HOMOGENEITY PROPERTIES ON

COMPACT SPACES

By

Ramiro H. de la Vega

A dissertation submitted in partial fulfillment of the

requirements for the degree of

Doctor of Philosophy

(Mathematics)

at the

UNIVERSITY OF WISCONSIN – MADISON

2005
 i

Abstract

A topological space X is said to be homogeneous if for any two points x and y in

X there is a homeomorphism f from X onto itself such that f (x) = y. An S-space

is a regular space such that all of its subspaces are separable but not all of them

are Lindelöf. First, we show that under CH there exists a compact S-space which

is homogeneous, thus answering a question of V.V. Fedorchuk.

By classic results, under CH , any compact homogeneous S-space must have

countable character and hence “small” cardinality (i.e. no more than 2ℵ0 ). How-
ℵ
ever we show here how to construct “big” (i.e. of size 22 0 ) compact S-spaces

while having certain control over their group of autohomeomorphisms. Using this

method, we get such spaces with various (non)homogeneity properties: i) We can

get the space to be zero-dimensional and h-homogeneous (i.e. all its non-empty

clopen subsets are homeomorphic) or ii) We can make the space rigid (i.e. the

only autohomeomorphism is the identity map) or iii) We can get a space which

is B-homogeneous (i.e. has a base of homeomorphic open subsets) but not h-

homogeneous. This last example partially answers a question of M.V. Matveev.

In the last chapter we give a new bound on the cardinality of homogeneous

spaces in terms of their tightness, Lindelöf degree and point-wise type. In particular

our result implies that any compact homogeneous space of countable tightness has

size no more than 2ℵ0 , thus proving a conjecture of A.V. Arkhangel’skiı̆.

 ii

Acknowledgements

First I would like to thank my advisor, Professor Kenneth Kunen, for his direction

and assistance during my research. Thanks to all those people who probably made

my stay in Madison a bit longer but much more enjoyable, especially Camilo, j-

posada, los tortolitos, Paul, Gautam, Erik and Mathilde. Thanks to my parents

for making me what I am today and just for being always there for me. Thanks

to Chabe for for all the spoiling and the good energy. Most of all I want to thank

Amelia for moving all the way to Madison, for all those Sundays and for becoming

the most important person in my life.

 iii

Contents

Abstract i

Acknowledgements ii

1 Introduction 1

1.1 Background . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1

1.2 General conventions and cardinal functions . . . . . . . . . . . . . . 5

2 A Compact Homogeneous S-space 10

2.1 A strong S-space . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10

2.2 Compactification . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13

2.3 Homogeneity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14

3 Homogeneity Properties in Large Compact S-spaces 16

3.1 Some technical facts and terminology . . . . . . . . . . . . . . . . . 17

3.1.1 Closed subspaces of C ω1 as inverse limits . . . . . . . . . . . 17

3.1.2 The Cantor space and group actions . . . . . . . . . . . . . 19

3.2 Lifting homeomorphisms . . . . . . . . . . . . . . . . . . . . . . . . 23

3.3 Main construction . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27

3.4 Applications . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32

4 Cardinality of Homogeneous Compacta 34

4.1 Elementary submodels . . . . . . . . . . . . . . . . . . . . . . . . . 35

 iv

4.2 Homogeneity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39

5 Final Remarks and Open Questions 42

Bibliography 45
 1

Chapter 1

Introduction

1.1 Background

A notion of homogeneity appears on practically every mathematical structure.

There are several reasonable ways in which one could try to define this notion for

topological spaces. The following is perhaps the most natural way to do this:

Definition 1.1 A topological space X is homogeneous if and only if for every

x, y ∈ X there is a homeomorphism f of X onto X with f (x) = y.

Every topological group is homogeneous. Recall that a topological group is a

triple hG, τ, ⋆i such that hG, τ i is a topological space, hG, ⋆i is a group, and the

function from G × G to G defined by ha, bi 7→ a ⋆ b−1 is continuous. Thus it is clear

that the function defined by a 7→ x ⋆ a−1 ⋆ y is continuous and sends x to y (and y

to x), showing that hG, τ i is homogeneous.

However not every homogeneous topological space, even a compact one, admits

a group structure compatible with its topology. A clasic example of this is the

“double arrow” space of Alexandrov and Urysohn (see [1]). This space is first

countable but not metrizable, while every topological group is metrizable if and

only if it is first countable. Another example is the Euclidian n-sphere S n ; it is

 2

homogeneous for every n, but it admits a compatible group structure if and only

if n is 0, 1 or 3.

Obviously the unit interval [0, 1] with its usual topology is not homogeneous.

However a very surprising result of Keller [19] shows that [0, 1]ω is homogeneous.

Examples like this make it natural to define power homogeneous spaces as the

ones for which some power of them is homogeneous. We should mention here

that the space [0, 1]ω does not admit a compatible group structure since it has the

fixed-point property.

To start describing our results, we need the following:

Definition 1.2 A topological space X is hereditarily separable (HS ) if and only

if every subspace is separable. An S-space is a regular Hausdorff HS space with a

non-Lindelöf subspace.

Under CH , several examples of S-spaces have been constructed, including topo-

logical groups (see [12]) and compact S-spaces (see [11, 16]). It is asked in [5]

(Problem I.5) and in [13] whether there are compact homogeneous S-spaces. The

main goal of Chapter 2 (see Theorem 2.8) is to show that there are under CH :

Theorem 1.3 (CH ) There is a compact homogeneous S-space.

This cannot be done in ZFC , since Szentmiklóssy showed in [26] that there are

no compact S-spaces under MA + ¬CH ; in fact, there are no S-spaces at all under

PFA (see [27]).

In Section 2.1, we use a slightly modified version of the construction in [16, 25]

to refine the topology of any given second-countable space, and turn it into a
 3

first-countable strong S-space (i.e., each of its finite powers is an S-space). In

Section 2.2, we show that if the original space is compact then there is a natural

compactification of the new space that is also a first-countable strong S-space.

If in addition the original space is zero-dimensional, then the ω th power of this

compactification will be homogeneous by Motorov [24], proving Theorem 2.8.

ℵ
In [11], Fedorchuk constructed under ♦ a compact S-space of size 22 0 . In

fact, his space has no non-trivial convergent sequences and every point of it has

character 2ℵ0 = ℵ1 . By classic results, it is not possible to make such a space

homogeneous, since any HS compact homogeneous space must have size at most

2ℵ0 (see Theorems 1.8 and 1.11 below). However, in Chapter 3 we show how to

construct these large S-spaces while keeping some control over their group of auto-

homeomorphisms. This will allow us to get such spaces with various homogeneity

(or non-homogeneity) properties. In order to keep the statements short, we de-

fine a Fedorchuk space as a compact hereditarily separable perfect space with no

non-trivial convergent sequences.

Definition 1.4 A Topological space is B-homogeneous if it has a base every ele-

ment of which can be mapped onto every other one by an autohomeomorphism of

the whole space. A zero-dimensional space is h-homogeneous if any two non-empty

clopen subsets are homeomorphic to each other. A topological space is rigid if the

only autohomeomorphism of the space is the identity map.

In Chapter 3 we prove the following (see Theorems 3.20 and 3.21):

Theorem 1.5 (♦) There is a rigid Fedorchuk space.

 4

Theorem 1.6 (♦) There is an h-homogeneous Fedorchuk space.

Matveev showed in [21] that any B-homogeneous zero-dimensional first count-

able compact space is h-homogeneous (and hence homogeneous). Matveev also

constructed an example of a pseudocompact zero-dimensional B-homogeneous

space which is not h-homogeneous, and he asked whether any B-homogeneous

zero-dimensional compact space is h-homogeneous. We show that under ♦ the

answer is no. In fact we will prove the following (see Theorem 3.22):

Theorem 1.7 (♦) There exists a zero-dimensional Fedorchuk space which is B-

homogeneous but not h-homogeneous.

In [7], van Douwen obtained several results concerning the homogeneity of

infinite powers of compact spaces. In particular he proved the following surprising

result:

Theorem 1.8 (E.K. van Douwen) If X is a power homogeneous Hausdorff space,

then |X| ≤ 2πw(X) .

He also suggested the possibility that there would be more relations between

cardinal functions on homogeneous spaces, which are not true without the assump-

tion of homogeneity. In [15] M. Ismail does a systematic study of cardinal functions

on homogeneous spaces and topological groups of point-countable type. Among

other results, he showed the following:

Theorem 1.9 (M. Ismail) If X is compact and homogeneous then |X| ≤ 2c(X)t(X) .
 5

It is known that any compact space of countable tightness contains a point

with character at most 2ℵ0 ; if the space is also homogeneous then it follows that
ℵ
|X| ≤ 22 0 . In [2], Arkhangel’skiı̆ asked if in fact |X| ≤ 2ℵ0 for any such space; he

later conjectured a positive answer to this question (see [5]). A well known result of

A. Dow (see [8]) states that under PFA any compact space of countable tightness

contains a point of countable character; from this it follows that Arkhangel’skiı̆’s

conjecture is true under PFA. The main goal of Chapter 4 is to give a proof in

ZF C of Arhangel’skiı̆’s conjecture. In fact we will prove (see Theorem 4.13) the

following theorem which generalizes Ismail’s result.

Theorem 1.10 If X is compact and homogeneous then |X| ≤ 2t(X) .

A related result of Arhangel’skiı̆ (see [2, 3]) states that if X is a sequential

compact homogeneous space then |X| ≤ 2ℵ0 . The previous theorem also generalizes

this result since any sequential space has countable tightness. As a corollary of our

result we also confirm a conjecture of I. Juhász, P. Nyikos and Z. Szentmiklóssy

(see [17]), stating that it is consistent that every homogeneous T5 compact space

is first countable.

1.2 General conventions and cardinal functions

We assume all spaces to be Hausdorff. If A ⊆ X we write cl(A) for the topological

closure of A in X. As usual [X]κ is the set of all subsets of X of cardinality κ and

[X]≤κ is the set of all subsets of X of size no more than κ. A set A ⊆ X is called a

Gκ -set or a Gκ -subset of X if it is the intersection of no more than κ open subsets

of X.
 6

In the rest of this section we have collected all the definitions and known facts

about cardinal functions that we will be using in the subsequent chapters. Most of

this section is based on [14], where the reader can find the proofs and authors of

the results. We start by fixing a Hausdorff space X, a subset Y ⊂ X and a point

p ∈ X.

A collection U of non-empty open subsets of X is called a π-base for X if for

every open V ⊆ X there is a U ∈ U with U ⊆ V . U is called a local π-base for p if

T
for every open neighborhood V of p there is a U ∈ U with U ⊆ V . If U = {p}

then U is called a pseudo-base for p. We say that U is a base for Y in X if for

every open V ⊂ X with Y ⊂ V there is a U ∈ U such that Y ⊆ U ⊆ V . We say

T
that U is a pseudo-base for Y in X if U = Y .

Now we have the following definitions:

weight
w(X) = min{|U| : U is a base for X}

cellularity
c(X) = sup{|U| : U is a collection of pairwise disjoint open subsets of X}

density
d(X) = min{|S| : S ⊆ X, cl(S) = X}

hereditary density
hd(X) = sup{d(Y ) : Y ⊆ X}

Lindelöf degree
L(X) = min{κ : every open cover of X has a subcover of size ≤ κ}

hereditary Lindelöf degree

hL(X) = sup{L(Y ) : Y ⊆ X}

π-weight
πw(X) = min{|U| : U is a π-base for X}
 7

character
χ(p, X) = min{|U| : U is a local base for p}
χ(Y, X) = min{|U| : U is a base for Y in X}
χ(X) = sup{χ(p, X) : p ∈ X}

pseudo-character
ψχ(p, X) = min{|U| : U is a pseudo-base for p}
ψχ(Y, X) = min{|U| : U is a pseudo-base for Y in X}
ψχ(X) = sup{ψχ(p, X) : p ∈ X}

π-character
πχ(p, X) = min{|U| : U is a π-base for p}
πχ(X) = sup{πχ(p, X) : p ∈ X}

tightness
t(p, X) = min{κ : ∀Y ⊆ X with p ∈ cl(Y ), ∃A ∈ [Y ]≤κ with p ∈ cl(A)}
t(X) = sup{t(p, X) : p ∈ X}

point-wise type
h(X) = min{κ : ∀x ∈ X∃K ⊆ X compact with x ∈ K and χ(K, X) ≤ κ}

Several relations between these cardinal functions are know to hold for compact

spaces. We list the ones that will be relevant for our work in the following:

Theorem 1.11 If X is a compact space then:

1. ψχ(X) = χ(X) ≤ hL(X).

2. πχ(X) ≤ t(X).

3. πw(X) ≤ hd(X).

4. |X| ≤ 2χ(X) .

5. If χ(x, X) ≥ κ for all x ∈ X, then |X| ≥ 2κ .

 8

We include a proof of the next well known result since we could not find one in

the literature. It allows us to generalize part 2 of the previous theorem for spaces

which are not compact, using instead the height of the space.

Lemma 1.12 Let K be a compact subset of the Hausdorff space X and let x ∈ K.

Then πχ(x, X) ≤ πχ(x, K)χ(K, X).

Proof. Since K is compact and hence regular, we can find a collection U of open

subsets of X with |U| = πχ(x, K) and with the property that for any open W ⊆ X

with x ∈ W , there is a U ∈ U such that U ∩ K 6= ∅ and cl(U) ∩ K ⊆ W . Just by

definition, we can fix a collection V of open subsets of X such that |V| = χ(K, X)

and whenever W is an open subset of X containing K, there is a V ∈ V such that

K ⊆ V ⊆ W . We let T = {U ∩ V : U ∈ U, V ∈ V} and prove that T is a local

π-base at x in X.

For this, fix an open W ⊆ X with x ∈ W . Then there is a U ∈ U such

that U ∩ K 6= ∅ and cl(U) ∩ K ⊆ W . Note that X \ cl(U) ∪ W is then an open

neighborhood of K and therefore there is a V ∈ V with K ⊆ V ⊆ X \ cl(U) ∪ W .

It is clear now that U ∩ V ⊆ W (and U ∩ V 6= ∅ since U ∩ K 6= ∅ and K ⊆ V ), so

we are done. ♠

Corollary 1.13 For any space X, πχ(X) ≤ t(X)h(X).

The following result is probably well known, but again we did not find an

explicit proof in the literature.

Theorem 1.14 If X is a Hausfdorff homogeneous space then |X| ≤ d(X)πχ(X) .

 9

Proof. Let κ = πχ(X) and λ = d(X). We fix a point p ∈ X and {Uα : α ∈ κ}

a local π-base at p in X. We also fix {dβ : β ∈ λ} a dense subset of X. Now for

each x ∈ X we let hx be any homeomorphism from X onto X such that hx (p) = x,

and we define ψx : κ → λ as follows: for each α ∈ κ, ψx (α) is the least β ∈ λ such

that dβ ∈ hx (Uα ). Now if x, y ∈ X and x 6= y, we can find an open neighborhood

W of p such that hx (W ) ∩ hy (W ) = ∅; then, if α ∈ κ is such that Uα ⊆ W , we

get that ψx (α) 6= ψy (α). This shows that the function defined by x 7→ ψx is an

injection and therefore |X| ≤ λκ . ♠

We finish this section with an easy well known result that characterizes hered-

itarily separable spaces. An ω1 -sequence hxα : α ∈ ω1 i is said to be left separated

if for each α ∈ ω1 we have that xα ∈

/ cl({xβ : β < α}).

Lemma 1.15 A topological space X contains a left separated ω1 -sequence if and

only if hd(X) > ℵ0 .

 10

Chapter 2

A Compact Homogeneous S-space

The goal of this chapter is to show that under CH there is a compact homogeneous

S-space. The key idea for this is to try to use the fact (proved by Dow and

Pearl in [9]) that any infinite power of a zero-dimensional first-countable space

is homogeneous. Naively one would then start by constructing a compact zero-

dimensional first-countable S-space Z and then Z ω would be the desired space.

The only problem with this is that being HS is not a productive property, so there

is no reason to expect that Z ω will be an S-space. Thus we actually want Z to be

what is called a strong S-space (i.e. an S-space such that Z ω is also HS ). We will

construct Z in two steps. First we will refine the topology on the Cantor space 2ω

to get a first-countable zero-dimensional locally compact strong S-space; it is for

this part of the construction that we need CH . The second task is to construct a

compactification Z of this new space, which is still zero-dimensional first-countable

and a strong S-space.

2.1 A strong S-space

If τ is the topology on a set X, we write τ I for the corresponding product topology

on X I ; likewise if τ ′ ⊆ τ is a base we write (τ ′ )I for the natural corresponding

 11

base for τ I . If E ⊆ X, then cl(E, τ ) denotes the closure of E with respect to

the topology τ . This notation will be used when we are discussing two different

topologies on the same set X.

The following two lemmas are well-known; the second is Lemma 7.2 in [25]:

Lemma 2.1 If X is HS and Y is second-countable, then X × Y is HS

Lemma 2.2 X ω is HS iff X n is HS for all n < ω.

The next lemma, an easy exercise, is used in the proof of Theorem 2.4:

Lemma 2.3 If (x, y) ∈ X × Y and S ⊆ X × Y , then (x, y) ∈ cl(S) iff y ∈

cl(π(S ∩(U ×Y ))) for all neighborhoods U of x, where π : X ×Y → Y is projection.

The following is proved (essentially) in [25], but our proof below may be a bit

simpler:

Theorem 2.4 Assume CH . Let ρ be a second-countable T3 topology on X, where

|X| = ℵ1 . Then there is a finer topology τ on X such that (ω1 , τ ) is a first-countable

locally compact strong S-space.

Proof. Without loss of generality, X = ω1 . For η < ω1 we write ρη for the

topology of η as a subspace of (ω1 , ρ). Applying CH , list 0<n<ω [(ω1 )n ]≤ω as

S

{Sµ : µ ∈ ω1 }, so that each Sµ ⊆ µn(µ) for some n(µ) with 0 < n(µ) < ω.

For η ≤ ω1 we construct a topology τη on η by induction on η so as to make

the following hold for all ξ < η ≤ ω1 :

1. τξ = τη ∩ P (ξ).
 12

2. τη is first-countable, locally compact, and T3 .

3. τη ⊇ ρη .

Note that (1) implies in particular that ξ ∈ τη ; that is, ξ is open. Thus, if

τ = τω1 , then (ω1 , τ ) is not Lindelöf. Also by (1), τη for limit η is determined from

the τξ for ξ < η. So, we need only specify what happens at successor ordinals.

For n ≥ 1 and ξ < ω1 , let Iseq(n, ξ) be the set of all f ∈ (ω1 )n that satisfy

f (0) < f (1) < · · · < f (n − 1) = ξ. The following condition states our requirement

on τξ+1 :

4. For each µ < ξ and each f ∈ Iseq(n, ξ), where n = nµ :

f ∈ cl(Sµ , (τξ+1 )n−1 × ρ) =⇒ f ∈ cl(Sµ , (τξ+1 )n ) .

If n = nµ = 1, then (τξ+1 )n−1 × ρ just denotes ρ. That is, (4) requires

ξ ∈ cl(E, ρ) =⇒ ξ ∈ cl(E, τξ+1 ) (∗)

for all E in the countable family {Sµ : µ < ξ & n(µ) = 1}. It is standard (see

[16]) that one may define τξ+1 so that this holds. Now, consider (4) in the case

n = nµ ≥ 2. By (2), τξ is second-countable, so let τξ′ be a countable base for τξ .

Applying Lemma 2.3, (4) will hold if whenever U = U0 × · · · × Un−2 ∈ (τξ′ )n−1 is a

neighborhood of f ↾(n − 1),

ξ ∈ cl(π(Sµ ∩ (U × (ξ + 1))), ρ) =⇒ ξ ∈ cl(π(Sµ ∩ (U × (ξ + 1))), τξ+1 ) ,

where π : ξ n−1 × (ξ + 1) → (ξ + 1) is projection. But this is just a requirement of

the form (∗) for countably many more sets E, so again there is no problem meeting

it.
 13

Now, we need to show that τ n is HS for each 0 < n < ω. We proceed by

induction, so assume that τ m is HS for all m < n. Fix A ⊆ (ω1 )n ; we need to show

that A is τ n -separable. Applying the induction hypothesis, we may assume that

each f ∈ A has all coordinates distinct. Also, since permutation of coordinates

induces a homeomorphism of (ω1 )n , we may assume that each f ∈ A is strictly

increasing; that is, f ∈ Iseq(n, ξ), where ξ = f (n−1). By the induction hypothesis

and Lemma 2.1, A is separable in τ n−1 × ρ. We can then fix µ such that n(µ) = n,

Sµ ⊆ A, and Sµ is τ n−1 × ρ-dense in A. Now, say f ∈ A with ξ = f (n − 1) > µ.

Applying (4), we have f ∈ cl(Sµ , τ n ). Thus, A \ cl(Sµ , τ n ) is countable, so A is

τ n -separable. ♠

2.2 Compactification

We need the following generalization of the Aleksandrov duplicate construction.

Similar generalizations have been described elsewhere; see in particular [6], which

also gives references to the earlier literature.

Definition 2.5 If ϕ is a continuous map from the T2 space Y into X, then Y ∪˙ ϕ X

denotes the disjoint union of X and Y , given the topology which has as a base:

a. All open subsets of Y , together with

b. All [U, K] := U ∪ (ϕ−1 U \ K), where U is open in X and K is compact in Y .

Our main interest here is in the case where X is compact and Y is locally

compact. Then, if |X| = 1, we have the 1-point compactification of Y , and if Y is

discrete and ϕ is a bijection we have the Aleksandrov duplicate of X.

 14

Lemma 2.6 Let Z = Y ∪˙ ϕ X, with X and Y Hausdorff:

1. X is closed in Z, Y is open in Z, and both X and Y inherit their original

topology as subspaces of Z.

2. If Y is locally compact, then Z is Hausdorff.

3. If X is compact, then Z is compact.

4. If X and Y are first-countable, X is compact, Y is locally compact, and each

ϕ−1 (x) is compact, then Z is first-countable.

5. If X and Y are zero-dimensional, X is compact, and Y is locally compact,

then Z is zero-dimensional.

6. If X is second-countable and Y ω is HS , then Z ω is HS .

Proof. For (3): If U is a basic open cover of Z, then there are n ∈ ω and
S S
[Ui , Ki] ∈ U for i < n such that i<n Ui = X. Thus, i<n [Ui , Ki ] contains all
S
points of Z except for (possibly) the points in the compact set i<n Ki ⊆ Y .

For (4): Z is compact Hausdorff and of countable pseudocharacter.

For (5): Z is compact Hausdorff and totally disconnected.

For (6): By Lemma 2.2, it is sufficient to prove that each Z n is HS . But Z n

is a finite union of subspaces of the form X j × Y k , which are HS by Lemma 2.1.

2.3 Homogeneity

The following was proved by Dow and Pearl [9]:

 15

Theorem 2.7 If Z is first-countable and zero-dimensional, then Z ω is homoge-

neous.

Actually, we only need here the special case of this result where Z is compact

and has a dense set of isolated points; this was announced (without proof) earlier

by Motorov [24].

Recall that any compact HS space must have countable π-weight (see part 3

of Theorem 1.11), so if the space is also homogeneous, it must have size at most

2ℵ0 by Theorem 1.8. Under CH this implies (see part 5 of Theorem 1.11) that the

space must be first-countable.

Theorem 2.8 (CH ) There exists a (necessarily first-countable) zero-dimensional

compact homogeneous strong S-space.

Proof. Let X be the Cantor set 2ω with its usual topology, let Y be 2ω with the

topology constructed in Theorem 2.4, let ϕ be the identity, and let Z = Y ∪˙ ϕ X. By

Lemma 2.6, Z, and hence also Z ω , are zero-dimensional first-countable compact

strong S-spaces; Z ω is homogeneous by Theorem 2.7. ♠

 16

Chapter 3

Homogeneity Properties in Large

Compact S-spaces

In this chapter we introduce a technique for constructing (under ♦) compact zero-

ℵ
dimensional S-spaces which have the biggest possible size, namely 22 0 , while keep-

ing some control over its group of homeomorphisms. Furthermore, our spaces will

have no isolated points and no convergent sequences.

We use an inverse limit approach similar to the one used in [10], where ♦ is

used to capture (and eventually destroy) all potential left separated ω1 -sequences,

making the space HS . In order to keep control over the autohomeomorphisms of

the space, we need to do two things: we need to destroy some potential autohome-

omorphisms (for which we use ♦ again), but we also want to preserve some of

them, for which we use a “lifting” technique used in [20] for getting certain kind of

L-spaces. In Section 3.2 we introduce this lifting technique, which will be used at

the successor stages of the inverse limit. The main construction is done in Section

3.3 and three applications are given in Section 3.4.

 17

3.1 Some technical facts and terminology

Given a collection S of subsets of a space X, we say that x ∈ X is a strong limit

point of S if and only if for all neighborhoods U of x there is an A ∈ S such that

/ A. If f : Y → X, we write f −1 (S) for the collection {f −1 (A) :

A ⊆ U and x ∈

A ∈ S}. If A ⊆ X I and J ⊆ I we write A↾J for the set {x↾J : x ∈ A}. We will

write Hom(X) for the group of homeomorphisms from X onto X. Throughout

this chapter we will denote the Cantor space 2ω by C. If K is a clopen subset of

C ω1 and α ∈ ω1 , we say that the support of K is contained in α if and only if K↾α

is clopen in C α .

3.1.1 Closed subspaces of C ω1 as inverse limits

Throughout this section, X is a fixed closed subspace of C ω1 . For α ≤ β ≤ ω1 we

write Xα = X↾α and we let παβ : Xβ → Xα be the obvious projection (we will often

drop the superscript when β = ω1 ). Note that each Xα is closed in C α and that

παβ (Xβ ) = Xα for α ≤ β ≤ ω1 . Moreover it is easy to see that if γ is a limit ordinal

and x ∈ C γ then x ∈ Xγ if and only if x↾α ∈ Xα for all α ∈ γ. In particular we

can always recover X from the sequence hXα : α ∈ ω1 i. Note that if each Xα is

perfect (and hence homeomorphic to C) then X is also perfect.

Definition 3.1 Given α ≤ β ≤ ω1 and S ⊆ P(Xα ), we say that Xβ preserves S

if whenever x ∈ Xα is a strong limit point of S, we have that every y ∈ (παβ )−1 (x)

is a strong limit point of (παβ )−1 (S).

Note that if S ⊆ P(Xα ) and η ≥ β ≥ α, then Xη preserves S if and only if Xβ

preserves S and Xη preserves (παβ )−1 (S). Also, if γ ≥ α is a limit ordinal, then Xγ
 18

preserves S if and only if Xξ preserves S for all α ≤ ξ < γ.

The following two lemmas were essentially proved in [10] and they (together

with Lemma 1.15) give a hint as to why ♦ will be relevant for our construction.

Lemma 3.2 Suppose that hxξ : ξ ∈ ω1 i is a sequence of distinct points in X. Then

there is a club Q ⊆ ω1 such that for all γ ∈ Q,

1. The xξ ↾γ for ξ ∈ γ are all distinct.

2. If η ≥ γ, then xη ↾γ is a limit point of hxξ ↾γ : ξ ∈ γi.

Lemma 3.3 Let hxξ : ξ ∈ ω1 i be a sequence of distinct points in X and suppose

that there is a stationary S ⊆ ω1 such that for each γ ∈ S, the collection {{xξ ↾γ} :

ξ ∈ γ} is preserved by X. Then hxξ : ξ ∈ ω1 i is not a left separated sequence.

The following should be clear:

Lemma 3.4 Let hxn : n ∈ ωi be a sequence of points in X. Suppose that there is

an α ∈ ω1 such that {{xn ↾α} : n ∈ ω} is preserved by X and has a strong limit

point x ∈ Xα such that |(πα )−1 (x)| > 1. Then hxn : n ∈ ωi is not a convergent

sequence.

If h is a homeomorphism from X onto X, then there is a club Qh ⊆ ω1 and

homeomorphisms hα from Xα onto Xα such that whenever α ∈ Qh , we have that

πα (h(x)) = hα (πα (x)) for all x ∈ X. Moreover, we have the following:

Lemma 3.5 Suppose that g and h are homeomorphisms from X onto X. Suppose

that X has a π-base every element of which is clopen and has its support contained

in α ∈ Qg ∩ Qh and gα = hα . Then g = h.
 19

Proof. If g 6= h, then there is a clopen U ⊆ X such that g(U) ∩ h(U) = ∅. By

hypothesis, there is a clopen V ⊆ U with support contained in α. This is already

a contradiction because gα = hα implies that g(V ) = h(V ). ♠

We finish this section with a result for which the proof is just a matter of

chasing the definitions.

Lemma 3.6 Suppose that h is a homeomorphisms from X onto X, α ∈ Qh , and

S ⊆ P(Xα ). Then X preserves S if and only if X preserves hα (S).

3.1.2 The Cantor space and group actions

In this chapter we will write K(C) for the collection of non-empty closed subsets

of C. Preserving (in the sense of Definition 3.1) strong limit points of countable

subsets of K(C) will play a key role in our construction. The next result will be

useful to handle the successor stages, but first we need some definitions.

Definition 3.7 Given S ⊆ K(C) and closed A0 , A1 ⊆ C, we say that (A0 , A1 )

preserves S if whenever x ∈ A0 ∩ A1 is a strong limit point of S and U is an open

neighborhood of x, there are K0 , K1 ∈ S such that K0 ⊆ U \ A1 and K1 ⊆ U \ A0 .

Definition 3.8 Given S ⊆ P(K(C)) and a convergent sequence B in C with limit

point b, we say that B avoids S, if whenever b is a strong limit point of S ∈ S and

U is an open neighborhood of b, there is a K ∈ S such that K ⊆ U \ (B ∪ {b}).

Note in particular that any constant sequence avoids any set S and hence

satisfies the hypothesis of the next lemma.

 20

Lemma 3.9 Suppose that S ⊆ [K(C)]ℵ0 is countable and that B is a convergent

sequence in C which avoids S. Let b ∈ C be the limit point of B. Then there are

closed A0 , A1 ⊆ C such that:

1. A0 ∪ A1 = C, A0 ∩ A1 = {b} and B ⊆ A0 .

2. (A0 , A1 ) preserves every S ∈ S.

Proof. The case in which B is a constant sequence (and hence essentially a

point) was proved in [20, proof of Theorem 0.1]. Thus we can assume here that

B = {bn : n ∈ ω} where the bn ’s are all distinct and all different from b. We

also fix an enumeration S = {{Sni : n ∈ ω} : i ∈ ω} and a decreasing sequence

T
hVm : m ∈ ωi of clopen subsets of C such that V0 = C and m∈ω Vm = B ∪ {b}.

We will construct inductively a decreasing sequence hWm : m ∈ ωi of clopen

subsets of C such that W0 = C, Wm ⊆ Vm and bm ∈ W2m \ W2m+1 for all m ∈ ω.

T S
Note that m∈ω Wm = {b} and that if we let A0 = {b} ∪ m∈ω (W2m \ W2m+1 ) and
S
A1 = {b} ∪ m∈ω (W2m+1 \ W2m+2 ), then condition (1) will be satisfied. Now we

explain how to construct the sequence in order to satisfy condition (2) as well.

Fix a local base {Oj : j ∈ ω} for b in C and a function (ψ, µ) : ω → ω × ω

such that (ψ, µ)−1(i, j) contains at least one odd and one even natural number

for each (i, j) ∈ ω × ω. If W2m is already defined and b is a strong limit point
ψ(2m)
of hSn : n ∈ ωi then (since B avoids S) there is an n = n(2m) ∈ ω such
ψ(2m) ψ(2m)
that Sn ⊆ W2m ∩ Oµ(2m) and Sn ∩ (B ∪ {b}) = ∅. Thus there is a k ∈ ω
ψ(2m)
and a clopen neighborhood U of bm such that Sn ⊆ W2m \ (Vk ∪ U) and

U ∩ B = {bm }; then we let W2m+1 = (W2m ∩ Vk ) \ U. If b wasn’t a strong limit

ψ(2m)
point of hSn : n ∈ ωi then we just let W2m+1 = (W2m ∩ V2m+1 ) \ U where
 21

U is any clopen with U ∩ B = {bm }. Note that in either case we have that

{bk : k > m} ⊆ W2m+1 ⊆ V2m+1 ∩ W2m and bm ∈

/ W2m+1 . If W2m+1 is already

defined we follow the same steps (now working with S ψ(2m+1) and Oµ(2m+1) ) except

that now we want to keep bm+1 inside W2m+2 , so we just let W2m+2 = W2m+1 ∩ Vk

in the first case and W2m+2 = W2m+1 ∩ V2m+2 in the second case.

To show that condition (2) is satisfied, suppose that b is a strong limit point

of hSni : n ∈ ωi and O is an open neighborhood of b. Then there are j, m0 , m1 ∈ ω

such that b ∈ Oj ⊆ O and (ψ, µ)(2m0 ) = (ψ, µ)(2m1 + 1) = (i, j). But now it is
i i
clear from the construction that Sn(2m0)
⊆ O \ A1 and Sn(2m1 +1)
⊆ O \ A0 , showing

that (A0 , A1 ) preserves {Sni : n ∈ ω}. ♠

Remember that a group G acts freely (or the action is free) on a set X if

g(x) 6= x whenever x ∈ X and g ∈ G \ {id}. If H is a subgroup of G and g ∈ G,

we use hH, gi to denote the subgroup of G generated by H and g.

The following is Lemma 6.4 in [20].

Lemma 3.10 Suppose G is a countable subgroup of Hom(C) acting freely on C

and suppose a, b ∈ C are distinct. Then there is an h ∈ Hom(C) such that h(a) = b

and hG, hi acts freely on C.

Almost the same proof as the one found in [20] works for showing the following:

Lemma 3.11 Suppose G is a countable subgroup of Hom(C) acting freely on C

and suppose K and L are disjoint non-empty clopen subsets of C. Then there is

an h ∈ Hom(C) such that h(K) = L and hG, hi acts freely on C.

Throughout the rest of this section, X is a fixed Hausdorff space and G is a

group acting on X.
 22

Definition 3.12 We say that f ∈ Hom(X) is locally in G if for every x ∈ X

there is an open neighborhood U of x and a g ∈ G such that f (u) = g(u) for all

u ∈ U. We write H(X, G) for the set of autohomeomorphisms of X which are

locally in G.

Note that if X is compact and zero-dimensional then f is locally in G if and

only if there is a finite partition {Pi : i ∈ I} of X into clopen subsets and there is

a finite {gi : i ∈ I} ⊆ G such that f ↾Pi = gi for all i ∈ I. Also note that H(X, G)

is actually a subgroup of Hom(X) and that H(X, H(X, G)) = H(X, G).

Definition 3.13 We say that two subsets A, B ⊆ X are G-equivalent if there is

an f ∈ H(X, G) such that f (A) = B.

We finish this section with a lemma that makes the task of verifying that a

space is B-homogeneous somewhat simpler.

Lemma 3.14 Suppose that U ⊆ X is clopen and let B be the set of all clopen
S
subsets of X which are G-equivalent to U. If B is a π-base for X and X = B

then B is a base for X.

Proof. Fix a clopen V ⊆ X and x ∈ V . We want to find an element of B

contained in V and containing x. Without loss of generality there is a B ∈ B such

S
that B ⊆ X \ V . Since X = B, there is an f ∈ H(X, G) such that x ∈ f B, and

we can fix W ⊆ B clopen such that x ∈ f (W ) ⊂ V . Since B is a π-base, there

is an f ′ ∈ H(X, G) such that f ′ (B) ⊆ V \ f (W ). Now we define h ∈ H(X, G)

as follows: if x ∈ W we let h(x) = f (x), if x ∈ B \ W we let h(x) = f ′ (x), if

x ∈ f (W ) we let h(x) = f −1 (x), if x ∈ f ′ (B \ W ) we let h(x) = f ′−1 (x) and we let

 23

h(x) = x otherwise. Then clearly h ∈ H(X, G) and x ∈ hB ⊆ V . Thus hB is the

desired element of B. ♠

3.2 Lifting homeomorphisms

In this section we present a construction used in [20] for getting certain kind of

L-spaces. This construction will be the key for handling the successor stages of

our main construction.

Throughout this section, G is a countable group acting freely on C. Given two

closed A0 , A1 ⊆ C and p ∈ C such that A0 ∩ A1 = {p} and A0 ∪ A1 = C, we let

Y = Y (A0 , A1 , G) be the space

Y = (x, ϕ) ∈ C × 2G : (∀g ∈ G) x ∈ gAϕ(g) .



We let G act on 2G by (gϕ)(h) = ϕ(g −1h) where g and h denote elements of G

while ϕ and gϕ are in 2G . Furthermore G acts on Y by g(x, ϕ) = (gx, gϕ).

For any choice of Ai ’s we have that Y is a closed subspace of C × 2G and the

action described above is free and continuous. We also have that the projection

into the first coordinate π : Y → C is surjective.

Given two disjoint clopen P0 , P1 ⊆ Y and x ∈ C, we will say that {P0 , P1 } splits

π −1 (x) if both π −1 (x) ∩ P0 and π −1 (x) ∩ P1 are non-empty. Since A0 ∩ A1 = {p}

and the action of G is free, it is easy to see that |π −1(x)| ≤ 2 for all x ∈ C. This

fact is the key for the following:

Lemma 3.15 Suppose that P is a finite partition of Y into clopen subsets. Then

there is a finite partition Q of C into clopen subsets and a function Ψ : Q → [P]≤2

 24

such that:

1. π −1 Q ⊆
S
Ψ(Q) for all Q ∈ Q and

2. Whenever |Ψ(Q)| = 2, there is a unique x ∈ Q such that Ψ(Q) splits π −1 (x).

Proof. For each clopen Q ⊆ C, let Ψ(Q) = {P ∈ P : P ∩ π −1 Q 6= ∅}. Given

x ∈ C, since |π −1 (x)| ≤ 2, there is a clopen neighborhood Qx of x such that

|Ψ(Q)| ≤ 2. Now using compactness of C and taking boolean combinations we

can get a finite partition Q′ of C with |Ψ(Q)| ≤ 2 for all Q ∈ Q′ . Now suppose

that Q ∈ Q′ is such that |Ψ(Q)| = 2. Since the two clopens in Ψ(Q) only mention

finitely many elements of G and Ψ(Q) splits π −1 (x) only if (maybe not even if)

x = gp for some g mentioned, we can further partition Q to isolate the finitely

many such possible x’s. We then let Q be the corresponding refinement of Q′ . ♠

Note that by the same argument, we can in fact get Q in the previous lemma so

that it refines any finite partition of C given in advance. This allows us to prove the

next result which says that non-G-equivalent clopens of C remain non-G-equivalent

in Y .

Lemma 3.16 Suppose that U, V are disjoint clopen subsets of C such that π −1 U

is G-equivalent to π −1 V . Then U and V are G-equivalent.

Proof. Fix f ∈ H(Y, G) such that f (π −1 U) = π −1 V . Without loss of general-

ity we may assume that f 2 = id and that f ↾(Y \ π −1 (U ∪ V )) = id. Fix a finite

partition P of Y into clopen sets so that f acts as a different element of G in each

element of the partition. Let Q be a partition of C refining {U, V, C \ (U ∪ V )} as

in the previous lemma with the corresponding Ψ : Q → [P]≤2 .

 25

Let S = {x ∈ C : π −1 (x) is split by P}. Since |S ∩ Q| ≤ 1 for each Q ∈ Q,

we have that S is finite. On the other hand, since the action of G in Y commutes

with π, it is easy to see that f (π −1S) = π −1 S, and hence |S ∩ V | = |S ∩ U|. Then

we can write S ∩ U = {gip : i ∈ n} and S ∩ V = {hj p : j ∈ n} for some n ∈ ω and

some gi , hj ’s in G.

Now fix W a clopen neighborhood of p such that each gi W and each hj W is

S S
contained in an element of Q. Let WU = i∈n giW and WV = i∈n hi W . We

claim that f (π −1 WU ) = π −1 WV .

Suppose that (x, ϕ) ∈ π −1 WU and fix i ∈ n and Q ∈ Q such that x ∈ gi W ⊆ Q.

Let ψ ∈ 2G such that (gi p, ψ) and (x, ϕ) belong to the same element P of Ψ(Q),

let g ∈ G such that f ↾P = g↾P , and fix j ∈ n such that π(f (gip, ψ)) = hj p. Thus

ggip = hj p and since the action of G is free we get that ggi = hj and therefore

π(f (x, ϕ)) = π(gx, gϕ) = gx = hj gi−1x ∈ hj W . This shows that f (x, ϕ) ∈ π −1 WV ,

hence proving that f (π −1 WU ) ⊆ π −1 WV . By a similar argument it follows that

f (π −1 WV ) ⊆ π −1 WU and since f 2 = id we get f (π −1 WU ) = π −1 WV .

Finally we apply the last lemma once more to get a partition R of C \(WU ∪WV )

such that for each R ∈ R there is a gR ∈ G with f ↾π −1 R = gR ↾π −1 R. Now we

can define h : C → C as follows: if x ∈ R ∈ R let h(x) = gR (x), if x ∈ gi W

let h(x) = hi gi−1(x) and if x ∈ hj W let h(x) = gj h−1

j (x). It is clear that h is an

autohomeomorphism of C, h is locally in G and hU = V . ♠

Definition 3.17 Suppose (x, ϕ) ∈ Y and |π −1 (x)| = 2. Then x = gp for some

unique g ∈ G. We say that (x, ϕ) is of 0-type if ϕ(g) = 0 and (x, ϕ) is of 1-type

if ϕ(g) = 1.
 26

The next result shows that not all the clopens of Y are G-equivalent.

Lemma 3.18 Let K = {(x, ϕ) ∈ Y : ϕ(e) = 0} where e ∈ G is the identity

element. Then K and Y \ K are not G-equivalent.

Proof. By contradiction, fix a locally in G autohomeomorphism f of Y such

that f K = Y \ K. Without loss of generality we may assume that f 2 = id. Let

S be the set of all x ∈ C such that π −1 (x) has two points and f is defined by

different elements of G at those two points. Then f (π −1 S) = π −1 S and in fact

f (π −1 S ∩ K) = π −1 S ∩ (Y \ K). Note that p ∈ S and that for any x ∈ S with

x 6= p we either have π −1 (x) ⊆ π −1 S ∩ K or π −1 (x) ⊆ π −1 S ∩ (Y \ K). Thus

π −1 S ∩ K has n + 1 points of 0-type and n points of 1-type while π −1 S ∩ (Y \ K)

has n points of 0-type and n + 1 points of 1-type, where |S| = 2n + 1. But this is

impossible since the action of G preserves the type of the points. ♠

We finish this section with a technical lemma that shows us how to preserve

(in the sense of Definition 3.1) countable collections of closed subsets of C at the

successor stages.

Lemma 3.19 Suppose that S ∈ [K(C)]ℵ0 and that (A0 , A1 ) preserves gS for each

g ∈ G. Suppose also that x ∈ C is a strong limit point of S. Then every point in

π −1 (x) is a strong limit point of π −1 S.

Proof. Fix (a, ϕ) ∈ Y such that a is a strong limit point of S, and fix U × W

a basic open neighborhood of (a, ϕ). Thus we can write W = {ψ ∈ 2G : σ ⊆ ψ}

for some finite σ ⊆ ϕ. If g ∈ dom(σ) and a 6= gp then a ∈

/ gA1−σ(g) so there is an
 27

T
open Vg with a ∈ Vg ⊆ U and Vg ∩ gA1−σ(g) = ∅. Let V = {Vg : g ∈ dom(σ) and

a 6= gp}. Now we consider two cases:

i) If a = gp for some g ∈ dom(σ) (since G acts freely, there is at most one such

g) then p = g −1a ∈ g −1 V and p is a strong limit point of g −1S. Therefore, since

(A0 , A1 ) preserves g −1S, there is a K ∈ S such that g −1 K ⊆ g −1V \ A1−σ(g) . Thus

K ⊆ V \ gA1−σ(g) .

ii) if a 6= gp for all g ∈ dom(σ) then we just use the fact that a is a strong limit

point of S to get a K ∈ S such that K ⊆ V \ {a}.

In either case we get that π −1 K ⊆ (U × W ) \ {(a, ϕ)}, showing that (a, ϕ) is a

strong limit point of π −1 S. ♠

3.3 Main construction

We construct simultaneously three sequences hXα : α ∈ ω1 i, hGα : α ∈ ω1 i and

hSα+1 : α ∈ ω1 i. Each Xα is a closed subspace of C α , and Gα is a countable group

acting freely on Xα . Following the notation in Section 3.1.1, we will also have the

natural projections παβ : Xβ → Xα for α ≤ β. Then we will let X ⊆ C ω1 be the

S
inverse limit of the sequence < Xα : α ∈ ω1 > and let G = α∈ω1 Gα act on X in

the natural way. At the end, we want the following:

i) X is compact and has no isolated points.

ii) X is hereditarily separable and has no non-trivial convergent sequences.

iii) Hom(X) = H(X, G)

 28

The following requirements completely determine the construction at limit ordi-

nals. They also ensure that condition (i) will be satisfied. For all 0 < α ≤ β < ω1 ,

we will require that:

R1) Xα is closed in C α and has no isolated points.

R2) παβ (Xβ ) = Xα .

R3) Gα acts freely on Xα .

R4) Gα ⊆ Gβ and παβ (g(x)) = g(παβ (x)) for all x ∈ Xβ and g ∈ Gα .

S
R5) If α is a limit ordinal then Gα = ξ∈α Gξ .

In order to satisfy conditions (ii) and (iii), we need to make sure that X pre-

serves strong limit points of certain sequences of closed subsets of the Xα ’s. This

is the role of our third sequence hSα+1 : α ∈ ω1 i. For all 0 < α ≤ β < ω1 , we will

require that (here, K(Xα ) denotes the set of all non-empty closed subsets of Xα ):

R6) Sα+1 ⊆ [K(Xα )]ℵ0 is countable.

R7) Xα+1 preserves each S ∈ Sα+1 .

R8) If S ∈ Sα+1 then (παβ )−1 (S) ∈ Sβ+1 .

Note that the last two requirements actually guarantee that X preserves each

S in every Sα+1 . Our notation might seem a bit artificial, but the idea is simple:

Sα+1 is the collection of sequences in Xα that we want to preserve, but essentially

we want Xα+1 to preserve them. For limit ordinals the preservation is automatic

(see the comment after Definition 3.1), so we don’t need to have Sα defined when

α is a limit ordinal. To simplify notation, given S ⊆ [K(Xα )]ℵ0 we will write

(παβ )−1 (S) for the set {(παβ )−1 (S) : S ∈ S}. So for instance, requirement (R8) just

says (παβ )−1 (Sα+1 ) ⊆ Sβ+1 .

 29

The following requirement ensures that the space X will have a π-base, every

element of which is of the form (π1 )−1 (U) for some U clopen subset of X1 . This

is not absolutely necessary for our construction, but it simplifies some of the ar-

guments. In particular it automatically implies that all the Xα ’s have no isolated

points. We fix a countable base W of clopen subsets of X1 , and require that:

R9) W ∈ S2 .


Using ♦ we now fix a set {zξα : α ∈ ω1 , ξ ∈ α} such that each zξα : ξ ∈ α

is an α-sequence in C α and for any ω1 -sequence hxξ : ξ ∈ ω1 i in C ω1 , the set


α ∈ ω1 : (∀ξ ∈ α)(xξ ↾α = zξα ) is stationary. We now make two more require-

ments for each α ∈ ω1 :


R10) If {zξα : ξ ∈ α} ⊆ Xα then {zξα } : ξ ∈ α ∈ Sα+1 .

R11) If hznα : n ∈ ωi is a convergent sequence in Xα with limit point x, then

{{znα } : n ∈ ω} ∈ Sα+1 and |(παα+1 )−1 (x)| > 1.

In view of Lemmas 3.3 and 3.4, it is clear now that our requirements so far guar-

antee that X will be hereditarily separable and will have no convergent sequences.

That is, condition (ii) is satisfied. We now turn to condition (iii).

Since X will be separable, any homeomorphism from X onto X is determined

by its restriction to a countable subset of X. This allows us to use our ♦-sequence

to capture and destroy if necessary these homeomorphisms.

We let Ω be the set of all α ∈ ω1 such that {znα : n ∈ ω} is a dense subset of

Xα and for which there is a (necessarily unique) homeomorphism f from Xα onto

Xα such that f (znα ) = zω+n

α
for all n ∈ ω, and f ∈
/ H(Xα , Gα ). Formally, the set Ω
 30

is determined only at the end of the construction; however, after Xα and Gα have

been constructed, we can already decide whether α is in Ω or not.

Now, for each α ∈ Ω, we will also construct two sequences haαn : n ∈ ωi and

hbαn : n ∈ ωi in Xα converging to aα and bα respectively, satisfying the following

requirements:

R12) Xα+1 does not preserve {{bαn } : n ∈ ω}.

R13) {{aαn } : n ∈ ω} ∈ Sα+1 .

R14) f (aαn ) = bαn for all n ∈ ω.

These three requirements now guarantee that condition (iii) will be satisfied.

To see this, fix h ∈ Hom(X). Using the fact that X is separable, we can fix a

dense {xn : n ∈ ω} ⊆ X. Then there is an α ∈ Qh such that znα = xn ↾α and

α
zω+n = h(xn )↾α for all n ∈ ω. Note that, in view of Lemma 3.6, h witnesses the

fact that f = hα did not get destroyed at stage α (i.e. requirements (R12)-(R14)

were not met at stage α). This means that α ∈

/ Ω and therefore f ∈ H(Xα , Gα ).

But then by Lemma 3.5, we get that h ∈ H(X, Gα ) and hence h ∈ H(X, G).

We finish this section showing how to construct the sequences hXα : α ∈ ω1 i,

hGα : α ∈ ω1 i and hSα+1 : α ∈ ω1 i, in such a way that all the requirements are met.

We start by putting X1 = C and we can let G1 be any countable group acting

freely on C. For a limit ordinal α all the requirements are automatically satisfied
S
by letting Xα be the inverse limit of the previous Xξ ’s, and letting Gα := ξ∈α Gξ

act on Xα in the natural way. Note that we don’t need to construct Sα .

For the successor stage, we fix α ∈ ω1 and assume Xξ and Gξ have been

constructed for all ξ ≤ α. We also assume that Sξ+1 has been constructed for all
 31

ξ < α. Now we construct Xα+1 , Gα+1 and Sα+1 .

We collect in S ′′′ all the (πξα )−1 (Sξ+1 ) for ξ < α; if the hypothesis of (R10) is

satisfied, we also include {zξα } : ξ ∈ α in S ′′′ and if the hypothesis of (R11) is

satisfied we include {{znα } : n ∈ ω} as well. Then we let S ′′ be the closure of S ′′′

under Gα . What we do next depends on whether α ∈ Ω or not.

Suppose first that α ∈ Ω and let f : Xα → Xα be as in the definition of Ω. Fix

aα ∈ Xα a witness of the fact that f ∈

/ H(Xα , Gα ) (i.e. such that there is no clopen

neighborhood of aα on which f acts as an element of Gα ) and let bα = f (aα ). If

there is a g0 ∈ Gα such that g0 (aα ) = bα then find a sequence hOn : n ∈ ωi of

clopen subsets of Xα converging to aα such that g0 (On ) ∩ f (On ) = ∅ for all n ∈ ω.

Otherwise (if there is no such a g0 ) choose any sequence hOn : n ∈ ωi of clopen

subsets of Xα converging to aα .

Let S ′ = S ′′ ∪ {{f (On ) : n ∈ ω}}. Look at {bα } as a constant sequence and

apply Lemma 3.9 with S ′ to get corresponding closed A′0 and A′1 subsets of Xα

such that (A′0 , A′1 ) preserves each sequence in S ′ . Note that by requirement (R1)

we have that Xα ∼
= C so we can indeed use Lemma 3.9.

In particular, (A′0 , A′1 ) preserves {f (On ) : n ∈ ω} and therefore there is a

subsequence hf (Oni ) : i ∈ ωi converging to bα such that f (Oni ) ⊆ A′0 for all i ∈ ω.

Now we choose bαi ∈ f (Oni ) and let aαi = f −1 (bαi ) for i ∈ ω.

The reason for choosing the bαi ’s in this way is that the sequence {bαi : i ∈ ω}

avoids the collection Sα+1 := S ′ ∪ {{{g(aαn)} : n ∈ ω} : g ∈ Gα }. Thus we can use

again Lemma 3.9 to get closed A0 , A1 ⊆ Xα with A0 ∪ A1 = Xα , A0 ∩ A1 = {bα },

{bαi : i ∈ ω} ⊆ A0 and (A0 , A1 ) preserving each element of Sα+1 .

 32

Suppose now that the hypothesis of (R11) are satisfied (in particular α ∈
/ Ω).

Then we just apply Lemma 3.9 directly to the set S ′′ to get closed A0 , A1 ⊆ Xα

such that A0 ∪ A1 = Xα , A0 ∩ A1 = {x} and (A0 , A1 ) preserves each element of

Sα+1 := S ′′ . If the hypothesis of (R11) are not satisfied and α ∈

/ Ω, we do the same

with an arbitrary x ∈ Xα .

We can now identify Xα with C and construct Y (A0 , A1 , Gα ) ⊆ Xα × 2Gα as in

Section 3.2. Then identifying 2Gα with C we get Xα+1 ⊆ Xα × C ⊆ C α+1 . Then

Gα acts on Xα+1 and we are free to choose any Gα+1 acting freely on Xα+1 for

which requirement (R4) is satisfied (with β = α + 1).

It is clear from the construction that all the requirements are preserved. For

(R7) just use Lemma 3.19 and for (R12) use the fact that {bαi : i ∈ ω} ⊆ A0 .

3.4 Applications

We present here three applications of our main construction. Recall that we define

a Fedorchuk space as a compact hereditarily separable perfect space with no non-

trivial convergent sequences.

Theorem 3.20 (♦) There is a rigid Fedorchuk space.

Proof. In the main construction just let Gα = 1 for all α ≤ ω1 . ♠

Theorem 3.21 (♦) There is an h-homogeneous Fedorchuk space.

Proof. Let {(Kξ , Lξ ) : ξ ∈ ω1 } be an enumeration of all pairs of disjoint non-

empty clopen subsets of C ω1 . In the main construction let G1 = 1 and at stage α do

 33

the following: let ξ0 be the smallest ξ not considered at a previous stage for which

both Kξ and Lξ have their supports contained in α. Let K = Xα+1 ∩ Kξ0 ↾(α + 1)

and L = Xα+1 ∩ Lξ0 ↾(α + 1). If both K and L are non-empty (and hence clopen in

Xα+1 ), apply Lemma 3.11 to get Gα+1 = hGα , gi acting freely on Xα+1 and with

g(K) = L. This clearly ensures that X is h-homogeneous. ♠

Theorem 3.22 (♦) There is a zero-dimensional Fedorchuk space which is B-

homogeneous but not h-homogeneous.

Proof. Start with G1 acting freely on C such that any two disjoint clopen

subsets are G1 -equivalent (for example one could use Lemma 3.11 ℵ0 -many times).

Then keep Gα = G1 for all α ≤ ω1 , in particular G = G1 . By Lemma 3.18,

X2 already has two disjoint non-G-equivalent clopen subsets and by Lemma 3.16

these remain non-G-equivalent (and hence non homeomorphic to each other) in X.

However by Lemma 3.14 and requirement (R9), X is B-homogeneous. ♠

 34

Chapter 4

Cardinality of Homogeneous

Compacta

The goal of this chapter is to give a proof in ZFC of the fact that |X| ≤ 2ℵ0

for compact homogeneous spaces of countable tightness. In fact we will prove the

more general inequality |X| ≤ 2t(X)L(X)h(X) for any regular homogeneous space.

Our main tool will be the “Elementary Submodel technique”: Given a topo-

logical space (X, τ ), one let M be an elementary submodel of H(θ) (the set of all

sets of hereditary cardinality less than θ) for a “large enough” regular cardinal θ.

Usually one asks for M to be “small” and to contain X and τ as elements. Then

one uses closure properties of M to get results about X ∩ M, τ ∩ M and ultimately

about (X, τ ). A model M is said to be κ-closed if any κ-sequence of elements of M

is in M (i.e. M κ ⊆ M). For more details and a good introduction to the technique

see [8]. Let us just say that in each specific application, one takes θ large enough

for H(θ) to contain all sets of interest in the context under discussion. In this

sense we will just say that M ≺ V. In Section 4.1 we prove some basic facts in

the context of elementary submodels of compact spaces with t(X)L(X) ≤ κ; we

also give answer (Theorem 4.7) to a question of L.R. Junqueira and F. Tall.
 35

4.1 Elementary submodels

We start by fixing an infinite cardinal κ, a T3 topological space (X, τ ) such that

L(X)t(X) ≤ κ, and a κ-closed M ≺ V with X, τ ∈ M. We let Z = cl(X ∩M) ⊆ X

with the subspace topology.

One of the goals of this section is to show that if X is compact, then Z is a

retract of X. The following result suggests what the retraction is going to be.

Lemma 4.1 For every x ∈ X there is a qx ∈ Z such that for all U ∈ τ ∩ M either

qx ∈
/ U or x ∈ U.

Proof. Fix x ∈ X and assume there is not such a qx . Then for each q ∈ Z we

can fix a Uq ∈ τ ∩ M such that q ∈ Uq and x ∈

/ Uq . Since L(Z) ≤ κ we get that

Z ⊆ U := q∈Q Uq for some Q ∈ [Z]≤κ . Since M is κ-closed we have that U ∈ M

S

/ U we get by elementarity that there is an x′ ∈ (X ∩ M) \ U; but

and since x ∈

this is impossible since Z ⊆ U. ♠

Just by elementarity and the fact that X is Hausdorff, it is immediate that

τ ∩ M separates points in X ∩ M. We prove now that in fact τ ∩ M separates

points in Z.

Lemma 4.2 If p0 , p1 ∈ Z and p0 6= p1 then there are U0 , U1 ∈ τ ∩ M such that

p0 ∈ U0 , p1 ∈ U1 and U0 ∩ U1 = ∅.

Proof. Given p0 , p1 ∈ Z with p0 6= p1 we use regularity of X to fix Ui′ , Vi ∈ τ

for i ∈ 2 such that pi ∈ Vi , cl(Vi ) ⊆ Ui′ and U0′ ∩ U1′ = ∅. Since t(X) ≤ κ, there

are A0 , A1 ⊆ M such that |Ai | ≤ κ, Ai ⊆ Vi and pi ∈ cl(Ai ) for i ∈ 2. Since M

 36

is κ-closed we have that A0 and A1 are both in M and therefore by elementarity

we have that M |= ∃U0 , U1 ∈ τ [ cl(A0 ) ⊆ U0 ∧ cl(A1 ) ⊆ U1 ∧ U0 ∩ U1 = ∅ ]. Then,

again by elementarity, any U0 , U1 ∈ τ ∩ M witnessing this fact will satisfy the

conclusion of the lemma. ♠

Corollary 4.3 For every x ∈ X there is a unique qx ∈ Z such that for each

U ∈ τ ∩ M either qx ∈
/ U or x ∈ U.

Proof. Immediate from Lemmas 4.1 and 4.2. ♠

Definition 4.4 In view of the last result, we define the function rM : X → Z by

rM (x) = qx for x ∈ X.

By the way in which it was defined, it is very natural to think that rM should

be continuous. However, we were only able to prove this for the compact case.

Lemma 4.5 If X is compact, then rM is continuous.

−1
Proof. Fix W ∈ τ (not necessarily in M) with W ∩ Z 6= ∅, fix x ∈ rM (W ) and
−1
let q = rM (x). We need to show that there is a V ∈ τ such that x ∈ V ⊆ rM (W ).

For each p ∈ Z \ W we use Lemma 4.2 to get Up , Vp ∈ τ ∩ M such that p ∈ Up ,

q ∈ Vp and Up ∩ Vp = ∅. Since Z \ W is closed (and hence compact), we get that

S T
Z \ W ⊆ U := p∈P Up for some finite P ⊆ Z \ W . Clearly q ∈ V := p∈P Vp and

U ∩ V = ∅. Also by elementarity we have that U, V ∈ M.

Since q ∈ V ∈ M we get (from the definition of rM ) that x ∈ V . To see that

−1
V ⊆ rM (W ), fix y ∈ V and note that if rM (y) was in Z \ W then it would be in U
 37

and since U ∈ M we would get that y ∈ U; but this is impossible since U ∩ V = ∅

and hence rM (y) ∈ W . ♠

It is clear that rM (x) = x for all x ∈ Z. So we actually have that rM is a

retraction whenever X is compact. Looking closer at the last proof, we see that

q ∈ V ∩ Z ⊆ W ∩ Z, so we also showed the following:

Corollary 4.6 If X is compact then the set {U ∩ Z : U ∈ τ ∩ M} is a base for

the topology of Z as a subspace of X.

In [18], Junqueira and Tall define the space XM as the set X ∩ M with the

topology τM generated by {U ∩ M : U ∈ τ ∩ M}. They ask (Problem 7.22) if there

is a consistent example of a compact T2 space X with countable tightness and a

countably closed M such that XM is not normal. A consequence of Corollary 4.6

is that XM is a subspace of X; this was already proved in [18] (Theorem 2.11).

In the next theorem we make use of this fact to give a negative answer to their

question.

Theorem 4.7 If (X, τ ) is a compact space with t(X) ≤ κ and M ≺ V is κ-closed

then X ∩ M (= XM ) is normal.

Proof. Fix two disjoint closed E, F ⊆ X ∩ M. We claim that cl(E) and cl(F )

are still disjoint. Therefore, since cl(X ∩ M) is compact (and hence normal), cl(E)

and cl(F ) (and hence E and F ) can be separated.

Now suppose (seeking a contradiction) that p ∈ cl(E) ∩ cl(F ). Since t(X) ≤ κ,

there are CE ⊆ E and CF ⊆ F such that |CE |, |CF | ≤ κ and p ∈ cl(CE ) ∩ cl(CF ).

Since CE and CF have size no more than κ, they are in M and thus by elementarity
 38

there is a q ∈ X ∩ M with q ∈ cl(CE ) ∩ cl(CF ). But this is impossible since E and

F are disjoint and closed in X ∩ M. Hence cl(E) ∩ cl(F ) = ∅. ♠

We finish this section with two technical results to be used latter in this chapter.

Lemma 4.8 For all A ∈ [X]≤κ there is a B ⊆ X ∩ M with B ∈ M and

cl(A) ∩ Z ⊆ cl(B).

Proof. Fix A ∈ [X]≤κ . For each a ∈ A we use that t(X) ≤ κ to find a

Ba ∈ [X ∩ M]≤κ such that rM (a) ∈ cl(Ba ) and we let B = a∈A Ba . Note that
S

since A and each Ba have size no more than κ we have that B ∈ [X ∩ M]≤κ

and since M is κ-closed, B ∈ M. Now fix x ∈ Z \ cl(B); we want to show that

x∈
/ cl(A).

Since X is regular, there are U, V ′ , W ′ ∈ τ such that x ∈ U, cl(U) ⊆ V ′ ,

cl(B) ⊆ W ′ and V ′ ∩ W ′ = ∅. Since x ∈ Z and t(X) ≤ κ, there is a C ∈ [U ∩ M]≤κ

such that x ∈ cl(C). Since M is κ-closed we have that C ∈ M and therefore

by elementarity there are V, W ∈ τ ∩ M such that cl(C) ⊆ V , cl(B) ⊆ W and

V ∩ W = ∅.

Now if a ∈ A then rM (a) ∈ cl(Ba ) ⊆ cl(B) ⊆ W so by definition of rM we get

that a ∈ W , showing that A ⊆ W . This implies that cl(A) ⊆ cl(W ) ⊆ X \ V and

therefore, since x ∈ cl(C) ⊆ V , we get that x ∈

/ cl(A). ♠

Lemma 4.9 If H is a Gκ -subset of X and q ∈ H ∩ Z, then there is a Gκ -set

J ∈ M such that q ∈ J ∩ Z ⊆ H ∩ Z.

T
Proof. Fix a collection {Wα : α ∈ κ} such that H = α∈κ Wα . For each α ∈ κ

and each p ∈ Z \ Wα we use Lemma 4.2 to get Uαp , Vαp ∈ τ ∩ M such that p ∈ Uαp ,
 39

q ∈ Vαp and Uαp ∩ Vαp = ∅. Since Z \ Wα is closed (and hence of Lindelöf degree no

more that κ), we get that Z \ Wα ⊆ p∈P Uαp for some P ∈ [Z \ W ]≤κ . Clearly
S

q ∈ Jα := p∈P Vαp and Jα ∩ Z ⊆ Wα . Since M is κ-closed we have that each Jα

T

T
and hence J := α∈κ Jα is in M. It is clear that q ∈ J ∩ Z ⊆ H ∩ Z. ♠

4.2 Homogeneity

Recall that for a topological space X, h(X) is the least cardinal λ with the prop-

erty that for every point x ∈ X there is a compact K ⊆ X such that x ∈ K

and χ(K, X) ≤ λ. The following lemma was proved in [4] for X compact. The

requirement h(X) ≤ κ is obviously enough, since the tightness of any subspace of

X is no more than the tightness of X.

Lemma 4.10 If X is a space with t(X)h(X) ≤ κ, then there is a set A ∈ [X]≤κ

and a non-empty closed Gκ -subset H of X such that H ⊆ cl(A).

In general, given a point p ∈ X one cannot expect to get p ∈ H in the previous

lemma, even if X is compact. For example if X = κ+ ∪ {∞} is the one point

compactification of a discrete κ+ then p = ∞ is a counter-example. In this space

the closure of any set of size κ has size κ, but on the other hand any Gκ -subset

containing p must have size κ+ . However the following is obvious.

Corollary 4.11 If X is a homogeneous space with t(X)h(X) ≤ κ and q ∈ X,

then there is a set A ∈ [X]≤κ and a closed Gκ -subset H of X such that q ∈ H and

H ⊆ cl(A).
 40

Now we are ready to prove our first important result.

Theorem 4.12 If X is a homogeneous space then d(X) ≤ 2t(X)L(X)h(X) .

Proof. Let τ be the topology of X, let κ = t(X)L(X)h(X), and fix a κ-closed

M ≺ V with X, τ ∈ M and |M| ≤ 2κ . Let Z = cl(X ∩ M) and rM : X → Z as in

Section 4.1. We will show that in fact X = Z and hence d(X) ≤ |X ∩ M| ≤ 2κ .

Fix now x ∈ X and let q = rM (x) ∈ Z. By Corollary 4.11, we can find a set

A ∈ [X]≤κ and a closed Gκ -subset H of X such that q ∈ H ⊆ cl(A). By Lemma

4.9, there is a Gκ -set J ∈ M such that q ∈ J ∩ Z ⊆ H ∩ Z. By Lemma 4.8,

there is a B ⊆ X ∩ M with B ∈ M and cl(A) ∩ Z ⊆ cl(B). Thus we have that

J ∩ Z ⊆ cl(B), but since both J and B are in M, we get by elementarity that in

fact J ⊆ cl(B). Finally, since q ∈ J, by definition of rM we get that x ∈ J and

hence x ∈ cl(B) ⊆ Z, which is what we wanted. ♠

Since πχ(X) ≤ t(X)h(X) for any space X (see Corollary 1.13), we can use

Lemma 1.14 to get our main result of this chapter:

Theorem 4.13 If X is a homogeneous space then |X| ≤ 2t(X)L(X)h(X) .

Corollary 4.14 If X is a countably tight Lindelöf homogeneous space of point-

countable type then |X| ≤ 2ℵ0 .

Corollary 4.15 If X is a compact homogeneous space of countable tightness then

|X| ≤ 2ℵ0 .

Since |X| = 2χ(X) for compact homogeneous spaces (see parts 4 and 5 of The-

orem 1.11) and the tightness of a point (in any space) never exceeds its character,

we get:
 41

Corollary 4.16 (GCH ) In every compact homogeneous space the tightness and

the character coincide.

Corollary 4.17 (2ℵ0 < 2ℵ1 ) A compact homogeneous space has countable tightness

if and only if it is first countable.

In [23], J. van Mill asked whether every T5 (i.e. hereditarily normal) homo-

geneous compact space has cardinality 2ℵ0 . In [17], I. Juhász, P. Nyikos and Z.

Szentmiklóssy proved that the answer is yes in forcing extensions resulting by

adding (2ℵ1 )V Cohen reals. They also showed that after adding ℵ2 Cohen reals,

every T5 homogeneous compact space has countable tightness. Putting this to-

gether with Corollary 4.17 and assuming for example that 2ℵ0 = ℵ2 and 2ℵ1 = ℵ3

in the ground model, we get a confirmation of a conjecture proposed by them in

the same paper.

Theorem 4.18 It is consistent that every homogeneous T5 compact space is first

countable.
 42

Chapter 5

Final Remarks and Open

Questions

It is known that no compact topological group can be either an S-space or an L-

space. However, under CH , Kunen [20] has constructed compact L-spaces that are

right topological groups (i.e. they admit a group operation such that multiplication

on the right by a fixed element defines a continuous map). It is natural to ask:

Question 5.1 Is there a compact S-space which is a right topological group?

Note that the compact homogeneous S-space constructed in Chapter 2 is an

infinite power and therefore it is necessarily a strong S-space. In relation to this,

J. T. Moore asked the following:

Question 5.2 Is there a compact homogeneous S-space which is not a strong S-

space?

As we mentioned before, any compact homogeneous HS space must have size

no more than 2ℵ0 . Under CH this implies that the space is first-countable. On the

other hand, under MA + ¬CH any compact HS space is hereditarily Lindelöf and

hence (by part 1 of Theorem 1.11) first-countable. Thus we ask:

 43

Question 5.3 Is there a consistent example of a compact homogeneous HS space

which is not first-countable?

If we change the condition of being HS by having countable π-weight in the

previous question, then we know that the answer is positive. In [22], J. van Mill

presented a beautiful construction of a compact space of countable π-weight and

uncountable character which is homogeneous under MA+¬CH but not under CH .

His space is not HS and in fact it has uncountable tightness, so the following is

still open:

Question 5.4 (Arkhangel’skiı̆) Is there a consistent example of a countably tight

compact homogeneous space which is not first-countable?

In Chapter 4 we proved that |X| ≤ 2t(X) for any compact homogeneous X. We

know that the π-character of a compact space does not exceed its tightness, but

in principle it could be smaller. Thus we ask the following question, also asked in

[23] in a slightly different way.

Question 5.5 If X is compact and homogeneous, is it true that |X| ≤ 2πχ(X) ?

In Chapter 3 we have showed that under ♦ there is a zero-dimensional B-

homogeneous space which is not h-homogeneous. In addition, our space is HS and

for this reason it could not have been constructed in ZFC . Again, under MA+¬CH

any compact HS space is first-countable, so if it is also B-homogeneous, it would

have to be h-homogeneous by [21]. We don’t know if there is a counterexample to

Matveev’s question in ZFC :

 44

Question 5.6 Is there a ZFC example of a zero-dimensional B-homogeneous not

h-homogeneous compact space?

We finish with two well known questions which, although not directly related

with our work, should be included in any list of problems concerning compact

homogeneous spaces. They are open under any set-theoretical assumptions.

Question 5.7 (W. Rudin) Is there a compact homogeneous space without non-

trivial converging sequences?

Question 5.8 (E.K. van Douwen) If X is compact and homogeneous, is it true

that c(X) ≤ 2ℵ0 ?

 45

Bibliography

[1] P. Alexandroff and P. Urysohn, Mémoire sur les espaces topologiques compacts,

Verh. Konink. Akad. Wetensch. Afd. Natuurk., Sectie 1 (Amsterdam 1920), 14,

1-96.

[2] A. V. Arkhangel’skiı̆, A survey of some recent advances in general topology,

old and new problems, Proc. International Congress of Mathematicians (Nice

1970), Tome 2, 19-26.

[3] A. V. Arkhangel’skiı̆, The Suslin number and cardinality. Character of points

in sequential bicompacta, Soviet Math. Dokl. 11 (1970) 597-601.

[4] A. V. Arkhangel’skiı̆, Structure and classification of topological spaces and

cardinal invariants, Russian Math. Surveys 33:6 (1978) 33-96.

[5] A. V. Arkhangel’skiı̆, Topological homogeneity. Topological groups and their

continuous images (Russian), Uspekhi Mat. Nauk 42:2 (1987) 69-105 (English

translation: Russian Math. Surveys 42:2 (1987) 83-131).

[6] R. E. Chandler, G. D. Faulkner, J. P. Guglielmi, and M. C. Memory, General-

izing the Alexandroff-Urysohn double circumference construction, Proc. Amer.

Math. Soc. 83 (1981) 606-608.

[7] E. K. van Douwen, Nonhomogeneity of products of preimages and π-weight,

Proc. Amer. Math. Soc. 69 (1978) 183-192.

 46

[8] A. Dow, An introduction to applications of elementary submodels to topology,

Topology Proc. 13 (1988) 17-72.

[9] A. Dow and E. Pearl, Homogeneity in powers of zero-dimensional first-

countable spaces, Proc. Amer. Math. Soc. 125 (1997) 2503-2510.

[10] M. Dzamonja, K. Kunen, Measures on compact HS spaces, Fund. Math. 143

(1993) 41-54.

[11] V.V. Fedorchuk, On the cardinality of hereditarily separable compact spaces

(Russian), Dokl. Akad. Nauk SSSR 222 (1975) 302-305.

[12] A. Hajnal and I. Juhász, A separable normal topological group need not be

Lindelöf, Gen. Top. and App. 6 (1976) 199-205.

[13] K. P. Hart, Review of [20], Mathematical Reviews, 2003a:54040.

[14] R. Hodel, Cardinal functions, I, in Handbook of Set-Theoretic Topology, North-

Holland, 1984, pp. 1-61.

[15] M. Ismail, Cardinal functions of homogeneous spaces and topological groups,

Math. Japon. 26 (1981) 635-646.

[16] I. Juhász, K. Kunen and M.E. Rudin, Two more hereditarily separable non-

Lindelöf spaces, Can. J. Math. 28 (1976) 998-1005.

[17] I. Juhász, P. Nyikos and Z. Szentmiklóssy, Cardinal restrictions on some ho-

mogeneous compacta, Preprint.

 47

[18] L.R. Junqueira and F.D. Tall, The topology of elementary submodels, Top.

Appl. 82 (1998) 239-266.

[19] O.H. Keller, Die Homoiomorphie der kompakten konvexen Mengen im

Hilbertschen Raum, Mathematische Annalen 105 (1931) 748-758.

[20] K. Kunen, Compact L-spaces and right topological groups, Top. Proc. 24

(1999) 295-327.

[21] M.V. Matveev, Basic homogeneity in the class of zero-dimensional spaces,

Filomat 12:2 (1998) 13-20.

[22] J. van Mill, On the character and π-weight of homogeneous compacta, Israel

J. Math. 133 (2003) 321-338.

[23] J. van Mill, On the cardinality of power homogeneous compacta, Top. Appl.

146-147 (2005) 421-428.

[24] D. B. Motorov, Zero-dimensional and linearly ordered bicompacta: properties

of homogeneity type, Russian Math. Surveys 44:6 (1989) 190-191.

[25] S. Negrepontis, Banach spaces and topology, in Handbook of set-theoretic

topology, North-Holland, Amsterdam, 1984, pp. 1045-1142.

[26] Z. Szentmiklóssy, S-spaces and L-spaces under Martin’s axiom, Topology, Vol.

II, Colloq. Math. Soc. János Bolyai 23, North-Holland, 1980, pp. 1139-1145.

[27] S. Todorčević, Partition Problems in Topology, Contemporary Mathematics,

84, American Mathematical Society, 1989.