Energy-Efficient Integrated Electro-Optic Memristors

Area of Science:

• Materials Science
• Photonics
• Computer Engineering

Background:

• Neuromorphic photonic processors leverage memristors for high-speed, in-memory information processing.
• Electro-optic memristors integrate electronic efficiency with photonic bandwidth but lack scalable, CMOS-compatible designs.
• Existing technologies face challenges in achieving efficient and integrated electro-optic functionalities.

Purpose of the Study:

• To develop efficient, scalable, and CMOS-compatible electro-optic memristors.
• To enable dual electrical and optical programmability and readability in a single device.
• To advance high-performance, energy-efficient integrated electro-optic neuromorphic computing.

Main Methods:

• Structuring phase-change material into a nanoscale constriction.
• Geometrically confining electrical heat profiles to overlap with optical fields.
• Demonstrating electrical switching and electro-optical modulation.

Main Results:

• Achieved sub-10 pJ electrical switching energy.
• Demonstrated a high electro-optical modulation efficiency of 0.15 nJ/dB.
• Successfully integrated electrical and optical programmability and readability.

Conclusions:

• The developed electro-optic memristors offer a promising solution for integrated neuromorphic computing.
• The nanoscale constriction approach enables efficient dual-domain functionality.
• This work paves the way for high-performance, energy-efficient photonic processors.