Fully photonic controlled flexible synapse for bionic machine vision and reconfigurable logic applications

Abstract

Optoelectronic synapses integrating sensing and synaptic functions are promising for neuromorphic computing, particularly in visual information processing. Traditional designs rely on electrical stimulation for bidirectional weight updating, limiting speed, bandwidth, and integration density. This work presents a wafer-scale, flexible silicon-based fully optical synaptic device capable of bidirectional optical response. This improvement facilitates excitatory and inhibitory synaptic behaviors under illumination with 465 nm and 785 nm wavelengths, respectively. The device demonstrates a range of optical synaptic features, including short-term plasticity, long-term plasticity, paired-pulse facilitation, paired-pulse depression, short-term memory (STM), long-term memory, and cognitive processes such as learning, forgetting, and relearning, particularly under 465 nm light stimulus. The system enables real-time image detection, in situ memorization, and processing within a single memory cell, reducing energy overhead and latency from traditional data conversion and transmission. Additionally, the device functions as a nonvolatile, reconfigurable logic gate. By leveraging three distinct wavelengths 465 nm and 532 nm, and 785 nm the system successfully implements logical operations such as “AND”, “OR”, “NAND” and “NOR”. It also integrates associative learning into the optical synaptic framework. This breakthrough marks a key step toward optogenetics-inspired neuromorphic computing, enabling adaptive processing networks and advancing next-generation wearable electronics and efficient computational systems.