Physics-Informed Machine Learning Assisted Liquid Crystals µWave Phase Shifters Design and Synthesis

Abstract

Liquid crystal (LC) has proven to be a promising material for microwave (µWave) phase shifters at GHz ranges, due to their continuous and wide tunability, as well as reasonably low absorption loss. However, designing LC phase shifters that meet specific application requirements (e.g., SpaceTech) is a challenging task that entails a complex trade-off between various parameters. Physics-informed machine learning (PI-ML) combines the power of machine learning with the underlying physics to develop a more accurate and interpretable model. Leveraging PI-ML to inform LC µWave device design is a relatively new area, with tremendous opportunities for exploration and innovation. In this article, a deep learning assisted LC µWave phase shifter design and synthesis framework is proposed. By incorporating physical constraints and knowledge into deep neural networks, one can effectively balance the trade-off between different design parameters and synthesize LC phase shifter structures that meet specific performance requirements (e.g., insertion loss, insertion loss balancing, phase tuning range, tuning speed, power consumption). The framework is envisaged to allow for the efficient and effective exploration of the design space, resulting in improved accuracy and efficiency compared to traditional two-stage design methods.