Topology Proceedings: Semigroup Actions: Proximities, Compactifications and Normality

TOPOLOGY
PROCEEDINGS
Volume 35 (2010)
Pages 1-35
http://topology.auburn.edu/tp/
E-Published on April xx, 2009
SEMIGROUP ACTIONS: PROXIMITIES,
COMPACTIFICATIONS AND NORMALITY
LEONID GOOGLE AND MICHAEL MEGRELISHVILI
This work is dedicated to the memory of Yu.M. Smirnov.
Abstract. We study equivariant compactications of con-
tinuous actions of topological semigroups. We give a trans-
parent description of such compactications in terms of
S-proximities, special action compatible proximities on X.
This leads us to a dynamical generalization of classical
Smirnovs theorem. As a strictly related concept we investi-
gate equivariant normality of actions and give some relevant
examples.
Contents
1. Introduction 2
2. Main results 3
3. Topological Background 4
3.1. Compactications of topological spaces 4
3.2. Uniform spaces 5
3.3. Proximities and proximity spaces. 8
3.4. Uniform spaces and the corresponding proximity 9
3.5. Proximity mappings 11
3.6. Smirnovs Theorem 11
4. Semigroup actions 12
2000 Mathematics Subject Classication. 54H15, 54H20.
Key words and phrases. Equivariant compactication, equivariant normality,
proximity space, semigroup action.
c 2009 Topology Proceedings.
1
2 LEONID GOOGLE AND MICHAEL MEGRELISHVILI
4.1. Actions and uniformities 14
5. Generalized Smirnovs Theorem for semigroup actions 17
5.1. Proximities for semigroup actions 17
5.2. Generalized Smirnovs theorem 19
5.3. Algebras of -uniform functions for semigroup actions 20
6. Equivariant normality of semigroup actions 22
6.1. Urysohns Theorem for semigroup actions 24
7. Some examples 26
8. Actions of topological groups 28
References 33
1. Introduction
A topological transformation semigroup (S-space, or an S-ow)
is a continuous action of a topological semigroup S on a topological
space X. For the particular case of a topological group action we
reserve the symbol G.
In the present work we investigate the following problems:
(A) Characterize equivariant compactications of semigroup
actions. For instance what about corresponding proximities and
a possible equivariant generalization of Smirnovs theorem?
(B) When a given action satises the normality conditions in the
spirit of the classical Urysohn and Tietze theorems?
One of our main objects is the dynamical analogue of the com-
pactication concept. Compactiabilityof topological spaces means
the existence of topological embeddings into compact Hausdor
spaces. For the compactiability of ows we require in addition the
continuous extendability of the original action. Compactiable G-
spaces are known also as G-Tychono spaces. See Fact 4.3 below
and also a recent review paper [18].
The compactications of a Tychono space X can be described
in several ways:
Banach subalgebras of C(X) (Gelfand-Kolmogoro 1-1 cor-
respondence [14]);
Completion of totally bounded uniformities on X (Samuel
compactications);
Proximities on X (Smirnovs Theorem).
SEMIGROUP ACTIONS: PROXIMITIES, COMPACTIFICATIONS... 3
It is well known (see for example [9, 37, 3, 22, 19]) that the
rst two correspondences admit dynamical generalizations in the
category of G-ows. Instead of continuous bounded functions,
we should use special subalgebras of -uniform functions C
(X)
(Denition 5.9) and instead of precompact uniformities, we need
now precompact equiuniformities (Denition 4.4).
Now about the third case the description by proximities. In the
case of G-spaces, group actions, it was initiated by Smirnov himself
in [5]. We will see that there exists a natural generalization of
Smirnovs Theorem for semigroup actions in terms of S-proximities,
special action compatible proximities on X.
Our second direction is a closely related theme of equivariant
normality of actions. The existing functional and topological char-
acterizations of usual normality admits a dynamical generalization
for group actions; see [21, 26, 24]. However full proofs never have
been published before. Later some natural generalizations for semi-
group case appeared in [7].
We give here a self-contained and unied exposition of some new
and old results around two problems (A) and (B) mentioned above.
We obtain several results which concentrate on similarities between
group and semigroup cases. On the other hand we give also relevant
contrasting counterexamples.
2. Main results
For every subset A of a topological space X we denote by N
A
or
N
A
(X) the set of all neighborhoods (in short: nbds) of A in X.
Denition 2.1. Let X be an S-space where S is a topological
semigroup.
(1) Subsets A, B X are -disjoint if for every s
0
S there
exists U N
s
0
(S) such that U
1
A U
1
B = . Notation:
A
B. We write A
B if A and B are not -disjoint. We

write A
B if A
(X B).
(2) Assume that is a continuous proximity on X. We say that
is an S-proximity if the following condition is satised: for
every pair A
B of -far subsets A, B in X and every s

0
S
there exists a nbd U N
s
0
such that U
1
A
U
1
B.
Theorem 2.2. (Generalized Smirnovs theorem for semigroup
actions) In the canonical 1-1 correspondence between continuous
S-proximities on X and compactications of X the S-compactica-
tions are in 1-1 correspondence with continuous S-proximities.
monoid and the action is monoidal. We say that X is S-normal if
for every pair A, B X of closed -disjoint subsets, there are open
disjoint nbds O
A
N
A
and O
B
N
B
, such that A
O
A
and
B
O
B
.
If S is discrete then this is equivalent to the usual (topological)
normality of X.
Theorem 2.4. (Normality for monoidal actions) Let X be an S-
space. The following are equivalent:
(1) X is S-normal.
(2) The relation
A
B cl(A)
cl(B)
is an S-proximity on the set X.
(3) For every pair A, B of closed -disjoint subsets in X there
exists f C
(X) such that f(A) = 0, f(B) = 1.

(4) For every pair A, B X of closed -disjoint subsets, there
are -disjoint open nbds O
A
N
A
and O
B
N
B
, such that
A
O
A
and B
O
B
.
(5) (Urysohns Small Lemma for S-spaces) For every closed
subset A and its open nbd O such that A
O there exists
an open nbd O
1
of A such that A
O
1
and cl(O
1
)
O.
Furthermore, if one of these equivalent conditions is satised then
, the greatest continuous S-proximity on the S-space X.

3. Topological Background
3.1. Compactications of topological spaces. Let X be a topo-
logical space, Y be a compact Hausdor space and let f : X Y
be a function such that f(X) is dense in Y . If f is continuous,
then Y (and sometimes a pair (Y, f)) is called a compactication of
X. If f is a homeomorphic embedding, then Y is called a proper
compactication of X. Denote by (C(X), ) the partially ordered
set of all compactications of X up to the standard equivalence.
Let C(X) be the algebra of all real valued bounded continuous
functions on X with respect to the supremum norm. Recall that
the unital (containing the constants) closed subalgebras of C(X)
determine the compactications of X.
Fact 3.1. (Gelfand-Kolmogoro [14]) There exists a natural order
preserving bijective correspondence between C(X) (dierent com-
pactications of X) and closed unital subalgebras of C(X). In par-
ticular, C(X) determines the greatest compactication : X
X.
3.2. Uniform spaces.
Denition 3.2. A uniformity on a set X is a subset of the set of
all relations on X which satises the following conditions (axioms
of uniformity):
(U1) If and , then ;
(U2) If
1
,
2
, then
1

2
;
(U3) If , then
1
;
(U4) For every there exists such that
2
:= .
An element of the uniformity is called an entourage of and the
pair (X, ) is called a uniform structure. If in addition it satises
the following axiom:
(U5)

=
X
(where
X
is the diagonal of X X);
then will be referred to as a separated (or, Hausdor) uniformity.
Let x
0
be a point of X and ; the set
B(x
0
, ) := y X : (y, x
o
)
is called the ball with center x
0
and radius , or -ball about x
0
.
The family
= top() := O X : x O such that B(x, ) O

is a completely regular topology on X. Every entourage
yields the cover B(x, )
xX
of X which is said to be -uniform
cover with respect to . About equivalent approach to the theory
of uniform spaces through coverings see [16].
Denition 3.3. Let (X, ) be a topological space and be a uni-
formity on the set X. Then we say that is continuous if

and compatible if
= .
A nite subset A X is an -net with respect to an entourage
if B(x, ) : x A still is a covering of X. If admits an
-net then we say that is totally bounded. A uniform space (X, )
is totally bounded or precompact if every is totally bounded.
A Hausdor uniform space (X, ) is precompact i the completion
is a compact Hausdor space. On a compact Hausdor space X
there exists a unique compatible uniformity
X
. The family of all
neighborhoods of the diagonal constitutes a uniform basis of
X
.
Remark 3.4. (Initial uniformity) Let f : X (Y, ) be an arbi-
trary map from the set X into a uniform space (Y, ). The system
of entourages
(f f)
1
() X X :
is a basis of some uniformity
f
on X, the corresponding initial
uniformity. Then
(1)
f
is the coarsest uniformity on X such that f : (X,
f
)
(Y, ) is uniform.
(2) The uniform topology
f
coincides with the initial topology
on X with respect to the map f : X (Y,
). Therefore if
is a topology on the set X such that f : (X, ) (Y,
)
is continuous then
f
is -continuous. That is,
f
.
This fact is a particular case of a more general assertion about
initial uniformities. See for example [32, Proposition 0.17].
Remark 3.5. (Associated Hausdor uniformity) For every pseu-
dometric d on X one can dene a natural metric space (X
, d
)
dening X
as the quotient set X/ with respect to the equivalence

relation xy i d(x, y) = 0. Then d
is dened as d
([x], [y]) =
d(x, y). Analogously if a uniform space (X, ) is not necessarily
separated then the associated Hausdor uniform space (X
) is
dened in a similar way, [32, p.18]. More precisely, X
:= X/
where := : is always an equivalence relation on
X. Denote by q : X X
, x [x] the natural map. Now the

following conditions are satised:
(1) The family (q q)()
is a base of some Hausdor uni-

formity
on X
.
(2) (q q)
1
(q q)() for every .
(3) is an initial uniformity w.r.t. the map q : (X, )
(X
).
Remark 3.6. (Samuel Compactication) Now we recall the def-
inition of Samuel compactication of a uniform space (X, ). We
consider separately two cases:
(a) is Hausdor. Every uniform structure contains the pre-
compact replica of . It is the nest uniformity among all
coarser totally bounded uniformities. In particular,
fin

. It is well known that
fin
is just the family of all to-
tally bounded entourages of . Denote by i
fin
: (X, )
(X,
fin
), x x the corresponding uniform map. This
map is a homeomorphism because top() = top(
fin
). The
uniformity
fin
is separated and hence the corresponding
completion (X,
fin
) (
X,
fin
) = (uX,
u
) (or simply
uX) is a proper compactication of the topological space
(X, top()). The compactication
u
X
= u
(X,)
: X uX
is the well known Samuel compactication (or, universal
uniform compactication) of (X, ) (see [33, 16]).
(b) is not Hausdor.
Then rst we pass to the associated Hausdor uniformity.
That is consider the map q : (X, ) (X
). Now we
can apply part (a). That is, consider the usual Samuel com-
pactication of (X
). The resulting composition map

u q : (X, ) (uX
u
) is the Samuel compactication
of (X, ).
Clearly the uniformity is Hausdor if and only if the corre-
sponding Samuel compactication u : X uX is proper. In both
cases (a) or (b) the corresponding algebra /
u
C(X) consists with
all -uniformly continuous real valued bounded functions on X.
Fact 3.7. There exists a natural order preserving 1-1 correspon-
dence between continuous (Hausdor) totally bounded uniformities
on X and (resp., proper) compactications of X.
Remark 3.8. This principal result is standard for Hausdor unifor-
mities and proper compactications. In general this fact, due to Gal
[13], requires a more careful analysis. For the sake of completeness
we include a sketch of the proof.
Proof. (of Fact 3.7) Let : X Y be a compactication of (X, ).
The system of entourages
( )
1
() X X :
Y
is a base of a uniformity
on X (the corresponding initial unifor-

mity). The compactness of Y implies that
Y
is precompact. Then
is also precompact. It is also easy (Remark 3.4) that
is a
continuous uniformity on the topological space (X, ).
Conversely, let be a totally bounded -continuous uniformity
on X. The corresponding Samuel compactication
: (X, ) (
)
is dened as the composition of two uniform maps: q : (X, )
(X
) (the associated Hausdor uniformity) and the usual com-

pletion j : (X
) (
). Denote by Y the compact Haus-

dor space

X
with its unique uniformity

Y
:=
. Now observe
that the initial uniformity on X w.r.t. map : X (Y,
Y
) is just
(see Remarks 3.5 and 3.6 above).
Alternative proof can be derived also making use Fact 3.1.
Example 3.9. Let G be a topological group and H is a subgroup.
The following system of entourages
U := (xH, yH) G/H G/H : xy

1
U (U N
e
(G))
is a basis of a compatible uniformity
R
on the coset space G/H,
the so-called right uniformity.
3.3. Proximities and proximity spaces. F. Riesz in 1908 rst
formulated a set of axioms to describe the notion of closeness of a
pair of sets. The most useful version of proximity was introduced
by V.A. Efremovich [11] (see also [29] and [6, Chapter 1.5]).
Denition 3.10. Let X be a nonempty set and be a relation in
the set of all its subsets. We write AB if A and B are -related
and AB if not. The relation will be called a proximity on X
provided that the following conditions are satised:
(P1) A B ,= implies AB.
(P2) AB implies BA;
(P3) AB implies A ,= ;
(P4) A(B C) i AB or AC;
(P5) If AB then there exist C X such that AC and (XC)B.
A pair (X, ) is called a proximity space. Two sets A, B X are
near (or proximal ) in (X, ) if AB and remote (or, far) if AB.
Every proximity space (X, ) induces a topology := top() on X
by the closure operator:
cl
(A) := x X : xA.
The topology top() is Hausdor i the following condition is sat-
ised
(P6) If x, y X and xy then x = y.
Every (separated) proximity space (X, ) is completely regular
(resp., Tychono) with respect to the topology top() (can be de-
rived from Remark 3.13).
Let (X, ) be a topological space. A proximity of X is called
continuous (or, more precisely, -continuous proximity) if top()
. In the case of top() = , we say that is a compatible proximity
on the topological space (X, ).
We say a subset A X is strongly contained in B X with
respect to , if A(X B) and write: A B. A nite cover A =
A
i
n
i=1
is called a -cover if and only if there exists another nite
cover B = B
i
n
i=1
, such that B
i
A
i
for every i = 1, . . . , n.
Lemma 3.11. (1) In Denition 3.10 one can replace (P5) by
one of the following axioms:
(P5.a) If AB then there exist C, D X such that CD = X
and AC, BD.
(P5.b) If AB then there exist subsets A
1
and B
1
of X such
that A A
1
, B B
1
and A
1
B
1
= .
(P5.c) If AB then there exist subsets A
1
and B
1
of X such
that A A
1
, B B
1
and A
1
B
1
.
3.4. Uniform spaces and the corresponding proximity.
Denition 3.12. Let be a uniformity on X. Then the corre-
sponding relation
dened by
A
B (AB) ,=
is a proximity on X which is called the proximity induced by the
uniformity .
Always, top() = top(
). Proximity
is Hausdor i is
Hausdor.
Conversely every proximity on a topological space denes canon-
ically a totally bounded compatible uniformity
. It is well known
for Hausdor proximities (cf. for example Engelking [12]). For not
necessarily Hausdor case see Gal [13].
Remark 3.13. Let (X, ) be a proximity space. The collection B of
all sets V of the form:
V :=
k
i=1
(A
i
A
i
),
where A
i
k
i=1
is a -cover of X, denes a basis of the unifor-
mity
. Uniformity
is totally bounded and =
. More-
over, top() = top(
). We call
the induced uniformity of . Its

induced proximity in the sense of Denition 3.12 gives back the
original proximity .
Example 3.14. Let (X, d) be a metric space. The corresponding
proximity on X is dened as follows
A
d
B d(A, B) = 0.
Example 3.15. Let (G, ) be a topological group with a subgroup H
and
R
is the right uniformity (Example 3.9). The corresponding
compatible proximity
R
on G/H is:
A
R
B U N
e
(G) : UA B ,= .
Example 3.16. (1) Let X be a Tychono space. The relation
dened by
A
B f C(X) such that f(A) = 0 and f(B) = 1

is a proximity which corresponds to the greatest compatible
uniformity on X. The proximity
B comes from the Stone-
Cech compactication : X X.
(2) A Hausdor topological space X is normal i the relation
A
n
B i cl(A) cl(B) ,=
denes a proximity relation on the set X.
(3) Let Y be a compact Hausdor space. Then there exists a
unique compatible proximity on the space Y dened by
AB cl(A) cl(B) ,= .
3.5. Proximity mappings.
Denition 3.17. [29, Denition 4.1] Let (X,
1
) and (Y,
2
) be
two proximity spaces. A mapping f : X Y is called a proximity
mapping if one of the following equivalent conditions are satised:
(C1) A
1
B = f(A)
2
f(B).
(C2) C
2
D = f
1
(C)
1
f
1
(D).
(C3) C D = f
1
(C) f
1
(D).
Fact 3.18. (1) f : (X,
1
) (Y,
2
) is a proximity mapping if
and only if it is uniformly continuous with respect to the
uniformities
1
and
2
.
(2) In a proximity space (X, ), A
B if and only if there exists

a proximity mapping f : X [0, 1] such that f(A) = 0 and
f(B) = 1.
(3) Let A be any subspace of a proximity space (X, ) and let
f : A [a, b] be a proximity mapping. Then f can be
extended to a proximity mapping f : X [a, b].
Remark 3.19. Let (Y, ) be a proximity space and f : X Y be
an arbitrary mapping. Dene the following natural relation on
X X by the following rule:
AB f(A)f(B).
Then :=
f
is the smallest proximity on X such that the map f is
proximity. We call it the initial proximity. There exists a natural
link between initial proximity and initial uniformity. Precisely, let
be the induced uniformity of and (
)
f
be the initial uniformity
on X. Then the induced proximity of (
)
f
is just =
f
.
3.6. Smirnovs Theorem. Let : X Y be a compactication.
Denote by
the corresponding initial proximity on X dened via

the canonical admissible proximity
Y
of Y . More precisely for
subsets A, B of X we dene A

B if (A)
Y
(B), i.e., if cl((A))
cl((B)) = .
Conversely every continuous proximity on a topological space
induces a totally bounded uniformity
(Remark 3.13). Now com-

pletion gives a compactication which we denote by c
. It is equiv-
alent to the Samuel compactication with respect to the uniformity
. This leads to a description of compactications in terms of prox-

imities. This approach is the so-called Smirnovs compactication
(see for example [34, 12, 29]).
Theorem 3.20. (Smirnovs classical theorem) Let X be a
topological space. By assigning to any compactication : X Y
the proximity
on X gives rise to a natural one-to-one order

preserving correspondence between all compactications of X and
all continuous proximities on the space X.
Note that originally Smirnovs theorem deals only with proper
compactications and compatible proximities. The present general
version is due to Gal [13].
4. Semigroup actions
Let (S, ) be a semigroup and assume that also S has a topological
structure . We say that (S, ) is a topological semigroup if the
multiplication map: S S S is continuous.
Topological S-ow (or an S-space) is a triple S, X, ) where
: S X X, (s, x) = sx = s(x) = x(s)
is a jointly continuous left action of a topological semigroup S on a
topological space X; we write it also as a pair (S, X), or simply, X
(when and S are understood). Action means that the following
condition is satised:
s
1
(s
2
x) = (s
1
s
2
)x (s
1
, s
2
, x) S S X.
We dene for every x X the corresponding orbit map x : S
X, by x(s) := sx and also for every s S, s-translations s : X
X, by s(x) := sx.
If S is a monoid and e is the identity of S then the action is
monoidal will mean that e is the identity mapping of X. In partic-
ular, for topological group actions we always require that the action
is monoidal. For acting group we reserve the symbol G. As usual
topological group means that in addition the inverse mapping is
also continuous.
For U S and A X dene UA := ua : (u, a) U A. For
every (s, x) S X we dene s
1
x := y X : sy = x and for
A X,
s
1
A :=
s
1
a : a A = y X : sy A,
U
1
A :=
u
1
A : u U = x X : Ux A ,= ,
U A :=
u
1
A : u U.
Examples 4.1. We recall some natural ways getting topological
monoids and monoidal actions (see for example [1], [19]).
(1) Let (Y, ) be a uniform space. Denote by
sup
the uni-
formity of uniform convergence on the set Unif(Y, Y ) of
all uniformly continuous self-maps Y Y . Then under
the corresponding topology top(
sup
) on Unif(Y, Y ) and
the usual composition we get a topological monoid. For
every subsemigroup S Unif(Y, Y ) the induced action
S Y Y denes a topological ow.
(2) For instance, for every compact space Y the semigroup
C(Y, Y ) endowed with the compact open topology is a topo-
logical monoid. The subset Homeo(Y ) in C(Y, Y ) of all
selfhomeomorphisms Y Y is a topological group.
(3) For every metric space (M, d) the semigroup (M, d) of
all d-contractive maps f : X X (that is, d(f(x), f(y))
d(x, y)) is a topological monoid with respect to the topology
of pointwise convergence. Furthermore, (M, d) M M
is a continuous monoidal action.
(4) For every normed space (V, [[ [[) the semigroup (V ) of
all contractive linear operators V V endowed with the
strong operator topology (being a topological submonoid of
(V, d) where d(x, y) := [[x y[[) is a topological monoid.
(5) For every normed space V and a subsemigroup S (V )
op
,
where (V )
op
is the opposite semigroup of (V ), the in-
duced action S B
on the weak star compact unit

ball B
of the dual space V
is continuous (see [19, Lemma

2.4]).
(6) Every normed algebra A treated as a multiplicative monoid
is a topological monoid. The subset B
A
is a topological
submonoid. In particular, for every normed space V the
monoids L(V ) and B
L(V )
of all bounded and, respectively,
of all contractive linear operators V V are topological
monoids endowed with the norm topology.
Let X, Y be S-ows. A function f : X Y is an S-map if
f(sx) = sf(x) for all (s, x) S X. An S-compactication of X is
an S-map : X Y where is a compactication of X.
We dene an analogue of Tychono spaces in the class of S-
spaces.
Denition 4.2. A ow (S, X) is said to be compactiable, or
S-Tychono if there exist a compact S-ow Y and a proper S-
compactication : X Y .
In [36], de Vries posed the compactication problem: is it true
that for arbitrary Hausdor topological group G every Tychono
G-space is compactiable? This question is negatively answered in
[23], even for Polish G-spaces.
Fact 4.3. Recall some useful sucient conditions (compare [18])
when G-spaces are compactiable:
(1) if G is locally compact then every Tychono G-space is
G-Tychono (de Vries [38]);
(2) every coset G-space G/H (de Vries [36]);
(3) every metric G-space (X, d) with the G-invariant metric d
(Ludescher and de Vries [17]);
(4) every metric G-space (X, d), where G is second category and
g : X X is d-uniformly continuous for every g G, and
also every linear G-space X (Megrelishvili [22]);
(5) every G-space X, where X is Baire, G is uniformly Lindelof
and acts transitively on X (Uspenskij [35]);
(6) every rst countable G-space (more generally, every G-space
with the b
f
-property), where G is locally pseudocompact
(S. Antonyan and M. Sanchis [4, Theorem 5]).
For some results related to Fact 4.3.5 see Chatyrko and
Kozlov [10]. For some new directions in the theory of G-compacti-
cations we refer to van Mill [27, 28] and the references thereof.
Several well known results for G-spaces cannot be generalized to
the case of S-spaces. As a typical example note that there exists a
discrete S-space X with a compact monoid S such that X is not
S-Tychono [19].
4.1. Actions and uniformities.
Denition 4.4. Let be a uniformity on an S-space X. We call
the action:
(1) -saturated if every s-translation s : X X is -uniform;
(2) -bounded at s
0
if for every there exists a neighbor-
hood U N
s
0
such that (s
0
x, sx) for each x X and
s U. If this condition holds for every s
0
S then we
simply say -bounded or is a bounded uniformity;
(3) -equiuniform if it is -saturated and -bounded. Some-
times we say also that is an equiuniformity.
For topological group actions this concept is well known. See for
instance [9, 38, 25].
Proposition 4.5. Let S, (X, ), ) be a uniform S-space, then the
following conditions are equivalent:
(1) is an equiuniformity;
(2) is saturated and the corresponding homomorphism
h
: S Unif(X, X), s s
is continuous;
(3) For every and s
0
S, there exist and U
N
s
0
(S), such that if (x, y) , then (s
1
x, s
2
y) s
1
, s
2

U;
(4) For every and s
0
S, there exist
and U
N
s
0
(S), such that if A and B are subsets in X with the
property (A B) = , then (U
1
A U
1
B)
=
(that is if A and B are -far then U
1
A and U
1
B are
-far).
Proof. (1) (2) : Is trivial.
(1) (3) : We choose s
0
S and . There exists

such that
.
Since the action is -saturated for
we can choose such

that
(x, y) = (s
0
x, s
0
y)
.
Also from the boundedness of the action we can choose U N
s
0
(S)
such that
(s
1
x, s
0
x)
, (s
0
y, s
2
y)
s
1
, s
2
U.
Now for (x, y) we obtain (s
1
x, s
2
y)
.
(3) (4) : The condition (AB) = , means that
(a, b) A B : (a, b) / .
Let s
0
S. By (3) for we can choose
and U N
s
0
(S)
such that
(x, y)
= (s
1
x, s
2
y) , s
1
, s
2
U. (4.1)
Now we claim that (U
1
AU
1
B)
= . Assuming the contrary

we get
(U
1
AU
1
B) ,= = (x, y) (U
1
AU
1
B) : (x, y)
.
Therefore by denition of U
1
A and U
1
B we conclude:
s
, s
U : (s
x, s
y) AB.
On the other hand by Formula 4.1 for s
, s
U we have:
(x, y)
= (s
x, s
y) .
This means
: (s
x, s
y) (AB) .
Hence (AB) ,= , a contradiction.
(4) (3) : Choose
and U(s
0
) such that (4) is satised.
Then we claim that (s
1
x, s
2
y) s
1
, s
2
U; whenever (x, y)
.
Assuming the contrary let (s
1
x, s
2
y) / . Denote A := s
1
x
and B := s
2
x. Then (A B) = . Hence necessarily
(U
1
A U
1
B)
= . On the other hand x U

1
A and
y U
1
B. Therefore (x, y) /
, a contradiction.
(3) (1) : Is trivial.
The following simple lemma provides two important examples.
In fact the second assertion can be derived from the rst. The rst
assertion follows from Proposition 4.5. Alternatively they can also
be easily veryed directly.
Lemma 4.6. (1) For every separated uniform space (X, ) the
natural action of the topological semigroup S := Unif(X, X)
on X is -equiuniform.
(2) For every compact S-space Y the action is equiuniform with
respect to the canonical uniformity
Y
.
Proposition 4.7. Let f : (X, ) (Y, ) be a uniform S-map.
Suppose that is the initial uniformity (Remark 3.4). Then if is
an equiuniformity then the same is true for .
Proof. It is straightforward using the fact that the system of en-
tourages
(f f)
1
() X X :
is a base of the uniformity .
Proposition 4.8. Let (X
) be the associated uniform space

(Remark 3.5) of the uniform space (X, ). Assume that is equiu-
niform. Then
is also equiuniform.
Proof. It is straightforward using the fact that the system of en-
tourages (q q)()
is a base of the uniformity
on X
.
The following theorem is well known for group actions and sepa-
rated uniformities (see for example, [9, 25]). For semigroup actions
and separated uniformities it appears in [19].
Theorem 4.9. Let X be an S-space.
(1) Assume that : S X X is a (separated) -equiuniform
semigroup action. Then the induced action
u
: S uX
uX on the Samuel compactication uX := u(X, ) is a
(resp., proper) S-compactication of X.
(2) There exists a natural one-to-one correspondence between
S-compactications of X and continuous totally bounded
equiuniformities on X.
Proof. (2) follows from (1).
For separated equiuniformities (1) is exactly [19, Prop. 4.9]. It
is easy to extend this result for a not necessarily separated case by
Propositions 4.7 and 4.8
5. Generalized Smirnovs Theorem for semigroup
actions
5.1. Proximities for semigroup actions.
Denition 5.1. Let X be an S-space.
(1) The subsets A, B of X are -disjoint at s
0
S if there exists
U N
so
(S) such that U
1
A U
1
B = . If this condition
holds for every s
0
S then we simply say: -disjoint sets.
Notation: A
B.
(2) We write A
B if A and B are not -disjoint.

(3) We write A
B if sets A and B
c
are -disjoint (where
B
c
:= X B).
Lemma 5.2. (1) A
B i there exists s
0
S such that for
every neighborhood U of s
0
one may choose x
0
X such
that Ux
0
A ,= and Ux
0
B ,= .
(2) Let S be a monoid and A B ,= . Then A
B. Hence
-disjoint subsets are disjoint.
(3) If S is discrete and A and B are disjoint then they are -
disjoint. If in addition S is a monoid then the converse is
also true.
(4) A
B i B
A;
(5)
X;
(6) A
(B C) i A
B or A
C;
(7) The relation
is a proximity on the set X for every topo-

logical group action.
(8) A
B i for every s
0
S there exists U N
s
0
(G)
such that s
1
A t
1
B for every s, t U. It is also
equivalent to saying that U
1
A U B, where U B :=
u
1
B : u U.
Proof. The proof is straightforward. For example, for (6) use the
equality U
1
B U
1
C = U
1
(B C).
The following denition is a generalized version of Smirnovs con-
cept from [5].
semigroup. Assume that is a proximity on X. We say that is
an S-proximity if for every pair A
B of -far subsets A, B in X and

every s
0
S there exists a nbd U N
s
0
such that U
1
AU
1
B.
If S is discrete then this condition simply means that the trans-
lations s : X X are -proximity mapping (see Denition 3.17).
Proposition 5.4. Let be an equiuniformity on an S-space X.
Then
is an S-proximity.
Proof. Let A
B, i.e. by Denition 3.12 there exists an entourage

, such that (AB) = . Fix s
0
S. Then by Proposition
4.5 there exist
and a neighborhood U of s
0
in S such that
U
1
A and U
1
B are
-far. This means that U

1
A
U
1
B.
Proposition 5.5. (1) Let : X Y be an S-compactication.
The corresponding initial proximity
on X is a (continu-
ous) S-proximity on X.
(2) For every compact S-space Y the canonical proximity
Y
is
an S-proximity.
Proof. (1): By Lemma 4.6.2 the unique compatible uniformity
of Y is equiuniform. By Proposition 4.7 the corresponding initial
uniformity
on X is equiuniform, too. Then the induced prox-

imity by Proposition 5.4 is an S-proximity. Finally observe that by
Remark 3.19 this proximity is just
.
(2): Easily follows from (1).
5.2. Generalized Smirnovs theorem. Our aim here is to prove
an equivariant generalization of the classical Smirnovs Theorem
3.20 for the case of semigroups actions. For group actions it was
done (in a dierent but equivalent form) by Smirnov himself in [5].
Theorem 5.6. (Smirnovs theorem for semigroup actions) In the
canonical 1-1 correspondence between continuous proximities on
X and compactications of X the S-compactications are in 1-1
correspondence with continuous S-proximities.
Proof. Let : X Y be an S-compactication. Then by Proposi-
tion 5.5 the corresponding proximity
on X is an S-proximity
on X. Converse direction will follow by Proposition 5.8 below
which states that if is a continuous S-proximity, then
is an
S-compactication.
Lemma 5.7. Let X be an S-space and be a proximity on X. The
following are equivalent :
(1) is an S-proximity on X (for every pair A
B of -far sub-
sets A, B in X and every s
0
S there exists U N
s
0
such
that U
1
AU
1
B).
(2) The following two conditions are satised:
(i) every s-translation s : X X is a proximity mapping.
(ii) for every pair A
B of -far subsets A, B in X we have

A
B (that is for every s

0
S there exists U N
s
0
such that U
1
A U
1
B = ).
(3) The induced uniformity
is an equiuniformity on X.
Proof. (1) (2):
(i) Since s
0
U(s
0
) we have
AB s
1
0
(A)s
1
0
(B).
This condition means by Denition 3.17 that s
0
: X X is a
proximity mapping. (ii) is trivial because -far subsets are always
disjoint by axiom (P1) of Denition 3.10.
(2) (3): Consider the corresponding induced precompact uni-
formity
on X (Remark 3.13). We have to show that (X,
) is
equiuniform. Observe that the action on X is
-saturated (be-
cause every translation s : X X is
-uniform by (i)). Hence

we have only to show that it is also
-bounded. If not then there

exist:
and s
0
S such that for every U N
s
0
we can choose
(u, x) U X with the property (s
0
x, ux) / . Then ux / (s
0
x).
On the other hand, by the properties of
(see for example [12,

Theorem 8.4.8]) there exist two nite covers T := A
1
, A
2
, , A
n
and Q := B
1
, B
2
, , B
n
such that, B
i
A
i
and
(A
i
A
i
) .
Then by condition (ii) there exists a nbd V of s
0
such that
V
1
B
i
V
1
(X A
i
) = i = 1, . . . , n.
By our assumption on the pair (, s
0
) there exists a pair (v, x)
V X such that vx / (s
0
x). Choose i
0
such that s
0
x B
i
0
. We
get s
0
x B
i
0
A
i
0
(s
0
x). Then clearly vx / A
i
0
. Equivalently,
vx X A
i
0
. Therefore, x V
1
(X A
i
0
). On the other hand,
x V
1
B
i
0
because s
0
x B
i
0
. Thus, x V
1
B
i
0
V
1
(X A
i
0
).
This contradicts the fact V
1
B
i
0
V
1
(X A
i
0
) = .
(3) (1): Directly follows from Proposition 5.4.
Proposition 5.8. Let be a continuous S-proximity on an S-ow
X. Then the associated Smirnovs compactication
: X Y is
an S-compactication of X.
Proof. Let be a continuous S-proximity on X. Consider the cor-
responding continuous precompact uniformity
on X. Then the
Smirnovs compactication of X dened by the proximity is just
the Samuel compactication
:= u
(X,
)
: X u
X
of
. By Lemma 5.7,
is an equiuniformity on X. Now Propo-

sition 4.9 says that
is an S-compactication of X.
5.3. Algebras of -uniform functions for semigroup actions.
Denition 5.9. Let X be an S-space. A function f : X R
is -uniform if f is continuous, bounded and for every > 0 and
s
0
S there exists a nbd U(s
0
) such that
[f(s
1
x) f(s
2
x)[ < (x, s
1
, s
2
) X U U.
Family of all -uniform functions on X constitutes Banach unital
S-invariant subalgebra of C(X) which is denoted by C
(X).
By the standard compactness argument we have:
Lemma 5.10. For every compact S-space X we have C
(X) =
C(X).
There exists a natural 1-1 correspondence between S-compacti-
cations of X and closed unital S-invariant subalgebras of C
(X).
In particular, C
(X) determines the maximal S-compactication
S
: X
S
X. These facts and Denition 5.9 are well known for
group actions [38]. For semigroup actions see for example [7, 19].
We say that a bounded function f : X R weakly separates
subsets A, B of X if cl(f(A)) and cl(f(B)) are disjoint. If f(A) =
a ,= b = f(B) for some dierent points a, b R then we simply
say that f separates A and B. For a uniformly continuous function
: R R we have f C
(X) i the composition f C
(X).
It follows that A and B are weakly separated by C
(X) i they are

separated by C
(X). Furthermore we can suppose that a = 0 and

b = 1.
Lemma 5.11. If C
(X) (weakly) separates sets A and B in X,

then cl(A) and cl(B) are -disjoint. In particular, if a -uniform
function separates A and X B, then cl(A)
int(B
c
).
Proof. If not then by Lemma 5.2.1 there exists s
0
S such that
for every U(s
0
) we can choose x
0
X and u
1
, u
2
U with the
property
u
1
x
0
cl(A), u
2
x
0
cl(B).
If C
(X) (weakly) separates A and B then we can assume that

f(cl(A)) = 0, f(cl(B)) = 1 for some f C
(X). For s
0
choose
V N
s
0
such that
[f(v
x) f(v
x)[ < 1 (v
, v
, x) V V X.
On the other hand v
1
x
0
cl(A) and v
2
x
0
cl(B) for certain
v
1
, v
2
V . Therefore we get
[f(v
1
x
0
) f(v
2
x
0
)[ = 1
which contradicts the previous fact.
Remark 5.12. Compactication
S
is an analogue of the standard
maximal (Stone-
Cech) compactication for S-spaces. Clearly, X

is S-Tychono i the maximal S-compactication
is proper i
C
(X) separates points and closed subsets.

6. Equivariant normality of semigroup actions
In this section we study S-normality of monoidal actions apply-
ing S-proximities.
First recall that the topological normality condition can be refor-
mulated in terms of proximity spaces. Namely, a Hausdor topo-
logical space is normal i the relation
A
n
B i cl(A) cl(B) ,=
denes a proximity relation on X. Moreover then this proximity is
exactly the proximity of the Stone-
Cech compactication (Example

3.16) of X (compare Theorem 6.2 below).
In the present section we assume that S is a monoid with a neu-
tral element e. All actions are monoidal, that is, e acts as the
identity mapping.
Denition 6.1. Let S be a monoid. An S-space X is S-normal (or,
equinormal ) if for every pair A, B X of closed -disjoint subsets,
there are open disjoint nbds O
A
N
A
(X) and O
B
N
B
(X), such
that A
O
A
and B
O
B
.
If S is discrete then S-normality is equivalent to the usual (topo-
logical) normality of X.
In the case of group actions, G-normality was introduced in [26,
24]. It is also studied in [21]. Note that this denition (in fact in
a dierent but equivalent form) also appears in a work by Ball and
Hagler [7].
Theorem 6.2. Let X be an S-space. The following are equivalent:
(1) X is S-normal.
(2) The relation
A
B cl(A)
cl(B)
denes an S-proximity on the set X.
1
(3) C
(X) separates closed -disjoint subsets in X.

(4) For every pair A, B X of closed -disjoint subsets, there
are open -disjoint nbds O
A
N
A
(X) and O
B
N
B
(X),
such that A
O
A
and B
O
B
.
1
Recall that cl(A)cl(B) means by Denition 5.1 that there exists s0 S
such that for every neighborhood U of s0 we have U
1
cl(A) U
1
cl(B) = .
(5) ( Urysohns Small Lemma for S-spaces) For every closed
subset A and its open nbd O such that A
O there exists
an open nbd O
1
of A such that A
O
1
and cl(O
1
)
O.
Furthermore, if one of these equivalent conditions is satised then
, the greatest continuous S-proximity on the space X.

Proof. (1) (2): The axioms (P1)(P4) of Denition 3.10 easily
follow from Lemma 5.2 (namely by the assertions (2), (4), (5) and
(6)). For (P1) we use the assumption that S is a monoid. By
Lemma 3.11 instead of (P5) it suces to check (P5.b). This directly
follows by the denition of S-normality because if C is closed and
O is its open nbd then C
O if and only if C O with respect

to
.
Now we check that
is an S-proximity. By Lemma 5.7 we have

to show
(i) every s-translation s : X X is a
-proximity mapping.
(ii) for every pair A

B of
-far subsets A, B in X we have

A
B.
The rst condition is straightforward by Denition 3.17 (C2)
using the inclusion
cl(t
1
C) t
1
cl(C)
for every C X and t S.
If A
B then cl(A)
cl(B). Hence necessarily A
B. This
means that
is an S-proximity.
(2) (3): We show rst that
is a continuous proximity on the

space X. Let a be a point and B a subset in X such that a cl(B).
Since S is a monoid then the denition of
for s
0
:= e implies
that a
B. Thus, top(
) .
By generalized Smirnovs theorem (see Theorem 5.6) the prox-
imity
corresponds to an S-compactication of X. In particu-

lar,
. Let A, B be closed -disjoint subsets in X. Then

they are
-far. Then they are also
-far. There exists a
-
uniform bounded function separating A and B. Now observe that
-uniform function means exactly that it lies in C
(X) (see Re-

mark 5.12).
(3) (4): Easily follows from Lemma 5.11. Take f C
(X)
with f(A) = 0, f(B) = 1 O
A
:= x X : f(x) <
1
3
and dene
O
B
:= x X : f(x) >
2
3
.
(4) (5): Dene B := XO. Then A
B. By (4) we can
choose -disjoint open nbds O
A
N
A
(X) and O
B
N
B
(X), such
that A
O
A
and B
O
B
. Then (XO
B
)
B. Since the open

subsets O
A
and O
B
are disjoint we have cl(O
A
) XO
B
. Therefore
we get cl(O
A
)
B. This means that cl(O

A
)
XB = O.
(5) (1): Use (5) twice.
Thus, we see that all four conditions are equivalent. We already
established above that
is a continuous proximity on the space X

and also
. By the characterization of
if two subsets A
and B are
-far then A and B are separated by a function from

C
(X). Then cl(A) and cl(B) are separated by the same function.
By Lemma 5.11 we get that cl(A)
cl(B). This means that A
B.
Thus
. So we get
, as desired.
Corollary 6.3. For monoidal actions every compact S-space is
S-normal.
Proof. Use the fact that by Lemma 5.10 we have C
(X) =
C(X).
Remark 6.4. Since
we get that
is compatible with the

topology of X i X is S-Tychono i every singleton a and a
closed subset B with a / B are -disjoint. If S := G is a topological
group then the latter condition always holds. Therefore, every G-
normal space is G-Tychono. It is not always true in general for
semigroup actions (Example 7.5).
6.1. Urysohns Theorem for semigroup actions. In this sub-
section we give a dynamical generalization of Urysohns classical
topological result about extensions of functions. We deal with -
uniform functions and monoidal actions. For group actions a similar
result was obtained rst in [24] (see also [2]).
Denition 6.5. Let =
r
: r R be a family of nonempty
open subsets on a topological space X where R is a dense subset
of the closed unit interval I := [0, 1]. We say that is a Urysohn
system (u-system, in short) if it satises the following condition:
(r
1
, r
2
) R R, r
1
< r
2
cl(
r
1
)
r
2
.
For every u-system naturally dened function f
from X to
[0, 1], by:
f
(x) := infr R : x
r
(6.1)
is continuous on X. We call it a u-function of the system . Con-
versely, for every continuous function f : X [0, 1] and a countable
dense set R [0, 1] the family
f
:=
r
= f
1
[0, r) : r R is a
u-system and f
f
= f.
Denition 6.6. Let : S X X be a continuous monoidal
action. Assume that =
r
: r R is a u-system. We say that
is stable (with respect to the action ) if it satises the following
condition:
r
1
< r
2
s
0
S U N
s
0
(S) : U
1
r
1
U
r
2
or
r
1

r
2
(see Lemma 5.2.8).
Theorem 6.7. Let : S X X be a continuous monoidal
action. Then the u-function f
of a u-system =
r
: r R is
in C
(X) if and only if is stable.

Proof.
u-system =
r
: r R is stable. We have to show that the
u-function f
(Formula 6.1) is -uniform at every s

0
S. Let
> 0. Then we need to choose a neighborhood U N
s
0
, such that
f
(s
0
x) < f
(sx) < f
(s
o
x) + : s U, x X. (6.2)
Without restriction of generality we assume that R := Q
2
, the
set of rational dyadic numbers and for every n N dene: Q
(n)
2
=
m
2
n
: m = 0, 1, 2, . . . , 2
n
, then R = Q
2
=

nN
Q
(n)
2
.
Choose n
0
N big enough such that for every t [0, 1] there
exists m
t
= 2
nt
, n
t
1, . . . , n
0
such that:
t <
m
t
2
2
n
0
<
m
t
1
2
n
0
< t <
m
t
+ 1
2
n
0
<
m
t
+ 2
2
n
0
< t + , (6.3)
where
mt1
2
n
0
,
mt2
2
n
0
Q
(n
0
)
2
.
Since is stable system at s
0
, then for every pair r
1
< r
2
(with
r
1
, r
2
Q
(n
0
)
2
) there exists a neighborhood U
r
1
r
2
N
s
0
, such that
U
1
r
1
r
2
r
1
U
r
1
r
2

r
2
. The nite intersection
U :=
U
r
i
r
i+1
, r
i
, r
i+1
Q
(n
0
)
2
is a neighborhood of s
0
. For every r
1
, r
2
Q
(n
0
)
2
we have:
U
1
r
1
U
r
2
. (6.4)
Let x X, y := s
0
x X and f
(y) = t I. By condition 6.3

we have
y m
t
+1
2
n
0
(X m
t
1
2
n
0
). (6.5)
Then x s
1
0
y U
1
m
t
+1
2
n
0
. By Formula 6.4 we have U
1
m
t
+1
2
n
0
U m
t
+2
2
n
0
. Hence x U m
t
+2
2
n
0
. This implies sx m
t
+2
2
n
0
for every
s U. It follows that for every s U we have
f
(sx)
m
t
+ 2
2
n
0
. (6.6)
By condition 6.5 we have y / m
t
1
2
n
0
. Therefore, s
1
0
y
U
m
t
1
2
n
0
= . In particular, x / U m
t
1
2
n
0
. By Formula 6.4,
U
1
m
t
2
2
n
0
Um
t
1
2
n
0
. Hence x / U
1
m
t
2
2
n
0
. We get s U, sx /
m
t
2
2
n
0
. For every s U we can conclude that
f
(sx)
m
t
2
2
n
0
. (6.7)
Since s
0
x = y and f
(s
0
x) = t, from conditions 6.6 and 6.7 we
conclude:
f
(s
0
x) <
m
t
2
2
n
0
f
(sx)
m
t
+ 2
2
n
0
< f
(s
0
x) + s U.
We obtain condition 6.2. This means that the u-function f
is -
uniform at s
0
.
Similar to the proof of Lemma 5.11.

7. Some examples
Example 7.1. Let X be a locally compact Hausdor space. Then
the following relation:
A
a
B cl(A) cl(B) = where cl(A) or cl(B) is compact
denes a compatible proximity on X which denes the (one-point)
Alexandro compactication.
If we have a group action then this proximity is a G-proximity.
It can be veried directly or using a result of de Vries [36], which
states that the one point compactication is always a proper G-
compactication. For semigroup actions it is not true in general.
See for instance [8, Example 3.1.10], [7, Example 7.2], or Example
7.5 below, where the points are not -disjoint.
Example 7.2. Let (X, d) be a metric space. Dene a binary relation
on subsets of X as follows:
A
0
B
A
d
B
or
diam(A) = & diam(B) =
where
d
is the standard metric proximity: A
d
B d(A, B) = 0
(Example 3.14).
For every metric space (X, d),
0
is a separated proximity,
d

0
and top(
0
) = top(
d
). In particular, we can consider the proximity
0
for Banach spaces.
Example 7.3. Let X = V be a Banach space. Then the group of all
linear continuous automorphisms GL(V ) L(V, V) is a topological
group with respect to the operator norm. The natural action
: GL(V ) V V, (L, v) = Lv
is continuous and
0
is a GL(V )-proximity. In particular, V is a
GL(V )-Tychono space.
Example 7.4. Let V be a Banach space. Consider the action of
the topological monoid S := ((V ), ||) (with the operator norm
topology) on the unit ball (B
V
, ||). Then
(1) The norm uniformity
d
on B
V
is an equiuniformity.
(2) The norm proximity
d
on B
V
is an S-proximity.
(3) (S, B
V
) is S-Tychono.
Example 7.5. (S-normal which is not S-Tychonoff). Consider
the (linear) action of the compact multiplicative monoid S :=
([1, 1], ) on X := R. It is easy to see that every pair of points
are not -disjoint. Indeed, let a, b R. For every nbd U of 0
in S choose z R big enough such that
a
z
U and
b
z
U.
Then clearly, z U
1
a U
1
b ,= . Therefore, by Lemma 5.11,
every f C
(X) is a constant function. The corresponding prox-

imity
on X is trivial (i.e., the proximity of the trivial compacti-

cation).
Recall in contrast that for group action case every G-normal is
necessarily G-Tychono (Remark 6.4).
Remark 7.6. From Example 3.15 we know that the relation
R
dened by
A
R
B U N
e
(G) : UA B ,=
is a compatible proximity on G/H for every topological group G
and its subgroup H. In fact it is a G-proximity for the left action
: GG/H G/H. The proximity
R
is separated if and only if
H is closed in G.
8. Actions of topological groups
Lemma 8.1. Let X be a G-space with respect to the continuous
action : G X X of a topological group G. Let N
e
(G) be the
set of all nbds of the identity e G.
(1) The relation
(Denition 5.1) can be described in a sim-

pler way:
A
B if and only if UA B ,= for every U N

e
(G). (8.1)
Furthermore
is a proximity on the set X and top(
)
top(X) holds.
In particular we have
(a) A and B are -disjoint if and only if UA B = for
some U N
e
(G).
(b) A
D if and only if UA D for some U N

e
(G).
(2) The family of entourages
U
UNe(G)
where
U
:= (u
1
x, u
2
x) X X[ x X, u
1
, u
2
U
is a base of a uniformity
on the set X such that the

corresponding proximity
is exactly
.
Proof. Straightforward using some elementary properties of N
e
(G).

Lemma 8.2. Let f : X R be a continuous bounded function on
a G-space X. The following are equivalent :
(1) f : X R is -uniform.
(2) For every > 0 there exists a nbd U N
e
(G) such that
[f(ux) f(x)[ < (u, x) U X.
(3) f : X R is
-uniformly continuous.
(4) f : X R is
-uniform mapping.
Proof. For (3) (4) use Fact 3.18 and Lemma 8.1.2.
Denition 8.3. Let A be a (not necessarily G-invariant) subspace
of a G-space X.
(1) The subspace proximity on A X induced by
is de-
noted by
A
.
(2) Let f : A R be a bounded continuous function on A. We
say that f is -uniform if it is a proximal mapping on the
subspace (A,
A
). Precisely this means that for every > 0

there exists a nbd U N
e
(G) such that
[f(ua) f(a)[ < (u, a, ua) U A A.
Remark 8.4. If A is a G-invariant subspace of X then Denitions
8.3.2 and 5.9 agree.
Theorem 8.5. Let X be a G-space such that X is a Tychono
space. The following are equivalent:
(1) X is G-normal (that is for every pair of -disjoint closed
subsets A and B there exist disjoint open nbds O
1
and O
2
such that A
O
1
and B
O
2
).
(2) For every pair of -disjoint closed subsets A and B there
exist -disjoint open nbds O
1
and O
2
such that A
O
1
and B
O
2
.
(3) For every pair of -disjoint closed subsets A and B there
exist -disjoint nbds O
1
and O
2
.
(4) The relation
A
B Ucl(A) cl(B) ,= U N
e
(G)
is a proximity on X.
(5) C
(X) separates closed -disjoint subsets in X.

(6) Every -uniform function f : A R (in the sense of De-
nition 8.3.2) on a closed subset A X is a restriction of a
-uniform function F : X R.
Furthermore, if one of these equivalent conditions is sat-
ised then
is a compatible G-proximity on the space X

and
.
Proof. (1) (4) (5) follows from Theorem 6.2 taking into ac-
count Lemma 8.1.
(2) (3): Is trivial.
(3) (1): There exist U(e) such that UO
1
UO
2
= , where
O
1
and O
2
are nbds of A and B respectively. We can suppose
that O
1
and O
2
are open (passing to the interiors if necessary).
Take a symmetric nbd V (e) s.t. V V U. Dene O
1
:= V O
1
and O
2
:= V O
2
. Then still O
1
and O
2
are -disjoint (because
V O
1
V O
2
= ) and also A
1
and B
2
by Lemma 8.1
(because, V A V O
1
= O
1
and V B V O
2
= O
2
).
(4) (6): Use Fact 3.18.3 taking into account Remark 8.4.
(6) (5): Let A and B be -disjoint closed subsets of X. Dene
the function f : A B [0, 1] by f(x) = 0 for every x A and
f(x) = 1 for every x B. It is easy to see that f is a -uniform
function on A B in the sense of Denition 8.3.2. By assumption
(6), f is a restriction of a -uniform function F : X R. The
latter function clearly separates A and B. This proves (5).
Finally the compatability of
follows from the fact that every

singleton a and a closed subset B with a / B are -disjoint for
every G-space (Remark 6.4).
Some versions of the Urysohn lemma for G-spaces appears in [15,
Theorem 3.9] and [2].
Proposition 8.6. Every G-normal G-space X is G-Tychono.
Proof. By Theorem 8.5 there exists a compatible G-proximity on
X.
Lemma 8.7. [24] Suppose that a G-space X, as a topological space,
is normal. The following are equivalent:
(1) Every pair of -disjoint closed subsets A and B in the G-
space X there exists U N
e
(G) such that cl(UA)cl(UB) =
.
(2) X, as a G-space, is G-normal.
Proof. (1) (2): By the normality there exist disjoint nbds O
1
and O
2
of the closed disjoint subsets cl(UA) and cl(UB). Then
A
O
1
and B
O
2
.
(2) (1): Use Theorem 6.2.5 and Lemma 8.1.1.
According to a fundamental result of de Vries [36] every
Tychono G-space is G-Tychono for any locally compact group
G. If X is normal then we can prove a stronger result.
Proposition 8.8. [26] Let G be a locally compact group and X is
a G-space such that X (as a topological space) is normal. Then
(1) X is G-normal.
(2) For every closed G-subspace A in X the compactications
A
G
A and A cl(A)
G
X are equivalent.
Proof. (1): Let A and B be -disjoint closed subsets. Then there
exists a nbd U(e) such that UAUB = . We can suppose that U
is compact. Then the subsets UA and UB are closed. We can now
apply Lemma 8.7.
(2): By (1) and Theorem 8.5.6 every -uniform function on A
can be extended to a -uniform function on X. This implies that
the G-compactications A
G
A and A cl(A)
G
X are
equivalent (both correspond to the same algebra C
(A)).
Proposition 8.8.2 answers a question of Yu.M. Smirnov (private
communication).
Proposition 8.9. [26] For every Hausdor topological group G and
every closed subgroup H the corresponding coset G-space G/H is
G-normal.
Proof. Observe that
is exactly the right uniformity on G/H and
is the proximity associated to
(see Remark 7.6). It follows

that assertion (4) of Theorem 8.5 is satised.
Since every G-normal is G-Tychono Proposition 8.9 strengthens
a result of de Vries [36] which asserts that any coset G-space G/H
is G-Tychono. There exists a G-normal G-space X which is not
normal as a topological space. Indeed take a topological group G
which is not normal and consider the G-space G under the left
action.
Corollary 8.10. The following actions, being coset spaces, are
equinormal:
(1) (U(H), S
H
), where S
H
is the unit sphere of a Hilbert space
H;
(2) (Is (U), U) where Is (U) is the isometry group of the Urysohn
space
2
U with the pointwise topology);
(3) (GL(V ), V 0) for every normed space V (see [20]);
(4) (GL(V ), P
V
) for every normed space V and its projective
space P
V
.
It follows in particular by (4) that P
V
is GL(V )-Tychono. This
fact was well known among experts and easy to prove using equiu-
niformities (cf. e.g. Pestov [31]).
By [20, Proposition 2.3], (GL(V ), V 0) is the coset space for
every normed space V . In order to see that the projective space P
V
for every normed space V is the coset space of the group GL(V )
observe that the projective space is an open G-quotient of the G-
space S
V
, the sphere of V . It suces to show that S
V
is a coset
G-space with respect to the action
(g, v)
g(v)
[[g(v)[[
.
As in [20, Proposition 2.3] it is easy to see that for every <
1
2
the set P
z is a nbd of z in S
V
, where
P
:= A
f,y
[ [[f[[ 1, [[y[[ <
and A
f,y
(x) := x + f(x)y for every functional f V
and y V .
Observe that |A
f,y
I| = |f| |y|.
Question 8.11. [18] Is it true that the following (G-Tychono)
actions are G-normal: (U(
2
),
2
), (Is (
p
), S
p
)), p > 1, (p ,= 2)?
The following concrete example shows that there exist G-spaces
X admitting a G-invariant metric (hence X is G-Tychono by Fact
4.3.3) such that X is not G-normal.
Example 8.12. [24, page 60] The action of the group Q of rational
numbers on R by translations is G-Tychono but not G-normal.
The idea of this example leads to a generalized version.
2
For the denition and properties of the Urysohn space see for example [31].
Fact 8.13. [21, Proposition 2.5] Let G be an arbitrary topological
group which is not Raikov complete. Then there exists a normal
G-space X (of weight w(X) = w(G)) which is not G-normal.
One can characterize locally compact groups in terms of G-
normality.
Fact 8.14. [21, Theorem 5.2] For every topological group G the
following are equivalent:
(1) Every normal G-space is G-normal.
(2) G is locally compact.
Recall that X is weakly G-normal (see [26, 24, 21]) if every pair
of -disjoint closed G-invariant subsets in X can be separated by
a function from C
(X).
Question 8.15. Is every second countable G-space weakly G-normal
for the group G := Q of rational numbers?
If not, then by [21, Theorem 3.2] one can construct for G := Q
a Tychono G-space X which is not G-Tychono. That is, it will
followthat Q is not a V-group (resolving the Question [18, Question
2.3]).
References
1. E. Akin, Recurrence in topological dynamics: Furstenberg families and Ellis
actions, University Series in Mathematics, 1997.
2. N. Antonyan, An intrinsic characterization of G-pseudocompact spaces,
Houston J. Math. 33:2(2007), 519 - 530.
3. S. Antonyan, The classication of bicompact G-extensions by means of rings
of equivariant maps, Doklady Acad. Sci. Arm. SSR, 69(1979), 260-264.
4. S. Antonyan and M. Sanchis, Extension of localy pseudocompact group ac-
tions, Annali di Matematica 181(2002), 239-246.
5. S. Antonyan and Yu.M. Smirnov, Universal objects and compact extensions
for topological transformation groups, Dokl. Akad. Nauk SSSR, 257(1981),
521-526.
6. C.E. Aull and R. Lowen (eds.), Handbook of the History of General Topol-
ogy, Kluwer Academic Publishers, 1998, vol.2.
7. R. Ball and J. Hagler, Real-Valued functions on ows, Preprint, 1996.
8. J.F. Berglund, H.D. Junghenn and P. Milnes, Analysis on Semigroups, Wi-
ley, New York, 1989.
9. R.B. Brook, A Construction of the Greatest ambit, Math. System Theory,
4:4(1970), 343-348.
10. V.A. Chatyrko and K.L. Kozlov, On G-compactications, Math. Notes, 78,
No. 5, 2005, p.649-661.
11. V.A. Efremovich, The geometry of proximity I, Mat. Sbornik N.S. 31 (73),
189-200 (1952).
12. R. Engelking, General Topology, Heldermann Verlag, Berlin, 1989.
13. I.S. Gal, Proximity relations and precompact structures, Proc. Kon. Ned.
Akad. Wetenschappen, 62(1959), 304-326.
14. I.M. Gelfand and A.N. Kolmogoro, On the rings of continuous functions
on topological spaces, Doklady Acad. Sci. USSR, 22(1939), 11-15.
15. E. Glasner and B. Weiss, Spatial and non-spatial actions of Polish groups,
Ergodic Theory Dyn. Syst., 25(2005), 1521-1538.
16. J.R. Isbell, Uniform Spaces, Mathematical Surveys 12 AMS, Providence,
Rhod Island, 1964.
17. H. Ludescher and J. de Vries, A sucient condition for the existence of
a G-compactication, Nederl. Akad. Wetensch Proc. ser. A 83, 1980, p.
263-268.
18. M. Megrelishvili, Topological transformation groups: selected topics. In:
Open Problems In Topology (Second edition, Elliott Pearl, ed.), Elsevier
Science, 2007, p. 423-438.
19. M. Megrelishvili, Compactications of semigroups and semigroup actions,
Topology Proceedings, 31:2(2007), 611-650.
20. M. Megrelishvili, G-Minimal topological groups, In: Abelian Groups, Mod-
ule Theory and Topology, Lecture Notes in Pure and Applied Algebra,
Marcel Dekker, 201(1998), 289-300.
21. M. Megrelishvili and T. Scarr, Constructing Tychono G-spaces which are
not G-Tychono, Topology and its Applications, 86:1(1998), 69-81.
22. M. Megrelishvili, Compactication and factorization in the category of G-
spaces, Categorical Topology and its Relation to Analysis, Algebra and
Combinatorics, Prague, 1988, p. 220-237.
23. M. Megrelishvili, A Tychono G-space which has no compact G-extensions
and G-linearizations, Russ. Math. Surv, 43(1988), 145-146.
24. M. Megrelishvili, Uniformity and Topological Transformation Groups, Ph.
D. Dissertation, Tbilisi State University, 1985 [in Russian].
25. M. Megrelishvili, Equivariant completions and compact extensions, Bull. Ac.
Sc. of Georgian SSR, 1984, no. 115, p. 21-24 [in Russian].
26. M. Megrelishvili, Equivariant normality, Bull. Ac. Sc. Georgian SSR,
111:1(1983), 43-46 [in Russian].
27. J. van Mill, Homogeneous spaces and transitive actions by Polish groups,
Israel J. Math. 165(2008), 133-159.
28. J. van Mill, On the G-compactications of the rational numbers, to appear
in Monatsh. Math.
29. S.A. Naimpally and B.D. Warrack, Proximity spaces, Cambridge University
Press, 1970.
30. R. Palais, The classication of G-spaces, Memories Amer. Math. Soc., no.
36, 1960.
31. V. Pestov, Dynamics of innite-dimensional groups, the Ramsey-Dvoretzky-
Milman phenomenonAmerican Math. Society, University Lecture Series 40,
2006.
32. W. Roelcke and S. Dierolf, Uniform structures on topological groups and
their quotients, Mc Graw-hill, New York, 1981.
33. P. Samuel, Ultralters and compactications of uniform spaces, Trans.
Amer. Math. Soc., 64(1948), 100-132.
34. Yu.M. Smirnov, On proximity spaces in the sense of V.A. Efremovich, Dok-
lady Akad. Nauk SSSR (N.S.) 84(1952), 895-898.
35. V.V. Uspenskij, Topological groups and Dugundji compacta, Math. Sb.,
67:2(1990), 555-580.
36. J. de Vries, Can every Tychono G-space equivariantly be embedded in a
compact Hausdor G-spaces?, Math Centrum, Amsterdam, Afd. Zuivere
Wisk. 36, 1975.
37. J. de Vries, Equivariant embeddings of G-spaces, in: J. Novak (ed.), General
Topology and its Relations to Modern Analysis and Algebra IV, Part B,
Prague, 1977, 485-493.
38. J. de Vries, On the existence of G-compactications, Bull. Acad. Polon. Sci.
Math., 26(1978), 275-280.
Department of Mathematics, Bar-Ilan University, 52900 Ramat-
Gan, Israel
E-mail address: [email protected]
E-mail address: [email protected]
URL: http://www.math.biu.ac.il/
megereli

Topology Proceedings: Semigroup Actions: Proximities, Compactifications and Normality

Uploaded by

Copyright:

Available Formats

Topology Proceedings: Semigroup Actions: Proximities, Compactifications and Normality

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Topology Proceedings: Semigroup Actions: Proximities, Compactifications and Normality

Uploaded by

Copyright:

Available Formats

TOPOLOGY

B if A and B are not -disjoint. We

B of -far subsets A, B in X and every s

(X) such that f(A) = 0, f(B) = 1.

, the greatest continuous S-proximity on the S-space X.

= top() := O X : x O such that B(x, ) O

as the quotient set X/ with respect to the equivalence

, x [x] the natural map. Now the

is a base of some Hausdor uni-

). The resulting composition map

on X (the corresponding initial unifor-

is also precompact. It is also easy (Remark 3.4) that

) (the associated Hausdor uniformity) and the usual com-

). Denote by Y the compact Haus-

with its unique uniformity

U := (xH, yH) G/H G/H : xy

is totally bounded and =

the induced uniformity of . Its

B f C(X) such that f(A) = 0 and f(B) = 1

B comes from the Stone-

B if and only if there exists

be the induced uniformity of and (

the corresponding initial proximity on X dened via

(Remark 3.13). Now com-

. This leads to a description of compactications in terms of prox-

on X gives rise to a natural one-to-one order

on the weak star compact unit

of the dual space V

is continuous (see [19, Lemma

we can choose such

= . Assuming the contrary

= . On the other hand x U

) be the associated uniform space

is a base of the uniformity

B if A and B are not -disjoint.

is a proximity on the set X for every topo-

B of -far subsets A, B in X and

B, i.e. by Denition 3.12 there exists an entourage

-far. This means that U

on X is equiuniform, too. Then the induced prox-

B of -far subsets A, B in X we have

B (that is for every s

on X (Remark 3.13). We have to show that (X,

-uniform by (i)). Hence

-bounded. If not then there

(see for example [12,

is an equiuniformity on X. Now Propo-

(X) determines the maximal S-compactication

(X) i the composition f C

(X) i they are

(X). Furthermore we can suppose that a = 0 and

(X) (weakly) separates sets A and B in X,

(X) (weakly) separates A and B then we can assume that

Cech) compactication for S-spaces. Clearly, X

(X) separates points and closed subsets.

Cech compactication (Example

(X) separates closed -disjoint subsets in X.

, the greatest continuous S-proximity on the space X.

O if and only if C O with respect