Fast and accurate defect classification for CMP process monitoring

Abstract

Chemical mechanical planarization (CMP) has become one of the most critical processes in advanced integrated circuit manufacturing. The CMP process involves a complex physical and chemical reaction for film thickness removal and surface roughness improvement. The control of defects associated with CMP processes is critical to avoiding yield loss. As device critical dimensions keep shrinking, CMP defect control limits are getting stricter than in the past. In addition to controlling the total defect count, individual defect type counts also need to be controlled. In this paper, an automated defect classification algorithm, inLine Defect Organizer (iDO™), is used in conjunction with laser scattering inspection technology to classify scratch, ring pit and particle/residue defects generated by CMP processes with different shapes (concave, concave-convex mixed, and convex) on unpatterned monitor wafers. Using iDO, an extremely high value for the purity (>80%) of the classified defects can be achieved, enabling tracking of the variation in individual defect counts with different CMP process flows. This method can also be applied for inline monitoring to shorten the partition time of excursion wafers by 40% without scanning electron microscope (SEM) defect review. Furthermore, engineers can also easily investigate the defect formation mechanism based on the classified results. In this paper, the defect formation mechanism of scratch, ring pit and particle/residue defect types are discussed and proved by an iDO result. It is proven to be helpful for the CMP process optimization to minimize the killer defects.