Event-Driven CsPbBr₃ Perovskite Visual Neurons for Motion Direction Recognition in Low-Light Scenes

The event-driven visual neuron system can enhance the efficiency of visual data processing at sensor terminals, due to its capability for sensing and encoding information with event-driven characteristics for neuromorphic computing. Here, we demonstrate an event-driven CsPbBr<sub xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"><roman>3</roman></sub> perovskite visual neuron system consisting solely of a single CsPbBr<sub xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"><roman>3</roman></sub> perovskite photodiode and simple circuit overhead. Compared with conventional event-based cameras with Si photodiode, this system enables energy-efficient event-driven light intensity encoding with a greatly simplified sensor circuit design. Our visual neurons utilize self-powered CsPbBr<sub xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"><roman>3</roman></sub> perovskite photodiode, which inherently features a high dynamic range of 160 dB. The CsPbBr<sub xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"><roman>3</roman></sub> visual neurons can detect dynamic changes in dim visual scenes with the light intensity as low as 0.89 μW/cm<sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"><roman>2</roman></sup>. Besides, they achieve the fastest event respond with a temporal resolution of 1.4 μs. Finally, CsPbBr<sub xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"><roman>3</roman></sub> perovskite visual neurons are incorporated as input neurons in a simulated three-layer spiking neural network (SNN), achieving 99% accuracy in motion direction recognition tasks.