Optimal Feedforward Recipe Adjustment for CD Control in Semiconductor Patterning

Acceptable yields for nanofabrication will require significant improvement in CD control. One method to achieve better runto-run CD control is through inter-process feedforward control. The potential benefits of feedforward control include reduced run-to-run post-etch CD variance, rework, and scrap. However, measurement noise poses a significant threat to the success of feedforward control. Since the stakes are high, an incorrect control action is unacceptable. To answer this concern, this paper will focus on how to properly use the available sensor measurement in a run-to-run feedforward recipe adjustment controller. We have developed a methodology based in probability theory that detunes the controller based on the confidence in the sensor's accuracy. Properly detuning the controller has the effect of filtering out the noise from the SEM We will simulate this control strategy on industrial gate-etch data.