Error-Smoothing Exponentially Weighted Moving Average for Improving Critical Dimension Performance in Photolithography Process

The increasingly stringent tolerance of linewidths is a result of shrinking feature size of integrated circuits, and thus the manufacturing process in wafer fabrication should be accurately controlled to maintain process yields. Critical dimension (CD) is defined as the minimum width of a photoresist line or space printed on an exposure pattern by a stepper or scanner in photolithography. The CD is measured using metrology equipment and is compensated by modifying the corresponding equipment setup parameters. A feedback message is then sent to the next wafer for pre-adjustment and a feedforward message is sent to the previous wafer for post-adjustment. This study aimed to address a manufacturing intelligence framework to improve CD performance in a photolithography process. The input recipe is updated for the next run that is based on recently measured process data through the modified controller called as an error-smoothing exponentially weighted moving average (E-EWMA); both process and information flows are considered. A case study with a run-to-run process control is conducted to compensate the process variation to demonstrate the proposed framework. The results demonstrate that the proposed E-EWMA outperforms the conventional EWMA used in the company.