Deep photonic network platform enabling arbitrary and broadband optical functionality.

Area of Science:

• Photonics and Optical Engineering
• Integrated Optics
• Machine Learning Applications in Science

Background:

• Growing demand for high-performance integrated photonic components in optical communications, computing, and sensing.
• Limitations of traditional design methods and current machine learning approaches due to computational demands and scalability issues.
• Need for advanced design tools that offer flexibility and accuracy for complex on-chip optical systems.

Purpose of the Study:

• To present a novel, highly-scalable, physics-informed design platform for arbitrary on-chip optical systems.
• To demonstrate the platform's capability in designing advanced photonic devices rapidly and efficiently.
• To enable systematic, large-scale design of photonic systems with tailored functionalities.

Main Methods:

• Development of a physics-informed design platform utilizing deep photonic networks.
• Custom design of Mach-Zehnder interferometers as the core building blocks.
• Leveraging the platform for rapid design of specific optical components like power splitters and spectral duplexers.

Main Results:

• Demonstration of ultra-broadband power splitters and a spectral duplexer designed in under two minutes.
• Achieved state-of-the-art experimental performance with insertion losses below 0.66 dB.
• Exhibited 1-dB bandwidths exceeding 120 nm for the designed devices.

Conclusions:

• The presented platform offers a scalable and efficient solution for designing complex integrated photonic devices.
• Enables custom control over power, phase, and dispersion profiles for diverse applications.
• Provides a tractable pathway for advancing high-throughput communications, quantum information processing, and sensing technologies.