ASCENT: TUNA: TUnable randomness for NAtural computing

A new class of computers that rely on the natural relaxation process to find the ground state, dubbed as “natural computers”, has been gaining interest. Natural computers can find the solution to hard optimization problems using room-temperature electronics, with far greater speed and efficiency than conventional digital computers. A critical challenge for natural computers is preventing the system from getting stuck in an undesirable local minima point. This necessitates intrinsic noise in the system to perturb the intermediate solutions, however, noise parameters must be carefully tuned for the system to resolve to a good ground state. This project aims at developing new tunable nanoscale magnets, modeling and benchmarking tools, and demonstrator systems for enabling practical natural computers, which is only possible through an integrated program such as ASCENT. The ability to solve hard optimization problems using natural computers will have a direct and transformative impact on our daily lives. Many industries, from manufacturing and finance to transportation and resource management, are poised to benefit from the new optimization capabilities brought about by natural computers. This inter-disciplinary project is providing growth opportunities across the entire microelectronic design stack, from device fabrication to application development.The technological advances being made by this ASCENT project is threefold. First, novel nanoscale magnetic devices with auxiliary terminals for tuning the noise characteristics are being developed. To ensure good compatibility with state-of-the-art silicon technology, the project team is investigating tuning schemes that can be integrated seamlessly into a silicon fabrication process. Second, simulation models and benchmarking tools are being developed for evaluating the tuning capabilities afforded by the new nanomagnets. Using the evaluation framework, the team is investigating various tuning modalities, such as temporal or spatial or both, to understand the impact on the computation results and energy-efficiency. Finally, several practical demonstrator systems are being built in traditional and emerging semiconductor manufacturing platforms to validate the tunable device concepts using real-world application drivers.This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.