Solvable groups, free divisors and nonisolated matrix singularities. II: Vanishing topology (Q2444361): Difference between revisions

In this interesting paper the authors describe how to compute the Milnor number of matrix singularities. Let \(\mathcal V\) be the variety of singular matrices in a space \(M\) of matrices. A map \(f:(\mathbb C^n,0) \to(M,0)\) defines a matrix singularity \(f^{-1}(\mathcal V)\). In particular, if \(M\) is the space of square matrices, the singularity is given by \(\det(f)\). Also skew-symmetric and general matrices are considered. In general matrix singularities are non-isolated singularities and a singular Milnor fibre is considered, which is obtained by perturbing the map \(f\) instead of the defining equations for the matrix singularity. This singular Milnor fibre is homotopy equivalent to a bouquet of spheres, and coincides with the usual Milnor fibre in the isolated singularity case. The authors give an inductive method to compute the singular Milnor number, analogous to that of Lê-Greuel for complete intersections, not using linear sections, but free divisors. The discriminants \(\mathcal W\) involved are not free divisors, but the authors find a free completion: a free divisor \(\mathcal V\) such that \(\mathcal W \cap \mathcal V\) is free. In the end one has to calculate the length of a number of determinantal modules. Reduction strategies to simplify the computation are given. The authors work out the procedure to obtain explicit formulas for symmetric and general \(2\times2\) and \(3\times 3\) matrices, and \(4\times4\) skew-symmetric matrices. The results are also applied to codimension 2 determinantal singularities, given as \(2\times 3\) matrix singularities. Although not a complete intersection, all deformations are determinantal, so it makes sense to talk about the singular vanishing Euler characteristic. The work of Pereira on the Milnor number of simple surface singularities is extended and her conjecture that \(\tau = \mu +1\) is verified for all remaining discrete cases and the start of series of singularities. For all simple 3-fold singularities in the list of \textit{A. Frühbis-Krüger} and \textit{A. Neumer} [Commun. Algebra 38, No. 2, 454--495 (2010; Zbl 1193.32015)] (for series of course only the first few terms are computed) the difference \(b_3-b_2\) is computed, as well as \(\tau\). Sometimes \(b_3-b_2\) seems to be constant in a series, but it also happens that \(\tau-(b_3-b_2)\) is constant. For Part I, see Ann. Inst. Fourier 65, No. 3, 1251--1300 (2015; Zbl 1371.14056).