Leveraging Dual Resistive Switching in Quasi-2D Perovskite Memristors for Integrated Non-volatile Memory, Synaptic...

Area of Science:

• Materials Science
• Computer Engineering
• Neuroscience

Background:

• Artificial intelligence (AI) computational demands challenge conventional computing.
• Memristors with diverse resistive switching (RS) offer novel computing architectures.
• Need for advanced materials for next-generation computing.

Purpose of the Study:

• Investigate coexistence of volatile and nonvolatile RS in quasi-2D perovskite memristors (Q-2DPM).
• Evaluate Q-2DPM performance for nonvolatile memory and synaptic functions.
• Demonstrate Q-2DPM suitability for neuromorphic and reservoir computing.

Main Methods:

• Fabrication and characterization of quasi-2D perovskite memristors.
• Simulation of synaptic functions (ePSC, PPF, LTP/LTD).
• Implementation in artificial neural networks for MNIST classification and reservoir computing experiments.

Main Results:

• Achieved coexistence of volatile and nonvolatile RS behaviors in Q-2DPM.
• Q-2DPM demonstrated competitive nonvolatile memory performance.
• Successful simulation of multiple synaptic functions and high accuracy in MNIST classification.
• Demonstrated Q-2DPM suitability for reservoir computing via digital classification.

Conclusions:

• Elucidation of dual RS mechanisms in Q-2DPM provides new insights.
• Single Q-2DPM devices can achieve diverse computational units.
• Paves the way for physical neuromorphic hardware and advanced computational primitives.