The parity of Farey denominators and the Farey index (Q817251): Difference between revisions

For a positive integer \(N\) let \({\mathcal F}_N=(x_0,x_1,\dots,x_R)= (\frac 01,\frac 1N,\dots,\frac 11)\) denote the Farey sequence of order \(N\), where \(R\) equals the sum of the values of the Euler function up to \(N\). If \(x_{i-1}=\frac ar\), \(x_i= \frac bs\), \(x_{i+1}=\frac ct\) in lowest terms then the integer \(\nu(x_i)=\frac {r+t}{s}=\frac{a+c}{b}=\lfloor \frac{N+r}{s}\rfloor\) is called the Farey index of \(x_i\) in \({\mathcal F}_N\). In an earlier paper [the author and \textit{P. Stein}, Mich. Math. J. 51, 209--223 (2003; Zbl 1038.11062)], it was shown that \(\sum_{1\leq i\leq R}\nu(x_i)=3R-1\). One of the results of the present paper is \(\sum_{1\leq i\leq R,s=s(x_i)\text{even}}\nu(x_i)=R-1\), and the corresponding formula for the sum with odd denominators \(s(x_i)\). There is a relation between Farey indices and a function \(E(b,s)\) on pairs of coprime integers \(b,s\) which is defined by a complicated recursion but which conjecturally is given by an explicit formula resembling the formula for Dedekind sums. There is a reciprocity formula for \(E(b,s)\) similar to that for Dedekind sums. The author proves sharp bounds for \(E(b,s)\) which imply that weighted partial sums of Farey indices are smaller than expected.