Arithmetic linear series with base conditions (Q1925725): Difference between revisions

The purpose of the paper under review is to give a comparison of the arithmetic volume of a big arithmetic \(\mathbb{R}\)-Cartier divisor of \(C^0\)-type on an arithmetic variety with the arithmetic volume of certain arithmetic linear series, and to apply this result to the study of the problem of Zariski decomposition on arithmetic varieties. Let \(\overline{D}\) be an arithmetic \(\mathbb{R}\)-divisor of \(C^0\)-type on an arithmetic variety \(X\), for \(\xi_1,\cdots,\xi_l\in X\) and \(\mu_1,\cdots,\mu_l\in \mathbb{R}_{\geq0}\), the arithmetic linear series of \(\overline{D}\) with base conditions \(\mu_1\xi_1,\cdots,\mu_l\xi_l\) is defined to be \[ \widehat{H}^0(X,\overline{D};\mu_1\xi_1,\cdots,\mu_l\xi_l):=\{\phi\in \widehat{H}^0(X,\overline{D})\backslash\{0\}\mid\text{mult}_{\xi_i}(D+(\phi))\geq\mu_i (\forall i)\}\cup\{0\}. \] Similar to the arithmetic volume of an arithmetic divisor, one can define \[ \widehat{\text{vol}}(\overline{D};\mu_1\xi_1,\cdots,\mu_l\xi_l):=\limsup_{n\to \infty}\frac{\log\#\widehat{H}^0(X,n\overline{D};n\mu_1\xi_1,\cdots,n\mu_l\xi_l)}{n^{d+1}/(d+1)!} \] where \(d\) is the dimension of \(X\). The main result of the paper under review states that if \(\xi_1,\cdots,\xi_l\in X_\mathbb{Q}\), \(\overline{D}\) is big and \(\mu_i>\mu_{\mathbb{R},\xi_i}(\overline{D})\) for some \(i\), then \(\widehat{\text{vol}}(\overline{D};\mu_1\xi_1,\cdots,\mu_l\xi_l)<\widehat{\text{vol}}(\overline{D})\). Here \(\mu_{\mathbb{R},\xi_i}(\overline{D})\) is the \(\mathbb{R}\)-asymptotic multiplicity of \(\overline{D}\) at \(\xi_i\). Concerning the problem of Zariski decomposition on \(X\), the author gave the following applications: (i) If \(\overline{D}\) is big and \(\overline{D}=\overline{P}+\overline{N}\) is a Zariski decomposition of \(\overline{D}\), then \(\mu_{\mathbb{R},\xi}(\overline{D})=\text{mult}_\xi(N)\) for all \(\xi\in X_\mathbb{Q}\). (ii) Let \(d\geq2\) and \(X=\mathbb{P}_\mathbb{Z}^d=\text{Proj}(\mathbb{Z}[T_0,T_1,\cdots,T_d])\). Suppose that \(\overline{D}\) is of the form \[ \big(H_0,\log(a_0+a_1\mid z_1\mid^2+\cdots+a_d\mid z_d\mid^2)\big) \] where \(H_0:=\{T_0=0\}, z_i=T_i/{T_0} (i=1,\cdots d)\) and \(a_0,a_1,\cdots,a_d\in \mathbb{R}_{>0}\). If \(\overline{D}\) is big and not nef, then for any birational morphism \(f: Y\to \mathbb{P}_\mathbb{Z}^d\) of generically smooth, normal and projective arithmetic varieties, there is no Zariski decomposition of \(f^*(\overline{D})\) on \(Y\). (iii) Let \(d=1\) and assume that \(X\) is regular. If \(\overline{D}\) is big, then there is a unique Zariski decomposition of \(\overline{D}\).