Polynomials with general \(\mathbb{C}^ 2\)-fibers are variables. (Q700677)

By the embedding theorem of \textit{S.S.Abhyankar} and \textit{T.T. Moh} [J. Reine Angew. Math. 276, 148--166 (1975; Zbl 0332.14004)] and \textit{M. Suzuki} [J. Math. Soc. Japan 26, 241--257 (1974; Zbl 0276.14001)], an irreducible polynomial \(p\in\mathbb{C}[x,y]\) with a fibre \(p^{-1}(0)\simeq \mathbb{C}\) must be a variable. The analogous problem in higher dimensions is still open and related to many other classical problems in affine algebraic geometry, as the embedding problem, the linearization problem or the existence of exotic stuctures. There are special results - for instance, the proof of the linearization conjecture in dimension three [\textit{M. Koras} and \textit{P. Russell}, J. Algebr. Geom. 6, 671--695 (1997; Zbl 0882.14013); \textit{S. Kaliman, M. Koras, L. Makar-Limanov, P. Russell}, Electron. Res. Announc. Am. Math. Soc. 3, 63--71 (1997; Zbl 0890.14026)] implies that an irreducible polynomial on \(\mathbb{C}^3\) with zero fibre isomorphic to \(\mathbb{C}^2\) must be a variable if it is additionally quasi-invariant with respect to a regular \(\mathbb{C}^*\)-action. One should also note that recently \textit{S. Vénéreau}, in his Ph.D. thesis [\texttt{http://www-fourier.ujf-grenoble.fr/THESE/ps/t111.ps.gz}; see also \textit{S. Kaliman} and \textit{M. Zaidenberg}, J. Pure Appl. Algebra 192, 275--286 (2004; Zbl 1062.14074)], found a very interesting class of polynomials \(v_n:=y+x^n(xz+y(yu+z^2))\) on \(\mathbb{C}^4\) that have all fibers isomorphic to the affine space \(\mathbb{C}^3\) which are variables for \(n\geq 3\) but nothing is known for \(n=1,2\). In the paper under review, the author follows results of \textit{M. Miyanishi} [in: Algebraic Geometry Seminar, Singapore, 1987, 53--67 (World Sci. Publishing, Singapore) (1988)] and \textit{A.Sathaye} [Invent. Math. 74, 159--168 (1983; Zbl 0538.13006)] which imply that \(p\in \mathbb{C}[x,y,z]\) is a variable if a generic fibre is isomorphic to \(\mathbb{C}^2\) and all fibres are UFDs. The aim is to substitute the condition on the generic fibre by topological assumptions. The author proves the following main theorem: Let \(X\) be an affine algebraic variety of dimension three such that the coordinate ring of \(X\) is a UFD and has only constants as units, and there is an open subset \(U'\subseteq X\) and a morphism \(p: U'\to U\), \(U\) an affine smooth curve whose fibres are isomorphic to \(\mathbb{C}^2\). If either \((1)\) \(X\) is smooth, \(H_3(X)=0\) and each irreducible component of \(X\setminus U'\) has at most isolated singularities, or \((1')\) the Euler characteristic of \(X\) equals one and each irreducible component of \(X\setminus U'\) is a UFD, then \(U\) is an open subset of \(\mathbb{C}\) and \(p\) can be extended to a morphism \(\bar{p}:X\to \mathbb{C}\). Furthermore, \(X\simeq \mathbb{C}^3\) and \(\bar{p}\) is a variable.