All-Optically Regulated ITO/Cu2O/WO3/ITO Memristor for Optoelectronic Neuromorphic Computing

Abstract

Optoelectronic memristors are becoming increasingly attractive compared to traditional electrical memristors because they can effectively integrate the benefits of both photonics and electronics. However, the inability to reversibly tune the conductance with light has significantly hindered the development of optoelectronic neuromorphic computing. In this work, all-optically regulated resistive switching characteristics are demonstrated in the CMOS process-compatible ITO/Cu2O/WO3/ITO optoelectronic memristor. The device exhibits an average transmittance of 70.14% under visible light. After electroforming, it achieves stable bipolar analogue switching, data retention beyond 107 s, and endurance of 106 cycles. An obvious increase in current is noticeable under 405 or 532 nm wavelength light irradiation, and the current decreases under 633 nm wavelength light irradiation. The light-tuned range is in μA. The synaptic plasticity learning behavior can be emulated in this memristor by electrical or optical stimulation. The learning, forgetting, erasure, and recovery processes of artificial intelligence are successfully implemented. Furthermore, based on the optical or electrical regulation of the conductances, the neural network simulation for supervising learning presents an online learning pattern recognition, and the accuracies can be achieved at 89.96 and 93.03% for optical and electric spikes, respectively. These results suggest that this memristor has high potential for optoelectronic neuromorphic computing applications.