Full-vdW Heterosynaptic Memtransistor with the Ferroelectric Inserted Functional Layer and its Neuromorphic Applications

The emergence of 2D van der Waals (vdW) materials, owing to highly tunable electrical conductivity, remarkably free stackability, and excellent compatibility for heterojunction integration, has provided abundant research diversity of artificial synapses. Here, an innovative 2D vdW heterosynaptic memtransistor (vdW-HT) synapse is proposed with a ferroelectric CuInP₂S₆ inserted layer. Neuromorphic synaptic weight change of the vdW-HT synapse in this work are modulated by the synergistic effects of interlayer coupling and ferroelectric polarization reversal. It is the first time to evaluate the required initial energy consumption of ferroelectric vdW-HT synapses with matrixed biomimetic characteristics of “Learning–Forgetting–Re-learning–Memorizing.” The required initial energy consumption is only ≈3.06 pJ, which provided supporting evidence to indicate the promising potential of vdW-HT synapses in exploring neuromorphic applications. A vdW-HT computing system with artificial recognition capability for intelligent automobiles is established and demonstrated outstanding recognition abilities for multiple targets in various environments. The highest recognition accuracy for pedestrians is 98%. In addition, the excellent recognition property is integrated with a robotic arm, successfully achieving high-precision grasping behavior and designated position transmission for identified targets. These results provide promising strategies for the integrated development of emerging neuromorphic electronics and industrial applications.