Vertical components in fibre powers of analytic spaces. (Q1425101)

The purpose of this paper is to discuss the relationship between degeneracies of the family of fibres of an analytic mapping (as expressed by a failure of openness or flatness) and the existence of vertical components in fibre powers of the mapping. There are in fact two natural notions of a vertical component (algebraic and geometric) and the author is interested also in the relationship between them. Let \(f_\xi: X_\xi\to Y_\eta\), be a morphism of germs of analytic spaces. An irreducible (isolated or embedded) component \(W\) of \(X_\xi\) is called algebraic vertical if there exists a nonzero element \(a\in\mathcal O_{Y,\eta}\) such that (the pullback of) \(a\) belongs to the associated prime \(\mathfrak p\) in \(\mathcal O_{X,\,\xi}\) corresponding to \(W\). We say that \(W\) is geometric vertical if an arbitrarily small representative of \(W\) is mapped into a nowhere dense subset of a neighbourhood of \(\eta\) in \(Y\). The main result of the paper is the criterion for openness of complex analytic maps in terms of vertical components in fibre powers under the assumption that the source be puredimensional: Let \(X_\xi\) be puredimensional and let \(Y_\eta\) be irreducible of dimension \(n\). Then the following conditions are equivalent: (i) \(f_\xi\) is open, (ii) the \(n\)th fibre power \(X_{\xi^{\{n\}}}^{\{n\}}\) has no isolated geometric vertical components, (iii) \(X_{\xi^{\{n\}}}^{\{n\}}\) has no isolated algebraic vertical components. The equivalence (i) \(\Leftrightarrow\) (iii) could be restated as follows: \(f_\xi \) is open iff the reduced local ring \((\mathcal O_{X^{\{n\}},\,\xi^{\{n\}}})_{\text{red}}\) is a torsionfree \(\mathcal O_{Y,\eta}\)-module. This proposition is interesting in itself and plays an important role in providing a relationship between the isolated irreducible components in the fibre power \(X_{\xi^{\{n\}}}^{\{n\}}\) of \(X_\xi\) over \(Y_\eta\), and a filtration of \(Y\) by the fibre dimension of \(f\). As the author shows in Example 2.5 of the paper, the geometric and the algebraic approaches are not equivalent in general; that is, there are examples of bad behaviour of analytic mappings that can be detected by means of geometric vertical components but not by the algebraic ones. The fundamental reasons for this are that, in general, the algebraic vertical components do not detect a hidden Gabrielov irregularity, and that a fibre power of a regular map may itself be irregular. These are the main obstructions to extending results on geometric vertical components to algebraic vertical ones. The author briefly discusses criteria for flatness in terms of vertical components of fibre powers and gives a generalization of results on flatness by Auslander and by the author.