\(K\)-stability of cubic threefolds (Q2328044)

\(K\)-stability and its relation to the existence of Kähler-Einstein metrics is currently the most intensively studied subjects in the theory of (not too) singular Fano varieties. An important application of this theory is the construction of proper good moduli spaces for Fano manifolds that are Kähler-Einstein, the existence of such a moduli space for \(\mathbb Q\)-Gorenstein smoothable \(K\)-polystable Fano varieties was shown recently [\textit{C. Li} et al., Duke Math. J. 168, No. 8, 1387--1459 (2019; Zbl 1469.14087)]. In general it is quite difficult to verify which Fano varieties are \(K\)-semistable, so it is not easy to identify the boundary points of the moduli space. The first named author [Compos. Math. 154, No. 6, 1131--1158 (2018; Zbl 1397.14052)] showed that if \(X\) is a \(K\)-semistable Fano variety of dimension \(n\), then \[ \widehat{\mbox{vol}}(x, X) \cdot \left( \frac{n+1}{n} \right)^n \geq (-K_X)^n \] where \(\widehat{\mbox{vol}}(x, X)\) is the local volume of the klt singularity \(x \in X\) (see [\textit{C. Li}, Math. Z. 289, No. 1--2, 491--513 (2018; Zbl 1423.14025)] for the details on volume of singularities). In particular one may hope that for Fano varieties with large degree, the inequality leads to strong restrictions on the singularities and ultimately to a description of the points of the moduli space. In this paper, the authors carry out this strategy for cubic hypersurfaces in \(\mathbb P^4\). The main result of their study is that a cubic threefold is \(K\)-polystable (resp. \(K\)-semistable) if and only if it is GIT-polystable (resp. GIT-semistable). Since GIT-stability is much better understood this leads to a complete classification of \(K\)-semistable cubic threefolds and an explicit description of their moduli space. In particular this shows that all smooth cubic threefolds admit a Kähler-Einstein metric. On a technical level, the key ingredient is to compute explicitly the volume of klt threefold singularities, for example the authors shows that \(\widehat{\mbox{vol}}(x, X) \leq 16\) and equality holds if and only if it is an \(A_1\)-singularity. This result (and other interesting estimates e.g. for quotient singularities) are based on classification results for three-dimensional Gorenstein singularities, so a generalisation to higher dimension seems very difficult.