Combinatorial optimization with Kerr solitons.

Area of Science:

• * Physics and Applied Sciences
• * Photonics and Optical Engineering
• * Computer Science and Engineering

Background:

• * Digital computing faces scaling challenges, driving innovation in physical systems mimicking neural networks and optimization problems.
• * Light, as an efficient information carrier, presents opportunities for direct information processing if effective interactions can be harnessed.
• * Kerr microresonator solitons offer a stable and controllable platform for optical information processing.

Purpose of the Study:

• * To develop and implement an Ising machine using an ensemble of Kerr microresonator solitons.
• * To demonstrate programmable all-to-all interactions for solving complex computational problems.
• * To evaluate the performance of the optical Ising machine against digital solvers for the Boolean satisfiability problem.

Main Methods:

• * Harnessing hundreds of Kerr microresonator solitons in an analog feedback network.
• * Implementing programmable all-to-all interactions by tuning feedback parameters.
• * Utilizing spin-like bifurcation in solitons for universal interactions.
• * Solving the Boolean satisfiability problem (SAT) using the developed Ising machine.

Main Results:

• * Successfully created an Ising machine with fully programmable interactions using Kerr solitons.
• * Demonstrated rapid and precise solutions for complex SAT instances.
• * Achieved low energy and time costs per feedback step (~0.15 mW/soliton and 1 µs).
• * Exceeded the performance of benchmark digital SAT solvers in >10,000 trials on >100 SAT instances.

Conclusions:

• * The optical Ising machine shows significant potential for computation acceleration.
• * Convergence of optical nonlinearity, ultralow loss photonics, and optoelectronics is key for future computing tasks.
• * This approach offers a viable alternative for solving computationally intensive problems like SAT.