Self-powered optoelectronic synaptic device for both static and dynamic reservoir computing

Self-powered optoelectronic reservoir computing offers a powerful solution for efficient computation while significantly reducing power consumption, making it an ideal candidate for advanced real-time applications in artificial intelligence and big data analytics. Therefore, exploring the broader application of optoelectronic synaptic devices in optical reservoir computing is of great importance. In this work, we present a self-powered optoelectronic synaptic device designed for optoelectronic reservoir computing. To highlight the device's self-powered capability and nonlinear characteristics, the device features a p-n junction composed of n-type InGaZnO and p-type NiO, enabling it to exhibit synaptic-like properties such as excitatory postsynaptic current, paired-pulse facilitation, and short-term plasticity in response to light pulse stimulation under self-powered conditions. By innovating the data preprocessing methods and inputting signal patterns for optoelectronic reservoir computing, both static and dynamic reservoir computing were successfully realized. In static reservoir computing, the device achieved an image classification accuracy of 95.2 %, while in dynamic reservoir computing, it attained 98.2 % accuracy in spoken signal recognition and 93.5 % in electromyographic signal classification. These results demonstrate the potential of our self-powered optoelectronic synaptic device for broader applications, offering a significant advancement in real-time data processing and adaptive learning systems.