Research In Time-varying EWMA Run-to-run Control In High-mix Process

As the semiconductor manufacturing industry evolving rapidly recently, the wafer size becomes larger and larger; on the other hand, the critical dimension shrinks because of the new technology. In this circumstance, the semiconductor manufacturing industries introduce some new control theory to maintain profitable. Run-to-run control is a form of discrete process and machine control, which adjusts the recipe precisely and timely to control the critical dimension and other wafer characters, so as to minimize the process shift, drift and variability. Run-to-run control has been widely used in semiconductor manufacturing nowadays. Exponent weighted moving average (EWMA) is one of the most important control theories, which uses post-process output as controller's input to adjust the parameters between runs to control the current process. Discount factor is the most important parameter of EWMA controller.Actually, there are different types of products or different processes of the same product in the same product line. The tool's parameters will change obviously after several runs of one product, and this will induce low convergence rate, large output varying or output instable. This thesis is based on the high-mix process in fabrication, using tool-based control theory and product-based control theory to analysis the result of different control theories. In order to decrease the numbers of non-product wafers and rework rate which are caused by output low-convergence rate, we introduce the time-varying discount factor into product-based EWMA control theory to optimize the capability or fast responding. Because the time-varying discount factor converges to a fixed value rapidly, so this control theory just increases the response's speed and won't affect the stability. In the end of this thesis there are simulations of different control theory using in the real process to expound their feasible and preponderance.