Positivity functions for curves on algebraic varieties (Q2324667)

In this stunning paper, the authors develop the positivity theory of curves in algebraic varieties. In order to present the main result, let us recall the basics. Let \(X\) be a smooth complex projective variety of dimension \(n\) and let \(L\) be a line bundle on \(X\). We define the volume of \(L\) by \[ \mathrm{vol}(L) = \lim_{m \rightarrow \infty} \frac{\dim H^{0}(X,mL)}{m^{n}/n!}. \] When \(X\) is a surface, then the volume can be calculated with use of the Zariski decomposition which is a very useful tool in the theory. Recently, there has been interest in extending the theory of positivity to subvarieties of arbitrary codimension -- a special role in this picture is played by curve classes. In this case, one can define a volume-type functions for curve classes. Definition. Let \(X\) be a projective variety of dimension \(n\) and let \(\alpha \in \overline{\mathrm{Eff}}_{1}(X)\) be a pseudo-effective curve class. Then the volume of \(\alpha\) is defined by \[ \widehat{\alpha} = \inf_{A \text{ big and nef divisor class}} \bigg( \frac{A\cdot \alpha}{\mathrm{vol}(A)^{1/n}} \bigg)^{n/n-1}. \] Another function captures the asymptotic geometry of curves, the so-called mobility function. Definition. Let \(X\) be a projective variety of dimension \(n\) and let \(\alpha \in N_{1}(X)\) be a curve class with integer coefficients. The mobility of \(\alpha\) is defined by \[ \mathrm{mob}(\alpha) = \limsup_{m \rightarrow \infty} \frac{\max\{b \in \mathbb{Z}_{\geq 0} : \text{any } b \text{ general points are contained in an effective curve of class} \, m \alpha \}} {m^{n/(n-1)} /n!}. \] There is also a closely related function, the weighted mobility, which counts singular points of the curve. If \(\alpha \in N_{1}(X)\) with integer coefficients, then we define \begin{multline*} \mathrm{wmc}(\alpha) = \sup \max \big\{b \in \mathbb{Z}_{\geq 0}\ \big| \text{ there is an effective cycle of class } \mu \alpha \\ \text{ through any } b \text{ points of } X \text{ with multiplicity at least } \mu \text{ at each point} \big\}. \end{multline*} It is proved that the above supremum exists. We define the weighted mobility as \[ \mathrm{wmob}(\alpha) =\limsup_{m \rightarrow \infty} \frac{\mathrm{wmc}(m \alpha)}{m^{n / (n-1)}}. \] The main result of the paper can be formulated as follows. Main Result. Let \(X\) be a smooth projective variety of dimension \(n\) and let \(\alpha \in \overline{\mathrm{Eff}}_{1}(X)\) be a pseudo-effective curve class. Then i) \(\widehat{\mathrm{vol}}(\alpha) = \mathrm{wmob}(\alpha)\). ii) \(\widehat{\mathrm{vol}}(\alpha) \leq\mathrm{mob}(\alpha) \leq n! \widehat{\mathrm{vol}}(\alpha)\). iii) Assume that for any ample divisor \(A\) one has \(\mathrm{mob}(A^{n-1}) = A^{n}\), then \(\mathrm{mob}(\alpha) = \widehat{\mathrm{vol}}(\alpha)\). The main result is really interesting -- it suggests that the mobility function of any curve class is optimized by complete intersection curves. If \(\alpha\) denotes the class of a line in \(\mathbb{P}^{3}\), then the mobility count of \(\alpha\) boils down to the following question: what is the minimal degree of a curve through \(b\) general points in \(\mathbb{P}^{3}\). The answer to this question is unknown (which is very surprising since the question is really natural). Perrin conjectured that the optimal curves (maximizing the number of points relative to their degree to \(\frac{3}{2}\)) are complete intersectons of two divisors of the same degree. In this context, Main Result supports a vast generalization of Perrin's conjecture to all big curve classes on all smooth projective varieties. In order to prove the main result, the authors develop, en passant, the theory of volume functions, polar transforms of the volume function for divisors, etc. Let us recall the most interesting result, from my very subjective point of view. Theorem. Let \(X\) be a smooth projective variety of dimension \(n\). For any two big divisor classes \(L_{1}, L_{2}\), the inequality \[ \mathrm{vol}(L_{1} + L_{2})^{1/n} \geq\mathrm{vol}(L_{1})^{1/n} + \mathrm{vol}(L_{2})^{1/n} \] is an equality if and only if the numerical positive part \(P_{\sigma}(L_{1})\), \(P_{\sigma}(L_{2})\) are proportional. Thus the function \(L \mapsto\mathrm{vol}(L)\) is strictly log concave on the cone of big and movable divisors. Since one can find a plethora of interesting and new results therein, the reviewer strongly encourages to read the paper.