"Image sensors are so sensitive to metal contamination that they can detect even one metal atom per pixel. To reduce the metal contamination, the plasma shower using RF (radio frequency) plasma generation is a representative example. The electrostatic angular energy filter after the mass analyzing magnet is a highly effective method to remove energetic metal contamination. The protection layer on the silicon is needed to protect the silicon wafer against the physisorbed metals. The thickness of the protection layer should be determined by considering the knock-on depth. The damage by ion implantation also causes blemishes. It becomes larger in the following conditions if the other conditions are the same; a. higher energy; b. larger dose; c. smaller beam size (higher beam current density); d. longer ion beam irradiation time; e. larger ion mass. To reduce channeling, the most effective method is to choose proper tilt and twist angles. For P+ pinning layer formation, the low-energy B+ implantation method might have less metal contamination and damage, compared with the BF2+ method."
This is very similar to DRAM, where the tail of the distribution functions of retention times (inverse of leakage) was always non-Gaussian, and was typically associated with discrete defects / metal impurities. There was also an interesting phenomenon of a variable retention time - where retention time switches between a high and low states, randomly in time.
ReplyDeleteIt's probably not surprise that some DRAM companies were also very successful with image sensors as well...