Particles Contaminations Detection during Plasma Etching Process by using k-Nearest Neighbors and Fuzzy k-Nearest Neighbors

This paper present the particle contamination detection during plasma etching process by using k-nearest neighbor (kNN) and Fuzzy k-nearest neighbor (FkNN). In the process of manufacturing semiconductor devices, detecting particle contamination in process tool is a vital factor for determining for product yield. In situ particle is an accurate and cost effective method of contamination control in a production environment which possible to measure particles under actual conditions in real time. Data were collected from two sources Statistical Process Control (SPC) database and Advance Process Control (APC) database. There are four features which are Standard Deviation of voltage bias, Range between minimum and maximum of voltage bias, average of voltage bias and Radio frequency (RF) per Hour. These data are analyzed to identify important features that able to correlate with the particle contamination count during plasma etching process. In this research there are two part of analysis, individual parameter analysis and combination several parameter analysis by using kNN and FkNN. This analysis, used to classify into two levels of contamination, that are low and high particles contamination. By analysis results, kNN method is highest accuracy 83.33% by using standard deviation of voltage bias and FkNN show highest accuracy on combination parameters analysis 80.56% from combination between RF hour and standard deviation of voltage bias.