Topology and its Applications 221 (2017) 524–533

Contents lists available at ScienceDirect

Topology and its Applications

www.elsevier.com/locate/topol

Distal points for Borel measures ✩

Keonhee Lee a , C.A. Morales b,∗
a
Department of Mathematics, Chungnam National University, Daejeon 305-764, Republic of Korea
b
Instituto de Matematica, Universidade Federal do Rio de Janeiro, C. P. 68.530, CEP 21.945-970,
Rio de Janeiro, Brazil

a r t i c l e i n f o a b s t r a c t

Article history: We introduce the notion of distal point for a Borel measure with respect to a given
Received 31 January 2017 homeomorphism. We prove that a point is distal for every nonatomic Borel prob-
Received in revised form 13 ability measure if and only if it is countably-distal for the given homeomorphism.
February 2017
We prove that the distal point set of a measure is a Borel set. We study the distal
Accepted 16 February 2017
Available online 24 February 2017 measures (i.e. measures for which every point is distal) and prove that they are
approximated with respect to the weak* topology by ones with invariant support.
MSC: Furthermore, the distal measures are dense in the space of measures just when
primary 54H20 the ones with full support are. Afterwards, we consider the almost distal measures
secondary 37B05 (i.e. measures for which almost every point is distal) and exhibit one which is not
distal. Moreover, a circle homeomorphism has an almost distal measure with full
Keywords: support if and only if it is distal. In particular, every countably-distal circle home-
Distal homeomorphism omorphism is distal. We prove for circle homeomorphisms with rational rotation
Distal measure number that the almost distal measures are precisely the distal ones. Finally, we
Almost distal measure
prove that every homeomorphism with distal measures has uncountably many al-
most periodic points and those with almost distal measures on compact spaces have
infinitely many nonwandering points.
© 2017 Elsevier B.V. All rights reserved.

1. Introduction

Let f : X → X be a homeomorphism of a metric space X. A point x ∈ X is distal if

inf d(f n (x), f n (y)) > 0

n∈Z

for every y ∈ X \ {x}. Equivalently, if P (x) = {x} where

 
P (x) = y ∈ X : inf d(f (x), f (y)) = 0
n n
(1)
n∈Z

✩
The first author was supported by the NRF grant funded by the Korea government (MSIP) (No. NRF-2015R1A2A2A01002437).
The second was partially supported by CNPq-Brazil-303389/2015-0.
* Corresponding author.
E-mail address: morales@impa.br (C.A. Morales).

http://dx.doi.org/10.1016/j.topol.2017.02.069
0166-8641/© 2017 Elsevier B.V. All rights reserved.
 K. Lee, C.A. Morales / Topology and its Applications 221 (2017) 524–533 525

is the proximal cell of x (occasionally we write Pf (x) to indicate the dependence on f ). Consequently,
a necessary condition for x to be distal is that μ(P (x)) = 0 for every Borel probability measure μ which is
nonatomic (i.e. without points of positive mass).
It is natural to ask if the latter condition is also sufficient for a point x to be distal. Nevertheless, the
answer is negative by the following counterexample.

Example 1.1. There is a compact metric space X, a homeomorphism f : X → X and a point x ∈ X which
is not distal for f though μ(P (x)) = 0 for every nonatomic Borel probability measure μ of X.

Proof. Denote by A = {z = (θ, r) ∈ C : 1 ≤ |z| ≤ 2} the closed annulus of C. Define F : A → A by

F (θ, r) = (θ + α (mod 1), (r − 1)2 + 1) where α is an irrational number. Clearly F is a homeomorphism of A.
Define X = ∂A ∪ {pn : n ∈ Z} where ∂A denotes the boundary of A and pn = F n (0, 13 ) for n ∈ Z. Hence X
is a compact invariant set of F . Now define the homeomorphism f : X → X by f = F |X . Then, x = (0, 0)
is not distal since P (x) = {(0, 0), (0, 13 )} but μ(P (x)) = 0 for every nonatomic Borel probability measure μ
is since P (x) has two elements. 2

However, we will show that the condition μ(P (x)) = 0 for every nonatomic measure μ still implies some
distality for x. For this we introduce the following definition closely related to the n-expansive and countable
expansive homeomorphisms [5,7,9].

Definition 1.2. Let f : X → X be a homeomorphism of a metric space X. We say that x ∈ X is a

countable-distal point of f if P (x) is a countable subset of X. We say that f is countable-distal if every
point x is.

Every distal point is clearly countable-distal but not conversely (by Example 1.1). The following re-
sult represents the central motivation of this work. It decomposes the distality of a given point into the
corresponding property for measures. It also parallels the one for expansive systems [2].

Theorem 1.3. Let f : X → X be a homeomorphism of a Polish metric space X. Then, x ∈ X is a countable-

distal point of f if and only if μ(P (x)) = 0 for every nonatomic Borel probability measure μ of X.

This theorem suggests the study of those points x satisfying μ(P (x)) = 0 for a given measure μ. More
precisely, we have the following definition.

Definition 1.4. Let μ be a Borel measure of a metric space X. A distal point of μ with respect to a home-
omorphism f : X → X (or simply a distal point of μ) is a point x ∈ X such that μ(P (x)) = 0. The distal
point set of μ with respect to f (or simply the distal point set of μ) is the set formed by the distal points
of μ with respect to f .

Let us illustrate this definition with the following examples.

Example 1.5. Let X be a metric space and x0 ∈ X. The Dirac measure supported on x0 is the measure δx0
whose value at A ⊆ X is 1 or 0 depending on whether x0 ∈ A or not. For every homeomorphism f : X → X
we have that x is a distal point of δx0 ⇔ δx0 (P (x)) = 0 ⇔ x0 ∈ / P (x) ⇔ x ∈ / P (x0 ). Consequently, the distal
point set of δx0 is X \ P (x0 ). In particular, the distal point set of δx0 is either ∅ or X \ {x0 } depending on
whether f is proximal (i.e. P (x) = X for every x ∈ X) or distal. In the latter case by choosing a nonisolated
point x0 we obtain a Borel probability measure for which the distal point set is a proper open and dense
subset of X.
526 K. Lee, C.A. Morales / Topology and its Applications 221 (2017) 524–533

Remark 1.6. It is not difficult to prove that the distal point set of any non-trivial convex combination of
Borel probability measures is the intersection of the distal point sets of the involved measures.

Let M(X) denote the set of Borel probability measures of a metric space X. Such a space is equipped
 
with the weak* topology defined by the convergence μn → μ if and only if φdμn → φdμ for every
continuous map φ : X → R.

Example 1.7. Theorem 6.3 in [11] together with Examples 1.5 and Remark 1.6 implies that the set of Borel
probability measures without distal points of a proximal homeomorphism of a separable metric space X is
dense in M(X).

These examples motivate the study of the distal point set of a given measure. In this direction we obtain
the following result.

Theorem 1.8. For every homeomorphism of a compact metric space, the distal point set of every Borel
probability measure is a Borel set.

We can also try to determine how large the distal point set of a given measure may be. This motivates
the following definition.

Definition 1.9. A distal measure of a homeomorphism of a metric space X is a Borel probability measure
whose distal point set is X.

Recall that a Borel probability measure μ of a metric space X is invariant respect to a continuous map
f : X → X if μ(f −1 (B)) = μ(B) for every Borel set B.
The invariance of the measure is not required in the previous definition. On the other hand, not every
Borel probability measure is distal. Actually, a necessary condition for a measure to be distal is that the
measure be nonatomic. This condition is also sufficient for countable-distal homeomorphisms namely for
such homeomorphisms the distal measures are precisely the nonatomic ones. In particular there are distal
homeomorphisms without distal measures—just take the identity of a finite space. Another example will be
given in Example 1.16.
On the other hand, the distal measures are dense in M(X) for countable-distal homeomorphisms on
complete separable metric spaces without isolated points. This follows because the distal measures are
precisely the nonatomic ones (as remarked above) which in turn are dense in M(X) by Corollary 8.1, p. 55
in [11]. Although every space X exhibiting a homeomorphism f : X → X whose distal measures are dense
in M(X) has no isolated points, we don’t know if such a homeomorphism f is countable-distal. However,
it is possible to obtain some information of the set of distal measures of a given homeomorphism.
The support of a Borel probability measure μ is the set supp(μ) of points x ∈ X such that μ(U ) > 0
for any neighborhood U of x. It follows that supp(μ) is a nonempty compact subset of X. We say that a
Borel probability measure μ of X has full or invariant support if supp(μ) = X or f (supp(μ)) = supp(μ)
respectively. Clearly an invariant measure has invariant support but not conversely. Also we do not known
if every homeomorphism with distal measures has necessarily an invariant distal measure. A partial answer
for this question is given below.

Theorem 1.10. Every distal measure of a homeomorphism of a compact metric space X can be approximated
by ones with invariant support. In addition, the distal measures are dense in M(X) if and only if the ones
with full support are.
 K. Lee, C.A. Morales / Topology and its Applications 221 (2017) 524–533 527

Fig. 1. An almost distal measure which is not distal.

We can also consider those Borel probability measures for which the distal point set has full measure.
This yields the following definition.

Definition 1.11. An almost distal measure of a homeomorphism of a metric space X is a Borel probability
measure whose distal point set has full measure.

Again the invariance of the measure is not required here. Obviously, every distal measure is almost distal.
As we shall see later in Lemma 2.8, every almost distal measure of a homeomorphism is nonatomic. Hence,
for countable-distal homeomorphisms, the almost distal measures are precisely the distal ones. This is false
for general homeomorphism by the following counterexample. Simultaneously, it shows that an almost distal
measure need not be distal.

Example 1.12. There is a homeomorphism of the two-torus T 2 exhibiting an almost distal measure which
is not distal.

Proof. We obtain T 2 by identifying the circles C1 and C2 in Fig. 1. Denote by C the circle in T 2 resulting
by such an identification. Consider the flow in T 2 as in that figure. The circle C is the set of equilibrium
points of this flow, with the remainder orbits spiraling forward and backward to C as described in the figure.
The homeomorphism is precisely the time-1 map of this flow. If suitable chosen, the flow has the property
that this time-1 map satisfies P (x) ∩ C = {x} (for x ∈ C) and C ⊆ P (x) (for x ∈ T 2 \ C). Let μ be the
Lebesgue measure of the circle C. It follows that μ(P (x)) = 0 (for x ∈ C) and μ(P (x)) = 1 (for x ∈ T 2 \ C).
From this we get that μ is almost distal. Since C is a proper subset of T 2 , μ is not distal. 2

However, we will prove the following result in the circle.

Theorem 1.13. A circle homeomorphism has an almost distal measure with full support if and only if it is
distal. In particular, every countably-distal circle homeomorphism is distal. Every almost distal measure of
a circle homeomorphism with rational rotation number is distal.

To finish we will give two dynamical consequences of the existence of distal or almost distal mea-
sures. First recall that a stable class of a homeomorphism f : X → X is a subset equals to {y ∈ X :
limn→∞ d(f n (x), f n (y)) = 0} for some x ∈ X. On the other hand, a Borel measure μ of X is positively
expansive if there is δ > 0 such that μ({y ∈ X : d(f n (x), f n (y)) ≤ δ for all n ∈ N}) = 0 for every x ∈ X.
Every continuous map with positively expansive invariant measures has uncountably many stable
classes [1]. Below we will obtain the same conclusion for homeomorphisms with almost distal measures
(invariant or not).
528 K. Lee, C.A. Morales / Topology and its Applications 221 (2017) 524–533

Theorem 1.14. Every homeomorphism with almost distal measures of a metric space has uncountably many
stable classes.

For the second consequence, we recall that a subset F ⊂ Z is syndetic if there is l > 0 such that
[n, n + l] ∩ F = ∅ for every n ∈ Z. We say that x ∈ X is an almost periodic point of a homeomorphism
f : X → X if for every neighborhood U of x there is a F ⊂ Z syndetic such that f n (x) ∈ U for every n ∈ F .
 
We say that x is a nonwandering point of f if U ∩ n∈N+ f n (U ) = ∅ for every neighborhood U of x. The
set of almost periodic (resp. nonwandering) points is denoted by AP (f ) (resp. Ω(f )). The latter is often
referred to as the nonwandering set of f . Every almost periodic point is nonwandering.
Every distal homeomorphism of a compact metric space satisfies that every point is almost periodic (and
so nonwandering too) [3]. In light of this result it is tempting to say that every homeomorphism with distal
or almost distal has many almost periodic or nonwandering points. This will be obtained in our last result.

Theorem 1.15. Every homeomorphism with distal measures of a metric space has uncountably many almost
periodic points. Every homeomorphism with almost distal measures of a compact metric space has infinitely
many nonwandering points.

Example 1.16. By Theorem 1.15 every Morse–Smale diffeomorphism of a compact manifold [10] has no
almost distal measures.

It seems that the Anosov diffeomorphisms [4] have no almost distal measures too. It is natural to ask
if Theorem 1.10 holds for almost distal measures instead of the distal ones. What can be said about those
homeomorphisms without distal or almost distal measures? We can also ask if the above results can be
extended to general group actions.

2. Proof of the theorems

We start with some basic notation. Let f : X → X be a homeomorphism of a metric space X. For all
x ∈ X and  > 0 we write P [x, ] = {y ∈ X : d(f i (x), f i (y)) ≤  for some i ∈ Z} and

Pi [x, ] = {y ∈ X : d(f j (x), f j (y)) ≤  for some − i ≤ j ≤ i}, ∀i ∈ N.

Lemma 2.1. For every x ∈ X and  > 0 one has:

1. Pi [x, ] is closed;

2. P [x, ] = i∈N Pi [x, ];

3. P (x) = n∈N+ P [x, n−1 ].

Proof. Item (3) is clear so we have to prove (1) and (2) only.
To prove (1) take a sequence xk ∈ Pi [x, ] with xk → z for some z ∈ X. Since xk ∈ Pi [x, ], there exists
sequence −i ≤ jk ≤ i satisfying

d(f jk (x), f jk (xk )) ≤ , ∀k ∈ N.

Clearly we can assume jk = j for some fixed −i ≤ j ≤ i yielding

d(f j (x), f j (xk )) ≤ , ∀k ∈ N.

Letting k → ∞ above we obtain d(f j (x), f j (z)) ≤ . Since −i ≤ j ≤ i, this proves z ∈ Pi [x, ] yielding (1). To
prove (2), take y ∈ P [x, ]. Then, d(f j (x), f j (y)) ≤  for some j ∈ Z. By taking i = |j| we obtain −i ≤ j ≤ i
 K. Lee, C.A. Morales / Topology and its Applications 221 (2017) 524–533 529


satisfying d(f j (x), f j (y)) ≤ . Hence y ∈ Pi [x, ] for some i ∈ N thus proving P [x, ] ⊆ i∈N Pi [x, ]. The
reversed inclusion is obvious so (2) holds. 2

Recall that on a metric space an Fσ subset is a countable union of closed subsets; and an Fσδ subset is
a countable intersection of Fσ subsets. A direct consequence of Lemma 2.1 is as follows.

Corollary 2.2. For every homeomorphism f : X → X of a metric space X and every x ∈ X one has that
P (x) is an Fσδ subset of X. In particular, P (x) is a Borel set.

Proof of Theorem 1.3. It is clear that every countable-distal point x satisfies μ(P (x)) = 0 for every
nonatomic Borel probability measure μ. Conversely, suppose that μ(P (x)) = 0 for every nonatomic Borel
probability measure μ but P (x) is uncountable. We have that P (x) is a Borel set by Corollary 2.2. Then, by
Theorem 2.8 on p. 12 of [11], there is a Cantor set C ⊆ P (x). On the other hand, by Theorem 8.1 on p. 53
of [11], there is a nonatomic Borel probability μ with supp(μ) ⊆ C. However, μ is nonatomic so μ(P (x)) = 0
thus μ(C) = 0 which is absurd. This contradiction ends the proof. 2

Again by Lemma 2.1 we have that for every Borel probability measure μ of X,  > 0 and i ∈ N, the maps
ϕ , ϕ,i : X → X defined by ϕ (x) = μ(P [x, ]) and ϕ,i (x) = μ(Pi [x, ]) are well-defined. About the former
one we have the following lemma.

Lemma 2.3. If f : X → X is a homeomorphism of a compact metric space X, then ϕ is a Borel map for
every  > 0.

Proof. First note that Pi [x, ] ⊆ Pi [x, ] whenever i ≤ i . Then, ϕ = supi∈N ϕ,i by Lemma 2.1. It then
suffices to show that ϕ,i is measurable for any  > 0 and i ∈ N. For this we have to prove that the
set ϕ−1
,i (] − ∞, a[) is a Borel set, ∀a > 0. Actually, we shall prove that this set is open ∀a > 0. Take
x ∈ ϕ−1 −1
,i (] − ∞, a[) and a sequence xk ∈ X with xk → x. Because x ∈ ϕ,i (] − ∞, a[), we have μ(Pi [x, ]) < a.
As μ is regular [11], there is O open with Pi [x, ] ⊆ O such that μ(O) < a.
We claim that Pi [xk , ] ⊆ O for k large.
Otherwise, by compactness, we can choose a sequence yk ∈ Pi [xk , ] \ O such that yk → y for some
y ∈ X. As yk ∈ / O and O is open, we get y ∈ / O. However, yk ∈ Pi [xk , ] and then there is a sequence
−i ≤ jk ≤ i satisfying d(f jk (xk ), f jk (yk )) ≤  for all k ∈ N. Clearly we can assume jk = j for some fixed
−i ≤ j ≤ i hence d(f j (xk ), f j (yk )) ≤  for all k ∈ N. Letting k → ∞ above we obtain d(f j (x), f j (y)) ≤ .
Since −i ≤ j ≤ i, this prove y ∈ Pi [x, ] and so y ∈ O. This is a contradiction which proves the claim.
From this claim one has φ,i (xk ) = μ(P,i (xk )) ≤ μ(O) < a, and so xk belongs to ϕ−1 ,i (] − ∞, a[) for k
−1
large. Hence ϕ,i (] − ∞, a[) is open and we are done. 2

Proof of Theorem 1.8. By Corollary 2.2 we have that the map ϕ : X → R is defined by ϕ(x) = μ(P (x)) for
p.p.
x ∈ X is well-defined. On the other hand, the set of distal points of μ is precisely ϕ−1 (0) and ϕn−1 → ϕ
by Lemma 2.1. Since each ϕn−1 is Borel by Lemma 2.3, we obtain that ϕ is Borel too. Hence ϕ−1 (0) is a
Borel set and we are done. 2

To prove Theorem 1.10 we need the following facts. A topological space Y is a Baire space if the inter-
section of each countable family of open and dense subsets in Y is dense in Y . We say that A ⊆ Y is a Baire
subset if A is a Baire space with respect to the topology induced by Y . Hereafter we denote by Mdis (f )
the set of distal measures of f .

Lemma 2.4. If f : X → X is a homeomorphism of a compact metric space X, then Mdis (f ) is a Baire

subset of M(X).
530 K. Lee, C.A. Morales / Topology and its Applications 221 (2017) 524–533

Proof. For all δ,  > 0 and i ∈ N we define

C(, i, δ) = {μ ∈ M(X) : μ(Pi [x, ]) ≥ δ for some x ∈ X}.

By Lemma 2.1 we have

μ ∈ M(X) \ Mdis (f ) ⇐⇒ μ(P (x)) > 0 for some x ∈ X

⇐⇒ ∃n ∈ N+ such that lim μ(P [x, m−1 ]) ≥ n−1
m→∞
for some x ∈ X
⇐⇒ ∃n, m ∈ N+ , x ∈ X such that μ(P [x, l−1 ]) ≥ n−1
for all l ≥ m
⇐⇒ ∃n, m ∈ N+ and x ∈ X
such that ∀l ≥ m∃i ∈ N such that
μ(Pi [x, l−1 ]) ≥ n−1


⇐⇒ μ∈ C(l−1 , i, n−1 )
n∈N+ m∈N+ l≥m i∈N

and so


M(X) \ Mdis (f ) = C(l−1 , i, n−1 ). (2)
n∈N+ m∈N+ l≥m i∈N

Next we prove that C(, i, δ) is closed in M(X) for any , δ > 0 and any i ∈ N.
Fix , δ > 0 and i ∈ N. Take a sequence μn ∈ C(, i, δ) converging to μ ∈ M(X) with respect to the
weak* topology.
As μn ∈ C(, i, δ) there is a sequence xn ∈ X with xn → x such that μ(Pi [xk , ]) ≥ δ for any k ∈ N.
If we prove μ(Pi [x, ]) ≥ δ we would obtain μ ∈ C(, i, δ) and then C(, i, δ) is closed in M(X). Hence it
suffices to prove

μ(Pi [x, ]) ≥ δ.

Take any compact neighborhood V of Pi [x, ]. We claim that Pi [xn , ] ⊆ Int(V ) for k large. If not, there
is a sequence yn ∈ Pi [xn , ] \ Int(V ) which, by compactness, we can further assume yk → y for some y ∈ X.
On the one hand, we have y ∈ / Int(V ) because Int(V ) is open and, on the other, yn ∈ Pi [xn , ] for any n
hence there is a sequence −i ≤ jn ≤ i satisfying

d(f jn (xn )f jn (yn )) ≤ , ∀n ∈ N.

We can assume jn = j for some −i ≤ j ≤ i hence

d(f j (xn ), f j (yn )) ≤ , ∀n ∈ N.

Letting n → ∞ above we obtain

d(f j (x), f j (y)) ≤ ,

and so y ∈ Pi [x, ]. Since V is a neighborhood of Pi [x, ], we obtain y ∈ Int(V ) which is a contradiction.
This contradiction shows Pi [xn , ] ⊆ Int(V ) for k large and, then,
 K. Lee, C.A. Morales / Topology and its Applications 221 (2017) 524–533 531

δ ≤ μn (Pi [xn , ]) ≤ μn (V ), ∀n large.

As μn → μ and V is closed, Theorem 6.1 in [11] implies

δ ≤ lim sup μn (V ) ≤ μ(V ).

n→∞

Hence μ(V ) ≥ δ for any compact neighborhood V of Pi [x, ]. From this we get

μ(Pi [x, ]) ≥ δ

and so C(, i, δ) is closed in M(X), for all , δ > 0 and i ∈ N.

  
Then, i∈N C(l−1 , i, n−1 ) is an Fσ subset of M(X) (∀n, l ∈ N+ ) from which we get that l≥m i∈N C(l−1 ,
i, n−1 ) is an Fσδ subset of M(X) (∀m, n ∈ N+ ). Since the space M(X) of Borel probability measures
equipped with the weak* topology is a compact (hence complete), every Fσδ subset of it is Baire (cf.
 
Lemma 1 in [8]). Consequently, l≥m i∈N C(l−1 , i, n−1 ) is a Baire subset of M(X), ∀m, n ∈ N+ . Therefore,
M(X) \ Mdis (f ) (and so Mdis (f )) are Baire subsets by (2). This finishes the proof. 2

By a meagre subset of a topological space Y we mean the countable union of nowhere dense subsets of Y .

Lemma 2.5. For every homeomorphism with distal measures f : X → X of a compact metric space X there
is a meagre subset D of Mdis (f ) such that

supp(μ) = supp(ν), ∀μ ∈ Mdis (f ) \ D.
ν∈Mdis (f )

Proof. Let 2Xc denote the set of compact subsets of X. This set is a compact metric space if endowed with the
Hausdorff distance. Define Ψ : Mdis (f ) → 2X c by Ψ(μ) = supp(μ). It is clear that Ψ is lower-semicontinuous
so the set D of discontinuity points of Ψ is meagre (by Corollary 1 p. 71 in [6]). Let us prove that this D
satisfies the desired property.
Take μ ∈ Mdis (f ) \ D and ν ∈ Mdis (f ). Define μt = (1 − t)μ + tν for t ∈]0, 1[. Clearly μt ∈ Mdis (f ) and
μt → μ as t → 0. Since μ ∈ Mdis (f ) \ D, Ψ is continuous at μ and so Ψ(μt ) = supp(μt ) = supp(μ) ∪ supp(ν)
converges to Ψ(μ) = supp(μ). From this we obtain supp(ν) ⊆ supp(μ) and the proof follows. 2

The notion of measure center [12] motivates the following one: The measure-distal center of a homeo-
morphism f is defined by

D(f ) = supp(ν).
ν∈Mdis (f )

A property to be used later is given below.

Lemma 2.6. For every homeomorphism with distal measures f : X → X of a compact metric space X there
is a dense subset R of Mdis (f ) such that supp(μ) = D(f ) for all μ ∈ R.

Proof. By Lemma 2.4 we have that Mdis (f ) equipped with the weak* topology is a Baire space. Now let
D be the meagre subset of Mdis (f ) given by Lemma 2.5. Since Mdis (f ) is Baire, Mdis (f ) \ D is dense
in Mdis (f ). Then, Lemma 2.5 implies the result by taking R = Mdis (f ) \ D. 2

Corollary 2.7. The measure-distal center is a compact invariant set.

532 K. Lee, C.A. Morales / Topology and its Applications 221 (2017) 524–533

Proof. Let f : X → X be a homeomorphism of a compact metric space X. If Mdis (f ) = ∅ there is

nothing to prove. Otherwise, by Lemma 2.6, the set of distal measures μ satisfying supp(μ) = D(f ) is dense
in Mdis (f ). Since the support of any measure is compact, D(f ) is compact too. To obtain that D(f ) is
invariant, we simply observe that f (supp(ν)) = supp(f∗ (ν)) for all ν ∈ M(X) and that f∗ (ν) ∈ Mdis (f )
for all ν ∈ Mdis (f ). This completes the proof. 2

Proof of Theorem 1.10. Let f : X → X be a homeomorphism with distal measures of a compact metric
space X. By Lemma 2.6 there is a dense subset of the space of distal measures all of whose elements have
support equal to D(f ). Since D(f ) is invariant by Corollary 2.7, we are done.
Now suppose that Mdis (f ) is dense in M(X). Then, D(f ) = X (by Lemma 5 in [8]) so, by Lemma 2.6,
there is a dense subset R of Mdis (f ) such that supp(μ) = X for all μ ∈ R. Since Mdis (f ) is dense in
M(X), we have that R is dense in M(X) and we are done. 2

To prove Theorem 1.13 we will need three short lemmas.

Lemma 2.8. Every almost distal measure of a homeomorphism f : X → X of a metric space X is nonatomic.

Proof. Suppose by contradiction that μ({x0 }) > 0 for some x0 ∈ X. Since μ is almost distal, there is a
measurable set E with μ(E) = 1 such that μ(x) = 0 for all x ∈ E. As μ({x0 }) > 0 and μ(E) = 1, one has
x0 ∈ E and so μ({x0 }) = 0 which is absurd. This completes the proof. 2

Lemma 2.9. If μ is an almost distal measure of a circle homeomorphism f , then supp(μ) ⊆ Ω(f ).

Proof. We have that S 1 \ Ω(f ) is a disjoint collection of open intervals J. It follows that diam(f n (J)) → 0
as n → ±∞ hence J ⊆ P (x) for every x ∈ J. Now, assume by contradiction that there is an almost distal
measure μ with supp(μ)  Ω(f ). Then, μ(J) > 0 for some interval J as above. Hence μ(P (x)) ≥ μ(J) > 0
for all x ∈ J from which we get J ⊆ S 1 \ Dist(f, μ). It then follows that μ(S 1 \ Dist(f, μ)) > 0 and so
μ(Dist(f, μ)) < 1 which is absurd. 2

Lemma 2.10. For every homeomorphism f : X → X of a metric space X and every k ∈ N+ , every distal
point of a Borel measure μ with respect to f is a distal point of μ with respect to f k . In particular, every
almost distal measure of f is an almost distal measure of f k .

Proof. Since inf n∈Z d(f kn (x), f kn (y)) = 0 implies inf n∈Z d(f n (x), f n (y)) = 0, one has Pf k (x) ⊆ P (x) and
the proof follows. 2

Proof of Theorem 1.13. Let f : S 1 → S 1 be a circle homeomorphism. First assume that f has an almost
distal measure with full support. Then, Ω(f ) = S 1 by Lemma 2.9 and so f is topologically conjugated to a
circle rotation. From this we obtain that f is distal and the proof follows. If f is countably distal, then f
has a distal measure with full support (e.g. the Lebesgue measure) and so f is distal by the previous part.
Next assume that f has rational rotation number and, by contradiction, that f has an almost distal
measure μ. Since f has rational rotation number, the periodic points of f have a common period k ∈ N+
(say). In addition, by using Lemma 2.10 and replacing f by f k if necessary we can assume that Ω(f ) =
F ix(f ) = {x ∈ S 1 : f (x) = x}. Then, we have
⎧
⎪
⎨ {x} if x ∈ Int(F ix(f ));
P (x) = Cl(J) if x ∈ J for some component J of S 1 \ F ix(f ); (3)
⎪
⎩ J ∪ {x} if x ∈ F r(J) for some component J of S 1 \ F ix(f ),

where F r(·) and Cl(·) denote the boundary and closure operations in S 1 .
 K. Lee, C.A. Morales / Topology and its Applications 221 (2017) 524–533 533

Suppose by contradiction that f has an almost distal measure μ which is not distal. Then, μ(P (x0 )) > 0
for some x0 ∈ S 1 . Since μ is nonatomic by Lemma 2.8, (3) applied to x = x0 would imply μ(J) > 0 for some
connected component J of S 1 \ F ix(f ). However, supp(μ) ⊆ Ω(f ) (by Lemma 2.9) so supp(μ) ⊆ F ix(f )
thus μ(J) = 0 since J ⊆ S 1 \ F ix(f ). This is a contradiction which proves the result. 2

Proof of Theorem 1.14. Let f : X → X be a homeomorphism of a compact metric space X exhibiting

an almost distal measure μ. Suppose by contradiction that the set of stable classes of f is countable. It
is apparent that the stable classes form a partition of X. Moreover, such classes are Borel sets (see the
proof of Lemma 2.4 in [1]). It follows that μ(W s (x)) > 0 for some x ∈ X where W s (z) = {y ∈ X :
limn→∞ d(f n (z), f n (y)) = 0} for every z ∈ X. Since μ is almost distal, there is y ∈ W s (x) such that y is
a distal point of μ. Now fix x̄ ∈ W s (x). Then, d(f n (x̄), f n (x)) → 0 and, since y ∈ W s (x), we also have
d(f n (x), f n (y)) → 0 as n → ∞. From this we obtain d(f n (x̄), f n (y)) → 0 as n → ∞. We conclude that
x̄ ∈ W s (y) and so W s (x) ⊆ W s (y). But clearly W s (y) ⊆ P (y) so W s (x) ⊆ P (y). As y is a distal point
of μ, we get μ(P (y)) = 0 and so μ(W s (x)) = 0 which is absurd. Therefore, f has uncountably many stable
classes. 2

Recall the definition of proximal cell P (x) in (1).

Proof of Theorem 1.15. Let f : X → X is a homeomorphism of a metric space X. Assume that f has a distal
measure μ. By Theorem 3, p. 67 in [3] we have that for every x ∈ X there is x∗ ∈ AP (f ) such that x∗ ∈ P (x).

On the other hand, since x∗ ∈ P (x), we have x ∈ P (x∗ ) and so X = x∗ ∈AP (f ) P (x∗ ). Now suppose by
contradiction that AP (f ) is countable. Since the proximal cells are all Borel sets by Corollary 2.2, we would

have μ(X) ≤ x∗ ∈AP (f ) μ(P (x∗ )). But μ is distal so μ(P (x∗ )) = 0 for all x∗ ∈ AP (f ) thus μ(X) = 0 which
is absurd. Therefore, AP (f ) is uncountable.
Now assume that X is compact and that f has an almost distal measure. If Ω(f ) were finite, then the
number of stable classes of f is finite. Applying Theorem 1.14 we obtain a contradiction. Therefore, Ω(f )
is infinite. 2

References

[1] A. Arbieto, C.A. Morales, Some properties of positive entropy maps, Ergod. Theory Dyn. Syst. 34 (3) (2014) 765–776.
[2] A. Artigue, D. Carrasco-Olivera, A note on measure-expansive diffeomorphisms, J. Math. Anal. Appl. 428 (1) (2015)
713–716.
[3] J. Auslander, Minimal flows and their extensions, in: North-Holland Mathematics Studies, vol. 153, in: Notas de Matemática
(Mathematical Notes), vol. 122, North-Holland Publishing Co., Amsterdam, 1988.
[4] M. Brin, G. Stuck, Introduction to Dynamical Systems, Cambridge University Press, Cambridge, 2002.
[5] B. Carvalho, W. Cordeiro, n-expansive homeomorphisms with the shadowing property, J. Differ. Equ. 261 (6) (2016)
3734–3755.
[6] K. Kuratowski, Topology. Vol. II, new edition, revised and augmented. Translated from the French by A. Kirkor, Academic
Press, New York, London; Państwowe Wydawnictwo Naukowe Polish Scientific Publishers, Warsaw, 1968.
[7] J. Li, R. Zhang, Levels of generalized expansiveness, J. Dyn. Differ. Equ. (2017), http://dx.doi.org/10.1007/s10884-015-
9502-6, in press.
[8] C.A. Morales, On supports of expansive measures, preprint, arXiv:1601.03618 [math.DS].
[9] C.A. Morales, A generalization of expansivity, Discrete Contin. Dyn. Syst. 32 (1) (2012) 293–301.
[10] J. Palis, W. de Melo, Geometric Theory of Dynamical Systems. An Introduction, Springer-Verlag, New York, Berlin, 1982,
translated from the Portuguese by A.K. Manning.
[11] K.R. Parthasarathy, Probability Measures on Metric Spaces, Probab. Math. Stat., vol. 3, Academic Press, Inc., New York,
London, 1967.
[12] Z.L. Zhou, Weakly almost periodic point and measure centre, Sci. China Ser. A 36 (2) (1993) 142–153.