Molecular crystal memristors

Area of Science:

• Materials Science
• Nanotechnology
• Computer Engineering

Background:

• Current memristors face challenges with material degradation, leading to high energy use and limited endurance.
• Resistive switching in memristors is crucial for in-memory computing, but material stability remains a bottleneck.

Purpose of the Study:

• To develop a stable and energy-efficient memristor for in-memory computing applications.
• To explore the potential of molecular crystal structures in memristor channel materials.

Main Methods:

• Fabrication of a memristor utilizing a molecular crystal material, antimony trioxide (Sb2O3).
• Investigation of ion migration through the van der Waals interconnected molecular cages.
• Characterization of resistive switching behavior, endurance, and energy consumption.
• Demonstration of device scalability and implementation in reservoir computing.

Main Results:

• The molecular crystal memristor exhibits low energy consumption (26 zJ/operation) and high endurance (>10^9 cycles).
• The device shows reconfigurable volatile and non-volatile switching across various scales (micrometers to nanometers).
• Successful fabrication of large crossbar arrays on an 8-inch wafer and implementation of reservoir computing with 100% accuracy in dynamic vision recognition.

Conclusions:

• Molecular crystal memristors offer a promising solution to overcome material degradation issues in current devices.
• This technology enables efficient, scalable in-memory computing and advanced AI applications like dynamic vision recognition.