An ultra energy-efficient hardware platform for neuromorphic computing enabled by 2D-TMD tunnel-FETs.

Area of Science:

• Materials Science
• Computer Engineering
• Artificial Intelligence

Background:

• Achieving brain-like energy efficiency in neuromorphic (NM) circuits remains a significant challenge for current hardware platforms.
• Conventional silicon-based technologies, such as 7nm FinFETs, struggle to meet the demands for highly energy-efficient computing.

Purpose of the Study:

• To introduce a novel digital neuromorphic circuit design that significantly enhances energy efficiency.
• To explore the potential of two-dimensional (2D) transition metal dichalcogenide (TMD) materials in neuromorphic computing.

Main Methods:

• Implementation of neuromorphic circuits using 2D transition metal dichalcogenide (TMD) layered channel material-based tunnel-field-effect transistors (TFETs).
• Development of a novel leaky-integrate-fire (LIF) based digital NM circuit incorporating Hebbian learning circuitry.

Main Results:

• The proposed 2D-TFET based NM circuit demonstrates two orders of magnitude higher energy efficiency compared to conventional silicon FinFET technology.
• The circuit operates effectively across a wide range of supply voltages, frequencies, and activity factors.

Conclusions:

• The innovative 2D-TFET based neuromorphic circuit design offers a viable path toward achieving brain-like energy-efficient computing.
• This advancement holds the potential to revolutionize future artificial intelligence (AI) and data analytics platforms.