Fast and accurate prediction of spurious modes in aluminum nitride MEMS resonators using artificial neural network algorithm

This paper introduces the use of artificial neural networks (ANNs) as a method to guide the design of high-performance MEMS resonators and applies it to the prediction of spurious modes (SMs) in the proximity of the series resonance of aluminum nitride (AlN) contour-mode MEMS resonators (CMRs). ANN enables both fast and accurate prediction of spurious modes, hence facilitating the MEMS designer in the selection of the most appropriate geometry. In this work, the ANN can predict 1,000,000 cases within 0.3 seconds and with an expected accuracy of 98.9% given a margin, Δp, of 0.15 (out of 1). This study clearly outlines how powerful the ANN algorithm can be in helping designers identify the optimal device geometry. As a point of comparison, it would have taken hundreds of years for conventional 3-dimensional (3D) finite element method (FEM) to complete the same study.