Dielectric-Engineered Monolayer MoS₂ Memtransistors for Brain-Inspired Computing with High Recognition Accuracy

Area of Science:

• Materials Science
• Nanotechnology
• Computer Engineering

Background:

• Two-dimensional transition metal dichalcogenides (2D-TMDs) are promising for neuromorphic hardware due to their synaptic emulation capabilities.
• Existing 2D-TMDs memtransistors often use polycrystalline channels, leading to material integrity issues and performance variability.

Purpose of the Study:

• To develop high-performance memtransistors for neuromorphic applications using monocrystalline 2D-TMDs.
• To address the limitations of polycrystalline channels in current memtransistor designs.

Main Methods:

• Fabrication of a monocrystalline monolayer MoS2 memtransistor on a silicon nitride (SiNx) substrate.
• Characterization of device performance, including resistive switching ratio, dynamic range, and weight update linearity and variability.
• Evaluation of the memtransistor's performance on the MNIST handwritten digits dataset.

Main Results:

• Achieved a large resistive switching ratio (10^4) and dynamic range (>90).
• Demonstrated highly linear and symmetric weight updates with minimal cycle-to-cycle and device-to-device variability.
• Attained over 97% recognition accuracy on the MNIST dataset using the artificial synapses.

Conclusions:

• Monocrystalline monolayer MoS2 memtransistors fabricated via dielectric engineering offer a viable path for high-performance neuromorphic hardware.
• The developed approach overcomes limitations of polycrystalline materials, enabling reliable and efficient artificial synapses.
• This work presents a promising platform for next-generation neuromorphic computing systems.