Perspective on the potential of 2D materials in programmable diffractive optical networks

Abstract The rapid development of artificial intelligence (AI) has highlighted the potential of neuromorphic computing based on photonic architectures, which offer high bandwidth, low latency, and energy efficiency. Diffractive optical networks (DONs), particularly three-dimensional (3D) DONs, have demonstrated exceptional capabilities in parallel processing optical information at light speed. However, current DONs are typically fixed post-manufacture and operate on the centimetre scale, presenting challenges in terms of dynamic tunability and miniaturization. In this perspective, we explore the potential of using addressable two-dimensional (2D) materials as a platform for creating dynamically tunable, compact DONs. This approach could pave the way for the next generation of on-chip reconfigurable photonic chips. We also discuss the future directions and challenges in this field.