A Short Study of Alexandroff Spaces

arXiv:0708.2136v1 [math.GN] 16 Aug 2007

by
Timothy Speer
Department of Mathematics
New York University
[email protected]
July 2007

Abstract
In this paper, we discuss the basic properties of Alexandroff spaces. Several
examples of Alexandroff spaces are given. We show how to construct new
Alexandroff spaces from given ones. Finally, two invariants for compact
Alexandroff spaces are defined and calculated for the given examples.

c Timothy Speer
All Rights Reserved, 2007
Contents
1 Introduction 1

2 Minimal Open Neighborhoods 2

3 New Alexandroff Spaces From Old Spaces 4

4 Continuous Maps 7

5 Compact Alexandroff Spaces 8

1 Introduction
The general definition of a topology is based off the properties of the standard
Euclidean topology. The goal of this paper is to study spaces that have
topologies which satisfy a stronger condition. Namely, arbitrary intersections
of open sets are open. With this restriction, we lose important spaces such
as Euclidean spaces, but the specialized spaces in turn display interesting
properties that are not necessary for a standard topological space.
Definition 1. Let X be a topological space, then X is an Alexandroff space
if arbitrary intersections of open sets are open.
Our first theorem shows that the condition of being an Alexandroff space
severely limits the type of spaces we are considering. So we must look to more
exotic examples of topologies to find useful spaces that are Alexandroff. A
theorem proved later will allow us to construct Alexandroff spaces easily.
Lemma 1. Any discrete topological space is an Alexandroff space.
Proof. This is clear because in a discrete space any subset is open.

Theorem 1. Let X be a metric space, then X is an Alexandroff space iff X

has the discrete topology.
Proof. ⇒) Suppose X is an Alexandroff space. Let x be a point in X. Then
the open balls B(x, n1 ) with radius n1 and centerTx, n a natural number, are
∞ 1
open in X. Since X is an Alexandroff T∞space, 1 n=1 B(x, n ) is an open set.
But by the properties of the metric, n=1 B(x, n ) = {x}. So we have shown
that singletons are open. Hence, X has the discrete topology.

⇐) The reverse direction follows from Lemma 1.

The original definition given for an Alexandroff space is easy to state,

however it is not too useful for proving theorems about Alexandroff spaces.
To fix this we will use a different, yet equivalent definition. Namely, in an
Alexandroff space, each point must have a minimal open set containing it.
Theorem 2. X is an Alexandroff space iff each point in X has a minimal
open neighborhood.

1
Proof. ⇒) Suppose X is an Alexandroff space with xT∈ X. Let O(x) = {U ⊂
X : U is an open neighborhood of x}. Take S(x)= U for U ∈ O(x), then
S(x) is an open neighborhood of x because X is Alexandroff. And from the
definition of S(x), it is clear that S(x) is a minimal open neighborhood of x.

⇐) Suppose each x ∈ X has a minimal open T neighborhood S(x). Consider

an arbitrary intersection of open sets, V = α∈A Uα , where each Uα is open
in X. If V = ∅, then V is open and we are done. If V 6= ∅, then pick x ∈ V
and we have x ∈ Uα for all α ∈ A. Hence, S(x) ⊂ Uα for all α because S(x)
is the minimal open neighborhood of x. Therefore, S(x) ⊂ V . Hence, V is
open because it contains an open set around each of it’s points.

From here on, we will write S(x) to denote the minimal open neighbor-
hood of a point x in an Alexandroff space.

2 Minimal Open Neighborhoods

In order to understand the properties of Alexandroff spaces, we will study
the minimal open neighborhoods of these spaces. The proofs of the theorems
show us that the minimal open neighborhoods of an Alexandroff space are
the natural objects to study.

Theorem 3. If β is a collection of subsets of X such that for each x ∈ X

there is a minimal set m(x) ∈ β containing x, then β is a basis for a topology
on X and X is an Alexandroff space with this topology. In addition, S(x) =
m(x).

Proof. Clearly the sets in β cover X. Suppose U, V ∈ β and that x ∈ U ∩ V .

Then m(x) is a minimal set in β containing x so we must have m(x) ⊂ U
and m(x) ⊂ V . Hence, m(x) ⊂ U ∩ V . So β is a basis for a topology on X.
S take any x ∈ X and
To show that X is an Alexandroff space with this basis,
let U be any open set in X containing x. Then U = α∈A Vα where Vα ∈ β.
But x must be in Vα for at least one α, which means m(x) ⊂ Vα ⊂ U. Hence,
m(x) is a minimal open set containing x. Therefore, X is Alexandroff and
S(x) = m(x).

2
 Theorem 3 is a powerful tool that will allow us to construct Alexandroff
spaces by specifying a basis.

Theorem 4. If X is an Alexandroff space with topology T , then β = {S(x) :

x ∈ X} is a basis for T .

Proof. By Theorem 3, β is a basis for a topology T ′ on X.SWe must show

T ′ = T . T ′ ⊂ T because β ⊂ T . Suppose U ∈ T . Let U ′ = x∈U S(x), then
U ′ ∈ T ′ and we also have U = U ′ . Hence, T ⊂ T ′ so we have T = T ′ .

Corollary 1. If T , T ′ are two topologies on X such that X is an Alexandroff

space and ST (x) = ST ′ (x) for all x in X, then T = T ′ .

Proof. From Theorem 4, we have β = {ST (x) : x ∈ X} and β ′ = {ST ′ (x) :

x ∈ X} are bases for T and T ′ respectively. But β = β ′ so we have T = T ′ .

We are now ready to construct some examples of Alexandroff spaces.

Example 1. (An Alexandroff Topology on Rn )

Take X to be Rn and let β = {B(0, r) : r ∈ R+ ∪ {0}}. Note that B(0, r) is
the closed ball with center 0 and radius r and that B(0, 0) = {0}. If x ∈ X
then B(0, |x|) is a minimal set in β containing x. β satisfies the conditions
of Theorem 3 so it is a basis for an Alexandroff topology on X. Notice that
this space is not Hausdorff and not compact.

Example 2. (An Alexandroff Topology on D n )

This topology is similar to the previous example on Rn , except this time we
take X = D n and β = {B(0, r) : r ∈ [0, 1]}. With the topology generated by
β, D n is compact because any open cover of D n must contain D n . Hence,
D n is a finite subcover. However, this space is also not Hausdorff.

Example 3. (Disjoint Minimal Open Neighborhoods)

Take X = R \ Z and β = {(n, n + 1) : n ∈ Z}. Then X is an Alexandroff
space with S(x) = (n, n + 1) where n < x < n + 1. For any two minimal
open neighborhoods S(x) 6= S(y) we have that S(x) and S(y) are disjoint.

Example 4. (An Alexandroff Topology on S 1 )

Take X = S 1 and define the sets Rn = {z ∈ S 1 : z n = 1} for n ∈ N ∪ {0}.
Let β = {Rn : n ∈ N ∪ {0}}. The point 1 ∈ S 1 has the set R1 = {1} as a

3
minimal set in β containing it. Suppose z ∈ S 1 and z 6= 1. If z n = 1 holds
only for n = 0, then R0 = S 1 is a minimal set in β containing z. If z n = 1
for some n 6= 0 then let m = min{n ∈ N : z n = 1}. Suppose z ∈ Rp for some
p > m, then p = qm + r where 0 ≤ r < m. Then 1 = z p = (z m )q z r = z r .
But r < m so it must be that r = 0. Hence, p = qm. If ρ ∈ Rm , then
ρp = (ρm )q = 1 so ρ ∈ Rp . Hence, Rm ⊂ Rp and so Rm is a minimal set in
β containing z. This space is compact because any open cover contains S 1 .
We conclude this section with a simple statement about minimal open
neighborhoods.
Theorem 5. If X is an Alexandroff space, then S(x) is compact for all
x ∈ X.
Proof. Let {Vα }α∈A be an open cover of S(x). Then x is in Vα for some
α ∈ A. So we must have S(x) ⊂ Vα . Hence, Vα is a finite subcover of
{Vα }α∈A .

3 New Alexandroff Spaces From Old Spaces

The tools in this section will allow us to construct new Alexandroff spaces
from given ones.
Theorem 6. If X and Y are Alexandroff spaces, then X × Y is also an
Alexandroff space, with S(x, y) = S(x) × S(y).
Proof. X × Y has as basis β = {U × V : U is open in X and V is open in
Y }. Let (x, y) ∈ X × Y , then S(x) × S(y) is in β and the claim is that this
is a minimal set in β containing (x, y). If (x, y) ∈ U × V ∈ β, then x ∈ U
and y ∈ V so S(x) ⊂ U and S(y) ⊂ V . Therefore, S(x) × S(y) is contained
in U × V . So by applying Theorem 3, we know that X × Y is an Alexandroff
space and S(x, y) = S(x) × S(y).

Corollary 2. If X1 , ..., Xn are Alexandroff spaces, then so is X1 × ... × Xn .

Furthermore, S(x1 , ..., xn ) = S(x1 ) × ... × S(xn ).
Proof. Use induction and apply Theorem 6.

4
Theorem 7. If A is a subspace of the Alexandroff space X, then A is an
Alexandroff space. In addition, SA (x) = A ∩ SX (x).

Proof. Let x ∈ A and suppose U is an open neighborhood of x in A. Then

U = A ∩ V where V is open in X. This means that SX (x) must be contained
in V , so that A ∩ SX (x) ⊂ A ∩ V = U. Hence, A is an Alexandroff space
with SA (x) = A ∩ SX (x).

Theorem 8. If X/ ∼ is a quotient space of the Alexandroff space X, then

X/ ∼ is an Alexandroff space.

Proof. Let q T: X → X/ ∼ be the quotient map. Consider T an arbitrary

intersection, α∈A Uα , of open sets in X/ ∼. We have q ( α∈A Uα ) =
T
−1

α∈A q (Uα ). Now, q T (Uα ) is open in X for each α ∈ A because

T q is the
−1 −1

quotient map. Hence, α∈A q (Uα ) is open in X and therefore α∈A Uα is

−1

open in X/ ∼ by definition of the quotient topology.

Example 5. If X is an Alexandroff space, then we can define an equivalence

relation ∼ on X by, x ∼ y iff S(x) = S(y). We can then from the quotient
space X/ ∼, which is an Alexandroff space by Theorem 8. The following
theorem will tell us exactly when X/ ∼ is a discrete space, but we first need
a definition.

Definition 2. If X is an Alexandroff space and x ∈ X, then S(x) is called

irreducible if S(y) ⊂ S(x) implies S(y) = S(x).

Theorem 9. X/ ∼ is a discrete space iff S(x) is irreducible for all x in X.

Proof. ⇒) Suppose X/ ∼ is discrete. If q is the quotient map, then q −1 ([x])

is open in X for all x. Then S(x) ⊂ q −1 ([x]) because x is in q −1 ([x]). If
y ∈ q −1 ([x]), then x ∼ y which means S(x) = S(y). Therefore, y is in
S(x) which gives us q −1 ([x]) = S(x). Now suppose S(z) ⊂ S(x), then z is
in q −1 ([x]). So we must have S(z) = S(x) because z ∼ x. Hence, S(x) is
irreducible.

⇐) Now suppose S(x) is irreducible for all x in X. Let y ∈ q −1 ([x]), then

x ∼ y so S(x) = S(y). This gives us q −1 ([x]) ⊂ S(x). Now if y ∈ S(x),
then S(y) ⊂ S(x) and since S(x) is irreducible, we must have S(y) = S(x).

5
Therefore, x ∼ y so that y ∈ q −1 ([x]). So we have q −1 ([x]) = S(x) is open in
X. This means that [x] is open in X/ ∼ and so it is discrete.

Theorem 10. If S(x) and S(y) are distinct irreducible subsets of X, then
S(x) ∩ S(y) = ∅.

Proof. Suppose S(x) and S(y) are distinct irreducible subsets of X. Suppose
that z ∈ S(x) ∩ S(y). Then S(z) ⊂ S(x) ∩ S(z) which implies that S(z) ⊂
S(x) and S(z) ⊂ S(y). Using the irreducibility of S(x) and S(y) gives us
S(x) = S(z) = S(y), which is a contradiction. Hence, S(x) ∩ S(y) = ∅.

Hausdorff Alexandroff Spaces

For general topological spaces, the spaces that satisfy the Hausdorff prop-
erty are the nicest to study. For Alexandroff spaces, the spaces that are
Hausdorff are not very interesting to study as the following theorem shows.

Theorem 11. X is a Hausdorff Alexandroff space iff for x 6= y in X, we

have S(x) ∩ S(y) = ∅.

Proof. ⇒) If X is Hausdorff we can find disjoint open sets U, V of X such

that x ∈ U and y ∈ V . Then S(x) ⊂ U and S(y) ⊂ V , so S(x) and S(y)
must also be disjoint.

⇐) This is trivial, just take S(x) and S(y) to be the disjoint open sets
containing x and y respectively.

Corollary 3. X is a Hausdorff Alexandroff space iff X is discrete.

Proof. ⇒) If X is Hausdorff, then we claim that S(x) = {x}. To see this

suppose y ∈ S(x). Then S(y) ⊂ S(x), which means S(y) ∩ S(x) = S(y).
And since S(y) 6= ∅, we must have y = x by Theorem 11. Hence, {x} is open
in X, so X must be discrete.

⇐). If X is discrete, then it is clearly Hausdorff.

6
4 Continuous Maps
It seems reasonable to believe that being Alexandroff is a topological property
of a space, and this is indeed the case. However, if f : X → Y is a continuous
map and X is an Alexandroff space, then f (X) is not necessarily Alexandroff.
For example, take X = N with the discrete topology and let Y = Q with
the subspace topology from R. Pick a bijection f : N → Q, then f must
always be continuous but f (N) = Q is not an Alexandroff space. We need a
stronger condition on f to insure that f (X) is an Alexandroff space.

Theorem 12. Let f : X → Y be an open and continuous map. If X

is an Alexandroff space, then so is f (X). In addition, if y ∈ f (X), then
S(y) = f (S(x)) where f (x) = y.

Proof. Let y ∈ f (X) and pick x ∈ X such that f (x) = y. Then f (S(x)) is
open in f (X) because f is an open map. Suppose that y is contained in an
open set U in f (X). This means x ∈ f −1 (U) and since f −1 (U) is open in
X, we have S(x) ⊂ f −1 (U). Therefore, f (S(x)) ⊂ U, which shows f (X) is
Alexandroff with S(y) = f (S(x)).

Corollary 4. If X is homeomorphic to Y and X is an Alexandroff space,

then so is Y .

Proof. If f is a homeomorphism between X and Y , then f is open and

continuous with f (X) = Y . By Theorem 12, Y is an Alexandroff space.

Theorem 13. If X and Y are Alexandroff spaces that satisfy the following:
i.) There is a bijection b : {S(x) : x ∈ X} → {S(y) : y ∈ Y }
ii.) There are homeomorphisms fx : S(x) → b(S(x)) for each x in X.
iii.) If S(x1 ) ∩ S(x2 ) 6= ∅ then fx1 (S(x1 ) ∩ S(x2 )) = fx2 (S(x1 ) ∩ S(x2 )).
then X and Y are homeomorphic.

Proof. Define h : X → Y by h(x) = fz (x) where x ∈ S(z). We must show

that h is well-defined. If x ∈ S(z1 ) and x ∈ S(z2 ), then fz1 (x) = fz2 (x)
by iii.). So the choice for h(x) does not depend on which z we choose.
Now suppose y ∈ Y , then there exists x ∈ X such that b(S(x)) = S(y).
Then fx : S(x) → S(y) is a homeomorphism so there exists z ∈ S(x) such

7
that fx (z) = y. So h(z) = fx (z) = y, which means h is onto. Suppose
h(x1 ) = h(x2 ) = y, then y ∈ b(S(x1 )) so we must have S(y) ⊂ b(S(x1 )).
But since fx−11
(S(y)) is open in S(x1 ) and x1 ∈ fx−1 1
(S(y)), we must have
S(x1 ) = fx1 (S(y)). Therefore, b(S(x1 )) = S(y). Similarly, we get that
−1

b(S(x2 )) = S(y). And b is injective so S(x1 ) = S(x2 ). This tells us that

fx1 = fx2 , so that fx1 (x1 ) = fx1 (x2 ). But this map is injective so we must
have x1 = x2 . So h is also injective. To show continuity, we only need to
use the basis of Y of minimal open neighborhoods. h−1 (S(y)) = S(x), where
h(x) = y so this is an open set. Hence, h is continuous. Similarly, h−1 is
continuous.

5 Compact Alexandroff Spaces

We will begin the study of compact Alexandroff spaces. If X is a compact
Alexandroff space, then it is covered by the set {S(x) : x ∈ X}. Hence, it
must be covered by a finite number of minimal open neighborhoods. This
key property will allow us to define several invariants of compact Alexandroff
spaces.
Definition 3. If X is a compact Alexandroff space, then we define min(X) =
min{|V | : V is a finite cover of X by minimal open neighborhoods}.
Example 6. The topologies given in Example 2 and Example 4 on D n and
S 1 are both compact. We have that min(D n ) = 1 and min(S 1 ) = 1.
The following theorem shows us that the min(X) of an Alexandroff space,
is an invariant of that space. However, it is not enough to distinguish between
the two spaces given in examples 2 and 4.
Theorem 14. If X and Y are compact Alexandroff spaces that are homeo-
morphic, then min(X) = min(Y ).
Proof. Let h : X → Y be a homeomorphism. Let {S(y1), ..., S(ymin(Y ) )}
be an open cover of Y by minimal open neighborhoods. Then {S(x1 ) =
h−1 (S(y1 )), ..., S(xmin(Y ) ) = h−1 (S(ymin(Y ) )}, where h(xi ) = yi , is an open
cover of X by minimal open neighborhoods. Therefore, min(X) ≤ min(Y ).
By a similar argument in the other direction, we get min(Y ) ≤ min(X).
Hence, min(X) = min(Y ).

8
 As seen in Theorem 13, we can define a homeomorphism between two
Alexandroff spaces by specifying homeomorphisms between the minimal open
neighborhoods of the spaces. This shows that minimal open neighborhoods
are the natural objects of study in Alexandroff spaces. For compact Alexan-
droff spaces, there is an even more basic object of study.

Definition 4. If X is an Alexandroff space, then S(x) is called basic if for

any S(y); if S(x) ⊂ S(y) and S(z) ⊂ S(y) then S(x) ⊂ S(z) and if S(x) is
not contained in S(y), then S(x) and S(y) are disjoint.

Example 7. In Example 1 and Example 2, the only basic set is the set {0}.
In Example 4, the only basic set is {1}. Example 3 has an infinite number
of basic sets because each minimal open neighborhood is also a basic set.

Theorem 15. If S(x) is a basic subset of X, then S(x) is irreducible.

Proof. Suppose that S(y) ⊂ S(x). Since S(x) is basic, if S(x) is not contained
in S(y) we must have S(x) ∩ S(y) = ∅. This can’t be because y is in S(y)
and S(x). Hence, S(x) ⊂ S(y), which implies S(y) = S(x).

Corollary 5. If S(x) and S(y) are basic subsets of X, then they are disjoint.

Proof. First apply Theorem 15 and then Theorem 10.

Theorem 16. If X is an Alexandroff space, then S(x) contains at most one

basic set for each x ∈ X.

Proof. Suppose S(x) contains two basic sets S(y) and S(z). Then by defini-
tion of basic, S(y) ⊂ S(z) and S(z) ⊂ S(y). Hence, S(y) = S(z).

The number of basic sets in a compact Alexandroff space will give us

another invariant of the space. We must first show that there is always a
finite number of basic sets in a compact Alexandroff space.

Theorem 17. If X is a compact Alexandroff space, then the number of basic

sets is less than or equal to min(X).

9
Proof. Cover X by {S(x1 ), ..., S(xmin(X) )}. Suppose the number of basic
subsets of X is greater than min(X). We claim that each basic set, S(y), is
contained in S(xi ) for some i. Suppose not, then S(y) ∩ S(xi ) = ∅ for all i.
This is a contradiction because the S(xi ) cover X. Hence, S(y) ⊂ S(xi ) for
some i. If the number of basic subsets of X is greater than min(X), then
some S(xi ) contains more than one basic set which contradicts Theorem 16.
So we must have the number of basic subsets of X less than or equal to
min(X).

Definition 5. If X is a compact Alexandroff space, then define index(X) to

be the number of basic subsets of X.

Example 8. Then index of the spaces in examples 2 and 4 is 1.

Theorem 18. If X and Y are compact Alexandroff spaces that are homeo-
morphic, then index(X) = index(Y ).

Proof. Let h : X → Y be a homeomorphism. Suppose S(x) is a basic

subset of X. Then we claim that h(S(x)) = S(h(x)) is a basic subset of Y .
Suppose S(h(x)) ⊂ S(y) and S(z) ⊂ S(y), then S(h−1 (y)) = h−1 (S(y)) ⊃
h−1 (S(z)) = S(h−1 (z)) and S(x) ⊂ S(h−1 (y)). So we must have S(x) ⊂
S(h−1 (z)) which means S(h(x)) ⊂ S(z). If S(y) does not contain S(h(x)),
then S(h−1 (y)) does not contain S(x), which implies S(h−1 (y)) ∩ S(x) = ∅.
Therefore, S(y) ∩ S(h(x)) = ∅. So S(h(x)) is a basic subset of Y . This gives
us index(X) ≤ index(Y ). By a similar argument, we obtain index(Y ) ≤
index(X). So index(X) = index(Y ).

10