Development of endpoint detection algorithm in the multi-step plasma etching process

An endpoint detection algorithm based on principal component regression is developed for the multi-step plasma etching process with the whole optical emission spectra data. Because many endpoint detection techniques use a few manually selected wavelengths, noise render them ineffective and it is hard to select important wavelengths. Furthermore, the smaller the open area changes, the more difficult this single wavelength method detection the endpoint. In this paper, the principal component regression between two wafers was used for the real-time endpoint detection In case study, we applied our multiple models to the multi-step plasma etching process, which consisted of continuous polysilicon etching after the bottom anti-reflective coating etching. So we could obtain the simple and clear information for the more effect endpoint detection, which can be used for the improved process monitoring afterwards.