Control of a semiconductor dry etch process using variation and correlation analyses

Abstract

Statistical process control (SPC) is one of the traditional quality control methods that, if correctly applied, can be effective to improve and maintain quality and yield in any manufacturing facility. The purpose of this project is to demonstrate how to effectively apply SPC to a dry etch process (in this case plasma ashing), at Analog Devices, Inc., a company that runs large-scale fabrication sites in the Boston area. This thesis focuses on spatial and run-to-run variation across multiple measurement sites on a wafer and validates the assumptions of normality and correlation between sites within a wafer in order to justify and confirm the value of employing SPC theories to the plasma ashing process. By plotting control charts on past data, outlier data points are detected using Analog's current monitoring system. Further, irregularities in the process that would not have been detected using traditional x-bar Shewhart charts are detected by monitoring non-uniformity. Finally, cost analysis suggests that implementing SPC would be a modest investment relative to the potential savings.