Successive minima of line bundles (Q2308301)

In this very interesting paper, the authors study a linkage between the Seshadri contants for ample line bundles on an algebraic variety \(X\) and the geometry of convex bodies. Let us recall that the Seshadri constant of an ample line bundle at a point \(x \in X\) is defined as \[ \varepsilon(X,L;x) =\mathrm{inf}_{C \subset X} \frac{L.C}{\mathrm{mult}_{x}(C)}, \] where the infimum is taken over all irreducible curves passing through the point \(x\). The Seshadri constant obtains its maximum at a very general point \(x\), and in this situation we denote this value as \(\varepsilon(X,L)\) and call the maximal Seshadri constant. The idea standing behind this paper comes from toric geometry and the following result. Theorem A. Let \(X = T_{N}(\mathrm{emb}(\triangle))\) be a proper toric variety of dimension \(d\) and let \(L\) be a torus invariant Cartier divisor on \(X\) with the Seshadri constant equal to \(\varepsilon\) at a general point of \(X\). Let \(\square_{L} \subset M_{\mathbb{R}}\) be the moment polytope and let us denote by \(\omega\) the lattice width of \(\square_{L}\). Then \[ \frac{\omega}{d} \leq \varepsilon \leq \omega. \] The strategy presented above can be generalized to a wider setting and it is done by introducing a sequence of successive minima for a line bundle on an algebraic variety having the property that the last minimum coincides with the classical Seshadri constant. Let \(X\) be a proper complex variety, \(L\) be a Cartier divisor on \(X\), and let \(x \in X\) be a point. For a real number \(t\geq 0\) one defines \(\mathrm{Bs}|\mathcal{I}_{x}^{t+}L|_{\mathbb{Q}}\) to be the common zero of sections \(s \in H^{0}(X,nL)\) with \(n\geq 1\) such that \(s\) vanishes at \(x\) of order strictly larger than \(nt\). For integers \(i\geq 1\) the \(i\)-th successive minimum of \(L\) at \(x\) is defined as \[ \varepsilon_{i}(L,x) =\mathrm{inf} \{t \geq 0 : \mathrm{codim} \, \mathrm{Bs} \, | \mathcal{I}_{x}^{t+}L|_{\mathbb{Q}} < i\}. \] We obtain the sequence \[ \varepsilon_{1}(L,x) \geq \varepsilon_{2}(L,x) \geq ... \geq \varepsilon_{d}(L,x) \geq 0 = \varepsilon_{d+1}(L,x), \] where \(d =\dim X\). It is worth emphasizing that the invariant \(\varepsilon_{i}(L,x)\) does not depend on a very general point \(x\), and then is denoted by \(\varepsilon_{i}(L)\). Moreover, one can show that \(\varepsilon_{i}(L)\)'s are numerical invariants in the case when \(L\) is big, and if \(L\) is semiample and \(x \in X\) is a smooth point, then \(\varepsilon_{d}(L)\) coincides with the maximal Seshadri constant of \(L\). The invariant \(\varepsilon_{1}(L,x)\) is the largest asymptotic multiplicity that can be imposed at \(x\) on sections of multiplies of \(L\). If \(x \in X\) is a smooth point, it coincides with the width of \(L\) at the geometric valuation induced by the exceptional divisor on the blow-up at \(x\). Due to this property, one calls \(\varepsilon_{1}(L,x)\) the width of \(L\) at \(x\). Theorem B. Let \(X\) be a proper complex variety of dimension \(d\), let \(L\) be a Cartier divisor on \(X\). Let \(\mathrm{vol}(L)\) be the volume of \(L\) and \(\varepsilon_{i}(L)\) the successive minima of \(L\) at very general points. Then \[ \prod_{i=1}^{d} \varepsilon_{i}(L) \leq \mathrm{vol}(L) \leq d! \cdot \prod_{i=1}^{d}\varepsilon_{i}(L). \] In particular, for a big divisor \(L\) on \(X\) we have inequalities \[ 1 \leq \frac{\mathrm{vol}(L)}{\varepsilon_{d}(L)^{d}} \leq d! \cdot \bigg( \frac{\varepsilon_{1}(L)}{\varepsilon_{d}(L)} \bigg)^{d}. \] The last main result of the paper tells us that for line bundles on toric varieties successive minima are equivalent to the reciprocal of usual ones. Theorem C. Let \(X = T_{N}(\mathrm{emb}(\triangle))\) be a proper toric variety of dimension \(d\). Let \(L\) be a torus invariant Cartier divisor on \(X\) with successive minima \(\varepsilon_{i}\) at a very generall point. Let \(\square_{L} \subset M_{\mathbb{R}}\) be the moment polytope, let \(\lambda_{1}, ..., \lambda_{d}\) be the successive minima of the \(0\)-symmetric convex body \(\square_{L} - \square_{L} \subset M_{\mathbb{R}}\), and \(\lambda_{1}^{*}, ..., \lambda_{d}^{*}\) be the successive minima of the polar body \((\square_{L} - \square_{L})^{*} \subset N_{\mathbb{R}}\). Then \(\varepsilon_{i}, \lambda_{i}^{-1}, \lambda_{d-i+1}^{*}\) are equivalent. More precisely, \[ 1 \leq \varepsilon_{i} \cdot \lambda_{i} \leq d \cdot \frac{\varepsilon_{i}}{\lambda_{d-i+1}^{*}} \leq d(d-i+1). \]