Solid State Technology and Devices Seminar: New Optimization Strategies in Inverse Electromagnetic Design

Seminar | October 4 | 1-2 p.m. | Cory Hall, The Hogan Room, 521

 Jonathan Fan, Assistant Professor, Department of Electrical Engineering at Stanford University

 Electrical Engineering and Computer Sciences (EECS)

In this talk, I will discuss new advances in the inverse design of nanophotonic devices. As a model
system, I will focus on the application of these design modalities to high efficiency metasurfaces, though
the concepts are general and broadly apply to passive electromagnetic systems. First, I will show how
freeform geometric design can be achieved using the adjoint variables method. The resulting devices
utilize qualitatively new types of light-matter interactions based on strong near-field interactions
between nanostructures, enabling new diffractive optics phenomena. I will then discuss two ways
generative neural networks can augment and generalize the freeform inverse design process. The first
is with generative adversarial networks, which can learn from images of topology-optimized devices.
The second is from global topology optimization networks, termed GLONets, in which the global
optimization process is reframed as the training of a generative neural network. These ideas help set
the stage for data-driven approaches to be used in defining the next generation of high performance
electromagnetic technologies.

 dadevera@berkeley.edu, 510-642-3214