Vanishing sequences and Okounkov bodies (Q2339323): Difference between revisions

The authors study the so-called vanishing sequences, which appear naturally in the theory of algebraic curves, and the main aim is to generalize this notion to higher dimensional projective varieties. Let \(X\) be a normal projective variety over an algebraically closed field \(\mathbb{K}\) of arbitrary characteristic. Let \(L\) be a line bundle such that \(H^{0}(X,L) \neq 0\). For any real valuation \(v\) on \(X\) one defines the following real filtration \[ \mathcal{F}_{v}^{t} H^{0}(X,L) = \bigg\{ s \in H^{0}(X,L) : v(s) \geq t \bigg\}, \,\, t \in \mathbb{R}, \] and we then have \[ a_{j}(L,v) = inf \bigg\{t \in \mathbb{R} : \mathrm{codim} \mathcal{F}_{v}^{t} H^{0}(X,L) \geq j \bigg\} \] for \(j = 1,\dots, N\) with \(N = h^{0}(X,L)\). Using the above numbers \(a_{j}\) we can define the \textit{vanishing sequence} of \(H^{0}(X,L)\) along \(v\) as \[ a_{min}(L,v) = a_{1}(L,v) \leq \dots \leq a_{N}(L,v) = a_{\max}(L,v). \] Under the assumption that \(L\) is big, which means that \(H^{0}(X, mL) \neq 0\) for \(m >>1\), one can show the existence of \[ a_{\min}(||L||,v) := \lim_{m \rightarrow \infty} m^{-1}a_{\min}(mL,v) \in [0, +\infty) \] and \[ a_{\max}(||L||,v) := \lim_{m \rightarrow \infty} m^{-1}a_{\max}(mL,v) \in (0, +\infty]. \] Moreover, we say that \(v\) has linear growth when \(a_{\max}(||L||,v)\) is finite, i.e. when the values of \(v\) on sections in \(H^{0}(X,mL)\) grow at most linearly with \(m\). One of the first results of this paper tells us that if \(v\) is the so-called Abhyankar valuation, then \(v\) has linear growth. The main result of this note can be formulated as follows. Theorem. Let \(L\) be a big line bundle on a normal projective variety \(X\) and denote \(N_{m} = h^{0}(X,mL)\). {\parindent=6mm \begin{itemize} \item[i)] For any real valuation \(v\) on \(X\), the scaled vanishing sequence \[ \bigg(m^{-1}a_{j}(mL,v)\bigg)_{1 \leq j \leq N_{m}} \] equidistributes as \(m\rightarrow \infty\), in the sense that the sequence of discrete probability measures \[ v_{k} := \frac{1}{N_{m}} \sum_{j} \delta_{m^{-1}a_{j}(mL,v)} \] converges weakly to a positive measure \(\mu_{L,v}\) on \(\mathbb{R}\). \item [ii)] If \(v\) has linear growth, then \(\mu_{L,v}\) is a probability measure supported on the interval \[ [a_{\min}(||L||,v); a_{\max}(||L||,v)], \] and its singular part with respect to the Lebesgue measure consists of at most a Dirac mass at \(a_{\max}(||L||,v)\). \item [iii)] When \(v\) does not have linear growth we have \(\mu_{L,v} = 0\). \end{itemize}} Assuming that \(v\) is divisorial, then one can provide an explicit formulae for the density of \(\mu_{L,v}\) in terms of restricted volumes [\textit{L. Ein} et al., Am. J. Math. 131, No. 3, 607--651 (2009; Zbl 1179.14006)], and if \(v\) has linear growth with \(L\), then one can also interpret the limit measure \(\mu_{L,v}\) as the push-forward of the normalized Lebesgue measure on the Okounkov body \(\triangle(L)_{\mathcal{Y}_{\bullet}} \subset \mathbb{R}^{n}\) with respect to any given admissible flag \(\mathcal{Y}_{\bullet}\) by a concave non-negative upper-semicontinuous function \[ G_{L,v} : \triangle(L)_{\mathcal{Y}_{\bullet}} \rightarrow \mathbb{R}_{+}, \] which is called the \textit{concave transform}, or sometimes the \textit{Okounkov function} on \(\triangle(L)_{\mathcal{Y}_{\bullet}}\). Another result of this paper tells us that \(G_{L,v}\) depends only on the numerical equivalence class of \(L\) (just like Okounkov body itself) and also the authors characterize continuity of concave transforms. It turns out that for surfaces all concave transforms on Okounkov bodies are continuous, and in general these concave transforms are continuous up to the boundary (see Example 3.9 therein).