Related Experiment Video
Updated: Jun 28, 2025
A Fabrication and Measurement Method for a Flexible Ferroelectric Element Based on Van Der Waals Heteroepitaxy
Published on: April 8, 2018
Pure ZrO2 Ferroelectric Thin Film for Nonvolatile Memory and Neural Network Computing
Zijian Wang1, Zeyu Guan1, He Wang1
1Hefei National Research Center for Physical Sciences at the Microscale, Department of Physics and CAS Key Laboratory of Strongly-Coupled Quantum Matter Physics, University of Science and Technology of China, Hefei, Anhui 230026, People's Republic of China.
Ferroelectric hafnium oxide (ZrO2) thin films show promise for nonvolatile memory and artificial neural networks. Devices exhibit excellent polarization, low leakage, and enable high accuracy in image recognition simulations.
Area of Science:
- Materials Science
- Solid State Physics
- Device Engineering
Background:
- Ferroelectricity in pure ZrO2 is a recent discovery attracting significant interest.
- The potential of ZrO2-based devices for information storage and processing requires further investigation.
Purpose of the Study:
- To fabricate and characterize a ZrO2-based ferroelectric capacitor.
- To investigate the ferroelectric orthorhombic phase evolution under electric field cycling.
- To evaluate the performance of a ferroelectric field-effect transistor (FeFET) for memory and computing applications.
Main Methods:
- Fabrication of a ~8 nm ZrO2 ferroelectric capacitor with RuO2 electrodes.
- Electrical characterization of ferroelectric properties, including remnant polarization and leakage current.
- Integration of the ferroelectric capacitor with a transistor to form an FeFET.
- Simulation of object image recognition using a convolutional neural network (CNN) based on FeFET conductance manipulation.
Main Results:
- Achieved a ferroelectric remnant polarization (2Pr) of >30 μC/cm2 and low leakage current density (~2.79 × 10-8 A/cm2 at 1 MV/cm).
- Estimated polarization retention exceeding 10 years.
- Demonstrated an FeFET with a memory window of ~0.8 V and eight distinct states.
- Attained high object image recognition accuracy (~93.32%) in CNN simulations, comparable to conventional methods.
Conclusions:
- ZrO2-based ferroelectric devices exhibit promising characteristics for nonvolatile memory applications.
- FeFETs show potential for efficient artificial neural network computing, particularly in image recognition tasks.
- These findings highlight the viability of ferroelectric ZrO2 for next-generation electronic devices.
