary set X with the discrete topology TX = P(X), where
P h(X) is the power set of X, i.e., the set of all subsets of X. In the
discrete topology, all subsets are open and hence all subsets are also
closed.
• On an arbitrary set X, there is always the coarse topology TX =
{∅, X}.
• Let (X, d) be a metric space. Then we can imitate the construction of
the standard topology on Rn and define the induced topology as the
set of all U ⊆ X such that for each x ∈ U there exists an r > 0 so that
B(x, r) ⊆ U . Here B(x, r) = {y ∈ X : d(x, y) < r} is the metric ball of
radius r centred at x.
• Let (X,TX ) be a topological space, let Y ⊂ X be an arbitrary subset.
The induced topology or subspace topology of Y is defined by
Warning
It is important to note that that the property of being an open subset
really depends on the bigger space we are looking at. Hence open always
refers to being open in some given space.
For example, a set can be open in a space X ⊂ R2 , but not be open in R2 ,
see the picture.
Open sets are nice for a lot of reasons. First of all, they provide us with a
way to talk about things that happen close to a point.
The type of maps that preserve open sets are the continuous maps:
21
Definition: Homeomorphisms
A continuous map f : X → Y is a homeomorphism if it is one-to-one and
onto, and its inverse f −1 is continuous as well. Homeomorphisms preserve
the topology in the sense that U ⊂ X is open in X if and only if f (U ) ⊂ Y
is open in Y .
Homeomorphisms are the isomorphisms in the category of topological
spaces.
Definition: Compactness
Let X be a topological space. A subset Z ⊂ X S is called compact if for
any collection {Ui }i∈I , Ui ⊂ X open, with Z ⊂ i∈I Ui there exist finitely
many i1 , . . . , in ∈ I such that Z ⊂ Ui1 ∪ · · · ∪ Uin .
In other words, a subset Z in a topological space is compact iff every open
cover {Ui }i of Z has a finite subcover.
Definition: Connectedness
A topological space X is called connected if it is not possible to split it
into the union of two non-empty, disjoint subsets which are both open and
closed at the same time.
In other words, a space is connected if and only if the empty set and the
whole space are the only subsets which are both open and closed.
Note that the image f (X) of a connected space X under a continuous map
f : X → Y is again connected.
Let us prove the first assertion. The other one is left as a little exercise.
Proof: Let {Ui }i∈I be an open cover of Z. We set U := X \ Z. Then
{U, Ui }i∈I is an open cover of X. Since X is compact, there exist i1 , . . . ,in such
that X ⊂ U ∪ Ui1 ∪ . . . ∪ Uin and hence, by the definition of U , we have Z ⊂
Ui1 ∪ . . . ∪ Uin . QED
Another useful fact:
Torus
We start with the square
S := {(x,y) ∈ R2 : 0 ≤ x ≤ 1, 0 ≤ y ≤ 1} ⊂ R2
with the subspace topology induced from the topology of R2 . Now we would
like to glue opposite sides to each. This corresponds to taking the quotient
T := S/((x,0) ∼ (x,1) and (0,y) ∼ (1,y)).
Sphere as a quotient
For every n ≥ 1, there is a homeomorphism
≈
ρ̄ : Dn /∂Dn −
→ S n.
There are in fact many different ways to construct such a homeomorphism.
Let us write down one in concrete terms for the special case n = 2. The
general case follows by throwing in more coordinates.
We define a continuous map ρ : D2 → S 2 such that
(
ρ(0,0) = (0,0, − 1) and
ρ(x,y) = (0,0, + 1) for all (x,y) ∈ ∂D2 = S 1 .
Since ρ will be constant on ∂Dn , it will induce a map ρ̄ on the quotient
D2 /∂D2 .
We define ρ as a rotation invariant map which sends the inner part of D2 of
points with radius less than 1/2 mapping onto the lower hemisphere of S 2
and the outer part of D2 of points with radius greater than 1/2 mapping
29
• Compactifications
The concrete maps we wrote down in the previous example are kind of ugly.
But there is another way to show that there is such a homeomorphism Dn /∂Dn ≈
S n.
For we can also consider S n as the one-point compactification of Rn . Let
us first say what that means:
30 CHAPTER 2. CELL COMPLEXES AND HOMOTOPY
• Cell complexes
Another way to think of the above procedure is the following. The sphere
consists of two parts that we glue together:
• an open n-disk, i.e., the open interior Dn \ ∂Dn ,
• and a single point, which corresponds to the class of the boundary ∂Dn ;
on S 2 we can picture this point as the northpole (the light blue dot in
the above picture).
Topologists think of such building blocks as the cells of a space. However,
not all spaces can be built this way. So let us make precise what is needed:
Note that there is a direct way to define the Euler characteristic of cell
complexes. We will later see the reason why this is the correct definition using
homology. Right now we can already check at the example of a tetrahedron that
this definition agrees with Euler’s formula we saw in the first lecture.
34 CHAPTER 2. CELL COMPLEXES AND HOMOTOPY
χ(T ) = 1 − 2 + 1 = 0.
To compare this definition with Euler’s formula we used in the first lecture,
let us look at the tetrahedron which is also a cell complex:
• Homotopy
Homotopy is a fundamental notion in topology. Let us start with a definition
and then try to make sense of this.
35
Definition: Homotopies
Let f0 ,f1 : X → Y be two continuous maps. Then f0 and f1 are called
homotopic, denoted f0 ≃ f1 , if there is a continuous map h : X ×[0,1] → Y
such that, for all x ∈ X,
h(x,0) = f0 (x), and h(x,1) = f1 (x).
Homotopy defines an equivalence relation (exercise!) on the set of continu-
ous maps from X to Y . The set of equivalence classes of continuous maps
from X to Y modulo homotopy is denoted by [X,Y ].
defines a homotopy between the identity map on Rn and the constant map Rn →
{0} ⊂ Rn to the one-point space consisting of the origin. For the same reason,
the n-disk Dn is contractible.
However, it is not always obvious which spaces are homotopy equivalent to
each other. So it will be useful to develop some intuition for homotopy equiva-
lences. There is a particular type that is easier to spot:
For a deformation retraction, one can think of the homotopy h as a map which
during the time from 0 to 1 pulls back all the points of X into the subspace A,
and leaves the whole time the points in A fixed. Here are some examples:
• The origin {0} is a strong deformation retract fo Rn and of the n-disk
Dn .
• For any topological space Y , the product Y ×{0} is a strong deformation
retract of Y × Rn and Y × Dn . For example, the circle S 1 × {0} is a
strong deformation retract of the solid torus S 1 × D2 .
• The n-sphere S n is a strong deformation retract of the punctured disk
Dn+1 \ {0} and also of Rn+1 \ {0}.
37
Why homotopy?
The simplest reason why we consider the homotopy relation is that it
works. It is fine enough such that all the tools that we are going to
define are invariant under homotopy, i.e., they are constant on equivalence
classes. But it is also coarse enough that it identifies enough things such
that many problems become simpler and in fact solvable.
With respect to first point, one can consider the homotopy category
hoTop of spaces, i.e., the category whose objects are topological spaces
and whose sets of morphisms from X to Y are the sets of homotopy classes
of maps [X,Y ], satisfies a universal property for invariants.
With respect to the second point, we just indicate that life in hoTop is
much easier because there are much fewer morphisms. For example,
there are many and complicated continuous maps S 1 → C \ {0}. But there
are very few homotopy classes of such maps, since [S 1 ,C \ {0}] = Z, up to
homotopy a map S 1 → C \ {0} is determined by the winding number, i.e.,
the number of times it goes around the origin.
To convince ourselves that homotopy actually works, we remark that homo-
topy is even fine enough to detect diffeomorphism classes between smooth
manifolds and helped for example to classify manifolds up to bordism. But
this is a story we save for a future lecture/class.
If you are still not convinced, then let us remark that to study things up-
to-homotopy is so useful that mathematicians work hard to find analogs
of the homotopy relation and the homotopy category in many different
areas. If you want to learn more about this, have a look at Quillen’s highly
influential book on Homotopical Algebra. You will also see an example
in homological algebra where one talks about homotopies between chain
complexes.
LECTURE 3
The standard simplices are related by face maps for 0 ≤ i ≤ n which can be
described as
ϕni (t0 , . . . ,tn−1 ) = (t0 , . . . ,ti−1 ,0,ti , . . . ,tn−1 )
with the 0 inserted at the ith coordinate (t0 is the 0th coordinate).
39
40 CHAPTER 3. SINGULAR CHAINS AND HOMOLOGY
Using the standard basis, ϕni can be described as the affine linear map (a
translation plus a linear map)
(
ej j<i
ϕni : ∆n−1 ,→ ∆n determined by ϕni (ej ) =
ej+1 j ≥ i.
A short way of expressing the above formula for ϕni is that it embedds ∆n−1
into ∆n by omitting the ith vertex (that is what the hat in the following formula
means):
ϕni = [e0 , . . . ,ei−1 ,b
ei ,ei+1 , . . . ,en ] : ∆n−1 → ∆n .
Definiton: Faces
Note that ϕni maps ∆n−1 onto the subsimplex opposite to the ith corner, or
in the standard basis, opposite to ei . We call the image of ϕni the ith face
of ∆n (which is opposite to ei ).
Note that the union of the images of all the face inclusions is the boundary
of ∆n .
The face maps satisfy a useful identity, sometimes called simplicial identity:
But, so far, Sing0 (X) is just a set and we are not allowed to add or subtract
elements. We are now going to remedy this defect, since algebraic operations
make life much easier. Therefore, we formally extend Singn (X) into an abelian
group.
The general way to turn a set B into an abelian group, is to form the associated
free abelian group. The idea is to add the minimal amount of structure
and relations to turn B into an abelian group. Since this is an important
construction, we recall how this works:
Note: If n < 0, Singn (X) is defined to be empty and Sn (X) is the trivial
abelian group {0}. So whenever we talk about n-chains, n will be assumed to be
nonnegative.
Definition: Cycles
An n-cycle in X is an n-chain c ∈ Sn (X) with ∂n c = 0. We denote the
group of n-cycles by
Zn (X) := Ker (∂n : Sn (X) → Sn−1 (X))
= {c ∈ Sn (X) : ∂n (c) = 0} ⊆ Sn (X).
Note that the group of 0-cycles is all of S0 (X), since every 0-chain is mapped
to 0:
Z0 (X) = S0 (X).
As the notation suggests, we are going to think of a chain of the form ∂(c) as
the boundary of c:
45
Definition: Boundaries
An n-dimensional boundary in X is an n-chain c ∈ Sn (X) such that
there exists an (n + 1)-chain b with ∂n+1 b = c. We denote the group of
n-boundaries by
Bn (X) := Im (∂n+1 : Sn+1 (X) → Sn (X))
= {c ∈ Sn (X) : there is a b ∈ Sn+1 (X) with ∂n+1 (b) = c}.
In general, we can think of the signs as giving the faces of the simplices
an orientation. And if an n-simplex is a face of an (n + 1)-simplex, then it
inherits an induced orientation which is determined by how it fits into the
bigger simplex. Going down two steps of inherited signs means things
cancel out.
However, thinking of signs as orientations is formally not correct as we will
notice in an example below. But, as we will see soon, we can algebraically
remedy this defect.
This basic resut shows that the sequence {Sn (X), ∂n }n has an important prop-
erty:
Hence we have shown that we obtain for any topological space X a complex
of (free) abelian groups
∂ n ∂ ∂n−1
2 1∂ 0 ∂ ∂
··· →
− Sn (X) −→ Sn−1 (X) −−−→ · · · −
→ S1 (X) −
→ S0 (X) −
→ 0.
47
It is called the singular chain complex of X. We will see next lecture what
such chain complexes are good for.
LECTURE 4
Recall that we constructed, for any topological space X, the singular chain
complex of X
∂ n∂ ∂n−1 2∂ 1 ∂ 0 ∂
··· →
− Sn (X) −→ Sn−1 (X) −−−→ · · · −
→ S1 (X) −
→ S0 (X) −
→ 0.
The homomorphisms ∂n satisfy the fundamental rule: ∂ ◦ ∂ = 0.
The following definition of homology groups applies to any chain complex.
However we formulate it only for the singular chain complex:
Singular what?
In algebra, homology is a way to measure the difference between cycles
and boundaries. Singular homology is an application of homology in order
to understand the structure of a space.
Given a space X, the group of n-cycles measures how often we can map
an n-dimensional simplex into X without collapsing it to any of its n − 1-
dimensional faces.
Let σ(∆n ) be the image in X of such a cycle. If we can even map an (n + 1)-
dimensional simplex σ ′ (∆n+1 ) into X whose boundary is σ(∆n ), then we can
continuously collapse all of σ(∆n ) to a point. In this case, we would like to
49
50 CHAPTER 4. SINGULAR HOMOLOGY, FUNCTORIALITY AND H0
forget about this σ. For, from an n-dimensional point of view, this σ(∆n ) is
not interesting. That is what it means geometrically/topologically to take
the quotient by Bn (X).
But if we cannot find an n + 1-dimensional simplex such that σ(∆n ) is
its boundary, then σ(∆n ) potentially carries interesting n-dimensional
information about X.
The slogan is: Hn (X) measures n-dimensional wholes in X.
Orientations revisited
Let X be some space, and suppose we have a one-simplex σ : ∆1 → X.
Define
ϕ : ∆1 → ∆1 , (t,1 − t) 7→ (1 − t,t).
Precomposing with ϕ gives another singular simplex σ̄ = σ◦ϕ which reverses
the orientation of σ. It is not true that σ̄ = −σ in S1 (X).
However, we claim that
σ̄ ≡ −σ mod B1 (X).
This means that there is a 2-chain in X whose boundary is σ̄ + σ. If d0 (σ) =
d1 (σ) such that σ ∈ Z1 (X) is a 1-cycle, then σ̄ and σ are homologous and
[σ̄] = [σ] in H1 (X).
To prove the claim we need to construct an appropriate 2-chain. Let
π : ∆2 → ∆1 be the affine map determined by sending e0 and e2 to e0
51
Now we calculate
∂(σ ◦ π) = σ ◦ π ◦ ϕ20 − σ ◦ π ◦ ϕ21 + σ ◦ π ◦ ϕ22 = σ̄ − κ1σ(0) + σ.
Hence up to the term −κ1σ(0) we get what we want. So we would like to elim-
inate this term. To do that we define the constant 2-simplex κ2σ(0) : ∆2 → X
at σ(0). Its boundary is
∂(κ2σ(0) ) = κ1σ(0) − κ1σ(0) + κ1σ(0) = κ1σ(0) .
Thus
σ̄ + σ = ∂(π ◦ σ + κ2σ(0) )
which proves the claim.
and
(
Z · σn n odd or
Bn (X) =
0 n even.
The induced map is compatible with the boundary operator in the following
way:
53
Proof: We just calculate and check that both ways have the same outcome
for any singular n-simplex σ on X:
n
X
∂Y (Sn (f ))(σ) = (−1)i (f ◦ σ) ◦ ϕni
i=0
Xn
= (−1)i f ◦ (σ ◦ ϕni )
i=0
n
!
X
i
= Sn−1 (f ) (−1) σ ◦ ϕni
i=0
= Sn−1 (f )(∂X σ).
□
The lemma has the important consequence that
Sn (f )(Zn (X)) ⊂ Zn (Y ) and Sn (f )(Bn (X)) ⊂ Bn (Y ).
To summarize our observations: Sn (−) and Hn (−) are functors from the
category of topological spaces to the category of abelian groups. For the sequence
of all Sn (−) even more is true: S∗ (−) is a functor from the category of topological
spaces to the category of chain complexes of abelian groups (with chain maps as
morphisms).
Invariance
As a consequence, if f : X → Y is a homeomorphism, then Hn (f ) is an
isomorphism of abelian groups. In other words, homology groups only
depend on the topology of a space.
In fact, we will soon see that homology is a coarser invariant in the sense that
homotopic maps induce the same map in homology.
• The homology group H0
Let us try to understand the simplest of the homology groups.
Lemma: Augmentation
For any topological space X, there is a homomorphism
ϵ : H0 (X) → Z
which is nontrivial whenever X ̸= ∅.
ϵ : H0 (X) → H0 (pt) = Z.
□
Let us try to understand this ϵ a bit better. The map X → pt induces a
homomorphism of chain complexes S∗ (X) → S∗ (pt) which sends any 0-simplex
55
and γj ◦ ϕ11 are 0-simplices and are sent to 1 by ϵ̃. Thus we get
X X X
ϵ(b) = ϵ(∂c) = ϵ( aj (γj ◦ ϕ10 − γj ◦ ϕ11 )) = aj − aj = 0.
j j j
(where we identify 0-simplices and their image points). Hence the 0-simplices κ0x
and κ0y are homologous. Since 0-simplices generate H0 (X) and ϵ is a homomor-
phism, this implies that ϵ is injective. QED
In other words, H0 (X) is the free abelian group generated by the set of
path-components of X.
Since singular n-chains are freely generated by n-simplices, this shows that
the singular chain complex of X splits into a direct sum
M
S∗ (X) = S∗ (Xi ).
i∈I
For the same reason the boundary operators
∂ : Sn (X) ∼ Sn−1 (Xi ) ∼
M M
= Sn (Xi ) → = Sn−1 (X)
i∈I i∈I
split into components ∂Xi : Sn (Xi ) → Sn−1 (Xi ). Hence we get an isomorphism.
The statement for n = 0 then follows from the previous result on path-connected
spaces. □
LECTURE 5
the set of continuous maps which respect the subspaces. In fact, we get a category
Top2 of pairs of topological spaces.
The induced map Sn (A) → Sn (X) is injective. Hence we are going to identify
Sn (A) with its image in Sn (X) and consider Sn (A) as a subgroup of Sn (X).
59
60 CHAPTER 5. RELATIVE HOMOLOGY AND LONG EXACT SEQUENCES
Since the boundary operator is defined via composition with the face maps,
it satisfies
∂(Sn (A)) ⊂ Sn−1 (A) ⊂ Sn−1 (X).
For, if the image of σ : ∆n → X lies in A, then so does the image of the composite
∆n−1 ,→ ∆n → X.
Thus ∂ induces a homomorphism ∂¯ on Sn (X,A) and we have a commuta-
tive diagram
Sn (X) / Sn (X,A)
∂ ∂¯
Sn−1 (X) / Sn−1 (X,A).
Since ∂ ◦ ∂ = 0 and since Sn (X) → Sn (X,A) is surjective, we also have
∂¯ ◦ ∂¯ = 0.
Now we observe that Zn (X,A) = Zn′ (X,A)/Sn (A) (since Sn (X,A) is Sn (X)/Sn (A))
and Bn (X,A) = Bn′ (X,A)/Sn (A). Hence we get
Zn (X,A) Z ′ (X,A)/Sn (A) Z ′ (X,A)
Hn (X,A) = = n′ = n′ .
Bn (X,A) Bn (X,A)/Sn (A) Bn (X,A)
In other words, we could also have used the latter quotient to define Hn (X,A).
Empty subspaces
As a special case with A = ∅ we get
Zn′ (X,∅) = Zn (X), Bn′ (X,∅) = Bn (X), and Hn (X,∅) = Hn (X).
Now let us have a look at two examples to see how the images of simplices
in Hn (X) and Hn (X,A) can differ.
62 CHAPTER 5. RELATIVE HOMOLOGY AND LONG EXACT SEQUENCES
However, each of these singular simplices lies in ∂∆n , and hence ∂(ιn ) ∈
Sn−1 (∂∆n ).
Thus the class ῑn ∈ Sn (∆n ,∂∆n ) is a relative cycle. We will see later
that the relative homology group Hn (∆n ,∂∆n ) is an infinite cyclic group
generated by [ῑn ].
64 CHAPTER 5. RELATIVE HOMOLOGY AND LONG EXACT SEQUENCES
• Long exact sequences
Back to the general case. So let (X,A) be a pair of spaces. We know that
the inclusion map i : A ,→ X induces a homomorphism Hn (i) : Hn (A) → Hn (X).
Moreover, the map of pairs j : (X,∅) → (X,A) induces a homomorphism
Hn (j) : Hn (X) ∼
= Hn (X,∅) → Hn (X,A).
Connectng homomorphism
For all n, there is a connecting homomorphism, which is often also called
boundary operator and therefore usually also denoted by ∂,
∂ : Hn (X,A) → Hn−1 (A), [c] 7→ [∂(c)]
with c ∈ Zn′ (X,A).
Let us try to make sense of this definition: We just learned that we can
represent an element in Hn (X,A) by an element c ∈ Zn′ (X,A). Then ∂(c) is an
element in Sn−1 (A). In fact, ∂(c) is a cycle, since it is a boundary and therefore
∂(∂(c)) = 0.
In particular, ∂(c) represents a class in the homology Hn−1 (A). Hence we can
send [c] under the connecting homomorphism to be the class [∂c] ∈ Hn−1 (A).
It remains to check that this is well-defined, i.e., if we choose another
representative for the class [c] we need to show that we obtain the same class
[∂(c)].
Another representative of [c] in Zn′ (X,A) has the form c + ∂(b) + a with
b ∈ Sn+1 (X) and a ∈ Sn (A). Then we get
∂(c + ∂(b) + a) = ∂(c) + ∂(a).
But, since ∂(a) ∈ Bn−1 (A), we get
[∂(c)] = [∂(c) + ∂(a)] in Hn−1 (A).
Thus, the connecting map is well-defined. And it is a homomorphism, since ∂
is a homomorphism.
Hence we get a sequence of homomorphisms
Hn (i) Hn (j) ∂
Hn (A) −−−→ Hn (X) −−−→ Hn (X,A) →
− Hn−1 (A).
65
In fact, the existence of the connecting map, the above sequence and its
properties can be deduced by a purely algebraic process, that we will recall below.
For, the relative chain complex fits into the short exact sequence of chain
complexes
0 → S∗ (A) → S∗ (X) → S∗ (X,A) → 0.
Such a sequence induces a long exact sequence in homology of the form
Hn+1 (i)
··· / Hn+1 (X,A)
∂
t Hn (j)
Hn (A) / Hn (X) / Hn (X,A)
Hn (i)
∂
t
Hn−1 (A) / ···
Hn−1 (i)
But we assumed ∂B (b) = 0. Thus fn−1 (∂A (a)) = 0. Since fn−1 is injective,
this implies ∂A (a) = 0. Hence a is indeed a cycle, and therefore represents a
homology class [a] ∈ Hn (A∗ ). It also follows
Hn (f )([a]) = [b − ∂B (b̄)] = [b].
67
t Hn (g∗ )
Hn (A∗ ) / Hn (B∗ ) / Hn (C∗ )
Hn (f∗ )
∂
t
Hn−1 (A∗ ) / ···
Hn−1 (f∗ )
is exact.
Furthermore, there is the famous Five Lemma (here in one of its variations):
Five Lemma
Suppose we have a commutative diagram
A1 / A2 / A3 / A4 / A5
f1 f2 f3 f4 f5
B1 / B2 / B3 / B4 / B5
with exact rows. Then
• If f2 and f4 are surjective and f5 injective, then f3 is surjective.
• If f2 and f4 are injective and f1 surjective, then f3 is injective.
In particular, if f1 , f2 , f4 , and f5 are isomorphisms, then f3 is an isomor-
phism.
69
The proof is another diagram chase and left as an exercise. You should defi-
nitely do it, it’s fun!
Here we rather state two consequences:
• Given a map of short exact sequences
0 / A′ / A / A′′ / 0
f′ f f ′′
0 / B′ / B / B ′′ / 0
in which f ′ and f ′′ are isomorphisms. Then f is an isomorphism.
• Back in topology, let f : (X,A) → (Y,B) be a map of pairs of spaces. If
any two of A → B, X → Y and (X,A) → (Y,B) induce isomorphisms,
then so does the third. This observation will simplify our life a lot.
LECTURE 6
For A ⊂ X, we call
[
A◦ = U with U open in X
U ⊂A
the closure of A.
Eilenberg-Steenrod Axioms
A homology theory (for topological spaces) h consists of:
• a sequence of functors hn : Top2 → Ab for all n ∈ Z and
• a sequence of functorial connecting homomorphisms
∂ : hn (X,A) → hn−1 (A,∅) =: hn−1 (A)
which satisfy the following properties:
71
72 CHAPTER 6. THE EILENBERG-STEENROD AXIOMS
We have already shown that singular homology satisfies the dimension axiom
and the connectiong homomorphism fits into long exact sequences. It remains to
check homotopy invariance and excision. But before we do that we will assume
these properties for a moment and use them to compute some homology groups.
First an important consequence of the homotopy axiom:
H0 (i)
H0 (S 0 ) / H0 (D1 ) / H0 (D1 ,S 0 ) / 0
Z⊕Z / Z / ? / 0.
This implies that H0 (i) is surjective. Since the above sequence is exact, this
implies
H0 (D1 ,S 0 ) = 0.
Z / Z / ? / 0.
Since both S n−1 and Dn are path-connected, their 0th homology is isomorphic
to Z and the generator of H0 (S n−1 ), the class of any constant map κ0x : ∆0 → S n−1 ,
is sent to the generator of H0 (Dn ), the class of κ0x : ∆0 → Dn corresponding to
the image point x ∈ S n−1 ⊂ Dn . Hence H0 (i) is surjective and we have again
H0 (Dn , S n−1 ) = 0.
This finishes the argument for H0 .
For q = 1, we start with the exact sequence
H0 (i)
(5) H1 (D1 ) / H1 (D1 , S 0 ) / H0 (S 0 ) / H0 (D1 )
0 / ? / Z⊕Z / Z.
Since the sequence is exact, this shows that H1 (D1 ,S 0 ) is isomorphic to the kernel
of
Hn (i) : Z ⊕ Z → Z, (a,b) 7→ a + b.
Thus
(6) H1 (D1 , S 0 ) ∼
= Z.
∼
0 / ? / Z
= / Z.
Since the right most map is an isomorphism, we get
(7) H1 (Dn ,S n−1 ) = 0 for all n ≥ 2.
76 CHAPTER 6. THE EILENBERG-STEENROD AXIOMS
In order to study further groups, we consider the subspaces
n
D+ := {(x0 , . . . ,xn ) ∈ S n : x0 ≥ 0} and D−
n
:= {(x0 , . . . ,xn ) ∈ S n : x0 ≤ 0}
which correspond to the upper and lower hemisphere (including the equator),
respectively, of S n .
For n ≥ 1, we have the exact sequence
∼ ∼
n
H1 (D− ) / H1 (S n )
= / H1 (S n ,D−
n
)
∂=0 / n
H0 (D− )
= / H0 (S n )
∼
0 Z
= / Z.
n n
Since D− is contractible, we know H1 (D− ) = 0. Hence the map H1 (S n ) →
n n
H1 (S ,D− ) is injective. Since the map
Z∼ = H0 (Dn ) → H0 (S n ) ∼
=Z
−
0 0.
0 0.
Generators for Hn (S n )
Last time we calculated the homology groups of S n and the pair (Dn ,S n−1 ).
To make this calculation a bit more concrete, let us try to figure out the generators
of the infinite cyclic groups Hn (Dn , S n−1 ) and Hn (S n ):
• On the standard n-simplex, there is a special n-chain Sn (∆n ), called the fun-
damental n-simplex, given by the identity map ιn : ∆n → ∆n . We observed
in a previous lecture that ιn is not a cycle, since its boundary ∂(ιn ) ∈ Sn−1 (∆n )
is the alternating sum of the faces of the n-simplex each of which is a generator
in Sn−1 (∆n ).
X
∂(ιn ) = (−1)i ϕni (∆n−1 ) ̸= 0.
i
Constructing φn
1
For each ∆n the point c = (t0 , . . . ,tn ) with ti = n+1 for all i is the barycen-
n
ter of ∆ .
For every point x ∈ ∆n which is not c, there is a unique ray from c to x.
We denote the unique point where this ray hits ∂∆n by f (x). In particular,
if x ∈ ∂∆n , then f (x) = x.
81
A concrete generator of H1 (S 1 )
We just learned that the class [∂(φ2 )] is a generator of H1 (S 1 ). We can
describe this class as follows:
By definition, ∂(φ2 ) is the 1-cycle
∂(φ2 ) = d0 φ2 − d1 φ2 + d2 φ2
= φ2 ◦ ϕ20 − φ2 ◦ ϕ21 + φ2 ◦ ϕ22 .
Recall that φ2 maps ∂∆2 homeomorphically to S 1 . With this in mind, the
summands look like
1 2
φ2 ◦ ϕ20 (1 − t, t) = eiπ(− 6 +t 3 )
7 2
φ2 ◦ ϕ21 (1 − t, t) = eiπ( 6 −t 3 )
7 2
φ2 ◦ ϕ22 (1 − t, t) = eiπ( 6 +t 3 ) .
82 CHAPTER 7. GENERATORS FOR HN (S N ) AND FIRST APPLICATIONS
First applications
f : Rn \ {0} → Rm \ {f (0)}
is also a homeomorphism, since these are open subsets and f|Rn \{0} and (f −1 )|Rm \{f (0)}
are still continuous mutual inverses.
We showed as an exercise that, for any k ≥ 1, S k−1 is a strong deformation
retract of Rk \{0}. In particular, we showed S k−1 ≃ Rn \{0}. Since the translation
Rn → Rn , y 7→ y + x is a homeomorphism for any x ∈ Rn , this implies that
But by our calculation of the Hq (S n−1 ), such an isomorphism can only exist
if n − 1 = q = m − 1. This contradicts the assumption n ̸= m. QED
We can also give a short proof of Brouwer’s famous Fixed-Point Theorem:
Before we prove the theorem, let us have a look at dimension one, where the
result is very familiar:
If g(0) = 0 or g(1) = 0, we are done. But if g(0) > 0 and g(1) < 0, then the
Intermediate Value Theorem implies that there is an x0 ∈ (0, 1) with
g(x0 ) = 0, i.e., f (x0 ) = x0 .
g : Dn → ∂Dn .
Let i : S n−1 = ∂Dn ,→ Dn denote the inclusion map. Note that if x ∈ ∂Dn ,
then g(x) = x. In other words,
g ◦ i = idS n−1 .
idH n−1 )
n−1 (S
Z∼
= Hn−1 (S n−1 ) / Hn−1 (S n−1 ) ∼
=Z
5
Hn (i) ) Hn (g)
Hn−1 (Dn ) = 0.
Before we start the proof, let us have a look at what happens for the reflection
map
r : D1 = [−1, 1] → [−1, 1] = D1 , t 7→ −t
and its restriction to S 0 . Recall that S 0 consists of just two points, x = 1 and
y = −1 (on the real line R). The effect of r on S 0 is to interchange x and y.
The inclusion maps ix and iy induce an isomorphism
∼
=
→ H0 (S 0 ).
H0 ({x}) ⊕ H0 ({y}) −
Thus H0 (r) can be viewed as
H0 (r) : H0 (S 0 ) → H0 (S 0 ), (a, b) 7→ (b, a).
Z / Z⊕Z
ϵ / Z / 0.
We know that the left hand and central squares commute, since the inclusion
and the reflection commute. We know that the horizontal maps are isomorphisms
from the calculation of these groups.
Thus, knowing H0 (r) = −1 on Ker (ϵ), we see that H1 (r) is also multipli-
cation by −1 on H1 (S 1 ).
Now we can proceed by induction: For n ≥ 2, we have again a commutative
diagram from the calculation of Hn (S n ):
∼ ∼ ∼
Hn (S n )
= / n o
Hn (S n , D+ )
= n
Hn (D− , S n−1 )
= / Hn−1 (S n−1 )
Hn (r) Hn (r) Hn (r) Hn−1 (r)
∼ ∼ ∼
Hn (S n )
= / n o
Hn (S n , D+ )
= n
Hn (D− , S n−1 )
= / Hn−1 (S n−1 ).
The right most square commutes by an exercise from last week. The left hand
and central squares commute, since the inclusion and the reflection commute.
88 CHAPTER 7. GENERATORS FOR HN (S N ) AND FIRST APPLICATIONS
Assuming the assertion for n − 1, i.e., Hn−1 (r) is multiplication by −1, we see
that Hn (r) is also multiplication by −1. □
LECTURE 8
Calculating degrees
In last week’s exercises we showed many useful properties of the degree and
calculated the degree of some interesting maps. Today, we are going to continue
our study of the degree.
But before we move on, another reason why the degree is so important:
Brouwer degree
Let p be an arbitrary point in S n . We consider p as the base point of
S n . Let C(S n ,S n )∗ be the set of pointed continuous maps, i.e., maps
f : S n → S n with f (p) = p. Pointed homotopy defines an equivalence
relation on this set. Hence we can define the quotient set
[S n , S n ]∗ := C(S n , S n )∗ / ≃
where we identify f and g if they are homotopic to each other f ≃ g.
Now the degree defines a function from C(S n ,S n )∗ to the integers Z. Since
the degree is invariant under homotopy, i.e., f0 ≃ f1 implies deg(f0 ) =
deg(f1 ), it induces a function
deg : [S n , S n ]∗ → Z, f 7→ deg(f ).
This function is actually an isomorphism of abelian groups. In fancier
language, we write πn (S n ) = [S n , S n ]∗ , call it the nth homotopy group
of S n and say that the degree completely determines πn (S n ).
Now let us see what kind of maps between spheres there are. Actually, such
maps arise quite naturally. For, every invertible real n × n-matrix A defines a
≈
homeomorphism between Rn − → Rn . It extends to a homeomorphism on the
one-point compactification S n of Rn and therefore defines a map
A : S n → S n.
89
90 CHAPTER 8. CALCULATING DEGREES
Proof: This follows from the fact that every othogonal matrix is the product
of reflections (at appropriate hyperplanes in Rn ). A reflection has determinant
−1, but it also has degree −1 as we have shown before. Since both deg and det
are multiplicative, the result follows. QED
Now let A ∈ GLn (R) be an invertible n × n-matrix. It defines a map
f : Rn \ {0} → Rn \ {0}, f (x) := Ax.
It induces a map
Hn−1 (f ) : Hn−1 (Rn \ {0}) → Hn−1 (Rn \ {0}).
Proof: Recall from linear algebra that any invertible matrix A has a po-
lar decomposition A = BC with B a symmetric matrix with only positive
eigenvalues and C ∈ O(n). Since we already know that the assertion is true
if A ∈ O(n), it suffices to show that B is homotopic to the identity as maps
Rn → Rn .
Since all eigenvalues of B are positive, we know det(B) > 0. Hence B and I
lie both in the component GLn (R)+ of the matrices with det > 0. The continuous
91
map
Γ : [0,1] → GLn (R)+ , t 7→ tI + (1 − t)B
defines a homotopy between I and B.
To check that Γ(t) is in GLn (R)+ for all t, we observe that the eigenvalues of
Γ(t) are all strictly positive. For, let λ be an eigenvalue of B. Then t + (1 − t)λ
is an eigenvalue of Γ(t), since all nonzero vectors are eigenvectors of tI. This
implies det(Γ(t)) > 0. QED
For n > 1, we know Hn (Rn , Rn \ {0}) ∼
= Z, since (Dn , S n−1 ) ,→ (Rn , Rn \ {0})
is a deformation retract. For A as above, we obtain a commutative diagram from
the long exact sequences of pairs
Hn (A)
Hn (Rn , Rn \ {0}) / Hn (Rn , Rn \ {0})
Hn−1 (Rn \ {0}) / Hn−1 (Rn \ {0}).
Hn−1 (A)
The vertical connecting homomorphisms are isomorphisms, since they are iso-
morphisms for the pair (Dn , S n−1 ). Since the diagram commutes, we deduce the
following consequence from the previous result:
Corollary
The effect of the map
Hn (A) : Hn (Rn , Rn \ {0}) → Hn (Rn , Rn \ {0})
is given by multiplication with sign(det(A)).
Now we would like to apply this to a situation familiar from Calculus. First
a brief observation:
Lemma
Let U ⊂ Rn be an open subset and x ∈ U . Then
Hn (U, U \ {x}) ∼
= Z.
• In particular, we can assume |g(x)| < |x|/2 for x small enough. By excision,
we can shrink U to become small enough such that still 0 ∈ U and |g(x)| < |x|/2
for all x ∈ U .
• We can further assume that A = I is the identity. For if not, we can replace
f with A−1 f and use the functoriality of Hn .
• Now we have |f (x) − x| < x/2 for all x ∈ U . Hence the map
satisfies h(x, t) ̸= 0 for all (x,t). This implies that Dh(0, t) is in GLn (R)+ for all
t. Thus h defines a homotopy between f and the identity map and the effect of
Hn (f ) is the same as the one of the identity map. □
Local degree
Often the effect of a map can be studied by focussing on the neighborhood
of certain interesting points. We would like to exploit this idea for studying the
degree.
93
For any given i, the obvious inclusions of pairs induce the following diagram:
Hn (f|Ui )
(10) Hn (Ui , Ui \ {xi }) / Hn (V, V \ {y})
deg(f |xi )
∼
=
ki ∼
=
t pi
Hn (f )
Hn (S n , S n \ {xi }) o Hn (S n , S n \ f −1 (y)) / Hn (S n , S n \ {y})
j O O
j ∼
=
∼
=
Hn (f )
Hn (S n ) / Hn (S n ).
By the excision axiom applied as in the proof of the lemma below and by
an exercise, we know that the diagonal maps on the left and the vertical maps
on the right are isomorphisms, as indicated in (10).
We are going to prove this result in the next lecture. In the diagram above
we claimed that some maps are isomorphisms. Here is an explanation why:
Lemma
(a) Let U ⊂ S n be an open subset and x ∈ U . Then there is an isomorphism
∼
→ Hn (S n , S n \ {x}) ∼
=
Hn (U, U \ {x}) − = Z.
(b) Let x1 , . . . , xm be m distinct points in S n and U1 , . . . , Um disjoint open
neighborhoods with xi ∈ Ui . Then there is an isomorphism
∼
→ Hn (S n , S n \ {x1 , . . . , xm }) ∼
=
⊕m Hn (Ui , Ui \ {xi }) −
i=1 = ⊕m Z.
i=1
Last time we defined the local degree of a map. The situation was as follows:
For n ≥ 1, let f : S n → S n be a map with the property that there is a point
y ∈ S n such that f −1 (y) = {x1 , . . . ,xm } consists of finitely many points.
We choose small disjoint open neighborhoods U1 , . . . ,Um of each xi such that
each Ui is mapped into an open neighborhood V of y in S n . (We could choose V
first, and then intersect f −1 (V ) with small open disks around xi ...).
Since xi ∈ Ui and the different Uj s are disjoint, we have
For any given i, the obvious inclusions of pairs induce the following diagram:
Hn (f|Ui )
(11) Hn (Ui , Ui \ {xi }) / Hn (V, V \ {y})
deg(f |xi )
∼
=
ki ∼
=
t pi
Hn (f )
Hn (S n , S n \ {xi }) o Hn (S n , S n \ f −1 (y)) / Hn (S n , S n \ {y})
j O O
j ∼
=
∼
=
Hn (f )
Hn (S n ) / Hn (S n ).
By the excision axiom applied as in the proof of the lemma below and by
an exercise, we know that the diagonal maps on the left and the vertical maps on
the right are isomorphisms, as indicated in (11). Then we made the following
definition:
95
96 CHAPTER 9. LOCAL VS GLOBAL DEGREES
We have the following result which connects global and local degrees:
• Since the lower square in diagram (11) commutes and since the lower
horizontal map is given by the degree of f , the asserted formula follows. QED
We have used the following lemma in diagram (11) the above proof:
Lemma
(a) Let U ⊂ S n be an open subset and x ∈ U . Then there is an isomorphism
∼
→ Hn (S n , S n \ {x}) ∼
=
Hn (U, U \ {x}) − = Z.
(b) Let x1 , . . . , xm be m distinct points in S n and U1 , . . . , Um disjoint open
neighborhoods with xi ∈ Ui . Then there is an isomorphism
∼
→ Hn (S n , S n \ {x1 , . . . , xm }) ∼
=
⊕m Hn (Ui , Ui \ {xi }) −
i=1 = ⊕m Z.
i=1
We proved this lemma in a previous lecture. Now let us apply what we learned
in an example:
• We know this is true for for k = 0 when f0 is the constant map and for
k = 1 when f1 is the identity.
• We know it also for k = −1, since z 7→ z −1 is a reflection at the real
axis.
• It suffices to check the remaining cases for k > 0, since the cases for k < 0
follow from composition with z 7→ z −1 and the multiplicativity of the
degree.
• So let k > 0. For any y ∈ S n , fk−1 (y) consists of k distinct points
x1 , . . . ,xk . Each point xi has an open neighborhood Ui which is mapped
homeomorphically by f to an open neighboorhood V of y. This local
homeomorphism is given by stretching (by the factor k) and a rotation
in positive direction.
• Stretching by a factor is homotopic to the identity near xi . Hence the
local degree of the stretching is +1.
A rotation is a homeomorphism and its global and local degree at any point
agree.
Since the rotation is in the positive direction, it is homotopic to the identity
and has therefore degree +1.
• Hence deg(f |xi ) = 1.
• Thus we can conclude by the proposition that
k
X
deg(f ) = deg(f |xi ) = k.
i=1
LECTURE 10
We still need to prove the Homotopy Axiom and Excision Axiom for singular
homology. The proof will follow from constructing a homotopy between chain
complexes, a concept we are now going to explore.
Recall that a chain complex K∗ = (K∗ ,∂ K ) constists of a sequence of abelian
groups
K K K
∂n+2 ∂n+1 n ∂K ∂n−1
· · · −−−→ Kn+1 −−−→ Kn −→ Kn−1 −−−→ · · ·
together with homomorphisms ∂nK : Kn → Kn−1 with the property that ∂n−1 ◦∂n =
0. Our main example is the singular chain complex.
Just to make sure that we understand the definition, let us look at an example
of a sequence of groups that is not a chain complex. Consider the sequence of
maps
2 2 2 2 2
··· →
− Z→
− Z→
− Z→
− Z→
− ···
99
100 CHAPTER 10. HOMOTOPIES OF CHAIN COMPLEXES
We need to check that this is well-defined. Since we hopped over this point
in Lecture 5, let us do it now.
There are two things we should check. We need to know that f sends cycles
to cycles and boundaries to boundaries.
First, let x ∈ Kn be a cycle in K, i.e., ∂nK (x) = 0. Then (12) implies
∂nL (fn (x)) = fn−1 (∂nK ((x)) = fn−1 (0) = 0.
Thus fn (x) is a cycle in L and we get fn (Zn (K∗ )) ⊂ Zn (L∗ ).
K
Second, let a x ∈ Kn be a boundary, say ∂n+1 (y) = x. Then (12) implies
K L
fn (x) = fn (∂n+1 (y)) = ∂n+1 (fn+1 (y)).
Thus fn (x) is a boundary in L and we get fn (Bn (K∗ )) ⊂ Bn (L∗ ). This shows
that f induces a well-defined homomrphism between the homologies of K∗ and
L∗ .
We would like to transfer the notion of homotopies between maps of spaces
to the homotopies between maps of chain complexes. This follows the general
slogan: Homotopy is a smart thing to do.
Why? The notion of an isomorphism in a category, e.g. the category of
topological spaces or the category of chain complexes, is often too rigid. There
are too few isomorphism such that classifying objects up to isomorphism is too
difficult. Therefore, one would like to relax the conditions. For many situations,
homotopy turns out to provide the right amount of flexibility and rigidity
at the same time. Moreover, many invariants, in fact all invariants in Algebraic
Topology, do not change if we alter a map by a homotopy.
In other words, our invariants only see the homotopy type.
Actually, this is exactly what we are going to show for singular homology
today. It is also true in Homological Algebra. The homology of a chain complex
only depends on the homotopy type of the complex.
So let us define homotopies between chain maps:
such that
L
(13) fn − gn = ∂n+1 ◦ hn + hn−1 ◦ ∂nK for all n ∈ Z.
Kn+1 / L
; n+1
K
hn L
∂n+1 ∂n+1
fn −gn
Kn / Ln
;
K
∂n L
∂n
hn−1
Kn−1 / Ln−1
If such a homotopy exists, we are going to say that f and g are homotopic
and write f ≃ g.
We say that f is null-homotopic if f ≃ 0.
fn − kn = fn − gn + gn − kn
L
= ∂n+1 ◦ hn + hn−1 ◦ ∂nK + ∂n+1
L
◦ h′n + h′n−1 ◦ ∂nK
L
= ∂n+1 ◦ (hn + h′n ) + (hn−1 + h′n−1 ) ◦ ∂nK .
(2) Let h be a homotopy which shows f ≃ f ′ and k be a homotopy which
shows g ≃ g ′ . Composition with g on the left and using that g is a chain map
102 CHAPTER 10. HOMOTOPIES OF CHAIN COMPLEXES
yields
gn ◦ (fn − fn′ ) = gn ◦ (∂n+1
L
◦ hn + hn−1 ◦ ∂nK )
M
= ∂n+1 ◦ (gn+1 ◦ hn ) + (gn ◦ hn−1 ) ◦ ∂nK .
This shows that the sequence of maps gn+1 ◦ hn defines a homotopy g ◦ f ≃ g ◦ f ′ .
Composition with f ′ on the right and using that f ′ is a chain map yields
(gn − gn′ ) ◦ fn′ = (∂n+1
M
◦ kn + kn−1 ◦ ∂nL ) ◦ fn′
M
= ∂n+1 ′
◦ (kn ◦ fn′ ) + (kn−1 ◦ fn−1 ) ◦ ∂nK .
This shows that the sequence of maps kn ◦ fn′ defines a homotopy g ◦ f ′ ≃ g ′ ◦ f ′ .
Summarizing we have shown
g ◦ f ≃ g ◦ f ′ ≃ g′ ◦ f ′.
By transitivity, this shows the desired result. □
Now we are ready to show an important fact in homological algebra:
Proof: This follows immediately from the fact that fn − gn is just given by
boundaries which, by definition, vanish in homology.
More concretely, let x ∈ Kn be an arbitrary cycle in Kn and let h be a
homotopy which gives f ≃ g. Then we get by using the definition of homotopies
L
Hn (f )([x]) = [fn (x)] = [gn (x) + ∂n+1 (hn (x)) + hn−1 (∂nK (x))] = [gn (x)] = Hn (g)([x])
L
where we use that ∂n+1 (hn (x)) is obviously a boundary in Ln and that hn−1 (∂nK (x)) =
0, since x is a cycle in Kn by assumption. □
Now we can also mimic the notion of homotopy equivalences.
0
··· → 0 → Z →
− Z → 0 → ··· .
Since all maps are trivial, we have Hn (K∗ ) = Kn for all n. Hence K∗ has
exactly two nonzero homology groups, both being Z. In particular, it is
not contractible.
• Let K∗ be the chain complex
1
··· → 0 → Z →
− Z → 0 → ··· .
104 CHAPTER 10. HOMOTOPIES OF CHAIN COMPLEXES
This complex is actually an exact sequence. Thus Hn (K∗ ) = 0 for all n.
Moreover it is contractible. We can write down a homotopy by
0
id / 0
?
0
id /
Z ?Z
1
1 1
id /
Z ?Z
0
0
id / 0.
• Let K∗ be the chain complex
2
··· → 0 → Z →
− Z → 0 → ··· .
This complex has one nonzero homology group H1 (K∗ ) = Z/2. It is
therefore not contractible.
• Let K∗ be the chain complex
2 2 2 2 2
··· →
− Z/4 →
− Z/4 →
− Z/4 →
− Z/4 →
− ··· .
The homology of K∗ vanishes, since, at each stage, the image and the
kernel of the differential is 2Z/4. Nevertheless, K∗ is not contractible.
For if there was a homotopy between idK∗ and the zero map, it would
like this
Z/4
id / Z/4
=
hn
2 2
Z/4
id / Z/4
=
2 2
hn−1
Z/4 / Z/4
id
and satisfy id = 2hn + hn−1 2. But 2hn + hn−1 2 can only produce even
numbers modulo 4. Hence it cannot be the identity map on Z/4.
After all this abstract stuff we should better demonstrate that the notion of
chain homotopies is useful for our purposes. We are going to do this by showing
that homotopies between maps of spaces induces a chain homotopy. By what we
have just seen, this will prove the Homotopy Axiom for singular homology.
LECTURE 11
We are going to prove the Homotopy Axiom for singular homology. The proof
will follow from constructing a homotopy between chain complexes.
Let f, g : X → Y be two homotopic maps and let h : X × [0,1] → Y be a
homotopy between them. Let σ : ∆n → X be an n-simplex on X. Then h induces
a map
σ×id h
∆n × [0,1] −−−→ X × [0,1] →
− Y
which defines a homotopy between f ◦ σ and g ◦ σ.
Our goal is to turn this into a geometrically induced chain homotopy
between Sn (f ) and Sn (g). By our result from the previous lecture, this will
imply the Homotopy Axiom.
So let us have a closer look at the space ∆n × [0,1]. For n = 1, it looks just
like a square. Via the diagonal we can divide it into two triangles which look like
∆2 . For n = 2, ∆2 × [0,1] looks like a prism which we can divide into three copies
of ∆3 .
In general, ∆n × [0,1] looks like a higher dimensional prism which we can
divide into n + 1 copies of ∆n+1 . We should make this idea more precise:
Simplices on a prism
For every n ≥ 0 and 0 ≤ i ≤ n, we define an injective map
pni : ∆n+1 → ∆n × [0,1],
(t0 , . . . ,tn+1 ) 7→ ((t0 , . . . ,ti−1 ,ti + ti+1 ,ti+2 , . . . ,tn+1 ), ti+1 + · · · + tn+1 ).
We can consider each pni as an n + 1-simplex on the space ∆n × [0,1]. When
n is clear we will often drop it from the notation.
Let us have a look at what happens for n = 1: Then the effect of p10 and
p11 is given by
and
In fact, we could define pni as the unique affine map which satisfies (14).
Let j0 and j1 be the two inclusions
j0 : ∆n ,→ ∆n × [0,1], x 7→ (x,0)
j1 : ∆n ,→ ∆n × [0,1], x 7→ (x,1)
determined by the endpoints of [0,1].
For the next result, recall our formulae for the face maps on standard sim-
plices:
For 0 ≤ i ≤ n + 1 which can be described as
ϕn+1
i (t0 , . . . ,tn ) = (t0 , . . . ,ti−1 ,0,ti , . . . ,tn )
with the 0 inserted at the ith coordinate.
Using the standard basis, ϕin+1 can be described as the affine map (a trans-
lation plus a linear map)
(
ek k<i
(15) ϕn+1
i : ∆n ,→ ∆n+1 determined by ϕn+1
i (ek ) =
ek+1 k ≥ i.
pn0 (ϕn+1
0 (t0 , . . . ,tn )) = p0 (0,t0 , . . . ,tn )
n
!
X
= (t0 , . . . ,tn ), ti
i=0
= (t0 , . . . ,tn ,1) = j1 (t0 , . . . ,tn ).
pnn (ϕn+1 n
n+1 (t0 , . . . ,tn )) = pn (t0 , . . . ,tn ,0)
= (t0 , . . . ,tn ,0) = j0 (t0 , . . . ,tn ).
pni (ϕn+1
i (t0 , . . . ,tn ))
= pni (t0 , . . . ,ti−1 ,0,ti , . . . ,tn )
= ((t0 , . . . ,ti−1 ,0 + ti ,ti+1 , . . . ,tn ), 0 + ti + · · · + tn )
n
!
X
= (t0 , . . . ,tn ), tj ,
j=i
109
and
pni−1 (ϕn+1
i (t0 , . . . ,tn ))
= pni−1 (t0 , . . . ,ti−1 ,0,ti , . . . ,tn )
= ((t0 , . . . ,ti−2 ,ti−1 + 0,ti ,ti+1 , . . . ,tn ), ti + · · · + tn )
n
!
X
= (t0 , . . . ,tn ), tj .
j=i
∆n
pn
j+1 &
∆n−1 × [0,1] / ∆n × [0,1].
ϕn
i ×id
To check this, we are going to use formulae (14) and (15). Since the affine maps
involved are determined by their effect on the ek s, this will suffice to prove the
formulae.
For k < i, we get
pnj+1 ◦ ϕn+1
i (ek ) = pnj+1 (ek ) = (ek ,0)
= (ϕni × id)(ek ,0) = (ϕni × id) ◦ pn−1
j (ek ).
For i ≤ k ≤ j, we get
pnj+1 ◦ ϕn+1
i (ek+1 ) = pnj+1 (ek ) = (ek+1 ,0)
= (ϕni × id)(ek ,0) = (ϕni × id) ◦ pn−1
j (ek ).
Let jtX denote the inclusion X ,→ X × [0, 1], x 7→ (x, t). The prism operator
is the desired chain homotopy:
n
!
X
n−1 n−1
P ◦ ∂n (σ) = P (−1)i σ ◦ ϕni
i=0
X
= (−1)i+j (σ × id) ◦ (ϕni × id) ◦ pn−1
j
0≤j<i≤n
X
+ (−1)i+j (σ × id) ◦ (ϕni × id) ◦ pn−1
j
0≤i≤j≤n−1
X
=− (−1)i+j+1 (σ × id) ◦ pnj+1 ◦ ϕn+1
i by (20)
0≤j<i≤n
X
− (−1)i+j+1 (σ × id) ◦ pnj ◦ ϕn+1
i+1 by (19).
0≤i≤j≤n−1
n
!
X
∂n+1 ◦ P n (σ) = ∂n+1 (−1)j (σ × id) ◦ pnj
j=0
n X
X n+1
= (−1)i+j (σ × id) ◦ pnj ◦ ϕn+1
i
j=0 i=0
X
= (−1)i+j (σ × id) ◦ pnj ◦ ϕn+1
i (i < j)
0≤i<j≤n
Xn
+ (σ × id) ◦ pni ◦ ϕn+1
i (i = j)
i=0
n+1
X
− (σ × id) ◦ pni−1 ◦ ϕn+1
i (i = j + 1)
i=1
X
+ (−1)i+j (σ × id) ◦ pnj ◦ ϕn+1
i (i > j + 1)
1≤j+1≤i≤n+1
′
X
= (−1)i+j +1 (σ × id) ◦ pnj′ +1 ◦ ϕn+1
i
0≤i≤j ′ ≤n−1
+ (σ × id) ◦ j1 − (σ × id) ◦ j0
′
X
+ (−1)i +j+1 (σ × id) ◦ pnj ◦ ϕn+1
i′ +1 .
0≤j≤i′ ≤n
112 CHAPTER 11. HOMOTOPY INVARIANCE OF SINGULAR HOMOLOGY
For the final step we used again the trick to relabel the indices and wrote j ′ = j −1
and i′ = i − 1. By comparing the two calculations, we see that all summands
cancel out except for (σ × id) ◦ j1 − (σ × id) ◦ j0 .
Thus we can conclude:
Sn (f ) − Sn (g)
= Sn (h) ◦ Sn (j1X ) − Sn (h) ◦ Sn (j0X )
= Sn (h) ◦ (∂n+1 ◦ P n ) + Sn (h) ◦ (P n−1 ◦ ∂n )
= ∂n+1 ◦ (Sn+1 (h) ◦ P n ) + (Sn (h) ◦ P n−1 ) ◦ ∂n since S∗ (h) is a chain map.
Hence Hn (i) is injective for all n. That means that the sequence
Hn (i)
0 → Hn (A) −−−→ Hn (X)
∂ Hn−1 (i)
· · · → Hn (X,A) →
− Hn−1 (A) −−−−→ Hn−1 (X) → · · ·
implies that the connecting homomorphism ∂ : Hn (X,A) → Hn−1 (A) is the zero
map. Thus we get a short exact sequence
Hn (i) Hn (j)
(21) 0 → Hn (A) −−−→ Hn (X) −−−→ Hn (X,A) → 0.
is an isomorphism.
We can describe an inverse of this map as
Hn
hoTop.
We are going to discuss the Excision Axiom for singular homology and some
consequences. Let us first recall what it says:
We are going to deduce the excision property of homology from the following
locality principle.
Let X be a topological space and let A = {Aj }j∈J be a cover of X, i.e., a
collection of subsets Aj ⊆ X such that X is the union of the interiors of the
Aj s.
A-small chains
• An n-simplex σ : ∆n → X is called A-small if the image of σ is contained
in one of the Aj s. P
• An n-chain c = i ni σi if X is called A-small if, for every i, there is a
Aj such that σi (∆n ) ⊂ Aj .
• We are going to denote the subgroup of A-small n-chains by
SnA (X) := {c ∈ Sn (X) : σ is A − small}.
• For a subspace A ⊂ X, we write
SnA (X)
SnA (X,A) := .
SnA (A)
115
116 CHAPTER 12. LOCALITY AND THE MAYER-VIETORIS SEQUENCE
If, for each j, ιj : Aj ,→ X denotes the inclusion map, then we can describe
SnA (X) also as
!
M ⊕j Sn (ιj )
SnA (X) = Im Sn (Aj ) −−−−−→ Sn (X) .
j∈J
The point of A-small chains is that we can use their chain complex to compute
singular homology:
The proof of this theorem takes quite an effort and we will postpone it for a
moment. Instead we will now explain how the excision property follows from the
theorem.
• Proof of the Excision Axiom using small chains:
Since Z̄ ⊆ A◦ , we have (X − Z)◦ ∪ A◦ = X. Thus, if we set B := X − Z,
A = {A, B} is a cover of X.
Moreover, we can rewrite
(X − Z, A − Z) = (B, A ∩ B).
0 / S∗ (A) / S∗ (X) / S∗ (X)/S∗ (A) / 0.
the Five-Lemma imply that the right-hand vertical map induces an isomor-
phism in homology as well. Thus we are reduced to compare S∗ (B,A ∩ B) and
S∗A (X)/S∗ (A).
Now we observe
S∗A (X) = S∗ (A) + S∗ (B) ⊂ S∗ (X)
and hence
S∗ (B) S∗ (B) = S∗ (A) + S∗ (B)
∼ S A (X)
= −
→ = ∗
S∗ (A ∩ B) S∗ (A) ∩ S∗ (B) S∗ (A) S∗ (A)
where the middle isomorphism follows from the general comparison of quo-
tients of sums and intersections of abelian groups.
Thus the chain map
S∗ (B,A ∩ B) → S∗A (X)/S∗ (A)
induces an isomorphism in homology and the excison axiom holds. QED
r βn
Hn (A ∩ B) / Hn (A) ⊕ Hn (B) / Hn (X)
αn
MV
∂n
r
Hn−1 (A ∩ B) / ···
αn−1
Proof: From the proof of the Excision Axiom we remember that there is a
short exact sequence of chain complexes
S∗ (jA ) h i
−S∗ (jB ) S∗ (iA ) S∗ (iB )
0 → S∗ (A ∩ B) −−−−−−−−→ S∗ (A) ⊕ S∗ (B) −−−−−−−−−−−−→ S∗A (X) → 0.
Note that the exactness at the right-hand term was part of the proof of the
Excision Axiom and the exacntess at the middle term can be easily checked by
looking long enough at the commutative diagram
∆n
$ jA "
A∩B / A
jB iA
% /
B iB
X.
119
s A
βn
Hn (A ∩ B) / Hn (A) ⊕ Hn (B) / HnA (X)
αn
A
∂n
r
Hn−1 (A ∩ B) / ···
αn−1
Thus we can apply the inverse of the isomorphism of the Small Chain The-
orem and define ∂nM V to be
∼
= ∂A
→ HnA (X) −→
∂nM V : Hn (X) − n
Hn−1 (A ∩ B).
0 / H1 (S 1 ) / Z⊕Z / Z⊕Z
1 1
−1 −1
where we obtain the lower right-hand map by observing that all summands are
of the form H0 (pt) and hence each generator in H0 (A ∩ B) is sent to (1, − 1) by
[H0 (jA ), −H0 (jB )]. Thus H1 (S 1 ) is the kernel of this map:
1 ∼ 1 1
H1 (S ) = Ker = {(x, −x) ∈ Z ⊕ Z} ∼
= Z.
−1 −1
Cell complexes
B / X ∪f B
φ
by
X ∪f B := (X ⊔ B)/(a ∼ f (a) for all a ∈ A).
We say that X ∪f B arises from attaching B to X along f , or along A,
and f is called an attaching map.
B / X ∪f B
#,
Y
then there is a unique dotted arrow which makes all triangles commute. We
can reformulate this fact by saying that X ∪f B is the pushout of the solid
diagram.
For example:
• if X = ∗ consists of just a point, then
X ∪f B = ∗ ∪f B = B/A;
• if A = ∅, then X ∪f B = X ⊔ B is just a disjoint union.
A more important example is the following:
Attaching a cell
We consider the pair (Dn , S n−1 ) of an n-disk and its boundary. We are
going to think of Dn as an n-cell.
Suppose we are given a map f : S n−1 → X. Then we can attach an n-cell
to X via f as
f
S n−1
_ / X
Dn / X ∪f Dn .
125
` `
n / X ∪f Dαn .
α∈J Dα α∈J
f
S 0 ⊔ _ S 0 / ∗
D1 ⊔ D1 / ∗ ∪f (D1 ⊔ D1 ).
Since there is only one choice for f , we get a figure eight: we start with
two 1-disks D1 and then we identify all four boundary points with the
0-cell. We denote this space by S 1 ∨ S 1 .
• We continue with this space and attach one 2-cell: We can think of S 1 ∨S 1
as an empty square where we glue together the horizontal edges and the
vertical edges. Then we glue in a 2-cell into the square by attaching its
126 CHAPTER 13. CELL COMPLEXES
boundary to the edges a, b, a−1 , and b−1 , i.e., by walking clockwise:
f =aba−1 b−1
S 1 _ / S1 ∨ S1
D2 / (S 1 ∨ S 1 ) ∪f D2 = T 2 .
Cell complex
A cell complex, or CW-complex, is a space X equipped with a sequence
of subspaces
∅ = Sk−1 X ⊆ Sk0 X ⊆ Sk1 X ⊆ Sk2 X ⊆ · · · X
such that
• X is the union of the Skn Xs,
• for all n, Skn X arises from Skn−1 by attaching n-cells, i.e., there
is a pushout diagram
n−1 fn /
`
α∈Jn Sα Skn−1 X
_
n /
`
α∈Jn Dα φn Skn X.
The space Skn X is called the n-skeleton of X.
Before we study more examples, we fix more terminology and list some facts
which should help clarify the picture:
• The topology of a cell complex is determined by its skeleta, i.e., a subset
U ⊂ X is open (closed) if and only if U ∩ Skn X is open (closed) for all
n.
• For any n-cell Dαn , the induced map φα : Dαn → X is called the charac-
teristic map of the cell. As we explained before, the restriction to the
open interior (Dαn )◦ = Dαn − Sαn−1
(φα )|(Dαn )◦ → X
is a homeomorphism onto its image.
• The C in CW-complex stands for closure finite which means that, for
every cell, φα (Sαn−1 ) is contained in finitely many cells (of dimension at
most n − 1).
• Note that every nonempty cell complex must have at least one 0-cell.
• The cell structure of a cell complex is in general not unique. Often there
are many different cell structures. We will observe this for example for
the n-sphere.
Here is an important theorem which demonstrates the wide range and impor-
tance of cell complexes:
• The sphere S n is a cell complex with just two cells: one 0-cell e0 (that
is a point) and one n-cell which is attached to e0 via the constant map
S n−1 → e0 . Geometrically, this corresponds to expressing S n as Dn /∂Dn :
we take the open n-disk Dn \ ∂Dn and collapse the boundary ∂Dn to a
single point which is, say, the north pole N = e0 .
• Real projective space RPn is a cell complex with one cell in each
dimension up to n. To show this we proceed inductively. We know that
RP0 consists of a single point, since it is S 0 whose two antipodal points
are identified.
130 CHAPTER 13. CELL COMPLEXES
Now we would like to understand how RPn can be constructed from
n−1
RP :
We embed Dn as the upper hemisphere into S n , i.e., we consider
D as {(x0 , . . . ,xn ) ∈ S n : x0 ≥ 0}. Then
n
• We can continue this process and build the infinite projective space
RP∞ := n RPn . It is a cell complex with one cell S
S
in each dimension.
∞ ∞
We can think of RP as the space of lines in R = n Rn .
n−1
X
(z0 , . . . ,zn−1 ) 7→ (z0 : z1 : . . . : zn−1 : 1 − ( |zi2 |)1/2 ).
i=0
D2n / CPn−1 ∪f D2n
g
φ '
-
CPn .
The induced map
g : D2n ∪f CPn−1 → CPn
131
Since we can rescale the nth coordinate, this map is bijective. Hence
it is a continuous bijection defined on a compact space. We learned
earlier that this implies that g is a homeomorphism.
We conclude that CPn is a cell complex with exactly one i-cell in
each even dimension up to 2n.
• Again we could continue this process and build infinite complex pro-
jective space CP∞ which is a cell complex with one i-cell in each even
dimension.
Finally, we would like to have a good notion of subspace in a cell complex
which respects the cell structure. It turns out that it is not sufficient to just
require to have a subspace. Though not much more is actually required. For,
a subspace A ⊂ X is subcomplex, or sub-CW-complex, if it is closed and a
union of cells of X.
These conditions imply that A is a cell complex on its own. For, since A
is closed the characteristic maps of each cell of A has image in A and so does
each attaching map. Hence the cells with their characteristic maps which lie in
A provide A with a cell structure.
A more technical definition sounds like this:
Subcomplexes
Let X be a cell complex with attaching maps {fα : Sαn−1 → Skn−1 X : α ∈
Jn , n ≥ 0}.
A subcomplex A of X is a closed subspace A ⊆ X such that for all n ≥ 0,
there is a subset Jn′ ⊂ Jn so that Skn A := A ∩ Skn X turns A into a cell
complex with attaching maps {fβ : β ∈ Jn′ , n ≥ 0}.
A pair (X,A) which consists of a CW-complex X and a subcomplex A is
called a CW-pair.
j
(X,A)
i / (X,B) o (X − A, B − A)
k
j̄
(X/A,∗)
ī / (X/A,B/A) o (X/A − ∗, B/A − ∗).
Our goal is to show that the left-hand vertical map induces an isomorphism
in homology. We will achieve this by showing that all the other maps induce
isomorphisms in homology:
• The map k is a homeomorphism of pairs and hence induces an iso-
morphism in homology.
• The map j induces an isomorphism in homology by the assumption (a)
and excision.
133
134 CHAPTER 14. HOMOLOGY OF CELL COMPLEXES
• The map i induces a homomorphism of long exact sequences
take Uα = Dαn − {0}). Hence we can apply the previous result to conclude
a a ∼
=
_
H∗ ( Dαk , Sαk−1 ) −
→ H∗ ( Sαk ,∗).
α α α
Hence we are reduced to calculate the relative homology on the left-hand side.
To do this, we can apply the long exact sequence of a pair to deduce that
a a ∼
=
a
∂ : Hq ( Dαk , Sαk−1 ) −
→ Hq−1 ( Sαk−1 ,∗)
α α α
is an
Lisomorphism for all q. Finally, we know that the latter group is isomorphic
to α∈J Z = Z[J] when q = k and 0 otherwise. □
Now we would like to apply this observation to a cell complex X. If we write
Xk = Skk X for the k-skeleton of X, then we get the following commutative
diagram
` k−1
/ k /
` W k
(22) α Sα α Dα α Sα
f φ φ̄
Xk−1 / Xk = Xk−1 ∪f ( α Dαk ) /
`
Xk /Xk−1 .
where the right-hand vertical map is induced by φ and taking quotients. Since
the restriction of φ to the open interior of the n-disks is a homeomorphism
onto its image, this implies that the dotted arrow φ̄ is a homeomorphism.
Hence we deduce from the previous result on bouquets of spheres:
(
Z[Jn ] if q = k
Hq (Xk , Xk−1 ) ∼
= Hq (Xk /Xk−1 ,∗) ∼
=
0 if q ̸= k
where Jn denotes the indexing set of the attached k-cells.
In other words, the relative homology group Hk (Xk , Xk−1 ) keeps track of
the k-cells of X.
This group will play a crucial role for us today. Let us analyse some conse-
quences of what we have found out about this group.
Let us look at a piece of the long exact sequence of the pair (Xk ,Xk−1 ):
Hq+1 (Xk , Xk−1 ) → Hq (Xk−1 ) → Hq (Xk ) → Hq (Xk , X k−1 ).
For q ̸= k, the last term Hq (Xk , Xk−1 ) = 0 vanishes and hence the map
Hq (Xk−1 ) → Hq (Xk ) is surjective.
136 CHAPTER 14. HOMOLOGY OF CELL COMPLEXES
For q ̸= k − 1, the first term Hq+1 (Xk , Xk−1 ) = 0 vanishes and hence the map
Hq (Xk−1 ) → Hq (Xk ) is injective.
Hence, for a fixed q > 0, we can observe how Hq (Xk ) varies when we let Xk
go through all skeleta of X:
• Hq (X0 ) = 0 since X0 is a discrete set and the higher homology groups
of points vanish.
∼
=
• Still, for X finite-dimensional, since Hq (Xq+1 ) −
→ Hq (X) and since
Hq (Xq ) → Hq (Xq+1 ) → Hq (Xq+1 , Xq ) = 0
is exact, we see that
Hq (Xq ) → Hq (X) is surjective.
137
Now we would like to find an efficient way to calculate the homology of our
cell complex X. Apparently, the group Hn (Xn , Xn−1 ) carries crucial information
about X. Therefore, we are going to give it a new name:
Cellular n-chains
The group of cellular n-chains in a cell complex X is defined to be
Cn (X) := Hn (Xn , Xn−1 ).
We claim that these groups sit inside a sequence of homomorphisms who form
a chain complex. The differential
∂n ) jn−1
Hn−1 (Xn−1 )
where ∂n is the connecting homomorphism in the long exact sequence of pairs and
jn−1 is the homomorphism induced by the inclusion (Xn−1 , ∅) ,→ (Xn−1 , Xn−2 ).
To show that dn ◦ dn+1 = 0 we consider the commutative diagram:
0 = Hn (Xn+1 , Xn )
Since j and ∂ are part of long exact sequences, we know j ◦ ∂ = 0 and get
dn ◦ dn+1 = (jn−1 ◦ ∂n ) ◦ (jn ◦ ∂n+1 ) = 0.
Since the left-hand column in the above big diagram is exact, we know
Hn (Xn )/Im (∂n+1 ) ∼
= Hn (Xn+1 ).
In this theorem we are referring to maps which preserve the skeleton structure
of cell complexes. We should better make this concept precise:
In other words, if we are given two cell complexes and care about their cell
structure, we should only consider filtration-preserving maps between them.
An immediate and very useful consequence of the above theorem is:
140 CHAPTER 14. HOMOLOGY OF CELL COMPLEXES
For example, recall that complex projective n-space CPn has exactly one cell
in each even dimension up to 2n. Hence as an application we can read off the
homology of complex projective space:
(
Z for 0 ≤ k ≤ 2n and k even
Hk (CPn ) =
0 for k odd.
(W
n
( S , ∗).
We have shown that all these maps induce isomorphisms in homology. In partic-
ular,
a a ∼
=
(24) Hn ( Dn , S n−1 ) −
→ Cn (X) = Hn (Xn , Xn−1 ).
We are now going to exploit this fact for the computation of H∗ (RPn ).
Recall that the cell structure of RPn is such that
• Skk (RPn ) = RPk and
• there is exactly one k-cell in each dimension k = 0, . . . ,n. We denote
this k-cell by ek .
Hence the cellular chain complex looks like this:
dn d2 d1
0 / Cn (RPn ) / ··· / C1 (RPn ) / C0 (RPn ) / 0
141
142
CHAPTER 15. COMPUTATIONS OF CELL HOMOLOGIES AND EULER CHARACTERISTIC
In order to compute the homology of this chain complex, we need to determine
the differentials dn :
• We know that H0 (RPn ) is Z = Z[e0 ]. That implies that the differential
d1 must be trivial.
• For k > 1, the differential dk is defined as the top row in the following
commutative diagram
∂k jn−1
Ck (RPn ) = Hk (RPk , RPk−1 ) / Hk−1 (RPk−1 ) / Hk−1 (RPk−1 , RPk−2 ) = Ck−1 (RPn )
O O O
∼
= Hk−1 (π) ∼
=
∂k
Hk (Dk , S k−1 ) / Hk−1 (S k−1 ) / Hk−1 (S k−1 ,∗)
q
(25) RPOk−1 / RPk−1 /RPk−2
π ≈
S k−1 / S k−1
g
q
S k−1
π / RPk−1 / S k−1
5
µ
pinch equator )
S k−1 /S k−2 = S−k−1 ∨ S+k−1 .
143
(S k−1 − S k−2 )+
(q◦π)+
)
α RPk−1
5 − RP
k−2
(q◦π)−
(S k−1 − S k−2 )−
Hence the map µ is the identity on one copy of S k−1 and the an-
tipodal map α on the other copy of S k−1 .
But we know what the effect of Hk−1 (α) is. Namely, it is given by
Hk−1 (α) = (−1)k−1 . Hence
(
2 if k is even
Hk−1 (g) = 1 + (−1)k =
0 if k is odd.
144
CHAPTER 15. COMPUTATIONS OF CELL HOMOLOGIES AND EULER CHARACTERISTIC
Summarizing, we have shown that the cellular chain complex of RPn looks
like:
2 0 0 2 0
0→Z→
− Z→
− ··· →
− Z→
− Z→
− Z → 0 if n is even
0 2 0 2 0
0→Z→
− Z→
− ··· →
− Z→
− Z→
− Z → 0 if n is odd
Homomology of RPn
The homology of real projective n-space is given by
Z k=0
Z k = n is odd
Hk (RPn ) =
Z/2 0 < k < n and k is odd
0 otherwise.
Knowing these maps, up to homotopy, determines the homotopy type of the cell
complex X.
However, homology does not record all of the information of the attaching
maps. For, homology only sees the effect of the composite obtained by pinching
145
out Xn−1 :
Sαn−1 / /
`
α∈Jn Xn−1 Xn−1 /Xn−2
≈
W' w
β∈Jn−1 Sβn−1 .
Hn (φ̄) ∼
= ∼
=
Hn (Xn , Xn−1 ) / Hn−1 (Xn−1 ) / Hn−1 (Xn−1 , Xn−2 ).
∂n jn−1 4
dn
The main result on χ(X) we are going to prove today is that it only de-
pends on the homotopy type of X and is, in particular, independent of the
given cell structure of X. We are going to prove this by showing that χ(X)
can be computed using the singular homology of X.
Recall that we have seen an Euler number for polyhedra in the first lecture. It
was defined as the number of vertices minus the number of edges plus the number
of faces. This fits well with the above definition for a finite cell complex.
For, if we assume the invariance of χ for a moment, then we get χ(S 2 ) = 2
using the standard cell structure on S 2 , i.e., one 0-cell and one 2-cell. This
implies that Euler’s polyhedra formula holds.
therefore isomorphic to Zr for some integer r. The number r is called the rank
of A denoted by rank(A).
In fact, by choosing generators of A/Torsion(A), we can construct a homomo-
prhism A/Torsion(A) → A which splits the projection map A → A/Torsion(A).
Thus if A is finitely generated abelian, then
A∼
= Torsion(A) ⊕ Zr .
We are going to use the following lemma from elementary algebra without
proving it:
Now we are euqipped for the proof of the above mentioned result:
Proof: Let ck be again the number of k-cells in the given finite cell structure
of X. Let C∗ := C∗ (X) denote the cellular chain complex of X. To simplify
the notation let us denote by Z∗ , B∗ , and H∗ the cycles, boundaries and homology,
respectively, in this complex.
148
CHAPTER 15. COMPUTATIONS OF CELL HOMOLOGIES AND EULER CHARACTERISTIC
By their definition, they fit into two short exact sequences:
0 → Zk → Ck → Bk−1 → 0
and
0 → Bk → Zk → Hk → 0,
When we take the sum over all k, the summands rank(Bk ) and rank(Bk−1 )
will cancel out. Thus we get
X
χ(X) = (−1)k rank(Hk ).
k
But by the theorem on the homology of the cellular chain complex, Hk is exactly
the singular homology group Hk (X) of X. QED
Note that the numbers rank(Hk (X) are called the Betti numbers of X.
They had already played an important role in mathematics, before homology
groups had been systematically developed. As the theorem shows, these numbers
are an interesting invariant of a space.
The description of the Euler number in the theorem now generalizes easily:
H̃∗ ( Xα ) ∼
_ M
= H̃∗ (Xα ).
α α
Proof: Let us start with just a single abelian group A. By choosing generators
for A, we can define a surjective homomorphism
F0 → A
151
152
CHAPTER 16. DESIGNING HOMOLOGY GROUPS AND HOMOLOGY WITH COEFFICIENTS
from a free abelian group F0 . The kernel of this homomorphism, denoted by
F1 , is also free, since it is a subgroup of a free abelian group.
We write J0 for a minimal set of generators of F0 such that we have a surjection
F0 → A and J1 for a minimal set of generators of F1 .
For n ≥ 1, we define Xn to be
_
Xn := Sαn .
α∈J0
Taking the disjoint union of all these maps as attaching maps, we get a cell
complex X whose cellular chain complex looks like
0 → F1 → F0 → 0
with F0 in dimension n and F1 in dimension n + 1, and whose homology is
(
A for q = n
H̃q (X) =
0 for q ̸= n.
We write M (A, n) for the CW -complex produced this way and call it a Moore
space of type A and n.
Finally, for a graded abelian group A∗ as in the theorem, we define X to
be the wedge of all the M (An , n). QED
The proof also shows why Topology is particularly well suited for this
endeavour. The gluing construction we used for building cell complexes
is unique for topological spaces. Requiring any additional structure
usually stops us from producing cell complexes.
For example, there are no cell complexes of smooth manifolds or
algebraic varieties. Nevertheless, there are some inventive procedures to
remedy this defect...
More generally, if M is any abelian group, we can form the tensor product
M
Sn (X; M ) = Sn (X) ⊗Z M = M
σ∈Singn (X)
( )
X
= mj σj : σj ∈ Singn (X), mj ∈ M .
j
More explicitly, ∂nM is given by the formula in (26) with rj s replaced with mj s.
Since ∂ ◦ ∂ = 0, we get ∂ M ◦ ∂ M = 0.
f : (X,A) → (Y,B)
Since everything we proved for singular homology was based on these proper-
ties, we can transfer basically all our work to homology with coefficients.
Let us point out two crucial facts:
• The calculations for spheres can be transfered and we get
(
n M for k = n
H̃k (S ; M ) =
0 otherwise.
where the sum is taken over the n-cells of X. As for M = Z, the nth
homology of C∗ (X; M ) is isomorphic to Hn (X; M ).
The reduced homology groups H̃n (X; M ) with coefficients in M are defined
as the homoogy groups of the augmented chain complex
ϵ
. . . → S1 (X; M ) → S0 (X; M ) →
− M →0
P P
where ϵ is the homomorphism which sends j mj σj to j mj ∈ M .
For a homomorphism of groups φ : M → N there is an induced morphism
of chain complexes S∗ (X,A; M ) → S∗ (X,A; N ) which induces a homomorphism
in homology
φ∗ : H∗ (X,A; M ) → H∗ (X,A; N ).
This homomorphism is compatible with f∗ for maps of pairs and with long exact
sequences of pairs.
For the calculations using the cellular chain complex, we need to check the
following lemma:
157
That the outer diagram commutes follows from the way we compute the homology
groups of S n with coefficients Z and M via the Eilenberg-Steenrod axioms. QED
• Why coefficients?
The coefficients that are most often used are the fields Fp , for a prime number
p, and the field Q, R and sometimes C.
In order to get an idea of what happens when we use different coefficients, let
us look at the homology of RPn for R = F2 . We use the cellular chain complex
which looks like this
0 → F2 → F2 → . . . → F2 → F2 → 0.
We showed that the differentials alternated between multiplication by 2 and 0.
But in F2 , 2 = 0 which means that all differentials vanish and we get
(
F2 for 0 ≤ k ≤ n
Hk (RPn ; F2 ) =
0 otherwise.
In other words, away from 2, real projective n-space looks for R-homology
like a point if n is even and like an n-sphere if n is odd.
This teaches us already that different coefficients can tell quite different
stories.
This notwithstanding one might wonder whether integral homology is the
finest invariant and all other homologies are just coarser variations. This is not
the case, and it is indeed possible that homology with coefficients detects
more than integral homology. Let us look at an example:
This example demonstrates that homology groups with coefficients are similar,
but often a bit different than integral homology groups. This raises the question
how different they can be. More generally, we could ask:
Question
Given an R-module M H∗ (X; R), what can we deduce about H∗ (X; M )?
For example, let M be the Z-module Z/m. One might wonder if Hn (X; Z/m)
is just the quotient Hn (X; Z)/mHn (X; Z), since the latter is isomorphic to the
tensor product Hn (X; Z) ⊗ Z/m.
160
CHAPTER 16. DESIGNING HOMOLOGY GROUPS AND HOMOLOGY WITH COEFFICIENTS
But we have to be careful. For, we do have a short exact sequence of chain
complexes
m
0 → S∗ (X; Z) −
→ S∗ (X; Z) → S∗ (X; Z/m) → 0.
Such a short exact sequence induces a long exact sequence of the respective
homology groups a part of which looks like
m m
Hn (X; Z) −
→ Hn (X; Z) → Hn (X; Z/m) → Hn−1 (X; Z) −
→ Hn−1 (X; Z).
Using the exactness of this sequence yields a short exact sequence
(27)
0 → Hn (X; Z)/mHn (X; Z) → Hn (X; Z/m) → m-Torsion(Hn−1 (X; Z)) → 0
where m-Torsion(Hn−1 (X; Z)) denotes the m-torsion, i.e., the kernel of the map
m
Hn−1 (X; Z) −
→ Hn−1 (X; Z) given by multiplication by m.
In fact, the short exact sequence (27) provides a tool to determine Hn (X; Z/m)
when we know both Hn (X; Z) and Hn−1 (X; Z). However, in general, we will
need a more sophisticated method to understand the relationship of H∗ (X; Z)⊗
M and H∗ (X; M ).
As a first generalization, we have the following result:
r
Hn (X,A; M ′ ) / Hn (X,A; M ) / Hn (X,A; M ′′ )
∂
r
Hn−1 (X,A; M ′ ) / ···
LECTURE 17
Our goal for this lecture is to prove the Universal Coefficient Theorem
for singular homology with coefficients. This will require some preparations in
homological algebra. For some this will be a review. Though to keep everybody
on board, this is what we have to do.
We will not treat the most general cases, but rather focus on the main ideas.
Any text book in homological algebra will provide more general results.
• Tensor products
Let A and B be abelian groups. We would like to combine A and B into just
one object, denoted A ⊗ B, in such a way that having a bilinear homomorphism
f: A×B →C
is the same as having a homomorphism from A ⊗ B into C.
That f is bilinear means
f (a1 + a2 ,b) = f (a1 ,b) + f (a2 ,b)
f (a,b1 + b2 ) = f (a,b1 ) + f (a,b2 ).
Tensor product
For, we can construct A ⊗ B as the quotient of the free abelian group
generated by the set A × B modulo the subgroup generated by {(a + a′ ,b) −
(a,b) − (a′ ,b)} and {(a,b + b′ ) − (a,b) − (a,b′ )} for all a,a′ ∈ A and b,b′ ∈ B.
We denote the equivalence class of (a,b) in A ⊗ B by a ⊗ b.
We call A ⊗ B the tensor product of A and B.
%
C.
• The universal property of the tensor product implies that we have an
isomorphism
Aα ) ⊗ B ∼
M M
( = (Aα ⊗ B).
α α
Proof: The composition is zero and if (xi )i is sent to 0 in i Mi′′ , then each
L
xi must be sent to 0 in Mi′′ . Hence each xi comes from some x′i , and hence (xi )i
comes from (x′i )i . We just need to remember to choose x′i = 0 whenever xi = 0.
□
Now let A be any abelian group. As we did in the previous lecture, we choose
a free abelian group F0 mapping surjectively onto A
F0 ↠ A.
The kernel F1 of this map is also free abelian as a subgroup of a free abelian
group. Hence we get an exact sequence of the form
0 → F1 ,→ F0 ↠ A.
Free resolutions
Such an exact sequence with F1 and F0 free abelian groups, is called a free
resolution of A of length two.
(Note that the fact that we can always choose such a free resolution of length
two is particular to the case of abelian groups, i.e., Z-modules. For R-modules
over other rings, one might only be able to find projective resolutions of higher
length. The fact that Z is a principal ideal domain, a PID, does the trick.)
For any abelian group M , tensoring these maps with M yields an exact se-
quence
F1 ⊗ M → F0 ⊗ M → A ⊗ M → 0.
Definition: Tor
The kernel of the map A ⊗ F1 → A ⊗ F0 is called Tor(A,M ). Hence by
definition we have an exact sequence
0 → Tor(A, M ) → A ⊗ F1 → A ⊗ F0 → A ⊗ M → 0.
This group measures how far − ⊗ M is from being exact.
Note that if we replace abelian groups with R-modules over other rings than
Z and take tensor products over R, we might have to consider higher Tor-terms.
Hence we should really write TorZ1 (A, M ) for Tor(A, M ). But we are going to
keep things simple and focus on the idea rather than general technicalities.
It is again time to see some examples:
• If M is a free abelian group, then Tor(A,M ) = 0 for any abelian group
A. That follows from the lemma above.
• For a concrete case, let us try to calculate Tor(Z/4, Z/6). We use the
free resolution
6
Z→
− Z → Z/6 → 0.
Tensoring with Z/4 yields
2
Z/4 →
− Z/4 → Z/4 ⊗ Z/6 → 0
2
where we use 6 = 2 in Z/4. The kernel of Z/4 →
− Z/4 is Z/2. Thus
Tor(Z/4, Z/6) = Z/2.
• More generally, we get
Tor(Z/n, Z/m) = Z/ gcd(n, m)
167
f0
E0 −
→ F0 such that f ◦ p = q ◦ f0 .
Proof: We just apply the previous result to the identity of M to get that,
with whatever resolution we calculate Tor, there is an isomorphism between any
two different ways. And this isomorphism is unique by the theorem on chain
homotopies and their induced maps on homologies. □
There are other properties of Tor the proof of which we are going to omit:
• Tor is functorial: For any homomorphisms of abelian groups A → A′
and M → M ′ , there is a homomorphism
Tor(A,M ) → Tor(A′ ,M ′ ).
• L
Since the direct sum of free resolutions of each Ai is a free resolution of
i Ai , we know that Tor commutes with direct sums:
Ai ,M ) ∼
M M
Tor( = Tor(Ai ,M ),
i i
Proof: We write Zn for the kernel and Bn−1 for the image of the differential
d : Cn → Cn−1 . Since Cn and Cn−1 are free, both Zn and Bn−1 are free as well.
Together with the differential in C∗ , this yields a morphism of short exact
sequences
dn
0 / Zn / Cn / Bn−1 / 0
dn dn dn−1
dn−1
0 / Zn−1 / Cn−1 / Bn−2 / 0.
Since all groups in these chain complexes are free, tensoring with M yields
again a short exact sequence of chain complexes
0 → Z∗ ⊗ M → C∗ ⊗ M → B∗−1 ⊗ M → 0.
CHAPTER
170 17. TENSOR PRODUCTS, TOR AND THE UNIVERSAL COEFFICIENT THEOREM
This can be checked as in the above lemma on direct sums of exact sequences.
Since the differentials in Z∗ and B∗ are trivial, the associated long exact
sequence in homology looks like
∂
n ∂n−1
· · · → Bn ⊗ M −→ Zn ⊗ M → Hn (C∗ ⊗ M ) → Bn−1 ⊗ M −−−→ Zn−1 ⊗ M → · · ·
∂
n
The connecting homomorphism Bn ⊗ M −→ Zn ⊗ M in this sequence is in ⊗ 1,
where in : Bn ,→ Zn denotes the inclusion and 1 denotes the identity on M . This
can be easily checked using the definition of the connecting homomorphism.
A long exact sequence can always be cut into short exact sequences of
the form
0 → Coker(in ⊗ 1) → Hn (C∗ ⊗ M ) → Ker (in−1 ⊗ 1) → 0.
Since the tensor product preserves cokernels, the cokernel on the left-
hand side is just
Coker(in ⊗ 1) = Coker(in ) ⊗ M = Zn /Bn ⊗ M = Hn (C∗ ) ⊗ M.
Finally, to obtain the asserted splitting we use that subgroups of free abelian
groups are free. That implies that sequence (28) splits and we have
Cn ∼
= Zn ⊕ Bn−1 .
Tensoring with M yields
Cn ⊗ M ∼
= (Zn ⊗ M ) ⊕ (Bn−1 ⊗ M ).
Now one has to work a little bit more to get that this induces a direct sum
decomposition in homology. We skip this here. □
Since the singular chain complex S∗ (X,A) is an example of a chain complex
of free abelian groups, the theorem implies:
171
Künneth Theorem
For any pair of spaces X and Y and every n, there is a split short exact
sequence
M M
0→ (Hp (X) ⊗ Hq (Y )) → Hn (X × Y ) → Tor(Hp (X), Hq (Y )) → 0.
p+q=n p+q=n−1
Singular cohomology
173
174 CHAPTER 18. SINGULAR COHOMOLOGY
• As an example, let us look at the case M = Z. Then an n-cochain on X
is just a function which assigns to any n-simplex σ : ∆n → X a number
in Z.
We know that simplices in different dimensions are connected via the face
maps. As for chains, the face maps induce an operator between cochains in dif-
ferent dimensions. But note that, for cochains, the degree will increase instead
of decrease.
Definition: Coboundaries
The coboundary operator
δ : S n (X; M ) → S n+1 (X; M )δ(c)(σ) = c(∂σ)
is defined as follows:
Given an n-cochain c and an n + 1-simplex σ : ∆n+1 → X. Then we define
the n + 1-cochain δ(c) as
n+1
X
δ n (c)(σ) = c(∂n+1 (σ)) = (−1)i c(σ ◦ ϕn+1
i )
i=0
where ϕn+1
i is the ith face map. This defines δ n (c) as a function on
Singn+1 (X).
• Let us have another look at the example from calculus we started with.
A function
f : R2 → R
Then the 1-cochain δ(f ) is the function which assigns to a (smooth) path γ
the number f (γ(1)) − f (γ(0)):
δ(f ) : γ 7→ f (γ(1)) − f (γ(0)).
Back to the general case. The coboundary operator turns S ∗ (X; M ) into a
cochain complex, since δ ◦ δ = 0 which follows from our previous calculation.
For, given an n + 1-simplex σ and an n − 1-cochain c, we get
(δ n ◦ δ n−1 (c))(σ) = (δ n c)(∂n (σ)) = c(∂n ◦ ∂n+1 (σ)) = 0.
Now what does it mean for such a function to be a cocycle? To figure this out
we need to calculate δ(f ). Since δ(f ) is defined on 1-simplices, let σ : ∆1 → X
be a 1-simplex on X. The effect of δ(f ) is to evaulate f on the boundary of
σ:
δ(f )(σ) = f (∂σ) = f (σ(e0 )) − f (σ(e1 )).
Cohomology of a point
If X is just a point, then Singn (pt) consists just of the constant map for
each n. Hence an n-cochain c ∈ S n (pt; M ) is completely determined by its
value mc on the constant map, and therefore Hom(Singn (pt),M ) ∼ = M for all n.
The coboundary operator takes c ∈ Hom(Singn (pt),M ) to the alternating sum
n+1
X n+1
X
i
δ(c) = (−1) c(constant map ◦ ϕni ) = (−1)i mc .
i=0 i=0
(g ◦ f )∗ = f ∗ ◦ g ∗ .
Why cohomology?
At first glance it seems like cohomology and homology are the same
guys, just wrapped up in slightly different cloths and reversing the
arrows. In fact, this is kind of true as we will see in the next lecture.
However, there is also a striking difference which is due to the
innocent looking fact that cohomology is contravariant. We are
going to exploit this fact as follows:
Assume that R is a ring, and let
∆
X−
→ X × X, x 7→ (x,x)
be the diagonal map. It induces a homomorphism in cohomology
∆∗
H ∗ (X × X; R) −→ H ∗ (X; R).
Now we only need to construct a suitable map H p (X; R) ⊗
H q (X; R) → H p+q (X L
× X; R) to get a multiplication on the di-
∗
rect sum H (X; R) = q H q (X; R):
∆∗
H p (X; R) ⊗ H q (X; R) → H p+q (X × X; R) −→ H p+q (X; R).
It will still take some effort to make this idea work. Nevertheless,
this gives us a first idea of how contravariance can be useful.
H ∗ ( Xα ; M ) ∼
a Y
= H ∗ (Xα ; M ).
α α
i∗
A
∗ ∗ ∗
h i
d −i B jA jB d
− H n (X; M ) −−−−→ H n (A; M ) ⊕ H n (B; M ) −−−−−−→ H n (A ∩ B; M ) →
··· → − H n+1 (X; M ) → · · ·
Note that the maps go in the other direction and the degree of
the connecting homomorphism increases.
Here we used the inclusion maps
jA
A ∩ _ B / A _
jB iA
/
B iB
X.
For, the singular chain complexes involved in the above short exact sequence
consist of free abelian groups in each dimension. And exactness is indeed pre-
served by Hom for such sequences.
More precisely, we would like to use the following fact:
We are going to deduce this result from the Universal Coefficient Theorem
in cohomology which we will prove in the next lecture. Roughly speaking, it
will tell us how homology and cohomology are related.
As a first approach, we observe the following phenomenon.
183
Kronecker homomorphism
Thus, applied to the integral singular chain complex and the cochain com-
plex with coefficients in M , this pairing yields a natural homomorphism
κ : H n (X; M ) → Hom(Hn (X),M ),
which sends the class [c] of a cocycle to the homomorphism κ([c]) defined
by
κ([c]) : Hn (X) → M, [σ] 7→ ⟨c, σ⟩ = c(σ).
In the previous lecture, we introuced the singular cochain complex and defined
singular chomology. Along the way we ran into some exact sequences to which
applied the Hom-functor. In particular, we constructed the Kronecker map
Our goal for this lecture is to study the Hom-functor in more detail and to
prove the Universal Coefficient Theorem for singular cohomology which will
tell us that κ is surjective. However, κ is not injective in general, but the UCT
will tell us what the kernel is.
Again, for some this will be a review of known results in homological algebra.
Nevertheless, those who have not seen this before, should get a chance to catch
up.
We will again focus on the main ideas.
Hom(Z,Z/n) ∼
= Z/n, but Hom(Z/n,Z) = 0.
185
186
CHAPTER 19. EXT AND THE UNIVERSAL COEFFICIENT THEOREM FOR COHOMOLOGY
The right-hand map is not necessarily surjective, or in other words, the cok-
ernel of the right-hand map is not necessarily zero.
This leads to the following important definition:
188
CHAPTER 19. EXT AND THE UNIVERSAL COEFFICIENT THEOREM FOR COHOMOLOGY
Definition: Ext
The cokernel of the map Hom(F0 ,M ) → Hom(F1 ,M ) is called Ext(A,M ).
Hence by definition we have an exact sequence
0 → Hom(A,M ) → Hom(F0 ,M ) → Hom(F1 ,M ) → Ext(A,M ) → 0.
Roughly speaking, the group Ext(A,−) measures how far Hom(A,−) is from
being exact.
• For a concrete case, let us calculate Ext(Z/2, Z/2). We use the free
resolution
2
Z→
− Z → Z/2 → 0.
Applying Hom(−,Z/2) yields
2
0 → Hom(Z/2,Z/2) → Hom(Z,Z/2) →
− Hom(Z,Z/2).
This sequence is isomorphic to
2=0
0 → Z/2 → Z/2 −−→ Z/2.
Since 2 = 0 in Z/2, the second map is trivial. Hence the cokernel of
this map is just Z/2. Thus
Ext(Z/2, Z/2) = Z/2.
• More generally, one can show
Ext(Z/n,Z/m) = Z/ gcd(n, m)
where gcd(n, m) denotes the greatest common divisor of n and m.
189
Now we should study Ext in more detail. As a first step we show that it can
be viewed as a cohomology group:
We should check that Ext does not depend on the choice of a free resolution.
To do this, we are going to apply the lemma we proved for the Tor-case which
states that maps can be lifted to resolutions and any two lifts are chain homotopic
in a suitable sense.
0 / E1
i / E0 / A / 0
f1 f0
j
0 / F1 / F0 / A / 0.
0 / E1
i / E0 / A / 0
g1 g0
j
0 / F1 / F0 / A / 0.
QED
s
Ext(C,M ) / Ext(B,M ) / Ext(A,M ) / 0.
0 → E1 ⊕ F1 → E0 ⊕ F0 → B → 0.
By the result of the previous lecture, we can lift the maps in the short exact
sequence to maps of resolutions
0 / E1 / E1 ⊕ F1 / F1 / 0
0 / E0 / E0 ⊕ F0 / F0 / 0
0 / A / B / C / 0
0 0 0
192
CHAPTER 19. EXT AND THE UNIVERSAL COEFFICIENT THEOREM FOR COHOMOLOGY
The horizontal sequences are short exact, since the middle term is a direct
sum of the other terms. Hence we get a short exact sequence of chain complexes
0 → E∗ → E∗ ⊕ F∗ → F∗ → 0.
Since all three complexes consist of free abelian groups, applying Hom(−,M )
yields a short exact sequence of cochain complexes
0 → Hom(F∗ ,M ) → Hom(E∗ ⊕ F∗ ,M ) → Hom(E∗ ,M ) → 0.
• Note that we can always construct a trivial extension by taking the direct
sum of A and M :
(1,0)
0 → A −−→ A ⊕ M → M → 0.
Recall that we say that such a sequence splits.
• The group Ext(A,M ) measures how far extensions of M by A can be from
being from the trivial extension. For, we have
Ext(A,M ) = 0 ⇐⇒ every extension of M by A splits.
Proof: Given an extension, applying Hom(−,M ) yields an exact sequence
Hom(B,M ) → Hom(M,M ) → Ext(A,M ).
1
Thus the identity map M → − M lifts to a map B → M if Ext(A,M ) = 0.
But that is equivalent to that the initial short exact sequence splits. QED
193
• Now one can show in general that Ext(A,M ) is in bijection with the set
of all equivalence classes of extensions of M by A.
• For example, we computed Ext(Z/2,Z/2) = Z/2. The trivial element in
Ext corresponds to the trivial extension
0 → Z/2 → Z/2 ⊕ Z/2 → Z/2 → 0
whereas the non-trivial element corresponds to the extension
2
0 → Z/2 →
− Z/4 → Z/2 → 0.
• Let A be a finitely generated abelian group and let T (A) denote its
torsion subgroup. Since Ext(Z/m,Z) = Z/m, the structure theorem for
finitely generated abelian groups and the previous two points imply that
Ext(A,Z) ∼
= T (A).
Now we prove the main result which connects homology and cohomology and
answers the question we raised last time about the Kronecker map κ:
for all n. These sequences split, but the splitting is not natural.
The proof builds on the same ideas as for the UCT in homology. But let us
do it anyway to get more practice.
Proof: • We write Zn for the kernel and Bn−1 for the image of the differ-
ential d : Cn → Cn−1 . Since Cn and Cn−1 are free, both Zn and Bn−1 are free as
well.
By definition of Zn and Bn , the restriction of the differentials to these groups
vanish. This implies that (Z∗ ,d) and (B∗ ,d) are chain complexes (with trivial
differentials).
Hence we get a short exact sequence of chain complexes
d
(29) 0 → Z∗ → C∗ →
− B∗−1 → 0.
• Since all groups in these chain complexes are free, applying the functor
Hom(−,M ) yields again a short exact sequence of cochain complexes
0 → Hom(B∗−1 ,M ) → Hom(C∗ ,M ) → Hom(Z∗ ,M ) → 0.
This follows from the lemma we proved in the previous lecture.
• Since the differentials in Z∗ and B∗ are trivial, the nth cohomology of
Hom(B∗−1 ,M ) is just Hom(Bn−1 ,M ), and the nth cohomology of Hom(Z∗ ,M ) is
just Hom(Zn ,M ).
Hence the long exact sequence in cohomology associated to the short exact
sequence (29) looks like
∂ d∗ i∗ ∂
· · · → Hom(Zn−1 ,M ) → → H n (Hom(C∗ ,M )) −
− Hom(Bn−1 ,M ) − → Hom(Zn ,M ) →
− Hom(Bn ,M ) → · · ·
∂
• The connecting homomorphism Hom(Zn ,M ) → − Hom(Bn ,M ) in this
sequence is i∗n = Hom(in ,M ), where in : Bn ,→ Zn denotes the inclusion. For, the
connecting homomorphism is defined as follows. Consider the maps
Hom(Cn ,M ) / Hom(Zn ,M )
δ
Hom(Bn ,M ) / Hom(Cn+1 ,M ).
δ
16 M
φ FO
ψ
Zn / CO n φ̃ φ
d
in
/ Bn /
d
Cn+1 Zn
• A long exact sequence can always be cut into short exact sequences of
the from
0 → Coker(Hom(in−1 ,M )) → Hn (C ∗ ) → Ker (Hom(in ,M )) → 0.
Since the functor Hom(−,M ) sends cokernels to kernels, the kernel on the
right-hand side is just
Ker (Hom(in ,M )) = Hom(Coker(in ),M ) = Hom(Zn /Bn ,M ) = Hom(Hn (C∗ ),M ).
Finally, to obtain the asserted splitting we use that subgroups of free abelian
groups are free. That implies that sequence (29) splits and we have
Cn ∼
= Zn ⊕ Bn−1 .
196
CHAPTER 19. EXT AND THE UNIVERSAL COEFFICIENT THEOREM FOR COHOMOLOGY
Applying Hom(−,M ) yields
Hom(Cn ,M ) ∼
= Hom(Zn ,M ) ⊕ Hom(Bn−1 ,M ).
Now one has to work a little bit more to get that this induces a direct sum
decomposition in homology.
It remains to check that the right-hand map in the theorem is in fact the
previously defined map κ. We leave this as an exercise. QED
Now we can prove the result we claimed in the previous lecture:
Proof: Since the construction of the long exact sequence we used in the proof
of the theorem is functorial, we see that φ induces a commutative diagram
The assumption that φ∗ induces an isomorphism implies that the two outer
vertical maps are isomomorphisms. The Five-Lemma implies that the middle
vertical map φ∗ is an isomorphism as well. QED
Our previous oberservations about Ext and torsion subgroups together with
the theorem imply:
197
As a final remark, we mention that there are versions of Ext for the category of
R-modules for any ring. The corresponding Ext-groups ExtR (M,N ) will depend
on the ring R as well as on the modules M and N . Moreover, there might be
non-trivial higher Ext-groups ExtiR (M,N ) for i ≥ 2, in general.
But the theory is very similar to the case of abelian groups, i.e., Z-modules,
as long as R is a principal ideal domain (PID). For, then submodules of free R-
modules are still free over R (which is not true in general). Hence free resolutions
of length two exist, and higher Ext groups vanish also in this case.
For example, fields are examples of PIDs. However, note that, for example,
Ext(Z/2,Z/2) = Ext1Z (Z/2,Z/2) = Z/2 whereas Ext1Z/2 (Z/2,Z/2) = 0. Hence the
base rings matter.
LECTURE 20
199
200 CHAPTER 20. CUP PRODUCTS IN COHOMOLOGY
Similarly, the symbol σ|[ep ,...,ep+q ] refers to the restriction of σ to the back
face of ∆p+q
σ
σ|[ep ,...,ep+q ] : ∆q ,→ ∆p+q →
− X, σ|[ep ,...,ep+q ] (t0 , . . . , tq ) = σ(0, . . . , 0, t0 , . . . , tq ).
For the next proof and the remaining lecture, recall that the notation êi means
that the vertex ei is omitted.
Proof: By definition, for a simplex σ ∈ ∆p+q+1 → X, we have
where the split into the two sums is justified by the fact that the last term of the
first sum is exactly (−1)-times the first term of the second sum.
Now it remains to observe that these two sums are exactly the definition of
(δφ ∪ ψ)(σ) and (−1)p (φ ∪ δψ)(σ). □
We would like this construction to descend to cohomology. Therefore, we
need to check:
201
Proof: We can check this formula already on the level of cochains. For, given
a simplex σ : ∆p+q → X, we get
(f ∗ φ ∪ f ∗ ψ)(σ) = f ∗ φ(σ|[e0 ,...,ep ] )f ∗ ψ(σ|[ep ,...,ep+q ] )
= φ(f ◦ σ|[e0 ,...,ep ] )ψ(f ◦ σ|[ep ,...,ep+q ] )
= (φ ∪ ψ)(f ◦ σ)
= f ∗ (φ ∪ ψ)(σ).
The proof of this result will require some efforts. Before we think about it,
let us collect some consequences of this theorem and of the construction of
the cup product in general.
• Many cup products are trivial just for degree reasons. For classes φ ∈
H p (X; R) and ψ ∈ H q (X; R) with p + q such that H p+q (X; R) = 0, then
φ ∪ ψ = 0 no matter what.
• This can happen for example if X is a finite cell complex.
• If φ ∈ H p (X; R) and p is odd, then
2
φ2 = (−1)p φ2 = −φ2 .
Therefore, 2φ2 = 0 in H 2p (X; R).
If R is torsion-free or if R is a field of characteristic different
from 2, this implies
φ2 = 0.
Proof for a special case: In order to find a strategy for the proof of the
theorem, let us look at a special case. So let [φ], [ψ] ∈ H 1 (X; R), and let σ : ∆2 →
X be a 2-simplex.
203
The respective cup products are then determined by their effect on a 2-simplex
σ : ∆2 → X:
(φ ∪ ψ)(σ) = φ(σ|[e0 ,e1 ] )ψ(σ|[e1 ,e2 ] ).
and
(ψ ∪ φ)(σ) = ψ(σ|[e0 ,e1 ] )φ(σ|[e1 ,e2 ] )
= φ(σ|[e1 ,e2 ] )ψ(σ|[e0 ,e1 ] )
where we use that R is commutative.
Hence in order to show that these two expressions are related, we would like
to reshuffle the vertices. As a first attempt we are going to reverse the order of
all vertices, i.e., we replace σ with σ̄ defined by
σ(ei ) = σ(e2−i ).
Recall that, a long time ago, we showed that reversing the order of vertices
on a 1-simplex corresponds, at least up to boundaries, multiplying the simplex
with (−1).
Hence we should consider inserting a sign as well. So we define maps
ρ1 : S1 (X) → S1 (X) and ρ2 : S2 (X) → S2 (X)
both defined by sending a simplex σ to −σ̄.
Surprisingly, the comparison of the two cup products after taking pullbacks
along the ρs becomes easier. For,
(ρ∗1 φ ∪ ρ∗1 ψ)(σ) = φ(−σ|[e1 ,e0 ] )ψ(−σ|[e2 ,e1 ] )
= φ(σ|[e1 ,e0 ] )ψ(σ|[e2 ,e1 ] )
and
(ρ∗2 (ψ ∪ φ))(σ) = −ψ(σ|[e2 ,e1 ] )φ(σ|[e1 ,e0 ] )
= −φ(σ|[e1 ,e0 ] )ψ(σ|[e2 ,e1 ] )
using that R is commutative.
204 CHAPTER 20. CUP PRODUCTS IN COHOMOLOGY
Hence we get
ρ∗1 φ ∪ ρ∗1 ψ = −ρ∗2 (ψ ∪ φ).
In other words, up to ρ∗1 and ρ∗2 we have shown the desired equality.
Now we remember that we are still on the level of cochains. The theorem is
about an equality of cohomology classes. Hence all we need to show is that ρ∗1
and ρ∗2 will vanish once we pass to cohomology.
This leads to the idea to show that ρ1 and ρ2 are part of a chain map which
is chain homotopic to the identity. So let us try to do this.
First, we want that ρ1 and ρ2 commute with the boundary operator:
(ρ1 ◦ ∂)(σ) = ρ(σ|[e1 ,e2 ] − σ|[e0 ,e2 ] + σ|[e0 ,e1 ] )
= −σ|[e2 ,e1 ] + σ|[e2 ,e0 ] − σ|[e1 ,e0 ] )
= ∂(−σ|[e2 ,e1 ,e0 ] )
= (∂ ◦ ρ2 )(σ).
Now we would like to construct a chain homotopy between ρ and the identity
chain map.
The idea is to interpolate between the identity and ρ by permuting the vertices
one after the other until the order is completely reversed. Then we sum up all
these maps. Along the way we need to introduce some signs.
Before we can define maps, we need to recall the prism operator we used to
construct a chain homotopy which showed that singular homology is homotopy
invariant.
These were maps
pni : ∆n+1 → ∆n × [0,1]
determined by
(
(ek ,0) if 0 ≤ k ≤ i
pni (ek ) =
(ek−1 ,1) if k > i.
Let us write e0k := (ek ,0) and e1k := (ek ,1). Given an n-simplex σ, we would
like to compose it with pni and also permute vertices.
Consider the permutation of simplices
s
∆n+1 −
→i
∆n+1 , (e0 , . . . ,en+1 ) 7→ (e0 , . . . ,ei ,en+1 , . . . ,ei+1 ).
205
This gives
(∂ ◦ h2 + h1 ◦ ∂)(σ) = ρ2 (σ) − σ.
This indicates that ρ1 and ρ2 are part of a chain map which is chain homotopic
to the identity.
or
σ̄(t0 , . . . , tn ) = σ(tn , . . . , t0 ).
Now we observe
(p + q)(p + q + 1) p2 + 2pq + q 2 + p + q
=
2 2
p(p + 1) q(q + 1) 2pq
= + +
2 2 2
p(p + 1) q(q + 1)
= + + pq.
2 2
Thus
ϵp+q = (−1)pq ϵp ϵq .
208 CHAPTER 20. CUP PRODUCTS IN COHOMOLOGY
We conclude from these two computations
ρ∗ φ ∪ ρ∗ ψ = (−1)pq ρ∗ (ψ ∪ φ).
Now we use that ρ is chain homotopic to the identity. That implies that
when we pass to cohomology classes, ρ∗ is the identity and we obtain the desired
identity
φ ∪ ψ = (−1)pq (ψ ∪ φ).
to some vertex while the others remain fixed. Then we throw in some signs to
make things work.
We define homomorphisms hn for each n by
hn : Sn (X) → Sn+1 (X)
X
σ 7→ (−1)i ϵn−i σ|[e0 ,...,ei ,en ,...,ei ] .
i=0
Now we can show by calculating ∂ ◦hn and hn−1 ◦∂ that h is a chain homotopy,
i.e., we have
∂ ◦ hn + hn−1 ◦ ∂ = ρ − id.
We have
X
(∂ ◦ hn )(σ) = ∂( (−1)i ϵn−i σ|[e0 ,...,ei ,en ,...,ei ] )
i=0
X
= (−1)i (−1)j ϵn−i σ|[e0 ,...,êj ,...,ei ,en ,...,ei ]
j≤i
X
+ (−1)i (−1)i+1+n−j ϵn−i σ|[e0 ,...,ei ,en ,...,êj ,...,ei ] .
j≥i
The overlap of the summation indices is necessary. For, at j = i, only the two
sums together yield all the summands we need:
X
ϵn σ|[en ,...,e0 ] + ϵn−i σ|[e0 ,...,ei−1 ,en ,...,ei ]
i>0
X
n+i+1
+ (−1) ϵn−i σ|[e0 ,...,ei ,en ,...,ei+1 ] − σ|[e0 ,...,en ] .
i<n
Now we observe that the two sums in the last expression cancel out, since if
we replace i by i − 1 in the second sum turns the sign into
(−1)n+i ϵn−i+1 = −ϵn−i .
Hence it suffices to show that the terms with j ̸= i in (∂ ◦ hn )(σ) cancel out
with (hn−1 ◦ ∂)(σ). So we calculate
210 CHAPTER 20. CUP PRODUCTS IN COHOMOLOGY
X
(hn−1 ◦ ∂)(σ) = hn−1 ( (−1)j σ|[e0 ,...,êj ,...,en ] )
j=0
X
= (−1)i−1 (−1)j ϵn−i σ|[e0 ,...,êj ,...,ei ,en ,...,ei ]
j<i
X
+ (−1)i (−1)j ϵn−i−1 σ|[e0 ,...,ei ,en ,...,êj ,...,ei ] .
j>i
Since ϵn−i = (−1)n−i ϵn−i−1 , the two sums cancel with the two corresponding
sums in (∂ ◦ hn )(σ). Hence h is a chain homotopy between ρ and the identity.
□
LECTURE 21
More generally, assume we have two open subsets A and B of X. Then the
formula for φ ∪ ψ on cochains implies that cup product yields a map
S n (X,A ∪ B; R) ,→ S n (X,A + B; R)
211
212 CHAPTER 21. APPLICATIONS OF CUP PRODUCTS IN COHOMOLOGY
induces an isomorphism in cohomology. For we have a map of long exact coho-
mology sequences
H n (A ∪ B) / H n (X) / H n (X,A ∪ B) / H n+1 (A ∪ B) / H n+1 (X)
H n (A + B) / H n (X) / H n (X,A + B) / H n+1 (A + B) / H n+1 (X)
where we omit the coefficients. The small chain theorem and ∼
our results on
=
cohomology of free chain complexes imply that H n (A ∪ B; R) −
→ H n (A + B; R)
is an isomorphism for every n. Thus, the Five-Lemma implies that
∼
=
H n (X,A ∪ B; R) −
→ H n (X,A + B; R)
is an isomorphism as well.
Thus composition with this isomorphism gives a cup product map
H p (X,A; R) × H q (X,B; R) → H p+q (X,A ∪ B; R).
Now one can check that all the formulae we proved for the cup product also
hold for the relative cup products.
Cohomology ring
All we are going to say now also works for relative cohomology. But to keep
things simple, we just describe the absolute case.
We will now often drop the symbol ∪ to denote the cup product and just
write
αβ = α ∪ β.
as the direct sum of all cohomology groups. Note that, while the symbol ∗
previously often indicated that something holds for an arbitrary degree, we now
use it to denote the direct sum over all degrees.
The product of two sums is defined as
X X X
( αi )( βj ) = α i βj .
i j i,j
213
This turns H ∗ (X; R) into a ring with unit, i.e., multiplication is associative,
there is a multiplicatively neutral element 1, and addition and multiplication
satisfy the distributive law.
We consider the cohomological degree n in H n (X; R) as a grading of H ∗ (X; R).
If an element α is in H p (X; R) we call p the degree of α and denote it also by |α|.
Since multiplication respects this grading in the sense that it defines a map
H p (X; R) × H q (X; R) → H p+q (X,A; R),
we call H ∗ (X; R) a graded ring.
Moreover, as we have shown with a lot of effort last time, the multiplication
is commutative up to a sign which depends on the grading:
αβ = (−1)|α||β| βα.
Hence H ∗ (X; R) a graded commutative ring.
Moreover, there is an obvious scalar multiplication by elements in R which
turns H ∗ (X; R) into a graded R-algebra.
Finally, if f : X → Y is a continuous map, then the induced map on coho-
mology
f ∗ : H ∗ (Y ; R) → H ∗ (X; R)
is a homomorphism of graded R-algebras.
Now we should determine some ring structures and see what they can tell us.
As a first, though disappointing, example, let us note that the product in
the cohomology of a sphere S n (with n ≥ 1) is boring, since H 0 (S n ; R) is just R
and the product on H n (S n ; R) is trivial for reasons of degrees:
H n (S n ; R) × H n (S n ; R) → H 2n (S n ; R) = 0.
2 d 1 d
0→Z−
→ Z⊕Z−
→ Z→0
where d1 (a,b) = a + b and d2 (s) = (s, − s) (the attaching map of the 2-cell to the
two 1-cells was aba−1 b−1 ). This yields the homology of T .
We can then apply the UCT to deduce that the singular cohomology of T is
given by
Z
if i = 0
i
H (T ; Z) = Z ⊕ Z if i = 1
Z if i = 2
H 1 (T ; Z) = Hom(H1 (T ; Z),Z).
where the funny brackets denote the Kronecker pairing we had defined earlier.
Since multiplication is graded commutative, we have
2α2 = 0 = 2β 2 .
α2 = 0 = β 2 .
It is a cycle, since
∂(σ − τ ) = ∂(σ) − ∂(τ ) = b − d + a − (a − d + b) = 0
where d denotes the diagonal.
Now we can calculate
(α ∪ β)(σ − τ ) = α(σ|[e0 ,e1 ] )β(σ|[e1 ,e2 ] ) − α(τ|[e0 ,e1 ] )β(τ|[e1 ,e2 ] )
= α(a)β(b) − α(b)β(a)
= 1 − 0 = 1.
In general, the exterior algebra ΛR [α1 , . . . ,αn ] over a commutative ring R with
unit is defined as the free R-module with generators αi1 · · · αik for i1 < · · · < ik
with associative and distributive multiplication defined by the rules
αi αj = −αj αi if i ̸= j, and αi2 = 0.
Setting Λ0 = R, ΛR [α1 , . . . ,αn ] becomes a graded commutative ring with odd
degrees for the αi s and unit 1 ∈ R.
216 CHAPTER 21. APPLICATIONS OF CUP PRODUCTS IN COHOMOLOGY
For the n-torus T n = S 1 × · · · × S 1 , defined as the n-fold product of S 1 , we
then get
H ∗ (T n ; Z) = ΛZ [α1 , . . . ,αn ].
The proof of this result requires some efforts. We will postpone its proof and
rather see some consequences of it.
Cup products detect more
Consider the wedge of spheres S 2 ∨ S 4 . We know that its homology is given
by
H̃∗ (S 2 ∨ S 4 ; Z) = H̃∗ (S 2 ; Z) ⊕ H̃∗ (S 4 ; Z).
In other words,
(
Z if i = 0, 2, 4
Hi (S 2 ∨ S 4 ; Z) =
0 else.
Hence the homologies of CP2 and S 2 ∨ S 4 are the same. Since all the groups
are free, this also implies that the cohomology groups of the two spaces are the
same. Thus, neither homology nor cohomology groups can distinguish between
these two spaces.
The cup product, however, can.
217
H̃ ∗ (CP2 ; Z) ∼
= H̃ ∗ (S 2 ∨ S 4 ; Z)
Hopf maps
As an important application of what we just learned, we consider the following
situation.
Many problems can be reduced to checking whether a map is null-homotopic,
i.e., homotopic to a constant map, or not.
Given a map f : X → Y , we can form the mapping cone Cf = CX ∪f Y
(which we introduced in the exercises). It is the pushout of the diagram
X × {1} / CX
f
Y
i / Cf .
pt / CX/(X × {1})
Y
i / SX ∨ Y
where we use that CX/(X × {1}) is the suspension SX of X.
218 CHAPTER 21. APPLICATIONS OF CUP PRODUCTS IN COHOMOLOGY
Thus, if f is null-homotopic, then there is a homotopy equivalence
≃
Cf −
→ SX ∨ Y.
But we just showed that such a homotopy equivalence cannot exits. Thus, η
is not null-homotopic.
Is there a multiplication on Rn ?
For the next application, we are going to assume one more result, namely that
the cohomology ring of the product of RPn × RPn is given
H ∗ (RPn × RPn ; Z/2) ∼
= F2 [α1 ,α2 ]/(α1n+1 ,α2n+1 ).
Theorem: Multiplication on Rn
Assume there is a R-bilinear map
µ : Rn × Rn → Rn
such that µ(x, y) = 0 implies x = 0 or y = 0.
Then n must be a power of 2.
But to show that there are no other such algebra structures on Rn is a much
harder task. The only known proofs of this fact are using algebraic topology!
In fact, for showing this we need to study the famous Hopf Invariant One-
Problem. This is beyond the scope of this lecture. So let us be modest and just
prove the result stated above.
Proof: • Since µ is linear in both variables, it induces a continuous map
µ̄ : RPn−1 × RPn−1 → RPn−1 .
Since both rings are polynomial algebras, µ̄∗ is completely determined by this
identity.
• Since αn = 0, we must have µ̄∗ (α)n = 0, i.e.,
X n
n
(α1 + α2 ) = α1k α2n−k = 0.
k
k
The sum on the right-hand side can only be zero if all the coefficients of the
monomials α1k α2n−k vanish for 0 < k < n. Since we are working over F2 , this
means that all the numbers nk for 0 < k < n must be even.
To prove this fact is equivalent to proving the following claim about the poly-
nomial ring F2 [x]:
• Claim: In F2 [x], we have
(1 + x)n = 1 + xn ⇐⇒ n is a power of 2.
Then
rm r r
(1 + x)n = (1 + x)2 = (1 + x2 )m = 1 + mx2 + . . . + xn ̸= 1 + xn in F2 [x]
since m is odd. □
LECTURE 22
We are going to meet an important class of topological spaces and study one of
their fundamental cohomological properties. This lecture will be short of proofs,
but rather aims to see an important theorem and structures at work.
Manifolds
We start with defining an important class of spaces.
In this lecture, the word manifold will always mean a topological manifold.
You know many examples of manifolds, most notably Rn itself, any open
subset of Rn , n-spheres S n , tori, Klein bottle, projective spaces. Even though the
definition does not refer to this information, any manifold M can be embedded
in some RN for some large N (which depends on M ).
Though it is a crucial point that N and n can and usually are different. For
example, S 2 is a subset of R3 , but each point on S 2 has a neighborhood which
looks like a plane, i.e., is homeomorphic to R2 .
221
222 CHAPTER 22. POINCARÉ DUALITY AND INTERSECTION FORM
There are many reasons why manifolds are important. One of them is that
we understand and can study them locally, while they can be very complicated
globally.
Poincaré duality
In this lecture, all homology and cohomology groups will be with F2 -coefficients.
Recall that there is a pairing
⟨−,−⟩
H k (X; F2 ) ⊗ Hk (X; F2 ) −−−→ F2
is an isomorphism.
Here are some first consequences of this theorem:
• Since cohomology vanishes in negative dimensions, we must have H p (X; F2 ) =
0 for p > n as well.
• Since M is assumed to be compact, we know that π0 (M ), the set of
connected components of M , is finite. Moreover, we once showed that
H 0 (M ; F2 ) equals Map(π0 (M ),F2 ). Hence we get
(Note that we formulated the UCT with the roles of homology and cohomology
reversed. But, since the map arose from the Kronecker pairing, we can also
produce the claimed version of the UCT. As mentioned in the intro to this lecture,
we rush through some points for the sake of telling a good story.)
Composition with the above pairing yields an isomorphism
∼ ∼
H p (X; F2 )
= / Hom(H n−p (M ; F2 ),F2 ) o
=
Hn−p (M ; F2 ).
2
∼
=
224 CHAPTER 22. POINCARÉ DUALITY AND INTERSECTION FORM
Intersection pairing
Combining this isomorphism for different dimensions, we can write the cup
product pairing in cohomology as a pairing in homology (where we drop the
coefficients which are still F2 )
Hp (M ) ⊗ Hq (M )
⋔ / Hp+q−n (M )
∼
= ∼
=
∪ /
H n−p (M ) ⊗ H n−q (M ) H 2n−p−q (M ).
α = i∗ [Y ] and β = j∗ [Z]
Intersection form
Let us look at a particular case of Poincaré duality. Let us assume that M
is even-dimensional, say of dimension n = 2p. Then Poincaré duality implies
that we have a symmetric bilinear form on the F2 -vector space H p (M ):
H p (M ) ⊗F2 H p (M ) → H 2p (M ) ∼
= F2 .
W ⊥ = {v ∈ V : v · w = 0 for all w ∈ W }.
Lemma
The restriction of a nondegenerate symmetric bilinear form on V to a sub-
space W is nondegenerate if and only if W ∩ W ⊥ = 0.
227
We are going to continue the study of forms and get back to topology in the
next lecture.
LECTURE 23
Classification of surfaces
We will first continue the study of bilinear forms, and then use this knowl-
dege to classify all compact connected surfaces, i.e., compact connected two-
dimensional manifolds. Then we are going to contemplate a bit more on Poincaré
duality. In this lecture, all vector spaces, homology and cohomology groups will
be over F2 .
Lemma
Over F2 we have the similarity
0 1 0 1 0 0
1 0 0 ∼ 0 1 0 .
0 0 1 0 0 1
Intersection form
Let us look at a particular case of Poincaré duality. Let us assume that M is
even-dimensional, say of dimension n = 2p. Then Poincaré duality defines a a
231
H p (M ) ⊗F2 H p (M ) → H p (M ).
Intersection form
For a compact manifold of dimension n = 2p, the intersection pairing
Hp (M ; F2 ) ⊗ Hp (M ; F2 ) → F2 , α · β := ⟨a ∪ b, [M ]⟩
defines a nondegenerate symmetric bilinear form on Hp (M ; F2 ), called the
intersection form. Here a and b are Poincaré dual to α and β, respectively.
• We have seen the example of the torus for which the intersection form
is hyperbolic, i.e., can be described in terms of the basis α and β by
the matrix
0 1
H= .
1 0
Note that the open Möbius band itself is a two-dimensional manifold, but
it is not compact. While the closed Möbius is compact, it is not a manifold
according to the definition we stated last time. Though it is a manifold with
boundary. The story is different if we allow boundaries.
Connected sums
There is an interesting geometric operation on manifolds which produces
new ones out of old:
Given two compact connected manifolds M1 and M2 both of dimension n.
Then we can
• cut out a small open n-dimensional disk Dn of each one, and
• sew them together along the resulting boundary spheres S n−1 , i.e., iden-
tify the boundaries via a homeomorphism.
• The resulting space is called the connected sum of M1 and M2 and
is denoted by M1 #M2 . Note M1 #M2 is a connected compact n-
dimensional manifold.
Let us see two examples:
• There is not much happening if we take S 2 #S 2 as it is homeomorphic
to S 2 :
233
Proof: We start with the pair (M1 #M2 , S n−1 ). Since M1 and M2 are man-
ifolds of dimension n, there is an open neighborhood around S n−1 in M1 #M2
which retracts onto S n−1 . Thus, by a result we showed some time ago when we
discussed cell complexes and wedge sums, we know
H∗ (M1 #M2 , S n−1 ) ∼
= H∗ ((M1 #M2 )/S n−1 ,pt) ∼
= H̃∗ (M1 ∨ M2 ).
Now we consider the long exact sequence of the pair (M1 #M2 , S n−1 ):
· · · → H̃i (S n−1 ) → Hi (M1 #M2 ) → Hi (M1 #M2 , S n−1 ) → H̃i−1 (S n−1 ) → · · ·
We also see that Hn (M1 #M2 ) is the kernel of the map in (30).
QED
Proof: Fundamental classes are natural in the sense that the homomorphism
∼
= ∼
=
Hn (M1 #M2 ) −
→ Hn (M1 ∨ M2 ) −
→ Hn (M1 ) ⊕ Hn (M2 ), [M1 #M2 ] 7→ [M1 ] + [M2 ]
sends the fundamental class of [M1 #M2 ] to the sum of the fundamental classes
of M1 and M2 .
Moreover, the cup product is natural so that we get a commutative diagram
∪ ⟨−,[M1 #M2 ]⟩
H p (M1 #M2 ) ⊗ H p (M1 #M2 ) / H 2p (M1 #M2 ) / F2
⟨−,[M1 ]⟩+⟨−,[M2 ]⟩
∪ / /
H p (M1 ) ⊗ H p (M1 ) ⊕ H p (M2 ) ⊗ H p (M2 ) H 2p (M1 ) ⊕ H 2p (M2 ) F2 .
Now it remains to translate this into the intersection pairing in homology which
proves the claim. QED
Classification of surfaces
Motivated by the examples of the torus and real projective plane we are going
to focus now on the case n = 2, i.e., two-dimensional manifolds which we are
235
until Mike Hill, Mike Hopkins, and Douglas Ravenel finally solved the
mystery (almost completely as there is one dimension left, it is 126) in a
groundbreaking work in 2009 (published in 2016) using highly sophisticated
methods in equivariant stable homotopy theory.
LECTURE 24
239
240 CHAPTER 24. MORE ON POINCARÉ DUALITY
Since real projective space is a compact connected n-dimensional manifold,
Poincaré duality applies. And, in fact, we can use this result to deduce the algebra
structure on the cohomology of real projective space:
Lemma
Let f : RPm → RPn be a continuous map which induces a nontrivial map
f∗ ̸= 0 : H1 (RPm ; F2 ) → H1 (RPn ; F2 )
Then m ≤ n.
Proof: First, that π∗ (σ) is a cycle on RPn just follows from the fact
[π(σ(e0 ))] = [−π(σ(e0 ))] = [π(σ(e1 ))] in RPn .
It remains to show that it is not a boundary.
Recall that there is a cell structure on S n with skeleta
S 0 ⊂ S 1 ⊂ · · · ⊂ S n−1 ⊂ S n .
By symmetry, we can assume that p and −p are the points of S 0 .
• For n = 1, we have a homeomorphism RP1 ≈ S 1 (for example, one could
use the stereographic projection). Since σ connects p and −p on S 1 , there is an
242 CHAPTER 24. MORE ON POINCARÉ DUALITY
integer k such that σ walks around S 1 (k + 1/2)-many times. Thus π∗ (σ) walks
around RP1 (2k + 1)-many times, i.e., an odd number of times.
Now recall that we showed
∼
=
→ H1 (S 1 ; Z), m 7→ (z 7→ z m )
Z−
where we use the identification π(S 1 ) = H1 (S 1 ; Z) that we showed in the ex-
ercises. This implies that with F2 -coefficients, even numbers correspond to 0 in
H1 (RP1 ; F2 ) and odd numbers correspond to the nonzero element in H1 (RP1 ; F2 ).
Thus, the image of π∗ (σ) in H1 (RP1 ; F2 ) is nonzero.
• For n > 1, we first choose a path τ on S 1 ⊂ S n which connects p and −p on
S . By the previous case, we know [π∗ (τ )] ̸= 0 in H1 (RP1 ; F2 ). The inclusion
1
Proof: Assume there was such a map f . Since f (−p) = f (−p) for all p, f
induces a map
f¯: RPn+1 → RPn
which fits into a commutative diagram
f
S n+1 / Sn
π n+1 πn
RPn+1 / RPn .
f¯
243
Now we take take a 1-simplex σ which connects two antipodal points on S n+1 .
Its image f∗ (σ) = f ◦ σ is then a 1-simplex which connects two antipodal points
on S n , since f (−p) = f (−p).
By the previous lemma, π∗n (f∗ (σ)) ̸= 0 in H1 (RPn ; F2 ). Thus
f¯∗ (π n+1 (σ)) = π n (f∗ (σ)) ̸= 0.
∗ ∗
In other words,
f¯∗ ̸= 0 : H1 (RPn+1 ; F2 ) → H1 (RPn ; F2 )
is nontrivial. By the other lemma, this is not possible. Hence f cannot exist.
□
Local orientation
The group Hn (M,M − {x}; Z) is often called the local homology of M
at x. It is an infinite cyclic group and therefore has two generators.
A choice of a generator µx ∈ Hn (M,M − {x}; Z) is a local orientation of
M at x.
For every point x ∈ M , we can choose such a generator. Note that such a
choice was not necessary in F2 , since there is only one generator. That makes
F2 -coefficients quite special.
245
The natural question is how all these choices are related. In other words,
is it possible to choose these generators in a compatible way?
More precisely, let x and y be two points in M which both lie in some subset
U ⊂ M . The inclusions ix : {x} ,→ M and iy : {y} ,→ M induce maps
ix∗ iy∗
Hn (M, M − {x}; Z) o Hn (M, M − U ; Z) / Hn (M, M − {y}; Z).
Orientations
Let M be a compact connected n-dimensional manifold.
• An orientation of M is a function x 7→ µx , where µx ∈ Hn (M, M −
{x}; Z) is a generator, which satisfies the following condition:
At any point x ∈ M , there is a neighborhood U around x and
an element µU ∈ Hn (M, M − U ; Z) such that iy∗ (µU ) = µy for all
y ∈ U.
• If such an orientation exists, we say that M is orientable.
The cap product is important for us, since (one form of) Poincaré duality can
be formulated by saying that the cap product with the fundamental class is an
isomorphism:
Poincaré duality
Let M be a compact n-dimensional oriented manifold. Let [M ] ∈ Hn (M ; Z)
be its fundamental class. Then taking the cap product with [M ] yields an
isomorphism
∼
=
D : H p (M ; Z) −
→ Hn−p (M ; Z), φ 7→ φ ∩ [M ].
Note that there are many different ways to formulate Poincaré duality. In
particular, there is also the stronger statement in terms of perfect pairings on
cohomology groups that we have seen in the mod 2-case.
The idea of the proof of this theorem is to study the case of open subsets of
Rn first. Then we use that every point in M has an open neighborhood which is
homeomorphic to an open subset in Rn . Since M is compact, we only need to take
finitely many such open neighborhoods to cover M . The Mayer-Vietoris sequence
then allows to patch the overlapping open subsets together. Unfortunately, there
are some technical difficulties to take care of along the way, e.g., that certain
diagrams actually commute.
Bibliography
[1] Christian Baer, Algebraic Topology, Lecture Notes, Universität Potsdam, Winter Term
2016/17.
[2] Glen E. Bredon, Topology and geometry, Graduate Texts in Mathematics, vol. 139,
Springer-Verlag, New York, 1993. MR1224675
[3] William Fulton, Algebraic topology, Graduate Texts in Mathematics, vol. 153, Springer-
Verlag, New York, 1995. A first course. MR1343250
[4] Allen Hatcher, Algebraic topology, Cambridge University Press, Cambridge, 2002.
MR1867354
[5] J. P. May, A concise course in algebraic topology, Chicago Lectures in Mathematics, Uni-
versity of Chicago Press, Chicago, IL, 1999. MR1702278
[6] Haynes Miller, Lectures on Algebraic Topology, Lecture Notes, MIT, Fall 2016.
[7] John W. Milnor and James D. Stasheff, Characteristic classes, Princeton University Press,
Princeton, N. J.; University of Tokyo Press, Tokyo, 1974. Annals of Mathematics Studies,
No. 76. MR0440554
[8] James R. Munkres, Topology, Prentice Hall, Inc., Upper Saddle River, NJ, 2000. Second
edition of [ MR0464128]. MR3728284
[9] , Elements of algebraic topology, Addison-Wesley Publishing Company, Menlo Park,
CA, 1984. MR755006
[10] Christoph Schweigert, Algebraic Topology, Lecture Notes, Universität Hamburg, Summer
Term 2016.
[11] James W. Vick, Homology theory, 2nd ed., Graduate Texts in Mathematics, vol. 145,
Springer-Verlag, New York, 1994. An introduction to algebraic topology. MR1254439
249
APPENDIX A
Exercises
2 Draw a picture of S 2 as a cell complex with six 0-cells, twelve 1-cells and
eight 2-cells.
251
252 CHAPTER A. EXERCISES
2. Exercises after Lecture 6
1 Let f ∈ C((X,A),(Y,B)) be a map of pairs.
a) Show that, for every n ≥ 0, f induces a homomorphism Hn (X,A) →
Hn (Y,B).
b) Show that the connecting homomorphisms fit into a commuative
diagram
Hn (f )
Hn (X,A) / Hn (Y,B)
∂ ∂
Hn−1 (A) / Hn−1 (B).
Hn−1 (f|A )
3 For every n ≥ 2, show that S n−1 is not a deformation retract of the unit
disk Dn .
5 In this bonus exercise we show that the additivity axiom is needed only
for infinite disjoint unions:
For two topological spaces X and Y , let iX : X ,→ X⊔Y and iY : Y ,→
X ⊔ Y be the inclusions into the disjoint union of X and Y . Without
referring to the additivity axiom show that the remaining Eilenberg-
Steenrod axioms imply that the induced map
Hn (iX ) ⊕ Hn (iY ) : Hn (X) ⊕ Hn (Y ) → Hn (X ⊔ Y )
is an isomorphism for every n. (Hint: You may want to apply the long
exact sequence and excision with U = X ⊂ X ⊔ Y .)
254 CHAPTER A. EXERCISES
3. Exercises after Lecture 7
5 With this exercise we would like to refresh our memory on real projective
spaces and connect it to questions on the existence of fixed points.
Recall from Lecture 2 that the real projective space RPk is defined
to be the quotient of Rk+1 \ {0} under the equivalence relation x ∼ λx
for λ ∈ R \ {0}. The topology on RPk is the quotient topology.
a) Show that any invertible R-linear map F : Rk+1 → Rk+1 induces a
continuous map f : RPk → RPk .
b) Show that for any invertible R-linear map F : Rk+1 → Rk+1 with an
eigenvector, the induced map f : RPk → RPk has a fixed point.
c) Show that any continuous map f : RP2n → RP2n that is induced by
an invertible R-linear map F : R2n+1 → R2n+1 has a fixed point.
4. EXERCISES AFTER LECTURE 9 257
2n−1 2n−1
d) Show that there are continuous maps f : RP → RP without
fixed points.
This will be a guided tour to the fundamental group and the Hurewicz ho-
momorphism. You can read about this topic in almost every textbook. But you
could also take the time to solve the following exercises and enjoy the fun of
developing the maths on your own.
Note that we already seen some of the following problems in previous exercises.
But feel free to do them again. :)
We fix the following notation:
Let X be a nonempty topological space and let x0 be a point in X. We write
I = [0,1] for the unit interval and ∂I = {0,1}.
We denote by
On this exercise set, we will always use the word loop to denote a loop based
at x0 and say that two loops are homotopic when they are homotopic relative to
∂I.
Let γ1 and γ2 be two loops based at x0 . We define the loop γ1 ∗ γ2 to be
the loop given by first walking along γ1 and then walking along γ2 with doubled
speed, i.e., the map
(
γ1 (2s) for 0 ≤ s ≤ 21
γ1 ∗ γ2 : [0,1] → X, s 7→
γ2 (2s − 1) for 12 ≤ s ≤ 1.
1 In this exercise we are going to show that π1 (X,x0 ) together with the
above described operation is a group.
Let γ, γ ′ , ξ, ξ ′ , ζ denote loops based at x0 . Let ϵx0 with ϵx0 (t) = x0
for all t be the constant loop at x0 . Recall that we use the notation
γ ≃ γ ′ to say that the two loops γ and γ ′ are homotopic relative to ∂I.
a) Show that if γ ≃ γ ′ and ξ ≃ ξ ′ , then γ ∗ ξ ≃ γ ′ ∗ ξ ′ .
f1 g
Mf
where f1 maps x to (x,1) and g maps (x,t) to f (x) for all x ∈ X and t ∈ [0,1]
and y ∈ Y to y.
On this exercise set we are going to explore some additional, important topics.
We will use them later in the lectures.
We start with an application of the excision property:
Reduced homology does not convey any new information, but is convenient
for stating things. It also helps focusing on the important information, since it
disregards the contribution in H0 (X) which comes from a single point.
Here are some basic properties of reduced homology:
2 Show the statement of the lecture that the isomorphism between the ho-
mology of the cellular chain complex is functorial in the following sense:
Let f : X → Y be a cellular (or filtration-preserving) map between cell
complexes, i.e., f (Xn ) ⊆ Yn for all n. Show that f induces a homomor-
phism of cellular chain complexes C∗ (f ) : C∗ (X) → C∗ (Y ) which fits into
a commutative diagram
H∗ (C∗ (f ))
H∗ (C∗ (X)) / H∗ (C∗ (X))
∼
= ∼
=
H∗ (X) / H∗ (Y ).
H∗ (f )
4 Consider S 1 with its standard cell structure, i.e. one 0-cell e0 and one
1-cell e1 . Let X be a cell complex obtained from S 1 by attaching two
2-cells e21 and e22 to S 1 by maps f2 and f3 of degree 2 and 3, respectively.
We may express this construction as
For the next exercise, note that if X and Y are cell complexes, then X × Y
is a cell complex with cells the products enα,X × em n
β,Y where eα,X ranges over the
m
cells of X and eβ,Y ranges over the cells of Y .
We start with proving some properties about the Tor-functor that we men-
tioned in the lecture.
f∗ i∗
· · · → H̃n+1 (Cf ; M ) → H̃n (X; M ) −
→ H̃n (Y ; M ) −
→ H̃n (Cf ; M ) → H̃n−1 (X; M ) → · · ·
(Hint: Relate Cf to the mapping cylinder Mf from a previous exer-
cise set.)
b) Show that f induces an isomorphism in homology with coefficients
in M if and only if H̃∗ (Cf ; M ) = 0.
The first part of the next exercise requires some familiarity with Tor beyond
the discussion of the lecture. But you should think about it anyway and definitely
note the statement.
4 Let X be a finite cell complex, and let Fp be a field with p elements. Show
that the Euler characteristic χ(X) can be computed by the formula
X
χ(X) = dimFp (−1)i Hi (X; Fp ).
i
jn * ∂n
H n (Xn ; M )
where ∂ n is the connecting homomorphism in the long exact sequence of
cohomology groups of pairs and j n is the homomorphism induced by the
inclusion (Xn ,∅) ,→ (Xn ,Xn−1 ). Define the cellular cochain complex
of X with coefficients with M to be the cochain complex (C ∗ (X; M ),d∗ ).
Note that the cup product defines a product on the cellular cochain
complex.
a) Show that C ∗ (X; M ) is in fact a complex, i.e., dn ◦ dn−1 = 0.
b) Show that C ∗ (X; M ) is isomorphic to the cochain complex Hom(C∗ (X),M )
where C∗ (X) is the cellular chain complex of X.
(Hint: Remember the Kronecker map κ.)
c) Use the UCT for cohomology and the isomorphism between Hn (X)
and Hn (C∗ (X)) to show
H n (X; M ) ∼
= H n (C ∗ (X; M )).
Note that the isomorphism we produce this way is not functorial.
Recall that we showed that q induces a trivial map on H̃i (−; Z) for
all i.
a) Show H n+1 (X; Z/m) ∼ = Z/m and that H n+1 (q; Z/m) is nontrivial.
(Hint: Use the UCT for cohomology.)
b) Use the previous example to show that the splitting in the UCT for
cohomology cannot be functorial.
(Hint: You need to show that a certain square induced by the UCT
does not commute.)
4 Show that there does not exist a homotopy equivalence between RP3 and
RP2 ∨ S 3 .