Properties of rational arithmetic functions (Q2368514)

\((r,s)\)-rational arithmetic functions \(f\) [abbreviated \(f \in {\mathcal C}(r,s)\)] are arithmetical functions which can be written als \[ f = g_1 \ast \dots \ast g_r \ast h_1^{-1} \ast \dots \ast h_s^{-1}, \] where the \(g_\rho\) and the \(h_\sigma\) are completely multiplicative functions, with Dirichlet convolution \(\ast\), and where \(h^{-1}\) denotes the convolution inverse of \(h\). Firstly the authors give some [complicated] characterizations of functions in \({\mathcal C}(r,s)\). For example, in the very special case \(g_1 = \dots = g_r = g\), \(h_1 = \dots = h_s = h\) [such functions are called \((r,s)\)-totients], the authors show: \(f\) is an \((r,s)\)-totient if and only if for any prime \(p\) and any \(\alpha>2\) there are complex numbers \(a(p),\, b(p)\), for which \[ f(p^\alpha) = (-1)^\alpha \sum_{i=0}^\alpha {{-r} \choose {\alpha - i}} {s \choose i} a(p)^{\alpha - i}\, b(p)^i. \] Results from this section are used to prove Busche-Ramanujan identities, for example for functions in \({\mathcal C}(1,s)\): There exists a multiplicative function \(F\) such that for each \(s\)-excessive pair \((m,n)\) [that means, for every prime \(p\) dividing \((m,n)\) either \(\nu_p(m) \geq \nu_p(n) + s\) or \(\nu_p(n) \geq \nu_p(m) + s\)] \[ f(mn) = \sum_{d \mid (m,n)} f\left({m\over d}\right) \cdot f\left({n\over d}\right) \cdot F(d), \] if (for \(g(p)\not=0\)) \[ \sum_{k=0}^s g(p)^{s-k} \cdot \left\{ (-1)^k \sum_{1\leq i_1< i_2 < \dots < i_k \leq s} h_{i_1}(p) \cdots h_{i_k}(p) \right\} \not= 0. \] For functions in \({\mathcal C}(3,0)\) necessary conditions are given for the existence of a [certain] Busche-Ramanujan identity. For functions in \({\mathcal C}(2,0)\), \( f = g_1 \ast g_2\), \textit{binomial identities}\ are known, \[ f(p^k) = \sum_{j=0}^{[k/2]} (-1)^j {{k-j}\choose j} f(p)^{k-2j} \Bigl(g_1(p)g_2(p)\Bigr)^j. \] The authors give [complicated] binomial identities for functions in \({\mathcal C}(2,s)\), which cannot be reproduced here. The authors state \textit{that it seems unlikely for functions in \({\mathcal C}(r,s)\) with \(r>2\) to have similar binomial-type identities}. In the last section the authors prove that for functions \(f \in {\mathcal C}(r,s)\) the Kesava Menon norm \(Nf\) is again in \({\mathcal C}(r,s)\), and they deduce several corollaries. This Kesava Menon norm is an arithmetical function defined by \[ N(f): n \mapsto (f \ast \lambda\cdot f)(n^2),\; \text{ with the Liouville function }\lambda = (-1)^\Omega. \] The authors state that the theorem just quoted was already given in P. Haukkanen's review of a paper of \textit{S. Thajoddin} and \textit{S. Vangipuram} in [Indian J. Pure Appl. Math. 19, 1156--1161 (1988; Zbl 0664.10001)].