Secret key rate enhancement through joint precoding and power loading for downstream CV-QKD over passive optical networks

Passive optical network (PON)-based continuous-variable quantum key distribution (CV-QKD) enables an optical line terminal (OLT) to establish multiple information-theoretically secure keys with optical network units (ONUs) over untrusted quantum channels. However, the average secret key rate (SKR) of PON systems is severely constrained by heterogeneous channel conditions among ONUs. Hardware-based solutions for PON systems, such as wavelength-division multiplexing and multi-core fiber, are also burdened by prohibitive costs and implementation complexity. Conventional signal-processing approaches that rely solely on power regulation are suboptimal for CV-QKD because the SKR exhibits a non-monotonic dependence on the channel transmit power. Although unitary precoding can effectively enhance the SKR of point-to-point links by exploiting the inherently non-flat noise frequency spectrum of practical CV-QKD systems, extending this technique to multi-user PONs introduces unique challenges: all users share a single global transmit power budget and a common precoding matrix, thereby requiring joint optimization to maximize the average SKR. To address these challenges, we propose what we believe is a novel time-domain precoding scheme that integrates precoding with power loading and jointly optimizes the unitary precoding matrix and power allocation profile by exploiting the inherently non-flat noise frequency spectrum of practical CV-QKD systems. Numerical simulations demonstrate that, for a standard 20 km link with 64 ONUs, the proposed scheme improves the SKR by factors of 2.52 to 4.57 over conventional unoptimized systems and by factors of 1.02 to 1.90 over PON systems employing a discrete Fourier transform (DFT) precoding matrix with optimized power loading.