On K-closedness, BMO-regularity and real interpolation of Hardy-type spaces

TL;DR

This paper proves the equivalence between BMO-regularity and K-closedness for Hardy-type spaces within certain Banach and quasi-Banach lattice couples, and characterizes when a specific interpolation formula holds.

Contribution

It establishes the equivalence of BMO-regularity and K-closedness for Hardy-type spaces in a broad setting, extending previous conjectures and providing conditions for interpolation formulas.

Findings

BMO-regularity is equivalent to K-closedness in certain lattice couples.

The 'good interpolation' formula holds if and only if X is BMO-regular.

Results apply to both discrete and arbitrary measure spaces.

Abstract

Let $(X, Y)$ be a suitable couple of quasi-Banach lattices of measurable functions on $\mathbb T \times \Omega$, and let $(X_A, Y_A)$ be the couple of the corresponding Hardy-type spaces. It has long been suspected that the BMO-regularity property of $(X, Y)$ is not only sufficient for the $\mathrm K$-closedness of $(X_A, Y_A)$ in $(X, Y)$ but also necessary. We establish the equivalence of these two properties for a general couple of Banach lattices having the Fatou property when $\Omega$ is a discrete measurable space, and also for couples $(X, Y)$ where $X$ is allowed to be quasi-Banach but $Y$ is assumed to be $p$-convex with some $p > 1$ (here $\Omega$ is arbitrary). We show under certain mild restrictions that the "good interpolation" formula $$ \left(X_A, \mathrm H_q\right)_{\theta, p} = \left[\left(X, \mathrm L_q\right)_{\theta, p}\right]_A $$ holds true if and only if $X$ is BMO-regular.