Some properties and applications of the new quantum $f$-divergences
Salman Beigi, Christoph Hirche, Marco Tomamichel
TL;DR
The paper studies the new quantum f-divergences defined via an integral representation, focusing on deriving alternative trace-like expressions, monotonicity, and convexity properties to unlock their full potential. It develops trace representations for H_α and D_α, proves inequalities comparing the new Renyi divergences to the Petz and sandwiched variants, and uses these tools to provide an alternative achievability proof for the quantum Chernoff bound. Additional integral representations yield a suite of inequalities, including reverse Pinsker-type bounds and χ^2-based controls, and a Taylor expansion framework clarifies higher-order corrections. Overall, the work broadens the analytical toolkit for quantum divergences, enabling tighter bounds in hypothesis testing, state discrimination, and information-processing tasks while clarifying the relationships among various quantum divergence families.
Abstract
Recently, a new definition for quantum $f$-divergences was introduced based on an integral representation. These divergences have shown remarkable properties, for example when investigating contraction coefficients under noisy channels. At the same time, many properties well known for other definitions have remained elusive for the new quantum $f$-divergence because of its unusual representation. In this work, we investigate alternative ways of expressing these quantum $f$-divergences. We leverage these expressions to prove new properties of these $f$-divergences and demonstrate some applications. In particular, we give a new proof of the achievability of the quantum Chernoff bound by establishing a strengthening of an inequality by Audenaert et al. We also establish inequalities between some previously known Renyi divergences and the new Renyi divergence. We further investigate some monotonicity and convexity properties of the new $f$-divergences, and prove inequalities between these divergences for various functions.
