FIP Seminar: Deep Dreaming the Next 20 Years of Metamaterials

Electromagnetic metamaterials obtain their properties from geometry rather than chemistry, and have demonstrated novel scattering not realizable with conventional materials. However, as metamaterials and metasurfaces have become more complex, the structure-property relationship is increasingly less understood or sometimes completely unknown. For complex metamaterials the only means to determine metamaterial properties is through computational electromagnetic simulations, making it challenging to achieve optimal results. Deep neural networks have been shown to effectively infer the relationship between metamaterial parameters and resulting electromagnetic properties using simulated training data. I present several examples of deep learning techniques applied to both forward and inverse metamaterial design problems, giving an overview of the field, including open challenges and potential future directions along this research line.

Willie Padilla is the Dr. Paul Wang Distinguished Professor of Electrical & Computer Engineering and the Director of the Duke Metamaterials Center. Dr. Padilla was an author on the "discovery" paper on "left-handed" or Negative Index (NI) materials, a main contributor demonstrating artificial magnetic response at THz frequencies, and pioneered the metamaterial "perfect absorber". He was awarded a Director's Postdoctoral Fellowship from Los Alamos National Laboratory, received a Young Investigator Award from the Office of Naval Research, and Presidential Early Career Award for Scientists and Engineers. Dr. Padilla is a Fellow of the American Physical Society, Optical Society of America, and Kavli Frontiers of Science, and is a Web of Science Highly Cited Researcher in Physics for 2018 and 2019. He heads a group working in the area of metamaterials with a focus on machine learning, computational imaging, spectroscopy and energy, and has more than 230 peer-reviewed journal articles.