Coherent state coding approaches the capacity of non-Gaussian bosonic channels

TL;DR

This paper investigates the classical capacity of non-Gaussian bosonic channels, demonstrating that simple coherent state coding nearly achieves optimal capacity and that entanglement offers limited capacity gains.

Contribution

It extends the additivity results to broader classes of bosonic channels, showing minimal benefits from entanglement and the near-optimality of coherent state modulation.

Findings

Additivity violations are minor for studied channels.

Classical capacity gain is bounded independently of input energy.

Coherent state modulation is nearly optimal.

Abstract

The additivity problem asks if the use of entanglement can boost the information-carrying capacity of a given channel beyond what is achievable by coding with simple product states only. This has recently been shown not to be the case for phase-insensitive one-mode Gaussian channels, but remains unresolved in general. Here we consider two general classes of bosonic noise channels, which include phase-insensitive Gaussian channels as special cases: these are beamsplitters with general, potentially non-Gaussian environment states and classical noise channels with general probabilistic noise. We show that additivity violations, if existent, are rather minor for all these channels: the maximal gain in classical capacity is bounded by a constant independent of the input energy. Our proof shows that coding by simple classical modulation of coherent states is close to optimal.