Reservoir computing (RC) system based on photoelectronic memristors has attracted increasing attention because of its inherently nonlinearity, history-dependent dynamics, and fading memory for real-time processing complex task. However, relying on its natural relaxation process leads to the RC system with low efficiency. Here, we propose an Au/CuI/TiOx/FTO heterojunction optoelectronic synergistic memristor and demonstrate its application as a single-node physical reservoir computing (PRC) system. The device synergistically integrates oxygen-vacancy migration in TiOx with the photosensitivity of CuI, enabling precise. This approach generates rich, high-dimensional reservoir states with enhanced separability. The memristor exhibits diverse synaptic plasticity and supports over 32 distinct conductance states (>5-bits precision). Furthermore, a predictive model was established to capture device dynamics, achieving an accuracy of 82.8% for the standard RC model, while the accuracy of the heterojunction memristor-assisted RC model increased to 97.26%. The research demonstrates that heterojunction optoelectronic collaborative memristors exhibit significant potential in efficient temporal encoding and high-precision single-node PRC, providing a promising solution for edge intelligence and low-power neuromorphic hardware.
Fan et al. (Sun,) studied this question.