07/14/2025
By Danielle Fretwell

The Francis College of Engineering, Department of Electrical and Computer Engineering, invites you to attend a Doctoral Dissertation defense by Yunxi Dong on: "Deep Learning on Metasurface Design and Optimization."

Candidate Name: Yunxi Dong
Degree: Doctoral
Defense Date: Monday, July 28, 2025
Time: 9 - 11 a.m.
Location: Perry Hall, Room 315

Committee:
Advisor: Hualiang Zhang, Professor, Electrical & Computer Engineering, UMass Lowell (UML)

Committee Members*
Wei Guo, Associate Professor, Physics Engineering, UML
Anhar Uddin Bhuiyan, Assistant Professor, Electrical and Computer Engineering, UML
Xingwei Wang, Professor, Electrical and Computer Engineering, UML

Abstract:
Metamaterials, which are engineered on the sub-wavelength scale, exhibit properties not found in naturally occurring materials and are a key aspect of micro-/nanotechnology. They can manipulate electromagnetic waves in novel ways, leading to groundbreaking applications in various fields. As surface counterparts of metamaterials, metasurfaces can tune the amplitude and phase of incident wavefronts by adjusting the layout of their sub-wavelength components: the meta-atoms. Metasurfaces are promising to replace conventional bulky wavefront control devices with high design flexibility and compact size. Their potential applications span from advanced imaging systems to next-generation communication technologies, making them critical for research and development.
Meanwhile, the integration of deep learning into metasurfaces design represents a significant advancement in the field. Traditional design methods often rely on time-consuming trial-and-error processes and complex simulations that are computationally expensive. Deep learning offers a transformative approach by leveraging large datasets and powerful neural network architectures to predict electromagnetic responses and optimize design parameters efficiently. Furthermore, deep learning enables real-time end-to-end optimization, which is crucial for dynamic applications optical metasurface devices.
This paper outlines the integration of deep learning with metasurfaces design and optimization, aiming to enhance optical engineering through advanced computational techniques. First, a brief introduction of metasurfaces background theory and metasurfaces design methodology is presented. Second, the recent advancements of the implementation of deep learning to assist metasurfaces design are addressed. Third, a data-driven approach for the design of metasurfaces using deep learning algorithms is proposed. Lastly the use of deep learning algorithms to optimize optical metasurfaces imaging is discussed.