Birational geometry of Fano hypersurfaces of index two (Q330877)

Let \(V \subset \mathbb{P}^{M + 1}\) be a hypersurface given by sufficiently (Zariski) general polynomial over \(\mathbb{C}\) of degree \(M\). Let also \(\chi: V \dashrightarrow Y\) be a rational map with rationally connected fibers. In the paper under review, the author proves that for \(M \geq 16\) the map \(\chi\) is induced by the linear projection \(\mathbb{P}^{M + 1} \dashrightarrow \mathbb{P}^1\) from a codimension \(2\) subspace, so that in particular \(Y = \mathbb{P}^1\) (see Theorem 1.1 in the text). The argument goes by the precise description of maximal singularities of a mobile linear system \(\Sigma \subset |2n\mathcal{O}_V(1)|\). This is Theorem 1.3 in the text. More precisely, it states that if \(\Sigma\) satisfies \(c_{\text{virt}}(\Sigma) < n\) (here \(c_{\text{virt}}\) is the \textit{virtual threshold of canonical adjunction} defined in Sec. 2.1 of [\textit{A. Pukhlikov}, Birationally rigid varieties. Providence, RI: American Mathematical Society (2013; Zbl 1297.14001)]), then there exists a codimension two subspace \(P \subset \mathbb{P}^{M + 1}\) such that \(\Sigma\) has multiplicity \(>n\) along \(P \cap V\), whereas for the strict transform \(\Sigma_P\) on the blowup \(V_P \longrightarrow V\) at \(P \cap V\) one has \(c_{\text{virt}}(\Sigma) = c_{\text{virt}}(\Sigma_P) = c(V_P,\Sigma_P)\). The main result easily follows from Theorem 1.3 (cf. Lemmas 2.1 and 2.2). Theorem 1.3 is proved in two steps (cf. Theorems 1.4 and 1.5). First, the author shows that once \(\text{mult}_B(\Sigma) > n\) for some codimension two subvariety \(B \subset V\), then \(B = P \cap V\) as above. Second, he proves that if \(c_{\text{virt}}(\Sigma) < n\), then there exists such a \(B\). These two claims imply the equality \(c_{\text{virt}}(\Sigma_P) = c(V_P,\Sigma_P) =: m\) (which is precisely Theorem 1.3) as follows. Suppose that actually \(c_{\text{virt}}(\Sigma_P) < m\). The author reduces to proving that there is no effective divisor \(D \sim \mathcal{O}_W(m)\) in a fiber \(W\) of the resolved projection \(V_P \longrightarrow \mathbb{P}^1\) from \(P\) such that the pair \((W,\frac{1}{m}D)\) is not canonical at a singular point \(o \in W\) (this is standard by the results in \textit{op.\,cit}). Note that by the generality assumption \(W\) is a hypersurface of degree \(M\) in \(\mathbb{P}^M\) with at most isolated quadratic singularities. By blowing up \(o\) the argument is reduced to the case of a smooth quadric in place of \(W\) (cf. Proposition 2.3). Then one concludes as in [\textit{A. V. Pukhlikov}, Proc. Steklov Inst. Math. 264, 159--177 (2009; Zbl 1312.14097)]. The above two claims about the maximal singularity \(B\) are proved in Sections 3 -- 6. Firstly, if \(B\) were contained in a hyperplane, the claim that \(B = P \cap V\) is immediate. In turn, the case when \(B\) is not contained in a hyperplane is excluded by a thorough study of the secant variety \(\text{Sec}(B)\) (cf. Proposition 3.2). Further, the exclusion of the infinitely near singularity is given in Sections 4, 5 and 6. The argument reduces to the case when \(B\) is a point. The proof then goes by restricting \(\Sigma\) to a general hyperplane section \(P \subset V\) such that the \(8n^2\)-inequality is realized for \(\Sigma\big|_P\) (cf. Proposition 4.4). The argument concludes by applying the multiplicity counting technique (see e.g. Chapter 2 in [\textit{A. Pukhlikov}, Birationally rigid varieties. Providence, RI: American Mathematical Society (2013; Zbl 1297.14001)]).