Modelling and decision support system for intelligent manufacturing: An empirical study for feedforward-feedback learning-based run-to-run controller for semiconductor dry-etching process

Abstract

Shrinkage in semiconductor devices affects the process window of all wafer fabrication steps including plasma etching. Drifts or shifts are most significant effects on the etching process due to shrinkage in semiconductor devices. Any drift or shift affects on critical dimensions (CD) of the wafer and changes the thickness and the width over time. Therefore, there would be an essential need for estimation and minimization of CD variation on a wafer-to-wafer basis by optimization techniques. This study aims to design a learning-based control system for monitoring the CD in Dry-Etching process. Feedforward-feedback control technique is used to reduce CD variation. Among all learning-based control systems, the Iterative Learning Control (ILC) integrated with Virtual Meteorology (VM) data, as a well-known system which can involve both feedforward signal from the past events, and feedback signals from the output of the current event is used to learn the behavior of the system and enhance the performance of the controller run-by-run. The proposed control model is optimized by gradient learning approach. The result is validated through the simulated study manipulated from empirical data and shows the advantage of the proposed feedforward-feedback learning controller than the common run-to-run exponentially weighted moving average (EWMA) control design.