R\'enyi relative entropies and noncommutative $L_p$-spaces

TL;DR

This paper extends Re9nyi relative entropies to general von Neumann algebras using noncommutative Lp-spaces, establishing their properties and connections to quantum information measures like the Araki-Masuda divergences.

Contribution

It introduces a new extension of Re9nyi relative entropies for von Neumann algebras, aligning with existing divergences and proving key properties such as data processing inequality.

Findings

Extension coincides with Araki-Masuda divergences

Proves data processing inequality for these divergences

Shows equality characterizes sufficiency of quantum channels

Abstract

We propose an extension of the sandwiched R\'enyi relative $\alpha$-entropy to normal positive functionals on arbitrary von Neumann algebras, for the values $\alpha>1$. For this, we use Kosaki's definition of noncommutative $L_p$-spaces with respect to a state. We show that these extensions coincide with the previously defined Araki-Masuda divergences [M. Berta et al., Annales Henri Poincar\'e, 19:1843--1867, 2018] and prove some of their properties, in particular the data processing inequality with respect to positive normal unital maps. As a consequence, we obtain monotonicity of the Araki relative entropy with respect to such maps, extending the results of [A. M\"uller-Hermes and D. Reeb. Annales Henri Poincar\'e,18:1777--1788, 2017] to arbitrary von Neumann algebras. It is also shown that equality in data processing inequality characterizes sufficiency (reversibility) of quantum channels.