以卡曼濾波器及最小變異控制器實現微影製程控制

Abstract

本論文主要目的為設計一套應用於微影製程之覆蓋誤差(Overlay，OL)的先進製程控制方法。經由歷史資料的分析找出曝光機覆蓋誤差之輸入變數與輸出變數之間的關係，建立批次控制之預測模型，然後利用卡曼濾波器(Kalman Filter)，做模型截距的動態調變，以適應製程隨時間變化之變異及干擾，再將已調變完成之模型代入最小變異控制器(Minimum Variance Controller，MVC)，使機台的輸入變數可以在改變量最小的限制下，即製程的變動最小的情形下，讓輸出變數快速地到達所設定的目標值，以提升製程之良率。

This thesis presents a new run-to-run (R2R) controller for photolithography process. The controller, termed Kalman minimum variance controller, can act both as a dynamical model optimizer and as a controller for given models. In this thesis, the relationships and run-to-run control model among input recipes and output overlay variables are formed by analyzing the historical data. Then on-line (Recursive) model identification and recipes generation are performed by using Kalman filter and minimum variance controller, respectively. Improvements due to advanced control have been quantified in simulations and actual FAB operations.