1: %\nonstopmode
2: \magnification=1200
3:
4: \def\og{\leavevmode\raise.3ex\hbox{$\scriptscriptstyle
5: \langle\!\langle\,$}}
6: \def \fg {\leavevmode\raise.3ex\hbox{$\scriptscriptstyle
7: \rangle\!\rangle\,\,$}}
8:
9:
10: \def\Q{{\bf {Q}}}
11: \def\K{{\bf {K}}}
12: \def\O{ {O_{\K}}} \def\U{{U_K}}
13: \def\N{{\bf N}} \def\L{{\bf L}}
14: \def\Z{{\bf Z}} \def\A{{\bf A}} \def\tA{{{\tilde\A}}} \def\tN{{{\tilde N}}}
15: \def\R{{\bf R}} \def\M{{\bf M}} \def\cK{{\bf B}} \def\tw{{\tilde w}}
16: \def\tcK{{{\tilde\cK}}} \def\U{{\bf U}} \def\tU{{{\tilde\U}}}
17: \def\tPsi{{\tilde \Psi}}
18: \def\C{{\bf C}} \def\ord{{\rm ord}}
19: \def\P{{\cal {P}}}
20: \def\cK{{\cal {K}}} \def\cC{{\cal {C}}} \def\cS{{\cal {S}}}
21: \def\cH{{\cal {H}}} \def\cJ{{\cal {J}}} \def\cV{{\cal {V}}}
22: \def\cA{{\cal {A}}} \def\cB{{\cal {B}}} \def\cN{{\cal {N}}}
23: \def\cE{{\cal {E}}} \def\cHI{{\cal {H}}_{\infty}}
24: \def\card{{\rm card}}
25: \def\x{{\underline x}} \def\si{{\sigma}}
26: \def\vep{{\varepsilon}} \def\dzeta{{\zeta}}
27:
28: \input typpo
29:
30: %
31: \catcode`@=11
32: \def\bibliographym@rk{\bgroup}
33: %
34: % \bye est modifie pour la biblio et la table des matieres
35: %
36: \outer\def\bye{ \par\vfill\supereject\end}
37:
38:
39: \def\dem{\noindent {\bf{D\'emonstration : }}}
40: \def\rem{\noindent {\bf{Remarque : }}}
41: \def\cf{{\it cf. }} \def\resp{{\it resp. }}
42: \def\thi{{\thinspace}} \def\diam{{\rm diam}}
43:
44:
45:
46: \def\house#1{\setbox1=\hbox{$\,#1\,$}%
47: \dimen1=\ht1 \advance\dimen1 by 2pt \dimen2=\dp1 \advance\dimen2 by 2pt
48: \setbox1=\hbox{\vrule height\dimen1 depth\dimen2\box1\vrule}%
49: \setbox1=\vbox{\hrule\box1}%
50: \advance\dimen1 by .4pt \ht1=\dimen1
51: \advance\dimen2 by .4pt \dp1=\dimen2 \box1\relax}
52:
53: \def\H{{\rm H}}
54: \def\MM{{\rm M}} \def\h{{\rm h}} \def\J{{\rm J}}
55: \def\Norm{{\rm Norm}} \def\de{\delta _{\K}} \def\res{{\rm Res}}
56: \def\NK{{\rm N}_{\K / \Q}} \def\NL{{\rm N}_{\L / \Q}}
57: \def\NS{{\rm N}_{\K(s) / \Q}} \def\NLL{{\rm N}_{\L(e_1) / \Q}}
58: \def\Df{{\Delta_f}} \def\NLK{{\rm N}_{\L / \K}}
59:
60: \def\Ga{{\Gamma}} \def\al{{\alpha}}
61: \def\ka{{\kappa}} \def\eps{{\varepsilon}}
62: \def\ga{{\gamma}} \def\La{{\Lambda}}
63: \def\be{{\beta}} \def\de{{\delta}}
64: \def\la{{\lambda}} \def\th{{\theta}}
65:
66: \def\cH{{\cal H}}
67: \def\uv{{\underline{v}}} \def\ux{{\underline{x}}} \def\uy{{\underline y}}
68:
69: \def\disp{\displaystyle} \def\scr{\scriptstyle}
70: \def\sm{\smallskip} \def\ens{\enspace} \def\noi{\noindent}
71:
72: \def\build#1_#2^#3{\mathrel{\mathop{\kern 0pt#1}\limits_{#2}^{#3}}}
73:
74: \def\date {le\ {\the\day}\ \ifcase\month\or janvier
75: \or fevrier\or mars\or avril\or mai\or juin\or juillet\or
76: ao\^ut\or septembre\or octobre\or novembre
77: \or d\'ecembre\fi\ {\oldstyle\the\year}}
78:
79: \font\fivegoth=eufm5 \font\sevengoth=eufm7 \font\tengoth=eufm10
80:
81: \newfam\gothfam \scriptscriptfont\gothfam=\fivegoth
82: \textfont\gothfam=\tengoth \scriptfont\gothfam=\sevengoth
83: \def\goth{\fam\gothfam\tengoth}
84:
85: \def\pp{{\goth p}} \def\aa{{\goth a}} \def\bb{{\goth b}}
86: \def\cc{{\goth c}} \def\qq{{\goth q}} \def\PP{{\goth P}}
87:
88: %\def\cqfd{\unskip\kern 6pt\penalty 500 \raise 0pt
89: % \hbox{\vrule\vbox to6pt{\hrule width 6pt \vfill\hrule}\vrule}\par}
90:
91: \def\pro{\noindent {\bf Proof : }}
92:
93: \def\smallsquare{\vbox{\hrule\hbox{\vrule height 1 ex\kern 1 ex\vrule}\hrule}}
94: \def\cqfd{\hfill \smallsquare\vskip 3mm}
95:
96:
97: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
98:
99: \centerline{}
100:
101: \vskip 4mm
102:
103:
104:
105:
106: \centerline{
107: \bf Continued fractions and transcendental numbers}
108:
109: \vskip 8mm
110: \centerline{Boris A{\sevenrm DAMCZEWSKI},
111: Yann B{\sevenrm UGEAUD}, and Les D{\sevenrm AVISON}}
112:
113:
114: \vskip 6mm
115:
116: \vskip 8mm
117:
118: \centerline{\bf 1. Introduction}
119:
120: \vskip 6mm
121:
122: It is widely believed that the
123: continued fraction expansion of every irrational algebraic number
124: $\alpha$ either is eventually periodic
125: (and we know that this is the case if and only if $\alpha$ is a
126: quadratic irrational), or it contains arbitrarily large partial quotients.
127: Apparently, this question was first considered by
128: Khintchine in \cite{Khintchine}
129: (see also \cite{Allouche,Shallit_survey,Waldschmidt}
130: for surveys including a discussion on this subject).
131: A preliminary step towards its resolution consists in providing explicit
132: examples of transcendental continued fractions.
133:
134: The first result of this type goes back to the
135: pioneering work of
136: Liouville \cite{Liouville}, who constructed transcendental
137: real numbers with a very fast growing sequence of partial quotients.
138: Indeed, the so-called `Liouville inequality' implies the
139: transcendence of real numbers with very large partial quotients.
140: Replacing it by Roth's theorem yields refined results,
141: as shown by Davenport and Roth \cite{DavRoth}. In \cite{AdBuMB},
142: the argument of Davenport and Roth is slightly improved and Roth's
143: theorem is replaced by a more recent result of Evertse \cite{Evertse97}
144: to obtain the best known result of this type. Note that the constant $e$,
145: whose continued fraction expansion is given by
146: $$
147: e=[2;1,2,1,1,4,1,1,6,1,1,8,1,1,10,\ldots, 1, 1, 2n, 1, 1, \ldots],
148: $$
149: (see for instance \cite{Lang}) provides an explicit example of a transcendental number
150: with unbounded partial quotients; however, its transcendence does not
151: follow from the criteria from \cite{Liouville,DavRoth,AdBuMB}.
152:
153:
154:
155: At the opposite side, there is a quest for finding
156: explicit examples of transcendental continued fractions
157: with bounded partial quotients. The first examples of such
158: continued fractions were found by Maillet \cite{Maillet}
159: (see also Section 34 of Perron \cite{Perron}).
160: The proof of Maillet's results is based on a general form
161: of the Liouville inequality
162: which limits the approximation of algebraic numbers by quadratic irrationals.
163: They were subsequently improved upon
164: by A. Baker \cite{Baker62,Baker64}, who used
165: the Roth theorem for number fields obtained by LeVeque \cite{Leveque56}.
166: Later on, Davison \cite{Dav89} applied a result of W. M. Schmidt \cite{Schm67},
167: saying that a real algebraic number cannot
168: be too well approximable by quadratic numbers, to show the transcendence of
169: some specific continued fractions (see Section 4). With the
170: same auxiliary tool, M. Queff\'elec \cite{Queffelec98}
171: established the nice result that the
172: Thue--Morse continued fraction is transcendental (see Section 5). This
173: method has then been made more explicit, and combinatorial
174: transcendence criteria based on Davison's approach were given in
175: \cite{ADQZ,Dav,Bax03,LiSt}.
176:
177:
178: \medskip
179:
180:
181: Recently, Adamczewski and Bugeaud \cite{AdBuAM} obtained two new
182: combinatorial transcendence criteria for continued fractions
183: that are recalled in Section 2.
184: The main novelty in their approach is the use of a stronger Diophantine result
185: of W. M. Schmidt \cite{Schmidt72a,Schmidt80}, commonly known as
186: the Subspace Theorem. After some work, it
187: yields considerable improvements upon the
188: criteria from \cite{ADQZ,Dav,Bax03,LiSt}.
189: These allowed them to prove that the continued fraction
190: expansion of every real algebraic number of degree at least three
191: cannot be `too simple', in various senses.
192: It is the purpose of the present work to give further
193: applications of their criteria to well-known (families of) continued
194: fractions. In addition, we provide a slight refinement of their
195: second criterion.
196: Note also that a significative improvement of
197: the results of Maillet and of Baker mentioned above is
198: obtained in \cite{AdBuMB} thanks to a similar use of the Subspace Theorem.
199:
200: \medskip
201:
202: The present paper is organized as follows. In Section 2,
203: we state the two main transcendence criteria of \cite{AdBuAM},
204: namely Theorems A and B. Then, Section 3 is devoted to a slight
205: sharpening of Theorem B. In the subsequent Sections, we give various
206: examples of applications of these transcendence criteria.
207: We first solve in Section 4 a problem originally tackled by
208: Davison in \cite{Dav89} and later considered by several
209: authors in \cite{ADQZ,Dav,Bax03}. The
210: Rudin--Shapiro and the Baum--Sweet continued fractions are
211: proved to be transcendental in Section 5.
212: Then, Sections 6 and 7 are respectively devoted to folded continued
213: fractions and continued fractions arising from perturbed
214: symmetries (these last sequences were introduced
215: by Mend\`es France \cite{Mendes}).
216: In the last Section, we show through the study of another family of
217: continued fractions that, in some cases, rather than
218: applying roughly Theorem B, it is much better to go into its
219: proof and to evaluate continuants carefully.
220:
221:
222: \vskip 8mm
223:
224: \centerline{\bf 2. The transcendence criteria}
225:
226: \vskip 6mm
227:
228: In this Section, we recall the transcendence criteria,
229: namely Theorems A and B, obtained by the first two authors in \cite{AdBuAM}.
230:
231: \medskip
232:
233: Before stating the new criteria, we need to introduce some notation.
234: Let ${\cal A}$ be a given set, not necessarily finite.
235: The length of a word
236: $W$ on the alphabet ${\cal A}$, that is, the number of letters
237: composing $W$, is denoted by $\vert W\vert$.
238: For any positive integer $k$, we write
239: $W^k$ for the word $W\ldots W$ ($k$ times repeated concatenation
240: of the word $W$). More generally, for any positive rational number
241: $x$, we denote by $W^x$ the word
242: $W^{[x]}W'$, where $W'$ is the prefix of
243: $W$ of length $\left\lceil(x- [x])\vert W\vert\right\rceil$.
244: Here, and in all what follows, $[y]$ and
245: $\lceil y\rceil$ denote, respectively, the integer part and the upper
246: integer part of the real number $y$.
247: Let ${\bf a}=(a_\ell)_{\ell \ge 1}$
248: be a sequence of elements from ${\cal A}$,
249: that we identify with the infinite word $a_1 a_2 \ldots a_\ell \ldots$.
250: Let $w$ be a rational number with $w>1$.
251: We say that ${\bf a}$
252: satisfies Condition $(*)_w$ if ${\bf a}$ is not
253: eventually periodic and if there exists
254: a sequence of finite words $(V_n)_{n \ge 1}$ such that:
255:
256: \medskip
257:
258:
259: \item{\rm (i)} For any $n \ge 1$, the word $V_n^w$ is a prefix
260: of the word ${\bf a}$;
261:
262:
263: \smallskip
264:
265: \item{\rm (ii)} The sequence $(\vert V_n\vert)_{n \ge 1}$ is
266: increasing.
267:
268: \medskip
269:
270:
271: Roughly speaking, ${\bf a}$ satisfies Condition $(*)_w$
272: if ${\bf a}$ is not eventually periodic and if there exist infinitely
273: many `non-trivial' repetitions (the size of which is measured
274: by $w$) at the beginning of the infinite word $a_1 a_2 \ldots a_\ell \ldots$
275:
276: A first transcendence criterion for `purely' stammering continued
277: fractions is given in \cite{AdBuAM}.
278:
279:
280: \proclaim Theorem A.
281: Let ${\bf a}=(a_\ell)_{\ell \ge 1}$ be a sequence of positive integers.
282: Let $(p_\ell/q_\ell)_{\ell \ge 1}$ denote the sequence of convergents to
283: the real number
284: $$
285: \alpha:= [0; a_1, a_2, \ldots, a_\ell,\ldots].
286: $$
287: If there exists a rational number $w \ge 2$ such that
288: ${\bf a}$ satisfies Condition $(*)_w$, then $\alpha$ is transcendental.
289: If there exists a rational number $w>1$ such that
290: ${\bf a}$ satisfies Condition $(*)_w$, and if
291: the sequence $(q_\ell^{1/\ell})_{\ell \ge 1}$ is bounded
292: (which is in particular the case when the sequence ${\bf a}$ is bounded),
293: then $\alpha$ is transcendental.
294:
295:
296:
297: Unfortunately, in the statement of Theorem A,
298: the repetitions must appear at the very beginning of ${\bf a}$.
299: When this is not the case, but the repetitions occur not
300: too far from the beginning of ${\bf a}$, then we have another
301: criterion.
302:
303: Keep the above notation.
304: Let $w$ and $w'$ be non-negative rational numbers with $w>1$.
305: We say that ${\bf a}$
306: satisfies Condition $(**)_{w, w'}$ if ${\bf a}$ is not
307: eventually periodic and if there exist
308: two sequences of finite words $(U_n)_{n \ge 1}$,
309: $(V_n)_{n \ge 1}$ such that:
310:
311: \medskip
312:
313:
314: \item{\rm (i)} For any $n \ge 1$, the word $U_nV_n^w$ is a prefix
315: of the word ${\bf a}$;
316:
317: \smallskip
318:
319: \item{\rm (ii)} The sequence
320: $({\vert U_n\vert} / {\vert V_n\vert})_{n \ge 1}$ is bounded
321: from above by $w'$;
322:
323: \smallskip
324:
325: \item{\rm (iii)} The sequence $(\vert V_n\vert)_{n \ge 1}$ is
326: increasing.
327:
328: \medskip
329:
330: We are now ready to present a transcendence criterion for (general)
331: stammering continued fractions, as stated in \cite{AdBuAM}.
332:
333:
334:
335: \proclaim Theorem B.
336: Let ${\bf a}=(a_\ell)_{\ell \ge 1}$ be a sequence of positive integers.
337: Let $(p_\ell/q_\ell)_{\ell \ge 1}$ denote the sequence of convergents to
338: the real number
339: $$
340: \alpha:= [0; a_1, a_2, \ldots,a_\ell,\ldots].
341: $$
342: Assume that the sequence $(q_\ell^{1/\ell})_{\ell \ge 1}$ is bounded
343: and set $M = \limsup_{\ell \to + \infty} \, q_{\ell}^{1/\ell}$ and
344: $m = \liminf_{\ell \to + \infty} \, q_{\ell}^{1/\ell}$.
345: Let $w$ and $w'$ be non-negative real numbers with
346: $$
347: w > w' \biggl( 2 {\log M \over \log m} - 1 \biggr)
348: + {\log M \over \log m}. \eqno (2.1)
349: $$
350: If ${\bf a}$ satisfies Condition $(**)_{w, w'}$,
351: then $\alpha$ is transcendental.
352:
353: It turns out that condition (2.1) can be slightly
354: weakened by means of a careful consideration of continuants.
355: This is the purpose of Section 3.
356:
357:
358:
359:
360: \vskip 6mm
361:
362: \centerline{\bf 3. A slight sharpening of Theorem B}
363:
364: \vskip 8mm
365:
366: In this Section,
367: we establish the following improvement of Theorem B.
368:
369:
370: \proclaim Theorem 3.1.
371: Let ${\bf a}=(a_\ell)_{\ell \ge 1}$ be a sequence of positive integers.
372: Let $(p_\ell/q_\ell)_{\ell \ge 1}$ denote the sequence of convergents to
373: the real number
374: $$
375: \alpha:= [0; a_1, a_2, \ldots,a_\ell,\ldots].
376: $$
377: Assume that the sequence $(q_\ell^{1/\ell})_{\ell \ge 1}$ is bounded
378: and set $M = \limsup_{\ell \to + \infty} \, q_{\ell}^{1/\ell}$ and
379: $m = \liminf_{\ell \to + \infty} \, q_{\ell}^{1/\ell}$.
380: Let $w$ and $w'$ be non-negative real numbers with
381: $$
382: w > w' \biggl( 2 {\log M \over \log m} - 1 \biggr) + 1. \eqno (3.1)
383: $$
384: If ${\bf a}$ satisfies Condition $(**)_{w, w'}$,
385: then $\alpha$ is transcendental.
386:
387: We note that the right hand side of (3.1) is always smaller than or
388: equal to the right hand side of (2.1).
389:
390: Before proceeding with the proof, we recall some useful facts
391: on continuants. For positive integers $a_1, \ldots, a_m$, denote
392: by $K_m (a_1, \ldots, a_m)$ the denominator of the rational number
393: $[0; a_1, \ldots, a_m]$. It is commonly called a {\it continuant}.
394:
395: \medskip
396:
397: \proclaim Lemma 3.2.
398: For any positive integers $a_1, \ldots, a_m$ and any integer $k$ with
399: $1 \le k \le m-1$, we have
400: $$
401: K_m (a_1, \ldots , a_m) = K_m (a_m, \ldots, a_1)
402: $$
403: and
404: $$
405: \eqalign{
406: K_k (a_1, \ldots, a_k) \cdot K_{m-k} (a_{k+1}, \ldots, a_m)
407: & \le K_m (a_1, \ldots , a_m) \cr
408: & \le 2 \, K_k (a_1, \ldots, a_k) \cdot K_{m-k} (a_{k+1}, \ldots, a_m). \cr}
409: $$
410:
411: \bigskip
412:
413: \noindent {\bf Proof of Theorem 3.1.}
414: We follow step by step the proof of Theorem 2 from \cite{AdBuAM},
415: with a single modification. We keep the notation from \cite{AdBuAM}.
416: Recall that, for any $n \ge 1$, we set
417: $r_n = |U_n|$ and $s_n = |V_n|$.
418:
419: Let $n$ be a positive integer. Let $\delta > 0$ be a (small)
420: real number. Since $w>1$ and $r_n \le w' s_n$, we get
421: $$
422: {2 r_n + \delta s_n \over r_n + (w-1) s_n} \le {2 w' s_n + \delta s_n
423: \over w' s_n + (w-1) s_n}
424: = {2 w' + \delta \over w' + w - 1} < {\log m \over \log M},
425: $$
426: by (3.1), if $\delta$ is sufficiently small. Consequently, there
427: exist positive real numbers $\eta$ and $\eta'$ with $\eta < 1$ such that
428: $$
429: 2 (1 +\eta)(1 + \eta')r_n + \eta (1 + \eta') s_n <
430: (1 - \eta') (r_n + w s_n - s_n) {\log m \over \log M}, \eqno (3.2)
431: $$
432: for any $n \ge 1$.
433:
434: In the course of the proof of Theorem 4.1
435: from \cite{AdBuAM} we had to bound from
436: above the quantity $q_{r_n}^{1 + \eta} \, q_{r_n + s_n}^{1 + \eta}
437: \, q_{r_n + [w s_n]}^{-1}$.
438: Our new observation is that the estimate
439: $$
440: q_{r_n + s_n} \, q_{r_n + [(w-1) s_n]} \ll
441: q_{r_n} \, q_{r_n + [w s_n]}
442: $$
443: follows from Lemma 3.2 (here and below, the
444: numerical constant implied by $\ll$ does not depend on $n$).
445: Consequently, we get
446: $$
447: q_{r_n}^{1 + \eta} \, q_{r_n + s_n}^{1 + \eta} \, q_{r_n + [w s_n]}^{-1}
448: \ll
449: q_{r_n}^{2 + \eta} \, q_{r_n + s_n}^{\eta} \, q_{r_n + [(w-1) s_n]}^{-1}.
450: \eqno (3.3)
451: $$
452:
453:
454: Assuming $n$ sufficiently large, we then have
455: $$
456: q_{r_n} \le M^{(1 + \eta') r_n}, \quad
457: q_{r_n + s_n} \le M^{(1 + \eta') (r_n+s_n)}, \quad {\rm and}
458: \quad q_{r_n + [(w-1) s_n]} \ge m^{(1 - \eta') (r_n + w s_n - s_n)},
459: $$
460: with $\eta'$ as in (3.2). Consequently, we get
461: $$
462: q_{r_n}^{2 + \eta} \, q_{r_n + s_n}^{\eta} \, q_{r_n + [(w-1) s_n]}^{-1}
463: \le M^{2(1 +\eta)(1 + \eta')r_n + \eta (1 + \eta') s_n}
464: \, m^{- (1 - \eta') (r_n + w s_n - s_n)}
465: \le 1,
466: $$
467: by our choice of $\eta$ and $\eta'$. It then follows from (3.2) and (3.3) that
468: $$
469: q_{r_n}^{1 + \eta} \, q_{r_n + s_n}^{1 + \eta} \, q_{r_n + [w s_n]}^{-1} \ll 1,
470: $$
471: and, with the notation from \cite{AdBuAM}, we get the upper bound
472: $$
473: \prod_{1 \le j \le 4} \, |L_j ({\underline z_n})|
474: \ll ( q_{r_n} \, q_{r_n + s_n})^{-\eta}
475: $$
476: for any positive integer $n$. We then conclude as in that paper. \cqfd
477:
478: \vskip 6mm
479:
480: \centerline{\bf 4. Davison's continued fractions}
481:
482: \vskip 8mm
483:
484: Let $\theta$ be an irrational number with $0 < \theta < 1$.
485: Under some mild assumptions,
486: Davison \cite{Dav89} established the transcendence of the real number
487: $\alpha_{\theta}=[0;d_1,d_2,\ldots]$, with $d_n=1+([n\theta] \hbox{ mod }
488: 2)$ for any $n \ge 1$. These extra assumptions were subsequently removed in
489: \cite{ADQZ}. Then, Davison \cite{Dav} and Baxa \cite{Bax03}
490: studied the more general question of the transcendence of the real number
491: $\alpha_{k, \theta}=[0;d_1,d_2,\ldots]$, where $k \ge 2$
492: is an integer and $d_n=1+([n\theta] \hbox{ mod }
493: k)$ for any $n\ge 1$. The two authors obtained
494: some partial results but their methods did not allow them to cover all the
495: cases. It turns out that Theorem A yields a complete
496: answer to this question.
497:
498: \proclaim Theorem 4.1. Let $\theta$ be an irrational number with
499: $0<\theta<1$ and let $k$ be an integer at
500: least equal to $2$. Let ${\bf d}=(d_n)_{n\geq 1}$ be defined by
501: $d_n=1+([n\theta] \hbox{ mod }k)$ for any $n \ge 1$. Then, the number
502: $\alpha_{k, \theta}=[0;d_1,d_2,\ldots]$ is transcendental.
503:
504:
505: In order to prove Theorem 4.1,
506: we need two auxiliary results (Lemmas 4.3 and 4.4 below),
507: which will be deduced from the following proposition obtained
508: in \cite{Dav}. Throughout this Section, $\theta
509: = [0; a_1, a_2, \ldots, a_n, \ldots]$ denotes an
510: irrational number in $(0, 1)$ and $(p_{n}/ q_{n})_{n \ge 0}$ is
511: the sequence of its convergents.
512:
513: \proclaim Proposition 4.2. For any non-negative integers $n$ and $r$
514: with $1\leq r\leq q_{n+1}-1$, we have
515: $$
516: [(q_n+r)\theta]=p_n +[r\theta].
517: $$
518:
519: Our first auxiliary result is the following.
520:
521: \proclaim Lemma 4.3. For any integers $n$, $r$ and $s$
522: with $n\geq 1$, $0\leq s\leq a_{n+1}$ and $1\leq
523: r\leq q_n+q_{n-1}-1$, we have
524: $$
525: [(sq_n+r)\theta]=sp_n+[r\theta].
526: $$
527:
528: \noi{\bf Proof.} The proof goes by induction on $s$. If $s=0$, the
529: statement is a tautology. Let us assume that the result holds for a
530: given $s$ with $0\leq s<a_{n+1}$. Then, we have $s+1\leq a_{n+1}$ and
531: $$
532: (s+1)q_n+r\leq a_{n+1}q_n+q_n+q_{n-1}-1=q_{n+1}+q_n-1.
533: $$
534: Hence, $sq_n+r\leq q_{n+1}-1$ and Proposition 4.2 implies that
535: $$
536: [((s+1)q_n+r)\theta]=[(q_n+(sq_n+r))\theta]
537: =p_n+[(sq_n+r)\theta].
538: $$
539: By our inductive assumption, we thus obtain
540: $$
541: [((s+1)q_n+r)\theta]=p_n+sp_n+[r\theta]=(s+1)p_n+[r\theta],
542: $$
543: concluding the proof of the lemma.
544: \cqfd
545:
546:
547:
548: \proclaim Lemma 4.4. For any integers $n$, $\ell$ and $r$
549: with $n\geq 1$, $\ell \geq 0$ and $1\leq
550: r\leq q_{n+1}-1$, we have
551: $$
552: [(q_{n+\ell}+q_{n+\ell-1}+\ldots+q_n+r)\theta]
553: = p_{n+\ell}+p_{n+\ell-1}+\ldots+p_n+[r\theta].
554: $$
555:
556: \noi{\bf Proof.} The proof goes by induction on $\ell$. The case $\ell=0$ is
557: given by Proposition 4.2. Let us assume that the desired result holds for a
558: given non-negative integer $\ell$.
559: Let $n$ and $r$ be as in the statement of the lemma.
560: Since $q_{m+2} \ge q_{m+1} + q_m$ holds
561: for any non-negative integer $m$, we easily get that
562: $$
563: q_{n+\ell}+\ldots+q_{n+1}+q_n+r\leq q_{n+\ell+2}-1.
564: $$
565: Then, we infer from Proposition 4.2 that
566: $$
567: \eqalign{
568: [(q_{n+\ell+1}+q_{n+\ell}+\ldots+q_n+r)\theta] &
569: = [(q_{n+\ell+1}+(q_{n+\ell}+\ldots+q_n+r))\theta] \cr
570: & =p_{n+\ell+1}+[(q_{n+\ell}+\ldots+q_n+r))\theta]. \cr}
571: $$
572: Using the inductive assumption, this shows that
573: the desired result holds for $\ell+1$
574: and proves the lemma. \cqfd
575:
576: {\noi{\bf Proof of Theorem 4.1.} We are now ready to prove Theorem
577: 4.1 by showing that the sequence $(d_n)_{n\geq 1}$ satisfies
578: Condition $(*)_w$ for a suitable real number $w > 1$.
579: We first remark that the sequence ${\bf d}$ is not
580: eventually periodic since $\theta$ is irrational.
581: We have to distinguish two cases.
582:
583: First, let us assume that there are
584: infinitely many integers $n$ such that $a_{n+1}\geq k$.
585: For such an $n$, we infer from Lemma 4.3
586: with $s=k$ that
587: $$
588: [(kq_n+r)\theta] \equiv [r\theta] \hbox{ mod }k,
589: \hbox{ for }1\leq r\leq q_n+q_{n-1}-1. \eqno (4.1)
590: $$
591: Set $V_n=d_1\ldots d_{kq_n}$ and
592: $W_n=d_{kq_n+1}d_{kq_n+2}\ldots
593: d_{(k+1)q_n+q_{n-1}-1}$ and view $V_n$ and $W_n$ as
594: words on the alphabet $\{1, 2, \ldots, k\}$.
595: It follows from $(4.1)$ that $W_n$ is a prefix
596: of $V_n$. Furthermore, since
597: $$
598: {\vert W_n\vert \over \vert V_n\vert}
599: ={q_n+q_{n-1}-1\over kq_n}\geq{1\over k},
600: $$
601: the infinite word ${\bf d}$ begins in $V_n^{1+1/k}$.
602: Thus, ${\bf d}$ satisfies Condition $(*)_{1 + 1/k}$.
603:
604: Now, let us assume that there exists an integer $n_0$ such that
605: $a_n < k$ for all $n\geq n_0$. Let $n \ge n_0$ be an integer.
606: At least two among the $k+1$ integers
607: $p_n,p_n+p_{n+1},\ldots,p_n+p_{n+1}+\ldots+p_{n+k}$ are
608: congruent modulo $k$. Consequently, there exist
609: integers $n'\geq n+1$ and $\ell$, with
610: $0 \le \ell \le k-1$ and $n' + \ell \le n + k$, such that
611: $$
612: p_{n'+\ell}+p_{n'+\ell-1}+\ldots+p_{n'} \equiv 0\hbox{ mod }k.
613: $$
614: Set $N=q_{n'+\ell}+q_{n'+\ell-1}+\ldots+q_{n'}$.
615: Then, Lemma 4.4 implies that
616: $$
617: [(N+r)\theta] \equiv [r\theta] \quad \hbox{ mod }k, \eqno (4.2)
618: $$
619: for $1\leq r\leq q_{n'+1}-1$.
620: Set $V_n=d_1\ldots d_{N}$ and $W_n=d_{N+1}\ldots d_{N+q_{n'+1}-1}$
621: and view $V_n$ and $W_n$ as words on the alphabet $\{1, 2, \ldots, k\}$.
622: Then, $(4.2)$ implies that $W_n$ is a prefix
623: of $V_n$. Since $q_{n'+\ell}\leq k^{\ell} q_{n'}$ (this follows
624: from the assumption $a_m < k$ for any $m \ge n_0$), we obtain
625: $$
626: {\vert W_n\vert \over \vert V_n\vert}
627: ={q_{n'+1}-1\over N}\geq{q_{n'}\over kq_{n'+\ell}} \ge
628: {1 \over k^{\ell+1}}\geq {1 \over k^{k}}.
629: $$
630: It thus follows that the infinite word ${\bf d}$ begins in
631: $V_n^{1+ 1/k^k}$.
632:
633: Consequently, the sequence ${\bf d}$ satisfies
634: in both cases Condition $(*)_{1+1/k^k}$. By Theorem A, the real number
635: $\alpha_{k, \theta}$ is transcendental, as claimed. \cqfd
636:
637:
638:
639: \vskip 8mm
640:
641: \centerline{\bf 5. Automatic continued fractions}
642:
643: \vskip 6mm
644:
645: Let $k \ge 2$ be an integer.
646: An infinite sequence ${\bf a}=(a_n)_{n\geq 0}$ is
647: said to be $k$-automatic if $a_n$ is a finite-state function of the base-$k$
648: representation of $n$. This means that there exists
649: a finite automaton starting
650: with the $k$-ary expansion of $n$ as input and producing the term $a_n$ as
651: output. Finite automata are one of the most basic models of computation and
652: take place at the bottom
653: of the hierarchy of Turing machines. A nice reference on this topic is the
654: book of Allouche and Shallit \cite{Allouche_Shallit}.
655: We refer the reader to it for more details about this notion.
656:
657: Motivated by the Hartmanis-Stearns problem \cite{Hartmanis_Stearns},
658: the following question
659: was addressed in \cite{AdBuAM}: do there exist algebraic numbers
660: of degree at least three
661: whose continued fraction expansion can be produced by a finite automaton?
662:
663: The first result towards this problem is due to
664: Queff\'elec \cite{Queffelec98}, who proved the transcendence
665: of the Thue--Morse continued fractions. A more general statement
666: can be found in \cite{Queffelec00}.
667: As it is shown in \cite{AdBuAM}, the transcendence of a large class
668: of automatic continued fractions can be derived from Theorem A and B.
669: In the present Section, we show how our transcendence
670: criteria apply to two others emblematic automatic sequences:
671: the Rudin--Shapiro and the Baum--Sweet sequences.
672:
673:
674: \medskip
675:
676: Before proving such results, we recall some classical facts about morphisms and morphic sequences.
677: For a finite set ${\cal A}$, we denote by ${\cal A}^*$ the free monoid
678: generated by ${\cal A}$. The empty word $\varepsilon$ is the neutral element
679: of ${\cal A}^*$. Let ${\cal A}$ and ${\cal B}$ be two finite sets. An
680: application from ${\cal A}$ to ${\cal B}^*$ can be uniquely extended to a
681: homomorphism between the free monoids ${\cal A}^*$ and ${\cal B}^*$. We
682: call morphism from ${\cal A}$ to ${\cal B}$ such a
683: homomorphism. If there is a positive integer
684: $k$ such that each element of ${\cal A}$ is mapped to a word of
685: length $k$, then the morphism is called $k$-uniform or simply
686: uniform.
687:
688: A morphism
689: $\sigma$ from ${\cal A}$ into itself is said to
690: be prolongable if there exists a
691: letter $a$ such that $\sigma(a)=aW$, where the word $W$ is such
692: that $\sigma^n(W)$ is a non-empty word for every $n\geq 0$.
693: In that case, the
694: sequence of finite words $(\sigma^n(a))_{n\geq 1}$ converges in
695: ${\cal A}^{\Z_{\ge 0}}$ (endowed with the product topology of the discrete
696: topology on each copy of ${\cal A}$)
697: to an infinite word ${\bf a}$. Such an infinite word is clearly
698: a fixed point for the map $\sigma$.
699:
700:
701:
702: \vskip 6mm
703:
704: {\bf 5.1 The Rudin--Shapiro continued fractions}
705:
706: \vskip 6mm
707:
708: Let ${\bf \varepsilon}=(\varepsilon_n)_{n\geq 0}$ be a
709: sequence with values in the set $\{+1,-1\}$. It is not difficult to
710: see that
711: $$
712: \sup_{\theta\in [0,1]}\biggl \vert\sum_{0 \le n<N}
713: \varepsilon_ne^{2i\pi n\theta}\biggr \vert\geq\sqrt N,
714: $$
715: for any positive integer $N$.
716: In 1950, Salem asked the following question, related to some problems
717: in harmonic analysis:
718: do there exist a sequence ${\bf \varepsilon}=(\varepsilon_n)_{n\geq
719: 0}$ in $\{+1,-1\}^{\Z_{\ge 0}}$ and a positive constant $c$ such that
720: $$
721: \sup_{\theta\in [0,1]}\biggl \vert\sum_{0 \le n<N}
722: \varepsilon_ne^{2i\pi n\theta}\biggr \vert\leq c\sqrt N \eqno (5.1)
723: $$
724: holds for any positive integer $N$?
725: A positive answer to this problem was given by Shapiro
726: \cite{Shapiro} and Rudin \cite{Rudin}, who provided an explicit solution
727: which is now known
728: as the Rudin--Shapiro sequence. This sequence is
729: a famous
730: example of a $2$-automatic sequence and can be defined as follows: $r_n$
731: is equal to $+1$ (respectively $-1$) if the number of occurrences of
732: the pattern `$11$' in the binary representation of $n$ is even
733: (respectively odd). Theorem A yields the
734: transcendence of the Rudin--Shapiro continued fractions.
735:
736:
737: \proclaim Theorem 5.1. Let $a$ and $b$ two distinct
738: positive integers, and let ${\bf r}=(r_n)_{n\geq 0}$ be the
739: Rudin--Shapiro
740: sequence on the alphabet $\{a,b\}$ (that is the symbol
741: $1$ is replaced by $a$ and the symbol $-1$ is replaced
742: by $b$ in the usual Rudin--Shapiro sequence). Then, the real number
743: $\alpha=[0;r_0,r_1,r_2,\ldots]$
744: is transcendental.
745:
746: {\noi{\bf Proof.} We first infer from (5.1) that the Rudin--Shapiro
747: sequence is not eventually periodic.
748: We present now a useful description of this sequence.
749: Let $\sigma$ be a morphism defined from $\{1,2,3,4\}^*$ into itself by:
750: $\sigma(1)=12$, $\sigma(2)=42$, $\sigma(3)=13$ and
751: $\sigma(4)=43$. Let
752: $$
753: {\bf u}=1242434213 \ldots
754: $$
755: be the fixed point of $\sigma$ begining with $1$ and let $\varphi$
756: be the morphism defined from $\{1,2,3,4\}^*$ to $\{a,b\}^*$ by:
757: $\varphi(1)=\varphi(2)=a$ and $\varphi(3)=\varphi(4)=b$.
758: It is known (see for instance \cite{Fog}, Ch. 5)
759: that
760: $$
761: {\bf r}=\varphi({\bf u}). \eqno (5.2)
762: $$
763: Since $\sigma^5 (1) = 1242434213$, ${\bf u}$ begins with $V^{1+1/8}$,
764: where $V=12424342$. Then, it follows from (5.2)
765: that for any positive integer $n$, the word ${\bf r}$ begins
766: with $\varphi(\sigma^n(V^{1+1/8}))$. The morphism $\sigma$
767: being a uniform morphism, we easily check that
768: $$
769: \varphi(\sigma^n(V^{1+1/8}))=(\varphi(\sigma^n(V)))^{1+1/8}.
770: $$
771: The sequence ${\bf r}$ thus satisfies the condition $(*)_{1+1/8}$.
772: This ends the proof thanks to Theorem~A. \cqfd
773:
774: \vskip 6mm
775:
776: {\bf 5.2 The Baum--Sweet continued fractions}
777:
778: \vskip 6mm
779:
780: In 1976, Baum and Sweet \cite{Baum_Sweet1} proved that, unlike
781: what is expected in the real case, the function field
782: ${\bf F}_2((X^{-1}))$ contains a cubic element (over ${\bf F}_2(X)$)
783: with bounded partial quotient in its continued fraction expansion.
784: This element is
785: $\sum_{n\ge 0} s_n X^{-n}$ , where $s_n$
786: is equal to $0$ if the binary representation of $n$
787: contains at least one string of $0'$s of odd length
788: and $s_n$ is equal to $1$
789: otherwise. The sequence ${\bf s}=(s_n)_{n\geq 0}$ is now usually referred
790: to as the Baum--Sweet sequence.
791: It follows from \cite{ABL,AdBuAoM} that,
792: for any integer $b\ge 2$, the real number $\sum_{n\ge 0}s_n/b^n$
793: is transcendental.
794: Here, we use Theorem 3.1 to prove a similar result
795: for the continued fraction expansion.
796:
797:
798: \proclaim Theorem 5.2. Let $a$ and $b$ be distinct
799: positive integers, and let ${\bf s}=(s_n)_{n\geq 0}$ be the
800: Baum--Sweet sequence on the alphabet $\{a,b\}$
801: (that is, the symbol $0$ is replaced by $a$ and the symbol $1$
802: is replaced by $b$ in the usual Baum--Sweet sequence).
803: Then, the real number
804: $\alpha=[0;s_0,s_1,s_2,\ldots]$
805: is transcendental.
806:
807:
808:
809: \noi{\bf Proof.} Let us first remark that the sequence ${\bf s}$
810: is not eventually periodic. Indeed, as shown in \cite{Baum_Sweet1},
811: the formal power series $\sum_{n\ge 0} s_n X^{-n}$ is
812: a cubic element over ${\bf F}_2 (X)$, thus, it is
813: not a rational function.
814:
815: Let us now recall a useful description of
816: the Baum--Sweet sequence on the alphabet $\{a, b\}$. Let
817: $\sigma$ be the morphism defined from $\{1,2,3,4\}^*$ into itself by:
818: $\sigma(1)=12$,
819: $\sigma(2)=32$,
820: $\sigma(3)=24$ and $\sigma(4)=44$. Let also $\varphi$
821: be the morphism defined from $\{1,2,3,4\}^*$
822: to $\{a,b\}^*$ by: $\varphi(1)=\varphi(2)=b$ and $\varphi(3)=\varphi(4)=a$.
823: Let $${\bf u} = 123224323244 \ldots$$
824: denote the fixed point of $\sigma$ begining with $1$. It is known (see for instance \cite{Allouche_Shallit}, Ch. 6)
825: that
826: $${\bf s}=\varphi({\bf u}). \eqno (5.3)$$
827: Observe that ${\bf u}$ begins in the word $UV^{3/2}$,
828: where $U=1$ and $V=232243$. Since the morphism $\sigma$ is
829: uniform, it follows from $(5.3)$ that, for any positive
830: integer $n$, the word ${\bf s}$ begins with $U_nV_n^{3/2}$, where
831: $U_n=\varphi(\sigma^n(U))$ and $V_n=\varphi(\sigma^n(V))$. In
832: particular, we have $\vert U_n\vert/\vert V_n\vert=1/6$ and
833: ${\bf s}$ thus satisfies Condition $(**)_{3/2, 1/6}$.
834: We further observe that the frequency of $a$ in the word $U_n$
835: tends to $1$ as $n$ tends to infinity. By Theorem 5 from \cite{ADQZ},
836: this implies that the sequence $(q_{\ell})^{1/\ell}_{\ell \ge 1}$
837: converges, where
838: $q_\ell$ denotes the denominator of
839: the $\ell$-th convergent of $\alpha$,
840: for any positive integer $\ell$.
841: Thus, with the notation of Theorem 3.1, we have $M=m$.
842: Since $3/2 > 1 + 1/6$,
843: we derive from Theorem 3.1 that $\alpha$ is transcendental.
844: This finishes the proof of our theorem. \cqfd
845: \bigskip
846:
847:
848:
849: \vskip 6mm
850:
851: \centerline{\bf 6. Folded continued fractions}
852:
853: \vskip 8mm
854:
855: Numerous papers, including the survey \cite{DeMevdP},
856: are devoted to paperfolding sequences.
857: In this Section we consider folded continued fractions
858: and we prove that they are always
859: transcendental.
860:
861: A sheet of paper can be folded in half lengthways in
862: two ways: right half over left (the positive way) or left half over
863: right (the negative way). After having
864: been folded an infinite number of times, the sheet of paper can be
865: unfolded to display an infinite sequence of creases formed by hills
866: and valleys. For convenience, we will denote by $+1$ the hills and by
867: $-1$ the valleys. The simplest choice, that is to fold always in the
868: positive way, gives the well-known regular paperfolding sequence over
869: the alphabet $\{+1,-1\}$. More
870: generally, if ${\bf e}=(e_n)_{n \ge 0}$ is in $\{+1,-1\}^{\Z_{\ge 0}}$,
871: the associated paperfolding
872: sequence on the alphabet $\{+1,-1\}$ is obtained accordingly to the
873: sequence ${\bf e}$ of folding instructions, that is, the $n$-th fold
874: is positive if $e_n=+1$ and it is negative otherwise.
875:
876:
877: Among the numerous studies concerned with paperfolding sequences, much
878: attention has been brought on the way (quite intriguing) they are
879: related to
880: the continued fraction expansion of some formal power series.
881: Indeed, it can be shown that for any sequence
882: ${\bf e}=(e_n)_{n\geq 0}\in\{+1,-1\}^{\Z_{\ge 0}}$ of folded instructions,
883: the continued fraction expansion of the formal power series
884: $X\sum_{n\geq 0}e_nX^{-2^n}$ can be deduced from the associated paperfolding
885: sequence ${\bf f}$.
886: As a consequence, the authors of \cite{vdP_Shallit} precisely
887: described the continued fraction expansion
888: of the real number $\xi = 2\sum_{n\geq 0}e_n2^{-2^n}$.
889: In particular, they proved that such an expansion is also closely
890: related to the sequence ${\bf f}$ and called it a
891: `folded continued fraction'.
892: However, we point out that these `folded continued fractions'
893: are not the same as those we consider in Theorem 6.1.
894: Real numbers such as $\xi$
895: were shown to be transcendental in
896: \cite{LvdP77} thanks to the so-called Mahler method
897: (the observation that Ridout's Theorem \cite{Ridout} also implies
898: that such numbers are transcendental was
899: for instance done in \cite{Adam}).
900: Consequently, we get the transcendence of a family of continued
901: fractions whose shape arises from paperfolding sequences;
902: a fact mentioned in \cite{vdP_Shallit}. The following result
903: has the same flavour though it is obtained in a totally different way.
904: It deals with another family of continued
905: fractions arising from paperfolding sequences.
906:
907: \proclaim Theorem 6.1. Let $a$ and $b$ be two positive distinct
908: integers, $(e_n)_{n\geq 0}
909: \in\{+1,-1\}^{\Z_{\ge 0}}$ be a sequence of folding instructions
910: and let ${\bf f}=(f_n)_{n\geq 0}$ be the associated paperfolding
911: sequence over the alphabet $\{a,b\}$ (that is, the symbol $+1$
912: is replaced by $a$ and the symbol $-1$ is replaced by $b$
913: in the usual paperfolding sequence associated with $(e_n)_{n\geq 0}$).
914: Then, the number
915: $\alpha=[0;f_0,f_1,f_2,\ldots]$ is transcendental.
916:
917:
918:
919:
920: {\noi{\bf Proof.} Let ${\bf f}=(f_n)_{n\geq 0}$ be the
921: paperfolding sequence on the alphabet $\{a,b\}$ associated with the
922: sequence $(e_n)_{n\geq 0}$ in $\{+1,-1\}^{\Z_{\ge 0}}$ of folding instructions.
923: We first notice that ${\bf f}$ is not eventually periodic
924: since no paperfolding sequence is eventually periodic (see
925: \cite{DeMevdP}).
926:
927: We present now another useful description of the sequence ${\bf f}$.
928: Let ${\cal F}_i$, $i\in\{+1,-1\}$, be the map defined from
929: the set $\{+1,-1\}^*$
930: into itself by
931: $$
932: {\cal F}_i:\;\;w\longmapsto wi-({\overline w }),
933: $$
934: where $\overline{ w_1w_2\ldots w_m} := w_m w_{m-1}\ldots w_1$
935: denotes the mirror image of the word $w_1 w_2 \ldots w_m$, and
936: $-(w_1w_2\ldots w_m):=w'_1w'_2\ldots w'_m$, with $w'_i=+1$
937: (resp. $w'_i=-1$) if $w_i=-1$ (resp. $w_i=+1$).
938: Let $\varphi$ be the morphism defined from $\{+1,-1\}^*$
939: to $\{a,b\}^*$ by $\varphi(+1)=a$
940: and $\varphi(-1)=b$. One can easily verify
941: (see for instance \cite{DeMevdP}) that
942: $$
943: {\bf f}=\lim_{n\to + \infty}\varphi({\cal F}_{e_0}{\cal F}_{e_1}\ldots{\cal
944: F}_{e_n}(\varepsilon)), \eqno (6.1)
945: $$
946: where $\varepsilon$ denotes the empty word.
947:
948: Let $n \ge 2$ be an integer and set $V_n={\cal
949: F}_{e_2}\ldots
950: {\cal F}_{e_n}(\varepsilon)$. By (6.1), the finite word
951: $$
952: F_n=\varphi({\cal F}_{e_0}{\cal F}_{e_1}\ldots{\cal
953: F}_{e_n}(\varepsilon))=\varphi({\cal F}_{e_0}{\cal F}_{e_1}(V_n))
954: $$
955: is a prefix of ${\bf f}$. Moreover, we have
956: $$
957: {\cal F}_{e_0}{\cal F}_{e_1}(V_n)={\cal F}_{e_0}(V_n e_1-\overline{
958: V_n})=(V_n e_1-\overline{
959: V_n})e_0-(\overline{ V_n e_1-\overline{
960: V_n}}).
961: $$
962: This gives
963: $$
964: {\cal F}_{e_0}{\cal F}_{e_1}(V_n)=(V_n e_1-\overline{
965: V_n}) e_0 V_n-(\overline{V_n e_1}).
966: $$
967: In particular, the word
968: $$
969: \varphi(V_n e_1(-\overline{ V_n}) e_0 V_n)
970: $$
971: is a prefix of ${\bf f}$.
972: Moreover, we have
973: $$
974: \varphi(V_n e_1(-\overline{ V_n}) e_0 V_n)=
975: (\varphi(V_ne_1(-\overline{ V_n}) e_0 ))^{w_n},
976: $$
977: where $w_n=1+\vert \varphi(V_n)\vert / (2\vert \varphi(V_n)\vert +2)$.
978: Thus, ${\bf f}$ begins with the word
979: $(\varphi(V_ne_1(-\overline{ V_n}) e_0 ))^{5/4}$. Since $|\varphi(V_n)|$
980: tends to infinity with $n$,
981: this proves
982: that ${\bf f}$ satisfies Condition $(*)_{5/4}$. Consequently,
983: Theorem A implies that the real number $\alpha = [0; f_0, f_1, f_2, \ldots]$
984: is transcendental. This ends the proof. \cqfd
985: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
986:
987: \vskip 6mm
988:
989: \centerline{\bf 7. Generalized perturbed symmetry systems}
990:
991: \vskip 8mm
992:
993: A perturbed symmetry is a map defined from ${\cal A}^*$ into itself
994: by $S_X(W)=WX\overline W$, where
995: ${\cal A}$ is a finite set, $W$ and $X$ are finite words
996: on ${\cal A}$, and $\overline W$ is
997: defined as in the previous Section.
998: Since $S_X(W)$ begins in $W$, we can
999: iterate the map $S_X$ to obtain the infinite sequence
1000: $$
1001: S^{\infty}_X(W)=WX\overline{W}XW\overline{X}\overline{W}\ldots
1002: $$
1003: Such sequences were introduced by Mend\`es France \cite{Mendes}, who
1004: proved that they are periodic if and only if the word $X$ is a
1005: palindrome, that is, $X={\overline X}$.
1006:
1007: More recently, Allouche and Shallit \cite{Allouche_Shallit98}
1008: generalized this notion as follows. For convenience, for any finite word $W$,
1009: we set $W^E := W$ and $W^R := \overline{W}$.
1010: Let $k$ be a positive integer. Let
1011: $X_1,X_2,\ldots,X_{k}$ be finite words of the same length $s$, and
1012: let $e_1,e_2,\ldots,e_{k}$ be in $\{E,R\}$. Then, the associated
1013: generalized perturbed symmetry is defined by
1014: $$
1015: S(W)=W\prod_{i=1}^{k} X_i W^{e_i}.
1016: $$
1017: Again, since $S(W)$ begins in $W$, we get
1018: an infinite word $S^{\infty}(W)$ by iterating $S$.
1019: In \cite{Allouche_Shallit98}, the authors proved
1020: that $S^{\infty}(W)$ is $(k+1)$-automatic and gave a necessary and
1021: sufficient condition for this sequence to be eventually periodic.
1022:
1023: We introduce now a generalization of this process.
1024: First, let us assume in the previous definition that the
1025: $X_i$ are finite and possibly empty words,
1026: without any condition on their length.
1027: We can similarly define a map $S$ that we again call a generalized perturbed
1028: symmetry. If all the $X_i$ have the same length, as previously,
1029: we may then call this map a uniform generalized perturbed symmetry.
1030: We point out that
1031: the definition we consider here is more general than
1032: the one given in \cite{Allouche_Shallit98}.
1033: A generalized perturbed symmetry system is defined as a $4$-tuple
1034: $({\cal A},W,{\cal S},(S_n)_{n\geq 0})$, where ${\cal A}$ is a finite
1035: set, $W\in{\cal A}^*$, $W\not=\varepsilon$,
1036: ${\cal S}$ is a finite set of generalized
1037: perturbed symmetries and $(S_n)_{n\geq 0}\in{\cal S}^{\Z_{\ge 0}}$.
1038: Then, it is easily verified that the sequence of finite
1039: words
1040: $$
1041: S_nS_{n-1}\ldots S_0(W)
1042: $$ converges to an infinite
1043: sequence, which we call the sequence produced by the generalized
1044: perturbed symmetry system $({\cal A},W,{\cal S},(S_n)_{n\geq 0})$.
1045: When ${\cal A}$ is a subset of $\Z_{\ge 1}$, we obtain an infinite sequence
1046: of positive integers that we can view as the continued fraction
1047: expansion of a real number.
1048:
1049:
1050: \proclaim Theorem 7.1. Generalized perturbed symmetry systems generate
1051: either quadratic or transcendental continued fractions.
1052:
1053: We point out that the set of numbers generated by
1054: generalized perturbed symmetry systems is not countable.
1055:
1056: \medskip
1057:
1058: {\noi{\bf Proof.} For any integer $n\geq 0$, we consider a
1059: generalized perturbed symmetry $S_n$ asso\-ciated with the parameters
1060: $X_{1,n},X_{2,n},\ldots,X_{k_n,n}$ and $e_{1,n},e_{2,n},\ldots,e_{k_n,n}$
1061: in $\{E, R\}$.
1062: Let ${\bf u}$ be the sequence generated by the generalized perturbed
1063: symmetry system $({\cal A},W,{\cal S},(S_n)_{n\geq 0})$. For every positive
1064: integer $n$ we define the finite word
1065: $$
1066: W_n=S_{n-1}\ldots S_0(W).
1067: $$
1068: Then, $W_n$ is a prefix of ${\bf u}$ and the sequence $(\vert
1069: W_n\vert)_{n\geq 1}$ tends to infinity.
1070: To prove Theorem 7.1, we will show that there exists a rational $w$
1071: greater than $1$, such that ${\bf u}$ satisfies
1072: Condition $(*)_w$. In order to do this, we have to distinguish two
1073: different cases.
1074:
1075: First, assume that for
1076: infinitely many integers $n$, we have $e_{1,n}=E$. In this case,
1077: $S_n(W_n)$ begins in $W_nX_{1,n} W_n^{e_{1,n}}=W_n X_{1,n} W_n$, thus the
1078: sequence ${\bf u}$ begins in $W_n X_{1,n} W_n$. Since $\vert W_n\vert$ tends
1079: to infinity and $\vert
1080: X_{1,n} \vert$ lies in a finite set, we have that ${\bf u}$ begins in
1081: $(W_n X_{1,n})^{3/2}$ as soon as $n$ is large enough. Thus, ${\bf u}$
1082: satisfies Condition $(*)_{3/2}$.
1083:
1084: Now, assume that there exists an integer $n_0$ such that
1085: for all $n\geq n_0$, we have $e_{1,n}=R$. Let $n$ be an integer greater
1086: than $n_0$. Then, $S_n(W_n)$ begins in $W_nX_{1,n}W_n^R$, where
1087: $$
1088: W_n=S_{n-1}(W_{n-1}) = W_{n-1}
1089: \prod_{i=1}^{k_{n-1}} X_{i,n-1}W_{n-1}^{e_{i, n-1}}.
1090: $$
1091: This implies that $S_n (W_n)$ begins with
1092: $$
1093: \eqalign{
1094: & \biggl(W_{n-1}\prod_{i=1}^{k_{n-1}}
1095: X_{i, n-1}W_{n-1}^{e_{i, n-1}}\biggr)
1096: X_{1,n} \biggl(W_{n-1}\prod_{i=1}^{k_{n-1}}
1097: X_{i,n-1} W_{n-1}^{e_{i,n-1}}\biggr)^R \cr
1098: & =\biggl(W_{n-1}\prod_{i=1}^{k_{n-1}}X_{i,n-1} W_{n-1}^{e_{i,n-1}}\biggr)
1099: X_{1,n}
1100: \biggl(\, \prod_{i=2}^{k_{n-1}} X_{i,n-1} W_{n-1}^{e_{i, n-1}}\biggr)^R
1101: \biggl(W_{n-1} X_{1, n-1} W_{n-1}^{e_{1, n-1}}\biggr)^R. \cr}
1102: $$
1103: Since $n>n_0$, we have $e_{1,n-1}=R$. Consequently, $S_n(W_n)$ begins in
1104: $$
1105: W_{n-1}\biggl(\, \prod_{i=1}^{k_{n-1}}
1106: X_{i,n-1} W_{n-1}^{e_{i,n-1}}\biggr) X_{1,n}
1107: \biggl(\, \prod_{i=1}^{k_{n-1}}
1108: X_{i,n-1} W_{n-1}^{e_{i,n-1}} \biggr)^R
1109: W_{n-1} =: W_{n-1} Y_{n-1} W_{n-1}.
1110: $$
1111: Since both the $k_j$ and the $X_{i,j}$ lie in a finite set, and since
1112: $\vert W_n\vert$ tends to infinity with $n$, we get
1113: that $S_n(W_n)$, and thus ${\bf u}$,
1114: begins in $(W_{n-1} Y_{n-1})^{1+ 1 / (3k)}$ for $n$
1115: large enough, where we have set
1116: $k=\max\{k_n : n \ge 1\}$. This implies
1117: that ${\bf u}$ satifies Condition $(*)_{1 + 1/(3k)}$,
1118: and we conclude the proof by applying Theorem A. \cqfd
1119: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
1120:
1121:
1122: \vskip 6mm
1123:
1124: \centerline{\bf 8. Inside the proof of Theorem 3.1:
1125: an example of the use of continuants}
1126:
1127: \vskip 8mm
1128:
1129: The purpose of this Section is to point out that, in some cases, rather than
1130: applying roughly Theorem 3.1, it is much better to go into its
1131: proof and to evaluate continuants carefully.
1132: In order to illustrate this idea, we introduce a new
1133: family of continued fractions.
1134:
1135: \medskip
1136:
1137: Throughout this Section, we use the following notation.
1138: Let $k \ge 3$ be an odd integer. Let $b_1, \ldots , b_k$
1139: be positive integers with $b_1 < \ldots < b_k$. We
1140: consider words defined on the alphabet $\Sigma = \{b_1, \ldots , b_k\}$.
1141: The character $b_j$ denotes either the letter $b_j$, or the integer $b_j$,
1142: according to the context. For any finite word $W$ on $\Sigma$
1143: and for $j=1, \ldots, k$, let denote by $|W|_{b_j}$ the number of
1144: occurrences of the letter $b_j$ in $W$. Set
1145: $$
1146: \Sigma^+_{\hbox{\sevenrm equal}} = \{ W \in \Sigma^+ :
1147: |W|_{b_j} = |W|_{b_1} \quad \hbox{for $2 \le j \le k$} \}.
1148: $$
1149:
1150: Let $\lambda > 1$ be real and, for any $n \ge 1$, let $W_n$ be in
1151: $\Sigma^+_{\hbox{\sevenrm equal}}$ with $|W_{n+1}| > \lambda |W_n|$.
1152: Consider the
1153: infinite word ${\bf a} = (a_{\ell})_{\ell \ge 1}$ defined by
1154: $$
1155: {\bf a} = W_1 W_2^2 W_3^2 \ldots W_n^2 \ldots
1156: $$
1157: Observe that ${\bf a}$ is either ultimately periodic, or it
1158: satisfies Condition $(*)_{w, w'}$ with $w=2$ and
1159: $w' = 2/(\lambda - 1)$.
1160:
1161: If ${\bf a}$ is not eventually periodic, then
1162: Theorem 3.1 implies the transcendence of the real number
1163: $\alpha = [0; a_1, a_2, \ldots ]$, provided
1164: $\lambda$ is sufficiently large in terms of $b_k$.
1165: It turns out that the method
1166: of proof of Theorem 3.1 is flexible enough to yield
1167: in some cases much better results: we can get a
1168: condition on $\lambda$ that does not depend on the values of the $b_j$'s.
1169:
1170: \proclaim Theorem 8.1. Keep the preceding notation.
1171: Assume furthermore that $|W_n|$ is odd for all
1172: sufficiently large $n$ and that $\lambda > 3.26$. Then, the real number
1173: $$
1174: \alpha = [0; a_1, a_2, \ldots ]
1175: $$
1176: is either quadratic, or transcendental.
1177:
1178: In order to establish Theorem 8.1, we need three lemmas.
1179: We keep the notation from Section 3 of \cite{ADQZ}.
1180: In particular, for an integer matrix $A$, we denote by $\rho(A)$
1181: its spectral radius and by $||A||$ its $L^2$-norm.
1182: Recall that $\rho(A) = ||A||$ when $A$ is symmetrical.
1183: Our first auxiliary result is extracted from \cite{ADQZ}.
1184: In all what follows, we set $\gamma = 0.885$.
1185:
1186: \proclaim Lemma 8.2. If $A = \pmatrix{ a & 1 \cr 1 & 0 \cr}$
1187: and $B = \pmatrix{ b & 1 \cr 1 & 0 \cr}$ where $a$ and $b$
1188: are distinct positive integers, then we have
1189: $\rho(AB) > \bigl( \rho(A) \rho(B) \bigr)^{\gamma}$.
1190:
1191:
1192:
1193:
1194: For $j=1, \ldots, k$, set $B_j = \pmatrix{ b_j & 1 \cr 1 & 0 \cr}$.
1195: Set also
1196: $$
1197: X = {1 \over k} \, \sum_{j=1}^k \, \log \rho (B_j).
1198: $$
1199: For any finite word $V$ on $\Sigma$, denote by $K(V)$ the
1200: corresponding continuant.
1201:
1202: \proclaim Lemma 8.3. If $V$ is in $\Sigma^+_{\hbox{\sevenrm equal}}$,
1203: then we have
1204: $$
1205: {1 \over |V|} \, \log K(V) \le X.
1206: $$
1207:
1208: \medskip
1209:
1210: \noindent {\bf Proof.} Let $V = d_1 d_2 \ldots d_m$ be a finite word
1211: defined over the alphabet $\Sigma^+_{\hbox{\sevenrm equal}}$. Set
1212: $p_{m-1}/q_{m-1} = [0; d_1, \ldots, d_{m-1}]$ and
1213: $p_m/q_m = [0; d_1, \ldots, d_m]$.
1214: Then, we have $K(V) = q_m$ and
1215: $$
1216: K(V) \le \biggl \Vert \pmatrix{ q_m & q_{m-1} \cr p_m & p_{m-1} \cr}
1217: \biggr \Vert.
1218: $$
1219: Setting $h_j = |V|_{b_j}$ for $j=1, \ldots, k$, it follows from
1220: the theory of continued fractions that
1221: $$
1222: K(V) \le ||B_1||^{h_1} \ldots || B_k||^{h_k}.
1223: $$
1224: Since the $B_j$'s are symmetrical and
1225: $h_1 = \ldots = h_k = h$, we have
1226: $$K(V) \le \rho(B_1)^{h_1} \ldots \rho(B_k)^{h_k}. $$
1227: Hence, the proof. \cqfd
1228:
1229: Our last auxiliary result is the following.
1230:
1231: \proclaim Lemma 8.4. If $V$ is in $\Sigma^+_{\hbox{\sevenrm equal}}$
1232: with $|V|$ odd, then we have
1233: $$
1234: {1 \over |V|} \, \log K(V) > \gamma X - {\log 4 \over |V|}.
1235: $$
1236:
1237: \medskip
1238:
1239: \noindent {\bf Proof.} We use repeatedly a particular case
1240: of Theorem 3.4 from \cite{Dav}.
1241: It asserts that if $W$ is the product of an odd number $m$
1242: of matrices $B_1, \ldots , B_k$, each of which occurring
1243: exactly $\ell$ times, then we have
1244: $$
1245: {\rm tr} (W) \ge \rho(B_1 B_k)^{\ell} \,
1246: {\rm tr} (W'),
1247: $$
1248: where $W'$ is the product arising from $W$ by replacing the
1249: matrices $B_1$ and $B_k$ by the identity matrix. As usual,
1250: ${\rm tr}(M)$ denotes the trace of the matrix $M$.
1251:
1252: With the notation of the proof of Lemma 8.3, we then get that
1253: $$
1254: \eqalign{
1255: K(V) & \ge {1 \over 2} \, {\rm tr}
1256: \pmatrix{ q_m & q_{m-1} \cr p_m & p_{m-1}}
1257: \ge {1 \over 2} \, \rho(B_1 B_k)^h \ldots
1258: \rho (B_{(k-1)/2} B_{(k+3)/2})^h \, {\rm tr} (B_{(k+1)/2}^h) \cr
1259: & \ge {1 \over 4} \, \rho(B_1 B_k)^h \ldots
1260: \rho (B_{(k-1)/2} B_{(k+3)/2})^h \rho (B_{(k+1)/2}^h) \cr
1261: & > {1 \over 4} \, \bigl( \rho(B_1) \ldots \rho (B_k) \bigr)^{\gamma h}, \cr}
1262: $$
1263: by Lemma 8.2. The lemma follows. \cqfd
1264:
1265: \medskip
1266:
1267: We have now all the tools needed to establish Theorem 8.1.
1268:
1269: \medskip
1270:
1271: \noindent {\bf Proof of Theorem 8.1.}
1272: For any $n \ge 2$, set
1273: $$
1274: U_n = W_1 W_2^2 \ldots W_{n-1}^2 \quad
1275: {\rm and} \quad V_n = W_n.
1276: $$
1277: Clearly, ${\bf a}$ begins in $U_n V_n^2$.
1278: Denote by $K(U_n)$ and by $K(V_n)$ the continuants associated to
1279: the words $U_n$ and $V_n$, respectively. In view of Lemma 3.2 and of (3.3),
1280: the theorem is proved as soon as we establish that there exists
1281: a positive real number $\eps$ such that
1282: $$
1283: \log K(V_n) > (1 + \eps) \log K(U_n), \eqno (8.1)
1284: $$
1285: for any sufficiently large integer $n$.
1286:
1287: To prove (8.1), we first infer from Lemmas 3.2
1288: and 8.3 that
1289: $$
1290: \log K(U_n) < \log K(V_1) + 2 \, \sum_{j=2}^{n-1} \,
1291: \log K(V_j) + 2 n < 2 X \, \sum_{j=1}^{n-1} \, |V_j| + 2 n.
1292: $$
1293: Consequently, we get
1294: $$
1295: {1 \over |V_n|} \, \log K(U_n) < {2X \over \lambda - 1} +
1296: {2 n \over |V_n|}. \eqno (8.2)
1297: $$
1298: On the other hand, Lemma 8.4 gives us that
1299: $$
1300: {1 \over |V_n|} \, \log K(V_n) > \gamma X - {\log 4 \over |V_n|}. \eqno (8.3)
1301: $$
1302: We then infer from (8.2) and (8.3) that (8.1) is satisfied for some
1303: positive $\eps$ as soon as we have $\gamma > 2/(\lambda - 1)$,
1304: that is, $\lambda > 1 + 2 \, \gamma^{-1} \asymp 3.25 \ldots$
1305: This completes the proof of the theorem. \cqfd
1306:
1307:
1308:
1309:
1310: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
1311:
1312:
1313:
1314:
1315: \vskip 12mm
1316:
1317: \centerline{\bf References}
1318:
1319: \vskip 7mm
1320:
1321: \beginthebibliography{999}
1322:
1323: \bibitem{Adam}
1324: B. Adamczewski,
1325: {\it Transcendance \og \`a la Liouville \fg de certain nombres r\'eels},
1326: C. R. Acad. Sci. Paris {\bf 338} (2004), 511--514.
1327:
1328: \bibitem{AdBuAoM}
1329: B. Adamczewski \& Y. Bugeaud,
1330: {\it On the complexity of algebraic numbers I.
1331: Expansions in integer bases}.
1332: Annals of Math. To appear.
1333:
1334: \bibitem{AdBuAM}
1335: B. Adamczewski \& Y. Bugeaud,
1336: {\it On the complexity of algebraic numbers II. Continued
1337: fractions}, Acta Math. To appear.
1338:
1339: \bibitem{AdBuMB}
1340: B. Adamczewski \& Y. Bugeaud,
1341: {\it On the Maillet--Baker continued fractions}. Preprint.
1342:
1343: \bibitem{ABL}
1344: B. Adamczewski, Y. Bugeaud \& F. Luca,
1345: {\it Sur la complexit\'e des nombres alg\'ebriques},
1346: C. R. Acad. Sci. Paris {\bf 339} (2004), 11--14.
1347:
1348:
1349: \bibitem{Allouche}
1350: J.-P. Allouche,
1351: {\it Nouveaux r\'esultats de transcendance de r\'eels \`a
1352: d\'eveloppements non al\'eatoire},
1353: Gaz. Math. {\bf 84} (2000), 19--34.
1354:
1355: \bibitem{ADQZ}
1356: J.-P. Allouche, J. L. Davison, M. Queff\'elec \& L. Q. Zamboni,
1357: {\it Transcendence of Sturmian or morphic continued fractions},
1358: J. Number Theory {\bf 91} (2001), 39--66.
1359:
1360: \bibitem{Allouche_Shallit98}
1361: J.-P. Allouche \& J. O. Shallit,
1362: {\it Generalized Pertured Symmetry},
1363: Europ. J. Combinatorics {\bf 19} (1998), 401--411.
1364:
1365: \bibitem{Allouche_Shallit}
1366: J.-P. Allouche \& J. O. Shallit,
1367: Automatic Sequences: Theory, Applications, Generalizations,
1368: Cambridge University Press, Cambridge, 2003.
1369:
1370:
1371: \bibitem{Baker62}
1372: A. Baker,
1373: {\it Continued fractions of transcendental numbers},
1374: Mathematika {\bf 9} (1962), 1--8.
1375:
1376: \bibitem{Baker64}
1377: A. Baker,
1378: {\it On Mahler's classification of transcendental numbers},
1379: Acta Math. {\bf 111} (1964), 97--120.
1380:
1381: \bibitem{Baum_Sweet1}
1382: L. E. Baum \& M. M. Sweet,
1383: {\it Continued fractions of algebraic power series in characteristic
1384: $2$},
1385: Annals of Math. {\bf 103} (1976), 593--610.
1386:
1387: \bibitem{Bax03}
1388: C. Baxa,
1389: {\it Extremal values of continuants and transcendence of certain
1390: continued fractions},
1391: Adv. in Appl. Math. {\bf 32} (2004), 754--790.
1392:
1393:
1394: \bibitem{DavRoth}
1395: H. Davenport \& K. F. Roth,
1396: {\it Rational approximations to algebraic numbers},
1397: Mathematika {\bf 2} (1955), 160--167.
1398:
1399:
1400: \bibitem{Dav89}
1401: J. L. Davison,
1402: {\it A class of transcendental numbers with bounded partial quotients}.
1403: In R. A. Mollin, ed., Number Theory and Applications, pp. 365--371, Kluwer
1404: Academic Publishers, 1989.
1405:
1406:
1407: \bibitem{Dav}
1408: J. L. Davison,
1409: {\it Continued fractions with bounded partial quotients},
1410: Proc. Edinburgh Math. Soc. {\bf 45} (2002), 653--671.
1411:
1412:
1413: \bibitem{DeMevdP}
1414: F. M. Dekking, M. Mend\`es France \& A. J. van der Poorten,
1415: {\it Folds!}, Math. Intelligencer {\bf 4} (1982), 130--138, 173--181,
1416: 190--195. Erratum, {\bf 5} (1983), 5.
1417:
1418: \bibitem{Evertse97}
1419: J.-H. Evertse,
1420: {\it The number of algebraic numbers of given degree approximating
1421: a given algebraic number}. In: Analytic number theory (Kyoto, 1996),
1422: 53--83, London Math. Soc. Lecture Note Ser. 247,
1423: Cambridge Univ. Press, Cambridge, 1997.
1424:
1425: \bibitem{Fog}
1426: N. Pytheas Fogg,
1427: Substitutions in Dynamics, Arithmetics and Combinatorics,
1428: Lecture Notes in Mathematics 1794, Springer-Verlag, 2002.
1429:
1430: \bibitem{Hartmanis_Stearns}
1431: J. Hartmanis \& R. E. Stearns,
1432: {\it On the computational complexity of algorithms},
1433: Trans. Amer. Math. Soc. {\bf 117} (1965), 285--306.
1434:
1435:
1436: \bibitem{Khintchine}
1437: A. Ya. Khintchine,
1438: Continued fractions, Gosudarstv. Izdat. Tehn.-Theor. Lit.
1439: Moscow-Leningrad, 2nd edition, 1949 (in Russian).
1440:
1441:
1442: \bibitem{Lang}
1443: S. Lang,
1444: Introduction to Diophantine Approximations, Sprin\-ger-Verlag, 1995.
1445:
1446: \bibitem{Leveque56}
1447: W. J. LeVeque,
1448: Topics in number theory, Vol. II, Addison-Wesley, 1956.
1449:
1450: \bibitem{LiSt}
1451: P. Liardet \& P. Stambul,
1452: {\it S\'eries de Engel et fractions continu\'ees},
1453: J. Th\'eor. Nombres Bordeaux {\bf 12} (2000), 37--68.
1454:
1455: \bibitem{Liouville} J.~Liouville, {\it Sur des classes
1456: tr\`es \'etendues de quantit\'es
1457: dont la valeur n'est ni alg\'ebri\-que, ni m\^eme r\'eductible
1458: \`a des irrationelles}
1459: {\it alg\'e\-bri\-ques},
1460: C. R. Acad. Sci. Paris {\bf 18} (1844), 883--885; 910-911.
1461:
1462:
1463: \bibitem{LvdP77}
1464: J. H. Loxton \& A. J. van der Poorten,
1465: {\it Arithmetic properties of certain functions in several variables
1466: III}, Bull. Austral. Math. Soc. {\bf 16} (1977), 15--47.
1467:
1468:
1469:
1470:
1471: \bibitem{Maillet}
1472: E. Maillet,
1473: Introduction \`a la th\'eorie des nombres transcendants et des propri\'et\'es
1474: arithm\'etiques des fonctions, Gauthier-Villars, Paris, 1906.
1475:
1476:
1477: \bibitem{Mendes}
1478: M. Mend\`es France,
1479: {\it Principe de la sym\'etrie perturb\'ee}.
1480: In: {S\'eminaire de Th\'eorie des
1481: Nombres, Paris 1979-80}, M.-J. Bertin (\'ed.), Birkh\"auser, Boston, 1981,
1482: pp. 77--98.
1483:
1484:
1485:
1486: \bibitem{Perron}
1487: O. Perron,
1488: Die Lehre von den Ketterbr\"uchen.
1489: Teubner, Leipzig, 1929.
1490:
1491:
1492: \bibitem{vdP_Shallit}
1493: A. J. van der Poorten \& J. O. Shallit,
1494: {\it Folded continued fractions},
1495: J. Number Theory {\bf 40} (1992), 237--250.
1496:
1497: \bibitem{Queffelec98}
1498: M. Queff\'elec,
1499: {\it Transcendance des fractions continues de Thue--Morse},
1500: J. Number Theory {\bf 73} (1998), 201--211.
1501:
1502:
1503: \bibitem{Queffelec00}
1504: M. Queff\'elec,
1505: {\it Irrational number with automaton-generated continued fraction expansion},
1506: In J.-M. Gambaudo, P. Hubert, P. Tisseur, and S. Vaienti, editors,
1507: {\it Dynamical Systems: From Crystal to Chaos}, World Scientific,
1508: 2000, 190--198.
1509:
1510:
1511:
1512:
1513: \bibitem{Ridout}
1514: D. Ridout,
1515: {\it Rational approximations to algebraic numbers},
1516: Mathematika {\bf 4} (1957), 125--131.
1517:
1518:
1519:
1520: \bibitem{Rudin}
1521: W. Rudin,
1522: {\it Some theorems on Fourier coefficients},
1523: Proc. Amer. Math. Soc. {\bf 10} (1959), 855--859.
1524:
1525:
1526: \bibitem{Schm67}
1527: W. M. Schmidt,
1528: {\it On simultaneous approximations of two algebraic numbers by rationals},
1529: Acta Math. {\bf 119} (1967), 27--50.
1530:
1531: \bibitem{Schmidt72a}
1532: W. M. Schmidt,
1533: {\it Norm form equations},
1534: Annals of Math. {\bf 96} (1972), 526--551.
1535:
1536: \bibitem{Schmidt80}
1537: W. M. Schmidt,
1538: {\it Diophantine approximation},
1539: Lecture Notes in Mathematics 785, Springer, Berlin, 1980.
1540:
1541:
1542: \bibitem{Shallit_survey}
1543: J. O. Shallit,
1544: {\it Real numbers with bounded partial quotients},
1545: Enseign. Math. {\bf 38} (1992), 151--187.
1546:
1547:
1548: \bibitem{Shapiro}
1549: H. S. Shapiro,
1550: {\it Extremal problems for polynomials and power series},
1551: Master's thesis, MIT, 1952.
1552:
1553:
1554: \bibitem{Waldschmidt}
1555: M. Waldschmidt,
1556: {\it Un demi-si\`ecle de transcendance}. In:
1557: Development of mathematics 1950--2000, pp. 1121--1186,
1558: Birkh\"auser, Basel, 2000.
1559:
1560: \endthebibliography
1561:
1562:
1563: \vskip 6mm
1564:
1565: \noindent Boris Adamczewski \hfill{Yann Bugeaud}
1566:
1567: \noindent CNRS, Institut Camille Jordan
1568: \hfill{Universit\'e Louis Pasteur}
1569:
1570: \noindent Universit\'e Claude Bernard Lyon 1
1571: \hfill{U. F. R. de math\'ematiques}
1572:
1573: \noindent B\^at. Braconnier, 21 avenue Claude Bernard
1574: \hfill{7, rue Ren\'e Descartes}
1575:
1576: \noindent 69622 VILLEURBANNE Cedex (FRANCE)
1577: \hfill{67084 STRASBOURG Cedex (FRANCE)}
1578:
1579: \vskip2mm
1580:
1581: \noindent {\tt Boris.Adamczewski@math.univ-lyon1.fr}
1582: \hfill{{\tt bugeaud@math.u-strasbg.fr}}
1583:
1584: \vskip 6mm
1585:
1586: \centerline{Les Davison}
1587:
1588: \centerline{Department of Mathematics and Computer Science}
1589:
1590: \centerline{Laurentian University}
1591:
1592: \centerline{Sudbury, Ontario}
1593:
1594: \centerline{CANADA P3E 2C6}
1595:
1596: \vskip2mm
1597:
1598: \centerline{{\tt ldavison@cs.laurentian.ca}}
1599:
1600:
1601: \bye
1602:
1603:
1604: