Model Quality Evaluation in Semiconductor Manufacturing Process With EWMA Run-to-Run Control

Abstract

Exponentially weighted moving average run-to-run (EWMA-RtR) controller is frequently adopted in semiconductor manufacturing process. As a model-based control method, the performance of EWMA-RtR controller greatly depends on the process model quality, hence the process model-plant mismatch is an important factor influencing the control performance. In this paper, a model quality evaluation method for semiconductor manufacturing process with EWMA-RtR control is proposed without introducing any external excitation. The white noise sequences are first obtained from the routine closed-loop data by an orthogonal projection method and a disturbance model is estimated from the output error by recursive extended least square method. A model quality variable is then defined and calculated using the disturbance model and the output error. The model evaluation index is proposed using the ratio between the variance of the estimated white noise and that of the model quality variable. The effectiveness of the proposed method is demonstrated by a simulated chemical mechanical polishing process and a shallow trench isolation etch process based on industrial data.