NONLOCAL NANOSCALE OPTICAL COMPUTATIONAL LAYERS
Abstract
Our society is hungry for data, and our overall consumption continues to rise exponentially year over-year. Bolstered by new technologies such as 5G, artificial intelligence, streaming platforms, telemedicine, etc., the rapid growth puts a strain on technologies to keep pace with demand. Paired with the slowing of conventional semiconductor scaling, new approaches to transmit and process information have been explored in recent years including quantum systems and neuromorphic systems with the goals to push speed, efficiency, size, and security limits. One promising approach has been the use of optical technologies to process information. Photonics provides an extremely fast method of computation, generally on the terahertz level, high parallelism, and direct compatibility with the large number of optical technologies such as imaging systems, optical fibers, and numerous sensors. However, the approaches developed thus far rely entirely on linear optical interactions to process data. While useful for a range of problems, most physical effects in the natural world are nonlinear, whose dynamics represent a significant challenge to model/solve using only linear systems. This limiting the potential to scale optical processing technologies to many classes of problems that were previously out of reach. This proposal will directly address this problem, seeking to realize a new highly nonlinear and inherently interconnected optical processing layer by combining films with a refractive index near zero with a nanostructured layer exhibiting a narrow resonance. The device is shown to have an overall mathematical model that directly resembles a nonlinear perceptron – the building block of artificial neural networks – and will facilitate the proof-of-concept of all-optical nonlinear processing. The tasks of the proposal to 1) realize the proposed processing layer, 2) characterize its performance under both linear and nonlinear optical scenarios, 3) develop a model through which one may create a digital-twin of the physical system, and 4) demonstrate hybrid nonlinear information processing using the digital-twin model. The effort will enable the exploration of new regimes of optical data processing that include nonlinearities, providing a platform for NextG smart sensors, analog all-optical modeling of nonlinear equations, and future on-chip neural networks for high-speed data routing, among many others.
Document Details
- Document Type
- DoD Grant Award
- Publication Date
- Apr 20, 2023
- Source ID
- FA95502210383
Entities
People
- Nathaniel Kinsey
Organizations
- Air Force Office of Scientific Research
- United States Air Force
- Virginia Commonwealth University