Comparing numerical dimensions (Q380352): Difference between revisions

Consider \(L\) a pseudoeffective divisor on a smooth complex projective variety of dimension \(n\). There are, in the literature, several proposed definitions for a concept of \textit{numerical dimension of \(L\)}, measuring its positivity. One classical way to do it is to use the rate of growth of global sections of \(mL\) as \(m\) increase, the so called Iitaka dimension. Since a numerical invariant is needed, one must perturbate \(mL\) by adding a sufficiently ample divisor and compute its rate of growth. This provides definition (1) in Theorem 1.1 of the paper under review. A second way to proceed is by using intersection products. If \(L\) is nef, an interesting invariant is the maximal \(k\) such that there is an ample divisor \(A\) for which \(L^k\cdot A^{n-k} \neq 0\). But when \(L\) is not nef we have to discount the contribution of the (diminished) base locus of \(L\) to these products. This leads to definition (5) in Theorem 1.1. A third way will be to compute volumes, a loose analogue to the top self intersection \(L^n\), see (2) in the same Theorem 1.1. Finally one can produce a set of Seshadri-type constants, say the Nakayama constants, which measure the positivity of a line bundle along a subvariety and lead also to a notion of numerical dimension, see definition (7) of Theorem 1.1. This Nakayama constant is defined as the maximal value of the \(\tau\)'s such that \(\phi^*L-\tau E\) is pseudoeffective, being \(\phi\) a smooth resolution of the considered subvariety and \(E\) its corresponding exceptional divisor. In the paper under review, the author proves equality of all the different notions of numerical dimension, giving furthermore some additional characterizations of it. Several new properties of the numerical dimension are also provided.