Continued fractions and numeration in the Fibonacci base (Q860450)

This work is devoted to the numeration systems in a non-integral base, introduced firstly by \textit{A. Renyi} [Acta Math. Acad. Sci. Hung. 8, 477--493 (1957; Zbl 0079.08901)] and \textit{W. Parry} [Acta Math. Acad. Sci. Hung. 11, 401--416 (1960; Zbl 0099.28103)]. Let \(\beta >1\) and \(x\in[0,1]\). Then the expansion \(=\sum_{k\in N*}\nu_{k}\beta^{-k}\) , where \(\nu_{k}\in{0,\dots, [\beta]}\) , is called expansion of \(x\) in base \(\beta\). The greatest sequence for the lexicographical order \(d_{\beta}(x)\) is called \(\beta\)-expansion of \(x\). Any \(x>1\) can be uniquely expanded as \(x=\sum^{n}_{k=0}\beta^{k}+\sum_{k\in N^*}\beta^{-k}\); the sum of non-negative powers of \(\beta\) is called \({\beta}\)-integer part of \(x\) and is denoted by \([x]_{\beta}\), the corresponding sum of negative power of \(\beta\) is called \(\beta\) -fractionary part of \(x\) and is denoted by \({x}_{\beta}\). The numbers \([x]_{\beta}\), which are non-negative, are called \({\beta}\)-integers and their set is denoted by \(Z^{+}_{\beta}\). The author concentrates about the base \(\phi=(1+\sqrt 5)/2\) (the golden mean). The final goal is to prove a conjecture of Akiyama, that the positive real numbers with finite continued \(\phi\)-fractions are the positive elements of \(Q(\phi)\) . If \(x>0\) and if there are only finitely many non-zero elements in \(d_{\phi}(x)\), then one say that \(x\) has a finite \(\phi\)-expansion (omitting the ending of consecutive zeros) and the set of such \(x\) is denoted by \(Fin(\phi)\). If \(0\neq x\in Fin(\phi)\) and \(d_{\phi}(x)=v_{N}v_{N-1}\ldots v_{1}v_{0}.v_{-1}\ldots v_{-N}\), then the quantity \(N+1\) is called \(\phi\)-integer length of \(x\); \(N'\) denotes the \({\phi}\)-fractional part of \(x\); \(N+N'+1\) denotes the global length of \(x\). If \(p, q\in Z^{+}_{\phi}\), then \(p/q\in Q(\phi)\) is called \(\phi\)-fraction, and the set of \(\phi\)-fractions is denoted by \(Q^{+}_{\phi}\). Continued fractions are constructed by convenient generalization of the algorithm of Euclid: for every \(x\in R_{+}\), let \(x_{0}=x\) and \(x_{1}=1/(x_{0}-[x_{0}]_{\phi})\) if \(x_{0}-[x_{0}]_{\phi}>0\), otherwise the algorithm ends, or more generally, at step \(i, x_{i+1}=1/(x_{i}-[x_{i}]_{\phi})\) if \(x_{i}\) is not a \(\phi\)-integer, otherwise the algorithm ends. Analogous to the existing terminology, the \(\phi\)-integer \(a_{n}[x_{n}]_{\phi}\) is called \(n\)th partial \(\phi\)-quotient of \(x\), \(c_{n}=[a_{0};a_{1}\ldots,a_{n}]\) is called \(n\)th partial \(\phi\)-convergent of \(x\). The expansion \(a_{0}+1/(a_{1}+\ldots 1/(a_{n}+\ldots))\) is called the continued \(\phi\)-fraction of \(x\). The author proves that the sequence of \(\phi\)-convergents of \(x\) tends to \(x\) (Proposition 4.5) and that \(Q^{+}_{\phi}=Q(\phi)^{+}\) (Proposition 4.6) and others. The proof of the Akiyama conjecture (Theorem 5.3) consists of two steps. First step: The author proves that the continued \(\phi\)-fraction of any \(x\in Q(\phi)^{+}\) is either ultimately periodic or finite. This step needs existence of cycles in a special algorithm, which constructs a sequence of \(\phi\)-fractionary expansions \((p_{i},q_{i})_{i}\) such that for all \(i\), \(x_{i}=p_{i}/q_{i}\), where \((x_{i})_{i}\) is a sequence of elements of \(Q(\phi)^{+}\). Second step consists of computation of these cycles. The author shows that these cycles correspond to quadratic numbers over \(Q(\phi)\) which do not belong to \(Q(\phi)\) itself. He proves that an eventually periodic case is not possible. In such a way, Theorem 5.3 (Akiyama) says more categorically: the continued \(\phi\)-fraction of \(x\) is finite if and only if \(x\in Q(\phi)^{+}\).