On the liftability of the automorphism group of smooth hypersurfaces of the projective space (Q6165170)

As the title of this article suggests, the objects studied here are smooth projective hypersurfaces over \(\mathbb{C}\), that is, the zeros of a homogeneous polynomial \(F\) in projective space. Let \(X\) be such a hypersurface, of dimension \(n\geq 1\) and of degree \(d\geq 3\). Under the assumption that \((n,d)\neq (1,3), (2,4)\), it is well-known that the automorphism group \(\mathrm{Aut}(X)\) can be seen as a subgroup of \(\mathrm{PGL}_{n+2}(\mathbb{C})\). The question on ``liftability'' is simply whether this group can be lifted to a subgroup of \(\mathrm{GL}_{n+2}(\mathbb{C})\), isomorphic to the original. A finer version asks whether the form \(F\) stays invariant by such a lift (it could change by a scalar in general). This is what the authors call an \(F\)-lifting, inspired by work of \textit{K. Oguiso} and \textit{X. Yu} [Asian J. Math. 23, No. 2, 201--256 (2019; Zbl 1433.14035)] and \textit{L. Wei} and \textit{X. Yu} [J. Math. Soc. Japan 72, No. 4, 1327--1343 (2020; Zbl 1460.14091)]. The main result of the article (Theorem 3.5) states that, if \(n\) and \(d+2\) are relatively prime, then \(\mathrm{Aut}(X)\) is \(F\)-liftable for \textit{every} \(X\). Moreover, the converse is true in the sense that, if \(n\) and \(d+2\) have a common divisor, then one can always find a hypersurface, defined by a form \(F\), of dimension \(n\) and degree \(d\), whose automorphism group is not \(F\)-liftable. A precise example is given by the Klein hypersurface, defined by the homogeneous form \[ K=x_0^{d-1}x_1+x_1^{d-1}x_2+\cdots + x_n^{d-1}x_{n+1}+x_{n+1}^{d-1}x_0. \] This hypersurface admits automorphisms that are liftable, yet not \(K\)-liftable, and this is the crucial fact used in order to contradict the liftability of the whole automorphism group by some elementary considerations on the determinant. The positive implication of the main result involves lifting first the \(p\)-Sylow subgroups of \(\mathrm{Aut}(X)\). One then proves that the subgroup generated by these liftings is indeed a lifting of \(\mathrm{Aut}(X)\). Thus, the authors turn their attention first to the question about lifting automorphisms of order \(p^r\) for some prime \(p\). This is addressed in Theorem 2.1, where the authors prove that \(p^r\) is the order of an \(F\)-liftable automorphism for \textit{every} smooth hypersurface of dimension \(n\) and degree \(d\) if and only if one of the following holds: \begin{itemize} \item[1.] \(p\) divides \(d-1\) and \(r\leq k(n+1)\), where \(k\) is the \(p\)-adic valuation of \(d-1\); \item[2.] \(p\) divides \(d\) and \((1-d)^\ell\equiv 1\pmod{p^r}\) for some \(1\leq \ell\leq n+1\); \item[3.] \(p\) does not divide \(d(d-1)\) and \((1-d)^\ell\equiv 1\pmod{p^r}\) for some \(1\leq \ell\leq n+2\). \end{itemize} This result is used to obtain an explicit description of the orders of automorphisms of smooth cubic hypersurfaces in low dimensions (up to \(n=5\)) and smooth quartic threefolds. Finally, there is a second application of all these results in the last section of the article, giving a criterion for the order of \(\mathrm{Aut}(X)\) not to be divisible by \(p^2\) (Proposition 4.1). It is worth mentioning that a second article dealing with similar questions [\textit{Z. Zheng}, Isr. J. Math. 247, No. 1, 479--498 (2022; Zbl 1499.14068)] appeared on the arXiv almost at the same time as this one. The authors explain the relations between the two papers in the introduction. It is also worth mentioning that there is a misprint in the published version of this article (corrected in the final arXiv version, as communicated to me by an author of the paper). It concerns the statement of Lemma 1.6 and its application in Lemma 1.8.