STAY HOME, STAY SAFE and SAVE LIVES

Lecture Notes On
Nets and Filters
Dhananjoy Mandal
Department of Pure Mathematics
University of Calcutta

1 Motivation : Nets
Question: Give an example of a function f : (X, τ ) → (Y, σ) with the property that
for any sequence {xn } in X converging to a point x in X, the sequence {f (xn )} in Y
converging to f (x) in Y , but f is not continuous.
Example: Let τc be the cocountable topology on the real number set R and i : (R, τc ) →
(R, τu ) be the identity map, where τu is the usual topology on R. Let {xn } be a sequence
in (R, τc ) with lim xn = x. Then we know that ∃ m ∈ N such that xn = x for all n ≥ m.
Hence i(xn ) = x for all n ≥ m i.e., lim i(xn ) = x = i(x). But i is not continuous as
(0, 1) is open in (R, τu ) and i−1 (0, 1) = (0, 1) is not open in (R, τc ).
Remark: We cannot decide the continuity of a function via convergence of a sequence.
This is an inadequacy of a sequence for the study of topology.
Example: We know that in a T2 space, every convergent sequence has a unique limit.
But a space X in which every convergent sequence has a unique limit may not be T2 . For
example, if τc is the cocountable topology on the set R, then every convergent sequence
in (R, τc ) has a unique limit and (R, τc ) is not T2 .
Remark : We cannot decide the T2 -ness of a space via convergence of a sequence. This
is another inadequacy of a sequence for the study of topology.

To look for some type of convergence in a topological space which is free from the
inadequacy of sequence −− the notion of nets introduced and developed by E. H. Moore
and H. L. Smith in the year 1922.

1
2 Nets
Definition 2.1 A binary relation ≥ on a nonempty set D is said to be a directed set,
denoted by (D, ≥) if the following conditions are satisfied:
(i) ‘ ≥’ is reflexive and transitive,
(ii) given m, n ∈ D, ∃ p ∈ D such that p ≥ m and p ≥ n (positive orientation property).

Example 2.2 1. Every linearly ordered set is a directed set. In particular, R, Q, N

endowed with the natural orderings are directed sets.
2. For any topological space X, if x ∈ X, then the neighbourhood system ηx at x is a
directed set under the natural inclusion ⊆.

Definition 2.3 A net in a set X is simply a mapping x : D → X, where D is a directed

set. The element x(λ), for λ ∈ D, is usually denoted by xλ and we write {xλ : λ ∈ D}
for the net x.

Example 2.4 1. (N, ≥) is a directed set where ≥ is usual arithmetic order in N and
therefore any sequence x : N → X is a net in a set X; obviously a net may not be
a sequence.

2. Consider the neighbourhood system ηx of any point x in a topological space X.

Then (ηx , ≥) becomes a directed set if we write U ≥ V iff U ⊆ V , for all U, V ∈ ηx
(check). For each U ∈ ηx , we chose a point xu ∈ U . Then the map U 7−→ xu ,
U ∈ ηx defines a net in X.

3. Let f : X → Y be a mapping and x : D → X be a net in X. Then f ox : D → Y

is a net in Y .

4. Let x : D → X be a net in X and f : E → D be a function, where D, E are

directed sets. Then xof : E → X is a net in X.

5. Let f : [a, b] → R be a bounded function. Let D be the set of all partitions of [a, b].
Then (D, ≥) becomes a directed set if we write for any P, Q ∈ D, P ≥ Q iff P is
a refinement of Q. For any P ∈ D, let L(f, P ) = the Riemann lower sum of f for
P . Then P 7−→ L(P, f ) defines a net in R (check). Similarly P 7−→ U (P, f )(=the
Riemann upper sum of f for P ) defines a net in R(check).

Definition 2.5 A net {xλ : λ ∈ D} in a topological space X is said to be

(i) convergent to a point x ∈ X if for each neighbourhood U of x in X, there is p ∈ D
such that xλ ∈ U for all λ ≥ p and we say that the net {xλ : λ ∈ D} lies eventually in

2
U . It is denoted by lim xλ = x and also we call x is a limit of the net {xλ : λ ∈ D}.
(ii) cluster at a point x ∈ X if given any neighbourhood U of x in X and given any
λ ∈ D, there exists µ ∈ D such that µ ≥ λ and xµ ∈ U and we say that the net
{xλ : λ ∈ D} lies frequently in U .

Remark 2.6 1. If a net {xλ : λ ∈ D} converges to a point x in a space X then

{xλ : λ ∈ D} cluster at that point x in X; but the converse is false (check).
2. A net {xλ : λ ∈ D} in a set X cannot lie eventually in both U and V , where U , V
are two disjoint nonempty subsets of X (check).

Theorem 2.7 A topological space X is T2 iff every convergent net in X has a single
limit.
Proof. Let X be T2 and {xλ : λ ∈ D} be a convergent net in X. If possible, let
{xλ : λ ∈ D} has two distinct limits x and y in X. Since X is T2 , there exist two
disjoint open sets U and V such that x ∈ U and y ∈ V and the rest is left to the
students.
Conversely, suppose that X is not T2 . Then there exist two distinct points x and y of X
such that ∀ U ∈ ηx and ∀ V ∈ ηy , we have U ∩ V 6= φ.
Now ηx × ηy = {(U, V ) : U ∈ ηx , V ∈ ηy } is a directed set with respect to ‘ ≥’ where ‘ ≥’
is defined by (U1 , V1 ) ≥ (U2 , V2 ) iff U1 ⊆ U2 and V1 ⊆ V2 . For any (U, V ) ∈ ηx × ηy ,
U ∩V 6= φ, we chose a point xu,v ∈ U ∩V . Then P : ηx ×ηy → X defied by P (U, V ) = xu,v
is a net in X.
Claim: This net converges to x and y both. Verify this and proves became complete.

Theorem 2.8 A function f : X → Y is continuous at a point x ∈ X iff whenever a net

{xλ : λ ∈ D} in X converges to x ∈ X, the net {f (xλ ) : λ ∈ D} in Y converges to f (x)
in Y .
Proof. One part is easy and is left to the students.
To prove the converse, let f : X → Y be not continuous at x ∈ X. Then there exists
a neighbourhood W of f (x) in Y such that f (U ) 6⊆ W for all U ∈ ηx , i.e., for any
U ∈ ηx , there exists xu ∈ U such that f (xu ) 6∈ W . It is clear that {xu : U ∈ ηx } becomes
a net in X and converging to x (check). But f (xu ) 6∈ W for all U ∈ ηx , so the net
{f (xu ) : U ∈ ηx } in Y cannot converges to f (x) in Y .

Theorem 2.9 A point x ∈ A in a topological space X iff there exists a net {xλ : λ ∈ D}
lying in A such that lim xλ = x.
Proof. Suppose x ∈ A. Then for all U ∈ ηx , U ∩ A 6= φ. Chose a point xu ∈ U ∩ A.
Then {xu : U ∈ ηx } is a net in A. Let G be any open neighbourhood of x in X. Then

3
for any U ∈ ηx with U ≥ G, we get xu ∈ U ⊆ G. This shows that this net converges to
x in X.
Converse is trivial.

Corollary 2.10 A subset A of a space X is closed iff no net in A can converges to a

point in X \ A.
Proof. If A is closed, then no net in A can converge to any point of X \ A, because
X \ A is a neighbourhood of each of its points containing no terms of the net.
Conversely, if A is not closed then there is x ∈ A\A and so there is a net in A converging
to a point in X \ A.

Corollary 2.11 A subset B of a space X is open iff no net in X \ B can converge to a

point in B.

Definition 2.12 Let P = {xλ : λ ∈ D} be a net in X. Then the composite map

P ◦ ψ : E → X is said to be subnet of P if ψ : E → D satisfies the following conditions:
(i) for all µ1 , µ2 ∈ E, µ1 ≥ µ2 ⇒ ψ(µ1 ) ≥ ψ(µ2 ) (In this case ψ is called increasing)
(ii) given λ ∈ D, there exist µ ∈ E such that ψ(µ) ≥ λ(In this case ψ is called cofinal).
This subnet of P, we shall write as {xψ(µ) : µ ∈ E}.

Note 2.13 1. A subsequence of a sequence is a subnet of the sequence. But a subnet of

a sequence may not be its subsequence. For example, let P : N → X be a net in X and
E = [1, ∞). We define ψ : E → N by ψ(x) = [x], x ∈ E. Then P ◦ ψ : E → X is a
subnet of the net P : N → X which is not a subsequence. This illustrated the fact that a
subnet can have a richer index than than the original net.
2. If a net {xλ : λ ∈ D} is eventually in a set U , then every subnet of it will be eventually
in U . In fact, let {xλ : λ ∈ D} be eventually in a set U . Then there exists λ0 ∈ D such
that xλ ∈ U for all λ ≥ λ0 . Let {xψ(µ) : µ ∈ E} be any subnet of {xλ : λ ∈ D}. Now
λ0 ∈ D ⇒ there exists µ0 ∈ E such that ψ(µ0 ) ≥ λ0 . Then for all µ ≥ µ0 , we have
xψ(µ) ∈ U . This shows that the subnet {xψ(µ) : µ ∈ E} lies eventually in U .

Theorem 2.14 A net {xλ : λ ∈ D} in a space X has a cluster point x ∈ X iff there
exists a subnet of it converging to x in X.
Proof. Let P = {xλ : λ ∈ D} be cluster at x in X. Consider E = {(λ, U ) : U ∈
ηx , xλ ∈ U }. For any (λ, U ), (λ0 , U0 ) ∈ E, we write (λ, U ) ≥ (λ0 , U0 ) iff λ ≥ λ0 and
U ⊆ U0 . Then (E, ≥) becomes a directed set. Let ψ : E → D be defined by ψ(λ, U ) = λ,
(λ, U ) ∈ E. Then ψ is increasing and cofinal in D and so P ◦ ψ : E → X is a subnet
of P . We now show that P ◦ ψ is convergent to x in X. Let U0 ∈ ηx and chose λ0 ∈ D

4
with xλ0 ∈ U0 . So (λ0 , U0 ) ∈ E. Now for any (λ, U ) ∈ E with (λ, U ) ≥ (λ0 , U0 ) we have
xλ ∈ U ⊆ U0 i.e., (P ◦ ψ)(λ, U ) = xλ ∈ U0 for all (λ, U ) ≥ (λ0 , U0 ). This shows that
P ◦ ψ converges to x.
Conversely, let x be not a cluster point of the net {xλ : λ ∈ D}. Then there exists U ∈ ηx
and a λ ∈ D such that xµ 6∈ U for all µ ≥ λ i.e., xµ ∈ X \ U for all µ ≥ λ. This shows
that the net lies eventually in X \ U . So all subnets of this net must be eventually in
X \ U and hence no subnet of {xλ : λ ∈ D} can converge to x.

Theorem 2.15 A space X is compact iff every net in X has a cluster point in X.
Proof. Let X be compact. If possible let there exists a net {xλ : λ ∈ D} having no
cluster point in X. Then for each x ∈ X, there exists an open neighbourhood Ux of x
and λx ∈ D such that for all µ ≥ λx , xµ 6∈ Ux . Since X is compact and {Ux : x ∈ X}
is an open cover of X, there exist finite number of points x1 , x2 , ..., xn of X such that
X = ∪ni=1 Uxi . Since D is a directed set, there is λ ∈ D such that λ ≥ λx1 , λx2 , ..., λxn .
Thus xλ 6∈ ∪ni=1 Uxi = X, a contradiction.
Conversely, let every net in X has a cluster point. Let F be a family of closed sets having
finite intersection property. To show X is compact, it is sufficient to show that ∩F 6= φ.
Let B be the family of all finite intersections of members of F and it is sufficient to
show that ∩B 6= φ ( why ?). Now (B, ≥) becomes a directed set if we define B1 ≥ B2 iff
B1 ⊆ B2 , B1 , B2 ∈ B. For each B ∈ B, chose xB ∈ B. Then {xB : B ∈ B} is a net in
X and so by hypothesis, it has a cluster point say, y in X. We claim that y ∈ ∩B (prove
it ).

Corollary 2.16 A space X is compact iff every net in X has a convergent subnet.

Definition 2.17 A net {xλ : λ ∈ D} in a set X is said to be an ultra net or a universal

net if for each subset A of X, the net is either eventually in A or eventually in X \ A.

Example 2.18 For any directed set D and any set X, the map P : D → X defined by
P (λ) = x, for all λ ∈ D ( where x is a fixed point in X ) is an ultra net in X, called
trivial ultra net.

3 Motivation: Filter
Let {xλ : λ ∈ D} be a net in a space X and x ∈ X. For any λ ∈ D, let Tλ = {xµ : µ ≥ λ}
and Tλ is called a tail of the net. Now {xλ : λ ∈ D} converges to x in X iff given any
U ∈ ηx , there exists a tail Tλ0 such that Tλ0 ⊆ U . Again {xλ : λ ∈ D} cluster at x in X
iff given any U ∈ ηx and any tail Tλ0 , U ∩ Tλ0 6= φ.

5
Remark 3.1 A net {xλ : λ ∈ D} in a space X converges to x in X iff every neighbour-
hood of x contains a tail of the net and a net cluster at x iff every neighbourhood of x
inteersects every tail of the net.

Observation 3.2 Let F = {F ⊆ X : there exists λ ∈ D with Tλ ⊆ F }. Then

(i) φ 6∈ F,
(ii) F1 , F2 ∈ F ⇒ F1 ∩ F2 ∈ F (check),
(iii) F ∈ F and F ⊆ G ⊆ X ⇒ G ∈ F.

This leads to a definition as follows:

Definition 3.3 Let X be a nonempty set. A nonempty collection F of subsets of X is

called a filter on X if the following conditions hold:
(i) φ 6∈ F,
(ii) F1 , F2 ∈ F ⇒ F1 ∩ F2 ∈ F and
(iii) F ∈ F and F ⊆ G ⊆ X ⇒ G ∈ F.

Remark 3.4 1. The family {F ⊆ X : there exists λ0 ∈ D such that Tλ0 ⊆ F } of all
supersets of the tail of the net {xλ : λ ∈ D} in X is a filter on X which is called the
filter generated by the net.
2. We see that every filter has the f.i.p. The converse is not true in general, because a
family having f.i.p need not be closed under finite intersection ( justify ).

Definition 3.5 Let F be a filter on X. Then a subfamily F0 of F is called a filterbase

for F if for any F ∈ F, there exists F0 ∈ F0 such that F0 ⊆ F .

Example 3.6 The family {Tλ : λ ∈ D} of all tails of the net {xλ : λ ∈ D} in X is a
filterbase for the filter F = {F ⊆ X : there exists λ0 ∈ D with Tλ0 ⊆ F } on X.

Theorem 3.7 A nonempty collection B of nonempty subsets of X is a filterbase for

some filter on X iff B1 , B2 ∈ B ⇒ there exists B3 ∈ B such that B3 ⊆ B1 ∩ B2 .

Proof. Let B be a filterbase for some filter F on X. Let B1 , B2 ∈ B. Then B1 ,

B2 ∈ F and so B1 ∩ B2 ∈ F. Since B is a filterbase for F, there exists B3 ∈ B such that
B3 ⊆ B1 ∩ B2 .
Conversely, let the condition holds. Consider F = {F ⊆ X : B ⊆ F for some B ∈ B}.
Then it can be easily checked that F is a filter on X and B is a filterbase for the filter
F on X.

6
Theorem 3.8 If B ⊆ P(X) has f.i.p, then B can be enlarge to a filter on X and
moreover, it is the smallest filter containing B on X.

Proof. Consider F(B) = {F ⊆ X : ∩m

i=1 Bi ⊆ F for some finite subfamily {B1 , B2 , ..., Bm }
of B}. Then F(B) is a filter on X and is the smallest filter containing B ( verify ) on X.

Example 3.9 1. For any set X, X is a member of each filter on it and {X} is a
filter on X and in fact it is the smallest filter on X.

2. For any A ⊆ X with A 6= φ, FA = {F ⊆ X : A ⊆ F } is a filter on X (check). In

particular, for any x ∈ X, Fx = {F ⊆ X : x ∈ F } is a filter on X.

3. In any topological space X, the neighbourhood system ηx at x is a filter on X

(check).

4. Let X be an infinite set. Then CF (X) = {F ⊆ X : X \ F = at most a finite set }

is a filter on X (check).

5. The collection B = {(a, ∞) : a ∈ R} is not a filter on R ( why ?), but it is a

filterbase for some filter on R.

Definition 3.10 A filter U on X is said to be an ultra filter on X if there does not exist
any filter F on X such that U $ F.

Example 3.11 For any x ∈ X, Ux = {F ⊆ X : x ∈ F } is an ultra filter on X. In fact,

if B is a subfamily of P(X) with Ux $ B, then there exists B ∈ B such that B 6∈ Ux so
that x 6∈ B. Again, {x} ∈ Ux ⊆ B ⇒ {x} ∈ B. Since {x} ∩ B = φ, B is not a filter on
X. Thus Ux is an ultra filter on X.

Theorem 3.12 Every filter F on X can be extended to an ultra filter on X.

Proof. Let Ω = {G ⊆ P(X) : G is a filter on X and F ⊆ G}. Then Ω 6= φ as F ∈ Ω.
Now Ω becomes a partially ordered set with respect to set inclusion. Let {Fα : α ∈ Λ}
be any chain in Ω. Then ∪α∈Λ Fα is an upper bound of this chain in Ω and hence by
Zorn’s Lemma, there exists a maximal element U in Ω. Clearly U is an ultra filter on X
containing F.

Corollary 3.13 If B ⊆ P(X) has f.i.p, then B can be extended to an ultra filter on X.

We have seen that every filter on a set X can be extended to an ultra filter on X. Is this
extension unique? The answer is no and find a suitable example of this claim.

7
Theorem 3.14 Let U be a filter on X. Then TFAE:
(a) U is an ultra filter on X.
(b) A ⊆ X and A ∩ F 6= φ, ∀ F ∈ U ⇒ A ∈ U.
(c) A, B ⊆ X and A ∪ B ∈ U ⇒ either A ∈ U or B ∈ U.
(d) for any A ⊆ X, either A ∈ U or X \ A ∈ U.

Proof. (a) ⇒ (b) : Let A ⊆ X and A ∩ F 6= φ ∀ F ∈ U. Then B = {A ∩ F : F ∈ U} is

a filterbase for some filter G on X with U ⊆ G (check). Since U is an ultra filter, U = G
and hence A ∈ U.
(b) ⇒ (c) : Let A ∪ B ∈ U. If A, B 6∈ U, then by (b),there exist P, Q ∈ U such that
A ∩ P = B ∩ Q = φ ⇒ (A ∪ B) ∩ (P ∩ Q) = φ, a contradiction ( why?). Thus either
A ∈ U or B ∈ U.
(c) ⇒ (d) : For any A ⊆ X, A ∪ (X \ A) = X ∈ U and rest follows.
(d) ⇒ (a) : If possible, let U be not an ultra filter on X. Then there exists a filter F on
such that U $ F. So there exists A ⊆ X such that A ∈ F, but A 6∈ U. Then by (d),
X \ A ∈ U and so X \ A ∈ F and hence A ∩ (X \ A) = φ ∈ F, a contradiction.

Definition 3.15 A filter F on a set X is called

(i) fixed if ∩F = ∩{F ⊆ X : F ∈ F} =
6 φ.
(ii) free if ∩F = φ.
Thus to show a filter F on X is fixed, we have to show that ∃ x ∈ X such that x ∈ F ,
∀ F ∈ F and to show a filter F on X is free, we have to show that for each x ∈ X, there
exists F ∈ F such that x 6∈ F .

Example 3.16 In the Example 3.9, the filters (1), (2) and (3) are fixed while the filters
(4) and (5) are free.

Observation 3.17 We have seen in Example 3.11 that for any x ∈ X, Ux = {F ⊆ X :

x ∈ F } is an ultra filter on X which is fixed.
Now let U be any fixed ultra filter on X. Then there exists x ∈ X such that x ∈ ∩{F :
F ∈ U}. Thus for all F ∈ U, x ∈ F implies F ∈ Ux and hence U ⊆ Ux . Again by
maximality of U, Ux = U.

By above observation we have the following result:

Theorem 3.18 A filter U on a set X is a fixed ultra filter on X iff U = Ux = {F ⊆

X : x ∈ F } for some x ∈ X.

8
Observation 3.19 Suppose U is an ultra filter on X and A1 ∪ A2 ∪ ... ∪ An ∈ U. Then
there exists at least one Ai ∈ U. Thus if a finite set A = {x1 , x2 , ..., xn } is a member of
an ultra filter U, then there exists i ∈ {1, 2, ..., n} such that {xi } ∈ U and hence U = Uxi .
So that U is a fixed ultra filter on X.

The above observation leads the following result:

Theorem 3.20 Every ultra filter on a finite set is fixed.

Corollary 3.21 Every filter on a finite set is fixed.

Proof. Left to the students.

Question: Is the converse of the above result true ? i.e., Let every filter on a set X
is fixed. Is X finite?
The answer is yes.
To answer this question it is sufficient to show that there exists a free filter on an
infinite set X. For example, let X be any infinite set and CF (X) = {F ⊆ X : X \ F =at
most a finite }. Then CF (X) is a free filter on X.

Theorem 3.22 For a set X, the TFAE:

(a) X is finite.
(b) every filter on X is fixed.
(c) every ultra filter on X is fixed.

Question: What about the free ultra filter on a set X ?

Observation 3.23 Every member of a free ultra filter U on a set X is at least an infinite
subset of X.

Moreover:

Theorem 3.24 Let F be a free filter on a set X. Then each member of F is at least
an infinite subset of X.
Proof. Left to the students.

Definition 3.25 A filter F on a space X is said to be

(i) convergent to x ∈ X if every neighbourhood of x in X belongs to F i.e., if ηx ⊆ F.
(ii) cluster at x ∈ X if every neighbourhood of x intersects every member of F i.e., if
for all U ∈ ηx and for all F ∈ F, U ∩ F 6= φ i.e., if x ∈ F , for each F ∈ F.

9
Definition 3.26 A filterbase F0 on a space X is said to be
(i) convergent to x ∈ X if for each neighbourhood U of x in X, F0 ⊆ U for some U ∈ F0 .
(ii) cluster at x ∈ X if every neighbourhood of x intersects every member of F0 i.e., if
for all U ∈ ηx and for all F0 ∈ F0 , U ∩ F0 6= φ i.e., if x ∈ F 0 , for each F0 ∈ F0 .

Remark 3.27 If a filter F on a space X converges to x, then F cluster at x. Is the

converse true? Justify.

Example 3.28 1. Let X be a space and ηx = the neighbourhood system at x. Then ηx

is a filter on X which converges to x.
2. Let X be a space, A ⊆ X and FA = {F ⊆ X : A ⊆ F }. Then every point of A is a
cluster point of the filter FA (check).
3. Consider co-finite topology on an infinite set X and CF (X) = {F ⊆ X : X \ F = at
most finite }. Then the filter CF (X) converges every point of X (check).
4. The filterbase B = {(a, ∞) : a ∈ R} on the real number space R has no cluster point
on R (why?).

Theorem 3.29 Let F and G be filters on a space X with F ⊆ G. If F converges to

some point of X then so does G.
Proof. Obvious.

Theorem 3.30 A filter F on a space X cluster at x ∈ X iff there exists a filter G ⊇ F

such that G converges to x in X.
Proof. Let F be cluster at x ∈ X. Then for all U ∈ ηx and for all F ∈ F, U ∩ F 6= φ.
Let B = {U ∩ F : U ∈ ηx and F ∈ U}. Then B is a filter base for some filter G on X
with F ⊆ G. Also G converges to x (check).
Converse is trivial.

Theorem 3.31 An ultra filter U on a space X converges to x ∈ X iff it is cluster at

x ∈ X.
Proof. Exercise.

Theorem 3.32 Let A be a subset in a space X. Then x ∈ A iff there exists a filter F
on X such that A ∈ F and F converges to x.
Proof. Let x ∈ A. Then U ∩ A 6= φ, for all U ∈ ηx . Let B = {U ∩ A : U ∈ ηx } is a filter
base for some filter F on X with B ⊆ F. Then A ∈ F and F converges to x.(check)
The rest is left to the students.

10
Theorem 3.33 A space X is T2 iff every convergent filter on X has a single limit.
Proof. Let X be T2 and F be any convergent filter on X. If possible, let F converges
to two distinct points x and y of X. Then ηx ⊆ F and ηy ⊆ F. Thus for all U ∈ ηx and
for all V ∈ ηy , we have U ∩ V 6= φ which contradicts that X is T2 .
Conversely, let X be not T2 . Then there exist two distinct points x and y of X such that
for all U ∈ ηx and for all V ∈ ηy , U ∩ V 6= φ. Let B = {U ∩ V : U ∈ ηx , V ∈ ηy }. Then
B is a filterbase for some filter F on X converges to x and y both (why?).

Theorem 3.34 A space X is compact iff every filter on X has a cluster point in X.
Proof. Let X be compact and F be any filter on X. Then F has f.i.p and hence
T
F = {F : F ∈ F} has f.i.p. Since X is compact, {F : F ∈ F} 6= φ. Thus any point
belonging to this nonempty set is a cluster point of F.
Conversely, let every filter on X has a cluster point in X. Let G = {Gα : α ∈ Λ} be
any family of closed sets having f.i.p. To show X is compact, it is sufficient to show
T
that α∈Λ Gα 6= φ. Since G has f.i.p, G can be extended to a filter F on X. Then by
hypothesis, F has cluster point at x in X and hence x ∈ F , for all F ∈ F. In particular,
T
x ∈ Gα = Gα for all Gα ∈ G and thus α∈Λ Gα 6= φ.

Theorem 3.35 A space X is compact iff every ultra filter on X converges to a point of
X.
Proof. Exercise.

Definition 3.36 Let f : X → Y be a function and F be a filter on X. Then f (F) =

{f (F ) : F ∈ F} may not be filter on Y ( why ?). But f (F) is a filterbase for some filter
on Y . The filter generated by f (F) is called the image filter of F under f , denoted by
f # (F) and thus f # (F) = {E ⊆ Y : f (F ) ⊆ E for some F ∈ F}.

Theorem 3.37 Let X and Y be two spaces. A function f : X → Y is continuous at

x ∈ X iff whenever a filter F on X converges to x ∈ X, the filterbase f (F) in Y con-
verges to f (x) in Y .
Proof. Let f be continuous at x and a filter F on X converges to x. Then for any
U ∈ ηx , U ∈ F. Let W be any neighbourhood of f (x) in Y . Then there exists a neigh-
bourhood V of x such that f (V ) ⊆ W i.e., f (V ) ⊆ W where V ∈ F. This shows that
the filterbase f (F) converges to f (x) in Y .
Conversely, let the condition holds. To show f is continuous at x, let W be any neigh-
bourhood of f (x) in Y . Consider the neighbourhood system ηx at x in X. Then ηx is a
filter on X converging to x. Then by hypothesis, the filterbase f (ηx ) converges to f (x)
in Y . Since W is a neighbourhood of f (x), there exists a U ∈ ηx such that f (U ) ⊆ W .
This shows that f is continuous at x.

11
4 Relation Between Nets and Filters
In this section we shall study the formal bridge between filters and nets.

Definition 4.1 Let {xλ : λ ∈ D} be a net in a set X. For each λ ∈ D, let Tλ =

{xµ : µ ≥ λ}. Then B = {Tλ : λ ∈ D} is a filterbase on X and thus the family
F = {F ⊆ X : Tλ0 ⊆ F for some Tλ0 } is a filter on X. This filter F on X is called the
filter generated by the net {xλ : λ ∈ D} on X.

Definition 4.2 Let F be a filter on a set X. Let D = {(x, F ) : x ∈ F ∈ F}. Then

(D, ≥) becomes a directed set if we write “(x1 , F1 ) ≥ (x2 , F2 ) iff F1 ⊆ F2 ”, for all
(x1 , F1 ), (x2 , F2 ) ∈ D. Then a map P : D → X defined by P (x, F ) = x is a net in X,
called the net based on the filter F or the net generated by the filter F on X.

Theorem 4.3 Let P = {xλ : λ ∈ D} be a net and F be a filter generated by the net P
on a space X. Then
(i) P converges to x in X iff the filter F converges to x in X.
(ii) P cluster at x in X iff the filter F cluster at x in X.
Proof. Let F be a filter generated by the net P = {xλ : λ ∈ D} on X. Then F = {F ⊆
X : Tλ ⊆ F for some Tλ } where Tλ = {xµ : µ ≥ λ}.
(i) Suppose P converges to x in X. Then for all U ∈ ηx , there exists λ0 ∈ D such that
xλ ∈ U for all λ ≥ λ0 . Thus Tλ0 ⊆ U which implies U ∈ F and hence F converges to x
in X.
Conversely, suppose F converges to x in X. Then for all U ∈ ηx , U ∈ F. Thus Tλ0 ⊆ U
for some λ0 ∈ D i.e., xλ ∈ U for all λ ≥ λ0 . This shows that P converges to x in X.
(ii) Let P be cluster at x in X. Then for all U ∈ ηx and for all λ ∈ D, there exists
µ ∈ D such that µ ≥ λ and xµ ∈ U . Thus for all λ ∈ D, Tλ ∩ U 6= φ and hence for all
F ∈ F, F ∩ U 6= φ. This shows that the filter F cluster at x.
Conversely, let F be cluster at x in X. Then for all U ∈ ηx and for all F ∈ F, U ∩F 6= φ.
Thus for all λ ∈ D, Tλ ∩ U 6= φ i.e., for all U ∈ ηx and for all λ ∈ D, there exists µ ∈ D
such that µ ≥ and xµ ∈ U . Hence the net P cluster at x in X.

Theorem 4.4 Let F be a filter and P be the net based on the filter F on a space X.
Then
(i) F converges to x in X iff the net P converges to x in X.
(ii) F cluster at x in X iff the net P cluster at x in X.
Proof. Exercise.

12
 Exercises

1. Let X be a topological space with a subbase S and x ∈ X. Prove that a net in X

converges to x in X iff the condition in the definition of convergence of nets holds
for all neighbouhood of x which are members of S.

2. Let T be a subnet of a net P in a space X. Prove that the limit points of P are
also so of T , and that the cluster points of T are also so of P .

3. Prove that a space X is an indiscrete space iff every net convergent to every point
of it.

4. Obtain a characterization of a discrete space in terms of convergence of nets in

it. Can a discrete space be characterize in terms of convergence of sequence in it?
Justify your answer.

5. Let τ1 , τ2 be two topologies on a set X. Prove that τ2 ⊆ τ1 iff whenever a net

converges to a point with respect to τ1 , it so does with respect to τ2 also.

6. Prove or disprove: The set of all cluster points of a net in a space X is closed.

7. Prove that the intersection of any family of filters on a set is again a filter on that
set.

8. Let X be a space and x ∈ X. Prove that the intersection of all filters on X

converges to x ∈ X is precisely the neighbourhood system at x in X.

9. Prove that a space X is Hausdorff iff every ultrafilter on X converges to at most

one point of X.

10. Prove that a filter F on a space X extended to a unique ultra filter on X iff F is
an ultra filter.

11. Let X be an infinite set and CF (X) = {F ⊆ X : X \ F = at most finite}. Prove

that the filter CF (X) can be extended to infinitely many ultra filter on X.

12. Let f : X → Y be a function and U an ultra filter on X. Prove that f # (U) =

{E ⊆ X : f (F ) ⊆ E for some F ∈ U} is an ultra filter on Y .

13
References
[1] R. Engelking : General Topology, Polish Scientific Publishers.

[2] K. D. Joshi : Introduction to General Topology, New Age International Publishers.

[3] J. R. Munkres : Topology− A First Course, Prentice-Hall, New Delhi.

[4] L. Steen and J. Seebach : Counterexamples in Topology, Holt, Rinehart and Winston,
New York, 1970.

[5] S. Wilard : General Topology, Addison-Wesley, Reading, 1970.

14