\(K\)-trivial structures on Fano complete intersections (Q2435981)

In the paper under review, the author studies Fano complete intersections \(V = (f_1 = \ldots = f_k) \subset \mathbb{P}^{M + k} =: \mathbb{P}\), where polynomials \(f_i\) are sufficiently general, of degree \(d_i\) satisfying \(d_1 + \ldots + d_k = M + k\). He proves that once \(M \geq 2k + 1\), except possibly for a finite number of \(k,d_1,\ldots,d_k\), every (birational) \textit{fiber space structure of nonmaximal relative Kodaira dimension} comes from a pencil of hyperplanes on \(V\) (see Theorem 1 in the text). More precisely, if \(\chi: V \dashrightarrow W\) is a birational map, where \(W\) admits a morphism \(\beta: W\longrightarrow S\) with Kodaira dimensions of \(S\) and of generic fiber summing up to be less than \(\dim W = M\), then \(\chi\circ\beta\) is simply a linear projection \(V\dashrightarrow\mathbb{P}^1\) and \(S = \mathbb{P}^1\) (there is a slight generalization of this result, stated in Theorem 2, for varieties with a pencil of Fano complete intersections). Here is a brief overview of the author's arguments. First, generality of \(V\) means that for \textit{any} linear subspace \(\Lambda\subset\mathbb{P}\) of codimension \(k + 1\) the intersection \(V\cap\Lambda\) is a reduced irreducible variety of dimension \(M - k - 1\) (the existence of such \(V\) is Proposition 1 in the text, proved in Section 5 by the standard parameter count). Now, for \(\chi,W\), etc. as above, the pullback to \(V\) of a mobile linear system on \(S\) provides a mobile linear subsystem \(\Sigma\subseteq |-nK_V|\), some \(n\geq 1\), such that \(\Sigma\) is formed by a pencil, the pair \((V,(1/n)\Sigma)\) is not terminal and there is an irreducible subvariety \(B \subset V\) of codimension two, for which \(\text{mult}_B\Sigma = n\) (see Proposition 3.5 in [\textit{A. V. Pukhlikov}, J. Reine Angew. Math. 688, 209--218 (2014; Zbl 1304.14016)]). The crucial technical result (= Proposition 2) of the paper is as follows. Let \(\mu: V \dashrightarrow \mathbb{P}^{k + 1}\) be the restriction to \(V\) of the linear projection from the tangent space \(T_bB\) to \(B\) at some general point \(b\). Then the fibers of \(\mu\) are reduced and irreducible, and \(\Sigma\) is the pull-back of a mobile linear system on \(\mathbb{P}^{k + 1}\). One can then push everything forward to \(\mathbb{P}^{k + 1}\), so that \(B\) becomes a codimension one cycle on \(\mathbb{P}^{k + 1}\) and \(\Sigma\) turns into a mobile linear system of degree \(n\) hypersurfaces, still satisfying \(\text{mult}_B\Sigma = n\). The latter is only possible for \(\Sigma\) made out of hyperplane sections and the proof is completed. Finally, the technical result above is proved in Section 4, using the following arguments. Let \(p\in V\) be a general point and \(p\in D\in\Sigma\) for some \(D\). Then it suffices to prove that \(T_p\mu^{-1}(\mu(p))\subset T_pD\). For the latter, the author observes (using the results from \textit{A. V. Pukhlikov} [Manuscr. Math. 121, No. 4, 491--526 (2006; Zbl 1135.14032)]) that the intersection of \(V\) with the cone \(C(Y,x)\) over \(Y\), for an appropriate \(k\)-dimensional subvariety \(Y \subset B\) (not passing through \(p\) and varying in a family) and a point \(x\in\mathbb{P}\) (the vertex of (\(C(Y,x)\)), is an irreducible curve \(R(Y,x) := V \cap C(Y,x)\) which is smooth at \(p\) and does not meet \(\text{Sing}\,Y\). The basic property of this curve is given by the identity \[ (R(Y,x)\cdot Y)_{C(Y,x)} = \deg R(Y,x). \] Then, since \((D\cdot R(Y,x)) = n\deg R(Y,x)\), it follows from \(\text{mult}_Y\Sigma = \text{mult}_B\Sigma = n\) and \(D \cap R(Y,x) \ni p \not\in Y\) that \(R(Y,x)\subset D\). Finally, since \(Y\) was chosen generic in a family, the author concludes that \(T_p\mu^{-1}(\mu(p))\subset T_pD\), as wanted.