Learning Control for Motion Coordination in Wafer Scanners: Toward Gain Adaptation

Abstract

Accurate pattern transfer in wafer scanners necessitates the wafer stage and the reticle stage executing a coordinated motion with the synchronization error in terms of nanometers. In an attempt to cope with this challenging issue, a cross-coupling iterative learning control (ILC) with two inputs and two outputs is proposed and then decomposed into two ILC with the same convergence condition, a master–slave ILC for the reticle stage and an independent ILC for the wafer stage. To handle the inevitable stochastic disturbance, which inhibits the achievable ILC performance, an adaptive gain is involved in the proposed method for the sake of accelerated convergence as well as enhanced robustness. It remains constant or is decreased adaptively along the iteration axis, depending on the proportion that the stochastic term accounts for in the error signal. Moreover, a phase leader is chosen as the learning filter and tuned along with the initial learning gain by a frequency-domain approach. Experimental comparisons with existing close yet distinctive approaches highlight the effectiveness and superiority of the proposed method.