betaFeedback

WorksMathematics

Topological notions of Shadowing under Semi-Conjugacy

Abstract

We study various notions of shadowing and investigate whether they are preserved under semi-conjugacy and inverse limits. In particular, we study Inverse shadowing, HH-shadowing, Periodic shadowing, Ramsey shadowing, Density shadowing and Ergodic shadowing.

Introduction

Let (X,d)(X,d) be a compact metric space and f:X→Xf:X \to X be a continuous function. Given a point z∈Xz \in X, we define the orbit of zz as the set {z,f(z),f2(z),…}\{z, f(z), f^{2}(z), \ldots\}. Let δ>0\delta> 0. A δ\delta-pseudo-orbit is a sequence (xi)i∈N0(x_i)_{i\in\mathbb{N}_0} such that d(f(xi),xi+1)<δd(f(x_i),x_{i+1}) < \delta for all i∈N0i \in\mathbb{N}_0. Intuitively a δ\delta-pseudo orbit is a sequence that behaves like an orbit, however at every stage of the iteration there is an error of at most δ\delta. Given ϵ>0\epsilon> 0, a sequence (yi)i∈N0(y_i)_{i\in\mathbb{N}_0} in XX is said to ϵ\epsilon-shadow the sequence (xi)i∈N0(x_i)_{i\in\mathbb{N}_0} if d(yi,xi)<ϵd(y_i,x_i) < \epsilon for all i∈N0i \in\mathbb{N}_0. We say that ff has shadowing if every ϵ>0\epsilon> 0, there exists a δ>0\delta> 0, such that every δ\delta-pseudo-orbit is ϵ\epsilon-shadowed by a true orbit.

Shadowing has become an important tool in the study of dynamical systems. Anosov’s shadowing lemma shows that diffeomorphisms restricted to a hyperbolic set exhibit shadowing [1]. This result has been used to study the stability of hyperbolic dynamical systems. Bowen used shadowing to study Axiom A diffeomorphisms and their ω\omega-limit sets [3].

Shadowing has also emerged as an important tool in applied contexts, particularly in numerical modelling. Computers operate with finite precision, hence the iteration of a function typically introduces small rounding errors. The resulting numerical trajectory can therefore be viewed as a pseudo-orbit rather than an exact orbit. If the dynamical system possesses the shadowing property, then this pseudo-orbit remains close to a genuine orbit. Thus, despite the presence of numerical errors, the computed trajectory can still be interpreted as representing a valid behaviour of the underlying dynamical system. For this reason, shadowing has been studied extensively in the context of numerical modelling and the reliability of numerical simulations [HJ05].

In [8] it was shown that shadowing can be formulated in purely topological terms and therefore does not require the underlying space to be metrizable. Building on this observation, they introduced the almost lifting property, which characterizes when a factor map preserves the shadowing property.

A factor map between two dynamical systems (X,f),(Y,g)(X,f),(Y,g) is a continuous surjection, ψ:X⟶Y\psi:X \longrightarrow Y, such that the following holds, ψ∘f=g∘ψ\psi\circ f = g \circ\psi. The authors also showed that shadowing was preserved under inverse limits provided it satisfied the Mittag-Leffler condition.

Since shadowing was first introduced, various other notions of shadowing have been developed. These include inverse shadowing, orbital shadowing and ergodic shadowing. Work has been done on studying whether these differing notions of shadowing are preserved under various actions such as factor maps and inverse limits [9].

In this paper, we expand upon the work of [9] by considering several other notions of shadowing and addressing some of the unresolved questions in the paper. In particular, we look at inverse shadowing, hh-shadowing, periodic shadowing, Ramsey shadowing, density shadowing and ergodic shadowing. We characterize the property that determines whether they are preserved under factor maps and investigate whether they are preserved under inverse limits, and the hyperspace topology.

Preliminaries

Definition 2.1. A dynamical system is a compact Hausdorff space XX, with a continuous function f:X→Xf:X \to X.

The orbit of a point x∈Xx \in X is defined to be the set {x,f(x),f2(x),…}\{x, f(x), f^{2}(x), \ldots\}. Given two dynamical systems (X,f)(X, f), (Y,g)(Y, g), a factor map from (X,f)(X, f) to (Y,g)(Y, g) is a continuous surjection, ψ\psi, such that the following holds, ψ∘f=g∘ψ\psi\circ f = g \circ\psi. ψ\psi is often also called a semi-conjugacy and we say that (X,f)(X, f) and (Y,g)(Y, g) are semi-conjugate. If ψ\psi is also a homeomorphism, we call it a conjugacy, and say that (X,f)(X, f) and (Y,g)(Y, g) are conjugate.

Definition 2.2. Given a dynamical system (X,f)(X, f), let 2X2^{X} denote the set of non-empty compact subsets of XX. Given a subset BB of XX define Bδ:={z:z∈B(x,δ),x∈B}B_{\delta} := \{z : z \in B(x,\delta), x \in B\}. For subsets AA, BB of XX, we define the Hausdorff metric by dH(A,B):=inf⁡{δ:A⊆Bδ and B⊆Aδ}d_{H}(A, B) := \inf\{\delta: A \subseteq B_{\delta} \text{ and } B \subseteq A_{\delta}\}. This forms a compact topological space under the Hausdorff metric. Then ff induces a continuous function, FF from 2X2^{X} to 2X2^{X} defined by sending B∈2XB \in2^{X} to f(B)f(B). Hence (2X,F)(2^{X}, F) is also a dynamical system.

Definition 2.3. Consider a collection of dynamical systems (Xn,Fn)n∈N0(X_{n}, F_{n})_{n \in\mathbb{N}_{0}}. Suppose for all m≤nm \leq n we have maps ψmn:Xn⟶Xm\psi_{m}^{n} : X_{n} \longrightarrow X_{m} which satisfy:

  • ψnn\psi_{n}^{n} is the identity.

  • For all l≤m≤nl \leq m \leq n we have that ψln=ψlm∘ψmn\psi_{l}^{n} = \psi_{l}^{m} \circ\psi_{m}^{n}.

  • Fm∘ψmn=ψmn∘FnF_{m} \circ\psi_{m}^{n} = \psi_{m}^{n} \circ F_{n}

We call this an inverse system (ψmn,Xn)(\psi_{m}^{n}, X_{n}) and its inverse limit space is given by: lim⁡⟵{ψmn,Xn}={(xi)∈∏Xn:∀m≤n ψmn(xn)=xm}\lim\limits_{\longleftarrow}\{\psi_{m}^{n}, X_{n}\} = \{(x_{i}) \in\prod X_{n} : \forall m \leq n\ \psi_{m}^{n}(x_{n}) = x_{m}\}

The inverse limit is the dynamical system given by (lim⁡⟵{ψmn,Xn},F∗)(\lim\limits_{\longleftarrow}\{\psi_{m}^{n}, X_{n}\}, F^{*}) where F∗(⟨xn⟩)=(F1(x1),F2(x2),F3(x3),…)F^{*}(\langle x_{n}\rangle) = (F_{1}(x_{1}), F_{2}(x_{2}), F_{3}(x_{3}), \ldots)

An inverse system of dynamical systems satisfies the Mittag-Leffler condition provided that for all ll, we can find an m≥lm \geq l such that for all n≥mn \geq m we have that ψln(Xn)=ψlm(Xm)\psi_{l}^{n}(X_{n}) = \psi_{l}^{m}(X_{m}). Clearly if each ψ\psi is surjective, the Mittag-Leffler condition is satisfied.

Definition 2.4 (Shadowing). Let δ>0\delta> 0. A sequence (xi)i∈N0(x_{i})_{i \in\mathbb{N}_{0}} is called a δ\delta-pseudo-orbit if d(f(xi),xi+1)<δd(f(x_{i}), x_{i+1}) < \delta for all i∈N0i \in\mathbb{N}_{0}. Given ϵ>0\epsilon> 0, a sequence (yi)i∈N0(y_{i})_{i \in\mathbb{N}_{0}} in XX is said to ϵ\epsilon-shadow the sequence (xi)i∈N0(x_{i})_{i \in\mathbb{N}_{0}} if d(yi,xi)<ϵd(y_{i}, x_{i}) < \epsilon for all i∈N0i \in\mathbb{N}_{0}. We say that ff has shadowing if for every ϵ>0\epsilon> 0, there exists a δ>0\delta> 0, such that if (xi)i∈N0(x_{i})_{i \in\mathbb{N}_{0}} is a δ\delta-pseudo-orbit, then there exists a zz such that (xi)i∈N0(x_{i})_{i \in\mathbb{N}_{0}} is ϵ\epsilon-shadowed by a true orbit.

The definition given for shadowing is given in a metric space. However in [8], Good and Meddaugh showed that an equivalent definition of shadowing could be formulated on topological spaces.

Definition 2.5. Let (X,f)(X, f) be a dynamical system, where XX is a topological compact space. Let FOC(X)FOC(X) denote the set of finite open covers of XX. Let U∈FOC(X)\mathcal{U} \in FOC(X). We say a sequence (xi)i∈N0(x_{i})_{i \in\mathbb{N}_{0}} is a U\mathcal{U}-pseudo orbit provided that for all i∈N0i \in\mathbb{N}_{0}, there exists a Ui+1∈UU_{i+1} \in\mathcal{U} such that f(xi),xi+1∈Ui+1f(x_{i}), x_{i+1} \in U_{i+1}.

We say a sequence (xi)i∈N0(x_{i})_{i \in\mathbb{N}_{0}} is U\mathcal{U}-shadowed by a sequence (yi)i∈N0(y_{i})_{i \in\mathbb{N}_{0}} if for all i∈N0i \in\mathbb{N}_{0}, there exists a Ui∈UU_{i} \in\mathcal{U} such that xi,yi∈Uix_{i}, y_{i} \in U_{i}. We say (X,f)(X, f) has topological shadowing if for all U∈FOC(X)\mathcal{U} \in FOC(X), there exists Y∈FOC(X)\mathcal{Y} \in FOC(X) such that if (xi)i∈N0(x_{i})_{i \in\mathbb{N}_{0}} is a U\mathcal{U}-pseudo orbit, then there exists a z∈Xz \in X such that the orbit of zz Y\mathcal{Y}-shadows (xi)i∈N0(x_{i})_{i \in\mathbb{N}_{0}}.

Theorem 2.1 ([8], Lemma 6). Let XX be a compact metric space. A dynamical system (X,f)(X, f) has shadowing if and only if for every U∈FOC(X)\mathcal{U} \in FOC(X), there exists a V∈FOC(X)\mathcal{V} \in FOC(X) such that every V\mathcal{V}-pseudo orbit is U\mathcal{U}-shadowed by a point $z \in X.

Analogously to [8], for each metric definition of an alternative notion of shadowing, we also provide the corresponding topological definition. We will show that these definitions are equivalent to their corresponding metric definition, provided the underlying metric space XX is compact.

Inverse shadowing

Definition 3.1. Inverse shadowing- Metric definition:

Let (X,f)(X, f) be a dynamical system, and let XωX^{\omega} denote the set of all infinite sequences of XX. For δ>0\delta> 0, let Φf(δ)⊆Xω\Phi_{f}(\delta) \subseteq X^{\omega} denote the set of all sequences given by a δ\delta-pseudo orbit. A δ\delta-method for ff is a map ρ:X→Φf(δ)\rho: X \to\Phi_{f}(\delta), which sends every x∈Xx \in X to a pseudo orbit sequence which begins with xx. We let T(f,δ)\mathcal{T}(f,\delta) denote the set of all δ\delta-methods of ff.

A dynamical system (X,f)(X, f) has inverse shadowing if for all ϵ>0\epsilon> 0 there exist δ>0\delta> 0 such that for all x∈Xx \in X and all δf∈T(f,δ)\delta_{f} \in\mathcal{T}(f,\delta), there exists yy such that δf(y)\delta_{f}(y) ϵ\epsilon-shadows the orbit of xx.

Under the standard notion of shadowing, we are interested in whether pseudo-orbits can be approximated by a real orbit. Inverse shadowing consider the converse problem: whether any orbit can be approximated using a suitable pseudo-orbit from any δ\delta-method of ff. Inverse shadowing arises when studying the structural stability and hyperbolicity of various dynamical systems [11]. In particular, it has been shown that on compact manifolds, hyperbolic homeomorphisms have inverse shadowing under certain conditions on the set of δ\delta-methods [13].

Definition 3.2. Inverse shadowing- Topological definition:

A dynamical system (X,f)(X,f) has inverse shadowing if for all VX∈FOC(X)V_X \in FOC(X), there exists an WX∈FOC(X)W_X \in FOC(X), such that for all x∈Xx \in X and all δf∈T(f,WX)\delta_f \in\mathcal{T}(f,W_X), there exists an yy such that, δf(y)\delta_f(y) VXV_X-shadows xx.

We use the following well-known theorem to prove the equivalence between the topological and metric notions of inverse-shadowing in a compact metric space.

Theorem 3.1. (Lebesgue’s number lemma) Let XX be a compact metric space. Let U\mathcal{U} be an open cover of XX. There exists a δ>0\delta> 0 such that for every x∈Xx \in X, there exists a Uk∈UU_k \in\mathcal{U} such that B(x,δ)⊂UkB(x,\delta) \subset U_k. We call this δ\delta the Lebesgue number of U\mathcal{U}.

Theorem 3.2. Let (X,f)(X,f) be a dynamical system where XX is a compact metric space. The topological and metric definitions of inverse shadowing are equivalent.

Proof. Suppose the topological definition of inverse shadowing holds in (X,f)(X,f). Let ϵ>0\epsilon> 0. Since XX is compact we can choose an open cover U\mathcal{U} of XX which consists of a finite number of ϵ4\frac{\epsilon}{4} balls. Choose V∈FOC(X)\mathcal{V} \in FOC(X) which satisfies Definition 3.2 with respect to U\mathcal{U}. Let δ>0\delta> 0 denote the Lebesgue number of V\mathcal{V}. Pick an arbitrary x∈Xx \in X and an arbitrary δf∈T(f,δ)\delta_f \in\mathcal{T}(f,\delta). Then δf\delta_f is a map ρ:X→Φf(δ)\rho:X \to\Phi_f(\delta), which sends every y∈Xy \in X to a δ\delta-pseudo orbit sequence which begins with yy. For all i∈N0i \in\mathbb{N}_0 we have d(f(yi),yi−1)<δd(f(y_i),y_{i-1}) < \delta. As δ\delta is the Lebesgue number of V\mathcal{V}, we know that there exists Vk∈VV_k \in\mathcal{V} such that f(yi),yi−1∈Vkf(y_i),y_{i-1} \in V_k. Hence Φ(y)\Phi(y) is also a V\mathcal{V}-pseudo orbit which starts with yy. Therefore δf\delta_f is also a member of T(f,V)\mathcal{T}(f,\mathcal{V}). Choose a yy which satisfies Definition 3.2 with respect to xx and δf\delta_f. As we have topological inverse shadowing, we know that for all i∈N0i \in\mathbb{N}_0, there exists a Uj∈UU_j \in\mathcal{U} such that fi(x)f^i(x) and (δf(y))i(\delta_f(y))_i both belong to UjU_j. Since U\mathcal{U} consists of ϵ4\frac{\epsilon}{4}-balls, we know that d(fi(x),(δf(y))i)<ϵ2d(f^i(x),(\delta_f(y))_i) < \frac{\epsilon}{2}. Hence the metric definition of inverse shadowing is satisfied.

Now we consider the case where the metric definition holds. Consider an finite open cover U\mathcal{U} of XX. Let ϵ>0\epsilon> 0 denote the Lebesgue number of U\mathcal{U}. Choose δ>0\delta> 0 which satisfies Definition 3.1 with respect to ϵ\epsilon. As XX is compact we can choose a finite open cover V\mathcal{V} of XX which consists of δ2\frac{\delta}{2}-balls. Pick an arbitrary x∈Xx \in X and an arbitrary δf∈T(f,V)\delta_f \in\mathcal{T}(f,\mathcal{V}).

Then δf\delta_f is a map ρ:X→Φf(V)\rho:X \to\Phi_f(\mathcal{V}), which sends every y∈Xy \in X to a V\mathcal{V}-pseudo orbit sequence which begins with yy. For all i∈N0i \in\mathbb{N}_0, there exists Vk∈VV_k \in\mathcal{V} such that f(yi),yi−1∈Vkf(y_i),y_{i-1} \in V_k. V\mathcal{V} consists of δ2\frac{\delta}{2}-balls, so we know that d(f(yi),yi−1)<δd(f(y_i),y_{i-1}) < \delta. Hence Φ(y)\Phi(y) is also a δ\delta-pseudo orbit which starts with yy. Therefore δf\delta_f is also a member of T(f,δ)\mathcal{T}(f,\delta). Choose yy which satisfies Definition 3.1. For i∈N0i \in\mathbb{N}_0 we have d(fi(x),yi)<ϵd(f^i(x),y_i) < \epsilon. Hence there exists a UjU_j such that fi(x),xi∈Ujf^i(x),x_i \in U_j. Hence the topological definition of inverse shadowing is also satisfied.

Definition 3.3. Almost lifting delta method property (ALDM)- Metric definition:

A factor map ψ:X⟶Y\psi:X \longrightarrow Y has almost lifting delta method property, if for all ϵ>0\epsilon> 0 and η>0\eta> 0 there exist δ>0\delta> 0, such that if δg∈T(g,δ)\delta_g \in\mathcal{T}(g,\delta), then there exists δf∈T(f,η)\delta_f \in\mathcal{T}(f,\eta) such that ψ(δf)\psi(\delta_f) ϵ\epsilon-shadows δg\delta_g.

When we say that ψ(δf)\psi(\delta_f) ϵ\epsilon-shadows δg\delta_g, we mean that for all x∈Xx \in X we have ψ(δf(x))\psi(\delta_f(x)) ϵ\epsilon-shadows some element of δg\delta_g.

Definition 3.4. Almost lifting delta method property (ALDM)- Topological definition:

ψ:X⟶Y\psi:X \longrightarrow Y has almost lifting delta method property, if for all VY∈FOC(Y)V_Y \in FOC(Y) and VX∈FOC(X)V_X \in FOC(X) there exist WY∈FOC(Y)W_Y \in FOC(Y), such that if δg∈T(g,WY)\delta_g \in\mathcal{T}(g,W_Y), then there exists δf∈T(f,VX)\delta_f \in\mathcal{T}(f,V_X) such that ψ(δf)\psi(\delta_f) VYV_Y-shadows δg\delta_g.

It is straightforward to show that the topological and metric definitions of ALDM are equivalent, hence we omit the proof.

Theorem 3.3. Suppose ψ\psi is a factor map from (X,f)(X,f) to (Y,g)(Y,g). We will prove that if (Y,g)(Y,g) has inverse shadowing, then ψ\psi must have ALDM.

Proof. Let ϵ>0\epsilon> 0 and η>0\eta> 0. Choose δ>0\delta> 0 such that (Y,g)(Y,g) witnesses inverse shadowing with respect to ϵ\epsilon. Let y∈Yy \in Y be arbitrary. Let δg∈T(g,δ)\delta_g \in\mathcal{T}(g,\delta) be arbitrary. For each x∈Xx \in X, consider its orbit. This clearly define a η\eta-pseudo orbit for each x∈Xx \in X. So consider δf∈T(f,η)\delta_f \in\mathcal{T}(f,\eta) defined by the orbits on XX. Let x∈Xx \in X be arbitrary. Now consider δf(x)\delta_f(x). Then clearly ψ(δf(x))\psi(\delta_f(x)) is the orbit of ψ(x)\psi(x). But since (Y,g)(Y,g) has inverse shadowing, we know there exists a yy such that δg(y)\delta_g(y) ϵ\epsilon-shadows the orbit of ψ(x)\psi(x), and hence ψ(δf(x))\psi(\delta_f(x)) ϵ\epsilon-shadows δg(y)\delta_g(y). Hence ψ\psi has the ALDM property.

Theorem 3.4. Let (X,f)(X,f) and (Y,g)(Y,g) be two dynamical systems. suppose (X,f)(X,f) has inverse shadowing and ψ:X⟶Y\psi:X\longrightarrow Y has the ALDM property, then (Y,g)(Y,g) has inverse shadowing.

Proof. Fix an open cover UY\mathcal{U}_Y of YY. Let WY\mathcal{W}_Y be an open cover of YY such that if W,W′∈WYW,W' \in\mathcal{W}_Y and W∩W′≠∅W \cap W' \ne\varnothing, there exist U∈UYU \in\mathcal{U}_Y such that W∪W′⊂UW \cup W' \subset U. Let UX\mathcal{U}_X be an open cover of XX such that ψ(UX)\psi(\mathcal{U}_X) refines WY\mathcal{W}_Y. Since (X,f)(X,f) has inverse shadowing, let VX\mathcal{V}_X witness this with respect to UX\mathcal{U}_X. Since ψ\psi has ALDM, let VY\mathcal{V}_Y witness this with respect to WY\mathcal{W}_Y and VX\mathcal{V}_X.

Let y∈Yy \in Y. Let δg∈T(g,VY)\delta_g \in\mathcal{T}(g,\mathcal{V}_Y) be arbitrary. Consider z∈ψ−1(y)z \in\psi^{-1}(y). By inverse shadowing we know that for all δf∈T(f,VX)\delta_f \in\mathcal{T}(f,\mathcal{V}_X), there exists a x′∈Xx' \in X, such that δf(x′)\delta_f(x') UX\mathcal{U}_X-shadows zz. Take δf∗∈T(f,VX)\delta_f^* \in\mathcal{T}(f,\mathcal{V}_X), such that it satisfies the conditions of the ALDM property. Hence, we know that zz is UX\mathcal{U}_X-shadowed by δf∗(x)\delta_f^*(x) for some x∈Xx \in X. Since ψ(UX)\psi(\mathcal{U}_X) refines WY\mathcal{W}_Y. We must have that for all ii, ψ(fi(z))\psi(f^i(z)) and (ψ(δf∗(x)))i(\psi(\delta_f^*(x)))_i are both in WYkW_{Y_k}, for some kk. But notice that ψ(fi(z))=gi(ψ(z))=gi(y)\psi(f^i(z)) = g^i(\psi(z)) = g^i(y). Hence for all ii, gi(y)g^i(y) and (ψ(δf∗(x)))i(\psi(\delta_f^*(x)))_i are both in WYkW_{Y_k}.

But by the ALDM property ψ(δf∗(x))\psi(\delta_f^*(x)) WY\mathcal{W}_Y-shadows δg(y′)\delta_g(y') for some y′∈Yy' \in Y. Hence for all ii, (ψ(δf∗(x)))i(\psi(\delta_f^*(x)))_i and (δg(y′))i(\delta_g(y'))_i belong to WYtW_{Y_t} for some tt. So for all ii we must have that WYkW_{Y_k} and WYtW_{Y_t} both contain (ψ(δf∗(x)))i(\psi(\delta_f^*(x)))_i, and hence their intersection is non empty. But by construction of WY\mathcal{W}_Y, we know that there exists a ss such that gi(y′)g^i(y') and (δg(y′))i(\delta_g(y'))_i are both contained in UYsU_{Y_s}. Hence yy is UY\mathcal{U}_Y-shadowed by δg(y′)\delta_g(y'). And so we have shown that (Y,g)(Y,g) has inverse shadowing.

In [9] it was shown that that inverse shadowing is preserved under inverse limits and the induced hyperspace.

H-shadowing

Definition 4.1. HH-shadowing - Metric definition: We say that (X,f)(X,f) has HH-shadowing if for all ϵ>0\epsilon> 0 there exists δ>0\delta> 0 such that if (x0,…,xn)(x_0,\ldots,x_n) is a finite δ\delta-pseudo orbit, then there exists a z∈Xz \in X such that d(fi(z),xi)<ϵd(f^i(z),x_i) < \epsilon and fn(z)=xnf^n(z) = x_n.

Definition 4.2. HH-shadowing- Topological definition: A dynamical system (X,f)(X,f) has HH-shadowing if for all U∈FOC⁡(X)\mathcal{U} \in\operatorname{FOC}(X), there exists a V∈FOC⁡(X)\mathcal{V} \in\operatorname{FOC}(X) such that for all finite V\mathcal{V}-pseudo orbits (xi)(x_i), we can find z∈Xz \in X such that fi(z),xi∈Uif^i(z),x_i \in U_i and fn(z)=xnf^n(z) = x_n.

HH-shadowing arises naturally when studying shadowing in shift spaces. In a shift space, if a δ\delta-pseudo orbit (xn)(x_n) is shadowed by a point zz, then there exists an N∈NN \in\mathbb{N} such that for all n≥Nn \ge N we have that fn(z)=xnf^n(z) = x_n. If a dynamical systems has HH-shadowing, this can be used to characterize their ω\omega-limit sets [2].

Theorem 4.1. Let (X,f)(X,f) be a dynamical system where XX is a compact metric space. The topological and metric definitions of HH-shadowing are equivalent.

Proof. Suppose the topological definition of HH-shadowing holds in (X,f)(X,f). Let ϵ>0\epsilon> 0. Since XX is compact we can choose an open cover U\mathcal{U} of XX which consists of a finite number of ϵ4\frac{\epsilon}{4}-balls. Let V∈FOC⁡(X)\mathcal{V} \in\operatorname{FOC}(X) which satisfies Definition 4.2 with respect to U\mathcal{U}. Let δ>0\delta> 0 denote the Lebesgue number of V\mathcal{V}. Let x0,x1,…,xnx_0,x_1,\ldots,x_n be a finite δ\delta-pseudo orbit. Then for all i∈{0,1,…,n}i \in\{0,1,\ldots,n\} we have d(f(xi−1),xi)<δd(f(x_{i-1}),x_i) < \delta. Hence there exists a Vk∈VV_k \in\mathcal{V} such that f(xi−1),xi∈Vkf(x_{i-1}),x_i \in V_k. Hence this δ\delta pseudo orbit is also a V\mathcal{V} pseudo orbit. Choose a zz which satisfies Definition 4.2. As we have topological HH-shadowing, we know that for all i∈{0,1,…,n}i \in\{0,1,\ldots,n\}, there exists a Uj∈UU_j \in\mathcal{U} such that fi(z)f^i(z) and (xi)(x_i) both belong to UjU_j with fn(z)=xnf^n(z) = x_n. Since U\mathcal{U} consists of ϵ4\frac{\epsilon}{4}-balls, we know that d(fi(z),xi)<ϵ2d(f^i(z),x_i) < \frac{\epsilon}{2}. Hence the metric definition of HH-shadowing holds in (X,f)(X,f).

Now we consider the case where the metric definition of HH-shadowing holds. Consider an finite open cover U\mathcal{U} of XX. Let ϵ>0\epsilon> 0 denote the Lebesgue number of U\mathcal{U}. Choose δ>0\delta> 0 which satisfies Definition 4.1 with respect to ϵ\epsilon. As XX is compact we can choose a finite open cover V\mathcal{V} of XX which consists of δ2\frac{\delta}{2}-balls. Let x0,x1,…,xnx_0,x_1,\ldots,x_n be a finite V\mathcal{V}-pseudo orbit. Hence for all i∈{0,1,…,n}i \in\{0,1,\ldots,n\}, we have that f(xi−1),xi∈Vkf(x_{i-1}),x_i \in V_k for some Vk∈VV_k \in\mathcal{V}. As V\mathcal{V} consists of δ2\frac{\delta}{2}-balls, clearly x0,x1,…,xnx_0,x_1,\ldots,x_n is also a finite δ\delta-pseudo orbit. Choose a zz which satisfies Definition 4.1. zz satisfies d(fi(z),xi)<ϵd(f^i(z),x_i) < \epsilon for i∈{0,1,…,n}i \in\{0,1,\ldots,n\} with fn(z)=xnf^n(z) = x_n. Hence there exists a UjU_j such that fi(z),xi∈Ujf^i(z),x_i \in U_j. Hence (X,f)(X,f) also exhibits topological HH-shadowing.

Definition 4.3. Almost lifting hit pseudo orbit property (ALhP) - Topological definition: We say a factor map ψ:X→Y\psi: X \to Y has almost lifting hit pseudo orbit property if for all VX∈FOC(X)V_X \in FOC(X), and WY∈FOC(Y)W_Y \in FOC(Y), there exists VY∈FOC(Y)V_Y \in FOC(Y), such that if y0,…,yny_0, \ldots, y_n is a finite VYV_Y pseudo orbit in YY, then there exists a VXV_X pseudo orbit in XX, x0,…,xnx_0, \ldots, x_n such that ϕ(xi),yi∈Wk\phi(x_i), y_i \in W_k, for some kk and ψ(xn)=yn\psi(x_n) = y_n.

Definition 4.4. Almost lifting hit pseudo orbit property (ALhP) - Metric definition: We say a factor map ψ:X→Y\psi: X \to Y has almost lifting hit pseudo orbit property if for all ϵ>0\epsilon> 0 and η>0\eta> 0, there exists δ>0\delta> 0 such that if y0,…,yny_0, \ldots, y_n is a finite δ\delta pseudo orbit in YY, then there exists a η\eta pseudo orbit in XX, x0,…,xnx_0, \ldots, x_n such that d(ϕ(xi),yi)<ϵd(\phi(x_i),y_i) < \epsilon, and ψ(xn)=yn\psi(x_n) = y_n.

Theorem 4.2. Let (X,f)(X,f) and (Y,g)(Y,g) be two dynamical systems. suppose (X,f)(X,f) has HH-shadowing and ψ:X→Y\psi: X \to Y has the ALhP property, then (Y,g)(Y,g) has HH-shadowing.

Proof. Fix an open cover UYU_Y of YY. Let WYW_Y be an open cover of YY such that if W,W′∈WYW,W' \in W_Y and W∩W′≠∅W \cap W' \ne\emptyset, there exist U∈UYU \in U_Y such that W∪W′⊂UW \cup W' \subset U. Let UXU_X be an open cover of XX such that ψ(UX)\psi(U_X) refines WYW_Y. Since (X,f)(X,f) has hh-shadowing, let VXV_X witness this with respect to UXU_X. Since ψ\psi has ALhP, let VYV_Y witness this with respect to WYW_Y and VXV_X.

Let y0,…,yny_0, \ldots, y_n be a finite VYV_Y pseudo orbit. Let x0,…,xnx_0, \ldots, x_n be the VXV_X pseudo orbit, such that ψ((xi))\psi((x_i)) WYW_Y-shadows (yi)(y_i) and ψ(xn)=yn\psi(x_n) = y_n.

Since (X,f)(X,f) has HH-shadowing, we know that xix_i is UXU_X shadowed by a point say z∈Xz \in X, such that fn(z)=xnf^n(z) = x_n. Hence for all ii we have that ψ(fi(z)),ψ(xi)∈ψ(Ui)\psi(f^i(z)), \psi(x_i) \in\psi(U_i) for some Ui∈UXU_i \in U_X. Since ψ\psi is a factor map we know that ψ(fi(z))=gi(ψ(z))\psi(f^i(z)) = g^i(\psi(z)). And since ψ(UX)\psi(U_X) refines WYW_Y, it follows that there exists WiW_i such that gi(ψ(z))g^i(\psi(z)) and ψ(xi)\psi(x_i) lie in WiW_i. But we also know that ψ(xi)\psi(x_i) and yiy_i belong to WjW_j for some jj. Hence ψ(xi)\psi(x_i) is in Wi∩Wj′W_i \cap W'_j. But by construction of WYW_Y we know there exists a UiU_i such that gi(ψ(z))g^i(\psi(z)) and yiy_i lie in it. Hence yiy_i is UYU_Y shadowed by ψ(z)\psi(z), and gn(ψ(z))=ψ(fn(z))=ψ(xn)=yng^n(\psi(z)) = \psi(f^n(z)) = \psi(x_n) = y_n. Hence (Y,g)(Y,g) has HH-shadowing.

Theorem 4.3. Suppose ψ\psi is a factor map from (X,f)(X,f) to (Y,g)(Y,g). We will prove that if (Y,g)(Y,g) has HH-shadowing, then ψ\psi must have the ALhP property.

Proof. Let ϵ>0\epsilon> 0 and η>0\eta> 0. Choose δ>0\delta> 0 such that (Y,g)(Y,g) witnesses HH-shadowing with respect to ϵ\epsilon. Let y0,…,yny_0, \ldots, y_n be a finite δ\delta pseudo orbit. Then it must be ϵ\epsilon-shadowed by a point say zz, such that gn(z)=yng^n(z) = y_n. Since ψ\psi is surjective we know there exists xx such that ψ(x)=z\psi(x) = z. Now consider x,f(x),…,fn(x)x, f(x), \ldots, f^n(x). This is trivially a η\eta pseudo orbit. Note that for each ii we must have that ψ(fi(x))=gi(ψ(x))=gi(z)\psi(f^i(x)) = g^i(\psi(x)) = g^i(z), so ψ(fi(x))\psi(f^i(x)) ϵ\epsilon-shadows y0,…,yny_0, \ldots, y_n. In particular ψ(fn(x))=gn(ψ(x))=gn(z)=yn\psi(f^n(x)) = g^n(\psi(x)) = g^n(z) = y_n. And so (Y,g)(Y,g) has the ALhP property.

In [9], it was shown that HH-shadowing is preserved under the induced hyperspace.

Periodic shadowing

Definition 5.1. Periodic shadowing - Metric definition: We say that (X,f)(X,f) has periodic shadowing if for all ϵ>0\epsilon> 0 there exists δ>0\delta> 0 such that if (x1,…,xn)(x_1, \ldots, x_n) is a periodic δ\delta-pseudoorbit, then there exists a periodic z∈Xz \in X such that d(fi(z),xi)<ϵd(f^i(z),x_i) < \epsilon.

Definition 5.2. Periodic shadowing- Topological definition: A dynamical system (X,f)(X,f) has periodic shadowing if for all U∈FOC(X)U \in FOC(X), there exists a V∈FOC(X)V \in FOC(X) such that if (xi)(x_i) is a periodic VV-pseudoorbits (xi)(x_i), we can find a periodic z∈Xz \in X such that fi(z),xi∈Uif^i(z),x_i \in U_i.

Periodic shadowing considers pseudo-orbits that approximate the behaviour of periodic orbits. It states that periodic pseudo-orbits can be shadowed by a genuine periodic point. It has been shown that there is a close relationship between a dynamical system exhibiting periodic shadowing, and it being Ω\Omega-stable [12, 5].

Theorem 5.1. Let (X,f)(X,f) be a dynamical system where XX is a compact metric space. The topological and metric definitions of periodic shadowing are equivalent.

Proof. Suppose the topological definition of periodic shadowing holds in (X,f)(X,f). Let ϵ>0\epsilon> 0. Since XX is compact we can choose an open cover UU of XX which consists of a finite number of ϵ4\frac{\epsilon}{4}-balls. Choose V∈FOC(X)V \in FOC(X) which satisfies Definition 5.2 with respect to UU. Let δ>0\delta> 0 denote the Lebesgue number of VV. Let x0,x1,…,xnx_0,x_1,\ldots,x_n be a δ\delta-periodic pseudo orbit. Then for all i∈{0,1,…,n}i \in\{0,1,\ldots,n\} we have d(f(xi−1),xi)<δd(f(x_{i-1}),x_i)<\delta. Hence there exists a Vk∈VV_k \in V such that f(xi−1),xi∈Vkf(x_{i-1}),x_i \in V_k. Hence this δ\delta-periodic pseudo orbit is also a VV-periodic pseudo orbit. Choose a periodic zz which satisfies Definition 5.2. As we have topological shadowing, we know that for all i∈{0,1,…,n}i \in\{0,1,\ldots,n\}, there exists a Uj∈UU_j \in U such that fi(z)f^i(z) and (xi)(x_i) both belong to UjU_j. Since UU consists of ϵ4\frac{\epsilon}{4}-balls, we know that d(fi(z),xi)<ϵ2d(f^i(z),x_i)<\frac{\epsilon}{2}. Hence the metric definition is satisfied.

Now we consider the case where the metric definition holds. Consider an finite open cover UU of XX. Let ϵ>0\epsilon>0 denote the Lebesgue number of UU. Choose δ>0\delta>0 which satisfies Definition 5.1 with respect to ϵ\epsilon. As XX is compact we can choose a finite open cover VV of XX which consists of δ2\frac{\delta}{2}-balls. Let x0,x1,…,xnx_0,x_1,\ldots,x_n be a VV-periodic pseudo orbit. Hence for all i∈{0,1,…,n}i \in\{0,1,\ldots,n\}, we have that f(xi−1),xi∈Vkf(x_{i-1}),x_i \in V_k for some Vk∈VV_k \in V. As VV consists of δ2\frac{\delta}{2}-balls, clearly x0,x1,…,xnx_0,x_1,\ldots,x_n is also a δ\delta-periodic pseudo orbit. Choose zz which satisfies Definition 5.1. zz is periodic and satisfies d(fi(z),xi)<ϵd(f^i(z),x_i)<\epsilon for i∈{0,1,…,n}i \in\{0,1,\ldots,n\}. Hence there exists a UjU_j such that fi(z),xi∈Ujf^i(z),x_i \in U_j. Hence the topological definition is also satisfied.

Definition 5.3. Almost Lifting Periodic Pseudo orbit property (ALPP) Metric definition : We say a factor map has ALPP if for all ϵ>0\epsilon>0 and for all η>0\eta>0, there exist a δ>0\delta>0, such that if (yi)i∈ω(y_i)_{i\in\omega} is a δ\delta periodic pseudo orbit in YY, then there exists a η\eta periodic pseudo orbit in XX, such that d(ϕ(xi),yi)<ϵd(\phi(x_i),y_i)<\epsilon.

We can extend this definition more generally to topological spaces by using the following definition:

Definition 5.4. Almost Lifting Periodic Pseudo orbit property (ALPP) Topological definition: We say a factor map has ALPP if for all VX∈FOC(X)V_X \in FOC(X), and WY∈FOC(Y)W_Y \in FOC(Y), there exists VY∈FOC(Y)V_Y \in FOC(Y), such that if (yi)i∈ω(y_i)_{i\in\omega} is a VYV_Y periodic pseudo orbit in YY, then there exists a VXV_X periodic pseudo orbit in XX, such that ϕ(xi),yi∈Wk\phi(x_i),y_i \in W_k, for some kk. In this definition FOCFOC just means finite open cover.

Theorem 5.2. Let (X,f)(X,f) and (Y,g)(Y,g) be two dynamical systems. suppose (X,f)(X,f) has periodic shadowing and ψ:X⟶Y\psi:X\longrightarrow Y has the ALPP property, then (Y,g)(Y,g) has periodic shadowing.

Proof. Fix an open cover UYU_Y of YY. Let WYW_Y be an open cover of YY such that if W,W′∈WYW,W' \in W_Y and W∩W′≠∅W\cap W'\ne\emptyset, there exist U∈UYU\in U_Y such that W∪W′⊂UW\cup W'\subset U. Let UXU_X be an open cover of XX such that ψ(UX)\psi(U_X) refines WYW_Y. Since (X,f)(X,f) has periodic shadowing, let VXV_X witness this with respect to UXU_X. Since ψ\psi has ALPP, let VYV_Y witness this with respect to WYW_Y and VXV_X.

Let (yi)i∈ω(y_i)_{i\in\omega} be a periodic VYV_Y periodic pseudo orbit. Let (xi)i∈ω(x_i)_{i\in\omega} be the periodic VXV_X pseudo orbit, such that ψ((xi)i∈ω)\psi((x_i)_{i\in\omega}) WYW_Y shadows (yi)i∈ω(y_i)_{i\in\omega}.

Since (X,f)(X,f) has periodic shadowing, we know that xix_i is UXU_X shadowed by a periodic point say z∈Xz\in X. Hence for all ii we have that ψ(fi(z)),ψ(xi)∈ψ(Ui)\psi(f^i(z)),\psi(x_i)\in\psi(U_i) for some Ui∈UXU_i\in U_X. Since ψ\psi is a factor map we know that ψ(fi(z))=gi(ψ(z))\psi(f^i(z))=g^i(\psi(z)). And since ψ(UX)\psi(U_X) refines WYW_Y, it follows that there exists WiW_i such that gi(ψ(z))g^i(\psi(z)) and ψ(xi)\psi(x_i) lie in WiW_i. But we also know that ψ(xi)\psi(x_i) and yiy_i belong to WjW_j for some jj. Hence ψ(xi)\psi(x_i) is in Wi∩WjW_i\cap W_j. But by construction of WYW_Y we know there exists a UiU_i such that gi(ψ(z))g^i(\psi(z)) and yiy_i lie in it. Hence (yi)(y_i) is UYU_Y shadowed by ψ(z)\psi(z), and since the image of periodic points under factor maps are periodic, we have shown that (Y,g)(Y,g) has periodic shadowing.

:::

Theorem 5.3. Suppose ψ\psi is a factor map from (X,f)(X,f) to (Y,g)(Y,g). We will prove that if (Y,g)(Y,g) has the periodic shadowing property, then ψ\psi must have the ALPP property.

Proof. Let ϵ>0\epsilon>0 and η>0\eta>0. Choose δ>0\delta>0 such that it witnesses periodic shadowing with respect to ϵ\epsilon. Let (yi)(y_i) be a δ\delta periodic pseudo orbit. Then it must be shadowed by a periodic point say zz. So there exists an nn such that gn(z)=zg^n(z)=z. Consider a covering of XX by open balls of radius η2\frac{\eta}{2}. Since XX is compact this covering has a finite sub cover. Hence ψ−1\psi^{-1} is contained in finite number of these open balls of radius η2\frac{\eta}{2}. Since ψ\psi is surjective we know there exists x∗x^* such that ψ(x∗)=z\psi(x^*)=z. Consider the sequence x∗,f(x∗),f2(x∗),…x^*,f(x^*),f^2(x^*),\ldots. Note that x∗,fn(x∗)f2n(x∗),f3n(x∗),…x^*,f^n(x^*)f^{2n}(x^*),f^{3n}(x^*),\ldots all belong in ψ−1(z)\psi^{-1}(z). By the pigeonhole principle, at least one of the open balls of

radius η2\frac{\eta}{2} that cover ψ−1(z)\psi^{-1}(z) contains fin(x∗)f^{in}(x^*), fkn(x∗)f^{kn}(x^*) where i<ki<k. But then consider the η\eta periodic pseudo orbit given by fin(x)f^{in}(x), fin+1(x)f^{in+1}(x), …\ldots fkn−1(x)f^{kn-1}(x), fin(x)f^{in}(x), fin+1(x)f^{in+1}(x), …\ldots fkn−1(x∗)f^{kn-1}(x^*), fin(x∗)f^{in}(x^*), fin+1(x∗)f^{in+1}(x^*), …\ldots

Clearly this η\eta periodic pseudo orbit satisfies the condition needed for the ALPP property to hold. Hence the map ψ\psi satisfies the ALPP property.

Ramsey shadowing

Definition 6.1. Furstenberg family:

We call F\mathcal{F} a Furstenberg family, if it is a nonempty set of subsets of N\mathbb{N} and it is closed under supersets.

Definition 6.2. Ramsey property:

We say F\mathcal{F}, a Furstenberg family, has the Ramsey property, if whenever A∈FA \in\mathcal{F} and A=⋃i<nAiA = \bigcup_{i<n} A_i, then Ai∈FA_i \in\mathcal{F} for some ii.

Ramsey shadowing has shown itself to be strongly connected to minimal points and other fundamental concepts in dynamical systems, including recurrence. In particular, it has been shown that Ramsey shadowing is equivalent to the existence of a dense set of minimal points [4].

Definition 6.3. Ramsey shadowing- Metric definition: A dynamical system (X,f)(X,f) has Ramsey shadowing if for all ϵ>0\epsilon>0, for all F\mathcal{F} with the Ramsey property, any sequence (xi)(x_i) can be ϵ\epsilon-shadowed on a set in F\mathcal{F}.

Definition 6.4. Ramsey shadowing- Topological definition: A dynamical system (X,f)(X,f) has Ramsey shadowing if for all U\mathcal{U} finite open cover of XX, for all F\mathcal{F} with the Ramsey property, any sequence (xi)(x_i) can be U\mathcal{U}-shadowed on a set in F\mathcal{F}.

Theorem 6.1. Let (X,f)(X,f) be a dynamical system where XX is a compact metric space. The topological and metric definitions of Ramsey shadowing are equivalent.

Proof. Suppose the topological definition of Ramsey shadowing holds in (X,f)(X,f). Let ϵ>0\epsilon>0. Since XX is compact we can choose an open cover U\mathcal{U} of XX which consists of a finite number of ϵ4\frac{\epsilon}{4}-balls. Pick an arbitrary Furstenberg family with the Ramsey property and an arbitrary sequence (xi)i∈N0(x_i)_{i\in\mathbb{N}_0}. Choose a zz which satisfies Definition 6.4. As we have topological Ramsey shadowing, we know that there exists a z∈Xz\in X and an A∈FA\in\mathcal{F} such that for all i∈Ai\in A, there exists a Uj∈UU_j\in\mathcal{U} such that fi(z)f^i(z) and (xi)(x_i) both belong to UjU_j. Since U\mathcal{U} consists of ϵ4\frac{\epsilon}{4}-balls, we know that d(fi(z),xi)<ϵ2d(f^i(z),x_i)<\frac{\epsilon}{2}. Hence the metric definition of Ramsey shadowing holds in (X,f)(X,f), as (xi)i∈N0(x_i)_{i\in\mathbb{N}_0} is ϵ\epsilon-shadowed by zz on AA.

Now we consider the case where the metric definition of Ramsey-shadowing holds. Consider a finite open cover U\mathcal{U} of XX. Let ϵ>0\epsilon>0 denote the Lebesgue number of U\mathcal{U}. Pick an arbitrary sequence (xi)i∈N0(x_i)_{i\in\mathbb{N}_0} and an arbitrary Furstenberg family with the Ramsey property. Choose zz which satisfies Definition 6.3. We know that there exists a z∈Xz\in X and a B∈FB\in\mathcal{F} such that for all i∈Ai\in A, we have d(fi(z),xi)<ϵd(f^i(z),x_i)<\epsilon. Therefore (xi)i∈N0(x_i)_{i\in\mathbb{N}_0} is U\mathcal{U}-shadowed on BB, and so (X,f)(X,f) also exhibits topological Ramsey shadowing.

:::

Theorem 6.2. Suppose ψ\psi is a factor map from (X,f)(X,f) to (Y,g)(Y,g). We will show that if (X,f)(X,f) has Ramsey shadowing, then so does (Y,g)(Y,g)

Proof. Fix an open cover UY\mathcal{U}_Y of YY. Let WY\mathcal{W}_Y be an open cover of YY such that if W,W′∈WYW,W'\in\mathcal{W}_Y and W∩W′≠∅W\cap W'\ne\emptyset, there exist U∈UYU\in\mathcal{U}_Y such that W∪W′⊂UW\cup W'\subset U. Let UX\mathcal{U}_X be an open cover of XX such that ψ(UX)\psi(\mathcal{U}_X) refines WY\mathcal{W}_Y.

Let (yi)∈Y(y_i)\in Y. Let F\mathcal{F} be arbitrary with the Ramsey property. Let (xi)(x_i) be the sequence defined by xi=ψ−1(yi)x_i=\psi^{-1}(y_i). By Ramsey shadowing we know that xix_i is UX\mathcal{U}_X-shadowed by a point z∈Xz\in X on some A∈FA\in\mathcal{F}. We note that ψ(xi)\psi(x_i) certainly WY\mathcal{W}_Y-shadows (yi)(y_i) on AA. From now on we assume the subscript is on AA. But then ψ(xi)\psi(x_i) and (yi)(y_i) belong to some WkW_k. But also ψ(xi)\psi(x_i) and ψ(fi(z))\psi(f^i(z)) also belong to some WtW_t, hence WtW_t and WkW_k is non empty so by construction we must have that ψ(fi(z))\psi(f^i(z)) and (yi)(y_i) belong to some UsU_s. But note that ψ(fi(z))=gi(ψ(z))\psi(f^i(z))=g^i(\psi(z)). And so ψ(z)\psi(z) UY\mathcal{U}_Y-shadows (yi)(y_i) on AA.

:::

Theorem 6.3. If the hyperspace (2X,2f)(2^X,2^f) has Ramsey shadowing then so does the (X,f)(X,f).

Proof. Let ϵ>0\epsilon> 0. Let F\mathcal{F} be an arbitrary Furstenberg family with the Ramsey property. Consider an arbitrary sequence (xi)(x_i). Now consider the sets {x1},{x2},…\{x_1\}, \{x_2\}, \ldots. These are compact sets hence belong to 2X2^X. Since the hyperspace has Ramsey shadowing we know that there exist a compact set A∈2XA \in2^X which ϵ\epsilon-shadows x1,x2,…x_1, x_2, \ldots on a set BB in F\mathcal{F}. But then by the definition of Hausdorff distance we see that any element of AA must shadow (xi)(x_i) on B∈FB \in\mathcal{F}. Hence (X,f)(X,f) has Ramsey shadowing.

Theorem 6.4. The inverse limit of a sequence of dynamical systems, each exhibiting the Ramsey shadowing property and satisfying the Mittag-Leffler condition, also has the Ramsey shadowing property.

Proof. Suppose (X,f)(X,f) is conjugate to the inverse limit system (Xn,fn,gkn)(X_n,f_n,g_k^n) where each (Xn,fn)(X_n,f_n) has the Ramsey shadowing property. Let W\mathcal{W} be a finite open cover of XX. Let F\mathcal{F} be a Furstenberg family with the Ramsey property. We can find a kk and a finite open cover Bk\mathcal{B}_k of XkX_k such that the B∗={πk−1(B)∩X:B∈Bk}\mathcal{B}^* = \{\pi_k^{-1}(B) \cap X : B \in\mathcal{B}_k\} refine W\mathcal{W}.

Using the Mittag-Leffler condition we know there exists a γ\gamma which satisfies the condition with respect to kk. Bγ={(gkγ)−1(B):B∈Bk}\mathcal{B}_\gamma= \{(g_k^\gamma)^{-1}(B) : B \in\mathcal{B}_k\}. Let (xi)(x_i) be an arbitrary sequence of point in XX. This means that (πγ(xi))(\pi_\gamma(x_i)) is a sequence of points in XγX_\gamma. Hence we know that there exists a point xγx_\gamma which Bγ\mathcal{B}_\gamma-shadows on it on some A∈FA \in\mathcal{F}. Hence it follows that gkγ(xγ)g_k^\gamma(x_\gamma) Bk\mathcal{B}_k-shadows (πk(xi))(\pi_k(x_i)) on A∈FA \in\mathcal{F}.

Note that we have that

gkγ(x∗)∈gkγ(⋂i∈Af−i(Bγi))⊆⋂i∈Af−i(Bki)g_k^\gamma(x^*) \in g_k^\gamma\left(\bigcap_{i\in A}f^{-i}(B_\gamma^i)\right) \subseteq\bigcap_{i\in A}f^{-i}(B_k^i)

And so, using the Mittag-Leffler condition we know there exists an zz such that (z)k∈gkγ(⋂i∈Af−i(Bγi))(z)_k \in g_k^\gamma\left(\bigcap_{i\in A}f^{-i}(B_\gamma^i)\right). And so we must have that fi(z),xi∈πk−1(Wki)∩Xf^i(z),x_i \in\pi_k^{-1}(W_k^i) \cap X. Hence (xi)(x_i) is B∗\mathcal{B}^*-shadowed on A∈FA \in\mathcal{F}. But since B∗\mathcal{B}^* refines W\mathcal{W}, we must also have that (xi)(x_i) is W\mathcal{W}-shadowed on A∈FA \in\mathcal{F}.

Ergodic Shadowing

Definition 7.1. Ergodic pseudo-orbit - Metric Definition:

Let δ>0\delta> 0. Let (xi)(x_i) be a sequence. Let N((xi),δ)={i:d(f(xi),xi+1)≥δ}N((x_i),\delta) = \{i : d(f(x_i),x_{i+1}) \ge\delta\}. Let N((xi),δ,n)={i:d(f(xi),xi+1)≥δ,i=0,…,n}N((x_i),\delta,n) = \{i : d(f(x_i),x_{i+1}) \ge\delta, i = 0,\ldots,n\}. Let M((xi),δ,z)={i:d(fi(z),xi)≥δ}M((x_i),\delta,z) = \{i : d(f^i(z),x_i) \ge\delta\} and M((xi),δ,z,n)={i:d(fi(z),xi)≥δ,i=0,…,n}M((x_i),\delta,z,n) = \{i : d(f^i(z),x_i) \ge\delta, i = 0,\ldots,n\}. We call (xi)(x_i) a δ\delta-ergodic pseudo orbit provided that

lim⁡n→∞∣N((xi),δ,n)∣n=0.\lim_{n\to\infty}\frac{|N((x_i),\delta,n)|}{n}=0.

We say that a δ\delta-ergodic pseudo orbit (xi)(x_i) is ϵ\epsilon-ergodic shadowed by a point zz provided that

lim⁡n→∞∣M((xi),ϵ,z,n)∣n=0.\lim_{n\to\infty}\frac{|M((x_i),\epsilon,z,n)|}{n}=0.

Definition 7.2. Ergodic pseudo-orbit - Topological definition:

Let V∈FOC(X)\mathcal{V} \in FOC(X), we call (xi)(x_i) a V\mathcal{V}-ergodic pseudo-orbit if:

lim⁡n→∞∣{i∈[0,…,n]:∀Vk∈V,f(xi)∉Vk or xi+1∉Vk}∣n=0\lim_{n\to\infty}\frac{|\{i\in[0,\ldots,n]:\forall V_k\in\mathcal{V}, f(x_i)\notin V_k\text{ or }x_{i+1}\notin V_k\}|}{n}=0

Definition 7.3. Ergodic shadowing - Metric definition:

We say the dynamical system has ergodic-shadowing if for all ϵ>0\epsilon> 0, there exists δ>0\delta> 0 such that if (xi)(x_i) is a δ\delta-ergodic pseudo orbit, then it is ϵ\epsilon ergodic shadowed by a point zz.

Ergodic shadowing was introduced in [6], and has shown itself to be closely linked to various notions of mixing. In fact, in [6], they showed that ergodic shadowing was equivalent to having both the normal shadowing property as well as topological mixing.

Definition 7.4. Ergodic shadowing- Topological definition: A dynamical system (X,f)(X,f) has ergodic shadowing if for all U∈FOC(X)\mathcal{U} \in FOC(X), there exists a V∈FOC(X)\mathcal{V} \in FOC(X) such that for all ergodic V\mathcal{V}-pseudo orbits, we can find a z∈Xz \in X such that:

lim⁡n→∞∣{i∈[0,…,n]:∀Uk∈U,fi(z)∉Uk or xi∉Uk}∣n=0\lim_{n\to\infty}\frac{|\{i\in[0,\ldots,n]:\forall U_k\in\mathcal{U}, f^i(z)\notin U_k\text{ or }x_i\notin U_k\}|}{n}=0

As in the previous cases, the metric and topological definitions of the Ergodic shadowing can be shown to be equivalent. We omit the proof, as it follows essentially the same arguments as before.

Definition 7.5. Ergodic Lifting Property (ELP) - Topological definition:

We say a factor map has ELP if for all UX∈FOC(X)U_X \in FOC(X) and for all WY∈FOC(Y)W_Y \in FOC(Y), there exists VY∈FOC(Y)V_Y \in FOC(Y), such that if (yi)(y_i) is a VYV_Y-ergodic pseudo orbit, there exists a UXU_X-ergodic pseudo orbit (xi)(x_i) in XX such that:

lim⁡n→∞∣{i∈[0,…,n]:∀Wt∈WY, ψ(xi)∉Wt or yi∉Wt}∣n=0.\lim_{n\to\infty}\frac{\left|\left\{i\in[0,\ldots,n]:\forall W_t\in W_Y,\ \psi(x_i)\notin W_t\ \text{or}\ y_i\notin W_t\right\}\right|}{n}=0.

Definition 7.6. Ergodic Lifting Property (ELP) - Metric definition:

We say the dynamical system has ELP if for all ϵ>0\epsilon>0 and for all η>0\eta>0, there exists δ>0\delta>0, such that if (yi)(y_i) δ\delta-ergodic pseudo orbit, there exists a η\eta-ergodic pseudo orbit (xi)(x_i) in XX such that:

lim⁡n→∞∣{i∈[0,…,n]:d(ψ(xi),yi)>ϵ}∣n=0.\lim_{n\to\infty}\frac{\left|\left\{i\in[0,\ldots,n]:d(\psi(x_i),y_i)>\epsilon\right\}\right|}{n}=0.

Theorem 7.1. Suppose ψ\psi is a factor map from (X,f)(X,f) to (Y,g)(Y,g). We will prove that if (Y,g)(Y,g) has ergodic shadowing, then ψ\psi must have ELP.

Proof. Let ϵ>0\epsilon>0 and η>0\eta>0. Choose δ>0\delta>0, such that YY satisfies ergodic shadowing with respect to ϵ\epsilon. Let (yi)(y_i) be a δ\delta-ergodic pseudo orbit. We know it is ϵ\epsilon-ergodic shadowed by a point say zz.

So we must have that

lim⁡n→∞∣M((xi),ϵ,z,n)∣n=0.\lim_{n\to\infty}\frac{|M((x_i),\epsilon,z,n)|}{n}=0.

Take x∈ψ−1(z)x\in\psi^{-1}(z) and consider the sequence (xi)(x_i) given by its orbit (x,f(x),…)(x,f(x),\ldots). This is clearly a η\eta-ergodic pseudo orbit. And by semi-conjugacy we know that when ψ(fi(x))=gi(z)\psi(f^i(x))=g^i(z). And since zz ϵ\epsilon-ergodic shadows yiy_i we know that:

lim⁡n→∞∣{i∈[0,…,n]:d(ψ(xi),yi)>ϵ}∣n=0.\lim_{n\to\infty}\frac{\left|\left\{i\in[0,\ldots,n]:d(\psi(x_i),y_i)>\epsilon\right\}\right|}{n}=0.

Hence ψ\psi must have ELP.

:::

Theorem 7.2. Let (X,f)(X,f) and (Y,g)(Y,g) be two dynamical systems. suppose (X,f)(X,f) has ergodic shadowing and ψ:X⟶Y\psi:X\longrightarrow Y has ELP, then (Y,g)(Y,g) has ergodic shadowing.

Proof. Fix an open cover UYU_Y of YY. Let WYW_Y be an open cover of YY such that if W,W′∈WYW,W'\in W_Y and W∩W′≠∅W\cap W'\ne\emptyset, there exist U∈UYU\in U_Y such that W∪W′⊂UW\cup W'\subset U. Let UXU_X be an open cover of XX such that ψ(UX)\psi(U_X) refines WYW_Y. Since XX has ergodic-shadowing, let VXV_X be an open cover that satisfies this with respect to UXU_X. Since ψ\psi has ELP let VYV_Y respect it with regards to WYW_Y and VXV_X.

Let (yi)∈Y(y_i)\in Y be a VYV_Y-ergodic pseudo orbit. Let (xi)(x_i) be the VXV_X-ergodic pseudo orbit in XX that satisfies ELP. By ergodic-shadowing we know that xix_i is UXU_X-shadowed by a point z∈Xz\in X, in such a way that:

lim⁡n→∞∣{i∈[0,…,n]:∀A∈UX, fi(z)∉A or xi∉A}∣n=0\lim_{n\to\infty}\frac{\left|\left\{i\in[0,\ldots,n]:\forall A\in U_X,\ f^i(z)\notin A\ \text{or}\ x_i\notin A\right\}\right|}{n}=0

But by the ELP property we note that ψ(xi)\psi(x_i) WYW_Y-shadows (yi)(y_i) in such a way that:

lim⁡n→∞∣{i∈[0,…,n]:∀B∈WY, yi∉B or ψ(xi)∉B}∣n=0\lim_{n\to\infty}\frac{\left|\left\{i\in[0,\ldots,n]:\forall B\in W_Y,\ y_i\notin B\ \text{or}\ \psi(x_i)\notin B\right\}\right|}{n}=0

Since ψ\psi refines WYW_Y, we know that whenever (xi)(x_i) and (fi(z))(f^i(z)) belong the same UXkU_{X_k}, then ψ(xi)\psi(x_i) and ψ(fi(z))=gi(ψ(z))\psi(f^i(z))=g^i(\psi(z)) must also belong to some WtW_t.

We want to show that:

lim⁡n→∞∣{i∈[0,…,n]:∀B∈WY, yi∉B or ψ(fi(z))∉B}∣n=0.\lim_{n\to\infty}\frac{\left|\left\{i\in[0,\ldots,n]:\forall B\in W_Y,\ y_i\notin B\ \text{or}\ \psi(f^i(z))\notin B\right\}\right|}{n}=0.

Let ϵ>0\epsilon>0. We know there exists M∈NM\in\mathbb{N} such that for all n>Mn>M:

∣{i∈[0,…,n]:∀A∈UX, fi(z)∉A or xi∉A}∣n<ϵ2\frac{\left|\left\{i\in[0,\ldots,n]:\forall A\in U_X,\ f^i(z)\notin A\ \text{or}\ x_i\notin A\right\}\right|}{n}<\frac{\epsilon}{2}

and

∣{i∈[0,…,n]:∀B∈WY, yi∉B or ψ(xi)∉B}∣n<ϵ2.\frac{\left|\left\{i\in[0,\ldots,n]:\forall B\in W_Y,\ y_i\notin B\ \text{or}\ \psi(x_i)\notin B\right\}\right|}{n}<\frac{\epsilon}{2}.

We note that {i∈[0,…,n]:∀B∈UY, yi∉B or ψ(fi(z))∉B}⊂{i∈[0,…,n]:∀A∈UX, xi∉A or fi(z)∉A}∪{i∈[0,…,n]:∀B∈WY, yi∉B or ψ(xi)∉B}\{i\in[0,\ldots,n]:\forall B\in U_Y,\ y_i\notin B\ \text{or}\ \psi(f^i(z))\notin B\}\subset\{i\in[0,\ldots,n]:\forall A\in U_X,\ x_i\notin A\ \text{or}\ f^i(z)\notin A\}\cup\{i\in[0,\ldots,n]:\forall B\in W_Y,\ y_i\notin B\ \text{or}\ \psi(x_i)\notin B\} Hence we must have that:

∣{i∈[0,…,n]:∀B∈UY, yi∉B or ψ(fi(z))∉B}∣n<ϵ2+ϵ2=ϵ.\frac{\left|\left\{i\in[0,\ldots,n]:\forall B\in U_Y,\ y_i\notin B\ \text{or}\ \psi(f^i(z))\notin B\right\}\right|}{n}<\frac{\epsilon}{2}+\frac{\epsilon}{2}=\epsilon.

Hence: lim⁡n→∞∣{i∈[0,…,n]:∀B∈UY, yi∉B or ψ(fi(z))∉B}∣n=0\lim_{n\to\infty}\frac{\left|\left\{i\in[0,\ldots,n]:\forall B\in U_Y,\ y_i\notin B\ \text{or}\ \psi(f^i(z))\notin B\right\}\right|}{n}=0.

But then ψ(fi(z))=gi(ψ(z))\psi(f^i(z))=g^i(\psi(z)). And so ψ(z)\psi(z) UYU_Y-ergodic shadows (yi)(y_i), hence we have ergodic shadowing.

Theorem 7.3. If the hyperspace (2X,2f)(2^X,2^f) has ergodic shadowing then so does the (X,f)(X,f).

Proof. Let ϵ>0\epsilon>0. Choose δ\delta such that (2X,2f)(2^X,2^f) has ergodic shadowing. Now consider a δ\delta-ergodic pseudo orbit in XX, (xi)(x_i). Now consider the sets {x1},{x2},…\{x_1\},\{x_2\},\ldots. These are compact sets hence belong to 2X2^X. Since the hyperspace has ergodic-shadowing we know that there exist a compact set A∈2XA\in2^X such that:

lim⁡n→∞M({xi},ϵ,(2f)i(A),n)n=0.\lim_{n\to\infty}\frac{M(\{x_i\},\epsilon,(2^f)^i(A),n)}{n}=0.

But then by the definition of Hausdorff distance we see that any element of a∈Aa\in A must satisfy:

lim⁡n→∞M(xi,ϵ,fi(a),n)n=0.\lim_{n\to\infty}\frac{M(x_i,\epsilon,f^i(a),n)}{n}=0.

Hence (X,f)(X,f) also has ergodic shadowing.

Density shadowing

Definition 8.1. d-shadowing- Metric definition :

We say the dynamical system has d-shadowing if for all ϵ>0\epsilon>0, there exists δ>0\delta>0 such that if (xi)(x_i) is a δ\delta-ergodic pseudo orbit, then there is a point zz such that:

lim inf⁡n→∞∣{i∈[0,…,n]:d(fi(z),xi)<ϵ}∣n>0\liminf_{n\to\infty}\frac{\left|\left\{i\in[0,\ldots,n]:d(f^i(z),x_i)<\epsilon\right\}\right|}{n}>0

Density shadowing was introduced in [7] and has been shown to be a useful tool in the study of equicontinuity and transitivity [14].

Definition 8.2. d-shadowing- Topological definition: A dynamical system (X,f)(X,f) has d shadowing if for all U∈FOC(X)U\in FOC(X), there exists a V∈FOC(X)V\in FOC(X) such that for all ergodic VV-pseudo orbits, we can find a z∈Xz\in X such that:

lim inf⁡n→∞∣{i∈[0,…,n]:fi(z) and xi∈Uk for some k}∣n>0\liminf_{n\to\infty}\frac{\left|\left\{i\in[0,\ldots,n]:f^i(z)\ \text{and}\ x_i\in U_k\ \text{for some }k\right\}\right|}{n}>0

Theorem 8.1. Let (X,f)(X,f) be a dynamical system where XX is a compact metric space. The topological and metric definitions of density shadowing are equivalent.

Proof. Suppose the topological definition of Density-shadowing holds in (X,f)(X,f). Let ϵ>0\epsilon>0. Since XX is compact we can choose an open cover UU of XX which consists of a finite number of ϵ4\frac{\epsilon}{4} balls. Choose V∈FOC(X)V\in FOC(X) which satisfies Definition 8.2 with respect to UU. Let δ>0\delta>0 denote the Lebesgue number of VV.

N((xi),δ,n)={i:d(f(xi),xi+1)>δ, i=0,…,n}N((x_i),\delta,n)=\{i:d(f(x_i),x_{i+1})>\delta,\ i=0,\ldots,n\}. N((xi),V,n)={i:∄j such that f(xi),xi+1∈Vj, i=0,…,n}N((x_i),V,n)=\{i:\nexists j\text{ such that }f(x_i),x_{i+1}\in V_j,\ i=0,\ldots,n\}.

Let (xi)i∈N0(x_i)_{i\in\mathbb{N}_0} be a δ\delta-ergodic pseudo orbit. Then lim⁡n→∞∣N((xi),δ,n)∣n=0\lim_{n\to\infty}\frac{|N((x_i),\delta,n)|}{n}=0. Pick k∈N((xi),V,n)k\in N((x_i),V,n), then there is no Vj∈VV_j\in V such that f(xk),xk+1∈Vjf(x_k),x_{k+1}\in V_j. As δ\delta is the Lebesgue number of VV this implies that d(f(xk),xk+1)>δd(f(x_k),x_{k+1})>\delta and so N((xi),V,n)⊆N((xi),δ,n)N((x_i),V,n)\subseteq N((x_i),\delta,n). Hence lim⁡n→∞∣N((xi),V,n)∣n=0\lim_{n\to\infty}\frac{|N((x_i),V,n)|}{n}=0 and so this δ\delta ergodic pseudo orbit is also a VV-ergodic pseudo orbit. Choose a zz which satisfies Definition 8.2. As we have topological density-shadowing, we know that

lim inf⁡n→∞∣{i∈[0,…,n]:d(fi(z),xi)∈U}∣n>0.\liminf_{n\to\infty}\frac{\left|\left\{i\in[0,\ldots,n]:d(f^i(z),x_i)\in U\right\}\right|}{n}>0.

Since UU consists of ϵ4\frac{\epsilon}{4}-balls, we know that whenever fi(z),xi∈Ujf^i(z),x_i\in U_j we must also have that d(fi(z),xi)<ϵ2d(f^i(z),x_i)<\frac{\epsilon}{2}. Hence lim inf⁡n→∞∣{i∈[0,…,n]:d(fi(z),xi)<ϵ}∣n>0\liminf_{n\to\infty}\frac{\left|\left\{i\in[0,\ldots,n]:d(f^i(z),x_i)<\epsilon\right\}\right|}{n}>0, and the metric definition of density shadowing also holds.

Now we consider the case where the metric definition of density shadowing holds. Consider an finite open cover UU of XX. Let ϵ>0\epsilon>0 denote the Lebesgue number of UU. Choose δ>0\delta>0 which satisfies Definition 8.1 with respect to ϵ\epsilon. As XX is compact we can choose a finite open cover VV of XX which consists of δ2\frac{\delta}{2}-balls. Let (xi)i∈N0(x_i)_{i\in\mathbb{N}_0} be a VV ergodic pseudo orbit. Then lim⁡n→∞∣N((xi),V,n)∣n=0\lim_{n\to\infty}\frac{|N((x_i),V,n)|}{n}=0. As VV consists of δ2\frac{\delta}{2}-balls, clearly {i:f(xi),xi+1∈V, i=0,…,n}⊆{i:d(f(xi),xi+1)<δ, i=0,…,n}\{i:f(x_i),x_{i+1}\in V,\ i=0,\ldots,n\}\subseteq\{i:d(f(x_i),x_{i+1})<\delta,\ i=0,\ldots,n\}. By taking complements we see that N((xi),δ,n)⊆N((xi),V,n)N((x_i),\delta,n)\subseteq N((x_i),V,n), and so lim⁡n→∞∣N((xi),δ,n)∣n=0\lim_{n\to\infty}\frac{|N((x_i),\delta,n)|}{n}=0. Hence (xi)i∈N0(x_i)_{i\in\mathbb{N}_0} also a δ\delta-ergodic pseudo orbit. Choose zz which satisfies Definition 8.1. Then lim inf⁡n→∞∣{i∈[0,…,n]:d(fi(z),xi)<ϵ}∣n>0\liminf_{n\to\infty}\frac{|\{i\in[0,\ldots,n]:d(f^{i}(z),x_i)<\epsilon\}|}{n}>0. As ϵ\epsilon is the Lebesgue number of UU, we know that whenever d(fi(z),xi)<ϵd(f^{i}(z),x_i)<\epsilon we must also have that fi(z),xi∈Ujf^{i}(z),x_i\in U_j for some Uj∈UU_j\in U. This implies that lim inf⁡n→∞∣{i∈[0,…,n]:fi(z),xi∈U}∣n>0\liminf_{n\to\infty}\frac{|\{i\in[0,\ldots,n]:f^{i}(z),x_i\in U\}|}{n}>0 and so (X,f)(X,f) also exhibits topological dd-shadowing.

Theorem 8.2. If the hyperspace (2X,2f)(2^X,2^f) has dd-shadowing then so does the (X,f)(X,f).

Proof. Let ϵ>0\epsilon>0. Choose δ\delta such that (2X,2f)(2^X,2^f) has dd-shadowing. Consider an ergodic δ\delta pseudo orbit (xi)(x_i). Now consider the sets {x1},{x2},…\{x_1\},\{x_2\},\ldots These are compact sets hence belong to 2X2^X. Since the hyperspace has dd-shadowing we know that there exist a compact set A∈2XA\in2^X such that:

lim inf⁡n→∞∣{i∈[0,…,n]:d((2f)i(A),{xi})<ϵ}∣n>0\liminf_{n\to\infty}\frac{|\{i\in[0,\ldots,n]:d((2^f)^i(A),\{x_i\})<\epsilon\}|}{n}>0

But then by the definition of Hausdorff distance we see that any element of a∈Aa\in A must satisfy:

lim inf⁡n→∞∣{i∈[0,…,n]:d(fi(a),{xi})<ϵ}∣n>0.\liminf_{n\to\infty}\frac{|\{i\in[0,\ldots,n]:d(f^i(a),\{x_i\})<\epsilon\}|}{n}>0.

Hence (X,f)(X,f) has dd-shadowing.

Theorem 8.3. The inverse limit of a sequence of dynamical systems, each exhibiting dd-shadowing and also satisfying the Mittag-Leffler condition, also has dd shadowing.

Proof. Suppose (X,f)(X,f) is conjugate to the inverse limit system (Xn,fn,gkγ)(X_n,f_n,g_k^\gamma) where each (Xn,fn)(X_n,f_n) has the dd shadowing property. Let WW be a finite open cover of XX. We can find a kk and a finite open cover BkB_k of XkX_k such that the B∗={πk−1(B)∩X:B∈Bk}B^*=\{\pi_k^{-1}(B)\cap X:B\in B_k\} refine WW.

Using the Mittag-Leffler condition we know there exists a γ\gamma which satisfies the condition with respect to kk. Consider the open cover of WγW_\gamma given by Bγ={(gkγ)−1(B):B∈Bk}B_\gamma=\{(g_k^\gamma)^{-1}(B):B\in B_k\}. Since (Xγ,fγ)(X_\gamma,f_\gamma) has dd-shadowing, let Vγ∈FOC(Xγ)V_\gamma\in FOC(X_\gamma) be the open cover that satisfies the definition. Let V={πγ−1(V)∩X:V∈Vγ}V=\{\pi_\gamma^{-1}(V)\cap X:V\in V_\gamma\}. Let (xi)(x_i) be an ergodic VV-pseudo orbit sequence in XX, which is given by the sequence (πγ−1(Vi)∩X)(\pi_\gamma^{-1}(V_i)\cap X). This means that (πγ(xi))(\pi_\gamma(x_i)) is an ergodic VγV_\gamma-pseudo orbit in XγX_\gamma with sequence (Viγ)(V_i^\gamma). Since (Xγ,fγ)(X_\gamma,f_\gamma) has dd-shadowing we know that there exists an point xγx_\gamma such that:

lim inf⁡n→∞∣{i∈[0,…,n]:d(fγi(z),xi)∈Bγ}∣n>0\liminf_{n\to\infty}\frac{|\{i\in[0,\ldots,n]:d(f_\gamma^i(z),x_i)\in B_\gamma\}|}{n}>0

Hence it follows that gkγ(x∗)g_k^\gamma(x^\ast) satisfies:

lim inf⁡n→∞∣{i∈[0,…,n]:d(gkγ(z),(xi)k)∈Bk}∣n>0\liminf_{n\to\infty}\frac{|\{i\in[0,\ldots,n]:d(g_k^\gamma(z),(x_i)_k)\in B_k\}|}{n}>0

Let A∗A^\ast denotes the values for which d(fi(z),xi)∈Bγd(f^i(z),x_i)\in B_\gamma. Note that we have that gkγ(x∗)∈gkγ(⋂i∈A∗(f−i(Biγ)))⊆⋂(f−i(Bik))g_k^\gamma(x^\ast)\in g_k^\gamma\left(\bigcap_{i\in A^\ast}(f^{-i}(B_i^\gamma))\right)\subseteq\bigcap(f^{-i}(B_i^k)).

And so, using the Mittag-Leffler condition we know there exists an zz such that (z)k∈gkγ(⋂i∈A∗(f−i(Byγ)))(z)_k\in g_k^\gamma\left(\bigcap_{i\in A^\ast}(f^{-i}(B_y^\gamma))\right). And so we must have that fi(z),xi∈πk−1(Wki)∩Xf^i(z),x_i\in\pi_k^{-1}(W_k^i)\cap X for i∈A∗i\in A^\ast.

Hence (xi)(x_i) is dd-shadowed. But since B∗B^\ast refines WW, we must also have that (xi)(x_i) is dd-shadowed.

Acknowledgments

The first author, Bura, is thankful to EPSRC grants EP/S02297X/1 for financial support during the development of this article.

References

  1. [1]Dmitri V. Anosov. Geodesic flows on closed Riemannian manifolds with negative curvature, volume 90 of Proceedings of the Steklov Institute of Mathematics. American Mathematical Society, Providence, RI, 1969. Translation of the 1967 Russian original.
  2. [2]Andrew D. Barwell, Chris Good, Piotr Op̨rocha, and Brian E. Raines. Characterizations of ω-limit sets in topologically hyperbolic systems. Discrete & Continuous Dynamical Systems, 33(5):1819–1833, 2013.DOI
  3. [3]Rufus Bowen. ω-limit sets for axiom a diffeomorphisms. Journal of Differential Equations, 18(2):333–339, 1975.DOI
  4. [4]W. R. Brian. Ramsey shadowing and minimal points. Proceedings of the American Mathematical Society, 144(6):2697–2703, 2016.DOI
  5. [5]Jifeng Chu, Zhaosheng Feng, and Ming Li. Periodic shadowing of vector fields. Discrete & Continuous Dynamical Systems, 36(7):3623–3638, 2016.DOI
  6. [6]A. Fakhari and F. H. Ghane. On shadowing: Ordinary and ergodic. Journal of Mathematical Analysis and Applications, 364(1):151–155, 2010.DOI
  7. [7]A. Fakhari and F. H. Ghane. Subshadowings. Nonlinear Analysis: Theory, Methods & Applications, 72(9-10):3759–3766, 2010.
  8. [8]C. Good and J. Meddaugh. Shifts of finite type as fundamental objects in the theory of shadowing. Invent. math., 220:715–736, 2020.arxiv.org/abs/1702.05170
  9. [9]Chris Good, Joel Mitchell, and Joe Thomas. Preservation of shadowing in discrete dynamical systems. Journal of Mathematical Analysis and Applications, 485(1):123767, 2020.DOI
  10. [11]Keonhee Lee. Continuous inverse shadowing and hyperbolicity. Bulletin of the Australian Mathematical Society, 79(2):315–323, 2009.
  11. [12]Alexey V. Osipov, Sergei Yu. Pilyugin, and Sergey B. Tikhomirov. Periodic shadowing and ω-stability. Regular and Chaotic Dynamics, 15(4):404–417, 2010.DOI
  12. [13]S. Yu. Pilyugin. Inverse shadowing by continuous methods. Discrete and Continuous Dynamical Systems, 8(1):29–38, 2002.DOI
  13. [14]Xinxing Wu, Piotr Oprocha, and Guanrong Chen. On various definitions of shadowing with average error in tracing. Nonlinearity, 29(7):1942–1972, 2014.arxiv.org/abs/1406.5822

Paper details

vlnc:W_g59n6br8
Version
v1 · latest
Published
Fresh — added to the Hub within the last 30 days
Added
License
CC BY 4.0
Usage

views · 5

downloads · 0

A view counts once per browser per day, a download each time the PDF is requested.

  • Mathematics

Not recorded for this paper

View statement
Looking for related papers…
Contents