On a class of arithmetic functions related to the divisors of an integer (Q5935914)

The author's paper deals with deep and difficult results concerning divisors of integers -- see for example the monographs \textit{Sets of Multiples}. Cambridge Tracts Math. 118. Cambridge Univ. Press (1996; Zbl 0871.11001), by \textit{R. R. Hall}, and \textit{Divisors}. Cambridge Tracts Math. 90. Cambridge Univ. Press (1988; Zbl 0653.10001) by \textit{R. R. Hall} and \textit{G. Tenenbaum}. Denote by \( {\mathcal D} (n) = \{d_1 < d_2 < \cdots < d_{\tau(n)} \}\) the set of divisors of \(n\). Define by \({\mathcal E} \) the class of arithmetical functions \(F\), where \[ F(n) = \# \bigl\{d; d|n,\;\text{gcd}\bigl(d,g_n (d)\bigr) = 1, \text{and } n \equiv 0 \bmod g_n(d)\bigr\}, \] where \( g_n : {\mathcal D}(n) \to {\mathbb Z} \) is some injective function with the property: If \( d|n, t|n\), \(\gcd(d,g_n (d)) = \text{gcd}(t, g_n (t)) = 1\), \(d g_n(d) = t g_n (t)\), and \( dg_n(d) |n\), then \( d= t.\) For example, Erdős' function \[ f(n) = \# \bigl\{i \in [1, \tau(n) [,\;\text{gcd}(d_i, d_{i+1}) = 1 \bigr\}, \] associated with the function \( g_n : d_i \mapsto d_{i+1}\) for \(1 \leq i < \tau(n)\), \(g_n(n) = n+1\), is in \({\mathcal E}\). The author proves, at first, that the function \[ E(n) = \max_{F\in {\mathcal E}} F(n) \] satisfies \( E(n) \cdot E(m) \leq E(nm)\), if gcd\((n,m)=1\) (Theorem 1). Next, the author improves G. Tenenbaum's result \[ E(n) \leq 3 \tau (n)^{c^\prime}, \text{ where } c^\prime = 1 - {1\over 2+36 \log\left({3\over 2}\right)} = 0.93974721 \dots \] to \[ E(n) \leq \tau(n)^c, \text{ where } c = {\log 3 \over\log 2} - {2\over 3} = 0.9182958 \dots, \] and he shows that \[ E(n) \leq \tau(n)\cdot \prod_{p^\nu\|n} \left( {2\sqrt{\nu}\over \nu+1} \right) . \] These estimates imply \[ E(n) = {\mathcal O}\left({\tau(n) \over \Omega(n)}\right). \] On the other hand, by a result of \textit{P. Erdős} and \textit{G. Tenenbaum} [Proc. Lond. Math. Soc. (3) 59, 417-438 (1989; Zbl 0694.10040)], \[ E(n_j) \geq {\tau(n_j) \over 2\Omega(n_j)} \] for some sequence \(n_j\) satisfying \(\tau(n_j) \to\infty\). \smallskip The ``proximity'' of the divisors, first studied by \textit{G. Tenenbaum} [Indag. Math., New Ser. 2, 105-114 (1991; Zbl 0728.11045)], is displayed by the estimate \[ \sum_{1 \leq i < j \leq n} {1 \over d_j - d_i} \ll \tau(n)^c \cdot \log\bigl(\tau(n)\bigr) \cdot \log\log \bigl( 2 \tau(n)\bigr). \] Finally, lower estimates (Theorem 3) show that the estimate \( E(n) \leq \tau^c(n) \) is optimal with respect to the exponent, since \[ \limsup_{\tau(n)\to\infty} \left( {\log E(n) \over \log \tau(n)} \right) = c . \] For the proofs, the author skilfully uses methods of Gerald Tenenbaum and a combinatorial lemma of \textit{Z. Baranyai} [Colloq. Math. Soc. János Bolyai 10, 91-108 (1975; Zbl 0306.05137)].