Block-counting sequences are not purely morphic

Given a positive integer $m$ larger than $1$ and a finite word $w$ over $\left\{0,1,2,...,m-1\right\}$, the block-counting sequence $(e_{m;w}(n))_{n\in\mathbf{N}}$ counts the number of occurrences of the word $w$ in the $m$-expansion of $n$ for each non-negative integer $n$. Let us define $(a_{m;w}(n))_{n\in\mathbf{N}}$ to be a sequence over $\left\{0,1,2,...,m-1\right\}$ such that $a_{m;w}(n)\equiv e_{m;w}(n)\mod(m)$ for all non-negative integer $n$. The analytical as well as the combinatorial properties of these sequences have been studied since 1900’s after Thue and some well-known sequences are strongly related to this notion. Recall that the $0,1$-Thue-Morse sequence can be defined as $(a_{2;1}(n))_{n\in\mathbf{N}}$ (see, for example, Page 15 in [4]) and the $0,1$-Rudin-Shapiro sequence can also be defined as $(a_{2;11}(n))_{n\in\mathbf{N}}$ (see, for example, Example 3.3.1 in [4]). In this article, we review some common properties of usual block-counting sequences and generalize them to all block-counting sequences.

To be able to announce our results, here we recall some definitions and notation. Let $A$ be a finite set. It will be called an alphabet and its elements will be called letters. Let $A^{*}$ denote the free monoid generated by $A$ under concatenations and let $A^{\mathbf{N}}$ denote the set of infinite concatenations of elements in $A$. Let $A^{\infty}=A^{*}\cup A^{\mathbf{N}}$. A finite word over the alphabet $A$ is an element in $A^{*}$ and an infinite word over $A$ is an element in $A^{\mathbf{N}}$. Particularly, the empty word is an element in $A^{*}$ and it is denoted by $\epsilon$. The length of a word $w$, denoted by $|w|$, is the number of letters that it contains. The length of the empty word is $0$ and the length of any infinite word is infinite. For any non-empty word $w\in A^{\infty}$, it can be denoted by $w[0]w[1]w[2]...$, where $w[i]$ are elements in $A$. A word $x$ is called a factor of $w$ if there exist two integers $0\leq i\leq j\leq|w|-1$ such that $x=w[i]w[i+1]...w[j]$, this factor can also be denoted by $w[i..j]$. A factor $x$ is called a prefix (resp. a suffix ) of the word $w$ if there exists a positive integer $i$ such that $0\leq i\leq|w|$ and $x=w[0..i]$ (resp. $x=w[i..|w|-1]$). For any finite word $w$ and any positive integer $n$, let $w^{n}$ denote the concatenation of $n$ copies of $w$, i.e. $w^{n}=ww...w$ $n$ times. Particularly, $w^{0}=\epsilon$. For any pair of words $w,v$ such that $v$ is a factor of $w$, let $|w|_{v}$ denote the number of occurrences of $v$ in $w$.

Let $A$ and $B$ be two alphabets, a morphism $\phi$ from $A$ to $B$ is a map from $A^{\infty}$ to $B^{\infty}$ satisfying $\phi(xy)=\phi(x)\phi(y)$ for any pair of elements $x,y$ in $A^{\infty}$. The morphism $\phi$ is called $k$-uniform if for all elements $a\in A$, $|\phi(a)|=k$ and it is called non-uniform otherwise. A morphism $\phi$ is called a coding function if it is $1$-uniform and it is called non-erasing if $\phi(a)\neq\epsilon$ for all $a\in A$.

Let $A$ be a finite alphabet and let $(a_{n})_{n\in\mathbf{N}}$ be an infinite sequence over $A$, it is called morphic if there exists an alphabet $B$, an infinite sequence $(b_{n})_{n\in\mathbf{N}}$ over $B$, a non-erasing morphism $\phi$ from $B^{\infty}$ to $B^{\infty}$ and a coding function $\psi$ from $B^{\infty}$ to $A^{\infty}$, such that $(b_{n})_{n\in\mathbf{N}}$ is a fixed point of $\phi$ and $(a_{n})_{n\in\mathbf{N}}=\psi((b_{n})_{n\in\mathbf{N}})$. Moreover, the sequence $(a_{n})_{n\in\mathbf{N}}$ is called uniformly morphic if $\phi$ is $k$-uniform for some integer $k$, and it is called non-uniformly morphic otherwise. The sequence $(a_{n})_{n\in\mathbf{N}}$ is called purely morphic if $A=B$ and $\psi=Id$.

For any positive integer $m$, let $[\![m]\!]=\left\{0,1,2,...,m-1\right\}$. For any $t\in[\![m]\!]$ let $t^{+}\equiv t+1\mod m$; for any $w\in[\![m]\!]^{*}$, let $w^{+}=w[0]^{+}w[1]^{+}...w[|w|-1]^{+}$.

In Section2, we give a fast algorithm to generate all block-counting sequences. It is well-known that the Thue-Morse sequence can be generated by the following algorithm (see, for example, [12, A008277]):

In Section2, we prove that the Rudin-Shapiro sequence can also be generalized by a similar algorithm, see 4. More generally, we find fast algorithms to generate all block-counting sequences. These algorithms are given by 3 and 5 in Section2.

From the definitions recalled as above, any morphic word can be classified as either a uniformly morphic word or a non-uniformly morphic word. However, from a recent article [5], Allouche and Shallit proved that all uniformly morphic sequences are also non-uniformly morphic. This result implies that all sequences in the family of morphic sequences are also in its subfamily of non-uniformly morphic sequences. Indeed, many works can be found in the literature in the direction of characterizing all those non-uniformly morphic sequences which are not uniformly morphic, for example, one can find [2][13][7][1][8][9][6]. However, in [5], it is proved actually that all uniformly morphic sequences are also non-uniformly non-purely morphic. In other words, from the construction of the proof given in [5], a nontrivial coding function is required. In Section 3, we investigate all those uniformly morphic sequences which are not purely morphic. It is already known that the Rudin-Shapiro sequence is in this case (Example 26 in [3]). In section 3, we prove that all other sequences in the form of $(a_{m,w}(n))_{n\in\mathbf{N}}$ have the same property when $|w|\neq 1$ and $m$ is a prime. The result is announced as follows:

In Section 4, under the assumption that $p$ is a prime number, we prove that the formal power series $f_{p;w}=\sum_{i=0}^{\infty}a_{m;w}(n)t^{n}$ is algebraic and of degree $p$ over $\mathbb{F}_{p}(t)$. Indeed, from Christol’s theorem [11], we know that the power series $f_{p;w}$ is algebraic over $\mathbb{F}_{p}(t)$. In Section 4, we prove that $f$ is algebraic of degree $p$.

For any positive integer $m$ and non-negative integer $n$, let $[n]_{m}$ denote the expansion of $n$ in the base $m$. For a given word $w\in[\![m]\!]^{*}=\{0,1,\cdots,m-1\}^{*}$, $w=w[0]w[1]...w[|w|-1]$, let $(w)_{m}=\sum_{i=0}^{|w|-1}w[i]m^{|w|-1-i}$ and let $w^{\prime}=w[1]w[2]\cdots w[|w|-1]$. A word $w$ is called a $x$-word if $w[0]=x$. For a given string $w$, let $\alpha_{w}=\frac{(w^{\prime})_{m}}{m^{|w|-1}}$, $\beta_{w}=\frac{(w^{\prime})_{m}+1}{m^{|w|-1}}$ and let $\phi_{w}:[\![m]\!]^{*}\to[\![m]\!]^{*}$ be a function such that for any $v\in[\![m]\!]^{*}$, $\phi_{w}(v)$ satisfies the following propriety:

From now on, we work with $p$ a prime number.

We first prove that $(a_{p;w}(n))_{n\in\mathbf{N}}$ is not purely morphic when $|w|>1$. We will need a simple notation, for $w=w[0]\cdots w[|w|-1]$, let $w^{\diamond}=w[0]\cdots w[|w|-2]$.

For the sequence $(a_{p;w}(n))_{n\in\mathbf{N}}$, let us define a $p$-block to be a sub-sequence of the form $(a_{p;w}(pn+i))_{0\leq i\leq p-1}$ for some integer $n$. From the previous proposition, a $p$-block is either of type 1 or type 2. For a $p$-block $(a_{p;w}(pn+i))_{0\leq i\leq p-1}$ of type 2, let us define its index to be an integer $i\in[\![p]\!]$ such that $a_{p;w}(pn+i)\neq a_{p;w}(pn+j)$ for all $j\neq i$.

This proposition will be proved with the help of the following lemmas.

Now we are able to prove Proposition 10.

Now we are able to prove the principle theorem in most cases:

Here we prove the $p$ particular cases.