A Robust Memristor Based on Epitaxial Vertically Aligned Nanostructured BaTiO₃ -CeO₂ Films on Silicon

Area of Science:

• Materials Science
• Nanotechnology
• Solid-State Physics

Background:

• Growing demand for high-density, non-volatile memory solutions.
• Need for silicon-based ferroelectric devices capable of complex computations.
• Exploration of epitaxial nanostructures for enhanced memory characteristics.

Purpose of the Study:

• To develop a silicon-based memristor utilizing epitaxial ferroelectric nanostructures.
• To investigate the device's potential for multi-value storage and algebraic operations.
• To demonstrate the application of the memristor in neural network simulations.

Main Methods:

• Fabrication of a La$_{0.67}$Sr$_{0.33}$MnO$_{3}$ /SrTiO$_{3}$ /Si substrate with epitaxial BaTiO$_{3}$-CeO$_{2}$ vertically aligned nanostructures.
• Optimization of ferroelectric polarization reversal through controlled growth temperature.
• Testing of resistance characteristics, endurance, synaptic plasticity functions, and algebraic operations using pulse voltages.
• Implementation of a convolutional neural network for CIFAR-10 dataset recognition.

Main Results:

• Achieved a five-state multi-value storage capability with robust endurance up to 10⁹ cycles.
• Successfully simulated synaptic plasticity functions, including spiking-time-dependent plasticity and paired-pulse facilitation.
• Demonstrated realization of basic algebraic operations (addition, subtraction, multiplication, division) using nanosecond pulses.
• Attained high recognition rates (90.03% online, 92.55% offline) in CIFAR-10 dataset identification using a convolutional neural network.

Conclusions:

• The developed silicon-based epitaxial ferroelectric memristor offers a promising platform for next-generation non-volatile memory.
• The device exhibits significant potential for multi-value storage, in-memory computing, and artificial neural network applications.
• This research paves the way for advanced silicon-based computing chips with integrated memory and processing capabilities.