Unveiling the Nanoscale Mechanism(imaging) of 2D Nanomaterial-Based Memristive Devices

A high-performance memristive device is a key component in neuromorphic computing. Conventional memristors based on metal oxides such as TiO_x can only mimic a single characteristic of neuromorphic layers and thus, it does not allow for the multi-terminal state-switching control which enables to simulate diverse neuron-synapse signal transfer processes more realistically. Recently, two-dimensional (2D) nanomaterials such as layered molybdenum disulfides (MoS₂) have drawn a great attention in simulating a dynamic process of neuromorphic computing with their superior and unique properties. Here, we demonstrate two 2D nanomaterial systems for developing memristive devices with various performances: h-BN/graphene/h-BN heterostructure system and MoS₂ system with a nano-gap. First, van der Waals h-BN/graphene/h-BN with interfacial charged impurities enabled the controllable carrier charge trap and release resulting in the gate bias-dependent hysteresis of electrical transfer curves. The trapped charge density of h-BN/graphene/h-BN field-effect transistors can be easily modulated by the magnitude and the polarity of the applied gate bias. The ‘set’ process (trap) can be achieved with applying a negative gate bias while the ‘reset’ process (release) can be done by applying a positive gate bias, showing the retention performance with the stable ratio of the currents for the set to the reset, and both the states were maintained more than few days. Secondly, to develop a high-performance memristive device, we demonstrate a MoS₂ nano-gap structure for mimicking actual human brain synapse structures. Completely separated MoS₂ flakes maintain high resistance state (HRS) but current tunneling occurred through the nano-gap by the field-emission of ions above a certain gate voltage. Continuous application of drain-source voltage formed the ion bridge at the nano-gap, as revealed by electrostatic force microscopy (EFM), and the bridge electrically connected each flake, significantly increasing the drain-source current, which can be a ‘set’ (low-resistance state, LRS) process. This study provides a fundamental understanding for how 2D nanomaterials can be utilized to design and develop various performance and characteristics of memristive devices for more realistically mimicking neuron-synapse signal transfer mechanisms.