Phase space analysis of the Hermite semigroup and applications to nonlinear global well-posedness

TL;DR

This paper analyzes the Hermite operator and its fractional powers in phase space, establishing estimates for the associated semigroup and applying these results to prove global well-posedness for nonlinear heat equations with small initial data.

Contribution

It provides the first complete range of fixed-time estimates for the Hermite semigroup on modulation spaces and applies these to nonlinear PDEs, demonstrating optimal decay and phase-space smoothing.

Findings

Established optimal decay rates for the Hermite semigroup on modulation spaces.

Proved global well-posedness for nonlinear heat equations with small initial data.

Showed solutions decay exponentially at the same rate as the linear equation.

Abstract

We study the Hermite operator $H=-\Delta+|x|^2$ in $\mathbb{R}^d$ and its fractional powers $H^\beta$, $\beta>0$ in phase space. Namely, we represent functions $f$ via the so-called short-time Fourier, alias Fourier-Wigner or Bargmann transform $V_g f$ ($g$ being a fixed window function), and we measure their regularity and decay by means of mixed Lebesgue norms in phase space of $V_g f$, that is in terms of membership to modulation spaces $M^{p,q}$, $0< p,q\leq \infty$. We prove the complete range of fixed-time estimates for the semigroup $e^{-tH^\beta}$ when acting on $M^{p,q}$, for every $0< p,q\leq \infty$, exhibiting the optimal global-in-time decay as well as phase-space smoothing. As an application, we establish global well-posedness for the nonlinear heat equation for $H^{\beta}$ with power-type nonlinearity (focusing or defocusing), with small initial data in modulation spaces or in Wiener amalgam spaces. We show that such a global solution exhibits the same optimal decay $e^{-c t}$ as the solution of the corresponding linear equation, where $c=d^\beta$ is the bottom of the spectrum of $H^\beta$. This is in sharp contrast to what happens for the nonlinear focusing heat equation without potential, where blow-up in finite time always occurs for (even small) constant initial data - hence in $M^{\infty,1}$.