Spectral clustering algorithm for real‐time endpoint detection of silicon nitride plasma etching

Abstract

AbstractThe spectral clustering algorithm (SCA) is developed for endpoint detection (EPD) of Si3N4 plasma etching using optical emission spectroscopy (OES). OES signals are collected in real‐time and filtered using discrete wavelet transform (DWT). The SCA using 3648 full‐spectrum wavelengths with DWT filtering improves signal‐to‐noise ratio (SNR) by 2.78 times compared to single‐wavelength related to N2 molecule in 1.0% relative area etching. The wavelengths related to reactants and products are selected to enhance the SNR of the SCA. The SCA using 87 selected wavelengths with DWT filtering improves SNR by 3.57 times compared to SCA using full‐spectrum wavelengths. This study demonstrates that the SCA improves the etching EPD sensitivity and can be applied for fault detection of various plasma processes.