1: \documentclass{amsart}
2:
3: \usepackage{a4wide,vaucanson-g}
4:
5: \newtheorem{lemma}{Lemma}[section]
6: \newtheorem{thm}[lemma]{Theorem}
7: \newtheorem{definition}[lemma]{Definition}
8: \newtheorem{proposition}[lemma]{Proposition}
9: \newtheorem{corollary}[lemma]{Corollary}
10:
11: \def\R{\mathcal R}
12: \def\El{\mathcal E_\ell}
13: \def\Er{\mathcal E_r}
14:
15: \title{Sequences with constant number of return words}
16:
17: \author{L\!'ubom\'ira Balkov\'a}
18: \author{Edita Pelantov\'a}
19: \address{Doppler Institute for Mathematical Physics and Applied
20: Mathematics, and Department of Mathematics, FNSPE, Czech Technical
21: University, Trojanova 13, 120~00 Praha~2, Czech Republic}
22: \email{l.balkova@centrum.cz, Pelantova@km1.fjfi.cvut.cz}
23:
24: \author{Wolfgang Steiner}
25: \address{LIAFA, CNRS, Universit\'e Paris Diderot -- Paris 7, Case 7014,
26: 75205 Paris Cedex 13, France}
27: \email{steiner@liafa.jussieu.fr}
28: \date\today
29:
30: \begin{document}
31: \begin{abstract}
32: An infinite word has the property $R_m$ if every factor has exactly
33: $m$ return words.
34: Vuillon showed that $R_2$ characterizes Sturmian words.
35: We prove that a word satisfies $R_m$ if its complexity function is
36: $(m-1)n+1$ and if it contains no weak bispecial factor.
37: These conditions are necessary for $m=3$, whereas for $m=4$ the
38: complexity function need not be $3n+1$.
39: New examples of words satisfying $R_m$ are given by words related to
40: digital expansions in real bases.
41: \end{abstract}
42:
43: \maketitle
44:
45: \section{Introduction}
46: Recently, return words have been intensively studied in (symbolic)
47: dynamical systems, combinatorics on words and number theory.
48: Roughly speaking, for a~given factor $w$ of an infinite word $u$,
49: a~return word of $w$ is a~word between two successive occurrences of
50: the factor~$w$.
51: This can be seen as a symbolic version of the first return map in a
52: dynamical system.
53: This notion was introduced by Durand~\cite{Durand} to give a~nice
54: characterization of primitive substitutive sequences.
55: A~slightly different notion of return words was used by Ferenczi,
56: Mauduit and Nogueira~\cite{FeMaNo}.
57:
58: Sturmian words are aperiodic words over a~biliteral alphabet with the
59: lowest possible factor complexity; they were defined by Morse and
60: Hedlund~\cite{MoHe}.
61: Using return words, Vuillon~\cite{Vuillon} found a~new equivalent
62: definition of Sturmian words.
63: He showed that an infinite word $u$ over a~biliteral alphabet is
64: Sturmian if and only if any factor of $u$ has exactly two return words.
65: A short proof of this fact is given in Section~\ref{R2R3}.
66:
67: A natural generalization of Sturmian words to $m$-letter alphabets is
68: constituted by infinite words with every factor having exactly $m$
69: return words.
70: This property is called $R_m$.
71: It covers other generalizations of Sturmian words:
72: Justin and Vuillon~\cite{JusVui} proved that Arnoux-Rauzy words of
73: order $m$ satisfy $R_m$, Vuillon~\cite{VuIET} proved this property for
74: words coding regular $m$-interval exchange transformations.
75:
76: The factor complexity, i.e., the number of different factors of length
77: $n$, of the two classes of words with property $R_m$ in the preceding
78: paragraph is $(m-1)n+1$ for all $n\ge0$.
79: Vuillon~\cite{VuIET} observed that this condition is not sufficient to
80: describe words satisfying $R_m$, $m\ge3$: the fixed point of a certain
81: recoding of the Chacon substitution, which has complexity $2n+1$ by
82: Ferenczi~\cite{Fer}, has factors with more than $3$ return words.
83:
84: A deeper inspection of the two classes of words with property $R_m$
85: shows that not only the first difference of complexity is constant, but
86: also that the bilateral order of every factor (see
87: Cassaigne~\cite{Cass} and Section~\ref{sufficient}) is zero.
88: We show that this condition is indeed sufficient to have the property
89: $R_m$, and provide a less known class of words satisfying this
90: condition.
91: If a word satisfies $R_3$, then we can show that no factor is weak
92: bispecial, i.e., no factor has negative bilateral order.
93: Therefore the words with $R_3$ are characterized by complexity $2n+1$
94: and the absence of weak bispecial factors.
95:
96: In Section~\ref{cex}, we provide a word satisfying $R_4$ with an even
97: number of factors of every positive length (containing infinitely many
98: weak bispecial factors).
99: Therefore words satisfying $R_m$ do not necessarily have complexity
100: $(m-1)n+1$, and it is an open question whether there exists a nice
101: characterization of words satisfying $R_m$ for $m\ge4$.
102:
103: We conclude the article by exhibiting a large class of purely
104: substitutive words satisfying $R_m$.
105: Every word in this class codes the sequence of distances between
106: consecutive $\beta$-integers for some real number $\beta>1$.
107:
108: In this article we focus only on the number of return words
109: corresponding to a given factor of an infinite word.
110: We do not study the ordering of return words in the infinite word,
111: i.e., we do not study derivated sequences (see~\cite{Durand} for the
112: precise definition).
113: Let us just mention here that a derivated sequence of a word with
114: property $R_m$ is again a word satisfying $R_m$.
115: A~description of derivated sequences of Sturmian words can be found
116: in~\cite{ArBr}.
117:
118:
119:
120: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
121: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
122:
123: \section{Basic definitions} \label{Preliminaries}
124:
125: An {\em alphabet} $\mathcal A$ is a~finite set of symbols called
126: {\em letters}. A~(possibly empty) concatenation of letters is a~{\em
127: word}. The set $\mathcal A^{*}$ of all finite words provided with
128: the operation of concatenation is a~free monoid. The {\em length} of
129: a~word $w$ is denoted by $|w|$. A~finite word $w$ is called a~{\em
130: factor} (or {\em subword}) of the (finite or right infinite) word
131: $u$ if there exist a~finite word $v$ and a~word $v'$ such that
132: $u=vwv'$. The word $w$ is a~{\em prefix} of $u$ if $v$ is the empty
133: word. Analogously, $w$ is a~{\em suffix} of $u$ if $v'$ is the empty
134: word. A~concatenation of $k$ words $w$ will be denoted by $w^k$.
135:
136: The {\em language} $\mathcal{L}(u)$ is the set of all factors of the
137: word $u$, and $\mathcal{L}_n(u)$ is the set of all factors of $u$ of
138: length $n$.
139: Let $w$ be a~factor of an infinite word $u$ and let $a,b\in\mathcal A$.
140: If $aw$ is a~factor of $u$, then we call $a$ a~{\em left extension} of
141: $w$.
142: Analogously, we call $b$ a~{\em right extension} of~$w$ if
143: $wb\in\mathcal L(u)$.
144: We will denote by $\El(w)$ the set of all left extensions of $w$, and
145: by $\Er(w)$ the set of right extensions.
146: A~factor $w$ is {\em left special} if $\#\El(w)\ge2$, {\em right
147: special} if $\#\Er(w)\ge2$ and {\em bispecial} if $w$ is both left
148: special and right special.
149:
150: Let $w$ be a~factor of an infinite word $u=u_0u_1\cdots$ (with
151: $u_j\in\mathcal A$), $|w|=\ell$.
152: An integer $j$ is called an {\em occurrence} of $w$ in $u$ if
153: $u_ju_{j+1}\cdots u_{j+\ell-1}=w$.
154: Let $j,k$, $j<k$, be successive occurrences of $w$.
155: Then $u_ju_{j+1}\cdots u_{k-1}$ is a~{\em return word} of $w$.
156: The set of all return words of $w$ is denoted by $\R(w)$,
157: $$
158: \R(w)=\{u_ju_{j+1}\dots u_{k-1}\mid j,k \mbox{ being successive
159: occurrences of } w \mbox{ in }u\}.
160: $$
161: If $v$ is a return word of $w$, then the word $vw$ is called
162: {\em complete return word}.
163:
164: An infinite word is {\em recurrent} if any of its factors occurs
165: infinitely often or, equivalently, if any of its factors occurs at
166: least twice.
167: It is {\em uniformly recurrent} if, for any $n\in\mathbb N$, every
168: sufficiently long factor contains all factors of length $n$.
169: It is not difficult to see that a recurrent word on a finite alphabet
170: is uniformly recurrent if and only if the set of return words of any
171: factor is finite.
172:
173: The variability of local configurations in $u$ is expressed by the
174: {\em factor complexity function} (or simply {\em complexity})
175: $ C(n)=\#\mathcal L_n(u)$. It is well known that a~word $u$ is
176: aperiodic if and only if $ C(n)\ge n+1$ for all $n \in \mathbb N$.
177: Infinite aperiodic words with the minimal complexity $ C(n)=n+1$ for
178: all $n\in \mathbb N$ are called {\em Sturmian words}. These words
179: have been studied extensively, and several equivalent definitions of
180: Sturmian words can be found in Berstel~\cite{Berstel}.
181:
182: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
183: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
184:
185: \section{Simple facts for return words}\label{HandyRules}
186:
187: \subsection{Restriction to bispecial factors}\label{sectbisp}
188: If a~factor $w$ is not right special, i.e., if it has a~unique right
189: extension $b\in\mathcal A$, then the sets of occurrences of $w$ and
190: $wb$ coincide, and
191: $$
192: \R(w)=\R(wb).
193: $$
194: If a~factor $w$ has a~unique left extension $a\in\mathcal A$, then
195: $j\ge 1$ is an occurrence of $w$ in the infinite word $u$ if and only
196: if $j-1$ is an occurrence of $bw$. This statement does not hold for
197: $j=0$.
198: Nevertheless, if $u$ is a recurrent infinite word, then the set of
199: return words of $w$ stays the same no matter whether we include the
200: return word corresponding to the prefix $w$ of $u$ or not.
201: Consequently, we have
202: $$
203: \R(aw)=a\R(w)a^{-1}=\{ava^{-1}\mid v\in\R(w)\},
204: $$
205: where $ava^{-1}$ means that the word $v$ is prolonged to the left by
206: the letter $a$ and it is shortened from the right by erasing the
207: letter $a$ (which is always the suffix of $v$ for $v\in\R(w)$).
208:
209: For an aperiodic uniformly recurrent infinite word $u$, each factor
210: $w$ can be extended to the left and to the right to a~bispecial factor.
211: To describe the cardinality and the structure of $\R(w)$ for
212: arbitrary $w$, it suffices therefore to consider bispecial factors $w$.
213:
214: \subsection{Tree of return words}\label{tree}
215: It is convenient to consider a tree (or trie) constructed in the
216: following way:
217: Label the root with a factor $w$, and attach $\#\Er(w)$ children,
218: with labels $wb$, $b\in \Er(w)$.
219: Repeat this recursively with every node labeled by $v$, except if $w$
220: is a suffix of $v$.
221: If $u$ is uniformly recurrent, then this algorithm stops, and it is
222: easy to see that the labels of the leaves of this tree are exactly the
223: complete return words of $w$.
224: Therefore we have
225: \begin{equation}\label{leaves}
226: \#\R(w)=\#\{\text{leaves}\}=1+\sum_{\text{non-leaves }v}(\#\Er(v)-1).
227: \end{equation}
228: In particular, if $w$ is the unique right special factor of its length,
229: then $\#\R(w)=\#\Er(w)$.
230:
231: \begin{figure}
232: {\small\VCDraw{\begin{VCPicture}{(0,-3)(20,3)}
233: \StateVar[01]{(0,0)}{a}
234: \StateVar[010]{(2.5,1.5)}{b}
235: \StateVar[011]{(2.5,-1.5)}{c}
236: \StateVar[0100]{(5,2.5)}{d}
237: \StateVar[0101]{(5,.5)}{e}
238: \StateVar[0110]{(5,-1.5)}{f}
239: \StateVar[01001]{(8,2.5)}{g}
240: \StateVar[01100]{(8,-.5)}{h}
241: \StateVar[01101]{(8,-2.5)}{i}
242: \StateVar[011001]{(11,-.5)}{j}
243:
244: \SetEdgeArrowStyle{-}
245: \EdgeL ab{}
246: \EdgeL ac{}
247: \EdgeL bd{}
248: \EdgeL be{}
249: \EdgeL cf{}
250: \EdgeL dg{}
251: \EdgeL fh{}
252: \EdgeL fi{}
253: \EdgeL hj{}
254:
255: \SmallState
256: \State{(14,0)}{k}
257: \State{(16,0)}{l}
258: \State{(18,1)}{m}
259: \State{(18,-1)}{n}
260: \State{(20,2)}{o}
261: \State{(20,.5)}{p}
262: \State{(20,-.5)}{q}
263: \State{(20,-2)}{r}
264:
265: \EdgeL kl{01}
266: \EdgeL lm{0}
267: \EdgeL ln{10}
268: \EdgeL mo{01}
269: \EdgeL mp{1}
270: \EdgeL nq{01}
271: \EdgeL nr{1}
272: \end{VCPicture}}}
273: \caption{The tree of return words of $01$ in the Thue-Morse sequence
274: and its trie representation.}
275: \end{figure}
276:
277: A similar construction can be done with left extensions, yielding
278: similar formulae.
279: Since we can restrict our attention to bispecial factors $w$ by
280: Section~\ref{sectbisp}, we obtain the following proposition.
281:
282: \begin{proposition}\label{onespecial}
283: Let $u$ be a recurrent word and $m\in\mathbb N$.
284: Suppose that for every $n\in\mathbb N$ at least one of the following
285: conditions is satisfied:
286: \begin{itemize}
287: \item
288: There is a unique left special factor $w\in\mathcal L_n(u)$, and
289: $\#\El(w)=m$.
290: \item
291: There is a unique right special factor $w\in\mathcal L_n(u)$, and
292: $\#\Er(w)=m$.
293: \end{itemize}
294: Then $u$ satisfies property $R_m$, i.e., every factor has exactly $m$
295: return words.
296: \end{proposition}
297:
298: Recall that Arnoux-Rauzy words of order $m$ are defined as uniformly
299: recurrent infinite words which have for every $n\in\mathbb N$ exactly
300: one right special factor $w$ of length $n$ with $\#\Er(w)=m$ and
301: exactly one left special factor $w$ of length $n$ with $\#\El(w)=m$.
302: They are also called strict episturmian words.
303: It is easy to see that Sturmian words are recurrent, and we obtain the
304: following corollary to Proposition~\ref{onespecial}.
305:
306: \begin{corollary}\label{ArnouxRauzy}
307: Arnoux-Rauzy words of order $m$ satisfy $R_m$, in particular Sturmian
308: words satisfy $R_2$.
309: \end{corollary}
310:
311: \section{Sufficient conditions for property $R_m$}\label{sufficient}
312:
313: This section is devoted to sufficient conditions for a word $u$ having
314: the property $R_m$, but we mention first two evident necessary
315: conditions.
316:
317: The alphabet $\mathcal A$ of $u$ must have $m$ letters since the
318: occurrences of the empty word are all integers $n\ge0$, and its return
319: words are therefore all letters $u_n$.
320: Furthermore, $u$ must be uniformly recurrent since every factor has a
321: return word and only finitely many of them.
322:
323: An important role in our further considerations is played by weak
324: bispecial factors.
325:
326: \begin{definition}\label{weak}
327: A factor $w$ of a recurrent word is {\em weak bispecial} if $B(w)<0$,
328: where
329: $$
330: B(w)=
331: \#\{awb\in\mathcal L(u)\mid a,b\in\mathcal A\}-\#\El(w)-\#\Er(w)+1
332: $$
333: is the {\em bilateral order} of $w$.
334: \end{definition}
335:
336: Since
337: $\#\{awb\in\mathcal L(u)\mid a,b\in\mathcal A\}=
338: \sum_{a\in \El(w)}\#\Er(aw)=\sum_{b\in \Er(w)}\#\El(wb)$,
339: the inequality $B(w)<0$ is equivalent to
340: $$
341: \sum_{a\in \El(w)}(\#\Er(aw)-1)<\#\Er(w)-1
342: $$
343: and to
344: $$
345: \sum_{b\in \Er(w)}(\#\El(wb)-1)<\#\El(w)-1.
346: $$
347:
348: The bilateral order was defined by Cassaigne~\cite{Cass} in order to
349: calculate the second complexity difference.
350: If we set $\Delta C(n)=C(n+1)-C(n)$, then we have
351: $$
352: \Delta C(n) = \sum_{w \in \mathcal{L}_n(u)} \bigl(\#\El(w)-1\bigr)
353: = \sum_{w \in \mathcal{L}_n(u)} \bigl(\#\Er(w)-1\bigr)
354: $$
355: and therefore
356: \begin{multline*}
357: \Delta C(n+1)-\Delta C(n)=\sum_{w\in\mathcal L_n(u)}\sum_{a\in \El(w)}
358: (\#\Er(aw)-1)-\sum_{w\in\mathcal L_n(u)}(\#\Er(w)-1) \\
359: =\sum_{w\in\mathcal L_n(u)}\big(\#\{awb\in\mathcal L(u)\mid a,b\in
360: \mathcal A\}-\#\El(w)-\#\Er(w)+1\big)=\sum_{w\in\mathcal L_n(u)}B(w).
361: \end{multline*}
362:
363: If $B(w)=0$ for all factors $w$, then the first complexity difference
364: is constant.
365: If no factor is weak bispecial, then $\Delta C(n)$ is non-decreasing.
366: Since $\Delta C(0)=\#\mathcal A-1$ and $\#A=m$, we obtain the following
367: lemma.
368:
369: \begin{lemma}\label{l0}
370: If $u$ satisfies $R_m$ and no factor is weak bispecial, then
371: $\Delta C(n)\ge m-1$ for all $n\ge0$.
372: \end{lemma}
373:
374: %\begin{definition} A set $S$ is called {\em representative system of right special factors} if \begin{itemize} \item $v$ is not a suffix of $w$ for all $v,w\in S$, $v\ne w$, and \item for every right special factor $w$ there exists $v\in S$ such that either $v$ is a suffix of $w$ or $w$ is a suffix of $v$. \end{itemize} \end{definition}
375:
376: The number of return words can be bounded by the following lemmas.
377:
378: \begin{lemma}\label{l1}
379: If $u$ is a uniformly recurrent word with no weak bispecial factor,
380: then
381: $$
382: \#\R(w)\ge1+\Delta C(|w|)
383: $$
384: for every factor $w\in\mathcal L(u)$.
385: \end{lemma}
386:
387: \noindent{\em Proof.} Let $w\in\mathcal L(u)$ and denote by
388: $v_1,v_2,\ldots,v_r$ the right special factors of length $|w|$.
389: Since no factor is weak bispecial and $u$ is uniformly recurrent,
390: every $v_j$ can be extended to the left without decreasing the total
391: amount of ``right branching'' until $w$ is reached.
392: More precisely, we have (mutually different) right special factors
393: $v_j^{(1)},v_j^{(2)},\ldots,v_j^{(s_j)}$ with suffix $v_j$, prefix
394: $w$ and no other occurrence of $w$ such that
395: $\#\Er(v_j)-1\le\sum_{i=1}^{s_j}(\#\Er(v_j^{(i)})-1)$. Since all
396: $v_j^{(i)}$ are nodes in the tree of return words and $v_j^{(i)}\ne
397: v_{j'}^{(i')}$ if $(j,i)\ne(j',i')$, we can use (\ref{leaves}) and
398: obtain
399: $$
400: \qquad\#\R(w)\ge1+\sum_{j=1}^r\sum_{i=1}^{s_j}(\#\Er(v_j^{(i)})-1)\ge
401: 1+\sum_{j=1}^r(\#\Er(v_j)-1)=1+\Delta C(|w|).\qquad\qed
402: $$
403: %\end{proof}
404:
405: \begin{lemma}\label{l2}
406: If $u$ has no weak bispecial factor and $\Delta C(n)<m$ for all
407: $n\ge0$, then
408: $$
409: \#\R(w)\le m
410: $$
411: for every factor $w\in\mathcal L(u)$.
412: \end{lemma}
413:
414: %\begin{proof}
415: \noindent{\em Proof.} Let $v_1,v_2,\ldots,v_r$ denote the right
416: special factors which are labels of non-leave nodes in the tree of
417: return words of $w$, and $n=\max_{1\le j\le r}|v_j|$. Since no
418: bispecial factor is weak, every $v_j$ can be extended to the left to
419: factors of length $n$ without decreasing the total amount of ``right
420: branching''. More precisely, we have (mutually different) right
421: special factors $v_j^{(1)},v_j^{(2)},\ldots,v_j^{(s_j)}$ of length
422: $n$ with suffix $v_j$ such that
423: $\#\Er(v_j)-1\le\sum_{i=1}^{s_j}(\#\Er(v_j^{(i)})-1)$. Since $w$
424: occurs in $v_j$ only as prefix, no $v_j$ can be a proper suffix of
425: $v_{j'}$. Hence we have $v_j^{(i)}\ne v_{j'}^{(i')}$ if
426: $(j,i)\ne(j',i')$ and
427: $$
428: \#\R(w)=1+\sum_{j=1}^r\big(\#\Er(v_j)-1\big)\le 1+\sum_{j=1}^r
429: \sum_{i=1}^{s_j}\big(\#\Er(v_j^{(i)})-1\big)\le1+\Delta C(n)\le m.\qed
430: $$
431: %\end{proof}
432:
433: For words with no weak bispecial factors, these three lemmas give a
434: very simple characterization of the property $R_m$.
435:
436: \begin{thm}\label{th1}
437: If $u$ is a uniformly recurrent word with no weak bispecial factor,
438: then it satisfies $R_m$ if and only if $ C(n)=(m-1)n+1$ for all $n\ge0$.
439: \end{thm}
440:
441: \section{Properties $R_2$ and $R_3$}\label{R2R3}
442:
443: For $m=2$ and $m=3$, we can completely characterize the words with
444: property $R_m$.
445:
446: \begin{definition}
447: Let $v$ be a return word of $w\in\mathcal L(u)$. We say that return
448: word $v$ starts with $b$ if $wb$ is a prefix of the complete return
449: word $vw$ and that it ends with $a$ if $aw$ is a suffix of $vw$.
450: \end{definition}
451:
452: A right special factor $w$ is called {\em maximal right special} if $w$
453: is not a proper suffix of any right special factor, i.e.,
454: $\sum_{a\in \El(w)}(\#\Er(aw)-1)=0$.
455: Any maximal right special factor is therefore weak bispecial.
456:
457: \begin{lemma}\label{lemma1}
458: If $w\in\mathcal L(u)$ is a maximal right special factor such that
459: for any $b\in \Er(w)$ there exists a unique $v\in\R(w)$ starting
460: with $b$, then $u$ is eventually periodic.
461: \end{lemma}
462:
463: \begin{proof}
464: Denote the return words of $w$ by $v_1,v_2,\ldots,v_r$, where,
465: w.l.o.g., $v_j$ starts with $b_j$, ends with $a_j$ and $b_{j+1}$ is
466: the only letter in $\Er(a_jw)$ for $1\le j<r$. Then $b_1$ is the
467: only letter in $\Er(a_rw)$ and $u=p(v_1v_2\cdots v_r)^\infty$ for
468: some prefix $p$.
469: \end{proof}
470:
471: \begin{corollary}\label{cormax}
472: If $u$ satisfies $R_2$, then it has no maximal right special factor.
473: \end{corollary}
474:
475: \begin{proof}
476: Assume that $w$ is a maximal right special factor. Then the two
477: return words of $w$ have different starting letters, hence $u$ is
478: eventually periodic by Lemma~\ref{lemma1} and $\#\R(wa)=1$.
479: \end{proof}
480:
481: On a binary alphabet, the notions ``weak bispecial'' and ``maximal
482: right special'' coincide.
483: Therefore Corollaries~\ref{ArnouxRauzy}, \ref{cormax} and
484: Lemma~\ref{l1} provide a short proof of the following theorem.
485:
486: \begin{thm}[Vuillon~\cite{Vuillon}]
487: An infinite word $u$ satisfies $R_2$ if and only if it is Sturmian.
488: \end{thm}
489:
490: For words with property $R_3$, we need the following lemma.
491:
492: \begin{lemma}\label{lemma2}
493: Let $w$ be a weak bispecial factor with a unique $a\in \El(w)$ such
494: that more than one return word of $w$ starts with a letter in
495: $\Er(aw)$, then $\#\R(aw)<\#\R(w)$.
496: \end{lemma}
497:
498: \begin{proof}
499: Any return word of $aw$ has the form $av_1v_2\cdots v_ra^{-1}$ for
500: some $r\ge1$ and $v_j\in\R(w)$, $1\le j\le r$. If $v_1$ ends with
501: $a$, then $r=1$. If $v_1$ ends with $a'\ne a$, then the assumption
502: of the lemma implies that there is a unique return word of $w$
503: starting with a letter in $\Er(a'w)$ (and $\#\Er(a'w)=1$).
504: Therefore $v_2$ and inductively the sequence of words
505: $v_2,\ldots,v_r$ are completely determined by the choice of $v_1$.
506: This implies that $\#\R(aw)$ equals the number of return words of
507: $w$ starting with a letter in $\#\Er(aw)$. Since $w$ is weak
508: bispecial, we have $\#\Er(aw)<\#\Er(w)$ and thus $\#\R(aw)<\#\R(w)$.
509: \end{proof}
510:
511: \noindent{\it Remark.}\
512: There are two cases for Lemma~\ref{lemma2}:
513: Either $aw$ is right special or there is more than one return word of
514: $w$ starting with the unique right extension of $aw$.
515:
516: \begin{corollary}\label{nomaximal}
517: If $u$ satisfies $R_3$, then it has no weak bispecial factor.
518: \end{corollary}
519:
520: \begin{proof}
521: Assume that $w$ is a weak bispecial factor.
522:
523: If $\#\Er(w)=3$, then every return word of $w$ starts with a
524: different letter in $\Er(w)$. Since at most for one $a\in \El(w)$,
525: the factor $aw$ is right special, we obtain a contradiction to $R_3$ by
526: Lemma~\ref{lemma1} or \ref{lemma2}.
527:
528: If $\#\Er(w)=2$, then $\Er(aw)=\{b\}$ and $\Er(a'w)=\{b'\}$.
529: Since, w.l.o.g., two return words of $w$ start with $b$ and one
530: starts with $b'$, we obtain a contradiction to $R_3$ by
531: Lemma~\ref{lemma2}.
532: \end{proof}
533:
534: By combining Corollary~\ref{nomaximal} and Theorem~\ref{th1}, we
535: obtain the following theorem.
536:
537: \begin{thm}
538: A uniformly recurrent word $u$ satisfies $R_3$ if and only if
539: $ C(n)=2n+1$ for all $n\ge0$ and $u$ has no weak bispecial factor.
540: \end{thm}
541:
542: \noindent{\it Remarks.}
543: \begin{itemize}
544: \item
545: The theorem remains true if ``weak bispecial'' is replaced by
546: ``maximal right special'':
547: If $\Delta C(n)=2$ for all $n\ge0$, then every factor $w$ with
548: $\#\Er(w)=3$ is the unique right special factor of its length, and it
549: cannot be weak bispecial.
550: If $\#\Er(w)=2$, then the two notions coincide.
551: \item
552: By symmetry, ``weak bispecial'' can be replaced by ``maximal left
553: special''.
554: \item
555: The condition on weak bispecial factors cannot be omitted.
556: Ferenczi~\cite{Fer} showed that the fixed point $\sigma^\infty(1)$
557: of the substitution given by
558: $\sigma:1\mapsto12,2\mapsto312,3\mapsto3312$, a recoding of the
559: Chacon substitution, has complexity $2n+1$ and it contains weak
560: bispecial factors.
561: \end{itemize}
562:
563: \section{Property $R_4$}\label{R_4}
564:
565: \subsection{A word with complexity $\ne3n+1$}\label{cex}
566: The following proposition shows that $C(n)$ need not be $(m-1)n+1$ for
567: all $n\ge0$ if $u$ satisfies $R_m$.
568:
569: \begin{proposition}
570: Define the substitution $\sigma$ by
571: \begin{align*}
572: \sigma: 1&\mapsto 13231 \\
573: 2&\mapsto 13231424131 \\
574: 3&\mapsto 42324131424 \\
575: 4&\mapsto 42324
576: \end{align*}
577: Then the fixed point $\sigma^\infty(1)$ satisfies $R_4$.
578: \end{proposition}
579:
580: \begin{proof}
581: By Section~\ref{sectbisp}, it is sufficient to consider bispecial
582: factors of $u=\sigma^\infty(1)$.
583: The factors of length $2$ are
584: $\mathcal L_2(u)=\{13,14,23,24,31,32,41,42\}$.
585: For the bispecial factors $1,2,23,2413$, the return words are easily
586: determined:
587: \begin{align*}
588: \R(1) & = \{13,1323,1424,142324\} \\
589: \R(2) & = \{23,2314,2413,241314\} \\
590: \R(23) & = \{2314,2314241314,232413,232413142413\} \\
591: \R(2413) & = \{241314,24131423,24132314,2413231423\}
592: \end{align*}
593:
594: The language of $u$ is closed under the morphism $\varphi$ defined by
595: $\varphi:1\leftrightarrow4,2\leftrightarrow3$, since
596: $\sigma\varphi(w)=\varphi\sigma(w)$ for all factors $w$.
597: Therefore we have $\R(\varphi(w))=\varphi(\R(w))$.
598:
599: The only factors of the form $a1b$, $a,b\in\mathcal A$ are $314$ and
600: $413$, hence $1$ is a weak bispecial factor, and $1,4$ are the only
601: bispecial factors with prefix or suffix $1$ or $4$.
602: Similarly, $23$ and $32$ are weak bispecial factors and no other
603: bispecial factor has prefix or suffix $23$ or $32$.
604:
605: The return words of the weak bispecial factor $2413142$ are factors of
606: $\sigma(v)$, with a factor $v$ of length $|v|\ge2$ having prefix $2$ or $3$,
607: suffix $2$ or $3$ and no other occurrence of $2$ and $3$.
608: Since the only possibilites for $v$ are $23,2413,32,3142$, we obtain
609: $$
610: \R(2413142)=
611: \{24131423,241314232413231423,24131424132314,241314241323142324132314\}.
612: $$
613:
614: All remaining bispecial factors $w$ have prefix $24132$ or $31423$ and
615: suffix $23142$ or $32413$, and therefore a decomposition
616: $w=t\,\sigma(v)\,t'$ with $t\in\{24,31\}$, $t'\in\{1323142,4232413\}$
617: and a unique bispecial factor $v$.
618: If $v$ is empty, then we have w.l.o.g. $w=241323142$ and
619: $$
620: \R(w)=\{2413231423,2413231423241314,2413231424131423,
621: 24132314241314232413242\}.
622: $$
623: If $v$ is not empty, then the uniqueness of $v$ implies that the set of
624: complete return words of $w$ is $t\,\sigma(\R(v)v)\,t'$.
625: Since $v$ is shorter than $w$, we obtain inductively that all bispecial
626: factors have exactly $4$ return words.
627: \end{proof}
628:
629: \subsection{Weak bispecial factors}
630:
631: The preceding example shows that weak bispecial factors cannot be
632: excluded in words $u$ satisfying $R_4$.
633: Nevertheless, we can show that the existence of a weak bispecial
634: factor imposes strong restrictions on the structure of the word $u$.
635:
636: \begin{lemma}
637: Let $w$ be a weak bispecial factor of a word $u$ satisfying $R_4$.
638: Then there exist factors $w_1,w_2\in\mathcal Aw\cup w\mathcal A$ and
639: $v_1,v_2,v_3,v_4$ such that
640: \begin{equation}\label{ret1234}
641: \R(w_1)=\{v_1v_3,v_1v_4,v_2v_3,v_2v_4\}\mbox{ and }
642: \R(w_2)=\{v_3v_1,v_3v_2,v_4v_1,v_4v_2\}.
643: \end{equation}
644: \end{lemma}
645:
646: \begin{proof}
647: Let $w$ be a weak bispecial factor. Since the problem is symmetric,
648: assume, w.l.o.g., $\#\Er(w)\ge\#\El(w)$. We have three different
649: situations:
650: \begin{itemize}
651: \item
652: $\#\Er(w)=2$: We have $\Er(aw)=1$ for all $a\in \El(w)$. For both
653: $b\in \Er(w)$, there must be more than one return word of $w$
654: starting with $b$ by Lemma~\ref{lemma2}. Therefore, there exists two
655: return words of $w$ starting with each $b$. Let $w_1=wb_1$,
656: $w_2=wb_2$.
657: \item
658: $\#\Er(w)=3$: There exists a unique $b\in \Er(w)$ such that two
659: return words of $w$ start with~$b$. By Lemma~\ref{lemma2}, there
660: exists therefore some $a\in \El(w)$ such that $\#\Er(aw)\ge2$.
661: Since $w$ is weak bispecial with $\#\Er(w)=3$, we have $\#\Er(aw)=2$
662: and $\Er(a'w)=1$ for all other $a'\in \El(w)$.
663: Let $w_1=aw$, $w_2=wb$.
664: \item
665: $\#\Er(w)=4$: For every $b\in \Er(w)$, there is a unique return
666: word of $w$ starting with~$b$. By Lemma~\ref{lemma2}, we have
667: $a_1,a_2\in \El(w)$ with $\#\Er(a_iw)=2$. Let $w_1=a_1w$,
668: $w_2=a_2w$.
669: \end{itemize}
670:
671: Consider ``complete return words of the set $\{w_1,w_2\}$'': words
672: which have either $w_1$ or $w_2$ as prefix, either $w_1$ or $w_2$ as
673: suffix, and no other occurrence of $w_1$ and $w_2$. By the
674: definitions of $w_1$ and $w_2$, there are exactly two such words
675: $v_1w_{i_1},v_2w_{i_2}$ with prefix $w_1$ and two words
676: $v_3w_{i_3},v_4w_{i_4}$ with prefix $w_2$.
677:
678: If $i_1=i_2=2$ and $i_3=i_4=1$, then $R_4$ implies that (\ref{ret1234})
679: holds.
680:
681: If $i_1=i_2=1$, then $w_1$ has only the two return words $v_1,v_2$.
682: If $i_2=i_3=i_4=1$, then the return words of $w_1$ are
683: $v_1v_3,v_1v_4,v_2$.
684: Similarly, $i_3=i_4=2$ and $i_1=i_2=i_3=2$ are not possible.
685:
686: The only remaining case is $i_1=i_4=1$, $i_2=i_3=2$.
687: Then the return words of $w_1$ are $v_1$ and $v_2v_3^{r_i}v_4$,
688: $i\in\{1,2,3\}$, $0\le r_1<r_2<r_3$. The return words of $w_2$ are
689: $v_3$ and $v_4v_1^{s_i}v_2$, $i\in\{1,2,3\}$, $0\le s_1<s_2<s_3$.
690:
691: The return words of $v_2w_2$ are therefore of the form
692: $v_2v_3^{r_i}v_4v_1^{s_j}$. Let $S_1$ be the set of these $4$ pairs
693: $(r_i,s_j)$. Similarly, let $S_2$ be the set of the $4$ pairs
694: $(s_i,r_j)$ such that $v_4v_1^{s_i}v_2v_3^{r_j}$ is a return word of
695: $v_4w_1$.
696:
697: We show that there must be some $i\in\{1,2,3\}$ such that
698: $(r_i,s_2)\in S_1$ and $(r_i,s_3)\in S_1$, by considering the return
699: words of $v_1^{s_2}w_1$ and of $v_1^{s_2}v_2w_2$. The return words
700: of $v_1^{s_2}v_2w_2$ are of the form
701: $v_1^{s_2}v_2tv_3^{r_i}v_4v_1^{s_j-s_2}$ with
702: $t\in(v_3^*v_4v_1^{s_1}v_2)^*$, $i\in\{1,2,3\}$ and $j\in\{2,3\}$.
703: For these $t$ and $r_i$, $v_1^{s_2}v_2tv_3^{r_i}v_4$ is a return
704: word of $v_1^{s_2}w_1$. If there was no $r_i$ with $(r_i,s_2)\in
705: S_1$ and $(r_i,s_3)\in S_1$, then these words would provide $4$
706: different return words of $v_1^{s_2}w_1$, wich contradicts $R_4$
707: since $v_1$ is another return word.
708:
709: Similarly, we must have some $i\in\{1,2,3\}$ such that $(s_i,r_2)\in
710: S_2$ and $(s_i,r_3)\in S_2$. By considering the return words of
711: $v_1^{s_2}w_1$ and $v_4v_1^{s_2}w_1$, we obtain as well the
712: existence of some $i\in\{1,2,3\}$ such that $(r_2,s_i)\in S_1$ and
713: $(r_3,s_i)\in S_1$. Finally, we must also have some $i\in\{1,2,3\}$
714: such that $(s_2,r_i)\in S_2$ and $(s_3,r_i)\in S_2$.
715:
716: Putting everything together, we have two possibilities for $S_1$.
717: Either it contains $(r_1,s_1)$ and no other $(r_i,s_j)$ with $i=1$
718: or $j=1$, or $S_1=\{(r_1,s_2),(r_1,s_3),(r_2,s_1),(r_3,s_1)\}$.
719: Analogously, $S_2$ contains $(s_1,r_1)$ and no other $(s_i,r_j)$
720: with $i=1$ or $j=1$, or
721: $S_2=\{(s_1,r_2),(s_1,r_3),(s_2,r_1),(s_3,r_1)\}$.
722:
723: If $(r_1,s_1)\in S_1$ and $(s_1,r_1)\in S_2$, then
724: $v_2v_3^{r_1}v_4w_1$ has only one return word,
725: $v_2v_3^{r_1}v_4v_1^{s_1}$. If $(r_1,s_1)\not\in S_1$ and
726: $(s_1,r_1)\not\in S_2$, then $v_2v_3^{r_1}v_4w_1$ has only two
727: return words, $v_2v_3^{r_1}v_4v_1^{s_2}$ and
728: $v_2v_3^{r_1}v_4v_1^{s_3}$. If $(r_1,s_1)\in S_1$ and
729: $(s_1,r_1)\not\in S_2$, then the return words of
730: $v_2v_3^{r_1}v_4w_1$ are of the form
731: $v_2v_3^{r_1}v_4v_1^{s_1}v_2v_3^{r_i}v_4v_1^{s_j}$ with
732: $(r_i,s_j)\in S_1\setminus\{(r_1,s_1)\}$, thus there are only three
733: words. Similarly, $v_4v_1^{s_1}v_2w_2$ has only three return words
734: if $(r_1,s_1)\not\in S_1$ and $(s_1,r_1)\in S_2$.
735:
736: This shows that $i_1=i_4=1$, $i_2=i_3=2$ is impossible, and
737: the lemma is proved.
738: \end{proof}
739:
740: \section{Words associated with $\beta$-integers}\label{beta}
741:
742: In this section, we describe a new class of infinite words with
743: the property $R_m$. The language of these words is not necessarily
744: closed under reversal.
745:
746: Consider the fixed point $u=\sigma^\infty(0)$ of a primitive
747: substitution of the form
748: \begin{equation}\label{subst}
749: \begin{array}{rcl}\sigma:\qquad 0 & \mapsto & 0^{t_1}1 \\
750: 1 & \mapsto & 0^{t_2}2 \\ & \vdots \\
751: m-2 & \mapsto & 0^{t_{m-1}}(m-1) \\ m-1 & \mapsto & 0^{t_m}\end{array}
752: \end{equation}
753: for some integers $m\ge2$, $t_1,t_m\ge1$ and $t_2,\ldots,t_{m-1}\ge0$.
754: The incidence matrix of $\sigma$ is a companion matrix of the
755: polynomial $x^m-t_1x^{m-1}-\cdots-t_m$.
756: Let $\beta>1$ be the dominant root of this polynomial (the
757: Perron-Frobenius eigenvalue of the matrix).
758: If
759: $$
760: t_j\cdots t_m\prec t_1\cdots t_m\quad\mbox{for all }j\in\{2,\ldots,m\},
761: $$
762: where $\preceq$ denotes the lexicographic ordering, then $\sigma$ is a
763: {\em $\beta$-substitution} and $\beta$ is a simple Parry number.
764: It is easy to see that $u$ codes in this case the sequence of
765: distances between consecutive nonnegative $\beta$-integers
766: $$
767: \mathbb Z_\beta^+=\Big\{\sum_{j=0}^J x_j\beta^j\mid J\ge0,\,
768: x_j\in\mathbb Z,\, x_j\ge0,\, x_j\cdots x_0\prec t_1\cdots t_m
769: \text{ for all }j,\,0\le j\le J\Big\},
770: $$
771: and a letter $k$ corresponds to the distance
772: $t_{k+1}/\beta+\cdots+t_m/\beta^{m-k}$.
773: ($0$~corresponds to distance~$1$.)
774:
775: \medskip\noindent{\em Remark.}
776: The most prominent example of a $\beta$-substitution is the Fibonacci
777: substitution ($m=2$, $t_1=t_2=1$), where $\beta$ is the golden mean.
778:
779: \medskip
780: It is not difficult to show that all prefixes of $u$ are left special
781: factors, with all $m$ letters being left extensions (see
782: e.g. Frougny, Mas\'akov\'a and Pelantov\'a~\cite{FrMaPe}).
783: For every factor $w$, the tree of return words constructed by the left
784: extensions (see Section~\ref{tree}) contains therefore a node with $m$
785: children, the shortest prefix of $u$ having $w$ as suffix.
786: The word $u$ is uniformly recurrent since all fixed points of primitive
787: substitutions have this property (Queff\'elec~\cite{Queffelec}).
788: Therefore every factor $w$ has at least $m$ return words.
789: If there exists a left special factor which is not a prefix of $u$,
790: then this factor has more than $m$ return words.
791: By Proposition~\ref{onespecial}, we obtain the following proposition.
792:
793: \begin{proposition}
794: If $u=\sigma^\infty(0)$ for some substitution $\sigma$ of the form
795: (\ref{subst}), then it satisfies $R_m$ if and only if
796: $ C(n)=(m-1)n+1$ for all $n\ge0$.
797: \end{proposition}
798:
799: Bernat, Mas\'akov\'a and Pelantov\'a~\cite{BeMaPe} characterized the
800: fixed points of $\beta$-substitutions satisfying $\Delta C(n)=m-1$ for
801: all $n\ge0$.
802: The techniques of their proof can also be used to construct non-prefix
803: left special factors if $\sigma$ is a substitution of the form
804: (\ref{subst}) which is not a $\beta$-substitution, and their conditions
805: can be reformulated as in the following corollary.
806:
807: \begin{corollary}\label{affineRm}
808: If $u=\sigma^\infty(0)$ for some substitution $\sigma$ of the form
809: (\ref{subst}), then it has the property $R_m$ if and only if
810: \begin{itemize}
811: \item $t_m=1$ and
812: \item $t_j\cdots t_{m-1}t_1\cdots t_{j-1}\preceq t_1\cdots t_{m-1}$
813: for all $j\in\{2,\ldots,m-1\}$.
814: \end{itemize}
815: \end{corollary}
816:
817: Note that the language of $u$ is closed under reversal if and only if
818: $t_1=t_2=\cdots=t_{m-1}$.
819: In this case, $u$ is an Arnoux-Rauzy word of order $m$.
820:
821: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
822: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
823:
824: \section*{Acknowledgements}
825:
826: The authors acknowledge financial support by the Czech Science
827: Foundation GA \v{C}R 201/05/0169 and by the grant LC06002 of the
828: Ministry of Education, Youth and Sports of the Czech Republic.
829:
830: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
831: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
832: \begin{thebibliography}{88}
833: \addcontentsline{toc}{section}{Bibliography}
834:
835: %\bibitem{AlSh} J.-P. Allouche, J. Shallit, {\em The ubiquitous Prouhet-Thue-Morse sequence}, Sequences and their applications (Singapore, 1998), 1--16, Springer Ser. Discrete Math. Theor. Comput. Sci., Springer, London (1999)
836:
837: \bibitem{ArBr} I. M. Ara\'ujo, V. Bruy\`ere, {\em Words derivated from
838: Sturmian words}, Theor. Comput. Sci. {\bf 340} (2005), 204--219.
839:
840: \bibitem{BeMaPe} J. Bernat, Z. Mas\'akov\'a, E. Pelantov\'a, {\em On
841: a~class of infinite words with affine factor complexity}, to appear in
842: Theor. Comput. Sci.
843:
844: \bibitem{Berstel} J. Berstel, {\em Recent results on extensions of
845: Sturmian words}, Int. J. Algebra Comput. {\bf 12} (2002), 371--385.
846:
847: %\bibitem{Cassaigne} J. Cassaigne, {\em An algorithm to test if a~given circular HD0L-language avoids a~pattern}, in: IFIP World Computer Congress'94 {\bf 1}, 459--474, Elsevier (North Holland)
848:
849: \bibitem{Cass} J. Cassaigne, {\em Complexit\'e et facteurs sp\'eciaux},
850: Bull. Belg. Math. Soc. Simon Stevin {\bf 4} (1997), 67--88.
851:
852: %\bibitem{Cass2} J. Cassaigne, {\em Recurrence in infinite words}, Proceedings of the 18th Symposium on Theoretical Aspects of Computer Science (STACS 2001), Lect. Notes Comput. Sci. {\bf 2010}, Springer, Berlin, 2001, 1--11.
853:
854: %\bibitem{DuThese} F. Durand, {\em Contributions \`a l' \'etude des suites et syst\`emes dynamiques substitutifs}, Th\`ese, Universit\'e de la M\'editerran\'ee, Aix-Marseille II. (1996)
855:
856: \bibitem{Durand} F. Durand, {\em A~characterization of substitutive
857: sequences using return words}, Discrete Math. {\bf 179} (1998), 89--101
858:
859: %\bibitem{Fabre} S. Fabre, {\em Substitutions et $\beta$-syst\`emes de num\'eration}, Theor. Comput. Sci. {\bf 137} (1995), 219--236
860:
861: \bibitem{Fer} S. Ferenczi, {\em Les transformations de Chacon:
862: combinatoire, structure g\'eom\'etrique, lien avec les syst\`emes de
863: complexit\'e $2n+1$}, Bull. Soc. Math. Fr. {\bf 123} (1995),
864: 271--292.
865:
866: \bibitem{FeHoZa} S. Ferenczi, Ch. Holton, L. Q. Zamboni, {\em Structure
867: of three interval exchange transformations. II. A combinatorial
868: description of the trajectories.}, J. Anal. Math. {\bf 89} (2003),
869: 239--276.
870:
871: \bibitem{FeMaNo} S. Ferenczi, C. Maudit, A. Nogueira, {\em Substitution
872: dynamical systems: algebraic characterization of eigenvalues}, Ann.
873: Sci. \'Ec. Norm. Sup\'er. {\bf 29} (1996), 519--533.
874:
875: %\bibitem{Frid} A. Frid, {\em Applying a uniform marked morphism to a word}, Discrete Math. Theor. Comput. Sci. {\bf 3} (1999), 125--140
876:
877: \bibitem{FrMaPe} Ch. Frougny, Z. Mas\'akov\'a, E. Pelantov\'a,
878: {\em Complexity of infinite words associated with beta-expansions},
879: Theor. Inform. Appl. {\bf 38} (2004), 163--185; Corrigendum, Theor.
880: Inform. Appl. {\bf 38} (2004), 269--271.
881:
882: %\bibitem {Justin} J. Justin, G. Pirillo, {\em On a combinatorial property of Sturmian words}, Theor. Comput. Sci. {\bf 154} (1996), 387--394
883:
884: \bibitem{JusVui} J. Justin, L. Vuillon, {\em Return words in Sturmian
885: and episturmian words}, Theor. Inform. Appl. {\bf 34} (2000), 343--356.
886:
887: %\bibitem{MiSe} F. Mignosi, P. S\'eebold, {\em If a~D0L-language is $k$-power-free then it is circular}, in: A. Lingas, R. Karlsson, S.~Carlsson (editors), ICALP'93, Lect. Notes Comp. Sci. {\bf 700}, 507--518, Springer
888:
889: \bibitem{MoHe} M. Morse, G. A. Hedlund, {\em Symbolic dynamics II.
890: Sturmian trajectories}, Amer. J. Math. {\bf 62} (1940), 1--42.
891:
892: %\bibitem{Parry} W. Parry, {\em On the {$\beta$}-expansions of real numbers}, Acta Math. Acad. Sci. Hung. {\bf 11} (1960), 401--416
893:
894: %\bibitem{Renyi} A. R\'enyi, {\em Representations for real numbers and their ergodic properties,} Acta Math. Acad. Sci. Hung. {\bf 8} (1957), 477--493
895:
896: \bibitem{Queffelec} M. Queff\'elec, {\em Substitution Dynamical Systems
897: -- Spectral Analysis}, Lecture Notes in Math. {\bf 1294}, Springer,
898: Berlin, 1987.
899:
900: %\bibitem{Thurston} W. P. Thurston, {\em Groups, tilings, and finite state automata}, AMS Colloquium lectures, 1989
901: %Geometry supercomputer project research report GCG1, University of Minnesota (1989)
902:
903: \bibitem{Vuillon} L. Vuillon, {\em A~characterization of Sturmian words
904: by return words}, Eur. J. Comb. {\bf 22} (2001), 263--275.
905:
906: \bibitem{VuIET} L. Vuillon, {\em On the number of return words in
907: infinite words with complexity $2n+1$}, LIAFA Research Report 2000/15.
908: \end{thebibliography}
909:
910: \end{document}
911: