Some non-special cubic fourfolds (Q1650672)

Consider \(D_{\text{V-ap}}\), the set of complex cubic fourfolds in \(\mathbb P^5\) which are apolar to a Veronese surface; \(D_{\text{IR}}\), the set of cubics apolar to a quartic scroll; and \(D_{\text{rk3}}\), the closure of the set of cubics apolar to the union of a plane and a disjoint hyperplane. In [Doc. Math. 22, 455--504 (2017; Zbl 1369.14056)] \textit{K. Ranestad} and \textit{C. Voisin} showed that these are irreducible divisors in the moduli space of cubics and they proved that \(D_{\text{V-ap}}\) is not a Noether-Lefschetz divisor, i.e. \(H^4(X,\mathbb Q)_{\mathrm{prim}}\) contains no algebraic part for the general point \(X \in D\). The present paper demonstrates that being not Noether-Lefschetz holds true also for the other two divisors. The idea is to use reduction modulo \(2\), it works to give a new proof for the case of \(D_{\text{V-ap}}\,\). The strategy is first to produce an explicit equation for one cubic \(X\) in each divisor and to check that \(X\) has good reduction modulo \(2\). Next one counts the number of points in the reduction over \(F_{2^q}\), which can be done effectively for small \(q\) by exploiting the conic bundle structure of the cubic \(X\). These numbers, for \( q= 1 , \dots, 11 \), are then plugged in the formula from the Weil hypothesis. In this way the authors compute the characteristic polynomial for the action of Frobenius on the Tate twist of the primitive étale cohomology \( V := H^4_{\text{ét, prim}}(X_{\bar k}, \mathbb Q_\ell(2))\). The polynomial so determined is found to be irreducible. On the other hand the authors show that if \(H^4(X,\mathbb Q)_{\mathrm{prim}}\) had non trivial algebraic part then so would \(V\), but this implies that the characteristic polynomial should contain a cyclotomic factor. The required computations are performed using mostly the software system Macaulay2.