Discrete wavelet monitoring of plasma impedance matching for process control

Using a discrete wavelet transformation (DWT), anomalies in equipment states are monitored for diagnosis and control of plasma processes. Two variables governing DWT, type of filter function and level, were optimized on the basis of a factor-to-factor variation. Experimental data were collected with a real-time RF match monitor system from a plasma etch equipment commercially available. The factors that were varied incrementally include radio frequency power, pressure, and Ar flow rates. For a quantitative characterization and comparison of DWT filtered patterns, two simple metrics are introduced such as percent sensitivity (PS) and total percent sensitivity (TPS). For each variation in RF power level, DWT provided improved sensitivity over raw data. In four cases and six cases, DWT yielded an enhancement for variations in pressure and Ar flow rates, respectively. Particularly, DWT enabled identification of an anomaly induced by Ar flow rates, which was hardly discernable by monitoring raw data. As compared to raw data, DWT in TPS yielded an improvement of about 11% for variations in either power or pressure, respectively. For Ar flow rates, improvement of about 26% was achieved