Adaptive Virtual Metrology applied to a CVD process

Micron Technology and Applied Materials engaged in a project to determine if Virtual Metrology (VM) is viable in a semiconductor manufacturing environment, focusing on six chambers of a CVD process with a target metrology parameter of thickness. An adaptive modeling process was utilized that related process fault detection data to metrology data. Results indicate that (1), VM models can be developed and used to improve productivity and reduce cost of the CVD process by providing data to support more intelligent strategies and possibly by supporting a move from lot level to wafer level run-to-run control, and (2) the adaptive component of the models makes the models robust to drift and shift in process characteristics. Model robustness could be improved by utilizing multiple adaptive techniques depending on the process dynamics.