Diffractive interconnects: all-optical permutation operation using diffractive networks

Area of Science:

• Optics and Photonics
• Deep Learning
• Computational Science

Background:

• Permutation matrices are crucial for communications, information security, and data processing.
• Optical implementation of permutation operators requires power-efficient, fast, and compact platforms for large interconnections.

Purpose of the Study:

• To develop diffractive optical networks capable of all-optical permutation operations.
• To enable scalable interconnections using passive transmissive layers structured at the wavelength scale.

Main Methods:

• Engineered diffractive optical networks using deep learning.
• Designed wavelength-scale structured passive transmissive layers.
• Developed misalignment-tolerant diffractive designs.
• Experimentally demonstrated operation at Terahertz (THz) frequencies.

Main Results:

• Diffractive optical networks can perform permutation operations with hundreds of thousands of interconnections.
• Network capacity scales with the number of diffractive layers and trainable elements.
• Demonstrated misalignment-tolerant designs for arbitrary permutation operations.
• First experimental demonstration of a diffractive permutation network operating at THz spectrum.

Conclusions:

• Diffractive permutation networks offer a scalable and efficient method for all-optical permutation operations.
• These networks have potential applications in security, image encryption, data processing, and telecommunications, particularly in THz wireless networks.
• The developed designs address practical challenges like alignment and diffraction efficiency.