Polar transform and local positivity for curves (Q2202780)

In this very interesting paper, the authors study some positivity questions for \((n-1,n-1)\)-classes via newly defined variants on Nakayama and Seshadri constants. Let \(X\) be a smooth projective variety of dimension \(n\) and let \(L\) be a nef line bundle on \(X\). The first important invariant which measures the local positivity of \(L\) at a point \(x \in X\) is the local Seshadri constant \[\varepsilon(X,L;x) =\mathrm{inf}{C \subset X} \frac{L.C}{\mathrm{mult}_{x}(C)},\] where the infimum is taken over all irreducible and reduced curves \(C\subset X\) passing through \(x \in X\). One can also define the local Seshadri constant as \[(\star): \quad \varepsilon(X,L;x) = \mathrm{sup} \{ t \geq 0 : \pi^{*}L - tE \text{ is nef} \},\] where \(\pi : Y \rightarrow X\) is the blow up of \(X\) at \(x \in X\) and \(E\) is the exceptional divisor. One can also define the global Seshadri constant as \[\varepsilon(L) = \mathrm{inf}_{x \in X} \varepsilon(X,L;x).\] Denoting the cone of nef \((1,1)\) classes of \(X\) by \(\mathrm{Nef}^{1}(X)\), it is not hard to see that the Seshadri constant can be defined for any class in \(\mathrm{Nef}^{1}(X)\) using \((\star)\). Due to this fact we can consider \(\varepsilon(X,\cdot \,;x)\) as a function on the cone \(\mathrm{Nef}^{1}(X)\) having values in \(\mathbb{R}\), leading to study of the pair \((\mathrm{Nef}^{1}(X), \varepsilon(X, \cdot \,x))\). Another invariant is defined defined for any pseudo-effective \((1,1)\) class \(L\), namely \[n_{x}(L) :=\mathrm{sup} \{t \geq 0 : \, \pi^{*}L - tE \text{ is pseudo-effective}\},\] where \(\pi : Y \rightarrow X\) is the blow-up of \(X\) at \(x\) and \(E\) is the exceptional divisor. We can then define the global Nakayama constant by \[n(L) =\mathrm{inf}_{x \in X} n_{x}(L).\] In the same spirit as in the case of local Seshadri constants, denoting the cone of pseudo-effective \((1,1)\) classes of \(X\) as \(\overline{\mathrm{Eff}}^{1}(X)\), we can consider \(n_{x}\) as a function of \(\overline{\mathrm{Eff}}^{1}(X)\) having values in \(\mathbb{R}\) and then we can study the pair \((\overline{\mathrm{Eff}}^{1}(X), n_{x})\). In the first step, via polar transforms, the authors introduce Nakayama and Seshadri constants for \((n-1,n-1)\) classes. Consider \[\mathrm{Nef}^{1}(X) = \overline{\mathrm{Eff}}_{1}(X)\] the cone of pseudo-effective \((n-1,n-1)\)-classes. Then for \(\alpha \in \overline{\mathrm{Eff}}_{1}(X)\) its local Nakayama constant at \(x\) defined as the polar of \(\varepsilon(x, \cdot \, ;x)\) as \[N_{x}(\alpha) =\mathrm{inf}_{L \in\mathrm{Nef}^{1}(X)^{\circ}} \frac{\alpha L}{\varepsilon(X,L;x)}.\] Next, consider \[\overline{\mathrm{Eff}}^{1}(X)^{*} = \mathrm{Mov}_{1}(X),\] where \(\mathrm{Mov}_{1}(X)\) is the cone of movable \((n-1,n-1)\)-classes. Then for any \(\alpha \in\mathrm{Mov}_{1}(X)\) its local Seshadri constant at \(x\), denoted by \(S_{x}(\alpha)\), defined be the polar of \(n_{x}\) as \[S_{x}(\alpha) =\mathrm{inf}_{L \in \overline{\mathrm{Eff}}^{1}(X)^{\circ}} \frac{\alpha L}{n_{x}(L)}.\] We can then define the global variants of those constants, namely the global Nakayama constant is defined as \[N(\alpha) = \mathrm{inf}_{x \in X} N_{x}(\alpha),\] and the global Seshadri constant \[S(\alpha) =\mathrm{inf}_{x \in X} S_{x}(\alpha).\] In order to formulate the main results of the paper under review, we need to recall additionally definitions of \(\mathfrak{m}(\cdot )\) and \(\widehat{\mathrm{vol}}(\cdot)\). Let \(X\) be a projective variety of dimension \(n\) and let \(\alpha \in \overline{\mathrm{Eff}}_{1}(X)\). Then the volume of \(\alpha\) is defined as \[\widehat{\mathrm{vol}}(\alpha) =\mathrm{inf}_{A} \bigg( \frac{A \cdot\alpha}{\mathrm{vol}(A)^{1/n}}\bigg)^{n/(n-1)},\] where the infimum is taken over big and nef divisor classes. In the same setting, for any \(\alpha \in\mathrm{Mov}_{1}(X)\) we define \[\mathfrak{m}(\alpha) =\mathrm{inf}_{L} \bigg( \frac{L \cdot\alpha}{\mathrm{vol}(L)^{1/n}}\bigg)^{n/(n-1)},\] where \(L\) is a big divisor class. Theorem A. For any \(\alpha \in \overline{\mathrm{Eff}}_{1}(X)\) its local Nakayama constant satisfies: i) \(N_{x}(\alpha) \geq \widehat{\mathrm{vol}}(\alpha)^{(n-1)/n}\), ii) \(N_{x}(\alpha)\) has the following geometric characterization: \[N_{x}(\alpha) =\mathrm{sup} \{ t \geq 0 : \, \pi^{*} \alpha + te \text{ is pseudo-effective}\},\] where \(\pi: Y \rightarrow X\) is the blow-up of \(X\) at \(x\) and \(e = (-E)^{n-1}\). Theorem B. The local Seshadri constant has the following properties: i) for any \(\alpha \in \mathrm{Mov}_{1}(X)\), \(S_{x}(\alpha) \leq \mathfrak{m}(\alpha)^{(n-1)/n}\), ii) \(S_{x}\) has the following geometric characterization: \[S_{x}(\alpha) =\mathrm{sup} \{t \geq 0 : \, \pi^{*}\alpha + te \text{ is movable}\},\] where \(\pi : Y \rightarrow X\) is the blow-up of \(X\) at \(x\) and \(e=(-E)^{n-1}\). iii) Suppose that \(\alpha \in\mathrm{Mov}_{1}(X)\), then there is some \(\delta > 0\) such that \(S_{x}(\alpha) > \delta\) holds for every point iff \(\alpha \in \mathrm{Mov}_{1}(X)^{\circ}\). Corollary. Let \(\alpha \in\mathrm{Mov}_{1}(X)\), then \(S_{x}(\alpha) \leq N_{x}(\alpha)\) for any \(x \in X\). Theorem C. Let \(\alpha \in \mathrm{Mov}_{1}(X)\) be a class on the boundary of \(\mathrm{Mov}_{1}(X)\) with \(\mathfrak{m}(\alpha)>0\). Then we have \[x \in \{x \in X: \, S_{x}(\alpha) = 0\} \iff x \in \text{ the divisorial components of the non-Kaehler locus } E_{nK}(L_{\alpha}),\] where \(L_{\alpha}\) is a unique big movable \((1,1)\) class such that \(\alpha = \langle L_{\alpha}^{n-1}\rangle\).