Harnessing Atomic Vacancy at Semiconductor Heterojunction for Programmable Photonic Memory and Reservoir Computing

ABSTRACT Artificial photonic synapses (APSs) are emerging as promising components for next‐generation neuromorphic computing due to their low crosstalk and high bandwidth. However, existing APSs either exhibit short‐lived photoconductivity (volatile), suitable for transient signal processing but inadequate for memory, or non‐volatile states that lack the dynamic decay needed for adaptive learning. In this study, we break this dilemma by introducing a semi‐volatile APS based on a sulfur vacancy (Sv)‐engineered ZnIn 2 S 4 /MoS 2 heterojunction, which exhibits dynamically adjustable multi‐photoconductivity states. At the heterojunction, a charge transfer interface (CTI) dynamically orchestrates the routing of photoexcited charges into long‐lived Sv traps. This optically gated process unlocks programmable memory times spanning from 10 to 10 3 s, enabling a seamless transition from short‐term to long‐term synaptic plasticity, filling the gap between volatile and non‐volatile memory. Leveraging the distinguishable multi‐photoconductivity states, we demonstrate a high‐security information encryption system and a dynamic cryptographic authentication protocol. Furthermore, the device serves as a reservoir computing unit, achieving efficient handwritten digit recognition. This work not only presents an advanced synaptic device but also establishes a new paradigm of interface engineering for multifunctional neuromorphic computing and secure photonic communications.