=================================================================================

Manual inking, a popular screening technique based on the observation that manufacturing defects are spatially correlated on the wafer surface [2], is being used to filter out the devices that are likely to exhibit latent defects. In this method, the assumption behind this common practice is that clusters of failing die on a wafer suggest a systematic local discrepancy. This technique is based on the proximity of each die to neighboring failed die, and the types of failure in that cluster. On the other hand, manual inking is inconsistently performed between different engineers, and often with slight discrepancies between inspections by the same engineer.

Table 4220. Patterns on wafers commonly reflecting specific process failure information.

Pattern	Failure information	Map	Reference
Blob	Contain patches of failing die at several locations on the wafer but not solely on the edge		[3]
Center	The mechanical polishing is uneven, or the pressure of the liquid is abnormal		[1]
Edge	Indicated by clusters of failing die near the edge of the wafer:    i) Wafers are often carried between processing steps in plastic carriers. The wafers are held between grooves in carriers such that only wafer edges touch the grooves. Particles can enter the carriers from sides only and are electrostatically attracted to the nearest edge of the wafer. Thus particles, clustered at the wafer edges, are likely to result in the formation of defects at the edges with a higher density than on any other location on the wafer.    ii) Photolithography defect arises due to diffraction caused by mask edges of the ultraviolet light used for exposure.		[3, 4]
Edge-Ring	Abnormal temperature control during annealing		[1]
Scratch	Indicates an exception in the moving or cutting processes; identified by streaks of failing die along the surface of the wafer and are usually caused by human error; thus, they are far rarer than the other two types		[1, 3]
None pattern	Contains defective grains with random distribution. This is caused by cleaning problems in cleaning rooms, which are expensive to eliminate completely; therefore, these defective grains are often considered noise		[1]

============================================

[1] Qiao Xu, Naigong Yu and Firdaous Essaf, Improved Wafer Map Inspection Using Attention Mechanism and Cosine Normalization, Machines, https://doi.org/10.3390/machines10020146, 10, 146, 2022.
[2] M. P. L. Ooi, E. K. J. Sim, Y. C. Kuang, L. Kleeman, C. Chan, and S. Demidenko, “Automatic defect cluster extraction for semiconductor wafers,” in Proc. IEEE Instrum. Meas. Technol. Conf., Austin, TX, USA, 2010, pp. 1024–1029.
[3] Constantinos Xanthopoulos, Arnold Neckermann, Paulus List, Klaus-Peter Tschernay, Peter Sarson and Yiorgos Makris, Automated Die Inking, IEEE TRANSACTIONS ON DEVICE AND MATERIALS RELIABILITY, 20(2), 295, 2020.
[4] Agha Irtaza Mirza, Spatial Yield Modeling for Semiconductor Wafers, thesis, 1995.