使用非線性D-EWMA控制器改善化學氣相沈積製程

Abstract

本論文應用一種新的R2R(Run-to-Run) 製程控制器-非線性D-EWMA控制器(Non-linear Double Exponentially Weighted Moving Average Controller)來改善化學氣相沈積之硬遮罩硼矽玻璃製程品質。首先由實驗設計得到影響製程品質特性之主要控制因子，再藉由非線性多元迴歸法建立製程控制模型，經由D-EWMA估測器更新每批次之控制模型截距項，再透過最小變異控制器(Minimum Variance Controller)來更新製程配方。本論文利用化學氣相沈積歷史資料來進行模擬及比較非線性D-EWMA與線性D-EWMA控制器之性能，結果顯示非線性D-EWMA可大幅改善硬遮罩硼矽玻璃製中因為製程干擾所產生的製程變異，最後本論文將非線性D-EWMA控制器實際應用於8吋半導體廠中以驗證控制器的效能。

This thesis proposes a new Run-to-Run process controller: Non-linear Double Exponentially Weighted Moving Average (D-EWMA) Controller to improve the process variance of Pad-Boron Silicon Glass (Pad-BSG) of Plasma Enhanced-Chemical Vapor Deposition (PE-CVD) process. Firstly, we use Design of Experiment (DOE) method to obtain the control factors which can affect the process characteristic, then the intercept of control model which is established by non-linear multi-regression method can be updated by D-EWMA every batch. The recipe of PE-CVD process is generated by MVC finally. This thesis presents the well performance of a Run-to-Run process controller in simulation with historical process data and experiment it on PE-CVD process in 8-inch fabrication. The simulation and experiment results show that the process variance converges fast and stably to the target value; moreover, the stability of process is also increased.