Itération contrôlée de la fonction \(\sigma\) (Q1166554)

Sequences \(n_0, n_1, \ldots, n_i, n_{i+1}, \ldots \) of positive integers \(\ge 3\) are studied, defined by \(n_{i+1}:= S(n_i)\), \(i\ge 0\), where \(S(n):= \sigma(n/2)\) if \(n\) is even, \(\sigma(n)\) if \(n\) is odd, \(\sigma(\cdot)\) being the sum of divisors function. Various computational and theoretical results are given concerning the possible behaviour of such sequences. We mention: the 1498 sequences with starting term \(n_0 \in [3,1500]\) become periodic with period length 1 or 2, or coincide from some term onwards with one of the eight sequences with starting term \(n_0 \in \{135, 199, 315, 441, 567, 651, 1029, 1203\}\). The latter eight sequences grow quite rapidly and show no periodic part among their first 45 terms. For the sequence with starting term\(n_0 =135\) the author gives \(n_{44} =1479010806\) \((= 2.3^37^213.19.31.73)\). [The reviewer has pursued this sequence until \(n_{75} = 2^93^55^37^213^337.127.757\). It turns out that the sequences studied here are easier to compute than other `iterating' sequences like, e.g., aliquot sequences [cf. \textit{R. K. Guy} and \textit{J. L. Selfridge}, Math. Comput. 29, 101--107 (1975; Zbl. 296.10007); ibid. 34, 319--321 (1980; Zbl 0423.10005)]. The reason for this lies in the way the sum of divisors function \(\sigma\) is involved in the definition of \(S\) above, which very often causes the odd prime factors of \(n_i\) to be transformed by \(S\) into smaller prime factors of \(n_{i+1}\)]. This is a nice, clearly written short paper which leaves room for further research on an interesting topic.