2D MoS₂-based reconfigurable analog hardware

Area of Science:

• Materials Science
• Neuroscience
• Computer Engineering

Background:

• Biological neural circuits exhibit remarkable adaptability through dynamic connection adjustments.
• Current two-dimensional (2D) materials-based neuromorphic hardware often focuses on mimicking individual neural components (synapse, soma).
• Integrating multiple 2D material devices for brain-like functions represents a significant research trend.

Purpose of the Study:

• To demonstrate a reconfigurable 2D MoS2-based analog hardware.
• To emulate synaptic, heterosynaptic, and somatic functionalities within a single platform.
• To showcase the hardware's potential for versatile, brain-inspired computing.

Main Methods:

• Fabrication of a 2D MoS2-based analog hardware.
• Integration of modules to emulate synaptic, heterosynaptic, and somatic functions.
• Co-encoding of inner states and inter-connections for diverse computational tasks.

Main Results:

• The hardware successfully emulates synaptic, heterosynaptic, and somatic functionalities.
• Versatile functions including analog-to-digital conversion, linear/nonlinear computations (integration, vector-matrix multiplication, convolution) were achieved.
• Experimental demonstrations include medical image reconstruction/sharpening and imitation of attention-switching/visual residual mechanisms.

Conclusions:

• The developed MoS2-based hardware offers high adaptability and flexibility for multiple tasks.
• This innovation advances the development of general-purpose computing machines with brain-like capabilities.
• The integrated approach enables complex functions for smart perception and medical diagnostics.