Homology of groups

Antoine Touzé

June 22, 2023

1
CONTENTS A. Touzé – 2023

This is not a book. . .

This document is not a book, but the lecture notes of a Master 2 course
on the homology of groups, taught in spring 2023 at the university of Lille.
The reader will find here all the definitions and statements of the theorems
given in the course, without proofs.
The student is assumed to have a knowledge of general topology, fun-
damental groups and covering spaces (although some recollections are done
on these subjects). The first part of the course covers singular homology
of topological spaces (including simplicial homology and cellular homology).
The second part of the course deals with the definition of Ext and Tor
over rings. Finally, the third part of the course introduces the homology of
groups in a purely algebraic fashion (Tor over group rings) and explains the
equivalent topological definition (Homology of K(G, 1) spaces).
A few "vista" sections were explained without proofs, they are just in-
tended as a brief introduction to more advanced topics and they are not
part of the material for the exam.
The whole course can be seen as an introduction to homological algebra.
Indeed all the most basic concepts of homological algebra are introduced
during the course: complexes, quasi-isomorphisms, homotopies, long exact
sequences, the fundamental theorem of homological algebra, bicomplexes,
universal coefficient theorems, the Künneth formula (over fields), the bar
construction.

Contents

I Topology and homology 6

1 Some algebraic topology 6

1.1 Homotopy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
1.2 Two basic homotopy invariants . . . . . . . . . . . . . . . . . 7
1.3 The language of categories . . . . . . . . . . . . . . . . . . . . 8
1.4 Singular homology . . . . . . . . . . . . . . . . . . . . . . . . 9
1.5 Singular homology of spheres and a few applications . . . . . 11

2 Complexes and singular homology 12

2.1 The world of complexes . . . . . . . . . . . . . . . . . . . . . 12
2.2 Homology of complexes . . . . . . . . . . . . . . . . . . . . . 13
2.3 Construction of singular homology . . . . . . . . . . . . . . . 14

3 Homology of pairs 17

2
CONTENTS A. Touzé – 2023

4 Homology of combinatorial topological spaces 19

4.1 Geometric simplicial complexes . . . . . . . . . . . . . . . . . 19
4.2 CW-complexes . . . . . . . . . . . . . . . . . . . . . . . . . . 21

5 A word on the ground ring R of homology 25

6 An overview of cohomology 27
6.1 Cochain complexes . . . . . . . . . . . . . . . . . . . . . . . . 27
6.2 Duality . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28
6.3 Singular cohomology . . . . . . . . . . . . . . . . . . . . . . . 29
6.4 The cup product . . . . . . . . . . . . . . . . . . . . . . . . . 30
6.5 Vista : some applications of singular cohomology . . . . . . . 31

II Algebra and homology 33

7 Projective modules and Ext 33

7.1 Projective modules . . . . . . . . . . . . . . . . . . . . . . . . 33
7.2 Projective resolutions . . . . . . . . . . . . . . . . . . . . . . . 34
7.3 Ext . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34

8 Injective modules and Ext 38

8.1 Injective modules . . . . . . . . . . . . . . . . . . . . . . . . . 38
8.1.1 Injective modules for fields . . . . . . . . . . . . . . . 39
8.1.2 Injective modules for PIDs . . . . . . . . . . . . . . . 39
8.1.3 Injective modules for general rings . . . . . . . . . . . 39
8.2 Injective coresolutions . . . . . . . . . . . . . . . . . . . . . . 40
8.3 Ext with injective coresolutions . . . . . . . . . . . . . . . . . 41
8.4 An application to semi-simplicity . . . . . . . . . . . . . . . . 41

9 Bicomplexes and their applications 42

9.1 Cochain bicomplexes . . . . . . . . . . . . . . . . . . . . . . . 42
9.2 The proof of theorem 128 . . . . . . . . . . . . . . . . . . . . 44
9.3 The Künneth formula . . . . . . . . . . . . . . . . . . . . . . 44
9.4 Chain bicomplexes . . . . . . . . . . . . . . . . . . . . . . . . 45
9.5 Vista: the singular homology of products of spaces . . . . . . 45

10 Flat modules and Tor 47

10.1 Tensor products over noncommutative rings . . . . . . . . . . 47
10.2 Flatness . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49
10.3 Tor . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49

11 Universal coefficient formulas 53

3
REFERENCES A. Touzé – 2023

III Homology of groups 54

12 Basic notions on representations 54

12.1 Group algebras and representations . . . . . . . . . . . . . . . 54
12.2 Tensor product and Hom . . . . . . . . . . . . . . . . . . . . 55
12.3 Invariants and coinvariants . . . . . . . . . . . . . . . . . . . 56
12.4 Induction and coinduction . . . . . . . . . . . . . . . . . . . . 56

13 Homology of groups 58
13.1 Homology of groups and Tor . . . . . . . . . . . . . . . . . . 58
13.2 The standard complex . . . . . . . . . . . . . . . . . . . . . . 59
13.3 Functoriality with respect to G . . . . . . . . . . . . . . . . . 60
13.3.1 Schapiro’s lemma . . . . . . . . . . . . . . . . . . . . . 61
13.3.2 The Cartan-Eilenberg formula for normal subgroups . 61

14 Homology of groups and topology 64

14.1 Recollections of covering spaces . . . . . . . . . . . . . . . . . 64
14.2 K(G, 1) and group homology . . . . . . . . . . . . . . . . . . 65
14.3 Projective resolutions of ktriv via G-complexes . . . . . . . . . 67
14.4 Semi-simplicial sets . . . . . . . . . . . . . . . . . . . . . . . . 68
14.5 Construction of K(G, 1)-spaces . . . . . . . . . . . . . . . . . 69

15 Vista: homotopy groups and EML spaces 70

15.1 Homotopy groups . . . . . . . . . . . . . . . . . . . . . . . . . 70
15.2 K(G, 1)-spaces and EML spaces . . . . . . . . . . . . . . . . . 72

16 Vista: the cohomology of groups 75

IV Exercises 78

The course and the exercise sheets are mainly based on the following
excellent textbooks. The curious reader may consult these references for the
proofs and further developments.

References
[Félix-Tanré] Y. Felix, D. Tanré, Topologie Algébrique, Dunod.

[Hatcher] A. Hatcher, Algebraic Topology, CUP, available online:

https://pi.math.cornell.edu/~hatcher/AT/ATpage.html

[Weibel] C. Weibel, An introduction to homological algebra, CUP

[Rotman] J. Rotman, An introduction to homological algebra, Springer

4
REFERENCES A. Touzé – 2023

[Brown] K. S. Brown, Cohomology of groups, Springer

To be more specific, part I on topology and homology is based on chap-

ters 5 and 6 of [Félix-Tanré], with a few additional facts borrowed from
[Hatcher], namely the formula for the differential in cellular homology, and
some facts relative to cohomology.
Part II on algebra and homology is based on chapters 1 and 3 of [Weibel],
with a few additional facts taken from [Rotman]. A nice account of the
universal coefficient theorem can be also found in [Félix-Tanré].
Part III on the (co)homology of groups is mainly based on chapter 6
of [Weibel], and on chapters 1–3 of [Brown]. The part relative to ∆-sets
and the general construction of K(G, 1) spaces is taken from the begining
of chapter 2 and the additional topics of chapter 1 of [Hatcher].

5
 A. Touzé – 2023

Part I
Topology and homology
1 Some algebraic topology
1.1 Homotopy
Definition 1. Given two continuous maps f, g : X → Y , we say that f is
homotopic to g if there is a continuous map H : X × [0, 1] → Y such that
H(x, 0) = f (x) and H(x, 1) = g(x) for all x ∈ X. The map H is called
a homotopy between f and g. If f is homotopic to g, we write f ∼ g, or
f ∼H g if we want to specify the homotopy H.

Lemma 2. 1. The binary relation ∼ is an equivalence relation on the

set C(X, Y ) of continuous maps from X to Y .

2. If f1 , g1 : X → Y are homotopic and f2 , g2 : Y → Z are homotopic

then f2 ◦ f1 , g2 ◦ g1 : X → Z are homotopic.

Definition 3. A continuous map f : X → Y is a homotopy equivalence if

there is a map g : Y → X such that g ◦ f ∼ idX and f ◦ g ∼ idY . Then g is
called a homotopy inverse of f .
A topological space X is homotopy equivalent to Y (or has the same
homotopy type as Y ) if there is a homotopy equivalence f : X → Y . If X is
homotopy equivalent to Y , we write X ∼ Y .
A space with the same homotopy type as the one point space {∗} is
called contractible.

Lemma 4. The binary relation ∼ defines an equivalence relation on the

collection of topological spaces.

Examples 5. 1. Homeomorphisms are homotopy equivalences.

2. Let Dn denote the n-dimensional disk, i.e. the unit ball in Rn , and
S n−1 the unit sphere. The inclusion Dn \ S n−1 ,→ Dn is a homotopy
equivalence, but not an homeomorphism.

3. If A is a subspace of X, a deformation retraction of X onto A is a

continuous map:
r : X × [0, 1] → X
such that r(x, 0) = idX , r(x, 1) ∈ A for all x ∈ X, and r(a, t) = a for
all a ∈ A and all t ∈ [0, 1]. If there is a deformation retraction of X
onto A, the inclusion A ⊂ X is a homotopy equivalence.

6
1.2 Two basic homotopy invariants A. Touzé – 2023

Remark 6. Deformation retractions are a very visual way of producing

homotopy equivalences. In fact, one can prove that every homotopy equiva-
lence can be realized by deformation retractions: X has the same homotopy
type as Y if and only if there is a third space Z which contains X and Y as
deformation retracts, see e.g. Hatcher, cor 0.21.

1.2 Two basic homotopy invariants

Informally speaking, a "homotopy invariant" is a quantity defined for all
topological spaces, and which is equal for two topological spaces of the same
homotopy type.
A first homotopy invariant that you already know is the fundamental
group of a (pointed) topological space (X, x). The fundamental group is
usually denoted by π1 (X, x). We will use it later, but for the moment we
only recall the following general facts.

1. Every continuous map f : X → Y induces a morphism of groups

π1 (f, x) : π1 (X, x) → π1 (Y, f (x))

and if f is a homotopy equivalence, then π1 (f, x) is an isomorphism.

(This point shows that the fundamental group is a homotopy invari-
ant).

if n = 1
(
Z
2. Basic calculation: for all x ∈ Sn, π1 (S n , x) = .
{1} if n =
6 1

Another (easier) homotopy invariant is the set of arcwise connected com-

ponents of a space. We study it into details now.

Definition 7. For all topological spaces X we let π0 (X) denote the set of
arcwise connected components of X. We let x denote the arcwise component
of x ∈ X. For all continuous maps f : X → Y we define a map

π0 (f ) : π0 (X) → π0 (Y )
.
x 7→ f (x)

Proposition 8. The following holds:

1. a) π0 (idX ) = idπ0 (X) ,

b) π0 (g ◦ f ) = π0 (g) ◦ π0 (f ),

2. if f ∼ g, then π0 (f ) = π0 (g).

7
1.3 The language of categories A. Touzé – 2023

1.3 The language of categories

The language of categories is convenient to give compact descriptions of the
properties of topological invariants.
Definition 9. A category is the data of:
• a collection Ob(C) of objects X, Y, . . . ,

• for all pairs of objects X, Y a set of morphisms HomC (X, Y )

• a composition law

◦ : HomC (Y, Z) × HomC (X, Y ) → HomC (X, Z)

which is associative, and which has units: for all objects X there exists
a morphism idX ∈ HomC (X, X) such that f ◦ idX = f and idX ◦ g = g
for all f ∈ HomC (X, Y ).
Examples 10. We have the following examples of categories. In each ex-
ample, the composition law is given by the usual composition.
Set: objects: the sets, morphisms: the maps.

Top: objects: the topological spaces, morphisms: the continuous maps.

Gp: objects: the groups, morphisms: the morphisms of groups.

R−Mod: objects: the R-modules, morphisms: the R-linear maps.

Definition 11. A functor F : C → D is the data:
• for each object X of C, an object F (X) of D.

• for each morphism f : X → Y in C, a morphism F (f ) : F (X) → F (Y )

in D, such that F preserves composition and identity morphisms.
Exercise 12. An isomorphism in a category C is a morphism f : X → Y
such that there exists a morphism g : Y → X such that g ◦ f = idX and
f ◦ g = idY . Show that every functor preserves the isomorphisms.
The definition and properties of π0 can then be reformulated in a com-
pact way by saying that π0 is a functor π0 : Top → Set which sends homo-
topic maps to the same map.
Proposition 13. Let F : Top → C be a functor. The following assertions
are equivalent: (i) for all continuous maps f, g if f ∼ g then F (f ) = F (g),
(ii) F sends the homotopy equivalences to isomorphisms in C.
Corollary 14. If f is a homotopy equivalence, then π0 (f ) is a bijection.
Exercise 15. Describe the fundamental group as a functor.

8
1.4 Singular homology A. Touzé – 2023

1.4 Singular homology

Definition 16. Given a set S and a ring R, we denote by RS the free R-
module on S. This is the set of formal sums s∈S rs bs with only finitely
P

many nonzero coefficients rs ∈ R, with R-linear module structure given by

! !
0
rs bs + ρ · rs0 bs := (ρrs + ρ0 rs0 )bs .
X X X
ρ·
s∈S s∈S s∈S

It is a free R-module with basis given by the symbols bs , s ∈ S. Every map

φ : S → T induces an R-linear map Rφ : RS → RT such that φ(bs ) = bφ(s) .

Remark 17. The free R-module construction defines a functor: R− : Set →

R−Mod.

Definition-Theorem 18. Let R be a ring. For all i ≥ 0 we can associate

to each topological space X a R-module Hi (X), and to each continuous map
f : X → Y an R-linear morphism Hi (f ) : Hi (X) → Hi (Y ) in such a way
that the following properties are satisfied.

1. Degree zero: we have H0 (X) = Rπ0 (X), and for all continuous maps
f : X → Y there H0 (f ) = Rπ0 (f ).

2. Functoriality:

a) Hi (idX ) = idHi (X) ,

b) Hi (f ◦ g) = Hi (f ) ◦ Hi (g).

3. Homotopy: if f ∼ g then Hi (f ) = Hi (g).

0 if i > 0
(
4. Homology of the point: Hi ({∗}) = .
R if i = 0

5. Additivity: if the Xα are the arcwise connected components of X

the inclusions Xα ,→ X induce an isomorphism:

Hi (Xα ) ' Hi (X)

M

6. Mayer-Vietoris: If U and V are two open subsets of X such that

X = U ∪ V , we have a long exact sequence of R-modules (i.e. a
sequence of R-modules in which the image of each morphism is equal
to the kernel of the following morphism):

∂ (1) (2) ∂
→ Hi (U ∩ V ) −−→ Hi (U ) ⊕ Hi (V ) −−→ Hi (X) −
... − → Hi−1 (U ∩ V ) → · · ·
∂ (1) (2)
→ H0 (U ∩ V ) −−→ H0 (U ) ⊕ H0 (V ) −−→ H0 (X) → 0 .
··· −

9
1.4 Singular homology A. Touzé – 2023

where the coordinates of the maps (1) are equal to Hi (U ∩ V ,→ U )

and Hi (U ∩ V ,→ V ), and where the components of the maps (2) are
equal to −Hi (U ,→ X) and Hi (V ,→ X).
Bonus: if X 0 = U 0 ∪ V 0 is another decomposition of a space into
two open subsets, and if f : X → X 0 is a continuous map such that
f (U ) ⊂ U 0 and f (V ) ⊂ V 0 , then the following squares commute:

∂
Hi (X) Hi−1 (U ∩ V )
Hi (f ) Hi−1 (f ) .
∂
Hi (X) Hi−1 (U ∩ V )

Proof. The proof will be given later, see section 2.

Let us comment on the properties listed in definition-theorem 18.

1. Dependence on R. The definition of singular homology depends on

a ring R. Most often, we take R a field of R = Z. We shall write
Hi (X; R) and Hi (f ; R) if we want to emphasize the dependence on
R, but usually R is clear from the context and we do not use these
heavier notations.

2. Homotopy functors. The module H0 contains essentially the same

information as π0 . All the Hi are homotopy functors like π0 . In
particular, if f is a homotopy equivalence, then all the R-linear maps
Hi (f ) are R-module isomorphisms.

3. Contactible spaces. All contractible spaces (such as Dn , Rn , [0, 1],

etc.) have the same homology as the point. They are the space with
the simplest possible homology.

4. Arcwise connected spaces. Because of the additivity property, we

can always restrict ourselves to computing the homology of arcwise
connected spaces.

5. Cut and paste. The Mayer-Vietoris property is the analogue for

singular homology of the Van Kampen theorem for the fundamental
group. It allows to compute the homology of a space inductively by
cutting it into simpler (ideally contractible) pieces.
Note that, in constrast to the Van Kampen theorem, we do not assume
that U ∩ V is arcwise connected. In particular, in the next section we
will compute Hi (S 1 ) with the Mayer-Vietoris property, whereas it is
impossible to compute π1 (S 1 , ∗) using the Van Kampen theorem.

10
1.5 Singular homology of spheres and a few applications A. Touzé – 2023

1.5 Singular homology of spheres and a few applications

Theorem 19. Recall that S n is the unit sphere in the euclidean space Rn+1 .
We have:

if i = 0 if i = 0, n
( (
R2 R
Hi (S ) =
0
, and for n > 0, Hi (S ) =
n
.
0 if i > 0 0 if i =
6 0, n

Corollary 20 (Invariance of dimension). If Rn is homeomorphic to Rm then

n = m.

Corollary 21 (Brouwer fixed point). Every continuous f : Dn → Dn has

a fixed point.

Degree theory.
Definition 22. Let n > 0. Since Hn (S n ; Z) = Z, for all continuous maps
f : S n → S n the map Hn (f ; Z) is given by multiplication by an integer
which is called the degree of f and denoted by deg(f ).

Proposition 23. Let r be the restriction of a linear reflexion to S n , n > 1.

Then deg(r) = −1.

Corollary 24. Let ι : S n → S n be the antipodal map (i.e. ι(x) = −x).

Then deg(f ) = (−1)n+1

Corollary 25 (Hairy ball). The sphere S n has a nonsingular tangent vector

field if and only if n is odd.

11
2. Complexes and singular homology A. Touzé – 2023

2 Complexes and singular homology

Complexes of R-modules are "algebraic models" of topological spaces. In
this section, we first develop the basic definitions and properties of com-
plexes, and then we define the homology of complexes. To finish, we will
explicitly contruct for each topological space X a complex C sing (X) which
is a algebraic model of X. The singular homology of X is defined as
Hi (X) := Hi (C sing (X)).

2.1 The world of complexes

Definition 26 (Complexes and their morphisms). Let R be a ring.
A complex (of R-modules) C is a the data of a family of R-modules
{Ci }i∈Z together with maps R-linear maps di : Ci → Ci−1 , such that di ◦
di+1 = 0 for all i ∈ Z.
The elements of Ci are called homogeneous of degree i. The maps di are
called the differentials, or the boundary maps of C. Usually the index is
dropped and all the di are simply denoted by the same letter ‘d’, so that the
equality above becomes d ◦ d = 0.
A morphism of complexes f : C → D is the data of a family of R-
linear morphisms fi : Ci → Di , i ∈ Z which preserves the differentials:
d ◦ fi = fi−1 ◦ d for all i.
Morphisms of complexes can be composed : f ◦g is given by the family of
R-linear maps fi ◦ gi , i ∈ Z. We denote by Ch(R) the category of complexes
and morphisms of complexes. A morphism f : C → D is an isomorphism if
there is a morphism g : D → C such that g ◦ f = idC and f ◦ g = idD .

Exercise 27. Show that f is an isomorphism if and only if all its components
fi are R-linear isomorphisms.

Definition 28. Two morphisms f, g : C → D are homotopic is there is a

family of maps hi : Ci → Di+1 such that fi −gi = d◦si +si−1 ◦d for all i. The
family h = (hi )i∈Z is called a homotopy between f and g. If f is homotopic
to g, we write f ∼ g, or f ∼h g if we want to specify the homotopy.

Exercise 29.

1. The binary relation ∼ is an equivalence relation on the set

HomCh(R) (C, D) of morphisms of chain complexes from C to D.

2. If f1 , g1 : C → D are homotopic and f2 , g2 : D → E are homotopic

then f2 ◦ f1 , g2 ◦ g1 : C → E are homotopic.

Definition 30. A morphism f : C → D is a homotopy equivalence if there

is a map g : D → C such that g ◦ f ∼ idC and f ◦ g ∼ idD . Then g is called
a homotopy inverse of f .

12
2.2 Homology of complexes A. Touzé – 2023

A complex C is homotopy equivalent to D (or has the same homotopy

type as D) if there is a homotopy equivalence f : C → D. If C is homotopy
equivalent to D, we write C ∼ D.

Exercise 31. The binary relation ∼ defines an equivalence relation on the

collection of complexes.

Examples 32. 1. Isomorphisms of complexes are homotopy equiva-

lences.
[ 10 ]
2. Let C = R −− →R ⊕R ,
| {z }
D = |{z}
R . The morphism π : C → D such
deg 0 deg 0
that π0 = [0 1] is a homotopy equivalence (but not an isomorphism!).

2.2 Homology of complexes

Definition 33 (Homology). Given a complex of R-modules C, we define
two submodules of Ci :
 
d
Zi (C) = Ker Ci −
→ Ci−1
 
d
Bi (C) = Im Ci+1 −
→ Ci

The elements of Zi (C) are called the cycles of degree i, the elements of
Bi (C) are called the boundaries of degree i. We have Bi (C) ⊂ Zi (C) and
the quotient
Hi (C) := Zi (C)/Bi (C)
is called the homology of degree i of the complex. Every morphism f : C → D
restricts to R-linear maps fi : Zi (C) → Zi (D) and fi : Bi (C) → Bi (D),
hence induces a morphism on the level of homology

Hi (f ) : Hi (C) → Hi (D) .

Explicitly, if z is a cycle of Ci we have Hi (f ) : [z] 7→ [fi (z)] where the

brackets denote the class of a cycle in homology. A morphism of chain
complexes is a quasi-isomorphism if it induces an isomorphism in homology.
r·
Example 34. Let r ∈ R and let C = R − R and D = R/rR . Then
→ |{z}
| {z }
deg 0 deg 0
the morphism q : C → D such that q0 : R → R/rR is the quotient map is a
quasi-isomorphism.

Exercise 35. Show that each Hi defines a functor from Ch(R) to R−Mod.

Proposition 36. If f ∼ g then Hi (f ) = Hi (g) for all i ∈ Z.

13
2.3 Construction of singular homology A. Touzé – 2023

Corollary 37. If f is a homotopy equivalence, then Hi (f ) is an isomor-

phism for all i ∈ Z. In other words, homotopy equivalences are quasi-
isomorphisms.

Definition 38. A short exact sequence of complexes is a sequence of com-

f g
plexes 0 → C −
→D→ − E → 0 such that for all i,
f g
0 → Ci −
→i
Di −
→i
Ei → 0

is a short exact sequence of R-modules (i.e. fi injective, Ker gi = Im fi and

gi surjective).
f g
Theorem 39. Every short exact sequence of complexes 0 → C − → D →
−
E → 0 yields a long exact sequence of R-modules in homology:

∂ Hi (f ) Hi (g) ∂
→ Hi (C) −−−→ Hi (D) −−−→ Hi (E) −
... − → Hi−1 (C) → · · ·
∂ H0 (f ) H0 (g) ∂
→ H0 (C) −−−−→ H0 (D) −−−→ H0 (D) −
··· − → H−1 (C) → · · · .

f0 g0
Bonus: if 0 → C 0 −→ D0 − → E 0 → 0 is another short exact sequence of
complexes and if we have a commutative diagram:

f g
0 C D E 0
αC αD αE
f0 g0
0 C0 D0 E0 0

then the following squares commute:

∂
Hi (E) Hi−1 (C)
Hi (αE ) Hi−1 (αC ) .
∂0
Hi (E 0 ) Hi−1 (C 0 )

2.3 Construction of singular homology

We are going to construct a complex C sing (X) of R-modules for each topo-
logical space X, which contains some geometrical information relative to
X.

Definition 40. Let (e0 , . . . , en ) denote the canonical basis of Rn+1 .

• The standard n-simplex ∆n is the convex hull of the vectors of the

canonical basis of Rn+1 :

∆n :=< e0 , . . . , en > .

14
2.3 Construction of singular homology A. Touzé – 2023

• The simplex < e0 , . . . , ei−1 , ei+1 , . . . , en > is denoted by ∂i ∆n . We call

it the i-th face of ∆n .

• If 0 ≤ i ≤ n, there is a unique affine map i : ∆n−1 → ∆n such that

si k ≤ i − 1
(
ek
 (ek ) =
i
ek+1 si k ≥ i + 1

This map di induces an affine isomorphism of ∆n−1 onto ∂i ∆n .

i j
Lemma 41. If i < j the composite maps ∆n−1 −
→ ∆n −
→ ∆n+1 and
j−1 i
∆n−1 −−−→ ∆n −
→ ∆n+1 are equal.

Definition-Theorem 42. Let X be a topological space and let R be a ring.

The singular complex C sing (X) is defined by:

(0) For n < 0, Cnsing (X) = 0 .

(1) For n ≥ 0, Cnsing (X) is the free R-module with basis the continuous
maps σ : ∆n → X. These maps σ are called the singular simplices of
X.
sing
(2) For n ≥ 1, the boundary map dn : Cnsing (X) → Cn−1 (X) sends each
σ : ∆n → X to n
dn (σ) := (−1)i (σ ◦ i ) .
X

i=0

For all continuous map f : X → Y , we let C sing (f ) : C sing (X) → C sing (Y ) be

the morphism of complexes which sends each singular simplex σ ∈ Cnsing (X)
to f ◦ σ ∈ Cnsing (Y ). Thus, the singular complex defines a functor:

C sing : Top → Ch(R) .

Examples 43. 1. C sing (∅) = 0.

2. C sing ({∗}) is a free R-module of rank 1 in each degree, generated by

the unique map ∆n → {pt}. The boundary map dn equals idR if n is
even and 0 if n is odd.

3. In general C sing (X) = α C sing (Xα ) where the Xα are the arcwise
L

connected components of X.

Definition 44. We define the singular homology of a topological space X

by Hi (X) = Hi (C sing (X)) and for all continuous map f : X → Y we let
Hi (f ) = Hi (C sing (f )). This yields functors:

Hi : Top → R−Mod .

15
2.3 Construction of singular homology A. Touzé – 2023

Proposition 45. We have H0 (X) = Rπ0 (X).

Proposition 46. If f ∼H g then C sing (f ) ∼ C sing (g).

Corollary 47. If f ∼H g then Hi (f ) = Hi (g) for all i ∈ Z.

Theorem 48 (U-small chains). Let U denote an open covering of X, and

let Cnsing,U (X) denote the submodule of Cnsing (X) generated by the singular
simplices whose image is contained in V , for some V ∈ U. Then:

1. the R-modules Cnsing,U (X) yield a subcomplex C sing,U (X) of C sing (X),

2. the inclusion C sing,U (X) ,→ C sing (X) is a quasi-isomorphism (in fact,

it is even a homotopy equivalence).

Corollary 49. The Mayer-Vietoris long exact sequence of theorem 18 holds.

16
3. Homology of pairs A. Touzé – 2023

3 Homology of pairs
Definition 50. Let X be a topological space and let A ⊂ X be a subspace.
The singular homology of the pair of (X, A) is the homology of the quotient
complex C sing (X)/C sing (A). We denote by Hi (X, A) the i-th homology
module of the pair (X, A).
If (Y, B) is another pair, every continuous map f : X → Y such that
f (A) ⊂ B induces a morphism in homology:

Hi (f ) : Hi (X, A) → Hi (Y, B) .

Examples 51 (two extreme examples.). 1. If A = ∅, then C sing (A) = 0

hence we have Hi (X, ∅) = Hi (X) in all degrees i.

2. If A = X then C sing (X) = C sing (A) hence Hi (X, X) = 0 for all i.

The homology of (X, A) is somehow like the homology of X modulo A.

We will give a clearer topological interpretation in theorem 59. Before this,
we give two major tools to compute the homology of a pair.

Theorem 52 (Long exact sequence of a pair). Every pair (X, A) gives rise
to a long exact sequence in homology:

∂ Hi (incl) Hi (incl) ∂
→ Hi (A) −−−−−→ Hi (X) −−−−−→ Hi (X, A) −
... − → Hi−1 (A) → · · ·
∂ H0 (incl) H0 (incl)
→ H0 (A) −−−−−→ H0 (X) −−−−−→ H0 (X, A) → 0 .
··· −

Bonus: if (Y, B) is another pair, and if f : X → Y is a continuous map

such that f (A) ⊂ B, then the following squares commute:

∂
Hi (X, A) Hi−1 (A)
Hi (f ) Hi−1 (f ) .
∂
Hi (Y, B) Hi−1 (B)

Examples 53. 1. Let x ∈ X. For all i > 0 the inclusion (X, ∅) ,→

'
(X, {x}) induces an isomorphism Hi (X) −
→ Hi (X, {x}), and in degree
zero there is a short exact sequence:

0 → H0 ({x}) → H0 (X) → H0 (X, {x}) → 0 .

if i = n,
(
R
2. For n > 0 we have Hi (Dn , Dn \ {0}) =
0 otherwise.

The next proposition is useful to produce further examples. The proof

relies on the five lemma.

17
3. Homology of pairs A. Touzé – 2023

Proposition 54. Let (X, A) and (Y, B) be two pairs, and let f : X → Y
be a continuous map such that f (A) ⊂ B. Assume that f : X → Y and
f : A → B both induce an isomorphism in singular homology. Then

Hi (f ) : Hi (X, A) → Hi (Y, B)

is an isomorphism for all i.

Example 55. If A ⊂ B ⊂ X and if A is a deformation retract of B, then

Hi (X, A) ' Hi (X, B) for all i.

Theorem 56 (Excision). Let B ⊂ A ⊂ X be topological spaces. Assume

that B ⊂ int(A). Then the inclusion X \ B ,→ X induces an isomorphism:
'
Hi (X \ B, A \ B) −
→ Hi (X, A) .

Example 57. If U is an open subset of Rn , and x ∈ U then

if i = n,
(
R
Hi (U, U \ {x}) =
0 otherwise.

If X is a topological space and x ∈ X then Hi (X, X \ {x}) is usually

called the local homology of X at x. The previous example shows that if U
is an open subset of Rn , then the local homology of the topological space
U remembers that U can be embedded as an open subset of Rn . The next
result generalizes the invariance of dimension.

Corollary 58 (Invariance of domain). Let U be an open subset of Rn and

let V be an open subset of Rm . If there is a homeomorphism between U and
V , then n = m.

We finish by a topological interpretation of the homology of a pair. Note

that by example 53.1, if ∗ = A/A denotes a point in the quotient X/A, then
the homology Hi (X/A, ∗) is isomorphic to Hi (X/A) if i 6= 0.

Theorem 59. Let X be a topological space, let A be a nonempty closed

subset of X and assume that there is a neighborhood V of A such that A is
a deformation retract of V . Then for all i, the quotient map q : X → X/A
induces an isomorphism:
'
Hi (X, A) −−−→ Hi (X/A, A/A) .
Hi (q)

18
4. Homology of combinatorial topological spaces A. Touzé – 2023

4 Homology of combinatorial topological spaces

4.1 Geometric simplicial complexes
Definition 60. A set of n + 1 points S = {x0 , . . . , xn } of RN is affinely
independent if the affine space generated by S has dimension n.
An affine n-simplex is the convex hull of a set of n+1 affinely independent
points. These points x0 , . . . , xn are called the vertices of the affine simplex,
and the affine simplex is denoted by < x0 , . . . , xn >.
If {xi0 , . . . , xik } is a subset of {x0 , . . . , xn }, then the affine k-simplex
< xi0 , . . . , xik > is called a k-face of < x0 , . . . , xn >. Thus the vertices are
the 0-faces. The (n − 1)-faces are often denoted by < x0 , . . . , x ck , . . . , xn >
where x ck means that xk is deleted.

Example 61. The standard n-simplex ∆n is an affine n-simplex of Rn+1 .

In fact, ∆ is an affine n-simplex of RN if and only if there is an affine map
'
Rn+1 → RN which restricts to a bijection ∆n − → ∆.
Definition 62. A geometric simplicial complex (of finite dimension) is a
nonempty set K of affine simplices of RN satisfying the following axioms.
(i) If ∆ ∈ K then all the faces of ∆ are also in K.

(ii) If ∆ and ∆0 are in K, then either ∆ ∩ ∆0 = ∅, or ∆ ∩ ∆0 is a face of

both ∆ and ∆0 .

(iii) Every ball in Rn meets a finite number of simplicies of K.

The dimension of K is the maximal n such that there is an n-simplex in
K. The 0-simplices of K are called the vertices of K. We denote by |K|
the realization of K, or the polyedron associated to K. It is the topological
subspace of RN :
|K| = ∆.
[

∆∈K

Remark 63. Because of the local finiteness axiom (iii), |K| is compact
if and only if K has a finite number of simplices. The last axiom also
guaranties that every geometric simplicial complex has a countable number
of simplicies.
Remark 64. A triangulation of a topological space X is the data of a geo-
metric simplicial complex K and a homeomorphism
≈
h : |K| −
→X.

Polyedra of RN admit a triangulation (by definition), but there are many

other examples, such as differentiable manifolds or semi-algebraic subsets
of Rn . Thus a wide variety of geometric objects can be described as ge-
ometric simplicial complexes. For example, here is a beautiful image of a

19
4.1 Geometric simplicial complexes A. Touzé – 2023

triangulated duck (image taken from a course on simplicial homology on M.

Mandell’s page):

For further beautiful explanations relative to triangulations, I recommend

to visit the website "Analysis Situs".

From a geometric simplicial complex K, we can cook up a complex

C simpl (K) called the simplicial complex of K, as follows.

1. Choose a total order on the vertices of K. Thus every simplex of K is

of the form < x0 , . . . , xn > where x0 < x1 < · · · < xn .

2. Define Cnsimpl (K) as the free R-module with basis the n-simplices of
K.
simpl
3. Define the boundary maps d : Cnsimpl (K) → Cn−1 (K) by the formula:
n
(−1)k < x0 , · · · , x
X
d < x0 , · · · , xn >= ck , · · · , xn > .
k=0

Theorem 65. The homology of the complex C simpl (K) is isomorphic to the
singular homology of |K|.

Proof. This will be a special case of theorem 73 for CW-complexes proved

later. A direct proof for finite geometric simplicial complexes may be found
in the exercises (exercise 30)

20
4.2 CW-complexes A. Touzé – 2023

4.2 CW-complexes
The geometric realization X = |K| of a geometric simplicial complex can be
constructed step by step: consider first the 0-simplicies (this is called the
0-skeleton of X, and usually denoted by X0 ) , then paste the 1-simplicies
on the 0 skeleton (this produces the 1-skeleton X1 ⊂ X), then paste the
2-simplicies onto the 1-skeleton to obtain the 2-skeleton X2 and so on.
The notion of a CW-complex generalize this step-by-step construction of
spaces. To formulate the definition of a CW-complex, we must first formalize
the operation of pasting a space onto another one.

Definition 66 (Pushout squares). Let C be a category. A commutative

square in C:
A B

C D
is called a pushout square if it satisfies the following universal property. For
all morphisms φB and φC making the outer square (ABZC) commute, there
is a unique map φ making the two triangle commute:

f
A B
g
φB
C D
∃!φ
φC
Z

Interpretation 67 (in the category Top). In Top, every diagram

f
A B
g

can be completed into a pushout square. Indeed, if A = ∅, we may take

D = B t C. If A 6= ∅ then we let ∼ be the smallest equivalence relation such
that for all a ∈ A, f (a) ∈ B is identified with g(a) ∈ C. Then we may take
D = B t D/ ∼. Thus, in a pushout square, the space D is the glueing of
the spaces B and C ‘along A’ (in a way prescribed by the maps f : A → B
and g : A → C).

Definition 68. Let X be a topological space. A CW-complex structure on

X is the following data.

21
4.2 CW-complexes A. Touzé – 2023

1. An increasing sequence of subspaces X0 ⊂ · · · ⊂ Xk ⊂ . . . such that

(i) X0 is a discrete space, (ii) X = k≥0 Xk , and (iii) a subset U is
S

open in X ⇔ for all k U ∩ Xk is an open subset of Xk .

2. For k positive, a family of maps fα : Dαk → Xk , α ∈ A which restrict

to maps f∂α : Sαk−1 → Xk−1 , (where Dαk denote a copy of the k-disk
and Sαk−1 is its boundary) such that the following square is a pushout
square.
F k−1 tf∂α
α∈A Sα Xk−1
.
F k tfα
α∈A Dα Xk

Terminology.

• The subspace Xk is called the k-skeleton of X. If X = Xk then we say

that X has dimension k.

• We let ekα = fα (Dαk \ Sαk−1 ). These subspaces of X are called the cells
(of dimension k). The cells form a partition of X.

• The map fα : Dαk → Xk ⊂ X is the characteristic map of the cell ekα ,

and its restriction f∂α : Sαk−1 → Xk−1 is the attaching map of ekα .

Remark 69. A CW-complex X is constructed inductively by attaching

Cells. Condition 1(iii) says that X has the Weak topology with respect to
the Xk . These two facts explain the letters CW in "CW-complexes".

Examples 70. 1. Polyedra have a natural structure of a CW-complex

(Yes, polyedra are obtained by glueing affine simplicies, but an affine
simplex of dimension n is homeomorphic to Dn .)

2. The n-sphere S n can be considered as an n-dimensional CW-complex

with one 0-cell (the south pole ∗), and an n-cell attached on it via the
constant map S n−1 → {∗}. In this cell structure, X0 = X1 = · · · =
Xn−1 = {∗} and Xn = S n .

3. The n-sphere S n can be considered as an n-dimensional CW-complex,

with i-skeleton S i := S n ∩ {(x0 , . . . , xi , 0, . . . , 0) ⊂ Rn+1 , and each
S i+1 is obtained by pasting two copies of Di+1 via the identity maps
=
Si −→ Si.

4. A topological graph is a CW-complex of dimension 1. The vertices are

the 0-cells, the edges are the 1-cells.

5. Let g ∈ N∗ . We denote by Sg the quotient of D2 (the unit disc of C

by the identification depicted by the following drawing.

22
4.2 CW-complexes A. Touzé – 2023

The quotient space Sg is the compact orientable surface of genus g:

Then Sg has the structure of a CW-complex of dimension 2, with

one 0-cell (the image of the extremities of the segments a1 , b1 , etc,
which are all identified in the quotient), with 1-skeleton the image of
the boundary of D2 in Sg (this 1-skeleton is a bouquet of 2g circles
a1 ,. . . ,ag , b1 ,. . . ,bg ).
Proposition 71 (Topology of CW complexes). Let X be a CW-complex.

(i) A ⊂ X is open (resp. closed) ⇔ for all α, fα−1 (A) is open (resp.
closed) in Dαn .
(ii) X is Hausdorff.
(iii) A ⊂ X is compact ⇔ A is closed and it is contained in a finite union
of cells.

Remark 72. The last fact of the proposition has many consequences:
- A CW complex is compact iff it has a finite number of cells,
- the closure of every cell enα meets only a finite number of cells,
- If a subset A of X is compact, then it is contained in Xn for some n.

23
4.2 CW-complexes A. Touzé – 2023

From a CW-complex X we can cook up a complex C cell (X) called the

cellular complex of X, as follows.

1. For each n, C cell (X) is the free R-module with basis the n-cells of X.

2. Define the boundary map d : C1cell (X) → C0simpl (X) by sending each
1-cell e1α with attaching map fα : Dα1 = [−1, 1] → X1 to

d1 (e1α ) = fα (1) − fα (−1) .

3. For n ≥ 1, define the boundary map d : Cn+1cell (X) → C simpl (X) by

n
sending each n + 1-cell eα with attaching map fα : Dαn+1 → Xn+1 to
n+1

α )=
X
d(en+1 dα,β enβ
β

where dαβ ∈ Z is the degree of the composition

f fβ
Sαn −−→ Xn  Xn /(Xn \ enβ ) ←− Dβn /Sβn−1 = S n .
∂α
'

Theorem 73. The homology of the complex C cell (X) is isomorphic to the
singular homology of X.

24
5. A word on the ground ring R of homology A. Touzé – 2023

5 A word on the ground ring R of homology

The homology of the spheres does not really depend on the ring R. But this
phenomenon is in fact quite exceptional.

Example 74. The real projective plane RP 2 = R3 /R∗ has a CW structure

with one 0-cell, with one 1-cell and with one 2-cell. To be more specific,

(RP 2 )1 = (R2 × {0})/R∗ = RP 1 ' S 1

and the attaching map of 2-cell has degree 2. Thus the associated cellular
complex of RP 2 has the form:
2 0
Re2 →
− Re1 →
− Re0 .

In particular, Hi (RP 2 ) is zero if i ≥ 3 and we have:

2
H0 (RP 2 ) = R , H1 (RP 2 ) = R/2R , H2 (RP 2 ) = Ker R →


− R .

Thus we see that the homology of RP 2 strongly depends on the ring R:

• if R is a field of characteristic 6= 2, then RP 2 has the same homology

as a point (zero everywhere except in degree zero),

• if R is a field of characteristic 2, then RP 2 has homology equal to R

in degrees 0, 1, and 2, and it is zero elsewhere,

• if R = Z, then RP 2 has nonzero homology in degrees 0 and 1 only.

Natural question 75. given a topological space X and a ring R, if we

know Hi (X, R), is it possible to predict what Hi (X, S) will be for other
rings S ?

We shall give a partial answer to this question here. If F is a functor

such that F (0) = 0 then for all complexes C of R-modules, we can apply F
to obtain a complex1 F (C) of S-modules
F (d) F (d) F (d) F (d)
F (C) := . . . −−−→ F (Ci ) −−−→ F (Ci−1 ) −−−→ F (Ci−2 ) −−−→ . . . .

Definition 76. A functor F : R−Mod → S−Mod is called exact if for all

short exact sequence of R-modules 0 → M − − P → 0, the diagram:
→N →
f g

0 → F (M ) −−−→ F (N ) −−−→ F (P ) → 0
F (f ) F (g)

is also an exact sequence.

1
The hypothesis F (0) = 0 guaranties that F (d) ◦ F (d) = F (d ◦ d) = F (0) = 0.

25
5. A word on the ground ring R of homology A. Touzé – 2023

Proposition 77. If F is exact, then it preserves surjective morphisms, in-

jective morphisms, kernels, cokernels, images, and F (0) = 0.

Proposition 78. If F : R−Mod → S−Mod is exact, then it preserves

homology. To be more specific, for all i we have

Hi (F (C)) ' F (Hi (C)) .

The previous proposition applies in particular if F = −⊗R S : R−Mod →

S−Mod is the extension of scalars along a morphism of rings R → S, such
that S is flat as an R-module. We will recall flatness in details later, but we
mention two elementary cases in the next corollary.

Corollary 79. Assume that R is a field and that S is an R-algebra, or that

R = Z and that S = Q or a field of characteristic zero. Then there are
isomorphisms for all i:

Hi (C ⊗R S) ' Hi (C) ⊗R S .

In particular, if X is a topological space, there are isomorphisms

Hi (X, S) ' Hi (X, R) ⊗R S .

26
6. An overview of cohomology A. Touzé – 2023

6 An overview of cohomology
6.1 Cochain complexes
The complexes that we have used until now are also called chain complexes.
We are now going to introduce an analogous notion, in which the differentials
(=boundary morphisms) raise the degree by one instead of lowering it by
one.
Definition 80. A cochain complex C is a family of R-modules (C i )i∈Z
equipped with morphisms:
d : C i → C i+1
satisfying d ◦ d = 0. These morphisms are called the differentials of C, and
C i is called the R-module of homogeneous elements of cohomological degree
i of C. A morphism of cochain complexes f : C → D is a family of R-linear
morphisms f i : C i → Di , i ∈ Z which preserve the differentials, that is:
d ◦ f i = f i+1 ◦ i for all i. Morphisms can be composed: f ◦ g is the family
(fi ◦ gi )i∈Z . We denote by Ch• (R) the category of cochain complexes.
Example 81. If M is a manifold, the De Rham complex of M (see the
differential geometry course) is a cochain complex of R-vector spaces.

Important Observation

Cochain complexes are nothing but chain complexes with

a different notation. To be more specific, one passes from
cochain complexes to chain complexes (and from morphisms
of chain complexes to morphisms of cochain complexes) by
the formulas:

Ci = C −i and fi = f −i .

In particular, all the concepts and all the results relative to chain com-
plexes can be converted into concepts and results relative to cochain com-
plexes, simply by changing the notations.
1. The cohomology of degree i a cochain complex is the subquotient of C i
defined by:
H i (C) = Z i /B i
where Z i = Ker d : C i → C i+1 and B i = Im d : C i−1 → C i .

Cohomology defines functors:

H i : Ch• (R) → R−Mod .

27
6.2 Duality A. Touzé – 2023

A morphism f : C → D is called a quasi-isomorphism if H i (f ) is an

isomorphism for all i.
Theorem 82. If 0 → C → D → E → 0 is a short exact sequence of
cochain complexes, there is a long exact sequence in cohomology:
∂ ∂
→ H i (C) → H i (D) → H i (E) −
... − → H i+1 (C) → . . . .

Bonus: every morphism of short exact sequence of complexes induces

a morphism of cohomology long exact sequences.

2. Two morphisms of cochain complexes f, g : C → D are homotopic if

there exists a family of maps hi : C i → Di−1 such that f i − g i =
d ◦ hi + hi+1 ◦ d. Notation : f ∼ g. We say that f : C → D is a
homotopy equivalence if there is a morphism g : D → C such that
f ◦ g ∼ idD and g ◦ f ∼ idC .
Proposition 83. Two homotopic morphisms induce the same mor-
phism in cohomology. In particular, every homotopy equivalence is a
quasi-isomorphism.

6.2 Duality
If R is a commutative ring, every R-module M has a dual module M ∨ =
HomR (M, R). Every R-linear map f : M → N induces an R-linear map
f ∨ : N ∨ → M ∨ , such that f ∨ (u) = u ◦ f . The map f ∨ is also called
the transpose map of f . Duality does not define a functor, but rather a
contravariant functor in the following sense.
Definition 84. Let C and D two categories. A contravariant functor F :
C → D is the data
• for all objects X of C, of an object F (X) of D,

• for all morphisms f : X → Y , of a ’backward morphism’ F (f ) :

F (Y ) → F (X), such that this assignment is associative: F (f ◦ g) =
F (g) ◦ F (f ), and preserves identities: F (idX ) = idF (X) .
For all chain complexes C of R-modules, we can use duality to construct
a cochain complex of R-modules:
d∨ d∨ d∨ d∨ d∨ d∨
C ∨ = · · · −→ C0∨ −→ C1∨ −→ · · · →−→ Cn∨ −→ ∨
Cn+1 −→ . . .
|{z} |{z} |{z} | {z }
=:(C ∨ )0 =:(C ∨ )1 =:(C ∨ )n =:(C ∨ )n+1

One may expect that the cohomology of C ∨ can be computed from the ho-
mology of C by a nice formula. This is indeed the case. The next proposition
gives such a formula when R is a field. We will see later a formula for more
general rings R.

28
6.3 Singular cohomology A. Touzé – 2023

Proposition 85. Assume that R is a field. Then for all i we have an

isomorphism:
H i (C ∨ ) ' Hi (C)∨
If R is not a field, the following example shows that the formula must
be more complicated than that of proposition 85
Example 86. Let R = Z. Let C denote the complex with only nonzero ob-
jects C0 = C1 = Z, and with differential d : C1 → C0 given by multiplication
by a positive integer n. Then H0 (C) = Z/p and H 1 (C) = 0, hence
H0 (C)∨ = H1 (C)∨ = 0 .
But C ∨ is the complex Z → Z with differential given by multiplication by
n, hence
H0 (C ∨ ) = 0 and H1 (C ∨ ) = Z/n .

6.3 Singular cohomology

Definition 87. Let R be a commutative ring and let X be a topologi-
cal space. The dual C sing (X)∨ of the singular chain complex is a cochain
complex and for all i ≥ 0 we define the cohomology module of degree i by:
H i (X) := H i (C sing (X)∨ ) .
Every continuous map f : X → Y induces a ’backward map’:
H i (f ) := H i (C sing (f )∨ ) : H i (Y ) → H i (X) .
In particular, singular cohomology of degree i defines a contravariant func-
tor:
H i : Top → R−Mod .
By applying duality ∨ to C sing (X), one can prove that cohomology sat-
isfies properties which are dual to the properties of homology. In particular:
• Two homotopic continuous maps f, g : X → Y induce the same mor-
phism in cohomology. Hence, homotopy equivalences yield isomor-
phisms in cohomology.
• If U and V are two open subsets of X such that U ∪ V = X there is
a long exact sequence:
∂
0 → H 0 (X) → H 0 (U ) ⊕ H 0 (V ) → H 0 (U ∩ V ) −
→ H 1 (X) → . . .
∂ ∂
· · · → H i−1 (U ∩ V ) −
→ H i (X) → H i (U ) ⊕ H i (V ) → H i (U ∩ V ) −
→ ...

Natural question 88. Since cohomology has the ’same’ formal properties
as homology, and since cohomology can even be computed from homology
by a nice formula, what is the point in introducing cohomology ?
The answer to this will be given in the next section.

29
6.4 The cup product A. Touzé – 2023

6.4 The cup product

Definition 89. Let R be a commutative ring. A graded R-algebra A is a
R-algebra A equipped with an R-module decomposition:

A=
M
Ai
i≥0

and such that the product is compatible with this decomposition, that is, if
a ∈ Ai and b ∈ Aj then ab ∈ Ai+j . An element of A is homogeneous if it
belongs to an Ai . In this case, i is called the degree of a and denoted by |a|

Example 90. Tensor algebras, symmetric algebras, exterior algebras, can

be equipped with the structure of graded algebras.

Definition 91. A differential graded R-algebra is a graded algebra A,

equipped with differentials d : Ai → Ai+1 satisfying d ◦ d = 0, and the
derivation rule:
d(ab) = d(a)b + (−1)|a| ad(b) .

Since a differential graded algebra A is a complex of R modules, we can

take its homology.

Proposition 92. Let A be a differential graded R-algebra. Set H ∗ (A) =

i≥0 H (A). Then H (A) is a graded algebra with product:
L i ∗

H i (A) × H j (A) → H i+j (A)

.
([y], [z]) 7→ [yz]

Now let us go back to singular cohomology. An element of degree i of

the cochain complex C sing (X)∨ is an R-linear function φ : Cising (X) → R.
Since Cising (X) is the free R-module with basis the set of singular i-simplices
of X,such a function is uniquely determined by its linearity and its values
on the i-simplices.

Definition 93. We define the cup product

Cising (X)∨ × Cjsing (X)∨ → Ci+j

sing
(X)∨
(φ, ψ) 7→ φ^ψ

by the formula:

(φ ^ ψ)(σ) = φ(σ|<e0 ,...,ei > )ψ(σ|<ei ,...,ei+j > ) .

Theorem 94. The cup product endows C sing (X)∨ with the structure of a
differential graded algebra. In particular, the cohomology H ∗ (X) is a graded
R-algebra. Moreover, every continuous map f : X → Y induces a morphism
of graded R-algebras H ∗ (f ) : H ∗ (Y ) → H ∗ (X).

30
6.5 Vista : some applications of singular cohomology A. Touzé – 2023

6.5 Vista : some applications of singular cohomology

Our purpose here is just to give a glance at what one can do with singular
cohomology. We first give three elementary ways of using cup products.

1. Distinguish spaces with the same (co)homology. The torus T 2

and the wedge of spheres S 1 ∨ S 1 ∨ S 2 have the same homology and
cohomology modules, namely

R

 if i = 0, 2,
H (T ) = H (S ∨ S ∨ S ) =
i 2 i 1 2 1
R2 if i = 1,

0

otherwise.

But they have different cohomology rings, namely H ∗ (T 2 ) is an ex-

terior algebra on two generators a and b of degree 1. In particular
the product a ^ b is nonzero. On the contrary, if x ^ y = 0 for all
elements x, y in H 1 (S 1 ∨ S 2 ∨ S 1 ).

2. Give a better control of maps in homology. One computes that

H ∗ (RP n , F2 ) = F2 [x]/xn+1 with deg(x) = 1 .

Algebraically, every morphism of F2 -algebras φ : F2 [x]/xn+1 → A is

completely determined by the element φ(x) ∈ A. Thus for all con-
tinuous map f : X → RP n , if we know the ring H ∗ (X, F2 ) then it
suffices to compute the map H 1 (f, F2 ) to completely determine the
ring morphism H ∗ (f, F2 ).
Here is a slightly different way of using this ring description. If N > n,
then every morphism of rings

φ : F2 [x]/xn+1 → F2 [x]/xN +1

is zero in positive degrees. We deduce that for all continuous maps

f : RP N → RP n we have H i (f, F2 ) = 0 for all positive i.
Similar considerations can be adapted to complex projective spaces
since for all rings R, H ∗ (CP n , R) = R[x]/xn+1 where deg x = 2.

3. Study coverings by contractible sets. If X can be covered by n

contractible sets then for all cohomology classes c1 , . . . , cn we have

c1 ^ · · · ^ cn = 0

For example RP n and CP n can’t be covered by less than (n + 1)

contractible sets.

We also have a comparison theorem with De Rham cohomology.

31
6.5 Vista : some applications of singular cohomology A. Touzé – 2023

Theorem 95 (De Rham). Let M be a manifold. There is an isomorphism

of R-algebras, natural with respect to M (the right hand side is the De Rham
cohomology of M ):
H ∗ (M, R) ' HDR
∗
(M ) .

This theorem tells us that despite its definition relying on differential

geometry, de Rham cohomology only measures properties of the underlying
topological spaces of differentiable manifolds. Thus we can use topological
methods to determine De Rham cohomology.

32
 A. Touzé – 2023

Part II
Algebra and homology
7 Projective modules and Ext
7.1 Projective modules
Definition 96. A functor F : R−Mod → S−Mod is left exact if for all
short exact sequences 0 → M 0 − − M 00 → 0, the diagram
→M →
f g

F (f ) F (g)
0 → F (M 0 ) −−−→ F (M ) −−−→ F (M 00 )

is an exact sequence. It is exact if in addition F (g) is surjective.

Let R be a ring and let M be an R-module. We consider the functor2

HomR (M, −) : R−Mod → Z−Mod

which sends an R-module M to the abelian group HomR (M, N ), and a mor-
phism f : N → N 0 to the morphism f ◦ − : HomR (M, N ) → HomR (M, N 0 ).
This functor is left exact but not exact in general.

Example 97. Let R = Z, and let M = Z/n. Then HomZ (M, −) is not
exact, as one can see it by considering the image of 0 → Z → Z → Z/n → 0.

Definition 98. An R-module M is called projective if the functor

HomR (M, −) is exact.

Proposition 99. Let P be an R-module. The following are equivalent.

1. P is projective.

2. Lifting property. For all surjective morphisms g : N  N 0 and all

u : P → N 0 there is a morphism u : P → N such that the following
diagram commutes (we say that u is a lifting of u):

N
u g .
u
P N0

3. P is a direct summand of a free R-module.

2
If R is a ring, then HomR (M, N ) is only an abelian group in general. Indeed, one would
like to define the linear combination of R-linear maps by (λf + µg)(x) = λf (x) + µg(x),
but if R is not commutative, (λf + µg)(rx) = λrf (x) + µrg(x) may differ from r(λf + µg),
so that the linear combination may fail to be an R-linear map.

33
7.2 Projective resolutions A. Touzé – 2023

7.2 Projective resolutions

Not all R-modules are projective, but every R-module can be approximated
by projective R-modules. Such approximations are called projective resolu-
tions.

Definition 100. A resolution of M is a chain complex of R-modules P

(zero in negative degrees) equipped with a map  : P0 → − M such that the
following complex is exact (i.e. has homology zero in each degree):
d d d d d 
... −
→ Pn −
→ Pn−1 −
→ ... −
→ P1 −
→ P0 →
− M.

A projective resolution is a resolution such that every Pi is a projective

R-module.

Proposition 101. Every R-module M admits a projective resolution.

The next result is sometimes called the "fundamental theorem of homo-

logical algebra".

Theorem 102. Let P be a complex of projective objects (zero in negative

degrees), equipped with a map M : P0 → M , and let (R, N ) be a resolution
of N . Then for all R-linear maps f : M → N there is a morphism of chain
complexes f : P → R such that the following square commutes:

f0
P0 R0
M N .
f
M N

Such a morphism f is unique up to homotopy. (And the morphism f is

called a lifting of the R-linear map f ).

Corollary 103 (Uniqueness of projective resolutions). If (P, ) and (Q, )

are two projective resolutions of M , there is a homotopy equivalence P → Q
which lifts idM .

7.3 Ext
Let k be a commutative ring. Then a k-algebra is a ring R equipped with
a morphism of rings k → R with image contained in the center of R (thus,
a ring is nothing but a Z-algebra, and every commutative ring is an alge-
bra over itself via the identity morphism of R). If R is a k-algebra, then
HomR (M, N ) is not only an abelian group, but also a k-module (the action
of a scalar λ ∈ k is defined by (λf )(m) = λf (m) for all m ∈ M ).

34
7.3 Ext A. Touzé – 2023

Definition 104. Let R be a k-algebra. Let M be an R-module, and let P

be a projective resolution of M . Then for all N , the k-module extensions of
degree i between M and N is defined for all i ≥ 0 by:

ExtiR (M, N ) = H i (HomR (P, N )) .

For all R-linear maps f : M → M 0 we denote by

ExtiR (f, N ) : ExtiR (M 0 , N ) → ExtiR (M, N )

the k-linear map induced in cohomology by the morphism of complexes

HomR (f , N ) : HomR (P 0 , N ) → HomR (P, N ), where f is a lifting of f on
the level of the projective resolutions. For all R-linear maps g : N → N 0 we
denote by
ExtiR (M, g) : ExtiR (M, N ) → ExtiR (M, N 0 )
the k-linear map induced in cohomology of degree i by the morphism of
complexes HomR (P, g) : HomR (P, N ) → HomR (P, N ).

Remark 105. The Ext-modules do not depend on the choice of the pro-
jective resolution P . If P 0 another choice, there is a homotopy equivalence
f : P → P 0 which induces a canonical isomorphism between the two ver-
sions of Ext. Similarly, the map ExtiR (f, N ) does not depend on the choice
0
of the lifting f . Indeed, two such liftings f and f are homotopic, hence
0
the morphisms of complexes HomR (f , N ) and HomR (f , N ) are homotopic,
hence they induce the same map in homology.

Theorem 106 (fundamental properties of Ext).

1. Bifunctoriality. For all M and N , and all i, the k-modules of degree

i extensions define a functor

ExtiR (M, −) : R−Mod → k−Mod ,

and a contravariant functor

ExtiR (−, N ) : R−Mod → k−Mod .

Moreover for all R-linear maps f : M → M 0 and g : N → N 0 there is

a commutative square:

ExtiR (M 0 ,g)
ExtiR (M 0 , N ) ExtiR (M 0 , N 0 )
ExtiR (f,N ) ExtiR (f,N 0 )
ExtiR (M,g)
ExtiR (M, N ) ExtiR (M, N 0 ) .

35
7.3 Ext A. Touzé – 2023

2. Degree zero. There is an isomorphism, natural with respect to M

and N :
Ext0R (M, N ) ' HomR (M, N ) .

3. Direct sums and products. There are isomorphisms, natural with

respect to M , Mα , N and Nβ :

ExtiR ( Mα , N ) ' ExtiR (Mα , N ) ,

M Y

α∈A α∈A

ExtR (M, Nβ ) ExtiR (M, Nβ ) .

Y Y
i
'
β∈B β∈B

4. Projectives. The following assertions are equivalent.

(a) M is projective,
(b) Ext1R (M, N ) = 0 for all R-modules N ,
(c) ExtiR (M, N ) = 0 for all positive i and all N .
f g
5. Long exact sequences. Every short exact sequence 0 → N 0 −
→N →
−
N → 0 induces a long exact sequence:
00

Ext0 (M,f ) Ext0 (M,g)

0 → Ext0R (M, N 0 ) −−−−
R
−−−→ Ext0R (M, N ) −−−−
R
−−−→ . . .
Exti (M,g) ∂ Exti+1 (M,f )
R
. . . −−−−−−−→ ExtiR (M, N 00 ) −
→ Exti+1 0
R (M, N ) −
R
−−− −−−−→ . . .

Bonus. Moreover, every morphism f : M1 → M2 and every morphism

of short exact sequences

0 N10 N1 N100 0
α0 α α00

0 N20 N2 N200 0

induces a morphism between the corresponding exact sequences. In

particular we have commutative squares:
∂
ExtiR (M, N100 ) Exti+1
R (M, N1 )
0

ExtiR (M,α00 ) Exti+1 0

R (M,α )
,
∂
ExtiR (M, N200 ) Exti+1
R (M, N2 )
0

∂
ExtiR (M2 , N 00 ) Exti+1
R (M2 , N )
0

ExtiR (f,N 00 ) Exti+1 0

R (f,N )
.
∂
ExtiR (M1 , N 00 ) Exti+1
R (M1 , N )
0

36
7.3 Ext A. Touzé – 2023

Now we justify the name Ext for the homological notion introduced here.

Definition 107. Let M and N be two R-modules. An extension of N by

M is a short exact sequence:
u v
0→M − →N →0.
→E−

Two extensions of N by M are called equivalent is there is an isomorphism

φ making the following diagram commutative:
u v
0 M E N 0
' φ .
u0 v0
0 M E0 N 0

An extension is called split if it is equivalent to the trivial extension (here

ιM is the canonical injection and πM the canonical projection):
ι π
0 → M −−
M
→N →0.
→ M ⊕ N −−N

We let E 1 (M, N ) be the set of equivalent classes of extensions of N by M .

Theorem 108. There is a bijection, which sends 0 to the class of the split
extension:
Ext1R (M, N ) ' E 1 (M, N ) .

Example 109. Take R = Z and let p be a prime integer. Every extension

of Z/p by Z is either trivial, or equivalent to the following exact sequence,
for a uniquely determined a ∈ Z/p \ {0}, where πa denotes the unique map
such that π(1) = a:
p πa
0→Z→ − Z −→ Z/p → 0 .

37
8. Injective modules and Ext A. Touzé – 2023

8 Injective modules and Ext

8.1 Injective modules
Definition 110. A contravariant functor F : R−Mod → S−Mod is left
exact if for all short exact sequences 0 → M 0 − − M 00 → 0, the diagram
→M →
f g

F (g) F (f )
0 → F (M 00 ) −−−→ F (M ) −−−→ F (M 0 )

is an exact sequence. It is exact if in addition F (f ) is surjective.

Let k be a commutative ring, let R be a k-algebra, and let M be an

R-module. The contravariant functor

HomR (−, M ) : R−Mod → k−Mod

is left exact but not exact in general.

Example 111. Let R = k = Z, and let M = Z/n. Then HomZ (−, M ) is not
exact, as one can see it by considering the image of 0 → Z → Z → Z/n → 0.

Definition 112. An R-module M is called injective if the functor

HomR (−, M ) is exact.

Proposition 113. Let J be an R-module. The following are equivalent.

1. J is injective.

2. Extension property. For all injective morphisms g : N ,→ N 0 and

all u : N → J there is a morphism u : N → J such that the following
diagram commutes (we say that u is an extension of u to N 0 ):

N0
g u .
u
N J

3. Restricted extension property. For all left ideal I ⊂ R, every map

u : I → J extends to a map u : R → J.

Remarks 114. It is clear from the extension property that a product of

injective R-modules is again injective.
The last criterion is known as "Baer’s criterion". It says that instead of
testing the extension property for all submodules of all modules, one may
test the extension property for all submodules of one module, namely R.

We now describe in more details the injective modules for various classes
of rings.

38
8.1 Injective modules A. Touzé – 2023

8.1.1 Injective modules for fields

Proposition 115. If R is a field, then every R-module is injective.

8.1.2 Injective modules for PIDs

Lemma 116 ("Injective = divisible" for PID). Let R be a PID. Then J is
injective if and only if it is divisible, that is, if and only if for all nonzero r
in R and all a ∈ J there is an b ∈ J such that rb = a.
Proposition 117. Let R be a PID, let F denote its fraction field, and
assume that R 6= F (i.e. R is not a field).
(i) The R-modules F and F/R are injective.
(ii) An R-module M is injective if and only if it is a direct summand of a
product of copies of F/R.
(iii) Every R-module is a submodule of a product of copies of F/R.

8.1.3 Injective modules for general rings

Definition 118. One says that J is an injective cogenerator of R-modules
if J is injective and if every R-module embeds into a product of copies of J.
Lemma 119. If J is an injective cogenerator of R-modules, then every
injective R-module is a direct summand of a product of copies of J.
Example 120. If R is a field, then R is an injective cogenerator. If R is a
PID with fraction field F 6= R, then F/R is an injective cogenerator.
Now we are going to prove that if R → S is a morphism of rings, then
one can construct an injective cogenerator of S-modules from an injective
cogenerator of R-modules. The main tool for this is coinduction.
Definition 121. Let S → R be a morphism of rings. If M is an S-module,
we define an R-module structure on the abelian group HomS (R, M ) by let-
ting (λf )(x) := f (xλ). This defines acoinduction functor:
HomS (R, −) : S−Mod → R−Mod .
Proposition 122 (Properties of coinduction.).

1. Left exactness. Coinduction is a left exact functor.

2. Adjunction. Let resR S : R−Mod → S−Mod denote the functor
which sends every R-module M to the underlying S-module. There
is an isomorphism, natural with respect to the R-module N and the
S-module N :

HomS (resR
S N, M ) ' HomR (N, HomS (R, M )) .

39
8.2 Injective coresolutions A. Touzé – 2023

3. Injectives. If J is injective, then HomS (R, J) is injective as well.

4. Injective cogenerators. If J is an injective cogenerator of S-

modules, then HomS (R, J) is an injective cogenerator of R-modules.

Corollary 123. 1. For all rings R, the coinduced module HomZ (R, Q/Z)
is an injective cogenerator of R-modules.

2. For all k-algebras R over a field k, the coinduced module Homk (R, k)
is an injective cogenerator of R-modules.

8.2 Injective coresolutions

The treatment of injective coresolutions is parallel to that of projective res-
olutions.

Definition 124. A coresolution of M is a cochain complex of R-modules J

(zero in negative degrees) equipped with a map η : M → − J 0 such that the
following complex is exact (i.e. has homology zero in each degree):
η d d d d d
→ J0 −
M− → J1 − → Jn −
→ ... − → J n+1 −
→ ... .

An injective coresolution is a coresolution such that every J k is an injective

R-module.

Proposition 125. Every R-module M admits an injective coresolution.

Theorem 126. Let J be a complex of injective objects (zero in negative

degrees), equipped with a map ηN : N → J 0 , and let (R, M ) be a coresolution
of M . Then for all R-linear maps f : M → N there is a morphism of chain
complexes f : P → R such that the following square commutes:
0
f
R0 J0
ηM ηN .
f
M N

Such a morphism f is unique up to homotopy. (And the morphism f is

called a lifting of the R-linear map f ).

Corollary 127 (Uniqueness of injective coresolutions). If (J, η) and (Q, η)

are two injective coresolutions of M , there is a homotopy equivalence P → Q
which lifts idM .

40
8.3 Ext with injective coresolutions A. Touzé – 2023

8.3 Ext with injective coresolutions

Theorem 128. Let M and N be R-modules, and let (Q, η) be an injec-
tive coresolution of N . Then for all i, ExtiR (M, N ) is isomorphic to the
cohomology of the cochain complex HomR (M, Q).
Moreover, for all R-linear maps f : M → M 0 , the map ExtiR (f, N ) is
the map induced in cohomology by the morphism of complexes HomR (f, Q).
Similarly, for all R-linear maps g : N → N 0 , the map ExtiR (M, g) is the map
induced in cohomology by the morphism of complexes HomR (M, g), where g
denotes a lifting of g to the level of the injective coresolutions.

The proof of theorem 128 will be given later (see section 9.2.)

Corollary 129. Every short exact sequence of modules 0 → M 0 → M →

M 00 → 0 induces a long exact sequence:
∂
0 → Ext0R (M 00 , N ) → Ext0R (M, , N ) → Ext0 (M 0 , N ) −
→ Ext1R (M 00 , N ) → . . .
∂
· · · → ExtiR (M 00 , N ) → ExtiR (M, N ) → ExtiR (M 0 , N ) −
→ Exti+1 00
R (M , N ) → . . .

Proof. We use the definition of Ext from an injective coresolution (Q, η) of

N . We have a short exact sequence of complexes

0 → HomR (M 00 , Q) → HomR (M, Q) → HomR (M 0 , Q) → 0 ,

and the associated long exact sequence gives the result.

8.4 An application to semi-simplicity

Proposition 130 (Homological characterization of semi-simplicity.). Let R
be an algebra over a field k. Then the following statements are equivalent.

(1) Every R-module M of finite dimension over k is a sum of simple R-

modules (i.e. the category of finite dimensional R-modules is semi-
simple).

(2) For all R-modules M, N of finite dimension over k, Ext1R (M, N ) = 0.

(3) For all simple R-modules M, N of finite dimension over k,

Ext1R (M, N ) = 0.

41
9. Bicomplexes and their applications A. Touzé – 2023

9 Bicomplexes and their applications

9.1 Cochain bicomplexes
Definition 131. A (cochain) bicomplex of R-modules is a family of R-
modules (C i,j )(i,j)∈Z2 equipped with R-linear maps d : C i,j → C i+1,j and
δ : C i,j → C i,j+1 which satisfy:

d◦d=0=δ◦δ , d◦δ =δ◦d.

The elements of C i,j are called the homogeneous elements of bidegree (i, j),
the maps d and δ are the differentials of the bicomplex. Cochain bicomplexes
are depicted as commutative diagrams:

.. ..
. .
δ δ
d d d
... C i,j+1 C i+1,j+1 ...
δ δ .
d d d
... C i,j C i+1,j ...
δ δ

.. ..
. .

A morphism of bicomplexes f : C → D is a family of R-linear maps f i,j :

C i,j → Di,j which preserves the differentials, i.e.

d ◦ f i,j = f i+1,j ◦ d , δ ◦ f i,j = f i,j+1 ◦ δ .

Morphims of bicomplexes can be composed degreewise (i.e. (f ◦ g)i,j =

f i,j ◦ g i,j ), so that cochain bicomplexes and morphisms of cochain complexes
form a category Ch•,• (R).
Examples 132. 1. Every cochain complex can be regarded as a bicom-
plex concentrated in the 0-th row, i.e. as a cochain bicomplex with
C i,0 := C i and C i,j = 0 for j 6= 0, and with only nonzero differential
given by the differential of the complex.

2. Every cochain complex can be regarded as a bicomplex concentrated

in the 0-th column, i.e. as a cochain bicomplex with C 0,j := C j and
C i,j = 0 for i 6= 0, and with only nonzero differential given by the
differential of the complex.

3. If C is a chain complex and D is a cochain complex, we can form a

cochain bicomplex HomR (C, D) with HomR (Ci , Dj ) placed in bidegree
(i, j), and with differentials induced by those of C and D:

42
9.1 Cochain bicomplexes A. Touzé – 2023

.. ..
. .

HomR (d,Dj+1 )
... HomR (Ci , Dj+1 ) HomR (Ci+1 , Dj+1 ) ...
HomR (Ci ,d) HomR (Ci+1 ,d) .
HomR (d,Dj )
... HomR (Ci , Dj ) HomR (Ci+1 , Dj ) ...

.. ..
. .

Definition 133. The total complex associated to a bicomplex (C, d, δ) is

the complex Tot C defined in each degree by:

(Tot C)k =
M
C i,j
i+j=k

and whose differential dTot : (Tot C)k → (Tot C)k+1 sends x ∈ C i,j to

dTot (x) = d(x) + (−1)i δ(x) .

Every morphism of bicomplexes f : C → D induces a morphism of complexes

Tot f : Tot C → Tot D, namely, if x ∈ C i,j then (Tot f )(x) = f i,j (x).

Example 134. If C is a cochain complex viewed as a bicomplex concen-

trated in the 0-th row or the 0-th column, the total complex associated to
this bicomplex is isomorphic to the cochain complex C.

Theorem 135. Let f : C → D be a morphism of bicomplexes. Assume that

for all i, the morphism of chain complexes given by restricting f to the i-th
column
f i,• : (C i,• , δ) → (Di,• , δ)
is a quasi-isomorphism. Assume furthermore that C and D are such that
for all k the direct sums
M M
C i,j and Di,j
i+j=k i+j=k

are finite. Then Tot f is a quasi-isomorphism.

Remark 136. The hypothesis that f is a column-wise quasi-isomorphism

can be replaced by the hypothesis that f is a row-wise quasi-isomorphism
(i.e for all fixed j the morphism f •,j : (C •,j , d) → (D•,j , d) between the
j-th rows is a quasi-isomorphism), with the same conclusion that Tot f is a
quasi-isomorphism (and a similar proof).

43
9.2 The proof of theorem 128 A. Touzé – 2023

9.2 The proof of theorem 128

Theorem 135 can be used to prove theorem 128 in the following way. Let
(P, ) be a projective resolution of M and (Q, η) be an injective coresolution
of N . Then we have three cochain bicomplexes C, C 0 and C 00 whose objects
are given by
C i,j = Hom(Pi , Qj ),
HomR (Pi , N ) if j = 0,
(
0 i,j
(C ) =
0 if j =
6 0,
HomR (M, Qj ) if i = 0,
(
00 i,j
(C ) =
0 if i =
6 0.
The map  and η induce morphisms of bicomplexes
HomR (η,P ) HomR (,Q)
C 0 −−−−−−−→ C ←−−−−−−− C 00 .
And theorem 135 shows that these morphisms induces quasi-isomorphisms
on the level of the associated total complexes. Now the cohomology of Tot C 0
is the definition of Ext via projective resolutions while the cohomology of
Tot C 00 is the definition of Ext via injective resolutions.

9.3 The Künneth formula

Definition 137. If (C, dC ) and (D, dD ) are two cochain complexes of R-
modules over a commutative ring R, their tensor product is the cochain
complex (C ⊗ D, d) such that
(C ⊗ D)k =
M
C i ⊗ Dj ,
i+j=k

d(x ⊗ y) = dC (x) ⊗ y + (−1)i x ⊗ dD (y) if x ⊗ y ∈ C i ⊗ Dj .

If z is a cycle of degree i in C and z 0 is a cycle of degree j in D, then z ⊗z 0
is a cycle of degree i + j of C ⊗ D. Moreover, if z or z 0 is a boundary, then
z ⊗ z 0 is a boundary. Hence we have a map, called the Künneth morphism:
κ: H i (C) ⊗ H j (D) → H k (C ⊗ D)
L
i+j=k
.
[z] ⊗ [z 0 ] 7→ [z ⊗ z 0 ]
The following theorem is a consequence of theorem 135.
Theorem 138 (Künneth formula). Assume that R is a field. Then κ is an
isomorphism for all k.
Remark 139. The theorem fails if R is not a field. For example, take
×n
R = Z, C = Z −−→ Z (viewed as a complex in degrees 0 and 1) and
D = Z/nZ (viewed as a complex concentrated in degree 0). Then in degree
0, the domain of the Künneth morphism is zero, whereas its codomain is
Z/nZ.

44
9.4 Chain bicomplexes A. Touzé – 2023

9.4 Chain bicomplexes

There is an obvious notion of a chain bicomplex, where degrees are de-
noted by indices and differentials lower the degree. Chain complexes form a
category Ch•,• (R). Just as in the case of complexes, the difference between
cochain bicomplexes and chain bicomplexes is just a matter of notation. One
can convert cochain bicomplexes and their morphisms into chain complexes
and their morphisms by the formulas

C i,j = C−i,−j , f i,j = f−i,−j ,

so that we have an equality Ch•,• (R) = Ch•,• (R).

If we use the chain complex notation instead of the cochain complex
notation, then theorems 135 and 138 translate as follows.

• Let f : C → D be a morphism of bicomplexes, such that f is column-

wise a quasi-isomorphism of complexes, or such that f is row-wise a
quasi-isomorphism of complexes. Assume furthermore that for all k
the direct sums
Ci,j and
M M
Di,j
i+j=k i+j=k

are finite. Then Tot f is a quasi-isomorphism.

• If C and D are chain complexes of R-modules over a field R, the

Künneth morphism induces an isomorphism of R-modules:

κ: Hi (C) ⊗ Hj (D) → Hk (C ⊗ D)
L
i+j=k
.
[z] ⊗ [z 0 ] 7→ [z ⊗ z 0 ]

9.5 Vista: the singular homology of products of spaces

Let us briefly mention an application of the Künneth theorem 138 in the
context of singular homology of topological spaces.
If X and Y are two topological spaces and R is a commutative ring, one
can prove that there is an homotopy equivalence:

C sing (C) ⊗ C sing (Y ) → C sing (X × Y ) .

If R is a field, the homology of the left-hand-side can be computed by the

Künneth theorem. Hence we obtain the following statement.

Theorem 140 (Topological Künneth theorem). Assume that R is a field.

Then for all topological spaces X and Y and all integers k there is an iso-
morphism:
Hi (X) ⊗ Hj (Y ) ' Hk (X × Y ) .
M

i+j=k

45
9.5 Vista: the singular homology of products of spaces A. Touzé – 2023

Example 141. The torus T 2 is homeomorphic to S 1 × S 1 . The topological

Künneth theorem allows one to compute its homology from that of S 1 :

H0 (T 2 ) =H0 (S 1 ) ⊗ H0 (S 1 ) = R ,
H1 (T 2 ) =H1 (S 1 ) ⊗ H0 (S 1 ) ⊕ H0 (S 1 ) ⊗ H1 (S 1 ) = R2 ,
H2 (T 2 ) =H1 (S 1 ) ⊗ H1 (S 1 ) = R .

More material on this topic can be found e.g. in [Hatcher, Section 3.B].

46
10. Flat modules and Tor A. Touzé – 2023

10 Flat modules and Tor

10.1 Tensor products over noncommutative rings
I assume that you are familiar with the tensor product of modules over a
commutative ring. Here is the more general definition when the ring is not
commutative.

Definition 142. Let R be a k-algebra over a commutative ring k. For all

right R-modules M and all left R-modules N , the tensor product M ⊗R N
is the k-module M ⊗R N defined by:
M ⊗k N
M ⊗R N = .
hmr ⊗ n − m ⊗ rn | (m, n, r) ∈ M × N × Ri

If f : M → M 0 and g : N → N 0 are R-linear, they define a k-linear map:

f ⊗ g : M ⊗R N → M 0 ⊗R N 0

such that (f ⊗ g)(m ⊗ n) = f (m) ⊗ g(n). The composition (f 0 ⊗ g 0 ) ◦ (f ⊗ g)

equals (f 0 ◦ f ) ⊗ (g 0 ◦ g). In particular, we have functors:

M ⊗R − : R−Mod → k−Mod ,
− ⊗R N : Mod − R → k−Mod .

Remark 143. If R is a k-algebra, then it is also a ring, hence a Z-algebra.

Thus we have two tensor products M ⊗R N : the one defined using the k-
algebra structure and the one defined using the Z-algebra structure. One
can show that there is in fact a canonical isomorphism of abelian groups
between these two objects (see exercise 55 in exercise sheet 8).

This construction has the same basic properties as the tensor product
over commutative rings, namely:

1. The universal property,

2. The adjunction property,

3. Additivity and unitality,

4. Right exactness.

We now review this properties in detail.

Definition 144. Let M be a right R-module, let N be a left R-module,

and let P be a k-module. A map f : M × N → P is (R, k)-bilinear if it is
k-linear with respect to each variable, and if moreover for all r ∈ R we have:

f (mr, n) = f (m, rn).

47
10.1 Tensor products over noncommutative rings A. Touzé – 2023

Example 145. The map π : M × N → M ⊗R N is (R, k)-bilinear.

(m, n) 7→ m ⊗ n
Proposition 146 (Universal property). For all (R, k)-bilinear maps f :
M × N → P there is a unique k-linear map f : M ⊗R N → P such that
f = f ◦ π:
f
M ×N P
π .
∃!f
M ⊗R N

Proposition 147 (Adjunction). There are isomorphisms of k-modules, nat-

ural with respect to M , N , P :

HomR (M, Homk (N, P )) ' Homk (M ⊗R N, P ) ' HomR (N, Homk (M, P )) .

Proposition 148 (Additivity and unitality). We have isomorphisms, nat-

ural with respect to the modules M , Mα , N , Nβ :

(Mα ⊗R N ) ' ( Mα ) ⊗R N ,
M M

α∈A α∈A

(M ⊗R Nβ ) ' M ⊗R ( Nβ ) ,
M M

β∈B β∈B

R ⊗R N ' N ,
M ⊗R R ' M .

One can prove the isomorphisms of proposition 148 by using the adjunc-
tion property together with the following useful lemma.
Lemma 149. Let S be a ring, and let θ : HomS (M, P ) → HomS (M 0 , P ) be
a morphism of abelian groups which is natural with respect to P . Then there
exists a unique S-linear morphism f : M 0 → M such that θ(g) = g ◦ f for
all g. Moreover, θ is an isomorphism if and only if f is an isomorphism.
Definition 150. A functor F : R−Mod → S−Mod is right exact if for all
short exact sequences 0 → M 0 − − M 00 → 0, the diagram
→M →
f g

F (M 0 ) −−−→ F (M ) −−−→ F (M 00 ) → 0
F (f ) F (g)

is exact. It is exact if in addition F (f ) is injective.

In the previous definition we are considering left modules, but the defi-
nition applies as well to right modules, since Mod − R = Rop −Mod where
Rop is the opposite ring of R.
Proposition 151 (Right exactness). The functors M ⊗R − and − ⊗R N
are right exact.

48
10.2 Flatness A. Touzé – 2023

One can prove proposition 151 by using the adjunction property together
with the following useful lemma.

Lemma 152. A sequence of S-modules M 0 → M → M 00 → 0 is exact if

and only if for all S-modules P , the following sequence of modules is exact:

0 → HomS (M 00 , P ) → HomS (M, P ) → HomS (M 0 , P ) .

10.2 Flatness
Definition 153. One says that a left R-module M is flat if the functor
− ⊗R M : Mod − R → Z−Mod is exact.

Proposition 154 (First properties of flatness).

1. If M is projective, then M is flat.

2. If (Mα )α∈A is a family of flat R-modules, then

L
α∈A Mα is flat.

Proposition 155 (Localizations). If R is a commutative ring and if S is

a multiplicative part of R, then S −1 R is a flat R-module.

Example 156. 1. If R is a field, every R-module is free, hence flat.

2. If R = Z, then Z/nZ is not flat and Q is flat but not projective.

10.3 Tor
Definition 157. Let R be a k-algebra. Let M be a right R-module, and let
P be a projective resolution of M . Then for all N , the k-module TorR
i (M, N )
is defined for all i ≥ 0 by:

TorR
i (M, N ) = Hi (P ⊗R N ) .

For all R-linear maps f : M → M we denote by

0
TorR
i (f, N ) : Tori (M, N ) → Tori (M , N )
R R

the k-linear map induced in homology by the morphism of complexes f ⊗R

idN : P ⊗R N → P 0 ⊗R N , where f is a lifting of f on the level of the
projective resolutions. For all R-linear maps g : N → N 0 we denote by
0
TorR
i (M, g) : Tori (M, N ) → Tori (M, N )
R R

the k-linear map induced on homology of degree i by the morphism of com-

plexes P ⊗R g : P ⊗R N → P ⊗R N 0 .

49
10.3 Tor A. Touzé – 2023

Remark 158. The definition of Tor depends on choices. First one chooses a
projective resolution P . However, since two projective resolutions are homo-
topy equivalent, changing the projective resolution only changes the complex
P ⊗R N up to a homotopy equivalence, hence it only changes TorR i (M, N )
up to an isomorphism. Second, the definition of Tori (f, N ) depends on the
R

choice of a lifting f , but since any two liftings are homotopic, they actually
define the same map TorR i (f, N ).

The following result is the Tor-analogue of theorem 128, and it has an

analogous proof.

Theorem 159. Let M be a right R-module, let N be a left R-module and

let (P, ) be a projective resolution of N . There is an isomorphism

i (M, N ) ' Hi (M ⊗R P ) .
TorR

Moreover, for all R-linear maps f : M → M 0 , the map TorR i (f, N ) is the
map induced in homology by the morphism of complexes f ⊗R id. Similarly,
for all R-linear maps g : N → N 0 , the map TorR i (M, g) is the map induced
in homology by the morphism of complexes M ⊗R g, where g denotes a lifting
of g to the level of the projective resolutions.

All the basic properties of Ext have a counterpart for Tor. We gather
these properties in the following long theorem.

Theorem 160. 1. Bifunctoriality. For all M and N , and all i, the

Tor-modules define functors

TorR
i (M, −) : R−Mod → k−Mod ,
TorR
i (−, N ) : Mod − R → k−Mod ,

and moreover for all R-linear maps f : M → M 0 and g : N → N 0

there is a commutative square:

TorR
i (M,g)
TorR
i (M, N ) TorR
i (M, N )
0

TorR TorR 0
i (f,N ) i (f,N )
TorR 0
i (M ,g)
TorR
i (M , N )
0 TorR
i (M , N ) .
0 0

2. Degree zero. There is an isomorphism of k-modules, natural with

respect to M and N :

TorR
0 (M, N ) ' M ⊗R N .

50
10.3 Tor A. Touzé – 2023

3. Additivity. There are isomorphisms of k-modules, natural with

respect to M , Mα , N , Nβ :

TorR
i ( Mα , N ) ' TorR
i (Mα , N ) ,
M M

α∈A α∈A

Tori (M, Nβ ) TorR

i (M, Nβ ) .
M M
R
'
β∈B β∈B

4. Long exact sequences (first variable). Every short exact sequence

f g
0 → M0 − − M 00 → 0 induces a long exact sequence in Tor:
→M →
TorR (g,N )00 0 ∂
i (M, N ) −−−−−−→ Tori (M , N ) −
· · · → TorR i R
→ TorR
i−1 (M , N ) → . . .

0 TorR (f,N ) 00TorR (g,N )

· · · → TorR
0 (M , N ) −−−−−−→ Tor0 (M, N ) −−−−−−→ Tor0 (M , N ) → 0.
0 R 0 R

Bonus. Every morphism f : N1 → N2 and every morphism of short

exact sequences:

0 M10 M1 M100 0
α0 α α00

0 M20 M2 M200 0

induces morphisms between the corresponding long exact sequences. In

particular the following squares commute:

∂
TorR
i (M1 , N )
00 TorR
i−1 (M1 , N )
0

TorR 00
i (α ,N ) TorR 0
i−1 (α ,N )
,
∂
TorR
i (M2 , N )
00 TorR
i−1 (M2 , N )
0

∂
TorR
i (M , N1 )
00 TorR
i−1 (M , N1 )
0

TorR 00
i (M ,f ) TorR 0
i−1 (M ,f )
.
∂
TorR
i (M , N2 )
00 TorR
i−1 (M , N2 )
0

5. Long exact sequences (second variable). Every short exact se-

f g
quence 0 → N 0 − − N 00 → 0 induces a long exact sequence in
→ N →
Tor:
TorR (M,g) 00 0 ∂
· · · → TorR
i (M, N ) −
i
−−− −−−→ TorR
i (M, N ) −
→ TorR
i−1 (M, N ) → . . .

0 TorR (M,f ) TorR (M,g)

00
· · · → TorR
0 (M, N ) −
0
−−− −−−→ TorR
0 (M, N ) −
0
−−− −−−→ TorR
0 (M, N ) → 0.

51
10.3 Tor A. Touzé – 2023

Bonus. As in the case of the first variable, these long exact sequences
are natural with respect to M and with respect to the short exact se-
quence 0 → N 0 → N → N 00 → 0.

6. Flatness. The following assertions are equivalent.

(a) M is flat,
(b) for all N , TorR
1 (M, N ) = 0,
(c) for all N , TorR
i (M, N ) = 0 for all i > 0.

Similarly, N is flat iff TorR

1 (M, N ) = 0 for all M iff Tori (M, N ) = 0
R

for all i > 0 and all M .

The name ’Tor’ comes from the following relation with torsion elements
of R-modules.

Definition 161. Let r ∈ R \ {0}. We denote by r N the module of r-torsion

elements of N :
r M = {m ∈ M | rm = 0} .

If R is a domain, the subset r∈R\{0} r N of all torsion elements of N is a

S

submodule of N that we denote by t(N ).

Remark 162. If R is not a domain, then the subset of torsion elementsof

M may fail to be a submodule of M . For example, take R = Z/6Z and let
M = R. Then {0, 2, 3, 4} is the set of all torsion elements of M , but this set
is not stable under addition : 2 + 3 = 5 is not a torsion element of M !

Proposition 163. For all rings R, and for all elements r ∈ R \ {0} which
is not a zero divisor, there is an isomorphism:

TorR
1 (R/rR, N ) ' r N .

If R is a domain with fraction field F there is an isomorphism:

TorR
1 (F/R, N ) ' t(N ) .

52
11. Universal coefficient formulas A. Touzé – 2023

11 Universal coefficient formulas

Theorem 164. Assume that R is a PID, and let R → S be a morphism of
rings. For all complexes C of free R-modules, we have short exact sequences:

0 → Hi (C) ⊗R S → Hi (C ⊗R S) → TorR
1 (Hi−1 (C), S) → 0 .

The first map in the short exact sequence is given by sending [z] ⊗ s to
[z ⊗ s]. This short exact sequence is natural with respect to C. It splits,
non-naturally with respect to C.

Example 165. Assume that R = Z, and S = Z/nZ. If C is the complex

×n
Z −−→ Z concentrated in homological degrees 0 and 1 then H1 (C) = 0 and
H0 (C) = Z/nZ. The universal coefficient theorem gives isomorphisms:

Z/nZ ' Z/nZ ⊗Z Z/nZ ' H0 (C ⊗Z Z/nZ) ,

H1 (C ⊗Z Z/nZ) ' TorZ1 (Z/nZ, Z/nZ) ' Z/nZ .

Yhis agrees with the direct computation of the homology of C⊗Z Z/nZ which
0
is nothing but the complex Z/nZ → − Z/nZ concentrated in homological
degrees 0 and 1.

The universal coefficient theorem has the following application for sin-
gular homology, which shows that homology with integral coefficients allows
to compute the homology with any coefficient ring S.

Corollary 166. Let X be a topological space. For all rings S, there is a

split exact sequence:

0 → Hi (X, Z) ⊗Z S → Hi (X, S) → TorZ1 (Hi−1 (X, Z), S) → 0 .

There is an analogous statement involving HomR instead of ⊗R .

Theorem 167. Assume that R is a PID, and let R → S be a morphism of

rings. For all complexes C of free R-modules, we have short exact sequences:

0 → Ext1R (Hi−1 (C), S) → H i (HomR (C, S)) → HomR (Hi (C), S) → 0 .

This short exact sequence is natural with respect to C. It splits, non-

naturally with respect to C.

Corollary 168. Let X be a topological space. For all rings S, there is a

split short exact sequence:

0 → Ext1R (Hi−1 (X, Z), S) → H i (X, S) → HomR (Hi (X, Z), S) → 0 .

53
 A. Touzé – 2023

Part III
Homology of groups
12 Basic notions on representations
12.1 Group algebras and representations
Definition 169. Let G be a group and let k be a commutative ring.
The group algebra kG is the free k-module with basis (eg )g∈G . Thus an
element of kG is a sum g∈G λg eg where only a finite number of λg are
P

nonzero. The product of two elements of kG is defined by:

  

λ0h eh  = λg λ0h egh .

X X X
 λg eg  
g∈G h∈G (g,h)∈G2

The category kG−Mod is called the category of k-linear representations of

G.

Proposition 170. A left kG-module is the same as a k-module M together

with a left action of G on M by k-linear maps. A morphism of left kG
modules is the same as a k-linear map f : M → M 0 which is G-equivariant,
that is, f (gm) = gf (m) for all m ∈ M and all g ∈ G.

Examples 171. 1. The (left) regular representation of G : kG.

The (left) regular representation is the k-module kG, with G-action:
 

λg eg  =
X X
h· λg ehg .
g∈G g∈G

Lemma 172. The (left) regular representation kG is a projective gener-

ator of kG−Mod.

2. The trivial representations.

A trivial representation is a k-module M , together with the action given
by gm = m for all g ∈ G and all m ∈ M .
A k-module M with trivial action may be denoted simply by M or by
M triv if we want to emphasize that the action is trivial.

3. The dual representation: V ] .

The dual of a representation V is the k-module V ] = Homk (V, k) with
G-action given by:
(gf )(v) := f (g −1 v) .

54
12.2 Tensor product and Hom A. Touzé – 2023

4. The representation Map(X, M ).

If X is a set with left action of G and if W is a k-module, then we can
endow the k-module Map(X, W ) of maps from X to M with an action
of G by:
(gf )(x) = f (g −1 x) .

Lemma 173. Consider G as a set with G-action given by left transla-

tions. Let J be an injective cogenerator of k−Mod. Then Map(G, J triv )
is an injective cogenerator of kG−Mod.

Remark 174. Without further notification, our representations will be left

representations, i.e. left kG-modules. Sometimes we will also need right rep-
resentations, i.e. right kG-modules. Observe that every left representation
V gives rise to a right representation with the same underlying k-module,
and with right action given by vg := g −1 v. A k-linear morphism f : V → W
is equivariant for the left action of G if and only if it is equivariant for the
right action of G.

12.2 Tensor product and Hom

Definition 175. Let V and W be two k-linear representations of G. The
k-modules V ⊗k W and Homk (V, W ) are representations of G, with G-action:

g(v ⊗ w) := (gv) ⊗ (gw) , (gf )(v) := gf (g −1 v) .

Example 176. Consider the k-module Mn (k) of n × n-matrices as a

GLn (k) representation, with GLn (k) acting by conjugation. Then Mn (k) '
Homk (kn , kn ) where kn is the standard representation of GLn (k), i.e.
gv := g(v).

Proposition 177. There is a canonical morphism of representations, which

is an isomorphism if V or W are free of finite rank as k-modules :

V ] ⊗k W → Homk (V, W ) .
f ⊗w 7 → (v 7→ f (v)w)

Proposition 178. For all representations V there is an isomorphism of

representations:
V ⊗k kG ' V triv ⊗k kG

Corollary 179. If V is projective as a k-module and P is a projective

representation of G, then V ⊗k P is a projective representation. In particular,
the tensor product of projective representations is projective.

55
12.3 Invariants and coinvariants A. Touzé – 2023

12.3 Invariants and coinvariants

Definition 180. Let V be a k-linear representation of G. The k-module of
invariants is the submodule V G ⊂ V defined by:
V G = {v ∈ V | gv = v ∀g ∈ G} .
The k-module of coinvariants is the quotient module VG defined by:
V
VG = .
hgm − mi
Example 181. Let k be a commutative ring and let Mn (k) denote the
k-module of n × n-matrices with conjugation action of GLn (k). Then
Mn (k)GLn (k) is a free k-module of rank one, generated by the identity ma-
trix, and (Mn (k)] )GLn (k) is a free k-module of rank one, generated by the
trace map.
Proposition 182. We have equalities:
Homk (V, W )G = HomkG (V, W ) , (V ⊗k W )G = V ⊗kG W ,
where V denotes the right representation associated to the left representa-
tion V (i.e. the right representation with the same underlying k-module,
and with right action defined by vg := g −1 v). As a consequence, we have
isomorphisms, natural with respect to V :
V G ' HomkG (ktriv , V ) , VG ' ktriv ⊗kG V .

12.4 Induction and coinduction

Let H be a subgroup of G. We have a restriction functor:
resG
H : kG−Mod → kH−Mod

which sends a representation of G to the representation of H obtained by

restricting the action to H. There are two functors going in the other direc-
tion, namely the induction and the coninduction:
indG
H = kG ⊗kH − : kH−Mod → kG−Mod

H = HomkH (k[G], −) : kH−Mod → kG−Mod

coindG
Proposition 183. Let H be a subgroup of G. Induction is exact, preserves
projectives, and satisfies the following adjunction isomorphism, natural with
respect to V and W :
HomkG (indG
H V, W ) ' HomkH (V, resH W ) .
G

Coinduction is exact, preserves injectives, and satisfies the following adjunc-

tion isomorphism, natural with respect to V and W :
HomkH (resG
H V, W ) ' HomkG (V, coindH W ) .
G

56
12.4 Induction and coinduction A. Touzé – 2023

Proposition 184. If H has finite index in G, there is an isomorphism

natural with respect to V :

indG
H V ' coindH V .
G

57
13. Homology of groups A. Touzé – 2023

13 Homology of groups
13.1 Homology of groups and Tor
Definition 185. Let G be a group, and let V be a (left) k-linear representa-
tion of G. Far all nonnegative degree i, the homology of G with coefficients
in V is defined by:
Hi (G, V ) = TorkG
i (k
triv
,V ) .
The following properties are simply properties of Tor that we restate in
our context.
1. There is a natural isomorphism: H0 (G, V ) ' VG .
2. Homology is additive: Hi (G, α Vα ) = α Hi (G, Vα )
L L

3. Hi (G, kG) = 0 if i > 0,

4. Every short exact sequence of representations 0 → V 0 → V → V 00 → 0
induces a short exact sequence in homology:
∂
· · · → Hi (G, V ) → Hi (G, V 00 ) −
→ Hi−1 (G, V 0 ) → Hi−1 (G, V ) → . . . .

Definition 185 shows homology of a group as a special case of Tor. Con-

versely Tor can be presented as a special case of homology.
Proposition 186. Let G be a group, let V and W be two left representations
of G and let V denote the right representation associated to V . Assume that
V is flat as a k-module. Then for all i, there is an isomorphism, natural
with respect to V and W :

Hi (G, V ⊗k W ) ' TorkG

i (V , W ) .

Example 187 (groups with cardinal invertible in k). Let G be a finite

group with cardinal |G| invertible in k. Then ktriv is a direct summand of
kG. Consequently:
VG if i = 0 ,
(
Hi (G, V ) =
0 if i > 0
Example 188 (infinite cyclic groups). Let C denote the infinite cyclic
group, with generator g. There is a projective resolution:
1−g 
0 → kC −−→ kC →
− k

In particular for all representations V we have


V G

 if i = 0 ,
Hi (G, V ) = V G if i = 1 ,

0

if i > 1 .

58
13.2 The standard complex A. Touzé – 2023

Example 189 (finite cyclic groups). Let Cn denote the finite cyclic group
of order n, with generator g. There is a projective resolution:
1+g+···+g n−1 1−g 1+g+···+g n−1 1−g 
. . . −−−−−−−−−→ kCn −−→ kCn −−−−−−−−−→ kCn −−→ kCn →
− k
In particular for all representations V we have

VG

 if i = 0 ,
Hi (G, V ) = Ker(N : VG → V G)
if i is even and positive ,

Coker(N : V → V G ) if i is odd

G

where N denotes the norm map, which sends [v] to g∈G gv. For example,
P

if V is a trivial representation, then VG = V G = V and N is multiplication

by |G|. Consequently:

V

 if i = 0 ,
Hi (G, V ) = V /|G|V if i is even and positive ,

if i is odd .

|G| V


13.2 The standard complex

Definition 190. Let R be a k-algebra, let M be a right R-module and let
N be a left R-module. The bar construction B(M, R, N ) is the complex of
k-modules whose term of degree k is the k-module Bk (M, R, N ) defined by:

0
 if k < 0
Bk (M, R, N ) = M ⊗k R ⊗k · · · ⊗k R ⊗k N if k ≥ 0

 | {z }
k terms

and with differential given by the formula:

k
d(m ⊗ r1 ⊗ · · · ⊗ rk ⊗ n) = (−1)k dk (m ⊗ r1 ⊗ · · · ⊗ rk ⊗ n)
X

i=0

where

mr1 ⊗ r2 ⊗ · · · ⊗ rk ⊗ n

 if i = 0,
di (m⊗r1 ⊗· · ·⊗rk ⊗n) = m ⊗ r ⊗ ··· ⊗ r r
1 i i+1 ⊗ · · · ⊗ rk ⊗ n if 0 < i < k,

r ⊗ · · · ⊗ r ⊗ if i = k .

1 k−1 rk n
Remark 191. To check that B(M, R, N ) is indeed a complex, one checks
that for i < j we have di dj = dj−1 di . Then the proof is similar to the proof
that the singular chain complex of a topological space is indeed a chain
complex. There is actually a notion hidden behind this similarity, namely
the notion of a simplicial object. You may have a look e.g. at [Weibel, Chap
8] for more details relative to simplicial objects. The closely related notion
of a semi-simplicial set is explained in section 14.4.

59
13.3 Functoriality with respect to G A. Touzé – 2023

Proposition 192. Assume that R and M are projective as k-modules. Then

for all R-modules M and N and all nonnegative integers i, we have

TorR
i (M, N ) = Hi (B(M, R, N )) .

Corollary 193. For all k-linear representations V , Hi (G, V ) is the homol-

ogy of the standard complex B(k, kG, V ).

Corollary 194. For all groups G we have H1 (G, Z) = Gab .

13.3 Functoriality with respect to G

Let V be a left representation of G and let U be a right representation of
G. Every morphism of groups α : H → G induces a surjective morphism
α0 : U ⊗kH V → U ⊗kG V such that α0 (u ⊗ v) = u ⊗ v.

Definition 195. If (P, ) is a projective resolution of k as a kH-module,

and if (Q, 0 ) is a projective resolution of k as a kG-module, then for all
representations V of G and for all morphisms α : H → G there is a morphism
of complexes
id⊗id
V 0 α
P ⊗kH V −−−−→ Q ⊗kH V −→ Q ⊗kG V
and we let αi : Hi (H, V ) → Hi (G, V ) be the induced application in homology
of degree i. (Here id : P → Q is a lift of the identity map of k to the level
of the resolutions).
α β
Lemma 196. Given morphisms of groups K − → G we have (β ◦ α)i =
→H−
βi ◦ αi for all i.

Remark 197. In the definition of αi , we have again denoted by V the

restriction to H of the representation of G. It is an abuse of notation that
we will often make. To be perfectly correct we should have used the heavy
notation resGH V , so that αi is actually a map Hi (H, resH V ) → Hi (G, V ).
G

In fact, even the heavy notation resH V is not completely precise, because
G

α does not appear whereas there may be many morphisms of groups α :

H → G. In the situations where we want to be very precise, we will denote
by α∗ V the representation of H obtained by restriction along α. Here is a
situation where we need to be precise to avoid confusion: if α : G → G is a
morphism of groups, definition 195 yields morphisms of k-modules:

αi : Hi (G, α∗ V ) → Hi (G, V ) .

Functoriality with respect to G is often used to compute the homology of

G from the homology of its subgroups (which is simpler to understand). In
the remainder of this section we explain two basic examples of this approach.

60
13.3 Functoriality with respect to G A. Touzé – 2023

13.3.1 Schapiro’s lemma

Let G be a group, let H be a subgroup, and let V be a representation
of H. There is an H-equivariant morphism, natural with respect to the
H-representation V
ηV : V → indG
H V = kG ⊗kH V
v 7 → 1⊗v
Theorem 198 (Schapiro’s lemma). Let G be a group, let H be a subgroup,
and let V be a representation of H. The following composition is an iso-
morphism for all i:
Hi (H,ηV ) incl
Hi (H, V ) −−−−−−→ Hi (H, indG
H V ) −−→ Hi (G, indH V ) .
i G

One says that a ring R has finite (left) homological dimension if there
is an integer d such that every left R-module has a projective resolution of
the form

0 → Pd → Pd−1 → · · · → P0 → − M →0.
If such a d does not exist, then R has infinite (left) homological dimension.
The following result is a direct consequence of Schapiro’s lemma.
Corollary 199. Let k be a field of characteristic p > 0. If G has an element
of order p, then kG has infinite homological dimension.

13.3.2 The Cartan-Eilenberg formula for normal subgroups

Let H be a normal subgroup of G. Then for all g ∈ G, conjugation by g
defines an isomorphism of groups:
cg : H → H
.
h 7→ cg (h) = ghg −1
Now if V is a representation of G, then for all g ∈ G, the action of g on V
defines an equivariant isomorphism:
g· : V → c∗g V
.
v 7→ gv
For all integers i, we denote by g? : Hi (H, V ) → Hi (H, V ) the isomorphism
of k-modules given by the composition:
Hi (H,g·) (cg )i
g? : Hi (H, V ) −−−−−→ Hi (H, c∗g V ) −−−→ Hi (H, V ) .
' '
Proposition 200. The map x 7→ g ? x defines a k-linear action of G on
Hi (H, V ). Moreover the elements h ∈ H act trivially through this action, so
that the action factors as an action of the quotient group G/H on Hi (H, V ).
incl
Moreover, the map Hi (H, V ) −−→ i
Hi (G, V ) passes to the quotient and in-
duces a map
incli : Hi (H, V )G/H → Hi (G, V ) .

61
13.3 Functoriality with respect to G A. Touzé – 2023

Proof. Let (P, ) be a projective resolution of k as a (right) representation of

H. Then by its definition g? is induced by the morphism of chain complexes
id⊗(g·) (cg )0
P ⊗kH V −−−−→ c∗g P ⊗kH c∗g V −−−→ P ⊗kH V , (∗)
=

where id : P → c∗g P is a lift of the identity of k to the level of the resolutions.

If we choose P to be the restriction to H of a projective resolution of k as
a (right) representation of G, then we may take id to be given by the right
action of g −1 , i.e. id(x) = xg −1 . With this choice, the morphism g? is
simply given on the level of chain complexes by the map
P ⊗kH V → P ⊗kH V
.
x⊗v 7→ xg −1 ⊗ gv
with this expression the statement of the proposition is easy to prove.

Theorem 201 (Cartan-Eilenberg formula). Assume that V is a represen-

tation of G, and that H is a normal subgroup of G such that |G : H| is
invertible in k. Then the morphism

incli : Hi (H, V )G/H → Hi (G, V ) .

is an isomorphism of k-modules.
The proof of the Cartan-Eilenberg formula relies on the following lemma
(which is a more precise version of proposition 78).
Lemma 202. Let K be a group, let C be a complex of kK-modules. Then
CK is a quotient complex of C considered as a complex of k-modules, and
the morphism Hi (π) : Hi (C) → Hi (CK ) induces a comparison morphism

Hi (C)K → Hi (CK ) .

If |K| is invertible in k, this comparison morphism is an isomorphism.

Example 203. Let us consider the symmetric group S3 on three letters,
with normal subgroup C3 and quotient S3 /C3 ' C2 . The Cartan-Eilenberg
formula tells us that

Hi (S3 , F3 ) ' Hi (C3 , F3 )C2 .

We know that Hi (C3 , F3 ) ' F3 , so in order to compute the homology of S3 ,
we only have to determine the action of C2 , hence of a transposition τ ∈ S3 .
For this purpose, we use the expression (∗) of the action of τ , together with
the resolution of example 189. We have a diagram:
P = k  kC3 1−c
kC3 N
kC3 1−c
kC3 N
kC3
= = −c −c c 2
c2 .
c∗τ P = k  kC3 kC3 kC3 kC3 kC3
1−c2 N 1−c2 N

62
13.3 Functoriality with respect to G A. Touzé – 2023

Hence we see that τ acts as (−1)i on H2i−1 (C3 , k) and H2i (C3 , k). Therefore
we obtain that
i= 0 1 2 3 4 5 6 7 8 9 ...
Hi (S3 , F3 ) = F3 0 0 F3 F3 0 0 F3 F3 0 . . .

Remark 204. The Cartan-Eilenberg formula has a generalization when the

subgroup H is not normal in G. The proof of this generalized statement is
not much harder, but this generalized statement is a bit more complicated
to state. Indeed, the conjugation map cg : G → G now takes H to the
conjugate subgroup gHg −1 . Hence g ∈ G does not act on Hi (H, V ) anymore,
but induces an isomorphism between Hi (H, V ) and Hi (gHg −1 , V ). So one
has to generalize the meaning of coinvariants. With this new meaning, the
Cartan-Eilenberg formula stays valid, i.e. there is an isomorphism

Hi (H, V )G/H ' Hi (G, V ) .

63
14. Homology of groups and topology A. Touzé – 2023

14 Homology of groups and topology

14.1 Recollections of covering spaces
Definition 205. A covering map is a continuous map p : E → X such that
every x ∈ X is contained in an open set U such that p−1 U is a disjoint union
of open sets Ui and the restriction maps p|Ui : Ui → U are homeomorphisms.
The space E is called a covering space of X, but the term covering space
often also refers to the map p.

Examples 206. 1. The protopype of a covering map is the map R → S 1 ,

t 7→ ei2πt .

2. Let Y be a topological space endowed with a free action by homeo-

morphisms of a group G. Assume that every y ∈ Y is contained in an
open set U such that gU ∩ U 6= ∅ for all g ∈ G. Then the quotient
map
q : Y → Y /G
is a covering map. (Covering maps constructed in this way are called
Galois covering maps)

One of the most important theorem of covering theory if the following

lifting theorem. Given a continuous map f : Y → X and a covering p :
E → X, a lifting of f is a continuous map f such that the following diagram
commutes
E
f
p .
f
Y X
Recall also that a space Y is locally arcwise connected if for all y ∈ Y and all
open subset U containing y, one can find an arcwise connected neighborhood
V of y which is contained in U . For a covering map p : E → X the base
space X is locally arcwise connected if and only if the total space E is locally
arcwise connected.

Theorem 207. Let p : E → X be a covering map, let f : Y → X be a

continuous map. Assume that Y is arcwise connected, and locally arcwise
connected. Let us fix basepoints e, x, y such that p(e) = x and f (y) = x.
Then f admits a lifting f such that f (y) = e if and only if the image of
π1 (f ) : π1 (Y, y) → π1 (X, x) is contained in the image of π1 (p) : π1 (E, e) →
π1 (X, x). Moreover, if the lifting f exists, then it is unique.

Definition 208. A covering p : Xe → X is called universal if X

e is locally
arcwise connected and simply connected.

64
14.2 K(G, 1) and group homology A. Touzé – 2023

If p : (X,
e xe) → (X, x) is a universal covering, it follows from the lifting
theorem 207 that it satisfies the following universal property. For all cover-
ings q : (E, e) → (X, x) there is a unique continuous map p0 : (X, e) → (E, e)
e x
making the following triangle commute:

p0
(X,
e xe) (E, e)
p q

(X, x)

One can deduce from the universal property that a universal covering is
unique up to homeomorphism. One also deduces from the universal property
the following structure result.

Proposition 209. If p : X e → X is a universal covering, then it has the

form of example 206 2), and G ' π1 (X, x).

14.2 K(G, 1) and group homology

Definition 210. Let G be a group. A topological space X is called a
K(G, 1) space if (i) it is arcwise connected and locally arcwise connected,
(ii) π1 (X, x) ' G and (iii) X admits a universal covering p : X
e → X such
that Xe is contractible.

Remark 211. The "1" in the name "K(G, 1)" refers to the fact that π1 (X, x)
is involved in the definition. See section 15 for more details.

Examples 212. 1. The circle S 1 is a K(Z, 1). Indeed the universal cov-
ering is R → S 1 , t 7→ e2iπt and R is contractible.

2. If n 6= 1, the sphere S n is not a K(G, 1).

3. The wedge of two circles S 1 ∨ S 1 is a K(F2 , 1). Indeed, the universal

covering is p : C → S 1 ∨ S 1 where C is the Cayley graph of F2 , which
is contractible (it is a tree).
Here is a picture (taken from wikipedia) of the Cayley graph of F2 . It
is the topological graph whose vertices are labelled by the elements of
F2 , and which has one edge between vertex x and vertex y if there is
` ∈ {a, b, a−1 , b−1 } such that y = x`.

65
14.2 K(G, 1) and group homology A. Touzé – 2023

4. The torus T 2 ' S 1 × S 1 is a K(Z2 , 1). Indeed the universal covering

is R2 → S 1 × S 1 , (s, t) 7→ (e2iπs , e2iπt ) and R2 is contractible.

5. More generally the oriented compact surface Sg of positive genus g

is a K(G, 1) with G = F2g /[a1 , b1 ] · · · · · [ag , bg ]. Indeed, its universal
covering space is the Poincaré disk D2 :

6. Let f : S 1 → S 3 be a knot (i.e. a smooth embedding). Then S 3 \ Im f

is a K(G, 1). (This is a nontrivial theorem of 3-manifold theory).

Theorem 213. Let X be a K(G, 1). Then C sing (X, e k) is a projective reso-
lution of ktriv as a kG-module, and there is an isomorphism of k-modules:

Hising (X, k) ' Hi (G, k) .

66
14.3 Projective resolutions of ktriv via G-complexes A. Touzé – 2023

Comment 214 (Singular homology with local coefficients). Assume that

X is an arcwise connected space, with π1 (X, x) ' G and which admits a
universal covering. Then X is maybe not a K(G, 1). However, C sing (X, e k)
is still a complex of projective kG-modules, and for all left representations
V of G, one may consider the complex of k-modules
C sing (X,
e k) ⊗kG V .

The homology of this complex is called the singular homology of X with local
coefficients V , and denoted by Hising (X, V ). This generalization of singular
homology appears in several places of algebraic topology.
The previous theorem extends to the setting of singular homology with
local coefficients. Namely, if X is a K(G, 1) then for all degrees i there is
an isomorphism
Hising (X, V ) ' Hi (G, V ) .

14.3 Projective resolutions of ktriv via G-complexes

Definition 215. A free G-complex is a CW-complex Y , equipped with a
free action of the group G which freely permutes the open cells of Y .
Theorem 216. If Y is a free G-complex, then C cell (Y ) is a complex of
projective kG-modules. Moreover, assume that Y is arcwise connected, and
that there is an n ∈]0, +∞] such that Hi (Y, k) = 0 for 0 < i < n. Then
C cell (Y ) is a projective resolution of ktriv up to degree n − 1, i.e. we have
an exact complex:

Cncell (Y, k) → · · · → C1cell (Y, k) → C0cell (Y, k) →
− ktriv → 0,
and in addition, for all i < n we have
Hi (Y /G) ' Hi (G, k) .
Examples 217. 1. Y = S n with 2 cells in each dimension. This is a free
G-complex, for G = C2 with antipodal action. The associated cellular
complex coincides with the projective resolution given in example 189
up to degree n − 1.
2. Y = R with action of C by translation. This is a free G-complex. The
associated cellular complex coincides with the projective resolution
given in example 188.
Corollary 218. Let X be an arcwise connected CW-complex of dimension
n with contractible universal covering. Then π1 (X, ∗) contains no element
of finite order. In particular, the group π1 (X, ∗) is either trivial or infinite.
(This corollary applies in particular to arcwise connected manifolds of
dimension n. It shows that all the nice examples of K(G, 1) coming from
geometry have infinite groups G in a strong sense.)

67
14.4 Semi-simplicial sets A. Touzé – 2023

14.4 Semi-simplicial sets

In this section, we briefly introduce semi-simplicial sets 3 , which we will use
later to construct K(G, 1) spaces. Semi-simplicial sets are a generalization
of geometric simplicial complexes. They can be thought of as combinatorial
instructions to build CW-complexes by pasting standard simplices together.
Definition 219. A semi-simplicial set S is a family of sets (Sn )n≥0 equiped
with maps di : Sn → Sn−1 , 0 ≤ i ≤ n such that:

di ◦ dj = dj−1 ◦ di if i < j.

A morphism of semi-simplicial sets f : S → T is a collection of maps

fn : Sn → Tn such that fn−1 ◦ di = di ◦ fn for all i and all n.
Example 220. 1. If K is a geometric simplicial complex (see definition
62) gives rise to a semi-simplicial set ∆(K) such that ∆(K)n is the set
of n-simplices of K and di : ∆(K)n → ∆(K)n−1 sends every n-simplex
onto its i-th face.

2. There is a semi-simplicial set such that Si = ∅ if i > 1, S1 and S0 have

exactly one element and the d0 and d1 are equal to the unique map
S1 → S0 .
Definition 221. The realization of a semi-simplicial set S is the topological
space
n≥0 Sn × ∆
F n
|S| =
(di s, t) ∼ (s, i t)
where Sn is considered as a discrete set (i.e. Sn × ∆n is homeomorphic to
the disjoint union of copies of the standard n-simplex ∆n indexed by the
elements of Sn ) and where i : ∆n−1 → ∆n is the map which identifies the
standard (n − 1)-simplex with the i-th face of the standard n-simplex (see
definition 40).
If f : S → T is a morphism of semi-simplicial sets, the continuous map

fn × id∆n : Sn × ∆n → Tn × ∆n
G G G

n≥0 n≥0 n≥0

induces a continuous map |f | : |S| → |T | on the quotients, which is called

the realization of f .
Example 222. 1. Let K be a geometric simplicial complex. Then there
is a homeomorphism:

|∆(K)| ' |K| =

[
σ.
σ∈K
3
The terminology and notation varies a lot in the literature. Semi-simplicial sets are
also called ∆-sets, ∆+ -sets, ∆inj -sets...

68
14.5 Construction of K(G, 1)-spaces A. Touzé – 2023

2. If S is the semi-simplicial set such that Si = ∅ for i > 1 and S1 and

S0 have one element, there is a homeomorphism |S| ' S 1 .

The realization of a semi-simplicial set produces a CW-complex. To

prove this, we need to introduce characteristic maps.

Definition 223. For all s ∈ Sn , we define the characteristic map fs : ∆n →

|S| be the continuous map given by the composition
t7→(s,t)
fs := ∆n −−−−→ Sn × ∆n  |S| .
G

n≥0

By definition of the disjoint union topology and of the quotient topology,

a subset U ⊂ |S| is open if and only if for all s, fs−1 (U ) is opened in ∆n .
Also, the equivalence relation defining |S| only involves the boundary of ∆n .
Hence the restriction of fs to ∆n \ ∂∆n is injective.

Proposition 224. Let S be a semi-simplicial set and let S≤n denote the
semi-simplicial set with
(
Si if i ≤ n,
(S≤n )i =
∅ if > n

Then |S| has a CW-complex structure with n-th skeleton |S≤n |, and with
characteristic maps fs , i.e. for all positive n we have a pushout square:

tfs|∂∆n
ts∈Sn ∂∆n |S≤n−1 |
.
tfs
ts∈Sn ∆n |S≤n |

14.5 Construction of K(G, 1)-spaces

Definition 225. For all groups G, we can define a semi-simplicial set EG
with EGn = Gn+1 and di : EGn → EGn−1 is defined by

di (g0 , g1 , . . . , gn ) = (g0 , . . . , gi−1 , gi+1 , . . . gn ) .

Every g ∈ G defines a morphism of simplicial sets EG → EG which sends

the n-simplex (g0 , g1 , . . . , gn ) to (gg0 , gg1 , . . . , ggn ).

Theorem 226. The topological space |EG| is a contractible free G-CW-

complex. In particular, |EG|/G is a K(G, 1) space.

69
15. Vista: homotopy groups and EML spaces A. Touzé – 2023

15 Vista: homotopy groups and EML spaces

We defined K(G, 1)-spaces as spaces with fundamental group G and con-
tractible universal cover. In this section, we give an equivalent definition of
K(G, 1)-spaces, relying on the notion of homotopy groups.

15.1 Homotopy groups

Given two pairs of topological spaces (X, A) and (Y, B), we denote by
Map(X, A; Y, B) the set of maps f : X → Y such that f (A) ⊂ B. We
define an equivalence relation ∼ on this set by f ∼ g if and only if there is
a homotopy H : X × [0, 1] → Y between f and g such that H(A, t) ⊂ B for
all t ∈ [0, 1]. We use the following notation for the quotient set:

Map(X, A; Y, B)
[(X, A), (Y, B)] := .
∼
The equivalence class of a map f is denoted by [f ]. If A = B = ∅ then we
recover the homotopy classes of maps from X to Y .

Definition 227. Let (X, ∗) be a pointed topological space. For all n ≥ 1, let
C n = [0, 1]n denote the n-th dimensional cube. The n-th homotopy group
πn (X, ∗) is the set:

πn (X, ∗) = [ (C n , ∂C n ) , (X, ∗) ]

equipped with the multiplication [f ] · [g] = [f g], where f g : C n → Y is the

concatenation of f and g along the first coordinate:

f (2t1 , t2 , . . . , tn ) if t ≤ 12 ,
(
(f g)(t1 , . . . , tn ) =
g(2t1 − 1, t2 , . . . , tn ) if t ≥ 12 .

70
15.1 Homotopy groups A. Touzé – 2023

One should make small verifications to ensure definition 227 is consitent,

that is, the group operation is well-defined, associative, with a unit. The
unit element is the class of the constant map C n → X with value ∗.
If n = 1, definition 227 is the usual definition of the fundamental group.
Hence for n ≥ 2, the homotopy groups can be thought of as higher di-
mensional analogues of the fundamental group. The basic properties of the
fundamental group generalize to higher homotopy groups:

1. If X is arcwise connected, the group πn (X, ∗) does not depend on the

basepoint: different basepoints yield isomorphic groups.

2. Every pointed map φ : (X, ∗) → (Y, ∗) induces a group morphism

πn (φ) : πn (X, ∗) → πn (Y, ∗) such that φ([f ]) = [φ ◦ f ].

3. If φ : X → Y is a homotopy equivalence, then for all ∗ ∈ X the map

πn (φ) : πn (X, ∗) → πn (Y, φ(∗)) is an isomorphism of groups.

A small surprise with higher homotopy groups is the following additional

basic fact, which is of course not true if n = 1.

Proposition 228. For all n ≥ 2 and for all pointed spaces (X, ∗), the group
πn (X, ∗) is abelian.

In view of proposition 228, one may think that higher homotopy groups
are nicer and easier to compute than the fundamental group. But this is not
the case. The reason is that the powerful Van Kampen theorem, which is
used to compute the fundamental groups of many spaces, has no equivalent
in higher dimensions. As an illustration of the difficulty of computing homo-
topy groups, at the current time we don’t know all the homotopy groups of
the spheres S d , d ≥ 2, although we computations in low dimension n, as well
as good qualitative information in all dimensions. We refer the reader e.g.
to chap 0 of Neisendorfer’s book "Algebraic methods in unstable homotopy
theory" for more details on this subject.
To finish our quick tour of homotopy groups, let us mention two funda-
mental results relating homotopy groups and singular homology. Firstly, for
all n ≥ 1, the integral homology of the pair (C n , ∂C n ) in degree n is equal
to Z. Choose a generator c of this homology group. For all pointed spaces
(X, ∗), the Hurewicz map is the map:

hn : πn (X, ∗) → Hn (X, ∗; Z) = Hn (X; Z)

such that hn ([f ]) = Hn (f )(c). One checks that this map is a group mor-
phism, natural with respect to X.

Theorem 229 (Hurewicz). Assume that (X, ∗) is a pointed, arcwise con-

nected space.

71
15.2 K(G, 1)-spaces and EML spaces A. Touzé – 2023

1. The map h1 vanishes on the commutator subgroup of π1 (X, ∗), and

induces an isomorphism of abelian groups:
π1 (X, ∗)ab ' H1 (X, Z) .

2. If n ≥ 2 and if πi (X, ∗) = 0 for i < n then hi is an isomorphism in

degrees i ≤ n.
If X is simply connected, the Hurewicz theorem is often used to compute
the first nontrivial homotopy group since the homology groups of X are
usually easier to compute than its homotopy groups.
Example 230. We have πn (S d ) = 0 if 0 < n < d and πd (S d ) = Z.
Another fundamental theorem is the following algebraic criterion to rec-
ognize homotopy equivalences. In part I of the course, we have introduced
homology groups as a tool to show that two spaces have different homotopy
types. The Whitehead theorem allows us to use homology groups to prove
that two spaces have the same homotopy type.
Theorem 231 (Whitehead). Let φ : X → Y be a map between two arcwise
connected CW-complexes, and let x be a basepoint in X. The following
statements are equivalent.
1. For all n ≥ 1, πn (φ) : πn (X, x) → πn (Y, φ(x)) is an isomorphism.
2. The map π1 (φ) : π1 (X, x) → π1 (Y, φ(x)) is an isomorphism, and for
all n ≥ 1 the map Hn (φ) : Hn (X; Z) → Hn (Y ; Z) is an isomorphism.
3. The map f : X → Y is a homotopy equivalence.
Observe that the Whitehead theorem does not say that two spaces with
the same homotopy or homology groups are homotopy equivalent: one really
needs a map between the two spaces. However, there is one situation where
such a map always exists, namely when we want to compare a space X with
the point ∗. In that case, the Whitehead theorem applied to the unique map
φ : X → ∗ gives the following result.
Corollary 232. Let (X, ∗) be a pointed arcwise connected CW-complex.
Then: πn (X, ∗) is trivial for all n ≥ 1 ⇔ π1 (X, ∗) and the Hn (X; Z) for all
n ≥ 1 are trivial ⇔ X is contractible.

15.2 K(G, 1)-spaces and EML spaces

We are now going to characterize K(G, 1)-spaces with homotopy groups.
The characterization relies on the following lemma, which is an easy exercise4
on the lifting properties of coverings.
4
For this exercise, you may use that πn (X, ∗) is isomorphic to the set [ (S n , ∗) , (X, ∗) ]
of pointed homotopy classes of maps S n → X, and that π1 (S n ) is trivial if n ≥ 2.

72
15.2 K(G, 1)-spaces and EML spaces A. Touzé – 2023

Lemma 233. Let p : E → B be a covering map, and let ∗ ∈ E. Then

πn (p) : πn (E, ∗) → πn (B, ∗) is an isomorphism of groups for all n ≥ 2.

Let X be a K(G, 1)-space with contractible universal cover X.

e Then all
the homotopy groups of X are trivial. By applying the previous lemma to
e
the covering map p : X
e → X we obtain

if n = 1,
(
G
πn (X, ∗) =
0 if n > 1.

Conversely, assume that X is an arcwise connected space with trivial higher

homotopy groups and with fundamental group equal to G. If X is a CW-
complex then its universal cover X
e is a CW-complex, and by applying lemma
233 and corollary 232, we obtain that X e is contractible, hence that X is a
K(G, 1) space. The next result summarizes this discussion.

Proposition 234. Let X be an arcwise connected CW -complex. Then X is

a K(G, 1)-space if and only if π1 (X, ∗) ' G and πn (X, ∗) = 0 for all n ≥ 2.

Remark 235. The circle S 1 is a K(Z, 1), hence it has fundamental group
isomorphic to Z and trivial higher homotopy groups. Among the spheres
S d , d ≥ 1, the circle is the only one whose homotopy groups are completely
known.

The characterization given in proposition 234 justifies the notation

K(G, 1) and lead to the following generalization of the concept.

Definition 236. Let n ≥ 1 and let G be a group, which is assumed to

be abelian if n ≥ 2. A CW-complex X is called a K(G, n)-space if it is
arcwise connected and if its homotopy groups πk (X, ∗) are trivial if k 6= n
and πn (X, ∗) ' G.

The K(G, n)-spaces were first studied by Eilenberg and MacLane, and
for this reason, they are often refered to as Eilenberg-MacLane spaces or
EML spaces for short.
In classical homotopy theory, there is a kind of symmetry between ho-
motopy groups and homology groups. On the one hand, spheres are the
topological spaces with the simplest homology groups, but their homotopy
groups are very complicated. On the other hand, Eilenberg-MacLane spaces
are the topological spaces with the simplest homotopy groups, but their have
complicated homology groups5
5
The symmetry is not perfect though: whereas the homotopy groups of spheres are still
mysterious, the homology groups of Eilenberg-Mac Lane spaces K(G, n) for G abelian,
were completely computed by Serre and Cartan in the 1950es, see http://www.numdam.
org/volume/SHC_1954-1955__7_1/.

73
15.2 K(G, 1)-spaces and EML spaces A. Touzé – 2023

This symmetry can be pushed further. Spheres play a fundamental role

in the study of topological spaces. Indeed, every CW-complex can be con-
structed as an increasing union of spaces, its skelettons Xn , n ≥ 0, and the
difference between the n-th skeleton and the (n − 1)-st skeleton behaves like
a wedge of spheres of dimension n:
_
Xn /Xn−1 ' Sn .
n-cells of X

This construction behaves particularly well with homology groups, in fact it

is at the heart of the definition of the cellular chain complex, which computes
the homology of X. But it is not adapted to the computation of homotopy
groups, let alone because the homotopy groups of spheres are not known.
Likewise, Eilenberg-MacLane spaces play a fundamental role in the study
of topological spaces. Indeed, every arcwise connected CW complex X can
be decomposed as a tower of spaces

X0  X1  X2  X3  · · ·

where the maps in the tower are fibrations (in particular surjective maps)
and the difference between X n−1 and X n behaves like a K(G, n)-space. This
is called the Postnikov decomposition of X, and we refer the reader to chap
4.3 of Hatcher’s book for further details on this. This decomposition is
useful to compute homotopy groups, but it is not well-adapted to compute
homology groups in general.

74
16. Vista: the cohomology of groups A. Touzé – 2023

16 Vista: the cohomology of groups

So far, we have introduced the homology of groups, studied its basic proper-
ties and its relations with singular homology. In this section we give a quick
overview of the dual theory, namely of group cohomology.

Definition 237. Let M be a k-linear representation of G. For all n ≥ 0,

the degree n cohomology of G with coefficients in M is defined by

H n (G, M ) = ExtnkG (k, M ) ,

where k is equipped with the trivial action of G.

It follows directly from the definition that cohomology of degree zero is

isomorphic to invariants:

H 0 (G, M ) = HomkG (k, M ) ' M G .

In higher degrees, one may try to compute H n (G, M ) by taking a projective

resolution P of the trivial kG-module, and by computing the homology of
the complex of k-modules HomkG (P, M ). If X is a K(G, 1), the singular
chain complex C∗sing (X;
e k) is a projective resolution of k by theorem 213.
If in addition M = k is the trivial module, then we have isomorphisms of
cochain complexes

HomkG (C∗sing (X;

e k), k) ' Homk (C sing (X;
∗
e k)G , k)

' Homk (C∗sing (X; k), k)

∗
= Csing (X; k) ,

which proves the following analogue of theorem 213.

Theorem 238. Assume that X is a K(G, 1). There is an an isomorphism

in every degree i ≥ 0:
i
Hsing (X, k) ' H i (G, k) .

All the properties of group homology that we have explained in this

course (functoriality, Schapiro’s lemma, the Cartan-Eilenberg formula. . . )
have an analogue in cohomology. But, in the same way as singular cohomol-
ogy of spaces, cohomology of groups has a very useful additional structure,
namely is has cup products. To be more specific, if M and N are two k-linear
representations of G, the cup product is a family of k-linear maps

H i (G, M ) ⊗k H j (G, N ) → H i+j (G, M ⊗k N )

a⊗b 7→ a^b

which has nice properties:

75
16. Vista: the cohomology of groups A. Touzé – 2023

• associativity: (a ^ b) ^ c = a ^ (b ^ c),

• unitality: if 1 denotes the unit of k = H 0 (G, k) and if we identify

M with λ : k ⊗k M ' M and λ0 : M ⊗k k → M are the canonical
isomorphisms then for all a ∈ H i (G, M ) we have

H i (G, λ)(a ^ 1) = a = H i (G, λ0 )(1 ^ a) .

• graded commutativity: if τ : M ⊗k N ' N ⊗k M denotes the isomor-

phism such that τ (x ⊗ y) = y ⊗ x, then for all a ∈ H i (G, M ) and all
b ∈ H j (G, N ) we have

b ^ a = (−1)ij H i+j (G, τ )(a ^ b) .

In particular we have the following statement.

Theorem 239. Let G be a group.

1. The cup product makes H ∗ (G, k) = i (G, k)

L
i≥0 H into a graded com-
mutative k-algebra.

2. Moreover, for all k-linear representations M , the cup product makes

H ∗ (G, M ) = i≥0 H i (G, M ) a module over H ∗ (G, k).
L

Products are of great help in explicit computations of group homology.

They are also involved in many qualitative results. For example, a result of
Venkov (1959), Evens (1961) and Quillen (1971) shows that if G is a finite
group and k is a field, then H ∗ (G, k) is a finitely generated algebra, and
H ∗ (G, M ) is a finitely generated module over it for all finite dimensional
representations M .
This finite generation result plays a role in a deep connection between
finite groups and algebraic geometry. Namely, since H ∗ (G, k) is graded com-
mutative and finitely generated, its subalgebra H ev (G, k) = i≥0 H 2i (G, k)
L

is commutative and finitely generated. Thus it is the coordinate algebra of

an algebraic variety V (G). By its definition, the geometry of this algebraic
variety contain significant information on the group G. For example Quillen
proved (1971) that if k has positive characteristic p, then the dimension of
V (G) equals the p-rank of G, i.e. the dimension of the maximal Fp -vector
subspace of G. One can push this correspondence between finite groups and
algebraic varieties further and associate to every k-linear representation M a
subvariety V (G)M ⊂ V (G). This is a starting point of the theory of support
varieties, which establishes a dictionnary between the geometric properties
of the varieties V (G)M and the algebraic properties of the representations
M , and which is very active nowadays.
In a different direction, let us mention a recent progress on the under-
standing of the cohomology ring H ∗ (G, k). Namely, Peter Symonds has

76
16. Vista: the cohomology of groups A. Touzé – 2023

proved (2010) that the generators of H ∗ (G, k) are in degrees less than |G|,
and the relations in degrees less than 2|G| − 1. In particular, the knowl-
edge of the truncated ring i<2|G|−1 H i (G, k) is sufficient to understand the
L

whole cohomology ring. The proof of this spectacular result is spectacular

as well. It relies on differential geometry, namely on the study of abelian
group actions on stratified manifolds. At the very end, this purely algebraic
result relies on the existence of tubular neighborhoods!

77
 A. Touzé – 2023

Part IV
Exercises

78
 A. Touzé – 2023

Exercise sheet n°1

Homotopy and categories

1. Homotopy
In the euclidean space Rn , we let S n−1 denote the unit sphere and Dn denote
the unit ball (= the n-dimensional disc).
Exercice 1. Some explicit homeomorphisms. Show that [0, 1]n is
homeomorphic to Dn . Let (Dn )+ = {x = (x1 , . . . , xn ) | xn ≥ 0 , x ∈ Dn }.
Show that Dn is homeomorphic to (Dn )+ .

Exercice 2. Alphabet. In the following shortened alphabet, we consider

the letters as topological subspaces of R2 . Classify the letters into groups
according to: 1) their homotopy type, 2) their homeomorphism type.

ABCDEFGHIJKLMNOP

Exercice 3. Half homotopy inverses. We say that f : X → Y has a left

homotopy inverse if there is g : Y → X such that g ◦ f ∼ idX . We say that
f has a right homotopy inverse if there is h : Y → X such that f ◦ h ∼ idY .

1. Show that if f has a left homotopy inverse g and a right homotopy

inverse h, then g ∼ h.

2. Show that f is a homotopy equivalence if and only if it has a left

homotopy inverse and a right homotopy inverse.

3. Show that if f is a homotopy equivalence, then the homotopy inverse

of f is unique up to homotopy (i.e. two homotopy inverses are always
homotopic).

f g
Exercice 4. The "2 out of 3" property. Let X − → Y → − Z be two
continuous maps. Prove that if any two maps among f , g and g ◦ f , are
homotopy equivalences, then so is the third map.

Exercice 5. Examples of homotopy equivalences.

1. Show that for all positive n, the inclusion S n−1 ,→ Rn \ {0} is a

homotopy equivalence.

1
 A. Touzé – 2023

2. Let X and Y be two contractible spaces. Show that every map f :

X → Y is a homotopy equivalence.
3. Let C be a contractible space. Show that for all X, the canonical
projection C × X → X is a homotopy equivalence.

Exercice 6. Convexity and contractibility. Let C be a convex subset

of Rn with the induced topology. Show that C is contractible. Find a
contractible subset of R2 which is not convex.

Exercice 7. Cones. Given a topological space X, the cone CX is defined

as the quotient6 of the cylinder X × [0, 1] in which all the points of X × {1}
are identified.
1. Show that CS n−1 is homeomorphic7 to Dn , and that CDn is homeo-
morphic to Dn+1 .
2. Show that for all topological spaces X, CX is contractible8 .

2. The language of categories

Exercice 8. Uniqueness of identity morphisms. Show that in a

category C the identity morphisms idX are uniquely determined (i.e. any
txo identity morphisms of X must be equal).

Exercice 9. Isomorphisms. An isomorphism in a category C is a mor-

phism f : X → Y such that there exists a morphism g : Y → X such that
g ◦ f = idX and f ◦ g = idY . The morphism g is called an inverse of f .
6
Recall the definition of the quotient topology. Let X be a space, let ∼ be an
equivalence relation on X, and let p : X → X/ ∼ be the quotient map. The quotient
topology on X/ ∼ is such that U is open in X/ ∼ if and only if p−1 (U ) is open in X.
Set Y = X/ ∼. Then the quotient map p : X → Y satisfies the following universal
property: for all continuous f : X → Z such that f is constant on the subsets p−1 ({y}),
y ∈ Y , there is a unique continuous map f : Y → Z such that f = f ◦ p.
In fact every continuous map p : X → Y satisfying the previous universal property
is a quotient map, and Y is then the quotient of X by the equivalence relation whose
equivalence classes are the subsets p−1 ({y}), y ∈ Y .
7
A very convenient way to contruct homeomorphisms in the compact case is the fol-
lowing lemma. Lemma: if f : X → Y is a continuous bijection, if X is compact (i.e.
every covering of X by open subsets contains a covering of X by a finite number of open
subsets.) and Y is separated, then f is a homeomorphism. (The proof is elementary)
8
A very convenient way to construct homotopies from a quotient space is the following
fact. Proposition: Let π : X → Y be a quotient map, and let I denote a locally compact
separated space, then π × idI : X × I → Y × I is a quotient map. (The proof is not that
easy. See e.g. Bredon, Topology and Geometry, Chapter I, prop. 13.19)

2
 A. Touzé – 2023

1. Show that if f is an isomorphism, then its inverse is unique.

2. Show that f : X → Y is an isomorphism if and only if there is g, h :

Y → X such that f ◦ g = idY and h ◦ f = idX .
f g
3. The "2 out of 3" property. Let X −→ Y → − Z be two morphisms in
a category C. Show that if two morphisms among f , g, g ◦ f , are
isomorphisms, then the third morphism is also an isomorphism.

4. Show that a functor preserves isomorphisms.

Exercice 10. Homotopy functors. Let F : Top → C be a functor.

Show that the following assertions are equivalent: (i) for all continuous maps
f, g if f ∼ g then F (f ) = F (g), (ii) F sends the homotopy equivalences
to isomorphisms in C. (A functor satisfying these conditions is called a
homotopy functor.)

Exercice 11. The homotopy category. The homotopy category hTop

is the category whose objects are topological spaces, whose set of morphisms
HomhTop (X, Y ) is the set of homotopy classes of continuous maps f : X →
Y , and whose composition is given by [f ] ◦ [g] = [f ◦ g] (the brackets indicate
the homotopy class of a map).

1. Check that the composition law in hTop is well-defined and that hTop
is indeed a category.

2. Let π : Top → hTop be the functor which is the identity on objects

and such that π(f ) = [f ]. Show that a functor F : Top → C is a
homotopy functor if and only if there is a functor F : hTop → C such
that F = F ◦ π.

3. Redo exercises 3 and 4 by using the category hTop and exercise 8.

Exercice 12. Categorical coproducts. Let (Xα )α∈A be a family

of objects of a category C. An object X of C, equipped with morphisms
iα : Xα → X, is called a coproduct of the Xα if it satisfies the following
universal property.
"For all families of morphisms fα : Xα → Y there is a unique f : X → Y
such that f ◦ iα = fα ."

1. Show that if X equipped with the iα is a coproduct of the Xα and if

X 0 equipped with the i0α is another coproduct of the Xα then there is
'
a unique isomorphism φ : X − → X 0 such that φ ◦ ια = ι0α for all α.

3
 A. Touzé – 2023

2. Show that coproducts always exist (and describe them!) in the follow-
ing categories: Ens, Top, R−Mod.

Exercice 13. Pushouts. A commutative square in C:

f
A B
g p
q
C D

is called a pushout square if it satisfies the following universal property. For

all morphisms φB and φC making the outer square (ABZC) commute, there
is a unique map φ making the two triangle commute:

A B
φB
C D .
∃!φ
φC
Z

f
A B
1. Show that if g p0 is another pushout with the same maps f and
q0 0
C D
'
g, then there is a unique isomorphism φ : D −
→ D0 , such that p0 ◦φ = p
and q 0 ◦ φ = q.

2. Show that if X is a topological space and if F and G are two closed

subsets, then the square with F ∩ G = A, F = B, G = C, X = D and
where the morphisms are given by the inclusions, is a pushout square.
Show that this may not be the case if we don’t assume that both F
and G are closed.
f
A B
3. Show that every diagram g in Ens, in Top or in R−Mod can
C
be completed into a pushout square.

4
 A. Touzé – 2023

Exercise sheet n°2

First computations in singular homology

3. First singular homology calculations

Exercice 14. General properties of degrees. Let n > 1.

1. Let f, g : S n → S n . Show that deg(f ◦ g) = deg(f ) deg(g).

2. Show that if f : S n → S n is a homotopy equivalence then deg(f ) = ±1.

3. Show that (deg f 6= 0) ⇒ (f is surjective).

Exercice 15. Degrees of homogeneous polynomials. Let S 1 be the

unit circle of C. Let 0 <  < 2n
1
, and let H + and H − be the two open
hemispheres of S :
1

H + = {eiθ , θ ∈] − , π + [} , H − = {eiθ , θ ∈]π − , 2π + [} .

Let n > 1 and consider the following two open subsets of S 1 :

G  2iπ  (2i + 1)π 

U +
= iθ
e , − <θ< + ,
0≤i<n
n n n n
G  (2i + 1)π  (2i + 2)π 

U− = eiθ , − <θ< + .
0≤i<n
n n n n

Let f : S 1 → S 1 be such that f (z) = z n .

By considering the Mayer-Vietoris sequences associated to the decom-
positions S 1 = U + ∪ U − and S 1 = H + ∪ H − , show that deg(f ) = n.

Exercice 16. A proof of the fundamental theorem of algebra. Let

P : C → C be a polynomial of degree d > 0.

1. Assume that P has no zeroes in the unit disc. Prove that the map f :
S 1 → S 1 , such that f (z) = P (z)/|P (z)| is homotopic to the constant
map with value P (0)/|P (0)|, hence that it has degree 0.

2. Assume that P has no zeroes in {z ∈ C , |z| ≥ 1}. Prove that f : S 1 →

S 1 , f (z) = P (z)/|P (z)| is homotopic to z 7→ z d .

3. Use degree theory to prove that P must have at least a zero in C.

1
 A. Touzé – 2023

Exercice 17. Bouquet. Let (Xα , xα )α∈A be a family of pointed topolog-

ical spaces. Their bouquet or their wedge is the topological space α∈A Xα
W

obtained as the quotient of α∈A Xα by the smallest equivalence relation9

F

which identifies all the xα .

1. Let (X, x) and (Y, y) be two pointed topological spaces. Assume that
{x} is closed in X and that {y} is closed in Y . Show that X ∨ Y is
homeomorphic to the subspace X × {y} ∪ {x} × Y of X × Y .

2. Let (X, x) and (Y, y) be two pointed topological spaces. Assume that
{x} and {y} are strong deformation retracts of open subsets U ⊂ X
and V ⊂ Y . Show that Y , X and Y ∩ X (viewed as subspaces of
X ∨ Y ) are strong deformation retracts of the open subsets U ∨ Y ,
X ∨ V , and U ∨ V .

3. Show that for all positive i, the inclusions X ,→ X ∨ Y and Y ,→

X ∨ Y induce an isomorphism Hi (X) ⊕ Hi (Y ) ' Hi (X ∨ Y ), and that
H0 (X ∨ Y ) is a free R-module of rank ]π0 (X) + ]π0 (Y ) − 1.

Exercice 18. The Torus and the Klein Bottle. Let C = S 1 × [0, 1]
be the cylinder. Let 0 <  < 61 . View S 1 as the unit circle of C. Let
I =]1/3 − , 2/3 + [⊂ [0, 1] and let J = [0, 1/3 + [∪]2/3 − , 1] ⊂ [0, 1].

1. The Torus is the quotient of C by the smallest equivalence relation

such that (z, 0) ≡ (z, 1). Draw a picture.

(a) Let U , V be the images in the Torus of the subsets S 1 × I, S 1 × J

of C. Show that U and V are two open subsets of the Torus.
Show furthermore that there is a commutative square:

Hi (U ∩ V ) Hi (U ) ⊕ Hi (V )
' '

R⊕R R⊕R

in which the top horizontal map is induced by the inclusions of

U ∩ V in U and V and the bottom horizontal map is given by the
matrix [ 11 11 ].
(b) Use the Mayer-Vietoris property to compute the homology of the
Torus.
9
i.e. by the equivalence relation which identifies all the xα , and which makes no other
identification. In general the smallest equivalence relation means the equivalence relation
with the smallest possible equivalent classes.

2
 A. Touzé – 2023

2. The Klein Bottle is the quotient of C by the smallest equivalence

relation such that (z, 0) ≡ (z, 1) (z is the conjugate of z). Draw a
picture.

(a) Let U and V be the images of S 1 × I and S 1 × J in the Klein

bottle. Show that U and V are two open subsets of the Torus.
Show furthermore that there is a commutative square:

Hi (U ∩ V ) Hi (U ) ⊕ Hi (V )
' '

R⊕R R⊕R

in which the top horizontal map is induced by the inclusions of

U ∩ V in U and V and the bottom horizontal map is given by the
matrix 11 −1
1 .


(b) Use the Mayer-Vietoris property to compute the homology of the

Klein bottle. Compare this homology for R = F2 , Fp with p odd,
and Z.

Exercice 19. Suspension. The suspension ΣX of a topological space X

is the quotient of the cylinder X × [0, 1] by the smallest equivalence relation
which identifies together the points of X × {0} on the one hand, and which
identifies together the points of X × {1} on the other hand. Every map
f : X → Y induces a map Σf : ΣX → ΣY .

1. Show that ΣS n is homeomorphic to S n+1 and that ΣDn is homeomor-

phic to Dn+1 .

2. Show that ΣX is always arcwise connected.

3. Show that H1 (ΣX) ' Ker(H0 (X) → H0 ({∗})), and that for all i > 0
there is an isomorphism Hi (X) ' Hi+1 (ΣX), which is natural with
respect to f , i.e. there are commutative squares
'
Hi (X) Hi+1 (ΣX)
Hi (f ) Hi+1 (Σf )
'
Hi (Y ) Hi+1 (ΣY )

4. Show that for all n > 0 and all k ∈ Z there exists a map f : S n → S n
of degree k.

3
 A. Touzé – 2023

Exercise sheet n°3

Complexes

4. Complexes

Exercice 20. Injectivity and surjectivity in homology.

1. Find an explicit complex C and an explicit subcomplex D ⊂ C such

that the map H1 (C) → H1 (D) induced by the inclusion is neither
injective, nor surjective.

2. Find an explicit complex C and an explicit quotient complex C  E

such that the map H1 (C) → H1 (E) induced by the inclusion is neither
injective, nor surjective.

Exercice 21. Nullhomotopic complexes. Let C be a complex and

assume that idC ∼ 0 (such a complex is called nullhomotopic). Show that
Hi (C) = 0 for all i ∈ Z.

Exercice 22. Euler characteristic. Assume that k is a field. Let C be a

complex with only a finite number of nonzero k-vector spaces Cn , and such
that these vector spaces are finite-dimensional. The Euler characteristic of
C is the number of χ(C) = n∈Z (−1)n dim Cn .
P

1. Show that χ(C) = n∈Z (−1) dim Hn (C).

n (Use the rank-nullity the-
P

orem).

2. Show that homotopy equivalent complexes have the same Euler char-
acteristic.

3. Show that in a short exact sequence of complexes 0 → C → D → E →

0 one has χ(C) + χ(E) = χ(D).

Exercice 23. The nine lemma. Consider the commutative diagram in

1
 A. Touzé – 2023

which the rows are short exact sequences and the columns are complexes.
0 0 0

0 A B C 0

0 A0 B0 C0 0

0 A00 B 00 C 00 0

0 0 0
Show that if any two of the columns are short exact sequences, then so is
the third one.

Exercice 24. A universal coefficient theorem. Let C be a complex

whose objects are free Z-modules. Fix an integer q, and denote by C ⊗Z Z/qZ
the complex which is equal to the free Z/qZ-module Cn ⊗Z Z/qZ in degree
n, and whose differential is equal to d ⊗ idZ/qZ .
q
1. An example. Let C = Z →
− Z . Compute the homology of
|{z}
degree 0
C ⊗Z Z/qZ. Compare with the homology of C tensored with Z/qZ.
2. The universal coefficient theorem. Show that for all complexes
C and all i, there is a short exact sequence of Z-modules:
0 → Hi (C) ⊗Z Z/qZ → Hi (C ⊗Z Z/qZ) → (q) Hi−1 (C) → 0 .
where (q) A denotes the q-torsion subgroup of an abelian group A, that
is: (q) A = {a ∈ A ; qa = 0}.
q
[Hint: use the short exact sequence 0 → C −
→ C → C/qC → 0.]

Exercice 25. The five lemma and a typical application of it.

1. Consider the following commutative diagram of R-modules in which
the rows are exact sequences.
H1 H2 H3 H4 H5
f1 f2 f3 f4 f5

K1 K2 K3 K4 K5
Show that if f1 , f2 , f4 and f5 are isomorphisms, then f3 is an isomor-
phism. (This is the "five lemma")

2
 A. Touzé – 2023

2. Consider the following commutative diagram of complexes of R-

modules in which the rows are short exact sequences.

0 C C0 C 00 0
f f0 f 00

0 D D0 D00 0

Show that if two of the morphisms of chain complexes f f 0 and f 00 are

quasi-isomorphisms, then so is the third one. isomorphisms, then f3
is an isomorphism.

Exercice 26. Homology of posets. Let P be a finite poset. For

all i ≥ −1, an i-chain in P is an increasing sequence of i + 1 elements in
P : (x0 < · · · < xi ). (By convention there is only one −1-chain, namely
the empty chain ().) We let Ci (P ) be the free R-module generated by the
i-chains in P , and we define R-linear maps di : Ci (P ) → Ci−1 (P ) by
j
di (x0 < · · · < xi ) = (−1)k (x0 < · · · < x
ck < · · · < xi )
X

k=0

where x
ck means that xk is deleted.

1. Check that the Ci (P ) together with the maps di yield a complex of

R-modules. ( The homology of this complex is called the homology of
the poset P .)

2. Compute the homology of the poset P1 with one element.

3. A vanishing property. Let P be a finite poset having a least element

m (i.e. for all x in P , x ≥ m).
Let hi : Ci (P ) → Ci+1 (P ) be the R-linear map such that si (x0 < · · · <
xi ) = (m < x0 < · · · < xi ) if x0 > m and 0 if x0 = m. Show that h is
a homotopy between idX and 0, and deduce Hi (P ) for all i.

4. Consider the poset P with 7 elements and Hasse diagram10 :

5 6 7

3 4 .

1 2
10
In the Hasse diagram of a poset P , there is one vertex for each element of P , and
there is a line x − y with y above x if y > x.

3
 A. Touzé – 2023

(a) What is the homology of the poset P \ {1} ?

(b) The complex C(P \ {1}) identifies with a subcomplex of C(P ).
Compute the homology of the quotient complex C(P )/C(P \{1}).
(c) Deduce the homology of P .

4
 A. Touzé – 2023

Exercise sheet n°4

Complexes

5. Singular and simplicial homology.

Exercice 27. Reduced singular homology11 . Let X be a topological

space, and let p : X → {∗} denote the unique map to the one-point space.
The reduced singular homology of degree i of X is the R-module H i (X)
defined by
Hi (p)
 
H i (X) := Ker Hi (X) −−−→ Hi ({x}) .

1. Show that H i (X) is a functor of the variable X.

2. Show that the composition H i (X) → Hi (X) → Hi (X, {x}) is an iso-

morphism for all x ∈ X.

3. Show that the Mayer-Vietoris sequence and the long exact sequence of
a pair remain valid if the singular homology is replaced by the reduced
singular homology.
sing
4. Let C (X) be the complex which equals the singular cohomology
complex C sing (X) in degrees ≥ 0, and which has an additional object
sing
C −1 (X) = R, and an additional differential  : C0sing (X) → R such
that (σ) = 1 for all 0-singular simplices ∆0 → X:
sing d d d 
C → C2sing (X) −
(X) := . . . − → C1sing (X) −
→ C0sing (X) →
− R
sing
Show that H i (X) = Hi (C (X)) for all i.

Exercice 28. The Mayer-Vietoris sequence for simplicial homology.

Let K be a geometric simplicial complex. A simplicial subcomplex of K is
a subset L ⊂ K which is a geometric simplicial complex. Show that if L
and L0 are two simplicial subcomplexes of K, then L ∩ L0 is also a simplicial
subcomplex of K. Show moreover that if L ∪ L0 = K there is a long exact
sequence in simplicial homology:
∂
· · · → Hisimpl (L∩L0 ) → Hisimpl (L)⊕Hisimpl (L0 ) → Hisimpl (K) − simpl
→ Hi−1 (L∩L0 ) → . . .

11
The reduced singular homology contains the same information as the homology, and
it is often more convenient in computations. To understand why, compute the reduced
homology of the spheres with the Mayer-Vietoris exact sequence, and compare with the
computation of the (unreduced) homology of spheres done in the course!

1
 A. Touzé – 2023

Exercice 29. Collapsing pairs. Let K be a geometric simplicial com-

plex. A collapsing pair of K is a pair (sk , sk+1 ) of simplices of K such that
(i) sk+1 has dimension k + 1 ≥ 1, (ii) sk is a face of sk+1 and of no other
simplex of dimension k + 1 (iii) sk+1 is not a face of any simplex of K.

1. Show that L = K \ {sk+1 , sk } is a geometric simplicial complex.

2. Show that the inclusion C∗simpl (L) ⊂ C∗simpl (K) is a quasi-isomorphism.

The drawing below represents on the left a geometric simplicial com-
plex K with collapsing pair (hx1 , x4 i, hx1 , x2 , x4 i) and on the right the
geometric simplicial complex L.

3. Application : compute the singular homology of the geometric simpli-

cial complex corresponding to ∆3 .

Exercice 30. Simplicial homology vs singular homology. Recall the

morphism of complexes φK : C simpl (K) → C sing (|K|) from the course. In
this exercise we give a direct proof that this morphism is a quasi-isomorphism
when the geometric simplicial complex K is finite.
1. Show that when K has dimension zero, φK is a quasi-isomorphism.

2. Assume that φK is an isomorphism for geometric simplicial complexes

of dimension n − 1. We now turn our attention to the geometric
simplicial complexes of degree n.

(a) Let K the geometric simplicial complex corresponding12 to an

affine n-simplex ha0 , . . . , an i and let ∂K be the geometric simpli-
cial complex corresponding to its boundary.
i. Show that φK induces an isomorphism in degree 0 homology.
ii. Show that C simpl (K) is isomorphic to the part of degree ≥ 0
of the complex of the poset {a0 < · · · < an } (see exercise 26)
and deduce that Hisimpl (K) is zero for i > 0.
iii. Deduce that φK is a quasi-isomorphism.
12
That is, K is the set of all the simplices which are convex hulls of
nonempty subsets of {a0 , . . . , an }, and ∂K = K \ {ha0 , . . . , an i}. For example, if
n = 2 then K = {ha0 , a1 , a2 i, ha0 , a1 i, ha0 , a2 i, ha1 , a2 i, ha0 i, ha1 i, ha2 i} and ∂K =
{ha0 , a1 i, ha0 , a2 i, ha1 , a2 i, ha0 i, ha1 i, ha2 i}.

2
 A. Touzé – 2023

iv. Check that there is a commutative diagram of complexes:

C simpl (∂K) C simpl (K) Rha0 , . . . , an i

φ∂K φK φK

C sing (|∂K|) C sing (|K|) C sing (|K|)/C sing (|∂K|)

where the horizontal morphisms are inclusions and quotient

morphisms and where φK is obtained by passing φK to the
quotient. Deduce that φK is a quasi-isomorphism.
(b) Assume now that φK is a quasi-isomorphism for geometric sim-
plicial complexes of dimension n with less than N n-simplices.
Let M be a geometric simplicial complex with N simplicies, and
let L be the geometric simplicial complex obtained from M by
removing one n-simplex K = ha0 , . . . , an i.
i. Check that there is a commutative diagram of complexes:

C simpl (L) C simpl (M ) Rha0 , . . . , an i

φL φM φM,L

C sing (|L|) C sing (|M |) C sing (|M |)/C sing (|L|)

where the horizontal morphisms are inclusions and quotient

morphisms and where φK is obtained by passing φK to the
quotient.
ii. Check that there is a commutative triangle in which ψ is
induced by the inclusion C sing (|K|) ⊂ C sing (|M |) :

Rha0 , . . . , an i
φK
φM,L
ψ
C sing (|K|)/C sing (|∂K|) C sing (|M |)/C sing (|L|)

Deduce that φM,L is a quasi-isomorphism.

iii. Deduce that φM is a quasi-isomorphism.
3. Conclude the proof that φK is an isomorphism for all finite geometric
simplicial complex.

Exercice 31. The Euler formula. Let K be a geometric simplicial

complex of dimension n, such that |K| is homeomorphic to S n . Let ni
denote the number of i-simplices of K. Show that

(−1)i ni = 1 + (−1)n .
X

0≤i≤n

3
 A. Touzé – 2023

Deduce the Euler Formula, i.e. if P is a regular polyedron of dimension 2,

then:
#vertices − #edges + #faces = 2 .

4
 A. Touzé – 2023

Exercise sheet n°5

CW-complexes

6. CW-structures

Exercice 32. Real projective spaces. If k is a field, we denote by kP n

the projective space of dimension n over k.
As a set, kP n is the set of lines of kn+1 (passing through the origin 0).
Since every line is determined by a nonzero vector, and since two nonzero
vectors determine the same line if and only if they are colinear, we can also
view kP n as the set of nonzero vectors module the action of the multiplica-
tive group k∗ by multiplication:

kP n = kn+1 \ {0} /k∗ .

We denote by [x0 : · · · : xn ] ∈ kP n the class of (x1 , . . . , xn ) ∈ kn+1 \ {0}.

Then kP n−1 identifies with the subset of points of kP n such that xn = 0.
If k = R of C, we endow kP n with the quotient topology relative to the
canonical quotient map kn+1 \ {0} → kP n .
In the sequel of this exercise, we assume that k = R.

1. Prove that RP n is homeomorphic to the quotient of S n by the equiv-

alence relation ∼
= which identifies antipodal points (i.e. x ∼
= −x for all
x ∈ S ).
n

2. Let S n−1 = S n ∩ Rn × {0}, and let H + = S n ∩ Rn × R+ . Prove that

there is a pushout square (where q denotes the quotient map):
q
S n−1 S n−1 / ∼
=
.
q
H+ Sn/ ∼
=

3. Deduce that RP n has a CW-complex structure with i-skeleton equal

to RP i for 0 ≤ i ≤ n, and with exactly one cell in each dimension i.

4. Show that RP 1 is homeomorphic to S 1 , and that the attaching map

of the 2-cell S 1 → RP 1 ' S 1 has degree 2.

5. Show that for all n ≥ 2, RP n−1 /RP n−2 ' S n−1 , and that the attach-
ing map of the n-cell has degree (−1)n + 1.

6. Compute the homology of RP n .

1
 A. Touzé – 2023

Exercice 33. Complex projective spaces. We view S 2n+1 as the unit

sphere of the euclidean space Cn+1 . We let U denote the multiplicative group
of complex numbers of module 1. Thus U acts on S 2n+1 by λ · (z0 , . . . , zn ) =
(λz0 , . . . , λzn ).
1. Prove that CP n is homeomorphic to the quotient of S 2n+1 by the
action of U .

2. Let S 2n−1 = S 2n+1 ∩Cn ×{0}, and let H + = S 2n+1 ∩ Cn ×R+ . Prove

that there is a pushout square (where q denotes the quotient map):

q
S 2n−1 S 2n−1 /U
.
q
H+ CP n

3. Deduce that CP n has a CW-complex structure with i-skeleton equal

to CP i for 0 ≤ i ≤ n, and with exactly one cell in each dimension 2i
for 0 ≤ i ≤ n.

4. Compute the homology of CP n .

Exercice 34. CW-structures and coverings. Let p : E → B be

a covering space. Assume that B is a CW-complex. Prove that E is a
CW-complex as well, with i-skeleton Ei = p−1 (Bi ).

7. A solution of question 5 of exercise 32.

Let us fix the notations.
• S n = {(x0 , . . . , xn−1 , xn ) | kxk2 + x2n = 1} ⊂ Rn+1
| {z }
=:x

• H + = {(x, xn ) ∈ S n | xn ≥ 0}

• H − = {(x, xn ) ∈ S n | xn ≥ 0}

• S n−1 = H + ∩ H −

• We denote by a : S n → S n the antipodal map, i.e. a(x) = −x.

Sn
• RP n =
a(x) ∼ x ∀x ∈ S n
S n−1
• RP n−1 = ⊂ RP n
a(x) ∼ x ∀x ∈ S n−1

2
 A. Touzé – 2023

7.1. Let us show that RP n /RP n−1 is homeomorphic to S n .

We consider the continuous map induced by the inclusion H + ⊂ S n :

H+ Sn RP n
= .
S n−1 a(x) ∼ x ∀x ∈ S n RP n−1
+ all points of S n−1
identified

It is bijective, its source is compact and its target is Hausdorff hence it is a

≈
homeomorphism. Thus, to construct a homeomorphism h : RP n /RP n−1 − →
≈
S n , it suffices to construct a homeomorphism h+ : H + /S n−1 − → Sn.

Let us give an explicit formula for h+ . We first introduce a map

π + : S n → S n by the following formulas:
q 
 4xn
x, 2x n − 1 if xn ≥ 0,
π + (x, x) = 1+xn
(0, −1) if xn ≤ 0.

Then π + is continuous on H + and on H − , hence on S n . Moreover,

• the restriction of π + to H − is constant equal to (0, −1),

• the restriction of π + to S n \ H − is a bijection 13 onto S n \ {(0, −1)}.

Thus π + : H + → S n induces a a continuous bijection:

h+ : H + /S n−1 → S n .

The domain of h+ is compact, its codomain is Hausdorff, hence h+ is a

homeomorphism.

Remark. We only need to introduce the restriction π + : H + → S n to define

h+ , but the map π + defined on the whole sphere will be useful later in the
solution.

7.2 A degree computation.

Now we have to compute deg(f ), where f is the following composition (the
first two maps are the quotient maps):
h
f := S n  RP n  RP n /RP n−1 −
→ Sn .
13
The inverse bijection is the map:
r 
yn + 3 1
(y, yn ) 7→ y, (yn + 1) .
4(1 + yn ) 2

3
 A. Touzé – 2023

Let us give a concrete formula for f . We start by a formula for the

restriction of f to H + . For this purpose we consider the following commu-
tative diagram, in which the vertical double arrows are the quotient maps
and the horizontal arrows are induced by the inclusion H + ⊂ S n :

H+ Sn

H+ Sn
= RP n
a(x) ∼ x ∀x ∈ S n−1 a(x) ∼ x ∀x ∈ S n

.
H+ Sn RP n
=
S n−1 a(x) ∼ x ∀x ∈ S n RP n−1
+ all points of S n−1

identified
h
h+
Sn

By definition, the restriction of f to H + is the composition given by the

top right corner of the diagram. By commutativity of the diagram, this is
equal to the composition given by the bottom left corner of the diagram. By
definition of h+ , the latter sends x ∈ H + to π + (x) ∈ S n . Thus we obtain:

f (x) = π + (x) if x ∈ H + .

Moreover, we have f (−x) = f (x) for all x ∈ S n , because x and −x are

identified in RP n . But if x ∈ H − then −x ∈ H + , thus we obtain:

f (x) = π + (−x) if x ∈ H − .

We note that f (x) is equal to (0, −1) for all x ∈ S n−1 .

Let us compute the degree of f . The following lemma allows us to

compute the degree of a map which is constant on S n−1 , from the restrictions
to H + and H − .

Lemma. Let α : S n → Z be a continuous map, such that α(x) = z for all

x ∈ S n−1 . Define α+ , α− : S n → Z by
( (
α(x) if x ∈ H + , − α(x) if x ∈ H − ,
α (x) =
+
and α (x) =
z if x ∈ H − , z if x ∈ H + .

Then Hn (α) = Hn (α+ ) + Hn (α− ). In particular, if Z = S n , then deg α =

deg α+ + deg α− .

4
 A. Touzé – 2023

Now we apply the previous lemma to our situation. The expression of f

that we have computed shows that
f + = π+ , f − = π+ ◦ a .
Thus we have
deg f = deg(π + ) + deg(π + ◦ a)
= deg(π + ) + deg(π + ) deg(a)
= deg(π + )(1 + (−1)n ) .
Now we claim that π + is homotopic to the identity. Indeed, π + (x) 6= −x
for all x ∈ S n , hence 0 6∈ [x, π + (x)], hence tx + (1 − t)π + (x) is never zero.
Thus the continuous map:
tx + (1 − t)π + (x)
H(x, t) =
ktx + (1 − t)π + (x)k
is well defined an yields a homotopy between π + and id. Thus deg(π + ) = 1
hence:
deg(f ) = 1 + (−1)n .
QED.
Proof of the Lemma. We proceed in several steps.
1) Let ι± : H ± /S n−1 → S n /S n−1 be the maps induced by the inclusions
H ± ⊂ S n . Then it follows from the Mayer-Vietoris exact sequence that
the R-linear map
! !
H+ H− Sn
 
Hn (ι+ )
h i
Φ= Hn (ι− )
: Hn ⊕ Hn →
− Hn
S n−1 S n−1 S n−1
is an isomorphism. (see also exercise 17.)
2) Consider the composition:

Sn Sn ≈ H ±
p± :
 ←
− n−1 .
S n−1 H∓ S
Then p ◦ι = id and p ◦ι is constant. Thus by using the functoriality
± ± ± ∓

of Hn , we see that the composition (in dots)

! !
H+ H− Sn
 
Φ
Hn ⊕ Hn Hn
S n−1 S n−1 S n−1
[Hn (p+ ),Hn (p− )]
! !
H+ H−
Hn ⊕ Hn
S n−1 S n−1

is equal to the identity, hence Φ−1 = [Hn (p+ ), Hn (p− )].

5
 A. Touzé – 2023

3) We have a commutative diagram (this diagram commutes because it

comes from a commutative diagram on the level of topological spaces.
The upper horizontal bars denote the maps induced on the quotients.)
! !
Sn H+ H−
 
Hn (quot.) n Φ
H n (S n ) H Hn ⊕ Hn
S n−1 S n−1 S n−1
Hn (α)
Hn (α)
[Hn (α|H + ),Hn (α|H − )]
Z

Let Ψ denote the composition:

Ψ = Φ−1 ◦ Hn (quot.) .

The commutativity of the above diagram implies that Hn (α) equals the
composition
! !
Ψ H+ H− [Hn (α|H + ),Hn (α|H − )]
H (S ) −
n n
→ Hn ⊕ Hn −−−−−−−−−−−−−−→ Z
S n−1 S n−1

4) We write down explicitly the components of Ψ.

h
Hn (q + )
i Sn ≈ H ±
We have Ψ = Hn (q − )
where q ± = S n  ←
− n−1 .
H∓ S
Thus we have

Hn (α) = Hn (α|H + ◦ q + ) + Hn (α|H − ◦ q − )

To finish, we observe that α± = α|H ± ◦ q ± , whence the result.

6
 A. Touzé – 2023

Exercise sheet n°6

Extensions

8. Ext and projectives

Exercice 36. Split exact sequences.

f g
1. Let 0 → M 0 −
→M → − M 00 → 0 be a short exact sequence of R-modules.
Show that the following assertions are equivalent.

(i) There exists a morphism i : M 00 → M such that g ◦ i = id.

(ii) There exists a morphism p : M → M 0 such that p ◦ f = id.
(iii) The exact sequence is isomorphic to the trivial exact sequence,
that is, there is an isomorphism φ making the following diagram
commutative:
f g
0 M 00 M M0 0
 id  φ
0 [0,id]
0 M 00 M 00 ⊕ M 0 M0 0

When a short exact sequence satisfies these conditions, one says that
it splits.

2. Give an example of a short exact sequence which does not split.

3. Show that an R-module P is projective if and only if every short exact

sequence 0 → M → N → P → 0 splits.

Exercice 37. Some Ext-computations over a PID. Let R be a PID.

1. Write an explicit projective resolution of a cyclic module C = R/rR.

2. Let C = R/rR and C 0 = R/sR be two cyclic modules. Compute

Exti (C, C 0 ) for all i.

3. Let M be a finitely generated R-module. Prove that Ext1R (M, R) = 0

if and only if M is free14 .
14
For R = Z, one can prove that the property still holds for M countable instead of
finitely generated. The Whitehead problem asks if this holds for all abelian groups M ,
without any finiteness or countability hypothesis. It was proved by Shelah in 1974 that
the answer to this question depends on the axioms of set theory used. In particular, under
the continuum hypothesis (i.e. there is no set cardinal strictly between the cardinal of N
and the cardinal of R) then the equivalence holds for all abelian groups M .

1
 A. Touzé – 2023

Exercice 38. Some Ext-computations over k[x]/xn+1 . Let k be a

field and let R = k[x]/xn+1 . There is a morphism of k-algebras  : R → k
such that (x) = 0. In this way, we see k as an R-module.

1. Write an explicit projective resolution of k.

2. Compute ExtiR (k, k) for all i.

Exercice 39. Some Ext-computations over k[x, y]. Let k be a field

and let R = k[x, y]. There is a morphism of k-algebras  : R → k such that
(x) = 0 = (y). In this way, we see k as an R-module.

1. We consider the free R-module F = R ⊗k Λ(x, y), that is, F is the free
R-module with basis given by the symbols 1 ⊗ 1, 1 ⊗ x, y ⊗ 1, 1 ⊗ (x ∧ y):

F = R1⊗1 ⊕ R1⊗x ⊕ R1⊗y ⊕ R1⊗(x∧y) .

We let d : F → F be the R-linear map such that

d(1 ⊗ x ∧ y) = x(1 ⊗ y) − y(1 ⊗ x)

d(1 ⊗ x) = x(1 ⊗ 1)
d(1 ⊗ y) = y(1 ⊗ 1)
d(1 ⊗ 1) = 0

We decide that the summand R1⊗1 has degree zero, that the sum-
mands R1⊗x and R1⊗y have degree 1 and that the summand R1⊗(x∧y)
has degree 2. Show that (F, d) is a chain complex of R-modules and
compute its homology.

2. Compute ExtiR (k, k) for all i.

Exercice 40. Degree shifting.

1. Assume that there is a short exact sequence 0 → K → P → M → 0

with P projective. Show that ExtiR (M, N ) ' Exti+1
R (K, N ) for all N
and all positive i.

2. Assume that there is an exact sequence

0 → K → Pn → Pn−1 → · · · → P0 → M → 0

in which every Pk is projective. Prove that there is an isomorphism

ExtiR (M, N ) ' Exti+n
R (K, N ) for all positive i.

2
 A. Touzé – 2023

Exercice 41. Projective dimension of a module. We say that a

module M has projective dimension less or equal to n if it admits a projective
resolution (P, ) with Pi = 0 for i > n.
Prove that the following assertions are equivalent (Hint: you may use
the previous exercise on degree shifting).

1. M has projective dimension less or equal to n,

2. ExtiR (M, N ) = 0 for all i > n and all R-modules N ,

3. Extn+1
R (M, N ) = 0 for all R-modules N .

Exercice 42. Finiteness of Ext over noetherian rings. Let R be a

left noetherian ring. Let M and N be finitely generated R-modules. Show
that for all i, ExtiR (M, N ) is a finitely generated R-module.

Exercice 43. The Horseshoe lemma.

f g
1. Let 0 → X − →Y → − Z → 0 be a short exact sequence of R-modules.
Let (P , X ) be a projective resolution of X and let (P Z , Z ) be a
X

projective resolution of Z.

Prove that there exists a projective resolution (P Y , Y ) of Y satisfying

the following three properties.

a) PiY = PiX ⊕ PiZ for all i ≥ 0.

b) The maps f i = id : PiX → PiX ⊕ PiZ define a morphism of
 
0
complexes f : P → P Y which lifts f .
X

c) The maps g i = [0, id] : PiX ⊕ PiZ → PiZ define a morphism of

complexes g : P Y → P Z which lifts g.

2. Prove that every short exact sequence 0 → M 00 → M → M 0 → 0

induces a long exact sequence15 in Ext:
00 ∂ ∂
· · · → Exti−1 → ExtiR (M 0 , N ) → ExtiR (M, N ) → ExtiR (M 00 , N ) −
R (M , N ) − → ... .

15
In the course, we constructed the long exact sequence by using injective coresolutions
- this exercise gives a contruction which uses only projectives.

3
 A. Touzé – 2023

9. Injectives

Exercice 44. Injectives and splitting.

1. Show that an R-module J is injective if and only if every short exact
sequence 0 → J → M → N → 0 splits.

2. Let 0 → Jn → Jn−1 → · · · → J0 → M → 0 be an exact sequence in

which all the Jk are injective. Show that M is a direct summand of
J0 (hence M is injective).

Exercice 45. Injectives and domains. Let R be a domain, with

fraction field F .
1. Let I be a nonzero ideal of R and let f : I → F be an R-linear map.

(a) Show that the map I \ {0} → F , a 7→ f (a)/a is constant. Let c

denote its value, hence f (r) = rc for all r ∈ I.
(b) Deduce that F is an injective R-module.

2. Let R be a domain. Prove that R is injective as an R-module if and

only if R is a field. (Hint: R cannot be a proper summand of F )

Exercice 46. Direct sums of injectives. We know from the course that
a product of injectives is injective. It is a natural question to ask whether
a direct sum of injectives is injective. In this exercise we prove that this is
indeed the case for noetherian rings16 Let R be a left noetherian ring and
let (Jα )α∈A be a family of injectives R-modules.
1. Let I be an ideal of R and let f : I → α∈A Jα be an R-linear map.
L

Show that there is a finite subset B ⊂ A such that Im f ⊂ α∈B Jα .

L

(Hint: use that J is a finitely generated R-module).

2. Use Baer’s criterion to prove that is injective.

L
α∈A Jα

Exercice 47. Self-injective rings. A ring R is called self-injective if the

left module R is injective.
1. Show that for n > 0, Z/nZ is self-injective (use Baer’s criterion).

2. Show that if R is self-injective and left noetherian, then every projec-

tive R-module is injective. (use the previous exercise on direct sums
of injectives)
16
In fact the converse is also true: a ring R is left noetherian if and only if every direct
sum of injectives is injective (Bass-Papp theorem).

4
 A. Touzé – 2023

3. Let R be a self-injective and noetherian ring. Let M be an R-module.

Show that the following assertion are equivalent.

(i) M is projective
(ii) For all i > 0 and all R-modules N , ExtiR (M, N ) = 0.
(iii) There is a positive integer i such that for all R-modules N ,
ExtiR (M, N ) = 0.

10. Extensions

Exercice 48. Classification of extensions of Z/pZ by Z. Let p be

a prime number. Show that every extension of Z/p by Z is either trivial,
or equivalent to the following exact sequence, for a uniquely determined
a ∈ Z/pZ \ {0}, where πa denotes the unique map such that π(1) = a:
×p π
0 → Z −−→ Z −
→ Z/pZ → 0 .

Exercice 49. Proof of the "Ext1 = extensions" theorem. Let

E (M, N ) denote the set of isomorphism classes of extensions of M by
1

N . Let us fix a projective resolution (P, ) of M , and let Ext1R (M, N ) =

H 1 (HomR (P, N )).
π
1. Every extension 0 → N → E −→ M → 0 can be viewed as a resolution
(R, π) of M (with R0 = E, R1 = N and Ri = 0 for i > 1), hence
by the fundamental theorem of homological algebra there is a lifting
f : P → R of idM .

(a) Show that f 1 : P1 → N is a cycle of the complex HomR (P, N ).

0 0
(b) Show that f , f are two different choices of liftings, then f 1 − f 1
is a boundary of the complex HomR (P, N ).
π
(c) Show that if 0 → N → E − → M → 0 is the trivial extension, then
f 1 is a boundary of the complex HomR (P, N ).
(d) Show that the map

Φ : E 1 (M, N ) → Ext1R (M, N )

which sends the isomorphism class of 0 → N → E → M → 0 to

[f 1 ] is well-defined, and sends that trivial extension to 0.
h i
2. Let f : P1 → N be a cycle of HomR (P, N ). Set ∆f = d
−f : P1 →
P0 ⊕ N .

5
 A. Touzé – 2023

(a) Show that the map [, 0] : P0 ⊕ N → M induces a morphism

π : (P0 ⊕ N )/Im ∆f → M .
(b) Show that the diagram
0
π
0 → N −−id−→ P0 ⊕ N/Im ∆f −
→M →0 (∗)

is a short exact sequence.

(c) Show that if f and f 0 are two cycles such that f −f 0 is a boundary
then the associated exact sequences are isomorphic.
(d) Show that the map

Ψ : Ext1R (M, N ) → E 1 (M, N )

which sends [f ] to the isomorphism class of (∗), is well-defined.

3. Check that Φ and ψ are mutually inverse bijections.

6
 A. Touzé – 2023

Exercise sheet n°7

Bicomplexes

11. Bicomplexes

Exercice 46. Short exact sequences of bicomplexes. A short exact

f g
sequence of bicomplexes is a diagram of bicomplexes 0 → C 0 − →C→ − C 00 → 0
such that for all indices i, j, the following diagram is a short exact sequence
of R-modules:
i,j f i,j g i,j i,j
0 → C 0 −−→ C i,j −−→ C 00 → 0

1. Show such an exact sequence of bicomplexes induces a long exact se-

quence in homology:
∂
· · · → H i (Tot C 0 ) → H i (Tot C) → H i (Tot C 00 ) −
→ H i+1 (Tot C 0 ) → . . .

2. Let C be a bicomplex. A sub-bicomplex of C i.e. a family C 0 i,j of

submodules of the C i,j which is preserved by the differentials of C.
Define a quotient bicomplex C/C 0 , in such a way that there is a short
exact sequence:
0 → C 0 → C → C/C 00 → 0
in which the morphism of bicomplexes C → C 0 is given by the inclu-
sions of R-modules C 0 i,j ,→ C i,j .

Exercice 47. The mapping cone. Let f : C → D be a morphism of

complexes. Let B denote the bicomplex whose −1-rst column is equal to C,
whose 0-th column is equal to D, whose other columns are zero, and whose
horizontal differential is given by f . The mapping cone of f is the complex
C(f ) := Tot B.

1. Show that there is a long exact sequence in cohomology:

H i+1 (f )
· · · → H i (D) → H i (C(f )) → H i+1 (C) −−−−−→ H i+1 (D) → . . .

(Hint: show that D is a subcomplex of C(f ) and identify the quotient

complex. In the associated long exact sequence, use the expression of
the connecting morphism to get an explicit expression of it).

2. Show that f is a quasi-isomorphism if and only if its mapping cone

has trivial cohomology (i.e. equal to zero in each degree).

1
 A. Touzé – 2023

Exercice 48. A generalization of the nine lemma. Let C be a first

quadrant cochain bicomplex, that is a cochain bicomplex such that C i,j = 0
if i < 0 or j < 0.

1. Assume that all the rows of C are complexes with trivial cohomol-
ogy (i.e. equal to zero in each degree). Show that Tot C has trivial
cohomology.

2. Assume that Tot C has trivial cohomology, and that in addition all the
columns but one are also cochain complexes with trivial cohomology.
Show that the last column also has trivial cohomology.

Exercice 49. Tensor products of algebras and Ext. Let R and S

be two k-algebras over a commutative ring k. Their tensor product algebra
R ⊗k S is the k-module R ⊗k S, equipped with the product (r ⊗ s)(r0 ⊗ s0 ) :=
(rr0 ) ⊗ (ss0 ). If M is an R-module and M 0 is an S-module, the tensor
product M ⊗k M 0 , equipped with the action (r ⊗ s)(m ⊗ m0 ) := (rm) ⊗ (sm0 )
is naturally an R ⊗k S-module.

1. Prove that if M is projective as an R-module and if M 0 is projective

as an S-module, then M ⊗k M 0 is projective as an R ⊗k S-module.

2. Assume that k is a field. Prove that if (Q, ) is a projective resolution

of an R-module M and if (Q0 , 0 ) is a projective resolution of an S-
module M 0 then (Q ⊗k Q0 ,  ⊗ 0 ) is a projective resolution of M ⊗k M 0
as a R ⊗k S-module.

3. Assume that k is a field, that R and S are noetherian k-algebras and

that M and M 0 are finitely generated k-modules. Show that for all
nonnegative k there is an isomorphism

ExtiR (M, N ) ⊗k ExtjS (M 0 , N 0 ) ' ExtkR⊗k S (M ⊗k M 0 , N ⊗k N 0 ) .

M

i+j=k

Hint: start by proving the case of free finitely generated modules M

and M 0 , and then take free resolutions.

Exercice 50. Some Ext-computations over k[x1 , . . . , xn ]. Let k be a

field and let R = k[x1 , . . . , xn ]. There is a morphism of k-algebras  : R → k
such that (xi ) = 0 for all i. In this way, we see k as an R-module. Compute
ExtiR (k, k) for all i. (Hint: start with n = 1, and for higher n, use the
previous exercise)

2
 A. Touzé – 2023

Exercise sheet n°8

Left modules, right modules and tensor products.

12. Left modules versus right modules: examples.

Exercice 52. Group rings. Let G be a group, and let k be a commutative

ring. Define kG as the free k-module with basis (eg )g∈G . Thus every element
of kG can be written as a linear combination g∈G λg eg in which only a finite
P

number of scalars λg are nonzero. We make this k-module into a k-algebra

by letting:    

λ0h eh  = λg λ0h egh .

X X X
 λg eg  · 
g∈G h∈G g,h∈G2

1. Prove that kG is a commutative ring if and only if G is abelian.

2. Prove that defining a left kG-module M is equivalent to specifying a k-

module M together with a left action of G on M by k-linear morphisms
(i.e. an action such that g · − : M → M is k-linear).

3. Prove that defining a morphism of kG-modules f : M → N is equiva-

lent to specifying a k-linear map which is G-equivariant, i.e. such that
f (g · m) = g · f (m) for all g ∈ G and all m ∈ M .

4. Let M be a left kG-module. Thus M is a k-module equipped with a

left k-linear action of G "·".

i) Prove that the map:

M × G → M , (m, g) 7→ g −1 · m

defines a right action of G on M , and in fact a right kG-module

structure on M . We let M right denote the k-module M considered
as a right module in this way.
ii) Prove that sending M to M right defines an isomorphism of cate-
gories kG−Mod ' Mod − kG (i.e. a functor which is a bijection
between the objects of the two categories and which is a bijection
on the morphisms.)

Exercice 53. Rings of matrices. Let R be the ring of (2 × 2)-matrices

with coefficients in a field k. Let C1 ⊂ R denote the subset of matrices of
the form [ ∗∗ 00 ], and let C2 denote the subset of matrices of the form [ 00 ∗∗ ]

1
 A. Touzé – 2023

1. Show that C1 and C2 are left R-submodules of the left R-module R.

Are they right R-submodules of the right R-module R ?

2. Show that C1 and C2 are simple and projective as left R-modules.

3. Show that C1 and C2 are isomorphic, and that C1 is, up to isomor-

phism, the only simple left R-module.

4. Show that every finite dimensional left R-module is semi-simple, i.e.

is isomorphic to a direct sum of copies of C1 . (Hint: an easy way to
prove this is by using the Ext-criterion for semi-simplicity)

5. If M is a left R-module, we may define a right action of R on it by

the map:
M × R → M , (m, r) 7→ t r · m
where t refers to the transpose of a matrix. We let M right denote the
right R-module obtained from M in this way. Prove that sending M
to M right defines an isomorphism of categories R−Mod ' Mod − R.

Exercice 54. A ring with different left and right modules. Let R
be the ring of (3 × 3)-matrices of the form

∗ 0 ∗
 

0 ∗ ∗
 
0 0 ∗

with coefficients in a field k. Let Ci ⊂ R denote the subset of R whose

elements are the matrices of R with entries zero in the columns of index 6= i.
Let Ri ⊂ R denote the subset of matrices of R whose entries are zero in the
rows of index different from i.

1. Show that R = C1 ⊕ C2 ⊕ C3 as left R-modules and that R = R1 ⊕

R2 ⊕ R3 as right R-modules.

2. Show that the left R-modules C1 and C2 are simple, non-isomorphic

and projective.

3. Show that every simple projective right R-module is isomorphic to R3 .

(Hint: show that there is up to isomorphism only 3 simple R modules,
which are the simple quotients of the Ri , and show that only one of
them is projective.)

4. Show that there is no isomorphism of categories between R−Mod and

Mod − R.

2
 A. Touzé – 2023

13. Tensor products.

Exercice 55. Two constructions of the same tensor product. Let

R be a k-algebra, let M be a right R-module and let N be a left R-module.
Let T be the abelian group obtained by tensoring M and N over the ring
R and let T 0 denote the k-module obtained by tensoring M and N over the
k-algebra R:
M ⊗Z N M ⊗k N
T = , T0 = .
hmr ⊗ n − m ⊗ rni hmr ⊗ n − m ⊗ rni

1. Show that there exists a unique map α : k×T → T such that α(λ, m⊗
n) = m ⊗ λn, and that this map endows T with the structure of a k-
module.

2. Show that the map π : M × N → T such that π(m, n) = m ⊗ n is

(R, k)-bilinear, and that for all k-modules P and all (R, k)-bilinear
map f : M × N → P there is a unique k-linear map f : T → P such
that f = f ◦ π.
'
3. Deduce that there is a k-linear isomorphism φ : T −
→ T 0 such that
φ(m ⊗ n) = m ⊗ n.

Exercice 56. Extension of scalars. Let R → S be a morphism of rings.

We view S as a right R-module via this morphism.

1. Let M be a left R-module. Show that there exists a unique map

α : S × S ⊗R M → S ⊗R M such that α(s, s0 ⊗ m) = ss0 ⊗ m. Show
that this map defines a left S-module structure on S ⊗R M . Show
moreover that if f : M → M 0 is R-linear then idS ⊗ f is S-linear.

2. Show that there is an isomorphism of abelian groups, natural with

respect to the R-module M and the S-module N :

HomS (S ⊗R M, N ) ' HomR (M, resSR N ) .

3. Show that if M is a projective R-module, then S ⊗R M is a projective

S-module. Show that if M is a flat R-module then S ⊗R M is a flat
S-module.

3
 A. Touzé – 2023

Exercise sheet n°9

Tor.

14. Tensor products and Tor.

Exercice 57. Computation of Tor over Z. Compute TorZi (A, B)

for all i, when A and B belong to the following list of abelian groups:
{finite cyclic groups , Q , Q/Z}.

Exercice 58. Left exactness of Tor over Z. Show that for all abelian
groups A, Tor1 defines a left exact functor:

TorZ1 (A, −) : Z−Mod → Z−Mod .

Exercice 59. Computation of Tor over k[x]. Let k be a commutative

ring. Let n and m be positive integers. Compute Tori (k[x]/xn , k[x]/xm )
k[x]

for all i.

Exercice 60. Computation of Tor over k[x]/xd . Let k be a commu-

tative ring, and let d be a positive integer. Let n and m be two positive
k[x]/xd
integers less or equal to d. Compute Tori (k[x]/xn , k[x]/xm ) for all i.

Exercice 61. Commutativity of Tor. Let R be a k-algebra and let

Rop denote the opposite k-algebra (i.e. the same k-module with product
·op : R × R → R given by r ·op s := sr). Every left R-module N can
be considered as a right Rop -module, and every right R-module M can be
considered as a left Rop -module.

1. Show that the isomorphism M ⊗k N ' N ⊗k M , m ⊗ n 7→ n ⊗ m

induces an isomorphism: M ⊗R N ' N ⊗Rop M .

2. Show that for all i, there is an isomorphism

op
TorR
i (M, N ) ' Tori
R
(N, M ) .

Exercice 62. Flatness versus torsion-freeness.

1. Let R be a domain. Show that a module M is flat, then M is torsion-

free (i.e. 0 is its only torsion element).

1
 A. Touzé – 2023

2. Assume now that R is a PID.

(a) Show that every finitely generated torsion-free module is flat.

(b) Let M be a torsion free-module, let N be a module with free
d
resolution P1 −
→ P0 . Let
id⊗d
x ∈ TorR
1 (M, N ) = Ker (M ⊗ P1 −
−−→ M ⊗ P0 ).

i. Show that there is a submodule S ⊂ M such that x is in the

image of the map id ⊗ incl : S ⊗ P1 → M ⊗ P1 .
ii. In the following commutative square, show that the maps (∗)
and (∗∗) are injective.

incl⊗id
S ⊗R P1 M ⊗R P1
(∗) id⊗d id⊗d .
incl⊗id
S ⊗R P0 M ⊗R P0
(∗∗)

iii. Deduce that x = 0.

(c) Show that over a PID, being flat is equivalent to being torsion-
free.

3. Take R = k[x, y] and let I be the ideal generated by x and y.

k[x,y]
(a) Show that Tor2 (k, k) 6= 0. (Hint: you may use the projective
resolution of k given in exercise 35 in the exercise sheet 6)
(b) Use the short exact sequence 0 → I → R → k → 0 to deduce
that I is not flat (although I is torsion-free).

2
 A. Touzé – 2023

Exercise sheet n°10

Homology of groups

15. Homology of groups

Exercice 63. Independence from the ground ring k. Let k be a

commutative ring.
1. Let C be a complex of projective k-modules, such that Ci = 0 for
i < 0.
(a) Prove that Hi (C) = 0 for all i if and only if the identity morphism
idC : C → C is homotopic to the zero morphism.
(b) Assume that Hi (C) = 0 for all i. Prove that for all k-modules
M , the homology of the complex C ⊗k M is zero in all degrees.
2. Let V be a k-linear representation of a group G. Then we can forget
the action of k, and see V as a Z-linear representation of G.
(a) Show that for all free ZG-modules F there is an an isomorphism
of abelian groups, natural with respect to F :
F ⊗ZG V ' (F ⊗Z k) ⊗kG V .
(b) Show that there is an isomorphism of abelian groups:
TorZG
i (Z
triv
, V ) ' TorkG
i (k
triv
,V ) .
[Hint: consider first a free resolution (P, ) of Ztriv , and show
that (P ⊗Z k,  ⊗ idk ) is a free resolution of ktriv . Use the latter
to compute TorkGi (k
triv , V ).]

Exercice 64. An exact sequence involving tensor products. Let k

be a commutative ring, let V and W be two k-modules and let I ⊂ V and
J ⊂ W be two submodules.
1. Let B denote the bicomplex whose rows are given by the Tor-exact
sequences, and whose vertical morphisms are given by functoriality
with respect to the canonical inclusion ι : J ,→ W :
∂ ∂
... Tork1 (I, J) Tork1 (V, J) Tork1 (V /I, J) I ⊗J V ⊗J 0

∂ ∂
... Tork1 (I, W ) Tork1 (V, W ) Tork1 (V /I, W ) I ⊗W V ⊗W 0
id⊗ι
Show that TotB is quasi-isomorphic to the complex V /I ⊗ J −−−→
V /I ⊗ W .

1
 A. Touzé – 2023

2. Show that H0 (TotB) ' V /I ⊗ W/J and deduce that there is a short
exact sequence, natural with respect to V and W

V ⊗ J ⊕ I ⊗ W → V ⊗ W → V /I ⊗ W/J → 0 .

Exercice 65. The Künneth theorem. If V is a k-linear representation

of G and W is a k-linear representation of H, then V ⊗k W is a k-linear
representation of G × H, with action given by (g, h) · (v ⊗ w) := (gv, hw).

1. Show that we have an isomorphism (V ⊗ W )G×H ' VG ⊗ WH natural

with respect to V and W .
[Hint: use the previous exercise, with I = hgv − v | (g, v) ∈ G × V i
and J = hhw − w | (h, w) ∈ H × W i.]

2. Assume that k is a field. Show that

Hi (G × H, V ⊗ W ) ' Hk (G, V ) ⊗ H` (H, W ) .

M

k+`=i

Exercice 66. Morphisms between cyclic groups. Let f : Cqr → Cq

be a surjection of cyclic groups, which sends the generator g of Cqr to the
generator g 0 of Cq . Compute the map fi : Hi (Cqr , k) → Hi (Cq , k) induced
in homology by f .

Exercice 67. Homology of dihedral groups. Let D2q = Cq oC2 denote

the dihedral group with 2q elements, q odd. Use the Cartan-Eilenberg stable
element formula to compute the homology of D2q with coefficients in Fp , p
odd.

Exercice 68. The universal coefficient theorem. Show that for all i
there is a short exact sequence natural with respect to G:

0 → Hi (G, Z) ⊗Z k → Hi (G, k) → TorZ1 (Hi−1 (G, Z), k) → 0

which splits (not naturally with resepct to G). Deduce that H1 (G, k) '
Gab ⊗Z k.