Convergence in Topological Spaces.

Nets.
Stefano Ferri

Abstract
Sequences are not always sufficient to describe convergence in topological spaces.
In what follows we shall introduce a generalisation to sequence which can satisfactorly
describe convergence in any topological space.

1 Preliminaries.
It is well-known that many topological properties of metric spaces can be described using
sequences (see [Ru, Chapters 3-4]). For istance:

Theorem 1.1. 1. A subset A of a metric space X is open if and only if all sequences
hxn i in X which converge to a point x ∈ A are eventually in A.

2. A subset A of a metric space X is closed if and only if, for all sequences hxn i in A, if
lim xn = x, then x ∈ A.
n→∞

3. If A is a subset of a metric space, then x ∈ A if and only if there exists a sequence

hxn i in A which converges to x.

4. A function f : X −→ Y between metric spaces is continuous if and only if, whenever

hxn i is a sequence which converges to x in X, then hf (xn )i converges to f (x) in Y .

As we shall see, in general topological spaces the above statements are no longer true
unless certain conditions are satisfied. We start by extending the definition of convergence
(of sequences) to general topological spaces.

Definition 1.2. A sequence in a topological space X is a function f : N −→ X. It is

common practice to denote a sequence by hxn i meaning that, for n ∈ N, f (n) = xn .

Definition 1.3. A sequence hxn i in a topological space X converges to L ∈ X if, for all
(basic) neighbourhood U of L, there exists nU ∈ N such that, whenever m > nU , we have
xm ∈ U .

Remebering proof of Theorem 1.1 it should be clear that when we can find countably
many “arbitrarly small” neighbourhoods around all points of X, we can still replicate this
proof it a topological space. In order to do so we start with a definition which formalises
the concept or “arbitrarly small” neighbourhood to topological spaces.

Definition 1.4. A topological space X is first countable if every point has a countable
neighbourhood base.

1
2 Nets

For first countable spaces the four statements of Theorem 1.1 are true. The proofs are
very similar to those given for metric spaces. Here we prove only one of the propositions
(the other three are left as exercises) in order to illustrate which kind of changes are needed.

Theorem 1.5. Let X be a first countable space and A ⊆ X. Then, x ∈ A if and only if
there exists a sequence hxn i in A converging to x.

Proof. Let x ∈ A and let {Un : n ∈ N} be a countable base of neighbourhoods of x.

\i
Let Vi := Uk . It is not difficult to see that {Vn : n ∈ N} is also a countable base of
k=1
neighbourhoods of x and that:
U1 ⊃ U2 ⊃ U3 ⊃ . . . .
Since x ∈ A, for all i ∈ ω, Vi ∩ A 6= ∅. So, for all i ∈ N, we can pick xi ∈ Vi ∩ A. The
sequence hxn i obtained this way clearly converges to x.
Conversely, if hxn i converges to x, then, every neighbourhood U of x contains infinitely
many points of this sequence. Hence U it has not empty intersection with A, i.e. x ∈ A.

The previous proof relies on the fact that all points of X have a countable neighbourhood
base. If this is not the case it may happen1 that sequences are no longer sufficient to describe
the topology of the space. The following examples (which is left as an exercise) illustrate
this.

Example 1.6. Let X = RR with the product topology. Let us define a subset E of RR as
follows:

E = {f ∈ RR : f (x) = 0 or f (x) = 1 and f (x) = 0 only for finitely many x ∈ R}.

Let g be the function in RR which is identically 0 and U (g) a basic neighbourhood of g,

U (g) = {h ∈ RR : |h(y) − g(y)| < ε if y ∈ F },

where F is some finite subset of R.

The set U (g) ∩ E is not empty, since it contains the function which is 0 on E and 1
elsewhere, then g ∈ E.
However, no sequence in E converges to g.

2 Nets
In this section we shall introduce the concept of net and show that this is a valid sobstitute
to sequences to describe convergence in topological spaces.

Definition 2.1. Let I be a set partially ordered by ≤. We say that I is directed by ≤ if,
for every, α, β ∈ I, there exists γ ∈ I such that α ≤ γ and β ≤ γ.

Definition 2.2. A net in a topological space X is a function f from a directed set I to X.

Usually, a net is denoted by hxα iα∈I meaning that f (α) = xα .
1
One may think that being first countable is a necessary and sufficient condition in order for sequences
to be enough to describe the topology. However, this is not true. Many mathematicians have studied
the problem of characterising the topological spaces for which sequences sufficies in order to describe their
topology and obtained a somhow wider class than that of first countable spaces.
 Stefano Ferri 3

Definition 2.3. A net hxα iα∈I in a topological space X, is said to be frequently in a subset
A of X if, for all α0 ∈ I, there exists some α ≥ α0 such that xα ∈ A. The same net is said
to be eventually in A if there exists some α0 ∈ I such that, for all α ≥ α0 , xα ∈ A.
Definition 2.4. A net hxα iα∈I , in a topological space X is said to have a point x of X as
a cluster point if the net is frequently in every neighbourhood of x. If the net is eventually
in every neighbourhood of x, then it is said to converge to x and x is said to be a limit of
the net. In this case we write x ∈ lim xα . Sometimes, if there is no danger of confusion we
α∈I
simply write x ∈ lim xα or also xα −→ x. When the limit is unique we write x = lim xα
α
instead of {x} = lim xα .
Nets are a usefull tool to describe convergence in metric spaces, however to formulate
the concept of subnet is rather cumbersome.
Definition 2.5. Let I and J be directed sets. A function f : J −→ I is said to be increasing
and cofinal if:
1. For all α, β ∈ J such that α ≤ β, f (α) ≤ f (β), and;
2. For all δ ∈ I, there exists α ∈ J such that f (α) ≥ δ.
Definition 2.6. Let hxα i be a net in X indexed by I and let f : I −→ X be the function
such that f (α) = xα . If g : J −→ I is an increasing cofinal function between the directed
sets J and I, then the composition f ◦ g : J −→ X is called a subnet of hxα i.
Theorem 2.7. Let hxα i be a net in a topological space X, and hxαβ i a subnet of hxα i (by
this notation we mean that the increasing cofinal function which defines the subnet maps β
into αβ ). Then the following statements hold:
1. If x is a cluster point of hxαβ i, then it is also a cluster point of hxα i;
2. If x is a limit of hxα i then it is a limit of hxαβ i;
3. If x is a cluster point of hxα i then there exists a subnet hxαγ i of hxα i, which admits x
as a limit.
Proof. 1. Let U be a neighbourhood of x and α0 ∈ I. Since hxαβ i is a subnet of hxα i
there is β0 ∈ J such that αβ0 ≥ α0 . Since x is a cluster point of hxαβ i, there exists
β ′ ∈ J such that αβ ′ ≥ αβ0 and xαβ ′ ∈ U . This means that x is a cluster point of hxα i.
2. If x is a limit of hxα i, then, for all neighbourhood U of x, hxα i is eventually in U . In
particular, hxαβ i is eventually in U .
3. Let x′ be a cluster point of hxα i. Consider
K = {(β, U ) : β ∈ I and U is a neighbourhood of x, xβ ∈ U }.
We define a partial order on K as follows:
(β1 , U1 ) ≤ (β2 , U2 ) if β1 ≤ β2 and U1 ⊇ U2 .
It is not difficult to check that K is directed by ≤. We define an increasing cofinal
map between K and I by (β, U ) 7−→ β and a subnet of hxα i by xα(α,U ) = xα . For
all neighbourhood U of x, there exists some β0 in I such that xβ0 ∈ U , so for all
(β0 , U0 ) ≤ (β, U ), xβ(β,U ) ∈ U ⊂ U0 , i.e. x ∈ lim xα(α,U ) , as requested.
4 Nets

Theorem 2.8. Let A be a subset of a topological space X. A point x of X belongs to A,

if and only if there exists a net in A converging to x.

Proof. Let S = hxα iα∈I be a net in A converging to x. Then any neighbourhood of x

contains points of A, so x is in A. Conversely, let x ∈ A and let Nx be the set of the
neighbourhoods of x. The set Nx is directed by the order ≤ given by:

U ≤ V if U ⊇ V.

We construct a net of points of A by choosing for every U ∈ Nx a point xU ∈ U ∩ A. It is

easy to check that the net hxU iU ∈Nx , constructed in this way, has x as a limit.

The proofs of the following theorems are left as exercises.

Theorem 2.9. Let X and Y be topological spaces. A map f : X −→ Y is continuous if

and only if, for all net hxα iα∈I in X:

f (lim xα ) ⊆ lim f (xα ).

α∈I α∈I

Theorem 2.10. A topological space X is Hausdorff if and only if every net in X has at
most one limit.

Nets correspond, for topological spaces, to sequences and, of course, sequences are a
particular example of nets in which the directed index set is N. However, when we pass
from the definition of subsequence to the definition of subnet we have some differences. For
istance, given a sequence there may be subnets which are not subsequences. For this reason,
in spite of the fact that by using nets it is possible to extend the concept of convergence to
any topological spaces, when dealing with nets we have to be be very careful.

Example 2.11. Consider the sequence h(−1)n in∈N in the space R with its standard topol-
ogy. Clearly, this sequence has 1 and −1 as cluster points. We use the construction given
in theorem 2.7 to construct a finer net converging to 1. We take as index set:

K = {(n, U ) : n ∈ N and U is a neighbourhood of 1 andxn ∈ U } =

= {(n, U ) : n even and U is a neighbourhood of 1},

directed by the order:
(n, U ) ≤ (m, V ) if n ≤ m and U ⊇ V,
and construct a net by defining, for (n, U ) in K, x(n,U ) = (−1)n = 1. This is a subnet of
the net we began with and it converges to 1, but, since |Σ| > ω, this is not a subsequence.
Of course h(−1)2m im∈N is a subsequence converging to 1.

We conclude this short note with a theorem concerning convergence in product spaces.
(This theorem will allow for a fast proof of Tychonoff Theorem.)
Y
Theorem 2.12. A net hxα i in a product Xγ converges to x in the product topology if
γ∈I
and only if, for all α ∈ I, the net hπα (xγ )i converges to πα (x) in Xα .
 Stefano Ferri 5

Y
Proof. If xγ −→ x in Xγ , then, since πα is continuous, πα (xγ ) −→ πα (x). In order to
γ∈I
show the converse, suppose that, for all α ∈ I, πα (xγ ) −→ πα (x) and let

πα−1
1
(Uα1 ) ∩ πα−1
2
(Uα2 ) ∩ · · · ∩ πα−1
n
(Uαn )

be a basic neighbourhood of x. Then, for all i ∈ {1, 2, . . . , n}, there exists γi such that,
when γ ≥ γi , then παi (xγ ) ∈ Uαi . Hence, if γ0 is greater than all the γi ’s (such a γ exists
because the index set of the net is directed) we haveTthat, whenever γ ≥ γ0 , for all α ∈ I,
παi (xγ ) ∈ Uαi . This means that, when γ ≥ γ0 , xγ ∈ πα−1 i
(Uαi ), i.e. xγ −→ x.

References
[Ru] Rudin W., “Principles of Mathematical Analysis” (third edition), Mc Graw – Hill,
Inc., New York, 1975.