All-optically controlled terahertz memristor for multidimensional neuromorphic computing

Optically controlled memristors are considered as promising components for optical computing and storage functions, which can revolutionize operational efficiency and information transmission security in the next generation of artificial intelligence (AI) systems. Here, we present a neuromorphic all-optically controlled terahertz memristor (AOCTM) based on vanadium dioxide nanowires/polydimethylsiloxane (VO 2 NWs/PDMS), which possesses all-optical storage and processing for multi-dimensional optical computing terahertz (THz) systems. This memristor with near-infrared (NIR) optical synaptic plasticity, short-term plasticity (STP), and long-term plasticity (LTP) can mimic artificial synapses, arising from thermal hysteresis and photothermal effect. An imaging encryption method using AOCTM is proposed to enhance data security for all-optical communication links. Moreover, when integrated with the array AOCTM and artificial neural networks, the all-optical computational strategy for image recognition can significantly upgrade the recognition rate to 93.5% under 57% Gaussian noise. Our work provides a master direction for advancing multi-dimensional optical AI computing.