Normal numbers with given limits of multiple ergodic averages
Extracted source text ยท arXiv 1904.12143
\documentclass[a4paper,11pt]{article}
\usepackage{amsfonts,amsthm,amsmath,amssymb,graphicx}
%\usepackage[polish]{babel}
\usepackage{color}
\title{Normal sequences with given limits of multiple ergodic averages}
\date{}
\author{Lingmin Liao\\
\normalsize LAMA UMR 8050, CNRS, Universit\'e Paris-Est Cr\'eteil,\\
\normalsize 61 Avenue du G\'en\'eral de Gaulle, 94010 Cr\'eteil Cedex, France\\
\normalsize e-mail: lingmin.liao@u-pec.fr\\
Micha\l \ Rams \\%\thanks{supported by MNiSW grant N201 607640 (Poland). This paper was written during the visit of M.R. in Universit\'e Paris-Est Cr\'eteil.}\\
\normalsize Institute of Mathematics, Polish Academy of Sciences\\
\normalsize ul. \'Sniadeckich 8, 00-656 Warszawa, Poland\\
\normalsize e-mail: rams@impan.pl}
\theoremstyle{plain}
\newtheorem{lem}{Lemma}[section]
\newtheorem{prop}[lem]{Proposition}
\newtheorem{thm}[lem]{Theorem}
\newtheorem{cor}[lem]{Corollary}
\theoremstyle{definition}
\newtheorem{defn}[lem]{Definition}
\newtheorem{ex}[lem]{Example}
\theoremstyle{remark}
\newtheorem*{rem}{Remark}
\newtheorem*{notat}{Notation}
\numberwithin{equation}{section}
\DeclareMathOperator{\cl}{cl}
\DeclareMathOperator{\diam}{diam}
\DeclareMathOperator{\vol}{vol}
\DeclareMathOperator{\dist}{dist}
\DeclareMathOperator{\var}{var}
\newcommand{\C}{\mathbb C}
\newcommand{\clC}{\widehat \C}
\newcommand{\R}{\mathbb R}
\newcommand{\Z}{\mathbb Z}
\newcommand{\N}{\mathbb N}
\newcommand{\bd}{\partial}
\renewcommand{\epsilon}{\varepsilon}
\begin{document}
\maketitle
\def\thefootnote{}
\footnote{2010 {\it Mathematics
Subject Classification}: Primary 28A80, Secondary 11K16, 37B40}
\def\thefootnote{\arabic{footnote}}
\setcounter{footnote}{0}
\begin{abstract}
We are interested in the set of normal sequences in the space $\{0,1\}^\N$ with a given frequency of the pattern $11$ in the positions $k, 2k$. The topological entropy of such sets is determined. %As a special example, the set of nomad numbers to the base $2$ with the frequency of the pattern $11$ in the positions $k, 2k$ of their dyadic expansions.
%We also prove that the topological entropy of the set of normal sequences with equal digits in the positions $k, 2k$ is ${1 \over 2}\log 2$.
\end{abstract}
\section{Introduction and statement of results}
%Initiated by the paper of Fan Liao and Ma \cite{FLM}, the study of the multiple ergodic average from a point view of multifractal analysis have attracted much attention. The major achievements have been made by Fan, Kenyon, Peres, Schmeling, Seuret, Solomyak, Wu and et al. (\cite{KPS11, FSW11, KPS12, PS, PeSo12, FSW12, FanSW12, Kifer13 PSSS}). For a short history, we refer the readers to the paper of Peres and Solomyak \cite{PeSo12}.
Recently, Fan, Liao, Ma \cite{FLM}, and Kifer \cite{K12} proposed to calculate the topological entropy spectrum of level sets of multiple ergodic averages. Here, the topological entropy means Bowen's topological entropy (in the sense of \cite{B73}, see the definition in Section \ref{sec:pre}) which can be defined for any subset, not necessarily invariant or closed.
Let $\Sigma=\{0,1\}^{\mathbb{N}}$. Among other questions, Fan, Liao, Ma \cite{FLM} asked for the topological entropy of
\[
A_{\alpha}:=\Big\{(\omega_k)_1^{\infty}\in \Sigma : \lim_{n\to\infty} \frac{1}{n}\sum_{k=1}^{n} \omega_k\omega_{2k}=\alpha \Big\} \qquad (\alpha\in [0,1]).
\]
As a first step to solve the question, they also suggested to study a subset of $A_0$:
\[ A:=\Big\{(\omega_k)_1^{\infty}\in \Sigma : \omega_k\omega_{2k}=0 \quad \text{for all } k\geq 1\Big\}.\]
The topological entropy of $A$ was later given by Kenyon, Peres and Solomyak \cite{KPS12}.
\begin{thm}[Kenyon-Peres-Solomyak]\label{thm:KPS}
We have
\[
h_{\rm top} (A)=-\log (1-p)=0.562399..., \]
where $p\in [0,1]$ is the unique solution of $$p^2=(1-p)^3.$$
\end{thm}
Enlightened by the idea of \cite{KPS12}, the question about the topological entropy of $A_\alpha$ was finally answered by Peres and Solomyak \cite{PeSo12}, and then in higher generality by Fan, Schmeling and Wu \cite{FSW16}.
\begin{thm}[Peres-Solomyak, Fan-Schmeling-Wu]\label{thm:PS}
For any $\alpha\in[0,1]$, we have
\[
h_{\rm top}( A_{\alpha})=-\log (1-p)- \frac{\alpha}{2} \log \frac{q(1-p)} {p(1-q)},
\] where $(p,q)\in [0,1]^2$ is the unique solution of the system % \[p^2(1-q)=(1-p)^3, \qquad \alpha=\frac {2pq} {2+p-q}.\]
\[
\left\{
\begin{array}{ll}
\displaystyle p^2(1-q)=(1-p)^3,\\
\displaystyle {2pq} =\alpha({2+p-q}).
\end{array}\right.
\]
In particular, $h_{\rm top}( A_{0})= h_{\rm top}( A)$.
\end{thm}
Another, interesting, related set is
\[
B:=\Big\{(\omega_k)_1^{\infty}\in \Sigma : \omega_k=\omega_{2k} \quad \text{for all } k\geq 1\Big\}.
\]
%It is well-known that each number in $[0,1]$ can be written as a dyadic expansion
%\[
%x={w_1\over 2}+{w_2 \over 2^2}+\cdots+ {w_n \over 2^n}+\cdots,
%\]
%with $w_j=w_j(x)\in \{0,1\}$ for all $j\geq 1$.
%We also know that except for the dyadic rational numbers who have exactly two dyadic expansions, the dyadic expansion is unique.
A sequence $ { \omega} \in \{0,1\}^\N$ is said to be simple normal if the frequency of the digit $0$ in the sequence is $1/2$. It is said to be normal if for all $n\in \mathbb{N}$, each word in { $\{0,1\}^{n}$} has frequency $1/2^n$. We denote the set of normal sequences by $\mathcal{N}$.
We are interested in the intersection of $\mathcal{N}$ with the set $A_{\alpha}$ of given frequency of the pattern $11$ in $\omega_k\omega_{2k}$. For the usual ergodic (Birkhoff) averages the normal sequences all belong to one set in the multifractal decomposition -- the situation for multiple ergodic averages turns out to be very different.
Our results are as follows:
\begin{thm} \label{thm:norm}
For $\alpha \leq 1/2$ we have
\[
h_{\rm top} (\mathcal{N} \cap A_{\alpha}) = {1 \over 2} \log 2+ {1 \over 2} H(2\alpha),
\]
where $H(t)=-t\log t -(1-t)\log (1-t)$. For $\alpha>1/2$ the set $\mathcal{N} \cap A_{\alpha}$ is empty.
Further,
\[ h_{\rm top} (\mathcal{N} \cap A)=h_{\rm top} (\mathcal{N} \cap A_{0}) ={1\over 2} \log 2.\]
Moreover, ${\mathcal N}\cap B \subset A_{1/2}$ and
\[
h_{\rm top} (\mathcal{N} \cap B)=h_{\rm top} (\mathcal{N} \cap A_{1/2}) = h_{\rm top} (B) ={1\over 2} \log 2.
\]
\end{thm}
The last statement of Theorem \ref{thm:norm} was recently proved, in higher generality, by Aistleitner, Becher and Carton \cite{ABC}.
\smallskip
Let us now define the set of sequences with prescribed frequency of $0$'s and $1$'s:
\[
E_\theta:=\{ { \omega} \in \Sigma: \lim_{n\to\infty} {\omega_1({ \omega})+\cdots +\omega_n({\omega}) \over n}=\theta\}.
\]
In particular, $E_{1/2}$ is the set of simple normal sequences.
\begin{thm} \label{thm:freq}
We have
\[
h_{\rm top} (E_\theta \cap A_{\alpha})=\big(1-{\theta \over 2}\big)H\Big({2\theta- \alpha \over 2-\theta}\Big) +{\theta \over 2}H\Big({\theta- \alpha \over \theta}\Big)
\]
for $\alpha \leq \theta \leq (2+\alpha)/3$, otherwise $E_\theta \cap A_{\alpha}=\emptyset$.
Further,
\[ h_{\rm top} ( E_\theta \cap A) = h_{\rm top} ( E_\theta \cap A_0)={2-\theta\over 2} H\Big({2\theta \over 2-\theta}\Big).\]
\end{thm}
Note that
\[ h_{\rm top} (E_{1/2} \cap A)={3\over 4} H\Big({2 \over 3}\Big)>h_{\rm top} (\mathcal{N} \cap A).\]
%By \cite[Theorem 1]{PeSo12}, the topological entropy of $A_\alpha$ is attained by the measure $\mu_{p,q}$ when $p,q$ satisfy the equation
%\[
%p^2(1-q)=(1-p)^3.
%\]
%Hence if our $p=(2\theta-\alpha)/(2-\theta)$ and $q=\alpha / \theta$ satisfy the equation, i.e.,
Applying the results of \cite{PeSo12}, we have the following corollary.
\begin{cor}\label{cor}
The equality
\[
h_{\rm top} (E_\theta \cap A_{\alpha})=h_{\rm top} ( A_{\alpha})
\]
holds if and only if $\alpha, \theta$ satisfy the relation
\begin{align}\label{eq:max}
(2\theta-\alpha)^2(\theta-\alpha)(2-\theta)=\theta(2-3\theta+\alpha)^3.
\end{align}
In particular, when
\[\theta={2\over 3}\left(1+\left({2 \over 23}\right)^{2/3} \sqrt[3]{3\sqrt{69}-23}- \left({2 \over 23}\right)^{2/3} \sqrt[3]{3\sqrt{69}+23}\right)=0.354...,\]
i.e., the unique real solution of the equation $4 \theta^2 (2 - \theta) =(2 - 3 \theta)^3$, % or $23 x^3 - 46 x^2 + 36 x - 8 = 0$.
we have
\[\dim_H E_\theta \cap A= \dim_H A.
\]
\end{cor}
We organize our paper as follows. In Section \ref{sec:pre}, we give some preliminaries. Section \ref{sec:3} is devoted to the proof of Theorem \ref{thm:norm}. The proofs of Theorem \ref{thm:freq} and Corollary \ref{cor} are given in Section \ref{sec:4}.
\bigskip
\section{Preliminary}\label{sec:pre}
\subsection{Bowen's topological entropy}
In 1973, Bowen \cite{B73} introduced a definition of topological entropy for any subset which is not necessarily invariant or closed. %for a topological dynamical system.
Though the original definition of Bowen's topological entropy is for any topological dynamical systems, we recall, for simplicity, the definition of Bowen's entropy for a topological dynamical system $(X,T)$ equipped with a metric $d$. For $x\in X$, $n\in \mathbb{N}$, $n\geq 1$, denote by $B_n(x, \epsilon)$ the Bowen ball defined by
\[
B_n(x,\epsilon):=\left\{y\in X: d(T^kx, T^ky)<\epsilon, \ \forall k=0, \dots, n-1\right\}.
\]
For $E\subset X$, $s\geq 0$, $N\geq 1$ and $\epsilon>0$, set
$$
\mathcal{H}^s_N (E, \epsilon) = \inf \sum_i \exp(-s n_i),
$$
where the infimum is taken over all finite or countable families $\{B_{n_i}(x_i, \epsilon)\}$ such that $x_i\in X, n_i\geq N$ and $E \subset \bigcup_i B_{n_i}(x_i,\epsilon)$.
The quantity $ \mathcal{H}^s_N (E, \epsilon)$ is non-decreasing as $N$ increases, so the following limit exists
$$
\mathcal{H}^s (E, \epsilon) = \lim_{N \to \infty} \mathcal{H}^s_N (E, \epsilon).%, \quad \mathcal{H}^s(E)=\lim_{\epsilon\to 0} \mathcal{H}^s(E, \epsilon).
$$
For the quantity $ \mathcal{H}^s (E, \epsilon)$ considered as a function of
$s$, there exists a critical value, which we denote by $h_{\rm
top} (E, \epsilon)$, such that
$$
\mathcal{H}^s (E, \epsilon) =\left\{ \begin{array} {ll} +\infty, & s <
h_{\rm top} (E, \epsilon) \\ 0 , & s> h_{\rm
top} (E, \epsilon). \end{array} \right.
$$
One can prove that the following limit exists
$$
h_{\rm top} (E) = \lim_{\epsilon \to 0} h_{\rm top} (E,
\epsilon).
$$
The quantity $h_{\rm top} (E)$ is called the {\em topological
entropy} of $E$.
We remark that for the symbolic dynamical system $(\Sigma, \sigma)$ where the space $\Sigma=\{0,1\}^{\mathbb{N}}$ is equipped with the usual metric defined by
\[
\forall \omega, \tau \in \Sigma, \quad d(\omega, \tau):=2^{-\min\{n\geq 0: \ \omega_{n+1} \neq \tau_{n+1}\}},
\]
and $\sigma$ is the left shift defined by
\[
\sigma : \omega_1\omega_2 \dots \mapsto \omega_2\omega_3 \dots,
\]
the Bowen ball $B_n(\omega, \epsilon)$ ($\epsilon<1$) is nothing but the cylinder $C_n(\omega_1, \cdots, \omega_n)$ of order $n$ defined by
\[
C_n(\omega_1, \cdots, \omega_n):=\{\tau\in \Sigma : \tau_1=\omega_1, \cdots, \tau_n=\omega_n\},
\]
and Bowen's topological entropy $h_{\rm top} (E) $ and Hausdorff dimension $\dim_H(E)$ of a subset $E\subset X$ are different only with a constant:
\[
h_{\rm top} (E) = (\log 2)\cdot \dim_H(E).
\]
We refer to Falconer's book \cite{F90} for the details on Hausdorff dimension.
%Let $\mu $ be a measure supported on $X$.
%Similarly to the Hausdorff dimension of $\mu$, we define the topological entropy $h_\mu$ of $\mu$ by
%\[
%h_\mu:=\inf \{h_{\rm top} (E) : E \subset X, \ \mu(E)=1\}.
%\]
\subsection{Billingsley's lemma for Bowen's entropy}
The Mass Distribution Principle (\cite[Principle 4.2]{F90}), or more generally, Billingsley's lemma (\cite{Bi61}) for the Hausdorff dimension has the following topological entropy version (\cite{MW08}).
Let $\mu$ be a Borel probability measure on $X$. The local entropy $h_{\mu}(x)$ at a point $x\in X$ is defined as
\[
h_{\mu}(x)=\lim_{\epsilon \to 0}\liminf_{n\to\infty} {- \log \mu (B_n(x,\epsilon)) \over n}.
\]
\begin{thm}[Ma-Wen 2008]\label{lem:Bi}
Let $\mu$ be a Borel probability measure on $X$, $E\subset X$ be a Borel subset and $0<h<\infty$. Then
\begin{enumerate}
\item[i)] if $h_\mu(x) \leq h$ for all $x\in E$, then $h_{\rm top} (E) \leq h$,
\item[ii)] if $h_\mu(x) \geq h$ for all $x\in E$, and $\mu(E)>0$, then $h_{\rm top} (E) \geq h$.
\end{enumerate}
\end{thm}
We remark that in the symbolic dynamical system $(\Sigma, \sigma)$, the local entropy $h_\mu(\omega)$ at a point $\omega\in \Sigma$ is
\[
h_\mu(\omega)=\liminf_{n\to\infty} {-\log \mu(C_n(\omega)) \over n},
\]
where $C_n(\omega)$ is the cylinder of order $n$ containing the point $\omega$.
\subsection{A lemma of elementary analysis}
The following lemma of Peres and Solomyak (\cite[Lemma 5]{PeSo12}) will be applied several times.
\begin{lem}[Peres and Solomyak] \label{lem:PS}
Suppose that $\{z_n\}$ is a bounded real sequence and there exists $c > 0$ such that
$|z_n-z_{n+m}|\leq c{{m \over n}}$ for all $m,n \in \mathbb{N}$.
If $z_{2^kn}\rightarrow \gamma$ as $k\to \infty$ for all $n\in \mathbb{N}$, then $z_n\to \gamma$.
\end{lem}
\subsection{A family of measures on $\{0,1\}^\mathbb{N}$}\label{subsec:measures}
For the lower bound estimations of the topological entropy, the following family of measures on $\{0,1\}^\mathbb{N}$ will be used.
Let $(p_0, p_1)$ be a probability vector, i.e., $p_0, p_1\geq0$ and $p_0+p_1=1$. Let
\[
\begin{pmatrix}
p_{00} & p_{01} \\
p_{10} & p_{11}
\end{pmatrix}
\]
be a transition matrix with all coefficients $p_{ij}\geq0$ and $p_{i0}+p_{i1}=1$ for $i=0,1$.
We also assume the following condition which will guarantee our measures to be non-trivial:
\begin{align}\label{cond:nontrivial}
|p_{01}-p_{11}| \neq 2.
\end{align}
With the data $p_i, p_{ij}$ ($i,j\in \{0,1\}$), we define a Borel measure $\mu$ on the space $\Sigma= \{0,1\}^\mathbb{N}$ as follows
\begin{itemize}
\item[--] if $k$ is odd then $\omega_k=1$ with probability $p_1$,
\item[--] if $k$ is even and $\omega_{k/2}=1$ then $\omega_k=1$ with probability $p_{11}$,
\item[--] if $k$ is even and $\omega_{k/2}=0$ then $\omega_k=1$ with probability $p_{01}$,
\end{itemize}
with the events $\{\omega_k=1\}$ and $\{\omega_\ell =1\}$ independent except when $k/\ell$ is a power of 2.
More precisely,
%recalling that for $n\in\mathbb{N}$, a cylinder $C_n(\omega_1\cdots\omega_{n})$ of order $n$ is
%\[
%C_n(\omega_1, \cdots, \omega_n):=\{\omega'\in \Sigma: \omega'_1=\omega_1, \cdots, \omega'_n=\omega_n\},
%\]
we define the measure $\mu$ on any cylinder $C_n(\omega_1\cdots\omega_{n})$ of order $n\in\mathbb{N}$ by
\[
\mu(C_n(\omega_1\cdots\omega_{n}))=\prod_{k=1}^{\lceil n/2\rceil}p_{\omega_{2k-1}} \cdot \prod_{k=1}^{\lfloor n/2\rfloor} p_{\omega_{k}\omega_{2k}}.%={1 \over 2^{\lceil n/2\rceil}} \cdot \prod_{k=1}^{\lfloor n/2\rfloor} p_{\omega_{k}\omega_{2k}},
\]
where $\lceil \cdot \rceil, {\lfloor \cdot\rfloor}$ denote the ceiling function and the integer part function.
Then by Kolmogorov consistence theorem, $\mu $ is well defined on $\Sigma$. We remark that the measure $\mu$ depends on the given data $p_i, p_{ij}$ ($i,j\in \{0,1\}$). We will see late that by suitablely choosing these data, we can find suitable measures supported on the sets, using which we calculate the topological entropy.
\smallskip
For $\omega\in \Sigma$ and $n\in \mathbb{N}$, set
\[
x_n(\omega) = \frac 2 n \sum_{k=n/2+1}^n \omega_k.
\]
The following Lemmas \ref{lem:LLN}, \ref{lem:wk-w2k} and \ref{lem:h-n} implied by the strong Law of Large Numbers will be useful.
\begin{lem}\label{lem:LLN}
For $\mu$-almost all $\omega$ and for big enough $n$,
\[
x_{2n}(\omega) = {\frac 1 2}p_1 + \frac {x_n(\omega)} 2 p_{10} + \frac {1-x_n(\omega)} 2 p_{01} + o(1).
\]
\end{lem}
\begin{proof}
Recall that, by the definition of the measure $\mu$, the events $\{\omega_k=1\}$ and $\{\omega_\ell =1\}$ are independent except when $k/\ell$ is a power of $2$.
While in the average of $x_n(\omega)$ there is no different $k$'s with quotient bing a power of $2$. Thus for those $n/2+1 \leq k \leq n$, $\omega_k$ are independent. They are also identically distributed. Further, it is evident that the expectation and variance of these $\omega_k$ are finite. Thus by the classical strong Law of Large Numbers, %for $\mu_\alpha$-a.e. $\omega$, $\lim_{n\to\infty}x_n(\omega)$ exists.
\begin{align}\label{eq:x-n}
x_n(\omega)=\mathbb{E}_\mu(\omega_k)+o(1)=\mu (\omega_k=1)+o(1), \quad \mu-a.e.
\end{align}
where $n/2+1 \leq k \leq n$.
Note that for $n+1\leq k \leq 2n$, we have
\begin{align*}
&\mu (\omega_k=1)\\
=&\mathbb{P}(\{k \text{ is odd} \}) \cdot \mu( \omega_k=1| \{k \text{ is odd} \}) +\mathbb{P}(\{k \text{ is even} \}) \cdot \mu( \omega_k=1| \{k \text{ is even} \})\\
=&{1 \over 2} p_1 + {1 \over 2} \mu (\omega_m=1) \cdot \mu(\omega_k=1|k=2m, \omega_m=1)\\
&+ {1 \over 2} \mu (\omega_m=0) \cdot \mu(\omega_k=1|k=2m, \omega_m=0)\\
=&{1 \over 2} p_1 + {1 \over 2} \mu (\omega_m=1) \cdot p_{11}+ {1 \over 2} \mu (\omega_m=0) \cdot p_{01},
\end{align*}
where $m=k/2$, so $n/2+1\leq m\leq n$.
Therefore, for $\mu$-almost all $\omega$ and for big enough $n$,
\[
x_{2n}(\omega) = {\frac 1 2}p_1 + \frac {x_n(\omega)} 2 p_{11} + \frac {1-x_n(\omega)} 2 p_{01} + o(1).
\]
\end{proof}
Applying Lemma \ref{lem:PS}, we can determine the $\mu$-almost sure limit of $x_n(\omega)$.
\begin{cor}\label{cor:lim-x-n}
For $\mu$-almost all sequence $\omega$,
\[
\lim_{n\to\infty} x_n(\omega) =\xi:= {p_1+p_{01} \over 2-p_{11}+p_{01}}.
\]
\end{cor}
\begin{proof}
Note that by the condition (\ref{cond:nontrivial}), $$\left|{p_{01}-p_{11} \over 2}\right|<1.$$
Thus by Lemma \ref{lem:LLN}, $\mu$-almost surely, as $k\to \infty$,
\[
x_{2^k n}(\omega) \to {p_1+p_{01} \over 2-p_{11}+p_{01}}.
\]
By Lemma \ref{lem:PS}, this implies that $\mu$-almost surely
\begin{equation} \label{eq:freq}
\lim_{n\to\infty} x_n(\omega) = {p_1+p_{01} \over 2-p_{11}+p_{01}}.
\end{equation}
\end{proof}
\begin{prop}\label{prop:lim-wk}
For $\mu$-almost all sequence $\omega$,
\[
\lim_{n\to\infty} {1 \over n} \sum_{j=1}^n\omega_j =\xi= {p_1+p_{01} \over 2-p_{11}+p_{01}}.
\]
\end{prop}
\begin{proof}
By Corollary \ref{cor:lim-x-n}, for $\mu$-almost all sequence $\omega$, for all $n$,
\begin{align*}
\lim_{k\to\infty} {1 \over 2^kn} \sum_{j=1}^{2^kn}\omega_j=\lim_{k\to\infty} {1 \over 2^kn} \left(\sum_{i=1}^{k}2^{i-1}n \cdot x_{2^i n}(\omega) + \sum_{j=1}^n\omega_j\right)=\xi.
\end{align*}
Applying Lemma \ref{lem:PS}, we have
\begin{align*}
\lim_{n\to\infty} {1 \over n} \sum_{j=1}^n\omega_j=\xi, \qquad \mu-a.e.
\end{align*}
\end{proof}
\begin{lem}\label{lem:wk-w2k}
We have for $(i,j)\in \{0,1\}^2$, for $\mu$-almost all $\omega$ and for big enough $n$,
\[
\frac 2 n \sum_{k=n/2+1}^n \mathbf{1}_{\{\omega_k \omega_{2k}=ij\}}(\omega) = \left(\frac 2 n \sum_{k=n/2+1}^n \mathbf{1}_{\{\omega_k =i\}}(\omega)\right) \cdot p_{ij}+o(1),
\]
and in particular,
\[
\frac 2 n \sum_{k=n/2+1}^n \omega_k \omega_{2k} = x_n(\omega) \cdot p_{11}+o(1).
\]
\end{lem}
\begin{proof}
Note that for $n/2+1 \leq k \leq n$ the variables $\omega_k\omega_{2k}$ are independent and identically distributed. Thus for $(i,j)\in \{0,1\}^2$, by the classical strong Law of Large Numbers, for $\mu$-almost all $\omega\in \Sigma$,
\[
\frac 2 n \sum_{k=n/2+1}^n \mathbf{1}_{\{\omega_k \omega_{2k}=ij\}}(\omega) = \mathbb{E}_\mu(\mathbf{1}_{\{\omega_k \omega_{2k}=ij\}})=\mu(\omega_k\omega_{2k}=ij)+o(1).
\]
Note that
\[
\mu(\omega_k\omega_{2k}=ij)=\mu(\omega_k=i)\cdot \mu(\omega_k\omega_{2k}=ij|\omega_k=i)=\mu(\omega_k=i)\cdot p_{ij}.
\]
Hence, by \eqref{eq:x-n}, we complete the proof.
\end{proof}
By Lemma \ref{lem:wk-w2k} and Corollary \ref{cor:lim-x-n}, we immediately obtain the following corollary.
\begin{cor}\label{cor:wk-w2k}
We have for $j\in \{0,1\}$, for $\mu$-almost all $\omega$,
\[
\lim_{n\to\infty}\frac 2 n \sum_{k=n/2+1}^n \mathbf{1}_{\{\omega_k \omega_{2k}=0j\}}(\omega) = (1-\xi) \cdot p_{0j},
\]
and
\[
\lim_{n\to\infty}\frac 2 n \sum_{k=n/2+1}^n \mathbf{1}_{\{\omega_k \omega_{2k}=1j\}}(\omega) = \xi \cdot p_{1j},
\]
and in particular,
\[
\lim_{n\to\infty}\frac 2 n \sum_{k=n/2+1}^n \omega_k \omega_{2k} = \xi \cdot p_{11}.
\]
\end{cor}
\begin{prop}\label{prop:wk-w2k}
For $\mu$-almost all sequence $\omega$,
\[
\lim_{n\to\infty} {1 \over n} \sum_{j=1}^n\omega_j\omega_{2j} =\xi \cdot p_{11}.
\]
\end{prop}
\begin{proof}
The proof is the same as that of Proposition \ref{prop:lim-wk} by using Corollary \ref{cor:wk-w2k} and Lemma \ref{lem:PS}.
\end{proof}
%For $\omega\in \Sigma$, denote
%\[
%h_{\mu}(\omega)= \lim_{n\to\infty} {-\log \mu(C_n(\omega)) \over n },
%\]
%where $C_n(\omega)$ is the cylinder of order $n$ containing $\omega$.
\smallskip
For $n\in \mathbb{N}$ and $\omega\in \Sigma$, denote
\[
h_n(\omega) := \log \mu(C_{2n}(\omega)) - \log \mu(C_n(\omega)).
\]
\begin{lem}\label{lem:h-n}
For $\mu$-almost all sequence $\omega$ and for big enough $n$, we have
\begin{align*}
\frac 2 n h_n(\omega) = &p_0\log p_0+p_1\log p_1+ (1-x_n(\omega))( p_{00} \log p_{00}+ p_{01} \log p_{01}) \\
&+ x_n(\omega)( p_{10} \log p_{10}+ p_{11} \log p_{11})+ o(1).
\end{align*}
\end{lem}
\begin{proof}
Following \cite{PeSo12}, for positive integers $m<n$, we write $\omega_m^n$ for the word $\omega_m\omega_{m+1}\cdots \omega_n$. For $i,j\in \{0,1\}$ and $\omega\in \Sigma$, denote
\[
N_i(\omega_m^n)=\sharp \{m\leq k \leq n: \omega_k=i\},
\]
and
\[
N_{ij}(\omega_m^n)=\sharp \{m\leq k \leq n: \omega_k\omega_{2k}=ij\}.
\]
We also denote
\[
N_{i, {\rm odd}}(\omega_m^n)=\sharp \{m\leq k \leq n: k \ \text{odd}, \ \omega_k=i\}.
\]
Then we have
\[
{\mu(C_{2n}(\omega)) \over \mu(C_{n}(\omega))} = p_0^{N_{0, {\rm odd}}}p_1^{N_{1, {\rm odd}}}p_{00}^{N_{00}}p_{01}^{N_{01}}p_{10}^{N_{10}}p_{11}^{N_{11}},
\]
with $N_{i, {\rm odd}}=N_{i, {\rm odd}}(\omega_{n+1}^{2n})$, and $N_{ij}=N_{ij}(\omega_{n/2+1}^n)$.
Thus
\begin{align*}
\frac 2 n h_n(\omega) = &{N_{0, {\rm odd}} \over n/2} \log p_0+ {N_{1, {\rm odd}} \over n/2} \log p_1 \\
&+ {N_{00} \over n/2} \log p_{01} + {N_{01} \over n/2} \log p_{01} + {N_{10} \over n/2} \log p_{10} + {N_{11} \over n/2} \log p_{11} .
%=& x_n(\omega)( (1-2\alpha)\log(1-2\alpha)) +(2\alpha)\log (2\alpha))+ o(1).
\end{align*}
By the classical strong Law of Large Numbers, for $\mu$-almost all $\omega$, for large enough $n$,
\begin{align*}
{N_{0, {\rm odd}} \over n/2} = p_0 + o(1), \quad {N_{1, {\rm odd}} \over n/2} = p_1 + o(1).
\end{align*}
By Lemma \ref{lem:wk-w2k}, for $\mu$-almost all $\omega$, for large enough $n$,
\begin{align*}
{N_{00} \over n/2} =(1-x_n(\omega))\cdot p_{00} + o(1), \quad {N_{01} \over n/2} =(1-x_n(\omega))\cdot p_{01} + o(1),
\end{align*}
and
\begin{align*}
{N_{10} \over n/2} =x_n(\omega)\cdot p_{10} + o(1), \quad {N_{11} \over n/2} =x_n(\omega)\cdot p_{11} + o(1).
\end{align*}
Therefore,
\begin{align*}
\frac 2 n h_n(\omega) = &p_0\log p_0+p_1\log p_1+ (1-x_n(\omega))( p_{00} \log p_{00}+ p_{01} \log p_{01}) \\
&+ x_n(\omega)( p_{10} \log p_{10}+ p_{11} \log p_{11})+ o(1).
\end{align*}
\end{proof}
By Lemma \ref{lem:h-n} and Corollary \ref{cor:lim-x-n}, we immediately have the following corollary.
\begin{cor}\label{cor:lim-hn}
For $\mu$-almost all sequence $\omega$,
\begin{align*}
\lim_{n\to\infty}\frac 2 n h_n(\omega) = &p_0\log p_0+p_1\log p_1+ (1-\xi)( p_{00} \log p_{00}+ p_{01} \log p_{01}) \\
&+ \xi( p_{10} \log p_{10}+ p_{11} \log p_{11}).
\end{align*}
\end{cor}
We close this section by the following corollary which gives the local entropy $h_\mu(\omega)$ of the measure $\mu$ for generic sequence $\omega$.
\begin{prop}\label{prop:local-ent}
For $\mu$-almost all sequence $\omega$,
\begin{align*}
h_\mu(\omega)
=& -{1 \over 2} \Big(p_0\log p_0+p_1\log p_1+ (1-\xi)( p_{00} \log p_{00}+ p_{01} \log p_{01}) \\
&+ \xi( p_{10} \log p_{10}+ p_{11} \log p_{11})\Big).
\end{align*}
%We have for $\mu$ almost all $\omega\in \Sigma$, for all large enough $n$
%\begin{align*}
%&{-\log \mu(C_n(\omega)) \over n }\\
%=& -{1 \over 2} \Big(p_0\log p_0+p_1\log p_1+ (1-x_n(\omega))( p_{00} \log \cdot p_{00}+ p_{01} \log \cdot p_{01}) \\
%&+ x_n(\omega)( p_{10} \log \cdot p_{10}+ p_{11} \log \cdot p_{11})+ o(1)\Big).
%\end{align*}
\end{prop}
\begin{proof}
By Corollary \ref{cor:lim-hn}, we need only to show that for all large enough $n$,
\[
{\log \mu(C_n(\omega)) \over n }={h_n(\omega) \over n} +o(1).
\]
In fact, for all $k, n\in \N$
\[
{1 \over 2^kn} \log \mu(C_{2^kn}(\omega))= {1 \over 2^kn} \left(\sum_{i=1}^{k-1} h_{2^in}(\omega)+ \log \mu(C_n(\omega)) \right).
\]
Then $\mu$-almost surely, for all $n\in \N$, for large $k$,
\[
{1 \over 2^kn} \log \mu(C_{2^kn}(\omega))= {h_n(\omega) \over n}+o(1).
\]
Applying Lemma \ref{lem:PS}, we complete the proof.
\end{proof}
%The following corollary is now evident.
%\begin{cor}
%For $\mu$-almost all $\omega$,
%\begin{align*}
%h_\mu(\omega)
%=& -{1 \over 2} \Big(p_0\log p_0+p_1\log p_1+ (1-\xi)( p_{00} \log \cdot p_{00}+ p_{01} \log \cdot p_{01}) \\
%&+ \xi( p_{10} \log \cdot p_{10}+ p_{11} \log \cdot p_{11})\Big).
%\end{align*}
%\end{cor}
%Let $\mu$ be a Borel probability measure supported on $E\subset X$. Denote by $|\cdot |$ the diameter of subset. Suppose that for some $s\geq 0$, there exist $c>0, \epsilon>0$, such that
%\[
%\mu(U) \leq c |U|^s,
%\]
%for any subset $U$ with $|U|\leq \epsilon$. Then
%Let
%\[
%D_\beta=\{x: \|2^nx\|<2^{-\beta n} \ \text{for infinitely many} \ n\}.
%\]
%\begin{thm}
%We have $D_{\beta}\cap A_\alpha=\emptyset$ if $\alpha>0$.
%\[
%\dim_H D_{\beta}\cap A = {1 \over 2(\beta +1)} ?
%\]
%\[
%\dim_H D_{\beta}\cap A _\alpha \in [{1 \over (\beta +1)} \dim_H A, \quad \dim_HA].
%\]
%\end{thm}
\bigskip
\section{Proof of Theorem \ref{thm:norm}}\label{sec:3}
%The same measures that were used to calculate the entropy spectrum (see \cite{PeSo12}) will be useful for the Hausdorff spectrum as well.
We first prove the lower bound. We will need the measures defined in Subsection \ref{subsec:measures} and we will conclude by applying Billingsley's lemma for Bowen's entropy: Theorem \ref{lem:Bi}.
Given $\alpha\in [0,1]$, let $\mu_\alpha$ be the probability measure on $\Sigma$ constructed in Subsection \ref{subsec:measures}, by using the data
$(p_0, p_1):=(1/2, 1/2)$ and
\[
\begin{pmatrix}
p_{00} & p_{01} \\
p_{10} & p_{11}
\end{pmatrix}
:=\begin{pmatrix}
2\alpha & 1-2 \alpha \\
1-2 \alpha & 2 \alpha
\end{pmatrix}.
\]
Then
\[
\mu_\alpha(C_n(\omega_1\cdots\omega_{n})={1 \over 2^{\lceil n/2\rceil}} \cdot \prod_{k=1}^{\lfloor n/2\rfloor} p_{\omega_{k}\omega_{2k}}.
\]
%Let $\nu_p = \pi_* \mu_p$.
\smallskip
We will prove that the measure $\mu_\alpha$ is supported on the set $\mathcal{N}\cap A_\alpha$.
\begin{lem} \label{typical}
We have
\[
\mu_{\alpha}(\mathcal{N} \cap A_\alpha) =1.
\]
\end{lem}
\begin{proof}
%Recall
%\[
%x_n(\omega) = \frac 2 n \sum_{k=n/2+1}^n \omega_k.
%\]
%By Lemma \ref{lem:LLN}, for $\mu_\alpha$-a.e. $\omega$,
%\[
%x_{2n}(\omega) = \frac 1 4 + \frac {x_n(\omega)} 2 2\alpha + \frac {1-x_n(\omega)} 2 (1-2\alpha) + o(1).
%\]
Note that by our choice of data,
\[
\xi= {p_1+p_{01} \over 2-p_{11}+p_{01}}= {{1\over 2}+1-2\alpha \over 2-2\alpha+1-2\alpha}={1 \over 2}.
\]
%\[
%x_{2^k n}(\omega) \to \frac {1} {2}.
%\]
Hence by Proposition \ref{prop:wk-w2k}, for $\mu_{\alpha}$-almost all sequence $\omega$,
\[
\lim_{n\to\infty}\frac 1 n \sum_{k=1}^n \omega_k \omega_{2k} = \xi \cdot p_{11}={1 \over 2} \cdot 2\alpha=\alpha.
\]
Thus $\mu_\alpha(A_\alpha)=1$.
Now, we show $\mu_\alpha(\mathcal{N})=1$.
We can divide the set of natural numbers into infinitely many subsets of the form
$$A_k=\{2k-1, 4k-2, \ldots, 2^\ell (2k-1), \ldots\} \quad (k\geq 1).$$
Let ${\cal B}_k$ be the $\sigma$-field generated by the events $\{\omega_{2^\ell(2k-1)}=1\}$, $\ell\geq 0$. Observe that for the measure $\mu_\alpha$ the $\sigma$-fields ${\cal B}_k$ are independent.
Observe further that $\mu_\alpha(\omega_{2^\ell(2k-1)}=1)=1/2$ for every $k, \ell$. Indeed, for $\ell=0$, it follows from the definition of $\mu_\alpha$. For the general $\ell$, it is proved by induction:
\begin{align*}
&\mu_\alpha(\omega_{2^\ell(2k-1)}=1)\\
=& \mu_\alpha(\omega_{2^\ell(2k-1)}=1 \wedge\omega_{2^{\ell-1}(2k-1)}=1) + \mu_\alpha(\omega_{2^\ell(2k-1)}=1 \wedge \omega_{2^{\ell-1}(2k-1)}=0) \\
=& \mu_\alpha(\omega_{2^{\ell-1}(2k-1)}=1)\cdot\mu_\alpha(\omega_{2^\ell(2k-1)}=1|\omega_{2^{\ell-1}(2k-1)}=1)\\
&+ \mu_\alpha(\omega_{2^{\ell-1}(2k-1)}=0) \cdot \mu_\alpha(\omega_{2^\ell(2k-1)}=1 |\omega_{2^{\ell-1}(2k-1)}=0)\\
=& {1\over 2} \cdot 2\alpha + {1\over 2} \cdot (1-2\alpha) = {1\over 2}.
\end{align*}
Consider now, for any $n \geq 1$, the word $\omega_{m+1}\cdots\omega_{m+n}$. If $m\geq n$ then the positions $m+1,\ldots,m+n$ come all from different $A_k$'s. Thus $\omega_{m+1},\ldots, \omega_{m+n}$ are independent and each of them takes values $0,1$ with probability $1/2$ respectively. That is, the measure $\mu$ restricted to such subset of positions is $(\frac 12, \frac 12)$-Bernoulli, and for any word $\eta\in \{0,1\}^n$ with $n\leq m$, the probability that we have $\omega_{m+i}=\eta_i$ for $i=1,\ldots,n$ equals $2^{-n}$. Thus, for a given word $\eta\in \{0,1\}^n$ we can divide $\N$ into intervals $[2^j+1, 2^{j+1}]$, inside all except initial finitely many of them (with $j< \log_2 n$), for any $\mu_\alpha$-generic sequence $\omega$, the frequency of appearance of $\eta$ equals $2^{-n}+O(2^{-j/2}j\log j)$, and this means that the $\mu_\alpha$-generic sequence $\omega$ is normal.
\end{proof}
%Next, we will calculate the topological entropy $h_{\mu_\alpha}$ of the measure $\mu_\alpha$ with the help of Mass Distribution Principle (\cite[Principle 4.2]{F90}). Recall that the topological entropy of a measure is defined as the infimum of the topological entropy of the sets of full measure.
Next, we will calculate the local entropy $h_{\mu_\alpha}(\omega)$ of the measure $\mu_\alpha$ for generic sequence $\omega$.
We denote for $t\in[0,1]$,
\[
H(t) = -t\log t -(1-t)\log (1-t),
\]
with convention $H(0)=H(1)=0$.
\begin{lem} \label{entropy-mu-alpha}
We have
\[
h_{\mu_\alpha}(\omega) = \frac 1 2 \log 2+ {H(2\alpha)}, \qquad \mu_{\alpha}-a.e.
\]
\end{lem}
\begin{proof}
By Proposition \ref{prop:local-ent}, we have for $\mu_\alpha$-almost all sequence $\omega$,
\begin{align*}
h_{\mu_\alpha}(\omega)
=& -{1 \over 2} \Big(p_0\log p_0+p_1\log p_1+ (1-\xi)( p_{00} \log p_{00}+ p_{01} \log p_{01}) \\
&+ \xi( p_{10} \log p_{10}+ p_{11} \log p_{11})\Big)\\
=& -{1 \over 2} \Big({1 \over 2}\log {1 \over 2}+{1 \over 2}\log {1 \over 2}+ (1-{1 \over 2})\big( 2\alpha \log (2\alpha)+ (1-2\alpha) \log (1-2\alpha)\big) \\
&+ {1 \over 2}\big( (1-2\alpha) \log (1-2\alpha)+ 2\alpha \log (2\alpha)\big)\Big)\\
=& \frac 1 2 \log 2+ {H(2\alpha)}.
\end{align*}
% For $\omega\in \Sigma$, denote by
%$
%C_n(\omega) = \{\tau\in \Sigma: \tau_k=\omega_k \ \forall k\leq n\}
%$ the cylinder of order $n$ containing $\omega$.
%Let
%\[
%h_n(\omega) := \log \mu_{\alpha}(C_{2n}(\omega)) - \log \mu_{\alpha}(C_n(\omega)).
%\]
%By the Law of Large Numbers, for $\mu_{\alpha}$-typical $\omega$ and for big enough $n$ we have
%\begin{align*}
%\frac 2 n h_n(\omega) = &-\log 2+ (1-x_n(\omega))(2\alpha \log (2\alpha) + (1-2\alpha)\log (1-2\alpha)) \\
%&+ x_n(\omega)( (1-2\alpha)\log(1-2\alpha)) +(2\alpha)\log (2\alpha))+ o(1).
%\end{align*}
%Thus,
%\[
%\lim_{n\to\infty} -\frac 1 n h_n(\omega)= {1 \over 2} \log 2+ {1 \over 2} H(2\alpha)\qquad \mu_{\alpha}-a.e.
%\]
%Note that for all $k, n\in \N$
%\[
%{1 \over 2^kn} \log \mu_{\alpha}(C_{2^kn}(\omega))= {1 \over 2^kn} \sum_{i=1}^{k-1} h_{2^in}.
%\]
%Then for all $n\in \N$
%\[
%\lim_{k\to \infty}-{1 \over 2^kn} \log \mu_{\alpha}(C_{2^kn}(\omega))= {1 \over 2} \log 2+ {1 \over 2} H(2\alpha)\qquad \mu_{\alpha}-a.e.
%\]
%Hence, by Lemma \ref{lem:PS},
%\[
%\liminf_{n\to\infty} -\frac 1 n {\color{red} \log} \mu_{\alpha}(C_{n}(\omega))= {1 \over 2} \log 2+ {1 \over 2} H(2\alpha)\qquad \mu_{\alpha}-a.e.
%\]
%%By the Mass Distribution Principle, the left hand side of this equation is equal to the topological entropy of the measure $\mu_\alpha$.
%This completes the proof.
\end{proof}
Applying Theorem \ref{lem:Bi}, by Lemmas \ref{typical} and \ref{entropy-mu-alpha}, we immediately obtain
\[
h_{\rm top} (\mathcal{N} \cap A_{\alpha}) \geq {1 \over 2} \log 2+ {1 \over 2} H(2\alpha).
\]
To finish the proof of the lower bound we note that $A\subset A_0$ but the measure $\mu_0$ is actually supported on $A$, that the measure $\mu_{1/2}$ is supported on $B$, and that the relation ${\mathcal N}\cap B \subset A_{1/2}$ follows from
\[
\frac 1n\sharp\{n+1\leq j \leq 2n: \omega_j = \omega_{2j} =1\} =\frac 1n\sharp\{n+1\leq j \leq 2n: \omega_j =1\}\to \frac 12
\]
being satisfied for every $\omega\in {\mathcal N}\cap B$.
\medskip
For the upper bound, let us first observe that
\[
\frac 1n \sum_{k=1}^n \omega_k \omega_{2k} \leq \frac 1n \sum_{k=1}^n \omega_k
\]
and the right hand side converges to $1/2$ for every normal sequence $\omega$. Thus, the set $\mathcal{N} \cap A_{\alpha}$ is empty for all $\alpha >1/2$.
To continue with the case $\alpha \leq 1/2$, we will need the following lemma.
\begin{lem} \label{lem:subseq}
Let $\omega$ be a normal sequence and let $(n_k = \ell_1 + k \ell_2)$ be an arithmetic subsequence of $\N$. Then $\omega$ restricted to the positions $(n_k)$ is normal.
\end{lem}
\begin{proof}
The result is originally due to Wall \cite{W49}. See also Kamae \cite{K73}.
\end{proof}
Let us fix some $m>0$. For $N>m$ and $i=0,1,\ldots,m$, denote by $R(N,i)$ the set $\{2^i(2k-1), k\leq 2^{N-i-1}\}$ (for example, $R(N,0)$ is the set of odd numbers smaller than $2^N$). Further, let $$R(N,i,I):=R(N-2,i),$$ $$R(N,i,II):=R(N-1,i)\setminus R(N-2,i),$$ and $$R(N,i,III):=R(N,i)\setminus R(N-1,i).$$ Note here the following obvious relations
\begin{align*}
& 2 R(N,i,I) = R(N, i+1, I) \cup R(N,i+1, II),\\
& 2R(N,i,II) = R(N,i+1,III),\\
& 2R(N,i,III) \cap R(N, i+1) =\emptyset.
\end{align*}
We denote by $\mathcal{N}(N,m,\varepsilon)$ the set of sequences $\omega$ such that for all $n\geq N$ in each $R(n,i,\ast)$, $i=0,\ldots,m$, $\ast\in\{I, II, III\}$ the frequency of 1's is between $1/2-\varepsilon$ and $1/2+\varepsilon$. By Lemma \ref{lem:subseq},
\[
{\cal N} \subset \bigcap_{\varepsilon>0} \bigcap_{m=1}^\infty \bigcup_{N=m+1}^\infty \mathcal{N}(N,m,\varepsilon).
\]
Similarly, let us denote by $A(\alpha, N, \varepsilon)$ the set of sequences $\omega$ such that for all $n\geq N$ we have
\[
\alpha - \varepsilon <2^{-n+1} \sum_{j=1}^{2^{n-1}} \omega_j \omega_{2j} < \alpha+\varepsilon.
\]
Then
\[
A_\alpha = \bigcap_{\varepsilon>0} \bigcup_{N=1}^\infty A(\alpha, N,\varepsilon).
\]
To obtain the upper bound, we will estimate from above the number of cylinders $C_{2^N}(\omega_1,\ldots,\omega_{2^N})$ needed to cover the set $\mathcal{N}(N,m,\varepsilon)\cap A(\alpha,N,\varepsilon)$.
Let us fix $N, m, \varepsilon$. To find the cylinders $C_{2^N}(\omega_1,\ldots,\omega_{2^N})$, we should determine for each position $1\leq n \leq 2^N$, which value ($0$ or $1$), $\omega_n$ will take. To this end, it would be convenient that we partition the positions from $1$ to $2^N$ by several classes according to the values of $\omega_n$ and that of the couple $(\omega_n, \omega_{2n})$. We will introduce the following notations concerning the number of positions in each such class.
For $i=1,\ldots,m$, $k_1, k_2\in \{0,1\}$, and $\ast\in \{I, II\}$, we denote
\[
X_{k_1k_2, \ast}^i(\omega) := \sharp \{n\in R(N,i-1,\ast): \omega_n = k_1, \omega_{2n} = k_2\}.
\]
For example, $X_{01,I}^1(\omega)$ denotes the number of odd positions $n$ smaller than $2^{N-2}$ such that $\omega_n=0, \omega_{2n}=1$. Similarly, let
\[
X_{k_1,\ast}^i(\omega) := \sharp \{n\in R(N,i,\ast); \omega_n = k_1\}.
\]
The following relations are obvious: for any $i$,
\begin{align}
X_{10,I}^i + X_{11,I}^i = X_{1,I}^{i-1} \label{eq:1}
\\
X_{00,I}^i + X_{01,I}^i = X_{0,I}^{i-1} \label{eq:2}
\\
X_{10,II}^i + X_{11,II}^i = X_{1,II}^{i-1} \label{eq:3}
\\
X_{00,II}^i + X_{01,II}^i = X_{0,II}^{i-1} \label{eq:4}
\\
X_{01,I}^i + X_{11,I}^i = X_{1,I}^{i} + X_{1,II}^{i} \label{eq:5}
\\
X_{00,I}^i + X_{10,I}^i = X_{0,I}^{i} + X_{0,II}^{i} \label{eq:6}
\\
X_{01,II}^i + X_{11,II}^i = X_{1,III}^{i} \label{eq:7}
\\
X_{00,II}^i + X_{10,II}^i = X_{0,III}^{i}. \label{eq:8}
\end{align}
%\begin{align}\label{eq:1}
%X_{10,I}^i + X_{11,I}^i = X_{1,I}^{i-1}
%\end{align}
%\begin{align}\label{eq:2}
%X_{00,I}^i + X_{01,I}^i = X_{0,I}^{i-1}
%\end{align}
%\begin{align}\label{eq:3}
%X_{10,II}^i + X_{11,II}^i = X_{1,II}^{i-1}
%\end{align}
%\begin{align}\label{eq:4}
%X_{00,II}^i + X_{01,II}^i = X_{0,II}^{i-1}
%\end{align}
%\begin{align}\label{eq:5}
%X_{01,I}^i + X_{11,I}^i = X_{1,I}^{i} + X_{1,II}^{i}
%\end{align}
%\begin{align}\label{eq:6}
%X_{00,I}^i + X_{10,I}^i = X_{0,I}^{i} + X_{0,II}^{i}
%\end{align}
%\begin{align}\label{eq:7}
%X_{01,II}^i + X_{11,II}^i = X_{1,III}^{i}
%\end{align}
%\begin{align}\label{eq:8}
%X_{00,II}^i + X_{10,II}^i = X_{0,III}^{i}.
%\end{align}
%\[
%X_{10,I}^i + X_{11,I}^i = X_{1,I}^{i-1}
%\]
%\[
%X_{00,I}^i + X_{01,I}^i = X_{0,I}^{i-1}
%\]
%\[
%X_{10,II}^i + X_{11,II}^i = X_{1,II}^{i-1}
%\]
%\[
%X_{00,II}^i + X_{01,II}^i = X_{0,II}^{i-1}
%\]
%\[
%X_{01,I}^i + X_{11,I}^i = X_{1,I}^{i} + X_{1,II}^{i}
%\]
%\[
%X_{00,I}^i + X_{10,I}^i = X_{0,I}^{i} + X_{0,II}^{i}
%\]
%\[
%X_{01,II}^i + X_{11,II}^i = X_{1,III}^{i}
%\]
%\[
%X_{00,II}^i + X_{10,II}^i = X_{0,III}^{i}.
%\]
Note that for a sequence $\omega\in \mathcal{N}(N,m,\varepsilon)$ the right hand sides in all these relations are in range $2^{N-3-i}\cdot (1-\varepsilon, 1+\varepsilon)$. Thus by \eqref{eq:1}, \eqref{eq:2}, \eqref{eq:5}, \eqref{eq:6}, we have
\begin{equation} \label{eqn:nnorm}
\Big|X_{11,I}^i-X_{00,I}^i \Big| \leq \varepsilon \cdot 2^{N-1-i},
\end{equation}
and by \eqref{eq:3}, \eqref{eq:4}, \eqref{eq:7}, \eqref{eq:8}, we have %the same for $X_{11,II}^i, X_{00,II}^i$.
\begin{equation} \label{eqn:nnorm2}
\Big|X_{11,II}^i-X_{00,II}^i \Big| \leq \varepsilon \cdot 2^{N-1-i}.
\end{equation}
Note that once we know the values of the sequence
\begin{equation} \label{sequence-X}
(X_{00,I}^i, X_{11,I}^i, X_{00,II}^i, X_{11,II}^i)_{i=1}^m,
\end{equation}
then by \eqref{eq:1}-\eqref{eq:4}, the values of $(X_{0,I}^i, X_{1,I}^i, X_{0,II}^i, X_{1,II}^i)_{i=1}^{m-1}$ are also determined.
Let us start the counting of the possible $(\omega_1, \dots, \omega_{2^N})$. The idea is as follows: we will first describe the sequences that can appear in $\mathcal{N}(N,m,\varepsilon)$, starting with what can happen on the odd positions $2k-1$, after that what can happen on positions of the form $2(2k-1)$ provided that the odd positions are already decided, and so on. Finally we will go back and add another condition, for our sequences to belong to $A(\alpha,N,\varepsilon)$.
Now assume that we know the values of the sequence \eqref{sequence-X} of number of positions. We will count how many possible $(\omega_1, \dots, \omega_{2^N})$ we can have, based on the information of the values of \eqref{sequence-X}.
The values of $\{\omega_n: n\in R(N,0)\}$ can be chosen in no more than $2^{2^{N-1}}$ ways.
After we have chosen $\{\omega_n: n\in R(N,i-1)\}$, we can choose $\{\omega_n: n\in R(N,i)\}$ in no more than
\[
{X_{1,I}^{i-1} \choose X_{11,I}^i} \cdot {X_{0,I}^{i-1} \choose X_{00,I}^i} \cdot {X_{1,II}^{i-1} \choose X_{11,II}^i} \cdot {X_{0,II}^{i-1} \choose X_{00,II}^i}
\]
ways. Finally, after we have chosen $\{\omega_n: n\in R(N,i)\}$ for all $i\leq m$, we will still have $2^{N-m-1}$ positions left, which we can choose in no more than $2^{2^{N-m-1}}$ ways.
To continue our estimation, we will use the following fact: for $k,n\in \mathbb{N}, 0\leq k\leq n$,
\begin{align}\label{eq:approx}
\begin{split}
\log {n\choose k} &= n \left(- \frac kn \log \frac kn - \frac {n-k}n \log \frac {n-k}n\right) + O(\log n)\\
&= n\cdot H\big(\frac kn\big)+O(\log n),
\end{split}
\end{align}
where we recall that $H(t)=-t\log t-(1-t)\log(1-t)$ is a concave analytic function on $[0,1]$.
Note that $X_{1,I}^{i-1}$ and $X_{1,I}^{i-1}$ are in the range $2^{N-3-i}\cdot (1-\varepsilon, 1+\varepsilon)$. Applying \eqref{eq:approx}, by \eqref{eqn:nnorm}, we have
\[
\log {X_{1,I}^{i-1} \choose X_{11,I}^i} +\log {X_{0,I}^{i-1} \choose X_{00,I}^i} = 2 \log{2^{N-3-i} \choose X_{11,I}^i} + 2^{N-3-i}O(\varepsilon).
\]
Similarly, by \eqref{eqn:nnorm2}, we have
\[
\log {X_{1,II}^{i-1} \choose X_{11,II}^i} +\log {X_{0,II}^{i-1} \choose X_{00,II}^i} = 2 \log{2^{N-3-i} \choose X_{11,II}^i} + 2^{N-3-i}O(\varepsilon).
\]
%Recall that in the construction, we obtain also the sequence $$(X_{00,I}^i, X_{11,I}^i, X_{00,II}^i, X_{11,II}^i)_{i=1}^m.$$
Thus for fixed values of the sequence \eqref{sequence-X}, the logarithm of the total number $Z$ of cylinders needed to cover the corresponding part of $\mathcal{N}(N,m,\varepsilon)$ is not larger than
\begin{align*}
& \log Z\big((X_{00,I}^i, X_{11,I}^i, X_{00,II}^i, X_{11,II}^i)_{i=1}^m\big)\\
\leq & (2^{N-1}+2^{N-m-1}) \log 2 \\
& + \sum_{i=1}^m \left(2 \log{2^{N-3-i} \choose X_{11,I}^i}
+ 2 \log{2^{N-3-i} \choose X_{11,II}^i} + 2^{N-3-i}O(\varepsilon)\right).
\end{align*}
By \eqref{eq:approx}, applying the Jensen inequality for the concave function $H$, we get
\begin{align*}
&
\log Z\big((X_{00,I}^i, X_{11,I}^i, X_{00,II}^i, X_{11,II}^i)_{i=1}^m\big)\\
\leq & (2^{N-1}+2^{N-m-1}) \log 2 \\
&+ \sum_{i=1}^m 2^{N-i-1} \cdot H\left(\frac 1 {\sum_{i=1}^m 2^{N-i-1}} \cdot \sum_{i=1}^m 2^{N-i-2} \frac {X_{11,I}^i+X_{11,II}^i}{2^{N-i-3}}\right) \\
&+ \sum_{i=1}^m 2^{N-i-3} \cdot O(\varepsilon).
\end{align*}
Hence,
\begin{align}\label{estimate-logZ}
\begin{split}
&
\log Z\big((X_{00,I}^i, X_{11,I}^i, X_{00,II}^i, X_{11,II}^i)_{i=1}^m\big)\\
\leq & 2^{N-1} \log 2 + 2^{N-1} H\left(2^{-N+2} \sum_{i=1}^m (X_{11,I}^i+X_{11,II}^i)\right) \\
&+ 2^N \cdot (O(\varepsilon + 2^{-m})).
\end{split}
\end{align}
%where we use \eqref{eqn:nnorm} to describe everything in terms of $(X_{11,I}^i, X_{11,II}^i)_{i=1}^m$.
Moreover, there are no more than
\begin{align}\label{eq:nb-choice}
\prod_{i=1}^m 2^{4(N-i-3)} < 2^{4mN}=o(2^{2^N})
\end{align}
possible values of $(X_{00,I}^i, X_{11,I}^i, X_{00,II}^i, X_{11,II}^i)_{i=1}^m$. We remark here that the number of possibilities is much less, because of \eqref{eqn:nnorm} and \eqref{eqn:nnorm2}. But the estimate \eqref{eq:nb-choice} is enough for us.
%We estimate
%\[
%\log {n\choose k} \approx n \left(- \frac kn \log \frac kn - \frac {n-k}n \log \frac {n-k}n\right) = n\cdot H\big(\frac kn\big),
%\]
%and observe that $H$ is a concave function.
On the other hand, for all $\omega\in A(\alpha,N,\varepsilon)$,
\[
\left|2^{-N+1} \sum_{i=1}^m (X_{11,I}^i+X_{11,II}^i) - \alpha\right| < \varepsilon + 2^{-m}.
\]
Thus by the estimates \eqref{estimate-logZ} and \eqref{eq:nb-choice}, the logarithm of the number of cylinders $C_{2^N}(\omega_1,\ldots,\omega_{2^N})$ needed to cover the set $\mathcal{N}(N,m,\varepsilon)\cap A(\alpha,N,\varepsilon)$ is less than
\begin{align*}
&o({2^N}) +\log Z\big((X_{00,I}^i, X_{11,I}^i, X_{00,II}^i, X_{11,II}^i)_{i=1}^m\big)\\
\leq & 2^{N-1} \log 2 + 2^{N-1} \left(H(2\alpha) +O(\varepsilon + 2^{-m})\right)+ 2^N \cdot (O(\varepsilon + 2^{-m})).
\end{align*}
Dividing the above value by $2^N$, and passing with $m,N$ to infinity and with $\varepsilon$ to $0$, we finish the proof of the upper bound.
\section{Proofs of Theorem \ref{thm:freq} and Corollary \ref{cor}}\label{sec:4}
%The same measures that were used to calculate the entropy spectrum (see \cite{PeSo12}) will be useful for the Hausdorff spectrum as well.
Given $p,q\in [0,1]$, let $\mu_{p,q}$ be the probability measure on $\Sigma$ constructed in Subsection \ref{subsec:measures}, by using the data
$(p_0, p_1):=(1-p, p)$ and
\[
\begin{pmatrix}
p_{00} & p_{01} \\
p_{10} & p_{11}
\end{pmatrix}
:=\begin{pmatrix}
1-p & p \\
1-q & q
\end{pmatrix}.
\]
%Then the measure $\mu_{p,q}$ of a cylinder is given by
%\[
%\mu_{p,q}(C_n(\omega_1\cdots\omega_{n}))=\prod_{k=1}^{\lceil n/2\rceil}p_{\omega_{2k-1}} \cdot \prod_{k=1}^{\lfloor n/2\rfloor} p_{\omega_{k}\omega_{2k}}.
%\]
%
%%Let $\nu_p = \pi_* \mu_p$.
%
%%We will prove that the measure $\mu_{p,q}$ is support on the set $E_\theta\cap A_\alpha$.
%
%
%For positive integers $m<n$, write $\omega_m^n$ for the word $\omega_m\omega_{m+1}\cdots \omega_n$. For $i,j\in \{0,1\}$ and $\omega\in \Sigma$, denote
%\[
%N_i(\omega_m^n)=\sharp \{m\leq k \leq n: \omega_k=i\},
%\]
%and
%\[
%N_{ij}(\omega_m^n)=\sharp \{m\leq k \leq n: \omega_k\omega_{2k}=ij\}.
%\]
%We also denote
%\[
%N_{i, {\rm odd}}(\omega_m^n)=\sharp \{m\leq k \leq n: k \ \text{odd}, \ \omega_k=i\}.
%\]
%Then we have
%\[
%\mu_{p,q}(C_n(\omega))= (1-p)^{N_{0, {\rm odd}}}p^{N_{1, {\rm odd}}}(1-p)^{N_{00}}p^{N_{01}}(1-q)^{N_{10}}q^{N_{11}},
%\]
%with $N_{i, {\rm odd}}=N_{i, {\rm odd}}(\omega_1^n)$, and $N_{ij}=N_{ij}(\omega_1^{n/2})$.
%Thus
%\begin{align}\label{measure-p-q}
%\begin{split}
%{-\log \mu_{p,q}(C_n(\omega)) \over n }= &-{1 \over 2} \Big( {N_{0, {\rm odd}} \over n/2} \log (1-p)+ {N_{1, {\rm odd}} \over n/2} \log p \\
%&+{N_{00} \over n/2} \log (1-p) +{N_{01} \over n/2} \log p \\
%&+{N_{10} \over n/2} \log (1-q)+{N_{11} \over n/2} \log q \Big).
%\end{split}
%\end{align}
The proof of Theorem \ref{thm:freq} is based on the following lemmas.
\begin{lem} \label{typical-2}
If $p=(2\theta-\alpha)/(2-\theta)$ and $q=\alpha / \theta$, then
\[
\mu_{p,q}(E_\theta\cap A_\alpha) =1.
\]
\end{lem}
\begin{proof}
%Recall that
%\[
%x_n(\omega) ={N_1(\omega_{n/2+1}^n) \over n/2}= \frac 2 n \sum_{k=n/2+1}^n \omega_k.
%\]
%By Lemma \ref{lem:LLN}, for $\mu_{p,q}$-almost all $\omega$
%\[
%x_{2n}(\omega) = \frac p 2 + \frac {x_n(\omega)} 2 q + \frac {1-x_n(\omega)} 2 p + o(1)= p + \frac {x_n(\omega)} 2 \cdot {q-p \over 2}+o(1).
%\]
%Note that $|{q-p \over 2}|<1$.
%Then, as $k\to \infty$,
%\[
%x_{2^k n}(\omega) \to \frac {2p} {2+p-q}.
%\]
%By Lemma \ref{lem:PS}, it implies that
By Proposition \ref{prop:lim-wk}, $\mu_{p,q}$-almost surely
\begin{equation} \label{eq:freq}
\lim_{n\to\infty} {1 \over n}\sum_{k=1}^n\omega_k =\xi= {p_1+p_{01} \over 2-p_{11}+p_{01}}= \frac {2p} {2+p-q}=\theta,
\end{equation}
where the last equality comes from the choices of $p$ and $q$.
%Similarly to the proof of Corollary \ref{cor:local-ent}, we
Thus $\mu_{p,q}(E_\theta)=1$.
On the other hand, by Proposition \ref{prop:wk-w2k}, for $\mu_{p,q}$-almost all sequence $\omega$,
\[
\lim_{n\to\infty}\frac 1 n \sum_{k=1}^n \omega_k \omega_{2k} = \xi \cdot p_{11} = \theta \cdot q= \alpha.
\]
By Lemma \ref{lem:PS}, we conclude $\mu_{p,q}(A_\alpha)=1$.
\end{proof}
\begin{lem} \label{entropy-measure-2}
For $p=(2\theta-\alpha)/(2-\theta)$ and $q=\alpha / \theta$, we have
\[
h_{\mu_{p,q}}(\omega)=\big(1-{\theta \over 2}\big)H\Big({2\theta- \alpha \over 2-\theta}\Big) +{\theta \over 2}H\Big({\theta- \alpha \over \theta}\Big), \qquad \mu_{p,q}-a.e.
\]
\end{lem}
\begin{proof}
By Proposition \ref{prop:local-ent}, we have for $\mu_{p,q}$ almost all $\omega\in \Sigma$,
\begin{align*}
h_{\mu_{p,q}}(\omega)= %&\lim_{n\to\infty} {-\log \mu_{p,q}(C_n(\omega)) \over n }\\
& -{1 \over 2} \Big( (1-p) \log (1-p)+p \log p+(1-\theta)(1-p) \log (1-p) \\
&+(1-\theta)p \log p +\theta(1-q) \log (1-q)+\theta q \log q \Big)\\
=& {1 \over 2} \Big((2-\theta)H(p)+ \theta H(q)\Big)\\
=& \big(1-{\theta \over 2}\big)H\Big({2\theta- \alpha \over 2-\theta}\Big) +{\theta \over 2}H\Big({\theta- \alpha \over \theta}\Big).
\end{align*}
%We conclude by applying the Mass Distribution Principle.
\end{proof}
\begin{lem} \label{entropy-measure-all}
If $\theta\notin [\alpha, (2+\alpha)/3]$ we have $E_\theta\cap A_\alpha=\emptyset$, otherwise
for $p=(2\theta-\alpha)/(2-\theta)$ and $q=\alpha / \theta$, we have for all $\omega\in E_\theta\cap A_\alpha$,
\[
h_{\mu_{p,q}}(\omega)=\big(1-{\theta \over 2}\big)H\Big({2\theta- \alpha \over 2-\theta}\Big) +{\theta \over 2}H\Big({\theta- \alpha \over \theta}\Big).
\]
\end{lem}
\begin{proof}
Observe that for any $\omega\in E_\theta\cap A_\alpha$, for any small $\epsilon>0$, for $n$ large enough, we have
\begin{align*}
N_1(\omega_{n/2}^n) &\in \Big[{\theta n \over 2}(1-\epsilon), \ {\theta n \over 2}(1+\epsilon)\Big]\\
N_1(\omega_n^{2n}) &\in \Big[{\theta n}(1-\epsilon), \ {\theta n}(1+\epsilon)\Big]\\
N_{11}(\omega_n^{2n}) &\in \Big[{\alpha n \over 2}(1-\epsilon), \ {\alpha n \over 2}(1+\epsilon)\Big].
\end{align*}
The obvious inequalities $$N_{11}(\omega_n^{2n}) \leq N_1(\omega_{n/2}^n) $$ and $$N_1(\omega_n^{2n})-N_{11}(\omega_n^{2n}) \leq n/2 + N_0(\omega_{n/2}^n) =n-N_1(\omega_{n/2}^n)$$ imply $\theta\in [\alpha, (2+\alpha)/3]$. Furthermore, we have
\begin{align*}
&\log \mu_{p,q}(C_{2n}(\omega)) - \log \mu_{p,q}(C_n(\omega))\\
=& {N_{11}(\omega_n^{2n})}\log q +\big({N_1(\omega_{n/2}^n)-N_{11}(\omega_n^{2n})}\big)\log (1-q)\\
&+\big({N_1(\omega_n^{2n})-N_{11}(\omega_n^{2n})}\big)\log p \\
&+\big({n-N_1(\omega_{n/2}^n)-N_1(\omega_n^{2n})+N_{11}(\omega_n^{2n})}\big)\log (1-p)\\
=& n\left( \big(1-{\theta \over 2}\big)H\Big({2\theta- \alpha \over 2-\theta}\Big) +{\theta \over 2}H\Big({\theta- \alpha \over \theta}\Big) + \epsilon \cdot O(1)\right).
\end{align*}
Hence, by the same argument as in the proof of Proposition \ref{prop:local-ent}, %Lemma \ref{entropy-mu-alpha},
we have
for all $\omega\in E_\theta\cap A_\alpha$,
\[
h_{\mu_{p,q}}(\omega)=\lim_{n\to\infty} {-\log \mu_{p,q}(C_n(\omega)) \over n }=\big(1-{\theta \over 2}\big)H\Big({2\theta- \alpha \over 2-\theta}\Big) +{\theta \over 2}H\Big({\theta- \alpha \over \theta}\Big).
\]
\end{proof}
\begin{proof}[Proof of Theorem \ref{thm:freq}]
We note that by Theorem \ref{lem:Bi}, the lower bound of Theorem \ref{thm:freq} follows from Lemmas \ref{typical-2} and \ref{entropy-measure-2} and the upper bound follows from Lemma \ref{entropy-measure-all}. We thus have completed our proof.
\end{proof}
\begin{proof}[Proof of Corollary \ref{cor}]
%Remark that we have used the same family of measures $\mu_{p,q}$ in the calculation of $h_{\rm top}(E_\theta \cap A_\alpha)$ as in the calculation of $h_{\rm top}(A_\alpha)$ in \cite{PeSo12}. By Lemma \ref{typical-2}, $h_{\rm top}(E_\theta \cap A_\alpha)$ is realized by the measure $\mu_{p,q}$ with $p=(2\theta-\alpha)/(2-\theta)$ and $q=\alpha / \theta$.
% Note however, that $h_{\rm top}(A_\alpha)$ is realized by $\mu_{p,q}$ with
% \begin{align}\label{eq:PS}
% \left\{
% \begin{array}{ll}
%\displaystyle p^2(1-q)=(1-p)^3,\\
% \displaystyle {2pq} =\alpha({2+p-q}).
%\end{array}\right.
%\end{align}
%Thus if the couple $(p,q)$ such that $p=(2\theta-\alpha)/(2-\theta)$ and $q=\alpha / \theta$ also satisfies (\ref{eq:PS}), that is (\ref{eq:max}) holds,
%%\[
%%(2\theta-\alpha)^2(\theta-\alpha)(2-\theta)=\theta(2-3\theta+\alpha)^3.
%%\]
%then the two realizing measures coincide and $h_{\rm top}(E_\theta \cap A_\alpha)=h_{\rm top}(A_\alpha)$.
Note that for any $\theta$, $E_\theta \cap A_\alpha \subset A_\alpha $. Thus, if $h_{\rm top}(E_\theta \cap A_\alpha)=h_{\rm top}(A_\alpha)$ for some $\theta$, then this $\theta$ is the maximal point for the entropy formula of $h_{\rm top}(E_\theta \cap A_\alpha)$. Remark that the entropy formula of $h_{\rm top}(E_\theta \cap A_\alpha)$ in Theorem \ref{thm:freq} is analytic and concave with respect to the variable $\theta$ and the partial derivative
\[
{\partial h_{\rm top}(E_\theta \cap A_\alpha) \over \partial \theta}= {1 \over 2} \log {\theta(2-3\theta+\alpha)^3 \over (2\theta-\alpha)^2(\theta-\alpha)(2-\theta) }=0
\]
if and only if (\ref{eq:max}) holds. We then have proved the first assertion.
%\[
%(2\theta-\alpha)^2(\theta-\alpha)(2-\theta)=\theta(2-3\theta+\alpha)^3.
%\]
By Theorems \ref{thm:KPS} and \ref{thm:PS}, $h_{\rm top}(A)=h_{\rm top}(A_0)$.
By Theorem \ref{thm:freq}, $h_{\rm top}(E_\theta \cap A)=h_{\rm top}(E_\theta \cap A_0)$. Then the second assertion follows by taking $\alpha=0$.
\end{proof}
%\begin{rem}
%Equations \eqref{eq:exact1} and \eqref{eq:exact2} do not easily show that our formulas reduce to ones given in \cite{PeSo} in case $\lambda_1=\lambda_0=1$. However, \eqref{eq:exact1} together with \eqref{eq:aux} easily give
%
%\[
%\dim_H \nu_{p,q} = \left( \alpha \log \frac {p(1-q)} {(1-p)q} -2\log (1-p)\right) \left(\frac 1 {4\lambda_0} + \frac 1 {4\lambda_1} \right) - \frac 1 {4\lambda_1} \log \frac {p^2 (1-q)} {(1-p)^3}.
%\]
%
%\end{rem}
\section*{Acknowlegdements} We thank the referees for the valuable remarks and suggestions which significantly improves the presentation of the paper. We thank one of the referees for pointing out us the reference \cite{W49} of Wall's Ph.D. thesis where we can find the original version of Lemma \ref{lem:subseq}.
\bibliographystyle{alpha}
\begin{thebibliography}{PSSS12}
\bibitem[ABC19]{ABC} C. Aistleitner, V. Becher and O. Carton,
{\em Normal numbers with digit dependencies}, Trans. Amer. Math. Soc., {\bf 372} (2019), 4425--4446.
\bibitem[B61]{Bi61} P. Billingsley, {\em Hausdorff dimension in probability theory. II}, Illinois J.
Math., {\bf 5}, 291--298.
\bibitem[B73]{B73}
R. Bowen.
\newblock {\em Topological entropy for noncompact sets},
\newblock Trans. Amer. Math. Soc., {\bf 184}:125--136, 1973.
\bibitem[F90]{F90} K. Falconer, {\it Fractal Geometry, Mathematical Foundations and Application}, Wiley, 1990.
\bibitem[FLM12]{FLM} A.H. Fan, L.M. Liao and J.H. Ma, {\em Level sets of multiple ergodic averages},
\newblock Monatsh. Math. \textbf{168} (2012), 17--26.
%\bibitem{FSW} A.H. Fan, J. Schmeling and M. Wu, {\em Multifractal analysis of multiple ergodic averages}.
%Comptes Rendus Math\'ematique,
%volume \textbf{349}, num\'ero 17--18 (2011), 961--964.
\bibitem[FSW16]{FSW16} A.H. Fan, J. Schmeling and M. Wu, {\em Multifractal analysis of some multiple ergodic averages}, Adv. Math. \textbf{295} (2016), 271--333.
\bibitem[K73]{K73} T. Kamae, {\em Subsequences of normal sequences}, Israel J. Math. {\bf 16} (1973), 121-149.
\bibitem[KPS12]{KPS12}
R. Kenyon, Y. Peres and B. Solomyak, {\em Hausdorff dimension for fractals invariant under the multiplicative integers}, Ergodic Theory and Dynamical Systems,
\textbf{32} (2012), no. 5, 1567--1584.
\bibitem[K12]{K12}
Y. Kifer, {\em A nonconventional strong law of large numbers and fractal dimensions of some multiple recurrence sets}, Stoch. Dyn., {\bf 12} (3) (2012),
21p.
\bibitem[MW08]{MW08} J.H. Ma, Z.Y. Wen, {\em A Billingsley type theorem for Bowen entropy}, C. R. Math. Acad. Sci. Paris {\bf 346} (2008), no. 9-10, 503--507.
\bibitem[PS12]{PeSo12}
Y. Peres and B. Solomyak, {\em Dimension spectrum for a nonconventional ergodic average}, Real Analysis Exchange, \textbf{37} (2011), no. 2, 375-388.
\bibitem[W49]{W49} D. D. Wall, Normal Numbers. Ph.D. thesis,
University of California, Berkeley, California, USA, 1949.
\end{thebibliography}
%
%\bibitem[Bow73]{Bowen73}
%Rufus Bowen.
%\newblock Topological entropy for noncompact sets.
%\newblock {\em Trans. Amer. Math. Soc.}, 184:125--136, 1973.
%
%\bibitem[FLM]{FLM}
%Ai-hua Fan, Lingmin Liao, and Ji-hua Ma.
%\newblock Level sets of multiple ergodic averages. , (5 november 2011, pp.
% 1-10, doi:10.1007/s00605-011-0358-5).
%\newblock {\em Monatshe. Math.}, 148, 2011 online.
%
%\bibitem[FSW11]{FSW11}
%Aihua Fan, J{\"o}rg Schmeling, and Meng Wu.
%\newblock Multifractal analysis of multiple ergodic averages.
%\newblock {\em C. R. Math. Acad. Sci. Paris}, 349(17-18):961--964, 2011.
%
%\bibitem[FSW12a]{FSW12}
%Aihua Fan, J{\"o}rg Schmeling, and Meng Wu.
%\newblock Multifractal analysis of multiple ergodic averages.
%\newblock {\em preprint}, 2012.
%
%\bibitem[FSW12b]{FanSW12}
%Aihua Fan, J{\"o}rg Schmeling, and Meng Wu.
%\newblock The multifractal spectra of v-statistics.
%\newblock In {\em New developments in Fractals and related Fields}.
% Birkh\"auser, Boston, 2012, to appear.
%
%\bibitem[KPS11]{KPS11}
%Richard Kenyon, Yuval Peres, and Boris Solomyak.
%\newblock Hausdorff dimension of the multiplicative golden mean shift.
%\newblock {\em C. R. Math. Acad. Sci. Paris}, 349(11-12):625--628, 2011.
%
%\bibitem[KPS12]{KPS12}
%Richard Kenyon, Yuval Peres, and Boris Solomyak.
%\newblock Hausdorff dimension for fractals invariant under the multiplicative
% integers.
%\newblock {\em Ergodic Theory Dynamical Systems}, arXiv:1102.5136, 2012.
%
%\bibitem[PS12b]{PeSo12}
%Yuval Peres and Boris Solomyak.
%\newblock Dimension spectrum for a nonconventional ergodic average.
%\newblock {\em Real Anal. Ex.}, arXiv:1107.1749, 2012.
%
%\bibitem[PS12a]{PS}
%Yuval Peres and Boris Solomyak.
%\newblock The multiplicative golden mean shift has infinite hausdorff measure.
%\newblock In {\em New developments in Fractals and related Fields}.
% Birkh\"auser, Boston, 2012 to apear, arXiv:1201.5842.
%
%\bibitem[PSSS12]{PSSS}
%Yuval Peres, J{\"o}rg Schmeling, St\'ephane Seuret, and Boris Solomyak.
%\newblock Dimensions of some fractals defined via the semi-group generated by
% $2$ and $3$.
%\newblock {\em preprint}, 2012.
%
%\end{thebibliography}
%\bibliography{multi}
\end{document}