半導體批次控制研究及其應用於混貨製程

Abstract

近年來，指數加權移動平均(EWMA)控制器已成為相當普遍的批次(Run-to-Run)控制器，但其對於量測機台之量測雜訊處理方面尚未有相關之探討。本文利用卡曼濾波器(Kalman filter, KF)來處理批次控制中量測雜訊的問題，並且研究其穩態時系統之輸出特性以及閉迴路之穩定性。文中針對半導體製程建構了五種不同型態之製程干擾(process disturbance)並加以分析；其製程干擾分別為：deterministic trend, DT, random walk with drift, RWD, integrated moving average process, IMA(1,1), autoregressive moving average, ARMA(1,1), and autoregressive integrated moving average, ARIMA(1,1,1)。分析結果顯示出KF可視為EWMA控制器之延伸形式，亦即EWMA控制器是KF在未考慮量測雜訊下之特殊型式。模擬結果也顯示出KF在處理量測雜訊方面相較於EWMA控制器有較佳的表現。另一方面，本文亦提出一套混貨批次控制器modified double EWMA (m-dEWMA) controller，並且應用於化學機械研磨(Chemical mechanical planarization, CMP)製程與黃光微影堆疊誤差(lithography overlay)製程上。m-dEWMA控制器是結合double EWMA控制器以及製程飄移干擾補償機制來處理機台所產生之飄移干擾(drift)以及產品特性所產生之位移(shift)干擾。文中亦針對三種不同之製程干擾(DT, RWD, IMA(1,1) with drift)加以分析m-dEWMA控制器在特定混貨生產排程下之穩態性質以及閉迴路穩定性。藉由真實混貨歷史製程資料與近期所發表之混貨控制方法(threaded EWMA controller, JADE control and threaded predicted correction controller (threaded PCC))模擬比較，m-dEWMA控制器皆有較佳的控制表現；同時，在CMP均勻度線上實驗部分，本文所提出之m-dEWMA控制器亦展現出顯著之控制效果。

Recently, the exponentially weighted moving average (EWMA) controller had become a popular control method in Run-to-run (RtR) process control, but the issue of measurement noise from metrology tools had not been addressed in RtR EWMA controllers yet. This dissertation utilized a Kalman filter (KF) controller to deal with measurement noise in RtR process control and investigated the system output properties for steady-state mean and variance, and for closed-loop stability. Five disturbance models modeling semiconductor process disturbances were investigated. They are deterministic trend, DT, random walk with drift, RWD, integrated moving average process, IMA(1,1), autoregressive moving average, ARMA(1,1), and autoregressive integrated moving average, ARIMA(1,1,1), disturbances. Analytical results showed that a KF controller could be considered as an extended version of an RtR EWMA controller. In particular, the EWMA controller was a special case of KF in a filtering form without the capability of measuring noise. Simulation results also showed that the KF had a better ability to deal with measurement noise than the EWMA controller. This thesis also proposed a new RtR control scheme, modified double EWMA controller (m-dEWMA), which is adaptive to the mixed product Chemical mechanical planarization (CMP) and lithography overlay processes. The m-dEWMA controller combined threaded dEWMA with the drift compensation scheme to deal with the drift and shift disturbances caused by tool and products, respectively. Moreover, we also analyzed the steady-state stability and property of m-dEWMA controller under three types of disturbance, DT, RWD and IMA(1,1) with drift, in specific mixed product process schedule. Comparing recently mixed product control schemes: threaded EWMA controller, JADE control and threaded prediction correction controller (threaded PCC), the simulation results showed that the m-dEWMA had better control performance than other controllers. The experiment results revealed that the m-dEWMA controller improved the wafer uniformity significantly in mixed product CMP process.