On the number of vertices of Newton-Okounkov polygons (Q2104836)

In the paper under review the authors study a very natural question regarding the number of vertices of Newton-Okounkov polygons associated with big divisors on smooth algebraic surfaces. Let \(S\) be a normal projective variety of dimension \(d\) and let \(D\) be a big divisor class on \(S\). We define the admissible flag for \(S\) which is \[ Y_{\bullet} : S= Y_{0} \supset \dots \supset Y_{d} = \{pt\}, \] where \(Y_{i}\) is an irreducible subvariety of codimension \(i\) and \(Y_{d}\) is smooth for each subvariety \(Y_{i}\). Let \(g_{i}\) be a local equation for \(Y_{i}\) in \(Y_{i-1}\) around the point \(Y_{d}\). Then the flag \(Y_{\bullet}\) determines a rank \(d\) valuation \(\nu_{Y_{\bullet}}\) on the field of rational functions of \(S\), namely \(\nu_{Y_{\bullet}}(f) = (v_{1}(f), \dots, v_{d}(f))\), where \(v_{i}\) are defined recursively setting \(f_{1}=f\) and \[ v_{i}(f) = \mathrm{ord}(f_{i}), \quad i=1, \dots, d, \] \[ f_{i+1} = (f_{i}/g_{i}^{v_{i}(f)})_{|Y_{i}}, \quad i = 1, \dots, d-1. \] Based on that the Newton-Okounkov body of \(D\) with respect to \(Y_{\bullet}\) is the convex body \[ \triangle_{Y_{\bullet}}(D) = \overline{\bigg\{\frac{\nu_{Y_{\bullet}}(s)}{k} : s \in H^{0}(S, \mathcal{O}_{S}(kD)) \bigg\}}. \] Whereas the volume of the Newton-Okounkov body is well-known, still its shape, and its dependence on \(Y_{\bullet}\), is an intriguing subject of study. The main aim of the paper is to understand this question when \(d=2\), where the prominent role is played by the Zariski decomposition of \(D\). In order to formulate the main result of the paper under review, recall the following notations. Let \(C_{1}, \dots, C_{k} \subset S\) be a configuration of negative curves. Assume that \(N\) is effective and the associated intersection matrix is negative definite. We denote by \(\mathrm{mc}(N)\) the largest number of irreducible components of a connected divisor contained in \(N\), and by \(\mathrm{mv}(N)\) we mean \(k+\mathrm{mc}(N)+4\) if \(k < \rho(S)-1\) and \(k+\mathrm{mc}(N)+3\) if \(k=\rho(S)-1\), where \(\rho(S)\) denotes the Picard number of \(S\). Given a smooth projective surface \(S\), let \[ \mathrm{mv}(S) =\max \{\mathrm{mv}(N):N = C_{1} +\dots+C_{k} \text{ is negative definite}\}. \] Main Result. On every smooth projective surface \(S\) and for every big divisor \(D\) \[ \max_{Y_{\bullet}} \{ \# \text{ vertices of } \triangle_{Y_{\bullet}}(D)\} \leq \mathrm{mv}(S), \] where the maximum is taken over all admissible flags \(Y\). If \(D\) is ample, then for every \(3 \leq r \leq\mathrm{mv}(S)\) there exists a flag \(Y_{\bullet}\) such that \(\triangle_{Y_{\bullet}}(D)\) has exactly \(r\) vertices. Corollary. Given a positive integer \(\rho\) there exists a projective smooth surface \(S\) with Picard number \(\rho = \rho(S)\), an admissible flag \(Y_{\bullet}\), and a divisor \(D\) such that \(\triangle_{Y_{\bullet}}(D)\) has exactly \(2\rho + 1\) vertices.