From volume cone to metric cone in the nonsmooth setting (Q730024)

In Riemannian geometry, a volume cone is a Riemannian manifold such that the volume of metric balls \(B(x,r)\) is proportional to \(r^n\). On the other hand, one says that \(B_R(x)\setminus B_r(x)\) is a metric cone if it is a warped product \((N\times\left[r,R\right],dt^2+t^2g_N)\) for some closed manifold \((N,g_N)\). A tangent cone of a Riemannian manifold \((M,g)\) means a Gromov-Hausdorff limit of \((M,r_j^{-2}g)\) for a sequence \(r_j\to\infty\). The volume comparison theorem for Riemannian manifolds \(M\) of nonnegative Ricci curvature asserts that the volume of metric balls \(B_r(x)\) grows more slowly than the volume of metric balls in Euclidean space: \[ \frac{d}{dr}\left(\frac{\mathrm{vol }(B_r(x)\subset M)}{\mathrm{vol } (B_r\subset{\mathbb R}^n)}\right)\leq 0. \] The equality case \(\frac{d}{dr}\left(\frac{\mathrm{vol }(B_r(x)\subset M)}{\mathrm{vol } (B_r\subset{\mathbb R}^n)}\right)=0\) means exactly that one has a volume cone and the rigidity part of the volume comparison theorem asserts that if a Riemannian manifold of nonpositive Ricci curvature is a volume cone, then it is a metric cone. In [Ann. Math. (2) 144, No. 1, 189--237 (1996; Zbl 0865.53037)], \textit{J. Cheeger} and \textit{T. H. Colding} proved that a Riemannian manifold with \(\mathrm{Ric}\geq 0\) and almost maximal volume must be Gromov-Hausdorff close to a metric cone. In particular, if \(\mathrm{Ric}\geq 0\) and \(\mathrm{vol }(B,x,r)\geq cr^n\) for some \(c>0\), then each tangent cone is a metric cone. The purpose of the paper under review is to generalize these results to the non-smooth setting, that is, to metric-measure spaces which have nonpositive Ricci curvature in a synthetic sense. A natural setting for such a generalisation would have been the \(CD(0,n)\) and \(CD^*(0,n)\) spaces introduced by Lott-Villani and Bacher-Sturm. However, they include some Finsler geometries for which the wanted rigidity does not always hold. For this reason the authors restrict to a class of spaces called \(RCD^*(0,n)\) which has been introduced by the second author [Mem. Am. Math. Soc. 1113, iii-v, 91 p. (2015; Zbl 1325.53054)] and can be characterised by a finite-dimensional Bochner inequality. The main result of the paper is then that for a ball in a \(RCD^*(0,n)\) space, being a volume cone implies being locally isometric to the cone over a ball in an \(RCD^*(n-2,n-1)\) space. (There are two exceptional cases in which the ball is just \(1\)-dimensional.) The consequences for tangent cones will be analyzed in subsequent papers.