High-performance organic synaptic transistors for multi-wavelength perception simulation and neuromorphic computing

• The device is prepared based on the simplest solution spin-coating process without the need for complex processes such as doping or mixing structures, which facilitates future exploration of photonic synaptic devices for multifunctional applications, lifting the constraints of material selection and complex fabrication processes. • Thanks to the broad spectral response of PDVT-10 from ultraviolet to near-infrared light, the device realizes photo-synaptic response under different wavelength stimuli. • Based on the good and stable long-term plasticity of PDVT-10 devices, the light-induced high-pass filtering properties and the application of pain sensing, sensitization and memory functions have been realized. • Utilizing the P-type characteristics of the PDVT-10 semiconductor, applying an electrical pulse in combination with a light pulse enables simple logic circuit applications. • A polyvinyl alcohol light-absorbing layer is introduced between the gate dielectric layer and the channel layer, and the photoelectric synergistic effect is utilized to realize weight updating, successfully constructing a convolutional neural network and realizing the MNIST digit recognition task. Inspired by biological neuromorphic systems, biomimetic artificial synaptic devices based on organic transistors have become a prominent research direction. The polymer PDVT-10, which is commonly used as channel layer in devices, has excellent stability and high mobility. However, previous research in simulating photonic synapses has focused on doping and hybrid structures, which is limited by the choice of materials and complex fabrication processes in exploring the multifunctional applications of photonic synaptic devices in the future. Here, we successfully constructed individual PDVT-10 photoelectric synaptic devices to simulate biological synaptic plasticity under different wavelengths of light pulse stimulation for the first time. Furthermore, the application of light-induced high-pass filtering characteristics, pain sensing, sensitization, as well as memory functions were realized. In addition, the application of logic circuits was realized based on the photoelectric synergistic effect. Moreover, the introduction of a polyvinyl alcohol light-absorbing layer endowed the device with light enhancement and electrical inhibition function. Subsequently, a simple convolutional neural network was successfully constructed and implemented for the MNIST number recognition task. This work not only contributes to an in-depth understanding of the response mechanism of the device to different wavelengths of light, but also enriches the application areas of the device and provides important support for the practical applications of neuromorphic computing in the future.