Sufficient conditions for distributional chaos of type I

Abstract.

Distributional chaos of type I (DC1) is a stronger variant of Li–Yorke chaos. In this paper, we consider the fact that the time-one map of a mixing Anosov flow exhibits DC1 and generalize it to obtain simple sufficient conditions for DC1.

1. Introduction

The concept of chaos plays a central role in the modern theory of dynamical systems. A mathematical definition of chaos, so-called Li–Yorke chaos, was given in [12]. In [16], three statistical variants of Li–Yorke chaos, collectively called distributional chaos, were introduced (initially) for interval maps. Among the three, distributional chaos of type I (DC1) is the strongest one, representing a stronger variant of Li–Yorke chaos. While the Li–Yorke type chaos has been generalized in terms of Furstenberg families, this paper focuses specifically on DC1 (see [11, 15] for background). We show that the time-one map of a mixing Anosov flow exhibits DC1 and generalize it to obtain sufficient conditions for DC1.

We begin by defining DC1 which is generalized for $n$-tuples, $n\geq 2$, in [10, 17]. Throughout, $X$ denotes a compact metric space endowed with a metric $d$.

Definition 1.1.

Given a continuous map $f\colon X\to X$, an $n$-tuple $(x_{1},x_{2},\dots,x_{n})\in X^{n}$, $n\geq 2$, is said to be DC1-$n$-$\delta$-scrambled for $\delta>0$ if

$$\limsup_{m\to\infty}\frac{1}{m}|\{0\leq i\leq m-1\colon\min_{1\leq j<k\leq n}d(f^{i}(x_{j}),f^{i}(x_{k}))>\delta\}|=1,$$

and

$$\limsup_{m\to\infty}\frac{1}{m}|\{0\leq i\leq m-1\colon\max_{1\leq j<k\leq n}d(f^{i}(x_{j}),f^{i}(x_{k}))<\epsilon\}|=1$$

for all $\epsilon>0$. Let ${\rm DC1}_{n}^{\delta}(X,f)$ denote the set of DC1-$n$-$\delta$-scrambled $n$-tuples and let

$${\rm DC1}_{n}(X,f)=\bigcup_{\delta>0}{\rm DC1}_{n}^{\delta}(X,f).$$

A subset $S$ of $X$ is said to be DC1-$n$-scrambled (resp. DC1-$n$-$\delta$-scrambled) if

$$(x_{1},x_{2},\dots,x_{n})\in{\rm DC1}_{n}(X,f)\>\text{(resp.\>${\rm DC1}_{n}^{\delta}(X,f)$)}$$

for all distinct $x_{1},x_{2},\dots,x_{n}\in S$. We say that $f$ exhibits the distributional $n$-chaos of type I (${\rm DC1}_{n}$) if there is an uncountable DC1-$n$-scrambled subset of $X$.

We recall a simplified version of a theorem of Mycielski [14]. Mycielski’s theorem is used in [4] to prove that positive topological entropy implies Li–Yorke chaos. A comprehensive treatment of Mycielski’s theorem and some of its applications to topological dynamics are given in [1].

For a topological space $Z$, a subset $S$ of $Z$ is said to be residual if $S$ contains a countable intersection of dense open subsets of $Z$. By Baire category theorem, if $Z$ is a complete metric space, then every countable intersection of residual subsets of $Z$ is dense in $Z$. A topological space $Z$ is said to be perfect if $Z$ has no isolated point. For a complete metric space $Z$, a subset $S$ of $Z$ is said to be a Mycielski set if $S$ is a union of countably many Cantor sets (see [4]). Note that for a Mycielski set $S$ in $Z$ and an open subset $U$ of $Z$ with $S\cap U\neq\emptyset$, $S\cap U$ is an uncountable set.

Theorem (Mycielski).

Let $Z$ be a perfect complete separable metric space. If $R_{n}$ is a residual subset of $Z^{n}$ for each $n\geq 2$, then there is a Mycielski set $S$ which is dense in $Z$ and satisfies $(x_{1},x_{2},\dots,x_{n})\in R_{n}$ for all $n\geq 2$ and distinct $x_{1},x_{2},\dots,x_{n}\in S$.

Remark 1.1.

Let $f\colon X\to X$ be a continuous map. By the above theorem, for a sequence $(\delta_{n})_{2\leq n<b+1}$ of positive numbers, where $2\leq b\leq\infty$, if

$${\rm DC1}_{n}^{\delta_{n}}(X,f)$$

is a residual subset of $X^{n}$ for each $2\leq n<b+1$, then there is a dense Mycielski set $S$ in $X$ such that $S$ is DC1-$n$-$\delta_{n}$-scrambled for all $2\leq n<b+1$, in particular, $f$ exhibits ${\rm DC1}_{n}$ for all $2\leq n<b+1$.

Let $M$ be a closed differentiable manifold endowed with a Riemannian metric $d$ and let $F\colon\mathbb{R}\times M\to M$ be a mixing Anosov flow. Anosov means that $M$ is a hyperbolic set for $F$ (see [6] for details and background information). Let $F^{t}(x)=F(t,x)$ for all $(t,x)\in\mathbb{R}\times M$ and let $g=F^{1}\colon M\to M$, the time-one map for $F$. Let

$$W^{ss}(x)=\{y\in M\colon\lim_{t\to\infty}d(F^{t}(x),F^{t}(y))=0\},$$

$x\in M$, and note that

$$W^{ss}(x)=\{y\in M\colon\lim_{i\to\infty}d(g^{i}(x),g^{i}(y))=0\}$$

for all $x\in M$. By results in Chapter 6 of [6], a periodic point $p$ for $F$ satisfies $M=\overline{W^{ss}(p)}$. Let $\tau$ denote the period of $p$, i.e.,

$$\tau=\min\{t>0\colon F^{t}(p)=p\}>0,$$

and let $\Lambda$ denote the orbit of $p$:

$$\Lambda=\{F^{t}(p)\colon t\in\mathbb{R}\}=\{F^{t}(p)\colon t\in[0,\tau)\}.$$

We have

$$\overline{W^{ss}(F^{t}(p))}=\overline{F^{t}(W^{ss}(p))}=F^{t}(\overline{W^{ss}(p)})=F^{t}(M)=M$$

for all $t\in\mathbb{R}$; therefore, $M=\overline{W^{ss}(x)}$ for all $x\in\Lambda$. Let $S^{1}=\{z\in\mathbb{C}\colon|z|=1\}$ and note that

$$g|_{\Lambda}\colon\Lambda\to\Lambda$$

is topologically conjugate to the circle rotation

$$R_{\tau^{-1}}\colon S^{1}\to S^{1}$$

defined by $R_{\tau^{-1}}(z)=z\cdot e^{2\pi i\tau^{-1}}$ for all $z\in S^{1}$, i.e., there is a homeomorphism $h\colon\Lambda\to S^{1}$ such that

$$h\circ g|_{\Lambda}=R_{\tau^{-1}}\circ h.$$

If $\tau\in\mathbb{Q}$, then every $z\in S^{1}$ is a periodic point for $R_{\tau^{-1}}$ and so every $x\in\Lambda$ is a periodic point for $g|_{\Lambda}$. If $\tau\notin\mathbb{Q}$, then $R_{\tau^{-1}}$ is minimal and so is $g|_{\Lambda}$. In both cases, $R_{\tau^{-1}}$ is equicontinuous and so is $g|_{\Lambda}$.

We prove the following claim.

Claim.

There is a sequence $(\delta_{n})_{n\geq 2}$ of positive numbers such that

$${\rm DC1}_{n}^{\delta_{n}}(M,g)$$

is a residual subset of $M^{n}$ for all $n\geq 2$.

Remark 1.2.

As stated in Remark 1.1, it follows from Mycielski’s theorem and the above claim that $g$ exhibits ${\rm DC1}_{n}$ for all $n\geq 2$.

We say that a map $f\colon X\to X$ is equicontinuous if for every $\epsilon>0$, there is $\delta>0$ such that $d(x,y)\leq\delta$ implies

$$\sup_{i\geq 0}d(f^{i}(x),f^{i}(y))\leq\epsilon$$

for all $x,y\in X$. We know that if an equicontinuous map $f\colon X\to X$ is surjective, then $f$ is a homeomorphism and $f^{-1}$ is also equicontinuous (cf. [2, 13]). If $f\colon X\to X$ is an equicontinuous homeomorphism, then $f$ is distal, i.e.,

$$\inf_{i\geq 0}\min_{1\leq j<k\leq n}d(f^{i}(x_{j}),f^{i}(x_{k}))>0$$

for all $n\geq 2$ and distinct $x_{1},x_{2},\dots,x_{n}\in X$.

We recall the basic definition of minimality.

Definition 1.2.

For a continuous map $f\colon X\to X$, a subset $S$ of $X$ is said to be $f$-invariant if $f(S)\subset S$. A closed $f$-invariant subset $K$ of $X$ is called a minimal set for $f$ if closed $f$-invariant subsets of $K$ are only $\emptyset$ and $K$. This is equivalent to that $K=\overline{\{f^{i}(x)\colon i\geq 0\}}$ for all $x\in K$. We say that a continuous map $f\colon X\to X$ is minimal if $X$ is a minimal set for $f$.

Remark 1.3.

Since a minimal continuous map $f\colon X\to X$ is surjective, every minimal equicontinuous map $f\colon X\to X$ is an equicontinuous homeomorphism. We know that if $f\colon X\to X$ is an equicontinuous homeomorphism, then $X$ is a disjoint union of minimal sets for $f$. We also know that every minimal equicontinuous homeomorphism $f\colon X\to X$ is topologically conjugate to a minimal rotation

$$R_{a}\colon G\to G$$

of a compact Abelian group $(G,+)$ where $a\in G$ and $R_{a}(z)=z+a$ for all $z\in G$ (see Theorem 2.42 of [8]).

Let $f\colon X\to X$ be a continuous map. We say that a subset $S$ of $X$ is a distal set for $f$ if

$$\inf_{i\geq 0}\min_{1\leq j<k\leq n}d(f^{i}(x_{j}),f^{i}(x_{k}))>0$$

for all $n\geq 2$ and distinct $x_{1},x_{2},\dots,x_{n}\in S$. Note that a subset $S$ of $X$ with $|S|\leq 1$ is by definition a distal set for $f$. For $x\in X$ and $\epsilon>0$, we define a subset $V_{\epsilon}^{s}(x)$ of $X$ by

$$V_{\epsilon}^{s}(x)=\{y\in X\colon\limsup_{i\to\infty}d(f^{i}(x),f^{i}(y))\leq\epsilon\}.$$

In order to prove the above claim, it is sufficient to show the following lemma.

Lemma 1.1.

Let $f\colon X\to X$ be a continuous map and let $S$ be a subset of $X$. If

• $S$ is a distal set for $f$,

• $X=\overline{V_{\epsilon}^{s}(x)}$ for all $x\in S$ and $\epsilon>0$,

then for any $2\leq n<|S|+1$, ${\rm DC1}_{n}^{\delta_{n}}(X,f)$ is a residual subset of $X^{n}$ for some $\delta_{n}>0$.

Proof.

Given $2\leq n<|S|+1$, let

$$A_{\delta} =\{(x_{1},x_{2},\dots,x_{n})\in X^{n}\colon$$

$$\qquad\qquad\qquad\limsup_{m\to\infty}\frac{1}{m}|\{0\leq i\leq m-1\colon\min_{1\leq j<k\leq n}d(f^{i}(x_{j}),f^{i}(x_{k}))>\delta\}|=1\}$$

$$=\bigcap_{p\geq 1}\bigcap_{q\geq 1}\bigcup_{m\geq q}\{(x_{1},x_{2},\dots,x_{n})\in X^{n}\colon$$

$$\qquad\qquad\qquad\frac{1}{m}|\{0\leq i\leq m-1\colon\min_{1\leq j<k\leq n}d(f^{i}(x_{j}),f^{i}(x_{k}))>\delta\}|>1-\frac{1}{p}\},$$

$\delta>0$, and let

$$B =\bigcap_{\epsilon>0}\{(x_{1},x_{2},\dots,x_{n})\in X^{n}\colon$$

$$\qquad\qquad\qquad\qquad\limsup_{m\to\infty}\frac{1}{m}|\{0\leq i\leq m-1\colon\max_{1\leq j<k\leq n}d(f^{i}(x_{j}),f^{i}(x_{k}))<\epsilon\}|=1\}$$

$$=\bigcap_{l\geq 1}\bigcap_{p\geq 1}\bigcap_{q\geq 1}\bigcup_{m\geq q}\{(x_{1},x_{2},\dots,x_{n})\in X^{n}\colon$$

$$\qquad\qquad\qquad\qquad\frac{1}{m}|\{0\leq i\leq m-1\colon\max_{1\leq j<k\leq n}d(f^{i}(x_{j}),f^{i}(x_{k}))<\frac{1}{l}\}|>1-\frac{1}{p}\}.$$

Note that

$${\rm DC1}_{n}^{\delta}(X,f)=A_{\delta}\cap B$$

for all $\delta>0$. Since $S$ is a distal set for $f$, there are $(a_{1},a_{2},\dots,a_{n})\in S^{n}$ and $\delta_{n},\epsilon_{n}>0$ such that

$$\inf_{i\geq 0}\min_{1\leq j<k\leq n}d(f^{i}(a_{j}),f^{i}(a_{k}))>\delta_{n}+\epsilon_{n}.$$

Since

$$V_{\epsilon_{n}}^{s}(a_{1})\times V_{\epsilon_{n}}^{s}(a_{2})\times\cdots\times V_{\epsilon_{n}}^{s}(a_{n})$$

is a dense subset of $X^{n}$ and contained in

$$\bigcup_{m\geq q}\{(x_{1},x_{2},\dots,x_{n})\in X^{n}\colon$$

$$\qquad\qquad\qquad\frac{1}{m}|\{0\leq i\leq m-1\colon\min_{1\leq j<k\leq n}d(f^{i}(x_{j}),f^{i}(x_{k}))>\delta_{n}\}|>1-\frac{1}{p}\}$$

for all $p,q\geq 1$, we see that $A_{\delta_{n}}$ is a residual subset of $X^{n}$. By taking $a\in S$, since $V_{\frac{1}{4l}}^{s}(a)^{n}$, $l\geq 1$, is a dense subset of $X^{n}$ and contained in

$$\bigcup_{m\geq q}\{(x_{1},x_{2},\dots,x_{n})\in X^{n}\colon$$

$$\qquad\qquad\qquad\frac{1}{m}|\{0\leq i\leq m-1\colon\max_{1\leq j<k\leq n}d(f^{i}(x_{j}),f^{i}(x_{k}))<\frac{1}{l}\}|>1-\frac{1}{p}\}$$

for all $l,p,q\geq 1$, we see that $B$ is a residual subset of $X^{n}$. It follows that ${\rm DC1}_{n}^{\delta_{n}}(X,f)$ is a residual subset of $X^{n}$, thus the proof has been completed. ∎

Remark 1.4.

We should note that, in the context of shadowing, a similar argument was presented by Li, Li, and Tu in Section 3 of [9] (see, in particular, Lemma 3.2; Lemma 3.3; and Theorem 3.4 of [9]).

Remark 1.5.

For a continuous map $f\colon X\to X$, let $Per(f)$ denote the set of periodic points for $f$:

$$Per(f)=\{y\in X\colon f^{i}(y)=y\>\>\text{for some $i>0$}\}.$$

For $x\in X$, the $\omega$-limit set $\omega(x,f)$ is defined as the set of $y\in X$ such that

$$\lim_{j\to\infty}f^{i_{j}}(x)=y$$

for some sequence $0\leq i_{1}<i_{2}<\cdots$. Given any $x\in X$, $\{f^{i}(x)\colon i\geq 0\}$ is a distal set for $f$ exactly if

• $|\{f^{i}(x)\colon i\geq 0\}|<\infty$ and $x\in Per(f)$; or

• $|\{f^{i}(x)\colon i\geq 0\}|=\infty$ and $\omega(x,f)\cap Per(f)=\emptyset$.

If $f$ is minimal, then $\{f^{i}(x)\colon i\geq 0\}$ is a distal set for $f$ for all $x\in X$. We also remark that if $f$ is surjective, then for any $x\in X$ and $\epsilon>0$, $X=\overline{V_{\epsilon}^{s}(x)}$ implies $X=\overline{V_{\epsilon}^{s}(f^{i}(x))}$ for all $i\geq 0$.

By this remark, we obtain the following corollary of Lemma 1.1.

Corollary 1.1.

Let $f\colon X\to X$ be a continuous map and let $\Lambda$ be a minimal set for $f$. If

• $f$ is surjective,

• there is $x\in\Lambda$ such that $X=\overline{V_{\epsilon}^{s}(x)}$ for all $\epsilon>0$,

then for any $2\leq n<|\Lambda|+1$, ${\rm DC1}_{n}^{\delta_{n}}(X,f)$ is a residual subset of $X^{n}$ for some $\delta_{n}>0$.

We have observed that the time-one map of a mixing Anosov flow exhibits ${\rm DC1}_{n}$ for all $n\geq 2$. The proof presented above is based on the fact that the strong stable manifold of a periodic point is dense in the phase space. By relaxing this condition, we can derive sufficient conditions for ${\rm DC1}_{n}$, $n\geq 2$.

The main result of this paper is the following theorem. A continuous map $f\colon X\to X$ is said to be transitive if for any non-empty open subsets $U,V$ of $X$, it holds that $f^{i}(U)\cap V\neq\emptyset$ for some $i\geq 0$.

Theorem 1.1.

Given a continuous map $f\colon X\to X$ and a closed $f$-invariant subset $\Lambda$ of $X$, if the following conditions are satisfied

• $f|_{\Lambda}\colon\Lambda\to\Lambda$ is minimal and equicontinuous,

• $X=\overline{\bigcup_{x\in\Lambda}V_{\epsilon}^{s}(x)}$ for all $\epsilon>0$,

• $f|_{\Lambda}\times f\colon\Lambda\times X\to\Lambda\times X$ is transitive,

then for any $2\leq n<|\Lambda|+1$, ${\rm DC1}_{n}^{\delta_{n}}(X,f)$ is a residual subset of $X^{n}$ for some $\delta_{n}>0$.

Remark 1.6.

In [3], it is shown that for a non-trivial transitive continuous map $f\colon X\to X$, if there is a closed $f$-invariant subset $\Lambda$ of $X$ such that

$$f|_{\Lambda}\times f\colon\Lambda\times X\to\Lambda\times X$$

is transitive, then $f$ exhibits (dense and uniform) Li–Yorke chaos (see Theorem 3.1 of [3]).

Remark 1.7.

Let $f\colon X\to X,g\colon Y\to Y$ be continuous self-maps of compact metric spaces $X,Y$. It is known that $f\times g\colon X\times Y\to X\times Y$ is transitive if

• $f$ is totally transitive; and $Y$ is a periodic orbit or an odometer,

• $f$ is weakly scattering; and $g$ is minimal and equicontinuous,

• $f$ is scattering; and $g$ is minimal

(see, e.g., [3, 7] and also Remark 1.8 for details).

We recall the notion of Furstenberg families.

Definition 1.3.

Let $\mathbb{N}_{0}=\{0\}\cup\mathbb{N}=\{0,1,2,\dots\}$ and let $\mathcal{F}\subset 2^{\mathbb{N}_{0}}$. We say that $\mathcal{F}$ is a Furstenberg family if the following conditions are satisfied

• (hereditary upward) For any $A,B\subset\mathbb{N}_{0}$, $A\in\mathcal{F}$ and $A\subset B$ implies $B\in\mathcal{F}$,

• (proper) $\mathcal{F}\neq\emptyset$ and $\mathcal{F}\neq 2^{\mathbb{N}_{0}}$.

For a Furstenberg family $\mathcal{F}$, we define its dual family $\mathcal{F}^{\ast}$ by

$$\mathcal{F}^{\ast}=\{A\subset\mathbb{N}_{0}\colon A\cap B\neq\emptyset\>\>\text{for all $B\in\mathcal{F}$}\},$$

which is also a Furstenberg family. For $A\subset\mathbb{N}_{0}$ with $0\in A$, let $\Delta(A)$ (or $A-A$) denote the set of $i\in\mathbb{N}_{0}$ such that $i=k-j$ for some $j,k\in A$. For a Furstenberg family $\mathcal{F}$, we define a Furstenberg family $\Delta(\mathcal{F})$ by

$$\Delta(\mathcal{F})=\{B\subset\mathbb{N}_{0}\colon\text{$\Delta(A)\subset B$ for some $A\in\mathcal{F}$}\}.$$

Let $f\colon X\to X$ be a continuous map and let $\mathcal{F}$ be a Furstenberg family. For $x\in X$ and $\epsilon>0$, let

$$N(x,\epsilon)=\{i\in\mathbb{N}_{0}\colon d(x,f^{i}(x))\leq\epsilon\}.$$

We say that $x\in X$ is $\mathcal{F}$-recurrent if $N(x,\epsilon)\in\mathcal{F}$ for all $\epsilon>0$. We denote by $R(f,\mathcal{F})$ the set of $\mathcal{F}$-recurrent points for $f$. For subsets $A,B$ of $X$, let

$$N(A,B)=\{i\in\mathbb{N}_{0}\colon f^{i}(A)\cap B\neq\emptyset\}.$$

We say that $f$ is $\mathcal{F}$-transitive if $N(U,V)\in\mathcal{F}$ for all non-empty open subsets $U,V$ of $X$.

We obtain the following corollary of Theorem 1.1.

Corollary 1.2.

Given a continuous map $f\colon X\to X$, a closed $f$-invariant subset $\Lambda$ of $X$, and a Furstenberg family $\mathcal{F}$, if the following conditions are satisfied

• $f|_{\Lambda}\colon\Lambda\to\Lambda$ is minimal and equicontinuous,

• $X=\overline{\bigcup_{x\in\Lambda}V_{\epsilon}^{s}(x)}$ for all $\epsilon>0$,

• $\Lambda\cap R(f,\mathcal{F})\neq\emptyset$,

• $f$ is $\Delta(\mathcal{F})^{\ast}$-transitive,

then for any $2\leq n<|\Lambda|+1$, ${\rm DC1}_{n}^{\delta_{n}}(X,f)$ is a residual subset of $X^{n}$ for some $\delta_{n}>0$.

In fact, this corollary is a direct consequence of Theorem 1.1 and the following two lemmas.

Lemma 1.2.

Let $f\colon X\to X$ be a continuous map and let $\Lambda$ be a closed $f$-invariant subset of $X$. If $f|_{\Lambda}\colon\Lambda\to\Lambda$ is minimal and equicontinuous, then for any Furstenberg family $\mathcal{F}$, $\Lambda\cap R(f,\mathcal{F})\neq\emptyset$ implies $\Lambda\subset R(f,\mathcal{F})$.

Lemma 1.3.

Let $f\colon X\to X,g\colon Y\to Y$ be continuous self-maps of compact metric spaces $X,Y$. For a Furstenberg family $\mathcal{F}$, if

• $f$ is $\Delta(\mathcal{F})^{\ast}$-transitive,

• $g$ is transitive and satisfies $Y=R(g,\mathcal{F})$,

then $f\times g\colon X\times Y\to X\times Y$ is transitive.

In Section 2, we prove these lemmas. We should note that Lemma 1.3 can be prove by a similar argument as in the proof of Theorem 4.9 in [7], however we prove it for the sake of completeness.

Remark 1.8.

Let $f\colon X\to X$ be a continuous map and let $\Lambda$ be a closed $f$-invariant subset of $X$.

• For $k\geq 1$, let $k\mathbb{N}_{0}=\{0,k,2k,3k,\dots\}$. We define a Furstenberg family $\mathcal{F}_{\rm rr}$ by

  $$\mathcal{F}_{\rm rr}=\{A\subset\mathbb{N}_{0}\colon\text{$k\mathbb{N}_{0}\subset A$ for some $k\geq 1$}\}.$$

  Every $x\in R(f,\mathcal{F}_{\rm rr})$ is called a regularly recurrent point for $f$. If $f|_{\Lambda}$ is minimal and equicontinuous, then by Lemma 1.2, $\Lambda\cap R(f,\mathcal{F}_{\rm rr})\neq\emptyset$ implies $\Lambda\subset R(f,\mathcal{F}_{rr})$. We know that whenever $f|_{\Lambda}$ is minimal, $\Lambda\subset R(f,\mathcal{F}_{\rm rr})$ holds exactly if $\Lambda$ is a periodic orbit for $f$ (if $|\Lambda|<\infty$) or an odometer (if $|\Lambda|=\infty$) (see Corollary 2.5 of [5] in which an odometer is called an adding machine). In both cases, $f|_{\Lambda}$ is equicontinuous. It is easy to see that $f$ is $\mathcal{F}_{\rm rr}^{\ast}$-transitive if and only if $f$ is totally transitive, i.e., $f^{k}$ is transitive for all $k\geq 1$. We also see that $f$ is totally transitive if and only if

  $$f\times g\colon X\times Y\to X\times Y$$

  is transitive for every continuous self-map $g\colon Y\to Y$ such that $Y$ is a periodic orbit or an odometer.

• We define a Furstenberg family $\mathcal{F}_{\rm b}$ as the set of $A\subset\mathbb{N}_{0}$ such that there are

  • a minimal equicontinuous self-map $g\colon Y\to Y$ of a compact metric space $Y$,

  • $y\in Y$ and $\delta>0$

  such that $N(y,\delta)\subset A$. If $f|_{\Lambda}$ is minimal and equicontinuous, then $\Lambda\subset R(f,\mathcal{F}_{\rm b})$. We know that $f$ is $\Delta(\mathcal{F}_{\rm b})^{\ast}$-transitive if and only if $f$ is weakly scattering, i.e.,

  $$f\times g\colon X\times Y\to X\times Y$$

  is transitive for every minimal equicontinuous self-map $g\colon Y\to Y$ of a compact metric space $Y$ (see Theorem 4.12 of [7]).

• We define two Furstenberg families $\mathcal{F}_{\rm s}$ and $\mathcal{F}_{\rm t}$ by

  $$\mathcal{F}_{\rm s}=\{A\subset\mathbb{N}_{0}\colon\text{$\exists k\geq 1$ s.t.\>$A\cap\{i,i+1,\dots,i+k-1\}\neq\emptyset$ for $\forall i\geq 0$}\}$$

  and

  $$\mathcal{F}_{\rm t}=\{B\subset\mathbb{N}_{0}\colon\text{for $\forall j\geq 1$ $\exists i_{j}\geq 0$ s.t.\>$\{i_{j},i_{j}+1,\dots,i_{j}+j-1\}\subset B$}\}.$$

  Note that $\mathcal{F}_{\rm s}^{\ast}=\mathcal{F}_{\rm t}$. It is well-known that if $f|_{\Lambda}$ is minimal, then $\Lambda\subset R(f,\mathcal{F}_{\rm s})$. We know that

  • $f$ is $\Delta(\mathcal{F}_{\rm s})^{\ast}$-transitive if and only if $f$ is scattering, i.e.,

    $$f\times g\colon X\times Y\to X\times Y$$

    is transitive for every minimal continuous self-map $g\colon Y\to Y$ of a compact metric space $Y$ (see Theorem 4.10 of [7]),

  • $f$ is $\mathcal{F}_{\rm t}$-transitive if and only if $f$ is weakly mixing, i.e.,

    $$f\times f\colon X\times X\to X\times X$$

    is transitive (see Proposition 7.2 of [1]).

• Since

  $$\mathcal{F}_{\rm rr}\subset\Delta(\mathcal{F}_{\rm b})\subset\Delta(\mathcal{F}_{\rm s})\subset\mathcal{F}_{\rm s},$$

  we have

  $$\mathcal{F}_{\rm t}\subset\Delta(\mathcal{F}_{\rm s})^{\ast}\subset\Delta(\mathcal{F}_{\rm b})^{\ast}\subset\mathcal{F}_{rr}^{\ast},$$

  which implies that for any continuous map $f\colon X\to X$, $f$ is

  $$\text{weakly mixing $\implies$ scattering $\implies$ weakly scattering $\implies$ totally transitive}.$$

This paper consists of two sections. In the next section, we prove Theorem 1.1, Lemmas 1.2 and 1.3.

2. Proofs of Theorems 1.1, Lemmas 1.2 and 1.3

In this section, we prove Theorems 1.1, Lemmas 1.2 and 1.3. The proof of Theorem 1.1 is structured as a step-by-step proof of a series of lemmas, the meaning of each of which should be clear.

Let $f\colon X\to X$ be a continuous map and let $\Lambda$ be a closed $f$-invariant subset of $X$ such that $f|_{\Lambda}\colon\Lambda\to\Lambda$ is minimal and equicontinuous.

Lemma 2.1.

For every $\epsilon>0$, $X=\overline{\bigcup_{x\in\Lambda}V_{\epsilon}^{s}(x)}$ implies $X=\bigcup_{x\in\Lambda}\overline{V_{2\epsilon}^{s}(x)}$.

Proof.

If $X=\overline{\bigcup_{x\in\Lambda}V_{\epsilon}^{s}(x)}$, then for every $y\in X$, there are sequences $x_{j}\in\Lambda$, $y_{j}\in V_{\epsilon}^{s}(x_{j})$, $j\geq 1$, and $x\in\Lambda$ such that $\lim_{j\to\infty}x_{j}=x$ and $\lim_{j\to\infty}y_{j}=y$. Since $f|_{\Lambda}\colon\Lambda\to\Lambda$ is equicontinuous, we have $\delta>0$ such that $d(x,z)\leq\delta$ implies

$$\sup_{i\geq 0}d(f^{i}(x),f^{i}(z))\leq\epsilon$$

and so $V_{\epsilon}^{s}(z)\subset V_{2\epsilon}^{s}(x)$ for all $z\in\Lambda$. It follows that $y_{j}\in V_{\epsilon}^{s}(x_{j})\subset V_{2\epsilon}^{s}(x)$ for all sufficiently large $j\geq 1$, which implies $y\in\overline{V_{2\epsilon}^{s}(x)}$. Since $y\in X$ is arbitrary, we obtain $X=\bigcup_{x\in\Lambda}\overline{V_{2\epsilon}^{s}(x)}$, proving the lemma. ∎

Lemma 2.2.

Let $\Gamma$ be a countable dense subset of $\Lambda$. For every $\epsilon>0$, $X=\bigcup_{x\in\Lambda}\overline{V_{\epsilon}^{s}(x)}$ implies $X=\bigcup_{x\in\Gamma}\overline{V_{2\epsilon}^{s}(x)}$.

Proof.

Let $\epsilon>0$. Given any $x\in\Lambda$, since $f|_{\Lambda}\colon\Lambda\to\Lambda$ is equicontinuous, there is $\delta>0$ such that $d(x,z)\leq\delta$ implies

$$\sup_{i\geq 0}d(f^{i}(x),f^{i}(z))\leq\epsilon$$

and so $V_{\epsilon}^{s}(x)\subset V_{2\epsilon}^{s}(z)$ for all $z\in\Lambda$. Since $\Gamma$ is dense in $\Lambda$, by taking $z\in\Gamma$ with $d(x,z)\leq\delta$, we obtain $V_{\epsilon}^{s}(x)\subset V_{2\epsilon}^{s}(z)$ and so

$$\overline{V_{\epsilon}^{s}(x)}\subset\overline{V_{2\epsilon}^{s}(z)}.$$

Since $x\in\Lambda$ is arbitrary, it follows that

$$\bigcup_{x\in\Lambda}\overline{V_{\epsilon}^{s}(x)}\subset\bigcup_{z\in\Gamma}\overline{V_{2\epsilon}^{s}(z)}.$$

If $X=\bigcup_{x\in\Lambda}\overline{V_{\epsilon}^{s}(x)}$, then we obtain

$$X=\bigcup_{x\in\Lambda}\overline{V_{\epsilon}^{s}(x)}=\bigcup_{z\in\Gamma}\overline{V_{2\epsilon}^{s}(z)},$$

completing the proof. ∎

Lemma 2.3.

For any $x\in\Lambda$ and $\epsilon>0$, if ${\rm int}[\overline{V_{\epsilon}^{s}(x)}]\neq\emptyset$, and if

$$f|_{\Lambda}\times f\colon\Lambda\times X\to\Lambda\times X$$

is transitive, then $X=\overline{V_{2\epsilon}^{s}(x)}$.

Proof.

Let $V$ be a non-empty open subset of $X$. Since $x\in\Lambda$ and $f|_{\Lambda}$ is equicontinuous, there is $\delta>0$ such that $d(x,b)\leq\delta$ implies

$$\sup_{i\geq 0}d(f^{i}(x),f^{i}(b))\leq\epsilon$$

and so $V_{\epsilon}^{s}(b)\subset V_{2\epsilon}^{s}(x)$ for all $b\in\Lambda$. Again since $x\in\Lambda$ and $f|_{\Lambda}$ is equicontinuous, there is $\gamma>0$ such that $d(x,a)\leq\gamma$ implies

$$\sup_{i\geq 0}d(f^{i}(x),f^{i}(a))\leq\delta/2$$

for all $a\in\Lambda$. Since ${\rm int}[\overline{V_{\epsilon}^{s}(x)}]\neq\emptyset$ and $f|_{\Lambda}\times f$ is transitive, there are $z\in\Lambda$, $p\in X$, and $i\geq 0$ such that

• $d(x,z)\leq\gamma$ and $d(x,f^{i}(z))\leq\delta/2$,

• $p\in{\rm int}[\overline{V_{\epsilon}^{s}(x)}]$ and $f^{i}(p)\in V$.

By $f^{i}(x)\in\Lambda$ and

$$d(x,f^{i}(x))\leq d(x,f^{i}(z))+d(f^{i}(x),f^{i}(z))\leq\delta/2+\delta/2=\delta,$$

we obtain $V_{\epsilon}^{s}(f^{i}(x))\subset V^{s}_{2\epsilon}(x)$. Since

$$p\in{\rm int}[\overline{V_{\epsilon}^{s}(x)}]\subset\overline{V_{\epsilon}^{s}(x)}$$

we have $f^{i}(q)\in V$ for some $q\in V_{\epsilon}^{s}(x)$. It follows that

$$f^{i}(q)\in f^{i}(V_{\epsilon}^{s}(x))\subset V_{\epsilon}^{s}(f^{i}(x))\subset V_{2\epsilon}^{s}(x)$$

and so

$$f^{i}(q)\in V\cap V_{2\epsilon}^{s}(x).$$

Since $V$ is arbitrary, we obtain $X=\overline{V_{2\epsilon}^{s}(x)}$, proving the lemma. ∎

Lemma 2.4.

Let $x,y\in\Lambda$ and $\epsilon>0$. If $X=\overline{V_{\epsilon}^{s}(x)}$, and if $f$ is surjective, then $X=\overline{V_{2\epsilon}^{s}(y)}$.

Proof.

Let $V$ be a non-empty open subset of $X$. Since $y\in\Lambda$ and $f|_{\Lambda}$ is equicontinuous, there is $\delta>0$ such that $d(y,z)\leq\delta$ implies

$$\sup_{i\geq 0}d(f^{i}(y),f^{i}(z))\leq\epsilon$$

and so $V_{\epsilon}^{s}(z)\subset V_{2\epsilon}^{s}(y)$ for all $z\in\Lambda$. Since $x\in\Lambda$ and $f|_{\Lambda}$ is minimal, we have

$$\Lambda=\overline{\{f^{i}(x)\colon i\geq 0\}},$$

which implies $d(y,f^{i}(x))\leq\delta$ and so $V_{\epsilon}^{s}(f^{i}(x))\subset V_{2\epsilon}^{s}(y)$ for some $i\geq 0$. Since $f$ is surjective, we have $f^{i}(p)\in V$ for some $p\in X$. By $X=\overline{V_{\epsilon}^{s}(x)}$, we obtain $f^{i}(q)\in V$ for some $q\in V_{\epsilon}^{s}(x)$. It follows that

$$f^{i}(q)\in f^{i}(V_{\epsilon}^{s}(x))\subset V_{\epsilon}^{s}(f^{i}(x))\subset V_{2\epsilon}^{s}(y);$$

therefore,

$$f^{i}(q)\in V\cap V_{2\epsilon}^{s}(y).$$

Since $V$ is arbitrary, we obtain $X=\overline{V_{2\epsilon}^{s}(y)}$, completing the proof. ∎

Lemma 2.5.

If $X=\overline{\bigcup_{x\in\Lambda}V_{\epsilon}^{s}(x)}$ for all $\epsilon>0$, and if

$$f|_{\Lambda}\times f\colon\Lambda\times X\to\Lambda\times X$$

is transitive, then $X=\overline{V_{\epsilon}^{s}(x)}$ for all $x\in\Lambda$ and $\epsilon>0$.

Proof.

Given any $\epsilon>0$, by Lemma 2.1, we have $X=\bigcup_{x\in\Lambda}\overline{V_{2\epsilon}^{s}(x)}$. By Lemma 2.2, taking a countable dense subset $\Gamma$ of $\Lambda$, we have $X=\bigcup_{x\in\Gamma}\overline{V_{4\epsilon}^{s}(x)}$. By Baire category theorem, we obtain

$${\rm int}[\overline{V_{4\epsilon}^{s}(p)}]\neq\emptyset$$

for some $p\in\Gamma$. Since

$$f|_{\Lambda}\times f\colon\Lambda\times X\to\Lambda\times X$$

is transitive, Lemma 2.3 implies $X=\overline{V_{8\epsilon}^{s}(p)}$. Note that $f$ is surjective because $f|_{\Lambda}\times f$ and so $f$ is transitive. By Lemma 2.4, we obtain $X=\overline{V_{16\epsilon}^{s}(x)}$ for all $x\in\Lambda$. Since $\epsilon>0$ is arbitrary, we obtain $X=\overline{V_{\epsilon}^{s}(x)}$ for all $x\in\Lambda$ and $\epsilon>0$, thus the lemma has been proved. ∎

Let us prove Theorem 1.1.

Proof of Theorem 1.1.

If a continuous map $f\colon X\to X$ and a closed $f$-invariant subset $\Lambda$ of $X$ satisfy conditions (1)–(3) in Theorem 1.1, then by Lemma 2.5, we have $X=\overline{V_{\epsilon}^{s}(x)}$ for all $x\in\Lambda$ and $\epsilon>0$. From Lemma 1.1, it follows that for any $2\leq n<|\Lambda|+1$, ${\rm DC1}_{n}^{\delta_{n}}(X,f)$ is a residual subset of $X^{n}$ for some $\delta_{n}>0$. This completes the proof of Theorem 1.1. ∎

Finally, we prove Lemmas 1.2 and 1.3.

Proof of Lemma 1.2.

Given any $q\in\Lambda$ and $\epsilon>0$, since $f|_{\Lambda}\colon\Lambda\to\Lambda$ is equicontinuous, there is $\delta>0$ such that $d(q,y)\leq\delta$ implies

$$\sup_{j\geq 0}d(f^{j}(q),f^{j}(y))\leq\epsilon/3$$

for all $y\in\Lambda$. Note that $f(x)\in R(f,\mathcal{F})$ for all $x\in R(f,\mathcal{F})$. By taking $p\in\Lambda\cap R(f,\mathcal{F})$, we obtain $f^{i}(p)\in R(f,\mathcal{F})$ for all $i\geq 0$. Since $f|_{\Lambda}\colon\Lambda\to\Lambda$ is minimal, $p$ satisfies

$$\Lambda=\overline{\{f^{i}(p)\colon i\geq 0\}}$$

and so $d(q,f^{i}(p))\leq\delta$ for some $i\geq 0$. By $f^{i}(p)\in R(f,\mathcal{F})$, we obtain

$$N(f^{i}(p),\epsilon/3)\in\mathcal{F}.$$

Since $f^{i}(p)\in\Lambda$ and $d(q,f^{i}(p))\leq\delta$, we have

$$d(q,f^{j}(q))\leq d(q,f^{i}(p))+d(f^{i}(p),f^{j}(f^{i}(p)))+d(f^{j}(f^{i}(p)),f^{j}(q))\leq\epsilon/3+\epsilon/3+\epsilon/3=\epsilon$$

for all $j\in N(f^{i}(p),\epsilon/3)$. It follows that

$$N(f^{i}(p),\epsilon/3)\subset N(q,\epsilon)$$

and so

$$N(q,\epsilon)\in\mathcal{F}.$$

Since $q\in\Lambda$ and $\epsilon>0$ are arbitrary, we obtain $\Lambda\subset R(f,\mathcal{F})$, completing the proof. ∎

Proof of Lemma 1.3.

For $A\subset\mathbb{N}_{0}$ and $j\geq 0$, let $A+j=\{i+j\colon i\in A\}$. Let $U_{1},V_{1}$ be non-empty open subsets of $Y$. Since $g$ is transitive, there are $y\in U_{1}$ and $j\geq 0$ such that $g^{j}(y)\in V_{1}$. We take $\delta>0$ such that $B_{\delta}(y)\in U_{1}$ and $g^{j}(B_{\delta}(y))\subset V_{1}$ where $B_{\delta}(y)$ is the closed $\delta$-ball centered at $y$. It follows that

$$N(B_{\delta}(y),B_{\delta}(y))+j\subset N(U_{1},V_{1})$$

because for every $i\in N(B_{\delta}(y),B_{\delta}(y))$, there is $p\in B_{\delta}(y)$ such that $g^{i}(p)\in B_{\delta}(y)$ which implies $p\in U_{1}$ and $g^{i+j}(p)=g^{j}(g^{i}(p))\in g^{j}(B_{\delta}(y))\subset V_{1}$; therefore, $i+j\in N(U_{1},V_{1})$. Note that $0\in N(y,\delta)$. We have

$$\Delta(N(y,\delta))\subset N(B_{\delta}(y),B_{\delta}(y))$$

because it holds for any $k,l\in N(y,\delta)$ with $k\leq l$, $g^{k}(y)\in B_{\delta}(y)$ and $g^{l-k}(g^{k}(y))=g^{l}(y)\in B_{\delta}(y)$; therefore, $l-k\in N(B_{\delta}(y),B_{\delta}(y))$. Let $U,V$ be non-empty open subsets of $X$. Since $y\in Y$ and $Y=R(g,\mathcal{F})$, we have $N(y,\delta)\in\mathcal{F}$ and so $\Delta(N(y,\delta))\in\Delta(\mathcal{F})$. Since $f$ is $\Delta(\mathcal{F})^{\ast}$-transitive, we obtain

$$N(U,f^{-j}(V))\cap\Delta(N(y,\delta))\neq\emptyset$$

and so

$$N(U,f^{-j}(V))\cap N(B_{\delta}(y),B_{\delta}(y))\neq\emptyset,$$

implying

$$[N(U,f^{-j}(V))+j]\cap[N(B_{\delta}(y),B_{\delta}(y))+j]\neq\emptyset$$

and so

$$[N(U,f^{-j}(V))+j]\cap N(U_{1},V_{1})\neq\emptyset.$$

We have

$$N(U,f^{-j}(V))+j\subset N(U,V)$$

because for every $i\in N(U,f^{-j}(V))$, there is $x\in U$ such that $f^{i}(x)\in f^{-j}(V)$ and so $f^{i+j}(x)=f^{j}(f^{i}(x))\in V$; therefore $i+j\in N(U,V)$. It follows that

$$N(U,V)\cap N(U_{1},V_{1})\neq\emptyset.$$

Since $U_{1},V_{1},U,V$ are arbitrary, we conclude that

$$f\times g\colon X\times Y\to X\times Y$$

is transitive, proving the lemma. ∎