Birationally rigid hypersurfaces (Q1955774)

The author shows that for \(N \geq 4\) every smooth hypersurface of degree \(N\) in \(\mathbb{P}^N_{\mathbb{C}}\) is birationally superrigid. Here, \(X\) is birationally superrigid if for every birational map \(\phi : X \dasharrow Y\) to a Mori fiber space is an isomorphism preserving the fibration. In particular, if \(X\) is a birationally superrigid Mori fiber space, then it is the unique Mori fiber space in its birational equivalence class. Recall that Mori fiber spaces are the (partially conjectured) main steps of running the Minimal Model Program for varieties with negative Kodaira dimension. Unfortunately, their appearance is not unique as one runs the MMP. An easy example is the blowup of \(\mathbb{P}^2\) at two points. One run of MMP yields \(\mathbb{P}^2\) as a Mori fiber space over the ground field. On the other hand, another run of MMP yieds \(\mathbb{P}^1 \times \mathbb{P}^1\) as a Mori fiber space over one of the factors. By the theorem of the article such behavior does not happen for varieties birational to smooth hypersurfaces of degree \(N\) in \(\mathbb{P}^N\). Note also that, if \(X\) is birationally superrigid, then \(\mathrm{Aut}(X)= \mathrm{Bir(X)}\). This condition is refered to as birational rigidity sometimes. The question of birational superrigidity of hypersurfaces has been a fertile area for a long time. Its roots can be traced back to the work of \textit{G. Fano} [Torino Atti 43, 973--984 (1908; JFM 39.0718.03), Torino Atti 50, 1067--1072 (1915; JFM 45.1357.03)]. The \(N=4\) case was shown in [\textit{V. A. Iskovskikh} and \textit{Yu. I. Manin}, Math. USSR, Sb. 15, 141--166 (1972; Zbl 0249.14001)]. The question answered by the author was conjectured in [\textit{A. V. Pukhlikov}, Invent. Math. 134, No. 2, 401--426 (1998; Zbl 0964.14011)], where the statement of the article was shown for general hypersurfaces. So, the main contribution of the author is removing the genericity assumption from Pukhlikov's theorem. The proof of the main theorem follows the method of maximal singularities. Assume that there is a birational rational map \(\phi : X \to Y\) where \(X\) is a smooth hypesurface of degree \(N\) in \(\mathbb{P}^N\) and \(Y\) is a Mori fiber space. The main idea is to examine the singularities of the pair \((X, B)\), where \(B\) is the base locus of the pull back \(\mathcal{H}\) of a very ample linear system on \(Y\) via \(\phi\). Essentially due to [\textit{V. A. Iskovskikh} and \textit{Yu. I. Manin}, Math. USSR, Sb. 15, 141--166, (1972; Zbl 0249.14001)] we have that the canonical threshold of \((X, B)\) is less than \(1/r\), where \(r\) is the degree of \(\mathcal{H}\). The main contribution of the article is bounding this threshold from below by \(1/r\). The main idea is to apply generic projections, homogeneous degerations and cutting by linear subspaces to \((X,B)\) to get a pair \((\mathbb{P}^2, \overline{\Delta})\), such that \(\deg \overline{\Delta} = c^2 N r^2 / 2\) (here \(c\) is the above canonical threshold) and \(\mathcal{O}_{\mathbb{P}^2}/ \mathcal{J}(\mathbb{P}^2, \overline{\Delta})\) contains a submodule of length at least \(n-4\) supported at one point. In paricular, then Nadel vanishing yields that \(N-4 \leq \lceil \deg (K_{\mathbb{P}^2} + \overline{\Delta}) + \varepsilon \rceil \leq c^2 N r^2 / 2 -2\), which yields the required inequality on the canonical threshold. The subtle part, is finding a sequence of generic projections, degenerations and cuts by linear subspaces where singularities can be controlled adequatly. This is done using the method of arc spaces.