Intelligent Resonator Design Framework Integrating Vector Fitting, Spurious Score Learning with DNN, and Pareto Optimization for Spurious-Free High-Q Performances

This work presents a learning-guided framework for designing MEMS resonators with high quality factor ($Q$) and spurious-free admittance responses. Acoustic resonators are widely utilized in radio-frequency (RF) front-end components due to their scalability and inherently high$Q$values. The proposed pipeline automatically discovers optimal acoustic resonator designs, specifically 42° Y-cut, X-propagating Lithium Tantalate ($\text{LiTaO}_{3}$) -based SAW devices. A weighted spurious score helps to achieve this architecture, which fuses vector-fitting poles with non-dominant peaks and valleys into a unified quantitative indicator. The metric is further integrated with regression and deep neural network models to steer design exploration and suppress spurious modes. In contrast to conventional workflows relying on FEM-based multiphysics simulations and labor-intensive fabrication trials, the proposed framework leverages artificial intelligence and tape-out data to accelerate resonator design and validation.