Chapter 48

Topology

48.1 Main definitions

As for notations, when we have a topological space X and a point x P X, we denote by Vpxq
the set of neighbourhoods of x. If F is a subset of X, we denote by F the closure of F and by
Int F , its interior.

Définition 48.1.
Let O be a topology on a set X. A subset B Ä O (the elements of B are subsets of X) is a basis
of topology O of X when any element of O is an union of elements of B.

Définition 48.2.
If X is a topological space, a subset A Ä X is a topological subspace when we consider on A the
induced topology from X. This is the topology on A where then open sets are the subsets of A
which can be written as A X O for an open set O Ä X in the sense of the topology on X.

Remarque 48.3.
In some literature, a topological subspace of X is just a subset A endowed with a topology for
which the inclusion ÿ : A Ñ X is continuous. This is a less restrictive condition.

From this definition, we see that the topology on a topological subspace is at least the induced
one. Indeed if O is an open subset of X, the set i´1 pOq must be open in A, but it is clear that
i´1 pOq “ A X O.

48.1.1 Separability axioms

The following definitions are the separability axioms. A topological space X is said to be
T0 if for every pair of points in X, at least one of them has an open neighbourhood which does
not contain the other.
The topological space X is T1 if every point is closed or, equivalently, when for every two
points, each has a neighbourhood which does not contain the other.
The topological space X is T2 if every two points belong to disjoint neighbourhoods.

48.1.2 Axioms of countability

A topological space X is
(1) first countable if every point has a countable neighbourhood basis,
(2) second countable if the topology has a countable basis,
(3) separable if there exists a countable dense subspace.

Lemme 48.4.
If Ï : X Ñ Y is a linear map between two normed vector spaces such that
}ÏpAq} § }A},

2075
2076 CHAPTER 48. TOPOLOGY

then Ï is continuous.

Proof. Let U be open in Y , namely, consider that it is a ball centered at y P Y with ray ”; if
x P Ï´1 pUq, we have to prove that A posses a neighbourhood contained in Ï´1 pUq. Let’s see the
condition on a P X for x ` a to belongs to Ï´1 pUq. We have

}Ïpx ` aq ´ y} § }Ïpxq ´ y} ` }a},

but }Ïpxq´y} § ”. Taking }a} suitably small, it is always possible to x`a to belong to Ï´1 pUq.

Définition 48.5.
If V and W are topological vector spaces, the map f : V Ñ W is uniformly continuous when
for every neighbourhood B of zero in W , there exists a neighbourhood A of zero in V such that

v1 ´ v2 P A ñ f pv1 q ´ f pv2 q P B.

Définition 48.6.
A topological space is separable when it admits a countable basis.

Lemme 48.7. î
Let be a family of subsets of a set X such that X “ AP A.
There exists a topology on X of basis if and only if the intersection of any finite subfamily
of can be written as the union of elements of .

Proof. The topology is simply the one where the open sets are the unions of elements of . The
difficulty is to show that it defines actually a topology. We don’t show it here.

Lemme 48.8.
A metric space pE, dq is separable if and only if it contains a dense sequence.

Proof. Necessary condition. If the countable basis is pBi q, we take a xi in each Bi . For any x P E
and open set A containing x, there is a Bi Ä A and then a xi P A. Thus the sequence pxi q is dense
in E.
Sufficient condition. Let pxi q be the dense sequence. We consider Bi,k , the open ball centred in xi
of radius 1{k. This is a countable system of open set in E. Let A be an open set in E and S, the
union of all the Bi,k which are contained in A. It is clear that S Ä A. Let x P A and D, an open
ball of radius ” chosen in order to have D Ä A. We choose k P such that 1{k † ” and xi such
that dpxi , xq † 1{k. Then x P S; namely x belongs to a ball of center xi and of radius 1{k.

Définition 48.9.
A subset of a topological space is connected if it cannot be written as an union of disjoint open
sets.

If A is a part of the topological space X, the boundary is the subset of X defined by

FrpAq “ A X AA

Lemme 48.10.
In a topological space, any path connected part is connected.

Théorème 48.11.
Let E be a topological space and A, B two connected parts of E. If A intersects B and AB, then A
intersects FrpBq.

Proposition 48.12.
Let X be a topological space and A, B two subsets of X such that AYB is closed and there exist A1 ,
B 1 , two disjoint sets containing respectively the closure of A and B. Then A and B are separately
closed.
48.1. MAIN DEFINITIONS 2077

Proof. We consider a converging sequence in A. If x is the limit, x P A Y B and x P A Ä A1 .

But it is clear from the assumptions that A1 X pA Y Bq “ A. Then x P A, which proves that A is
closed.

Définition 48.13.
A topological space E is compact when from any open covering, one can extract a finite subcov-
ering.

Lemme 48.14.
Let X be a compact topological space and A Ä X. In this case, A is closed in X if and only if A
is compact.

Proposition 48.15.
Let X be a topological space, K Ä X a compact and „ : K Ñ X a continuous map. Then „pKq is
compact in X.

Définition 48.16.
A family tXi uiPI of subsets of a topological space E is said to be locally finite if each x P E has
a neighbourhood V such that V X Xi ‰ H for only finitely many values of i.

Définition 48.17.
A topological space E is paracompact if every open covering admits a locally finite subcovering.

Définition 48.18.
A topological space E is locally compact if @x P E, there exists a compact neighbourhood of x in
E.

For example, a (non-empty) open subset of n is never compact 1 but always locally compact.

Définition 48.19.
A subset A of a metric space is relatively compact when A is compact.

Définition 48.20.
An Hausdorff space is a topological space in which any two distinct elements have disjoint
neighbourhoods.

Lemme 48.21.
If a set is Hausdorff for a topology ·1 and compact in a weaker topology ·2 , then ·1 “ ·2 .

Proposition 48.22.
If K is a compact topological space and E a separated 2 one and if f : K Ñ E is a continuous
bijection, then f is an homeomorphism.

A space is normal when for all disjoint closed subset E and F of the topological space X,
there exist neighbourhoods U and V of E and F which are disjoint too. A topological result says
that a compact Hausdorff space is normal.

Lemme 48.23.
If X is a normal space, F a closed subset of X and U an open set containing F , then there exists
a continuous function f : X Ñ r0, 1s such that f pF q “ 0 and f pXzUq “ 1.

Définition 48.24.
A metric space is complete when any Cauchy sequence is converging in the space.

Pay attention not to be confused with a characterization of a closed subset: the limit of any
converging sequence lies in the space.
1. The compacts subsets of n are the closed and bounded subsets.
2. Comment on dit un espace séparé?
2078 CHAPTER 48. TOPOLOGY

Définition 48.25.
Let E and F be two spaces in duality 3 ; if A Ä E and B Ä F , we say that A separates B if
@x ‰ y P B, Da P A such that âpxq ‰ âpyq where the hat denotes the duality.

For an Hausdorff X space, we denote by CpXq the space of all continuous functions on X. For
a locally compact space X, the set C0 pXq contains continuous functions which decrease to zero at
infinity in the sense that for any Á ° 0, there exists a compact set K such that |f pXq| † Á for all
x outside K.

Définition 48.26.
If X is an Hausdorff compact space, one can define the supremum norm on CpXq by

}f }8 :“ sup |f pxq|. (48.1)

xPX

Définition 48.27.
Let E be a vector space on or . A seminorm on E is a map p : E Ñ such that @x, y P E,
@⁄ P ,
(1) ppxq • 0
(2) pp⁄xq “ |⁄|ppxq
(3) ppx ` yq § ppxq ` ppyq.

A seminorm is a norm without the condition ppxq “ 0 if and only if x “ 0.

Définition 48.28.
An ecart 4 on a set E is a map d : E ˆ E Ñ such that @x, y, z P E,
(1) dpx, xq “ 0
(2) dpx, yq • 0
(3) dpx, yq “ dpy, xq
(4) dpx, zq § dpx, yq ` dpy, zq.

It is a distance minus the condition dpx, yq “ 0 if and only if x “ y. It is clear that dpx, yq “
ppx´yq is an ecart on E when p is a seminorm. Moreover, this d is invariant under the translations
and fulfils dp⁄x, ⁄yq “ |⁄|dpx, yq.

Proposition 48.29.
Let p be a seminorm on a topological vector space. Then p is continuous if and only if there exists
a neighbourhood of o in which p is bounded.

From a family of ecarts, we can define a topology. Let pd– q–PI be a family of ecarts on the set
E. For an element a P E, a finite subset p–j qj“1...m of I and some numbers prj qj“1...m ° 0, we
write
Bpa; p–j q, prj qq “ tx P E ˚ d–j pa, xq † rj , @j “ 1, . . . , mu.
A subset U Ä E is said open by respect of the family p–i qiPI when for any x P U , there exists a
sub-family p–j q and some numbers prj q ° 0 such that Bpx; p–j q, prj qq Ä U .

Définition 48.30.
When a topology can be defined from a family of ecarts, it is said an uniformisable topology.

If we have a family pp– q–PI of seminorm, we can build the ecarts d– px, yq “ p– px ´ yq. One
can show that this topological structure is compatible with the vector space structure.
3. I’m not sure what it means.
4. C’est pas joli la traduction du mot “Écart“.
48.2. TOPOLOGY AND CONVERGENCE 2079

Définition 48.31.
A topological vector space whose topology can be defined from a family of seminorm is said locally
convex.

Définition 48.32.
A metric space is separable if there exists a subset at most countable which is everywhere dense.

Proposition 48.33.
If E is a metric locally compact space, then the following are equivalent:
(1) There exists an increasing
î sequence pAn q of open relatively compacts subset of E such that
An Ä An`1 and E “ n An ,
(2) E is a countable union of compact subsets,
(3) E is separable.

If A is a topological space with an equivalence relation „ and the canonical projection Ï : A Ñ

A{ „, then V Ä A{ „ is open in then quotient topology if and only if Ï´1 pV q Ä A is open.

48.2 Topology and convergence

When we have a topological space X, we can define a convergence notion for the sequences.

Définition 48.34.
We say that the sequence pxn q in X converges to x P X if @V P Vpxq, Dn0 such that n • n0 implies
xn P V .

Définition 48.35.
If X and Y are two topological spaces, a map f : X Ñ Y is said continuous if f ´1 pUq is open in
Y for any open set U in X.

If f is continuous and U open in X, it is not true that f pUq is open in Y . Simple counter-
example are given by the constants functions from to . So it is false that continuity preserves
the openness. But it preserves the convergence.

Proposition 48.36.
Let ak be a converging (ak Ñ a) sequence in X and g : X Ñ Y , a continuous map. Then gpaq “
limkÑ8 gpak q.

Proof. If U is an open in Y which contains gpaq, g ´1 pUq is an open subset of X which contains
a. Then from a certain N , ak P g ´1 pUq, so that gpak q P U; it proves that the sequence gpak q in Y
converges to gpaq in the sense of the topology of Y .

Let us define a notion which depend only on the convergence notion.

Définition 48.37.
An usinage around x P X is a subset U Ä X which contains x P U and such that for any
convergent sequence pan q Ñ x, there exists n0 such that n ° n0 implies an P U.

Lemme 48.38.
Let X be a metric topological space. The topology is completely determined by his convergence
notion.
An other way to state this proposition: there is only one metric topology which gives rise to a
given convergence notion.

Proof. We will show that, provided that the topology comes from a metric, the notion of neigh-
bourhood is the same as the notion of usinage.
On the one hand, from topological definition 48.34 of convergence, it is clear that any neigh-
bourhood is an usinage.
2080 CHAPTER 48. TOPOLOGY

On the other hand, any usinage must contains a metric ball centred at x and then is a neigh-
bourhood. Indeed, if U doesn’t contains any ball, then

@‘ ° 0, Dap‘q such that dpap‘q, xq † ‘ and ap‘q R U.

In such a case, any sequence pap‘1 q, . . .q with ‘n Ñ 0 contradict the fact that U is an usinage.

Proposition 48.39.
For a given notion of usinage (i.e. for a given notion of convergence) on any set, there exists a
topology (maybe no metric!) for which the usinage form a basis.

Proof. This is a direct application of lemma 48.7 and the fact that the intersection of two usinages
is an usinage.

The fact that a convergence notion gives rise to a topology is summarized by the following
scheme:

Basis of topology (always)

3

Convergence / Usinage

+
Unique metric topology (when it exists)

Let U be an open subset of n.

Définition 48.40.
A increasing sequence pKm q of compacts subset of U is fundamental if Km Ä Int Km`1 for all
m and §
Km “ U
m

Proposition 48.41.
On E pUq, there exists one and only one structure of locally convex metrisable Hausdorff space
whose convergence notion is:
pfk q Ñ 0 in E pUq when for any compact K Ä U and any multi-indices ‹, the sequence of
restriction pD‹ fk |K q converges uniformly to 0.
Here the D‹ denotes the multi-derivation with respect to the coordinates contained in the multi-
index ‹.

Proof. The unicity part is given by lemma 48.38. Now, we will build such a topology. First remark
that we can find a fundamental sequence Km in U. Next, for any couple of integer s • 0, m ° 0
and for any function f P E prq pUq with r • s, we define

ps,m pf q “ sup |pD‹ f qpxq|.

xPKm
|‹|§s

One can see that these ps,m are seminorms, moreover ps,m § pr,m . Remark that if p0,m pf q “ 0
for any m, f “ 0 on any Km , and then f “ 0 in U. Thus p0,m “ 0 for all m implies ps,m pf q “ 0.
Now, we consider only the ps,m with 0 § s § m and the induced topology on E pUq.
First, we prove that this topology is Hausdorff. Let us consider f , g P E pUq, and suppose that
any ball centred at f contains g. Then we must obtain f “ g. For any sequences psi q, pri q with
0 § si § ri and for any sequence p‘j q ° 0,

g P Bpf ; ppsi ,ri q, p‘j qq “ tx P E pUq| dpsi ,ri pf, xq † ‘j , @ju.

In particular, for a sequence ‘j Ñ 0 and si “ 0, the condition becomes p0,ri pf ´ gq † ‘j for any f .
Then p0,ri pf ´ gq “ 0 and f “ g.
48.3. METRISABLE GROUPS 2081

By the way, remark that this topology is almost the “supremum norm“ topology, then one can
guess that the convergence notion will be something as the uniform convergence. There are just
some subtleties as “for any compact“ or “for any derivatives“. But these are just what we need in
the statement. Let us be more precise.
Definition 48.13 of a compact subset makes that any compact subset of U is a subset of one of
the Km . This makes our topology independent of the choice of the fundamental sequence Km .
Let us consider pfk q, a sequence in E pUq. The condition to converge to 0 in the sense of our
new topology is that if A is an open subset of E pUq which contains 0, there exists a k0 such that
k ° k0 implies fk P A.
Then for any real sequences p‘j q ° 0 and psi ,ri with 0 § si § ri , fk P Bp0; ppsi ,ri q, p‘j qq. Thus,
Dk0 such that k ° k0 implies that @sj , rj ,

sup |D‹ fk pxq| † ‘j .

xPKrj
|‹|§sj

Taking a sequence ‘j Ñ 0, we get the thesis.

Définition 48.42.
A subset H Ä E pUq is bounded is each of the seminorm ps,m is bounded in H.

One can see that this notion only depend on the topology of E pUq. Let us once again give
some properties without proof.

Proposition 48.43.
There exists a sequence of functions in E pUq with compact support which is dense in each E prq pUq
and in E pUq. With others words, the spaces E prq pUq and E pUq are Fréchet separable spaces.

Proposition 48.44.
Any bounded subset of E pUq is relatively compact in E pUq.

Proposition 48.45.
For any multi-index ‹, the linear map D‹ : E pUq Ñ E pUq, f Ñ D‹ f is continuous.

48.3 Metrisable groups

Let G be any group. The function f : G ˆ G Ñ is said left invariant if for any x, y, z P G,
f pxy, xzq “ f py, zq. If G is abelian, the notion of left invariant and of right invariant are the same;
in this case, we say it to be translation invariant.
For example, when a distance is left invariant, the left translations are isometries. For example,
in a normed space vector, the distance dpx, yq “ }x ´ y} is always translation invariant: }px ` aq ´
py ` aq} “ }x ´ y}.

Proposition 48.46.
Let G be a topological group.
— G is metrisable if and only if there exists a countable fundamental system of neighbourhoods
of e (the neutral) whose intersection is only e.
— In this case, the topology of G ca be defined by a left invariant distance or a right invariant
distance.

On a metrisable topological group, one has thus two types of distance. But in general, one can
not find a distance which defines the topology being left and right invariant.

Proposition 48.47.
Let G be a metrisable group. A sequence pan q in G is a Cauchy sequence for a left invariant
distance if and only if
@ V P Vpeq, Dn0 ° 0 such that x´1
n xm P V @n, m • n0 .
2082 CHAPTER 48. TOPOLOGY

The point is that if a sequence is a Cauchy sequence for a left invariant distance, then it
is a Cauchy sequence for any left invariant distance. This yields the important conclusion: in
metrisable groups, the notion of Cauchy sequence is a topological notion which only depends on
the neighbourhood notion.
A right invariant Cauchy sequence is the same but with xn x´1
m instead of xn xm . By definition
´1

of a topological group, the map x Ñ x´1 is continuous. Thus, if all the left Cauchy sequences are
convergent, then all the right Cauchy sequences are convergent. In this case, the group G is said
complete.
If the group is abelian, we have only one notion of Cauchy sequence, and it is independent of
the metric.

48.4 One point compactification

Définition 48.48.
Let X be a non compact topological space. We consider the “extended” space X̃ :“ X Y t8u on
which we put the following topology. Open sets of X are open in X̃ and when O Ä X̃ contains 8,
it is open if and only if the complementary in is compact in X. It is easy to see that X̃ is compact.
It is the one point compactification of X.

Let us now consider X̃, any compact space and X “ X̃zt8u where 8 is a point in X̃. If we
put on X the induced topology from X̃, we get a locally compact space whose compactification is
X̃.

Exemple 48.49
When X “ we add one infinity to get the compactification ˜ “ Y t8u which is isomorphic
1
to the circle S .
This is not the same space as ¯ “ Y t`8, ´8u which is not compact. —

In this context we make the difference between 8 and `8. When we write `8 we are speaking
of the infinity in ¯ while writing 8 with no sign we are speaking about the infinity in the one
point compactification. The difference is important since for the one point compactification of
we have the limit
1
lim “ 8 (48.2)
xÑ0 x

while for the topology on ¯ the limite does not exist.

48.4.1 Example: the Riemann sphere

We already defined the Riemann sphere as a projective space in the definition 24.56.

Définition 48.50.
The Riemann sphere is the one point compactification of . We denote it by ˆ .

See the definition 48.48 for the topology.

Proposition 48.51.
Let “ : Ñ ˆ be a path in ˆ . We have limtÑ8 “ptq “ 8 if and only if }“ptq} Ñ `8.

Proof. To be clear: the first limit is a limit in ˆ for its one point compactification topology while
the second limit is an usual limit in .
ñ Let M ° 0. Since “ptq Ñ 8, there exists a T such that t ° T implies
c
“ptq P BpO, M q (48.3)

because this is an open set around 8. For that t ° T we thus have }“ptq} ° M . This
proves that }“ptq} Ñ `8.
 48.5. TOPOLOGICAL APPROXIMATION 2083

 Let O be an open neighbourhood of 8 in ˆ . This Oc is compact in ; Oc is then bounded 5

and there exists M such that Oc Ä Bp0, M q.
There also exists T such that t ° T implies }“ptq} ° M . Thus for t ° T we also have
“ptq P Bp0, M qc Ä O.

1 1
As far as arithmetic is concerned, the proposition 48.51 makes us define 0 “ 8 and 8 “ 0.

Remarque 48.52.

2 SageMath Version 7.0 , Release Date : 2016 -01 -19

3 Type " n o t e b o o k ( ) " f o r the browser - based notebook interface .
4 Type " h e l p ( ) " f o r h e l p .
5

6 sage : f ( x ) =1/ x
7 sage : limit (f , x =0)
8 x | - - > Infinity
9 sage : g ( x ) = x **2
10 sage : limit (g , x = oo )
11 x | - - > + Infinity
tex/sage/sageSnip002.sage

Sage makes the difference between `8, ´8 and 8.

In the first case, 1{x is seen in ˆ and the limit is 8, while in the second case, the limit is
understood in and sage provides `8.

Définition 48.53 (Pole of a complex function [437]).

Suppose U is an open subset of the complex plane , p is an element of U and f : U ztpu Ñ is
an holomorphic function over its domain. If there exists a holomorphic function g : U Ñ , such
that gppq is nonzero, and a positive integer n, such that for all z P U ztpu,

gpzq
f pzq “ (48.4)
pz ´ pqn

holds, then p is called a pole of f . The smallest such n is called the order. A pole of order 1 is
called a simple pole.
A function f : ˆ Ñ which is holomorphic on a neighbourhood of 8 has a pole at z “ 8 if
the function z fiÑ f p1{zq has a pole at z “ 0.

48.5 Topological approximation

48.5.1 Introduction
In this section, we follow [438]. Let consider a particle moving on a circle S 1 , and suppose that
we have three detectors, each of them cover a surface, let
fi 2fi fi 4fi
U1 “s ´ , r U2 “s , r, U3 “sfi, 2fir.
3 3 3 3
If the detectors U1 and U2 are react, we know that the particle is in U1 X U2 (which is an open
set), but no more; while if the detector U1 reacts alone, we know that the particle is not in U2 or
U3 , so that the particle belongs to ` ˘
U1 z U2 X U3 ,
2
5. The theorem 7.7 applies because has the topology of .
2084 CHAPTER 48. TOPOLOGY

which is closed. Our experimental setting allows us to distinguish six parts of S 1 that we name in
the following way

U1 X U3 ބ РU1 zpU2 Y U3 q ބ a
U1 X U2 fiÑ — U2 zpU1 Y U3 q fiÑ b
U2 X U3 fiÑ “ U3 zpU1 Y U2 q fiÑ c.

Notice that each point in S 1 belongs to one and only one of these sets. One says that a point of
S 1 is open when it belongs to –, — or “ and that it is closed when it belongs to a, b, c. Now let us
consider the six point space
P “ t–, —, “, a, b, cu,
on which we put the topology induced from S 1 . A basis of that topology is given by

t–u, t—u, t“u, t–, a, —u, t–, c, “u, t—, b, “u.

48.5.2 Generalization
Let pX, U⁄ q, a topological space. We define the equivalence relation x „ y if and only if
x P U⁄ ô y P U⁄ for every ⁄. In other words, x is equivalent to y when they cannot be distinguished
by the topology. Now we consider the quotient space

PU pXq “ X{ „

with its quotient topology: A Ä PU pXq is open if and only if fi ´1 pAq is open in X where fi : X Ñ
PU pXq is the canonical projection. It is the finest topology in which fi is continuous.
When X is compact, the covering U⁄ can be finite, so that PU pXq is finite. If X is only locally
finite, the space PU pXq will be countable and it is said to be unitary. Notice that PU pXq is not
Hausdorff in general, for example in P6 pS 1 q, the point a cannot be separated from –. Neither it is
T1 because points –, — and “ are open (in genera, it is even possible to get points which are open
neither closed). The space PU pXq is however always T0 .

48.5.3 Order and topology

When P is a finite topological space, the collections · of open sets in P is closed under arbitrary
unions and intersections. For each x P P we can consider the set
£
pxq “ tO P · tel que x P Ou (48.5)

which is the smallest neighbourhood of x. We define the relation x ® y if and only if pxq Ñ pyq,
or equivalently if and only if every open neighbourhood of y contains x. Yet another way to express
it is x ® y if and only if y P txu where the bar denotes the closure. The relation ® is reflexive and
transitive.
The space P is T0 by assumptions, so that for every couple of points x, y P P , there always is
a neighbourhood of x which does not contain y (or vice versa). That makes the relation antisym-
metric:
x ® y and y ® ñ x “ y.
So the relation ® is reflexive, antisymmetric and transitive. Such a relation is a partial order; a
partially ordered set is often called poset for short.
The construction can be carried in the inverse sense: if pP, ®q is a poset, we define pxq “
ty P P tel que y ® xu and we define an open set in P as an union of such subsets. The resulting
topological space is T0 .
A map f : P Ñ Q between two posets is continuous if and only if it preserves order:

x ®P y ñ f pxq ®Q f pyq. (48.6)

48.6. DIRECTED SETS AND NET 2085

When x ® y and x ‰ y, we write x † y.

With every poset, there is a Hasse diagram associated. It is obtained by drawing one point
for each element of the poset on different levels and lines between them respecting the following
two rules:
(1) if x † y, then the point associated with x has to be one level below the one of y,
(2) if x † y and if there does not exist z such that x † z † y then the point of x in one level
bollow the one of y and the two points are connected by a line.
Let us draw the Hasse diagram for P6 pS 1 q “ ta, b, c, –, —, “u. We have

p–q “ t–u p—q “ t—u p“q “ t“u

paq “ t–, a, —u pbq “ t—, b, “u pcq “ t–, c, “u

and

– † a, –†c —†a —†b “†b “ † c,

so that the Hasse diagram of P6 pS 1 q is

a b c
‚ ‚ ‚

‚ ‚ ‚
– — “

The smallest open set containing a point x is the set of points below x which are connected to x
by a sequence of lines.

48.6 Directed sets and net

A directed set is a pre-ordered set (i.e. a set with a reflexive and transitive binary relation)
such that every pair of elements has an upper bound. As a consequence, when a1 , . . . an are
elements of the directed set A, then there exists a a P A such that a • ai .
A net is a map A Ñ X from a directed set to a topological space. We denote by x– the element
of X which corresponds to – P A.
As example, if S is any set, the set A of finite subsets of S with the inclusion is an example of
net.
There is a notion of convergence of net. We say that the net Рބ xРconverges to x and we
write x– Ñ x if and only if for every open set U Ñ X containing x, there exists a – P A such that
–1 • – implies x–1 P U.
So a topology implies a convergence notion for nets, as well as for sequences. However, there ex-
ists different topologies which have the same notion of convergence of sequences, but two topologies
having the same notion of convergence of nets are the same.

48.7 Homotopy group

Let X be a topological space with a base point b, and S n be the n-sphere. The nth group of
homotopy on the point b of X is

fin pX, bq “ thomotopy classes of maps f : S n Ñ X such that f paq “ bu. (48.7)

The classes are taken up to homotopy, i.e. continuous deformations. In an equivalent way, fin pX, bq
can be seen as the set of classes of maps p : r0, 1sn Ñ X which sent the whole border of the cube
to b.
2086 CHAPTER 48. TOPOLOGY

48.8 Covering spaces

Let X be a topological space. A topological space C is a covering of X if we have a continuous
surjective map fl : C Ñ X such that @x P X, there exists an open neighbourhood U such that
fl´1 pUq is an union of disjoint open sets Si on which the restricted map fl|Si : Si Ñ U is an
homeomorphism.

Proposition 48.54 (lifting property). ` ˘

Let fl : C Ñ X be a covering and “ : r0, 1s Ñ X, a continuous map. Let c P fl´1 “p0q . Then there
exists one unique path ‡ in C such that ‡ ˝ fl “ “ and ‡p0q “ 0.

Proof. No proof.

If x and y in X are connected by a path, the lifted path provides a bijection between the fibres
fl´1 pxq and fl´1 pyq.

48.8.1 Universal covering

One says that a covering q : D Ñ X is universal if D is simply connected. The following
proposition states that an universal covering is a covering that covers all other coverings.

Proposition 48.55.
Let q : D Ñ X be an universal covering, and fl : C Ñ X be a covering of X with C being connected.
Then there exists a covering map f : D Ñ C such that fl ˝ f “ q.

The following proposition states that the universal covering is essentially unique.

Proposition 48.56.
Let qi : Di Ñ X (with i “ 1, 2) be two universal coverings of the topological space X. Then there
exists an homeomorphism f : D1 Ñ D2 such that q2 ˝ f “ q1 .

48.8.2 Monodromy action

Let fl : C Ñ X be a covering with C being connected and locally arc connected. First, that
shows that X has these two properties too. Now, let x P X and c P fl´1 pxq and a path “ : r0, 1s Ñ X
with “p0q “ “p1q “ x. By the lifting property, that path lifts to an unique path in C starting at
c, while it is not guarantee that the lifted path will end at x. One only knows that the lifted path
will end in fl´1 pxq.
It turns out that the end point of the lifted path only depends on the class of “ in fipX, xq.
Thus we define an action if fipX, xq on the fibre over x. This is the monodromy action. Notice
that by taking that action pointwise on X, the group fipXq acts on C.