Superconcentration, and randomized Dvoretzky's theorem for spaces with 1-unconditional bases

Abstract

Let $n$ be a sufficiently large natural number and let $B$ be an origin-symmetric convex body in $R^n$ in the $\ell$-position, and such that the normed space $(R^n,\Vert \cdot {\Vert }_B)$ admits a $1$-unconditional basis. Then for any $\epsilon \in (0,1/2]$, and for random $c\epsilon \log n/\log \frac{1}{\epsilon }$-dimensional subspace $E$ distributed according to the rotation-invariant (Haar) measure, the section $B\cap E$ is $(1+\epsilon )$-Euclidean with probability close to one. This shows that the "worst-case" dependence on $\epsilon$ in the randomized Dvoretzky theorem in the $\ell$-position is significantly better than in John's position. It is a previously unexplored feature, which has strong connections with the concept of superconcentration introduced by S. Chatterjee. In fact, our main result follows from the next theorem: Let $B$ be as before and assume additionally that $B$ has a smooth boundary and ${\mathbb{E}}_{{\gamma }_n}\Vert \cdot {\Vert }_B\le n^c{\mathbb{E}}_{{\gamma }_n}\Vert {\mathrm{grad}}_B(\cdot ){\Vert }_2$ for a small universal constant $c>0$, where ${\mathrm{grad}}_B(\cdot )$ is the gradient of $\Vert \cdot {\Vert }_B$ and ${\gamma }_n$ is the standard Gaussian measure in $R^n$. Then for any $p\in [1,c\log n]$ the $p$-th power of the norm $\Vert \cdot {\Vert }_B^p$ is $\frac{C}{\log n}$--superconcentrated in the Gauss space.

removed the smoothness assumption in the main theorem