An inequality between multipoint Seshadri constants (Q1027767)

Let \(L\) be a nef divisor on a projective variety \(X\) of dimension \(n\). The Seshadri constant of dimension \(d\leq n\) at \(r\) points \(p_1, \dots, p_r\) in the smooth locus of \(X\) is defined to be the real number \[ \epsilon_d(p_1, \dots, p_r;X,L)=\inf \left\{ \left( \frac{L^d\cdot Z}{\sum \text{mult}_{p_i}Z}\right)^{1/d}: Z\subset X \,\,\text{ effective \,\, cycle \,\, of \,\, dimension}\, d \right\}. \] By semicontinuity of multiplicities if the points are in very general position then the Seshadri constant does not depend on the actual points chosen. Thus one can define \(\epsilon_d(r;X,L)=\epsilon_d(p_1, \dots, p_n;X,L)\) where \(p_1, \dots, p_r\) is any collection of \(r\) very general points in \(X\). The authors prove the following inequality comparing Seshadri constants of very general points in \(X\) and those in projective space. Theorem. Let \(X\) be an \(n-\)dimensional projective variety, \(L\) be a nef divisor on \(X\) and \(r,s,d,\geq 1\) be integers, with \(d \leq n\). Then \[ \epsilon_d(rs;X,L)\geq \epsilon_d(r;X,L) \cdot \epsilon_d(s; \mathbb{P}^n, \mathcal{O}_{\mathbb{P}^n}(1)). \] It is well known and not hard to see that \(\epsilon_d(r;X,L)\leq (L^n/r)^{\frac{1}{n}}\) for \(d=1\) and \(d=n-1\). On the other hand, explicit values or even lower bounds are in general hard to compute. Thus the previous result should be seen as a lover bound on \(\epsilon_d(rs;X,L)\), which will be useful if Seshadri constants on projective spaces are known. The theorem implies that if the Nagata-Biran-Szemberg conjecture is true for \(d-\)dimensional Seshadri constants on \(\mathbb{P}^n\) and \(\epsilon_d(1;X,L)=(L^n)^{\frac{1}{n}}\) , then the Nagata-Biran-Szemberg conjecture is true for the \(d-\)dimensional Seshadri constants of \((X,L)\). It is important to notice that, when \(d=1\), the Seshadri constant \(\epsilon_1\) reduces to the usual Seshadri constant as introduced by Demailly and in this case, the previous Theorem is due to Biran. Also the case \(r=1\), \(d=n-1\) was already known and due to Roé. \vskip0.2cm Since, in the proof of the Theorem, it is not necessary to divide the collections of \(rs\) points evenly, the authors obtain another result. Theorem. Let \(X\) be an \(n-\)dimensional projective variety, \(L\) be a nef divisor on \(X\). Suppose \(1 \leq s_1 \leq s_2 \leq \cdots \leq s_r\) are integers and set \(s=\sum_{i=1}^rs_i\). Then \[ \epsilon_d(s;X,L)\geq \epsilon_d(r;X,L) \cdot \epsilon_d(s_r; \mathbb{P}^n, \mathcal{O}_{\mathbb{P}^n}(1)). \] The authors, moreover, consider the case of weighted Seshadri constants. In addition to the nef divisor \(L\) on \(X\) and \(r\) points \(p_1, \dots, p_r\) in the smooth locus of \(X\), let a nonzero real vector \(\ell=(l_{1}, \dots, l_{r})\in \mathbb{R}^r_{+}\) be given with each \(l_i\geq 0\). The Seshadri constant of dimension \(d\leq n\) at these points with weights \({\ell}\) is defined to be the real number \[ \epsilon_d(l_1p_1, \dots, l_rp_r;X,L)=\inf \left\{ \left( \frac{L^d\cdot Z}{\sum l_i\text{mult}_{p_i}Z}\right)^{1/d}: Z\subset X \text{ effective \, cycle \, of \, dimension}\, d \right\}. \] And the following result is proved. Theorem. Let \(X\) be an \(n-\)dimensional projective variety, \(L\) be a nef divisor on \(X\) and \(r,d,\geq 1\) be integers, with \(d \leq n\). Suppose \(1 \leq s_1 \leq s_2 \leq \cdots \leq s_r\) are integers and set \(s=\sum_{i=1}^rs_i\). For each \(i\) let \(\ell_i\) be a vector in \(\mathbb{R}^{s_i}_+\) and let \(\ell=(\ell_1, \dots, \ell_r) \in \mathbb{R}^{s}_+\). Then \[ \epsilon_d(s,\ell;X,L)\geq \epsilon_d(r;X,L) \cdot \inf_{i=1,\dots, r} \epsilon_d(s_i, \ell_i; \mathbb{P}^n, \mathcal{O}_{\mathbb{P}^n}(1)). \]