More cubic surfaces violating the Hasse principle (Q719162)

This paper presents a class of counter-examples to the Hasse Principle for cubic surfaces, generalizing those given by \textit{H.P.F. Swinnerton-Dyer} [Mathematika, Lond. 9, 54--56 (1962; Zbl 0103.38302)] and \textit{L. J. Mordell} [J. Lond. Math. Soc. 40, 149--158 (1965; Zbl 0124.02605)]. Let \(p\equiv 1\pmod{3}\) be a prime, and let \(K\) be the unique cubic subfield of the cyclotomic field \(\mathbb{Q}(\zeta_p)\). The cubic surfaces considered then take the form \[ T_3(a_1T_0+d_1T_0)(a_2T_0+d_2T_0)=N_{K/\mathbb{Q}}(T_0+\theta T_1+\theta^2 T_2), \] where \(\theta=\text{tr}_{\mathbb{Q}(\zeta_p)/K}(\zeta_p-1)\), and \(a_1,a_2,d_1,d_2\in\mathbb{Z}\), as in the above-mentioned works. It is shown that the above surface has points everywhere locally, but has no rational point, under suitable conditions on \(a_1,a_2,d_1.d_2\). Specifically one requires (i) \(p\nmid d_1d_2\); (ii) \(\text{g.c.d.}(a_1,d_1)\) and \(\text{g.c.d.}(a_2,d_2)\) contain only primes which split in \(K\); (iii) \(T(a_1+d_1T)(a_2+d_2T)-1\in\mathbb{F}_p[T]\) has at least one simple zero in \(\mathbb{F}_p\); and (iv) if \(s\) is any zero of \(T(a_1+d_1T)(a_2+d_2T)-1\) in \(\mathbb{F}_p\) then \(a_1s^{-1}+d_1\) is a non-cube in \(\mathbb{F}_p\).