Spectral multipliers in group algebras and noncommutative Calderón-Zygmund theory

Abstract

In this paper we solve three problems in noncommutative harmonic analysis which are related to endpoint inequalities for singular integrals. In first place, we prove that an L 2 -form of Hörmander's kernel condition suffices for the weak type (1,1) of Calderón-Zygmund operators acting on matrix-valued functions. To that end, we introduce an improved CZ decomposition for martingale filtrations in von Neumann algebras, and apply a very simple unconventional argument which notably avoids pseudolocalization. In second place, we establish as well the weak L 1 endpoint for matrix-valued CZ operators over nondoubling measures of polynomial growth, in the line of the work of Tolsa and Nazarov/Treil/Volberg. The above results are valid for other von Neumann algebras and solve in the positive two open problems formulated in 2009. An even more interesting problem is the lack of L 1 endpoint inequalities for singular Fourier and Schur multipliers over nonabelian groups. Given a locally compact group G equipped with a conditionally negative length ψ : G → R + , we prove that Herz-Schur multipliers with symbol m • ψ satisfying a Mikhlin condition in terms of the ψ-cocycle dimension are of weak type (1, 1). Our result extends to Fourier multipliers for amenable groups and impose sharp regularity conditions on the symbol. The proof crucially combines our new CZ methods with novel forms of recent transference techniques. This L 1 endpoint gives a very much expected inequality which complements the L∞ → BMO estimates proved in 2014 by Junge, Mei and Parcet.

Schur multipliers in Schatten-von Neumann classes

“…In particular, the kernel representation above is meaningful when z / ∈ supp R n f , the Euclidean support of the matrix-valued function f . This is exactly the framework of the noncommutative Calderón-Zygmund theory developed in [8,26,39]. In our use of CZ methods, it will be crucial to weaken the kernel condition given in [26,Lemma 2.3].…”