FINITE TOPOLOGICAL SPACES

NOTES FOR REU BY J.P. MAY

Contents
1. Alexandroff spaces and finite spaces 1
2. Alexandroff spaces, preorders, and partial orders 4
3. Continuous maps and order-preserving functions 4
4. Finite spaces and homeomorphisms 5
5. Spaces with at most four points 7
6. Connectivity and path connectivity 9
7. Function spaces and homotopies 11
8. Homotopy equivalences 12
References 14

1. Alexandroff spaces and finite spaces

It is a standard saying that one picture is worth a thousand words. Since the
author is not good at drawing pictures, there will not be as many as there should
be. The reader should draw lots of them!
In mathematics, it is perhaps fair to say that one good definition is worth a
thousand calculations. The author likes to make up definitions and to see relations
between seemingly unrelaed concepts, so we will do lots of that.
However, to quote a slogan from a T-shirt worn by one of the author’s students,
“calculation is the way to the truth”. There is a need for more calculational un-
derstanding of the subject here, and the author, being too old and lazy to compute
himself, hopes that readers will be inspired.
The intuitive notion of a set in which there is a prescribed description of nearness
of points is obvious. Formulating the “right” general abstract notion of what a
“topology” on a set should be is not. Distance functions lead to metric spaces,
which is how we usually think of spaces. Hausdorff came up with a much more
abstract and general notion that is now universally accepted.
Definition 1.1. A topology on a set X consists of a set U of subsets of X, called
the “open sets of X in the topology U ”, with the following properties.
(i) The empty set ∅ and the set X are in U .
(ii) A finite intersection of sets in U is in U .
(iii) An arbitrary union of sets in U is in U .
A complement of an open set is called a closed set. The closed sets include ∅ and
X and are closed under finite unions and arbitrary intersections.

Date: Summer, 2003; revised summers 2008 and 2010.

1
2 NOTES FOR REU BY J.P. MAY

There are many standard operations on spaces that we shall have occasion to
use. We record three of them now and will come back to others later.
Definition 1.2. The subspace topology on A ⊂ X is the set of all intersections
A ∩ U for open sets U of X.
Definition 1.3. The topology of the union on X q Y has as open sets the unions
of an open set of X and an open set of Y .
Definition 1.4. The product topology on X × Y is the topology with basis the
products U × V of an open set U in X and an open set V in Y .
It is very often interesting to see what happens when one takes a standard
definition and tweaks it a bit. The following tweaking of the notion of a topology
is due to Alexandroff [1], except that he used a different name for the notion.
Definition 1.5. A topological space X is an A-space if the set U is closed under
arbitrary intersections.
Remark 1.6. The notion of an A-space has a pleasing complementarity. If X is an
A-space, then the closed subsets of X give it a new A-space topology. We write
X op for X with this opposite topology. Then (X op )op is the space X back again.
A space is finite if the set X is finite, and the following observation is clear.
Lemma 1.7. A finite space is an A-space.
It turns out that a great deal of what can be proven for finite spaces applies
equally well more generally to A-spaces. However, the finite spaces have recently
captured people’s attention. Since digital processing and image processing start
from finite sets of observations and seek to understand pictures that emerge from a
notion of nearness of points, finite topological spaces seem a natural tool in many
such scientific applications. There are many papers on the subject, but few of any
mathematical depth, dating from the 1980’s and 1990’s.
There was a brief early flurry of beautiful mathematical work on this subject.
Two independent papers, by McCord and Stong [11, 15], both published in 1966, are
especially interesting. We will work through them. We are especially interested in
questions raised by the union of these papers that are answered in neither and were
not pursued until quite recently. We are also interested in calculational questions
about the enumeration of finite topologies.
There is a hierarchy of “separation properties” on spaces, and intuition about
finite spaces is impeded by too much habituation to the stronger of them.
Definition 1.8. Let (X, U ) be a topological space.
(i) X is a T0 -space if for any two points of X, there is an open neighborhood
of one that does not contain the other.
(ii) X is a T1 -space if each point of X is a closed subset.
(iii) X is a T2 -space, or Hausdorff space, if any two points of X have disjoint
open neighborhoods.1
Lemma 1.9. T2 =⇒ T1 =⇒ T0 .
1The German word for separation is “Trennung”, hence the letter T for the hierarchy of
separation properties.
 FINITE TOPOLOGICAL SPACES 3

We shall omit proofs of standard and elementary results, such as this, that are
part of basic point-set topology. However, for the reader’s convenience, we give a
summary outline of that subject in Chapter ??.
In most of topology, the spaces considered are Hausdorff. For example, metric
spaces are Hausdorff. Intuition gained from thinking about such spaces is rather
misleading when one thinks about finite spaces.

Definition 1.10. The discrete topology on X is the topology in which all sets are
open. The trivial or coarse topology on X is the topology on X in which ∅ and X
are the only open sets. We write Dn and Cn for the discrete and coarse topologies
on a set with n elements. They are the largest and the smallest possible topologies
(in terms of the number of open subsets).

Lemma 1.11. If a finite space is T1 , then it is discrete.

Proof. Every subset is a union of finitely many points, hence is closed. Therefore
every set is open. 

In contrast, finite T0 -spaces are very interesting. The following problem might
be a bit difficult right now, but its solution will shortly become evident.

Exercise 1.12. Show (by induction) that a finite T0 space has at least one point
which is a closed subset.

Finite spaces have canonical minimal “bases”, which we describe next.

Definition 1.13. A basis B for a topological space X is a set of open sets, called
basic open sets, with the following properties.
(i) Every point of X is in some basic open set.
(ii) If x is in basic open sets B1 and B2 , then x is in a basic set B3 that is
contained in B1 ∩ B2 .
If B is a set satisfying these two properties, the topology generated by B is the set
U of subsets U of X such that, for each point x ∈ U , there is a set B in B such
that x ∈ B ⊂ U .

Example 1.14. The set of singleton sets {x} is a basis for the discrete topology
on X. The set of disks Dr (x) = {y|d(x, y) < r} is a basis for the topology on a
metric space X.

Lemma 1.15. B is a basis for a topology U if and only if for each x ∈ U ∈ U ,

there is a B ∈ B such that x ∈ B ⊂ U .

Definition 1.16. Let X be an Alexandroff space. For x ∈ X, define Ux to be the

intersection of the open sets that contain x. Define a relation ≤ on the set X by
x ≤ y if x ∈ Uy or, equivalently, Ux ⊂ Uy . Write x < y if the inclusion is proper.

Lemma 1.17. The set of open sets Ux is a basis for X. Indeed, it is the unique
minimal basis for X.

Proof. The first statement is clear. If C is another basis and x ∈ X, there is a

C ∈ C such that x ∈ C ⊂ Ux . This implies C = Ux , so that Ux ∈ C . 
4 NOTES FOR REU BY J.P. MAY

2. Alexandroff spaces, preorders, and partial orders

Here we relate Alexandroff spaces to the combinatorial notions of preorder and
partial order.
Definition 2.1. A preorder on a set X is a reflexive and transitive relation ≤; thus
x ≤ x and if x ≤ y and y ≤ z, then x ≤ z. A preorder is a partial order if it is
antisymmetric, which means that x ≤ y and y ≤ x implies x = y.
Lemma 2.2. The relation ≤ on an A-space X is reflexive and transitive, hence
(X ≤) is a preorder; it is a partial order if and only if the space X is T0 .
Proof. The first statement is clear. For the second, x ≤ y and y ≤ x means that
Ux = Uy . This holds if and only if every open set that contains either x or y also
contains the other. 
Lemma 2.3. A preorder (X, ≤) determines a topology U with basis the set of all
sets Ux = {y|y ≤ x}, and (X, U ) is an A-space; it is T0 if and only if (X, ≤) is a
partial order.
Proof. If x ∈ Uy and x ∈ Uz , then x ≤ y and x ≤ z, hence x ∈ UX ⊂ Uy ∩ Uz .
Therefore {Ux } is a basis for a topology. The intersection U of a set {Ui } of open
subsets is open since if x ∈ U , then Ux ⊂ Ui for each i and therefore U is the union
of these Ux . For the second statement, x ≤ y and y ≤ x if and only if Ux = Uy ,
and the T0 property then gives that x = y. 
We put things together to obtain the following conclusion.
Proposition 2.4. For a set X, the A-space topologies on X are in bijective cor-
respondence with the preorders on X. The topology U corresponding to ≤ is T0 if
and only if the relation ≤ is a partial order.
Remark 2.5. If ≤ is a preorder on X, the opposite preorder is given by x ≤op y if
and only if y ≤ x. The corresponding A-space is X op .

3. Continuous maps and order-preserving functions

The real force of the comparison between A-spaces and preorders comes from the
fact that, with the appropriate definitions, continuous maps correspond precisely
to order-preserving functions.
Definition 3.1. Let X and Y be topological spaces. A function f : X −→ Y is
continuous if f −1 (V ) is open in X for each open set V in Y . We call continuous
functions “maps”. A map f is a homeomorphism if f is one–to-one and onto and
its inverse function is continuous.
Definition 3.2. Let X and Y be preorders. A function f : X −→ Y is order-
preserving if x ≤ y implies f (x) ≤ f (y).
Lemma 3.3. A function f : X −→ Y between A-spaces is continuous if and only
if it is order preserving.
Proof. Let f be continuous and suppose that x ≤ y. Then x ∈ Uy ⊂ f −1 Uf (y) and
thus f (x) ∈ Uf (y) , which means that f (x) ≤ f (y). For the converse, let V be open
in Y . If f (y) ∈ V , then Uf (y) ⊂ V . If x ∈ Uy , then x ≤ y and thus f (x) ≤ f (y)
and f (x) ∈ Uf (y) ⊂ V , so that x ∈ f −1 (V ). Thus f −1 (V ) is the union of these Uy
and is therefore open. 
 FINITE TOPOLOGICAL SPACES 5

4. Finite spaces and homeomorphisms

From now on, except where stated otherwise, X is a finite space. We write
|X| for the number of points in X. We have chosen to work with finite spaces
for simplicity and clarity. However, just as in the sections above, most of what
we do applies verbatim, or with minor changes, to A-spaces. At first sight, one
might think that finite spaces are uninteresting since they are just finite preorders
in disguise, but that turns out to be far from the case.
Topologists are only interested in spaces up to homeomorphism, and we proceed
to classify finite spaces up to homeomorphism. Let X and Y be finite spaces in
what follows.

Lemma 4.1. A map f : X −→ X is a homeomorphism if and only if f is either

one–to–one or onto.

Proof. By finiteness, one–to–one and onto are equivalent. Assume they hold. Then
f induces a bijection 2f from the set 2X of subsets of f to itself. Since f is
continuous, if f (U ) is open, then so is U . Therefore the bijection 2f must restrict
to a bijection from the topology U to itself. 

The previous lemma fails if we allow different topologies on X: there are con-
tinuous bijections between different topologies. We proceed to describe how to
enumerate the distinct topologies up to homeomorphism. There are quite a few
papers on this enumeration problem in the literature, although some of them focus
on enumeration of all topologies, rather than homeomorphism classes of topolo-
gies [3, 4, 6, 5, 9, 7, 8, 10, 13, 14]. The difference already appears for two point
spaces, where there are four distinct topologies but three inequivalent topologies,
that is three non-homeomorphic two point spaces. Here is a table lifted straight
from Wikipedia that gives an idea of the enumeration.

n Distinct Distinct Inequivalent Inequivalent

topologies T0 -topologies topologies T0 -topologies
1 1 1 1 1
2 4 3 3 2
3 29 19 9 5
4 355 219 33 16
5 6942 4231 139 63
6 209,527 130,023 718 318
7 9,535,241 6,129,859 4,535 2,045
8 642,779,354 431,723,379 35,979 16,999
9 63,260,289,423 44,511,042,511 363,083 183,231
10 8,977,053,873,043 6,611,065,248,783 4,717,687 2,567,284

Through n = 9, a published source for the fourth column is [9]. However, this
is not the kind of enumeration problem for which one expects to obtain a precise
answer for all n. Rather, one expects bounds and asymptopics. There is a precise
formula relating the second column to the first column, but we are really only
interested in the last column. In fact, we are far more interested in refinements of
the last column that shrink its still inordinately large numbers to smaller numbers
of far greater interest to an algebraic topologist.
6 NOTES FOR REU BY J.P. MAY

We shall explain how to reduce the determination of the fourth column to a

matrix computation, using minimal bases. For this purpose, it is convenient to
describe minimal bases for the topology on X without reference to their enumeration
by the elements x ∈ X, since the latter can give redundant information.
Lemma 4.2. A set B of nonempty subsets of X is the minimal base for a topology
if and only if
(i) Every point of X is in some set B in B.
(ii) The intersection of two sets in B is a union of sets in B.
(iii) If a union of sets Bi in B is again in B, then the union is equal to one of
the Bi .
Proof. Conditions (i) and (ii) are equivalent to saying that B is a basis, and then
the minimal basis is contained in B. If (iii) also holds, then each B in B, being a
union of sets of the form Ux , must be one of the Ux . Conversely, if B is the minimal
basis and Ux is in B and is the union of sets Uy , then x is in Uy for some y and
thus Ux = Uy , so (iii) holds. 
This result implies the following descriptions of the relationship between minimal
bases and subspaces, disjoint unions, and products.
Lemma 4.3. If A is a subspace of X, the minimal basis of A consists of the
intersections A ∩ U , where U is in the minimal basis of X.
Lemma 4.4. The minimal basis of X q Y is the union of the minimal basis of X
and the minimal basis of Y .
Lemma 4.5. The minimal basis of X × Y is the set of products U × V , where U
and V are in the minimal bases of X and Y .
Definition 4.6. Consider square matrixes M = (ai,j ) with integer entries that
satisfy the following properties.
(i) ai,i ≥ 1.
(ii) ai,j is −1, 0, or 1 if i 6= j.
(iii) ai,j = −aj,i if i 6= j.
(iv) ai1 ,is = 0 if there is a sequence of distinct indices {i1 , · · · , is } such that
s > 2 and aik ,ik+1 = 1 for 1 ≤ k ≤ s − 1.
Say that two such matrices M and N are equivalent if there is a permutation matrix
T such that T −1 M T = N and let M denote the set of equivalence classes of such
matrices.
Theorem 4.7. The homeomorphism classes of finite spaces are in bijective cor-
respondence with M . The number of sets in a minimal basis for a finite space X
determines the size of the corresponding matrix, and the trace of the matrix is the
number of elements of X. Moreover, X is a T0 -space if and only if the diagonal
entries ai,i are all one.
Proof. We work with minimal bases for the topologies rather than with elements of
the set. For a minimal basis U1 , · · · , Ur of a topology U on a finite set X, define
an r × r matrix M = (ai,j ) as follows. If i = j, let ai,i be the number of elements
x ∈ X such that Ux = Ui . Define ai,j = 1 and aj,i = −1 if Ui ⊂ Uj and there
is no k (other than i or j) such that Ui ⊂ Uk ⊂ Uj . Define ai,j = 0 otherwise.
Clearly (i)–(iv) hold, and a reordering of the basis results in a permutation matrix
 FINITE TOPOLOGICAL SPACES 7

that conjugates M into the matrix determined by the reordered basis. Thus X
determines an element of M .
If f : X −→ Y is a homeomorphism, then f determines a bijection from the basis
for X to the basis for Y that preserves inclusions and the number of elements that
determine corresponding basic sets, hence X and Y determine the same element
of M . Conversely, suppose that X and Y have minimal bases {U1 , · · · , Ur } and
{V1 , · · · , Vr } that give rise to the same element of M . Reordering bases if necessary,
we can assume that they give rise to the same matrix. For each i, choose a bijection
fi from the set of elements x ∈ X such that Ux = Ui and the set of elements y ∈ Y
such that Vy = Vi . We read off from the matrix that the fi together specify
a homeomorphism f : X −→ Y . Therefore our mapping from homeomorphism
classes to M is one–to–one.
To see that our mapping is onto, consider an r × r-matrix M of the sort under
consideration and let X be the set of pairs of integers (u, v) with 1 ≤ u ≤ r and
1 ≤ v ≤ ai,i . Define subsets Ui of X by letting Ui have elements those (u, v) ∈ X
such that either u = i or u 6= i but u = i1 for some sequence of distinct indices
{i1 , · · · , is } such that s ≥ 2, aik ,ik+1 = 1 for 1 ≤ k ≤ s − 1, and is = i. We see
that the Ui give a minimal basis for a topology on X by verifying the conditions
specified in Lemma 4.2. Condition (i) is clear since (u, v) ∈ Uu . To verify (ii) and
(iii), we observe that if (u, v) ∈ Ui and u 6= i, then Uu ⊂ Ui . Indeed, we certainly
have (u, v) ∈ Ui for all v, and if (k, v) ∈ Uu with k 6= u, we must have a sequence
connecting k to u and a sequence connecting u to i which can be concatenated to
give a sequence connecting k to i that shows that (k, v) is in Ui . To see (ii), if
(u, v) ∈ Ui ∩ Uj , then Uu ⊂ Ui ∩ Uj , which implies that Ui ∩ Uj is a union of sets Uu .
To see (iii), if a union of sets Ui is a set Uj , there is an element of Uj in some Ui
and then Uj ⊂ Ui , so that Uj = Ui . A counting argument for the diagonal entries
and consideration of chains of inclusions show that the matrix associated to the
topology whose minimal basis is {Ui } is the matrix M that we started with. 

5. Spaces with at most four points

We describe the homeomorphism classes of spaces with at most four points, with
just a start on taxonomy.
There is a unique space with one point, namely C1 = D1 .
There are three spaces with two points, namely C2 , P2 = CD1 , D2 .
Proper subsets of X are those not of the form ∅ or X. We often restrict to
proper subsets when specifying topologies. The following definitions prescribe the
two names for the second space in the short list just given.
Definition 5.1. For a set with n elements, let Pn = P1,n be the space (unique up
to homeomorphism) which has only one proper open set, containing only one point;
for 1 < m < n, let Pm,n be the space whose proper open subsets are the non-empty
subsets of a given subset with m elements.
Definition 5.2. For a space X define the non-Hausdorff cone CX by adjoining
a new point ∗ and letting the proper open subsets of CX be the non-empty open
subsets of X. Thus, if |X| = n − 1, then CX ∼
= Pn−1,n .
We shall see that CX is contractible in Lemma 8.2 below.
Here is a table of the nine homeomorphism classes of topologies on a three point
set X = {a, b, c}.
8 NOTES FOR REU BY J.P. MAY

Proper open sets Name T0 ? connected?

all D3 yes no
a, b, (a,b), (b,c) D1 q P2 yes no
a, b, (a,b) P (2, 3) ∼
= CD2 yes yes
a P3 no yes
a, (a,b) CP2 ∼= (CP2 )op yes yes
a, (b,c) D1 q C2 no no
a, (a,b), (a,c) (CD2 )op yes yes
(a,b) CC2 ∼ = P3
op
no yes
op
none C3 = D3 no yes

Here is a tabulation of bases for the proper open subsets of the 33 homeomor-
phism classes of topologies on a four point space X = {a, b, c, d}. That is, the
topologies are obtained by adding in the empty set, the whole set, and all unions
of the listed sets. The list is ordered by decreasing number of singleton sets in the
topology, and, when that is fixed, by increasing number of two-point subsets and
then by increasing number of three-point subsets.
 FINITE TOPOLOGICAL SPACES 9

1 all
2 a, b, c, (a,b), (a,c), (b,c), (a,b,c)
3 a, b, c, (a,b), (a,c), (b,c), (a,b,c), (a,b,d)
4 a, b, c, (a,b), (a,c), (b,c), (a,d), (a,b,c), (a,b,d), (a,c,d)
5 a, b, (a,b)
6 a, b, (a,b), (a,b,c)
7 a, b, (a,b), (a,c,d)
8 a, b, (a,b), (a,b,c), (a,b,d)
9 a, b, (a,b), (a,c), (a,b,c)
10 a, b, (a,b), (a,c), (a,b,c), (a,c,d)
11 a, b, (a,b), (a,c), (a,b,c), (a,b,d)
12 a, b, (a,b), (c,d), (a,c,d), (b,c,d)
13 a, b, (a,b), (a,c), (a,d), (a,b,c), (a,b,d)
14 a, b, (a,b), (a,c), (a,d), (a,b,c), (a,b,d), (a,c,d)
15 a
16 a, (a,b)
17 a, (a,b), (a,b,c)
18 a, (b,c), (a,b,c)
19 a, (a,b), (a,c,d)
20 a, (a,b), (a,b,c), (a,b,d)
21 a, (b,c), (a,b,c), (b,c,d)
22 a, (a,b), (a,c), (a,b,c)
23 a, (a,b), (a,c), (a,b,c), (a,b,d)
24 a, (c,d), (a,b), (a,c,d)
25 a, (a,b), (a,c), (a,d), (a,b,c), (a,b,d), (a,c,d)
26 a, (a,b,c)
27 a, (b,c,d)
28 (a,b)
29 (a,b), (c,d)
30 (a,b), (a,b,c)
31 (a,b), (a,b,c), (a,b,d)
32 (a,b,c)
33 none

Problem 5.3. Determine which of these spaces are T0 and which are connected.
Give a taxonomy in terms of explicit general constructions that accounts for all of
these topologies. That is, determine appropriate “names” for all of these spaces.

6. Connectivity and path connectivity

We begin the exploration of homotopy properties of finite spaces by discussing
connectivity and path connectivity. We recall the general definitions.
Definition 6.1. A space X is connected it is not the disjoint union of two non-
empty open subsets. Equivalently, X is connected if the only open and closed
subsets of X are ∅ and X. Define an equivalence relation ∼ on X by x ∼ y if x and
y are elements of some connected subspace of X. An equivalence class under ∼ is
called a component of X.
10 NOTES FOR REU BY J.P. MAY

Lemma 6.2. The components of X are connected, X is the disjoint union of its
components, and any connected subspace of X is contained in a component.

Proof. Left as an exercise. (Or see Munkres [12, 3.3.1].) 

Lemma 6.3. If f : X −→ Y is a map and X is connected, then f (X) is a connected

subspace of Y .

Proof. Left as an exercise. (Or see Munkres [12, 3.1.5].) 

Let I = [0, 1] with its usual metric topology as a subspace of Rn . It is a connected

space, hence so is its image under any map. A map p : I −→ X is called a path
from p(0) to p(1) in X.

Definition 6.4. A space X is path connected if any two points can be connected
by a path. Define a second equivalence relation ' on X by x ' y if there is a path
connecting x to y. An equivalence class under ' is called a path component of X.
Note that x ' y implies x ∼ y, but not conversely in general.

Lemma 6.5. The path components of X are path connected, X is the disjoint
union of its path components, and any path connected subspace of X is contained
in a path omponent. Each path component is contained in a component.

Proof. Left as an exercise. (Or see Munkres [12, 3.3.2].) 

Now return to finite spaces X. At first sight, one might imagine that there are
no continuous maps from I to a finite space, but that is far from the case. The
most important feature of finite spaces is that they are surprisingly richly related
to the “real” spaces that algebraic topologists care about.

Lemma 6.6. Each Ux is connected. If X is connected and x, y ∈ X, there is a

sequence of points zi , 1 ≤ i ≤ s, such that z1 = x, zs = y and either zi ≤ zi+1 or
zi+1 ≤ zi for i < s.

Proof. If Ux = A q B, A and B open, say x ∈ A, then Ux ⊂ A and therefore B = ∅.

Fix x and consider the set A of points y that are connected to x by some sequence
zi . We see that A is open since z ≤ z 0 implies Uz ⊂ Uz0 . We see that A is closed
since if y is not so connected to x, then neither is any point of Uy , so that the
complement of A is open. Since X is connected, it follows that A = X. 

Lemma 6.7. If x ≤ y, then there is a path p connecting x and y.

Proof. Define p(t) = x if t < 1 and p(1) = y. We claim that p is continuous. Let V
be an open set of X. If neither x nor y is in V , then p−1 (V ) = ∅. If x is in V and
y is not in V , then p−1 (V ) = [0, 1). If y is in V , then x is in Vy ⊂ V since x ≤ y,
hence p−1 (V ) = I. 

Proposition 6.8. A finite space is connected if and only if it is path connected.

Proof. The previous two lemmas imply that x ∼ y if and only if x ' y. 
 FINITE TOPOLOGICAL SPACES 11

7. Function spaces and homotopies

Definition 7.1. A space is compact if every open cover has a finite subcover.
Definition 7.2. Let X and Y be spaces and consider the set Y X of maps X −→ Y .
The compact–open topology on Y X is the topology in which a subset is open if and
only if it is a union of finite intersections of sets W (C, U ) = {f |f (C) ⊂ U }, where
C is compact in X and U is open in Y . This means that the set of all W (C, U ) is
a subbasis for the topology.
We insert a small but non-standard technical condition. Experts will recognize
that it codifies a standard property of locally compact Hausdorff spaces, but it is
also true trivially for all finite spaces.
Definition 7.3. A space X is locally compact if for each x ∈ X, there is a compact
subspace C of X and an open subspace U such that x ∈ U ⊂ C; X is smally compact
if every open subset V is locally compact. When X is finite, X is smally compact
since every subset is compact and we can take U = C = V .
Ignoring topology, for sets X, Y , and Z, functions f : X ×Y −→ Z are in bijective
correspondence with functions fˆ: X −→ Z Y via the relation f (x, y) = f (x)(y).
Returning to topology, and so restricting Z Y to consist of the continuous maps
Y −→ Z, inspection of the proof of similar statements in any standard text, for
example [12, 7.5.3, 7.5.4], shows that the following result holds.
Proposition 7.4. For spaces X, Y , and Z such that X is smally compact, a
function f : X × Y −→ Z is continuous if and only if fˆ: X −→ Z Y is continuous.
Definition 7.5. A homotopy h : f ' g is a map h : X ×I −→ Y such that h(x, 0) =
f (x) and h(x, 1) = g(x). Two maps are homotopic, written f ' g if there is a
homotopy between them.
Lemma 7.6. If X is smally compact, then homotopies h : X × I −→ Y correspond
bijectively to continuous maps j : X −→ Y I via h ↔ j if h(x, t) = j(x)(t). The
homotopy classes of maps X −→ Y are in canonical bijective correspondence with
the path components of Y X .
Definition 7.7. If Y is finite, define the pointwise ordering of maps X −→ Y by
f ≤ g if f (x) ≤ g(x) for all x ∈ X.
Proposition 7.8. If Y is finite, then the intersection of the open sets in Y X that
contain a map g is {f |f ≤ g}.
Proof. Let Vg be the cited intersection and let Zg = {f |f ≤ g}. Let f ∈ Vg
and x ∈ X. Since g ∈ W ({x}, Ug(x) ), f ∈ W ({x}, Ug(x) ), so f (x) ∈ Ug(x) and
f (x) ≤ g(x). Since x was arbitrary, f is in Zg . Conversely, let f ≤ g. Consider any
W (C, U ) which contains g and let x ∈ C. Then g(x) ∈ U and, since f (x) ≤ g(x),
f (x) ∈ Ug(x) ⊂ U . Therefore f ∈ W (C, U ) and f is in all open subsets of Y X that
contain g. 
Corollary 7.9. If X and Y are finite, then the pointwise ordering on Y X coincides
with the ordering given by its compact open topology.
Proposition 7.10. If Y is finite and f ≤ g, then f ' g by a homotopy h such that
h(x, t) = f (x) for all t and all points x ∈ X such that f (x) = g(x).
12 NOTES FOR REU BY J.P. MAY

Proof. We have the path p connecting f to g in Y X specified by p(t) = f if t < 1

and p(1) = g. Indeed, with V = W (C, U ), the proof that p is continuous is a direct
adaptation of the proof of Lemma 6.7, the key point being that if g ∈ V , then
f ∈ V by Proposition 7.8. 

8. Homotopy equivalences
We have seen that enumeration of finite sets with reflexive and transitive relations
≤ amounts to enumeration of the topologies on finite sets. We have refined this
to consideration of homeomorphism classes of finite spaces. We are much more
interested in the enumeration of the homotopy types of finite spaces. We will come
to a still weaker and even more interesting enumeration problem later.
Definition 8.1. Two spaces X and Y are homotopy equivalent if there are maps
f : X −→ Y and g : Y −→ X such that g ◦ f ' idX and f ◦ g ' idY . A space is
contractible if it is homotopy equivalent to a point.
This relationship can change the number of points. We have a first example.
Lemma 8.2. If X is a space containing a point y such that the only open (or only
closed) subset of X containing y is X itself, then X is contractible. In particular,
the non-Hausdorff cone CX is contractible for any X.
Proof. This is a variation on a theme we have already seen twice. Let ∗ denote a
space with a single point, also denoted ∗. Define r : X −→ ∗ by r(x) = ∗ for all
x and define i : ∗ −→ X by i(∗) = y. Clearly r ◦ i = id. Define h : X × I −→ X
by h(x, t) = x if t < 1 and h(x, 1) = y. Then h is continuous. Indeed, let U be
open in X. If y ∈ U , then U = X and h−1 (U ) = X × I, while if y ∈ / U , then
h−1 (U ) = U × [0, 1). The argument when X is the only closed subset containing y
is the same. Clearly h is a homotopy id ' i ◦ r. 
Corollary 8.3. If X is finite, then Ux is contractible.
Proof. The only open subset of Ux that contains x is Ux itself. 
The following result of McCord [11, Thm. 4] says that, when studying finite
spaces up to homotopy type, there is no loss of generality if we restrict attention
to T0 -spaces, that is, to finite posets (poset = partially ordered set).
Theorem 8.4. Let X be a finite space. There is a quotient T0 -space X0 such that
the quotient map qX : X −→ X0 is a homotopy equivalence. For a map f : X −→ Y
of finite spaces, there is a unique map f0 : X0 −→ Y0 such that qY ◦ f = f0 ◦ qX .
Proof. Define x ∼ y if Ux = Uy , or, equivalently, if x ≤ y and y ≤ x. Let X0 be the
set of equivalence classes and let q = qX send x to its equivalence class [x]. Give
X0 the quotient topology. This means that a subset V of X0 is open if and only
if q −1 (V ) is open in X. Clearly q is continuous. The relation ≤ on X induces a
relation ≤ on X0 . Since X0 is finite, we have the open set Uq(x) for x ∈ X. Observe
that q −1 q(Ux ) = Ux since if q(y) = q(z) where z ∈ Ux , then y ∈ Uy = Uz ⊂ Ux .
Therefore q(Ux ) is open, hence contains Uq(x) . Conversely, Ux ⊂ q −1 (Uq(x) ) by
continuity and thus q(Ux ) ⊂ Uq(x) . This proves that q(Ux ) = Uq(x) . It follows that
[x] ≤ [y] if and only if x ≤ y. Indeed, q(x) ≤ q(y) implies q(x) ∈ Uq(y) = q(Uy ).
Thus q(x) = q(z) for some z ∈ Uy and Ux = Uz ⊂ Uy , so that x ≤ y. Conversely,
 FINITE TOPOLOGICAL SPACES 13

if x ≤ y, then Ux ⊂ Uy and therefore Uq(x) ⊂ Uq(y) , so that q(x) ≤ q(y). It follows

that ≤ is antisymmetric on X0 , so that X0 is a T0 -space.
We must prove that q is a homotopy equivalence. Let f : X0 −→ X be any
function such that q ◦ f = id. That is, we choose a point from each equivalence
class. By what we have just proven, f preserves ≤ and is therefore continuous. Let
g = f ◦ q. We must show that g is homotopic to the identity. We see that g is
obtained by first choosing one xu with Uxu = U for each U in the minimal basis for
X and then letting g(x) = xu if Ux = U . Thus Ug(x) = Ux and g(x) ∈ Ux , which
means that g ≤ id. Now Proposition 7.10 gives the required homotopy h : id ' g.
Note that h(g(x), t) = g(x) for all t.
For the last statement, a map f : X −→ Y is a function that preserves ≤, and it
follows that it induces a unique function f0 : X0 −→ Y0 such that qY ◦ f = f0 ◦ qX .
Clearly f0 preserves ≤ and is thus continuous. 
The space T0 is called the Kolmogorov quotient of T . The construction is classical
and has many other applications. We conclude that to classify finite spaces up to
homotopy equivalence, it suffices to classify T0 -spaces up to homotopy equivalence.
Stong [15, §4] has given an interesting way of studying this. We change his language
a bit in the following exposition.
Definition 8.5. Let X be a finite space.
(a) A point x ∈ X is upbeat if there is a y > x such that z > x implies z ≥ y.
(b) A point x ∈ X is downbeat if there is a y < x such that z < x implies z ≤ y.
X is a minimal finite space if it is a T0 -space and has no upbeat or downbeat
points. A core of a finite space X is a subspace Y that is a minimal finite space
and a deformation retract of X. That is, if i : Y −→ X is the inclusion, there is a
map r : X −→ Y such that r ◦ i = id together with a homotopy h : X × I −→ X
from id to i ◦ r such that h(y, t) = y if y ∈ Y .
Remark 8.6. If we draw a graph of a poset by drawing a line upwards from x to y
if x < y, we see that, above an upbeat point x, the graph looks like
z1 B z2 ··· m zs
BB
BB mmmmm
BB mmm
B mmmmm
ym

x
Turning the picture upside down, we see what the graph below a downbeat point
looks like.
Intuitively, identifying x and y and erasing the line between them should not
change the homotopy type. We say this another way in the proof of the following
result, looking at inclusions rather than quotients in accordance with our definition
of a core.
Theorem 8.7. Any finite (or finite based) space X has a core.
Proof. With the notations of the proof of Theorem 8.4, identify X0 with its image
g(X0 ) ⊂ X. The proof of Theorem 8.4 shows that X0 , so interpreted, is a defor-
mation retract of X. Thus we may as well assume that X is T0 . Suppose that
14 NOTES FOR REU BY J.P. MAY

X has an upbeat point x. We claim that the subspace X − {x} is a deformation

retract of X. To see this define f : X −→ X − {x} ⊂ X by f (z) = z if z 6= x and
f (x) = y, where y > x is such that z > x implies z ≥ y. Clearly f ≥ id. We
claim that f preserves order and is therefore continuous. Thus suppose that u ≤ v.
We must show that f (u) ≤ f (v). If u = v = x or if neither u nor v is x, there is
nothing to prove. When u = x < v, f (u) = y and f (v) = v ≥ y. When u < x = v,
f (u) = u < x < y = f (v). Now Proposition 7.10 gives the required deformation.
A similar argument applies to show that X − {x} is a deformation retract of X
if x is a downbeat point. Starting with X0 , define Xi from Xi−1 by deleting one
upbeat or downbeat point. After finitely many stages, there are no more upbeat or
downbeat points left, and we arrive at the required core. 
Theorem 8.8. If X is a minimal finite space and f : X −→ X is homotopic to the
identity, then f is the identity.
Proof. First suppose that f ≥ id. For all x, f (x) ≥ x. If x is a maximal point, then
f (x) = x. Let x be any point of X and suppose inductively that f (z) = z for all
z > x. Then, by continuity, z > x implies z = f (z) ≥ f (x) ≥ x. If f (x) 6= x, this
implies that x is an upbeat point, contradicting the minimality of X. Therefore
f (x) = x. By induction, f (x) = x for all x. A similar argument shows that f ≤ id
implies f = id. By Lemma 6.6, it now follows that the component of the identity
map in the finite space X X consists only of the identity map. That is, any map
homotopic to the identity is the identity. 
Corollary 8.9. If f : X −→ Y is a homotopy equivalence of minimal finite spaces,
then f is a homeomorphism.
Proof. If g : Y −→ X is a homotopy inverse, then g ◦ f ' id and f ◦ g ' id. By the
theorem, g ◦ f = id and f ◦ g = id. 
Corollary 8.10. Finite spaces X and Y are homotopy equivalent if and only if
they have homeomorphic cores. In particular, the core of X is unique up to home-
omorphism.
Proof. This is immediate since the cores of X and Y are minimal finite spaces that
are homotopy equivalent to X and Y . 
Remark 8.11. In any homotopy class of finite spaces, there is a representative with
the least possible number of points. This representative must be a minimal finite
space, since its core is a homotopy equivalent subspace. The minimal representative
is homeomorphic to a core of any finite space in the given homotopy class.

References
[1] P.S. Alexandroff. Diskrete Räume. MatematiceskiiSbornik (N.S.) 2(1937), 501–518.
[2] J. Anusiak and K.P., Shum. Remarks on finite topological spaces. Colloquium Mathematicum
23(1971), 217–223.
[3] R. Bumby, R. Fisher, H. Levinson and R. Silverman. Topologies on Finite Sets. Proc. 9th
S-E Conf. Combinatorics, Graph Theory, and Computing (1978), 163–170.
[4] J.C. Culberson and G.J.E. Rawlins. New Results from an Algorithm for Counting Posets.
Order 7 (90/91), no 4, p 361-374.
[5] M. Erne and K. Stege. Counting Finite Posets and Topologies. Order 8 (91), p 247-265.
[6] J.W. Evans, F. Harary, and M.S. Lynn. On the computer enumeration of finite topologies.
Communications of the ACM 10(1967), 295–297, 313.
 FINITE TOPOLOGICAL SPACES 15

[7] D. Kleitman and B. Rothschild. The number of finite topologies. Proc. AMS, 25, 1970, 276-
282.
[8] D. Kleitman and B. Rothschild. Asymptotic enumeration of partial orders on a finite set.
Trans. Amer. Math. Soc. 205 (1975), 205–220.
[9] Y. Koda. The numbers of finite lattices and finite topologies. Bulletin of the Institute of
Combinatorics and its applications 10 (1994), 83–89.
[10] V. Krishnamurthy. On the number of topologies on a finite set. Amer. Math. Monthly
73(1966), 154-157.
[11] M.C. McCord. Singular homology groups and homotopy groups of finite topological spaces.
Duke Mathematical Journal 33(1966), 465–474.
[12] J.R. Munkres. Topology. Prentice–Hall. 1975.
[13] P. Renteln. On the enumeration of finite topologies. Journal of Combinatorics, Information,
and System Sciences 19(1994), 201-206.
[14] P. Renteln. Geometrical Approaches to the Enumeration of Finite Posets: An Introductory
Survey. Nieuw Archief voor Wiskunde 14 (1996), 349-371.
[15] R.E. Stong. Finite topological spaces. Trans. Amer. Math. Soc. 123(1966), 325–340.