Power-Delay Area-Efficient Processing-In-Memory Based on Nanocrystalline Hafnia Ferroelectric Field-Effect...

Area of Science:

• Materials Science
• Electrical Engineering
• Computer Science

Background:

• Ferroelectric field-effect transistors (FeFETs) are crucial for low-power, high-density nonvolatile memory in processing-in-memory (PIM) applications.
• Existing FeFETs suffer from small memory windows and limited endurance, hindering PIM device efficiency and reliability.

Purpose of the Study:

• To overcome the limitations of small memory windows and limited endurance in FeFET-based PIM devices.
• To enhance the area efficiency and reliability of PIM devices through material, device, and array architecture innovations.

Main Methods:

• Demonstrated high ferroelectricity in thick (≥30 nm) ferroelectric materials by considering thermodynamics and kinetics.
• Utilized a metal-ferroelectric-metal-insulator-semiconductor (MFSMI) architecture for improved voltage division.
• Fabricated an FeFET-PIM array to verify multiply-and-accumulation (MAC) operations and energy efficiency.

Main Results:

• Achieved a large memory window (∼11 V), fast operation speed (<20 ns), and high endurance (∼10¹¹ cycles).
• Verified reliable and energy-efficient MAC operations in the fabricated FeFET-PIM array.
• System-level simulations confirmed high energy efficiency attributed to charge-domain computing.

Conclusions:

• The developed FeFET-PIM architecture significantly improves memory window, endurance, and operational speed.
• The FeFET-PIM array demonstrates high accuracy for signed weight MAC computations, suitable for AI tasks like image recognition.