Multifunctional computing-in-memory SRAM cells based on two-surface-channel MoS₂ transistors.

Area of Science:

• Materials Science
• Computer Engineering
• Electrical Engineering

Background:

• The von Neumann architecture faces significant energy efficiency and throughput limitations due to advancements in machine learning, artificial intelligence, and the Internet of Things.
• The development of novel computer architectures is crucial to overcome these challenges.

Purpose of the Study:

• To propose and demonstrate two novel computing-in-memory (CIM) designs based on two-surface-channel MoS2 transistors.
• To address the limitations of traditional computer architectures by integrating memory and processing units.

Main Methods:

• Design and fabrication of a symmetrical 4T2R Static Random-Access Memory (SRAM) cell.
• Design and fabrication of a skewed 3T3R SRAM cell, both utilizing two-surface-channel MoS2 transistors.
• Experimental verification of in-memory logical computations.

Main Results:

• The symmetrical SRAM cell successfully performed in-memory XNOR and XOR computations.
• The skewed SRAM cell achieved in-memory NAND and NOR computations.
• Both designs demonstrated high area efficiency due to the use of fewer transistors.

Conclusions:

• The proposed MoS2-based CIM SRAM cells offer a promising approach to alleviate the von Neumann bottleneck.
• These designs pave the way for highly area-efficient and multifunctional computing chips with integrated memory and processing capabilities.