Answers are grounded in this paper and cite their sources.
Emulating biological sensory processing in hardware requires synaptic devices with tunable dynamics and multimodal integration capability. However, heterogeneous physical signals are difficult to encode uniformly in a single synaptic device. Herein, we construct an optically programmable photonic synapse (CZH‐PS) based on Cs 2 AgBiBr 6 /ZnO x (defect‐type) heterostructure, where abundant oxygen vacancy (Ov) deep traps in ZnO x dominate the synaptic plasticity. Photoexcited electrons injected from the perovskite into ZnO x are stored in Ov trap states and thermally released, producing a persistent photocurrent whose decay kinetics are optically tunable. This mechanism underpins a rich repertoire of synaptic behaviors, including short‐term/long‐term memory (STM/LTM), Ebbinghaus forgetting, and the habituation/sensitization of the Aplysia gill‐withdrawal reflex under dual‐wavelength programming. Moreover, the nonlinear, history‐dependent excitatory postsynaptic current (EPSC) serves as a physical reservoir for temporal information processing. With 4‐bit optical pulse coding, a reservoir computing system is constructed that achieves recognition accuracies of 91.7% and 95.0% for letter and audio classification tasks, respectively. These results establish trap‐mediated photonic synapses as a versatile hardware platform for bimodal bio‐inspired computing.
No takes yet. Share an insight, caveat, or question.
Gao et al. (2026) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: