Multiplier ideals in two-dimensional local rings with rational singularities (Q737120): Difference between revisions

Consider an arbitrary ideal \(a\) on a regular complex surface germ \(X\), or, more generally, on the germ of a rational surface singularity \(X\). The main result in this paper is an efficient algorithm to compute sequentially the jumping numbers of \(a\), and also the whole chain of multiplier ideals of \(a\). More precisely, the \textsl{multiplier ideals} \(O_X \supset J(a^\lambda), \lambda \in \mathbb Q_{>0},\) are well known (singularity) invariants of \(a\), satisfying \(J(a^{\lambda}) \supseteq J(a^{\lambda'})\) when \(\lambda < \lambda'\). The rational numbers \(0<\lambda_1 < \lambda_2 < \dots\) where these ideals change are the \textsl{jumping numbers} of \(a\). There is a known complete list of candidate jumping numbers in terms of the minimal log principalization \(\pi:Y\to X\) of \(a\). Write the (effective) divisor associated to the principal ideal \(a O_Y\) as \(\sum_i e_i E_i\), where \(E_i, i\in I,\) are the irreducible components of its support. Write also the relative canonical divisor \(K_\pi\) as \(\sum_i k_i E_i\). Then the jumping numbers of \(a\) are included in the list \(\frac {k_i +m}{e_i}\), where \(i\in I\) and \(m\in \mathbb Z_{>0}\). \textit{K. Tucker} [Trans. Am. Math. Soc. 362, No. 6, 3223--3241 (2010; Zbl 1194.14026)] developed an algorithm, deciding whether such a candidate is a jumping number or not, yielding then the list of actual jumping numbers. (This list is periodic, one only has to investigate the range \((0,2]\).) The authors proceed in a different and more efficient way; they construct an algorithm that computes any jumping number in terms of the previous one (thus avoiding \` candidates\'). Their main technical result to achieve this is as follows. For a given \(\lambda \in \mathbb Q_{>0}\), let \(D_\lambda= \sum_i e_i^\lambda E_i\) be the \textsl{antinef closure} of the round down of the (\(\mathbb Q\)-)divisor \(\lambda\text{div}(a O_Y) - K_\pi\). Then the jumping number consecutive to \(\lambda\) is the minimum of the numbers \(\frac{k_i+1+e_i^\lambda}{e_i}, i\in I\). In addition, any multiplier ideal is computed from the \` previous one\'. Important in the arguments is the new concept of \textsl{minimal jumping divisor}. In the meantime, \textit{H. Baumers} and \textit{F. Dachs-Cadefau} extended some results of the present paper partly to higher dimensions [``Computing jumping numbers in higher dimensions'', \url{arXiv:1603.00787}].