Discrete behavior of Seshadri constants on surfaces (Q2463860)

Given a smooth projective variety \(X\), with \(\dim X = N > 1\), a nef divisor class \(L\) on it, a real vector \({\mathbf l} = (l_1,\dots,l_n)\), \(l_i>0\), and \(p_1,\dots,p_n\in X\), we define the multipoint Seshadri constant as: \[ \epsilon(X,L,l_1p_1,\dots,l_np_n) = \inf \left\{ {L\cdot C \over \sum_{i=1} ^n \text{mult}_{p_i}C } \right\}, \] where \(C\) varies among all curves which passes at least through one of the points. The study of Seshadri constants has been the subject of work of many authors in the last years. If \(\pi > Y \rightarrow X\) is the blow up of \(X\) at distinct \(p_1,\dots,p_n\), then\(\epsilon(X,L,l_1p_1,\dots,l_np_n)\) is the largest \(t\) such that \(F_t = \pi ^*L - t(l_1E_1+\dots+l_nE_n)\) is nef (where \(E_i=\pi^*(p_i)\)). It is known that \(\epsilon(X,L,l_1p_1,\dots,l_np_n) \leq {}^N\sqrt{{L^N\over {\mathbf l}^2}}\) and if \(l_i\in \mathbb{Q}\), then for any \(0\leq t \leq {}^N\sqrt{{L^N\over {\mathbf l}^2}}\), \(t\in \mathbb{Q}\), we have \(t\leq \epsilon(X,L,l_1p_1,\dots,l_np_n) \) if and only if for general points \(p_i\) there are no reduced and irreducible curves \(C \subset X\) such that \(F_t\cdot \tilde C < 0\). The idea in the paper is to construct , given \(F_t\), a set of real numbers \(o_n(F_t)\) which contains any value \({(\pi^*L\cdot C)\over -(l_1E_1+\dots+l_nE_n)\cdot D}\) for every effective reduced irreducible \(D\) in \(Y\) with \(F_t\cdot D < 0\), if any. Then the aim is to cancel from \(o_n(F_t)\) as many values \(\nu\) as possible by showing that there is no \(D\) for which \(\nu = {(\pi^*L\cdot C)\over -(l_1E_1+\dots+l_nE_n)\cdot D}\) so that if \(c\) is the \(\inf\) of what is left in \(o_n(F_t)\) it is possible to conclude that \(F_c\) is nef and so \(c < \epsilon(X,L,l_1p_1,\dots,l_np_n)\). Given \(L\) big and nef and a vector \({\mathbf l}\), the authors determine a first set \(o_n(F_c)\) for each \(c < {}^N\sqrt{{L^N\over {\mathbf l}^2}}\) which depends only on \({\mathbf l}\), \(c\), \(L^2\) and the semigrup \(\{cL \mid C \text{ is effective}\}\). So they get the main result in the paper, which is quite remarkable: The set of possible values for \(\epsilon(X,L,l_1p_1,\dots,l_np_n)\) is either finite or an increasing discrete sequence with \({}^N\sqrt{{L^N\over {\mathbf l}^2}}\) as its only limit point (hence, apart from \({}^N\sqrt{{L^N\over {\mathbf l}^2}}\), it is a discrete set). This result is used to get various lower bounds for Seshadri constants and also results on ample divisors on low ups of \(\mathbb{P}^2\).