Accelerating Photonic Inverse Design

Abstract
“Inverse design” is the technique of using large-scale optimization, over thousands or even millions of parameters, to allow the computer to “discover” the best arrangement of materials for a given application, often leading to non-intuitive freeform geometries. In contrast to machine learning, this is typically model-driven optimization, based on accurate first-principles physics simulations, rather than data-driven.  However, there are many problems for which simply plugging the most straightforward computational model into an optimization algorithm yields an intractable problem, and non-obvious transformations are required to obtain a good solution with reasonable effort. In this talk, Professor Johnson will illustrate these ideas with several design problems from photonics, including incoherent emission, laser optimization, and the design of optical filters.

Steven G. Johnson is a professor of applied mathematics at the Massachusetts Institute of Technology, where he joined the faculty in 2004. He received his doctoral degree in physics from MIT in 2001, where he also received his bachelor’s degrees in physics, mathematics, and EECS in 1995. He works on the influence of complex geometries, particularly in the nanoscale, on the solutions of partial differential equations, especially for wave phenomena and electromagnetism. This includes analytical theory, numerics, and design of devices and phenomena. He is also known for his work in computational science, including both algorithmic advances and popular software packages, such as his development of the FFTW fast Fourier transform library, for which he co-received the 1999 J. H. Wilkinson Prize for Numerical Software, and the Meep and MPB software for electromagnetic simulation.