This paper present an analysis on in-situ particle monitor using virtual metrology system for particle contamination measurement. In the process of manufacturing semiconductor devices, detecting particle contamination in process tools is a vital factor for determining the product yield. In-situ monitoring of particle contamination can be as accurate and cost effective method of contamination control depend on type of particle monitoring sensor selection and its data used for virtual metrology. In this study, data samples are obtained from three system; Statistical Process Control (SPC) data base, Advanced Process Control (APC) data base and Hamamatsu Multiband Plasma Process Monitor system. Then, an artificial neural network based classifier called multilayer perceptron (MLP) network is applied to measure the particle contamination level from the given dataset. The performance of MLP network is compared using two different algorithms namely Levernberg-Marquad (LM) and resilient back-propagation (RP) algorithm. Based on the simulation results, it can be concluded that the MLP network using LM algorithm gives the best regression result of 0.999 and 0.54 during the training and testing respectively. The outcome of this project is in-situ particle monitor would be able to detect particle in the oxide etch chamber as alternative for Surf-scan methodology for each processed wafers.
展开▼
