Ultralow Energy Optoelectronic Synapse Using Halide Perovskite/Organic Semiconductor Heterostructure for Neuromorphic...

With the advancement of artificial intelligence, the emulation of biological neural processes through neuromorphic computing has gained significant attention. Artificial optoelectronic synapses have emerged as promising components for neuromorphic computing due to their simple structure, low energy consumption, and ability to overcome the von Neumann bottleneck. Here, we design a multifunctional, energy-efficient optoelectronic synapse based on a formamidinium cesium lead iodide (FA_xCs_1-xPbI₃)/Poly(3-hexylthiophene) (P3HT) heterojunction in a two-terminal vertical structure. The synaptic device exhibits key synaptic characteristics, such as excitatory postsynaptic current (EPSC), paired-pulse facilitation (PPF), and achieves a transition from short-term to long-term memory with an exceptionally low energy consumption of 0.59 fJ per synaptic event, and successfully emulates biological learning behavior, such as learning-forgetting-relearning. Long-term potentiation (LTP) enables efficient visual object recognition with 90.31% accuracy on the Modified National Institute of Standards and Technology (MNIST) data set using an artificial neural network (ANN). In addition, light logic functions ("AND", "OR") and associative learning (Pavlov's dog experiment) are demonstrated using 405 and 532 nm pulses. More significantly, optical wireless communication is experimentally performed using Morse code for words such as IITG, 2025, HELP, and SOS. Moreover, the device achieves 86.76% pixel-wise accuracy in the semantic segmentation of urban street scenes using a U-Net model. Finally, the working mechanism of the device, attributed to the efficient photogeneration of carriers and accumulation of electrons at the perovskite side, offers deep insights into the optoelectronic plasticity. These findings show the path toward the development of a highly integrated, photonic neuromorphic device for future intelligent systems.