Base manifolds for fibrations of projective irreducible symplectic manifolds (Q1006315)

By results of Matsushita, any surjective holomorphic map \(f: M \rightarrow X\) with connected fibers from a smooth holomorphic symplectic \(2n\)-fold \(M\) to a projective manifold \(X\), with \(0 < \dim X < 2n\), is a Lagrangian fibration (i.e. with fibers being Lagrangian subvarieties of \(M\)), and the general fiber of \(f\) is an abelian \(n\)-fold; in particular \(\dim\;X = n\). Moreover the base \(X\) has the same Betti numbers as the projective \(n\)-space \({\mathbb{P}}^n\), i.e. with all odd Betti numbers \(b_{2k+1}(X)\) equal to \(0\) and all even Betti numbers \(b_{2k}(X)\) equal to \(1\), see [\textit{D. Matsushita}, Topology 38, No.1, 79--83 (1999); addendum ibid. 40, 431--432 (2001; Zbl 0932.32027); Math. Res. Lett. 7, No. 4, 389--391 (2000; Zbl 1002.53050)]. In this paper is proved the following important complement to the results of Matsushita: In the above setting, the base \(X\) is in fact the projective space \({\mathbb{P}}^n\) (see Theorem 1.2). In other words: Any not-trivial surjective morphism \(f: M \rightarrow X\) with connected fibers from a holomorphic symplectic \(2n\)-fold \(M\) to a smooth manifold \(X\) is a Lagrangian fibration (Matshushita), and the base manifold \(X\) of \(f\) is the projective space \(X = {\mathbb{P}}^n\) (Hwang, the paper under review). Here we reproduce the following excerpt from the author's introduction that best summarizes the way used in the proof of Theorem 1.2: ``There are two geometric ingredients in the proof of Theorem 1.2: the theory of varieties of minimal rational tangents and the theory of Lagrangian fibrations. On the one hand, the theory of varieties of minimal rational tangents describes a certain geometric structure arising from minimal rational curves at general points of a Fano manifold \(X\) with \(b_2(X)=1\). This geometric structure has differential geometric properties reflecting special features of the deformation theory of minimal rational curves. On the other hand, the theory of Lagrangian fibrations, or equivalently, the theory of completely integrable Hamiltonian systems, provides an affine structure at general points of the base manifold \(X\) via the classical action variables. Our strategy to prove Theorem 1.2 is to exploit the interplay of these two geometric structures on the base manifold \(X\). Under the assumption that \(X\) is different from \({\mathbb{P}}^n\), the condition \(b_2(X)=1\) forces the geometric structure arising from the variety of minimal rational tangents to be `non-flat', while the affine structure arising from the action variables is naturally `flat'. These two structures interact via the monodromy of the Lagrangian fibration, leading to a contradiction. To be precise, two separate arguments are needed depending on whether the dimension \(p\) of the variety of minimal rational tangents is positive or zero. The easier case of \(p>0\) is handled by a topological argument using \(b_4(X) =1\). The more difficult case of \(p=0\) needs a deeper argument, depending on the local differential geometry of the variety of minimal rational tangents.''