$\mathbf{RP}^n \# \mathbf{RP}^n$ and some others admit no real projective structure
Suhyoung Choi
TL;DR
The paper proves that $\mathbb{RP}^n \# \mathbb{RP}^n$ (and a class of related manifolds) admit no real projective structure for $n\ge 3$, by refining Benoist’s classification of manifolds with infinite-cyclic holonomy and leveraging octantizability. The authors develop a streamlined approach that combines deformation-space analysis, diagonalizable holonomy perturbations, and Kuiper completion to derive topological obstructions, culminating in a constructive brick decomposition into Benoist- or Hopf-type pieces. This reduces the problem to ruling out compatible global actions on the developing image, and the method extends to generalizations beyond the $n=3$ Cooper–Goldman/Çoban cases. The work thus clarifies the landscape of real projective structures on high-dimensional manifolds and provides a robust toolkit for excluding such structures via holonomy-end geometry and deformation theory, with a focus on diagonalizable and nilpotent holonomy cases.
Abstract
A manifold $M$ possesses a real projective structure if it has an atlas consisting of charts mapping to $\mathbf{S}^n$, where the transition maps lie in $\mathrm{SL}_\pm(n+1, \mathbf{R})$. In this context, we present a concise proof demonstrating that $\mathbf{RP}^n\#\mathbf{RP}^n$ and a few other manifolds do not possess a real projective structure when $n\geq3$. Notably, our proof is shorter than those provided by Cooper-Goldman for $n=3$ and Çoban for $n\geq 4$. To do this, we reprove the classification of closed real projective manifolds with infinite-cyclic holonomy groups by Benoist due to a small error. We will leverage the concept of the octantizability of real projective manifolds with nilpotent holonomy groups, as introduced by Benoist and Smillie, which serves as a powerful tool.
