Proof of the projective Lichnerowicz-Obata conjecture (Q874688)

Recall that a diffeomorphism between two Riemannian manifolds is said to be projective if it maps the geodesics of the first manifold onto the geodesics of the second. An affine diffeomorphism between two Riemannian manifolds is a diffeomorphism that takes the Levi-Civita connection of the first manifold to the Levi-Civita connection of the second manifold. With this terminology, the main theorem of the paper under review says that if \(n\geq2\) and \((M^n,g)\) is a complete connected Riemannian manifold acted on by a connected Lie group \(G\) by means of projective transformations, then either all of these transformations are affine or the manifold \((M^n,g)\) has constant positive sectional curvature. This theorem solves the so-called projective Lichnerowicz conjecture. In the special case when \((M^n,g)\) is actually a closed manifold, it either has constant positive sectional curvature or is acted on by means of isometries, which solves the projective Obata conjecture. The paper includes a detailed historical discussion on these conjectures along with an extensive bibliography. It is noteworthy that the aforementioned statements had been already proved in the case \(n=2\) in the author's earlier paper [Comment. Math. Helv. 80, No.~3, 541--570 (2005; Zbl 1113.53025)], using methods completely different from the ones that work in the case \(n\geq3\). The method of proof in higher dimensions relies on constructing commuting integrals for the geodesic flow as well as on a number of results on warped decompositions and Levi-Civita coordinates for Riemannian metrics which hold true only in dimensions higher than~3. The present paper also includes an interesting list of counterexamples illustrating the fact that all of the hypotheses in the main theorem are necessary.