DIAGNOSTICS OF PLASMA ETCH: PCA WITHADAPTIVE CENTERING AND SCALING

摘要

In this research we show that static principal component analysis (PCA) is not sufficient for monitoring thernperformance of a plasma etch chamber. An adaptive scheme has been developed whereby the centeringrnand scaling coefficients in a static PCA model are updated, instead of using a dynamic PCA model orrnfrequent model rebuilding. This scheme is able to maintain a stable Q statistic for a long horizon below orrnnear the original model confidence limit. When the scheme is applied to data from a second chamber afterrnrunning only a single cassette of wafers, the resulting Q statistic lies below or near the original modelrnconfidence limit. A metric describing the relative movement of the adaptations is proposed in order tornhighlight variables with significant changes between two identified runs.