Laser Ablation Quality Study of Silicon Nitride during CMOS-MEMS post Processing by Using Machine Learning and Data Science

Abstract

Laser processing could be applied to the process of CMOS-MEMS fabrication. The TSMC/TSRI D35 common process is an example in this study. There are different laser wavelengths, laser energy, interval time, and light targets for different material fabrication. This study proposes the laser ablation quality of silicon nitride as the measurement of laser processing. According to the experimental results, the best quality of the green laser is 93% while the energy is 0.228 mJ, the interval time is 60 s and the light target size is 30 x 30 μm2. On the other hand, the best quality of the ultraviolet ablation is 92% which is generated at an energy of 0.48 mJ, an interval of 30 s, and an aperture size of 30 x 30 μm2. As the energy increases, the ablation quality becomes large. The results demonstrate the Fraunhofer diffraction is a dominant role in this study of laser ablation quality. This study simultaneously investigates the ablation phenomenon of microfabrication in green laser applied to CMOS-MEMS components by machine learning and data science. That proposes the approaching methodology for the optimal operation of the laser processing. The experimental results show that the pulse interval time is 90 s and the energy density is 57 J/m2, which has a good quality of ablation. Data science and machine learning successfully predict the quality level of ablation by using the random forest algorithm to achieve a mean accuracy of 98.04%.