Variable convolution-based prediction of surface profiles in polishing pad conditioning

Abstract

Polishing pad conditioning is a crucial step in the chemical mechanical polishing (CMP) process, where establishing a relationship between conditioning parameters and pad wear profile is essential for process control. Traditional models based on diamond abrasive trajectory analysis are computationally intensive and time-consuming, creating a significant bottleneck for practical applications and forcing a reliance on inefficient trial-and-error process adjustments. To address this, a novel kinematic model is proposed, utilizing spatially varying convolution to predict pad wear profiles. The accuracy and effectiveness are validated through conditioning experiments. Benefiting from the improvement in prediction efficiency, extensive numerical simulations are conducted to investigate the influence of different conditioning parameters. The results indicate that matching the rotational speeds of the conditioner and the polishing pad, combined with reducing the conditioner size, produces pad wear profiles that closely align with the sweep time distribution. By adjusting the sweep mode, diverse pad wear profiles can be achieved. This study provides an efficient and accurate analytical framework for precise conditioning control, offering valuable insights for optimization conditioning process of CMP.