Spurious Modes Metric Definition for Machine Learning Aided MEMS Design

This work reports on a novel metric to assess the severity of in-band spurious modes in high performance microelectromechanical acoustic resonators. Different approaches are discussed to design an informative standard to be implemented as a target function for unsupervised machine learning aided MEMS design algorithms. According to this preliminary analysis, an absolute metric (M), based on the in-band spurious modes admittance excursion, and a regularized version of M (MR) are defined. Both standards are finally tested on the frequency response of sub-GHz X-cut lithium niobate laterally vibrating resonators, succeeding in quantifying the severity of in-band unwanted modes with minimal computational complexity.