On Lie's approach to the study of translation manifolds (Q1080493)

A generalized double translation manifold is, by definition, an analytic hypersurface \(S\subset {\mathbb{C}}^{n+1}\) which has two distinct analytic parametrizations: \[ x_ i=\alpha_ i(t_ 1)+A_ i(t_ 2,...,t_ n)=\beta_ i(u_ 1)+B_ i(u_ 2,...,u_ n)\quad (i-1,...,n+1). \] Equivalently, a generalized double translation manifold S can be generated in two different ways by translating an analytic curve (\(\alpha\) or \(\beta)\) along an analytic subvariety of codimension 2 in \({\mathbb{C}}^{n+1}\) (A, or respectively, B). This paper presents a complete study of generalized double translation 3- folds in \({\mathbb{C}}^ 4\), following some ideas of Sophus Lie. Our major result is that if S is not ''developable'' (that is, the Gauß mapping on S is nondegenerate) and the generating curves and surfaces of a generalized double translation 3-fold S are in suitably general position, then S must in fact be a double translation 3-fold. Indeed, the generating surfaces A and B must be translation surfaces, so S has two other parametrizations of the even more special form: \[ X_ i=\alpha_ i(t_ 1)+A_{i2}(t_ 2')+A_{i3}(t_ 3')=\beta_ i(u_ 1)+B_{i2}(u_ 2')+B_{i3}(u_ 3')\quad (i=1,...,4), \] where \(t_ 2'\) and \(t_ 3'\) depend only on \(t_ 2\) and \(t_ 3\), and similarly for \(u_ 2'\) and \(u_ 3'.\) This result yields a new characterization of the Jacobian varieties of nonhyperelliptic curves of genus 4 among all 4-dimensional principally poarized abelian varieties. Namely, a principally polarized abelian variety \((A,\theta)\) is the Jacobian of a nonhyperelliptic curve if and only if \(\theta\) is a generalized double translation manifold. The proof consists of combining the main result mentioned above and the Lie- Wirtinger theorem on double translation manifolds. The results mentioned above may all be extended to generalized double translation manifolds of dimension n for all \(n\geq 3\); details of this will appear subsequently.