Christian Bär

GEOMETRIE IN POTSDAM

Algebraic Topology
Lecture Notes, Summer Term 2022

Version of August 5, 2022

© Christian Bär 2022. All rights reserved.

Picture on title page taken from pixabay.com.

Contents

Preface 1

1. Set Theoretic Topology 3

1.1. Typical problems in topology . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
1.2. Some basic definitions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
1.3. Compactness . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
1.4. Hausdorff spaces . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
1.5. Quotient spaces . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
1.6. Product spaces . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
1.7. Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13

2. Homotopy Theory 15
2.1. Homotopic maps . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
2.2. The fundamental group . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20
2.3. The fundamental group of the circle . . . . . . . . . . . . . . . . . . . . . . . 27
2.4. The Seifert-van Kampen theorem . . . . . . . . . . . . . . . . . . . . . . . . . 40
2.5. The fundamental group of surfaces . . . . . . . . . . . . . . . . . . . . . . . . 51
2.6. Higher homotopy groups . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 58
2.7. Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 74

3. Homology Theory 79
3.1. Singular homology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 79
3.2. Relative homology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 83
3.3. The Eilenberg-Steenrod axioms and applications . . . . . . . . . . . . . . . . 88
3.4. The degree of a continuous map . . . . . . . . . . . . . . . . . . . . . . . . . 96
3.5. Homological algebra . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 106
3.6. Proof of the homotopy axiom . . . . . . . . . . . . . . . . . . . . . . . . . . . 113
3.7. Proof of the excision axiom . . . . . . . . . . . . . . . . . . . . . . . . . . . . 116
3.8. The Mayer-Vietoris sequence . . . . . . . . . . . . . . . . . . . . . . . . . . . 124
3.9. Generalized Jordan curve theorem . . . . . . . . . . . . . . . . . . . . . . . . 126
3.10. CW-complexes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 131
3.11. Homology of CW-complexes . . . . . . . . . . . . . . . . . . . . . . . . . . . 135
3.12. Betti numbers and the Euler number . . . . . . . . . . . . . . . . . . . . . . . 141
3.13. Incidence numbers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 143
3.14. Homotopy versus homology . . . . . . . . . . . . . . . . . . . . . . . . . . . 146
3.15. Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 152

i
Contents

A. Appendix 157
A.1. Free module generated by a set . . . . . . . . . . . . . . . . . . . . . . . . . . 157

List of Figures 161

Bibliography 163

Index 165

ii
Preface
These are the lecture notes of an introductory course on algebraic topology which I taught at
the University of Potsdam during the summer term 2022. The aim was to introduce the basic
tools from homotopy and homology theory. Choices concerning the material had to be made.
Since time was too short for a reasonable discussion of cohomology theory after homology
had been treated, I decided to skip cohomology altogether and instead included more material
from homotopy theory than is often done. In particular, there is a detailed discussion of higher
homotopy groups and the long exact sequence for Serre fibrations.
The necessary prerequisites of the students were rather modest. The course contains a quick
introduction to set theoretic topology but a certain acquaintance with these concepts was certainly
helpful. Familiarity with basic algebraic notions like rings, modules, linear maps etc. was
assumed.
These notes are based on the lecture notes of a course I taught back in 2010. I am grateful to
Volker Branding who wrote a first draft of those lecture notes and created most of the figures
and to Ramona Ziese who improved those notes considerably. Moreover, I would like to thank
the participants of the 2022 course, especially for the valuable feedback that they provided.

Potsdam, August 2022

Christian Bär

1
1. Set Theoretic Topology

1.1. Typical problems in topology

Topology is rough geometry. For example, a sphere is topologically the same as a cube, even
though the sphere is smooth and curved while the cube is piecewise flat and has corners. In
more precise mathematical terms this means that they are homeomorphic. On the other hand,
the sphere is different from a torus even topologically.

Figure 1. (Non-) homeomorphic spaces

Here are four versions of a typical question that one asks in mathematics. Fix integers 1 ≤ 𝑛 < 𝑚.

1.) Does there exist a linear isomorphism 𝜑 : R𝑚 → R𝑛 ?

No, since linear algebra tells us that isomorphic vector spaces have the same dimension.

2.) Does there exist a diffeomorphism 𝜑 : R𝑚 → R𝑛 ?

No, otherwise the differential 𝑑𝜑(0) : R𝑚 → R𝑛 would be a linear isomorphism.

3.) Does there exist a homeomorphism 𝜑 : R𝑚 → R𝑛 ?

We cannot answer that question yet, but we will develop the necessary tools to find the answer.

4.) Does there exist a bijective map 𝜑 : R𝑚 → R𝑛 ?

Yes. We will now explicitly construct an example for such a map in the case 𝑛 = 1 and 𝑚 = 2.

Example 1.1. Since the exponential function maps R bijectively onto R+ = (0, ∞) it suffices to
construct a bijective map 𝜑 : R+ → R+ × R+ .

3
1. Set Theoretic Topology

Given 𝑥 > 0 write 𝑥 as infinite decimal fraction.

𝑥 = . . . 𝑎3 𝑎2 𝑎1 , 𝑏1 𝑏2 𝑏3 . . .
Here 𝑎 𝑗 ∈ {0, 1, . . . , 9} and almost all 𝑎 𝑗 are equal to 0. The 𝑏 𝑗 are blocks of the form 0 . . . 0𝑐 𝑗
with 𝑐 𝑗 ∈ {1, . . . , 9}. This representation of 𝑥 is unique. Now define 𝜑(𝑥) = (𝜑1 (𝑥), 𝜑2 (𝑥))
where
𝜑1 (𝑥) = . . . 𝑎5 𝑎3 𝑎1 , 𝑏1 𝑏3 𝑏5 . . . 𝜑2 (𝑥) = . . . 𝑎6 𝑎4 𝑎2 , 𝑏2 𝑏4 𝑏6 . . .
One easily checks that 𝜑 maps R+ bijectively onto R+ × R+ .
Let us evaluate the construction for an example. Let 𝑥 = 1987, 30500735 . . .. Then we can read
off the 𝑎 𝑗 and the 𝑏 𝑗 as
𝑎1 = 7, 𝑎2 = 8, 𝑎3 = 9, 𝑎4 = 1, 𝑎 𝑗 = 0 for 𝑗 ≥ 5
𝑏1 = 3, 𝑏2 = 05, 𝑏3 = 007, 𝑏4 = 3, 𝑏5 = 5, . . .
Hence 𝜑1 (𝑥) = 97, 30075 . . . and 𝜑2 (𝑥) = 18, 053 . . .

Remark 1.2. In the continuous world counter-intuitive things can happen which are not possible
in the differentiable world. For example, there exist continuous maps
𝜑 : [0, 1] → [0, 1] × [0, 1]
which are surjective. Such a map is called a plane-filling curve.

Example 1.3. The first example goes back to Peano [6]. The following example was shortly
after given by Hilbert [3] and is known as the Hilbert curve. It is defined by
𝜑(𝑥) := lim 𝜑 𝑛 (𝑥)
𝑛→∞

where the 𝜑 𝑛 are defined recursively as indicated in the pictures:

𝜑1 𝜑2 𝜑3

𝜑4 𝜑5 𝜑6

Figure 2. Hilbert’s curve1

1 Basedon an illustration by Zbigniew Fiedorowicz, see

https://commons.wikimedia.org/wiki/File:Hilbert_curve.png

4
 1.1. Typical problems in topology

Remark 1.4. This cannot happen for smooth curves due to a Theorem by Sard which states that
for smooth 𝜑 : [0, 1] → R2 the image 𝜑([0, 1]) ⊂ R2 is a zero set.

Many typical problems in topology are of the following form:

1.) Given two spaces, are they homeomorphic? To show that they are, construct a homeo-
morphism. To show that they are not, find topological invariants, which are different for given
spaces.

2.) Classify all spaces in a certain class up to homeomorphisms.

3.) Fixed point theorems.

Example 1.5 (Classification theorem for surfaces). The classification theorem for surfaces
states that each orientable compact connected surface is homeomorphic to exactly one in the
following infinite list:

𝑆2 (sphere), genus 0

𝑇 2 (torus), genus 1

𝐹2 (Pretzel surface), genus 2

𝐹3 (true pretzel), genus 3

b
b
b

Figure 3. Surface classification2

The genus is the number of “holes” in the surface. We will give a more precise definition later. So
the classification theorem says that to each 𝑔 ∈ N0 there exists an orientable compact connected
surface 𝐹𝑔 with genus 𝑔 and each orientable compact connected surface is homeomorphic to
exactly one 𝐹𝑔 .
2 Images from https://pixabay.com

5
1. Set Theoretic Topology

Example 1.6 (Brouwer fixed point theorem). Any continuous map 𝑓 : 𝐷 𝑛 → 𝐷 𝑛 with
𝐷 𝑛 = {𝑥 ∈ R𝑛 |k𝑥 k ≤ 1} has a fixed point, i.e., there exists an 𝑥 ∈ 𝐷 𝑛 such that

𝑓 (𝑥) = 𝑥.

We give a proof for 𝑛 = 1. Consider the continuous function 𝑔 : [−1, 1] → R defined by

𝑔(𝑥) := 𝑓 (𝑥) − 𝑥. Now | 𝑓 | ≤ 1 implies

𝑔(−1) ≥ −1 − (−1) = 0, 𝑔(1) ≤ 1 − 1 = 0.

By the intermediate value theorem we can find an 𝑥 with 𝑔(𝑥) = 0 which is equivalent to 𝑓 (𝑥) = 𝑥.

1.2. Some basic definitions

First of all let us recall the following

Definition 1.7. A subset 𝑈 ⊂ R𝑛 is called open

iff
∀𝑥 ∈ 𝑈 ∃𝑟 > 0 : 𝐵(𝑥, 𝑟) ⊂ 𝑈
𝑟
where 𝐵(𝑥, 𝑟) = {𝑦 ∈ R𝑛 | 𝑑 (𝑥, 𝑦) = k𝑥 − 𝑦k < 𝑟}. b 𝑥

Figure 4. Open subset

The set of open subsets of R𝑛 is called the standard topology of R𝑛 . We write

TR𝑛 := {𝑈 ⊂ R𝑛 open} ⊂ P (R𝑛 ).

One easily checks

Proposition 1.8. TR𝑛 has the following properties:

(i) ∅, R𝑛 ∈ TR𝑛 ;
Ð
(ii) 𝑈 𝑗 ∈ TR𝑛 , 𝑗 ∈ 𝐽 ⇒ 𝑗∈𝐽 𝑈 𝑗 ∈ TR𝑛 ;

(iii) 𝑈1 , 𝑈2 ∈ TR𝑛 ⇒ 𝑈1 ∩ 𝑈2 ∈ TR𝑛 .

The importance of the concept of open subsets comes from the fact that one can characterize
continuous maps entirely in terms of open subsets. Namely, a map 𝑓 : R𝑛 → R𝑚 is continuous

6
 1.2. Some basic definitions

iff for each 𝑈 ∈ TR𝑚 the preimage 𝑓 −1 (𝑈) is open, 𝑓 −1 (𝑈) ∈ TR𝑛 . This motivates taking open
subsets axiomatically as the starting point in topology.

Definition 1.9. A topological space is a pair (𝑋, T𝑋 ) with T𝑋 ⊂ P (𝑋) such that

(i) ∅, 𝑋 ∈ T𝑋 ;
Ð
(ii) 𝑈 𝑗 ∈ T𝑋 , 𝑗 ∈ 𝐽 ⇒ 𝑗∈𝐽 𝑈 𝑗 ∈ T𝑋 ;

(iii) 𝑈1 , 𝑈2 ∈ T𝑋 ⇒ 𝑈1 ∩ 𝑈2 ∈ T𝑋 .

Definition 1.10. Let (𝑋, T𝑋 ) be a topological space. Elements of T𝑋 are called open subsets
of 𝑋. Subsets of the form 𝑋 \ 𝑈 with 𝑈 ∈ T𝑋 are called closed.

Examples 1.11. 1.) 𝑋 arbitrary set, T𝑋 = P (𝑋) (discrete topology). All subsets of 𝑋 are open
and they are also all closed.

2.) 𝑋 arbitrary set, T𝑋 = {∅, 𝑋 } (coarse topology). Only 𝑋 and ∅ are open subsets of 𝑋. They
are also the only closed subsets.

3.) 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 T𝑌 := {𝑈 ∩ 𝑌 | 𝑈 ∈ T𝑋 }.

4.) Let (𝑋, 𝑑) be a metric space. Then we can imitate the construction of the standard topology
on R𝑛 and define the induced metric 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
centered at 𝑥.

Remark 1.12. Let two metrics 𝑑1 and 𝑑2 be given on a set 𝑋. If 𝑑1 and 𝑑2 are equivalent, i.e.,
∃ 𝐶 ≥ 0 with
𝐶 −1 𝑑2 (𝑥, 𝑦) ≤ 𝑑1 (𝑥, 𝑦) ≤ 𝐶𝑑2 (𝑥, 𝑦) ∀𝑥, 𝑦 ∈ 𝑋,

then 𝑑1 and 𝑑2 induce the same topology.

Remark 1.13. Openness is a relative concept. If 𝑌 ⊂ 𝑋 be an arbitrary subset of a topological

space 𝑋, then 𝑌 is in general not an open subset of 𝑋 but it is always open as a subset of itself
(w.r.t. the induced topology). The same remark applies to closed subsets.

It is now clear how to define continuous maps between topological spaces.

7
1. Set Theoretic Topology

Definition 1.14. Let (𝑋, T𝑋 ) and (𝑌 , T𝑌 ) be topological spaces. Then a map 𝑓 : 𝑋 → 𝑌 is

called continuous iff ∀ 𝑈 ∈ T𝑌 : 𝑓 −1 (𝑈) ∈ T𝑋 .

Example 1.15. If 𝑋 carries the discrete topology then every map 𝑓 : 𝑋 → 𝑌 is continuous.

Example 1.16. If 𝑌 carries the coarse topology then every map 𝑓 : 𝑋 → 𝑌 is continuous.

Remark 1.17. In general, a continuous bijective map need not have a continuous inverse. For
example, let #𝑋 > 1 and consider 𝑓 = id𝑋 : (𝑋, Tdiscrete ) → (𝑋, Tcoarse ).

Remark 1.18. 𝑓 : 𝑋 → 𝑌 is continuous iff 𝑓 −1 ( 𝐴) ⊂ 𝑋 is closed for all closed subsets 𝐴 ⊂ 𝑌 .

Definition 1.19. Let 𝑋 and 𝑌 be topological spaces. A bijective continuous map 𝑓 : 𝑋 → 𝑌 is

called a homeomorphism iff 𝑓 −1 : 𝑌 → 𝑋 is again continuous. If there exists a homeomorphism
𝑓 : 𝑋 → 𝑌 then 𝑋 and 𝑌 are called homeomorphic. We then write 𝑋 ≈ 𝑌 .

Remark 1.20. If 𝑓 : 𝑋 → 𝑌 and 𝑔 : 𝑌 → 𝑍 are continuous then the composition 𝑔 ◦ 𝑓 : 𝑋 → 𝑍

is also continuous.

1.3. Compactness

Definition 1.21. Let 𝑋 be a topological space. A subset 𝑌 ⊂ 𝑋 is called compact iff for

any collection {𝑈𝑖 }𝑖∈ 𝐼 , 𝑈𝑖 ⊂ 𝑋 open, with 𝑌 ⊂ ∪𝑖∈ 𝐼 𝑈𝑖 there exist 𝑖1 , . . . , 𝑖 𝑛 ∈ 𝐼 such that
𝑌 ⊂ 𝑈𝑖1 ∪ · · · ∪ 𝑈𝑖𝑛 .

Examples 1.22. 1.) Finite sets are always compact. Namely, et 𝑌 = {𝑦 1 , . . . , 𝑦 𝑛 } and let {𝑈𝑖 }𝑖∈ 𝐼
be an open cover of 𝑌 . Then choose 𝑖 𝑗 such that 𝑦 𝑖 ∈ 𝑈𝑖 𝑗 . Then {𝑈𝑖1 , . . . , 𝑈𝑖𝑛 } still covers 𝑌 .

2.) If 𝑋 carries the discrete topology then a subset 𝑌 ⊂ 𝑋 is compact if and only if it is finite.
We have already seen that finite sets are always compact. Conversely, let 𝑌 be a compact subset
of 𝑋. We cover 𝑌 by {{𝑦} | 𝑦 ∈ 𝑌 }. These one-point sets are open because 𝑋 carries the discrete
topology. Since 𝑌 is compact this cover must be finite and therefore #𝑌 < ∞.

3.) If 𝑋 carries the coarse topology then every 𝑌 ⊂ 𝑋 is compact

4.) A subset 𝑌 ⊂ R𝑛 is compact iff 𝑌 is closed and bounded (Heine-Borel theorem).

8
 1.4. Hausdorff spaces

Example 1.23. In particular, R is not compact. We can see this directly by looking at the open
cover {(𝑥 − 1, 𝑥 + 1)|𝑥 ∈ R} of R. Since all intervals in this cover have length 2, any finite
subcover can cover only a bounded subset of R.

Proposition 1.24. Let 𝑋 be a compact topological space. Let 𝑌 ⊂ 𝑋 be a closed subset. Then
𝑌 is compact.

Proof. Let {𝑈𝑖 }𝑖∈ 𝐼 be an open cover of 𝑌 . Put 𝑈 := 𝑋 \ 𝑌 ∈ T𝑋 . Then {𝑈, 𝑈𝑖 }𝑖∈ 𝐼 is an open
cover of 𝑋. Since 𝑋 is compact, there exist 𝑖 1 , . . . , 𝑖 𝑛 such that 𝑋 ⊂ 𝑈 ∪ 𝑈𝑖1 ∪ · · · ∪ 𝑈𝑖𝑛 and we
conclude that 𝑌 ⊂ 𝑈𝑖1 ∪ · · · ∪ 𝑈𝑖𝑛 . 

Proposition 1.25. Let 𝑓 : 𝑋 → 𝑌 be continuous. Let 𝐾 ⊂ 𝑋 be compact. Then 𝑓 (𝐾) ⊂ 𝑌 is

compact.

Proof. Let {𝑈𝑖 }𝑖∈ 𝐼 be an open cover of 𝑓 (𝐾), i.e. 𝑓 (𝐾) ⊂ ∪𝑖∈ 𝐼 𝑈𝑖 . Then we have

𝐾 ⊂ 𝑓 −1 ( 𝑓 (𝐾)) ⊂ 𝑓 −1 (∪𝑖∈ 𝐼 𝑈𝑖 ) = ∪𝑖∈ 𝐼 𝑓 −1 (𝑈𝑖 ) .

| {z }
open

Since 𝐾 is compact there exist 𝑖 1 , . . . , 𝑖 𝑛 such that for

𝐾 ⊂ 𝑓 −1 (𝑈𝑖1 ) ∪ · · · ∪ 𝑓 −1 (𝑈𝑖𝑛 ) = 𝑓 −1 (𝑈𝑖1 ∪ · · · ∪ 𝑈𝑖𝑛 )

and we conclude that

𝑓 (𝐾) ⊂ 𝑓 ( 𝑓 −1 (𝑈𝑖1 ∪ · · · ∪ 𝑈𝑖𝑛 )) ⊂ 𝑈𝑖1 ∪ · · · ∪ 𝑈𝑖𝑛 . 

1.4. Hausdorff spaces

Definition 1.26. A topological space 𝑋 is called Hausdorff iff ∀𝑥1 , 𝑥2 ∈ 𝑋 with 𝑥1 ≠ 𝑥2

∃𝑈𝑖 ⊂ 𝑋 open with 𝑥𝑖 ∈ 𝑈𝑖 , such that 𝑈1 ∩ 𝑈2 = ∅.

9
1. Set Theoretic Topology

b b

The Hausdorff property says that any two distinct points 𝑥1 𝑥2

can be separated by disjoint open neighborhoods.

𝑈1 𝑈2
Figure 5. Hausdorff property

Examples 1.27. 1.) Spaces with discrete topology are Hausdorff spaces, simply put 𝑈𝑖 = {𝑥𝑖 }.

2.) Let #𝑋 ≥ 2. Then 𝑋 with the coarse topology is not a Hausdorff space.

3.) If the topology of 𝑋 is induced by a metric 𝑑, then 𝑋 is Hausdorff. Namely, for 𝑥1 ≠ 𝑥2 put
𝑟 := 𝑑 (𝑥1 , 𝑥2 ) > 0. Then the open balls 𝑈𝑖 = 𝐵(𝑥𝑖 , 𝑟/2) separate 𝑥1 and 𝑥2 .

4.) Let 𝑋 be Hausdorff and let 𝑌 ⊂ 𝑋 any subset. Then 𝑌 with its induced topology is again
Hausdorff.

Proposition 1.28. Let 𝑋 be a Hausdorff space. Let 𝑌 ⊂ 𝑋 be a compact subset. Then 𝑌 is a

closed subset.

Proof. Let 𝑝 ∈ 𝑋 \ 𝑌 . Then for every 𝑦 ∈ 𝑌 the Hausdorff property tells us that there exist open
subsets 𝑈 𝑦, 𝑝 , 𝑉𝑦, 𝑝 ⊂ 𝑋 such that 𝑦 ∈ 𝑈 𝑦, 𝑝 , 𝑝 ∈ 𝑉𝑦, 𝑝 and 𝑈 𝑦, 𝑝 ∩ 𝑉𝑦, 𝑝 = ∅. Since 𝑌 is compact
there exist 𝑦 1 , . . . , 𝑦 𝑛 ∈ 𝑌 such that 𝑌 ⊂ 𝑈 𝑦1 , 𝑝 ∪ · · · ∪ 𝑈 𝑦𝑛 , 𝑝 . Now put 𝑉 𝑝 := ∩𝑛𝑗=1𝑉𝑦 𝑗 , 𝑝 . Since
𝑉 𝑝 is a finite intersection of open subsets it is open itself. Moreover, 𝑝 ∈ 𝑉 𝑝 . Now

𝑌 ∩ 𝑉 𝑝 ⊂ (𝑈 𝑦1 , 𝑝 ∪ · · · ∪ 𝑈 𝑦𝑛 , 𝑝 ) ∩ 𝑉 𝑝 ⊂ (𝑈 𝑦1 , 𝑝 ∩ 𝑉𝑦1 , 𝑝 ) ∪ · · · ∪ (𝑈 𝑦𝑛 , 𝑝 ∩ 𝑉𝑦𝑛 , 𝑝 ) = ∅,

hence 𝑉 𝑝 ⊂ 𝑋 \ 𝑌 . Therefore

𝑋 \ 𝑌 = ∪ 𝑝∈𝑋\𝑌 𝑉 𝑝 ⊂ 𝑋 is open.

Thus 𝑌 ⊂ 𝑋 is closed. 

Corollary 1.29. Let 𝑋 be compact, 𝑌 Hausdorff. If 𝑓 : 𝑋 → 𝑌 is continuous and bijective,

then 𝑓 is a homeomorphism.

Proof. We have to show that ∀𝐴 ⊂ 𝑋 closed 𝑓 ( 𝐴) ⊂ 𝑌 is closed. Now let 𝐴 ⊂ 𝑋 be closed.

Since 𝑋 is compact, 𝐴 is compact and also the image 𝑓 ( 𝐴) is compact. Since 𝑌 is a Hausdorff
space we conclude that 𝑓 ( 𝐴) ⊂ 𝑌 is closed. 

10
 1.5. Quotient spaces

1.5. Quotient spaces

Let 𝑋 be a topological space. Let ∼ be an equivalence relation on 𝑋. For any 𝑥 ∈ 𝑋 let [𝑥] be
the equivalence class of 𝑥. Denote the set of equivalence classes by

𝑋/∼ := [𝑥] | 𝑥 ∈ 𝑋

Let 𝜋 : 𝑋 → 𝑋/∼ , 𝑥 ↦→ [𝑥]. Define 𝑈 ⊂ 𝑋/∼ to be open iff 𝜋 −1 (𝑈) ⊂ 𝑋 is open. One can
easily check that this defines a topology on 𝑋/∼ . Observe that 𝜋 : 𝑋 → 𝑋/∼ is continuous by
definition.
We now state the universal property of the quotient topology: For any topological space 𝑌 and
for any maps 𝑓 : 𝑋 → 𝑌 and 𝑓¯ : 𝑋/∼ → 𝑌 such that the diagram

𝑓
𝑋 /𝑌
⑤⑤=
⑤
𝜋 ⑤⑤⑤
 ⑤ ¯
⑤ 𝑓
𝑋/∼

commutes, 𝑓 is continuous iff 𝑓¯ is continuous.

Example 1.30
Let 𝑋 = [0, 1] and let the equivalence relation be given by 𝑥 ∼ 𝑥 ∀𝑥 ∈ 𝑋 and 0 ∼ 1.

This equivalence relation identifies the end points of the

interval. We expect to obtain a topological space homeo- 0 1
morphic to the circle.

Figure 6. Interval with end-

points identified
Indeed, we can construct such a homeomorphism. Consider 𝑓 : 𝑋 → 𝑆1 given by
𝑓 (𝑥) = (cos(2𝜋𝑥), sin(2𝜋𝑥)). Since 𝑓 (0) = 𝑓 (1) there exists a unique 𝑓¯ : 𝑋/∼ → 𝑆1 such
that the diagram
𝑓
[0, 1] / 𝑆1
① ;
①①①
𝜋 ①
①①
 ①① 𝑓¯
[0, 1]/∼

commutes. From the universal property we know that 𝑓¯ is continuous because 𝑓 is continuous.
Moreover, 𝑓¯ : 𝑋/∼ → 𝑆1 is bijective. Since 𝑋 is compact (Heine Borel) 𝜋(𝑋) = 𝑋/∼ is
compact as well. Since R2 is Hausdorff, 𝑆1 is also Hausdorff. Hence Corollary 1.29 applies and
𝑓¯ : 𝑋/∼ → 𝑆1 is a homeomorphism.

11
1. Set Theoretic Topology

If 𝑋 is a topological space and 𝑌 ⊂ 𝑋 a subset then 𝑋/𝑌 := 𝑋/∼ where

𝑥1 ∼ 𝑥2 ⇔ 𝑥1 = 𝑥2 or 𝑥1 , 𝑥2 ∈ 𝑌 .

This equivalence relation identifies all points in 𝑌 to one point and performs no further identifi-
cations. Example 1.30 is of this form.

Example 1.31. R/ [0,1] is homeomorphic to R.

Example 1.32. R/ (0,∞) is not a Hausdorff space because the points [0] and [1] cannot be
separated.

1.6. Product spaces

Let 𝑋1 , . . . , 𝑋𝑛 be topological spaces. We set

𝑋 := 𝑋1 × · · · × 𝑋𝑛

Definition 1.33. A subset 𝑈 ⊂ 𝑋 is called open (for the product topology) iff for all
𝑝 = ( 𝑝 1 , . . . , 𝑝 𝑛 ) ∈ 𝑈 there exist open subsets 𝑈𝑖 ⊂ 𝑋𝑖 with 𝑝 𝑖 ∈ 𝑈𝑖 and 𝑈1 × · · · × 𝑈𝑛 ⊂ 𝑈.

It is easy to check that this defines a topology on 𝑋.

Examples 1.34. 1.) If all 𝑋𝑖 carry the discrete topology then 𝑋 carries the discrete topology.
Namely, the sets {( 𝑝 1 , . . . , 𝑝 𝑛 )} are open, hence all subsets of the product space are open.

2.) If all 𝑋𝑖 carry the coarse topology then 𝑋 carries the coarse topology.

3.) R𝑛 ×R𝑚 ≈ R𝑛+𝑚 . To see this it is convenient to use the maximum metric on R𝑛 to characterize
the standard topology.

Now we list some important properties of the product topology:

1.) The projection maps 𝜋𝑖 : 𝑋 → 𝑋𝑖 , 𝜋𝑖 (𝑥) = 𝑥𝑖 , are continuous because for any open subset
𝑈𝑖 ⊂ 𝑋𝑖
𝜋𝑖−1 (𝑈𝑖 ) = 𝑋1 × · · · × 𝑋𝑖−1 × 𝑈𝑖 × 𝑋𝑖+1 × · · · × 𝑋𝑛
is open in 𝑋.

2.) Fix 𝑥𝑖 ∈ 𝑋𝑖 where 𝑖 ∈ {1, . . . , 𝑖 0 − 1, 𝑖 0 + 1, . . . , 𝑛}. Then the map 𝜄 : 𝑋𝑖0 → 𝑋 is continuous
where 𝜄(𝜉) = (𝑥1 , . . . , 𝑥𝑖0 −1 , 𝜉, 𝑥𝑖0 +1 , . . . , 𝑥 𝑛 ).

12
 1.7. Exercises

3.) Universal property: for all topological spaces 𝑌 and for all maps

𝑓 = ( 𝑓 1 , . . . , 𝑓 𝑛 ) : 𝑌 → 𝑋 = 𝑋1 × · · · × 𝑋 𝑛

the map 𝑓 is continuous if and only if all 𝑓𝑖 : 𝑌 → 𝑋𝑖 are continuous.

4.) If all 𝑋𝑖 are compact then 𝑋 is also compact.

5.) If all 𝑋𝑖 are Hausdorff then 𝑋 is also Hausdorff.

1.7. Exercises
1.1. Determine all topologies on the set {1, 2, 3} and investigate which ones are homeomorphic.

1.2. Let 𝑊 𝑛 = [−1, 1] × · · · × [−1, 1] ⊂ R𝑛 and 𝐷 𝑛 = {𝑥 ∈ R𝑛 | k𝑥 k ≤ 1}. Show that 𝐷 𝑛 and

𝑊 𝑛 are homeomorphic.

1.3. Let T be a topology on R which contains all half-open intervals (𝑥, 𝑦] and [𝑥, 𝑦), 𝑥 < 𝑦.
Show that T is the discrete topology.

1.4. Show that (0, 1) and [0, 1]

a) are not homeomorphic when equipped with the standard topologies induced by R;

b) are homeomorphic when equipped with the discrete topology.

1.5. Let 𝐷 𝑛 be as in Exercise 1.2. Let 𝑥, 𝑦 ∈ 𝐷˚ 𝑛 , i.e., k𝑥 k, k𝑦k < 1. Construct a homeomorphism
𝜑 : 𝐷 𝑛 → 𝐷 𝑛 with 𝜑(𝑥) = 𝑦 and 𝜑(𝑧) = 𝑧 for all 𝑧 ∈ 𝜕𝐷 𝑛 , i.e., k𝑧k = 1.

1.6. Let 𝑋 be a topological space, let “∼” be an equivalence relation on 𝑋. On page 11 we

defined that 𝑈 ⊂ 𝑋/∼ is called open if and only if 𝜋 −1 (𝑈) ⊂ 𝑋 is open where 𝜋 : 𝑋 → 𝑋/∼ is
the standard projection.

a) Show that 𝑋/∼ equipped with this system of open sets is a topological space.

b) Show that the universal property on page 11 determines the topology on 𝑋/∼ uniquely, i.e.,
if T is a topology on 𝑋/∼ such that 𝜋 : (𝑋, T𝑋 ) → (𝑋/∼, T ) is continuous and if the universal
property holds for (𝑋/∼, T ) then T is the topology defined above.

1.7. Let 𝐷 𝑛 and 𝑊 𝑛 be as in Exercise 1.2. Show:

a) 𝐷 𝑛 /𝜕𝐷 𝑛 ≈ 𝑆 𝑛 ;

b) 𝑊 𝑛 /𝜕𝑊 𝑛 ≈ 𝑆 𝑛 .

13
1. Set Theoretic Topology

1.8. Let 𝑋 = R/Q, i.e., the quotient space of R under the equivalence relation 𝑥 ∼ 𝑦 iff 𝑥 − 𝑦 ∈ Q.
Show that 𝑋 has uncountably many elements but carries the coarse topology.

1.9. Let 𝑋 = R2 /Z2 , i.e., the quotient space of R2 under the equivalence relation 𝑥 ∼ 𝑦 iff
𝑥 − 𝑦 ∈ Z2 . Let 𝑌 = 𝑆1 × 𝑆1 with the product topology. Show that 𝑋 and 𝑌 are homeomorphic.

1.10. Let 𝑋 be a topological space. One obtains the cone 𝐶 𝑋 over 𝑋 by considering the cylinder
𝑋 × [0, 1] and then passing to the quotient 𝐶 𝑋 = (𝑋 × [0, 1])/∼ where the equivalence relation
∼ is given by (𝑥, 𝑡) ∼ (𝑥 ′ , 𝑡 ′ ) if and only if (𝑥, 𝑡) = (𝑥 ′ , 𝑡 ′ ) or 𝑡 = 𝑡 ′ = 1. Show:

a) 𝐶𝑆 𝑛 ≈ 𝐷 𝑛+1 .

b) If 𝑋 is compact so is 𝐶 𝑋.

c) If 𝑋 is Hausdorff so is 𝐶 𝑋.

1.11. Let 𝑋 be a topological space. One obtains the suspension Σ𝑋 of 𝑋 as the quotient
Σ𝑋 = (𝑋 × [0, 1])/∼ where the equivalence relation ∼ is given by (𝑥, 𝑡) ∼ (𝑥 ′ , 𝑡 ′ ) if and only if
(𝑥, 𝑡) = (𝑥 ′ , 𝑡 ′ ) or 𝑡 = 𝑡 ′ = 1 or 𝑡 = 𝑡 ′ = 0.
If furthermore 𝑓 : 𝑋 → 𝑌 is a map then 𝑓 × id : 𝑋 × [0, 1] → 𝑌 × [0, 1] induces a map
Σ 𝑓 : Σ𝑋 → Σ𝑌 . Show:

a) If 𝑓 is continuous so is Σ 𝑓 .

b) Σ𝐷 𝑛 ≈ 𝐷 𝑛+1 .

c) Σ𝑆 𝑛 ≈ 𝑆 𝑛+1 .

14
2. Homotopy Theory

2.1. Homotopic maps

Definition 2.1. Let 𝑋 and 𝑌 be topological spaces. Let 𝐴 ⊂ 𝑋 be a subset. Two continuous
maps 𝑓0 , 𝑓1 : 𝑋 → 𝑌 are called homotopic relative to 𝐴 iff there exists a continuous map
𝐹 : 𝑋 × [0, 1] → 𝑌 such that

(i) 𝐹 (𝑥, 0) = 𝑓0 (𝑥) ∀𝑥 ∈ 𝑋;

(ii) 𝐹 (𝑥, 1) = 𝑓1 (𝑥) ∀𝑥 ∈ 𝑋;

(iii) 𝐹 (𝑎, 𝑡) = 𝑓0 (𝑎) ∀𝑎 ∈ 𝐴, ∀𝑡 ∈ [0, 1].

In symbols, 𝑓0 ≃ 𝐴 𝑓1 . The map 𝐹 is then called a homotopy relative to 𝐴.

Remark 2.2. If 𝑓0 ≃ 𝐴 𝑓1 then 𝑓0 𝐴

= 𝑓1 𝐴 .

Remark 2.3. If 𝐴 = ∅, then we say “ 𝑓0 and 𝑓1 are homotopic” instead of “ 𝑓0 and 𝑓1 are homotopic
relative to ∅”. Similarly, we write “ 𝑓0 ≃ 𝑓1 ” instead of “ 𝑓0 ≃ ∅ 𝑓1 ”.

Examples 2.4. 1.) Let 𝑓0 , 𝑓1 : 𝑋 → R𝑛 be continuous with 𝑓0 𝐴 = 𝑓1 𝐴. Put

𝑛
𝐹 : 𝑋 × [0, 1] → R with 𝐹 (𝑥, 𝑡) := 𝑡 𝑓1 (𝑥) + (1 − 𝑡) 𝑓0 (𝑥). Then 𝐹 is a homotopy from 𝑓0
to 𝑓1 relative to 𝐴, hence 𝑓0 ≃ 𝐴 𝑓1 .

2.) Let 𝑓0 : R𝑛 → 𝑌 be continuous. Put 𝐹 (𝑥, 𝑡) := 𝑓0 ((1 − 𝑡)𝑥), then 𝑓0 ≃ const map.

3.) Lef 𝑓 = Exp : R → 𝑆1 ∈ C, Exp(𝑥) = 𝑒2 𝜋𝑖𝑥 . From the previous example we know
𝑓 ≃ const map, but we will shortly see that 𝑓 ;Z const map.
Given two topological spaces 𝑋, 𝑌 we set

𝐶 (𝑋, 𝑌 ) := { 𝑓 : 𝑋 → 𝑌 | 𝑓 is continuous}.

Lemma 2.5. Let 𝑋, 𝑌 be topological spaces, let 𝐴 ⊂ 𝑋 and let 𝜑 ∈ 𝐶 ( 𝐴,𝑌 ). Then “≃ 𝐴” is
an equivalence relation on { 𝑓 ∈ 𝐶 (𝑋, 𝑌 ) | 𝑓 | 𝐴 = 𝜑}.

15
2. Homotopy Theory

Proof. a) “≃ 𝐴” is reflexive:
𝑓 ≃ 𝐴 𝑓 , because we can put 𝐹 (𝑥, 𝑡) := 𝑓 (𝑥).

b) “≃ 𝐴” is symmetric:
Let 𝑓 ≃ 𝐴 𝑔. We have to show 𝑔 ≃ 𝐴 𝑓 . Let 𝐹 : 𝑋 × [0, 1] → 𝑋 be a homotopy relative to 𝐴
from 𝑓 to 𝑔. Put 𝐺 (𝑥, 𝑡) := 𝐹 (𝑥, 1 − 𝑡). This is a homotopy relative to 𝐴 from 𝑔 to 𝑓 , therefore
𝑔 ≃𝐴 𝑓 .

c) “≃ 𝐴” is transitive:
Let 𝑓 ≃ 𝐴 𝑔 and 𝑔 ≃ 𝐴 ℎ. We have to show 𝑓 ≃ 𝐴 ℎ. Let 𝐹 : 𝑋 × [0, 1] → 𝑌 be homotopy relative
to 𝐴 from 𝑓 to 𝑔 and let 𝐺 : 𝑌 × [0, 1] → 𝑋 be homotopy relative to 𝐴 from 𝑔 to ℎ. Then put
(
𝐻 : 𝑋 × [0, 1] → 𝑌 ,
𝐹 (𝑥, 2𝑡), 0 ≤ 𝑡 ≤ 1/2
𝐻 (𝑥, 𝑡) :=
𝐺 (𝑥, 2𝑡 − 1), 1/2 ≤ 𝑡 ≤ 1

This is a homotopy relative to 𝐴 from 𝑓 to ℎ, thus 𝑓 ≃ 𝐴 ℎ. 

Lemma 2.6. Let 𝑋, 𝑌 , 𝑍 be topological spaces. Let 𝐴 ⊂ 𝑋, 𝐵 ⊂ 𝑌 be subsets. Let 𝑓0 , 𝑓1 ∈

𝐶 (𝑋,𝑌 ) be such that 𝑓0 ≃ 𝐴 𝑓1 , 𝑓𝑖 ( 𝐴) ⊂ 𝐵 and let 𝑔0 , 𝑔1 ∈ 𝐶 (𝑌 , 𝑍) be such that 𝑔0 ≃ 𝐵 𝑔1 .
Then 𝑔0 ◦ 𝑓0 ≃ 𝐴 𝑔1 ◦ 𝑓1 .

Proof. Let 𝐹 : 𝑋 × [0, 1] → 𝑌 be homotopy relative to 𝐴 from 𝑓0 to 𝑓1 and 𝐺 : 𝑌 × [0, 1] → 𝑍

be homotopy relative to 𝐵 from 𝑔0 to 𝑔1 Then 𝐻 : 𝑋 × [0, 1] → 𝑍 is a homotopy relative to 𝐴
from 𝑔0 ◦ 𝑓0 to 𝑔1 ◦ 𝑓1 where 𝐻 (𝑥, 𝑡) = 𝐺 (𝐹 (𝑥, 𝑡), 𝑡). 

Definition 2.7. A map 𝑓 ∈ 𝐶 (𝑋, 𝑌 ) is called a homotopy equivalence iff there exists a 𝑔 ∈

𝐶 (𝑌 , 𝑋) such that 𝑔◦ 𝑓 ≃ id𝑋 and 𝑓 ◦𝑔 ≃ id𝑌 If there exists a homotopy equivalence 𝑓 : 𝑋 → 𝑌
then 𝑋 and 𝑌 are homotopy equivalent. In symbols, 𝑋 ≃ 𝑌 .

Remark 2.8. This defines an equivalence relation on the class of topological spaces.

Remark 2.9. Obviously, homeomorphic implies homotopy equivalent, in short,

𝑋 ≈ 𝑌 ⇒ 𝑋 ≃ 𝑌.

Example 2.10. Euclidean space is homotopy equivalent to a point, R𝑛 ≃ {0}. Namely, put
𝑓 : {0} → R𝑛 , 𝑓 (0) = 0, and 𝑔 : R𝑛 → {0}, 𝑔(𝑥) = 0. Then 𝑔 ◦ 𝑓 = id {0} and 𝑓 ◦ 𝑔 ≃ idR𝑛 by
Example 2.4.1. Remark 2.8 implies R𝑛 ≃ R𝑚 for all 𝑛, 𝑚 ∈ N.

16
 2.1. Homotopic maps

Definition 2.11. A topological space is called contractible iff it is homotopy equivalent to

{point}.

Definition 2.12. Let 𝐴 ⊂ 𝑋 and let 𝜄 : 𝐴 → 𝑋 be the inclusion map. Then 𝐴 is called

1.) a retract of 𝑋 iff there exists 𝑟 ∈ 𝐶 (𝑋, 𝐴) such that 𝑟 𝐴

= id 𝐴, i.e. 𝑟 ◦ 𝜄 = id 𝐴. Then 𝑟 is
called a retraction from 𝑋 to 𝐴.

2.) a deformation retract of 𝑋 iff there exists a retraction 𝑟 : 𝑋 → 𝐴 such that 𝜄 ◦ 𝑟 ≃ id𝑋 .

3.) a strong deformation retract of 𝑋 iff there exists a retraction 𝑟 : 𝑋 → 𝐴 such that

𝜄 ◦ 𝑟 ≃ 𝐴 id𝑋 .

Examples 2.13. 1.) Let 𝑋 any topological space, let 𝐴 = {𝑥0 } ⊂ 𝑋 consist of just one point.
Then 𝑟 : 𝑋 → 𝐴, 𝑟 (𝑥) = 𝑥0 , is a retraction, hence 𝐴 is a retract of 𝑋. The one-pointed set 𝐴 is a
deformation retract of 𝑋 iff 𝑋 is contractible.

2.) Let 𝑋 = R𝑛+1 \ {0} and 𝐴 = 𝑆 𝑛 . Consider the map 𝑟 : 𝑋 → 𝐴 with 𝑟 (𝑥) = k 𝑥𝑥 k .
The composition 𝜄 ◦ 𝑟 : R𝑛+1 \ {0} → R𝑛+1 \ {0} then satisfies 𝜄 ◦ 𝑟 = k 𝑥𝑥 k . The map 𝐹 ∈
𝐶 (R𝑛+1 \ {0} × [0, 1], R𝑛+1 \ {0}) given by

𝑥
𝐹 (𝑥, 𝑡) = 𝑡𝑥 + (1 − 𝑡)
k𝑥 k

is continuous and satisfies

𝐹 (𝑥, 0) = (𝜄 ◦ 𝑟) (𝑥), 𝐹 (𝑥, 1) = id(𝑥)

for all 𝑥 ∈ R𝑛+1 \ {0} and

𝐹 (𝑎, 𝑡) = 𝑎, 𝑎 ∈ 𝑆𝑛

We thus conclude that 𝜄 ◦ 𝑟 ≃𝑆 𝑛 idR𝑛+1 \{0} , hence 𝑆 𝑛 is a strong deformation retract of R𝑛+1 \ {0}.

The difference between a deformation retract and a strong deformation retract is rather subtle.

17
2. Homotopy Theory

𝐴
Example 2.14 1
We consider the comb space given by
n o
𝑋 = (𝑥, 𝑦) ∈ R2 | 0 ≤ 𝑦 ≤ 1 and (𝑥 = 0 or 𝑥 = 1
for some 𝑛 ∈ N) ......

𝑛

∪ (𝑥, 𝑦) ∈ R2 | 𝑦 = 0 and 0 ≤ 𝑥 ≤ 1

The space 𝑋 is a bounded and closed subset of R2 , hence compact. Let

the set 𝐴 be given by 𝐴 = {(0, 1)}. 𝑍
1 1 1
0 4 3 2 1
Figure 7. Comb space

First we show that 𝐴 is a deformation retract of 𝑋. The map 𝐹 : 𝑋 × [0, 1] → 𝑋 given by

𝐹 (𝑥, 𝑦, 𝑡) := (𝑥, (1 − 𝑡)𝑦) is continuous and satisfies

𝐹 (𝑥, 𝑦, 0) = (𝑥, 𝑦), 𝐹 (𝑥, 𝑦, 1) = (𝑥, 0) .

Therefore id𝑋 ≃ Inclusion 𝑍→𝑋 ◦ 𝜋 where 𝜋 : 𝑋 → [0, 1] × {0} =: 𝑍 is the projection 𝜋(𝑥, 𝑦) =
(𝑥, 0). Moreover, we have 𝜋 ◦ Inclusion 𝑍→𝑋 = id 𝑍 . This shows that 𝜋 is a homotopy equivalence
between 𝑋 and 𝑍. Hence 𝑋 ≃ 𝑍 ≈ [0, 1] ≃ {pt}, which means that 𝑋 is contractible. Therefore
𝐴 is a deformation retract of 𝑋.

Now we show that 𝐴 is not a strong deformation retract of 𝑋. Suppose it were, then there would
exist a continuous map 𝐺 : 𝑋 × [0, 1] → 𝑋, such that for all 𝑡 ∈ [0, 1] and all (𝑥, 𝑦) ∈ 𝑋

𝐺 (𝑥, 𝑦, 0) = (𝑥, 𝑦), 𝐺 (0, 1, 𝑡) = (0, 1).

Since 𝑋 × [0, 1] is compact the map 𝐺 would be uniformly continuous. Therefore for 𝜀 = 1/2
we can find 𝛿 > 0 such that

k𝐺 (𝑥, 𝑦, 𝑡) − 𝐺 (𝑥 ′ , 𝑦 ′ , 𝑡 ′ ) k < 1
2

whenever |𝑥 − 𝑥 ′ | < 𝛿, |𝑦 − 𝑦 ′ | < 𝛿 and |𝑡 − 𝑡 ′ | < 𝛿.

18
 2.1. Homotopic maps

𝐴
1 1
Now choose 𝑛 so large that 𝑛 < 𝛿 and consider
 
(𝑥, 𝑦) = 1
𝑛, 1 , (𝑥 ′ , 𝑦 ′ ) = (0, 1), 𝑡 = 𝑡′.

Then
 
1
𝐺 𝑛 , 1, 𝑡 − 𝐺 (0, 1, 𝑡) < 21 , ∀𝑡 ∈ [0, 1] .

Hence 𝐺 ( 𝑛1 , 1, 𝑡) ∈ 𝐵((0, 1), 21 ) for all 𝑡 ∈ [0, 1]. 𝑍

1
0 𝑛 1
Figure 8. 𝐴 is not a strong
deformation retract
of 𝑋

On the other hand the mapping 𝑡 ↦→ 𝐺 ( 𝑛1 , 1, 𝑡) is a continuous path in 𝑋 from ( 𝑛1 , 1) to (0, 1) and
must take values in 𝑍 for some 𝑡.

Remark 2.15. We have the following scheme of implications:

𝐴 is a strong deformation retract of 𝑋

⇐

𝐴 is a deformation retract of 𝑋
⇐
⇐

:
𝐴 is a retract of 𝑋 ; 𝐴≃𝑋

That both possible implications in the bottom row do not hold in general can be seen by
counterexamples. Let 𝐴 = {𝑥0 } be a point in 𝑋 and let 𝑋 be not contractible. Then 𝐴 is a retract
of 𝑋 but 𝐴 and 𝑋 are not homotopically equivalent. This is a counterexample for the implication
“ ⇒ ”.

A counterexample for the other direction “ ⇐ ” is given by 𝑋 = [0, 1] × [0, 1] ⊂ R2 and

𝐴 = comb space. Then 𝑋 and 𝐴 are contractible, hence 𝑋 ≃ 𝐴, but one can show that there is
no retraction 𝑋 → 𝐴.

19
2. Homotopy Theory

2.2. The fundamental group

Definition 2.16. Let 𝑋 be a topological space and let 𝑥0 , 𝑥1 , 𝑥2 ∈ 𝑋. Put

Ω (𝑋; 𝑥0 , 𝑥1 ) := {𝜔 ∈ 𝐶 ([0, 1], 𝑋) | 𝜔(0) = 𝑥0 , 𝜔(1) = 𝑥1 } and

Ω (𝑋; 𝑥0 ) := Ω (𝑋; 𝑥0 , 𝑥0 ) .

Elements of Ω(𝑋; 𝑥0 , 𝑥1 ) are called paths and elements of Ω(𝑋; 𝑥0 ) loops.

For 𝜔 ∈ Ω(𝑋; 𝑥0 , 𝑥1 ) and 𝜂 ∈ Ω(𝑋; 𝑥1 , 𝑥2 ) define 𝜔 ★ 𝜂 ∈ Ω(𝑋; 𝑥0 , 𝑥2 ) by
(
𝜔(2𝑡), 0 ≤ 𝑡 ≤ 1/2,
(𝜔 ★ 𝜂) (𝑡) =
𝜂(2𝑡 − 1), 1/2 ≤ 𝑡 ≤ 1.

Moreover, we consider 𝜔 −1 ∈ Ω(𝑋; 𝑥1 , 𝑥0 ) with 𝜔 −1 (𝑡) = 𝜔(1 − 𝑡) and 𝜀 𝑥0 ∈ Ω(𝑋; 𝑥0 ) where

𝜀 𝑥0 (𝑡) = 𝑥0 .

𝑥1
𝑋

𝜂 𝑥0 𝑥2
𝜔

0 1 0 1
𝑡 → 2𝑡 𝑡 → 2𝑡 − 1

0 1
Figure 9. Concatenation of paths

20
 2.2. The fundamental group

Definition 2.17. For 𝜔 ∈ Ω(𝑋; 𝑥0 ) denote by [𝜔] the homotopy class of 𝜔 relative to {0, 1}.
Then
𝜋1 (𝑋; 𝑥0 ) := {[𝜔] | 𝜔 ∈ Ω(𝑋; 𝑥0 )}
is called the fundamental group of (𝑋, 𝑥0 ).

Lemma 2.18. For 𝜔, 𝜔′ , 𝜂, 𝜂 ′ , 𝜁 ∈ Ω(𝑋, 𝑥0 ) we have

(i) If 𝜔 ≃ {0,1} 𝜔′ and 𝜂 ≃ {0,1} 𝜂 ′ , then 𝜔 ★ 𝜂 ≃ {0,1} 𝜔′ ★ 𝜂 ′ ;

(ii) 𝜀 𝑥0 ★ 𝜔 ≃ {0,1} 𝜔 ≃ {0,1} 𝜔 ★ 𝜀 𝑥0 ;

(iii) 𝜔 ★ 𝜔 −1 ≃ {0,1} 𝜀 𝑥0 ≃ {0,1} 𝜔 −1 ★ 𝜔;

(iv) (𝜔 ★ 𝜂) ★ 𝜁 ≃ {0,1} 𝜔 ★ (𝜂 ★ 𝜁).

Proof. The proof will be given graphically. In the following diagrams we draw the domain of
the required homotopies. The horizontal axis denotes the loop parameter whereas the vertical
axis represents the deformation parameter. The interpolations are piecewise linear. The red area
gets mapped to 𝑥0 .

(i) We run the loop parameter with twice the speed.

𝜔′ 𝜂′

𝜔 𝜂
Figure 10. Concatenations of homotopic paths are homotopic.

In formulas, if we denote the homotopy between 𝜔 and 𝜔′ by 𝐹 and the one between 𝜂 and 𝜂 ′ by
𝐺, then the homotopy 𝐻 : [0, 1] × [0, 1] → 𝑋 between 𝜔 ★ 𝜂 and 𝜔′ ★ 𝜂 ′ is given by
(
𝐹 (2𝑡, 𝑠), 0 ≤ 𝑡 ≤ 12 ,
𝐻 (𝑡, 𝑠) =
𝐺 (2𝑡 − 1, 𝑠), 21 ≤ 𝑡 ≤ 1.
(ii) The first diagram in Figure ?? shows 𝜀 𝑥0 ★𝜔 ≃ {0,1} 𝜔, the second proves that 𝜔 ≃ {0,1} 𝜔★𝜀 𝑥0 .

21
2. Homotopy Theory

𝜔 𝜔

𝜀 𝑥0 𝜔 𝜔 𝜀 𝑥0
Figure 11. Concatenation with constant path is homotopic to original path

In formulas, the homotopy 𝐻 : [0, 1] × [0, 1] → 𝑋 between 𝜀 𝑥0 ★ 𝜔 and 𝜔 is given by

(
𝑥0 , 0 ≤ 𝑡 ≤ 1−𝑠
2 ,
𝐻 (𝑡, 𝑠) = 2𝑡 1−𝑠 1−𝑠
𝜔( 1+𝑠 − 1+𝑠 ), 2 ≤ 𝑡 ≤ 1,

and similarly for 𝜔 ★ 𝜀 𝑥0 .

(iii) The statement 𝜔 ★ 𝜔 −1 ≃ {0,1} 𝜀 𝑥0 is proven by the following diagram
𝜀 𝑥0

𝜔(1 − 𝑠)
𝑠

𝜔 𝜔 −1
Figure 12. Inverse modulo homotopy

For any 𝑠 ∈ [0, 1] the corresponding ”blue“ line segment gets mapped to 𝜔(1 − 𝑠). In formulas,
the homotopy 𝐻 : [0, 1] × [0, 1] → 𝑋 between 𝜔 ★ 𝜔 −1 and 𝜀 𝑥0 is given by





𝜔(2𝑡), 0 ≤ 𝑡 ≤ 1−𝑠
2 ,
1−𝑠 1+𝑠

𝐻 (𝑡, 𝑠) = 𝜔(1 − 𝑠), 2 ≤ 𝑡 ≤ 2 ,

 𝜔(2 − 2𝑡),

1+𝑠
2 ≤ 𝑡 ≤ 1.

To prove the statement 𝜀 𝑥0 ≃ {0,1} 𝜔 −1 ★ 𝜔 interchange the roles of 𝜔 and 𝜔 −1 .

(iv) The last assertion follows from the homotopy indicated in Figure 13.


22
 2.2. The fundamental group

𝜔 𝜂 𝜁

𝜔 𝜂 𝜁
Figure 13. Associativity

Lemma 2.18 has the following implications:

(i) ⇒ [𝜔] · [𝜂] := [𝜔 ★ 𝜂] is well defined;

(ii) ⇒ [𝜀 𝑥0 ] · [𝜔] = [𝜔] = [𝜔] · [𝜀 𝑥0 ];

(iii) ⇒ [𝜔] · [𝜔 −1 ] = [𝜀 𝑥0 ] = [𝜔 −1 ] · [𝜔];

(iv) ⇒ ([𝜔] · [𝜂]) · [𝜁] = [𝜔] · ([𝜂] · [𝜁]).

Hence 𝜋1 (𝑋; 𝑥0 ) together with “ · ” is a group with neutral element 1 := [𝜀 𝑥0 ] and inverse element
[𝜔] −1 = [𝜔 −1 ].
To any topological space with a preferred point we have associated a group, the fundamental group
of the space with that point. Now we consider continuous maps. Let 𝑓 ∈ 𝐶 (𝑋, 𝑌 ) and 𝑥0 ∈ 𝑋.
We put 𝑓 (𝑥0 ) =: 𝑦 0 ∈ 𝑌 . If 𝜔 ≃ {0,1} 𝜔′ and 𝐻 is a homotopy between them relative to {0, 1}
then 𝑓 ◦ 𝐻 is a homotopy between 𝑓 ◦ 𝜔 and 𝑓 ◦ 𝜔′ relative to {0, 1}. Hence 𝑓 ◦ 𝜔 ≃ {0,1} 𝑓 ◦ 𝜔′ .
Therefore we have a well-defined map 𝑓# : 𝜋1 (𝑋; 𝑥0 ) → 𝜋1 (𝑌 ; 𝑦 0 ), 𝑓# ([𝜔]) = [ 𝑓 ◦ 𝜔].

Lemma 2.19. (i) The map 𝑓# : 𝜋1 (𝑋; 𝑥0 ) → 𝜋1 (𝑌 ; 𝑦 0 ) is a group homomorphism.

(ii) It has the functorial properties

a) ( 𝑓 ◦ 𝑔)# = 𝑓# ◦ 𝑔# ;
b) (id𝑋 )# = id 𝜋1 (𝑋;𝑥0 ) .

(iii) If 𝑓 ≃ { 𝑥0 } 𝑓 ′ then 𝑓# = 𝑓#′ .

Proof. (i) From the definitions we have 𝑓 ◦ (𝜔 ★ 𝜂) = ( 𝑓 ◦ 𝜔) ★ ( 𝑓 ◦ 𝜂) and hence

𝑓# ([𝜔] · [𝜂]) = 𝑓# ([𝜔 ★ 𝜂]) = [ 𝑓 ◦ (𝜔 ★ 𝜂)] = [( 𝑓 ◦ 𝜔) ★ ( 𝑓 ◦ 𝜂)] = 𝑓# ([𝜔]) · 𝑓# ([𝜂]).

23
2. Homotopy Theory

(ii) Assertion b) being obvious we compute a):

( 𝑓 ◦ 𝑔)# ([𝜔]) = [( 𝑓 ◦ 𝑔) ◦ 𝜔] = [ 𝑓 ◦ (𝑔 ◦ 𝜔)] = 𝑓# ([𝑔 ◦ 𝜔]) = 𝑓# (𝑔# ([𝜔])) = ( 𝑓# ◦ 𝑔# ) ([𝜔]) .

(iii) By Lemma 2.6, 𝑓 ≃ {0,1} 𝑓 ′ implies 𝑓 ◦ 𝜔 ≃ {0,1} 𝑓 ′ ◦ 𝜔. We conclude

𝑓# ([𝜔]) = [ 𝑓 ◦ 𝜔] = [ 𝑓 ′ ◦ 𝜔] = 𝑓#′ ([𝜔])

which proves the statement. 

Corollary 2.20. If 𝑓 : 𝑋 → 𝑌 is a homeomorphism with 𝑓 (𝑥0 ) = 𝑦 0 then

𝑓# : 𝜋1 (𝑋, 𝑥0 ) → 𝜋1 (𝑌 , 𝑦 0 )

is a group isomorphism.

Proof. We only use the functorial properties:

( 𝑓 −1 )# ◦ ( 𝑓# ) = ( 𝑓 −1 ◦ 𝑓 )# = (id𝑋 )# = id 𝜋1 (𝑋;𝑥0 )

and similarly one sees that ( 𝑓# ) ◦ ( 𝑓 −1 )# = id 𝜋1 (𝑌 ;𝑦0 ) . Thus 𝑓♯ is an isomorphism with ( 𝑓# ) −1 =

( 𝑓 −1 )# . 

Now we want to deal with the question to what extent 𝜋1 (𝑋; 𝑥0 ) depends on the choice of 𝑥0 .

𝑋
To study this question let 𝑥0 , 𝑥1 ∈ 𝑋 and assume 𝜔
there exists a path 𝛾 ∈ Ω(𝑋; 𝑥0 , 𝑥1 ). If such a
𝑥1
path does not exist 𝜋1 (𝑋; 𝑥0 ) and 𝜋1 (𝑋; 𝑥1 ) are not
related. 𝛾
𝑥0
Figure 14. Dependence on base point
Look at the map Φ𝛾 : Ω(𝑋; 𝑥1 ) → Ω(𝑋; 𝑥0 ) where 𝜔 ↦→ (𝛾 ★ 𝜔) ★ 𝛾 −1 . Applying
Lemma 2.18 twice we know that if 𝜔 ≃ {0,1} 𝜔′ then 𝛾 ★ 𝜔 ≃ {0,1} 𝛾 ★ 𝜔′ and hence
(𝛾 ★ 𝜔) ★ 𝛾 −1 ≃ {0,1} (𝛾 ★ 𝜔′ ) ★ 𝛾 −1 . Thus the map

Φ̂𝛾 : 𝜋1 (𝑋; 𝑥1 ) → 𝜋1 (𝑋; 𝑥0 ),

[𝜔] ↦→ [Φ𝛾 (𝜔)] = [(𝛾 ★ 𝜔) ★ 𝛾 −1 ],

is well defined.

24
 2.2. The fundamental group

Proposition 2.21. Let 𝑋 be a topological space, assume 𝑥0 , 𝑥1 ∈ 𝑋 and 𝛾, 𝛾 ′ ∈ Ω(𝑋; 𝑥0 , 𝑥1 ).

Then

1.) The map Φ̂𝛾 : 𝜋1 (𝑋; 𝑥1 ) → 𝜋1 (𝑋; 𝑥0 ) is a group isomorphism.

2.) If 𝛾 ≃ {0,1} 𝛾 ′ then Φ̂𝛾 = Φ̂𝛾′ .

3.) For 𝛽 ∈ Ω(𝑋; 𝑥1 , 𝑥2 ) we have Φ̂𝛾★𝛽 = Φ̂𝛾 ◦ Φ̂𝛽 .

4.) For the constant path we have Φ̂ 𝜀𝑥0 = id 𝜋1 (𝑋;𝑥0 ) .

5.) For any [𝜔] ∈ 𝜋1 (𝑋; 𝑥1 ) we have Φ̂𝛾′ ([𝜔]) = 𝜅 · Φ̂𝛾 ([𝜔]) · 𝜅 −1 where 𝜅 = [𝛾 ′ ★ 𝛾 −1 ] ∈
𝜋1 (𝑋; 𝑥0 ).

Proof. a) Assertions 2., 3., and 4. follow directly from Lemma 2.18 and the definitions.
b) The map Φ̂𝛾 is a group homomorphism because


 
Φ̂𝛾 [𝜔] · [𝜂] = Φ̂𝛾 ([𝜔 ★ 𝜂])
= (𝛾 ★ (𝜔 ★ 𝜂)) ★ 𝛾 −1


= 𝛾 ★ ((𝜔 ★ (𝛾 −1 ★ 𝛾)) ★ 𝜂) ★ 𝛾 −1



= (𝛾 ★ 𝜔) ★ 𝛾 −1 ★ (𝛾 ★ 𝜂) ★ 𝛾 −1
= Φ̂𝛾 ([𝜔]) · Φ̂𝛾 ([𝜂]).
The map Φ̂𝛾 is bijective, because
3. 2. 4.
Φ̂𝛾 ◦ Φ̂𝛾 −1 = Φ̂𝛾★𝛾 −1 = Φ̂ 𝜀𝑥0 = id 𝜋1 (𝑋;𝑥0 ) .

and similarly Φ̂𝛾 −1 ◦ Φ̂𝛾 = id 𝜋1 (𝑋;𝑥1 ) . This proves 1.

c) We compute
     
𝜅 · Φ̂𝛾 ([𝜔]) · 𝜅 −1 = 𝛾 ′ ★ 𝛾 −1 · 𝛾 ★ 𝜔 ★ 𝛾 −1 · 𝛾 ★ (𝛾 ′ ) −1
 
= 𝛾 ′ ★ 𝛾 −1 ★ 𝛾 ★ 𝜔 ★ 𝛾 −1 ★ 𝛾 ★ (𝛾 ′ ) −1
 
= 𝛾 ′ ★ 𝜔 ★ (𝛾 ′ ) −1
= Φ̂𝛾′ ([𝜔]). 

Proposition 2.22. Let 𝑋, 𝑌 be topological spaces and 𝑥0 ∈ 𝑋. Let 𝑓 , 𝑔 ∈ 𝐶 (𝑋, 𝑌 ) and let
𝐻 : 𝑋 × [0, 1] → 𝑌 be a homotopy from 𝑓 to 𝑔. Define 𝜂 ∈ Ω(𝑌 ; 𝑓 (𝑥0 ), 𝑔(𝑥0 )) by

𝜂(𝑠) := 𝐻 (𝑥0 , 𝑠).

25
2. Homotopy Theory

𝑌
𝑔◦𝜔

𝜂
𝑠 𝐹

𝑓 ◦𝜔
𝑡

Figure 15. Auxiliary homotopy

Then the following diagram commutes:

𝜋1 (𝑋; 𝑥0 )P
PPP
𝑔#
PPP𝑓#
PPP
 PP(
𝜋1 (𝑌 ; 𝑔(𝑥0 )) / 𝜋1 (𝑌 ; 𝑓 (𝑥 0 ))
Φ̂ 𝜂

Proof. Let [𝜔] ∈ 𝜋1 (𝑋; 𝑥0 ) and define 𝐹 : [0, 1] × [0, 1] → 𝑌 by 𝐹 (𝑡, 𝑠) := 𝐻 (𝜔(𝑡), 𝑠).

The deformation indicated in Figure 16 yields a homotopy re-

lative to {0, 1} from 𝑓 ◦ 𝜔 to 𝜂 ★ (𝑔 ◦ 𝜔) ★ 𝜂 −1 . We conclude
that

𝑓# [𝜔] = [ 𝑓 ◦ 𝜔] = [𝜂 ★ (𝑔 ◦ 𝜔) ★ 𝜂 −1 ] = Φ̂ 𝜂 (𝑔# ([𝜔])) .

| |
1 2
3 3
Figure 16. The deformation


Now we can improve Corollary 2.20 and show that homotopy equivalent spaces have isomorphic
fundamental groups.

Theorem 2.23. Let 𝑓 : 𝑋 → 𝑌 be a homotopy equivalence. Then

𝑓# : 𝜋1 (𝑋; 𝑥0 ) → 𝜋1 (𝑌 ; 𝑓 (𝑥0 ))

is an isomorphism for all 𝑥0 ∈ 𝑋.

26
 2.3. The fundamental group of the circle

Proof. Let 𝑔 : 𝑌 → 𝑋 be a homotopy inverse of 𝑓 , i.e., 𝑔 ◦ 𝑓 ≃ id𝑋 and 𝑓 ◦ 𝑔 ≃ id𝑌 . We know

by Proposition 2.22 that for a suitable 𝜂 ∈ Ω(𝑋; 𝑥0 , 𝑓 (𝑔(𝑥0 )))

𝑔# ◦ 𝑓# = (𝑔 ◦ 𝑓 )# = Φ̂ 𝜂 ◦ (id𝑋 )# = Φ̂ 𝜂 ◦ id 𝜋1 (𝑋;𝑥0 ) = Φ̂ 𝜂 .

Hence 𝑔# ◦ 𝑓# is an isomorphism. In particular, 𝑓# is injective. Similarly, we can show that 𝑓# ◦ 𝑔#

is an isomorphism, hence 𝑓# is surjective. Therefore 𝑓# is an isomorphism. 

Corollary 2.24. If 𝐴 ⊂ 𝑋 is a deformation retract then the inclusion 𝜄 : 𝐴 → 𝑋 induces an

isomorphism 𝜄# : 𝜋1 ( 𝐴, 𝑥0 ) → 𝜋1 (𝑋; 𝑥0 ) for any 𝑥0 ∈ 𝐴.

Proof. If 𝐴 ⊂ 𝑋 is a deformation retract then the map 𝜄 : 𝐴 → 𝑋 is a homotopy equivalence.

Theorem 2.23 yields the claim. 

Example 2.25. If 𝑋 is contractible then the one-point set 𝐴 = {𝑥0 } ⊂ 𝑋 is a deformation retract.
Hence 𝜋1 (𝑋; 𝑥0 ) ≃ 𝜋1 ( 𝐴; 𝑥0 ) = {[𝜀 𝑥0 ]} = {1}. Thus contactible spaces have trivial fundamental
group.

2.3. The fundamental group of the circle

Recall the map Exp : R → 𝑆1 ⊂ C where Exp(𝑡) = 𝑒2 𝜋𝑖𝑡 .

Lemma 2.26. Let 𝑡0 ∈ R and 𝑧0 = Exp(𝑡0 ) ∈ 𝑆1 . Then for all 𝑓 ∈ 𝐶 (𝑆1 , 𝑆1 ) with 𝑓 (1) = 𝑧0
there exists a unique 𝑓ˆ ∈ 𝐶 (R, R) with 𝑓ˆ(0) = 𝑡0 such that the following diagram commutes:

𝑓ˆ
R /R

Exp Exp
 
𝑆1 / 𝑆1
𝑓

Proof. a) First we show uniqueness of the map 𝑓ˆ.

Assume that besides 𝑓ˆ there is a second map 𝑓˜ ∈ 𝐶 (R, R) with the same proper-
ties as 𝑓ˆ. The equality of Exp(𝑥) = Exp(𝑥 ′ ) is equivalent to 𝑥 − 𝑥 ′ ∈ Z. Since
Exp( 𝑓ˆ(𝑡)) = 𝑓 (Exp(𝑡)) = Exp( 𝑓˜(𝑡)) we deduce that 𝑓ˆ(𝑡) − 𝑓˜(𝑡) ∈ Z for all 𝑡 ∈ R. Since both
𝑓˜, 𝑓ˆ are continuous it follows that 𝑓ˆ − 𝑓˜ is constant. Finally, we know that 𝑓ˆ(0) = 𝑡 0 = 𝑓˜(0),
hence 𝑓˜ = 𝑓ˆ.

27
2. Homotopy Theory

b) Now we show existence of 𝑓ˆ.

Since 𝑆1 is compact, 𝑓 is uniformly continuous. Since Exp is also uniformly continuous, the
composition 𝑓 ◦ Exp : R → 𝑆1 is uniformly continuous. Hence there exists 𝜀 > 0 such that
𝑓 (Exp(𝐼)) ⊂ 𝑆1 is contained in a semi-circle for every interval 𝐼 ⊂ R with length ≤ 𝜀.
For any closed semi-circle 𝐶 ⊂ 𝑆1 the pre-image Exp −1 (𝐶) ⊂ R is the disjoint union of compact
intervals of length 12 . More precisely,
Ø 1

−1
Exp (𝐶) = 𝑡 1 + 𝑘, 𝑡 1 + 𝑘 + .
𝑘 ∈Z
2

Moreover, the restriction of Exp to each of these intervals is a homeomorphism onto its image,
 
1 ≈
Exp [𝑡1 +𝑘,𝑡1 +𝑘+ 1 ] : 𝑡 1 + 𝑘, 𝑡 1 + 𝑘 + −→ 𝐶.
2 2
Its inverse can written down explicitly in terms of logarithms but we will not need this.
Now we construct 𝑓ˆ step by step.

Since 𝑓 (Exp([0, 𝜀])) is contained in a closed semi-circle 𝐶0 we can define 𝑓ˆ on [0, 𝜀] by

𝑓ˆ := (Exp | 𝐼0 ) −1 ◦ 𝑓 ◦ Exp,

1
where 𝐼0 ⊂ R is the compact interval of length 2 with Exp(𝐼0 ) = 𝐶0 and 𝑡0 ∈ 𝐼0 . This insures
that

𝑓ˆ(0) = (Exp | 𝐼0 ) −1 ◦ 𝑓 ◦ Exp(0) = (Exp | 𝐼0 ) −1 ◦ 𝑓 (1) = (Exp | 𝐼0 ) −1 (𝑧0 ) = 𝑡 0 .

Put 𝑡 1 := 𝑓ˆ(𝜀).
Next, 𝑓 (Exp([𝜀, 2𝜀])) is contained in a closed semi-circle 𝐶1 and we define 𝑓ˆ on [𝜀, 2𝜀] by

𝑓ˆ := (Exp | 𝐼1 ) −1 ◦ 𝑓 ◦ Exp

where 𝐼1 ⊂ R is the compact interval of length 12 with Exp(𝐼1 ) = 𝐶1 and 𝑡1 ∈ 𝐼1 . This insures
that the two definitions of 𝑓ˆ at 𝜀 agree so that we obtain a continuous function 𝑓ˆ : [0, 2𝜀] → R.
Repeating this procedure infinitely many times we can extend 𝑓ˆ continuously to [0, ∞) → R.
The extension to the left is done similarly so that we obtain a continuous function 𝑓ˆ : R → R.
Commutativity of the diagram holds by construction. 

Definition 2.27. For a map 𝑓 ∈ 𝐶 (𝑆1 , 𝑆1 ) a map 𝑓ˆ ∈ 𝐶 (R, R) for which the diagram in
Lemma 2.26 commutes is called a lift of 𝑓 .

28
 2.3. The fundamental group of the circle

Example 2.28. Consider the map 𝑓𝑛 : 𝑆1 → 𝑆1 with 𝑓𝑛 = 𝑧 𝑛 , 𝑛 ∈ Z. Then we have

𝑓𝑛 Exp(𝑡) = Exp(𝑡) 𝑛 = Exp(𝑛𝑡) .

Hence the map 𝑓ˆ𝑛 given by 𝑓ˆ𝑛 (𝑡) = 𝑛𝑡 is a lift of 𝑓𝑛 .

Definition 2.29. For 𝑓 ∈ 𝐶 (𝑆1 , 𝑆1 ) we call

deg( 𝑓 ) := 𝑓ˆ(1) − 𝑓ˆ(0)

the degree of 𝑓 , where 𝑓ˆ is a lift of the map 𝑓 .

Remark 2.30
1.) We have seen that different lifts of 𝑓 differ by an additive constant in Z. Hence the degree
deg( 𝑓 ) is well defined, independently of the choice of lift 𝑓ˆ.

2.) The degree deg( 𝑓 ) is an integer because

 

 
Exp 𝑓ˆ(1) = 𝑓 Exp(1) = 𝑓 (1) = 𝑓 Exp(0) = Exp 𝑓ˆ(0)

Hence we have deg( 𝑓 ) = 𝑓ˆ(1) − 𝑓ˆ(0) ∈ Z.

3.) The map 𝑡 ↦→ 𝑓ˆ(𝑡 + 1) − 𝑓ˆ(𝑡) is continuous and takes values in Z, by the same argument as
above. We conclude that 𝑓ˆ(𝑡 + 1) − 𝑓ˆ(𝑡) = deg( 𝑓 ) for all 𝑡 ∈ R.

4.) For 𝑘 ∈ Z we compute

𝑓ˆ(𝑡 0 + 𝑘) − 𝑓ˆ(𝑡0 ) = 𝑓ˆ(𝑡 0 + 𝑘) − 𝑓ˆ(𝑡 0 + (𝑘 − 1))

+ 𝑓ˆ(𝑡 0 + (𝑘 − 1)) − 𝑓ˆ(𝑡0 + (𝑘 − 2))
+··· +
+ 𝑓ˆ(𝑡 0 + 1) − 𝑓ˆ(𝑡 0 )
(3)
= 𝑘 deg( 𝑓 ) .

5.) For 𝑓 , 𝑔 ∈ 𝐶 (𝑆1 , 𝑆1 ) and lifts 𝑓ˆ, 𝑔ˆ we compute






Exp 𝑓ˆ + 𝑔ˆ (𝑡) = Exp 𝑓ˆ(𝑡) Exp ( 𝑔(𝑡)) ˆ = 𝑓 Exp(𝑡) 𝑔 Exp(𝑡) .

Hence 𝑓ˆ + 𝑔ˆ is a lift of 𝑓 𝑔 and we get the following formula for the degree


deg( 𝑓 𝑔) = 𝑓ˆ(1) + 𝑔(1)
ˆ − 𝑓ˆ(0) + 𝑔(0)
ˆ = deg( 𝑓 ) + deg(𝑔) .

6.) For 𝑓 , 𝑔 ∈ 𝐶 (𝑆1 , 𝑆1 ) and lifts 𝑓ˆ, 𝑔ˆ we compute


 
 

Exp 𝑓ˆ 𝑔(𝑡)
ˆ = 𝑓 Exp 𝑔(𝑡)
ˆ = 𝑓 𝑔 Exp(𝑡)

29
2. Homotopy Theory

and therefore 𝑓ˆ ◦ 𝑔ˆ is a lift of 𝑓 ◦ 𝑔. For the degree of 𝑓 ◦ 𝑔 this means

(4)
deg( 𝑓 ◦ 𝑔) = 𝑓ˆ 𝑔(1)
ˆ − 𝑓ˆ 𝑔(0)
ˆ = 𝑓ˆ 𝑔(0)
ˆ + deg(𝑔) − 𝑓ˆ 𝑔(0)
ˆ = deg(𝑔) deg( 𝑓 ) ,

hence deg( 𝑓 ◦ 𝑔) = deg(𝑔) deg( 𝑓 ).

7.) Let 𝑓 ∈ 𝐶 (𝑆1 , 𝑆1 ) with deg( 𝑓 ) ≠ 0. We show that 𝑓 must then be surjective.
Namely, let 𝑓ˆ be a lift of 𝑓 . Then deg( 𝑓 ) = 𝑓ˆ(1) − 𝑓ˆ(0) is an integer, not equal to 0. Then, if
𝑓ˆ(1) > 𝑓ˆ(0), the whole interval 𝐼 := [ 𝑓ˆ(0), 𝑓ˆ(1)] must be contained in the image of 𝑓ˆ by the
intermediate value theorem. If 𝑓ˆ(1) < 𝑓ˆ(0) this holds for 𝐼 := [ 𝑓ˆ(1), 𝑓ˆ(0)]. In either case 𝐼 is
an interval of length at least 1, hence Exp(𝐼) = 𝑆1 . We conclude

𝑆1 = Exp(𝐼) ⊂ Exp(im( 𝑓ˆ)) = im( 𝑓 ).

Thus 𝑓 is onto.

Example 2.31. For the map 𝑓𝑛 : 𝑆1 → 𝑆1 with 𝑓𝑛 (𝑧) = 𝑧 𝑛 a lift is given by 𝑓ˆ(𝑡) = 𝑛𝑡 so that its
degree is deg( 𝑓𝑛 ) = 𝑓ˆ𝑛 (1) − 𝑓ˆ𝑛 (0) = 𝑛.

Lemma 2.32. Let 𝑓 , 𝑔 ∈ 𝐶 (𝑆1 , 𝑆1 ). If 𝑓 ≃ 𝑔, then we have deg( 𝑓 ) = deg(𝑔).

Proof. Let 𝐹 : 𝑆1 × [0, 1] → 𝑆1 be a homotopy from 𝑓 to 𝑔. Since 𝑆1 × [0, 1] is compact the

map 𝐹 is uniformly continuous. Hence, there exists a 𝛿 > 0 such that

|𝐹 (𝑧, 𝑠) − 𝐹 (𝑧, 𝑠′ )| < 1

whenever 𝑧 ∈ 𝑆1 and 𝑠, 𝑠′ ∈ [0, 1] with |𝑠 − 𝑠′ | < 𝛿. For such 𝑠, 𝑠′ the map

𝐹 (𝑧, 𝑠)
𝑧 ↦→ : 𝑆1 → 𝑆1
𝐹 (𝑧, 𝑠′ )


is continuous  and not surjective because −1 is not contained in the image. Hence, by Remark 2.30,
𝐹 (·,𝑠)
deg 𝐹 (·,𝑠′ ) = 0. We now compute using 2.30.5.):
   
𝐹 (·, 𝑠) ′ 𝐹 (·, 𝑠)
deg(𝐹 (·, 𝑠)) = deg · 𝐹 (·, 𝑠 ) = deg + deg 𝐹 (·, 𝑠′ ) = deg 𝐹 (·, 𝑠′ ) .
𝐹 (·, 𝑠′ ) 𝐹 (·, 𝑠′ )
We see inductively that

deg( 𝑓 ) = deg 𝐹 (·, 0) = deg 𝐹 (·, 𝑠1 ) = · · · = deg 𝐹 (·, 1) = deg(𝑔)

where 0 = 𝑠0 < 𝑠1 < · · · < 𝑠𝑟 = 1 is a partition of the unit interval [0, 1] satisfying
|𝑠𝑖+1 − 𝑠𝑖 | < 𝛿. 

30
 2.3. The fundamental group of the circle

Corollary 2.33. Let 𝑓 ∈ 𝐶 (𝑆1, 𝑆1 ) be such that 𝑓 = 𝑔 𝑆 1 where 𝑔 ∈ 𝐶 (𝐷 2 , 𝑆1 ) then

deg( 𝑓 ) = 0.

Proof. The map 𝐹 ∈ 𝐶 (𝑆1 × [0, 1], 𝑆1 ), 𝐹 (𝑧, 𝑠) := 𝑔(𝑠𝑧), defines a homotopy from a constant
map to 𝑓 . By Lemma 2.32 we conclude that deg( 𝑓 ) = deg(const) = 0. 

Let us give several applications of the concept of the degree.

Theorem 2.34 (Fundamental theorem of algebra). Every non-constant complex polyno-

mial has a root.

Proof. Suppose we are given a non-constant polynomial

𝑝(𝑧) = 𝑎 𝑛 𝑧 𝑛 + 𝑎 𝑛−1 𝑧 𝑛−1 + · · · + 𝑎1 𝑧 + 𝑎0 , 𝑎 𝑗 ∈ C, 𝑎 𝑛 ≠ 0, 𝑛 ≥ 1.

Since dividing by 𝑎 𝑛 does not change the roots, we assume without loss of generality that 𝑎 𝑛 = 1.
Now assume that 𝑝 has no roots. Then the map 𝑓 : C → 𝑆1 with 𝑓 (𝑧) = | 𝑝𝑝 (𝑧)
(𝑧) | is a well-defined
continuous map. To compute deg( 𝑓 𝑆 1 ) consider

𝑠𝑛 𝑝(𝑧/𝑠) 𝑧 𝑛 + 𝑠𝑎 𝑛−1 𝑧 𝑛−1 + · · · + 𝑠𝑛 𝑎0

𝐹 (𝑧, 𝑠) := 𝑛
= 𝑛 .
|𝑠 𝑝(𝑧/𝑠)| |𝑧 + 𝑠𝑎 𝑛−1 𝑧 𝑛−1 + · · · + 𝑠𝑛 𝑎0 |

The map 𝐹 ∈ 𝐶 (𝑆1 × [0, 1], 𝑆1 ) is a homotopy from 𝑓𝑛 (𝑧) = 𝑧 𝑛 to 𝑓 𝑆 1 . Computing its degree
we find deg( 𝑓 𝑆 1 ) = deg( 𝑓𝑛 ) = 𝑛 ≥ 1.
On the other hand, 𝑓 is a continuous map defined on all of C, hence Corollary 2.33 implies
deg( 𝑓 𝑆 1 ) = 0. This is a contradiction, thus 𝑝 must have a root. 

Lemma 2.35. Suppose the map 𝑓 ∈ 𝐶 (𝑆1 , 𝑆1 ) satisfies 𝑓 (−𝑧) = − 𝑓 (𝑧) for all 𝑧 ∈ 𝑆1 . Then
the degree deg( 𝑓 ) is odd.

Proof. Lef 𝑓ˆ be a lift of the map 𝑓 . We compute

         
𝑓 − Exp(𝑡) = 𝑓 Exp 12 Exp(𝑡) = 𝑓 Exp 𝑡 + 21 = Exp 𝑓ˆ 𝑡 + 12 .

Moreover,

       
− 𝑓 Exp(𝑡) = − Exp 𝑓ˆ(𝑡) = Exp 21 Exp 𝑓ˆ(𝑡) = Exp 𝑓ˆ(𝑡) + 21 .

31
2. Homotopy Theory

    
From 𝑓 (− Exp(𝑡)) = − 𝑓 (Exp(𝑡)) we conclude Exp 𝑓ˆ 𝑡 + 12 = Exp 𝑓ˆ(𝑡) + 12 and hence

1
1

𝑓ˆ 𝑡 + 2 − 𝑓ˆ(𝑡) + 2 =: 𝑘 (𝑡)

is an integer for every 𝑡. Due to the continuity of the map, 𝑘 (𝑡) it is constant, 𝑘 (𝑡) = 𝑘. Hence
𝑓ˆ(𝑡 + 1 ) − 𝑓ˆ(𝑡) = 𝑘 + 1 for all 𝑡 ∈ R. Now we can compute
2 2

      
deg( 𝑓 ) = 𝑓ˆ(1) − 𝑓ˆ(0) = 𝑓ˆ(1) − 𝑓ˆ 12 + 𝑓ˆ 21 − 𝑓ˆ(0)
= (𝑘 + 21 ) + (𝑘 + 12 ) = 2𝑘 + 1

which proves the assertion. 

Theorem 2.36 (Borsuk-Ulam). Let 𝑓 ∈ 𝐶 (𝑆2 , R2 ) satisfy 𝑓 (−𝑥) = − 𝑓 (𝑥) for all 𝑥 ∈ 𝑆2 .
Then 𝑓 has a zero.

Proof. Assume that the map 𝑓 has no zero. Then the map 𝑔 : 𝑆2 → 𝑆1 with 𝑔(𝑥) := k 𝑓𝑓 (( 𝑥) 𝑥) k
p
is defined and continuous. Moreover, 𝑔 satisfies 𝑔(−𝑥) = −𝑔(𝑥) for all 𝑥 ∈ 𝑆 2 . Now define a

map 𝐺 : 𝐷 2 → 𝑆1 by 𝐺 (𝑦) = 𝑔(𝑦, 1 − k𝑦k 2). The map 𝐺 ∈ 𝐶 (𝐷 2 , 𝑆1 ) has the property that
𝐺 𝑆 1 = 𝑔| 𝑆 1 . By Corollary 2.33 we know that deg(𝑔 𝑆 1 ) = 0. On the other hand we know by
Lemma 2.35 that deg(𝑔 𝑆 1 ) is odd, which gives a contradiction. 

Corollary 2.37. Let 𝑓 ∈ 𝐶 (𝑆2, R2 ). Then there exists a point 𝑥0 ∈ 𝑆2 with 𝑓 (−𝑥0 ) = 𝑓 (𝑥0 ).

Proof. Put 𝑔(𝑥) := 𝑓 (𝑥) − 𝑓 (−𝑥). Then the map 𝑔 ∈ 𝐶 (𝑆2 , R2 ) satisfies 𝑔(−𝑥) = −𝑔(𝑥)
for all 𝑥 ∈ 𝑆2 . Hence by the Borsuk-Ulam Theorem 2.36 there exists an 𝑥0 ∈ 𝑆2 with
0 = 𝑔(𝑥0 ) = 𝑓 (𝑥0 ) − 𝑓 (−𝑥0 ), which proves the theorem. 

Remark 2.38. In particular, the map 𝑓 ∈ 𝐶 (𝑆2 , R2 ) cannot be injective. Thus the sphere 𝑆2
cannot be homeomorphic to a subset of R2 . This also shows that R3 cannot be homeomorphic to
R2 .

32
 2.3. The fundamental group of the circle

Now suppose you have a sandwich consisting of bread,

ham, and cheese. You want to share it evenly with your
friend. Can you cut the sandwich into two pieces such
that each piece contains the same amount of bread, the
same amount of ham and the same amount of cheese?
The following theorem tells us that it is possible.

Figure 17. Cutting a sandwich1

Theorem 2.39 (Ham-Sandwich-Theorem). Let 𝐴, 𝐵, 𝐶 ⊂ R3 be open and bounded. Then

there exists an affine hyperplane 𝐻 ⊂ R3 such that each of the sets is divided into pieces of
equal volume.

Proof. The proof consists of four steps.

𝐻 𝑥,𝑡
a) For each 𝑥 ∈ 𝑆2 and 𝑡 ∈ R we define the
affine hyperplane
𝑡𝑥 • 𝐻 +𝑥,𝑡
3
𝐻 𝑥,𝑡 := {𝑦 ∈ R | h𝑦, 𝑥i = 𝑡} . 𝑆2
•𝑥
It is clear that 𝐻 − 𝑥,−𝑡 = 𝐻 𝑥,𝑡 . We define the 0•
half space 𝐻 +𝑥,𝑡 by

𝐻 +𝑥,𝑡 := {𝑦 ∈ R3 | h𝑦, 𝑥i ≥ 𝑡}.

Figure 18. The hyperplane function

𝐻 𝑥,𝑡 ′
𝐻 𝑥,𝑡

b) Now fix 𝑥 ∈ 𝑆2 . Look at the func-

tion 𝑎 𝑥 : R → R defined by
𝐵(0, 𝑅 𝐴)
𝑎 𝑥 (𝑡) := vol( 𝐴 ∩ 𝐻 +𝑥,𝑡 ).
𝐴
It satisfies 𝑎 − 𝑥 (𝑡) + 𝑎 𝑥 (−𝑡) = vol( 𝐴)
and is monotonically decreasing.
Since 𝐴 is bounded there exists
|𝑡 − 𝑡 ′ |
𝑅 𝐴 > 0 such that 𝐴 ⊂ 𝐵(0, 𝑅 𝐴).

Figure 19. The volume function

1 Based on public domain images from http://www.sxc.hu

33
2. Homotopy Theory

For 𝑡 < 𝑡 ′ ∈ R we have

|𝑎 𝑥 (𝑡 ′ ) − 𝑎 𝑥 (𝑡)| = vol( 𝐴 ∩ (𝐻 +𝑥,𝑡 \ 𝐻 +𝑥,𝑡 ′ )) ≤ 𝜋𝑅 2𝐴 · |𝑡 − 𝑡 ′ | .

Thus 𝑎 𝑥 is Lipschitz continuous.

vol( 𝐴)
𝛼𝑥

Moreover, 𝑎 𝑥 (𝑡) = vol( 𝐴) for 𝑡 ≪ 0

and 𝑎 𝑥 (𝑡) = 0 for 𝑡 ≫ 0.

𝑡 𝑥− 𝑡 +𝑥
Figure 20. Monotonicity of the volume function

c) By continuity and monotonicity the pre-image of any value under 𝛼 𝑥 is a closed interval. In
particular, 𝑎 −1 − +
𝑥 (vol( 𝐴)/2) = [𝑡 𝑥 , 𝑡 𝑥 ].
𝑡 − +𝑡 +
Now put 𝛼(𝑥) := 𝑥 2 𝑥 . Hence 𝐻 𝑥, 𝛼( 𝑥) divides 𝐴 into two pieces of equal volume. Moreover,
𝛼(−𝑥) = −𝛼(𝑥) for all 𝑥 ∈ 𝑆2 and it is not hard to check that 𝛼 is continuous. Similarly, define
the functions 𝛽 for 𝐵 and 𝛾 for 𝐶.

d) Consider the map 𝑓 ∈ 𝐶 (𝑆2 , R2 ) with 𝑓 (𝑥) = (𝛼(𝑥) − 𝛽(𝑥), 𝛼(𝑥) − 𝛾(𝑥)). We have

𝑓 (−𝑥) = (𝛼(−𝑥) − 𝛽(−𝑥), 𝛼(−𝑥) − 𝛾(−𝑥))

= (−𝛼(𝑥) + 𝛽(𝑥), −𝛼(𝑥) + 𝛾(𝑥))
= − 𝑓 (𝑥).

Thus the Borsuk-Ulam Theorem 2.36 applies and there exists a point 𝑥0 ∈ 𝑆2 with

(0, 0) = 𝑓 (𝑥0 ) = (𝛼(𝑥0 ) − 𝛽(𝑥0 ), 𝛼(𝑥0 ) − 𝛾(𝑥0 )).

Hence 𝛼(𝑥0 ) = 𝛽(𝑥0 ) = 𝛾(𝑥0 ). Thus the hyperplance 𝐻 𝑥0 , 𝛼( 𝑥0 ) does the job. 

Remark 2.40. The ham-sandwich theorem is optimal in the sense that it fails for more than three
sets in R3 .

Example 2.41. A ball 𝐵(𝑥, 𝑟) ⊂ R3 with center 𝑥 is cut into two halves of equal volume exactly
by those planes that contain 𝑥. If you choose four balls in R3 in such a way that their centers are
not contained in one plane, then no plane will cut them all into halves of equal volume.

In the following we want to use the concept of degree to determine 𝜋1 (𝑆1 ; 1).

34
 2.3. The fundamental group of the circle

𝑓𝜔

Exp 𝜔
!
a) For 𝜔 ∈ Ω(𝑆1 ; 1) and 𝐼 = [0, 1] consider the 𝑆1 a❈o 𝐼 / 𝑆1
❈❈ ④=
following diagram: ❈❈≈ ④④④
❈❈ ④④
Exp ❈  ④④ 𝜔
𝐼/𝜕𝐼
Here 𝜔 and Exp are the continuous maps induced on the quotient space. They exist because
𝜔(0) = 𝜔(1) and Exp(0) = Exp(1), compare Section 1.5. By Example 1.30 we know that
Exp : 𝐼/𝜕𝐼 → 𝑆1 is a homeomorphism. Now put


𝑓 𝜔 := 𝜔 ◦ (Exp) −1 ∈ 𝐶 𝑆1 , 𝑆1

and define deg(𝜔) := deg( 𝑓 𝜔 ). We have obtained a map

deg : Ω(𝑆1 ; 1) → Z.

b) Now suppose 𝜔 ≃ {0,1} 𝜔′ . We choose a homotopy 𝐹 : 𝐼 × 𝐼 → 𝑆1 from 𝜔 to 𝜔′ relativ to

{0, 1}. Then the map 𝐺 : 𝑆1 × 𝐼 → 𝑆1 defined by 𝐺 (𝑧, 𝑠) := 𝐹 ((Exp) −1 (𝑧), 𝑠) is a homotopy
from 𝑓 𝜔 to 𝑓 𝜔 ′ , hence 𝑓 𝜔 ≃ 𝑓 𝜔 ′ . Therefore

deg(𝜔) = deg( 𝑓 𝜔 ) = deg( 𝑓 𝜔 ′ ) = deg(𝜔′ ) .

Hence we get a well-defined map

deg : 𝜋1 (𝑆1 ; 1) → Z where [𝜔] ↦→ deg(𝜔).

c) This map is surjective because for 𝑛 ∈ Z we can consider the map 𝜔(𝑡) = Exp(𝑛𝑡). The
commutative diagram
𝑧↦→ 𝑧 𝑛 = 𝑓 𝑛 (𝑧)
𝑆1 f◆◆ / 𝑆1
O
◆◆◆
◆◆◆ 𝑡↦→ 𝜔 (𝑡 )=Exp (𝑛𝑡 )=Exp (𝑡 ) 𝑛
◆
𝑡↦→Exp(𝑡 ) ◆◆◆
◆
𝐼
shows 𝑓 𝜔 (𝑧) = 𝑧 𝑛 = 𝑓𝑛 (𝑧) and we get deg(𝜔) = deg( 𝑓𝑛 ) = 𝑛.

d) Now let 𝜔, 𝜔′ ∈ Ω(𝑆1 ; 1) and consider the map

𝑓 𝜔★𝜔 ′ (Exp(𝑡)) = 𝜔 ★ 𝜔′ (𝑡)

(
𝜔(2𝑡), 0 ≤ 𝑡 ≤ 1/2
=
𝜔′ (2𝑡 − 1), 1/2 ≤ 𝑡 ≤ 1
(
𝑓 𝜔 (Exp(2𝑡)), 0 ≤ 𝑡 ≤ 1/2
=
𝑓 𝜔 ′ (Exp(2𝑡 − 1)), 1/2 ≤ 𝑡 ≤ 1

35
2. Homotopy Theory

Let 𝑓ˆ𝜔 , 𝑓ˆ𝜔 ′ be lifts of 𝑓 𝜔 , 𝑓 𝜔 ′ with 𝑓ˆ𝜔 ′ (0) = 𝑓ˆ𝜔 (1). Then we have
(
Exp( 𝑓ˆ𝜔 (2𝑡)), 0 ≤ 𝑡 ≤ 1/2
𝑓 𝜔★𝜔 ′ (Exp(𝑡)) =
Exp( 𝑓ˆ𝜔 ′ (2𝑡 − 1)), 1/2 ≤ 𝑡 ≤ 1
( !
𝑓ˆ𝜔 (2𝑡), 0 ≤ 𝑡 ≤ 1/2
= Exp
𝑓ˆ𝜔 ′ (2𝑡 − 1), 1/2 ≤ 𝑡 ≤ 1
| {z }
=:𝑔(𝑡 )

Note that the map 𝑔(𝑡) is continuous because of 𝑓ˆ𝜔 ′ (0) = 𝑓ˆ𝜔 (1). Thus 𝑔(𝑡) is a lift of 𝑓 𝜔★𝜔 ′ .
Now we compute the degree of 𝜔 ★ 𝜔′ .

deg(𝜔 ★ 𝜔′ ) = 𝑔(1) − 𝑔(0)

= 𝑓ˆ𝜔 ′ (1) − 𝑓ˆ𝜔 (0)
= 𝑓ˆ𝜔 ′ (1) − 𝑓ˆ𝜔 ′ (0) + 𝑓ˆ𝜔 (1) − 𝑓ˆ𝜔 (0)
= deg(𝜔′ ) + deg(𝜔)

Hence the map deg : 𝜋1 (𝑆1 ; 1) → (Z, +) is a group homomorphism.

e) Finally we compute its kernel. Let 𝜔 ∈ Ω(𝑆1 ; 1) with deg(𝜔) = 0. Let 𝑓ˆ𝜔 be the lift of 𝑓 𝜔
with 𝑓ˆ𝜔 (0) = 0. Since 0 = deg(𝜔) = 𝑓ˆ𝜔 (1) − 𝑓ˆ𝜔 (0) we have 𝑓ˆ𝜔 (1) = 𝑓ˆ𝜔 (0) = 0. Next consider
the continuous map 𝐹 : 𝐼 × 𝐼 → 𝑆1 with 𝐹 (𝑡, 𝑠) := Exp(𝑠 𝑓ˆ𝜔 (𝑡)). It satisfies:

𝐹 (𝑡, 0) = 1 = 𝜀 1 (𝑡) ,
𝐹 (𝑡, 1) = 𝑓 𝜔 (Exp(𝑡)) = 𝜔(𝑡) ,
𝐹 (0, 𝑠) = Exp(𝑠 · 0) = 1 ,
𝐹 (1, 𝑠) = Exp(𝑠 · 𝑓ˆ𝜔 (1)) = Exp(𝑠 0) = 1 .

We conclude that 𝜔 ≃ {0,1} 𝜀 1 , hence [𝜔] = [𝜀 1 ] = 1 ∈ 𝜋1 (𝑆1 ; 1). Therefore the kernel is trivial
and the map deg : 𝜋1 (𝑆1 ; 1) → Z is injective.

We summarize the discussion in the following

Theorem 2.42. The map deg : 𝜋1 (𝑆1 ; 1) → Z is a group isomorphism.

Example 2.43. We already know that 𝑆1 is a strong deformation retract of C \ {0}. Hence the
inclusion 𝜄 : 𝑆1 → C \ {0} induces an isomorphism

𝜄# : 𝜋1 (𝑆1 ; 1) → 𝜋1 (C \ {0}; 1)

and therefore 𝜋1 (C \ {0}; 1)  Z.

36
 2.3. The fundamental group of the circle

𝑆1 = 𝜕𝐷 2

Example 2.44 𝐷2
We show that 𝑆1 = 𝜕𝐷 2 is not a retract of 𝐷 2 . Suppose the
map 𝑟 : 𝐷 2 → 𝑆1 is a retraction and denote the inclusion by
𝜄 : 𝑆1 → 𝐷 2 .
Figure 21. The disk and its
boundary
Then we would have 𝑟 ◦ 𝜄 = id𝑆 1 , hence 𝑟 # ◦ 𝜄# = (𝑟 ◦ 𝜄)# = (id𝑆 1 )# = id 𝜋1 (𝑆 1 ;1) . We then get a
contradiction because of the following diagram
𝜄# 𝑟#
Z  𝜋1 (𝑆1 ; 1) / 𝜋1 (𝐷 2 ; 1)  {1} / 𝜋1 (𝑆 1 ; 1)  Z
:

id 𝜋 1
1 (𝑆 ;1)

As a corollary we get a proof of Brouwer’s fixed point theorem in dimension two. See page 93
for the theorem in general dimensions.

Theorem 2.45 (Brouwer’s fixed point theorem in 2 dimensions). Let 𝑓 : 𝐷 2 → 𝐷 2 be a

continuous map. Then 𝑓 has a fixed point, i.e., there exists an 𝑥 ∈ 𝐷 2 such that 𝑓 (𝑥) = 𝑥.

Proof. Assume that 𝑓 ∈ 𝐶 (𝐷 2 , 𝐷 2 ) has no fixed point. Then 𝑓 (𝑥) ≠ 𝑥 for all 𝑥 ∈ 𝐷 2 so that we
can consider the half line emanating from 𝑓 (𝑥) through 𝑥. We let 𝑟 (𝑥) be its intersection point
with 𝜕𝐷 2 = 𝑆1 as indicated in the picture.

𝑥
• 𝑟 (𝑥)
𝑓 (𝑥)

Figure 22. Constructing a retraction

This yields a retraction 𝑟 : 𝐷 2 → 𝑆1 and we get a contradiction. 

Example 2.46. We show that the system of equations

1 + 21 sin(𝑥)𝑦 − 2𝑥 = 0,
1 𝑥2
2 cos(𝑥)𝑦 + 2 − 2𝑦 = 0, (2.1)

37
2. Homotopy Theory

has a solution. We rewrite this as a fixed point equation and apply the Brouwer fixed point
theorem. To do this we put
 
1 sin(𝑥)𝑦 cos(𝑥)𝑦 𝑥 2
𝑓 (𝑥, 𝑦) := + , +
2 4 4 4
Fixed points of 𝑓 (𝑥, 𝑦) are then the same as solutions to the above system of equations (2.1). To
apply Brouwer’s fixed point theorem we have to show that 𝑓 (𝐷 2 ) ⊂ 𝐷 2 . Let 𝑥, 𝑦 ∈ 𝐷 2 . Then
 2  2
1 sin(𝑥)𝑦 cos(𝑥)𝑦 𝑥 2
k 𝑓 (𝑥, 𝑦) k 2 = + + +
2 4 4 4
1 sin(𝑥)𝑦 sin(𝑥) 𝑦 2 2 cos(𝑥) 2 𝑦 2 cos(𝑥)𝑦𝑥 2 𝑥 4
= + + + + +
4 4 16 16 8 16
1 sin(𝑥)𝑦 𝑦 2 cos(𝑥)𝑦𝑥 2 𝑥 4
= + + + +
4 4 16 8 16
1 |𝑦| 𝑦 2 |𝑦|𝑥 2 𝑥 4
≤ + + + +
4 4 16 8 16
1 1 1 1 1
≤ + + + +
4 4 16 8 16
3
=
4
≤ 1.

Example 2.47. Consider 𝑓𝑛 ∈ 𝐶 (𝑆1 , 𝑆1 ) with 𝑓𝑛 (𝑧) = 𝑧 𝑛 . We check that the diagram
deg
𝜋1 (𝑆1 ; 1) 
/Z

( 𝑓 𝑛 )# 𝑛·
deg 
𝜋1 (𝑆1 ; 1)  /Z

commutes. Let [𝜔] ∈ 𝜋1 (𝑆1 ; 1). We compute

deg ( 𝑓𝑛 )# ([𝜔]) = deg [ 𝑓𝑛 ◦ 𝜔] = deg( 𝑓𝑛 ◦ 𝜔)


= deg( 𝑓𝑛 ◦ 𝑓 𝜔 ) = deg( 𝑓𝑛 ) deg( 𝑓 𝜔 ) = 𝑛 deg [𝜔] .

Remark 2.48. From Exercise 2.5 we know that for 𝑋1 , 𝑋2 and 𝑥1 ∈ 𝑋1 , 𝑥2 ∈ 𝑋2 we have

𝜋1 (𝑋1 × 𝑋2 ; (𝑥1 , 𝑥2 ))  𝜋1 (𝑋1 ; 𝑥1 ) × 𝜋1 (𝑋2 ; 𝑥2 ) .

For the two-dimensional torus 𝑇 2 = 𝑆1 × 𝑆1 we get 𝜋1 (𝑇 2 ; 𝑥)  Z × Z = Z2 . More generally, we

get inductively for the 𝑛-torus 𝑇 𝑛 = 𝑆1 × · · · × 𝑆1 that 𝜋1 (𝑇 𝑛 ; 𝑥)  Z𝑛 . In particular, 𝑇 𝑛 ; 𝑇 𝑚
| {z }
𝑛
for 𝑛 ≠ 𝑚.

38
 2.3. The fundamental group of the circle

Definition 2.49. Let 𝑋 be a topological space.

1.) We call 𝑋 connected iff 𝑋 and ∅ are the only subsets of 𝑋 which are both open and closed.

2.) We call 𝑋 path-connected iff for all 𝑥1 , 𝑥2 ∈ 𝑋 there exists a path 𝜔 ∈ Ω(𝑋; 𝑥1 , 𝑥2 ).

3.) Let 𝑋 be path-connected. Then 𝑋 is called simply-connected (or 1-connected) iff

𝜋1 (𝑋; 𝑥0 ) = {1} for some (and hence all) 𝑥0 ∈ 𝑋.

Example 2.50. The interval [0, 1] is connected. To see this let 𝐼 ⊂ [0, 1] be open and closed
and assume that 𝐼 is neither empty nor all of [0, 1]. Then there exists 𝑡 0 ∈ 𝐼 and 𝑡 1 ∈ [0, 1] \ 𝐼.
W.l.o.g. let 𝑡 0 < 𝑡1 , the other case being analogous. We put 𝑇 := sup(𝐼 ∩ [0, 𝑡1 )). Then
0 ≤ 𝑡 0 ≤ 𝑇 ≤ 𝑡 1 ≤ 1. Since 𝐼 is closed 𝑇 ∈ 𝐼.
If 𝑇 = 1 then 𝑇 = 𝑡 1 which contradicts 𝑡1 ∉ 𝐼. If 𝑇 < 1 then there exists 𝜀 > 0 such that
[𝑇 , 𝑇 + 𝜀) ⊂ 𝐼 because 𝐼 is open. This contradicts the maximality of 𝑇 .

Remark 2.51. If 𝑋 is 1-connected then it is path-connected by definition but the converse is not
true. For example, 𝑆1 is path-connected but not 1-connected.

Remark 2.52. If 𝑋 is path-connected then 𝑋 is connected.

Proof. Let 𝑋 be path-connected and let 𝑈 ⊂ 𝑋 be open and closed, 𝑈 ≠ ∅. We show 𝑈 = 𝑋.

Since 𝑈 is non-empty we can find 𝑥1 ∈ 𝑈. Let 𝑥2 ∈ 𝑋 be any point.
Since 𝑋 is path-connected there is a path 𝜔 ∈ Ω(𝑋; 𝑥1 , 𝑥2 ). Then 𝐼 := 𝜔 −1 (𝑈) is an open and
closed subset of [0, 1]. Since 𝜔(0) = 𝑥1 ∈ 𝑈 we have 0 ∈ 𝐼 and hence 𝐼 is non-empty. Thus
𝐼 = [0, 1] because [0, 1] is connected. We conclude 1 ∈ 𝐼 and hence 𝑥2 = 𝜔(1) ∈ 𝑈. This shows
that 𝑈 contains all points of 𝑋. 

Again, the converse implication does not hold in general. Consider for example the space

𝑋 := {(𝑡, sin(1/𝑡)) | 𝑡 > 0} ∪ {(0, 𝑠) | −1 ≤ 𝑠 ≤ 1},

see Figure 23. Then 𝑋 is connected but not path-connected.

Remark 2.53. Let 𝑋 be path-connected. Then the following are equivalent (see Exercise 2.2):

(i) 𝑋 is simply connected;

(ii) Every 𝜔 ∈ 𝐶 (𝑆1 , 𝑋) is homotopic to a constant map;

(iii) Every 𝜔 ∈ 𝐶 (𝑆1 , 𝑋) has a continuous extension to a map 𝐷 2 → 𝑋.

39
2. Homotopy Theory

Figure 23. Connected but not path-connected

2.4. The Seifert-van Kampen theorem

The Seifert-van Kampen theorem will allow us to compute the fundamental group of spaces
which are built out of simpler spaces whose fundamental groups we already know. We start with
an excursion to group theory.

Definition 2.54. Let 𝐺 be a group. A subgroup 𝐻 ⊂ 𝐺 is called normal iff

𝑔 · 𝐻 = 𝐻 · 𝑔 for all 𝑔 ∈ 𝐺 .

The condition on a subgroup of being normal can be reformulated in various ways. It is equivalent
to any of the following:

(i) 𝑔 · 𝐻 · 𝑔 −1 = 𝐻 for all 𝑔 ∈ 𝐺;

(ii) 𝑔 · 𝐻 · 𝑔 −1 ⊂ 𝐻 for all 𝑔 ∈ 𝐺;
(iii) 𝑔 · ℎ · 𝑔 −1 ∈ 𝐻 for all 𝑔 ∈ 𝐺 and ℎ ∈ 𝐻.

Example 2.55. Consider the cartesian product of two groups 𝐺 = 𝐺 1 × 𝐺 2 = {(𝑔1 , 𝑔2 ) | 𝑔 𝑗 ∈

𝐺 𝑗 } with componentwise multiplication. Then {(𝑔1 , 1) | 𝑔1 ∈ 𝐺 1 }  𝐺 1 is a normal subgroup
of 𝐺 because



(𝑔1 , 𝑔2 ) · ( 𝑔˜ 1 , 1) · (𝑔1 , 𝑔2 ) −1 = 𝑔1 𝑔˜1 𝑔1−1 , 𝑔2 1𝑔2−1 = 𝑔1 𝑔˜1 𝑔1−1 , 1 .

Example 2.56. Let 𝜑 : 𝐺 → 𝐾 be a group homomorphism. Then 𝐻 = ker(𝜑) is a normal

subgroup because for ℎ ∈ ker(𝜑) and 𝑔 ∈ 𝐺 we have

𝜑(𝑔 · ℎ · 𝑔 −1 ) = 𝜑(𝑔) · 𝜑(ℎ) ·𝜑(𝑔) −1 = 1 ,

|{z}
1

hence 𝑔 · ℎ · 𝑔 −1 ∈ ker(𝜑).

40
 2.4. The Seifert-van Kampen theorem

Remark 2.57. If 𝐻 ⊂ 𝐺 is a normal subgroup then 𝐺/𝐻 is again a group via

(𝑔 · 𝐻) · ( 𝑔˜ · 𝐻) = (𝑔 𝑔)
˜ · 𝐻.

Normality of 𝐻 ensures that this multiplication is well defined. The group 𝐻 is then the kernel of
𝐺 → 𝐺/𝐻 with 𝑔 ↦→ 𝑔 · 𝐻. Thus the normal subgroups are exactly those which arise as kernels
of group homomorphisms.

Now let 𝑆 ⊂ 𝐺 be any subset. Then

Ù
N (𝑆) := 𝐻
𝐻 ⊂𝐺 normal subgroup,
𝐻 ⊃𝑆

is the smallest normal subgroup containing 𝑆. We call N (𝑆) the normal subgroup generated by
𝑆.

Example 2.58. N (∅) = {1}.

Definition 2.59
Let 𝐺 1 and 𝐺 2 be groups. A group 𝐺 is called free product of 𝐺1 ❆
𝐺 1 and 𝐺 2 iff there exist homomorphisms 𝑖 𝑗 : 𝐺 𝑗 → 𝐺 such 𝑖1 ⑥⑥ ❆❆ 𝜑
⑥ ❆❆ 1
⑥⑥ ❆❆
that for all groups 𝐻 and for all homomorphisms 𝜑 𝑗 : 𝐺 𝑗 → 𝐻 ~ ⑥⑥⑥
𝜑1★𝜑2
❆
there exists a unique homomorphism 𝐺 `❆ /𝐻
❆❆ >
❆❆ ⑥⑥⑥
❆ ⑥
𝜑1 ★ 𝜑2 : 𝐺 → 𝐻 𝑖2 ❆❆ ⑥⑥ 𝜑
⑥⑥ 2
𝐺2
such that the diagram to the right commutes.

Remark 2.60. Here we have characterized free products by

their universal property. This universal property implies for 𝐺 1 ❇❇
𝑖1 ⑦⑦ ❇❇❇❇❇
example that the maps 𝑖 𝑗 : 𝐺 𝑗 → 𝐺 are injective. ⑦⑦ ❇❇❇id❇
⑦⑦ ❇❇❇❇
~⑦ ❇
Namely, choose 𝐻 = 𝐺 1 , 𝜑1 = id𝐺1 and 𝜑2 (𝑔2 ) = 1 for all ⑦ id ★𝜑2
/ 𝐺1
𝐺 `❅
𝑔2 ∈ 𝐺 2 . The diagram now tells us that the map 𝑖 1 must be ❅❅ ⑤ =
❅❅ ⑤⑤
injective because the identity is injective. Similarly, we see that ❅
𝑖2 ❅❅
⑤
⑤⑤ 𝜑
𝑖2 is injective. ⑤⑤ 2
𝐺2

Remark 2.61. The free product of 𝐺 1 and 𝐺 2 is unique up to isomorphism.

41
2. Homotopy Theory

𝐺1 ❇
Namely, let 𝐺 ′ be another free product of 𝐺 1 and 𝐺 2 with 𝑖1 ⑥⑥⑥ ❇❇ 𝑖′
❇❇1
𝑖 ′𝑗 : 𝐺 𝑗 → 𝐺 ′ the corresponding homomorphisms. By the ⑥⑥⑥ ❇❇
~⑥⑥ ′ ′
𝑖1 ★𝑖2 !
universal property of 𝐺 with 𝐻 = 𝐺 ′ and 𝜑 𝑗 = 𝑖 ′𝑗 we get the 𝐺 `❆ / 𝐺′
❆❆ ⑤=
following commutative diagram: ❆❆
❆ ⑤⑤⑤
𝑖2 ❆❆ ⑤⑤ ′
⑤⑤ 𝑖2
𝐺2

𝐺1 ❆❆
𝑖1′ ⑤⑤ ❆❆ 𝑖1
⑤ ❆❆
⑤ ⑤⑤ ❆❆
} ⑤ 𝑖1★𝑖2
Interchanging the roles of 𝐺 and 𝐺 ′ we get another commutative ′
𝐺 a❇ /𝐺
diagram: ❇❇ ⑥⑥>
❇❇ ⑥⑥
❇ ⑥⑥𝑖2
𝑖2′ ❇❇ ⑥⑥
𝐺2

𝐺1 ❆
𝑖1 ⑥⑥ ′ ❆❆❆ 𝑖1
⑥⑥⑥ 𝑖1 ❆❆
⑥ ❆
Combining both diagrams we obtain ~⑥⑥𝑖1′ ★𝑖2′  𝑖1★𝑖2 ❆
𝐺 `❆ / 𝐺O ′ /𝐺
❆❆ ⑥⑥>
❆❆ ′ ⑥
❆ 𝑖 ⑥⑥
𝑖2 ❆❆2 ⑥⑥⑥ 𝑖2
𝐺2

𝐺1 ❆
𝑖1 ⑥⑥ ❆❆ 𝑖
⑥⑥ ❆❆1
On the other hand, this diagram commute as well. ⑥⑥ ❆❆
❆
~⑥
⑥ id
𝐺 `❆ /𝐺
❆❆ ⑥⑥>
❆❆ ⑥⑥
❆ ⑥⑥
𝑖2 ❆❆ ⑥⑥ 𝑖2
𝐺1
By the uniqueness of the induced homomorphisms we have

(𝑖 1 ★ 𝑖 2 ) ◦ 𝑖 1′ ★ 𝑖 2′ = id𝐺

and similarly


𝑖 1′ ★ 𝑖 2′ ◦ (𝑖 1 ★ 𝑖2 ) = id𝐺 ′ .

Hence the map 𝑖 1 ★ 𝑖 2 : 𝐺 → 𝐺 ′ is a group isomorphism with inverse 𝑖1′ ★ 𝑖 2′ .

Next we show the existence of the free product of two groups by a direct construction. For this
purpose let 𝐺 1 and 𝐺 2 be groups. For formal reasons we assume without loss of generality that

42
 2.4. The Seifert-van Kampen theorem

𝐺 1 ∩ 𝐺 2 = ∅.2 We define


𝐺 1 ★ 𝐺 2 := (𝑥1 , . . . 𝑥 𝑛 ) | 𝑛 ∈ N0 , 𝑥 𝑗 ∈ 𝐺 1 \ {1𝐺1 )} ∪ (𝐺 2 \ {1𝐺2 } such that
if 𝑥𝑖 ∈ 𝐺 1 then 𝑥𝑖+1 ∈ 𝐺 2 or conversely .

The group multiplication in 𝐺 1 ★ 𝐺 2 is then inductively defined as

(𝑥1 , . . . , 𝑥 𝑛 ) · (𝑥 𝑛+1 , . . . , 𝑥 𝑛+𝑚 )





(𝑥1 , . . . , 𝑥 𝑛−1 , 𝑥 𝑛 · 𝑥 𝑛+1 , 𝑥 𝑛+2 , . . . , 𝑥 𝑛+𝑚 ) if (𝑥 𝑛 , 𝑥 𝑛+1 ∈ 𝐺 1 or 𝑥 𝑛 , 𝑥 𝑛+1 ∈ 𝐺 2 )






and 𝑥 𝑛 · 𝑥 𝑛+1 ≠ 1,


:= (𝑥1 , . . . , 𝑥 𝑛−1 ) · (𝑥 𝑛+2 , . . . , 𝑥 𝑛+𝑚 ) if (𝑥 𝑛 , 𝑥 𝑛+1 ∈ 𝐺 1 or 𝑥 𝑛 , 𝑥 𝑛+1 ∈ 𝐺 2 )





and 𝑥 𝑛 · 𝑥 𝑛+1 = 1,

 (𝑥1 , . . . , 𝑥 𝑛 , 𝑥 𝑛+1 , . . . , 𝑥 𝑛+𝑚 )
 otherwise.

This turns 𝐺 1 ★𝐺 2 into a group with neutral element the empty sequence (). The inverse element
for (𝑥1 , . . . 𝑥 𝑛 ) is given by (𝑥 𝑛−1 , . . . , 𝑥1−1 ) because of



(𝑥1 , . . . , 𝑥 𝑛 ) · 𝑥 𝑛−1 , . . . , 𝑥1−1 = (𝑥1 , . . . , 𝑥 𝑛−1 ) · 𝑥 𝑛−1
−1
, . . . , 𝑥1−1 = · · · = (𝑥1 ) (𝑥1−1 ) = ().

Now consider the map 𝑖 𝑗 : 𝐺 𝑗 → 𝐺 1 ★ 𝐺 2 given by

(
(𝑥), 𝑥 ≠ 1
𝑖 𝑗 (𝑥) =
(), 𝑥 = 1

For 𝜑 𝑗 : 𝐺 𝑗 → 𝐻 homomorphisms we put

(𝜑1 ★ 𝜑2 ) (𝑥1 , . . . , 𝑥 𝑛 ) := 𝜑𝑖1 (𝑥1 ) · 𝜑𝑖2 (𝑥2 ) · . . . · 𝜑𝑖𝑛 (𝑥 𝑛 )

where 𝑖 𝑗 is chosen such that 𝑥 𝑗 ∈ 𝐺 𝑖 𝑗 .

Remark 2.62
1.) The subset 𝑖 𝑗 (𝐺 𝑗 ) ⊂ 𝐺 1 ★ 𝐺 2 is a subgroup isomorphic to 𝐺 𝑗 . The intersection 𝑖1 (𝐺 1 ) ∩
𝑖2 (𝐺 2 ) = {1} is trivial. The union 𝑖1 (𝐺 1 ) ∪ 𝑖 2 (𝐺 2 ) generates 𝑖1 (𝐺 1 ) ★ 𝑖 2 (𝐺 2 ) as a group.

2.) If 𝐺 1 = {1} then 𝐺 1 ★ 𝐺 2 = 𝑖 2 (𝐺 2 )  𝐺 2 . Similarly, if 𝐺 2 = {1} then 𝐺 1 ★ 𝐺 2  𝐺 1 .

3.) If 𝐺 1 ≠ {1} and 𝐺 2 ≠ {1} then we may choose 𝑥 ∈ 𝐺 1 \ {1} and 𝑦 ∈ 𝐺 2 \ {1}. Then

(𝑥), (𝑥, 𝑦), (𝑥, 𝑦, 𝑥), (𝑥, 𝑦, 𝑥, 𝑦), . . .

yields infinitely many pairwise different elements in 𝐺 1 ★ 𝐺 2 , hence |𝐺 1 ★ 𝐺 2 | = ∞ (even if

|𝐺 𝑗 | < ∞). In addition, we have

(𝑥) · (𝑦) = (𝑥, 𝑦) ≠ (𝑦, 𝑥) = (𝑦) · (𝑥) ,

hence the group 𝐺 1 ★ 𝐺 2 is not abelian, even if 𝐺 1 and 𝐺 2 are.

2 Otherwise replace 𝐺 2 by an isomorphic group which is disjoint to 𝐺 1 .

43
2. Homotopy Theory

Remark 2.63. Usually one identifies 𝐺 1 with 𝑖 1 (𝐺 1 ) and 𝐺 2 with 𝑖2 (𝐺 2 ) and writes

𝑥1 · 𝑥2 · . . . · 𝑥 𝑛 instead of (𝑥1 , 𝑥2 , . . . , 𝑥 𝑛 ).

Example 2.64. Consider 𝐺 1 = Z/2Z = {1, −1} and 𝐺 2 = Z/2Z = {1′ , −1′ }. Elements of
Z/2Z ★ Z/2Z are for example

𝑎 = (−1) · (−1′ ) · (−1) · (−1′ )

𝑏 = (−1′ ) · (−1) · (−1′ ) · (−1) · (−1′ ) .

Now we calculate

𝑎 · 𝑏 = (−1) · (−1′ ) · (−1) ·(−1′ ) · (−1′ ) ·(−1) · (−1′ ) · (−1) · (−1′ )

| {z }
1′
= (−1) · (−1′ ) ·(−1) · (−1) ·(−1′ ) · (−1) · (−1′ )
| {z }
1
= (−1) ·(−1′ ) · (−1′ ) ·(−1) · (−1′ )
| {z }
1′
= (−1) · (−1) ·(−1′ )
| {z }
1
′
= (−1 ) .

Now we are ready to return to topology.

Proposition 2.65. Let 𝑋 be a topological space and let 𝑈, 𝑉 ⊂ 𝑋 be open such that 𝑈 ∪𝑉 = 𝑋.
Let 𝑥0 ∈ 𝑈 ∩ 𝑉. Furthermore, assume that 𝑈, 𝑉 and 𝑈 ∩ 𝑉 are path-connected. Then 𝑋 is
path-connected and the map

𝑖 # ★ 𝑗 # : 𝜋1 (𝑈; 𝑥0 ) ★ 𝜋1 (𝑉; 𝑥0 ) → 𝜋1 (𝑋; 𝑥0 )

is onto where 𝑖 : 𝑈 → 𝑋 and 𝑗 : 𝑉 → 𝑋 are the corresponding inclusion maps.

Proof. First of all we note that the space 𝑋 is path-connected because each point in 𝑋 lies in 𝑈
or in 𝑉 and can therefore be connected to 𝑥0 by a path.
The statement of the proposition is equivalent to saying that



𝑖 # 𝜋1 (𝑈; 𝑥0 ) ∪ 𝑗 # 𝜋1 (𝑉; 𝑥0 )

generates 𝜋1 (𝑋; 𝑥0 ) as a group. Now let [𝜔] ∈ 𝜋1 (𝑋; 𝑥0 ) and subdivide the unit interval 𝐼 = [0, 1]
by 0 = 𝑡 0 ≤ 𝑡 1 · · · < 𝑡 𝑛 = 1 such that 𝜔([𝑡𝑖 , 𝑡 𝑖+1 ]) ⊂ 𝑈 or ⊂ 𝑉.

44
 2.4. The Seifert-van Kampen theorem

𝑋
𝜔0 𝜔2

𝑈
𝑥0

𝑈 ∩𝑉 𝜂1 𝜂2 𝜂3 𝑈 ∩𝑉
𝜔(𝑡1 ) 𝜔(𝑡2 ) 𝜔(𝑡 3 )

𝜔1

Figure 24. Subdividing 𝜔

By removing subdivision points if necessary we can assume that if 𝜔([𝑡𝑖−1 , 𝑡𝑖 ]) ⊂ 𝑈 then

𝜔([𝑡𝑖 , 𝑡𝑖+1 ]) ⊂ 𝑉 or conversely. Reparametrize 𝜔𝑖 (𝑡) = 𝜔((1 − 𝑡)𝑡𝑖 + 𝑡𝑡 𝑖+1 ). Then 𝜔𝑖 ∈
Ω(𝑈 or 𝑉; 𝜔(𝑡 𝑖 ), 𝜔(𝑡 𝑖+1 )) and in particular 𝜔(𝑡 𝑖 ) ∈ 𝑈 ∩ 𝑉. Since by assumption 𝑈 ∩ 𝑉 is
path-connected there exist 𝜂 𝑗 ∈ Ω(𝑈 ∩ 𝑉; 𝑥0 , 𝜔(𝑡 𝑗 )). Then

𝜔0 ★ 𝜂1−1 , 𝜂1 ★ 𝜔1 ★ 𝜂2−1 , 𝜂2 ★ 𝜔2 ★ 𝜂3−1 , . . . , 𝜂 𝑛−1 ★ 𝜔 𝑛−1

are loops with base point 𝑥0 contained entirely in 𝑈 or 𝑉. We then calculate

𝜔0 ★ 𝜂1−1 ★ 𝜂1 ★ 𝜔1 ★ 𝜂2−1 ★ 𝜂2 ★ 𝜔2 ★ 𝜂3−1 ★ · · · ★ (𝜂 𝑛−1 ★ 𝜔 𝑛−1 )
≃ {0,1} 𝜔0 ★ 𝜔1 ★ · · · ★ 𝜔 𝑛−1 ≃ {0,1} 𝜔

in 𝑋. Denoting the homotopy class of a loop 𝜔 in 𝑋, 𝑈, 𝑉, or 𝑈 ∩ 𝑉 by [𝜔] 𝑋 , [𝜔]𝑈 , [𝜔] 𝑉 , and

[𝜔]𝑈∩𝑉 respectively, we have
     
[𝜔] 𝑋 = 𝜔0 ★ 𝜂1−1 𝑋 · 𝜂1 ★ 𝜔1 ★ 𝜂2−1 𝑋 · . . . · 𝜂 𝑛−1 ★ 𝜔 𝑛−1 𝑋

and it follows that

        
[𝜔] 𝑋 = 𝑖# 𝜔0 ★ 𝜂1−1 𝑈 · 𝑗 # 𝜂1 ★ 𝜔1 ★ 𝜂2−1 𝑉 · 𝑖 # 𝜂2 ★ 𝜔2 ★ 𝜂3−1 𝑈 . . .
    h  
= (𝑖 # ★ 𝑗# ) 𝜔0 ★ 𝜂1−1 𝑈 · 𝜂1 ★ 𝜔1 ★ 𝜂2−1 𝑉 · 𝜂2 ★ 𝜔2 ★ 𝜂3−1 𝑈 . . .
| {z }
∈ 𝜋1 (𝑈;𝑥0 )★𝜋1 (𝑉;𝑥0 )

which proves the assertion. 

45
2. Homotopy Theory

Corollary 2.66. Let 𝑋 is a topological space and let 𝑈, 𝑉 ⊂ 𝑋 be open such that 𝑈 ∪ 𝑉 = 𝑋
and 𝑈 ∩ 𝑉 ≠ ∅. If 𝑈 and 𝑉 are 1-connected and 𝑈 ∩ 𝑉 is path-connected, then the space 𝑋 is
1-connected.

Proof. Since
{1} = {1} ★ {1} = 𝜋1 (𝑈; 𝑥0 ) ★ 𝜋1 (𝑉; 𝑥0 ) → 𝜋1 (𝑋; 𝑥0 )
is onto we get that 𝜋1 (𝑋; 𝑥0 ) = {1}. 

Example 2.67. Consider 𝑋 = 𝑆 𝑛 and put 𝑈 = 𝑆 𝑛 \ {𝑒1 }. The stereographic projection yields
a homeomorphism 𝑈 → R𝑛 . Hence 𝑈 is 1-connected. Similarly, 𝑉 = 𝑆 𝑛 \ {−𝑒1 } is also
1-connected. Now
𝑈 ∩ 𝑉 = 𝑆 𝑛 \ {𝑒1 , −𝑒1 } ≈ R𝑛 \ {0}.
For 𝑛 ≥ 2 the space 𝑈 ∩ 𝑉 is path-connected. Corollary 2.66 shows that 𝑆 𝑛 is simply connected
for 𝑛 ≥ 2.
Recall that we know already that this is not true for 𝑛 = 1 because 𝜋1 (𝑆1 ; 1)  Z.

To determine 𝜋1 (𝑋; 𝑥0 ) more precisely we compute the kernel of the homomorphism

𝑖 # ★ 𝑗# : 𝜋1 (𝑈; 𝑥0 ) ★ 𝜋1 (𝑉; 𝑥0 ) → 𝜋1 (𝑋; 𝑥0 ).
𝑖 /𝑋
𝑈O O
Consider the inclusion maps 𝑖 ′
: 𝑈 ∩ 𝑉 → 𝑈 and : 𝑈 ∩ 𝑉 → 𝑉. 𝑗′
𝑖′ 𝑗
Clearly the diagram on the right commutes. This implies
𝑈 ∩𝑉 ′ /𝑉
𝑗
𝑗 # ◦ 𝑗 #′ = 𝑖 # ◦ 𝑖#′ .
For 𝛼 ∈ 𝜋1 (𝑈 ∩ 𝑉; 𝑥0 ) we calculate


 −1



1 = 𝑖 # 𝑖 #′ (𝛼) · 𝑗 # 𝑗#′ (𝛼) = 𝑖 # 𝑖 #′ (𝛼) · 𝑗 # 𝑗#′ (𝛼) −1 = (𝑖 # ★ 𝑗 # ) 𝑖 #′ (𝛼) · 𝑗 #′ (𝛼) −1 .
| {z }
∈ 𝜋1 (𝑈;𝑥0 )★𝜋1 (𝑉;𝑥0 )

Hence 𝑖 #′ (𝛼) · 𝑗 #′ (𝛼) −1 ∈ ker(𝑖 # ★ 𝑗# ) and it follows that



N 𝑖 #′ (𝛼) · 𝑗 #′ (𝛼) −1 | 𝛼 ∈ 𝜋1 (𝑈 ∩ 𝑉; 𝑥0 ) ⊂ ker(𝑖 # ★ 𝑗# ) .

Theorem 2.68 (Seifert-van Kampen). Let 𝑋 be a topological space and let 𝑈, 𝑉 ⊂ 𝑋 be open
subsets such that 𝑈 ∪ 𝑉 = 𝑋 and 𝑥0 ∈ 𝑈 ∩ 𝑉. Let 𝑈, 𝑉 and 𝑈 ∩ 𝑉 be path connected. Then
the map 𝑖# ★ 𝑗 # induces an isomorphism

′
 𝜋1 (𝑋; 𝑥0 ).
𝜋1 (𝑈; 𝑥0 ) ★ 𝜋1 (𝑉; 𝑥0 )
N 𝑖 # (𝛼) · 𝑗 #′ (𝛼) −1 | 𝛼 ∈ 𝜋1 (𝑈 ∩ 𝑉; 𝑥0 )

46
 2.4. The Seifert-van Kampen theorem

Proof. It remains to show that



ker(𝑖 # ★ 𝑗 # ) ⊂ N 𝑖#′ (𝛼) · 𝑗 #′ (𝛼) −1 | 𝛼 ∈ 𝜋1 (𝑈 ∩ 𝑉; 𝑥0 ) .

Let 𝜔1 , 𝜔3 , · · · ∈ Ω(𝑈; 𝑥0 ) and 𝜔2 , 𝜔4 , · · · ∈ Ω(𝑉; 𝑥0 ) be such that

1 = 𝑖 # ★ 𝑗 # [𝜔1 ]𝑈 · [𝜔2 ] 𝑉 · [𝜔3 ]𝑈 · . . .
= 𝑖 # ([𝜔1 ]𝑈 ) · 𝑗 # ([𝜔2 ] 𝑉 ) · 𝑖 # ([𝜔3 ]𝑈 ) · . . .
= [𝜔1 ] 𝑋 · [𝜔2 ] 𝑋 · [𝜔3 ] 𝑋 · . . .
= [𝜔1 ★ 𝜔2 ★ 𝜔3 ★ . . . ] 𝑋

1 −

Then there exists a homotopy 𝐻 : [0, 1] ×

[0, 1] → 𝑋 such that

𝐻 (𝑡, 0) = (𝜔1 ★ 𝜔2 ★ 𝜔3 ★ . . . ) (𝑡),

𝐻 (𝑡, 1) = 𝑥0 ,
𝐻 (0, 𝑠) = 𝐻 (1, 𝑠) = 𝑥0 .
0 − | | |
In Figure 25 the red area gets mapped to 𝑥0 . 𝜔1 𝜔2 𝜔𝑁
| |
𝑡
0 1
Figure 25. The homotopy to start with

Now subdivide [0, 1] × [0, 1] further such that 𝑣

𝐻 maps each closed subsquare entirely to 𝑈 or
to 𝑉.

| | |
𝜔1 𝜔2 𝜔𝑁
Figure 26. Subdividing the homotopy

47
2. Homotopy Theory

𝑢
>

𝑙 𝑟
∧ ∧
Considering the edges of the subsquares we
get homotopies

>
𝑑

Figure 27. Deforming the homotopy in each

square

and hence the relation

𝑑 ★ 𝑟 ≃ {0,1} 𝑙 ★ 𝑢 (2.2)
resulting in
(𝐻 ◦ 𝑑) ★ (𝐻 ◦ 𝑟) ≃ {0,1} (𝐻 ◦ 𝑙) ★ (𝐻 ◦ 𝑢) in 𝑈 or in 𝑉 .
For each vertex 𝑣 in this subdivision of [0, 1] × [0, 1] choose 𝜂 𝑣 ∈ Ω(𝑋; 𝑥0 , 𝐻 (𝑣)) in such a way
that 𝜂 𝑣 ∈ Ω(𝑊; 𝑥0 , 𝐻 (𝑣)) if 𝐻 (𝑣) ∈ 𝑊 where 𝑊 = 𝑈, 𝑉 or 𝑈 ∩ 𝑉. This is possible because
𝑥0 ∈ 𝑊 and 𝑊 is path-connected by assumption. If 𝐻 (𝑣) = 𝑥0 choose 𝜂 𝑣 = 𝜀 𝑥0 . For each edge
with endpoints 𝑣 0 and 𝑣 1 we obtain a loop in 𝑈 or in 𝑉 by
𝜂 𝑣0 ★ (𝐻 ◦ 𝑐) ★ 𝜂 𝑣−1
1
∈ Ω(𝑈 or 𝑉; 𝑥0 ) .
Now look at one row of the subdivision, see Figure 28. We find that
𝐷 𝑖 := 𝜂 𝑑𝑖 (0) ★ (𝐻 ◦ 𝑑𝑖 ) ★ 𝜂 𝑑−1𝑖 (1) ∈ Ω(𝑈 or 𝑉; 𝑥0 ).
Similarly we define 𝐿 𝑖 , 𝑅𝑖 , 𝑈𝑖 ∈ Ω(𝑈 or 𝑉; 𝑥0 ) and we conclude that by (2.2)
[𝐷 𝑖 ] 𝑊𝑖 · [𝑅𝑖 ] 𝑊𝑖 = [𝐿 𝑖 ] 𝑊𝑖 · [𝑈𝑖 ] 𝑊𝑖 ∈ 𝜋1 (𝑊𝑖 ; 𝑥0 ) (2.3)
where 𝑊𝑖 = 𝑈or 𝑉. We now compute in 𝜋1 (𝑈; 𝑥0 ) ★ 𝜋1 (𝑉; 𝑥0 ) modulo N ({𝑖 #′ (𝛼) · 𝑗 #′ (𝛼) −1 | 𝛼 ∈
𝜋1 (𝑈 ∩ 𝑉; 𝑥0 )}):
[𝐷 1 ] 𝑊1 · [𝐷 2 ] 𝑊2 · . . . · [𝐷 𝑁 ] 𝑊𝑁 = [𝐷 1 ] 𝑊1 · [𝐷 2 ] 𝑊2 · . . . · [𝐷 𝑁 ] 𝑊𝑁 · [𝑅 𝑁 ] 𝑊𝑁
= [𝐷 1 ] 𝑊1 · . . . · [𝐷 𝑁 −1 ] 𝑊𝑁 −1 · [𝐿 𝑁 ] 𝑊𝑁 · [𝑈 𝑁 ] 𝑊𝑁
= [𝐷 1 ] 𝑊1 · . . . · [𝐷 𝑁 −1 ] 𝑊𝑁 −1 · [𝑅 𝑁 −1 ] 𝑊𝑁 · [𝑈 𝑁 ] 𝑊𝑁 .
If now 𝑊 𝑁 −1 = 𝑊 𝑁 then we can apply (2.3) once more and get
[𝐷 1 ] 𝑊1 · [𝐷 2 ] 𝑊2 ·. . .· [𝐷 𝑁 ] 𝑊𝑁 = [𝐷 1 ] 𝑊1 ·. . .· [𝐷 𝑁 −2 ] 𝑊𝑁 −2 · [𝐿 𝑁 −1 ] 𝑊𝑁 −1 · [𝑈 𝑁 −1 ] 𝑊𝑁 −1 · [𝑈 𝑁 ] 𝑊𝑁
(2.4)

48
 2.4. The Seifert-van Kampen theorem

𝑢1 𝑢2 𝑢𝑁
> > >
𝜀 𝑥0 = 𝑙 1 ∧ ∧ ∧ ∧ ∧ 𝑟 𝑁 = 𝜀 𝑥0
> > >
𝑑1 𝑑2 𝑑𝑁

Figure 28. Deforming along one row

In case 𝑊 𝑁 −1 ≠ 𝑊 𝑁 , say 𝑊 𝑁 −1 = 𝑈 and 𝑊 𝑁 = 𝑉, then 𝐿 𝑁 = 𝑅 𝑁 −1 ∈ Ω(𝑈 ∩ 𝑉; 𝑥0 ). Hence

mod N (... ) ′
[𝑅 𝑁 −1 ] 𝑉 = 𝑗 #′ ([𝑅 𝑁 −1 ]𝑈∩𝑉 ) = 𝑖 # ([𝑅 𝑁 −1]𝑈∩𝑉 ) = [𝑅 𝑁 −1 ]𝑈 .

In
𝜋1 (𝑈; 𝑥0 ) ★ 𝜋1 (𝑉; 𝑥0 )
𝐺 :=
N ({𝑖 #′ (𝛼) · 𝑗 #′ (𝛼) −1 | 𝛼 ∈ 𝜋1 (𝑈 ∩ 𝑉; 𝑥0 )})
the computation (2.4) is still possible. By induction on all squares in the row from right to the
left we find that

z }| {
=1

[𝐷 1 ] 𝑊1 · . . . · [𝐷 𝑁 ] 𝑊𝑁 = [𝐿 1 ] 𝑊1 ·[𝐷 1 ] 𝑊1 · . . . · [𝐷 𝑁 ] 𝑊𝑁 = [𝑈1 ] 𝑊1 · . . . · [𝑈 𝑁 ] 𝑊𝑁 .

A second induction on all rows from bottom to top yields in 𝐺

mod N (... )
[𝜔1 ]𝑈 · [𝜔2 ] 𝑉 · [𝜔3 ]𝑈 · . . . = [𝜀 𝑥0 ] 𝑊1 · [𝜀 𝑥0 ] 𝑊2 · [𝜀 𝑥0 ] 𝑊3 · . . . = 1 .

We have shown that in 𝜋1 (𝑈; 𝑥0 ) ★ 𝜋1 (𝑉; 𝑥0 )

[𝜔1 ]𝑈 · [𝜔2 ] 𝑉 · [𝜔3 ]𝑈 · . . . ∈ N ({𝑖 #′ (𝛼) · 𝑗 #′ (𝛼) −1 | 𝛼 ∈ 𝜋1 (𝑈 ∩ 𝑉; 𝑥0 )})

and hence 

ker(𝑖 # ★ 𝑗# ) ⊂ N 𝑖 #′ (𝛼) · 𝑗 #′ (𝛼) −1 | 𝛼 ∈ 𝜋1 (𝑈 ∩ 𝑉; 𝑥0 ) . 

49
2. Homotopy Theory

Corollary 2.69. Let 𝑋 be a topological space. Let 𝑈, 𝑉 ⊂ 𝑋 be open subsets such that
𝑈 ∪ 𝑉 = 𝑋 and let 𝑥0 ∈ 𝑈 ∩ 𝑉. Assume that 𝑈 and 𝑉 are path connected and that 𝑈 ∩ 𝑉 is
1-connected. Then
𝜋1 (𝑋; 𝑥0 )  𝜋1 (𝑈; 𝑥0 ) ★ 𝜋1 (𝑉; 𝑥0 )
where the isomorphism is induced by the inclusion maps.

Proof. By assumption 𝜋1 (𝑈 ∩ 𝑉, 𝑥0 ) = {1} and hence

N ({𝑖 #′ (𝛼) · 𝑗 #′ (𝛼) −1 | 𝛼 ∈ 𝜋1 (𝑈 ∩ 𝑉; 𝑥0 )}) = {1} .

The assertion then follows from Theorem 2.68. 

Example 2.70. Consider the figure 8 space and the two subsets 𝑈 and 𝑉 as indicated in the
picture:
𝑋
𝑉 𝑈
b b

𝑥0 𝑥0

Figure 29. Covering the figure 8

It is easy to see that 𝑈 ≃ 𝑆1 and also 𝑉 ≃ 𝑆1 . Moreover, the intersection

𝑈 ∩ 𝑉 ≃ point

is 1-connected. Hence we have

𝜋1 (𝑋; 𝑥0 )  𝜋1 (𝑆1 ; 𝑥0 ) ★ 𝜋1 (𝑆1 ; 𝑥0 )  Z ★ Z .

Example 2.71. Let 𝑋 be a connected 𝑛-dimensional manifold with 𝑛 ≥ 3.

𝑀
𝑉
b
𝑝

Figure 30. Punctured manifold

Let 𝑝 ∈ 𝑋 and 𝑈 := 𝑋 \ {𝑝}. Now let 𝑉 be an open neighborhood of 𝑝 homeomorphic to R𝑛 ,

which is contractible and hence 1-connected. Then the space

𝑈 ∩ 𝑉 ≈ R𝑛 \ {0} ≃ 𝑆 𝑛−1

50
 2.5. The fundamental group of surfaces

is 1-connected. By Corollary 2.69 we then deduce

𝜋1 (𝑋; 𝑥0 )  𝜋1 (𝑈; 𝑥0 ) ★ 𝜋1 (𝑉; 𝑥0 ) = 𝜋1 (𝑋 \ {𝑝}; 𝑥0 ) .

Removing a point from a manifold of dimension at least 3 does not change its fundamental group.

Example 2.72. Let 𝑀 and 𝑁 be two connected manifold of dimension 𝑛 ≥ 3. Let 𝑋 = 𝑀#𝑁.

𝑀 𝑁

Figure 31. Start with two manifolds...

Now choose 𝑈 and 𝑉 as in Figure 32. Then we have that 𝑈 ≈ 𝑀 \ {𝑝} and 𝑉 ≈ 𝑁 \ {𝑞}.

𝑈 ∩𝑉

| {z }
=𝑈 | {z }
=𝑉
Figure 32. ... and consider their connected sum.

Since
𝑈 ∩ 𝑉 ≈ 𝑆 𝑛−1 × (0, 1) ≃ 𝑆 𝑛−1
is 1-connected we find by Corollary 2.69 once again that

𝜋1 (𝑀#𝑁)  𝜋1 (𝑈) ★ 𝜋1 (𝑉)  𝜋1 (𝑀) ★ 𝜋1 (𝑁) .

For example, if 𝑀 = 𝑁 = 𝑇 3 then 𝜋1 (𝑇 3#𝑇 3 ) = (Z3 ) ★ (Z3 ).

Remark 2.73. We now see easily that the torus 𝑇 𝑛 with 𝑛 ≥ 3 cannot be homotopy equivalent
to the connected sum 𝑇 𝑛 ≃ 𝑀#𝑁 of two non-1-connected manifolds 𝑀 and 𝑁.3 If it were
possible then 𝜋1 (𝑇 𝑛 )  𝜋1 (𝑀) ★ 𝜋1 (𝑁) would not be abelian but we know that 𝜋1 (𝑇 𝑛 )  Z𝑛 , a
contradiction.

2.5. The fundamental group of surfaces

Our aim is to prove that orientable compact connected surfaces of different genus (as depicted
by the pastries in Example 1.5) are not homotopy equivalent and therefore not homeomorphic.
3 If one allows simply connected summands then it is of course possible, 𝑇 𝑛 ≈ 𝑇 𝑛 #𝑆 𝑛 .

51
2. Homotopy Theory

Definition 2.74. We call 𝐹𝑔 := 𝑇 2 # · · · #𝑇 2 a surface of genus 𝑔 ≥ 1.

| {z }
𝑔−times

𝐹𝑔 = ···

| {z }
𝑔−times

Figure 33. Surface of genus 𝑔

Remark 2.75. We also put 𝐹0 := 𝑆2 .

We now want to compute 𝜋1 (𝐹𝑔 ) and show that the fundamental groups for surfaces of different
genus are not isomorphic. This then shows in particular that they are not homotopy equivalent.

Proposition 2.76. For any 𝑔 ≥ 1

𝐹𝑔 = 𝑇 2 # · · · #𝑇 2 ≈ 𝐷 2 /∼
| {z }
𝑔−times

where 𝑥 ∼ 𝑦 iff 𝑥 = 𝑦 or 𝑥, 𝑦 ∈ 𝜕𝐷 2 and are identified according to the following scheme:

𝑏 𝑔−1 𝑎1
𝑎 𝑔−1 𝑏1
𝑎1−1

𝑏1−1
··· ≈ 𝐷2
𝑎2

𝑏2
𝑎2−1

Figure 34. Building a surface from the disk

This identification is to be understood as follows: Each line segment labelled by 𝑎𝑖 (or 𝑏 𝑗

respectively) is identified with with 𝑎𝑖−1 (or 𝑏 −1
𝑗 ) with respect to the direction indicated by the
arrow. Note that there are 2𝑔 labels 𝑎𝑖 , 𝑏𝑖 , 𝑖 = 1, .., 𝑔.

52
 2.5. The fundamental group of surfaces

Proof. We do an induction on 𝑔.

Induction basis for 𝑔 = 1: We see that the labelled disk 𝐷 2 /∼ is homeomorphic to the labelled
𝑊 2 /∼ which is homeomorphic to the two-dimensional torus, i.e. to 𝐹1 .

𝑎1
𝑏1−1 𝑎1 𝑎1

≈ 𝑏1−1 𝑏1 ≈ 𝑏1−1 𝑏1 ≈

𝑎1−1 𝑏1

𝑎1−1
Figure 35. Starting the induction with the torus

Inductive step, 𝑔 − 1 ⇒ 𝑔: We perform a cut along the line 𝑐.

𝑏 𝑔−1 𝑎1 𝑎1
𝑏1 𝑏1
𝑎 𝑔−1 𝑏 𝑔−1
𝑎1−1 𝑎 𝑔−1 𝑎1−1
𝑐
𝑏1−1 𝑏1−1
≈ 𝑐
𝑎2
𝑎2
𝑏2
𝑏2
𝑎2−1
𝑎2−1
Figure 36. Induction step by cutting off a segment

The endpoints of 𝑐 in the two pieces are identified. This yields a homeomorphism where the
interior of the now closed loop 𝑐 is cut out of the two remaining disks.

53
2. Homotopy Theory

𝑎1
𝑏1
𝑏 𝑔−1
𝑎1−1
𝑎 𝑔−1 𝑏 𝑔−1 𝑎2
𝑎 𝑔−1 𝑏2
𝑐 𝑏1−1 𝑎2−1 𝑏1−1 𝑎1
𝑐
𝑐 𝑏2−1 𝑐
≈
𝑎2

𝑏2 𝑎1−1 𝑏1

𝑎2−1
Figure 37. Closing the loop 𝑐

By induction hypothesis the left-hand disc is homeomorphic to a surface of genus 𝑔 − 1 with

a disc removed that is bounded by 𝑐. The right-hand disc becomes a torus, also with a disc
removed that is bounded by 𝑐.
Gluing these two spaces together along 𝑐, we obtain a space which is homeomorphic to a surface
of genus 𝑔:

𝑐 𝑐 ≈

| {z } | {z }
(𝑔−1) −times 𝑔−times

Figure 38. Completing the induction step

Remark 2.77. We note that 𝐹0 ≈ 𝐷 2 /𝜕𝐷 2 .

Remark 2.78. Let 𝐺 be a group. Assume 𝐺 is generated by 𝑔1 , ..., 𝑔𝑛 ∈ 𝐺 as a group. We have

group homomorphisms 𝜑𝑖 : Z → 𝐺, 𝜑𝑖 (𝑘) = 𝑔𝑖𝑘 . Repeated application of the universal property
of free products of groups yields the group homomorphism

(...(𝜑1 ★ 𝜑2 ) ★ 𝜑3 ) ★ · · · ★ 𝜑 𝑛 =: 𝜑1 ★ · · · ★ 𝜑 𝑛 : (...(Z ★ Z) ★ Z) ★ · · · ★ Z =: Z ★ · · · ★ Z → 𝐺

with
(𝜑1 ★ · · · ★ 𝜑 𝑛 ) ((𝑘 1 , 𝑖 1 ), ..., (𝑘 𝑟 , 𝑖𝑟 )) = 𝑔𝑖𝑘11 · · · 𝑔𝑖𝑘𝑟𝑟 ,
where 𝑖 𝑗 ∈ {1, ..., 𝑛} is an index used to make the Z-factors formally disjoint.

54
 2.5. The fundamental group of surfaces

The fact that 𝑔1 , ..., 𝑔𝑛 generate 𝐺 is equivalent to the fact that 𝜑1 ★ · · · ★ 𝜑 𝑛 : Z ★ · · · ★ Z → 𝐺

is onto. It follows that
Z★···★Z
𝐺
ker(𝜑1 ★ · · · ★ 𝜑 𝑛 )

If the normal subgroup ker(𝜑1 ★ · · · ★ 𝜑 𝑛 ) is also finitely generated as a group, with generators
𝑥1 , ..., 𝑥 𝑚 , then we call 𝐺 finitely presentable and

h𝑔1 , ..., 𝑔𝑛 | 𝑥1 , ..., 𝑥 𝑚 i

a presentation of 𝐺.

Example 2.79. For the cartesian product Z2 of two copies of Z we have the presentation

Z2  h𝑎, 𝑏 | 𝑎𝑏𝑎 −1 𝑏 −1 i.

The cyclic group Z/2Z of order 2 is presentable as

Z/2Z  h𝑎 | 𝑎2 i.

Remark 2.80. A word of caution: isomorphic groups may have several, quite different presen-
tations. For example

h𝑥, 𝑦 | 𝑥𝑦𝑥𝑦 −1 𝑥 −1 𝑦 −1 i  h𝑥, 𝑦, 𝑤, 𝑧 | 𝑥𝑦𝑥𝑦 −1𝑥 −1 𝑦 −1 , 𝑥𝑦𝑥𝑤 −1 , 𝑧𝑦 −1 𝑥 −1 i

because the generators 𝑥𝑦𝑥𝑤 −1 and 𝑧𝑦 −1 𝑥 −1 on the right-hand side can be used to eleminate 𝑤
and 𝑧. It is therefore often not obvious whether two presentations give rise to isomorphic groups.

Theorem 2.81. For any 𝑔 ∈ N we have

𝜋1 (𝐹𝑔 )  h𝑎1 , 𝑏1 , ..., 𝑎 𝑔 , 𝑏 𝑔 | 𝑎1 𝑏1 𝑎1−1 𝑏1−1 · · · 𝑎 𝑔 𝑏 𝑔 𝑎 𝑔−1 𝑏 𝑔−1 i.

Proof. Recall that 𝐹𝑔 ≈ 𝐷 2 /∼ as in Proposition 2.76. To apply the Seifert-van Kampen theo-
rem 2.68 we put 𝑈 := 𝐷˚ 2 and 𝑉 := (𝐷 2 /∼) \ 𝐷 2 ( 21 ). We have 𝐹𝑔 = 𝑈 ∪ 𝑉.

55
2. Homotopy Theory

𝑏 𝑔−1 𝑎1
𝑏1
𝑎 𝑔−1
𝑎1−1

𝑏1−1
𝑉
𝑎2

𝑏2

𝑎2−1
Figure 39. Applying the Seifert-van Kampen theorem to the disk representation

The subset 𝑈 is contractible and therefore 𝜋1 (𝑈) = {1}. The subset 𝑉 is homotopy equivalent
to the boundary 𝜕𝐷 subject to the identifications of the equivalence relation, 𝑉 ≃ 𝜕𝐷/∼. The
identifications induced by ∼ generate a bouquet of 2𝑔 circles, one for each relation 𝑎𝑖 and 𝑏𝑖 .
The bouquet is denoted by 𝑆1 ∨ · · · ∨ 𝑆1 , where the wedge sum “∨” of two topological spaces 𝑋
and 𝑌 is defined to be the disjoint union of 𝑋 and 𝑌 with identification of two base points 𝑥0 ∈ 𝑋,
𝑦 0 ∈ 𝑌 such that 𝑋 ∨ 𝑌 := 𝑋 ∪ 𝑌 /{𝑥0 ∼ 𝑦 0 }. For the bouquet of circles all 𝑆1 ’s are joined at the
same base point. Graphically we can depict the bouquet of circles as follows

𝑆1 ∨ · · · ∨ 𝑆1 =

Figure 40. Bouquet of circles

So we have
𝑉 ≃ 𝜕𝐷 2 /∼ ≈ 𝑆1 ∨ · · · ∨ 𝑆1 .
| {z }
2𝑔 times

The fundamental group of 𝑉 follows immediately from Example 2.70 by induction:

| {z }
𝜋1 (𝑉)  Z ★ Z ★ · · · ★ Z
2𝑔 times

with generators again denoted by 𝑎1 , 𝑏1 , 𝑎2 , 𝑏2 , ..., 𝑎 𝑔 , 𝑏 𝑔 . For the intersection of 𝑈 and 𝑉 we

have 𝑈 ∩ 𝑉 = 𝐷˚ 2 \ 𝐷¯ 2 ( 21 ) ≃ 𝑆1 and thus by Theorem 2.42 we have 𝜋1 (𝑈 ∩ 𝑉)  Z. A generator

56
 2.5. The fundamental group of surfaces

of 𝜋1 (𝑈 ∩ 𝑉) is given by a loop 𝑐 of degree 1.

𝑏 𝑔−1 𝑎1
𝑏1
𝑎 𝑔−1
𝑎1−1

𝑏1−1
𝑐
𝑎2

𝑏2

𝑎2−1

Figure 41. Finding the generator

The inclusion map 𝑖 ′ : 𝑈 ∩𝑉 → 𝑈 induces the trivial homomorphism 𝑖 #′ because 𝜋1 (𝑈) is trivial.
For the inclusion map 𝑗 ′ : 𝑈 ∩ 𝑉 → 𝑉 we note that the induced isomorphism maps 𝑐 onto
𝑎1 𝑏1 𝑎1−1 𝑏1−1 · · · 𝑎 𝑔 𝑏 𝑔 𝑎 𝑔−1 𝑏 𝑔−1 .
By the Seifert-van Kampen theorem 2.68 we find

𝜋1 (𝑈) ★ 𝜋1 (𝑉)
𝜋1 (𝐹𝑔 ) 
N ( 𝑗 #′ (𝛼)
· 𝑖 #′ (𝛼) −1 | 𝛼 ∈ 𝜋1 (𝑈 ∩ 𝑉))
𝜋1 (𝑉)
= ′
N ( 𝑗# (𝛼) | 𝛼 ∈ 𝜋1 (𝑈 ∩ 𝑉))
Z★···★Z
=
N ( 𝑗#′ (𝑐))
= h𝑎1 , 𝑏1 , ..., 𝑎 𝑔 , 𝑏 𝑔 | 𝑎1 𝑏1 𝑎1−1 𝑏1−1 · · · 𝑎 𝑔 𝑏 𝑔 𝑎 𝑔−1 𝑏 𝑔−1 i. 

Example 2.82. For the two-dimensional torus 𝑇 2 we find with Example 2.79

𝜋1 (𝑇 2) = 𝜋1 (𝐹1 ) = h𝑎, 𝑏 | 𝑎𝑏𝑎 −1 𝑏 −1 i  Z2 ,

in agreement with Remark 2.48.

Corollary 2.83. For 𝑔, 𝑔′ ∈ N, 𝑔 ≠ 𝑔′ we have 𝐹𝑔 ; 𝐹𝑔′ and hence 𝐹𝑔 0 𝐹𝑔′ .

57
2. Homotopy Theory

Proof. The statement follows once we see that 𝜋1 (𝐹𝑔 )  𝜋1 (𝐹𝑔′ ). Attention here: as noted in
2.80 different presentations can yield isomorphic groups.
For any group 𝐺 let [𝐺, 𝐺] be the normal subgroup generated by all commutators 𝑎𝑏𝑎 −1 𝑏 −1 ,
𝑎, 𝑏 ∈ 𝐺. The abelian factor group 𝐺/[𝐺, 𝐺] is called the abelianization of 𝐺. We now calculate
the abelianization of 𝜋1 (𝐹𝑔 ).
𝜋1 (𝐹𝑔 )
= h𝑎1 , 𝑏1 , ..., 𝑎 𝑔 , 𝑏 𝑔 | 𝑎1 𝑏1 𝑎1−1 𝑏1−1 · · · 𝑎 𝑔 𝑏 𝑔 𝑎 𝑔−1 𝑏 𝑔−1 , 𝑎1 𝑏1 𝑎1−1 𝑏1−1 , ...
[𝜋1 (𝐹𝑔 ), 𝜋1 (𝐹𝑔 )]
..., 𝑎1 𝑎2 𝑎1−1 𝑎2−1 , ...i
= h𝑎1 , 𝑏1 , ..., 𝑎 𝑔 , 𝑏 𝑔 | 𝑎1 𝑏1 𝑎1−1 𝑏1−1 , ..., 𝑎1 𝑎2 𝑎1−1 𝑎2−1 , ...i
 Z2𝑔 ,

where the second equality follows because the simple commutators 𝑎𝑖 𝑏𝑖 𝑎𝑖−1 𝑏𝑖−1 , 𝑖 = 1, ..., 𝑔
generate the relation 𝑎1 𝑏1 𝑎1−1 𝑏1−1 · · · 𝑎 𝑔 𝑏 𝑔 𝑎 𝑔−1 𝑏 𝑔−1 .
Hence if 𝐹𝑔 ≃ 𝐹𝑔′ then 𝜋1 (𝐹𝑔 )  𝜋1 (𝐹𝑔′ ) and thus

𝜋1 (𝐹𝑔 )/[𝜋1 (𝐹𝑔 ), 𝜋1 (𝐹𝑔 )]  𝜋1 (𝐹𝑔′ )/[𝜋1 (𝐹𝑔′ ), 𝜋1 (𝐹𝑔′ )].

′
Thus Z2𝑔  Z2𝑔 and therefore 𝑔 = 𝑔′ . 

Remark 2.84. This proves the uniqueness part of the classification for surfaces, see Example 1.5.

2.6. Higher homotopy groups

We now generalize the definition of 𝜋1 (𝑋; 𝑥0 ).

Definition 2.85. Let 𝑊 𝑛 = [0, 1] × . . . × [0, 1] be the 𝑛-cube. Let 𝑋 be a topological space
| {z }
𝑛 times
and 𝑥0 ∈ 𝑋. Then

𝜋 𝑛 (𝑋; 𝑥0 ) := {[𝜎] 𝜕𝑊 𝑛 | 𝜎 ∈ 𝐶 (𝑊 𝑛 , 𝑋), 𝜎 (𝜕𝑊 𝑛 ) = {𝑥0 }}

is called the 𝑛-th homotopy group of 𝑋 with base point 𝑥0 . Here [𝜎] 𝜕𝑊 𝑛 denotes the homotopy
class of 𝜎 relative to 𝜕𝑊 𝑛 .

The group structure on 𝜋 𝑛 (𝑋; 𝑥0 ) is obtained as follows: For 𝜎, 𝜏 ∈ 𝐶 (𝑊 𝑛 , 𝑋) with 𝜎 (𝜕𝑊 𝑛 ) =

𝜏(𝜕𝑊 𝑛 ) = {𝑥0 } define 𝜎 ★ 𝜏 by
(
𝜎 (2𝑡 1 , 𝑡2 , . . . , 𝑡 𝑛 ), 0 ≤ 𝑡1 ≤ 1/2,
(𝜎 ★ 𝜏) (𝑡 1 , . . . , 𝑡 𝑛 ) =
𝜏(2𝑡 1 − 1, 𝑡2 , . . . , 𝑡 𝑛 ), 1/2 ≤ 𝑡 1 ≤ 1.

Then 𝜎 ★ 𝜏 ∈ 𝐶 (𝑊 𝑛, 𝑋) with (𝜎 ★ 𝜏) (𝜕𝑊 𝑛 ) = {𝑥0 }.

58
 2.6. Higher homotopy groups

𝑊𝑛 𝑊𝑛 𝑊𝑛

𝜎 𝜎★𝜏

𝑋 𝑡2 , . . . , 𝑡 𝑛
b
𝑡1
𝑥0

Figure 42. Concatenation

Now put [𝜎] 𝜕𝑊 𝑛 · [𝜏] 𝜕𝑊 𝑛 := [𝜎 ★ 𝜏] 𝜕𝑊 𝑛 . The proof that this yields a well-defined group
multiplication on 𝜋 𝑛 (𝑋; 𝑥0 ) for 𝑛 ≥ 2 is literally the same as in the case for 𝑛 = 1. The neutral
element is represented by the constant map 𝜀 𝑛𝑥0 : 𝑊 𝑛 → 𝑋, 𝜀 𝑛𝑥0 (𝑡1 , . . . , 𝑡 𝑛 ) = 𝑥0 , and [𝜎] 𝜕𝑊
−1 is
𝑛
represented by 𝜎 (1 − 𝑡 1 , 𝑡 2 , . . . , 𝑡 𝑛 ).

Unlike for the case 𝑛 = 1 the higher homotopy groups are abelian:

Proposition 2.86. Let 𝑋 be a topological space and 𝑥0 ∈ 𝑋. Then for 𝑛 ≥ 2 the group
𝜋 𝑛 (𝑋; 𝑥0 ) is abelian.

Proof. For 𝜎, 𝜏 ∈ 𝐶 (𝑊 𝑛, 𝑋) with 𝜎 (𝜕𝑊 𝑛 ) = 𝜏(𝜕𝑊 𝑛 ) = {𝑥0 } we need to show that 𝜎 ★ 𝜏 and
𝜏 ★ 𝜎 are homotopic relative to 𝜕𝑊 𝑛 . The homotopy is obtained by precomposing with the
homotopy of the 𝑛-cube indicated in the following picture (which illustrates the case 𝑛 = 2):

59
2. Homotopy Theory

𝜎 𝜏 𝜎 𝜎 𝜎
deform deform deform
−→ −→ −→
𝜏 𝜏 𝜏

𝜏 𝜎
deform
−→

Figure 43. Commutativity of higher homotopy groups

Remark 2.87. This proof also shows that replacing 𝑡 1 by any other variable in the definition of
𝜎 ★ 𝜏 gives the same group multiplication on 𝜋 𝑛 (𝑋; 𝑥0 ).

For 𝑓 ∈ 𝐶 (𝑋, 𝑌 ) with 𝑓 (𝑥0 ) = 𝑦 0 we get a group homomorphism

𝑓# : 𝜋 𝑛 (𝑋; 𝑥0 ) → 𝜋 𝑛 (𝑌 ; 𝑦 0 )

defined by [𝜎] 𝜕𝑊 𝑛 ↦→ [ 𝑓 ◦ 𝜎] 𝜕𝑊 𝑛 .

Remark 2.88. Lemma 2.18, 2.19, Corollary 2.20, Propositions 2.21, 2.22, Theorem 2.23 and
Corollary 2.24 also hold for 𝜋 𝑛 (𝑋; 𝑥0 ). In particular, if 𝑓 : 𝑋 → 𝑌 is a homotopy equivalence
then the map 𝑓# : 𝜋 𝑛 (𝑋; 𝑥0 ) → 𝜋 𝑛 (𝑌 ; 𝑓 (𝑥0 )) is an isomorphism. For 𝛾 ∈ Ω(𝑋; 𝑥0 , 𝑥1 ) there is
an isomorphism
Φ𝛾 : 𝜋 𝑛 (𝑋; 𝑥1 ) → 𝜋 𝑛 (𝑋; 𝑥0 )
given by [𝜎] 𝜕𝑊 𝑛 ↦→ [𝜎 ′ ] 𝜕𝑊 𝑛 where

𝛾
( 𝑥0 𝑥1 𝜎
𝜎 (2𝑡), k𝑡 k max ≤ 12 ,
𝜎 ′ (𝑡) = 1
𝛾(2 − 2k𝑡 k max ), 2 ≤ k𝑡 k max ≤ 1.

Figure 44. “Independence” of base point

60
 2.6. Higher homotopy groups

Remark 2.89. By Exercise 1.7 we have the maps

𝜋 𝜑
𝑊𝑛 / 𝑊 𝑛 /𝜕𝑊 𝑛
≈
/ 𝑆𝑛

and we see that 𝜎 ∈ 𝐶 (𝑊 𝑛, 𝑋) with 𝜎 (𝜕𝑊 𝑛 ) = {𝑥0 } corresponds uniquely to 𝑓 𝜎 ∈ 𝐶 (𝑆 𝑛 , 𝑋)

with 𝑓 𝜎 (𝑠0 ) = 𝑥0 , where 𝑠0 = 𝜑(𝜕𝑊 𝑛 ) such that 𝜎 = 𝑓 𝜎 ◦ 𝜑 ◦ 𝜋. Thus there is a canonical
bijection
1:1
𝜋 𝑛 (𝑋; 𝑥0 ) ←→ {[ 𝑓 ] {𝑠0 } | 𝑓 ∈ 𝐶 (𝑆 𝑛 , 𝑋), 𝑓 (𝑠0 ) = 𝑥0 } .

Remark 2.90. The Seifert-van-Kampen theorem for 𝜋 𝑛 works only under very restrictive as-
sumptions. For this reason the computation of 𝜋 𝑛 for explicit examples can be very difficult. We
will be able to compute 𝜋 𝑛 (𝑆 𝑚 ) for 𝑛 ≤ 𝑚 but many of the 𝜋 𝑛 (𝑆 𝑚 ) for 𝑛 > 𝑚 are actually still
unknown.

Remark 2.91. The definition of 𝜋 𝑛 (𝑋; 𝑥0 ) also works for 𝑛 = 0. A map 𝜎 ∈ 𝐶 (𝑊 0, 𝑋)

corresponds to the point 𝜎 (𝑊 0) ∈ 𝑋. Two such maps are homotopic iff the corresponding points
can be joined by a path. Hence

𝜋0 (𝑋; 𝑥0 ) = {path components of 𝑋 }.

But there is no (natural) group structure on 𝜋0 (𝑋; 𝑥0 ). More precisely, 𝜋0 (𝑋; 𝑥0 ) is a pointed set,
i.e., a set with a distinguished point, namely the path component containing 𝑥0 . This corresponds
to the neutral element in 𝜋 𝑛 (𝑋; 𝑥0 ) for 𝑛 ≥ 1.

Definition 2.92. Let 𝑊, 𝐸 and 𝐵 be topological spaces and let 𝑝 ∈ 𝐶 (𝐸, 𝐵). We say that 𝑝
has the homotopy lifting property (HLP for short) for 𝑊 iff for every 𝑓 ∈ 𝐶 (𝑊, 𝐸) and every
ℎ ∈ 𝐶 (𝑊 × [0, 1], 𝐵) with ℎ(𝑤, 0) = 𝑝( 𝑓 (𝑤)) for all 𝑤 ∈ 𝑊 there exists 𝐻 ∈ 𝐶 (𝑊 × [0, 1], 𝐸)
such that
𝐻 (𝑤, 0) = 𝑓 (𝑤) ∀𝑤 ∈ 𝑊 and ℎ= 𝑝◦𝐻.
In other words, there exists an 𝐻 ∈ 𝐶 (𝑊 × [0, 1], 𝐸) such that the diagram

𝑓
𝑤 ∈ 𝑊 /
❴ ✉: 𝐸
𝐻✉ ✉
𝑝
✉
  ✉ ℎ

(𝑤, 0) ∈ 𝑊 × [0, 1] /𝐵

commutes.

61
2. Homotopy Theory

b
𝑒0
Example 2.93. Consider the spaces 𝐸 = {point},
𝐵 = R, 𝑊 = 𝑊 0 and the map given by 𝑝(𝑒) = 0.
No ℎ : 𝑊 × [0, 1] → 𝐵 except the constant path 𝜀 0
can be lifted because it leaves the image of 𝑝. Here
the problem is the lack of surjectivity of 𝑝. | R
ℎ
0
Figure 45. Failure of HLP due to
lack of surjectivity

Example 2.94. Now consider 𝐸 = [0, ∞) × {0} ∪ (−∞, 0] × {1} ⊂ R2 , 𝐵 = R and let 𝑝 be the
projection onto the first factor, 𝑝(𝑡, 𝑠) = 𝑡.
𝑓
b
𝐸

𝑊 = {𝑤 0 }

| 𝐵
ℎ
Figure 46. Surjective but HLP still fails

The map 𝑝 is surjective but still does not have the HLP for 𝑊 = 𝑊 0 . For example, choose
ℎ(𝑤 0 , 𝑡) = 𝑡, 𝑓 (𝑤 0 ) = (0, 1).

Definition 2.95. A map 𝑝 ∈ 𝐶 (𝐸, 𝐵) is called a Serre fibration or weak fibration iff it has the
HLP for all 𝑊 𝑛 , 𝑛 ≥ 0. The space 𝐸 is called the total space and 𝐵 is called the base space of
the fibration. For 𝑏0 ∈ 𝐵 we call 𝑝 −1 (𝑏0 ) the fiber over 𝑏0 .

Example 2.96. For topological spaces 𝐹 and 𝐵 put 𝐸 := 𝐵 × 𝐹 and 𝑝 = 𝑝𝑟 1 , the projection on
the 𝐵-factor. Then 𝑝 has the HLP for all 𝑊. In particular, 𝑝 is a Serre fibration. Namely, let
𝑓 ∈ 𝐶 (𝑊, 𝐵 × 𝐹) and ℎ ∈ 𝐶 (𝑊 × [0, 1], 𝐵) be given such that 𝑝( 𝑓 (𝑤)) = ℎ(𝑤, 0) for all 𝑤 ∈ 𝑊.
Now write 𝑓 (𝑤) = (𝛽(𝑤), 𝜑(𝑤)) with 𝛽 ∈ 𝐶 (𝑊, 𝐵) and 𝜑 ∈ 𝐶 (𝑊, 𝐹). Hence

ℎ(𝑤, 0) = 𝑝( 𝑓 (𝑤)) = 𝛽(𝑤).

Now put 𝐻 (𝑤, 𝑡) := (ℎ(𝑤, 𝑡), 𝜑(𝑤)). Then 𝐻 ∈ 𝐶 (𝑊 × [0, 1], 𝐵 × 𝐹) and

𝑝(𝐻 (𝑤, 𝑡)) = 𝑝𝑟 1 (ℎ(𝑤, 𝑡), 𝜑(𝑤)) = ℎ(𝑤, 𝑡) ,

𝐻 (𝑤, 0) = (ℎ(𝑤, 0), 𝜑(𝑤)) = (𝛽(𝑤), 𝜑(𝑤)) = 𝑓 (𝑤) ,

as required. Hence 𝑝 has the HLP for any 𝑊.

62
 2.6. Higher homotopy groups

𝑊 𝑛 × [0, 1]

[0, 1]

𝑄 11
𝑊𝑛
Figure 47. Subdivision for which the fiber bundle is trivial over each subcube

Definition 2.97. A map 𝑝 ∈ 𝐶 (𝐸, 𝐵) is called fiber bundle with fiber 𝐹 iff for each 𝑏 ∈ 𝐵
there exists an open subset 𝑈 ⊂ 𝐵 with 𝑏 ∈ 𝑈 and a homeomorphism Φ : 𝑝 −1 (𝑈) → 𝑈 × 𝐹
such that the diagram
Φ
𝑈×𝐹 o ≈ 𝑝 −1 (𝑈)
tt
𝑝𝑟1 ttt
t
 t tt 𝑝 | 𝑝 −1 (𝑈)
y t
𝑈
commutes.

Lemma 2.98. Every fiber bundle is a Serre fibration.

Proof. Let 𝑓 ∈ 𝐶 (𝑊 𝑛 , 𝐸) and ℎ ∈ 𝐶 (𝑊 𝑛 × [0, 1], 𝐵) such that ℎ(𝑤, 0) = 𝑝( 𝑓 (𝑤)) for all 𝑤 ∈ 𝑊.
Now subdivide 𝑊 𝑛 × [0, 1] into small subcubes such that ℎ maps each subcube entirely into an
open subset 𝑈 as in the definition of the fiber bundle, see Figure 47.
By Example 2.96, products have the HLP, hence we can extend the map 𝑓 to a continuous map
𝐻11 : (𝑊 𝑛 × {0}) ∪ 𝑄 11 → 𝐸 such that 𝑝 ◦ 𝐻11 = ℎ.
Next we want to extend the lift to 𝑄 12 , see Figure 48. Now there seems to be a problem because
the required lift 𝐻12 need not only coincide with 𝑓 along the edge 𝑄 12 ∩ (𝑊 𝑛 × {0}) but also
with 𝐻11 along 𝑄 11 ∩ 𝑄 12 .
But there are homeomorphisms of a cube onto itself mapping two edges onto one as indicated in
Figure 49.

63
2. Homotopy Theory

𝑊 𝑛 × [0, 1]

[0, 1]

𝑄 11 𝑄 12
𝑊𝑛
Figure 48. Lift over second cube

Figure 49. Homeomorphism mapping two faces to one

64
 2.6. Higher homotopy groups

Thus we can apply the HLP and extend the lift to (𝑊 𝑛 × {0}) ∪ 𝑄 11 ∪ 𝑄 12 . Iteration of this
procedure proves the assertion. 

Example 2.99. Let 𝐺 be a Lie group, e.g. a closed subgroup of GL(𝑛; 𝐾) with 𝐾 = R or 𝐾 = C.
Let 𝐻 ⊂ 𝐺 be a closed subgroup. We equip the space 𝐵 = 𝐺/𝐻 = {𝑔 · 𝐻 | 𝑔 ∈ 𝐺} with the
quotient topology. Such a space is called a homogeneous space. Then 𝐺 → 𝐺/𝐻 with 𝑔 ↦→ 𝑔 · 𝐻
is a fiber bundle with fiber 𝐻. The proof of this fact requires some technical work, see [8, p. 120
ff].

Example 2.100. Let 𝐺 = SO(𝑛 + 1) and

  
𝐵 0
𝐻= 𝐵 ∈ SO(𝑛) .
0 1
Then 𝐻 is a closed subgroup of 𝐺 isomorphic to SO(𝑛). We now show that 𝐺/𝐻 ≈ 𝑆 𝑛 . Consider
the map 𝑓 : 𝐺 → 𝑆 𝑛 with 𝐴 ↦→ 𝐴 · 𝑒 𝑛+1 where 𝑒 𝑛+1 the (𝑛 + 1)-st unit vector of the canonical
basis. The map 𝑓 is continuous and surjective. We observe
˜
𝑓 ( 𝐴) = 𝑓 ( 𝐴) ⇐⇒ 𝐴 · 𝑒 𝑛+1 = 𝐴˜ · 𝑒 𝑛+1
𝐴˜ −1 · 𝐴 · 𝑒 𝑛+1 = 𝑒 𝑛+1
 
⇐⇒
−1 ★ 0
⇐⇒ 𝐴˜ · 𝐴 =
★ 1
−1
⇐⇒ 𝐴˜ · 𝐴 ∈ 𝐻
⇐⇒ ˜
𝜋( 𝐴) = 𝜋( 𝐴)
where the map 𝜋 : 𝐺 → 𝐺/𝐻 is the canonical projection. We conclude that the map 𝑓 descends
to a bijective map 𝑓¯ : 𝐺/𝐻 → 𝑆 𝑛 . By the universal property of the quotient topology the map
𝑓¯ : 𝐺/𝐻 → 𝑆 𝑛 is continuous. Since 𝐺 is compact the space 𝐺/𝐻 is also compact. Moreover,
the sphere 𝑆 𝑛 is a Hausdorff space, hence the map 𝑓¯ is a homeomorphism. Thus we obtain a
fiber bundle SO(𝑛 + 1) → 𝑆 𝑛 with fiber SO(𝑛).

Let 𝑝 : 𝐸 → 𝐵 be any Serre fibration and fix 𝑒0 ∈ 𝐸. Put 𝑏0 := 𝑝(𝑒0 ) ∈ 𝐵 and let 𝐹 := 𝑝 −1 (𝑏0 ).
Then 𝑒0 ∈ 𝐹. Let 𝜄 : 𝐹 → 𝐸 be the inclusion map. We obtain the following two homomorphisms:
𝜄# : 𝜋 𝑛 (𝐹; 𝑒0 ) → 𝜋 𝑛 (𝐸; 𝑒0 ),
𝑝 # : 𝜋 𝑛 (𝐸; 𝑒0 ) → 𝜋 𝑛 (𝐵; 𝑏0 ).
Now we construct a map 𝜕 : 𝜋 𝑛 (𝐵; 𝑏0 ) → 𝜋 𝑛−1 (𝐹; 𝑒0 ). We define
Box0 := {1} and Box 𝑘 := (𝑊 𝑘 × {1}) ∪ (𝜕𝑊 𝑘 × [0, 1]) for 𝑘 ≥ 1.
Then we have
𝜕𝑊 𝑛 = (𝑊 𝑛−1 × {0}) ∪ Box𝑛−1 ,
(𝑊 𝑛−1 × {0}) ∩ Box𝑛−1 = 𝜕𝑊 𝑛−1 × {0} .

65
2. Homotopy Theory

b
𝑥

b
𝜂 𝑘 (𝑥)

b
( 21 , . . . , 12 , −1)

Figure 50. Mapping the box to the (bottom) cube

Consider the homeomorphism 𝜂 𝑘 : Box𝑘 → 𝑊 𝑘 obtained by the central projection from

( 21 , . . . , 21 , −1). This homeomorphism maps the faces out of which Box𝑘 is built onto the
regions depicted in Figure 50.
In order to define 𝜕 : 𝜋 𝑛 (𝐵; 𝑏0 ) → 𝜋 𝑛−1 (𝐹; 𝑒0 ) let 𝜎 ∈ 𝐶 (𝑊 𝑛, 𝐵) with 𝜎 (𝜕𝑊 𝑛 ) = {𝑏0 }.
Since (𝑡 1 , . . . , 𝑡 𝑛−1 ) ↦→ 𝜎 (𝑡1, . . . , 𝑡 𝑛−1 , 0) = 𝑏0 is constant we can lift it to the constant map
(𝑡 1 , . . . , 𝑡 𝑛−1 ) ↦→ 𝑒0 , see Figure 51. Now the HLP of 𝑝 for 𝑊 𝑛−1 yields a continuous map
Σ : 𝑊 𝑛 → 𝐸 with Σ(𝑡1 , . . . , 𝑡 𝑛−1 , 0) = 𝑒0 and 𝑝 ◦ Σ = 𝜎. From 𝜎 (𝜕𝑊 𝑛 ) = {𝑏0 } we
have Σ(𝜕𝑊 𝑛 ) ⊂ 𝐹. We put 𝜎 ˜ := Σ ◦ 𝜂 𝑛−1 −1 : 𝑊 𝑛−1 → 𝐹. We want to define the map

𝜕 ([𝜎] 𝜕𝑊 𝑛 ) := [ 𝜎] ˜ 𝜕𝑊 𝑛−1 . We have to check well-definedness of this map:

a) We have to show that [ 𝜎]

˜ 𝜕𝑊 𝑛−1 does not depend on the particular choice of the lift Σ.
Let Σ′ ∈ 𝐶 (𝑊 𝑛 , 𝐸) be another lift of 𝜎 with Σ′ (𝑡1 , . . . , 𝑡 𝑛−1 , 0) = 𝑒0 . Then Σ −1 • Σ′ is a lift
of 𝜎 −1 • 𝜎. Here • denotes the concatenation with respect to the variable 𝑡 𝑛 , Σ −1 (𝑡1 , . . . , 𝑡 𝑛 ) =
Σ(𝑡1 , . . . , 𝑡 𝑛−1 , 1 − 𝑡 𝑛 ) and similarly for 𝜎 −1 . Since 𝜎 −1 • 𝜎 ≃𝜕𝑊 𝑛 𝜀 𝑛𝑏0 we can find a homotopy
ℎ : 𝑊 𝑛+1 → 𝐵 relative to 𝜕𝑊 𝑛 with

ℎ(𝑡 1 , . . . , 𝑡 𝑛 , 0) = (𝜎 −1 • 𝜎) (𝑡 1, . . . , 𝑡 𝑛 ) and ℎ(𝑡 1 , . . . , 𝑡 𝑛 , 1) = 𝑏0 .

Then ℎ(Box𝑛 ) = {𝑏0 }. We apply the HLP of 𝑝 for 𝑊 𝑛+1 to get a lift 𝐻 ∈ 𝐶 (𝑊 𝑛+1 , 𝐸) of ℎ with

𝐻 (𝑡 1 , . . . , 𝑡 𝑛 , 0) = Σ −1 • Σ′ (𝑡1 , . . . , 𝑡 𝑛 ).

From ℎ(Box𝑛 ) = {𝑏0 } we have 𝐻 (Box𝑛 ) ⊂ 𝐹. Then we get a homotopy in 𝐹 relative to 𝜕𝑊 𝑛−1
−1 to 𝜎
˜ = Σ ◦ 𝜂 𝑛−1
from 𝜎 ˜ ′ = Σ′ ◦ 𝜂 𝑛−1
−1 as shown in Figure 52.

b) We also have to show that 𝜎 ≃𝜕𝑊 𝑛 𝜎 ′ in 𝐵 implies 𝜎 ˜ ′ in 𝐹.

˜ ≃𝜕𝑊 𝑛−1 𝜎
Let ℎ : 𝑊 𝑛 × [0, 1] → 𝐵 be a homotopy in 𝐵 from 𝜎 to 𝜎 ′ relative to 𝜕𝑊 𝑛 . The HLP for 𝑊 𝑛

66
 2.6. Higher homotopy groups

𝐹 𝐸

𝑡𝑛
b 𝑒0

𝜎 𝑝

b
𝑏0
𝑡 1 , . . . , 𝑡 𝑛−1

Figure 51. Lift 𝜎

Σ′

Σ −1
𝑡 𝑛+1 𝑡𝑛

𝑡1 , . . . , 𝑡 𝑛−1

Figure 52. Homotopy between projected lifts

67
2. Homotopy Theory

𝜎′

𝐵
ℎ
b
𝑏0

𝑡 𝑛+1 𝑡𝑛

𝑡 1 , . . . , 𝑡 𝑛−1
𝜎
Figure 53. Homotopy invariance

yields a lift 𝐻 : 𝑊 𝑛+1 → 𝐸 of ℎ with 𝐻 (𝑡 1 , . . . , 𝑡 𝑛−1 , 0, 𝑡 𝑛+1 ) = 𝑒0 , see Figure 53. We thus obtain
a homotopy
−1
𝐻ˆ (𝑡1 , . . . , 𝑡 𝑛−1 , 𝑠) = 𝐻 (𝜂 𝑛−1 (𝑡 1 , . . . , 𝑡 𝑛−1 ), 𝑠)
in 𝐹 from 𝜎 ˜ ′ relative to 𝜕𝑊 𝑛−1 and hence
˜ to 𝜎
˜ ′ ] 𝜕𝑊 𝑛−1 .
˜ 𝜕𝑊 𝑛−1 = [ 𝜎
[ 𝜎]

We have shown that the map 𝜕 : 𝜋 𝑛 (𝐵; 𝑏0 ) → 𝜋 𝑛−1 (𝐹, 𝑒0 ) is well defined.

Lemma 2.101. For 𝑛 ≥ 2 the map 𝜕 : 𝜋 𝑛 (𝐵; 𝑏0 ) → 𝜋 𝑛−1 (𝐹, 𝑒0 ) is a group homomorphism.

Proof. Let 𝜎, 𝜏 ∈ 𝐶 (𝑊 𝑛 , 𝐵) such that 𝜎 (𝜕𝑊 𝑛 ) = 𝜏(𝜕𝑊 𝑛 ) = {𝑏0 }. Choose a lift 𝐻 ∈ 𝐶 (𝑊 𝑛, 𝐸)

of 𝜎 ★ 𝜏 with 𝐻 (𝑡 1 , . . . , 𝑡 𝑛−1 , 0) = 𝑒0 . Restriction yields lifts Σ of 𝜎 and 𝑇 of 𝜏 up to stretching
in the 𝑡1 -direction. Moreover, we have
−1 −1 −1
(Σ ◦ 𝜂 𝑛−1 ) ★ (𝑇 ◦ 𝜂 𝑛−1 ) ≃𝜕𝑊 𝑛−1 𝐻 ◦ 𝜂 𝑛−1 .
The homotopy is given by shrinking the marked region in the 𝑡1 -direction, see Figure 54. We
conclude that
−1 −1
𝜕 ([𝜎] 𝜕𝑊 𝑛 ) · 𝜕 ([𝜏] 𝜕𝑊 𝑛 ) = [Σ ◦ 𝜂 𝑛−1 ] 𝜕𝑊 𝑛−1 · [𝑇 ◦ 𝜂 𝑛−1 ] 𝜕𝑊 𝑛−1
−1 −1
= [(Σ ◦ 𝜂 𝑛−1 ) ★ (𝑇 ◦ 𝜂 𝑛−1 )] 𝜕𝑊 𝑛−1
−1
= [𝐻 ◦ 𝜂 𝑛−1 ] 𝜕𝑊 𝑛−1
= 𝜕 ([𝜎 ★ 𝜏] 𝜕𝑊 𝑛 )
= 𝜕 ([𝜎] 𝜕𝑊 𝑛 · [𝜏] 𝜕𝑊 𝑛 ). 

68
 2.6. Higher homotopy groups

Σ ★𝑇 𝐻

−1 ) ★ (𝑇 ◦ 𝜂 −1 )
(Σ ◦ 𝜂 𝑛−1 −1
𝐻 ◦ 𝜂 𝑛−1
𝑛−1

b b b

Figure 54. Boundary map is a group homomorphism

Hence 𝜕 : 𝜋 𝑛 (𝐵; 𝑏0 ) → 𝜋 𝑛−1 (𝐹, 𝑒0 ) is a homomorphism if 𝑛 ≥ 2. For 𝑛 = 1 this statement

does not make sense because 𝜋0 (𝐹, 𝑒0 ) is not a group. But 𝜕 still maps the neutral element of
𝜋1 (𝐵, 𝑏0 ) to the distinguished element of 𝜋0 (𝐹, 𝑒0 ).

Theorem 2.102 (Long exact homotopy sequence of a Serre fibration). Let 𝑝 : 𝐸 → 𝐵 be

a Serre fibration, 𝑒0 ∈ 𝐸, 𝑏0 = 𝑝(𝑒0 ) ∈ 𝐵 and 𝐹 = 𝑝 −1 (𝑏0 ). Let 𝜄 : 𝐹 → 𝐸 be the inclusion
map. Then the following sequence is exact:

𝜕 𝜄# 𝑝# 𝜕 𝜄#
... / 𝜋 𝑛 (𝐹; 𝑒 𝑜 ) / 𝜋 𝑛 (𝐸; 𝑒 0 ) / 𝜋 𝑛 (𝐵; 𝑏 0 ) / 𝜋 𝑛−1 (𝐹; 𝑒 0 ) / ...

𝜕 𝜄# 𝑝#
... / 𝜋0 (𝐹; 𝑒 0 ) / 𝜋0 (𝐸; 𝑒 0 ) / 𝜋0 (𝐵; 𝑏 0 )

Remark 2.103. Exactness means that the image of the incoming map equals the kernel of the
outgoing map. The question arises what this means on the 𝜋0 -level where we do not have
homomorphisms. The image is defined for an arbitrary map. For the kernel we recall that 𝜋0 is
a set together with a distinguished element which corresponds to the neutral element of a group.
It is therefore natural to define the kernel of a map to be the set of all elements in the domain of
the map which are mapped to the distinguished element. Having clearified this, exactness of the
above sequence also makes sense on the 𝜋0 -level.

69
2. Homotopy Theory

Proof. a) Exactness at 𝜋 𝑛 (𝐸; 𝑒0 ) for 𝑛 ≥ 0:

i) im(𝜄# ) ⊂ ker( 𝑝 # ):
Since 𝑝 ◦ 𝜄 is the constant map, we have for any [𝜎] 𝜕𝑊 𝑛 ∈ 𝜋 𝑛 (𝐹; 𝑒0 ):

𝑝 # (𝜄# ([𝜎] 𝜕𝑊 𝑛 )) = ( 𝑝 ◦ 𝜄)# ([𝜎] 𝜕𝑊 𝑛 ) = [ 𝑝 ◦ 𝜄 ◦ 𝜎] 𝜕𝑊 𝑛 = [𝜀 𝑛𝑏0 ] 𝜕𝑊 𝑛 = 0.

ii) ker( 𝑝 # ) ⊂ im(𝜄# ):

Let [𝜏] 𝜕𝑊 𝑛 ∈ 𝜋 𝑛 (𝐸; 𝑒0 ) with 𝑝 # ([𝜏] 𝜕𝑊 𝑛 ) = [ 𝑝 ◦ 𝜏] 𝜕𝑊 𝑛 = 0. Hence 𝑝 ◦ 𝜏 ≃𝜕𝑊 𝑛 𝜀 𝑛𝑏0 . Let
ℎ : 𝑊 𝑛 × [0, 1] → 𝐵 be a homotopy in 𝐵 relative to 𝜕𝑊 𝑛 from 𝑝 ◦ 𝜏 to 𝜀 𝑛𝑏0 . Lift the map
ℎ to a homotopy 𝐻 : 𝑊 𝑛 × [0, 1] → 𝐸 with initial conditions 𝜏, i.e. 𝐻 (·, 0) = 𝜏. The red
area in the diagram gets mapped to 𝐹 by 𝐻 because it gets mapped to 𝑏0 by ℎ.

[0, 1]

𝑊𝑛
Figure 55. Bottom-to-box homotopy
We obtain a homotopy in 𝐸 relative to 𝜕𝑊 𝑛 from 𝜏 to 𝐻 ◦ 𝜂 𝑛−1 . Hence 𝜏 ≃𝜕𝑊 𝑛 𝐻 ◦ 𝜂 𝑛−1
and we conclude that
[𝜏] 𝐸,𝜕𝑊 𝑛 = [𝐻 ◦ 𝜂 𝑛−1 ] 𝐸,𝜕𝑊 𝑛 = 𝜄# ([𝐻 ◦ 𝜂 𝑛−1 ] 𝐹,𝜕𝑊 𝑛 ) ∈ im 𝜄# .

b) Exactness at 𝜋 𝑛 (𝐹; 𝑒0 ) for 𝑛 ≥ 0:

i) im 𝜕 ⊂ ker 𝜄# :
Let [𝜎] 𝜕𝑊 𝑛+1 ∈ 𝜋 𝑛+1 (𝐵; 𝑏0 ). Then we have 𝜕 ([𝜎] 𝜕𝑊 𝑛+1 ) = [Σ ◦ 𝜂 𝑛−1 ] 𝐹,𝜕𝑊 𝑛 where Σ is a
lift of 𝜎 with initial conditions 𝜀 𝑛𝑒0 . The map Σ yields a homotopy relative to 𝜕𝑊 𝑛 in 𝐸
from 𝜀 𝑛𝑒0 to Σ ◦ 𝜂 𝑛−1 , see Figure 56. Hence

𝜄# (𝜕 ([𝜎] 𝜕𝑊 𝑛+1 )) = 𝜄# ([Σ ◦ 𝜂 𝑛−1 ] 𝐹,𝜕𝑊 𝑛 ) = [Σ ◦ 𝜂 𝑛−1 ] 𝐸,𝜕𝑊 𝑛 = [𝜀 𝑛𝑒0 ] 𝐸,𝜕𝑊 𝑛 = 0.

ii) ker 𝜄# ⊂ im 𝜕:
Let [𝜏] 𝜕𝑊 𝑛 ∈ 𝜋 𝑛 (𝐹; 𝑒0 ) with 0 = 𝜄# ([𝜏] 𝜕𝑊 𝑛 ). Hence 𝜏 ≃𝜕𝑊 𝑛 𝜀 𝑛𝑒0 in 𝐸. Let 𝐻 be a
homotopy in 𝐸 relative to 𝜕𝑊 𝑛 from 𝜀 𝑛𝑒0 to 𝜏, see Figure 57. Then 𝐻 is a lift of ℎ := 𝑝 ◦ 𝐻.
Since 𝐻 maps 𝜕𝑊 𝑛+1 to 𝐹, the map ℎ maps 𝜕𝑊 𝑛+1 to 𝑏0 . Thus ℎ represents an element in
𝜋 𝑛+1 (𝐵; 𝑏0 ). By definition of 𝜕, we have [𝐻 ◦ 𝜂 𝑛−1 ] 𝜕𝑊 𝑛 = 𝜕 ([ℎ] 𝜕𝑊 𝑛+1 ).
In the image of Box𝑛 under 𝜂 𝑛 we let the interior cube grow and thereby obtain a homotopy
in 𝐹 relative to 𝜕𝑊 𝑛 from 𝐻 ◦𝜂 𝑛−1 to 𝜏, see Figure 58. Therefore [𝜏] 𝜕𝑊 𝑛 = [𝐻 ◦𝜂 𝑛−1
−1 ]
𝜕𝑊 𝑛 =
𝜕 [ℎ] 𝜕𝑊 𝑛+1 ∈ im(𝜕).

70
 2.6. Higher homotopy groups

𝜀 𝑛𝑒0
Figure 56. Bottom-to-box homotopy, again

[0, 1]
𝜀 𝑛𝑒0

𝑊𝑛
Figure 57. Homotopy between 𝜏 and constant map

c) Exactness at 𝜋 𝑛 (𝐵; 𝑏0 ) for 𝑛 ≥ 1:

i) im 𝑝 # ⊂ ker 𝜕:
Let [𝜏] 𝜕𝑊 𝑛 ∈ 𝜋 𝑛 (𝐸; 𝑒0 ). Then 𝜏 is a lift of 𝑝 ◦ 𝜏 with initial conditions 𝜀 𝑛−1
𝑒0 . Since 𝜏 maps
the boundary of 𝑊 𝑛 to 𝑒0 we have 𝜏 ◦ 𝜂 𝑛−1 −1 = 𝜀 𝑛−1 . Thus
𝑒0

−1
𝜕 ( 𝑝 # ([𝜏] 𝜕𝑊 𝑛 ) = 𝜕 ([ 𝑝 ◦ 𝜏] 𝜕𝑊 𝑛 )) = [𝜏 ◦ 𝜂 𝑛−1 ] 𝜕𝑊 𝑛−1 = [𝜀 𝑛−1
𝑒0 ] 𝜕𝑊 𝑛−1 = 0.

ii) ker 𝜕 ⊂ im 𝑝 # :
Let [𝜎] 𝜕𝑊 𝑛 ∈ 𝜋 𝑛 (𝐵; 𝑏0 ) with 𝜕 ([𝜎] 𝜕𝑊 𝑛 ) = 0. Let Σ be a lift of 𝜎 with initial condition
𝜀 𝑛−1 −1
𝑒0 . Then Σ ◦ 𝜂 𝑛−1 represents 𝜕 ([𝜎] 𝜕𝑊 𝑛 ) = 0. Hence we have in 𝐹

−1
Σ ◦ 𝜂 𝑛−1 ≃𝜕𝑊 𝑛−1 𝜀 𝑛−1
𝑒0

Now choose a homotopy in 𝐹 relative to 𝜕𝑊 𝑛−1 from Σ ◦ 𝜂 𝑛−1

−1 to 𝜀 𝑛−1 . Use this homotopy
𝑒0
to continuously extend Σ to the larger cube with boundary values 𝑒0 , see Figure 59. Call
this extension 𝜏. We then have

[𝜏] 𝜕𝑊 𝑛 ∈ 𝜋 𝑛 (𝐸; 𝑒0 ) and 𝑝 # ([𝜏] 𝐸,𝜕𝑊 𝑛 ) = [ 𝑝 ◦ 𝜏] 𝜕𝑊 𝑛 .

71
2. Homotopy Theory

𝑒0
𝑒0

𝑒0 𝜏 𝑒0 −→ 𝑒0
deform 𝜏 𝑒0 deform
−→ 𝜏

𝑒0
𝑒0

Figure 58. Shrink “constant region” to boundary

𝑒0

maps to 𝐹

𝑒0 𝐹 𝐹 𝑒0
Σ

𝑒0
Figure 59. Extension of Σ

Now 𝑝 ◦ 𝜏 is homotopic relative to 𝜕𝑊 𝑛 to 𝜎 as shown in Figure 60. Thus [𝜎] 𝜕𝑊 𝑛 =

[ 𝑝 ◦ 𝜏] 𝜕𝑊 𝑛 ∈ im 𝑝 # . 

Definition 2.104. A fiber bundle with discrete fiber is called a covering.

Corollary 2.105. If 𝑝 : 𝐸 → 𝐵 is a covering with 𝑒0 ∈ 𝐹, 𝑏0 = 𝑝(𝑒0 ) ∈ 𝐵, then the map

𝑝 # : 𝜋 𝑛 (𝐸, 𝑒0 ) → 𝜋 𝑛 (𝐵; 𝑏0 )

is an isomorphism for all 𝑛 ≥ 2.

72
 2.6. Higher homotopy groups

𝑏0
deform deform
−→ −→

𝑏0 𝜎 𝑏0 𝜎 𝜎

Figure 60. Homotopy between 𝑝 ◦ 𝜏 and 𝜎

Proof. The assertion follows from the long exact sequence:

𝜄# 𝑝# 𝜕
{0} = 𝜋 𝑛 (𝐹; 𝑒 𝑜 ) / 𝜋 𝑛 (𝐸; 𝑒 0 ) / 𝜋 𝑛 (𝐵; 𝑏 0 ) / 𝜋 𝑛−1 (𝐹; 𝑒 0 ) = {0}. 

Example 2.106. The map Exp : R → 𝑆1 with 𝑡 ↦→ 𝑒2 𝜋𝑖𝑡 is a covering with fiber Z. Hence

𝜋 𝑘 (𝑆1 ; 1)  𝜋 𝑘 (R, 0) = {0}

for all 𝑘 ≥ 2. More generally, Exp : R𝑛 → 𝑇 𝑛 = 𝑆1 × . . . × 𝑆1 with

(𝑡1 , . . . , 𝑡 𝑛 ) ↦→ (𝑒2 𝜋𝑖𝑡1 , . . . , 𝑒2 𝜋𝑖𝑡𝑛 )

is a covering with fiber Z𝑛 . Hence 𝜋 𝑘 (𝑇 𝑛 )  𝜋 𝑘 (R𝑛 ) = {0} for all 𝑘 ≥ 2.

Example 2.107. Consider the real projective space, defined by RP𝑛 := 𝑆 𝑛 /∼, where 𝑥 ∼ 𝑦
⇐⇒ 𝑥 = 𝑦 or 𝑥 = −𝑦. Then the map 𝑝 : 𝑆 𝑛 → RP𝑛 with 𝑥 ↦→ [𝑥] ∼ is a covering with fiber
Z/2Z. Hence 𝜋 𝑘 (RP𝑛 )  𝜋 𝑘 (𝑆 𝑛 ) for all 𝑘 ≥ 2. For 𝑛 ≥ 2 we investigate the sequence:

{0} = 𝜋1 (𝑆 𝑛 ) / 𝜋1 (RP𝑛 ) / 𝜋0 (Z/2Z) / 𝜋0 (𝑆 𝑛 ) = {0}

We deduce that 𝜋1 (RP𝑛 )  𝜋0 (Z/2Z)  Z/2Z as sets. But then 𝜋1 (RP𝑛 )  Z/2Z also as groups
because there is only one group of order 2 (up to isomorphism).

Example 2.108. We know that SO(𝑛 + 1)/SO(𝑛) ≈ 𝑆 𝑛 from Example 2.99. In the case of 𝑛 = 2
this means that SO(3)/SO(2) ≈ 𝑆2 . We also know that SO(2) ≈ 𝑆1 . For 𝑘 ≥ 3 we consider the
long exact sequence

··· / 𝜋 𝑘 (SO(2)) / 𝜋 𝑘 (SO(3)) / 𝜋 𝑘 (𝑆 2 ) / 𝜋 𝑘−1 (SO(2))

We know that 𝜋 𝑘 (𝑆1 ) = {0} and 𝜋 𝑘−1 (𝑆1 ) = {0}. Hence we find

𝜋 𝑘 (SO(3))  𝜋 𝑘 (𝑆2 ) for all 𝑘 ≥ 3.

73
2. Homotopy Theory

The map 𝑓 : 𝑆3 → SO(3) given by

𝑥 2 + 𝑦 2 − 𝑢2 − 𝑣 2
© ª
2(𝑦𝑢 − 𝑥𝑣) 2(𝑦𝑣 − 𝑥𝑢)
𝑓 (𝑥, 𝑦, 𝑢, 𝑣) = ­ 2(𝑦𝑢 + 𝑥𝑣) 2 2
𝑥 −𝑦 +𝑢 −𝑣2 2 2(𝑢𝑣 − 𝑥𝑦) ®
« 2(𝑦𝑣 − 𝑥𝑢) 2(𝑢𝑣 + 𝑥𝑦) 𝑥 2 − 𝑦 2 − 𝑢2 + 𝑣 2 ¬

satisfies 𝑓 (−𝑥, −𝑦, −𝑧, −𝑣) = 𝑓 (𝑥, 𝑦, 𝑧, 𝑣) and therefore induces a continuous map
𝑓¯ : RP3 → SO(3). This map is bijective and thus a homeomorphism. Hence we have
SO(3) ≈ RP3 . It follows that

𝜋 𝑘 (𝑆2 )  𝜋 𝑘 (SO(3))  𝜋 𝑘 (RP3 )  𝜋 𝑘 (𝑆3 )

for all 𝑘 ≥ 3. Later we will see (Example 3.121 on page 151) that 𝜋3 (𝑆3 )  Z and consequently
𝜋3 (𝑆2 )  Z.

Example 2.109. By the same proof as for SO(𝑛 + 1)/SO(𝑛) ≈ 𝑆 𝑛 we get that

SU(𝑛 + 1)/SU(𝑛) ≈ 𝑆2𝑛+1 .

Therefore there is a fiber bundle SU(𝑛 + 1) → 𝑆2𝑛+1 with fiber SU(𝑛). For 𝑛 ≥ 1 we consider
the following part of the long exact homotopy sequence:
𝜄# 𝑝#
𝜋0 (SU(𝑛)) / 𝜋0 (SU(𝑛 + 1)) / 𝜋0 (𝑆 2𝑛+1 ) = {0}.

Hence the map 𝜄# : 𝜋0 (SU(𝑛)) → 𝜋0 (SU(𝑛 + 1)) is onto. Since SU(1) = {1} we have
𝜋0 (SU(1)) = {0} and thus 𝜋0 (SU(𝑛)) = {0} by induction on 𝑛. Thus SU(𝑛) is path-connected
for all 𝑛 ≥ 1.
Now let us analyze 𝜋1 (SU(𝑛)). Consider
𝜄# 𝑝#
𝜋1 (SU(𝑛)) / 𝜋1 (SU(𝑛 + 1)) / 𝜋1 (𝑆 2𝑛+1 ) = {0}.

Again, we conclude that the map 𝜄# : 𝜋1 (SU(𝑛)) → 𝜋1 (SU(𝑛+1)) is onto. By the same induction
as before we find 𝜋1 (SU(𝑛)) = {0} for all 𝑛 ≥ 1. Thus SU(𝑛) is simply connected for all 𝑛 ≥ 1.

2.7. Exercises
2.1. Let 𝑋 be a set and let 𝑥0 ∈ 𝑋. Determine 𝜋1 (𝑋; 𝑥0 ) where

a) 𝑋 carries the discrete topology;

b) 𝑋 carries the coarse topology.

2.2. Let 𝑋 be a topological space and let 𝜔 : 𝑆1 → 𝑋 be continuous. Show that the following
are equivalent:

(i) 𝜔 is homotopic to a constant map.

74
 2.7. Exercises

(ii) 𝜔 has a continuous extension 𝐷 2 → 𝑋.

2.3. Let 𝑋 be a topological space and let 𝑓 , 𝑔 : 𝑋 → 𝑆 𝑛 be continuous. Assume 𝑓 (𝑥) ≠ −𝑔(𝑥)
for all 𝑥 ∈ 𝑋. Show that 𝑓 and 𝑔 are homotopic.

2.4. Let 𝑋 and 𝑌 be topological spaces. Show that 𝑋 × 𝑌 is contractible if and only if 𝑋 and 𝑌
are contractible.

2.5. Let 𝑋1 , 𝑋2 be topological spaces, 𝑥𝑖 ∈ 𝑋𝑖 . Put 𝑋 := 𝑋1 × 𝑋2 and 𝑥 := (𝑥1 , 𝑥2 ) ∈ 𝑋. Let

𝑝 𝑖 : 𝑋 → 𝑋𝑖 be the canonical projections. Show that

( 𝑝 1# , 𝑝 2# ) : 𝜋1 (𝑋; 𝑥) → 𝜋1 (𝑋; 𝑥1 ) × 𝜋1 (𝑋2 ; 𝑥2 )

is a group isomorphism.

2.6. Let 𝑋 = [0, 1] × [0, 1] ⊂ R2 and let 𝐴 ⊂ 𝑋 be the comb space. Show that there is no
retraction 𝑋 → 𝐴.

2.7. For 𝑓 ∈ 𝐶 (𝑆1 , R2 ) and 𝑝 ∈ R2 \ 𝑓 (𝑆1 ) consider 𝑓 𝑝 ∈ 𝐶 (𝑆1 , 𝑆1 ) given by

𝑓 (𝑧) − 𝑝
𝑓 𝑝 (𝑧) = .
| 𝑓 (𝑧) − 𝑝|

Then
𝑈 ( 𝑓 , 𝑝) := deg( 𝑓 𝑝 )
is called the winding number of 𝑓 around 𝑝.

a) Show that for 𝑝, 𝑞 ∈ R2 \ 𝑓 (𝑆1 ) which can be joined by a continuous path in R2 \ 𝑓 (𝑆1 ) we
have 𝑈 ( 𝑓 , 𝑝) = 𝑈 ( 𝑓 , 𝑞).

b) Compute 𝑈 ( 𝑓𝑛 , 𝑝) for all 𝑝 ∈ R2 \ 𝑓 (𝑆1 ) and all 𝑛 ∈ Z where 𝑓𝑛 (𝑧) = 𝑧 𝑛 .

2.8. Show that the system of equations

cos 1 + 𝑥 2 𝑦 3 + sin(𝑥𝑦 2 ) − 𝑥 2 = 0,
1
𝑦+ = 0,
cosh(𝑥 + 𝑦 + 10)

has a solution (𝑥, 𝑦) ∈ R2 .

2.9. a) Compute 𝜋1 (𝐷 2 \ {𝑥0 }; 𝑥1 ) for all 𝑥0 ≠ 𝑥1 ∈ 𝐷 2 .

b) Show that each homeomorphism 𝑓 : 𝐷 2 → 𝐷 2 maps the boundary onto itself, 𝑓 (𝜕𝐷 2 ) =
𝜕𝐷 2 .

75
2. Homotopy Theory

2.10. On [0, 1] × [−1, 1] consider the equivalence relation ∼ given by (𝑡, 𝑠) ∼ (𝑡 ′ , 𝑠′ ) iff

(𝑡, 𝑠) = (𝑡 ′ , 𝑠′ ) or |𝑡 − 𝑡 ′ | = 1 and 𝑠′ = −𝑠. The quotient space 𝑀 := [0, 1] × [−1, 1]/∼ is called
the Möbius strip. The image 𝑆 of [0, 1] × {0} is called the chord of 𝑀, that of [0, 1] × {−1, 1}
is the boundary 𝜕 𝑀 of 𝑀.

a) Show that 𝜕 𝑀 is homeomorphic to 𝑆1 .

b) Show that 𝑆 is a strong deformation retract of 𝑀.

c) Determine 𝜋1 (𝑀; 𝑥0 ) for some 𝑥0 ∈ 𝜕 𝑀 and the subgroup 𝜄# (𝜋1 (𝜕 𝑀; 𝑥0 )) ⊂ 𝜋1 (𝑀; 𝑥0 )

where 𝜄 : 𝜕 𝑀 ↩→ 𝑀 is the inclusion map.

d) Show that 𝜕 𝑀 is not a retract of 𝑀.

2.11. Let 𝑋 = {(𝑡, 𝑛𝑡 ) | 0 ≤ 𝑡 ≤ 1, 𝑛 ∈ N} ∪ {(𝑠, 0) | 21 ≤ 𝑠 ≤ 1} ⊂ R2 equipped with the

induced topology. Show that 𝑋 is connected but not path-connected.

2.12. Let 𝐺 1 and 𝐺 2 be groups. Show that 𝐺 1 ∗ 𝐺 2  𝐺 2 ∗ 𝐺 1

a) using the construction of the free product;

b) using the universal property.

2.13. Let 𝐺 1 and 𝐺 2 be groups. Show:

a) If 𝑔 ∈ 𝐺 1 ∗ 𝐺 2 has finite order then 𝑔 is conjugate to an element in 𝑖1 (𝐺 1 ) or in 𝑖2 (𝐺 2 ).

b) If 𝐺 1 and 𝐺 2 are nontrivial then 𝐺 1 ∗ 𝐺 2 contains elements of infinite order.

2.14. Provide a presentation for the following group 𝐺:

a) 𝐺 = Z/2Z × Z/3Z;

b) Z ∗ Z;

c) 𝐺 = (Z ∗ Z) × (Z ∗ Z);

d) 𝐺 = Z ∗ Z/[Z ∗ Z, Z ∗ Z].

2.15. Decide whether or not the groups 𝐺 and 𝐻 are isomorphic where

a) 𝐺 = h𝑎, 𝑏 | 𝑎2 𝑏2 i and 𝐻 = h𝑥, 𝑦, 𝑧 | 𝑥𝑦 2 , 𝑥𝑧2 i;

b) 𝐺 = h𝑎, 𝑏 | 𝑎𝑏i and 𝐻 = h𝑥, 𝑦 | 𝑥 2 i.

2.16. Let 𝑋 be a path-connected topological space. Show that the suspension Σ𝑋 (see Exer-
cise 1.11) is also path-connected.

76
 2.7. Exercises

2.17. Show that the map C → C, 𝑧 ↦→ 𝑧2 , has the homotopy lifting property for 𝑊 0 but not for
𝑊 1.

2.18. Let 𝑋 = R3 \ (𝑆1 × {0}) = R3 \ {(𝑥, 𝑦, 𝑧) ∈ R3 | 𝑥 2 + 𝑦 2 = 1, 𝑧 = 0}. Show:

𝜋1 (𝑋, {0})  Z

and draw a generator of the fundamental group.

2.19. On 𝑆 𝑛 consider the equivalence relation ∼ given by 𝑥 ∼ 𝑦 ⇔ 𝑥 = 𝑦 or 𝑥 = −𝑦. The

quotient RP𝑛 := 𝑆 𝑛 /∼ is called the 𝑛-dimensional real projective space.
Show inductively using the Seifert-van Kampen theorem that for 𝑛 ≥ 2

𝜋1 (RP𝑛 )  Z/2Z.

Hint: Exercise 2.10 may be helpful for the induction base 𝑛 = 2.

2.20. Decide by proof or counter-example whether or not the following assertion holds true:
The Seifert-van Kampen theorem also holds if one only assumes that 𝑈 ∩ 𝑉 is connected rather
than path-connected.

2.21. Let 𝐸 = 𝐵 × 𝐹 and 𝑝 = pr1 : 𝐸 → 𝐵 be the product fibration. Show that the boundary
map 𝜕 : 𝜋 𝑛+1 (𝐵; 𝑏0 ) → 𝜋 𝑛 (𝐹; 𝑒0 ) is trivial in this case.

2.22. Show that the inclusion SU(𝑛) ↩→ U(𝑛) induces an isomorphism

𝜋 𝑘 (SU(𝑛))  𝜋 𝑘 (U(𝑛))

for all 𝑘 ≥ 2.

2.23. The complex projective space is defined as C𝑃 𝑛 := 𝑆2𝑛+1 /∼ where 𝑧 ∼ 𝑤 iff there exists
𝑢 ∈ 𝑆1 ⊂ C such that 𝑧 = 𝑢 · 𝑤. Here we have regarded 𝑆2𝑛+1 = {𝑧 ∈ C𝑛+1 | |𝑧| = 1} as a subset
of C𝑛+1 . The map 𝑝 : 𝑆2𝑛+1 → C𝑃 𝑛 , 𝑧 ↦→ [𝑧] ∼ , is called the Hopf fibration.
Compute
𝜋 𝑘 (C𝑃 𝑛 )
for all 𝑘 ≤ 2𝑛.
Hint: You can use 𝜋 𝑘 (𝑆 𝑚 ) = {0} for all 1 ≤ 𝑘 < 𝑚 and 𝑚 ≥ 2.

2.24. Let 𝑝 : 𝑋 → 𝑌 and 𝑞 : 𝑌 → 𝑍 be Serre fibrations. Prove that 𝑞 ◦ 𝑝 : 𝑋 → 𝑍 is a Serre

fibration as well.

2.25. Let 𝑝 : 𝐸 → 𝐵 be a Serre fibration, 𝑒0 ∈ 𝐸, 𝑏0 = 𝑝(𝑒0 ) and 𝐹 = 𝑝 −1 (𝑏0 ). Show

a) If 𝐹 is simply connected then 𝜋1 (𝐸; 𝑒0 )  𝜋1 (𝐵; 𝑏0 ).

77
2. Homotopy Theory

b) If 𝐸 is contractible then 𝜋 𝑛+1 (𝐵; 𝑏0 )  𝜋 𝑛 (𝐹; 𝑒0 ) for all 𝑛 ≥ 1.

𝑖 𝑝
2.26. Let 0 → 𝐴 → − 𝐴′ − → 𝐴′′ → 0 be an exact sequence of abelian groups. Show that the
following three conditions are equivalent:

(i) There exists an isomorphism Ψ : 𝐴′ → 𝐴 × 𝐴′′ such that the following diagram commutes:

𝑖 𝑝
0 /𝐴 / 𝐴′ / 𝐴′′ /0

= Ψ =
  
0 /𝐴 / 𝐴 × 𝐴′′ / 𝐴′′ /0

where the arrows 𝐴 → 𝐴 × 𝐴′′ and 𝐴 × 𝐴′′ → 𝐴′′ are given by the canonical maps
𝑎 ↦→ (𝑎, 0) and (𝑎, 𝑎′′ ) ↦→ 𝑎′′ , respectively.

(ii) There exists a homomorphism 𝑝 ′ : 𝐴′′ → 𝐴′ such that 𝑝 ◦ 𝑝 ′ = id 𝐴′′ .

(iii) There exists a homomorphism 𝑟 : 𝐴′ → 𝐴 such that 𝑟 ◦ 𝑖 = id 𝐴.

If these conditions hold then we say that the exact sequence is split.

𝑖 𝑝
2.27. a) Show that every exact sequence 0 → 𝐴 → − 𝐴′ −→ 𝐴′′ → 0 of vector spaces (i.e. all
spaces are vector spaces over some fixed field and all homomorphisms are linear maps) is split.
2
b) Show that the exact sequence 0 → Z →
− Z → Z/2Z → 0 is not split.

78
3. Homology Theory
Homotopy groups are in general hard to compute. For instance, for spheres not all homotopy
groups are known even now. In this chapter we introduce rougher invariants which are much
easier to determine, the homology groups.

3.1. Singular homology

We will use the notation

𝑒0 = (0, . . . , 0) ∈ R𝑛
𝑒1 = (1, . . . , 0) ∈ R𝑛
..
.
𝑒 𝑛 = (0, . . . , 1) ∈ R𝑛

Now we define
( )
Õ
𝑛 Õ
𝑛
Δ𝑛 := convex hull of 𝑒0 , . . . , 𝑒 𝑛 = 𝑡 𝑖 𝑒𝑖 𝑡 𝑖 ≥ 0, 𝑡𝑖 ≤ 1 .
𝑖=0 𝑖=0

Then Δ𝑛 is called 𝑛-dimensional standard simplex.

Example 3.1. For 𝑛 = 0, 1, 2, and 3 the standard simplices are familiar, compare Figure 61:

𝑛=0: Δ0 = {𝑒0 } = point

𝑛=1: Δ1 = [0, 1] = line segment
𝑛=2: Δ2 = triangle
𝑛=3: Δ3 = tetrahedron

Definition 3.2. Let 𝑋 be a topological space. A singular 𝑛-simplex in 𝑋 is a continuous map

Δ𝑛 → 𝑋.

Now we fix a commutative ring 𝑅 with 1. The most important examples will be

𝑅 = R, Q, C, Z, Z/𝑛Z.

79
3. Homology Theory

𝑒2 b b

Δ3

b b b b

𝑒0 𝑒1

Figure 61. 2 and 3-dimensional standard simplices

Definition 3.3. We define the set of singular 𝑛-chains 𝑆 𝑛 (𝑋; 𝑅) as the free 𝑅-module generated
by 𝐶 (Δ𝑛 , 𝑋).

Í𝑚
Then 𝑆𝑛 (𝑋; 𝑅) is an 𝑅-module. Elements of 𝑆𝑛 (𝑋; 𝑅) are formal linear combinations 𝑖=1 𝛼𝑖 𝜎𝑖
where 𝛼𝑖 ∈ 𝑅 and 𝜎𝑖 ∈ 𝐶 (Δ𝑛 , 𝑋). See Appendix A.1 for more on free modules generated by
sets.
For 𝑛 ≥ 0 we consider the affine linear map 𝐹𝑛+1 𝑖 : Δ𝑛 → Δ𝑛+1 given by
(
𝑖 𝑒 𝑗, 𝑗 <𝑖
𝐹𝑛+1 (𝑒 𝑗 ) = (3.1)
𝑒 𝑗+1 , 𝑗 ≥ 𝑖

Note that 𝐹𝑛𝑖 maps Δ𝑛 to the face of Δ𝑛+1 opposite to 𝑒𝑖 .

Example 3.4. Consider the special case 𝐹21 :

𝑋
𝑒2 𝜎
b

𝐹21

b
b

𝑒0 𝑒1 𝑒0 𝑒1

𝜎 (𝑖)
Figure 62. Face maps

Definition 3.5. If 𝜎 is an (𝑛 + 1)-dimensional singular simplex in 𝑋 then 𝜎 (𝑖) := 𝜎 ◦ 𝐹𝑛𝑖 is

called the 𝑖-th face of 𝜎.

80
 3.1. Singular homology

The boundary of a singular 𝑛-simplex in 𝑋 is given by:

Õ
𝑛
𝜕𝜎 := (−1) 𝑖 𝜎 (𝑖) .
𝑖=0

We see that the boundary of a singular 𝑛-simplex is a singular (𝑛 − 1)-chain. We extend 𝜕 to

chains by linearity. The boundary of a singular 𝑛-chain in 𝑋 is thus given by
Õ
𝑚  Õ
𝑚
𝜕 𝛼 𝑗 𝜎𝑗 = 𝛼 𝑗 𝜕𝜎 𝑗 .
𝑗=0 𝑗=0

Hence we obtain a linear map 𝜕 : 𝑆𝑛 (𝑋; 𝑅) → 𝑆𝑛−1 (𝑋; 𝑅) and we set 𝜕 (0-chain) := 0.

Lemma 3.6. 𝜕 ◦ 𝜕 = 0.

Proof. It suffices to prove 𝜕𝜕𝜎 = 0 for all 𝑛-simplices 𝜎. For 𝑗 < 𝑖 we have
𝑗 𝑗
𝐹𝑛𝑖 ◦ 𝐹𝑛−1 = 𝐹𝑛 ◦ 𝐹𝑛−1
𝑖−1
. (3.2)

We compute
!
Õ
𝑛
𝜕𝜕𝜎 = 𝜕 (−1) 𝑖 𝜎 ◦ 𝐹𝑛𝑖
𝑖=0
Õ
𝑛


= (−1) 𝑖 𝜕 𝜎 ◦ 𝐹𝑛𝑖
𝑖=0
Õ𝑛 Õ
𝑛−1
𝑖 𝑗
= (−1) (−1) 𝑗 𝜎 ◦ 𝐹𝑛𝑖 ◦ 𝐹𝑛−1
Õ Õ
𝑖=0 𝑗=0
𝑗 𝑗
= (−1) 𝑖+ 𝑗 𝜎 ◦ 𝐹𝑛𝑖 ◦ 𝐹𝑛−1 + (−1) 𝑖+ 𝑗 𝜎 ◦ 𝐹𝑛𝑖 ◦ 𝐹𝑛−1
Õ Õ
0≤ 𝑗<𝑖≤𝑛 0≤𝑖≤ 𝑗 ≤𝑛−1
𝑗 ′ ′ −1 𝑗′ ′
= (−1) 𝑖+ 𝑗 𝜎 ◦ 𝐹𝑛 𝑖−1
◦ 𝐹𝑛−1 + (−1) 𝑗 +𝑖 𝑖 −1
𝜎 ◦ 𝐹𝑛 ◦ 𝐹𝑛−1
0≤ 𝑗<𝑖≤𝑛 0≤ 𝑗 ′ <𝑖′ ≤𝑛
= 0.

In the last step we used (3.2) for the first sum and changed the summation indices from 𝑖 → 𝑗 ′
and 𝑗 → 𝑖 ′ − 1 in the second sum. 

Now we define the set of singular 𝑛-cycles by

𝑍 𝑛 (𝑋; 𝑅) := ker(𝜕 : 𝑆𝑛 (𝑋; 𝑅) → 𝑆𝑛−1 (𝑋; 𝑅))

={𝑐 ∈ 𝑆𝑛 (𝑋; 𝑅) | 𝜕𝑐 = 0}

81
3. Homology Theory

and the set of singular 𝑛-boundaries by

𝐵𝑛 (𝑋; 𝑅) := im(𝜕 : 𝑆𝑛+1 (𝑋; 𝑅) → 𝑆𝑛 (𝑋; 𝑅))

={𝑐 ∈ 𝑆𝑛 (𝑋; 𝑅) | ∃ 𝑏 ∈ 𝑆𝑛+1 (𝑋; 𝑅) such that 𝑐 = 𝜕𝑏} .

Lemma 3.6 says 𝐵𝑛 (𝑋; 𝑅) ⊂ 𝑍 𝑛 (𝑋; 𝑅).

Definition 3.7. The quotient

𝑍 𝑛 (𝑋; 𝑅)
𝐻𝑛 (𝑋; 𝑅) :=
𝐵𝑛 (𝑋; 𝑅)
is called 𝑛-th singular homology of 𝑋 with coefficients in 𝑅.

Remark 3.8. The 𝑛-th homology 𝐻𝑛 (𝑋; 𝑅) is an 𝑅-module.

Example 3.9. Assume that 𝑋 = {point}. Then there is only one singular 𝑛-simplex, namely
the constant map 𝜎𝑛 : Δ𝑛 → 𝑋. In other words, 𝑆𝑛 (𝑋; 𝑅) = 𝑅 · 𝜎𝑛 . Consequently,
𝜎𝑛(𝑖) = 𝜎𝑛 ◦ 𝐹𝑛𝑖 = 𝜎𝑛−1 and



Õ
𝑛 


0, 𝑛 odd,
𝑖

𝜕𝜎𝑛 = (−1) 𝜎𝑛−1 = 𝜎𝑛−1 , 𝑛 even, 𝑛 ≠ 0,

 0,

𝑖=0 𝑛 = 0.

This implies (
𝑅 · 𝜎𝑛 , 𝑛 odd or 𝑛 = 0,
𝑍 𝑛 (𝑋; 𝑅) =
0, 𝑛 even and 𝑛 ≠ 0,
and (
𝑅 · 𝜎𝑛 , 𝑛 odd,
𝐵𝑛 (𝑋; 𝑅) =
0, 𝑛 even.
We conclude that (
𝑅, if 𝑛 = 0,
𝐻𝑛 (𝑋; 𝑅) 
0, otherwise.

As in homotopy theory we not only associate groups to spaces but also homomorphisms to
maps. Let 𝑓 : 𝑋 → 𝑌 be a continuous map. Then we obtain an 𝑅-module homomorphism
𝑆𝑛 ( 𝑓 ) : 𝑆𝑛 (𝑋; 𝑅) → 𝑆𝑛 (𝑌 ; 𝑅) by setting
Õ
𝑚  Õ
𝑚
𝑆𝑛 ( 𝑓 ) 𝛼𝑖 · 𝜎𝑖 := 𝛼𝑖 · ( 𝑓 ◦ 𝜎𝑖 ).
𝑖=1 𝑖=1

82
 3.2. Relative homology

Lemma 3.10. The diagram

𝑆𝑛 ( 𝑓 )
𝑆𝑛 (𝑋; 𝑅) / 𝑆 𝑛 (𝑌 ; 𝑅)

𝜕 𝜕
 𝑆𝑛−1 ( 𝑓 ) 
𝑆𝑛−1 (𝑋; 𝑅) / 𝑆 𝑛−1 (𝑌 ; 𝑅)

commutes for all 𝑛 ≥ 0.

Proof. We compute

𝜕 (𝑆𝑛 ( 𝑓 ) (𝜎)) = 𝜕 ( 𝑓 ◦ 𝜎)
Õ𝑛
= (−1) 𝑖 ( 𝑓 ◦ 𝜎) ◦ 𝐹𝑛𝑖
𝑖=0
Õ
𝑛
= (−1) 𝑖 𝑓 ◦ (𝜎 ◦ 𝐹𝑛𝑖 )
𝑖=0
Õ
𝑛 
= 𝑆𝑛−1 ( 𝑓 ) (−1) 𝑖 𝜎 ◦ 𝐹𝑛𝑖
𝑖=0
= 𝑆𝑛−1 ( 𝑓 ) (𝜕𝜎). 

This lemma implies 𝑆𝑛 ( 𝑓 ) (𝑍 𝑛 (𝑋; 𝑅)) ⊂ 𝑍 𝑛 (𝑌 ; 𝑅) and 𝑆𝑛 ( 𝑓 ) (𝐵𝑛 (𝑋; 𝑅)) ⊂ 𝐵𝑛 (𝑌 ; 𝑅). Hence
we obtain a well-defined 𝑅-module homomorphism 𝐻𝑛 ( 𝑓 ) : 𝐻𝑛 (𝑋; 𝑅) → 𝐻𝑛 (𝑌 ; 𝑅) where
𝐻𝑛 ( 𝑓 ) ([𝑥]) := [𝑆𝑛 ( 𝑓 ) (𝑥)]. Here the square brackets denote the homology classes of the
𝑛-cycles. One sees directly from the definition that 𝐻𝑛 (·) has the functorial properties

(i) 𝐻𝑛 (id𝑋 ) = id 𝐻𝑛 (𝑋;𝑅) ,

(ii) 𝐻𝑛 ( 𝑓 ◦ 𝑔) = 𝐻𝑛 ( 𝑓 ) ◦ 𝐻𝑛 (𝑔).

Exactly as for homotopy groups these functorial properties imply that a homeomorphism 𝑓 :
𝑋 → 𝑌 induces an isomorphism 𝐻𝑛 ( 𝑓 ) : 𝐻𝑛 (𝑋; 𝑅) → 𝐻𝑛 (𝑌 ; 𝑅). Homeomorphic spaces have
isomorphic homology groups.

3.2. Relative homology

For a topological space 𝑋 and 𝐴 ⊂ 𝑋 we call (𝑋, 𝐴) a pair of spaces and set

𝐶 ((𝑋, 𝐴), (𝑌 , 𝐵)) := { 𝑓 ∈ 𝐶 (𝑋,𝑌 ) | 𝑓 ( 𝐴) ⊂ 𝐵} .

We abbreviate 𝑆𝑛 (𝑋) = 𝑆𝑛 (𝑋; 𝑅) if 𝑅 is understood. We observe that 𝑆𝑛 ( 𝐴) ⊂ 𝑆𝑛 (𝑋) and

𝜕 (𝑆𝑛 ( 𝐴)) ⊂ 𝑆𝑛−1 ( 𝐴). Writing 𝑆𝑛 (𝑋, 𝐴) = 𝑆𝑛 (𝑋, 𝐴; 𝑅) := 𝑆𝑛 (𝑋; 𝑅)/𝑆𝑛 ( 𝐴; 𝑅), the map 𝜕

83
3. Homology Theory

induces a well-defined homomorphism 𝜕¯ such that

𝑆𝑛 (𝑋) / 𝑆 𝑛 (𝑋, 𝐴) (3.3)

𝜕 𝜕¯
 
𝑆𝑛−1 (𝑋) / 𝑆 𝑛−1 (𝑋, 𝐴)

commutes. Since 𝜕 ◦ 𝜕 = 0 and since 𝑆𝑛 (𝑋) → 𝑆𝑛 (𝑋, 𝐴) is onto we also have that 𝜕¯ ◦ 𝜕¯ = 0.
Set


𝑍 𝑛 (𝑋, 𝐴) = 𝑍 𝑛 (𝑋, 𝐴; 𝑅) := ker 𝜕¯ : 𝑆𝑛 (𝑋, 𝐴) → 𝑆𝑛−1 (𝑋, 𝐴) ,


𝐵𝑛 (𝑋, 𝐴) = 𝐵𝑛 (𝑋, 𝐴; 𝑅) := im 𝜕¯ : 𝑆𝑛+1 (𝑋, 𝐴) → 𝑆𝑛 (𝑋, 𝐴) .

We define the relative singular homology of (𝑋, 𝐴) by

𝑍 𝑛 (𝑋, 𝐴; 𝑅)
𝐻𝑛 (𝑋, 𝐴) = 𝐻𝑛 (𝑋, 𝐴; 𝑅) := .
𝐵𝑛 (𝑋, 𝐴; 𝑅)
Now consider the preimage of 𝑍 𝑛 (𝑋, 𝐴) under 𝑆𝑛 (𝑋) → 𝑆𝑛 (𝑋, 𝐴) and set

𝑍 𝑛′ (𝑋, 𝐴) := {𝑐 ∈ 𝑆𝑛 (𝑋) | 𝜕𝑐 ∈ 𝑆𝑛−1 ( 𝐴)},

𝐵′𝑛 (𝑋, 𝐴) := {𝑐 ∈ 𝑆𝑛 (𝑋) | ∃𝑏 ∈ 𝑆𝑛+1 (𝑋) such that 𝑐 + 𝜕𝑏 ∈ 𝑆𝑛 ( 𝐴)}.

Since 𝑍 𝑛 (𝑋, 𝐴) = 𝑍 𝑛′ (𝑋, 𝐴)/𝑆𝑛 ( 𝐴) and 𝐵𝑛 (𝑋, 𝐴) = 𝐵′𝑛 (𝑋, 𝐴)/𝑆𝑛 ( 𝐴) we obtain
𝑍 𝑛′ (𝑋, 𝐴)/𝑆𝑛 ( 𝐴) 𝑍 𝑛′ (𝑋, 𝐴)
𝐻𝑛 (𝑋, 𝐴) = = .
𝐵′𝑛 (𝑋, 𝐴)/𝑆𝑛 ( 𝐴) 𝐵′𝑛 (𝑋, 𝐴)

Remark 3.11. For 𝐴 = ∅ we have the special cases

𝑍 𝑛′ (𝑋, ∅) = 𝑍 𝑛 (𝑋),
𝐵′𝑛 (𝑋, ∅) = 𝐵𝑛 (𝑋),
𝐻𝑛 (𝑋, ∅) = 𝐻𝑛 (𝑋).

Example 3.12. Let 𝑋 = 𝑆1 × [0, 1] be the cylinder over 𝑆1 and 𝐴 = 𝑆1 × {0} ⊂ 𝑋, see Figure 63.
To construct an element in the relative homology 𝐻1 (𝑋, 𝐴) take the 1-simplex

𝜎 : Δ1 → 𝑋,
(𝑡𝑒1 + (1 − 𝑡)𝑒0 ) ↦→ (cos (2𝜋𝑡), sin (2𝜋𝑡), 1),

see Figure 64. Since 𝜎 is a closed curve in 𝑋 we find for its boundary

𝜕𝜎 = 𝜎 (𝑒1 ) − 𝜎 (𝑒0 ) = 0.

Therefore 𝜎 ∈ 𝑍1 (𝑋) ⊂ 𝑍1′ (𝑋, 𝐴) and, as we will see later, represents a nontrivial element in
𝐻1 (𝑋). For the 2-simplices 𝜎1 and 𝜎2 defined as indicated in Figure 65.
we find for the 2-chain 𝜎1 + 𝜎2 the boundary 𝜕 (𝜎1 + 𝜎2 ) = 𝜎 + 𝑎 where 𝑎 ∈ 𝑆1 ( 𝐴). Hence,
modulo 𝑎 ∈ 𝑆1 ( 𝐴), we have 𝜎 ∈ 𝐵1 (𝑋, 𝐴) and therefore 0 = [𝜎] ∈ 𝐻1 (𝑋, 𝐴).

84
 3.2. Relative homology

𝐴
Figure 63. Cylinder relative to bottom

|
>
𝜎 𝑋

Δ1
𝐴

Figure 64. The representative 𝜎

85
3. Homology Theory

𝜎
>
<
𝜎1 𝑋

Δ2

𝜎2

𝐴
Figure 65. 𝜎 is null-homologous

Let (𝑋, 𝐴) and (𝑌 , 𝐵) be pairs of spaces and 𝑓 ∈ 𝐶 ((𝑋, 𝐴), (𝑌 , 𝐵)). Then, for each 𝑛 ∈ N0 , we
have the commutative diagram
𝑆𝑛 ( 𝑓 )
𝑆𝑛 (𝑋) / 𝑆 𝑛 (𝑌 )
O O

? 𝑆𝑛 ( 𝑓 | 𝐴 ) ?
𝑆𝑛 ( 𝐴) / 𝑆 𝑛 (𝐵)

Thus we obtain a well-defined homomorphism 𝑆𝑛 ( 𝑓 ) : 𝑆𝑛 (𝑋, 𝐴) → 𝑆𝑛 (𝑌 , 𝐵) such that

𝑆𝑛 ( 𝑓 )
𝑆𝑛 (𝑋) / 𝑆 𝑛 (𝑌 )

 𝑆𝑛 ( 𝑓 ) 
𝑆𝑛 (𝑋, 𝐴) / 𝑆 𝑛 (𝑌 , 𝐵)

commutes. Combining with Lemma 3.10 and diagram (3.3) we get the commutative diagram

𝑆𝑛 ( 𝑓 )

𝑆𝑛 ( 𝑓 ) )
𝑆𝑛 (𝑋, 𝐴) o 𝑆𝑛 (𝑋) / 𝑆 𝑛 (𝑌 ) / 𝑆 𝑛 (𝑌 , 𝐵)

𝜕¯ 𝜕 𝜕 𝜕¯
  𝑆𝑛−1 ( 𝑓 )  
𝑆𝑛−1 (𝑋, 𝐴) o 𝑆𝑛−1 (𝑋) / 𝑆 𝑛−1 (𝑌 ) / 𝑆 𝑛−1 (𝑌 , 𝐵)
5
𝑆𝑛−1 ( 𝑓 )

We have extended Lemma 3.10 to relative homology. In particular, we obtain a well-defined

86
 3.2. Relative homology

homomorphism 𝐻𝑛 ( 𝑓 ) : 𝐻𝑛 (𝑋, 𝐴) → 𝐻𝑛 (𝑌 , 𝐵) such that

𝑆𝑛 ( 𝑓 )
𝑍 𝑛 (𝑋, 𝐴) / 𝑍 𝑛 (𝑌 , 𝐵)

 𝐻𝑛 ( 𝑓 ) 
𝐻𝑛 (𝑋, 𝐴) / 𝐻𝑛 (𝑌 , 𝐵)

commutes.
Let (𝑋, 𝐴) be a pair of spaces. The inclusion map 𝑖 : 𝐴 ↩→ 𝑋 induces a homomorphism
𝐻𝑛 (𝑖) : 𝐻𝑛 ( 𝐴) → 𝐻𝑛 (𝑋). Furthermore we have the inclusion map 𝑗 : (𝑋, ∅) → (𝑋, 𝐴) which
induces the homomorphism 𝐻𝑛 ( 𝑗) : 𝐻𝑛 (𝑋) = 𝐻𝑛 (𝑋, ∅) → 𝐻𝑛 (𝑋, 𝐴).
We define the connecting homomorphism or boundary operator

𝜕 : 𝐻𝑛 (𝑋, 𝐴) → 𝐻𝑛−1 ( 𝐴), (3.4)

[𝑐] ↦→ [𝜕𝑐],

where 𝑐 ∈ 𝑍 𝑛′ (𝑋, 𝐴). Note that 𝜕𝑐 ∈ 𝑍 𝑛−1 ( 𝐴) since 𝜕 2 = 0. The connecting homomorphism is
well defined because replacing 𝑐 by another representative 𝑐 + 𝜕𝑏 + 𝑎 where 𝑏 ∈ 𝑆𝑛+1 (𝑋) and
𝑎 ∈ 𝑆𝑛 ( 𝐴) yields

𝑐 + 𝜕𝑏 + 𝑎 ↦→ 𝜕 (𝑐 + 𝜕𝑏 + 𝑎) = 𝜕𝑐 + 𝜕𝑎.

Since 𝜕𝑎 ∈ 𝐵𝑛−1 ( 𝐴) we get [𝜕𝑐 + 𝜕𝑎] = [𝜕𝑐] ∈ 𝐻𝑛−1 ( 𝐴). Since 𝜕 : 𝑍 𝑛′ (𝑋, 𝐴) → 𝑆𝑛−1 ( 𝐴) is a
homomorphism, the connecting homomorphism is also a homomorphism.

Lemma 3.13. The connecting homomorphism is natural, i.e. the diagram

𝐻𝑛 ( 𝑓 )
𝐻𝑛 (𝑋, 𝐴) / 𝐻𝑛 (𝑌 , 𝐵)

𝜕 𝜕
 𝐻𝑛−1 ( 𝑓 | 𝐴 ) 
𝐻𝑛−1 ( 𝐴) / 𝐻𝑛−1 (𝐵)

commutes for every 𝑓 ∈ 𝐶 ((𝑋, 𝐴), (𝑌 , 𝐵)) and 𝑛 ∈ N.

Proof. We compute
Õ 

 Õ 

𝐻𝑛−1 ( 𝑓 | 𝐴) 𝜕 𝛼𝑖 𝜎𝑖 = 𝐻𝑛−1 ( 𝑓 | 𝐴) 𝜕 𝛼𝑖 𝜎𝑖
Õ Õ 

𝑖 𝑖
𝑗
= 𝐻𝑛−1 ( 𝑓 | 𝐴) 𝛼𝑖 (−1) 𝑗 𝜎 𝑗 ◦ 𝐹𝑛−1
Õ Õ 
𝑖 𝑗
𝑗
= 𝛼𝑖 (−1) 𝑗 ( 𝑓 | 𝐴) ◦ (𝜎𝑖 ◦ 𝐹𝑛−1 )
Õ Õ 
𝑖 𝑗
𝑗
= 𝛼𝑖 (−1) 𝑗 ( 𝑓 ◦ 𝜎𝑖 ) ◦ 𝐹𝑛−1
𝑖 𝑗

87
3. Homology Theory

 Õ 
= 𝜕 𝛼𝑖 ( 𝑓 ◦ 𝜎𝑖 )
Õ 
𝑖

=𝜕 𝛼𝑖 ( 𝑓 ◦ 𝜎𝑖 )
Õ 


𝑖

= 𝜕 𝐻𝑛 ( 𝑓 ) 𝛼𝑖 𝜎𝑖 . 
𝑖

3.3. The Eilenberg-Steenrod axioms and applications

Now we list the most important properties of homology theory known as Eilenberg-Steenrod
axioms. The first axiom is exactly Example 3.9.

Dimension Axiom. 
𝑅, 𝑛 = 0
𝐻𝑛 ({point}; 𝑅) 
0, otherwise
The next axiom deals with homotopy invariance. Two continuous maps 𝑓0 , 𝑓1 : (𝑋, 𝐴) → (𝑌 , 𝐵)
of pairs of spaces are called homotopic (in symbols 𝑓0 ≃ 𝑓1 ) iff there exists an 𝐻 ∈ 𝐶 (𝑋×[0, 1], 𝑌 )
such that for all 𝑥 ∈ 𝑋

𝐻 (𝑥, 0) = 𝑓0 (𝑥),
𝐻 (𝑥, 1) = 𝑓1 (𝑥),
𝐻 ( 𝐴 × [0, 1]) ⊂ 𝐵.

Homotopy Axiom.
Let 𝑓0 , 𝑓1 ∈ 𝐶 ((𝑋, 𝐴), (𝑌 , 𝐵)) be homotopic, 𝑓0 ≃ 𝑓1 . Then the induced maps on homology
coincide, i.e.,
𝐻𝑛 ( 𝑓0 ) = 𝐻𝑛 ( 𝑓1 ) : 𝐻𝑛 (𝑋, 𝐴) → 𝐻𝑛 (𝑌 , 𝐵)
holds for all 𝑛.

Remark 3.14. Let (𝑋, 𝐴) ≃ (𝑌 , 𝐵), i.e., there exist 𝑓 ∈ 𝐶 ((𝑋, 𝐴), (𝑌 , 𝐵)) and
𝑔 ∈ 𝐶 ((𝑌 , 𝐵), (𝑋, 𝐴)) such that 𝑓 ◦ 𝑔 ≃ id (𝑌 ,𝐵) and 𝑔 ◦ 𝑓 ≃ id (𝑋, 𝐴) . It then follows as be-
fore that 𝐻𝑛 (𝑋, 𝐴; 𝑅)  𝐻𝑛 (𝑌 , 𝐵; 𝑅).

88
 3.3. The Eilenberg-Steenrod axioms and applications

Exactness Axiom.
For any pair of spaces (𝑋, 𝐴) and inclusion maps 𝑖 : 𝐴 ↩→ 𝑋, 𝑗 : (𝑋, ∅) ↩→ (𝑋, 𝐴), the sequence

𝐻𝑛+1 ( 𝑗)
𝐻𝑛+1 (𝑋, 𝐴) o ···
𝜕
 𝐻𝑛 (𝑖) 𝐻𝑛 ( 𝑗)
𝐻𝑛 ( 𝐴) / 𝐻𝑛 (𝑋) / 𝐻𝑛 (𝑋, 𝐴)

𝜕
𝐻𝑛−1 (𝑖) 
··· o 𝐻𝑛−1 ( 𝐴)

is exact and natural.

Notation: For 𝐴 ⊂ 𝑋 we call

Ø
𝐴˚ = 𝑈 with 𝑈 open in 𝑋
𝑈⊂ 𝐴

the interior of 𝐴 and Ù

𝐴¯ = 𝐵 with 𝐵 closed in 𝑋
𝐵⊂𝑋

the closure of 𝐴.

Excision Axiom.
For every pair of spaces (𝑋, 𝐴) and every 𝑈 ⊂ 𝐴 with 𝑈¯ ⊂ 𝐴˚ the homomorphism

𝐻𝑛 ( 𝑗) : 𝐻𝑛 (𝑋 \ 𝑈, 𝐴 \ 𝑈) → 𝐻𝑛 (𝑋, 𝐴)

induced by the inclusion map

𝑗 : (𝑋 \ 𝑈, 𝐴 \ 𝑈) ↩→ (𝑋, 𝐴)

is an isomorphism.
The proofs of axioms A2, A3, and A4 will be given later. Before that we will show their
usefulness by studying some basic examples.

Remark 3.15. (cf. Exercise 3.1)

Í
1.) If 𝑋 ≠ ∅ is path-connected then 𝐻0 (𝑋; 𝑅)  𝑅 and generators are represented by every 0-
Í
simplex 𝜎 : Δ0 → 𝑋. In other words, the isomorphism 𝐻0 (𝑋; 𝑅) → 𝑅 is given by 𝑗 𝛼 𝑗 𝜎 𝑗 ↦→
𝑗 𝛼𝑗.
É
2.) If 𝑋 𝑘 , 𝑘 ∈ 𝐾, are the path-components of 𝑋 then 𝐻𝑛 (𝑋; 𝑅)  𝑘 ∈𝐾 𝐻 𝑛 (𝑋 𝑘 ; 𝑅).

89
3. Homology Theory

Theorem 3.16. For 𝑛 ≥ 1 we have


𝑛 𝑅, if 𝑚 = 0 or 𝑚 = 𝑛
𝐻𝑚 (𝑆 ; 𝑅) 
0, otherwise

𝑅, if 𝑚 = 𝑛
𝐻𝑚 (𝐷 𝑛 , 𝑆 𝑛−1 ; 𝑅) 
0, otherwise

Proof. a) For 𝑛 ≥ 1 the sphere 𝑆 𝑛 is path-connected and therefore 𝐻0 (𝑆 𝑛 ; 𝑅)  𝑅.

For 𝑛 = 0 we have that 𝑆0 = {𝑥, 𝑦} with the discrete topology and therefore 𝐻0 (𝑆0 ; 𝑅)  𝑅 ⊕ 𝑅.

b) We have the exact sequence

𝐻0 (𝑆0 ) −→ 𝐻0 (𝐷 1 ) −→ 𝐻0 (𝐷 1 , 𝑆0 ) −→ 0


𝑅2 𝑅
∈

(𝑎, 𝑏) ↦−→ 𝑎+𝑏

Since the map (𝑎, 𝑏) ↦→ 𝑎 + 𝑏 is onto, the map 𝐻0 (𝐷 1 ) → 𝐻0 (𝐷 1 , 𝑆0 ) must be zero. Hence
𝐻0 (𝐷 1 , 𝑆0 ) = 0.
For 𝑛 ≥ 2 we have the following exact sequence

𝐻0 (𝑆 𝑛−1 ) −→ 𝐻0 (𝐷 𝑛 ) −→ 𝐻0 (𝐷 𝑛 , 𝑆 𝑛−1 ) −→ 0
id


𝑅 −→ 𝑅

Again the second arrow has to be trivial and therefore 𝐻0 (𝐷 𝑛 , 𝑆 𝑛−1 ) = 0. This settles the case
𝑚 = 0.

c) We will now use the exact sequence

𝐻1 (𝐷 𝑛 ) −→ 𝐻1 (𝐷 𝑛 , 𝑆 𝑛−1 ) −→ 𝐻0 (𝑆 𝑛−1 ) −→ 𝐻0 (𝐷 𝑛 )

For 𝑛 = 1 we have

𝐻1 (𝐷 1 ) −→ 𝐻1 (𝐷 1 , 𝑆0 ) −→ 𝐻0 (𝑆0 ) −→ 𝐻0 (𝐷 1 )


𝐻1 ({point}) 𝑅2 𝑅
=

0 (𝑎, 𝑏) ↦−→ 𝑎+𝑏

which implies 𝐻1 (𝐷 1 , 𝑆0 )  ker((𝑎, 𝑏) ↦→ 𝑎 + 𝑏)  𝑅.

90
 3.3. The Eilenberg-Steenrod axioms and applications

For 𝑛 ≥ 2 we have

0 
𝐻1 (𝐷 𝑛 ) −→ 𝐻1 (𝐷 𝑛 , 𝑆 𝑛−1 ) −→ 𝐻0 (𝑆 𝑛−1 ) −→ 𝐻0 (𝐷 𝑛 )
=
0

from which we get 𝐻1 (𝐷 𝑛 , 𝑆 𝑛−1 ) = 0.

d) Consider the pair of spaces (𝑆 𝑛 , 𝐷 𝑛− ) where 𝐷 𝑛− = {𝑥 ∈ 𝑆 𝑛 | 𝑥0 ≤ 0} is the lower hemisphere.1

𝑆𝑛

𝐷 𝑛−

Figure 66. Sphere relative to southern hemisphere

For 𝑛 ≥ 1 both 𝐷 𝑛− and 𝑆 𝑛 are path-connected, hence the inclusion 𝐷 𝑛− ↩→ 𝑆 𝑛 induces an
isomorphism 𝐻0 (𝐷 𝑛− ) → 𝐻0 (𝑆 𝑛 ). From

0 = 𝐻1 (𝐷 𝑛− ) −→ 𝐻1 (𝑆 𝑛 ) −→ 𝐻1 (𝑆 𝑛 , 𝐷 𝑛− ) −→ 𝐻0 (𝐷 𝑛− ) −→ 𝐻0 (𝑆 𝑛 )

we see that the connecting homomorphism 𝐻1 (𝑆 𝑛 , 𝐷 𝑛− ) → 𝐻0 (𝐷 𝑛− ) must be zero. Therefore

𝐻1 (𝑆 𝑛 ) → 𝐻1 (𝑆 𝑛 , 𝐷 𝑛− ) is onto. Since 𝐻1 (𝐷 𝑛− ) = 0 we get 𝐻1 (𝑆 𝑛 )  𝐻1 (𝑆 𝑛 , 𝐷 𝑛− ).

e) Put 𝑈 −𝑛 := {𝑥 ∈ 𝑆 𝑛 | 𝑥0 < − 21 }.

𝑆𝑛

𝐷 𝑛−

𝑈 −𝑛

Figure 67. Excising a southern cap

By the excision axiom the inclusion

(𝑆 𝑛 \ 𝑈 −𝑛 , 𝐷 𝑛− \ 𝑈 −𝑛 ) ↩→ (𝑆 𝑛 , 𝐷 𝑛− )
1 For later use, we also define the upper hemisphere 𝐷 +𝑛 = {𝑥 ∈ 𝑆 𝑛 | 𝑥0 ≥ 0}.

91
3. Homology Theory

induces an isomorphism

𝐻𝑚 (𝑆 𝑛 \ 𝑈 −𝑛 , 𝐷 𝑛− \ 𝑈 −𝑛 )  𝐻𝑚 (𝑆 𝑛 , 𝐷 𝑛− )

since 𝑈¯ −𝑛 ⊂ 𝐷˚ 𝑛− . We also have that

(𝑆 𝑛 \ 𝑈 −𝑛 , 𝐷 𝑛− \ 𝑈 −𝑛 ) ≃ (𝐷 +𝑛 , 𝑆 𝑛−1 ) ≈ (𝐷 𝑛 , 𝑆 𝑛−1 )

where the homeomorphism is given by a vertical projection. This gives us the isomorphism

𝐻𝑚 (𝑆 𝑛 \ 𝑈 −𝑛 , 𝐷 𝑛− \ 𝑈 −𝑛 )  𝐻𝑚 (𝐷 𝑛 , 𝑆 𝑛−1 ).

In particular,

𝑛 𝑛 𝑅, for 𝑛 = 1
𝐻1 (𝑆 )  𝐻1 (𝑆 , 𝐷 𝑛− ) 𝑛
 𝐻1 (𝐷 , 𝑆 𝑛−1
)= (3.5)
0, otherwise

This concludes the case 𝑚 = 1.

f) Finally we treat the case 𝑚 ≥ 2 by induction. Observe that

𝐻𝑚 (𝐷 𝑛− ) −→ 𝐻𝑚 (𝑆 𝑛 ) −→ 𝐻𝑚 (𝑆 𝑛 , 𝐷 𝑛− ) −→ 𝐻𝑚−1 (𝐷 𝑛 )
=

=
0 0

and
𝐻𝑚 (𝐷 𝑛 ) −→ 𝐻𝑚 (𝐷 𝑛 , 𝑆 𝑛−1 ) −→ 𝐻𝑚−1 (𝑆 𝑛−1 ) −→ 𝐻𝑚−1 (𝐷 𝑛 )
=

=
0 0
This yields
𝐻𝑚 (𝑆 𝑛 )  𝐻𝑚 (𝑆 𝑛 , 𝐷 𝑛− )  𝐻𝑚 (𝐷 𝑛 , 𝑆 𝑛−1 )  𝐻𝑚−1 (𝑆 𝑛−1 ) (3.6)
Induction over 𝑚 concludes the proof. 

Remark 3.17. Let us describe a generator of 𝐻1 (𝑆1 )  Z geometrically. We use the isomor-
phisms in (3.5) and start with 𝐻1 (𝐷 1 , 𝑆0 ). The map 𝑐 : Δ1 → 𝐷 1 , 𝑡 ↦→ cos(𝜋(1 − 𝑡)), is a
singular 1-simplex with 𝜕𝑐 = const1 − const −1 . Under the isomorphism 𝐻0 (𝑆0 )  𝑅 2 it maps
to (1, −1) which generates the kernel of the map 𝑅 2 → 𝑅 given by (𝑎, 𝑏) ↦→ 𝑎 + 𝑏. Hence 𝑐
represents a generator of 𝐻1 (𝐷 1 , 𝑆0 ).
The isomorphism 𝐻1 (𝐷 1+ , 𝑆0 ) → 𝐻1 (𝐷 1 , 𝑆0 ) induced by vertical projection gives us a generator
of 𝐻1 (𝐷 1+ , 𝑆0 ), namely the homology class represented by 𝑐′ : Δ1 → 𝐷 1+ , 𝑡 ↦→ 𝑒𝑖 𝜋 (1−𝑡 ) .
The isomorphism 𝐻1 (𝐷 1+ , 𝑆0 ) → 𝐻1 (𝑆1 , 𝐷 1− ) is induced by the inclusion. Hence [𝑐′ ] ∈
𝐻1 (𝑆1 , 𝐷 1− ) is a generator. Now

𝐹 : Δ1 × [0, 1] → 𝑆1 , (𝑡, 𝑠) ↦→ 𝑒𝑖 𝜋 (1− (𝑠+1)𝑡 ) ,

92
 3.3. The Eilenberg-Steenrod axioms and applications

is continuous and satisfies

𝐹 (𝑡, 0) = 𝑐′ (𝑡),
𝐹 (𝑡, 1) = 𝑒𝑖 𝜋 (1−2𝑡 ) =: 𝑐′′ (𝑡),
𝐹 (0, 𝑠) = −1 ∈ 𝐷 1− ,
𝐹 (1, 𝑠) = 𝑒 −𝑖 𝜋𝑠 ∈ 𝐷 1− .

Thus 𝑐′ ≃ 𝑐′′ as maps (Δ1 , {0, 1}) → (𝑆1 , 𝐷 1− ). Using the homotopy axiom we compute

[𝑐′ ] = [𝑐′ ◦ idΔ1 ] = 𝐻1 (𝑐′ ) [idΔ1 ] = 𝐻1 (𝑐′′ ) [idΔ1 ] = [𝑐′′ ◦ idΔ1 ] = [𝑐′′ ].

Therefore [𝑐′′ ] ∈ 𝐻1 (𝑆1 , 𝐷 1− ) is a generator. Since 𝜕𝑐′′ = const𝑐′′ (1) − const 𝑐′′ (0) = const −1 −
const −1 = 0, the 1-simplex 𝑐′′ represents a homology class in 𝐻1 (𝑆1 ). Moreover, since inclusion
induces an isomorphism 𝐻1 (𝑆1 ) → 𝐻1 (𝑆1 , 𝐷 1− ) we find that [𝑐′′ ] ∈ 𝐻1 (𝑆1 ) is a generator.
Using the homotopy (𝑡, 𝑠) ↦→ 𝑒𝑖 𝜋 (𝑠−2𝑡 ) we see that 𝑡 ↦→ 𝑒 −2𝑖 𝜋𝑡 also represents a generator of
𝐻1 (𝑆1 ). Finally, playing the same game with lower and upper hemispheres interchanged shows
that 𝑡 ↦→ 𝑒2𝑖 𝜋𝑡 represents a generator of 𝐻1 (𝑆1 ) as well.

As a first application of Theorem 3.16 we now prove Brouwer’s fixed point theorem in all
dimensions.

Theorem 3.18 (Brouwer’s fixed point theorem). Let 𝑓 : 𝐷 𝑛 → 𝐷 𝑛 , 𝑛 ≥ 1, be a continuous

map. Then 𝑓 has a fixed point, i.e., there exists an 𝑥 ∈ 𝐷 𝑛 such that 𝑓 (𝑥) = 𝑥.

Proof. We assume that the map 𝑓 does not have a fixed point and then derive a contradiction.
Let 𝑛 ≥ 2 (the case 𝑛 = 1 having been treated in Remark 1.6).

𝐷𝑛

Now consider the continuous map 𝑔 : 𝐷 𝑛 → 𝑆 𝑛−1 as in 𝑥

• 𝑔(𝑥)
𝑓 (𝑥)
the picture. Denote the inclusion map by 𝜄 : 𝑆 𝑛−1 → 𝐷 𝑛
and note that 𝑔 ◦ 𝜄 = id𝑆 𝑛−1 . 𝜕𝐷 𝑛 = 𝑆 𝑛−1

Figure 68. Constructing a retraction

By the functorial properties of homology we obtain the following commutative diagram:

𝐻𝑛−1 (id)=id
𝐻𝑛−1 (𝑆 𝑛−1 ; Z)❙ Z / 𝐻𝑛−1 (𝑆 𝑛−1 ; Z)  Z
5
❙❙❙❙
❙❙❙❙ ❦❦ ❦❦❦❦❦
❙❙
𝐻𝑛−1 ( 𝜄) ❙❙❙❙) ❦❦❦❦
❦❦❦❦ 𝐻𝑛−1 (𝑔)
𝐻𝑛−1 (𝐷 𝑛 ; Z) = 0

93
3. Homology Theory

Since the identity Z → Z does not factor through 0 we run into a contradiction. 

Now we are in the position to answer the third question on page 3.

Theorem 3.19. For 𝑛 ≠ 𝑚 the space R𝑛 is not homeomorphic to R𝑚 .

Proof. Let us assume there exists a homeomorphism 𝑓 : R𝑛 → R𝑚 . Then

𝑓 : R𝑛 \ {0} → R𝑚 \ { 𝑓 (0)}
is also a homeomorphism. Since R𝑛 \ {point} ≃ 𝑆 𝑛−1 we obtain an isomorphism on the level of
homology groups:
𝐻 𝑗 (𝑆 𝑛−1 )  𝐻 𝑗 (R𝑛 \ {0})  𝐻 𝑗 (R𝑚 \ { 𝑓 (0)})  𝐻 𝑗 (𝑆 𝑚−1 ).
By Theorem 3.16 this is a contradiction for 𝑗 = 𝑛 − 1 or 𝑗 = 𝑚 − 1 unless 𝑛 = 𝑚. 

Proposition 3.20. For 𝑛 ≥ 1 let 𝑠 : 𝑆 𝑛 → 𝑆 𝑛 be the reflection given by

𝑠(𝑥0 , 𝑥1 , . . . , 𝑥 𝑛 ) = (−𝑥0 , 𝑥1 , . . . , 𝑥 𝑛 ).

Then the map

𝐻𝑛 (𝑠) : 𝐻𝑛 (𝑆 𝑛 ) → 𝐻𝑛 (𝑆 𝑛 )
is given by 𝐻𝑛 (𝑠) = − id.

Proof. The proof is given by induction. First consider the case 𝑛 = 1. Let 𝑐 : Δ1 → 𝑆1 be a
1-simplex generating 𝐻1 (𝑆1 ) and let 𝑆1 (𝑠) (𝑐) be its image under the induced homomorphism on
1-chains.

𝑆1 (𝑠)

b
∧ b
∨

Figure 69. The reflection in one dimension

We need to show that 𝐻1 (𝑠) [𝑐] = [𝑆1 (𝑠)𝑐] = −[𝑐]. We construct two 2-simplices. The first one
is as indicated in Figure 70.

94
 3.3. The Eilenberg-Steenrod axioms and applications

𝑒2
b 𝜎1

b
∧

b b

𝑒0 𝑒1

Figure 70. First 2-simplex 𝜎1

Applying the boundary operator yields

𝜕𝜎1 = 𝑆1 (𝑠)𝑐 − const + 𝑐.
The second 2-simplex is the constant map.
𝑒2
b 𝜎2

b b

𝑒0 𝑒1
Figure 71. Second 2-simplex 𝜎2

We apply the boundary operator again and we get

𝜕𝜎2 = const − const + const = const
It follows that for the chain 𝜎1 + 𝜎2 that
𝜕 (𝜎1 + 𝜎2 ) = 𝑆1 (𝑠)𝑐 + 𝑐
and hence
0 = [𝜕 (𝜎1 + 𝜎2 )] = [𝑆1 (𝑠)𝑐 + 𝑐] = [𝑆1 (𝑠)𝑐] + [𝑐] .
This yields the desired result in the case of 𝑛 = 1.
The induction step 𝑛 − 1 ⇒ 𝑛 follows from the following commutative diagram:
  
𝐻𝑛 (𝑆 𝑛 ) / 𝐻𝑛 (𝑆 𝑛 , 𝐷 𝑛 ) o
− 𝐻𝑛 (𝐷 +𝑛 , 𝑆 𝑛−1 ) / 𝐻𝑛−1 (𝑆 𝑛−1 )

𝐻𝑛 (𝑠) 𝐻𝑛 (𝑠) 𝐻𝑛 (𝑠) 𝐻𝑛−1 (𝑠)

   
  
𝐻𝑛 (𝑆 𝑛 ) / 𝐻𝑛 (𝑆 𝑛 , 𝐷 𝑛 ) o
− 𝐻𝑛 (𝐷 +𝑛 , 𝑆 𝑛−1 ) / 𝐻𝑛−1 (𝑆 𝑛−1 )

where the horizontal isomorphisms are the ones in (3.6). Commutativity of the last square
follows from Lemma 3.13 and that of the first and the second one from the fact that the horizontal
isomorphisms are induced by inclusion maps (which commute with 𝑠). Note here that we have to
choose the lower hemisphere with respect to a different coordinate than the one which is reflected
by 𝑠 so that 𝑠(𝐷 𝑛− ) = 𝐷 𝑛− . 

95
3. Homology Theory

Remark 3.21. Let 𝑎 : 𝑆 𝑛 → 𝑆 𝑛 with 𝑎(𝑥) = −𝑥 for all 𝑥 ∈ 𝑆 𝑛 be the antipodal map then
𝐻𝑛 (𝑎) = (−1) 𝑛+1 . This follows from the fact that 𝑎 is the composition of 𝑛 + 1 reflections.

Definition 3.22. A vector field on 𝑆 𝑛 is a map 𝑣 : 𝑆 𝑛 → R𝑛+1 such that 𝑣(𝑥) ⊥ 𝑥 for all 𝑥 ∈ 𝑆 𝑛 .

𝑥
𝑣(𝑥)
b

Figure 72. Vector field on 𝑆 𝑛

Theorem 3.23 (Hairy ball theorem). The 𝑛-dimensional sphere 𝑆 𝑛 admits a continuous vec-
tor field without zeros iff 𝑛 is odd.
In particular, every continuous vector field on 𝑆2 has a zero.

Loosely speaking, this means that every continuously combed hedgehog has a “bald” spot.

Proof. If 𝑛 is odd we simply set 𝑣(𝑥) := (−𝑥1 , 𝑥0 , −𝑥3 , 𝑥2 , . . . , −𝑥 𝑛 , 𝑥 𝑛−1 ). This defines a nowhere
vanishing continuous vector field.
Now let 𝑛 be even and let 𝑣 : 𝑆 𝑛 → R𝑛+1 be a continuous vector
field without a zero. Then we can put 𝑤 (𝑥) := | 𝑣𝑣 (( 𝑥)
𝑥) | . We define
𝑥
the continuous map 𝐹 : 𝑆 𝑛 × [0, 1] → 𝑆 𝑛 by
∧ b

𝐹 (𝑥, 𝑡) := 𝑥 cos(𝜋𝑡) + 𝑤 (𝑥) sin(𝜋𝑡) . 𝑤 (𝑥)

−𝑥
Since 𝐹 (𝑥, 0) = 𝑥 and 𝐹 (𝑥, 1) = −𝑥 = 𝑎(𝑥) with 𝑎(𝑥) the
antipodal map we have found that 𝑎 ≃ id. Hence 𝐻𝑛 (𝑎) =
𝐻𝑛 (id) = 1.
Figure 73. The retraction
This contradicts 𝐻𝑛 (𝑎) = (−1) 𝑛+1 = −1. 

3.4. The degree of a continuous map

For 𝑛 ≥ 1 we consider a continuous map 𝑓 : 𝑆 𝑛 → 𝑆 𝑛 . Then the homomorphism

𝐻𝑛 ( 𝑓 ) : 𝐻𝑛 (𝑆 𝑛 ; Z)  Z → 𝐻𝑛 (𝑆 𝑛 ; Z)  Z

is given by multiplication with a number which we denote deg( 𝑓 ) ∈ Z.

96
 3.4. The degree of a continuous map

Definition 3.24. The number deg( 𝑓 ) is called the degree of the map 𝑓 . The degree can be
defined in the same way for continuous maps 𝑓 : (𝐷 𝑛 , 𝑆 𝑛−1 ) → (𝐷 𝑛 , 𝑆 𝑛−1 ).

Examples 3.25. 1.) For a reflection we have deg(𝑠) = −1.

2.) For the antipodal map we have seen that deg(𝑎) = (−1) 𝑛+1 .

Lemma 3.26. The degree of a function has the following properties

(i) deg(id) = 1;

(ii) deg(const) = 0;

(iii) deg( 𝑓 ◦ 𝑔) = deg( 𝑓 ) deg(𝑔);

(iv) If 𝑓 ≃ 𝑔 then deg( 𝑓 ) = deg(𝑔);

(v) If the map 𝑓 is a homotopy equivalence then deg( 𝑓 ) = ±1;

(vi) For 𝑓 : (𝐷 𝑛+1 , 𝑆 𝑛 ) → (𝐷 𝑛+1 , 𝑆 𝑛 ) we have deg( 𝑓 ) = deg( 𝑓 𝑆𝑛

).

Proof. The first and third assertion follow directly from the functorial property of 𝐻𝑛 ( 𝑓 ). The
fourth and fifth statement follow from the homotopy axiom. The second assertion follows from
the fact that the homomorphism induced by a constant map factors through 𝐻𝑛 ( 𝑝𝑡) = 0. The last
statement of the lemma follows fromm the commutativity of the following diagram:

𝐻𝑛+1 ( 𝑓 )
𝐻𝑛+1 (𝐷 𝑛+1 , 𝑆 𝑛 ) / 𝐻𝑛+1 (𝐷 𝑛+1 , 𝑆 𝑛 )

 𝜕 𝜕 
 𝐻𝑛+1 ( 𝑓 | 𝑆 𝑛 ) 
𝐻𝑛 (𝑆 𝑛 ) / 𝐻𝑛 (𝑆 𝑛 )

Theorem 3.27. (i) Every 𝑓 ∈ 𝐶 (𝑆 𝑛 , 𝑆 𝑛 ) without fixed points satisfies deg( 𝑓 ) = (−1) 𝑛+1 .

(ii) Every 𝑓 ∈ 𝐶 (𝑆 𝑛 , 𝑆 𝑛 ) without an antipodal point, i.e., 𝑓 (𝑥) ≠ −𝑥 for all 𝑥 ∈ 𝑆 𝑛 , satisfies
deg( 𝑓 ) = 1.

(iii) For 𝑛 even every 𝑓 ∈ 𝐶 (𝑆 𝑛 , 𝑆 𝑛 ) has a fixed point or an antipodal point.

97
3. Homology Theory

Proof. (i) Let 𝑓 ∈ 𝐶 (𝑆 𝑛 , 𝑆 𝑛 ) be without a fixed point. Then the line segment joining 𝑓 (𝑥)
and −𝑥 does not contain the origin.

𝑥 b

b
0

𝑓 (𝑥) b b
−𝑥

Hence we can define the continuous map

(1 − 𝑡) 𝑓 (𝑥) − 𝑡𝑥
𝐹 : 𝑆 𝑛 × [0, 1] → 𝑆 𝑛 , 𝐹 (𝑥, 𝑡) := .
|(1 − 𝑡) 𝑓 (𝑥) − 𝑡𝑥|

Since 𝐹 (𝑥, 0) = 𝑓 (𝑥) and 𝐹 (𝑥, 1) = −𝑥 = 𝑎(𝑥) with 𝑎 being the antipodal map the map 𝐹
is a homotopy for 𝑓 ≃ 𝑎. Thus

deg( 𝑓 ) = deg(𝑎) = (−1) 𝑛+1 .

(ii) Now let 𝑓 ∈ 𝐶 (𝑆 𝑛 , 𝑆 𝑛 ) be without antipodal points. Then we can define

(1 − 𝑡) 𝑓 (𝑥) + 𝑡𝑥
𝐺 (𝑥, 𝑡) := .
|(1 − 𝑡) 𝑓 (𝑥) + 𝑡𝑥|

𝑥 b

b
0

b
−𝑥
b

𝑓 (𝑥)

Since 𝐺 (𝑥, 0) = 𝑓 (𝑥) and 𝐺 (𝑥, 1) = 𝑥 we have that 𝑓 ≃ id and deg( 𝑓 ) = deg(id) = 1
follows.

(iii) Finally, assume that 𝑓 ∈ 𝐶 (𝑆 𝑛 , 𝑆 𝑛 ) has neither fixed points nor antipodal points. Then
deg( 𝑓 ) = (−1) 𝑛+1 by ((i)) and by deg( 𝑓 ) = 1 by ((ii)). Thus 𝑛 must be odd. 

98
 3.4. The degree of a continuous map

Let 𝜇 : 𝑆 𝑛 × 𝑆 𝑛 → 𝑆 𝑛 for 𝑛 ≥ 1 be a continuous map. We choose 𝑝 ∈ 𝑆 𝑛 arbitrarily and define

𝑗1 : 𝑆𝑛 → 𝑆𝑛 × 𝑆𝑛 , 𝑥 ↦→ (𝑥, 𝑝),
𝑛 𝑛 𝑛
𝑗2 : 𝑆 → 𝑆 × 𝑆 , 𝑥 ↦→ ( 𝑝, 𝑥) .

We then get the following diagram:

(𝐻𝑛 ( 𝑗1 ),𝐻𝑛 ( 𝑗2 ) ) 𝐻𝑛 ( 𝜇)
Z2  𝐻𝑛 (𝑆 𝑛 ) ⊕ 𝐻𝑛 (𝑆 𝑛 ) / 𝐻𝑛 (𝑆 𝑛 × 𝑆 𝑛 ) / 𝐻𝑛 (𝑆 𝑛 )  Z
6

(𝑑1 ,𝑑2 )

with 𝑑 𝜇 ∈ Z.

Definition 3.28. The pair of numbers (𝑑1 , 𝑑2 ) ∈ Z2 is called the bidegree of the map 𝜇.

Remark 3.29. The bidegree does not depend on the choice of 𝑝. If one chooses another 𝑝 ′ , then
a path from 𝑝 to 𝑝 ′ will yield a homotopy between the corresponding embedding maps 𝑗 𝜈 and 𝑗 𝜈′ .
Hence they induce the same homomorphisms on homology and therefore the same bidegrees.

Examples 3.30. 1.) Consider the case 𝑛 = 1, 𝑆1 ⊂ C and let the map 𝜇 be given by 𝜇(𝑧1 , 𝑧2 ) =
𝑧1 𝑧2 . Choose 𝑝 = 1 ∈ 𝑆1 . Then 𝜇 ◦ 𝑗 1 = id and thus

𝑑1 = deg(𝜇 ◦ 𝑗1 ) = deg(id) = 1.

Similarly, we get 𝑑2 = 1. Hence the bidegree by (𝑑1 , 𝑑2 ) = (1, 1).

2.) In the case of 𝑛 = 3, 𝑆3 ⊂ H we consider the map 𝜇 given by quaternionic multiplication,

𝜇(ℎ1 , ℎ2 ) = ℎ1 ℎ2 . Similar reasoning shows that the bidegree is again given by (𝑑1 , 𝑑2 ) = (1, 1).

Remark 3.31. The quaternions H form a division algebra isomorphic to R4 as a vector space.
The algebra H is associate and noncommutative. The standard vector space basis we be denoted
by 1, 𝑖, 𝑗, 𝑘. Therefore any ℎ ∈ H can be uniquely written as

ℎ = ℎ0 + ℎ1𝑖 + ℎ2 𝑗 + ℎ3 𝑘 .

Quaternionic multiplication is now determined by the relations 𝑖 2 = 𝑗 2 = 𝑘 2 = 𝑖 𝑗 𝑘 = −1. With

the help of the conjugation
ℎ∗ = ℎ0 − ℎ1 𝑖 − ℎ2 𝑗 − ℎ3 𝑘
√
we can define |ℎ| := ℎ∗ ℎ. We regard 𝑆3 ⊂ H as the set of unit-length quaternions.

99
3. Homology Theory

Proposition 3.32. Let 𝑛 = 1 or 𝑛 = 3 and let 𝑘 ∈ Z. The map 𝑓 𝑘 : 𝑆 𝑛 → 𝑆 𝑛 with 𝑓 𝑘 : 𝑧 ↦→ 𝑧 𝑘

given by complex multiplication in the case 𝑛 = 1 and by quaternionic multiplication in the
case 𝑛 = 3 has degree 𝑘.

Proof. The proof is by induction on 𝑘. For 𝑘 = 0 and 𝑘 = 1 the statement is trivial because
constant maps have degree 0 while the identity has degree 1. The case of 𝑘 = −1 has already
been shown for 𝑛 = 1, since here 𝑧 ↦→ 𝑧 −1 = 𝑧¯ is a reflection and hence deg( 𝑓 −1 ) = −1. On the
other hand for 𝑘 = −1, 𝑛 = 3 we note that 𝑆1 ⊂ 𝑆2 ⊂ 𝑆3 regarding C ⊂ H. Now the following
commutative diagram

𝐻1 (𝑆1 )  𝐻2 (𝑆2 )  𝐻3 (𝑆3 )

𝐻1 ( 𝑓−1 | 𝑆 1 )=deg ( 𝑓 −1 | 𝑆 1 )=−1 𝐻3 ( 𝑓−1 )=deg ( 𝑓−1 )
 
𝐻1 (𝑆1 )  𝐻2 (𝑆2 )  𝐻3 (𝑆3 )

implies that deg( 𝑓 −1 ) = −1 in the quaternionic case too. The horizontal isomorphisms are
obtained as the composition of the isomorphisms

𝐻 𝑘 (𝑆 𝑘 )  𝐻 𝑘+1 (𝐷 𝑘+1 , 𝑆 𝑘 )  𝐻 𝑘+1 (𝑆 𝑘+1 , 𝐷 −𝑘+1 )  𝐻 𝑘+1 (𝑆 𝑘+1 ).

We used that the quaternionic multiplication restricted to the complex numbers C is just complex
multiplication.
Now we are ready to carry out the induction over 𝑘. We consider 𝑘 > 0 and we show that the
statement for 𝑘 − 1 implies that for 𝑘. Let 𝜇 be complex (resp. quaternionic) multiplication. Then

 
deg( 𝑓 𝑘 ) = deg(𝜇 ◦ ( 𝑓 𝑘−1 , 𝑓1 ))
deg( 𝑓 𝑘−1 )
= (1, 1) ·
deg( 𝑓1 )
= deg( 𝑓 𝑘−1 ) + deg( 𝑓1 )
= 𝑘 − 1 + 1 = 𝑘.

The case 𝑘 < 0 is treated similarly. 

Now we can show that the fundamental theorem of algebra 2.34 also holds for quaternionic
polynomials.

Theorem 3.33 (Fundamental theorem of algebra for quaternions). Every quaternionic

polynomial
𝑝(𝑧) = 𝑧 𝑘 + 𝛼1 𝑧 𝑘−1 + . . . + 𝛼 𝑘

100
 3.4. The degree of a continuous map

of positive degree 𝑘 has a quaternionic root.

Proof. Suppose the polynomial 𝑝 has no root. Then we can define the continuous map
𝑝ˆ : 𝑆3 → 𝑆3 with 𝑝(𝑧)
ˆ := | 𝑝𝑝 (𝑧)
(𝑧) 3 3 𝑝 (𝑡 𝑧)
| . Now consider 𝐹 (𝑧, 𝑡) : 𝑆 × [0, 1] → 𝑆 with 𝐹 (𝑧, 𝑡) := | 𝑝 (𝑡 𝑧) | .
We observe that 𝐹 (𝑧, 0) = const and 𝐹 (𝑧, 1) = 𝑝(𝑧), ˆ hence we have 𝑝(𝑧)
ˆ ≃ const and conse-
quently deg( 𝑝)
ˆ = 0.
On the other hand, we can put for 𝑧 ∈ 𝑆3 and 𝑡 > 0


𝑡 𝑘 𝑝 𝑧𝑡
𝐺 (𝑧, 𝑡) := 𝑘 𝑧

|𝑡 𝑝 𝑡 |
and observe that the expression
𝑧
𝑡𝑘 𝑝 = 𝑧 𝑘 + 𝑡𝛼1 𝑧 𝑘−1 + . . . + 𝑡 𝑘 𝛼
𝑡
extends continuously to 𝑡 = 0. The map 𝐺 : 𝑆3 × [0, 1] → 𝑆3 satisfies 𝐺 (𝑧, 0) = 𝑓 𝑘 (𝑧) and again
𝐺 (𝑧, 1) = 𝑝(𝑧).
ˆ Thus 𝑝ˆ ≃ 𝑓 𝑘 and hence deg( 𝑝)
ˆ = 𝑘. This contradicts 𝑘 > 0. 

Now let 𝑈 ⊂ 𝑆 𝑛 be open, 𝑛 ≥ 1. Consider a continuous map 𝑓 : 𝑈 → 𝑆 𝑛 and a point 𝑝 ∈ 𝑆 𝑛

such that 𝑓 −1 ( 𝑝) is compact. Then we have the following diagram (with deg 𝑝 ( 𝑓 ) ∈ Z):

(𝑖) (𝑖𝑖)
Z  𝐻𝑛 (𝑆 𝑛 ; Z) / 𝐻𝑛 (𝑆 𝑛 , 𝑆 𝑛 \ 𝑓 −1 ( 𝑝); Z) o

𝐻𝑛 (𝑈, 𝑈 \ 𝑓 −1 ( 𝑝); Z) (3.7)
deg 𝑝 ( 𝑓 ) 𝐻𝑛 ( 𝑓 )
 (𝑖𝑖𝑖) 
Z  𝐻𝑛 (𝑆 𝑛 ; Z) 
/ 𝐻𝑛 (𝑆 𝑛 , 𝑆 𝑛 \ {𝑝}; Z)

We observe:
1. Concerning (𝑖): This homomorphism is induced by the inclusion
𝑆 𝑛 = (𝑆 𝑛 , ∅) ↩→ (𝑆 𝑛 , 𝑆 𝑛 \ 𝑓 −1 ( 𝑝)).

2. Concerning (𝑖𝑖): The inclusion (𝑈, 𝑈 \ 𝑓 −1 ( 𝑝)) ↩→ (𝑆 𝑛 , 𝑆 𝑛 \ 𝑓 −1 ( 𝑝)) induces an isomor-

phism by the excision axiom.
3. Concerning (𝑖𝑖𝑖): Consider the exact homology sequence of (𝑆 𝑛 , 𝑆 𝑛 \ {𝑝})

0 = 𝐻𝑛 (𝑆 𝑛 \ {𝑝}) → 𝐻𝑛 (𝑆 𝑛 ) → 𝐻𝑛 (𝑆 𝑛 , 𝑆 𝑛 \ {𝑝}) → 𝐻𝑛−1 (𝑆 𝑛 \ {𝑝}) → 𝐻𝑛−1 (𝑆 𝑛 ) .
| {z }




0 for 𝑛 ≥ 2

=

Z for 𝑛 = 1


Hence in both cases 𝑛 = 1 and 𝑛 ≥ 2 the homomorphism (𝑖𝑖𝑖) is an isomorphism.

101
3. Homology Theory

Definition 3.34. The number deg 𝑝 ( 𝑓 ) is called the local degree of 𝑓 over 𝑝.

Examples 3.35. 1.) If 𝑝 ∉ im( 𝑓 ) then deg 𝑝 ( 𝑓 ) = 0 because 𝐻𝑛 (𝑆 𝑛 , 𝑆 𝑛 ) = 0.

2.) If 𝑓 : 𝑈 → 𝑆 𝑛 is the inclusion map then deg 𝑝 ( 𝑓 ) = 1 for all 𝑝 ∈ 𝑈. Namely, in this case the
homomorphisms (i) and (iii) in (3.7) coincide and so do 𝐻𝑛 ( 𝑓 ) and (ii).

3.) For a homeomorphism 𝑓 : 𝑈 → 𝑓 (𝑈) ⊂ 𝑆 𝑛 we have that deg 𝑝 ( 𝑓 ) = ±1 for all 𝑝 ∈ 𝑓 (𝑈).
Namely, the homomorphisms (i) and 𝐻𝑛 ( 𝑓 ) in (3.7) are isomorphisms in this case, hence
multiplication by deg 𝑝 ( 𝑓 ) is an isomorphism, thus deg 𝑝 ( 𝑓 ) = ±1.

Proposition 3.36. Assume that 𝑓 −1 ( 𝑝) ⊂ 𝐾 ⊂ 𝑉 ⊂ 𝑈 where 𝐾 is compact and 𝑉 open. Then

the degree deg 𝑝 ( 𝑓 ) is given by:


𝐻𝑛 (𝑆 𝑛 ) / 𝐻𝑛 (𝑆 𝑛 , 𝑆 𝑛 \ 𝐾) o 𝐻𝑛 (𝑉, 𝑉 \ 𝐾)
deg 𝑝 ( 𝑓 ) 𝐻𝑛 ( 𝑓 | 𝑉 )
 
𝐻𝑛 (𝑆 𝑛 ) 
/ 𝐻𝑛 (𝑆 𝑛 , 𝑆 𝑛 \ {𝑝})

Hence we can replace 𝑓 −1 ( 𝑝) by a larger compact set in 𝑈 and also 𝑈 by a smaller open
neighborhood of 𝑓 −1 ( 𝑝). For this reason we call deg 𝑝 ( 𝑓 ) local.

Proof. The assertion follows from the commutativity of the following diagram:


𝐻𝑛 (𝑆 𝑛6 , 𝑆 𝑛 O \ 𝑓 −1 ( 𝑝)) o 𝐻𝑛 (𝑈, 𝑈 \O 𝑓 −1❙( 𝑝))
♠ ❙❙❙
♠♠♠ ❙❙❙𝐻❙𝑛 ( 𝑓 )
♠♠♠♠♠ ❙❙❙❙
♠ ❙❙❙
♠♠♠ ) 
𝐻𝑛 (𝑆 𝑛 ) ◗ 𝐻𝑛 (𝑆
5
𝑛 , 𝑆 𝑛 \ {𝑝}) o 𝐻𝑛 (𝑆 𝑛 )
◗◗◗ ❦❦❦
◗◗◗
◗◗◗ ❦❦❦❦❦❦❦
❦
❦❦❦❦ 𝐻𝑛 ( 𝑓 | 𝑉 )
◗◗◗
( 
𝐻𝑛 (𝑆 𝑛 , 𝑆 𝑛 \ 𝐾) o 𝐻𝑛 (𝑉, 𝑉 \ 𝐾)

where all but two arrows are induced by inclusions. 

Corollary 3.37. For 𝑓 : 𝑆 𝑛 → 𝑆 𝑛 we have that deg( 𝑓 ) = deg 𝑝 ( 𝑓 ) for all 𝑝 ∈ 𝑆 𝑛 .

Proof. Choose 𝐾 = 𝑉 = 𝑆 𝑛 in Proposition 3.36. 

102
 3.4. The degree of a continuous map

Lemma 3.38. Let 𝑓 : (𝐷 +𝑛 , 𝑆 𝑛−1 ) → (𝐷 +𝑛 , 𝑆 𝑛−1 ) be continuous and 𝑝 ∈ 𝐷˚ +𝑛 such that

𝑓 −1 ( 𝑝) ⊂ 𝐷˚ +𝑛 is compact. Then
 
deg( 𝑓 ) = deg 𝑝 𝑓 | 𝐷˚ +𝑛 .

Proof. We extend 𝑓 to a continuous map 𝐹 : (𝑆 𝑛 , 𝐷 𝑛− ) → (𝑆 𝑛 , 𝐷 𝑛− ) by mapping the circular arc

from a point 𝑥 on the equator 𝑆 𝑛−1 to the south pole to the corresponding arc from 𝑓 (𝑥) to the
south pole. In formulas,




 𝑓 (𝑥0 , 𝑥 )
′


if 𝑥0 ≥ 0,
′


′
′ ′ ′

𝐹 (𝑥0 , 𝑥 ) = 𝑥0 , k𝑥 k · 𝑓 (0, 𝑥 /k𝑥 k) if − 1 < 𝑥0 < 0,



 (−1, 0)
 if 𝑥0 = −1.

Here we wrote 𝑥 = (𝑥0 , 𝑥 ′ ) = (𝑥0 , 𝑥1 , . . . , 𝑥 𝑛 ) ∈ 𝑆 𝑛 ⊂ R𝑛+1 . Then by Corollary 3.37 and

Proposition 3.36 with 𝐾 = 𝑓 −1 ( 𝑝), 𝑉 = 𝐷˚ +𝑛 and 𝑈 = 𝑆 𝑛 we find

deg(𝐹) = deg 𝑝 (𝐹) = deg 𝑝 (𝐹 | 𝐷˚ +𝑛 ) = deg 𝑝 ( 𝑓 | 𝐷˚ +𝑛 ) . (3.8)

On the other hand, the commutative diagram

(𝐷 +𝑛 , 𝑆 𝑛−1 ) / (𝑆 𝑛 , 𝐷 𝑛 ) o
− 𝑆𝑛
𝑓 𝐹 𝐹
  
(𝐷 +𝑛 , 𝑆 𝑛−1 ) / (𝑆 𝑛 , 𝐷 𝑛 ) o
− 𝑆𝑛

yields on the level of homology

 
𝐻𝑛 (𝐷 +𝑛 , 𝑆 𝑛−1 ) / 𝐻𝑛 (𝑆 𝑛 , 𝐷 𝑛 ) o
− 𝐻𝑛 (𝑆 𝑛 )
· deg( 𝑓 ) 𝐻𝑛 (𝐹 ) · deg(𝐹 )
  
 
𝐻𝑛 (𝐷 +𝑛 , 𝑆 𝑛−1 ) / 𝐻𝑛 (𝑆 𝑛 , 𝐷 𝑛 ) o
− 𝐻𝑛 (𝑆 𝑛 )

Hence deg(𝐹) = deg( 𝑓 ). This together with (3.8) concludes the proof. 

Proposition 3.39. Let 𝑓 : 𝑈 ⊂ 𝑆 𝑛 → 𝑆 𝑛 be continuous, let 𝑝 ∈ 𝑆 𝑛 with 𝑓 −1 ( 𝑝) compact and

let 𝑔 ∈ 𝐶 (𝑆 𝑛 , 𝑆 𝑛 ). Then
deg 𝑝 ( 𝑓 ◦ 𝑔) = deg 𝑝 ( 𝑓 ) · deg(𝑔).

103
3. Homology Theory

Proof. This follows from the commutative diagram:

𝐻𝑛 (𝑆 𝑛) / 𝐻𝑛 (𝑆 𝑛 , 𝑆 𝑛 \ 𝑓 −1 ( 𝑝))  𝐻𝑛 (𝑈, 𝑈 \ 𝑓 −1 ( 𝑝))

O
deg 𝑝 ( 𝑓 ) 𝐻𝑛 ( 𝑓 )
- 
deg(𝑔)=𝐻𝑛 (𝑔) 1 𝐻𝑛 (𝑆 𝑛 )  𝐻𝑛 (𝑆 𝑛 , 𝑆 𝑛 \ {𝑝})
O
deg 𝑝 ( 𝑓 ◦𝑔)
𝐻𝑛 ( 𝑓 ◦𝑔)

𝐻𝑛 (𝑆 𝑛 ) / 𝐻𝑛 (𝑆 𝑛 , 𝑆 𝑛 \ ( 𝑓 ◦ 𝑔) −1 ( 𝑝))  𝐻𝑛 (𝑔 −1 (𝑈), 𝑔 −1 (𝑈) \ ( 𝑓 ◦ 𝑔) −1 ( 𝑝))

Remark 3.40. If 𝑋𝑖 are the path-components of 𝑋 then

Ê
𝐻𝑚 (𝑋)  𝐻𝑚 (𝑋𝑖 ),
𝑖

see Exercise 3.1. The isomorphism is induced by the inclusion maps of the connected components
into 𝑋. Similarly one sees that for 𝐴 ⊂ 𝑋 and 𝐴𝑖 = 𝐴 ∩ 𝑋𝑖

Ê
𝐻𝑚 (𝑋, 𝐴)  𝐻𝑚 (𝑋𝑖 , 𝐴𝑖 ) .
𝑖

Proposition 3.41 (Additivity of the local degree). Let 𝑓 : 𝑈 → 𝑆 𝑛 be a continuous map. Let
𝑝 ∈ 𝑆 𝑛 be such that 𝑓 −1 ( 𝑝) is compact. Let 𝑈𝜆 ⊂ 𝑈 be open and put 𝑓𝜆 := 𝑓 |𝑈𝜆 , 𝜆 = 1, . . . , 𝑟.
Assume that 𝑓 −1 ( 𝑝) is the disjoint union of the 𝑓𝜆−1 ( 𝑝), i.e. 𝑓 −1 ( 𝑝) = ⊔𝑟𝜆=1 𝑓𝜆−1 ( 𝑝). Then

Õ
𝑟
deg 𝑝 ( 𝑓 ) = deg 𝑝 ( 𝑓𝜆 ) .
𝜆=1

Before proving the proposition we give some examples.

104
 3.4. The degree of a continuous map

𝑈
Example 3.42
Assume that 𝑓 −1 ( 𝑝) is a finite set, i.e.
b

𝑓 −1 ( 𝑝) = {𝑝 1 , . . . , 𝑝𝑟 }. Now choose 𝑈𝜆 such

that 𝑝 𝜆 ∈ 𝑈𝜆 and 𝑝 𝜈 ∉ 𝑈𝜆 for 𝜈 ≠ 𝜆. It fol-
lows that
Õ
b
𝑟
deg 𝑝 ( 𝑓 ) = deg 𝑝 ( 𝑓𝜆 ) .
𝜆=1

If the map 𝑓 is a local homeomorphism, then b

deg 𝑝 ( 𝑓𝜆 ) = ±1 for every 𝜆.

𝑈𝜆

Figure 74. The case of finite preimage

Example 3.43. Consider the map 𝑓 𝑘 : 𝑆1 → 𝑆1 with 𝑓 (𝑧) = 𝑧 𝑘 , 𝑘 > 0, and set 𝑝 = 1. We write

𝑓 𝑘−1 (1) = {𝑘-th unit roots} = {𝜉1 , . . . , 𝜉 𝑘 }

and find that 𝑓 𝑘 | small neighborhood of 𝜉𝜆 is a homeomorphism. Hence deg( 𝑓 𝑘,𝜆 ) = ±1. Since 𝑘 > 0,
the restriction of 𝑓 𝑘 to a small neighborhood of 𝜉𝜆 is homotopic to an embedding of a (𝑘 times
larger) neighborhood of 𝜉𝜆 into 𝑆1 . Hence deg1 ( 𝑓 𝑘,𝜆 ) = 1. We conclude deg1 ( 𝑓 𝑘 ) = 𝑘.

Proof of Proposition 3.41. Choose open neighborhoods 𝑉𝜆 such that 𝑓𝜆−1 ( 𝑝) ⊂ 𝑉𝜆 ⊂ 𝑈𝜆 with
𝑉𝜆 ∩ 𝑉𝜇 = ∅ for 𝜆 ≠ 𝜇. Now put 𝑉 = ∪𝑟𝜆=1𝑉𝜆 . Proposition 3.36 tells us deg 𝑝 ( 𝑓 ) = deg 𝑝 ( 𝑓 | 𝑉 ).
The commutative diagram

deg 𝑝 ( 𝑓 | 𝑉 )

𝐻𝑛 ( 𝑓 | 𝑉 ) ,
𝐻𝑛 (𝑆 𝑛 ) / 𝐻𝑛 (𝑉, 𝑉 \ 𝑓 −1 ( 𝑝))
O
/ 𝐻𝑛 (𝑆 𝑛 , 𝑆 𝑛 \ {𝑝}) o
O 
𝐻𝑛 (𝑆
O
𝑛)

1
© ª
­.®
­.®
­.®  (1,...,1) (1,...,1)
­ ®
«1¬
É𝑟  É𝑟 ⊕𝜆 𝐻𝑛 ( 𝑓𝜆 ) É𝑟  É𝑟
𝜆=1 𝐻𝑛 (𝑆 𝑛 ) /
𝜆=1 𝐻𝑛 (𝑉𝜆 , 𝑉𝜆 \ 𝑓𝜆−1 ( 𝑝)) /
𝜆=1 𝐻𝑛 (𝑆 𝑛 , 𝑆 𝑛 \ {𝑝}) o 2 𝜆=1 𝐻𝑛 (𝑆 𝑛 )

©deg 𝑝 ( 𝑓1 ) ª
­ ®
­ ®
­ .. ®
­ . ®
­ ®
«
deg 𝑝 ( 𝑓𝑟 ) ¬

105
3. Homology Theory

yields

© ª ©.ª
deg 𝑝 ( 𝑓1 ) 1
deg 𝑝 ( 𝑓 | 𝑉 ) = (1, . . . , 1) · ­­ ..
. ® ­ ®
® · ­ .. ® = deg 𝑝 ( 𝑓1 ) + . . . + deg 𝑝 ( 𝑓𝑟 ). 
« deg 𝑝 ( 𝑓𝑟 ) ¬ «1¬

3.5. Homological algebra

Before continuing with topological considerations we clarify some of the underlying algebra.
Throughout this section let 𝑅 be a commutative ring with unit element.

Definition 3.44. A complex of 𝑅-modules 𝐾∗ is a sequence

𝜕𝑛+1 𝜕𝑛
... / 𝐾𝑛+1 / 𝐾𝑛 / 𝐾𝑛−1 / ...

of 𝑅-modules 𝐾𝑛 together with homomorphisms 𝜕𝑛 such that

𝜕𝑛 ◦ 𝜕𝑛+1 = 0 (3.9)

for all 𝑛 ∈ Z. We define the space of 𝑛-cycles

𝑍 𝑛 𝐾∗ := {𝑥 ∈ 𝐾𝑛 | 𝜕𝑛 𝑥 = 0} = ker(𝜕𝑛 ) ,

and the space of 𝑛-boundaries

𝐵𝑛 𝐾∗ := {𝜕𝑛+1 𝑥 ∈ 𝐾𝑛 | 𝑥 ∈ 𝐾𝑛+1 } = im(𝜕𝑛+1 ).

By (3.9) 𝐵𝑛 𝐾 ⊂ 𝑍 𝑛 𝐾 so that we can define the 𝑛-th homology by

𝑍 𝑛 𝐾∗
𝐻𝑛 𝐾∗ := .
𝐵 𝑛 𝐾∗

Example 3.45. For

(
𝑆𝑛 (𝑋, 𝐴; 𝑅) = 𝑆𝑛 (𝑋; 𝑅)/𝑆𝑛 ( 𝐴; 𝑅), 𝑛 ≥ 0
𝐾𝑛 =
0, 𝑛<0

we obtain relative singular homology 𝐻𝑛 𝐾∗ = 𝐻𝑛 (𝑋, 𝐴; 𝑅).

106
 3.5. Homological algebra

Definition 3.46. Given two complexes 𝐾∗ and 𝐾∗′ a chain map 𝜑∗ : 𝐾∗ → 𝐾∗′ is a sequence of
homomorphisms 𝜑 𝑛 : 𝐾𝑛 → 𝐾𝑛′ such that
𝜑𝑛+1
𝐾𝑛+1 / 𝐾′
𝑛+1
′
𝜕𝑛+1
𝜕𝑛+1
 𝜑𝑛 
𝐾𝑛 / 𝐾′
𝑛

commutes for all 𝑛 ∈ Z.

Example 3.47. For a continuous map 𝑓 : (𝑋, 𝐴) → (𝑌 , 𝐵) the homomorphisms 𝜑 𝑛 = 𝑆𝑛 ( 𝑓 )

constitute a chain map.

Given a general chain map 𝜑∗ we conclude from

′
𝜑 𝑛 ◦ 𝜕𝑛+1 = 𝜕𝑛+1 ◦ 𝜑 𝑛+1

that 𝜑 𝑛 (𝑍 𝑛 𝐾∗ ) ⊂ 𝑍 𝑛 𝐾∗′ and also that 𝜑 𝑛 (𝐵𝑛 𝐾∗ ) ⊂ 𝐵𝑛 𝐾∗′ . Hence 𝜑∗ induces homomorphisms
𝐻𝑛 (𝜑∗) : 𝐻𝑛 𝐾∗ → 𝐻𝑛 𝐾∗′ by 𝐻𝑛 (𝜑∗) ([𝑧]) = [𝜑 𝑛 𝑧]. This construction is functorial in the sense
that
𝐻𝑛 (𝜑∗ ◦ 𝜓∗ ) = 𝐻𝑛 (𝜑∗) ◦ 𝐻𝑛 (𝜓∗ ) and 𝐻𝑛 (id𝐾∗ ) = id 𝐻𝑛 𝐾∗ .

𝜑∗ 𝜓∗
Definition 3.48. A sequence · · · −→ 𝐾∗′ −→ 𝐾∗ −→ 𝐾∗′′ −→ · · · of chain maps is called
𝜑𝑛 𝜓𝑛
exact iff the sequence · · · −→ 𝐾𝑛′ −→ 𝐾𝑛 −→ 𝐾𝑛′′ −→ · · · is exact for every 𝑛 ∈ Z.

𝑖∗ 𝑝∗
Proposition 3.49. If 0 → 𝐾∗′ −→ 𝐾∗ −→ 𝐾∗′′ → 0 is an exact sequence of complexes then
𝐻𝑛 (𝑖∗ ) 𝐻𝑛 ( 𝑝∗ )
the sequence 𝐻𝑛 𝐾∗′ −→ 𝐻𝑛 𝐾∗ −→ 𝐻𝑛 𝐾∗′′ is exact for every 𝑛 ∈ Z.

Proof. a) By assumption we have that 𝑝 ∗ ◦ 𝑖∗ = 0 and hence

0 = 𝐻𝑛 ( 𝑝 ∗ ◦ 𝑖 ∗ ) = 𝐻𝑛 ( 𝑝 ∗ ) ◦ 𝐻𝑛 (𝑖 ∗ ).

Therefore im 𝐻𝑛 (𝑖 ∗ ) ⊂ ker 𝐻𝑛 ( 𝑝 ∗ ).

b) It remains to show that ker 𝐻𝑛 ( 𝑝 ∗ ) ⊂ im 𝐻𝑛 (𝑖 ∗ ).

Let 𝑧 ∈ 𝑍 𝑛 𝐾∗ represent [𝑧] ∈ ker 𝐻𝑛 ( 𝑝 ∗ ). Hence 𝑝 𝑛 𝑧 = 𝜕 ′′ 𝑥 ′′ for some 𝑥 ′′ ∈ 𝐾𝑛+1
′′ . Since 𝑝
𝑛+1
′′
is surjective we can choose 𝑥 ∈ 𝐾𝑛+1 with 𝑝 𝑛+1 𝑥 = 𝑥 . We compute

𝑝 𝑛 (𝑧 − 𝜕𝑥) = 𝜕 ′′ 𝑥 ′′ − 𝜕 ′′ 𝑝 𝑛+1 𝑥 = 𝜕 ′′ 𝑥 ′′ − 𝜕 ′′ 𝑥 ′′ = 0 .

107
3. Homology Theory

By exactness of the complex there exists 𝑦 ′ ∈ 𝐾𝑛′ such that 𝑧 − 𝜕𝑥 = 𝑖 𝑛 𝑦 ′ . Now we get

𝑖 𝑛−1 𝜕 ′ 𝑦 ′ = 𝜕𝑖 𝑛 𝑦 ′ = 𝜕 [𝑧 − 𝜕𝑥] = 𝜕𝑧 − 𝜕𝜕𝑥 = 0 .

Since 𝑖 𝑛−1 is injective it follows that 𝜕 ′ 𝑦 ′ = 0, i.e., 𝑦 ′ ∈ 𝑍 𝑛 𝐾∗′ represents an element in homology.
Finally we see
𝐻𝑛 (𝑖 ∗ ) ([𝑦 ′ ]) = [𝑖 𝑛 𝑦 ′ ] = [𝑧 − 𝜕𝑥] = [𝑧].
This shows ker 𝐻𝑛 ( 𝑝 ∗ ) ⊂ im 𝐻𝑛 (𝑖 ∗ ). 

𝑖∗ 𝑝∗
Definition 3.50. Let 0 → 𝐾∗′ −→ 𝐾∗ −→ 𝐾∗′′ → 0 be an exact sequence of complexes. We
construct the connecting homomorphism

𝜕∗ : 𝐻𝑛 𝐾∗′′ → 𝐻𝑛−1 𝐾∗′

as follows: Let 𝑧′′ ∈ 𝑍 𝑛 𝐾∗′′ represent an element [𝑧′′ ] ∈ 𝐻𝑛 𝐾∗′′ . Since 𝑝 𝑛 is surjective we can
choose 𝑥 ∈ 𝐾𝑛 with 𝑝 𝑛 𝑥 = 𝑧′′ . For 𝜕𝑥 ∈ 𝐾𝑛−1 we observe

𝑝 𝑛−1 𝜕𝑥 = 𝜕 ′′ 𝑝 𝑛 𝑥 = 𝜕 ′′ 𝑧′′ = 0.

By exactness there is a unique 𝑦 ′ ∈ 𝐾𝑛−1

′ with 𝑖 𝑛−1 𝑦 ′ = 𝜕𝑥. Moreover,

𝑖 𝑛−2 𝜕 ′ 𝑦 ′ = 𝜕𝑖 𝑛−1 𝑦 ′ = 𝜕𝜕𝑥 = 0.

Since 𝑖 𝑛−2 is injective we have 𝜕 ′ 𝑦 ′ = 0, i.e., 𝑦 ′ ∈ 𝑍 𝑛−1 𝐾∗′ . Now put

𝜕∗ [𝑧′′ ] := [𝑦 ′ ].

Lemma 3.51. The conncecting homomorphism 𝜕∗ : 𝐻𝑛 𝐾∗′′ → 𝐻𝑛−1 𝐾∗′ is well defined, i.e.,
independent of the choices made in its construction.

Proof. There are two choices in the construction of the connecting homomorphism: that of the
preimage 𝑥 with 𝑝 𝑛 𝑥 = 𝑧′′ and that of the representing cycle 𝑧′′ itself.
As to the choice of 𝑥, let 𝑝 𝑛 𝑥 = 𝑝 𝑛 𝑥¯ = 𝑧′′ . For the corresponding elements 𝑦 ′ , 𝑦¯ ′ ∈ 𝐾𝑛−1
′ we
′ ′
have and 𝑖 𝑛−1 𝑦 = 𝜕𝑥 and 𝑖 𝑛−1 𝑦¯ = 𝜕 𝑥. ¯ Since 𝑝 𝑛 (𝑥 − 𝑥)¯ = 0 there exists an 𝜔 ∈ 𝐾𝑛′ with
′

𝑥 − 𝑥¯ = 𝑖 𝑛 (𝜔′ ). We compute

𝑖 𝑛−1 (𝑦 ′ − 𝑦¯ ′ ) = 𝜕 (𝑥 − 𝑥)
¯ = 𝜕 (𝑖 𝑛 𝜔′ ) = 𝑖 𝑛−1 𝜕𝜔′ ,

from which we conclude that 𝑦 − 𝑦¯ ′ = 𝜕𝜔′ and therefore [𝑦 ′ ] = [ 𝑦¯ ′ ].

108
 3.5. Homological algebra

As to the choice of the representing cycle 𝑧′′ , it suffices to show that if 𝑧′′ is a boundary then
[𝑦 ′ ] = 0. Let 𝑧′′ = 𝜕 ′′ 𝜁 ′′ be a boundary. Since 𝑝 𝑛+1 is onto we can choose 𝜉 ∈ 𝐾𝑛+1 with
𝑝 𝑛+1 𝜉 = 𝜁 ′′ . Then 𝑥 = 𝜕𝜉 is an admissible choice because

𝑝 𝑛 𝑥 = 𝑝 𝑛 𝜕𝜉 = 𝜕 ′′ 𝑝 𝑛+1 𝜉 = 𝜕 ′′ 𝜁 ′′ = 𝑧′′ .

Thus 𝜕𝑥 = 𝜕𝜕𝜉 = 0 and hence 𝑦 ′ = 0. 

It is easy to see that the connecting homomorphism is indeed a homomorphism. Now we are
ready to prove the Exactness Axiom.

𝑖∗ 𝑝∗
Proposition 3.52. If 0 → 𝐾∗′ −→ 𝐾∗ −→ 𝐾∗′′ → 0 is an exact sequence of complexes then
the long homology sequence

𝐻𝑛 (𝑖∗ )
··· / 𝐻𝑛+1 𝐾 ′′ 𝜕∗
/ 𝐻𝑛 𝐾 ′ / 𝐻𝑛 𝐾∗ 𝐻𝑛 ( 𝑝∗ )/ 𝐻𝑛 𝐾 ′′ 𝜕∗
/ 𝐻𝑛−1 𝐾 ′ / ···
∗ ∗ ∗ ∗

is also exact.

Proof. In view of Proposition 3.49 it remains to show ker 𝐻 (𝑖) = im 𝜕∗ and ker 𝜕∗ = im 𝐻 ( 𝑝).

a) im 𝜕∗ ⊂ ker 𝐻 (𝑖):
Using the notation of Definition 3.50 we compute

𝐻 (𝑖)𝜕∗ [𝑧′′ ] = 𝐻 (𝑖)𝜕∗ [ 𝑝𝑥] = 𝐻 (𝑖) [𝑖 −1 𝜕𝑥] = [𝜕𝑥] = 0 .

b) ker 𝐻 (𝑖) ⊂ im 𝜕∗ :
Let 𝐻 (𝑖) [𝑧′ ] = 0. Then we have [𝑖𝑧′ ] = 0 and therefore 𝑖𝑧′ = 𝜕𝑥. Put 𝑧′′ := 𝑝𝑥. Then
𝜕 ′′ 𝑧′′ = 𝜕 ′′ 𝑝𝑥 = 𝑝𝜕𝑥 = 𝑝𝑖𝑧′ = 0. Hence 𝑧′′ ∈ 𝑍 𝑛 𝐾∗ represents an element in homology. We
compute
𝜕∗ [𝑧′′ ] = 𝜕∗ [ 𝑝𝑥] = [𝑖 −1 𝜕𝑥] = [𝑧′ ] .
Hence [𝑧′ ] ∈ im 𝜕∗ .

c) im 𝐻 ( 𝑝) ⊂ ker 𝜕∗ :
This follows from
𝜕∗ 𝐻 ( 𝑝) [𝑧] = 𝜕∗ [ 𝑝𝑧] = [𝑖 −1 𝜕𝑧 ] = 0 .
|{z}
=0

d) ker 𝜕∗ ⊂ im 𝐻 ( 𝑝):
Let 𝜕∗ [𝑧′′ ] = 0. We write 𝑧′′ = 𝑝𝑥 and compute

0 = 𝜕∗ [𝑧′′ ] = 𝜕∗ [ 𝑝𝑥] = [𝑖 −1 𝜕𝑥] .

109
3. Homology Theory

This implies that 𝑖 −1 𝜕𝑥 is a boundary in 𝐾 ′ . Hence we have that 𝑖 −1 𝜕𝑥 = 𝜕 ′ 𝑥 ′ and therefore

𝜕 (𝑥 − 𝑖𝑥 ′ ) = 𝜕𝑥 − 𝑖𝜕 ′ 𝑥 ′ = 0 .

This finally leads to

𝐻 ( 𝑝) [𝑥 − 𝑖𝑥 ′ ] = [ 𝑝𝑥 − 𝑝𝑖𝑥 ′ ] = [ 𝑝𝑥] = [𝑧′′ ] ,
hence [𝑧′′ ] ∈ im 𝐻 ( 𝑝). 

Example 3.53. For 𝐾𝑛′ = 𝑆𝑛 ( 𝐴), 𝐾𝑛 = 𝑆𝑛 (𝑋) and

𝑆𝑛 (𝑋)
𝐾𝑛′′ = 𝑆𝑛 (𝑋, 𝐴) =
𝑆𝑛 ( 𝐴)
Proposition 3.52 now yields the Exactness Axiom for singular homology.

Example 3.54. Consider a triple of spaces (𝑋, 𝐴, 𝐵) with 𝐵 ⊂ 𝐴 ⊂ 𝑋 and set

𝐾𝑛′ = 𝑆𝑛 ( 𝐴)/𝑆𝑛 (𝐵) = 𝑆𝑛 ( 𝐴, 𝐵)

𝐾𝑛 = 𝑆𝑛 (𝑋)/𝑆𝑛 (𝐵) = 𝑆𝑛 (𝑋, 𝐵)
𝐾𝑛′′ = 𝑆𝑛 (𝑋)/𝑆𝑛 ( 𝐴) = 𝑆𝑛 (𝑋, 𝐴) .

Then the canonical sequence

0 / 𝐾′ / 𝐾𝑛 / 𝐾 ′′ /0
𝑛 𝑛

is again exact. By Proposition 3.52 we obtain the long exact homology sequence for a triple:

... / 𝐻 (𝑋, 𝐴)
❣ ❣❣❣ 𝑛+1
𝜕∗ ❣❣❣❣ ❣
❣❣ ❣❣❣❣❣❣
s❣❣❣ ❣
𝐻𝑛 ( 𝐴, 𝐵) / 𝐻𝑛 (𝑋, 𝐵) / 𝐻𝑛 (𝑋, 𝐴)
❣❣ ❣❣❣
𝜕∗ ❣❣❣❣❣
❣
❣❣❣❣❣❣❣❣
s❣❣
𝐻𝑛−1 ( 𝐴, 𝐵) / ...

Proposition 3.55. The long homology sequence is natural, i.e., if the diagram of chain maps

𝑖∗ 𝑝∗
0 / 𝐾′ / 𝐾∗ / 𝐾 ′′ /0
∗ ∗

𝜑∗′ 𝜑∗ 𝜑∗′′
 𝑗∗  𝑞∗ 
0 / 𝐿′ / 𝐿∗ / 𝐿 ′′ /0
∗ ∗

110
 3.5. Homological algebra

is commutative with exact rows then the diagram

𝐻𝑛+1 ( 𝑝∗ ) 𝐻𝑛 (𝑖∗ )
... / 𝐻𝑛+1 𝐾 ′′ 𝜕∗
/ 𝐻𝑛 𝐾 ′ / 𝐻𝑛 𝐾∗ 𝐻𝑛 ( 𝑝∗ )/ 𝐻𝑛 𝐾 ′′ 𝜕∗
/ ...
∗ ∗ ∗

𝐻𝑛+1 ( 𝜑∗′′ ) 𝐻𝑛 ( 𝜑∗′ ) 𝐻 𝑛 ( 𝜑∗ ) 𝐻𝑛 ( 𝜑∗′′ )

𝐻𝑛+1 (𝑞∗ )  𝜕∗  𝐻𝑛 ( 𝑗∗ )  𝐻𝑛 (𝑞∗ )  𝜕∗
... / 𝐻𝑛+1 𝐿 ′′ / 𝐻𝑛 𝐿 ′ / 𝐻𝑛 𝐿 ∗ / 𝐻𝑛 𝐿 ′′ / ...
∗ ∗ ∗

is commutative as well.

Proof. By assumption we have 𝜑 ◦ 𝑖 = 𝑗 ◦ 𝜑′ and 𝜑′′ ◦ 𝑝 = 𝑞 ◦ 𝜑, hence

𝐻𝑛 (𝜑) ◦ 𝐻𝑛 (𝑖) = 𝐻𝑛 ( 𝑗) ◦ 𝐻𝑛 (𝜑′ ),

𝐻𝑛 (𝜑′′ ) ◦ 𝐻𝑛 ( 𝑝) = 𝐻𝑛 (𝑞) ◦ 𝐻𝑛 (𝜑).

We are left to show that the diagram

𝜕∗
𝐻𝑛 𝐾 ′′ / 𝐻𝑛−1 𝐾 ′

𝐻𝑛 ( 𝜑 ′′ ) 𝐻𝑛−1 ( 𝜑 ′ )
 𝜕∗ 
𝐻𝑛 𝐿 ′′ / 𝐻𝑛 𝐿 ′

commutes. We calculate

𝐻𝑛−1 (𝜑′ )𝜕∗ [ 𝑝𝑥] = 𝐻𝑛−1 (𝜑′ ) [𝑖 −1 𝜕𝑥] = [𝜑′𝑖 −1 𝜕𝑥] = [ 𝑗 −1 𝜑𝜕𝑥]
= [ 𝑗 −1 𝜕𝜑𝑥] = 𝜕∗ [ 𝑝𝜑𝑥] = 𝜕∗ [𝜑′′ 𝑝𝑥] = 𝜕∗ 𝐻𝑛 (𝜑′′ ) [ 𝑝𝑥] ,

which proves the assertion. 

Example 3.56. Let 𝑓 ∈ 𝐶 ((𝑋, 𝐴), (𝑌 , 𝐵)). For 𝐾𝑛′ = 𝑆𝑛 ( 𝐴), 𝐾𝑛 = 𝑆𝑛 (𝑋), 𝐾𝑛′′ = 𝑆𝑛 (𝑋, 𝐴)
and 𝜑′𝑛 = 𝑆𝑛 ( 𝑓 | 𝐴), 𝜑 𝑛 = 𝑆𝑛 ( 𝑓 ), and 𝜑′′𝑛 the induced homomorphism on 𝑆𝑛 (𝑋, 𝐴) we recover
Lemma 3.13.

Definition 3.57. Let 𝜑, 𝜓 : 𝐾 → 𝐾 ′ be chain maps. A chain homotopy between 𝜑 and 𝜓 is a

sequence of homomorphisms
′
ℎ𝑛 : 𝐾𝑛 → 𝐾𝑛+1
such that
′
𝜑 𝑛 − 𝜓 𝑛 = 𝜕𝑛+1 ℎ𝑛 + ℎ𝑛−1 𝜕𝑛 ,
for all 𝑛 ∈ Z. The maps 𝜑 and 𝜓 are called homotopic if such a homotopy exists. In this case
we write 𝜑 ≃ 𝜓.

111
3. Homology Theory

Proposition 3.58. (i) The relation “≃” is an equivalence relation on the set of all chain
maps 𝐾 → 𝐾 . ′

(ii) If 𝜑 ≃ 𝜓 : 𝐾 → 𝐾 ′ and 𝜑′ ≃ 𝜓 ′ : 𝐾 ′ → 𝐾 ′′ then 𝜑′ 𝜑 ≃ 𝜓 ′ 𝜓 : 𝐾 → 𝐾 ′′ .

Proof. (i) First we show that “≃” is an equivalence relation.

a) To show that 𝜑 ≃ 𝜑 choose ℎ = 0.
b) The statement 𝜑 ≃ 𝜓 ⇒ 𝜓 ≃ 𝜑 follows from replacing ℎ by −ℎ.
c) If 𝜑 ≃ 𝜓 and 𝜓 ≃ 𝜒 then 𝜑 ≃ 𝜒 .
ℎ 𝑘 ℎ+𝑘

(ii) Assume that 𝜑 ≃ 𝜓 and 𝜑′ ≃′ 𝜓 ′ . Now we calculate

ℎ ℎ

𝜑′ 𝜑 − 𝜑′ 𝜓 = 𝜑′ (𝜕 ′ ℎ + ℎ𝜕) = 𝜕𝜑′ ℎ + 𝜑′ ℎ𝜕 ,

hence 𝜑′ 𝜑 ≃′ 𝜑′ 𝜓. Similarly we see

𝜑 ℎ

𝜑′ 𝜓 − 𝜓 ′ 𝜓 = (𝜕 ′ ℎ′ + ℎ′ 𝜕)𝜓 = 𝜕 ′ ℎ′ 𝜓 + ℎ′ 𝜓𝜕 ,

hence 𝜑′ 𝜓 ≃
′
𝜓 ′ 𝜓. Combining both equivalences we get the desired result, namely
ℎ 𝜓

𝜑′ 𝜑 ≃ 𝜑′ 𝜓 ≃ 𝜓 ′ 𝜓 .

Homotopic chain maps induce the same homomorphisms on homology:

Proposition 3.59. If 𝜑 ≃ 𝜓 : 𝐾 → 𝐾 ′ then 𝐻𝑛 (𝜑) = 𝐻𝑛 (𝜓) : 𝐻𝑛 𝐾 → 𝐻𝑛 𝐾 ′ for all 𝑛.

Proof. Let 𝑧 ∈ 𝑍 𝑛 𝐾. We compute

𝐻𝑛 (𝜑) [𝑧] = [𝜑𝑧] = [𝜓𝑧 + 𝜕 ′ ℎ𝑧 + ℎ𝜕𝑧] = [𝜓𝑧] = 𝐻𝑛 (𝜓) [𝑧]

because 𝜕𝑧 = 0 and [𝜕 ′ ℎ𝑧] = 0. 

Definition 3.60. A chain map 𝜑 : 𝐾 → 𝐾 ′ is called a homotopy equivalence if there exists a

chain map 𝜓 : 𝐾 ′ → 𝐾 such that 𝜓𝜑 ≃ id𝐾 and 𝜑𝜓 ≃ id𝐾 ′ . If such a homotopy equivalence
exists we write 𝐾 ≃ 𝐾 ′ .

112
 3.6. Proof of the homotopy axiom

Corollary 3.61. If 𝐾 ≃ 𝐾 ′ then 𝐻𝑛 𝐾  𝐻𝑛 𝐾 ′ for all 𝑛.

𝜑 𝐻𝑛 ( 𝜑)

3.6. Proof of the homotopy axiom

We put 𝐼 := [0, 1] and define the affine linear map
𝑗
𝑇𝑛 : Δ𝑛+1 → Δ𝑛 × 𝐼
for 𝑗 = 0, . . . , 𝑛 by (
𝑗 (𝑒 𝑘 , 0), 𝑘 ≤ 𝑗,
𝑇𝑛 (𝑒 𝑘 ) =
(𝑒 𝑘−1 , 1), 𝑘 > 𝑗.
In the case 𝑛 = 0 we have
𝑇00 : Δ1 → Δ0 × 𝐼 = {𝑒0 } × 𝐼 ,
which is visualized in Figure 75.
b
(𝑒0 , 0)
𝑇00

b b b
(𝑒0 , 1)
𝑒0 𝑒1

Figure 75. 𝑇00

In the case 𝑛 = 1 we find

𝑒2 (𝑒0 , 1) (𝑒1 , 1)
b b b

𝑇10

b b b b

𝑒0 𝑒1 (𝑒0 , 0) (𝑒1 , 0)

𝑒2 (𝑒0 , 1) (𝑒1 , 1)
b b b

𝑇11

b b b b

𝑒0 𝑒1 (𝑒0 , 0) (𝑒1 , 0)

Figure 76. 𝑇10 and 𝑇01

113
3. Homology Theory

Lemma 3.62. The composition of the operators 𝑇 and the affine linear map 𝐹 (defined in (3.1)
on page 80) yields:
𝑗+1 𝑖 𝑗
𝑇𝑛 ◦ 𝐹𝑛+1 = (𝐹𝑛𝑖 × id) ◦ 𝑇𝑛−1 ( 𝑗 ≥ 𝑖),
𝑗 𝑖+1 𝑗
𝑇𝑛 ◦ 𝐹𝑛+1 = (𝐹𝑛𝑖 × id) ◦ 𝑇𝑛−1 ( 𝑗 < 𝑖),
𝑇𝑛𝑖 ◦ 𝐹𝑛+1
𝑖
= 𝑇𝑛𝑖−1 ◦ 𝐹𝑛+1
𝑖
(1 ≤ 𝑖 ≤ 𝑛),
𝑇𝑛0 ◦ 𝐹𝑛+1
0
= 𝑖1 ,
𝑇𝑛𝑛 ◦ 𝐹𝑛+1
𝑛+1
= 𝑖0 ,

with 𝑖 𝑡 : Δ𝑛 → Δ𝑛 × 𝐼 being the inclusion map 𝑥 ↦→ (𝑥, 𝑡).

Proof. Let us prove the first formula where 𝑗 ≥ 𝑖. If 𝑘 < 𝑖 then

𝑗+1 𝑖 𝑗+1 𝑗
𝑇𝑛 ◦ 𝐹𝑛+1 (𝑒 𝑘 ) = 𝑇𝑛 (𝑒 𝑘 ) = (𝑒 𝑘 , 0) = (𝐹𝑛𝑖 × id) (𝑒 𝑘 , 0) = (𝐹𝑛𝑖 × id) ◦ 𝑇𝑛−1 (𝑒 𝑘 ).

If 𝑖 ≤ 𝑘 ≤ 𝑗 then
𝑗+1 𝑖 𝑗+1 𝑗
𝑇𝑛 ◦ 𝐹𝑛+1 (𝑒 𝑘 ) = 𝑇𝑛 (𝑒 𝑘+1 ) = (𝑒 𝑘+1 , 0) = (𝐹𝑛𝑖 × id) (𝑒 𝑘 , 0) = (𝐹𝑛𝑖 × id) ◦ 𝑇𝑛−1 (𝑒 𝑘 ).

If 𝑘 > 𝑗 then
𝑗+1 𝑖 𝑗+1 𝑗
𝑇𝑛 ◦ 𝐹𝑛+1 (𝑒 𝑘 ) = 𝑇𝑛 (𝑒 𝑘+1 ) = (𝑒 𝑘 , 1) = (𝐹𝑛𝑖 × id) (𝑒 𝑘−1 , 1) = (𝐹𝑛𝑖 × id) ◦ 𝑇𝑛−1 (𝑒 𝑘 ).

The proofs of the other formulas are similar exercises in index shifting. 

𝑗
Now let 𝜎 : Δ𝑛 → 𝑋 be a singular 𝑛-simplex. Note that (𝜎 × id) ◦ 𝑇𝑛 ∈ 𝐶 (Δ𝑛+1 , 𝑋 × 𝐼). We
define 𝑃𝜎 ∈ 𝑆𝑛+1 (𝑋 × 𝐼) by

Õ
𝑛
𝑗
𝑃𝜎 := (−1) 𝑗 (𝜎 × id) ◦ 𝑇𝑛 .
𝑗=0

We extend 𝑃 linearly to chains and get a linear map

𝑃 : 𝑆𝑛 (𝑋) → 𝑆𝑛+1 (𝑋 × 𝐼).

This homomorphism descends to relative chains

𝑃 : 𝑆𝑛 (𝑋, 𝐴) → 𝑆𝑛+1 (𝑋 × 𝐼, 𝐴 × 𝐼).

The operator 𝑃 is called prism operator.

114
 3.6. Proof of the homotopy axiom

Lemma 3.63. Let (𝑋, 𝐴) be a pair of spaces. The operator 𝑃 is a chain homotopy for

𝑆(𝑖0𝑋 ), 𝑆(𝑖 1𝑋 ) : 𝑆(𝑋, 𝐴) → 𝑆(𝑋 × 𝐼, 𝐴 × 𝐼) ,

where 𝑖𝑡𝑋 : 𝑋 → 𝑋 × 𝐼 denotes the inclusion map 𝑥 ↦→ (𝑥, 𝑡).

Proof. We have to show that

𝑆(𝑖 0𝑋 ) − 𝑆(𝑖1𝑋 ) = 𝑃𝜕 + 𝜕𝑃. (3.10)
Let 𝜎 : Δ𝑛 → 𝑋 be a singular 𝑛-simplex. We compute, using Lemma 3.62:
Õ
𝑛
𝑃𝜕𝜎 = 𝑃 (−1) 𝑖 (𝜎 ◦ 𝐹𝑛𝑖 )
𝑖=0
Õ
𝑛 Õ
𝑛−1
𝑖 𝑗
= (−1) (−1) 𝑗 (𝜎 ◦ 𝐹𝑛𝑖 × id) ◦ 𝑇𝑛−1
Õ
𝑖=0 𝑗=0
𝑗
= (−1) 𝑖+ 𝑗 (𝜎 × id) ◦ (𝐹𝑛𝑖 × id) ◦ 𝑇𝑛−1
Õ
0≤ 𝑗<𝑖≤𝑛
𝑗
+ (−1) 𝑖+ 𝑗 (𝜎 × id) ◦ (𝐹𝑛𝑖 × id) ◦ 𝑇𝑛−1
Õ
0≤𝑖≤ 𝑗 ≤𝑛−1
𝑗
= − (−1) 𝑖+ 𝑗+1 (𝜎 × id) ◦ 𝑇𝑛 ◦ 𝐹𝑛+1
𝑖+1

Õ
0≤ 𝑗<𝑖≤𝑛
𝑗+1
− (−1) 𝑖+ 𝑗+1 (𝜎 × id) ◦ 𝑇𝑛 𝑖
◦ 𝐹𝑛+1 .
0≤𝑖≤ 𝑗 ≤𝑛−1

On the other hand, we find

Õ
𝑛
𝑗
𝜕𝑃𝜎 = 𝜕 (−1) 𝑗 (𝜎 × id) ◦ 𝑇𝑛
𝑗=0
Õ
𝑛 Õ
𝑛+1
𝑗 𝑗
= (−1) (−1) 𝑖 (𝜎 × id) ◦ 𝑇𝑛 ◦ 𝐹𝑛+1
𝑖

Õ
𝑗=0 𝑖=0
𝑗
= (−1) 𝑖+ 𝑗 (𝜎 × id) ◦ 𝑇𝑛 ◦ 𝐹𝑛+1
𝑖
(𝑖 < 𝑗)
0≤𝑖< 𝑗 ≤𝑛
Õ𝑛
+ (𝜎 × id) ◦ 𝑇𝑛𝑖 ◦ 𝐹𝑛+1
𝑖
(𝑖 = 𝑗)
𝑖=0
Õ
𝑛+1
− (𝜎 × id) ◦ 𝑇𝑛𝑖−1 ◦ 𝐹𝑛+1
𝑖
(𝑖 = 𝑗 + 1)
𝑖=1
Õ
𝑗
+ (−1) 𝑖+ 𝑗 (𝜎 × id) ◦ 𝑇𝑛 ◦ 𝐹𝑛+1
𝑖
(𝑖 > 𝑗 + 1)
1≤ 𝑗+1<𝑖≤𝑛+1

115
3. Homology Theory

Õ ′ +1 𝑗 ′ +1
= (−1) 𝑖+ 𝑗 (𝜎 × id) ◦ 𝑇𝑛 𝑖
◦ 𝐹𝑛+1
0≤𝑖≤ 𝑗 ′ ≤𝑛−1
+(𝜎 Õ
× id) ◦ 𝑖1 − (𝜎 × id) ◦ 𝑖0
′ 𝑗 ′
+ (−1) 𝑖 + 𝑗+1 (𝜎 × id) ◦ 𝑇𝑛 ◦ 𝐹𝑛+1
𝑖 +1

0≤ 𝑗<𝑖′ ≤𝑛

In the last step we changed the summation indices to 𝑗 ′ = 𝑗 − 1 and 𝑖 ′ = 𝑖 − 1. We see that
𝑃𝜕𝜎 + 𝜕𝑃𝜎 = (𝜎 × id) ◦ 𝑖1 − (𝜎 × id) ◦ 𝑖0
= 𝑖 1𝑋 ◦ 𝜎 − 𝑖 0𝑋 ◦ 𝜎
= 𝑆𝑛 (𝑖 1𝑋 )𝜎 − 𝑆𝑛 (𝑖 0𝑋 )𝜎
which proves the lemma. 

Proposition 3.64. If 𝑓 ≃ 𝑔 : (𝑋, 𝐴) → (𝑌 , 𝐵) then we have

𝑆( 𝑓 ) ≃ 𝑆(𝑔) : 𝑆(𝑋, 𝐴) → 𝑆(𝑌 , 𝐵)

Proof. Let 𝐹 be a homotopy for 𝑓 and 𝑔, i.e., 𝐹 ∈ 𝐶 ((𝑋 × 𝐼, 𝐴 × 𝐼), (𝑌 , 𝐵)) with
𝑓 = 𝐹 ◦ 𝑖 1𝑋 , 𝑔 = 𝐹 ◦ 𝑖 0𝑋 .
Then by (3.10) we get
𝑆( 𝑓 ) − 𝑆(𝑔) = 𝑆(𝐹)𝑆(𝑖1𝑋 ) − 𝑆(𝐹)𝑆(𝑖 0𝑋 ) = 𝑆(𝐹)𝑃𝜕 + 𝑆(𝐹)𝜕𝑃 = 𝑆(𝐹)𝑃𝜕 + 𝜕𝑆(𝐹)𝑃.
Hence 𝑆(𝐹)𝑃 is a chain homotopy for 𝑆( 𝑓 ) and 𝑆(𝑔). 

Corollary 3.65. The homotopy axiom holds for singular homology.

3.7. Proof of the excision axiom

We want to show that the inclusion (𝑋 \ 𝑈, 𝐴 \ 𝑈) ↩→ (𝑋, 𝐴) induces an isomorphism
𝐻𝑛 (𝑋 \ 𝑈, 𝐴 \ 𝑈)  𝐻𝑛 (𝑋, 𝐴)
˚
provided 𝑈¯ ⊂ 𝐴. 𝑋

𝐴 𝑈

Figure 77. Setup for the excision axiom

116
 3.7. Proof of the excision axiom

Í
Let 𝑋 be a topological space and let U = {𝑈𝑖 }𝑖∈ 𝐼 be an open cover of 𝑋. A chain
𝜎 = 𝑗 𝛼 𝑗 𝜎 𝑗 ∈ 𝑆𝑛 (𝑋) is called U-small iff for each 𝑗 there exists an 𝑖 such that 𝜎 𝑗 (Δ𝑛 ) ⊂ 𝑈𝑖 .
We denote
𝑆𝑛U (𝑋) := {𝜎 ∈ 𝑆𝑛 (𝑋) | 𝜎 is U-small}
!
Ê ⊕𝑖 𝑆𝑛 ( 𝑗𝑖 )
= im 𝑆𝑛 (𝑈𝑖 ) −−−−−−−−−→ 𝑆𝑛 (𝑋)
𝑖∈ 𝐼

with 𝑗 𝑖 : 𝑈𝑖 → 𝑋 the inclusion map. For 𝐴 ⊂ 𝑋 we put

𝑆𝑛U (𝑋)
𝑆𝑛U (𝑋, 𝐴) := .
𝑆𝑛U ( 𝐴)

Theorem 3.66 (Small chain theorem). The inclusion 𝑆𝑛U (𝑋, 𝐴) → 𝑆𝑛 (𝑋, 𝐴) induces an iso-
morphism in homology.

Before coming to the proof we show that the small chain theorem implies the excision axiom.
˚ Now set 𝑈1 := 𝐴˚
To this extent let (𝑋, 𝐴) be a pair of spaces and let 𝑈 ⊂ 𝐴 be such that 𝑈¯ ⊂ 𝐴.
¯
and 𝑈2 := 𝑋 \ 𝑈. Then U = {𝑈1 , 𝑈2 } forms an open cover of the space 𝑋. We compute
𝑆𝑛U (𝑋)
𝑆𝑛U (𝑋, 𝐴) =
𝑆𝑛U ( 𝐴)
˚ + 𝑆𝑛 (𝑋 \ 𝑈)
𝑆𝑛 ( 𝐴) ¯
=
˚ + 𝑆𝑛 ( 𝐴 \ 𝑈)
𝑆𝑛 ( 𝐴) ¯
¯
𝑆𝑛 (𝑋 \ 𝑈)
=
˚ + 𝑆𝑛 ( 𝐴 \ 𝑈))
(𝑆𝑛 ( 𝐴) ¯ ∩ 𝑆𝑛 (𝑋 \ 𝑈)
¯
¯
𝑆𝑛 (𝑋 \ 𝑈)
= .
¯
𝑆𝑛 ( 𝐴 \ 𝑈)
Similarly we get
𝑆𝑛 ( 𝐴˚ \ 𝑈) + 𝑆𝑛 (𝑋 \ 𝑈)
¯ ¯
𝑆𝑛 (𝑋 \ 𝑈)
𝑆𝑛U (𝑋 \ 𝑈, 𝐴 \ 𝑈) = = .
𝑆𝑛 ( 𝐴˚ \ 𝑈) + 𝑆𝑛 ( 𝐴 \ 𝑈)
¯ ¯
𝑆𝑛 ( 𝐴 \ 𝑈)
Thus
𝑆𝑛U (𝑋, 𝐴) = 𝑆𝑛U (𝑋 \ 𝑈, 𝐴 \ 𝑈).
In the following commutative diagram all arrows are induced by inclusions.
U (𝑋 \ 𝑈, 𝐴 \ 𝑈) = / 𝑆 U (𝑋, 𝐴)
𝑆𝑚 𝑛

 
𝑆𝑚 (𝑋 \ 𝑈, 𝐴 \ 𝑈) / 𝑆 𝑛 (𝑋, 𝐴)

117
3. Homology Theory

By the small chain theorem the vertical arrows induce isomorphisms on homology and the
excision theorem is proved.
It remains to prove Theorem 3.66. We define the barycenter of Δ𝑛 by

1 Õ
𝑛
𝐵𝑛 := 𝑒𝑗.
𝑛 + 1 𝑗=0

Example 3.67. For example, 𝑒2

𝐵1 𝐵2
𝐵0 = 𝑒 0 ,
1 𝑒0 𝑒1 𝑒0 𝑒1
𝐵1 = (𝑒0 + 𝑒1 ),
2
1 Figure 78. Barycenters in 1 and 2 di-
𝐵2 = (𝑒0 + 𝑒1 + 𝑒2 ). mensions
3

For an affine map 𝜎 : Δ𝑛 → Δ𝑛+1 we define the affine map 𝐶𝑛 𝜎 : Δ𝑛+1 → Δ𝑛+1 by
(
𝐵𝑛+1 , 𝑘 = 0,
(𝐶𝑛 𝜎) (𝑒 𝑘 ) =
𝜎 (𝑒 𝑘−1 ), 𝑘 ≥ 1.

Example 3.68. For the example 𝜎 = 𝐹20 : Δ1 → Δ2 see Figure 79.

Now we set
 Õ
𝑛 𝑛
𝑆aff
𝑘 (Δ ) := 𝜎 ∈ 𝑆 𝑘 (Δ ) 𝜎 = 𝛼 𝑗 𝜎 𝑗 and each 𝜎 𝑗 is affine .
𝑗

By linear extension we get a homomorphism

𝑛+1 𝑛+1
𝐶𝑛 : 𝑆aff
𝑛 (Δ ) → 𝑆aff
𝑛+1 (Δ ).

Lemma 3.69. The homomorphism 𝐶𝑛 has the following properties:

Í Í
(i) 𝜕𝐶0 (𝑐) = 𝑐 − ( 𝑗 𝛼 𝑗 )𝐵0 where 𝑐 = 𝑗 𝛼 𝑗 𝜎 𝑗 ;

(ii) 𝜕𝐶𝑛 (𝑐) = 𝑐 − 𝐶𝑛−1 𝜕𝑐 for 𝑛 ≥ 1.

118
 3.7. Proof of the excision axiom

𝑒2
b

𝐺
b

b
𝑒0 𝑒1

𝑒0 𝑒1 𝐶1 𝐺
𝑒2
b

b b
𝑒0 𝑒1

Figure 79. 𝐶1

Proof. (i) It suffices to show the assertion for an affine simplex 𝑐. We then find

(𝐶0 𝑐) (𝑒0 ) = 𝐵0 , (𝐶0 𝑐) (𝑒1 ) = 𝑐(𝑒0 )

and hence

𝜕𝐶0 (𝑐) (𝑒0 ) = (𝐶0 (𝑐)) (𝑒1 ) − (𝐶0 (𝑐)) (𝑒0 ) = 𝑐(𝑒0 ) − 𝐵0 = (𝑐 − 1𝐵0 ) (𝑒0 )

as desired.

(ii) is left as an exercise. 

Lemma 3.70. To each topological space 𝑋 and each 𝑛 ∈ N0 we can associate homomorphisms

Sd𝑛 : 𝑆𝑛 (𝑋) → 𝑆𝑛 (𝑋),

𝑄 𝑛 : 𝑆𝑛 (𝑋) → 𝑆𝑛+1 (𝑋),

such that

(i) Sd∗ is a chain map, i.e., 𝜕 ◦ Sd𝑛 = Sd𝑛−1 ◦ 𝜕;

(ii) 𝑄 ∗ is a chain homotopy between id and Sd∗ , i.e., id −Sd𝑛 = 𝜕 ◦ 𝑄 𝑛 + 𝑄 𝑛−1 ◦ 𝜕;

119
3. Homology Theory

(iii) Sd∗ and 𝑄 ∗ are natural, i.e., for every 𝑓 ∈ 𝐶 (𝑋,𝑌 ) the following diagrams commute:

Sd𝑛 𝑄𝑛
𝑆𝑛 (𝑋) / 𝑆 𝑛 (𝑋) 𝑆𝑛 (𝑋) / 𝑆 𝑛+1 (𝑋)

𝑆𝑛 ( 𝑓 ) 𝑆𝑛 ( 𝑓 ) 𝑆𝑛 ( 𝑓 ) 𝑆𝑛+1 ( 𝑓 )
 Sd𝑛   𝑄𝑛 
𝑆𝑛 (𝑌 ) / 𝑆 𝑛 (𝑌 ) 𝑆𝑛 (𝑌 ) / 𝑆 𝑛+1 (𝑌 )

(iv) If the map 𝜎 : Δ𝑛 → Δ𝑛 is affine then each simplex 𝜎 𝑗 occuring in Sd𝑛 (𝜎) or in 𝑄 𝑛 (𝜎)
is affine and
𝑛
diam(𝜎 𝑗 ) ≤ diam(𝜎).
𝑛+1

Proof. The construction of Sd𝑛 and 𝑄 𝑛 will be done recursively, the proof is by induction over
𝑛. We start by considering the case 𝑛 = 0. We put Sd0 := id : 𝑆0 (𝑋) → 𝑆0 (𝑋) and 𝑄 0 := 0. It
is obvious that the four assertions hold.
In the case 𝑛 ≥ 1 we assume that Sd𝑛−1 and 𝑄 𝑛−1 are already defined and we set for a singular
simplex 𝜎 : Δ𝑛 → 𝑋:

Sd𝑛 (𝜎) := 𝑆𝑛 (𝜎) (𝐶𝑛−1 (Sd𝑛−1 (𝜕 (idΔ𝑛 ) ) ) )

|{z}

| {z }
∈𝑆𝑛aff (Δ𝑛 )

aff (Δ𝑛 )
∈𝑆𝑛−1
| {z }
aff (Δ𝑛 )
∈𝑆𝑛−1
| {z }
∈𝑆𝑛aff (Δ𝑛 )

and
𝑄 𝑛 (𝜎) = 𝑆𝑛+1 (𝜎) (𝐶𝑛 (idΔ𝑛 −Sd𝑛 (idΔ𝑛 ) − 𝑄 𝑛−1 𝜕 idΔ𝑛 ) ).
| {z }

| {z }
∈𝑆𝑛aff (Δ𝑛 )

aff (Δ𝑛 )
∈𝑆𝑛+1

We have to verify assertions (i)-(iv). We check (iii):

Sd𝑛 (𝑆𝑛 ( 𝑓 ) (𝜎)) = 𝑆𝑛 (𝑆𝑛 ( 𝑓 ) (𝜎)) (𝐶𝑛−1 (Sd𝑛−1 (𝜕 (idΔ𝑛 ))))

= 𝑆𝑛 ( 𝑓 ◦ 𝜎) (𝐶𝑛−1 (Sd𝑛−1 (𝜕 (idΔ𝑛 ))))
= 𝑆𝑛 ( 𝑓 ) ◦ 𝑆𝑛 (𝜎) (𝐶𝑛−1 (Sd𝑛−1 (𝜕 (idΔ𝑛 ))))
= 𝑆𝑛 ( 𝑓 ) (Sd𝑛 (𝜎)).

The computation for 𝑄 𝑛 is similar.

120
 3.7. Proof of the excision axiom

Next we check (i): First we consider the case that 𝑋 = Δ𝑛 and 𝜎 = idΔ𝑛 .
𝜕 (Sd𝑛 (idΔ𝑛 )) = 𝜕 (𝐶𝑛−1 (Sd𝑛−1 (𝜕 idΔ𝑛 )))
L. 3.69
= Sd𝑛−1 (𝜕 idΔ𝑛 ) − 𝐶𝑛−2 𝜕 (Sd𝑛−1 (𝜕 idΔ𝑛 ))
ind. hyp.
|{z}
= Sd𝑛−1 (𝜕 idΔ𝑛 ) − 𝐶𝑛−2 Sd𝑛−2 𝜕𝜕 idΔ𝑛
=0
= Sd𝑛−1 (𝜕 idΔ𝑛 ).
For a general simplex 𝜎 : Δ𝑛 → 𝑋 we then find
𝜕 (Sd𝑛 (𝜎)) = 𝜕 (𝑆𝑛 (𝜎) (Sd𝑛 (idΔ𝑛 )))
= 𝑆𝑛 (𝜎) (𝜕 (Sd𝑛 (idΔ𝑛 )))
= 𝑆𝑛 (𝜎) (Sd𝑛−1 (𝜕 idΔ𝑛 ))
(𝑖𝑖𝑖)
= Sd𝑛−1 (𝑆𝑛 (𝜎) (𝜕 (idΔ𝑛 )))
= Sd𝑛−1 (𝜕𝑆𝑛 (𝜎) (idΔ𝑛 ))
= Sd𝑛−1 (𝜕𝜎).
Now we check (ii): Again we first consider the case 𝑋 = Δ𝑛 and 𝜎 = idΔ𝑛 .
𝜕𝑄 (idΔ𝑛 ) = 𝜕𝐶𝑛 (idΔ𝑛 −Sd𝑛 (idΔ𝑛 ) − 𝑄 𝑛−1 𝜕 idΔ𝑛 )
L. 3.69
 
= idΔ𝑛 −Sd𝑛 (idΔ𝑛 ) − 𝑄 𝑛−1 𝜕 idΔ𝑛
−𝐶𝑛−1 𝜕 idΔ𝑛 −Sd𝑛 (idΔ𝑛 ) − 𝑄 𝑛−1 𝜕 idΔ𝑛
ind. hyp.

= idΔ𝑛 −Sd𝑛 (idΔ𝑛 ) − 𝑄 𝑛−1 𝜕 idΔ𝑛
−𝐶𝑛−1 𝜕 idΔ𝑛 −𝜕Sd𝑛 (idΔ𝑛 ) − (𝜕 idΔ𝑛 −Sd𝑛−1 (𝜕 idΔ𝑛 )

|{z}
+ 𝑄 𝑛−2 ( 𝜕𝜕 idΔ𝑛 ))
=0
(𝑖)
= idΔ𝑛 −Sd𝑛 (idΔ𝑛 ) − 𝑄 𝑛−1 𝜕 idΔ𝑛 .
The passage to general 𝜎 can now be done as before.
Finally we check (iv): It is clear from the recursive definition of Sd𝑛 and of 𝑄 𝑛 that each simplex
𝜎 𝑗 occuring in Sd𝑛 (𝜎) or in 𝑄 𝑛 (𝜎) is again affine. The diameter of an affine simplex is the
maximal distance of any two of its vertices. We distinguish two cases:
1. The vertices 𝑝, 𝑞 of 𝜎 𝑗 of maximal distance lie on 𝜕𝜎.

𝑝 𝑞
Figure 80. Vertices of maximal distance

121
3. Homology Theory

Then we find by the induction hypothesis

𝑛−1 𝑛
𝑑 ( 𝑝, 𝑞) ≤ diam(face of 𝜎) < diam(𝜎) .
𝑛 𝑛+1

2. One vertex is 𝐵𝑛 :

𝐵𝑛
𝑝𝑖
Figure 81. Barycenter is a vertex
Then we find

1 Õ
𝑛
𝑑 ( 𝑝 𝑖 , 𝐵𝑛 ) = 𝑝 𝑖 − 𝑝𝑗
𝑛 + 1 𝑗=0

1 Õ
𝑛
= ( 𝑝𝑖 − 𝑝 𝑗 )
𝑛 + 1 𝑗=0

1 Õ
𝑛
≤ | 𝑝𝑖 − 𝑝 𝑗 |
𝑛 + 1 𝑗=0 | {z }
≤diam( 𝜎) and =0 for 𝑗=𝑖
𝑛
≤ diam(𝜎) . 
𝑛+1

Conclusion 3.71. We have that Sd ≃ id and therefore

Sd − Sd2 = Sd ◦ 𝜕 ◦ 𝑄 + Sd ◦ 𝑄 ◦ 𝜕
= 𝜕 ◦ Sd ◦ 𝑄 + Sd ◦ 𝑄 ◦ 𝜕.

We conclude that Sd2 ≃ Sd ≃ id. Iterating this procedure we find that Sd𝑟 ≃ id for all 𝑟 ∈ N.
Hence there exist homomorphisms 𝑄 𝑛(𝑟 ) : 𝑆𝑛 (𝑋) → 𝑆𝑛+1 (𝑋) with

Sd𝑟𝑛 − id = 𝜕 ◦ 𝑄 𝑛(𝑟 ) + 𝑄 𝑛−1

(𝑟 )
◦𝜕.

For 𝜎 affine, we have for every 𝜎 𝑗 occuring in Sd𝑟 (𝜎) that

 𝑛 𝑟
diam(𝜎 𝑗 ) ≤ diam(𝜎)
𝑛+1

122
 3.7. Proof of the excision axiom

Lemma 3.72. For 𝜎 : Δ𝑛 → 𝑋 continuous there exists a 𝜀 > 0 such that all 𝜀-balls ∩Δ𝑛 are
completely contained in 𝜎 −1 (𝑈𝑖 ) for some 𝑈𝑖 ∈ U.

Proof. Assume the assertion were false. Then for 𝜀 𝑘 = 1/𝑘 there exists a point 𝑝 𝑘 ∈ Δ𝑛 such
that
𝐵 1 ( 𝑝 𝑘 ) = {𝑥 ∈ Δ𝑛 | |𝑥 − 𝑝 𝑘 | < 𝜀 𝑘 }
𝑘

is not contained in any of the 𝜎 −1 (𝑈𝑖 ). After passing to a subsequence we have that 𝑝 𝑘 → 𝑝 ∈ Δ𝑛
by compactness of Δ𝑛 . Choose 𝑖0 with 𝑝 ∈ 𝜎 −1 (𝑈𝑖0 ). Since 𝜎 −1 (𝑈𝑖0 ) is open there exists a 𝛿 > 0
such that 𝐵 𝛿 ( 𝑝) ⊂ 𝜎 −1 (𝑈𝑖0 ). Now choose 𝑘 so large, that | 𝑝 𝑘 − 𝑝| < 𝛿/2 and 𝜀 𝑘 = 𝑘1 < 𝛿/2.
We then find
𝐵 1 ( 𝑝 𝑘 ) ⊂ 𝐵 𝛿 ( 𝑝) ⊂ 𝜎 −1 (𝑈𝑖0 ),
𝑘

a contradiction. 

Corollary 3.73. Assume that 𝜎 : Δ𝑛 → 𝑋 is continuous. Then there exists an 𝑟 (𝜎) ∈ N such
that every simplex 𝜎 𝑗 occuring in Sd𝑟 (𝜎) or in 𝑄 (𝑟 ) (𝜎) for 𝑟 ≥ 𝑟 (𝜎) is completely contained
in one of the 𝑈𝑖 , i.e., Sd𝑟 (𝜎), 𝑄 (𝑟 ) (𝜎) ∈ 𝑆𝑛U (𝑋).

Finally we can prove the small chain theorem.

Proof of Theorem 3.66. a) First we consider the case 𝐴 = ∅.

i) We show injectivity: Assume that 𝑧 ∈ 𝑍 𝑛U (𝑋) with 𝐻𝑛 ( 𝑗) ([𝑧] 𝐻𝑛U ) = 0. Then there exists
an 𝑥 ∈ 𝑆𝑛+1 (𝑋) with 𝜕𝑥 = 𝑧. Now we calculate

Sd𝑟 𝑥 = Sd𝑟 𝜕𝑥
|{z}
𝜕
U (𝑋) for large 𝑟
∈𝑆𝑛+1

= Sd𝑟 𝑧
= 𝑧 − 𝜕𝑄 (𝑟 ) 𝑧 − 𝑄 (𝑟 ) 𝜕𝑧 .
|{z}
=0

Hence
𝑧 = 𝜕 ( Sd𝑟 𝑥 + 𝑄 (𝑟 ) 𝑧 )
| {z }
∈𝑆𝑛U (𝑋) for large 𝑟

and therefore [𝑧] 𝐻𝑛U = 0.

123
3. Homology Theory

ii) We show surjectivity: Let [𝑧] ∈ 𝐻𝑛 (𝑋) be given. We know that Sd𝑟 𝑧 ∈ 𝑆𝑛U (𝑋) for 𝑟 large
enough. We compute

[Sd𝑟𝑛 𝑧] = [𝑧 − 𝜕𝑄 𝑛(𝑟 ) 𝑧 − 𝑄 𝑛−1

(𝑟 )
|{z}
𝜕𝑧 ] = [𝑧].
| {z }
∈ 𝐻𝑛 ( 𝑗) (𝐻𝑛U (𝑋) ) =0

b) Now we pass to general (𝑋, 𝐴). Consider the commutative diagram with exact rows:

𝐻𝑛U ( 𝐴) / 𝐻 U (𝑋) / 𝐻 U (𝑋, 𝐴) / 𝐻 U ( 𝐴) / 𝐻 U (𝑋)

𝑛 𝑛 𝑛−1 𝑛−1

   
    
𝐻𝑛 ( 𝐴) / 𝐻𝑛 (𝑋) / 𝐻𝑛 (𝑋, 𝐴) / 𝐻𝑛−1 ( 𝐴) / 𝐻𝑛−1 (𝑋)

By part a) of the proof we know that the outer four arrows are isomorphisms. The Five Lemma
(Exercise 3.11) implies that the map 𝐻𝑛U (𝑋, 𝐴) → 𝐻𝑛 (𝑋, 𝐴) is also an isomorphism. 

3.8. The Mayer-Vietoris sequence

Let 𝑋 be a topological space and let 𝐴 ⊂ 𝑋 be a subset. Assume that 𝑈1 , 𝑈2 ⊂ 𝑋 are open with
𝑈1 ∪ 𝑈2 = 𝑋, hence U = {𝑈1 , 𝑈2 } forms an open cover of 𝑋. Consider the exact sequence of
chain complexes
 
𝑆𝑛 (𝑖1 )

0 / 𝑆 𝑛 (𝑈1 ∩ 𝑈2 ) −𝑆𝑛 (𝑖2 )

/ 𝑆 𝑛 (𝑈1 ) ⊕ 𝑆 𝑛 (𝑈2 ) (𝑆𝑛 ( 𝑗1 ),𝑆𝑛 ( 𝑗2 ) )/ 𝑆 U (𝑋) /0
𝑛

with the inclusion maps 𝑖 𝜈 : 𝑈1 ∩ 𝑈2 → 𝑈𝜈 and 𝑗 𝜈 : 𝑈𝜈 → 𝑋. We then get the following long
exact homology sequence:
 
𝐻𝑛 (𝑖1 )

· · · → 𝐻𝑛 (𝑈1 ∩ 𝑈2 )
−𝐻𝑛 (𝑖2 )
/ 𝐻𝑛 (𝑈1 ) ⊕ 𝐻𝑛 (𝑈2 ) (𝐻𝑛 ( 𝑗1 ),𝐻𝑛 ( 𝑗2 ) ) / 𝐻 U (𝑋) →
𝜕
𝐻𝑛−1 (𝑈1 ∩ 𝑈2 ) → · · ·
𝑛

By the small chain theorem 𝐻𝑛U (𝑋) is canonically isomorphic to 𝐻𝑛 (𝑋). Using this isomorphism
we can replace 𝐻𝑛U (𝑋) in the above exact homology sequence by 𝐻𝑛 (𝑋).

(𝐻𝑛 ( 𝑗1 ),𝐻𝑛 ( 𝑗2 ) ) 𝜕
· · · → 𝐻𝑛 (𝑈1 ) ⊕ 𝐻❱𝑛 (𝑈2 ) / 𝐻 U (𝑋) / 𝐻𝑛−1 (𝑈1 ∩ 𝑈2 ) → · · ·
𝑛
❱❱❱❱ ✐✐4
❱❱❱❱
❱❱❱❱ ✐✐✐ ✐✐✐✐
 ✐✐
(𝐻𝑛 ( 𝑗1 ),𝐻𝑛 ( 𝑗2 ) )❱❱❱❱❱❱ ✐✐✐✐ 𝜕 𝑀𝑉
❱*  ✐✐✐✐
𝐻𝑛 (𝑋)

The same reasoning applies to relative homology and we obtain

124
 3.8. The Mayer-Vietoris sequence

Theorem 3.74 (Mayer-Vietoris sequence). Let 𝑋 be a topological space, let 𝐴 ⊂ 𝑋 and let
𝑈1 , 𝑈2 ⊂ 𝑋 be open such that 𝑈1 ∪ 𝑈2 = 𝑋. Set 𝐴𝜈 := 𝐴 ∩ 𝑈𝜈 and let

𝑖 𝜈 :(𝑈1 ∩ 𝑈2 , 𝐴1 ∩ 𝐴2 ) → (𝑈𝜈 , 𝐴𝜈 ),
𝑗 𝜈 :(𝑈𝜈 , 𝐴𝜈 ) → (𝑋, 𝐴),

be the inclusion maps, 𝜈 = 1, 2. Then the following sequence is exact and natural
 
𝐻𝑛 (𝑖1 )
𝜕 𝑀𝑉 / 𝐻𝑛 (𝑈1 ∩ 𝑈2 , 𝐴1 ∩ 𝐴2 ) −𝐻𝑛 (𝑖2 )
/ 𝐻𝑛 (𝑈1 , 𝐴1 ) ⊕ 𝐻𝑛 (𝑈2 , 𝐴2 )
···
❝ ❝ ❝ ❝ ❝❝❝❝
❝❝❝❝❝❝❝❝
❝❝❝❝❝❝❝ ❝❝❝❝❝❝❝❝(𝐻𝑛 ( 𝑗1 ),𝐻𝑛 ( 𝑗2 ) )
❝❝❝
q ❝❝❝❝❝❝❝
❝
𝐻𝑛 (𝑋, 𝐴) 𝑀𝑉 / 𝐻𝑛−1 (𝑈1 ∩ 𝑈2 , 𝐴1 ∩ 𝐴2 ) / ···
𝜕

Example 3.75. We give a new computation of the homology of 𝑆1 . To this extent we cover the
circle as indicated in the picture.

𝑈1 𝑈2

Figure 82. Open cover of 𝑆1

We directly see that

𝑈1 ≈ 𝑈2 ≈ (0, 1) ≃ {𝑝}

and
𝑈1 ∩ 𝑈2 ≈ (0, 1) ⊔ (0, 1) ≃ {𝑝 1 , 𝑝 2 }

Consider the following part of the Mayer-Vietoris sequence.

𝐻𝑛 (𝑈1 ∩ 𝑈2 ) = 𝐻𝑛 ({𝑝 1 , 𝑝 2 }) / 𝐻𝑛 (𝑈1 ) ⊕ 𝐻𝑛 (𝑈2 ) = 𝐻𝑛 ({𝑝}) ⊕ 𝐻𝑛 ({𝑝}) / 1

❜❜❜❜❜❜ ❜❜❜❜❜❜❜❜❜❜ 𝐻𝑛 (𝑆 )
❜❜❜❜❜❜❜
❜❜❜❜❜❜❜❜ ❜❜❜❜❜❜❜❜❜❜
❜❜❜❜ ❜❜ 𝑀𝑉
q❜❜❜❜❜❜❜❜❜❜ 𝜕
𝐻𝑛−1 (𝑈1 ∩ 𝑈2 ) = 𝐻𝑛−1 ({𝑝 1 , 𝑝 2 }) / 𝐻𝑛−1 (𝑈1 ) ⊕ 𝐻𝑛−1 (𝑈2 ) = 𝐻𝑛−1 ({𝑝}) ⊕ 𝐻𝑛−1 ({𝑝})

If 𝑛 ≥ 2 then all homologies of the point occuring in this diagram vanish. Hence 𝐻𝑛 (𝑆1 ) = 0 for
all 𝑛 ≥ 2. Since 𝑆1 is path-connected we have that 𝐻0 (𝑆1 )  𝑅. In the case 𝑛 = 1 we find
 
11
11
0 / 𝐻1 (𝑆 1 ) / 𝑅2 / 𝑅2

125
3. Homology Theory

To compute 𝐻1 (𝑆1 ) we calculate


   
1 1 1 −𝑥
𝐻1 (𝑆 )  ker = 𝑥 ∈ 𝑅  𝑅.
1 1 𝑥

Example 3.76. Now we consider the space 𝑋 = 𝐺 2 as given in the picture.

𝑈1 𝑈2

Figure 83. Open cover of figure 8

It is easy to see that 𝑈1 ≈ 𝑈2 ≃ 𝑆1 and 𝑈1 ∩ 𝑈2 ≃ {𝑝}. Since 𝐺 2 is path-connected we find

𝐻0 (𝐺 2 )  𝑅.

𝐻𝑛 (𝑈1 ∩ 𝑈2 ) → 𝐻𝑛 (𝑈1 ) ⊕ 𝐻𝑛 (𝑈2 ) → 𝐻𝑛 (𝐺 2 ) → 𝐻𝑛−1 (𝑈1 ∩ 𝑈2 )

| {z } | {z }
𝐻𝑛 (𝑆 1 ) ⊕𝐻𝑛 (𝑆 1 ) 𝐻𝑛−1 ( { 𝑝} )

In the case of 𝑛 ≥ 2 we find the exact sequence

0 / 𝐻𝑛 (𝐺 2 ) /0

and hence 𝐻𝑛 (𝐺 2 ) = 0. For 𝑛 = 1 we find

𝐻1 ({𝑝}) → 𝐻1 (𝑆1 ) ⊕ 𝐻1 (𝑆1 ) → 𝐻1 (𝐺 2 ) → 𝐻0 ({𝑝}) → 𝐻0 (𝑆1 ) ⊕ 𝐻0 (𝑆1 ) .

| {z } | {z } | {z } | {z }
=0 𝑅 2 𝑅 𝑅 2

The last map in this diagram is given by

𝐻0 ({𝑝})  𝑅 /𝐻 (𝑆 1) ⊕: 𝐻0 (𝑆1 )  𝑅 2
  0
1
1

and hence injective. Consequently we find the isomorphism

𝐻1 (𝐺 2 )  𝐻1 (𝑈1 ) ⊕ 𝐻1 (𝑈2 )  𝑅 2 .

3.9. Generalized Jordan curve theorem

In this section we will use the Mayer-Vietoris sequence to prove the Jordan curve theorem in
arbitrary dimensions.

126
 3.9. Generalized Jordan curve theorem

Lemma 3.77. Let 𝑋 be a topological space and let 𝑈𝑖 ⊂ 𝑋 be open with 𝑈𝑖 ⊂ 𝑈𝑖+1 and
∪𝑖∈N𝑈𝑖 = 𝑋. Furthermore let 𝜄𝑛 : 𝑈𝑛 ↩→ 𝑋 and 𝜄𝑛,𝑚 : 𝑈𝑛 ↩→ 𝑈𝑚 for 𝑚 ≥ 𝑛 be the inclusion
maps. Then we have:

(i) For each 𝛼 ∈ 𝐻 𝑘 (𝑋; 𝑅) there exists an 𝑛0 such that 𝛼 ∈ im(𝐻 𝑘 (𝜄𝑛 )) for all 𝑛 ≥ 𝑛0 .

(ii) For each 𝛼𝑛 ∈ 𝐻 𝑘 (𝑈𝑛 ; 𝑅) with 𝐻 𝑘 (𝜄𝑛 ) (𝛼𝑛 ) = 0 there exists an 𝑚 0 such that
𝐻 𝑘 (𝜄𝑛,𝑚 ) (𝛼𝑛 ) = 0 for all 𝑚 ≥ 𝑚 0 .

Proof. (i) Let 𝛼 ∈ 𝐻 𝑘 (𝑋). We represent 𝛼 by

Õ
𝑙
𝛼 𝑗 𝜎 𝑗 ∈ 𝑍 𝑘 (𝑋; 𝑅), 𝛼 𝑗 ∈ 𝑅, 𝜎 𝑗 ∈ 𝐶 (Δ𝑘 , 𝑋).
𝑗=1

Ð
Note that 𝜎 𝑗 (Δ𝑘 ) ⊂ 𝑋 is a compact subset and therefore 𝐶 := 𝑙𝑗=1 𝜎 𝑗 (Δ𝑘 ) ⊂ 𝑋 is
compact. Hence there exists an 𝑛0 such that for all 𝑛 ≥ 𝑛0 we have 𝐶 ⊂ 𝑈𝑛 . We conclude
that
Õ 𝑙
𝛼 𝑗 𝜎 𝑗 ∈ 𝑍 𝑘 (𝑈𝑛 ; 𝑅)
𝑗=1

for all 𝑛 ≥ 𝑛0 and therefore

hÕ i  hÕ i 
𝛼= 𝛼 𝑗 𝜎𝑗 = 𝐻𝑛 (𝜄𝑛 ) 𝛼 𝑗 𝜎𝑗 .
𝐻𝑘 (𝑋) 𝐻𝑘 (𝑈𝑛 )

Í
(ii) Again represent 𝛼𝑛 ∈ 𝐻 𝑘 (𝑈𝑛 ) by 𝑙𝑗=1 𝛼 𝑗 𝜎 𝑗 . From 𝐻𝑛 (𝜄𝑛 ) (𝛼𝑛 ) = 0 we know that there
Í
exists a 𝛽 ∈ 𝐶 𝑘+1 (𝑋; 𝑅) such that 𝑙𝑗=1 𝛼 𝑗 𝜎 𝑗 = 𝜕 𝛽. As before there exists a compact
subset 𝐶 ′ ⊂ 𝑋 such that 𝛽 ∈ 𝐶 𝑘+1 (𝐶 ′ ; 𝑅). Since there exists an 𝑚 0 with 𝐶 ′ ⊂ 𝑈𝑚 for all
𝑚 ≥ 𝑚 0 and thus 𝛽 ∈ 𝐶 𝑘+1 (𝑈𝑚 ; 𝑅) we have that 𝐻 𝑘 (𝜄𝑛,𝑚 ) (𝛼𝑛 ) = 0. 

By an embedding we mean a continuous map 𝑓 : 𝑋 → 𝑌 which is a homeomorphism onto its

image. In other words, 𝑓 is continuous, open and injective. In the considerations which follow
the domain will be compact and the target will be Hausdorff so that 𝑓 is automatically open.

Proposition 3.78. Let 𝑛 ∈ N and let 𝑌 be a compact topological space such that for any
embedding 𝑓 : 𝑌 → 𝑆 𝑛
𝐻∗ (𝑆 𝑛 \ 𝑓 (𝑌 ); 𝑅)  𝐻∗ (point; 𝑅) .
Then the space [0, 1] × 𝑌 also has this property.

127
3. Homology Theory

Proof. Let 𝑓 : [0, 1] × 𝑌 → 𝑆 𝑛 be an embedding. Suppose 0 ≠ 𝛼 ∈ 𝐻𝑖 (𝑆 𝑛 \ 𝑓 ([0, 1] × 𝑌 )). Put

𝑈0 := 𝑆 𝑛 \ 𝑓 ([0, 21 ] × 𝑌 ) and 𝑈1 := 𝑆 𝑛 \ 𝑓 ([ 12 , 1] × 𝑌 ) .
| {z } | {z }
compact compact

Both 𝑈0 and 𝑈1 are open. We also find

𝑈0 ∩ 𝑈1 = 𝑆 𝑛 \ 𝑓 ([0, 1] × 𝑌 ) and 𝑈0 ∪ 𝑈1 = 𝑆 𝑛 \ 𝑓 ({ 12 } × 𝑌 ) .
| {z }
≈𝑌

The Mayer-Vietoris sequence yields:

0 = 𝐻𝑖+1 (𝑆 𝑛 \ 𝑓 ({ 12 } × 𝑌 )) / 𝐻𝑖 (𝑆 𝑛 \ 𝑓 ([0, 1] × 𝑌 ))


𝐻𝑖 (𝑆 𝑛 \ 𝑓 ([0, 12 ] × 𝑌 )) ⊕ 𝐻𝑖 (𝑆 𝑛 \ 𝑓 ([ 21 , 1] × 𝑌 ))

Thus the inclusions 𝑆 𝑛 \ 𝑓 ([0, 1] × 𝑌 ) ↩→ 𝑆 𝑛 \ 𝑓 ([0, 12 ] × 𝑌 ) and 𝑆 𝑛 \ 𝑓 ([0, 1] × 𝑌 ) ↩→

𝑆 𝑛 \ 𝑓 ([ 21 , 1] × 𝑌 ) induce an injective homomorphism

𝐻𝑖 (𝑆 𝑛 \ 𝑓 ([0, 1] × 𝑌 )) → 𝐻𝑖 (𝑆 𝑛 \ 𝑓 ([0, 21 ] × 𝑌 ) ⊕ 𝐻𝑖 (𝑆 𝑛 \ 𝑓 ([ 12 , 1] × 𝑌 )) .

Hence

0 ≠ 𝐻𝑖 (inclusion) (𝛼) ∈ 𝐻𝑖 (𝑆 𝑛 \ 𝑓 ([0, 21 ] × 𝑌 )) or

0 ≠ 𝐻𝑖 (inclusion) (𝛼) ∈ 𝐻𝑖 (𝑆 𝑛 \ 𝑓 ([ 12 , 1] × 𝑌 )) .

By iterating this procedure we obtain a sequence of intervals 𝐼 𝑘 such that

𝐼0 = [0, 1], 𝐼 𝑘+1 ⊂ 𝐼 𝑘 , |𝐼 𝑘 | = 2−𝑘

and
0 ≠ 𝐻𝑖 (𝜄0,𝑘 ) (𝛼) ∈ 𝐻𝑖 (𝑆 𝑛 \ 𝑓 (𝐼 𝑘 × 𝑌 ))
for all 𝑘. Here 𝑉𝑘 := 𝑆 𝑛 \ 𝑓 (𝐼 𝑘 × 𝑌 ) is open in 𝑆 𝑛 and 𝜄 𝑘,𝑙 : 𝑉𝑘 ↩→ 𝑉𝑙 for 𝑙 ≥ 𝑘 denotes the
inclusion map. We find

∩𝑘 ∈N 𝐼 𝑘 = {𝑡}
∪𝑘 ∈N𝑉𝑘 = 𝑆 𝑛 \ 𝑓 ({𝑡} × 𝑌 ) =: 𝑋

Now we apply the previous Lemma 3.77 for the inclusion 𝜄 : 𝑉0 ↩→ 𝑋. Hence, for 𝑖 ≥ 1,

0 ≠ 𝐻𝑖 (𝜄) (𝛼) ∈ 𝐻𝑖 (𝑋) = 𝐻𝑖 (𝑆 𝑛 \ 𝑓 ({𝑡} × 𝑌 ) ) = 0

| {z }
≈𝑌

giving a contradiction. The proof for 𝑖 = 0 is similar (or in fact the same if one uses augmented
homology). 

128
 3.9. Generalized Jordan curve theorem

Corollary 3.79. If 𝑓 : 𝐷 𝑟 → 𝑆 𝑛 is an embedding then

𝐻∗ (𝑆 𝑛 \ 𝑓 (𝐷 𝑟 ))  𝐻∗ ({point})

Proof. The proof is by induction on 𝑟. For 𝑟 = 0 we find

𝑆 𝑛 \ 𝑓 (𝐷 0 ) ≈ R𝑛 ≃ {point}

and hence 𝐻∗ (𝑆 𝑛 \ 𝑓 (𝐷 𝑟 ))  𝐻∗ ({point}).

For the induction step “𝑟 − 1 ⇒ 𝑟” we observe 𝐷 𝑟 ≈ 𝑊 𝑟 = [0, 1] × 𝑊 𝑟 −1 ≈ [0, 1] × 𝐷 𝑟 −1 and
hence Proposition 3.78 applies. 

Theorem 3.80. Let 𝑟 < 𝑛 and let 𝑓 : 𝑆𝑟 → 𝑆 𝑛 be an embedding. Then

𝐻∗ (𝑆 𝑛 \ 𝑓 (𝑆𝑟 ))  𝐻∗ (𝑆 𝑛−𝑟 −1 ).

Proof. Again the proof is done by induction on 𝑟. For 𝑟 = 0 we find

𝑆 𝑛 \ 𝑓 (𝑆0 ) = 𝑆 𝑛 \ {𝑝, 𝑞} ≈ R𝑛 \ {0} ≃ 𝑆 𝑛−1 .

For the induction step “𝑟 − 1 ⇒ 𝑟” we write 𝑆𝑟 = 𝐷 𝑟+ ∪ 𝐷 𝑟− . Then 𝐷 𝑟+ ∩ 𝐷 𝑟− = 𝑆𝑟 −1 . We put

𝑈+ := 𝑆 𝑛 \ 𝑓 (𝐷 𝑟+ ) and 𝑈 − := 𝑆 𝑛 \ 𝑓 (𝐷 𝑟− ). Both sets 𝑈+ , 𝑈 − are open because 𝑓 (𝐷 𝑟± ) is compact.
In addition
𝑈+ ∩ 𝑈 − = 𝑆 𝑛 \ 𝑓 (𝑆𝑟 ) and 𝑈+ ∪ 𝑈 − = 𝑆 𝑛 \ 𝑓 (𝑆𝑟 −1 ).
Now we look at the Mayer-Vietoris sequence and use Corollary 3.79:

𝐻𝑖+1 (𝑈+) ⊕ 𝐻𝑖+1 (𝑈 − ) / 𝐻𝑖+1 (𝑆 𝑛 \ 𝑓 (𝑆𝑟 −1 )) / 𝐻𝑖 (𝑆 𝑛 \ 𝑓 (𝑆𝑟 ))

| {z }
𝐻𝑖+1 (point) ⊕𝐻𝑖+1 (point)=0


𝐻𝑖 (𝑈+ ) ⊕ 𝐻𝑖 (𝑈 − )
| {z }
𝐻𝑖 (point) ⊕𝐻𝑖 (point) =0
if 𝑖 ≥1

Thus for 𝑖 ≥ 1 (and by similar reasoning also for 𝑖 = 0) we get an isomorphism

𝐻𝑖 (𝑆 𝑛 \ 𝑓 (𝑆𝑟 ))  𝐻𝑖+1 (𝑆 𝑛 \ 𝑓 (𝑆𝑟 −1 ))  𝐻𝑖+1 (𝑆 𝑛−𝑟 )  𝐻𝑖 (𝑆 𝑛−𝑟 −1 ).

This proves the theorem. 

129
3. Homology Theory

Theorem 3.81 (Jordan-Brouwer separation theorem). For any embedding 𝑓 : 𝑆 𝑛−1 → 𝑆 𝑛

the complement 𝑆 𝑛 \ 𝑓 (𝑆 𝑛−1 ) consists of exactly two path-components 𝑈 and 𝑉. Both 𝑈 and
𝑉 are open in 𝑆 𝑛 and 𝜕𝑈 = 𝜕𝑉 = 𝑓 (𝑆 𝑛−1 ).

Proof. a) We know that 𝐻0 (𝑆 𝑛 \ 𝑓 (𝑆 𝑛−1 ))  𝐻0 (𝑆0 )  𝑅 2 , hence 𝑆 𝑛 \ 𝑓 (𝑆 𝑛−1 ) has exactly two
path-components 𝑈 and 𝑉.

b) Since 𝑓 (𝑆 𝑛−1 ) is compact the union 𝑈 ∪𝑉 = 𝑆 𝑛 \ 𝑓 (𝑆 𝑛−1 ) is open and therefore both connected
components 𝑈 and 𝑉 are open.

c) Since 𝑈 and 𝑉 are open they contain no boundary point of 𝑈, thus 𝜕𝑈 ⊂ 𝑓 (𝑆 𝑛−1 ). Assume
that there exists 𝑝 ∈ 𝑆 𝑛−1 with 𝑓 ( 𝑝) ∉ 𝜕𝑈. Then there exists 𝑊 ⊂ 𝑆 𝑛 open with 𝑓 ( 𝑝) ∈ 𝑊 and
𝑊 ∩ 𝑈 = ∅. Since 𝑓 is continuous we can choose an open ball 𝐵 ⊂ 𝑆 𝑛−1 with 𝑓 (𝐵) ⊂ 𝑊. Since
𝑆 𝑛−1 \ 𝐵 ≈ 𝐷 𝑛−1 we find that the space 𝑌 := 𝑆 𝑛 \ 𝑓 (𝑆 𝑛−1 \ 𝐵) is path-connected because of
Corollary 3.79:
𝐻0 (𝑌 ) = 𝐻0 (𝑆 𝑛 \ 𝑓 (𝑆 𝑛−1 \ 𝐵))  𝐻0 (point)  𝑅 .
Moreover, we have

𝑌 = 𝑈 ∪ 𝑉 ∪ 𝑓 (𝐵) ⊂ 𝑈 ∪ 𝑉 ∪ 𝑊 and 𝑈 ∩ (𝑉 ∪ 𝑊) = ∅

and hence
𝑌 = (𝑌 ∩ 𝑈 ) ⊔ (𝑌 ∩ (𝑉 ∪ 𝑊))
| {z } | {z }
𝑈 ⊃𝑉

can be written as a disjoint union of open non-empty subsets. This contradicts 𝑌 being path-
connected. 

Corollary 3.82 (Generalized Jordan curve theorem). For every embedding 𝑓 : 𝑆 𝑛−1 → R𝑛
the complement R𝑛 \ 𝑓 (𝑆 𝑛−1 ) consists of exactly two path components 𝑈 and 𝑉. Both 𝑈 and
𝑉 are open, 𝑈 is bounded, 𝑉 is unbounded and 𝜕𝑈 = 𝜕𝑉 = 𝑓 (𝑆 𝑛−1 ).

Proof. We use the stereographic projection to identify R𝑛 with a subset of 𝑆 𝑛 ,

𝑆 𝑛 = R𝑛 ∪ {∞}, 𝑓 : 𝑆 𝑛−1 → R𝑛 ⊂ 𝑆 𝑛 .

By the Jordan-Brouwer separation theorem the two path-components 𝑈˜ and 𝑉˜ of 𝑆 𝑛 \ 𝑓 (𝑆 𝑛−1 )

are both open and 𝜕𝑈˜ = 𝜕𝑉˜ = 𝑓 (𝑆 𝑛−1 ). Let 𝑉˜ be the component containing ∞. Then 𝑈 = 𝑈˜
and 𝑉 = 𝑉˜ \ {∞} are the two path-components of R𝑛 \ 𝑓 (𝑆 𝑛−1 ). Clearly, 𝑉 is unbounded and
𝜕𝑈 = 𝜕𝑈˜ = 𝜕𝑉 = 𝜕𝑉˜ = 𝑓 (𝑆 𝑛−1 ). If 𝑈 were unbounded then ∞ ∈ 𝜕𝑈 which is not the case.
Hence 𝑈 is bounded. 

130
 3.10. CW-complexes

Remark 3.83. Consider an embedding 𝑓 : 𝑆 𝑛−1 → 𝑆 𝑛 . If 𝑖 ≥ 1 we find for the homology of the
components of the complement
𝐻 (𝑈 ⊔ 𝑉) = 𝐻𝑖 (𝑈) ⊕ 𝐻𝑖 (𝑉)  𝐻𝑖 (𝑆0 ) = 0
and hence 𝑈 and 𝑉 have the same homology as a point. This makes us suspect that 𝑈, 𝑉 ≈ 𝐷˚ 𝑛 .
For 𝑛 = 2 this is indeed true, but it is false for 𝑛 ≥ 3. The Alexander horned sphere is an example
of an embedding of 𝑆2 into 𝑆3 where one component of the complement is not even simply
connected:

Figure 84. Alexander horned sphere2

The red circle in the picture is a non-contractible loop giving rise to a non-trivial element in
the fundamental group. A very nice video illustrating this embedding of 𝑆2 can be found at
http://www.youtube.com/watch?v=d1Vjsm9pQlc.

3.10. CW-complexes
We now describe a type of topological spaces for which there is a particularly efficient way to
compute their homology. These space are obtained by gluing together balls of various dimensions.

Ý
Definition 3.84. A finite CW-complex is a pair (𝑋, X) where 𝑋 is a Hausdorff space,
X = 𝑛∈N0 X𝑛 , X𝑛 ⊂ P (𝑋) and |X| < ∞ with the following properties:
Ý
(i) 𝑋 = 𝜎 ∈ X 𝜎.

(ii) Set 𝑋 𝑛 := ∪ 𝜎 ∈ X𝑚 𝜎 ⊂ 𝑋. For every 𝜎 ∈ X𝑛 we have 𝜎

¯ \ 𝜎 ⊂ 𝑋 𝑛−1 .
𝑚≤𝑛

(iii) For every 𝜎 ∈ X𝑛 there exists a surjective continuous map

𝜑 𝜎 : 𝐷𝑛 → 𝜎
¯ ⊂𝑋

2 Taken from https://matteocapucci.wordpress.com/2019/02/05/you-wont-believe-what-this-space-is-homeomorphic-to

131
3. Homology Theory

such that 𝜑 𝜎 | 𝐷˚ 𝑛 : 𝐷˚ 𝑛 → 𝜎 is a homeomorphism.

Definition 3.85. An element 𝜎 ∈ X𝑛 is called an n-cell. The map 𝜑 𝜎 is called the characteris-
tic map of 𝜎 and 𝑋 𝑛 is called the n-skeleton of (𝑋, X). The map 𝜑 𝜎 𝑆 𝑛−1 =𝜕𝐷 𝑛 : 𝑆 𝑛−1 → 𝑋 𝑛−1
is called the attaching map of 𝜎.

𝑋2

b b
𝑋0

𝑋1

b b

Figure 85. A 2-dimensional CW-complex

Example 3.86. Consider 𝑋 = 𝑆 𝑛 for 𝑛 ≥ 1. Then the choice

X0 = {{𝑒1 }},
X𝑛 = {𝜎𝑛 = 𝑆 𝑛 \ {𝑒1 }},
X𝑚 = ∅, otherwise,
turns the 𝑛-sphere into a CW-complex.
𝑒1
𝐷𝑛 b
𝑆𝑛

𝜑 𝜎𝑛

Figure 86. Attaching an 𝑛-cell to a point to obtain an 𝑛-sphere

The attaching map to the 𝑛-cell is the constant map 𝑆 𝑛−1 → {𝑒1 }. We have
𝑋 0 = 𝑋 1 = 𝑋 2 = . . . = 𝑋 𝑛−1 = {𝑒1 },
𝑆 𝑛 = 𝑋 𝑛 = 𝑋 𝑛+1 = . . .

132
 3.10. CW-complexes

Example 3.87. If a space 𝑋 has the structure of a CW-complex there are in general many different
ways to write 𝑋 as a CW-complex, i.e., there are many different X for the same 𝑋. Let us look
again at 𝑋 = 𝑆 𝑛 . We start with the case 𝑛 = 0. Here the CW-structure is unique:
(
{{𝑒1 }, {−𝑒1 }}, 𝑚 = 0
X𝑚 =
∅, 𝑚 > 0.

For 𝑛 > 0 we use that 𝑆 𝑛−1 ⊂ 𝑆 𝑛 and define recursively




𝑛−1
X𝑆 ,
 𝑚

𝑚 ≤ 𝑛−1
𝑛
X𝑚𝑆 := { 𝐷˚ +𝑛 , 𝐷˚ 𝑛− },

𝑚=𝑛

 ∅,
 𝑚>𝑛

𝑆𝑛
◦+
𝐷𝑛

𝑆 𝑛−1

◦−
𝐷𝑛

Figure 87. Cell decomposition of 𝑆 𝑛 with two 𝑛-cells

Therefore we have in this case

𝑋 0 = 𝑆0 , 𝑋 1 = 𝑆1 , . . . , 𝑋 𝑛 = 𝑆 𝑛 .

Example 3.88. Real projective space RP𝑛 . We define the real projective space as

RP𝑛 = {1-dimensional real vector subspace of R𝑛+1 } = R𝑛+1 \ {0}/∼

where
𝑥 = (𝑥0 , . . . , 𝑥 𝑛 ) ∼ 𝑦 = (𝑦 0 , . . . , 𝑦 𝑛 )
iff there exists a 𝑡 ≠ 0 such that 𝑥 = 𝑡 𝑦. We consider the canonical projection map

𝜋 : R𝑛+1 \ {0} → RP𝑛 , 𝑥 ↦→ [𝑥0 , . . . , 𝑥 𝑛 ].

The restriction 𝜓 := 𝜋 𝑆 𝑛 : 𝑆 𝑛 → RP𝑛 is continuous and surjective. Thus RP𝑛 is compact. Clearly
𝜓 (𝑥) = 𝜓 (𝑦) iff 𝑥 = ±𝑦. We set

𝜎𝑘 := {[𝑥0 , . . . , 𝑥 𝑛 ] ∈ RP𝑛 | 𝑥 𝑘+1 = . . . = 𝑥 𝑛 = 0, 𝑥 𝑘 ≠ 0} .

133
3. Homology Theory

Then
¯ 𝑘 = {[𝑥0 , . . . , 𝑥 𝑛 ] ∈ RP𝑛 | 𝑥 𝑘+1 = . . . = 𝑥 𝑛 = 0} ≈ RP𝑘 .
𝜎
For the characteristic map 𝜑 𝑘 : 𝐷 𝑘 → 𝜎 ¯ 𝑘 we can take
q
𝜉 ↦→ [𝜉1 , . . . , 𝜉 𝑘 , 1 − |𝜉 | 2 , 0, . . . , 0] .

It is clear that the map 𝜑 𝑘 is continuous. Now we check that it is also surjective. Let
[𝑥 ′ , 𝑥 𝑘 , 0] ∈ 𝜎
¯ 𝑘 where 𝑥 ′ = (𝑥0 , . . . , 𝑥 𝑘−1 ). Without loss of generality we assume that 𝑥 𝑘 ≥ 0.
′
Set 𝜉 := √ ′ 𝑥2 2
∈ 𝐷 𝑘 . We compute
| 𝑥 | +| 𝑥 𝑘 |

" s #
𝑥′ |𝑥 ′ | 2
𝜑 𝑘 (𝜉) = p , 1− ′ 2 ,0
|𝑥 ′ | 2 + |𝑥 𝑘 | 2 |𝑥 | + |𝑥 𝑘 | 2
" s #
𝑥′ |𝑥 𝑘 | 2
= p , ,0
|𝑥 ′ | 2 + |𝑥 𝑘 | 2 |𝑥 ′ | 2 + |𝑥 𝑘 | 2
= [𝑥 ′ , |𝑥 𝑘 |, 0]
= [𝑥 ′ , 𝑥 𝑘 , 0] .

Next we show that 𝜑 𝑘 | 𝐷˚ 𝑘 is injective. Let 𝜑 𝑘 (𝜉) = 𝜑 𝑘 (𝜂) for 𝜉, 𝜂 ∈ 𝐷˚ 𝑘 . This leads to the two
equations: q q
𝜉 = 𝑡𝜂 and 1 − |𝜉 | = 𝑡 1 − |𝜂| 2
2

for some 𝑡 ≠ 0. Squaring and adding both equations we find

|𝜉 | 2 + 1 − |𝜉 | 2 = 𝑡 2 |𝜂| 2 + 𝑡 2 (1 − |𝜂| 2 )
p
p
leading to 𝑡 2 = 1 and consequently 𝑡 = ±1. Since |𝜉 |, |𝜂| < 1 it follows that 1 − |𝜉 | 2 > 0 and
1 − |𝜂| 2 > 0 and therefore 𝑡 > 0. Hence 𝑡 = 1 and thus 𝜉 = 𝜂.
The map 𝜑 𝑘 : 𝐷 𝑘 → 𝜎 ¯ 𝑘 is closed, therefore the restriction

𝜑𝑘 ˚𝑘
𝐷
: 𝐷˚ → 𝜎𝑘

is bijective, continuous and closed, hence a homeomorphism. We find:

𝑋 0 = {point} ⊂ 𝑋 1 ⊂ 𝑋 2 ⊂ · · · ⊂ 𝑋 𝑛 .
|{z} |{z} |{z}
≈RP1 ≈RP2 =RP𝑛

Finally let us discuss the gluing map. For 𝜉 ∈ 𝜕𝐷 𝑘 = 𝑆 𝑘−1 , i.e. |𝜉 | = 1, the gluing map is given
by 𝜑 𝑘 (𝜉) = [𝜉, 0, 0] and hence 𝜑 𝑘 = 𝜓 : 𝑆 𝑘−1 → 𝑋 𝑘−1 ≈ RP𝑘−1 .

Example 3.89. Complex projective space CP𝑛 . For the complex projective space the discussion
is analogous to the real case with complex parameters instead of real parameters,

CP𝑛 = {1-dimensional complex vector subspace of C𝑛+1 } = C𝑛+1 \ {0}/∼

134
 3.11. Homology of CW-complexes

with 𝑥 ∼ 𝑦 iff 𝑥 = 𝑡 𝑦 for some 𝑡 ∈ C, 𝑡 ≠ 0. We find

(
1, 𝑚 even and 0 ≤ 𝑚 ≤ 2𝑛
|X𝑚 | =
0, otherwise

and
𝑋 0 = {point} = 𝑋 1 ⊂ 𝑋 2 = 𝑋 3 ⊂ 𝑋 4 = 𝑋 5 ⊂ · · · ⊂ 𝑋 2𝑛−1 = 𝑋 2𝑛 .
|{z} |{z} |{z}
≈CP1 ≈CP2 =CP𝑛

Example 3.90. Quaternionic projective space HP𝑛 . Similarly, for the quaternionic projective
space

HP𝑛 = {1-dimensional quaternionic vector subspace of H𝑛+1 } = H𝑛+1 \ {0}/∼

with 𝑥 ∼ 𝑦 iff 𝑥 = 𝑡 𝑦 for some 𝑡 ∈ H, 𝑡 ≠ 0, we find that

(
1, 𝑚 divisible by 4 and 0 ≤ 𝑚 ≤ 4𝑛,
|X𝑚 | =
0, otherwise,

and
𝑋 0 = {point} = 𝑋 1 = 𝑋 2 = 𝑋 3 ⊂ 𝑋 4 = . . . ⊂ 𝑋 4𝑛−3 = · · · = 𝑋 4𝑛 .
|{z} |{z}
≈HP1 =HP𝑛

Remark 3.91. Every compact differentiable manifold can be triangulated and is consequently a
finite CW-complex.

3.11. Homology of CW-complexes

Throughout this section we assume that (𝑋, X) is a finite CW-complex. Our goal is to prove
Theorem 3.100 which will provide us with an efficient way to compute the homology of 𝑋.

Lemma 3.92. The map

É ⊕ 𝜎∈X𝑛 𝐻𝑖 ( 𝜑 𝜎 )
𝐻𝑖 (𝐷 𝑛 , 𝑆 𝑛−1 ) / 𝐻𝑖 (𝑋 𝑛 , 𝑋 𝑛−1 )
𝜎 ∈ X𝑛

is an isomorphism.

Once we have this lemma, Theorem 3.16 implies

135
3. Homology Theory

Corollary 3.93. We have the isomorphisms

(
𝑅 | X𝑛 | , for 𝑖 = 𝑛
𝐻𝑖 (𝑋 𝑛 , 𝑋 𝑛−1 ) 
0, otherwise

Proof of Lemma 3.92. We set 𝐷¤ 𝑛 := 𝐷 𝑛 \ {0} and

Ø
𝑋¤ 𝑛 := 𝑋 𝑛−1 ∪ 𝜑 𝜎 ( 𝐷¤ 𝑛 ) = 𝑋 𝑛 \ {𝜑 𝜎 (0) | 𝜎 ∈ X𝑛 }.
𝜎 ∈ X𝑛

𝑥
The inclusion 𝑆 𝑛−1 ↩→ 𝐷¤ 𝑛 is a homotopy equivalence with homotopy inverse 𝑥 ↦→ |𝑥| .

𝐷𝑛

Figure 88. Punctured 𝑛-disk is homotopy equivalent to 𝑆 𝑛−1

We define Þ Þ Þ
𝑌 𝑛 := 𝐷𝑛, 𝑌 𝑛−1 := 𝑆 𝑛−1 , 𝑌¤ 𝑛 := 𝐷¤ 𝑛
𝜎 ∈ X𝑛 𝜎 ∈ X𝑛 𝜎 ∈ X𝑛

The inclusions yield homotopy equivalences 𝑌 𝑛−1 → 𝑌¤ 𝑛 and 𝑋 𝑛−1 → 𝑋¤ 𝑛 .

𝑋2
b

𝑋1

Figure 89. Punctured 𝑛-skeleton is homotopy equivalent to (𝑛 − 1)-skeleton

Now consider:

 Φ:=∪ 𝜎∈X𝑛 𝜑 𝜎
(𝑌 𝑛 , 𝑌 𝑛−1 )  / (𝑌 𝑛 , 𝑌¤ 𝑛 ) / (𝑋 𝑛 , 𝑋¤ 𝑛 ) o ? _ (𝑋 𝑛 , 𝑋 𝑛−1 )

136
 3.11. Homology of CW-complexes

Both inclusions induce isomorphisms on 𝐻𝑖 . Due to the diagram

𝐻𝑖 (𝑌 𝑛−1 ) / 𝐻𝑖 (𝑌 𝑛 ) / 𝐻𝑖 (𝑌 𝑛 , 𝑌 𝑛−1 ) / 𝐻𝑖−1 (𝑌 𝑛−1 ) / 𝐻𝑖−1 (𝑌 𝑛 )

 =  =
    
𝐻𝑖 (𝑌¤ 𝑛 ) / 𝐻𝑖 (𝑌 𝑛 ) / 𝐻𝑖 (𝑌 𝑛 , 𝑌¤ 𝑛 ) / 𝐻𝑖−1 (𝑌¤ 𝑛 ) / 𝐻𝑖−1 (𝑌 𝑛 )

and the Five Lemma the map 𝐻𝑖 (𝑌 𝑛 , 𝑌 𝑛−1 ) → 𝐻𝑖 (𝑌 𝑛 , 𝑌¤ 𝑛 ) is also an isomorphism. Similarly,
we get that the inclusion (𝑋 𝑛 , 𝑋 𝑛−1 ) ↩→ (𝑋 𝑛 , 𝑋¤ 𝑛 ) induces an isomorphism 𝐻𝑖 (𝑋 𝑛 , 𝑋 𝑛−1 ) →
𝐻𝑖 (𝑋 𝑛 , 𝑋¤ 𝑛 ). The inclusions

(𝑌 𝑛 \ 𝑌 𝑛−1 , 𝑌¤ 𝑛 \ 𝑌 𝑛−1 ) ↩→ (𝑌 𝑛 , 𝑌¤ 𝑛 ),
(𝑋 𝑛 \ 𝑋 𝑛−1 , 𝑋¤ 𝑛 \ 𝑋 𝑛−1 ) ↩→ (𝑋 𝑛 , 𝑋¤ 𝑛 ),

induce isomorphisms on homology by the excision axiom. In addition, we have

(𝑌 𝑛 \ 𝑌 𝑛−1 , 𝑌¤ 𝑛 \ 𝑌 𝑛−1 ) ≈ (𝑋 𝑛 \ 𝑋 𝑛−1 , 𝑋¤ 𝑛 \ 𝑋 𝑛−1 )

Hence we find
Ê
𝐻𝑖 (𝐷 𝑛 , 𝑆 𝑛−1 )  𝐻𝑖 (𝑌 𝑛 , 𝑌 𝑛−1 )
𝜎 ∈ X𝑛
 𝐻𝑖 (𝑌 𝑛 , 𝑌¤ 𝑛 )
 𝐻𝑖 (𝑌 𝑛 \ 𝑌 𝑛−1 , 𝑌¤ 𝑛 \ 𝑌 𝑛−1 )
 𝐻𝑖 (𝑋 𝑛 \ 𝑋 𝑛−1 , 𝑋¤ 𝑛 \ 𝑋 𝑛−1 )
 𝐻𝑖 (𝑋 𝑛 , 𝑋¤ 𝑛 )
 𝐻𝑖 (𝑋 𝑛 , 𝑋 𝑛−1 ) . 

Set 𝐾𝑛 (𝑋, X) := 𝐻𝑛 (𝑋 𝑛 , 𝑋 𝑛−1 )  𝑅 | X𝑛 | . We define a homomorphism

𝜕𝑛 : 𝐾𝑛 (𝑋, X) → 𝐾𝑛−1 (𝑋, X)

by the following diagram:

/ 𝐻𝑛 (𝑋 𝑛 , 𝑋 𝑛−1 ) 𝜕 / 𝐻𝑛−1 (𝑋 𝑛−1 ) / 𝐻𝑛−1 (𝑋 𝑛 ) / ···

···
= 𝜕𝑛
 *
··· / 𝐻𝑛−1 (𝑋 𝑛−2 ) / 𝐻𝑛−1 (𝑋 𝑛−1 ) / 𝐻𝑛−1 (𝑋 𝑛−1 , 𝑋 𝑛−2 ) / ···

Lemma 3.94. The sequence of groups 𝐾𝑛 (𝑋, X) together with 𝜕𝑛 forms a complex.

The pair (𝐾∗ (𝑋, X), 𝜕∗ ) is called the cellular complex of (𝑋, X).

137
3. Homology Theory

Proof. The diagram

𝜕
𝐻𝑛 (𝑋 𝑛 , 𝑋 𝑛−1 ) / 𝐻𝑛−1 (𝑋 𝑛−1 )

= 𝜕𝑛
 *
𝜕
𝐻𝑛−1 (𝑋 𝑛−1 ) / 𝐻𝑛−1 (𝑋 𝑛−1 , 𝑋 𝑛−2 ) /
2 𝐻𝑛−1 (𝑋
𝑛−2 )

=0 = 𝜕𝑛−1
 *
𝐻𝑛−1 (𝑋 𝑛−2 ) / 𝐻𝑛−2 (𝑋 𝑛−2 , 𝑋 𝑛−3 )

shows 𝜕𝑛−1 ◦ 𝜕𝑛 = 0 which is the claim. 

Lemma 3.95. For 𝑝 ≥ 𝑞 ≥ 𝑛 or 𝑛 > 𝑝 ≥ 𝑞 we have 𝐻𝑛 (𝑋 𝑝 , 𝑋 𝑞 ) = 0.

Proof. The proof is by induction on 𝑝 −𝑞. The assertion is certainly true for 𝑝 −𝑞 = 0. To analyze
the situation 𝑝 − 𝑞 > 0 we look at the exact homology sequence for the triple (𝑋 𝑝 , 𝑋 𝑞+1 , 𝑋 𝑞 ):
ind. hyp.
𝐻𝑛 (𝑋 𝑞+1 , 𝑋 𝑞 ) / 𝐻𝑛 (𝑋 𝑝 , 𝑋 𝑞 ) / 𝐻𝑛 (𝑋 𝑝 , 𝑋 𝑞+1 ) = 0.
Since either 𝑞 ≥ 𝑛 or 𝑞 < 𝑝 < 𝑛 we have 𝑛 ≠ 𝑞 + 1. Thus 𝐻𝑛 (𝑋 𝑞+1 , 𝑋 𝑞 ) = 0 by Lemma 3.92
and hence 𝐻𝑛 (𝑋 𝑝 , 𝑋 𝑞 ) = 0. 

Corollary 3.96. For 𝑛 > 𝑝 we have that 𝐻𝑛 (𝑋 𝑝 ) = 0.

Proof. Lemma 3.95 with 𝑞 = 0 says 𝐻𝑛 (𝑋 𝑝 , 𝑋 0 ) = 0. The claim now follows from the exact
sequence
0 = 𝐻𝑛 (𝑋 0 ) −→ 𝐻𝑛 (𝑋 𝑝 ) −→ 𝐻𝑛 (𝑋 𝑝 , 𝑋 0 ) = 0.

Corollary 3.97. For 𝑞 ≥ 𝑛 we have 𝐻𝑛 (𝑋, 𝑋 𝑞 ) = 0.

Proof. Choose 𝑝 ≥ 𝑞 so large that 𝑋 𝑝 = 𝑋 and use Lemma 3.95. 

Corollary 3.98. For 𝑟 > 𝑛 the inclusion 𝑋 𝑟 ↩→ 𝑋 induces an isomorphism

𝐻𝑛 (𝑋)  𝐻𝑛 (𝑋 𝑟 ) .

138
 3.11. Homology of CW-complexes

Proof. The assertion follows from Corollary 3.97 and the exact sequence

0 = 𝐻𝑛+1 (𝑋, 𝑋 𝑟 ) −→ 𝐻𝑛 (𝑋 𝑟 ) −→ 𝐻𝑛 (𝑋) −→ 𝐻𝑛 (𝑋, 𝑋 𝑟 ) = 0.

Lemma 3.99. For 𝑟 > 𝑛 and 𝑟 ≥ 𝑞 the inclusion induces an isomorphism

𝐻𝑛 (𝑋, 𝑋 𝑞 )  𝐻𝑛 (𝑋 𝑟 , 𝑋 𝑞 ) .

Proof. Since 𝑟 ≥ 𝑛 + 1, Corollary 3.97 gives us 𝐻𝑛+1 (𝑋, 𝑋 𝑟 ) = 𝐻𝑛 (𝑋, 𝑋 𝑟 ) = 0. The assertion
now follows from the exact homology sequence of the triple (𝑋, 𝑋 𝑟 , 𝑋 𝑞 ):

𝐻𝑛+1 (𝑋, 𝑋 𝑟 ) −→ 𝐻𝑛 (𝑋 𝑟 , 𝑋 𝑞 ) −→ 𝐻𝑛 (𝑋, 𝑋 𝑞 ) −→ 𝐻𝑛 (𝑋, 𝑋 𝑟 ).

Theorem 3.100. We have the following isomorphism:

𝐻𝑛 𝐾∗ (𝑋, X)  𝐻𝑛 (𝑋)

Proof. Consider the commutative diagram with exact columns and rows

𝐻𝑛+1 (𝑋 𝑛+1 , 𝑋
❘
𝑛) 𝐻𝑛−1 (𝑋 𝑛−2 ) = 0
❘❘❘
❘❘❘𝜕𝑛
𝜕 ❘❘❘
❘❘❘
 ( 
𝑖∗
0 = 𝐻𝑛 (𝑋 𝑛−1 ) / 𝐻𝑛 (𝑋 ) 𝑛 / 𝐻𝑛 (𝑋 𝑛 , 𝑋 𝑛−1 )
❘❘❘
/ 𝐻𝑛−1 (𝑋 𝑛−1 )
❘❘❘𝜕𝑛−1
❘❘❘
❘❘❘
 ❘) 
𝐻𝑛 (𝑋 𝑛+1 ) 𝐻𝑛−1 (𝑋 𝑛−1 , 𝑋 𝑛−2 )


𝐻𝑛 (𝑋 𝑛+1 , 𝑋 𝑛 ) = 0

Now we compute

𝐻𝑛 (𝑋)  𝐻𝑛 (𝑋 𝑛+1 )
𝐻𝑛 (𝑋 𝑛 )

𝜕𝐻𝑛+1 (𝑋 𝑛+1 , 𝑋 𝑛 )

139
3. Homology Theory

𝑖∗ 𝐻𝑛 (𝑋 𝑛 )

𝑖 ∗ 𝜕𝐻𝑛+1 (𝑋 𝑛+1 , 𝑋 𝑛 )
ker(𝐻𝑛 (𝑋 𝑛 , 𝑋 𝑛−1 ) → 𝐻𝑛−1 (𝑋 𝑛−1 ))

𝜕𝑛 𝐻𝑛+1 (𝑋 𝑛+1 , 𝑋 𝑛 )
ker(𝜕𝑛−1 )

𝜕𝑛 𝐻𝑛+1 (𝑋 𝑛+1 , 𝑋 𝑛 )
= 𝐻𝑛 𝐾∗ (𝑋, X)

and the theorem is proved. 

In the following examples we set 𝛼𝑛 := |X𝑛 |.

Example 3.101. We consider 𝑋 = 𝑆 𝑛 for 𝑛 ≥ 2. The CW-decomposition of 𝑆 𝑛 from Exam-

ple 3.86 has
𝛼0 = 𝛼𝑛 = 1, 𝛼 𝑗 = 0 otherwise.
Hence we have to compute the homology of the complex

𝑅 ←− 0 ←− · · · ←− 0 ←− 𝑅 ←− 0 ←− · · ·

Since all arrows are 0 the homology coincides with the complex, hence
(
𝑛 𝑅, 𝑗 = 0 or 𝑛,
𝐻 𝑗 (𝑆 ) 
0, otherwise,

which confirms our earlier findings.

Example 3.102. Now look at 𝑋 = CP𝑛 . Then we have for the CW-decomposition from Exam-
ple 3.89
𝛼0 = 𝛼2 = . . . = 𝛼2𝑛 = 1, 𝛼 𝑗 = 0 otherwise.
Again all arrows in the complex

𝑅 ←− 0 ←− · · · ←− 0 ←− 𝑅 ←− 0 ←− · · ·

must be zero, hence the homology is given by

(
𝑅, 𝑗 = 0, 2, . . . , 2𝑛,
𝐻 𝑗 (CP𝑛 ) 
0, otherwise.

Example 3.103. Consider 𝑋 = HP𝑛 . For the CW-decomposition from Example 3.90 we have

𝛼0 = 𝛼4 = 𝛼8 = . . . = 𝛼4𝑛 = 1, 𝛼 𝑗 = 0 otherwise

and all arrows in

0 ←− 𝑅 ←− 0 ←− 0 ←− 0 ←− 𝑅 ←− · · · ←− 0 ←− 𝑅 ←− 0 ←− · · ·

140
 3.12. Betti numbers and the Euler number

must be zero. We find for the homology

(
𝑅, 𝑗 = 0, 4, 8, . . . , 4𝑛,
𝐻 𝑗 (HP𝑛 ) 
0, otherwise.

3.12. Betti numbers and the Euler number

Throughout this section let 𝑅 be a field.

Definition 3.104. The dimension 𝑏 𝑗 (𝑋; 𝑅) := dim 𝑅 𝐻 𝑗 (𝑋; 𝑅) is called 𝑗-th Betti number of
the space 𝑋 (over the field 𝑅).

Now let (𝑋, X) be a finite CW-complex. Denote the number of 𝑗-cells in (𝑋, X) by 𝛼 𝑗 . Then
we get the following estimate for the Betti numbers:
𝑏 𝑗 (𝑋; 𝑅) = dim 𝑅 𝐻 𝑗 (𝑋; 𝑅)
ker(𝜕 𝑗 : 𝐾 𝑗 (𝑋, X) → 𝐾 𝑗−1 (𝑋, X))
= dim 𝑅
im(𝜕 𝑗+1 : 𝐾 𝑗+1 (𝑋, X) → 𝐾 𝑗 (𝑋, X))
≤ dim 𝑅 ker(𝜕 𝑗 : 𝐾 𝑗 (𝑋, X) → 𝐾 𝑗−1 (𝑋, X))
≤ dim 𝑅 𝐾 𝑗 (𝑋, X)
= 𝛼𝑗.
We conclude that 𝑏 𝑗 (𝑋; 𝑅) ≤ 𝛼 𝑗 . In particular, the Betti numbers are finite, 𝑏 𝑗 (𝑋; 𝑅) < ∞.

Definition 3.105. For a finite CW-complex (𝑋, X) we call

Õ
∞
𝜒(𝑋, X) = (−1) 𝑖 𝛼𝑖
𝑖=0

the Euler number or Euler-Poincaré characteristic.

Note that the sum in this definition is finite. It ends at the highest dimension occuring in the cell
decomposition.

Proposition 3.106. We have the following relation between Euler and Betti numbers:
Õ
∞
𝜒(𝑋, X) = (−1) 𝑖 𝑏𝑖 (𝑋; 𝑅).
𝑖=0

141
3. Homology Theory

In particular, the Euler number does not depend on the cell decomposition because the Betti
numbers don’t. On the other hand, the Euler number does not depend on the coefficient field
because the 𝛼𝑖 ’s don’t. We will henceforth write 𝜒(𝑋) instead of 𝜒(𝑋, X).
In order to prove the proposition we use the following

Lemma 3.107. Let

𝑑1 𝑑2 𝑑𝑛
0o 𝑉0 o 𝑉1 o ··· o 𝑉𝑛 o 0o 0o ···

be a complex of finite-dimensional 𝑅-vector spaces. Then

Õ
∞ Õ
∞
(−1) 𝑗 dim 𝐻 𝑗 𝑉∗ = (−1) 𝑗 dim 𝑉 𝑗 .
𝑗=0 𝑗=0

Proof of Lemma 3.107. To show the lemma we compute, using the dimension formula from
linear algebra,
Õ Õ
(−1) 𝑗 dim 𝑉 𝑗 = (−1) 𝑗 (dim(𝑑 𝑗 𝑉 𝑗 ) + dim ker(𝑑 𝑗 ))
Õ
𝑗 𝑗

= (−1) 𝑗 (dim ker(𝑑 𝑗 ) − dim(𝑑 𝑗+1𝑉 𝑗+1 ))

 
𝑗
Õ ker(𝑑 𝑗 )
𝑗
= (−1) dim
𝑑 𝑗+1 𝑉 𝑗+1
Õ
𝑗

= (−1) 𝑗 dim 𝐻 𝑗 𝑉∗ . 
𝑗

Proof of Proposition 3.106. This follows from Lemma 3.107 with 𝑉 𝑗 = 𝐾 𝑗 (𝑋, X)  𝑅 𝛼 𝑗 . 

Example 3.108. For 𝑋 = 𝑆 𝑛 we have

𝑏0 = 𝑏 𝑛 = 1, 𝑏 𝑗 = 0 otherwise

and therefore (
2, 𝑛 even
𝜒(𝑆 𝑛 ) =
0, 𝑛 odd .
The case 𝑛 = 2 contains Euler’s classical formula for polyhedra as a special case. It says that for
the alternating sum of the number of vertices, edges and faces of a polyhedron is always equal to
2. In particular, for platonic solids we have the following list:

142
 3.13. Incidence numbers

𝛼0 𝛼1 𝛼2
Tetrahedron 4 6 4
Cube 8 12 6
Octahedron 6 12 8
Dodecahedron 20 30 12
Icosahedron 12 30 20

Figure 90. Platonic solids

Examples 3.109. We compute the Euler numbers of the projective spaces.

1.) For 𝑋 = CP𝑛 we have 𝑏0 = 𝑏2 = . . . = 𝑏2𝑛 = 1 and 𝑏 𝑗 = 0 otherwise. Hence 𝜒(CP𝑛 ) = 𝑛 +1.

2.) For 𝑋 = HP𝑛 we have 𝑏0 = 𝑏4 = . . . = 𝑏4𝑛 = 1 and 𝑏 𝑗 = 0 otherwise. Hence 𝜒(HP𝑛 ) = 𝑛+1.

3.) In the case of 𝑋 = RP𝑛 we do not know the Betti numbers yet. So we use Proposition 3.106
to compute the Euler number. For the cell decomposition described in Example 3.88 we have

𝛼0 = . . . = 𝛼𝑛 = 1, 𝛼 𝑗 = 0 otherwise.

Hence (
1, 𝑛 even,
𝜒(RP𝑛 ) =
0, 𝑛 odd.

3.13. Incidence numbers

We return to a general commutative ring 𝑅 with unit 1. Throughout this section let (𝑋, X) be
a finite CW-complex. In order to compute the cellular homology of (𝑋, X) we need a better
understanding of the homomorphism 𝜕𝑛+1 : 𝐾𝑛+1 (𝑋, X) → 𝐾𝑛 (𝑋, X). In the case of 𝑛 = 0 we
find
𝜕1
𝐾1 (𝑋, X) / 𝐾0 (𝑋, X)

𝜕
𝐻1 (𝑋 1 , 𝑋 0 ) / 𝐻0 (𝑋 0 )


É É
𝐻1 (𝜑 𝜏 ) (𝐻1 (𝐷 1 , 𝑆0 )) 𝑅
𝜏 ∈ X1 𝜎 ∈ X0

143
3. Homology Theory

Hence 𝜕1 is given by the (𝛼0 × 𝛼1 )-matrix (𝜕𝜎𝜏 ) where 𝜏 ∈ X1 and 𝜎 ∈ X0 . The entries 𝜕𝜎𝜏 ∈ 𝑅
of this matrix are easily computed. Namely, recall that a generator of 𝐻1 (𝐷 1 , 𝑆0 ) is represented
by 𝑐 : Δ1 = [0, 1] → 𝐷 1 = [−1, 1] with 𝑐(𝑡) = 2𝑡 − 1. Then

𝜕1 𝐻1 (𝜑 𝜏 ) ([𝑐]) = 𝜕 [𝜑 𝜏 ◦ 𝑐] = 𝜑 𝜏 (𝑐(1)) − 𝜑 𝜏 (𝑐(0)) = 𝜑 𝜏 (1) − 𝜑 𝜏 (−1).

Thus if 𝜑 𝜏 (−1) = 𝜑 𝜏 (1) then 𝜕𝜎𝜏 = 0 for all 𝜎. If 𝜑 𝜏 (−1) ≠ 𝜑 𝜏 (1) then






1, for 𝜎 = 𝜑 𝜏 (1),
𝜏

𝜕𝜎 = −1, for 𝜎 = 𝜑 𝜏 (−1),

 0,
 otherwise.

Example 3.110. We compute the homology of the following CW-complex consisting of two
0-cells and three 1-cells:
𝜏2
𝜏1 𝜎1 b b 𝜎2
𝜏3
Figure 91. A cell decomposition of the figure 8

Clearly, 𝜕𝜎𝜏11 = 𝜕𝜎𝜏12 = 0. Depending on how the characteristic maps parametrize the 1-cells 𝜏2
and 𝜏3 we get
𝜕𝜎𝜏21 = 𝜕𝜎𝜏31 = 1 and 𝜕𝜎𝜏22 = 𝜕𝜎𝜏32 = −1,

or possibly different signs which will not affect the homology however. Thus the cellular complex
is
©0 1 1ª
­ ®
0 −1 −1¬
··· o 0o 𝑅 2 «o 𝑅3 o 0o ···

The image of the matrix is {(𝑥, −𝑥) | 𝑥 ∈ 𝑅} = 𝑅 · (1, −1) and hence

𝐻0 (𝑋; 𝑅)  𝑅 2 /𝑅 · (1, −1)  𝑅

where the latter isomorphism is induced by 𝑅 2 → 𝑅, (𝑥, 𝑦) ↦→ 𝑥 + 𝑦. Moreover,

 
0 1 1
𝐻1 (𝑋; 𝑅)  ker = {(𝑥, 𝑦, −𝑦) | 𝑥, 𝑦 ∈ 𝑅}  𝑅 2 .
0 −1 −1

We summarize





𝑅 if 𝑗 = 0,
𝐻 𝑗 (𝑋; 𝑅)  𝑅 2

if 𝑗 = 1,

0
 else.

144
 3.13. Incidence numbers

Returning to our general CW-complex (𝑋, X) we find for 𝑛 ≥ 1:

𝜕𝑛+1
𝐾𝑛+1 (𝑋, X) / 𝐾𝑛 (𝑋, X)

𝐻𝑛+1 (𝑋 𝑛+1 , 𝑋 𝑛 ) 𝐻𝑛 (𝑋 𝑛 , 𝑋 𝑛−1 )

É É
𝐻𝑛+1 (𝜑 𝜏 ) (𝐻𝑛+1 (𝐷 𝑛+1 , 𝑆 𝑛 )) 𝐻𝑛 (𝜑 𝜎 ) (𝐻𝑛 (𝐷 𝑛 , 𝑆 𝑛−1 ))
𝜏 ∈ X𝑛+1 𝜎 ∈ X𝑛
 
 𝜏 )
(𝜕𝜎 
𝑅 𝛼𝑛+1 / 𝑅 𝛼𝑛

Hence 𝜕𝑛+1 is given by the (𝛼𝑛 × 𝛼𝑛+1 )-matrix (𝜕𝜎𝜏 ) where 𝜏 ∈ X𝑛+1 and 𝜎 ∈ X𝑛 . The entries
(𝜕𝜎𝜏 ) ∈ 𝑅 of this matrix are called the incidence numbers. We want to see how we can compute
them.
Fix 𝜎 ∈ X𝑛 , 𝜏 ∈ X𝑛+1 , and 𝑝 ∈ 𝐷˚ 𝑛 . From the commutative diagram

𝜕𝑛+1

* pr 𝜎
𝜕
𝐻𝑛+1 (𝑋 𝑛+1
O
, 𝑋 𝑛) / 𝐻𝑛 (𝑋 𝑛 )
O
/ 𝐻𝑛 (𝑋 𝑛 , 𝑋 𝑛−1 ) / 𝐻𝑛 (𝜑 𝜎 )𝐻𝑛 (𝐷 𝑛 , 𝑆 𝑛−1 )
O
𝐻𝑛 ( 𝜑 𝜎 ) 

𝐻𝑛+1 ( 𝜑 𝜏 ) 𝐻𝑛 ( 𝜑 𝜏 | 𝑆 𝑛 ) 𝐻𝑛 (𝐷 𝑛O , 𝑆 𝑛−1 )

deg 𝑝 ( 𝜑 −1 −1 𝑛 𝑛
𝜕 𝜎 ◦𝜑 𝜏 :𝜑 𝜏 ( 𝜎)→𝐷 ⊂𝑆 )
𝐻𝑛+1 (𝐷 𝑛+1 , 𝑆 𝑛 ) 
/ 𝐻𝑛 (𝑆 𝑛 ) / 𝐻𝑛 (𝑆 𝑛 )
O
𝐻𝑛 ( 𝜑 −1
𝜎 ◦𝜑 𝜏 )


𝐻𝑛 (𝑆 𝑛 , 𝑆 𝑛 \ 𝜑 −1
𝜏 (𝜑 𝜎 ( 𝑝)))
o 𝐻𝑛 (𝜑 −1 −1 −1
𝜏 (𝜎), 𝜑 𝜏 (𝜎) \ 𝜑 𝜏 (𝜑 𝜎 ( 𝑝)))

we conclude that

𝜕𝜎𝜏 = deg 𝑝 (𝜑 −1 −1 𝑛 𝑛
𝜎 ◦ 𝜑 𝜏 | 𝜑 𝜏−1 ( 𝜎) : 𝜑 𝜏 (𝜎) → 𝐷 ⊂ 𝑆 ).

We used the canonical isomorphism 𝐻𝑛 (𝑆 𝑛 )  𝐻𝑛 (𝐷 𝑛 , 𝑆 𝑛−1 ). We have intepreted the incidence

numbers as certain local mapping degrees. In applications they can often be computed by
counting preimages.

Example 3.111. Look at 𝑋 = RP𝑛 . We want to compute the homology of

𝜕1 𝜕2 𝜕3 𝜕𝑛
0o 𝑅o 𝑅o 𝑅o ··· o 𝑅o 0o ···

The operator 𝜕 𝑗+1 is given by the (1 × 1)-matrix (𝜕𝜎𝜏 ) with 𝜎 ∈ X 𝑗 and 𝜏 ∈ X 𝑗+1 . The gluing map
𝜑 𝜏 | 𝑆 𝑗 : 𝑆 𝑗 → 𝑋 𝑗 ≈ RP 𝑗 is given by the canonical projection. The image point 𝜑 𝜎 ( 𝑝) ∈ 𝑋 𝑗 has

145
3. Homology Theory

exactly two preimages under 𝜑 𝜏 which we call 𝑥, −𝑥 ∈ 𝑆 𝑗 . Put Φ := 𝜑 −1 −1 𝑗

𝜎 ◦ 𝜑 𝜏 : 𝜑 𝜏 (𝜎) → 𝐷 .
From the additivity of the local degree we obtain
𝑗 𝑗 𝑗
deg 𝑝 (Φ : 𝜑 −1
𝜏 (𝜎) → 𝐷 ) = deg 𝑝 (Φ 𝑈 ( 𝑥) : 𝑈 (𝑥) → 𝐷 ) + deg 𝑝 (Φ 𝑈 (− 𝑥) : 𝑈 (−𝑥) → 𝐷 )

where 𝑈 (𝑥) is a small neighborhood of 𝑥. W.l.o.g. we assume 𝑈 (−𝑥) = −𝑈 (𝑥). Since Φ 𝑈 ( 𝑥) is

a homeomorphism onto its image we have
deg 𝑝 (Φ 𝑈 ( 𝑥) : 𝑈 (𝑥) → 𝐷 𝑗 ) = ±1 =: 𝜀 .

Let 𝑎 : 𝑆 𝑗 → 𝑆 𝑗 be the antipodal map. Then Φ = Φ ◦ 𝑎 and hence

deg 𝑝 (Φ 𝑈 (− 𝑥) : 𝑈 (−𝑥) → 𝐷 𝑗 ) = deg 𝑝 (Φ ◦ 𝑎 : 𝑈 (−𝑥) → 𝐷 𝑗 )
= deg 𝑝 (Φ : 𝑈 (𝑥) → 𝐷 𝑛 ) · deg(𝑎)
= 𝜀 · (−1) 𝑗+1 .
We conclude that 𝜕 𝑗+1 = 𝜀 · (1 + (−1) 𝑗+1 ). For 𝑛 even we obtain the complex
0 ±2 0 ±2
0o 𝑅o 𝑅o ··· o 𝑅o 𝑅o 0
whereas for 𝑛 odd we find
0 ±2 0
0o 𝑅o 𝑅o ··· o 𝑅o 𝑅o 0.
For 𝑅 = Z/2Z all arrows are zero so that
(
Z/2Z, 𝑗 = 0, . . . , 𝑛,
𝐻 𝑗 (RP𝑛 ; Z/2Z) =
0, otherwise.
For 𝑅 = Z the homology is computed to




Z, 𝑗 = 0,






Z/2Z, 𝑗 = 1, 3, 𝑛 − 1 or 𝑛 − 2 resp.,
𝑛

𝐻 𝑗 (RP ; Z) = Z, 𝑗 = 𝑛 odd,





0, 𝑗 = 𝑛 even,

 0,
 otherwise.
For 𝑅 = Q we find (
Q, 𝑗 = 0 or 𝑗 = 𝑛 odd,
𝐻 𝑗 (RP𝑛 ; Q) =
0, else.

3.14. Homotopy versus homology

The final goal of this chapter is to compare homotopy groups and homology groups. We start by
examining the fundamental group of CW-complexes.

146
 3.14. Homotopy versus homology

Proposition 3.112. Let (𝑋, X) be a finite CW-complex and let 𝑥0 ∈ 𝑋 0 . Then the inclusion
map 𝑗 : 𝑋 2 ↩→ 𝑋 induces an isomorphism

𝑗 # : 𝜋1 (𝑋 2 ; 𝑥0 ) → 𝜋1 (𝑋; 𝑥0 ) .

Proof. We have to show that attaching a 𝑘−cell for 𝑘 ≥ 3 does not alter 𝜋1 in the sense that the
inclusion induces an isomorphism on 𝜋1 . We assume that 𝑋 2 is path-connected, since otherwise
we may replace 𝑋 2 by the path-component that contains 𝑥0 .
Let 𝑌 be a path-connected finite CW-complex and let 𝑌˜ be obtained from 𝑌 by attaching a 𝑘-cell.
More precisely, 𝑌˜ = 𝑌 ∪ 𝜑 𝐷 𝑘 = 𝑌 ⊔ 𝐷 𝑘 /∼ where 𝑥 ∼ 𝜑(𝑥) for all 𝑥 ∈ 𝑆 𝑘−1 . Here 𝜑 : 𝑆 𝑘−1 → 𝑌
is a continuous map. We have to show that the inclusion map induces an isomorphism
˜ 𝑥0 )
𝑗 # : 𝜋1 (𝑌 ; 𝑥0 ) → 𝜋(𝑌;
if 𝑘 ≥ 3. Let 𝐷 𝑘 ( 21 ) ⊂ 𝐷˚ 𝑘 be the closed 𝑘-dimensional subball of radius 12 . Cover 𝑌˜ by the two
open subsets 𝑈1 = 𝐷˚ 𝑘 and 𝑈2 = 𝑌˜ \ 𝐷 𝑘 ( 21 ) ≃ 𝑌 .

𝑌 𝑈2

Figure 92. Attaching a cell of dimension ≥ 3

We then find 𝑈1 ∩ 𝑈2 = 𝐷˚ 𝑘 \ 𝐷 𝑘 ( 21 ) ≃ 𝑆 𝑘−1 . To apply the Seifert-van-Kampen Theorem 2.68

we calculate
𝜋1 (𝑈1 ) = {1},
𝜋1 (𝑈2 )  𝜋1 (𝑌 ),
𝜋1 (𝑈1 ∩ 𝑈2 )  𝜋1 (𝑆 𝑘−1 ) = {1}, (here we use 𝑘 − 1 ≥ 2)
and we find
𝜋1 (𝑈1 ) ★ 𝜋1 (𝑈2 )
𝜋1 (𝑌˜ )   𝜋1 (𝑌 )
im 𝜋1 (𝑈1 ∩ 𝑈2 )

147
3. Homology Theory

the isomorphisms being induced by inclusions. 

Example 3.113. Consider complex-projective space 𝑋 = CP𝑛 . We use the cell decomposition
from Example 3.89:
𝑋 0 = {point} = 𝑋 1 ⊂ 𝑋 2 = 𝑋 3 ⊂ 𝑋 4 = 𝑋 5 ⊂ . . . ⊂ 𝑋 2𝑛−1 ⊂ 𝑋 2𝑛
|{z} |{z} |{z}
≈CP1 ≈CP2 =CP𝑛

We are now able to calculate the fundamental group

𝜋1 (CP𝑛 )  𝜋1 (𝑋 2 )  𝜋1 (CP1 ) = 𝜋1 (𝑆2 ) = {1}.
Hence complex-projective space CP𝑛 is simply connected.

Example 3.114. Similarly, for 𝑋 = HP𝑛 we use the cell decomposition from Example 3.90:
𝑋 0 = {point} = 𝑋 1 = 𝑋 2 = 𝑋 3 ⊂ 𝑋 4 = . . . ⊂ 𝑋 4𝑛−3 = . . . ⊂ 𝑋 4𝑛
|{z} |{z}
≈HP1 =HP𝑛

For the fundamental group we find

𝜋1 (HP𝑛 ) = 𝜋1 (𝑋 2 ) = 𝜋1 ({point}) = {1}.
Thus HP𝑛 is also simply connected.

Remark 3.115. Proposition 3.112 can be generalized as follows: For a finite CW-complex the
inclusion map 𝑗 : 𝑋 𝑛+1 ↩→ 𝑋 always induces an isomorphism 𝑗# : 𝜋 𝑛 (𝑋 𝑛+1 ; 𝑥0 ) → 𝜋 𝑛 (𝑋; 𝑥0 )
where 𝑥0 ∈ 𝑋 0 .

Now we relate homotopy and homology groups. Recall that for the 𝑛-dimensional cube
𝑊 𝑛 = [0, 1] 𝑛 we have (𝑊 𝑛 , 𝜕𝑊 𝑛 ) ≈ (𝐷 𝑛 , 𝑆 𝑛−1 ). Fix a generator 𝛼𝑛 ∈ 𝐻𝑛 (𝑊 𝑛 , 𝜕𝑊 𝑛 ; Z) 
𝐻𝑛 (𝐷 𝑛 , 𝑆 𝑛−1 ; Z)  Z. The elements of 𝜋 𝑛 (𝑋; 𝑥0 ) are homotopy classes relative to 𝜕𝑊 𝑛 of
maps 𝑓 : 𝑊 𝑛 → 𝑋 with 𝑓 (𝜕𝑊 𝑛 ) = {𝑥0 }. Then 𝐻𝑛 ( 𝑓 ) (𝛼𝑛 ) ∈ 𝐻𝑛 (𝑋, {𝑥0 }; Z). The long exact
homology sequence of the pair (𝑋, {𝑥0 }) yields for 𝑛 ≥ 1

/ 𝐻𝑛 (𝑋; Z) / 𝐻𝑛 (𝑋, {𝑥 0 }; Z) 0 / 𝐻𝑛−1 ({𝑥 0 }; Z).

0 = 𝐻𝑛 ({𝑥0 }; Z)
Namely, if 𝑛 ≥ 2 then 𝐻𝑛−1 ({𝑥0 }; Z) = 0. For 𝑛 = 1 the arrow emanating from 𝐻0 ({𝑥0 }; Z)
is injective so that the incoming arrow must again be zero. In either case the inclusion

𝑗 : (𝑋, ∅) ↩→ (𝑋, {𝑥0 }) induces an isomorphism 𝐻𝑛 ( 𝑗) : 𝐻𝑛 (𝑋; Z) −→ 𝐻𝑛 (𝑋, {𝑥0 }; Z). Now
set
ℎ([ 𝑓 ]) := 𝐻𝑛 ( 𝑗) −1 𝐻𝑛 ( 𝑓 ) (𝛼𝑛 ) ∈ 𝐻𝑛 (𝑋; Z) .
Due to homotopy invariance the expression ℎ([ 𝑓 ]) only depends on the homotopy class of the
map 𝑓 . Hence we have constructed a well-defined map
ℎ : 𝜋 𝑛 (𝑋; 𝑥0 ) → 𝐻𝑛 (𝑋; Z), 𝑛 ≥ 1.

148
 3.14. Homotopy versus homology

Proposition 3.116. The map ℎ : 𝜋 𝑛 (𝑋; 𝑥0 ) → 𝐻𝑛 (𝑋; Z) is a homomorphism.

Proof. Consider the map 𝑠1 : 𝑊1𝑛 = [0, 21 ] × [0, 1] 𝑛−1 → 𝑊 𝑛 given by (𝑥1 , . . . , 𝑥 𝑛 ) ↦→
(2𝑥1 , 𝑥2 , . . . , 𝑥 𝑛 ) and the map 𝑠2 : 𝑊2𝑛 = [ 21 , 1] × [0, 1] 𝑛−1 → 𝑊 𝑛 defined by (𝑥1 , . . . , 𝑥 𝑛 ) ↦→
(2𝑥1 − 1, 𝑥2 , . . . , 𝑥 𝑛 ).
𝑊𝑛 𝑊1𝑛 𝑊2𝑛 𝑊 𝑛
𝑠1

𝑠2

Figure 93. The maps 𝑠1 and 𝑠2

For [ 𝑓1 ], [ 𝑓2 ] ∈ 𝜋 𝑛 (𝑋; 𝑥0 ) we have [ 𝑓1 ] + [ 𝑓2 ] = [𝑔] with

(
𝑓1 (𝑠1 (𝑥)), 𝑥1 ≤ 12 ,
𝑔(𝑥) =
𝑓2 (𝑠2 (𝑥)), 𝑥1 ≥ 12 .

We represent 𝛼𝑛 by 𝑐1 + 𝑐2 ∈ 𝑆𝑛 (𝑊 𝑛 ; Z), where 𝑐 𝜈 ∈ 𝑆 𝑛 (𝑊𝜈𝑛 ; Z) and 𝑐 𝜈 represents the generator

𝐻𝑛 (𝑠 𝜈 ) −1 (𝛼𝑛 ) of 𝐻𝑛 (𝑊𝜈𝑛 , 𝜕𝑊𝜈𝑛 ; Z).

𝑛 = 1: 𝑛 = 2:

𝜎1 𝜎3 𝑐1 = 𝜎1 + 𝜎2
𝑐2 = 𝜎3 + 𝜎4
𝜎2 𝜎4

𝑐1 𝑐2

Figure 94. Representative of 𝛼1 and 𝛼2

Now the proposition follows from

ℎ([ 𝑓1 ] · [ 𝑓2 ]) = ℎ([𝑔])
= 𝐻𝑛 ( 𝑗) −1 𝐻𝑛 (𝑔) (𝛼𝑛)
= 𝐻𝑛 ( 𝑗) −1 𝐻𝑛 (𝑔) ([𝑐1 + 𝑐2 ])
= 𝐻𝑛 ( 𝑗) −1 𝐻𝑛 (𝑔) ([𝑐1 ]) + 𝐻𝑛 ( 𝑗) −1 𝐻𝑛 (𝑔) ([𝑐2 ])
= 𝐻𝑛 ( 𝑗) −1 𝐻𝑛 ( 𝑓1 ◦ 𝑠1 ) ([𝑐1 ]) + 𝐻𝑛 ( 𝑗) −1 𝐻𝑛 ( 𝑓2 ◦ 𝑠2 ) ([𝑐2 ])

149
3. Homology Theory

= 𝐻𝑛 ( 𝑗) −1 𝐻𝑛 ( 𝑓1 )𝐻𝑛 (𝑠1 ) ([𝑐1 ]) + 𝐻𝑛 ( 𝑗) −1 𝐻𝑛 ( 𝑓2 )𝐻𝑛 (𝑠2 ) ([𝑐1 ])

= 𝐻𝑛 ( 𝑗) −1 𝐻𝑛 ( 𝑓1 ) (𝛼𝑛 ) + 𝐻𝑛 ( 𝑗) −1 𝐻𝑛 ( 𝑓2 ) (𝛼𝑛 )
= ℎ( 𝑓1 ) + ℎ( 𝑓2 ). 

Definition 3.117. The map ℎ : 𝜋 𝑛 (𝑋; 𝑥0 ) → 𝐻𝑛 (𝑋; Z) is called Hurewicz homomorphism.

Proposition 3.118. The Hurewicz homomorphism ℎ is natural, i.e., for every 𝑓 ∈ 𝐶 (𝑋, 𝑌 )
with 𝑓 (𝑥0 ) = 𝑦 0 the following diagram commutes:

ℎ / 𝐻𝑛 (𝑋; Z)
𝜋 𝑛 (𝑋; 𝑥0 )
𝑓# 𝐻𝑛 ( 𝑓 )
 ℎ

𝜋 𝑛 (𝑌 ; 𝑦 0 ) / 𝐻𝑛 (𝑌 ; Z)

Proof. The inclusion maps

𝑗 𝑋 : (𝑋, ∅) ↩→ (𝑋, {𝑥0 })
𝑗𝑌 : (𝑌 , ∅) ↩→ (𝑌 , {𝑦 0 })
satisfy
( 𝑗𝑌 ◦ 𝑓 ) (𝑥) = 𝑗𝑌 ( 𝑓 (𝑥)) = ( 𝑓 (𝑥), 𝑦 0 )
( 𝑓 ◦ 𝑗 𝑋 ) (𝑥) = 𝑓 ((𝑥, 𝑥0 )) = ( 𝑓 (𝑥), 𝑓 (𝑥0 )) = ( 𝑓 (𝑥), 𝑦 0 )
and therefore 𝑗𝑌 ◦ 𝑓 = 𝑓 ◦ 𝑗 𝑋 . On the level of homology groups we find
𝐻𝑛 ( 𝑗𝑌 ) ◦ 𝐻𝑛 ( 𝑓 ) = 𝐻𝑛 ( 𝑓 ) ◦ 𝐻𝑛 ( 𝑗 𝑋 )
and consequently
𝐻𝑛 ( 𝑓 ) ◦ 𝐻𝑛 ( 𝑗 𝑋 ) −1 = 𝐻𝑛 ( 𝑗𝑌 ) −1 ◦ 𝐻𝑛 ( 𝑓 ) .
Thus
(𝐻𝑛 ( 𝑓 ) ◦ ℎ) ([𝜎]) = 𝐻𝑛 ( 𝑓 ) ◦ 𝐻𝑛 ( 𝑗 𝑋 ) −1 ◦ 𝐻𝑛 (𝜎) (𝛼𝑛 )
= 𝐻𝑛 ( 𝑗𝑌 ) −1 ◦ 𝐻𝑛 ( 𝑓 ) ◦ 𝐻𝑛 (𝜎) (𝛼𝑛 )
= 𝐻𝑛 ( 𝑗𝑌 ) −1 ◦ 𝐻𝑛 ( 𝑓 ◦ 𝜎) (𝛼𝑛 )
= ℎ([ 𝑓 ◦ 𝜎])
= (ℎ ◦ 𝑓# ) ([𝜎])
as claimed. 

150
 3.14. Homotopy versus homology

Theorem 3.119 (Hurewicz). Let 𝑋 be a topological space, 𝑥0 ∈ 𝑋 and 𝑛 ≥ 2. Assume that

𝜋0 (𝑋; 𝑥0 ) = 𝜋1 (𝑋; 𝑥0 ) = . . . = 𝜋 𝑛−1 (𝑋; 𝑥0 ) = {1} .

Then
ℎ : 𝜋 𝑛 (𝑋; Z) → 𝐻𝑛 (𝑋; Z)
is an isomorphism.

In particular, we conclude that 𝐻1 (𝑋; Z) = . . . = 𝐻𝑛−1 (𝑋; Z) = 0. For a proof of this theorem
see e.g. [2, Sec. 4.2].

Remark 3.120. For 𝑛 = 1 this theorem cannot be true as it stands because 𝐻1 (𝑋; Z) is always
abelian while 𝜋1 (𝑋; 𝑥0 ) is not in general. However, ℎ induces a homomorphism

ℎ̄ : 𝜋1 (𝑋; 𝑥0 )/ [ 𝜋1 (𝑋;𝑥0 ), 𝜋1 (𝑋;𝑥0 ) ] → 𝐻1 (𝑋; Z).

Already Poincaré showed that if 𝜋0 (𝑋, 𝑥0 ) = {1} (i.e., 𝑋 is path-connected) then the map ℎ̄ is an
isomorphism.

Example 3.121. Consider 𝑋 = 𝑆 𝑛 for 𝑛 ≥ 2. We already know that 𝑆 𝑛 is 1-connected. Now

apply the Hurewicz isomorphism with 𝑛 = 2:
(
𝑛 𝑛 Z, 𝑛 = 2,
𝜋2 (𝑆 )  𝐻2 (𝑆 ; Z) =
0, 𝑛 ≥ 3.

For 𝑛 ≥ 3 we find due to Hurewicz:

(
𝑛 𝑛 Z, 𝑛 = 3,
𝜋3 (𝑆 )  𝐻3 (𝑆 ; Z) =
0, 𝑛 ≥ 4.

By induction we then deduce that 𝜋1 (𝑆𝑛 ) = . . . = 𝜋 𝑛−1 (𝑆 𝑛 ) = 0 and 𝜋 𝑛 (𝑆 𝑛 )  Z.

Example 3.122. We know that 𝑋 = CP𝑛 is 1-connected. With the help of the Hurewicz
isomorphism we calculate
𝜋2 (CP𝑛 )  𝐻2 (CP𝑛 ; Z)  Z .

Example 3.123. We also know that 𝑋 = HP𝑛 is 1-connected. We apply the Hurewicz isomor-
phism three times and we get

𝜋2 (HP𝑛 )  𝐻2 (HP𝑛 ; Z) = 0,
𝜋3 (HP𝑛 )  𝐻3 (HP𝑛 ; Z) = 0,
𝜋4 (HP𝑛 )  𝐻4 (HP𝑛 ; Z)  Z.

151
3. Homology Theory

Example 3.124. Now let us analyze the space 𝑋 = RP𝑛 for 𝑛 ≥ 2. The map 𝜓 : 𝑆 𝑛 → RP𝑛 is a
two-fold covering and by Theorem 2.102
𝜓♯
𝜋1 (𝑆 𝑛 ) −−→ 𝜋1 (RP𝑛 ) → 𝜋0 ({𝑝, 𝑞}) → 𝜋0 (𝑆 𝑛 )

is exact. Since 𝑆 𝑛 is simply connected 𝜋1 (RP𝑛 ) → 𝜋0 ({𝑝, 𝑞}) is an isomorphism of pointed

sets, thus 𝜋1 (RP𝑛 ) has exactly two elements. There is only one group with two elements, hence
𝜋1 (RP𝑛 )  Z/2Z. By Corollary 2.105 𝜓# : 𝜋 𝑘 (𝑆 𝑛 ) → 𝜋 𝑘 (RP𝑛 ) is an isomorphism for all 𝑘 ≥ 2.
We conclude that 𝜋2 (RP𝑛 ) = . . . = 𝜋 𝑛−1 (RP𝑛 ) = 0 and 𝜋 𝑛 (RP𝑛 )  Z.

Remark 3.125. Under the assumptions of the theorem of Hurewicz 3.119 not much can be
said about ℎ : 𝜋 𝑘 (𝑋, 𝑥0 ) → 𝐻 𝑘 (𝑋; Z) for 𝑘 > 𝑛. For example, consider the Hopf fibration
𝑆3 → 𝑆2 with fiber 𝑆1 . By Theorem 2.102 it induces an isomorphism 𝜋3 (𝑆2 )  𝜋3 (𝑆3 )  Z.
But 𝐻3 (𝑆2 ; Z) = 0 and hence ℎ : 𝜋3 (𝑆2 ) → 𝐻3 (𝑆2 ; Z) is not injective.
On the other hand, for the 2-torus 𝑇 2 we have again by Corollary 2.105 𝜋2 (𝑇 2 )  𝜋2 (R2 ) = 0
while one can compute 𝐻2 (𝑇 2 ; Z)  Z. Thus ℎ : 𝜋2 (𝑇 2 ) → 𝐻2 (𝑇 2 ; Z) is not surjective.

Remark 3.126. We have seen that 𝐻 𝑘 (𝑆 𝑛 ; Z) = 0 whenever 𝑘 > 𝑛. But in general this is not
true for the higher homotopy groups of the sphere, e.g. 𝜋3 (𝑆2 )  Z. The computation of 𝜋 𝑘 (𝑆 𝑛 )
for 𝑘 > 𝑛 is a difficult problem and many of these groups are not known to date.

3.15. Exercises
3.1. Let 𝑋 be a topological space. Show:
a) If 𝑋 is path-connected then
𝐻0 (𝑋; 𝑅)  𝑅

b) If 𝑋 𝑘 , 𝑘 ∈ 𝐾, are the path-components of 𝑋 then

Ê
𝐻𝑛 (𝑋; 𝑅)  𝐻𝑛 (𝑋 𝑘 ; 𝑅)
𝑘 ∈𝐾

3.2. Let 𝑌𝑘 = {1, ..., 𝑘} be equipped with the discrete topology. Compute 𝐻𝑛 (𝑌𝑘 ; 𝑅) without
using Exercise 3.1. Instead use the Eilenberg-Steenrod axioms.
Hint: Consider the pair (𝑌𝑘 , 𝑌𝑘−1 ).

3.3. Let 𝑋 be a topological space. The augmented boundary operator is defined by

𝜕 # : 𝑆0 (𝑋; 𝑅) → 𝑅,
Õ  Õ
𝜕# 𝛼𝑖 𝜎𝑖 = 𝛼𝑖 ,

where 𝜎𝑖 ∈ 𝐶 (Δ0 , 𝑋).

152
 3.15. Exercises

a) Verify 𝜕 # ◦ 𝜕 = 0.
b) Compute the augmented homology
ker(𝜕 # : 𝑆0 (𝑋; 𝑅) → 𝑅)
𝐻0# (𝑋; 𝑅) :=
im(𝜕 : 𝑆1 (𝑋; 𝑅) → 𝑆0 (𝑋; 𝑅))
for 𝑋 = {point}.

3.4. a) Show that homeomorphisms 𝜎 : Δ𝑛 → 𝐷 𝑛 represent generators of 𝐻𝑛 (𝐷 𝑛 , 𝑆 𝑛−1 ; 𝑅).

b) Describe generators of 𝐻𝑛 (𝑆 𝑛 ; Z). Make a drawing for 𝑛 = 2.

3.5 (Topological invariance of the dimension). Let 𝑈 ⊂ R𝑛 and 𝑉 ⊂ R𝑚 be open and

nonempty. Show: If 𝑈 and 𝑉 are homeomorphic then 𝑛 = 𝑚.
Hint: For 𝑝 ∈ 𝑈 and 𝑞 ∈ 𝑉 consider the pairs (𝑈, 𝑈 \ {𝑝}) and (𝑉, 𝑉 \ {𝑞}).

3.6. Let 𝑝 be a complex polynomial without zeros on the unit circle 𝑆1 ⊂ C. Show: The degree
of the map
𝑝(𝑧)
𝑝ˆ : 𝑆1 → 𝑆1 , 𝑝(𝑧)ˆ = ,
| 𝑝(𝑧)|
coincides with the number of zeros of 𝑝 in the interior of the unit disk (counted with multiplicities).

3.7 (Homotopy invariance of the local mapping degree). Let 𝑉 ⊂ 𝑆 𝑛 be open and 𝐹 : 𝑉 ×
Ð
[0, 1] → 𝑆 𝑛 continuous. We put 𝑓𝑡 (𝑥) := 𝐹 (𝑥, 𝑡). Let 𝑝 ∈ 𝑆 𝑛 such that 𝑡 ∈ [0,1] 𝑓𝑡−1 ( 𝑝) is
compact. Show:
deg 𝑝 ( 𝑓1 ) = deg 𝑝 ( 𝑓0 ).

3.8. Let (𝑋, 𝐴) be a pair such that 𝐴 is closed and a strong deformation retract of an open
neighborhood 𝑈. Show that 𝐻𝑛 (𝑋, 𝐴) = 𝐻𝑛 (𝑋/𝐴) for 𝑛 ≠ 0.

3.9. Let 𝑍 = 𝑆1 × [0, 1] be the cylinder. Compute 𝐻𝑛 (𝑍, 𝑆1 × {0} ∪ 𝑆1 × {1}) for all 𝑛. Sketch
generators of the nontrivial homology groups.
Hint: Use the homology sequence of the triple
(𝑍, 𝑆1 × {0} ∪ 𝑆1 × {1}, 𝑆1 × {0}).

3.10. Let (𝑋, 𝐴) be a pair. Describe the 0th singular relative homology group 𝐻0 (𝑋, 𝐴).

3.11 (Five lemma). Let the rows in the following commutative diagram of abelian groups be
exact:
𝑓1 𝑓2 𝑓3 𝑓4
𝐴1 / 𝐴2 / 𝐴3 / 𝐴4 / 𝐴5

𝜑1 𝜑2 𝜑3 𝜑4 𝜑5
 𝑔1  𝑔2  𝑔3  𝑔4 
𝐵1 / 𝐵2 / 𝐵3 / 𝐵4 / 𝐵5

153
3. Homology Theory

Show that if 𝜑1 , 𝜑2 , 𝜑4 , and 𝜑5 are isomorphisms then so is 𝜑3 .

Hint: 𝜑3 is injective if 𝜑1 is surjective and 𝜑2 , 𝜑4 are injective; 𝜑3 is surjective if 𝜑5 is injective
and 𝜑2 , 𝜑4 are surjective.

3.12. Suppose
𝑓 𝑛+1 𝑓𝑛
· · · → 𝐺 𝑛+1 −−−−−−−−→ 𝐺 𝑛 −−−−−−→ 𝐺 𝑛−1 → · · ·
is a long exact sequence of abelian groups and
𝑓′ ′
𝑓𝑛
𝑛+1
· · · → 𝐺 ′𝑛+1 −−−−−−−−→ 𝐺 ′𝑛 −−−−−−→ 𝐺 ′𝑛−1 → · · ·

is a subsequence, i.e., 𝐺 ′𝑛 ⊂ 𝐺 𝑛 and 𝑓𝑛′ = 𝑓𝑛 | 𝐺𝑛′ . Prove that the subsequence is exact if and only
if the quotient sequence

· · · → 𝐺 𝑛+1 /𝐺 ′𝑛+1 → 𝐺 𝑛 /𝐺 ′𝑛 → 𝐺 𝑛−1 /𝐺 ′𝑛−1 → · · ·

is exact.

3.13. Let 𝑛 ∈ N and 𝑚 ∈ Z. Show that there exists 𝑓 ∈ 𝐶 (𝑆 𝑛 , 𝑆 𝑛 ) with deg( 𝑓 ) = 𝑚.

3.14. Let 𝑓 ∈ 𝐶 (𝑆 𝑛 , 𝑆 𝑛 ) with 𝑓 (𝐷 +𝑛 ) ⊂ 𝐷 +𝑛 and 𝑓 (𝐷 𝑛− ) ⊂ 𝐷 𝑛− . We identify 𝑆 𝑛−1 with 𝐷 +𝑛 ∩ 𝐷 𝑛−

and therefore have 𝑓 (𝑆 𝑛−1 ) ⊂ 𝑆 𝑛−1 . Show

deg( 𝑓 ) = deg( 𝑓 | 𝑆 𝑛−1 ).

3.15. Show that 𝐻1 (R, Q; Z) is a free abelian group and find a basis as a Z-module.

3.16. Let 𝑀 be an 𝑛-dimensional manifold, 𝑛 ≥ 3. Let 𝑝 ∈ 𝑀. Show that the inclusion map
𝑀 \ {𝑝} ↩→ 𝑀 induces an isomorphism

𝐻 𝑗 (𝑀 \ {𝑝}; 𝑅)  𝐻 𝑗 (𝑀; 𝑅)

for all 𝑗 ∈ {1, ..., 𝑛 − 2}.

3.17. a) Show that for disjoint closed subsets 𝐴, 𝐵 ⊂ R2 we have

𝐻1 (R2 \ ( 𝐴 ∪ 𝐵))  𝐻1 (R2 \ 𝐴) ⊕ 𝐻1 (R2 \ 𝐵).

b) Let 𝑝 1 , ..., 𝑝 𝑛 ∈ R2 be pairwise distinct. Compute 𝐻1 (R2 \{𝑝 1 , ..., 𝑝 𝑛 }) and sketch generators.

3.18. Use the Mayer-Vietoris sequence to compute

a) the homology of the 2-torus;

b) the homology of surfaces 𝐹𝑔 of genus 𝑔 ≥ 2.

154
 3.15. Exercises

Sketch generators.

3.19. Let (𝑋, X) be a finite CW-complex. Show that 𝑋 is compact.

3.20. Describe a CW-decomposition for the surfaces of genus 𝑔 ≥ 1.

3.21. Show:

a) Each nonempty CW-complex has at least one 0-cell.

b) Each CW-complex consisting of exactly two cells is homeomorphic to a sphere.

3.22. Let 𝑛 ≥ 2 and 𝑘 ≥ 1. Let 𝑋 be the topological space obtained from 𝑘 copies of 𝑆 𝑛 by
identifying them all at one point. More formally,

Ø
𝑘


𝑋= { 𝑗 } × 𝑆𝑛 ∼
𝑗=1

where ( 𝑗, 𝑥) ∼ ( 𝑗 ′ , 𝑥 ′ ) iff 𝑥 = 𝑥 ′ = 𝑒1 . Compute the homology of 𝑋.

3.23. Find a CW-decomposition of the 2-torus with exactly one 0-cell, two 1-cells, and one
2-cell and use it to compute the homology.

155
A. Appendix

A.1. Free module generated by a set

Let 𝑅 be a commutative ring with unit 1 and let 𝑆 be a set. Then the set 𝑋 of all maps from 𝑆 to
𝑅 forms an 𝑅-module.1 Addition and multiplication with scalars are defined pointwise, for any
𝑓 , 𝑔 : 𝑆 → 𝑅 and 𝛼 ∈ 𝑅 we have, by definition,
( 𝑓 + 𝑔) (𝑠) = 𝑓 (𝑠) + 𝑔(𝑠), (𝛼 · 𝑓 ) (𝑠) = 𝛼 · 𝑓 (𝑠)
for all 𝑠 ∈ 𝑆.
Now let 𝑌 ⊂ 𝑋 be the set of all 𝑓 ∈ 𝑋 for which 𝑓 (𝑠) = 0 for all but finitely many 𝑠 ∈ 𝑆. Then 𝑌
is an 𝑅-submodule of 𝑋. The module 𝑌 has a natural basis. Namely, for each 𝑠 ∈ 𝑆 define
(
1, if 𝑠′ = 𝑠,
𝑓𝑠 (𝑠′ ) :=
0, if 𝑠′ ≠ 𝑠.
Then for each 𝑓 ∈ 𝑌 we have Õ
𝑓 = 𝑓 (𝑠) 𝑓𝑠 . (A.1)
𝑠∈𝑆
Note that we need to sum only over those 𝑠 for which 𝑓 (𝑠) ≠ 0 which leaves us with a finite sum.
Thus the set { 𝑓𝑠 | 𝑠 ∈ 𝑆} generates 𝑌 . The set is also linearly independent. Namely, if
𝛼1 𝑓𝑠1 + . . . + 𝛼𝑚 𝑓𝑠𝑚 = 0
for pairwise different 𝑠 𝑗 then by inserting 𝑠𝑖 we find
0 = (𝛼1 𝑓𝑠1 + . . . + 𝛼𝑚 𝑓𝑠𝑚 ) (𝑠𝑖 ) = 𝛼1 𝑓𝑠1 (𝑠𝑖 ) + . . . + 𝛼𝑚 𝑓𝑠𝑚 (𝑠𝑖 ) = 𝛼𝑖 .
Thus { 𝑓𝑠 | 𝑠 ∈ 𝑆} is indeed a basis of 𝑌 . Hence the dimension of 𝑌 is given by the cardinality of
𝑆. In particular, 𝑌 is infinite-dimensional if 𝑆 is an infinite set.

Definition A.1. The 𝑅-module 𝑌 is called the free 𝑅-module generated by 𝑆.

We usually use a somewhat sloppy notation and will not distinuish between an element 𝑠 ∈ 𝑆
and the corresponding function 𝑓𝑠 : 𝑆 → 𝑅. Thus, instead of (A.1) we will write
Õ
𝑓 = 𝑓 (𝑠)𝑠.
𝑠∈𝑆
1 If you
are not familiar with modules over rings think of the special case that 𝑅 is a field. Then an 𝑅-module is just
an 𝑅-vector space.

157
A. Appendix

In Section 3.1 we consider the free 𝑅-module generated by 𝑆 = 𝐶 (Δ𝑛 , 𝑋) and write 𝑆𝑛 (𝑋; 𝑅)
instead of 𝑌 .

158
List of Figures
1. (Non-) homeomorphic spaces . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
2. Hilbert’s curve2 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
3. Surface classification3 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
4. Open subset . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
5. Hausdorff property . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
6. Interval with endpoints identified . . . . . . . . . . . . . . . . . . . . . . . . . 11

7. Comb space . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18
8. 𝐴 is not a strong deformation retract of 𝑋 . . . . . . . . . . . . . . . . . . . . 19
9. Concatenation of paths . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20
10. Concatenations of homotopic paths are homotopic. . . . . . . . . . . . . . . . 21
11. Concatenation with constant path is homotopic to original path . . . . . . . . . 22
12. Inverse modulo homotopy . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22
13. Associativity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
14. Dependence on base point . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24
15. Auxiliary homotopy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26
16. The deformation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26
17. Cutting a sandwich4 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33
18. The hyperplane function . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33
19. The volume function . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33
20. Monotonicity of the volume function . . . . . . . . . . . . . . . . . . . . . . . 34
21. The disk and its boundary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37
22. Constructing a retraction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37
23. Connected but not path-connected . . . . . . . . . . . . . . . . . . . . . . . . 40
24. Subdividing 𝜔 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45
25. The homotopy to start with . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47
26. Subdividing the homotopy . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47
27. Deforming the homotopy in each square . . . . . . . . . . . . . . . . . . . . . 48
28. Deforming along one row . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49
29. Covering the figure 8 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 50
30. Punctured manifold . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 50
31. Start with two manifolds... . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51
32. ... and consider their connected sum. . . . . . . . . . . . . . . . . . . . . . . . 51
33. Surface of genus 𝑔 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52
34. Building a surface from the disk . . . . . . . . . . . . . . . . . . . . . . . . . 52
35. Starting the induction with the torus . . . . . . . . . . . . . . . . . . . . . . . 53
36. Induction step by cutting off a segment . . . . . . . . . . . . . . . . . . . . . . 53

159
List of Figures

37. Closing the loop 𝑐 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 54

38. Completing the induction step . . . . . . . . . . . . . . . . . . . . . . . . . . 54
39. Applying the Seifert-van Kampen theorem to the disk representation . . . . . . 56
40. Bouquet of circles . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 56
41. Finding the generator . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 57
42. Concatenation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 59
43. Commutativity of higher homotopy groups . . . . . . . . . . . . . . . . . . . 60
44. “Independence” of base point . . . . . . . . . . . . . . . . . . . . . . . . . . . 60
45. Failure of HLP due to lack of surjectivity . . . . . . . . . . . . . . . . . . . . 62
46. Surjective but HLP still fails . . . . . . . . . . . . . . . . . . . . . . . . . . . 62
47. Subdivision for which the fiber bundle is trivial over each subcube . . . . . . . 63
48. Lift over second cube . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 64
49. Homeomorphism mapping two faces to one . . . . . . . . . . . . . . . . . . . 64
50. Mapping the box to the (bottom) cube . . . . . . . . . . . . . . . . . . . . . . 66
51. Lift 𝜎 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 67
52. Homotopy between projected lifts . . . . . . . . . . . . . . . . . . . . . . . . 67
53. Homotopy invariance . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 68
54. Boundary map is a group homomorphism . . . . . . . . . . . . . . . . . . . . 69
55. Bottom-to-box homotopy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 70
56. Bottom-to-box homotopy, again . . . . . . . . . . . . . . . . . . . . . . . . . 71
57. Homotopy between 𝜏 and constant map . . . . . . . . . . . . . . . . . . . . . 71
58. Shrink “constant region” to boundary . . . . . . . . . . . . . . . . . . . . . . 72
59. Extension of Σ . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 72
60. Homotopy between 𝑝 ◦ 𝜏 and 𝜎 . . . . . . . . . . . . . . . . . . . . . . . . . 73

61. 2 and 3-dimensional standard simplices . . . . . . . . . . . . . . . . . . . . . 80

62. Face maps . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 80
63. Cylinder relative to bottom . . . . . . . . . . . . . . . . . . . . . . . . . . . . 85
64. The representative 𝜎 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 85
65. 𝜎 is null-homologous . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 86
66. Sphere relative to southern hemisphere . . . . . . . . . . . . . . . . . . . . . . 91
67. Excising a southern cap . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 91
68. Constructing a retraction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 93
69. The reflection in one dimension . . . . . . . . . . . . . . . . . . . . . . . . . 94
70. First 2-simplex 𝜎1 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 95
71. Second 2-simplex 𝜎2 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 95
72. Vector field on 𝑆 𝑛 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 96
73. The retraction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 96
74. The case of finite preimage . . . . . . . . . . . . . . . . . . . . . . . . . . . . 105
75. 𝑇00 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 113
76. 𝑇10 and 𝑇01 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 113
77. Setup for the excision axiom . . . . . . . . . . . . . . . . . . . . . . . . . . . 116
78. Barycenters in 1 and 2 dimensions . . . . . . . . . . . . . . . . . . . . . . . . 118
79. 𝐶1 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 119

160
 List of Figures

80. Vertices of maximal distance . . . . . . . . . . . . . . . . . . . . . . . . . . . 121

81. Barycenter is a vertex . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 122
82. Open cover of 𝑆1 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 125
83. Open cover of figure 8 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 126
84. Alexander horned sphere5 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 131
85. A 2-dimensional CW-complex . . . . . . . . . . . . . . . . . . . . . . . . . . 132
86. Attaching an 𝑛-cell to a point to obtain an 𝑛-sphere . . . . . . . . . . . . . . . 132
87. Cell decomposition of 𝑆 𝑛 with two 𝑛-cells . . . . . . . . . . . . . . . . . . . . 133
88. Punctured 𝑛-disk is homotopy equivalent to 𝑆 𝑛−1 . . . . . . . . . . . . . . . . 136
89. Punctured 𝑛-skeleton is homotopy equivalent to (𝑛 − 1)-skeleton . . . . . . . . 136
90. Platonic solids . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 143
91. A cell decomposition of the figure 8 . . . . . . . . . . . . . . . . . . . . . . . 144
92. Attaching a cell of dimension ≥ 3 . . . . . . . . . . . . . . . . . . . . . . . . 147
93. The maps 𝑠1 and 𝑠2 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 149
94. Representative of 𝛼1 and 𝛼2 . . . . . . . . . . . . . . . . . . . . . . . . . . . . 149

161
Bibliography
[1] Greenberg, M. J.; Harper, J. R.: Algebraic topology. A first course, Benjamin/Cummings
Publishing, 1981

[2] Hatcher, A.: Algebraic topology, Cambridge University Press, 2002

[3] Hilbert, D.: Ueber die stetige Abbildung einer Line auf ein Flächenstück, Math. Ann. 38
(1891), 459–460

[4] Lück, W.: Algebraische Topologie. Homologie und Mannigfaltigkeiten, Vieweg, 2004

[5] Ossa, E.: Topologie, Vieweg, 1992

[6] Peano, G.: Sur une courbe, qui remplit toute une aire plane, Math. Ann. 36 (1890), 157–160

[7] tom Dieck, T.: Topologie, Walter de Gruyter, 1991

[8] Warner, F. W.: Foundations of differentiable manifolds and Lie groups, Springer-Verlag,
New York, 1983

163
Index

Symbols H, quaternions . . . . . . . . . . . . . . . . . . . . . . . . 99
ℎ, Hurewicz homomorphism . . . . . . . . . . 149
𝑎, antipodal map . . . . . . . . . . . . . . . . . . . . . . 96 𝐻𝑛 ( 𝑓 ), map induced on homology . . . . . 83
¯ closure of 𝐴 . . . . . . . . . . . . . . . . . . . . . . . 89
𝐴, 𝐻𝑛 𝐾∗ , 𝑛-th homology . . . . . . . . . . . . . . . . 106
˚ interior of 𝐴 . . . . . . . . . . . . . . . . . . . . . . . 89
𝐴, 𝐻𝑛 (𝜑∗), map induced on homology . . . 107
𝑏 𝑗 (𝑋; 𝑅), Betti number . . . . . . . . . . . . . . 141 𝐻𝑛 (𝑋, 𝐴; 𝑅), relative 𝑛-th singular
𝐵𝑛 , barycenter . . . . . . . . . . . . . . . . . . . . . . . 118 homology . . . . . . . . . . . . . . . . . . . 84
𝐵𝑛 𝐾∗ , 𝑛-boundaries . . . . . . . . . . . . . . . . . . 106 𝐻𝑛 (𝑋; 𝑅), 𝑛-th singular homology . . . . . . 82
𝐵𝑛 (𝑋, 𝐴; 𝑅), relative singular 𝑛-boundaries HP𝑛 , quaternionic projective space . . . . 135
84 𝐾𝑛 (𝑋, X), cellular complex . . . . . . . . . . . 137
𝐵𝑛 (𝑋; 𝑅), singular 𝑛-boundaries . . . . . . . 82 N (𝑆), normal subgroup generated by S . 41
𝐵(𝑥, 𝑟), open ball of radius 𝑟 centered at 𝑥 6 Ω (𝑋; 𝑥0 , 𝑥1 ), set of paths . . . . . . . . . . . . . . 20
𝜒(𝑋, X), Euler number. . . . . . . . . . . . . . . 141 Ω (𝑋; 𝑥0 ), set of loops . . . . . . . . . . . . . . . . . 20
𝐶𝑛 , barycentric subdivision . . . . . . . . . . . 118 𝑃, prism operator . . . . . . . . . . . . . . . . . . . . 114
CP𝑛 , complex projective space. . . . .77, 134 𝜑 𝜎 , attaching map . . . . . . . . . . . . . . . . . . . 132
𝐶 𝑋, cone over 𝑋 . . . . . . . . . . . . . . . . . . . . . . 14 𝜋1 (𝑋; 𝑥0 ), fundamental group . . . . . . . . . . 21
𝐶 ((𝑋, 𝐴), (𝑌 , 𝐵)), set of continuous maps 𝜋 𝑛 (𝑋; 𝑥0 ), 𝑛-th homotopy group . . . . . . . . 58
between pairs of spaces . . . . . . . 83 𝑄 𝑛 , chain homotopy for subdivision map
𝐶 (𝑋, 𝑌 ), set of continuous maps . . . . . . . 15 119
deg, degree . . . . . . . . . . . . . . . . . . . . . . . 29, 97 RP𝑛 , real projective space . . . . . 73, 77, 133
deg 𝑝 , local degree . . . . . . . . . . . . . . . . . . . 102 Sd𝑛 , subdivision chain map . . . . . . . . . . . 119
𝐷 𝑛 , closed 𝑛-dimensional unit ball . . . . . . . 6 𝑆aff 𝑛
𝑘 (Δ ), affine chains . . . . . . . . . . . . . . . . 118
Δ𝑛 , standard simplex . . . . . . . . . . . . . . . . . . 79 𝑆𝑛 ( 𝑓 ), map induced on singular chains. . 82
𝐷 𝑛− , lower hemisphere . . . . . . . . . . . . . . . . . 91 𝑆𝑛 (𝑋, 𝐴; 𝑅), relative singular 𝑛-chains . . 83
𝐷 +𝑛 , upper hemisphere . . . . . . . . . . . . . . . . . 91 𝑆𝑛 (𝑋; 𝑅), singular 𝑛-chains . . . . . . . . . . . . 80
𝜕, boundary of singular chain . . . . . . . . . . 81 𝑆𝑛U (𝑋, 𝐴), small singular 𝑛-chains. . . . .117
𝜕, connecting homomorphism for Σ𝑋, suspension of 𝑋 . . . . . . . . . . . . . . . . . . 14
homology groups . . . . . . . . 87, 108 𝜎 (𝑖) , 𝑖-th face of singular simplex . . . . . . 80
𝜕, connecting homomorphism for 𝑇𝑛𝑖 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 113
homotopy groups . . . . . . . . . . . . . 66 TR𝑛 , standard topology of 𝑅 𝑛 . . . . . . . . . . . . 6
¯
𝜕, boundary of relative singular chain . . . 84 T𝑋 , topology of 𝑋 . . . . . . . . . . . . . . . . . . . . . . 7
𝜀 𝑥0 , constant loop . . . . . . . . . . . . . . . . . . . . . 20 𝑈 ( 𝑓 , 𝑝), winding number . . . . . . . . . . . . . . 75
Exp, covering map of 𝑆1 . . . . . . . . . . . . . . . 27 𝑊 𝑛 , 𝑛-dimensional cube . . . . . . . . . . . . . . . 13
𝑓# , map induced on fundamental group . . 23 𝑋 𝑛 , 𝑛-skeleton . . . . . . . . . . . . . . . . . . . . . . 132
𝐹𝑔 , surface of genus 𝑔 . . . . . . . . . . . . . . . . . . 5 𝑍 𝑛 𝐾∗ , 𝑛-cycles . . . . . . . . . . . . . . . . . . . . . . 106
𝐹𝑛𝑖 , face map . . . . . . . . . . . . . . . . . . . . . . . . . 80 𝑍 𝑛 (𝑋, 𝐴; 𝑅), relative singular 𝑛-cycles . . 84

165
Index

𝑍 𝑛 (𝑋; 𝑅), singular 𝑛-cycles . . . . . . . . . . . . 82 cone . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14

≃, homotopic . . . . . . . . . . . . . . . . . . . . 15, 111 connected . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39
≃ 𝐴, homotopic relative to 𝐴 . . . . . . . . . . . . 15 1-connected . . . . . . . . . . . . . . . . . . . . . . . . . . 39
≈, homeomorphic . . . . . . . . . . . . . . . . . . . . . . 8 connecting homomorphism . . . . . . . 87, 108
𝐺/𝐻, homogeneous space . . . . . . . . . . . . . 65 continuous map . . . . . . . . . . . . . . . . . . . . . . . . 8
𝑀#𝑁, connected sum . . . . . . . . . . . . . . . . . 51 contractible . . . . . . . . . . . . . . . . . . . . . . . . . . 17
𝐺 1 ★ 𝐺 2 , free product . . . . . . . . . . . . . . . . . 43 covering . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 72
𝜔 ★ 𝜂, concatenation . . . . . . . . . . . . . . . . . . 20 cycle . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 106
cylinder . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
A
D
Alexander horned sphere . . . . . . . . . . . . . 131
antipodal map . . . . . . . . . . . . . . . . . . . . . . . . 96 deformation retract . . . . . . . . . . . . . . . . . . . . 17
attaching map . . . . . . . . . . . . . . . . . . . . . . . 132 degree. . . . . . . . . . . . . . . . . . . . . . . . . . . .29, 97
augmented boundary operator . . . . . . . . . 152 dimension axiom . . . . . . . . . . . . . . . . . . . . . 88
augmented homology . . . . . . . . . . . . . . . . 153 discrete topology . . . . . . . . . . . . . . . . . . . . . . 7

B E

barycenter . . . . . . . . . . . . . . . . . . . . . . . . . . 118 Eilenberg-Steenrod axioms . . . . . . . . . . . . 88

base space . . . . . . . . . . . . . . . . . . . . . . . . . . . 62 embedding . . . . . . . . . . . . . . . . . . . . . . . . . . 127
Betti number . . . . . . . . . . . . . . . . . . . . . . . . 141 Euler number . . . . . . . . . . . . . . . . . . . . . . . 141
bidegree . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 99 Euler’s formula for polyhedra . . . . . . . . . 142
Borsuk-Ulam theorem . . . . . . . . . . . . . . . . . 32 Euler-Poincaré characteristic . . . . . . . . . . 141
boundary . . . . . . . . . . . . . . . . . . . . . . . . . . . 106 exact sequence of complexes . . . . . . . . . . 107
boundary of a singular simplex . . . . . . . . . 81 exactness axiom . . . . . . . . . . . . . . . . . . . . . . 89
boundary operator . . . . . . . . . . . . . . . . . . . . 87 excision axiom . . . . . . . . . . . . . . . . . . . . . . . 89
augmented . . . . . . . . . . . . . . . . . . . . . . 152
F
Brouwer’s fixed point theorem . . . . . . 37, 93
face of a singular simplex . . . . . . . . . . . . . . 80
C fiber . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 62
fiber bundle . . . . . . . . . . . . . . . . . . . . . . . . . . 63
cell . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 132
finite CW-complex . . . . . . . . . . . . . . . . . . . 131
cellular complex . . . . . . . . . . . . . . . . . . . . . 137
finitely presentable group . . . . . . . . . . . . . . 55
chain homotopy . . . . . . . . . . . . . . . . . . . . . 111
free 𝑅-module generated by a set . . . . . . 157
chain map . . . . . . . . . . . . . . . . . . . . . . . . . . 107
free product . . . . . . . . . . . . . . . . . . . . . . . . . . 41
characteristic map . . . . . . . . . . . . . . . . . . . 132
fundamental group . . . . . . . . . . . . . . . . . . . . 21
chord . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 76
closed subset . . . . . . . . . . . . . . . . . . . . . . . . . . 7 H
closure . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 89
coarse topology . . . . . . . . . . . . . . . . . . . . . . . . 7 hairy ball theorem . . . . . . . . . . . . . . . . . . . . 96
comb space . . . . . . . . . . . . . . . . . . . . . . . 18, 75 ham-sandwich theorem . . . . . . . . . . . . . . . . 33
compact subset . . . . . . . . . . . . . . . . . . . . . . . . 8 Hausdorff space . . . . . . . . . . . . . . . . . . . . . . . 9
complex . . . . . . . . . . . . . . . . . . . . . . . . . . . . 106 Heine-Borel theorem . . . . . . . . . . . . . . . . . . . 8
complex projective space. . . . . . . . . .77, 134 Hilbert curve . . . . . . . . . . . . . . . . . . . . . . . . . . 4

166
 Index

HLP . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 61 O
homeomorphic . . . . . . . . . . . . . . . . . . . . . . . . 8
open (for the product topology) . . . . . . . . 12
homeomorphism . . . . . . . . . . . . . . . . . . . . . . . 8
open subset of R𝑛 . . . . . . . . . . . . . . . . . . . . . . 6
homogeneous space . . . . . . . . . . . . . . . . . . . 65
open subset of a topological space . . . . . . . 7
homology . . . . . . . . . . . . . . . . . . . . . . . . . . . 106
augmented . . . . . . . . . . . . . . . . . . . . . . 153 P
relative singular . . . . . . . . . . . . . . . . . . 84
singular . . . . . . . . . . . . . . . . . . . . . . . . . .82 pair of spaces. . . . . . . . . . . . . . . . . . . . . . . . . 83
homotopic chain maps . . . . . . . . . . . . . . . 111 path . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20
homotopic relative to 𝐴 . . . . . . . . . . . . . . . 15 path-connected . . . . . . . . . . . . . . . . . . . . . . . 39
homotopy axiom . . . . . . . . . . . . . . . . . . . . . . 88 plane-filling curve . . . . . . . . . . . . . . . . . . . . . 4
homotopy equivalence . . . . . . . . . . . . 16, 112 pointed set . . . . . . . . . . . . . . . . . . . . . . . . . . . 61
homotopy equivalent . . . . . . . . . . . . . . . . . . 16 presentation of a group . . . . . . . . . . . . . . . . 55
homotopy group . . . . . . . . . . . . . . . . . . . . . . 58 prism operator. . . . . . . . . . . . . . . . . . . . . . .114
homotopy lifting property . . . . . . . . . . . . . 61 projective space
homotopy relative to 𝐴 . . . . . . . . . . . . . . . . 15 complex . . . . . . . . . . . . . . . . . . . . . 77, 134
Hopf fibration . . . . . . . . . . . . . . . . . . . . . . . . 77 quaternionic . . . . . . . . . . . . . . . . . . . . 135
Hurewicz homomorphism . . . . . . . . . . . . 150 real . . . . . . . . . . . . . . . . . . . . . 73, 77, 133
Q
I
quaternionic projective space . . . . . . . . . 135
incidence numbers . . . . . . . . . . . . . . . . . . . 145 quaternions . . . . . . . . . . . . . . . . . . . . . . . . . . 99
induced topology of metric space . . . . . . . . 7
induced topology of subspace . . . . . . . . . . . 7 R
interior . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 89 real projective space . . . . . . . . . . 73, 77, 133
L relative singular homology . . . . . . . . . . . . . 84
retract . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
lift . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28 retraction . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
local degree . . . . . . . . . . . . . . . . . . . . . . . . . 102
S
long exact homology sequence of a pair of
spaces . . . . . . . . . . . . . . . . . . . . . . . 89 Seifert-van Kampen theorem . . . . . . . . . . . 46
long exact homotopy sequence of a Serre Serre fibration . . . . . . . . . . . . . . . . . . . . . . . . 62
fibration . . . . . . . . . . . . . . . . . . . . . 69 simply-connected . . . . . . . . . . . . . . . . . . . . . 39
loop . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20 singular 𝑛-boundary . . . . . . . . . . . . . . . . . . . 82
singular 𝑛-chain . . . . . . . . . . . . . . . . . . . . . . 80
M singular 𝑛-cycle . . . . . . . . . . . . . . . . . . . . . . 81
singular 𝑛-simplex . . . . . . . . . . . . . . . . . . . . 79
Mayer-Vietoris sequence . . . . . . . . . . . . . 125
singular homology . . . . . . . . . . . . . . . . . . . . 82
Möbius strip . . . . . . . . . . . . . . . . . . . . . . . . . 76
skeleton . . . . . . . . . . . . . . . . . . . . . . . . . . . . 132
N small chain . . . . . . . . . . . . . . . . . . . . . . . . . 117
small chain theorem . . . . . . . . . . . . . . . . . 117
natural . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 87 split exact sequence . . . . . . . . . . . . . . . . . . . 78
normal. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .40 standard simplex . . . . . . . . . . . . . . . . . . . . . . 79
normal subgroup generated by a set . . . . . 41 standard topology of R𝑛 . . . . . . . . . . . . . . . . 6

167
Index

strong deformation retract . . . . . . . . . . . . . 17 of product topology . . . . . . . . . . . . . . . 13

subspace topology . . . . . . . . . . . . . . . . . . . . . 7 of quotient topology . . . . . . . . . . . . . . 11
surface of genus 𝑔 . . . . . . . . . . . . . . . . . . . . 52
suspension . . . . . . . . . . . . . . . . . . . . . . . . . . . 14 V

T vector field . . . . . . . . . . . . . . . . . . . . . . . . . . . 96
total space . . . . . . . . . . . . . . . . . . . . . . . . . . . 62 W
U
weak fibration . . . . . . . . . . . . . . . . . . . . . . . . 62
universal property winding number . . . . . . . . . . . . . . . . . . . . . . 75

168