A rough curvature-dimension condition for metric measure spaces (Q2440580)

The goal of this article is to adapt the curvature-dimension condition, a substitute for lower Ricci curvature bounds in geodesic metric measure spaces, to a wider class of metric measure spaces, which may be discrete. The \(CD(K,N)\) condition of \textit{K.-T. Sturm} [Acta Math. 196, No. 1, 133--177 (2006; Zbl 1106.53032)] invovles the consideration of geodesics in the space of probability measures equipped with the \(L^2\) Wasserstein distance. In order to obtain a rough version of this condition, suitable for discrete spaces, the author introduces a parameter \(h>0\), which has the meaning of the mesh size. The notion of a geodesic is replaced by \(h\)-rough \(t\)-intermediate points, which can be introduced in two inequivalent ways. The combination of inequalities \(d(x_0,x_t)\leq td(x_0,x_1)+h\) and \(d(x_t,x_1)\leq (1-t)d(x_0,x_1)+h\) qualifies \(x_t\) as an \(h\)-rough \(t\)-intermediate point. If the inequality \((1-t)d^2(x_0,x_t)+t d^2(x_t,x_1) \leq t(1-t) d^2(x_0,x_1)+h^2\) holds, \(x_t\) is called an \(h\)-rough \(t\)-intermediate point in the strong sense. It turns out that the first definition is more stable under discretization while the stronger version leads to more powerful geometric consequences. Lastly, instead of an optimal coupling of probability measures the rough curvature-dimension condition involves \(\delta h\)-optimal coupling. The author shows that the rough curvature-dimension condition is stable under convergence of metric measure spaces, stable under their discretization, and leads to geometric results such as a form of the Brunn-Minkowski inequality.