ON METRIZABLE ENVELOPING SEMIGROUPS

ELI GLASNER, MICHAEL MEGRELISHVILI, AND VLADIMIR V. USPENSKIJ

Abstract. When a topological group G acts on a compact space X, its enveloping
semigroup E(X) is the closure of the set of g-translations, g G, in the compact
space X
X
. Assume that X is metrizable. It has recently been shown by the rst two
authors that the following conditions are equivalent: (1) X is hereditarily almost
equicontinuous; (2) X is hereditarily non-sensitive; (3) for any compatible metric
d on X the metric d
G
(x, y) := supd(gx, gy) : g G denes a separable topol-
ogy on X; (4) the dynamical system (G, X) admits a proper representation on an
Asplund Banach space. We prove that these conditions are also equivalent to the
following: the enveloping semigroup E(X) is metrizable.
1. Introduction
A dynamical system, or a G-space, in this paper is a compact space X (compact
will mean compact and Hausdor) on which a topological group G acts continuously.
We denote such a system by (G, X). For g G the g-translation (or g-shift) is the
self-homeomorphism x gx of X. If a nonempty subset Y X is G-invariant, i.e. if
Y is closed under g-shifts, then Y is a G-subspace. The enveloping semigroup (or Ellis
semigroup) of (G, X) is the closure of the set of g-shifts (g G) in the compact space
X
X
, equipped with the product topology. Even for simple dynamical systems on a
compact metric space the enveloping semigroup may be non-metrizable. For example,
for the classical Bernoulli shift (with G := Z) on the Cantor space X = 0, 1
Z
,
the enveloping semigroup E(X) is homeomorphic to N (see [18, Exercise 1.25]).
If X is the unit interval [0, 1] and G = H
+
[0, 1] is the group of all orientation-
preserving homeomorphisms, then E(X) is the non-metrizable space of increasing and
end-points-preserving self-maps of [0, 1]. If X is a compact manifold without boundary
of dimension > 1 and G = Homeo (X) is the group of all self-homeomorphisms of X,
then E(X) is X
X
.
On the other hand, if G is an equicontinuous group of homeomorphisms of a com-
pact metric space X, then E(X) consists of continuous self-maps of X and hence is
metrizable. The same is true, more generally, if (G, X) is WAP (= Weakly Almost
Date: May 14, 2007.
Key words and phrases. almost equicontinuous, Asplund space, Baire 1 function, barely contin-
uous, enveloping semigroup, locally equicontinuous, Radon-Nikod ym system, Rosenthal compact,
semigroup compactication, sensitive dependence, weakly almost periodic, weak

-topology.
2000 Mathematics Subject Classication: Primary: 54H20. Secondary 22A25, 22F05, 37B05,
46B10, 46B22, 54H15.
1
2 E. GLASNER, M. MEGRELISHVILI, AND V.V. USPENSKIJ
Periodic). Recall that a function f C(X) is weakly almost periodic if its G-orbit

g
f : g G lies in a weakly compact subset of the Banach space C(X), and (G, X)
is WAP if every f C(X) is WAP. A dynamical system (G, X) is WAP if and only
if E(X) consists of continuous self-maps of X [12, 14].
A generalization of WAP systems, called RadonNikodym (RN for short) systems,
was studied in [25, 20]. To dene this notion, note that with every Banach space V
one can associate a dynamical system S
V
= (H, Y ) as follows: H = Iso (V ) is the
group of all linear isometries of V onto itself, equipped with pointwise convergence
topology (or the compact-open topology, the two topologies coincide on H), and Y
is the unit ball of the dual space V

, equipped with the weak

-topology. The action

of H on Y is dened by g(v) = (g
1
(v)), g H, Y , v V . The continuity
of this action can be easily veried. A representation of a dynamical system (G, X)
on a Banach space V is a homomorphism of (G, X) to S
V
= (H, Y ), that is, a pair
of continuous maps (h, ), h : G Iso (V ) and : X Y , such that h is a group
homomorphism and (gx) = h(g)(x) for all g G and x X. A representation is
proper if is a topological embedding.
A compact metric G-space X is WAP if and only if (G, X) admits a proper rep-
resentation on a reexive Banach space [25, Corollary 6.10], [20, Theorem 7.6(1)]. A
dynamical system is RadonNikodym (RN) if it admits a proper representation on an
Asplund Banach space [25, Denition 3.10], [20, Denition 7.5.2]. (If G = 1, we get
the class of RadonNikod ym compact spaces in the sense of Namioka [29].) Recall
that a Banach space V is Asplund if for every separable subspace E V the dual E

is separable. Reexive spaces and spaces of the form c

0
() are Asplund. About the
history and importance of Asplund spaces see for example [7, 8, 16].
Now assume that X is a metrizable compact space. One of the main results of [20]
was a characterization of RN-systems as those which are close to equicontinuous.
To give a precise statement we recall a few denitions from [21, 3, 25, 20].
Let d be a compatible metric on X. We say that (G, X) is non-sensitive if for
every > 0 there exists a non-empty open set O X such that for every g G
the set gO has d-diameter < . (This property does not depend on the choice of
a compatible metric d.) A system (G, X) is hereditarily non-sensitive (HNS) if all
closed G-subsystems are non-sensitive.
A system (G, X) is equicontinuous at p X if for every > 0 there exists a
neighborhood O of p such that for every x O and every g G we have d(gx, gp) < .
A system is almost equicontinuous (AE) if it is equicontinuous at a dense set of points,
and hereditarily almost equicontinuous (HAE) if every closed subsystem is AE.
Denote by Eq

the union of all open sets O X such that for every g G the
set gO has diameter < . Then Eq

is open and G-invariant. Let Eq =

>0
Eq

.
Note that a system (G, X) is non-sensitive if and only if Eq

,= for every > 0, and

(G, X) is equicontinuous at p X if and only if p Eq. Suppose that Eq

is dense
ON METRIZABLE ENVELOPING SEMIGROUPS 3
for every > 0. Then Eq is dense, in virtue of the Baire category theorem. It follows
that (G, X) is AE.
If (G, X) is non-sensitive and x X is a transitive point that is, Gx is dense
then for every > 0 the open invariant set Eq

meets Gx and hence contains Gx. Thus

x Eq. If, in addition, (G, X) is minimal (= all points are transitive), then Eq = X.
Thus minimal non-sensitive systems are equicontinuous (see [5], [21, Theorem 1.3],
[2], or [20, Corollary 5.15]).
Theorem 1.1 ([20, Theorem 9.14]). For a compact metric G-space X the following
conditions are equivalent:
(1) X is RN;
(2) X is HNS;
(3) X is HAE;
(4) every nonempty closed G-subspace Y of X has a point of equicontinuity;
(5) for any compatible metric d on X the metric d
G
(x, y) := sup
gG
d(gx, gy)
denes a separable topology on X.
It was proved in [20] that the equivalent conditions of Theorem 1.1 imply that the
enveloping semigroup E(X) must be of cardinality 2

. In fact, it was established

in [20, Theorem 14.8] that E(X) is Rosenthal compact (see the rst paragraph of
Section 6 for a denition), and the question was posed whether this conclusion can be
strengthened to E(X) is metrizable. This question was repeated in [26, Question
7.7]. The aim of the present paper is to answer this question in the armative.
Moreover, it turns out that metrizablity of E(X) in fact is equivalent to the conditions
of Theorem 1.1:
Theorem 1.2. Let X be a compact metric G-space. The following conditions are
equivalent:
(1) the dynamical system (G, X) is hereditarily almost equicontinuous (HAE);
(2) the dynamical system (G, X) is RN, that is, admits a proper representation
on an Asplund Banach space;
(3) the enveloping semigroup E(X) is metrizable.
Note that, as the enveloping semigroup depends only on the image of Gin Homeo (X),
we can deduce that for metrizable ows the RN property likewise depends only on the
image of G in Homeo (X), and is therefore independent of the topology of G. Of
course we can obtain these observations also from Theorem 1.1 since HNS (even for
non-metrizable ows) has the same property.
After providing a few facts from general topology in Section 2, we prove in Sec-
tion 3 the implication (2) (3) of Theorem 1.2, in other words, that for every RN
compact metric G-space X the enveloping semigroup E(X) is metrizable. We prove
the implication (3) (1) in Section 4. Since (1) and (2) are known to be equivalent
(Theorem 1.1), this proves Theorem 1.2. The implication (1) (3) is thus proved
4 E. GLASNER, M. MEGRELISHVILI, AND V.V. USPENSKIJ
via representations on Banach spaces; we give an alternative direct proof in Section 5.
Some corollaries of the main theorem are discussed in Section 6.
We thank the anonymous referee for a careful reading of the paper and for many
useful remarks.
2. General topology: prerequisites
A subset of a topological space is meagre if it can be covered by a countable family
of closed sets with empty interior. A space is Baire if every meagre set has empty
interior, or, equivalently, if the intersection of any countable family of dense open
sets is dense. Let us say that a (not necessarily continuous) function f : X Y is
Baire 1 if the inverse image of every open set in Y is F

(= the union of countably

many closed sets) in X. According to this denition, Baire 1 functions need not be
limits of continuous functions. However, if the target space Y is metrizable (or, more
generally, perfectly normal), then the limit of every pointwise converging sequence of
continuous functions is Baire 1:
Proposition 2.1 (R. Baire). If Y is a metric space and f
n
: X Y is a sequence
of continuous functions converging pointwise to f : X Y then f is Baire 1.
Proof. Let U Y be open. There is a sequence F
n
of closed sets such that U =

F
n
=

Int F
n
, where Int denotes the interior. Then f
1
(U) is the union over n
and k of the closed sets

i>n
f
1
i
(F
k
).
Proposition 2.2 (R. Baire). Let f : X Y be Baire 1. If X is Baire and Y is
separable and metrizable then there exists a dense G

-subset A of X such that f is

continuous at every x A.
Proof. Let U
n
: n be a countable base for Y . Write f
1
(U
n
) =

k
F
nk
, where
each F
nk
is closed, and consider the union D of the boundaries of all the F
nk
s. Then
D is meagre, and it is easy to see that f is continuous at every point of the dense
G

-set A = X D.
Proposition 2.3. Let f : X Y be a (not necessarily continuous) function from a
topological space X to a separable metric space Y . Suppose that the inverse image of
every closed ball in Y is closed in X. Then f is Baire 1.
Proof. Every open set U in Y is the union of a countable family of closed balls, hence
f
1
(U) is F

.
We denote by C(X, Y ) the space of continuous maps from X to Y , equipped
with the compact-open topology. If X is compact and Y is metric, this topology is
generated by the sup-metric. If X is compact metrizable then the group Homeo (X)
C(X, X) of all self-homeomorphisms of X is a separable and metrizable topological
group.
ON METRIZABLE ENVELOPING SEMIGROUPS 5
Proposition 2.4. Let X be Baire, L separable metrizable, K compact metrizable, Y
dense in K. If f : X C(K, L) is a (not necessarily continuous) function such that
for every y Y the function x f(x)(y) from X to L is continuous, then there
exists a dense G

-subset A of X such that f is continuous at every x A.

The same result is true under the following assumptions: Y = K, K is compact
but not necessarily metrizable, X is regular and strongly countably complete in the
sense of Namioka [28]. For an easier proof of Namiokas theorem that works under
less restrictive assumptions, see [30].
Proof. Equip C = C(K, L) with the sup-metric using a compatible metric d on L.
Then C is a separable metric space, and the inverse image under f of the closed ball
of radius r > 0 centered at h C is closed, being the intersection of the closed sets
x X : d(f(x)(y), h(y)) r, y Y . Thus Propositions 2.2 and 2.3 apply.
A function f : X Y is barely continuous if for every closed non-empty A X
the restriction f[A has a point of continuity. (This pun originates in a 1976 paper
of E. Michael and I. Namioka, [27].) It is a classical fact (contained in R. Baires
Thesis, 1899) that a function between Polish spaces is barely continuous if and only
if it is Baire 1 (see e.g. [23, Theorem 24.15]). If f : X Y is an onto barely
continuous function between metric spaces and X is separable, then so is Y [27] (see
also [20, Lemma 6.5 and Proposition 6.7]). We will need later a G-space version of
this statement.
If X and Y are G-spaces, let us say that f : X Y is G-barely continuous if
the restriction f[A has a point of continuity for every G-invariant closed non-empty
subset A X. A G-map between G-spaces is a map commuting with the action of G.
Proposition 2.5. Let X and Y be metric spaces. Suppose that a (discrete) group G
acts on X by homeomorphisms and on Y by isometries. Let f : X Y be an onto
G-map. If f is G-barely continuous and X is separable, then Y is separable.
Proof. Pick > 0. Let be the collection of all open subsets U of X such that f(U)
can be covered by countably many sets of diameter . Then is G-invariant and
closed under countable unions. Since there exists a countable subfamily such
that = , the family has a largest element, namely V = . Let A = X V .
If a A is a point of continuity of f[A, there exists an open set O X such that
a O and f(O A) has diameter . Then f(O V ) = f(O A) f(V ) can be
covered by countably many sets of diameter . Thus O V , in contradiction
with the fact that O meets the complement of = V . We have proved that f[A
has no points of continuity. Since A is closed and G-invariant, and f is G-barely
continuous, it follows that A is empty.
Thus X = V , and Y can be covered by countably many sets of diameter .
Since was arbitrary, Y is separable.
6 E. GLASNER, M. MEGRELISHVILI, AND V.V. USPENSKIJ
Proposition 2.6. The Banach dual V

of a non-separable Banach space V is non-

separable.
Proof. Construct a transnite sequence x

: <
1
of unit vectors in V such that
for each <
1
the vector x

does not belong to the closed linear space L

spanned
by the vectors x

, < . For every <

1
nd a functional f

such that
f
a
L

and f

(x

) = 1. All the pairwise distances between distinct f

s are 1. It
follows that V

, considered with its norm topology, is not separable.

Proposition 2.7. Let f : X Y be a continuous onto map between compact spaces.
If X is metrizable, then so is Y .
Proof. A compact space K is metrizable if and only if it has a countable base if and
only if the Banach space C(K) is separable. Note that C(Y ) is isometric to a subspace
of C(X) and hence is separable if C(X) is separable.
Alternatively, one can use Arhangelskiis theorem on coincidence of the network
weight and weight in compact spaces [15, Theorem 3.1.19]. This approach yields a
stronger result: a compact space is metrizable if it is the image under a continuous
mapping of any space with a countable base, compact or not.
3. Proof of Theorem 1.2: Part 1
In this section we prove that for every RN compact metric G-space X the enveloping
semigroup E(X) is metrizable. Recall that X being RN means that (G, X) has a
proper representation on an Asplund Banach space.
For a Banach space V we denote by S
V
the dynamical system (Iso (V ), Y ), where
Y is the unit ball of the dual space V

, equipped with the weak

topology.
We rst prove the special case of the implication (2) (3) of Theorem 1.2, when
the dynamical system is of the form S
V
, where V is a Banach space with a separable
dual:
Proposition 3.1. Let V be a Banach space with a separable dual, G = Iso (V ), Y
the compact unit ball of V

with the weak

topology, considered as a G-space. Then

the enveloping semigroup E(Y ) is metrizable.
Proof. Let K be the set of all linear operators of norm 1 on the Banach space V

.
Consider the topology on K inherited from the product (V

)
V

, where each factor

V

is equipped with the weak

topology. Then K is compact, being a closed subset

of the product

fV

|f| Y . We claim that K is metrizable. Indeed, V is separable

(Proposition 2.6), hence Y is metrizable, and so is each ball rY , r > 0. If C is a norm-
dense countable subset of V

, the restriction A A[C denes a homeomorphism of K

onto a subspace of the product

fC
|f| Y of countably many metrizable compacta.
This proves our claim that K is metrizable.
Restricting each operator A K to Y , we obtain a homeomorphism of K with a
compact subset L of Y
Y
. The enveloping semigroup E(Y ) is the closure of the set
T

[Y : T G in L. Since K is metrizable, so are L and E(Y ).

ON METRIZABLE ENVELOPING SEMIGROUPS 7
Proposition 3.2. Let G be a separable topological group, X a compact metric G-
space. If X is RN then (G, X) has a proper representation on a Banach space with a
separable dual.
Proof. There exists a proper representation (h, ) : (G, X) S
V
= (H, Y ) for some
Asplund V . Since (X) is metrizable, there exists a countable subset A V that
separates points of (X). Let W be the closed linear subspace of V spanned by the
union of G-orbits of all points of A. Then W is separable (note that the G-orbit of
any point v V is separable, being a continuous image of G), G-invariant, and the
restriction map V

is one-to-one on (X). It follows that (G, X) admits a

proper representation on W. Since V is Asplund and W is separable, the dual of W
is separable.
Proposition 3.3. Let X be a compact G-space. Suppose that G
1
is a subgroup of G
and X
1
is a closed G
1
-invariant subset of X. If E(G, X) is metrizable then E(G
1
, X
1
)
also is metrizable.
Proof. Consider the dynamical systems D = (G, X), D
1
= (G
1
, X), and D
2
=
(G
1
, X
1
). The enveloping semigroup E(D
1
) is a subspace of E(D), and there is a
natural onto map E(D
1
) E(D
2
). If E(D) is metrizable, then so are E(D
1
) and
E(D
2
) (Proposition 2.7).
We now show that for every RN compact metric G-space X the enveloping semi-
group E(G, X) is metrizable. Since E(G, X) depends only on the image

G of G in
Homeo (X), we may assume that G is separable. One way to see this is to choose a
countable dense subset

A

G, choose a set of representatives A = a G : a
a, a

A, and let G
0
be the countable subgroup of G generated by A. Then clearly
(G
0
, X) is RN and E(G, X) = E(G
0
, X).
By Proposition 3.2 there exists a proper representation (h, ) : (G, X) S
V
=
(H, Y ) for some Banach space V with a separable dual. In virtue of Proposition 3.1,
the enveloping semigroup of the system S
V
= (H, Y ) is metrizable. Consider the dy-
namical system (h(G), (X)). Its enveloping semigroup is metrizable by Proposition
3.3. It remains to note that E(h(G), (X)) and E(G, X) are isomorphic.
4. Proof of Theorem 1.2: Part 2
Let X be a compact metric G-space such that E(X) is metrizable. We prove that
X is HAE (= Hereditarily Almost Equicontinuous).
For every closed G-subsystem Y of X the enveloping semigroup E(Y ) is metrizable,
being a continuous image of E(X). Thus it suces to prove that X is AE, that is,
that the system (G, X) is equicontinuous at a dense set of points.
Consider the metric space C = C(E, X) of all continuous maps from E = E(X)
to X, equipped with the sup-metric. For each x X let x

C be the evaluation map

dened by x

(e) = e(x), e E. It is easy to see that the map f : X C dened by

8 E. GLASNER, M. MEGRELISHVILI, AND V.V. USPENSKIJ
f(x) = x

is continuous at a point x X if and only if (G, X) is equicontinuous at x.

Thus we must prove that f has a dense set of points of continuity. This follows from
Proposition 2.4, where K = E, L = X and Y K is the set of all G-translations.
5. An alternative proof of the implication (1) (3) in Theorem 1.2
The implication (1) (3) in Theorem 1.2: if X is metric and HAE, then E(X) is
metrizable was obtained in an indirect way, via representations on Banach spaces.
In this section we give a direct proof in the spirit of Section 4.
Consider the same evaluation map f : X C(E, X) as in Section 4. The as-
sumption that X is HAE implies that for every non-empty closed G-invariant subset
Y of X the restriction f[Y has a point of continuity. In other words, f is G-barely
continuous in the sense of Section 2.
Consider the action of G on E given by ge(x) = e(g
1
x) (g G, e E, x X),
and the action of G on C(E, X) given by gh(e) = h(g
1
e) (g G, h C(E, X),
e E). (We consider here G as a group without topology; these actions need not be
continuous if G is considered with its original topology.) Then G acts on C(E, X)
by isometries. The evaluation map f : X C(E, X) is a G-map. Therefore we can
apply Proposition 2.5: f(X) is a separable subset of C(E, X). Pick a dense countable
subset A of f(X). Since f(X) separates points of E, so does A. The diagonal product
A : E X
A
is therefore an embedding. Since X is metrizable and A is countable,
X
A
is metrizable, and so is E.
6. Some applications and remarks
6.1. Tame dynamical systems. For a topological space X denote by B
1
(X) the
space of all Baire 1 real-valued functions on X, equipped with the pointwise conver-
gence topology. A compact space K is Rosenthal if it is homeomorphic to a subspace
of B
1
(X) for some Polish X.
In [20, Theorem 3.2] the following dynamical Bourgain-Fremlin-Talagrand dichotomy
was established.
Theorem 6.1 (A dynamical BFT dichotomy). Let (G, X) be a metric dynamical
system and let E(X) be its enveloping semigroup. We have the following dichotomy.
Either
(1) E(X) is separable Rosenthal compact, hence with cardinality card E(X) 2

;
or
(2) the compact space E contains a homeomorphic copy of N, hence card E(X) =
2
2

.
In [19] a dynamical system is called tame if the rst alternative occurs, i.e. E(X) is
Rosenthal compact. It is shown in [19] that a minimal metrizable tame system with
a commutative acting group is PI. (For the denition of PI and for more details on
the structure theory of minimal dynamical systems see e.g. [17].) The authors of two
ON METRIZABLE ENVELOPING SEMIGROUPS 9
recent works [22] and [24] improve this result to show that under the same conditions
the system is in fact an almost 1-1 extension of an equicontinuous system.
Under the stronger assumption that E(X) is metrizable Theorem 1.2 now shows
that the commutativity assumption can be dropped and that the system is actually
equicontinuous. We get the following denitive result in the spirit of R. Ellis joint
continuity theorem [10].
Theorem 6.2. A metric minimal system (G, X) is equicontinuous if and only if its
enveloping semigroup E(X) is metrizable.
Proof. It is well known that the enveloping semigroup of a metric equicontinuous
system is a metrizable compact topological group (see e.g. [18, Exercise 1.26]). Con-
versely, if E(X) is metrizable then, by Theorem 1.2, (G, X) is HAE and being also
minimal it is equicontinuous (see the paragraph before Theorem 1.1).
Our characterization of metrizable HNS systems as those having metrizable en-
veloping semigroups should be compared with the following theorem:
Theorem 6.3. A compact metric dynamical system (G, X) is tame if and only if
every element of E(X) is a Baire 1 function from X to itself.
Proof. If Y is a separable metric space and B
1
(X, Y ) Y
X
is the space of Baire 1
functions from X to Y , then every compact subset of B
1
(X, Y ) is Rosenthal. Indeed,
Y embeds in R
N
, hence B
1
(X, Y ) embeds in B
1
(X, R
N
) = B
1
(X N). In particular,
if E(X) B
1
(X, X), then E(X) is Rosenthal, which means that (G, X) is tame.
Conversely, if E(X) is Rosenthal, then by the Bourgain-Fremlin-Talagrand theorem
it is Frechet [6]. (Recall that a topological space K is Frechet if for every A K
and every x A there exists a sequence of elements of A which converges to x.) In
particular, every p E(X) = G (we may assume that G Homeo (X)) is the limit
of a sequence of elements of G and therefore of Baire class 1 (Proposition 2.1).
Remarks 6.4. (1) Note that Theorem 1.2 resolves Problem 15.3 in [20]. In fact, since
the Glasner-Weiss examples are metric and HNS (see [20, Section 11]) we now know
that their enveloping semigroups are metrizable.
(2) Theorem 6.2 answers negatively Problem 3.3 in [19].
(3) In his paper [13] Ellis, following Furstenbergs classical work, investigates the
projective action of GL(n, R) on the projective space P
n1
. It follows from his results
that the corresponding enveloping semigroup is not rst countable. In a later work
[1], Akin studies the action of G = GL(n, R) on the sphere S
n1
and shows that
here the enveloping semigroup is rst countable (but not metrizable). The dynamical
systems D
1
= (G, P
n1
) and D
2
= (G, S
n1
) are tame but not RN. Note that E(D
1
)
is Frechet, being a continuous image of a rst countable space, namely E(D
2
).
6.2. Distality and equicontinuity. A dynamical system (G, X) is distal if for any
two distinct points x, y X the closure of the set (gx, gy) : g G in X
2
is disjoint
10 E. GLASNER, M. MEGRELISHVILI, AND V.V. USPENSKIJ
from the diagonal. If X is metrizable and d is a compatible metric on X, this condition
means that inf
gG
d(gx, gy) > 0. Every equicontinuous system is distal. By a theorem
of Ellis a dynamical system (G, X) is distal if and only if its enveloping semigroup
E(X) is (algebraically) a group, see [11]. Note that this characterization implies that
for any distal system (G, X) the phase space X is the disjoint union of its minimal
subsets. In particular it follows that a point transitive distal system is minimal. (A
dynamical system (G, X) is point transitive if there is some x X for which the orbit
Gx is dense in X.) As we have already observed, when X is equicontinuous E(X) is
actually a compact topological group.
One version of Ellis famous joint continuity theorem says that a compact dy-
namical system (G, X) such that E(X) is a group of continuous maps is necessarily
equicontinuous (see [10] and [4, page 60]). Using Elliss characterizations of WAP and
distality this can be reformulated as follows: A distal WAP system is equicontinuous.
We will now show that the WAP condition can not be much relaxed.
Example 6.5. The following is an example of a dynamical system (Z, X) which is
distal, HAE, and its enveloping semigroup E(X) is a compact topological group
isomorphic to the 2-adic integers. However, (Z, X) is not WAP and a fortiori not
equicontinuous.
Let S = R/Z (reals mod 1) be the circle. Let X = S (N), where N
is the one point compactication of the natural numbers. Let T : X X be dened
by:
T(s, n) = (s + 2
n
, n), T(s, ) = (s, ).
It is not hard to see that E(X) is isomorphic to the compact topological group Z
2
of 2-adic integers. The fact that X is not WAP can be veried directly by observing
that E(X) contains discontinuous maps. Indeed, the map f
a
E(X) corresponding
to the 2-adic integer
a = . . . 10101 = 1 + 4 + 16 +. . .
can be described as follows: f
a
(s, n) = (s +a
n
, n), where
a
2k
=
2
2k
1
3 2
2k

1
3
, a
2k+1
=
2
2k+2
1
3 2
2k+1

2
3
.
Geometrically this means that half of the circles are turned by approximately 2/3,
while the other half are turned by approximately the same angle in the opposite
direction. The map f
a
is discontinuous at the points of the limit circle.
For a point transitive HAE system distality is equivalent to equicontinuity because,
as we have seen, a distal point transitive system must be minimal and a minimal HAE
system is equicontinuous.
6.3. Semigroup compactications of groups. A semigroup S is right topological
if it is equipped with such a topology that for every y S the map x xy from
S to itself is continuous. (Some authors use the term left topological for this.) If
ON METRIZABLE ENVELOPING SEMIGROUPS 11
for every y S the self-maps x xy and x yx of S both are continuous,
S is a semitopological semigroup. A right topological semigroup compactication of
a topological group G is a compact right topological semigroup S together with a
continuous semigroup morphism G S with a dense range such that the induced
action GS S is continuous. A typical example is the enveloping semigroup E(X)
of a dynamical system (G, X) together with the natural map G E(X).
Semitopological semigroup compactications are dened analogously.
We have the following direct corollaries of Theorem 1.2.
Corollary 6.6. For a metric HAE system (G, X) its enveloping semigroup E(X) is
again a metrizable HAE system.
Proof. This follows from Theorem 1.2 because the enveloping semigroup of the ow
(G, E(G, X)) is isomorphic to E(G, X).
Corollary 6.7. The following three classes coincide:
(1) Metrizable enveloping semigroups of G-systems.
(2) Enveloping semigroups of HAE metrizable G-systems.
(3) Metrizable right topological semigroup compactications of G.
Proof. A dynamical system has the structure of a right topological semigroup com-
pactication of G if and only if it is the enveloping semigroup of some dynamical
system (see e.g. [18, Section 1.4] or [20, Section 2]).
Remark 6.8. It is well known that the enveloping semigroup of a WAP dynamical
system is a semitopological semigroup compactication of G (see e.g. [18, Section
1.4] or [20, Section 2]). Thus a WAP version of Corollary 6.7 (omitting part (1))
can be obtained by changing HAE to WAP and right topological semigroup to
semitopological semigroup. Moreover, as was shown in [9] (see also [18, Theorem
1.48]), when the acting group G is commutative, a point transitive WAP system is
always isomorphic to its enveloping semigroup, which in this case is a commutative
semitopological semigroup. Thus for such G the class of all metric, point transitive,
WAP systems coincides with that of all metrizable, commutative, semitopological
semigroup compactications of G.
6.4. Semigroup actions. Our main result (Theorem 1.2) remains true for semigroup
actions up to a more exible version of HAE. Namely, we say that a continuous
action of a topological semigroup S on a metric space (X, d) is HAE if for every (not
necessarily S-invariant) closed nonempty subset Y there exists a dense subset Y
0
Y
such that every point y
0
Y
0
is a point of continuity of the natural inclusion map
(Y, d[
Y
) (X, d
S
), where d
S
(x, y) := sup
sS
d(sx, sy). (It is not hard to see that
for G-group actions on compact metric spaces this denition is equivalent to our old
denition which involved only G-invariant closed subsets.) Then again HAE, RN and
the metrizability of E(X) are equivalent. We omit the details.
12 E. GLASNER, M. MEGRELISHVILI, AND V.V. USPENSKIJ
References
[1] E. Akin, Enveloping linear maps, Topological dynamics and applications, Contemporary Math-
ematics 215, a volume in honor of R. Ellis, 1998, pp. 121-131.
[2] E. Akin, J. Auslander, and K. Berg, When is a transitive map chaotic , Convergence in Ergodic
Theory and Probability, Walter de Gruyter & Co. 1996, pp. 25-40.
[3] E. Akin, J. Auslander, and K. Berg, Almost equicontinuity and the enveloping semigroup, Topo-
logical dynamics and applications, Contemporary Mathematics 215, a volume in honor of R. El-
lis, 1998, pp. 75-81.
[4] J. Auslander, Minimal Flows and their Extensions, Mathematics Studies 153, Notas de
Matematica, 1988.
[5] J. Auslander and J. Yorke, Interval maps, factors of maps, and chaos, Tohoku Math. J. 32
(1980), 177-188.
[6] J. Bourgain, D.H. Fremlin and M. Talagrand, Pointwise compact sets of Baire-measurable func-
tions, Amer. J. Math. 100 (1978), 845-886.
[7] R.D. Bourgin, Geometric Aspects of Convex Sets with the Radon-Nikodym Property, Lecture
Notes in Math., 993, Springer-Verlag, 1983.
[8] R. Deville, G. Godefroy and V. Zizler, Smoothness and renormings in Banach spaces, Pitman
Monographs and Surveys in Pure and Applied Mathematics, 64, Longman Scientic Technical,
1993.
[9] T. Downarowicz, Weakly almost periodic ows and hidden eigenvalues, Topological dynamics
and applications, Contemporary Mathematics 215, a volume in honor of R. Ellis, 1998, pp.
101-120.
[10] R. Ellis, Locally compact transformation groups, Duke Math. J. 24, (1957), 119-126.
[11] R. Ellis, Distal transformation groups, Pacic J. Math. 8, (1957), 401-405.
[12] R. Ellis, Equicontinuity and almost periodic functions, Proc. Amer. Math. Soc. 10 (1959), 637-
643.
[13] R. Ellis, The enveloping semigroup of projective ows, Ergod. Th. Dynam. Sys. 13 (1993),
635-660.
[14] R. Ellis and M. Nerurkar, Weakly almost periodic ows, Trans. Amer. Math. Soc. 313 (1989),
103-119.
[15] R. Engelking, General topology, revised and completed edition, Heldermann Verlag, Berlin,
1989.
[16] M. Fabian, Gateaux dierentiability of convex functions and topology. Weak Asplund spaces,
Canadian Math. Soc. Series of Monographs and Advanced Texts, A Wiley-Interscience Publi-
cation, New York, 1997.
[17] E. Glasner, Structure theory as a tool in topological dynamics, Descriptive set theory and dy-
namical systems, LMS Lecture note Series 277, Cambridge University Press, Cambridge, 2000,
173-209.
[18] E. Glasner, Ergodic Theory via joinings, Math. Surveys and Monographs, AMS, 101, 2003.
[19] E. Glasner, On tame dynamical systems, Colloq. Math. 105 (2006), 283-295.
[20] E. Glasner and M. Megrelishvili, Hereditarily non-sensitive dynamical systems and linear rep-
resentations, Colloq. Math. 104 (2006), no. 2, 223-283.
[21] E. Glasner and B. Weiss, Sensitive dependence on initial conditions, Nonlinearity 6 (1993),
1067-1075.
[22] W. Huang, Tame systems and scrambled pairs under an abelian group action, Ergod. Th. Dy-
nam. Sys. 26 (2006), 1549-1567.
[23] A.S. Kechris, Classical descriptive set theory, Springer-Verlag, Graduate texts in mathematics
156, 1991.
ON METRIZABLE ENVELOPING SEMIGROUPS 13
[24] D. Kerr and H. Li, Independence in topological and C

-dynamics, to appear.
[25] M. Megrelishvili, Fragmentability and representations of ows, Topology Proceedings 27 (2003),
no. 2, 497-544. See also: www.math.biu.ac.il/megereli.
[26] M. Megrelishvili, Topological transformation groups: selected topics, in book: Second edition of
Open Problems in Topology (Elliott Pearl, editor), to appear.
[27] E. Michael and I. Namioka, Barely continuous functions, Bull. Acad. Polon. Sci. Ser. Sci. Math.
Astronom. Phys. 24 (1976), no. 10, 889-892.
[28] I. Namioka, Separate continuity and joint continuity, Pacif. J. Math. 51 (1974), 515-531.
[29] I. Namioka, Radon-Nikodym compact spaces and fragmentability, Mathematika 34 (1987), 258-
281.
[30] J. Saint Raymond, Jeux topologiques et espaces de Namioka, Proc. Amer. Math. Soc. 87 (1983),
499-504.
Department of Mathematics, Tel-Aviv University, Tel Aviv, Israel
E-mail address: [email protected]
URL: http://www.math.tau.ac.il/

glasner
Department of Mathematics, Bar-Ilan University, 52900 Ramat-Gan, Israel
E-mail address: [email protected]
URL: http://www.math.biu.ac.il/

megereli
Department of Mathematics, 321 Morton Hall, Ohio University, Athens, Ohio
45701, USA
E-mail address: [email protected]