Slabs are more noise stable than balls at large Gaussian measure, for every correlation
Among symmetric subsets of $\mathbb{R}^n$ of prescribed Gaussian measure, which one maximizes the probability that two $\rho$-correlated standard Gaussian vectors both lie in the set? A centred ball or its complement was suggested, and Heilman showed that this fails in the plane to second order at $\rho=0$. We show that it fails for every $\rho\in(0,1)$ and every $n\ge2$: at measure close to $1$ the slab $\{|x_1|\le t\}$ is strictly more noise stable than every cylinder $B^k\times\mathbb{R}^{n-k}$ with $k\ge2$ and every cylinder complement, and at measure close to $0$ the complement of the slab wins. Computations in dimensions one, two and six are consistent with cylinders and their complements being the optimal sets.
Authors
- Yitzchak Shmalo (ORCID: https://orcid.org/0000-0002-2886-3049)
Institutions
- Hebrew University of Jerusalem (IL)
Publication Details
- Journal
- Zenodo (CERN European Organization for Nuclear Research)
- Published
- 2026-11-15
- DOI
- https://doi.org/10.5281/zenodo.24233414
- Primary Topic
- Point processes and geometric inequalities
- Type
- preprint