Abstract: Optical metasurfaces are thin large-area structures with aperiodic subwavelength patterns, designed for focusing light and a variety of other wave transformation. Because of their irregularity and large scale, they are one of the most challenging tasks for computational design. This talk will present ways to harness the full computational power of modern large-scale optimization in order to design metasurfaces with thousands or millions of free parameters. To that end, we exploit domain-decomposition approximations and 'surrogate' models. We will also present some recent experimental results including lenses with extended depth of field. Finally, we will discuss recent progress towards holistic 'end-to-end' optimization that combines optical design with lensless image processing.
Biography: Raphaël has been a postdoctoral associate at MIT Mathematics since 2020. His current research leverages machine learning for modeling and inverse design for PDE-constrained problems. Originally from France, Raphaël earned a dual degree between ESSEC Business School and École Centrale Paris. He earned a Master's of research in Nanosciences at Université Paris Saclay. He conducted his master's research on metamaterials at UC Berkeley in the lab of Prof. Xiang Zhang. He earned a PhD in applied mathematics from Harvard University as a fellow from Fulbright France. His PhD research was about inverse design for metasurfaces, and he was co-advised between Federico Capasso and Steven Johnson at MIT. Raphaël has also been a resident affiliate of Quincy House at Harvard College for four years, where he lives among and advises Harvard undergraduate students.
Location: Online Seminar
