Adaptive ferroelectric memristors with high-throughput BaTiO₃ thin films for neuromorphic computing

Area of Science:

• Materials Science
• Nanotechnology
• Solid State Physics

Background:

• Ferroelectric tunnel junctions (FTJs) and ferroelectric diodes (FDs) are promising for neuromorphic computing.
• Current limitations include strong thickness dependence and laborious optimization of ferroelectric layers.

Purpose of the Study:

• To develop high-performance electronic synapses using a high-throughput ferroelectric BaTiO3 (BTO) thin film.
• To integrate multiple resistive switching behaviors within a single device for enhanced functionality.

Main Methods:

• Fabrication of two-terminal ferroelectric memristors on a thickness-gradient BTO film (1-30 unit cells).
• Investigation of ferroelectricity and resistive switching behaviors across varying BTO thicknesses.
• Development of an on-chip integration configuration for adaptive conductance compensation.

Main Results:

• Intrinsic ferroelectricity observed in BTO films thicker than 5 unit cells.
• Sequential occurrence of resistor, FTJ, and FD behaviors with increasing BTO thickness.
• Improved handwritten digit recognition accuracy from 91.3% to 95.7% using FTJ synapses with adaptive compensation.
• Demonstrated fault tolerance against Gaussian noise interference.

Conclusions:

• The developed approach enables the integration of resistor, FTJ, and FD components for advanced electronic synapses.
• High-performance and fault-tolerant neuromorphic computing systems can be realized.
• The technology is suitable for memory, logic processing, and neuromorphic computing applications.