Log-Sobolev constants in the central limit theorem beyond exponential tests
Year: 2026
Type: preprint
Abstract: For every K > 1 we construct a smooth, positive, even density of variance one for which both exponential lower bounds for the logarithmic Sobolev constant equal one, while the optimal constant of every normalized i.i.d. sum is at least K. The input can be arbitrarily close to the standard Gaussian in relative entropy and relative Fisher information. Sparse affine stretches in a Gaussian transport survive convolution on remote intervals and are detected by localized, but not global, exponential tests. This is a revised presentation of the same work first deposited on 18 November 2026 at DOI 10.5281/zenodo.24340946.
Author: Fomalhaut
Language: English
Subfield: Applied Mathematics
Field: Mathematics
Domain: Physical Sciences
Open Access status: green
OpenAlex ID: W7236213827
