Emanuel Milman (Technion)

Tuesday, May 26, 2015, 14:30 – 15:30, Math -101


Given an n-dimensional Riemannian manifold endowed with a probability density, we are interested in studying its isoperimetric, spectral and concentration properties. To this end, the Curvature-Dimension condition CD(K,N), introduced by Bakry and Emery in the 80’s, is a very useful tool. Roughly put, the parameter K serves as a lower bound on the weighted manifold’s “generalized Ricci curvature”, whereas N serves as an upper bound on its “generalized dimension”. Traditionally, the range of admissible values for the generalized dimension N has been confined to [n,infty]. In this talk, we present some recent developments in extending this range to N < 1, allowing in particular negative (!) generalized dimensions. We will mostly be concerned with obtaining sharp isoperimetric inequalities under the Curvature-Dimension condition, identifying new one-dimensional model-spaces for the isoperimetric problem. Of particular interest is when curvature is strictly positive, yielding a new single model space (besides the previously known N-sphere and Gaussian measure): the sphere of (possibly negative) dimension N<1, which enjoys a spectral-gap and improved exponential concentration.

Time permitting, we will also discuss the case when curvature is only assumed non-negative. When N is negative, we confirm that such spaces always satisfy an N-dimensional Cheeger isoperimetric inequality and N-degree polynomial concentration, and establish that these properties are in fact equivalent. In particular, this renders equivalent various weak Sobolev and Nash inequalities for different exponents on such spaces.