T. T. Zeng, S. E. Liu, R. Wu, V. K. Sangwan, M. C. Hersam
Northwestern University,
United States
Keywords: two-dimensional, memtransistor, low-voltage, back-gated, neuromorphic computing
Summary:
The rapid ascent of artificial intelligence in the digital era imposes unprecedented energy demands for supporting electronics hardware. Neuromorphic computing aims to revolutionize data processing and communication with brain-inspired paradigms of parallel processing, in-memory computing, and classification of analog signals at the edge. For neuromorphic hardware, two-dimensional memtransistors have shown to embody bio-realistic functionalities. However, memtransistors to date have not achieved sub-1 V operating voltage without compromising the gate-tunability. Here we report short-channel length (100-370 nm), low-voltage MoS2 memtransistors with field-effect tuning of low and high resistance states over four orders of magnitude. This is achieved by integrating a back-gate design with channel length and width (20 μm) that are smaller and larger respectively compared to average grain size (≈ 1 μm) of monolayer MoS2 grown by chemical vapor deposition. Finite-element simulations confirm enhanced electrostatic effects in the back-gated design compared to incumbent top-gated memtransistors that are critical for voltage reduction without compromising resistive switching and field-effect ON/OFF ratios.