Counterexamples to the nonsimply connected double soul conjecture (Q6140365): Difference between revisions

A double disk bundle is a closed smooth manifold obtained by gluing the total spaces of two closed disk bundles whose total spaces of their respective sphere-bundle boundaries are diffeomorphic. Double disk bundles crop up, among other places, in the study of closed smooth Riemannian manifolds with cohomogeneity one isometric actions, since in the case where the orbit space is a closed interval, the tubular neighborhood theorem provides such a decomposition where the common boundary sphere bundle is diffeomorphic to a principal orbit. \textit{K. Grove} conjectured in [\textit{S.-Y. A. Chang} et al., Conformal, Riemannian and Lagrangian geometry. The 2000 Barrett lectures. Edited by Alexandre Freire. Providence, RI: American Mathematical Society (AMS) (2002; Zbl 1019.53001)] that any closed, simply connected, non-negatively curved Riemannian manifold is a double disk bundle. Grove himself demonstrated that if one drops the condition of simple connectivity from the conjecture, it becomes false by showing the Poincaré dodecahedral space \(S^3/I^*\), (where \(I^*\) denotes the binary icosahedral group) is positively curved but not a double disk bundle. The author expands on Grove's idea by providing two other positively curved, non-simply connected 3-manifolds that are not double disk bundles, namely \(S^3/T^*\) and \(S^3/O^*\), where \(T^*\) and \(O^*\) denote the binary tetrahedral and binary octahedral groups, respectively. In the last section, the author builds on an idea that Belegradek expressed in an online forum of using an old result of \textit{L. Auslander} and \textit{M. Kuranishi} [Ann. Math. (2) 65, 411--415 (1957; Zbl 0079.38304)] that \(\pi\) is the fundamental group of a closed flat \(n\)-manifold precisely when it is torsion-free and fits into a short exact sequence \(0\rightarrow\mathbb{Z}^n \rightarrow \pi \rightarrow \phi \rightarrow 0\) where \(\phi\) is a finite group. If \(0\rightarrow\mathbb{Z}^n \rightarrow \pi \rightarrow \phi \rightarrow 0\) is a short exact sequence, then \(0\rightarrow\mathbb{Z}^n \rightarrow [\pi,\pi] \rightarrow [\phi,\phi] \rightarrow 0\) is as well, and if \([\phi,\phi]=\phi\) then \([\pi,\pi]=\pi\). Using L. Auslander and M. Kuranshi's work [loc. cit.], such a closed flat \(n\)-manifold \(M_\phi\) has trivial first homology since \(H_1(M_\phi) = \pi/[\pi,\pi]\) is trivial. But then \(H_1(M_\phi)\) has no surjection onto \(\mathbb{Z}/2\mathbb{Z}\), so by the reasoning of Proposition 5.2, \(M_\phi\) is not a double disk bundle. By taking \(\phi=A_n\) to be the alternating group on \(n\) letters, where \(n \geq 4\), one has \([A_n,A_n]=A_n\). Thus, the author obtains infinity many closed flat \(n\)-manifolds \(M_{A_n}\) that are not disk bundles. Namely, he provides infinitely many further counterexamples to the non-simply connected version of Grove's conjecture above.