Isometric immersions of Riemannian products revisited (Q1343879)

Let \(M^n=M_1^{n_1}\times M_2^{n_2}\) be a Riemannian product of two connected complete Riemannian manifolds. Assume \(\dim M^{n_i}_i=n_i\geq 2\), \(1\leq i\leq 2\), and that no \(M_i\) either is flat everywhere or contains an ``Euclidean strip'', that is, an open submanifold which is isometric to the Riemannian product \(I\times \mathbb{R}^{n_i-1}\), where \(I\subset \mathbb{R}\) denotes an open interval. Under these assumptions, it was proved in a beautiful paper by \textit{S. Alexander} and \textit{R. Maltz} [J. Differ. Geom. 11, 47-57 (1976; Zbl 0334.53053)] that any isometric immersion \(f:M^n\to \mathbb{R}^{n+2}\) is a Riemannian product of hypersurface immersions. This means that there exist an orthogonal factorization \(\mathbb{R}^{n+2}=\mathbb{R}^{n_1+1}\times \mathbb{R}^{n_2+1}\) and isometric immersions \(f_i:M^{n_i}_i\to \mathbb{R}^{n_i+1}\), \(1\leq i\leq 2\), such that \(f(x_1,x_2)=(f_1(x_1),f_2(x_2))\). The main goal of this paper is to provide an understanding of the possible cases for which \(f:M^n=M^{n_1}_1\times M^{n_2}_2\to \mathbb{R}^{n+2}\) may fail to be a Riemannian product of hypersurface immersions.