InSe Ferroelectric Field-Effect Transistor for Simultaneous Perception–Memory–Computation All-in-One Broadband Motion Detection

Abstract

With the ardent use of artificially intelligent robots, broadband motion detection and recognition (BMDR) technology with simultaneous perception–memory–computation functions is becoming essentially important. However, the existing state-of-the-art all-in-one BMDR remains insufficient. Herein, we demonstrate an InSe ferroelectric field-effect transistor exhibiting broadband responsivity from 450 to 973 nm, attributed to the moderate band gap of approximately 1.3 eV in InSe. The InSe device displays memory functionality derived from the ferroelectric properties of Pb(Zr0.2Ti0.8)O3 (PZT) and computational capability enabled by its progressive nonvolatile modulation of conductance, which facilitates artificial synapse-weight updating. Hence, the InSe ferroelectric field-effect transistor has a broadband perception–memory–computation functionality. Importantly, by integrating bias-based nonvolatile positive-direction/negative-direction photoconductive mechanisms with interframe differencing computations, the proposed InSe ferroelectric field-effect transistor achieves efficient motion detection and recognition (MDR) of a moving rabbit. The detected moving-rabbit images are subsequently fed into a neural network built on conductance mapping principles for classification, achieving 100% accuracy after only 40 training epochs, outperforming comparable results. The proposed work enables the development of the next-generation all-in-one BMDR technology based on two-dimensional devices.