| 학회 |
한국화학공학회 |
| 학술대회 |
2022년 봄 (04/20 ~ 04/23, 제주국제컨벤션센터) |
| 권호 |
28권 1호, p.133 |
| 발표분야 |
[주제 2] 기계학습 |
| 제목 |
Development of Robust Endpoint Detection Algorithm for Cyclic Plasma Etching Process |
| 초록 |
Endpoint detection (EPD) is required in the plasma etching process to produce finely etched wafers during semiconductor device manufacture. EPD has traditionally been conducted using multivariate data from optical emission spectroscopy (OES), therefore there is a procedure that selects specific wavelengths that include important endpoint information. Furthermore, because increased device density on integrated circuit chips results in a low signal-to-noise ratio, several supplemental analysis approaches have been utilized to improve the spectral resolution of OES data linked with endpoint throughout the previous few decades. By merging data-driven wavelength selection and GMM-based EPD algorithm, this study introduces an end-to-end real-time EPD algorithm. For real-time EPD, wavelengths with a high correlation with a sigmoid function that separates data at the endpoint are chosen. The sigmoid function is then used to the endpoint training data in order to distinguish between features before and after the endpoint. Endpoint is defined as a series of deviations from a specified cluster in the real-time EPD method. |
| 저자 |
노해랑1, 김채선1, 이혜지1, 박태균2, 이용석2, 이찬민2, 윤국한2, 손영우1, 최원혁1, 이종민1
|
| 소속 |
1서울대, 2삼성전자 |
| 키워드 |
공정시스템(Process Systems Engineering) |
| E-Mail |
|
| 원문파일 |
초록 보기 |
