Converging State Distributions for Discrete Modulated CVQKD Protocols

Abstract

Consider the problem of using a finite set of coherent states to distribute secret keys over a quantum channel. It is known that computing the exact secret key rate in this scenario is intractable due to the infinite dimensionality of the Hilbert spaces and usually one computes a lower bound using a Gaussian equivalent bipartite state in the entangled based version of the protocol, which leads to underestimating the actual protocol capability of generating secret keys for the sake of security. Here, we define the QKD protocol's non-Gaussianity, a function quantifying the amount of secret key rate lost due to assuming a Gaussian model when a non-Gaussian modulation was used, and develop relevant properties for it. We show that if the set of coherent states is induced by a random variable approaching the AWGN channel capacity, then the protocol's non-Gaussianity vanishes, meaning that there is no loss of secret key rate due to the use of a Gaussian model for computing bound on the secret key rate. The numerical results show that by using a 256-QAM with Gauss-Hermite shaping, the loss of secret key rate quickly falls below $10^{-5}$ as the distance increases.