Application of a cyclic determinant in the theory of cubic residues (Q788762)

Let \(p\equiv 1 \bmod 3\) be a prime, \(\rho =\exp(2\pi i/3)\), \(\zeta =\exp(2\pi i/p)\), \(\omega\) a prime divisor of \(p\) in \(\mathbb Z[\rho]\) with \(\omega\equiv -1 \bmod 3\) and \(\text{Im}\,\omega >0\), and \(f\equiv \rho \bmod \omega\), \(f\in\mathbb Z\). Denote \[ \Omega =p^{-1}\prod^{p- 1}_{s=1}(\zeta^ s+\rho \zeta^{sf}+\rho^ 2\zeta^{sf^ 2}). \] It is easy to see that \(\Omega =\alpha^ 3\omega\) for some \(\alpha\) in \(\mathbb Z[\rho]\). In an earlier paper [Mat. Zametki 7, 469--476 (1970; Zbl 0194.35007)] the author conjectured that \(\text{Im}\,\Omega >0\), and a partial result to that effect has been obtained by \textit{J. H. Loxton} [J. Reine Angew. Math. 268/269, 53--67 (1974; Zbl 0293.10019)]. In the present paper the author computes an expression for \(\Omega\) by means of a cyclic determinant. There is a further conjecture stating that \(| \arg \Omega -\arg(-{\bar \omega})|<\pi /3\) where \(\arg z\) satisfies \(-\pi<\arg z\leq \pi\). The conjecture is true for sufficiently large \(p\) by Loxton's result. The author has verified by direct computation that it is also true for \(p<1000\).