Two-dimensional materials, especially transition metal dichalcogenides like MoS2, show great promise for neuromorphic computing due to their highly tunable electronic properties. However, the role of defects such as sulfur vacancies (VS), oxygen vacancies (VO), and other in-plane defects in device operation mechanisms remains poorly understood, limiting performance and integration potential. In this work, we leverage a competitive-synergistic model of these defects to modulate resistive switching modes in MoS2-based neuromorphic devices, transitioning from filamentary non-volatile to interface-dominated volatile and hybrid regimes. Our interface-type devices demonstrate exceptional endurance (>105 pulses) and large synaptic weight modulation range up to >103, while hybrid-type devices achieve stable multilevel states with a low variation of each state down to 10-4). The transition of switching mechanism is attributed to the relative concentration of defects in MoS2 through defect engineering protocol, which changes the dominant type of vacancy for migration. More importantly, we integrate volatile and non-volatile MoS2 devices into a spike-response model circuit that emulates different stimulation modes and more advanced neuronal functions. This work establishes a defect engineering strategy for designing both high-performance neuromorphic and memory devices, enabling homogeneous integration for complex neural network applications.
Pan et al. (Sun,) studied this question.