Statistical enhancement of a reflectometry metrology system

The increasing demands for in-situ statistical process control (SPC) in the semiconductor industry have elevated the need for high performance metrology. As we move to deep submicron technology and larger wafers, the accuracy and repeatability of the metrology system becomes more and more important for capturing the temporal and spatial process variation. In this paper we describe a set of statistical techniques for improving and evaluating a reflectometry metrology system. Our techniques include experiment design and Bayesian analysis that help us to focus only on the active factors while ignoring those that are statistically insignificant. Bootstrap evaluation of the metrology is performed based on those active factors. It is expected that this methodology will greatly facilitate the speed and characterization of in-situ metrology, which is essential to modern control applications