Design of a self-encoding and all-optical decoding spatially multiplexed free-space optical communication channel through optical turbulence

Free-space optical communication (FSOC) faces persistent challenges from optical turbulence and interception. We present an initial simulated study of a spatially multiplexed FSOC architecture that reframes turbulence as a passive encoding mechanism, leveraging its statistical structure to obfuscate transmitted signals without additional optics. Signal recovery is achieved through a compact, all-optical diffractive neural network inspired by the Gerchberg–Saxton algorithm (GS-D2NN), which uses alternating real- and Fourier-domain phase masks to decouple turbulence effects. We demonstrate range-selective decoding across weak-to-moderate turbulence and investigate SNR versus turbulence performance for practical deployment, identifying turbulence-driven off-bit leakage as the primary barrier to closing the gap with conventional FSOC.