Presentation description

Microwave and millimeter wave integrated circuits have become a rapidly growing field due to the potential for high-bandwidth and low-loss signal processing for wireless and wireline communications and sensing . In recent years, inverse designed electromagnetic surfaces have emerged as a potential solution to miniaturize the size of passive components and therefore increase the functionality and complexity of such circuits. Designing the complex surfaces needed for this application has traditionally been challenging, but several studies have explored the potential for optimization and machine-learning techniques to systematically create surfaces tailored to a particular function. We build upon prior studies in photonic integrated circuits to show that using the binary-additive reinforcement learning algorithm (b-ARLA) is a potentially effective solution to creating novel pixelated bandpass filters of arbitrary response. We developed a design pipeline combining b-ARLA optimization with custom Python/Visual Basic scripting for automated design generation and frequency simulation. The proposed framework displays promising potential for speeding up the design of pixelated metasurfaces for advanced signal-processing applications.