Optoelectronic Memristor Based on SrTiO 3 /HfO x Heterostructure for Neuromorphic Visual Processing

This study presents a transparent optoelectronic memristive device utilizing a heterostructure, ITO/STO (5 nm)/ HfOx(10 nm)/ITO, for neuromorphic computing applications. The device operated at low voltages (set: 1.3 V; reset: -1.9 V), demonstrates good endurance exceeding 104cycles, stable retention beyond 104s, consistent ON/OFF ratio of approximately 10. Under 405 nm violet light illumination, it exhibits positive photoconductivity and emulates synaptic functions, including long-term potentiation and depression, excitatory postsynaptic current, paired-pulse facilitation, and short-term to long-term memory transitions. It shows near-ideal nonlinearity values of 1.73 and 0.14 for potentiation and depression, respectively, enabling precise synaptic weight modulation. Applying a convolutional neural network model simulated for facial recognition using the ORL dataset, the device achieves over 90% accuracy under low noise conditions. It indicates robust performance against Gaussian and salt-and-pepper noises. According to the results of material characterizations via TEM, EDS, and XPS, it proposes an oxygen vacancy-driven filamentary switching mechanism for the device. In addition, under 405 nm violet light illumination, this optoelectronic memristor with the synaptic properties offers high potential for energy-efficient neuromorphic visual processing applications.