S-Expansiveness and Zip Shift Maps in Symbolic Dynamics (original) (raw)

, P. Mehdipour Departamento de Matematica-Universidade Federal de Viçosa pouya@ufv.br and W. Vargas Departamento de Matematica-Universidade Federal de Viçosa walterhv@ufv.br

Abstract.

In this paper, we introduce the concept of S-expansiveness for local homeomorphisms and demonstrate that a class of extended symbolic dynamics, known as zip shift maps, are S-expansive and possess the shadowing property. Furthermore, we prove that any S-expansive local homeomorphism is a factor of a zip shift map.

Key words and phrases:

local homeomorphisms, extended symbolic dynamics, shadowing property, expansivity

1991 Mathematics Subject Classification:

Primary: 37Dxx. Secondary: 37D20, 37D45, 26A18, 28D80, 28D05.

1. Introduction

Topological dynamics, which emerged between 1920 and 1930, studies the orbit trajectories within the framework of their topological behavior. After the 1960, research on the topological dynamics of homeomorphisms [1] led to significant advances and played a crucial role in the study of differentiable dynamical systems and ergodic theory [3].

Key properties studied in topological dynamics include the shadowing property and expansivity of homeomorphisms on compact metric spaces. The latter indicates sensitivity to initial conditions, which is a hallmark of chaotic systems. Notably, for expansive systems, the shadowing property is equivalent to topological stability, and all such properties are preserved under topological conjugacy.

One of the earliest examples of expansive homeomorphisms exhibiting the shadowing property is the two-sided shift maps defined on shift spaces [1], [10]. The shift maps are suitable examples in which demonstrate the interplay between expansiveness and the shadowing property in dynamical systems. The topological conjugacy (or semi-conjugacy) with shift maps serves as a primary tool for investigating these properties within dynamical systems theory.

In [5, 6], the authors extend the two-sided shift spaces into a new construct called zip shift space. The associated map, referred to as the zip shift, serves as a local homeomorphism and shows significant potential in preserving key properties of topological dynamics. In this paper, we introduce the concept of bilateral S-expansivity and demonstrate that zip shift maps are not only S-expansive but also exhibit the shadowing property. Moreover, we show that S-expansive m-to-1 local homeomorphisms are, in fact, factors of zip shift maps. We hope that these initial properties highlight the robustness and significance of zip shifts in the study of topological dynamics, making them a valuable tool for further exploration and understanding of the dynamical behavior of non-invertible maps.

2. Preliminary and Main results

Let (X,d)(X,d) be a compact metric space and f:X→Xf:X\to X a continuous map. We set, Xf={x~=(xn)|f​(xn)=xn+1,n∈ℤ}X^{f}=\{\tilde{x}=(x_{n})|f(x_{n})=x_{n+1},\,n\in\mathbb{Z}\} as the inverse limit space corresponded to XX and f~:Xf→Xf\tilde{f}:X^{f}\to X^{f} as its (bilateral) natural extension. The future orbit of a point over f,f, the O+​(x)={fn​(x)|n∈ℤ+}O^{+}(x)=\{f^{n}(x)|n\in\mathbb{Z}^{+}\} is unique, but with infinitely many, different pre-histories, in which creates infinitely many points in π−1​(x)⊂Xf.\pi^{-1}(x)\subset X^{f}. Let d(.,.)d(.,.) represent the (Riemannian) metric on XX. For x~=(xn),y~=(yn)∈Xf\tilde{x}=(x_{n}),\tilde{y}=(y_{n})\in X^{f}, we consider d~\tilde{d} as a natural metric on XfX^{f}.

| d~​(x~,y~):=∑n=−∞∞2−|n|​d​(xn,yn).\tilde{d}(\tilde{x},\tilde{y}):=\sum_{n=-\infty}^{\infty}2^{-|n|}d(x_{n},y_{n}). | | -------------------------------------------------------------------------------------------------------------------------------- |

The metric topology induced by d~\tilde{d} and the product topology are equivalent on XfX^{f}.

Definition 2.1 (m-to-1 local homeomorphism).

We say that, the continuous map f:X→Xf:X\to X is an m-to-1 local homeomorphism, whenever XX is a compact metric space and there exist m∈ℕm\in\mathbb{N} disjoint connected subsets Xi⊂XX_{i}\subset X that cover XX with mm local maps fi:Xi→Xf_{i}:X_{i}\to X such that (for all i=1,…,mi=1,\dots,m) fif_{i} is a homeomorphism.

Once working with closed Riemannian manifolds, the m-covering maps are examples of such local homeomorphisms. We have the following Theoem from Eilenberg [1].

Theorem 2.2.

Let XX be compact and f:X→Xf:X\to X a continuous surjection. If ff is a local homeomorphism, then there exist two positive numbers λ\lambda and μ\mu such that each D⊂XD\subset X with diameter less than λ\lambdadetermines a decomposition of the set f−1​(D)f^{-1}(D) with the following properties:

  1. (1)
    f−1​(D)=D1∪⋯∪Dk;f^{-1}(D)=D_{1}\cup\cdots\cup D_{k};
  2. (2)
    ff maps each DiD_{i} homeomorphically onto DD,
  3. (3)
    if i≠ji\neq j then no point of DiD_{i} is closer than 2​μ2\mu to a point of DjD_{j};
  4. (4)
    for every θ>0\theta>0 there exists some 0<ϵ<λ0<\epsilon<\lambda such that if d​i​a​m​(D)<ϵdiam(D)<\epsilon (diameter), then d​i​a​m​(Dj)<θdiam(D_{j})<\theta for all i=1,⋯,ki=1,\cdots,k.

If, in addition, XX is connected, then there is a constant m>0m>0 such that f:X→Xf:X\to X is an m-to-1 map.

We may use the following well known Lemma in the proof of Theorems. Let (X,d)(X,d) be a metric space. Then by diameter of a subset AA of XX, we mean: d​i​a​m​(A)=sup{d​(a,b):a,b∈A}.diam(A)=\sup\{d(a,b):a,b\in A\}.

Lemma 2.3 (Lebesgue number).

[1]Let XX be a compact metric space and α\alpha a finite open cover ofXX. Then there exists some δ=δ​(α)\delta=\delta(\alpha) such that for any A⊂XA\subset X if the diameter of AA is less than δ\delta then A⊂BA\subset B for some B∈αB\in\alpha. Such a number is called the Lebesgue number of α\alpha.

Definition 2.4 (Principal Domain).

We say that Y⊂XY\subset X is a principal domain when ϕ​(Y)=X\phi(Y)=X and ϕ|Y:Y→X\phi_{|_{Y}}:Y\to X is a homeomorphism.

Definition 2.5.

Let (X,d)(X,d) be a compact metric space. We say that a finite cover of XX denoted by 𝒟={D0,D1,…,DN−1}\mathcal{D}=\{D_{0},D_{1},...,D_{N-1}\} is a topological partition if:

  1. (1)
    Each DiD_{i} is open.
  2. (2)
    Di∩Dj=∅,i≠jD_{i}\cap D_{j}=\emptyset,i\neq j.
  3. (3)
    X=D0¯∪D1¯∪…∪DN−1¯X=\overline{D_{0}}\cup\overline{D_{1}}\cup...\cup\overline{D_{N-1}}.
Definition 2.6 (Domain topological partition [8]).

Consider an m-to-1 local homeomorphism ϕ\phi defined on a compact connected metric space XX with principal domains Pi,i=1,…,mP_{i},i=1,\dots,m. Let {Pi1,…,Piki},ki∈ℕ\{P_{i_{1}},...,P_{i_{k_{i}}}\},k_{i}\in\mathbb{N} be a topological partition for PiP_{i}. We define the Domain Topological Partition (DTP) as the collection of all elements in 𝒫={Pi1,…,Piki}i=1m\mathcal{P}=\{P_{i_{1}},...,P_{i_{k_{i}}}\}_{i=1}^{m}. Note that by Definition 2.4, for all kj∈ℕ,i=1,2,…,m,k_{j}\in\mathbb{N},i=1,2,\dots,m, the map f:Pikj→f​(Pikj)f:P_{i_{k_{j}}}\to f(P_{i_{k_{j}}}) is a homeomorphism.

Definition 2.7 (Image topological partition [8]).

Let ϕ:X→X\phi:X\rightarrow X be an m-to-1 local homeomorphism with a domain topological partition 𝒫={Pi1,…,Piki}i=1m\mathcal{P}=\{P_{i_{1}},...,P_{i_{k_{i}}}\}_{i=1}^{m} (ki∈ℕk_{i}\in\mathbb{N}). We say that 𝒬={𝒬1,…,𝒬k}\mathcal{Q}=\{\mathcal{Q}_{1},...,\mathcal{Q}_{k}\} is an Image topological partition (ITP) associated to 𝒫\mathcal{P}, if 𝒬j∈⋁ijϕ​(Pij)\mathcal{Q}_{j}\in\bigvee_{i_{j}}\phi(P_{i_{j}}) for Pij∈𝒫P_{i_{j}}\in\mathcal{P} is a topological partition for XX (by definition ϕ​(Pi)=X\phi(P_{i})=X when PiP_{i} is a principal domain).

Definition 2.8 (Extended topological partition [8]).

Let ϕ:X→X\phi:X\rightarrow X be an m-to-1 local homeomorphism with an image topological partition 𝒬={𝒬1,…,𝒬k}\mathcal{Q}=\{\mathcal{Q}_{1},...,\mathcal{Q}_{k}\}. The pre-image of 𝒬\mathcal{Q}, induces a domain topological partition which we call Extended Topological Partition (ETP) associated to 𝒬\mathcal{Q}.

The following Definition is adapted from [1] and [13].

Definition 2.10 (Generator/Weak generator).

Let f:X→Xf:X\to X be a local homeomorphism of a compact metric space. We say that a finite open cover α\alpha of XX is a generator (weak generator) for ff if for every bi-sequence {An}\{A_{n}\} of elements of α,\alpha, the⋂n=−∞∞f−n​(A¯n),\bigcap_{n=-\infty}^{\infty}f^{-n}(\overline{A}_{n}), ( ⋂n=−∞∞f−n​(An)\bigcap_{n=-\infty}^{\infty}f^{-n}(A_{n})) is at most one point. Here B¯\overline{B} denotes the closure of a subset BB and as ff is non-invertible, f−n​(x)=(fn)−1​(x)f^{-n}(x)=(f^{n})^{-1}(x).

Equivalently, one can say that α\alpha is a generator if for any ϵ>0\epsilon>0exists N>0N>0 such that the cover ⋁n=−NNf−n​(α)\bigvee_{n=-N}^{N}f^{-n}(\alpha); consists of open sets AnA_{n} each of which with diameter at most ϵ\epsilon, i.e. supid​i​a​m​(An)<ϵ\sup_{i}\,\,diam(A_{n})<\epsilon.

Definition 2.11.

Let α={A1,…,Ak}\mathcal{\alpha}=\{A_{1},\dots,A_{k}\} and β={B1,…,Bm}\mathcal{\beta}=\{B_{1},\dots,B_{m}\} be two topological partitions. We say that β\mathcal{\beta} is finer than α\mathcal{\alpha} (or β\beta is a refinement of α\alpha) if #​(α)<#​(β)\#(\mathcal{\alpha})<\#(\mathcal{\beta}) and for any Bi∈βB_{i}\in\mathcal{\beta} there exists some Aj∈αA_{j}\in\mathcal{\alpha} such that Bi⊆AjB_{i}\subseteq A_{j}.

For continuous non-invertible maps, it is commonly used the notions of positive expansivity and c-expansivity [1]. The latter means for any x~=(xi),y~=(yi)∈Xf\tilde{x}=(x_{i}),\tilde{y}=(y_{i})\in X^{f} if d​(xi,yi)<cd(x_{i},y_{i})<c for all i∈ℤi\in\mathbb{Z} then x~=y~\tilde{x}=\tilde{y}.

Definition 2.12.

Let (X,d)(X,d) be a compact metric space and f:X→Xf:X\to X a continuous dynamical system.

Definition 2.14.

Let (X,d)(X,d) be a compact metric space and f:X→Xf:X\to X a continuous dynamical system (local hommeomorphism).

Definition 2.15.

Let (X,d)(X,d) be a compact connected metric space and f:X→Xf:X\to X a continuous dynamical system.

Lemma 2.16.

Let MM be a compact metric space (without isolated points) and f:M→Mf:M\to M an onto continuous map. Then Transitivity implies the Topological transitivity and pre-transitivity implies the Topological pre-transitivity.

Proof.

As the first claim is a well known fact we just prove the secound part [2]. Let U,VU,V be two disjoint non-empty open subsets of MM. Then by pre-transitivity, there exists positive integers k1,k2k_{1},k_{2} such that f−k1​(x)∈Uf^{-k_{1}}(x)\in U and f−k2​(x)∈Vf^{-k_{2}}(x)\in V. One can without any loss of generality assume that k1<k2k_{1}<k_{2}. Then f−k3​(U)∩V≠∅,f^{-k_{3}}(U)\cap V\neq\emptyset, where k3=k2−k1k_{3}=k_{2}-k_{1}. ∎

Definition 2.17 (Sensitive dependence on initial conditions).

Let (X,d)(X,d) be a compact metric space and f:X→Xf:X\to X a continuous dynamical system. We say that ff has "sensitive dependence on initial conditions", if there is δ>0\delta>0 such that for each z∈Xz\in X and each neighborhood UU of zz there exist y∈Uy\in U and n∈ℤn\in\mathbb{Z} such that d​(fn​(z),fn​(y))>δd(f^{n}(z),f^{n}(y))>\delta. From definition it follows that X has no isolated points.

In [1] the authors show the following for hommeomorphisms. Here we show it for ff being a local homomorphism and XX a compact metric space. For n>0n>0, let Pern={x∈X;fn​(x)=x}\textrm{Per}_{n}=\{x\in X;f^{n}(x)=x\} and Per​(f)=⋃n=1∞Pern\textrm{Per}(f)=\displaystyle\bigcup_{n=1}^{\infty}\textrm{Per}_{n}.

Proposition 2.18.

Let f:X→Xf:X\to X be a local homeomorphism of a compact metric space. Suppose XX is an infinite set. If ff is topologically transitive and Per​(f)\textrm{Per}(f) is dense in XX, then ff has sensitive dependence on initial conditions.

Proof.

We first prove that Per​(f)≠X\textrm{Per}(f)\neq X. Otherwise, there exists n>0n>0such that int​(Pern)≠∅\textrm{int}(\textrm{Per}_{n})\neq\emptyset. Since ff is topologically transitive, then Pern=X\textrm{Per}_{n}=X and XX must be a finite set which is a contradiction.

Now we claim that there is δ0>0\delta_{0}>0 such that for any x∈Xx\in X, there existsp∈Per​(f)p\in\textrm{Per}(f) such that d​(x,𝒪f​(p))>δ0d(x,\mathcal{O}_{f}(p))>\delta_{0}. Otherwise, let q1q_{1}, q2q_{2} be two arbitrary disjoint periodic points. Let δ0=1/2​(𝒪f​(q1),𝒪f​(q2))\delta_{0}=1/2(\mathcal{O}_{f}(q_{1}),\mathcal{O}_{f}(q_{2})). For any x∈Xx\in X withd​(x,𝒪f​(q1))<δ0d(x,\mathcal{O}_{f}(q_{1}))<\delta_{0}, then d​(x,𝒪f​(q2))>δ0d(x,\mathcal{O}_{f}(q_{2}))>\delta_{0} which is a contradiction.

We will proof that ff has sensitive dependence on initial conditions with sensitivity constant δ=δ0/4\delta=\delta_{0}/4.

Let xx be an arbitrary point in XX and NN be some neighborhood of xx. Since the periodic points are dense, there exists a periodic point pp of period nn in U=N∩B​(x,δ)U=N\cap B(x,\delta), where B​(x,δ)B(x,\delta) is a ball of radius δ\delta. As we showed above, there exists a periodic point q∈Xq\in X satisfying d​(x,𝒪f​(q))>4​δd(x,\mathcal{O}_{f}(q))>4\delta. Let

V=⋂i=1nf−i​(B​(fi​(q,δ)))V=\displaystyle\bigcap_{i=1}^{n}f^{-i}(B(f^{i}(q,\delta)))

As ff is a local homeomorphism, VV is open and non-empty. Consequently, since ff is transitive, there exists y∈Uy\in U and a natural number kk such that fk​(y)∈Vf^{k}(y)\in V.

Now let j>0j>0 be the integer part of kn+1\dfrac{k}{n}+1. So 1≤n​j−k≤n1\leq nj-k\leq n. By construction, one has

fn​j​(y)=fn​j−k​(fk​(y))∈fn​j−k​(V)⊂B​(fn​j−k​(q),δ).f^{nj}(y)=f^{nj-k}(f^{k}(y))\in f^{nj-k}(V)\subset B(f^{nj-k}(q),\delta).

Now fn​j​(p)=pf^{nj}(p)=p,

d​(fn​j​(p),fn​j​(y))=d​(p,fn​j​(y))d(f^{nj}(p),f^{nj}(y))=d(p,f^{nj}(y))

and

d​(x,fn​j−k​(q))≤d​(x,p)+d​(p,fn​j​(y))+d​(fn​j​(y),fn​j−k​(q)).d(x,f^{nj-k}(q))\leq d(x,p)+d(p,f^{nj}(y))+d(f^{nj}(y),f^{nj-k}(q)).

Therefore, since p∈B​(x,δ)p\in B(x,\delta),

d​(fn​j​(p),fn​j​(y))>4​δ−δ−δ=2​δ.d(f^{nj}(p),f^{nj}(y))>4\delta-\delta-\delta=2\delta.

Thus, using triangle inequality again, we have either d​(fn​j​(x),fn​j​(y))>δd(f^{nj}(x),f^{nj}(y))>\delta or d​(fn​j​(x),fn​j​(p))>δd(f^{nj}(x),f^{nj}(p))>\delta. This completes the proof. ∎

It is easy to verify that topological pre-transitivity implies the topological transitivity. Indeed Proposition 2.18 is valid if one considers pre-transitivity instead of transitivity.

Proposition 2.19.

Let f:X→Xf:X\to X be a local homeomorphism of a compact metric space. Suppose XX is an infinite set. If ff is topologically pre-transitive and the set P​e​r​(f)Per(f) is dense in XX, then ff has sensitive dependence on initial conditions.

Theorem 2.20.

The following properties is valid for S-expansive maps.

  1. (1)
    If f:X→Xf:X\to X is S-expansive and YY is a closed invariant subset of XX (i.e. f±1​(Y)=Yf^{\pm 1}(Y)=Y), then f|Y:Y→Yf_{|_{Y}}:Y\to Y is S-expansive.
  2. (2)
    If fi:Xi→Xif_{i}:X_{i}\to X_{i} (i=1,2i=1,2) are S-expansive, then the local homeomorphismf1×f2:X1×X2→X1×X2f_{1}\times f_{2}:X_{1}\times X_{2}\to X_{1}\times X_{2} defined byf1×f2​(x,y)=(f1​(x),f2​(y))f_{1}\times f_{2}(x,y)=(f_{1}(x),f_{2}(y))is S-expansive. Every finite direct product of expansive homeomorphisms is expansive.
  3. (3)
    If XX is compact and f:X→Xf:X\to X is S-expansive, h∘f∘h−1:Y→Yh\circ f\circ h^{-1}:Y\to Yis S-expansive where h:X→Yh:X\to Y is a homeomorphism.
Proof.

The items (1)(1) and (2)(2) have a trivial proof. We prove item (3)(3). Let g=h∘f∘h−1g=h\circ f\circ h^{-1}. It is not difficult to verify that, h−1∘gn=fn∘h−1h^{-1}\circ g^{n}=f^{n}\circ h^{-1} for all n∈ℤn\in\mathbb{Z}. Let γ\gamma be the S-expansiveness constant of ff. If there exists x,y∈Xx,y\in X, such that d​(fn​(x),fn​(y))<γd(f^{n}(x),f^{n}(y))<\gamma, for all n∈ℤn\in\mathbb{Z}, then x=yx=y. By uniform continuity of hh and h−1h^{-1}, for all ϵ>0,\epsilon>0, in particular ϵ=γ,\epsilon=\gamma, there exists δ>0\delta>0 such that d​(gn​(x),gn​(y))<δd(g^{n}(x),g^{n}(y))<\delta implies d​(h−1∘gn​(x),h−1∘gn​(y))<γd(h^{-1}\circ g^{n}(x),h^{-1}\circ g^{n}(y))<\gamma.

We have

d​(h−1∘gn​(x),h−1∘gn​(y))=d​(fn∘h−1​(x),fn∘h−1​(y))<γd(h^{-1}\circ g^{n}(x),h^{-1}\circ g^{n}(y))=d(f^{n}\circ h^{-1}(x),f^{n}\circ h^{-1}(y))<\gamma

for all n∈ℤn\in\mathbb{Z}. Therefore, h−1​(x)=h−1​(y)h^{-1}(x)=h^{-1}(y), which implies x=yx=y. ∎

2.1. Zip shift spaces

In this section, we describe the zip shift maps[6]. Let Z={a1,a2,⋯,ak}Z=\{a_{1},a_{2},\cdots,a_{k}\} and S={0,1,…,N−1},S=\{0,1,\dots,N-1\}, be two collections of symbols that N≥kN\geq k and τ:S→Z\tau:S\to Z a surjective map. Consider ∏−∞+∞S\prod_{-\infty}^{+\infty}S and let t¯=(ti)∈∏−∞+∞S\bar{t}=(t_{i})\in\prod_{-\infty}^{+\infty}S. Then to any such point t¯=(ti)i∈ℤ\bar{t}=(t_{i})_{i\in\mathbb{Z}} correspond a point x¯=(xi)i∈ℤ\bar{x}=(x_{i})_{i\in\mathbb{Z}}, such that

xi={ti∈S ∀i≥0τ(ti)∈Z ∀i<0.x_{i}=\left\{\begin{tabular}[]{ll}$t_{i}\in S\hskip 19.91692pt$\,\,\,\,\quad\quad\quad$\forall i\geq 0$\\ tau(ti)inZhskip19.91692pt\tau(t_{i})\in Z\hskip 19.91692pttau(ti)inZhskip19.91692pt\,\,\,\quad\quad$\forall i<0$.\end{tabular}\right. (2.1)

DefineΣ={x¯=(xi)i∈ℤ:xi​ satisfies​(2.1)}.\Sigma=\{\bar{x}=(x_{i})_{i\in\mathbb{Z}}\ :x_{i}\textrm{\ satisfies}\,(\ref{Z})\}.We define στ:Σ→Σ\sigma_{\tau}:\Sigma\to\Sigma as follows.

(στ(x¯))i={xi+1ifi≠−1,τ(x0)ifi=−1.\displaystyle(\sigma_{\tau}(\bar{x}))_{i}=\left\{\begin{tabular}[]{ll}$x_{i+1}\,\,\,\quad\quad\text{if}\,\,i\neq-1,$\\ tau(x0)quadquadtextif,,i=−1.\tau(x_{0})\quad\quad\text{if}\,\,i=-1.tau(x0)quadquadtextif,,i=−1.\end{tabular}\right. (2.4)

We call this map zip shift. If Σ\Sigma is closed and invariant under στ\sigma_{\tau}, call it Zip shift Space. One can equip Σ\Sigma with a metric dd. Let M:Σ×Σ→ℕ∪{0}M:\Sigma\times\Sigma\rightarrow\mathbb{N}\cup\{0\} be given as,

| M​(x,y)={∞, if ​x=ymin⁡{|i|;{xi}≠{yi}}, if ​x≠y\displaystyle M(x,y)=\left\{\begin{array}[]{lr}\infty,&\text{ if }x=y\\ \min\{|i|;\,\{x_{i}\}\neq\{y_{i}\}\},&\text{ if }x\neq y\end{array}\right. | (2.7) | | ----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- | ----- |

then d​(x,y)=1λM​(x,y)d(x,y)=\frac{1}{\lambda^{M(x,y)}} defines a metric on Σ\Sigma, where we choose λ=#​(S)\lambda=\#(S). Throughout the remainder of this paper, we work within the metric space(Σ,d)(\Sigma,d).

Proposition 2.21.

[9]The followings are valid for στ:Σ→Σ\sigma_{\tau}:\Sigma\to\Sigma.

Proposition 2.22.

[5]Periodic points of a zip shift map are dense in the zip shift space. P​e​r​(στk)=(l)kPer(\sigma_{\tau}^{k})=(l)^{k}. Moreover the zip shift map is topologically mixing.

The basic cylinders on a zip shift space are defined as follows.

| Cisi={x∈Σ|xi=si,i∈ℤ,si∈Z,if​i<0,and​si∈S,if​i≥0}.C_{i}^{s_{i}}=\{x\in\Sigma\,|\,x_{i}=s_{i},i\in\mathbb{Z},\,s_{i}\in Z,\,\text{if}\,i<0,\,\text{and}\,s_{i}\in S,\,\text{if}\,i\geq 0\}. | (2.8) | | ----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- | ----- |

The set CisiC_{i}^{s_{i}} presents the set of all sequences, that have sis_{i} in the i−i-th entry. For i,n∈ℤi,n\in\mathbb{Z} and ℓ∈ℕ∪{0}\ell\in\mathbb{N}\cup\{0\}, one can define a general cylinder set as follows.

| Ci1​…,iks1,…,sk={(tn)∈Σ|ti1=s1,⋯,tik=sk;s1,…,sk∈S∪Z},C_{i_{1}\dots,i_{k}}^{s_{1},\dots,s_{k}}=\{(t_{n})\in\Sigma|t_{i_{1}}=s_{1},\cdots,t_{i_{k}}=s_{k};s_{1},\dots,s_{k}\in S\cup Z\}, | | ------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- |

where sj∈Z,if​ij<0,and​sj∈S,if​ij≥0s_{j}\in Z,\,\text{if}\,i_{j}<0,\,\text{and}\,s_{j}\in S,\,\text{if}\,i_{j}\geq 0 (1≤j≤k1\leq j\leq k).

Such cylinder sets are clopen subsets in the metric topology of (Σ,d)(\Sigma,d). The set of all such cylinder sets, form a basis for the topology. In fact, (Σ,d)(\Sigma,d) is compact, totally disconnected and perfect, which means it is a Cantor set.

Example 2.23.

Let S={0,1,2,3}S=\{0,1,2,3\} and Z={a,b}Z=\{a,b\}. Define the corresponding transition map τ:S→Z\tau:S\to Z as τ​(0)=τ​(2)=a\tau(0)=\tau(2)=a and τ​(1)=τ​(3)=b\tau(1)=\tau(3)=b. Let x¯=(xn)n∈ℤ=(⋯​a​b​a​b​b​.​ 1 0 3 1 1 2​⋯).\bar{x}=(x_{n})_{n\in\mathbb{Z}}=(\cdots a\,b\,a\,b\,b\,\textbf{.}\,1\,0\,3\,1\,1\,2\,\cdots). One can verify that

στ​((⋯​a​b​a​b​b​.​ 1 0 3 1 1 2​⋯))=(⋯​a​b​a​b​b​b​.​ 0 3 1 1 2​⋯).\sigma_{\tau}((\cdots\,a\,b\,a\,b\,b\,\textbf{.}\,1\,0\,3\,1\,1\,2\,\cdots))=(\cdots\,a\,b\,a\,b\,b\,b\,\textbf{.}\,0\,3\,1\,1\,2\,\cdots).

Note that one can consider S=S1∪S2S=S_{1}\cup S_{2}, where, S1={0,1},S2={2,3}S_{1}=\{0,1\},S_{2}=\{2,3\} and define τ\tau as follows.

τ(x)={τ1(x)ifx∈S1,τ2(x)ifx∈S2.\displaystyle\tau(x)=\left\{\begin{tabular}[]{ll}$\tau_{1}(x)\,\,\,\quad\quad\text{if}\,\,x\in S_{1},$\\ tau2(x)quadquadtextif,,xinS2.\tau_{2}(x)\quad\quad\text{if}\,\,x\in S_{2}.tau2(x)quadquadtextif,,xinS2.\end{tabular}\right. (2.11)

Then, στi:Σi→Σ\sigma_{\tau_{i}}:\Sigma_{i}\to\Sigma will be the local dynamics of στ\sigma_{\tau} (see Definition 2.1) with Σi=C0sm∪C0sn,sm≠sn∈Si\Sigma_{i}=C_{0}^{s_{m}}\cup C_{0}^{s_{n}},s_{m}\neq s_{n}\in S_{i} for i=1,2i=1,2.■\hfill\blacksquare

Example 2.24.

Let X=[0,1]X=[0,1] and T​(x)=3​x​m​o​d​ 1T(x)=3x\,\,mod\,1 be the expanding map. For S={0,1,2},Z={a}S=\{0,1,2\},\,Z=\{a\}, this map is conjugate (mod 0) with a zip shift map with τ:S→Z\tau:S\to Z that τ​(0)=τ​(1)=τ​(2)=a\tau(0)=\tau(1)=\tau(2)=a. ■\hfill\blacksquare

Proposition 2.25.

The zip shift maps are S-expansive local homeomorphisms.

Proof.

Let γ=1/2\gamma=1/2. For any x,y∈Σx,y\in\Sigma recall that d​(x,y)=1λM​(x,y)d(x,y)=\frac{1}{\lambda^{M(x,y)}} as in (2.7). For x≠yx\neq y there exists some i∈ℤi\in\mathbb{Z} such that xi≠yix_{i}\neq y_{i}. Let ii be the least in modulus of such ii. If i>0i>0, then n=i+1n=i+1 and d​(στi+1​x,στi+1​y)=1λ0>1/2d(\sigma_{\tau}^{i+1}x,\sigma_{\tau}^{i+1}y)=\frac{1}{\lambda^{0}}>1/2. If i<0i<0, then d​(στ−i​x,στ−i​y)=d​((στi)−1​x,(στi)−1​y)=1λ0>1/2d(\sigma_{\tau}^{-i}x,\sigma_{\tau}^{-i}y)=d((\sigma_{\tau}^{i})^{-1}x,(\sigma_{\tau}^{i})^{-1}y)=\frac{1}{\lambda^{0}}>1/2. Indeed the zip shift map is an S-expansive dynamical system with γ=1/2\gamma=1/2. ∎

Theorem 2.26.

Let στ:ΣZ,S→ΣZ,S\sigma_{\tau}:\Sigma_{Z,S}\to\Sigma_{Z,S} be a full zip shift map. Then στ\sigma_{\tau} has shadowing property.

Proof.

Let (ΣZ,S,στ)(\Sigma_{Z,S},\sigma_{\tau}) be a zip shift space. Let ϵ>0\epsilon>0 and m≥1m\geq 1 such that λ−m<ϵ\lambda^{-m}<\epsilon. Then if x,y∈ΣZ,Sx,y\in\Sigma_{Z,S} such that d​(x,y)<λ−(m+1)d(x,y)<\lambda^{-(m+1)}, we have xi=yix_{i}=y_{i} for |i|≤m|i|\leq m. In particular, x0=y0x_{0}=y_{0}. Let {zj}j=−∞+∞\{z^{j}\}_{j=-\infty}^{+\infty} be a δ=λ−(m+1)\delta=\lambda^{-(m+1)}-pseudo orbit for στ\sigma_{\tau}. Therefore by definition, d​(στ​(zj),zj+1)<δd(\sigma_{\tau}(z^{j}),z^{j+1})<\delta and indeed, zk+1j=zkj+1z^{j}_{k+1}=z^{j+1}_{k} for all j∈ℤ,|k|≤mj\in\mathbb{Z},|k|\leq m. Let x=(xn)x=(x_{n}) where

x=(…​z−m−2−2​z−m−1−1​z−m−10​…​z−10;z00​…​zm0​zm+10​zm+11​zm+12​…).x=(\dots\,z_{-m-2}^{-2}\,z_{-m-1}^{-1}\,z_{-m-1}^{0}\,\dots\,z_{-1}^{0}\,;z_{0}^{0}\,\dots\,z_{m}^{0}\,z_{m+1}^{0}z_{m+1}^{1}\,z_{m+1}^{2}\,\dots).

As {zj}j=−∞+∞\{z^{j}\}_{j=-\infty}^{+\infty} is a δ\delta pseudo orbit, so for all |j|≤m|j|\leq m and n∈ℤn\in\mathbb{Z}, we have xn+j=zjnx_{n+j}=z_{j}^{n}, which means

d​(στn​(x),zn)<λ−m<ϵ.d(\sigma_{\tau}^{n}(x),z^{n})<\lambda^{-m}<\epsilon.

Thus στ\sigma_{\tau} has the shadowing property. ∎

In [7] the authors show that a zip shift map has shadowing property iff it is a zip shift of finite type (SFT-zip shift).

Theorem 2.27.

Let f:X→Xf:X\to X be an m-to-1 local homeomorphism of a compact metric space. Then the following are equivalent.

  1. (1)
    ff is S-expansive;
  2. (2)
    ff has a generator;
  3. (3)
    ff has a weak generator;
Proof.

(1)⇒(2)(1)\Rightarrow(2): Let ff be S-expansive with expansivity constant γ\gamma. Let α={A1,…,Ak}\alpha=\{A_{1},\dots,A_{k}\} be a finite cover consisting of open balls of radius less than γ/2\gamma/2. Suppose that x,y∈⋂−∞∞f−n(A¯n)}x,y\in\bigcap_{-\infty}^{\infty}f^{-n}(\overline{A}_{n})\}. As by S-expansivity, d​(fn​(x),fn​(y))≤γ/2<γd(f^{n}(x),f^{n}(y))\leq\gamma/2<\gamma, for all n∈ℤn\in\mathbb{Z}, therefore x=yx=y.

(2)⇒(3)(2)\Rightarrow(3): Suppose that ff has a generator α\alpha which is not a weak generator. Therefore let exists x,y∈⋂−∞∞f−n​(An)≠∅x,y\in\bigcap_{-\infty}^{\infty}f^{-n}(A_{n})\neq\emptyset with x≠yx\neq y. As for any n,n, the An⊂A¯n,A_{n}\subset\overline{A}_{n}, it means that x,y∈⋂−∞∞f−n​(A¯n)≠∅,x,y\in\bigcap_{-\infty}^{\infty}f^{-n}(\overline{A}_{n})\neq\emptyset, with x≠yx\neq y, which is a contradiction.

(3)⇒(1)(3)\Rightarrow(1): Let β\beta be a weak generator for ff with Lebesgue number δ\delta. Therefore, for every bi-sequence {Bn}\{B_{n}\} of elements of β,\beta, the⋂n=−∞∞f−n​(Bn)\bigcap_{n=-\infty}^{\infty}f^{-n}(B_{n}) is at most one point. Suppose that d​(fn​(x),fn​(y))<δd(f^{n}(x),f^{n}(y))<\delta for all n∈ℤn\in\mathbb{Z}. Then by Lemma 2.3 for each nn there is some Bn∈βB_{n}\in\beta such thatfn​(x),fn​(y)∈Bnf^{n}(x),f^{n}(y)\in B_{n}, and so x,y∈⋂−∞∞{f−n​(Bn)}x,y\in\bigcap_{-\infty}^{\infty}\{f^{-n}(B_{n})\} which is at most one point. Thereupon, x=yx=y and ff is S-expansive. ∎

Theorem 2.28.

Let f:X→Xf:X\to X be an m-to-1 local homeomorphism of a compact metric space and let k≠0k\neq 0 be a positive integer. Then ff is S-expansive if and only if fkf^{k} is S-expansive.

Proof.

Let γ\gamma be the S-expansivity constant of ff. Then by Theorem 2.27 ff has a generator. Call it α\alpha. Note that ⋁i=0k−1fi​(α)\bigvee_{i=0}^{k-1}f^{i}(\alpha) will be a generator for fkf^{k}. Moreover, if α\alpha is a generator for fkf^{k}, then it is also a generator for ff. ∎

Theorem 2.29 (Main).

Let f:X→Xf:X\to X be an S-expansive m-to-1 local homeomorphism of a compact connected metric space. Then there exist finite symbolic sets S,Z,S,Z, a closed invariant subset Σ\Sigma, a zip shift map στ:Σ→Σ\sigma_{\tau}:\Sigma\to\Sigma and a continuous surjectionπ:Σ→X\pi:\Sigma\to X satisfying π∘στ=f∘π\pi\circ\sigma_{\tau}=f\circ\pi, i.e. the following diagram commutes.

Σ{\Sigma}Σ{\Sigma}↻{\circlearrowright}X{X}X{X}στ\scriptstyle{\sigma_{\tau}}π\scriptstyle{\pi}π\scriptstyle{\pi}f\scriptstyle{f} (2.12)

Moreover, if XX is connected, the above result is valid for any S-expansive local homeomorphism f:X→Xf:X\to X.

Proof.

Let γ>0\gamma>0 be the S-expansive constant for ff. As ff is an m-to-1 local homeomorphism, there exist mm principal domains DiD_{i} such that fi=f|Di:Di→Xf_{i}=f_{{}_{|_{D_{i}}}}:D_{i}\to X is a homeomorphism. Note that the collection of principal domains 𝒟={D1,…,Dm}\mathcal{D}=\{D_{1},\dots,D_{m}\} is a topological partition for XX. We choose a finite topological partition α={A0,⋯,Ak−1},\alpha=\{A_{0},\cdots,A_{k-1}\},such that for all A∈αA\in\alpha, the d​i​a​m​(Ai)<γ,i=0,…,k−1diam(A_{i})<\gamma,i=0,\dots,k-1. Refine this topological partition to β=α∨D\beta=\alpha\vee D if necessary to obtain a DTP. Let β={B0,⋯,Bl−1}\beta=\{B_{0},\cdots,B_{l-1}\} and ∂∞=⋃n=−∞+∞fn​(∂Bi)\partial_{\infty}=\bigcup\limits_{n=-\infty}^{+\infty}f^{n}(\partial{B_{i}}). Then ∂∞\partial_{\infty}is a set of first category and so X¯=X∖∂∞\bar{X}=X\setminus\partial_{\infty} is dense in XX. Suppose that ΣZ,S\Sigma_{Z,S} is a zip shift space in which S,ZS,Z and the transition map τ\tau are deduced by the help of β\beta and some induced image topological partition (Remark 2.9). We show that for any x∈X¯x\in\bar{X}, a unique member of ΣZ,S\Sigma_{Z,S} is assigned. Let Q={Q1,⋯,Qs}Q=\{Q_{1},\cdots,Q_{s}\} be the image topological partition associated to β\beta. Consider the main topological partition MM, associated with QQ. If necessary one can refine the partition QQ (take Q∨βQ\vee\beta) and M,M, to obtain a good image and consequently main topological partition with d​i​a​m​(Mi)<γdiam(M_{i})<\gamma for all Mi∈MM_{i}\in M. Note that the elements of MM are pre-images of QQ and by Theorem 2.27 MM is a generator. The good image topological partitions QQ and the main topological partitions M,M, induce two finite symbolic sets, which can be represented by SS (for MM) and ZZ (for QQ). Furthermore, as the elements of MM are pre-images of the elements of QQ, there are surjective maps τ^:M→Q\hat{\tau}:M\to Q and τ:S→Z\tau:S\to Z that relate the domain and image partitions, as well as their associated symbolic sets. Let ΣZ,S\Sigma_{Z,S} be the full zip shift map associated with S,ZS,Z and with transition map τ\tau. Let (#​(Z)=s¯)≤(k¯=#​(S))(\#(Z)=\bar{s})\leq(\bar{k}=\#(S)) and consider Z={a1,…,as¯}Z=\{a_{1},\dots,a_{\bar{s}}\} and S={b0,b1,…,bk¯−1}S=\{b_{0},b_{1},\dots,b_{{\bar{k}}-1}\}. Then to any x∈X¯x\in\bar{X} can be assigned a unique element of the form (xn)∈ΣZ,S(x_{n})\in\Sigma_{Z,S}. Let Mi∈MM_{i}\in M and Qi∈QQ_{i}\in Q. Then for n≥0,n\geq 0, xn=bix_{n}=b_{i} if x∈f−n​(Mi)x\in f^{-n}(M_{i}) i.e. fn​(x)∈Mif^{n}(x)\in M_{i} and for n<0,n<0, xn=aix_{n}=a_{i} if x∈f−n−1​(Qi)x\in f^{-n-1}(Q_{i}) (−n>0-n>0). Suppose that Λ⊂ΣZ,S\Lambda\subset\Sigma_{Z,S} denotes the set of all such points. In other words, one can say that x∈[⋂−∞0f−n​(Qxn−1)]¯∩[⋂0∞f−n​(Mxn)]¯x\in\overline{[\bigcap_{-\infty}^{0}f^{-n}(Q_{x_{n-1}})]}\cap\overline{[\bigcap_{0}^{\infty}f^{-n}(M_{x_{n}})]}. Note that as domain topological partition MM is a generator and we have the S-expansivity, the non-empty intersection [⋂−∞0f−n​(Qxn−1)]¯∩[⋂0∞f−n​(Mxn)]¯\overline{[\bigcap_{-\infty}^{0}f^{-n}(Q_{x_{n-1}})]}\cap\overline{[\bigcap_{0}^{\infty}f^{-n}(M_{x_{n}})]} is a single point. One needs to guarantee that Λ\Lambda is a zip shift space. As ΣZ,S\Sigma_{Z,S} is a compact metric space and[⋂−n+10ϕ−k​(Qxk−1)]¯∩[⋂0nϕ−k​(Mxk)]¯\overline{[\bigcap_{-n+1}^{0}\phi^{-k}(Q_{x_{k-1}})]}\cap\overline{[\bigcap_{0}^{n}\phi^{-k}(M_{x_{k}})]} has the finite intersection property, the subset Λ\Lambda of ΣZ,S\Sigma_{Z,S} is closed. Let στ:Λ→Λ\sigma_{\tau}:\Lambda\to\Lambda and consider the injective map ϕ:Λ→X¯\phi:\Lambda\to\bar{X}, in which ϕ−1​(x)=(zn)n∈ℤ\phi^{-1}(x)=(z_{n})_{n\in\mathbb{Z}}. Note that, στ∘ϕ−1​(x)=ϕ−1∘f​(x)\sigma_{\tau}\circ\phi^{-1}(x)=\phi^{-1}\circ f(x), for all x∈X¯x\in\bar{X}. This means that Λ\Lambda is στ−\sigma_{\tau}-invariant as well. Indeed, it is a (sub)zip shift space.

In order to extend ϕ\phi to a continuous map ϕ~:Λ→X\tilde{\phi}:\Lambda\to X satisfying ϕ~∘στ​(zn)=f∘ϕ~​(zn)\tilde{\phi}\circ\sigma_{\tau}(z_{n})=f\circ\tilde{\phi}(z_{n}) for all z∈Λz\in\Lambda, it is enough to show that ϕ\phi is a uniformly continuous map. Let ϵ>0\epsilon>0 be arbitrary. As ff is an S-expansive m-to-1 local homeomorphism and MM is a generator, there exists some N>0N>0 such that each member of ⋁k=−NNfk​(M)\bigvee_{k=-N}^{N}f^{k}(M) has diameter less than ϵ.\epsilon. Moreover, if a=(an),b=(bn)a=(a_{n}),b=(b_{n}) be elements of Λ\Lambda with an=bna_{n}=b_{n} for |n|≤N|n|\leq N, i.e. for 0<δ≤12N0<\delta\leq\frac{1}{2^{N}}, then ϕ​(a)\phi(a) and ϕ​(b)\phi(b) belong to the same element ∨k=−N+10f−k(Qxk−1¯)∨(∨k=0Nf−k​(Mxk)¯∈⋁k=−NNfk(M)\overline{\vee_{k=-N+1}^{0}f^{-k}(Q_{x_{k-1}}})\vee(\overline{\vee_{k=0}^{N}f^{-k}(M_{x_{k}})}\in\bigvee_{k=-N}^{N}f^{k}(M) and thus d​(ϕ​(a),ϕ​(b))<ϵ.d(\phi(a),\phi(b))<\epsilon. Thereupon, ϕ\phi is uniformly continuous. Moreover, the last assertion is correct by Theorem 2.2.

∎

2.2. Dataavailability

Data sharing not applicable to this article as no data sets were generated or analysed during the current study.

2.3. Conflicts of interest

The authors declare no conflict of interest.

2.4. Founding

The second author gratefully acknowledges partial financial support from FAPEMIG, Brazil.

References