The Petersson-Knopp identity for homogeneous Dedekind sums (Q1912285)

Homogeneous Dedekind sums are defined by \[ S(a, b, q)= \sum_{r\text{ mod } q}\Biggr(\Biggr({ar\over q}\Biggr)\Biggr) \Biggl(\Biggl( {br\over q}\Biggr)\Biggr), \] where \(((x))= x- [x]- {1\over 2}\) if \(x\neq\) integer, \(((x))= 0\) otherwise. It is shown that for integers \(a\), \(b\), \(q\) with \(q> 0\), we have \[ \sum_{d|n} \sum^d_{r_1= 1} \sum^d_{r_2=1} S\Biggl( {n\over d} a+ r_1 q, {n\over d} b+ r_2 q, dq)= n\sigma(n) S(a, b, q), \] where \(\sigma(n)\) is the sum of the positive divisors of \(n\). This generalizes a result of \textit{M. Knopp} [J. Number Theory 12, 2-9 (1980; Zbl 0423.10015)] who treated the case \(b= 1\). A similar generalization holds for sums of Dedekind type introduced by the reviewer and \textit{T. Vu} [J. Number Theory 14, 391-396 (1982; Zbl 0487.10012)]. The methods used are elementary.