Basics on homotopy theory

Estanislao Herscovich

1 Conventions 1

2 Basic topological definitions I 1

3 Basic topological definitions II: Connectedness 3

4 Basic topological definitions III: Path connectedness 4

5 Algebraic interlude I: Basics on category theory 5

6 Homotopy I: The basics 8

7 Homotopy II: The fundamental group 10

8 Algebraic interlude II: The category of transitive G-sets 14

9 Covering spaces I: Basic properties 16

10 Covering spaces II: Automorphisms of coverings 21

11 Covering spaces III: Simply connectedness 23

12 Covering spaces IV: Classification of coverings 25

13 Algebraic interlude III: Free amalgamated products 27

14 Homotopy III: The Seifert-Van Kampen theorem 28

15 Surfaces I: General theory 30

16 Surfaces II: Homotopy groups 33

References 35

Index 36
 §1. Conventions

We will denote by N = {1, 2, . . .} the set of (strictly) positive integers, N0 = {0, 1, 2, . . .} the set of nonneg-
ative integers, Q the set of rational numbers and R the set of all real numbers. Given two integers m ≤ n,
we define Jm, nK = { x ∈ Z : m ≤ x ≤ n}. Given two real numbers a ≤ b, [ a, b] = {t ∈ R : a ≤ t ≤ b} is
the usual closed interval, whereas if a < b, ] a, b[ = {t ∈ R : a < t < b} is the usual open interval. The
half-open intervals [ a, b[ and ] a, b] are defined analogously. More generally, for n ∈ N, given x0 ∈ Rn
and r > 0, Br ( x0 ) = { x ∈ Rn : k x − x0 k < r } is the open ball and B̄r ( x0 ) = { x ∈ Rn : k x − x0 k ≤ r } is the
closed ball (for the Euclidean norm). Given x, y ∈ Rn , we denote by [ x, y] = {tx + (1 − t)y : t ∈ [0, 1]} the
segment they form. Recall also that we set R0 = {∗}.
If I is a set, #( I ) denotes its cardinal. For a map f : X → Y, the image of f will be denoted by Img( f ).
Moreover, given a set X, idX : X → X will usually denote the identity on X. If A is a ring, A× denotes the
group of invertible elements of A.

§2. Basic topological definitions I

2.1. We assume that the reader is familiar with the basic notions of topology, such as topological space,
continuous maps, etc. We refer the reader to [Kel1975] or [Mun2000] for the basics on set topology.
We will review however some of the basic notions that we will need in the sequel, in particular those
for which we use a slightly different definition. We will embark afterwards in our study of homotopy
theory, in particular of the fundamental group, making use of covering spaces. Our exposition will be as
self-contained as possible, but some of the results are left as exercises, many of which will be discussed in
class. Our exposition is greatly influenced by the nice presentation in [Wed2016]. For further details, see
[GH1981] and [Mun1984].
2.2. We will sometimes denote a topological space ( X, T ) simply by X if the topology is clear. Topological
spaces are not necessarily Hausdorff, unless it is explicitly stated. Moreover, to avoid confusion we remark
that a neighborhood of a point x in a topological space is a subset Y ⊆ X such that there is an open set
U ⊆ X satisfying that x ∈ U ⊆ Y. In particular, a neighborhood of a point is not necessarily open. We also
recall that a map (regardless of continuity) f : X → Y between topological spaces is called closed (resp.,
open) if f ( A) ⊆ Y is closed (resp., open) for every closed (resp., open) subset A ⊆ X.
2.3. Exercise. Let X and Y be topological spaces and let f : X → Y be a continuous map. Let {Yi }i∈ I be a family
of subsets of Y such that one of the following two conditions is satisfied:
(i) ∪i∈ I Yi◦ = Y, where Yi◦ denotes the interior of Yi in Y;
(ii) ∪i∈ I Yi = Y, Yi is closed for all i ∈ I, and for all y ∈ Y there exist an open set V ⊆ Y and a finite subset
J ⊆ I such that y ∈ V and V ∩ Yi = ∅ for i ∈ J.
We recall that a covering of a topological space satisfying the last condition of (ii) is said to be locally finite. Prove
that f is closed (resp., open) if and only if the map f | f −1 (Yi ) : f −1 (Yi ) → Yi is closed (resp., open) for all i ∈ I.
2.4. Recall that a subset A ⊆ X of a topological space is called relatively Hausdorff if given x, y ∈ A,
there exist open sets U, V ⊆ X such that U ∩ V = ∅, x ∈ U and y ∈ V. Recall that a continuous map
f : X → Y between topological spaces is called separated if the induced map

∆ f : X → X ×Y X = {( x, x 0 ) ∈ X × X : f ( x ) = f ( x 0 )}

1
sending x ∈ X to ( x, x ) is closed, where X ×Y X has the subspace topology of X × X. Using the continuous
map X ×Y X → X given by ( x, x 0 ) 7→ x, we see that f is separated if and only if Img(∆ f ) ⊆ X ×Y X is a
closed subset. We will also denote this image by ∆ f if there is no possible confusion.
2.5. Exercise. Let f : X → Y be a continuous map of topological spaces. Prove that the following condition are
equivalent:
(i) f is separated;
(ii) f −1 ( A) ⊆ X is relatively Hausdorff for every relatively Hausdorff A ⊆ Y;
(iii) f −1 ({y}) ⊆ X is relatively Hausdorff for every y ∈ Y.
2.6. Using the previous exercise we see easily that, given f : X → Y and g : Y → Z two continuous maps
of topological spaces, then g ◦ f is separated if f and g are so, and f is separated if g ◦ f is so.
2.7. Recall that a continuous map f : X → Y of topological spaces is called an embedding if the induced
map f : X → Img( f ) is a homeomorphism, where the image of f is endowed with the subspace topology.
2.8. Exercise. Let f : X → Y be a continuous map of topological spaces. Assume that there exists a continuous map
g : Y → X such that g ◦ f = idX . Prove that f is an embedding.
2.9. Recall that a topological space X compact if given any open covering {Ui }i∈ I of X, i.e. ∪i∈ I Ui = X,
there exists a finite subset I0 ⊆ I such that ∪i∈ I0 Ui = X.
2.10. Exercise. Let X and Y be topological spaces, and let A ⊆ X and B ⊆ Y be subspaces.
(i) Prove that if A and B are compact subspaces and W ⊇ A × B is an open subset of X × Y, there exist open
sets U ⊆ X and V ⊆ Y such that A ⊆ U, B ⊆ V and U × V ⊆ W.
(ii) Prove that X is compact if and only if the projection map X × Y → Y is closed for all topological spaces Y.
Hint. For the converse implication, suppose that X is not compact, pick a covering {Ui }i∈ I of X such that
∪i∈ I 0 Ui 6= X for every finite subset I 0 ⊆ I, consider the space I = { I 0 ⊆ I : I 0 6= ∅ is finite} ∪ { I }
endowed with the topology generated by the family {U J : J ⊆ I such that J is finite } where

U J = { I 0 ∈ I : I 0 ⊇ J },

as well as the set A ⊆ X × I given by

A = {( x, I 0 ) ∈ X × I : x ∈
[
/ Ui }.
i∈ I 0

2.11. Recall that a continuous map f : X → Y between topological spaces is called proper if given any
continuous map g : Z → Y of topological spaces, the projection X ×Y Z → Z on the second factor is
closed, where X ×Y Z = {( x, z) ∈ X × Z : f ( x ) = g(z)} is endowed with the subspace topology of X × Z.
2.12. Exercise. Let f : X → Y be a continuous map of topological spaces. Prove that the following condition are
equivalent:
(i) f is proper;
(ii) f is closed and f −1 ({y}) ⊆ X is a compact subspace for every y ∈ Y;
(iii) f is closed and f −1 ( A) ⊆ X is a compact subspace for every compact subspace A ⊆ Y;
(iv) the map f × idZ : X × Z → Y × Z is closed, for every topological space Z.
Hint. Prove (iv) ⇒ (iii) ⇒ (ii) ⇒ (i) ⇒ (iv). For the third implication, consider A ⊆ X ×Y Z
closed and z0 ∈ Z \ π2 ( A), where π2 : X ×Y Z → Z is the canonical projection. After considering the
simple case z0 ∈ / Img(π2 ) = g−1 ( f ( X )), suppose the opposite and construct open subsets U ⊆ X and
V ⊆ Y such that (U × V ) ∩ A = ∅ by means of Exercise 2.10, (i). Take finally the open neighborhood
V ∩ g−1 (Y \ f ( X \ U )) ⊆ Z \ π2 ( A) of z0 .

2
2.13. Using the previous exercise we see easily that, given f : X → Y and g : Y → Z two continuous maps
of topological spaces, then g ◦ f is proper if f and g are so.
2.14. Exercise. Prove that any closed embedding is proper.
2.15. Exercise. Let f : X → Y and g : Y → Z be two continuous maps of topological spaces.
(i) Define Γ f : X → X × Z Y as the continuous map sending x ∈ X to ( x, f ( x )), where

X × Z Y = ( x, y) ∈ X × Y : ( g ◦ f )( x ) = g(y)

is endowed with the subspace topology. Use the canonical projection X × Z Y → X to prove that Γ f is an
embedding.
(ii) Show that Img(Γ f ) = ( f × idY )−1 (Img(∆ g )). Deduce that Γ f is a closed map if g is separated.
(iii) Show that f is the composition of Γ f and the canonical projection X × Z Y → Y. Deduce that f is proper if
g ◦ f is so and g is separated.
(iv) Prove that g is proper if g ◦ f is so and f is surjective.

§3. Basic topological definitions II: Connectedness

3.1. We recall that a topological space X is said to be connected if X 6= ∅ and for every pair U, V ⊆ X of
open sets of X such that U ∩ V = ∅, X = U ∪ V, either U = ∅ or V = ∅. 1
3.2. Exercise. Prove that, for every pair of real numbers a ≤ b, the interval [ a, b] ⊆ R is connected, for the subspace
topology of the usual topology of R. Unless otherwise stated, the topology of Rn , for n ∈ N will always be the usual
one (i.e., the one induced by any norm), and the topology of any of the intervals of R or balls of Rn will always be
the subspace topology.
3.3. Exercise. (i) Prove that, given a topological space X and a family {Yi }i∈ I of connected subspaces of X (i.e.
Yi ⊆ X is a connected topological space for the subspace topology) such that ∩i∈ I Yi 6= ∅, then ∪i∈ I Yi is a
connected subspace of X.
(ii) Prove that, if f : X → Y is a surjective continuous map between topological spaces X and Y, and X is
connected, then Y is connected as well.
3.4. Given a topological space X and a point x ∈ X, we define the (connected) component of x as the
union of all the connected subspaces of X containing x. This is a connected subspace of X, by item (i) of
the previous exercise, that we will typically denote by Cx .
3.5. Exercise. (i) Let A ⊆ X be a connected subspace of a topological space X. Prove that Ā ⊆ X is also
connected.
(ii) Prove that Cx is closed for a topological space X and a point x ∈ X.
(iii) Consider the topological space Q, with the subspace topology of R. Prove that the connected component of
any x ∈ Q is { x }, which is not open in Q.
1 Thisis the first deviation we take: our definition of connected space explicitly excludes the case
when it is empty, in contrast to some classical textbooks, such as [Kel1975]. Since the case of an empty
topological space is somehow pathological, for it forces us to explicitly exclude it from many equivalent
conditions to that of being connected as well as many properties satisfied by them, we have to decided to
exclude it from the definition.

3
3.6. Exercise. A nonempty topological space X is called irreducible (or hyperconnected) if given any pair of
nonempty open subsets U, V ⊆ X, U ∩ V 6= ∅.
(i) Prove that every irreducible space is connected.
(ii) Prove that a nonempty topological space is irreducible if and only if every nonempty subset is nowhere dense
(i.e. the interior of its closure is empty), if and only if every nonempty open subset is dense.
3.7. A topological space X is said to be locally connected if for every x ∈ X there is a fundamental system
of connected neighborhoods of x, i.e. for every x ∈ X and every open set V ⊆ X including x there exists a
connected subspace W of X and an open set U such that x ∈ U ⊆ W ⊆ V.
3.8. Exercise. Let
   
{1/2n } × [0, 1] ⊆ R2 .
 [
X = (0, 1) ∪ [0, 1] × {0} ∪
n∈N0

We give X the subspace topology of R2 . Prove that X is connected but it is not locally connected. Give an example of
a locally connected topological space that is not connected.

§4. Basic topological definitions III: Path connectedness

4.1. Given a topological space X, a (continuous) path is a continuous map γ : [0, 1] → X. The point γ(0)
is called the initial (or starting) point of γ and γ(1) is called the end point of γ. Moreover, we say that γ
is a loop (based at γ(0)) if γ(0) = γ(1).
4.2. Let γ, η : [0, 1] → X be two paths in X such that γ(1) = η (0). These paths are called composable.
Define γ− : [0, 1] → X by γ− (t) = γ(1 − t), for t ∈ [0, 1], and γ · η : [0, 1] → X by
(
γ(2t), if t ∈ [0, 1/2],
(γ · η )(t) =
η (2t − 1), if t ∈ [1/2, 1].

It is clear that γ− and γ · η are paths in X. They are called the inverse path of γ and the concatenation
(or product) of γ and η, respectively.
4.3. Remark. Note that the order of the paths in the definition of the concatenation does not follow the order used
for the composition of functions. This is however the most common convention in the literature, so we decided to
stick to it.
4.4. Define the following relation on a topological space X: given x, y ∈ X we say that x ∼ y if there is a
path in X with starting point x and end point y. It is easy to see that this is an equivalence relation, so it
defines the set π0 ( X ) formed by all equivalence classes. An element of π0 ( X ) (i.e. an equivalence class of
X under the relation ∼) is called a path connected component of X. We will say that X is path connected
if π0 ( X ) has cardinal 1. Using Exercise 3.2, one sees that any path connected space is connected.
4.5. Remark. Note that a path connected space is a fortiori nonempty.
4.6. Exercise. Let
n o
A = x, sin(1/x ) : x ∈ ]0, π ] ⊆ R2

be a subset of R2 , provided with the usual topology. Prove that A is path connected (so a fortiori connected) but Ā is
not path connected. Note however that Ā is connected by Exercise 3.5, (i).

4
4.7. A topological space X is said to be locally path connected if for every x ∈ X there is a fundamental
system of path connected neighborhoods of x.
4.8. Exercise. Let
     
n
{1/2 } × [0, 1] ⊆ R2 .
[
X = {0} × [0, 1] ∪ [0, 1] × {0} ∪
n∈N0

We give X the subspace topology of R2 . Prove that X is path connected but it is not locally path connected. Give an
example of a locally path connected space that is not path connected.
4.9. Proposition. Let X be a topological space such that every point x ∈ X has a path connected neighborhood.
Then, a subspace Y ⊆ X is a connected component of X if and only if it is a path connected component of X.
Moreover, all the (path) connected components of X are open (and closed).
Proof. Since connected and path connected components are nonempty, we assume without loss of
generality that X 6= ∅. Let x ∈ X be any element, and let Cx (resp., PCx ) denote the (resp., path) connected
component of x. We first show that PCx is open. Let Vx ⊆ X be a path connected neighborhood of x.
Since Vx is path connected, we see that PCx ∪ Vx is also path connected, so by definition of path connected
component we obtain that PCx ∪ Vx = PCx , which in turn implies that the latter is open.
Since any path connected set is connected, we see that PCy ⊆ Cx for all y ∈ Cx . Consider the
canonical projection p : Cx → π0 ( X ) sending y to PCy . By the axiom of choice there is an injective
map s : Img( p) → Cx such that p ◦ s = idImg( p) . Let A x = Img(s). Then, Cx = ty∈ Ax PCy . Since Cx is
component, A x has cardinal one. Indeed, if #( A) ≥ 2, Cx = PCy t (∪y0 ∈ A\{y} PCy0 ) gives a contradiction.
Hence, Cx is path connected, and the inclusion PCx ⊆ Cx tells us that PCx = Cx , as was to be shown.
Since the connected components are closed (by Exercise 3.5, (ii)), the (path) connected components are
also closed.
4.10. Corollary. Let X be a locally path connected topological space. Then, the path connected components coincide
with the connected components, which are open and closed. Moreover, every point has a fundamental system of open
connected neighborhoods.
Proof. The first part is a direct consequence of the previous result. For the last part, it suffices to prove that
given any open neighborhood W of a point x ∈ X, there is an open connected set V such that x ∈ V ⊆ W.
Then, the subspace W of X satisfies the hypotheses of the previous proposition, which in turn implies that
W = ti∈ I Pi is the disjoint union of its path connected components Pi , which are path connected open
subsets of W, so a fortiori path connected open subsets of X. Since x ∈ W = ti∈ I Pi , there is i0 ∈ I such
that x ∈ Pi0 ⊆ W. Taking V = Pi0 This proves the corollary.

§5. Algebraic interlude I: Basics on category theory

We present the following definition, which provides a very useful general language to deal with the
different situations we will encounter. For a more detailed and comprehensive exposition, see [Mac1971].
For a discussion on how to safely frame category theory within set theory (or first order logic) see
[Mur2006] and the references therein.
5.1. Definition. A category C is a tuple (C0 , C1 , sC , tC , ◦C , iC ) where
(i) C0 is a collection of elements, called the objects of C and typically denoted by X, Y, Z, . . . ;
(ii) C1 is a collection of elements, called the morphisms of C and typically denoted by f , g, h, . . . ;

5
 (iii) sC , tC : C1 → C0 are two maps, called the source map and the target map, and whose inverse image
−1 −1
sC ({ X }) ∩ tC ({Y })

is denoted by HomC ( X, Y ) and is called the space of morphisms from X to Y;

(iv) ◦C : C1 ×s,t C1 → C1 is a map called the composition, sending a pair ( f , g) to f ◦C g, where

C1 ×s,t C1 = ( f , g) ∈ C1 × C1 : sC ( f ) = tC ( g) ;


(v) iC : C0 → C1 is a map, called the identity, sending every element X to a distinguished morphism
iC ( X ) = idX , called the identity morphism;
satisfying that
(CAT.1) given f , g, h ∈ C1 such that sC ( f ) = tC ( g) and sC ( g) = tC (h), then

( f ◦C g ) ◦C h = f ◦C ( g ◦C h ).

(CAT.2) For every X ∈ C0 and every f , g ∈ C1 such that sC ( f ) = X = tC ( g), we have

f ◦C idX = f and idX ◦C g = g.

5.2. Remark. More intuitively, a category can be simply defined as a collection of objects C0 and for every
pair of objects ( X, Y ) ∈ C02 a set HomC ( X, Y ), called the space of morphisms from X to Y, satisfying that
HomC ( X, Y ) = HomC ( X 0 , Y 0 ) if and only if X = X 0 and Y = Y 0 , as well as maps

◦ X,Y,Z = ◦C |HomC (Y,Z)×HomC (X,Y ) : HomC (Y, Z ) × HomC ( X, Y ) → HomC ( X, Z )

for every triple of objects ( X, Y, Z ) ∈ C03 , and a distinguished morphism idX ∈ HomC ( X, X ) for every object
X ∈ C0 satisfying the axioms (CAT.1) and (CAT.2). If the category is clear, we will denote its composition ◦C
simply by ◦.
5.3. Exercise. (i) Let C0 be the collection of all sets, HomC ( X, Y ) the set of all maps from X to Y, ◦C the
usual composition of maps and idX is the identity map from X to itself. Prove that this forms a category,
denoted by Set.
(ii) Let C0 be the collection of all groups, HomC ( X, Y ) the set of all morphisms of groups from X to Y, ◦C the
usual composition of morphisms of groups and idX is the identity map from X to itself. Prove that this forms
a category, denoted by Grp.
(iii) Let C0 be the collection of all topological spaces, HomC ( X, Y ) the set of all continuous maps from X to Y,
◦C the usual composition of maps and idX is the identity map from X to itself. Prove that this forms a
category, denoted by Top.
(iv) Let G be a group and C0 be the collection of all sets provided with a right action of G, HomC ( X, Y ) the
set of all G-linear (or G-equivariant) maps from X to Y, ◦C the usual composition of maps and idX is the
identity map from X to itself (see Section 8 for more details). Prove that this forms a category, denoted by
Set-G.
5.4. Let C be a category. A morphism f ∈ HomC ( X, Y ) is called an isomorphism if there exists a
morphism g ∈ HomC (Y, X ) such that g ◦ f = idX and f ◦ g = idY . In this case, we also say that X and
Y are isomorphic objects, and we will write it X ∼ = Y. The set of isomorphisms from X to Y is denoted
by IsoC ( X, Y ), and we define AutC ( X, Y ) as the group IsoC ( X, X ) under the composition ◦C , which is
called the group of automorphisms of X.
5.5. Exercise. Prove that there exists at most one morphism g ∈ HomC (Y, X ) satisfying the previous condition.
This morphism, if it exists, is called the inverse of f .

6
5.6. Exercise. Prove that a morphism in Set or in Grp is an isomorphism if and only if it is bijective. Show that
this does not hold in Top.
5.7. Definition. Given two categories C and D, a functor F from C to D, usually written as F : C → D, is a
pair of maps ( F0 , F1 ) of the form F0 : C0 → D0 and F1 : C1 → D1 such that tD ◦ F1 = F0 ◦ tC , sD ◦ F1 = F0 ◦ sC ,
F1 ◦ iC = iD and F1 ◦ (◦C ) = (◦D ) ◦ F.
5.8. Remark. More intuitively, a functor F : C → D is given by a map F0 : C0 → D0 together with maps

FX,Y = F1 |HomC ( X,Y ) : HomC ( X, Y ) → HomD ( F0 ( X ), F0 (Y ))

for all pairs ( X, Y ) ∈ C02 satisfying that

FX,X (idX ) = idF0 ( X ) and FX,Z ( f ◦ g) = FY,Z ( f ) ◦ FX,Y ( g),

for all X, Y, Z ∈ C0 and for all morphisms f ∈ HomC (Y, Z ) and g ∈ HomC ( X, Y ). By abuse of notation, and if
it is clear from the context, it is usual to omit the subscripts and to denote both F0 and F1 simply by F.
5.9. Exercise. Let C be a category, and consider F = ( F0 , F1 ) : C → C given as follows. For every object X ∈ C0 ,
set F ( X ) = X, and for every morphism f : X → Y in C1 between the objects X and Y, define F1 ( f ) = f . Prove
that this defines a functor, called the identity functor.
5.10. Exercise. Let F = ( F0 , F1 ) : Top → Set be defined as follows. For every topological space X, set
F ( X ) = π0 ( X ), and for every continuous map f : X → Y between topological spaces X and Y, define
F1 ( f ) : π0 ( X ) → π0 (Y ) as the map sending PCx to PC f ( x) . Prove that this defines a functor, called π0 .
5.11. Exercise. Let F = ( F0 , F1 ) : C → D be a functor between two categories C and D, and let f ∈ C1 be an
isomorphism. Prove that F1 ( f ) is an isomorphism.
5.12. A functor F = ( F0 , F1 ) : C → D is an equivalence if FX,Y : HomC ( X, Y ) → HomD ( F0 ( X ), F0 (Y )) is
bijective for all objects X and Y in C0 , and for every object Z in D, there is an object X in C0 such that
F0 ( X ) ∼
= Z. If a functor satisfies the first part of the definition it is called fully faithful, whereas a functor
satisfies the second part it is called dense.
5.13. Given functors F = ( F0 , F1 ) : C → D and G = ( G0 , G1 ) : D → E , it is easy to see that ( G0 ◦ F0 , G1 ◦ F1 )
is a functor from C to E , called the composition functor, and it is denoted by G ◦ F.
5.14. Given two functors F = ( F0 , F1 ) : C → D and G = ( G0 , G1 ) : C → D, a natural transforma-
tion from F to G is a map h : C0 → D1 such that sD ◦ h = F0 , tD ◦ h = G0 , and for every f ∈ C1 ,
G1 ( f ) ◦ h(sC ( f )) = h(tC ( f )) ◦ F1 ( f ). Equivalently, a natural transformation from F to G is a collection of
morphisms {h X = h( X ) ∈ HomD ( F0 ( X ), G0 ( X )) : X ∈ C0 } such that G1 ( f ) ◦ h X = hY ◦ F1 ( f ), for every
morphism f ∈ HomC ( X, Y ), which can be represented graphically as the commutation of the following
diagram

hX
F0 ( X ) G0 ( X )
F1 ( f ) G1 ( f )
hY
F0 (Y ) G0 (Y )

A natural isomorphism is a natural transformation h from F to G such that h( X ) is an isomorphism in D,

for all X ∈ C0 . We say in this case that F and G are naturally isomorphic.
5.15. Exercise. Let F = ( F0 , F1 ) : C → D be a functor between two categories C and D. Prove that F is an
equivalence if and only if there exists a functor G = ( G0 , G1 ) : D → C such that F ◦ G is naturally isomorphic to
the identity functor of D and G ◦ F is naturally isomorphic to the identity functor of C . The functor G is called a
quasi-inverse.

7
5.16. Given a category C , a subset S ⊆ C0 is called a set of isomorphic classes of objects of C if for every
Y ∈ C0 there is X ∈ S such that X ∼ = Y, and given X, X 0 ∈ S, X ∼= S0 implies that S = S0 .
5.17. Exercise. Let C be a category with a set S of isomorphic classes of objects.
(i) Assume that S0 ⊆ C0 is another set of isomorphic classes of objects of C . Prove that S and S0 are in bijection.
(ii) and let F = ( F0 , F1 ) : C → D be an equivalence between two categories C and D. Prove that the set
{ F ( X ) : X ∈ S} ⊆ D0 is a set of isomorphic classes of objects of D.

§6. Homotopy I: The basics

6.1. Let f , g : X → Y be two continuous maps between topological spaces and let A ⊆ X be a subset. We
say that f is homotopic to g relative to A, a we will denote it by f ' g (rel A), 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, and H ( x, t) = f ( x ) = g( x ),
for all x ∈ A and all t ∈ [0, 1]. If A = ∅ we will simply say that homotopic to g, and we will denote it by
f ' g.
6.2. Remark. Note that f ' g (rel A) implies that f ( x ) = g( x ), for all x ∈ A.
6.3. Lemma. Let X, Y and Z be topological spaces, and A ⊆ X be a subset.
(i) Consider the relation on the set HomTop ( X, Y ) of continuous maps from X to Y given by f ' g (rel A)
defined before for f , g : X → Y. This is an equivalence relation.
(ii) Given continuous maps f , g : X → Y and f 0 , g0 : Y → Z such that f ' g (rel A) and f 0 ' g0 (rel f ( A)),
then f 0 ◦ f ' g0 ◦ g (rel A).
Proof. (i) It is clear that f ' f (rel A), for all f ∈ HomTop ( X, Y ), using the homotopy H ( x, t) = f ( x ),
for all x ∈ X and t ∈ [0, 1]. On the other hand, if f ' g (rel A) by means of the homotopy
H : X × [0, 1] → Y, then G : X × [0, 1] → Y given by G ( x, t) = H ( x, 1 − t), for all x ∈ X and
t ∈ [0, 1], is a homotopy from g to f relative to A, i.e. g ' f (rel A). Finally, if f , g, h ∈ HomTop ( X, Y )
satisfy that f ' g (rel A) and g ' h (rel A) by means of homotopies H : X × [0, 1] → Y and
G : X × [0, 1] → Y, respectively, then the map K : X × [0, 1] → Y given by
(
H ( x, 2t), if t ∈ [0, 1/2],
K ( x, t) =
G ( x, 2t − 1), if t ∈ [1/2, 1],

is a homotopy from f to h relative to A.

(ii) Let H : X × [0, 1] → Y be a homotopy from f to g relative to A and G : Y × [0, 1] → Z be a homotopy
from f 0 to g0 relative to f ( A). Define the map K : X × [0, 1] → Z by K ( x, t) = G ( H ( x, t), t), for all
x ∈ X and t ∈ [0, 1]. This gives a homotopy from f 0 ◦ f to g0 ◦ g relative to A, as was to be shown.

6.4. Exercise. Let U ⊆ Rn for n ∈ N, provided with the subspace topology, A ⊆ U and f , g : X → U be two
continuous functions defined on a topological space X such that f ( x ) = g( x ) for all x ∈ A and that

[ f ( x ), g( x )] = {t f ( x ) + (1 − t) g( x ) : t ∈ [0, 1]} ⊆ U

for all x ∈ X. Prove that f ' g relative to A.

8
6.5. Exercise. (i) Let C0 be the collection of all topological spaces, HomC ( X, Y ) the set of all equivalence
classes of continuous maps from X to Y under the equivalence relation given by homotopy, and idX is the
equivalence class of the identity map from X to itself. Prove that usual composition of maps induces a map

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

for all topological spaces X, Y, Z. Moreover, prove that the previous data forms a category, denoted by hTop,
that is called the (naive) homotopy category.
(ii) Prove that the maps ( Q0 , Q1 ) : Top → hTop given by Q0 ( X ) = X for all topological spaces X and such
that

Q X,Y : HomTop ( X, Y ) → HomhTop ( X, Y )

sends any continuous map f : X → Y to the unique equivalence class under the homotopy equivalence
relation including f , form a functor, called the quotient functor and denoted by Q.
(iii) Prove that there exists a unique functor π̄0 : hTop → Set such that π̄0 ◦ Q = π0 , where π0 is the functor
defined in Exercise 5.10. By abuse of notation, we will denote the functor π̄0 also by π0 .
6.6. Let f : X → Y be a continuous map between topological spaces. We say that f is null homotopic if
there exists y0 ∈ Y such that f is homotopic to the constant map cy0 : X → Y that sends every x ∈ X to y0 .
On the other hand, we say that f is a homotopy equivalence if there exists a continuous map g : Y → X
such that f ◦ g ' idY and g ◦ f ' idY . This is tantamount to say the image of f under the quotient functor
Q : Top → hTop defined before is an isomorphism.
6.7. Exercise. Let n ∈ N and Sn ⊆ Rn+1 be the unit sphere. Prove that the inclusion Sn → Rn+1 \ {0Rn+1 } is a
homotopy equivalence that is not a homeomorphism.
6.8. We say that a topological space X is contractible if X is contractible if the identity map idX is null
homotopic. Equivalently, X is contractible if it is isomorphic to the singleton space {∗} in the category
hTop. A topological space X is said to be locally contractible if every point has a fundamental system of
open contractible neighborhoods.
6.9. Exercise. Recall that a set A ⊆ Rn is called convex (resp., star-shaped) if for every pair of points x, y ∈ A,
tx + (1 − t)y ∈ A for all t ∈ [0, 1] (resp., if there exists x0 ∈ A such that tx0 + (1 − t) x ∈ A for all t ∈ [0, 1] and
x ∈ A). We call x0 a star center of the star-shaped set.
(i) Prove that any convex set is star-shaped, and that any star-shaped set is contractible.
(ii) Prove that X = S1 \ {1} ⊆ R2 , with the subspace topology, is contractible but not star-shaped.
6.10. Exercise (Or why this definition of homotopy is (almost) useless in algebraic geometry). A topological
space X is called sober if for every nonempty irreducible closed subset Y ⊆ X there exists a unique y ∈ X, called
the generic point of Y, such that Y = {y}. Let X be a nonempty topological space having a point x0 ∈ X such
that { x0 } is dense in X. Prove that X is contractible. Deduce that an irreducible and sober topological space is
contractible.
Hint. Consider H : X × [0, 1] → X as H ( x, t) = x0 for x ∈ X and t ∈ ]0, 1] and H ( x, 0) = x for x ∈ X.

9
 §7. Homotopy II: The fundamental group

7.1. Given a point x in a topological space, we will denote by ε x : [0, 1] → X the constant path, i.e.
ε x (t) = x for all t ∈ [0, 1]. Moreover, if two paths γ and η in X satisfy that γ ' η (rel{0, 1}), we will say
that γ is path-homotopic to η (or that there is a homotopy as paths from γ to η).
7.2. Lemma. Let X be a nonempty topological space.
(i) Let γ be a path and set x0 = γ(0) and x1 = γ(1). Then

ε x0 · γ ' γ ' γ · ε x1 (rel{0, 1}).

(ii) Given paths γ1 , γ2 , η1 , η2 in X such that γ2 (0) = γ1 (1), η2 (0) = η1 (1) and γi ' ηi (rel{0, 1}) for
i = 1, 2, we have

γ1 · γ2 ' η1 · η2 (rel{0, 1}).

(iii) Given any path γ in X with x0 = γ(0), γ · γ− ' ε x0 (rel{0, 1}).

(iv) Given paths γ1 , γ2 , γ3 in X such that γi+1 (0) = γi (1), for i = 1, 2, we have

(γ1 · γ2 ) · γ3 ' γ1 · (γ2 · γ3 ) (rel{0, 1}).

Proof. (i) Consider the maps f i : [0, 1] → [0, 1] for i = 1, 2, 3 given by

( (
0, if t ∈ [0, 1/2], 2t, if t ∈ [0, 1/2],
f 1 (t) = f 3 (t) =
2t − 1, if t ∈ [1/2, 1], 1, if t ∈ [1/2, 1],

and f 2 (t) = t, for all t ∈ [0, 1], which are clearly continuous. By Exercise 6.4, the three maps are
homotopic relative to {0, 1}, which in turn implies that the three paths γ ◦ f i are homotopic relative
to {0, 1}, by Lemma 6.3, (ii). Since γ ◦ f 1 = ε x0 · γ, γ ◦ f 2 = γ and γ ◦ f 3 = γ · ε x1 , the claim follows.
(ii) Let Hi : [0, 1] × [0, 1] → X be the homotopy from γi to ηi relative to {0, 1}) for i = 1, 2. Then, the
map H : [0, 1] × [0, 1] → X given by
(
H1 (2s, t), if s ∈ [0, 1/2] and t ∈ [0, 1],
H (s, t) =
H2 (2s − 1, t), if s ∈ [1/2, 1] and t ∈ [0, 1],

gives a homotopy from γ1 · γ2 to η1 · η2 relative to {0, 1}.

(iii) The map H : [0, 1] × [0, 1] → [0, 1] given by
(

γ 2s(1 − t) , if s ∈ [0, 1/2] and t ∈ [0, 1],
H (s, t) =

γ 2(1 − s)(1 − t) , if s ∈ [1/2, 1] and t ∈ [0, 1],

which is clearly continuous, gives a homotopy from γ · γ− to ε x0 relative to {0, 1}.

(iv) Let γ : [0, 1] → X be the continuous path given by

γ1 (3t),

 if t ∈ [0, 1/3],
γ(t) = γ2 (3t − 1), if t ∈ [1/3, 2/3],


γ (3t − 2), if t ∈ [2/3, 1],
3

10
 and consider the maps gi : [0, 1] → [0, 1] for i = 1, 2 given by
( (
4t/3, if t ∈ [0, 1/2], 2t/3, if t ∈ [0, 1/2],
g1 ( t ) = g2 ( t ) =
(2t + 1)/3, if t ∈ [1/2, 1], (4t − 1)/3, if t ∈ [1/2, 1],

which are clearly continuous. It is a simple computation to verify that γ ◦ g1 = (γ1 · γ2 ) · γ3 and
γ ◦ g2 = γ1 · (γ2 · γ3 ). By Exercise 6.4, the maps g1 and g2 are homotopic relative to {0, 1}, which
in turn implies that the paths γ ◦ g1 and γ ◦ g2 are homotopic relative to {0, 1}, by Lemma 6.3, (ii),
and the claim follows.

7.3. Let X be a topological space and x0 ∈ X be a fixed point. Define the space of pointed loops as

Ω( X, x0 ) = γ : [0, 1] → X : γ is continuous and γ(0) = γ(1) = x0 .



Consider the equivalence relation on Ω( X, x0 ) given by γ ' γ0 (rel{0, 1}) and define π1 ( X, x0 ) as the
corresponding set of equivalence classes. By Lemma 7.2, the concatenation of loops gives a group structure
on π1 ( X, x0 ), which is called the fundamental group. We will denote by [γ] ∈ π1 ( X, x0 ) the equivalence
class of a loop γ based at x0 , and the product of two elements in [γ], [γ0 ] ∈ π1 ( X, x0 ) will be denoted by
[ γ ] · [ γ 0 ] = [ γ · γ 0 ].
7.4. Exercise. (i) Let C0 be the collection of all pairs ( X, x ), where X is a nonempty topological spaces and
x ∈ X, HomC (( X, x ), (Y, y)) the set of all continuous maps f from X to Y such that f ( x ) = y, ◦C the
usual composition of maps and idX is the identity map from X to itself. Prove that this forms a category,
denoted by Top• , called the category of pointed topological spaces.
(ii) Let C0 be the collection of all pairs ( X, x ), where X is a nonempty topological spaces and x ∈ X,
HomC (( X, x ), (Y, y)) the set of all equivalence classes of continuous maps f from X to Y such that
f ( x ) = y with respect to homotopy relative to { x }, ◦C the usual composition of maps and idX is the
equivalence class of the identity map from X to itself. Prove that this forms a category, denoted by hTop• .
(iii) Prove that the maps ( F0 , F1 ) : Top• → hTop• given by F0 ( X, x ) = ( X, x ) and such that



FX,Y : HomTop• ( X, x ), (Y, y) → HomhTop• ( X, x ), (Y, y)

sends any continuous map f : X → Y satisfying that f ( x ) = y to the unique equivalence class under the
homotopy equivalence relation relative to { x } including f , form a functor, called the pointed quotient
functor and denoted by Q• .
7.5. Given two pointed topological spaces ( X, x ) and (Y, y), one defines the wedge product ( X, x ) ∧ (Y, y)
as the following pointed topological space. Consider Z = X t Y, endowed with the disjoint union topology.
Let ∼ be the equivalence relation on Z generated by x ∼ y. The underlying topological space X ∧ Y
of ( X, x ) ∧ (Y, y) is Z/ ∼ endowed with the quotient topology. The distinguished point of X ∧ Y is the
equivalence class { x, y}.
7.6. Exercise. Consider the X± = ∂B1 (±1, 0) = S1 ± (1, 0) ⊆ R2 and X = X+ ∪ X− ⊆ R2 be endowed with the
subspace topology. Prove that the pointed topological spaces ( X, 0) and (S1 , 1) ∧ (S1 , 1) are isomorphic.
7.7. Let F = ( F0 , F1 ) : Top• → Grp be defined as follows. For every pointed topological space ( X, x ), set
F ( X ) = π1 ( X, x ), and for every continuous map f : X → Y between topological spaces X and Y, define
F1 ( f ) : π1 ( X, x ) → π1 (Y, f ( x )) as the map sending the equivalence class [γ] of a loop based at x to [ f ◦ γ].
It is easy to see that this defines a functor, called π1 .
7.8. Exercise. Prove that there exists a unique functor π̄1 : hTop• → Grp such that π̄1 ◦ Q• = π1 . By abuse of
notation, we will denote the functor π̄1 also by π1 .
7.9. Lemma. Let X be a path connected topological space and let x, x 0 ∈ X be two different points. Then, there is an
isomorphism of groups π1 ( X, x ) ' π1 ( X, x 0 ).

11
Proof. Let η : [0, 1] → X be a path satisfying that η (0) = x 0 and η (1) = x, which exists by assumption.
Define the map

Adη : π1 ( X, x ) → π1 ( X, x0 )

sending [γ] ∈ π1 ( X, x ), where γ is a loop based at x, to [η · γ · η − ] = [η ] · [γ] · [η − ]. This is well defined by

Lemma 7.2, (ii), and it is clearly a morphism of groups. Moreover, by Lemma 7.2, (iii), it is also a bijection,
since its inverse is the map

Adη − : π1 ( X, x0 ) → π1 ( X, x )

sending [γ0 ] ∈ π1 ( X, x 0 ), where γ0 is a loop based at x 0 , to [η − · γ0 · η ] = [η − ] · [γ0 ] · [η ].

7.10. Exercise. Let X1 and X2 be two topological spaces, and let x1 ∈ X1 and x2 ∈ X2 be two elements.
Consider the canonical projections pi : X1 × X2 → Xi for i = 1, 2. They induce morphisms of groups
π1 ( pi ) : π1 ( X1 × X2 , ( x1 , x2 )) → π1 ( Xi , xi ) for i = 1, 2.
(i) Prove that the map


π 1 X1 × X2 , ( x 1 , x 2 ) → π 1 ( X1 , x 1 ) × π 1 ( X2 , x 2 )

given by [γ] 7→ (π1 ( p1 )([γ]), π1 ( p2 )([γ])) is a morphism of groups.

(ii) Given paths γi : [0, 1] → Xi for i = 1, 2, define the path (γ1 , γ2 ) : [0, 1] → X1 × X2 by means of
(γ1 , γ2 )(t) = (γ1 (t), γ2 (t)), for all t ∈ [0, 1]. Show that the map


π 1 ( X1 , x 1 ) × π 1 ( X2 , x 2 ) → π 1 X1 × X2 , ( x 1 , x 2 )

given by ([γ1 ], [γ2 ]) 7→ [(γ1 , γ2 )] is well defined and it is the inverse of the map in the previous item.
Conclude that they are isomorphisms of groups.
7.11. Exercise. Recall that a topological group is a topological space G together with a group structure given
by product µ : G × G → G that is also a continuous map, such that the map i : G → G sending g to g−1 is
continuous. Let eG denote the unit of G. By the previous exercise, π1 ( G × G, (eG , eG )) ∼
= π1 ( G, eG ) × π1 ( G, eG ),
and π1 (µ) : π1 ( G, eG ) × π1 ( G, eG ) → π1 ( G, eG ) is a morphism of groups. Prove that π1 (µ) coincides with the
product of π1 ( G, eG ), and deduce that that π1 ( G, eG ) is an abelian group.
7.12. A topological space X is called simply connected if it is path connected and π1 ( X, x ) = 1 for all
x ∈ X (or, equivalently, for one x ∈ X). We say that a topological space X is locally simply connected if
every point x ∈ X has a fundamental system of open simply connected neighborhoods.
7.13. Lemma. Let f 1 , f 2 : X → Y be two continuous maps and let H : X × [0, 1] → Y be a homotopy from f 1 to f 2 .
Let x ∈ X, and set yi = f i ( x ), for i = 1, 2. Let η : [0, 1] → Y be the path given by η (t) = H ( x, t), for t ∈ [0, 1].
Then, π1 ( f 1 ) = Adη ◦π1 ( f 2 ), i.e. the following triangle

π1 (Y, y2 )
π1 ( f 2 )

π1 ( X, x ) Adη
π1 ( f 1 )

π1 (Y, y1 )

commutes.

12
Proof. Let γ be a loop based at x. We have to show that ( f 1 ◦ γ) · η ' η · ( f 2 ◦ γ) (rel{0, 1}). Define
h : [0, 1] × [0, 1] → [0, 1] by h(s, t) = H (γ(s), t), for s, t ∈ [0, 1]. Consider the maps hi : [0, 1] → [0, 1] × [0, 1]
for i = 1, 2 given by
( (
(2t, 0), if t ∈ [0, 1/2], (0, 2t), if t ∈ [0, 1/2],
h1 ( t ) = h2 ( t ) =
(1, 2t − 1), if t ∈ [1/2, 1], (2t − 1, 1), if t ∈ [1/2, 1],

which are clearly continuous. It is a simple computation to verify that h ◦ h1 = ( f 1 ◦ γ) · η and

h ◦ h2 = η · ( f 2 ◦ γ). By Exercise 6.4, the maps h1 and h2 are homotopic relative to {0, 1}, which in
turn implies that the paths h ◦ h1 and h ◦ h2 are homotopic relative to {0, 1}, by Lemma 6.3, (ii), and the
lemma follows.
7.14. Corollary. Let X and Y be two nonempty topological spaces, and let x ∈ X. If f : X → Y is a homotopy
equivalence, then π1 ( f ) : π1 ( X, x ) → π1 (Y, f ( x )) is an isomorphism of groups. In particular, X is simply
connected if and only if Y is.
Proof. Let g : Y → X be a continuous map such that g ◦ f ' idX and f ◦ g ' idY . By the previous lemma
applied to g ◦ f and f ◦ g, π1 ( g ◦ f ) = π1 ( g) ◦ π1 ( f ) and π1 ( f ◦ g) = π1 ( f ) ◦ π1 ( g) are bijections, so a
fortiori π1 ( g) ◦ π1 ( f ) is an injection and π1 ( f ) ◦ π1 ( g) is a surjection. This implies that π1 ( f ) is injective
and π1 ( f ) is surjective, respectively, so π1 ( f ) is an isomorphism as claimed. The last part is an immediate
consequence.
7.15. Exercise. Let Y = [−1, 1]× ] − 1, 1[ and ∼ be the equivalence relation on Y generated by (−1, y) ∼ (1, −y),
for all y ∈ ] − 1, 1[. Define M = Y/ ∼ the set of equivalence classes, provide with the quotient topology. It is
usually called the (topological) Möbius band. Let X 0 ⊆ M be the image of [−1, 1] × {0} under the canonical
projection Y → Y/ ∼.
(i) Prove that X 0 is homeomorhic to S1 .
(ii) Prove that the continuous map p : Y → X 0 given by ( x, y) 7→ ( x, 0), for all ( x, y) ∈ Y induces a continuous
map p : M → X 0 such that the composition of p and the inclusion X 0 ⊆ M is homotopic to the identity map
of M. Deduce that π1 (M, 0̄R2 ) ∼
= π1 ( S1 , 1).
7.16. Exercise. (i) Let n ∈ N be a positive integer and U ⊆ Rn be a subset. A path γ : [0, 1] → U is said to be
piecewise affine if there exists a finite set S = {0 = t0 < t1 < · · · < tm = 1} ⊆ [0, 1] such that γ|[ti ,ti+1 ]
is an affine map for all i ∈ J0, m − 1K, i.e.
t − ti
γ(t) = γ(ti ) + (γ(ti+1 ) − γ(ti )) ,
t i +1 − t i

for all t ∈ [ti , ti+1 ] and i ∈ J0, m − 1K. Prove that, given any path γ : [0, 1] → U in an open subset of Rn
there exists a piecewise affine path γ0 : [0, 1] → U such that γ ' γ0 (rel{0, 1}).
Hint. Consider δ = inf {d(γ(t), Rn \ U ) : t ∈ [0, 1]} ∈ R>0 , where d( x0 , A) = inf {k x0 − x k : x ∈ A}
is the usual Euclidean distance between a point x0 ∈ Rn and a subset A ⊆ Rn , and, by uniform continuity
of γ, let m ∈ N be such that kγ(t) − γ(s)k ≤ d/2 if |t − s| ≤ 1/m and t, s ∈ [0, 1]. Define ti = i/m for
i ∈ J0, mK and note that γ(ti+1 ) ∈ Bd (γ(ti )) ⊆ U, for all i ∈ J0, m − 1K.
(ii) Let n ∈ N be a positive integer and x ∈ Rn+1 be a nonzero element. Prove that Rn+1 \ {λx : λ ∈ R≥0 } is
star-shaped with star center − x and in consequence it is contractible. Deduce that Sn \ { x } is contractible
for any x ∈ Sn .
(iii) Let n ≥ 2 be an integer. Prove that given any piecewise affine loop γ : [0, 1] → Rn+1 \ {0Rn+1 } there is a
nonzero x ∈ Rn+1 such that Img(γ) ⊆ Rn+1 \ {λx : λ ∈ R≥0 }. Deduce that Rn+1 \ {0Rn+1 } (and hence
Sn ) is simply connected. One can also prove that Rn+1 \ {0Rn+1 } (and hence Sn ) is not contractible, but it
is outside the scope of these notes.

13
7.17. Since the singleton space {∗} is simply connected, the previous corollary tells us that any contractible
space is simply connected. We have the following implications:

convex star-shaped contractible

simply connected path connected connected

Note that none of the converses hold in general (see Exercises 4.6, 6.9, 7.16, 9.20 and 10.8).

§8. Algebraic interlude II: The category of transitive G-sets

8.1. Let G be a group, whose product we write by a dot or by a simple yuxtaposition, and whose unit we
denote by eG . We will denote its opposite group by Gop , whose underlying set is G and whose product
is given by x ·op x 0 = x 0 · x, for x, x 0 ∈ G. It is clear that G and Gop are isomorphic groups via g 7→ g−1 .
Recall that a right (resp., left) group action of G on an object X in a category C is an morphism of groups
ρ : G → AutC ( X )op (resp., ρ : G → AutC ( X )). The action is called effective if ρ is injective. In the case of
Set, Top, or other concrete categories, we typically write x · g (resp., g · x) instead of ρ( g)( x ), for g ∈ G
and x ∈ X. We will focus in this section on the category Set, but the reader can check that all the results
(and exercises) in this section also hold in other concrete categories, such as Top.
8.2. If X is a set provided with a right (resp., left) action of G is typically called a right (resp., left) G-set.
Given a nonempty right (resp., left) G-set X and x ∈ X we denote by x · G = { x · g : g ∈ G } (resp.,
G · x = { g · x : g ∈ G }) the orbit of x, and by X/G (resp., G \ X) the set of all orbits of all elements of X.
For simplicity, we will mainly deal in this section with right G-sets, and will call them just G-sets for
shortness. This assumption is harmless since any left G-set X is a right G-set via x · g = g−1 · x, for g ∈ G
and x ∈ X, and vice versa. Equivalently, a left G-set is equivalent to a right Gop -set.
8.3. Recall that a nonempty G-set (or the corresponding action) is transitive if there is only one orbit.
Note that any G-set X is a disjoint union of transitive G-sets, the orbits of X. Moreover, if H ≤ G is a
subgroup, then the product of G gives a left action of H on G and H \ G has a right action of G given by
( H · x ) · g = H · ( xg), for x, g ∈ G, which turns H \ G into a transitive G-set.
8.4. Given x in a G-set X, the isotropy group of x is

Gx = { g ∈ G : x · g = x }.

It is clearly a subgroup of G. Moreover, it is clear that

Gx· g = g−1 Gx g, (1)

for all x ∈ X and g ∈ G. We recall that a group action of G on X is called free if the isotropy group Gx of
every x ∈ X is trivial.
8.5. Given two G-sets X and X 0 , a G-equivariant map (or simply a morphism of G-sets) from X to X 0 is
a map f : X → X 0 such that f ( x · g) = f ( x ) · g, for all x ∈ X and g ∈ G. Note that any transitive G-set X
is isomorphic to a G-set of the form H \ G, for some subgroup H ≤ G. Indeed, given x ∈ X, the surjective
G-equivariant map G → X sending g to x · g induces an isomorphism Gx \ G → X of G-sets.
8.6. Lemma. Let G be a group and X, X 0 two nonempty G-sets.

14
 (i) If X is transitive and f , f 0 : X → X 0 are two G-equivariant maps such that f ( x0 ) = f 0 ( x0 ) for some
x0 ∈ X, then f = g.
(ii) If X 0 is transitive and f : X → X 0 is a G-equivariant map, then it is surjective.
(iii) If X and X 0 are transitive, given x ∈ X and x 0 ∈ X 0 , then there exists a (resp., bijective) G-equivariant map
f : X → X 0 such that f ( x ) = x 0 if and only if Gx ⊆ Gx0 (resp., Gx = Gx0 ).
Proof. For item (i), note that, for every x ∈ X, there exists g ∈ G such that x0 · g = x, so

f ( x ) = f ( x0 · g ) = f ( x0 ) · g = f 0 ( x0 ) · g = f 0 ( x0 · g ) = f 0 ( x ).

For item (ii), note that, since X is nonempty, there exists x0 ∈ X. By transitivity of X 0 , for every x 0 ∈ X,
there ,there exists g ∈ G such that f ( x0 ) · g = x 0 , so f ( x0 · g) = f ( x0 ) · g = x 0 .
We now prove the first part of item (iii). If a G-equivariant map f : X → X 0 such that f ( x ) = x 0 exists,
then, given g ∈ Gx , x 0 = f ( x ) = f ( x · g) = f ( x ) · g = x 0 · g, so g ∈ Gx0 . Conversely, assume that Gx ⊆ Gx0 ,
and define f : X → X 0 by f ( x · g) = x 0 · g, for all g ∈ G. This is well-defined since x · g = x · g0 if and only
if gg0−1 ∈ Gx ⊆ Gx0 , which implies that x 0 · g = x 0 · g0 . Moreover, since X is transitive, every element of X
is of the form x · g, for some g ∈ G. The map f is clearly G-equivariant and f ( x ) = x 0 .
The statement concerning the bijective property in item (iii) follows directly. Indeed, a G-equivariant
map f is bijective if and only if it has a G-equivariant inverse map. Hence, applying the first part of item
(iii), we see that a bijective G-equivariant map f sending x to x 0 tells us that Gx = Gx0 . Conversely, if
Gx = Gx0 , there exists a G-equivariant map f sending x to x 0 and another f 0 sending x 0 to x. Then, the
G-equivariant map f 0 ◦ f send x to itself, whereas f ◦ f 0 sends x 0 to itself, so they must be the identity of
X and of X 0 respectively by (i).
8.7. Recall that, given a subgroup H ≤ G of a group G, the normalizer NG ( H ) of H in G is the subgroup
of G given by the element g ∈ G such that g−1 Hg = H. Equivalently, it is the largest subgroup of G
including H such that H is normal in it.
8.8. Proposition. Let G be a group, X a (nonempty) transitive G-set and x0 ∈ X. Denote by AutG ( X ) the group
of G-equivariant automorphisms of X. Define the map e x0 : NG ( Gx0 ) → AutG ( X ) sending g to the unique
G-equivariant automorphism of X satisfying that e x0 ( g)( x0 ) = x0 · g. Then e x0 is surjective morphism of groups
with kernel Gx0 , so it induces an isomorphism of groups

ē x0 : NG ( Gx0 )/Gx0 → AutG ( X ).

Proof. We first note that the map e x0 is well defined. Let g ∈ NG ( Gx0 ), so g−1 ∈ NG ( Gx0 ). Then,
Gx0 · g = g−1 Gx0 g = Gx0 , and there exists a unique G-equivariant automorphism e x0 ( g) : X → X sending
x0 to x0 · g by Lemma 8.6, (iii). Moreover, e x0 is a morphism of groups, since

e x0 ( gg0 )( x0 ) = x0 · ( gg0 ) = ( x0 · g) · g0 = e x0 ( g)( x0 ) · g0 = e x0 ( g)( x0 · g0 )

= ex0 ( g)(ex0 ( g0 )( x0 )) = (ex0 ( g) ◦ (ex0 ( g0 ))( x0 ),

for all g, g0 ∈ NG ( Gx0 ), so e x0 ( gg0 ) = e x0 ( g) ◦ (e x0 ( g0 ) for all g, g0 ∈ NG ( Gx0 ), by Lemma 8.6, (i).
The morphism of groups e x0 is also surjective. Indeed, given f ∈ AutG ( X ), let x00 = f ( x0 ). Since X is
transitive there exists g ∈ G such that x00 = x0 · g, and in particular Gx00 = g−1 Gx0 g. By Lemma 8.6, (iii),
Gx0 = Gx00 , so Gx0 = g−1 Gx0 g, which implies that g ∈ NG ( Gx0 ). Since e x0 ( g)( x0 ) = x0 · g = x00 = f ( x0 ),
Lemma 8.6, (i) tells us thet e x0 = f .
We will finally prove that the kernel of e x0 is Gx0 . It is clear that g ∈ Ker(e x0 ) if and only if e x0 ( g) = idX ,
which, by Lemma 8.6, (i), is tantamount to e x0 ( g)( x0 ) = x0 · g = x0 , which is equivalent to g ∈ Gx0 .
8.9. Exercise. Let X be a right G-set and Y a left G-set. Consider the left action of G on X × Y given by
g · ( x, y) = ( x · g−1 , g · y), for g ∈ G, x ∈ X and y ∈ Y. Define X ×G Y to be G \( X × Y ).

15
 (i) Prove that if X = ti∈ I Yi , with Xi right G-sets, then the left G-set X × Y is isomorphic to the disjoint union
left G-set ti∈ I ( X × Yi ). Conclude that the sets X ×G Y and ti∈ I ( X ×G Yi ) are in bijection.
(ii) Let X = H \ G.

(a) Prove that the map Y → X ×G Y given by y 7→ [( H, y)] induces a map ι : H \Y → X ×G Y.

(b) Prove that the map G × Y → Y given by ( g, y) 7→ g · y is G-equivariant, where G × Y has the
regular left action given by g · ( g0 , y) = ( gg0 , y), for all g, g0 ∈ G and y ∈ Y. Prove that the previous
mapping induces a map µ : X ×G Y → H \Y.
(c) Finally, prove that ι and µ are inverse to each other.

8.10. Let G and H be two groups. Recall that a G-H-biset is a left ( G × H op )-set, i.e. a set X provided with
a left action of G and a right action of H such that g · ( x · h) = ( g · x ) · h, for all g ∈ G, h ∈ H and x ∈ X.
Equivalently, a G-H-biset is a set X together with a morphism of groups H → AutG ( X )op . Morphisms of
G-H-bisets are just morphisms of ( G × H op )-sets.
8.11. Assume that X is a G-H-biset such that its underlying left G-set structure is isomorphic to G. In
other words, we assume that X = G with the regular left action of G, and there is morphism of groups
H → AutG ( G )op . Since the map Gop → AutG ( G ) sending g to x 7→ x · g for all x, g ∈ G is an isomorphism
of groups, we see that the right H-set structure on G provided with the regular left G-set structure is given
by a morphism of groups H → G.

§9. Covering spaces I: Basic properties

9.1. Let p : X̃ → X be a continuous map of topological spaces. We say that p is a covering if for every
x ∈ X there are an open set U ⊆ X such that x ∈ U as well as decomposition

p −1 (U ) =
G
Ũi
i∈ I

for some nonempty index set I (depending on U) such that Ũi ⊆ X̃ is open and p|Ũi : Ũi → U is a
homeomorphism for all i ∈ I. Equivalently, p is a covering if for every x ∈ X there are an open set
U ⊆ X such that x ∈ U as well as a nonempty discrete topological space F (depending on U) and a
homeomorphism

τ : p −1 (U ) → F × U

such that such that πU ◦ τ = p, where πU : F × U → U is the canonical projection. Note that this implies
that p is surjective. Moreover, Exercise 2.3 tells us that p is open. If the cardinal of the set I (or F) is
independent of the open sets U considered, we will say the covering p : X̃ → X has degree #( I ) = #( F )
(see Exercise 9.3).
9.2. If p : X̃ → X is a covering and x ∈ X, we call Fx = p−1 ({ x }) the fiber of p at x. By definition, the
subspace topology of p−1 ({ x }) induced from that of X̃ is discrete. We will say that a covering p : X̃ → X
is universal if X̃ is simply connected.
9.3. Exercise. Let p : X̃ → X be a covering such that X is connected. Prove that Fx is homeomorphic to Fx0 , for all
x, x 0 ∈ X.
9.4. Exercise. Prove that the quotient topology of R/Q coincides with the trivial topology. Deduce that the canonical
projection R → R/Q is not a covering.

16
9.5. The definitions in paragraphs 9.1 and 9.2 can also be carried out mutatis mutandi for pointed topological
spaces, by replacing topological spaces and their morphisms by pointed ones and their morphisms, giving
the notion of pointed covering.
9.6. Exercise. (i) Let X be fixed topological space. Let C0 be the collection of all pairs ( X̃, p), where X̃ is a
topological space and p : X̃ → X is a covering, HomC (( X̃, p), ( X̃ 0 , p0 )) the set of all continuous maps from
f : X̃ → X̃ 0 such that p0 ◦ f = p, ◦C the usual composition of maps and id( X̃,p) is the identity map from X̃
to itself. Prove that this forms a category, denoted by Cov( X ).
(ii) Suppose now that ( X, x ) is a fixed pointed topological space. Consider the same construction as before by
replacing topological spaces and their morphisms by pointed topological spaces ones and their morphisms.
Prove that this forms a category, denoted by Cov( X, x ).
9.7. A covering p : X̃ → X of X is called trivial if there exists a nonempty discrete topological space F
such that ( X̃, p) is isomorphic in Cov( X ) to the covering p I : X × F → X, where X × F has the product
topology and p I sends ( x, f ) ∈ X × F to x ∈ X. The same definition makes perfect sense in the category
of pointed coverings.
9.8. Exercise. (i) Let p : R → S1 be the map sending t ∈ R to ei2πt , where S1 ⊆ R2 ∼
= C has the subspace
topology. Prove that

p−1 S1 \ {ei2πt0 } =

G
]t0 + n, t0 + n + 1[,
n ∈Z

for all t0 ∈ R. Conclude that this gives a (universal) covering of S1 .

(ii) Let p : C → C× be the map sending z to ez . Prove that this gives a (universal) covering of C× .
(iii) Let n ∈ N and p : C× → C× be the map sending z to zn . Prove that this gives a covering of C× of degree n.
(iv) Recall that the (real) projective space of dimension n ∈ N is defined as RPn = (Rn+1 \ {0Rn+1 })/ ∼,
where ∼ is the equivalence relation on Rn+1 \ {0Rn+1 } given by x ∼ y if and only if there exists λ ∈ R×
such that y = λx. Let n ∈ N be an integer and let p : Sn → RPn be the canonical projection sending
x ∈ Sn to the equivalence class including x ∈ Rn+1 . Show that this gives a covering of RPn of degree 2,
which is universal if n ≥ 2. For n = 1, prove that RP1 is homeomorphic to S1 .
(v) Let G and G 0 be topological groups and let p : G → G 0 be a continuous morphism of groups. Prove that p is
a covering if and only if p is open, surjective and the kernel of p is a discrete subset of G.
9.9. Exercise. Let p : X̃ → X be a covering and let f : Y → X be a continuous map. Prove that p̂ : X̃ × X Y → Y,
where X̃ × X Y = {( x̃, y) ∈ X̃ × Y : p( x̃ ) = f (y)} ⊆ X̃ × Y has the subspace topology, is a covering as well.
9.10. Let p : X̃ → X be a covering, where X is locally path connected and path connected. Then, X̃ is also
locally path connected, so its path connected components { X̃i }i∈ I are open in X̃. Define pi = p| X̃i : X̃i → X.
We claim that p| X̃i : X̃i → X is a covering, for all i ∈ I, i.e. any covering over a locally path connected
and path connected space X decomposes as a disjoint union of covering of path connected spaces over
X. To prove it we proceed as follows. Since X is locally path connected, for every x ∈ X we can pick
an open neighborhood U of x that is path connected and that p−1 (U ) = t j∈ J Ũj , with Uj ⊆ X̃ open and
p|Ũj : Ũj → U a homeomorphism. Hence, Ũj is path connected, so for every j ∈ J, there is i ∈ I such
that Ũj ⊆ X̃i . If we prove that p| X̃i : X̃i → X is surjective, then it is a covering. To do so, note that the
image Vi = p( X̃i ) is open, since p is open, and it coincides with { x ∈ X : p−1 ({ x }) ∩ X̃i 6= ∅}. However,
the subset { x ∈ X : p−1 ({ x }) ∩ X̃i = ∅} ⊆ X is also open. Since X is connected and Vi 6= ∅, we get
that X = Vi = p( X̃i ), as was to be shown. The covering p : X̃ → X is then a disjoint union of coverings
pi : X̃i → X of path connected topological spaces.

17
9.11. Let G be a group. A covering p : X̃ → X of topological spaces is called a G-covering if X̃ is provided
with a free left group action of G such that p( g · x̃ ) = p( x̃ ) for all x̃ ∈ X̃ and g ∈ G, and the restriction
of the action of G to the fiber p−1 ({ x }) is transitive. Note in particular that the fiber of p at every point
x ∈ X is (noncanonically) isomorphic to G as left G-sets. A morphism of G-coverings from a G-covering
p : X̃ → X to another one p0 : X̃ 0 → X is a continuous map f : X̃ → X̃ 0 that is G-equivariant and
p0 ◦ f = p.
9.12. Exercise. (i) Let X be fixed topological space and G a group. Let C0 be the collection of all pairs ( X̃, p),
where X̃ is a topological space and p : X̃ → X is a G-covering, HomC (( X̃, p), ( X̃ 0 , p0 )) the set of all
continuous and G-equivariant maps f : X̃ → X̃ 0 such that p0 ◦ f = p, ◦C the usual composition of maps
and id( X̃,p) is the identity map from X̃ to itself. Prove that this forms a category, denoted by Cov G ( X ).
(ii) Suppose now that ( X, x ) is a fixed pointed topological space and G is a group. Consider the same construction
as before by replacing topological spaces and their morphisms by pointed topological spaces ones and their
morphisms. Prove that this forms a category, denoted by Cov G ( X, x ).
9.13. Given morphisms p : X̃ → X and f : Z → X in a category C , a lifting of f along p is a morphism
f˜ : Z → X̃ in C such that p ◦ f˜ = f , i.e. the diagram

X̃
f˜

Z p
f

commutes.
9.14. Proposition. Let p : ( X̃, x̃0 ) → ( X, x0 ) be a covering of pointed topological spaces and let ( Z, z0 ) be a
connected pointed topological space together with a morphism f : ( Z, z0 ) → ( X, x0 ) of pointed topological spaces.
Let f˜1 , f˜2 : ( Z, z0 ) → ( X̃, x̃0 ) be two liftings of f along p, i.e. p ◦ f˜i = f , for i = 1, 2. Then f˜1 = f˜2 .
Proof. Since p ◦ f˜1 = f = p ◦ f˜2 , we get a continuous map

f˜ : Z → X̃ × X X̃ = {( x̃, x̃ 0 ) ∈ X̃ × X̃ : p( x̃ ) = p( x̃ 0 )}

given by f˜(z) = ( f˜1 (z), f˜2 (z)), for all z ∈ Z. Since p is a covering, it is separated (using Exercise 2.5 and
the fact that p−1 ({ x }) is discrete so a fortiori relatively Hausdorff) and a local homeomorphism, so

∆ p = {( x̃, x̃ ) ∈ X̃ × X̃ } ⊆ X̃ × X X̃

is both closed and open in X̃ × X X̃. Hence, f˜−1 (∆ p ) is both closed and open in Z. Since it contains z0 and
Z is connected we get that f˜−1 (∆ p ) = Z, which proves the statement.
9.15. Exercise. Let X̃ be a topological space and let G be a group acting on the left on X̃ via a morphism
ρ : G → AutTop ( X̃ ) of groups. We consider in this case the quotient space G \ X̃ provided with the quotient topology
under the canonical projection pG : X̃ → G \ X̃. We say that that ρ is a covering group action if for every x̃ ∈ X̃
there is an open neighborhood U such that U ∩ ρ( g)(U ) = ∅ for all g ∈ G \ {eG }, where eG denotes the unit of G.
2 Note that this implies that the action is free, and, in particular, the action ρ is effective.

(i) Assume that X̃ is connected. Let p : X̃ → X be a covering and let G = AutCov( X ) ( X̃, p). Prove that the
left action of G on X̃ given by ρ( g)( x̃ ) = g( x̃ ) for all g ∈ G and x̃ ∈ X̃ is a covering group action.
2 Thisis called a properly discontinuous action in some classical textbooks, but we refrain from using
this oxymoronic expression which simultaneously conveys that the action is continuous and discontinuous.

18
 (ii) Prove that, if ρ : G → AutTop ( X̃ ) is a covering group action and H ≤ G is a subgroup of G, then
ρ| H : H → AutTop ( X̃ ) is a covering group action.
(iii) Assume that we have an effective action of a group G on a connected topological space X̃. Prove that if pG is
a covering, then ρ is a covering group action.
(iv) Prove that if ρ is a covering group action, then pG is a covering. Conclude in this case that pG : X̃ → G \ X̃
is a G-covering.
(v) Prove that if p : X̃ → X is a G-covering, then there is a homeomorphism f : G \ X̃ → X such that
p = f ◦ pG .
9.16. Proposition. Let p : X̃ → X be a covering of topological spaces, Z a topological space together with
a continuous map Z → X̃, and A ⊆ Z a subset. Let H : Z × [0, 1] → X be continuous map such that
H (z, 0) = p ◦ f˜(z) for all z ∈ Z, and H (z, t) = p ◦ f˜(z) for all z ∈ A and t ∈ [0, 1]. Then, there exists a unique
continuous map H̃ : Z × [0, 1] → X̃ such that p ◦ H̃ = H, H̃ (z, 0) = f˜(z) for all z ∈ Z, and H̃ (z, t) = f˜(z) for
all z ∈ A and t ∈ [0, 1].
Proof. We first prove the uniqueness. Given two continuous maps H̃1 , H̃2 : Z × [0, 1] → X̃ as in
the statement, then, for every z ∈ Z, H̃1 (z, −), H̃2 (z, −) : ([0, 1], 0) → ( X̃, f˜(z)) are two liftings of
H (z, −) : ([0, 1], 0) → ( X, p( f˜(z))) along p, which should coincide by Proposition 9.14.
Finally we prove the existence. We will prove that there is an open covering U of Z such that for all
U ∈ U there is a continuous map H̃U : U × [0, 1] → X̃ such that p ◦ H̃U = H |U ×[0,1] , H̃ (z, 0) = f˜(z) for
all z ∈ U, and H̃ (z, t) = f˜(z) for all z ∈ A ∩ U and t ∈ [0, 1]. By the previous uniqueness result, H̃U is in
fact unique. Moreover, we claim that the proposition follows directly from this local result. Indeed, given
two open sets U, U 0 ∈ U and the corresponding maps H̃U , H̃U 0 , the previous uniqueness also tells us that
H̃U |(U ∩U 0 )×[0,1] = H̃U 0 |(U ∩U 0 )×[0,1] , and in consequence the unique continuous map H̃ : Z × [0, 1] → X̃
such that H̃ |U ×[0,1] = H̃U satisfies the conditions of the statement.
It thus remains to prove that given any z0 ∈ Z, there is an open set U ⊆ Z including z0 and a
continuous map H̃U : U × [0, 1] → X̃ satisfying the previous conditions. Let (Vj ) j∈ J be an open covering
of X such that p−1 (Vj ) = ti∈ Ij Ṽi,j for some nonempty set Ij , with Ṽi,j ⊆ X̃ open and p|Ṽi,j : Ṽi,j → Vj
a homeomorphism. Since H is continuous, for every t ∈ [0, 1], there exist a neighborhood Ut of z0 ,
an open connected neighborhood Jt ⊆ [0, 1] of t and jt ∈ J such that H (Ut × Jt ) ⊆ Vjt . By the com-
pactness of [0, 1], there is a finite set T ⊆ [0, 1] such that { Jt }t∈T is a finite covering of [0, 1]. Define
U = ∩t∈T Ut . Suppose T = {t1 < · · · < tn } for some integer n ≥ 2. To simplify, we will write
jk instead of jtk , for k ∈ J1, nK. Pick sk ∈ Jtk ∩ Jtk+1 for all k ∈ J1, n − 1K. By defining s0 = 0 and
sn = 1 we see that H (U × [sk , sk+1 ]) ⊆ Vjk+1 , for all k ∈ J0, n − 1K. Assume that a continuous map
H̃U,k : U × [0, sk ] → X̃ satisfying the conditions of the proposition was constructed, for some k ∈ J1, n − 2K.
We will prove that there is a continuous map H̃U,k+1 : U × [0, sk+1 ] → X̃ also satisfying the conditions
of the proposition and H̃U,k+1 |U ×[0,sk ] = H̃U,k . Since H (U × [sk , sk+1 ]) ⊆ Vjk+1 , H (z0 , sk ) ∈ Vjk+1 . Let
i ∈ Ijk+1 such that H̃U,k (z0 , sk ) ∈ Ṽi,jk+1 and set pi = p|Ṽi,j : Ṽi,jk+1 → Vjk+1 . Define now H̃U,k+1 such
k +1

that H̃U,k+1 |U ×[0,sk ] = H̃U,k and H̃U,k+1 |U ×[sk ,sk+1 ] = pi−1 ◦ H |U ×[sk ,sk+1 ] . This gives the desired map. The
proposition is thus proved.
9.17. Remark. The previous proposition tells us that, if p : ( X̃, x̃0 ) → ( X, x0 ) is a covering of pointed topological
spaces, then π1 ( p) : π1 ( X̃, x̃0 ) → π1 ( X, x0 ) in injective.
9.18. Remark. By taking Z = {∗} in the previous proposition we see that, given a continuous map γ : [0, 1] → X
such that γ(0) = x0 ) and a pointed covering p : ( X̃, x̃0 ) → ( X, x0 ) of topological spaces, there exists a unique
continuous map γ̃ : [0, 1] → X̃ such that p ◦ γ̃ = γ and γ̃(0) = x̃0 .
9.19. Theorem. Let p : ( X̃, x̃0 ) → ( X, x0 ) be a covering of pointed topological spaces and let ( Z, z0 ) be a path
connected and locally path connected pointed topological space together with a morphism f : ( Z, z0 ) → ( X, x0 ) of

19
pointed topological spaces. Then, there exists a lifting f˜ : ( Z, z0 ) → ( X̃, x̃0 ) of f along p, i.e. p ◦ f˜ = f , if and only
if π1 ( f )(π1 ( Z, z0 )) ⊆ π1 ( p)(π1 ( X̃, x̃0 )).
Proof. Assume first that there exists a lifting f˜ : ( Z, z0 ) → ( X̃, x̃0 ) of f along p, i.e. p ◦ f˜ = f . Then, since π1
is a functor we get π1 ( p) ◦ π1 ( f˜) = π1 ( f ), which in particular gives us π1 ( f )(π1 ( Z, z0 )) ⊆ π1 ( p)( X̃, x̃0 ).
Assume now that π1 ( f )(π1 ( Z, z0 )) ⊆ π1 ( p)( X̃, x̃0 ). Define a map f˜ : Z → X̃ as follows. Given z ∈ Z,
let η : [0, 1] → Z be a path satisfying that η (0) = z0 and η (1) = z, which exists since Z is path connected.
By choosing Z = {∗} in Proposition 9.16, we see that the path f ◦ η in X has a lift η̃ : ([0, 1], 0) → ( X̃, x̃0 )
along p, i.e. p ◦ η̃ = f ◦ η. Define now f˜(z) = η̃ (1). We claim that f˜ is well defined, i.e. it is independent
of the choice of path η, and it is continuous.
We first prove the good definition of f˜. Let η 0 : [0, 1] → Z be another path satisfying that η 0 (0) = z0
and η 0 (1) = z, and let η̃ 0 : [0, 1] → X̃ be the unique path such that p ◦ η̃ 0 = f ◦ η 0 . We have to
show that η̃ 0 (1) = η̃ (1). Consider the loop γ = η · η 0− in Z based at z0 and thus the loop f ◦ γ
in X based at x0 . Since π1 ( f )(π1 ( Z, z0 )) ⊆ π1 ( p)(π1 ( X̃, x̃0 )), there exists a loop γ̃ in X̃ based at
x̃0 such that p ◦ γ̃ ' f ◦ γ (rel{0, 1}) by means of a homotopy H. By Proposition 9.16, there exists
thus a homotopy of paths H̃ such that H̃ (t, 0) = γ̃(t) for all t ∈ [0, 1] and p ◦ H̃ = H. In particular,
p ◦ H̃ (t, 1) = H (t, 1) = f ◦ γ(t) for all t ∈ [0, 1], which means that the path γ̃0 : [0, 1] → X̃ given by
γ̃0 (t) = H̃ (t, 1) is a lifting of f ◦ γ. Since γ̃ is a loop and H̃ is a homotopy of paths, γ̃0 is a loop as well. As
−
p ◦ γ̃0 = f ◦ (η · η 0− ), is now clear that γ̃0 = η̃ · η˜0 , which in turn implies that η̃ 0 (1) = η̃ (1).
˜
We finally prove the continuity of f . Let z ∈ Z be any point. Pick an open neighborhood U of
f (z) such that p−1 (U ) = ti∈ I Ũi , for a nonempty set I, Ũi ⊆ X̃ open and pi = p|Ũi : Ũi → U is a
homeomorphism for all i ∈ I. Since f is continuous and Z is locally path connected, there exists a path
connected open neighborhood V of z such that f (V ) ⊆ U. Assume that f˜(z) ∈ Ũi0 . Note now that
f˜|V = pi−0 1 ◦ f |V , which follows from uniqueness of liftings by Proposition 9.14. Then, f˜|V is clearly
continuous, and in consequence f˜ is also continuous.
9.20. Exercise. Let G be a topological group that is locally path connected and path connected as a topological space
and eG be its unit. Let p : G̃ → G be a covering of path connected topological spaces and pick ẽ ∈ G̃ such that
p(ẽ) = eG .
(i) Prove that there exists a unique group structure on G̃ with unit ẽ such that it becomes a topological group
for the given topology and p is a morphism of groups.
(ii) Prove that G is abelian if and only if G̃ is so.
(iii) Prove that the sequence of groups
π1 ( p ) ẽ ∂
1 → π1 ( G̃, ẽ) −→ π1 ( G, eG ) −→ Ker( p) → 1

is exact, where ∂ẽ ([γ]) = γ̃(1), where γ̃ : ([0, 1], 0) → ( G̃, ẽ) is the unique lifting of γ.
(iv) Prove that π1 (S1 , 1) ∼
= Z and thus π1 (T2 , (1, 1)) ∼
= Z × Z, where T2 = S1 × S1 is the 2-dimensional torus.
Deduce that π1 (M, (0, 0)) ∼= Z (see Exercise 7.15).
(v) Prove that there is no continuous map r : B̄1 (0, 0) → S1 such that r |S1 = idS1 .
(vi) Prove that every continuous map f : B̄1 (0, 0) → B̄1 (0, 0) has a fixed point.
(vii) Prove that for any continuous function f : S2 → R2 there is a point x ∈ S2 such that f ( x ) = f (− x ).
Hint. Consider g : S2 → S1 given by g( x ) = ( f ( x ) − ( f (− x )))/k( f ( x ) − ( f (− x )))k for x ∈ S2 ,
γ : [0, 1] → S2 defined as γ(t) = (cos(2πt), sin(2πt), 0) for t ∈ [0, 1] and γ0 = g ◦ γ. Note γ0 is null
homotopic, since γ is. Note on the other hand that γ0 (s + 1/2) = −γ0 (s) for s ∈ [0, 1/2] and deduce from
this that any lifting γ̃0 : [0, 1] → R of γ0 is not a loop, so γ0 is not null homotopic.

20
9.21. By Theorem 9.19, the existence of a lifting f : ( Z, z0 ) → ( X, x0 ) along a covering p is assured if Z is
simply connected. Moreover, if ( X, x0 ) is a path connected and locally path connected topological space,
then the previous theorem tells us that the map
 
 isomorphism classes of pointed 
 ( )
 subgroups of
path connected covering spaces −→ (2)
π1 ( X, x0 )
p : ( X̃, x̃0 ) → ( X, x0 )

 


sending a covering p : ( X̃, x̃0 ) → ( X, x0 ) of pointed path connected topological spaces to π1 ( p)(π1 ( X̃, x̃0 ))
is injective.
9.22. Exercise. (i) Classify all coverings of S1 , up to isomorphism of coverings.
(ii) Prove that every subgroup of Z × Z has at most two generators. Conclude that it coincides with one of the
subgroups below:
(1) the trivial subgroup {(0, 0)} of Z × Z;
(2) {Z.( a, b) : for all a, b ∈ Z such that ( a, b) 6= (0, 0)};
(3) {Z.( a, b) ⊕ Z.(c, d) : for all a, b, c, d ∈ Z such that ad 6= bc}.
(iii) Classify all coverings of T2 = S1 × S1 , up to isomorphism of coverings.
Hint. Given a, b, c, d ∈ Z, consider the maps p : S1 × R → S1 × S1 given by p(z, x ) = (z a , zb e2πix ) for
z ∈ S1 and x ∈ R, as well as p0 : S1 × S1 → S1 × S1 given by p0 (z, z0 ) = (z a z0c , zb z0d ) for z, z0 ∈ S1 .

§10. Covering spaces II: Automorphisms of coverings

10.1. We will generalize the construction in paragraph 9.21 as follows. Let X be a path connected and
locally path connected topological space and p : X̃ → X a covering. For x ∈ X, consider the fiber
Fx = p−1 ({ x }) ⊆ X̃, which is a discrete subspace. Fix x0 ∈ X and define the map

ρ : Fx0 × π1 ( X, x0 ) → Fx0 (3)

by ρ( x̃, [γ]) = γ̃(1), where γ̃ : [0, 1] → X̃ is a path satisfying that p ◦ γ̃ ' γ (rel{0, 1}) and γ̃(0) = x̃. It
is well defined by Proposition 9.16. Then, ρ is a right action of π1 ( X, x0 ) on Fx0 , called the monodromy
action at x0 . If p : X̃ → X and p0 : X̃ 0 → X are two coverings and f : X̃ → X̃ 0 is a morphism of coverings,
then it induces a map f | p−1 ({ x0 }) : p−1 ({ x0 }) → p0−1 ({ x0 }) which is clearly equivariant with respect to
the right action of π1 ( X, x0 ). As a consequence, one obtains a functor

Fibx0 : Cov( X ) → Set -π1 ( X, x0 )

sending the covering p : X̃ → X to the fiber Fx0 = p−1 ({ x0 }), and a morphism f : X̃ → X̃ 0 to the
π1 ( X, x0 )-equivariant map f | p−1 ({ x0 }) : p−1 ({ x0 }) → p0−1 ({ x0 }).
10.2. Note that the monodromy action (3) at x0 ∈ X is transitive if and only if X̃ is path connected. Indeed,
if X̃ is path connected, given any two points x̃ and x̃ 0 of Fx0 , there is a path η̃ in X̃ starting at x̃ and ending
at x̃ 0 . Then, x̃ 0 = ρ( x̃, [ p ◦ η̃ ]). Conversely, fix a point x̃ ∈ Fx0 and let x̃ 0 ∈ X̃ be a general point. It suffices
to prove that the transitivity of the action at x0 implies that there is path in X̃ starting at x̃ 0 and ending at
x̃. Since X is connected, there is a path η in X starting at p( x̃ 0 ) and ending at x0 = p( x̃ ). By Proposition
9.16, there is a path X̃ starting at x̃ 0 and ending at a point x̃ 00 in Fx0 . Since the monodromy action at x0 is
transitive, then two points in Fx0 can be connected by a path in X̃, so there is a path η̃ 0 starting at x̃ 00 and
ending at a point x̃. The path η̃ · η̃ 0 proves our claim.

21
10.3. Note also that, given x̃ ∈ X̃ such that p( x̃ ) = x0 , then the isotropy group Γ x̃ of x̃ for the monodromy
action (3) is exactly π1 ( p)(π1 ( X̃, x̃ )). Indeed, by definition Γ x̃ is the subgroup of homotopy classes of
loops at x0 (relative to {0, 1}) whose lift to X̃ (along p) are also loops at x̃. In this sense, the functor Fibx0
also encodes the information given by the map (2).
10.4. Finally, notice that, if X̃ is path connected, the monodromy action (3) at x0 ∈ X is free if and only if
X̃ is simply connected. Indeed, by paragraph 10.2, X̃ is path connected is equivalent to the monodromy
action being transitive. Then, using (1), the action is free if and only if the isotropy group Γ x̃ of any
x̃ ∈ X̃ is trivial, i.e. if π1 ( p)(π1 ( X̃, x̃ )) is trivial. By Proposition 9.16 (or Remark 9.17), this is equivalent to
π1 ( X̃, x̃ ) being trivial, i.e. X̃ is simply connected.
10.5. The functoriality of Fibx0 tells us that, if f : X̃ → X̃ is an automorphism of a covering p : X̃ → X,
where X is a path connected and locally path connected topological space and x0 ∈ X, then Fibx0 ( f ) is an
automorphism of the π1 ( X, x0 )-set Fx0 = Fibx0 ( X̃, p). We have even the following stronger result.
10.6. Theorem. Let X be a path connected and locally path connected topological space and x0 ∈ X. Let p : X̃ → X
be a covering, where X̃ is path connected. Then, the functor Fibx0 restricts to an isomorphism of groups

Fibx0 |AutCov(X) (X̃ ) : AutCov(X ) ( X̃ ) → Autπ1 (X,x0 ) ( Fx0 ). (4)

Proof. Note first that, by Theorem 9.19, Fibx ( f )( x̃ ) = x̃ for some x̃ ∈ Fx0 if and only if f is the identity.
This implies that the map (4) is injective. As explained in paragraph 10.2, the path connectedness
assumption on X̃ implies that Fx0 is a transitive π1 ( X, x0 )-set. Let f¯ : Fx0 → Fx0 be a π1 ( X, x0 )-equivariant
automorphism and pick x̃ ∈ Fx0 . By Lemma 8.6, (iii), the isotropy groups Γ x̃ and Γ f¯( x̃) coincide. Since
Γ x̃ = π1 ( p)(π1 ( X̃, x̃ )) and Γ ¯ = π1 ( p)(π1 ( X̃, f¯( x̃ ))), by paragraph 10.3, Theorem 9.19 tells us that there
f ( x̃ )
exists a morphism f : X̃ → X̃ of the covering p : X̃ → X such that f ( x̃ ) = f¯( x̃ ). The same arguments
apply to f¯−1 , giving a morphism f 0 : X̃ → X̃ of the covering p : X̃ → X such that f 0 ( x̃ ) = f¯−1 ( x̃ ). Since
f 0 ◦ f ( x̃ ) = ( x̃ ) and f ◦ f 0 ( x̃ ) = ( x̃ ), Proposition 9.14 implies that f and f 0 are inverses to each other.
Moreover, since f¯( x0 ) = Fibx0 ( f )( x0 ) and Fx0 is transitive, Lemma 8.6, (iii), tells us that f¯ = Fibx0 ( f ). The
theorem is proved.
10.7. Combining Proposition 8.8 and Theorem 10.6, we get the main result of this section, namely that,
given a p : ( X̃, x̃0 ) → ( X, x0 ) of pointed path connected topological spaces with X locally path connected,
then there is an isomorphism of groups


AutCov(X ) ( X̃ ) ∼


= Nπ1 (X,x0 ) π1 ( p) π1 ( X̃, x̃0 ) /π1 ( p) π1 ( X̃, x̃0 ) .

In particular, if X̃ is simply connected, then

AutCov(X ) ( X̃ ) ∼
= π1 ( X, x0 ), (5)

which gives in particular a left action of π1 ( X, x0 ) on X̃ by automorphisms of coverings.

10.8. Exercise. (i) Give another proof that π1 (S1 , 1) ∼
= Z.
(ii) Prove that π1 (RPn , ∗) ∼
= Z/2Z, for ∗ any point in RPn and n ≥ 2.
10.9. We say that a covering p : ( X̃, x̃0 ) → ( X, x0 ) of pointed topological spaces, where X is a path con-
nected and locally path connected topological space, is normal (or regular) if the action of AutCov( X ) ( X̃ )
on Fx0 is transitive. If X̃ is path connected this is equivalent to π1 ( p)(π1 ( X̃, x̃0 )) being a normal sub-
group of π1 ( X, x0 ). Indeed, by Theorem 9.19, AutCov( X ) ( X̃ ) acts transitively on Fx0 if and only if
π1 ( p)(π1 ( X̃, x̃ )) = π1 ( p)(π1 ( X̃, x̃ 0 )) for any two points x̃, x̃ 0 ∈ Fx0 . However, since the latter subgroups
are precisely the isotropy subgroups of x̃, x̃ 0 ∈ Fx under the monodromy action, by paragraph 10.3, then
π1 ( p)(π1 ( X̃, x̃ )) = π1 ( p)(π1 ( X̃, x̃ 0 )) for any two points x̃, x̃ 0 ∈ Fx0 is equivalent to π1 ( p)(π1 ( X̃, x̃ )) being
a normal subgroup of π1 ( X, x0 ), by (1).

22
10.10. Exercise. Let X̃ be a topological space and let G be a group acting (on the left) on X̃ (see Exercise 9.15).
Assume that the action is covering and consider the quotient space G \ X̃ provided with the quotient topology under
the canonical projection pG : X̃ → G \ X̃.
(i) Show that the morphism of groups ι G : G → AutCov(G\ X̃ ) ( X̃ ) sending g to x̃ 7→ g · x̃ is injective.
(ii) Let G = Z/n, X a topological space and X̃ = G × X be the product space, where G has the discrete
topology. Endow X̃ with the regular action g · ( g0 , x ) = ( gg0 , x ). Prove that there is a homeomorphism
G \ X̃ ∼
= X, pG : X̃ → X is the projection on the second component under the previous homeomorphism, and
AutCov(G\X̃ ) ( X̃ ) ∼
= Sn , the group of permutations of n letters. Conclude that ι G is not surjective in this
case.
(iii) Assume that X̃ is path connected and locally path connected. Prove that pG is a normal covering and that ι G
is surjective giving an isomorphism G ∼
= AutCov(G\X̃ ) ( X̃ ) of groups.
10.11. Exercise. Let Si = S1 + (2i − 1, 0) = ∂B1 (2i − 1, 0) ⊆ R2 ∼
= C for i ∈ J1, 4K. Let X = S2 ∪ S3 ⊆ R2 and
let

X̃ = S1 ∪ S2 ∪ S3 ∪ S4 ⊆ R2 ∼
=C
provided with the subspace topology of R2 ∼ = C. Define p : X̃ → X such that


 2 + z, if z ∈ S1 ,

5 − (z − 3)2 , if z ∈ S ,

2
p(z) =
3 + ( z − 5 ) 2 , if z ∈ S ,
3




−2 + z, if z ∈ S4 .


(i) Prove that p : X̃ → X is a covering of degree 3.

(ii) Prove that the only automorphism of the covering p : X̃ → X is the identity.
Hint. Prove that any automorphism f˜ : X̃ → X̃ of the covering satisfies that f˜(S1 ) = S1 and f˜(S4 ) = S4 ,
so f˜(4) = 4.
(iii) Using the previous item deduce that π1 ( p)(π1 ( X̃, x̃ )) is not a normal subgroup of π1 ( X, x ).

§11. Covering spaces III: Simply connectedness

11.1. We say that a topological space X is semilocally simply connected if every x ∈ X has an open
neighborhood U such that the image of π1 (inc) : π1 (U, x ) → π1 ( X, x ) is trivial, where inc : U → X is the
inclusion.
11.2. Exercise. (i) Let X be a topological space provided with an open covering U such that U is simply
connected, for all U ∈ U . Prove that X is semilocally simply connected.
(ii) Define X = ∪n∈N ∂B1/n (1/n, 0) ⊆ R2 with the subspace topology. This space is called the Hawaiian
earring. Prove that X is not semilocally simply connected.
(iii) Let B̄1 (0, 0) ⊆ R2 be provided with the subspace topology and let ∼ be the equivalence relation on
B̄1 (0, 0) generated by (0, 0) ∼ (1, 0). Define X = B̄1 (0, 0)/ ∼, endowed with the quotient topology. Let
Xn = X × {n} and denote the equivalence class of an element ( x, y) ∈ B̄1 (0, 0) by [( x, y)]. Define now
X̂ = tn∈N0 Xn and let ∼0 be the equivalence relation on X̂ generated by

([( x, 0)], n) ∼0 ([(cos(2πx ), sin(2πx ))], n + 1),

23
 for all x ∈ [0, 1] and n ∈ N0 . Endow X̂ with the topology given by
   G  
G
U= Un t Xn : for all Un ⊆ Xn open and n0 ∈ N0 .
n ∈ N0 n ∈ N0
n ≤ n0 n > n0

Define X = X̂/ ∼0 and endow it with the quotient topology. Prove that X is semilocally simply connected
but no open set U ⊆ X including the equivalence class of ([(1, 0)], 0) ∈ X̂ is simply connected.
11.3. Theorem. Let X be a path connected and locally path connected topological space. The following conditions
are equivalent:
(i) X is semilocally simply connected;
(ii) X has a universal covering.
(iii) the map (2) is bijective.
Proof. It is clear that (iii) implies (ii), by taking the inverse image of the trivial subgroup under (2).
Moreover, (ii) clearly implies (i). Indeed, let p : X̃ → X be an universal and x ∈ X. Take an open
neighborhood U of x such that p−1 (U ) = ti∈ I Ũi , for a nonempty set I, Ũi ⊆ X̃ open and pi : p|Ũi : Ũi → U
is a homeomorphism for all i ∈ I. Fix i ∈ and let inc : U → X and inc ˜ : Ũi → X̃ be the inclusions. Then,
π1 (inc ◦ pi ) = π1 ( p ◦ inci ) = π1 ( p) ◦ π1 (inci ) together with the fact that π1 ( pi ) is an isomorphism tells
us that




Img π1 (inc) = Img π1 ( p) ◦ π1 (inci ) ⊆ Img π1 ( p) ,
which is a trivial subgroup of π1 ( X, x0 ), since π1 ( X̃, x̃0 ) is trivial.
Let us finally prove that (i) implies (iii). Let C ≤ π1 ( X, x0 ) be a subgroup. Define X̂ the set of
homotopy classes relative to {0, 1} of all paths η : [0, 1] → X starting at x0 , and let p̂ : X̂ → X be the map
p̂([η ]) = η (1), for every such path η. Let x̂0 be the homotopy class of the constant path with value x0 .
Note that the map ρ : π1 ( X, x0 ) × X̂ → X̂ given by ρ([γ], [η ]) = [γ] · η gives a well defined left action. Let
X̃ = C \ X̂ be the set of orbits of X̂ under this action and let x̃0 be the orbit containing x̂0 . We will denote
the orbit of a homotopy equivalence class [η ] also by [η ]. It is clear that p( x̂ ) = p( x̂ 0 ) if x̂ and x̂ 0 are in the
same orbit, so we have a well defined map p : X̃ → X.
Let U be the set of all open path connected subsets of X such that the image of the mapping
π1 (incU ) : π1 (U, x ) → π1 ( X, x ) is trivial for one (or, equivalently, for all) x ∈ U, where incU : U → X is
the inclusion. It is easy to see that U is a basis of the topology of X. Indeed, given U, U 0 ∈ U such that
U ∩ U 0 6= ∅, we see that U ∩ U 0 ∈ U , since π1 (incU ∩U 0 ) factors through π1 (incU ), so the image of the
former is included in the image of the latter, which is trivial. Now, given U ∈ U and η : [0, 1] → X a path
such that η (0) = x0 and η (1) ∈ U, define

Û ([η ]) = [η ] · [η 0 ] ∈ X̂ : η 0 : [0, 1] → U such that η 0 (0) = η (1) ⊆ X̂



The set

Û = Û ([η ]) : U ∈ U , η : ([0, 1], 0) → ( X, x0 ), η (1) = x1 ∈ U
defines a topology on X̂. Indeed, given Û ([η ]), Û 0 ([η 0 ]) ∈ Û such that Û ([η ]) ∩ Û 0 ([η 0 ]) 6= ∅, then
U 00 = U ∩ U 0 6= ∅ and there exists η 00 : ([0, 1], 0) → ( X, x0 ) such that η (1) ∈ U ∩ U 0 . It is clear that
Û 00 ([η 00 ]) ⊆ Û ([η ]), Û 0 ([η 0 ]). It is easy to see that X̂ is path connected. Indeed, given [η ] ∈ X̂ and a
representative η : [0, 1] → X, the map η̂ : [0, 1] → X̂ given by η̂ (s)(t) = [η (st)] is a path from x̂0 to [η ]. We
give X̃ the quotient topology for the canonical projection πC : X̂ → X̃. Since X̂ is path connected, X̃ is
also path connected. Define Ũ ([η ]) to be the image of Û ([η ]) under the canonical projection πC . Then,

πC−1 (Ũ ([η ])) =

[
Û ([γ] · [η ]),
[γ]∈C

24
so it is an open subset of X̃. It is clearly a basis of the topology of X̃.
We claim that the map p : ( X̃, x̃0 ) → ( X, x0 ) is a covering. Given x ∈ X, take U ∈ U such that x ∈ U.
Then,
p −1 (U ) =
G
Ũ ([η ]),
[η ] ∈ X̃
η (1) ∈ U

and p|Ũ ([η ]) : Ũ ([η ]) → U is clearly a homeomorphism, so p is a covering as claimed.

Finally, we will prove that C = π1 ( p)(π1 ( X̃, x̃0 )). Note first that X̂ is simply connected. Indeed,
if γ̂ : [0, 1] → X̂ is a loop based at x̂0 , consider γ = p ◦ γ̂ and define the map γ̂0 : [0, 1] → X̂ given
by γ̂(s)(t) = [γ(st)]. By Proposition 9.14, γ̂ = γ̂0 . Since γ̂ is a loop, x̂0 = γ̂(1) = γ̂0 (1) = [γ], i.e.
γ ' ε x0 (rel{0, 1}), which implies that γ̂0 = γ̂ ' ε x̂0 (rel{0, 1}), by Proposition 9.16. Finally, using Exercise
10.10 for the quotient X̂ → C \ X̂ as well as (5), we see that π1 ( p)(π1 ( X̃, x̃0 )) = C, as was to be shown.
11.4. Exercise. Let X be the topological space in Exercise 10.11. Prove that X is semilocally simply connected and
compute a universal covering.

§12. Covering spaces IV: Classification of coverings

12.1. Let X be a semilocally simply connected topological space with a point x0 ∈ X. Fix a universal
covering p : X̃ → X, which exists by Theorem 11.3. Recall from (5) that there is a left action of π1 ( X, x0 )
on X̃ by automorphisms of coverings, given by x̃ · [γ] = γ̃(1), where γ̃ is a lift of γ to X̃ along p such
that γ̃(0) = x̃. By Theorem 10.6, we can identify π1 ( X, x0 ) with the fiber p−1 ({ x0 }) once we fix a point
x̃ ∈ Fx0 , since the action of π1 ( X, x0 ) on p−1 ({ x0 }) is free and transitive (see 10.2 and 10.4). Given a set
F provided with a right action of π1 ( X, x0 ), define Γ p ( F ) = F ×π1 ( X,x0 ) X̃ as the quotient of F × X̃ by
the right action of π1 ( X, x0 ) given by ( f , x̃ 0 ) · [γ] = ( f · [γ], [γ]−1 · x̃ 0 ), for f ∈ F, x̃ 0 and [γ] ∈ π1 ( X, x ).
Endow F × X̃ with the product topology, where F has the discrete topology, and consider the quotient
topology on F ×π1 ( X,x0 ) X̃. It is easy to see that the canonical projection F × X̃ → X given by composing
the projection on the second component and p induces a continuous map p F : F ×π1 ( X,x0 ) X̃ → X.
12.2. Proposition. Let X be a semilocally simply connected topological space with a point x ∈ X and let p : X̃ → X
be a universal covering. Then, the continuous map p F : F ×π1 ( X,x0 ) X̃ → X constructed in the previous paragraph
is a covering.
Proof. Since p : X̃ → X is a covering, there exists an open neighborhood U of x such that there is a
homeomorphism p−1 (U ) ∼ = π1 ( X, x0 ) × U, where π1 ( X, x0 ) has the discrete topology and p identifies
with the projection on the second factor under this homeomorphism. Then, we get a homeomorphism
p− 1 ∼
F (U ) = F × U, where F has the discrete topology and p F identifies with the projection on the second
factor under this homeomorphism.
12.3. If f : F → F 0 is a π1 ( X, x0 )-equivariant map, then f × idX̃ : F × X̃ → F 0 × X̃ induces a map
Γ p ( f ) : F ×π1 ( X,x0 ) X̃ → F 0 ×π1 ( X,x0 ) X̃ such that p F0 ◦ Γ p ( f ) = p F . Moreover, f × idX̃ is clearly continuous,
for the corresponding topologies, which in turn implies that Γ p ( f ) is continuous.
12.4. Proposition. Let X be a semilocally simply connected topological space with a point x0 ∈ X and let p : X̃ → X
be a universal covering. Then, the maps given by sending a right π1 ( X, x0 )-set to F 7→ Γ p ( F ) = F ×π1 ( X,x0 ) X̃
defined in paragraph 12.1, and a π1 ( X, x0 )-equivariant map f : F → F 0 between right π1 ( X, x0 )-sets to the
morphism of coverings Γ p ( f ) : F ×π1 ( X,x0 ) X̃ → F 0 ×π1 ( X,x0 ) X̃ defined in the previous paragraph, is a functor
Γ p : Set -π1 ( X, x0 ) → Cov( X ).

25
Moreover, Γ p and the functor Fibx0 defined in paragraph 10.1 are quasi-inverse to each other.
Proof. The fact that Γ p is a functor is clear. We will prove that Γ p and Fibx0 are quasi-inverse functors. We
first note that Fibx0 ◦Γ p is clearly the identity functor of Set-π1 ( X, x0 ). On the other hand, Γ p ◦ Fibx0 is
naturally isomorphic to the identity functor of Cov( X ), which implies that Fibx0 and Γ p are quasi-inverse
equivalences (see Exercise 5.15). Indeed, given any covering p̂ : X̂ → X, it decomposes as the disjoint
union of the coverings p̂i = p̂| X̂i : X̂i → X of path connected topological spaces by paragraph 9.10, where
{ X̂i }i∈ I are the path connected components of X̂. The result now follows from Exercise 8.9, using that
every covering p̂ : X̂ → X of path connected topological spaces is isomorphic to one of the form H \ X̃ → X,
where H ≤ π1 ( X, x0 ) ∼ = AutCov(X ) ( X̃ ), by Exercise 9.15 for the quotient X̃ → H \ X̃ as well as (5).
12.5. Exercise. Let G be a group and C0 be the collection of all pairs ( X, x ), where X is a nonempty right G-set
and x ∈ X, HomC (( X, x ), (Y, y)) the set of all G-equivariant maps f from X to Y such that f ( x ) = y, ◦C the
usual composition of maps and idX is the identity map from X to itself. Prove that this forms a category, denoted by
Set• -G, called the category of pointed right G-sets.
12.6. Let X be a semilocally simply connected topological space, x0 ∈ X and let p : ( X̃, x̃0 ) → ( X, x0 ) be a
universal covering. The functors Γ p and Fibx0 considered in Proposition 12.4 can also be easily extended
to functors

Γ p,• : Set• -π1 ( X, x0 ) → Cov( X, x0 )

and

Fibx0 ,• : Cov( X, x0 ) → Set• -π1 ( X, x0 )

as follows. Given a right π1 ( X, x0 )-set F and v ∈ F, we will write v̂ for the class in F ×π1 ( X,x0 ) X̃ of
(v, x̃0 ) ∈ F × X̃. We define Γ p,• ( F, v) = (Γ p ( F ), v̂) for any right π1 ( X, x0 )-set F with v ∈ F, and note
that the covering map p F : F ×π1 ( X,x0 ) X̃ → X sends v̂ to x0 . If f : ( F, v) → ( F 0 , v0 ) is a morphism of
pointed right π1 ( X, x0 )-sets, then set Γ p,• ( f ) = Γ p ( f ) and note that Γ p ( f ) sends v̂ to v̂0 . Analogously,
given a covering p̂ : ( X̂, x̂ ) → ( X, x0 ) of pointed topological spaces, we set Fibx0 ,• ( p̂) = ( p̂−1 ({ x }), x̂ )
and for any morphism f : X̂ → X̂ 0 from a covering p̂ : ( X̂, x̂ ) → ( X, x0 ) of pointed topological spaces
to another one p̂0 : ( X̂, x̂ ) → ( X, x0 ) of pointed topological spaces to the π1 ( X, x0 )-equivariant map
f | p̂−1 ({ x0 }) : p̂−1 ({ x0 }) → p̂0−1 ({ x0 }) that sends x̂ to x̂ 0 .
12.7. Proposition. Let X be a semilocally simply connected topological space with a point x0 ∈ X and let
p : ( X̃, x̃0 ) → ( X, x0 ) be a universal covering. Then the functors Γ p,• and Fibx0 ,• defined in paragraph 12.6 are
quasi-inverse to each other.
Proof. The proof is the exactly the same as that of Proposition 12.4.
12.8. It is easy to verify that the functors Γ p,• and Fibx0 ,• in Proposition 12.7 restrict to functors

Γ p,• : G-ft Set• -π1 ( X, x0 ) → Cov G ( X, x0 ) (6)

and

Fibx0 ,• : CovG ( X, x0 ) → G-ft Set• -π1 ( X, x0 ), (7)

which are quasi-inverse to other, where G-ft Set• -π1 ( X, x0 ) denotes the category of pointed G-π1 ( X, x0 )-
bisets such that the left action of G is free and transitive (see paragraph 8.10 and Exercise 12.5).
12.9. Proposition. Let X be a semilocally simply connected topological space with a point x0 ∈ X. Then there is a
bijection
 
 isomorphism classes of pointed 
 ( )
 morphisms of groups
G-covering spaces −→ .
π1 ( X, x0 ) → G
p̂ : ( X̂, x̂0 ) → ( X, x0 )

 


26
Proof. By Exercise 5.17, the equivalence given by the quasi-inverse functors (6) and (7) in paragraph 12.8
induces an isomorphism between the sets of isomorphism classes of the categories G-ft Set• -π1 ( X, x0 )
and Cov G ( X, x0 ). By paragraph 8.11, a G-π1 ( X, x0 )-biset structure on the left regular G-set G is exactly the
same as a morphism of groups π1 ( X, x0 ) → G. Moreover, it is easy to verify that there are no morphisms
between two pointed G-π1 ( X, x0 )-bisets corresponding to different morphisms of groups π1 ( X, x0 ) → G.
The proposition thus follows.

§13. Algebraic interlude III: Free amalgamated products

13.1. Given a collection of objects { Xi }i∈ I in a category C , a coproduct is an object C in C together with
morphisms u : Xi → C for all i ∈ I such that for every other object C 0 in C together with morphisms
ui0 : Xi → C 0 there exists a unique morphism w : C → C 0 such that ui0 = w ◦ ui for all i ∈ I.
13.2. Exercise. Let C be a category, and let { Xi }i∈ I be a collection of objects. Assume that (C, (ui )i∈ I ) and
(C 0 , (ui0 )i∈ I ) are two coproducts { Xi }i∈ I . Prove that there exists a unique isomorphism w : C → C 0 such that
ui0 = w ◦ ui for all i ∈ I. We denote the thus uniquely defined coproduct by äi∈ I Xi .
13.3. Let { Gi }i∈ I be a collection of groups, with I nonempty. By replacing if necessary Gi by Gi × {i } for
i ∈ I we assume that the previous family is disjoint. For i ∈ I, set Gi0 = Gi \ {eGi }, where eGi denotes the
neutral element of Gi . Given n ∈ N, define Injn = { f : J1, nK → I : f injective} and
 
Grn = g : J1, nK → Gi0 : there exists f ∈ Injn such that g(`) ∈ G0f (`) for all ` ∈ J1, nK .
G

i∈ I

We will denote an element g of Grn by ( g(1), . . . , g(n)). For n = 0, set Grn = {∗}. Define further
G
Gr = Grn .
n∈N0

We define a map µ : Gr × Gr → Gr as follows. First, µ(∗, w) = µ(w, ∗) = w, for all w ∈ Gr. Moreover,
given n, m ∈ N, g = ( g(1), . . . , g(n)) ∈ Grn and h = (h(1), . . . , h(m)) ∈ Grm we define µ( g, h) recursively
as follows. If g(n) ∈ Gi and h(1) ∈ Gj with i 6= j, we set




µ( g, h) = µ g(1), . . . , g(n) , h(1), . . . , h(m) = g(1), . . . , g(n), h(1), . . . , h(m) ∈ Grn+m .

If g(n), h(1) ∈ Gi but g(n)h(1) 6= eGi , where , we set





µ( g, h) = µ g(1), . . . , g(n) , h(1), . . . , h(m) = g(1), . . . , g(n)h(1), . . . , h(m) ∈ Grn+m−1 .

Finally, if g(n), h(1) ∈ Gi and g(n)h(1) = eGi , we define µ( g, h) recursively by



 


µ( g, h) = µ g(1), . . . , g(n) , h(1), . . . , h(m) = µ g(1), . . . , g(n − 1) , h(2), . . . , h(m) ,

where we set ( g(1), . . . , g(n − 1)) = ∗ if n = 1 and (h(2), . . . , h(m)) = ∗ if m = 1.

13.4. Exercise. Suppose the same assumptions as in the previous paragraph.
(i) Prove that Gr endowed with the operation µ is a group.
(ii) Given i ∈ I, let ui : Gi → Gr be the map sending eGi to ∗ ∈ Gr0 , and x ∈ Gi0 to the element g ∈ Gr1
satisfying that g(1) = x. Prove that ui is a morphism of groups for all i ∈ I.
(iii) Let {vi : Gi → G : i ∈ I } be a family of morphisms of groups. Prove that there exists a unique morphism of
groups v : Gr → G such that v ◦ ui = vi for all i ∈ I. Conclude that coproducts exist in the category of
groups.

27
13.5. In case of only two groups G1 and G2 , the coproduct of them is typically called the free product of
groups, and is denoted by G1 ∗ G2 . Given a set X and the family of groups { Gx } x∈ X with Gx = Z for
every x ∈ X, the coproduct ä x∈ X Gx is called the free group generated by X, and it is typically denoted
by h X i. A quotient of the free group h X i by the smallest normal subgroup including a subset R ⊆ h X i
will be denoted by h X | Ri.
13.6. Exercise. Let X be a set and h X i be the free group generated by X. Denote by u x : Z = Gx → h X i be the
morphism of groups given by the definition of coproduct. Let i X : X → h X i be the map sending x ∈ X to u x (1), for
all x ∈ X. Prove that, given any map f : X → G, where G is a group, there exists a unique morphism of groups
F : h X i → G such that F ◦ i X = f .
13.7. Given morphisms f : X → Y and g : X → Z in a category C , a push-out is an object P in C together
with morphisms u : Y → P and v : Z → P such that u ◦ f = v ◦ g, and such that for every other object
P0 in C together with morphisms u0 : Y → P0 and v0 : Z → P0 such that u0 ◦ f = v0 ◦ g there is a unique
morphism w : P → P0 such that u0 = w ◦ u and v0 = w ◦ v. Graphically,

f
X Y

g u
u0
v
Z P
∃w

v0
P0

13.8. Exercise. Let C be a category, and let f : Y → X and g : Z → X be two morphisms. Assume that ( P, u, v)
and ( P0 , u0 , v0 ) are two push-outs for ( f , g). Prove that there exists a unique isomorphism w : P → P0 such that
u0 = w ◦ u and v0 = w ◦ v.
13.9. By the previous exercise, if push-out exists, they are uniquely determined (up to unique isomor-
phism). On the other hand, push-outs also exist in the category Grp, as we now show. Indeed, given
morphisms f 1 : G → G1 and f 2 : G → G2 of groups, then
G1 ∗ G2
P= ,
N
where G1 ∗ G2 is the free product of the groups G1 and G2 , and N is the normal closure of the set
X = { f 1 ( x ) f 2 ( x )−1 : x ∈ G } ⊆ G1 ∗ G2 , i.e. the smallest normal subgroup of G1 ∗ G2 containing X. The
push-out of groups is sometimes called free product with amalgamation.

§14. Homotopy III: The Seifert-Van Kampen theorem

14.1. In this section we give a very simple proof of the Seifert-Van Kampen theorem, following [Ful1995].
The latter reference attributes the proof to A. Grothendieck. See [Mas1977], Chap. IV, Section 2, for more
general results.
14.2. Theorem. Let X be a path connected, locally path connected and semilocally simply connected space. Let
U, V ⊆ X be two nonempty open subsets such that U ∪ V = X, and U, V and U ∩ V are path connected and
semilocally simply connected. Then for every x ∈ U ∩ V the diagram

28
 π1 (incU ∩V,U )
π1 (U ∩ V, x0 ) π1 (U, x0 )

π1 (incU ∩V,V ) π1 (incU,X )

π1 (incV,X )
π1 (V, x0 ) π1 ( X, x0 )

is a push-out in the category of groups, where incY,Z : Y → Z is the inclusion map of a subset Y ⊆ Z.
Proof. We have to prove that given any group G and any commutative diagram
π1 (incU ∩V,U )
π1 (U ∩ V, x0 ) π1 (U, x0 )

π1 (incU ∩V,V ) π1 (incU,X )

u
π1 (incV,X )
π1 (V, x0 ) π1 ( X, x0 )

v G
of morphisms of groups, there exists a unique morphism of groups w : π1 ( X, x0 ) → G such that
u = w ◦ π1 (incU,X ) and v = w ◦ π1 (incV,X ). By Proposition 12.9, there exist unique pointed G-
coverings pU : (Ũ, x̃U ) → (U, x0 ) and pV : (Ṽ, x̃V ) → (V, x0 ) corresponding to the morphisms of
−1 −1
groups π1 (U, x0 ) → G and π1 (V, x0 ) → G, respectively. Let Ũ 0 = pU (U ∩ V ) and Ṽ 0 = pV (U ∩ V ), as
0 0
well as the pointed G-coverings pU |Ũ 0 : (Ũ , x̃U ) → (U ∩ V, x0 ) and pV |Ṽ 0 : (Ṽ , x̃V ) → (U ∩ V, x0 ). Under
the bijection in Proposition 12.9, the last pointed G-coverings correspond to the morphisms of groups
given by the compositions of

π1 (U ∩ V, x0 ) → π1 (U, x0 ) → G and π1 (U ∩ V, x0 ) → π1 (V, x0 ), → G, (8)

respectively. Since the composition of the maps in the left of (8) coincides with the composition of
the maps in the right, pU |Ũ 0 : (Ũ 0 , x̃U ) → (U ∩ V, x0 ) and pV |Ṽ 0 : (Ṽ 0 , x̃V ) → (U ∩ V, x0 ) are isomor-
phic. In consequence, there exists a unique G-equivariant morphism of pointed topological spaces
f : (Ũ 0 , x̃U ) → (Ṽ 0 , x̃V ) such that pV |Ṽ 0 ◦ f = pU |Ũ 0 . Define X̃ = (Ũ t Ṽ )/ ∼, where ∼ is the equivalence
relation on X̃ generated by u ∼ f (u) for u ∈ Ũ 0 , and endow it with the quotient topology. Denote by
y∼ ∈ (Ũ t Ṽ )/ ∼ the equivalence class of y ∈ Ũ t Ṽ, and set x̃0 = x̃V ∼ = x̃ ∼ . Let p : X̃ → X be the unique
U
map induced by the mapping Ũ t Ṽ → X sending u ∈ Ũ to pU (u) and v ∈ Ṽ to pV (v). It is clear that
p is continuous map, sends x̃0 to x0 and it is a G-covering. By Proposition 12.9, there exists a unique
morphism of groups w : π1 ( X, x0 ) → G such that the diagram
π1 (incU ∩V,U )
π1 (U ∩ V, x0 ) π1 (U, x0 )

π1 (incU ∩V,V ) π1 (incU,X )

u
π1 (incV,X )
π1 (V, x0 ) π1 ( X, x0 )
w

v G
commutes. The theorem follows.
14.3. Exercise. Let X be the topological space given in Exercise 10.11. Prove that π1 ( X, 4) = Z ∗ Z = h x, yi, the
free group on two generators.
14.4. Exercise. Let Y = [−1, 1] × [−1, 1] and let ∼ be the equivalence relation on Y generated by ( x, 1) ∼ ( x, −1)
and (−1, y) ∼ (1, −y), for all x, y ∈ [−1, 1]. Define K = Y/ ∼ the set of equivalence classes, provided with the

29
quotient topology. It is usually called the (topological) Klein bottle. Let U ⊆ K (resp., V ⊆ K) be the image of
[−1, 1]× ] − 1, 1[ (resp., Y \ ([−1, 1] × {0})) under the canonical projection Y → Y/ ∼. Let p ∈ U ∩ V.
(i) Prove that U and V are homeomorphic to the Möbius band M (see Exercises 7.15 and 9.20).
(ii) Prove that U ∩ V is homeomorphic to a cylinder S1 × ] − 1, 1[.
Hint. Prove that the map f : [−1, 1] × ( ] − 1, 1[ \{0}) → S1 × ] − 1, 1[ given by
(
(eπix/2 , 2y + 1), if ( x, y) ∈ [−1, 1]× ] − 1, 0[,
f ( x, y) =
(e πi ( 2x + 1 ) /2 , −2y + 1), if ( x, y) ∈ [−1, 1]× ]0, 1[

induces a homeomorphism U ∩ V → S1 × ] − 1, 1[.

(iii) Prove that the group morphisms

Z∼
= π1 (U ∩ V, p) → π1 (U, p) ∼
=Z and Z∼
= π1 (U ∩ V, p) → π1 (V, p) ∼
=Z
send 1 to 2.
Hint. Use Exercise 9.20, (iii).
(iv) Deduce the isomorphism of groups π1 (K, p) ∼
=h x, y| x2 y−2 i.
14.5. Given topological spaces X and Y, a closed subspace A ⊆ X and a continuous map f : A → Y,
consider the equivalence relation ∼ on X t Y generated by a ∼ f ( a) for all a ∈ A, and define X ∪ f Y as
the set of all equivalence classes endowed with the quotient topology. It is called the adjunction space
given by attaching Y to X along f .
14.6. Exercise. Let n ∈ N, X = S1 and Y = B̄1 (0, 0) ⊆ R2 . Consider the map f : X → Y sending z ∈ S1
to zn . Let Z = X t Y and ∼ the equivalence relation on Z defined in paragraph 14.5. Let Ũ = B1 (0, 0) and
Ṽ = ( B̄1 (0, 0) \ {(0, 0)}) t X. Define U and V as the image of Ũ and Ṽ under the canonical projection Z → Z/ ∼.
Let p ∈ U ∩ V.
(i) Prove that the group morphism Z ∼ = π1 (U ∩ V, p) → π1 (V, p) ∼ = Z sends 1 to n.
(ii) Deduce that π1 ( X ∪ f Y, p) ∼
= Z/nZ.

§15. Surfaces I: General theory

15.1. We will recall first some basic facts on the theory of surfaces. We refer the reader to the wonderful
book [Lee2011], or the nice exposition [Mas1977], for a more comprehensive treatment.
15.2. We recall that a (topological) manifold with boundary of dimension n ∈ N0 (or n-manifold
with boundary for short) is a Hausdorff second countable topological space M together with an
open covering U such that there are homeomorhisms of φ : U → V where V is an open subset of
Hn = {( x1 , . . . , xn ) ∈ Rn : xn ≥ 0}, for all U ∈ U . An element p ∈ M is called a boundary point if there
is a (equivalently, if for every) homeomorhism of φ : U → V with p ∈ U ∈ U such that φ( p) ∈ ∂Hn .
Denote by ∂M the set of boundary points. It is easy to see that ∂M is a closed subset. An n-manifold with
boundary M is called a (topological) manifold of dimension n (or n-manifold for short) if ∂M = ∅.
15.3. A coordinate ball of an n-manifold M is a homeomorphism φ : B → Br ( x0 ) ⊆ Rn , where B is an
open subset of M. We will typically denote a coordinate ball simply by B. We say that a coordinate ball
φ : B → Br ( x0 ) ⊆ Rn is regular if there exists an open set B0 ⊇ B̄ in M including the closure B̄ of B, r 0 > r
and a homeomorphism φ0 : B0 → Br0 ( x0 ) ⊆ Rn such that φ0 ( B) = Br ( x0 ). Note that the fact that φ0 is a
homeomorphism implies that φ0 ( B̄) = B̄r ( x0 ). A (topological) surface is a manifold of dimension 2.

30
15.4. Remark. We remark that there exist nonregular coordinate balls. For instance, there exists a subset S ⊆ R3 ,
called the Alexander horned sphere, such that
(i) S is homeomorphic to the usual sphere S2 ⊆ R3 ;
(ii) R3 \ S is the disjoint union of two connected components, B and C, with B bounded;
(iii) B t S is homeomorphic to the closed unit ball B̄1 (0R3 );
(iv) there exists s ∈ S such that for every open set U including s, the image of the map
π 1 (U ∩ C ) → π 1 ( C )
is not trivial.
The last condition implies that B ⊆ R3 is a nonregular coordinate ball of R3 .
15.5. Exercise. (i) Prove that S2 is a surface.
(ii) Prove that the torus T2 = S1 × S1 is a surface.
(iii) Prove that the (real) projective plane RP2 is a surface.
15.6. Given M and N be two n-manifolds with nonempty boundaries and f : ∂M → ∂N a homeomorphism,
one can prove that the adjunction M ∪ f N, recalled in paragraph 14.5, is an n-manifold (see [Lee2011],
Thm. 3.79). Given two connected n-manifolds M1 and M2 , regular coordinate balls Bi ⊆ Mi for
i = 1, 2, then the subspaces Mi0 = Mi \ Bi are manifolds with boundaries. Given a homeomorphism
f : ∂M1 → ∂M2 , by the previous comments, M10 ∪ f M20 is an n-manifold, called connected sum of
M1 and M2 , and is denoted by M1 #M2 . This definition depends in principle on the choice of regular
coordinate balls and the homeomorphism f between the boundaries, although one can prove that in fact
at most two nonhomeomorphic manifolds can be obtained when picking different coordinate balls and
homeomorphisms. This makes use of the so-called annulus theorem, a very hard theorem that states
that if B is a regular coordinate ball embedded in Br (0Rn ) ⊆ Rn , then B̄r (0Rn ) \ B is homeomorphic to
B̄2 (0Rn ) \ B1 (0Rn ). If n = 2, the two possible connected sums are in fact homeomorphic to each other
(see [Lee2011], Problem 10-8, for the compact case). One can also show that M1 #M2 is compact (resp.,
connected) if M1 and M2 are so (resp., and n > 1) (see [Lee2011], Problem 4-18).
15.7. We will describe a combinatorial way to construct surfaces. Let S be a set, S±1 = S × {±1},
S± = S+1 t S−1 and W = tk∈N S± k , the set of all words in S. An element of ( a, 1) is written simply a,

whereas ( a, −1) will be written a . We define ( a−1 )−1 = a. An element w = (w1 , . . . , wk ) ∈ S±

− 1 k for k ∈ N

will be simply written w = w1 . . . wk , whose length is `(w) = k. The set W has a natural structure of
(associative) monoid (without unit), given by concatenation of words, that we denote by yuxtaposition,
i.e. (w1 . . . wk ) · (w10 . . . wk0 0 ) = w1 . . . wk w10 . . . wk0 0 . A polygonal presentation P , written hS|w1 , . . . , wk i, is
a finite set S together with a finite tuple of words w1 , . . . , wk , each of which has length at least 3, such
that each element a ∈ S appears in at least one word (as a or a−1 ). A surface presentation is a polygonal
presentation such that every letter a ∈ S appear exactly twice in the whole list w1 , . . . , wk (as a or a−1 ). We
will also allow the surface presentations h a| aa−1 i and h a| a−1 ai (associated to the sphere S2 ), as well as
h a| aai and h a| a−1 a−1 i (associated to the projective plane RP2 ).
15.8. For n ∈ N and k ∈ Z, define ζ k,n = ie2πki/n ∈ C ∼ = R2 . For an integer n ≥ 3, we denote by Pn ⊆ R2
the regular convex polygon with n edges and set of vertices Vn = {ζ k,n : k ∈ J0, n − 1K}. It is endowed
with the subspace topology of R2 . Let En = {ek = [ζ k,n , ζ k+1,n ] : k ∈ J0, n − 1K} the set edges of Pn and
Ēn = {ēk = (ζ k,n , ζ k+1,n ) : k ∈ J0, n − 1K} be the set of oriented edges of Pn . If e is an usual edge, ē will thus
( x 0 ,y0 ) ( x 0 ,y0 )
denote the corresponding oriented edge. Given x, y, x 0 , y0 ∈ R2 , denote by L( x,y) , L̄( x,y) : [ x, y] → [ x 0 , y0 ]
the maps
( x 0 ,y0 )
L( x,y) (tx + (1 − t)y) = (tx 0 + (1 − t)y0 ),
( x 0 ,y0 ) (y0 ,x 0 )
for t ∈ [0, 1], and L̄( x,y) = L( x,y) .

31
15.9. Given a polygonal presentation hS|w1 , . . . , wk i, where wi = w11 . . . wi`(wi ) for all i ∈ J1, kK, consider
P = tik=1 P`(wi ) as well as the maps φ : tik=1 E`(wi ) → S± sending the j-th edge eij of P`(wi ) to wij , for i ∈ J1, kK
and j ∈ J1, `(wi )K. The geometric realization |P | of P = hS|w1 , . . . , wk i is the quotient topological space
of P, which is endowed with the disjoint union topology, by the equivalence relation ∼ on P generated by
0
x ∼ Lēē ( x ),

for all x ∈ e and for all pairs of edges e, e0 in P such that φ(e) = φ(e0 ), and by
0
x ∼ L̄ēē ( x ),

for all x ∈ e and for all pairs of edges e, e0 in P such that φ(e) = φ(e0 )−1 .
15.10. We will typically draw a polygonal presentations of the form hS|wi (or its geometric realizations)
as follows. We draw the polygon P`(w) . For every edge e in P`(w) , we orient it adding an arrow ē together
with the label φ(e), if φ(e) ∈ S+1 , and an arrow −ē together with the label φ(e)−1 , if φ(e) ∈ S−1 . For
example, you find below the corresponding graphical representation of h a1 , b1 , . . . , a5 , b5 | ∏5i=1 [ ai , bi ]i,
where we recall that [ x, y] = xyx −1 y−1 is the commutator of the elements x, y in a group.

a1 b5
b1 a5
a1 b5

b1 a5

a2 b4

b2 a4

a2 b4

b2 a4
a3 b3
b3 a3

15.11. Exercise. (i) Prove that S2 is homeomorphic to the geometric realization of the surface presentation
h a, b| abb−1 a−1 i.
(ii) Prove that T2 is homeomorphic to the geometric realization of the surface presentation h a, b| aba−1 b−1 i.
(iii) Prove that RP2 is homeomorphic to the geometric realization of the surface presentation h a, b| ababi.
(iv) Prove that Klein bottle K (see Exercise 14.4) is homeomorphic to the geometric realization of the surface
presentation h a, b| abab−1 i.
15.12. The following operations are called elementary transformations of the polygonal presentation.
(i) Relabeling : the result in P of applying a bijection S → S0 or applying the bijection S± → S±
sending a0±1 to a0∓1 for a fixed a0 ∈ S and a±1 to itself, for a ∈ S \ { a0 }.
(ii) Subdividing : replacing every occurrence in P of some a0 ∈ S by a0 e and a0−1 by e−1 a0−1 , where e
is a new element not in S± .
(iii) Consolidating : if elements a and b of S always occur in the words of P either as ab or b−1 a−1 ,
replacing every occurrence of ab by a and every occurrence of b−1 a−1 by a−1 , provided that the
result is one or more words of length at least 3 or a single word of length 2.

32
 (iv) Reflecting : hS| x1 . . . xn , w1 , . . . , wk i transforms into hS| xn−1 . . . x1−1 , w1 , . . . , wk i, for x1 , . . . , xn ∈ S± .
(v) Rotating : hS| x1 . . . xn , w1 , . . . , wk i transforms into hS| x2 . . . xn x1 , w1 , . . . , wk i, for x1 , . . . , xn ∈ S± .
(vi) Cutting : hS|w1 w2 , . . . , wk i transforms into hS t {e}|w1 e, e−1 w2 , . . . , wk i, provided w1 and w2 have
both length at least 2.
(vii) Pasting : hS t {e}|w1 e, e−1 w2 , . . . , wk i transforms into hS|w1 w2 , . . . , wk i.
(viii) Folding : hS t {e}|w1 ee−1 , w2 , . . . , wk i transforms into hS|w1 , w2 , . . . , wk i, provided w1 has length
at least 3, or length 2 if k = 1.
(ix) Unfolding : hS|w1 , w2 , . . . , wk i transforms into hS t {e}|w1 ee−1 , w2 , . . . , wk i.
15.13. A lengthy but elementary verification of topological invariance of each of the previous elementary
transformations provides the following result.
15.14. Proposition ([Lee2011], Prop. 6.10). Let P and P 0 be two polygonal presentations that can be linked
through a finite sequence of elementary transformations. Then, |P | is homeomorphic to |P 0 |.
15.15. Using the previous result for clever representations of the surfaces and some technical results on
triangulations of surfaces we get the following result.
15.16. Proposition ([Lee2011], Prop. 6.12). Let M be a surface homeomorphic to |P | for a surface presentation
P = hS|wi, and M0 be a surface homeomorphic to |P 0 | for a surface presentation P 0 = hS0 |w0 i. Then M#M0 is
homeomorphic to the geometric realization of the surface presentation hS t S0 |ww0 i.
15.17. Exercise. (i) Let M be a connected sum of g ∈ N copies of the torus T2 = S1 × S1 . Prove that M is
g
homeomorphic to the geometric realization of the surface presentation h a1 , b1 , . . . , a g , bg | ∏i=1 [ ai , bi ]i.
(ii) Let M be a connected sum of n ∈ N copies of the torus RP2 . Prove that M is homeomorphic to the geometric
realization of the surface presentation hc1 , . . . , cn | ∏in=1 c2i i.
15.18. Theorem ([Lee2011], Prop. 6.14, Thm. 6.15, or [Mas1977], Thm. I.5.1). Given a any compact surface
M there exists a surface presentation P = hS|w1 , . . . , wk i such that M is homeomorphic to |P |. Moreover, any
compact connected surface is homeomorphic to one of the following
(i) the sphere S2 ;
(ii) a connected sum of one or more copies of the torus T2 = S1 × S1 ;
(iii) a connected sum of one or more copies of the projective plane RP2 .
15.19. Exercise. (i) Prove that the Klein bottle K (see Exercise 14.4) is homeomorphic to RP2 #RP2 .
(ii) Prove that the T2 #RP2 is homeomorphic to RP2 #RP2 #RP2 .

§16. Surfaces II: Homotopy groups

16.1. We conclude these short notes with the statement of the fundamental group of any compact connected
surface M. Let P = h x1 , . . . , xm |wi be its surface presentation, according to Theorem 15.18, and Exercises
15.11 and 15.17. Define M̂ = im=1 (S1 , 1), the wedge product of m copies of (S1 , 1). By the previous
W

results, there is a continuous map f : ∂ B̄1 (0R2 ) → M̂ such that M is homeomorphic to B̄1 (0R2 ) ∪ f M̂.
Denote M̄ = B̄1 (0R2 ) t M̂. Take then U, V ⊆ B̄1 (0R2 ) ∪ f M̂ to be the image under the canonical projection
M̄ → B̄1 (0R2 ) ∪ f M̂ of ( B̄1 (0R2 ) \ {0R2 }) t M̂ and B1 (0R2 ), respectively. Then V is contractible, so simply
connected, and using Theorem 14.2 we get the following result.

33
16.2. Theorem ([Lee2011], Cor. 10.17, or [Mas1977], Chap. IV, Section 5). Omitting the general point of the
following compact connected surfaces, the corresponding fundamental group is given by
(i) π1 (S2 ) = 1;
g
(ii) π1 (T2 # . . . #T2 ) ∼
=h a1 , b1 , . . . , a g , bg | ∏i=1 [ ai , bi ]i, where we are considering g factors in the connected sum
and we recall that [ x, y] = xyx −1 y−1 is the commutator of the elements x, y in a group;
(iii) π1 (RP2 # . . . #RP2 ) ∼
=hc1 , . . . , cn | ∏in=1 c2i i, where we are considering n factors in the connected sum.

34
 References
[Ful1995] William Fulton, Algebraic topology, Graduate Texts in Mathematics, vol. 153, Springer-Verlag,
New York, 1995. A first course. MR1343250
[GH1981] Marvin J. Greenberg and John R. Harper, Algebraic topology, Mathematics Lecture Note Series,
vol. 58, Benjamin/Cummings Publishing Co., Inc., Advanced Book Program, Reading, Mass.,
1981. A first course. MR643101
[Kel1975] John L. Kelley, General topology, Springer-Verlag, New York-Berlin, 1975. Reprint of the 1955
edition [Van Nostrand, Toronto, Ont.]; Graduate Texts in Mathematics, No. 27. MR0370454
[Lee2011] John M. Lee, Introduction to topological manifolds, 2nd ed., Graduate Texts in Mathematics,
vol. 202, Springer, New York, 2011. MR2766102
[Mac1971] Saunders MacLane, Categories for the working mathematician, Springer-Verlag, New York-Berlin,
1971. Graduate Texts in Mathematics, Vol. 5. MR0354798
[Mas1977] William S. Massey, Algebraic topology: an introduction, Springer-Verlag, New York-Heidelberg,
1977. Reprint of the 1967 edition; Graduate Texts in Mathematics, Vol. 56. MR0448331
[Mun1984] James R. Munkres, Elements of algebraic topology, Addison-Wesley Publishing Company, Menlo
Park, CA, 1984. MR755006
[Mun2000] James R. Munkres, Topology, Prentice Hall, Inc., Upper Saddle River, NJ, 2000. Second edition
of [ MR0464128]. MR3728284
[Mur2006] Daniel Murfet, Foundations for category theory (2006), 9 pp.
[Wed2016] Torsten Wedhorn, Manifolds, sheaves, and cohomology, Springer Studium Mathematik—Master,
Springer Spektrum, Wiesbaden, 2016. MR3560225

35
 Index
G-H-biset, 15 Group action, 13
G-covering, 17 covering, 17
G-set, 13 effective, 13
pointed, 25 free, 14
transitive, 13
Automorphism, 6
Hawaiian earring, 22
Category, 5 Homotopy, 7
source map, 5 as paths, 9
target map, 5 Homotopy category (naive), 8
with set of isomorphic classes of objects, 7
Component Isomorphism, 6
connected, 3 Isotropy group, 13
path connected, 4
Connected sum, 29 Klein bottle, 28
Coordinate ball, 29
Loop, 4
regular, 29
Coproduct, 26 Möbius band, 12
Covering, 15 Manifold, 29
locally finite, 1 with boundary, 29
normal, 21 Map
pointed, 16 G-equivariant, 14
regular, 21 closed, 1
trivial, 16 homotopy equivalence, 8
universal, 16 null homotopic, 8
open, 1
Degree (of a covering), 15
proper, 2
separated, 1
Fiber, 16
Monodromy action, 20
Free group, 26
Morphism, 5
Free product, 26
identity morphism, 5
with amalgamation, 27
inverse, 6
Functor, 6
lifting, 17
composition, 7
of G-coverings, 17
dense, 7
of G-sets, 14
equivalence, 7
fully faithful, 7 Natural isomorphism, 7
identity, 6 Natural transformation, 7
naturally isomorphic, 7 Normalizer, 14
pointed quotient, 10
quasi-inverse, 7 Object, 5
quotient, 8 isomorphic, 6
Fundamental group, 10 Orbit, 13

36
Path, 4 Wedge product, 11
composable, 4 Word, 30
concatenation, 4 length, 30
end point, 4
initial point, 4
path-homotopic, 9
piecewise affine, 12
product, 4
starting point, 4
Point
boundary, 29
generic, 9
star center, 9
Preojective space, 16
Presentation
polygonal, 30
surface, 30
Push-out, 26

Realization
geometric, 30

Space
ajunction, 29
compact, 2
connected, 3
contractible, 9
hyperconnected, 3
irreducible, 3
locally connected, 3
locally contractible, 9
locally path connected, 4
locally simply connected, 12
of pointed loops, 10
path connected, 4
pointed, 10
semilocally simply connected, 22
simply connected, 12
sober, 9
Space of morphisms, 5
Subset
convex, 9
relatively Hausdorff, 1
star-shaped, 9
Surface, 29

Topological group, 11
Transformation
elementary, 31

37