Ensemble and Unsupervised Machine Learning Applied on Laser Ablation Quality Study of Silicon Nitride during CMOS-MEMS Post Processing

Abstract

This work proposes a brand-new approach for the laser ablation study of Si3 N4 film based upon unsupervised machine learning (ML) in the CMOS-MEMS process. The study demonstrates that energy and interval time dominate laser ablation quality for green light 532nm. There are four rational classes for this task by the k-means algorithm. While the interval time is longer than 70 s, the mean laser ablation quality (reb) is more than 80%. The interval time is shorter than the 50s which reb is less than 74%. The result shows energy 0.318mJ, interval time 84 seconds, pulse shots 5 times, and left pad position to have the maximum reb of 88.64% compared to other conditions. Finally, there is a statistically significant relationship between energy and reb based on the P-value of OLS regression. Typical ensemble learners Decision Tree and Random Forest have the appropriate classification ability.