Small defect clusters leave more dies clean
Supplied working report; original verdicts not validated.
Review status: Needs review. This is an archival source note, not a passed proof.
notes.md
# Small defect clusters leave more dies clean **Verdict: Supported.** At fixed mean density, tightening the cluster radius raises yield in every one of the four density cells: the tightest clusters keep 0.72 of dies clean at D=1.0 where the widest keep 0.39, and every clustered case beats the Poisson reference. ## Design Neyman-Scott cluster-defect simulation, 200 x 200 mm wafer with 400 dies, 120 wafers per cell; mean defect density held at 0.1-1.0 per cm2 while cluster radius varies from 0.2 to 16 mm (0.02-1.6 of a die width). ## Result - Change in yield, widest minus narrowest cluster: -0.32 to -0.06 - Densities tested: 4 - Wafer dies: 400 ## Limits - Yield here means die with zero defects; killer-defect size and spatial correlation with die boundaries are not modelled. - The cluster radius is a Gaussian sigma, not a hard cluster diameter, so the 'cluster size' in the hypothesis is only loosely mapped. - At very large radii the whole wafer is one cluster, which is a different limit from the small-cluster comparison. - This tests the statistical geometry of clustering, not a particular fab process; real yield models also include systematic defect bands.