High-Performance Electrode-Controlled MoTe2 Memristor for Dual-Mode Synaptic Plasticity in Neuromorphic Systems

Abstract

This paper introduces a molybdenum ditelluride (MoTe₂)-based memtransistor featuring an innovative bottom-embedded electrode architecture, which mitigates the persistent Fermi-level pinning effect in conventional two-dimensional (2D) material devices. The embedded electrode design significantly enhances synaptic emulation performance and device stability by optimizing interfacial contact and enabling precise carrier transport modulation. This breakthrough allows for high-accuracy, multi-level conductance tuning, supporting adsorption-controlled memory characteristics and energy-efficient operation. Through coordinated pulse modulation of drain and gate terminals, the device successfully mimics biological synaptic functionalities, including homosynaptic and heterosynaptic plasticity, and faithfully replicates dynamic behaviors such as long-term potentiation (LTP) and long-term depression (LTD). These capabilities establish a robust hardware foundation for constructing complex neuromorphic systems. Furthermore, an artificial neural network (ANN) leveraging MoTe 2 heterosynaptic plasticity achieves an accuracy of 90.32% on the MNIST handwritten digit recognition task, demonstrating its potential for parallel information processing and in-memory computing architectures. This work not only provides a novel design paradigm for 2D material-based synaptic devices but also highlights the transformative potential of embedded electrode structures in advancing low-power neuromorphic computing and high-density brain-inspired integrated circuits.