Mori dream spaces and birational rigidity of Fano 3-folds (Q5965072)

A Mori fiber space is a morphism \(f : X \rightarrow S\) such that \(X\) is a \(\mathbb{Q}\)-factorial projective threefold with at worst terminal singularities, \(-K_X\) is \(f\)-ample, \(\dim S\leq 2\) and the map \(f\) has relative Picard number one. If \(S=\mathrm{Spec } k\), then \(X\) is called a Fano threefold. Two Mori fiber spaces \(f : X \rightarrow S\) and \(f^{\prime} : X^{\prime} \rightarrow S^{\prime}\) are called square birational if and only if there exists birational maps \(\phi : X\dasharrow X^{\prime}\) and \(\psi : S\dasharrow S^{\prime}\) such that \(f^{\prime}\phi =\psi f\). A Mori fiber space is called birationally rigid if and only if it is square birational only to itself. Let \(X\) be a Fano threefold. Then the anticanonical ring of \(X\) \[ R=\bigoplus_{n\geq 0}H^0(X,\mathcal{O}_X(-nK_X)) \] is a finitely generated Gorenstein ring and provides an embedding of \(X\) in the weighted projective space \(\mathbb{P}=\mathrm{Proj}(R)\). It is well known that if \(X\) is a complete intersection in \(\mathbb{P}\) then the codimension of \(X\) in \(\mathbb{P}\) is at most three. It is known that there are 95 families of codimension 1, 85 families of codimension 2 and 70 families of codimension three. It is also known that any quasi smooth member of any of the 95 families of codimension 1 is birationally rigid [\textit{A. Corti} et al., Lond. Math. Soc. Lect. Note Ser. 281, 175--258 (2000; Zbl 0960.14020)], that there are 18 families of codimension 2 whose general member is birationally rigid while the general member of the remaining codimension 2 families is not birationally rigid [\textit{T. Okada}, Proc. Lond. Math. Soc. (3) 109, No. 6, 1549--1600 (2014; Zbl 1349.14056)], [\textit{V. A. Iskovskikh} and \textit{A. V. Pukhlikov}, J. Math. Sci., New York 82, No. 4, 3528--3613 (1996; Zbl 0917.14007)]. The main result of this paper is the following. Let \(X\) be a quasi-smooth codimension 2 Fano threefold anticanonically embedded in a singular weighted projective space in one of the 85 families above. If the general member of the family is birationally rigid then \(X\) is birationally rigid. This result generalizes the result by Okada [Zbl 1349.14056] by lifting the generality assumption imposed there. The authors use a variation of the method used in [Zbl 0960.14020] in order to prove the main result. The main steps of the proof are the following. Suppose that \(\phi : X\dasharrow X^{\prime}\) is a birational map of Mori fiber spaces. Then \(\phi\) is a composition of Sarkisov links [\textit{A. Corti}, J. Algebr. Geom. 4, No. 2, 223--254, appendix 248--254 (1995; Zbl 0866.14007)]. If \(\phi\) is not an isomorphism, then \(X\) has a maximal center \(\Gamma \subset X\) and the first Sarkisov link starts with the corresponding maximal extraction \(f : Y \rightarrow X\). If this is the case then the authors show that \(-K_Y\in \mathrm{Int}\overline{\mathrm{Mob}}(Y)\). This is a variation of the test class method developed in [Zbl 0960.14020]. Then the authors use this criterion, the classification of divisorial contractions by Kawamata and Kawakita and an explicit case by case study in order to show that in some of the 18 families of codimension 2 Fanos studied by Okada [Zbl 1349.14056], \(X\) has no maximal centers and therefore in these cases \(\phi \) is an isomorphism and hence \(X\) is birationally rigid. In the remaining families where the maximal centers cannot be excluded using the previous criterion, the authors use the untwisting method developed in [Zbl 0917.14007] in order to show that any Sarisov link starting from \(X\) produces a birational automorphism of \(X\) and hence again it follows that \(X\) is birationally rigid.