Reconfigurable Ferroelectric Field-Effect Transistor Integrating Freestanding BaTiO3 and MoS2 for Neuromorphic Computing.

In the era of big data and artificial intelligence, the demand for computing capacity is growing exponentially, driving the need for transformative computing technologies. Neuromorphic computing, which adopts a non-von Neumann architecture, has emerged as a promising solution to overcome the limitations of conventional systems. As fundamental building blocks for neuromorphic hardware, artificial synapses are of great importance. Here, we report a reconfigurable ferroelectric field-effect transistor (FeFET) with a metal-ferroelectric-metal-insulator-semiconductor (MFMIS) structure that serves as a three-terminal artificial synapse, integrating memory and computing functionalities. By modulating the input pulse width, this FeFET-fabricated with a ferroelectric perovskite BaTiO3 and a 2D MoS2 channel-can be configured as either a non-volatile memory, exhibiting a counterclockwise hysteresis window larger than 3.8 V, or a volatile synaptic device with low power consumption of 1.22 fJ per synaptic event. Furthermore, by leveraging both non-volatile memory and volatile synaptic modes, we demonstrate the device's applications in convolutional neural network (CNN)-based traffic sign recognition and classification for autonomous driving, as well as motion direction decision-making via multi-level programming. These results provide a viable strategy for the development of neuromorphic devices in next-generation computing.