On torsion Kähler differential forms (Q1743007)

As a preliminary, the authors first discuss a series of questions related to the well-known Berger conjecture. In more appropriate terms, let \(X\) be an affine hypersurface over a field \(k\) of characteristic zero and \(\Omega^1_{X/k}\) the module of Kähler differential 1-forms. By their definition, for a given torsion differential \(\vartheta \in \text{Tors}\,\Omega^1_{X/k}\), the singular locus \(\text{Sing} (\vartheta)\) is the intersection of the vanishing locus of any lifting \(\widetilde{\vartheta}\) in the ambient affine space and the singular locus of \(X\). Using the language and the basic concepts from [\textit{E. R. Kolchin}, Differential algebra and algebraic groups. New York-London: Academic Press (1973; Zbl 0264.12102)], the authors extend this notion to the case of arbitrary affine algebraic varieties, describe five natural equivalent conditions under which a 1-form \(\vartheta\in \Omega^1_{X/k}\) is a torsion differential, and derive several statements concerning relationships between the singular loci of \(\vartheta\), \(\widetilde{\vartheta}\) and \(\text{Sing}\,X\). In conclusion, the main result of the paper is formulated and proved for a (reduced) hypersurface \(X\): if \(\vartheta \in \text{Tors}\, \Omega^1_{X/k}\), then for any lifting \(\widetilde{\vartheta}\) and for any ``noncylindrical'' closed point \(x\in X\), one has \(x\in \text{Sing} (\widetilde{\vartheta})\). However, simple examples show that this method does not work effectively when analyzing concrete situations that differ from the Whitney umbrella and its generalizations. In this connection it should be noted that the authors' notions and results one can present in a more clear form, taking into account currently available terminology and methods in the framework of the general theory of logarithmic differential forms developed by the reviewer in the late 1980s. More precisely, let \(h\) be a polynomial which determines the reduced hypersurface \(D\) in the ambient affine space \(V\) over the field of complex numbers, \(\Omega^1_V(\log D)\) the module of logarithmic differential 1-forms on \(V\) and \(\text{res}: \Omega^1_V(\log D)\rightarrow \omega^0_D\) the residue map. Then there is a natural correspondence between \(\text{Tors}\, \Omega^1_{D/k}\) and \(\Omega^1_V(\log D)\) given by the rule \({\vartheta \rightsquigarrow \widetilde{\vartheta}/h}\) (see, e.g., [\textit{A. G. Aleksandrov}, Funct. Anal. Appl. 22, No. 2, 131--133 (1988; Zbl 0663.32008); translation from Funkts. Anal. Prilozh. 22, No. 2, 59--60 (1988)], so that the set of singular points of the meromorphic function \(\text{res}(\widetilde{\vartheta}/h)\) is contained in the singular locus of \(D\). From this, one can deduce the required (or even improved) result in the hypersurface case, since \(\omega^0_D\) is the sheaf of locally bounded meromorphic functions in a punctured neighborhood of the distinguished point. Developing this approach, one can produce similar statements for the torsion modules \(\text{Tors}\, \Omega^p_{D/k}\), \(p\geqslant 1\), and, in a more general setting, for differential forms logarithmic along singular varieties over a perfect ground field (see, e.g., [\textit{A. G. Aleksandrov}, Funct. Anal. Appl. 51, No. 4, 245--254 (2017; Zbl 1386.58002); translation from Funkts. Anal. Prilozh. 51, No. 4, 3--15 (2017)]).