Fast prototyping of memristors for ReRAMs and neuromorphic computing

Area of Science:

• Materials Science
• Electrical Engineering
• Computer Science

Background:

• The increasing demand for energy-efficient computing in artificial intelligence (AI) necessitates advanced memory technologies.
• Memristors are promising for AI due to their information storage, synaptic mimicry, and in-memory computing capabilities, with low power consumption.

Purpose of the Study:

• To present a scalable and cost-effective method for fabricating Ag/MoS₂/Au memristors.
• To demonstrate the potential of these memristors for energy-efficient AI hardware.

Main Methods:

• Utilized roll-to-roll mechanical exfoliation of two-dimensional MoS₂.
• Employed inkjet printing for device fabrication.
• Simulated a fully-connected neural network using a virtual memristor crossbar array.

Main Results:

• Fabricated Ag/MoS₂/Au memristors exhibiting reliable non-volatile resistive switching.
• Observed switching behavior attributed to conductive filament formation/dissolution in MoS₂.
• Achieved high resistance ratios and robust retention times.
• Demonstrated high classification accuracy in neural network simulations even with limited bit-width precision.

Conclusions:

• The developed memristors are suitable for energy-efficient, high-throughput AI hardware.
• The combination of 2D material exfoliation and inkjet printing offers a scalable fabrication approach.
• These memristors show significant potential for next-generation AI computing applications.