Accelerating the Innovation Cycle of Nanophotonic Systems Design

Abstract:
The general process for nanophotonics systems innovation involves identifying/generating a new concept, proposing a device design that can capture the concept, and validating the device design with an electromagnetic simulator. The latter two steps are typically performed iteratively by a researcher with specialized domain knowledge until a satisfactory device is identified, thereby requiring significant expenditure in time and computational cost. We will discuss computational algorithms based on deep neural networks that can accelerate the design and simulation of nanophotonic devices.

We will discuss the use of generative networks to perform population-based optimization and elucidate how the neural network architecture can be tailored to effectively search for the global optimum in a non-convex design landscape. We will also discuss how physics-informed deep networks can be trained with a combination of data and physical constraints to serve as accurate surrogate electromagnetic solvers. We anticipate that the ability for deep learning models to accelerate and even automate the simulation and design of photonic systems will push the innovation cycle of photonics research in academia and industry.