Machine Learning Inspired Electromagnetic Metasurfaces
This project aims to design novel metasurfaces with various functionalities and to develop a platform integrating advanced electromagnetic theory and AI techniques toward a behavior-oriented era of designing such metasurfaces.
Various frequency bands, including the microwave, mid-IR, and optical bands, are utilized for numerous real-life applications, including but not limited to wireless communications, sensing, and energy harvesting. This poses different challenges and opens the door for numerous fields of research towards the pursuit of better understanding, thus easier design and better performance for different elements of such systems.
On the other hand, metasurfaces are the 2D versions of metamaterials that have gained so much attention over the past few years. Their great capabilities in manipulating electromagnetic waves in unprecedented ways merely by choosing proper structural design parameters and usually omitting the need for unconventional materials lend themselves as an important candidate for solving crucial design problems across the frequency spectra. The expectations from metasurfaces in terms of their behavior and consequently the complexity of their 2D or quasi-2D design call for novel behavior-oriented methods to synthesize such structures. New computational algorithms based on concepts like machine learning and deep learning lend themselves as promising tools that, if integrated with a deep understanding of the electromagnetic properties of metasurfaces, can lead to unprecedented designs. In doing so, we open a new gateway to predicting the solution to a problem in the fastest possible way based on a statistical analysis of the datasets rather than researching the whole solution space. This project aims to design novel metasurfaces with various functionalities and to develop a platform integrating advanced electromagnetic theory and AI techniques toward a behavior-oriented era of designing such metasurfaces.