Effective bounds of linear series on algebraic varieties and arithmetic varieties (Q1703673)

The authors use the rescaling method and numerical inequalities to prove effective upper bounds for the number of effective sections on line bundles on projective varieties (geometric case) and hermitian line bundles on arithmetic varieties (arithmetic case). The bounds represent effective versions of the Hilbert-Samuel formula and the arithmetic Hilbert-Samuel formula respectively. The idea of the proof in the geometric case is inspired by the rescaling method of the arithmetic case and uses the invariant \[ \varepsilon(L) = \inf_B \varepsilon(L,B), \] where the infimum is taken over all big base-point-free line bundles \(B\) on the projective variety \(X\); and the magnitude \(\varepsilon(L,B)=(\lambda_{L,B}+1)^{n-1}B^n\) is defined using the smallest non-negative number \(\lambda_{L,B}\) such that \(\lambda_{L,B}B-L\) is a pseudo-effective divisor. The proof uses properties of \(\varepsilon(L)\) as well as the reduction of the question to a fibration \(f : X \rightarrow C\) over a curve \(C\). The main result in the geometric case reads: let \(X\) be a geometrically integral projective variety of dimension \(n\) over a field \(k\), then for a line bundle \(L\) \[ h^0(L) \leq \frac{1}{n!} \mathrm{vol}(L) +n \varepsilon(L). \] For the arithmetic case, the authors consider a number field \(K\) and an arithmetic variety \(\mathcal{X}\) of dimension \(n \geq 2\), defined over \(\mathcal{O}_K\) together with a hermitian line bundle \(\bar{\mathcal{L}}\) on \(\mathcal{X}\). The new ingredient to use for the effective bound will be the volume derivative \[ \mathrm{dvol}(\mathcal{\bar{L}})=\frac{1}{n[K:\mathbb{Q}]}\lim_{t \rightarrow 0} \frac{1}{t} (\widehat{\mathrm{vol}}(\bar{\mathcal{L}}(t))-\widehat{\mathrm{vol}}(\bar{\mathcal{L}})). \] The differentiability of the arithmetic volume was first considered by \textit{H. Chen} [J. Lond. Math. Soc., II. Ser. 84, No. 2, 365--384 (2011; Zbl 1228.14022)]. In the nef case \(\mathrm{dvol}(\mathcal{\bar{L}})=L_K^{n-1}\), the proof of the theorem is simpler because we can use the numerical inequality of Lemma 3.9. However, the general case of a big line bundle \(\bar{L}\) (not necessarily nef) uses the arithmetic Fujita approximation of \textit{H. Chen} [Ann. Sci. Éc. Norm. Supér. (4) 43, No. 4, 555--578 (2010; Zbl 1202.14024)] and the first author [Compos. Math. 145, No. 6, 1447--1464 (2009; Zbl 1197.14023)] to be able to apply the inequality in Lemma 3.9. In general, the proof in the arithmetic case relies on the geometric case applied to generic fibre \(\mathcal{X}_K\) and the line bundle \(\mathcal{\bar{L}}_K\). The result in the arithmetic case reads: Let \(K\) be a number field and \(\mathcal{X}\) an arithmetic variety of dimension \(n \geq 2\), defined over \(\mathcal{O}_K\) together with a big line bundle \(\bar{\mathcal{L}}\) on \(\mathcal{X}\). Then: \[ \hat{h}^0(\bar{\mathcal{L}}) \leq \big(\frac{1}{n!}+\frac{(n-1)\varepsilon(\mathcal{L}_K)}{\mathrm{dvol}(\bar{L})}\big) \widehat{\mathrm{vol}}(\bar{\mathcal{L}}) + 4r \log(3r), \] where \(r=h^0(\mathcal{L}_{\mathbb{Q}})=[K:\mathbb{Q}]h^0(\mathcal{L}_{K})\). The authors also observe that for \(L\) small (\(\omega_X-L\) pseudo-effective), they can replace \(n!\) by \(2(n)!\) in the formulaes and obtain a much better approximations for the number of effective sections. The paper represents a high-dimensional generalization of the analogous work done by the authors on fibered surfaces and arithmetic surfaces [Duke Math. J. 162, No. 10, 1723--1770 (2013; Zbl 1281.14019); Adv. Math. 259, 89--115 (2014; Zbl 1297.14045)].