Analysis and reduction of variability in scanning electron microscopy measurements of critical dimensions

Abstract

This thesis describes work done during a Leaders for Manufacturing internship at Intel. At the broadest level, this work relates to the importance of controlling and monitoring measurement processes just as one controls the "fundamental" processes being measured. Without such control there can be no confidence in the integrity of the data describing the fundamental process. More specifically, the project assessed the variability that characterized the critical dimension measurements of one specific layer. It was shown that there was significant operator variability, related primarily to several common types of mismeasurement, that could not be monitored using standard production data. Other potential sources of variability were also investigated but were found to be less important. Various steps were undertaken to reduce the observed operator variability. As part of this effort an anonymous, automated feedback system was developed and piloted to give operators feedback on their measurements using a standard structure. Although the data from the pilot was inconclusive, the need to monitor measurement variability seems clear. Finally, the thesis recommends changing the production system so that information on measure­ment processes can be asceri.ained from standard production data. It also makes specific recommendations that while not addressing the control of the measurement process, could make the system less susceptible to variation.