Two Data Driven CMP Modeling Methods for DFM

Chemical Mechanical Polishing(CMP) plays an essential role in the fabrication of advanced integrated circuits, and the flatness of the wafer surface after CMP directly affects the stability of subsequent processes and chip yields. Mechanism- based modeling is difficult. Data-driven approaches offer a different approach to CMP modeling. In this paper, we use artificial neural networks and Group Method of Data Handling(GMDH) polynomial networks to build CMP models based on real silicon data, and compare and analyze the two approaches.