Beyond the mixture of generalized Pauli dephasing channels

TL;DR

This paper challenges the common belief that all generalized Pauli channels can be expressed as convex combinations of a fixed number of dephasing channels, revealing new properties and distinctions between qubit and higher-dimensional cases.

Contribution

It demonstrates that the intuitive convex decomposition of generalized Pauli channels is inaccurate and explores the properties of broader convex combinations beyond the traditional limits.

Findings

Mixtures of invertible generalized Pauli channels remain invertible.

The decomposition of Pauli channels as mixtures of dephasing channels is specific to qubits.

Non-invertibility does not prevent channels from forming a Markovian semigroup.

Abstract

In recent times, there has been a growing scholarly focus on investigating the intricacies of quantum channel mixing. It has been commonly believed, based on intuition in the literature, that every generalized Pauli channel with dimensionality $d$ could be represented as a convex combination of $(d+1)$ generalized Pauli dephasing channels (see [Phys. Rev. A 103, 022605 (2021)] as a reference). To our surprise, our findings indicate the inaccuracy of this intuitive perspective. This has stimulated our interest in exploring the properties of convex combinations of generalized Pauli channels, beyond the restriction to just $(d+1)$ generalized Pauli dephasing channels. We demonstrate that many previously established properties still hold within this broader context. For instance, any mixture of invertible generalized Pauli channels retains its invertibility. It's worth noting that this property doesn't hold when considering the Weyl channels setting. Additionally, we demonstrate that every Pauli channel (for the case of $d=2$) can be represented as a mixture of $(d+1)$ Pauli dephasing channels, but this generalization doesn't apply to higher dimensions. This highlights a fundamental distinction between qubit and general qudit cases. In contrast to prior understanding, we show that non-invertibility of mixed channels is not a prerequisite for the resulting mapping to constitute a Markovian semigroup.