Modeling and filtering of optical emission spectroscopy data for plasma etching systems

Full-spectrum in situ optical emission spectroscopy (OES) has emerged as a promising technology for in-line sensor systems for plasma etching process control. In this paper, we present a novel empirical model-based approach for OES data filtering based on statistical principal component analysis and jump linear filtering. The modeling procedure is demonstrated on a commercial multilayer Al/TiN/SiO2 etch process. For this process, we show that the proposed model provides a succinct representation of the OES signals and is in excellent agreement with the experimental data.