Subquantum Semimetal Bi and Oxygen Vacancy Filament Memristors for Neuromorphic Computing

Area of Science:

• Materials Science and Engineering
• Nanotechnology
• Computer Engineering

Background:

• Memristors are promising for non-von Neumann architectures due to scalability.
• Current filamentary memristors suffer from high variability and poor reliability caused by random filament formation.

Purpose of the Study:

• To propose and investigate a novel semimetal bismuth (Bi)-based memristor.
• To elucidate the switching mechanism of oxygen vacancy (V_O)-Bi filaments.
• To evaluate the performance of the proposed memristor in artificial neural network (ANN) simulations.

Main Methods:

• Fabrication and characterization of a Bi-based memristor.
• Spherical aberration-corrected scanning transmission electron microscopy (AC-STEM) for filament analysis.
• Density functional theory (DFT) calculations for mechanism elucidation.
• ANN simulations using back-propagation and reservoir computing (RC).

Main Results:

• The Bi-based memristor demonstrated a subquantum conductance change, high switching consistency, and controllable weight update linearity.
• AC-STEM and DFT revealed the formation mechanism of V_O and Bi clusters, clarifying the switching process.
• ANN simulations achieved high digit recognition accuracies of 95.77% (back-propagation) and 94.15% (RC).

Conclusions:

• The V_O-Bi filament mechanism in the Bi-based memristor offers a pathway to overcome the reliability issues of conventional memristors.
• The proposed memristor shows significant potential for advanced computing applications, including neuromorphic systems.