Effective bound of linear series on arithmetic surfaces (Q2393590)

The results of the article under review lie in the arithmetic intersection theory of Arakelov, Faltings, and Gillet and Soulé. It contains three main statements, the first two concern effective upper bounds on the number of effective sections of an arbitrary Hermitian line bundle and for a nef arithmetic divisor over an arithmetic surface, they can be viewed as effective versions of the arithmetic Hilbert-Samuel formula. The third statement improves the upper bound substantially for nef line bundles of small degree on the generic fiber. As a consequence, the authors obtained effective lower bounds on the Faltings height and on the sef-intersection of the canonical bundle in terms of the number of singular points on the fibers of the arithmetic surface. To be more precise, let \(\overline{\mathcal{L}}=(\mathcal{L},\parallel\cdot\parallel)\) be a Hermitian line bundle over an arithmetic surface \(X\) and let \[ \widehat{H}^0(\overline{\mathcal{L}})=\{s\in H^0(X,\mathcal{L}) : \parallel s\parallel_{\text{sup}}\leq 1\} \] denote the set of effective sections. Define \(\widehat{h}^0(\overline{\mathcal{L}})=\log \# \widehat{H}^0(\overline{\mathcal{L}})\) and \(\widehat{\text{vol}}(\overline{\mathcal{L}})=\limsup_{n\to \infty}\frac{2}{n^2}\widehat{h}^0(\overline{\mathcal{L}})\), Huayi Chen has showed that the ``\(\limsup\)'' is actually a limit and hence \[ \widehat{h}^0(n\overline{\mathcal{L}})=\frac{1}{2}\widehat{\text{vol}}(\overline{\mathcal{L}})n^2+\o(n^2)\quad\text{as}\quad n\to \infty. \] The authors of the article under review obtained an effective version of the above expansion in one direction: Theorem 1. Let \(X\) be a regular and geometrically connected arithmetic surface of genus \(g\) over \(\mathcal{O}_K\), where \(\mathcal{O}_K\) stands for the ring of integers in a number field \(K\). Let \(\overline{\mathcal{L}}\) be a Hermitian line bundle on \(X\). Denote \(d^0=\text{deg}(\mathcal{L}_K)\), and denote by \(r'\) the \(\mathcal{O}_K\)-rank of the \(\mathcal{O}_K\)-submodule of \(H^0(\mathcal{L})\) generated by \(\widehat{H}^0(\overline{\mathcal{L}})\). Assume that \(r'\geq2\). (i). If \(g>0\), then \(\widehat{h}^0(\overline{\mathcal{L}})\leq\frac{1}{2}\widehat{\text{vol}}(\overline{\mathcal{L}})+4d\log(3d)\). Here \(d=d^0[K:\mathbb{Q}]\); (ii). If \(g=0\), then \(\widehat{h}^0(\overline{\mathcal{L}})\leq(\frac{1}{2}+\frac{1}{2(r'-1)})\widehat{\text{vol}}(\overline{\mathcal{L}})+4r\log(3r)\). Here \(r=(d^0+1)[K:\mathbb{Q}]\). For nef Hermitian line bundle \(\overline{\mathcal{L}}\), \(d^0=\text{deg}(\mathcal{L}_K)\geq0\) and the self-intersection number \(\overline{\mathcal{L}}^2\geq0\), the authors also obtained similar results to Theorem 1 for such \(\overline{\mathcal{L}}\). Theorem 2. Let \(\overline{\mathcal{L}}\) be a nef Hermitian line bundle on \(X\) with \(d^0=\text{deg}(\mathcal{L}_K)>0\). (i). If \(g>0\) and \(d^0>1\), then \(\widehat{h}^0(\overline{\mathcal{L}})\leq\frac{1}{2}\overline{\mathcal{L}}^2+4d\log (3d)\). Here \(d=d^0[K:\mathbb{Q}]\); (ii). If \(g=0\) and \(d^0>0\), then \(\widehat{h}^0(\overline{\mathcal{L}})\leq(\frac{1}{2}+\frac{1}{2d^0})\overline{\mathcal{L}}^2+4r\log(3r)\). Here \(r=(d^0+1)[K:\mathbb{Q}]\). The authors also gave a proof of the fact that Theorem 2 implies Theorem 1, which is inspired by the arithmetic Zariski decomposition of Moriwaki. Theorem 2 may be too weak if \(\text{deg}(\mathcal{L}_K)\) is very small. The authors presented in this article a substantial improvement of Theorem 2 for the line bundles of small degree on the generic fiber, that's the following: Theorem 3. Let \(X\) be a regular and geometrically connected arithmetic surface of genus \(g>1\) over \(\mathcal{O}_K\). Let \(\overline{\mathcal{L}}\) be a nef Hermitian line bundle on \(X\). Assume that \(2\leq d^0\leq 2g-2\). Then \[ \widehat{h}^0(\overline{\mathcal{L}})\leq(\frac{1}{4}+\frac{2+\varepsilon}{4d^0})\overline{\mathcal{L}}^2+4d\log(3d). \] Here \(d=d^0[K:\mathbb{Q}]\). The number \(\varepsilon=1\) if \(X_K\) is hyperelliptic and \(d^0\) is odd; otherwise, \(\varepsilon=0\). Applying Theorem 3 to the case where \(\overline{\mathcal{L}}\) is the Arakelov canonical bundle \(\overline{\omega}_X=(\omega_X,\parallel\cdot\parallel_{\text{Ar}})\), it reads Theorem 4. \(\widehat{h}^0(\overline{\overline{\omega}}_X)\leq\frac{g}{4(g-1)}\overline{\omega}_X^2+4d\log(3d)\). Here \(d=(2g-2)[K:\mathbb{Q}]\). Theorem 4 can be combined with Faltings' arithmetic Noether formula to deduce effective lower bounds on the Faltings height and on the self-intersection of the Arakelov canonical bundle in terms of the number of singular points on fibers of \(X\). Finally, the authors indicated that the three main results in this article can be viewed as arithmetic version of Noether-type inequalities.