Randomization and the Gross-Pitaevskii hierarchy

TL;DR

This paper introduces a probabilistic approach to the Gross-Pitaevskii hierarchy on the torus, using randomization of Fourier coefficients to extend spacetime estimate ranges and analyze randomized collision operators.

Contribution

It extends the regularity range for spacetime estimates in the hierarchy using randomization, and studies convergence of randomized hierarchies in low regularity Sobolev spaces.

Findings

Averaged spacetime estimate holds for ta > 3/4

Convergence to zero of Duhamel expansions in low regularity spaces

Randomized collision operators may lead to understanding nonlinear randomization

Abstract

We study the Gross-Pitaevskii hierarchy on the spatial domain $\mathbb{T}^3$. By using an appropriate randomization of the Fourier coefficients in the collision operator, we prove an averaged form of the main estimate which is used in order to contract the Duhamel terms that occur in the study of the hierarchy. In the averaged estimate, we do not need to integrate in the time variable. An averaged spacetime estimate for this range of regularity exponents then follows as a direct corollary. The range of regularity exponents that we obtain is $\alpha>\frac{3}{4}$. It was shown in our previous joint work with Gressman that the range $\alpha>1$ is sharp in the corresponding deterministic spacetime estimate. This is in contrast to the non-periodic setting, which was studied by Klainerman and Machedon, in which the spacetime estimate is known to hold whenever $\alpha \geq 1$. The goal of our paper is to extend the range of $\alpha$ in this class of estimates in a \emph{probabilistic sense}. We use the new estimate and the ideas from its proof in order to study randomized forms of the Gross-Pitaevskii hierarchy. More precisely, we consider hierarchies similar to the Gross-Pitaevskii hierarchy, but in which the collision operator has been randomized. For these hierarchies, we show convergence to zero in low regularity Sobolev spaces of Duhamel expansions of fixed deterministic density matrices. We believe that the study of the randomized collision operators could be the first step in the understanding of a nonlinear form of randomization.