Ferroelectric Hf_0.5Zr_0.5O₂ with Enhanced Intermediate Polarization: A Platform for Neuromorphic and Logic-in-Memory...

Area of Science:

• Materials Science
• Solid State Physics
• Computer Engineering

Background:

• Ferroelectric materials offer nonvolatile, reversible polarization, crucial for advanced computing.
• Their synaptic emulation capabilities are key for neuromorphic networks and logic-in-memory (LiM).
• Exploiting ferroelectric switching for weight modulation is an underexplored area.

Purpose of the Study:

• To improve intermediate and cumulative polarization levels in Hf0.5Zr0.5O2 (HZO) via phase engineering.
• To demonstrate HZO-based ferroelectric field-effect transistors (FeFETs) for synaptic weight modulation.
• To integrate neuromorphic and LiM functionalities on a unified platform.

Main Methods:

• Phase engineering of Hf0.5Zr0.5O2 (HZO) to enhance polarization characteristics.
• Fabrication and characterization of HZO-based synaptic FeFETs.
• Implementation of FeFETs for neuromorphic (MNIST, Fashion-MNIST) and LiM (NOR, NAND logic) tasks.

Main Results:

• Achieved improved intermediate and cumulative polarization levels in engineered HZO.
• Demonstrated HZO FeFETs with a wide range of synaptic weights (up to 8 bits) and high linearity.
• Attained high classification accuracies: 98% for MNIST and 88% for Fashion-MNIST.
• Showcased reconfigurable in-memory NOR/NAND logic and 3-bit logic state generation using multigate FeFETs.

Conclusions:

• Engineered HZO exhibits enhanced polarization for efficient synaptic weight modulation.
• HZO-based FeFETs successfully enable high-accuracy neuromorphic computing and versatile LiM operations.
• This work presents a unified platform for integrating learning and logic, paving the way for next-generation computing systems.