Restrictions on Seshadri constants on surfaces (Q515104)

In the paper under review, the authors study possible restrictions on Seshadri constants on algebraic surfaces. Let us recall basic definitions. Let \(X\) be a smooth algebraic variety and let \(L\) be a nef line bundle on \(X\). For any point \(x \in X\) the real number \[ \varepsilon(L;x) = \inf_{x \in C} \frac{ L.C } {\text{mult}_{x}C}, \] where the infimum is taken over all irreducible curves passing through \(x\), is called the Seshadri constant of \(L\) at \(x\). Let us recall that the Seshadri constants were introduced by J. P. Demailly in the context of Fujita's conjecture and they have become a subject of considerable interest ever since. One of the most classical results in this subject, proved by \textit{L. Ein} and \textit{R. Lazarsfeld} [in: Journées de géométrie algébrique d'Orsay, France, juillet 20-26, 1992. Paris: Société Mathématique de France. 177--186 (1993; Zbl 0812.14027)], provides us a lower (sharp) bound on the Seshadri constant on algebraic surfaces taken at a very general point \(1\) -- usually one does not specify this point. Theorem. Let \(X\) be an algebraic surface and let \(L\) be an ample line bundle on \(X\). Then \[ \varepsilon(L;1) \geq 1. \] It is worth pointing out that this result, in general, cannot be improved, even under the assumption that the self-intersection \(d = L^{2}\) of \(L\) is very large. To observe this phenomenon, consider \(X = \mathbb{P}^{1} \times \mathbb{P}^{1}\) and \(L = sH+V\), where \(H\) is the class of the fiber of the projection from \(X\) onto the second factor and \(V\) is the class of the fiber of the projection onto the first factor. Then an easy computation reveals that \(L^{2} = 2s\), but \(\varepsilon(L;x) = 1\) for all points \(x \in X\). The main result of the paper shows that the set of potential values \(\varepsilon(L;1)\) can take on is surprisingly limited. In order to formulate the result, let us recall that for a non-square integer \(d\) and \((p,q)\) a solution to Pell's equation \[ y^{2} - dx^{2}=1 \] one defines the following set \[ \text{Exc}(d;p,q) = \{1,2, \dots, \lfloor\sqrt{d}\rfloor\}\cup\bigg\{ \frac{a}{b} \text{ such that } 1\leq \frac{a}{b} \leq \frac{p}{q}d \text{ and } 2\leq b \leq q^{2} \bigg\}. \] Main Result. Let \(X\) be a smooth projective surface, \(x \in X\), \(L\) an ample line bundle on \(X\) such that \(L^{2} =d\) is not a square. Let \((p,q)\) be an arbitrary solution to the above Pell's equation. Then either \[ \varepsilon(L;1) \geq \frac{p}{q}d, \] or \(\varepsilon(L;1) \in \text{Exc}(d;p,q)\). Let us also emphasize that the key point of the result is that possible values of \(\varepsilon(L;1)\) depend in a uniform way only on \(L^{2}\). The proof of Main Result is based on a rather new approach using infinitesimal Newton-Okounkov bodies -- this idea comes from the paper by \textit{A. Küronya} and \textit{V. Lozovanu} [``Infinitesimal Newton-Okounkov bodies and jet separation'', Duke Math. J. (to appear), \url{doi:10.1215/00127094-0000002X}]. At the end of the paper, the authors consider the following conjecture. Conjecture. Let \(X\) be a smooth projective surface with Picard number \(1\) and let \(L\) be the ample generator of \(NS(X)\) with \(L^{2} =d\). Assume furthermore that \(d\) is not a square. Then \[ \varepsilon(L;1) \geq \frac{p_{0}}{q_{0}}d, \] where \((p_{0},q_{0})\) is the primitive solution of the mentioned Pell's equation. It turns out that the authors are able to verify the above conjecture in two particular cases, namely when \(L^{2} = n^{2}-1\) or \(n^{2}-n\) for an arbitrary positive integer \(n\).