On the dynamics of the roots of polynomials under differentiation

TL;DR

This paper studies a nonlinear, nonlocal PDE related to polynomial roots under differentiation, proving local and global well-posedness results at critical regularity using sharp commutator estimates.

Contribution

It establishes local well-posedness in the critical space $H^{1/2}$ and extends to global well-posedness for positive initial data, addressing weak parabolicity.

Findings

Proved local well-posedness in $H^{1/2}$ space.

Established global well-posedness for positive initial data.

Developed a new strategy for lifespan depending on initial data profile.

Abstract

This article is devoted to the study of a nonlinear and nonlocal parabolic equation introduced by Stefan Steinerberger to study the roots of polynomials under differentiation; it also appeared in a work by Dimitri Shlyakhtenko and Terence Tao on free convolution. Rafael Granero-Belinch\'on obtained a global well-posedness result for positive initial data small enough in a Wiener space, and recently Alexander Kiselev and Changhui Tan proved a global well-posedness result for any positive initial data in the Sobolev space $H^s(\mathbb{S})$ with $s>3/2$. In this paper, we consider the Cauchy problem in the critical space $H^{1/2}(\mathbb{S})$. Two interesting new features, at this level of regularity, are that the equation can be written in the form $$ \partial_t u+V\partial_x u+\gamma \Lambda u=0, $$ where $\gamma$ is non-negative but not bounded from below and $V/\sqrt{\gamma}$ is not bounded. Therefore, the equation is only weakly parabolic. We prove that nevertheless the Cauchy problem is well posed locally in time and that the solutions are smooth for positive times. Combining this with the results of Kiselev and Tan, this gives a global well-posedness result for any positive initial data in $H^{1/2}(\mathbb{S})$. Our proof relies on sharp commutator estimates and introduces a strategy to prove a local well-posedness result in a situation where the lifespan depends on the profile of the initial data and not only on its norm.