H\"ormander functional calculus on UMD lattice valued $L^p$ spaces under generalised Gaussian estimates

TL;DR

This paper extends H"ormander functional calculus to UMD lattice-valued $L^p$ spaces under general Gaussian estimates, providing new spectral multiplier results and boundedness of maximal operators.

Contribution

It establishes H"ormander calculus for tensorized operators on UMD lattice-valued spaces under Gaussian estimates, including new boundedness results for Hardy-Littlewood maximal operators.

Findings

H"ormander calculus extends to UMD lattice-valued $L^p$ spaces.

Hardy-Littlewood maximal operator is bounded on these spaces.

Spectral multipliers satisfy square function estimates.

Abstract

We consider self-adjoint semigroups $T_t = \exp(-tA)$ acting on $L^2(\Omega)$ and satisfying (generalised) Gaussian estimates, where $\Omega$ is a metric measure space of homogeneous type of dimension $d$. The aim of the article is to show that $A \otimes \mathrm{Id}_Y$ admits a H\"ormander type $\mathcal{H}^\beta_2$ functional calculus on $L^p(\Omega;Y)$ where $Y$ is a UMD lattice, thus extending the well-known H\"ormander calculus of $A$ on $L^p(\Omega)$. We show that if $T_t$ is lattice positive (or merely admits an $H^\infty$ calculus on $L^p(\Omega;Y)$) then this is indeed the case. Here the derivation exponent has to satisfy $\beta > \alpha \cdot d + \frac12$, where $\alpha \in (0,1)$ depends on $p$, and on convexity and concavity exponents of $Y$. A part of the proof is the new result that the Hardy-Littlewood maximal operator is bounded on $L^p(\Omega;Y)$. Moreover, our spectral multipliers satisfy square function estimates in $L^p(\Omega;Y)$. In a variant, we show that if $e^{itA}$ satisfies a dispersive $L^1(\Omega) \to L^\infty(\Omega)$ estimate, then $\beta > \frac{d+1}{2}$ above is admissible independent of convexity and concavity of $Y$. Finally, we illustrate these results in a variety of examples.