Resolution except for minimal singularities. II. The case of four variables (Q715224)

This is a sequel to Part I of this article, \textit{E. Bierstone} and \textit{P. Milman} [Adv. Math. 231, 3022--3053 (2012; Zbl 1257.14002)], where the authors investigate ``partial resolutions'' of an algebraic variety \(X\) over a characteristic zero field. That is, in the process we avoid modifying points of a certain type (e.g. normal crossings.) More precisely, consider sets \(K\) of equivalence classes of singularities, where we identify two if they become isomorphic after base field extension (if necessary) and completion. Let \(X^K\) denote the set of points of \(X\) which are regular or have a singularity in \(K\). If \(K\) consists of normal crossings points, we write \(X^K=X^{nc}\). In the present paper, assuming \(\dim X = 3\), the authors determine the smallest set \(S\) of classes of singularities, such that there is a morphism \(\sigma:X' \to X\), which is a composition of permissible blowing-ups, such that \({(X')}^S=X'\) and \(\sigma\) induces an isomorphism over \(X^{nc}\). The problem can be reduced to that for embedded surfaces, and they give a list of representatives of the classes in \(S\), defined by particularly simple equations (in four variables), called normal forms. This list \(S\) has seven elements. It is not true that \(\sigma\) is necessarily an isomorphism over \(X^S\). To obtain such a result, it is necessary to enlarge \(S\). In fact, adding just one more normal form, they get a list \(S'\), so that one obtains a morphism \(\bar \sigma: {\bar X} \to X\) as above, with \({\bar X}^{S'}= {\bar X}\), inducing an isomorphism over \(X^{S'}\). In Part I the case \(\dim X <3\) was studied, in that situation \(S=S'\). The main technique used comes from \textit{E. Bierstone}--\textit{P. Milman}'s main resolution paper [Invent. Math. 128, No. 2, 207--302 (1997; Zbl 0896.14006)], specially their function \(inv\). But they need other auxiliary results, possibly of independent interest, which are proved in this paper. E.g., they give a characterization of singularities which are limits of triple normal crossings and cannot be eliminated by means of blowing-ups whose centers do not include normal crossings singularities. They also use a ``cleaning lemma'', proved in part I.