Discrete and Continuous Dynamical Systems Series S Volume 13, Number 4, April 2020
Discrete and Continuous Dynamical Systems Series S Volume 13, Number 4, April 2020
2020074
DYNAMICAL SYSTEMS SERIES S
Volume 13, Number 4, April 2020 pp. 1291–1317
Sergiı̆ Kolyada†
Institute of Mathematics, NASU
Tereshchenkivs’ka 3
01601 Kyiv, Ukraine
mathematical community, as a mathematician and as a person. Due to his untimely death, Profes-
sor Kolyada could not implement the changes to the first version of the manuscript as suggested in
the (positive) reviews. With the consent of Professor Kolyada’s family, Professor L’ubomı́r Snoha
(Matej Bel University, Banská Bystrica, Slovakia), a friend and colleague of Professor Kolyada,
assumed the responsibility of carrying out the revision. The editors thank Professor Snoha for this
invaluable contribution.
1291
1292 SERGIĬ KOLYADA
trivial if the space is a singleton. Throughout this paper we are only interested in
nontrivial dynamical systems. In Section 3 we additionally assume that X has no
isolated point and T is surjective.
When mathematicians are considering various classes of functions with a natural
topology, often one of the first questions that comes to their mind is about the
topological properties of those spaces. While those questions have been answered
long ago in Mathematical Analysis and Topology, and also in Ergodic Theory there
is a series of papers where topological properties of the group of measure preserving
bijections or homeomorphisms are investigated, see e.g. [25], [31], [15], [10], [43], [51],
it seems that they even have not been asked before in Dynamical Systems. There
are very few papers on topological properties of spaces of maps that have some
specific dynamical properties. Perhaps, the first of them were the paper of Farrell
and Gogolev [14] about the spaces of Anosov diffeomorphisms, the paper [32] that
was being written at the same time as [14] (and completely independently of it),
and the paper [33].
The area of Dynamical Systems where one investigates dynamical properties
that can be described in topological terms is Topological Dynamics. Investigating
topological properties of spaces of maps that can be described in dynamical terms
is in a sense the opposite idea. Therefore in [32] it was proposed to call this area
Dynamical Topology.
In the present paper we survey some recent results in the following areas of
Dynamical Topology:
(1) Topological properties of the space of all transitive maps of a compact interval
to itself and its subspaces.
(2) Dynamical compactness, especially transitive and sensitive compactnesses as
new concepts of chaoticity of a dynamical system.
(3) Slovak spaces, i.e. compact metric spaces whose homeomorphism group is
cyclic and generated by a minimal homeomorphism.
When some notion is not defined where it appears for the first time, the reader
should consult Appendix.
full n-horseshoe is a piecewise monotone map with constant slope and n laps, each
of which is mapped to the whole domain of the map.
We will use the following notation.
(1) I = [0, 1];
(2) T – (continuous !) transitive maps I → I
(3) TPM – piecewise monotone transitive maps I → I;
(4) TPL – piecewise linear transitive maps I → I;
(5) Tn – elements of TPM of modality n;
(6) Tn+ – elements of Tn increasing on the first lap;
(7) Tn− – elements of Tn decreasing on the first lap;
(8) CS n – piecewise linear maps I → I, with constant slope and of modality n;
(9) T CS n – transitive maps from CS n .
All those spaces are considered with the C 0 -metric d:
d(f, g) = sup |f (x) − g(x)|.
x∈I
2.1. How to recognize that a piecewise linear map with constant slope is
transitive. We use a coding of maps f ∈ CS n as follows.
(1) Code of f = (f (0), f (1st turning pt), . . . , f (nth turning pt), f (1)).
(2) When we consider a union of spaces CS n with different n0 s, we use the common
(largest) length of codes, say
code of tent map = (0, 1, 0) = (0, 0, 1, 0) = (0, 1, 0, 0) = . . . ,
One can show the following two properties. Pn+1
(3) If code of f = (a0 , . . . , an+1 ) then slope of f = j=1 |aj − aj−1 |.
Sn
(4) f ∈ i=1 CS i depends continuously on the parameters a0 , a1 , . . . , an+1
(jointly).
It is easy to describe the space T CS 1 .
Lemma 2.1 ([33]). A map f ∈ CS 1 is transitive if and only if it has the code
√
(1) (a, 1, 0) where a ∈ [0,
√ 2 − 2], or
(2) (1, 0, c) where c ∈ [ 2 − 1, 1].
If a map is in T CS 2 then one can show that it has one of the four codes (a, 1, 0, d),
(a, 0, 1, d), (1, 0, c, d), (a, b, 1, 0). To save the space, the results proved in several lem-
mas in [33] are presented in Table 1. Alternatively, one could use [45, Theorem 1].
Using these results, the space T CS 1 ∪ T CS 2 has been drawn in [33], see Figure 1.
1294 SERGIĬ KOLYADA
CS 2
code picture condition equivalent to transitivity
2 2
(a, 1, 0, d) d≤a−4+ a or 1 − a ≤ (1 − d) − 4 + 1−d
(a, 0, 1, d) a>d
1
(1, 0, c, d) d ≤ 2 + 2c − c
1
(a, b, 1, 0) 1 − a ≤ 2 + 2(1 − b) − 1−b
Table 1. Transitive maps in CS 2
Figure 3. Boxes Ii × Ij
ar = f (1) ∈ [0, 1]. Hence the set Z = {gf,1 : f ∈ T }, being homeomorphic to the
square, is contractible.
Our family gf,t can be treated as a homotopy joining
1. the identity idT : T → T , f 7→ f = gf,0 and
2. the map T → Z ⊂ T , f 7→ gf,1 .
Since Z is contractible, also T is contractible.
Theorem 2.6 ([32]). Each space Tn has two connected components, Tn+ and Tn− ,
and they are arcwise connected. The distance between T1+ and T1− is positive, while
the distance between Tn+ and Tn− is zero for all n ≥ 2.
2.3. Loops of transitive interval maps. We are going to consider loops of tran-
sitive maps with constant slopes (the reader may wish to check that transitivity
of maps considered below in this subsection always follows from results in Subsec-
tion 2.1; we will not emphasize it any more).
Theorem 2.7 ([32]). For every n ≥ 1 there is a loop in Tn ∪ Tn+1 , which is not
contractible in Tn ∪ Tn+1 .
In fact, we have found such a loop in T CS n ∪ T CS n+1 . We call it the basic loop
of order n and we denote it by Ln . It can be described as follows. We start with
the full (n + 1)-horseshoe which is increasing on the first lap, then an additional
rightmost lap starts to grow until we get a full (n + 2)-horseshoe. Next the leftmost
lap starts to disappear, until we get the full (n + 1)-horseshoe which is decreasing
on the first lap. Then we repeat those operations and we end up with the initial
(n + 1)-horseshoe. For n = 2 see Figure 4. A rigorous description of the basic loops
can be obtained by using codes, see [33].
The basic loop L2 consisting of four arcs has been deformed to the auxiliary loop
consisting of two arcs (formally four arcs, but the second and the third are constant,
so we can ignore them), see Figure 9.
The auxiliary loop consisting of two arcs can be seen in Figures 10 and 11.
Still we need to show that the auxiliary loop consisting of two arcs is contractible
in T CS 2 ∪ T CS 4 . The techniques for showing that are similar to those already used
1298 SERGIĬ KOLYADA
Figure 10. The 1st arc Figure 11. The 2nd arc
of the auxiliary loop ob- of the auxiliary loop ob-
tained from L2 . tained from L2 .
above and therefore we refer the reader to [33] for details. Here we only reveal the
idea: One can show that the first of these two arcs can be deformed, in T CS 2 ∪T CS 4 ,
to the second arc run backward. Moreover, during that homotopy the endpoints
of the arc are not modified and so we get contractibility of the auxiliary loop in
T CS 2 ∪ T CS 4 . Thus, L2 is really contractible in T CS 2 ∪ T CS 3 ∪ T CS 4 .
We worked with the basic loop L2 . More generally, in [33] the following is shown
(recall that Ln was a loop in T CS n ∪ T CS n+1 and not contractible there, even not
contractible in Tn ∪ Tn+1 ).
Theorem 2.8 ([33]). Let Ln be the basic loop constructed in the proof of Theo-
rem 2.7.
(1) If n ≥ 2 then Ln can be contracted in T CS n ∪ T CS n+1 ∪ T CS n+2 .
(2) If n ≥ 3 then Ln can be contracted also in T CS n−1 ∪ T CS n ∪ T CS n+1 .
(3) L1 can be contracted in T CS 1 ∪ T CS 2 ∪ T CS 4 .
The situation seems to be similar as for the following model, although we do not
know how far we can go with this analogy. Think about the sequence of spaces Rn ,
n = 0, 1, 2, . . . , where each space is a subset of the next one. Set Rn = Rn \ Rn−1
for n = 1, 2, 3, . . . . Then the fundamental group of the space
Rn ∪ Rn+1 = Rn+1 \ Rn−1 = (R2 \ {0}) × Rn−1
is nontrivial, while the fundamental group of the space
Rn ∪ Rn+1 ∪ Rn+2 = Rn+2 \ Rn−1 = (R3 \ {0}) × Rn−1
is trivial.
3. Dynamical compactness. This section is based on the papers [28] and [29].
Throughout this section, when considering a dynamical system (X, T ), we addition-
ally assume that the compact metric space X has no isolated point and that the
continuous map T is surjective.
DYNAMICAL TOPOLOGY 1299
Let us note that not always ωF (x) is a subset of the ω-limit set ωT (x), which is
defined as
\∞
{T k x : k ≥ n} = {z ∈ X : nT (x, G) ∈ B for every neighborhood G of z}.
n=1
For instance, if each element of F contains 0 then any point x ∈ ωF (x). Neverthe-
less, if a family F is free, then ωF (x) ⊂ ωT (x) for any point x ∈ X and if (X, T )
has a nonrecurrent point1 , then the converse is true.
A dynamical system (X, T ) is called compact with respect to F, or shortly dy-
namically compact, if the ωF -limit set ωF (x) is nonempty for all x ∈ X.
H. Furstenberg started a systematic study of transitive systems in his paper on
disjointness in topological dynamics and ergodic theory [17], and the theory was
further developed in [19] and [18]. Recall that the system (X, T ) is (topologically)
transitive if NT (U1 , U2 ) = {n ∈ Z+ : U1 ∩ T −n U2 6= ∅} (= {n ∈ Z+ : T n U1 ∩ U2 6=
∅}) ∈ P+ for any opene 2 subsets U1 , U2 ⊂ X, equivalently, NT (U1 , U2 ) ∈ B for any
opene subsets U1 , U2 ⊂ X.
3.2. Transitive compactness, sensitive compactness, FIP. Transitive com-
pactness is one of possible dynamical compactnesses. Let NT be the set of all
subsets of Z+ containing some NT (U, V ), where U, V are opene subsets of X. A
dynamical system (X, T ) is called transitively compact, if for any point x ∈ X the
ωNT -limit set ωNT (x) is nonempty, in other words, for any point x ∈ X there exists
a point z ∈ X such that
nT (x, G) ∩ NT (U, V ) 6= ∅
for any neighborhood G of z and any opene subsets U, V of X.
1A point x ∈ X is called recurrent if x ∈ ωT (x).
2 Because we so often have to refer to open, nonempty subsets, we will call such subsets opene.
1300 SERGIĬ KOLYADA
Let (X, T ) and (Y, S) be two dynamical systems and k ∈ N. The product system
(X × Y, T × S) is defined naturally, and denote by (X k , T (k) ) the product system of
k copies of the system (X, T ). Recall that the system (X, T ) is minimal if it does
not admit a nonempty, closed, proper subset K of X with T K ⊂ K, and is weakly
mixing if the product system (X 2 , T (2) ) is transitive. Any transitively compact
system is obviously topologically transitive, and observe that each weakly mixing
system is transitively compact ([3]).
Sensitive compactness is another possible dynamical compactness. The notion
of sensitivity was first used by Ruelle [46]. It captures the idea that in a chaotic
system a small change in the initial condition can cause a big change in the trajec-
tory. According to the works by Guckenheimer [24] and Auslander and Yorke [7],
a dynamical system (X, T ) is called sensitive if there exists δ > 0 such that for
every x ∈ X and every neighborhood Ux of x, there exist y ∈ Ux and n ∈ N with
d(T n x, T n y) > δ. Such a δ is called a sensitive constant of (X, T ). Recently in
[42] Moothathu initiated a way to measure the sensitivity of a dynamical system,
by checking how large is the set of nonnegative integers for which the sensitivity
occurs. For a positive δ and a subset U ⊂ X define
ST (U, δ) = {n ∈ Z+ : there are x1 , x2 ∈ U such that d(T n x1 , T n x2 ) > δ}.
Let ST (δ) be the set of all subsets of Z+ containing some ST (U, δ), where U is
an opene subset of X. A dynamical system (X, T ) is called sensitively compact, if
there exists δ > 0 such that for any point x ∈ X the ωST (δ) -limit set ωST (δ) (x) is
nonempty, in other words, for any point x ∈ X there exists a point z ∈ X such that
nT (x, G) ∩ ST (U, δ) 6= ∅
for any neighborhood G of z and any opene subset U in X.
A dynamical system (X, T ) is called multi-sensitive if there exists δ > 0 such
Tk
that i=1 ST (Ui , δ) 6= ∅ for any finite collection of opene U1 , . . . , Uk ⊂ X. Such a
δ is called a constant of multi-sensitivity of (X, T ).
Recall that a collection A of subsets of a set Y has the finite intersection property
(FIP) if the intersection of all sets in any finite subcollection of A is nonempty. If
such intersections are infinite, we speak on strong finite intersection property (SFIP).
It is well known that the FIP is useful in formulating an alternative definition of
compactness of a topological space: a topological space is compact if and only if
every collection of closed subsets satisfying the FIP has a nonempty intersection
itself.
We recall that if (X, T ) is weakly mixing then it is well known that the family
NT is a filter (see [17], [1]), and hence has FIP. If (X, T ) is a multi-sensitive system
with a constant of multi-sensitivity δ > 0 then obviously the family ST (δ) has FIP.
Since these families are also free, actually they have SFIP.
In fact we can say more — the FIP is useful in characterizing the dynamical
compactness.
Theorem 3.1 (Theorem FIP, [28]). All dynamical systems are dynamically compact
with respect to F if and only if the family F has the finite intersection property.
Proof. Sufficiency. Suppose that F has FIP. Take arbitrary dynamical system
(X, T ) and let x ∈ X. Obviously the family {T F x : F ∈ F} also has FIP, and
then by compactness of X the family {T F x : F ∈ F} has a nonempty intersection
itself, i.e., ωF (x) 6= ∅. Thus (X, T ) is dynamically compact with respect to F.
DYNAMICAL TOPOLOGY 1301
Necessity. Suppose that the family F does not have FIP. Then there is a collection
Tk
{F1 , . . . , Fk } ⊂ F with i=1 Fi = ∅. Let A = {a1 , . . . , ak } be an alphabet and let
(X, T ) := (Σ, σ) be the full (one-sided) A-shift. We are going to define a point
x ∈ X with ωF (x) = ∅. Let x0 = a1 . For any n ≥ 1 there is i with n ∈ / Fi , else
the intersection of F1 , . . . , Fk would be nonempty. Then define xn := ai . Finally,
let x = x0 x1 x2 x3 . . . and the construction is finished.
Assume the contrary that we can take z ∈ ωF (x), and that z begins with ai ∈ A.
Take Gz = C[ai ]. As z ∈ ωF (x) we have nT (x, Gz ) ∩ Fi 6= ∅. But if n ∈ nT (x, Gz ),
then xn = ai and so n ∈ / Fi by the construction, a contradiction.
3.3. Dynamical compactness with respect to an arbitrary family. As we
have mentioned, any filter has FIP; if (X, T ) is weakly mixing then the family NT
is a filter; if A is a weakly mixing subset of (X, T ) then the family NT (A) has FIP;
and if (X, T ) is a multi-sensitive system with a constant of multi-sensitivity δ > 0
then the family ST (δ) also has FIP.
A collection H ⊂ F will be called a base for F if for any F ∈ F there is H ∈ H
with H ⊂ F . We are interested in those families which have a countable base.
Note that not every Furstenberg family F has a countable base, for example, the
family B. Assume the contrary that B admits a countable base {Fn : n ∈ N}. We
take k1 ∈ F1 , and once km ∈ Fm , m ∈ N is defined we choose km+1 ∈ Fm+1 with
km+1 > km + m + 1. Set E = {kn : n ∈ N} and F = Z+ \ E. Then E ∩ Fn 6= ∅ for
all n ∈ N, and F ⊃ {km + m : m ∈ N} and hence F ∈ B, in particular, there exists
no n ∈ N with Fn ⊂ F , a contradiction.
It is not hard to show even the existence of a family with FIP, but without
a countable base. Nevertheless the families NT and ST (δ) have countable bases.
Indeed, we can consider a countable base U of open sets for the space X. Note that
U1 ⊂ U , V1 ⊂ V implies NT (U1 , V1 ) ⊂ NT (U, V ) and ST (U1 , δ) ⊂ ST (U, δ). Then
{NT (U, V ) : U, V ∈ U } and {ST (U, δ) : U ∈ U } are countable bases for NT and
ST (δ), respectively.
The following is a general result that will be especially useful for families with
countable bases. By TranF (X, T ) we denote the set of all F-transitive points in
(X, T ), see Appendix for the definition.
Proposition 3.2 ([29]). Let (X, T ) be a dynamical system and let F be a family
such that there exists x ∈ TranF (X, T ). Then orbT (x) ⊂ TranF (X, T ).
Proposition 3.3 ([29]). Assume that F admits a countable base H. Then
TrankF (X, T ) is a Gδ subset of X. Moreover, the following are equivalent:
(1) The system (X, T ) is kF-transitive,
(2) TrankF (X, T ) is a dense Gδ subset of X,
(3) TrankF (X, T ) 6= ∅.
3.4. Transitive sensitivity and sensitive compactness. A dynamical system
(X, T ) is transitively sensitive if there exists δ > 0 such that ST (W, δ) ∩ NT (U, V ) 6=
∅ for any opene subsets U, V, W of X. Recall that it is sensitively compact if there
exists δ > 0 such that for any point x ∈ X the set ωST (δ) (x) is nonempty. Sometimes
in such cases we will also say that (X, T ) is transitively sensitive with a sensitive
constant δ and (X, T ) is sensitively compact with a sensitive constant δ. Then
Theorem 3.4 ([29]). Let (X, T ) be a minimal system. Then the following condi-
tions are equivalent:
1. (X, T ) is multi-sensitive.
1302 SERGIĬ KOLYADA
Recall that (X, T ) is Li-Yorke sensitive if there exists δ > 0 such that for every
x ∈ X and every neighborhood Ux of x, there exist y ∈ Ux such that (x, y) is a
proximal pair while lim supn→+∞ d(T n x, T n y) > δ.
Let (X, T ) be a dynamical system which is compact with respect to a (dynamical)
Furstenberg family F. Say, F is NT or ST . Then it is not so hard to show that in
that case for any point x ∈ X there exist a minimal subset Mx of X and a point
z ∈ Mx such that
nT (x, Gz ) ∩ F 6= ∅
for any neighborhood G of z in X and any F ∈ F. If A is a set in X, by Bε [A]
we will denote the union of all open balls of radius ε > 0 whose centers run over
A. Similarly as in the proof of [28, Lemma 3.12] one can show that for any x ∈ X
and any Bε [Mx ] the set nT (x, Bε [Mx ]) is thickly syndetic. Let nT be the set of all
subsets of Z+ containing some nT (x, Bε [Mx ]), where x is a point of X, ε > 0 and
Mx ⊂ ωT (x) is a minimal subset of X.
Notice that Figure 12 presents also an open question. Are all transitively com-
pact systems sensitively compact? More precisely, does there exist a proximal,
transitively compact system which is not sensitively compact? Nevertheless, for
topologically transitive, non-proximal dynamical systems we have more clear situ-
ation, see Figure 13.
4. Slovak spaces. This section is mostly based on the papers [35] and [13].
compact-open topology. Similarly, let H(X) be the topological group of all home-
omorphisms X → X.
A dynamical system (X, T ) gives rise to several “hyper-systems”. For example,
T acts naturally on 2X , the space of all compact subsets of X equipped with the
Hausdorff metric. The induced map here is the continuous map T̃ : 2X → 2X given
by T̃ (A) = T (A) for each A ∈ 2X .
In the same spirit (although now the phase space is not necessarily compact),
jointly with J. Auslander and L’. Snoha in [6], we introduced and started to study
the “hyper-system” (S(X), FT ), where the transformation
FT : S(X) → S(X) is defined by FT ϕ = T ◦ ϕ for any ϕ ∈ S(X).
We called this system the functional envelope of (X, T ). Since the metric space X is
compact, the compact-open topology on S(X) is compatible both with the uniform
metric and with the Hausdorff metric (applied to the graphs of maps). If S(X) is
viewed as a metric space with the uniform or Hausdorff metric, the notation SU (X)
or SH (X) is used. If the properties under consideration do not depend on which of
the two metrics is used in S(X), we will sometimes keep the notation S(X) instead
of SU (X) or SH (X).
Why the name “functional envelope”? Because the system (S(X), FT ) contains
(properly, if card X ≥ 2) an isomorphic copy of (X, T ). The map ι : X → S(X)
sending a point a ∈ X to the constant map x 7→ a, x ∈ X, is an isometry (regardless
of whether the uniform metric or the Hausdorff metric is used in S(X)) and also a
topological conjugacy, i.e., ι ◦ T = FT ◦ ι.
DYNAMICAL TOPOLOGY 1307
The “uniform” functional envelope (SU (X), FT ) and the “Hausdorff” functional
envelope (SH (X), FT ) may differ in some dynamical properties which depend on
the metric (for example the topological entropy in non-compact systems does). On
the other hand, the essential dynamical properties of these functional envelopes do
not depend on the metric.
Now suppose additionally that the map T : X → X is a homeomorphism. Then
the space H(X) is invariant under FT and (H(X), FT ) becomes another kind of a
“functional envelope”. Here again
FT : H(X) → H(X) is defined by FT ϕ = T ◦ ϕ for any ϕ ∈ H(X).
One of the main topics of our research in this section is to study relations between
some properties (structure) of the topological semigroup S(X), group H(X) and
possible values of the topological entropy (and/or some other chaotic properties) of
its elements (continuous maps and homeomorphisms, respectively).
In [35] we have mostly considered the following two questions: 1) when does
a compact metric space admit a continuous map (homeomorphism) with positive
topological entropy? 2) when does the existence of a positive-entropy continuous
map on a compact metric space imply the existence of a +∞-entropy continuous
map? We have proved there the following.
Theorem 4.1 ([35]). Let X be a compact metric space. If S(X) is compact, then
for any f ∈ S(X) topological entropy of (X, f ) and topological entropy of the func-
tional envelope (S(X), Ff ) are zero. If H(X) is compact, then for any f ∈ H(X)
topological entropy of (X, f ) is also zero.
At present, not so much is known about the topological structure of compact
semigroups S(X) and groups H(X). The compactness of S(X) and H(X) is not
a very strict condition and occurs sometimes, because both of them may be “very
small”. Recall that a topological space X is said to be rigid (for homeomorphisms),
if the full topological homeomorphism group H(X) contains only the identity. De
Groot and Wille in [22] showed the existence of rigid spaces even as locally con-
nected one-dimensional continua (Peano curves) of the plane. The main idea of the
construction of such a space is similar to that of the Sierpinski carpet, which is a
square with interiors of a dense family of subsquares removed, see Figure 14. The
Sierpinski carpet, also known as the Sierpinski universal curve, is a one-dimensional
planar Peano continuum.
Figure 14. The first 4 steps in the construction of the Sierpinski carpet.
One of the De Groot – Wille rigid plane locally connected one-dimensional con-
tinua is a disc with interiors of a dense family of propellers (with different numbers
of blades) removed, see Figure 15. To be more precise, by [22], consider a disc D in
the plane. Let {ai } be a countable dense subset of the interior of D. We define a
sequence of “propellers” in D. The first is bounded by a two-bladed curve having
a1 as its centre, which avoids the boundary of D. Suppose the first n − 1 propellers
have already been defined. Let a0n be the first member of the sequence {ai } which
1308 SERGIĬ KOLYADA
As a curiosity, notice that the Julia set from Figure 16 (a picture by Volodymyr
Nekrashevych), though not being rigid, reminds the De Groot – Wille rigid conti-
nuum.
Figure 16. The Julia set for the map z 7→ (z 2 + 0.3 + 0.05i)/(z 2 − 1).
Take the suspension flow over h (with ceiling function ≡ 1), Φt , t ∈ R defined on
SOL by the formula
Φt (y, s) := (h[t+s] y, {t + s}),
where [ · ] and {·} denote the integer and fractional parts of a real number, respec-
tively.
Step 3. It is well known that there exists t0 ∈ R such that the map
T := time t0 -map of the flow Φ
is a minimal homeomorphism on SOL (see e.g. [16]).
Step 4. This is the main technical step. The Slovak space constructed in [13] will
be a subset of the cylinder SOL × [0, 1]. The main element of this construction is a
topologist’s sine curve (a subset of the plane that is the union of the graph of the
function f (x) = sin(1/x), 0 < x ≤ 1 with the segment −1 ≤ y ≤ 1 of the y-axis).
1. The continuum SOL has uncountably many composants (orbits of the flow
Φ); choose a composant γ and a point x0 ∈ γ. (Recall that the composant of
a point p in a continuum A is the union of all proper subcontinua of A that
contain p. If a continuum is indecomposable, then its composants are pairwise
disjoint. The composants of a continuum are dense in that continuum.)
2. On a closed arc around x0 (minus the point x0 itself) and lying in γ, we define
a function which looks like a one-sided topologist’s sine curve (values in [0, 1],
wiggles of height 1 in any left neighbourhood of x0 , constant value 0 to the
right of x0 ). It is continuous and not defined at x0 .
3. Extend itPto a continuous function f : SOL \ {x0 } → [0, 1].
n
P F = n∈Z an f ◦ T , where the coefficients an are all strictly positive with
4. Let
n∈Z an = 1 and satisfying some technical assumptions (F is defined on SOL
minus the T -orbit of x0 ).
5. Then one can show that both the mapping (x, F (x)) 7→ (T x, F (T x)) and its
inverse are uniformly continuous homeomorphisms of the graph of F . There-
fore, the map (x, F (x)) 7→ (T x, F (T x)) extends to a homeomorphism T̄ of F̄
(=the closure of the graph of F ).
6. F̄ ⊆ SOL × [0, 1] is our S lovak space. The composant γ̄ of F̄ “above” γ has
basicly the shape as in Figure 18. The other composants of F̄ are continuous
In [4] Akin and Rautio considered the related problem of when a space X admits
a homeomorphism f so that (X, f ) is the omega limit set in a larger system. As we
will see, their results are somewhat different from the map case. They reinterpreted
the problem by using some other notion. Given ε ≥ 0, a finite or infinite sequence
{xn ∈ X} with at least two terms is an ε-chain for (X, f ) if d(f xk , xk+1 ) ≤ ε for all
terms xk of the sequence (except the last one). The system (X, f ) is called chain
transitive when every pair of points of X can be connected by some finite ε-chain
for every positive ε. A subset A ⊂ X is called a chain transitive subset when it is
closed and f -invariant (i.e., f (A) = A) and the subsystem (A, f ) is chain transitive.
It is well known that chain transitivity and f -connectedness are equivalent con-
cepts (see [4] for details). Akin and Rautio generalized several known results (in
particular, regarding rigid spaces). They proved the existence of the following com-
pact metric spaces:
1. Suppose that G is a finitely generated (and hence countable) group. Then
there exists a space X such that the homeomorphism group H(X) is isomor-
phic to G and every f ∈ H(X) is chain transitive.
2. There exists a space X such that the homeomorphism group H(X) contains a
nontrivial path-connected subgroup and every f ∈ H(X) is chain transitive.
3. There exists a space X such that the homeomorphism group H(X) is isomor-
phic to the homeomorphism group of the Cantor set and every f ∈ H(X) is
chain transitive.
Akin and Rautio in [4] also extended the result of Downarowicz, Snoha and Ty-
woniuk on the existence of Slovak spaces by showing the existence of spaces they
call Slovakian spaces. These are infinite compact metric spaces with H(X) non-
trivial and such that every homeomorphism other than the identity is topologically
transitive. The only Slovakian spaces known to them are variations on the original
construction of [13]; all of these have homeomorphism group isomorphic to Z.
5. Appendix.
5.1. Basic concepts in topological dynamics. Given a dynamical system (X, T ),
recall that x ∈ X is a fixed point if T (x) = x, and an F-transitive point of (X, T )
[39] if NT (x, U ) ∈ F for any opene subset U of X. It is a trivial observation that
if a family F admits an F-transitive dynamical system (X, T ) without isolated
points, then F is free. Since k(kF) = F, it is easy to see that x ∈ X is an F-
transitive point of (X, T ) if and only if ωkF (x) = X. Denote by TranF (X, T ) the
set of all F-transitive points of (X, T ). The system (X, T ) is F-point transitive if
TranF (X, T ) 6= ∅, and is F-transitive if NT (U, V ) ∈ F for any opene subsets U, V of
X. Write Tran(X, T ) = TranP+ (X, T ) for short, and we also call the point x transi-
tive if x ∈ Tran(X, T ) or, equivalently, if its orbit orbT (x) = {T n x : x = 0, 1, 2, . . . }
is dense in X. The system (X, T ) is transitive if and only if Tran(X, T ) is a dense
Gδ subset of X.
A subset A of X is T -invariant if T (A) = A, and positively T -invariant if
T (A) ⊂ A. If A is a closed, nonempty, T -invariant subset then (A, T |A ) is called
the associated subsystem. A minimal subset of X is a closed, nonempty, T -invariant
subset such that the associated subsystem is minimal. Clearly, (X, T ) is minimal if
and only if Tran(X, T ) = X, if and only if it admits no proper, closed, nonempty,
positively T -invariant subset. A point x ∈ X is called minimal if it lies in some
minimal subset. Zorn’s Lemma implies that every closed, nonempty, positively
T -invariant set contains a minimal set.
DYNAMICAL TOPOLOGY 1313
Proposition 5.2 ([29]). Let (X, T ) be a dynamical system and let F be a family.
(i) If F is free, then ωF (x) ⊂ ωT (x) for any x ∈ X. Moreover, if (X, T ) has a
nonrecurrent point, then the converse implication is true.
(ii) If F is free and has FIP then it has SFIP.
5.3. Almost one-to-one maps. Let ϕ : X → Y be a continuous surjective map
from a compact metric space X onto a compact Hausdorff space Y . Recall that ϕ
is almost open if ϕ(U ) has a nonempty interior in Y for any opene U ⊂ X. Note
that each factor map between minimal systems is almost open [5, Theorem 1.15].
In particular, for a minimal system (X, T ) the map T : X → X is almost open [38].
Denote by Y0 the set of all points y ∈ Y whose preimage is a singleton. Then Y0
is a Gδ subset of Y , because
\ 1
−1 −1
Y0 = {y ∈ Y : ϕ (y) is a singleton} = y ∈ Y : diam(ϕ (y)) <
n
n∈N
−1
and the map y 7→ diam(ϕ (y)) is upper semi-continuous. Here, we denote by
diam(A) the diameter of a subset A ⊂ X. Recall that the function f : Y → R+ is
upper semi-continuous if lim sup f (y) ≤ f (y0 ) for each y0 ∈ Y .
y→y0
Denote by X0 the set of all points x ∈ X such that the preimage of ϕ(x) is a
singleton. Then X0 = ϕ−1 (Y0 ) is a Gδ subset of X.
We call ϕ weakly almost one-to-one if Y0 is dense in Y , and almost one-to-one 3
if X0 is dense in X. It is not hard to show that: if ϕ is weakly almost one-to-one,
then for any δ > 0 and any opene subset U of Y there exists opene V ⊂ U with
diam(ϕ−1 V ) < δ; and if ϕ is almost one-to-one, then for any opene subset U ∗ of
X there exists an opene subset V ∗ of Y with ϕ−1 V ∗ ⊂ U ∗ . Clearly almost one-
to-one is stronger than weakly almost one-to-one. For instance, define T x = 2x
for x ∈ [0, 12 ] and T x = 1 for x ∈ [ 12 , 1]. Then T : [0, 1] → [0, 1] is weakly almost
one-to-one but not almost one-to-one.
For each minimal system (X, T ), the map T : X → X is weakly almost one-to-one
[38, Theorem 2.7], and in fact almost one-to-one [30, Proposition 2.3]. The following
result characterizes the relationship between weakly almost one-to-one and almost
one-to-one, which extends [30, Proposition 2.3].
Proposition 5.3 ([29]). Let ϕ : X → Y be a continuous surjective map from a
compact metric space X onto a compact Hausdorff space Y . Then ϕ is almost
one-to-one if and only if it is almost open and weakly almost one-to-one.
As a direct corollary, we have:
Corollary 5.4 ([29]). Let ϕ : X → Y and π : Y → Z be continuous surjective
maps between compact metric spaces. Then the composition map π ◦ ϕ : X → Z is
almost one-to-one if and only if both ϕ and π are almost one-to-one.
Let π : (X, T ) → (Y, S) be a factor map between dynamical systems. If the
map π : X → Y is almost one-to-one (weakly almost one-to-one, respectively),
then we also say that (X, T ) is an almost one-to-one extension (a weakly almost
one-to-one extension, respectively) of (Y, S). The main result of [30] states that a
minimal system is either multi-sensitive or a weakly almost one-to-one extension of
its maximal equicontinuous factor. This is an analog of the well-known Auslander-
Yorke dichotomy theorem: a minimal system is either sensitive or equicontinuous.
3 Here we use the concept of almost one-to-one following [2]. The concept of almost one-to-one
Acknowledgments. The author thanks very much Danylo Khilko and L’ubomı́r
Snoha for useful remarks and discussions. The author was supported by Max-
Planck-Institut für Mathematik (MPIM, Bonn) and Institut des Hautes Études
Scientifiques (IHES, Bures-sur-Yvette); he acknowledges the hospitality of the In-
stitutes.
1316 SERGIĬ KOLYADA
REFERENCES
[1] E. Akin, Recurrence in Topological Dynamics. Furstenberg families and Ellis actions, The
University Series in Mathematics, Plenum Press, New York, 1997.
[2] E. Akin and E. Glasner, Residual properties and almost equicontinuity, J. Anal. Math., 84
(2001), 243–286.
[3] E. Akin and S. Kolyada, Li-Yorke sensitivity, Nonlinearity, 16 (2003), 1421–1433.
[4] E. Akin and J. Rautio, Chain transitive homeomorphisms on a space: All or none, Pacific J.
Math., 291 (2017), 1–49.
[5] J. Auslander, Minimal Flows and Their Extensions, North-Holland Mathematics Studies, vol.
153, North-Holland Publishing Co., Amsterdam, 1988, Notas de Matemática [Mathematical
Notes], 122.
[6] J. Auslander, S. Kolyada and L’. Snoha, Functional envelope of a dynamical system, Nonlin-
earity, 20 (2007), 2245–2269.
[7] J. Auslander and J. A. Yorke, Interval maps, factors of maps, and chaos, Tôhoku Math. J.,
(2) 32 (1980), 177–188.
[8] H. Cook, Continua which admit only the identity mapping onto non-degenerate subcontinua,
Fund. Math., 60(1967), 241–249.
[9] T. Das, E. Shah and L’. Snoha, (Non-)expansivity in functional envelopes, J. Math. Anal.
Appl., 410 (2014), 1043–1048.
[10] T. Dobrowolski, Examples of topological groups homeomorphic to l2f , Proc. Amer. Math.
Soc., 98 (1986), 303–311.
[11] Y. N. Dowker and F. G. Friedlander, On limit sets in dynamical systems, Proc. London Math.
Soc., (3) 4 (1954), 168–176.
[12] T. Downarowicz, Survey of odometers and Toeplitz flows, Algebraic and Topological Dynam-
ics, Contemp. Math., vol. 385, Amer. Math. Soc., Providence, RI, 2005, 7–37.
[13] T. Downarowicz, L’. Snoha and D. Tywoniuk, Minimal spaces with cyclic group of homeo-
morphisms, J. Dynam. Differential Equations, 29 (2017), 243–257.
[14] F. T. Farrell and A. Gogolev, The space of Anosov diffeomorphisms, J. Lond. Math. Soc., (2)
89 (2014), 383–396.
[15] A. Fathi, Structure of the group of homeomorphisms preserving a good measure on a compact
manifold, Ann. Sci. École Norm. Sup., 13 (1980), 45–93.
[16] B. R. Fayad, Topologically mixing and minimal but not ergodic, analytic transformation on
T5 , Bol. Soc. Brasil. Mat. (N.S.), 31 (2000), 277–285.
[17] H. Furstenberg, Disjointness in ergodic theory, minimal sets, and a problem in Diophantine
approximation, Math. Systems Theory, 1 (1967), 1–49.
[18] H. Furstenberg, Recurrence in Ergodic Theory and Combinatorial Number Theory, Princeton
University Press, Princeton, N.J., 1981, M. B. Porter Lectures.
[19] H. Furstenberg and B. Weiss, Topological dynamics and combinatorial number theory, J.
Analyse Math., 34 (1978), 61–85 (1979).
[20] P. Gartside and A. Glyn, Autohomeomorphism groups, Topology Appl., 129 (2003), 103–110.
[21] E. Glasner and B. Weiss, Sensitive dependence on initial conditions, Nonlinearity, 6 (1993),
1067–1075.
[22] J. de Groot and R. J. Wille, Rigid continua and topological group-pictures, Arch. Math., 9
(1958), 441–446.
[23] J. de Groot, Groups represented by homeomorphism groups, Math. Ann., 138 (1959), 80–102.
[24] J. Guckenheimer, Sensitive dependence to initial conditions for one-dimensional maps, Comm.
Math. Phys., 70 (1979), 133–160.
[25] S. Harada, Remarks on the topological group of measure preserving transformations, Proc.
Japan Acad., 27 (1951), 523–526.
[26] K. H. Hofmann and S. A. Morris, Compact homeomorphism groups are profinite, Topology
Appl., 159 (2012), 2453–2462.
[27] K. H. Hofmann and S. A. Morris, Representing a profinite group as the homeomorphism
group of a continuum, preprint, arXiv:1108.3876.
[28] W. Huang, D. Khilko, S. Kolyada and G. Zhang, Dynamical compactness and sensitivity, J.
Differential Equations, 260 (2016), 6800–6827.
[29] W. Huang, D. Khilko, S. Kolyada, A. Peris and G. Zhang, Finite intersection property and
dynamical compactness, J. Dynam. Differential Equations, 30 (2018), 1221–1245.
DYNAMICAL TOPOLOGY 1317
[30] W. Huang, S. Kolyada and G. Zhang, Analogues of Auslander-Yorke theorems for multi-
sensitivity, Ergodic Theory Dynam. Systems, 38 (2018), 651–665.
[31] M. Keane, Contractibility of the automorphism group of a nonatomic measure space, Proc.
Amer. Math. Soc., 26 (1970), 420–422.
[32] S. Kolyada, M. Misiurewicz and L’. Snoha, Spaces of transitive interval maps, Ergodic Theory
Dynam. Systems, 35 (2015), 2151–2170.
[33] S. Kolyada, M. Misiurewicz and L’. Snoha, Loops of transitive interval maps, Dynamics and
numbers, Contemp. Math., Amer. Math. Soc., Providence, RI, 669 (2016), 137–154.
[34] S. Kolyada and O. Rybak, On the Lyapunov numbers, Colloq. Math., 131 (2013), 209–218.
[35] S. Kolyada and J. Semikina, On topological entropy: When positivity implies +infinity, Proc.
Amer. Math. Soc., 143 (2015), 1545–1558.
[36] S. Kolyada and L’. Snoha, Topological entropy of nonautonomous dynamical systems, Random
Comput. Dynam., 4 (1996), 205–233.
[37] S. Kolyada and L’. Snoha, Some aspects of topological transitivity – a survey, Iteration Theory
(ECIT 94) (Opava), Grazer Math. Ber., 334 (1997), 3–35.
[38] S. Kolyada, L’. Snoha and S. Trofimchuk, Noninvertible minimal maps, Fund. Math., 168
(2001), 141–163.
[39] J. Li, Transitive points via Furstenberg family, Topology Appl., 158 (2011), 2221–2231.
[40] J. Li and X. Ye, Recent development of chaos theory in topological dynamics, Acta Math.
Sin. (Engl. Ser.), 32 (2016), 83–114.
[41] M. Matviichuk, On the dynamics of subcontinua of a tree, J. Difference Equ. Appl., 19 (2013),
223–233.
[42] T. K. S. Moothathu, Stronger forms of sensitivity for dynamical systems, Nonlinearity, 20
(2007), 2115–2126.
[43] N. T. Nhu, The group of measure preserving transformations of the unit interval is an absolute
retract, Proc. Amer. Math. Soc., 110 (1990), 515–522.
[44] K. E. Petersen, Disjointness and weak mixing of minimal sets, Proc. Amer. Math. Soc., 24
(1970), 278–280.
[45] P. Raith, Topological transitivity for expanding piecewise monotonic maps on the interval,
Aequationes Math., 57 (1999), 303–311.
[46] D. Ruelle, Dynamical systems with turbulent behavior, Mathematical Problems in Theoretical
Physics (Proc. Internat. Conf., Univ. Rome, Rome, 1977), Lecture Notes in Phys., vol. 80,
Springer, Berlin-New York, 1978, 341–360.
[47] A. N. Šarkovskiı̆, On attracting and attracted sets, Dokl. Akad. Nauk SSSR, 160 (1965),
1036–1038. In Russian; translated in Soviet Math. Dokl., 6 (1965), 268–270.
[48] A. N. Šarkovskiı̆, Continuous mapping on the limit points of an iteration sequence, Ukrain.
Mat. Ž., 18 (1966), 127–130.
[49] A. N. Šarkovskiı̆, S. F. Kolyada, A. G. Sivak and V. V. Fedorenko, Dynamics of One-
Dimensional Maps, Kiev, 1989.
[50] P. Walters, An Introduction to Ergodic Theory, Springer-Verlag, New York-Berlin, 1982.
[51] T. Yagasaki, Weak extension theorem for measure-preserving homeomorphisms of noncompact
manifolds, J. Math. Soc. Japan, 61 (2009), 687–721.