Multivariate method for the monitoring of etch chamber insitu cleaning

In plasma etching, the etch byproduct deposition on the chamber wall plays an influential role in controlling the density of reactive species. Both recombination and release of reactive species occur depending on the wall conditions such as: temperature, thickness, and composition of the deposited film. The stability of the wall conditions affects the etch output such as critical dimension and selectivity to the exposed films. A well known practice to maintain the process chamber stability and prevent process drift is to season the plasma chamber with conditions similar to the ones used for etching product wafers. Periodical insitu cleaning to remove byproduct films has also been used. In order to control such processes, a monitoring system is needed. Optical emission spectroscopy (OES) has been extensively used in plasma etching and specific set of wavelengths monitoring has been established for several etch applications. In the case of monitoring the insitu cleaning, literature is very limited due the uniqueness of each case. The byproduct accumulation on the chamber wall depends on the etch product mix. In this paper we developed a multivariate method that combines machine learning algorithm (MLA) and principal component analysis (PCA). MLA is used to reduce the input variables to the few ones that are contributing to the differentiation between clean and chamber with polymer buildup while PCA has been used to build a control chart to monitor the state of the etch chamber.