04/06/2022
By Abantika Ghosh

The Kennedy College of Science, Department of Physics & Applied Physics, invites you to attend a doctoral thesis defense by Abantika Ghosh on "Physics-Informed Machine Learning for Optical Meta-materials."

The defense will be held on April 14 at 11 a.m. via Zoom. Please contact at Abantika_Ghosh@student.uml.edu for meeting information if you are interested in attending.

Committee Chair/Advisor: Viktor A. Podolskiy, Ph.D., Professor, Department of Physics and Applied Physics, University of Massachusetts Lowell

Committee Members:

  • Archana Kamal, Ph.D., Assistant Professor, Department of Physics and Applied Physics, University of Massachusetts Lowell
  • Timothy Cook, Ph.D., Associate Professor, Department of Physics and Applied Physics, University of Massachusetts Lowell

Abstract:

In this dissertation, implementation of Physics-driven knowledge in Machine Learning (ML) algorithms will be presented to advance the design and characterization of optical metamaterials. A comprehensive analysis of machine learning algorithms has been performed to gain insight into the learning process and to understand the flow of sub-wavelength information in diffractive imaging. A new technique for image characterization based on a single far-field intensity measurement is presented which is highly capable of fast and accurate characterization of two-dimensional structures with sub-wavelength resolution. This technique can be applied to new optical characterization tools with high spatial resolution, fast data acquisition, such as high-speed nanoscale metrology and quality control and can be further developed to high-resolution spectroscopy. Separately, Physics-informed learning is developed and illustrated on the example of analysis of optical modes propagating through a spatially periodic optical composite. Physics-informed learning can be used to enhance the accuracy and generalizability of the ML algorithms and improve optical metamaterial design, optimization, and characterization. The approach presented here can be readily utilized in other situations mapped onto an eigenvalue problem, a known bottleneck of computational electrodynamics.