On a generalization of the Dedekind-Hölder theorem (Q539116): Difference between revisions

Ramanujan's sum is defined as \[ C(k, n)=\sum_{(x,n)=1} e^{2\pi ikx/n}, \] where \(x\) runs through a reduced residue system (mod \(n\)). The Von Sterneck function is defined as \[ \Phi(k, n)={\varphi(n)\over \varphi\left({n/(k, n)}\right)}\mu\left({n/(k, n)}\right), \] where \(\varphi\) is the Euler totient function and \(\mu\) is the Möbius function. The function \(\Phi(k, n)\) was encountered by Von Sterneck in 1902 in a study of restricted partitions with summands reduced to their least residues modulo \(k\). It is known that \[ C(k,n)=\Phi(k,n). \] R. Dedekind was the first mathematician to state this connection between \(C(k,n)\) and \(\Phi(k,n)\), and O. Hölder was the first to offer a proof. A survey of generalizations of this identity is given by \textit{P. Haukkanen} [Ann. Acad. Sci. Fenn., Ser. A I, Diss. 68, 69 p. (1988; Zbl 0651.10005)]. The present authors develop a further generalization arising from Schemmel's function. Schemmel's function \(\Phi_r(n)\) counts the number of sequences of \(r\) consecutive positive integers less than or equal to \(n\) and relatively prime to \(n\). The authors generalize the Von Sterneck function \(\Phi(k,n)\), using Schemmel's function, as \[ \Phi_r(k,n)={\varphi_r(n)\over\varphi_r\left({n/(k, n)}\right)}\mu\left({n/(k, n)}\right), \] where \(\varphi_r({n\over (k, n)})\neq 0\). Note that \(\varphi_1(k,n)=\varphi(k,n)\). On the other hand, the usual greatest common divisor function can be extended by defining the \((k,r)\)-gcd of \(x\) and \(n\) as \[ (x,n)_{k,r}=\gcd\left(x\left(x+{n\over n_{(k)}}\right)\dots\left(x+(r-1){n\over n_{(k)}}\right),n\right), \] where \ \[ n_{(k)}=\prod_{p^\alpha \| n \atop p^\alpha | k} p^\alpha. \] The authors generalize Ramanujan's sum \(C(k,n)\) as \[ C_r(k,n)=\sum_{(x,n)_{k,r}=1} e^{2\pi ikx/n}, \] where the sum is taken over all \(x\) with \(1\leq x \leq n\) and \((x,n)_{k,r}=1\). The main result of this paper is a generalized Dedekind-Hölder theorem: If the smallest prime factor of \(n\) is \(>r\), then \(C_r(k,n)=\Phi_r(k,n)\). The authors also present interesting identities for the function \(C_r(k,n)\) or \(\Phi_r(k,n)\).