Depolarization Field Controllable HfZrO-Based Ferroelectric Capacitors for Physical Reservoir Computing System

Area of Science:

• Materials Science
• Artificial Intelligence
• Computer Engineering

Background:

• Reservoir computing, a type of artificial neural network, excels at modeling time-varying data due to its energy efficiency and minimal training requirements.
• Ferroelectric memory devices, particularly those utilizing hafnium zirconium oxide (HZO) films, are being explored for advanced computing applications.

Purpose of the Study:

• To investigate the impact of HZO film thickness and interlayer thickness on the decay time of ferroelectric memory devices for reservoir computing.
• To optimize HZO film properties for enhanced performance in physical reservoir computing systems.

Main Methods:

• Fabrication and characterization of HZO-based ferroelectric memory devices with varying HZO film thicknesses (10, 15, and 20 nm).
• Analysis of ferroelectric properties, including orthorhombic phase formation and short-term memory characteristics.
• Integration of optimized HZO devices into a reservoir computing system for performance evaluation.

Main Results:

• The 15 nm HZO film thickness demonstrated optimal ferroelectric properties and reliable short-term memory.
• Optimized HZO devices exhibited enhanced orthorhombic phase formation.
• The reservoir computing system utilizing the optimized HZO device achieved an average accuracy of 93.42% in image recognition tasks.

Conclusions:

• Hafnium zirconium oxide (HZO) films offer significant advantages for reservoir computing applications by enabling control over the depolarization field.
• Optimized HZO film thickness is crucial for achieving excellent ferroelectric properties and reliable short-term memory.
• HZO-based ferroelectric memory devices are capable of high-precision computations, as evidenced by their performance in image recognition tasks.