Single-crystal PMN-PT gated ferroelectric field-effect transistor for artificial visual application

Abstract

The rapid advancement of artificial intelligence (AI) has fueled a growing demand for neuromorphic devices that integrate high-speed performance with non-volatile storage. Ferroelectric field-effect transistors (FeFETs) have emerged as promising candidates, offering low power consumption, reliable non-destructive readout, and synaptic functionality. In this study, we present a high-performance FeFET incorporating 0.7Pb(Mg1/3Nb2/3)O3–0.3PbTiO3 (PMN-PT) as the dielectric layer and monolayer graphene as the channel material. The combination of PMN-PT's high dielectric constant and strong spontaneous polarization with graphene's exceptional conductivity and carrier mobility delivers superior electrical and optoelectronic properties. Experimental results demonstrate the device's capability to modulate channel conductance, achieve multi-level memory states, and exhibit synaptic plasticity, including long-term potentiation and depression under variable gate voltages and light intensities. Furthermore, the proposed FeFETs can be trained and recognize handwritten digit images from the MNIST dataset with an excellent recognition accuracy of 94.8%. These findings highlight the potential of PMN-PT/graphene-based FeFETs for energy-efficient neuromorphic computing, offering a pathway to next-generation AI hardware.