Sensitivity Enhancement of Dielectric Plasma Etching Endpoint Detection by Optical Emission Spectra With Modified $K$ -Means Cluster Analysis

The sensitivity enhancement of etching endpoint detection is demonstrated by analyzing the optical emission spectra with the modified K-means cluster analysis for small area dielectric etching processes in inductively coupled plasma. K-means cluster analysis algorithm is modified for real-time endpoint detection and applied to enhance the sensitivity of optical emission signals. The proposed technique demonstrates endpoint detection as low as 1.0% open area SiN
_x
etching and five times improved sensitivity compared to the average of normalized single wavelengths related to CN radical. This technique is expected to be applied not only to endpoint detection but also to fault detection of semiconductor fabrication processes.