Abstract
We investigate the zero set of a stationary Gaussian process on the real line, and in particular give lower bounds for the variance of the number of points and of linear statistics on a large interval, in all generality. We prove that this point process is never hyperuniform, i.e. the variance is at least linear, and give a necessary condition to have linear variance, which is close to be sufficient. We study the class of symmetric Bernoulli convolutions and give an example where the zero set is maximally rigid, weakly mixing, and not hyperuniform.
On étudie l’ensemble formé par les zéros d’un processus Gaussien stationnaire sur la droite réelle, et on donne en particulier des bornes inférieures sur la variance du nombre de zéros et des statistiques linéaires sur un grand intervalle, en toute généralité. On montre que ce processus de points n’est jamais hyperuniforme, i.e. la variance est toujours au moins linéaire, et on donne une condition nécessaire pour avoir une variance linéaire, cette condition est proche d’être suffisante. On étudie la classe des convolutions symmétriques de Bernoulli et donnons un exemple où l’ensemble des zéros est maximalement rigide, faiblement mélangeant, et pas hyperuniforme.
Funding Statement
This study was supported by the IdEx Université de Paris, ANR-18-IDEX-0001.
Acknowledgements
I am indebted to M. A. Klatt, who introduced to me the fascinating topic of hyperuniformity and with whom I had many fruitful discussions on this subject. I warmly thank Anne Estrade, with whom I had many discussions along the elaboration of this work, about Gaussian processes and their zeros. I also wish to thank Jose Léon for a discussion around the variance of Gaussian zeros.
Citation
Raphaël Lachièze-Rey. "Variance linearity for real Gaussian zeros." Ann. Inst. H. Poincaré Probab. Statist. 58 (4) 2114 - 2128, November 2022. https://doi.org/10.1214/21-AIHP1228
Information