A direct approach to the S-functional calculus for closed operators

TL;DR

This paper introduces a direct Cauchy integral approach to the S-functional calculus for unbounded quaternionic operators, broadening its applicability and establishing key properties like the product rule and spectral projections.

Contribution

It defines the S-functional calculus directly via a Cauchy integral, allowing for operators without real points in their resolvent set and considering functions on non-connected sets.

Findings

The product rule and spectral mapping theorem hold under the new definition.

The calculus can generate spectral projections without previous restrictions.

Left and right slice hyperholomorphic calculi become inconsistent for certain functions.

Abstract

The $S$-functional calculus for slice hyperholomorphic functions generalizes the Riesz-Dunford-functional calculus for holomorphic functions to quaternionic linear operators and to $n$-tuples of noncommuting operators. For an unbounded closed operator, it is defined, as in the classical case, using an appropriate transformation and the $S$-functional calculus for bounded operators. This is however only possible if the $S$-resolvent set of the operator contains a real point. In this paper, we define the $S$-functional calculus directly via a Cauchy integral, which allows us to consider also operators whose resolvent sets do not contain real points. We show that the product rule and the spectral mapping theorem also hold true with this definition and that the $S$-functional calculus is compatible with polynomials, although polynomials are not included in the set of admissible functions if the operator is unbounded. We also prove that the $S$-functional calculus is able to create spectral projections. In order to do this, we remove another assumption usually made: we do not assume that admissible functions are defined on slice domains. Instead, we also consider functions that are defined on not necessarily connected sets. This leads to an unexpected phenomenon: the $S$-functional calculi for left and right slice hyperholomorphic functions become inconsistent and give different operators for functions that are both left and right slice hyperholomorphic. We show that any such function is the sum of a locally constant and an intrinsic function. For intrinsic functions both functional calculi agree, but for locally constant functions they must give different operators unless any spectral projection commutes with arbitrary scalars.