Very general monomial valuations of \(\mathbb{P}^2\) and a Nagata-type conjecture (Q1688570)

In the paper under review, the authors present a Nagata-type statement in the language of valuation. More precisely, instead of asking whether the celebrated Nagata conjecture holds for \(s\) points in the complex projective plane, one would like to find a way to formulate this conjecture for real values \(t\geq 1\) of the number of points. At the very beginning it seems to be somehow surprising, but this attempt is a rather natural step which is taken in many situations ever since Zariski's pioneering works. For a valuation \(v\) centered at the origin of \(\mathbb{A}^{2} = \text{Spec} \,\mathbb{C}[x,y]\) one has by definition \[ \hat{\mu}(v) = \lim_{d \rightarrow \infty} \frac{\max\{v(f) : f \in \mathbb{C}[x,y], \deg f \leq d\}}{d}. \] All such invariants encode essentially the same information as the Seshadri constant does in the case of points and they turn out to be extremely hard to compute. The authors study the invariant \(\hat{\mu}\) as a function on the space \(\mathcal{V}\) of plane valuations of rank \(1\). This invariant turns out to be lower semicontinuous and continuous along the so-called arcs in \(\mathcal{V}\). Let \(\hat{\mu}(t) = \hat{\mu}(v_{t})\) with \(t \in [1, \infty)\), where \(v_{t}\) is a very general quasimonomial valuation with characteristic exponent \(t\) and denote the minimal proper birational model \(X_{t}\) where the center of \(v_{t}\) is a divisor. When \(X_{t}\) supports an effective anticanonical divisor, one can derive some interesting geometric properties. Theorem. Let \(v_{t}\) be a very general quasimonomial valuation on \(\mathbb{P}^{2}\) with characteristic exponent \(t \in \mathbb{Q}\), and let \(X_{t}\) be the minimal model where \(v_{t}\) has divisorial center, \(X_{t}\) supports an effective anticanonical divisor iff \(1 \leq t \leq 7\), \(t = 7 + \frac{1}{n}\) for some natural number \(n\), or \(t=9\). If \(1\leq t \leq 7\), then the Mori cone \(\overline{\mathrm{NE}}(X_{t})\) is a polyhedral cone spanned by the classes of the exceptional components of \(X_{t} \rightarrow \mathbb{P}^{2}\), the class of a particular nodal cubic, and finitely many \((-1)\)-curves. If \(t = 7 + \frac{1}{n}\) for natural \(n\), then the only prime divisors \(C\) in \(X_{t}\) with \(C^{2} \leq -2\) are exceptional components, and \(\overline{\mathrm{NE}}(X_{t})\) is a polyderal cone iff \(n \leq 8\). If \(1 \leq t \leq 3\), \(t = 3 + \frac{1}{n}\) for natural \(n\), \(t=5\), then the monoid of effective classes can be generated by the classes of the components of the exceptional divisor, a particular conic, and the \((-1)\)-curves. It is not difficult to see that \(\hat{\mu}(t) \geq \sqrt{t}\). However, the authors predict the following behaviour of \(\hat{\mu}\). Conjecture. For every \(t \geq 8 + \frac{1}{36}\) one has \(\hat{\mu}(t) = \sqrt{t}\). It turns out that Nagata's conjecture is just a special case of the above conjecture, so it makes sense to speak about \textit{fractional Nagata's conjecture}. The authors provide some evidence that the conjecture might be true. In order to do so, they use, for instance, the so-called Orevkov's rational cuspidal curves.